Merge branch 'drm-next' of git://people.freedesktop.org/~airlied/linux

Pull drm merge (part 1) from Dave Airlie:
 "So first of all my tree and uapi stuff has a conflict mess, its my
  fault as the nouveau stuff didn't hit -next as were trying to rebase
  regressions out of it before we merged.

  Highlights:
   - SH mobile modesetting driver and associated helpers
   - some DRM core documentation
   - i915 modesetting rework, haswell hdmi, haswell and vlv fixes, write
     combined pte writing, ilk rc6 support,
   - nouveau: major driver rework into a hw core driver, makes features
     like SLI a lot saner to implement,
   - psb: add eDP/DP support for Cedarview
   - radeon: 2 layer page tables, async VM pte updates, better PLL
     selection for > 2 screens, better ACPI interactions

  The rest is general grab bag of fixes.

  So why part 1? well I have the exynos pull req which came in a bit
  late but was waiting for me to do something they shouldn't have and it
  looks fairly safe, and David Howells has some more header cleanups
  he'd like me to pull, that seem like a good idea, but I'd like to get
  this merge out of the way so -next dosen't get blocked."

Tons of conflicts mostly due to silly include line changes, but mostly
mindless.  A few other small semantic conflicts too, noted from Dave's
pre-merged branch.

* 'drm-next' of git://people.freedesktop.org/~airlied/linux: (447 commits)
  drm/nv98/crypt: fix fuc build with latest envyas
  drm/nouveau/devinit: fixup various issues with subdev ctor/init ordering
  drm/nv41/vm: fix and enable use of "real" pciegart
  drm/nv44/vm: fix and enable use of "real" pciegart
  drm/nv04/dmaobj: fixup vm target handling in preparation for nv4x pcie
  drm/nouveau: store supported dma mask in vmmgr
  drm/nvc0/ibus: initial implementation of subdev
  drm/nouveau/therm: add support for fan-control modes
  drm/nouveau/hwmon: rename pwm0* to pmw1* to follow hwmon's rules
  drm/nouveau/therm: calculate the pwm divisor on nv50+
  drm/nouveau/fan: rewrite the fan tachometer driver to get more precision, faster
  drm/nouveau/therm: move thermal-related functions to the therm subdev
  drm/nouveau/bios: parse the pwm divisor from the perf table
  drm/nouveau/therm: use the EXTDEV table to detect i2c monitoring devices
  drm/nouveau/therm: rework thermal table parsing
  drm/nouveau/gpio: expose the PWM/TOGGLE parameter found in the gpio vbios table
  drm/nouveau: fix pm initialization order
  drm/nouveau/bios: check that fixed tvdac gpio data is valid before using it
  drm/nouveau: log channel debug/error messages from client object rather than drm client
  drm/nouveau: have drm debugging macros build on top of core macros
  ...

72 files changed, 35165 insertions, 0 deletions

diff --git a/drivers/gpu/drm/nouveau/core/engine/bsp/nv84.c b/drivers/gpu/drm/nouveau/core/engine/bsp/nv84.c
new file mode 100644
index 0000000..66f7dfd
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/bsp/nv84.c
@@ -0,0 +1,175 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <engine/bsp.h>
+
+struct nv84_bsp_priv {
+	struct nouveau_bsp base;
+};
+
+struct nv84_bsp_chan {
+	struct nouveau_bsp_chan base;
+};
+
+/*******************************************************************************
+ * BSP object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv84_bsp_sclass[] = {
+	{},
+};
+
+/*******************************************************************************
+ * BSP context
+ ******************************************************************************/
+
+static int
+nv84_bsp_context_ctor(struct nouveau_object *parent,
+		      struct nouveau_object *engine,
+		      struct nouveau_oclass *oclass, void *data, u32 size,
+		      struct nouveau_object **pobject)
+{
+	struct nv84_bsp_chan *priv;
+	int ret;
+
+	ret = nouveau_bsp_context_create(parent, engine, oclass, NULL,
+					 0, 0, 0, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static void
+nv84_bsp_context_dtor(struct nouveau_object *object)
+{
+	struct nv84_bsp_chan *priv = (void *)object;
+	nouveau_bsp_context_destroy(&priv->base);
+}
+
+static int
+nv84_bsp_context_init(struct nouveau_object *object)
+{
+	struct nv84_bsp_chan *priv = (void *)object;
+	int ret;
+
+	ret = nouveau_bsp_context_init(&priv->base);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int
+nv84_bsp_context_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv84_bsp_chan *priv = (void *)object;
+	return nouveau_bsp_context_fini(&priv->base, suspend);
+}
+
+static struct nouveau_oclass
+nv84_bsp_cclass = {
+	.handle = NV_ENGCTX(BSP, 0x84),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv84_bsp_context_ctor,
+		.dtor = nv84_bsp_context_dtor,
+		.init = nv84_bsp_context_init,
+		.fini = nv84_bsp_context_fini,
+		.rd32 = _nouveau_bsp_context_rd32,
+		.wr32 = _nouveau_bsp_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * BSP engine/subdev functions
+ ******************************************************************************/
+
+static void
+nv84_bsp_intr(struct nouveau_subdev *subdev)
+{
+}
+
+static int
+nv84_bsp_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	      struct nouveau_oclass *oclass, void *data, u32 size,
+	      struct nouveau_object **pobject)
+{
+	struct nv84_bsp_priv *priv;
+	int ret;
+
+	ret = nouveau_bsp_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x04008000;
+	nv_subdev(priv)->intr = nv84_bsp_intr;
+	nv_engine(priv)->cclass = &nv84_bsp_cclass;
+	nv_engine(priv)->sclass = nv84_bsp_sclass;
+	return 0;
+}
+
+static void
+nv84_bsp_dtor(struct nouveau_object *object)
+{
+	struct nv84_bsp_priv *priv = (void *)object;
+	nouveau_bsp_destroy(&priv->base);
+}
+
+static int
+nv84_bsp_init(struct nouveau_object *object)
+{
+	struct nv84_bsp_priv *priv = (void *)object;
+	int ret;
+
+	ret = nouveau_bsp_init(&priv->base);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int
+nv84_bsp_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv84_bsp_priv *priv = (void *)object;
+	return nouveau_bsp_fini(&priv->base, suspend);
+}
+
+struct nouveau_oclass
+nv84_bsp_oclass = {
+	.handle = NV_ENGINE(BSP, 0x84),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv84_bsp_ctor,
+		.dtor = nv84_bsp_dtor,
+		.init = nv84_bsp_init,
+		.fini = nv84_bsp_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/copy/fuc/nva3.fuc b/drivers/gpu/drm/nouveau/core/engine/copy/fuc/nva3.fuc
new file mode 100644
index 0000000..219850d
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/copy/fuc/nva3.fuc
@@ -0,0 +1,872 @@
+/* fuc microcode for copy engine on nva3- chipsets
+ *
+ * Copyright 2011 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+/* To build for nva3:nvc0
+ *    m4 -DNVA3 nva3_copy.fuc | envyas -a -w -m fuc -V nva3 -o nva3_copy.fuc.h
+ *
+ * To build for nvc0-
+ *    m4 -DNVC0 nva3_copy.fuc | envyas -a -w -m fuc -V nva3 -o nvc0_copy.fuc.h
+ */
+
+ifdef(`NVA3',
+.section #nva3_pcopy_data
+,
+.section #nvc0_pcopy_data
+)
+
+ctx_object:                   .b32 0
+ifdef(`NVA3',
+ctx_dma:
+ctx_dma_query:                .b32 0
+ctx_dma_src:                  .b32 0
+ctx_dma_dst:                  .b32 0
+,)
+.equ #ctx_dma_count 3
+ctx_query_address_high:       .b32 0
+ctx_query_address_low:        .b32 0
+ctx_query_counter:            .b32 0
+ctx_src_address_high:         .b32 0
+ctx_src_address_low:          .b32 0
+ctx_src_pitch:                .b32 0
+ctx_src_tile_mode:            .b32 0
+ctx_src_xsize:                .b32 0
+ctx_src_ysize:                .b32 0
+ctx_src_zsize:                .b32 0
+ctx_src_zoff:                 .b32 0
+ctx_src_xoff:                 .b32 0
+ctx_src_yoff:                 .b32 0
+ctx_src_cpp:                  .b32 0
+ctx_dst_address_high:         .b32 0
+ctx_dst_address_low:          .b32 0
+ctx_dst_pitch:                .b32 0
+ctx_dst_tile_mode:            .b32 0
+ctx_dst_xsize:                .b32 0
+ctx_dst_ysize:                .b32 0
+ctx_dst_zsize:                .b32 0
+ctx_dst_zoff:                 .b32 0
+ctx_dst_xoff:                 .b32 0
+ctx_dst_yoff:                 .b32 0
+ctx_dst_cpp:                  .b32 0
+ctx_format:                   .b32 0
+ctx_swz_const0:               .b32 0
+ctx_swz_const1:               .b32 0
+ctx_xcnt:                     .b32 0
+ctx_ycnt:                     .b32 0
+.align 256
+
+dispatch_table:
+// mthd 0x0000, NAME
+.b16 0x000 1
+.b32 #ctx_object                     ~0xffffffff
+// mthd 0x0100, NOP
+.b16 0x040 1
+.b32 0x00010000 + #cmd_nop           ~0xffffffff
+// mthd 0x0140, PM_TRIGGER
+.b16 0x050 1
+.b32 0x00010000 + #cmd_pm_trigger    ~0xffffffff
+ifdef(`NVA3', `
+// mthd 0x0180-0x018c, DMA_
+.b16 0x060 #ctx_dma_count
+dispatch_dma:
+.b32 0x00010000 + #cmd_dma           ~0xffffffff
+.b32 0x00010000 + #cmd_dma           ~0xffffffff
+.b32 0x00010000 + #cmd_dma           ~0xffffffff
+',)
+// mthd 0x0200-0x0218, SRC_TILE
+.b16 0x80 7
+.b32 #ctx_src_tile_mode              ~0x00000fff
+.b32 #ctx_src_xsize                  ~0x0007ffff
+.b32 #ctx_src_ysize                  ~0x00001fff
+.b32 #ctx_src_zsize                  ~0x000007ff
+.b32 #ctx_src_zoff                   ~0x00000fff
+.b32 #ctx_src_xoff                   ~0x0007ffff
+.b32 #ctx_src_yoff                   ~0x00001fff
+// mthd 0x0220-0x0238, DST_TILE
+.b16 0x88 7
+.b32 #ctx_dst_tile_mode              ~0x00000fff
+.b32 #ctx_dst_xsize                  ~0x0007ffff
+.b32 #ctx_dst_ysize                  ~0x00001fff
+.b32 #ctx_dst_zsize                  ~0x000007ff
+.b32 #ctx_dst_zoff                   ~0x00000fff
+.b32 #ctx_dst_xoff                   ~0x0007ffff
+.b32 #ctx_dst_yoff                   ~0x00001fff
+// mthd 0x0300-0x0304, EXEC, WRCACHE_FLUSH
+.b16 0xc0 2
+.b32 0x00010000 + #cmd_exec          ~0xffffffff
+.b32 0x00010000 + #cmd_wrcache_flush ~0xffffffff
+// mthd 0x030c-0x0340, various stuff
+.b16 0xc3 14
+.b32 #ctx_src_address_high           ~0x000000ff
+.b32 #ctx_src_address_low            ~0xffffffff
+.b32 #ctx_dst_address_high           ~0x000000ff
+.b32 #ctx_dst_address_low            ~0xffffffff
+.b32 #ctx_src_pitch                  ~0x0007ffff
+.b32 #ctx_dst_pitch                  ~0x0007ffff
+.b32 #ctx_xcnt                       ~0x0000ffff
+.b32 #ctx_ycnt                       ~0x00001fff
+.b32 #ctx_format                     ~0x0333ffff
+.b32 #ctx_swz_const0                 ~0xffffffff
+.b32 #ctx_swz_const1                 ~0xffffffff
+.b32 #ctx_query_address_high         ~0x000000ff
+.b32 #ctx_query_address_low          ~0xffffffff
+.b32 #ctx_query_counter              ~0xffffffff
+.b16 0x800 0
+
+ifdef(`NVA3',
+.section #nva3_pcopy_code
+,
+.section #nvc0_pcopy_code
+)
+
+main:
+   clear b32 $r0
+   mov $sp $r0
+
+   // setup i0 handler and route fifo and ctxswitch to it
+   mov $r1 #ih
+   mov $iv0 $r1
+   mov $r1 0x400
+   movw $r2 0xfff3
+   sethi $r2 0
+   iowr I[$r1 + 0x300] $r2
+
+   // enable interrupts
+   or $r2 0xc
+   iowr I[$r1] $r2
+   bset $flags ie0
+
+   // enable fifo access and context switching
+   mov $r1 0x1200
+   mov $r2 3
+   iowr I[$r1] $r2
+
+   // sleep forever, waking for interrupts
+   bset $flags $p0
+   spin:
+      sleep $p0
+      bra #spin
+
+// i0 handler
+ih:
+   iord $r1 I[$r0 + 0x200]
+
+   and $r2 $r1 0x00000008
+   bra e #ih_no_chsw
+      call #chsw
+   ih_no_chsw:
+   and $r2 $r1 0x00000004
+   bra e #ih_no_cmd
+      call #dispatch
+
+   ih_no_cmd:
+   and $r1 $r1 0x0000000c
+   iowr I[$r0 + 0x100] $r1
+   iret
+
+// $p1 direction (0 = unload, 1 = load)
+// $r3 channel
+swctx:
+   mov $r4 0x7700
+   mov $xtargets $r4
+ifdef(`NVA3', `
+   // target 7 hardcoded to ctx dma object
+   mov $xdbase $r0
+', ` // NVC0
+   // read SCRATCH3 to decide if we are PCOPY0 or PCOPY1
+   mov $r4 0x2100
+   iord $r4 I[$r4 + 0]
+   and $r4 1
+   shl b32 $r4 4
+   add b32 $r4 0x30
+
+   // channel is in vram
+   mov $r15 0x61c
+   shl b32 $r15 6
+   mov $r5 0x114
+   iowrs I[$r15] $r5
+
+   // read 16-byte PCOPYn info, containing context pointer, from channel
+   shl b32 $r5 $r3 4
+   add b32 $r5 2
+   mov $xdbase $r5
+   mov $r5 $sp
+   // get a chunk of stack space, aligned to 256 byte boundary
+   sub b32 $r5 0x100
+   mov $r6 0xff
+   not b32 $r6
+   and $r5 $r6
+   sethi $r5 0x00020000
+   xdld $r4 $r5
+   xdwait
+   sethi $r5 0
+
+   // set context pointer, from within channel VM
+   mov $r14 0
+   iowrs I[$r15] $r14
+   ld b32 $r4 D[$r5 + 0]
+   shr b32 $r4 8
+   ld b32 $r6 D[$r5 + 4]
+   shl b32 $r6 24
+   or $r4 $r6
+   mov $xdbase $r4
+')
+   // 256-byte context, at start of data segment
+   mov b32 $r4 $r0
+   sethi $r4 0x60000
+
+   // swap!
+   bra $p1 #swctx_load
+      xdst $r0 $r4
+      bra #swctx_done
+   swctx_load:
+      xdld $r0 $r4
+   swctx_done:
+   xdwait
+   ret
+
+chsw:
+   // read current channel
+   mov $r2 0x1400
+   iord $r3 I[$r2]
+
+   // if it's active, unload it and return
+   xbit $r15 $r3 0x1e
+   bra e #chsw_no_unload
+      bclr $flags $p1
+      call #swctx
+      bclr $r3 0x1e
+      iowr I[$r2] $r3
+      mov $r4 1
+      iowr I[$r2 + 0x200] $r4
+      ret
+
+   // read next channel
+   chsw_no_unload:
+   iord $r3 I[$r2 + 0x100]
+
+   // is there a channel waiting to be loaded?
+   xbit $r13 $r3 0x1e
+   bra e #chsw_finish_load
+      bset $flags $p1
+      call #swctx
+ifdef(`NVA3',
+      // load dma objects back into TARGET regs
+      mov $r5 #ctx_dma
+      mov $r6 #ctx_dma_count
+      chsw_load_ctx_dma:
+         ld b32 $r7 D[$r5 + $r6 * 4]
+         add b32 $r8 $r6 0x180
+         shl b32 $r8 8
+         iowr I[$r8] $r7
+         sub b32 $r6 1
+         bra nc #chsw_load_ctx_dma
+,)
+
+   chsw_finish_load:
+   mov $r3 2
+   iowr I[$r2 + 0x200] $r3
+   ret
+
+dispatch:
+   // read incoming fifo command
+   mov $r3 0x1900
+   iord $r2 I[$r3 + 0x100]
+   iord $r3 I[$r3 + 0x000]
+   and $r4 $r2 0x7ff
+   // $r2 will be used to store exception data
+   shl b32 $r2 0x10
+
+   // lookup method in the dispatch table, ILLEGAL_MTHD if not found
+   mov $r5 #dispatch_table
+   clear b32 $r6
+   clear b32 $r7
+   dispatch_loop:
+      ld b16 $r6 D[$r5 + 0]
+      ld b16 $r7 D[$r5 + 2]
+      add b32 $r5 4
+      cmpu b32 $r4 $r6
+      bra c #dispatch_illegal_mthd
+      add b32 $r7 $r6
+      cmpu b32 $r4 $r7
+      bra c #dispatch_valid_mthd
+      sub b32 $r7 $r6
+      shl b32 $r7 3
+      add b32 $r5 $r7
+      bra #dispatch_loop
+
+   // ensure no bits set in reserved fields, INVALID_BITFIELD
+   dispatch_valid_mthd:
+   sub b32 $r4 $r6
+   shl b32 $r4 3
+   add b32 $r4 $r5
+   ld b32 $r5 D[$r4 + 4]
+   and $r5 $r3
+   cmpu b32 $r5 0
+   bra ne #dispatch_invalid_bitfield
+
+   // depending on dispatch flags: execute method, or save data as state
+   ld b16 $r5 D[$r4 + 0]
+   ld b16 $r6 D[$r4 + 2]
+   cmpu b32 $r6 0
+   bra ne #dispatch_cmd
+      st b32 D[$r5] $r3
+      bra #dispatch_done
+   dispatch_cmd:
+      bclr $flags $p1
+      call $r5
+      bra $p1 #dispatch_error
+      bra #dispatch_done
+
+   dispatch_invalid_bitfield:
+   or $r2 2
+   dispatch_illegal_mthd:
+   or $r2 1
+
+   // store exception data in SCRATCH0/SCRATCH1, signal hostirq
+   dispatch_error:
+   mov $r4 0x1000
+   iowr I[$r4 + 0x000] $r2
+   iowr I[$r4 + 0x100] $r3
+   mov $r2 0x40
+   iowr I[$r0] $r2
+   hostirq_wait:
+      iord $r2 I[$r0 + 0x200]
+      and $r2 0x40
+      cmpu b32 $r2 0
+      bra ne #hostirq_wait
+
+   dispatch_done:
+   mov $r2 0x1d00
+   mov $r3 1
+   iowr I[$r2] $r3
+   ret
+
+// No-operation
+//
+// Inputs:
+//    $r1: irqh state
+//    $r2: hostirq state
+//    $r3: data
+//    $r4: dispatch table entry
+// Outputs:
+//    $r1: irqh state
+//    $p1: set on error
+//       $r2: hostirq state
+//       $r3: data
+cmd_nop:
+   ret
+
+// PM_TRIGGER
+//
+// Inputs:
+//    $r1: irqh state
+//    $r2: hostirq state
+//    $r3: data
+//    $r4: dispatch table entry
+// Outputs:
+//    $r1: irqh state
+//    $p1: set on error
+//       $r2: hostirq state
+//       $r3: data
+cmd_pm_trigger:
+   mov $r2 0x2200
+   clear b32 $r3
+   sethi $r3 0x20000
+   iowr I[$r2] $r3
+   ret
+
+ifdef(`NVA3',
+// SET_DMA_* method handler
+//
+// Inputs:
+//    $r1: irqh state
+//    $r2: hostirq state
+//    $r3: data
+//    $r4: dispatch table entry
+// Outputs:
+//    $r1: irqh state
+//    $p1: set on error
+//       $r2: hostirq state
+//       $r3: data
+cmd_dma:
+   sub b32 $r4 #dispatch_dma
+   shr b32 $r4 1
+   bset $r3 0x1e
+   st b32 D[$r4 + #ctx_dma] $r3
+   add b32 $r4 0x600
+   shl b32 $r4 6
+   iowr I[$r4] $r3
+   ret
+,)
+
+// Calculates the hw swizzle mask and adjusts the surface's xcnt to match
+//
+cmd_exec_set_format:
+   // zero out a chunk of the stack to store the swizzle into
+   add $sp -0x10
+   st b32 D[$sp + 0x00] $r0
+   st b32 D[$sp + 0x04] $r0
+   st b32 D[$sp + 0x08] $r0
+   st b32 D[$sp + 0x0c] $r0
+
+   // extract cpp, src_ncomp and dst_ncomp from FORMAT
+   ld b32 $r4 D[$r0 + #ctx_format]
+   extr $r5 $r4 16:17
+   add b32 $r5 1
+   extr $r6 $r4 20:21
+   add b32 $r6 1
+   extr $r7 $r4 24:25
+   add b32 $r7 1
+
+   // convert FORMAT swizzle mask to hw swizzle mask
+   bclr $flags $p2
+   clear b32 $r8
+   clear b32 $r9
+   ncomp_loop:
+      and $r10 $r4 0xf
+      shr b32 $r4 4
+      clear b32 $r11
+      bpc_loop:
+         cmpu b8 $r10 4
+         bra nc #cmp_c0
+            mulu $r12 $r10 $r5
+            add b32 $r12 $r11
+            bset $flags $p2
+            bra #bpc_next
+         cmp_c0:
+         bra ne #cmp_c1
+            mov $r12 0x10
+            add b32 $r12 $r11
+            bra #bpc_next
+         cmp_c1:
+         cmpu b8 $r10 6
+         bra nc #cmp_zero
+            mov $r12 0x14
+            add b32 $r12 $r11
+            bra #bpc_next
+         cmp_zero:
+            mov $r12 0x80
+         bpc_next:
+         st b8 D[$sp + $r8] $r12
+         add b32 $r8 1
+         add b32 $r11 1
+         cmpu b32 $r11 $r5
+         bra c #bpc_loop
+      add b32 $r9 1
+      cmpu b32 $r9 $r7
+      bra c #ncomp_loop
+
+   // SRC_XCNT = (xcnt * src_cpp), or 0 if no src ref in swz (hw will hang)
+   mulu $r6 $r5
+   st b32 D[$r0 + #ctx_src_cpp] $r6
+   ld b32 $r8 D[$r0 + #ctx_xcnt]
+   mulu $r6 $r8
+   bra $p2 #dst_xcnt
+   clear b32 $r6
+
+   dst_xcnt:
+   mulu $r7 $r5
+   st b32 D[$r0 + #ctx_dst_cpp] $r7
+   mulu $r7 $r8
+
+   mov $r5 0x810
+   shl b32 $r5 6
+   iowr I[$r5 + 0x000] $r6
+   iowr I[$r5 + 0x100] $r7
+   add b32 $r5 0x800
+   ld b32 $r6 D[$r0 + #ctx_dst_cpp]
+   sub b32 $r6 1
+   shl b32 $r6 8
+   ld b32 $r7 D[$r0 + #ctx_src_cpp]
+   sub b32 $r7 1
+   or $r6 $r7
+   iowr I[$r5 + 0x000] $r6
+   add b32 $r5 0x100
+   ld b32 $r6 D[$sp + 0x00]
+   iowr I[$r5 + 0x000] $r6
+   ld b32 $r6 D[$sp + 0x04]
+   iowr I[$r5 + 0x100] $r6
+   ld b32 $r6 D[$sp + 0x08]
+   iowr I[$r5 + 0x200] $r6
+   ld b32 $r6 D[$sp + 0x0c]
+   iowr I[$r5 + 0x300] $r6
+   add b32 $r5 0x400
+   ld b32 $r6 D[$r0 + #ctx_swz_const0]
+   iowr I[$r5 + 0x000] $r6
+   ld b32 $r6 D[$r0 + #ctx_swz_const1]
+   iowr I[$r5 + 0x100] $r6
+   add $sp 0x10
+   ret
+
+// Setup to handle a tiled surface
+//
+// Calculates a number of parameters the hardware requires in order
+// to correctly handle tiling.
+//
+// Offset calculation is performed as follows (Tp/Th/Td from TILE_MODE):
+//    nTx = round_up(w * cpp, 1 << Tp) >> Tp
+//    nTy = round_up(h, 1 << Th) >> Th
+//    Txo = (x * cpp) & ((1 << Tp) - 1)
+//     Tx = (x * cpp) >> Tp
+//    Tyo = y & ((1 << Th) - 1)
+//     Ty = y >> Th
+//    Tzo = z & ((1 << Td) - 1)
+//     Tz = z >> Td
+//
+//    off  = (Tzo << Tp << Th) + (Tyo << Tp) + Txo
+//    off += ((Tz * nTy * nTx)) + (Ty * nTx) + Tx) << Td << Th << Tp;
+//
+// Inputs:
+//    $r4: hw command (0x104800)
+//    $r5: ctx offset adjustment for src/dst selection
+//    $p2: set if dst surface
+//
+cmd_exec_set_surface_tiled:
+   // translate TILE_MODE into Tp, Th, Td shift values
+   ld b32 $r7 D[$r5 + #ctx_src_tile_mode]
+   extr $r9 $r7 8:11
+   extr $r8 $r7 4:7
+ifdef(`NVA3',
+   add b32 $r8 2
+,
+   add b32 $r8 3
+)
+   extr $r7 $r7 0:3
+   cmp b32 $r7 0xe
+   bra ne #xtile64
+   mov $r7 4
+   bra #xtileok
+   xtile64:
+   xbit $r7 $flags $p2
+   add b32 $r7 17
+   bset $r4 $r7
+   mov $r7 6
+   xtileok:
+
+   // Op = (x * cpp) & ((1 << Tp) - 1)
+   // Tx = (x * cpp) >> Tp
+   ld b32 $r10 D[$r5 + #ctx_src_xoff]
+   ld b32 $r11 D[$r5 + #ctx_src_cpp]
+   mulu $r10 $r11
+   mov $r11 1
+   shl b32 $r11 $r7
+   sub b32 $r11 1
+   and $r12 $r10 $r11
+   shr b32 $r10 $r7
+
+   // Tyo = y & ((1 << Th) - 1)
+   // Ty  = y >> Th
+   ld b32 $r13 D[$r5 + #ctx_src_yoff]
+   mov $r14 1
+   shl b32 $r14 $r8
+   sub b32 $r14 1
+   and $r11 $r13 $r14
+   shr b32 $r13 $r8
+
+   // YTILE = ((1 << Th) << 12) | ((1 << Th) - Tyo)
+   add b32 $r14 1
+   shl b32 $r15 $r14 12
+   sub b32 $r14 $r11
+   or $r15 $r14
+   xbit $r6 $flags $p2
+   add b32 $r6 0x208
+   shl b32 $r6 8
+   iowr I[$r6 + 0x000] $r15
+
+   // Op += Tyo << Tp
+   shl b32 $r11 $r7
+   add b32 $r12 $r11
+
+   // nTx = ((w * cpp) + ((1 << Tp) - 1) >> Tp)
+   ld b32 $r15 D[$r5 + #ctx_src_xsize]
+   ld b32 $r11 D[$r5 + #ctx_src_cpp]
+   mulu $r15 $r11
+   mov $r11 1
+   shl b32 $r11 $r7
+   sub b32 $r11 1
+   add b32 $r15 $r11
+   shr b32 $r15 $r7
+   push $r15
+
+   // nTy = (h + ((1 << Th) - 1)) >> Th
+   ld b32 $r15 D[$r5 + #ctx_src_ysize]
+   mov $r11 1
+   shl b32 $r11 $r8
+   sub b32 $r11 1
+   add b32 $r15 $r11
+   shr b32 $r15 $r8
+   push $r15
+
+   // Tys = Tp + Th
+   // CFG_YZ_TILE_SIZE = ((1 << Th) >> 2) << Td
+   add b32 $r7 $r8
+   sub b32 $r8 2
+   mov $r11 1
+   shl b32 $r11 $r8
+   shl b32 $r11 $r9
+
+   // Tzo = z & ((1 << Td) - 1)
+   // Tz  = z >> Td
+   // Op += Tzo << Tys
+   // Ts  = Tys + Td
+   ld b32 $r8 D[$r5 + #ctx_src_zoff]
+   mov $r14 1
+   shl b32 $r14 $r9
+   sub b32 $r14 1
+   and $r15 $r8 $r14
+   shl b32 $r15 $r7
+   add b32 $r12 $r15
+   add b32 $r7 $r9
+   shr b32 $r8 $r9
+
+   // Ot = ((Tz * nTy * nTx) + (Ty * nTx) + Tx) << Ts
+   pop $r15
+   pop $r9
+   mulu $r13 $r9
+   add b32 $r10 $r13
+   mulu $r8 $r9
+   mulu $r8 $r15
+   add b32 $r10 $r8
+   shl b32 $r10 $r7
+
+   // PITCH = (nTx - 1) << Ts
+   sub b32 $r9 1
+   shl b32 $r9 $r7
+   iowr I[$r6 + 0x200] $r9
+
+   // SRC_ADDRESS_LOW   = (Ot + Op) & 0xffffffff
+   // CFG_ADDRESS_HIGH |= ((Ot + Op) >> 32) << 16
+   ld b32 $r7 D[$r5 + #ctx_src_address_low]
+   ld b32 $r8 D[$r5 + #ctx_src_address_high]
+   add b32 $r10 $r12
+   add b32 $r7 $r10
+   adc b32 $r8 0
+   shl b32 $r8 16
+   or $r8 $r11
+   sub b32 $r6 0x600
+   iowr I[$r6 + 0x000] $r7
+   add b32 $r6 0x400
+   iowr I[$r6 + 0x000] $r8
+   ret
+
+// Setup to handle a linear surface
+//
+// Nothing to see here.. Sets ADDRESS and PITCH, pretty non-exciting
+//
+cmd_exec_set_surface_linear:
+   xbit $r6 $flags $p2
+   add b32 $r6 0x202
+   shl b32 $r6 8
+   ld b32 $r7 D[$r5 + #ctx_src_address_low]
+   iowr I[$r6 + 0x000] $r7
+   add b32 $r6 0x400
+   ld b32 $r7 D[$r5 + #ctx_src_address_high]
+   shl b32 $r7 16
+   iowr I[$r6 + 0x000] $r7
+   add b32 $r6 0x400
+   ld b32 $r7 D[$r5 + #ctx_src_pitch]
+   iowr I[$r6 + 0x000] $r7
+   ret
+
+// wait for regs to be available for use
+cmd_exec_wait:
+   push $r0
+   push $r1
+   mov $r0 0x800
+   shl b32 $r0 6
+   loop:
+      iord $r1 I[$r0]
+      and $r1 1
+      bra ne #loop
+   pop $r1
+   pop $r0
+   ret
+
+cmd_exec_query:
+   // if QUERY_SHORT not set, write out { -, 0, TIME_LO, TIME_HI }
+   xbit $r4 $r3 13
+   bra ne #query_counter
+      call #cmd_exec_wait
+      mov $r4 0x80c
+      shl b32 $r4 6
+      ld b32 $r5 D[$r0 + #ctx_query_address_low]
+      add b32 $r5 4
+      iowr I[$r4 + 0x000] $r5
+      iowr I[$r4 + 0x100] $r0
+      mov $r5 0xc
+      iowr I[$r4 + 0x200] $r5
+      add b32 $r4 0x400
+      ld b32 $r5 D[$r0 + #ctx_query_address_high]
+      shl b32 $r5 16
+      iowr I[$r4 + 0x000] $r5
+      add b32 $r4 0x500
+      mov $r5 0x00000b00
+      sethi $r5 0x00010000
+      iowr I[$r4 + 0x000] $r5
+      mov $r5 0x00004040
+      shl b32 $r5 1
+      sethi $r5 0x80800000
+      iowr I[$r4 + 0x100] $r5
+      mov $r5 0x00001110
+      sethi $r5 0x13120000
+      iowr I[$r4 + 0x200] $r5
+      mov $r5 0x00001514
+      sethi $r5 0x17160000
+      iowr I[$r4 + 0x300] $r5
+      mov $r5 0x00002601
+      sethi $r5 0x00010000
+      mov $r4 0x800
+      shl b32 $r4 6
+      iowr I[$r4 + 0x000] $r5
+
+   // write COUNTER
+   query_counter:
+   call #cmd_exec_wait
+   mov $r4 0x80c
+   shl b32 $r4 6
+   ld b32 $r5 D[$r0 + #ctx_query_address_low]
+   iowr I[$r4 + 0x000] $r5
+   iowr I[$r4 + 0x100] $r0
+   mov $r5 0x4
+   iowr I[$r4 + 0x200] $r5
+   add b32 $r4 0x400
+   ld b32 $r5 D[$r0 + #ctx_query_address_high]
+   shl b32 $r5 16
+   iowr I[$r4 + 0x000] $r5
+   add b32 $r4 0x500
+   mov $r5 0x00000300
+   iowr I[$r4 + 0x000] $r5
+   mov $r5 0x00001110
+   sethi $r5 0x13120000
+   iowr I[$r4 + 0x100] $r5
+   ld b32 $r5 D[$r0 + #ctx_query_counter]
+   add b32 $r4 0x500
+   iowr I[$r4 + 0x000] $r5
+   mov $r5 0x00002601
+   sethi $r5 0x00010000
+   mov $r4 0x800
+   shl b32 $r4 6
+   iowr I[$r4 + 0x000] $r5
+   ret
+
+// Execute a copy operation
+//
+// Inputs:
+//    $r1: irqh state
+//    $r2: hostirq state
+//    $r3: data
+//       000002000 QUERY_SHORT
+//       000001000 QUERY
+//       000000100 DST_LINEAR
+//       000000010 SRC_LINEAR
+//       000000001 FORMAT
+//    $r4: dispatch table entry
+// Outputs:
+//    $r1: irqh state
+//    $p1: set on error
+//       $r2: hostirq state
+//       $r3: data
+cmd_exec:
+   call #cmd_exec_wait
+
+   // if format requested, call function to calculate it, otherwise
+   // fill in cpp/xcnt for both surfaces as if (cpp == 1)
+   xbit $r15 $r3 0
+   bra e #cmd_exec_no_format
+      call #cmd_exec_set_format
+      mov $r4 0x200
+      bra #cmd_exec_init_src_surface
+   cmd_exec_no_format:
+      mov $r6 0x810
+      shl b32 $r6 6
+      mov $r7 1
+      st b32 D[$r0 + #ctx_src_cpp] $r7
+      st b32 D[$r0 + #ctx_dst_cpp] $r7
+      ld b32 $r7 D[$r0 + #ctx_xcnt]
+      iowr I[$r6 + 0x000] $r7
+      iowr I[$r6 + 0x100] $r7
+      clear b32 $r4
+
+   cmd_exec_init_src_surface:
+   bclr $flags $p2
+   clear b32 $r5
+   xbit $r15 $r3 4
+   bra e #src_tiled
+      call #cmd_exec_set_surface_linear
+      bra #cmd_exec_init_dst_surface
+   src_tiled:
+      call #cmd_exec_set_surface_tiled
+      bset $r4 7
+
+   cmd_exec_init_dst_surface:
+   bset $flags $p2
+   mov $r5 #ctx_dst_address_high - #ctx_src_address_high
+   xbit $r15 $r3 8
+   bra e #dst_tiled
+      call #cmd_exec_set_surface_linear
+      bra #cmd_exec_kick
+   dst_tiled:
+      call #cmd_exec_set_surface_tiled
+      bset $r4 8
+
+   cmd_exec_kick:
+   mov $r5 0x800
+   shl b32 $r5 6
+   ld b32 $r6 D[$r0 + #ctx_ycnt]
+   iowr I[$r5 + 0x100] $r6
+   mov $r6 0x0041
+   // SRC_TARGET = 1, DST_TARGET = 2
+   sethi $r6 0x44000000
+   or $r4 $r6
+   iowr I[$r5] $r4
+
+   // if requested, queue up a QUERY write after the copy has completed
+   xbit $r15 $r3 12
+   bra e #cmd_exec_done
+      call #cmd_exec_query
+
+   cmd_exec_done:
+   ret
+
+// Flush write cache
+//
+// Inputs:
+//    $r1: irqh state
+//    $r2: hostirq state
+//    $r3: data
+//    $r4: dispatch table entry
+// Outputs:
+//    $r1: irqh state
+//    $p1: set on error
+//       $r2: hostirq state
+//       $r3: data
+cmd_wrcache_flush:
+   mov $r2 0x2200
+   clear b32 $r3
+   sethi $r3 0x10000
+   iowr I[$r2] $r3
+   ret
+
+.align 0x100
diff --git a/drivers/gpu/drm/nouveau/core/engine/copy/fuc/nva3.fuc.h b/drivers/gpu/drm/nouveau/core/engine/copy/fuc/nva3.fuc.h
new file mode 100644
index 0000000..c92520f
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/copy/fuc/nva3.fuc.h
@@ -0,0 +1,620 @@
+static u32 nva3_pcopy_data[] = {
+/* 0x0000: ctx_object */
+	0x00000000,
+/* 0x0004: ctx_dma */
+/* 0x0004: ctx_dma_query */
+	0x00000000,
+/* 0x0008: ctx_dma_src */
+	0x00000000,
+/* 0x000c: ctx_dma_dst */
+	0x00000000,
+/* 0x0010: ctx_query_address_high */
+	0x00000000,
+/* 0x0014: ctx_query_address_low */
+	0x00000000,
+/* 0x0018: ctx_query_counter */
+	0x00000000,
+/* 0x001c: ctx_src_address_high */
+	0x00000000,
+/* 0x0020: ctx_src_address_low */
+	0x00000000,
+/* 0x0024: ctx_src_pitch */
+	0x00000000,
+/* 0x0028: ctx_src_tile_mode */
+	0x00000000,
+/* 0x002c: ctx_src_xsize */
+	0x00000000,
+/* 0x0030: ctx_src_ysize */
+	0x00000000,
+/* 0x0034: ctx_src_zsize */
+	0x00000000,
+/* 0x0038: ctx_src_zoff */
+	0x00000000,
+/* 0x003c: ctx_src_xoff */
+	0x00000000,
+/* 0x0040: ctx_src_yoff */
+	0x00000000,
+/* 0x0044: ctx_src_cpp */
+	0x00000000,
+/* 0x0048: ctx_dst_address_high */
+	0x00000000,
+/* 0x004c: ctx_dst_address_low */
+	0x00000000,
+/* 0x0050: ctx_dst_pitch */
+	0x00000000,
+/* 0x0054: ctx_dst_tile_mode */
+	0x00000000,
+/* 0x0058: ctx_dst_xsize */
+	0x00000000,
+/* 0x005c: ctx_dst_ysize */
+	0x00000000,
+/* 0x0060: ctx_dst_zsize */
+	0x00000000,
+/* 0x0064: ctx_dst_zoff */
+	0x00000000,
+/* 0x0068: ctx_dst_xoff */
+	0x00000000,
+/* 0x006c: ctx_dst_yoff */
+	0x00000000,
+/* 0x0070: ctx_dst_cpp */
+	0x00000000,
+/* 0x0074: ctx_format */
+	0x00000000,
+/* 0x0078: ctx_swz_const0 */
+	0x00000000,
+/* 0x007c: ctx_swz_const1 */
+	0x00000000,
+/* 0x0080: ctx_xcnt */
+	0x00000000,
+/* 0x0084: ctx_ycnt */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0100: dispatch_table */
+	0x00010000,
+	0x00000000,
+	0x00000000,
+	0x00010040,
+	0x00010160,
+	0x00000000,
+	0x00010050,
+	0x00010162,
+	0x00000000,
+	0x00030060,
+/* 0x0128: dispatch_dma */
+	0x00010170,
+	0x00000000,
+	0x00010170,
+	0x00000000,
+	0x00010170,
+	0x00000000,
+	0x00070080,
+	0x00000028,
+	0xfffff000,
+	0x0000002c,
+	0xfff80000,
+	0x00000030,
+	0xffffe000,
+	0x00000034,
+	0xfffff800,
+	0x00000038,
+	0xfffff000,
+	0x0000003c,
+	0xfff80000,
+	0x00000040,
+	0xffffe000,
+	0x00070088,
+	0x00000054,
+	0xfffff000,
+	0x00000058,
+	0xfff80000,
+	0x0000005c,
+	0xffffe000,
+	0x00000060,
+	0xfffff800,
+	0x00000064,
+	0xfffff000,
+	0x00000068,
+	0xfff80000,
+	0x0000006c,
+	0xffffe000,
+	0x000200c0,
+	0x00010492,
+	0x00000000,
+	0x0001051b,
+	0x00000000,
+	0x000e00c3,
+	0x0000001c,
+	0xffffff00,
+	0x00000020,
+	0x00000000,
+	0x00000048,
+	0xffffff00,
+	0x0000004c,
+	0x00000000,
+	0x00000024,
+	0xfff80000,
+	0x00000050,
+	0xfff80000,
+	0x00000080,
+	0xffff0000,
+	0x00000084,
+	0xffffe000,
+	0x00000074,
+	0xfccc0000,
+	0x00000078,
+	0x00000000,
+	0x0000007c,
+	0x00000000,
+	0x00000010,
+	0xffffff00,
+	0x00000014,
+	0x00000000,
+	0x00000018,
+	0x00000000,
+	0x00000800,
+};
+
+static u32 nva3_pcopy_code[] = {
+/* 0x0000: main */
+	0x04fe04bd,
+	0x3517f000,
+	0xf10010fe,
+	0xf1040017,
+	0xf0fff327,
+	0x12d00023,
+	0x0c25f0c0,
+	0xf40012d0,
+	0x17f11031,
+	0x27f01200,
+	0x0012d003,
+/* 0x002f: spin */
+	0xf40031f4,
+	0x0ef40028,
+/* 0x0035: ih */
+	0x8001cffd,
+	0xf40812c4,
+	0x21f4060b,
+/* 0x0041: ih_no_chsw */
+	0x0412c472,
+	0xf4060bf4,
+/* 0x004a: ih_no_cmd */
+	0x11c4c321,
+	0x4001d00c,
+/* 0x0052: swctx */
+	0x47f101f8,
+	0x4bfe7700,
+	0x0007fe00,
+	0xf00204b9,
+	0x01f40643,
+	0x0604fa09,
+/* 0x006b: swctx_load */
+	0xfa060ef4,
+/* 0x006e: swctx_done */
+	0x03f80504,
+/* 0x0072: chsw */
+	0x27f100f8,
+	0x23cf1400,
+	0x1e3fc800,
+	0xf4170bf4,
+	0x21f40132,
+	0x1e3af052,
+	0xf00023d0,
+	0x24d00147,
+/* 0x0093: chsw_no_unload */
+	0xcf00f880,
+	0x3dc84023,
+	0x220bf41e,
+	0xf40131f4,
+	0x57f05221,
+	0x0367f004,
+/* 0x00a8: chsw_load_ctx_dma */
+	0xa07856bc,
+	0xb6018068,
+	0x87d00884,
+	0x0162b600,
+/* 0x00bb: chsw_finish_load */
+	0xf0f018f4,
+	0x23d00237,
+/* 0x00c3: dispatch */
+	0xf100f880,
+	0xcf190037,
+	0x33cf4032,
+	0xff24e400,
+	0x1024b607,
+	0x010057f1,
+	0x74bd64bd,
+/* 0x00dc: dispatch_loop */
+	0x58005658,
+	0x50b60157,
+	0x0446b804,
+	0xbb4d08f4,
+	0x47b80076,
+	0x0f08f404,
+	0xb60276bb,
+	0x57bb0374,
+	0xdf0ef400,
+/* 0x0100: dispatch_valid_mthd */
+	0xb60246bb,
+	0x45bb0344,
+	0x01459800,
+	0xb00453fd,
+	0x1bf40054,
+	0x00455820,
+	0xb0014658,
+	0x1bf40064,
+	0x00538009,
+/* 0x0127: dispatch_cmd */
+	0xf4300ef4,
+	0x55f90132,
+	0xf40c01f4,
+/* 0x0132: dispatch_invalid_bitfield */
+	0x25f0250e,
+/* 0x0135: dispatch_illegal_mthd */
+	0x0125f002,
+/* 0x0138: dispatch_error */
+	0x100047f1,
+	0xd00042d0,
+	0x27f04043,
+	0x0002d040,
+/* 0x0148: hostirq_wait */
+	0xf08002cf,
+	0x24b04024,
+	0xf71bf400,
+/* 0x0154: dispatch_done */
+	0x1d0027f1,
+	0xd00137f0,
+	0x00f80023,
+/* 0x0160: cmd_nop */
+/* 0x0162: cmd_pm_trigger */
+	0x27f100f8,
+	0x34bd2200,
+	0xd00233f0,
+	0x00f80023,
+/* 0x0170: cmd_dma */
+	0x012842b7,
+	0xf00145b6,
+	0x43801e39,
+	0x0040b701,
+	0x0644b606,
+	0xf80043d0,
+/* 0x0189: cmd_exec_set_format */
+	0xf030f400,
+	0xb00001b0,
+	0x01b00101,
+	0x0301b002,
+	0xc71d0498,
+	0x50b63045,
+	0x3446c701,
+	0xc70160b6,
+	0x70b63847,
+	0x0232f401,
+	0x94bd84bd,
+/* 0x01b4: ncomp_loop */
+	0xb60f4ac4,
+	0xb4bd0445,
+/* 0x01bc: bpc_loop */
+	0xf404a430,
+	0xa5ff0f18,
+	0x00cbbbc0,
+	0xf40231f4,
+/* 0x01ce: cmp_c0 */
+	0x1bf4220e,
+	0x10c7f00c,
+	0xf400cbbb,
+/* 0x01da: cmp_c1 */
+	0xa430160e,
+	0x0c18f406,
+	0xbb14c7f0,
+	0x0ef400cb,
+/* 0x01e9: cmp_zero */
+	0x80c7f107,
+/* 0x01ed: bpc_next */
+	0x01c83800,
+	0xb60180b6,
+	0xb5b801b0,
+	0xc308f404,
+	0xb80190b6,
+	0x08f40497,
+	0x0065fdb2,
+	0x98110680,
+	0x68fd2008,
+	0x0502f400,
+/* 0x0216: dst_xcnt */
+	0x75fd64bd,
+	0x1c078000,
+	0xf10078fd,
+	0xb6081057,
+	0x56d00654,
+	0x4057d000,
+	0x080050b7,
+	0xb61c0698,
+	0x64b60162,
+	0x11079808,
+	0xfd0172b6,
+	0x56d00567,
+	0x0050b700,
+	0x0060b401,
+	0xb40056d0,
+	0x56d00160,
+	0x0260b440,
+	0xb48056d0,
+	0x56d00360,
+	0x0050b7c0,
+	0x1e069804,
+	0x980056d0,
+	0x56d01f06,
+	0x1030f440,
+/* 0x0276: cmd_exec_set_surface_tiled */
+	0x579800f8,
+	0x6879c70a,
+	0xb66478c7,
+	0x77c70280,
+	0x0e76b060,
+	0xf0091bf4,
+	0x0ef40477,
+/* 0x0291: xtile64 */
+	0x027cf00f,
+	0xfd1170b6,
+	0x77f00947,
+/* 0x029d: xtileok */
+	0x0f5a9806,
+	0xfd115b98,
+	0xb7f000ab,
+	0x04b7bb01,
+	0xff01b2b6,
+	0xa7bbc4ab,
+	0x105d9805,
+	0xbb01e7f0,
+	0xe2b604e8,
+	0xb4deff01,
+	0xb605d8bb,
+	0xef9401e0,
+	0x02ebbb0c,
+	0xf005fefd,
+	0x60b7026c,
+	0x64b60208,
+	0x006fd008,
+	0xbb04b7bb,
+	0x5f9800cb,
+	0x115b980b,
+	0xf000fbfd,
+	0xb7bb01b7,
+	0x01b2b604,
+	0xbb00fbbb,
+	0xf0f905f7,
+	0xf00c5f98,
+	0xb8bb01b7,
+	0x01b2b604,
+	0xbb00fbbb,
+	0xf0f905f8,
+	0xb60078bb,
+	0xb7f00282,
+	0x04b8bb01,
+	0x9804b9bb,
+	0xe7f00e58,
+	0x04e9bb01,
+	0xff01e2b6,
+	0xf7bbf48e,
+	0x00cfbb04,
+	0xbb0079bb,
+	0xf0fc0589,
+	0xd9fd90fc,
+	0x00adbb00,
+	0xfd0089fd,
+	0xa8bb008f,
+	0x04a7bb00,
+	0xbb0192b6,
+	0x69d00497,
+	0x08579880,
+	0xbb075898,
+	0x7abb00ac,
+	0x0081b600,
+	0xfd1084b6,
+	0x62b7058b,
+	0x67d00600,
+	0x0060b700,
+	0x0068d004,
+/* 0x0382: cmd_exec_set_surface_linear */
+	0x6cf000f8,
+	0x0260b702,
+	0x0864b602,
+	0xd0085798,
+	0x60b70067,
+	0x57980400,
+	0x1074b607,
+	0xb70067d0,
+	0x98040060,
+	0x67d00957,
+/* 0x03ab: cmd_exec_wait */
+	0xf900f800,
+	0xf110f900,
+	0xb6080007,
+/* 0x03b6: loop */
+	0x01cf0604,
+	0x0114f000,
+	0xfcfa1bf4,
+	0xf800fc10,
+/* 0x03c5: cmd_exec_query */
+	0x0d34c800,
+	0xf5701bf4,
+	0xf103ab21,
+	0xb6080c47,
+	0x05980644,
+	0x0450b605,
+	0xd00045d0,
+	0x57f04040,
+	0x8045d00c,
+	0x040040b7,
+	0xb6040598,
+	0x45d01054,
+	0x0040b700,
+	0x0057f105,
+	0x0153f00b,
+	0xf10045d0,
+	0xb6404057,
+	0x53f10154,
+	0x45d08080,
+	0x1057f140,
+	0x1253f111,
+	0x8045d013,
+	0x151457f1,
+	0x171653f1,
+	0xf1c045d0,
+	0xf0260157,
+	0x47f10153,
+	0x44b60800,
+	0x0045d006,
+/* 0x0438: query_counter */
+	0x03ab21f5,
+	0x080c47f1,
+	0x980644b6,
+	0x45d00505,
+	0x4040d000,
+	0xd00457f0,
+	0x40b78045,
+	0x05980400,
+	0x1054b604,
+	0xb70045d0,
+	0xf1050040,
+	0xd0030057,
+	0x57f10045,
+	0x53f11110,
+	0x45d01312,
+	0x06059840,
+	0x050040b7,
+	0xf10045d0,
+	0xf0260157,
+	0x47f10153,
+	0x44b60800,
+	0x0045d006,
+/* 0x0492: cmd_exec */
+	0x21f500f8,
+	0x3fc803ab,
+	0x0e0bf400,
+	0x018921f5,
+	0x020047f1,
+/* 0x04a7: cmd_exec_no_format */
+	0xf11e0ef4,
+	0xb6081067,
+	0x77f00664,
+	0x11078001,
+	0x981c0780,
+	0x67d02007,
+	0x4067d000,
+/* 0x04c2: cmd_exec_init_src_surface */
+	0x32f444bd,
+	0xc854bd02,
+	0x0bf4043f,
+	0x8221f50a,
+	0x0a0ef403,
+/* 0x04d4: src_tiled */
+	0x027621f5,
+/* 0x04db: cmd_exec_init_dst_surface */
+	0xf40749f0,
+	0x57f00231,
+	0x083fc82c,
+	0xf50a0bf4,
+	0xf4038221,
+/* 0x04ee: dst_tiled */
+	0x21f50a0e,
+	0x49f00276,
+/* 0x04f5: cmd_exec_kick */
+	0x0057f108,
+	0x0654b608,
+	0xd0210698,
+	0x67f04056,
+	0x0063f141,
+	0x0546fd44,
+	0xc80054d0,
+	0x0bf40c3f,
+	0xc521f507,
+/* 0x0519: cmd_exec_done */
+/* 0x051b: cmd_wrcache_flush */
+	0xf100f803,
+	0xbd220027,
+	0x0133f034,
+	0xf80023d0,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/copy/fuc/nvc0.fuc.h b/drivers/gpu/drm/nouveau/core/engine/copy/fuc/nvc0.fuc.h
new file mode 100644
index 0000000..0d98c6c
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/copy/fuc/nvc0.fuc.h
@@ -0,0 +1,606 @@
+static u32 nvc0_pcopy_data[] = {
+/* 0x0000: ctx_object */
+	0x00000000,
+/* 0x0004: ctx_query_address_high */
+	0x00000000,
+/* 0x0008: ctx_query_address_low */
+	0x00000000,
+/* 0x000c: ctx_query_counter */
+	0x00000000,
+/* 0x0010: ctx_src_address_high */
+	0x00000000,
+/* 0x0014: ctx_src_address_low */
+	0x00000000,
+/* 0x0018: ctx_src_pitch */
+	0x00000000,
+/* 0x001c: ctx_src_tile_mode */
+	0x00000000,
+/* 0x0020: ctx_src_xsize */
+	0x00000000,
+/* 0x0024: ctx_src_ysize */
+	0x00000000,
+/* 0x0028: ctx_src_zsize */
+	0x00000000,
+/* 0x002c: ctx_src_zoff */
+	0x00000000,
+/* 0x0030: ctx_src_xoff */
+	0x00000000,
+/* 0x0034: ctx_src_yoff */
+	0x00000000,
+/* 0x0038: ctx_src_cpp */
+	0x00000000,
+/* 0x003c: ctx_dst_address_high */
+	0x00000000,
+/* 0x0040: ctx_dst_address_low */
+	0x00000000,
+/* 0x0044: ctx_dst_pitch */
+	0x00000000,
+/* 0x0048: ctx_dst_tile_mode */
+	0x00000000,
+/* 0x004c: ctx_dst_xsize */
+	0x00000000,
+/* 0x0050: ctx_dst_ysize */
+	0x00000000,
+/* 0x0054: ctx_dst_zsize */
+	0x00000000,
+/* 0x0058: ctx_dst_zoff */
+	0x00000000,
+/* 0x005c: ctx_dst_xoff */
+	0x00000000,
+/* 0x0060: ctx_dst_yoff */
+	0x00000000,
+/* 0x0064: ctx_dst_cpp */
+	0x00000000,
+/* 0x0068: ctx_format */
+	0x00000000,
+/* 0x006c: ctx_swz_const0 */
+	0x00000000,
+/* 0x0070: ctx_swz_const1 */
+	0x00000000,
+/* 0x0074: ctx_xcnt */
+	0x00000000,
+/* 0x0078: ctx_ycnt */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0100: dispatch_table */
+	0x00010000,
+	0x00000000,
+	0x00000000,
+	0x00010040,
+	0x0001019f,
+	0x00000000,
+	0x00010050,
+	0x000101a1,
+	0x00000000,
+	0x00070080,
+	0x0000001c,
+	0xfffff000,
+	0x00000020,
+	0xfff80000,
+	0x00000024,
+	0xffffe000,
+	0x00000028,
+	0xfffff800,
+	0x0000002c,
+	0xfffff000,
+	0x00000030,
+	0xfff80000,
+	0x00000034,
+	0xffffe000,
+	0x00070088,
+	0x00000048,
+	0xfffff000,
+	0x0000004c,
+	0xfff80000,
+	0x00000050,
+	0xffffe000,
+	0x00000054,
+	0xfffff800,
+	0x00000058,
+	0xfffff000,
+	0x0000005c,
+	0xfff80000,
+	0x00000060,
+	0xffffe000,
+	0x000200c0,
+	0x000104b8,
+	0x00000000,
+	0x00010541,
+	0x00000000,
+	0x000e00c3,
+	0x00000010,
+	0xffffff00,
+	0x00000014,
+	0x00000000,
+	0x0000003c,
+	0xffffff00,
+	0x00000040,
+	0x00000000,
+	0x00000018,
+	0xfff80000,
+	0x00000044,
+	0xfff80000,
+	0x00000074,
+	0xffff0000,
+	0x00000078,
+	0xffffe000,
+	0x00000068,
+	0xfccc0000,
+	0x0000006c,
+	0x00000000,
+	0x00000070,
+	0x00000000,
+	0x00000004,
+	0xffffff00,
+	0x00000008,
+	0x00000000,
+	0x0000000c,
+	0x00000000,
+	0x00000800,
+};
+
+static u32 nvc0_pcopy_code[] = {
+/* 0x0000: main */
+	0x04fe04bd,
+	0x3517f000,
+	0xf10010fe,
+	0xf1040017,
+	0xf0fff327,
+	0x12d00023,
+	0x0c25f0c0,
+	0xf40012d0,
+	0x17f11031,
+	0x27f01200,
+	0x0012d003,
+/* 0x002f: spin */
+	0xf40031f4,
+	0x0ef40028,
+/* 0x0035: ih */
+	0x8001cffd,
+	0xf40812c4,
+	0x21f4060b,
+/* 0x0041: ih_no_chsw */
+	0x0412c4ca,
+	0xf5070bf4,
+/* 0x004b: ih_no_cmd */
+	0xc4010221,
+	0x01d00c11,
+/* 0x0053: swctx */
+	0xf101f840,
+	0xfe770047,
+	0x47f1004b,
+	0x44cf2100,
+	0x0144f000,
+	0xb60444b6,
+	0xf7f13040,
+	0xf4b6061c,
+	0x1457f106,
+	0x00f5d101,
+	0xb6043594,
+	0x57fe0250,
+	0x0145fe00,
+	0x010052b7,
+	0x00ff67f1,
+	0x56fd60bd,
+	0x0253f004,
+	0xf80545fa,
+	0x0053f003,
+	0xd100e7f0,
+	0x549800fe,
+	0x0845b600,
+	0xb6015698,
+	0x46fd1864,
+	0x0047fe05,
+	0xf00204b9,
+	0x01f40643,
+	0x0604fa09,
+/* 0x00c3: swctx_load */
+	0xfa060ef4,
+/* 0x00c6: swctx_done */
+	0x03f80504,
+/* 0x00ca: chsw */
+	0x27f100f8,
+	0x23cf1400,
+	0x1e3fc800,
+	0xf4170bf4,
+	0x21f40132,
+	0x1e3af053,
+	0xf00023d0,
+	0x24d00147,
+/* 0x00eb: chsw_no_unload */
+	0xcf00f880,
+	0x3dc84023,
+	0x090bf41e,
+	0xf40131f4,
+/* 0x00fa: chsw_finish_load */
+	0x37f05321,
+	0x8023d002,
+/* 0x0102: dispatch */
+	0x37f100f8,
+	0x32cf1900,
+	0x0033cf40,
+	0x07ff24e4,
+	0xf11024b6,
+	0xbd010057,
+/* 0x011b: dispatch_loop */
+	0x5874bd64,
+	0x57580056,
+	0x0450b601,
+	0xf40446b8,
+	0x76bb4d08,
+	0x0447b800,
+	0xbb0f08f4,
+	0x74b60276,
+	0x0057bb03,
+/* 0x013f: dispatch_valid_mthd */
+	0xbbdf0ef4,
+	0x44b60246,
+	0x0045bb03,
+	0xfd014598,
+	0x54b00453,
+	0x201bf400,
+	0x58004558,
+	0x64b00146,
+	0x091bf400,
+	0xf4005380,
+/* 0x0166: dispatch_cmd */
+	0x32f4300e,
+	0xf455f901,
+	0x0ef40c01,
+/* 0x0171: dispatch_invalid_bitfield */
+	0x0225f025,
+/* 0x0174: dispatch_illegal_mthd */
+/* 0x0177: dispatch_error */
+	0xf10125f0,
+	0xd0100047,
+	0x43d00042,
+	0x4027f040,
+/* 0x0187: hostirq_wait */
+	0xcf0002d0,
+	0x24f08002,
+	0x0024b040,
+/* 0x0193: dispatch_done */
+	0xf1f71bf4,
+	0xf01d0027,
+	0x23d00137,
+/* 0x019f: cmd_nop */
+	0xf800f800,
+/* 0x01a1: cmd_pm_trigger */
+	0x0027f100,
+	0xf034bd22,
+	0x23d00233,
+/* 0x01af: cmd_exec_set_format */
+	0xf400f800,
+	0x01b0f030,
+	0x0101b000,
+	0xb00201b0,
+	0x04980301,
+	0x3045c71a,
+	0xc70150b6,
+	0x60b63446,
+	0x3847c701,
+	0xf40170b6,
+	0x84bd0232,
+/* 0x01da: ncomp_loop */
+	0x4ac494bd,
+	0x0445b60f,
+/* 0x01e2: bpc_loop */
+	0xa430b4bd,
+	0x0f18f404,
+	0xbbc0a5ff,
+	0x31f400cb,
+	0x220ef402,
+/* 0x01f4: cmp_c0 */
+	0xf00c1bf4,
+	0xcbbb10c7,
+	0x160ef400,
+/* 0x0200: cmp_c1 */
+	0xf406a430,
+	0xc7f00c18,
+	0x00cbbb14,
+/* 0x020f: cmp_zero */
+	0xf1070ef4,
+/* 0x0213: bpc_next */
+	0x380080c7,
+	0x80b601c8,
+	0x01b0b601,
+	0xf404b5b8,
+	0x90b6c308,
+	0x0497b801,
+	0xfdb208f4,
+	0x06800065,
+	0x1d08980e,
+	0xf40068fd,
+	0x64bd0502,
+/* 0x023c: dst_xcnt */
+	0x800075fd,
+	0x78fd1907,
+	0x1057f100,
+	0x0654b608,
+	0xd00056d0,
+	0x50b74057,
+	0x06980800,
+	0x0162b619,
+	0x980864b6,
+	0x72b60e07,
+	0x0567fd01,
+	0xb70056d0,
+	0xb4010050,
+	0x56d00060,
+	0x0160b400,
+	0xb44056d0,
+	0x56d00260,
+	0x0360b480,
+	0xb7c056d0,
+	0x98040050,
+	0x56d01b06,
+	0x1c069800,
+	0xf44056d0,
+	0x00f81030,
+/* 0x029c: cmd_exec_set_surface_tiled */
+	0xc7075798,
+	0x78c76879,
+	0x0380b664,
+	0xb06077c7,
+	0x1bf40e76,
+	0x0477f009,
+/* 0x02b7: xtile64 */
+	0xf00f0ef4,
+	0x70b6027c,
+	0x0947fd11,
+/* 0x02c3: xtileok */
+	0x980677f0,
+	0x5b980c5a,
+	0x00abfd0e,
+	0xbb01b7f0,
+	0xb2b604b7,
+	0xc4abff01,
+	0x9805a7bb,
+	0xe7f00d5d,
+	0x04e8bb01,
+	0xff01e2b6,
+	0xd8bbb4de,
+	0x01e0b605,
+	0xbb0cef94,
+	0xfefd02eb,
+	0x026cf005,
+	0x020860b7,
+	0xd00864b6,
+	0xb7bb006f,
+	0x00cbbb04,
+	0x98085f98,
+	0xfbfd0e5b,
+	0x01b7f000,
+	0xb604b7bb,
+	0xfbbb01b2,
+	0x05f7bb00,
+	0x5f98f0f9,
+	0x01b7f009,
+	0xb604b8bb,
+	0xfbbb01b2,
+	0x05f8bb00,
+	0x78bbf0f9,
+	0x0282b600,
+	0xbb01b7f0,
+	0xb9bb04b8,
+	0x0b589804,
+	0xbb01e7f0,
+	0xe2b604e9,
+	0xf48eff01,
+	0xbb04f7bb,
+	0x79bb00cf,
+	0x0589bb00,
+	0x90fcf0fc,
+	0xbb00d9fd,
+	0x89fd00ad,
+	0x008ffd00,
+	0xbb00a8bb,
+	0x92b604a7,
+	0x0497bb01,
+	0x988069d0,
+	0x58980557,
+	0x00acbb04,
+	0xb6007abb,
+	0x84b60081,
+	0x058bfd10,
+	0x060062b7,
+	0xb70067d0,
+	0xd0040060,
+	0x00f80068,
+/* 0x03a8: cmd_exec_set_surface_linear */
+	0xb7026cf0,
+	0xb6020260,
+	0x57980864,
+	0x0067d005,
+	0x040060b7,
+	0xb6045798,
+	0x67d01074,
+	0x0060b700,
+	0x06579804,
+	0xf80067d0,
+/* 0x03d1: cmd_exec_wait */
+	0xf900f900,
+	0x0007f110,
+	0x0604b608,
+/* 0x03dc: loop */
+	0xf00001cf,
+	0x1bf40114,
+	0xfc10fcfa,
+/* 0x03eb: cmd_exec_query */
+	0xc800f800,
+	0x1bf40d34,
+	0xd121f570,
+	0x0c47f103,
+	0x0644b608,
+	0xb6020598,
+	0x45d00450,
+	0x4040d000,
+	0xd00c57f0,
+	0x40b78045,
+	0x05980400,
+	0x1054b601,
+	0xb70045d0,
+	0xf1050040,
+	0xf00b0057,
+	0x45d00153,
+	0x4057f100,
+	0x0154b640,
+	0x808053f1,
+	0xf14045d0,
+	0xf1111057,
+	0xd0131253,
+	0x57f18045,
+	0x53f11514,
+	0x45d01716,
+	0x0157f1c0,
+	0x0153f026,
+	0x080047f1,
+	0xd00644b6,
+/* 0x045e: query_counter */
+	0x21f50045,
+	0x47f103d1,
+	0x44b6080c,
+	0x02059806,
+	0xd00045d0,
+	0x57f04040,
+	0x8045d004,
+	0x040040b7,
+	0xb6010598,
+	0x45d01054,
+	0x0040b700,
+	0x0057f105,
+	0x0045d003,
+	0x111057f1,
+	0x131253f1,
+	0x984045d0,
+	0x40b70305,
+	0x45d00500,
+	0x0157f100,
+	0x0153f026,
+	0x080047f1,
+	0xd00644b6,
+	0x00f80045,
+/* 0x04b8: cmd_exec */
+	0x03d121f5,
+	0xf4003fc8,
+	0x21f50e0b,
+	0x47f101af,
+	0x0ef40200,
+/* 0x04cd: cmd_exec_no_format */
+	0x1067f11e,
+	0x0664b608,
+	0x800177f0,
+	0x07800e07,
+	0x1d079819,
+	0xd00067d0,
+	0x44bd4067,
+/* 0x04e8: cmd_exec_init_src_surface */
+	0xbd0232f4,
+	0x043fc854,
+	0xf50a0bf4,
+	0xf403a821,
+/* 0x04fa: src_tiled */
+	0x21f50a0e,
+	0x49f0029c,
+/* 0x0501: cmd_exec_init_dst_surface */
+	0x0231f407,
+	0xc82c57f0,
+	0x0bf4083f,
+	0xa821f50a,
+	0x0a0ef403,
+/* 0x0514: dst_tiled */
+	0x029c21f5,
+/* 0x051b: cmd_exec_kick */
+	0xf10849f0,
+	0xb6080057,
+	0x06980654,
+	0x4056d01e,
+	0xf14167f0,
+	0xfd440063,
+	0x54d00546,
+	0x0c3fc800,
+	0xf5070bf4,
+/* 0x053f: cmd_exec_done */
+	0xf803eb21,
+/* 0x0541: cmd_wrcache_flush */
+	0x0027f100,
+	0xf034bd22,
+	0x23d00133,
+	0x0000f800,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/copy/nva3.c b/drivers/gpu/drm/nouveau/core/engine/copy/nva3.c
new file mode 100644
index 0000000..4df6da0
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/copy/nva3.c
@@ -0,0 +1,222 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/enum.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <subdev/fb.h>
+#include <subdev/vm.h>
+
+#include <engine/fifo.h>
+#include <engine/copy.h>
+
+#include "fuc/nva3.fuc.h"
+
+struct nva3_copy_priv {
+	struct nouveau_copy base;
+};
+
+struct nva3_copy_chan {
+	struct nouveau_copy_chan base;
+};
+
+/*******************************************************************************
+ * Copy object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nva3_copy_sclass[] = {
+	{ 0x85b5, &nouveau_object_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * PCOPY context
+ ******************************************************************************/
+
+static int
+nva3_copy_context_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nva3_copy_chan *priv;
+	int ret;
+
+	ret = nouveau_copy_context_create(parent, engine, oclass, NULL, 256, 0,
+					  NVOBJ_FLAG_ZERO_ALLOC, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static struct nouveau_oclass
+nva3_copy_cclass = {
+	.handle = NV_ENGCTX(COPY0, 0xa3),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nva3_copy_context_ctor,
+		.dtor = _nouveau_copy_context_dtor,
+		.init = _nouveau_copy_context_init,
+		.fini = _nouveau_copy_context_fini,
+		.rd32 = _nouveau_copy_context_rd32,
+		.wr32 = _nouveau_copy_context_wr32,
+
+	},
+};
+
+/*******************************************************************************
+ * PCOPY engine/subdev functions
+ ******************************************************************************/
+
+static const struct nouveau_enum nva3_copy_isr_error_name[] = {
+	{ 0x0001, "ILLEGAL_MTHD" },
+	{ 0x0002, "INVALID_ENUM" },
+	{ 0x0003, "INVALID_BITFIELD" },
+	{}
+};
+
+static void
+nva3_copy_intr(struct nouveau_subdev *subdev)
+{
+	struct nouveau_fifo *pfifo = nouveau_fifo(subdev);
+	struct nouveau_engine *engine = nv_engine(subdev);
+	struct nouveau_object *engctx;
+	struct nva3_copy_priv *priv = (void *)subdev;
+	u32 dispatch = nv_rd32(priv, 0x10401c);
+	u32 stat = nv_rd32(priv, 0x104008) & dispatch & ~(dispatch >> 16);
+	u64 inst = nv_rd32(priv, 0x104050) & 0x3fffffff;
+	u32 ssta = nv_rd32(priv, 0x104040) & 0x0000ffff;
+	u32 addr = nv_rd32(priv, 0x104040) >> 16;
+	u32 mthd = (addr & 0x07ff) << 2;
+	u32 subc = (addr & 0x3800) >> 11;
+	u32 data = nv_rd32(priv, 0x104044);
+	int chid;
+
+	engctx = nouveau_engctx_get(engine, inst);
+	chid   = pfifo->chid(pfifo, engctx);
+
+	if (stat & 0x00000040) {
+		nv_error(priv, "DISPATCH_ERROR [");
+		nouveau_enum_print(nva3_copy_isr_error_name, ssta);
+		printk("] ch %d [0x%010llx] subc %d mthd 0x%04x data 0x%08x\n",
+		       chid, inst << 12, subc, mthd, data);
+		nv_wr32(priv, 0x104004, 0x00000040);
+		stat &= ~0x00000040;
+	}
+
+	if (stat) {
+		nv_error(priv, "unhandled intr 0x%08x\n", stat);
+		nv_wr32(priv, 0x104004, stat);
+	}
+
+	nv50_fb_trap(nouveau_fb(priv), 1);
+	nouveau_engctx_put(engctx);
+}
+
+static int
+nva3_copy_tlb_flush(struct nouveau_engine *engine)
+{
+	nv50_vm_flush_engine(&engine->base, 0x0d);
+	return 0;
+}
+
+static int
+nva3_copy_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	bool enable = (nv_device(parent)->chipset != 0xaf);
+	struct nva3_copy_priv *priv;
+	int ret;
+
+	ret = nouveau_copy_create(parent, engine, oclass, enable, 0, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00802000;
+	nv_subdev(priv)->intr = nva3_copy_intr;
+	nv_engine(priv)->cclass = &nva3_copy_cclass;
+	nv_engine(priv)->sclass = nva3_copy_sclass;
+	nv_engine(priv)->tlb_flush = nva3_copy_tlb_flush;
+	return 0;
+}
+
+static int
+nva3_copy_init(struct nouveau_object *object)
+{
+	struct nva3_copy_priv *priv = (void *)object;
+	int ret, i;
+
+	ret = nouveau_copy_init(&priv->base);
+	if (ret)
+		return ret;
+
+	/* disable all interrupts */
+	nv_wr32(priv, 0x104014, 0xffffffff);
+
+	/* upload ucode */
+	nv_wr32(priv, 0x1041c0, 0x01000000);
+	for (i = 0; i < sizeof(nva3_pcopy_data) / 4; i++)
+		nv_wr32(priv, 0x1041c4, nva3_pcopy_data[i]);
+
+	nv_wr32(priv, 0x104180, 0x01000000);
+	for (i = 0; i < sizeof(nva3_pcopy_code) / 4; i++) {
+		if ((i & 0x3f) == 0)
+			nv_wr32(priv, 0x104188, i >> 6);
+		nv_wr32(priv, 0x104184, nva3_pcopy_code[i]);
+	}
+
+	/* start it running */
+	nv_wr32(priv, 0x10410c, 0x00000000);
+	nv_wr32(priv, 0x104104, 0x00000000); /* ENTRY */
+	nv_wr32(priv, 0x104100, 0x00000002); /* TRIGGER */
+	return 0;
+}
+
+static int
+nva3_copy_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nva3_copy_priv *priv = (void *)object;
+
+	nv_mask(priv, 0x104048, 0x00000003, 0x00000000);
+	nv_wr32(priv, 0x104014, 0xffffffff);
+
+	return nouveau_copy_fini(&priv->base, suspend);
+}
+
+struct nouveau_oclass
+nva3_copy_oclass = {
+	.handle = NV_ENGINE(COPY0, 0xa3),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nva3_copy_ctor,
+		.dtor = _nouveau_copy_dtor,
+		.init = nva3_copy_init,
+		.fini = nva3_copy_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/copy/nvc0.c b/drivers/gpu/drm/nouveau/core/engine/copy/nvc0.c
new file mode 100644
index 0000000..06d4a87
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/copy/nvc0.c
@@ -0,0 +1,265 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/enum.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <engine/fifo.h>
+#include <engine/copy.h>
+
+#include "fuc/nvc0.fuc.h"
+
+struct nvc0_copy_priv {
+	struct nouveau_copy base;
+};
+
+struct nvc0_copy_chan {
+	struct nouveau_copy_chan base;
+};
+
+/*******************************************************************************
+ * Copy object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nvc0_copy0_sclass[] = {
+	{ 0x90b5, &nouveau_object_ofuncs },
+	{},
+};
+
+static struct nouveau_oclass
+nvc0_copy1_sclass[] = {
+	{ 0x90b8, &nouveau_object_ofuncs },
+	{},
+};
+
+/*******************************************************************************
+ * PCOPY context
+ ******************************************************************************/
+
+static int
+nvc0_copy_context_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nvc0_copy_chan *priv;
+	int ret;
+
+	ret = nouveau_copy_context_create(parent, engine, oclass, NULL, 256,
+					  256, NVOBJ_FLAG_ZERO_ALLOC, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static struct nouveau_ofuncs
+nvc0_copy_context_ofuncs = {
+	.ctor = nvc0_copy_context_ctor,
+	.dtor = _nouveau_copy_context_dtor,
+	.init = _nouveau_copy_context_init,
+	.fini = _nouveau_copy_context_fini,
+	.rd32 = _nouveau_copy_context_rd32,
+	.wr32 = _nouveau_copy_context_wr32,
+};
+
+static struct nouveau_oclass
+nvc0_copy0_cclass = {
+	.handle = NV_ENGCTX(COPY0, 0xc0),
+	.ofuncs = &nvc0_copy_context_ofuncs,
+};
+
+static struct nouveau_oclass
+nvc0_copy1_cclass = {
+	.handle = NV_ENGCTX(COPY1, 0xc0),
+	.ofuncs = &nvc0_copy_context_ofuncs,
+};
+
+/*******************************************************************************
+ * PCOPY engine/subdev functions
+ ******************************************************************************/
+
+static const struct nouveau_enum nvc0_copy_isr_error_name[] = {
+	{ 0x0001, "ILLEGAL_MTHD" },
+	{ 0x0002, "INVALID_ENUM" },
+	{ 0x0003, "INVALID_BITFIELD" },
+	{}
+};
+
+static void
+nvc0_copy_intr(struct nouveau_subdev *subdev)
+{
+	struct nouveau_fifo *pfifo = nouveau_fifo(subdev);
+	struct nouveau_engine *engine = nv_engine(subdev);
+	struct nouveau_object *engctx;
+	int idx = nv_engidx(nv_object(subdev)) - NVDEV_ENGINE_COPY0;
+	struct nvc0_copy_priv *priv = (void *)subdev;
+	u32 disp = nv_rd32(priv, 0x10401c + (idx * 0x1000));
+	u32 intr = nv_rd32(priv, 0x104008 + (idx * 0x1000));
+	u32 stat = intr & disp & ~(disp >> 16);
+	u64 inst = nv_rd32(priv, 0x104050 + (idx * 0x1000)) & 0x0fffffff;
+	u32 ssta = nv_rd32(priv, 0x104040 + (idx * 0x1000)) & 0x0000ffff;
+	u32 addr = nv_rd32(priv, 0x104040 + (idx * 0x1000)) >> 16;
+	u32 mthd = (addr & 0x07ff) << 2;
+	u32 subc = (addr & 0x3800) >> 11;
+	u32 data = nv_rd32(priv, 0x104044 + (idx * 0x1000));
+	int chid;
+
+	engctx = nouveau_engctx_get(engine, inst);
+	chid   = pfifo->chid(pfifo, engctx);
+
+	if (stat & 0x00000040) {
+		nv_error(priv, "DISPATCH_ERROR [");
+		nouveau_enum_print(nvc0_copy_isr_error_name, ssta);
+		printk("] ch %d [0x%010llx] subc %d mthd 0x%04x data 0x%08x\n",
+		       chid, (u64)inst << 12, subc, mthd, data);
+		nv_wr32(priv, 0x104004 + (idx * 0x1000), 0x00000040);
+		stat &= ~0x00000040;
+	}
+
+	if (stat) {
+		nv_error(priv, "unhandled intr 0x%08x\n", stat);
+		nv_wr32(priv, 0x104004 + (idx * 0x1000), stat);
+	}
+
+	nouveau_engctx_put(engctx);
+}
+
+static int
+nvc0_copy0_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		struct nouveau_oclass *oclass, void *data, u32 size,
+		struct nouveau_object **pobject)
+{
+	struct nvc0_copy_priv *priv;
+	int ret;
+
+	if (nv_rd32(parent, 0x022500) & 0x00000100)
+		return -ENODEV;
+
+	ret = nouveau_copy_create(parent, engine, oclass, true, 0, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00000040;
+	nv_subdev(priv)->intr = nvc0_copy_intr;
+	nv_engine(priv)->cclass = &nvc0_copy0_cclass;
+	nv_engine(priv)->sclass = nvc0_copy0_sclass;
+	return 0;
+}
+
+static int
+nvc0_copy1_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		struct nouveau_oclass *oclass, void *data, u32 size,
+		struct nouveau_object **pobject)
+{
+	struct nvc0_copy_priv *priv;
+	int ret;
+
+	if (nv_rd32(parent, 0x022500) & 0x00000200)
+		return -ENODEV;
+
+	ret = nouveau_copy_create(parent, engine, oclass, true, 1, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00000080;
+	nv_subdev(priv)->intr = nvc0_copy_intr;
+	nv_engine(priv)->cclass = &nvc0_copy1_cclass;
+	nv_engine(priv)->sclass = nvc0_copy1_sclass;
+	return 0;
+}
+
+static int
+nvc0_copy_init(struct nouveau_object *object)
+{
+	int idx = nv_engidx(object) - NVDEV_ENGINE_COPY0;
+	struct nvc0_copy_priv *priv = (void *)object;
+	int ret, i;
+
+	ret = nouveau_copy_init(&priv->base);
+	if (ret)
+		return ret;
+
+	/* disable all interrupts */
+	nv_wr32(priv, 0x104014 + (idx * 0x1000), 0xffffffff);
+
+	/* upload ucode */
+	nv_wr32(priv, 0x1041c0 + (idx * 0x1000), 0x01000000);
+	for (i = 0; i < sizeof(nvc0_pcopy_data) / 4; i++)
+		nv_wr32(priv, 0x1041c4 + (idx * 0x1000), nvc0_pcopy_data[i]);
+
+	nv_wr32(priv, 0x104180 + (idx * 0x1000), 0x01000000);
+	for (i = 0; i < sizeof(nvc0_pcopy_code) / 4; i++) {
+		if ((i & 0x3f) == 0)
+			nv_wr32(priv, 0x104188 + (idx * 0x1000), i >> 6);
+		nv_wr32(priv, 0x104184 + (idx * 0x1000), nvc0_pcopy_code[i]);
+	}
+
+	/* start it running */
+	nv_wr32(priv, 0x104084 + (idx * 0x1000), idx);
+	nv_wr32(priv, 0x10410c + (idx * 0x1000), 0x00000000);
+	nv_wr32(priv, 0x104104 + (idx * 0x1000), 0x00000000); /* ENTRY */
+	nv_wr32(priv, 0x104100 + (idx * 0x1000), 0x00000002); /* TRIGGER */
+	return 0;
+}
+
+static int
+nvc0_copy_fini(struct nouveau_object *object, bool suspend)
+{
+	int idx = nv_engidx(object) - NVDEV_ENGINE_COPY0;
+	struct nvc0_copy_priv *priv = (void *)object;
+
+	nv_mask(priv, 0x104048 + (idx * 0x1000), 0x00000003, 0x00000000);
+	nv_wr32(priv, 0x104014 + (idx * 0x1000), 0xffffffff);
+
+	return nouveau_copy_fini(&priv->base, suspend);
+}
+
+struct nouveau_oclass
+nvc0_copy0_oclass = {
+	.handle = NV_ENGINE(COPY0, 0xc0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvc0_copy0_ctor,
+		.dtor = _nouveau_copy_dtor,
+		.init = nvc0_copy_init,
+		.fini = nvc0_copy_fini,
+	},
+};
+
+struct nouveau_oclass
+nvc0_copy1_oclass = {
+	.handle = NV_ENGINE(COPY1, 0xc0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvc0_copy1_ctor,
+		.dtor = _nouveau_copy_dtor,
+		.init = nvc0_copy_init,
+		.fini = nvc0_copy_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/copy/nve0.c b/drivers/gpu/drm/nouveau/core/engine/copy/nve0.c
new file mode 100644
index 0000000..2017c15
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/copy/nve0.c
@@ -0,0 +1,156 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/enum.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <engine/copy.h>
+
+struct nve0_copy_priv {
+	struct nouveau_copy base;
+};
+
+struct nve0_copy_chan {
+	struct nouveau_copy_chan base;
+};
+
+/*******************************************************************************
+ * Copy object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nve0_copy_sclass[] = {
+	{ 0xa0b5, &nouveau_object_ofuncs },
+	{},
+};
+
+/*******************************************************************************
+ * PCOPY context
+ ******************************************************************************/
+
+static int
+nve0_copy_context_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nve0_copy_chan *priv;
+	int ret;
+
+	ret = nouveau_copy_context_create(parent, engine, oclass, NULL, 256,
+					  256, NVOBJ_FLAG_ZERO_ALLOC, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static struct nouveau_ofuncs
+nve0_copy_context_ofuncs = {
+	.ctor = nve0_copy_context_ctor,
+	.dtor = _nouveau_copy_context_dtor,
+	.init = _nouveau_copy_context_init,
+	.fini = _nouveau_copy_context_fini,
+	.rd32 = _nouveau_copy_context_rd32,
+	.wr32 = _nouveau_copy_context_wr32,
+};
+
+static struct nouveau_oclass
+nve0_copy_cclass = {
+	.handle = NV_ENGCTX(COPY0, 0xc0),
+	.ofuncs = &nve0_copy_context_ofuncs,
+};
+
+/*******************************************************************************
+ * PCOPY engine/subdev functions
+ ******************************************************************************/
+
+static int
+nve0_copy0_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		struct nouveau_oclass *oclass, void *data, u32 size,
+		struct nouveau_object **pobject)
+{
+	struct nve0_copy_priv *priv;
+	int ret;
+
+	if (nv_rd32(parent, 0x022500) & 0x00000100)
+		return -ENODEV;
+
+	ret = nouveau_copy_create(parent, engine, oclass, true, 0, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00000040;
+	nv_engine(priv)->cclass = &nve0_copy_cclass;
+	nv_engine(priv)->sclass = nve0_copy_sclass;
+	return 0;
+}
+
+static int
+nve0_copy1_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		struct nouveau_oclass *oclass, void *data, u32 size,
+		struct nouveau_object **pobject)
+{
+	struct nve0_copy_priv *priv;
+	int ret;
+
+	if (nv_rd32(parent, 0x022500) & 0x00000200)
+		return -ENODEV;
+
+	ret = nouveau_copy_create(parent, engine, oclass, true, 1, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00000080;
+	nv_engine(priv)->cclass = &nve0_copy_cclass;
+	nv_engine(priv)->sclass = nve0_copy_sclass;
+	return 0;
+}
+
+struct nouveau_oclass
+nve0_copy0_oclass = {
+	.handle = NV_ENGINE(COPY0, 0xe0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nve0_copy0_ctor,
+		.dtor = _nouveau_copy_dtor,
+		.init = _nouveau_copy_init,
+		.fini = _nouveau_copy_fini,
+	},
+};
+
+struct nouveau_oclass
+nve0_copy1_oclass = {
+	.handle = NV_ENGINE(COPY1, 0xe0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nve0_copy1_ctor,
+		.dtor = _nouveau_copy_dtor,
+		.init = _nouveau_copy_init,
+		.fini = _nouveau_copy_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/crypt/fuc/nv98.fuc b/drivers/gpu/drm/nouveau/core/engine/crypt/fuc/nv98.fuc
new file mode 100644
index 0000000..629da02
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/crypt/fuc/nv98.fuc
@@ -0,0 +1,698 @@
+/*
+ *  fuc microcode for nv98 pcrypt engine
+ *  Copyright (C) 2010  Marcin Kościelnicki
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 of the License, or
+ *  (at your option) any later version.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program; if not, write to the Free Software
+ *  Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+
+.section #nv98_pcrypt_data
+
+ctx_dma:
+ctx_dma_query:		.b32 0
+ctx_dma_src:		.b32 0
+ctx_dma_dst:		.b32 0
+.equ #dma_count 3
+ctx_query_address_high:	.b32 0
+ctx_query_address_low:	.b32 0
+ctx_query_counter:	.b32 0
+ctx_cond_address_high:	.b32 0
+ctx_cond_address_low:	.b32 0
+ctx_cond_off:		.b32 0
+ctx_src_address_high:	.b32 0
+ctx_src_address_low:	.b32 0
+ctx_dst_address_high:	.b32 0
+ctx_dst_address_low:	.b32 0
+ctx_mode:		.b32 0
+.align 16
+ctx_key:		.skip 16
+ctx_iv:			.skip 16
+
+.align 0x80
+swap:
+.skip 32
+
+.align 8
+common_cmd_dtable:
+.b32 #ctx_query_address_high + 0x20000 ~0xff
+.b32 #ctx_query_address_low + 0x20000 ~0xfffffff0
+.b32 #ctx_query_counter + 0x20000 ~0xffffffff
+.b32 #cmd_query_get + 0x00000 ~1
+.b32 #ctx_cond_address_high + 0x20000 ~0xff
+.b32 #ctx_cond_address_low + 0x20000 ~0xfffffff0
+.b32 #cmd_cond_mode + 0x00000 ~7
+.b32 #cmd_wrcache_flush + 0x00000 ~0
+.equ #common_cmd_max 0x88
+
+
+.align 8
+engine_cmd_dtable:
+.b32 #ctx_key + 0x0 + 0x20000 ~0xffffffff
+.b32 #ctx_key + 0x4 + 0x20000 ~0xffffffff
+.b32 #ctx_key + 0x8 + 0x20000 ~0xffffffff
+.b32 #ctx_key + 0xc + 0x20000 ~0xffffffff
+.b32 #ctx_iv + 0x0 + 0x20000 ~0xffffffff
+.b32 #ctx_iv + 0x4 + 0x20000 ~0xffffffff
+.b32 #ctx_iv + 0x8 + 0x20000 ~0xffffffff
+.b32 #ctx_iv + 0xc + 0x20000 ~0xffffffff
+.b32 #ctx_src_address_high + 0x20000 ~0xff
+.b32 #ctx_src_address_low + 0x20000 ~0xfffffff0
+.b32 #ctx_dst_address_high + 0x20000 ~0xff
+.b32 #ctx_dst_address_low + 0x20000 ~0xfffffff0
+.b32 #crypt_cmd_mode + 0x00000 ~0xf
+.b32 #crypt_cmd_length + 0x10000 ~0x0ffffff0
+.equ #engine_cmd_max 0xce
+
+.align 4
+crypt_dtable:
+.b16 #crypt_copy_prep #crypt_do_inout
+.b16 #crypt_store_prep #crypt_do_out
+.b16 #crypt_ecb_e_prep #crypt_do_inout
+.b16 #crypt_ecb_d_prep #crypt_do_inout
+.b16 #crypt_cbc_e_prep #crypt_do_inout
+.b16 #crypt_cbc_d_prep #crypt_do_inout
+.b16 #crypt_pcbc_e_prep #crypt_do_inout
+.b16 #crypt_pcbc_d_prep #crypt_do_inout
+.b16 #crypt_cfb_e_prep #crypt_do_inout
+.b16 #crypt_cfb_d_prep #crypt_do_inout
+.b16 #crypt_ofb_prep #crypt_do_inout
+.b16 #crypt_ctr_prep #crypt_do_inout
+.b16 #crypt_cbc_mac_prep #crypt_do_in
+.b16 #crypt_cmac_finish_complete_prep #crypt_do_in
+.b16 #crypt_cmac_finish_partial_prep #crypt_do_in
+
+.align 0x100
+
+.section #nv98_pcrypt_code
+
+	// $r0 is always set to 0 in our code - this allows some space savings.
+	clear b32 $r0
+
+	// set up the interrupt handler
+	mov $r1 #ih
+	mov $iv0 $r1
+
+	// init stack pointer
+	mov $sp $r0
+
+	// set interrupt dispatch - route timer, fifo, ctxswitch to i0, others to host
+	movw $r1 0xfff0
+	sethi $r1 0
+	mov $r2 0x400
+	iowr I[$r2 + 0x300] $r1
+
+	// enable the interrupts
+	or $r1 0xc
+	iowr I[$r2] $r1
+
+	// enable fifo access and context switching
+	mov $r1 3
+	mov $r2 0x1200
+	iowr I[$r2] $r1
+
+	// enable i0 delivery
+	bset $flags ie0
+
+	// sleep forver, waking only for interrupts.
+	bset $flags $p0
+	spin:
+	sleep $p0
+	bra #spin
+
+// i0 handler
+ih:
+	// see which interrupts we got
+	iord $r1 I[$r0 + 0x200]
+
+	and $r2 $r1 0x8
+	cmpu b32 $r2 0
+	bra e #noctx
+
+		// context switch... prepare the regs for xfer
+		mov $r2 0x7700
+		mov $xtargets $r2
+		mov $xdbase $r0
+		// 128-byte context.
+		mov $r2 0
+		sethi $r2 0x50000
+
+		// read current channel
+		mov $r3 0x1400
+		iord $r4 I[$r3]
+		// if bit 30 set, it's active, so we have to unload it first.
+		shl b32 $r5 $r4 1
+		cmps b32 $r5 0
+		bra nc #ctxload
+
+			// unload the current channel - save the context
+			xdst $r0 $r2
+			xdwait
+			// and clear bit 30, then write back
+			bclr $r4 0x1e
+			iowr I[$r3] $r4
+			// tell PFIFO we unloaded
+			mov $r4 1
+			iowr I[$r3 + 0x200] $r4
+
+		bra #noctx
+
+		ctxload:
+			// no channel loaded - perhaps we're requested to load one
+			iord $r4 I[$r3 + 0x100]
+			shl b32 $r15 $r4 1
+			cmps b32 $r15 0
+			// if bit 30 of next channel not set, probably PFIFO is just
+			// killing a context. do a faux load, without the active bit.
+			bra nc #dummyload
+
+				// ok, do a real context load.
+				xdld $r0 $r2
+				xdwait
+				mov $r5 #ctx_dma
+				mov $r6 #dma_count - 1
+				ctxload_dma_loop:
+					ld b32 $r7 D[$r5 + $r6 * 4]
+					add b32 $r8 $r6 0x180
+					shl b32 $r8 8
+					iowr I[$r8] $r7
+					sub b32 $r6 1
+				bra nc #ctxload_dma_loop
+
+			dummyload:
+			// tell PFIFO we're done
+			mov $r5 2
+			iowr I[$r3 + 0x200] $r5
+
+	noctx:
+	and $r2 $r1 0x4
+	cmpu b32 $r2 0
+	bra e #nocmd
+
+		// incoming fifo command.
+		mov $r3 0x1900
+		iord $r2 I[$r3 + 0x100]
+		iord $r3 I[$r3]
+		// extract the method
+		and $r4 $r2 0x7ff
+		// shift the addr to proper position if we need to interrupt later
+		shl b32 $r2 0x10
+
+		// mthd 0 and 0x100 [NAME, NOP]: ignore
+		and $r5 $r4 0x7bf
+		cmpu b32 $r5 0
+		bra e #cmddone
+
+		mov $r5 #engine_cmd_dtable - 0xc0 * 8
+		mov $r6 #engine_cmd_max
+		cmpu b32 $r4 0xc0
+		bra nc #dtable_cmd
+		mov $r5 #common_cmd_dtable - 0x80 * 8
+		mov $r6 #common_cmd_max
+		cmpu b32 $r4 0x80
+		bra nc #dtable_cmd
+		cmpu b32 $r4 0x60
+		bra nc #dma_cmd
+		cmpu b32 $r4 0x50
+		bra ne #illegal_mthd
+
+			// mthd 0x140: PM_TRIGGER
+			mov $r2 0x2200
+			clear b32 $r3
+			sethi $r3 0x20000
+			iowr I[$r2] $r3
+			bra #cmddone
+
+		dma_cmd:
+			// mthd 0x180...: DMA_*
+			cmpu b32 $r4 0x60+#dma_count
+			bra nc #illegal_mthd
+			shl b32 $r5 $r4 2
+			add b32 $r5 ((#ctx_dma - 0x60 * 4) & 0xffff)
+			bset $r3 0x1e
+			st b32 D[$r5] $r3
+			add b32 $r4 0x180 - 0x60
+			shl b32 $r4 8
+			iowr I[$r4] $r3
+			bra #cmddone
+
+		dtable_cmd:
+			cmpu b32 $r4 $r6
+			bra nc #illegal_mthd
+			shl b32 $r4 3
+			add b32 $r4 $r5
+			ld b32 $r5 D[$r4 + 4]
+			and $r5 $r3
+			cmpu b32 $r5 0
+			bra ne #invalid_bitfield
+			ld b16 $r5 D[$r4]
+			ld b16 $r6 D[$r4 + 2]
+			cmpu b32 $r6 2
+			bra e #cmd_setctx
+			ld b32 $r7 D[$r0 + #ctx_cond_off]
+			and $r6 $r7
+			cmpu b32 $r6 1
+			bra e #cmddone
+			call $r5
+			bra $p1 #dispatch_error
+			bra #cmddone
+
+		cmd_setctx:
+			st b32 D[$r5] $r3
+			bra #cmddone
+
+
+		invalid_bitfield:
+			or $r2 1
+		dispatch_error:
+		illegal_mthd:
+			mov $r4 0x1000
+			iowr I[$r4] $r2
+			iowr I[$r4 + 0x100] $r3
+			mov $r4 0x40
+			iowr I[$r0] $r4
+
+			im_loop:
+				iord $r4 I[$r0 + 0x200]
+				and $r4 0x40
+				cmpu b32 $r4 0
+			bra ne #im_loop
+
+		cmddone:
+		// remove the command from FIFO
+		mov $r3 0x1d00
+		mov $r4 1
+		iowr I[$r3] $r4
+
+	nocmd:
+	// ack the processed interrupts
+	and $r1 $r1 0xc
+	iowr I[$r0 + 0x100] $r1
+iret
+
+cmd_query_get:
+	// if bit 0 of param set, trigger interrupt afterwards.
+	setp $p1 $r3
+	or $r2 3
+
+	// read PTIMER, beware of races...
+	mov $r4 0xb00
+	ptimer_retry:
+		iord $r6 I[$r4 + 0x100]
+		iord $r5 I[$r4]
+		iord $r7 I[$r4 + 0x100]
+		cmpu b32 $r6 $r7
+	bra ne #ptimer_retry
+
+	// prepare the query structure
+	ld b32 $r4 D[$r0 + #ctx_query_counter]
+	st b32 D[$r0 + #swap + 0x0] $r4
+	st b32 D[$r0 + #swap + 0x4] $r0
+	st b32 D[$r0 + #swap + 0x8] $r5
+	st b32 D[$r0 + #swap + 0xc] $r6
+
+	// will use target 0, DMA_QUERY.
+	mov $xtargets $r0
+
+	ld b32 $r4 D[$r0 + #ctx_query_address_high]
+	shl b32 $r4 0x18
+	mov $xdbase $r4
+
+	ld b32 $r4 D[$r0 + #ctx_query_address_low]
+	mov $r5 #swap
+	sethi $r5 0x20000
+	xdst $r4 $r5
+	xdwait
+
+	ret
+
+cmd_cond_mode:
+	// if >= 5, INVALID_ENUM
+	bset $flags $p1
+	or $r2 2
+	cmpu b32 $r3 5
+	bra nc #return
+
+	// otherwise, no error.
+	bclr $flags $p1
+
+	// if < 2, no QUERY object is involved
+	cmpu b32 $r3 2
+	bra nc #cmd_cond_mode_queryful
+
+		xor $r3 1
+		st b32 D[$r0 + #ctx_cond_off] $r3
+	return:
+		ret
+
+	cmd_cond_mode_queryful:
+	// ok, will need to pull a QUERY object, prepare offsets
+	ld b32 $r4 D[$r0 + #ctx_cond_address_high]
+	ld b32 $r5 D[$r0 + #ctx_cond_address_low]
+	and $r6 $r5 0xff
+	shr b32 $r5 8
+	shl b32 $r4 0x18
+	or $r4 $r5
+	mov $xdbase $r4
+	mov $xtargets $r0
+
+	// pull the first one
+	mov $r5 #swap
+	sethi $r5 0x20000
+	xdld $r6 $r5
+
+	// if == 2, only a single QUERY is involved...
+	cmpu b32 $r3 2
+	bra ne #cmd_cond_mode_double
+
+		xdwait
+		ld b32 $r4 D[$r0 + #swap + 4]
+		cmpu b32 $r4 0
+		xbit $r4 $flags z
+		st b32 D[$r0 + #ctx_cond_off] $r4
+		ret
+
+	// ok, we'll need to pull second one too
+	cmd_cond_mode_double:
+	add b32 $r6 0x10
+	add b32 $r5 0x10
+	xdld $r6 $r5
+	xdwait
+
+	// compare COUNTERs
+	ld b32 $r5 D[$r0 + #swap + 0x00]
+	ld b32 $r6 D[$r0 + #swap + 0x10]
+	cmpu b32 $r5 $r6
+	xbit $r4 $flags z
+
+	// compare RESen
+	ld b32 $r5 D[$r0 + #swap + 0x04]
+	ld b32 $r6 D[$r0 + #swap + 0x14]
+	cmpu b32 $r5 $r6
+	xbit $r5 $flags z
+	and $r4 $r5
+
+	// and negate or not, depending on mode
+	cmpu b32 $r3 3
+	xbit $r5 $flags z
+	xor $r4 $r5
+	st b32 D[$r0 + #ctx_cond_off] $r4
+	ret
+
+cmd_wrcache_flush:
+	bclr $flags $p1
+	mov $r2 0x2200
+	clear b32 $r3
+	sethi $r3 0x10000
+	iowr I[$r2] $r3
+	ret
+
+crypt_cmd_mode:
+	// if >= 0xf, INVALID_ENUM
+	bset $flags $p1
+	or $r2 2
+	cmpu b32 $r3 0xf
+	bra nc #crypt_cmd_mode_return
+
+		bclr $flags $p1
+		st b32 D[$r0 + #ctx_mode] $r3
+
+	crypt_cmd_mode_return:
+	ret
+
+crypt_cmd_length:
+	// nop if length == 0
+	cmpu b32 $r3 0
+	bra e #crypt_cmd_mode_return
+
+	// init key, IV
+	cxset 3
+	mov $r4 #ctx_key
+	sethi $r4 0x70000
+	xdst $r0 $r4
+	mov $r4 #ctx_iv
+	sethi $r4 0x60000
+	xdst $r0 $r4
+	xdwait
+	ckeyreg $c7
+
+	// prepare the targets
+	mov $r4 0x2100
+	mov $xtargets $r4
+
+	// prepare src address
+	ld b32 $r4 D[$r0 + #ctx_src_address_high]
+	ld b32 $r5 D[$r0 + #ctx_src_address_low]
+	shr b32 $r8 $r5 8
+	shl b32 $r4 0x18
+	or $r4 $r8
+	and $r5 $r5 0xff
+
+	// prepare dst address
+	ld b32 $r6 D[$r0 + #ctx_dst_address_high]
+	ld b32 $r7 D[$r0 + #ctx_dst_address_low]
+	shr b32 $r8 $r7 8
+	shl b32 $r6 0x18
+	or $r6 $r8
+	and $r7 $r7 0xff
+
+	// find the proper prep & do functions
+	ld b32 $r8 D[$r0 + #ctx_mode]
+	shl b32 $r8 2
+
+	// run prep
+	ld b16 $r9 D[$r8 + #crypt_dtable]
+	call $r9
+
+	// do it
+	ld b16 $r9 D[$r8 + #crypt_dtable + 2]
+	call $r9
+	cxset 1
+	xdwait
+	cxset 0x61
+	xdwait
+	xdwait
+
+	// update src address
+	shr b32 $r8 $r4 0x18
+	shl b32 $r9 $r4 8
+	add b32 $r9 $r5
+	adc b32 $r8 0
+	st b32 D[$r0 + #ctx_src_address_high] $r8
+	st b32 D[$r0 + #ctx_src_address_low] $r9
+
+	// update dst address
+	shr b32 $r8 $r6 0x18
+	shl b32 $r9 $r6 8
+	add b32 $r9 $r7
+	adc b32 $r8 0
+	st b32 D[$r0 + #ctx_dst_address_high] $r8
+	st b32 D[$r0 + #ctx_dst_address_low] $r9
+
+	// pull updated IV
+	cxset 2
+	mov $r4 #ctx_iv
+	sethi $r4 0x60000
+	xdld $r0 $r4
+	xdwait
+
+	ret
+
+
+crypt_copy_prep:
+	cs0begin 2
+		cxsin $c0
+		cxsout $c0
+	ret
+
+crypt_store_prep:
+	cs0begin 1
+		cxsout $c6
+	ret
+
+crypt_ecb_e_prep:
+	cs0begin 3
+		cxsin $c0
+		cenc $c0 $c0
+		cxsout $c0
+	ret
+
+crypt_ecb_d_prep:
+	ckexp $c7 $c7
+	cs0begin 3
+		cxsin $c0
+		cdec $c0 $c0
+		cxsout $c0
+	ret
+
+crypt_cbc_e_prep:
+	cs0begin 4
+		cxsin $c0
+		cxor $c6 $c0
+		cenc $c6 $c6
+		cxsout $c6
+	ret
+
+crypt_cbc_d_prep:
+	ckexp $c7 $c7
+	cs0begin 5
+		cmov $c2 $c6
+		cxsin $c6
+		cdec $c0 $c6
+		cxor $c0 $c2
+		cxsout $c0
+	ret
+
+crypt_pcbc_e_prep:
+	cs0begin 5
+		cxsin $c0
+		cxor $c6 $c0
+		cenc $c6 $c6
+		cxsout $c6
+		cxor $c6 $c0
+	ret
+
+crypt_pcbc_d_prep:
+	ckexp $c7 $c7
+	cs0begin 5
+		cxsin $c0
+		cdec $c1 $c0
+		cxor $c6 $c1
+		cxsout $c6
+		cxor $c6 $c0
+	ret
+
+crypt_cfb_e_prep:
+	cs0begin 4
+		cenc $c6 $c6
+		cxsin $c0
+		cxor $c6 $c0
+		cxsout $c6
+	ret
+
+crypt_cfb_d_prep:
+	cs0begin 4
+		cenc $c0 $c6
+		cxsin $c6
+		cxor $c0 $c6
+		cxsout $c0
+	ret
+
+crypt_ofb_prep:
+	cs0begin 4
+		cenc $c6 $c6
+		cxsin $c0
+		cxor $c0 $c6
+		cxsout $c0
+	ret
+
+crypt_ctr_prep:
+	cs0begin 5
+		cenc $c1 $c6
+		cadd $c6 1
+		cxsin $c0
+		cxor $c0 $c1
+		cxsout $c0
+	ret
+
+crypt_cbc_mac_prep:
+	cs0begin 3
+		cxsin $c0
+		cxor $c6 $c0
+		cenc $c6 $c6
+	ret
+
+crypt_cmac_finish_complete_prep:
+	cs0begin 7
+		cxsin $c0
+		cxor $c6 $c0
+		cxor $c0 $c0
+		cenc $c0 $c0
+		cprecmac $c0 $c0
+		cxor $c6 $c0
+		cenc $c6 $c6
+	ret
+
+crypt_cmac_finish_partial_prep:
+	cs0begin 8
+		cxsin $c0
+		cxor $c6 $c0
+		cxor $c0 $c0
+		cenc $c0 $c0
+		cprecmac $c0 $c0
+		cprecmac $c0 $c0
+		cxor $c6 $c0
+		cenc $c6 $c6
+	ret
+
+// TODO
+crypt_do_in:
+	add b32 $r3 $r5
+	mov $xdbase $r4
+	mov $r9 #swap
+	sethi $r9 0x20000
+	crypt_do_in_loop:
+		xdld $r5 $r9
+		xdwait
+		cxset 0x22
+		xdst $r0 $r9
+		cs0exec 1
+		xdwait
+		add b32 $r5 0x10
+		cmpu b32 $r5 $r3
+	bra ne #crypt_do_in_loop
+	cxset 1
+	xdwait
+	ret
+
+crypt_do_out:
+	add b32 $r3 $r7
+	mov $xdbase $r6
+	mov $r9 #swap
+	sethi $r9 0x20000
+	crypt_do_out_loop:
+		cs0exec 1
+		cxset 0x61
+		xdld $r7 $r9
+		xdst $r7 $r9
+		cxset 1
+		xdwait
+		add b32 $r7 0x10
+		cmpu b32 $r7 $r3
+	bra ne #crypt_do_out_loop
+	ret
+
+crypt_do_inout:
+	add b32 $r3 $r5
+	mov $r9 #swap
+	sethi $r9 0x20000
+	crypt_do_inout_loop:
+		mov $xdbase $r4
+		xdld $r5 $r9
+		xdwait
+		cxset 0x21
+		xdst $r0 $r9
+		cs0exec 1
+		cxset 0x61
+		mov $xdbase $r6
+		xdld $r7 $r9
+		xdst $r7 $r9
+		cxset 1
+		xdwait
+		add b32 $r5 0x10
+		add b32 $r7 0x10
+		cmpu b32 $r5 $r3
+	bra ne #crypt_do_inout_loop
+	ret
+
+.align 0x100
diff --git a/drivers/gpu/drm/nouveau/core/engine/crypt/fuc/nv98.fuc.h b/drivers/gpu/drm/nouveau/core/engine/crypt/fuc/nv98.fuc.h
new file mode 100644
index 0000000..09962e4
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/crypt/fuc/nv98.fuc.h
@@ -0,0 +1,584 @@
+static uint32_t nv98_pcrypt_data[] = {
+/* 0x0000: ctx_dma */
+/* 0x0000: ctx_dma_query */
+	0x00000000,
+/* 0x0004: ctx_dma_src */
+	0x00000000,
+/* 0x0008: ctx_dma_dst */
+	0x00000000,
+/* 0x000c: ctx_query_address_high */
+	0x00000000,
+/* 0x0010: ctx_query_address_low */
+	0x00000000,
+/* 0x0014: ctx_query_counter */
+	0x00000000,
+/* 0x0018: ctx_cond_address_high */
+	0x00000000,
+/* 0x001c: ctx_cond_address_low */
+	0x00000000,
+/* 0x0020: ctx_cond_off */
+	0x00000000,
+/* 0x0024: ctx_src_address_high */
+	0x00000000,
+/* 0x0028: ctx_src_address_low */
+	0x00000000,
+/* 0x002c: ctx_dst_address_high */
+	0x00000000,
+/* 0x0030: ctx_dst_address_low */
+	0x00000000,
+/* 0x0034: ctx_mode */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0040: ctx_key */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0050: ctx_iv */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0080: swap */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x00a0: common_cmd_dtable */
+	0x0002000c,
+	0xffffff00,
+	0x00020010,
+	0x0000000f,
+	0x00020014,
+	0x00000000,
+	0x00000192,
+	0xfffffffe,
+	0x00020018,
+	0xffffff00,
+	0x0002001c,
+	0x0000000f,
+	0x000001d7,
+	0xfffffff8,
+	0x00000260,
+	0xffffffff,
+/* 0x00e0: engine_cmd_dtable */
+	0x00020040,
+	0x00000000,
+	0x00020044,
+	0x00000000,
+	0x00020048,
+	0x00000000,
+	0x0002004c,
+	0x00000000,
+	0x00020050,
+	0x00000000,
+	0x00020054,
+	0x00000000,
+	0x00020058,
+	0x00000000,
+	0x0002005c,
+	0x00000000,
+	0x00020024,
+	0xffffff00,
+	0x00020028,
+	0x0000000f,
+	0x0002002c,
+	0xffffff00,
+	0x00020030,
+	0x0000000f,
+	0x00000271,
+	0xfffffff0,
+	0x00010285,
+	0xf000000f,
+/* 0x0150: crypt_dtable */
+	0x04db0321,
+	0x04b1032f,
+	0x04db0339,
+	0x04db034b,
+	0x04db0361,
+	0x04db0377,
+	0x04db0395,
+	0x04db03af,
+	0x04db03cd,
+	0x04db03e3,
+	0x04db03f9,
+	0x04db040f,
+	0x04830429,
+	0x0483043b,
+	0x0483045d,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+};
+
+static uint32_t nv98_pcrypt_code[] = {
+	0x17f004bd,
+	0x0010fe35,
+	0xf10004fe,
+	0xf0fff017,
+	0x27f10013,
+	0x21d00400,
+	0x0c15f0c0,
+	0xf00021d0,
+	0x27f10317,
+	0x21d01200,
+	0x1031f400,
+/* 0x002f: spin */
+	0xf40031f4,
+	0x0ef40028,
+/* 0x0035: ih */
+	0x8001cffd,
+	0xb00812c4,
+	0x0bf40024,
+	0x0027f167,
+	0x002bfe77,
+	0xf00007fe,
+	0x23f00027,
+	0x0037f105,
+	0x0034cf14,
+	0xb0014594,
+	0x18f40055,
+	0x0602fa17,
+	0x4af003f8,
+	0x0034d01e,
+	0xd00147f0,
+	0x0ef48034,
+/* 0x0075: ctxload */
+	0x4034cf33,
+	0xb0014f94,
+	0x18f400f5,
+	0x0502fa21,
+	0x57f003f8,
+	0x0267f000,
+/* 0x008c: ctxload_dma_loop */
+	0xa07856bc,
+	0xb6018068,
+	0x87d00884,
+	0x0162b600,
+/* 0x009f: dummyload */
+	0xf0f018f4,
+	0x35d00257,
+/* 0x00a5: noctx */
+	0x0412c480,
+	0xf50024b0,
+	0xf100df0b,
+	0xcf190037,
+	0x33cf4032,
+	0xff24e400,
+	0x1024b607,
+	0x07bf45e4,
+	0xf50054b0,
+	0xf100b90b,
+	0xf1fae057,
+	0xb000ce67,
+	0x18f4c044,
+	0xa057f14d,
+	0x8867f1fc,
+	0x8044b000,
+	0xb03f18f4,
+	0x18f46044,
+	0x5044b019,
+	0xf1741bf4,
+	0xbd220027,
+	0x0233f034,
+	0xf50023d0,
+/* 0x0103: dma_cmd */
+	0xb000810e,
+	0x18f46344,
+	0x0245945e,
+	0xfe8050b7,
+	0x801e39f0,
+	0x40b70053,
+	0x44b60120,
+	0x0043d008,
+/* 0x0123: dtable_cmd */
+	0xb8600ef4,
+	0x18f40446,
+	0x0344b63e,
+	0x980045bb,
+	0x53fd0145,
+	0x0054b004,
+	0x58291bf4,
+	0x46580045,
+	0x0264b001,
+	0x98170bf4,
+	0x67fd0807,
+	0x0164b004,
+	0xf9300bf4,
+	0x0f01f455,
+/* 0x015b: cmd_setctx */
+	0x80280ef4,
+	0x0ef40053,
+/* 0x0161: invalid_bitfield */
+	0x0125f022,
+/* 0x0164: dispatch_error */
+/* 0x0164: illegal_mthd */
+	0x100047f1,
+	0xd00042d0,
+	0x47f04043,
+	0x0004d040,
+/* 0x0174: im_loop */
+	0xf08004cf,
+	0x44b04044,
+	0xf71bf400,
+/* 0x0180: cmddone */
+	0x1d0037f1,
+	0xd00147f0,
+/* 0x018a: nocmd */
+	0x11c40034,
+	0x4001d00c,
+/* 0x0192: cmd_query_get */
+	0x38f201f8,
+	0x0325f001,
+	0x0b0047f1,
+/* 0x019c: ptimer_retry */
+	0xcf4046cf,
+	0x47cf0045,
+	0x0467b840,
+	0x98f41bf4,
+	0x04800504,
+	0x21008020,
+	0x80220580,
+	0x0bfe2306,
+	0x03049800,
+	0xfe1844b6,
+	0x04980047,
+	0x8057f104,
+	0x0253f000,
+	0xf80645fa,
+/* 0x01d7: cmd_cond_mode */
+	0xf400f803,
+	0x25f00131,
+	0x0534b002,
+	0xf41218f4,
+	0x34b00132,
+	0x0b18f402,
+	0x800136f0,
+/* 0x01f2: return */
+	0x00f80803,
+/* 0x01f4: cmd_cond_mode_queryful */
+	0x98060498,
+	0x56c40705,
+	0x0855b6ff,
+	0xfd1844b6,
+	0x47fe0545,
+	0x000bfe00,
+	0x008057f1,
+	0xfa0253f0,
+	0x34b00565,
+	0x131bf402,
+	0x049803f8,
+	0x0044b021,
+	0x800b4cf0,
+	0x00f80804,
+/* 0x022c: cmd_cond_mode_double */
+	0xb61060b6,
+	0x65fa1050,
+	0x9803f805,
+	0x06982005,
+	0x0456b824,
+	0x980b4cf0,
+	0x06982105,
+	0x0456b825,
+	0xfd0b5cf0,
+	0x34b00445,
+	0x0b5cf003,
+	0x800645fd,
+	0x00f80804,
+/* 0x0260: cmd_wrcache_flush */
+	0xf10132f4,
+	0xbd220027,
+	0x0133f034,
+	0xf80023d0,
+/* 0x0271: crypt_cmd_mode */
+	0x0131f400,
+	0xb00225f0,
+	0x18f40f34,
+	0x0132f409,
+/* 0x0283: crypt_cmd_mode_return */
+	0xf80d0380,
+/* 0x0285: crypt_cmd_length */
+	0x0034b000,
+	0xf4fb0bf4,
+	0x47f0033c,
+	0x0743f040,
+	0xf00604fa,
+	0x43f05047,
+	0x0604fa06,
+	0x3cf503f8,
+	0x47f1c407,
+	0x4bfe2100,
+	0x09049800,
+	0x950a0598,
+	0x44b60858,
+	0x0548fd18,
+	0x98ff55c4,
+	0x07980b06,
+	0x0878950c,
+	0xfd1864b6,
+	0x77c40568,
+	0x0d0898ff,
+	0x580284b6,
+	0x95f9a889,
+	0xf9a98958,
+	0x013cf495,
+	0x3cf403f8,
+	0xf803f861,
+	0x18489503,
+	0xbb084994,
+	0x81b60095,
+	0x09088000,
+	0x950a0980,
+	0x69941868,
+	0x0097bb08,
+	0x800081b6,
+	0x09800b08,
+	0x023cf40c,
+	0xf05047f0,
+	0x04fa0643,
+	0xf803f805,
+/* 0x0321: crypt_copy_prep */
+	0x203cf500,
+	0x003cf594,
+	0x003cf588,
+/* 0x032f: crypt_store_prep */
+	0xf500f88c,
+	0xf594103c,
+	0xf88c063c,
+/* 0x0339: crypt_ecb_e_prep */
+	0x303cf500,
+	0x003cf594,
+	0x003cf588,
+	0x003cf5d0,
+/* 0x034b: crypt_ecb_d_prep */
+	0xf500f88c,
+	0xf5c8773c,
+	0xf594303c,
+	0xf588003c,
+	0xf5d4003c,
+	0xf88c003c,
+/* 0x0361: crypt_cbc_e_prep */
+	0x403cf500,
+	0x003cf594,
+	0x063cf588,
+	0x663cf5ac,
+	0x063cf5d0,
+/* 0x0377: crypt_cbc_d_prep */
+	0xf500f88c,
+	0xf5c8773c,
+	0xf594503c,
+	0xf584623c,
+	0xf588063c,
+	0xf5d4603c,
+	0xf5ac203c,
+	0xf88c003c,
+/* 0x0395: crypt_pcbc_e_prep */
+	0x503cf500,
+	0x003cf594,
+	0x063cf588,
+	0x663cf5ac,
+	0x063cf5d0,
+	0x063cf58c,
+/* 0x03af: crypt_pcbc_d_prep */
+	0xf500f8ac,
+	0xf5c8773c,
+	0xf594503c,
+	0xf588003c,
+	0xf5d4013c,
+	0xf5ac163c,
+	0xf58c063c,
+	0xf8ac063c,
+/* 0x03cd: crypt_cfb_e_prep */
+	0x403cf500,
+	0x663cf594,
+	0x003cf5d0,
+	0x063cf588,
+	0x063cf5ac,
+/* 0x03e3: crypt_cfb_d_prep */
+	0xf500f88c,
+	0xf594403c,
+	0xf5d0603c,
+	0xf588063c,
+	0xf5ac603c,
+	0xf88c003c,
+/* 0x03f9: crypt_ofb_prep */
+	0x403cf500,
+	0x663cf594,
+	0x003cf5d0,
+	0x603cf588,
+	0x003cf5ac,
+/* 0x040f: crypt_ctr_prep */
+	0xf500f88c,
+	0xf594503c,
+	0xf5d0613c,
+	0xf5b0163c,
+	0xf588003c,
+	0xf5ac103c,
+	0xf88c003c,
+/* 0x0429: crypt_cbc_mac_prep */
+	0x303cf500,
+	0x003cf594,
+	0x063cf588,
+	0x663cf5ac,
+/* 0x043b: crypt_cmac_finish_complete_prep */
+	0xf500f8d0,
+	0xf594703c,
+	0xf588003c,
+	0xf5ac063c,
+	0xf5ac003c,
+	0xf5d0003c,
+	0xf5bc003c,
+	0xf5ac063c,
+	0xf8d0663c,
+/* 0x045d: crypt_cmac_finish_partial_prep */
+	0x803cf500,
+	0x003cf594,
+	0x063cf588,
+	0x003cf5ac,
+	0x003cf5ac,
+	0x003cf5d0,
+	0x003cf5bc,
+	0x063cf5bc,
+	0x663cf5ac,
+/* 0x0483: crypt_do_in */
+	0xbb00f8d0,
+	0x47fe0035,
+	0x8097f100,
+	0x0293f000,
+/* 0x0490: crypt_do_in_loop */
+	0xf80559fa,
+	0x223cf403,
+	0xf50609fa,
+	0xf898103c,
+	0x1050b603,
+	0xf40453b8,
+	0x3cf4e91b,
+	0xf803f801,
+/* 0x04b1: crypt_do_out */
+	0x0037bb00,
+	0xf10067fe,
+	0xf0008097,
+/* 0x04be: crypt_do_out_loop */
+	0x3cf50293,
+	0x3cf49810,
+	0x0579fa61,
+	0xf40679fa,
+	0x03f8013c,
+	0xb81070b6,
+	0x1bf40473,
+/* 0x04db: crypt_do_inout */
+	0xbb00f8e8,
+	0x97f10035,
+	0x93f00080,
+/* 0x04e5: crypt_do_inout_loop */
+	0x0047fe02,
+	0xf80559fa,
+	0x213cf403,
+	0xf50609fa,
+	0xf498103c,
+	0x67fe613c,
+	0x0579fa00,
+	0xf40679fa,
+	0x03f8013c,
+	0xb61050b6,
+	0x53b81070,
+	0xd41bf404,
+	0x000000f8,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/crypt/nv84.c b/drivers/gpu/drm/nouveau/core/engine/crypt/nv84.c
new file mode 100644
index 0000000..1d85e5b
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/crypt/nv84.c
@@ -0,0 +1,217 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/enum.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/gpuobj.h>
+
+#include <subdev/fb.h>
+
+#include <engine/fifo.h>
+#include <engine/crypt.h>
+
+struct nv84_crypt_priv {
+	struct nouveau_crypt base;
+};
+
+struct nv84_crypt_chan {
+	struct nouveau_crypt_chan base;
+};
+
+/*******************************************************************************
+ * Crypt object classes
+ ******************************************************************************/
+
+static int
+nv84_crypt_object_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nouveau_gpuobj *obj;
+	int ret;
+
+	ret = nouveau_gpuobj_create(parent, engine, oclass, 0, parent,
+				    16, 16, 0, &obj);
+	*pobject = nv_object(obj);
+	if (ret)
+		return ret;
+
+	nv_wo32(obj, 0x00, nv_mclass(obj));
+	nv_wo32(obj, 0x04, 0x00000000);
+	nv_wo32(obj, 0x08, 0x00000000);
+	nv_wo32(obj, 0x0c, 0x00000000);
+	return 0;
+}
+
+static struct nouveau_ofuncs
+nv84_crypt_ofuncs = {
+	.ctor = nv84_crypt_object_ctor,
+	.dtor = _nouveau_gpuobj_dtor,
+	.init = _nouveau_gpuobj_init,
+	.fini = _nouveau_gpuobj_fini,
+	.rd32 = _nouveau_gpuobj_rd32,
+	.wr32 = _nouveau_gpuobj_wr32,
+};
+
+static struct nouveau_oclass
+nv84_crypt_sclass[] = {
+	{ 0x74c1, &nv84_crypt_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * PCRYPT context
+ ******************************************************************************/
+
+static int
+nv84_crypt_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nv84_crypt_chan *priv;
+	int ret;
+
+	ret = nouveau_crypt_context_create(parent, engine, oclass, NULL, 256,
+					   0, NVOBJ_FLAG_ZERO_ALLOC, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static struct nouveau_oclass
+nv84_crypt_cclass = {
+	.handle = NV_ENGCTX(CRYPT, 0x84),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv84_crypt_context_ctor,
+		.dtor = _nouveau_crypt_context_dtor,
+		.init = _nouveau_crypt_context_init,
+		.fini = _nouveau_crypt_context_fini,
+		.rd32 = _nouveau_crypt_context_rd32,
+		.wr32 = _nouveau_crypt_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PCRYPT engine/subdev functions
+ ******************************************************************************/
+
+static const struct nouveau_bitfield nv84_crypt_intr_mask[] = {
+	{ 0x00000001, "INVALID_STATE" },
+	{ 0x00000002, "ILLEGAL_MTHD" },
+	{ 0x00000004, "ILLEGAL_CLASS" },
+	{ 0x00000080, "QUERY" },
+	{ 0x00000100, "FAULT" },
+	{}
+};
+
+static void
+nv84_crypt_intr(struct nouveau_subdev *subdev)
+{
+	struct nouveau_fifo *pfifo = nouveau_fifo(subdev);
+	struct nouveau_engine *engine = nv_engine(subdev);
+	struct nouveau_object *engctx;
+	struct nv84_crypt_priv *priv = (void *)subdev;
+	u32 stat = nv_rd32(priv, 0x102130);
+	u32 mthd = nv_rd32(priv, 0x102190);
+	u32 data = nv_rd32(priv, 0x102194);
+	u32 inst = nv_rd32(priv, 0x102188) & 0x7fffffff;
+	int chid;
+
+	engctx = nouveau_engctx_get(engine, inst);
+	chid   = pfifo->chid(pfifo, engctx);
+
+	if (stat) {
+		nv_error(priv, "");
+		nouveau_bitfield_print(nv84_crypt_intr_mask, stat);
+		printk(" ch %d [0x%010llx] mthd 0x%04x data 0x%08x\n",
+		       chid, (u64)inst << 12, mthd, data);
+	}
+
+	nv_wr32(priv, 0x102130, stat);
+	nv_wr32(priv, 0x10200c, 0x10);
+
+	nv50_fb_trap(nouveau_fb(priv), 1);
+	nouveau_engctx_put(engctx);
+}
+
+static int
+nv84_crypt_tlb_flush(struct nouveau_engine *engine)
+{
+	nv50_vm_flush_engine(&engine->base, 0x0a);
+	return 0;
+}
+
+static int
+nv84_crypt_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv84_crypt_priv *priv;
+	int ret;
+
+	ret = nouveau_crypt_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00004000;
+	nv_subdev(priv)->intr = nv84_crypt_intr;
+	nv_engine(priv)->cclass = &nv84_crypt_cclass;
+	nv_engine(priv)->sclass = nv84_crypt_sclass;
+	nv_engine(priv)->tlb_flush = nv84_crypt_tlb_flush;
+	return 0;
+}
+
+static int
+nv84_crypt_init(struct nouveau_object *object)
+{
+	struct nv84_crypt_priv *priv = (void *)object;
+	int ret;
+
+	ret = nouveau_crypt_init(&priv->base);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, 0x102130, 0xffffffff);
+	nv_wr32(priv, 0x102140, 0xffffffbf);
+	nv_wr32(priv, 0x10200c, 0x00000010);
+	return 0;
+}
+
+struct nouveau_oclass
+nv84_crypt_oclass = {
+	.handle = NV_ENGINE(CRYPT, 0x84),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv84_crypt_ctor,
+		.dtor = _nouveau_crypt_dtor,
+		.init = nv84_crypt_init,
+		.fini = _nouveau_crypt_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/crypt/nv98.c b/drivers/gpu/drm/nouveau/core/engine/crypt/nv98.c
new file mode 100644
index 0000000..9e3876c
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/crypt/nv98.c
@@ -0,0 +1,208 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/enum.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <subdev/timer.h>
+#include <subdev/fb.h>
+
+#include <engine/fifo.h>
+#include <engine/crypt.h>
+
+#include "fuc/nv98.fuc.h"
+
+struct nv98_crypt_priv {
+	struct nouveau_crypt base;
+};
+
+struct nv98_crypt_chan {
+	struct nouveau_crypt_chan base;
+};
+
+/*******************************************************************************
+ * Crypt object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv98_crypt_sclass[] = {
+	{ 0x88b4, &nouveau_object_ofuncs },
+	{},
+};
+
+/*******************************************************************************
+ * PCRYPT context
+ ******************************************************************************/
+
+static int
+nv98_crypt_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nv98_crypt_chan *priv;
+	int ret;
+
+	ret = nouveau_crypt_context_create(parent, engine, oclass, NULL, 256,
+					   256, NVOBJ_FLAG_ZERO_ALLOC, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static struct nouveau_oclass
+nv98_crypt_cclass = {
+	.handle = NV_ENGCTX(CRYPT, 0x98),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv98_crypt_context_ctor,
+		.dtor = _nouveau_crypt_context_dtor,
+		.init = _nouveau_crypt_context_init,
+		.fini = _nouveau_crypt_context_fini,
+		.rd32 = _nouveau_crypt_context_rd32,
+		.wr32 = _nouveau_crypt_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PCRYPT engine/subdev functions
+ ******************************************************************************/
+
+static const struct nouveau_enum nv98_crypt_isr_error_name[] = {
+	{ 0x0000, "ILLEGAL_MTHD" },
+	{ 0x0001, "INVALID_BITFIELD" },
+	{ 0x0002, "INVALID_ENUM" },
+	{ 0x0003, "QUERY" },
+	{}
+};
+
+static void
+nv98_crypt_intr(struct nouveau_subdev *subdev)
+{
+	struct nouveau_fifo *pfifo = nouveau_fifo(subdev);
+	struct nouveau_engine *engine = nv_engine(subdev);
+	struct nouveau_object *engctx;
+	struct nv98_crypt_priv *priv = (void *)subdev;
+	u32 disp = nv_rd32(priv, 0x08701c);
+	u32 stat = nv_rd32(priv, 0x087008) & disp & ~(disp >> 16);
+	u32 inst = nv_rd32(priv, 0x087050) & 0x3fffffff;
+	u32 ssta = nv_rd32(priv, 0x087040) & 0x0000ffff;
+	u32 addr = nv_rd32(priv, 0x087040) >> 16;
+	u32 mthd = (addr & 0x07ff) << 2;
+	u32 subc = (addr & 0x3800) >> 11;
+	u32 data = nv_rd32(priv, 0x087044);
+	int chid;
+
+	engctx = nouveau_engctx_get(engine, inst);
+	chid   = pfifo->chid(pfifo, engctx);
+
+	if (stat & 0x00000040) {
+		nv_error(priv, "DISPATCH_ERROR [");
+		nouveau_enum_print(nv98_crypt_isr_error_name, ssta);
+		printk("] ch %d [0x%010llx] subc %d mthd 0x%04x data 0x%08x\n",
+		       chid, (u64)inst << 12, subc, mthd, data);
+		nv_wr32(priv, 0x087004, 0x00000040);
+		stat &= ~0x00000040;
+	}
+
+	if (stat) {
+		nv_error(priv, "unhandled intr 0x%08x\n", stat);
+		nv_wr32(priv, 0x087004, stat);
+	}
+
+	nv50_fb_trap(nouveau_fb(priv), 1);
+	nouveau_engctx_put(engctx);
+}
+
+static int
+nv98_crypt_tlb_flush(struct nouveau_engine *engine)
+{
+	nv50_vm_flush_engine(&engine->base, 0x0a);
+	return 0;
+}
+
+static int
+nv98_crypt_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv98_crypt_priv *priv;
+	int ret;
+
+	ret = nouveau_crypt_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00004000;
+	nv_subdev(priv)->intr = nv98_crypt_intr;
+	nv_engine(priv)->cclass = &nv98_crypt_cclass;
+	nv_engine(priv)->sclass = nv98_crypt_sclass;
+	nv_engine(priv)->tlb_flush = nv98_crypt_tlb_flush;
+	return 0;
+}
+
+static int
+nv98_crypt_init(struct nouveau_object *object)
+{
+	struct nv98_crypt_priv *priv = (void *)object;
+	int ret, i;
+
+	ret = nouveau_crypt_init(&priv->base);
+	if (ret)
+		return ret;
+
+	/* wait for exit interrupt to signal */
+	nv_wait(priv, 0x087008, 0x00000010, 0x00000010);
+	nv_wr32(priv, 0x087004, 0x00000010);
+
+	/* upload microcode code and data segments */
+	nv_wr32(priv, 0x087ff8, 0x00100000);
+	for (i = 0; i < ARRAY_SIZE(nv98_pcrypt_code); i++)
+		nv_wr32(priv, 0x087ff4, nv98_pcrypt_code[i]);
+
+	nv_wr32(priv, 0x087ff8, 0x00000000);
+	for (i = 0; i < ARRAY_SIZE(nv98_pcrypt_data); i++)
+		nv_wr32(priv, 0x087ff4, nv98_pcrypt_data[i]);
+
+	/* start it running */
+	nv_wr32(priv, 0x08710c, 0x00000000);
+	nv_wr32(priv, 0x087104, 0x00000000); /* ENTRY */
+	nv_wr32(priv, 0x087100, 0x00000002); /* TRIGGER */
+	return 0;
+}
+
+struct nouveau_oclass
+nv98_crypt_oclass = {
+	.handle = NV_ENGINE(CRYPT, 0x98),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv98_crypt_ctor,
+		.dtor = _nouveau_crypt_dtor,
+		.init = nv98_crypt_init,
+		.fini = _nouveau_crypt_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/disp/nv04.c b/drivers/gpu/drm/nouveau/core/engine/disp/nv04.c
new file mode 100644
index 0000000..1c919f2
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/disp/nv04.c
@@ -0,0 +1,90 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <engine/disp.h>
+
+struct nv04_disp_priv {
+	struct nouveau_disp base;
+};
+
+static struct nouveau_oclass
+nv04_disp_sclass[] = {
+	{},
+};
+
+static void
+nv04_disp_intr_vblank(struct nv04_disp_priv *priv, int crtc)
+{
+	struct nouveau_disp *disp = &priv->base;
+	if (disp->vblank.notify)
+		disp->vblank.notify(disp->vblank.data, crtc);
+}
+
+static void
+nv04_disp_intr(struct nouveau_subdev *subdev)
+{
+	struct nv04_disp_priv *priv = (void *)subdev;
+	u32 crtc0 = nv_rd32(priv, 0x600100);
+	u32 crtc1 = nv_rd32(priv, 0x602100);
+
+	if (crtc0 & 0x00000001) {
+		nv04_disp_intr_vblank(priv, 0);
+		nv_wr32(priv, 0x600100, 0x00000001);
+	}
+
+	if (crtc1 & 0x00000001) {
+		nv04_disp_intr_vblank(priv, 1);
+		nv_wr32(priv, 0x602100, 0x00000001);
+	}
+}
+
+static int
+nv04_disp_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		  struct nouveau_oclass *oclass, void *data, u32 size,
+		  struct nouveau_object **pobject)
+{
+	struct nv04_disp_priv *priv;
+	int ret;
+
+	ret = nouveau_disp_create(parent, engine, oclass, "DISPLAY",
+				  "display", &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_engine(priv)->sclass = nv04_disp_sclass;
+	nv_subdev(priv)->intr = nv04_disp_intr;
+	return 0;
+}
+
+struct nouveau_oclass
+nv04_disp_oclass = {
+	.handle = NV_ENGINE(DISP, 0x04),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv04_disp_ctor,
+		.dtor = _nouveau_disp_dtor,
+		.init = _nouveau_disp_init,
+		.fini = _nouveau_disp_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/disp/nv50.c b/drivers/gpu/drm/nouveau/core/engine/disp/nv50.c
new file mode 100644
index 0000000..16a9afb
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/disp/nv50.c
@@ -0,0 +1,125 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <engine/software.h>
+#include <engine/disp.h>
+
+struct nv50_disp_priv {
+	struct nouveau_disp base;
+};
+
+static struct nouveau_oclass
+nv50_disp_sclass[] = {
+	{},
+};
+
+static void
+nv50_disp_intr_vblank(struct nv50_disp_priv *priv, int crtc)
+{
+	struct nouveau_disp *disp = &priv->base;
+	struct nouveau_software_chan *chan, *temp;
+	unsigned long flags;
+
+	spin_lock_irqsave(&disp->vblank.lock, flags);
+	list_for_each_entry_safe(chan, temp, &disp->vblank.list, vblank.head) {
+		if (chan->vblank.crtc != crtc)
+			continue;
+
+		nv_wr32(priv, 0x001704, chan->vblank.channel);
+		nv_wr32(priv, 0x001710, 0x80000000 | chan->vblank.ctxdma);
+
+		if (nv_device(priv)->chipset == 0x50) {
+			nv_wr32(priv, 0x001570, chan->vblank.offset);
+			nv_wr32(priv, 0x001574, chan->vblank.value);
+		} else {
+			if (nv_device(priv)->chipset >= 0xc0) {
+				nv_wr32(priv, 0x06000c,
+					upper_32_bits(chan->vblank.offset));
+			}
+			nv_wr32(priv, 0x060010, chan->vblank.offset);
+			nv_wr32(priv, 0x060014, chan->vblank.value);
+		}
+
+		list_del(&chan->vblank.head);
+		if (disp->vblank.put)
+			disp->vblank.put(disp->vblank.data, crtc);
+	}
+	spin_unlock_irqrestore(&disp->vblank.lock, flags);
+
+	if (disp->vblank.notify)
+		disp->vblank.notify(disp->vblank.data, crtc);
+}
+
+static void
+nv50_disp_intr(struct nouveau_subdev *subdev)
+{
+	struct nv50_disp_priv *priv = (void *)subdev;
+	u32 stat1 = nv_rd32(priv, 0x610024);
+
+	if (stat1 & 0x00000004) {
+		nv50_disp_intr_vblank(priv, 0);
+		nv_wr32(priv, 0x610024, 0x00000004);
+		stat1 &= ~0x00000004;
+	}
+
+	if (stat1 & 0x00000008) {
+		nv50_disp_intr_vblank(priv, 1);
+		nv_wr32(priv, 0x610024, 0x00000008);
+		stat1 &= ~0x00000008;
+	}
+
+}
+
+static int
+nv50_disp_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		  struct nouveau_oclass *oclass, void *data, u32 size,
+		  struct nouveau_object **pobject)
+{
+	struct nv50_disp_priv *priv;
+	int ret;
+
+	ret = nouveau_disp_create(parent, engine, oclass, "PDISP",
+				  "display", &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_engine(priv)->sclass = nv50_disp_sclass;
+	nv_subdev(priv)->intr = nv50_disp_intr;
+
+	INIT_LIST_HEAD(&priv->base.vblank.list);
+	spin_lock_init(&priv->base.vblank.lock);
+	return 0;
+}
+
+struct nouveau_oclass
+nv50_disp_oclass = {
+	.handle = NV_ENGINE(DISP, 0x50),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv50_disp_ctor,
+		.dtor = _nouveau_disp_dtor,
+		.init = _nouveau_disp_init,
+		.fini = _nouveau_disp_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/disp/nvd0.c b/drivers/gpu/drm/nouveau/core/engine/disp/nvd0.c
new file mode 100644
index 0000000..d93efbc
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/disp/nvd0.c
@@ -0,0 +1,118 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <subdev/bar.h>
+
+#include <engine/software.h>
+#include <engine/disp.h>
+
+struct nvd0_disp_priv {
+	struct nouveau_disp base;
+};
+
+static struct nouveau_oclass
+nvd0_disp_sclass[] = {
+	{},
+};
+
+static void
+nvd0_disp_intr_vblank(struct nvd0_disp_priv *priv, int crtc)
+{
+	struct nouveau_bar *bar = nouveau_bar(priv);
+	struct nouveau_disp *disp = &priv->base;
+	struct nouveau_software_chan *chan, *temp;
+	unsigned long flags;
+
+	spin_lock_irqsave(&disp->vblank.lock, flags);
+	list_for_each_entry_safe(chan, temp, &disp->vblank.list, vblank.head) {
+		if (chan->vblank.crtc != crtc)
+			continue;
+
+		nv_wr32(priv, 0x001718, 0x80000000 | chan->vblank.channel);
+		bar->flush(bar);
+		nv_wr32(priv, 0x06000c, upper_32_bits(chan->vblank.offset));
+		nv_wr32(priv, 0x060010, lower_32_bits(chan->vblank.offset));
+		nv_wr32(priv, 0x060014, chan->vblank.value);
+
+		list_del(&chan->vblank.head);
+		if (disp->vblank.put)
+			disp->vblank.put(disp->vblank.data, crtc);
+	}
+	spin_unlock_irqrestore(&disp->vblank.lock, flags);
+
+	if (disp->vblank.notify)
+		disp->vblank.notify(disp->vblank.data, crtc);
+}
+
+static void
+nvd0_disp_intr(struct nouveau_subdev *subdev)
+{
+	struct nvd0_disp_priv *priv = (void *)subdev;
+	u32 intr = nv_rd32(priv, 0x610088);
+	int i;
+
+	for (i = 0; i < 4; i++) {
+		u32 mask = 0x01000000 << i;
+		if (mask & intr) {
+			u32 stat = nv_rd32(priv, 0x6100bc + (i * 0x800));
+			if (stat & 0x00000001)
+				nvd0_disp_intr_vblank(priv, i);
+			nv_mask(priv, 0x6100bc + (i * 0x800), 0, 0);
+			nv_rd32(priv, 0x6100c0 + (i * 0x800));
+		}
+	}
+}
+
+static int
+nvd0_disp_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		  struct nouveau_oclass *oclass, void *data, u32 size,
+		  struct nouveau_object **pobject)
+{
+	struct nvd0_disp_priv *priv;
+	int ret;
+
+	ret = nouveau_disp_create(parent, engine, oclass, "PDISP",
+				  "display", &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_engine(priv)->sclass = nvd0_disp_sclass;
+	nv_subdev(priv)->intr = nvd0_disp_intr;
+
+	INIT_LIST_HEAD(&priv->base.vblank.list);
+	spin_lock_init(&priv->base.vblank.lock);
+	return 0;
+}
+
+struct nouveau_oclass
+nvd0_disp_oclass = {
+	.handle = NV_ENGINE(DISP, 0xd0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvd0_disp_ctor,
+		.dtor = _nouveau_disp_dtor,
+		.init = _nouveau_disp_init,
+		.fini = _nouveau_disp_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/disp/vga.c b/drivers/gpu/drm/nouveau/core/engine/disp/vga.c
new file mode 100644
index 0000000..5a1c684
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/disp/vga.c
@@ -0,0 +1,215 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/subdev.h>
+#include <core/device.h>
+#include <subdev/vga.h>
+
+u8
+nv_rdport(void *obj, int head, u16 port)
+{
+	struct nouveau_device *device = nv_device(obj);
+
+	if (device->card_type >= NV_50)
+		return nv_rd08(obj, 0x601000 + port);
+
+	if (port == 0x03c0 || port == 0x03c1 ||	/* AR */
+	    port == 0x03c2 || port == 0x03da ||	/* INP0 */
+	    port == 0x03d4 || port == 0x03d5)	/* CR */
+		return nv_rd08(obj, 0x601000 + (head * 0x2000) + port);
+
+	if (port == 0x03c2 || port == 0x03cc ||	/* MISC */
+	    port == 0x03c4 || port == 0x03c5 ||	/* SR */
+	    port == 0x03ce || port == 0x03cf) {	/* GR */
+		if (device->card_type < NV_40)
+			head = 0; /* CR44 selects head */
+		return nv_rd08(obj, 0x0c0000 + (head * 0x2000) + port);
+	}
+
+	nv_error(obj, "unknown vga port 0x%04x\n", port);
+	return 0x00;
+}
+
+void
+nv_wrport(void *obj, int head, u16 port, u8 data)
+{
+	struct nouveau_device *device = nv_device(obj);
+
+	if (device->card_type >= NV_50)
+		nv_wr08(obj, 0x601000 + port, data);
+	else
+	if (port == 0x03c0 || port == 0x03c1 ||	/* AR */
+	    port == 0x03c2 || port == 0x03da ||	/* INP0 */
+	    port == 0x03d4 || port == 0x03d5)	/* CR */
+		nv_wr08(obj, 0x601000 + (head * 0x2000) + port, data);
+	else
+	if (port == 0x03c2 || port == 0x03cc ||	/* MISC */
+	    port == 0x03c4 || port == 0x03c5 ||	/* SR */
+	    port == 0x03ce || port == 0x03cf) {	/* GR */
+		if (device->card_type < NV_40)
+			head = 0; /* CR44 selects head */
+		nv_wr08(obj, 0x0c0000 + (head * 0x2000) + port, data);
+	} else
+		nv_error(obj, "unknown vga port 0x%04x\n", port);
+}
+
+u8
+nv_rdvgas(void *obj, int head, u8 index)
+{
+	nv_wrport(obj, head, 0x03c4, index);
+	return nv_rdport(obj, head, 0x03c5);
+}
+
+void
+nv_wrvgas(void *obj, int head, u8 index, u8 value)
+{
+	nv_wrport(obj, head, 0x03c4, index);
+	nv_wrport(obj, head, 0x03c5, value);
+}
+
+u8
+nv_rdvgag(void *obj, int head, u8 index)
+{
+	nv_wrport(obj, head, 0x03ce, index);
+	return nv_rdport(obj, head, 0x03cf);
+}
+
+void
+nv_wrvgag(void *obj, int head, u8 index, u8 value)
+{
+	nv_wrport(obj, head, 0x03ce, index);
+	nv_wrport(obj, head, 0x03cf, value);
+}
+
+u8
+nv_rdvgac(void *obj, int head, u8 index)
+{
+	nv_wrport(obj, head, 0x03d4, index);
+	return nv_rdport(obj, head, 0x03d5);
+}
+
+void
+nv_wrvgac(void *obj, int head, u8 index, u8 value)
+{
+	nv_wrport(obj, head, 0x03d4, index);
+	nv_wrport(obj, head, 0x03d5, value);
+}
+
+u8
+nv_rdvgai(void *obj, int head, u16 port, u8 index)
+{
+	if (port == 0x03c4) return nv_rdvgas(obj, head, index);
+	if (port == 0x03ce) return nv_rdvgag(obj, head, index);
+	if (port == 0x03d4) return nv_rdvgac(obj, head, index);
+	nv_error(obj, "unknown indexed vga port 0x%04x\n", port);
+	return 0x00;
+}
+
+void
+nv_wrvgai(void *obj, int head, u16 port, u8 index, u8 value)
+{
+	if      (port == 0x03c4) nv_wrvgas(obj, head, index, value);
+	else if (port == 0x03ce) nv_wrvgag(obj, head, index, value);
+	else if (port == 0x03d4) nv_wrvgac(obj, head, index, value);
+	else nv_error(obj, "unknown indexed vga port 0x%04x\n", port);
+}
+
+bool
+nv_lockvgac(void *obj, bool lock)
+{
+	bool locked = !nv_rdvgac(obj, 0, 0x1f);
+	u8 data = lock ? 0x99 : 0x57;
+	nv_wrvgac(obj, 0, 0x1f, data);
+	if (nv_device(obj)->chipset == 0x11) {
+		if (!(nv_rd32(obj, 0x001084) & 0x10000000))
+			nv_wrvgac(obj, 1, 0x1f, data);
+	}
+	return locked;
+}
+
+/* CR44 takes values 0 (head A), 3 (head B) and 4 (heads tied)
+ * it affects only the 8 bit vga io regs, which we access using mmio at
+ * 0xc{0,2}3c*, 0x60{1,3}3*, and 0x68{1,3}3d*
+ * in general, the set value of cr44 does not matter: reg access works as
+ * expected and values can be set for the appropriate head by using a 0x2000
+ * offset as required
+ * however:
+ * a) pre nv40, the head B range of PRMVIO regs at 0xc23c* was not exposed and
+ *    cr44 must be set to 0 or 3 for accessing values on the correct head
+ *    through the common 0xc03c* addresses
+ * b) in tied mode (4) head B is programmed to the values set on head A, and
+ *    access using the head B addresses can have strange results, ergo we leave
+ *    tied mode in init once we know to what cr44 should be restored on exit
+ *
+ * the owner parameter is slightly abused:
+ * 0 and 1 are treated as head values and so the set value is (owner * 3)
+ * other values are treated as literal values to set
+ */
+u8
+nv_rdvgaowner(void *obj)
+{
+	if (nv_device(obj)->card_type < NV_50) {
+		if (nv_device(obj)->chipset == 0x11) {
+			u32 tied = nv_rd32(obj, 0x001084) & 0x10000000;
+			if (tied == 0) {
+				u8 slA = nv_rdvgac(obj, 0, 0x28) & 0x80;
+				u8 tvA = nv_rdvgac(obj, 0, 0x33) & 0x01;
+				u8 slB = nv_rdvgac(obj, 1, 0x28) & 0x80;
+				u8 tvB = nv_rdvgac(obj, 1, 0x33) & 0x01;
+				if (slA && !tvA) return 0x00;
+				if (slB && !tvB) return 0x03;
+				if (slA) return 0x00;
+				if (slB) return 0x03;
+				return 0x00;
+			}
+			return 0x04;
+		}
+
+		return nv_rdvgac(obj, 0, 0x44);
+	}
+
+	nv_error(obj, "rdvgaowner after nv4x\n");
+	return 0x00;
+}
+
+void
+nv_wrvgaowner(void *obj, u8 select)
+{
+	if (nv_device(obj)->card_type < NV_50) {
+		u8 owner = (select == 1) ? 3 : select;
+		if (nv_device(obj)->chipset == 0x11) {
+			/* workaround hw lockup bug */
+			nv_rdvgac(obj, 0, 0x1f);
+			nv_rdvgac(obj, 1, 0x1f);
+		}
+
+		nv_wrvgac(obj, 0, 0x44, owner);
+
+		if (nv_device(obj)->chipset == 0x11) {
+			nv_wrvgac(obj, 0, 0x2e, owner);
+			nv_wrvgac(obj, 0, 0x2e, owner);
+		}
+	} else
+		nv_error(obj, "wrvgaowner after nv4x\n");
+}
diff --git a/drivers/gpu/drm/nouveau/core/engine/dmaobj/base.c b/drivers/gpu/drm/nouveau/core/engine/dmaobj/base.c
new file mode 100644
index 0000000..e1f013d
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/dmaobj/base.c
@@ -0,0 +1,87 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/object.h>
+#include <core/class.h>
+
+#include <subdev/fb.h>
+#include <engine/dmaobj.h>
+
+int
+nouveau_dmaobj_create_(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass,
+		       void *data, u32 size, int len, void **pobject)
+{
+	struct nv_dma_class *args = data;
+	struct nouveau_dmaobj *object;
+	int ret;
+
+	if (size < sizeof(*args))
+		return -EINVAL;
+
+	ret = nouveau_object_create_(parent, engine, oclass, 0, len, pobject);
+	object = *pobject;
+	if (ret)
+		return ret;
+
+	switch (args->flags & NV_DMA_TARGET_MASK) {
+	case NV_DMA_TARGET_VM:
+		object->target = NV_MEM_TARGET_VM;
+		break;
+	case NV_DMA_TARGET_VRAM:
+		object->target = NV_MEM_TARGET_VRAM;
+		break;
+	case NV_DMA_TARGET_PCI:
+		object->target = NV_MEM_TARGET_PCI;
+		break;
+	case NV_DMA_TARGET_PCI_US:
+	case NV_DMA_TARGET_AGP:
+		object->target = NV_MEM_TARGET_PCI_NOSNOOP;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	switch (args->flags & NV_DMA_ACCESS_MASK) {
+	case NV_DMA_ACCESS_VM:
+		object->access = NV_MEM_ACCESS_VM;
+		break;
+	case NV_DMA_ACCESS_RD:
+		object->access = NV_MEM_ACCESS_RO;
+		break;
+	case NV_DMA_ACCESS_WR:
+		object->access = NV_MEM_ACCESS_WO;
+		break;
+	case NV_DMA_ACCESS_RDWR:
+		object->access = NV_MEM_ACCESS_RW;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	object->start = args->start;
+	object->limit = args->limit;
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/core/engine/dmaobj/nv04.c b/drivers/gpu/drm/nouveau/core/engine/dmaobj/nv04.c
new file mode 100644
index 0000000..9f4cc2f
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/dmaobj/nv04.c
@@ -0,0 +1,185 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/gpuobj.h>
+#include <core/class.h>
+
+#include <subdev/fb.h>
+#include <subdev/vm/nv04.h>
+
+#include <engine/dmaobj.h>
+
+struct nv04_dmaeng_priv {
+	struct nouveau_dmaeng base;
+};
+
+struct nv04_dmaobj_priv {
+	struct nouveau_dmaobj base;
+};
+
+static int
+nv04_dmaobj_bind(struct nouveau_dmaeng *dmaeng,
+		 struct nouveau_object *parent,
+		 struct nouveau_dmaobj *dmaobj,
+		 struct nouveau_gpuobj **pgpuobj)
+{
+	struct nv04_vmmgr_priv *vmm = nv04_vmmgr(dmaeng);
+	struct nouveau_gpuobj *gpuobj;
+	u32 flags0 = nv_mclass(dmaobj);
+	u32 flags2 = 0x00000000;
+	u64 offset = dmaobj->start & 0xfffff000;
+	u64 adjust = dmaobj->start & 0x00000fff;
+	u32 length = dmaobj->limit - dmaobj->start;
+	int ret;
+
+	if (dmaobj->target == NV_MEM_TARGET_VM) {
+		if (nv_object(vmm)->oclass == &nv04_vmmgr_oclass) {
+			struct nouveau_gpuobj *pgt = vmm->vm->pgt[0].obj[0];
+			if (!dmaobj->start)
+				return nouveau_gpuobj_dup(parent, pgt, pgpuobj);
+			offset  = nv_ro32(pgt, 8 + (offset >> 10));
+			offset &= 0xfffff000;
+		}
+
+		dmaobj->target = NV_MEM_TARGET_PCI;
+		dmaobj->access = NV_MEM_ACCESS_RW;
+	}
+
+	switch (dmaobj->target) {
+	case NV_MEM_TARGET_VRAM:
+		flags0 |= 0x00003000;
+		break;
+	case NV_MEM_TARGET_PCI:
+		flags0 |= 0x00023000;
+		break;
+	case NV_MEM_TARGET_PCI_NOSNOOP:
+		flags0 |= 0x00033000;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	switch (dmaobj->access) {
+	case NV_MEM_ACCESS_RO:
+		flags0 |= 0x00004000;
+		break;
+	case NV_MEM_ACCESS_WO:
+		flags0 |= 0x00008000;
+	case NV_MEM_ACCESS_RW:
+		flags2 |= 0x00000002;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	ret = nouveau_gpuobj_new(parent, parent, 16, 16, 0, &gpuobj);
+	*pgpuobj = gpuobj;
+	if (ret == 0) {
+		nv_wo32(*pgpuobj, 0x00, flags0 | (adjust << 20));
+		nv_wo32(*pgpuobj, 0x04, length);
+		nv_wo32(*pgpuobj, 0x08, flags2 | offset);
+		nv_wo32(*pgpuobj, 0x0c, flags2 | offset);
+	}
+
+	return ret;
+}
+
+static int
+nv04_dmaobj_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		 struct nouveau_oclass *oclass, void *data, u32 size,
+		 struct nouveau_object **pobject)
+{
+	struct nouveau_dmaeng *dmaeng = (void *)engine;
+	struct nv04_dmaobj_priv *dmaobj;
+	struct nouveau_gpuobj *gpuobj;
+	int ret;
+
+	ret = nouveau_dmaobj_create(parent, engine, oclass,
+				    data, size, &dmaobj);
+	*pobject = nv_object(dmaobj);
+	if (ret)
+		return ret;
+
+	switch (nv_mclass(parent)) {
+	case NV_DEVICE_CLASS:
+		break;
+	case NV03_CHANNEL_DMA_CLASS:
+	case NV10_CHANNEL_DMA_CLASS:
+	case NV17_CHANNEL_DMA_CLASS:
+	case NV40_CHANNEL_DMA_CLASS:
+		ret = dmaeng->bind(dmaeng, *pobject, &dmaobj->base, &gpuobj);
+		nouveau_object_ref(NULL, pobject);
+		*pobject = nv_object(gpuobj);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+static struct nouveau_ofuncs
+nv04_dmaobj_ofuncs = {
+	.ctor = nv04_dmaobj_ctor,
+	.dtor = _nouveau_dmaobj_dtor,
+	.init = _nouveau_dmaobj_init,
+	.fini = _nouveau_dmaobj_fini,
+};
+
+static struct nouveau_oclass
+nv04_dmaobj_sclass[] = {
+	{ 0x0002, &nv04_dmaobj_ofuncs },
+	{ 0x0003, &nv04_dmaobj_ofuncs },
+	{ 0x003d, &nv04_dmaobj_ofuncs },
+	{}
+};
+
+static int
+nv04_dmaeng_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		 struct nouveau_oclass *oclass, void *data, u32 size,
+		 struct nouveau_object **pobject)
+{
+	struct nv04_dmaeng_priv *priv;
+	int ret;
+
+	ret = nouveau_dmaeng_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	priv->base.base.sclass = nv04_dmaobj_sclass;
+	priv->base.bind = nv04_dmaobj_bind;
+	return 0;
+}
+
+struct nouveau_oclass
+nv04_dmaeng_oclass = {
+	.handle = NV_ENGINE(DMAOBJ, 0x04),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv04_dmaeng_ctor,
+		.dtor = _nouveau_dmaeng_dtor,
+		.init = _nouveau_dmaeng_init,
+		.fini = _nouveau_dmaeng_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/dmaobj/nv50.c b/drivers/gpu/drm/nouveau/core/engine/dmaobj/nv50.c
new file mode 100644
index 0000000..045d256
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/dmaobj/nv50.c
@@ -0,0 +1,173 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/gpuobj.h>
+#include <core/class.h>
+
+#include <subdev/fb.h>
+#include <engine/dmaobj.h>
+
+struct nv50_dmaeng_priv {
+	struct nouveau_dmaeng base;
+};
+
+struct nv50_dmaobj_priv {
+	struct nouveau_dmaobj base;
+};
+
+static int
+nv50_dmaobj_bind(struct nouveau_dmaeng *dmaeng,
+		 struct nouveau_object *parent,
+		 struct nouveau_dmaobj *dmaobj,
+		 struct nouveau_gpuobj **pgpuobj)
+{
+	u32 flags = nv_mclass(dmaobj);
+	int ret;
+
+	switch (dmaobj->target) {
+	case NV_MEM_TARGET_VM:
+		flags |= 0x00000000;
+		flags |= 0x60000000; /* COMPRESSION_USEVM */
+		flags |= 0x1fc00000; /* STORAGE_TYPE_USEVM */
+		break;
+	case NV_MEM_TARGET_VRAM:
+		flags |= 0x00010000;
+		flags |= 0x00100000; /* ACCESSUS_USER_SYSTEM */
+		break;
+	case NV_MEM_TARGET_PCI:
+		flags |= 0x00020000;
+		flags |= 0x00100000; /* ACCESSUS_USER_SYSTEM */
+		break;
+	case NV_MEM_TARGET_PCI_NOSNOOP:
+		flags |= 0x00030000;
+		flags |= 0x00100000; /* ACCESSUS_USER_SYSTEM */
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	switch (dmaobj->access) {
+	case NV_MEM_ACCESS_VM:
+		break;
+	case NV_MEM_ACCESS_RO:
+		flags |= 0x00040000;
+		break;
+	case NV_MEM_ACCESS_WO:
+	case NV_MEM_ACCESS_RW:
+		flags |= 0x00080000;
+		break;
+	}
+
+	ret = nouveau_gpuobj_new(parent, parent, 24, 32, 0, pgpuobj);
+	if (ret == 0) {
+		nv_wo32(*pgpuobj, 0x00, flags);
+		nv_wo32(*pgpuobj, 0x04, lower_32_bits(dmaobj->limit));
+		nv_wo32(*pgpuobj, 0x08, lower_32_bits(dmaobj->start));
+		nv_wo32(*pgpuobj, 0x0c, upper_32_bits(dmaobj->limit) << 24 |
+					upper_32_bits(dmaobj->start));
+		nv_wo32(*pgpuobj, 0x10, 0x00000000);
+		nv_wo32(*pgpuobj, 0x14, 0x00000000);
+	}
+
+	return ret;
+}
+
+static int
+nv50_dmaobj_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		 struct nouveau_oclass *oclass, void *data, u32 size,
+		 struct nouveau_object **pobject)
+{
+	struct nouveau_dmaeng *dmaeng = (void *)engine;
+	struct nv50_dmaobj_priv *dmaobj;
+	struct nouveau_gpuobj *gpuobj;
+	int ret;
+
+	ret = nouveau_dmaobj_create(parent, engine, oclass,
+				    data, size, &dmaobj);
+	*pobject = nv_object(dmaobj);
+	if (ret)
+		return ret;
+
+	switch (nv_mclass(parent)) {
+	case NV_DEVICE_CLASS:
+		break;
+	case NV50_CHANNEL_DMA_CLASS:
+	case NV84_CHANNEL_DMA_CLASS:
+	case NV50_CHANNEL_IND_CLASS:
+	case NV84_CHANNEL_IND_CLASS:
+		ret = dmaeng->bind(dmaeng, *pobject, &dmaobj->base, &gpuobj);
+		nouveau_object_ref(NULL, pobject);
+		*pobject = nv_object(gpuobj);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+static struct nouveau_ofuncs
+nv50_dmaobj_ofuncs = {
+	.ctor = nv50_dmaobj_ctor,
+	.dtor = _nouveau_dmaobj_dtor,
+	.init = _nouveau_dmaobj_init,
+	.fini = _nouveau_dmaobj_fini,
+};
+
+static struct nouveau_oclass
+nv50_dmaobj_sclass[] = {
+	{ 0x0002, &nv50_dmaobj_ofuncs },
+	{ 0x0003, &nv50_dmaobj_ofuncs },
+	{ 0x003d, &nv50_dmaobj_ofuncs },
+	{}
+};
+
+static int
+nv50_dmaeng_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		 struct nouveau_oclass *oclass, void *data, u32 size,
+		 struct nouveau_object **pobject)
+{
+	struct nv50_dmaeng_priv *priv;
+	int ret;
+
+	ret = nouveau_dmaeng_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	priv->base.base.sclass = nv50_dmaobj_sclass;
+	priv->base.bind = nv50_dmaobj_bind;
+	return 0;
+}
+
+struct nouveau_oclass
+nv50_dmaeng_oclass = {
+	.handle = NV_ENGINE(DMAOBJ, 0x50),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv50_dmaeng_ctor,
+		.dtor = _nouveau_dmaeng_dtor,
+		.init = _nouveau_dmaeng_init,
+		.fini = _nouveau_dmaeng_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/dmaobj/nvc0.c b/drivers/gpu/drm/nouveau/core/engine/dmaobj/nvc0.c
new file mode 100644
index 0000000..5baa086
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/dmaobj/nvc0.c
@@ -0,0 +1,99 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/gpuobj.h>
+
+#include <subdev/fb.h>
+#include <engine/dmaobj.h>
+
+struct nvc0_dmaeng_priv {
+	struct nouveau_dmaeng base;
+};
+
+struct nvc0_dmaobj_priv {
+	struct nouveau_dmaobj base;
+};
+
+static int
+nvc0_dmaobj_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		 struct nouveau_oclass *oclass, void *data, u32 size,
+		 struct nouveau_object **pobject)
+{
+	struct nvc0_dmaobj_priv *dmaobj;
+	int ret;
+
+	ret = nouveau_dmaobj_create(parent, engine, oclass, data, size, &dmaobj);
+	*pobject = nv_object(dmaobj);
+	if (ret)
+		return ret;
+
+	if (dmaobj->base.target != NV_MEM_TARGET_VM || dmaobj->base.start)
+		return -EINVAL;
+
+	return 0;
+}
+
+static struct nouveau_ofuncs
+nvc0_dmaobj_ofuncs = {
+	.ctor = nvc0_dmaobj_ctor,
+	.dtor = _nouveau_dmaobj_dtor,
+	.init = _nouveau_dmaobj_init,
+	.fini = _nouveau_dmaobj_fini,
+};
+
+static struct nouveau_oclass
+nvc0_dmaobj_sclass[] = {
+	{ 0x0002, &nvc0_dmaobj_ofuncs },
+	{ 0x0003, &nvc0_dmaobj_ofuncs },
+	{ 0x003d, &nvc0_dmaobj_ofuncs },
+	{}
+};
+
+static int
+nvc0_dmaeng_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		 struct nouveau_oclass *oclass, void *data, u32 size,
+		 struct nouveau_object **pobject)
+{
+	struct nvc0_dmaeng_priv *priv;
+	int ret;
+
+	ret = nouveau_dmaeng_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	priv->base.base.sclass = nvc0_dmaobj_sclass;
+	return 0;
+}
+
+struct nouveau_oclass
+nvc0_dmaeng_oclass = {
+	.handle = NV_ENGINE(DMAOBJ, 0xc0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvc0_dmaeng_ctor,
+		.dtor = _nouveau_dmaeng_dtor,
+		.init = _nouveau_dmaeng_init,
+		.fini = _nouveau_dmaeng_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/base.c b/drivers/gpu/drm/nouveau/core/engine/fifo/base.c
new file mode 100644
index 0000000..bbb43c6
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/base.c
@@ -0,0 +1,181 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/object.h>
+#include <core/handle.h>
+
+#include <engine/dmaobj.h>
+#include <engine/fifo.h>
+
+int
+nouveau_fifo_channel_create_(struct nouveau_object *parent,
+			     struct nouveau_object *engine,
+			     struct nouveau_oclass *oclass,
+			     int bar, u32 addr, u32 size, u32 pushbuf,
+			     u32 engmask, int len, void **ptr)
+{
+	struct nouveau_device *device = nv_device(engine);
+	struct nouveau_fifo *priv = (void *)engine;
+	struct nouveau_fifo_chan *chan;
+	struct nouveau_dmaeng *dmaeng;
+	unsigned long flags;
+	int ret;
+
+	/* create base object class */
+	ret = nouveau_namedb_create_(parent, engine, oclass, 0, NULL,
+				     engmask, len, ptr);
+	chan = *ptr;
+	if (ret)
+		return ret;
+
+	/* validate dma object representing push buffer */
+	chan->pushdma = (void *)nouveau_handle_ref(parent, pushbuf);
+	if (!chan->pushdma)
+		return -ENOENT;
+
+	dmaeng = (void *)chan->pushdma->base.engine;
+	switch (chan->pushdma->base.oclass->handle) {
+	case 0x0002:
+	case 0x003d:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (dmaeng->bind) {
+		ret = dmaeng->bind(dmaeng, parent, chan->pushdma, &chan->pushgpu);
+		if (ret)
+			return ret;
+	}
+
+	/* find a free fifo channel */
+	spin_lock_irqsave(&priv->lock, flags);
+	for (chan->chid = priv->min; chan->chid < priv->max; chan->chid++) {
+		if (!priv->channel[chan->chid]) {
+			priv->channel[chan->chid] = nv_object(chan);
+			break;
+		}
+	}
+	spin_unlock_irqrestore(&priv->lock, flags);
+
+	if (chan->chid == priv->max) {
+		nv_error(priv, "no free channels\n");
+		return -ENOSPC;
+	}
+
+	/* map fifo control registers */
+	chan->user = ioremap(pci_resource_start(device->pdev, bar) + addr +
+			     (chan->chid * size), size);
+	if (!chan->user)
+		return -EFAULT;
+
+	chan->size = size;
+	return 0;
+}
+
+void
+nouveau_fifo_channel_destroy(struct nouveau_fifo_chan *chan)
+{
+	struct nouveau_fifo *priv = (void *)nv_object(chan)->engine;
+	unsigned long flags;
+
+	iounmap(chan->user);
+
+	spin_lock_irqsave(&priv->lock, flags);
+	priv->channel[chan->chid] = NULL;
+	spin_unlock_irqrestore(&priv->lock, flags);
+
+	nouveau_gpuobj_ref(NULL, &chan->pushgpu);
+	nouveau_object_ref(NULL, (struct nouveau_object **)&chan->pushdma);
+	nouveau_namedb_destroy(&chan->base);
+}
+
+void
+_nouveau_fifo_channel_dtor(struct nouveau_object *object)
+{
+	struct nouveau_fifo_chan *chan = (void *)object;
+	nouveau_fifo_channel_destroy(chan);
+}
+
+u32
+_nouveau_fifo_channel_rd32(struct nouveau_object *object, u32 addr)
+{
+	struct nouveau_fifo_chan *chan = (void *)object;
+	return ioread32_native(chan->user + addr);
+}
+
+void
+_nouveau_fifo_channel_wr32(struct nouveau_object *object, u32 addr, u32 data)
+{
+	struct nouveau_fifo_chan *chan = (void *)object;
+	iowrite32_native(data, chan->user + addr);
+}
+
+static int
+nouveau_fifo_chid(struct nouveau_fifo *priv, struct nouveau_object *object)
+{
+	int engidx = nv_hclass(priv) & 0xff;
+
+	while (object && object->parent) {
+		if ( nv_iclass(object->parent, NV_ENGCTX_CLASS) &&
+		    (nv_hclass(object->parent) & 0xff) == engidx)
+			return nouveau_fifo_chan(object)->chid;
+		object = object->parent;
+	}
+
+	return -1;
+}
+
+void
+nouveau_fifo_destroy(struct nouveau_fifo *priv)
+{
+	kfree(priv->channel);
+	nouveau_engine_destroy(&priv->base);
+}
+
+int
+nouveau_fifo_create_(struct nouveau_object *parent,
+		     struct nouveau_object *engine,
+		     struct nouveau_oclass *oclass,
+		     int min, int max, int length, void **pobject)
+{
+	struct nouveau_fifo *priv;
+	int ret;
+
+	ret = nouveau_engine_create_(parent, engine, oclass, true, "PFIFO",
+				     "fifo", length, pobject);
+	priv = *pobject;
+	if (ret)
+		return ret;
+
+	priv->min = min;
+	priv->max = max;
+	priv->channel = kzalloc(sizeof(*priv->channel) * (max + 1), GFP_KERNEL);
+	if (!priv->channel)
+		return -ENOMEM;
+
+	priv->chid = nouveau_fifo_chid;
+	spin_lock_init(&priv->lock);
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/nv04.c b/drivers/gpu/drm/nouveau/core/engine/fifo/nv04.c
new file mode 100644
index 0000000..ea76e3e
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/nv04.c
@@ -0,0 +1,630 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/namedb.h>
+#include <core/handle.h>
+#include <core/ramht.h>
+
+#include <subdev/instmem.h>
+#include <subdev/instmem/nv04.h>
+#include <subdev/timer.h>
+#include <subdev/fb.h>
+
+#include <engine/fifo.h>
+
+#include "nv04.h"
+
+static struct ramfc_desc
+nv04_ramfc[] = {
+	{ 32,  0, 0x00,  0, NV04_PFIFO_CACHE1_DMA_PUT },
+	{ 32,  0, 0x04,  0, NV04_PFIFO_CACHE1_DMA_GET },
+	{ 16,  0, 0x08,  0, NV04_PFIFO_CACHE1_DMA_INSTANCE },
+	{ 16, 16, 0x08,  0, NV04_PFIFO_CACHE1_DMA_DCOUNT },
+	{ 32,  0, 0x0c,  0, NV04_PFIFO_CACHE1_DMA_STATE },
+	{ 32,  0, 0x10,  0, NV04_PFIFO_CACHE1_DMA_FETCH },
+	{ 32,  0, 0x14,  0, NV04_PFIFO_CACHE1_ENGINE },
+	{ 32,  0, 0x18,  0, NV04_PFIFO_CACHE1_PULL1 },
+	{}
+};
+
+/*******************************************************************************
+ * FIFO channel objects
+ ******************************************************************************/
+
+int
+nv04_fifo_object_attach(struct nouveau_object *parent,
+			struct nouveau_object *object, u32 handle)
+{
+	struct nv04_fifo_priv *priv = (void *)parent->engine;
+	struct nv04_fifo_chan *chan = (void *)parent;
+	u32 context, chid = chan->base.chid;
+	int ret;
+
+	if (nv_iclass(object, NV_GPUOBJ_CLASS))
+		context = nv_gpuobj(object)->addr >> 4;
+	else
+		context = 0x00000004; /* just non-zero */
+
+	switch (nv_engidx(object->engine)) {
+	case NVDEV_ENGINE_DMAOBJ:
+	case NVDEV_ENGINE_SW:
+		context |= 0x00000000;
+		break;
+	case NVDEV_ENGINE_GR:
+		context |= 0x00010000;
+		break;
+	case NVDEV_ENGINE_MPEG:
+		context |= 0x00020000;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	context |= 0x80000000; /* valid */
+	context |= chid << 24;
+
+	mutex_lock(&nv_subdev(priv)->mutex);
+	ret = nouveau_ramht_insert(priv->ramht, chid, handle, context);
+	mutex_unlock(&nv_subdev(priv)->mutex);
+	return ret;
+}
+
+void
+nv04_fifo_object_detach(struct nouveau_object *parent, int cookie)
+{
+	struct nv04_fifo_priv *priv = (void *)parent->engine;
+	mutex_lock(&nv_subdev(priv)->mutex);
+	nouveau_ramht_remove(priv->ramht, cookie);
+	mutex_unlock(&nv_subdev(priv)->mutex);
+}
+
+int
+nv04_fifo_context_attach(struct nouveau_object *parent,
+			 struct nouveau_object *object)
+{
+	nv_engctx(object)->addr = nouveau_fifo_chan(parent)->chid;
+	return 0;
+}
+
+static int
+nv04_fifo_chan_ctor(struct nouveau_object *parent,
+		    struct nouveau_object *engine,
+		    struct nouveau_oclass *oclass, void *data, u32 size,
+		    struct nouveau_object **pobject)
+{
+	struct nv04_fifo_priv *priv = (void *)engine;
+	struct nv04_fifo_chan *chan;
+	struct nv03_channel_dma_class *args = data;
+	int ret;
+
+	if (size < sizeof(*args))
+		return -EINVAL;
+
+	ret = nouveau_fifo_channel_create(parent, engine, oclass, 0, 0x800000,
+					  0x10000, args->pushbuf,
+					  (1 << NVDEV_ENGINE_DMAOBJ) |
+					  (1 << NVDEV_ENGINE_SW) |
+					  (1 << NVDEV_ENGINE_GR), &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	nv_parent(chan)->object_attach = nv04_fifo_object_attach;
+	nv_parent(chan)->object_detach = nv04_fifo_object_detach;
+	nv_parent(chan)->context_attach = nv04_fifo_context_attach;
+	chan->ramfc = chan->base.chid * 32;
+
+	nv_wo32(priv->ramfc, chan->ramfc + 0x00, args->offset);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x04, args->offset);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x08, chan->base.pushgpu->addr >> 4);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x10,
+			     NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES |
+			     NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES |
+#ifdef __BIG_ENDIAN
+			     NV_PFIFO_CACHE1_BIG_ENDIAN |
+#endif
+			     NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8);
+	return 0;
+}
+
+void
+nv04_fifo_chan_dtor(struct nouveau_object *object)
+{
+	struct nv04_fifo_priv *priv = (void *)object->engine;
+	struct nv04_fifo_chan *chan = (void *)object;
+	struct ramfc_desc *c = priv->ramfc_desc;
+
+	do {
+		nv_wo32(priv->ramfc, chan->ramfc + c->ctxp, 0x00000000);
+	} while ((++c)->bits);
+
+	nouveau_fifo_channel_destroy(&chan->base);
+}
+
+int
+nv04_fifo_chan_init(struct nouveau_object *object)
+{
+	struct nv04_fifo_priv *priv = (void *)object->engine;
+	struct nv04_fifo_chan *chan = (void *)object;
+	u32 mask = 1 << chan->base.chid;
+	unsigned long flags;
+	int ret;
+
+	ret = nouveau_fifo_channel_init(&chan->base);
+	if (ret)
+		return ret;
+
+	spin_lock_irqsave(&priv->base.lock, flags);
+	nv_mask(priv, NV04_PFIFO_MODE, mask, mask);
+	spin_unlock_irqrestore(&priv->base.lock, flags);
+	return 0;
+}
+
+int
+nv04_fifo_chan_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv04_fifo_priv *priv = (void *)object->engine;
+	struct nv04_fifo_chan *chan = (void *)object;
+	struct nouveau_gpuobj *fctx = priv->ramfc;
+	struct ramfc_desc *c;
+	unsigned long flags;
+	u32 data = chan->ramfc;
+	u32 chid;
+
+	/* prevent fifo context switches */
+	spin_lock_irqsave(&priv->base.lock, flags);
+	nv_wr32(priv, NV03_PFIFO_CACHES, 0);
+
+	/* if this channel is active, replace it with a null context */
+	chid = nv_rd32(priv, NV03_PFIFO_CACHE1_PUSH1) & priv->base.max;
+	if (chid == chan->base.chid) {
+		nv_mask(priv, NV04_PFIFO_CACHE1_DMA_PUSH, 0x00000001, 0);
+		nv_wr32(priv, NV03_PFIFO_CACHE1_PUSH0, 0);
+		nv_mask(priv, NV04_PFIFO_CACHE1_PULL0, 0x00000001, 0);
+
+		c = priv->ramfc_desc;
+		do {
+			u32 rm = ((1ULL << c->bits) - 1) << c->regs;
+			u32 cm = ((1ULL << c->bits) - 1) << c->ctxs;
+			u32 rv = (nv_rd32(priv, c->regp) &  rm) >> c->regs;
+			u32 cv = (nv_ro32(fctx, c->ctxp + data) & ~cm);
+			nv_wo32(fctx, c->ctxp + data, cv | (rv << c->ctxs));
+		} while ((++c)->bits);
+
+		c = priv->ramfc_desc;
+		do {
+			nv_wr32(priv, c->regp, 0x00000000);
+		} while ((++c)->bits);
+
+		nv_wr32(priv, NV03_PFIFO_CACHE1_GET, 0);
+		nv_wr32(priv, NV03_PFIFO_CACHE1_PUT, 0);
+		nv_wr32(priv, NV03_PFIFO_CACHE1_PUSH1, priv->base.max);
+		nv_wr32(priv, NV03_PFIFO_CACHE1_PUSH0, 1);
+		nv_wr32(priv, NV04_PFIFO_CACHE1_PULL0, 1);
+	}
+
+	/* restore normal operation, after disabling dma mode */
+	nv_mask(priv, NV04_PFIFO_MODE, 1 << chan->base.chid, 0);
+	nv_wr32(priv, NV03_PFIFO_CACHES, 1);
+	spin_unlock_irqrestore(&priv->base.lock, flags);
+
+	return nouveau_fifo_channel_fini(&chan->base, suspend);
+}
+
+static struct nouveau_ofuncs
+nv04_fifo_ofuncs = {
+	.ctor = nv04_fifo_chan_ctor,
+	.dtor = nv04_fifo_chan_dtor,
+	.init = nv04_fifo_chan_init,
+	.fini = nv04_fifo_chan_fini,
+	.rd32 = _nouveau_fifo_channel_rd32,
+	.wr32 = _nouveau_fifo_channel_wr32,
+};
+
+static struct nouveau_oclass
+nv04_fifo_sclass[] = {
+	{ NV03_CHANNEL_DMA_CLASS, &nv04_fifo_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * FIFO context - basically just the instmem reserved for the channel
+ ******************************************************************************/
+
+int
+nv04_fifo_context_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nv04_fifo_base *base;
+	int ret;
+
+	ret = nouveau_fifo_context_create(parent, engine, oclass, NULL, 0x1000,
+				          0x1000, NVOBJ_FLAG_HEAP, &base);
+	*pobject = nv_object(base);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static struct nouveau_oclass
+nv04_fifo_cclass = {
+	.handle = NV_ENGCTX(FIFO, 0x04),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv04_fifo_context_ctor,
+		.dtor = _nouveau_fifo_context_dtor,
+		.init = _nouveau_fifo_context_init,
+		.fini = _nouveau_fifo_context_fini,
+		.rd32 = _nouveau_fifo_context_rd32,
+		.wr32 = _nouveau_fifo_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PFIFO engine
+ ******************************************************************************/
+
+void
+nv04_fifo_pause(struct nouveau_fifo *pfifo, unsigned long *pflags)
+__acquires(priv->base.lock)
+{
+	struct nv04_fifo_priv *priv = (void *)pfifo;
+	unsigned long flags;
+
+	spin_lock_irqsave(&priv->base.lock, flags);
+	*pflags = flags;
+
+	nv_wr32(priv, NV03_PFIFO_CACHES, 0x00000000);
+	nv_mask(priv, NV04_PFIFO_CACHE1_PULL0, 0x00000001, 0x00000000);
+
+	/* in some cases the puller may be left in an inconsistent state
+	 * if you try to stop it while it's busy translating handles.
+	 * sometimes you get a CACHE_ERROR, sometimes it just fails
+	 * silently; sending incorrect instance offsets to PGRAPH after
+	 * it's started up again.
+	 *
+	 * to avoid this, we invalidate the most recently calculated
+	 * instance.
+	 */
+	if (!nv_wait(priv, NV04_PFIFO_CACHE1_PULL0,
+			   NV04_PFIFO_CACHE1_PULL0_HASH_BUSY, 0x00000000))
+		nv_warn(priv, "timeout idling puller\n");
+
+	if (nv_rd32(priv, NV04_PFIFO_CACHE1_PULL0) &
+			  NV04_PFIFO_CACHE1_PULL0_HASH_FAILED)
+		nv_wr32(priv, NV03_PFIFO_INTR_0, NV_PFIFO_INTR_CACHE_ERROR);
+
+	nv_wr32(priv, NV04_PFIFO_CACHE1_HASH, 0x00000000);
+}
+
+void
+nv04_fifo_start(struct nouveau_fifo *pfifo, unsigned long *pflags)
+__releases(priv->base.lock)
+{
+	struct nv04_fifo_priv *priv = (void *)pfifo;
+	unsigned long flags = *pflags;
+
+	nv_mask(priv, NV04_PFIFO_CACHE1_PULL0, 0x00000001, 0x00000001);
+	nv_wr32(priv, NV03_PFIFO_CACHES, 0x00000001);
+
+	spin_unlock_irqrestore(&priv->base.lock, flags);
+}
+
+static const char *
+nv_dma_state_err(u32 state)
+{
+	static const char * const desc[] = {
+		"NONE", "CALL_SUBR_ACTIVE", "INVALID_MTHD", "RET_SUBR_INACTIVE",
+		"INVALID_CMD", "IB_EMPTY"/* NV50+ */, "MEM_FAULT", "UNK"
+	};
+	return desc[(state >> 29) & 0x7];
+}
+
+static bool
+nv04_fifo_swmthd(struct nv04_fifo_priv *priv, u32 chid, u32 addr, u32 data)
+{
+	struct nv04_fifo_chan *chan = NULL;
+	struct nouveau_handle *bind;
+	const int subc = (addr >> 13) & 0x7;
+	const int mthd = addr & 0x1ffc;
+	bool handled = false;
+	unsigned long flags;
+	u32 engine;
+
+	spin_lock_irqsave(&priv->base.lock, flags);
+	if (likely(chid >= priv->base.min && chid <= priv->base.max))
+		chan = (void *)priv->base.channel[chid];
+	if (unlikely(!chan))
+		goto out;
+
+	switch (mthd) {
+	case 0x0000:
+		bind = nouveau_namedb_get(nv_namedb(chan), data);
+		if (unlikely(!bind))
+			break;
+
+		if (nv_engidx(bind->object->engine) == NVDEV_ENGINE_SW) {
+			engine = 0x0000000f << (subc * 4);
+			chan->subc[subc] = data;
+			handled = true;
+
+			nv_mask(priv, NV04_PFIFO_CACHE1_ENGINE, engine, 0);
+		}
+
+		nouveau_namedb_put(bind);
+		break;
+	default:
+		engine = nv_rd32(priv, NV04_PFIFO_CACHE1_ENGINE);
+		if (unlikely(((engine >> (subc * 4)) & 0xf) != 0))
+			break;
+
+		bind = nouveau_namedb_get(nv_namedb(chan), chan->subc[subc]);
+		if (likely(bind)) {
+			if (!nv_call(bind->object, mthd, data))
+				handled = true;
+			nouveau_namedb_put(bind);
+		}
+		break;
+	}
+
+out:
+	spin_unlock_irqrestore(&priv->base.lock, flags);
+	return handled;
+}
+
+void
+nv04_fifo_intr(struct nouveau_subdev *subdev)
+{
+	struct nouveau_device *device = nv_device(subdev);
+	struct nv04_fifo_priv *priv = (void *)subdev;
+	uint32_t status, reassign;
+	int cnt = 0;
+
+	reassign = nv_rd32(priv, NV03_PFIFO_CACHES) & 1;
+	while ((status = nv_rd32(priv, NV03_PFIFO_INTR_0)) && (cnt++ < 100)) {
+		uint32_t chid, get;
+
+		nv_wr32(priv, NV03_PFIFO_CACHES, 0);
+
+		chid = nv_rd32(priv, NV03_PFIFO_CACHE1_PUSH1) & priv->base.max;
+		get  = nv_rd32(priv, NV03_PFIFO_CACHE1_GET);
+
+		if (status & NV_PFIFO_INTR_CACHE_ERROR) {
+			uint32_t mthd, data;
+			int ptr;
+
+			/* NV_PFIFO_CACHE1_GET actually goes to 0xffc before
+			 * wrapping on my G80 chips, but CACHE1 isn't big
+			 * enough for this much data.. Tests show that it
+			 * wraps around to the start at GET=0x800.. No clue
+			 * as to why..
+			 */
+			ptr = (get & 0x7ff) >> 2;
+
+			if (device->card_type < NV_40) {
+				mthd = nv_rd32(priv,
+					NV04_PFIFO_CACHE1_METHOD(ptr));
+				data = nv_rd32(priv,
+					NV04_PFIFO_CACHE1_DATA(ptr));
+			} else {
+				mthd = nv_rd32(priv,
+					NV40_PFIFO_CACHE1_METHOD(ptr));
+				data = nv_rd32(priv,
+					NV40_PFIFO_CACHE1_DATA(ptr));
+			}
+
+			if (!nv04_fifo_swmthd(priv, chid, mthd, data)) {
+				nv_info(priv, "CACHE_ERROR - Ch %d/%d "
+					      "Mthd 0x%04x Data 0x%08x\n",
+					chid, (mthd >> 13) & 7, mthd & 0x1ffc,
+					data);
+			}
+
+			nv_wr32(priv, NV04_PFIFO_CACHE1_DMA_PUSH, 0);
+			nv_wr32(priv, NV03_PFIFO_INTR_0,
+						NV_PFIFO_INTR_CACHE_ERROR);
+
+			nv_wr32(priv, NV03_PFIFO_CACHE1_PUSH0,
+				nv_rd32(priv, NV03_PFIFO_CACHE1_PUSH0) & ~1);
+			nv_wr32(priv, NV03_PFIFO_CACHE1_GET, get + 4);
+			nv_wr32(priv, NV03_PFIFO_CACHE1_PUSH0,
+				nv_rd32(priv, NV03_PFIFO_CACHE1_PUSH0) | 1);
+			nv_wr32(priv, NV04_PFIFO_CACHE1_HASH, 0);
+
+			nv_wr32(priv, NV04_PFIFO_CACHE1_DMA_PUSH,
+				nv_rd32(priv, NV04_PFIFO_CACHE1_DMA_PUSH) | 1);
+			nv_wr32(priv, NV04_PFIFO_CACHE1_PULL0, 1);
+
+			status &= ~NV_PFIFO_INTR_CACHE_ERROR;
+		}
+
+		if (status & NV_PFIFO_INTR_DMA_PUSHER) {
+			u32 dma_get = nv_rd32(priv, 0x003244);
+			u32 dma_put = nv_rd32(priv, 0x003240);
+			u32 push = nv_rd32(priv, 0x003220);
+			u32 state = nv_rd32(priv, 0x003228);
+
+			if (device->card_type == NV_50) {
+				u32 ho_get = nv_rd32(priv, 0x003328);
+				u32 ho_put = nv_rd32(priv, 0x003320);
+				u32 ib_get = nv_rd32(priv, 0x003334);
+				u32 ib_put = nv_rd32(priv, 0x003330);
+
+				nv_info(priv, "DMA_PUSHER - Ch %d Get 0x%02x%08x "
+				     "Put 0x%02x%08x IbGet 0x%08x IbPut 0x%08x "
+				     "State 0x%08x (err: %s) Push 0x%08x\n",
+					chid, ho_get, dma_get, ho_put,
+					dma_put, ib_get, ib_put, state,
+					nv_dma_state_err(state),
+					push);
+
+				/* METHOD_COUNT, in DMA_STATE on earlier chipsets */
+				nv_wr32(priv, 0x003364, 0x00000000);
+				if (dma_get != dma_put || ho_get != ho_put) {
+					nv_wr32(priv, 0x003244, dma_put);
+					nv_wr32(priv, 0x003328, ho_put);
+				} else
+				if (ib_get != ib_put) {
+					nv_wr32(priv, 0x003334, ib_put);
+				}
+			} else {
+				nv_info(priv, "DMA_PUSHER - Ch %d Get 0x%08x "
+					     "Put 0x%08x State 0x%08x (err: %s) Push 0x%08x\n",
+					chid, dma_get, dma_put, state,
+					nv_dma_state_err(state), push);
+
+				if (dma_get != dma_put)
+					nv_wr32(priv, 0x003244, dma_put);
+			}
+
+			nv_wr32(priv, 0x003228, 0x00000000);
+			nv_wr32(priv, 0x003220, 0x00000001);
+			nv_wr32(priv, 0x002100, NV_PFIFO_INTR_DMA_PUSHER);
+			status &= ~NV_PFIFO_INTR_DMA_PUSHER;
+		}
+
+		if (status & NV_PFIFO_INTR_SEMAPHORE) {
+			uint32_t sem;
+
+			status &= ~NV_PFIFO_INTR_SEMAPHORE;
+			nv_wr32(priv, NV03_PFIFO_INTR_0,
+				NV_PFIFO_INTR_SEMAPHORE);
+
+			sem = nv_rd32(priv, NV10_PFIFO_CACHE1_SEMAPHORE);
+			nv_wr32(priv, NV10_PFIFO_CACHE1_SEMAPHORE, sem | 0x1);
+
+			nv_wr32(priv, NV03_PFIFO_CACHE1_GET, get + 4);
+			nv_wr32(priv, NV04_PFIFO_CACHE1_PULL0, 1);
+		}
+
+		if (device->card_type == NV_50) {
+			if (status & 0x00000010) {
+				nv50_fb_trap(nouveau_fb(priv), 1);
+				status &= ~0x00000010;
+				nv_wr32(priv, 0x002100, 0x00000010);
+			}
+		}
+
+		if (status) {
+			nv_info(priv, "unknown intr 0x%08x, ch %d\n",
+				status, chid);
+			nv_wr32(priv, NV03_PFIFO_INTR_0, status);
+			status = 0;
+		}
+
+		nv_wr32(priv, NV03_PFIFO_CACHES, reassign);
+	}
+
+	if (status) {
+		nv_info(priv, "still angry after %d spins, halt\n", cnt);
+		nv_wr32(priv, 0x002140, 0);
+		nv_wr32(priv, 0x000140, 0);
+	}
+
+	nv_wr32(priv, 0x000100, 0x00000100);
+}
+
+static int
+nv04_fifo_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv04_instmem_priv *imem = nv04_instmem(parent);
+	struct nv04_fifo_priv *priv;
+	int ret;
+
+	ret = nouveau_fifo_create(parent, engine, oclass, 0, 15, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nouveau_ramht_ref(imem->ramht, &priv->ramht);
+	nouveau_gpuobj_ref(imem->ramro, &priv->ramro);
+	nouveau_gpuobj_ref(imem->ramfc, &priv->ramfc);
+
+	nv_subdev(priv)->unit = 0x00000100;
+	nv_subdev(priv)->intr = nv04_fifo_intr;
+	nv_engine(priv)->cclass = &nv04_fifo_cclass;
+	nv_engine(priv)->sclass = nv04_fifo_sclass;
+	priv->base.pause = nv04_fifo_pause;
+	priv->base.start = nv04_fifo_start;
+	priv->ramfc_desc = nv04_ramfc;
+	return 0;
+}
+
+void
+nv04_fifo_dtor(struct nouveau_object *object)
+{
+	struct nv04_fifo_priv *priv = (void *)object;
+	nouveau_gpuobj_ref(NULL, &priv->ramfc);
+	nouveau_gpuobj_ref(NULL, &priv->ramro);
+	nouveau_ramht_ref(NULL, &priv->ramht);
+	nouveau_fifo_destroy(&priv->base);
+}
+
+int
+nv04_fifo_init(struct nouveau_object *object)
+{
+	struct nv04_fifo_priv *priv = (void *)object;
+	int ret;
+
+	ret = nouveau_fifo_init(&priv->base);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, NV04_PFIFO_DELAY_0, 0x000000ff);
+	nv_wr32(priv, NV04_PFIFO_DMA_TIMESLICE, 0x0101ffff);
+
+	nv_wr32(priv, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ |
+				       ((priv->ramht->bits - 9) << 16) |
+				        (priv->ramht->base.addr >> 8));
+	nv_wr32(priv, NV03_PFIFO_RAMRO, priv->ramro->addr >> 8);
+	nv_wr32(priv, NV03_PFIFO_RAMFC, priv->ramfc->addr >> 8);
+
+	nv_wr32(priv, NV03_PFIFO_CACHE1_PUSH1, priv->base.max);
+
+	nv_wr32(priv, NV03_PFIFO_INTR_0, 0xffffffff);
+	nv_wr32(priv, NV03_PFIFO_INTR_EN_0, 0xffffffff);
+
+	nv_wr32(priv, NV03_PFIFO_CACHE1_PUSH0, 1);
+	nv_wr32(priv, NV04_PFIFO_CACHE1_PULL0, 1);
+	nv_wr32(priv, NV03_PFIFO_CACHES, 1);
+	return 0;
+}
+
+struct nouveau_oclass
+nv04_fifo_oclass = {
+	.handle = NV_ENGINE(FIFO, 0x04),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv04_fifo_ctor,
+		.dtor = nv04_fifo_dtor,
+		.init = nv04_fifo_init,
+		.fini = _nouveau_fifo_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/nv04.h b/drivers/gpu/drm/nouveau/core/engine/fifo/nv04.h
new file mode 100644
index 0000000..496a4b4
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/nv04.h
@@ -0,0 +1,178 @@
+#ifndef __NV04_FIFO_H__
+#define __NV04_FIFO_H__
+
+#include <engine/fifo.h>
+
+#define NV04_PFIFO_DELAY_0                                 0x00002040
+#define NV04_PFIFO_DMA_TIMESLICE                           0x00002044
+#define NV04_PFIFO_NEXT_CHANNEL                            0x00002050
+#define NV03_PFIFO_INTR_0                                  0x00002100
+#define NV03_PFIFO_INTR_EN_0                               0x00002140
+#    define NV_PFIFO_INTR_CACHE_ERROR                          (1<<0)
+#    define NV_PFIFO_INTR_RUNOUT                               (1<<4)
+#    define NV_PFIFO_INTR_RUNOUT_OVERFLOW                      (1<<8)
+#    define NV_PFIFO_INTR_DMA_PUSHER                          (1<<12)
+#    define NV_PFIFO_INTR_DMA_PT                              (1<<16)
+#    define NV_PFIFO_INTR_SEMAPHORE                           (1<<20)
+#    define NV_PFIFO_INTR_ACQUIRE_TIMEOUT                     (1<<24)
+#define NV03_PFIFO_RAMHT                                   0x00002210
+#define NV03_PFIFO_RAMFC                                   0x00002214
+#define NV03_PFIFO_RAMRO                                   0x00002218
+#define NV40_PFIFO_RAMFC                                   0x00002220
+#define NV03_PFIFO_CACHES                                  0x00002500
+#define NV04_PFIFO_MODE                                    0x00002504
+#define NV04_PFIFO_DMA                                     0x00002508
+#define NV04_PFIFO_SIZE                                    0x0000250c
+#define NV50_PFIFO_CTX_TABLE(c)                        (0x2600+(c)*4)
+#define NV50_PFIFO_CTX_TABLE__SIZE                                128
+#define NV50_PFIFO_CTX_TABLE_CHANNEL_ENABLED                  (1<<31)
+#define NV50_PFIFO_CTX_TABLE_UNK30_BAD                        (1<<30)
+#define NV50_PFIFO_CTX_TABLE_INSTANCE_MASK_G80             0x0FFFFFFF
+#define NV50_PFIFO_CTX_TABLE_INSTANCE_MASK_G84             0x00FFFFFF
+#define NV03_PFIFO_CACHE0_PUSH0                            0x00003000
+#define NV03_PFIFO_CACHE0_PULL0                            0x00003040
+#define NV04_PFIFO_CACHE0_PULL0                            0x00003050
+#define NV04_PFIFO_CACHE0_PULL1                            0x00003054
+#define NV03_PFIFO_CACHE1_PUSH0                            0x00003200
+#define NV03_PFIFO_CACHE1_PUSH1                            0x00003204
+#define NV03_PFIFO_CACHE1_PUSH1_DMA                            (1<<8)
+#define NV40_PFIFO_CACHE1_PUSH1_DMA                           (1<<16)
+#define NV03_PFIFO_CACHE1_PUSH1_CHID_MASK                  0x0000000f
+#define NV10_PFIFO_CACHE1_PUSH1_CHID_MASK                  0x0000001f
+#define NV50_PFIFO_CACHE1_PUSH1_CHID_MASK                  0x0000007f
+#define NV03_PFIFO_CACHE1_PUT                              0x00003210
+#define NV04_PFIFO_CACHE1_DMA_PUSH                         0x00003220
+#define NV04_PFIFO_CACHE1_DMA_FETCH                        0x00003224
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_8_BYTES         0x00000000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_16_BYTES        0x00000008
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_24_BYTES        0x00000010
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_32_BYTES        0x00000018
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_40_BYTES        0x00000020
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_48_BYTES        0x00000028
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_56_BYTES        0x00000030
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_64_BYTES        0x00000038
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_72_BYTES        0x00000040
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_80_BYTES        0x00000048
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_88_BYTES        0x00000050
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_96_BYTES        0x00000058
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_104_BYTES       0x00000060
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_112_BYTES       0x00000068
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_120_BYTES       0x00000070
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES       0x00000078
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_136_BYTES       0x00000080
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_144_BYTES       0x00000088
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_152_BYTES       0x00000090
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_160_BYTES       0x00000098
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_168_BYTES       0x000000A0
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_176_BYTES       0x000000A8
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_184_BYTES       0x000000B0
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_192_BYTES       0x000000B8
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_200_BYTES       0x000000C0
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_208_BYTES       0x000000C8
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_216_BYTES       0x000000D0
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_224_BYTES       0x000000D8
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_232_BYTES       0x000000E0
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_240_BYTES       0x000000E8
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_248_BYTES       0x000000F0
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_256_BYTES       0x000000F8
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE                 0x0000E000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_32_BYTES        0x00000000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_64_BYTES        0x00002000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_96_BYTES        0x00004000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES       0x00006000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_160_BYTES       0x00008000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_192_BYTES       0x0000A000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_224_BYTES       0x0000C000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_256_BYTES       0x0000E000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS             0x001F0000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_0           0x00000000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_1           0x00010000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_2           0x00020000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_3           0x00030000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_4           0x00040000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_5           0x00050000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_6           0x00060000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_7           0x00070000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8           0x00080000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_9           0x00090000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_10          0x000A0000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_11          0x000B0000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_12          0x000C0000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_13          0x000D0000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_14          0x000E0000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_15          0x000F0000
+#    define NV_PFIFO_CACHE1_ENDIAN                         0x80000000
+#    define NV_PFIFO_CACHE1_LITTLE_ENDIAN                  0x7FFFFFFF
+#    define NV_PFIFO_CACHE1_BIG_ENDIAN                     0x80000000
+#define NV04_PFIFO_CACHE1_DMA_STATE                        0x00003228
+#define NV04_PFIFO_CACHE1_DMA_INSTANCE                     0x0000322c
+#define NV04_PFIFO_CACHE1_DMA_CTL                          0x00003230
+#define NV04_PFIFO_CACHE1_DMA_PUT                          0x00003240
+#define NV04_PFIFO_CACHE1_DMA_GET                          0x00003244
+#define NV10_PFIFO_CACHE1_REF_CNT                          0x00003248
+#define NV10_PFIFO_CACHE1_DMA_SUBROUTINE                   0x0000324C
+#define NV03_PFIFO_CACHE1_PULL0                            0x00003240
+#define NV04_PFIFO_CACHE1_PULL0                            0x00003250
+#    define NV04_PFIFO_CACHE1_PULL0_HASH_FAILED            0x00000010
+#    define NV04_PFIFO_CACHE1_PULL0_HASH_BUSY              0x00001000
+#define NV03_PFIFO_CACHE1_PULL1                            0x00003250
+#define NV04_PFIFO_CACHE1_PULL1                            0x00003254
+#define NV04_PFIFO_CACHE1_HASH                             0x00003258
+#define NV10_PFIFO_CACHE1_ACQUIRE_TIMEOUT                  0x00003260
+#define NV10_PFIFO_CACHE1_ACQUIRE_TIMESTAMP                0x00003264
+#define NV10_PFIFO_CACHE1_ACQUIRE_VALUE                    0x00003268
+#define NV10_PFIFO_CACHE1_SEMAPHORE                        0x0000326C
+#define NV03_PFIFO_CACHE1_GET                              0x00003270
+#define NV04_PFIFO_CACHE1_ENGINE                           0x00003280
+#define NV04_PFIFO_CACHE1_DMA_DCOUNT                       0x000032A0
+#define NV40_PFIFO_GRCTX_INSTANCE                          0x000032E0
+#define NV40_PFIFO_UNK32E4                                 0x000032E4
+#define NV04_PFIFO_CACHE1_METHOD(i)                (0x00003800+(i*8))
+#define NV04_PFIFO_CACHE1_DATA(i)                  (0x00003804+(i*8))
+#define NV40_PFIFO_CACHE1_METHOD(i)                (0x00090000+(i*8))
+#define NV40_PFIFO_CACHE1_DATA(i)                  (0x00090004+(i*8))
+
+struct ramfc_desc {
+	unsigned bits:6;
+	unsigned ctxs:5;
+	unsigned ctxp:8;
+	unsigned regs:5;
+	unsigned regp;
+};
+
+struct nv04_fifo_priv {
+	struct nouveau_fifo base;
+	struct ramfc_desc *ramfc_desc;
+	struct nouveau_ramht  *ramht;
+	struct nouveau_gpuobj *ramro;
+	struct nouveau_gpuobj *ramfc;
+};
+
+struct nv04_fifo_base {
+	struct nouveau_fifo_base base;
+};
+
+struct nv04_fifo_chan {
+	struct nouveau_fifo_chan base;
+	u32 subc[8];
+	u32 ramfc;
+};
+
+int  nv04_fifo_object_attach(struct nouveau_object *,
+			     struct nouveau_object *, u32);
+void nv04_fifo_object_detach(struct nouveau_object *, int);
+
+void nv04_fifo_chan_dtor(struct nouveau_object *);
+int  nv04_fifo_chan_init(struct nouveau_object *);
+int  nv04_fifo_chan_fini(struct nouveau_object *, bool suspend);
+
+int  nv04_fifo_context_ctor(struct nouveau_object *, struct nouveau_object *,
+			    struct nouveau_oclass *, void *, u32,
+			    struct nouveau_object **);
+
+void nv04_fifo_dtor(struct nouveau_object *);
+int  nv04_fifo_init(struct nouveau_object *);
+void nv04_fifo_pause(struct nouveau_fifo *, unsigned long *);
+void nv04_fifo_start(struct nouveau_fifo *, unsigned long *);
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/nv10.c b/drivers/gpu/drm/nouveau/core/engine/fifo/nv10.c
new file mode 100644
index 0000000..4ba7542
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/nv10.c
@@ -0,0 +1,171 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/ramht.h>
+
+#include <subdev/instmem.h>
+#include <subdev/instmem/nv04.h>
+#include <subdev/fb.h>
+
+#include <engine/fifo.h>
+
+#include "nv04.h"
+
+static struct ramfc_desc
+nv10_ramfc[] = {
+	{ 32,  0, 0x00,  0, NV04_PFIFO_CACHE1_DMA_PUT },
+	{ 32,  0, 0x04,  0, NV04_PFIFO_CACHE1_DMA_GET },
+	{ 32,  0, 0x08,  0, NV10_PFIFO_CACHE1_REF_CNT },
+	{ 16,  0, 0x0c,  0, NV04_PFIFO_CACHE1_DMA_INSTANCE },
+	{ 16, 16, 0x0c,  0, NV04_PFIFO_CACHE1_DMA_DCOUNT },
+	{ 32,  0, 0x10,  0, NV04_PFIFO_CACHE1_DMA_STATE },
+	{ 32,  0, 0x14,  0, NV04_PFIFO_CACHE1_DMA_FETCH },
+	{ 32,  0, 0x18,  0, NV04_PFIFO_CACHE1_ENGINE },
+	{ 32,  0, 0x1c,  0, NV04_PFIFO_CACHE1_PULL1 },
+	{}
+};
+
+/*******************************************************************************
+ * FIFO channel objects
+ ******************************************************************************/
+
+static int
+nv10_fifo_chan_ctor(struct nouveau_object *parent,
+		    struct nouveau_object *engine,
+		    struct nouveau_oclass *oclass, void *data, u32 size,
+		    struct nouveau_object **pobject)
+{
+	struct nv04_fifo_priv *priv = (void *)engine;
+	struct nv04_fifo_chan *chan;
+	struct nv03_channel_dma_class *args = data;
+	int ret;
+
+	if (size < sizeof(*args))
+		return -EINVAL;
+
+	ret = nouveau_fifo_channel_create(parent, engine, oclass, 0, 0x800000,
+					  0x10000, args->pushbuf,
+					  (1 << NVDEV_ENGINE_DMAOBJ) |
+					  (1 << NVDEV_ENGINE_SW) |
+					  (1 << NVDEV_ENGINE_GR), &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	nv_parent(chan)->object_attach = nv04_fifo_object_attach;
+	nv_parent(chan)->object_detach = nv04_fifo_object_detach;
+	nv_parent(chan)->context_attach = nv04_fifo_context_attach;
+	chan->ramfc = chan->base.chid * 32;
+
+	nv_wo32(priv->ramfc, chan->ramfc + 0x00, args->offset);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x04, args->offset);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x0c, chan->base.pushgpu->addr >> 4);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x14,
+			     NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES |
+			     NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES |
+#ifdef __BIG_ENDIAN
+			     NV_PFIFO_CACHE1_BIG_ENDIAN |
+#endif
+			     NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8);
+	return 0;
+}
+
+static struct nouveau_ofuncs
+nv10_fifo_ofuncs = {
+	.ctor = nv10_fifo_chan_ctor,
+	.dtor = nv04_fifo_chan_dtor,
+	.init = nv04_fifo_chan_init,
+	.fini = nv04_fifo_chan_fini,
+	.rd32 = _nouveau_fifo_channel_rd32,
+	.wr32 = _nouveau_fifo_channel_wr32,
+};
+
+static struct nouveau_oclass
+nv10_fifo_sclass[] = {
+	{ NV10_CHANNEL_DMA_CLASS, &nv10_fifo_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * FIFO context - basically just the instmem reserved for the channel
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv10_fifo_cclass = {
+	.handle = NV_ENGCTX(FIFO, 0x10),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv04_fifo_context_ctor,
+		.dtor = _nouveau_fifo_context_dtor,
+		.init = _nouveau_fifo_context_init,
+		.fini = _nouveau_fifo_context_fini,
+		.rd32 = _nouveau_fifo_context_rd32,
+		.wr32 = _nouveau_fifo_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PFIFO engine
+ ******************************************************************************/
+
+static int
+nv10_fifo_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv04_instmem_priv *imem = nv04_instmem(parent);
+	struct nv04_fifo_priv *priv;
+	int ret;
+
+	ret = nouveau_fifo_create(parent, engine, oclass, 0, 31, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nouveau_ramht_ref(imem->ramht, &priv->ramht);
+	nouveau_gpuobj_ref(imem->ramro, &priv->ramro);
+	nouveau_gpuobj_ref(imem->ramfc, &priv->ramfc);
+
+	nv_subdev(priv)->unit = 0x00000100;
+	nv_subdev(priv)->intr = nv04_fifo_intr;
+	nv_engine(priv)->cclass = &nv10_fifo_cclass;
+	nv_engine(priv)->sclass = nv10_fifo_sclass;
+	priv->base.pause = nv04_fifo_pause;
+	priv->base.start = nv04_fifo_start;
+	priv->ramfc_desc = nv10_ramfc;
+	return 0;
+}
+
+struct nouveau_oclass
+nv10_fifo_oclass = {
+	.handle = NV_ENGINE(FIFO, 0x10),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv10_fifo_ctor,
+		.dtor = nv04_fifo_dtor,
+		.init = nv04_fifo_init,
+		.fini = _nouveau_fifo_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/nv17.c b/drivers/gpu/drm/nouveau/core/engine/fifo/nv17.c
new file mode 100644
index 0000000..b96e6b0
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/nv17.c
@@ -0,0 +1,208 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/ramht.h>
+
+#include <subdev/instmem.h>
+#include <subdev/instmem/nv04.h>
+#include <subdev/fb.h>
+
+#include <engine/fifo.h>
+
+#include "nv04.h"
+
+static struct ramfc_desc
+nv17_ramfc[] = {
+	{ 32,  0, 0x00,  0, NV04_PFIFO_CACHE1_DMA_PUT },
+	{ 32,  0, 0x04,  0, NV04_PFIFO_CACHE1_DMA_GET },
+	{ 32,  0, 0x08,  0, NV10_PFIFO_CACHE1_REF_CNT },
+	{ 16,  0, 0x0c,  0, NV04_PFIFO_CACHE1_DMA_INSTANCE },
+	{ 16, 16, 0x0c,  0, NV04_PFIFO_CACHE1_DMA_DCOUNT },
+	{ 32,  0, 0x10,  0, NV04_PFIFO_CACHE1_DMA_STATE },
+	{ 32,  0, 0x14,  0, NV04_PFIFO_CACHE1_DMA_FETCH },
+	{ 32,  0, 0x18,  0, NV04_PFIFO_CACHE1_ENGINE },
+	{ 32,  0, 0x1c,  0, NV04_PFIFO_CACHE1_PULL1 },
+	{ 32,  0, 0x20,  0, NV10_PFIFO_CACHE1_ACQUIRE_VALUE },
+	{ 32,  0, 0x24,  0, NV10_PFIFO_CACHE1_ACQUIRE_TIMESTAMP },
+	{ 32,  0, 0x28,  0, NV10_PFIFO_CACHE1_ACQUIRE_TIMEOUT },
+	{ 32,  0, 0x2c,  0, NV10_PFIFO_CACHE1_SEMAPHORE },
+	{ 32,  0, 0x30,  0, NV10_PFIFO_CACHE1_DMA_SUBROUTINE },
+	{}
+};
+
+/*******************************************************************************
+ * FIFO channel objects
+ ******************************************************************************/
+
+static int
+nv17_fifo_chan_ctor(struct nouveau_object *parent,
+		    struct nouveau_object *engine,
+		    struct nouveau_oclass *oclass, void *data, u32 size,
+		    struct nouveau_object **pobject)
+{
+	struct nv04_fifo_priv *priv = (void *)engine;
+	struct nv04_fifo_chan *chan;
+	struct nv03_channel_dma_class *args = data;
+	int ret;
+
+	if (size < sizeof(*args))
+		return -EINVAL;
+
+	ret = nouveau_fifo_channel_create(parent, engine, oclass, 0, 0x800000,
+					  0x10000, args->pushbuf,
+					  (1 << NVDEV_ENGINE_DMAOBJ) |
+					  (1 << NVDEV_ENGINE_SW) |
+					  (1 << NVDEV_ENGINE_GR) |
+					  (1 << NVDEV_ENGINE_MPEG), /* NV31- */
+					  &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	nv_parent(chan)->object_attach = nv04_fifo_object_attach;
+	nv_parent(chan)->object_detach = nv04_fifo_object_detach;
+	nv_parent(chan)->context_attach = nv04_fifo_context_attach;
+	chan->ramfc = chan->base.chid * 64;
+
+	nv_wo32(priv->ramfc, chan->ramfc + 0x00, args->offset);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x04, args->offset);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x0c, chan->base.pushgpu->addr >> 4);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x14,
+			     NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES |
+			     NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES |
+#ifdef __BIG_ENDIAN
+			     NV_PFIFO_CACHE1_BIG_ENDIAN |
+#endif
+			     NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8);
+	return 0;
+}
+
+static struct nouveau_ofuncs
+nv17_fifo_ofuncs = {
+	.ctor = nv17_fifo_chan_ctor,
+	.dtor = nv04_fifo_chan_dtor,
+	.init = nv04_fifo_chan_init,
+	.fini = nv04_fifo_chan_fini,
+	.rd32 = _nouveau_fifo_channel_rd32,
+	.wr32 = _nouveau_fifo_channel_wr32,
+};
+
+static struct nouveau_oclass
+nv17_fifo_sclass[] = {
+	{ NV17_CHANNEL_DMA_CLASS, &nv17_fifo_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * FIFO context - basically just the instmem reserved for the channel
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv17_fifo_cclass = {
+	.handle = NV_ENGCTX(FIFO, 0x17),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv04_fifo_context_ctor,
+		.dtor = _nouveau_fifo_context_dtor,
+		.init = _nouveau_fifo_context_init,
+		.fini = _nouveau_fifo_context_fini,
+		.rd32 = _nouveau_fifo_context_rd32,
+		.wr32 = _nouveau_fifo_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PFIFO engine
+ ******************************************************************************/
+
+static int
+nv17_fifo_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv04_instmem_priv *imem = nv04_instmem(parent);
+	struct nv04_fifo_priv *priv;
+	int ret;
+
+	ret = nouveau_fifo_create(parent, engine, oclass, 0, 31, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nouveau_ramht_ref(imem->ramht, &priv->ramht);
+	nouveau_gpuobj_ref(imem->ramro, &priv->ramro);
+	nouveau_gpuobj_ref(imem->ramfc, &priv->ramfc);
+
+	nv_subdev(priv)->unit = 0x00000100;
+	nv_subdev(priv)->intr = nv04_fifo_intr;
+	nv_engine(priv)->cclass = &nv17_fifo_cclass;
+	nv_engine(priv)->sclass = nv17_fifo_sclass;
+	priv->base.pause = nv04_fifo_pause;
+	priv->base.start = nv04_fifo_start;
+	priv->ramfc_desc = nv17_ramfc;
+	return 0;
+}
+
+static int
+nv17_fifo_init(struct nouveau_object *object)
+{
+	struct nv04_fifo_priv *priv = (void *)object;
+	int ret;
+
+	ret = nouveau_fifo_init(&priv->base);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, NV04_PFIFO_DELAY_0, 0x000000ff);
+	nv_wr32(priv, NV04_PFIFO_DMA_TIMESLICE, 0x0101ffff);
+
+	nv_wr32(priv, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ |
+				       ((priv->ramht->bits - 9) << 16) |
+				        (priv->ramht->base.addr >> 8));
+	nv_wr32(priv, NV03_PFIFO_RAMRO, priv->ramro->addr >> 8);
+	nv_wr32(priv, NV03_PFIFO_RAMFC, priv->ramfc->addr >> 8 | 0x00010000);
+
+	nv_wr32(priv, NV03_PFIFO_CACHE1_PUSH1, priv->base.max);
+
+	nv_wr32(priv, NV03_PFIFO_INTR_0, 0xffffffff);
+	nv_wr32(priv, NV03_PFIFO_INTR_EN_0, 0xffffffff);
+
+	nv_wr32(priv, NV03_PFIFO_CACHE1_PUSH0, 1);
+	nv_wr32(priv, NV04_PFIFO_CACHE1_PULL0, 1);
+	nv_wr32(priv, NV03_PFIFO_CACHES, 1);
+	return 0;
+}
+
+struct nouveau_oclass
+nv17_fifo_oclass = {
+	.handle = NV_ENGINE(FIFO, 0x17),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv17_fifo_ctor,
+		.dtor = nv04_fifo_dtor,
+		.init = nv17_fifo_init,
+		.fini = _nouveau_fifo_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/nv40.c b/drivers/gpu/drm/nouveau/core/engine/fifo/nv40.c
new file mode 100644
index 0000000..559c3b4
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/nv40.c
@@ -0,0 +1,349 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/ramht.h>
+
+#include <subdev/instmem.h>
+#include <subdev/instmem/nv04.h>
+#include <subdev/fb.h>
+
+#include <engine/fifo.h>
+
+#include "nv04.h"
+
+static struct ramfc_desc
+nv40_ramfc[] = {
+	{ 32,  0, 0x00,  0, NV04_PFIFO_CACHE1_DMA_PUT },
+	{ 32,  0, 0x04,  0, NV04_PFIFO_CACHE1_DMA_GET },
+	{ 32,  0, 0x08,  0, NV10_PFIFO_CACHE1_REF_CNT },
+	{ 32,  0, 0x0c,  0, NV04_PFIFO_CACHE1_DMA_INSTANCE },
+	{ 32,  0, 0x10,  0, NV04_PFIFO_CACHE1_DMA_DCOUNT },
+	{ 32,  0, 0x14,  0, NV04_PFIFO_CACHE1_DMA_STATE },
+	{ 28,  0, 0x18,  0, NV04_PFIFO_CACHE1_DMA_FETCH },
+	{  2, 28, 0x18, 28, 0x002058 },
+	{ 32,  0, 0x1c,  0, NV04_PFIFO_CACHE1_ENGINE },
+	{ 32,  0, 0x20,  0, NV04_PFIFO_CACHE1_PULL1 },
+	{ 32,  0, 0x24,  0, NV10_PFIFO_CACHE1_ACQUIRE_VALUE },
+	{ 32,  0, 0x28,  0, NV10_PFIFO_CACHE1_ACQUIRE_TIMESTAMP },
+	{ 32,  0, 0x2c,  0, NV10_PFIFO_CACHE1_ACQUIRE_TIMEOUT },
+	{ 32,  0, 0x30,  0, NV10_PFIFO_CACHE1_SEMAPHORE },
+	{ 32,  0, 0x34,  0, NV10_PFIFO_CACHE1_DMA_SUBROUTINE },
+	{ 32,  0, 0x38,  0, NV40_PFIFO_GRCTX_INSTANCE },
+	{ 17,  0, 0x3c,  0, NV04_PFIFO_DMA_TIMESLICE },
+	{ 32,  0, 0x40,  0, 0x0032e4 },
+	{ 32,  0, 0x44,  0, 0x0032e8 },
+	{ 32,  0, 0x4c,  0, 0x002088 },
+	{ 32,  0, 0x50,  0, 0x003300 },
+	{ 32,  0, 0x54,  0, 0x00330c },
+	{}
+};
+
+/*******************************************************************************
+ * FIFO channel objects
+ ******************************************************************************/
+
+static int
+nv40_fifo_object_attach(struct nouveau_object *parent,
+			struct nouveau_object *object, u32 handle)
+{
+	struct nv04_fifo_priv *priv = (void *)parent->engine;
+	struct nv04_fifo_chan *chan = (void *)parent;
+	u32 context, chid = chan->base.chid;
+	int ret;
+
+	if (nv_iclass(object, NV_GPUOBJ_CLASS))
+		context = nv_gpuobj(object)->addr >> 4;
+	else
+		context = 0x00000004; /* just non-zero */
+
+	switch (nv_engidx(object->engine)) {
+	case NVDEV_ENGINE_DMAOBJ:
+	case NVDEV_ENGINE_SW:
+		context |= 0x00000000;
+		break;
+	case NVDEV_ENGINE_GR:
+		context |= 0x00100000;
+		break;
+	case NVDEV_ENGINE_MPEG:
+		context |= 0x00200000;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	context |= chid << 23;
+
+	mutex_lock(&nv_subdev(priv)->mutex);
+	ret = nouveau_ramht_insert(priv->ramht, chid, handle, context);
+	mutex_unlock(&nv_subdev(priv)->mutex);
+	return ret;
+}
+
+static int
+nv40_fifo_context_attach(struct nouveau_object *parent,
+			 struct nouveau_object *engctx)
+{
+	struct nv04_fifo_priv *priv = (void *)parent->engine;
+	struct nv04_fifo_chan *chan = (void *)parent;
+	unsigned long flags;
+	u32 reg, ctx;
+
+	switch (nv_engidx(engctx->engine)) {
+	case NVDEV_ENGINE_SW:
+		return 0;
+	case NVDEV_ENGINE_GR:
+		reg = 0x32e0;
+		ctx = 0x38;
+		break;
+	case NVDEV_ENGINE_MPEG:
+		reg = 0x330c;
+		ctx = 0x54;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	spin_lock_irqsave(&priv->base.lock, flags);
+	nv_engctx(engctx)->addr = nv_gpuobj(engctx)->addr >> 4;
+	nv_mask(priv, 0x002500, 0x00000001, 0x00000000);
+
+	if ((nv_rd32(priv, 0x003204) & priv->base.max) == chan->base.chid)
+		nv_wr32(priv, reg, nv_engctx(engctx)->addr);
+	nv_wo32(priv->ramfc, chan->ramfc + ctx, nv_engctx(engctx)->addr);
+
+	nv_mask(priv, 0x002500, 0x00000001, 0x00000001);
+	spin_unlock_irqrestore(&priv->base.lock, flags);
+	return 0;
+}
+
+static int
+nv40_fifo_context_detach(struct nouveau_object *parent, bool suspend,
+			 struct nouveau_object *engctx)
+{
+	struct nv04_fifo_priv *priv = (void *)parent->engine;
+	struct nv04_fifo_chan *chan = (void *)parent;
+	unsigned long flags;
+	u32 reg, ctx;
+
+	switch (nv_engidx(engctx->engine)) {
+	case NVDEV_ENGINE_SW:
+		return 0;
+	case NVDEV_ENGINE_GR:
+		reg = 0x32e0;
+		ctx = 0x38;
+		break;
+	case NVDEV_ENGINE_MPEG:
+		reg = 0x330c;
+		ctx = 0x54;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	spin_lock_irqsave(&priv->base.lock, flags);
+	nv_mask(priv, 0x002500, 0x00000001, 0x00000000);
+
+	if ((nv_rd32(priv, 0x003204) & priv->base.max) == chan->base.chid)
+		nv_wr32(priv, reg, 0x00000000);
+	nv_wo32(priv->ramfc, chan->ramfc + ctx, 0x00000000);
+
+	nv_mask(priv, 0x002500, 0x00000001, 0x00000001);
+	spin_unlock_irqrestore(&priv->base.lock, flags);
+	return 0;
+}
+
+static int
+nv40_fifo_chan_ctor(struct nouveau_object *parent,
+		    struct nouveau_object *engine,
+		    struct nouveau_oclass *oclass, void *data, u32 size,
+		    struct nouveau_object **pobject)
+{
+	struct nv04_fifo_priv *priv = (void *)engine;
+	struct nv04_fifo_chan *chan;
+	struct nv03_channel_dma_class *args = data;
+	int ret;
+
+	if (size < sizeof(*args))
+		return -EINVAL;
+
+	ret = nouveau_fifo_channel_create(parent, engine, oclass, 0, 0xc00000,
+					  0x1000, args->pushbuf,
+					  (1 << NVDEV_ENGINE_DMAOBJ) |
+					  (1 << NVDEV_ENGINE_SW) |
+					  (1 << NVDEV_ENGINE_GR) |
+					  (1 << NVDEV_ENGINE_MPEG), &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	nv_parent(chan)->context_attach = nv40_fifo_context_attach;
+	nv_parent(chan)->context_detach = nv40_fifo_context_detach;
+	nv_parent(chan)->object_attach = nv40_fifo_object_attach;
+	nv_parent(chan)->object_detach = nv04_fifo_object_detach;
+	chan->ramfc = chan->base.chid * 128;
+
+	nv_wo32(priv->ramfc, chan->ramfc + 0x00, args->offset);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x04, args->offset);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x0c, chan->base.pushgpu->addr >> 4);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x18, 0x30000000 |
+			     NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES |
+			     NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES |
+#ifdef __BIG_ENDIAN
+			     NV_PFIFO_CACHE1_BIG_ENDIAN |
+#endif
+			     NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8);
+	nv_wo32(priv->ramfc, chan->ramfc + 0x3c, 0x0001ffff);
+	return 0;
+}
+
+static struct nouveau_ofuncs
+nv40_fifo_ofuncs = {
+	.ctor = nv40_fifo_chan_ctor,
+	.dtor = nv04_fifo_chan_dtor,
+	.init = nv04_fifo_chan_init,
+	.fini = nv04_fifo_chan_fini,
+	.rd32 = _nouveau_fifo_channel_rd32,
+	.wr32 = _nouveau_fifo_channel_wr32,
+};
+
+static struct nouveau_oclass
+nv40_fifo_sclass[] = {
+	{ NV40_CHANNEL_DMA_CLASS, &nv40_fifo_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * FIFO context - basically just the instmem reserved for the channel
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv40_fifo_cclass = {
+	.handle = NV_ENGCTX(FIFO, 0x40),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv04_fifo_context_ctor,
+		.dtor = _nouveau_fifo_context_dtor,
+		.init = _nouveau_fifo_context_init,
+		.fini = _nouveau_fifo_context_fini,
+		.rd32 = _nouveau_fifo_context_rd32,
+		.wr32 = _nouveau_fifo_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PFIFO engine
+ ******************************************************************************/
+
+static int
+nv40_fifo_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv04_instmem_priv *imem = nv04_instmem(parent);
+	struct nv04_fifo_priv *priv;
+	int ret;
+
+	ret = nouveau_fifo_create(parent, engine, oclass, 0, 31, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nouveau_ramht_ref(imem->ramht, &priv->ramht);
+	nouveau_gpuobj_ref(imem->ramro, &priv->ramro);
+	nouveau_gpuobj_ref(imem->ramfc, &priv->ramfc);
+
+	nv_subdev(priv)->unit = 0x00000100;
+	nv_subdev(priv)->intr = nv04_fifo_intr;
+	nv_engine(priv)->cclass = &nv40_fifo_cclass;
+	nv_engine(priv)->sclass = nv40_fifo_sclass;
+	priv->base.pause = nv04_fifo_pause;
+	priv->base.start = nv04_fifo_start;
+	priv->ramfc_desc = nv40_ramfc;
+	return 0;
+}
+
+static int
+nv40_fifo_init(struct nouveau_object *object)
+{
+	struct nv04_fifo_priv *priv = (void *)object;
+	struct nouveau_fb *pfb = nouveau_fb(object);
+	int ret;
+
+	ret = nouveau_fifo_init(&priv->base);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, 0x002040, 0x000000ff);
+	nv_wr32(priv, 0x002044, 0x2101ffff);
+	nv_wr32(priv, 0x002058, 0x00000001);
+
+	nv_wr32(priv, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ |
+				       ((priv->ramht->bits - 9) << 16) |
+				        (priv->ramht->base.addr >> 8));
+	nv_wr32(priv, NV03_PFIFO_RAMRO, priv->ramro->addr >> 8);
+
+	switch (nv_device(priv)->chipset) {
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		nv_wr32(priv, 0x002230, 0x00000001);
+	case 0x40:
+	case 0x41:
+	case 0x42:
+	case 0x43:
+	case 0x45:
+	case 0x48:
+		nv_wr32(priv, 0x002220, 0x00030002);
+		break;
+	default:
+		nv_wr32(priv, 0x002230, 0x00000000);
+		nv_wr32(priv, 0x002220, ((pfb->ram.size - 512 * 1024 +
+					 priv->ramfc->addr) >> 16) |
+					0x00030000);
+		break;
+	}
+
+	nv_wr32(priv, NV03_PFIFO_CACHE1_PUSH1, priv->base.max);
+
+	nv_wr32(priv, NV03_PFIFO_INTR_0, 0xffffffff);
+	nv_wr32(priv, NV03_PFIFO_INTR_EN_0, 0xffffffff);
+
+	nv_wr32(priv, NV03_PFIFO_CACHE1_PUSH0, 1);
+	nv_wr32(priv, NV04_PFIFO_CACHE1_PULL0, 1);
+	nv_wr32(priv, NV03_PFIFO_CACHES, 1);
+	return 0;
+}
+
+struct nouveau_oclass
+nv40_fifo_oclass = {
+	.handle = NV_ENGINE(FIFO, 0x40),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv40_fifo_ctor,
+		.dtor = nv04_fifo_dtor,
+		.init = nv40_fifo_init,
+		.fini = _nouveau_fifo_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/nv50.c b/drivers/gpu/drm/nouveau/core/engine/fifo/nv50.c
new file mode 100644
index 0000000..536e763
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/nv50.c
@@ -0,0 +1,502 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/client.h>
+#include <core/engctx.h>
+#include <core/ramht.h>
+#include <core/class.h>
+#include <core/math.h>
+
+#include <subdev/timer.h>
+#include <subdev/bar.h>
+
+#include <engine/dmaobj.h>
+#include <engine/fifo.h>
+
+#include "nv50.h"
+
+/*******************************************************************************
+ * FIFO channel objects
+ ******************************************************************************/
+
+void
+nv50_fifo_playlist_update(struct nv50_fifo_priv *priv)
+{
+	struct nouveau_bar *bar = nouveau_bar(priv);
+	struct nouveau_gpuobj *cur;
+	int i, p;
+
+	cur = priv->playlist[priv->cur_playlist];
+	priv->cur_playlist = !priv->cur_playlist;
+
+	for (i = priv->base.min, p = 0; i < priv->base.max; i++) {
+		if (nv_rd32(priv, 0x002600 + (i * 4)) & 0x80000000)
+			nv_wo32(cur, p++ * 4, i);
+	}
+
+	bar->flush(bar);
+
+	nv_wr32(priv, 0x0032f4, cur->addr >> 12);
+	nv_wr32(priv, 0x0032ec, p);
+	nv_wr32(priv, 0x002500, 0x00000101);
+}
+
+static int
+nv50_fifo_context_attach(struct nouveau_object *parent,
+			 struct nouveau_object *object)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nv50_fifo_base *base = (void *)parent->parent;
+	struct nouveau_gpuobj *ectx = (void *)object;
+	u64 limit = ectx->addr + ectx->size - 1;
+	u64 start = ectx->addr;
+	u32 addr;
+
+	switch (nv_engidx(object->engine)) {
+	case NVDEV_ENGINE_SW   : return 0;
+	case NVDEV_ENGINE_GR   : addr = 0x0000; break;
+	case NVDEV_ENGINE_MPEG : addr = 0x0060; break;
+	default:
+		return -EINVAL;
+	}
+
+	nv_engctx(ectx)->addr = nv_gpuobj(base)->addr >> 12;
+	nv_wo32(base->eng, addr + 0x00, 0x00190000);
+	nv_wo32(base->eng, addr + 0x04, lower_32_bits(limit));
+	nv_wo32(base->eng, addr + 0x08, lower_32_bits(start));
+	nv_wo32(base->eng, addr + 0x0c, upper_32_bits(limit) << 24 |
+					upper_32_bits(start));
+	nv_wo32(base->eng, addr + 0x10, 0x00000000);
+	nv_wo32(base->eng, addr + 0x14, 0x00000000);
+	bar->flush(bar);
+	return 0;
+}
+
+static int
+nv50_fifo_context_detach(struct nouveau_object *parent, bool suspend,
+			 struct nouveau_object *object)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nv50_fifo_priv *priv = (void *)parent->engine;
+	struct nv50_fifo_base *base = (void *)parent->parent;
+	struct nv50_fifo_chan *chan = (void *)parent;
+	u32 addr, me;
+	int ret = 0;
+
+	switch (nv_engidx(object->engine)) {
+	case NVDEV_ENGINE_SW   : return 0;
+	case NVDEV_ENGINE_GR   : addr = 0x0000; break;
+	case NVDEV_ENGINE_MPEG : addr = 0x0060; break;
+	default:
+		return -EINVAL;
+	}
+
+	nv_wo32(base->eng, addr + 0x00, 0x00000000);
+	nv_wo32(base->eng, addr + 0x04, 0x00000000);
+	nv_wo32(base->eng, addr + 0x08, 0x00000000);
+	nv_wo32(base->eng, addr + 0x0c, 0x00000000);
+	nv_wo32(base->eng, addr + 0x10, 0x00000000);
+	nv_wo32(base->eng, addr + 0x14, 0x00000000);
+	bar->flush(bar);
+
+	/* HW bug workaround:
+	 *
+	 * PFIFO will hang forever if the connected engines don't report
+	 * that they've processed the context switch request.
+	 *
+	 * In order for the kickoff to work, we need to ensure all the
+	 * connected engines are in a state where they can answer.
+	 *
+	 * Newer chipsets don't seem to suffer from this issue, and well,
+	 * there's also a "ignore these engines" bitmask reg we can use
+	 * if we hit the issue there..
+	 */
+	me = nv_mask(priv, 0x00b860, 0x00000001, 0x00000001);
+
+	/* do the kickoff... */
+	nv_wr32(priv, 0x0032fc, nv_gpuobj(base)->addr >> 12);
+	if (!nv_wait_ne(priv, 0x0032fc, 0xffffffff, 0xffffffff)) {
+		nv_error(priv, "channel %d unload timeout\n", chan->base.chid);
+		if (suspend)
+			ret = -EBUSY;
+	}
+
+	nv_wr32(priv, 0x00b860, me);
+	return ret;
+}
+
+static int
+nv50_fifo_object_attach(struct nouveau_object *parent,
+			struct nouveau_object *object, u32 handle)
+{
+	struct nv50_fifo_chan *chan = (void *)parent;
+	u32 context;
+
+	if (nv_iclass(object, NV_GPUOBJ_CLASS))
+		context = nv_gpuobj(object)->node->offset >> 4;
+	else
+		context = 0x00000004; /* just non-zero */
+
+	switch (nv_engidx(object->engine)) {
+	case NVDEV_ENGINE_DMAOBJ:
+	case NVDEV_ENGINE_SW    : context |= 0x00000000; break;
+	case NVDEV_ENGINE_GR    : context |= 0x00100000; break;
+	case NVDEV_ENGINE_MPEG  : context |= 0x00200000; break;
+	default:
+		return -EINVAL;
+	}
+
+	return nouveau_ramht_insert(chan->ramht, 0, handle, context);
+}
+
+void
+nv50_fifo_object_detach(struct nouveau_object *parent, int cookie)
+{
+	struct nv50_fifo_chan *chan = (void *)parent;
+	nouveau_ramht_remove(chan->ramht, cookie);
+}
+
+static int
+nv50_fifo_chan_ctor_dma(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nv50_fifo_base *base = (void *)parent;
+	struct nv50_fifo_chan *chan;
+	struct nv03_channel_dma_class *args = data;
+	int ret;
+
+	if (size < sizeof(*args))
+		return -EINVAL;
+
+	ret = nouveau_fifo_channel_create(parent, engine, oclass, 0, 0xc00000,
+					  0x2000, args->pushbuf,
+					  (1 << NVDEV_ENGINE_DMAOBJ) |
+					  (1 << NVDEV_ENGINE_SW) |
+					  (1 << NVDEV_ENGINE_GR) |
+					  (1 << NVDEV_ENGINE_MPEG), &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	nv_parent(chan)->context_attach = nv50_fifo_context_attach;
+	nv_parent(chan)->context_detach = nv50_fifo_context_detach;
+	nv_parent(chan)->object_attach = nv50_fifo_object_attach;
+	nv_parent(chan)->object_detach = nv50_fifo_object_detach;
+
+	ret = nouveau_ramht_new(parent, parent, 0x8000, 16, &chan->ramht);
+	if (ret)
+		return ret;
+
+	nv_wo32(base->ramfc, 0x08, lower_32_bits(args->offset));
+	nv_wo32(base->ramfc, 0x0c, upper_32_bits(args->offset));
+	nv_wo32(base->ramfc, 0x10, lower_32_bits(args->offset));
+	nv_wo32(base->ramfc, 0x14, upper_32_bits(args->offset));
+	nv_wo32(base->ramfc, 0x3c, 0x003f6078);
+	nv_wo32(base->ramfc, 0x44, 0x01003fff);
+	nv_wo32(base->ramfc, 0x48, chan->base.pushgpu->node->offset >> 4);
+	nv_wo32(base->ramfc, 0x4c, 0xffffffff);
+	nv_wo32(base->ramfc, 0x60, 0x7fffffff);
+	nv_wo32(base->ramfc, 0x78, 0x00000000);
+	nv_wo32(base->ramfc, 0x7c, 0x30000001);
+	nv_wo32(base->ramfc, 0x80, ((chan->ramht->bits - 9) << 27) |
+				   (4 << 24) /* SEARCH_FULL */ |
+				   (chan->ramht->base.node->offset >> 4));
+	bar->flush(bar);
+	return 0;
+}
+
+static int
+nv50_fifo_chan_ctor_ind(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nv50_channel_ind_class *args = data;
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nv50_fifo_base *base = (void *)parent;
+	struct nv50_fifo_chan *chan;
+	u64 ioffset, ilength;
+	int ret;
+
+	if (size < sizeof(*args))
+		return -EINVAL;
+
+	ret = nouveau_fifo_channel_create(parent, engine, oclass, 0, 0xc00000,
+					  0x2000, args->pushbuf,
+					  (1 << NVDEV_ENGINE_DMAOBJ) |
+					  (1 << NVDEV_ENGINE_SW) |
+					  (1 << NVDEV_ENGINE_GR) |
+					  (1 << NVDEV_ENGINE_MPEG), &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	nv_parent(chan)->context_attach = nv50_fifo_context_attach;
+	nv_parent(chan)->context_detach = nv50_fifo_context_detach;
+	nv_parent(chan)->object_attach = nv50_fifo_object_attach;
+	nv_parent(chan)->object_detach = nv50_fifo_object_detach;
+
+	ret = nouveau_ramht_new(parent, parent, 0x8000, 16, &chan->ramht);
+	if (ret)
+		return ret;
+
+	ioffset = args->ioffset;
+	ilength = log2i(args->ilength / 8);
+
+	nv_wo32(base->ramfc, 0x3c, 0x403f6078);
+	nv_wo32(base->ramfc, 0x44, 0x01003fff);
+	nv_wo32(base->ramfc, 0x48, chan->base.pushgpu->node->offset >> 4);
+	nv_wo32(base->ramfc, 0x50, lower_32_bits(ioffset));
+	nv_wo32(base->ramfc, 0x54, upper_32_bits(ioffset) | (ilength << 16));
+	nv_wo32(base->ramfc, 0x60, 0x7fffffff);
+	nv_wo32(base->ramfc, 0x78, 0x00000000);
+	nv_wo32(base->ramfc, 0x7c, 0x30000001);
+	nv_wo32(base->ramfc, 0x80, ((chan->ramht->bits - 9) << 27) |
+				   (4 << 24) /* SEARCH_FULL */ |
+				   (chan->ramht->base.node->offset >> 4));
+	bar->flush(bar);
+	return 0;
+}
+
+void
+nv50_fifo_chan_dtor(struct nouveau_object *object)
+{
+	struct nv50_fifo_chan *chan = (void *)object;
+	nouveau_ramht_ref(NULL, &chan->ramht);
+	nouveau_fifo_channel_destroy(&chan->base);
+}
+
+static int
+nv50_fifo_chan_init(struct nouveau_object *object)
+{
+	struct nv50_fifo_priv *priv = (void *)object->engine;
+	struct nv50_fifo_base *base = (void *)object->parent;
+	struct nv50_fifo_chan *chan = (void *)object;
+	struct nouveau_gpuobj *ramfc = base->ramfc;
+	u32 chid = chan->base.chid;
+	int ret;
+
+	ret = nouveau_fifo_channel_init(&chan->base);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, 0x002600 + (chid * 4), 0x80000000 | ramfc->addr >> 12);
+	nv50_fifo_playlist_update(priv);
+	return 0;
+}
+
+int
+nv50_fifo_chan_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv50_fifo_priv *priv = (void *)object->engine;
+	struct nv50_fifo_chan *chan = (void *)object;
+	u32 chid = chan->base.chid;
+
+	/* remove channel from playlist, fifo will unload context */
+	nv_mask(priv, 0x002600 + (chid * 4), 0x80000000, 0x00000000);
+	nv50_fifo_playlist_update(priv);
+	nv_wr32(priv, 0x002600 + (chid * 4), 0x00000000);
+
+	return nouveau_fifo_channel_fini(&chan->base, suspend);
+}
+
+static struct nouveau_ofuncs
+nv50_fifo_ofuncs_dma = {
+	.ctor = nv50_fifo_chan_ctor_dma,
+	.dtor = nv50_fifo_chan_dtor,
+	.init = nv50_fifo_chan_init,
+	.fini = nv50_fifo_chan_fini,
+	.rd32 = _nouveau_fifo_channel_rd32,
+	.wr32 = _nouveau_fifo_channel_wr32,
+};
+
+static struct nouveau_ofuncs
+nv50_fifo_ofuncs_ind = {
+	.ctor = nv50_fifo_chan_ctor_ind,
+	.dtor = nv50_fifo_chan_dtor,
+	.init = nv50_fifo_chan_init,
+	.fini = nv50_fifo_chan_fini,
+	.rd32 = _nouveau_fifo_channel_rd32,
+	.wr32 = _nouveau_fifo_channel_wr32,
+};
+
+static struct nouveau_oclass
+nv50_fifo_sclass[] = {
+	{ NV50_CHANNEL_DMA_CLASS, &nv50_fifo_ofuncs_dma },
+	{ NV50_CHANNEL_IND_CLASS, &nv50_fifo_ofuncs_ind },
+	{}
+};
+
+/*******************************************************************************
+ * FIFO context - basically just the instmem reserved for the channel
+ ******************************************************************************/
+
+static int
+nv50_fifo_context_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nv50_fifo_base *base;
+	int ret;
+
+	ret = nouveau_fifo_context_create(parent, engine, oclass, NULL, 0x10000,
+				          0x1000, NVOBJ_FLAG_HEAP, &base);
+	*pobject = nv_object(base);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, nv_object(base), 0x0200, 0x1000,
+				 NVOBJ_FLAG_ZERO_ALLOC, &base->ramfc);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, nv_object(base), 0x1200, 0,
+				 NVOBJ_FLAG_ZERO_ALLOC, &base->eng);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, nv_object(base), 0x4000, 0, 0,
+				&base->pgd);
+	if (ret)
+		return ret;
+
+	ret = nouveau_vm_ref(nouveau_client(parent)->vm, &base->vm, base->pgd);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+void
+nv50_fifo_context_dtor(struct nouveau_object *object)
+{
+	struct nv50_fifo_base *base = (void *)object;
+	nouveau_vm_ref(NULL, &base->vm, base->pgd);
+	nouveau_gpuobj_ref(NULL, &base->pgd);
+	nouveau_gpuobj_ref(NULL, &base->eng);
+	nouveau_gpuobj_ref(NULL, &base->ramfc);
+	nouveau_gpuobj_ref(NULL, &base->cache);
+	nouveau_fifo_context_destroy(&base->base);
+}
+
+static struct nouveau_oclass
+nv50_fifo_cclass = {
+	.handle = NV_ENGCTX(FIFO, 0x50),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv50_fifo_context_ctor,
+		.dtor = nv50_fifo_context_dtor,
+		.init = _nouveau_fifo_context_init,
+		.fini = _nouveau_fifo_context_fini,
+		.rd32 = _nouveau_fifo_context_rd32,
+		.wr32 = _nouveau_fifo_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PFIFO engine
+ ******************************************************************************/
+
+static int
+nv50_fifo_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv50_fifo_priv *priv;
+	int ret;
+
+	ret = nouveau_fifo_create(parent, engine, oclass, 1, 127, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 128 * 4, 0x1000, 0,
+				&priv->playlist[0]);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 128 * 4, 0x1000, 0,
+				&priv->playlist[1]);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00000100;
+	nv_subdev(priv)->intr = nv04_fifo_intr;
+	nv_engine(priv)->cclass = &nv50_fifo_cclass;
+	nv_engine(priv)->sclass = nv50_fifo_sclass;
+	return 0;
+}
+
+void
+nv50_fifo_dtor(struct nouveau_object *object)
+{
+	struct nv50_fifo_priv *priv = (void *)object;
+
+	nouveau_gpuobj_ref(NULL, &priv->playlist[1]);
+	nouveau_gpuobj_ref(NULL, &priv->playlist[0]);
+
+	nouveau_fifo_destroy(&priv->base);
+}
+
+int
+nv50_fifo_init(struct nouveau_object *object)
+{
+	struct nv50_fifo_priv *priv = (void *)object;
+	int ret, i;
+
+	ret = nouveau_fifo_init(&priv->base);
+	if (ret)
+		return ret;
+
+	nv_mask(priv, 0x000200, 0x00000100, 0x00000000);
+	nv_mask(priv, 0x000200, 0x00000100, 0x00000100);
+	nv_wr32(priv, 0x00250c, 0x6f3cfc34);
+	nv_wr32(priv, 0x002044, 0x01003fff);
+
+	nv_wr32(priv, 0x002100, 0xffffffff);
+	nv_wr32(priv, 0x002140, 0xffffffff);
+
+	for (i = 0; i < 128; i++)
+		nv_wr32(priv, 0x002600 + (i * 4), 0x00000000);
+	nv50_fifo_playlist_update(priv);
+
+	nv_wr32(priv, 0x003200, 0x00000001);
+	nv_wr32(priv, 0x003250, 0x00000001);
+	nv_wr32(priv, 0x002500, 0x00000001);
+	return 0;
+}
+
+struct nouveau_oclass
+nv50_fifo_oclass = {
+	.handle = NV_ENGINE(FIFO, 0x50),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv50_fifo_ctor,
+		.dtor = nv50_fifo_dtor,
+		.init = nv50_fifo_init,
+		.fini = _nouveau_fifo_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/nv50.h b/drivers/gpu/drm/nouveau/core/engine/fifo/nv50.h
new file mode 100644
index 0000000..3a9ceb3
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/nv50.h
@@ -0,0 +1,36 @@
+#ifndef __NV50_FIFO_H__
+#define __NV50_FIFO_H__
+
+struct nv50_fifo_priv {
+	struct nouveau_fifo base;
+	struct nouveau_gpuobj *playlist[2];
+	int cur_playlist;
+};
+
+struct nv50_fifo_base {
+	struct nouveau_fifo_base base;
+	struct nouveau_gpuobj *ramfc;
+	struct nouveau_gpuobj *cache;
+	struct nouveau_gpuobj *eng;
+	struct nouveau_gpuobj *pgd;
+	struct nouveau_vm *vm;
+};
+
+struct nv50_fifo_chan {
+	struct nouveau_fifo_chan base;
+	u32 subc[8];
+	struct nouveau_ramht *ramht;
+};
+
+void nv50_fifo_playlist_update(struct nv50_fifo_priv *);
+
+void nv50_fifo_object_detach(struct nouveau_object *, int);
+void nv50_fifo_chan_dtor(struct nouveau_object *);
+int  nv50_fifo_chan_fini(struct nouveau_object *, bool);
+
+void nv50_fifo_context_dtor(struct nouveau_object *);
+
+void nv50_fifo_dtor(struct nouveau_object *);
+int  nv50_fifo_init(struct nouveau_object *);
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/nv84.c b/drivers/gpu/drm/nouveau/core/engine/fifo/nv84.c
new file mode 100644
index 0000000..b4fd26d
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/nv84.c
@@ -0,0 +1,420 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/client.h>
+#include <core/engctx.h>
+#include <core/ramht.h>
+#include <core/class.h>
+#include <core/math.h>
+
+#include <subdev/timer.h>
+#include <subdev/bar.h>
+
+#include <engine/dmaobj.h>
+#include <engine/fifo.h>
+
+#include "nv50.h"
+
+/*******************************************************************************
+ * FIFO channel objects
+ ******************************************************************************/
+
+static int
+nv84_fifo_context_attach(struct nouveau_object *parent,
+			 struct nouveau_object *object)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nv50_fifo_base *base = (void *)parent->parent;
+	struct nouveau_gpuobj *ectx = (void *)object;
+	u64 limit = ectx->addr + ectx->size - 1;
+	u64 start = ectx->addr;
+	u32 addr;
+
+	switch (nv_engidx(object->engine)) {
+	case NVDEV_ENGINE_SW   : return 0;
+	case NVDEV_ENGINE_GR   : addr = 0x0020; break;
+	case NVDEV_ENGINE_MPEG : addr = 0x0060; break;
+	case NVDEV_ENGINE_CRYPT: addr = 0x00a0; break;
+	case NVDEV_ENGINE_COPY0: addr = 0x00c0; break;
+	default:
+		return -EINVAL;
+	}
+
+	nv_engctx(ectx)->addr = nv_gpuobj(base)->addr >> 12;
+	nv_wo32(base->eng, addr + 0x00, 0x00190000);
+	nv_wo32(base->eng, addr + 0x04, lower_32_bits(limit));
+	nv_wo32(base->eng, addr + 0x08, lower_32_bits(start));
+	nv_wo32(base->eng, addr + 0x0c, upper_32_bits(limit) << 24 |
+					upper_32_bits(start));
+	nv_wo32(base->eng, addr + 0x10, 0x00000000);
+	nv_wo32(base->eng, addr + 0x14, 0x00000000);
+	bar->flush(bar);
+	return 0;
+}
+
+static int
+nv84_fifo_context_detach(struct nouveau_object *parent, bool suspend,
+			 struct nouveau_object *object)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nv50_fifo_priv *priv = (void *)parent->engine;
+	struct nv50_fifo_base *base = (void *)parent->parent;
+	struct nv50_fifo_chan *chan = (void *)parent;
+	u32 addr, save, engn;
+	bool done;
+
+	switch (nv_engidx(object->engine)) {
+	case NVDEV_ENGINE_SW   : return 0;
+	case NVDEV_ENGINE_GR   : engn = 0; addr = 0x0020; break;
+	case NVDEV_ENGINE_MPEG : engn = 1; addr = 0x0060; break;
+	case NVDEV_ENGINE_CRYPT: engn = 4; addr = 0x00a0; break;
+	case NVDEV_ENGINE_COPY0: engn = 2; addr = 0x00c0; break;
+	default:
+		return -EINVAL;
+	}
+
+	nv_wo32(base->eng, addr + 0x00, 0x00000000);
+	nv_wo32(base->eng, addr + 0x04, 0x00000000);
+	nv_wo32(base->eng, addr + 0x08, 0x00000000);
+	nv_wo32(base->eng, addr + 0x0c, 0x00000000);
+	nv_wo32(base->eng, addr + 0x10, 0x00000000);
+	nv_wo32(base->eng, addr + 0x14, 0x00000000);
+	bar->flush(bar);
+
+	save = nv_mask(priv, 0x002520, 0x0000003f, 1 << engn);
+	nv_wr32(priv, 0x0032fc, nv_gpuobj(base)->addr >> 12);
+	done = nv_wait_ne(priv, 0x0032fc, 0xffffffff, 0xffffffff);
+	nv_wr32(priv, 0x002520, save);
+	if (!done) {
+		nv_error(priv, "channel %d unload timeout\n", chan->base.chid);
+		if (suspend)
+			return -EBUSY;
+	}
+	return 0;
+}
+
+static int
+nv84_fifo_object_attach(struct nouveau_object *parent,
+			struct nouveau_object *object, u32 handle)
+{
+	struct nv50_fifo_chan *chan = (void *)parent;
+	u32 context;
+
+	if (nv_iclass(object, NV_GPUOBJ_CLASS))
+		context = nv_gpuobj(object)->node->offset >> 4;
+	else
+		context = 0x00000004; /* just non-zero */
+
+	switch (nv_engidx(object->engine)) {
+	case NVDEV_ENGINE_DMAOBJ:
+	case NVDEV_ENGINE_SW    : context |= 0x00000000; break;
+	case NVDEV_ENGINE_GR    : context |= 0x00100000; break;
+	case NVDEV_ENGINE_MPEG  :
+	case NVDEV_ENGINE_PPP   : context |= 0x00200000; break;
+	case NVDEV_ENGINE_ME    :
+	case NVDEV_ENGINE_COPY0 : context |= 0x00300000; break;
+	case NVDEV_ENGINE_VP    : context |= 0x00400000; break;
+	case NVDEV_ENGINE_CRYPT :
+	case NVDEV_ENGINE_UNK1C1: context |= 0x00500000; break;
+	case NVDEV_ENGINE_BSP   : context |= 0x00600000; break;
+	default:
+		return -EINVAL;
+	}
+
+	return nouveau_ramht_insert(chan->ramht, 0, handle, context);
+}
+
+static int
+nv84_fifo_chan_ctor_dma(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nv50_fifo_base *base = (void *)parent;
+	struct nv50_fifo_chan *chan;
+	struct nv03_channel_dma_class *args = data;
+	int ret;
+
+	if (size < sizeof(*args))
+		return -EINVAL;
+
+	ret = nouveau_fifo_channel_create(parent, engine, oclass, 0, 0xc00000,
+					  0x2000, args->pushbuf,
+					  (1 << NVDEV_ENGINE_DMAOBJ) |
+					  (1 << NVDEV_ENGINE_SW) |
+					  (1 << NVDEV_ENGINE_GR) |
+					  (1 << NVDEV_ENGINE_MPEG) |
+					  (1 << NVDEV_ENGINE_ME) |
+					  (1 << NVDEV_ENGINE_VP) |
+					  (1 << NVDEV_ENGINE_CRYPT) |
+					  (1 << NVDEV_ENGINE_BSP) |
+					  (1 << NVDEV_ENGINE_PPP) |
+					  (1 << NVDEV_ENGINE_COPY0) |
+					  (1 << NVDEV_ENGINE_UNK1C1), &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	ret = nouveau_ramht_new(parent, parent, 0x8000, 16, &chan->ramht);
+	if (ret)
+		return ret;
+
+	nv_parent(chan)->context_attach = nv84_fifo_context_attach;
+	nv_parent(chan)->context_detach = nv84_fifo_context_detach;
+	nv_parent(chan)->object_attach = nv84_fifo_object_attach;
+	nv_parent(chan)->object_detach = nv50_fifo_object_detach;
+
+	nv_wo32(base->ramfc, 0x08, lower_32_bits(args->offset));
+	nv_wo32(base->ramfc, 0x0c, upper_32_bits(args->offset));
+	nv_wo32(base->ramfc, 0x10, lower_32_bits(args->offset));
+	nv_wo32(base->ramfc, 0x14, upper_32_bits(args->offset));
+	nv_wo32(base->ramfc, 0x3c, 0x003f6078);
+	nv_wo32(base->ramfc, 0x44, 0x01003fff);
+	nv_wo32(base->ramfc, 0x48, chan->base.pushgpu->node->offset >> 4);
+	nv_wo32(base->ramfc, 0x4c, 0xffffffff);
+	nv_wo32(base->ramfc, 0x60, 0x7fffffff);
+	nv_wo32(base->ramfc, 0x78, 0x00000000);
+	nv_wo32(base->ramfc, 0x7c, 0x30000001);
+	nv_wo32(base->ramfc, 0x80, ((chan->ramht->bits - 9) << 27) |
+				   (4 << 24) /* SEARCH_FULL */ |
+				   (chan->ramht->base.node->offset >> 4));
+	nv_wo32(base->ramfc, 0x88, base->cache->addr >> 10);
+	nv_wo32(base->ramfc, 0x98, nv_gpuobj(base)->addr >> 12);
+	bar->flush(bar);
+	return 0;
+}
+
+static int
+nv84_fifo_chan_ctor_ind(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nv50_fifo_base *base = (void *)parent;
+	struct nv50_fifo_chan *chan;
+	struct nv50_channel_ind_class *args = data;
+	u64 ioffset, ilength;
+	int ret;
+
+	if (size < sizeof(*args))
+		return -EINVAL;
+
+	ret = nouveau_fifo_channel_create(parent, engine, oclass, 0, 0xc00000,
+					  0x2000, args->pushbuf,
+					  (1 << NVDEV_ENGINE_DMAOBJ) |
+					  (1 << NVDEV_ENGINE_SW) |
+					  (1 << NVDEV_ENGINE_GR) |
+					  (1 << NVDEV_ENGINE_MPEG) |
+					  (1 << NVDEV_ENGINE_ME) |
+					  (1 << NVDEV_ENGINE_VP) |
+					  (1 << NVDEV_ENGINE_CRYPT) |
+					  (1 << NVDEV_ENGINE_BSP) |
+					  (1 << NVDEV_ENGINE_PPP) |
+					  (1 << NVDEV_ENGINE_COPY0) |
+					  (1 << NVDEV_ENGINE_UNK1C1), &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	ret = nouveau_ramht_new(parent, parent, 0x8000, 16, &chan->ramht);
+	if (ret)
+		return ret;
+
+	nv_parent(chan)->context_attach = nv84_fifo_context_attach;
+	nv_parent(chan)->context_detach = nv84_fifo_context_detach;
+	nv_parent(chan)->object_attach = nv84_fifo_object_attach;
+	nv_parent(chan)->object_detach = nv50_fifo_object_detach;
+
+	ioffset = args->ioffset;
+	ilength = log2i(args->ilength / 8);
+
+	nv_wo32(base->ramfc, 0x3c, 0x403f6078);
+	nv_wo32(base->ramfc, 0x44, 0x01003fff);
+	nv_wo32(base->ramfc, 0x48, chan->base.pushgpu->node->offset >> 4);
+	nv_wo32(base->ramfc, 0x50, lower_32_bits(ioffset));
+	nv_wo32(base->ramfc, 0x54, upper_32_bits(ioffset) | (ilength << 16));
+	nv_wo32(base->ramfc, 0x60, 0x7fffffff);
+	nv_wo32(base->ramfc, 0x78, 0x00000000);
+	nv_wo32(base->ramfc, 0x7c, 0x30000001);
+	nv_wo32(base->ramfc, 0x80, ((chan->ramht->bits - 9) << 27) |
+				   (4 << 24) /* SEARCH_FULL */ |
+				   (chan->ramht->base.node->offset >> 4));
+	nv_wo32(base->ramfc, 0x88, base->cache->addr >> 10);
+	nv_wo32(base->ramfc, 0x98, nv_gpuobj(base)->addr >> 12);
+	bar->flush(bar);
+	return 0;
+}
+
+static int
+nv84_fifo_chan_init(struct nouveau_object *object)
+{
+	struct nv50_fifo_priv *priv = (void *)object->engine;
+	struct nv50_fifo_base *base = (void *)object->parent;
+	struct nv50_fifo_chan *chan = (void *)object;
+	struct nouveau_gpuobj *ramfc = base->ramfc;
+	u32 chid = chan->base.chid;
+	int ret;
+
+	ret = nouveau_fifo_channel_init(&chan->base);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, 0x002600 + (chid * 4), 0x80000000 | ramfc->addr >> 8);
+	nv50_fifo_playlist_update(priv);
+	return 0;
+}
+
+static struct nouveau_ofuncs
+nv84_fifo_ofuncs_dma = {
+	.ctor = nv84_fifo_chan_ctor_dma,
+	.dtor = nv50_fifo_chan_dtor,
+	.init = nv84_fifo_chan_init,
+	.fini = nv50_fifo_chan_fini,
+	.rd32 = _nouveau_fifo_channel_rd32,
+	.wr32 = _nouveau_fifo_channel_wr32,
+};
+
+static struct nouveau_ofuncs
+nv84_fifo_ofuncs_ind = {
+	.ctor = nv84_fifo_chan_ctor_ind,
+	.dtor = nv50_fifo_chan_dtor,
+	.init = nv84_fifo_chan_init,
+	.fini = nv50_fifo_chan_fini,
+	.rd32 = _nouveau_fifo_channel_rd32,
+	.wr32 = _nouveau_fifo_channel_wr32,
+};
+
+static struct nouveau_oclass
+nv84_fifo_sclass[] = {
+	{ NV84_CHANNEL_DMA_CLASS, &nv84_fifo_ofuncs_dma },
+	{ NV84_CHANNEL_IND_CLASS, &nv84_fifo_ofuncs_ind },
+	{}
+};
+
+/*******************************************************************************
+ * FIFO context - basically just the instmem reserved for the channel
+ ******************************************************************************/
+
+static int
+nv84_fifo_context_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nv50_fifo_base *base;
+	int ret;
+
+	ret = nouveau_fifo_context_create(parent, engine, oclass, NULL, 0x10000,
+				          0x1000, NVOBJ_FLAG_HEAP, &base);
+	*pobject = nv_object(base);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, nv_object(base), 0x0200, 0,
+				 NVOBJ_FLAG_ZERO_ALLOC, &base->eng);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, nv_object(base), 0x4000, 0,
+				 0, &base->pgd);
+	if (ret)
+		return ret;
+
+	ret = nouveau_vm_ref(nouveau_client(parent)->vm, &base->vm, base->pgd);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, nv_object(base), 0x1000, 0x400,
+				 NVOBJ_FLAG_ZERO_ALLOC, &base->cache);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, nv_object(base), 0x0100, 0x100,
+				 NVOBJ_FLAG_ZERO_ALLOC, &base->ramfc);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static struct nouveau_oclass
+nv84_fifo_cclass = {
+	.handle = NV_ENGCTX(FIFO, 0x84),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv84_fifo_context_ctor,
+		.dtor = nv50_fifo_context_dtor,
+		.init = _nouveau_fifo_context_init,
+		.fini = _nouveau_fifo_context_fini,
+		.rd32 = _nouveau_fifo_context_rd32,
+		.wr32 = _nouveau_fifo_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PFIFO engine
+ ******************************************************************************/
+
+static int
+nv84_fifo_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv50_fifo_priv *priv;
+	int ret;
+
+	ret = nouveau_fifo_create(parent, engine, oclass, 1, 127, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 128 * 4, 0x1000, 0,
+				&priv->playlist[0]);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 128 * 4, 0x1000, 0,
+				&priv->playlist[1]);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00000100;
+	nv_subdev(priv)->intr = nv04_fifo_intr;
+	nv_engine(priv)->cclass = &nv84_fifo_cclass;
+	nv_engine(priv)->sclass = nv84_fifo_sclass;
+	return 0;
+}
+
+struct nouveau_oclass
+nv84_fifo_oclass = {
+	.handle = NV_ENGINE(FIFO, 0x84),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv84_fifo_ctor,
+		.dtor = nv50_fifo_dtor,
+		.init = nv50_fifo_init,
+		.fini = _nouveau_fifo_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/nvc0.c b/drivers/gpu/drm/nouveau/core/engine/fifo/nvc0.c
new file mode 100644
index 0000000..6f21be6
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/nvc0.c
@@ -0,0 +1,647 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/client.h>
+#include <core/handle.h>
+#include <core/namedb.h>
+#include <core/gpuobj.h>
+#include <core/engctx.h>
+#include <core/class.h>
+#include <core/math.h>
+#include <core/enum.h>
+
+#include <subdev/timer.h>
+#include <subdev/bar.h>
+#include <subdev/vm.h>
+
+#include <engine/dmaobj.h>
+#include <engine/fifo.h>
+
+struct nvc0_fifo_priv {
+	struct nouveau_fifo base;
+	struct nouveau_gpuobj *playlist[2];
+	int cur_playlist;
+	struct {
+		struct nouveau_gpuobj *mem;
+		struct nouveau_vma bar;
+	} user;
+	int spoon_nr;
+};
+
+struct nvc0_fifo_base {
+	struct nouveau_fifo_base base;
+	struct nouveau_gpuobj *pgd;
+	struct nouveau_vm *vm;
+};
+
+struct nvc0_fifo_chan {
+	struct nouveau_fifo_chan base;
+};
+
+/*******************************************************************************
+ * FIFO channel objects
+ ******************************************************************************/
+
+static void
+nvc0_fifo_playlist_update(struct nvc0_fifo_priv *priv)
+{
+	struct nouveau_bar *bar = nouveau_bar(priv);
+	struct nouveau_gpuobj *cur;
+	int i, p;
+
+	cur = priv->playlist[priv->cur_playlist];
+	priv->cur_playlist = !priv->cur_playlist;
+
+	for (i = 0, p = 0; i < 128; i++) {
+		if (!(nv_rd32(priv, 0x003004 + (i * 8)) & 1))
+			continue;
+		nv_wo32(cur, p + 0, i);
+		nv_wo32(cur, p + 4, 0x00000004);
+		p += 8;
+	}
+	bar->flush(bar);
+
+	nv_wr32(priv, 0x002270, cur->addr >> 12);
+	nv_wr32(priv, 0x002274, 0x01f00000 | (p >> 3));
+	if (!nv_wait(priv, 0x00227c, 0x00100000, 0x00000000))
+		nv_error(priv, "playlist update failed\n");
+}
+
+static int
+nvc0_fifo_context_attach(struct nouveau_object *parent,
+			 struct nouveau_object *object)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nvc0_fifo_base *base = (void *)parent->parent;
+	struct nouveau_engctx *ectx = (void *)object;
+	u32 addr;
+	int ret;
+
+	switch (nv_engidx(object->engine)) {
+	case NVDEV_ENGINE_SW   : return 0;
+	case NVDEV_ENGINE_GR   : addr = 0x0210; break;
+	case NVDEV_ENGINE_COPY0: addr = 0x0230; break;
+	case NVDEV_ENGINE_COPY1: addr = 0x0240; break;
+	default:
+		return -EINVAL;
+	}
+
+	if (!ectx->vma.node) {
+		ret = nouveau_gpuobj_map_vm(nv_gpuobj(ectx), base->vm,
+					    NV_MEM_ACCESS_RW, &ectx->vma);
+		if (ret)
+			return ret;
+
+		nv_engctx(ectx)->addr = nv_gpuobj(base)->addr >> 12;
+	}
+
+	nv_wo32(base, addr + 0x00, lower_32_bits(ectx->vma.offset) | 4);
+	nv_wo32(base, addr + 0x04, upper_32_bits(ectx->vma.offset));
+	bar->flush(bar);
+	return 0;
+}
+
+static int
+nvc0_fifo_context_detach(struct nouveau_object *parent, bool suspend,
+			 struct nouveau_object *object)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nvc0_fifo_priv *priv = (void *)parent->engine;
+	struct nvc0_fifo_base *base = (void *)parent->parent;
+	struct nvc0_fifo_chan *chan = (void *)parent;
+	u32 addr;
+
+	switch (nv_engidx(object->engine)) {
+	case NVDEV_ENGINE_SW   : return 0;
+	case NVDEV_ENGINE_GR   : addr = 0x0210; break;
+	case NVDEV_ENGINE_COPY0: addr = 0x0230; break;
+	case NVDEV_ENGINE_COPY1: addr = 0x0240; break;
+	default:
+		return -EINVAL;
+	}
+
+	nv_wo32(base, addr + 0x00, 0x00000000);
+	nv_wo32(base, addr + 0x04, 0x00000000);
+	bar->flush(bar);
+
+	nv_wr32(priv, 0x002634, chan->base.chid);
+	if (!nv_wait(priv, 0x002634, 0xffffffff, chan->base.chid)) {
+		nv_error(priv, "channel %d kick timeout\n", chan->base.chid);
+		if (suspend)
+			return -EBUSY;
+	}
+
+	return 0;
+}
+
+static int
+nvc0_fifo_chan_ctor(struct nouveau_object *parent,
+		    struct nouveau_object *engine,
+		    struct nouveau_oclass *oclass, void *data, u32 size,
+		    struct nouveau_object **pobject)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nvc0_fifo_priv *priv = (void *)engine;
+	struct nvc0_fifo_base *base = (void *)parent;
+	struct nvc0_fifo_chan *chan;
+	struct nv50_channel_ind_class *args = data;
+	u64 usermem, ioffset, ilength;
+	int ret, i;
+
+	if (size < sizeof(*args))
+		return -EINVAL;
+
+	ret = nouveau_fifo_channel_create(parent, engine, oclass, 1,
+					  priv->user.bar.offset, 0x1000,
+					  args->pushbuf,
+					  (1 << NVDEV_ENGINE_SW) |
+					  (1 << NVDEV_ENGINE_GR) |
+					  (1 << NVDEV_ENGINE_COPY0) |
+					  (1 << NVDEV_ENGINE_COPY1), &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	nv_parent(chan)->context_attach = nvc0_fifo_context_attach;
+	nv_parent(chan)->context_detach = nvc0_fifo_context_detach;
+
+	usermem = chan->base.chid * 0x1000;
+	ioffset = args->ioffset;
+	ilength = log2i(args->ilength / 8);
+
+	for (i = 0; i < 0x1000; i += 4)
+		nv_wo32(priv->user.mem, usermem + i, 0x00000000);
+
+	nv_wo32(base, 0x08, lower_32_bits(priv->user.mem->addr + usermem));
+	nv_wo32(base, 0x0c, upper_32_bits(priv->user.mem->addr + usermem));
+	nv_wo32(base, 0x10, 0x0000face);
+	nv_wo32(base, 0x30, 0xfffff902);
+	nv_wo32(base, 0x48, lower_32_bits(ioffset));
+	nv_wo32(base, 0x4c, upper_32_bits(ioffset) | (ilength << 16));
+	nv_wo32(base, 0x54, 0x00000002);
+	nv_wo32(base, 0x84, 0x20400000);
+	nv_wo32(base, 0x94, 0x30000001);
+	nv_wo32(base, 0x9c, 0x00000100);
+	nv_wo32(base, 0xa4, 0x1f1f1f1f);
+	nv_wo32(base, 0xa8, 0x1f1f1f1f);
+	nv_wo32(base, 0xac, 0x0000001f);
+	nv_wo32(base, 0xb8, 0xf8000000);
+	nv_wo32(base, 0xf8, 0x10003080); /* 0x002310 */
+	nv_wo32(base, 0xfc, 0x10000010); /* 0x002350 */
+	bar->flush(bar);
+	return 0;
+}
+
+static int
+nvc0_fifo_chan_init(struct nouveau_object *object)
+{
+	struct nouveau_gpuobj *base = nv_gpuobj(object->parent);
+	struct nvc0_fifo_priv *priv = (void *)object->engine;
+	struct nvc0_fifo_chan *chan = (void *)object;
+	u32 chid = chan->base.chid;
+	int ret;
+
+	ret = nouveau_fifo_channel_init(&chan->base);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, 0x003000 + (chid * 8), 0xc0000000 | base->addr >> 12);
+	nv_wr32(priv, 0x003004 + (chid * 8), 0x001f0001);
+	nvc0_fifo_playlist_update(priv);
+	return 0;
+}
+
+static int
+nvc0_fifo_chan_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nvc0_fifo_priv *priv = (void *)object->engine;
+	struct nvc0_fifo_chan *chan = (void *)object;
+	u32 chid = chan->base.chid;
+
+	nv_mask(priv, 0x003004 + (chid * 8), 0x00000001, 0x00000000);
+	nvc0_fifo_playlist_update(priv);
+	nv_wr32(priv, 0x003000 + (chid * 8), 0x00000000);
+
+	return nouveau_fifo_channel_fini(&chan->base, suspend);
+}
+
+static struct nouveau_ofuncs
+nvc0_fifo_ofuncs = {
+	.ctor = nvc0_fifo_chan_ctor,
+	.dtor = _nouveau_fifo_channel_dtor,
+	.init = nvc0_fifo_chan_init,
+	.fini = nvc0_fifo_chan_fini,
+	.rd32 = _nouveau_fifo_channel_rd32,
+	.wr32 = _nouveau_fifo_channel_wr32,
+};
+
+static struct nouveau_oclass
+nvc0_fifo_sclass[] = {
+	{ NVC0_CHANNEL_IND_CLASS, &nvc0_fifo_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * FIFO context - instmem heap and vm setup
+ ******************************************************************************/
+
+static int
+nvc0_fifo_context_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nvc0_fifo_base *base;
+	int ret;
+
+	ret = nouveau_fifo_context_create(parent, engine, oclass, NULL, 0x1000,
+				          0x1000, NVOBJ_FLAG_ZERO_ALLOC |
+					  NVOBJ_FLAG_HEAP, &base);
+	*pobject = nv_object(base);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 0x10000, 0x1000, 0, &base->pgd);
+	if (ret)
+		return ret;
+
+	nv_wo32(base, 0x0200, lower_32_bits(base->pgd->addr));
+	nv_wo32(base, 0x0204, upper_32_bits(base->pgd->addr));
+	nv_wo32(base, 0x0208, 0xffffffff);
+	nv_wo32(base, 0x020c, 0x000000ff);
+
+	ret = nouveau_vm_ref(nouveau_client(parent)->vm, &base->vm, base->pgd);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static void
+nvc0_fifo_context_dtor(struct nouveau_object *object)
+{
+	struct nvc0_fifo_base *base = (void *)object;
+	nouveau_vm_ref(NULL, &base->vm, base->pgd);
+	nouveau_gpuobj_ref(NULL, &base->pgd);
+	nouveau_fifo_context_destroy(&base->base);
+}
+
+static struct nouveau_oclass
+nvc0_fifo_cclass = {
+	.handle = NV_ENGCTX(FIFO, 0xc0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvc0_fifo_context_ctor,
+		.dtor = nvc0_fifo_context_dtor,
+		.init = _nouveau_fifo_context_init,
+		.fini = _nouveau_fifo_context_fini,
+		.rd32 = _nouveau_fifo_context_rd32,
+		.wr32 = _nouveau_fifo_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PFIFO engine
+ ******************************************************************************/
+
+static const struct nouveau_enum nvc0_fifo_fault_unit[] = {
+	{ 0x00, "PGRAPH" },
+	{ 0x03, "PEEPHOLE" },
+	{ 0x04, "BAR1" },
+	{ 0x05, "BAR3" },
+	{ 0x07, "PFIFO" },
+	{ 0x10, "PBSP" },
+	{ 0x11, "PPPP" },
+	{ 0x13, "PCOUNTER" },
+	{ 0x14, "PVP" },
+	{ 0x15, "PCOPY0" },
+	{ 0x16, "PCOPY1" },
+	{ 0x17, "PDAEMON" },
+	{}
+};
+
+static const struct nouveau_enum nvc0_fifo_fault_reason[] = {
+	{ 0x00, "PT_NOT_PRESENT" },
+	{ 0x01, "PT_TOO_SHORT" },
+	{ 0x02, "PAGE_NOT_PRESENT" },
+	{ 0x03, "VM_LIMIT_EXCEEDED" },
+	{ 0x04, "NO_CHANNEL" },
+	{ 0x05, "PAGE_SYSTEM_ONLY" },
+	{ 0x06, "PAGE_READ_ONLY" },
+	{ 0x0a, "COMPRESSED_SYSRAM" },
+	{ 0x0c, "INVALID_STORAGE_TYPE" },
+	{}
+};
+
+static const struct nouveau_enum nvc0_fifo_fault_hubclient[] = {
+	{ 0x01, "PCOPY0" },
+	{ 0x02, "PCOPY1" },
+	{ 0x04, "DISPATCH" },
+	{ 0x05, "CTXCTL" },
+	{ 0x06, "PFIFO" },
+	{ 0x07, "BAR_READ" },
+	{ 0x08, "BAR_WRITE" },
+	{ 0x0b, "PVP" },
+	{ 0x0c, "PPPP" },
+	{ 0x0d, "PBSP" },
+	{ 0x11, "PCOUNTER" },
+	{ 0x12, "PDAEMON" },
+	{ 0x14, "CCACHE" },
+	{ 0x15, "CCACHE_POST" },
+	{}
+};
+
+static const struct nouveau_enum nvc0_fifo_fault_gpcclient[] = {
+	{ 0x01, "TEX" },
+	{ 0x0c, "ESETUP" },
+	{ 0x0e, "CTXCTL" },
+	{ 0x0f, "PROP" },
+	{}
+};
+
+static const struct nouveau_bitfield nvc0_fifo_subfifo_intr[] = {
+/*	{ 0x00008000, "" }	seen with null ib push */
+	{ 0x00200000, "ILLEGAL_MTHD" },
+	{ 0x00800000, "EMPTY_SUBC" },
+	{}
+};
+
+static void
+nvc0_fifo_isr_vm_fault(struct nvc0_fifo_priv *priv, int unit)
+{
+	u32 inst = nv_rd32(priv, 0x002800 + (unit * 0x10));
+	u32 valo = nv_rd32(priv, 0x002804 + (unit * 0x10));
+	u32 vahi = nv_rd32(priv, 0x002808 + (unit * 0x10));
+	u32 stat = nv_rd32(priv, 0x00280c + (unit * 0x10));
+	u32 client = (stat & 0x00001f00) >> 8;
+
+	switch (unit) {
+	case 3: /* PEEPHOLE */
+		nv_mask(priv, 0x001718, 0x00000000, 0x00000000);
+		break;
+	case 4: /* BAR1 */
+		nv_mask(priv, 0x001704, 0x00000000, 0x00000000);
+		break;
+	case 5: /* BAR3 */
+		nv_mask(priv, 0x001714, 0x00000000, 0x00000000);
+		break;
+	default:
+		break;
+	}
+
+	nv_error(priv, "%s fault at 0x%010llx [", (stat & 0x00000080) ?
+		 "write" : "read", (u64)vahi << 32 | valo);
+	nouveau_enum_print(nvc0_fifo_fault_reason, stat & 0x0000000f);
+	printk("] from ");
+	nouveau_enum_print(nvc0_fifo_fault_unit, unit);
+	if (stat & 0x00000040) {
+		printk("/");
+		nouveau_enum_print(nvc0_fifo_fault_hubclient, client);
+	} else {
+		printk("/GPC%d/", (stat & 0x1f000000) >> 24);
+		nouveau_enum_print(nvc0_fifo_fault_gpcclient, client);
+	}
+	printk(" on channel 0x%010llx\n", (u64)inst << 12);
+}
+
+static int
+nvc0_fifo_swmthd(struct nvc0_fifo_priv *priv, u32 chid, u32 mthd, u32 data)
+{
+	struct nvc0_fifo_chan *chan = NULL;
+	struct nouveau_handle *bind;
+	unsigned long flags;
+	int ret = -EINVAL;
+
+	spin_lock_irqsave(&priv->base.lock, flags);
+	if (likely(chid >= priv->base.min && chid <= priv->base.max))
+		chan = (void *)priv->base.channel[chid];
+	if (unlikely(!chan))
+		goto out;
+
+	bind = nouveau_namedb_get_class(nv_namedb(chan), 0x906e);
+	if (likely(bind)) {
+		if (!mthd || !nv_call(bind->object, mthd, data))
+			ret = 0;
+		nouveau_namedb_put(bind);
+	}
+
+out:
+	spin_unlock_irqrestore(&priv->base.lock, flags);
+	return ret;
+}
+
+static void
+nvc0_fifo_isr_subfifo_intr(struct nvc0_fifo_priv *priv, int unit)
+{
+	u32 stat = nv_rd32(priv, 0x040108 + (unit * 0x2000));
+	u32 addr = nv_rd32(priv, 0x0400c0 + (unit * 0x2000));
+	u32 data = nv_rd32(priv, 0x0400c4 + (unit * 0x2000));
+	u32 chid = nv_rd32(priv, 0x040120 + (unit * 0x2000)) & 0x7f;
+	u32 subc = (addr & 0x00070000) >> 16;
+	u32 mthd = (addr & 0x00003ffc);
+	u32 show = stat;
+
+	if (stat & 0x00200000) {
+		if (mthd == 0x0054) {
+			if (!nvc0_fifo_swmthd(priv, chid, 0x0500, 0x00000000))
+				show &= ~0x00200000;
+		}
+	}
+
+	if (stat & 0x00800000) {
+		if (!nvc0_fifo_swmthd(priv, chid, mthd, data))
+			show &= ~0x00800000;
+	}
+
+	if (show) {
+		nv_error(priv, "SUBFIFO%d:", unit);
+		nouveau_bitfield_print(nvc0_fifo_subfifo_intr, show);
+		printk("\n");
+		nv_error(priv, "SUBFIFO%d: ch %d subc %d mthd 0x%04x "
+			       "data 0x%08x\n",
+			 unit, chid, subc, mthd, data);
+	}
+
+	nv_wr32(priv, 0x0400c0 + (unit * 0x2000), 0x80600008);
+	nv_wr32(priv, 0x040108 + (unit * 0x2000), stat);
+}
+
+static void
+nvc0_fifo_intr(struct nouveau_subdev *subdev)
+{
+	struct nvc0_fifo_priv *priv = (void *)subdev;
+	u32 mask = nv_rd32(priv, 0x002140);
+	u32 stat = nv_rd32(priv, 0x002100) & mask;
+
+	if (stat & 0x00000100) {
+		nv_info(priv, "unknown status 0x00000100\n");
+		nv_wr32(priv, 0x002100, 0x00000100);
+		stat &= ~0x00000100;
+	}
+
+	if (stat & 0x10000000) {
+		u32 units = nv_rd32(priv, 0x00259c);
+		u32 u = units;
+
+		while (u) {
+			int i = ffs(u) - 1;
+			nvc0_fifo_isr_vm_fault(priv, i);
+			u &= ~(1 << i);
+		}
+
+		nv_wr32(priv, 0x00259c, units);
+		stat &= ~0x10000000;
+	}
+
+	if (stat & 0x20000000) {
+		u32 units = nv_rd32(priv, 0x0025a0);
+		u32 u = units;
+
+		while (u) {
+			int i = ffs(u) - 1;
+			nvc0_fifo_isr_subfifo_intr(priv, i);
+			u &= ~(1 << i);
+		}
+
+		nv_wr32(priv, 0x0025a0, units);
+		stat &= ~0x20000000;
+	}
+
+	if (stat & 0x40000000) {
+		nv_warn(priv, "unknown status 0x40000000\n");
+		nv_mask(priv, 0x002a00, 0x00000000, 0x00000000);
+		stat &= ~0x40000000;
+	}
+
+	if (stat) {
+		nv_fatal(priv, "unhandled status 0x%08x\n", stat);
+		nv_wr32(priv, 0x002100, stat);
+		nv_wr32(priv, 0x002140, 0);
+	}
+}
+
+static int
+nvc0_fifo_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nvc0_fifo_priv *priv;
+	int ret;
+
+	ret = nouveau_fifo_create(parent, engine, oclass, 0, 127, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 0x1000, 0x1000, 0,
+				&priv->playlist[0]);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 0x1000, 0x1000, 0,
+				&priv->playlist[1]);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 128 * 0x1000, 0x1000, 0,
+				&priv->user.mem);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_map(priv->user.mem, NV_MEM_ACCESS_RW,
+				&priv->user.bar);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00000100;
+	nv_subdev(priv)->intr = nvc0_fifo_intr;
+	nv_engine(priv)->cclass = &nvc0_fifo_cclass;
+	nv_engine(priv)->sclass = nvc0_fifo_sclass;
+	return 0;
+}
+
+static void
+nvc0_fifo_dtor(struct nouveau_object *object)
+{
+	struct nvc0_fifo_priv *priv = (void *)object;
+
+	nouveau_gpuobj_unmap(&priv->user.bar);
+	nouveau_gpuobj_ref(NULL, &priv->user.mem);
+	nouveau_gpuobj_ref(NULL, &priv->playlist[1]);
+	nouveau_gpuobj_ref(NULL, &priv->playlist[0]);
+
+	nouveau_fifo_destroy(&priv->base);
+}
+
+static int
+nvc0_fifo_init(struct nouveau_object *object)
+{
+	struct nvc0_fifo_priv *priv = (void *)object;
+	int ret, i;
+
+	ret = nouveau_fifo_init(&priv->base);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, 0x000204, 0xffffffff);
+	nv_wr32(priv, 0x002204, 0xffffffff);
+
+	priv->spoon_nr = hweight32(nv_rd32(priv, 0x002204));
+	nv_debug(priv, "%d subfifo(s)\n", priv->spoon_nr);
+
+	/* assign engines to subfifos */
+	if (priv->spoon_nr >= 3) {
+		nv_wr32(priv, 0x002208, ~(1 << 0)); /* PGRAPH */
+		nv_wr32(priv, 0x00220c, ~(1 << 1)); /* PVP */
+		nv_wr32(priv, 0x002210, ~(1 << 1)); /* PPP */
+		nv_wr32(priv, 0x002214, ~(1 << 1)); /* PBSP */
+		nv_wr32(priv, 0x002218, ~(1 << 2)); /* PCE0 */
+		nv_wr32(priv, 0x00221c, ~(1 << 1)); /* PCE1 */
+	}
+
+	/* PSUBFIFO[n] */
+	for (i = 0; i < priv->spoon_nr; i++) {
+		nv_mask(priv, 0x04013c + (i * 0x2000), 0x10000100, 0x00000000);
+		nv_wr32(priv, 0x040108 + (i * 0x2000), 0xffffffff); /* INTR */
+		nv_wr32(priv, 0x04010c + (i * 0x2000), 0xfffffeff); /* INTREN */
+	}
+
+	nv_mask(priv, 0x002200, 0x00000001, 0x00000001);
+	nv_wr32(priv, 0x002254, 0x10000000 | priv->user.bar.offset >> 12);
+
+	nv_wr32(priv, 0x002a00, 0xffffffff); /* clears PFIFO.INTR bit 30 */
+	nv_wr32(priv, 0x002100, 0xffffffff);
+	nv_wr32(priv, 0x002140, 0xbfffffff);
+	return 0;
+}
+
+struct nouveau_oclass
+nvc0_fifo_oclass = {
+	.handle = NV_ENGINE(FIFO, 0xc0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvc0_fifo_ctor,
+		.dtor = nvc0_fifo_dtor,
+		.init = nvc0_fifo_init,
+		.fini = _nouveau_fifo_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/nve0.c b/drivers/gpu/drm/nouveau/core/engine/fifo/nve0.c
new file mode 100644
index 0000000..36e81b6
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/nve0.c
@@ -0,0 +1,628 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/client.h>
+#include <core/handle.h>
+#include <core/namedb.h>
+#include <core/gpuobj.h>
+#include <core/engctx.h>
+#include <core/class.h>
+#include <core/math.h>
+#include <core/enum.h>
+
+#include <subdev/timer.h>
+#include <subdev/bar.h>
+#include <subdev/vm.h>
+
+#include <engine/dmaobj.h>
+#include <engine/fifo.h>
+
+#define _(a,b) { (a), ((1 << (a)) | (b)) }
+static const struct {
+	int subdev;
+	u32 mask;
+} fifo_engine[] = {
+	_(NVDEV_ENGINE_GR      , (1 << NVDEV_ENGINE_SW)),
+	_(NVDEV_ENGINE_VP      , 0),
+	_(NVDEV_ENGINE_PPP     , 0),
+	_(NVDEV_ENGINE_BSP     , 0),
+	_(NVDEV_ENGINE_COPY0   , 0),
+	_(NVDEV_ENGINE_COPY1   , 0),
+	_(NVDEV_ENGINE_VENC    , 0),
+};
+#undef _
+#define FIFO_ENGINE_NR ARRAY_SIZE(fifo_engine)
+
+struct nve0_fifo_engn {
+	struct nouveau_gpuobj *playlist[2];
+	int cur_playlist;
+};
+
+struct nve0_fifo_priv {
+	struct nouveau_fifo base;
+	struct nve0_fifo_engn engine[FIFO_ENGINE_NR];
+	struct {
+		struct nouveau_gpuobj *mem;
+		struct nouveau_vma bar;
+	} user;
+	int spoon_nr;
+};
+
+struct nve0_fifo_base {
+	struct nouveau_fifo_base base;
+	struct nouveau_gpuobj *pgd;
+	struct nouveau_vm *vm;
+};
+
+struct nve0_fifo_chan {
+	struct nouveau_fifo_chan base;
+	u32 engine;
+};
+
+/*******************************************************************************
+ * FIFO channel objects
+ ******************************************************************************/
+
+static void
+nve0_fifo_playlist_update(struct nve0_fifo_priv *priv, u32 engine)
+{
+	struct nouveau_bar *bar = nouveau_bar(priv);
+	struct nve0_fifo_engn *engn = &priv->engine[engine];
+	struct nouveau_gpuobj *cur;
+	u32 match = (engine << 16) | 0x00000001;
+	int i, p;
+
+	cur = engn->playlist[engn->cur_playlist];
+	if (unlikely(cur == NULL)) {
+		int ret = nouveau_gpuobj_new(nv_object(priv)->parent, NULL,
+					     0x8000, 0x1000, 0, &cur);
+		if (ret) {
+			nv_error(priv, "playlist alloc failed\n");
+			return;
+		}
+
+		engn->playlist[engn->cur_playlist] = cur;
+	}
+
+	engn->cur_playlist = !engn->cur_playlist;
+
+	for (i = 0, p = 0; i < priv->base.max; i++) {
+		u32 ctrl = nv_rd32(priv, 0x800004 + (i * 8)) & 0x001f0001;
+		if (ctrl != match)
+			continue;
+		nv_wo32(cur, p + 0, i);
+		nv_wo32(cur, p + 4, 0x00000000);
+		p += 8;
+	}
+	bar->flush(bar);
+
+	nv_wr32(priv, 0x002270, cur->addr >> 12);
+	nv_wr32(priv, 0x002274, (engine << 20) | (p >> 3));
+	if (!nv_wait(priv, 0x002284 + (engine * 4), 0x00100000, 0x00000000))
+		nv_error(priv, "playlist %d update timeout\n", engine);
+}
+
+static int
+nve0_fifo_context_attach(struct nouveau_object *parent,
+			 struct nouveau_object *object)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nve0_fifo_base *base = (void *)parent->parent;
+	struct nouveau_engctx *ectx = (void *)object;
+	u32 addr;
+	int ret;
+
+	switch (nv_engidx(object->engine)) {
+	case NVDEV_ENGINE_SW   : return 0;
+	case NVDEV_ENGINE_GR   :
+	case NVDEV_ENGINE_COPY0:
+	case NVDEV_ENGINE_COPY1: addr = 0x0210; break;
+	default:
+		return -EINVAL;
+	}
+
+	if (!ectx->vma.node) {
+		ret = nouveau_gpuobj_map_vm(nv_gpuobj(ectx), base->vm,
+					    NV_MEM_ACCESS_RW, &ectx->vma);
+		if (ret)
+			return ret;
+
+		nv_engctx(ectx)->addr = nv_gpuobj(base)->addr >> 12;
+	}
+
+	nv_wo32(base, addr + 0x00, lower_32_bits(ectx->vma.offset) | 4);
+	nv_wo32(base, addr + 0x04, upper_32_bits(ectx->vma.offset));
+	bar->flush(bar);
+	return 0;
+}
+
+static int
+nve0_fifo_context_detach(struct nouveau_object *parent, bool suspend,
+			 struct nouveau_object *object)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nve0_fifo_priv *priv = (void *)parent->engine;
+	struct nve0_fifo_base *base = (void *)parent->parent;
+	struct nve0_fifo_chan *chan = (void *)parent;
+	u32 addr;
+
+	switch (nv_engidx(object->engine)) {
+	case NVDEV_ENGINE_SW   : return 0;
+	case NVDEV_ENGINE_GR   :
+	case NVDEV_ENGINE_COPY0:
+	case NVDEV_ENGINE_COPY1: addr = 0x0210; break;
+	default:
+		return -EINVAL;
+	}
+
+	nv_wo32(base, addr + 0x00, 0x00000000);
+	nv_wo32(base, addr + 0x04, 0x00000000);
+	bar->flush(bar);
+
+	nv_wr32(priv, 0x002634, chan->base.chid);
+	if (!nv_wait(priv, 0x002634, 0xffffffff, chan->base.chid)) {
+		nv_error(priv, "channel %d kick timeout\n", chan->base.chid);
+		if (suspend)
+			return -EBUSY;
+	}
+
+	return 0;
+}
+
+static int
+nve0_fifo_chan_ctor(struct nouveau_object *parent,
+		    struct nouveau_object *engine,
+		    struct nouveau_oclass *oclass, void *data, u32 size,
+		    struct nouveau_object **pobject)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nve0_fifo_priv *priv = (void *)engine;
+	struct nve0_fifo_base *base = (void *)parent;
+	struct nve0_fifo_chan *chan;
+	struct nve0_channel_ind_class *args = data;
+	u64 usermem, ioffset, ilength;
+	int ret, i;
+
+	if (size < sizeof(*args))
+		return -EINVAL;
+
+	for (i = 0; i < FIFO_ENGINE_NR; i++) {
+		if (args->engine & (1 << i)) {
+			if (nouveau_engine(parent, fifo_engine[i].subdev)) {
+				args->engine = (1 << i);
+				break;
+			}
+		}
+	}
+
+	if (i == FIFO_ENGINE_NR)
+		return -ENODEV;
+
+	ret = nouveau_fifo_channel_create(parent, engine, oclass, 1,
+					  priv->user.bar.offset, 0x200,
+					  args->pushbuf,
+					  fifo_engine[i].mask, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	nv_parent(chan)->context_attach = nve0_fifo_context_attach;
+	nv_parent(chan)->context_detach = nve0_fifo_context_detach;
+	chan->engine = i;
+
+	usermem = chan->base.chid * 0x200;
+	ioffset = args->ioffset;
+	ilength = log2i(args->ilength / 8);
+
+	for (i = 0; i < 0x200; i += 4)
+		nv_wo32(priv->user.mem, usermem + i, 0x00000000);
+
+	nv_wo32(base, 0x08, lower_32_bits(priv->user.mem->addr + usermem));
+	nv_wo32(base, 0x0c, upper_32_bits(priv->user.mem->addr + usermem));
+	nv_wo32(base, 0x10, 0x0000face);
+	nv_wo32(base, 0x30, 0xfffff902);
+	nv_wo32(base, 0x48, lower_32_bits(ioffset));
+	nv_wo32(base, 0x4c, upper_32_bits(ioffset) | (ilength << 16));
+	nv_wo32(base, 0x84, 0x20400000);
+	nv_wo32(base, 0x94, 0x30000001);
+	nv_wo32(base, 0x9c, 0x00000100);
+	nv_wo32(base, 0xac, 0x0000001f);
+	nv_wo32(base, 0xe8, chan->base.chid);
+	nv_wo32(base, 0xb8, 0xf8000000);
+	nv_wo32(base, 0xf8, 0x10003080); /* 0x002310 */
+	nv_wo32(base, 0xfc, 0x10000010); /* 0x002350 */
+	bar->flush(bar);
+	return 0;
+}
+
+static int
+nve0_fifo_chan_init(struct nouveau_object *object)
+{
+	struct nouveau_gpuobj *base = nv_gpuobj(object->parent);
+	struct nve0_fifo_priv *priv = (void *)object->engine;
+	struct nve0_fifo_chan *chan = (void *)object;
+	u32 chid = chan->base.chid;
+	int ret;
+
+	ret = nouveau_fifo_channel_init(&chan->base);
+	if (ret)
+		return ret;
+
+	nv_mask(priv, 0x800004 + (chid * 8), 0x000f0000, chan->engine << 16);
+	nv_wr32(priv, 0x800000 + (chid * 8), 0x80000000 | base->addr >> 12);
+	nv_mask(priv, 0x800004 + (chid * 8), 0x00000400, 0x00000400);
+	nve0_fifo_playlist_update(priv, chan->engine);
+	nv_mask(priv, 0x800004 + (chid * 8), 0x00000400, 0x00000400);
+	return 0;
+}
+
+static int
+nve0_fifo_chan_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nve0_fifo_priv *priv = (void *)object->engine;
+	struct nve0_fifo_chan *chan = (void *)object;
+	u32 chid = chan->base.chid;
+
+	nv_mask(priv, 0x800004 + (chid * 8), 0x00000800, 0x00000800);
+	nve0_fifo_playlist_update(priv, chan->engine);
+	nv_wr32(priv, 0x800000 + (chid * 8), 0x00000000);
+
+	return nouveau_fifo_channel_fini(&chan->base, suspend);
+}
+
+static struct nouveau_ofuncs
+nve0_fifo_ofuncs = {
+	.ctor = nve0_fifo_chan_ctor,
+	.dtor = _nouveau_fifo_channel_dtor,
+	.init = nve0_fifo_chan_init,
+	.fini = nve0_fifo_chan_fini,
+	.rd32 = _nouveau_fifo_channel_rd32,
+	.wr32 = _nouveau_fifo_channel_wr32,
+};
+
+static struct nouveau_oclass
+nve0_fifo_sclass[] = {
+	{ NVE0_CHANNEL_IND_CLASS, &nve0_fifo_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * FIFO context - instmem heap and vm setup
+ ******************************************************************************/
+
+static int
+nve0_fifo_context_ctor(struct nouveau_object *parent,
+		    struct nouveau_object *engine,
+		    struct nouveau_oclass *oclass, void *data, u32 size,
+		    struct nouveau_object **pobject)
+{
+	struct nve0_fifo_base *base;
+	int ret;
+
+	ret = nouveau_fifo_context_create(parent, engine, oclass, NULL, 0x1000,
+				          0x1000, NVOBJ_FLAG_ZERO_ALLOC, &base);
+	*pobject = nv_object(base);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 0x10000, 0x1000, 0, &base->pgd);
+	if (ret)
+		return ret;
+
+	nv_wo32(base, 0x0200, lower_32_bits(base->pgd->addr));
+	nv_wo32(base, 0x0204, upper_32_bits(base->pgd->addr));
+	nv_wo32(base, 0x0208, 0xffffffff);
+	nv_wo32(base, 0x020c, 0x000000ff);
+
+	ret = nouveau_vm_ref(nouveau_client(parent)->vm, &base->vm, base->pgd);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static void
+nve0_fifo_context_dtor(struct nouveau_object *object)
+{
+	struct nve0_fifo_base *base = (void *)object;
+	nouveau_vm_ref(NULL, &base->vm, base->pgd);
+	nouveau_gpuobj_ref(NULL, &base->pgd);
+	nouveau_fifo_context_destroy(&base->base);
+}
+
+static struct nouveau_oclass
+nve0_fifo_cclass = {
+	.handle = NV_ENGCTX(FIFO, 0xe0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nve0_fifo_context_ctor,
+		.dtor = nve0_fifo_context_dtor,
+		.init = _nouveau_fifo_context_init,
+		.fini = _nouveau_fifo_context_fini,
+		.rd32 = _nouveau_fifo_context_rd32,
+		.wr32 = _nouveau_fifo_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PFIFO engine
+ ******************************************************************************/
+
+static const struct nouveau_enum nve0_fifo_fault_unit[] = {
+	{}
+};
+
+static const struct nouveau_enum nve0_fifo_fault_reason[] = {
+	{ 0x00, "PT_NOT_PRESENT" },
+	{ 0x01, "PT_TOO_SHORT" },
+	{ 0x02, "PAGE_NOT_PRESENT" },
+	{ 0x03, "VM_LIMIT_EXCEEDED" },
+	{ 0x04, "NO_CHANNEL" },
+	{ 0x05, "PAGE_SYSTEM_ONLY" },
+	{ 0x06, "PAGE_READ_ONLY" },
+	{ 0x0a, "COMPRESSED_SYSRAM" },
+	{ 0x0c, "INVALID_STORAGE_TYPE" },
+	{}
+};
+
+static const struct nouveau_enum nve0_fifo_fault_hubclient[] = {
+	{}
+};
+
+static const struct nouveau_enum nve0_fifo_fault_gpcclient[] = {
+	{}
+};
+
+static const struct nouveau_bitfield nve0_fifo_subfifo_intr[] = {
+	{ 0x00200000, "ILLEGAL_MTHD" },
+	{ 0x00800000, "EMPTY_SUBC" },
+	{}
+};
+
+static void
+nve0_fifo_isr_vm_fault(struct nve0_fifo_priv *priv, int unit)
+{
+	u32 inst = nv_rd32(priv, 0x2800 + (unit * 0x10));
+	u32 valo = nv_rd32(priv, 0x2804 + (unit * 0x10));
+	u32 vahi = nv_rd32(priv, 0x2808 + (unit * 0x10));
+	u32 stat = nv_rd32(priv, 0x280c + (unit * 0x10));
+	u32 client = (stat & 0x00001f00) >> 8;
+
+	nv_error(priv, "PFIFO: %s fault at 0x%010llx [", (stat & 0x00000080) ?
+		       "write" : "read", (u64)vahi << 32 | valo);
+	nouveau_enum_print(nve0_fifo_fault_reason, stat & 0x0000000f);
+	printk("] from ");
+	nouveau_enum_print(nve0_fifo_fault_unit, unit);
+	if (stat & 0x00000040) {
+		printk("/");
+		nouveau_enum_print(nve0_fifo_fault_hubclient, client);
+	} else {
+		printk("/GPC%d/", (stat & 0x1f000000) >> 24);
+		nouveau_enum_print(nve0_fifo_fault_gpcclient, client);
+	}
+	printk(" on channel 0x%010llx\n", (u64)inst << 12);
+}
+
+static int
+nve0_fifo_swmthd(struct nve0_fifo_priv *priv, u32 chid, u32 mthd, u32 data)
+{
+	struct nve0_fifo_chan *chan = NULL;
+	struct nouveau_handle *bind;
+	unsigned long flags;
+	int ret = -EINVAL;
+
+	spin_lock_irqsave(&priv->base.lock, flags);
+	if (likely(chid >= priv->base.min && chid <= priv->base.max))
+		chan = (void *)priv->base.channel[chid];
+	if (unlikely(!chan))
+		goto out;
+
+	bind = nouveau_namedb_get_class(nv_namedb(chan), 0x906e);
+	if (likely(bind)) {
+		if (!mthd || !nv_call(bind->object, mthd, data))
+			ret = 0;
+		nouveau_namedb_put(bind);
+	}
+
+out:
+	spin_unlock_irqrestore(&priv->base.lock, flags);
+	return ret;
+}
+
+static void
+nve0_fifo_isr_subfifo_intr(struct nve0_fifo_priv *priv, int unit)
+{
+	u32 stat = nv_rd32(priv, 0x040108 + (unit * 0x2000));
+	u32 addr = nv_rd32(priv, 0x0400c0 + (unit * 0x2000));
+	u32 data = nv_rd32(priv, 0x0400c4 + (unit * 0x2000));
+	u32 chid = nv_rd32(priv, 0x040120 + (unit * 0x2000)) & 0xfff;
+	u32 subc = (addr & 0x00070000) >> 16;
+	u32 mthd = (addr & 0x00003ffc);
+	u32 show = stat;
+
+	if (stat & 0x00200000) {
+		if (mthd == 0x0054) {
+			if (!nve0_fifo_swmthd(priv, chid, 0x0500, 0x00000000))
+				show &= ~0x00200000;
+		}
+	}
+
+	if (stat & 0x00800000) {
+		if (!nve0_fifo_swmthd(priv, chid, mthd, data))
+			show &= ~0x00800000;
+	}
+
+	if (show) {
+		nv_error(priv, "SUBFIFO%d:", unit);
+		nouveau_bitfield_print(nve0_fifo_subfifo_intr, show);
+		printk("\n");
+		nv_error(priv, "SUBFIFO%d: ch %d subc %d mthd 0x%04x "
+			       "data 0x%08x\n",
+			 unit, chid, subc, mthd, data);
+	}
+
+	nv_wr32(priv, 0x0400c0 + (unit * 0x2000), 0x80600008);
+	nv_wr32(priv, 0x040108 + (unit * 0x2000), stat);
+}
+
+static void
+nve0_fifo_intr(struct nouveau_subdev *subdev)
+{
+	struct nve0_fifo_priv *priv = (void *)subdev;
+	u32 mask = nv_rd32(priv, 0x002140);
+	u32 stat = nv_rd32(priv, 0x002100) & mask;
+
+	if (stat & 0x00000100) {
+		nv_warn(priv, "unknown status 0x00000100\n");
+		nv_wr32(priv, 0x002100, 0x00000100);
+		stat &= ~0x00000100;
+	}
+
+	if (stat & 0x10000000) {
+		u32 units = nv_rd32(priv, 0x00259c);
+		u32 u = units;
+
+		while (u) {
+			int i = ffs(u) - 1;
+			nve0_fifo_isr_vm_fault(priv, i);
+			u &= ~(1 << i);
+		}
+
+		nv_wr32(priv, 0x00259c, units);
+		stat &= ~0x10000000;
+	}
+
+	if (stat & 0x20000000) {
+		u32 units = nv_rd32(priv, 0x0025a0);
+		u32 u = units;
+
+		while (u) {
+			int i = ffs(u) - 1;
+			nve0_fifo_isr_subfifo_intr(priv, i);
+			u &= ~(1 << i);
+		}
+
+		nv_wr32(priv, 0x0025a0, units);
+		stat &= ~0x20000000;
+	}
+
+	if (stat & 0x40000000) {
+		nv_warn(priv, "unknown status 0x40000000\n");
+		nv_mask(priv, 0x002a00, 0x00000000, 0x00000000);
+		stat &= ~0x40000000;
+	}
+
+	if (stat) {
+		nv_fatal(priv, "unhandled status 0x%08x\n", stat);
+		nv_wr32(priv, 0x002100, stat);
+		nv_wr32(priv, 0x002140, 0);
+	}
+}
+
+static int
+nve0_fifo_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nve0_fifo_priv *priv;
+	int ret;
+
+	ret = nouveau_fifo_create(parent, engine, oclass, 0, 4095, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 4096 * 0x200, 0x1000,
+				 NVOBJ_FLAG_ZERO_ALLOC, &priv->user.mem);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_map(priv->user.mem, NV_MEM_ACCESS_RW,
+				&priv->user.bar);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00000100;
+	nv_subdev(priv)->intr = nve0_fifo_intr;
+	nv_engine(priv)->cclass = &nve0_fifo_cclass;
+	nv_engine(priv)->sclass = nve0_fifo_sclass;
+	return 0;
+}
+
+static void
+nve0_fifo_dtor(struct nouveau_object *object)
+{
+	struct nve0_fifo_priv *priv = (void *)object;
+	int i;
+
+	nouveau_gpuobj_unmap(&priv->user.bar);
+	nouveau_gpuobj_ref(NULL, &priv->user.mem);
+
+	for (i = 0; i < ARRAY_SIZE(priv->engine); i++) {
+		nouveau_gpuobj_ref(NULL, &priv->engine[i].playlist[1]);
+		nouveau_gpuobj_ref(NULL, &priv->engine[i].playlist[0]);
+	}
+
+	nouveau_fifo_destroy(&priv->base);
+}
+
+static int
+nve0_fifo_init(struct nouveau_object *object)
+{
+	struct nve0_fifo_priv *priv = (void *)object;
+	int ret, i;
+
+	ret = nouveau_fifo_init(&priv->base);
+	if (ret)
+		return ret;
+
+	/* enable all available PSUBFIFOs */
+	nv_wr32(priv, 0x000204, 0xffffffff);
+	priv->spoon_nr = hweight32(nv_rd32(priv, 0x000204));
+	nv_debug(priv, "%d subfifo(s)\n", priv->spoon_nr);
+
+	/* PSUBFIFO[n] */
+	for (i = 0; i < priv->spoon_nr; i++) {
+		nv_mask(priv, 0x04013c + (i * 0x2000), 0x10000100, 0x00000000);
+		nv_wr32(priv, 0x040108 + (i * 0x2000), 0xffffffff); /* INTR */
+		nv_wr32(priv, 0x04010c + (i * 0x2000), 0xfffffeff); /* INTREN */
+	}
+
+	nv_wr32(priv, 0x002254, 0x10000000 | priv->user.bar.offset >> 12);
+
+	nv_wr32(priv, 0x002a00, 0xffffffff);
+	nv_wr32(priv, 0x002100, 0xffffffff);
+	nv_wr32(priv, 0x002140, 0xbfffffff);
+	return 0;
+}
+
+struct nouveau_oclass
+nve0_fifo_oclass = {
+	.handle = NV_ENGINE(FIFO, 0xe0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nve0_fifo_ctor,
+		.dtor = nve0_fifo_dtor,
+		.init = nve0_fifo_init,
+		.fini = _nouveau_fifo_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/ctx.h b/drivers/gpu/drm/nouveau/core/engine/graph/ctx.h
new file mode 100644
index 0000000..e194701
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/ctx.h
@@ -0,0 +1,129 @@
+#ifndef __NOUVEAU_GRCTX_H__
+#define __NOUVEAU_GRCTX_H__
+
+struct nouveau_grctx {
+	struct nouveau_device *device;
+
+	enum {
+		NOUVEAU_GRCTX_PROG,
+		NOUVEAU_GRCTX_VALS
+	} mode;
+	void *data;
+
+	u32 ctxprog_max;
+	u32 ctxprog_len;
+	u32 ctxprog_reg;
+	int ctxprog_label[32];
+	u32 ctxvals_pos;
+	u32 ctxvals_base;
+};
+
+static inline void
+cp_out(struct nouveau_grctx *ctx, u32 inst)
+{
+	u32 *ctxprog = ctx->data;
+
+	if (ctx->mode != NOUVEAU_GRCTX_PROG)
+		return;
+
+	BUG_ON(ctx->ctxprog_len == ctx->ctxprog_max);
+	ctxprog[ctx->ctxprog_len++] = inst;
+}
+
+static inline void
+cp_lsr(struct nouveau_grctx *ctx, u32 val)
+{
+	cp_out(ctx, CP_LOAD_SR | val);
+}
+
+static inline void
+cp_ctx(struct nouveau_grctx *ctx, u32 reg, u32 length)
+{
+	ctx->ctxprog_reg = (reg - 0x00400000) >> 2;
+
+	ctx->ctxvals_base = ctx->ctxvals_pos;
+	ctx->ctxvals_pos = ctx->ctxvals_base + length;
+
+	if (length > (CP_CTX_COUNT >> CP_CTX_COUNT_SHIFT)) {
+		cp_lsr(ctx, length);
+		length = 0;
+	}
+
+	cp_out(ctx, CP_CTX | (length << CP_CTX_COUNT_SHIFT) | ctx->ctxprog_reg);
+}
+
+static inline void
+cp_name(struct nouveau_grctx *ctx, int name)
+{
+	u32 *ctxprog = ctx->data;
+	int i;
+
+	if (ctx->mode != NOUVEAU_GRCTX_PROG)
+		return;
+
+	ctx->ctxprog_label[name] = ctx->ctxprog_len;
+	for (i = 0; i < ctx->ctxprog_len; i++) {
+		if ((ctxprog[i] & 0xfff00000) != 0xff400000)
+			continue;
+		if ((ctxprog[i] & CP_BRA_IP) != ((name) << CP_BRA_IP_SHIFT))
+			continue;
+		ctxprog[i] = (ctxprog[i] & 0x00ff00ff) |
+			     (ctx->ctxprog_len << CP_BRA_IP_SHIFT);
+	}
+}
+
+static inline void
+_cp_bra(struct nouveau_grctx *ctx, u32 mod, int flag, int state, int name)
+{
+	int ip = 0;
+
+	if (mod != 2) {
+		ip = ctx->ctxprog_label[name] << CP_BRA_IP_SHIFT;
+		if (ip == 0)
+			ip = 0xff000000 | (name << CP_BRA_IP_SHIFT);
+	}
+
+	cp_out(ctx, CP_BRA | (mod << 18) | ip | flag |
+		    (state ? 0 : CP_BRA_IF_CLEAR));
+}
+#define cp_bra(c, f, s, n) _cp_bra((c), 0, CP_FLAG_##f, CP_FLAG_##f##_##s, n)
+#define cp_cal(c, f, s, n) _cp_bra((c), 1, CP_FLAG_##f, CP_FLAG_##f##_##s, n)
+#define cp_ret(c, f, s) _cp_bra((c), 2, CP_FLAG_##f, CP_FLAG_##f##_##s, 0)
+
+static inline void
+_cp_wait(struct nouveau_grctx *ctx, int flag, int state)
+{
+	cp_out(ctx, CP_WAIT | flag | (state ? CP_WAIT_SET : 0));
+}
+#define cp_wait(c, f, s) _cp_wait((c), CP_FLAG_##f, CP_FLAG_##f##_##s)
+
+static inline void
+_cp_set(struct nouveau_grctx *ctx, int flag, int state)
+{
+	cp_out(ctx, CP_SET | flag | (state ? CP_SET_1 : 0));
+}
+#define cp_set(c, f, s) _cp_set((c), CP_FLAG_##f, CP_FLAG_##f##_##s)
+
+static inline void
+cp_pos(struct nouveau_grctx *ctx, int offset)
+{
+	ctx->ctxvals_pos = offset;
+	ctx->ctxvals_base = ctx->ctxvals_pos;
+
+	cp_lsr(ctx, ctx->ctxvals_pos);
+	cp_out(ctx, CP_SET_CONTEXT_POINTER);
+}
+
+static inline void
+gr_def(struct nouveau_grctx *ctx, u32 reg, u32 val)
+{
+	if (ctx->mode != NOUVEAU_GRCTX_VALS)
+		return;
+
+	reg = (reg - 0x00400000) / 4;
+	reg = (reg - ctx->ctxprog_reg) + ctx->ctxvals_base;
+
+	nv_wo32(ctx->data, reg * 4, val);
+}
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/ctxnv40.c b/drivers/gpu/drm/nouveau/core/engine/graph/ctxnv40.c
new file mode 100644
index 0000000..e45035e
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/ctxnv40.c
@@ -0,0 +1,689 @@
+/*
+ * Copyright 2009 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/gpuobj.h>
+
+/* NVIDIA context programs handle a number of other conditions which are
+ * not implemented in our versions.  It's not clear why NVIDIA context
+ * programs have this code, nor whether it's strictly necessary for
+ * correct operation.  We'll implement additional handling if/when we
+ * discover it's necessary.
+ *
+ * - On context save, NVIDIA set 0x400314 bit 0 to 1 if the "3D state"
+ *   flag is set, this gets saved into the context.
+ * - On context save, the context program for all cards load nsource
+ *   into a flag register and check for ILLEGAL_MTHD.  If it's set,
+ *   opcode 0x60000d is called before resuming normal operation.
+ * - Some context programs check more conditions than the above.  NV44
+ *   checks: ((nsource & 0x0857) || (0x400718 & 0x0100) || (intr & 0x0001))
+ *   and calls 0x60000d before resuming normal operation.
+ * - At the very beginning of NVIDIA's context programs, flag 9 is checked
+ *   and if true 0x800001 is called with count=0, pos=0, the flag is cleared
+ *   and then the ctxprog is aborted.  It looks like a complicated NOP,
+ *   its purpose is unknown.
+ * - In the section of code that loads the per-vs state, NVIDIA check
+ *   flag 10.  If it's set, they only transfer the small 0x300 byte block
+ *   of state + the state for a single vs as opposed to the state for
+ *   all vs units.  It doesn't seem likely that it'll occur in normal
+ *   operation, especially seeing as it appears NVIDIA may have screwed
+ *   up the ctxprogs for some cards and have an invalid instruction
+ *   rather than a cp_lsr(ctx, dwords_for_1_vs_unit) instruction.
+ * - There's a number of places where context offset 0 (where we place
+ *   the PRAMIN offset of the context) is loaded into either 0x408000,
+ *   0x408004 or 0x408008.  Not sure what's up there either.
+ * - The ctxprogs for some cards save 0x400a00 again during the cleanup
+ *   path for auto-loadctx.
+ */
+
+#define CP_FLAG_CLEAR                 0
+#define CP_FLAG_SET                   1
+#define CP_FLAG_SWAP_DIRECTION        ((0 * 32) + 0)
+#define CP_FLAG_SWAP_DIRECTION_LOAD   0
+#define CP_FLAG_SWAP_DIRECTION_SAVE   1
+#define CP_FLAG_USER_SAVE             ((0 * 32) + 5)
+#define CP_FLAG_USER_SAVE_NOT_PENDING 0
+#define CP_FLAG_USER_SAVE_PENDING     1
+#define CP_FLAG_USER_LOAD             ((0 * 32) + 6)
+#define CP_FLAG_USER_LOAD_NOT_PENDING 0
+#define CP_FLAG_USER_LOAD_PENDING     1
+#define CP_FLAG_STATUS                ((3 * 32) + 0)
+#define CP_FLAG_STATUS_IDLE           0
+#define CP_FLAG_STATUS_BUSY           1
+#define CP_FLAG_AUTO_SAVE             ((3 * 32) + 4)
+#define CP_FLAG_AUTO_SAVE_NOT_PENDING 0
+#define CP_FLAG_AUTO_SAVE_PENDING     1
+#define CP_FLAG_AUTO_LOAD             ((3 * 32) + 5)
+#define CP_FLAG_AUTO_LOAD_NOT_PENDING 0
+#define CP_FLAG_AUTO_LOAD_PENDING     1
+#define CP_FLAG_UNK54                 ((3 * 32) + 6)
+#define CP_FLAG_UNK54_CLEAR           0
+#define CP_FLAG_UNK54_SET             1
+#define CP_FLAG_ALWAYS                ((3 * 32) + 8)
+#define CP_FLAG_ALWAYS_FALSE          0
+#define CP_FLAG_ALWAYS_TRUE           1
+#define CP_FLAG_UNK57                 ((3 * 32) + 9)
+#define CP_FLAG_UNK57_CLEAR           0
+#define CP_FLAG_UNK57_SET             1
+
+#define CP_CTX                   0x00100000
+#define CP_CTX_COUNT             0x000fc000
+#define CP_CTX_COUNT_SHIFT               14
+#define CP_CTX_REG               0x00003fff
+#define CP_LOAD_SR               0x00200000
+#define CP_LOAD_SR_VALUE         0x000fffff
+#define CP_BRA                   0x00400000
+#define CP_BRA_IP                0x0000ff00
+#define CP_BRA_IP_SHIFT                   8
+#define CP_BRA_IF_CLEAR          0x00000080
+#define CP_BRA_FLAG              0x0000007f
+#define CP_WAIT                  0x00500000
+#define CP_WAIT_SET              0x00000080
+#define CP_WAIT_FLAG             0x0000007f
+#define CP_SET                   0x00700000
+#define CP_SET_1                 0x00000080
+#define CP_SET_FLAG              0x0000007f
+#define CP_NEXT_TO_SWAP          0x00600007
+#define CP_NEXT_TO_CURRENT       0x00600009
+#define CP_SET_CONTEXT_POINTER   0x0060000a
+#define CP_END                   0x0060000e
+#define CP_LOAD_MAGIC_UNK01      0x00800001 /* unknown */
+#define CP_LOAD_MAGIC_NV44TCL    0x00800029 /* per-vs state (0x4497) */
+#define CP_LOAD_MAGIC_NV40TCL    0x00800041 /* per-vs state (0x4097) */
+
+#include "nv40.h"
+#include "ctx.h"
+
+/* TODO:
+ *  - get vs count from 0x1540
+ */
+
+static int
+nv40_graph_vs_count(struct nouveau_device *device)
+{
+
+	switch (device->chipset) {
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		return 8;
+	case 0x40:
+		return 6;
+	case 0x41:
+	case 0x42:
+		return 5;
+	case 0x43:
+	case 0x44:
+	case 0x46:
+	case 0x4a:
+		return 3;
+	case 0x4c:
+	case 0x4e:
+	case 0x67:
+	default:
+		return 1;
+	}
+}
+
+
+enum cp_label {
+	cp_check_load = 1,
+	cp_setup_auto_load,
+	cp_setup_load,
+	cp_setup_save,
+	cp_swap_state,
+	cp_swap_state3d_3_is_save,
+	cp_prepare_exit,
+	cp_exit,
+};
+
+static void
+nv40_graph_construct_general(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int i;
+
+	cp_ctx(ctx, 0x4000a4, 1);
+	gr_def(ctx, 0x4000a4, 0x00000008);
+	cp_ctx(ctx, 0x400144, 58);
+	gr_def(ctx, 0x400144, 0x00000001);
+	cp_ctx(ctx, 0x400314, 1);
+	gr_def(ctx, 0x400314, 0x00000000);
+	cp_ctx(ctx, 0x400400, 10);
+	cp_ctx(ctx, 0x400480, 10);
+	cp_ctx(ctx, 0x400500, 19);
+	gr_def(ctx, 0x400514, 0x00040000);
+	gr_def(ctx, 0x400524, 0x55555555);
+	gr_def(ctx, 0x400528, 0x55555555);
+	gr_def(ctx, 0x40052c, 0x55555555);
+	gr_def(ctx, 0x400530, 0x55555555);
+	cp_ctx(ctx, 0x400560, 6);
+	gr_def(ctx, 0x400568, 0x0000ffff);
+	gr_def(ctx, 0x40056c, 0x0000ffff);
+	cp_ctx(ctx, 0x40057c, 5);
+	cp_ctx(ctx, 0x400710, 3);
+	gr_def(ctx, 0x400710, 0x20010001);
+	gr_def(ctx, 0x400714, 0x0f73ef00);
+	cp_ctx(ctx, 0x400724, 1);
+	gr_def(ctx, 0x400724, 0x02008821);
+	cp_ctx(ctx, 0x400770, 3);
+	if (device->chipset == 0x40) {
+		cp_ctx(ctx, 0x400814, 4);
+		cp_ctx(ctx, 0x400828, 5);
+		cp_ctx(ctx, 0x400840, 5);
+		gr_def(ctx, 0x400850, 0x00000040);
+		cp_ctx(ctx, 0x400858, 4);
+		gr_def(ctx, 0x400858, 0x00000040);
+		gr_def(ctx, 0x40085c, 0x00000040);
+		gr_def(ctx, 0x400864, 0x80000000);
+		cp_ctx(ctx, 0x40086c, 9);
+		gr_def(ctx, 0x40086c, 0x80000000);
+		gr_def(ctx, 0x400870, 0x80000000);
+		gr_def(ctx, 0x400874, 0x80000000);
+		gr_def(ctx, 0x400878, 0x80000000);
+		gr_def(ctx, 0x400888, 0x00000040);
+		gr_def(ctx, 0x40088c, 0x80000000);
+		cp_ctx(ctx, 0x4009c0, 8);
+		gr_def(ctx, 0x4009cc, 0x80000000);
+		gr_def(ctx, 0x4009dc, 0x80000000);
+	} else {
+		cp_ctx(ctx, 0x400840, 20);
+		if (nv44_graph_class(ctx->device)) {
+			for (i = 0; i < 8; i++)
+				gr_def(ctx, 0x400860 + (i * 4), 0x00000001);
+		}
+		gr_def(ctx, 0x400880, 0x00000040);
+		gr_def(ctx, 0x400884, 0x00000040);
+		gr_def(ctx, 0x400888, 0x00000040);
+		cp_ctx(ctx, 0x400894, 11);
+		gr_def(ctx, 0x400894, 0x00000040);
+		if (!nv44_graph_class(ctx->device)) {
+			for (i = 0; i < 8; i++)
+				gr_def(ctx, 0x4008a0 + (i * 4), 0x80000000);
+		}
+		cp_ctx(ctx, 0x4008e0, 2);
+		cp_ctx(ctx, 0x4008f8, 2);
+		if (device->chipset == 0x4c ||
+		    (device->chipset & 0xf0) == 0x60)
+			cp_ctx(ctx, 0x4009f8, 1);
+	}
+	cp_ctx(ctx, 0x400a00, 73);
+	gr_def(ctx, 0x400b0c, 0x0b0b0b0c);
+	cp_ctx(ctx, 0x401000, 4);
+	cp_ctx(ctx, 0x405004, 1);
+	switch (device->chipset) {
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		cp_ctx(ctx, 0x403448, 1);
+		gr_def(ctx, 0x403448, 0x00001010);
+		break;
+	default:
+		cp_ctx(ctx, 0x403440, 1);
+		switch (device->chipset) {
+		case 0x40:
+			gr_def(ctx, 0x403440, 0x00000010);
+			break;
+		case 0x44:
+		case 0x46:
+		case 0x4a:
+			gr_def(ctx, 0x403440, 0x00003010);
+			break;
+		case 0x41:
+		case 0x42:
+		case 0x43:
+		case 0x4c:
+		case 0x4e:
+		case 0x67:
+		default:
+			gr_def(ctx, 0x403440, 0x00001010);
+			break;
+		}
+		break;
+	}
+}
+
+static void
+nv40_graph_construct_state3d(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int i;
+
+	if (device->chipset == 0x40) {
+		cp_ctx(ctx, 0x401880, 51);
+		gr_def(ctx, 0x401940, 0x00000100);
+	} else
+	if (device->chipset == 0x46 || device->chipset == 0x47 ||
+	    device->chipset == 0x49 || device->chipset == 0x4b) {
+		cp_ctx(ctx, 0x401880, 32);
+		for (i = 0; i < 16; i++)
+			gr_def(ctx, 0x401880 + (i * 4), 0x00000111);
+		if (device->chipset == 0x46)
+			cp_ctx(ctx, 0x401900, 16);
+		cp_ctx(ctx, 0x401940, 3);
+	}
+	cp_ctx(ctx, 0x40194c, 18);
+	gr_def(ctx, 0x401954, 0x00000111);
+	gr_def(ctx, 0x401958, 0x00080060);
+	gr_def(ctx, 0x401974, 0x00000080);
+	gr_def(ctx, 0x401978, 0xffff0000);
+	gr_def(ctx, 0x40197c, 0x00000001);
+	gr_def(ctx, 0x401990, 0x46400000);
+	if (device->chipset == 0x40) {
+		cp_ctx(ctx, 0x4019a0, 2);
+		cp_ctx(ctx, 0x4019ac, 5);
+	} else {
+		cp_ctx(ctx, 0x4019a0, 1);
+		cp_ctx(ctx, 0x4019b4, 3);
+	}
+	gr_def(ctx, 0x4019bc, 0xffff0000);
+	switch (device->chipset) {
+	case 0x46:
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		cp_ctx(ctx, 0x4019c0, 18);
+		for (i = 0; i < 16; i++)
+			gr_def(ctx, 0x4019c0 + (i * 4), 0x88888888);
+		break;
+	}
+	cp_ctx(ctx, 0x401a08, 8);
+	gr_def(ctx, 0x401a10, 0x0fff0000);
+	gr_def(ctx, 0x401a14, 0x0fff0000);
+	gr_def(ctx, 0x401a1c, 0x00011100);
+	cp_ctx(ctx, 0x401a2c, 4);
+	cp_ctx(ctx, 0x401a44, 26);
+	for (i = 0; i < 16; i++)
+		gr_def(ctx, 0x401a44 + (i * 4), 0x07ff0000);
+	gr_def(ctx, 0x401a8c, 0x4b7fffff);
+	if (device->chipset == 0x40) {
+		cp_ctx(ctx, 0x401ab8, 3);
+	} else {
+		cp_ctx(ctx, 0x401ab8, 1);
+		cp_ctx(ctx, 0x401ac0, 1);
+	}
+	cp_ctx(ctx, 0x401ad0, 8);
+	gr_def(ctx, 0x401ad0, 0x30201000);
+	gr_def(ctx, 0x401ad4, 0x70605040);
+	gr_def(ctx, 0x401ad8, 0xb8a89888);
+	gr_def(ctx, 0x401adc, 0xf8e8d8c8);
+	cp_ctx(ctx, 0x401b10, device->chipset == 0x40 ? 2 : 1);
+	gr_def(ctx, 0x401b10, 0x40100000);
+	cp_ctx(ctx, 0x401b18, device->chipset == 0x40 ? 6 : 5);
+	gr_def(ctx, 0x401b28, device->chipset == 0x40 ?
+			      0x00000004 : 0x00000000);
+	cp_ctx(ctx, 0x401b30, 25);
+	gr_def(ctx, 0x401b34, 0x0000ffff);
+	gr_def(ctx, 0x401b68, 0x435185d6);
+	gr_def(ctx, 0x401b6c, 0x2155b699);
+	gr_def(ctx, 0x401b70, 0xfedcba98);
+	gr_def(ctx, 0x401b74, 0x00000098);
+	gr_def(ctx, 0x401b84, 0xffffffff);
+	gr_def(ctx, 0x401b88, 0x00ff7000);
+	gr_def(ctx, 0x401b8c, 0x0000ffff);
+	if (device->chipset != 0x44 && device->chipset != 0x4a &&
+	    device->chipset != 0x4e)
+		cp_ctx(ctx, 0x401b94, 1);
+	cp_ctx(ctx, 0x401b98, 8);
+	gr_def(ctx, 0x401b9c, 0x00ff0000);
+	cp_ctx(ctx, 0x401bc0, 9);
+	gr_def(ctx, 0x401be0, 0x00ffff00);
+	cp_ctx(ctx, 0x401c00, 192);
+	for (i = 0; i < 16; i++) { /* fragment texture units */
+		gr_def(ctx, 0x401c40 + (i * 4), 0x00018488);
+		gr_def(ctx, 0x401c80 + (i * 4), 0x00028202);
+		gr_def(ctx, 0x401d00 + (i * 4), 0x0000aae4);
+		gr_def(ctx, 0x401d40 + (i * 4), 0x01012000);
+		gr_def(ctx, 0x401d80 + (i * 4), 0x00080008);
+		gr_def(ctx, 0x401e00 + (i * 4), 0x00100008);
+	}
+	for (i = 0; i < 4; i++) { /* vertex texture units */
+		gr_def(ctx, 0x401e90 + (i * 4), 0x0001bc80);
+		gr_def(ctx, 0x401ea0 + (i * 4), 0x00000202);
+		gr_def(ctx, 0x401ec0 + (i * 4), 0x00000008);
+		gr_def(ctx, 0x401ee0 + (i * 4), 0x00080008);
+	}
+	cp_ctx(ctx, 0x400f5c, 3);
+	gr_def(ctx, 0x400f5c, 0x00000002);
+	cp_ctx(ctx, 0x400f84, 1);
+}
+
+static void
+nv40_graph_construct_state3d_2(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int i;
+
+	cp_ctx(ctx, 0x402000, 1);
+	cp_ctx(ctx, 0x402404, device->chipset == 0x40 ? 1 : 2);
+	switch (device->chipset) {
+	case 0x40:
+		gr_def(ctx, 0x402404, 0x00000001);
+		break;
+	case 0x4c:
+	case 0x4e:
+	case 0x67:
+		gr_def(ctx, 0x402404, 0x00000020);
+		break;
+	case 0x46:
+	case 0x49:
+	case 0x4b:
+		gr_def(ctx, 0x402404, 0x00000421);
+		break;
+	default:
+		gr_def(ctx, 0x402404, 0x00000021);
+	}
+	if (device->chipset != 0x40)
+		gr_def(ctx, 0x402408, 0x030c30c3);
+	switch (device->chipset) {
+	case 0x44:
+	case 0x46:
+	case 0x4a:
+	case 0x4c:
+	case 0x4e:
+	case 0x67:
+		cp_ctx(ctx, 0x402440, 1);
+		gr_def(ctx, 0x402440, 0x00011001);
+		break;
+	default:
+		break;
+	}
+	cp_ctx(ctx, 0x402480, device->chipset == 0x40 ? 8 : 9);
+	gr_def(ctx, 0x402488, 0x3e020200);
+	gr_def(ctx, 0x40248c, 0x00ffffff);
+	switch (device->chipset) {
+	case 0x40:
+		gr_def(ctx, 0x402490, 0x60103f00);
+		break;
+	case 0x47:
+		gr_def(ctx, 0x402490, 0x40103f00);
+		break;
+	case 0x41:
+	case 0x42:
+	case 0x49:
+	case 0x4b:
+		gr_def(ctx, 0x402490, 0x20103f00);
+		break;
+	default:
+		gr_def(ctx, 0x402490, 0x0c103f00);
+		break;
+	}
+	gr_def(ctx, 0x40249c, device->chipset <= 0x43 ?
+			      0x00020000 : 0x00040000);
+	cp_ctx(ctx, 0x402500, 31);
+	gr_def(ctx, 0x402530, 0x00008100);
+	if (device->chipset == 0x40)
+		cp_ctx(ctx, 0x40257c, 6);
+	cp_ctx(ctx, 0x402594, 16);
+	cp_ctx(ctx, 0x402800, 17);
+	gr_def(ctx, 0x402800, 0x00000001);
+	switch (device->chipset) {
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		cp_ctx(ctx, 0x402864, 1);
+		gr_def(ctx, 0x402864, 0x00001001);
+		cp_ctx(ctx, 0x402870, 3);
+		gr_def(ctx, 0x402878, 0x00000003);
+		if (device->chipset != 0x47) { /* belong at end!! */
+			cp_ctx(ctx, 0x402900, 1);
+			cp_ctx(ctx, 0x402940, 1);
+			cp_ctx(ctx, 0x402980, 1);
+			cp_ctx(ctx, 0x4029c0, 1);
+			cp_ctx(ctx, 0x402a00, 1);
+			cp_ctx(ctx, 0x402a40, 1);
+			cp_ctx(ctx, 0x402a80, 1);
+			cp_ctx(ctx, 0x402ac0, 1);
+		}
+		break;
+	case 0x40:
+		cp_ctx(ctx, 0x402844, 1);
+		gr_def(ctx, 0x402844, 0x00000001);
+		cp_ctx(ctx, 0x402850, 1);
+		break;
+	default:
+		cp_ctx(ctx, 0x402844, 1);
+		gr_def(ctx, 0x402844, 0x00001001);
+		cp_ctx(ctx, 0x402850, 2);
+		gr_def(ctx, 0x402854, 0x00000003);
+		break;
+	}
+
+	cp_ctx(ctx, 0x402c00, 4);
+	gr_def(ctx, 0x402c00, device->chipset == 0x40 ?
+			      0x80800001 : 0x00888001);
+	switch (device->chipset) {
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		cp_ctx(ctx, 0x402c20, 40);
+		for (i = 0; i < 32; i++)
+			gr_def(ctx, 0x402c40 + (i * 4), 0xffffffff);
+		cp_ctx(ctx, 0x4030b8, 13);
+		gr_def(ctx, 0x4030dc, 0x00000005);
+		gr_def(ctx, 0x4030e8, 0x0000ffff);
+		break;
+	default:
+		cp_ctx(ctx, 0x402c10, 4);
+		if (device->chipset == 0x40)
+			cp_ctx(ctx, 0x402c20, 36);
+		else
+		if (device->chipset <= 0x42)
+			cp_ctx(ctx, 0x402c20, 24);
+		else
+		if (device->chipset <= 0x4a)
+			cp_ctx(ctx, 0x402c20, 16);
+		else
+			cp_ctx(ctx, 0x402c20, 8);
+		cp_ctx(ctx, 0x402cb0, device->chipset == 0x40 ? 12 : 13);
+		gr_def(ctx, 0x402cd4, 0x00000005);
+		if (device->chipset != 0x40)
+			gr_def(ctx, 0x402ce0, 0x0000ffff);
+		break;
+	}
+
+	cp_ctx(ctx, 0x403400, device->chipset == 0x40 ? 4 : 3);
+	cp_ctx(ctx, 0x403410, device->chipset == 0x40 ? 4 : 3);
+	cp_ctx(ctx, 0x403420, nv40_graph_vs_count(ctx->device));
+	for (i = 0; i < nv40_graph_vs_count(ctx->device); i++)
+		gr_def(ctx, 0x403420 + (i * 4), 0x00005555);
+
+	if (device->chipset != 0x40) {
+		cp_ctx(ctx, 0x403600, 1);
+		gr_def(ctx, 0x403600, 0x00000001);
+	}
+	cp_ctx(ctx, 0x403800, 1);
+
+	cp_ctx(ctx, 0x403c18, 1);
+	gr_def(ctx, 0x403c18, 0x00000001);
+	switch (device->chipset) {
+	case 0x46:
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		cp_ctx(ctx, 0x405018, 1);
+		gr_def(ctx, 0x405018, 0x08e00001);
+		cp_ctx(ctx, 0x405c24, 1);
+		gr_def(ctx, 0x405c24, 0x000e3000);
+		break;
+	}
+	if (device->chipset != 0x4e)
+		cp_ctx(ctx, 0x405800, 11);
+	cp_ctx(ctx, 0x407000, 1);
+}
+
+static void
+nv40_graph_construct_state3d_3(struct nouveau_grctx *ctx)
+{
+	int len = nv44_graph_class(ctx->device) ? 0x0084 : 0x0684;
+
+	cp_out (ctx, 0x300000);
+	cp_lsr (ctx, len - 4);
+	cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_swap_state3d_3_is_save);
+	cp_lsr (ctx, len);
+	cp_name(ctx, cp_swap_state3d_3_is_save);
+	cp_out (ctx, 0x800001);
+
+	ctx->ctxvals_pos += len;
+}
+
+static void
+nv40_graph_construct_shader(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	struct nouveau_gpuobj *obj = ctx->data;
+	int vs, vs_nr, vs_len, vs_nr_b0, vs_nr_b1, b0_offset, b1_offset;
+	int offset, i;
+
+	vs_nr    = nv40_graph_vs_count(ctx->device);
+	vs_nr_b0 = 363;
+	vs_nr_b1 = device->chipset == 0x40 ? 128 : 64;
+	if (device->chipset == 0x40) {
+		b0_offset = 0x2200/4; /* 33a0 */
+		b1_offset = 0x55a0/4; /* 1500 */
+		vs_len = 0x6aa0/4;
+	} else
+	if (device->chipset == 0x41 || device->chipset == 0x42) {
+		b0_offset = 0x2200/4; /* 2200 */
+		b1_offset = 0x4400/4; /* 0b00 */
+		vs_len = 0x4f00/4;
+	} else {
+		b0_offset = 0x1d40/4; /* 2200 */
+		b1_offset = 0x3f40/4; /* 0b00 : 0a40 */
+		vs_len = nv44_graph_class(device) ? 0x4980/4 : 0x4a40/4;
+	}
+
+	cp_lsr(ctx, vs_len * vs_nr + 0x300/4);
+	cp_out(ctx, nv44_graph_class(device) ? 0x800029 : 0x800041);
+
+	offset = ctx->ctxvals_pos;
+	ctx->ctxvals_pos += (0x0300/4 + (vs_nr * vs_len));
+
+	if (ctx->mode != NOUVEAU_GRCTX_VALS)
+		return;
+
+	offset += 0x0280/4;
+	for (i = 0; i < 16; i++, offset += 2)
+		nv_wo32(obj, offset * 4, 0x3f800000);
+
+	for (vs = 0; vs < vs_nr; vs++, offset += vs_len) {
+		for (i = 0; i < vs_nr_b0 * 6; i += 6)
+			nv_wo32(obj, (offset + b0_offset + i) * 4, 0x00000001);
+		for (i = 0; i < vs_nr_b1 * 4; i += 4)
+			nv_wo32(obj, (offset + b1_offset + i) * 4, 0x3f800000);
+	}
+}
+
+static void
+nv40_grctx_generate(struct nouveau_grctx *ctx)
+{
+	/* decide whether we're loading/unloading the context */
+	cp_bra (ctx, AUTO_SAVE, PENDING, cp_setup_save);
+	cp_bra (ctx, USER_SAVE, PENDING, cp_setup_save);
+
+	cp_name(ctx, cp_check_load);
+	cp_bra (ctx, AUTO_LOAD, PENDING, cp_setup_auto_load);
+	cp_bra (ctx, USER_LOAD, PENDING, cp_setup_load);
+	cp_bra (ctx, ALWAYS, TRUE, cp_exit);
+
+	/* setup for context load */
+	cp_name(ctx, cp_setup_auto_load);
+	cp_wait(ctx, STATUS, IDLE);
+	cp_out (ctx, CP_NEXT_TO_SWAP);
+	cp_name(ctx, cp_setup_load);
+	cp_wait(ctx, STATUS, IDLE);
+	cp_set (ctx, SWAP_DIRECTION, LOAD);
+	cp_out (ctx, 0x00910880); /* ?? */
+	cp_out (ctx, 0x00901ffe); /* ?? */
+	cp_out (ctx, 0x01940000); /* ?? */
+	cp_lsr (ctx, 0x20);
+	cp_out (ctx, 0x0060000b); /* ?? */
+	cp_wait(ctx, UNK57, CLEAR);
+	cp_out (ctx, 0x0060000c); /* ?? */
+	cp_bra (ctx, ALWAYS, TRUE, cp_swap_state);
+
+	/* setup for context save */
+	cp_name(ctx, cp_setup_save);
+	cp_set (ctx, SWAP_DIRECTION, SAVE);
+
+	/* general PGRAPH state */
+	cp_name(ctx, cp_swap_state);
+	cp_pos (ctx, 0x00020/4);
+	nv40_graph_construct_general(ctx);
+	cp_wait(ctx, STATUS, IDLE);
+
+	/* 3D state, block 1 */
+	cp_bra (ctx, UNK54, CLEAR, cp_prepare_exit);
+	nv40_graph_construct_state3d(ctx);
+	cp_wait(ctx, STATUS, IDLE);
+
+	/* 3D state, block 2 */
+	nv40_graph_construct_state3d_2(ctx);
+
+	/* Some other block of "random" state */
+	nv40_graph_construct_state3d_3(ctx);
+
+	/* Per-vertex shader state */
+	cp_pos (ctx, ctx->ctxvals_pos);
+	nv40_graph_construct_shader(ctx);
+
+	/* pre-exit state updates */
+	cp_name(ctx, cp_prepare_exit);
+	cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_check_load);
+	cp_bra (ctx, USER_SAVE, PENDING, cp_exit);
+	cp_out (ctx, CP_NEXT_TO_CURRENT);
+
+	cp_name(ctx, cp_exit);
+	cp_set (ctx, USER_SAVE, NOT_PENDING);
+	cp_set (ctx, USER_LOAD, NOT_PENDING);
+	cp_out (ctx, CP_END);
+}
+
+void
+nv40_grctx_fill(struct nouveau_device *device, struct nouveau_gpuobj *mem)
+{
+	nv40_grctx_generate(&(struct nouveau_grctx) {
+			     .device = device,
+			     .mode = NOUVEAU_GRCTX_VALS,
+			     .data = mem,
+			   });
+}
+
+void
+nv40_grctx_init(struct nouveau_device *device, u32 *size)
+{
+	u32 ctxprog[256], i;
+	struct nouveau_grctx ctx = {
+		.device = device,
+		.mode = NOUVEAU_GRCTX_PROG,
+		.data = ctxprog,
+		.ctxprog_max = ARRAY_SIZE(ctxprog)
+	};
+
+	nv40_grctx_generate(&ctx);
+
+	nv_wr32(device, 0x400324, 0);
+	for (i = 0; i < ctx.ctxprog_len; i++)
+		nv_wr32(device, 0x400328, ctxprog[i]);
+	*size = ctx.ctxvals_pos * 4;
+}
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/ctxnv50.c b/drivers/gpu/drm/nouveau/core/engine/graph/ctxnv50.c
new file mode 100644
index 0000000..552fdbd
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/ctxnv50.c
@@ -0,0 +1,3341 @@
+/*
+ * Copyright 2009 Marcin Kościelnicki
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include <core/gpuobj.h>
+
+#define CP_FLAG_CLEAR                 0
+#define CP_FLAG_SET                   1
+#define CP_FLAG_SWAP_DIRECTION        ((0 * 32) + 0)
+#define CP_FLAG_SWAP_DIRECTION_LOAD   0
+#define CP_FLAG_SWAP_DIRECTION_SAVE   1
+#define CP_FLAG_UNK01                 ((0 * 32) + 1)
+#define CP_FLAG_UNK01_CLEAR           0
+#define CP_FLAG_UNK01_SET             1
+#define CP_FLAG_UNK03                 ((0 * 32) + 3)
+#define CP_FLAG_UNK03_CLEAR           0
+#define CP_FLAG_UNK03_SET             1
+#define CP_FLAG_USER_SAVE             ((0 * 32) + 5)
+#define CP_FLAG_USER_SAVE_NOT_PENDING 0
+#define CP_FLAG_USER_SAVE_PENDING     1
+#define CP_FLAG_USER_LOAD             ((0 * 32) + 6)
+#define CP_FLAG_USER_LOAD_NOT_PENDING 0
+#define CP_FLAG_USER_LOAD_PENDING     1
+#define CP_FLAG_UNK0B                 ((0 * 32) + 0xb)
+#define CP_FLAG_UNK0B_CLEAR           0
+#define CP_FLAG_UNK0B_SET             1
+#define CP_FLAG_XFER_SWITCH           ((0 * 32) + 0xe)
+#define CP_FLAG_XFER_SWITCH_DISABLE   0
+#define CP_FLAG_XFER_SWITCH_ENABLE    1
+#define CP_FLAG_STATE                 ((0 * 32) + 0x1c)
+#define CP_FLAG_STATE_STOPPED         0
+#define CP_FLAG_STATE_RUNNING         1
+#define CP_FLAG_UNK1D                 ((0 * 32) + 0x1d)
+#define CP_FLAG_UNK1D_CLEAR           0
+#define CP_FLAG_UNK1D_SET             1
+#define CP_FLAG_UNK20                 ((1 * 32) + 0)
+#define CP_FLAG_UNK20_CLEAR           0
+#define CP_FLAG_UNK20_SET             1
+#define CP_FLAG_STATUS                ((2 * 32) + 0)
+#define CP_FLAG_STATUS_BUSY           0
+#define CP_FLAG_STATUS_IDLE           1
+#define CP_FLAG_AUTO_SAVE             ((2 * 32) + 4)
+#define CP_FLAG_AUTO_SAVE_NOT_PENDING 0
+#define CP_FLAG_AUTO_SAVE_PENDING     1
+#define CP_FLAG_AUTO_LOAD             ((2 * 32) + 5)
+#define CP_FLAG_AUTO_LOAD_NOT_PENDING 0
+#define CP_FLAG_AUTO_LOAD_PENDING     1
+#define CP_FLAG_NEWCTX                ((2 * 32) + 10)
+#define CP_FLAG_NEWCTX_BUSY           0
+#define CP_FLAG_NEWCTX_DONE           1
+#define CP_FLAG_XFER                  ((2 * 32) + 11)
+#define CP_FLAG_XFER_IDLE             0
+#define CP_FLAG_XFER_BUSY             1
+#define CP_FLAG_ALWAYS                ((2 * 32) + 13)
+#define CP_FLAG_ALWAYS_FALSE          0
+#define CP_FLAG_ALWAYS_TRUE           1
+#define CP_FLAG_INTR                  ((2 * 32) + 15)
+#define CP_FLAG_INTR_NOT_PENDING      0
+#define CP_FLAG_INTR_PENDING          1
+
+#define CP_CTX                   0x00100000
+#define CP_CTX_COUNT             0x000f0000
+#define CP_CTX_COUNT_SHIFT               16
+#define CP_CTX_REG               0x00003fff
+#define CP_LOAD_SR               0x00200000
+#define CP_LOAD_SR_VALUE         0x000fffff
+#define CP_BRA                   0x00400000
+#define CP_BRA_IP                0x0001ff00
+#define CP_BRA_IP_SHIFT                   8
+#define CP_BRA_IF_CLEAR          0x00000080
+#define CP_BRA_FLAG              0x0000007f
+#define CP_WAIT                  0x00500000
+#define CP_WAIT_SET              0x00000080
+#define CP_WAIT_FLAG             0x0000007f
+#define CP_SET                   0x00700000
+#define CP_SET_1                 0x00000080
+#define CP_SET_FLAG              0x0000007f
+#define CP_NEWCTX                0x00600004
+#define CP_NEXT_TO_SWAP          0x00600005
+#define CP_SET_CONTEXT_POINTER   0x00600006
+#define CP_SET_XFER_POINTER      0x00600007
+#define CP_ENABLE                0x00600009
+#define CP_END                   0x0060000c
+#define CP_NEXT_TO_CURRENT       0x0060000d
+#define CP_DISABLE1              0x0090ffff
+#define CP_DISABLE2              0x0091ffff
+#define CP_XFER_1      0x008000ff
+#define CP_XFER_2      0x008800ff
+#define CP_SEEK_1      0x00c000ff
+#define CP_SEEK_2      0x00c800ff
+
+#include "nv50.h"
+#include "ctx.h"
+
+#define IS_NVA3F(x) (((x) > 0xa0 && (x) < 0xaa) || (x) == 0xaf)
+#define IS_NVAAF(x) ((x) >= 0xaa && (x) <= 0xac)
+
+/*
+ * This code deals with PGRAPH contexts on NV50 family cards. Like NV40, it's
+ * the GPU itself that does context-switching, but it needs a special
+ * microcode to do it. And it's the driver's task to supply this microcode,
+ * further known as ctxprog, as well as the initial context values, known
+ * as ctxvals.
+ *
+ * Without ctxprog, you cannot switch contexts. Not even in software, since
+ * the majority of context [xfer strands] isn't accessible directly. You're
+ * stuck with a single channel, and you also suffer all the problems resulting
+ * from missing ctxvals, since you cannot load them.
+ *
+ * Without ctxvals, you're stuck with PGRAPH's default context. It's enough to
+ * run 2d operations, but trying to utilise 3d or CUDA will just lock you up,
+ * since you don't have... some sort of needed setup.
+ *
+ * Nouveau will just disable acceleration if not given ctxprog + ctxvals, since
+ * it's too much hassle to handle no-ctxprog as a special case.
+ */
+
+/*
+ * How ctxprogs work.
+ *
+ * The ctxprog is written in its own kind of microcode, with very small and
+ * crappy set of available commands. You upload it to a small [512 insns]
+ * area of memory on PGRAPH, and it'll be run when PFIFO wants PGRAPH to
+ * switch channel. or when the driver explicitely requests it. Stuff visible
+ * to ctxprog consists of: PGRAPH MMIO registers, PGRAPH context strands,
+ * the per-channel context save area in VRAM [known as ctxvals or grctx],
+ * 4 flags registers, a scratch register, two grctx pointers, plus many
+ * random poorly-understood details.
+ *
+ * When ctxprog runs, it's supposed to check what operations are asked of it,
+ * save old context if requested, optionally reset PGRAPH and switch to the
+ * new channel, and load the new context. Context consists of three major
+ * parts: subset of MMIO registers and two "xfer areas".
+ */
+
+/* TODO:
+ *  - document unimplemented bits compared to nvidia
+ *  - NVAx: make a TP subroutine, use it.
+ *  - use 0x4008fc instead of 0x1540?
+ */
+
+enum cp_label {
+	cp_check_load = 1,
+	cp_setup_auto_load,
+	cp_setup_load,
+	cp_setup_save,
+	cp_swap_state,
+	cp_prepare_exit,
+	cp_exit,
+};
+
+static void nv50_graph_construct_mmio(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_xfer1(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_xfer2(struct nouveau_grctx *ctx);
+
+/* Main function: construct the ctxprog skeleton, call the other functions. */
+
+static int
+nv50_grctx_generate(struct nouveau_grctx *ctx)
+{
+	cp_set (ctx, STATE, RUNNING);
+	cp_set (ctx, XFER_SWITCH, ENABLE);
+	/* decide whether we're loading/unloading the context */
+	cp_bra (ctx, AUTO_SAVE, PENDING, cp_setup_save);
+	cp_bra (ctx, USER_SAVE, PENDING, cp_setup_save);
+
+	cp_name(ctx, cp_check_load);
+	cp_bra (ctx, AUTO_LOAD, PENDING, cp_setup_auto_load);
+	cp_bra (ctx, USER_LOAD, PENDING, cp_setup_load);
+	cp_bra (ctx, ALWAYS, TRUE, cp_prepare_exit);
+
+	/* setup for context load */
+	cp_name(ctx, cp_setup_auto_load);
+	cp_out (ctx, CP_DISABLE1);
+	cp_out (ctx, CP_DISABLE2);
+	cp_out (ctx, CP_ENABLE);
+	cp_out (ctx, CP_NEXT_TO_SWAP);
+	cp_set (ctx, UNK01, SET);
+	cp_name(ctx, cp_setup_load);
+	cp_out (ctx, CP_NEWCTX);
+	cp_wait(ctx, NEWCTX, BUSY);
+	cp_set (ctx, UNK1D, CLEAR);
+	cp_set (ctx, SWAP_DIRECTION, LOAD);
+	cp_bra (ctx, UNK0B, SET, cp_prepare_exit);
+	cp_bra (ctx, ALWAYS, TRUE, cp_swap_state);
+
+	/* setup for context save */
+	cp_name(ctx, cp_setup_save);
+	cp_set (ctx, UNK1D, SET);
+	cp_wait(ctx, STATUS, BUSY);
+	cp_wait(ctx, INTR, PENDING);
+	cp_bra (ctx, STATUS, BUSY, cp_setup_save);
+	cp_set (ctx, UNK01, SET);
+	cp_set (ctx, SWAP_DIRECTION, SAVE);
+
+	/* general PGRAPH state */
+	cp_name(ctx, cp_swap_state);
+	cp_set (ctx, UNK03, SET);
+	cp_pos (ctx, 0x00004/4);
+	cp_ctx (ctx, 0x400828, 1); /* needed. otherwise, flickering happens. */
+	cp_pos (ctx, 0x00100/4);
+	nv50_graph_construct_mmio(ctx);
+	nv50_graph_construct_xfer1(ctx);
+	nv50_graph_construct_xfer2(ctx);
+
+	cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_check_load);
+
+	cp_set (ctx, UNK20, SET);
+	cp_set (ctx, SWAP_DIRECTION, SAVE); /* no idea why this is needed, but fixes at least one lockup. */
+	cp_lsr (ctx, ctx->ctxvals_base);
+	cp_out (ctx, CP_SET_XFER_POINTER);
+	cp_lsr (ctx, 4);
+	cp_out (ctx, CP_SEEK_1);
+	cp_out (ctx, CP_XFER_1);
+	cp_wait(ctx, XFER, BUSY);
+
+	/* pre-exit state updates */
+	cp_name(ctx, cp_prepare_exit);
+	cp_set (ctx, UNK01, CLEAR);
+	cp_set (ctx, UNK03, CLEAR);
+	cp_set (ctx, UNK1D, CLEAR);
+
+	cp_bra (ctx, USER_SAVE, PENDING, cp_exit);
+	cp_out (ctx, CP_NEXT_TO_CURRENT);
+
+	cp_name(ctx, cp_exit);
+	cp_set (ctx, USER_SAVE, NOT_PENDING);
+	cp_set (ctx, USER_LOAD, NOT_PENDING);
+	cp_set (ctx, XFER_SWITCH, DISABLE);
+	cp_set (ctx, STATE, STOPPED);
+	cp_out (ctx, CP_END);
+	ctx->ctxvals_pos += 0x400; /* padding... no idea why you need it */
+
+	return 0;
+}
+
+void
+nv50_grctx_fill(struct nouveau_device *device, struct nouveau_gpuobj *mem)
+{
+	nv50_grctx_generate(&(struct nouveau_grctx) {
+			     .device = device,
+			     .mode = NOUVEAU_GRCTX_VALS,
+			     .data = mem,
+			   });
+}
+
+int
+nv50_grctx_init(struct nouveau_device *device, u32 *size)
+{
+	u32 *ctxprog = kmalloc(512 * 4, GFP_KERNEL), i;
+	struct nouveau_grctx ctx = {
+		.device = device,
+		.mode = NOUVEAU_GRCTX_PROG,
+		.data = ctxprog,
+		.ctxprog_max = 512,
+	};
+
+	if (!ctxprog)
+		return -ENOMEM;
+	nv50_grctx_generate(&ctx);
+
+	nv_wr32(device, 0x400324, 0);
+	for (i = 0; i < ctx.ctxprog_len; i++)
+		nv_wr32(device, 0x400328, ctxprog[i]);
+	*size = ctx.ctxvals_pos * 4;
+	kfree(ctxprog);
+	return 0;
+}
+
+/*
+ * Constructs MMIO part of ctxprog and ctxvals. Just a matter of knowing which
+ * registers to save/restore and the default values for them.
+ */
+
+static void
+nv50_graph_construct_mmio_ddata(struct nouveau_grctx *ctx);
+
+static void
+nv50_graph_construct_mmio(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int i, j;
+	int offset, base;
+	u32 units = nv_rd32 (ctx->device, 0x1540);
+
+	/* 0800: DISPATCH */
+	cp_ctx(ctx, 0x400808, 7);
+	gr_def(ctx, 0x400814, 0x00000030);
+	cp_ctx(ctx, 0x400834, 0x32);
+	if (device->chipset == 0x50) {
+		gr_def(ctx, 0x400834, 0xff400040);
+		gr_def(ctx, 0x400838, 0xfff00080);
+		gr_def(ctx, 0x40083c, 0xfff70090);
+		gr_def(ctx, 0x400840, 0xffe806a8);
+	}
+	gr_def(ctx, 0x400844, 0x00000002);
+	if (IS_NVA3F(device->chipset))
+		gr_def(ctx, 0x400894, 0x00001000);
+	gr_def(ctx, 0x4008e8, 0x00000003);
+	gr_def(ctx, 0x4008ec, 0x00001000);
+	if (device->chipset == 0x50)
+		cp_ctx(ctx, 0x400908, 0xb);
+	else if (device->chipset < 0xa0)
+		cp_ctx(ctx, 0x400908, 0xc);
+	else
+		cp_ctx(ctx, 0x400908, 0xe);
+
+	if (device->chipset >= 0xa0)
+		cp_ctx(ctx, 0x400b00, 0x1);
+	if (IS_NVA3F(device->chipset)) {
+		cp_ctx(ctx, 0x400b10, 0x1);
+		gr_def(ctx, 0x400b10, 0x0001629d);
+		cp_ctx(ctx, 0x400b20, 0x1);
+		gr_def(ctx, 0x400b20, 0x0001629d);
+	}
+
+	nv50_graph_construct_mmio_ddata(ctx);
+
+	/* 0C00: VFETCH */
+	cp_ctx(ctx, 0x400c08, 0x2);
+	gr_def(ctx, 0x400c08, 0x0000fe0c);
+
+	/* 1000 */
+	if (device->chipset < 0xa0) {
+		cp_ctx(ctx, 0x401008, 0x4);
+		gr_def(ctx, 0x401014, 0x00001000);
+	} else if (!IS_NVA3F(device->chipset)) {
+		cp_ctx(ctx, 0x401008, 0x5);
+		gr_def(ctx, 0x401018, 0x00001000);
+	} else {
+		cp_ctx(ctx, 0x401008, 0x5);
+		gr_def(ctx, 0x401018, 0x00004000);
+	}
+
+	/* 1400 */
+	cp_ctx(ctx, 0x401400, 0x8);
+	cp_ctx(ctx, 0x401424, 0x3);
+	if (device->chipset == 0x50)
+		gr_def(ctx, 0x40142c, 0x0001fd87);
+	else
+		gr_def(ctx, 0x40142c, 0x00000187);
+	cp_ctx(ctx, 0x401540, 0x5);
+	gr_def(ctx, 0x401550, 0x00001018);
+
+	/* 1800: STREAMOUT */
+	cp_ctx(ctx, 0x401814, 0x1);
+	gr_def(ctx, 0x401814, 0x000000ff);
+	if (device->chipset == 0x50) {
+		cp_ctx(ctx, 0x40181c, 0xe);
+		gr_def(ctx, 0x401850, 0x00000004);
+	} else if (device->chipset < 0xa0) {
+		cp_ctx(ctx, 0x40181c, 0xf);
+		gr_def(ctx, 0x401854, 0x00000004);
+	} else {
+		cp_ctx(ctx, 0x40181c, 0x13);
+		gr_def(ctx, 0x401864, 0x00000004);
+	}
+
+	/* 1C00 */
+	cp_ctx(ctx, 0x401c00, 0x1);
+	switch (device->chipset) {
+	case 0x50:
+		gr_def(ctx, 0x401c00, 0x0001005f);
+		break;
+	case 0x84:
+	case 0x86:
+	case 0x94:
+		gr_def(ctx, 0x401c00, 0x044d00df);
+		break;
+	case 0x92:
+	case 0x96:
+	case 0x98:
+	case 0xa0:
+	case 0xaa:
+	case 0xac:
+		gr_def(ctx, 0x401c00, 0x042500df);
+		break;
+	case 0xa3:
+	case 0xa5:
+	case 0xa8:
+	case 0xaf:
+		gr_def(ctx, 0x401c00, 0x142500df);
+		break;
+	}
+
+	/* 2000 */
+
+	/* 2400 */
+	cp_ctx(ctx, 0x402400, 0x1);
+	if (device->chipset == 0x50)
+		cp_ctx(ctx, 0x402408, 0x1);
+	else
+		cp_ctx(ctx, 0x402408, 0x2);
+	gr_def(ctx, 0x402408, 0x00000600);
+
+	/* 2800: CSCHED */
+	cp_ctx(ctx, 0x402800, 0x1);
+	if (device->chipset == 0x50)
+		gr_def(ctx, 0x402800, 0x00000006);
+
+	/* 2C00: ZCULL */
+	cp_ctx(ctx, 0x402c08, 0x6);
+	if (device->chipset != 0x50)
+		gr_def(ctx, 0x402c14, 0x01000000);
+	gr_def(ctx, 0x402c18, 0x000000ff);
+	if (device->chipset == 0x50)
+		cp_ctx(ctx, 0x402ca0, 0x1);
+	else
+		cp_ctx(ctx, 0x402ca0, 0x2);
+	if (device->chipset < 0xa0)
+		gr_def(ctx, 0x402ca0, 0x00000400);
+	else if (!IS_NVA3F(device->chipset))
+		gr_def(ctx, 0x402ca0, 0x00000800);
+	else
+		gr_def(ctx, 0x402ca0, 0x00000400);
+	cp_ctx(ctx, 0x402cac, 0x4);
+
+	/* 3000: ENG2D */
+	cp_ctx(ctx, 0x403004, 0x1);
+	gr_def(ctx, 0x403004, 0x00000001);
+
+	/* 3400 */
+	if (device->chipset >= 0xa0) {
+		cp_ctx(ctx, 0x403404, 0x1);
+		gr_def(ctx, 0x403404, 0x00000001);
+	}
+
+	/* 5000: CCACHE */
+	cp_ctx(ctx, 0x405000, 0x1);
+	switch (device->chipset) {
+	case 0x50:
+		gr_def(ctx, 0x405000, 0x00300080);
+		break;
+	case 0x84:
+	case 0xa0:
+	case 0xa3:
+	case 0xa5:
+	case 0xa8:
+	case 0xaa:
+	case 0xac:
+	case 0xaf:
+		gr_def(ctx, 0x405000, 0x000e0080);
+		break;
+	case 0x86:
+	case 0x92:
+	case 0x94:
+	case 0x96:
+	case 0x98:
+		gr_def(ctx, 0x405000, 0x00000080);
+		break;
+	}
+	cp_ctx(ctx, 0x405014, 0x1);
+	gr_def(ctx, 0x405014, 0x00000004);
+	cp_ctx(ctx, 0x40501c, 0x1);
+	cp_ctx(ctx, 0x405024, 0x1);
+	cp_ctx(ctx, 0x40502c, 0x1);
+
+	/* 6000? */
+	if (device->chipset == 0x50)
+		cp_ctx(ctx, 0x4063e0, 0x1);
+
+	/* 6800: M2MF */
+	if (device->chipset < 0x90) {
+		cp_ctx(ctx, 0x406814, 0x2b);
+		gr_def(ctx, 0x406818, 0x00000f80);
+		gr_def(ctx, 0x406860, 0x007f0080);
+		gr_def(ctx, 0x40689c, 0x007f0080);
+	} else {
+		cp_ctx(ctx, 0x406814, 0x4);
+		if (device->chipset == 0x98)
+			gr_def(ctx, 0x406818, 0x00000f80);
+		else
+			gr_def(ctx, 0x406818, 0x00001f80);
+		if (IS_NVA3F(device->chipset))
+			gr_def(ctx, 0x40681c, 0x00000030);
+		cp_ctx(ctx, 0x406830, 0x3);
+	}
+
+	/* 7000: per-ROP group state */
+	for (i = 0; i < 8; i++) {
+		if (units & (1<<(i+16))) {
+			cp_ctx(ctx, 0x407000 + (i<<8), 3);
+			if (device->chipset == 0x50)
+				gr_def(ctx, 0x407000 + (i<<8), 0x1b74f820);
+			else if (device->chipset != 0xa5)
+				gr_def(ctx, 0x407000 + (i<<8), 0x3b74f821);
+			else
+				gr_def(ctx, 0x407000 + (i<<8), 0x7b74f821);
+			gr_def(ctx, 0x407004 + (i<<8), 0x89058001);
+
+			if (device->chipset == 0x50) {
+				cp_ctx(ctx, 0x407010 + (i<<8), 1);
+			} else if (device->chipset < 0xa0) {
+				cp_ctx(ctx, 0x407010 + (i<<8), 2);
+				gr_def(ctx, 0x407010 + (i<<8), 0x00001000);
+				gr_def(ctx, 0x407014 + (i<<8), 0x0000001f);
+			} else {
+				cp_ctx(ctx, 0x407010 + (i<<8), 3);
+				gr_def(ctx, 0x407010 + (i<<8), 0x00001000);
+				if (device->chipset != 0xa5)
+					gr_def(ctx, 0x407014 + (i<<8), 0x000000ff);
+				else
+					gr_def(ctx, 0x407014 + (i<<8), 0x000001ff);
+			}
+
+			cp_ctx(ctx, 0x407080 + (i<<8), 4);
+			if (device->chipset != 0xa5)
+				gr_def(ctx, 0x407080 + (i<<8), 0x027c10fa);
+			else
+				gr_def(ctx, 0x407080 + (i<<8), 0x827c10fa);
+			if (device->chipset == 0x50)
+				gr_def(ctx, 0x407084 + (i<<8), 0x000000c0);
+			else
+				gr_def(ctx, 0x407084 + (i<<8), 0x400000c0);
+			gr_def(ctx, 0x407088 + (i<<8), 0xb7892080);
+
+			if (device->chipset < 0xa0)
+				cp_ctx(ctx, 0x407094 + (i<<8), 1);
+			else if (!IS_NVA3F(device->chipset))
+				cp_ctx(ctx, 0x407094 + (i<<8), 3);
+			else {
+				cp_ctx(ctx, 0x407094 + (i<<8), 4);
+				gr_def(ctx, 0x4070a0 + (i<<8), 1);
+			}
+		}
+	}
+
+	cp_ctx(ctx, 0x407c00, 0x3);
+	if (device->chipset < 0x90)
+		gr_def(ctx, 0x407c00, 0x00010040);
+	else if (device->chipset < 0xa0)
+		gr_def(ctx, 0x407c00, 0x00390040);
+	else
+		gr_def(ctx, 0x407c00, 0x003d0040);
+	gr_def(ctx, 0x407c08, 0x00000022);
+	if (device->chipset >= 0xa0) {
+		cp_ctx(ctx, 0x407c10, 0x3);
+		cp_ctx(ctx, 0x407c20, 0x1);
+		cp_ctx(ctx, 0x407c2c, 0x1);
+	}
+
+	if (device->chipset < 0xa0) {
+		cp_ctx(ctx, 0x407d00, 0x9);
+	} else {
+		cp_ctx(ctx, 0x407d00, 0x15);
+	}
+	if (device->chipset == 0x98)
+		gr_def(ctx, 0x407d08, 0x00380040);
+	else {
+		if (device->chipset < 0x90)
+			gr_def(ctx, 0x407d08, 0x00010040);
+		else if (device->chipset < 0xa0)
+			gr_def(ctx, 0x407d08, 0x00390040);
+		else
+			gr_def(ctx, 0x407d08, 0x003d0040);
+		gr_def(ctx, 0x407d0c, 0x00000022);
+	}
+
+	/* 8000+: per-TP state */
+	for (i = 0; i < 10; i++) {
+		if (units & (1<<i)) {
+			if (device->chipset < 0xa0)
+				base = 0x408000 + (i<<12);
+			else
+				base = 0x408000 + (i<<11);
+			if (device->chipset < 0xa0)
+				offset = base + 0xc00;
+			else
+				offset = base + 0x80;
+			cp_ctx(ctx, offset + 0x00, 1);
+			gr_def(ctx, offset + 0x00, 0x0000ff0a);
+			cp_ctx(ctx, offset + 0x08, 1);
+
+			/* per-MP state */
+			for (j = 0; j < (device->chipset < 0xa0 ? 2 : 4); j++) {
+				if (!(units & (1 << (j+24)))) continue;
+				if (device->chipset < 0xa0)
+					offset = base + 0x200 + (j<<7);
+				else
+					offset = base + 0x100 + (j<<7);
+				cp_ctx(ctx, offset, 0x20);
+				gr_def(ctx, offset + 0x00, 0x01800000);
+				gr_def(ctx, offset + 0x04, 0x00160000);
+				gr_def(ctx, offset + 0x08, 0x01800000);
+				gr_def(ctx, offset + 0x18, 0x0003ffff);
+				switch (device->chipset) {
+				case 0x50:
+					gr_def(ctx, offset + 0x1c, 0x00080000);
+					break;
+				case 0x84:
+					gr_def(ctx, offset + 0x1c, 0x00880000);
+					break;
+				case 0x86:
+					gr_def(ctx, offset + 0x1c, 0x018c0000);
+					break;
+				case 0x92:
+				case 0x96:
+				case 0x98:
+					gr_def(ctx, offset + 0x1c, 0x118c0000);
+					break;
+				case 0x94:
+					gr_def(ctx, offset + 0x1c, 0x10880000);
+					break;
+				case 0xa0:
+				case 0xa5:
+					gr_def(ctx, offset + 0x1c, 0x310c0000);
+					break;
+				case 0xa3:
+				case 0xa8:
+				case 0xaa:
+				case 0xac:
+				case 0xaf:
+					gr_def(ctx, offset + 0x1c, 0x300c0000);
+					break;
+				}
+				gr_def(ctx, offset + 0x40, 0x00010401);
+				if (device->chipset == 0x50)
+					gr_def(ctx, offset + 0x48, 0x00000040);
+				else
+					gr_def(ctx, offset + 0x48, 0x00000078);
+				gr_def(ctx, offset + 0x50, 0x000000bf);
+				gr_def(ctx, offset + 0x58, 0x00001210);
+				if (device->chipset == 0x50)
+					gr_def(ctx, offset + 0x5c, 0x00000080);
+				else
+					gr_def(ctx, offset + 0x5c, 0x08000080);
+				if (device->chipset >= 0xa0)
+					gr_def(ctx, offset + 0x68, 0x0000003e);
+			}
+
+			if (device->chipset < 0xa0)
+				cp_ctx(ctx, base + 0x300, 0x4);
+			else
+				cp_ctx(ctx, base + 0x300, 0x5);
+			if (device->chipset == 0x50)
+				gr_def(ctx, base + 0x304, 0x00007070);
+			else if (device->chipset < 0xa0)
+				gr_def(ctx, base + 0x304, 0x00027070);
+			else if (!IS_NVA3F(device->chipset))
+				gr_def(ctx, base + 0x304, 0x01127070);
+			else
+				gr_def(ctx, base + 0x304, 0x05127070);
+
+			if (device->chipset < 0xa0)
+				cp_ctx(ctx, base + 0x318, 1);
+			else
+				cp_ctx(ctx, base + 0x320, 1);
+			if (device->chipset == 0x50)
+				gr_def(ctx, base + 0x318, 0x0003ffff);
+			else if (device->chipset < 0xa0)
+				gr_def(ctx, base + 0x318, 0x03ffffff);
+			else
+				gr_def(ctx, base + 0x320, 0x07ffffff);
+
+			if (device->chipset < 0xa0)
+				cp_ctx(ctx, base + 0x324, 5);
+			else
+				cp_ctx(ctx, base + 0x328, 4);
+
+			if (device->chipset < 0xa0) {
+				cp_ctx(ctx, base + 0x340, 9);
+				offset = base + 0x340;
+			} else if (!IS_NVA3F(device->chipset)) {
+				cp_ctx(ctx, base + 0x33c, 0xb);
+				offset = base + 0x344;
+			} else {
+				cp_ctx(ctx, base + 0x33c, 0xd);
+				offset = base + 0x344;
+			}
+			gr_def(ctx, offset + 0x0, 0x00120407);
+			gr_def(ctx, offset + 0x4, 0x05091507);
+			if (device->chipset == 0x84)
+				gr_def(ctx, offset + 0x8, 0x05100202);
+			else
+				gr_def(ctx, offset + 0x8, 0x05010202);
+			gr_def(ctx, offset + 0xc, 0x00030201);
+			if (device->chipset == 0xa3)
+				cp_ctx(ctx, base + 0x36c, 1);
+
+			cp_ctx(ctx, base + 0x400, 2);
+			gr_def(ctx, base + 0x404, 0x00000040);
+			cp_ctx(ctx, base + 0x40c, 2);
+			gr_def(ctx, base + 0x40c, 0x0d0c0b0a);
+			gr_def(ctx, base + 0x410, 0x00141210);
+
+			if (device->chipset < 0xa0)
+				offset = base + 0x800;
+			else
+				offset = base + 0x500;
+			cp_ctx(ctx, offset, 6);
+			gr_def(ctx, offset + 0x0, 0x000001f0);
+			gr_def(ctx, offset + 0x4, 0x00000001);
+			gr_def(ctx, offset + 0x8, 0x00000003);
+			if (device->chipset == 0x50 || IS_NVAAF(device->chipset))
+				gr_def(ctx, offset + 0xc, 0x00008000);
+			gr_def(ctx, offset + 0x14, 0x00039e00);
+			cp_ctx(ctx, offset + 0x1c, 2);
+			if (device->chipset == 0x50)
+				gr_def(ctx, offset + 0x1c, 0x00000040);
+			else
+				gr_def(ctx, offset + 0x1c, 0x00000100);
+			gr_def(ctx, offset + 0x20, 0x00003800);
+
+			if (device->chipset >= 0xa0) {
+				cp_ctx(ctx, base + 0x54c, 2);
+				if (!IS_NVA3F(device->chipset))
+					gr_def(ctx, base + 0x54c, 0x003fe006);
+				else
+					gr_def(ctx, base + 0x54c, 0x003fe007);
+				gr_def(ctx, base + 0x550, 0x003fe000);
+			}
+
+			if (device->chipset < 0xa0)
+				offset = base + 0xa00;
+			else
+				offset = base + 0x680;
+			cp_ctx(ctx, offset, 1);
+			gr_def(ctx, offset, 0x00404040);
+
+			if (device->chipset < 0xa0)
+				offset = base + 0xe00;
+			else
+				offset = base + 0x700;
+			cp_ctx(ctx, offset, 2);
+			if (device->chipset < 0xa0)
+				gr_def(ctx, offset, 0x0077f005);
+			else if (device->chipset == 0xa5)
+				gr_def(ctx, offset, 0x6cf7f007);
+			else if (device->chipset == 0xa8)
+				gr_def(ctx, offset, 0x6cfff007);
+			else if (device->chipset == 0xac)
+				gr_def(ctx, offset, 0x0cfff007);
+			else
+				gr_def(ctx, offset, 0x0cf7f007);
+			if (device->chipset == 0x50)
+				gr_def(ctx, offset + 0x4, 0x00007fff);
+			else if (device->chipset < 0xa0)
+				gr_def(ctx, offset + 0x4, 0x003f7fff);
+			else
+				gr_def(ctx, offset + 0x4, 0x02bf7fff);
+			cp_ctx(ctx, offset + 0x2c, 1);
+			if (device->chipset == 0x50) {
+				cp_ctx(ctx, offset + 0x50, 9);
+				gr_def(ctx, offset + 0x54, 0x000003ff);
+				gr_def(ctx, offset + 0x58, 0x00000003);
+				gr_def(ctx, offset + 0x5c, 0x00000003);
+				gr_def(ctx, offset + 0x60, 0x000001ff);
+				gr_def(ctx, offset + 0x64, 0x0000001f);
+				gr_def(ctx, offset + 0x68, 0x0000000f);
+				gr_def(ctx, offset + 0x6c, 0x0000000f);
+			} else if (device->chipset < 0xa0) {
+				cp_ctx(ctx, offset + 0x50, 1);
+				cp_ctx(ctx, offset + 0x70, 1);
+			} else {
+				cp_ctx(ctx, offset + 0x50, 1);
+				cp_ctx(ctx, offset + 0x60, 5);
+			}
+		}
+	}
+}
+
+static void
+dd_emit(struct nouveau_grctx *ctx, int num, u32 val) {
+	int i;
+	if (val && ctx->mode == NOUVEAU_GRCTX_VALS)
+		for (i = 0; i < num; i++)
+			nv_wo32(ctx->data, 4 * (ctx->ctxvals_pos + i), val);
+	ctx->ctxvals_pos += num;
+}
+
+static void
+nv50_graph_construct_mmio_ddata(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int base, num;
+	base = ctx->ctxvals_pos;
+
+	/* tesla state */
+	dd_emit(ctx, 1, 0);	/* 00000001 UNK0F90 */
+	dd_emit(ctx, 1, 0);	/* 00000001 UNK135C */
+
+	/* SRC_TIC state */
+	dd_emit(ctx, 1, 0);	/* 00000007 SRC_TILE_MODE_Z */
+	dd_emit(ctx, 1, 2);	/* 00000007 SRC_TILE_MODE_Y */
+	dd_emit(ctx, 1, 1);	/* 00000001 SRC_LINEAR #1 */
+	dd_emit(ctx, 1, 0);	/* 000000ff SRC_ADDRESS_HIGH */
+	dd_emit(ctx, 1, 0);	/* 00000001 SRC_SRGB */
+	if (device->chipset >= 0x94)
+		dd_emit(ctx, 1, 0);	/* 00000003 eng2d UNK0258 */
+	dd_emit(ctx, 1, 1);	/* 00000fff SRC_DEPTH */
+	dd_emit(ctx, 1, 0x100);	/* 0000ffff SRC_HEIGHT */
+
+	/* turing state */
+	dd_emit(ctx, 1, 0);		/* 0000000f TEXTURES_LOG2 */
+	dd_emit(ctx, 1, 0);		/* 0000000f SAMPLERS_LOG2 */
+	dd_emit(ctx, 1, 0);		/* 000000ff CB_DEF_ADDRESS_HIGH */
+	dd_emit(ctx, 1, 0);		/* ffffffff CB_DEF_ADDRESS_LOW */
+	dd_emit(ctx, 1, 0);		/* ffffffff SHARED_SIZE */
+	dd_emit(ctx, 1, 2);		/* ffffffff REG_MODE */
+	dd_emit(ctx, 1, 1);		/* 0000ffff BLOCK_ALLOC_THREADS */
+	dd_emit(ctx, 1, 1);		/* 00000001 LANES32 */
+	dd_emit(ctx, 1, 0);		/* 000000ff UNK370 */
+	dd_emit(ctx, 1, 0);		/* 000000ff USER_PARAM_UNK */
+	dd_emit(ctx, 1, 0);		/* 000000ff USER_PARAM_COUNT */
+	dd_emit(ctx, 1, 1);		/* 000000ff UNK384 bits 8-15 */
+	dd_emit(ctx, 1, 0x3fffff);	/* 003fffff TIC_LIMIT */
+	dd_emit(ctx, 1, 0x1fff);	/* 000fffff TSC_LIMIT */
+	dd_emit(ctx, 1, 0);		/* 0000ffff CB_ADDR_INDEX */
+	dd_emit(ctx, 1, 1);		/* 000007ff BLOCKDIM_X */
+	dd_emit(ctx, 1, 1);		/* 000007ff BLOCKDIM_XMY */
+	dd_emit(ctx, 1, 0);		/* 00000001 BLOCKDIM_XMY_OVERFLOW */
+	dd_emit(ctx, 1, 1);		/* 0003ffff BLOCKDIM_XMYMZ */
+	dd_emit(ctx, 1, 1);		/* 000007ff BLOCKDIM_Y */
+	dd_emit(ctx, 1, 1);		/* 0000007f BLOCKDIM_Z */
+	dd_emit(ctx, 1, 4);		/* 000000ff CP_REG_ALLOC_TEMP */
+	dd_emit(ctx, 1, 1);		/* 00000001 BLOCKDIM_DIRTY */
+	if (IS_NVA3F(device->chipset))
+		dd_emit(ctx, 1, 0);	/* 00000003 UNK03E8 */
+	dd_emit(ctx, 1, 1);		/* 0000007f BLOCK_ALLOC_HALFWARPS */
+	dd_emit(ctx, 1, 1);		/* 00000007 LOCAL_WARPS_NO_CLAMP */
+	dd_emit(ctx, 1, 7);		/* 00000007 LOCAL_WARPS_LOG_ALLOC */
+	dd_emit(ctx, 1, 1);		/* 00000007 STACK_WARPS_NO_CLAMP */
+	dd_emit(ctx, 1, 7);		/* 00000007 STACK_WARPS_LOG_ALLOC */
+	dd_emit(ctx, 1, 1);		/* 00001fff BLOCK_ALLOC_REGSLOTS_PACKED */
+	dd_emit(ctx, 1, 1);		/* 00001fff BLOCK_ALLOC_REGSLOTS_STRIDED */
+	dd_emit(ctx, 1, 1);		/* 000007ff BLOCK_ALLOC_THREADS */
+
+	/* compat 2d state */
+	if (device->chipset == 0x50) {
+		dd_emit(ctx, 4, 0);		/* 0000ffff clip X, Y, W, H */
+
+		dd_emit(ctx, 1, 1);		/* ffffffff chroma COLOR_FORMAT */
+
+		dd_emit(ctx, 1, 1);		/* ffffffff pattern COLOR_FORMAT */
+		dd_emit(ctx, 1, 0);		/* ffffffff pattern SHAPE */
+		dd_emit(ctx, 1, 1);		/* ffffffff pattern PATTERN_SELECT */
+
+		dd_emit(ctx, 1, 0xa);		/* ffffffff surf2d SRC_FORMAT */
+		dd_emit(ctx, 1, 0);		/* ffffffff surf2d DMA_SRC */
+		dd_emit(ctx, 1, 0);		/* 000000ff surf2d SRC_ADDRESS_HIGH */
+		dd_emit(ctx, 1, 0);		/* ffffffff surf2d SRC_ADDRESS_LOW */
+		dd_emit(ctx, 1, 0x40);		/* 0000ffff surf2d SRC_PITCH */
+		dd_emit(ctx, 1, 0);		/* 0000000f surf2d SRC_TILE_MODE_Z */
+		dd_emit(ctx, 1, 2);		/* 0000000f surf2d SRC_TILE_MODE_Y */
+		dd_emit(ctx, 1, 0x100);		/* ffffffff surf2d SRC_HEIGHT */
+		dd_emit(ctx, 1, 1);		/* 00000001 surf2d SRC_LINEAR */
+		dd_emit(ctx, 1, 0x100);		/* ffffffff surf2d SRC_WIDTH */
+
+		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect CLIP_B_X */
+		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect CLIP_B_Y */
+		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect CLIP_C_X */
+		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect CLIP_C_Y */
+		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect CLIP_D_X */
+		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect CLIP_D_Y */
+		dd_emit(ctx, 1, 1);		/* ffffffff gdirect COLOR_FORMAT */
+		dd_emit(ctx, 1, 0);		/* ffffffff gdirect OPERATION */
+		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect POINT_X */
+		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect POINT_Y */
+
+		dd_emit(ctx, 1, 0);		/* 0000ffff blit SRC_Y */
+		dd_emit(ctx, 1, 0);		/* ffffffff blit OPERATION */
+
+		dd_emit(ctx, 1, 0);		/* ffffffff ifc OPERATION */
+
+		dd_emit(ctx, 1, 0);		/* ffffffff iifc INDEX_FORMAT */
+		dd_emit(ctx, 1, 0);		/* ffffffff iifc LUT_OFFSET */
+		dd_emit(ctx, 1, 4);		/* ffffffff iifc COLOR_FORMAT */
+		dd_emit(ctx, 1, 0);		/* ffffffff iifc OPERATION */
+	}
+
+	/* m2mf state */
+	dd_emit(ctx, 1, 0);		/* ffffffff m2mf LINE_COUNT */
+	dd_emit(ctx, 1, 0);		/* ffffffff m2mf LINE_LENGTH_IN */
+	dd_emit(ctx, 2, 0);		/* ffffffff m2mf OFFSET_IN, OFFSET_OUT */
+	dd_emit(ctx, 1, 1);		/* ffffffff m2mf TILING_DEPTH_OUT */
+	dd_emit(ctx, 1, 0x100);		/* ffffffff m2mf TILING_HEIGHT_OUT */
+	dd_emit(ctx, 1, 0);		/* ffffffff m2mf TILING_POSITION_OUT_Z */
+	dd_emit(ctx, 1, 1);		/* 00000001 m2mf LINEAR_OUT */
+	dd_emit(ctx, 2, 0);		/* 0000ffff m2mf TILING_POSITION_OUT_X, Y */
+	dd_emit(ctx, 1, 0x100);		/* ffffffff m2mf TILING_PITCH_OUT */
+	dd_emit(ctx, 1, 1);		/* ffffffff m2mf TILING_DEPTH_IN */
+	dd_emit(ctx, 1, 0x100);		/* ffffffff m2mf TILING_HEIGHT_IN */
+	dd_emit(ctx, 1, 0);		/* ffffffff m2mf TILING_POSITION_IN_Z */
+	dd_emit(ctx, 1, 1);		/* 00000001 m2mf LINEAR_IN */
+	dd_emit(ctx, 2, 0);		/* 0000ffff m2mf TILING_POSITION_IN_X, Y */
+	dd_emit(ctx, 1, 0x100);		/* ffffffff m2mf TILING_PITCH_IN */
+
+	/* more compat 2d state */
+	if (device->chipset == 0x50) {
+		dd_emit(ctx, 1, 1);		/* ffffffff line COLOR_FORMAT */
+		dd_emit(ctx, 1, 0);		/* ffffffff line OPERATION */
+
+		dd_emit(ctx, 1, 1);		/* ffffffff triangle COLOR_FORMAT */
+		dd_emit(ctx, 1, 0);		/* ffffffff triangle OPERATION */
+
+		dd_emit(ctx, 1, 0);		/* 0000000f sifm TILE_MODE_Z */
+		dd_emit(ctx, 1, 2);		/* 0000000f sifm TILE_MODE_Y */
+		dd_emit(ctx, 1, 0);		/* 000000ff sifm FORMAT_FILTER */
+		dd_emit(ctx, 1, 1);		/* 000000ff sifm FORMAT_ORIGIN */
+		dd_emit(ctx, 1, 0);		/* 0000ffff sifm SRC_PITCH */
+		dd_emit(ctx, 1, 1);		/* 00000001 sifm SRC_LINEAR */
+		dd_emit(ctx, 1, 0);		/* 000000ff sifm SRC_OFFSET_HIGH */
+		dd_emit(ctx, 1, 0);		/* ffffffff sifm SRC_OFFSET */
+		dd_emit(ctx, 1, 0);		/* 0000ffff sifm SRC_HEIGHT */
+		dd_emit(ctx, 1, 0);		/* 0000ffff sifm SRC_WIDTH */
+		dd_emit(ctx, 1, 3);		/* ffffffff sifm COLOR_FORMAT */
+		dd_emit(ctx, 1, 0);		/* ffffffff sifm OPERATION */
+
+		dd_emit(ctx, 1, 0);		/* ffffffff sifc OPERATION */
+	}
+
+	/* tesla state */
+	dd_emit(ctx, 1, 0);		/* 0000000f GP_TEXTURES_LOG2 */
+	dd_emit(ctx, 1, 0);		/* 0000000f GP_SAMPLERS_LOG2 */
+	dd_emit(ctx, 1, 0);		/* 000000ff */
+	dd_emit(ctx, 1, 0);		/* ffffffff */
+	dd_emit(ctx, 1, 4);		/* 000000ff UNK12B0_0 */
+	dd_emit(ctx, 1, 0x70);		/* 000000ff UNK12B0_1 */
+	dd_emit(ctx, 1, 0x80);		/* 000000ff UNK12B0_3 */
+	dd_emit(ctx, 1, 0);		/* 000000ff UNK12B0_2 */
+	dd_emit(ctx, 1, 0);		/* 0000000f FP_TEXTURES_LOG2 */
+	dd_emit(ctx, 1, 0);		/* 0000000f FP_SAMPLERS_LOG2 */
+	if (IS_NVA3F(device->chipset)) {
+		dd_emit(ctx, 1, 0);	/* ffffffff */
+		dd_emit(ctx, 1, 0);	/* 0000007f MULTISAMPLE_SAMPLES_LOG2 */
+	} else {
+		dd_emit(ctx, 1, 0);	/* 0000000f MULTISAMPLE_SAMPLES_LOG2 */
+	}
+	dd_emit(ctx, 1, 0xc);		/* 000000ff SEMANTIC_COLOR.BFC0_ID */
+	if (device->chipset != 0x50)
+		dd_emit(ctx, 1, 0);	/* 00000001 SEMANTIC_COLOR.CLMP_EN */
+	dd_emit(ctx, 1, 8);		/* 000000ff SEMANTIC_COLOR.COLR_NR */
+	dd_emit(ctx, 1, 0x14);		/* 000000ff SEMANTIC_COLOR.FFC0_ID */
+	if (device->chipset == 0x50) {
+		dd_emit(ctx, 1, 0);	/* 000000ff SEMANTIC_LAYER */
+		dd_emit(ctx, 1, 0);	/* 00000001 */
+	} else {
+		dd_emit(ctx, 1, 0);	/* 00000001 SEMANTIC_PTSZ.ENABLE */
+		dd_emit(ctx, 1, 0x29);	/* 000000ff SEMANTIC_PTSZ.PTSZ_ID */
+		dd_emit(ctx, 1, 0x27);	/* 000000ff SEMANTIC_PRIM */
+		dd_emit(ctx, 1, 0x26);	/* 000000ff SEMANTIC_LAYER */
+		dd_emit(ctx, 1, 8);	/* 0000000f SMENATIC_CLIP.CLIP_HIGH */
+		dd_emit(ctx, 1, 4);	/* 000000ff SEMANTIC_CLIP.CLIP_LO */
+		dd_emit(ctx, 1, 0x27);	/* 000000ff UNK0FD4 */
+		dd_emit(ctx, 1, 0);	/* 00000001 UNK1900 */
+	}
+	dd_emit(ctx, 1, 0);		/* 00000007 RT_CONTROL_MAP0 */
+	dd_emit(ctx, 1, 1);		/* 00000007 RT_CONTROL_MAP1 */
+	dd_emit(ctx, 1, 2);		/* 00000007 RT_CONTROL_MAP2 */
+	dd_emit(ctx, 1, 3);		/* 00000007 RT_CONTROL_MAP3 */
+	dd_emit(ctx, 1, 4);		/* 00000007 RT_CONTROL_MAP4 */
+	dd_emit(ctx, 1, 5);		/* 00000007 RT_CONTROL_MAP5 */
+	dd_emit(ctx, 1, 6);		/* 00000007 RT_CONTROL_MAP6 */
+	dd_emit(ctx, 1, 7);		/* 00000007 RT_CONTROL_MAP7 */
+	dd_emit(ctx, 1, 1);		/* 0000000f RT_CONTROL_COUNT */
+	dd_emit(ctx, 8, 0);		/* 00000001 RT_HORIZ_UNK */
+	dd_emit(ctx, 8, 0);		/* ffffffff RT_ADDRESS_LOW */
+	dd_emit(ctx, 1, 0xcf);		/* 000000ff RT_FORMAT */
+	dd_emit(ctx, 7, 0);		/* 000000ff RT_FORMAT */
+	if (device->chipset != 0x50)
+		dd_emit(ctx, 3, 0);	/* 1, 1, 1 */
+	else
+		dd_emit(ctx, 2, 0);	/* 1, 1 */
+	dd_emit(ctx, 1, 0);		/* ffffffff GP_ENABLE */
+	dd_emit(ctx, 1, 0x80);		/* 0000ffff GP_VERTEX_OUTPUT_COUNT*/
+	dd_emit(ctx, 1, 4);		/* 000000ff GP_REG_ALLOC_RESULT */
+	dd_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
+	if (IS_NVA3F(device->chipset)) {
+		dd_emit(ctx, 1, 3);	/* 00000003 */
+		dd_emit(ctx, 1, 0);	/* 00000001 UNK1418. Alone. */
+	}
+	if (device->chipset != 0x50)
+		dd_emit(ctx, 1, 3);	/* 00000003 UNK15AC */
+	dd_emit(ctx, 1, 1);		/* ffffffff RASTERIZE_ENABLE */
+	dd_emit(ctx, 1, 0);		/* 00000001 FP_CONTROL.EXPORTS_Z */
+	if (device->chipset != 0x50)
+		dd_emit(ctx, 1, 0);	/* 00000001 FP_CONTROL.MULTIPLE_RESULTS */
+	dd_emit(ctx, 1, 0x12);		/* 000000ff FP_INTERPOLANT_CTRL.COUNT */
+	dd_emit(ctx, 1, 0x10);		/* 000000ff FP_INTERPOLANT_CTRL.COUNT_NONFLAT */
+	dd_emit(ctx, 1, 0xc);		/* 000000ff FP_INTERPOLANT_CTRL.OFFSET */
+	dd_emit(ctx, 1, 1);		/* 00000001 FP_INTERPOLANT_CTRL.UMASK.W */
+	dd_emit(ctx, 1, 0);		/* 00000001 FP_INTERPOLANT_CTRL.UMASK.X */
+	dd_emit(ctx, 1, 0);		/* 00000001 FP_INTERPOLANT_CTRL.UMASK.Y */
+	dd_emit(ctx, 1, 0);		/* 00000001 FP_INTERPOLANT_CTRL.UMASK.Z */
+	dd_emit(ctx, 1, 4);		/* 000000ff FP_RESULT_COUNT */
+	dd_emit(ctx, 1, 2);		/* ffffffff REG_MODE */
+	dd_emit(ctx, 1, 4);		/* 000000ff FP_REG_ALLOC_TEMP */
+	if (device->chipset >= 0xa0)
+		dd_emit(ctx, 1, 0);	/* ffffffff */
+	dd_emit(ctx, 1, 0);		/* 00000001 GP_BUILTIN_RESULT_EN.LAYER_IDX */
+	dd_emit(ctx, 1, 0);		/* ffffffff STRMOUT_ENABLE */
+	dd_emit(ctx, 1, 0x3fffff);	/* 003fffff TIC_LIMIT */
+	dd_emit(ctx, 1, 0x1fff);	/* 000fffff TSC_LIMIT */
+	dd_emit(ctx, 1, 0);		/* 00000001 VERTEX_TWO_SIDE_ENABLE*/
+	if (device->chipset != 0x50)
+		dd_emit(ctx, 8, 0);	/* 00000001 */
+	if (device->chipset >= 0xa0) {
+		dd_emit(ctx, 1, 1);	/* 00000007 VTX_ATTR_DEFINE.COMP */
+		dd_emit(ctx, 1, 1);	/* 00000007 VTX_ATTR_DEFINE.SIZE */
+		dd_emit(ctx, 1, 2);	/* 00000007 VTX_ATTR_DEFINE.TYPE */
+		dd_emit(ctx, 1, 0);	/* 000000ff VTX_ATTR_DEFINE.ATTR */
+	}
+	dd_emit(ctx, 1, 4);		/* 0000007f VP_RESULT_MAP_SIZE */
+	dd_emit(ctx, 1, 0x14);		/* 0000001f ZETA_FORMAT */
+	dd_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
+	dd_emit(ctx, 1, 0);		/* 0000000f VP_TEXTURES_LOG2 */
+	dd_emit(ctx, 1, 0);		/* 0000000f VP_SAMPLERS_LOG2 */
+	if (IS_NVA3F(device->chipset))
+		dd_emit(ctx, 1, 0);	/* 00000001 */
+	dd_emit(ctx, 1, 2);		/* 00000003 POLYGON_MODE_BACK */
+	if (device->chipset >= 0xa0)
+		dd_emit(ctx, 1, 0);	/* 00000003 VTX_ATTR_DEFINE.SIZE - 1 */
+	dd_emit(ctx, 1, 0);		/* 0000ffff CB_ADDR_INDEX */
+	if (device->chipset >= 0xa0)
+		dd_emit(ctx, 1, 0);	/* 00000003 */
+	dd_emit(ctx, 1, 0);		/* 00000001 CULL_FACE_ENABLE */
+	dd_emit(ctx, 1, 1);		/* 00000003 CULL_FACE */
+	dd_emit(ctx, 1, 0);		/* 00000001 FRONT_FACE */
+	dd_emit(ctx, 1, 2);		/* 00000003 POLYGON_MODE_FRONT */
+	dd_emit(ctx, 1, 0x1000);	/* 00007fff UNK141C */
+	if (device->chipset != 0x50) {
+		dd_emit(ctx, 1, 0xe00);		/* 7fff */
+		dd_emit(ctx, 1, 0x1000);	/* 7fff */
+		dd_emit(ctx, 1, 0x1e00);	/* 7fff */
+	}
+	dd_emit(ctx, 1, 0);		/* 00000001 BEGIN_END_ACTIVE */
+	dd_emit(ctx, 1, 1);		/* 00000001 POLYGON_MODE_??? */
+	dd_emit(ctx, 1, 1);		/* 000000ff GP_REG_ALLOC_TEMP / 4 rounded up */
+	dd_emit(ctx, 1, 1);		/* 000000ff FP_REG_ALLOC_TEMP... without /4? */
+	dd_emit(ctx, 1, 1);		/* 000000ff VP_REG_ALLOC_TEMP / 4 rounded up */
+	dd_emit(ctx, 1, 1);		/* 00000001 */
+	dd_emit(ctx, 1, 0);		/* 00000001 */
+	dd_emit(ctx, 1, 0);		/* 00000001 VTX_ATTR_MASK_UNK0 nonempty */
+	dd_emit(ctx, 1, 0);		/* 00000001 VTX_ATTR_MASK_UNK1 nonempty */
+	dd_emit(ctx, 1, 0x200);		/* 0003ffff GP_VERTEX_OUTPUT_COUNT*GP_REG_ALLOC_RESULT */
+	if (IS_NVA3F(device->chipset))
+		dd_emit(ctx, 1, 0x200);
+	dd_emit(ctx, 1, 0);		/* 00000001 */
+	if (device->chipset < 0xa0) {
+		dd_emit(ctx, 1, 1);	/* 00000001 */
+		dd_emit(ctx, 1, 0x70);	/* 000000ff */
+		dd_emit(ctx, 1, 0x80);	/* 000000ff */
+		dd_emit(ctx, 1, 0);	/* 000000ff */
+		dd_emit(ctx, 1, 0);	/* 00000001 */
+		dd_emit(ctx, 1, 1);	/* 00000001 */
+		dd_emit(ctx, 1, 0x70);	/* 000000ff */
+		dd_emit(ctx, 1, 0x80);	/* 000000ff */
+		dd_emit(ctx, 1, 0);	/* 000000ff */
+	} else {
+		dd_emit(ctx, 1, 1);	/* 00000001 */
+		dd_emit(ctx, 1, 0xf0);	/* 000000ff */
+		dd_emit(ctx, 1, 0xff);	/* 000000ff */
+		dd_emit(ctx, 1, 0);	/* 000000ff */
+		dd_emit(ctx, 1, 0);	/* 00000001 */
+		dd_emit(ctx, 1, 1);	/* 00000001 */
+		dd_emit(ctx, 1, 0xf0);	/* 000000ff */
+		dd_emit(ctx, 1, 0xff);	/* 000000ff */
+		dd_emit(ctx, 1, 0);	/* 000000ff */
+		dd_emit(ctx, 1, 9);	/* 0000003f UNK114C.COMP,SIZE */
+	}
+
+	/* eng2d state */
+	dd_emit(ctx, 1, 0);		/* 00000001 eng2d COLOR_KEY_ENABLE */
+	dd_emit(ctx, 1, 0);		/* 00000007 eng2d COLOR_KEY_FORMAT */
+	dd_emit(ctx, 1, 1);		/* ffffffff eng2d DST_DEPTH */
+	dd_emit(ctx, 1, 0xcf);		/* 000000ff eng2d DST_FORMAT */
+	dd_emit(ctx, 1, 0);		/* ffffffff eng2d DST_LAYER */
+	dd_emit(ctx, 1, 1);		/* 00000001 eng2d DST_LINEAR */
+	dd_emit(ctx, 1, 0);		/* 00000007 eng2d PATTERN_COLOR_FORMAT */
+	dd_emit(ctx, 1, 0);		/* 00000007 eng2d OPERATION */
+	dd_emit(ctx, 1, 0);		/* 00000003 eng2d PATTERN_SELECT */
+	dd_emit(ctx, 1, 0xcf);		/* 000000ff eng2d SIFC_FORMAT */
+	dd_emit(ctx, 1, 0);		/* 00000001 eng2d SIFC_BITMAP_ENABLE */
+	dd_emit(ctx, 1, 2);		/* 00000003 eng2d SIFC_BITMAP_UNK808 */
+	dd_emit(ctx, 1, 0);		/* ffffffff eng2d BLIT_DU_DX_FRACT */
+	dd_emit(ctx, 1, 1);		/* ffffffff eng2d BLIT_DU_DX_INT */
+	dd_emit(ctx, 1, 0);		/* ffffffff eng2d BLIT_DV_DY_FRACT */
+	dd_emit(ctx, 1, 1);		/* ffffffff eng2d BLIT_DV_DY_INT */
+	dd_emit(ctx, 1, 0);		/* 00000001 eng2d BLIT_CONTROL_FILTER */
+	dd_emit(ctx, 1, 0xcf);		/* 000000ff eng2d DRAW_COLOR_FORMAT */
+	dd_emit(ctx, 1, 0xcf);		/* 000000ff eng2d SRC_FORMAT */
+	dd_emit(ctx, 1, 1);		/* 00000001 eng2d SRC_LINEAR #2 */
+
+	num = ctx->ctxvals_pos - base;
+	ctx->ctxvals_pos = base;
+	if (IS_NVA3F(device->chipset))
+		cp_ctx(ctx, 0x404800, num);
+	else
+		cp_ctx(ctx, 0x405400, num);
+}
+
+/*
+ * xfer areas. These are a pain.
+ *
+ * There are 2 xfer areas: the first one is big and contains all sorts of
+ * stuff, the second is small and contains some per-TP context.
+ *
+ * Each area is split into 8 "strands". The areas, when saved to grctx,
+ * are made of 8-word blocks. Each block contains a single word from
+ * each strand. The strands are independent of each other, their
+ * addresses are unrelated to each other, and data in them is closely
+ * packed together. The strand layout varies a bit between cards: here
+ * and there, a single word is thrown out in the middle and the whole
+ * strand is offset by a bit from corresponding one on another chipset.
+ * For this reason, addresses of stuff in strands are almost useless.
+ * Knowing sequence of stuff and size of gaps between them is much more
+ * useful, and that's how we build the strands in our generator.
+ *
+ * NVA0 takes this mess to a whole new level by cutting the old strands
+ * into a few dozen pieces [known as genes], rearranging them randomly,
+ * and putting them back together to make new strands. Hopefully these
+ * genes correspond more or less directly to the same PGRAPH subunits
+ * as in 400040 register.
+ *
+ * The most common value in default context is 0, and when the genes
+ * are separated by 0's, gene bounduaries are quite speculative...
+ * some of them can be clearly deduced, others can be guessed, and yet
+ * others won't be resolved without figuring out the real meaning of
+ * given ctxval. For the same reason, ending point of each strand
+ * is unknown. Except for strand 0, which is the longest strand and
+ * its end corresponds to end of the whole xfer.
+ *
+ * An unsolved mystery is the seek instruction: it takes an argument
+ * in bits 8-18, and that argument is clearly the place in strands to
+ * seek to... but the offsets don't seem to correspond to offsets as
+ * seen in grctx. Perhaps there's another, real, not randomly-changing
+ * addressing in strands, and the xfer insn just happens to skip over
+ * the unused bits? NV10-NV30 PIPE comes to mind...
+ *
+ * As far as I know, there's no way to access the xfer areas directly
+ * without the help of ctxprog.
+ */
+
+static void
+xf_emit(struct nouveau_grctx *ctx, int num, u32 val) {
+	int i;
+	if (val && ctx->mode == NOUVEAU_GRCTX_VALS)
+		for (i = 0; i < num; i++)
+			nv_wo32(ctx->data, 4 * (ctx->ctxvals_pos + (i << 3)), val);
+	ctx->ctxvals_pos += num << 3;
+}
+
+/* Gene declarations... */
+
+static void nv50_graph_construct_gene_dispatch(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_m2mf(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_ccache(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk10xx(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk14xx(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_zcull(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_clipid(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk24xx(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_vfetch(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_eng2d(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_csched(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk1cxx(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_strmout(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk34xx(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_ropm1(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_ropm2(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_ropc(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_xfer_tp(struct nouveau_grctx *ctx);
+
+static void
+nv50_graph_construct_xfer1(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int i;
+	int offset;
+	int size = 0;
+	u32 units = nv_rd32 (ctx->device, 0x1540);
+
+	offset = (ctx->ctxvals_pos+0x3f)&~0x3f;
+	ctx->ctxvals_base = offset;
+
+	if (device->chipset < 0xa0) {
+		/* Strand 0 */
+		ctx->ctxvals_pos = offset;
+		nv50_graph_construct_gene_dispatch(ctx);
+		nv50_graph_construct_gene_m2mf(ctx);
+		nv50_graph_construct_gene_unk24xx(ctx);
+		nv50_graph_construct_gene_clipid(ctx);
+		nv50_graph_construct_gene_zcull(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 1 */
+		ctx->ctxvals_pos = offset + 0x1;
+		nv50_graph_construct_gene_vfetch(ctx);
+		nv50_graph_construct_gene_eng2d(ctx);
+		nv50_graph_construct_gene_csched(ctx);
+		nv50_graph_construct_gene_ropm1(ctx);
+		nv50_graph_construct_gene_ropm2(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 2 */
+		ctx->ctxvals_pos = offset + 0x2;
+		nv50_graph_construct_gene_ccache(ctx);
+		nv50_graph_construct_gene_unk1cxx(ctx);
+		nv50_graph_construct_gene_strmout(ctx);
+		nv50_graph_construct_gene_unk14xx(ctx);
+		nv50_graph_construct_gene_unk10xx(ctx);
+		nv50_graph_construct_gene_unk34xx(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 3: per-ROP group state */
+		ctx->ctxvals_pos = offset + 3;
+		for (i = 0; i < 6; i++)
+			if (units & (1 << (i + 16)))
+				nv50_graph_construct_gene_ropc(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strands 4-7: per-TP state */
+		for (i = 0; i < 4; i++) {
+			ctx->ctxvals_pos = offset + 4 + i;
+			if (units & (1 << (2 * i)))
+				nv50_graph_construct_xfer_tp(ctx);
+			if (units & (1 << (2 * i + 1)))
+				nv50_graph_construct_xfer_tp(ctx);
+			if ((ctx->ctxvals_pos-offset)/8 > size)
+				size = (ctx->ctxvals_pos-offset)/8;
+		}
+	} else {
+		/* Strand 0 */
+		ctx->ctxvals_pos = offset;
+		nv50_graph_construct_gene_dispatch(ctx);
+		nv50_graph_construct_gene_m2mf(ctx);
+		nv50_graph_construct_gene_unk34xx(ctx);
+		nv50_graph_construct_gene_csched(ctx);
+		nv50_graph_construct_gene_unk1cxx(ctx);
+		nv50_graph_construct_gene_strmout(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 1 */
+		ctx->ctxvals_pos = offset + 1;
+		nv50_graph_construct_gene_unk10xx(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 2 */
+		ctx->ctxvals_pos = offset + 2;
+		if (device->chipset == 0xa0)
+			nv50_graph_construct_gene_unk14xx(ctx);
+		nv50_graph_construct_gene_unk24xx(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 3 */
+		ctx->ctxvals_pos = offset + 3;
+		nv50_graph_construct_gene_vfetch(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 4 */
+		ctx->ctxvals_pos = offset + 4;
+		nv50_graph_construct_gene_ccache(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 5 */
+		ctx->ctxvals_pos = offset + 5;
+		nv50_graph_construct_gene_ropm2(ctx);
+		nv50_graph_construct_gene_ropm1(ctx);
+		/* per-ROP context */
+		for (i = 0; i < 8; i++)
+			if (units & (1<<(i+16)))
+				nv50_graph_construct_gene_ropc(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 6 */
+		ctx->ctxvals_pos = offset + 6;
+		nv50_graph_construct_gene_zcull(ctx);
+		nv50_graph_construct_gene_clipid(ctx);
+		nv50_graph_construct_gene_eng2d(ctx);
+		if (units & (1 << 0))
+			nv50_graph_construct_xfer_tp(ctx);
+		if (units & (1 << 1))
+			nv50_graph_construct_xfer_tp(ctx);
+		if (units & (1 << 2))
+			nv50_graph_construct_xfer_tp(ctx);
+		if (units & (1 << 3))
+			nv50_graph_construct_xfer_tp(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 7 */
+		ctx->ctxvals_pos = offset + 7;
+		if (device->chipset == 0xa0) {
+			if (units & (1 << 4))
+				nv50_graph_construct_xfer_tp(ctx);
+			if (units & (1 << 5))
+				nv50_graph_construct_xfer_tp(ctx);
+			if (units & (1 << 6))
+				nv50_graph_construct_xfer_tp(ctx);
+			if (units & (1 << 7))
+				nv50_graph_construct_xfer_tp(ctx);
+			if (units & (1 << 8))
+				nv50_graph_construct_xfer_tp(ctx);
+			if (units & (1 << 9))
+				nv50_graph_construct_xfer_tp(ctx);
+		} else {
+			nv50_graph_construct_gene_unk14xx(ctx);
+		}
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+	}
+
+	ctx->ctxvals_pos = offset + size * 8;
+	ctx->ctxvals_pos = (ctx->ctxvals_pos+0x3f)&~0x3f;
+	cp_lsr (ctx, offset);
+	cp_out (ctx, CP_SET_XFER_POINTER);
+	cp_lsr (ctx, size);
+	cp_out (ctx, CP_SEEK_1);
+	cp_out (ctx, CP_XFER_1);
+	cp_wait(ctx, XFER, BUSY);
+}
+
+/*
+ * non-trivial demagiced parts of ctx init go here
+ */
+
+static void
+nv50_graph_construct_gene_dispatch(struct nouveau_grctx *ctx)
+{
+	/* start of strand 0 */
+	struct nouveau_device *device = ctx->device;
+	/* SEEK */
+	if (device->chipset == 0x50)
+		xf_emit(ctx, 5, 0);
+	else if (!IS_NVA3F(device->chipset))
+		xf_emit(ctx, 6, 0);
+	else
+		xf_emit(ctx, 4, 0);
+	/* SEEK */
+	/* the PGRAPH's internal FIFO */
+	if (device->chipset == 0x50)
+		xf_emit(ctx, 8*3, 0);
+	else
+		xf_emit(ctx, 0x100*3, 0);
+	/* and another bonus slot?!? */
+	xf_emit(ctx, 3, 0);
+	/* and YET ANOTHER bonus slot? */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 3, 0);
+	/* SEEK */
+	/* CTX_SWITCH: caches of gr objects bound to subchannels. 8 values, last used index */
+	xf_emit(ctx, 9, 0);
+	/* SEEK */
+	xf_emit(ctx, 9, 0);
+	/* SEEK */
+	xf_emit(ctx, 9, 0);
+	/* SEEK */
+	xf_emit(ctx, 9, 0);
+	/* SEEK */
+	if (device->chipset < 0x90)
+		xf_emit(ctx, 4, 0);
+	/* SEEK */
+	xf_emit(ctx, 2, 0);
+	/* SEEK */
+	xf_emit(ctx, 6*2, 0);
+	xf_emit(ctx, 2, 0);
+	/* SEEK */
+	xf_emit(ctx, 2, 0);
+	/* SEEK */
+	xf_emit(ctx, 6*2, 0);
+	xf_emit(ctx, 2, 0);
+	/* SEEK */
+	if (device->chipset == 0x50)
+		xf_emit(ctx, 0x1c, 0);
+	else if (device->chipset < 0xa0)
+		xf_emit(ctx, 0x1e, 0);
+	else
+		xf_emit(ctx, 0x22, 0);
+	/* SEEK */
+	xf_emit(ctx, 0x15, 0);
+}
+
+static void
+nv50_graph_construct_gene_m2mf(struct nouveau_grctx *ctx)
+{
+	/* Strand 0, right after dispatch */
+	struct nouveau_device *device = ctx->device;
+	int smallm2mf = 0;
+	if (device->chipset < 0x92 || device->chipset == 0x98)
+		smallm2mf = 1;
+	/* SEEK */
+	xf_emit (ctx, 1, 0);		/* DMA_NOTIFY instance >> 4 */
+	xf_emit (ctx, 1, 0);		/* DMA_BUFFER_IN instance >> 4 */
+	xf_emit (ctx, 1, 0);		/* DMA_BUFFER_OUT instance >> 4 */
+	xf_emit (ctx, 1, 0);		/* OFFSET_IN */
+	xf_emit (ctx, 1, 0);		/* OFFSET_OUT */
+	xf_emit (ctx, 1, 0);		/* PITCH_IN */
+	xf_emit (ctx, 1, 0);		/* PITCH_OUT */
+	xf_emit (ctx, 1, 0);		/* LINE_LENGTH */
+	xf_emit (ctx, 1, 0);		/* LINE_COUNT */
+	xf_emit (ctx, 1, 0x21);		/* FORMAT: bits 0-4 INPUT_INC, bits 5-9 OUTPUT_INC */
+	xf_emit (ctx, 1, 1);		/* LINEAR_IN */
+	xf_emit (ctx, 1, 0x2);		/* TILING_MODE_IN: bits 0-2 y tiling, bits 3-5 z tiling */
+	xf_emit (ctx, 1, 0x100);	/* TILING_PITCH_IN */
+	xf_emit (ctx, 1, 0x100);	/* TILING_HEIGHT_IN */
+	xf_emit (ctx, 1, 1);		/* TILING_DEPTH_IN */
+	xf_emit (ctx, 1, 0);		/* TILING_POSITION_IN_Z */
+	xf_emit (ctx, 1, 0);		/* TILING_POSITION_IN */
+	xf_emit (ctx, 1, 1);		/* LINEAR_OUT */
+	xf_emit (ctx, 1, 0x2);		/* TILING_MODE_OUT: bits 0-2 y tiling, bits 3-5 z tiling */
+	xf_emit (ctx, 1, 0x100);	/* TILING_PITCH_OUT */
+	xf_emit (ctx, 1, 0x100);	/* TILING_HEIGHT_OUT */
+	xf_emit (ctx, 1, 1);		/* TILING_DEPTH_OUT */
+	xf_emit (ctx, 1, 0);		/* TILING_POSITION_OUT_Z */
+	xf_emit (ctx, 1, 0);		/* TILING_POSITION_OUT */
+	xf_emit (ctx, 1, 0);		/* OFFSET_IN_HIGH */
+	xf_emit (ctx, 1, 0);		/* OFFSET_OUT_HIGH */
+	/* SEEK */
+	if (smallm2mf)
+		xf_emit(ctx, 0x40, 0);	/* 20 * ffffffff, 3ffff */
+	else
+		xf_emit(ctx, 0x100, 0);	/* 80 * ffffffff, 3ffff */
+	xf_emit(ctx, 4, 0);		/* 1f/7f, 0, 1f/7f, 0 [1f for smallm2mf, 7f otherwise] */
+	/* SEEK */
+	if (smallm2mf)
+		xf_emit(ctx, 0x400, 0);	/* ffffffff */
+	else
+		xf_emit(ctx, 0x800, 0);	/* ffffffff */
+	xf_emit(ctx, 4, 0);		/* ff/1ff, 0, 0, 0 [ff for smallm2mf, 1ff otherwise] */
+	/* SEEK */
+	xf_emit(ctx, 0x40, 0);		/* 20 * bits ffffffff, 3ffff */
+	xf_emit(ctx, 0x6, 0);		/* 1f, 0, 1f, 0, 1f, 0 */
+}
+
+static void
+nv50_graph_construct_gene_ccache(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	xf_emit(ctx, 2, 0);		/* RO */
+	xf_emit(ctx, 0x800, 0);		/* ffffffff */
+	switch (device->chipset) {
+	case 0x50:
+	case 0x92:
+	case 0xa0:
+		xf_emit(ctx, 0x2b, 0);
+		break;
+	case 0x84:
+		xf_emit(ctx, 0x29, 0);
+		break;
+	case 0x94:
+	case 0x96:
+	case 0xa3:
+		xf_emit(ctx, 0x27, 0);
+		break;
+	case 0x86:
+	case 0x98:
+	case 0xa5:
+	case 0xa8:
+	case 0xaa:
+	case 0xac:
+	case 0xaf:
+		xf_emit(ctx, 0x25, 0);
+		break;
+	}
+	/* CB bindings, 0x80 of them. first word is address >> 8, second is
+	 * size >> 4 | valid << 24 */
+	xf_emit(ctx, 0x100, 0);		/* ffffffff CB_DEF */
+	xf_emit(ctx, 1, 0);		/* 0000007f CB_ADDR_BUFFER */
+	xf_emit(ctx, 1, 0);		/* 0 */
+	xf_emit(ctx, 0x30, 0);		/* ff SET_PROGRAM_CB */
+	xf_emit(ctx, 1, 0);		/* 3f last SET_PROGRAM_CB */
+	xf_emit(ctx, 4, 0);		/* RO */
+	xf_emit(ctx, 0x100, 0);		/* ffffffff */
+	xf_emit(ctx, 8, 0);		/* 1f, 0, 0, ... */
+	xf_emit(ctx, 8, 0);		/* ffffffff */
+	xf_emit(ctx, 4, 0);		/* ffffffff */
+	xf_emit(ctx, 1, 0);		/* 3 */
+	xf_emit(ctx, 1, 0);		/* ffffffff */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_CODE_CB */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_TIC */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_TSC */
+	xf_emit(ctx, 1, 0);		/* 00000001 LINKED_TSC */
+	xf_emit(ctx, 1, 0);		/* 000000ff TIC_ADDRESS_HIGH */
+	xf_emit(ctx, 1, 0);		/* ffffffff TIC_ADDRESS_LOW */
+	xf_emit(ctx, 1, 0x3fffff);	/* 003fffff TIC_LIMIT */
+	xf_emit(ctx, 1, 0);		/* 000000ff TSC_ADDRESS_HIGH */
+	xf_emit(ctx, 1, 0);		/* ffffffff TSC_ADDRESS_LOW */
+	xf_emit(ctx, 1, 0x1fff);	/* 000fffff TSC_LIMIT */
+	xf_emit(ctx, 1, 0);		/* 000000ff VP_ADDRESS_HIGH */
+	xf_emit(ctx, 1, 0);		/* ffffffff VP_ADDRESS_LOW */
+	xf_emit(ctx, 1, 0);		/* 00ffffff VP_START_ID */
+	xf_emit(ctx, 1, 0);		/* 000000ff CB_DEF_ADDRESS_HIGH */
+	xf_emit(ctx, 1, 0);		/* ffffffff CB_DEF_ADDRESS_LOW */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 000000ff GP_ADDRESS_HIGH */
+	xf_emit(ctx, 1, 0);		/* ffffffff GP_ADDRESS_LOW */
+	xf_emit(ctx, 1, 0);		/* 00ffffff GP_START_ID */
+	xf_emit(ctx, 1, 0);		/* 000000ff FP_ADDRESS_HIGH */
+	xf_emit(ctx, 1, 0);		/* ffffffff FP_ADDRESS_LOW */
+	xf_emit(ctx, 1, 0);		/* 00ffffff FP_START_ID */
+}
+
+static void
+nv50_graph_construct_gene_unk10xx(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int i;
+	/* end of area 2 on pre-NVA0, area 1 on NVAx */
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 4);		/* 0000007f VP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 0x80);		/* 0000ffff GP_VERTEX_OUTPUT_COUNT */
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_REG_ALLOC_RESULT */
+	xf_emit(ctx, 1, 0x80c14);	/* 01ffffff SEMANTIC_COLOR */
+	xf_emit(ctx, 1, 0);		/* 00000001 VERTEX_TWO_SIDE_ENABLE */
+	if (device->chipset == 0x50)
+		xf_emit(ctx, 1, 0x3ff);
+	else
+		xf_emit(ctx, 1, 0x7ff);	/* 000007ff */
+	xf_emit(ctx, 1, 0);		/* 111/113 */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+	for (i = 0; i < 8; i++) {
+		switch (device->chipset) {
+		case 0x50:
+		case 0x86:
+		case 0x98:
+		case 0xaa:
+		case 0xac:
+			xf_emit(ctx, 0xa0, 0);	/* ffffffff */
+			break;
+		case 0x84:
+		case 0x92:
+		case 0x94:
+		case 0x96:
+			xf_emit(ctx, 0x120, 0);
+			break;
+		case 0xa5:
+		case 0xa8:
+			xf_emit(ctx, 0x100, 0);	/* ffffffff */
+			break;
+		case 0xa0:
+		case 0xa3:
+		case 0xaf:
+			xf_emit(ctx, 0x400, 0);	/* ffffffff */
+			break;
+		}
+		xf_emit(ctx, 4, 0);	/* 3f, 0, 0, 0 */
+		xf_emit(ctx, 4, 0);	/* ffffffff */
+	}
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 4);		/* 0000007f VP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 0x80);		/* 0000ffff GP_VERTEX_OUTPUT_COUNT */
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_REG_ALLOC_TEMP */
+	xf_emit(ctx, 1, 1);		/* 00000001 RASTERIZE_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1900 */
+	xf_emit(ctx, 1, 0x27);		/* 000000ff UNK0FD4 */
+	xf_emit(ctx, 1, 0);		/* 0001ffff GP_BUILTIN_RESULT_EN */
+	xf_emit(ctx, 1, 0x26);		/* 000000ff SEMANTIC_LAYER */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+}
+
+static void
+nv50_graph_construct_gene_unk34xx(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	/* end of area 2 on pre-NVA0, area 1 on NVAx */
+	xf_emit(ctx, 1, 0);		/* 00000001 VIEWPORT_CLIP_RECTS_EN */
+	xf_emit(ctx, 1, 0);		/* 00000003 VIEWPORT_CLIP_MODE */
+	xf_emit(ctx, 0x10, 0x04000000);	/* 07ffffff VIEWPORT_CLIP_HORIZ*8, VIEWPORT_CLIP_VERT*8 */
+	xf_emit(ctx, 1, 0);		/* 00000001 POLYGON_STIPPLE_ENABLE */
+	xf_emit(ctx, 0x20, 0);		/* ffffffff POLYGON_STIPPLE */
+	xf_emit(ctx, 2, 0);		/* 00007fff WINDOW_OFFSET_XY */
+	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+	xf_emit(ctx, 1, 0x04e3bfdf);	/* ffffffff UNK0D64 */
+	xf_emit(ctx, 1, 0x04e3bfdf);	/* ffffffff UNK0DF4 */
+	xf_emit(ctx, 1, 0);		/* 00000003 WINDOW_ORIGIN */
+	xf_emit(ctx, 1, 0);		/* 00000007 */
+	xf_emit(ctx, 1, 0x1fe21);	/* 0001ffff tesla UNK0FAC */
+	if (device->chipset >= 0xa0)
+		xf_emit(ctx, 1, 0x0fac6881);
+	if (IS_NVA3F(device->chipset)) {
+		xf_emit(ctx, 1, 1);
+		xf_emit(ctx, 3, 0);
+	}
+}
+
+static void
+nv50_graph_construct_gene_unk14xx(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	/* middle of area 2 on pre-NVA0, beginning of area 2 on NVA0, area 7 on >NVA0 */
+	if (device->chipset != 0x50) {
+		xf_emit(ctx, 5, 0);		/* ffffffff */
+		xf_emit(ctx, 1, 0x80c14);	/* 01ffffff SEMANTIC_COLOR */
+		xf_emit(ctx, 1, 0);		/* 00000001 */
+		xf_emit(ctx, 1, 0);		/* 000003ff */
+		xf_emit(ctx, 1, 0x804);		/* 00000fff SEMANTIC_CLIP */
+		xf_emit(ctx, 1, 0);		/* 00000001 */
+		xf_emit(ctx, 2, 4);		/* 7f, ff */
+		xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
+	}
+	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 1, 4);			/* 0000007f VP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 4);			/* 000000ff GP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0);			/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 0x10);			/* 7f/ff VIEW_VOLUME_CLIP_CTRL */
+	xf_emit(ctx, 1, 0);			/* 000000ff VP_CLIP_DISTANCE_ENABLE */
+	if (device->chipset != 0x50)
+		xf_emit(ctx, 1, 0);		/* 3ff */
+	xf_emit(ctx, 1, 0);			/* 000000ff tesla UNK1940 */
+	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK0D7C */
+	xf_emit(ctx, 1, 0x804);			/* 00000fff SEMANTIC_CLIP */
+	xf_emit(ctx, 1, 1);			/* 00000001 VIEWPORT_TRANSFORM_EN */
+	xf_emit(ctx, 1, 0x1a);			/* 0000001f POLYGON_MODE */
+	if (device->chipset != 0x50)
+		xf_emit(ctx, 1, 0x7f);		/* 000000ff tesla UNK0FFC */
+	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 1, 1);			/* 00000001 SHADE_MODEL */
+	xf_emit(ctx, 1, 0x80c14);		/* 01ffffff SEMANTIC_COLOR */
+	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1900 */
+	xf_emit(ctx, 1, 0x8100c12);		/* 1fffffff FP_INTERPOLANT_CTRL */
+	xf_emit(ctx, 1, 4);			/* 0000007f VP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 4);			/* 000000ff GP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0);			/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 0x10);			/* 7f/ff VIEW_VOLUME_CLIP_CTRL */
+	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK0D7C */
+	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK0F8C */
+	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 1, 1);			/* 00000001 VIEWPORT_TRANSFORM_EN */
+	xf_emit(ctx, 1, 0x8100c12);		/* 1fffffff FP_INTERPOLANT_CTRL */
+	xf_emit(ctx, 4, 0);			/* ffffffff NOPERSPECTIVE_BITMAP */
+	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1900 */
+	xf_emit(ctx, 1, 0);			/* 0000000f */
+	if (device->chipset == 0x50)
+		xf_emit(ctx, 1, 0x3ff);		/* 000003ff tesla UNK0D68 */
+	else
+		xf_emit(ctx, 1, 0x7ff);		/* 000007ff tesla UNK0D68 */
+	xf_emit(ctx, 1, 0x80c14);		/* 01ffffff SEMANTIC_COLOR */
+	xf_emit(ctx, 1, 0);			/* 00000001 VERTEX_TWO_SIDE_ENABLE */
+	xf_emit(ctx, 0x30, 0);			/* ffffffff VIEWPORT_SCALE: X0, Y0, Z0, X1, Y1, ... */
+	xf_emit(ctx, 3, 0);			/* f, 0, 0 */
+	xf_emit(ctx, 3, 0);			/* ffffffff last VIEWPORT_SCALE? */
+	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 1, 1);			/* 00000001 VIEWPORT_TRANSFORM_EN */
+	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1900 */
+	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1924 */
+	xf_emit(ctx, 1, 0x10);			/* 000000ff VIEW_VOLUME_CLIP_CTRL */
+	xf_emit(ctx, 1, 0);			/* 00000001 */
+	xf_emit(ctx, 0x30, 0);			/* ffffffff VIEWPORT_TRANSLATE */
+	xf_emit(ctx, 3, 0);			/* f, 0, 0 */
+	xf_emit(ctx, 3, 0);			/* ffffffff */
+	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 2, 0x88);			/* 000001ff tesla UNK19D8 */
+	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1924 */
+	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 1, 4);			/* 0000000f CULL_MODE */
+	xf_emit(ctx, 2, 0);			/* 07ffffff SCREEN_SCISSOR */
+	xf_emit(ctx, 2, 0);			/* 00007fff WINDOW_OFFSET_XY */
+	xf_emit(ctx, 1, 0);			/* 00000003 WINDOW_ORIGIN */
+	xf_emit(ctx, 0x10, 0);			/* 00000001 SCISSOR_ENABLE */
+	xf_emit(ctx, 1, 0);			/* 0001ffff GP_BUILTIN_RESULT_EN */
+	xf_emit(ctx, 1, 0x26);			/* 000000ff SEMANTIC_LAYER */
+	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1900 */
+	xf_emit(ctx, 1, 0);			/* 0000000f */
+	xf_emit(ctx, 1, 0x3f800000);		/* ffffffff LINE_WIDTH */
+	xf_emit(ctx, 1, 0);			/* 00000001 LINE_STIPPLE_ENABLE */
+	xf_emit(ctx, 1, 0);			/* 00000001 LINE_SMOOTH_ENABLE */
+	xf_emit(ctx, 1, 0);			/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 0);		/* 00000001 */
+	xf_emit(ctx, 1, 0x1a);			/* 0000001f POLYGON_MODE */
+	xf_emit(ctx, 1, 0x10);			/* 000000ff VIEW_VOLUME_CLIP_CTRL */
+	if (device->chipset != 0x50) {
+		xf_emit(ctx, 1, 0);		/* ffffffff */
+		xf_emit(ctx, 1, 0);		/* 00000001 */
+		xf_emit(ctx, 1, 0);		/* 000003ff */
+	}
+	xf_emit(ctx, 0x20, 0);			/* 10xbits ffffffff, 3fffff. SCISSOR_* */
+	xf_emit(ctx, 1, 0);			/* f */
+	xf_emit(ctx, 1, 0);			/* 0? */
+	xf_emit(ctx, 1, 0);			/* ffffffff */
+	xf_emit(ctx, 1, 0);			/* 003fffff */
+	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 1, 0x52);			/* 000001ff SEMANTIC_PTSZ */
+	xf_emit(ctx, 1, 0);			/* 0001ffff GP_BUILTIN_RESULT_EN */
+	xf_emit(ctx, 1, 0x26);			/* 000000ff SEMANTIC_LAYER */
+	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1900 */
+	xf_emit(ctx, 1, 4);			/* 0000007f VP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 4);			/* 000000ff GP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0);			/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 0x1a);			/* 0000001f POLYGON_MODE */
+	xf_emit(ctx, 1, 0);			/* 00000001 LINE_SMOOTH_ENABLE */
+	xf_emit(ctx, 1, 0);			/* 00000001 LINE_STIPPLE_ENABLE */
+	xf_emit(ctx, 1, 0x00ffff00);		/* 00ffffff LINE_STIPPLE_PATTERN */
+	xf_emit(ctx, 1, 0);			/* 0000000f */
+}
+
+static void
+nv50_graph_construct_gene_zcull(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	/* end of strand 0 on pre-NVA0, beginning of strand 6 on NVAx */
+	/* SEEK */
+	xf_emit(ctx, 1, 0x3f);		/* 0000003f UNK1590 */
+	xf_emit(ctx, 1, 0);		/* 00000001 ALPHA_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
+	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_BACK_FUNC_FUNC */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_MASK */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_REF */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_MASK */
+	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_BACK_OP_FAIL, ZFAIL, ZPASS */
+	xf_emit(ctx, 1, 2);		/* 00000003 tesla UNK143C */
+	xf_emit(ctx, 2, 0x04000000);	/* 07ffffff tesla UNK0D6C */
+	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+	xf_emit(ctx, 1, 0);		/* 00000001 CLIPID_ENABLE */
+	xf_emit(ctx, 2, 0);		/* ffffffff DEPTH_BOUNDS */
+	xf_emit(ctx, 1, 0);		/* 00000001 */
+	xf_emit(ctx, 1, 0);		/* 00000007 DEPTH_TEST_FUNC */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
+	xf_emit(ctx, 1, 4);		/* 0000000f CULL_MODE */
+	xf_emit(ctx, 1, 0);		/* 0000ffff */
+	xf_emit(ctx, 1, 0);		/* 00000001 UNK0FB0 */
+	xf_emit(ctx, 1, 0);		/* 00000001 POLYGON_STIPPLE_ENABLE */
+	xf_emit(ctx, 1, 4);		/* 00000007 FP_CONTROL */
+	xf_emit(ctx, 1, 0);		/* ffffffff */
+	xf_emit(ctx, 1, 0);		/* 0001ffff GP_BUILTIN_RESULT_EN */
+	xf_emit(ctx, 1, 0);		/* 000000ff CLEAR_STENCIL */
+	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_FRONT_FUNC_FUNC */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_MASK */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_REF */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_MASK */
+	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_FRONT_OP_FAIL, ZFAIL, ZPASS */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
+	xf_emit(ctx, 1, 0);		/* ffffffff CLEAR_DEPTH */
+	xf_emit(ctx, 1, 0);		/* 00000007 */
+	if (device->chipset != 0x50)
+		xf_emit(ctx, 1, 0);	/* 00000003 tesla UNK1108 */
+	xf_emit(ctx, 1, 0);		/* 00000001 SAMPLECNT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
+	xf_emit(ctx, 1, 0x1001);	/* 00001fff ZETA_ARRAY_MODE */
+	/* SEEK */
+	xf_emit(ctx, 4, 0xffff);	/* 0000ffff MSAA_MASK */
+	xf_emit(ctx, 0x10, 0);		/* 00000001 SCISSOR_ENABLE */
+	xf_emit(ctx, 0x10, 0);		/* ffffffff DEPTH_RANGE_NEAR */
+	xf_emit(ctx, 0x10, 0x3f800000);	/* ffffffff DEPTH_RANGE_FAR */
+	xf_emit(ctx, 1, 0x10);		/* 7f/ff/3ff VIEW_VOLUME_CLIP_CTRL */
+	xf_emit(ctx, 1, 0);		/* 00000001 VIEWPORT_CLIP_RECTS_EN */
+	xf_emit(ctx, 1, 3);		/* 00000003 FP_CTRL_UNK196C */
+	xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK1968 */
+	if (device->chipset != 0x50)
+		xf_emit(ctx, 1, 0);	/* 0fffffff tesla UNK1104 */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK151C */
+}
+
+static void
+nv50_graph_construct_gene_clipid(struct nouveau_grctx *ctx)
+{
+	/* middle of strand 0 on pre-NVA0 [after 24xx], middle of area 6 on NVAx */
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* 00000007 UNK0FB4 */
+	/* SEEK */
+	xf_emit(ctx, 4, 0);		/* 07ffffff CLIPID_REGION_HORIZ */
+	xf_emit(ctx, 4, 0);		/* 07ffffff CLIPID_REGION_VERT */
+	xf_emit(ctx, 2, 0);		/* 07ffffff SCREEN_SCISSOR */
+	xf_emit(ctx, 2, 0x04000000);	/* 07ffffff UNK1508 */
+	xf_emit(ctx, 1, 0);		/* 00000001 CLIPID_ENABLE */
+	xf_emit(ctx, 1, 0x80);		/* 00003fff CLIPID_WIDTH */
+	xf_emit(ctx, 1, 0);		/* 000000ff CLIPID_ID */
+	xf_emit(ctx, 1, 0);		/* 000000ff CLIPID_ADDRESS_HIGH */
+	xf_emit(ctx, 1, 0);		/* ffffffff CLIPID_ADDRESS_LOW */
+	xf_emit(ctx, 1, 0x80);		/* 00003fff CLIPID_HEIGHT */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_CLIPID */
+}
+
+static void
+nv50_graph_construct_gene_unk24xx(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int i;
+	/* middle of strand 0 on pre-NVA0 [after m2mf], end of strand 2 on NVAx */
+	/* SEEK */
+	xf_emit(ctx, 0x33, 0);
+	/* SEEK */
+	xf_emit(ctx, 2, 0);
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 4);		/* 0000007f VP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
+	/* SEEK */
+	if (IS_NVA3F(device->chipset)) {
+		xf_emit(ctx, 4, 0);	/* RO */
+		xf_emit(ctx, 0xe10, 0); /* 190 * 9: 8*ffffffff, 7ff */
+		xf_emit(ctx, 1, 0);	/* 1ff */
+		xf_emit(ctx, 8, 0);	/* 0? */
+		xf_emit(ctx, 9, 0);	/* ffffffff, 7ff */
+
+		xf_emit(ctx, 4, 0);	/* RO */
+		xf_emit(ctx, 0xe10, 0); /* 190 * 9: 8*ffffffff, 7ff */
+		xf_emit(ctx, 1, 0);	/* 1ff */
+		xf_emit(ctx, 8, 0);	/* 0? */
+		xf_emit(ctx, 9, 0);	/* ffffffff, 7ff */
+	} else {
+		xf_emit(ctx, 0xc, 0);	/* RO */
+		/* SEEK */
+		xf_emit(ctx, 0xe10, 0); /* 190 * 9: 8*ffffffff, 7ff */
+		xf_emit(ctx, 1, 0);	/* 1ff */
+		xf_emit(ctx, 8, 0);	/* 0? */
+
+		/* SEEK */
+		xf_emit(ctx, 0xc, 0);	/* RO */
+		/* SEEK */
+		xf_emit(ctx, 0xe10, 0); /* 190 * 9: 8*ffffffff, 7ff */
+		xf_emit(ctx, 1, 0);	/* 1ff */
+		xf_emit(ctx, 8, 0);	/* 0? */
+	}
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 4);		/* 0000007f VP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
+	if (device->chipset != 0x50)
+		xf_emit(ctx, 1, 3);	/* 00000003 tesla UNK1100 */
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
+	xf_emit(ctx, 1, 0);		/* 0000000f VP_GP_BUILTIN_ATTR_EN */
+	xf_emit(ctx, 1, 0x80c14);	/* 01ffffff SEMANTIC_COLOR */
+	xf_emit(ctx, 1, 1);		/* 00000001 */
+	/* SEEK */
+	if (device->chipset >= 0xa0)
+		xf_emit(ctx, 2, 4);	/* 000000ff */
+	xf_emit(ctx, 1, 0x80c14);	/* 01ffffff SEMANTIC_COLOR */
+	xf_emit(ctx, 1, 0);		/* 00000001 VERTEX_TWO_SIDE_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 POINT_SPRITE_ENABLE */
+	xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
+	xf_emit(ctx, 1, 0x27);		/* 000000ff SEMANTIC_PRIM_ID */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 0000000f */
+	xf_emit(ctx, 1, 1);		/* 00000001 */
+	for (i = 0; i < 10; i++) {
+		/* SEEK */
+		xf_emit(ctx, 0x40, 0);		/* ffffffff */
+		xf_emit(ctx, 0x10, 0);		/* 3, 0, 0.... */
+		xf_emit(ctx, 0x10, 0);		/* ffffffff */
+	}
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* 00000001 POINT_SPRITE_CTRL */
+	xf_emit(ctx, 1, 1);		/* 00000001 */
+	xf_emit(ctx, 1, 0);		/* ffffffff */
+	xf_emit(ctx, 4, 0);		/* ffffffff NOPERSPECTIVE_BITMAP */
+	xf_emit(ctx, 0x10, 0);		/* 00ffffff POINT_COORD_REPLACE_MAP */
+	xf_emit(ctx, 1, 0);		/* 00000003 WINDOW_ORIGIN */
+	xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
+	if (device->chipset != 0x50)
+		xf_emit(ctx, 1, 0);	/* 000003ff */
+}
+
+static void
+nv50_graph_construct_gene_vfetch(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int acnt = 0x10, rep, i;
+	/* beginning of strand 1 on pre-NVA0, strand 3 on NVAx */
+	if (IS_NVA3F(device->chipset))
+		acnt = 0x20;
+	/* SEEK */
+	if (device->chipset >= 0xa0) {
+		xf_emit(ctx, 1, 0);	/* ffffffff tesla UNK13A4 */
+		xf_emit(ctx, 1, 1);	/* 00000fff tesla UNK1318 */
+	}
+	xf_emit(ctx, 1, 0);		/* ffffffff VERTEX_BUFFER_FIRST */
+	xf_emit(ctx, 1, 0);		/* 00000001 PRIMITIVE_RESTART_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 UNK0DE8 */
+	xf_emit(ctx, 1, 0);		/* ffffffff PRIMITIVE_RESTART_INDEX */
+	xf_emit(ctx, 1, 0xf);		/* ffffffff VP_ATTR_EN */
+	xf_emit(ctx, (acnt/8)-1, 0);	/* ffffffff VP_ATTR_EN */
+	xf_emit(ctx, acnt/8, 0);	/* ffffffff VTX_ATR_MASK_UNK0DD0 */
+	xf_emit(ctx, 1, 0);		/* 0000000f VP_GP_BUILTIN_ATTR_EN */
+	xf_emit(ctx, 1, 0x20);		/* 0000ffff tesla UNK129C */
+	xf_emit(ctx, 1, 0);		/* 000000ff turing UNK370??? */
+	xf_emit(ctx, 1, 0);		/* 0000ffff turing USER_PARAM_COUNT */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+	/* SEEK */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 0xb, 0);	/* RO */
+	else if (device->chipset >= 0xa0)
+		xf_emit(ctx, 0x9, 0);	/* RO */
+	else
+		xf_emit(ctx, 0x8, 0);	/* RO */
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* 00000001 EDGE_FLAG */
+	xf_emit(ctx, 1, 0);		/* 00000001 PROVOKING_VERTEX_LAST */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 0x1a);		/* 0000001f POLYGON_MODE */
+	/* SEEK */
+	xf_emit(ctx, 0xc, 0);		/* RO */
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* 7f/ff */
+	xf_emit(ctx, 1, 4);		/* 7f/ff VP_REG_ALLOC_RESULT */
+	xf_emit(ctx, 1, 4);		/* 7f/ff VP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0);		/* 0000000f VP_GP_BUILTIN_ATTR_EN */
+	xf_emit(ctx, 1, 4);		/* 000001ff UNK1A28 */
+	xf_emit(ctx, 1, 8);		/* 000001ff UNK0DF0 */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	if (device->chipset == 0x50)
+		xf_emit(ctx, 1, 0x3ff);	/* 3ff tesla UNK0D68 */
+	else
+		xf_emit(ctx, 1, 0x7ff);	/* 7ff tesla UNK0D68 */
+	if (device->chipset == 0xa8)
+		xf_emit(ctx, 1, 0x1e00);	/* 7fff */
+	/* SEEK */
+	xf_emit(ctx, 0xc, 0);		/* RO or close */
+	/* SEEK */
+	xf_emit(ctx, 1, 0xf);		/* ffffffff VP_ATTR_EN */
+	xf_emit(ctx, (acnt/8)-1, 0);	/* ffffffff VP_ATTR_EN */
+	xf_emit(ctx, 1, 0);		/* 0000000f VP_GP_BUILTIN_ATTR_EN */
+	if (device->chipset > 0x50 && device->chipset < 0xa0)
+		xf_emit(ctx, 2, 0);	/* ffffffff */
+	else
+		xf_emit(ctx, 1, 0);	/* ffffffff */
+	xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK0FD8 */
+	/* SEEK */
+	if (IS_NVA3F(device->chipset)) {
+		xf_emit(ctx, 0x10, 0);	/* 0? */
+		xf_emit(ctx, 2, 0);	/* weird... */
+		xf_emit(ctx, 2, 0);	/* RO */
+	} else {
+		xf_emit(ctx, 8, 0);	/* 0? */
+		xf_emit(ctx, 1, 0);	/* weird... */
+		xf_emit(ctx, 2, 0);	/* RO */
+	}
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* ffffffff VB_ELEMENT_BASE */
+	xf_emit(ctx, 1, 0);		/* ffffffff UNK1438 */
+	xf_emit(ctx, acnt, 0);		/* 1 tesla UNK1000 */
+	if (device->chipset >= 0xa0)
+		xf_emit(ctx, 1, 0);	/* ffffffff tesla UNK1118? */
+	/* SEEK */
+	xf_emit(ctx, acnt, 0);		/* ffffffff VERTEX_ARRAY_UNK90C */
+	xf_emit(ctx, 1, 0);		/* f/1f */
+	/* SEEK */
+	xf_emit(ctx, acnt, 0);		/* ffffffff VERTEX_ARRAY_UNK90C */
+	xf_emit(ctx, 1, 0);		/* f/1f */
+	/* SEEK */
+	xf_emit(ctx, acnt, 0);		/* RO */
+	xf_emit(ctx, 2, 0);		/* RO */
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK111C? */
+	xf_emit(ctx, 1, 0);		/* RO */
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* 000000ff UNK15F4_ADDRESS_HIGH */
+	xf_emit(ctx, 1, 0);		/* ffffffff UNK15F4_ADDRESS_LOW */
+	xf_emit(ctx, 1, 0);		/* 000000ff UNK0F84_ADDRESS_HIGH */
+	xf_emit(ctx, 1, 0);		/* ffffffff UNK0F84_ADDRESS_LOW */
+	/* SEEK */
+	xf_emit(ctx, acnt, 0);		/* 00003fff VERTEX_ARRAY_ATTRIB_OFFSET */
+	xf_emit(ctx, 3, 0);		/* f/1f */
+	/* SEEK */
+	xf_emit(ctx, acnt, 0);		/* 00000fff VERTEX_ARRAY_STRIDE */
+	xf_emit(ctx, 3, 0);		/* f/1f */
+	/* SEEK */
+	xf_emit(ctx, acnt, 0);		/* ffffffff VERTEX_ARRAY_LOW */
+	xf_emit(ctx, 3, 0);		/* f/1f */
+	/* SEEK */
+	xf_emit(ctx, acnt, 0);		/* 000000ff VERTEX_ARRAY_HIGH */
+	xf_emit(ctx, 3, 0);		/* f/1f */
+	/* SEEK */
+	xf_emit(ctx, acnt, 0);		/* ffffffff VERTEX_LIMIT_LOW */
+	xf_emit(ctx, 3, 0);		/* f/1f */
+	/* SEEK */
+	xf_emit(ctx, acnt, 0);		/* 000000ff VERTEX_LIMIT_HIGH */
+	xf_emit(ctx, 3, 0);		/* f/1f */
+	/* SEEK */
+	if (IS_NVA3F(device->chipset)) {
+		xf_emit(ctx, acnt, 0);		/* f */
+		xf_emit(ctx, 3, 0);		/* f/1f */
+	}
+	/* SEEK */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 2, 0);	/* RO */
+	else
+		xf_emit(ctx, 5, 0);	/* RO */
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* ffff DMA_VTXBUF */
+	/* SEEK */
+	if (device->chipset < 0xa0) {
+		xf_emit(ctx, 0x41, 0);	/* RO */
+		/* SEEK */
+		xf_emit(ctx, 0x11, 0);	/* RO */
+	} else if (!IS_NVA3F(device->chipset))
+		xf_emit(ctx, 0x50, 0);	/* RO */
+	else
+		xf_emit(ctx, 0x58, 0);	/* RO */
+	/* SEEK */
+	xf_emit(ctx, 1, 0xf);		/* ffffffff VP_ATTR_EN */
+	xf_emit(ctx, (acnt/8)-1, 0);	/* ffffffff VP_ATTR_EN */
+	xf_emit(ctx, 1, 1);		/* 1 UNK0DEC */
+	/* SEEK */
+	xf_emit(ctx, acnt*4, 0);	/* ffffffff VTX_ATTR */
+	xf_emit(ctx, 4, 0);		/* f/1f, 0, 0, 0 */
+	/* SEEK */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 0x1d, 0);	/* RO */
+	else
+		xf_emit(ctx, 0x16, 0);	/* RO */
+	/* SEEK */
+	xf_emit(ctx, 1, 0xf);		/* ffffffff VP_ATTR_EN */
+	xf_emit(ctx, (acnt/8)-1, 0);	/* ffffffff VP_ATTR_EN */
+	/* SEEK */
+	if (device->chipset < 0xa0)
+		xf_emit(ctx, 8, 0);	/* RO */
+	else if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 0xc, 0);	/* RO */
+	else
+		xf_emit(ctx, 7, 0);	/* RO */
+	/* SEEK */
+	xf_emit(ctx, 0xa, 0);		/* RO */
+	if (device->chipset == 0xa0)
+		rep = 0xc;
+	else
+		rep = 4;
+	for (i = 0; i < rep; i++) {
+		/* SEEK */
+		if (IS_NVA3F(device->chipset))
+			xf_emit(ctx, 0x20, 0);	/* ffffffff */
+		xf_emit(ctx, 0x200, 0);	/* ffffffff */
+		xf_emit(ctx, 4, 0);	/* 7f/ff, 0, 0, 0 */
+		xf_emit(ctx, 4, 0);	/* ffffffff */
+	}
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* 113/111 */
+	xf_emit(ctx, 1, 0xf);		/* ffffffff VP_ATTR_EN */
+	xf_emit(ctx, (acnt/8)-1, 0);	/* ffffffff VP_ATTR_EN */
+	xf_emit(ctx, acnt/8, 0);	/* ffffffff VTX_ATTR_MASK_UNK0DD0 */
+	xf_emit(ctx, 1, 0);		/* 0000000f VP_GP_BUILTIN_ATTR_EN */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+	/* SEEK */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 7, 0);	/* weird... */
+	else
+		xf_emit(ctx, 5, 0);	/* weird... */
+}
+
+static void
+nv50_graph_construct_gene_eng2d(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	/* middle of strand 1 on pre-NVA0 [after vfetch], middle of strand 6 on NVAx */
+	/* SEEK */
+	xf_emit(ctx, 2, 0);		/* 0001ffff CLIP_X, CLIP_Y */
+	xf_emit(ctx, 2, 0);		/* 0000ffff CLIP_W, CLIP_H */
+	xf_emit(ctx, 1, 0);		/* 00000001 CLIP_ENABLE */
+	if (device->chipset < 0xa0) {
+		/* this is useless on everything but the original NV50,
+		 * guess they forgot to nuke it. Or just didn't bother. */
+		xf_emit(ctx, 2, 0);	/* 0000ffff IFC_CLIP_X, Y */
+		xf_emit(ctx, 2, 1);	/* 0000ffff IFC_CLIP_W, H */
+		xf_emit(ctx, 1, 0);	/* 00000001 IFC_CLIP_ENABLE */
+	}
+	xf_emit(ctx, 1, 1);		/* 00000001 DST_LINEAR */
+	xf_emit(ctx, 1, 0x100);		/* 0001ffff DST_WIDTH */
+	xf_emit(ctx, 1, 0x100);		/* 0001ffff DST_HEIGHT */
+	xf_emit(ctx, 1, 0x11);		/* 3f[NV50]/7f[NV84+] DST_FORMAT */
+	xf_emit(ctx, 1, 0);		/* 0001ffff DRAW_POINT_X */
+	xf_emit(ctx, 1, 8);		/* 0000000f DRAW_UNK58C */
+	xf_emit(ctx, 1, 0);		/* 000fffff SIFC_DST_X_FRACT */
+	xf_emit(ctx, 1, 0);		/* 0001ffff SIFC_DST_X_INT */
+	xf_emit(ctx, 1, 0);		/* 000fffff SIFC_DST_Y_FRACT */
+	xf_emit(ctx, 1, 0);		/* 0001ffff SIFC_DST_Y_INT */
+	xf_emit(ctx, 1, 0);		/* 000fffff SIFC_DX_DU_FRACT */
+	xf_emit(ctx, 1, 1);		/* 0001ffff SIFC_DX_DU_INT */
+	xf_emit(ctx, 1, 0);		/* 000fffff SIFC_DY_DV_FRACT */
+	xf_emit(ctx, 1, 1);		/* 0001ffff SIFC_DY_DV_INT */
+	xf_emit(ctx, 1, 1);		/* 0000ffff SIFC_WIDTH */
+	xf_emit(ctx, 1, 1);		/* 0000ffff SIFC_HEIGHT */
+	xf_emit(ctx, 1, 0xcf);		/* 000000ff SIFC_FORMAT */
+	xf_emit(ctx, 1, 2);		/* 00000003 SIFC_BITMAP_UNK808 */
+	xf_emit(ctx, 1, 0);		/* 00000003 SIFC_BITMAP_LINE_PACK_MODE */
+	xf_emit(ctx, 1, 0);		/* 00000001 SIFC_BITMAP_LSB_FIRST */
+	xf_emit(ctx, 1, 0);		/* 00000001 SIFC_BITMAP_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 0000ffff BLIT_DST_X */
+	xf_emit(ctx, 1, 0);		/* 0000ffff BLIT_DST_Y */
+	xf_emit(ctx, 1, 0);		/* 000fffff BLIT_DU_DX_FRACT */
+	xf_emit(ctx, 1, 1);		/* 0001ffff BLIT_DU_DX_INT */
+	xf_emit(ctx, 1, 0);		/* 000fffff BLIT_DV_DY_FRACT */
+	xf_emit(ctx, 1, 1);		/* 0001ffff BLIT_DV_DY_INT */
+	xf_emit(ctx, 1, 1);		/* 0000ffff BLIT_DST_W */
+	xf_emit(ctx, 1, 1);		/* 0000ffff BLIT_DST_H */
+	xf_emit(ctx, 1, 0);		/* 000fffff BLIT_SRC_X_FRACT */
+	xf_emit(ctx, 1, 0);		/* 0001ffff BLIT_SRC_X_INT */
+	xf_emit(ctx, 1, 0);		/* 000fffff BLIT_SRC_Y_FRACT */
+	xf_emit(ctx, 1, 0);		/* 00000001 UNK888 */
+	xf_emit(ctx, 1, 4);		/* 0000003f UNK884 */
+	xf_emit(ctx, 1, 0);		/* 00000007 UNK880 */
+	xf_emit(ctx, 1, 1);		/* 0000001f tesla UNK0FB8 */
+	xf_emit(ctx, 1, 0x15);		/* 000000ff tesla UNK128C */
+	xf_emit(ctx, 2, 0);		/* 00000007, ffff0ff3 */
+	xf_emit(ctx, 1, 0);		/* 00000001 UNK260 */
+	xf_emit(ctx, 1, 0x4444480);	/* 1fffffff UNK870 */
+	/* SEEK */
+	xf_emit(ctx, 0x10, 0);
+	/* SEEK */
+	xf_emit(ctx, 0x27, 0);
+}
+
+static void
+nv50_graph_construct_gene_csched(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	/* middle of strand 1 on pre-NVA0 [after eng2d], middle of strand 0 on NVAx */
+	/* SEEK */
+	xf_emit(ctx, 2, 0);		/* 00007fff WINDOW_OFFSET_XY... what is it doing here??? */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1924 */
+	xf_emit(ctx, 1, 0);		/* 00000003 WINDOW_ORIGIN */
+	xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
+	xf_emit(ctx, 1, 0);		/* 000003ff */
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* ffffffff turing UNK364 */
+	xf_emit(ctx, 1, 0);		/* 0000000f turing UNK36C */
+	xf_emit(ctx, 1, 0);		/* 0000ffff USER_PARAM_COUNT */
+	xf_emit(ctx, 1, 0x100);		/* 00ffffff turing UNK384 */
+	xf_emit(ctx, 1, 0);		/* 0000000f turing UNK2A0 */
+	xf_emit(ctx, 1, 0);		/* 0000ffff GRIDID */
+	xf_emit(ctx, 1, 0x10001);	/* ffffffff GRIDDIM_XY */
+	xf_emit(ctx, 1, 0);		/* ffffffff */
+	xf_emit(ctx, 1, 0x10001);	/* ffffffff BLOCKDIM_XY */
+	xf_emit(ctx, 1, 1);		/* 0000ffff BLOCKDIM_Z */
+	xf_emit(ctx, 1, 0x10001);	/* 00ffffff BLOCK_ALLOC */
+	xf_emit(ctx, 1, 1);		/* 00000001 LANES32 */
+	xf_emit(ctx, 1, 4);		/* 000000ff FP_REG_ALLOC_TEMP */
+	xf_emit(ctx, 1, 2);		/* 00000003 REG_MODE */
+	/* SEEK */
+	xf_emit(ctx, 0x40, 0);		/* ffffffff USER_PARAM */
+	switch (device->chipset) {
+	case 0x50:
+	case 0x92:
+		xf_emit(ctx, 8, 0);	/* 7, 0, 0, 0, ... */
+		xf_emit(ctx, 0x80, 0);	/* fff */
+		xf_emit(ctx, 2, 0);	/* ff, fff */
+		xf_emit(ctx, 0x10*2, 0);	/* ffffffff, 1f */
+		break;
+	case 0x84:
+		xf_emit(ctx, 8, 0);	/* 7, 0, 0, 0, ... */
+		xf_emit(ctx, 0x60, 0);	/* fff */
+		xf_emit(ctx, 2, 0);	/* ff, fff */
+		xf_emit(ctx, 0xc*2, 0);	/* ffffffff, 1f */
+		break;
+	case 0x94:
+	case 0x96:
+		xf_emit(ctx, 8, 0);	/* 7, 0, 0, 0, ... */
+		xf_emit(ctx, 0x40, 0);	/* fff */
+		xf_emit(ctx, 2, 0);	/* ff, fff */
+		xf_emit(ctx, 8*2, 0);	/* ffffffff, 1f */
+		break;
+	case 0x86:
+	case 0x98:
+		xf_emit(ctx, 4, 0);	/* f, 0, 0, 0 */
+		xf_emit(ctx, 0x10, 0);	/* fff */
+		xf_emit(ctx, 2, 0);	/* ff, fff */
+		xf_emit(ctx, 2*2, 0);	/* ffffffff, 1f */
+		break;
+	case 0xa0:
+		xf_emit(ctx, 8, 0);	/* 7, 0, 0, 0, ... */
+		xf_emit(ctx, 0xf0, 0);	/* fff */
+		xf_emit(ctx, 2, 0);	/* ff, fff */
+		xf_emit(ctx, 0x1e*2, 0);	/* ffffffff, 1f */
+		break;
+	case 0xa3:
+		xf_emit(ctx, 8, 0);	/* 7, 0, 0, 0, ... */
+		xf_emit(ctx, 0x60, 0);	/* fff */
+		xf_emit(ctx, 2, 0);	/* ff, fff */
+		xf_emit(ctx, 0xc*2, 0);	/* ffffffff, 1f */
+		break;
+	case 0xa5:
+	case 0xaf:
+		xf_emit(ctx, 8, 0);	/* 7, 0, 0, 0, ... */
+		xf_emit(ctx, 0x30, 0);	/* fff */
+		xf_emit(ctx, 2, 0);	/* ff, fff */
+		xf_emit(ctx, 6*2, 0);	/* ffffffff, 1f */
+		break;
+	case 0xaa:
+		xf_emit(ctx, 0x12, 0);
+		break;
+	case 0xa8:
+	case 0xac:
+		xf_emit(ctx, 4, 0);	/* f, 0, 0, 0 */
+		xf_emit(ctx, 0x10, 0);	/* fff */
+		xf_emit(ctx, 2, 0);	/* ff, fff */
+		xf_emit(ctx, 2*2, 0);	/* ffffffff, 1f */
+		break;
+	}
+	xf_emit(ctx, 1, 0);		/* 0000000f */
+	xf_emit(ctx, 1, 0);		/* 00000000 */
+	xf_emit(ctx, 1, 0);		/* ffffffff */
+	xf_emit(ctx, 1, 0);		/* 0000001f */
+	xf_emit(ctx, 4, 0);		/* ffffffff */
+	xf_emit(ctx, 1, 0);		/* 00000003 turing UNK35C */
+	xf_emit(ctx, 1, 0);		/* ffffffff */
+	xf_emit(ctx, 4, 0);		/* ffffffff */
+	xf_emit(ctx, 1, 0);		/* 00000003 turing UNK35C */
+	xf_emit(ctx, 1, 0);		/* ffffffff */
+	xf_emit(ctx, 1, 0);		/* 000000ff */
+}
+
+static void
+nv50_graph_construct_gene_unk1cxx(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	xf_emit(ctx, 2, 0);		/* 00007fff WINDOW_OFFSET_XY */
+	xf_emit(ctx, 1, 0x3f800000);	/* ffffffff LINE_WIDTH */
+	xf_emit(ctx, 1, 0);		/* 00000001 LINE_SMOOTH_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1658 */
+	xf_emit(ctx, 1, 0);		/* 00000001 POLYGON_SMOOTH_ENABLE */
+	xf_emit(ctx, 3, 0);		/* 00000001 POLYGON_OFFSET_*_ENABLE */
+	xf_emit(ctx, 1, 4);		/* 0000000f CULL_MODE */
+	xf_emit(ctx, 1, 0x1a);		/* 0000001f POLYGON_MODE */
+	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
+	xf_emit(ctx, 1, 0);		/* 00000001 POINT_SPRITE_ENABLE */
+	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK165C */
+	xf_emit(ctx, 0x10, 0);		/* 00000001 SCISSOR_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
+	xf_emit(ctx, 1, 0);		/* 00000001 LINE_STIPPLE_ENABLE */
+	xf_emit(ctx, 1, 0x00ffff00);	/* 00ffffff LINE_STIPPLE_PATTERN */
+	xf_emit(ctx, 1, 0);		/* ffffffff POLYGON_OFFSET_UNITS */
+	xf_emit(ctx, 1, 0);		/* ffffffff POLYGON_OFFSET_FACTOR */
+	xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK1668 */
+	xf_emit(ctx, 2, 0);		/* 07ffffff SCREEN_SCISSOR */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1900 */
+	xf_emit(ctx, 1, 0xf);		/* 0000000f COLOR_MASK */
+	xf_emit(ctx, 7, 0);		/* 0000000f COLOR_MASK */
+	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
+	xf_emit(ctx, 1, 0x11);		/* 0000007f RT_FORMAT */
+	xf_emit(ctx, 7, 0);		/* 0000007f RT_FORMAT */
+	xf_emit(ctx, 8, 0);		/* 00000001 RT_HORIZ_LINEAR */
+	xf_emit(ctx, 1, 4);		/* 00000007 FP_CONTROL */
+	xf_emit(ctx, 1, 0);		/* 00000001 ALPHA_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000007 ALPHA_TEST_FUNC */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 3);	/* 00000003 UNK16B4 */
+	else if (device->chipset >= 0xa0)
+		xf_emit(ctx, 1, 1);	/* 00000001 UNK16B4 */
+	xf_emit(ctx, 1, 0);		/* 00000003 MULTISAMPLE_CTRL */
+	xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK0F90 */
+	xf_emit(ctx, 1, 2);		/* 00000003 tesla UNK143C */
+	xf_emit(ctx, 2, 0x04000000);	/* 07ffffff tesla UNK0D6C */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_MASK */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 SAMPLECNT_ENABLE */
+	xf_emit(ctx, 1, 5);		/* 0000000f UNK1408 */
+	xf_emit(ctx, 1, 0x52);		/* 000001ff SEMANTIC_PTSZ */
+	xf_emit(ctx, 1, 0);		/* ffffffff POINT_SIZE */
+	xf_emit(ctx, 1, 0);		/* 00000001 */
+	xf_emit(ctx, 1, 0);		/* 00000007 tesla UNK0FB4 */
+	if (device->chipset != 0x50) {
+		xf_emit(ctx, 1, 0);	/* 3ff */
+		xf_emit(ctx, 1, 1);	/* 00000001 tesla UNK1110 */
+	}
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 0);	/* 00000003 tesla UNK1928 */
+	xf_emit(ctx, 0x10, 0);		/* ffffffff DEPTH_RANGE_NEAR */
+	xf_emit(ctx, 0x10, 0x3f800000);	/* ffffffff DEPTH_RANGE_FAR */
+	xf_emit(ctx, 1, 0x10);		/* 000000ff VIEW_VOLUME_CLIP_CTRL */
+	xf_emit(ctx, 0x20, 0);		/* 07ffffff VIEWPORT_HORIZ, then VIEWPORT_VERT. (W&0x3fff)<<13 | (X&0x1fff). */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK187C */
+	xf_emit(ctx, 1, 0);		/* 00000003 WINDOW_ORIGIN */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_MASK */
+	xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
+	xf_emit(ctx, 1, 5);		/* 0000000f tesla UNK1220 */
+	xf_emit(ctx, 1, 0);		/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 1, 0);		/* 000000ff tesla UNK1A20 */
+	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 VERTEX_TWO_SIDE_ENABLE */
+	xf_emit(ctx, 4, 0xffff);	/* 0000ffff MSAA_MASK */
+	if (device->chipset != 0x50)
+		xf_emit(ctx, 1, 3);	/* 00000003 tesla UNK1100 */
+	if (device->chipset < 0xa0)
+		xf_emit(ctx, 0x1c, 0);	/* RO */
+	else if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 0x9, 0);
+	xf_emit(ctx, 1, 0);		/* 00000001 UNK1534 */
+	xf_emit(ctx, 1, 0);		/* 00000001 LINE_SMOOTH_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 LINE_STIPPLE_ENABLE */
+	xf_emit(ctx, 1, 0x00ffff00);	/* 00ffffff LINE_STIPPLE_PATTERN */
+	xf_emit(ctx, 1, 0x1a);		/* 0000001f POLYGON_MODE */
+	xf_emit(ctx, 1, 0);		/* 00000003 WINDOW_ORIGIN */
+	if (device->chipset != 0x50) {
+		xf_emit(ctx, 1, 3);	/* 00000003 tesla UNK1100 */
+		xf_emit(ctx, 1, 0);	/* 3ff */
+	}
+	/* XXX: the following block could belong either to unk1cxx, or
+	 * to STRMOUT. Rather hard to tell. */
+	if (device->chipset < 0xa0)
+		xf_emit(ctx, 0x25, 0);
+	else
+		xf_emit(ctx, 0x3b, 0);
+}
+
+static void
+nv50_graph_construct_gene_strmout(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	xf_emit(ctx, 1, 0x102);		/* 0000ffff STRMOUT_BUFFER_CTRL */
+	xf_emit(ctx, 1, 0);		/* ffffffff STRMOUT_PRIMITIVE_COUNT */
+	xf_emit(ctx, 4, 4);		/* 000000ff STRMOUT_NUM_ATTRIBS */
+	if (device->chipset >= 0xa0) {
+		xf_emit(ctx, 4, 0);	/* ffffffff UNK1A8C */
+		xf_emit(ctx, 4, 0);	/* ffffffff UNK1780 */
+	}
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 4);		/* 0000007f VP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	if (device->chipset == 0x50)
+		xf_emit(ctx, 1, 0x3ff);	/* 000003ff tesla UNK0D68 */
+	else
+		xf_emit(ctx, 1, 0x7ff);	/* 000007ff tesla UNK0D68 */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+	/* SEEK */
+	xf_emit(ctx, 1, 0x102);		/* 0000ffff STRMOUT_BUFFER_CTRL */
+	xf_emit(ctx, 1, 0);		/* ffffffff STRMOUT_PRIMITIVE_COUNT */
+	xf_emit(ctx, 4, 0);		/* 000000ff STRMOUT_ADDRESS_HIGH */
+	xf_emit(ctx, 4, 0);		/* ffffffff STRMOUT_ADDRESS_LOW */
+	xf_emit(ctx, 4, 4);		/* 000000ff STRMOUT_NUM_ATTRIBS */
+	if (device->chipset >= 0xa0) {
+		xf_emit(ctx, 4, 0);	/* ffffffff UNK1A8C */
+		xf_emit(ctx, 4, 0);	/* ffffffff UNK1780 */
+	}
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_STRMOUT */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_QUERY */
+	xf_emit(ctx, 1, 0);		/* 000000ff QUERY_ADDRESS_HIGH */
+	xf_emit(ctx, 2, 0);		/* ffffffff QUERY_ADDRESS_LOW QUERY_COUNTER */
+	xf_emit(ctx, 2, 0);		/* ffffffff */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+	/* SEEK */
+	xf_emit(ctx, 0x20, 0);		/* ffffffff STRMOUT_MAP */
+	xf_emit(ctx, 1, 0);		/* 0000000f */
+	xf_emit(ctx, 1, 0);		/* 00000000? */
+	xf_emit(ctx, 2, 0);		/* ffffffff */
+}
+
+static void
+nv50_graph_construct_gene_ropm1(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	xf_emit(ctx, 1, 0x4e3bfdf);	/* ffffffff UNK0D64 */
+	xf_emit(ctx, 1, 0x4e3bfdf);	/* ffffffff UNK0DF4 */
+	xf_emit(ctx, 1, 0);		/* 00000007 */
+	xf_emit(ctx, 1, 0);		/* 000003ff */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 0x11);	/* 000000ff tesla UNK1968 */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
+}
+
+static void
+nv50_graph_construct_gene_ropm2(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_QUERY */
+	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
+	xf_emit(ctx, 2, 0);		/* ffffffff */
+	xf_emit(ctx, 1, 0);		/* 000000ff QUERY_ADDRESS_HIGH */
+	xf_emit(ctx, 2, 0);		/* ffffffff QUERY_ADDRESS_LOW, COUNTER */
+	xf_emit(ctx, 1, 0);		/* 00000001 SAMPLECNT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 7 */
+	/* SEEK */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_QUERY */
+	xf_emit(ctx, 1, 0);		/* 000000ff QUERY_ADDRESS_HIGH */
+	xf_emit(ctx, 2, 0);		/* ffffffff QUERY_ADDRESS_LOW, COUNTER */
+	xf_emit(ctx, 1, 0x4e3bfdf);	/* ffffffff UNK0D64 */
+	xf_emit(ctx, 1, 0x4e3bfdf);	/* ffffffff UNK0DF4 */
+	xf_emit(ctx, 1, 0);		/* 00000001 eng2d UNK260 */
+	xf_emit(ctx, 1, 0);		/* ff/3ff */
+	xf_emit(ctx, 1, 0);		/* 00000007 */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 0x11);	/* 000000ff tesla UNK1968 */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
+}
+
+static void
+nv50_graph_construct_gene_ropc(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int magic2;
+	if (device->chipset == 0x50) {
+		magic2 = 0x00003e60;
+	} else if (!IS_NVA3F(device->chipset)) {
+		magic2 = 0x001ffe67;
+	} else {
+		magic2 = 0x00087e67;
+	}
+	xf_emit(ctx, 1, 0);		/* f/7 MUTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
+	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_BACK_FUNC_FUNC */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_MASK */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_MASK */
+	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_BACK_OP_FAIL, ZFAIL, ZPASS */
+	xf_emit(ctx, 1, 2);		/* 00000003 tesla UNK143C */
+	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+	xf_emit(ctx, 1, magic2);	/* 001fffff tesla UNK0F78 */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
+	xf_emit(ctx, 1, 0);		/* 00000007 DEPTH_TEST_FUNC */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
+	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_FRONT_FUNC_FUNC */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_MASK */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_MASK */
+	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_FRONT_OP_FAIL, ZFAIL, ZPASS */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
+	if (device->chipset >= 0xa0 && !IS_NVAAF(device->chipset))
+		xf_emit(ctx, 1, 0x15);	/* 000000ff */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
+	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK15B4 */
+	xf_emit(ctx, 1, 0x10);		/* 3ff/ff VIEW_VOLUME_CLIP_CTRL */
+	xf_emit(ctx, 1, 0);		/* ffffffff CLEAR_DEPTH */
+	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
+	if (device->chipset == 0x86 || device->chipset == 0x92 || device->chipset == 0x98 || device->chipset >= 0xa0) {
+		xf_emit(ctx, 3, 0);	/* ff, ffffffff, ffffffff */
+		xf_emit(ctx, 1, 4);	/* 7 */
+		xf_emit(ctx, 1, 0x400);	/* fffffff */
+		xf_emit(ctx, 1, 0x300);	/* ffff */
+		xf_emit(ctx, 1, 0x1001);	/* 1fff */
+		if (device->chipset != 0xa0) {
+			if (IS_NVA3F(device->chipset))
+				xf_emit(ctx, 1, 0);	/* 0000000f UNK15C8 */
+			else
+				xf_emit(ctx, 1, 0x15);	/* ff */
+		}
+	}
+	xf_emit(ctx, 1, 0);		/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
+	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_BACK_FUNC_FUNC */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_MASK */
+	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+	xf_emit(ctx, 1, 2);		/* 00000003 tesla UNK143C */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
+	xf_emit(ctx, 1, 0);		/* 00000007 DEPTH_TEST_FUNC */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_FRONT_FUNC_FUNC */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_MASK */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
+	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK15B4 */
+	xf_emit(ctx, 1, 0x10);		/* 7f/ff VIEW_VOLUME_CLIP_CTRL */
+	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1900 */
+	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_BACK_FUNC_FUNC */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_MASK */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_REF */
+	xf_emit(ctx, 2, 0);		/* ffffffff DEPTH_BOUNDS */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
+	xf_emit(ctx, 1, 0);		/* 00000007 DEPTH_TEST_FUNC */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 0000000f */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK0FB0 */
+	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_FRONT_FUNC_FUNC */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_MASK */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_REF */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
+	xf_emit(ctx, 1, 0x10);		/* 7f/ff VIEW_VOLUME_CLIP_CTRL */
+	xf_emit(ctx, 0x10, 0);		/* ffffffff DEPTH_RANGE_NEAR */
+	xf_emit(ctx, 0x10, 0x3f800000);	/* ffffffff DEPTH_RANGE_FAR */
+	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
+	xf_emit(ctx, 1, 0);		/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_BACK_FUNC_FUNC */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_MASK */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_REF */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_MASK */
+	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_BACK_OP_FAIL, ZFAIL, ZPASS */
+	xf_emit(ctx, 2, 0);		/* ffffffff DEPTH_BOUNDS */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
+	xf_emit(ctx, 1, 0);		/* 00000007 DEPTH_TEST_FUNC */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 000000ff CLEAR_STENCIL */
+	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_FRONT_FUNC_FUNC */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_MASK */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_REF */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_MASK */
+	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_FRONT_OP_FAIL, ZFAIL, ZPASS */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
+	xf_emit(ctx, 1, 0x10);		/* 7f/ff VIEW_VOLUME_CLIP_CTRL */
+	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
+	xf_emit(ctx, 1, 0x3f);		/* 0000003f UNK1590 */
+	xf_emit(ctx, 1, 0);		/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
+	xf_emit(ctx, 2, 0);		/* ffff0ff3, ffff */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK0FB0 */
+	xf_emit(ctx, 1, 0);		/* 0001ffff GP_BUILTIN_RESULT_EN */
+	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK15B4 */
+	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
+	xf_emit(ctx, 1, 0);		/* ffffffff CLEAR_DEPTH */
+	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK19CC */
+	if (device->chipset >= 0xa0) {
+		xf_emit(ctx, 2, 0);
+		xf_emit(ctx, 1, 0x1001);
+		xf_emit(ctx, 0xb, 0);
+	} else {
+		xf_emit(ctx, 1, 0);	/* 00000007 */
+		xf_emit(ctx, 1, 0);	/* 00000001 tesla UNK1534 */
+		xf_emit(ctx, 1, 0);	/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
+		xf_emit(ctx, 8, 0);	/* 00000001 BLEND_ENABLE */
+		xf_emit(ctx, 1, 0);	/* ffff0ff3 */
+	}
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f RT_FORMAT */
+	xf_emit(ctx, 7, 0);		/* 3f/7f RT_FORMAT */
+	xf_emit(ctx, 1, 0xf);		/* 0000000f COLOR_MASK */
+	xf_emit(ctx, 7, 0);		/* 0000000f COLOR_MASK */
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f */
+	xf_emit(ctx, 1, 0);		/* 00000001 LOGIC_OP_ENABLE */
+	if (device->chipset != 0x50) {
+		xf_emit(ctx, 1, 0);	/* 0000000f LOGIC_OP */
+		xf_emit(ctx, 1, 0);	/* 000000ff */
+	}
+	xf_emit(ctx, 1, 0);		/* 00000007 OPERATION */
+	xf_emit(ctx, 1, 0);		/* ff/3ff */
+	xf_emit(ctx, 1, 0);		/* 00000003 UNK0F90 */
+	xf_emit(ctx, 2, 1);		/* 00000007 BLEND_EQUATION_RGB, ALPHA */
+	xf_emit(ctx, 1, 1);		/* 00000001 UNK133C */
+	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_RGB */
+	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_RGB */
+	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_ALPHA */
+	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_ALPHA */
+	xf_emit(ctx, 1, 0);		/* 00000001 */
+	xf_emit(ctx, 1, magic2);	/* 001fffff tesla UNK0F78 */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
+	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
+	if (IS_NVA3F(device->chipset)) {
+		xf_emit(ctx, 1, 0);	/* 00000001 tesla UNK12E4 */
+		xf_emit(ctx, 8, 1);	/* 00000007 IBLEND_EQUATION_RGB */
+		xf_emit(ctx, 8, 1);	/* 00000007 IBLEND_EQUATION_ALPHA */
+		xf_emit(ctx, 8, 1);	/* 00000001 IBLEND_UNK00 */
+		xf_emit(ctx, 8, 2);	/* 0000001f IBLEND_FUNC_SRC_RGB */
+		xf_emit(ctx, 8, 1);	/* 0000001f IBLEND_FUNC_DST_RGB */
+		xf_emit(ctx, 8, 2);	/* 0000001f IBLEND_FUNC_SRC_ALPHA */
+		xf_emit(ctx, 8, 1);	/* 0000001f IBLEND_FUNC_DST_ALPHA */
+		xf_emit(ctx, 1, 0);	/* 00000001 tesla UNK1140 */
+		xf_emit(ctx, 2, 0);	/* 00000001 */
+		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
+		xf_emit(ctx, 1, 0);	/* 0000000f */
+		xf_emit(ctx, 1, 0);	/* 00000003 */
+		xf_emit(ctx, 1, 0);	/* ffffffff */
+		xf_emit(ctx, 2, 0);	/* 00000001 */
+		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
+		xf_emit(ctx, 1, 0);	/* 00000001 */
+		xf_emit(ctx, 1, 0);	/* 000003ff */
+	} else if (device->chipset >= 0xa0) {
+		xf_emit(ctx, 2, 0);	/* 00000001 */
+		xf_emit(ctx, 1, 0);	/* 00000007 */
+		xf_emit(ctx, 1, 0);	/* 00000003 */
+		xf_emit(ctx, 1, 0);	/* ffffffff */
+		xf_emit(ctx, 2, 0);	/* 00000001 */
+	} else {
+		xf_emit(ctx, 1, 0);	/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
+		xf_emit(ctx, 1, 0);	/* 00000003 tesla UNK1430 */
+		xf_emit(ctx, 1, 0);	/* ffffffff tesla UNK1A3C */
+	}
+	xf_emit(ctx, 4, 0);		/* ffffffff CLEAR_COLOR */
+	xf_emit(ctx, 4, 0);		/* ffffffff BLEND_COLOR A R G B */
+	xf_emit(ctx, 1, 0);		/* 00000fff eng2d UNK2B0 */
+	if (device->chipset >= 0xa0)
+		xf_emit(ctx, 2, 0);	/* 00000001 */
+	xf_emit(ctx, 1, 0);		/* 000003ff */
+	xf_emit(ctx, 8, 0);		/* 00000001 BLEND_ENABLE */
+	xf_emit(ctx, 1, 1);		/* 00000001 UNK133C */
+	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_RGB */
+	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_RGB */
+	xf_emit(ctx, 1, 1);		/* 00000007 BLEND_EQUATION_RGB */
+	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_ALPHA */
+	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_ALPHA */
+	xf_emit(ctx, 1, 1);		/* 00000007 BLEND_EQUATION_ALPHA */
+	xf_emit(ctx, 1, 0);		/* 00000001 UNK19C0 */
+	xf_emit(ctx, 1, 0);		/* 00000001 LOGIC_OP_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 0000000f LOGIC_OP */
+	if (device->chipset >= 0xa0)
+		xf_emit(ctx, 1, 0);	/* 00000001 UNK12E4? NVA3+ only? */
+	if (IS_NVA3F(device->chipset)) {
+		xf_emit(ctx, 8, 1);	/* 00000001 IBLEND_UNK00 */
+		xf_emit(ctx, 8, 1);	/* 00000007 IBLEND_EQUATION_RGB */
+		xf_emit(ctx, 8, 2);	/* 0000001f IBLEND_FUNC_SRC_RGB */
+		xf_emit(ctx, 8, 1);	/* 0000001f IBLEND_FUNC_DST_RGB */
+		xf_emit(ctx, 8, 1);	/* 00000007 IBLEND_EQUATION_ALPHA */
+		xf_emit(ctx, 8, 2);	/* 0000001f IBLEND_FUNC_SRC_ALPHA */
+		xf_emit(ctx, 8, 1);	/* 0000001f IBLEND_FUNC_DST_ALPHA */
+		xf_emit(ctx, 1, 0);	/* 00000001 tesla UNK15C4 */
+		xf_emit(ctx, 1, 0);	/* 00000001 */
+		xf_emit(ctx, 1, 0);	/* 00000001 tesla UNK1140 */
+	}
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f DST_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 DST_LINEAR */
+	xf_emit(ctx, 1, 0);		/* 00000007 PATTERN_COLOR_FORMAT */
+	xf_emit(ctx, 2, 0);		/* ffffffff PATTERN_MONO_COLOR */
+	xf_emit(ctx, 1, 0);		/* 00000001 PATTERN_MONO_FORMAT */
+	xf_emit(ctx, 2, 0);		/* ffffffff PATTERN_MONO_BITMAP */
+	xf_emit(ctx, 1, 0);		/* 00000003 PATTERN_SELECT */
+	xf_emit(ctx, 1, 0);		/* 000000ff ROP */
+	xf_emit(ctx, 1, 0);		/* ffffffff BETA1 */
+	xf_emit(ctx, 1, 0);		/* ffffffff BETA4 */
+	xf_emit(ctx, 1, 0);		/* 00000007 OPERATION */
+	xf_emit(ctx, 0x50, 0);		/* 10x ffffff, ffffff, ffffff, ffffff, 3 PATTERN */
+}
+
+static void
+nv50_graph_construct_xfer_unk84xx(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int magic3;
+	switch (device->chipset) {
+	case 0x50:
+		magic3 = 0x1000;
+		break;
+	case 0x86:
+	case 0x98:
+	case 0xa8:
+	case 0xaa:
+	case 0xac:
+	case 0xaf:
+		magic3 = 0x1e00;
+		break;
+	default:
+		magic3 = 0;
+	}
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 4);		/* 7f/ff[NVA0+] VP_REG_ALLOC_RESULT */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 1, 0);		/* 111/113[NVA0+] */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 0x1f, 0);	/* ffffffff */
+	else if (device->chipset >= 0xa0)
+		xf_emit(ctx, 0x0f, 0);	/* ffffffff */
+	else
+		xf_emit(ctx, 0x10, 0);	/* fffffff VP_RESULT_MAP_1 up */
+	xf_emit(ctx, 2, 0);		/* f/1f[NVA3], fffffff/ffffffff[NVA0+] */
+	xf_emit(ctx, 1, 4);		/* 7f/ff VP_REG_ALLOC_RESULT */
+	xf_emit(ctx, 1, 4);		/* 7f/ff VP_RESULT_MAP_SIZE */
+	if (device->chipset >= 0xa0)
+		xf_emit(ctx, 1, 0x03020100);	/* ffffffff */
+	else
+		xf_emit(ctx, 1, 0x00608080);	/* fffffff VP_RESULT_MAP_0 */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 2, 0);		/* 111/113, 7f/ff */
+	xf_emit(ctx, 1, 4);		/* 7f/ff VP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_REG_ALLOC_RESULT */
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0x80);		/* 0000ffff GP_VERTEX_OUTPUT_COUNT */
+	if (magic3)
+		xf_emit(ctx, 1, magic3);	/* 00007fff tesla UNK141C */
+	xf_emit(ctx, 1, 4);		/* 7f/ff VP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 1, 0);		/* 111/113 */
+	xf_emit(ctx, 0x1f, 0);		/* ffffffff GP_RESULT_MAP_1 up */
+	xf_emit(ctx, 1, 0);		/* 0000001f */
+	xf_emit(ctx, 1, 0);		/* ffffffff */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_REG_ALLOC_RESULT */
+	xf_emit(ctx, 1, 0x80);		/* 0000ffff GP_VERTEX_OUTPUT_COUNT */
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0x03020100);	/* ffffffff GP_RESULT_MAP_0 */
+	xf_emit(ctx, 1, 3);		/* 00000003 GP_OUTPUT_PRIMITIVE_TYPE */
+	if (magic3)
+		xf_emit(ctx, 1, magic3);	/* 7fff tesla UNK141C */
+	xf_emit(ctx, 1, 4);		/* 7f/ff VP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 0);		/* 00000001 PROVOKING_VERTEX_LAST */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 1, 0);		/* 111/113 */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 3);		/* 00000003 GP_OUTPUT_PRIMITIVE_TYPE */
+	xf_emit(ctx, 1, 0);		/* 00000001 PROVOKING_VERTEX_LAST */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK13A0 */
+	xf_emit(ctx, 1, 4);		/* 7f/ff VP_REG_ALLOC_RESULT */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+	xf_emit(ctx, 1, 0);		/* 111/113 */
+	if (device->chipset == 0x94 || device->chipset == 0x96)
+		xf_emit(ctx, 0x1020, 0);	/* 4 x (0x400 x 0xffffffff, ff, 0, 0, 0, 4 x ffffffff) */
+	else if (device->chipset < 0xa0)
+		xf_emit(ctx, 0xa20, 0);	/* 4 x (0x280 x 0xffffffff, ff, 0, 0, 0, 4 x ffffffff) */
+	else if (!IS_NVA3F(device->chipset))
+		xf_emit(ctx, 0x210, 0);	/* ffffffff */
+	else
+		xf_emit(ctx, 0x410, 0);	/* ffffffff */
+	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
+	xf_emit(ctx, 1, 3);		/* 00000003 GP_OUTPUT_PRIMITIVE_TYPE */
+	xf_emit(ctx, 1, 0);		/* 00000001 PROVOKING_VERTEX_LAST */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+}
+
+static void
+nv50_graph_construct_xfer_tprop(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int magic1, magic2;
+	if (device->chipset == 0x50) {
+		magic1 = 0x3ff;
+		magic2 = 0x00003e60;
+	} else if (!IS_NVA3F(device->chipset)) {
+		magic1 = 0x7ff;
+		magic2 = 0x001ffe67;
+	} else {
+		magic1 = 0x7ff;
+		magic2 = 0x00087e67;
+	}
+	xf_emit(ctx, 1, 0);		/* 00000007 ALPHA_TEST_FUNC */
+	xf_emit(ctx, 1, 0);		/* ffffffff ALPHA_TEST_REF */
+	xf_emit(ctx, 1, 0);		/* 00000001 ALPHA_TEST_ENABLE */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 1);	/* 0000000f UNK16A0 */
+	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_MASK */
+	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_BACK_OP_FAIL, ZFAIL, ZPASS */
+	xf_emit(ctx, 4, 0);		/* ffffffff BLEND_COLOR */
+	xf_emit(ctx, 1, 0);		/* 00000001 UNK19C0 */
+	xf_emit(ctx, 1, 0);		/* 00000001 UNK0FDC */
+	xf_emit(ctx, 1, 0xf);		/* 0000000f COLOR_MASK */
+	xf_emit(ctx, 7, 0);		/* 0000000f COLOR_MASK */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 LOGIC_OP_ENABLE */
+	xf_emit(ctx, 1, 0);		/* ff[NV50]/3ff[NV84+] */
+	xf_emit(ctx, 1, 4);		/* 00000007 FP_CONTROL */
+	xf_emit(ctx, 4, 0xffff);	/* 0000ffff MSAA_MASK */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_MASK */
+	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_FRONT_OP_FAIL, ZFAIL, ZPASS */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
+	xf_emit(ctx, 2, 0);		/* 00007fff WINDOW_OFFSET_XY */
+	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK19CC */
+	xf_emit(ctx, 1, 0);		/* 7 */
+	xf_emit(ctx, 1, 0);		/* 00000001 SAMPLECNT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
+	xf_emit(ctx, 1, 0);		/* ffffffff COLOR_KEY */
+	xf_emit(ctx, 1, 0);		/* 00000001 COLOR_KEY_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000007 COLOR_KEY_FORMAT */
+	xf_emit(ctx, 2, 0);		/* ffffffff SIFC_BITMAP_COLOR */
+	xf_emit(ctx, 1, 1);		/* 00000001 SIFC_BITMAP_WRITE_BIT0_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000007 ALPHA_TEST_FUNC */
+	xf_emit(ctx, 1, 0);		/* 00000001 ALPHA_TEST_ENABLE */
+	if (IS_NVA3F(device->chipset)) {
+		xf_emit(ctx, 1, 3);	/* 00000003 tesla UNK16B4 */
+		xf_emit(ctx, 1, 0);	/* 00000003 */
+		xf_emit(ctx, 1, 0);	/* 00000003 tesla UNK1298 */
+	} else if (device->chipset >= 0xa0) {
+		xf_emit(ctx, 1, 1);	/* 00000001 tesla UNK16B4 */
+		xf_emit(ctx, 1, 0);	/* 00000003 */
+	} else {
+		xf_emit(ctx, 1, 0);	/* 00000003 MULTISAMPLE_CTRL */
+	}
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
+	xf_emit(ctx, 8, 0);		/* 00000001 BLEND_ENABLE */
+	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_ALPHA */
+	xf_emit(ctx, 1, 1);		/* 00000007 BLEND_EQUATION_ALPHA */
+	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_ALPHA */
+	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_RGB */
+	xf_emit(ctx, 1, 1);		/* 00000007 BLEND_EQUATION_RGB */
+	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_RGB */
+	if (IS_NVA3F(device->chipset)) {
+		xf_emit(ctx, 1, 0);	/* 00000001 UNK12E4 */
+		xf_emit(ctx, 8, 1);	/* 00000007 IBLEND_EQUATION_RGB */
+		xf_emit(ctx, 8, 1);	/* 00000007 IBLEND_EQUATION_ALPHA */
+		xf_emit(ctx, 8, 1);	/* 00000001 IBLEND_UNK00 */
+		xf_emit(ctx, 8, 2);	/* 0000001f IBLEND_SRC_RGB */
+		xf_emit(ctx, 8, 1);	/* 0000001f IBLEND_DST_RGB */
+		xf_emit(ctx, 8, 2);	/* 0000001f IBLEND_SRC_ALPHA */
+		xf_emit(ctx, 8, 1);	/* 0000001f IBLEND_DST_ALPHA */
+		xf_emit(ctx, 1, 0);	/* 00000001 UNK1140 */
+	}
+	xf_emit(ctx, 1, 1);		/* 00000001 UNK133C */
+	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f RT_FORMAT */
+	xf_emit(ctx, 7, 0);		/* 3f/7f RT_FORMAT */
+	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
+	xf_emit(ctx, 1, 0);		/* 00000001 LOGIC_OP_ENABLE */
+	xf_emit(ctx, 1, 0);		/* ff/3ff */
+	xf_emit(ctx, 1, 4);		/* 00000007 FP_CONTROL */
+	xf_emit(ctx, 1, 0);		/* 00000003 UNK0F90 */
+	xf_emit(ctx, 1, 0);		/* 00000001 FRAMEBUFFER_SRGB */
+	xf_emit(ctx, 1, 0);		/* 7 */
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f DST_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 DST_LINEAR */
+	xf_emit(ctx, 1, 0);		/* 00000007 OPERATION */
+	xf_emit(ctx, 1, 0xcf);		/* 000000ff SIFC_FORMAT */
+	xf_emit(ctx, 1, 0xcf);		/* 000000ff DRAW_COLOR_FORMAT */
+	xf_emit(ctx, 1, 0xcf);		/* 000000ff SRC_FORMAT */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
+	xf_emit(ctx, 1, 0);		/* 7/f[NVA3] MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 8, 0);		/* 00000001 BLEND_ENABLE */
+	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_ALPHA */
+	xf_emit(ctx, 1, 1);		/* 00000007 BLEND_EQUATION_ALPHA */
+	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_ALPHA */
+	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_RGB */
+	xf_emit(ctx, 1, 1);		/* 00000007 BLEND_EQUATION_RGB */
+	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_RGB */
+	xf_emit(ctx, 1, 1);		/* 00000001 UNK133C */
+	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+	xf_emit(ctx, 8, 1);		/* 00000001 UNK19E0 */
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f RT_FORMAT */
+	xf_emit(ctx, 7, 0);		/* 3f/7f RT_FORMAT */
+	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
+	xf_emit(ctx, 1, 0xf);		/* 0000000f COLOR_MASK */
+	xf_emit(ctx, 7, 0);		/* 0000000f COLOR_MASK */
+	xf_emit(ctx, 1, magic2);	/* 001fffff tesla UNK0F78 */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f DST_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 DST_LINEAR */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
+	if (device->chipset == 0x50)
+		xf_emit(ctx, 1, 0);	/* ff */
+	else
+		xf_emit(ctx, 3, 0);	/* 1, 7, 3ff */
+	xf_emit(ctx, 1, 4);		/* 00000007 FP_CONTROL */
+	xf_emit(ctx, 1, 0);		/* 00000003 UNK0F90 */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000007 */
+	xf_emit(ctx, 1, 0);		/* 00000001 SAMPLECNT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
+	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
+	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f RT_FORMAT */
+	xf_emit(ctx, 7, 0);		/* 3f/7f RT_FORMAT */
+	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f DST_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 DST_LINEAR */
+	xf_emit(ctx, 1, 0);		/* 000fffff BLIT_DU_DX_FRACT */
+	xf_emit(ctx, 1, 1);		/* 0001ffff BLIT_DU_DX_INT */
+	xf_emit(ctx, 1, 0);		/* 000fffff BLIT_DV_DY_FRACT */
+	xf_emit(ctx, 1, 1);		/* 0001ffff BLIT_DV_DY_INT */
+	xf_emit(ctx, 1, 0);		/* ff/3ff */
+	xf_emit(ctx, 1, magic1);	/* 3ff/7ff tesla UNK0D68 */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
+	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK15B4 */
+	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000007 */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
+	xf_emit(ctx, 8, 0);		/* 0000ffff DMA_COLOR */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_GLOBAL */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_LOCAL */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_STACK */
+	xf_emit(ctx, 1, 0);		/* ff/3ff */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_DST */
+	xf_emit(ctx, 1, 0);		/* 7 */
+	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+	xf_emit(ctx, 8, 0);		/* 000000ff RT_ADDRESS_HIGH */
+	xf_emit(ctx, 8, 0);		/* ffffffff RT_LAYER_STRIDE */
+	xf_emit(ctx, 8, 0);		/* ffffffff RT_ADDRESS_LOW */
+	xf_emit(ctx, 8, 8);		/* 0000007f RT_TILE_MODE */
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f RT_FORMAT */
+	xf_emit(ctx, 7, 0);		/* 3f/7f RT_FORMAT */
+	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
+	xf_emit(ctx, 8, 0x400);		/* 0fffffff RT_HORIZ */
+	xf_emit(ctx, 8, 0x300);		/* 0000ffff RT_VERT */
+	xf_emit(ctx, 1, 1);		/* 00001fff RT_ARRAY_MODE */
+	xf_emit(ctx, 1, 0xf);		/* 0000000f COLOR_MASK */
+	xf_emit(ctx, 7, 0);		/* 0000000f COLOR_MASK */
+	xf_emit(ctx, 1, 0x20);		/* 00000fff DST_TILE_MODE */
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f DST_FORMAT */
+	xf_emit(ctx, 1, 0x100);		/* 0001ffff DST_HEIGHT */
+	xf_emit(ctx, 1, 0);		/* 000007ff DST_LAYER */
+	xf_emit(ctx, 1, 1);		/* 00000001 DST_LINEAR */
+	xf_emit(ctx, 1, 0);		/* ffffffff DST_ADDRESS_LOW */
+	xf_emit(ctx, 1, 0);		/* 000000ff DST_ADDRESS_HIGH */
+	xf_emit(ctx, 1, 0x40);		/* 0007ffff DST_PITCH */
+	xf_emit(ctx, 1, 0x100);		/* 0001ffff DST_WIDTH */
+	xf_emit(ctx, 1, 0);		/* 0000ffff */
+	xf_emit(ctx, 1, 3);		/* 00000003 tesla UNK15AC */
+	xf_emit(ctx, 1, 0);		/* ff/3ff */
+	xf_emit(ctx, 1, 0);		/* 0001ffff GP_BUILTIN_RESULT_EN */
+	xf_emit(ctx, 1, 0);		/* 00000003 UNK0F90 */
+	xf_emit(ctx, 1, 0);		/* 00000007 */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
+	xf_emit(ctx, 1, magic2);	/* 001fffff tesla UNK0F78 */
+	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
+	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+	xf_emit(ctx, 1, 2);		/* 00000003 tesla UNK143C */
+	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_ZETA */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
+	xf_emit(ctx, 2, 0);		/* ffff, ff/3ff */
+	xf_emit(ctx, 1, 0);		/* 0001ffff GP_BUILTIN_RESULT_EN */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_MASK */
+	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK15B4 */
+	xf_emit(ctx, 1, 0);		/* 00000007 */
+	xf_emit(ctx, 1, 0);		/* ffffffff ZETA_LAYER_STRIDE */
+	xf_emit(ctx, 1, 0);		/* 000000ff ZETA_ADDRESS_HIGH */
+	xf_emit(ctx, 1, 0);		/* ffffffff ZETA_ADDRESS_LOW */
+	xf_emit(ctx, 1, 4);		/* 00000007 ZETA_TILE_MODE */
+	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
+	xf_emit(ctx, 1, 0x400);		/* 0fffffff ZETA_HORIZ */
+	xf_emit(ctx, 1, 0x300);		/* 0000ffff ZETA_VERT */
+	xf_emit(ctx, 1, 0x1001);	/* 00001fff ZETA_ARRAY_MODE */
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
+	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 0);	/* 00000001 */
+	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f RT_FORMAT */
+	xf_emit(ctx, 7, 0);		/* 3f/7f RT_FORMAT */
+	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
+	xf_emit(ctx, 1, 0xf);		/* 0000000f COLOR_MASK */
+	xf_emit(ctx, 7, 0);		/* 0000000f COLOR_MASK */
+	xf_emit(ctx, 1, 0);		/* ff/3ff */
+	xf_emit(ctx, 8, 0);		/* 00000001 BLEND_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000003 UNK0F90 */
+	xf_emit(ctx, 1, 0);		/* 00000001 FRAMEBUFFER_SRGB */
+	xf_emit(ctx, 1, 0);		/* 7 */
+	xf_emit(ctx, 1, 0);		/* 00000001 LOGIC_OP_ENABLE */
+	if (IS_NVA3F(device->chipset)) {
+		xf_emit(ctx, 1, 0);	/* 00000001 UNK1140 */
+		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
+	}
+	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 1, 0);		/* 00000001 UNK1534 */
+	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+	if (device->chipset >= 0xa0)
+		xf_emit(ctx, 1, 0x0fac6881);	/* fffffff */
+	xf_emit(ctx, 1, magic2);	/* 001fffff tesla UNK0F78 */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
+	xf_emit(ctx, 1, 0x11);		/* 3f/7f DST_FORMAT */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK0FB0 */
+	xf_emit(ctx, 1, 0);		/* ff/3ff */
+	xf_emit(ctx, 1, 4);		/* 00000007 FP_CONTROL */
+	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
+	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK15B4 */
+	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK19CC */
+	xf_emit(ctx, 1, 0);		/* 00000007 */
+	xf_emit(ctx, 1, 0);		/* 00000001 SAMPLECNT_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
+	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
+	if (IS_NVA3F(device->chipset)) {
+		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
+		xf_emit(ctx, 1, 0);	/* 0000000f tesla UNK15C8 */
+	}
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
+	if (device->chipset >= 0xa0) {
+		xf_emit(ctx, 3, 0);		/* 7/f, 1, ffff0ff3 */
+		xf_emit(ctx, 1, 0xfac6881);	/* fffffff */
+		xf_emit(ctx, 4, 0);		/* 1, 1, 1, 3ff */
+		xf_emit(ctx, 1, 4);		/* 7 */
+		xf_emit(ctx, 1, 0);		/* 1 */
+		xf_emit(ctx, 2, 1);		/* 1 */
+		xf_emit(ctx, 2, 0);		/* 7, f */
+		xf_emit(ctx, 1, 1);		/* 1 */
+		xf_emit(ctx, 1, 0);		/* 7/f */
+		if (IS_NVA3F(device->chipset))
+			xf_emit(ctx, 0x9, 0);	/* 1 */
+		else
+			xf_emit(ctx, 0x8, 0);	/* 1 */
+		xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+		xf_emit(ctx, 8, 1);		/* 1 */
+		xf_emit(ctx, 1, 0x11);		/* 7f */
+		xf_emit(ctx, 7, 0);		/* 7f */
+		xf_emit(ctx, 1, 0xfac6881);	/* fffffff */
+		xf_emit(ctx, 1, 0xf);		/* f */
+		xf_emit(ctx, 7, 0);		/* f */
+		xf_emit(ctx, 1, 0x11);		/* 7f */
+		xf_emit(ctx, 1, 1);		/* 1 */
+		xf_emit(ctx, 5, 0);		/* 1, 7, 3ff, 3, 7 */
+		if (IS_NVA3F(device->chipset)) {
+			xf_emit(ctx, 1, 0);	/* 00000001 UNK1140 */
+			xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
+		}
+	}
+}
+
+static void
+nv50_graph_construct_xfer_tex(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	xf_emit(ctx, 2, 0);		/* 1 LINKED_TSC. yes, 2. */
+	if (device->chipset != 0x50)
+		xf_emit(ctx, 1, 0);	/* 3 */
+	xf_emit(ctx, 1, 1);		/* 1ffff BLIT_DU_DX_INT */
+	xf_emit(ctx, 1, 0);		/* fffff BLIT_DU_DX_FRACT */
+	xf_emit(ctx, 1, 1);		/* 1ffff BLIT_DV_DY_INT */
+	xf_emit(ctx, 1, 0);		/* fffff BLIT_DV_DY_FRACT */
+	if (device->chipset == 0x50)
+		xf_emit(ctx, 1, 0);	/* 3 BLIT_CONTROL */
+	else
+		xf_emit(ctx, 2, 0);	/* 3ff, 1 */
+	xf_emit(ctx, 1, 0x2a712488);	/* ffffffff SRC_TIC_0 */
+	xf_emit(ctx, 1, 0);		/* ffffffff SRC_TIC_1 */
+	xf_emit(ctx, 1, 0x4085c000);	/* ffffffff SRC_TIC_2 */
+	xf_emit(ctx, 1, 0x40);		/* ffffffff SRC_TIC_3 */
+	xf_emit(ctx, 1, 0x100);		/* ffffffff SRC_TIC_4 */
+	xf_emit(ctx, 1, 0x10100);	/* ffffffff SRC_TIC_5 */
+	xf_emit(ctx, 1, 0x02800000);	/* ffffffff SRC_TIC_6 */
+	xf_emit(ctx, 1, 0);		/* ffffffff SRC_TIC_7 */
+	if (device->chipset == 0x50) {
+		xf_emit(ctx, 1, 0);	/* 00000001 turing UNK358 */
+		xf_emit(ctx, 1, 0);	/* ffffffff tesla UNK1A34? */
+		xf_emit(ctx, 1, 0);	/* 00000003 turing UNK37C tesla UNK1690 */
+		xf_emit(ctx, 1, 0);	/* 00000003 BLIT_CONTROL */
+		xf_emit(ctx, 1, 0);	/* 00000001 turing UNK32C tesla UNK0F94 */
+	} else if (!IS_NVAAF(device->chipset)) {
+		xf_emit(ctx, 1, 0);	/* ffffffff tesla UNK1A34? */
+		xf_emit(ctx, 1, 0);	/* 00000003 */
+		xf_emit(ctx, 1, 0);	/* 000003ff */
+		xf_emit(ctx, 1, 0);	/* 00000003 */
+		xf_emit(ctx, 1, 0);	/* 000003ff */
+		xf_emit(ctx, 1, 0);	/* 00000003 tesla UNK1664 / turing UNK03E8 */
+		xf_emit(ctx, 1, 0);	/* 00000003 */
+		xf_emit(ctx, 1, 0);	/* 000003ff */
+	} else {
+		xf_emit(ctx, 0x6, 0);
+	}
+	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A34 */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_TEXTURE */
+	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_SRC */
+}
+
+static void
+nv50_graph_construct_xfer_unk8cxx(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	xf_emit(ctx, 1, 0);		/* 00000001 UNK1534 */
+	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 2, 0);		/* 7, ffff0ff3 */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE */
+	xf_emit(ctx, 1, 0x04e3bfdf);	/* ffffffff UNK0D64 */
+	xf_emit(ctx, 1, 0x04e3bfdf);	/* ffffffff UNK0DF4 */
+	xf_emit(ctx, 1, 1);		/* 00000001 UNK15B4 */
+	xf_emit(ctx, 1, 0);		/* 00000001 LINE_STIPPLE_ENABLE */
+	xf_emit(ctx, 1, 0x00ffff00);	/* 00ffffff LINE_STIPPLE_PATTERN */
+	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK0F98 */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
+	xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK1668 */
+	xf_emit(ctx, 1, 0);		/* 00000001 LINE_STIPPLE_ENABLE */
+	xf_emit(ctx, 1, 0x00ffff00);	/* 00ffffff LINE_STIPPLE_PATTERN */
+	xf_emit(ctx, 1, 0);		/* 00000001 POLYGON_SMOOTH_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 UNK1534 */
+	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1658 */
+	xf_emit(ctx, 1, 0);		/* 00000001 LINE_SMOOTH_ENABLE */
+	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE */
+	xf_emit(ctx, 1, 1);		/* 00000001 UNK15B4 */
+	xf_emit(ctx, 1, 0);		/* 00000001 POINT_SPRITE_ENABLE */
+	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK165C */
+	xf_emit(ctx, 1, 0x30201000);	/* ffffffff tesla UNK1670 */
+	xf_emit(ctx, 1, 0x70605040);	/* ffffffff tesla UNK1670 */
+	xf_emit(ctx, 1, 0xb8a89888);	/* ffffffff tesla UNK1670 */
+	xf_emit(ctx, 1, 0xf8e8d8c8);	/* ffffffff tesla UNK1670 */
+	xf_emit(ctx, 1, 0);		/* 00000001 VERTEX_TWO_SIDE_ENABLE */
+	xf_emit(ctx, 1, 0x1a);		/* 0000001f POLYGON_MODE */
+}
+
+static void
+nv50_graph_construct_xfer_tp(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	if (device->chipset < 0xa0) {
+		nv50_graph_construct_xfer_unk84xx(ctx);
+		nv50_graph_construct_xfer_tprop(ctx);
+		nv50_graph_construct_xfer_tex(ctx);
+		nv50_graph_construct_xfer_unk8cxx(ctx);
+	} else {
+		nv50_graph_construct_xfer_tex(ctx);
+		nv50_graph_construct_xfer_tprop(ctx);
+		nv50_graph_construct_xfer_unk8cxx(ctx);
+		nv50_graph_construct_xfer_unk84xx(ctx);
+	}
+}
+
+static void
+nv50_graph_construct_xfer_mpc(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int i, mpcnt = 2;
+	switch (device->chipset) {
+		case 0x98:
+		case 0xaa:
+			mpcnt = 1;
+			break;
+		case 0x50:
+		case 0x84:
+		case 0x86:
+		case 0x92:
+		case 0x94:
+		case 0x96:
+		case 0xa8:
+		case 0xac:
+			mpcnt = 2;
+			break;
+		case 0xa0:
+		case 0xa3:
+		case 0xa5:
+		case 0xaf:
+			mpcnt = 3;
+			break;
+	}
+	for (i = 0; i < mpcnt; i++) {
+		xf_emit(ctx, 1, 0);		/* ff */
+		xf_emit(ctx, 1, 0x80);		/* ffffffff tesla UNK1404 */
+		xf_emit(ctx, 1, 0x80007004);	/* ffffffff tesla UNK12B0 */
+		xf_emit(ctx, 1, 0x04000400);	/* ffffffff */
+		if (device->chipset >= 0xa0)
+			xf_emit(ctx, 1, 0xc0);	/* 00007fff tesla UNK152C */
+		xf_emit(ctx, 1, 0x1000);	/* 0000ffff tesla UNK0D60 */
+		xf_emit(ctx, 1, 0);		/* ff/3ff */
+		xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
+		if (device->chipset == 0x86 || device->chipset == 0x98 || device->chipset == 0xa8 || IS_NVAAF(device->chipset)) {
+			xf_emit(ctx, 1, 0xe00);		/* 7fff */
+			xf_emit(ctx, 1, 0x1e00);	/* 7fff */
+		}
+		xf_emit(ctx, 1, 1);		/* 000000ff VP_REG_ALLOC_TEMP */
+		xf_emit(ctx, 1, 0);		/* 00000001 LINKED_TSC */
+		xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+		if (device->chipset == 0x50)
+			xf_emit(ctx, 2, 0x1000);	/* 7fff tesla UNK141C */
+		xf_emit(ctx, 1, 1);		/* 000000ff GP_REG_ALLOC_TEMP */
+		xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
+		xf_emit(ctx, 1, 4);		/* 000000ff FP_REG_ALLOC_TEMP */
+		xf_emit(ctx, 1, 2);		/* 00000003 REG_MODE */
+		if (IS_NVAAF(device->chipset))
+			xf_emit(ctx, 0xb, 0);	/* RO */
+		else if (device->chipset >= 0xa0)
+			xf_emit(ctx, 0xc, 0);	/* RO */
+		else
+			xf_emit(ctx, 0xa, 0);	/* RO */
+	}
+	xf_emit(ctx, 1, 0x08100c12);		/* 1fffffff FP_INTERPOLANT_CTRL */
+	xf_emit(ctx, 1, 0);			/* ff/3ff */
+	if (device->chipset >= 0xa0) {
+		xf_emit(ctx, 1, 0x1fe21);	/* 0003ffff tesla UNK0FAC */
+	}
+	xf_emit(ctx, 3, 0);			/* 7fff, 0, 0 */
+	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1534 */
+	xf_emit(ctx, 1, 0);			/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
+	xf_emit(ctx, 4, 0xffff);		/* 0000ffff MSAA_MASK */
+	xf_emit(ctx, 1, 1);			/* 00000001 LANES32 */
+	xf_emit(ctx, 1, 0x10001);		/* 00ffffff BLOCK_ALLOC */
+	xf_emit(ctx, 1, 0x10001);		/* ffffffff BLOCKDIM_XY */
+	xf_emit(ctx, 1, 1);			/* 0000ffff BLOCKDIM_Z */
+	xf_emit(ctx, 1, 0);			/* ffffffff SHARED_SIZE */
+	xf_emit(ctx, 1, 0x1fe21);		/* 1ffff/3ffff[NVA0+] tesla UNk0FAC */
+	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A34 */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 1);		/* 0000001f tesla UNK169C */
+	xf_emit(ctx, 1, 0);			/* ff/3ff */
+	xf_emit(ctx, 1, 0);			/* 1 LINKED_TSC */
+	xf_emit(ctx, 1, 0);			/* ff FP_ADDRESS_HIGH */
+	xf_emit(ctx, 1, 0);			/* ffffffff FP_ADDRESS_LOW */
+	xf_emit(ctx, 1, 0x08100c12);		/* 1fffffff FP_INTERPOLANT_CTRL */
+	xf_emit(ctx, 1, 4);			/* 00000007 FP_CONTROL */
+	xf_emit(ctx, 1, 0);			/* 000000ff FRAG_COLOR_CLAMP_EN */
+	xf_emit(ctx, 1, 2);			/* 00000003 REG_MODE */
+	xf_emit(ctx, 1, 0x11);			/* 0000007f RT_FORMAT */
+	xf_emit(ctx, 7, 0);			/* 0000007f RT_FORMAT */
+	xf_emit(ctx, 1, 0);			/* 00000007 */
+	xf_emit(ctx, 1, 0xfac6881);		/* 0fffffff RT_CONTROL */
+	xf_emit(ctx, 1, 0);			/* 00000003 MULTISAMPLE_CTRL */
+	if (IS_NVA3F(device->chipset))
+		xf_emit(ctx, 1, 3);		/* 00000003 tesla UNK16B4 */
+	xf_emit(ctx, 1, 0);			/* 00000001 ALPHA_TEST_ENABLE */
+	xf_emit(ctx, 1, 0);			/* 00000007 ALPHA_TEST_FUNC */
+	xf_emit(ctx, 1, 0);			/* 00000001 FRAMEBUFFER_SRGB */
+	xf_emit(ctx, 1, 4);			/* ffffffff tesla UNK1400 */
+	xf_emit(ctx, 8, 0);			/* 00000001 BLEND_ENABLE */
+	xf_emit(ctx, 1, 0);			/* 00000001 LOGIC_OP_ENABLE */
+	xf_emit(ctx, 1, 2);			/* 0000001f BLEND_FUNC_SRC_RGB */
+	xf_emit(ctx, 1, 1);			/* 0000001f BLEND_FUNC_DST_RGB */
+	xf_emit(ctx, 1, 1);			/* 00000007 BLEND_EQUATION_RGB */
+	xf_emit(ctx, 1, 2);			/* 0000001f BLEND_FUNC_SRC_ALPHA */
+	xf_emit(ctx, 1, 1);			/* 0000001f BLEND_FUNC_DST_ALPHA */
+	xf_emit(ctx, 1, 1);			/* 00000007 BLEND_EQUATION_ALPHA */
+	xf_emit(ctx, 1, 1);			/* 00000001 UNK133C */
+	if (IS_NVA3F(device->chipset)) {
+		xf_emit(ctx, 1, 0);		/* 00000001 UNK12E4 */
+		xf_emit(ctx, 8, 2);		/* 0000001f IBLEND_FUNC_SRC_RGB */
+		xf_emit(ctx, 8, 1);		/* 0000001f IBLEND_FUNC_DST_RGB */
+		xf_emit(ctx, 8, 1);		/* 00000007 IBLEND_EQUATION_RGB */
+		xf_emit(ctx, 8, 2);		/* 0000001f IBLEND_FUNC_SRC_ALPHA */
+		xf_emit(ctx, 8, 1);		/* 0000001f IBLEND_FUNC_DST_ALPHA */
+		xf_emit(ctx, 8, 1);		/* 00000007 IBLEND_EQUATION_ALPHA */
+		xf_emit(ctx, 8, 1);		/* 00000001 IBLEND_UNK00 */
+		xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK1928 */
+		xf_emit(ctx, 1, 0);		/* 00000001 UNK1140 */
+	}
+	xf_emit(ctx, 1, 0);			/* 00000003 tesla UNK0F90 */
+	xf_emit(ctx, 1, 4);			/* 000000ff FP_RESULT_COUNT */
+	/* XXX: demagic this part some day */
+	if (device->chipset == 0x50)
+		xf_emit(ctx, 0x3a0, 0);
+	else if (device->chipset < 0x94)
+		xf_emit(ctx, 0x3a2, 0);
+	else if (device->chipset == 0x98 || device->chipset == 0xaa)
+		xf_emit(ctx, 0x39f, 0);
+	else
+		xf_emit(ctx, 0x3a3, 0);
+	xf_emit(ctx, 1, 0x11);			/* 3f/7f DST_FORMAT */
+	xf_emit(ctx, 1, 0);			/* 7 OPERATION */
+	xf_emit(ctx, 1, 1);			/* 1 DST_LINEAR */
+	xf_emit(ctx, 0x2d, 0);
+}
+
+static void
+nv50_graph_construct_xfer2(struct nouveau_grctx *ctx)
+{
+	struct nouveau_device *device = ctx->device;
+	int i;
+	u32 offset;
+	u32 units = nv_rd32 (ctx->device, 0x1540);
+	int size = 0;
+
+	offset = (ctx->ctxvals_pos+0x3f)&~0x3f;
+
+	if (device->chipset < 0xa0) {
+		for (i = 0; i < 8; i++) {
+			ctx->ctxvals_pos = offset + i;
+			/* that little bugger belongs to csched. No idea
+			 * what it's doing here. */
+			if (i == 0)
+				xf_emit(ctx, 1, 0x08100c12); /* FP_INTERPOLANT_CTRL */
+			if (units & (1 << i))
+				nv50_graph_construct_xfer_mpc(ctx);
+			if ((ctx->ctxvals_pos-offset)/8 > size)
+				size = (ctx->ctxvals_pos-offset)/8;
+		}
+	} else {
+		/* Strand 0: TPs 0, 1 */
+		ctx->ctxvals_pos = offset;
+		/* that little bugger belongs to csched. No idea
+		 * what it's doing here. */
+		xf_emit(ctx, 1, 0x08100c12); /* FP_INTERPOLANT_CTRL */
+		if (units & (1 << 0))
+			nv50_graph_construct_xfer_mpc(ctx);
+		if (units & (1 << 1))
+			nv50_graph_construct_xfer_mpc(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 1: TPs 2, 3 */
+		ctx->ctxvals_pos = offset + 1;
+		if (units & (1 << 2))
+			nv50_graph_construct_xfer_mpc(ctx);
+		if (units & (1 << 3))
+			nv50_graph_construct_xfer_mpc(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 2: TPs 4, 5, 6 */
+		ctx->ctxvals_pos = offset + 2;
+		if (units & (1 << 4))
+			nv50_graph_construct_xfer_mpc(ctx);
+		if (units & (1 << 5))
+			nv50_graph_construct_xfer_mpc(ctx);
+		if (units & (1 << 6))
+			nv50_graph_construct_xfer_mpc(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 3: TPs 7, 8, 9 */
+		ctx->ctxvals_pos = offset + 3;
+		if (units & (1 << 7))
+			nv50_graph_construct_xfer_mpc(ctx);
+		if (units & (1 << 8))
+			nv50_graph_construct_xfer_mpc(ctx);
+		if (units & (1 << 9))
+			nv50_graph_construct_xfer_mpc(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+	}
+	ctx->ctxvals_pos = offset + size * 8;
+	ctx->ctxvals_pos = (ctx->ctxvals_pos+0x3f)&~0x3f;
+	cp_lsr (ctx, offset);
+	cp_out (ctx, CP_SET_XFER_POINTER);
+	cp_lsr (ctx, size);
+	cp_out (ctx, CP_SEEK_2);
+	cp_out (ctx, CP_XFER_2);
+	cp_wait(ctx, XFER, BUSY);
+}
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/ctxnvc0.c b/drivers/gpu/drm/nouveau/core/engine/graph/ctxnvc0.c
new file mode 100644
index 0000000..0b7951a
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/ctxnvc0.c
@@ -0,0 +1,3039 @@
+/*
+ * Copyright 2010 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include "nvc0.h"
+
+void
+nv_icmd(struct nvc0_graph_priv *priv, u32 icmd, u32 data)
+{
+	nv_wr32(priv, 0x400204, data);
+	nv_wr32(priv, 0x400200, icmd);
+	while (nv_rd32(priv, 0x400700) & 2) {}
+}
+
+int
+nvc0_grctx_init(struct nvc0_graph_priv *priv, struct nvc0_grctx *info)
+{
+	struct nouveau_bar *bar = nouveau_bar(priv);
+	struct nouveau_object *parent = nv_object(priv);
+	struct nouveau_gpuobj *chan;
+	u32 size = (0x80000 + priv->size + 4095) & ~4095;
+	int ret, i;
+
+	/* allocate memory to for a "channel", which we'll use to generate
+	 * the default context values
+	 */
+	ret = nouveau_gpuobj_new(parent, NULL, size, 0x1000,
+				 NVOBJ_FLAG_ZERO_ALLOC, &info->chan);
+	chan = info->chan;
+	if (ret) {
+		nv_error(priv, "failed to allocate channel memory, %d\n", ret);
+		return ret;
+	}
+
+	/* PGD pointer */
+	nv_wo32(chan, 0x0200, lower_32_bits(chan->addr + 0x1000));
+	nv_wo32(chan, 0x0204, upper_32_bits(chan->addr + 0x1000));
+	nv_wo32(chan, 0x0208, 0xffffffff);
+	nv_wo32(chan, 0x020c, 0x000000ff);
+
+	/* PGT[0] pointer */
+	nv_wo32(chan, 0x1000, 0x00000000);
+	nv_wo32(chan, 0x1004, 0x00000001 | (chan->addr + 0x2000) >> 8);
+
+	/* identity-map the whole "channel" into its own vm */
+	for (i = 0; i < size / 4096; i++) {
+		u64 addr = ((chan->addr + (i * 4096)) >> 8) | 1;
+		nv_wo32(chan, 0x2000 + (i * 8), lower_32_bits(addr));
+		nv_wo32(chan, 0x2004 + (i * 8), upper_32_bits(addr));
+	}
+
+	/* context pointer (virt) */
+	nv_wo32(chan, 0x0210, 0x00080004);
+	nv_wo32(chan, 0x0214, 0x00000000);
+
+	bar->flush(bar);
+
+	nv_wr32(priv, 0x100cb8, (chan->addr + 0x1000) >> 8);
+	nv_wr32(priv, 0x100cbc, 0x80000001);
+	nv_wait(priv, 0x100c80, 0x00008000, 0x00008000);
+
+	/* setup default state for mmio list construction */
+	info->data = priv->mmio_data;
+	info->mmio = priv->mmio_list;
+	info->addr = 0x2000 + (i * 8);
+	info->priv = priv;
+	info->buffer_nr = 0;
+
+	if (priv->firmware) {
+		nv_wr32(priv, 0x409840, 0x00000030);
+		nv_wr32(priv, 0x409500, 0x80000000 | chan->addr >> 12);
+		nv_wr32(priv, 0x409504, 0x00000003);
+		if (!nv_wait(priv, 0x409800, 0x00000010, 0x00000010))
+			nv_error(priv, "load_ctx timeout\n");
+
+		nv_wo32(chan, 0x8001c, 1);
+		nv_wo32(chan, 0x80020, 0);
+		nv_wo32(chan, 0x80028, 0);
+		nv_wo32(chan, 0x8002c, 0);
+		bar->flush(bar);
+		return 0;
+	}
+
+	/* HUB_FUC(SET_CHAN) */
+	nv_wr32(priv, 0x409840, 0x80000000);
+	nv_wr32(priv, 0x409500, 0x80000000 | chan->addr >> 12);
+	nv_wr32(priv, 0x409504, 0x00000001);
+	if (!nv_wait(priv, 0x409800, 0x80000000, 0x80000000)) {
+		nv_error(priv, "HUB_SET_CHAN timeout\n");
+		nvc0_graph_ctxctl_debug(priv);
+		nouveau_gpuobj_ref(NULL, &info->chan);
+		return -EBUSY;
+	}
+
+	return 0;
+}
+
+void
+nvc0_grctx_data(struct nvc0_grctx *info, u32 size, u32 align, u32 access)
+{
+	info->buffer[info->buffer_nr]  = info->addr;
+	info->buffer[info->buffer_nr] +=  (align - 1);
+	info->buffer[info->buffer_nr] &= ~(align - 1);
+	info->addr = info->buffer[info->buffer_nr++] + size;
+
+	info->data->size = size;
+	info->data->align = align;
+	info->data->access = access;
+	info->data++;
+}
+
+void
+nvc0_grctx_mmio(struct nvc0_grctx *info, u32 addr, u32 data, u32 shift, u32 buf)
+{
+	struct nvc0_graph_priv *priv = info->priv;
+
+	info->mmio->addr = addr;
+	info->mmio->data = data;
+	info->mmio->shift = shift;
+	info->mmio->buffer = buf;
+	info->mmio++;
+
+	if (shift)
+		data |= info->buffer[buf] >> shift;
+	nv_wr32(priv, addr, data);
+}
+
+int
+nvc0_grctx_fini(struct nvc0_grctx *info)
+{
+	struct nvc0_graph_priv *priv = info->priv;
+	int i;
+
+	/* trigger a context unload by unsetting the "next channel valid" bit
+	 * and faking a context switch interrupt
+	 */
+	nv_mask(priv, 0x409b04, 0x80000000, 0x00000000);
+	nv_wr32(priv, 0x409000, 0x00000100);
+	if (!nv_wait(priv, 0x409b00, 0x80000000, 0x00000000)) {
+		nv_error(priv, "grctx template channel unload timeout\n");
+		return -EBUSY;
+	}
+
+	priv->data = kmalloc(priv->size, GFP_KERNEL);
+	if (priv->data) {
+		for (i = 0; i < priv->size; i += 4)
+			priv->data[i / 4] = nv_ro32(info->chan, 0x80000 + i);
+	}
+
+	nouveau_gpuobj_ref(NULL, &info->chan);
+	return priv->data ? 0 : -ENOMEM;
+}
+
+static void
+nvc0_grctx_generate_9097(struct nvc0_graph_priv *priv)
+{
+	u32 fermi = nvc0_graph_class(priv);
+	u32 mthd;
+
+	nv_mthd(priv, 0x9097, 0x0800, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0840, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0880, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x08c0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0900, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0940, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0980, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x09c0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0804, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0844, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0884, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x08c4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0904, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0944, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0984, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x09c4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0808, 0x00000400);
+	nv_mthd(priv, 0x9097, 0x0848, 0x00000400);
+	nv_mthd(priv, 0x9097, 0x0888, 0x00000400);
+	nv_mthd(priv, 0x9097, 0x08c8, 0x00000400);
+	nv_mthd(priv, 0x9097, 0x0908, 0x00000400);
+	nv_mthd(priv, 0x9097, 0x0948, 0x00000400);
+	nv_mthd(priv, 0x9097, 0x0988, 0x00000400);
+	nv_mthd(priv, 0x9097, 0x09c8, 0x00000400);
+	nv_mthd(priv, 0x9097, 0x080c, 0x00000300);
+	nv_mthd(priv, 0x9097, 0x084c, 0x00000300);
+	nv_mthd(priv, 0x9097, 0x088c, 0x00000300);
+	nv_mthd(priv, 0x9097, 0x08cc, 0x00000300);
+	nv_mthd(priv, 0x9097, 0x090c, 0x00000300);
+	nv_mthd(priv, 0x9097, 0x094c, 0x00000300);
+	nv_mthd(priv, 0x9097, 0x098c, 0x00000300);
+	nv_mthd(priv, 0x9097, 0x09cc, 0x00000300);
+	nv_mthd(priv, 0x9097, 0x0810, 0x000000cf);
+	nv_mthd(priv, 0x9097, 0x0850, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0890, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x08d0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0910, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0950, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0990, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x09d0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0814, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x0854, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x0894, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x08d4, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x0914, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x0954, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x0994, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x09d4, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x0818, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x0858, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x0898, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x08d8, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x0918, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x0958, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x0998, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x09d8, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x081c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x085c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x089c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x08dc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x091c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x095c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x099c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x09dc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0820, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0860, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x08a0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x08e0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0920, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0960, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x09a0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x09e0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2700, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2720, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2740, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2760, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2780, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x27a0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x27c0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x27e0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2704, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2724, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2744, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2764, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2784, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x27a4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x27c4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x27e4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2708, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2728, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2748, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2768, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2788, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x27a8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x27c8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x27e8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x270c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x272c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x274c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x276c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x278c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x27ac, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x27cc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x27ec, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2710, 0x00014000);
+	nv_mthd(priv, 0x9097, 0x2730, 0x00014000);
+	nv_mthd(priv, 0x9097, 0x2750, 0x00014000);
+	nv_mthd(priv, 0x9097, 0x2770, 0x00014000);
+	nv_mthd(priv, 0x9097, 0x2790, 0x00014000);
+	nv_mthd(priv, 0x9097, 0x27b0, 0x00014000);
+	nv_mthd(priv, 0x9097, 0x27d0, 0x00014000);
+	nv_mthd(priv, 0x9097, 0x27f0, 0x00014000);
+	nv_mthd(priv, 0x9097, 0x2714, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x2734, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x2754, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x2774, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x2794, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x27b4, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x27d4, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x27f4, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x1c00, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c10, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c20, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c30, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c40, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c50, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c60, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c70, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c80, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c90, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1ca0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cb0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cc0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cd0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1ce0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cf0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c04, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c14, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c24, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c34, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c44, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c54, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c64, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c74, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c84, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c94, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1ca4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cb4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cc4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cd4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1ce4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cf4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c08, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c18, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c28, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c38, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c48, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c58, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c68, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c78, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c88, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c98, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1ca8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cb8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cc8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cd8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1ce8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cf8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c0c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c1c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c2c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c3c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c4c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c5c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c6c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c7c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c8c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1c9c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cac, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cbc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1ccc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cdc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cec, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1cfc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d00, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d10, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d20, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d30, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d40, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d50, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d60, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d70, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d80, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d90, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1da0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1db0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1dc0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1dd0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1de0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1df0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d04, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d14, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d24, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d34, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d44, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d54, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d64, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d74, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d84, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d94, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1da4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1db4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1dc4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1dd4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1de4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1df4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d08, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d18, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d28, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d38, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d48, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d58, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d68, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d78, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d88, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d98, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1da8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1db8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1dc8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1dd8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1de8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1df8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d0c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d1c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d2c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d3c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d4c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d5c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d6c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d7c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d8c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1d9c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1dac, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1dbc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1dcc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1ddc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1dec, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1dfc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f00, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f08, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f10, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f18, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f20, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f28, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f30, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f38, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f40, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f48, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f50, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f58, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f60, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f68, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f70, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f78, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f04, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f0c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f14, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f1c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f24, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f2c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f34, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f3c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f44, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f4c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f54, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f5c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f64, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f6c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f74, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f7c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f80, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f88, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f90, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f98, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fa0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fa8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fb0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fb8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fc0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fc8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fd0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fd8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fe0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fe8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1ff0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1ff8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f84, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f8c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f94, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1f9c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fa4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fac, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fb4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fbc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fc4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fcc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fd4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fdc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fe4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1fec, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1ff4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1ffc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2200, 0x00000022);
+	nv_mthd(priv, 0x9097, 0x2210, 0x00000022);
+	nv_mthd(priv, 0x9097, 0x2220, 0x00000022);
+	nv_mthd(priv, 0x9097, 0x2230, 0x00000022);
+	nv_mthd(priv, 0x9097, 0x2240, 0x00000022);
+	nv_mthd(priv, 0x9097, 0x2000, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2040, 0x00000011);
+	nv_mthd(priv, 0x9097, 0x2080, 0x00000020);
+	nv_mthd(priv, 0x9097, 0x20c0, 0x00000030);
+	nv_mthd(priv, 0x9097, 0x2100, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x2140, 0x00000051);
+	nv_mthd(priv, 0x9097, 0x200c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x204c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x208c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x20cc, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x210c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x214c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x2010, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2050, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2090, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x20d0, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x2110, 0x00000003);
+	nv_mthd(priv, 0x9097, 0x2150, 0x00000004);
+	nv_mthd(priv, 0x9097, 0x0380, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x03a0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x03c0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x03e0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0384, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x03a4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x03c4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x03e4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0388, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x03a8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x03c8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x03e8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x038c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x03ac, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x03cc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x03ec, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0700, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0710, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0720, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0730, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0704, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0714, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0724, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0734, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0708, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0718, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0728, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0738, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2800, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2804, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2808, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x280c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2810, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2814, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2818, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x281c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2820, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2824, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2828, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x282c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2830, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2834, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2838, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x283c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2840, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2844, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2848, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x284c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2850, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2854, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2858, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x285c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2860, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2864, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2868, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x286c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2870, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2874, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2878, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x287c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2880, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2884, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2888, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x288c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2890, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2894, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2898, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x289c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28a0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28a4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28a8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28ac, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28b0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28b4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28b8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28bc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28c0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28c4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28c8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28cc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28d0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28d4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28d8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28dc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28e0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28e4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28e8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28ec, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28f0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28f4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28f8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x28fc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2900, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2904, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2908, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x290c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2910, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2914, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2918, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x291c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2920, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2924, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2928, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x292c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2930, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2934, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2938, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x293c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2940, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2944, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2948, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x294c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2950, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2954, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2958, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x295c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2960, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2964, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2968, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x296c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2970, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2974, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2978, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x297c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2980, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2984, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2988, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x298c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2990, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2994, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2998, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x299c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29a0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29a4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29a8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29ac, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29b0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29b4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29b8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29bc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29c0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29c4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29c8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29cc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29d0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29d4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29d8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29dc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29e0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29e4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29e8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29ec, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29f0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29f4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29f8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x29fc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a00, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a20, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a40, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a60, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a80, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0aa0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ac0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ae0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b00, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b20, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b40, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b60, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b80, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ba0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0bc0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0be0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a04, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a24, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a44, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a64, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a84, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0aa4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ac4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ae4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b04, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b24, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b44, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b64, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b84, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ba4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0bc4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0be4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a08, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a28, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a48, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a68, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a88, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0aa8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ac8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ae8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b08, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b28, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b48, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b68, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b88, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ba8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0bc8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0be8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a0c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a2c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a4c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a6c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a8c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0aac, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0acc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0aec, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b0c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b2c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b4c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b6c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b8c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0bac, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0bcc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0bec, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a10, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a30, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a50, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a70, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a90, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ab0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ad0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0af0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b10, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b30, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b50, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b70, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b90, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0bb0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0bd0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0bf0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a14, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a34, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a54, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a74, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0a94, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ab4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ad4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0af4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b14, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b34, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b54, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b74, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0b94, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0bb4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0bd4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0bf4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c00, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c10, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c20, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c30, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c40, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c50, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c60, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c70, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c80, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c90, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ca0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0cb0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0cc0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0cd0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ce0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0cf0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c04, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c14, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c24, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c34, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c44, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c54, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c64, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c74, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c84, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c94, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ca4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0cb4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0cc4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0cd4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ce4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0cf4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c08, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c18, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c28, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c38, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c48, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c58, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c68, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c78, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c88, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c98, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ca8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0cb8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0cc8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0cd8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ce8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0cf8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0c0c, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0c1c, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0c2c, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0c3c, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0c4c, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0c5c, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0c6c, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0c7c, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0c8c, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0c9c, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0cac, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0cbc, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0ccc, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0cdc, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0cec, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0cfc, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0d00, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d08, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d10, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d18, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d20, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d28, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d30, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d38, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d04, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d0c, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d14, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d1c, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d24, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d2c, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d34, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d3c, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e00, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0e10, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0e20, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0e30, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0e40, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0e50, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0e60, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0e70, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0e80, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0e90, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ea0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0eb0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ec0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ed0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ee0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ef0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0e04, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e14, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e24, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e34, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e44, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e54, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e64, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e74, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e84, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e94, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0ea4, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0eb4, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0ec4, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0ed4, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0ee4, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0ef4, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e08, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e18, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e28, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e38, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e48, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e58, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e68, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e78, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e88, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0e98, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0ea8, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0eb8, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0ec8, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0ed8, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0ee8, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0ef8, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d40, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d48, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d50, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d58, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d44, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d4c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d54, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d5c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1e00, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e20, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e40, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e60, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e80, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1ea0, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1ec0, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1ee0, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e04, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e24, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e44, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e64, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e84, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1ea4, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1ec4, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1ee4, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e08, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1e28, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1e48, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1e68, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1e88, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1ea8, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1ec8, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1ee8, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1e0c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e2c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e4c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e6c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e8c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1eac, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1ecc, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1eec, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e10, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e30, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e50, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e70, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e90, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1eb0, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1ed0, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1ef0, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e14, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1e34, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1e54, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1e74, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1e94, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1eb4, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1ed4, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1ef4, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1e18, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e38, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e58, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e78, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1e98, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1eb8, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1ed8, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1ef8, 0x00000001);
+	if (fermi == 0x9097) {
+		for (mthd = 0x3400; mthd <= 0x35fc; mthd += 4)
+			nv_mthd(priv, 0x9097, mthd, 0x00000000);
+	}
+	nv_mthd(priv, 0x9097, 0x030c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1944, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1514, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d68, 0x0000ffff);
+	nv_mthd(priv, 0x9097, 0x121c, 0x0fac6881);
+	nv_mthd(priv, 0x9097, 0x0fac, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1538, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x0fe0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0fe4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0fe8, 0x00000014);
+	nv_mthd(priv, 0x9097, 0x0fec, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x0ff0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x179c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1228, 0x00000400);
+	nv_mthd(priv, 0x9097, 0x122c, 0x00000300);
+	nv_mthd(priv, 0x9097, 0x1230, 0x00010001);
+	nv_mthd(priv, 0x9097, 0x07f8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x15b4, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x15cc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1534, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0fb0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x15d0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x153c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x16b4, 0x00000003);
+	nv_mthd(priv, 0x9097, 0x0fbc, 0x0000ffff);
+	nv_mthd(priv, 0x9097, 0x0fc0, 0x0000ffff);
+	nv_mthd(priv, 0x9097, 0x0fc4, 0x0000ffff);
+	nv_mthd(priv, 0x9097, 0x0fc8, 0x0000ffff);
+	nv_mthd(priv, 0x9097, 0x0df8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0dfc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1948, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1970, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x161c, 0x000009f0);
+	nv_mthd(priv, 0x9097, 0x0dcc, 0x00000010);
+	nv_mthd(priv, 0x9097, 0x163c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x15e4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1160, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x1164, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x1168, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x116c, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x1170, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x1174, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x1178, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x117c, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x1180, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x1184, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x1188, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x118c, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x1190, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x1194, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x1198, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x119c, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11a0, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11a4, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11a8, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11ac, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11b0, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11b4, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11b8, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11bc, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11c0, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11c4, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11c8, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11cc, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11d0, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11d4, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11d8, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x11dc, 0x25e00040);
+	nv_mthd(priv, 0x9097, 0x1880, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1884, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1888, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x188c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1890, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1894, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1898, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x189c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18a0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18a4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18a8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18ac, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18b0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18b4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18b8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18bc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18c0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18c4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18c8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18cc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18d0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18d4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18d8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18dc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18e0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18e4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18e8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18ec, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18f0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18f4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18f8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x18fc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0f84, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0f88, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x17c8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x17cc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x17d0, 0x000000ff);
+	nv_mthd(priv, 0x9097, 0x17d4, 0xffffffff);
+	nv_mthd(priv, 0x9097, 0x17d8, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x17dc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x15f4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x15f8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1434, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1438, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d74, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0dec, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x13a4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1318, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1644, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0748, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0de8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1648, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x12a4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1120, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1124, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1128, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x112c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1118, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x164c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1658, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1910, 0x00000290);
+	nv_mthd(priv, 0x9097, 0x1518, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x165c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1520, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1604, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1570, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x13b0, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x13b4, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x020c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1670, 0x30201000);
+	nv_mthd(priv, 0x9097, 0x1674, 0x70605040);
+	nv_mthd(priv, 0x9097, 0x1678, 0xb8a89888);
+	nv_mthd(priv, 0x9097, 0x167c, 0xf8e8d8c8);
+	nv_mthd(priv, 0x9097, 0x166c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1680, 0x00ffff00);
+	nv_mthd(priv, 0x9097, 0x12d0, 0x00000003);
+	nv_mthd(priv, 0x9097, 0x12d4, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1684, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1688, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0dac, 0x00001b02);
+	nv_mthd(priv, 0x9097, 0x0db0, 0x00001b02);
+	nv_mthd(priv, 0x9097, 0x0db4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x168c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x15bc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x156c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x187c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1110, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x0dc0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0dc4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0dc8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1234, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1690, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x12ac, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x02c4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0790, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0794, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0798, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x079c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x07a0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x077c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1000, 0x00000010);
+	nv_mthd(priv, 0x9097, 0x10fc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1290, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0218, 0x00000010);
+	nv_mthd(priv, 0x9097, 0x12d8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x12dc, 0x00000010);
+	nv_mthd(priv, 0x9097, 0x0d94, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x155c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1560, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1564, 0x00001fff);
+	nv_mthd(priv, 0x9097, 0x1574, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1578, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x157c, 0x003fffff);
+	nv_mthd(priv, 0x9097, 0x1354, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1664, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1610, 0x00000012);
+	nv_mthd(priv, 0x9097, 0x1608, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x160c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x162c, 0x00000003);
+	nv_mthd(priv, 0x9097, 0x0210, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0320, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0324, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0328, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x032c, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0330, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0334, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0338, 0x3f800000);
+	nv_mthd(priv, 0x9097, 0x0750, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0760, 0x39291909);
+	nv_mthd(priv, 0x9097, 0x0764, 0x79695949);
+	nv_mthd(priv, 0x9097, 0x0768, 0xb9a99989);
+	nv_mthd(priv, 0x9097, 0x076c, 0xf9e9d9c9);
+	nv_mthd(priv, 0x9097, 0x0770, 0x30201000);
+	nv_mthd(priv, 0x9097, 0x0774, 0x70605040);
+	nv_mthd(priv, 0x9097, 0x0778, 0x00009080);
+	nv_mthd(priv, 0x9097, 0x0780, 0x39291909);
+	nv_mthd(priv, 0x9097, 0x0784, 0x79695949);
+	nv_mthd(priv, 0x9097, 0x0788, 0xb9a99989);
+	nv_mthd(priv, 0x9097, 0x078c, 0xf9e9d9c9);
+	nv_mthd(priv, 0x9097, 0x07d0, 0x30201000);
+	nv_mthd(priv, 0x9097, 0x07d4, 0x70605040);
+	nv_mthd(priv, 0x9097, 0x07d8, 0x00009080);
+	nv_mthd(priv, 0x9097, 0x037c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x0740, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0744, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x2600, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1918, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x191c, 0x00000900);
+	nv_mthd(priv, 0x9097, 0x1920, 0x00000405);
+	nv_mthd(priv, 0x9097, 0x1308, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1924, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x13ac, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x192c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x193c, 0x00002c1c);
+	nv_mthd(priv, 0x9097, 0x0d7c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0f8c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x02c0, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1510, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1940, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ff4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0ff8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x194c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1950, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1968, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1590, 0x0000003f);
+	nv_mthd(priv, 0x9097, 0x07e8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x07ec, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x07f0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x07f4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x196c, 0x00000011);
+	nv_mthd(priv, 0x9097, 0x197c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0fcc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0fd0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x02d8, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x1980, 0x00000080);
+	nv_mthd(priv, 0x9097, 0x1504, 0x00000080);
+	nv_mthd(priv, 0x9097, 0x1984, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0300, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x13a8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x12ec, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1310, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1314, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1380, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1384, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1388, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x138c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1390, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1394, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x139c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1398, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1594, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1598, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x159c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x15a0, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x15a4, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x0f54, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0f58, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0f5c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x19bc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0f9c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0fa0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x12cc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x12e8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x130c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1360, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1364, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1368, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x136c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1370, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1374, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1378, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x137c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x133c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1340, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1344, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1348, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x134c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1350, 0x00000002);
+	nv_mthd(priv, 0x9097, 0x1358, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x12e4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x131c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1320, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1324, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1328, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x19c0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1140, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x19c4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x19c8, 0x00001500);
+	nv_mthd(priv, 0x9097, 0x135c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0f90, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x19e0, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x19e4, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x19e8, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x19ec, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x19f0, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x19f4, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x19f8, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x19fc, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x19cc, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x15b8, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1a00, 0x00001111);
+	nv_mthd(priv, 0x9097, 0x1a04, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1a08, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1a0c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1a10, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1a14, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1a18, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1a1c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d6c, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x0d70, 0xffff0000);
+	nv_mthd(priv, 0x9097, 0x10f8, 0x00001010);
+	nv_mthd(priv, 0x9097, 0x0d80, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d84, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d88, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d8c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0d90, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0da0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1508, 0x80000000);
+	nv_mthd(priv, 0x9097, 0x150c, 0x40000000);
+	nv_mthd(priv, 0x9097, 0x1668, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0318, 0x00000008);
+	nv_mthd(priv, 0x9097, 0x031c, 0x00000008);
+	nv_mthd(priv, 0x9097, 0x0d9c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x07dc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x074c, 0x00000055);
+	nv_mthd(priv, 0x9097, 0x1420, 0x00000003);
+	nv_mthd(priv, 0x9097, 0x17bc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x17c0, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x17c4, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1008, 0x00000008);
+	nv_mthd(priv, 0x9097, 0x100c, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x1010, 0x0000012c);
+	nv_mthd(priv, 0x9097, 0x0d60, 0x00000040);
+	nv_mthd(priv, 0x9097, 0x075c, 0x00000003);
+	nv_mthd(priv, 0x9097, 0x1018, 0x00000020);
+	nv_mthd(priv, 0x9097, 0x101c, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1020, 0x00000020);
+	nv_mthd(priv, 0x9097, 0x1024, 0x00000001);
+	nv_mthd(priv, 0x9097, 0x1444, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x1448, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x144c, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0360, 0x20164010);
+	nv_mthd(priv, 0x9097, 0x0364, 0x00000020);
+	nv_mthd(priv, 0x9097, 0x0368, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0de4, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0204, 0x00000006);
+	nv_mthd(priv, 0x9097, 0x0208, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x02cc, 0x003fffff);
+	nv_mthd(priv, 0x9097, 0x02d0, 0x00000c48);
+	nv_mthd(priv, 0x9097, 0x1220, 0x00000005);
+	nv_mthd(priv, 0x9097, 0x0fdc, 0x00000000);
+	nv_mthd(priv, 0x9097, 0x0f98, 0x00300008);
+	nv_mthd(priv, 0x9097, 0x1284, 0x04000080);
+	nv_mthd(priv, 0x9097, 0x1450, 0x00300008);
+	nv_mthd(priv, 0x9097, 0x1454, 0x04000080);
+	nv_mthd(priv, 0x9097, 0x0214, 0x00000000);
+	/* in trace, right after 0x90c0, not here */
+	nv_mthd(priv, 0x9097, 0x3410, 0x80002006);
+}
+
+static void
+nvc0_grctx_generate_9197(struct nvc0_graph_priv *priv)
+{
+	u32 fermi = nvc0_graph_class(priv);
+	u32 mthd;
+
+	if (fermi == 0x9197) {
+		for (mthd = 0x3400; mthd <= 0x35fc; mthd += 4)
+			nv_mthd(priv, 0x9197, mthd, 0x00000000);
+	}
+	nv_mthd(priv, 0x9197, 0x02e4, 0x0000b001);
+}
+
+static void
+nvc0_grctx_generate_9297(struct nvc0_graph_priv *priv)
+{
+	u32 fermi = nvc0_graph_class(priv);
+	u32 mthd;
+
+	if (fermi == 0x9297) {
+		for (mthd = 0x3400; mthd <= 0x35fc; mthd += 4)
+			nv_mthd(priv, 0x9297, mthd, 0x00000000);
+	}
+	nv_mthd(priv, 0x9297, 0x036c, 0x00000000);
+	nv_mthd(priv, 0x9297, 0x0370, 0x00000000);
+	nv_mthd(priv, 0x9297, 0x07a4, 0x00000000);
+	nv_mthd(priv, 0x9297, 0x07a8, 0x00000000);
+	nv_mthd(priv, 0x9297, 0x0374, 0x00000000);
+	nv_mthd(priv, 0x9297, 0x0378, 0x00000020);
+}
+
+static void
+nvc0_grctx_generate_902d(struct nvc0_graph_priv *priv)
+{
+	nv_mthd(priv, 0x902d, 0x0200, 0x000000cf);
+	nv_mthd(priv, 0x902d, 0x0204, 0x00000001);
+	nv_mthd(priv, 0x902d, 0x0208, 0x00000020);
+	nv_mthd(priv, 0x902d, 0x020c, 0x00000001);
+	nv_mthd(priv, 0x902d, 0x0210, 0x00000000);
+	nv_mthd(priv, 0x902d, 0x0214, 0x00000080);
+	nv_mthd(priv, 0x902d, 0x0218, 0x00000100);
+	nv_mthd(priv, 0x902d, 0x021c, 0x00000100);
+	nv_mthd(priv, 0x902d, 0x0220, 0x00000000);
+	nv_mthd(priv, 0x902d, 0x0224, 0x00000000);
+	nv_mthd(priv, 0x902d, 0x0230, 0x000000cf);
+	nv_mthd(priv, 0x902d, 0x0234, 0x00000001);
+	nv_mthd(priv, 0x902d, 0x0238, 0x00000020);
+	nv_mthd(priv, 0x902d, 0x023c, 0x00000001);
+	nv_mthd(priv, 0x902d, 0x0244, 0x00000080);
+	nv_mthd(priv, 0x902d, 0x0248, 0x00000100);
+	nv_mthd(priv, 0x902d, 0x024c, 0x00000100);
+}
+
+static void
+nvc0_grctx_generate_9039(struct nvc0_graph_priv *priv)
+{
+	nv_mthd(priv, 0x9039, 0x030c, 0x00000000);
+	nv_mthd(priv, 0x9039, 0x0310, 0x00000000);
+	nv_mthd(priv, 0x9039, 0x0314, 0x00000000);
+	nv_mthd(priv, 0x9039, 0x0320, 0x00000000);
+	nv_mthd(priv, 0x9039, 0x0238, 0x00000000);
+	nv_mthd(priv, 0x9039, 0x023c, 0x00000000);
+	nv_mthd(priv, 0x9039, 0x0318, 0x00000000);
+	nv_mthd(priv, 0x9039, 0x031c, 0x00000000);
+}
+
+static void
+nvc0_grctx_generate_90c0(struct nvc0_graph_priv *priv)
+{
+	int i;
+
+	for (i = 0; nv_device(priv)->chipset == 0xd9 && i < 4; i++) {
+		nv_mthd(priv, 0x90c0, 0x2700 + (i * 0x40), 0x00000000);
+		nv_mthd(priv, 0x90c0, 0x2720 + (i * 0x40), 0x00000000);
+		nv_mthd(priv, 0x90c0, 0x2704 + (i * 0x40), 0x00000000);
+		nv_mthd(priv, 0x90c0, 0x2724 + (i * 0x40), 0x00000000);
+		nv_mthd(priv, 0x90c0, 0x2708 + (i * 0x40), 0x00000000);
+		nv_mthd(priv, 0x90c0, 0x2728 + (i * 0x40), 0x00000000);
+	}
+	nv_mthd(priv, 0x90c0, 0x270c, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x272c, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x274c, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x276c, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x278c, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x27ac, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x27cc, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x27ec, 0x00000000);
+	for (i = 0; nv_device(priv)->chipset == 0xd9 && i < 4; i++) {
+		nv_mthd(priv, 0x90c0, 0x2710 + (i * 0x40), 0x00014000);
+		nv_mthd(priv, 0x90c0, 0x2730 + (i * 0x40), 0x00014000);
+		nv_mthd(priv, 0x90c0, 0x2714 + (i * 0x40), 0x00000040);
+		nv_mthd(priv, 0x90c0, 0x2734 + (i * 0x40), 0x00000040);
+	}
+	nv_mthd(priv, 0x90c0, 0x030c, 0x00000001);
+	nv_mthd(priv, 0x90c0, 0x1944, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x0758, 0x00000100);
+	nv_mthd(priv, 0x90c0, 0x02c4, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x0790, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x0794, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x0798, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x079c, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x07a0, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x077c, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x0204, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x0208, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x020c, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x0214, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x024c, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x0d94, 0x00000001);
+	nv_mthd(priv, 0x90c0, 0x1608, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x160c, 0x00000000);
+	nv_mthd(priv, 0x90c0, 0x1664, 0x00000000);
+}
+
+static void
+nvc0_grctx_generate_dispatch(struct nvc0_graph_priv *priv)
+{
+	int i;
+
+	nv_wr32(priv, 0x404004, 0x00000000);
+	nv_wr32(priv, 0x404008, 0x00000000);
+	nv_wr32(priv, 0x40400c, 0x00000000);
+	nv_wr32(priv, 0x404010, 0x00000000);
+	nv_wr32(priv, 0x404014, 0x00000000);
+	nv_wr32(priv, 0x404018, 0x00000000);
+	nv_wr32(priv, 0x40401c, 0x00000000);
+	nv_wr32(priv, 0x404020, 0x00000000);
+	nv_wr32(priv, 0x404024, 0x00000000);
+	nv_wr32(priv, 0x404028, 0x00000000);
+	nv_wr32(priv, 0x40402c, 0x00000000);
+	nv_wr32(priv, 0x404044, 0x00000000);
+	nv_wr32(priv, 0x404094, 0x00000000);
+	nv_wr32(priv, 0x404098, 0x00000000);
+	nv_wr32(priv, 0x40409c, 0x00000000);
+	nv_wr32(priv, 0x4040a0, 0x00000000);
+	nv_wr32(priv, 0x4040a4, 0x00000000);
+	nv_wr32(priv, 0x4040a8, 0x00000000);
+	nv_wr32(priv, 0x4040ac, 0x00000000);
+	nv_wr32(priv, 0x4040b0, 0x00000000);
+	nv_wr32(priv, 0x4040b4, 0x00000000);
+	nv_wr32(priv, 0x4040b8, 0x00000000);
+	nv_wr32(priv, 0x4040bc, 0x00000000);
+	nv_wr32(priv, 0x4040c0, 0x00000000);
+	nv_wr32(priv, 0x4040c4, 0x00000000);
+	nv_wr32(priv, 0x4040c8, 0xf0000087);
+	nv_wr32(priv, 0x4040d4, 0x00000000);
+	nv_wr32(priv, 0x4040d8, 0x00000000);
+	nv_wr32(priv, 0x4040dc, 0x00000000);
+	nv_wr32(priv, 0x4040e0, 0x00000000);
+	nv_wr32(priv, 0x4040e4, 0x00000000);
+	nv_wr32(priv, 0x4040e8, 0x00001000);
+	nv_wr32(priv, 0x4040f8, 0x00000000);
+	nv_wr32(priv, 0x404130, 0x00000000);
+	nv_wr32(priv, 0x404134, 0x00000000);
+	nv_wr32(priv, 0x404138, 0x20000040);
+	nv_wr32(priv, 0x404150, 0x0000002e);
+	nv_wr32(priv, 0x404154, 0x00000400);
+	nv_wr32(priv, 0x404158, 0x00000200);
+	nv_wr32(priv, 0x404164, 0x00000055);
+	nv_wr32(priv, 0x404168, 0x00000000);
+	nv_wr32(priv, 0x404174, 0x00000000);
+	nv_wr32(priv, 0x404178, 0x00000000);
+	nv_wr32(priv, 0x40417c, 0x00000000);
+	for (i = 0; i < 8; i++)
+		nv_wr32(priv, 0x404200 + (i * 4), 0x00000000); /* subc */
+}
+
+static void
+nvc0_grctx_generate_macro(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x404404, 0x00000000);
+	nv_wr32(priv, 0x404408, 0x00000000);
+	nv_wr32(priv, 0x40440c, 0x00000000);
+	nv_wr32(priv, 0x404410, 0x00000000);
+	nv_wr32(priv, 0x404414, 0x00000000);
+	nv_wr32(priv, 0x404418, 0x00000000);
+	nv_wr32(priv, 0x40441c, 0x00000000);
+	nv_wr32(priv, 0x404420, 0x00000000);
+	nv_wr32(priv, 0x404424, 0x00000000);
+	nv_wr32(priv, 0x404428, 0x00000000);
+	nv_wr32(priv, 0x40442c, 0x00000000);
+	nv_wr32(priv, 0x404430, 0x00000000);
+	nv_wr32(priv, 0x404434, 0x00000000);
+	nv_wr32(priv, 0x404438, 0x00000000);
+	nv_wr32(priv, 0x404460, 0x00000000);
+	nv_wr32(priv, 0x404464, 0x00000000);
+	nv_wr32(priv, 0x404468, 0x00ffffff);
+	nv_wr32(priv, 0x40446c, 0x00000000);
+	nv_wr32(priv, 0x404480, 0x00000001);
+	nv_wr32(priv, 0x404498, 0x00000001);
+}
+
+static void
+nvc0_grctx_generate_m2mf(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x404604, 0x00000015);
+	nv_wr32(priv, 0x404608, 0x00000000);
+	nv_wr32(priv, 0x40460c, 0x00002e00);
+	nv_wr32(priv, 0x404610, 0x00000100);
+	nv_wr32(priv, 0x404618, 0x00000000);
+	nv_wr32(priv, 0x40461c, 0x00000000);
+	nv_wr32(priv, 0x404620, 0x00000000);
+	nv_wr32(priv, 0x404624, 0x00000000);
+	nv_wr32(priv, 0x404628, 0x00000000);
+	nv_wr32(priv, 0x40462c, 0x00000000);
+	nv_wr32(priv, 0x404630, 0x00000000);
+	nv_wr32(priv, 0x404634, 0x00000000);
+	nv_wr32(priv, 0x404638, 0x00000004);
+	nv_wr32(priv, 0x40463c, 0x00000000);
+	nv_wr32(priv, 0x404640, 0x00000000);
+	nv_wr32(priv, 0x404644, 0x00000000);
+	nv_wr32(priv, 0x404648, 0x00000000);
+	nv_wr32(priv, 0x40464c, 0x00000000);
+	nv_wr32(priv, 0x404650, 0x00000000);
+	nv_wr32(priv, 0x404654, 0x00000000);
+	nv_wr32(priv, 0x404658, 0x00000000);
+	nv_wr32(priv, 0x40465c, 0x007f0100);
+	nv_wr32(priv, 0x404660, 0x00000000);
+	nv_wr32(priv, 0x404664, 0x00000000);
+	nv_wr32(priv, 0x404668, 0x00000000);
+	nv_wr32(priv, 0x40466c, 0x00000000);
+	nv_wr32(priv, 0x404670, 0x00000000);
+	nv_wr32(priv, 0x404674, 0x00000000);
+	nv_wr32(priv, 0x404678, 0x00000000);
+	nv_wr32(priv, 0x40467c, 0x00000002);
+	nv_wr32(priv, 0x404680, 0x00000000);
+	nv_wr32(priv, 0x404684, 0x00000000);
+	nv_wr32(priv, 0x404688, 0x00000000);
+	nv_wr32(priv, 0x40468c, 0x00000000);
+	nv_wr32(priv, 0x404690, 0x00000000);
+	nv_wr32(priv, 0x404694, 0x00000000);
+	nv_wr32(priv, 0x404698, 0x00000000);
+	nv_wr32(priv, 0x40469c, 0x00000000);
+	nv_wr32(priv, 0x4046a0, 0x007f0080);
+	nv_wr32(priv, 0x4046a4, 0x00000000);
+	nv_wr32(priv, 0x4046a8, 0x00000000);
+	nv_wr32(priv, 0x4046ac, 0x00000000);
+	nv_wr32(priv, 0x4046b0, 0x00000000);
+	nv_wr32(priv, 0x4046b4, 0x00000000);
+	nv_wr32(priv, 0x4046b8, 0x00000000);
+	nv_wr32(priv, 0x4046bc, 0x00000000);
+	nv_wr32(priv, 0x4046c0, 0x00000000);
+	nv_wr32(priv, 0x4046c4, 0x00000000);
+	nv_wr32(priv, 0x4046c8, 0x00000000);
+	nv_wr32(priv, 0x4046cc, 0x00000000);
+	nv_wr32(priv, 0x4046d0, 0x00000000);
+	nv_wr32(priv, 0x4046d4, 0x00000000);
+	nv_wr32(priv, 0x4046d8, 0x00000000);
+	nv_wr32(priv, 0x4046dc, 0x00000000);
+	nv_wr32(priv, 0x4046e0, 0x00000000);
+	nv_wr32(priv, 0x4046e4, 0x00000000);
+	nv_wr32(priv, 0x4046e8, 0x00000000);
+	nv_wr32(priv, 0x4046f0, 0x00000000);
+	nv_wr32(priv, 0x4046f4, 0x00000000);
+}
+
+static void
+nvc0_grctx_generate_unk47xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x404700, 0x00000000);
+	nv_wr32(priv, 0x404704, 0x00000000);
+	nv_wr32(priv, 0x404708, 0x00000000);
+	nv_wr32(priv, 0x40470c, 0x00000000);
+	nv_wr32(priv, 0x404710, 0x00000000);
+	nv_wr32(priv, 0x404714, 0x00000000);
+	nv_wr32(priv, 0x404718, 0x00000000);
+	nv_wr32(priv, 0x40471c, 0x00000000);
+	nv_wr32(priv, 0x404720, 0x00000000);
+	nv_wr32(priv, 0x404724, 0x00000000);
+	nv_wr32(priv, 0x404728, 0x00000000);
+	nv_wr32(priv, 0x40472c, 0x00000000);
+	nv_wr32(priv, 0x404730, 0x00000000);
+	nv_wr32(priv, 0x404734, 0x00000100);
+	nv_wr32(priv, 0x404738, 0x00000000);
+	nv_wr32(priv, 0x40473c, 0x00000000);
+	nv_wr32(priv, 0x404740, 0x00000000);
+	nv_wr32(priv, 0x404744, 0x00000000);
+	nv_wr32(priv, 0x404748, 0x00000000);
+	nv_wr32(priv, 0x40474c, 0x00000000);
+	nv_wr32(priv, 0x404750, 0x00000000);
+	nv_wr32(priv, 0x404754, 0x00000000);
+}
+
+static void
+nvc0_grctx_generate_shaders(struct nvc0_graph_priv *priv)
+{
+
+	if (nv_device(priv)->chipset == 0xd9) {
+		nv_wr32(priv, 0x405800, 0x0f8000bf);
+		nv_wr32(priv, 0x405830, 0x02180218);
+		nv_wr32(priv, 0x405834, 0x08000000);
+	} else
+	if (nv_device(priv)->chipset == 0xc1) {
+		nv_wr32(priv, 0x405800, 0x0f8000bf);
+		nv_wr32(priv, 0x405830, 0x02180218);
+		nv_wr32(priv, 0x405834, 0x00000000);
+	} else {
+		nv_wr32(priv, 0x405800, 0x078000bf);
+		nv_wr32(priv, 0x405830, 0x02180000);
+		nv_wr32(priv, 0x405834, 0x00000000);
+	}
+	nv_wr32(priv, 0x405838, 0x00000000);
+	nv_wr32(priv, 0x405854, 0x00000000);
+	nv_wr32(priv, 0x405870, 0x00000001);
+	nv_wr32(priv, 0x405874, 0x00000001);
+	nv_wr32(priv, 0x405878, 0x00000001);
+	nv_wr32(priv, 0x40587c, 0x00000001);
+	nv_wr32(priv, 0x405a00, 0x00000000);
+	nv_wr32(priv, 0x405a04, 0x00000000);
+	nv_wr32(priv, 0x405a18, 0x00000000);
+}
+
+static void
+nvc0_grctx_generate_unk60xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x406020, 0x000103c1);
+	nv_wr32(priv, 0x406028, 0x00000001);
+	nv_wr32(priv, 0x40602c, 0x00000001);
+	nv_wr32(priv, 0x406030, 0x00000001);
+	nv_wr32(priv, 0x406034, 0x00000001);
+}
+
+static void
+nvc0_grctx_generate_unk64xx(struct nvc0_graph_priv *priv)
+{
+
+	nv_wr32(priv, 0x4064a8, 0x00000000);
+	nv_wr32(priv, 0x4064ac, 0x00003fff);
+	nv_wr32(priv, 0x4064b4, 0x00000000);
+	nv_wr32(priv, 0x4064b8, 0x00000000);
+	if (nv_device(priv)->chipset == 0xd9)
+		nv_wr32(priv, 0x4064bc, 0x00000000);
+	if (nv_device(priv)->chipset == 0xc1 ||
+	    nv_device(priv)->chipset == 0xd9) {
+		nv_wr32(priv, 0x4064c0, 0x80140078);
+		nv_wr32(priv, 0x4064c4, 0x0086ffff);
+	}
+}
+
+static void
+nvc0_grctx_generate_tpbus(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x407804, 0x00000023);
+	nv_wr32(priv, 0x40780c, 0x0a418820);
+	nv_wr32(priv, 0x407810, 0x062080e6);
+	nv_wr32(priv, 0x407814, 0x020398a4);
+	nv_wr32(priv, 0x407818, 0x0e629062);
+	nv_wr32(priv, 0x40781c, 0x0a418820);
+	nv_wr32(priv, 0x407820, 0x000000e6);
+	nv_wr32(priv, 0x4078bc, 0x00000103);
+}
+
+static void
+nvc0_grctx_generate_ccache(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x408000, 0x00000000);
+	nv_wr32(priv, 0x408004, 0x00000000);
+	nv_wr32(priv, 0x408008, 0x00000018);
+	nv_wr32(priv, 0x40800c, 0x00000000);
+	nv_wr32(priv, 0x408010, 0x00000000);
+	nv_wr32(priv, 0x408014, 0x00000069);
+	nv_wr32(priv, 0x408018, 0xe100e100);
+	nv_wr32(priv, 0x408064, 0x00000000);
+}
+
+static void
+nvc0_grctx_generate_rop(struct nvc0_graph_priv *priv)
+{
+	int chipset = nv_device(priv)->chipset;
+
+	/* ROPC_BROADCAST */
+	nv_wr32(priv, 0x408800, 0x02802a3c);
+	nv_wr32(priv, 0x408804, 0x00000040);
+	if (chipset == 0xd9) {
+		nv_wr32(priv, 0x408808, 0x1043e005);
+		nv_wr32(priv, 0x408900, 0x3080b801);
+		nv_wr32(priv, 0x408904, 0x1043e005);
+		nv_wr32(priv, 0x408908, 0x00c8102f);
+	} else
+	if (chipset == 0xc1) {
+		nv_wr32(priv, 0x408808, 0x1003e005);
+		nv_wr32(priv, 0x408900, 0x3080b801);
+		nv_wr32(priv, 0x408904, 0x62000001);
+		nv_wr32(priv, 0x408908, 0x00c80929);
+	} else {
+		nv_wr32(priv, 0x408808, 0x0003e00d);
+		nv_wr32(priv, 0x408900, 0x3080b801);
+		nv_wr32(priv, 0x408904, 0x02000001);
+		nv_wr32(priv, 0x408908, 0x00c80929);
+	}
+	nv_wr32(priv, 0x40890c, 0x00000000);
+	nv_wr32(priv, 0x408980, 0x0000011d);
+}
+
+static void
+nvc0_grctx_generate_gpc(struct nvc0_graph_priv *priv)
+{
+	int chipset = nv_device(priv)->chipset;
+	int i;
+
+	/* GPC_BROADCAST */
+	nv_wr32(priv, 0x418380, 0x00000016);
+	nv_wr32(priv, 0x418400, 0x38004e00);
+	nv_wr32(priv, 0x418404, 0x71e0ffff);
+	nv_wr32(priv, 0x418408, 0x00000000);
+	nv_wr32(priv, 0x41840c, 0x00001008);
+	nv_wr32(priv, 0x418410, 0x0fff0fff);
+	nv_wr32(priv, 0x418414, chipset != 0xd9 ? 0x00200fff : 0x02200fff);
+	nv_wr32(priv, 0x418450, 0x00000000);
+	nv_wr32(priv, 0x418454, 0x00000000);
+	nv_wr32(priv, 0x418458, 0x00000000);
+	nv_wr32(priv, 0x41845c, 0x00000000);
+	nv_wr32(priv, 0x418460, 0x00000000);
+	nv_wr32(priv, 0x418464, 0x00000000);
+	nv_wr32(priv, 0x418468, 0x00000001);
+	nv_wr32(priv, 0x41846c, 0x00000000);
+	nv_wr32(priv, 0x418470, 0x00000000);
+	nv_wr32(priv, 0x418600, 0x0000001f);
+	nv_wr32(priv, 0x418684, 0x0000000f);
+	nv_wr32(priv, 0x418700, 0x00000002);
+	nv_wr32(priv, 0x418704, 0x00000080);
+	nv_wr32(priv, 0x418708, 0x00000000);
+	nv_wr32(priv, 0x41870c, chipset != 0xd9 ? 0x07c80000 : 0x00000000);
+	nv_wr32(priv, 0x418710, 0x00000000);
+	nv_wr32(priv, 0x418800, chipset != 0xd9 ? 0x0006860a : 0x7006860a);
+	nv_wr32(priv, 0x418808, 0x00000000);
+	nv_wr32(priv, 0x41880c, 0x00000000);
+	nv_wr32(priv, 0x418810, 0x00000000);
+	nv_wr32(priv, 0x418828, 0x00008442);
+	if (chipset == 0xc1 || chipset == 0xd9)
+		nv_wr32(priv, 0x418830, 0x10000001);
+	else
+		nv_wr32(priv, 0x418830, 0x00000001);
+	nv_wr32(priv, 0x4188d8, 0x00000008);
+	nv_wr32(priv, 0x4188e0, 0x01000000);
+	nv_wr32(priv, 0x4188e8, 0x00000000);
+	nv_wr32(priv, 0x4188ec, 0x00000000);
+	nv_wr32(priv, 0x4188f0, 0x00000000);
+	nv_wr32(priv, 0x4188f4, 0x00000000);
+	nv_wr32(priv, 0x4188f8, 0x00000000);
+	if (chipset == 0xd9)
+		nv_wr32(priv, 0x4188fc, 0x20100008);
+	else if (chipset == 0xc1)
+		nv_wr32(priv, 0x4188fc, 0x00100018);
+	else
+		nv_wr32(priv, 0x4188fc, 0x00100000);
+	nv_wr32(priv, 0x41891c, 0x00ff00ff);
+	nv_wr32(priv, 0x418924, 0x00000000);
+	nv_wr32(priv, 0x418928, 0x00ffff00);
+	nv_wr32(priv, 0x41892c, 0x0000ff00);
+	for (i = 0; i < 8; i++) {
+		nv_wr32(priv, 0x418a00 + (i * 0x20), 0x00000000);
+		nv_wr32(priv, 0x418a04 + (i * 0x20), 0x00000000);
+		nv_wr32(priv, 0x418a08 + (i * 0x20), 0x00000000);
+		nv_wr32(priv, 0x418a0c + (i * 0x20), 0x00010000);
+		nv_wr32(priv, 0x418a10 + (i * 0x20), 0x00000000);
+		nv_wr32(priv, 0x418a14 + (i * 0x20), 0x00000000);
+		nv_wr32(priv, 0x418a18 + (i * 0x20), 0x00000000);
+	}
+	nv_wr32(priv, 0x418b00, chipset != 0xd9 ? 0x00000000 : 0x00000006);
+	nv_wr32(priv, 0x418b08, 0x0a418820);
+	nv_wr32(priv, 0x418b0c, 0x062080e6);
+	nv_wr32(priv, 0x418b10, 0x020398a4);
+	nv_wr32(priv, 0x418b14, 0x0e629062);
+	nv_wr32(priv, 0x418b18, 0x0a418820);
+	nv_wr32(priv, 0x418b1c, 0x000000e6);
+	nv_wr32(priv, 0x418bb8, 0x00000103);
+	nv_wr32(priv, 0x418c08, 0x00000001);
+	nv_wr32(priv, 0x418c10, 0x00000000);
+	nv_wr32(priv, 0x418c14, 0x00000000);
+	nv_wr32(priv, 0x418c18, 0x00000000);
+	nv_wr32(priv, 0x418c1c, 0x00000000);
+	nv_wr32(priv, 0x418c20, 0x00000000);
+	nv_wr32(priv, 0x418c24, 0x00000000);
+	nv_wr32(priv, 0x418c28, 0x00000000);
+	nv_wr32(priv, 0x418c2c, 0x00000000);
+	if (chipset == 0xc1 || chipset == 0xd9)
+		nv_wr32(priv, 0x418c6c, 0x00000001);
+	nv_wr32(priv, 0x418c80, 0x20200004);
+	nv_wr32(priv, 0x418c8c, 0x00000001);
+	nv_wr32(priv, 0x419000, 0x00000780);
+	nv_wr32(priv, 0x419004, 0x00000000);
+	nv_wr32(priv, 0x419008, 0x00000000);
+	nv_wr32(priv, 0x419014, 0x00000004);
+}
+
+static void
+nvc0_grctx_generate_tp(struct nvc0_graph_priv *priv)
+{
+	int chipset = nv_device(priv)->chipset;
+
+	/* GPC_BROADCAST.TP_BROADCAST */
+	nv_wr32(priv, 0x419818, 0x00000000);
+	nv_wr32(priv, 0x41983c, 0x00038bc7);
+	nv_wr32(priv, 0x419848, 0x00000000);
+	if (chipset == 0xc1 || chipset == 0xd9)
+		nv_wr32(priv, 0x419864, 0x00000129);
+	else
+		nv_wr32(priv, 0x419864, 0x0000012a);
+	nv_wr32(priv, 0x419888, 0x00000000);
+	nv_wr32(priv, 0x419a00, 0x000001f0);
+	nv_wr32(priv, 0x419a04, 0x00000001);
+	nv_wr32(priv, 0x419a08, 0x00000023);
+	nv_wr32(priv, 0x419a0c, 0x00020000);
+	nv_wr32(priv, 0x419a10, 0x00000000);
+	nv_wr32(priv, 0x419a14, 0x00000200);
+	nv_wr32(priv, 0x419a1c, 0x00000000);
+	nv_wr32(priv, 0x419a20, 0x00000800);
+	if (chipset == 0xd9)
+		nv_wr32(priv, 0x00419ac4, 0x0017f440);
+	else if (chipset != 0xc0 && chipset != 0xc8)
+		nv_wr32(priv, 0x00419ac4, 0x0007f440);
+	nv_wr32(priv, 0x419b00, 0x0a418820);
+	nv_wr32(priv, 0x419b04, 0x062080e6);
+	nv_wr32(priv, 0x419b08, 0x020398a4);
+	nv_wr32(priv, 0x419b0c, 0x0e629062);
+	nv_wr32(priv, 0x419b10, 0x0a418820);
+	nv_wr32(priv, 0x419b14, 0x000000e6);
+	nv_wr32(priv, 0x419bd0, 0x00900103);
+	if (chipset == 0xc1 || chipset == 0xd9)
+		nv_wr32(priv, 0x419be0, 0x00400001);
+	else
+		nv_wr32(priv, 0x419be0, 0x00000001);
+	nv_wr32(priv, 0x419be4, 0x00000000);
+	nv_wr32(priv, 0x419c00, chipset != 0xd9 ? 0x00000002 : 0x0000000a);
+	nv_wr32(priv, 0x419c04, 0x00000006);
+	nv_wr32(priv, 0x419c08, 0x00000002);
+	nv_wr32(priv, 0x419c20, 0x00000000);
+	if (nv_device(priv)->chipset == 0xd9) {
+		nv_wr32(priv, 0x419c24, 0x00084210);
+		nv_wr32(priv, 0x419c28, 0x3cf3cf3c);
+		nv_wr32(priv, 0x419cb0, 0x00020048);
+	} else
+	if (chipset == 0xce || chipset == 0xcf) {
+		nv_wr32(priv, 0x419cb0, 0x00020048);
+	} else {
+		nv_wr32(priv, 0x419cb0, 0x00060048);
+	}
+	nv_wr32(priv, 0x419ce8, 0x00000000);
+	nv_wr32(priv, 0x419cf4, 0x00000183);
+	if (chipset == 0xc1 || chipset == 0xd9)
+		nv_wr32(priv, 0x419d20, 0x12180000);
+	else
+		nv_wr32(priv, 0x419d20, 0x02180000);
+	nv_wr32(priv, 0x419d24, 0x00001fff);
+	if (chipset == 0xc1 || chipset == 0xd9)
+		nv_wr32(priv, 0x419d44, 0x02180218);
+	nv_wr32(priv, 0x419e04, 0x00000000);
+	nv_wr32(priv, 0x419e08, 0x00000000);
+	nv_wr32(priv, 0x419e0c, 0x00000000);
+	nv_wr32(priv, 0x419e10, 0x00000002);
+	nv_wr32(priv, 0x419e44, 0x001beff2);
+	nv_wr32(priv, 0x419e48, 0x00000000);
+	nv_wr32(priv, 0x419e4c, 0x0000000f);
+	nv_wr32(priv, 0x419e50, 0x00000000);
+	nv_wr32(priv, 0x419e54, 0x00000000);
+	nv_wr32(priv, 0x419e58, 0x00000000);
+	nv_wr32(priv, 0x419e5c, 0x00000000);
+	nv_wr32(priv, 0x419e60, 0x00000000);
+	nv_wr32(priv, 0x419e64, 0x00000000);
+	nv_wr32(priv, 0x419e68, 0x00000000);
+	nv_wr32(priv, 0x419e6c, 0x00000000);
+	nv_wr32(priv, 0x419e70, 0x00000000);
+	nv_wr32(priv, 0x419e74, 0x00000000);
+	nv_wr32(priv, 0x419e78, 0x00000000);
+	nv_wr32(priv, 0x419e7c, 0x00000000);
+	nv_wr32(priv, 0x419e80, 0x00000000);
+	nv_wr32(priv, 0x419e84, 0x00000000);
+	nv_wr32(priv, 0x419e88, 0x00000000);
+	nv_wr32(priv, 0x419e8c, 0x00000000);
+	nv_wr32(priv, 0x419e90, 0x00000000);
+	nv_wr32(priv, 0x419e98, 0x00000000);
+	if (chipset != 0xc0 && chipset != 0xc8)
+		nv_wr32(priv, 0x419ee0, 0x00011110);
+	nv_wr32(priv, 0x419f50, 0x00000000);
+	nv_wr32(priv, 0x419f54, 0x00000000);
+	if (chipset != 0xc0 && chipset != 0xc8)
+		nv_wr32(priv, 0x419f58, 0x00000000);
+}
+
+int
+nvc0_grctx_generate(struct nvc0_graph_priv *priv)
+{
+	struct nvc0_grctx info;
+	int ret, i, gpc, tpc, id;
+	u32 fermi = nvc0_graph_class(priv);
+	u32 r000260, tmp;
+
+	ret = nvc0_grctx_init(priv, &info);
+	if (ret)
+		return ret;
+
+	r000260 = nv_rd32(priv, 0x000260);
+	nv_wr32(priv, 0x000260, r000260 & ~1);
+	nv_wr32(priv, 0x400208, 0x00000000);
+
+	nvc0_grctx_generate_dispatch(priv);
+	nvc0_grctx_generate_macro(priv);
+	nvc0_grctx_generate_m2mf(priv);
+	nvc0_grctx_generate_unk47xx(priv);
+	nvc0_grctx_generate_shaders(priv);
+	nvc0_grctx_generate_unk60xx(priv);
+	nvc0_grctx_generate_unk64xx(priv);
+	nvc0_grctx_generate_tpbus(priv);
+	nvc0_grctx_generate_ccache(priv);
+	nvc0_grctx_generate_rop(priv);
+	nvc0_grctx_generate_gpc(priv);
+	nvc0_grctx_generate_tp(priv);
+
+	nv_wr32(priv, 0x404154, 0x00000000);
+
+	/* generate per-context mmio list data */
+	mmio_data(0x002000, 0x0100, NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS);
+	mmio_data(0x008000, 0x0100, NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS);
+	mmio_data(0x060000, 0x1000, NV_MEM_ACCESS_RW);
+	mmio_list(0x408004, 0x00000000,  8, 0);
+	mmio_list(0x408008, 0x80000018,  0, 0);
+	mmio_list(0x40800c, 0x00000000,  8, 1);
+	mmio_list(0x408010, 0x80000000,  0, 0);
+	mmio_list(0x418810, 0x80000000, 12, 2);
+	mmio_list(0x419848, 0x10000000, 12, 2);
+	mmio_list(0x419004, 0x00000000,  8, 1);
+	mmio_list(0x419008, 0x00000000,  0, 0);
+	mmio_list(0x418808, 0x00000000,  8, 0);
+	mmio_list(0x41880c, 0x80000018,  0, 0);
+	if (nv_device(priv)->chipset != 0xc1) {
+		tmp = 0x02180000;
+		mmio_list(0x405830, tmp, 0, 0);
+		for (gpc = 0; gpc < priv->gpc_nr; gpc++) {
+			for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) {
+				u32 reg = TPC_UNIT(gpc, tpc, 0x0520);
+				mmio_list(reg, tmp, 0, 0);
+				tmp += 0x0324;
+			}
+		}
+	} else {
+		tmp = 0x02180000;
+		mmio_list(0x405830, 0x00000218 | tmp, 0, 0);
+		mmio_list(0x4064c4, 0x0086ffff, 0, 0);
+		for (gpc = 0; gpc < priv->gpc_nr; gpc++) {
+			for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) {
+				u32 reg = TPC_UNIT(gpc, tpc, 0x0520);
+				mmio_list(reg, 0x10000000 | tmp, 0, 0);
+				tmp += 0x0324;
+			}
+			for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) {
+				u32 reg = TPC_UNIT(gpc, tpc, 0x0544);
+				mmio_list(reg, tmp, 0, 0);
+				tmp += 0x0324;
+			}
+		}
+	}
+
+	for (tpc = 0, id = 0; tpc < 4; tpc++) {
+		for (gpc = 0; gpc < priv->gpc_nr; gpc++) {
+			if (tpc < priv->tpc_nr[gpc]) {
+				nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x698), id);
+				nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x4e8), id);
+				nv_wr32(priv, GPC_UNIT(gpc, 0x0c10 + tpc * 4), id);
+				nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x088), id);
+				id++;
+			}
+
+			nv_wr32(priv, GPC_UNIT(gpc, 0x0c08), priv->tpc_nr[gpc]);
+			nv_wr32(priv, GPC_UNIT(gpc, 0x0c8c), priv->tpc_nr[gpc]);
+		}
+	}
+
+	tmp = 0;
+	for (i = 0; i < priv->gpc_nr; i++)
+		tmp |= priv->tpc_nr[i] << (i * 4);
+	nv_wr32(priv, 0x406028, tmp);
+	nv_wr32(priv, 0x405870, tmp);
+
+	nv_wr32(priv, 0x40602c, 0x00000000);
+	nv_wr32(priv, 0x405874, 0x00000000);
+	nv_wr32(priv, 0x406030, 0x00000000);
+	nv_wr32(priv, 0x405878, 0x00000000);
+	nv_wr32(priv, 0x406034, 0x00000000);
+	nv_wr32(priv, 0x40587c, 0x00000000);
+
+	if (1) {
+		u8 tpcnr[GPC_MAX], data[TPC_MAX];
+
+		memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr));
+		memset(data, 0x1f, sizeof(data));
+
+		gpc = -1;
+		for (tpc = 0; tpc < priv->tpc_total; tpc++) {
+			do {
+				gpc = (gpc + 1) % priv->gpc_nr;
+			} while (!tpcnr[gpc]);
+			tpcnr[gpc]--;
+			data[tpc] = gpc;
+		}
+
+		for (i = 0; i < 4; i++)
+			nv_wr32(priv, 0x4060a8 + (i * 4), ((u32 *)data)[i]);
+	}
+
+	if (1) {
+		u32 data[6] = {}, data2[2] = {};
+		u8 tpcnr[GPC_MAX];
+		u8 shift, ntpcv;
+
+		/* calculate first set of magics */
+		memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr));
+
+		gpc = -1;
+		for (tpc = 0; tpc < priv->tpc_total; tpc++) {
+			do {
+				gpc = (gpc + 1) % priv->gpc_nr;
+			} while (!tpcnr[gpc]);
+			tpcnr[gpc]--;
+
+			data[tpc / 6] |= gpc << ((tpc % 6) * 5);
+		}
+
+		for (; tpc < 32; tpc++)
+			data[tpc / 6] |= 7 << ((tpc % 6) * 5);
+
+		/* and the second... */
+		shift = 0;
+		ntpcv = priv->tpc_total;
+		while (!(ntpcv & (1 << 4))) {
+			ntpcv <<= 1;
+			shift++;
+		}
+
+		data2[0]  = (ntpcv << 16);
+		data2[0] |= (shift << 21);
+		data2[0] |= (((1 << (0 + 5)) % ntpcv) << 24);
+		for (i = 1; i < 7; i++)
+			data2[1] |= ((1 << (i + 5)) % ntpcv) << ((i - 1) * 5);
+
+		/* GPC_BROADCAST */
+		nv_wr32(priv, 0x418bb8, (priv->tpc_total << 8) |
+					priv->magic_not_rop_nr);
+		for (i = 0; i < 6; i++)
+			nv_wr32(priv, 0x418b08 + (i * 4), data[i]);
+
+		/* GPC_BROADCAST.TP_BROADCAST */
+		nv_wr32(priv, 0x419bd0, (priv->tpc_total << 8) |
+				       priv->magic_not_rop_nr |
+				       data2[0]);
+		nv_wr32(priv, 0x419be4, data2[1]);
+		for (i = 0; i < 6; i++)
+			nv_wr32(priv, 0x419b00 + (i * 4), data[i]);
+
+		/* UNK78xx */
+		nv_wr32(priv, 0x4078bc, (priv->tpc_total << 8) |
+					priv->magic_not_rop_nr);
+		for (i = 0; i < 6; i++)
+			nv_wr32(priv, 0x40780c + (i * 4), data[i]);
+	}
+
+	if (1) {
+		u32 tpc_mask = 0, tpc_set = 0;
+		u8  tpcnr[GPC_MAX], a, b;
+
+		memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr));
+		for (gpc = 0; gpc < priv->gpc_nr; gpc++)
+			tpc_mask |= ((1 << priv->tpc_nr[gpc]) - 1) << (gpc * 8);
+
+		for (i = 0, gpc = -1, b = -1; i < 32; i++) {
+			a = (i * (priv->tpc_total - 1)) / 32;
+			if (a != b) {
+				b = a;
+				do {
+					gpc = (gpc + 1) % priv->gpc_nr;
+				} while (!tpcnr[gpc]);
+				tpc = priv->tpc_nr[gpc] - tpcnr[gpc]--;
+
+				tpc_set |= 1 << ((gpc * 8) + tpc);
+			}
+
+			nv_wr32(priv, 0x406800 + (i * 0x20), tpc_set);
+			nv_wr32(priv, 0x406c00 + (i * 0x20), tpc_set ^ tpc_mask);
+		}
+	}
+
+	nv_wr32(priv, 0x400208, 0x80000000);
+
+	nv_icmd(priv, 0x00001000, 0x00000004);
+	nv_icmd(priv, 0x000000a9, 0x0000ffff);
+	nv_icmd(priv, 0x00000038, 0x0fac6881);
+	nv_icmd(priv, 0x0000003d, 0x00000001);
+	nv_icmd(priv, 0x000000e8, 0x00000400);
+	nv_icmd(priv, 0x000000e9, 0x00000400);
+	nv_icmd(priv, 0x000000ea, 0x00000400);
+	nv_icmd(priv, 0x000000eb, 0x00000400);
+	nv_icmd(priv, 0x000000ec, 0x00000400);
+	nv_icmd(priv, 0x000000ed, 0x00000400);
+	nv_icmd(priv, 0x000000ee, 0x00000400);
+	nv_icmd(priv, 0x000000ef, 0x00000400);
+	nv_icmd(priv, 0x00000078, 0x00000300);
+	nv_icmd(priv, 0x00000079, 0x00000300);
+	nv_icmd(priv, 0x0000007a, 0x00000300);
+	nv_icmd(priv, 0x0000007b, 0x00000300);
+	nv_icmd(priv, 0x0000007c, 0x00000300);
+	nv_icmd(priv, 0x0000007d, 0x00000300);
+	nv_icmd(priv, 0x0000007e, 0x00000300);
+	nv_icmd(priv, 0x0000007f, 0x00000300);
+	nv_icmd(priv, 0x00000050, 0x00000011);
+	nv_icmd(priv, 0x00000058, 0x00000008);
+	nv_icmd(priv, 0x00000059, 0x00000008);
+	nv_icmd(priv, 0x0000005a, 0x00000008);
+	nv_icmd(priv, 0x0000005b, 0x00000008);
+	nv_icmd(priv, 0x0000005c, 0x00000008);
+	nv_icmd(priv, 0x0000005d, 0x00000008);
+	nv_icmd(priv, 0x0000005e, 0x00000008);
+	nv_icmd(priv, 0x0000005f, 0x00000008);
+	nv_icmd(priv, 0x00000208, 0x00000001);
+	nv_icmd(priv, 0x00000209, 0x00000001);
+	nv_icmd(priv, 0x0000020a, 0x00000001);
+	nv_icmd(priv, 0x0000020b, 0x00000001);
+	nv_icmd(priv, 0x0000020c, 0x00000001);
+	nv_icmd(priv, 0x0000020d, 0x00000001);
+	nv_icmd(priv, 0x0000020e, 0x00000001);
+	nv_icmd(priv, 0x0000020f, 0x00000001);
+	nv_icmd(priv, 0x00000081, 0x00000001);
+	nv_icmd(priv, 0x00000085, 0x00000004);
+	nv_icmd(priv, 0x00000088, 0x00000400);
+	nv_icmd(priv, 0x00000090, 0x00000300);
+	nv_icmd(priv, 0x00000098, 0x00001001);
+	nv_icmd(priv, 0x000000e3, 0x00000001);
+	nv_icmd(priv, 0x000000da, 0x00000001);
+	nv_icmd(priv, 0x000000f8, 0x00000003);
+	nv_icmd(priv, 0x000000fa, 0x00000001);
+	nv_icmd(priv, 0x0000009f, 0x0000ffff);
+	nv_icmd(priv, 0x000000a0, 0x0000ffff);
+	nv_icmd(priv, 0x000000a1, 0x0000ffff);
+	nv_icmd(priv, 0x000000a2, 0x0000ffff);
+	nv_icmd(priv, 0x000000b1, 0x00000001);
+	nv_icmd(priv, 0x000000b2, 0x00000000);
+	nv_icmd(priv, 0x000000b3, 0x00000000);
+	nv_icmd(priv, 0x000000b4, 0x00000000);
+	nv_icmd(priv, 0x000000b5, 0x00000000);
+	nv_icmd(priv, 0x000000b6, 0x00000000);
+	nv_icmd(priv, 0x000000b7, 0x00000000);
+	nv_icmd(priv, 0x000000b8, 0x00000000);
+	nv_icmd(priv, 0x000000b9, 0x00000000);
+	nv_icmd(priv, 0x000000ba, 0x00000000);
+	nv_icmd(priv, 0x000000bb, 0x00000000);
+	nv_icmd(priv, 0x000000bc, 0x00000000);
+	nv_icmd(priv, 0x000000bd, 0x00000000);
+	nv_icmd(priv, 0x000000be, 0x00000000);
+	nv_icmd(priv, 0x000000bf, 0x00000000);
+	nv_icmd(priv, 0x000000c0, 0x00000000);
+	nv_icmd(priv, 0x000000c1, 0x00000000);
+	nv_icmd(priv, 0x000000c2, 0x00000000);
+	nv_icmd(priv, 0x000000c3, 0x00000000);
+	nv_icmd(priv, 0x000000c4, 0x00000000);
+	nv_icmd(priv, 0x000000c5, 0x00000000);
+	nv_icmd(priv, 0x000000c6, 0x00000000);
+	nv_icmd(priv, 0x000000c7, 0x00000000);
+	nv_icmd(priv, 0x000000c8, 0x00000000);
+	nv_icmd(priv, 0x000000c9, 0x00000000);
+	nv_icmd(priv, 0x000000ca, 0x00000000);
+	nv_icmd(priv, 0x000000cb, 0x00000000);
+	nv_icmd(priv, 0x000000cc, 0x00000000);
+	nv_icmd(priv, 0x000000cd, 0x00000000);
+	nv_icmd(priv, 0x000000ce, 0x00000000);
+	nv_icmd(priv, 0x000000cf, 0x00000000);
+	nv_icmd(priv, 0x000000d0, 0x00000000);
+	nv_icmd(priv, 0x000000d1, 0x00000000);
+	nv_icmd(priv, 0x000000d2, 0x00000000);
+	nv_icmd(priv, 0x000000d3, 0x00000000);
+	nv_icmd(priv, 0x000000d4, 0x00000000);
+	nv_icmd(priv, 0x000000d5, 0x00000000);
+	nv_icmd(priv, 0x000000d6, 0x00000000);
+	nv_icmd(priv, 0x000000d7, 0x00000000);
+	nv_icmd(priv, 0x000000d8, 0x00000000);
+	nv_icmd(priv, 0x000000d9, 0x00000000);
+	nv_icmd(priv, 0x00000210, 0x00000040);
+	nv_icmd(priv, 0x00000211, 0x00000040);
+	nv_icmd(priv, 0x00000212, 0x00000040);
+	nv_icmd(priv, 0x00000213, 0x00000040);
+	nv_icmd(priv, 0x00000214, 0x00000040);
+	nv_icmd(priv, 0x00000215, 0x00000040);
+	nv_icmd(priv, 0x00000216, 0x00000040);
+	nv_icmd(priv, 0x00000217, 0x00000040);
+	if (nv_device(priv)->chipset == 0xd9) {
+		for (i = 0x0400; i <= 0x0417; i++)
+			nv_icmd(priv, i, 0x00000040);
+	}
+	nv_icmd(priv, 0x00000218, 0x0000c080);
+	nv_icmd(priv, 0x00000219, 0x0000c080);
+	nv_icmd(priv, 0x0000021a, 0x0000c080);
+	nv_icmd(priv, 0x0000021b, 0x0000c080);
+	nv_icmd(priv, 0x0000021c, 0x0000c080);
+	nv_icmd(priv, 0x0000021d, 0x0000c080);
+	nv_icmd(priv, 0x0000021e, 0x0000c080);
+	nv_icmd(priv, 0x0000021f, 0x0000c080);
+	if (nv_device(priv)->chipset == 0xd9) {
+		for (i = 0x0440; i <= 0x0457; i++)
+			nv_icmd(priv, i, 0x0000c080);
+	}
+	nv_icmd(priv, 0x000000ad, 0x0000013e);
+	nv_icmd(priv, 0x000000e1, 0x00000010);
+	nv_icmd(priv, 0x00000290, 0x00000000);
+	nv_icmd(priv, 0x00000291, 0x00000000);
+	nv_icmd(priv, 0x00000292, 0x00000000);
+	nv_icmd(priv, 0x00000293, 0x00000000);
+	nv_icmd(priv, 0x00000294, 0x00000000);
+	nv_icmd(priv, 0x00000295, 0x00000000);
+	nv_icmd(priv, 0x00000296, 0x00000000);
+	nv_icmd(priv, 0x00000297, 0x00000000);
+	nv_icmd(priv, 0x00000298, 0x00000000);
+	nv_icmd(priv, 0x00000299, 0x00000000);
+	nv_icmd(priv, 0x0000029a, 0x00000000);
+	nv_icmd(priv, 0x0000029b, 0x00000000);
+	nv_icmd(priv, 0x0000029c, 0x00000000);
+	nv_icmd(priv, 0x0000029d, 0x00000000);
+	nv_icmd(priv, 0x0000029e, 0x00000000);
+	nv_icmd(priv, 0x0000029f, 0x00000000);
+	nv_icmd(priv, 0x000003b0, 0x00000000);
+	nv_icmd(priv, 0x000003b1, 0x00000000);
+	nv_icmd(priv, 0x000003b2, 0x00000000);
+	nv_icmd(priv, 0x000003b3, 0x00000000);
+	nv_icmd(priv, 0x000003b4, 0x00000000);
+	nv_icmd(priv, 0x000003b5, 0x00000000);
+	nv_icmd(priv, 0x000003b6, 0x00000000);
+	nv_icmd(priv, 0x000003b7, 0x00000000);
+	nv_icmd(priv, 0x000003b8, 0x00000000);
+	nv_icmd(priv, 0x000003b9, 0x00000000);
+	nv_icmd(priv, 0x000003ba, 0x00000000);
+	nv_icmd(priv, 0x000003bb, 0x00000000);
+	nv_icmd(priv, 0x000003bc, 0x00000000);
+	nv_icmd(priv, 0x000003bd, 0x00000000);
+	nv_icmd(priv, 0x000003be, 0x00000000);
+	nv_icmd(priv, 0x000003bf, 0x00000000);
+	nv_icmd(priv, 0x000002a0, 0x00000000);
+	nv_icmd(priv, 0x000002a1, 0x00000000);
+	nv_icmd(priv, 0x000002a2, 0x00000000);
+	nv_icmd(priv, 0x000002a3, 0x00000000);
+	nv_icmd(priv, 0x000002a4, 0x00000000);
+	nv_icmd(priv, 0x000002a5, 0x00000000);
+	nv_icmd(priv, 0x000002a6, 0x00000000);
+	nv_icmd(priv, 0x000002a7, 0x00000000);
+	nv_icmd(priv, 0x000002a8, 0x00000000);
+	nv_icmd(priv, 0x000002a9, 0x00000000);
+	nv_icmd(priv, 0x000002aa, 0x00000000);
+	nv_icmd(priv, 0x000002ab, 0x00000000);
+	nv_icmd(priv, 0x000002ac, 0x00000000);
+	nv_icmd(priv, 0x000002ad, 0x00000000);
+	nv_icmd(priv, 0x000002ae, 0x00000000);
+	nv_icmd(priv, 0x000002af, 0x00000000);
+	nv_icmd(priv, 0x00000420, 0x00000000);
+	nv_icmd(priv, 0x00000421, 0x00000000);
+	nv_icmd(priv, 0x00000422, 0x00000000);
+	nv_icmd(priv, 0x00000423, 0x00000000);
+	nv_icmd(priv, 0x00000424, 0x00000000);
+	nv_icmd(priv, 0x00000425, 0x00000000);
+	nv_icmd(priv, 0x00000426, 0x00000000);
+	nv_icmd(priv, 0x00000427, 0x00000000);
+	nv_icmd(priv, 0x00000428, 0x00000000);
+	nv_icmd(priv, 0x00000429, 0x00000000);
+	nv_icmd(priv, 0x0000042a, 0x00000000);
+	nv_icmd(priv, 0x0000042b, 0x00000000);
+	nv_icmd(priv, 0x0000042c, 0x00000000);
+	nv_icmd(priv, 0x0000042d, 0x00000000);
+	nv_icmd(priv, 0x0000042e, 0x00000000);
+	nv_icmd(priv, 0x0000042f, 0x00000000);
+	nv_icmd(priv, 0x000002b0, 0x00000000);
+	nv_icmd(priv, 0x000002b1, 0x00000000);
+	nv_icmd(priv, 0x000002b2, 0x00000000);
+	nv_icmd(priv, 0x000002b3, 0x00000000);
+	nv_icmd(priv, 0x000002b4, 0x00000000);
+	nv_icmd(priv, 0x000002b5, 0x00000000);
+	nv_icmd(priv, 0x000002b6, 0x00000000);
+	nv_icmd(priv, 0x000002b7, 0x00000000);
+	nv_icmd(priv, 0x000002b8, 0x00000000);
+	nv_icmd(priv, 0x000002b9, 0x00000000);
+	nv_icmd(priv, 0x000002ba, 0x00000000);
+	nv_icmd(priv, 0x000002bb, 0x00000000);
+	nv_icmd(priv, 0x000002bc, 0x00000000);
+	nv_icmd(priv, 0x000002bd, 0x00000000);
+	nv_icmd(priv, 0x000002be, 0x00000000);
+	nv_icmd(priv, 0x000002bf, 0x00000000);
+	nv_icmd(priv, 0x00000430, 0x00000000);
+	nv_icmd(priv, 0x00000431, 0x00000000);
+	nv_icmd(priv, 0x00000432, 0x00000000);
+	nv_icmd(priv, 0x00000433, 0x00000000);
+	nv_icmd(priv, 0x00000434, 0x00000000);
+	nv_icmd(priv, 0x00000435, 0x00000000);
+	nv_icmd(priv, 0x00000436, 0x00000000);
+	nv_icmd(priv, 0x00000437, 0x00000000);
+	nv_icmd(priv, 0x00000438, 0x00000000);
+	nv_icmd(priv, 0x00000439, 0x00000000);
+	nv_icmd(priv, 0x0000043a, 0x00000000);
+	nv_icmd(priv, 0x0000043b, 0x00000000);
+	nv_icmd(priv, 0x0000043c, 0x00000000);
+	nv_icmd(priv, 0x0000043d, 0x00000000);
+	nv_icmd(priv, 0x0000043e, 0x00000000);
+	nv_icmd(priv, 0x0000043f, 0x00000000);
+	nv_icmd(priv, 0x000002c0, 0x00000000);
+	nv_icmd(priv, 0x000002c1, 0x00000000);
+	nv_icmd(priv, 0x000002c2, 0x00000000);
+	nv_icmd(priv, 0x000002c3, 0x00000000);
+	nv_icmd(priv, 0x000002c4, 0x00000000);
+	nv_icmd(priv, 0x000002c5, 0x00000000);
+	nv_icmd(priv, 0x000002c6, 0x00000000);
+	nv_icmd(priv, 0x000002c7, 0x00000000);
+	nv_icmd(priv, 0x000002c8, 0x00000000);
+	nv_icmd(priv, 0x000002c9, 0x00000000);
+	nv_icmd(priv, 0x000002ca, 0x00000000);
+	nv_icmd(priv, 0x000002cb, 0x00000000);
+	nv_icmd(priv, 0x000002cc, 0x00000000);
+	nv_icmd(priv, 0x000002cd, 0x00000000);
+	nv_icmd(priv, 0x000002ce, 0x00000000);
+	nv_icmd(priv, 0x000002cf, 0x00000000);
+	nv_icmd(priv, 0x000004d0, 0x00000000);
+	nv_icmd(priv, 0x000004d1, 0x00000000);
+	nv_icmd(priv, 0x000004d2, 0x00000000);
+	nv_icmd(priv, 0x000004d3, 0x00000000);
+	nv_icmd(priv, 0x000004d4, 0x00000000);
+	nv_icmd(priv, 0x000004d5, 0x00000000);
+	nv_icmd(priv, 0x000004d6, 0x00000000);
+	nv_icmd(priv, 0x000004d7, 0x00000000);
+	nv_icmd(priv, 0x000004d8, 0x00000000);
+	nv_icmd(priv, 0x000004d9, 0x00000000);
+	nv_icmd(priv, 0x000004da, 0x00000000);
+	nv_icmd(priv, 0x000004db, 0x00000000);
+	nv_icmd(priv, 0x000004dc, 0x00000000);
+	nv_icmd(priv, 0x000004dd, 0x00000000);
+	nv_icmd(priv, 0x000004de, 0x00000000);
+	nv_icmd(priv, 0x000004df, 0x00000000);
+	nv_icmd(priv, 0x00000720, 0x00000000);
+	nv_icmd(priv, 0x00000721, 0x00000000);
+	nv_icmd(priv, 0x00000722, 0x00000000);
+	nv_icmd(priv, 0x00000723, 0x00000000);
+	nv_icmd(priv, 0x00000724, 0x00000000);
+	nv_icmd(priv, 0x00000725, 0x00000000);
+	nv_icmd(priv, 0x00000726, 0x00000000);
+	nv_icmd(priv, 0x00000727, 0x00000000);
+	nv_icmd(priv, 0x00000728, 0x00000000);
+	nv_icmd(priv, 0x00000729, 0x00000000);
+	nv_icmd(priv, 0x0000072a, 0x00000000);
+	nv_icmd(priv, 0x0000072b, 0x00000000);
+	nv_icmd(priv, 0x0000072c, 0x00000000);
+	nv_icmd(priv, 0x0000072d, 0x00000000);
+	nv_icmd(priv, 0x0000072e, 0x00000000);
+	nv_icmd(priv, 0x0000072f, 0x00000000);
+	nv_icmd(priv, 0x000008c0, 0x00000000);
+	nv_icmd(priv, 0x000008c1, 0x00000000);
+	nv_icmd(priv, 0x000008c2, 0x00000000);
+	nv_icmd(priv, 0x000008c3, 0x00000000);
+	nv_icmd(priv, 0x000008c4, 0x00000000);
+	nv_icmd(priv, 0x000008c5, 0x00000000);
+	nv_icmd(priv, 0x000008c6, 0x00000000);
+	nv_icmd(priv, 0x000008c7, 0x00000000);
+	nv_icmd(priv, 0x000008c8, 0x00000000);
+	nv_icmd(priv, 0x000008c9, 0x00000000);
+	nv_icmd(priv, 0x000008ca, 0x00000000);
+	nv_icmd(priv, 0x000008cb, 0x00000000);
+	nv_icmd(priv, 0x000008cc, 0x00000000);
+	nv_icmd(priv, 0x000008cd, 0x00000000);
+	nv_icmd(priv, 0x000008ce, 0x00000000);
+	nv_icmd(priv, 0x000008cf, 0x00000000);
+	nv_icmd(priv, 0x00000890, 0x00000000);
+	nv_icmd(priv, 0x00000891, 0x00000000);
+	nv_icmd(priv, 0x00000892, 0x00000000);
+	nv_icmd(priv, 0x00000893, 0x00000000);
+	nv_icmd(priv, 0x00000894, 0x00000000);
+	nv_icmd(priv, 0x00000895, 0x00000000);
+	nv_icmd(priv, 0x00000896, 0x00000000);
+	nv_icmd(priv, 0x00000897, 0x00000000);
+	nv_icmd(priv, 0x00000898, 0x00000000);
+	nv_icmd(priv, 0x00000899, 0x00000000);
+	nv_icmd(priv, 0x0000089a, 0x00000000);
+	nv_icmd(priv, 0x0000089b, 0x00000000);
+	nv_icmd(priv, 0x0000089c, 0x00000000);
+	nv_icmd(priv, 0x0000089d, 0x00000000);
+	nv_icmd(priv, 0x0000089e, 0x00000000);
+	nv_icmd(priv, 0x0000089f, 0x00000000);
+	nv_icmd(priv, 0x000008e0, 0x00000000);
+	nv_icmd(priv, 0x000008e1, 0x00000000);
+	nv_icmd(priv, 0x000008e2, 0x00000000);
+	nv_icmd(priv, 0x000008e3, 0x00000000);
+	nv_icmd(priv, 0x000008e4, 0x00000000);
+	nv_icmd(priv, 0x000008e5, 0x00000000);
+	nv_icmd(priv, 0x000008e6, 0x00000000);
+	nv_icmd(priv, 0x000008e7, 0x00000000);
+	nv_icmd(priv, 0x000008e8, 0x00000000);
+	nv_icmd(priv, 0x000008e9, 0x00000000);
+	nv_icmd(priv, 0x000008ea, 0x00000000);
+	nv_icmd(priv, 0x000008eb, 0x00000000);
+	nv_icmd(priv, 0x000008ec, 0x00000000);
+	nv_icmd(priv, 0x000008ed, 0x00000000);
+	nv_icmd(priv, 0x000008ee, 0x00000000);
+	nv_icmd(priv, 0x000008ef, 0x00000000);
+	nv_icmd(priv, 0x000008a0, 0x00000000);
+	nv_icmd(priv, 0x000008a1, 0x00000000);
+	nv_icmd(priv, 0x000008a2, 0x00000000);
+	nv_icmd(priv, 0x000008a3, 0x00000000);
+	nv_icmd(priv, 0x000008a4, 0x00000000);
+	nv_icmd(priv, 0x000008a5, 0x00000000);
+	nv_icmd(priv, 0x000008a6, 0x00000000);
+	nv_icmd(priv, 0x000008a7, 0x00000000);
+	nv_icmd(priv, 0x000008a8, 0x00000000);
+	nv_icmd(priv, 0x000008a9, 0x00000000);
+	nv_icmd(priv, 0x000008aa, 0x00000000);
+	nv_icmd(priv, 0x000008ab, 0x00000000);
+	nv_icmd(priv, 0x000008ac, 0x00000000);
+	nv_icmd(priv, 0x000008ad, 0x00000000);
+	nv_icmd(priv, 0x000008ae, 0x00000000);
+	nv_icmd(priv, 0x000008af, 0x00000000);
+	nv_icmd(priv, 0x000008f0, 0x00000000);
+	nv_icmd(priv, 0x000008f1, 0x00000000);
+	nv_icmd(priv, 0x000008f2, 0x00000000);
+	nv_icmd(priv, 0x000008f3, 0x00000000);
+	nv_icmd(priv, 0x000008f4, 0x00000000);
+	nv_icmd(priv, 0x000008f5, 0x00000000);
+	nv_icmd(priv, 0x000008f6, 0x00000000);
+	nv_icmd(priv, 0x000008f7, 0x00000000);
+	nv_icmd(priv, 0x000008f8, 0x00000000);
+	nv_icmd(priv, 0x000008f9, 0x00000000);
+	nv_icmd(priv, 0x000008fa, 0x00000000);
+	nv_icmd(priv, 0x000008fb, 0x00000000);
+	nv_icmd(priv, 0x000008fc, 0x00000000);
+	nv_icmd(priv, 0x000008fd, 0x00000000);
+	nv_icmd(priv, 0x000008fe, 0x00000000);
+	nv_icmd(priv, 0x000008ff, 0x00000000);
+	nv_icmd(priv, 0x0000094c, 0x000000ff);
+	nv_icmd(priv, 0x0000094d, 0xffffffff);
+	nv_icmd(priv, 0x0000094e, 0x00000002);
+	nv_icmd(priv, 0x000002ec, 0x00000001);
+	nv_icmd(priv, 0x00000303, 0x00000001);
+	nv_icmd(priv, 0x000002e6, 0x00000001);
+	nv_icmd(priv, 0x00000466, 0x00000052);
+	nv_icmd(priv, 0x00000301, 0x3f800000);
+	nv_icmd(priv, 0x00000304, 0x30201000);
+	nv_icmd(priv, 0x00000305, 0x70605040);
+	nv_icmd(priv, 0x00000306, 0xb8a89888);
+	nv_icmd(priv, 0x00000307, 0xf8e8d8c8);
+	nv_icmd(priv, 0x0000030a, 0x00ffff00);
+	nv_icmd(priv, 0x0000030b, 0x0000001a);
+	nv_icmd(priv, 0x0000030c, 0x00000001);
+	nv_icmd(priv, 0x00000318, 0x00000001);
+	nv_icmd(priv, 0x00000340, 0x00000000);
+	nv_icmd(priv, 0x00000375, 0x00000001);
+	nv_icmd(priv, 0x00000351, 0x00000100);
+	nv_icmd(priv, 0x0000037d, 0x00000006);
+	nv_icmd(priv, 0x000003a0, 0x00000002);
+	nv_icmd(priv, 0x000003aa, 0x00000001);
+	nv_icmd(priv, 0x000003a9, 0x00000001);
+	nv_icmd(priv, 0x00000380, 0x00000001);
+	nv_icmd(priv, 0x00000360, 0x00000040);
+	nv_icmd(priv, 0x00000366, 0x00000000);
+	nv_icmd(priv, 0x00000367, 0x00000000);
+	nv_icmd(priv, 0x00000368, 0x00001fff);
+	nv_icmd(priv, 0x00000370, 0x00000000);
+	nv_icmd(priv, 0x00000371, 0x00000000);
+	nv_icmd(priv, 0x00000372, 0x003fffff);
+	nv_icmd(priv, 0x0000037a, 0x00000012);
+	nv_icmd(priv, 0x000005e0, 0x00000022);
+	nv_icmd(priv, 0x000005e1, 0x00000022);
+	nv_icmd(priv, 0x000005e2, 0x00000022);
+	nv_icmd(priv, 0x000005e3, 0x00000022);
+	nv_icmd(priv, 0x000005e4, 0x00000022);
+	nv_icmd(priv, 0x00000619, 0x00000003);
+	nv_icmd(priv, 0x00000811, 0x00000003);
+	nv_icmd(priv, 0x00000812, 0x00000004);
+	nv_icmd(priv, 0x00000813, 0x00000006);
+	nv_icmd(priv, 0x00000814, 0x00000008);
+	nv_icmd(priv, 0x00000815, 0x0000000b);
+	nv_icmd(priv, 0x00000800, 0x00000001);
+	nv_icmd(priv, 0x00000801, 0x00000001);
+	nv_icmd(priv, 0x00000802, 0x00000001);
+	nv_icmd(priv, 0x00000803, 0x00000001);
+	nv_icmd(priv, 0x00000804, 0x00000001);
+	nv_icmd(priv, 0x00000805, 0x00000001);
+	nv_icmd(priv, 0x00000632, 0x00000001);
+	nv_icmd(priv, 0x00000633, 0x00000002);
+	nv_icmd(priv, 0x00000634, 0x00000003);
+	nv_icmd(priv, 0x00000635, 0x00000004);
+	nv_icmd(priv, 0x00000654, 0x3f800000);
+	nv_icmd(priv, 0x00000657, 0x3f800000);
+	nv_icmd(priv, 0x00000655, 0x3f800000);
+	nv_icmd(priv, 0x00000656, 0x3f800000);
+	nv_icmd(priv, 0x000006cd, 0x3f800000);
+	nv_icmd(priv, 0x000007f5, 0x3f800000);
+	nv_icmd(priv, 0x000007dc, 0x39291909);
+	nv_icmd(priv, 0x000007dd, 0x79695949);
+	nv_icmd(priv, 0x000007de, 0xb9a99989);
+	nv_icmd(priv, 0x000007df, 0xf9e9d9c9);
+	nv_icmd(priv, 0x000007e8, 0x00003210);
+	nv_icmd(priv, 0x000007e9, 0x00007654);
+	nv_icmd(priv, 0x000007ea, 0x00000098);
+	nv_icmd(priv, 0x000007ec, 0x39291909);
+	nv_icmd(priv, 0x000007ed, 0x79695949);
+	nv_icmd(priv, 0x000007ee, 0xb9a99989);
+	nv_icmd(priv, 0x000007ef, 0xf9e9d9c9);
+	nv_icmd(priv, 0x000007f0, 0x00003210);
+	nv_icmd(priv, 0x000007f1, 0x00007654);
+	nv_icmd(priv, 0x000007f2, 0x00000098);
+	nv_icmd(priv, 0x000005a5, 0x00000001);
+	nv_icmd(priv, 0x00000980, 0x00000000);
+	nv_icmd(priv, 0x00000981, 0x00000000);
+	nv_icmd(priv, 0x00000982, 0x00000000);
+	nv_icmd(priv, 0x00000983, 0x00000000);
+	nv_icmd(priv, 0x00000984, 0x00000000);
+	nv_icmd(priv, 0x00000985, 0x00000000);
+	nv_icmd(priv, 0x00000986, 0x00000000);
+	nv_icmd(priv, 0x00000987, 0x00000000);
+	nv_icmd(priv, 0x00000988, 0x00000000);
+	nv_icmd(priv, 0x00000989, 0x00000000);
+	nv_icmd(priv, 0x0000098a, 0x00000000);
+	nv_icmd(priv, 0x0000098b, 0x00000000);
+	nv_icmd(priv, 0x0000098c, 0x00000000);
+	nv_icmd(priv, 0x0000098d, 0x00000000);
+	nv_icmd(priv, 0x0000098e, 0x00000000);
+	nv_icmd(priv, 0x0000098f, 0x00000000);
+	nv_icmd(priv, 0x00000990, 0x00000000);
+	nv_icmd(priv, 0x00000991, 0x00000000);
+	nv_icmd(priv, 0x00000992, 0x00000000);
+	nv_icmd(priv, 0x00000993, 0x00000000);
+	nv_icmd(priv, 0x00000994, 0x00000000);
+	nv_icmd(priv, 0x00000995, 0x00000000);
+	nv_icmd(priv, 0x00000996, 0x00000000);
+	nv_icmd(priv, 0x00000997, 0x00000000);
+	nv_icmd(priv, 0x00000998, 0x00000000);
+	nv_icmd(priv, 0x00000999, 0x00000000);
+	nv_icmd(priv, 0x0000099a, 0x00000000);
+	nv_icmd(priv, 0x0000099b, 0x00000000);
+	nv_icmd(priv, 0x0000099c, 0x00000000);
+	nv_icmd(priv, 0x0000099d, 0x00000000);
+	nv_icmd(priv, 0x0000099e, 0x00000000);
+	nv_icmd(priv, 0x0000099f, 0x00000000);
+	nv_icmd(priv, 0x000009a0, 0x00000000);
+	nv_icmd(priv, 0x000009a1, 0x00000000);
+	nv_icmd(priv, 0x000009a2, 0x00000000);
+	nv_icmd(priv, 0x000009a3, 0x00000000);
+	nv_icmd(priv, 0x000009a4, 0x00000000);
+	nv_icmd(priv, 0x000009a5, 0x00000000);
+	nv_icmd(priv, 0x000009a6, 0x00000000);
+	nv_icmd(priv, 0x000009a7, 0x00000000);
+	nv_icmd(priv, 0x000009a8, 0x00000000);
+	nv_icmd(priv, 0x000009a9, 0x00000000);
+	nv_icmd(priv, 0x000009aa, 0x00000000);
+	nv_icmd(priv, 0x000009ab, 0x00000000);
+	nv_icmd(priv, 0x000009ac, 0x00000000);
+	nv_icmd(priv, 0x000009ad, 0x00000000);
+	nv_icmd(priv, 0x000009ae, 0x00000000);
+	nv_icmd(priv, 0x000009af, 0x00000000);
+	nv_icmd(priv, 0x000009b0, 0x00000000);
+	nv_icmd(priv, 0x000009b1, 0x00000000);
+	nv_icmd(priv, 0x000009b2, 0x00000000);
+	nv_icmd(priv, 0x000009b3, 0x00000000);
+	nv_icmd(priv, 0x000009b4, 0x00000000);
+	nv_icmd(priv, 0x000009b5, 0x00000000);
+	nv_icmd(priv, 0x000009b6, 0x00000000);
+	nv_icmd(priv, 0x000009b7, 0x00000000);
+	nv_icmd(priv, 0x000009b8, 0x00000000);
+	nv_icmd(priv, 0x000009b9, 0x00000000);
+	nv_icmd(priv, 0x000009ba, 0x00000000);
+	nv_icmd(priv, 0x000009bb, 0x00000000);
+	nv_icmd(priv, 0x000009bc, 0x00000000);
+	nv_icmd(priv, 0x000009bd, 0x00000000);
+	nv_icmd(priv, 0x000009be, 0x00000000);
+	nv_icmd(priv, 0x000009bf, 0x00000000);
+	nv_icmd(priv, 0x000009c0, 0x00000000);
+	nv_icmd(priv, 0x000009c1, 0x00000000);
+	nv_icmd(priv, 0x000009c2, 0x00000000);
+	nv_icmd(priv, 0x000009c3, 0x00000000);
+	nv_icmd(priv, 0x000009c4, 0x00000000);
+	nv_icmd(priv, 0x000009c5, 0x00000000);
+	nv_icmd(priv, 0x000009c6, 0x00000000);
+	nv_icmd(priv, 0x000009c7, 0x00000000);
+	nv_icmd(priv, 0x000009c8, 0x00000000);
+	nv_icmd(priv, 0x000009c9, 0x00000000);
+	nv_icmd(priv, 0x000009ca, 0x00000000);
+	nv_icmd(priv, 0x000009cb, 0x00000000);
+	nv_icmd(priv, 0x000009cc, 0x00000000);
+	nv_icmd(priv, 0x000009cd, 0x00000000);
+	nv_icmd(priv, 0x000009ce, 0x00000000);
+	nv_icmd(priv, 0x000009cf, 0x00000000);
+	nv_icmd(priv, 0x000009d0, 0x00000000);
+	nv_icmd(priv, 0x000009d1, 0x00000000);
+	nv_icmd(priv, 0x000009d2, 0x00000000);
+	nv_icmd(priv, 0x000009d3, 0x00000000);
+	nv_icmd(priv, 0x000009d4, 0x00000000);
+	nv_icmd(priv, 0x000009d5, 0x00000000);
+	nv_icmd(priv, 0x000009d6, 0x00000000);
+	nv_icmd(priv, 0x000009d7, 0x00000000);
+	nv_icmd(priv, 0x000009d8, 0x00000000);
+	nv_icmd(priv, 0x000009d9, 0x00000000);
+	nv_icmd(priv, 0x000009da, 0x00000000);
+	nv_icmd(priv, 0x000009db, 0x00000000);
+	nv_icmd(priv, 0x000009dc, 0x00000000);
+	nv_icmd(priv, 0x000009dd, 0x00000000);
+	nv_icmd(priv, 0x000009de, 0x00000000);
+	nv_icmd(priv, 0x000009df, 0x00000000);
+	nv_icmd(priv, 0x000009e0, 0x00000000);
+	nv_icmd(priv, 0x000009e1, 0x00000000);
+	nv_icmd(priv, 0x000009e2, 0x00000000);
+	nv_icmd(priv, 0x000009e3, 0x00000000);
+	nv_icmd(priv, 0x000009e4, 0x00000000);
+	nv_icmd(priv, 0x000009e5, 0x00000000);
+	nv_icmd(priv, 0x000009e6, 0x00000000);
+	nv_icmd(priv, 0x000009e7, 0x00000000);
+	nv_icmd(priv, 0x000009e8, 0x00000000);
+	nv_icmd(priv, 0x000009e9, 0x00000000);
+	nv_icmd(priv, 0x000009ea, 0x00000000);
+	nv_icmd(priv, 0x000009eb, 0x00000000);
+	nv_icmd(priv, 0x000009ec, 0x00000000);
+	nv_icmd(priv, 0x000009ed, 0x00000000);
+	nv_icmd(priv, 0x000009ee, 0x00000000);
+	nv_icmd(priv, 0x000009ef, 0x00000000);
+	nv_icmd(priv, 0x000009f0, 0x00000000);
+	nv_icmd(priv, 0x000009f1, 0x00000000);
+	nv_icmd(priv, 0x000009f2, 0x00000000);
+	nv_icmd(priv, 0x000009f3, 0x00000000);
+	nv_icmd(priv, 0x000009f4, 0x00000000);
+	nv_icmd(priv, 0x000009f5, 0x00000000);
+	nv_icmd(priv, 0x000009f6, 0x00000000);
+	nv_icmd(priv, 0x000009f7, 0x00000000);
+	nv_icmd(priv, 0x000009f8, 0x00000000);
+	nv_icmd(priv, 0x000009f9, 0x00000000);
+	nv_icmd(priv, 0x000009fa, 0x00000000);
+	nv_icmd(priv, 0x000009fb, 0x00000000);
+	nv_icmd(priv, 0x000009fc, 0x00000000);
+	nv_icmd(priv, 0x000009fd, 0x00000000);
+	nv_icmd(priv, 0x000009fe, 0x00000000);
+	nv_icmd(priv, 0x000009ff, 0x00000000);
+	nv_icmd(priv, 0x00000468, 0x00000004);
+	nv_icmd(priv, 0x0000046c, 0x00000001);
+	nv_icmd(priv, 0x00000470, 0x00000000);
+	nv_icmd(priv, 0x00000471, 0x00000000);
+	nv_icmd(priv, 0x00000472, 0x00000000);
+	nv_icmd(priv, 0x00000473, 0x00000000);
+	nv_icmd(priv, 0x00000474, 0x00000000);
+	nv_icmd(priv, 0x00000475, 0x00000000);
+	nv_icmd(priv, 0x00000476, 0x00000000);
+	nv_icmd(priv, 0x00000477, 0x00000000);
+	nv_icmd(priv, 0x00000478, 0x00000000);
+	nv_icmd(priv, 0x00000479, 0x00000000);
+	nv_icmd(priv, 0x0000047a, 0x00000000);
+	nv_icmd(priv, 0x0000047b, 0x00000000);
+	nv_icmd(priv, 0x0000047c, 0x00000000);
+	nv_icmd(priv, 0x0000047d, 0x00000000);
+	nv_icmd(priv, 0x0000047e, 0x00000000);
+	nv_icmd(priv, 0x0000047f, 0x00000000);
+	nv_icmd(priv, 0x00000480, 0x00000000);
+	nv_icmd(priv, 0x00000481, 0x00000000);
+	nv_icmd(priv, 0x00000482, 0x00000000);
+	nv_icmd(priv, 0x00000483, 0x00000000);
+	nv_icmd(priv, 0x00000484, 0x00000000);
+	nv_icmd(priv, 0x00000485, 0x00000000);
+	nv_icmd(priv, 0x00000486, 0x00000000);
+	nv_icmd(priv, 0x00000487, 0x00000000);
+	nv_icmd(priv, 0x00000488, 0x00000000);
+	nv_icmd(priv, 0x00000489, 0x00000000);
+	nv_icmd(priv, 0x0000048a, 0x00000000);
+	nv_icmd(priv, 0x0000048b, 0x00000000);
+	nv_icmd(priv, 0x0000048c, 0x00000000);
+	nv_icmd(priv, 0x0000048d, 0x00000000);
+	nv_icmd(priv, 0x0000048e, 0x00000000);
+	nv_icmd(priv, 0x0000048f, 0x00000000);
+	nv_icmd(priv, 0x00000490, 0x00000000);
+	nv_icmd(priv, 0x00000491, 0x00000000);
+	nv_icmd(priv, 0x00000492, 0x00000000);
+	nv_icmd(priv, 0x00000493, 0x00000000);
+	nv_icmd(priv, 0x00000494, 0x00000000);
+	nv_icmd(priv, 0x00000495, 0x00000000);
+	nv_icmd(priv, 0x00000496, 0x00000000);
+	nv_icmd(priv, 0x00000497, 0x00000000);
+	nv_icmd(priv, 0x00000498, 0x00000000);
+	nv_icmd(priv, 0x00000499, 0x00000000);
+	nv_icmd(priv, 0x0000049a, 0x00000000);
+	nv_icmd(priv, 0x0000049b, 0x00000000);
+	nv_icmd(priv, 0x0000049c, 0x00000000);
+	nv_icmd(priv, 0x0000049d, 0x00000000);
+	nv_icmd(priv, 0x0000049e, 0x00000000);
+	nv_icmd(priv, 0x0000049f, 0x00000000);
+	nv_icmd(priv, 0x000004a0, 0x00000000);
+	nv_icmd(priv, 0x000004a1, 0x00000000);
+	nv_icmd(priv, 0x000004a2, 0x00000000);
+	nv_icmd(priv, 0x000004a3, 0x00000000);
+	nv_icmd(priv, 0x000004a4, 0x00000000);
+	nv_icmd(priv, 0x000004a5, 0x00000000);
+	nv_icmd(priv, 0x000004a6, 0x00000000);
+	nv_icmd(priv, 0x000004a7, 0x00000000);
+	nv_icmd(priv, 0x000004a8, 0x00000000);
+	nv_icmd(priv, 0x000004a9, 0x00000000);
+	nv_icmd(priv, 0x000004aa, 0x00000000);
+	nv_icmd(priv, 0x000004ab, 0x00000000);
+	nv_icmd(priv, 0x000004ac, 0x00000000);
+	nv_icmd(priv, 0x000004ad, 0x00000000);
+	nv_icmd(priv, 0x000004ae, 0x00000000);
+	nv_icmd(priv, 0x000004af, 0x00000000);
+	nv_icmd(priv, 0x000004b0, 0x00000000);
+	nv_icmd(priv, 0x000004b1, 0x00000000);
+	nv_icmd(priv, 0x000004b2, 0x00000000);
+	nv_icmd(priv, 0x000004b3, 0x00000000);
+	nv_icmd(priv, 0x000004b4, 0x00000000);
+	nv_icmd(priv, 0x000004b5, 0x00000000);
+	nv_icmd(priv, 0x000004b6, 0x00000000);
+	nv_icmd(priv, 0x000004b7, 0x00000000);
+	nv_icmd(priv, 0x000004b8, 0x00000000);
+	nv_icmd(priv, 0x000004b9, 0x00000000);
+	nv_icmd(priv, 0x000004ba, 0x00000000);
+	nv_icmd(priv, 0x000004bb, 0x00000000);
+	nv_icmd(priv, 0x000004bc, 0x00000000);
+	nv_icmd(priv, 0x000004bd, 0x00000000);
+	nv_icmd(priv, 0x000004be, 0x00000000);
+	nv_icmd(priv, 0x000004bf, 0x00000000);
+	nv_icmd(priv, 0x000004c0, 0x00000000);
+	nv_icmd(priv, 0x000004c1, 0x00000000);
+	nv_icmd(priv, 0x000004c2, 0x00000000);
+	nv_icmd(priv, 0x000004c3, 0x00000000);
+	nv_icmd(priv, 0x000004c4, 0x00000000);
+	nv_icmd(priv, 0x000004c5, 0x00000000);
+	nv_icmd(priv, 0x000004c6, 0x00000000);
+	nv_icmd(priv, 0x000004c7, 0x00000000);
+	nv_icmd(priv, 0x000004c8, 0x00000000);
+	nv_icmd(priv, 0x000004c9, 0x00000000);
+	nv_icmd(priv, 0x000004ca, 0x00000000);
+	nv_icmd(priv, 0x000004cb, 0x00000000);
+	nv_icmd(priv, 0x000004cc, 0x00000000);
+	nv_icmd(priv, 0x000004cd, 0x00000000);
+	nv_icmd(priv, 0x000004ce, 0x00000000);
+	nv_icmd(priv, 0x000004cf, 0x00000000);
+	nv_icmd(priv, 0x00000510, 0x3f800000);
+	nv_icmd(priv, 0x00000511, 0x3f800000);
+	nv_icmd(priv, 0x00000512, 0x3f800000);
+	nv_icmd(priv, 0x00000513, 0x3f800000);
+	nv_icmd(priv, 0x00000514, 0x3f800000);
+	nv_icmd(priv, 0x00000515, 0x3f800000);
+	nv_icmd(priv, 0x00000516, 0x3f800000);
+	nv_icmd(priv, 0x00000517, 0x3f800000);
+	nv_icmd(priv, 0x00000518, 0x3f800000);
+	nv_icmd(priv, 0x00000519, 0x3f800000);
+	nv_icmd(priv, 0x0000051a, 0x3f800000);
+	nv_icmd(priv, 0x0000051b, 0x3f800000);
+	nv_icmd(priv, 0x0000051c, 0x3f800000);
+	nv_icmd(priv, 0x0000051d, 0x3f800000);
+	nv_icmd(priv, 0x0000051e, 0x3f800000);
+	nv_icmd(priv, 0x0000051f, 0x3f800000);
+	nv_icmd(priv, 0x00000520, 0x000002b6);
+	nv_icmd(priv, 0x00000529, 0x00000001);
+	nv_icmd(priv, 0x00000530, 0xffff0000);
+	nv_icmd(priv, 0x00000531, 0xffff0000);
+	nv_icmd(priv, 0x00000532, 0xffff0000);
+	nv_icmd(priv, 0x00000533, 0xffff0000);
+	nv_icmd(priv, 0x00000534, 0xffff0000);
+	nv_icmd(priv, 0x00000535, 0xffff0000);
+	nv_icmd(priv, 0x00000536, 0xffff0000);
+	nv_icmd(priv, 0x00000537, 0xffff0000);
+	nv_icmd(priv, 0x00000538, 0xffff0000);
+	nv_icmd(priv, 0x00000539, 0xffff0000);
+	nv_icmd(priv, 0x0000053a, 0xffff0000);
+	nv_icmd(priv, 0x0000053b, 0xffff0000);
+	nv_icmd(priv, 0x0000053c, 0xffff0000);
+	nv_icmd(priv, 0x0000053d, 0xffff0000);
+	nv_icmd(priv, 0x0000053e, 0xffff0000);
+	nv_icmd(priv, 0x0000053f, 0xffff0000);
+	nv_icmd(priv, 0x00000585, 0x0000003f);
+	nv_icmd(priv, 0x00000576, 0x00000003);
+	if (nv_device(priv)->chipset == 0xc1 ||
+	    nv_device(priv)->chipset == 0xd9)
+		nv_icmd(priv, 0x0000057b, 0x00000059);
+	nv_icmd(priv, 0x00000586, 0x00000040);
+	nv_icmd(priv, 0x00000582, 0x00000080);
+	nv_icmd(priv, 0x00000583, 0x00000080);
+	nv_icmd(priv, 0x000005c2, 0x00000001);
+	nv_icmd(priv, 0x00000638, 0x00000001);
+	nv_icmd(priv, 0x00000639, 0x00000001);
+	nv_icmd(priv, 0x0000063a, 0x00000002);
+	nv_icmd(priv, 0x0000063b, 0x00000001);
+	nv_icmd(priv, 0x0000063c, 0x00000001);
+	nv_icmd(priv, 0x0000063d, 0x00000002);
+	nv_icmd(priv, 0x0000063e, 0x00000001);
+	nv_icmd(priv, 0x000008b8, 0x00000001);
+	nv_icmd(priv, 0x000008b9, 0x00000001);
+	nv_icmd(priv, 0x000008ba, 0x00000001);
+	nv_icmd(priv, 0x000008bb, 0x00000001);
+	nv_icmd(priv, 0x000008bc, 0x00000001);
+	nv_icmd(priv, 0x000008bd, 0x00000001);
+	nv_icmd(priv, 0x000008be, 0x00000001);
+	nv_icmd(priv, 0x000008bf, 0x00000001);
+	nv_icmd(priv, 0x00000900, 0x00000001);
+	nv_icmd(priv, 0x00000901, 0x00000001);
+	nv_icmd(priv, 0x00000902, 0x00000001);
+	nv_icmd(priv, 0x00000903, 0x00000001);
+	nv_icmd(priv, 0x00000904, 0x00000001);
+	nv_icmd(priv, 0x00000905, 0x00000001);
+	nv_icmd(priv, 0x00000906, 0x00000001);
+	nv_icmd(priv, 0x00000907, 0x00000001);
+	nv_icmd(priv, 0x00000908, 0x00000002);
+	nv_icmd(priv, 0x00000909, 0x00000002);
+	nv_icmd(priv, 0x0000090a, 0x00000002);
+	nv_icmd(priv, 0x0000090b, 0x00000002);
+	nv_icmd(priv, 0x0000090c, 0x00000002);
+	nv_icmd(priv, 0x0000090d, 0x00000002);
+	nv_icmd(priv, 0x0000090e, 0x00000002);
+	nv_icmd(priv, 0x0000090f, 0x00000002);
+	nv_icmd(priv, 0x00000910, 0x00000001);
+	nv_icmd(priv, 0x00000911, 0x00000001);
+	nv_icmd(priv, 0x00000912, 0x00000001);
+	nv_icmd(priv, 0x00000913, 0x00000001);
+	nv_icmd(priv, 0x00000914, 0x00000001);
+	nv_icmd(priv, 0x00000915, 0x00000001);
+	nv_icmd(priv, 0x00000916, 0x00000001);
+	nv_icmd(priv, 0x00000917, 0x00000001);
+	nv_icmd(priv, 0x00000918, 0x00000001);
+	nv_icmd(priv, 0x00000919, 0x00000001);
+	nv_icmd(priv, 0x0000091a, 0x00000001);
+	nv_icmd(priv, 0x0000091b, 0x00000001);
+	nv_icmd(priv, 0x0000091c, 0x00000001);
+	nv_icmd(priv, 0x0000091d, 0x00000001);
+	nv_icmd(priv, 0x0000091e, 0x00000001);
+	nv_icmd(priv, 0x0000091f, 0x00000001);
+	nv_icmd(priv, 0x00000920, 0x00000002);
+	nv_icmd(priv, 0x00000921, 0x00000002);
+	nv_icmd(priv, 0x00000922, 0x00000002);
+	nv_icmd(priv, 0x00000923, 0x00000002);
+	nv_icmd(priv, 0x00000924, 0x00000002);
+	nv_icmd(priv, 0x00000925, 0x00000002);
+	nv_icmd(priv, 0x00000926, 0x00000002);
+	nv_icmd(priv, 0x00000927, 0x00000002);
+	nv_icmd(priv, 0x00000928, 0x00000001);
+	nv_icmd(priv, 0x00000929, 0x00000001);
+	nv_icmd(priv, 0x0000092a, 0x00000001);
+	nv_icmd(priv, 0x0000092b, 0x00000001);
+	nv_icmd(priv, 0x0000092c, 0x00000001);
+	nv_icmd(priv, 0x0000092d, 0x00000001);
+	nv_icmd(priv, 0x0000092e, 0x00000001);
+	nv_icmd(priv, 0x0000092f, 0x00000001);
+	nv_icmd(priv, 0x00000648, 0x00000001);
+	nv_icmd(priv, 0x00000649, 0x00000001);
+	nv_icmd(priv, 0x0000064a, 0x00000001);
+	nv_icmd(priv, 0x0000064b, 0x00000001);
+	nv_icmd(priv, 0x0000064c, 0x00000001);
+	nv_icmd(priv, 0x0000064d, 0x00000001);
+	nv_icmd(priv, 0x0000064e, 0x00000001);
+	nv_icmd(priv, 0x0000064f, 0x00000001);
+	nv_icmd(priv, 0x00000650, 0x00000001);
+	nv_icmd(priv, 0x00000658, 0x0000000f);
+	nv_icmd(priv, 0x000007ff, 0x0000000a);
+	nv_icmd(priv, 0x0000066a, 0x40000000);
+	nv_icmd(priv, 0x0000066b, 0x10000000);
+	nv_icmd(priv, 0x0000066c, 0xffff0000);
+	nv_icmd(priv, 0x0000066d, 0xffff0000);
+	nv_icmd(priv, 0x000007af, 0x00000008);
+	nv_icmd(priv, 0x000007b0, 0x00000008);
+	nv_icmd(priv, 0x000007f6, 0x00000001);
+	nv_icmd(priv, 0x000006b2, 0x00000055);
+	nv_icmd(priv, 0x000007ad, 0x00000003);
+	nv_icmd(priv, 0x00000937, 0x00000001);
+	nv_icmd(priv, 0x00000971, 0x00000008);
+	nv_icmd(priv, 0x00000972, 0x00000040);
+	nv_icmd(priv, 0x00000973, 0x0000012c);
+	nv_icmd(priv, 0x0000097c, 0x00000040);
+	nv_icmd(priv, 0x00000979, 0x00000003);
+	nv_icmd(priv, 0x00000975, 0x00000020);
+	nv_icmd(priv, 0x00000976, 0x00000001);
+	nv_icmd(priv, 0x00000977, 0x00000020);
+	nv_icmd(priv, 0x00000978, 0x00000001);
+	nv_icmd(priv, 0x00000957, 0x00000003);
+	nv_icmd(priv, 0x0000095e, 0x20164010);
+	nv_icmd(priv, 0x0000095f, 0x00000020);
+	if (nv_device(priv)->chipset == 0xd9)
+		nv_icmd(priv, 0x0000097d, 0x00000020);
+	nv_icmd(priv, 0x00000683, 0x00000006);
+	nv_icmd(priv, 0x00000685, 0x003fffff);
+	nv_icmd(priv, 0x00000687, 0x00000c48);
+	nv_icmd(priv, 0x000006a0, 0x00000005);
+	nv_icmd(priv, 0x00000840, 0x00300008);
+	nv_icmd(priv, 0x00000841, 0x04000080);
+	nv_icmd(priv, 0x00000842, 0x00300008);
+	nv_icmd(priv, 0x00000843, 0x04000080);
+	nv_icmd(priv, 0x00000818, 0x00000000);
+	nv_icmd(priv, 0x00000819, 0x00000000);
+	nv_icmd(priv, 0x0000081a, 0x00000000);
+	nv_icmd(priv, 0x0000081b, 0x00000000);
+	nv_icmd(priv, 0x0000081c, 0x00000000);
+	nv_icmd(priv, 0x0000081d, 0x00000000);
+	nv_icmd(priv, 0x0000081e, 0x00000000);
+	nv_icmd(priv, 0x0000081f, 0x00000000);
+	nv_icmd(priv, 0x00000848, 0x00000000);
+	nv_icmd(priv, 0x00000849, 0x00000000);
+	nv_icmd(priv, 0x0000084a, 0x00000000);
+	nv_icmd(priv, 0x0000084b, 0x00000000);
+	nv_icmd(priv, 0x0000084c, 0x00000000);
+	nv_icmd(priv, 0x0000084d, 0x00000000);
+	nv_icmd(priv, 0x0000084e, 0x00000000);
+	nv_icmd(priv, 0x0000084f, 0x00000000);
+	nv_icmd(priv, 0x00000850, 0x00000000);
+	nv_icmd(priv, 0x00000851, 0x00000000);
+	nv_icmd(priv, 0x00000852, 0x00000000);
+	nv_icmd(priv, 0x00000853, 0x00000000);
+	nv_icmd(priv, 0x00000854, 0x00000000);
+	nv_icmd(priv, 0x00000855, 0x00000000);
+	nv_icmd(priv, 0x00000856, 0x00000000);
+	nv_icmd(priv, 0x00000857, 0x00000000);
+	nv_icmd(priv, 0x00000738, 0x00000000);
+	nv_icmd(priv, 0x000006aa, 0x00000001);
+	nv_icmd(priv, 0x000006ab, 0x00000002);
+	nv_icmd(priv, 0x000006ac, 0x00000080);
+	nv_icmd(priv, 0x000006ad, 0x00000100);
+	nv_icmd(priv, 0x000006ae, 0x00000100);
+	nv_icmd(priv, 0x000006b1, 0x00000011);
+	nv_icmd(priv, 0x000006bb, 0x000000cf);
+	nv_icmd(priv, 0x000006ce, 0x2a712488);
+	nv_icmd(priv, 0x00000739, 0x4085c000);
+	nv_icmd(priv, 0x0000073a, 0x00000080);
+	nv_icmd(priv, 0x00000786, 0x80000100);
+	nv_icmd(priv, 0x0000073c, 0x00010100);
+	nv_icmd(priv, 0x0000073d, 0x02800000);
+	nv_icmd(priv, 0x00000787, 0x000000cf);
+	nv_icmd(priv, 0x0000078c, 0x00000008);
+	nv_icmd(priv, 0x00000792, 0x00000001);
+	nv_icmd(priv, 0x00000794, 0x00000001);
+	nv_icmd(priv, 0x00000795, 0x00000001);
+	nv_icmd(priv, 0x00000796, 0x00000001);
+	nv_icmd(priv, 0x00000797, 0x000000cf);
+	nv_icmd(priv, 0x00000836, 0x00000001);
+	nv_icmd(priv, 0x0000079a, 0x00000002);
+	nv_icmd(priv, 0x00000833, 0x04444480);
+	nv_icmd(priv, 0x000007a1, 0x00000001);
+	nv_icmd(priv, 0x000007a3, 0x00000001);
+	nv_icmd(priv, 0x000007a4, 0x00000001);
+	nv_icmd(priv, 0x000007a5, 0x00000001);
+	nv_icmd(priv, 0x00000831, 0x00000004);
+	nv_icmd(priv, 0x0000080c, 0x00000002);
+	nv_icmd(priv, 0x0000080d, 0x00000100);
+	nv_icmd(priv, 0x0000080e, 0x00000100);
+	nv_icmd(priv, 0x0000080f, 0x00000001);
+	nv_icmd(priv, 0x00000823, 0x00000002);
+	nv_icmd(priv, 0x00000824, 0x00000100);
+	nv_icmd(priv, 0x00000825, 0x00000100);
+	nv_icmd(priv, 0x00000826, 0x00000001);
+	nv_icmd(priv, 0x0000095d, 0x00000001);
+	nv_icmd(priv, 0x0000082b, 0x00000004);
+	nv_icmd(priv, 0x00000942, 0x00010001);
+	nv_icmd(priv, 0x00000943, 0x00000001);
+	nv_icmd(priv, 0x00000944, 0x00000022);
+	nv_icmd(priv, 0x000007c5, 0x00010001);
+	nv_icmd(priv, 0x00000834, 0x00000001);
+	nv_icmd(priv, 0x000007c7, 0x00000001);
+	nv_icmd(priv, 0x0000c1b0, 0x0000000f);
+	nv_icmd(priv, 0x0000c1b1, 0x0000000f);
+	nv_icmd(priv, 0x0000c1b2, 0x0000000f);
+	nv_icmd(priv, 0x0000c1b3, 0x0000000f);
+	nv_icmd(priv, 0x0000c1b4, 0x0000000f);
+	nv_icmd(priv, 0x0000c1b5, 0x0000000f);
+	nv_icmd(priv, 0x0000c1b6, 0x0000000f);
+	nv_icmd(priv, 0x0000c1b7, 0x0000000f);
+	nv_icmd(priv, 0x0000c1b8, 0x0fac6881);
+	nv_icmd(priv, 0x0000c1b9, 0x00fac688);
+	nv_icmd(priv, 0x0001e100, 0x00000001);
+	nv_icmd(priv, 0x00001000, 0x00000002);
+	nv_icmd(priv, 0x000006aa, 0x00000001);
+	nv_icmd(priv, 0x000006ad, 0x00000100);
+	nv_icmd(priv, 0x000006ae, 0x00000100);
+	nv_icmd(priv, 0x000006b1, 0x00000011);
+	nv_icmd(priv, 0x0000078c, 0x00000008);
+	nv_icmd(priv, 0x00000792, 0x00000001);
+	nv_icmd(priv, 0x00000794, 0x00000001);
+	nv_icmd(priv, 0x00000795, 0x00000001);
+	nv_icmd(priv, 0x00000796, 0x00000001);
+	nv_icmd(priv, 0x00000797, 0x000000cf);
+	nv_icmd(priv, 0x0000079a, 0x00000002);
+	nv_icmd(priv, 0x00000833, 0x04444480);
+	nv_icmd(priv, 0x000007a1, 0x00000001);
+	nv_icmd(priv, 0x000007a3, 0x00000001);
+	nv_icmd(priv, 0x000007a4, 0x00000001);
+	nv_icmd(priv, 0x000007a5, 0x00000001);
+	nv_icmd(priv, 0x00000831, 0x00000004);
+	nv_icmd(priv, 0x0001e100, 0x00000001);
+	nv_icmd(priv, 0x00001000, 0x00000014);
+	nv_icmd(priv, 0x00000351, 0x00000100);
+	nv_icmd(priv, 0x00000957, 0x00000003);
+	nv_icmd(priv, 0x0000095d, 0x00000001);
+	nv_icmd(priv, 0x0000082b, 0x00000004);
+	nv_icmd(priv, 0x00000942, 0x00010001);
+	nv_icmd(priv, 0x00000943, 0x00000001);
+	nv_icmd(priv, 0x000007c5, 0x00010001);
+	nv_icmd(priv, 0x00000834, 0x00000001);
+	nv_icmd(priv, 0x000007c7, 0x00000001);
+	nv_icmd(priv, 0x0001e100, 0x00000001);
+	nv_icmd(priv, 0x00001000, 0x00000001);
+	nv_icmd(priv, 0x0000080c, 0x00000002);
+	nv_icmd(priv, 0x0000080d, 0x00000100);
+	nv_icmd(priv, 0x0000080e, 0x00000100);
+	nv_icmd(priv, 0x0000080f, 0x00000001);
+	nv_icmd(priv, 0x00000823, 0x00000002);
+	nv_icmd(priv, 0x00000824, 0x00000100);
+	nv_icmd(priv, 0x00000825, 0x00000100);
+	nv_icmd(priv, 0x00000826, 0x00000001);
+	nv_icmd(priv, 0x0001e100, 0x00000001);
+	nv_wr32(priv, 0x400208, 0x00000000);
+	nv_wr32(priv, 0x404154, 0x00000400);
+
+	nvc0_grctx_generate_9097(priv);
+	if (fermi >= 0x9197)
+		nvc0_grctx_generate_9197(priv);
+	if (fermi >= 0x9297)
+		nvc0_grctx_generate_9297(priv);
+	nvc0_grctx_generate_902d(priv);
+	nvc0_grctx_generate_9039(priv);
+	nvc0_grctx_generate_90c0(priv);
+
+	nv_wr32(priv, 0x000260, r000260);
+
+	return nvc0_grctx_fini(&info);
+}
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/ctxnve0.c b/drivers/gpu/drm/nouveau/core/engine/graph/ctxnve0.c
new file mode 100644
index 0000000..6d8c639
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/ctxnve0.c
@@ -0,0 +1,2788 @@
+/*
+ * Copyright 2010 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include "nvc0.h"
+
+static void
+nve0_grctx_generate_icmd(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x400208, 0x80000000);
+	nv_icmd(priv, 0x001000, 0x00000004);
+	nv_icmd(priv, 0x000039, 0x00000000);
+	nv_icmd(priv, 0x00003a, 0x00000000);
+	nv_icmd(priv, 0x00003b, 0x00000000);
+	nv_icmd(priv, 0x0000a9, 0x0000ffff);
+	nv_icmd(priv, 0x000038, 0x0fac6881);
+	nv_icmd(priv, 0x00003d, 0x00000001);
+	nv_icmd(priv, 0x0000e8, 0x00000400);
+	nv_icmd(priv, 0x0000e9, 0x00000400);
+	nv_icmd(priv, 0x0000ea, 0x00000400);
+	nv_icmd(priv, 0x0000eb, 0x00000400);
+	nv_icmd(priv, 0x0000ec, 0x00000400);
+	nv_icmd(priv, 0x0000ed, 0x00000400);
+	nv_icmd(priv, 0x0000ee, 0x00000400);
+	nv_icmd(priv, 0x0000ef, 0x00000400);
+	nv_icmd(priv, 0x000078, 0x00000300);
+	nv_icmd(priv, 0x000079, 0x00000300);
+	nv_icmd(priv, 0x00007a, 0x00000300);
+	nv_icmd(priv, 0x00007b, 0x00000300);
+	nv_icmd(priv, 0x00007c, 0x00000300);
+	nv_icmd(priv, 0x00007d, 0x00000300);
+	nv_icmd(priv, 0x00007e, 0x00000300);
+	nv_icmd(priv, 0x00007f, 0x00000300);
+	nv_icmd(priv, 0x000050, 0x00000011);
+	nv_icmd(priv, 0x000058, 0x00000008);
+	nv_icmd(priv, 0x000059, 0x00000008);
+	nv_icmd(priv, 0x00005a, 0x00000008);
+	nv_icmd(priv, 0x00005b, 0x00000008);
+	nv_icmd(priv, 0x00005c, 0x00000008);
+	nv_icmd(priv, 0x00005d, 0x00000008);
+	nv_icmd(priv, 0x00005e, 0x00000008);
+	nv_icmd(priv, 0x00005f, 0x00000008);
+	nv_icmd(priv, 0x000208, 0x00000001);
+	nv_icmd(priv, 0x000209, 0x00000001);
+	nv_icmd(priv, 0x00020a, 0x00000001);
+	nv_icmd(priv, 0x00020b, 0x00000001);
+	nv_icmd(priv, 0x00020c, 0x00000001);
+	nv_icmd(priv, 0x00020d, 0x00000001);
+	nv_icmd(priv, 0x00020e, 0x00000001);
+	nv_icmd(priv, 0x00020f, 0x00000001);
+	nv_icmd(priv, 0x000081, 0x00000001);
+	nv_icmd(priv, 0x000085, 0x00000004);
+	nv_icmd(priv, 0x000088, 0x00000400);
+	nv_icmd(priv, 0x000090, 0x00000300);
+	nv_icmd(priv, 0x000098, 0x00001001);
+	nv_icmd(priv, 0x0000e3, 0x00000001);
+	nv_icmd(priv, 0x0000da, 0x00000001);
+	nv_icmd(priv, 0x0000f8, 0x00000003);
+	nv_icmd(priv, 0x0000fa, 0x00000001);
+	nv_icmd(priv, 0x00009f, 0x0000ffff);
+	nv_icmd(priv, 0x0000a0, 0x0000ffff);
+	nv_icmd(priv, 0x0000a1, 0x0000ffff);
+	nv_icmd(priv, 0x0000a2, 0x0000ffff);
+	nv_icmd(priv, 0x0000b1, 0x00000001);
+	nv_icmd(priv, 0x0000ad, 0x0000013e);
+	nv_icmd(priv, 0x0000e1, 0x00000010);
+	nv_icmd(priv, 0x000290, 0x00000000);
+	nv_icmd(priv, 0x000291, 0x00000000);
+	nv_icmd(priv, 0x000292, 0x00000000);
+	nv_icmd(priv, 0x000293, 0x00000000);
+	nv_icmd(priv, 0x000294, 0x00000000);
+	nv_icmd(priv, 0x000295, 0x00000000);
+	nv_icmd(priv, 0x000296, 0x00000000);
+	nv_icmd(priv, 0x000297, 0x00000000);
+	nv_icmd(priv, 0x000298, 0x00000000);
+	nv_icmd(priv, 0x000299, 0x00000000);
+	nv_icmd(priv, 0x00029a, 0x00000000);
+	nv_icmd(priv, 0x00029b, 0x00000000);
+	nv_icmd(priv, 0x00029c, 0x00000000);
+	nv_icmd(priv, 0x00029d, 0x00000000);
+	nv_icmd(priv, 0x00029e, 0x00000000);
+	nv_icmd(priv, 0x00029f, 0x00000000);
+	nv_icmd(priv, 0x0003b0, 0x00000000);
+	nv_icmd(priv, 0x0003b1, 0x00000000);
+	nv_icmd(priv, 0x0003b2, 0x00000000);
+	nv_icmd(priv, 0x0003b3, 0x00000000);
+	nv_icmd(priv, 0x0003b4, 0x00000000);
+	nv_icmd(priv, 0x0003b5, 0x00000000);
+	nv_icmd(priv, 0x0003b6, 0x00000000);
+	nv_icmd(priv, 0x0003b7, 0x00000000);
+	nv_icmd(priv, 0x0003b8, 0x00000000);
+	nv_icmd(priv, 0x0003b9, 0x00000000);
+	nv_icmd(priv, 0x0003ba, 0x00000000);
+	nv_icmd(priv, 0x0003bb, 0x00000000);
+	nv_icmd(priv, 0x0003bc, 0x00000000);
+	nv_icmd(priv, 0x0003bd, 0x00000000);
+	nv_icmd(priv, 0x0003be, 0x00000000);
+	nv_icmd(priv, 0x0003bf, 0x00000000);
+	nv_icmd(priv, 0x0002a0, 0x00000000);
+	nv_icmd(priv, 0x0002a1, 0x00000000);
+	nv_icmd(priv, 0x0002a2, 0x00000000);
+	nv_icmd(priv, 0x0002a3, 0x00000000);
+	nv_icmd(priv, 0x0002a4, 0x00000000);
+	nv_icmd(priv, 0x0002a5, 0x00000000);
+	nv_icmd(priv, 0x0002a6, 0x00000000);
+	nv_icmd(priv, 0x0002a7, 0x00000000);
+	nv_icmd(priv, 0x0002a8, 0x00000000);
+	nv_icmd(priv, 0x0002a9, 0x00000000);
+	nv_icmd(priv, 0x0002aa, 0x00000000);
+	nv_icmd(priv, 0x0002ab, 0x00000000);
+	nv_icmd(priv, 0x0002ac, 0x00000000);
+	nv_icmd(priv, 0x0002ad, 0x00000000);
+	nv_icmd(priv, 0x0002ae, 0x00000000);
+	nv_icmd(priv, 0x0002af, 0x00000000);
+	nv_icmd(priv, 0x000420, 0x00000000);
+	nv_icmd(priv, 0x000421, 0x00000000);
+	nv_icmd(priv, 0x000422, 0x00000000);
+	nv_icmd(priv, 0x000423, 0x00000000);
+	nv_icmd(priv, 0x000424, 0x00000000);
+	nv_icmd(priv, 0x000425, 0x00000000);
+	nv_icmd(priv, 0x000426, 0x00000000);
+	nv_icmd(priv, 0x000427, 0x00000000);
+	nv_icmd(priv, 0x000428, 0x00000000);
+	nv_icmd(priv, 0x000429, 0x00000000);
+	nv_icmd(priv, 0x00042a, 0x00000000);
+	nv_icmd(priv, 0x00042b, 0x00000000);
+	nv_icmd(priv, 0x00042c, 0x00000000);
+	nv_icmd(priv, 0x00042d, 0x00000000);
+	nv_icmd(priv, 0x00042e, 0x00000000);
+	nv_icmd(priv, 0x00042f, 0x00000000);
+	nv_icmd(priv, 0x0002b0, 0x00000000);
+	nv_icmd(priv, 0x0002b1, 0x00000000);
+	nv_icmd(priv, 0x0002b2, 0x00000000);
+	nv_icmd(priv, 0x0002b3, 0x00000000);
+	nv_icmd(priv, 0x0002b4, 0x00000000);
+	nv_icmd(priv, 0x0002b5, 0x00000000);
+	nv_icmd(priv, 0x0002b6, 0x00000000);
+	nv_icmd(priv, 0x0002b7, 0x00000000);
+	nv_icmd(priv, 0x0002b8, 0x00000000);
+	nv_icmd(priv, 0x0002b9, 0x00000000);
+	nv_icmd(priv, 0x0002ba, 0x00000000);
+	nv_icmd(priv, 0x0002bb, 0x00000000);
+	nv_icmd(priv, 0x0002bc, 0x00000000);
+	nv_icmd(priv, 0x0002bd, 0x00000000);
+	nv_icmd(priv, 0x0002be, 0x00000000);
+	nv_icmd(priv, 0x0002bf, 0x00000000);
+	nv_icmd(priv, 0x000430, 0x00000000);
+	nv_icmd(priv, 0x000431, 0x00000000);
+	nv_icmd(priv, 0x000432, 0x00000000);
+	nv_icmd(priv, 0x000433, 0x00000000);
+	nv_icmd(priv, 0x000434, 0x00000000);
+	nv_icmd(priv, 0x000435, 0x00000000);
+	nv_icmd(priv, 0x000436, 0x00000000);
+	nv_icmd(priv, 0x000437, 0x00000000);
+	nv_icmd(priv, 0x000438, 0x00000000);
+	nv_icmd(priv, 0x000439, 0x00000000);
+	nv_icmd(priv, 0x00043a, 0x00000000);
+	nv_icmd(priv, 0x00043b, 0x00000000);
+	nv_icmd(priv, 0x00043c, 0x00000000);
+	nv_icmd(priv, 0x00043d, 0x00000000);
+	nv_icmd(priv, 0x00043e, 0x00000000);
+	nv_icmd(priv, 0x00043f, 0x00000000);
+	nv_icmd(priv, 0x0002c0, 0x00000000);
+	nv_icmd(priv, 0x0002c1, 0x00000000);
+	nv_icmd(priv, 0x0002c2, 0x00000000);
+	nv_icmd(priv, 0x0002c3, 0x00000000);
+	nv_icmd(priv, 0x0002c4, 0x00000000);
+	nv_icmd(priv, 0x0002c5, 0x00000000);
+	nv_icmd(priv, 0x0002c6, 0x00000000);
+	nv_icmd(priv, 0x0002c7, 0x00000000);
+	nv_icmd(priv, 0x0002c8, 0x00000000);
+	nv_icmd(priv, 0x0002c9, 0x00000000);
+	nv_icmd(priv, 0x0002ca, 0x00000000);
+	nv_icmd(priv, 0x0002cb, 0x00000000);
+	nv_icmd(priv, 0x0002cc, 0x00000000);
+	nv_icmd(priv, 0x0002cd, 0x00000000);
+	nv_icmd(priv, 0x0002ce, 0x00000000);
+	nv_icmd(priv, 0x0002cf, 0x00000000);
+	nv_icmd(priv, 0x0004d0, 0x00000000);
+	nv_icmd(priv, 0x0004d1, 0x00000000);
+	nv_icmd(priv, 0x0004d2, 0x00000000);
+	nv_icmd(priv, 0x0004d3, 0x00000000);
+	nv_icmd(priv, 0x0004d4, 0x00000000);
+	nv_icmd(priv, 0x0004d5, 0x00000000);
+	nv_icmd(priv, 0x0004d6, 0x00000000);
+	nv_icmd(priv, 0x0004d7, 0x00000000);
+	nv_icmd(priv, 0x0004d8, 0x00000000);
+	nv_icmd(priv, 0x0004d9, 0x00000000);
+	nv_icmd(priv, 0x0004da, 0x00000000);
+	nv_icmd(priv, 0x0004db, 0x00000000);
+	nv_icmd(priv, 0x0004dc, 0x00000000);
+	nv_icmd(priv, 0x0004dd, 0x00000000);
+	nv_icmd(priv, 0x0004de, 0x00000000);
+	nv_icmd(priv, 0x0004df, 0x00000000);
+	nv_icmd(priv, 0x000720, 0x00000000);
+	nv_icmd(priv, 0x000721, 0x00000000);
+	nv_icmd(priv, 0x000722, 0x00000000);
+	nv_icmd(priv, 0x000723, 0x00000000);
+	nv_icmd(priv, 0x000724, 0x00000000);
+	nv_icmd(priv, 0x000725, 0x00000000);
+	nv_icmd(priv, 0x000726, 0x00000000);
+	nv_icmd(priv, 0x000727, 0x00000000);
+	nv_icmd(priv, 0x000728, 0x00000000);
+	nv_icmd(priv, 0x000729, 0x00000000);
+	nv_icmd(priv, 0x00072a, 0x00000000);
+	nv_icmd(priv, 0x00072b, 0x00000000);
+	nv_icmd(priv, 0x00072c, 0x00000000);
+	nv_icmd(priv, 0x00072d, 0x00000000);
+	nv_icmd(priv, 0x00072e, 0x00000000);
+	nv_icmd(priv, 0x00072f, 0x00000000);
+	nv_icmd(priv, 0x0008c0, 0x00000000);
+	nv_icmd(priv, 0x0008c1, 0x00000000);
+	nv_icmd(priv, 0x0008c2, 0x00000000);
+	nv_icmd(priv, 0x0008c3, 0x00000000);
+	nv_icmd(priv, 0x0008c4, 0x00000000);
+	nv_icmd(priv, 0x0008c5, 0x00000000);
+	nv_icmd(priv, 0x0008c6, 0x00000000);
+	nv_icmd(priv, 0x0008c7, 0x00000000);
+	nv_icmd(priv, 0x0008c8, 0x00000000);
+	nv_icmd(priv, 0x0008c9, 0x00000000);
+	nv_icmd(priv, 0x0008ca, 0x00000000);
+	nv_icmd(priv, 0x0008cb, 0x00000000);
+	nv_icmd(priv, 0x0008cc, 0x00000000);
+	nv_icmd(priv, 0x0008cd, 0x00000000);
+	nv_icmd(priv, 0x0008ce, 0x00000000);
+	nv_icmd(priv, 0x0008cf, 0x00000000);
+	nv_icmd(priv, 0x000890, 0x00000000);
+	nv_icmd(priv, 0x000891, 0x00000000);
+	nv_icmd(priv, 0x000892, 0x00000000);
+	nv_icmd(priv, 0x000893, 0x00000000);
+	nv_icmd(priv, 0x000894, 0x00000000);
+	nv_icmd(priv, 0x000895, 0x00000000);
+	nv_icmd(priv, 0x000896, 0x00000000);
+	nv_icmd(priv, 0x000897, 0x00000000);
+	nv_icmd(priv, 0x000898, 0x00000000);
+	nv_icmd(priv, 0x000899, 0x00000000);
+	nv_icmd(priv, 0x00089a, 0x00000000);
+	nv_icmd(priv, 0x00089b, 0x00000000);
+	nv_icmd(priv, 0x00089c, 0x00000000);
+	nv_icmd(priv, 0x00089d, 0x00000000);
+	nv_icmd(priv, 0x00089e, 0x00000000);
+	nv_icmd(priv, 0x00089f, 0x00000000);
+	nv_icmd(priv, 0x0008e0, 0x00000000);
+	nv_icmd(priv, 0x0008e1, 0x00000000);
+	nv_icmd(priv, 0x0008e2, 0x00000000);
+	nv_icmd(priv, 0x0008e3, 0x00000000);
+	nv_icmd(priv, 0x0008e4, 0x00000000);
+	nv_icmd(priv, 0x0008e5, 0x00000000);
+	nv_icmd(priv, 0x0008e6, 0x00000000);
+	nv_icmd(priv, 0x0008e7, 0x00000000);
+	nv_icmd(priv, 0x0008e8, 0x00000000);
+	nv_icmd(priv, 0x0008e9, 0x00000000);
+	nv_icmd(priv, 0x0008ea, 0x00000000);
+	nv_icmd(priv, 0x0008eb, 0x00000000);
+	nv_icmd(priv, 0x0008ec, 0x00000000);
+	nv_icmd(priv, 0x0008ed, 0x00000000);
+	nv_icmd(priv, 0x0008ee, 0x00000000);
+	nv_icmd(priv, 0x0008ef, 0x00000000);
+	nv_icmd(priv, 0x0008a0, 0x00000000);
+	nv_icmd(priv, 0x0008a1, 0x00000000);
+	nv_icmd(priv, 0x0008a2, 0x00000000);
+	nv_icmd(priv, 0x0008a3, 0x00000000);
+	nv_icmd(priv, 0x0008a4, 0x00000000);
+	nv_icmd(priv, 0x0008a5, 0x00000000);
+	nv_icmd(priv, 0x0008a6, 0x00000000);
+	nv_icmd(priv, 0x0008a7, 0x00000000);
+	nv_icmd(priv, 0x0008a8, 0x00000000);
+	nv_icmd(priv, 0x0008a9, 0x00000000);
+	nv_icmd(priv, 0x0008aa, 0x00000000);
+	nv_icmd(priv, 0x0008ab, 0x00000000);
+	nv_icmd(priv, 0x0008ac, 0x00000000);
+	nv_icmd(priv, 0x0008ad, 0x00000000);
+	nv_icmd(priv, 0x0008ae, 0x00000000);
+	nv_icmd(priv, 0x0008af, 0x00000000);
+	nv_icmd(priv, 0x0008f0, 0x00000000);
+	nv_icmd(priv, 0x0008f1, 0x00000000);
+	nv_icmd(priv, 0x0008f2, 0x00000000);
+	nv_icmd(priv, 0x0008f3, 0x00000000);
+	nv_icmd(priv, 0x0008f4, 0x00000000);
+	nv_icmd(priv, 0x0008f5, 0x00000000);
+	nv_icmd(priv, 0x0008f6, 0x00000000);
+	nv_icmd(priv, 0x0008f7, 0x00000000);
+	nv_icmd(priv, 0x0008f8, 0x00000000);
+	nv_icmd(priv, 0x0008f9, 0x00000000);
+	nv_icmd(priv, 0x0008fa, 0x00000000);
+	nv_icmd(priv, 0x0008fb, 0x00000000);
+	nv_icmd(priv, 0x0008fc, 0x00000000);
+	nv_icmd(priv, 0x0008fd, 0x00000000);
+	nv_icmd(priv, 0x0008fe, 0x00000000);
+	nv_icmd(priv, 0x0008ff, 0x00000000);
+	nv_icmd(priv, 0x00094c, 0x000000ff);
+	nv_icmd(priv, 0x00094d, 0xffffffff);
+	nv_icmd(priv, 0x00094e, 0x00000002);
+	nv_icmd(priv, 0x0002ec, 0x00000001);
+	nv_icmd(priv, 0x000303, 0x00000001);
+	nv_icmd(priv, 0x0002e6, 0x00000001);
+	nv_icmd(priv, 0x000466, 0x00000052);
+	nv_icmd(priv, 0x000301, 0x3f800000);
+	nv_icmd(priv, 0x000304, 0x30201000);
+	nv_icmd(priv, 0x000305, 0x70605040);
+	nv_icmd(priv, 0x000306, 0xb8a89888);
+	nv_icmd(priv, 0x000307, 0xf8e8d8c8);
+	nv_icmd(priv, 0x00030a, 0x00ffff00);
+	nv_icmd(priv, 0x00030b, 0x0000001a);
+	nv_icmd(priv, 0x00030c, 0x00000001);
+	nv_icmd(priv, 0x000318, 0x00000001);
+	nv_icmd(priv, 0x000340, 0x00000000);
+	nv_icmd(priv, 0x000375, 0x00000001);
+	nv_icmd(priv, 0x00037d, 0x00000006);
+	nv_icmd(priv, 0x0003a0, 0x00000002);
+	nv_icmd(priv, 0x0003aa, 0x00000001);
+	nv_icmd(priv, 0x0003a9, 0x00000001);
+	nv_icmd(priv, 0x000380, 0x00000001);
+	nv_icmd(priv, 0x000383, 0x00000011);
+	nv_icmd(priv, 0x000360, 0x00000040);
+	nv_icmd(priv, 0x000366, 0x00000000);
+	nv_icmd(priv, 0x000367, 0x00000000);
+	nv_icmd(priv, 0x000368, 0x00000fff);
+	nv_icmd(priv, 0x000370, 0x00000000);
+	nv_icmd(priv, 0x000371, 0x00000000);
+	nv_icmd(priv, 0x000372, 0x000fffff);
+	nv_icmd(priv, 0x00037a, 0x00000012);
+	nv_icmd(priv, 0x000619, 0x00000003);
+	nv_icmd(priv, 0x000811, 0x00000003);
+	nv_icmd(priv, 0x000812, 0x00000004);
+	nv_icmd(priv, 0x000813, 0x00000006);
+	nv_icmd(priv, 0x000814, 0x00000008);
+	nv_icmd(priv, 0x000815, 0x0000000b);
+	nv_icmd(priv, 0x000800, 0x00000001);
+	nv_icmd(priv, 0x000801, 0x00000001);
+	nv_icmd(priv, 0x000802, 0x00000001);
+	nv_icmd(priv, 0x000803, 0x00000001);
+	nv_icmd(priv, 0x000804, 0x00000001);
+	nv_icmd(priv, 0x000805, 0x00000001);
+	nv_icmd(priv, 0x000632, 0x00000001);
+	nv_icmd(priv, 0x000633, 0x00000002);
+	nv_icmd(priv, 0x000634, 0x00000003);
+	nv_icmd(priv, 0x000635, 0x00000004);
+	nv_icmd(priv, 0x000654, 0x3f800000);
+	nv_icmd(priv, 0x000657, 0x3f800000);
+	nv_icmd(priv, 0x000655, 0x3f800000);
+	nv_icmd(priv, 0x000656, 0x3f800000);
+	nv_icmd(priv, 0x0006cd, 0x3f800000);
+	nv_icmd(priv, 0x0007f5, 0x3f800000);
+	nv_icmd(priv, 0x0007dc, 0x39291909);
+	nv_icmd(priv, 0x0007dd, 0x79695949);
+	nv_icmd(priv, 0x0007de, 0xb9a99989);
+	nv_icmd(priv, 0x0007df, 0xf9e9d9c9);
+	nv_icmd(priv, 0x0007e8, 0x00003210);
+	nv_icmd(priv, 0x0007e9, 0x00007654);
+	nv_icmd(priv, 0x0007ea, 0x00000098);
+	nv_icmd(priv, 0x0007ec, 0x39291909);
+	nv_icmd(priv, 0x0007ed, 0x79695949);
+	nv_icmd(priv, 0x0007ee, 0xb9a99989);
+	nv_icmd(priv, 0x0007ef, 0xf9e9d9c9);
+	nv_icmd(priv, 0x0007f0, 0x00003210);
+	nv_icmd(priv, 0x0007f1, 0x00007654);
+	nv_icmd(priv, 0x0007f2, 0x00000098);
+	nv_icmd(priv, 0x0005a5, 0x00000001);
+	nv_icmd(priv, 0x000980, 0x00000000);
+	nv_icmd(priv, 0x000981, 0x00000000);
+	nv_icmd(priv, 0x000982, 0x00000000);
+	nv_icmd(priv, 0x000983, 0x00000000);
+	nv_icmd(priv, 0x000984, 0x00000000);
+	nv_icmd(priv, 0x000985, 0x00000000);
+	nv_icmd(priv, 0x000986, 0x00000000);
+	nv_icmd(priv, 0x000987, 0x00000000);
+	nv_icmd(priv, 0x000988, 0x00000000);
+	nv_icmd(priv, 0x000989, 0x00000000);
+	nv_icmd(priv, 0x00098a, 0x00000000);
+	nv_icmd(priv, 0x00098b, 0x00000000);
+	nv_icmd(priv, 0x00098c, 0x00000000);
+	nv_icmd(priv, 0x00098d, 0x00000000);
+	nv_icmd(priv, 0x00098e, 0x00000000);
+	nv_icmd(priv, 0x00098f, 0x00000000);
+	nv_icmd(priv, 0x000990, 0x00000000);
+	nv_icmd(priv, 0x000991, 0x00000000);
+	nv_icmd(priv, 0x000992, 0x00000000);
+	nv_icmd(priv, 0x000993, 0x00000000);
+	nv_icmd(priv, 0x000994, 0x00000000);
+	nv_icmd(priv, 0x000995, 0x00000000);
+	nv_icmd(priv, 0x000996, 0x00000000);
+	nv_icmd(priv, 0x000997, 0x00000000);
+	nv_icmd(priv, 0x000998, 0x00000000);
+	nv_icmd(priv, 0x000999, 0x00000000);
+	nv_icmd(priv, 0x00099a, 0x00000000);
+	nv_icmd(priv, 0x00099b, 0x00000000);
+	nv_icmd(priv, 0x00099c, 0x00000000);
+	nv_icmd(priv, 0x00099d, 0x00000000);
+	nv_icmd(priv, 0x00099e, 0x00000000);
+	nv_icmd(priv, 0x00099f, 0x00000000);
+	nv_icmd(priv, 0x0009a0, 0x00000000);
+	nv_icmd(priv, 0x0009a1, 0x00000000);
+	nv_icmd(priv, 0x0009a2, 0x00000000);
+	nv_icmd(priv, 0x0009a3, 0x00000000);
+	nv_icmd(priv, 0x0009a4, 0x00000000);
+	nv_icmd(priv, 0x0009a5, 0x00000000);
+	nv_icmd(priv, 0x0009a6, 0x00000000);
+	nv_icmd(priv, 0x0009a7, 0x00000000);
+	nv_icmd(priv, 0x0009a8, 0x00000000);
+	nv_icmd(priv, 0x0009a9, 0x00000000);
+	nv_icmd(priv, 0x0009aa, 0x00000000);
+	nv_icmd(priv, 0x0009ab, 0x00000000);
+	nv_icmd(priv, 0x0009ac, 0x00000000);
+	nv_icmd(priv, 0x0009ad, 0x00000000);
+	nv_icmd(priv, 0x0009ae, 0x00000000);
+	nv_icmd(priv, 0x0009af, 0x00000000);
+	nv_icmd(priv, 0x0009b0, 0x00000000);
+	nv_icmd(priv, 0x0009b1, 0x00000000);
+	nv_icmd(priv, 0x0009b2, 0x00000000);
+	nv_icmd(priv, 0x0009b3, 0x00000000);
+	nv_icmd(priv, 0x0009b4, 0x00000000);
+	nv_icmd(priv, 0x0009b5, 0x00000000);
+	nv_icmd(priv, 0x0009b6, 0x00000000);
+	nv_icmd(priv, 0x0009b7, 0x00000000);
+	nv_icmd(priv, 0x0009b8, 0x00000000);
+	nv_icmd(priv, 0x0009b9, 0x00000000);
+	nv_icmd(priv, 0x0009ba, 0x00000000);
+	nv_icmd(priv, 0x0009bb, 0x00000000);
+	nv_icmd(priv, 0x0009bc, 0x00000000);
+	nv_icmd(priv, 0x0009bd, 0x00000000);
+	nv_icmd(priv, 0x0009be, 0x00000000);
+	nv_icmd(priv, 0x0009bf, 0x00000000);
+	nv_icmd(priv, 0x0009c0, 0x00000000);
+	nv_icmd(priv, 0x0009c1, 0x00000000);
+	nv_icmd(priv, 0x0009c2, 0x00000000);
+	nv_icmd(priv, 0x0009c3, 0x00000000);
+	nv_icmd(priv, 0x0009c4, 0x00000000);
+	nv_icmd(priv, 0x0009c5, 0x00000000);
+	nv_icmd(priv, 0x0009c6, 0x00000000);
+	nv_icmd(priv, 0x0009c7, 0x00000000);
+	nv_icmd(priv, 0x0009c8, 0x00000000);
+	nv_icmd(priv, 0x0009c9, 0x00000000);
+	nv_icmd(priv, 0x0009ca, 0x00000000);
+	nv_icmd(priv, 0x0009cb, 0x00000000);
+	nv_icmd(priv, 0x0009cc, 0x00000000);
+	nv_icmd(priv, 0x0009cd, 0x00000000);
+	nv_icmd(priv, 0x0009ce, 0x00000000);
+	nv_icmd(priv, 0x0009cf, 0x00000000);
+	nv_icmd(priv, 0x0009d0, 0x00000000);
+	nv_icmd(priv, 0x0009d1, 0x00000000);
+	nv_icmd(priv, 0x0009d2, 0x00000000);
+	nv_icmd(priv, 0x0009d3, 0x00000000);
+	nv_icmd(priv, 0x0009d4, 0x00000000);
+	nv_icmd(priv, 0x0009d5, 0x00000000);
+	nv_icmd(priv, 0x0009d6, 0x00000000);
+	nv_icmd(priv, 0x0009d7, 0x00000000);
+	nv_icmd(priv, 0x0009d8, 0x00000000);
+	nv_icmd(priv, 0x0009d9, 0x00000000);
+	nv_icmd(priv, 0x0009da, 0x00000000);
+	nv_icmd(priv, 0x0009db, 0x00000000);
+	nv_icmd(priv, 0x0009dc, 0x00000000);
+	nv_icmd(priv, 0x0009dd, 0x00000000);
+	nv_icmd(priv, 0x0009de, 0x00000000);
+	nv_icmd(priv, 0x0009df, 0x00000000);
+	nv_icmd(priv, 0x0009e0, 0x00000000);
+	nv_icmd(priv, 0x0009e1, 0x00000000);
+	nv_icmd(priv, 0x0009e2, 0x00000000);
+	nv_icmd(priv, 0x0009e3, 0x00000000);
+	nv_icmd(priv, 0x0009e4, 0x00000000);
+	nv_icmd(priv, 0x0009e5, 0x00000000);
+	nv_icmd(priv, 0x0009e6, 0x00000000);
+	nv_icmd(priv, 0x0009e7, 0x00000000);
+	nv_icmd(priv, 0x0009e8, 0x00000000);
+	nv_icmd(priv, 0x0009e9, 0x00000000);
+	nv_icmd(priv, 0x0009ea, 0x00000000);
+	nv_icmd(priv, 0x0009eb, 0x00000000);
+	nv_icmd(priv, 0x0009ec, 0x00000000);
+	nv_icmd(priv, 0x0009ed, 0x00000000);
+	nv_icmd(priv, 0x0009ee, 0x00000000);
+	nv_icmd(priv, 0x0009ef, 0x00000000);
+	nv_icmd(priv, 0x0009f0, 0x00000000);
+	nv_icmd(priv, 0x0009f1, 0x00000000);
+	nv_icmd(priv, 0x0009f2, 0x00000000);
+	nv_icmd(priv, 0x0009f3, 0x00000000);
+	nv_icmd(priv, 0x0009f4, 0x00000000);
+	nv_icmd(priv, 0x0009f5, 0x00000000);
+	nv_icmd(priv, 0x0009f6, 0x00000000);
+	nv_icmd(priv, 0x0009f7, 0x00000000);
+	nv_icmd(priv, 0x0009f8, 0x00000000);
+	nv_icmd(priv, 0x0009f9, 0x00000000);
+	nv_icmd(priv, 0x0009fa, 0x00000000);
+	nv_icmd(priv, 0x0009fb, 0x00000000);
+	nv_icmd(priv, 0x0009fc, 0x00000000);
+	nv_icmd(priv, 0x0009fd, 0x00000000);
+	nv_icmd(priv, 0x0009fe, 0x00000000);
+	nv_icmd(priv, 0x0009ff, 0x00000000);
+	nv_icmd(priv, 0x000468, 0x00000004);
+	nv_icmd(priv, 0x00046c, 0x00000001);
+	nv_icmd(priv, 0x000470, 0x00000000);
+	nv_icmd(priv, 0x000471, 0x00000000);
+	nv_icmd(priv, 0x000472, 0x00000000);
+	nv_icmd(priv, 0x000473, 0x00000000);
+	nv_icmd(priv, 0x000474, 0x00000000);
+	nv_icmd(priv, 0x000475, 0x00000000);
+	nv_icmd(priv, 0x000476, 0x00000000);
+	nv_icmd(priv, 0x000477, 0x00000000);
+	nv_icmd(priv, 0x000478, 0x00000000);
+	nv_icmd(priv, 0x000479, 0x00000000);
+	nv_icmd(priv, 0x00047a, 0x00000000);
+	nv_icmd(priv, 0x00047b, 0x00000000);
+	nv_icmd(priv, 0x00047c, 0x00000000);
+	nv_icmd(priv, 0x00047d, 0x00000000);
+	nv_icmd(priv, 0x00047e, 0x00000000);
+	nv_icmd(priv, 0x00047f, 0x00000000);
+	nv_icmd(priv, 0x000480, 0x00000000);
+	nv_icmd(priv, 0x000481, 0x00000000);
+	nv_icmd(priv, 0x000482, 0x00000000);
+	nv_icmd(priv, 0x000483, 0x00000000);
+	nv_icmd(priv, 0x000484, 0x00000000);
+	nv_icmd(priv, 0x000485, 0x00000000);
+	nv_icmd(priv, 0x000486, 0x00000000);
+	nv_icmd(priv, 0x000487, 0x00000000);
+	nv_icmd(priv, 0x000488, 0x00000000);
+	nv_icmd(priv, 0x000489, 0x00000000);
+	nv_icmd(priv, 0x00048a, 0x00000000);
+	nv_icmd(priv, 0x00048b, 0x00000000);
+	nv_icmd(priv, 0x00048c, 0x00000000);
+	nv_icmd(priv, 0x00048d, 0x00000000);
+	nv_icmd(priv, 0x00048e, 0x00000000);
+	nv_icmd(priv, 0x00048f, 0x00000000);
+	nv_icmd(priv, 0x000490, 0x00000000);
+	nv_icmd(priv, 0x000491, 0x00000000);
+	nv_icmd(priv, 0x000492, 0x00000000);
+	nv_icmd(priv, 0x000493, 0x00000000);
+	nv_icmd(priv, 0x000494, 0x00000000);
+	nv_icmd(priv, 0x000495, 0x00000000);
+	nv_icmd(priv, 0x000496, 0x00000000);
+	nv_icmd(priv, 0x000497, 0x00000000);
+	nv_icmd(priv, 0x000498, 0x00000000);
+	nv_icmd(priv, 0x000499, 0x00000000);
+	nv_icmd(priv, 0x00049a, 0x00000000);
+	nv_icmd(priv, 0x00049b, 0x00000000);
+	nv_icmd(priv, 0x00049c, 0x00000000);
+	nv_icmd(priv, 0x00049d, 0x00000000);
+	nv_icmd(priv, 0x00049e, 0x00000000);
+	nv_icmd(priv, 0x00049f, 0x00000000);
+	nv_icmd(priv, 0x0004a0, 0x00000000);
+	nv_icmd(priv, 0x0004a1, 0x00000000);
+	nv_icmd(priv, 0x0004a2, 0x00000000);
+	nv_icmd(priv, 0x0004a3, 0x00000000);
+	nv_icmd(priv, 0x0004a4, 0x00000000);
+	nv_icmd(priv, 0x0004a5, 0x00000000);
+	nv_icmd(priv, 0x0004a6, 0x00000000);
+	nv_icmd(priv, 0x0004a7, 0x00000000);
+	nv_icmd(priv, 0x0004a8, 0x00000000);
+	nv_icmd(priv, 0x0004a9, 0x00000000);
+	nv_icmd(priv, 0x0004aa, 0x00000000);
+	nv_icmd(priv, 0x0004ab, 0x00000000);
+	nv_icmd(priv, 0x0004ac, 0x00000000);
+	nv_icmd(priv, 0x0004ad, 0x00000000);
+	nv_icmd(priv, 0x0004ae, 0x00000000);
+	nv_icmd(priv, 0x0004af, 0x00000000);
+	nv_icmd(priv, 0x0004b0, 0x00000000);
+	nv_icmd(priv, 0x0004b1, 0x00000000);
+	nv_icmd(priv, 0x0004b2, 0x00000000);
+	nv_icmd(priv, 0x0004b3, 0x00000000);
+	nv_icmd(priv, 0x0004b4, 0x00000000);
+	nv_icmd(priv, 0x0004b5, 0x00000000);
+	nv_icmd(priv, 0x0004b6, 0x00000000);
+	nv_icmd(priv, 0x0004b7, 0x00000000);
+	nv_icmd(priv, 0x0004b8, 0x00000000);
+	nv_icmd(priv, 0x0004b9, 0x00000000);
+	nv_icmd(priv, 0x0004ba, 0x00000000);
+	nv_icmd(priv, 0x0004bb, 0x00000000);
+	nv_icmd(priv, 0x0004bc, 0x00000000);
+	nv_icmd(priv, 0x0004bd, 0x00000000);
+	nv_icmd(priv, 0x0004be, 0x00000000);
+	nv_icmd(priv, 0x0004bf, 0x00000000);
+	nv_icmd(priv, 0x0004c0, 0x00000000);
+	nv_icmd(priv, 0x0004c1, 0x00000000);
+	nv_icmd(priv, 0x0004c2, 0x00000000);
+	nv_icmd(priv, 0x0004c3, 0x00000000);
+	nv_icmd(priv, 0x0004c4, 0x00000000);
+	nv_icmd(priv, 0x0004c5, 0x00000000);
+	nv_icmd(priv, 0x0004c6, 0x00000000);
+	nv_icmd(priv, 0x0004c7, 0x00000000);
+	nv_icmd(priv, 0x0004c8, 0x00000000);
+	nv_icmd(priv, 0x0004c9, 0x00000000);
+	nv_icmd(priv, 0x0004ca, 0x00000000);
+	nv_icmd(priv, 0x0004cb, 0x00000000);
+	nv_icmd(priv, 0x0004cc, 0x00000000);
+	nv_icmd(priv, 0x0004cd, 0x00000000);
+	nv_icmd(priv, 0x0004ce, 0x00000000);
+	nv_icmd(priv, 0x0004cf, 0x00000000);
+	nv_icmd(priv, 0x000510, 0x3f800000);
+	nv_icmd(priv, 0x000511, 0x3f800000);
+	nv_icmd(priv, 0x000512, 0x3f800000);
+	nv_icmd(priv, 0x000513, 0x3f800000);
+	nv_icmd(priv, 0x000514, 0x3f800000);
+	nv_icmd(priv, 0x000515, 0x3f800000);
+	nv_icmd(priv, 0x000516, 0x3f800000);
+	nv_icmd(priv, 0x000517, 0x3f800000);
+	nv_icmd(priv, 0x000518, 0x3f800000);
+	nv_icmd(priv, 0x000519, 0x3f800000);
+	nv_icmd(priv, 0x00051a, 0x3f800000);
+	nv_icmd(priv, 0x00051b, 0x3f800000);
+	nv_icmd(priv, 0x00051c, 0x3f800000);
+	nv_icmd(priv, 0x00051d, 0x3f800000);
+	nv_icmd(priv, 0x00051e, 0x3f800000);
+	nv_icmd(priv, 0x00051f, 0x3f800000);
+	nv_icmd(priv, 0x000520, 0x000002b6);
+	nv_icmd(priv, 0x000529, 0x00000001);
+	nv_icmd(priv, 0x000530, 0xffff0000);
+	nv_icmd(priv, 0x000531, 0xffff0000);
+	nv_icmd(priv, 0x000532, 0xffff0000);
+	nv_icmd(priv, 0x000533, 0xffff0000);
+	nv_icmd(priv, 0x000534, 0xffff0000);
+	nv_icmd(priv, 0x000535, 0xffff0000);
+	nv_icmd(priv, 0x000536, 0xffff0000);
+	nv_icmd(priv, 0x000537, 0xffff0000);
+	nv_icmd(priv, 0x000538, 0xffff0000);
+	nv_icmd(priv, 0x000539, 0xffff0000);
+	nv_icmd(priv, 0x00053a, 0xffff0000);
+	nv_icmd(priv, 0x00053b, 0xffff0000);
+	nv_icmd(priv, 0x00053c, 0xffff0000);
+	nv_icmd(priv, 0x00053d, 0xffff0000);
+	nv_icmd(priv, 0x00053e, 0xffff0000);
+	nv_icmd(priv, 0x00053f, 0xffff0000);
+	nv_icmd(priv, 0x000585, 0x0000003f);
+	nv_icmd(priv, 0x000576, 0x00000003);
+	nv_icmd(priv, 0x00057b, 0x00000059);
+	nv_icmd(priv, 0x000586, 0x00000040);
+	nv_icmd(priv, 0x000582, 0x00000080);
+	nv_icmd(priv, 0x000583, 0x00000080);
+	nv_icmd(priv, 0x0005c2, 0x00000001);
+	nv_icmd(priv, 0x000638, 0x00000001);
+	nv_icmd(priv, 0x000639, 0x00000001);
+	nv_icmd(priv, 0x00063a, 0x00000002);
+	nv_icmd(priv, 0x00063b, 0x00000001);
+	nv_icmd(priv, 0x00063c, 0x00000001);
+	nv_icmd(priv, 0x00063d, 0x00000002);
+	nv_icmd(priv, 0x00063e, 0x00000001);
+	nv_icmd(priv, 0x0008b8, 0x00000001);
+	nv_icmd(priv, 0x0008b9, 0x00000001);
+	nv_icmd(priv, 0x0008ba, 0x00000001);
+	nv_icmd(priv, 0x0008bb, 0x00000001);
+	nv_icmd(priv, 0x0008bc, 0x00000001);
+	nv_icmd(priv, 0x0008bd, 0x00000001);
+	nv_icmd(priv, 0x0008be, 0x00000001);
+	nv_icmd(priv, 0x0008bf, 0x00000001);
+	nv_icmd(priv, 0x000900, 0x00000001);
+	nv_icmd(priv, 0x000901, 0x00000001);
+	nv_icmd(priv, 0x000902, 0x00000001);
+	nv_icmd(priv, 0x000903, 0x00000001);
+	nv_icmd(priv, 0x000904, 0x00000001);
+	nv_icmd(priv, 0x000905, 0x00000001);
+	nv_icmd(priv, 0x000906, 0x00000001);
+	nv_icmd(priv, 0x000907, 0x00000001);
+	nv_icmd(priv, 0x000908, 0x00000002);
+	nv_icmd(priv, 0x000909, 0x00000002);
+	nv_icmd(priv, 0x00090a, 0x00000002);
+	nv_icmd(priv, 0x00090b, 0x00000002);
+	nv_icmd(priv, 0x00090c, 0x00000002);
+	nv_icmd(priv, 0x00090d, 0x00000002);
+	nv_icmd(priv, 0x00090e, 0x00000002);
+	nv_icmd(priv, 0x00090f, 0x00000002);
+	nv_icmd(priv, 0x000910, 0x00000001);
+	nv_icmd(priv, 0x000911, 0x00000001);
+	nv_icmd(priv, 0x000912, 0x00000001);
+	nv_icmd(priv, 0x000913, 0x00000001);
+	nv_icmd(priv, 0x000914, 0x00000001);
+	nv_icmd(priv, 0x000915, 0x00000001);
+	nv_icmd(priv, 0x000916, 0x00000001);
+	nv_icmd(priv, 0x000917, 0x00000001);
+	nv_icmd(priv, 0x000918, 0x00000001);
+	nv_icmd(priv, 0x000919, 0x00000001);
+	nv_icmd(priv, 0x00091a, 0x00000001);
+	nv_icmd(priv, 0x00091b, 0x00000001);
+	nv_icmd(priv, 0x00091c, 0x00000001);
+	nv_icmd(priv, 0x00091d, 0x00000001);
+	nv_icmd(priv, 0x00091e, 0x00000001);
+	nv_icmd(priv, 0x00091f, 0x00000001);
+	nv_icmd(priv, 0x000920, 0x00000002);
+	nv_icmd(priv, 0x000921, 0x00000002);
+	nv_icmd(priv, 0x000922, 0x00000002);
+	nv_icmd(priv, 0x000923, 0x00000002);
+	nv_icmd(priv, 0x000924, 0x00000002);
+	nv_icmd(priv, 0x000925, 0x00000002);
+	nv_icmd(priv, 0x000926, 0x00000002);
+	nv_icmd(priv, 0x000927, 0x00000002);
+	nv_icmd(priv, 0x000928, 0x00000001);
+	nv_icmd(priv, 0x000929, 0x00000001);
+	nv_icmd(priv, 0x00092a, 0x00000001);
+	nv_icmd(priv, 0x00092b, 0x00000001);
+	nv_icmd(priv, 0x00092c, 0x00000001);
+	nv_icmd(priv, 0x00092d, 0x00000001);
+	nv_icmd(priv, 0x00092e, 0x00000001);
+	nv_icmd(priv, 0x00092f, 0x00000001);
+	nv_icmd(priv, 0x000648, 0x00000001);
+	nv_icmd(priv, 0x000649, 0x00000001);
+	nv_icmd(priv, 0x00064a, 0x00000001);
+	nv_icmd(priv, 0x00064b, 0x00000001);
+	nv_icmd(priv, 0x00064c, 0x00000001);
+	nv_icmd(priv, 0x00064d, 0x00000001);
+	nv_icmd(priv, 0x00064e, 0x00000001);
+	nv_icmd(priv, 0x00064f, 0x00000001);
+	nv_icmd(priv, 0x000650, 0x00000001);
+	nv_icmd(priv, 0x000658, 0x0000000f);
+	nv_icmd(priv, 0x0007ff, 0x0000000a);
+	nv_icmd(priv, 0x00066a, 0x40000000);
+	nv_icmd(priv, 0x00066b, 0x10000000);
+	nv_icmd(priv, 0x00066c, 0xffff0000);
+	nv_icmd(priv, 0x00066d, 0xffff0000);
+	nv_icmd(priv, 0x0007af, 0x00000008);
+	nv_icmd(priv, 0x0007b0, 0x00000008);
+	nv_icmd(priv, 0x0007f6, 0x00000001);
+	nv_icmd(priv, 0x0006b2, 0x00000055);
+	nv_icmd(priv, 0x0007ad, 0x00000003);
+	nv_icmd(priv, 0x000937, 0x00000001);
+	nv_icmd(priv, 0x000971, 0x00000008);
+	nv_icmd(priv, 0x000972, 0x00000040);
+	nv_icmd(priv, 0x000973, 0x0000012c);
+	nv_icmd(priv, 0x00097c, 0x00000040);
+	nv_icmd(priv, 0x000979, 0x00000003);
+	nv_icmd(priv, 0x000975, 0x00000020);
+	nv_icmd(priv, 0x000976, 0x00000001);
+	nv_icmd(priv, 0x000977, 0x00000020);
+	nv_icmd(priv, 0x000978, 0x00000001);
+	nv_icmd(priv, 0x000957, 0x00000003);
+	nv_icmd(priv, 0x00095e, 0x20164010);
+	nv_icmd(priv, 0x00095f, 0x00000020);
+	nv_icmd(priv, 0x00097d, 0x00000020);
+	nv_icmd(priv, 0x000683, 0x00000006);
+	nv_icmd(priv, 0x000685, 0x003fffff);
+	nv_icmd(priv, 0x000687, 0x003fffff);
+	nv_icmd(priv, 0x0006a0, 0x00000005);
+	nv_icmd(priv, 0x000840, 0x00400008);
+	nv_icmd(priv, 0x000841, 0x08000080);
+	nv_icmd(priv, 0x000842, 0x00400008);
+	nv_icmd(priv, 0x000843, 0x08000080);
+	nv_icmd(priv, 0x000818, 0x00000000);
+	nv_icmd(priv, 0x000819, 0x00000000);
+	nv_icmd(priv, 0x00081a, 0x00000000);
+	nv_icmd(priv, 0x00081b, 0x00000000);
+	nv_icmd(priv, 0x00081c, 0x00000000);
+	nv_icmd(priv, 0x00081d, 0x00000000);
+	nv_icmd(priv, 0x00081e, 0x00000000);
+	nv_icmd(priv, 0x00081f, 0x00000000);
+	nv_icmd(priv, 0x000848, 0x00000000);
+	nv_icmd(priv, 0x000849, 0x00000000);
+	nv_icmd(priv, 0x00084a, 0x00000000);
+	nv_icmd(priv, 0x00084b, 0x00000000);
+	nv_icmd(priv, 0x00084c, 0x00000000);
+	nv_icmd(priv, 0x00084d, 0x00000000);
+	nv_icmd(priv, 0x00084e, 0x00000000);
+	nv_icmd(priv, 0x00084f, 0x00000000);
+	nv_icmd(priv, 0x000850, 0x00000000);
+	nv_icmd(priv, 0x000851, 0x00000000);
+	nv_icmd(priv, 0x000852, 0x00000000);
+	nv_icmd(priv, 0x000853, 0x00000000);
+	nv_icmd(priv, 0x000854, 0x00000000);
+	nv_icmd(priv, 0x000855, 0x00000000);
+	nv_icmd(priv, 0x000856, 0x00000000);
+	nv_icmd(priv, 0x000857, 0x00000000);
+	nv_icmd(priv, 0x000738, 0x00000000);
+	nv_icmd(priv, 0x0006aa, 0x00000001);
+	nv_icmd(priv, 0x0006ab, 0x00000002);
+	nv_icmd(priv, 0x0006ac, 0x00000080);
+	nv_icmd(priv, 0x0006ad, 0x00000100);
+	nv_icmd(priv, 0x0006ae, 0x00000100);
+	nv_icmd(priv, 0x0006b1, 0x00000011);
+	nv_icmd(priv, 0x0006bb, 0x000000cf);
+	nv_icmd(priv, 0x0006ce, 0x2a712488);
+	nv_icmd(priv, 0x000739, 0x4085c000);
+	nv_icmd(priv, 0x00073a, 0x00000080);
+	nv_icmd(priv, 0x000786, 0x80000100);
+	nv_icmd(priv, 0x00073c, 0x00010100);
+	nv_icmd(priv, 0x00073d, 0x02800000);
+	nv_icmd(priv, 0x000787, 0x000000cf);
+	nv_icmd(priv, 0x00078c, 0x00000008);
+	nv_icmd(priv, 0x000792, 0x00000001);
+	nv_icmd(priv, 0x000794, 0x00000001);
+	nv_icmd(priv, 0x000795, 0x00000001);
+	nv_icmd(priv, 0x000796, 0x00000001);
+	nv_icmd(priv, 0x000797, 0x000000cf);
+	nv_icmd(priv, 0x000836, 0x00000001);
+	nv_icmd(priv, 0x00079a, 0x00000002);
+	nv_icmd(priv, 0x000833, 0x04444480);
+	nv_icmd(priv, 0x0007a1, 0x00000001);
+	nv_icmd(priv, 0x0007a3, 0x00000001);
+	nv_icmd(priv, 0x0007a4, 0x00000001);
+	nv_icmd(priv, 0x0007a5, 0x00000001);
+	nv_icmd(priv, 0x000831, 0x00000004);
+	nv_icmd(priv, 0x000b07, 0x00000002);
+	nv_icmd(priv, 0x000b08, 0x00000100);
+	nv_icmd(priv, 0x000b09, 0x00000100);
+	nv_icmd(priv, 0x000b0a, 0x00000001);
+	nv_icmd(priv, 0x000a04, 0x000000ff);
+	nv_icmd(priv, 0x000a0b, 0x00000040);
+	nv_icmd(priv, 0x00097f, 0x00000100);
+	nv_icmd(priv, 0x000a02, 0x00000001);
+	nv_icmd(priv, 0x000809, 0x00000007);
+	nv_icmd(priv, 0x00c221, 0x00000040);
+	nv_icmd(priv, 0x00c1b0, 0x0000000f);
+	nv_icmd(priv, 0x00c1b1, 0x0000000f);
+	nv_icmd(priv, 0x00c1b2, 0x0000000f);
+	nv_icmd(priv, 0x00c1b3, 0x0000000f);
+	nv_icmd(priv, 0x00c1b4, 0x0000000f);
+	nv_icmd(priv, 0x00c1b5, 0x0000000f);
+	nv_icmd(priv, 0x00c1b6, 0x0000000f);
+	nv_icmd(priv, 0x00c1b7, 0x0000000f);
+	nv_icmd(priv, 0x00c1b8, 0x0fac6881);
+	nv_icmd(priv, 0x00c1b9, 0x00fac688);
+	nv_icmd(priv, 0x00c401, 0x00000001);
+	nv_icmd(priv, 0x00c402, 0x00010001);
+	nv_icmd(priv, 0x00c403, 0x00000001);
+	nv_icmd(priv, 0x00c404, 0x00000001);
+	nv_icmd(priv, 0x00c40e, 0x00000020);
+	nv_icmd(priv, 0x00c500, 0x00000003);
+	nv_icmd(priv, 0x01e100, 0x00000001);
+	nv_icmd(priv, 0x001000, 0x00000002);
+	nv_icmd(priv, 0x0006aa, 0x00000001);
+	nv_icmd(priv, 0x0006ad, 0x00000100);
+	nv_icmd(priv, 0x0006ae, 0x00000100);
+	nv_icmd(priv, 0x0006b1, 0x00000011);
+	nv_icmd(priv, 0x00078c, 0x00000008);
+	nv_icmd(priv, 0x000792, 0x00000001);
+	nv_icmd(priv, 0x000794, 0x00000001);
+	nv_icmd(priv, 0x000795, 0x00000001);
+	nv_icmd(priv, 0x000796, 0x00000001);
+	nv_icmd(priv, 0x000797, 0x000000cf);
+	nv_icmd(priv, 0x00079a, 0x00000002);
+	nv_icmd(priv, 0x000833, 0x04444480);
+	nv_icmd(priv, 0x0007a1, 0x00000001);
+	nv_icmd(priv, 0x0007a3, 0x00000001);
+	nv_icmd(priv, 0x0007a4, 0x00000001);
+	nv_icmd(priv, 0x0007a5, 0x00000001);
+	nv_icmd(priv, 0x000831, 0x00000004);
+	nv_icmd(priv, 0x01e100, 0x00000001);
+	nv_icmd(priv, 0x001000, 0x00000008);
+	nv_icmd(priv, 0x000039, 0x00000000);
+	nv_icmd(priv, 0x00003a, 0x00000000);
+	nv_icmd(priv, 0x00003b, 0x00000000);
+	nv_icmd(priv, 0x000380, 0x00000001);
+	nv_icmd(priv, 0x000366, 0x00000000);
+	nv_icmd(priv, 0x000367, 0x00000000);
+	nv_icmd(priv, 0x000368, 0x00000fff);
+	nv_icmd(priv, 0x000370, 0x00000000);
+	nv_icmd(priv, 0x000371, 0x00000000);
+	nv_icmd(priv, 0x000372, 0x000fffff);
+	nv_icmd(priv, 0x000813, 0x00000006);
+	nv_icmd(priv, 0x000814, 0x00000008);
+	nv_icmd(priv, 0x000957, 0x00000003);
+	nv_icmd(priv, 0x000818, 0x00000000);
+	nv_icmd(priv, 0x000819, 0x00000000);
+	nv_icmd(priv, 0x00081a, 0x00000000);
+	nv_icmd(priv, 0x00081b, 0x00000000);
+	nv_icmd(priv, 0x00081c, 0x00000000);
+	nv_icmd(priv, 0x00081d, 0x00000000);
+	nv_icmd(priv, 0x00081e, 0x00000000);
+	nv_icmd(priv, 0x00081f, 0x00000000);
+	nv_icmd(priv, 0x000848, 0x00000000);
+	nv_icmd(priv, 0x000849, 0x00000000);
+	nv_icmd(priv, 0x00084a, 0x00000000);
+	nv_icmd(priv, 0x00084b, 0x00000000);
+	nv_icmd(priv, 0x00084c, 0x00000000);
+	nv_icmd(priv, 0x00084d, 0x00000000);
+	nv_icmd(priv, 0x00084e, 0x00000000);
+	nv_icmd(priv, 0x00084f, 0x00000000);
+	nv_icmd(priv, 0x000850, 0x00000000);
+	nv_icmd(priv, 0x000851, 0x00000000);
+	nv_icmd(priv, 0x000852, 0x00000000);
+	nv_icmd(priv, 0x000853, 0x00000000);
+	nv_icmd(priv, 0x000854, 0x00000000);
+	nv_icmd(priv, 0x000855, 0x00000000);
+	nv_icmd(priv, 0x000856, 0x00000000);
+	nv_icmd(priv, 0x000857, 0x00000000);
+	nv_icmd(priv, 0x000738, 0x00000000);
+	nv_icmd(priv, 0x000b07, 0x00000002);
+	nv_icmd(priv, 0x000b08, 0x00000100);
+	nv_icmd(priv, 0x000b09, 0x00000100);
+	nv_icmd(priv, 0x000b0a, 0x00000001);
+	nv_icmd(priv, 0x000a04, 0x000000ff);
+	nv_icmd(priv, 0x00097f, 0x00000100);
+	nv_icmd(priv, 0x000a02, 0x00000001);
+	nv_icmd(priv, 0x000809, 0x00000007);
+	nv_icmd(priv, 0x00c221, 0x00000040);
+	nv_icmd(priv, 0x00c401, 0x00000001);
+	nv_icmd(priv, 0x00c402, 0x00010001);
+	nv_icmd(priv, 0x00c403, 0x00000001);
+	nv_icmd(priv, 0x00c404, 0x00000001);
+	nv_icmd(priv, 0x00c40e, 0x00000020);
+	nv_icmd(priv, 0x00c500, 0x00000003);
+	nv_icmd(priv, 0x01e100, 0x00000001);
+	nv_icmd(priv, 0x001000, 0x00000001);
+	nv_icmd(priv, 0x000b07, 0x00000002);
+	nv_icmd(priv, 0x000b08, 0x00000100);
+	nv_icmd(priv, 0x000b09, 0x00000100);
+	nv_icmd(priv, 0x000b0a, 0x00000001);
+	nv_icmd(priv, 0x01e100, 0x00000001);
+	nv_wr32(priv, 0x400208, 0x00000000);
+}
+
+static void
+nve0_grctx_generate_a097(struct nvc0_graph_priv *priv)
+{
+	nv_mthd(priv, 0xa097, 0x0800, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0840, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0880, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x08c0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0900, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0940, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0980, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x09c0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0804, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0844, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0884, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x08c4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0904, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0944, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0984, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x09c4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0808, 0x00000400);
+	nv_mthd(priv, 0xa097, 0x0848, 0x00000400);
+	nv_mthd(priv, 0xa097, 0x0888, 0x00000400);
+	nv_mthd(priv, 0xa097, 0x08c8, 0x00000400);
+	nv_mthd(priv, 0xa097, 0x0908, 0x00000400);
+	nv_mthd(priv, 0xa097, 0x0948, 0x00000400);
+	nv_mthd(priv, 0xa097, 0x0988, 0x00000400);
+	nv_mthd(priv, 0xa097, 0x09c8, 0x00000400);
+	nv_mthd(priv, 0xa097, 0x080c, 0x00000300);
+	nv_mthd(priv, 0xa097, 0x084c, 0x00000300);
+	nv_mthd(priv, 0xa097, 0x088c, 0x00000300);
+	nv_mthd(priv, 0xa097, 0x08cc, 0x00000300);
+	nv_mthd(priv, 0xa097, 0x090c, 0x00000300);
+	nv_mthd(priv, 0xa097, 0x094c, 0x00000300);
+	nv_mthd(priv, 0xa097, 0x098c, 0x00000300);
+	nv_mthd(priv, 0xa097, 0x09cc, 0x00000300);
+	nv_mthd(priv, 0xa097, 0x0810, 0x000000cf);
+	nv_mthd(priv, 0xa097, 0x0850, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0890, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x08d0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0910, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0950, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0990, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x09d0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0814, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x0854, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x0894, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x08d4, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x0914, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x0954, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x0994, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x09d4, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x0818, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x0858, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x0898, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x08d8, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x0918, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x0958, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x0998, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x09d8, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x081c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x085c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x089c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x08dc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x091c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x095c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x099c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x09dc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0820, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0860, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x08a0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x08e0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0920, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0960, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x09a0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x09e0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c00, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c10, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c20, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c30, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c40, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c50, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c60, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c70, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c80, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c90, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1ca0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cb0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cc0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cd0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1ce0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cf0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c04, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c14, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c24, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c34, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c44, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c54, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c64, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c74, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c84, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c94, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1ca4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cb4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cc4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cd4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1ce4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cf4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c08, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c18, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c28, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c38, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c48, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c58, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c68, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c78, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c88, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c98, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1ca8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cb8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cc8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cd8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1ce8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cf8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c0c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c1c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c2c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c3c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c4c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c5c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c6c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c7c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c8c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1c9c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cbc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1ccc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cdc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1cfc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d00, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d10, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d20, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d30, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d40, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d50, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d60, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d70, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d80, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d90, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1da0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1db0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1dc0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1dd0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1de0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1df0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d04, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d14, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d24, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d34, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d44, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d54, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d64, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d74, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d84, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d94, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1da4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1db4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1dc4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1dd4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1de4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1df4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d08, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d18, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d28, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d38, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d48, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d58, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d68, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d78, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d88, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d98, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1da8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1db8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1dc8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1dd8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1de8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1df8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d0c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d1c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d2c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d3c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d4c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d5c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d6c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d7c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d8c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1d9c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1dac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1dbc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1dcc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1ddc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1dec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1dfc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f00, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f08, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f10, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f18, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f20, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f28, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f30, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f38, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f40, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f48, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f50, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f58, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f60, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f68, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f70, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f78, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f04, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f0c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f14, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f1c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f24, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f2c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f34, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f3c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f44, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f4c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f54, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f5c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f64, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f6c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f74, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f7c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f80, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f88, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f90, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f98, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fa0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fa8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fb0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fb8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fc0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fc8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fd0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fd8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fe0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fe8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1ff0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1ff8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f84, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f8c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f94, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1f9c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fa4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fb4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fbc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fc4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fcc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fd4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fdc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fe4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1fec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1ff4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1ffc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2000, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2040, 0x00000011);
+	nv_mthd(priv, 0xa097, 0x2080, 0x00000020);
+	nv_mthd(priv, 0xa097, 0x20c0, 0x00000030);
+	nv_mthd(priv, 0xa097, 0x2100, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x2140, 0x00000051);
+	nv_mthd(priv, 0xa097, 0x200c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x204c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x208c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x20cc, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x210c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x214c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x2010, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2050, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2090, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x20d0, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x2110, 0x00000003);
+	nv_mthd(priv, 0xa097, 0x2150, 0x00000004);
+	nv_mthd(priv, 0xa097, 0x0380, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x03a0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x03c0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x03e0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0384, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x03a4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x03c4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x03e4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0388, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x03a8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x03c8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x03e8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x038c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x03ac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x03cc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x03ec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0700, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0710, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0720, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0730, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0704, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0714, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0724, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0734, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0708, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0718, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0728, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0738, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2800, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2804, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2808, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x280c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2810, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2814, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2818, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x281c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2820, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2824, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2828, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x282c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2830, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2834, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2838, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x283c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2840, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2844, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2848, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x284c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2850, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2854, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2858, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x285c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2860, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2864, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2868, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x286c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2870, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2874, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2878, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x287c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2880, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2884, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2888, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x288c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2890, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2894, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2898, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x289c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28a0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28a4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28a8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28ac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28b0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28b4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28b8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28bc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28c0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28c4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28c8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28cc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28d0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28d4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28d8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28dc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28e0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28e4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28e8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28ec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28f0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28f4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28f8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x28fc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2900, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2904, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2908, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x290c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2910, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2914, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2918, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x291c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2920, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2924, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2928, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x292c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2930, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2934, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2938, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x293c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2940, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2944, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2948, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x294c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2950, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2954, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2958, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x295c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2960, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2964, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2968, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x296c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2970, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2974, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2978, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x297c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2980, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2984, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2988, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x298c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2990, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2994, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2998, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x299c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29a0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29a4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29a8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29ac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29b0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29b4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29b8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29bc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29c0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29c4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29c8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29cc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29d0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29d4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29d8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29dc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29e0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29e4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29e8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29ec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29f0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29f4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29f8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x29fc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a00, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a20, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a40, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a60, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a80, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0aa0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ac0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ae0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b00, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b20, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b40, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b60, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b80, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ba0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0bc0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0be0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a04, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a24, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a44, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a64, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a84, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0aa4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ac4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ae4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b04, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b24, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b44, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b64, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b84, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ba4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0bc4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0be4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a08, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a28, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a48, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a68, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a88, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0aa8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ac8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ae8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b08, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b28, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b48, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b68, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b88, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ba8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0bc8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0be8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a0c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a2c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a4c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a6c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a8c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0aac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0acc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0aec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b0c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b2c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b4c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b6c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b8c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0bac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0bcc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0bec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a10, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a30, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a50, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a70, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a90, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ab0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ad0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0af0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b10, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b30, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b50, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b70, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b90, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0bb0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0bd0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0bf0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a14, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a34, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a54, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a74, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0a94, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ab4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ad4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0af4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b14, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b34, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b54, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b74, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0b94, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0bb4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0bd4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0bf4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c00, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c10, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c20, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c30, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c40, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c50, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c60, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c70, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c80, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c90, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ca0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0cb0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0cc0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0cd0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ce0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0cf0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c04, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c14, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c24, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c34, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c44, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c54, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c64, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c74, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c84, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c94, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ca4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0cb4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0cc4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0cd4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ce4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0cf4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c08, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c18, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c28, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c38, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c48, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c58, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c68, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c78, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c88, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c98, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ca8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0cb8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0cc8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0cd8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ce8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0cf8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0c0c, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0c1c, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0c2c, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0c3c, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0c4c, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0c5c, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0c6c, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0c7c, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0c8c, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0c9c, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0cac, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0cbc, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0ccc, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0cdc, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0cec, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0cfc, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0d00, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d08, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d10, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d18, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d20, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d28, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d30, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d38, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d04, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d0c, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d14, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d1c, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d24, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d2c, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d34, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d3c, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e00, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0e10, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0e20, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0e30, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0e40, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0e50, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0e60, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0e70, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0e80, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0e90, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ea0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0eb0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ec0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ed0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ee0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ef0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0e04, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e14, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e24, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e34, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e44, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e54, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e64, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e74, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e84, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e94, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0ea4, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0eb4, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0ec4, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0ed4, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0ee4, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0ef4, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e08, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e18, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e28, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e38, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e48, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e58, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e68, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e78, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e88, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0e98, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0ea8, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0eb8, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0ec8, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0ed8, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0ee8, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0ef8, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d40, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d48, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d50, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d58, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d44, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d4c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d54, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d5c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1e00, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e20, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e40, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e60, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e80, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1ea0, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1ec0, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1ee0, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e04, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e24, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e44, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e64, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e84, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1ea4, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1ec4, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1ee4, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e08, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1e28, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1e48, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1e68, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1e88, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1ea8, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1ec8, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1ee8, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1e0c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e2c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e4c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e6c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e8c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1eac, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1ecc, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1eec, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e10, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e30, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e50, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e70, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e90, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1eb0, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1ed0, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1ef0, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e14, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1e34, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1e54, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1e74, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1e94, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1eb4, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1ed4, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1ef4, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1e18, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e38, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e58, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e78, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1e98, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1eb8, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1ed8, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1ef8, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x3400, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3404, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3408, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x340c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3410, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3414, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3418, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x341c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3420, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3424, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3428, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x342c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3430, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3434, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3438, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x343c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3440, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3444, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3448, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x344c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3450, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3454, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3458, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x345c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3460, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3464, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3468, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x346c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3470, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3474, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3478, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x347c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3480, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3484, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3488, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x348c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3490, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3494, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3498, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x349c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34a0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34a4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34a8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34ac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34b0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34b4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34b8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34bc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34c0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34c4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34c8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34cc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34d0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34d4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34d8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34dc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34e0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34e4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34e8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34ec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34f0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34f4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34f8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x34fc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3500, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3504, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3508, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x350c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3510, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3514, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3518, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x351c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3520, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3524, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3528, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x352c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3530, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3534, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3538, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x353c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3540, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3544, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3548, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x354c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3550, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3554, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3558, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x355c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3560, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3564, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3568, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x356c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3570, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3574, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3578, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x357c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3580, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3584, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3588, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x358c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3590, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3594, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x3598, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x359c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35a0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35a4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35a8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35ac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35b0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35b4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35b8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35bc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35c0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35c4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35c8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35cc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35d0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35d4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35d8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35dc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35e0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35e4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35e8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35ec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35f0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35f4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35f8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x35fc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x030c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1944, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1514, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d68, 0x0000ffff);
+	nv_mthd(priv, 0xa097, 0x121c, 0x0fac6881);
+	nv_mthd(priv, 0xa097, 0x0fac, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1538, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x0fe0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0fe4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0fe8, 0x00000014);
+	nv_mthd(priv, 0xa097, 0x0fec, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x0ff0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x179c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1228, 0x00000400);
+	nv_mthd(priv, 0xa097, 0x122c, 0x00000300);
+	nv_mthd(priv, 0xa097, 0x1230, 0x00010001);
+	nv_mthd(priv, 0xa097, 0x07f8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x15b4, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x15cc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1534, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0fb0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x15d0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x153c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x16b4, 0x00000003);
+	nv_mthd(priv, 0xa097, 0x0fbc, 0x0000ffff);
+	nv_mthd(priv, 0xa097, 0x0fc0, 0x0000ffff);
+	nv_mthd(priv, 0xa097, 0x0fc4, 0x0000ffff);
+	nv_mthd(priv, 0xa097, 0x0fc8, 0x0000ffff);
+	nv_mthd(priv, 0xa097, 0x0df8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0dfc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1948, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1970, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x161c, 0x000009f0);
+	nv_mthd(priv, 0xa097, 0x0dcc, 0x00000010);
+	nv_mthd(priv, 0xa097, 0x163c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x15e4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1160, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x1164, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x1168, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x116c, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x1170, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x1174, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x1178, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x117c, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x1180, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x1184, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x1188, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x118c, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x1190, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x1194, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x1198, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x119c, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11a0, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11a4, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11a8, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11ac, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11b0, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11b4, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11b8, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11bc, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11c0, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11c4, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11c8, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11cc, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11d0, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11d4, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11d8, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x11dc, 0x25e00040);
+	nv_mthd(priv, 0xa097, 0x1880, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1884, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1888, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x188c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1890, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1894, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1898, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x189c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18a0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18a4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18a8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18ac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18b0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18b4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18b8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18bc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18c0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18c4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18c8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18cc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18d0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18d4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18d8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18dc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18e0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18e4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18e8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18ec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18f0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18f4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18f8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x18fc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0f84, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0f88, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x17c8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x17cc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x17d0, 0x000000ff);
+	nv_mthd(priv, 0xa097, 0x17d4, 0xffffffff);
+	nv_mthd(priv, 0xa097, 0x17d8, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x17dc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x15f4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x15f8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1434, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1438, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d74, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0dec, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x13a4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1318, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1644, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0748, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0de8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1648, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x12a4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1120, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1124, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1128, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x112c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1118, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x164c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1658, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1910, 0x00000290);
+	nv_mthd(priv, 0xa097, 0x1518, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x165c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1520, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1604, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1570, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x13b0, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x13b4, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x020c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1670, 0x30201000);
+	nv_mthd(priv, 0xa097, 0x1674, 0x70605040);
+	nv_mthd(priv, 0xa097, 0x1678, 0xb8a89888);
+	nv_mthd(priv, 0xa097, 0x167c, 0xf8e8d8c8);
+	nv_mthd(priv, 0xa097, 0x166c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1680, 0x00ffff00);
+	nv_mthd(priv, 0xa097, 0x12d0, 0x00000003);
+	nv_mthd(priv, 0xa097, 0x12d4, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1684, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1688, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0dac, 0x00001b02);
+	nv_mthd(priv, 0xa097, 0x0db0, 0x00001b02);
+	nv_mthd(priv, 0xa097, 0x0db4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x168c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x15bc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x156c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x187c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1110, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x0dc0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0dc4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0dc8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1234, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1690, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x12ac, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x0790, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0794, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0798, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x079c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x07a0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x077c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1000, 0x00000010);
+	nv_mthd(priv, 0xa097, 0x10fc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1290, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0218, 0x00000010);
+	nv_mthd(priv, 0xa097, 0x12d8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x12dc, 0x00000010);
+	nv_mthd(priv, 0xa097, 0x0d94, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x155c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1560, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1564, 0x00000fff);
+	nv_mthd(priv, 0xa097, 0x1574, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1578, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x157c, 0x000fffff);
+	nv_mthd(priv, 0xa097, 0x1354, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1610, 0x00000012);
+	nv_mthd(priv, 0xa097, 0x1608, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x160c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x260c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x07ac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x162c, 0x00000003);
+	nv_mthd(priv, 0xa097, 0x0210, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0320, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0324, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0328, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x032c, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0330, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0334, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0338, 0x3f800000);
+	nv_mthd(priv, 0xa097, 0x0750, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0760, 0x39291909);
+	nv_mthd(priv, 0xa097, 0x0764, 0x79695949);
+	nv_mthd(priv, 0xa097, 0x0768, 0xb9a99989);
+	nv_mthd(priv, 0xa097, 0x076c, 0xf9e9d9c9);
+	nv_mthd(priv, 0xa097, 0x0770, 0x30201000);
+	nv_mthd(priv, 0xa097, 0x0774, 0x70605040);
+	nv_mthd(priv, 0xa097, 0x0778, 0x00009080);
+	nv_mthd(priv, 0xa097, 0x0780, 0x39291909);
+	nv_mthd(priv, 0xa097, 0x0784, 0x79695949);
+	nv_mthd(priv, 0xa097, 0x0788, 0xb9a99989);
+	nv_mthd(priv, 0xa097, 0x078c, 0xf9e9d9c9);
+	nv_mthd(priv, 0xa097, 0x07d0, 0x30201000);
+	nv_mthd(priv, 0xa097, 0x07d4, 0x70605040);
+	nv_mthd(priv, 0xa097, 0x07d8, 0x00009080);
+	nv_mthd(priv, 0xa097, 0x037c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x0740, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0744, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x2600, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1918, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x191c, 0x00000900);
+	nv_mthd(priv, 0xa097, 0x1920, 0x00000405);
+	nv_mthd(priv, 0xa097, 0x1308, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1924, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x13ac, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x192c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x193c, 0x00002c1c);
+	nv_mthd(priv, 0xa097, 0x0d7c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0f8c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x02c0, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1510, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1940, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ff4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0ff8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x194c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1950, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1968, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1590, 0x0000003f);
+	nv_mthd(priv, 0xa097, 0x07e8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x07ec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x07f0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x07f4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x196c, 0x00000011);
+	nv_mthd(priv, 0xa097, 0x02e4, 0x0000b001);
+	nv_mthd(priv, 0xa097, 0x036c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0370, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x197c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0fcc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0fd0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x02d8, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x1980, 0x00000080);
+	nv_mthd(priv, 0xa097, 0x1504, 0x00000080);
+	nv_mthd(priv, 0xa097, 0x1984, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0300, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x13a8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x12ec, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1310, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1314, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1380, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1384, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1388, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x138c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1390, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1394, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x139c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1398, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1594, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1598, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x159c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x15a0, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x15a4, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x0f54, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0f58, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0f5c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x19bc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0f9c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0fa0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x12cc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x12e8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x130c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1360, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1364, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1368, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x136c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1370, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1374, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1378, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x137c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x133c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1340, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1344, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1348, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x134c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1350, 0x00000002);
+	nv_mthd(priv, 0xa097, 0x1358, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x12e4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x131c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1320, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1324, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1328, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x19c0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1140, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x19c4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x19c8, 0x00001500);
+	nv_mthd(priv, 0xa097, 0x135c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0f90, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x19e0, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x19e4, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x19e8, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x19ec, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x19f0, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x19f4, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x19f8, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x19fc, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x19cc, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x15b8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1a00, 0x00001111);
+	nv_mthd(priv, 0xa097, 0x1a04, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1a08, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1a0c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1a10, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1a14, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1a18, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1a1c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d6c, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x0d70, 0xffff0000);
+	nv_mthd(priv, 0xa097, 0x10f8, 0x00001010);
+	nv_mthd(priv, 0xa097, 0x0d80, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d84, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d88, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d8c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0d90, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0da0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x07a4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x07a8, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1508, 0x80000000);
+	nv_mthd(priv, 0xa097, 0x150c, 0x40000000);
+	nv_mthd(priv, 0xa097, 0x1668, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0318, 0x00000008);
+	nv_mthd(priv, 0xa097, 0x031c, 0x00000008);
+	nv_mthd(priv, 0xa097, 0x0d9c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x0374, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0378, 0x00000020);
+	nv_mthd(priv, 0xa097, 0x07dc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x074c, 0x00000055);
+	nv_mthd(priv, 0xa097, 0x1420, 0x00000003);
+	nv_mthd(priv, 0xa097, 0x17bc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x17c0, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x17c4, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1008, 0x00000008);
+	nv_mthd(priv, 0xa097, 0x100c, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x1010, 0x0000012c);
+	nv_mthd(priv, 0xa097, 0x0d60, 0x00000040);
+	nv_mthd(priv, 0xa097, 0x075c, 0x00000003);
+	nv_mthd(priv, 0xa097, 0x1018, 0x00000020);
+	nv_mthd(priv, 0xa097, 0x101c, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1020, 0x00000020);
+	nv_mthd(priv, 0xa097, 0x1024, 0x00000001);
+	nv_mthd(priv, 0xa097, 0x1444, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x1448, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x144c, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0360, 0x20164010);
+	nv_mthd(priv, 0xa097, 0x0364, 0x00000020);
+	nv_mthd(priv, 0xa097, 0x0368, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0de4, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0204, 0x00000006);
+	nv_mthd(priv, 0xa097, 0x0208, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x02cc, 0x003fffff);
+	nv_mthd(priv, 0xa097, 0x02d0, 0x003fffff);
+	nv_mthd(priv, 0xa097, 0x1220, 0x00000005);
+	nv_mthd(priv, 0xa097, 0x0fdc, 0x00000000);
+	nv_mthd(priv, 0xa097, 0x0f98, 0x00400008);
+	nv_mthd(priv, 0xa097, 0x1284, 0x08000080);
+	nv_mthd(priv, 0xa097, 0x1450, 0x00400008);
+	nv_mthd(priv, 0xa097, 0x1454, 0x08000080);
+	nv_mthd(priv, 0xa097, 0x0214, 0x00000000);
+}
+
+static void
+nve0_grctx_generate_902d(struct nvc0_graph_priv *priv)
+{
+	nv_mthd(priv, 0x902d, 0x0200, 0x000000cf);
+	nv_mthd(priv, 0x902d, 0x0204, 0x00000001);
+	nv_mthd(priv, 0x902d, 0x0208, 0x00000020);
+	nv_mthd(priv, 0x902d, 0x020c, 0x00000001);
+	nv_mthd(priv, 0x902d, 0x0210, 0x00000000);
+	nv_mthd(priv, 0x902d, 0x0214, 0x00000080);
+	nv_mthd(priv, 0x902d, 0x0218, 0x00000100);
+	nv_mthd(priv, 0x902d, 0x021c, 0x00000100);
+	nv_mthd(priv, 0x902d, 0x0220, 0x00000000);
+	nv_mthd(priv, 0x902d, 0x0224, 0x00000000);
+	nv_mthd(priv, 0x902d, 0x0230, 0x000000cf);
+	nv_mthd(priv, 0x902d, 0x0234, 0x00000001);
+	nv_mthd(priv, 0x902d, 0x0238, 0x00000020);
+	nv_mthd(priv, 0x902d, 0x023c, 0x00000001);
+	nv_mthd(priv, 0x902d, 0x0244, 0x00000080);
+	nv_mthd(priv, 0x902d, 0x0248, 0x00000100);
+	nv_mthd(priv, 0x902d, 0x024c, 0x00000100);
+	nv_mthd(priv, 0x902d, 0x3410, 0x00000000);
+}
+
+static void
+nve0_graph_generate_unk40xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x404010, 0x0);
+	nv_wr32(priv, 0x404014, 0x0);
+	nv_wr32(priv, 0x404018, 0x0);
+	nv_wr32(priv, 0x40401c, 0x0);
+	nv_wr32(priv, 0x404020, 0x0);
+	nv_wr32(priv, 0x404024, 0xe000);
+	nv_wr32(priv, 0x404028, 0x0);
+	nv_wr32(priv, 0x4040a8, 0x0);
+	nv_wr32(priv, 0x4040ac, 0x0);
+	nv_wr32(priv, 0x4040b0, 0x0);
+	nv_wr32(priv, 0x4040b4, 0x0);
+	nv_wr32(priv, 0x4040b8, 0x0);
+	nv_wr32(priv, 0x4040bc, 0x0);
+	nv_wr32(priv, 0x4040c0, 0x0);
+	nv_wr32(priv, 0x4040c4, 0x0);
+	nv_wr32(priv, 0x4040c8, 0xf800008f);
+	nv_wr32(priv, 0x4040d0, 0x0);
+	nv_wr32(priv, 0x4040d4, 0x0);
+	nv_wr32(priv, 0x4040d8, 0x0);
+	nv_wr32(priv, 0x4040dc, 0x0);
+	nv_wr32(priv, 0x4040e0, 0x0);
+	nv_wr32(priv, 0x4040e4, 0x0);
+	nv_wr32(priv, 0x4040e8, 0x1000);
+	nv_wr32(priv, 0x4040f8, 0x0);
+	nv_wr32(priv, 0x404130, 0x0);
+	nv_wr32(priv, 0x404134, 0x0);
+	nv_wr32(priv, 0x404138, 0x20000040);
+	nv_wr32(priv, 0x404150, 0x2e);
+	nv_wr32(priv, 0x404154, 0x400);
+	nv_wr32(priv, 0x404158, 0x200);
+	nv_wr32(priv, 0x404164, 0x55);
+	nv_wr32(priv, 0x4041a0, 0x0);
+	nv_wr32(priv, 0x4041a4, 0x0);
+	nv_wr32(priv, 0x4041a8, 0x0);
+	nv_wr32(priv, 0x4041ac, 0x0);
+	nv_wr32(priv, 0x404200, 0x0);
+	nv_wr32(priv, 0x404204, 0x0);
+	nv_wr32(priv, 0x404208, 0x0);
+	nv_wr32(priv, 0x40420c, 0x0);
+}
+
+static void
+nve0_graph_generate_unk44xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x404404, 0x0);
+	nv_wr32(priv, 0x404408, 0x0);
+	nv_wr32(priv, 0x40440c, 0x0);
+	nv_wr32(priv, 0x404410, 0x0);
+	nv_wr32(priv, 0x404414, 0x0);
+	nv_wr32(priv, 0x404418, 0x0);
+	nv_wr32(priv, 0x40441c, 0x0);
+	nv_wr32(priv, 0x404420, 0x0);
+	nv_wr32(priv, 0x404424, 0x0);
+	nv_wr32(priv, 0x404428, 0x0);
+	nv_wr32(priv, 0x40442c, 0x0);
+	nv_wr32(priv, 0x404430, 0x0);
+	nv_wr32(priv, 0x404434, 0x0);
+	nv_wr32(priv, 0x404438, 0x0);
+	nv_wr32(priv, 0x404460, 0x0);
+	nv_wr32(priv, 0x404464, 0x0);
+	nv_wr32(priv, 0x404468, 0xffffff);
+	nv_wr32(priv, 0x40446c, 0x0);
+	nv_wr32(priv, 0x404480, 0x1);
+	nv_wr32(priv, 0x404498, 0x1);
+}
+
+static void
+nve0_graph_generate_unk46xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x404604, 0x14);
+	nv_wr32(priv, 0x404608, 0x0);
+	nv_wr32(priv, 0x40460c, 0x3fff);
+	nv_wr32(priv, 0x404610, 0x100);
+	nv_wr32(priv, 0x404618, 0x0);
+	nv_wr32(priv, 0x40461c, 0x0);
+	nv_wr32(priv, 0x404620, 0x0);
+	nv_wr32(priv, 0x404624, 0x0);
+	nv_wr32(priv, 0x40462c, 0x0);
+	nv_wr32(priv, 0x404630, 0x0);
+	nv_wr32(priv, 0x404640, 0x0);
+	nv_wr32(priv, 0x404654, 0x0);
+	nv_wr32(priv, 0x404660, 0x0);
+	nv_wr32(priv, 0x404678, 0x0);
+	nv_wr32(priv, 0x40467c, 0x2);
+	nv_wr32(priv, 0x404680, 0x0);
+	nv_wr32(priv, 0x404684, 0x0);
+	nv_wr32(priv, 0x404688, 0x0);
+	nv_wr32(priv, 0x40468c, 0x0);
+	nv_wr32(priv, 0x404690, 0x0);
+	nv_wr32(priv, 0x404694, 0x0);
+	nv_wr32(priv, 0x404698, 0x0);
+	nv_wr32(priv, 0x40469c, 0x0);
+	nv_wr32(priv, 0x4046a0, 0x7f0080);
+	nv_wr32(priv, 0x4046a4, 0x0);
+	nv_wr32(priv, 0x4046a8, 0x0);
+	nv_wr32(priv, 0x4046ac, 0x0);
+	nv_wr32(priv, 0x4046b0, 0x0);
+	nv_wr32(priv, 0x4046b4, 0x0);
+	nv_wr32(priv, 0x4046b8, 0x0);
+	nv_wr32(priv, 0x4046bc, 0x0);
+	nv_wr32(priv, 0x4046c0, 0x0);
+	nv_wr32(priv, 0x4046c8, 0x0);
+	nv_wr32(priv, 0x4046cc, 0x0);
+	nv_wr32(priv, 0x4046d0, 0x0);
+}
+
+static void
+nve0_graph_generate_unk47xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x404700, 0x0);
+	nv_wr32(priv, 0x404704, 0x0);
+	nv_wr32(priv, 0x404708, 0x0);
+	nv_wr32(priv, 0x404718, 0x0);
+	nv_wr32(priv, 0x40471c, 0x0);
+	nv_wr32(priv, 0x404720, 0x0);
+	nv_wr32(priv, 0x404724, 0x0);
+	nv_wr32(priv, 0x404728, 0x0);
+	nv_wr32(priv, 0x40472c, 0x0);
+	nv_wr32(priv, 0x404730, 0x0);
+	nv_wr32(priv, 0x404734, 0x100);
+	nv_wr32(priv, 0x404738, 0x0);
+	nv_wr32(priv, 0x40473c, 0x0);
+	nv_wr32(priv, 0x404744, 0x0);
+	nv_wr32(priv, 0x404748, 0x0);
+	nv_wr32(priv, 0x404754, 0x0);
+}
+
+static void
+nve0_graph_generate_unk58xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x405800, 0xf8000bf);
+	nv_wr32(priv, 0x405830, 0x2180648);
+	nv_wr32(priv, 0x405834, 0x8000000);
+	nv_wr32(priv, 0x405838, 0x0);
+	nv_wr32(priv, 0x405854, 0x0);
+	nv_wr32(priv, 0x405870, 0x1);
+	nv_wr32(priv, 0x405874, 0x1);
+	nv_wr32(priv, 0x405878, 0x1);
+	nv_wr32(priv, 0x40587c, 0x1);
+	nv_wr32(priv, 0x405a00, 0x0);
+	nv_wr32(priv, 0x405a04, 0x0);
+	nv_wr32(priv, 0x405a18, 0x0);
+	nv_wr32(priv, 0x405b00, 0x0);
+	nv_wr32(priv, 0x405b10, 0x1000);
+}
+
+static void
+nve0_graph_generate_unk60xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x406020, 0x4103c1);
+	nv_wr32(priv, 0x406028, 0x1);
+	nv_wr32(priv, 0x40602c, 0x1);
+	nv_wr32(priv, 0x406030, 0x1);
+	nv_wr32(priv, 0x406034, 0x1);
+}
+
+static void
+nve0_graph_generate_unk64xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x4064a8, 0x0);
+	nv_wr32(priv, 0x4064ac, 0x3fff);
+	nv_wr32(priv, 0x4064b4, 0x0);
+	nv_wr32(priv, 0x4064b8, 0x0);
+	nv_wr32(priv, 0x4064c0, 0x801a00f0);
+	nv_wr32(priv, 0x4064c4, 0x192ffff);
+	nv_wr32(priv, 0x4064c8, 0x1800600);
+	nv_wr32(priv, 0x4064cc, 0x0);
+	nv_wr32(priv, 0x4064d0, 0x0);
+	nv_wr32(priv, 0x4064d4, 0x0);
+	nv_wr32(priv, 0x4064d8, 0x0);
+	nv_wr32(priv, 0x4064dc, 0x0);
+	nv_wr32(priv, 0x4064e0, 0x0);
+	nv_wr32(priv, 0x4064e4, 0x0);
+	nv_wr32(priv, 0x4064e8, 0x0);
+	nv_wr32(priv, 0x4064ec, 0x0);
+	nv_wr32(priv, 0x4064fc, 0x22a);
+}
+
+static void
+nve0_graph_generate_unk70xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x407040, 0x0);
+}
+
+static void
+nve0_graph_generate_unk78xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x407804, 0x23);
+	nv_wr32(priv, 0x40780c, 0xa418820);
+	nv_wr32(priv, 0x407810, 0x62080e6);
+	nv_wr32(priv, 0x407814, 0x20398a4);
+	nv_wr32(priv, 0x407818, 0xe629062);
+	nv_wr32(priv, 0x40781c, 0xa418820);
+	nv_wr32(priv, 0x407820, 0xe6);
+	nv_wr32(priv, 0x4078bc, 0x103);
+}
+
+static void
+nve0_graph_generate_unk80xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x408000, 0x0);
+	nv_wr32(priv, 0x408004, 0x0);
+	nv_wr32(priv, 0x408008, 0x30);
+	nv_wr32(priv, 0x40800c, 0x0);
+	nv_wr32(priv, 0x408010, 0x0);
+	nv_wr32(priv, 0x408014, 0x69);
+	nv_wr32(priv, 0x408018, 0xe100e100);
+	nv_wr32(priv, 0x408064, 0x0);
+}
+
+static void
+nve0_graph_generate_unk88xx(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x408800, 0x2802a3c);
+	nv_wr32(priv, 0x408804, 0x40);
+	nv_wr32(priv, 0x408808, 0x1043e005);
+	nv_wr32(priv, 0x408840, 0xb);
+	nv_wr32(priv, 0x408900, 0x3080b801);
+	nv_wr32(priv, 0x408904, 0x62000001);
+	nv_wr32(priv, 0x408908, 0xc8102f);
+	nv_wr32(priv, 0x408980, 0x11d);
+}
+
+static void
+nve0_graph_generate_gpc(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x418380, 0x16);
+	nv_wr32(priv, 0x418400, 0x38004e00);
+	nv_wr32(priv, 0x418404, 0x71e0ffff);
+	nv_wr32(priv, 0x41840c, 0x1008);
+	nv_wr32(priv, 0x418410, 0xfff0fff);
+	nv_wr32(priv, 0x418414, 0x2200fff);
+	nv_wr32(priv, 0x418450, 0x0);
+	nv_wr32(priv, 0x418454, 0x0);
+	nv_wr32(priv, 0x418458, 0x0);
+	nv_wr32(priv, 0x41845c, 0x0);
+	nv_wr32(priv, 0x418460, 0x0);
+	nv_wr32(priv, 0x418464, 0x0);
+	nv_wr32(priv, 0x418468, 0x1);
+	nv_wr32(priv, 0x41846c, 0x0);
+	nv_wr32(priv, 0x418470, 0x0);
+	nv_wr32(priv, 0x418600, 0x1f);
+	nv_wr32(priv, 0x418684, 0xf);
+	nv_wr32(priv, 0x418700, 0x2);
+	nv_wr32(priv, 0x418704, 0x80);
+	nv_wr32(priv, 0x418708, 0x0);
+	nv_wr32(priv, 0x41870c, 0x0);
+	nv_wr32(priv, 0x418710, 0x0);
+	nv_wr32(priv, 0x418800, 0x7006860a);
+	nv_wr32(priv, 0x418808, 0x0);
+	nv_wr32(priv, 0x41880c, 0x0);
+	nv_wr32(priv, 0x418810, 0x0);
+	nv_wr32(priv, 0x418828, 0x44);
+	nv_wr32(priv, 0x418830, 0x10000001);
+	nv_wr32(priv, 0x4188d8, 0x8);
+	nv_wr32(priv, 0x4188e0, 0x1000000);
+	nv_wr32(priv, 0x4188e8, 0x0);
+	nv_wr32(priv, 0x4188ec, 0x0);
+	nv_wr32(priv, 0x4188f0, 0x0);
+	nv_wr32(priv, 0x4188f4, 0x0);
+	nv_wr32(priv, 0x4188f8, 0x0);
+	nv_wr32(priv, 0x4188fc, 0x20100018);
+	nv_wr32(priv, 0x41891c, 0xff00ff);
+	nv_wr32(priv, 0x418924, 0x0);
+	nv_wr32(priv, 0x418928, 0xffff00);
+	nv_wr32(priv, 0x41892c, 0xff00);
+	nv_wr32(priv, 0x418a00, 0x0);
+	nv_wr32(priv, 0x418a04, 0x0);
+	nv_wr32(priv, 0x418a08, 0x0);
+	nv_wr32(priv, 0x418a0c, 0x10000);
+	nv_wr32(priv, 0x418a10, 0x0);
+	nv_wr32(priv, 0x418a14, 0x0);
+	nv_wr32(priv, 0x418a18, 0x0);
+	nv_wr32(priv, 0x418a20, 0x0);
+	nv_wr32(priv, 0x418a24, 0x0);
+	nv_wr32(priv, 0x418a28, 0x0);
+	nv_wr32(priv, 0x418a2c, 0x10000);
+	nv_wr32(priv, 0x418a30, 0x0);
+	nv_wr32(priv, 0x418a34, 0x0);
+	nv_wr32(priv, 0x418a38, 0x0);
+	nv_wr32(priv, 0x418a40, 0x0);
+	nv_wr32(priv, 0x418a44, 0x0);
+	nv_wr32(priv, 0x418a48, 0x0);
+	nv_wr32(priv, 0x418a4c, 0x10000);
+	nv_wr32(priv, 0x418a50, 0x0);
+	nv_wr32(priv, 0x418a54, 0x0);
+	nv_wr32(priv, 0x418a58, 0x0);
+	nv_wr32(priv, 0x418a60, 0x0);
+	nv_wr32(priv, 0x418a64, 0x0);
+	nv_wr32(priv, 0x418a68, 0x0);
+	nv_wr32(priv, 0x418a6c, 0x10000);
+	nv_wr32(priv, 0x418a70, 0x0);
+	nv_wr32(priv, 0x418a74, 0x0);
+	nv_wr32(priv, 0x418a78, 0x0);
+	nv_wr32(priv, 0x418a80, 0x0);
+	nv_wr32(priv, 0x418a84, 0x0);
+	nv_wr32(priv, 0x418a88, 0x0);
+	nv_wr32(priv, 0x418a8c, 0x10000);
+	nv_wr32(priv, 0x418a90, 0x0);
+	nv_wr32(priv, 0x418a94, 0x0);
+	nv_wr32(priv, 0x418a98, 0x0);
+	nv_wr32(priv, 0x418aa0, 0x0);
+	nv_wr32(priv, 0x418aa4, 0x0);
+	nv_wr32(priv, 0x418aa8, 0x0);
+	nv_wr32(priv, 0x418aac, 0x10000);
+	nv_wr32(priv, 0x418ab0, 0x0);
+	nv_wr32(priv, 0x418ab4, 0x0);
+	nv_wr32(priv, 0x418ab8, 0x0);
+	nv_wr32(priv, 0x418ac0, 0x0);
+	nv_wr32(priv, 0x418ac4, 0x0);
+	nv_wr32(priv, 0x418ac8, 0x0);
+	nv_wr32(priv, 0x418acc, 0x10000);
+	nv_wr32(priv, 0x418ad0, 0x0);
+	nv_wr32(priv, 0x418ad4, 0x0);
+	nv_wr32(priv, 0x418ad8, 0x0);
+	nv_wr32(priv, 0x418ae0, 0x0);
+	nv_wr32(priv, 0x418ae4, 0x0);
+	nv_wr32(priv, 0x418ae8, 0x0);
+	nv_wr32(priv, 0x418aec, 0x10000);
+	nv_wr32(priv, 0x418af0, 0x0);
+	nv_wr32(priv, 0x418af4, 0x0);
+	nv_wr32(priv, 0x418af8, 0x0);
+	nv_wr32(priv, 0x418b00, 0x6);
+	nv_wr32(priv, 0x418b08, 0xa418820);
+	nv_wr32(priv, 0x418b0c, 0x62080e6);
+	nv_wr32(priv, 0x418b10, 0x20398a4);
+	nv_wr32(priv, 0x418b14, 0xe629062);
+	nv_wr32(priv, 0x418b18, 0xa418820);
+	nv_wr32(priv, 0x418b1c, 0xe6);
+	nv_wr32(priv, 0x418bb8, 0x103);
+	nv_wr32(priv, 0x418c08, 0x1);
+	nv_wr32(priv, 0x418c10, 0x0);
+	nv_wr32(priv, 0x418c14, 0x0);
+	nv_wr32(priv, 0x418c18, 0x0);
+	nv_wr32(priv, 0x418c1c, 0x0);
+	nv_wr32(priv, 0x418c20, 0x0);
+	nv_wr32(priv, 0x418c24, 0x0);
+	nv_wr32(priv, 0x418c28, 0x0);
+	nv_wr32(priv, 0x418c2c, 0x0);
+	nv_wr32(priv, 0x418c40, 0xffffffff);
+	nv_wr32(priv, 0x418c6c, 0x1);
+	nv_wr32(priv, 0x418c80, 0x20200004);
+	nv_wr32(priv, 0x418c8c, 0x1);
+	nv_wr32(priv, 0x419000, 0x780);
+	nv_wr32(priv, 0x419004, 0x0);
+	nv_wr32(priv, 0x419008, 0x0);
+	nv_wr32(priv, 0x419014, 0x4);
+}
+
+static void
+nve0_graph_generate_tpc(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x419848, 0x0);
+	nv_wr32(priv, 0x419864, 0x129);
+	nv_wr32(priv, 0x419888, 0x0);
+	nv_wr32(priv, 0x419a00, 0xf0);
+	nv_wr32(priv, 0x419a04, 0x1);
+	nv_wr32(priv, 0x419a08, 0x21);
+	nv_wr32(priv, 0x419a0c, 0x20000);
+	nv_wr32(priv, 0x419a10, 0x0);
+	nv_wr32(priv, 0x419a14, 0x200);
+	nv_wr32(priv, 0x419a1c, 0xc000);
+	nv_wr32(priv, 0x419a20, 0x800);
+	nv_wr32(priv, 0x419a30, 0x1);
+	nv_wr32(priv, 0x419ac4, 0x37f440);
+	nv_wr32(priv, 0x419c00, 0xa);
+	nv_wr32(priv, 0x419c04, 0x80000006);
+	nv_wr32(priv, 0x419c08, 0x2);
+	nv_wr32(priv, 0x419c20, 0x0);
+	nv_wr32(priv, 0x419c24, 0x84210);
+	nv_wr32(priv, 0x419c28, 0x3efbefbe);
+	nv_wr32(priv, 0x419ce8, 0x0);
+	nv_wr32(priv, 0x419cf4, 0x3203);
+	nv_wr32(priv, 0x419e04, 0x0);
+	nv_wr32(priv, 0x419e08, 0x0);
+	nv_wr32(priv, 0x419e0c, 0x0);
+	nv_wr32(priv, 0x419e10, 0x402);
+	nv_wr32(priv, 0x419e44, 0x13eff2);
+	nv_wr32(priv, 0x419e48, 0x0);
+	nv_wr32(priv, 0x419e4c, 0x7f);
+	nv_wr32(priv, 0x419e50, 0x0);
+	nv_wr32(priv, 0x419e54, 0x0);
+	nv_wr32(priv, 0x419e58, 0x0);
+	nv_wr32(priv, 0x419e5c, 0x0);
+	nv_wr32(priv, 0x419e60, 0x0);
+	nv_wr32(priv, 0x419e64, 0x0);
+	nv_wr32(priv, 0x419e68, 0x0);
+	nv_wr32(priv, 0x419e6c, 0x0);
+	nv_wr32(priv, 0x419e70, 0x0);
+	nv_wr32(priv, 0x419e74, 0x0);
+	nv_wr32(priv, 0x419e78, 0x0);
+	nv_wr32(priv, 0x419e7c, 0x0);
+	nv_wr32(priv, 0x419e80, 0x0);
+	nv_wr32(priv, 0x419e84, 0x0);
+	nv_wr32(priv, 0x419e88, 0x0);
+	nv_wr32(priv, 0x419e8c, 0x0);
+	nv_wr32(priv, 0x419e90, 0x0);
+	nv_wr32(priv, 0x419e94, 0x0);
+	nv_wr32(priv, 0x419e98, 0x0);
+	nv_wr32(priv, 0x419eac, 0x1fcf);
+	nv_wr32(priv, 0x419eb0, 0xd3f);
+	nv_wr32(priv, 0x419ec8, 0x1304f);
+	nv_wr32(priv, 0x419f30, 0x0);
+	nv_wr32(priv, 0x419f34, 0x0);
+	nv_wr32(priv, 0x419f38, 0x0);
+	nv_wr32(priv, 0x419f3c, 0x0);
+	nv_wr32(priv, 0x419f40, 0x0);
+	nv_wr32(priv, 0x419f44, 0x0);
+	nv_wr32(priv, 0x419f48, 0x0);
+	nv_wr32(priv, 0x419f4c, 0x0);
+	nv_wr32(priv, 0x419f58, 0x0);
+	nv_wr32(priv, 0x419f78, 0xb);
+}
+
+static void
+nve0_graph_generate_tpcunk(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x41be24, 0x6);
+	nv_wr32(priv, 0x41bec0, 0x12180000);
+	nv_wr32(priv, 0x41bec4, 0x37f7f);
+	nv_wr32(priv, 0x41bee4, 0x6480430);
+	nv_wr32(priv, 0x41bf00, 0xa418820);
+	nv_wr32(priv, 0x41bf04, 0x62080e6);
+	nv_wr32(priv, 0x41bf08, 0x20398a4);
+	nv_wr32(priv, 0x41bf0c, 0xe629062);
+	nv_wr32(priv, 0x41bf10, 0xa418820);
+	nv_wr32(priv, 0x41bf14, 0xe6);
+	nv_wr32(priv, 0x41bfd0, 0x900103);
+	nv_wr32(priv, 0x41bfe0, 0x400001);
+	nv_wr32(priv, 0x41bfe4, 0x0);
+}
+
+int
+nve0_grctx_generate(struct nvc0_graph_priv *priv)
+{
+	struct nvc0_grctx info;
+	int ret, i, gpc, tpc, id;
+	u32 data[6] = {}, data2[2] = {}, tmp;
+	u32 tpc_set = 0, tpc_mask = 0;
+	u32 magic[GPC_MAX][2], offset;
+	u8 tpcnr[GPC_MAX], a, b;
+	u8 shift, ntpcv;
+
+	ret = nvc0_grctx_init(priv, &info);
+	if (ret)
+		return ret;
+
+	nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
+	nv_wr32(priv, 0x400204, 0x00000000);
+	nv_wr32(priv, 0x400208, 0x00000000);
+
+	nve0_graph_generate_unk40xx(priv);
+	nve0_graph_generate_unk44xx(priv);
+	nve0_graph_generate_unk46xx(priv);
+	nve0_graph_generate_unk47xx(priv);
+	nve0_graph_generate_unk58xx(priv);
+	nve0_graph_generate_unk60xx(priv);
+	nve0_graph_generate_unk64xx(priv);
+	nve0_graph_generate_unk70xx(priv);
+	nve0_graph_generate_unk78xx(priv);
+	nve0_graph_generate_unk80xx(priv);
+	nve0_graph_generate_unk88xx(priv);
+	nve0_graph_generate_gpc(priv);
+	nve0_graph_generate_tpc(priv);
+	nve0_graph_generate_tpcunk(priv);
+
+	nv_wr32(priv, 0x404154, 0x0);
+
+	mmio_data(0x003000, 0x0100, NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS);
+	mmio_data(0x008000, 0x0100, NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS);
+	mmio_data(0x060000, 0x1000, NV_MEM_ACCESS_RW);
+	mmio_list(0x40800c, 0x00000000,  8, 1);
+	mmio_list(0x408010, 0x80000000,  0, 0);
+	mmio_list(0x419004, 0x00000000,  8, 1);
+	mmio_list(0x419008, 0x00000000,  0, 0);
+	mmio_list(0x4064cc, 0x80000000,  0, 0);
+	mmio_list(0x408004, 0x00000000,  8, 0);
+	mmio_list(0x408008, 0x80000030,  0, 0);
+	mmio_list(0x418808, 0x00000000,  8, 0);
+	mmio_list(0x41880c, 0x80000030,  0, 0);
+	mmio_list(0x4064c8, 0x01800600,  0, 0);
+	mmio_list(0x418810, 0x80000000, 12, 2);
+	mmio_list(0x419848, 0x10000000, 12, 2);
+	mmio_list(0x405830, 0x02180648,  0, 0);
+	mmio_list(0x4064c4, 0x0192ffff,  0, 0);
+	for (gpc = 0, offset = 0; gpc < priv->gpc_nr; gpc++) {
+		u16 magic0 = 0x0218 * priv->tpc_nr[gpc];
+		u16 magic1 = 0x0648 * priv->tpc_nr[gpc];
+		magic[gpc][0]  = 0x10000000 | (magic0 << 16) | offset;
+		magic[gpc][1]  = 0x00000000 | (magic1 << 16);
+		offset += 0x0324 * priv->tpc_nr[gpc];
+	}
+	for (gpc = 0; gpc < priv->gpc_nr; gpc++) {
+		mmio_list(GPC_UNIT(gpc, 0x30c0), magic[gpc][0], 0, 0);
+		mmio_list(GPC_UNIT(gpc, 0x30e4), magic[gpc][1] | offset, 0, 0);
+		offset += 0x07ff * priv->tpc_nr[gpc];
+	}
+	mmio_list(0x17e91c, 0x06060609, 0, 0);
+	mmio_list(0x17e920, 0x00090a05, 0, 0);
+
+	nv_wr32(priv, 0x418c6c, 0x1);
+	nv_wr32(priv, 0x41980c, 0x10);
+	nv_wr32(priv, 0x41be08, 0x4);
+	nv_wr32(priv, 0x4064c0, 0x801a00f0);
+	nv_wr32(priv, 0x405800, 0xf8000bf);
+	nv_wr32(priv, 0x419c00, 0xa);
+
+	for (tpc = 0, id = 0; tpc < 4; tpc++) {
+		for (gpc = 0; gpc < priv->gpc_nr; gpc++) {
+			if (tpc < priv->tpc_nr[gpc]) {
+				nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x0698), id);
+				nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x04e8), id);
+				nv_wr32(priv, GPC_UNIT(gpc, 0x0c10 + tpc * 4), id);
+				nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x0088), id++);
+			}
+
+			nv_wr32(priv, GPC_UNIT(gpc, 0x0c08), priv->tpc_nr[gpc]);
+			nv_wr32(priv, GPC_UNIT(gpc, 0x0c8c), priv->tpc_nr[gpc]);
+		}
+	}
+
+	tmp = 0;
+	for (i = 0; i < priv->gpc_nr; i++)
+		tmp |= priv->tpc_nr[i] << (i * 4);
+	nv_wr32(priv, 0x406028, tmp);
+	nv_wr32(priv, 0x405870, tmp);
+
+	nv_wr32(priv, 0x40602c, 0x0);
+	nv_wr32(priv, 0x405874, 0x0);
+	nv_wr32(priv, 0x406030, 0x0);
+	nv_wr32(priv, 0x405878, 0x0);
+	nv_wr32(priv, 0x406034, 0x0);
+	nv_wr32(priv, 0x40587c, 0x0);
+
+	/* calculate first set of magics */
+	memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr));
+
+	gpc = -1;
+	for (tpc = 0; tpc < priv->tpc_total; tpc++) {
+		do {
+			gpc = (gpc + 1) % priv->gpc_nr;
+		} while (!tpcnr[gpc]);
+		tpcnr[gpc]--;
+
+		data[tpc / 6] |= gpc << ((tpc % 6) * 5);
+	}
+
+	for (; tpc < 32; tpc++)
+		data[tpc / 6] |= 7 << ((tpc % 6) * 5);
+
+	/* and the second... */
+	shift = 0;
+	ntpcv = priv->tpc_total;
+	while (!(ntpcv & (1 << 4))) {
+		ntpcv <<= 1;
+		shift++;
+	}
+
+	data2[0]  = ntpcv << 16;
+	data2[0] |= shift << 21;
+	data2[0] |= (((1 << (0 + 5)) % ntpcv) << 24);
+	data2[0] |= priv->tpc_total << 8;
+	data2[0] |= priv->magic_not_rop_nr;
+	for (i = 1; i < 7; i++)
+		data2[1] |= ((1 << (i + 5)) % ntpcv) << ((i - 1) * 5);
+
+	/* and write it all the various parts of PGRAPH */
+	nv_wr32(priv, 0x418bb8, (priv->tpc_total << 8) | priv->magic_not_rop_nr);
+	for (i = 0; i < 6; i++)
+		nv_wr32(priv, 0x418b08 + (i * 4), data[i]);
+
+	nv_wr32(priv, 0x41bfd0, data2[0]);
+	nv_wr32(priv, 0x41bfe4, data2[1]);
+	for (i = 0; i < 6; i++)
+		nv_wr32(priv, 0x41bf00 + (i * 4), data[i]);
+
+	nv_wr32(priv, 0x4078bc, (priv->tpc_total << 8) | priv->magic_not_rop_nr);
+	for (i = 0; i < 6; i++)
+		nv_wr32(priv, 0x40780c + (i * 4), data[i]);
+
+
+	memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr));
+	for (gpc = 0; gpc < priv->gpc_nr; gpc++)
+		tpc_mask |= ((1 << priv->tpc_nr[gpc]) - 1) << (gpc * 8);
+
+	for (i = 0, gpc = -1, b = -1; i < 32; i++) {
+		a = (i * (priv->tpc_total - 1)) / 32;
+		if (a != b) {
+			b = a;
+			do {
+				gpc = (gpc + 1) % priv->gpc_nr;
+			} while (!tpcnr[gpc]);
+			tpc = priv->tpc_nr[gpc] - tpcnr[gpc]--;
+
+			tpc_set |= 1 << ((gpc * 8) + tpc);
+		}
+
+		nv_wr32(priv, 0x406800 + (i * 0x20), tpc_set);
+		nv_wr32(priv, 0x406c00 + (i * 0x20), tpc_set ^ tpc_mask);
+	}
+
+	for (i = 0; i < 8; i++)
+		nv_wr32(priv, 0x4064d0 + (i * 0x04), 0x00000000);
+
+	nv_wr32(priv, 0x405b00, 0x201);
+	nv_wr32(priv, 0x408850, 0x2);
+	nv_wr32(priv, 0x408958, 0x2);
+	nv_wr32(priv, 0x419f78, 0xa);
+
+	nve0_grctx_generate_icmd(priv);
+	nve0_grctx_generate_a097(priv);
+	nve0_grctx_generate_902d(priv);
+
+	nv_mask(priv, 0x000260, 0x00000001, 0x00000001);
+	nv_wr32(priv, 0x418800, 0x7026860a); //XXX
+	nv_wr32(priv, 0x41be10, 0x00bb8bc7); //XXX
+	return nvc0_grctx_fini(&info);
+}
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnvc0.fuc b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnvc0.fuc
new file mode 100644
index 0000000..b86cc60
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnvc0.fuc
@@ -0,0 +1,539 @@
+/* fuc microcode for nvc0 PGRAPH/GPC
+ *
+ * Copyright 2011 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+/* To build:
+ *    m4 gpcnvc0.fuc | envyas -a -w -m fuc -V fuc3 -o gpcnvc0.fuc.h
+ */
+
+/* TODO
+ * - bracket certain functions with scratch writes, useful for debugging
+ * - watchdog timer around ctx operations
+ */
+
+.section #nvc0_grgpc_data
+include(`nvc0.fuc')
+gpc_id:			.b32 0
+gpc_mmio_list_head:	.b32 0
+gpc_mmio_list_tail:	.b32 0
+
+tpc_count:		.b32 0
+tpc_mask:		.b32 0
+tpc_mmio_list_head:	.b32 0
+tpc_mmio_list_tail:	.b32 0
+
+cmd_queue:		queue_init
+
+// chipset descriptions
+chipsets:
+.b8  0xc0 0 0 0
+.b16 #nvc0_gpc_mmio_head
+.b16 #nvc0_gpc_mmio_tail
+.b16 #nvc0_tpc_mmio_head
+.b16 #nvc0_tpc_mmio_tail
+.b8  0xc1 0 0 0
+.b16 #nvc0_gpc_mmio_head
+.b16 #nvc1_gpc_mmio_tail
+.b16 #nvc0_tpc_mmio_head
+.b16 #nvc1_tpc_mmio_tail
+.b8  0xc3 0 0 0
+.b16 #nvc0_gpc_mmio_head
+.b16 #nvc0_gpc_mmio_tail
+.b16 #nvc0_tpc_mmio_head
+.b16 #nvc3_tpc_mmio_tail
+.b8  0xc4 0 0 0
+.b16 #nvc0_gpc_mmio_head
+.b16 #nvc0_gpc_mmio_tail
+.b16 #nvc0_tpc_mmio_head
+.b16 #nvc3_tpc_mmio_tail
+.b8  0xc8 0 0 0
+.b16 #nvc0_gpc_mmio_head
+.b16 #nvc0_gpc_mmio_tail
+.b16 #nvc0_tpc_mmio_head
+.b16 #nvc0_tpc_mmio_tail
+.b8  0xce 0 0 0
+.b16 #nvc0_gpc_mmio_head
+.b16 #nvc0_gpc_mmio_tail
+.b16 #nvc0_tpc_mmio_head
+.b16 #nvc3_tpc_mmio_tail
+.b8  0xcf 0 0 0
+.b16 #nvc0_gpc_mmio_head
+.b16 #nvc0_gpc_mmio_tail
+.b16 #nvc0_tpc_mmio_head
+.b16 #nvcf_tpc_mmio_tail
+.b8  0xd9 0 0 0
+.b16 #nvd9_gpc_mmio_head
+.b16 #nvd9_gpc_mmio_tail
+.b16 #nvd9_tpc_mmio_head
+.b16 #nvd9_tpc_mmio_tail
+.b8  0 0 0 0
+
+// GPC mmio lists
+nvc0_gpc_mmio_head:
+mmctx_data(0x000380, 1)
+mmctx_data(0x000400, 6)
+mmctx_data(0x000450, 9)
+mmctx_data(0x000600, 1)
+mmctx_data(0x000684, 1)
+mmctx_data(0x000700, 5)
+mmctx_data(0x000800, 1)
+mmctx_data(0x000808, 3)
+mmctx_data(0x000828, 1)
+mmctx_data(0x000830, 1)
+mmctx_data(0x0008d8, 1)
+mmctx_data(0x0008e0, 1)
+mmctx_data(0x0008e8, 6)
+mmctx_data(0x00091c, 1)
+mmctx_data(0x000924, 3)
+mmctx_data(0x000b00, 1)
+mmctx_data(0x000b08, 6)
+mmctx_data(0x000bb8, 1)
+mmctx_data(0x000c08, 1)
+mmctx_data(0x000c10, 8)
+mmctx_data(0x000c80, 1)
+mmctx_data(0x000c8c, 1)
+mmctx_data(0x001000, 3)
+mmctx_data(0x001014, 1)
+nvc0_gpc_mmio_tail:
+mmctx_data(0x000c6c, 1);
+nvc1_gpc_mmio_tail:
+
+nvd9_gpc_mmio_head:
+mmctx_data(0x000380, 1)
+mmctx_data(0x000400, 2)
+mmctx_data(0x00040c, 3)
+mmctx_data(0x000450, 9)
+mmctx_data(0x000600, 1)
+mmctx_data(0x000684, 1)
+mmctx_data(0x000700, 5)
+mmctx_data(0x000800, 1)
+mmctx_data(0x000808, 3)
+mmctx_data(0x000828, 1)
+mmctx_data(0x000830, 1)
+mmctx_data(0x0008d8, 1)
+mmctx_data(0x0008e0, 1)
+mmctx_data(0x0008e8, 6)
+mmctx_data(0x00091c, 1)
+mmctx_data(0x000924, 3)
+mmctx_data(0x000b00, 1)
+mmctx_data(0x000b08, 6)
+mmctx_data(0x000bb8, 1)
+mmctx_data(0x000c08, 1)
+mmctx_data(0x000c10, 8)
+mmctx_data(0x000c6c, 1)
+mmctx_data(0x000c80, 1)
+mmctx_data(0x000c8c, 1)
+mmctx_data(0x001000, 3)
+mmctx_data(0x001014, 1)
+nvd9_gpc_mmio_tail:
+
+// TPC mmio lists
+nvc0_tpc_mmio_head:
+mmctx_data(0x000018, 1)
+mmctx_data(0x00003c, 1)
+mmctx_data(0x000048, 1)
+mmctx_data(0x000064, 1)
+mmctx_data(0x000088, 1)
+mmctx_data(0x000200, 6)
+mmctx_data(0x00021c, 2)
+mmctx_data(0x000300, 6)
+mmctx_data(0x0003d0, 1)
+mmctx_data(0x0003e0, 2)
+mmctx_data(0x000400, 3)
+mmctx_data(0x000420, 1)
+mmctx_data(0x0004b0, 1)
+mmctx_data(0x0004e8, 1)
+mmctx_data(0x0004f4, 1)
+mmctx_data(0x000520, 2)
+mmctx_data(0x000604, 4)
+mmctx_data(0x000644, 20)
+mmctx_data(0x000698, 1)
+mmctx_data(0x000750, 2)
+nvc0_tpc_mmio_tail:
+mmctx_data(0x000758, 1)
+mmctx_data(0x0002c4, 1)
+mmctx_data(0x0006e0, 1)
+nvcf_tpc_mmio_tail:
+mmctx_data(0x0004bc, 1)
+nvc3_tpc_mmio_tail:
+mmctx_data(0x000544, 1)
+nvc1_tpc_mmio_tail:
+
+nvd9_tpc_mmio_head:
+mmctx_data(0x000018, 1)
+mmctx_data(0x00003c, 1)
+mmctx_data(0x000048, 1)
+mmctx_data(0x000064, 1)
+mmctx_data(0x000088, 1)
+mmctx_data(0x000200, 6)
+mmctx_data(0x00021c, 2)
+mmctx_data(0x0002c4, 1)
+mmctx_data(0x000300, 6)
+mmctx_data(0x0003d0, 1)
+mmctx_data(0x0003e0, 2)
+mmctx_data(0x000400, 3)
+mmctx_data(0x000420, 3)
+mmctx_data(0x0004b0, 1)
+mmctx_data(0x0004e8, 1)
+mmctx_data(0x0004f4, 1)
+mmctx_data(0x000520, 2)
+mmctx_data(0x000544, 1)
+mmctx_data(0x000604, 4)
+mmctx_data(0x000644, 20)
+mmctx_data(0x000698, 1)
+mmctx_data(0x0006e0, 1)
+mmctx_data(0x000750, 3)
+nvd9_tpc_mmio_tail:
+
+.section #nvc0_grgpc_code
+bra #init
+define(`include_code')
+include(`nvc0.fuc')
+
+// reports an exception to the host
+//
+// In: $r15 error code (see nvc0.fuc)
+//
+error:
+	push $r14
+	mov $r14 -0x67ec 	// 0x9814
+	sethi $r14 0x400000
+	call #nv_wr32		// HUB_CTXCTL_CC_SCRATCH[5] = error code
+	add b32 $r14 0x41c
+	mov $r15 1
+	call #nv_wr32		// HUB_CTXCTL_INTR_UP_SET
+	pop $r14
+	ret
+
+// GPC fuc initialisation, executed by triggering ucode start, will
+// fall through to main loop after completion.
+//
+// Input:
+//   CC_SCRATCH[0]: chipset (PMC_BOOT_0 read returns 0x0bad0bad... sigh)
+//   CC_SCRATCH[1]: context base
+//
+// Output:
+//   CC_SCRATCH[0]:
+//	     31:31: set to signal completion
+//   CC_SCRATCH[1]:
+//	      31:0: GPC context size
+//
+init:
+	clear b32 $r0
+	mov $sp $r0
+
+	// enable fifo access
+	mov $r1 0x1200
+	mov $r2 2
+	iowr I[$r1 + 0x000] $r2		// FIFO_ENABLE
+
+	// setup i0 handler, and route all interrupts to it
+	mov $r1 #ih
+	mov $iv0 $r1
+	mov $r1 0x400
+	iowr I[$r1 + 0x300] $r0		// INTR_DISPATCH
+
+	// enable fifo interrupt
+	mov $r2 4
+	iowr I[$r1 + 0x000] $r2		// INTR_EN_SET
+
+	// enable interrupts
+	bset $flags ie0
+
+	// figure out which GPC we are, and how many TPCs we have
+	mov $r1 0x608
+	shl b32 $r1 6
+	iord $r2 I[$r1 + 0x000]		// UNITS
+	mov $r3 1
+	and $r2 0x1f
+	shl b32 $r3 $r2
+	sub b32 $r3 1
+	st b32 D[$r0 + #tpc_count] $r2
+	st b32 D[$r0 + #tpc_mask] $r3
+	add b32 $r1 0x400
+	iord $r2 I[$r1 + 0x000]		// MYINDEX
+	st b32 D[$r0 + #gpc_id] $r2
+
+	// find context data for this chipset
+	mov $r2 0x800
+	shl b32 $r2 6
+	iord $r2 I[$r2 + 0x000]		// CC_SCRATCH[0]
+	mov $r1 #chipsets - 12
+	init_find_chipset:
+		add b32 $r1 12
+		ld b32 $r3 D[$r1 + 0x00]
+		cmpu b32 $r3 $r2
+		bra e #init_context
+		cmpu b32 $r3 0
+		bra ne #init_find_chipset
+		// unknown chipset
+		ret
+
+	// initialise context base, and size tracking
+	init_context:
+	mov $r2 0x800
+	shl b32 $r2 6
+	iord $r2 I[$r2 + 0x100]	// CC_SCRATCH[1], initial base
+	clear b32 $r3		// track GPC context size here
+
+	// set mmctx base addresses now so we don't have to do it later,
+	// they don't currently ever change
+	mov $r4 0x700
+	shl b32 $r4 6
+	shr b32 $r5 $r2 8
+	iowr I[$r4 + 0x000] $r5		// MMCTX_SAVE_SWBASE
+	iowr I[$r4 + 0x100] $r5		// MMCTX_LOAD_SWBASE
+
+	// calculate GPC mmio context size, store the chipset-specific
+	// mmio list pointers somewhere we can get at them later without
+	// re-parsing the chipset list
+	clear b32 $r14
+	clear b32 $r15
+	ld b16 $r14 D[$r1 + 4]
+	ld b16 $r15 D[$r1 + 6]
+	st b16 D[$r0 + #gpc_mmio_list_head] $r14
+	st b16 D[$r0 + #gpc_mmio_list_tail] $r15
+	call #mmctx_size
+	add b32 $r2 $r15
+	add b32 $r3 $r15
+
+	// calculate per-TPC mmio context size, store the list pointers
+	ld b16 $r14 D[$r1 + 8]
+	ld b16 $r15 D[$r1 + 10]
+	st b16 D[$r0 + #tpc_mmio_list_head] $r14
+	st b16 D[$r0 + #tpc_mmio_list_tail] $r15
+	call #mmctx_size
+	ld b32 $r14 D[$r0 + #tpc_count]
+	mulu $r14 $r15
+	add b32 $r2 $r14
+	add b32 $r3 $r14
+
+	// round up base/size to 256 byte boundary (for strand SWBASE)
+	add b32 $r4 0x1300
+	shr b32 $r3 2
+	iowr I[$r4 + 0x000] $r3		// MMCTX_LOAD_COUNT, wtf for?!?
+	shr b32 $r2 8
+	shr b32 $r3 6
+	add b32 $r2 1
+	add b32 $r3 1
+	shl b32 $r2 8
+	shl b32 $r3 8
+
+	// calculate size of strand context data
+	mov b32 $r15 $r2
+	call #strand_ctx_init
+	add b32 $r3 $r15
+
+	// save context size, and tell HUB we're done
+	mov $r1 0x800
+	shl b32 $r1 6
+	iowr I[$r1 + 0x100] $r3		// CC_SCRATCH[1]  = context size
+	add b32 $r1 0x800
+	clear b32 $r2
+	bset $r2 31
+	iowr I[$r1 + 0x000] $r2		// CC_SCRATCH[0] |= 0x80000000
+
+// Main program loop, very simple, sleeps until woken up by the interrupt
+// handler, pulls a command from the queue and executes its handler
+//
+main:
+	bset $flags $p0
+	sleep $p0
+	mov $r13 #cmd_queue
+	call #queue_get
+	bra $p1 #main
+
+	// 0x0000-0x0003 are all context transfers
+	cmpu b32 $r14 0x04
+	bra nc #main_not_ctx_xfer
+		// fetch $flags and mask off $p1/$p2
+		mov $r1 $flags
+		mov $r2 0x0006
+		not b32 $r2
+		and $r1 $r2
+		// set $p1/$p2 according to transfer type
+		shl b32 $r14 1
+		or $r1 $r14
+		mov $flags $r1
+		// transfer context data
+		call #ctx_xfer
+		bra #main
+
+	main_not_ctx_xfer:
+	shl b32 $r15 $r14 16
+	or $r15 E_BAD_COMMAND
+	call #error
+	bra #main
+
+// interrupt handler
+ih:
+	push $r8
+	mov $r8 $flags
+	push $r8
+	push $r9
+	push $r10
+	push $r11
+	push $r13
+	push $r14
+	push $r15
+
+	// incoming fifo command?
+	iord $r10 I[$r0 + 0x200]	// INTR
+	and $r11 $r10 0x00000004
+	bra e #ih_no_fifo
+		// queue incoming fifo command for later processing
+		mov $r11 0x1900
+		mov $r13 #cmd_queue
+		iord $r14 I[$r11 + 0x100]	// FIFO_CMD
+		iord $r15 I[$r11 + 0x000]	// FIFO_DATA
+		call #queue_put
+		add b32 $r11 0x400
+		mov $r14 1
+		iowr I[$r11 + 0x000] $r14	// FIFO_ACK
+
+	// ack, and wake up main()
+	ih_no_fifo:
+	iowr I[$r0 + 0x100] $r10	// INTR_ACK
+
+	pop $r15
+	pop $r14
+	pop $r13
+	pop $r11
+	pop $r10
+	pop $r9
+	pop $r8
+	mov $flags $r8
+	pop $r8
+	bclr $flags $p0
+	iret
+
+// Set this GPC's bit in HUB_BAR, used to signal completion of various
+// activities to the HUB fuc
+//
+hub_barrier_done:
+	mov $r15 1
+	ld b32 $r14 D[$r0 + #gpc_id]
+	shl b32 $r15 $r14
+	mov $r14 -0x6be8 	// 0x409418 - HUB_BAR_SET
+	sethi $r14 0x400000
+	call #nv_wr32
+	ret
+
+// Disables various things, waits a bit, and re-enables them..
+//
+// Not sure how exactly this helps, perhaps "ENABLE" is not such a
+// good description for the bits we turn off?  Anyways, without this,
+// funny things happen.
+//
+ctx_redswitch:
+	mov $r14 0x614
+	shl b32 $r14 6
+	mov $r15 0x020
+	iowr I[$r14] $r15	// GPC_RED_SWITCH = POWER
+	mov $r15 8
+	ctx_redswitch_delay:
+		sub b32 $r15 1
+		bra ne #ctx_redswitch_delay
+	mov $r15 0xa20
+	iowr I[$r14] $r15	// GPC_RED_SWITCH = UNK11, ENABLE, POWER
+	ret
+
+// Transfer GPC context data between GPU and storage area
+//
+// In: $r15 context base address
+//     $p1 clear on save, set on load
+//     $p2 set if opposite direction done/will be done, so:
+//		on save it means: "a load will follow this save"
+//		on load it means: "a save preceeded this load"
+//
+ctx_xfer:
+	// set context base address
+	mov $r1 0xa04
+	shl b32 $r1 6
+	iowr I[$r1 + 0x000] $r15// MEM_BASE
+	bra not $p1 #ctx_xfer_not_load
+		call #ctx_redswitch
+	ctx_xfer_not_load:
+
+	// strands
+	mov $r1 0x4afc
+	sethi $r1 0x20000
+	mov $r2 0xc
+	iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x0c
+	call #strand_wait
+	mov $r2 0x47fc
+	sethi $r2 0x20000
+	iowr I[$r2] $r0		// STRAND_FIRST_GENE(0x3f) = 0x00
+	xbit $r2 $flags $p1
+	add b32 $r2 3
+	iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x03/0x04 (SAVE/LOAD)
+
+	// mmio context
+	xbit $r10 $flags $p1	// direction
+	or $r10 2		// first
+	mov $r11 0x0000
+	sethi $r11 0x500000
+	ld b32 $r12 D[$r0 + #gpc_id]
+	shl b32 $r12 15
+	add b32 $r11 $r12	// base = NV_PGRAPH_GPCn
+	ld b32 $r12 D[$r0 + #gpc_mmio_list_head]
+	ld b32 $r13 D[$r0 + #gpc_mmio_list_tail]
+	mov $r14 0		// not multi
+	call #mmctx_xfer
+
+	// per-TPC mmio context
+	xbit $r10 $flags $p1	// direction
+	or $r10 4		// last
+	mov $r11 0x4000
+	sethi $r11 0x500000	// base = NV_PGRAPH_GPC0_TPC0
+	ld b32 $r12 D[$r0 + #gpc_id]
+	shl b32 $r12 15
+	add b32 $r11 $r12	// base = NV_PGRAPH_GPCn_TPC0
+	ld b32 $r12 D[$r0 + #tpc_mmio_list_head]
+	ld b32 $r13 D[$r0 + #tpc_mmio_list_tail]
+	ld b32 $r15 D[$r0 + #tpc_mask]
+	mov $r14 0x800		// stride = 0x800
+	call #mmctx_xfer
+
+	// wait for strands to finish
+	call #strand_wait
+
+	// if load, or a save without a load following, do some
+	// unknown stuff that's done after finishing a block of
+	// strand commands
+	bra $p1 #ctx_xfer_post
+	bra not $p2 #ctx_xfer_done
+	ctx_xfer_post:
+		mov $r1 0x4afc
+		sethi $r1 0x20000
+		mov $r2 0xd
+		iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x0d
+		call #strand_wait
+
+	// mark completion in HUB's barrier
+	ctx_xfer_done:
+	call #hub_barrier_done
+	ret
+
+.align 256
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnvc0.fuc.h b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnvc0.fuc.h
new file mode 100644
index 0000000..96050dd
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnvc0.fuc.h
@@ -0,0 +1,604 @@
+uint32_t nvc0_grgpc_data[] = {
+/* 0x0000: gpc_id */
+	0x00000000,
+/* 0x0004: gpc_mmio_list_head */
+	0x00000000,
+/* 0x0008: gpc_mmio_list_tail */
+	0x00000000,
+/* 0x000c: tpc_count */
+	0x00000000,
+/* 0x0010: tpc_mask */
+	0x00000000,
+/* 0x0014: tpc_mmio_list_head */
+	0x00000000,
+/* 0x0018: tpc_mmio_list_tail */
+	0x00000000,
+/* 0x001c: cmd_queue */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0064: chipsets */
+	0x000000c0,
+	0x012800c8,
+	0x01e40194,
+	0x000000c1,
+	0x012c00c8,
+	0x01f80194,
+	0x000000c3,
+	0x012800c8,
+	0x01f40194,
+	0x000000c4,
+	0x012800c8,
+	0x01f40194,
+	0x000000c8,
+	0x012800c8,
+	0x01e40194,
+	0x000000ce,
+	0x012800c8,
+	0x01f40194,
+	0x000000cf,
+	0x012800c8,
+	0x01f00194,
+	0x000000d9,
+	0x0194012c,
+	0x025401f8,
+	0x00000000,
+/* 0x00c8: nvc0_gpc_mmio_head */
+	0x00000380,
+	0x14000400,
+	0x20000450,
+	0x00000600,
+	0x00000684,
+	0x10000700,
+	0x00000800,
+	0x08000808,
+	0x00000828,
+	0x00000830,
+	0x000008d8,
+	0x000008e0,
+	0x140008e8,
+	0x0000091c,
+	0x08000924,
+	0x00000b00,
+	0x14000b08,
+	0x00000bb8,
+	0x00000c08,
+	0x1c000c10,
+	0x00000c80,
+	0x00000c8c,
+	0x08001000,
+	0x00001014,
+/* 0x0128: nvc0_gpc_mmio_tail */
+	0x00000c6c,
+/* 0x012c: nvc1_gpc_mmio_tail */
+/* 0x012c: nvd9_gpc_mmio_head */
+	0x00000380,
+	0x04000400,
+	0x0800040c,
+	0x20000450,
+	0x00000600,
+	0x00000684,
+	0x10000700,
+	0x00000800,
+	0x08000808,
+	0x00000828,
+	0x00000830,
+	0x000008d8,
+	0x000008e0,
+	0x140008e8,
+	0x0000091c,
+	0x08000924,
+	0x00000b00,
+	0x14000b08,
+	0x00000bb8,
+	0x00000c08,
+	0x1c000c10,
+	0x00000c6c,
+	0x00000c80,
+	0x00000c8c,
+	0x08001000,
+	0x00001014,
+/* 0x0194: nvd9_gpc_mmio_tail */
+/* 0x0194: nvc0_tpc_mmio_head */
+	0x00000018,
+	0x0000003c,
+	0x00000048,
+	0x00000064,
+	0x00000088,
+	0x14000200,
+	0x0400021c,
+	0x14000300,
+	0x000003d0,
+	0x040003e0,
+	0x08000400,
+	0x00000420,
+	0x000004b0,
+	0x000004e8,
+	0x000004f4,
+	0x04000520,
+	0x0c000604,
+	0x4c000644,
+	0x00000698,
+	0x04000750,
+/* 0x01e4: nvc0_tpc_mmio_tail */
+	0x00000758,
+	0x000002c4,
+	0x000006e0,
+/* 0x01f0: nvcf_tpc_mmio_tail */
+	0x000004bc,
+/* 0x01f4: nvc3_tpc_mmio_tail */
+	0x00000544,
+/* 0x01f8: nvc1_tpc_mmio_tail */
+/* 0x01f8: nvd9_tpc_mmio_head */
+	0x00000018,
+	0x0000003c,
+	0x00000048,
+	0x00000064,
+	0x00000088,
+	0x14000200,
+	0x0400021c,
+	0x000002c4,
+	0x14000300,
+	0x000003d0,
+	0x040003e0,
+	0x08000400,
+	0x08000420,
+	0x000004b0,
+	0x000004e8,
+	0x000004f4,
+	0x04000520,
+	0x00000544,
+	0x0c000604,
+	0x4c000644,
+	0x00000698,
+	0x000006e0,
+	0x08000750,
+};
+
+uint32_t nvc0_grgpc_code[] = {
+	0x03060ef5,
+/* 0x0004: queue_put */
+	0x9800d898,
+	0x86f001d9,
+	0x0489b808,
+	0xf00c1bf4,
+	0x21f502f7,
+	0x00f802ec,
+/* 0x001c: queue_put_next */
+	0xb60798c4,
+	0x8dbb0384,
+	0x0880b600,
+	0x80008e80,
+	0x90b6018f,
+	0x0f94f001,
+	0xf801d980,
+/* 0x0039: queue_get */
+	0x0131f400,
+	0x9800d898,
+	0x89b801d9,
+	0x210bf404,
+	0xb60789c4,
+	0x9dbb0394,
+	0x0890b600,
+	0x98009e98,
+	0x80b6019f,
+	0x0f84f001,
+	0xf400d880,
+/* 0x0066: queue_get_done */
+	0x00f80132,
+/* 0x0068: nv_rd32 */
+	0x0728b7f1,
+	0xb906b4b6,
+	0xc9f002ec,
+	0x00bcd01f,
+/* 0x0078: nv_rd32_wait */
+	0xc800bccf,
+	0x1bf41fcc,
+	0x06a7f0fa,
+	0x010321f5,
+	0xf840bfcf,
+/* 0x008d: nv_wr32 */
+	0x28b7f100,
+	0x06b4b607,
+	0xb980bfd0,
+	0xc9f002ec,
+	0x1ec9f01f,
+/* 0x00a3: nv_wr32_wait */
+	0xcf00bcd0,
+	0xccc800bc,
+	0xfa1bf41f,
+/* 0x00ae: watchdog_reset */
+	0x87f100f8,
+	0x84b60430,
+	0x1ff9f006,
+	0xf8008fd0,
+/* 0x00bd: watchdog_clear */
+	0x3087f100,
+	0x0684b604,
+	0xf80080d0,
+/* 0x00c9: wait_donez */
+	0x3c87f100,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d000,
+	0x081887f1,
+	0xd00684b6,
+/* 0x00e2: wait_done_wait_donez */
+	0x87f1008a,
+	0x84b60400,
+	0x0088cf06,
+	0xf4888aff,
+	0x87f1f31b,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00099,
+/* 0x0103: wait_doneo */
+	0xf100f800,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00099f0,
+	0x87f10089,
+	0x84b60818,
+	0x008ad006,
+/* 0x011c: wait_done_wait_doneo */
+	0x040087f1,
+	0xcf0684b6,
+	0x8aff0088,
+	0xf30bf488,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0099f094,
+	0xf80089d0,
+/* 0x013d: mmctx_size */
+/* 0x013f: nv_mmctx_size_loop */
+	0x9894bd00,
+	0x85b600e8,
+	0x0180b61a,
+	0xbb0284b6,
+	0xe0b60098,
+	0x04efb804,
+	0xb9eb1bf4,
+	0x00f8029f,
+/* 0x015c: mmctx_xfer */
+	0x083c87f1,
+	0xbd0684b6,
+	0x0199f094,
+	0xf10089d0,
+	0xb6071087,
+	0x94bd0684,
+	0xf405bbfd,
+	0x8bd0090b,
+	0x0099f000,
+/* 0x0180: mmctx_base_disabled */
+	0xf405eefd,
+	0x8ed00c0b,
+	0xc08fd080,
+/* 0x018f: mmctx_multi_disabled */
+	0xb70199f0,
+	0xc8010080,
+	0xb4b600ab,
+	0x0cb9f010,
+	0xb601aec8,
+	0xbefd11e4,
+	0x008bd005,
+/* 0x01a8: mmctx_exec_loop */
+/* 0x01a8: mmctx_wait_free */
+	0xf0008ecf,
+	0x0bf41fe4,
+	0x00ce98fa,
+	0xd005e9fd,
+	0xc0b6c08e,
+	0x04cdb804,
+	0xc8e81bf4,
+	0x1bf402ab,
+/* 0x01c9: mmctx_fini_wait */
+	0x008bcf18,
+	0xb01fb4f0,
+	0x1bf410b4,
+	0x02a7f0f7,
+	0xf4c921f4,
+/* 0x01de: mmctx_stop */
+	0xabc81b0e,
+	0x10b4b600,
+	0xf00cb9f0,
+	0x8bd012b9,
+/* 0x01ed: mmctx_stop_wait */
+	0x008bcf00,
+	0xf412bbc8,
+/* 0x01f6: mmctx_done */
+	0x87f1fa1b,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00199,
+/* 0x0207: strand_wait */
+	0xf900f800,
+	0x02a7f0a0,
+	0xfcc921f4,
+/* 0x0213: strand_pre */
+	0xf100f8a0,
+	0xf04afc87,
+	0x97f00283,
+	0x0089d00c,
+	0x020721f5,
+/* 0x0226: strand_post */
+	0x87f100f8,
+	0x83f04afc,
+	0x0d97f002,
+	0xf50089d0,
+	0xf8020721,
+/* 0x0239: strand_set */
+	0xfca7f100,
+	0x02a3f04f,
+	0x0500aba2,
+	0xd00fc7f0,
+	0xc7f000ac,
+	0x00bcd00b,
+	0x020721f5,
+	0xf000aed0,
+	0xbcd00ac7,
+	0x0721f500,
+/* 0x0263: strand_ctx_init */
+	0xf100f802,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00399f0,
+	0x21f50089,
+	0xe7f00213,
+	0x3921f503,
+	0xfca7f102,
+	0x02a3f046,
+	0x0400aba0,
+	0xf040a0d0,
+	0xbcd001c7,
+	0x0721f500,
+	0x010c9202,
+	0xf000acd0,
+	0xbcd002c7,
+	0x0721f500,
+	0x2621f502,
+	0x8087f102,
+	0x0684b608,
+	0xb70089cf,
+	0x95220080,
+/* 0x02ba: ctx_init_strand_loop */
+	0x8ed008fe,
+	0x408ed000,
+	0xb6808acf,
+	0xa0b606a5,
+	0x00eabb01,
+	0xb60480b6,
+	0x1bf40192,
+	0x08e4b6e8,
+	0xf1f2efbc,
+	0xb6085c87,
+	0x94bd0684,
+	0xd00399f0,
+	0x00f80089,
+/* 0x02ec: error */
+	0xe7f1e0f9,
+	0xe3f09814,
+	0x8d21f440,
+	0x041ce0b7,
+	0xf401f7f0,
+	0xe0fc8d21,
+/* 0x0306: init */
+	0x04bd00f8,
+	0xf10004fe,
+	0xf0120017,
+	0x12d00227,
+	0x3e17f100,
+	0x0010fe04,
+	0x040017f1,
+	0xf0c010d0,
+	0x12d00427,
+	0x1031f400,
+	0x060817f1,
+	0xcf0614b6,
+	0x37f00012,
+	0x1f24f001,
+	0xb60432bb,
+	0x02800132,
+	0x04038003,
+	0x040010b7,
+	0x800012cf,
+	0x27f10002,
+	0x24b60800,
+	0x0022cf06,
+/* 0x035f: init_find_chipset */
+	0xb65817f0,
+	0x13980c10,
+	0x0432b800,
+	0xb00b0bf4,
+	0x1bf40034,
+/* 0x0373: init_context */
+	0xf100f8f1,
+	0xb6080027,
+	0x22cf0624,
+	0xf134bd40,
+	0xb6070047,
+	0x25950644,
+	0x0045d008,
+	0xbd4045d0,
+	0x58f4bde4,
+	0x1f58021e,
+	0x020e4003,
+	0xf5040f40,
+	0xbb013d21,
+	0x3fbb002f,
+	0x041e5800,
+	0x40051f58,
+	0x0f400a0e,
+	0x3d21f50c,
+	0x030e9801,
+	0xbb00effd,
+	0x3ebb002e,
+	0x0040b700,
+	0x0235b613,
+	0xb60043d0,
+	0x35b60825,
+	0x0120b606,
+	0xb60130b6,
+	0x34b60824,
+	0x022fb908,
+	0x026321f5,
+	0xf1003fbb,
+	0xb6080017,
+	0x13d00614,
+	0x0010b740,
+	0xf024bd08,
+	0x12d01f29,
+/* 0x0401: main */
+	0x0031f400,
+	0xf00028f4,
+	0x21f41cd7,
+	0xf401f439,
+	0xf404e4b0,
+	0x81fe1e18,
+	0x0627f001,
+	0x12fd20bd,
+	0x01e4b604,
+	0xfe051efd,
+	0x21f50018,
+	0x0ef404c3,
+/* 0x0431: main_not_ctx_xfer */
+	0x10ef94d3,
+	0xf501f5f0,
+	0xf402ec21,
+/* 0x043e: ih */
+	0x80f9c60e,
+	0xf90188fe,
+	0xf990f980,
+	0xf9b0f9a0,
+	0xf9e0f9d0,
+	0x800acff0,
+	0xf404abc4,
+	0xb7f11d0b,
+	0xd7f01900,
+	0x40becf1c,
+	0xf400bfcf,
+	0xb0b70421,
+	0xe7f00400,
+	0x00bed001,
+/* 0x0474: ih_no_fifo */
+	0xfc400ad0,
+	0xfce0fcf0,
+	0xfcb0fcd0,
+	0xfc90fca0,
+	0x0088fe80,
+	0x32f480fc,
+/* 0x048f: hub_barrier_done */
+	0xf001f800,
+	0x0e9801f7,
+	0x04febb00,
+	0x9418e7f1,
+	0xf440e3f0,
+	0x00f88d21,
+/* 0x04a4: ctx_redswitch */
+	0x0614e7f1,
+	0xf006e4b6,
+	0xefd020f7,
+	0x08f7f000,
+/* 0x04b4: ctx_redswitch_delay */
+	0xf401f2b6,
+	0xf7f1fd1b,
+	0xefd00a20,
+/* 0x04c3: ctx_xfer */
+	0xf100f800,
+	0xb60a0417,
+	0x1fd00614,
+	0x0711f400,
+	0x04a421f5,
+/* 0x04d4: ctx_xfer_not_load */
+	0x4afc17f1,
+	0xf00213f0,
+	0x12d00c27,
+	0x0721f500,
+	0xfc27f102,
+	0x0223f047,
+	0xf00020d0,
+	0x20b6012c,
+	0x0012d003,
+	0xf001acf0,
+	0xb7f002a5,
+	0x50b3f000,
+	0xb6000c98,
+	0xbcbb0fc4,
+	0x010c9800,
+	0xf0020d98,
+	0x21f500e7,
+	0xacf0015c,
+	0x04a5f001,
+	0x4000b7f1,
+	0x9850b3f0,
+	0xc4b6000c,
+	0x00bcbb0f,
+	0x98050c98,
+	0x0f98060d,
+	0x00e7f104,
+	0x5c21f508,
+	0x0721f501,
+	0x0601f402,
+/* 0x054b: ctx_xfer_post */
+	0xf11412f4,
+	0xf04afc17,
+	0x27f00213,
+	0x0012d00d,
+	0x020721f5,
+/* 0x055c: ctx_xfer_done */
+	0x048f21f5,
+	0x000000f8,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc
new file mode 100644
index 0000000..7b715fd
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc
@@ -0,0 +1,451 @@
+/* fuc microcode for nve0 PGRAPH/GPC
+ *
+ * Copyright 2011 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+/* To build:
+ *    m4 nve0_grgpc.fuc | envyas -a -w -m fuc -V nva3 -o nve0_grgpc.fuc.h
+ */
+
+/* TODO
+ * - bracket certain functions with scratch writes, useful for debugging
+ * - watchdog timer around ctx operations
+ */
+
+.section #nve0_grgpc_data
+include(`nve0.fuc')
+gpc_id:			.b32 0
+gpc_mmio_list_head:	.b32 0
+gpc_mmio_list_tail:	.b32 0
+
+tpc_count:		.b32 0
+tpc_mask:		.b32 0
+tpc_mmio_list_head:	.b32 0
+tpc_mmio_list_tail:	.b32 0
+
+cmd_queue:		queue_init
+
+// chipset descriptions
+chipsets:
+.b8  0xe4 0 0 0
+.b16 #nve4_gpc_mmio_head
+.b16 #nve4_gpc_mmio_tail
+.b16 #nve4_tpc_mmio_head
+.b16 #nve4_tpc_mmio_tail
+.b8  0xe7 0 0 0
+.b16 #nve4_gpc_mmio_head
+.b16 #nve4_gpc_mmio_tail
+.b16 #nve4_tpc_mmio_head
+.b16 #nve4_tpc_mmio_tail
+.b8  0 0 0 0
+
+// GPC mmio lists
+nve4_gpc_mmio_head:
+mmctx_data(0x000380, 1)
+mmctx_data(0x000400, 2)
+mmctx_data(0x00040c, 3)
+mmctx_data(0x000450, 9)
+mmctx_data(0x000600, 1)
+mmctx_data(0x000684, 1)
+mmctx_data(0x000700, 5)
+mmctx_data(0x000800, 1)
+mmctx_data(0x000808, 3)
+mmctx_data(0x000828, 1)
+mmctx_data(0x000830, 1)
+mmctx_data(0x0008d8, 1)
+mmctx_data(0x0008e0, 1)
+mmctx_data(0x0008e8, 6)
+mmctx_data(0x00091c, 1)
+mmctx_data(0x000924, 3)
+mmctx_data(0x000b00, 1)
+mmctx_data(0x000b08, 6)
+mmctx_data(0x000bb8, 1)
+mmctx_data(0x000c08, 1)
+mmctx_data(0x000c10, 8)
+mmctx_data(0x000c40, 1)
+mmctx_data(0x000c6c, 1)
+mmctx_data(0x000c80, 1)
+mmctx_data(0x000c8c, 1)
+mmctx_data(0x001000, 3)
+mmctx_data(0x001014, 1)
+mmctx_data(0x003024, 1)
+mmctx_data(0x0030c0, 2)
+mmctx_data(0x0030e4, 1)
+mmctx_data(0x003100, 6)
+mmctx_data(0x0031d0, 1)
+mmctx_data(0x0031e0, 2)
+nve4_gpc_mmio_tail:
+
+// TPC mmio lists
+nve4_tpc_mmio_head:
+mmctx_data(0x000048, 1)
+mmctx_data(0x000064, 1)
+mmctx_data(0x000088, 1)
+mmctx_data(0x000200, 6)
+mmctx_data(0x00021c, 2)
+mmctx_data(0x000230, 1)
+mmctx_data(0x0002c4, 1)
+mmctx_data(0x000400, 3)
+mmctx_data(0x000420, 3)
+mmctx_data(0x0004e8, 1)
+mmctx_data(0x0004f4, 1)
+mmctx_data(0x000604, 4)
+mmctx_data(0x000644, 22)
+mmctx_data(0x0006ac, 2)
+mmctx_data(0x0006c8, 1)
+mmctx_data(0x000730, 8)
+mmctx_data(0x000758, 1)
+mmctx_data(0x000778, 1)
+nve4_tpc_mmio_tail:
+
+.section #nve0_grgpc_code
+bra #init
+define(`include_code')
+include(`nve0.fuc')
+
+// reports an exception to the host
+//
+// In: $r15 error code (see nve0.fuc)
+//
+error:
+	push $r14
+	mov $r14 -0x67ec 	// 0x9814
+	sethi $r14 0x400000
+	call #nv_wr32		// HUB_CTXCTL_CC_SCRATCH[5] = error code
+	add b32 $r14 0x41c
+	mov $r15 1
+	call #nv_wr32		// HUB_CTXCTL_INTR_UP_SET
+	pop $r14
+	ret
+
+// GPC fuc initialisation, executed by triggering ucode start, will
+// fall through to main loop after completion.
+//
+// Input:
+//   CC_SCRATCH[0]: chipset (PMC_BOOT_0 read returns 0x0bad0bad... sigh)
+//   CC_SCRATCH[1]: context base
+//
+// Output:
+//   CC_SCRATCH[0]:
+//	     31:31: set to signal completion
+//   CC_SCRATCH[1]:
+//	      31:0: GPC context size
+//
+init:
+	clear b32 $r0
+	mov $sp $r0
+
+	// enable fifo access
+	mov $r1 0x1200
+	mov $r2 2
+	iowr I[$r1 + 0x000] $r2		// FIFO_ENABLE
+
+	// setup i0 handler, and route all interrupts to it
+	mov $r1 #ih
+	mov $iv0 $r1
+	mov $r1 0x400
+	iowr I[$r1 + 0x300] $r0		// INTR_DISPATCH
+
+	// enable fifo interrupt
+	mov $r2 4
+	iowr I[$r1 + 0x000] $r2		// INTR_EN_SET
+
+	// enable interrupts
+	bset $flags ie0
+
+	// figure out which GPC we are, and how many TPCs we have
+	mov $r1 0x608
+	shl b32 $r1 6
+	iord $r2 I[$r1 + 0x000]		// UNITS
+	mov $r3 1
+	and $r2 0x1f
+	shl b32 $r3 $r2
+	sub b32 $r3 1
+	st b32 D[$r0 + #tpc_count] $r2
+	st b32 D[$r0 + #tpc_mask] $r3
+	add b32 $r1 0x400
+	iord $r2 I[$r1 + 0x000]		// MYINDEX
+	st b32 D[$r0 + #gpc_id] $r2
+
+	// find context data for this chipset
+	mov $r2 0x800
+	shl b32 $r2 6
+	iord $r2 I[$r2 + 0x000]		// CC_SCRATCH[0]
+	mov $r1 #chipsets - 12
+	init_find_chipset:
+		add b32 $r1 12
+		ld b32 $r3 D[$r1 + 0x00]
+		cmpu b32 $r3 $r2
+		bra e #init_context
+		cmpu b32 $r3 0
+		bra ne #init_find_chipset
+		// unknown chipset
+		ret
+
+	// initialise context base, and size tracking
+	init_context:
+	mov $r2 0x800
+	shl b32 $r2 6
+	iord $r2 I[$r2 + 0x100]	// CC_SCRATCH[1], initial base
+	clear b32 $r3		// track GPC context size here
+
+	// set mmctx base addresses now so we don't have to do it later,
+	// they don't currently ever change
+	mov $r4 0x700
+	shl b32 $r4 6
+	shr b32 $r5 $r2 8
+	iowr I[$r4 + 0x000] $r5		// MMCTX_SAVE_SWBASE
+	iowr I[$r4 + 0x100] $r5		// MMCTX_LOAD_SWBASE
+
+	// calculate GPC mmio context size, store the chipset-specific
+	// mmio list pointers somewhere we can get at them later without
+	// re-parsing the chipset list
+	clear b32 $r14
+	clear b32 $r15
+	ld b16 $r14 D[$r1 + 4]
+	ld b16 $r15 D[$r1 + 6]
+	st b16 D[$r0 + #gpc_mmio_list_head] $r14
+	st b16 D[$r0 + #gpc_mmio_list_tail] $r15
+	call #mmctx_size
+	add b32 $r2 $r15
+	add b32 $r3 $r15
+
+	// calculate per-TPC mmio context size, store the list pointers
+	ld b16 $r14 D[$r1 + 8]
+	ld b16 $r15 D[$r1 + 10]
+	st b16 D[$r0 + #tpc_mmio_list_head] $r14
+	st b16 D[$r0 + #tpc_mmio_list_tail] $r15
+	call #mmctx_size
+	ld b32 $r14 D[$r0 + #tpc_count]
+	mulu $r14 $r15
+	add b32 $r2 $r14
+	add b32 $r3 $r14
+
+	// round up base/size to 256 byte boundary (for strand SWBASE)
+	add b32 $r4 0x1300
+	shr b32 $r3 2
+	iowr I[$r4 + 0x000] $r3		// MMCTX_LOAD_COUNT, wtf for?!?
+	shr b32 $r2 8
+	shr b32 $r3 6
+	add b32 $r2 1
+	add b32 $r3 1
+	shl b32 $r2 8
+	shl b32 $r3 8
+
+	// calculate size of strand context data
+	mov b32 $r15 $r2
+	call #strand_ctx_init
+	add b32 $r3 $r15
+
+	// save context size, and tell HUB we're done
+	mov $r1 0x800
+	shl b32 $r1 6
+	iowr I[$r1 + 0x100] $r3		// CC_SCRATCH[1]  = context size
+	add b32 $r1 0x800
+	clear b32 $r2
+	bset $r2 31
+	iowr I[$r1 + 0x000] $r2		// CC_SCRATCH[0] |= 0x80000000
+
+// Main program loop, very simple, sleeps until woken up by the interrupt
+// handler, pulls a command from the queue and executes its handler
+//
+main:
+	bset $flags $p0
+	sleep $p0
+	mov $r13 #cmd_queue
+	call #queue_get
+	bra $p1 #main
+
+	// 0x0000-0x0003 are all context transfers
+	cmpu b32 $r14 0x04
+	bra nc #main_not_ctx_xfer
+		// fetch $flags and mask off $p1/$p2
+		mov $r1 $flags
+		mov $r2 0x0006
+		not b32 $r2
+		and $r1 $r2
+		// set $p1/$p2 according to transfer type
+		shl b32 $r14 1
+		or $r1 $r14
+		mov $flags $r1
+		// transfer context data
+		call #ctx_xfer
+		bra #main
+
+	main_not_ctx_xfer:
+	shl b32 $r15 $r14 16
+	or $r15 E_BAD_COMMAND
+	call #error
+	bra #main
+
+// interrupt handler
+ih:
+	push $r8
+	mov $r8 $flags
+	push $r8
+	push $r9
+	push $r10
+	push $r11
+	push $r13
+	push $r14
+	push $r15
+
+	// incoming fifo command?
+	iord $r10 I[$r0 + 0x200]	// INTR
+	and $r11 $r10 0x00000004
+	bra e #ih_no_fifo
+		// queue incoming fifo command for later processing
+		mov $r11 0x1900
+		mov $r13 #cmd_queue
+		iord $r14 I[$r11 + 0x100]	// FIFO_CMD
+		iord $r15 I[$r11 + 0x000]	// FIFO_DATA
+		call #queue_put
+		add b32 $r11 0x400
+		mov $r14 1
+		iowr I[$r11 + 0x000] $r14	// FIFO_ACK
+
+	// ack, and wake up main()
+	ih_no_fifo:
+	iowr I[$r0 + 0x100] $r10	// INTR_ACK
+
+	pop $r15
+	pop $r14
+	pop $r13
+	pop $r11
+	pop $r10
+	pop $r9
+	pop $r8
+	mov $flags $r8
+	pop $r8
+	bclr $flags $p0
+	iret
+
+// Set this GPC's bit in HUB_BAR, used to signal completion of various
+// activities to the HUB fuc
+//
+hub_barrier_done:
+	mov $r15 1
+	ld b32 $r14 D[$r0 + #gpc_id]
+	shl b32 $r15 $r14
+	mov $r14 -0x6be8 	// 0x409418 - HUB_BAR_SET
+	sethi $r14 0x400000
+	call #nv_wr32
+	ret
+
+// Disables various things, waits a bit, and re-enables them..
+//
+// Not sure how exactly this helps, perhaps "ENABLE" is not such a
+// good description for the bits we turn off?  Anyways, without this,
+// funny things happen.
+//
+ctx_redswitch:
+	mov $r14 0x614
+	shl b32 $r14 6
+	mov $r15 0x020
+	iowr I[$r14] $r15	// GPC_RED_SWITCH = POWER
+	mov $r15 8
+	ctx_redswitch_delay:
+		sub b32 $r15 1
+		bra ne #ctx_redswitch_delay
+	mov $r15 0xa20
+	iowr I[$r14] $r15	// GPC_RED_SWITCH = UNK11, ENABLE, POWER
+	ret
+
+// Transfer GPC context data between GPU and storage area
+//
+// In: $r15 context base address
+//     $p1 clear on save, set on load
+//     $p2 set if opposite direction done/will be done, so:
+//		on save it means: "a load will follow this save"
+//		on load it means: "a save preceeded this load"
+//
+ctx_xfer:
+	// set context base address
+	mov $r1 0xa04
+	shl b32 $r1 6
+	iowr I[$r1 + 0x000] $r15// MEM_BASE
+	bra not $p1 #ctx_xfer_not_load
+		call #ctx_redswitch
+	ctx_xfer_not_load:
+
+	// strands
+	mov $r1 0x4afc
+	sethi $r1 0x20000
+	mov $r2 0xc
+	iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x0c
+	call #strand_wait
+	mov $r2 0x47fc
+	sethi $r2 0x20000
+	iowr I[$r2] $r0		// STRAND_FIRST_GENE(0x3f) = 0x00
+	xbit $r2 $flags $p1
+	add b32 $r2 3
+	iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x03/0x04 (SAVE/LOAD)
+
+	// mmio context
+	xbit $r10 $flags $p1	// direction
+	or $r10 2		// first
+	mov $r11 0x0000
+	sethi $r11 0x500000
+	ld b32 $r12 D[$r0 + #gpc_id]
+	shl b32 $r12 15
+	add b32 $r11 $r12	// base = NV_PGRAPH_GPCn
+	ld b32 $r12 D[$r0 + #gpc_mmio_list_head]
+	ld b32 $r13 D[$r0 + #gpc_mmio_list_tail]
+	mov $r14 0		// not multi
+	call #mmctx_xfer
+
+	// per-TPC mmio context
+	xbit $r10 $flags $p1	// direction
+	or $r10 4		// last
+	mov $r11 0x4000
+	sethi $r11 0x500000	// base = NV_PGRAPH_GPC0_TPC0
+	ld b32 $r12 D[$r0 + #gpc_id]
+	shl b32 $r12 15
+	add b32 $r11 $r12	// base = NV_PGRAPH_GPCn_TPC0
+	ld b32 $r12 D[$r0 + #tpc_mmio_list_head]
+	ld b32 $r13 D[$r0 + #tpc_mmio_list_tail]
+	ld b32 $r15 D[$r0 + #tpc_mask]
+	mov $r14 0x800		// stride = 0x800
+	call #mmctx_xfer
+
+	// wait for strands to finish
+	call #strand_wait
+
+	// if load, or a save without a load following, do some
+	// unknown stuff that's done after finishing a block of
+	// strand commands
+	bra $p1 #ctx_xfer_post
+	bra not $p2 #ctx_xfer_done
+	ctx_xfer_post:
+		mov $r1 0x4afc
+		sethi $r1 0x20000
+		mov $r2 0xd
+		iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x0d
+		call #strand_wait
+
+	// mark completion in HUB's barrier
+	ctx_xfer_done:
+	call #hub_barrier_done
+	ret
+
+.align 256
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc.h b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc.h
new file mode 100644
index 0000000..26c2165
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc.h
@@ -0,0 +1,530 @@
+uint32_t nve0_grgpc_data[] = {
+/* 0x0000: gpc_id */
+	0x00000000,
+/* 0x0004: gpc_mmio_list_head */
+	0x00000000,
+/* 0x0008: gpc_mmio_list_tail */
+	0x00000000,
+/* 0x000c: tpc_count */
+	0x00000000,
+/* 0x0010: tpc_mask */
+	0x00000000,
+/* 0x0014: tpc_mmio_list_head */
+	0x00000000,
+/* 0x0018: tpc_mmio_list_tail */
+	0x00000000,
+/* 0x001c: cmd_queue */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0064: chipsets */
+	0x000000e4,
+	0x01040080,
+	0x014c0104,
+	0x000000e7,
+	0x01040080,
+	0x014c0104,
+	0x00000000,
+/* 0x0080: nve4_gpc_mmio_head */
+	0x00000380,
+	0x04000400,
+	0x0800040c,
+	0x20000450,
+	0x00000600,
+	0x00000684,
+	0x10000700,
+	0x00000800,
+	0x08000808,
+	0x00000828,
+	0x00000830,
+	0x000008d8,
+	0x000008e0,
+	0x140008e8,
+	0x0000091c,
+	0x08000924,
+	0x00000b00,
+	0x14000b08,
+	0x00000bb8,
+	0x00000c08,
+	0x1c000c10,
+	0x00000c40,
+	0x00000c6c,
+	0x00000c80,
+	0x00000c8c,
+	0x08001000,
+	0x00001014,
+	0x00003024,
+	0x040030c0,
+	0x000030e4,
+	0x14003100,
+	0x000031d0,
+	0x040031e0,
+/* 0x0104: nve4_gpc_mmio_tail */
+/* 0x0104: nve4_tpc_mmio_head */
+	0x00000048,
+	0x00000064,
+	0x00000088,
+	0x14000200,
+	0x0400021c,
+	0x00000230,
+	0x000002c4,
+	0x08000400,
+	0x08000420,
+	0x000004e8,
+	0x000004f4,
+	0x0c000604,
+	0x54000644,
+	0x040006ac,
+	0x000006c8,
+	0x1c000730,
+	0x00000758,
+	0x00000778,
+};
+
+uint32_t nve0_grgpc_code[] = {
+	0x03060ef5,
+/* 0x0004: queue_put */
+	0x9800d898,
+	0x86f001d9,
+	0x0489b808,
+	0xf00c1bf4,
+	0x21f502f7,
+	0x00f802ec,
+/* 0x001c: queue_put_next */
+	0xb60798c4,
+	0x8dbb0384,
+	0x0880b600,
+	0x80008e80,
+	0x90b6018f,
+	0x0f94f001,
+	0xf801d980,
+/* 0x0039: queue_get */
+	0x0131f400,
+	0x9800d898,
+	0x89b801d9,
+	0x210bf404,
+	0xb60789c4,
+	0x9dbb0394,
+	0x0890b600,
+	0x98009e98,
+	0x80b6019f,
+	0x0f84f001,
+	0xf400d880,
+/* 0x0066: queue_get_done */
+	0x00f80132,
+/* 0x0068: nv_rd32 */
+	0x0728b7f1,
+	0xb906b4b6,
+	0xc9f002ec,
+	0x00bcd01f,
+/* 0x0078: nv_rd32_wait */
+	0xc800bccf,
+	0x1bf41fcc,
+	0x06a7f0fa,
+	0x010321f5,
+	0xf840bfcf,
+/* 0x008d: nv_wr32 */
+	0x28b7f100,
+	0x06b4b607,
+	0xb980bfd0,
+	0xc9f002ec,
+	0x1ec9f01f,
+/* 0x00a3: nv_wr32_wait */
+	0xcf00bcd0,
+	0xccc800bc,
+	0xfa1bf41f,
+/* 0x00ae: watchdog_reset */
+	0x87f100f8,
+	0x84b60430,
+	0x1ff9f006,
+	0xf8008fd0,
+/* 0x00bd: watchdog_clear */
+	0x3087f100,
+	0x0684b604,
+	0xf80080d0,
+/* 0x00c9: wait_donez */
+	0x3c87f100,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d000,
+	0x081887f1,
+	0xd00684b6,
+/* 0x00e2: wait_done_wait_donez */
+	0x87f1008a,
+	0x84b60400,
+	0x0088cf06,
+	0xf4888aff,
+	0x87f1f31b,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00099,
+/* 0x0103: wait_doneo */
+	0xf100f800,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00099f0,
+	0x87f10089,
+	0x84b60818,
+	0x008ad006,
+/* 0x011c: wait_done_wait_doneo */
+	0x040087f1,
+	0xcf0684b6,
+	0x8aff0088,
+	0xf30bf488,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0099f094,
+	0xf80089d0,
+/* 0x013d: mmctx_size */
+/* 0x013f: nv_mmctx_size_loop */
+	0x9894bd00,
+	0x85b600e8,
+	0x0180b61a,
+	0xbb0284b6,
+	0xe0b60098,
+	0x04efb804,
+	0xb9eb1bf4,
+	0x00f8029f,
+/* 0x015c: mmctx_xfer */
+	0x083c87f1,
+	0xbd0684b6,
+	0x0199f094,
+	0xf10089d0,
+	0xb6071087,
+	0x94bd0684,
+	0xf405bbfd,
+	0x8bd0090b,
+	0x0099f000,
+/* 0x0180: mmctx_base_disabled */
+	0xf405eefd,
+	0x8ed00c0b,
+	0xc08fd080,
+/* 0x018f: mmctx_multi_disabled */
+	0xb70199f0,
+	0xc8010080,
+	0xb4b600ab,
+	0x0cb9f010,
+	0xb601aec8,
+	0xbefd11e4,
+	0x008bd005,
+/* 0x01a8: mmctx_exec_loop */
+/* 0x01a8: mmctx_wait_free */
+	0xf0008ecf,
+	0x0bf41fe4,
+	0x00ce98fa,
+	0xd005e9fd,
+	0xc0b6c08e,
+	0x04cdb804,
+	0xc8e81bf4,
+	0x1bf402ab,
+/* 0x01c9: mmctx_fini_wait */
+	0x008bcf18,
+	0xb01fb4f0,
+	0x1bf410b4,
+	0x02a7f0f7,
+	0xf4c921f4,
+/* 0x01de: mmctx_stop */
+	0xabc81b0e,
+	0x10b4b600,
+	0xf00cb9f0,
+	0x8bd012b9,
+/* 0x01ed: mmctx_stop_wait */
+	0x008bcf00,
+	0xf412bbc8,
+/* 0x01f6: mmctx_done */
+	0x87f1fa1b,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00199,
+/* 0x0207: strand_wait */
+	0xf900f800,
+	0x02a7f0a0,
+	0xfcc921f4,
+/* 0x0213: strand_pre */
+	0xf100f8a0,
+	0xf04afc87,
+	0x97f00283,
+	0x0089d00c,
+	0x020721f5,
+/* 0x0226: strand_post */
+	0x87f100f8,
+	0x83f04afc,
+	0x0d97f002,
+	0xf50089d0,
+	0xf8020721,
+/* 0x0239: strand_set */
+	0xfca7f100,
+	0x02a3f04f,
+	0x0500aba2,
+	0xd00fc7f0,
+	0xc7f000ac,
+	0x00bcd00b,
+	0x020721f5,
+	0xf000aed0,
+	0xbcd00ac7,
+	0x0721f500,
+/* 0x0263: strand_ctx_init */
+	0xf100f802,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00399f0,
+	0x21f50089,
+	0xe7f00213,
+	0x3921f503,
+	0xfca7f102,
+	0x02a3f046,
+	0x0400aba0,
+	0xf040a0d0,
+	0xbcd001c7,
+	0x0721f500,
+	0x010c9202,
+	0xf000acd0,
+	0xbcd002c7,
+	0x0721f500,
+	0x2621f502,
+	0x8087f102,
+	0x0684b608,
+	0xb70089cf,
+	0x95220080,
+/* 0x02ba: ctx_init_strand_loop */
+	0x8ed008fe,
+	0x408ed000,
+	0xb6808acf,
+	0xa0b606a5,
+	0x00eabb01,
+	0xb60480b6,
+	0x1bf40192,
+	0x08e4b6e8,
+	0xf1f2efbc,
+	0xb6085c87,
+	0x94bd0684,
+	0xd00399f0,
+	0x00f80089,
+/* 0x02ec: error */
+	0xe7f1e0f9,
+	0xe3f09814,
+	0x8d21f440,
+	0x041ce0b7,
+	0xf401f7f0,
+	0xe0fc8d21,
+/* 0x0306: init */
+	0x04bd00f8,
+	0xf10004fe,
+	0xf0120017,
+	0x12d00227,
+	0x3e17f100,
+	0x0010fe04,
+	0x040017f1,
+	0xf0c010d0,
+	0x12d00427,
+	0x1031f400,
+	0x060817f1,
+	0xcf0614b6,
+	0x37f00012,
+	0x1f24f001,
+	0xb60432bb,
+	0x02800132,
+	0x04038003,
+	0x040010b7,
+	0x800012cf,
+	0x27f10002,
+	0x24b60800,
+	0x0022cf06,
+/* 0x035f: init_find_chipset */
+	0xb65817f0,
+	0x13980c10,
+	0x0432b800,
+	0xb00b0bf4,
+	0x1bf40034,
+/* 0x0373: init_context */
+	0xf100f8f1,
+	0xb6080027,
+	0x22cf0624,
+	0xf134bd40,
+	0xb6070047,
+	0x25950644,
+	0x0045d008,
+	0xbd4045d0,
+	0x58f4bde4,
+	0x1f58021e,
+	0x020e4003,
+	0xf5040f40,
+	0xbb013d21,
+	0x3fbb002f,
+	0x041e5800,
+	0x40051f58,
+	0x0f400a0e,
+	0x3d21f50c,
+	0x030e9801,
+	0xbb00effd,
+	0x3ebb002e,
+	0x0040b700,
+	0x0235b613,
+	0xb60043d0,
+	0x35b60825,
+	0x0120b606,
+	0xb60130b6,
+	0x34b60824,
+	0x022fb908,
+	0x026321f5,
+	0xf1003fbb,
+	0xb6080017,
+	0x13d00614,
+	0x0010b740,
+	0xf024bd08,
+	0x12d01f29,
+/* 0x0401: main */
+	0x0031f400,
+	0xf00028f4,
+	0x21f41cd7,
+	0xf401f439,
+	0xf404e4b0,
+	0x81fe1e18,
+	0x0627f001,
+	0x12fd20bd,
+	0x01e4b604,
+	0xfe051efd,
+	0x21f50018,
+	0x0ef404c3,
+/* 0x0431: main_not_ctx_xfer */
+	0x10ef94d3,
+	0xf501f5f0,
+	0xf402ec21,
+/* 0x043e: ih */
+	0x80f9c60e,
+	0xf90188fe,
+	0xf990f980,
+	0xf9b0f9a0,
+	0xf9e0f9d0,
+	0x800acff0,
+	0xf404abc4,
+	0xb7f11d0b,
+	0xd7f01900,
+	0x40becf1c,
+	0xf400bfcf,
+	0xb0b70421,
+	0xe7f00400,
+	0x00bed001,
+/* 0x0474: ih_no_fifo */
+	0xfc400ad0,
+	0xfce0fcf0,
+	0xfcb0fcd0,
+	0xfc90fca0,
+	0x0088fe80,
+	0x32f480fc,
+/* 0x048f: hub_barrier_done */
+	0xf001f800,
+	0x0e9801f7,
+	0x04febb00,
+	0x9418e7f1,
+	0xf440e3f0,
+	0x00f88d21,
+/* 0x04a4: ctx_redswitch */
+	0x0614e7f1,
+	0xf006e4b6,
+	0xefd020f7,
+	0x08f7f000,
+/* 0x04b4: ctx_redswitch_delay */
+	0xf401f2b6,
+	0xf7f1fd1b,
+	0xefd00a20,
+/* 0x04c3: ctx_xfer */
+	0xf100f800,
+	0xb60a0417,
+	0x1fd00614,
+	0x0711f400,
+	0x04a421f5,
+/* 0x04d4: ctx_xfer_not_load */
+	0x4afc17f1,
+	0xf00213f0,
+	0x12d00c27,
+	0x0721f500,
+	0xfc27f102,
+	0x0223f047,
+	0xf00020d0,
+	0x20b6012c,
+	0x0012d003,
+	0xf001acf0,
+	0xb7f002a5,
+	0x50b3f000,
+	0xb6000c98,
+	0xbcbb0fc4,
+	0x010c9800,
+	0xf0020d98,
+	0x21f500e7,
+	0xacf0015c,
+	0x04a5f001,
+	0x4000b7f1,
+	0x9850b3f0,
+	0xc4b6000c,
+	0x00bcbb0f,
+	0x98050c98,
+	0x0f98060d,
+	0x00e7f104,
+	0x5c21f508,
+	0x0721f501,
+	0x0601f402,
+/* 0x054b: ctx_xfer_post */
+	0xf11412f4,
+	0xf04afc17,
+	0x27f00213,
+	0x0012d00d,
+	0x020721f5,
+/* 0x055c: ctx_xfer_done */
+	0x048f21f5,
+	0x000000f8,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnvc0.fuc b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnvc0.fuc
new file mode 100644
index 0000000..acfc457
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnvc0.fuc
@@ -0,0 +1,856 @@
+/* fuc microcode for nvc0 PGRAPH/HUB
+ *
+ * Copyright 2011 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+/* To build:
+ *    m4 hubnvc0.fuc | envyas -a -w -m fuc -V fuc3 -o hubnvc0.fuc.h
+ */
+
+.section #nvc0_grhub_data
+include(`nvc0.fuc')
+gpc_count:		.b32 0
+rop_count:		.b32 0
+cmd_queue:		queue_init
+hub_mmio_list_head:	.b32 0
+hub_mmio_list_tail:	.b32 0
+
+ctx_current:		.b32 0
+
+chipsets:
+.b8  0xc0 0 0 0
+.b16 #nvc0_hub_mmio_head
+.b16 #nvc0_hub_mmio_tail
+.b8  0xc1 0 0 0
+.b16 #nvc0_hub_mmio_head
+.b16 #nvc1_hub_mmio_tail
+.b8  0xc3 0 0 0
+.b16 #nvc0_hub_mmio_head
+.b16 #nvc0_hub_mmio_tail
+.b8  0xc4 0 0 0
+.b16 #nvc0_hub_mmio_head
+.b16 #nvc0_hub_mmio_tail
+.b8  0xc8 0 0 0
+.b16 #nvc0_hub_mmio_head
+.b16 #nvc0_hub_mmio_tail
+.b8  0xce 0 0 0
+.b16 #nvc0_hub_mmio_head
+.b16 #nvc0_hub_mmio_tail
+.b8  0xcf 0 0 0
+.b16 #nvc0_hub_mmio_head
+.b16 #nvc0_hub_mmio_tail
+.b8  0xd9 0 0 0
+.b16 #nvd9_hub_mmio_head
+.b16 #nvd9_hub_mmio_tail
+.b8  0 0 0 0
+
+nvc0_hub_mmio_head:
+mmctx_data(0x17e91c, 2)
+mmctx_data(0x400204, 2)
+mmctx_data(0x404004, 11)
+mmctx_data(0x404044, 1)
+mmctx_data(0x404094, 14)
+mmctx_data(0x4040d0, 7)
+mmctx_data(0x4040f8, 1)
+mmctx_data(0x404130, 3)
+mmctx_data(0x404150, 3)
+mmctx_data(0x404164, 2)
+mmctx_data(0x404174, 3)
+mmctx_data(0x404200, 8)
+mmctx_data(0x404404, 14)
+mmctx_data(0x404460, 4)
+mmctx_data(0x404480, 1)
+mmctx_data(0x404498, 1)
+mmctx_data(0x404604, 4)
+mmctx_data(0x404618, 32)
+mmctx_data(0x404698, 21)
+mmctx_data(0x4046f0, 2)
+mmctx_data(0x404700, 22)
+mmctx_data(0x405800, 1)
+mmctx_data(0x405830, 3)
+mmctx_data(0x405854, 1)
+mmctx_data(0x405870, 4)
+mmctx_data(0x405a00, 2)
+mmctx_data(0x405a18, 1)
+mmctx_data(0x406020, 1)
+mmctx_data(0x406028, 4)
+mmctx_data(0x4064a8, 2)
+mmctx_data(0x4064b4, 2)
+mmctx_data(0x407804, 1)
+mmctx_data(0x40780c, 6)
+mmctx_data(0x4078bc, 1)
+mmctx_data(0x408000, 7)
+mmctx_data(0x408064, 1)
+mmctx_data(0x408800, 3)
+mmctx_data(0x408900, 4)
+mmctx_data(0x408980, 1)
+nvc0_hub_mmio_tail:
+mmctx_data(0x4064c0, 2)
+nvc1_hub_mmio_tail:
+
+nvd9_hub_mmio_head:
+mmctx_data(0x17e91c, 2)
+mmctx_data(0x400204, 2)
+mmctx_data(0x404004, 10)
+mmctx_data(0x404044, 1)
+mmctx_data(0x404094, 14)
+mmctx_data(0x4040d0, 7)
+mmctx_data(0x4040f8, 1)
+mmctx_data(0x404130, 3)
+mmctx_data(0x404150, 3)
+mmctx_data(0x404164, 2)
+mmctx_data(0x404178, 2)
+mmctx_data(0x404200, 8)
+mmctx_data(0x404404, 14)
+mmctx_data(0x404460, 4)
+mmctx_data(0x404480, 1)
+mmctx_data(0x404498, 1)
+mmctx_data(0x404604, 4)
+mmctx_data(0x404618, 32)
+mmctx_data(0x404698, 21)
+mmctx_data(0x4046f0, 2)
+mmctx_data(0x404700, 22)
+mmctx_data(0x405800, 1)
+mmctx_data(0x405830, 3)
+mmctx_data(0x405854, 1)
+mmctx_data(0x405870, 4)
+mmctx_data(0x405a00, 2)
+mmctx_data(0x405a18, 1)
+mmctx_data(0x406020, 1)
+mmctx_data(0x406028, 4)
+mmctx_data(0x4064a8, 2)
+mmctx_data(0x4064b4, 5)
+mmctx_data(0x407804, 1)
+mmctx_data(0x40780c, 6)
+mmctx_data(0x4078bc, 1)
+mmctx_data(0x408000, 7)
+mmctx_data(0x408064, 1)
+mmctx_data(0x408800, 3)
+mmctx_data(0x408900, 4)
+mmctx_data(0x408980, 1)
+nvd9_hub_mmio_tail:
+
+.align 256
+chan_data:
+chan_mmio_count:	.b32 0
+chan_mmio_address:	.b32 0
+
+.align 256
+xfer_data: 		.b32 0
+
+.section #nvc0_grhub_code
+bra #init
+define(`include_code')
+include(`nvc0.fuc')
+
+// reports an exception to the host
+//
+// In: $r15 error code (see nvc0.fuc)
+//
+error:
+	push $r14
+	mov $r14 0x814
+	shl b32 $r14 6
+	iowr I[$r14 + 0x000] $r15	// CC_SCRATCH[5] = error code
+	mov $r14 0xc1c
+	shl b32 $r14 6
+	mov $r15 1
+	iowr I[$r14 + 0x000] $r15	// INTR_UP_SET
+	pop $r14
+	ret
+
+// HUB fuc initialisation, executed by triggering ucode start, will
+// fall through to main loop after completion.
+//
+// Input:
+//   CC_SCRATCH[0]: chipset (PMC_BOOT_0 read returns 0x0bad0bad... sigh)
+//
+// Output:
+//   CC_SCRATCH[0]:
+//	     31:31: set to signal completion
+//   CC_SCRATCH[1]:
+//	      31:0: total PGRAPH context size
+//
+init:
+	clear b32 $r0
+	mov $sp $r0
+	mov $xdbase $r0
+
+	// enable fifo access
+	mov $r1 0x1200
+	mov $r2 2
+	iowr I[$r1 + 0x000] $r2	// FIFO_ENABLE
+
+	// setup i0 handler, and route all interrupts to it
+	mov $r1 #ih
+	mov $iv0 $r1
+	mov $r1 0x400
+	iowr I[$r1 + 0x300] $r0	// INTR_DISPATCH
+
+	// route HUB_CHANNEL_SWITCH to fuc interrupt 8
+	mov $r3 0x404
+	shl b32 $r3 6
+	mov $r2 0x2003		// { HUB_CHANNEL_SWITCH, ZERO } -> intr 8
+	iowr I[$r3 + 0x000] $r2
+
+	// not sure what these are, route them because NVIDIA does, and
+	// the IRQ handler will signal the host if we ever get one.. we
+	// may find out if/why we need to handle these if so..
+	//
+	mov $r2 0x2004
+	iowr I[$r3 + 0x004] $r2 // { 0x04, ZERO } -> intr 9
+	mov $r2 0x200b
+	iowr I[$r3 + 0x008] $r2 // { 0x0b, ZERO } -> intr 10
+	mov $r2 0x200c
+	iowr I[$r3 + 0x01c] $r2 // { 0x0c, ZERO } -> intr 15
+
+	// enable all INTR_UP interrupts
+	mov $r2 0xc24
+	shl b32 $r2 6
+	not b32 $r3 $r0
+	iowr I[$r2] $r3
+
+	// enable fifo, ctxsw, 9, 10, 15 interrupts
+	mov $r2 -0x78fc		// 0x8704
+	sethi $r2 0
+	iowr I[$r1 + 0x000] $r2	// INTR_EN_SET
+
+	// fifo level triggered, rest edge
+	sub b32 $r1 0x100
+	mov $r2 4
+	iowr I[$r1] $r2
+
+	// enable interrupts
+	bset $flags ie0
+
+	// fetch enabled GPC/ROP counts
+	mov $r14 -0x69fc	// 0x409604
+	sethi $r14 0x400000
+	call #nv_rd32
+	extr $r1 $r15 16:20
+	st b32 D[$r0 + #rop_count] $r1
+	and $r15 0x1f
+	st b32 D[$r0 + #gpc_count] $r15
+
+	// set BAR_REQMASK to GPC mask
+	mov $r1 1
+	shl b32 $r1 $r15
+	sub b32 $r1 1
+	mov $r2 0x40c
+	shl b32 $r2 6
+	iowr I[$r2 + 0x000] $r1
+	iowr I[$r2 + 0x100] $r1
+
+	// find context data for this chipset
+	mov $r2 0x800
+	shl b32 $r2 6
+	iord $r2 I[$r2 + 0x000]		// CC_SCRATCH[0]
+	mov $r15 #chipsets - 8
+	init_find_chipset:
+		add b32 $r15 8
+		ld b32 $r3 D[$r15 + 0x00]
+		cmpu b32 $r3 $r2
+		bra e #init_context
+		cmpu b32 $r3 0
+		bra ne #init_find_chipset
+		// unknown chipset
+		ret
+
+	// context size calculation, reserve first 256 bytes for use by fuc
+	init_context:
+	mov $r1 256
+
+	// calculate size of mmio context data
+	ld b16 $r14 D[$r15 + 4]
+	ld b16 $r15 D[$r15 + 6]
+	sethi $r14 0
+	st b32 D[$r0 + #hub_mmio_list_head] $r14
+	st b32 D[$r0 + #hub_mmio_list_tail] $r15
+	call #mmctx_size
+
+	// set mmctx base addresses now so we don't have to do it later,
+	// they don't (currently) ever change
+	mov $r3 0x700
+	shl b32 $r3 6
+	shr b32 $r4 $r1 8
+	iowr I[$r3 + 0x000] $r4		// MMCTX_SAVE_SWBASE
+	iowr I[$r3 + 0x100] $r4		// MMCTX_LOAD_SWBASE
+	add b32 $r3 0x1300
+	add b32 $r1 $r15
+	shr b32 $r15 2
+	iowr I[$r3 + 0x000] $r15	// MMCTX_LOAD_COUNT, wtf for?!?
+
+	// strands, base offset needs to be aligned to 256 bytes
+	shr b32 $r1 8
+	add b32 $r1 1
+	shl b32 $r1 8
+	mov b32 $r15 $r1
+	call #strand_ctx_init
+	add b32 $r1 $r15
+
+	// initialise each GPC in sequence by passing in the offset of its
+	// context data in GPCn_CC_SCRATCH[1], and starting its FUC (which
+	// has previously been uploaded by the host) running.
+	//
+	// the GPC fuc init sequence will set GPCn_CC_SCRATCH[0] bit 31
+	// when it has completed, and return the size of its context data
+	// in GPCn_CC_SCRATCH[1]
+	//
+	ld b32 $r3 D[$r0 + #gpc_count]
+	mov $r4 0x2000
+	sethi $r4 0x500000
+	init_gpc:
+		// setup, and start GPC ucode running
+		add b32 $r14 $r4 0x804
+		mov b32 $r15 $r1
+		call #nv_wr32			// CC_SCRATCH[1] = ctx offset
+		add b32 $r14 $r4 0x800
+		mov b32 $r15 $r2
+		call #nv_wr32			// CC_SCRATCH[0] = chipset
+		add b32 $r14 $r4 0x10c
+		clear b32 $r15
+		call #nv_wr32
+		add b32 $r14 $r4 0x104
+		call #nv_wr32			// ENTRY
+		add b32 $r14 $r4 0x100
+		mov $r15 2			// CTRL_START_TRIGGER
+		call #nv_wr32			// CTRL
+
+		// wait for it to complete, and adjust context size
+		add b32 $r14 $r4 0x800
+		init_gpc_wait:
+			call #nv_rd32
+			xbit $r15 $r15 31
+			bra e #init_gpc_wait
+		add b32 $r14 $r4 0x804
+		call #nv_rd32
+		add b32 $r1 $r15
+
+		// next!
+		add b32 $r4 0x8000
+		sub b32 $r3 1
+		bra ne #init_gpc
+
+	// save context size, and tell host we're ready
+	mov $r2 0x800
+	shl b32 $r2 6
+	iowr I[$r2 + 0x100] $r1		// CC_SCRATCH[1]  = context size
+	add b32 $r2 0x800
+	clear b32 $r1
+	bset $r1 31
+	iowr I[$r2 + 0x000] $r1		// CC_SCRATCH[0] |= 0x80000000
+
+// Main program loop, very simple, sleeps until woken up by the interrupt
+// handler, pulls a command from the queue and executes its handler
+//
+main:
+	// sleep until we have something to do
+	bset $flags $p0
+	sleep $p0
+	mov $r13 #cmd_queue
+	call #queue_get
+	bra $p1 #main
+
+	// context switch, requested by GPU?
+	cmpu b32 $r14 0x4001
+	bra ne #main_not_ctx_switch
+		trace_set(T_AUTO)
+		mov $r1 0xb00
+		shl b32 $r1 6
+		iord $r2 I[$r1 + 0x100]		// CHAN_NEXT
+		iord $r1 I[$r1 + 0x000]		// CHAN_CUR
+
+		xbit $r3 $r1 31
+		bra e #chsw_no_prev
+			xbit $r3 $r2 31
+			bra e #chsw_prev_no_next
+				push $r2
+				mov b32 $r2 $r1
+				trace_set(T_SAVE)
+				bclr $flags $p1
+				bset $flags $p2
+				call #ctx_xfer
+				trace_clr(T_SAVE);
+				pop $r2
+				trace_set(T_LOAD);
+				bset $flags $p1
+				call #ctx_xfer
+				trace_clr(T_LOAD);
+				bra #chsw_done
+			chsw_prev_no_next:
+				push $r2
+				mov b32 $r2 $r1
+				bclr $flags $p1
+				bclr $flags $p2
+				call #ctx_xfer
+				pop $r2
+				mov $r1 0xb00
+				shl b32 $r1 6
+				iowr I[$r1] $r2
+				bra #chsw_done
+		chsw_no_prev:
+			xbit $r3 $r2 31
+			bra e #chsw_done
+				bset $flags $p1
+				bclr $flags $p2
+				call #ctx_xfer
+
+		// ack the context switch request
+		chsw_done:
+		mov $r1 0xb0c
+		shl b32 $r1 6
+		mov $r2 1
+		iowr I[$r1 + 0x000] $r2		// 0x409b0c
+		trace_clr(T_AUTO)
+		bra #main
+
+	// request to set current channel? (*not* a context switch)
+	main_not_ctx_switch:
+	cmpu b32 $r14 0x0001
+	bra ne #main_not_ctx_chan
+		mov b32 $r2 $r15
+		call #ctx_chan
+		bra #main_done
+
+	// request to store current channel context?
+	main_not_ctx_chan:
+	cmpu b32 $r14 0x0002
+	bra ne #main_not_ctx_save
+		trace_set(T_SAVE)
+		bclr $flags $p1
+		bclr $flags $p2
+		call #ctx_xfer
+		trace_clr(T_SAVE)
+		bra #main_done
+
+	main_not_ctx_save:
+		shl b32 $r15 $r14 16
+		or $r15 E_BAD_COMMAND
+		call #error
+		bra #main
+
+	main_done:
+	mov $r1 0x820
+	shl b32 $r1 6
+	clear b32 $r2
+	bset $r2 31
+	iowr I[$r1 + 0x000] $r2		// CC_SCRATCH[0] |= 0x80000000
+	bra #main
+
+// interrupt handler
+ih:
+	push $r8
+	mov $r8 $flags
+	push $r8
+	push $r9
+	push $r10
+	push $r11
+	push $r13
+	push $r14
+	push $r15
+
+	// incoming fifo command?
+	iord $r10 I[$r0 + 0x200]	// INTR
+	and $r11 $r10 0x00000004
+	bra e #ih_no_fifo
+		// queue incoming fifo command for later processing
+		mov $r11 0x1900
+		mov $r13 #cmd_queue
+		iord $r14 I[$r11 + 0x100]	// FIFO_CMD
+		iord $r15 I[$r11 + 0x000]	// FIFO_DATA
+		call #queue_put
+		add b32 $r11 0x400
+		mov $r14 1
+		iowr I[$r11 + 0x000] $r14	// FIFO_ACK
+
+	// context switch request?
+	ih_no_fifo:
+	and $r11 $r10 0x00000100
+	bra e #ih_no_ctxsw
+		// enqueue a context switch for later processing
+		mov $r13 #cmd_queue
+		mov $r14 0x4001
+		call #queue_put
+
+	// anything we didn't handle, bring it to the host's attention
+	ih_no_ctxsw:
+	mov $r11 0x104
+	not b32 $r11
+	and $r11 $r10 $r11
+	bra e #ih_no_other
+		mov $r10 0xc1c
+		shl b32 $r10 6
+		iowr I[$r10] $r11	// INTR_UP_SET
+
+	// ack, and wake up main()
+	ih_no_other:
+	iowr I[$r0 + 0x100] $r10	// INTR_ACK
+
+	pop $r15
+	pop $r14
+	pop $r13
+	pop $r11
+	pop $r10
+	pop $r9
+	pop $r8
+	mov $flags $r8
+	pop $r8
+	bclr $flags $p0
+	iret
+
+// Not real sure, but, MEM_CMD 7 will hang forever if this isn't done
+ctx_4160s:
+	mov $r14 0x4160
+	sethi $r14 0x400000
+	mov $r15 1
+	call #nv_wr32
+	ctx_4160s_wait:
+		call #nv_rd32
+		xbit $r15 $r15 4
+		bra e #ctx_4160s_wait
+	ret
+
+// Without clearing again at end of xfer, some things cause PGRAPH
+// to hang with STATUS=0x00000007 until it's cleared.. fbcon can
+// still function with it set however...
+ctx_4160c:
+	mov $r14 0x4160
+	sethi $r14 0x400000
+	clear b32 $r15
+	call #nv_wr32
+	ret
+
+// Again, not real sure
+//
+// In: $r15 value to set 0x404170 to
+//
+ctx_4170s:
+	mov $r14 0x4170
+	sethi $r14 0x400000
+	or $r15 0x10
+	call #nv_wr32
+	ret
+
+// Waits for a ctx_4170s() call to complete
+//
+ctx_4170w:
+	mov $r14 0x4170
+	sethi $r14 0x400000
+	call #nv_rd32
+	and $r15 0x10
+	bra ne #ctx_4170w
+	ret
+
+// Disables various things, waits a bit, and re-enables them..
+//
+// Not sure how exactly this helps, perhaps "ENABLE" is not such a
+// good description for the bits we turn off?  Anyways, without this,
+// funny things happen.
+//
+ctx_redswitch:
+	mov $r14 0x614
+	shl b32 $r14 6
+	mov $r15 0x270
+	iowr I[$r14] $r15	// HUB_RED_SWITCH = ENABLE_GPC, POWER_ALL
+	mov $r15 8
+	ctx_redswitch_delay:
+		sub b32 $r15 1
+		bra ne #ctx_redswitch_delay
+	mov $r15 0x770
+	iowr I[$r14] $r15	// HUB_RED_SWITCH = ENABLE_ALL, POWER_ALL
+	ret
+
+// Not a clue what this is for, except that unless the value is 0x10, the
+// strand context is saved (and presumably restored) incorrectly..
+//
+// In: $r15 value to set to (0x00/0x10 are used)
+//
+ctx_86c:
+	mov $r14 0x86c
+	shl b32 $r14 6
+	iowr I[$r14] $r15	// HUB(0x86c) = val
+	mov $r14 -0x75ec
+	sethi $r14 0x400000
+	call #nv_wr32		// ROP(0xa14) = val
+	mov $r14 -0x5794
+	sethi $r14 0x410000
+	call #nv_wr32		// GPC(0x86c) = val
+	ret
+
+// ctx_load - load's a channel's ctxctl data, and selects its vm
+//
+// In: $r2 channel address
+//
+ctx_load:
+	trace_set(T_CHAN)
+
+	// switch to channel, somewhat magic in parts..
+	mov $r10 12		// DONE_UNK12
+	call #wait_donez
+	mov $r1 0xa24
+	shl b32 $r1 6
+	iowr I[$r1 + 0x000] $r0	// 0x409a24
+	mov $r3 0xb00
+	shl b32 $r3 6
+	iowr I[$r3 + 0x100] $r2	// CHAN_NEXT
+	mov $r1 0xa0c
+	shl b32 $r1 6
+	mov $r4 7
+	iowr I[$r1 + 0x000] $r2 // MEM_CHAN
+	iowr I[$r1 + 0x100] $r4	// MEM_CMD
+	ctx_chan_wait_0:
+		iord $r4 I[$r1 + 0x100]
+		and $r4 0x1f
+		bra ne #ctx_chan_wait_0
+	iowr I[$r3 + 0x000] $r2	// CHAN_CUR
+
+	// load channel header, fetch PGRAPH context pointer
+	mov $xtargets $r0
+	bclr $r2 31
+	shl b32 $r2 4
+	add b32 $r2 2
+
+	trace_set(T_LCHAN)
+	mov $r1 0xa04
+	shl b32 $r1 6
+	iowr I[$r1 + 0x000] $r2		// MEM_BASE
+	mov $r1 0xa20
+	shl b32 $r1 6
+	mov $r2 0x0002
+	sethi $r2 0x80000000
+	iowr I[$r1 + 0x000] $r2		// MEM_TARGET = vram
+	mov $r1 0x10			// chan + 0x0210
+	mov $r2 #xfer_data
+	sethi $r2 0x00020000		// 16 bytes
+	xdld $r1 $r2
+	xdwait
+	trace_clr(T_LCHAN)
+
+	// update current context
+	ld b32 $r1 D[$r0 + #xfer_data + 4]
+	shl b32 $r1 24
+	ld b32 $r2 D[$r0 + #xfer_data + 0]
+	shr b32 $r2 8
+	or $r1 $r2
+	st b32 D[$r0 + #ctx_current] $r1
+
+	// set transfer base to start of context, and fetch context header
+	trace_set(T_LCTXH)
+	mov $r2 0xa04
+	shl b32 $r2 6
+	iowr I[$r2 + 0x000] $r1		// MEM_BASE
+	mov $r2 1
+	mov $r1 0xa20
+	shl b32 $r1 6
+	iowr I[$r1 + 0x000] $r2		// MEM_TARGET = vm
+	mov $r1 #chan_data
+	sethi $r1 0x00060000		// 256 bytes
+	xdld $r0 $r1
+	xdwait
+	trace_clr(T_LCTXH)
+
+	trace_clr(T_CHAN)
+	ret
+
+// ctx_chan - handler for HUB_SET_CHAN command, will set a channel as
+//            the active channel for ctxctl, but not actually transfer
+//            any context data.  intended for use only during initial
+//            context construction.
+//
+// In: $r2 channel address
+//
+ctx_chan:
+	call #ctx_4160s
+	call #ctx_load
+	mov $r10 12			// DONE_UNK12
+	call #wait_donez
+	mov $r1 0xa10
+	shl b32 $r1 6
+	mov $r2 5
+	iowr I[$r1 + 0x000] $r2		// MEM_CMD = 5 (???)
+	ctx_chan_wait:
+		iord $r2 I[$r1 + 0x000]
+		or $r2 $r2
+		bra ne #ctx_chan_wait
+	call #ctx_4160c
+	ret
+
+// Execute per-context state overrides list
+//
+// Only executed on the first load of a channel.  Might want to look into
+// removing this and having the host directly modify the channel's context
+// to change this state...  The nouveau DRM already builds this list as
+// it's definitely needed for NVIDIA's, so we may as well use it for now
+//
+// Input: $r1 mmio list length
+//
+ctx_mmio_exec:
+	// set transfer base to be the mmio list
+	ld b32 $r3 D[$r0 + #chan_mmio_address]
+	mov $r2 0xa04
+	shl b32 $r2 6
+	iowr I[$r2 + 0x000] $r3		// MEM_BASE
+
+	clear b32 $r3
+	ctx_mmio_loop:
+		// fetch next 256 bytes of mmio list if necessary
+		and $r4 $r3 0xff
+		bra ne #ctx_mmio_pull
+			mov $r5 #xfer_data
+			sethi $r5 0x00060000	// 256 bytes
+			xdld $r3 $r5
+			xdwait
+
+		// execute a single list entry
+		ctx_mmio_pull:
+		ld b32 $r14 D[$r4 + #xfer_data + 0x00]
+		ld b32 $r15 D[$r4 + #xfer_data + 0x04]
+		call #nv_wr32
+
+		// next!
+		add b32 $r3 8
+		sub b32 $r1 1
+		bra ne #ctx_mmio_loop
+
+	// set transfer base back to the current context
+	ctx_mmio_done:
+	ld b32 $r3 D[$r0 + #ctx_current]
+	iowr I[$r2 + 0x000] $r3		// MEM_BASE
+
+	// disable the mmio list now, we don't need/want to execute it again
+	st b32 D[$r0 + #chan_mmio_count] $r0
+	mov $r1 #chan_data
+	sethi $r1 0x00060000		// 256 bytes
+	xdst $r0 $r1
+	xdwait
+	ret
+
+// Transfer HUB context data between GPU and storage area
+//
+// In: $r2 channel address
+//     $p1 clear on save, set on load
+//     $p2 set if opposite direction done/will be done, so:
+//		on save it means: "a load will follow this save"
+//		on load it means: "a save preceeded this load"
+//
+ctx_xfer:
+	bra not $p1 #ctx_xfer_pre
+	bra $p2 #ctx_xfer_pre_load
+	ctx_xfer_pre:
+		mov $r15 0x10
+		call #ctx_86c
+		call #ctx_4160s
+		bra not $p1 #ctx_xfer_exec
+
+	ctx_xfer_pre_load:
+		mov $r15 2
+		call #ctx_4170s
+		call #ctx_4170w
+		call #ctx_redswitch
+		clear b32 $r15
+		call #ctx_4170s
+		call #ctx_load
+
+	// fetch context pointer, and initiate xfer on all GPCs
+	ctx_xfer_exec:
+	ld b32 $r1 D[$r0 + #ctx_current]
+	mov $r2 0x414
+	shl b32 $r2 6
+	iowr I[$r2 + 0x000] $r0	// BAR_STATUS = reset
+	mov $r14 -0x5b00
+	sethi $r14 0x410000
+	mov b32 $r15 $r1
+	call #nv_wr32		// GPC_BCAST_WRCMD_DATA = ctx pointer
+	add b32 $r14 4
+	xbit $r15 $flags $p1
+	xbit $r2 $flags $p2
+	shl b32 $r2 1
+	or $r15 $r2
+	call #nv_wr32		// GPC_BCAST_WRCMD_CMD = GPC_XFER(type)
+
+	// strands
+	mov $r1 0x4afc
+	sethi $r1 0x20000
+	mov $r2 0xc
+	iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x0c
+	call #strand_wait
+	mov $r2 0x47fc
+	sethi $r2 0x20000
+	iowr I[$r2] $r0		// STRAND_FIRST_GENE(0x3f) = 0x00
+	xbit $r2 $flags $p1
+	add b32 $r2 3
+	iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x03/0x04 (SAVE/LOAD)
+
+	// mmio context
+	xbit $r10 $flags $p1	// direction
+	or $r10 6		// first, last
+	mov $r11 0		// base = 0
+	ld b32 $r12 D[$r0 + #hub_mmio_list_head]
+	ld b32 $r13 D[$r0 + #hub_mmio_list_tail]
+	mov $r14 0		// not multi
+	call #mmctx_xfer
+
+	// wait for GPCs to all complete
+	mov $r10 8		// DONE_BAR
+	call #wait_doneo
+
+	// wait for strand xfer to complete
+	call #strand_wait
+
+	// post-op
+	bra $p1 #ctx_xfer_post
+		mov $r10 12		// DONE_UNK12
+		call #wait_donez
+		mov $r1 0xa10
+		shl b32 $r1 6
+		mov $r2 5
+		iowr I[$r1] $r2		// MEM_CMD
+		ctx_xfer_post_save_wait:
+			iord $r2 I[$r1]
+			or $r2 $r2
+			bra ne #ctx_xfer_post_save_wait
+
+	bra $p2 #ctx_xfer_done
+	ctx_xfer_post:
+		mov $r15 2
+		call #ctx_4170s
+		clear b32 $r15
+		call #ctx_86c
+		call #strand_post
+		call #ctx_4170w
+		clear b32 $r15
+		call #ctx_4170s
+
+		bra not $p1 #ctx_xfer_no_post_mmio
+		ld b32 $r1 D[$r0 + #chan_mmio_count]
+		or $r1 $r1
+		bra e #ctx_xfer_no_post_mmio
+			call #ctx_mmio_exec
+
+		ctx_xfer_no_post_mmio:
+		call #ctx_4160c
+
+	ctx_xfer_done:
+	ret
+
+.align 256
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnvc0.fuc.h b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnvc0.fuc.h
new file mode 100644
index 0000000..85a8d55
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnvc0.fuc.h
@@ -0,0 +1,927 @@
+uint32_t nvc0_grhub_data[] = {
+/* 0x0000: gpc_count */
+	0x00000000,
+/* 0x0004: rop_count */
+	0x00000000,
+/* 0x0008: cmd_queue */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0050: hub_mmio_list_head */
+	0x00000000,
+/* 0x0054: hub_mmio_list_tail */
+	0x00000000,
+/* 0x0058: ctx_current */
+	0x00000000,
+/* 0x005c: chipsets */
+	0x000000c0,
+	0x013c00a0,
+	0x000000c1,
+	0x014000a0,
+	0x000000c3,
+	0x013c00a0,
+	0x000000c4,
+	0x013c00a0,
+	0x000000c8,
+	0x013c00a0,
+	0x000000ce,
+	0x013c00a0,
+	0x000000cf,
+	0x013c00a0,
+	0x000000d9,
+	0x01dc0140,
+	0x00000000,
+/* 0x00a0: nvc0_hub_mmio_head */
+	0x0417e91c,
+	0x04400204,
+	0x28404004,
+	0x00404044,
+	0x34404094,
+	0x184040d0,
+	0x004040f8,
+	0x08404130,
+	0x08404150,
+	0x04404164,
+	0x08404174,
+	0x1c404200,
+	0x34404404,
+	0x0c404460,
+	0x00404480,
+	0x00404498,
+	0x0c404604,
+	0x7c404618,
+	0x50404698,
+	0x044046f0,
+	0x54404700,
+	0x00405800,
+	0x08405830,
+	0x00405854,
+	0x0c405870,
+	0x04405a00,
+	0x00405a18,
+	0x00406020,
+	0x0c406028,
+	0x044064a8,
+	0x044064b4,
+	0x00407804,
+	0x1440780c,
+	0x004078bc,
+	0x18408000,
+	0x00408064,
+	0x08408800,
+	0x0c408900,
+	0x00408980,
+/* 0x013c: nvc0_hub_mmio_tail */
+	0x044064c0,
+/* 0x0140: nvc1_hub_mmio_tail */
+/* 0x0140: nvd9_hub_mmio_head */
+	0x0417e91c,
+	0x04400204,
+	0x24404004,
+	0x00404044,
+	0x34404094,
+	0x184040d0,
+	0x004040f8,
+	0x08404130,
+	0x08404150,
+	0x04404164,
+	0x04404178,
+	0x1c404200,
+	0x34404404,
+	0x0c404460,
+	0x00404480,
+	0x00404498,
+	0x0c404604,
+	0x7c404618,
+	0x50404698,
+	0x044046f0,
+	0x54404700,
+	0x00405800,
+	0x08405830,
+	0x00405854,
+	0x0c405870,
+	0x04405a00,
+	0x00405a18,
+	0x00406020,
+	0x0c406028,
+	0x044064a8,
+	0x104064b4,
+	0x00407804,
+	0x1440780c,
+	0x004078bc,
+	0x18408000,
+	0x00408064,
+	0x08408800,
+	0x0c408900,
+	0x00408980,
+/* 0x01dc: nvd9_hub_mmio_tail */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0200: chan_data */
+/* 0x0200: chan_mmio_count */
+	0x00000000,
+/* 0x0204: chan_mmio_address */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0300: xfer_data */
+	0x00000000,
+};
+
+uint32_t nvc0_grhub_code[] = {
+	0x03090ef5,
+/* 0x0004: queue_put */
+	0x9800d898,
+	0x86f001d9,
+	0x0489b808,
+	0xf00c1bf4,
+	0x21f502f7,
+	0x00f802ec,
+/* 0x001c: queue_put_next */
+	0xb60798c4,
+	0x8dbb0384,
+	0x0880b600,
+	0x80008e80,
+	0x90b6018f,
+	0x0f94f001,
+	0xf801d980,
+/* 0x0039: queue_get */
+	0x0131f400,
+	0x9800d898,
+	0x89b801d9,
+	0x210bf404,
+	0xb60789c4,
+	0x9dbb0394,
+	0x0890b600,
+	0x98009e98,
+	0x80b6019f,
+	0x0f84f001,
+	0xf400d880,
+/* 0x0066: queue_get_done */
+	0x00f80132,
+/* 0x0068: nv_rd32 */
+	0x0728b7f1,
+	0xb906b4b6,
+	0xc9f002ec,
+	0x00bcd01f,
+/* 0x0078: nv_rd32_wait */
+	0xc800bccf,
+	0x1bf41fcc,
+	0x06a7f0fa,
+	0x010321f5,
+	0xf840bfcf,
+/* 0x008d: nv_wr32 */
+	0x28b7f100,
+	0x06b4b607,
+	0xb980bfd0,
+	0xc9f002ec,
+	0x1ec9f01f,
+/* 0x00a3: nv_wr32_wait */
+	0xcf00bcd0,
+	0xccc800bc,
+	0xfa1bf41f,
+/* 0x00ae: watchdog_reset */
+	0x87f100f8,
+	0x84b60430,
+	0x1ff9f006,
+	0xf8008fd0,
+/* 0x00bd: watchdog_clear */
+	0x3087f100,
+	0x0684b604,
+	0xf80080d0,
+/* 0x00c9: wait_donez */
+	0x3c87f100,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d000,
+	0x081887f1,
+	0xd00684b6,
+/* 0x00e2: wait_done_wait_donez */
+	0x87f1008a,
+	0x84b60400,
+	0x0088cf06,
+	0xf4888aff,
+	0x87f1f31b,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00099,
+/* 0x0103: wait_doneo */
+	0xf100f800,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00099f0,
+	0x87f10089,
+	0x84b60818,
+	0x008ad006,
+/* 0x011c: wait_done_wait_doneo */
+	0x040087f1,
+	0xcf0684b6,
+	0x8aff0088,
+	0xf30bf488,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0099f094,
+	0xf80089d0,
+/* 0x013d: mmctx_size */
+/* 0x013f: nv_mmctx_size_loop */
+	0x9894bd00,
+	0x85b600e8,
+	0x0180b61a,
+	0xbb0284b6,
+	0xe0b60098,
+	0x04efb804,
+	0xb9eb1bf4,
+	0x00f8029f,
+/* 0x015c: mmctx_xfer */
+	0x083c87f1,
+	0xbd0684b6,
+	0x0199f094,
+	0xf10089d0,
+	0xb6071087,
+	0x94bd0684,
+	0xf405bbfd,
+	0x8bd0090b,
+	0x0099f000,
+/* 0x0180: mmctx_base_disabled */
+	0xf405eefd,
+	0x8ed00c0b,
+	0xc08fd080,
+/* 0x018f: mmctx_multi_disabled */
+	0xb70199f0,
+	0xc8010080,
+	0xb4b600ab,
+	0x0cb9f010,
+	0xb601aec8,
+	0xbefd11e4,
+	0x008bd005,
+/* 0x01a8: mmctx_exec_loop */
+/* 0x01a8: mmctx_wait_free */
+	0xf0008ecf,
+	0x0bf41fe4,
+	0x00ce98fa,
+	0xd005e9fd,
+	0xc0b6c08e,
+	0x04cdb804,
+	0xc8e81bf4,
+	0x1bf402ab,
+/* 0x01c9: mmctx_fini_wait */
+	0x008bcf18,
+	0xb01fb4f0,
+	0x1bf410b4,
+	0x02a7f0f7,
+	0xf4c921f4,
+/* 0x01de: mmctx_stop */
+	0xabc81b0e,
+	0x10b4b600,
+	0xf00cb9f0,
+	0x8bd012b9,
+/* 0x01ed: mmctx_stop_wait */
+	0x008bcf00,
+	0xf412bbc8,
+/* 0x01f6: mmctx_done */
+	0x87f1fa1b,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00199,
+/* 0x0207: strand_wait */
+	0xf900f800,
+	0x02a7f0a0,
+	0xfcc921f4,
+/* 0x0213: strand_pre */
+	0xf100f8a0,
+	0xf04afc87,
+	0x97f00283,
+	0x0089d00c,
+	0x020721f5,
+/* 0x0226: strand_post */
+	0x87f100f8,
+	0x83f04afc,
+	0x0d97f002,
+	0xf50089d0,
+	0xf8020721,
+/* 0x0239: strand_set */
+	0xfca7f100,
+	0x02a3f04f,
+	0x0500aba2,
+	0xd00fc7f0,
+	0xc7f000ac,
+	0x00bcd00b,
+	0x020721f5,
+	0xf000aed0,
+	0xbcd00ac7,
+	0x0721f500,
+/* 0x0263: strand_ctx_init */
+	0xf100f802,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00399f0,
+	0x21f50089,
+	0xe7f00213,
+	0x3921f503,
+	0xfca7f102,
+	0x02a3f046,
+	0x0400aba0,
+	0xf040a0d0,
+	0xbcd001c7,
+	0x0721f500,
+	0x010c9202,
+	0xf000acd0,
+	0xbcd002c7,
+	0x0721f500,
+	0x2621f502,
+	0x8087f102,
+	0x0684b608,
+	0xb70089cf,
+	0x95220080,
+/* 0x02ba: ctx_init_strand_loop */
+	0x8ed008fe,
+	0x408ed000,
+	0xb6808acf,
+	0xa0b606a5,
+	0x00eabb01,
+	0xb60480b6,
+	0x1bf40192,
+	0x08e4b6e8,
+	0xf1f2efbc,
+	0xb6085c87,
+	0x94bd0684,
+	0xd00399f0,
+	0x00f80089,
+/* 0x02ec: error */
+	0xe7f1e0f9,
+	0xe4b60814,
+	0x00efd006,
+	0x0c1ce7f1,
+	0xf006e4b6,
+	0xefd001f7,
+	0xf8e0fc00,
+/* 0x0309: init */
+	0xfe04bd00,
+	0x07fe0004,
+	0x0017f100,
+	0x0227f012,
+	0xf10012d0,
+	0xfe05b917,
+	0x17f10010,
+	0x10d00400,
+	0x0437f1c0,
+	0x0634b604,
+	0x200327f1,
+	0xf10032d0,
+	0xd0200427,
+	0x27f10132,
+	0x32d0200b,
+	0x0c27f102,
+	0x0732d020,
+	0x0c2427f1,
+	0xb90624b6,
+	0x23d00003,
+	0x0427f100,
+	0x0023f087,
+	0xb70012d0,
+	0xf0010012,
+	0x12d00427,
+	0x1031f400,
+	0x9604e7f1,
+	0xf440e3f0,
+	0xf1c76821,
+	0x01018090,
+	0x801ff4f0,
+	0x17f0000f,
+	0x041fbb01,
+	0xf10112b6,
+	0xb6040c27,
+	0x21d00624,
+	0x4021d000,
+	0x080027f1,
+	0xcf0624b6,
+	0xf7f00022,
+/* 0x03a9: init_find_chipset */
+	0x08f0b654,
+	0xb800f398,
+	0x0bf40432,
+	0x0034b00b,
+	0xf8f11bf4,
+/* 0x03bd: init_context */
+	0x0017f100,
+	0x02fe5801,
+	0xf003ff58,
+	0x0e8000e3,
+	0x150f8014,
+	0x013d21f5,
+	0x070037f1,
+	0x950634b6,
+	0x34d00814,
+	0x4034d000,
+	0x130030b7,
+	0xb6001fbb,
+	0x3fd002f5,
+	0x0815b600,
+	0xb60110b6,
+	0x1fb90814,
+	0x6321f502,
+	0x001fbb02,
+	0xf1000398,
+	0xf0200047,
+/* 0x040e: init_gpc */
+	0x4ea05043,
+	0x1fb90804,
+	0x8d21f402,
+	0x08004ea0,
+	0xf4022fb9,
+	0x4ea08d21,
+	0xf4bd010c,
+	0xa08d21f4,
+	0xf401044e,
+	0x4ea08d21,
+	0xf7f00100,
+	0x8d21f402,
+	0x08004ea0,
+/* 0x0440: init_gpc_wait */
+	0xc86821f4,
+	0x0bf41fff,
+	0x044ea0fa,
+	0x6821f408,
+	0xb7001fbb,
+	0xb6800040,
+	0x1bf40132,
+	0x0027f1b4,
+	0x0624b608,
+	0xb74021d0,
+	0xbd080020,
+	0x1f19f014,
+/* 0x0473: main */
+	0xf40021d0,
+	0x28f40031,
+	0x08d7f000,
+	0xf43921f4,
+	0xe4b1f401,
+	0x1bf54001,
+	0x87f100d1,
+	0x84b6083c,
+	0xf094bd06,
+	0x89d00499,
+	0x0017f100,
+	0x0614b60b,
+	0xcf4012cf,
+	0x13c80011,
+	0x7e0bf41f,
+	0xf41f23c8,
+	0x20f95a0b,
+	0xf10212b9,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00799f0,
+	0x32f40089,
+	0x0231f401,
+	0x082921f5,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0799f094,
+	0xfc0089d0,
+	0x3c87f120,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d006,
+	0xf50131f4,
+	0xf1082921,
+	0xb6085c87,
+	0x94bd0684,
+	0xd00699f0,
+	0x0ef40089,
+/* 0x0509: chsw_prev_no_next */
+	0xb920f931,
+	0x32f40212,
+	0x0232f401,
+	0x082921f5,
+	0x17f120fc,
+	0x14b60b00,
+	0x0012d006,
+/* 0x0527: chsw_no_prev */
+	0xc8130ef4,
+	0x0bf41f23,
+	0x0131f40d,
+	0xf50232f4,
+/* 0x0537: chsw_done */
+	0xf1082921,
+	0xb60b0c17,
+	0x27f00614,
+	0x0012d001,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0499f094,
+	0xf50089d0,
+/* 0x0557: main_not_ctx_switch */
+	0xb0ff200e,
+	0x1bf401e4,
+	0x02f2b90d,
+	0x07b521f5,
+/* 0x0567: main_not_ctx_chan */
+	0xb0420ef4,
+	0x1bf402e4,
+	0x3c87f12e,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d007,
+	0xf40132f4,
+	0x21f50232,
+	0x87f10829,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00799,
+	0x110ef400,
+/* 0x0598: main_not_ctx_save */
+	0xf010ef94,
+	0x21f501f5,
+	0x0ef502ec,
+/* 0x05a6: main_done */
+	0x17f1fed1,
+	0x14b60820,
+	0xf024bd06,
+	0x12d01f29,
+	0xbe0ef500,
+/* 0x05b9: ih */
+	0xfe80f9fe,
+	0x80f90188,
+	0xa0f990f9,
+	0xd0f9b0f9,
+	0xf0f9e0f9,
+	0xc4800acf,
+	0x0bf404ab,
+	0x00b7f11d,
+	0x08d7f019,
+	0xcf40becf,
+	0x21f400bf,
+	0x00b0b704,
+	0x01e7f004,
+/* 0x05ef: ih_no_fifo */
+	0xe400bed0,
+	0xf40100ab,
+	0xd7f00d0b,
+	0x01e7f108,
+	0x0421f440,
+/* 0x0600: ih_no_ctxsw */
+	0x0104b7f1,
+	0xabffb0bd,
+	0x0d0bf4b4,
+	0x0c1ca7f1,
+	0xd006a4b6,
+/* 0x0616: ih_no_other */
+	0x0ad000ab,
+	0xfcf0fc40,
+	0xfcd0fce0,
+	0xfca0fcb0,
+	0xfe80fc90,
+	0x80fc0088,
+	0xf80032f4,
+/* 0x0631: ctx_4160s */
+	0x60e7f101,
+	0x40e3f041,
+	0xf401f7f0,
+/* 0x063e: ctx_4160s_wait */
+	0x21f48d21,
+	0x04ffc868,
+	0xf8fa0bf4,
+/* 0x0649: ctx_4160c */
+	0x60e7f100,
+	0x40e3f041,
+	0x21f4f4bd,
+/* 0x0657: ctx_4170s */
+	0xf100f88d,
+	0xf04170e7,
+	0xf5f040e3,
+	0x8d21f410,
+/* 0x0666: ctx_4170w */
+	0xe7f100f8,
+	0xe3f04170,
+	0x6821f440,
+	0xf410f4f0,
+	0x00f8f31b,
+/* 0x0678: ctx_redswitch */
+	0x0614e7f1,
+	0xf106e4b6,
+	0xd00270f7,
+	0xf7f000ef,
+/* 0x0689: ctx_redswitch_delay */
+	0x01f2b608,
+	0xf1fd1bf4,
+	0xd00770f7,
+	0x00f800ef,
+/* 0x0698: ctx_86c */
+	0x086ce7f1,
+	0xd006e4b6,
+	0xe7f100ef,
+	0xe3f08a14,
+	0x8d21f440,
+	0xa86ce7f1,
+	0xf441e3f0,
+	0x00f88d21,
+/* 0x06b8: ctx_load */
+	0x083c87f1,
+	0xbd0684b6,
+	0x0599f094,
+	0xf00089d0,
+	0x21f40ca7,
+	0x2417f1c9,
+	0x0614b60a,
+	0xf10010d0,
+	0xb60b0037,
+	0x32d00634,
+	0x0c17f140,
+	0x0614b60a,
+	0xd00747f0,
+	0x14d00012,
+/* 0x06f1: ctx_chan_wait_0 */
+	0x4014cf40,
+	0xf41f44f0,
+	0x32d0fa1b,
+	0x000bfe00,
+	0xb61f2af0,
+	0x20b60424,
+	0x3c87f102,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d008,
+	0x0a0417f1,
+	0xd00614b6,
+	0x17f10012,
+	0x14b60a20,
+	0x0227f006,
+	0x800023f1,
+	0xf00012d0,
+	0x27f11017,
+	0x23f00300,
+	0x0512fa02,
+	0x87f103f8,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00899,
+	0xc1019800,
+	0x981814b6,
+	0x25b6c002,
+	0x0512fd08,
+	0xf1160180,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00999f0,
+	0x27f10089,
+	0x24b60a04,
+	0x0021d006,
+	0xf10127f0,
+	0xb60a2017,
+	0x12d00614,
+	0x0017f100,
+	0x0613f002,
+	0xf80501fa,
+	0x5c87f103,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d009,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0599f094,
+	0xf80089d0,
+/* 0x07b5: ctx_chan */
+	0x3121f500,
+	0xb821f506,
+	0x0ca7f006,
+	0xf1c921f4,
+	0xb60a1017,
+	0x27f00614,
+	0x0012d005,
+/* 0x07d0: ctx_chan_wait */
+	0xfd0012cf,
+	0x1bf40522,
+	0x4921f5fa,
+/* 0x07df: ctx_mmio_exec */
+	0x9800f806,
+	0x27f18103,
+	0x24b60a04,
+	0x0023d006,
+/* 0x07ee: ctx_mmio_loop */
+	0x34c434bd,
+	0x0f1bf4ff,
+	0x030057f1,
+	0xfa0653f0,
+	0x03f80535,
+/* 0x0800: ctx_mmio_pull */
+	0x98c04e98,
+	0x21f4c14f,
+	0x0830b68d,
+	0xf40112b6,
+/* 0x0812: ctx_mmio_done */
+	0x0398df1b,
+	0x0023d016,
+	0xf1800080,
+	0xf0020017,
+	0x01fa0613,
+	0xf803f806,
+/* 0x0829: ctx_xfer */
+	0x0611f400,
+/* 0x082f: ctx_xfer_pre */
+	0xf01102f4,
+	0x21f510f7,
+	0x21f50698,
+	0x11f40631,
+/* 0x083d: ctx_xfer_pre_load */
+	0x02f7f01c,
+	0x065721f5,
+	0x066621f5,
+	0x067821f5,
+	0x21f5f4bd,
+	0x21f50657,
+/* 0x0856: ctx_xfer_exec */
+	0x019806b8,
+	0x1427f116,
+	0x0624b604,
+	0xf10020d0,
+	0xf0a500e7,
+	0x1fb941e3,
+	0x8d21f402,
+	0xf004e0b6,
+	0x2cf001fc,
+	0x0124b602,
+	0xf405f2fd,
+	0x17f18d21,
+	0x13f04afc,
+	0x0c27f002,
+	0xf50012d0,
+	0xf1020721,
+	0xf047fc27,
+	0x20d00223,
+	0x012cf000,
+	0xd00320b6,
+	0xacf00012,
+	0x06a5f001,
+	0x9800b7f0,
+	0x0d98140c,
+	0x00e7f015,
+	0x015c21f5,
+	0xf508a7f0,
+	0xf5010321,
+	0xf4020721,
+	0xa7f02201,
+	0xc921f40c,
+	0x0a1017f1,
+	0xf00614b6,
+	0x12d00527,
+/* 0x08dd: ctx_xfer_post_save_wait */
+	0x0012cf00,
+	0xf40522fd,
+	0x02f4fa1b,
+/* 0x08e9: ctx_xfer_post */
+	0x02f7f032,
+	0x065721f5,
+	0x21f5f4bd,
+	0x21f50698,
+	0x21f50226,
+	0xf4bd0666,
+	0x065721f5,
+	0x981011f4,
+	0x11fd8001,
+	0x070bf405,
+	0x07df21f5,
+/* 0x0914: ctx_xfer_no_post_mmio */
+	0x064921f5,
+/* 0x0918: ctx_xfer_done */
+	0x000000f8,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnve0.fuc b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnve0.fuc
new file mode 100644
index 0000000..138eeaa
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnve0.fuc
@@ -0,0 +1,780 @@
+/* fuc microcode for nve0 PGRAPH/HUB
+ *
+ * Copyright 2011 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+/* To build:
+ *    m4 nve0_grhub.fuc | envyas -a -w -m fuc -V nva3 -o nve0_grhub.fuc.h
+ */
+
+.section #nve0_grhub_data
+include(`nve0.fuc')
+gpc_count:		.b32 0
+rop_count:		.b32 0
+cmd_queue:		queue_init
+hub_mmio_list_head:	.b32 0
+hub_mmio_list_tail:	.b32 0
+
+ctx_current:		.b32 0
+
+chipsets:
+.b8  0xe4 0 0 0
+.b16 #nve4_hub_mmio_head
+.b16 #nve4_hub_mmio_tail
+.b8  0xe7 0 0 0
+.b16 #nve4_hub_mmio_head
+.b16 #nve4_hub_mmio_tail
+.b8  0 0 0 0
+
+nve4_hub_mmio_head:
+mmctx_data(0x17e91c, 2)
+mmctx_data(0x400204, 2)
+mmctx_data(0x404010, 7)
+mmctx_data(0x4040a8, 9)
+mmctx_data(0x4040d0, 7)
+mmctx_data(0x4040f8, 1)
+mmctx_data(0x404130, 3)
+mmctx_data(0x404150, 3)
+mmctx_data(0x404164, 1)
+mmctx_data(0x4041a0, 4)
+mmctx_data(0x404200, 4)
+mmctx_data(0x404404, 14)
+mmctx_data(0x404460, 4)
+mmctx_data(0x404480, 1)
+mmctx_data(0x404498, 1)
+mmctx_data(0x404604, 4)
+mmctx_data(0x404618, 4)
+mmctx_data(0x40462c, 2)
+mmctx_data(0x404640, 1)
+mmctx_data(0x404654, 1)
+mmctx_data(0x404660, 1)
+mmctx_data(0x404678, 19)
+mmctx_data(0x4046c8, 3)
+mmctx_data(0x404700, 3)
+mmctx_data(0x404718, 10)
+mmctx_data(0x404744, 2)
+mmctx_data(0x404754, 1)
+mmctx_data(0x405800, 1)
+mmctx_data(0x405830, 3)
+mmctx_data(0x405854, 1)
+mmctx_data(0x405870, 4)
+mmctx_data(0x405a00, 2)
+mmctx_data(0x405a18, 1)
+mmctx_data(0x405b00, 1)
+mmctx_data(0x405b10, 1)
+mmctx_data(0x406020, 1)
+mmctx_data(0x406028, 4)
+mmctx_data(0x4064a8, 2)
+mmctx_data(0x4064b4, 2)
+mmctx_data(0x4064c0, 12)
+mmctx_data(0x4064fc, 1)
+mmctx_data(0x407040, 1)
+mmctx_data(0x407804, 1)
+mmctx_data(0x40780c, 6)
+mmctx_data(0x4078bc, 1)
+mmctx_data(0x408000, 7)
+mmctx_data(0x408064, 1)
+mmctx_data(0x408800, 3)
+mmctx_data(0x408840, 1)
+mmctx_data(0x408900, 3)
+mmctx_data(0x408980, 1)
+nve4_hub_mmio_tail:
+
+.align 256
+chan_data:
+chan_mmio_count:	.b32 0
+chan_mmio_address:	.b32 0
+
+.align 256
+xfer_data: 		.b32 0
+
+.section #nve0_grhub_code
+bra #init
+define(`include_code')
+include(`nve0.fuc')
+
+// reports an exception to the host
+//
+// In: $r15 error code (see nve0.fuc)
+//
+error:
+	push $r14
+	mov $r14 0x814
+	shl b32 $r14 6
+	iowr I[$r14 + 0x000] $r15	// CC_SCRATCH[5] = error code
+	mov $r14 0xc1c
+	shl b32 $r14 6
+	mov $r15 1
+	iowr I[$r14 + 0x000] $r15	// INTR_UP_SET
+	pop $r14
+	ret
+
+// HUB fuc initialisation, executed by triggering ucode start, will
+// fall through to main loop after completion.
+//
+// Input:
+//   CC_SCRATCH[0]: chipset (PMC_BOOT_0 read returns 0x0bad0bad... sigh)
+//
+// Output:
+//   CC_SCRATCH[0]:
+//	     31:31: set to signal completion
+//   CC_SCRATCH[1]:
+//	      31:0: total PGRAPH context size
+//
+init:
+	clear b32 $r0
+	mov $sp $r0
+	mov $xdbase $r0
+
+	// enable fifo access
+	mov $r1 0x1200
+	mov $r2 2
+	iowr I[$r1 + 0x000] $r2	// FIFO_ENABLE
+
+	// setup i0 handler, and route all interrupts to it
+	mov $r1 #ih
+	mov $iv0 $r1
+	mov $r1 0x400
+	iowr I[$r1 + 0x300] $r0	// INTR_DISPATCH
+
+	// route HUB_CHANNEL_SWITCH to fuc interrupt 8
+	mov $r3 0x404
+	shl b32 $r3 6
+	mov $r2 0x2003		// { HUB_CHANNEL_SWITCH, ZERO } -> intr 8
+	iowr I[$r3 + 0x000] $r2
+
+	// not sure what these are, route them because NVIDIA does, and
+	// the IRQ handler will signal the host if we ever get one.. we
+	// may find out if/why we need to handle these if so..
+	//
+	mov $r2 0x2004
+	iowr I[$r3 + 0x004] $r2 // { 0x04, ZERO } -> intr 9
+	mov $r2 0x200b
+	iowr I[$r3 + 0x008] $r2 // { 0x0b, ZERO } -> intr 10
+	mov $r2 0x200c
+	iowr I[$r3 + 0x01c] $r2 // { 0x0c, ZERO } -> intr 15
+
+	// enable all INTR_UP interrupts
+	mov $r2 0xc24
+	shl b32 $r2 6
+	not b32 $r3 $r0
+	iowr I[$r2] $r3
+
+	// enable fifo, ctxsw, 9, 10, 15 interrupts
+	mov $r2 -0x78fc		// 0x8704
+	sethi $r2 0
+	iowr I[$r1 + 0x000] $r2	// INTR_EN_SET
+
+	// fifo level triggered, rest edge
+	sub b32 $r1 0x100
+	mov $r2 4
+	iowr I[$r1] $r2
+
+	// enable interrupts
+	bset $flags ie0
+
+	// fetch enabled GPC/ROP counts
+	mov $r14 -0x69fc	// 0x409604
+	sethi $r14 0x400000
+	call #nv_rd32
+	extr $r1 $r15 16:20
+	st b32 D[$r0 + #rop_count] $r1
+	and $r15 0x1f
+	st b32 D[$r0 + #gpc_count] $r15
+
+	// set BAR_REQMASK to GPC mask
+	mov $r1 1
+	shl b32 $r1 $r15
+	sub b32 $r1 1
+	mov $r2 0x40c
+	shl b32 $r2 6
+	iowr I[$r2 + 0x000] $r1
+	iowr I[$r2 + 0x100] $r1
+
+	// find context data for this chipset
+	mov $r2 0x800
+	shl b32 $r2 6
+	iord $r2 I[$r2 + 0x000]		// CC_SCRATCH[0]
+	mov $r15 #chipsets - 8
+	init_find_chipset:
+		add b32 $r15 8
+		ld b32 $r3 D[$r15 + 0x00]
+		cmpu b32 $r3 $r2
+		bra e #init_context
+		cmpu b32 $r3 0
+		bra ne #init_find_chipset
+		// unknown chipset
+		ret
+
+	// context size calculation, reserve first 256 bytes for use by fuc
+	init_context:
+	mov $r1 256
+
+	// calculate size of mmio context data
+	ld b16 $r14 D[$r15 + 4]
+	ld b16 $r15 D[$r15 + 6]
+	sethi $r14 0
+	st b32 D[$r0 + #hub_mmio_list_head] $r14
+	st b32 D[$r0 + #hub_mmio_list_tail] $r15
+	call #mmctx_size
+
+	// set mmctx base addresses now so we don't have to do it later,
+	// they don't (currently) ever change
+	mov $r3 0x700
+	shl b32 $r3 6
+	shr b32 $r4 $r1 8
+	iowr I[$r3 + 0x000] $r4		// MMCTX_SAVE_SWBASE
+	iowr I[$r3 + 0x100] $r4		// MMCTX_LOAD_SWBASE
+	add b32 $r3 0x1300
+	add b32 $r1 $r15
+	shr b32 $r15 2
+	iowr I[$r3 + 0x000] $r15	// MMCTX_LOAD_COUNT, wtf for?!?
+
+	// strands, base offset needs to be aligned to 256 bytes
+	shr b32 $r1 8
+	add b32 $r1 1
+	shl b32 $r1 8
+	mov b32 $r15 $r1
+	call #strand_ctx_init
+	add b32 $r1 $r15
+
+	// initialise each GPC in sequence by passing in the offset of its
+	// context data in GPCn_CC_SCRATCH[1], and starting its FUC (which
+	// has previously been uploaded by the host) running.
+	//
+	// the GPC fuc init sequence will set GPCn_CC_SCRATCH[0] bit 31
+	// when it has completed, and return the size of its context data
+	// in GPCn_CC_SCRATCH[1]
+	//
+	ld b32 $r3 D[$r0 + #gpc_count]
+	mov $r4 0x2000
+	sethi $r4 0x500000
+	init_gpc:
+		// setup, and start GPC ucode running
+		add b32 $r14 $r4 0x804
+		mov b32 $r15 $r1
+		call #nv_wr32			// CC_SCRATCH[1] = ctx offset
+		add b32 $r14 $r4 0x800
+		mov b32 $r15 $r2
+		call #nv_wr32			// CC_SCRATCH[0] = chipset
+		add b32 $r14 $r4 0x10c
+		clear b32 $r15
+		call #nv_wr32
+		add b32 $r14 $r4 0x104
+		call #nv_wr32			// ENTRY
+		add b32 $r14 $r4 0x100
+		mov $r15 2			// CTRL_START_TRIGGER
+		call #nv_wr32			// CTRL
+
+		// wait for it to complete, and adjust context size
+		add b32 $r14 $r4 0x800
+		init_gpc_wait:
+			call #nv_rd32
+			xbit $r15 $r15 31
+			bra e #init_gpc_wait
+		add b32 $r14 $r4 0x804
+		call #nv_rd32
+		add b32 $r1 $r15
+
+		// next!
+		add b32 $r4 0x8000
+		sub b32 $r3 1
+		bra ne #init_gpc
+
+	// save context size, and tell host we're ready
+	mov $r2 0x800
+	shl b32 $r2 6
+	iowr I[$r2 + 0x100] $r1		// CC_SCRATCH[1]  = context size
+	add b32 $r2 0x800
+	clear b32 $r1
+	bset $r1 31
+	iowr I[$r2 + 0x000] $r1		// CC_SCRATCH[0] |= 0x80000000
+
+// Main program loop, very simple, sleeps until woken up by the interrupt
+// handler, pulls a command from the queue and executes its handler
+//
+main:
+	// sleep until we have something to do
+	bset $flags $p0
+	sleep $p0
+	mov $r13 #cmd_queue
+	call #queue_get
+	bra $p1 #main
+
+	// context switch, requested by GPU?
+	cmpu b32 $r14 0x4001
+	bra ne #main_not_ctx_switch
+		trace_set(T_AUTO)
+		mov $r1 0xb00
+		shl b32 $r1 6
+		iord $r2 I[$r1 + 0x100]		// CHAN_NEXT
+		iord $r1 I[$r1 + 0x000]		// CHAN_CUR
+
+		xbit $r3 $r1 31
+		bra e #chsw_no_prev
+			xbit $r3 $r2 31
+			bra e #chsw_prev_no_next
+				push $r2
+				mov b32 $r2 $r1
+				trace_set(T_SAVE)
+				bclr $flags $p1
+				bset $flags $p2
+				call #ctx_xfer
+				trace_clr(T_SAVE);
+				pop $r2
+				trace_set(T_LOAD);
+				bset $flags $p1
+				call #ctx_xfer
+				trace_clr(T_LOAD);
+				bra #chsw_done
+			chsw_prev_no_next:
+				push $r2
+				mov b32 $r2 $r1
+				bclr $flags $p1
+				bclr $flags $p2
+				call #ctx_xfer
+				pop $r2
+				mov $r1 0xb00
+				shl b32 $r1 6
+				iowr I[$r1] $r2
+				bra #chsw_done
+		chsw_no_prev:
+			xbit $r3 $r2 31
+			bra e #chsw_done
+				bset $flags $p1
+				bclr $flags $p2
+				call #ctx_xfer
+
+		// ack the context switch request
+		chsw_done:
+		mov $r1 0xb0c
+		shl b32 $r1 6
+		mov $r2 1
+		iowr I[$r1 + 0x000] $r2		// 0x409b0c
+		trace_clr(T_AUTO)
+		bra #main
+
+	// request to set current channel? (*not* a context switch)
+	main_not_ctx_switch:
+	cmpu b32 $r14 0x0001
+	bra ne #main_not_ctx_chan
+		mov b32 $r2 $r15
+		call #ctx_chan
+		bra #main_done
+
+	// request to store current channel context?
+	main_not_ctx_chan:
+	cmpu b32 $r14 0x0002
+	bra ne #main_not_ctx_save
+		trace_set(T_SAVE)
+		bclr $flags $p1
+		bclr $flags $p2
+		call #ctx_xfer
+		trace_clr(T_SAVE)
+		bra #main_done
+
+	main_not_ctx_save:
+		shl b32 $r15 $r14 16
+		or $r15 E_BAD_COMMAND
+		call #error
+		bra #main
+
+	main_done:
+	mov $r1 0x820
+	shl b32 $r1 6
+	clear b32 $r2
+	bset $r2 31
+	iowr I[$r1 + 0x000] $r2		// CC_SCRATCH[0] |= 0x80000000
+	bra #main
+
+// interrupt handler
+ih:
+	push $r8
+	mov $r8 $flags
+	push $r8
+	push $r9
+	push $r10
+	push $r11
+	push $r13
+	push $r14
+	push $r15
+
+	// incoming fifo command?
+	iord $r10 I[$r0 + 0x200]	// INTR
+	and $r11 $r10 0x00000004
+	bra e #ih_no_fifo
+		// queue incoming fifo command for later processing
+		mov $r11 0x1900
+		mov $r13 #cmd_queue
+		iord $r14 I[$r11 + 0x100]	// FIFO_CMD
+		iord $r15 I[$r11 + 0x000]	// FIFO_DATA
+		call #queue_put
+		add b32 $r11 0x400
+		mov $r14 1
+		iowr I[$r11 + 0x000] $r14	// FIFO_ACK
+
+	// context switch request?
+	ih_no_fifo:
+	and $r11 $r10 0x00000100
+	bra e #ih_no_ctxsw
+		// enqueue a context switch for later processing
+		mov $r13 #cmd_queue
+		mov $r14 0x4001
+		call #queue_put
+
+	// anything we didn't handle, bring it to the host's attention
+	ih_no_ctxsw:
+	mov $r11 0x104
+	not b32 $r11
+	and $r11 $r10 $r11
+	bra e #ih_no_other
+		mov $r10 0xc1c
+		shl b32 $r10 6
+		iowr I[$r10] $r11	// INTR_UP_SET
+
+	// ack, and wake up main()
+	ih_no_other:
+	iowr I[$r0 + 0x100] $r10	// INTR_ACK
+
+	pop $r15
+	pop $r14
+	pop $r13
+	pop $r11
+	pop $r10
+	pop $r9
+	pop $r8
+	mov $flags $r8
+	pop $r8
+	bclr $flags $p0
+	iret
+
+// Again, not real sure
+//
+// In: $r15 value to set 0x404170 to
+//
+ctx_4170s:
+	mov $r14 0x4170
+	sethi $r14 0x400000
+	or $r15 0x10
+	call #nv_wr32
+	ret
+
+// Waits for a ctx_4170s() call to complete
+//
+ctx_4170w:
+	mov $r14 0x4170
+	sethi $r14 0x400000
+	call #nv_rd32
+	and $r15 0x10
+	bra ne #ctx_4170w
+	ret
+
+// Disables various things, waits a bit, and re-enables them..
+//
+// Not sure how exactly this helps, perhaps "ENABLE" is not such a
+// good description for the bits we turn off?  Anyways, without this,
+// funny things happen.
+//
+ctx_redswitch:
+	mov $r14 0x614
+	shl b32 $r14 6
+	mov $r15 0x270
+	iowr I[$r14] $r15	// HUB_RED_SWITCH = ENABLE_GPC, POWER_ALL
+	mov $r15 8
+	ctx_redswitch_delay:
+		sub b32 $r15 1
+		bra ne #ctx_redswitch_delay
+	mov $r15 0x770
+	iowr I[$r14] $r15	// HUB_RED_SWITCH = ENABLE_ALL, POWER_ALL
+	ret
+
+// Not a clue what this is for, except that unless the value is 0x10, the
+// strand context is saved (and presumably restored) incorrectly..
+//
+// In: $r15 value to set to (0x00/0x10 are used)
+//
+ctx_86c:
+	mov $r14 0x86c
+	shl b32 $r14 6
+	iowr I[$r14] $r15	// HUB(0x86c) = val
+	mov $r14 -0x75ec
+	sethi $r14 0x400000
+	call #nv_wr32		// ROP(0xa14) = val
+	mov $r14 -0x5794
+	sethi $r14 0x410000
+	call #nv_wr32		// GPC(0x86c) = val
+	ret
+
+// ctx_load - load's a channel's ctxctl data, and selects its vm
+//
+// In: $r2 channel address
+//
+ctx_load:
+	trace_set(T_CHAN)
+
+	// switch to channel, somewhat magic in parts..
+	mov $r10 12		// DONE_UNK12
+	call #wait_donez
+	mov $r1 0xa24
+	shl b32 $r1 6
+	iowr I[$r1 + 0x000] $r0	// 0x409a24
+	mov $r3 0xb00
+	shl b32 $r3 6
+	iowr I[$r3 + 0x100] $r2	// CHAN_NEXT
+	mov $r1 0xa0c
+	shl b32 $r1 6
+	mov $r4 7
+	iowr I[$r1 + 0x000] $r2 // MEM_CHAN
+	iowr I[$r1 + 0x100] $r4	// MEM_CMD
+	ctx_chan_wait_0:
+		iord $r4 I[$r1 + 0x100]
+		and $r4 0x1f
+		bra ne #ctx_chan_wait_0
+	iowr I[$r3 + 0x000] $r2	// CHAN_CUR
+
+	// load channel header, fetch PGRAPH context pointer
+	mov $xtargets $r0
+	bclr $r2 31
+	shl b32 $r2 4
+	add b32 $r2 2
+
+	trace_set(T_LCHAN)
+	mov $r1 0xa04
+	shl b32 $r1 6
+	iowr I[$r1 + 0x000] $r2		// MEM_BASE
+	mov $r1 0xa20
+	shl b32 $r1 6
+	mov $r2 0x0002
+	sethi $r2 0x80000000
+	iowr I[$r1 + 0x000] $r2		// MEM_TARGET = vram
+	mov $r1 0x10			// chan + 0x0210
+	mov $r2 #xfer_data
+	sethi $r2 0x00020000		// 16 bytes
+	xdld $r1 $r2
+	xdwait
+	trace_clr(T_LCHAN)
+
+	// update current context
+	ld b32 $r1 D[$r0 + #xfer_data + 4]
+	shl b32 $r1 24
+	ld b32 $r2 D[$r0 + #xfer_data + 0]
+	shr b32 $r2 8
+	or $r1 $r2
+	st b32 D[$r0 + #ctx_current] $r1
+
+	// set transfer base to start of context, and fetch context header
+	trace_set(T_LCTXH)
+	mov $r2 0xa04
+	shl b32 $r2 6
+	iowr I[$r2 + 0x000] $r1		// MEM_BASE
+	mov $r2 1
+	mov $r1 0xa20
+	shl b32 $r1 6
+	iowr I[$r1 + 0x000] $r2		// MEM_TARGET = vm
+	mov $r1 #chan_data
+	sethi $r1 0x00060000		// 256 bytes
+	xdld $r0 $r1
+	xdwait
+	trace_clr(T_LCTXH)
+
+	trace_clr(T_CHAN)
+	ret
+
+// ctx_chan - handler for HUB_SET_CHAN command, will set a channel as
+//            the active channel for ctxctl, but not actually transfer
+//            any context data.  intended for use only during initial
+//            context construction.
+//
+// In: $r2 channel address
+//
+ctx_chan:
+	call #ctx_load
+	mov $r10 12			// DONE_UNK12
+	call #wait_donez
+	mov $r1 0xa10
+	shl b32 $r1 6
+	mov $r2 5
+	iowr I[$r1 + 0x000] $r2		// MEM_CMD = 5 (???)
+	ctx_chan_wait:
+		iord $r2 I[$r1 + 0x000]
+		or $r2 $r2
+		bra ne #ctx_chan_wait
+	ret
+
+// Execute per-context state overrides list
+//
+// Only executed on the first load of a channel.  Might want to look into
+// removing this and having the host directly modify the channel's context
+// to change this state...  The nouveau DRM already builds this list as
+// it's definitely needed for NVIDIA's, so we may as well use it for now
+//
+// Input: $r1 mmio list length
+//
+ctx_mmio_exec:
+	// set transfer base to be the mmio list
+	ld b32 $r3 D[$r0 + #chan_mmio_address]
+	mov $r2 0xa04
+	shl b32 $r2 6
+	iowr I[$r2 + 0x000] $r3		// MEM_BASE
+
+	clear b32 $r3
+	ctx_mmio_loop:
+		// fetch next 256 bytes of mmio list if necessary
+		and $r4 $r3 0xff
+		bra ne #ctx_mmio_pull
+			mov $r5 #xfer_data
+			sethi $r5 0x00060000	// 256 bytes
+			xdld $r3 $r5
+			xdwait
+
+		// execute a single list entry
+		ctx_mmio_pull:
+		ld b32 $r14 D[$r4 + #xfer_data + 0x00]
+		ld b32 $r15 D[$r4 + #xfer_data + 0x04]
+		call #nv_wr32
+
+		// next!
+		add b32 $r3 8
+		sub b32 $r1 1
+		bra ne #ctx_mmio_loop
+
+	// set transfer base back to the current context
+	ctx_mmio_done:
+	ld b32 $r3 D[$r0 + #ctx_current]
+	iowr I[$r2 + 0x000] $r3		// MEM_BASE
+
+	// disable the mmio list now, we don't need/want to execute it again
+	st b32 D[$r0 + #chan_mmio_count] $r0
+	mov $r1 #chan_data
+	sethi $r1 0x00060000		// 256 bytes
+	xdst $r0 $r1
+	xdwait
+	ret
+
+// Transfer HUB context data between GPU and storage area
+//
+// In: $r2 channel address
+//     $p1 clear on save, set on load
+//     $p2 set if opposite direction done/will be done, so:
+//		on save it means: "a load will follow this save"
+//		on load it means: "a save preceeded this load"
+//
+ctx_xfer:
+	bra not $p1 #ctx_xfer_pre
+	bra $p2 #ctx_xfer_pre_load
+	ctx_xfer_pre:
+		mov $r15 0x10
+		call #ctx_86c
+		bra not $p1 #ctx_xfer_exec
+
+	ctx_xfer_pre_load:
+		mov $r15 2
+		call #ctx_4170s
+		call #ctx_4170w
+		call #ctx_redswitch
+		clear b32 $r15
+		call #ctx_4170s
+		call #ctx_load
+
+	// fetch context pointer, and initiate xfer on all GPCs
+	ctx_xfer_exec:
+	ld b32 $r1 D[$r0 + #ctx_current]
+	mov $r2 0x414
+	shl b32 $r2 6
+	iowr I[$r2 + 0x000] $r0	// BAR_STATUS = reset
+	mov $r14 -0x5b00
+	sethi $r14 0x410000
+	mov b32 $r15 $r1
+	call #nv_wr32		// GPC_BCAST_WRCMD_DATA = ctx pointer
+	add b32 $r14 4
+	xbit $r15 $flags $p1
+	xbit $r2 $flags $p2
+	shl b32 $r2 1
+	or $r15 $r2
+	call #nv_wr32		// GPC_BCAST_WRCMD_CMD = GPC_XFER(type)
+
+	// strands
+	mov $r1 0x4afc
+	sethi $r1 0x20000
+	mov $r2 0xc
+	iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x0c
+	call #strand_wait
+	mov $r2 0x47fc
+	sethi $r2 0x20000
+	iowr I[$r2] $r0		// STRAND_FIRST_GENE(0x3f) = 0x00
+	xbit $r2 $flags $p1
+	add b32 $r2 3
+	iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x03/0x04 (SAVE/LOAD)
+
+	// mmio context
+	xbit $r10 $flags $p1	// direction
+	or $r10 6		// first, last
+	mov $r11 0		// base = 0
+	ld b32 $r12 D[$r0 + #hub_mmio_list_head]
+	ld b32 $r13 D[$r0 + #hub_mmio_list_tail]
+	mov $r14 0		// not multi
+	call #mmctx_xfer
+
+	// wait for GPCs to all complete
+	mov $r10 8		// DONE_BAR
+	call #wait_doneo
+
+	// wait for strand xfer to complete
+	call #strand_wait
+
+	// post-op
+	bra $p1 #ctx_xfer_post
+		mov $r10 12		// DONE_UNK12
+		call #wait_donez
+		mov $r1 0xa10
+		shl b32 $r1 6
+		mov $r2 5
+		iowr I[$r1] $r2		// MEM_CMD
+		ctx_xfer_post_save_wait:
+			iord $r2 I[$r1]
+			or $r2 $r2
+			bra ne #ctx_xfer_post_save_wait
+
+	bra $p2 #ctx_xfer_done
+	ctx_xfer_post:
+		mov $r15 2
+		call #ctx_4170s
+		clear b32 $r15
+		call #ctx_86c
+		call #strand_post
+		call #ctx_4170w
+		clear b32 $r15
+		call #ctx_4170s
+
+		bra not $p1 #ctx_xfer_no_post_mmio
+		ld b32 $r1 D[$r0 + #chan_mmio_count]
+		or $r1 $r1
+		bra e #ctx_xfer_no_post_mmio
+			call #ctx_mmio_exec
+
+		ctx_xfer_no_post_mmio:
+
+	ctx_xfer_done:
+	ret
+
+.align 256
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnve0.fuc.h b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnve0.fuc.h
new file mode 100644
index 0000000..decf0c6
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnve0.fuc.h
@@ -0,0 +1,857 @@
+uint32_t nve0_grhub_data[] = {
+/* 0x0000: gpc_count */
+	0x00000000,
+/* 0x0004: rop_count */
+	0x00000000,
+/* 0x0008: cmd_queue */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0050: hub_mmio_list_head */
+	0x00000000,
+/* 0x0054: hub_mmio_list_tail */
+	0x00000000,
+/* 0x0058: ctx_current */
+	0x00000000,
+/* 0x005c: chipsets */
+	0x000000e4,
+	0x013c0070,
+	0x000000e7,
+	0x013c0070,
+	0x00000000,
+/* 0x0070: nve4_hub_mmio_head */
+	0x0417e91c,
+	0x04400204,
+	0x18404010,
+	0x204040a8,
+	0x184040d0,
+	0x004040f8,
+	0x08404130,
+	0x08404150,
+	0x00404164,
+	0x0c4041a0,
+	0x0c404200,
+	0x34404404,
+	0x0c404460,
+	0x00404480,
+	0x00404498,
+	0x0c404604,
+	0x0c404618,
+	0x0440462c,
+	0x00404640,
+	0x00404654,
+	0x00404660,
+	0x48404678,
+	0x084046c8,
+	0x08404700,
+	0x24404718,
+	0x04404744,
+	0x00404754,
+	0x00405800,
+	0x08405830,
+	0x00405854,
+	0x0c405870,
+	0x04405a00,
+	0x00405a18,
+	0x00405b00,
+	0x00405b10,
+	0x00406020,
+	0x0c406028,
+	0x044064a8,
+	0x044064b4,
+	0x2c4064c0,
+	0x004064fc,
+	0x00407040,
+	0x00407804,
+	0x1440780c,
+	0x004078bc,
+	0x18408000,
+	0x00408064,
+	0x08408800,
+	0x00408840,
+	0x08408900,
+	0x00408980,
+/* 0x013c: nve4_hub_mmio_tail */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0200: chan_data */
+/* 0x0200: chan_mmio_count */
+	0x00000000,
+/* 0x0204: chan_mmio_address */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0300: xfer_data */
+	0x00000000,
+};
+
+uint32_t nve0_grhub_code[] = {
+	0x03090ef5,
+/* 0x0004: queue_put */
+	0x9800d898,
+	0x86f001d9,
+	0x0489b808,
+	0xf00c1bf4,
+	0x21f502f7,
+	0x00f802ec,
+/* 0x001c: queue_put_next */
+	0xb60798c4,
+	0x8dbb0384,
+	0x0880b600,
+	0x80008e80,
+	0x90b6018f,
+	0x0f94f001,
+	0xf801d980,
+/* 0x0039: queue_get */
+	0x0131f400,
+	0x9800d898,
+	0x89b801d9,
+	0x210bf404,
+	0xb60789c4,
+	0x9dbb0394,
+	0x0890b600,
+	0x98009e98,
+	0x80b6019f,
+	0x0f84f001,
+	0xf400d880,
+/* 0x0066: queue_get_done */
+	0x00f80132,
+/* 0x0068: nv_rd32 */
+	0x0728b7f1,
+	0xb906b4b6,
+	0xc9f002ec,
+	0x00bcd01f,
+/* 0x0078: nv_rd32_wait */
+	0xc800bccf,
+	0x1bf41fcc,
+	0x06a7f0fa,
+	0x010321f5,
+	0xf840bfcf,
+/* 0x008d: nv_wr32 */
+	0x28b7f100,
+	0x06b4b607,
+	0xb980bfd0,
+	0xc9f002ec,
+	0x1ec9f01f,
+/* 0x00a3: nv_wr32_wait */
+	0xcf00bcd0,
+	0xccc800bc,
+	0xfa1bf41f,
+/* 0x00ae: watchdog_reset */
+	0x87f100f8,
+	0x84b60430,
+	0x1ff9f006,
+	0xf8008fd0,
+/* 0x00bd: watchdog_clear */
+	0x3087f100,
+	0x0684b604,
+	0xf80080d0,
+/* 0x00c9: wait_donez */
+	0x3c87f100,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d000,
+	0x081887f1,
+	0xd00684b6,
+/* 0x00e2: wait_done_wait_donez */
+	0x87f1008a,
+	0x84b60400,
+	0x0088cf06,
+	0xf4888aff,
+	0x87f1f31b,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00099,
+/* 0x0103: wait_doneo */
+	0xf100f800,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00099f0,
+	0x87f10089,
+	0x84b60818,
+	0x008ad006,
+/* 0x011c: wait_done_wait_doneo */
+	0x040087f1,
+	0xcf0684b6,
+	0x8aff0088,
+	0xf30bf488,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0099f094,
+	0xf80089d0,
+/* 0x013d: mmctx_size */
+/* 0x013f: nv_mmctx_size_loop */
+	0x9894bd00,
+	0x85b600e8,
+	0x0180b61a,
+	0xbb0284b6,
+	0xe0b60098,
+	0x04efb804,
+	0xb9eb1bf4,
+	0x00f8029f,
+/* 0x015c: mmctx_xfer */
+	0x083c87f1,
+	0xbd0684b6,
+	0x0199f094,
+	0xf10089d0,
+	0xb6071087,
+	0x94bd0684,
+	0xf405bbfd,
+	0x8bd0090b,
+	0x0099f000,
+/* 0x0180: mmctx_base_disabled */
+	0xf405eefd,
+	0x8ed00c0b,
+	0xc08fd080,
+/* 0x018f: mmctx_multi_disabled */
+	0xb70199f0,
+	0xc8010080,
+	0xb4b600ab,
+	0x0cb9f010,
+	0xb601aec8,
+	0xbefd11e4,
+	0x008bd005,
+/* 0x01a8: mmctx_exec_loop */
+/* 0x01a8: mmctx_wait_free */
+	0xf0008ecf,
+	0x0bf41fe4,
+	0x00ce98fa,
+	0xd005e9fd,
+	0xc0b6c08e,
+	0x04cdb804,
+	0xc8e81bf4,
+	0x1bf402ab,
+/* 0x01c9: mmctx_fini_wait */
+	0x008bcf18,
+	0xb01fb4f0,
+	0x1bf410b4,
+	0x02a7f0f7,
+	0xf4c921f4,
+/* 0x01de: mmctx_stop */
+	0xabc81b0e,
+	0x10b4b600,
+	0xf00cb9f0,
+	0x8bd012b9,
+/* 0x01ed: mmctx_stop_wait */
+	0x008bcf00,
+	0xf412bbc8,
+/* 0x01f6: mmctx_done */
+	0x87f1fa1b,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00199,
+/* 0x0207: strand_wait */
+	0xf900f800,
+	0x02a7f0a0,
+	0xfcc921f4,
+/* 0x0213: strand_pre */
+	0xf100f8a0,
+	0xf04afc87,
+	0x97f00283,
+	0x0089d00c,
+	0x020721f5,
+/* 0x0226: strand_post */
+	0x87f100f8,
+	0x83f04afc,
+	0x0d97f002,
+	0xf50089d0,
+	0xf8020721,
+/* 0x0239: strand_set */
+	0xfca7f100,
+	0x02a3f04f,
+	0x0500aba2,
+	0xd00fc7f0,
+	0xc7f000ac,
+	0x00bcd00b,
+	0x020721f5,
+	0xf000aed0,
+	0xbcd00ac7,
+	0x0721f500,
+/* 0x0263: strand_ctx_init */
+	0xf100f802,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00399f0,
+	0x21f50089,
+	0xe7f00213,
+	0x3921f503,
+	0xfca7f102,
+	0x02a3f046,
+	0x0400aba0,
+	0xf040a0d0,
+	0xbcd001c7,
+	0x0721f500,
+	0x010c9202,
+	0xf000acd0,
+	0xbcd002c7,
+	0x0721f500,
+	0x2621f502,
+	0x8087f102,
+	0x0684b608,
+	0xb70089cf,
+	0x95220080,
+/* 0x02ba: ctx_init_strand_loop */
+	0x8ed008fe,
+	0x408ed000,
+	0xb6808acf,
+	0xa0b606a5,
+	0x00eabb01,
+	0xb60480b6,
+	0x1bf40192,
+	0x08e4b6e8,
+	0xf1f2efbc,
+	0xb6085c87,
+	0x94bd0684,
+	0xd00399f0,
+	0x00f80089,
+/* 0x02ec: error */
+	0xe7f1e0f9,
+	0xe4b60814,
+	0x00efd006,
+	0x0c1ce7f1,
+	0xf006e4b6,
+	0xefd001f7,
+	0xf8e0fc00,
+/* 0x0309: init */
+	0xfe04bd00,
+	0x07fe0004,
+	0x0017f100,
+	0x0227f012,
+	0xf10012d0,
+	0xfe05b917,
+	0x17f10010,
+	0x10d00400,
+	0x0437f1c0,
+	0x0634b604,
+	0x200327f1,
+	0xf10032d0,
+	0xd0200427,
+	0x27f10132,
+	0x32d0200b,
+	0x0c27f102,
+	0x0732d020,
+	0x0c2427f1,
+	0xb90624b6,
+	0x23d00003,
+	0x0427f100,
+	0x0023f087,
+	0xb70012d0,
+	0xf0010012,
+	0x12d00427,
+	0x1031f400,
+	0x9604e7f1,
+	0xf440e3f0,
+	0xf1c76821,
+	0x01018090,
+	0x801ff4f0,
+	0x17f0000f,
+	0x041fbb01,
+	0xf10112b6,
+	0xb6040c27,
+	0x21d00624,
+	0x4021d000,
+	0x080027f1,
+	0xcf0624b6,
+	0xf7f00022,
+/* 0x03a9: init_find_chipset */
+	0x08f0b654,
+	0xb800f398,
+	0x0bf40432,
+	0x0034b00b,
+	0xf8f11bf4,
+/* 0x03bd: init_context */
+	0x0017f100,
+	0x02fe5801,
+	0xf003ff58,
+	0x0e8000e3,
+	0x150f8014,
+	0x013d21f5,
+	0x070037f1,
+	0x950634b6,
+	0x34d00814,
+	0x4034d000,
+	0x130030b7,
+	0xb6001fbb,
+	0x3fd002f5,
+	0x0815b600,
+	0xb60110b6,
+	0x1fb90814,
+	0x6321f502,
+	0x001fbb02,
+	0xf1000398,
+	0xf0200047,
+/* 0x040e: init_gpc */
+	0x4ea05043,
+	0x1fb90804,
+	0x8d21f402,
+	0x08004ea0,
+	0xf4022fb9,
+	0x4ea08d21,
+	0xf4bd010c,
+	0xa08d21f4,
+	0xf401044e,
+	0x4ea08d21,
+	0xf7f00100,
+	0x8d21f402,
+	0x08004ea0,
+/* 0x0440: init_gpc_wait */
+	0xc86821f4,
+	0x0bf41fff,
+	0x044ea0fa,
+	0x6821f408,
+	0xb7001fbb,
+	0xb6800040,
+	0x1bf40132,
+	0x0027f1b4,
+	0x0624b608,
+	0xb74021d0,
+	0xbd080020,
+	0x1f19f014,
+/* 0x0473: main */
+	0xf40021d0,
+	0x28f40031,
+	0x08d7f000,
+	0xf43921f4,
+	0xe4b1f401,
+	0x1bf54001,
+	0x87f100d1,
+	0x84b6083c,
+	0xf094bd06,
+	0x89d00499,
+	0x0017f100,
+	0x0614b60b,
+	0xcf4012cf,
+	0x13c80011,
+	0x7e0bf41f,
+	0xf41f23c8,
+	0x20f95a0b,
+	0xf10212b9,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00799f0,
+	0x32f40089,
+	0x0231f401,
+	0x07fb21f5,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0799f094,
+	0xfc0089d0,
+	0x3c87f120,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d006,
+	0xf50131f4,
+	0xf107fb21,
+	0xb6085c87,
+	0x94bd0684,
+	0xd00699f0,
+	0x0ef40089,
+/* 0x0509: chsw_prev_no_next */
+	0xb920f931,
+	0x32f40212,
+	0x0232f401,
+	0x07fb21f5,
+	0x17f120fc,
+	0x14b60b00,
+	0x0012d006,
+/* 0x0527: chsw_no_prev */
+	0xc8130ef4,
+	0x0bf41f23,
+	0x0131f40d,
+	0xf50232f4,
+/* 0x0537: chsw_done */
+	0xf107fb21,
+	0xb60b0c17,
+	0x27f00614,
+	0x0012d001,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0499f094,
+	0xf50089d0,
+/* 0x0557: main_not_ctx_switch */
+	0xb0ff200e,
+	0x1bf401e4,
+	0x02f2b90d,
+	0x078f21f5,
+/* 0x0567: main_not_ctx_chan */
+	0xb0420ef4,
+	0x1bf402e4,
+	0x3c87f12e,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d007,
+	0xf40132f4,
+	0x21f50232,
+	0x87f107fb,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00799,
+	0x110ef400,
+/* 0x0598: main_not_ctx_save */
+	0xf010ef94,
+	0x21f501f5,
+	0x0ef502ec,
+/* 0x05a6: main_done */
+	0x17f1fed1,
+	0x14b60820,
+	0xf024bd06,
+	0x12d01f29,
+	0xbe0ef500,
+/* 0x05b9: ih */
+	0xfe80f9fe,
+	0x80f90188,
+	0xa0f990f9,
+	0xd0f9b0f9,
+	0xf0f9e0f9,
+	0xc4800acf,
+	0x0bf404ab,
+	0x00b7f11d,
+	0x08d7f019,
+	0xcf40becf,
+	0x21f400bf,
+	0x00b0b704,
+	0x01e7f004,
+/* 0x05ef: ih_no_fifo */
+	0xe400bed0,
+	0xf40100ab,
+	0xd7f00d0b,
+	0x01e7f108,
+	0x0421f440,
+/* 0x0600: ih_no_ctxsw */
+	0x0104b7f1,
+	0xabffb0bd,
+	0x0d0bf4b4,
+	0x0c1ca7f1,
+	0xd006a4b6,
+/* 0x0616: ih_no_other */
+	0x0ad000ab,
+	0xfcf0fc40,
+	0xfcd0fce0,
+	0xfca0fcb0,
+	0xfe80fc90,
+	0x80fc0088,
+	0xf80032f4,
+/* 0x0631: ctx_4170s */
+	0x70e7f101,
+	0x40e3f041,
+	0xf410f5f0,
+	0x00f88d21,
+/* 0x0640: ctx_4170w */
+	0x4170e7f1,
+	0xf440e3f0,
+	0xf4f06821,
+	0xf31bf410,
+/* 0x0652: ctx_redswitch */
+	0xe7f100f8,
+	0xe4b60614,
+	0x70f7f106,
+	0x00efd002,
+/* 0x0663: ctx_redswitch_delay */
+	0xb608f7f0,
+	0x1bf401f2,
+	0x70f7f1fd,
+	0x00efd007,
+/* 0x0672: ctx_86c */
+	0xe7f100f8,
+	0xe4b6086c,
+	0x00efd006,
+	0x8a14e7f1,
+	0xf440e3f0,
+	0xe7f18d21,
+	0xe3f0a86c,
+	0x8d21f441,
+/* 0x0692: ctx_load */
+	0x87f100f8,
+	0x84b6083c,
+	0xf094bd06,
+	0x89d00599,
+	0x0ca7f000,
+	0xf1c921f4,
+	0xb60a2417,
+	0x10d00614,
+	0x0037f100,
+	0x0634b60b,
+	0xf14032d0,
+	0xb60a0c17,
+	0x47f00614,
+	0x0012d007,
+/* 0x06cb: ctx_chan_wait_0 */
+	0xcf4014d0,
+	0x44f04014,
+	0xfa1bf41f,
+	0xfe0032d0,
+	0x2af0000b,
+	0x0424b61f,
+	0xf10220b6,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00899f0,
+	0x17f10089,
+	0x14b60a04,
+	0x0012d006,
+	0x0a2017f1,
+	0xf00614b6,
+	0x23f10227,
+	0x12d08000,
+	0x1017f000,
+	0x030027f1,
+	0xfa0223f0,
+	0x03f80512,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0899f094,
+	0x980089d0,
+	0x14b6c101,
+	0xc0029818,
+	0xfd0825b6,
+	0x01800512,
+	0x3c87f116,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d009,
+	0x0a0427f1,
+	0xd00624b6,
+	0x27f00021,
+	0x2017f101,
+	0x0614b60a,
+	0xf10012d0,
+	0xf0020017,
+	0x01fa0613,
+	0xf103f805,
+	0xb6085c87,
+	0x94bd0684,
+	0xd00999f0,
+	0x87f10089,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00599,
+/* 0x078f: ctx_chan */
+	0xf500f800,
+	0xf0069221,
+	0x21f40ca7,
+	0x1017f1c9,
+	0x0614b60a,
+	0xd00527f0,
+/* 0x07a6: ctx_chan_wait */
+	0x12cf0012,
+	0x0522fd00,
+	0xf8fa1bf4,
+/* 0x07b1: ctx_mmio_exec */
+	0x81039800,
+	0x0a0427f1,
+	0xd00624b6,
+	0x34bd0023,
+/* 0x07c0: ctx_mmio_loop */
+	0xf4ff34c4,
+	0x57f10f1b,
+	0x53f00300,
+	0x0535fa06,
+/* 0x07d2: ctx_mmio_pull */
+	0x4e9803f8,
+	0xc14f98c0,
+	0xb68d21f4,
+	0x12b60830,
+	0xdf1bf401,
+/* 0x07e4: ctx_mmio_done */
+	0xd0160398,
+	0x00800023,
+	0x0017f180,
+	0x0613f002,
+	0xf80601fa,
+/* 0x07fb: ctx_xfer */
+	0xf400f803,
+	0x02f40611,
+/* 0x0801: ctx_xfer_pre */
+	0x10f7f00d,
+	0x067221f5,
+/* 0x080b: ctx_xfer_pre_load */
+	0xf01c11f4,
+	0x21f502f7,
+	0x21f50631,
+	0x21f50640,
+	0xf4bd0652,
+	0x063121f5,
+	0x069221f5,
+/* 0x0824: ctx_xfer_exec */
+	0xf1160198,
+	0xb6041427,
+	0x20d00624,
+	0x00e7f100,
+	0x41e3f0a5,
+	0xf4021fb9,
+	0xe0b68d21,
+	0x01fcf004,
+	0xb6022cf0,
+	0xf2fd0124,
+	0x8d21f405,
+	0x4afc17f1,
+	0xf00213f0,
+	0x12d00c27,
+	0x0721f500,
+	0xfc27f102,
+	0x0223f047,
+	0xf00020d0,
+	0x20b6012c,
+	0x0012d003,
+	0xf001acf0,
+	0xb7f006a5,
+	0x140c9800,
+	0xf0150d98,
+	0x21f500e7,
+	0xa7f0015c,
+	0x0321f508,
+	0x0721f501,
+	0x2201f402,
+	0xf40ca7f0,
+	0x17f1c921,
+	0x14b60a10,
+	0x0527f006,
+/* 0x08ab: ctx_xfer_post_save_wait */
+	0xcf0012d0,
+	0x22fd0012,
+	0xfa1bf405,
+/* 0x08b7: ctx_xfer_post */
+	0xf02e02f4,
+	0x21f502f7,
+	0xf4bd0631,
+	0x067221f5,
+	0x022621f5,
+	0x064021f5,
+	0x21f5f4bd,
+	0x11f40631,
+	0x80019810,
+	0xf40511fd,
+	0x21f5070b,
+/* 0x08e2: ctx_xfer_no_post_mmio */
+/* 0x08e2: ctx_xfer_done */
+	0x00f807b1,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/nvc0.fuc b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/nvc0.fuc
new file mode 100644
index 0000000..e6b2288
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/nvc0.fuc
@@ -0,0 +1,400 @@
+/* fuc microcode util functions for nvc0 PGRAPH
+ *
+ * Copyright 2011 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+define(`mmctx_data', `.b32 eval((($2 - 1) << 26) | $1)')
+define(`queue_init', `.skip eval((2 * 4) + ((8 * 4) * 2))')
+
+ifdef(`include_code', `
+// Error codes
+define(`E_BAD_COMMAND', 0x01)
+define(`E_CMD_OVERFLOW', 0x02)
+
+// Util macros to help with debugging ucode hangs etc
+define(`T_WAIT', 0)
+define(`T_MMCTX', 1)
+define(`T_STRWAIT', 2)
+define(`T_STRINIT', 3)
+define(`T_AUTO', 4)
+define(`T_CHAN', 5)
+define(`T_LOAD', 6)
+define(`T_SAVE', 7)
+define(`T_LCHAN', 8)
+define(`T_LCTXH', 9)
+
+define(`trace_set', `
+	mov $r8 0x83c
+	shl b32 $r8 6
+	clear b32 $r9
+	bset $r9 $1
+	iowr I[$r8 + 0x000] $r9		// CC_SCRATCH[7]
+')
+
+define(`trace_clr', `
+	mov $r8 0x85c
+	shl b32 $r8 6
+	clear b32 $r9
+	bset $r9 $1
+	iowr I[$r8 + 0x000] $r9		// CC_SCRATCH[7]
+')
+
+// queue_put - add request to queue
+//
+// In : $r13 queue pointer
+//	$r14 command
+//	$r15 data
+//
+queue_put:
+	// make sure we have space..
+	ld b32 $r8 D[$r13 + 0x0]	// GET
+	ld b32 $r9 D[$r13 + 0x4]	// PUT
+	xor $r8 8
+	cmpu b32 $r8 $r9
+	bra ne #queue_put_next
+		mov $r15 E_CMD_OVERFLOW
+		call #error
+		ret
+
+	// store cmd/data on queue
+	queue_put_next:
+	and $r8 $r9 7
+	shl b32 $r8 3
+	add b32 $r8 $r13
+	add b32 $r8 8
+	st b32 D[$r8 + 0x0] $r14
+	st b32 D[$r8 + 0x4] $r15
+
+	// update PUT
+	add b32 $r9 1
+	and $r9 0xf
+	st b32 D[$r13 + 0x4] $r9
+	ret
+
+// queue_get - fetch request from queue
+//
+// In : $r13 queue pointer
+//
+// Out:	$p1  clear on success (data available)
+//	$r14 command
+// 	$r15 data
+//
+queue_get:
+	bset $flags $p1
+	ld b32 $r8 D[$r13 + 0x0]	// GET
+	ld b32 $r9 D[$r13 + 0x4]	// PUT
+	cmpu b32 $r8 $r9
+	bra e #queue_get_done
+		// fetch first cmd/data pair
+		and $r9 $r8 7
+		shl b32 $r9 3
+		add b32 $r9 $r13
+		add b32 $r9 8
+		ld b32 $r14 D[$r9 + 0x0]
+		ld b32 $r15 D[$r9 + 0x4]
+
+		// update GET
+		add b32 $r8 1
+		and $r8 0xf
+		st b32 D[$r13 + 0x0] $r8
+		bclr $flags $p1
+queue_get_done:
+	ret
+
+// nv_rd32 - read 32-bit value from nv register
+//
+// In : $r14 register
+// Out: $r15 value
+//
+nv_rd32:
+	mov $r11 0x728
+	shl b32 $r11 6
+	mov b32 $r12 $r14
+	bset $r12 31			// MMIO_CTRL_PENDING
+	iowr I[$r11 + 0x000] $r12	// MMIO_CTRL
+	nv_rd32_wait:
+		iord $r12 I[$r11 + 0x000]
+		xbit $r12 $r12 31
+		bra ne #nv_rd32_wait
+	mov $r10 6			// DONE_MMIO_RD
+	call #wait_doneo
+	iord $r15 I[$r11 + 0x100]	// MMIO_RDVAL
+	ret
+
+// nv_wr32 - write 32-bit value to nv register
+//
+// In : $r14 register
+//      $r15 value
+//
+nv_wr32:
+	mov $r11 0x728
+	shl b32 $r11 6
+	iowr I[$r11 + 0x200] $r15	// MMIO_WRVAL
+	mov b32 $r12 $r14
+	bset $r12 31			// MMIO_CTRL_PENDING
+	bset $r12 30			// MMIO_CTRL_WRITE
+	iowr I[$r11 + 0x000] $r12	// MMIO_CTRL
+	nv_wr32_wait:
+		iord $r12 I[$r11 + 0x000]
+		xbit $r12 $r12 31
+		bra ne #nv_wr32_wait
+	ret
+
+// (re)set watchdog timer
+//
+// In : $r15 timeout
+//
+watchdog_reset:
+	mov $r8 0x430
+	shl b32 $r8 6
+	bset $r15 31
+	iowr I[$r8 + 0x000] $r15
+	ret
+
+// clear watchdog timer
+watchdog_clear:
+	mov $r8 0x430
+	shl b32 $r8 6
+	iowr I[$r8 + 0x000] $r0
+	ret
+
+// wait_done{z,o} - wait on FUC_DONE bit to become clear/set
+//
+// In : $r10 bit to wait on
+//
+define(`wait_done', `
+$1:
+	trace_set(T_WAIT);
+	mov $r8 0x818
+	shl b32 $r8 6
+	iowr I[$r8 + 0x000] $r10	// CC_SCRATCH[6] = wait bit
+	wait_done_$1:
+		mov $r8 0x400
+		shl b32 $r8 6
+		iord $r8 I[$r8 + 0x000]	// DONE
+		xbit $r8 $r8 $r10
+		bra $2 #wait_done_$1
+	trace_clr(T_WAIT)
+	ret
+')
+wait_done(wait_donez, ne)
+wait_done(wait_doneo, e)
+
+// mmctx_size - determine size of a mmio list transfer
+//
+// In : $r14 mmio list head
+//      $r15 mmio list tail
+// Out: $r15 transfer size (in bytes)
+//
+mmctx_size:
+	clear b32 $r9
+	nv_mmctx_size_loop:
+		ld b32 $r8 D[$r14]
+		shr b32 $r8 26
+		add b32 $r8 1
+		shl b32 $r8 2
+		add b32 $r9 $r8
+		add b32 $r14 4
+		cmpu b32 $r14 $r15
+		bra ne #nv_mmctx_size_loop
+	mov b32 $r15 $r9
+	ret
+
+// mmctx_xfer - execute a list of mmio transfers
+//
+// In : $r10 flags
+//		bit 0: direction (0 = save, 1 = load)
+//		bit 1: set if first transfer
+//		bit 2: set if last transfer
+//	$r11 base
+//	$r12 mmio list head
+//	$r13 mmio list tail
+//	$r14 multi_stride
+//	$r15 multi_mask
+//
+mmctx_xfer:
+	trace_set(T_MMCTX)
+	mov $r8 0x710
+	shl b32 $r8 6
+	clear b32 $r9
+	or $r11 $r11
+	bra e #mmctx_base_disabled
+		iowr I[$r8 + 0x000] $r11	// MMCTX_BASE
+		bset $r9 0			// BASE_EN
+	mmctx_base_disabled:
+	or $r14 $r14
+	bra e #mmctx_multi_disabled
+		iowr I[$r8 + 0x200] $r14 	// MMCTX_MULTI_STRIDE
+		iowr I[$r8 + 0x300] $r15 	// MMCTX_MULTI_MASK
+		bset $r9 1			// MULTI_EN
+	mmctx_multi_disabled:
+	add b32 $r8 0x100
+
+	xbit $r11 $r10 0
+	shl b32 $r11 16			// DIR
+	bset $r11 12			// QLIMIT = 0x10
+	xbit $r14 $r10 1
+	shl b32 $r14 17
+	or $r11 $r14			// START_TRIGGER
+	iowr I[$r8 + 0x000] $r11	// MMCTX_CTRL
+
+	// loop over the mmio list, and send requests to the hw
+	mmctx_exec_loop:
+		// wait for space in mmctx queue
+		mmctx_wait_free:
+			iord $r14 I[$r8 + 0x000] // MMCTX_CTRL
+			and $r14 0x1f
+			bra e #mmctx_wait_free
+
+		// queue up an entry
+		ld b32 $r14 D[$r12]
+		or $r14 $r9
+		iowr I[$r8 + 0x300] $r14
+		add b32 $r12 4
+		cmpu b32 $r12 $r13
+		bra ne #mmctx_exec_loop
+
+	xbit $r11 $r10 2
+	bra ne #mmctx_stop
+		// wait for queue to empty
+		mmctx_fini_wait:
+			iord $r11 I[$r8 + 0x000]	// MMCTX_CTRL
+			and $r11 0x1f
+			cmpu b32 $r11 0x10
+			bra ne #mmctx_fini_wait
+		mov $r10 2				// DONE_MMCTX
+		call #wait_donez
+		bra #mmctx_done
+	mmctx_stop:
+		xbit $r11 $r10 0
+		shl b32 $r11 16			// DIR
+		bset $r11 12			// QLIMIT = 0x10
+		bset $r11 18			// STOP_TRIGGER
+		iowr I[$r8 + 0x000] $r11	// MMCTX_CTRL
+		mmctx_stop_wait:
+			// wait for STOP_TRIGGER to clear
+			iord $r11 I[$r8 + 0x000] // MMCTX_CTRL
+			xbit $r11 $r11 18
+			bra ne #mmctx_stop_wait
+	mmctx_done:
+	trace_clr(T_MMCTX)
+	ret
+
+// Wait for DONE_STRAND
+//
+strand_wait:
+	push $r10
+	mov $r10 2
+	call #wait_donez
+	pop $r10
+	ret
+
+// unknown - call before issuing strand commands
+//
+strand_pre:
+	mov $r8 0x4afc
+	sethi $r8 0x20000
+	mov $r9 0xc
+	iowr I[$r8] $r9
+	call #strand_wait
+	ret
+
+// unknown - call after issuing strand commands
+//
+strand_post:
+	mov $r8 0x4afc
+	sethi $r8 0x20000
+	mov $r9 0xd
+	iowr I[$r8] $r9
+	call #strand_wait
+	ret
+
+// Selects strand set?!
+//
+// In: $r14 id
+//
+strand_set:
+	mov $r10 0x4ffc
+	sethi $r10 0x20000
+	sub b32 $r11 $r10 0x500
+	mov $r12 0xf
+	iowr I[$r10 + 0x000] $r12		// 0x93c = 0xf
+	mov $r12 0xb
+	iowr I[$r11 + 0x000] $r12		// 0x928 = 0xb
+	call #strand_wait
+	iowr I[$r10 + 0x000] $r14		// 0x93c = <id>
+	mov $r12 0xa
+	iowr I[$r11 + 0x000] $r12		// 0x928 = 0xa
+	call #strand_wait
+	ret
+
+// Initialise strand context data
+//
+// In : $r15 context base
+// Out: $r15 context size (in bytes)
+//
+// Strandset(?) 3 hardcoded currently
+//
+strand_ctx_init:
+	trace_set(T_STRINIT)
+	call #strand_pre
+	mov $r14 3
+	call #strand_set
+	mov $r10 0x46fc
+	sethi $r10 0x20000
+	add b32 $r11 $r10 0x400
+	iowr I[$r10 + 0x100] $r0	// STRAND_FIRST_GENE = 0
+	mov $r12 1
+	iowr I[$r11 + 0x000] $r12	// STRAND_CMD = LATCH_FIRST_GENE
+	call #strand_wait
+	sub b32 $r12 $r0 1
+	iowr I[$r10 + 0x000] $r12	// STRAND_GENE_CNT = 0xffffffff
+	mov $r12 2
+	iowr I[$r11 + 0x000] $r12	// STRAND_CMD = LATCH_GENE_CNT
+	call #strand_wait
+	call #strand_post
+
+	// read the size of each strand, poke the context offset of
+	// each into STRAND_{SAVE,LOAD}_SWBASE now, no need to worry
+	// about it later then.
+	mov $r8 0x880
+	shl b32 $r8 6
+	iord $r9 I[$r8 + 0x000]		// STRANDS
+	add b32 $r8 0x2200
+	shr b32 $r14 $r15 8
+	ctx_init_strand_loop:
+		iowr I[$r8 + 0x000] $r14	// STRAND_SAVE_SWBASE
+		iowr I[$r8 + 0x100] $r14	// STRAND_LOAD_SWBASE
+		iord $r10 I[$r8 + 0x200]	// STRAND_SIZE
+		shr b32 $r10 6
+		add b32 $r10 1
+		add b32 $r14 $r10
+		add b32 $r8 4
+		sub b32 $r9 1
+		bra ne #ctx_init_strand_loop
+
+	shl b32 $r14 8
+	sub b32 $r15 $r14 $r15
+	trace_clr(T_STRINIT)
+	ret
+')
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/nve0.fuc b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/nve0.fuc
new file mode 100644
index 0000000..f16a5d5
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/nve0.fuc
@@ -0,0 +1,400 @@
+/* fuc microcode util functions for nve0 PGRAPH
+ *
+ * Copyright 2011 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+define(`mmctx_data', `.b32 eval((($2 - 1) << 26) | $1)')
+define(`queue_init', `.skip eval((2 * 4) + ((8 * 4) * 2))')
+
+ifdef(`include_code', `
+// Error codes
+define(`E_BAD_COMMAND', 0x01)
+define(`E_CMD_OVERFLOW', 0x02)
+
+// Util macros to help with debugging ucode hangs etc
+define(`T_WAIT', 0)
+define(`T_MMCTX', 1)
+define(`T_STRWAIT', 2)
+define(`T_STRINIT', 3)
+define(`T_AUTO', 4)
+define(`T_CHAN', 5)
+define(`T_LOAD', 6)
+define(`T_SAVE', 7)
+define(`T_LCHAN', 8)
+define(`T_LCTXH', 9)
+
+define(`trace_set', `
+	mov $r8 0x83c
+	shl b32 $r8 6
+	clear b32 $r9
+	bset $r9 $1
+	iowr I[$r8 + 0x000] $r9		// CC_SCRATCH[7]
+')
+
+define(`trace_clr', `
+	mov $r8 0x85c
+	shl b32 $r8 6
+	clear b32 $r9
+	bset $r9 $1
+	iowr I[$r8 + 0x000] $r9		// CC_SCRATCH[7]
+')
+
+// queue_put - add request to queue
+//
+// In : $r13 queue pointer
+//	$r14 command
+//	$r15 data
+//
+queue_put:
+	// make sure we have space..
+	ld b32 $r8 D[$r13 + 0x0]	// GET
+	ld b32 $r9 D[$r13 + 0x4]	// PUT
+	xor $r8 8
+	cmpu b32 $r8 $r9
+	bra ne #queue_put_next
+		mov $r15 E_CMD_OVERFLOW
+		call #error
+		ret
+
+	// store cmd/data on queue
+	queue_put_next:
+	and $r8 $r9 7
+	shl b32 $r8 3
+	add b32 $r8 $r13
+	add b32 $r8 8
+	st b32 D[$r8 + 0x0] $r14
+	st b32 D[$r8 + 0x4] $r15
+
+	// update PUT
+	add b32 $r9 1
+	and $r9 0xf
+	st b32 D[$r13 + 0x4] $r9
+	ret
+
+// queue_get - fetch request from queue
+//
+// In : $r13 queue pointer
+//
+// Out:	$p1  clear on success (data available)
+//	$r14 command
+// 	$r15 data
+//
+queue_get:
+	bset $flags $p1
+	ld b32 $r8 D[$r13 + 0x0]	// GET
+	ld b32 $r9 D[$r13 + 0x4]	// PUT
+	cmpu b32 $r8 $r9
+	bra e #queue_get_done
+		// fetch first cmd/data pair
+		and $r9 $r8 7
+		shl b32 $r9 3
+		add b32 $r9 $r13
+		add b32 $r9 8
+		ld b32 $r14 D[$r9 + 0x0]
+		ld b32 $r15 D[$r9 + 0x4]
+
+		// update GET
+		add b32 $r8 1
+		and $r8 0xf
+		st b32 D[$r13 + 0x0] $r8
+		bclr $flags $p1
+queue_get_done:
+	ret
+
+// nv_rd32 - read 32-bit value from nv register
+//
+// In : $r14 register
+// Out: $r15 value
+//
+nv_rd32:
+	mov $r11 0x728
+	shl b32 $r11 6
+	mov b32 $r12 $r14
+	bset $r12 31			// MMIO_CTRL_PENDING
+	iowr I[$r11 + 0x000] $r12	// MMIO_CTRL
+	nv_rd32_wait:
+		iord $r12 I[$r11 + 0x000]
+		xbit $r12 $r12 31
+		bra ne #nv_rd32_wait
+	mov $r10 6			// DONE_MMIO_RD
+	call #wait_doneo
+	iord $r15 I[$r11 + 0x100]	// MMIO_RDVAL
+	ret
+
+// nv_wr32 - write 32-bit value to nv register
+//
+// In : $r14 register
+//      $r15 value
+//
+nv_wr32:
+	mov $r11 0x728
+	shl b32 $r11 6
+	iowr I[$r11 + 0x200] $r15	// MMIO_WRVAL
+	mov b32 $r12 $r14
+	bset $r12 31			// MMIO_CTRL_PENDING
+	bset $r12 30			// MMIO_CTRL_WRITE
+	iowr I[$r11 + 0x000] $r12	// MMIO_CTRL
+	nv_wr32_wait:
+		iord $r12 I[$r11 + 0x000]
+		xbit $r12 $r12 31
+		bra ne #nv_wr32_wait
+	ret
+
+// (re)set watchdog timer
+//
+// In : $r15 timeout
+//
+watchdog_reset:
+	mov $r8 0x430
+	shl b32 $r8 6
+	bset $r15 31
+	iowr I[$r8 + 0x000] $r15
+	ret
+
+// clear watchdog timer
+watchdog_clear:
+	mov $r8 0x430
+	shl b32 $r8 6
+	iowr I[$r8 + 0x000] $r0
+	ret
+
+// wait_done{z,o} - wait on FUC_DONE bit to become clear/set
+//
+// In : $r10 bit to wait on
+//
+define(`wait_done', `
+$1:
+	trace_set(T_WAIT);
+	mov $r8 0x818
+	shl b32 $r8 6
+	iowr I[$r8 + 0x000] $r10	// CC_SCRATCH[6] = wait bit
+	wait_done_$1:
+		mov $r8 0x400
+		shl b32 $r8 6
+		iord $r8 I[$r8 + 0x000]	// DONE
+		xbit $r8 $r8 $r10
+		bra $2 #wait_done_$1
+	trace_clr(T_WAIT)
+	ret
+')
+wait_done(wait_donez, ne)
+wait_done(wait_doneo, e)
+
+// mmctx_size - determine size of a mmio list transfer
+//
+// In : $r14 mmio list head
+//      $r15 mmio list tail
+// Out: $r15 transfer size (in bytes)
+//
+mmctx_size:
+	clear b32 $r9
+	nv_mmctx_size_loop:
+		ld b32 $r8 D[$r14]
+		shr b32 $r8 26
+		add b32 $r8 1
+		shl b32 $r8 2
+		add b32 $r9 $r8
+		add b32 $r14 4
+		cmpu b32 $r14 $r15
+		bra ne #nv_mmctx_size_loop
+	mov b32 $r15 $r9
+	ret
+
+// mmctx_xfer - execute a list of mmio transfers
+//
+// In : $r10 flags
+//		bit 0: direction (0 = save, 1 = load)
+//		bit 1: set if first transfer
+//		bit 2: set if last transfer
+//	$r11 base
+//	$r12 mmio list head
+//	$r13 mmio list tail
+//	$r14 multi_stride
+//	$r15 multi_mask
+//
+mmctx_xfer:
+	trace_set(T_MMCTX)
+	mov $r8 0x710
+	shl b32 $r8 6
+	clear b32 $r9
+	or $r11 $r11
+	bra e #mmctx_base_disabled
+		iowr I[$r8 + 0x000] $r11	// MMCTX_BASE
+		bset $r9 0			// BASE_EN
+	mmctx_base_disabled:
+	or $r14 $r14
+	bra e #mmctx_multi_disabled
+		iowr I[$r8 + 0x200] $r14 	// MMCTX_MULTI_STRIDE
+		iowr I[$r8 + 0x300] $r15 	// MMCTX_MULTI_MASK
+		bset $r9 1			// MULTI_EN
+	mmctx_multi_disabled:
+	add b32 $r8 0x100
+
+	xbit $r11 $r10 0
+	shl b32 $r11 16			// DIR
+	bset $r11 12			// QLIMIT = 0x10
+	xbit $r14 $r10 1
+	shl b32 $r14 17
+	or $r11 $r14			// START_TRIGGER
+	iowr I[$r8 + 0x000] $r11	// MMCTX_CTRL
+
+	// loop over the mmio list, and send requests to the hw
+	mmctx_exec_loop:
+		// wait for space in mmctx queue
+		mmctx_wait_free:
+			iord $r14 I[$r8 + 0x000] // MMCTX_CTRL
+			and $r14 0x1f
+			bra e #mmctx_wait_free
+
+		// queue up an entry
+		ld b32 $r14 D[$r12]
+		or $r14 $r9
+		iowr I[$r8 + 0x300] $r14
+		add b32 $r12 4
+		cmpu b32 $r12 $r13
+		bra ne #mmctx_exec_loop
+
+	xbit $r11 $r10 2
+	bra ne #mmctx_stop
+		// wait for queue to empty
+		mmctx_fini_wait:
+			iord $r11 I[$r8 + 0x000]	// MMCTX_CTRL
+			and $r11 0x1f
+			cmpu b32 $r11 0x10
+			bra ne #mmctx_fini_wait
+		mov $r10 2				// DONE_MMCTX
+		call #wait_donez
+		bra #mmctx_done
+	mmctx_stop:
+		xbit $r11 $r10 0
+		shl b32 $r11 16			// DIR
+		bset $r11 12			// QLIMIT = 0x10
+		bset $r11 18			// STOP_TRIGGER
+		iowr I[$r8 + 0x000] $r11	// MMCTX_CTRL
+		mmctx_stop_wait:
+			// wait for STOP_TRIGGER to clear
+			iord $r11 I[$r8 + 0x000] // MMCTX_CTRL
+			xbit $r11 $r11 18
+			bra ne #mmctx_stop_wait
+	mmctx_done:
+	trace_clr(T_MMCTX)
+	ret
+
+// Wait for DONE_STRAND
+//
+strand_wait:
+	push $r10
+	mov $r10 2
+	call #wait_donez
+	pop $r10
+	ret
+
+// unknown - call before issuing strand commands
+//
+strand_pre:
+	mov $r8 0x4afc
+	sethi $r8 0x20000
+	mov $r9 0xc
+	iowr I[$r8] $r9
+	call #strand_wait
+	ret
+
+// unknown - call after issuing strand commands
+//
+strand_post:
+	mov $r8 0x4afc
+	sethi $r8 0x20000
+	mov $r9 0xd
+	iowr I[$r8] $r9
+	call #strand_wait
+	ret
+
+// Selects strand set?!
+//
+// In: $r14 id
+//
+strand_set:
+	mov $r10 0x4ffc
+	sethi $r10 0x20000
+	sub b32 $r11 $r10 0x500
+	mov $r12 0xf
+	iowr I[$r10 + 0x000] $r12		// 0x93c = 0xf
+	mov $r12 0xb
+	iowr I[$r11 + 0x000] $r12		// 0x928 = 0xb
+	call #strand_wait
+	iowr I[$r10 + 0x000] $r14		// 0x93c = <id>
+	mov $r12 0xa
+	iowr I[$r11 + 0x000] $r12		// 0x928 = 0xa
+	call #strand_wait
+	ret
+
+// Initialise strand context data
+//
+// In : $r15 context base
+// Out: $r15 context size (in bytes)
+//
+// Strandset(?) 3 hardcoded currently
+//
+strand_ctx_init:
+	trace_set(T_STRINIT)
+	call #strand_pre
+	mov $r14 3
+	call #strand_set
+	mov $r10 0x46fc
+	sethi $r10 0x20000
+	add b32 $r11 $r10 0x400
+	iowr I[$r10 + 0x100] $r0	// STRAND_FIRST_GENE = 0
+	mov $r12 1
+	iowr I[$r11 + 0x000] $r12	// STRAND_CMD = LATCH_FIRST_GENE
+	call #strand_wait
+	sub b32 $r12 $r0 1
+	iowr I[$r10 + 0x000] $r12	// STRAND_GENE_CNT = 0xffffffff
+	mov $r12 2
+	iowr I[$r11 + 0x000] $r12	// STRAND_CMD = LATCH_GENE_CNT
+	call #strand_wait
+	call #strand_post
+
+	// read the size of each strand, poke the context offset of
+	// each into STRAND_{SAVE,LOAD}_SWBASE now, no need to worry
+	// about it later then.
+	mov $r8 0x880
+	shl b32 $r8 6
+	iord $r9 I[$r8 + 0x000]		// STRANDS
+	add b32 $r8 0x2200
+	shr b32 $r14 $r15 8
+	ctx_init_strand_loop:
+		iowr I[$r8 + 0x000] $r14	// STRAND_SAVE_SWBASE
+		iowr I[$r8 + 0x100] $r14	// STRAND_LOAD_SWBASE
+		iord $r10 I[$r8 + 0x200]	// STRAND_SIZE
+		shr b32 $r10 6
+		add b32 $r10 1
+		add b32 $r14 $r10
+		add b32 $r8 4
+		sub b32 $r9 1
+		bra ne #ctx_init_strand_loop
+
+	shl b32 $r14 8
+	sub b32 $r15 $r14 $r15
+	trace_clr(T_STRINIT)
+	ret
+')
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv04.c b/drivers/gpu/drm/nouveau/core/engine/graph/nv04.c
new file mode 100644
index 0000000..6185282
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv04.c
@@ -0,0 +1,1387 @@
+/*
+ * Copyright 2007 Stephane Marchesin
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/handle.h>
+#include <core/namedb.h>
+
+#include <subdev/fb.h>
+#include <subdev/instmem.h>
+#include <subdev/timer.h>
+
+#include <engine/fifo.h>
+#include <engine/graph.h>
+
+#include "regs.h"
+
+static u32
+nv04_graph_ctx_regs[] = {
+	0x0040053c,
+	0x00400544,
+	0x00400540,
+	0x00400548,
+	NV04_PGRAPH_CTX_SWITCH1,
+	NV04_PGRAPH_CTX_SWITCH2,
+	NV04_PGRAPH_CTX_SWITCH3,
+	NV04_PGRAPH_CTX_SWITCH4,
+	NV04_PGRAPH_CTX_CACHE1,
+	NV04_PGRAPH_CTX_CACHE2,
+	NV04_PGRAPH_CTX_CACHE3,
+	NV04_PGRAPH_CTX_CACHE4,
+	0x00400184,
+	0x004001a4,
+	0x004001c4,
+	0x004001e4,
+	0x00400188,
+	0x004001a8,
+	0x004001c8,
+	0x004001e8,
+	0x0040018c,
+	0x004001ac,
+	0x004001cc,
+	0x004001ec,
+	0x00400190,
+	0x004001b0,
+	0x004001d0,
+	0x004001f0,
+	0x00400194,
+	0x004001b4,
+	0x004001d4,
+	0x004001f4,
+	0x00400198,
+	0x004001b8,
+	0x004001d8,
+	0x004001f8,
+	0x0040019c,
+	0x004001bc,
+	0x004001dc,
+	0x004001fc,
+	0x00400174,
+	NV04_PGRAPH_DMA_START_0,
+	NV04_PGRAPH_DMA_START_1,
+	NV04_PGRAPH_DMA_LENGTH,
+	NV04_PGRAPH_DMA_MISC,
+	NV04_PGRAPH_DMA_PITCH,
+	NV04_PGRAPH_BOFFSET0,
+	NV04_PGRAPH_BBASE0,
+	NV04_PGRAPH_BLIMIT0,
+	NV04_PGRAPH_BOFFSET1,
+	NV04_PGRAPH_BBASE1,
+	NV04_PGRAPH_BLIMIT1,
+	NV04_PGRAPH_BOFFSET2,
+	NV04_PGRAPH_BBASE2,
+	NV04_PGRAPH_BLIMIT2,
+	NV04_PGRAPH_BOFFSET3,
+	NV04_PGRAPH_BBASE3,
+	NV04_PGRAPH_BLIMIT3,
+	NV04_PGRAPH_BOFFSET4,
+	NV04_PGRAPH_BBASE4,
+	NV04_PGRAPH_BLIMIT4,
+	NV04_PGRAPH_BOFFSET5,
+	NV04_PGRAPH_BBASE5,
+	NV04_PGRAPH_BLIMIT5,
+	NV04_PGRAPH_BPITCH0,
+	NV04_PGRAPH_BPITCH1,
+	NV04_PGRAPH_BPITCH2,
+	NV04_PGRAPH_BPITCH3,
+	NV04_PGRAPH_BPITCH4,
+	NV04_PGRAPH_SURFACE,
+	NV04_PGRAPH_STATE,
+	NV04_PGRAPH_BSWIZZLE2,
+	NV04_PGRAPH_BSWIZZLE5,
+	NV04_PGRAPH_BPIXEL,
+	NV04_PGRAPH_NOTIFY,
+	NV04_PGRAPH_PATT_COLOR0,
+	NV04_PGRAPH_PATT_COLOR1,
+	NV04_PGRAPH_PATT_COLORRAM+0x00,
+	NV04_PGRAPH_PATT_COLORRAM+0x04,
+	NV04_PGRAPH_PATT_COLORRAM+0x08,
+	NV04_PGRAPH_PATT_COLORRAM+0x0c,
+	NV04_PGRAPH_PATT_COLORRAM+0x10,
+	NV04_PGRAPH_PATT_COLORRAM+0x14,
+	NV04_PGRAPH_PATT_COLORRAM+0x18,
+	NV04_PGRAPH_PATT_COLORRAM+0x1c,
+	NV04_PGRAPH_PATT_COLORRAM+0x20,
+	NV04_PGRAPH_PATT_COLORRAM+0x24,
+	NV04_PGRAPH_PATT_COLORRAM+0x28,
+	NV04_PGRAPH_PATT_COLORRAM+0x2c,
+	NV04_PGRAPH_PATT_COLORRAM+0x30,
+	NV04_PGRAPH_PATT_COLORRAM+0x34,
+	NV04_PGRAPH_PATT_COLORRAM+0x38,
+	NV04_PGRAPH_PATT_COLORRAM+0x3c,
+	NV04_PGRAPH_PATT_COLORRAM+0x40,
+	NV04_PGRAPH_PATT_COLORRAM+0x44,
+	NV04_PGRAPH_PATT_COLORRAM+0x48,
+	NV04_PGRAPH_PATT_COLORRAM+0x4c,
+	NV04_PGRAPH_PATT_COLORRAM+0x50,
+	NV04_PGRAPH_PATT_COLORRAM+0x54,
+	NV04_PGRAPH_PATT_COLORRAM+0x58,
+	NV04_PGRAPH_PATT_COLORRAM+0x5c,
+	NV04_PGRAPH_PATT_COLORRAM+0x60,
+	NV04_PGRAPH_PATT_COLORRAM+0x64,
+	NV04_PGRAPH_PATT_COLORRAM+0x68,
+	NV04_PGRAPH_PATT_COLORRAM+0x6c,
+	NV04_PGRAPH_PATT_COLORRAM+0x70,
+	NV04_PGRAPH_PATT_COLORRAM+0x74,
+	NV04_PGRAPH_PATT_COLORRAM+0x78,
+	NV04_PGRAPH_PATT_COLORRAM+0x7c,
+	NV04_PGRAPH_PATT_COLORRAM+0x80,
+	NV04_PGRAPH_PATT_COLORRAM+0x84,
+	NV04_PGRAPH_PATT_COLORRAM+0x88,
+	NV04_PGRAPH_PATT_COLORRAM+0x8c,
+	NV04_PGRAPH_PATT_COLORRAM+0x90,
+	NV04_PGRAPH_PATT_COLORRAM+0x94,
+	NV04_PGRAPH_PATT_COLORRAM+0x98,
+	NV04_PGRAPH_PATT_COLORRAM+0x9c,
+	NV04_PGRAPH_PATT_COLORRAM+0xa0,
+	NV04_PGRAPH_PATT_COLORRAM+0xa4,
+	NV04_PGRAPH_PATT_COLORRAM+0xa8,
+	NV04_PGRAPH_PATT_COLORRAM+0xac,
+	NV04_PGRAPH_PATT_COLORRAM+0xb0,
+	NV04_PGRAPH_PATT_COLORRAM+0xb4,
+	NV04_PGRAPH_PATT_COLORRAM+0xb8,
+	NV04_PGRAPH_PATT_COLORRAM+0xbc,
+	NV04_PGRAPH_PATT_COLORRAM+0xc0,
+	NV04_PGRAPH_PATT_COLORRAM+0xc4,
+	NV04_PGRAPH_PATT_COLORRAM+0xc8,
+	NV04_PGRAPH_PATT_COLORRAM+0xcc,
+	NV04_PGRAPH_PATT_COLORRAM+0xd0,
+	NV04_PGRAPH_PATT_COLORRAM+0xd4,
+	NV04_PGRAPH_PATT_COLORRAM+0xd8,
+	NV04_PGRAPH_PATT_COLORRAM+0xdc,
+	NV04_PGRAPH_PATT_COLORRAM+0xe0,
+	NV04_PGRAPH_PATT_COLORRAM+0xe4,
+	NV04_PGRAPH_PATT_COLORRAM+0xe8,
+	NV04_PGRAPH_PATT_COLORRAM+0xec,
+	NV04_PGRAPH_PATT_COLORRAM+0xf0,
+	NV04_PGRAPH_PATT_COLORRAM+0xf4,
+	NV04_PGRAPH_PATT_COLORRAM+0xf8,
+	NV04_PGRAPH_PATT_COLORRAM+0xfc,
+	NV04_PGRAPH_PATTERN,
+	0x0040080c,
+	NV04_PGRAPH_PATTERN_SHAPE,
+	0x00400600,
+	NV04_PGRAPH_ROP3,
+	NV04_PGRAPH_CHROMA,
+	NV04_PGRAPH_BETA_AND,
+	NV04_PGRAPH_BETA_PREMULT,
+	NV04_PGRAPH_CONTROL0,
+	NV04_PGRAPH_CONTROL1,
+	NV04_PGRAPH_CONTROL2,
+	NV04_PGRAPH_BLEND,
+	NV04_PGRAPH_STORED_FMT,
+	NV04_PGRAPH_SOURCE_COLOR,
+	0x00400560,
+	0x00400568,
+	0x00400564,
+	0x0040056c,
+	0x00400400,
+	0x00400480,
+	0x00400404,
+	0x00400484,
+	0x00400408,
+	0x00400488,
+	0x0040040c,
+	0x0040048c,
+	0x00400410,
+	0x00400490,
+	0x00400414,
+	0x00400494,
+	0x00400418,
+	0x00400498,
+	0x0040041c,
+	0x0040049c,
+	0x00400420,
+	0x004004a0,
+	0x00400424,
+	0x004004a4,
+	0x00400428,
+	0x004004a8,
+	0x0040042c,
+	0x004004ac,
+	0x00400430,
+	0x004004b0,
+	0x00400434,
+	0x004004b4,
+	0x00400438,
+	0x004004b8,
+	0x0040043c,
+	0x004004bc,
+	0x00400440,
+	0x004004c0,
+	0x00400444,
+	0x004004c4,
+	0x00400448,
+	0x004004c8,
+	0x0040044c,
+	0x004004cc,
+	0x00400450,
+	0x004004d0,
+	0x00400454,
+	0x004004d4,
+	0x00400458,
+	0x004004d8,
+	0x0040045c,
+	0x004004dc,
+	0x00400460,
+	0x004004e0,
+	0x00400464,
+	0x004004e4,
+	0x00400468,
+	0x004004e8,
+	0x0040046c,
+	0x004004ec,
+	0x00400470,
+	0x004004f0,
+	0x00400474,
+	0x004004f4,
+	0x00400478,
+	0x004004f8,
+	0x0040047c,
+	0x004004fc,
+	0x00400534,
+	0x00400538,
+	0x00400514,
+	0x00400518,
+	0x0040051c,
+	0x00400520,
+	0x00400524,
+	0x00400528,
+	0x0040052c,
+	0x00400530,
+	0x00400d00,
+	0x00400d40,
+	0x00400d80,
+	0x00400d04,
+	0x00400d44,
+	0x00400d84,
+	0x00400d08,
+	0x00400d48,
+	0x00400d88,
+	0x00400d0c,
+	0x00400d4c,
+	0x00400d8c,
+	0x00400d10,
+	0x00400d50,
+	0x00400d90,
+	0x00400d14,
+	0x00400d54,
+	0x00400d94,
+	0x00400d18,
+	0x00400d58,
+	0x00400d98,
+	0x00400d1c,
+	0x00400d5c,
+	0x00400d9c,
+	0x00400d20,
+	0x00400d60,
+	0x00400da0,
+	0x00400d24,
+	0x00400d64,
+	0x00400da4,
+	0x00400d28,
+	0x00400d68,
+	0x00400da8,
+	0x00400d2c,
+	0x00400d6c,
+	0x00400dac,
+	0x00400d30,
+	0x00400d70,
+	0x00400db0,
+	0x00400d34,
+	0x00400d74,
+	0x00400db4,
+	0x00400d38,
+	0x00400d78,
+	0x00400db8,
+	0x00400d3c,
+	0x00400d7c,
+	0x00400dbc,
+	0x00400590,
+	0x00400594,
+	0x00400598,
+	0x0040059c,
+	0x004005a8,
+	0x004005ac,
+	0x004005b0,
+	0x004005b4,
+	0x004005c0,
+	0x004005c4,
+	0x004005c8,
+	0x004005cc,
+	0x004005d0,
+	0x004005d4,
+	0x004005d8,
+	0x004005dc,
+	0x004005e0,
+	NV04_PGRAPH_PASSTHRU_0,
+	NV04_PGRAPH_PASSTHRU_1,
+	NV04_PGRAPH_PASSTHRU_2,
+	NV04_PGRAPH_DVD_COLORFMT,
+	NV04_PGRAPH_SCALED_FORMAT,
+	NV04_PGRAPH_MISC24_0,
+	NV04_PGRAPH_MISC24_1,
+	NV04_PGRAPH_MISC24_2,
+	0x00400500,
+	0x00400504,
+	NV04_PGRAPH_VALID1,
+	NV04_PGRAPH_VALID2,
+	NV04_PGRAPH_DEBUG_3
+};
+
+struct nv04_graph_priv {
+	struct nouveau_graph base;
+	struct nv04_graph_chan *chan[16];
+	spinlock_t lock;
+};
+
+struct nv04_graph_chan {
+	struct nouveau_object base;
+	int chid;
+	u32 nv04[ARRAY_SIZE(nv04_graph_ctx_regs)];
+};
+
+
+static inline struct nv04_graph_priv *
+nv04_graph_priv(struct nv04_graph_chan *chan)
+{
+	return (void *)nv_object(chan)->engine;
+}
+
+/*******************************************************************************
+ * Graphics object classes
+ ******************************************************************************/
+
+/*
+ * Software methods, why they are needed, and how they all work:
+ *
+ * NV04 and NV05 keep most of the state in PGRAPH context itself, but some
+ * 2d engine settings are kept inside the grobjs themselves. The grobjs are
+ * 3 words long on both. grobj format on NV04 is:
+ *
+ * word 0:
+ *  - bits 0-7: class
+ *  - bit 12: color key active
+ *  - bit 13: clip rect active
+ *  - bit 14: if set, destination surface is swizzled and taken from buffer 5
+ *            [set by NV04_SWIZZLED_SURFACE], otherwise it's linear and taken
+ *            from buffer 0 [set by NV04_CONTEXT_SURFACES_2D or
+ *            NV03_CONTEXT_SURFACE_DST].
+ *  - bits 15-17: 2d operation [aka patch config]
+ *  - bit 24: patch valid [enables rendering using this object]
+ *  - bit 25: surf3d valid [for tex_tri and multitex_tri only]
+ * word 1:
+ *  - bits 0-1: mono format
+ *  - bits 8-13: color format
+ *  - bits 16-31: DMA_NOTIFY instance
+ * word 2:
+ *  - bits 0-15: DMA_A instance
+ *  - bits 16-31: DMA_B instance
+ *
+ * On NV05 it's:
+ *
+ * word 0:
+ *  - bits 0-7: class
+ *  - bit 12: color key active
+ *  - bit 13: clip rect active
+ *  - bit 14: if set, destination surface is swizzled and taken from buffer 5
+ *            [set by NV04_SWIZZLED_SURFACE], otherwise it's linear and taken
+ *            from buffer 0 [set by NV04_CONTEXT_SURFACES_2D or
+ *            NV03_CONTEXT_SURFACE_DST].
+ *  - bits 15-17: 2d operation [aka patch config]
+ *  - bits 20-22: dither mode
+ *  - bit 24: patch valid [enables rendering using this object]
+ *  - bit 25: surface_dst/surface_color/surf2d/surf3d valid
+ *  - bit 26: surface_src/surface_zeta valid
+ *  - bit 27: pattern valid
+ *  - bit 28: rop valid
+ *  - bit 29: beta1 valid
+ *  - bit 30: beta4 valid
+ * word 1:
+ *  - bits 0-1: mono format
+ *  - bits 8-13: color format
+ *  - bits 16-31: DMA_NOTIFY instance
+ * word 2:
+ *  - bits 0-15: DMA_A instance
+ *  - bits 16-31: DMA_B instance
+ *
+ * NV05 will set/unset the relevant valid bits when you poke the relevant
+ * object-binding methods with object of the proper type, or with the NULL
+ * type. It'll only allow rendering using the grobj if all needed objects
+ * are bound. The needed set of objects depends on selected operation: for
+ * example rop object is needed by ROP_AND, but not by SRCCOPY_AND.
+ *
+ * NV04 doesn't have these methods implemented at all, and doesn't have the
+ * relevant bits in grobj. Instead, it'll allow rendering whenever bit 24
+ * is set. So we have to emulate them in software, internally keeping the
+ * same bits as NV05 does. Since grobjs are aligned to 16 bytes on nv04,
+ * but the last word isn't actually used for anything, we abuse it for this
+ * purpose.
+ *
+ * Actually, NV05 can optionally check bit 24 too, but we disable this since
+ * there's no use for it.
+ *
+ * For unknown reasons, NV04 implements surf3d binding in hardware as an
+ * exception. Also for unknown reasons, NV04 doesn't implement the clipping
+ * methods on the surf3d object, so we have to emulate them too.
+ */
+
+static void
+nv04_graph_set_ctx1(struct nouveau_object *object, u32 mask, u32 value)
+{
+	struct nv04_graph_priv *priv = (void *)object->engine;
+	int subc = (nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR) >> 13) & 0x7;
+	u32 tmp;
+
+	tmp  = nv_ro32(object, 0x00);
+	tmp &= ~mask;
+	tmp |= value;
+	nv_wo32(object, 0x00, tmp);
+
+	nv_wr32(priv, NV04_PGRAPH_CTX_SWITCH1, tmp);
+	nv_wr32(priv, NV04_PGRAPH_CTX_CACHE1 + (subc<<2), tmp);
+}
+
+static void
+nv04_graph_set_ctx_val(struct nouveau_object *object, u32 mask, u32 value)
+{
+	int class, op, valid = 1;
+	u32 tmp, ctx1;
+
+	ctx1 = nv_ro32(object, 0x00);
+	class = ctx1 & 0xff;
+	op = (ctx1 >> 15) & 7;
+
+	tmp = nv_ro32(object, 0x0c);
+	tmp &= ~mask;
+	tmp |= value;
+	nv_wo32(object, 0x0c, tmp);
+
+	/* check for valid surf2d/surf_dst/surf_color */
+	if (!(tmp & 0x02000000))
+		valid = 0;
+	/* check for valid surf_src/surf_zeta */
+	if ((class == 0x1f || class == 0x48) && !(tmp & 0x04000000))
+		valid = 0;
+
+	switch (op) {
+	/* SRCCOPY_AND, SRCCOPY: no extra objects required */
+	case 0:
+	case 3:
+		break;
+	/* ROP_AND: requires pattern and rop */
+	case 1:
+		if (!(tmp & 0x18000000))
+			valid = 0;
+		break;
+	/* BLEND_AND: requires beta1 */
+	case 2:
+		if (!(tmp & 0x20000000))
+			valid = 0;
+		break;
+	/* SRCCOPY_PREMULT, BLEND_PREMULT: beta4 required */
+	case 4:
+	case 5:
+		if (!(tmp & 0x40000000))
+			valid = 0;
+		break;
+	}
+
+	nv04_graph_set_ctx1(object, 0x01000000, valid << 24);
+}
+
+static int
+nv04_graph_mthd_set_operation(struct nouveau_object *object, u32 mthd,
+			      void *args, u32 size)
+{
+	u32 class = nv_ro32(object, 0) & 0xff;
+	u32 data = *(u32 *)args;
+	if (data > 5)
+		return 1;
+	/* Old versions of the objects only accept first three operations. */
+	if (data > 2 && class < 0x40)
+		return 1;
+	nv04_graph_set_ctx1(object, 0x00038000, data << 15);
+	/* changing operation changes set of objects needed for validation */
+	nv04_graph_set_ctx_val(object, 0, 0);
+	return 0;
+}
+
+static int
+nv04_graph_mthd_surf3d_clip_h(struct nouveau_object *object, u32 mthd,
+			      void *args, u32 size)
+{
+	struct nv04_graph_priv *priv = (void *)object->engine;
+	u32 data = *(u32 *)args;
+	u32 min = data & 0xffff, max;
+	u32 w = data >> 16;
+	if (min & 0x8000)
+		/* too large */
+		return 1;
+	if (w & 0x8000)
+		/* yes, it accepts negative for some reason. */
+		w |= 0xffff0000;
+	max = min + w;
+	max &= 0x3ffff;
+	nv_wr32(priv, 0x40053c, min);
+	nv_wr32(priv, 0x400544, max);
+	return 0;
+}
+
+static int
+nv04_graph_mthd_surf3d_clip_v(struct nouveau_object *object, u32 mthd,
+			      void *args, u32 size)
+{
+	struct nv04_graph_priv *priv = (void *)object->engine;
+	u32 data = *(u32 *)args;
+	u32 min = data & 0xffff, max;
+	u32 w = data >> 16;
+	if (min & 0x8000)
+		/* too large */
+		return 1;
+	if (w & 0x8000)
+		/* yes, it accepts negative for some reason. */
+		w |= 0xffff0000;
+	max = min + w;
+	max &= 0x3ffff;
+	nv_wr32(priv, 0x400540, min);
+	nv_wr32(priv, 0x400548, max);
+	return 0;
+}
+
+static u16
+nv04_graph_mthd_bind_class(struct nouveau_object *object, u32 *args, u32 size)
+{
+	struct nouveau_instmem *imem = nouveau_instmem(object);
+	u32 inst = *(u32 *)args << 4;
+	return nv_ro32(imem, inst);
+}
+
+static int
+nv04_graph_mthd_bind_surf2d(struct nouveau_object *object, u32 mthd,
+			    void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx1(object, 0x00004000, 0);
+		nv04_graph_set_ctx_val(object, 0x02000000, 0);
+		return 0;
+	case 0x42:
+		nv04_graph_set_ctx1(object, 0x00004000, 0);
+		nv04_graph_set_ctx_val(object, 0x02000000, 0x02000000);
+		return 0;
+	}
+	return 1;
+}
+
+static int
+nv04_graph_mthd_bind_surf2d_swzsurf(struct nouveau_object *object, u32 mthd,
+				    void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx1(object, 0x00004000, 0);
+		nv04_graph_set_ctx_val(object, 0x02000000, 0);
+		return 0;
+	case 0x42:
+		nv04_graph_set_ctx1(object, 0x00004000, 0);
+		nv04_graph_set_ctx_val(object, 0x02000000, 0x02000000);
+		return 0;
+	case 0x52:
+		nv04_graph_set_ctx1(object, 0x00004000, 0x00004000);
+		nv04_graph_set_ctx_val(object, 0x02000000, 0x02000000);
+		return 0;
+	}
+	return 1;
+}
+
+static int
+nv01_graph_mthd_bind_patt(struct nouveau_object *object, u32 mthd,
+			  void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx_val(object, 0x08000000, 0);
+		return 0;
+	case 0x18:
+		nv04_graph_set_ctx_val(object, 0x08000000, 0x08000000);
+		return 0;
+	}
+	return 1;
+}
+
+static int
+nv04_graph_mthd_bind_patt(struct nouveau_object *object, u32 mthd,
+			  void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx_val(object, 0x08000000, 0);
+		return 0;
+	case 0x44:
+		nv04_graph_set_ctx_val(object, 0x08000000, 0x08000000);
+		return 0;
+	}
+	return 1;
+}
+
+static int
+nv04_graph_mthd_bind_rop(struct nouveau_object *object, u32 mthd,
+			 void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx_val(object, 0x10000000, 0);
+		return 0;
+	case 0x43:
+		nv04_graph_set_ctx_val(object, 0x10000000, 0x10000000);
+		return 0;
+	}
+	return 1;
+}
+
+static int
+nv04_graph_mthd_bind_beta1(struct nouveau_object *object, u32 mthd,
+			   void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx_val(object, 0x20000000, 0);
+		return 0;
+	case 0x12:
+		nv04_graph_set_ctx_val(object, 0x20000000, 0x20000000);
+		return 0;
+	}
+	return 1;
+}
+
+static int
+nv04_graph_mthd_bind_beta4(struct nouveau_object *object, u32 mthd,
+			   void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx_val(object, 0x40000000, 0);
+		return 0;
+	case 0x72:
+		nv04_graph_set_ctx_val(object, 0x40000000, 0x40000000);
+		return 0;
+	}
+	return 1;
+}
+
+static int
+nv04_graph_mthd_bind_surf_dst(struct nouveau_object *object, u32 mthd,
+			      void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx_val(object, 0x02000000, 0);
+		return 0;
+	case 0x58:
+		nv04_graph_set_ctx_val(object, 0x02000000, 0x02000000);
+		return 0;
+	}
+	return 1;
+}
+
+static int
+nv04_graph_mthd_bind_surf_src(struct nouveau_object *object, u32 mthd,
+			      void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx_val(object, 0x04000000, 0);
+		return 0;
+	case 0x59:
+		nv04_graph_set_ctx_val(object, 0x04000000, 0x04000000);
+		return 0;
+	}
+	return 1;
+}
+
+static int
+nv04_graph_mthd_bind_surf_color(struct nouveau_object *object, u32 mthd,
+				void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx_val(object, 0x02000000, 0);
+		return 0;
+	case 0x5a:
+		nv04_graph_set_ctx_val(object, 0x02000000, 0x02000000);
+		return 0;
+	}
+	return 1;
+}
+
+static int
+nv04_graph_mthd_bind_surf_zeta(struct nouveau_object *object, u32 mthd,
+			       void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx_val(object, 0x04000000, 0);
+		return 0;
+	case 0x5b:
+		nv04_graph_set_ctx_val(object, 0x04000000, 0x04000000);
+		return 0;
+	}
+	return 1;
+}
+
+static int
+nv01_graph_mthd_bind_clip(struct nouveau_object *object, u32 mthd,
+			  void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx1(object, 0x2000, 0);
+		return 0;
+	case 0x19:
+		nv04_graph_set_ctx1(object, 0x2000, 0x2000);
+		return 0;
+	}
+	return 1;
+}
+
+static int
+nv01_graph_mthd_bind_chroma(struct nouveau_object *object, u32 mthd,
+			    void *args, u32 size)
+{
+	switch (nv04_graph_mthd_bind_class(object, args, size)) {
+	case 0x30:
+		nv04_graph_set_ctx1(object, 0x1000, 0);
+		return 0;
+	/* Yes, for some reason even the old versions of objects
+	 * accept 0x57 and not 0x17. Consistency be damned.
+	 */
+	case 0x57:
+		nv04_graph_set_ctx1(object, 0x1000, 0x1000);
+		return 0;
+	}
+	return 1;
+}
+
+static struct nouveau_omthds
+nv03_graph_gdi_omthds[] = {
+	{ 0x0184, nv01_graph_mthd_bind_patt },
+	{ 0x0188, nv04_graph_mthd_bind_rop },
+	{ 0x018c, nv04_graph_mthd_bind_beta1 },
+	{ 0x0190, nv04_graph_mthd_bind_surf_dst },
+	{ 0x02fc, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static struct nouveau_omthds
+nv04_graph_gdi_omthds[] = {
+	{ 0x0188, nv04_graph_mthd_bind_patt },
+	{ 0x018c, nv04_graph_mthd_bind_rop },
+	{ 0x0190, nv04_graph_mthd_bind_beta1 },
+	{ 0x0194, nv04_graph_mthd_bind_beta4 },
+	{ 0x0198, nv04_graph_mthd_bind_surf2d },
+	{ 0x02fc, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static struct nouveau_omthds
+nv01_graph_blit_omthds[] = {
+	{ 0x0184, nv01_graph_mthd_bind_chroma },
+	{ 0x0188, nv01_graph_mthd_bind_clip },
+	{ 0x018c, nv01_graph_mthd_bind_patt },
+	{ 0x0190, nv04_graph_mthd_bind_rop },
+	{ 0x0194, nv04_graph_mthd_bind_beta1 },
+	{ 0x0198, nv04_graph_mthd_bind_surf_dst },
+	{ 0x019c, nv04_graph_mthd_bind_surf_src },
+	{ 0x02fc, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static struct nouveau_omthds
+nv04_graph_blit_omthds[] = {
+	{ 0x0184, nv01_graph_mthd_bind_chroma },
+	{ 0x0188, nv01_graph_mthd_bind_clip },
+	{ 0x018c, nv04_graph_mthd_bind_patt },
+	{ 0x0190, nv04_graph_mthd_bind_rop },
+	{ 0x0194, nv04_graph_mthd_bind_beta1 },
+	{ 0x0198, nv04_graph_mthd_bind_beta4 },
+	{ 0x019c, nv04_graph_mthd_bind_surf2d },
+	{ 0x02fc, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static struct nouveau_omthds
+nv04_graph_iifc_omthds[] = {
+	{ 0x0188, nv01_graph_mthd_bind_chroma },
+	{ 0x018c, nv01_graph_mthd_bind_clip },
+	{ 0x0190, nv04_graph_mthd_bind_patt },
+	{ 0x0194, nv04_graph_mthd_bind_rop },
+	{ 0x0198, nv04_graph_mthd_bind_beta1 },
+	{ 0x019c, nv04_graph_mthd_bind_beta4 },
+	{ 0x01a0, nv04_graph_mthd_bind_surf2d_swzsurf },
+	{ 0x03e4, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static struct nouveau_omthds
+nv01_graph_ifc_omthds[] = {
+	{ 0x0184, nv01_graph_mthd_bind_chroma },
+	{ 0x0188, nv01_graph_mthd_bind_clip },
+	{ 0x018c, nv01_graph_mthd_bind_patt },
+	{ 0x0190, nv04_graph_mthd_bind_rop },
+	{ 0x0194, nv04_graph_mthd_bind_beta1 },
+	{ 0x0198, nv04_graph_mthd_bind_surf_dst },
+	{ 0x02fc, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static struct nouveau_omthds
+nv04_graph_ifc_omthds[] = {
+	{ 0x0184, nv01_graph_mthd_bind_chroma },
+	{ 0x0188, nv01_graph_mthd_bind_clip },
+	{ 0x018c, nv04_graph_mthd_bind_patt },
+	{ 0x0190, nv04_graph_mthd_bind_rop },
+	{ 0x0194, nv04_graph_mthd_bind_beta1 },
+	{ 0x0198, nv04_graph_mthd_bind_beta4 },
+	{ 0x019c, nv04_graph_mthd_bind_surf2d },
+	{ 0x02fc, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static struct nouveau_omthds
+nv03_graph_sifc_omthds[] = {
+	{ 0x0184, nv01_graph_mthd_bind_chroma },
+	{ 0x0188, nv01_graph_mthd_bind_patt },
+	{ 0x018c, nv04_graph_mthd_bind_rop },
+	{ 0x0190, nv04_graph_mthd_bind_beta1 },
+	{ 0x0194, nv04_graph_mthd_bind_surf_dst },
+	{ 0x02fc, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static struct nouveau_omthds
+nv04_graph_sifc_omthds[] = {
+	{ 0x0184, nv01_graph_mthd_bind_chroma },
+	{ 0x0188, nv04_graph_mthd_bind_patt },
+	{ 0x018c, nv04_graph_mthd_bind_rop },
+	{ 0x0190, nv04_graph_mthd_bind_beta1 },
+	{ 0x0194, nv04_graph_mthd_bind_beta4 },
+	{ 0x0198, nv04_graph_mthd_bind_surf2d },
+	{ 0x02fc, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static struct nouveau_omthds
+nv03_graph_sifm_omthds[] = {
+	{ 0x0188, nv01_graph_mthd_bind_patt },
+	{ 0x018c, nv04_graph_mthd_bind_rop },
+	{ 0x0190, nv04_graph_mthd_bind_beta1 },
+	{ 0x0194, nv04_graph_mthd_bind_surf_dst },
+	{ 0x0304, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static struct nouveau_omthds
+nv04_graph_sifm_omthds[] = {
+	{ 0x0188, nv04_graph_mthd_bind_patt },
+	{ 0x018c, nv04_graph_mthd_bind_rop },
+	{ 0x0190, nv04_graph_mthd_bind_beta1 },
+	{ 0x0194, nv04_graph_mthd_bind_beta4 },
+	{ 0x0198, nv04_graph_mthd_bind_surf2d },
+	{ 0x0304, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static struct nouveau_omthds
+nv04_graph_surf3d_omthds[] = {
+	{ 0x02f8, nv04_graph_mthd_surf3d_clip_h },
+	{ 0x02fc, nv04_graph_mthd_surf3d_clip_v },
+	{}
+};
+
+static struct nouveau_omthds
+nv03_graph_ttri_omthds[] = {
+	{ 0x0188, nv01_graph_mthd_bind_clip },
+	{ 0x018c, nv04_graph_mthd_bind_surf_color },
+	{ 0x0190, nv04_graph_mthd_bind_surf_zeta },
+	{}
+};
+
+static struct nouveau_omthds
+nv01_graph_prim_omthds[] = {
+	{ 0x0184, nv01_graph_mthd_bind_clip },
+	{ 0x0188, nv01_graph_mthd_bind_patt },
+	{ 0x018c, nv04_graph_mthd_bind_rop },
+	{ 0x0190, nv04_graph_mthd_bind_beta1 },
+	{ 0x0194, nv04_graph_mthd_bind_surf_dst },
+	{ 0x02fc, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static struct nouveau_omthds
+nv04_graph_prim_omthds[] = {
+	{ 0x0184, nv01_graph_mthd_bind_clip },
+	{ 0x0188, nv04_graph_mthd_bind_patt },
+	{ 0x018c, nv04_graph_mthd_bind_rop },
+	{ 0x0190, nv04_graph_mthd_bind_beta1 },
+	{ 0x0194, nv04_graph_mthd_bind_beta4 },
+	{ 0x0198, nv04_graph_mthd_bind_surf2d },
+	{ 0x02fc, nv04_graph_mthd_set_operation },
+	{}
+};
+
+static int
+nv04_graph_object_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nouveau_gpuobj *obj;
+	int ret;
+
+	ret = nouveau_gpuobj_create(parent, engine, oclass, 0, parent,
+				    16, 16, 0, &obj);
+	*pobject = nv_object(obj);
+	if (ret)
+		return ret;
+
+	nv_wo32(obj, 0x00, nv_mclass(obj));
+#ifdef __BIG_ENDIAN
+	nv_mo32(obj, 0x00, 0x00080000, 0x00080000);
+#endif
+	nv_wo32(obj, 0x04, 0x00000000);
+	nv_wo32(obj, 0x08, 0x00000000);
+	nv_wo32(obj, 0x0c, 0x00000000);
+	return 0;
+}
+
+struct nouveau_ofuncs
+nv04_graph_ofuncs = {
+	.ctor = nv04_graph_object_ctor,
+	.dtor = _nouveau_gpuobj_dtor,
+	.init = _nouveau_gpuobj_init,
+	.fini = _nouveau_gpuobj_fini,
+	.rd32 = _nouveau_gpuobj_rd32,
+	.wr32 = _nouveau_gpuobj_wr32,
+};
+
+static struct nouveau_oclass
+nv04_graph_sclass[] = {
+	{ 0x0012, &nv04_graph_ofuncs }, /* beta1 */
+	{ 0x0017, &nv04_graph_ofuncs }, /* chroma */
+	{ 0x0018, &nv04_graph_ofuncs }, /* pattern (nv01) */
+	{ 0x0019, &nv04_graph_ofuncs }, /* clip */
+	{ 0x001c, &nv04_graph_ofuncs, nv01_graph_prim_omthds }, /* line */
+	{ 0x001d, &nv04_graph_ofuncs, nv01_graph_prim_omthds }, /* tri */
+	{ 0x001e, &nv04_graph_ofuncs, nv01_graph_prim_omthds }, /* rect */
+	{ 0x001f, &nv04_graph_ofuncs, nv01_graph_blit_omthds },
+	{ 0x0021, &nv04_graph_ofuncs, nv01_graph_ifc_omthds },
+	{ 0x0030, &nv04_graph_ofuncs }, /* null */
+	{ 0x0036, &nv04_graph_ofuncs, nv03_graph_sifc_omthds },
+	{ 0x0037, &nv04_graph_ofuncs, nv03_graph_sifm_omthds },
+	{ 0x0038, &nv04_graph_ofuncs }, /* dvd subpicture */
+	{ 0x0039, &nv04_graph_ofuncs }, /* m2mf */
+	{ 0x0042, &nv04_graph_ofuncs }, /* surf2d */
+	{ 0x0043, &nv04_graph_ofuncs }, /* rop */
+	{ 0x0044, &nv04_graph_ofuncs }, /* pattern */
+	{ 0x0048, &nv04_graph_ofuncs, nv03_graph_ttri_omthds },
+	{ 0x004a, &nv04_graph_ofuncs, nv04_graph_gdi_omthds },
+	{ 0x004b, &nv04_graph_ofuncs, nv03_graph_gdi_omthds },
+	{ 0x0052, &nv04_graph_ofuncs }, /* swzsurf */
+	{ 0x0053, &nv04_graph_ofuncs, nv04_graph_surf3d_omthds },
+	{ 0x0054, &nv04_graph_ofuncs }, /* ttri */
+	{ 0x0055, &nv04_graph_ofuncs }, /* mtri */
+	{ 0x0057, &nv04_graph_ofuncs }, /* chroma */
+	{ 0x0058, &nv04_graph_ofuncs }, /* surf_dst */
+	{ 0x0059, &nv04_graph_ofuncs }, /* surf_src */
+	{ 0x005a, &nv04_graph_ofuncs }, /* surf_color */
+	{ 0x005b, &nv04_graph_ofuncs }, /* surf_zeta */
+	{ 0x005c, &nv04_graph_ofuncs, nv04_graph_prim_omthds }, /* line */
+	{ 0x005d, &nv04_graph_ofuncs, nv04_graph_prim_omthds }, /* tri */
+	{ 0x005e, &nv04_graph_ofuncs, nv04_graph_prim_omthds }, /* rect */
+	{ 0x005f, &nv04_graph_ofuncs, nv04_graph_blit_omthds },
+	{ 0x0060, &nv04_graph_ofuncs, nv04_graph_iifc_omthds },
+	{ 0x0061, &nv04_graph_ofuncs, nv04_graph_ifc_omthds },
+	{ 0x0064, &nv04_graph_ofuncs }, /* iifc (nv05) */
+	{ 0x0065, &nv04_graph_ofuncs }, /* ifc (nv05) */
+	{ 0x0066, &nv04_graph_ofuncs }, /* sifc (nv05) */
+	{ 0x0072, &nv04_graph_ofuncs }, /* beta4 */
+	{ 0x0076, &nv04_graph_ofuncs, nv04_graph_sifc_omthds },
+	{ 0x0077, &nv04_graph_ofuncs, nv04_graph_sifm_omthds },
+	{},
+};
+
+/*******************************************************************************
+ * PGRAPH context
+ ******************************************************************************/
+
+static struct nv04_graph_chan *
+nv04_graph_channel(struct nv04_graph_priv *priv)
+{
+	struct nv04_graph_chan *chan = NULL;
+	if (nv_rd32(priv, NV04_PGRAPH_CTX_CONTROL) & 0x00010000) {
+		int chid = nv_rd32(priv, NV04_PGRAPH_CTX_USER) >> 24;
+		if (chid < ARRAY_SIZE(priv->chan))
+			chan = priv->chan[chid];
+	}
+	return chan;
+}
+
+static int
+nv04_graph_load_context(struct nv04_graph_chan *chan, int chid)
+{
+	struct nv04_graph_priv *priv = nv04_graph_priv(chan);
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(nv04_graph_ctx_regs); i++)
+		nv_wr32(priv, nv04_graph_ctx_regs[i], chan->nv04[i]);
+
+	nv_wr32(priv, NV04_PGRAPH_CTX_CONTROL, 0x10010100);
+	nv_mask(priv, NV04_PGRAPH_CTX_USER, 0xff000000, chid << 24);
+	nv_mask(priv, NV04_PGRAPH_FFINTFC_ST2, 0xfff00000, 0x00000000);
+	return 0;
+}
+
+static int
+nv04_graph_unload_context(struct nv04_graph_chan *chan)
+{
+	struct nv04_graph_priv *priv = nv04_graph_priv(chan);
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(nv04_graph_ctx_regs); i++)
+		chan->nv04[i] = nv_rd32(priv, nv04_graph_ctx_regs[i]);
+
+	nv_wr32(priv, NV04_PGRAPH_CTX_CONTROL, 0x10000000);
+	nv_mask(priv, NV04_PGRAPH_CTX_USER, 0xff000000, 0x0f000000);
+	return 0;
+}
+
+static void
+nv04_graph_context_switch(struct nv04_graph_priv *priv)
+{
+	struct nv04_graph_chan *prev = NULL;
+	struct nv04_graph_chan *next = NULL;
+	unsigned long flags;
+	int chid;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	nv04_graph_idle(priv);
+
+	/* If previous context is valid, we need to save it */
+	prev = nv04_graph_channel(priv);
+	if (prev)
+		nv04_graph_unload_context(prev);
+
+	/* load context for next channel */
+	chid = (nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR) >> 24) & 0x0f;
+	next = priv->chan[chid];
+	if (next)
+		nv04_graph_load_context(next, chid);
+
+	spin_unlock_irqrestore(&priv->lock, flags);
+}
+
+static u32 *ctx_reg(struct nv04_graph_chan *chan, u32 reg)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(nv04_graph_ctx_regs); i++) {
+		if (nv04_graph_ctx_regs[i] == reg)
+			return &chan->nv04[i];
+	}
+
+	return NULL;
+}
+
+static int
+nv04_graph_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nouveau_fifo_chan *fifo = (void *)parent;
+	struct nv04_graph_priv *priv = (void *)engine;
+	struct nv04_graph_chan *chan;
+	unsigned long flags;
+	int ret;
+
+	ret = nouveau_object_create(parent, engine, oclass, 0, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	if (priv->chan[fifo->chid]) {
+		*pobject = nv_object(priv->chan[fifo->chid]);
+		atomic_inc(&(*pobject)->refcount);
+		spin_unlock_irqrestore(&priv->lock, flags);
+		nouveau_object_destroy(&chan->base);
+		return 1;
+	}
+
+	*ctx_reg(chan, NV04_PGRAPH_DEBUG_3) = 0xfad4ff31;
+
+	priv->chan[fifo->chid] = chan;
+	chan->chid = fifo->chid;
+	spin_unlock_irqrestore(&priv->lock, flags);
+	return 0;
+}
+
+static void
+nv04_graph_context_dtor(struct nouveau_object *object)
+{
+	struct nv04_graph_priv *priv = (void *)object->engine;
+	struct nv04_graph_chan *chan = (void *)object;
+	unsigned long flags;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	priv->chan[chan->chid] = NULL;
+	spin_unlock_irqrestore(&priv->lock, flags);
+
+	nouveau_object_destroy(&chan->base);
+}
+
+static int
+nv04_graph_context_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv04_graph_priv *priv = (void *)object->engine;
+	struct nv04_graph_chan *chan = (void *)object;
+	unsigned long flags;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000);
+	if (nv04_graph_channel(priv) == chan)
+		nv04_graph_unload_context(chan);
+	nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001);
+	spin_unlock_irqrestore(&priv->lock, flags);
+
+	return nouveau_object_fini(&chan->base, suspend);
+}
+
+static struct nouveau_oclass
+nv04_graph_cclass = {
+	.handle = NV_ENGCTX(GR, 0x04),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv04_graph_context_ctor,
+		.dtor = nv04_graph_context_dtor,
+		.init = nouveau_object_init,
+		.fini = nv04_graph_context_fini,
+	},
+};
+
+/*******************************************************************************
+ * PGRAPH engine/subdev functions
+ ******************************************************************************/
+
+bool
+nv04_graph_idle(void *obj)
+{
+	struct nouveau_graph *graph = nouveau_graph(obj);
+	u32 mask = 0xffffffff;
+
+	if (nv_device(obj)->card_type == NV_40)
+		mask &= ~NV40_PGRAPH_STATUS_SYNC_STALL;
+
+	if (!nv_wait(graph, NV04_PGRAPH_STATUS, mask, 0)) {
+		nv_error(graph, "idle timed out with status 0x%08x\n",
+			 nv_rd32(graph, NV04_PGRAPH_STATUS));
+		return false;
+	}
+
+	return true;
+}
+
+static const struct nouveau_bitfield
+nv04_graph_intr_name[] = {
+	{ NV_PGRAPH_INTR_NOTIFY, "NOTIFY" },
+	{}
+};
+
+static const struct nouveau_bitfield
+nv04_graph_nstatus[] = {
+	{ NV04_PGRAPH_NSTATUS_STATE_IN_USE,       "STATE_IN_USE" },
+	{ NV04_PGRAPH_NSTATUS_INVALID_STATE,      "INVALID_STATE" },
+	{ NV04_PGRAPH_NSTATUS_BAD_ARGUMENT,       "BAD_ARGUMENT" },
+	{ NV04_PGRAPH_NSTATUS_PROTECTION_FAULT,   "PROTECTION_FAULT" },
+	{}
+};
+
+const struct nouveau_bitfield
+nv04_graph_nsource[] = {
+	{ NV03_PGRAPH_NSOURCE_NOTIFICATION,       "NOTIFICATION" },
+	{ NV03_PGRAPH_NSOURCE_DATA_ERROR,         "DATA_ERROR" },
+	{ NV03_PGRAPH_NSOURCE_PROTECTION_ERROR,   "PROTECTION_ERROR" },
+	{ NV03_PGRAPH_NSOURCE_RANGE_EXCEPTION,    "RANGE_EXCEPTION" },
+	{ NV03_PGRAPH_NSOURCE_LIMIT_COLOR,        "LIMIT_COLOR" },
+	{ NV03_PGRAPH_NSOURCE_LIMIT_ZETA,         "LIMIT_ZETA" },
+	{ NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD,       "ILLEGAL_MTHD" },
+	{ NV03_PGRAPH_NSOURCE_DMA_R_PROTECTION,   "DMA_R_PROTECTION" },
+	{ NV03_PGRAPH_NSOURCE_DMA_W_PROTECTION,   "DMA_W_PROTECTION" },
+	{ NV03_PGRAPH_NSOURCE_FORMAT_EXCEPTION,   "FORMAT_EXCEPTION" },
+	{ NV03_PGRAPH_NSOURCE_PATCH_EXCEPTION,    "PATCH_EXCEPTION" },
+	{ NV03_PGRAPH_NSOURCE_STATE_INVALID,      "STATE_INVALID" },
+	{ NV03_PGRAPH_NSOURCE_DOUBLE_NOTIFY,      "DOUBLE_NOTIFY" },
+	{ NV03_PGRAPH_NSOURCE_NOTIFY_IN_USE,      "NOTIFY_IN_USE" },
+	{ NV03_PGRAPH_NSOURCE_METHOD_CNT,         "METHOD_CNT" },
+	{ NV03_PGRAPH_NSOURCE_BFR_NOTIFICATION,   "BFR_NOTIFICATION" },
+	{ NV03_PGRAPH_NSOURCE_DMA_VTX_PROTECTION, "DMA_VTX_PROTECTION" },
+	{ NV03_PGRAPH_NSOURCE_DMA_WIDTH_A,        "DMA_WIDTH_A" },
+	{ NV03_PGRAPH_NSOURCE_DMA_WIDTH_B,        "DMA_WIDTH_B" },
+	{}
+};
+
+static void
+nv04_graph_intr(struct nouveau_subdev *subdev)
+{
+	struct nv04_graph_priv *priv = (void *)subdev;
+	struct nv04_graph_chan *chan = NULL;
+	struct nouveau_namedb *namedb = NULL;
+	struct nouveau_handle *handle = NULL;
+	u32 stat = nv_rd32(priv, NV03_PGRAPH_INTR);
+	u32 nsource = nv_rd32(priv, NV03_PGRAPH_NSOURCE);
+	u32 nstatus = nv_rd32(priv, NV03_PGRAPH_NSTATUS);
+	u32 addr = nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR);
+	u32 chid = (addr & 0x0f000000) >> 24;
+	u32 subc = (addr & 0x0000e000) >> 13;
+	u32 mthd = (addr & 0x00001ffc);
+	u32 data = nv_rd32(priv, NV04_PGRAPH_TRAPPED_DATA);
+	u32 class = nv_rd32(priv, 0x400180 + subc * 4) & 0xff;
+	u32 inst = (nv_rd32(priv, 0x40016c) & 0xffff) << 4;
+	u32 show = stat;
+	unsigned long flags;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	chan = priv->chan[chid];
+	if (chan)
+		namedb = (void *)nv_pclass(nv_object(chan), NV_NAMEDB_CLASS);
+	spin_unlock_irqrestore(&priv->lock, flags);
+
+	if (stat & NV_PGRAPH_INTR_NOTIFY) {
+		if (chan && (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD)) {
+			handle = nouveau_namedb_get_vinst(namedb, inst);
+			if (handle && !nv_call(handle->object, mthd, data))
+				show &= ~NV_PGRAPH_INTR_NOTIFY;
+		}
+	}
+
+	if (stat & NV_PGRAPH_INTR_CONTEXT_SWITCH) {
+		nv_wr32(priv, NV03_PGRAPH_INTR, NV_PGRAPH_INTR_CONTEXT_SWITCH);
+		stat &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH;
+		show &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH;
+		nv04_graph_context_switch(priv);
+	}
+
+	nv_wr32(priv, NV03_PGRAPH_INTR, stat);
+	nv_wr32(priv, NV04_PGRAPH_FIFO, 0x00000001);
+
+	if (show) {
+		nv_error(priv, "");
+		nouveau_bitfield_print(nv04_graph_intr_name, show);
+		printk(" nsource:");
+		nouveau_bitfield_print(nv04_graph_nsource, nsource);
+		printk(" nstatus:");
+		nouveau_bitfield_print(nv04_graph_nstatus, nstatus);
+		printk("\n");
+		nv_error(priv, "ch %d/%d class 0x%04x "
+			       "mthd 0x%04x data 0x%08x\n",
+			 chid, subc, class, mthd, data);
+	}
+
+	nouveau_namedb_put(handle);
+}
+
+static int
+nv04_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		struct nouveau_oclass *oclass, void *data, u32 size,
+		struct nouveau_object **pobject)
+{
+	struct nv04_graph_priv *priv;
+	int ret;
+
+	ret = nouveau_graph_create(parent, engine, oclass, true, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00001000;
+	nv_subdev(priv)->intr = nv04_graph_intr;
+	nv_engine(priv)->cclass = &nv04_graph_cclass;
+	nv_engine(priv)->sclass = nv04_graph_sclass;
+	spin_lock_init(&priv->lock);
+	return 0;
+}
+
+static int
+nv04_graph_init(struct nouveau_object *object)
+{
+	struct nouveau_engine *engine = nv_engine(object);
+	struct nv04_graph_priv *priv = (void *)engine;
+	int ret;
+
+	ret = nouveau_graph_init(&priv->base);
+	if (ret)
+		return ret;
+
+	/* Enable PGRAPH interrupts */
+	nv_wr32(priv, NV03_PGRAPH_INTR, 0xFFFFFFFF);
+	nv_wr32(priv, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF);
+
+	nv_wr32(priv, NV04_PGRAPH_VALID1, 0);
+	nv_wr32(priv, NV04_PGRAPH_VALID2, 0);
+	/*nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x000001FF);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x001FFFFF);*/
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x1231c000);
+	/*1231C000 blob, 001 haiku*/
+	/*V_WRITE(NV04_PGRAPH_DEBUG_1, 0xf2d91100);*/
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_1, 0x72111100);
+	/*0x72111100 blob , 01 haiku*/
+	/*nv_wr32(priv, NV04_PGRAPH_DEBUG_2, 0x11d5f870);*/
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_2, 0x11d5f071);
+	/*haiku same*/
+
+	/*nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0xfad4ff31);*/
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0xf0d4ff31);
+	/*haiku and blob 10d4*/
+
+	nv_wr32(priv, NV04_PGRAPH_STATE        , 0xFFFFFFFF);
+	nv_wr32(priv, NV04_PGRAPH_CTX_CONTROL  , 0x10000100);
+	nv_mask(priv, NV04_PGRAPH_CTX_USER, 0xff000000, 0x0f000000);
+
+	/* These don't belong here, they're part of a per-channel context */
+	nv_wr32(priv, NV04_PGRAPH_PATTERN_SHAPE, 0x00000000);
+	nv_wr32(priv, NV04_PGRAPH_BETA_AND     , 0xFFFFFFFF);
+	return 0;
+}
+
+struct nouveau_oclass
+nv04_graph_oclass = {
+	.handle = NV_ENGINE(GR, 0x04),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv04_graph_ctor,
+		.dtor = _nouveau_graph_dtor,
+		.init = nv04_graph_init,
+		.fini = _nouveau_graph_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv10.c b/drivers/gpu/drm/nouveau/core/engine/graph/nv10.c
new file mode 100644
index 0000000..92521c8
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv10.c
@@ -0,0 +1,1314 @@
+/*
+ * Copyright 2007 Matthieu CASTET <castet.matthieu@free.fr>
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/handle.h>
+
+#include <subdev/fb.h>
+
+#include <engine/fifo.h>
+#include <engine/graph.h>
+
+#include "regs.h"
+
+struct pipe_state {
+	u32 pipe_0x0000[0x040/4];
+	u32 pipe_0x0040[0x010/4];
+	u32 pipe_0x0200[0x0c0/4];
+	u32 pipe_0x4400[0x080/4];
+	u32 pipe_0x6400[0x3b0/4];
+	u32 pipe_0x6800[0x2f0/4];
+	u32 pipe_0x6c00[0x030/4];
+	u32 pipe_0x7000[0x130/4];
+	u32 pipe_0x7400[0x0c0/4];
+	u32 pipe_0x7800[0x0c0/4];
+};
+
+static int nv10_graph_ctx_regs[] = {
+	NV10_PGRAPH_CTX_SWITCH(0),
+	NV10_PGRAPH_CTX_SWITCH(1),
+	NV10_PGRAPH_CTX_SWITCH(2),
+	NV10_PGRAPH_CTX_SWITCH(3),
+	NV10_PGRAPH_CTX_SWITCH(4),
+	NV10_PGRAPH_CTX_CACHE(0, 0),
+	NV10_PGRAPH_CTX_CACHE(0, 1),
+	NV10_PGRAPH_CTX_CACHE(0, 2),
+	NV10_PGRAPH_CTX_CACHE(0, 3),
+	NV10_PGRAPH_CTX_CACHE(0, 4),
+	NV10_PGRAPH_CTX_CACHE(1, 0),
+	NV10_PGRAPH_CTX_CACHE(1, 1),
+	NV10_PGRAPH_CTX_CACHE(1, 2),
+	NV10_PGRAPH_CTX_CACHE(1, 3),
+	NV10_PGRAPH_CTX_CACHE(1, 4),
+	NV10_PGRAPH_CTX_CACHE(2, 0),
+	NV10_PGRAPH_CTX_CACHE(2, 1),
+	NV10_PGRAPH_CTX_CACHE(2, 2),
+	NV10_PGRAPH_CTX_CACHE(2, 3),
+	NV10_PGRAPH_CTX_CACHE(2, 4),
+	NV10_PGRAPH_CTX_CACHE(3, 0),
+	NV10_PGRAPH_CTX_CACHE(3, 1),
+	NV10_PGRAPH_CTX_CACHE(3, 2),
+	NV10_PGRAPH_CTX_CACHE(3, 3),
+	NV10_PGRAPH_CTX_CACHE(3, 4),
+	NV10_PGRAPH_CTX_CACHE(4, 0),
+	NV10_PGRAPH_CTX_CACHE(4, 1),
+	NV10_PGRAPH_CTX_CACHE(4, 2),
+	NV10_PGRAPH_CTX_CACHE(4, 3),
+	NV10_PGRAPH_CTX_CACHE(4, 4),
+	NV10_PGRAPH_CTX_CACHE(5, 0),
+	NV10_PGRAPH_CTX_CACHE(5, 1),
+	NV10_PGRAPH_CTX_CACHE(5, 2),
+	NV10_PGRAPH_CTX_CACHE(5, 3),
+	NV10_PGRAPH_CTX_CACHE(5, 4),
+	NV10_PGRAPH_CTX_CACHE(6, 0),
+	NV10_PGRAPH_CTX_CACHE(6, 1),
+	NV10_PGRAPH_CTX_CACHE(6, 2),
+	NV10_PGRAPH_CTX_CACHE(6, 3),
+	NV10_PGRAPH_CTX_CACHE(6, 4),
+	NV10_PGRAPH_CTX_CACHE(7, 0),
+	NV10_PGRAPH_CTX_CACHE(7, 1),
+	NV10_PGRAPH_CTX_CACHE(7, 2),
+	NV10_PGRAPH_CTX_CACHE(7, 3),
+	NV10_PGRAPH_CTX_CACHE(7, 4),
+	NV10_PGRAPH_CTX_USER,
+	NV04_PGRAPH_DMA_START_0,
+	NV04_PGRAPH_DMA_START_1,
+	NV04_PGRAPH_DMA_LENGTH,
+	NV04_PGRAPH_DMA_MISC,
+	NV10_PGRAPH_DMA_PITCH,
+	NV04_PGRAPH_BOFFSET0,
+	NV04_PGRAPH_BBASE0,
+	NV04_PGRAPH_BLIMIT0,
+	NV04_PGRAPH_BOFFSET1,
+	NV04_PGRAPH_BBASE1,
+	NV04_PGRAPH_BLIMIT1,
+	NV04_PGRAPH_BOFFSET2,
+	NV04_PGRAPH_BBASE2,
+	NV04_PGRAPH_BLIMIT2,
+	NV04_PGRAPH_BOFFSET3,
+	NV04_PGRAPH_BBASE3,
+	NV04_PGRAPH_BLIMIT3,
+	NV04_PGRAPH_BOFFSET4,
+	NV04_PGRAPH_BBASE4,
+	NV04_PGRAPH_BLIMIT4,
+	NV04_PGRAPH_BOFFSET5,
+	NV04_PGRAPH_BBASE5,
+	NV04_PGRAPH_BLIMIT5,
+	NV04_PGRAPH_BPITCH0,
+	NV04_PGRAPH_BPITCH1,
+	NV04_PGRAPH_BPITCH2,
+	NV04_PGRAPH_BPITCH3,
+	NV04_PGRAPH_BPITCH4,
+	NV10_PGRAPH_SURFACE,
+	NV10_PGRAPH_STATE,
+	NV04_PGRAPH_BSWIZZLE2,
+	NV04_PGRAPH_BSWIZZLE5,
+	NV04_PGRAPH_BPIXEL,
+	NV10_PGRAPH_NOTIFY,
+	NV04_PGRAPH_PATT_COLOR0,
+	NV04_PGRAPH_PATT_COLOR1,
+	NV04_PGRAPH_PATT_COLORRAM, /* 64 values from 0x400900 to 0x4009fc */
+	0x00400904,
+	0x00400908,
+	0x0040090c,
+	0x00400910,
+	0x00400914,
+	0x00400918,
+	0x0040091c,
+	0x00400920,
+	0x00400924,
+	0x00400928,
+	0x0040092c,
+	0x00400930,
+	0x00400934,
+	0x00400938,
+	0x0040093c,
+	0x00400940,
+	0x00400944,
+	0x00400948,
+	0x0040094c,
+	0x00400950,
+	0x00400954,
+	0x00400958,
+	0x0040095c,
+	0x00400960,
+	0x00400964,
+	0x00400968,
+	0x0040096c,
+	0x00400970,
+	0x00400974,
+	0x00400978,
+	0x0040097c,
+	0x00400980,
+	0x00400984,
+	0x00400988,
+	0x0040098c,
+	0x00400990,
+	0x00400994,
+	0x00400998,
+	0x0040099c,
+	0x004009a0,
+	0x004009a4,
+	0x004009a8,
+	0x004009ac,
+	0x004009b0,
+	0x004009b4,
+	0x004009b8,
+	0x004009bc,
+	0x004009c0,
+	0x004009c4,
+	0x004009c8,
+	0x004009cc,
+	0x004009d0,
+	0x004009d4,
+	0x004009d8,
+	0x004009dc,
+	0x004009e0,
+	0x004009e4,
+	0x004009e8,
+	0x004009ec,
+	0x004009f0,
+	0x004009f4,
+	0x004009f8,
+	0x004009fc,
+	NV04_PGRAPH_PATTERN,	/* 2 values from 0x400808 to 0x40080c */
+	0x0040080c,
+	NV04_PGRAPH_PATTERN_SHAPE,
+	NV03_PGRAPH_MONO_COLOR0,
+	NV04_PGRAPH_ROP3,
+	NV04_PGRAPH_CHROMA,
+	NV04_PGRAPH_BETA_AND,
+	NV04_PGRAPH_BETA_PREMULT,
+	0x00400e70,
+	0x00400e74,
+	0x00400e78,
+	0x00400e7c,
+	0x00400e80,
+	0x00400e84,
+	0x00400e88,
+	0x00400e8c,
+	0x00400ea0,
+	0x00400ea4,
+	0x00400ea8,
+	0x00400e90,
+	0x00400e94,
+	0x00400e98,
+	0x00400e9c,
+	NV10_PGRAPH_WINDOWCLIP_HORIZONTAL, /* 8 values from 0x400f00-0x400f1c */
+	NV10_PGRAPH_WINDOWCLIP_VERTICAL,   /* 8 values from 0x400f20-0x400f3c */
+	0x00400f04,
+	0x00400f24,
+	0x00400f08,
+	0x00400f28,
+	0x00400f0c,
+	0x00400f2c,
+	0x00400f10,
+	0x00400f30,
+	0x00400f14,
+	0x00400f34,
+	0x00400f18,
+	0x00400f38,
+	0x00400f1c,
+	0x00400f3c,
+	NV10_PGRAPH_XFMODE0,
+	NV10_PGRAPH_XFMODE1,
+	NV10_PGRAPH_GLOBALSTATE0,
+	NV10_PGRAPH_GLOBALSTATE1,
+	NV04_PGRAPH_STORED_FMT,
+	NV04_PGRAPH_SOURCE_COLOR,
+	NV03_PGRAPH_ABS_X_RAM,	/* 32 values from 0x400400 to 0x40047c */
+	NV03_PGRAPH_ABS_Y_RAM,	/* 32 values from 0x400480 to 0x4004fc */
+	0x00400404,
+	0x00400484,
+	0x00400408,
+	0x00400488,
+	0x0040040c,
+	0x0040048c,
+	0x00400410,
+	0x00400490,
+	0x00400414,
+	0x00400494,
+	0x00400418,
+	0x00400498,
+	0x0040041c,
+	0x0040049c,
+	0x00400420,
+	0x004004a0,
+	0x00400424,
+	0x004004a4,
+	0x00400428,
+	0x004004a8,
+	0x0040042c,
+	0x004004ac,
+	0x00400430,
+	0x004004b0,
+	0x00400434,
+	0x004004b4,
+	0x00400438,
+	0x004004b8,
+	0x0040043c,
+	0x004004bc,
+	0x00400440,
+	0x004004c0,
+	0x00400444,
+	0x004004c4,
+	0x00400448,
+	0x004004c8,
+	0x0040044c,
+	0x004004cc,
+	0x00400450,
+	0x004004d0,
+	0x00400454,
+	0x004004d4,
+	0x00400458,
+	0x004004d8,
+	0x0040045c,
+	0x004004dc,
+	0x00400460,
+	0x004004e0,
+	0x00400464,
+	0x004004e4,
+	0x00400468,
+	0x004004e8,
+	0x0040046c,
+	0x004004ec,
+	0x00400470,
+	0x004004f0,
+	0x00400474,
+	0x004004f4,
+	0x00400478,
+	0x004004f8,
+	0x0040047c,
+	0x004004fc,
+	NV03_PGRAPH_ABS_UCLIP_XMIN,
+	NV03_PGRAPH_ABS_UCLIP_XMAX,
+	NV03_PGRAPH_ABS_UCLIP_YMIN,
+	NV03_PGRAPH_ABS_UCLIP_YMAX,
+	0x00400550,
+	0x00400558,
+	0x00400554,
+	0x0040055c,
+	NV03_PGRAPH_ABS_UCLIPA_XMIN,
+	NV03_PGRAPH_ABS_UCLIPA_XMAX,
+	NV03_PGRAPH_ABS_UCLIPA_YMIN,
+	NV03_PGRAPH_ABS_UCLIPA_YMAX,
+	NV03_PGRAPH_ABS_ICLIP_XMAX,
+	NV03_PGRAPH_ABS_ICLIP_YMAX,
+	NV03_PGRAPH_XY_LOGIC_MISC0,
+	NV03_PGRAPH_XY_LOGIC_MISC1,
+	NV03_PGRAPH_XY_LOGIC_MISC2,
+	NV03_PGRAPH_XY_LOGIC_MISC3,
+	NV03_PGRAPH_CLIPX_0,
+	NV03_PGRAPH_CLIPX_1,
+	NV03_PGRAPH_CLIPY_0,
+	NV03_PGRAPH_CLIPY_1,
+	NV10_PGRAPH_COMBINER0_IN_ALPHA,
+	NV10_PGRAPH_COMBINER1_IN_ALPHA,
+	NV10_PGRAPH_COMBINER0_IN_RGB,
+	NV10_PGRAPH_COMBINER1_IN_RGB,
+	NV10_PGRAPH_COMBINER_COLOR0,
+	NV10_PGRAPH_COMBINER_COLOR1,
+	NV10_PGRAPH_COMBINER0_OUT_ALPHA,
+	NV10_PGRAPH_COMBINER1_OUT_ALPHA,
+	NV10_PGRAPH_COMBINER0_OUT_RGB,
+	NV10_PGRAPH_COMBINER1_OUT_RGB,
+	NV10_PGRAPH_COMBINER_FINAL0,
+	NV10_PGRAPH_COMBINER_FINAL1,
+	0x00400e00,
+	0x00400e04,
+	0x00400e08,
+	0x00400e0c,
+	0x00400e10,
+	0x00400e14,
+	0x00400e18,
+	0x00400e1c,
+	0x00400e20,
+	0x00400e24,
+	0x00400e28,
+	0x00400e2c,
+	0x00400e30,
+	0x00400e34,
+	0x00400e38,
+	0x00400e3c,
+	NV04_PGRAPH_PASSTHRU_0,
+	NV04_PGRAPH_PASSTHRU_1,
+	NV04_PGRAPH_PASSTHRU_2,
+	NV10_PGRAPH_DIMX_TEXTURE,
+	NV10_PGRAPH_WDIMX_TEXTURE,
+	NV10_PGRAPH_DVD_COLORFMT,
+	NV10_PGRAPH_SCALED_FORMAT,
+	NV04_PGRAPH_MISC24_0,
+	NV04_PGRAPH_MISC24_1,
+	NV04_PGRAPH_MISC24_2,
+	NV03_PGRAPH_X_MISC,
+	NV03_PGRAPH_Y_MISC,
+	NV04_PGRAPH_VALID1,
+	NV04_PGRAPH_VALID2,
+};
+
+static int nv17_graph_ctx_regs[] = {
+	NV10_PGRAPH_DEBUG_4,
+	0x004006b0,
+	0x00400eac,
+	0x00400eb0,
+	0x00400eb4,
+	0x00400eb8,
+	0x00400ebc,
+	0x00400ec0,
+	0x00400ec4,
+	0x00400ec8,
+	0x00400ecc,
+	0x00400ed0,
+	0x00400ed4,
+	0x00400ed8,
+	0x00400edc,
+	0x00400ee0,
+	0x00400a00,
+	0x00400a04,
+};
+
+struct nv10_graph_priv {
+	struct nouveau_graph base;
+	struct nv10_graph_chan *chan[32];
+	spinlock_t lock;
+};
+
+struct nv10_graph_chan {
+	struct nouveau_object base;
+	int chid;
+	int nv10[ARRAY_SIZE(nv10_graph_ctx_regs)];
+	int nv17[ARRAY_SIZE(nv17_graph_ctx_regs)];
+	struct pipe_state pipe_state;
+	u32 lma_window[4];
+};
+
+
+static inline struct nv10_graph_priv *
+nv10_graph_priv(struct nv10_graph_chan *chan)
+{
+	return (void *)nv_object(chan)->engine;
+}
+
+/*******************************************************************************
+ * Graphics object classes
+ ******************************************************************************/
+
+#define PIPE_SAVE(priv, state, addr)					\
+	do {								\
+		int __i;						\
+		nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, addr);		\
+		for (__i = 0; __i < ARRAY_SIZE(state); __i++)		\
+			state[__i] = nv_rd32(priv, NV10_PGRAPH_PIPE_DATA); \
+	} while (0)
+
+#define PIPE_RESTORE(priv, state, addr)					\
+	do {								\
+		int __i;						\
+		nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, addr);		\
+		for (__i = 0; __i < ARRAY_SIZE(state); __i++)		\
+			nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, state[__i]); \
+	} while (0)
+
+static struct nouveau_oclass
+nv10_graph_sclass[] = {
+	{ 0x0012, &nv04_graph_ofuncs }, /* beta1 */
+	{ 0x0019, &nv04_graph_ofuncs }, /* clip */
+	{ 0x0030, &nv04_graph_ofuncs }, /* null */
+	{ 0x0039, &nv04_graph_ofuncs }, /* m2mf */
+	{ 0x0043, &nv04_graph_ofuncs }, /* rop */
+	{ 0x0044, &nv04_graph_ofuncs }, /* pattern */
+	{ 0x004a, &nv04_graph_ofuncs }, /* gdi */
+	{ 0x0052, &nv04_graph_ofuncs }, /* swzsurf */
+	{ 0x005f, &nv04_graph_ofuncs }, /* blit */
+	{ 0x0062, &nv04_graph_ofuncs }, /* surf2d */
+	{ 0x0072, &nv04_graph_ofuncs }, /* beta4 */
+	{ 0x0089, &nv04_graph_ofuncs }, /* sifm */
+	{ 0x008a, &nv04_graph_ofuncs }, /* ifc */
+	{ 0x009f, &nv04_graph_ofuncs }, /* blit */
+	{ 0x0093, &nv04_graph_ofuncs }, /* surf3d */
+	{ 0x0094, &nv04_graph_ofuncs }, /* ttri */
+	{ 0x0095, &nv04_graph_ofuncs }, /* mtri */
+	{ 0x0056, &nv04_graph_ofuncs }, /* celcius */
+	{},
+};
+
+static struct nouveau_oclass
+nv15_graph_sclass[] = {
+	{ 0x0012, &nv04_graph_ofuncs }, /* beta1 */
+	{ 0x0019, &nv04_graph_ofuncs }, /* clip */
+	{ 0x0030, &nv04_graph_ofuncs }, /* null */
+	{ 0x0039, &nv04_graph_ofuncs }, /* m2mf */
+	{ 0x0043, &nv04_graph_ofuncs }, /* rop */
+	{ 0x0044, &nv04_graph_ofuncs }, /* pattern */
+	{ 0x004a, &nv04_graph_ofuncs }, /* gdi */
+	{ 0x0052, &nv04_graph_ofuncs }, /* swzsurf */
+	{ 0x005f, &nv04_graph_ofuncs }, /* blit */
+	{ 0x0062, &nv04_graph_ofuncs }, /* surf2d */
+	{ 0x0072, &nv04_graph_ofuncs }, /* beta4 */
+	{ 0x0089, &nv04_graph_ofuncs }, /* sifm */
+	{ 0x008a, &nv04_graph_ofuncs }, /* ifc */
+	{ 0x009f, &nv04_graph_ofuncs }, /* blit */
+	{ 0x0093, &nv04_graph_ofuncs }, /* surf3d */
+	{ 0x0094, &nv04_graph_ofuncs }, /* ttri */
+	{ 0x0095, &nv04_graph_ofuncs }, /* mtri */
+	{ 0x0096, &nv04_graph_ofuncs }, /* celcius */
+	{},
+};
+
+static int
+nv17_graph_mthd_lma_window(struct nouveau_object *object, u32 mthd,
+			   void *args, u32 size)
+{
+	struct nv10_graph_chan *chan = (void *)object->parent;
+	struct nv10_graph_priv *priv = nv10_graph_priv(chan);
+	struct pipe_state *pipe = &chan->pipe_state;
+	u32 pipe_0x0040[1], pipe_0x64c0[8], pipe_0x6a80[3], pipe_0x6ab0[3];
+	u32 xfmode0, xfmode1;
+	u32 data = *(u32 *)args;
+	int i;
+
+	chan->lma_window[(mthd - 0x1638) / 4] = data;
+
+	if (mthd != 0x1644)
+		return 0;
+
+	nv04_graph_idle(priv);
+
+	PIPE_SAVE(priv, pipe_0x0040, 0x0040);
+	PIPE_SAVE(priv, pipe->pipe_0x0200, 0x0200);
+
+	PIPE_RESTORE(priv, chan->lma_window, 0x6790);
+
+	nv04_graph_idle(priv);
+
+	xfmode0 = nv_rd32(priv, NV10_PGRAPH_XFMODE0);
+	xfmode1 = nv_rd32(priv, NV10_PGRAPH_XFMODE1);
+
+	PIPE_SAVE(priv, pipe->pipe_0x4400, 0x4400);
+	PIPE_SAVE(priv, pipe_0x64c0, 0x64c0);
+	PIPE_SAVE(priv, pipe_0x6ab0, 0x6ab0);
+	PIPE_SAVE(priv, pipe_0x6a80, 0x6a80);
+
+	nv04_graph_idle(priv);
+
+	nv_wr32(priv, NV10_PGRAPH_XFMODE0, 0x10000000);
+	nv_wr32(priv, NV10_PGRAPH_XFMODE1, 0x00000000);
+	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0);
+	for (i = 0; i < 4; i++)
+		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
+	for (i = 0; i < 4; i++)
+		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+
+	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0);
+	for (i = 0; i < 3; i++)
+		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
+
+	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80);
+	for (i = 0; i < 3; i++)
+		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+
+	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040);
+	nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000008);
+
+	PIPE_RESTORE(priv, pipe->pipe_0x0200, 0x0200);
+
+	nv04_graph_idle(priv);
+
+	PIPE_RESTORE(priv, pipe_0x0040, 0x0040);
+
+	nv_wr32(priv, NV10_PGRAPH_XFMODE0, xfmode0);
+	nv_wr32(priv, NV10_PGRAPH_XFMODE1, xfmode1);
+
+	PIPE_RESTORE(priv, pipe_0x64c0, 0x64c0);
+	PIPE_RESTORE(priv, pipe_0x6ab0, 0x6ab0);
+	PIPE_RESTORE(priv, pipe_0x6a80, 0x6a80);
+	PIPE_RESTORE(priv, pipe->pipe_0x4400, 0x4400);
+
+	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x000000c0);
+	nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+
+	nv04_graph_idle(priv);
+
+	return 0;
+}
+
+static int
+nv17_graph_mthd_lma_enable(struct nouveau_object *object, u32 mthd,
+			   void *args, u32 size)
+{
+	struct nv10_graph_chan *chan = (void *)object->parent;
+	struct nv10_graph_priv *priv = nv10_graph_priv(chan);
+
+	nv04_graph_idle(priv);
+
+	nv_mask(priv, NV10_PGRAPH_DEBUG_4, 0x00000100, 0x00000100);
+	nv_mask(priv, 0x4006b0, 0x08000000, 0x08000000);
+	return 0;
+}
+
+static struct nouveau_omthds
+nv17_celcius_omthds[] = {
+	{ 0x1638, nv17_graph_mthd_lma_window },
+	{ 0x163c, nv17_graph_mthd_lma_window },
+	{ 0x1640, nv17_graph_mthd_lma_window },
+	{ 0x1644, nv17_graph_mthd_lma_window },
+	{ 0x1658, nv17_graph_mthd_lma_enable },
+	{}
+};
+
+static struct nouveau_oclass
+nv17_graph_sclass[] = {
+	{ 0x0012, &nv04_graph_ofuncs }, /* beta1 */
+	{ 0x0019, &nv04_graph_ofuncs }, /* clip */
+	{ 0x0030, &nv04_graph_ofuncs }, /* null */
+	{ 0x0039, &nv04_graph_ofuncs }, /* m2mf */
+	{ 0x0043, &nv04_graph_ofuncs }, /* rop */
+	{ 0x0044, &nv04_graph_ofuncs }, /* pattern */
+	{ 0x004a, &nv04_graph_ofuncs }, /* gdi */
+	{ 0x0052, &nv04_graph_ofuncs }, /* swzsurf */
+	{ 0x005f, &nv04_graph_ofuncs }, /* blit */
+	{ 0x0062, &nv04_graph_ofuncs }, /* surf2d */
+	{ 0x0072, &nv04_graph_ofuncs }, /* beta4 */
+	{ 0x0089, &nv04_graph_ofuncs }, /* sifm */
+	{ 0x008a, &nv04_graph_ofuncs }, /* ifc */
+	{ 0x009f, &nv04_graph_ofuncs }, /* blit */
+	{ 0x0093, &nv04_graph_ofuncs }, /* surf3d */
+	{ 0x0094, &nv04_graph_ofuncs }, /* ttri */
+	{ 0x0095, &nv04_graph_ofuncs }, /* mtri */
+	{ 0x0099, &nv04_graph_ofuncs, nv17_celcius_omthds },
+	{},
+};
+
+/*******************************************************************************
+ * PGRAPH context
+ ******************************************************************************/
+
+static struct nv10_graph_chan *
+nv10_graph_channel(struct nv10_graph_priv *priv)
+{
+	struct nv10_graph_chan *chan = NULL;
+	if (nv_rd32(priv, 0x400144) & 0x00010000) {
+		int chid = nv_rd32(priv, 0x400148) >> 24;
+		if (chid < ARRAY_SIZE(priv->chan))
+			chan = priv->chan[chid];
+	}
+	return chan;
+}
+
+static void
+nv10_graph_save_pipe(struct nv10_graph_chan *chan)
+{
+	struct nv10_graph_priv *priv = nv10_graph_priv(chan);
+	struct pipe_state *pipe = &chan->pipe_state;
+
+	PIPE_SAVE(priv, pipe->pipe_0x4400, 0x4400);
+	PIPE_SAVE(priv, pipe->pipe_0x0200, 0x0200);
+	PIPE_SAVE(priv, pipe->pipe_0x6400, 0x6400);
+	PIPE_SAVE(priv, pipe->pipe_0x6800, 0x6800);
+	PIPE_SAVE(priv, pipe->pipe_0x6c00, 0x6c00);
+	PIPE_SAVE(priv, pipe->pipe_0x7000, 0x7000);
+	PIPE_SAVE(priv, pipe->pipe_0x7400, 0x7400);
+	PIPE_SAVE(priv, pipe->pipe_0x7800, 0x7800);
+	PIPE_SAVE(priv, pipe->pipe_0x0040, 0x0040);
+	PIPE_SAVE(priv, pipe->pipe_0x0000, 0x0000);
+}
+
+static void
+nv10_graph_load_pipe(struct nv10_graph_chan *chan)
+{
+	struct nv10_graph_priv *priv = nv10_graph_priv(chan);
+	struct pipe_state *pipe = &chan->pipe_state;
+	u32 xfmode0, xfmode1;
+	int i;
+
+	nv04_graph_idle(priv);
+	/* XXX check haiku comments */
+	xfmode0 = nv_rd32(priv, NV10_PGRAPH_XFMODE0);
+	xfmode1 = nv_rd32(priv, NV10_PGRAPH_XFMODE1);
+	nv_wr32(priv, NV10_PGRAPH_XFMODE0, 0x10000000);
+	nv_wr32(priv, NV10_PGRAPH_XFMODE1, 0x00000000);
+	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0);
+	for (i = 0; i < 4; i++)
+		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
+	for (i = 0; i < 4; i++)
+		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+
+	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0);
+	for (i = 0; i < 3; i++)
+		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
+
+	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80);
+	for (i = 0; i < 3; i++)
+		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+
+	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040);
+	nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000008);
+
+
+	PIPE_RESTORE(priv, pipe->pipe_0x0200, 0x0200);
+	nv04_graph_idle(priv);
+
+	/* restore XFMODE */
+	nv_wr32(priv, NV10_PGRAPH_XFMODE0, xfmode0);
+	nv_wr32(priv, NV10_PGRAPH_XFMODE1, xfmode1);
+	PIPE_RESTORE(priv, pipe->pipe_0x6400, 0x6400);
+	PIPE_RESTORE(priv, pipe->pipe_0x6800, 0x6800);
+	PIPE_RESTORE(priv, pipe->pipe_0x6c00, 0x6c00);
+	PIPE_RESTORE(priv, pipe->pipe_0x7000, 0x7000);
+	PIPE_RESTORE(priv, pipe->pipe_0x7400, 0x7400);
+	PIPE_RESTORE(priv, pipe->pipe_0x7800, 0x7800);
+	PIPE_RESTORE(priv, pipe->pipe_0x4400, 0x4400);
+	PIPE_RESTORE(priv, pipe->pipe_0x0000, 0x0000);
+	PIPE_RESTORE(priv, pipe->pipe_0x0040, 0x0040);
+	nv04_graph_idle(priv);
+}
+
+static void
+nv10_graph_create_pipe(struct nv10_graph_chan *chan)
+{
+	struct nv10_graph_priv *priv = nv10_graph_priv(chan);
+	struct pipe_state *pipe_state = &chan->pipe_state;
+	u32 *pipe_state_addr;
+	int i;
+#define PIPE_INIT(addr) \
+	do { \
+		pipe_state_addr = pipe_state->pipe_##addr; \
+	} while (0)
+#define PIPE_INIT_END(addr) \
+	do { \
+		u32 *__end_addr = pipe_state->pipe_##addr + \
+				ARRAY_SIZE(pipe_state->pipe_##addr); \
+		if (pipe_state_addr != __end_addr) \
+			nv_error(priv, "incomplete pipe init for 0x%x :  %p/%p\n", \
+				addr, pipe_state_addr, __end_addr); \
+	} while (0)
+#define NV_WRITE_PIPE_INIT(value) *(pipe_state_addr++) = value
+
+	PIPE_INIT(0x0200);
+	for (i = 0; i < 48; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x0200);
+
+	PIPE_INIT(0x6400);
+	for (i = 0; i < 211; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	NV_WRITE_PIPE_INIT(0x40000000);
+	NV_WRITE_PIPE_INIT(0x40000000);
+	NV_WRITE_PIPE_INIT(0x40000000);
+	NV_WRITE_PIPE_INIT(0x40000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x3f000000);
+	NV_WRITE_PIPE_INIT(0x3f000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	PIPE_INIT_END(0x6400);
+
+	PIPE_INIT(0x6800);
+	for (i = 0; i < 162; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	for (i = 0; i < 25; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x6800);
+
+	PIPE_INIT(0x6c00);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0xbf800000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x6c00);
+
+	PIPE_INIT(0x7000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	for (i = 0; i < 35; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x7000);
+
+	PIPE_INIT(0x7400);
+	for (i = 0; i < 48; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x7400);
+
+	PIPE_INIT(0x7800);
+	for (i = 0; i < 48; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x7800);
+
+	PIPE_INIT(0x4400);
+	for (i = 0; i < 32; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x4400);
+
+	PIPE_INIT(0x0000);
+	for (i = 0; i < 16; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x0000);
+
+	PIPE_INIT(0x0040);
+	for (i = 0; i < 4; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x0040);
+
+#undef PIPE_INIT
+#undef PIPE_INIT_END
+#undef NV_WRITE_PIPE_INIT
+}
+
+static int
+nv10_graph_ctx_regs_find_offset(struct nv10_graph_priv *priv, int reg)
+{
+	int i;
+	for (i = 0; i < ARRAY_SIZE(nv10_graph_ctx_regs); i++) {
+		if (nv10_graph_ctx_regs[i] == reg)
+			return i;
+	}
+	nv_error(priv, "unknow offset nv10_ctx_regs %d\n", reg);
+	return -1;
+}
+
+static int
+nv17_graph_ctx_regs_find_offset(struct nv10_graph_priv *priv, int reg)
+{
+	int i;
+	for (i = 0; i < ARRAY_SIZE(nv17_graph_ctx_regs); i++) {
+		if (nv17_graph_ctx_regs[i] == reg)
+			return i;
+	}
+	nv_error(priv, "unknow offset nv17_ctx_regs %d\n", reg);
+	return -1;
+}
+
+static void
+nv10_graph_load_dma_vtxbuf(struct nv10_graph_chan *chan, int chid, u32 inst)
+{
+	struct nv10_graph_priv *priv = nv10_graph_priv(chan);
+	u32 st2, st2_dl, st2_dh, fifo_ptr, fifo[0x60/4];
+	u32 ctx_user, ctx_switch[5];
+	int i, subchan = -1;
+
+	/* NV10TCL_DMA_VTXBUF (method 0x18c) modifies hidden state
+	 * that cannot be restored via MMIO. Do it through the FIFO
+	 * instead.
+	 */
+
+	/* Look for a celsius object */
+	for (i = 0; i < 8; i++) {
+		int class = nv_rd32(priv, NV10_PGRAPH_CTX_CACHE(i, 0)) & 0xfff;
+
+		if (class == 0x56 || class == 0x96 || class == 0x99) {
+			subchan = i;
+			break;
+		}
+	}
+
+	if (subchan < 0 || !inst)
+		return;
+
+	/* Save the current ctx object */
+	ctx_user = nv_rd32(priv, NV10_PGRAPH_CTX_USER);
+	for (i = 0; i < 5; i++)
+		ctx_switch[i] = nv_rd32(priv, NV10_PGRAPH_CTX_SWITCH(i));
+
+	/* Save the FIFO state */
+	st2 = nv_rd32(priv, NV10_PGRAPH_FFINTFC_ST2);
+	st2_dl = nv_rd32(priv, NV10_PGRAPH_FFINTFC_ST2_DL);
+	st2_dh = nv_rd32(priv, NV10_PGRAPH_FFINTFC_ST2_DH);
+	fifo_ptr = nv_rd32(priv, NV10_PGRAPH_FFINTFC_FIFO_PTR);
+
+	for (i = 0; i < ARRAY_SIZE(fifo); i++)
+		fifo[i] = nv_rd32(priv, 0x4007a0 + 4 * i);
+
+	/* Switch to the celsius subchannel */
+	for (i = 0; i < 5; i++)
+		nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(i),
+			nv_rd32(priv, NV10_PGRAPH_CTX_CACHE(subchan, i)));
+	nv_mask(priv, NV10_PGRAPH_CTX_USER, 0xe000, subchan << 13);
+
+	/* Inject NV10TCL_DMA_VTXBUF */
+	nv_wr32(priv, NV10_PGRAPH_FFINTFC_FIFO_PTR, 0);
+	nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2,
+		0x2c000000 | chid << 20 | subchan << 16 | 0x18c);
+	nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2_DL, inst);
+	nv_mask(priv, NV10_PGRAPH_CTX_CONTROL, 0, 0x10000);
+	nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001);
+	nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000);
+
+	/* Restore the FIFO state */
+	for (i = 0; i < ARRAY_SIZE(fifo); i++)
+		nv_wr32(priv, 0x4007a0 + 4 * i, fifo[i]);
+
+	nv_wr32(priv, NV10_PGRAPH_FFINTFC_FIFO_PTR, fifo_ptr);
+	nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2, st2);
+	nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2_DL, st2_dl);
+	nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2_DH, st2_dh);
+
+	/* Restore the current ctx object */
+	for (i = 0; i < 5; i++)
+		nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(i), ctx_switch[i]);
+	nv_wr32(priv, NV10_PGRAPH_CTX_USER, ctx_user);
+}
+
+static int
+nv10_graph_load_context(struct nv10_graph_chan *chan, int chid)
+{
+	struct nv10_graph_priv *priv = nv10_graph_priv(chan);
+	u32 inst;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(nv10_graph_ctx_regs); i++)
+		nv_wr32(priv, nv10_graph_ctx_regs[i], chan->nv10[i]);
+
+	if (nv_device(priv)->chipset >= 0x17) {
+		for (i = 0; i < ARRAY_SIZE(nv17_graph_ctx_regs); i++)
+			nv_wr32(priv, nv17_graph_ctx_regs[i], chan->nv17[i]);
+	}
+
+	nv10_graph_load_pipe(chan);
+
+	inst = nv_rd32(priv, NV10_PGRAPH_GLOBALSTATE1) & 0xffff;
+	nv10_graph_load_dma_vtxbuf(chan, chid, inst);
+
+	nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10010100);
+	nv_mask(priv, NV10_PGRAPH_CTX_USER, 0xff000000, chid << 24);
+	nv_mask(priv, NV10_PGRAPH_FFINTFC_ST2, 0x30000000, 0x00000000);
+	return 0;
+}
+
+static int
+nv10_graph_unload_context(struct nv10_graph_chan *chan)
+{
+	struct nv10_graph_priv *priv = nv10_graph_priv(chan);
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(nv10_graph_ctx_regs); i++)
+		chan->nv10[i] = nv_rd32(priv, nv10_graph_ctx_regs[i]);
+
+	if (nv_device(priv)->chipset >= 0x17) {
+		for (i = 0; i < ARRAY_SIZE(nv17_graph_ctx_regs); i++)
+			chan->nv17[i] = nv_rd32(priv, nv17_graph_ctx_regs[i]);
+	}
+
+	nv10_graph_save_pipe(chan);
+
+	nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10000000);
+	nv_mask(priv, NV10_PGRAPH_CTX_USER, 0xff000000, 0x1f000000);
+	return 0;
+}
+
+static void
+nv10_graph_context_switch(struct nv10_graph_priv *priv)
+{
+	struct nv10_graph_chan *prev = NULL;
+	struct nv10_graph_chan *next = NULL;
+	unsigned long flags;
+	int chid;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	nv04_graph_idle(priv);
+
+	/* If previous context is valid, we need to save it */
+	prev = nv10_graph_channel(priv);
+	if (prev)
+		nv10_graph_unload_context(prev);
+
+	/* load context for next channel */
+	chid = (nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR) >> 20) & 0x1f;
+	next = priv->chan[chid];
+	if (next)
+		nv10_graph_load_context(next, chid);
+
+	spin_unlock_irqrestore(&priv->lock, flags);
+}
+
+#define NV_WRITE_CTX(reg, val) do { \
+	int offset = nv10_graph_ctx_regs_find_offset(priv, reg); \
+	if (offset > 0) \
+		chan->nv10[offset] = val; \
+	} while (0)
+
+#define NV17_WRITE_CTX(reg, val) do { \
+	int offset = nv17_graph_ctx_regs_find_offset(priv, reg); \
+	if (offset > 0) \
+		chan->nv17[offset] = val; \
+	} while (0)
+
+static int
+nv10_graph_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nouveau_fifo_chan *fifo = (void *)parent;
+	struct nv10_graph_priv *priv = (void *)engine;
+	struct nv10_graph_chan *chan;
+	unsigned long flags;
+	int ret;
+
+	ret = nouveau_object_create(parent, engine, oclass, 0, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	if (priv->chan[fifo->chid]) {
+		*pobject = nv_object(priv->chan[fifo->chid]);
+		atomic_inc(&(*pobject)->refcount);
+		spin_unlock_irqrestore(&priv->lock, flags);
+		nouveau_object_destroy(&chan->base);
+		return 1;
+	}
+
+	NV_WRITE_CTX(0x00400e88, 0x08000000);
+	NV_WRITE_CTX(0x00400e9c, 0x4b7fffff);
+	NV_WRITE_CTX(NV03_PGRAPH_XY_LOGIC_MISC0, 0x0001ffff);
+	NV_WRITE_CTX(0x00400e10, 0x00001000);
+	NV_WRITE_CTX(0x00400e14, 0x00001000);
+	NV_WRITE_CTX(0x00400e30, 0x00080008);
+	NV_WRITE_CTX(0x00400e34, 0x00080008);
+	if (nv_device(priv)->chipset >= 0x17) {
+		/* is it really needed ??? */
+		NV17_WRITE_CTX(NV10_PGRAPH_DEBUG_4,
+					nv_rd32(priv, NV10_PGRAPH_DEBUG_4));
+		NV17_WRITE_CTX(0x004006b0, nv_rd32(priv, 0x004006b0));
+		NV17_WRITE_CTX(0x00400eac, 0x0fff0000);
+		NV17_WRITE_CTX(0x00400eb0, 0x0fff0000);
+		NV17_WRITE_CTX(0x00400ec0, 0x00000080);
+		NV17_WRITE_CTX(0x00400ed0, 0x00000080);
+	}
+	NV_WRITE_CTX(NV10_PGRAPH_CTX_USER, chan->chid << 24);
+
+	nv10_graph_create_pipe(chan);
+
+	priv->chan[fifo->chid] = chan;
+	chan->chid = fifo->chid;
+	spin_unlock_irqrestore(&priv->lock, flags);
+	return 0;
+}
+
+static void
+nv10_graph_context_dtor(struct nouveau_object *object)
+{
+	struct nv10_graph_priv *priv = (void *)object->engine;
+	struct nv10_graph_chan *chan = (void *)object;
+	unsigned long flags;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	priv->chan[chan->chid] = NULL;
+	spin_unlock_irqrestore(&priv->lock, flags);
+
+	nouveau_object_destroy(&chan->base);
+}
+
+static int
+nv10_graph_context_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv10_graph_priv *priv = (void *)object->engine;
+	struct nv10_graph_chan *chan = (void *)object;
+	unsigned long flags;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000);
+	if (nv10_graph_channel(priv) == chan)
+		nv10_graph_unload_context(chan);
+	nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001);
+	spin_unlock_irqrestore(&priv->lock, flags);
+
+	return nouveau_object_fini(&chan->base, suspend);
+}
+
+static struct nouveau_oclass
+nv10_graph_cclass = {
+	.handle = NV_ENGCTX(GR, 0x10),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv10_graph_context_ctor,
+		.dtor = nv10_graph_context_dtor,
+		.init = nouveau_object_init,
+		.fini = nv10_graph_context_fini,
+	},
+};
+
+/*******************************************************************************
+ * PGRAPH engine/subdev functions
+ ******************************************************************************/
+
+static void
+nv10_graph_tile_prog(struct nouveau_engine *engine, int i)
+{
+	struct nouveau_fb_tile *tile = &nouveau_fb(engine)->tile.region[i];
+	struct nouveau_fifo *pfifo = nouveau_fifo(engine);
+	struct nv10_graph_priv *priv = (void *)engine;
+	unsigned long flags;
+
+	pfifo->pause(pfifo, &flags);
+	nv04_graph_idle(priv);
+
+	nv_wr32(priv, NV10_PGRAPH_TLIMIT(i), tile->limit);
+	nv_wr32(priv, NV10_PGRAPH_TSIZE(i), tile->pitch);
+	nv_wr32(priv, NV10_PGRAPH_TILE(i), tile->addr);
+
+	pfifo->start(pfifo, &flags);
+}
+
+const struct nouveau_bitfield nv10_graph_intr_name[] = {
+	{ NV_PGRAPH_INTR_NOTIFY, "NOTIFY" },
+	{ NV_PGRAPH_INTR_ERROR,  "ERROR"  },
+	{}
+};
+
+const struct nouveau_bitfield nv10_graph_nstatus[] = {
+	{ NV10_PGRAPH_NSTATUS_STATE_IN_USE,       "STATE_IN_USE" },
+	{ NV10_PGRAPH_NSTATUS_INVALID_STATE,      "INVALID_STATE" },
+	{ NV10_PGRAPH_NSTATUS_BAD_ARGUMENT,       "BAD_ARGUMENT" },
+	{ NV10_PGRAPH_NSTATUS_PROTECTION_FAULT,   "PROTECTION_FAULT" },
+	{}
+};
+
+static void
+nv10_graph_intr(struct nouveau_subdev *subdev)
+{
+	struct nv10_graph_priv *priv = (void *)subdev;
+	struct nv10_graph_chan *chan = NULL;
+	struct nouveau_namedb *namedb = NULL;
+	struct nouveau_handle *handle = NULL;
+	u32 stat = nv_rd32(priv, NV03_PGRAPH_INTR);
+	u32 nsource = nv_rd32(priv, NV03_PGRAPH_NSOURCE);
+	u32 nstatus = nv_rd32(priv, NV03_PGRAPH_NSTATUS);
+	u32 addr = nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR);
+	u32 chid = (addr & 0x01f00000) >> 20;
+	u32 subc = (addr & 0x00070000) >> 16;
+	u32 mthd = (addr & 0x00001ffc);
+	u32 data = nv_rd32(priv, NV04_PGRAPH_TRAPPED_DATA);
+	u32 class = nv_rd32(priv, 0x400160 + subc * 4) & 0xfff;
+	u32 show = stat;
+	unsigned long flags;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	chan = priv->chan[chid];
+	if (chan)
+		namedb = (void *)nv_pclass(nv_object(chan), NV_NAMEDB_CLASS);
+	spin_unlock_irqrestore(&priv->lock, flags);
+
+	if (stat & NV_PGRAPH_INTR_ERROR) {
+		if (chan && (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD)) {
+			handle = nouveau_namedb_get_class(namedb, class);
+			if (handle && !nv_call(handle->object, mthd, data))
+				show &= ~NV_PGRAPH_INTR_ERROR;
+		}
+	}
+
+	if (stat & NV_PGRAPH_INTR_CONTEXT_SWITCH) {
+		nv_wr32(priv, NV03_PGRAPH_INTR, NV_PGRAPH_INTR_CONTEXT_SWITCH);
+		stat &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH;
+		show &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH;
+		nv10_graph_context_switch(priv);
+	}
+
+	nv_wr32(priv, NV03_PGRAPH_INTR, stat);
+	nv_wr32(priv, NV04_PGRAPH_FIFO, 0x00000001);
+
+	if (show) {
+		nv_error(priv, "");
+		nouveau_bitfield_print(nv10_graph_intr_name, show);
+		printk(" nsource:");
+		nouveau_bitfield_print(nv04_graph_nsource, nsource);
+		printk(" nstatus:");
+		nouveau_bitfield_print(nv10_graph_nstatus, nstatus);
+		printk("\n");
+		nv_error(priv, "ch %d/%d class 0x%04x "
+			       "mthd 0x%04x data 0x%08x\n",
+			 chid, subc, class, mthd, data);
+	}
+
+	nouveau_namedb_put(handle);
+}
+
+static int
+nv10_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		struct nouveau_oclass *oclass, void *data, u32 size,
+		struct nouveau_object **pobject)
+{
+	struct nv10_graph_priv *priv;
+	int ret;
+
+	ret = nouveau_graph_create(parent, engine, oclass, true, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00001000;
+	nv_subdev(priv)->intr = nv10_graph_intr;
+	nv_engine(priv)->cclass = &nv10_graph_cclass;
+
+	if (nv_device(priv)->chipset <= 0x10)
+		nv_engine(priv)->sclass = nv10_graph_sclass;
+	else
+	if (nv_device(priv)->chipset <  0x17 ||
+	    nv_device(priv)->chipset == 0x1a)
+		nv_engine(priv)->sclass = nv15_graph_sclass;
+	else
+		nv_engine(priv)->sclass = nv17_graph_sclass;
+
+	nv_engine(priv)->tile_prog = nv10_graph_tile_prog;
+	spin_lock_init(&priv->lock);
+	return 0;
+}
+
+static void
+nv10_graph_dtor(struct nouveau_object *object)
+{
+	struct nv10_graph_priv *priv = (void *)object;
+	nouveau_graph_destroy(&priv->base);
+}
+
+static int
+nv10_graph_init(struct nouveau_object *object)
+{
+	struct nouveau_engine *engine = nv_engine(object);
+	struct nouveau_fb *pfb = nouveau_fb(object);
+	struct nv10_graph_priv *priv = (void *)engine;
+	int ret, i;
+
+	ret = nouveau_graph_init(&priv->base);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, NV03_PGRAPH_INTR   , 0xFFFFFFFF);
+	nv_wr32(priv, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF);
+
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x00000000);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_1, 0x00118700);
+	/* nv_wr32(priv, NV04_PGRAPH_DEBUG_2, 0x24E00810); */ /* 0x25f92ad9 */
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_2, 0x25f92ad9);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0x55DE0830 | (1 << 29) | (1 << 31));
+
+	if (nv_device(priv)->chipset >= 0x17) {
+		nv_wr32(priv, NV10_PGRAPH_DEBUG_4, 0x1f000000);
+		nv_wr32(priv, 0x400a10, 0x03ff3fb6);
+		nv_wr32(priv, 0x400838, 0x002f8684);
+		nv_wr32(priv, 0x40083c, 0x00115f3f);
+		nv_wr32(priv, 0x4006b0, 0x40000020);
+	} else {
+		nv_wr32(priv, NV10_PGRAPH_DEBUG_4, 0x00000000);
+	}
+
+	/* Turn all the tiling regions off. */
+	for (i = 0; i < pfb->tile.regions; i++)
+		engine->tile_prog(engine, i);
+
+	nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(0), 0x00000000);
+	nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(1), 0x00000000);
+	nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(2), 0x00000000);
+	nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(3), 0x00000000);
+	nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(4), 0x00000000);
+	nv_wr32(priv, NV10_PGRAPH_STATE, 0xFFFFFFFF);
+
+	nv_mask(priv, NV10_PGRAPH_CTX_USER, 0xff000000, 0x1f000000);
+	nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10000100);
+	nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2, 0x08000000);
+	return 0;
+}
+
+static int
+nv10_graph_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv10_graph_priv *priv = (void *)object;
+	return nouveau_graph_fini(&priv->base, suspend);
+}
+
+struct nouveau_oclass
+nv10_graph_oclass = {
+	.handle = NV_ENGINE(GR, 0x10),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv10_graph_ctor,
+		.dtor = nv10_graph_dtor,
+		.init = nv10_graph_init,
+		.fini = nv10_graph_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv20.c b/drivers/gpu/drm/nouveau/core/engine/graph/nv20.c
new file mode 100644
index 0000000..8f3f619
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv20.c
@@ -0,0 +1,381 @@
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/handle.h>
+#include <core/enum.h>
+
+#include <subdev/timer.h>
+#include <subdev/fb.h>
+
+#include <engine/graph.h>
+#include <engine/fifo.h>
+
+#include "nv20.h"
+#include "regs.h"
+
+/*******************************************************************************
+ * Graphics object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv20_graph_sclass[] = {
+	{ 0x0012, &nv04_graph_ofuncs, NULL }, /* beta1 */
+	{ 0x0019, &nv04_graph_ofuncs, NULL }, /* clip */
+	{ 0x0030, &nv04_graph_ofuncs, NULL }, /* null */
+	{ 0x0039, &nv04_graph_ofuncs, NULL }, /* m2mf */
+	{ 0x0043, &nv04_graph_ofuncs, NULL }, /* rop */
+	{ 0x0044, &nv04_graph_ofuncs, NULL }, /* patt */
+	{ 0x004a, &nv04_graph_ofuncs, NULL }, /* gdi */
+	{ 0x0062, &nv04_graph_ofuncs, NULL }, /* surf2d */
+	{ 0x0072, &nv04_graph_ofuncs, NULL }, /* beta4 */
+	{ 0x0089, &nv04_graph_ofuncs, NULL }, /* sifm */
+	{ 0x008a, &nv04_graph_ofuncs, NULL }, /* ifc */
+	{ 0x0096, &nv04_graph_ofuncs, NULL }, /* celcius */
+	{ 0x0097, &nv04_graph_ofuncs, NULL }, /* kelvin */
+	{ 0x009e, &nv04_graph_ofuncs, NULL }, /* swzsurf */
+	{ 0x009f, &nv04_graph_ofuncs, NULL }, /* imageblit */
+	{},
+};
+
+/*******************************************************************************
+ * PGRAPH context
+ ******************************************************************************/
+
+static int
+nv20_graph_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nv20_graph_chan *chan;
+	int ret, i;
+
+	ret = nouveau_graph_context_create(parent, engine, oclass, NULL,
+					   0x37f0, 16, NVOBJ_FLAG_ZERO_ALLOC,
+					   &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	chan->chid = nouveau_fifo_chan(parent)->chid;
+
+	nv_wo32(chan, 0x0000, 0x00000001 | (chan->chid << 24));
+	nv_wo32(chan, 0x033c, 0xffff0000);
+	nv_wo32(chan, 0x03a0, 0x0fff0000);
+	nv_wo32(chan, 0x03a4, 0x0fff0000);
+	nv_wo32(chan, 0x047c, 0x00000101);
+	nv_wo32(chan, 0x0490, 0x00000111);
+	nv_wo32(chan, 0x04a8, 0x44400000);
+	for (i = 0x04d4; i <= 0x04e0; i += 4)
+		nv_wo32(chan, i, 0x00030303);
+	for (i = 0x04f4; i <= 0x0500; i += 4)
+		nv_wo32(chan, i, 0x00080000);
+	for (i = 0x050c; i <= 0x0518; i += 4)
+		nv_wo32(chan, i, 0x01012000);
+	for (i = 0x051c; i <= 0x0528; i += 4)
+		nv_wo32(chan, i, 0x000105b8);
+	for (i = 0x052c; i <= 0x0538; i += 4)
+		nv_wo32(chan, i, 0x00080008);
+	for (i = 0x055c; i <= 0x0598; i += 4)
+		nv_wo32(chan, i, 0x07ff0000);
+	nv_wo32(chan, 0x05a4, 0x4b7fffff);
+	nv_wo32(chan, 0x05fc, 0x00000001);
+	nv_wo32(chan, 0x0604, 0x00004000);
+	nv_wo32(chan, 0x0610, 0x00000001);
+	nv_wo32(chan, 0x0618, 0x00040000);
+	nv_wo32(chan, 0x061c, 0x00010000);
+	for (i = 0x1c1c; i <= 0x248c; i += 16) {
+		nv_wo32(chan, (i + 0), 0x10700ff9);
+		nv_wo32(chan, (i + 4), 0x0436086c);
+		nv_wo32(chan, (i + 8), 0x000c001b);
+	}
+	nv_wo32(chan, 0x281c, 0x3f800000);
+	nv_wo32(chan, 0x2830, 0x3f800000);
+	nv_wo32(chan, 0x285c, 0x40000000);
+	nv_wo32(chan, 0x2860, 0x3f800000);
+	nv_wo32(chan, 0x2864, 0x3f000000);
+	nv_wo32(chan, 0x286c, 0x40000000);
+	nv_wo32(chan, 0x2870, 0x3f800000);
+	nv_wo32(chan, 0x2878, 0xbf800000);
+	nv_wo32(chan, 0x2880, 0xbf800000);
+	nv_wo32(chan, 0x34a4, 0x000fe000);
+	nv_wo32(chan, 0x3530, 0x000003f8);
+	nv_wo32(chan, 0x3540, 0x002fe000);
+	for (i = 0x355c; i <= 0x3578; i += 4)
+		nv_wo32(chan, i, 0x001c527c);
+	return 0;
+}
+
+int
+nv20_graph_context_init(struct nouveau_object *object)
+{
+	struct nv20_graph_priv *priv = (void *)object->engine;
+	struct nv20_graph_chan *chan = (void *)object;
+	int ret;
+
+	ret = nouveau_graph_context_init(&chan->base);
+	if (ret)
+		return ret;
+
+	nv_wo32(priv->ctxtab, chan->chid * 4, nv_gpuobj(chan)->addr >> 4);
+	return 0;
+}
+
+int
+nv20_graph_context_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv20_graph_priv *priv = (void *)object->engine;
+	struct nv20_graph_chan *chan = (void *)object;
+	int chid = -1;
+
+	nv_mask(priv, 0x400720, 0x00000001, 0x00000000);
+	if (nv_rd32(priv, 0x400144) & 0x00010000)
+		chid = (nv_rd32(priv, 0x400148) & 0x1f000000) >> 24;
+	if (chan->chid == chid) {
+		nv_wr32(priv, 0x400784, nv_gpuobj(chan)->addr >> 4);
+		nv_wr32(priv, 0x400788, 0x00000002);
+		nv_wait(priv, 0x400700, 0xffffffff, 0x00000000);
+		nv_wr32(priv, 0x400144, 0x10000000);
+		nv_mask(priv, 0x400148, 0xff000000, 0x1f000000);
+	}
+	nv_mask(priv, 0x400720, 0x00000001, 0x00000001);
+
+	nv_wo32(priv->ctxtab, chan->chid * 4, 0x00000000);
+	return nouveau_graph_context_fini(&chan->base, suspend);
+}
+
+static struct nouveau_oclass
+nv20_graph_cclass = {
+	.handle = NV_ENGCTX(GR, 0x20),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv20_graph_context_ctor,
+		.dtor = _nouveau_graph_context_dtor,
+		.init = nv20_graph_context_init,
+		.fini = nv20_graph_context_fini,
+		.rd32 = _nouveau_graph_context_rd32,
+		.wr32 = _nouveau_graph_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PGRAPH engine/subdev functions
+ ******************************************************************************/
+
+void
+nv20_graph_tile_prog(struct nouveau_engine *engine, int i)
+{
+	struct nouveau_fb_tile *tile = &nouveau_fb(engine)->tile.region[i];
+	struct nouveau_fifo *pfifo = nouveau_fifo(engine);
+	struct nv20_graph_priv *priv = (void *)engine;
+	unsigned long flags;
+
+	pfifo->pause(pfifo, &flags);
+	nv04_graph_idle(priv);
+
+	nv_wr32(priv, NV20_PGRAPH_TLIMIT(i), tile->limit);
+	nv_wr32(priv, NV20_PGRAPH_TSIZE(i), tile->pitch);
+	nv_wr32(priv, NV20_PGRAPH_TILE(i), tile->addr);
+
+	nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0030 + 4 * i);
+	nv_wr32(priv, NV10_PGRAPH_RDI_DATA, tile->limit);
+	nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0050 + 4 * i);
+	nv_wr32(priv, NV10_PGRAPH_RDI_DATA, tile->pitch);
+	nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0010 + 4 * i);
+	nv_wr32(priv, NV10_PGRAPH_RDI_DATA, tile->addr);
+
+	if (nv_device(engine)->card_type == NV_20) {
+		nv_wr32(priv, NV20_PGRAPH_ZCOMP(i), tile->zcomp);
+		nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00ea0090 + 4 * i);
+		nv_wr32(priv, NV10_PGRAPH_RDI_DATA, tile->zcomp);
+	}
+
+	pfifo->start(pfifo, &flags);
+}
+
+void
+nv20_graph_intr(struct nouveau_subdev *subdev)
+{
+	struct nouveau_engine *engine = nv_engine(subdev);
+	struct nouveau_object *engctx;
+	struct nouveau_handle *handle;
+	struct nv20_graph_priv *priv = (void *)subdev;
+	u32 stat = nv_rd32(priv, NV03_PGRAPH_INTR);
+	u32 nsource = nv_rd32(priv, NV03_PGRAPH_NSOURCE);
+	u32 nstatus = nv_rd32(priv, NV03_PGRAPH_NSTATUS);
+	u32 addr = nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR);
+	u32 chid = (addr & 0x01f00000) >> 20;
+	u32 subc = (addr & 0x00070000) >> 16;
+	u32 mthd = (addr & 0x00001ffc);
+	u32 data = nv_rd32(priv, NV04_PGRAPH_TRAPPED_DATA);
+	u32 class = nv_rd32(priv, 0x400160 + subc * 4) & 0xfff;
+	u32 show = stat;
+
+	engctx = nouveau_engctx_get(engine, chid);
+	if (stat & NV_PGRAPH_INTR_ERROR) {
+		if (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD) {
+			handle = nouveau_handle_get_class(engctx, class);
+			if (handle && !nv_call(handle->object, mthd, data))
+				show &= ~NV_PGRAPH_INTR_ERROR;
+			nouveau_handle_put(handle);
+		}
+	}
+
+	nv_wr32(priv, NV03_PGRAPH_INTR, stat);
+	nv_wr32(priv, NV04_PGRAPH_FIFO, 0x00000001);
+
+	if (show) {
+		nv_info(priv, "");
+		nouveau_bitfield_print(nv10_graph_intr_name, show);
+		printk(" nsource:");
+		nouveau_bitfield_print(nv04_graph_nsource, nsource);
+		printk(" nstatus:");
+		nouveau_bitfield_print(nv10_graph_nstatus, nstatus);
+		printk("\n");
+		nv_info(priv, "ch %d/%d class 0x%04x mthd 0x%04x data 0x%08x\n",
+			chid, subc, class, mthd, data);
+	}
+
+	nouveau_engctx_put(engctx);
+}
+
+static int
+nv20_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv20_graph_priv *priv;
+	int ret;
+
+	ret = nouveau_graph_create(parent, engine, oclass, true, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 32 * 4, 16,
+				 NVOBJ_FLAG_ZERO_ALLOC, &priv->ctxtab);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00001000;
+	nv_subdev(priv)->intr = nv20_graph_intr;
+	nv_engine(priv)->cclass = &nv20_graph_cclass;
+	nv_engine(priv)->sclass = nv20_graph_sclass;
+	nv_engine(priv)->tile_prog = nv20_graph_tile_prog;
+	return 0;
+}
+
+void
+nv20_graph_dtor(struct nouveau_object *object)
+{
+	struct nv20_graph_priv *priv = (void *)object;
+	nouveau_gpuobj_ref(NULL, &priv->ctxtab);
+	nouveau_graph_destroy(&priv->base);
+}
+
+int
+nv20_graph_init(struct nouveau_object *object)
+{
+	struct nouveau_engine *engine = nv_engine(object);
+	struct nv20_graph_priv *priv = (void *)engine;
+	struct nouveau_fb *pfb = nouveau_fb(object);
+	u32 tmp, vramsz;
+	int ret, i;
+
+	ret = nouveau_graph_init(&priv->base);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, NV20_PGRAPH_CHANNEL_CTX_TABLE, priv->ctxtab->addr >> 4);
+
+	if (nv_device(priv)->chipset == 0x20) {
+		nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x003d0000);
+		for (i = 0; i < 15; i++)
+			nv_wr32(priv, NV10_PGRAPH_RDI_DATA, 0x00000000);
+		nv_wait(priv, 0x400700, 0xffffffff, 0x00000000);
+	} else {
+		nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x02c80000);
+		for (i = 0; i < 32; i++)
+			nv_wr32(priv, NV10_PGRAPH_RDI_DATA, 0x00000000);
+		nv_wait(priv, 0x400700, 0xffffffff, 0x00000000);
+	}
+
+	nv_wr32(priv, NV03_PGRAPH_INTR   , 0xFFFFFFFF);
+	nv_wr32(priv, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF);
+
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x00000000);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_1, 0x00118700);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0xF3CE0475); /* 0x4 = auto ctx switch */
+	nv_wr32(priv, NV10_PGRAPH_DEBUG_4, 0x00000000);
+	nv_wr32(priv, 0x40009C           , 0x00000040);
+
+	if (nv_device(priv)->chipset >= 0x25) {
+		nv_wr32(priv, 0x400890, 0x00a8cfff);
+		nv_wr32(priv, 0x400610, 0x304B1FB6);
+		nv_wr32(priv, 0x400B80, 0x1cbd3883);
+		nv_wr32(priv, 0x400B84, 0x44000000);
+		nv_wr32(priv, 0x400098, 0x40000080);
+		nv_wr32(priv, 0x400B88, 0x000000ff);
+
+	} else {
+		nv_wr32(priv, 0x400880, 0x0008c7df);
+		nv_wr32(priv, 0x400094, 0x00000005);
+		nv_wr32(priv, 0x400B80, 0x45eae20e);
+		nv_wr32(priv, 0x400B84, 0x24000000);
+		nv_wr32(priv, 0x400098, 0x00000040);
+		nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00E00038);
+		nv_wr32(priv, NV10_PGRAPH_RDI_DATA , 0x00000030);
+		nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00E10038);
+		nv_wr32(priv, NV10_PGRAPH_RDI_DATA , 0x00000030);
+	}
+
+	/* Turn all the tiling regions off. */
+	for (i = 0; i < pfb->tile.regions; i++)
+		engine->tile_prog(engine, i);
+
+	nv_wr32(priv, 0x4009a0, nv_rd32(priv, 0x100324));
+	nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA000C);
+	nv_wr32(priv, NV10_PGRAPH_RDI_DATA, nv_rd32(priv, 0x100324));
+
+	nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10000100);
+	nv_wr32(priv, NV10_PGRAPH_STATE      , 0xFFFFFFFF);
+
+	tmp = nv_rd32(priv, NV10_PGRAPH_SURFACE) & 0x0007ff00;
+	nv_wr32(priv, NV10_PGRAPH_SURFACE, tmp);
+	tmp = nv_rd32(priv, NV10_PGRAPH_SURFACE) | 0x00020100;
+	nv_wr32(priv, NV10_PGRAPH_SURFACE, tmp);
+
+	/* begin RAM config */
+	vramsz = pci_resource_len(nv_device(priv)->pdev, 0) - 1;
+	nv_wr32(priv, 0x4009A4, nv_rd32(priv, 0x100200));
+	nv_wr32(priv, 0x4009A8, nv_rd32(priv, 0x100204));
+	nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0000);
+	nv_wr32(priv, NV10_PGRAPH_RDI_DATA , nv_rd32(priv, 0x100200));
+	nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0004);
+	nv_wr32(priv, NV10_PGRAPH_RDI_DATA , nv_rd32(priv, 0x100204));
+	nv_wr32(priv, 0x400820, 0);
+	nv_wr32(priv, 0x400824, 0);
+	nv_wr32(priv, 0x400864, vramsz - 1);
+	nv_wr32(priv, 0x400868, vramsz - 1);
+
+	/* interesting.. the below overwrites some of the tile setup above.. */
+	nv_wr32(priv, 0x400B20, 0x00000000);
+	nv_wr32(priv, 0x400B04, 0xFFFFFFFF);
+
+	nv_wr32(priv, NV03_PGRAPH_ABS_UCLIP_XMIN, 0);
+	nv_wr32(priv, NV03_PGRAPH_ABS_UCLIP_YMIN, 0);
+	nv_wr32(priv, NV03_PGRAPH_ABS_UCLIP_XMAX, 0x7fff);
+	nv_wr32(priv, NV03_PGRAPH_ABS_UCLIP_YMAX, 0x7fff);
+	return 0;
+}
+
+struct nouveau_oclass
+nv20_graph_oclass = {
+	.handle = NV_ENGINE(GR, 0x20),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv20_graph_ctor,
+		.dtor = nv20_graph_dtor,
+		.init = nv20_graph_init,
+		.fini = _nouveau_graph_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv20.h b/drivers/gpu/drm/nouveau/core/engine/graph/nv20.h
new file mode 100644
index 0000000..2bea731
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv20.h
@@ -0,0 +1,31 @@
+#ifndef __NV20_GRAPH_H__
+#define __NV20_GRAPH_H__
+
+#include <core/enum.h>
+
+#include <engine/graph.h>
+#include <engine/fifo.h>
+
+struct nv20_graph_priv {
+	struct nouveau_graph base;
+	struct nouveau_gpuobj *ctxtab;
+};
+
+struct nv20_graph_chan {
+	struct nouveau_graph_chan base;
+	int chid;
+};
+
+extern struct nouveau_oclass nv25_graph_sclass[];
+int  nv20_graph_context_init(struct nouveau_object *);
+int  nv20_graph_context_fini(struct nouveau_object *, bool);
+
+void nv20_graph_tile_prog(struct nouveau_engine *, int);
+void nv20_graph_intr(struct nouveau_subdev *);
+
+void nv20_graph_dtor(struct nouveau_object *);
+int  nv20_graph_init(struct nouveau_object *);
+
+int  nv30_graph_init(struct nouveau_object *);
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv25.c b/drivers/gpu/drm/nouveau/core/engine/graph/nv25.c
new file mode 100644
index 0000000..b2b650d
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv25.c
@@ -0,0 +1,167 @@
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/enum.h>
+
+#include <subdev/timer.h>
+#include <subdev/fb.h>
+
+#include <engine/graph.h>
+
+#include "nv20.h"
+#include "regs.h"
+
+/*******************************************************************************
+ * Graphics object classes
+ ******************************************************************************/
+
+struct nouveau_oclass
+nv25_graph_sclass[] = {
+	{ 0x0012, &nv04_graph_ofuncs, NULL }, /* beta1 */
+	{ 0x0019, &nv04_graph_ofuncs, NULL }, /* clip */
+	{ 0x0030, &nv04_graph_ofuncs, NULL }, /* null */
+	{ 0x0039, &nv04_graph_ofuncs, NULL }, /* m2mf */
+	{ 0x0043, &nv04_graph_ofuncs, NULL }, /* rop */
+	{ 0x0044, &nv04_graph_ofuncs, NULL }, /* patt */
+	{ 0x004a, &nv04_graph_ofuncs, NULL }, /* gdi */
+	{ 0x0062, &nv04_graph_ofuncs, NULL }, /* surf2d */
+	{ 0x0072, &nv04_graph_ofuncs, NULL }, /* beta4 */
+	{ 0x0089, &nv04_graph_ofuncs, NULL }, /* sifm */
+	{ 0x008a, &nv04_graph_ofuncs, NULL }, /* ifc */
+	{ 0x0096, &nv04_graph_ofuncs, NULL }, /* celcius */
+	{ 0x009e, &nv04_graph_ofuncs, NULL }, /* swzsurf */
+	{ 0x009f, &nv04_graph_ofuncs, NULL }, /* imageblit */
+	{ 0x0597, &nv04_graph_ofuncs, NULL }, /* kelvin */
+	{},
+};
+
+/*******************************************************************************
+ * PGRAPH context
+ ******************************************************************************/
+
+static int
+nv25_graph_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nv20_graph_chan *chan;
+	int ret, i;
+
+	ret = nouveau_graph_context_create(parent, engine, oclass, NULL, 0x3724,
+					   16, NVOBJ_FLAG_ZERO_ALLOC, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	chan->chid = nouveau_fifo_chan(parent)->chid;
+
+	nv_wo32(chan, 0x0028, 0x00000001 | (chan->chid << 24));
+	nv_wo32(chan, 0x035c, 0xffff0000);
+	nv_wo32(chan, 0x03c0, 0x0fff0000);
+	nv_wo32(chan, 0x03c4, 0x0fff0000);
+	nv_wo32(chan, 0x049c, 0x00000101);
+	nv_wo32(chan, 0x04b0, 0x00000111);
+	nv_wo32(chan, 0x04c8, 0x00000080);
+	nv_wo32(chan, 0x04cc, 0xffff0000);
+	nv_wo32(chan, 0x04d0, 0x00000001);
+	nv_wo32(chan, 0x04e4, 0x44400000);
+	nv_wo32(chan, 0x04fc, 0x4b800000);
+	for (i = 0x0510; i <= 0x051c; i += 4)
+		nv_wo32(chan, i, 0x00030303);
+	for (i = 0x0530; i <= 0x053c; i += 4)
+		nv_wo32(chan, i, 0x00080000);
+	for (i = 0x0548; i <= 0x0554; i += 4)
+		nv_wo32(chan, i, 0x01012000);
+	for (i = 0x0558; i <= 0x0564; i += 4)
+		nv_wo32(chan, i, 0x000105b8);
+	for (i = 0x0568; i <= 0x0574; i += 4)
+		nv_wo32(chan, i, 0x00080008);
+	for (i = 0x0598; i <= 0x05d4; i += 4)
+		nv_wo32(chan, i, 0x07ff0000);
+	nv_wo32(chan, 0x05e0, 0x4b7fffff);
+	nv_wo32(chan, 0x0620, 0x00000080);
+	nv_wo32(chan, 0x0624, 0x30201000);
+	nv_wo32(chan, 0x0628, 0x70605040);
+	nv_wo32(chan, 0x062c, 0xb0a09080);
+	nv_wo32(chan, 0x0630, 0xf0e0d0c0);
+	nv_wo32(chan, 0x0664, 0x00000001);
+	nv_wo32(chan, 0x066c, 0x00004000);
+	nv_wo32(chan, 0x0678, 0x00000001);
+	nv_wo32(chan, 0x0680, 0x00040000);
+	nv_wo32(chan, 0x0684, 0x00010000);
+	for (i = 0x1b04; i <= 0x2374; i += 16) {
+		nv_wo32(chan, (i + 0), 0x10700ff9);
+		nv_wo32(chan, (i + 4), 0x0436086c);
+		nv_wo32(chan, (i + 8), 0x000c001b);
+	}
+	nv_wo32(chan, 0x2704, 0x3f800000);
+	nv_wo32(chan, 0x2718, 0x3f800000);
+	nv_wo32(chan, 0x2744, 0x40000000);
+	nv_wo32(chan, 0x2748, 0x3f800000);
+	nv_wo32(chan, 0x274c, 0x3f000000);
+	nv_wo32(chan, 0x2754, 0x40000000);
+	nv_wo32(chan, 0x2758, 0x3f800000);
+	nv_wo32(chan, 0x2760, 0xbf800000);
+	nv_wo32(chan, 0x2768, 0xbf800000);
+	nv_wo32(chan, 0x308c, 0x000fe000);
+	nv_wo32(chan, 0x3108, 0x000003f8);
+	nv_wo32(chan, 0x3468, 0x002fe000);
+	for (i = 0x3484; i <= 0x34a0; i += 4)
+		nv_wo32(chan, i, 0x001c527c);
+	return 0;
+}
+
+static struct nouveau_oclass
+nv25_graph_cclass = {
+	.handle = NV_ENGCTX(GR, 0x25),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv25_graph_context_ctor,
+		.dtor = _nouveau_graph_context_dtor,
+		.init = nv20_graph_context_init,
+		.fini = nv20_graph_context_fini,
+		.rd32 = _nouveau_graph_context_rd32,
+		.wr32 = _nouveau_graph_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PGRAPH engine/subdev functions
+ ******************************************************************************/
+
+static int
+nv25_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv20_graph_priv *priv;
+	int ret;
+
+	ret = nouveau_graph_create(parent, engine, oclass, true, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 32 * 4, 16,
+				 NVOBJ_FLAG_ZERO_ALLOC, &priv->ctxtab);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00001000;
+	nv_subdev(priv)->intr = nv20_graph_intr;
+	nv_engine(priv)->cclass = &nv25_graph_cclass;
+	nv_engine(priv)->sclass = nv25_graph_sclass;
+	nv_engine(priv)->tile_prog = nv20_graph_tile_prog;
+	return 0;
+}
+
+struct nouveau_oclass
+nv25_graph_oclass = {
+	.handle = NV_ENGINE(GR, 0x25),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv25_graph_ctor,
+		.dtor = nv20_graph_dtor,
+		.init = nv20_graph_init,
+		.fini = _nouveau_graph_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv2a.c b/drivers/gpu/drm/nouveau/core/engine/graph/nv2a.c
new file mode 100644
index 0000000..700462f
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv2a.c
@@ -0,0 +1,134 @@
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/enum.h>
+
+#include <subdev/timer.h>
+#include <subdev/fb.h>
+
+#include <engine/graph.h>
+
+#include "nv20.h"
+#include "regs.h"
+
+/*******************************************************************************
+ * PGRAPH context
+ ******************************************************************************/
+
+static int
+nv2a_graph_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nv20_graph_chan *chan;
+	int ret, i;
+
+	ret = nouveau_graph_context_create(parent, engine, oclass, NULL, 0x36b0,
+					   16, NVOBJ_FLAG_ZERO_ALLOC, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	chan->chid = nouveau_fifo_chan(parent)->chid;
+
+	nv_wo32(chan, 0x0000, 0x00000001 | (chan->chid << 24));
+	nv_wo32(chan, 0x033c, 0xffff0000);
+	nv_wo32(chan, 0x03a0, 0x0fff0000);
+	nv_wo32(chan, 0x03a4, 0x0fff0000);
+	nv_wo32(chan, 0x047c, 0x00000101);
+	nv_wo32(chan, 0x0490, 0x00000111);
+	nv_wo32(chan, 0x04a8, 0x44400000);
+	for (i = 0x04d4; i <= 0x04e0; i += 4)
+		nv_wo32(chan, i, 0x00030303);
+	for (i = 0x04f4; i <= 0x0500; i += 4)
+		nv_wo32(chan, i, 0x00080000);
+	for (i = 0x050c; i <= 0x0518; i += 4)
+		nv_wo32(chan, i, 0x01012000);
+	for (i = 0x051c; i <= 0x0528; i += 4)
+		nv_wo32(chan, i, 0x000105b8);
+	for (i = 0x052c; i <= 0x0538; i += 4)
+		nv_wo32(chan, i, 0x00080008);
+	for (i = 0x055c; i <= 0x0598; i += 4)
+		nv_wo32(chan, i, 0x07ff0000);
+	nv_wo32(chan, 0x05a4, 0x4b7fffff);
+	nv_wo32(chan, 0x05fc, 0x00000001);
+	nv_wo32(chan, 0x0604, 0x00004000);
+	nv_wo32(chan, 0x0610, 0x00000001);
+	nv_wo32(chan, 0x0618, 0x00040000);
+	nv_wo32(chan, 0x061c, 0x00010000);
+	for (i = 0x1a9c; i <= 0x22fc; i += 16) { /*XXX: check!! */
+		nv_wo32(chan, (i + 0), 0x10700ff9);
+		nv_wo32(chan, (i + 4), 0x0436086c);
+		nv_wo32(chan, (i + 8), 0x000c001b);
+	}
+	nv_wo32(chan, 0x269c, 0x3f800000);
+	nv_wo32(chan, 0x26b0, 0x3f800000);
+	nv_wo32(chan, 0x26dc, 0x40000000);
+	nv_wo32(chan, 0x26e0, 0x3f800000);
+	nv_wo32(chan, 0x26e4, 0x3f000000);
+	nv_wo32(chan, 0x26ec, 0x40000000);
+	nv_wo32(chan, 0x26f0, 0x3f800000);
+	nv_wo32(chan, 0x26f8, 0xbf800000);
+	nv_wo32(chan, 0x2700, 0xbf800000);
+	nv_wo32(chan, 0x3024, 0x000fe000);
+	nv_wo32(chan, 0x30a0, 0x000003f8);
+	nv_wo32(chan, 0x33fc, 0x002fe000);
+	for (i = 0x341c; i <= 0x3438; i += 4)
+		nv_wo32(chan, i, 0x001c527c);
+	return 0;
+}
+
+static struct nouveau_oclass
+nv2a_graph_cclass = {
+	.handle = NV_ENGCTX(GR, 0x2a),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv2a_graph_context_ctor,
+		.dtor = _nouveau_graph_context_dtor,
+		.init = nv20_graph_context_init,
+		.fini = nv20_graph_context_fini,
+		.rd32 = _nouveau_graph_context_rd32,
+		.wr32 = _nouveau_graph_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PGRAPH engine/subdev functions
+ ******************************************************************************/
+
+static int
+nv2a_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv20_graph_priv *priv;
+	int ret;
+
+	ret = nouveau_graph_create(parent, engine, oclass, true, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 32 * 4, 16,
+				 NVOBJ_FLAG_ZERO_ALLOC, &priv->ctxtab);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00001000;
+	nv_subdev(priv)->intr = nv20_graph_intr;
+	nv_engine(priv)->cclass = &nv2a_graph_cclass;
+	nv_engine(priv)->sclass = nv25_graph_sclass;
+	nv_engine(priv)->tile_prog = nv20_graph_tile_prog;
+	return 0;
+}
+
+struct nouveau_oclass
+nv2a_graph_oclass = {
+	.handle = NV_ENGINE(GR, 0x2a),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv2a_graph_ctor,
+		.dtor = nv20_graph_dtor,
+		.init = nv20_graph_init,
+		.fini = _nouveau_graph_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv30.c b/drivers/gpu/drm/nouveau/core/engine/graph/nv30.c
new file mode 100644
index 0000000..cedadaa
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv30.c
@@ -0,0 +1,238 @@
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/enum.h>
+
+#include <subdev/timer.h>
+#include <subdev/fb.h>
+
+#include <engine/graph.h>
+
+#include "nv20.h"
+#include "regs.h"
+
+/*******************************************************************************
+ * Graphics object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv30_graph_sclass[] = {
+	{ 0x0012, &nv04_graph_ofuncs, NULL }, /* beta1 */
+	{ 0x0019, &nv04_graph_ofuncs, NULL }, /* clip */
+	{ 0x0030, &nv04_graph_ofuncs, NULL }, /* null */
+	{ 0x0039, &nv04_graph_ofuncs, NULL }, /* m2mf */
+	{ 0x0043, &nv04_graph_ofuncs, NULL }, /* rop */
+	{ 0x0044, &nv04_graph_ofuncs, NULL }, /* patt */
+	{ 0x004a, &nv04_graph_ofuncs, NULL }, /* gdi */
+	{ 0x0062, &nv04_graph_ofuncs, NULL }, /* surf2d */
+	{ 0x0072, &nv04_graph_ofuncs, NULL }, /* beta4 */
+	{ 0x0089, &nv04_graph_ofuncs, NULL }, /* sifm */
+	{ 0x008a, &nv04_graph_ofuncs, NULL }, /* ifc */
+	{ 0x009f, &nv04_graph_ofuncs, NULL }, /* imageblit */
+	{ 0x0362, &nv04_graph_ofuncs, NULL }, /* surf2d (nv30) */
+	{ 0x0389, &nv04_graph_ofuncs, NULL }, /* sifm (nv30) */
+	{ 0x038a, &nv04_graph_ofuncs, NULL }, /* ifc (nv30) */
+	{ 0x039e, &nv04_graph_ofuncs, NULL }, /* swzsurf (nv30) */
+	{ 0x0397, &nv04_graph_ofuncs, NULL }, /* rankine */
+	{},
+};
+
+/*******************************************************************************
+ * PGRAPH context
+ ******************************************************************************/
+
+static int
+nv30_graph_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nv20_graph_chan *chan;
+	int ret, i;
+
+	ret = nouveau_graph_context_create(parent, engine, oclass, NULL, 0x5f48,
+					   16, NVOBJ_FLAG_ZERO_ALLOC, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	chan->chid = nouveau_fifo_chan(parent)->chid;
+
+	nv_wo32(chan, 0x0028, 0x00000001 | (chan->chid << 24));
+	nv_wo32(chan, 0x0410, 0x00000101);
+	nv_wo32(chan, 0x0424, 0x00000111);
+	nv_wo32(chan, 0x0428, 0x00000060);
+	nv_wo32(chan, 0x0444, 0x00000080);
+	nv_wo32(chan, 0x0448, 0xffff0000);
+	nv_wo32(chan, 0x044c, 0x00000001);
+	nv_wo32(chan, 0x0460, 0x44400000);
+	nv_wo32(chan, 0x048c, 0xffff0000);
+	for (i = 0x04e0; i < 0x04e8; i += 4)
+		nv_wo32(chan, i, 0x0fff0000);
+	nv_wo32(chan, 0x04ec, 0x00011100);
+	for (i = 0x0508; i < 0x0548; i += 4)
+		nv_wo32(chan, i, 0x07ff0000);
+	nv_wo32(chan, 0x0550, 0x4b7fffff);
+	nv_wo32(chan, 0x058c, 0x00000080);
+	nv_wo32(chan, 0x0590, 0x30201000);
+	nv_wo32(chan, 0x0594, 0x70605040);
+	nv_wo32(chan, 0x0598, 0xb8a89888);
+	nv_wo32(chan, 0x059c, 0xf8e8d8c8);
+	nv_wo32(chan, 0x05b0, 0xb0000000);
+	for (i = 0x0600; i < 0x0640; i += 4)
+		nv_wo32(chan, i, 0x00010588);
+	for (i = 0x0640; i < 0x0680; i += 4)
+		nv_wo32(chan, i, 0x00030303);
+	for (i = 0x06c0; i < 0x0700; i += 4)
+		nv_wo32(chan, i, 0x0008aae4);
+	for (i = 0x0700; i < 0x0740; i += 4)
+		nv_wo32(chan, i, 0x01012000);
+	for (i = 0x0740; i < 0x0780; i += 4)
+		nv_wo32(chan, i, 0x00080008);
+	nv_wo32(chan, 0x085c, 0x00040000);
+	nv_wo32(chan, 0x0860, 0x00010000);
+	for (i = 0x0864; i < 0x0874; i += 4)
+		nv_wo32(chan, i, 0x00040004);
+	for (i = 0x1f18; i <= 0x3088 ; i += 16) {
+		nv_wo32(chan, i + 0, 0x10700ff9);
+		nv_wo32(chan, i + 1, 0x0436086c);
+		nv_wo32(chan, i + 2, 0x000c001b);
+	}
+	for (i = 0x30b8; i < 0x30c8; i += 4)
+		nv_wo32(chan, i, 0x0000ffff);
+	nv_wo32(chan, 0x344c, 0x3f800000);
+	nv_wo32(chan, 0x3808, 0x3f800000);
+	nv_wo32(chan, 0x381c, 0x3f800000);
+	nv_wo32(chan, 0x3848, 0x40000000);
+	nv_wo32(chan, 0x384c, 0x3f800000);
+	nv_wo32(chan, 0x3850, 0x3f000000);
+	nv_wo32(chan, 0x3858, 0x40000000);
+	nv_wo32(chan, 0x385c, 0x3f800000);
+	nv_wo32(chan, 0x3864, 0xbf800000);
+	nv_wo32(chan, 0x386c, 0xbf800000);
+	return 0;
+}
+
+static struct nouveau_oclass
+nv30_graph_cclass = {
+	.handle = NV_ENGCTX(GR, 0x30),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv30_graph_context_ctor,
+		.dtor = _nouveau_graph_context_dtor,
+		.init = nv20_graph_context_init,
+		.fini = nv20_graph_context_fini,
+		.rd32 = _nouveau_graph_context_rd32,
+		.wr32 = _nouveau_graph_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PGRAPH engine/subdev functions
+ ******************************************************************************/
+
+static int
+nv30_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv20_graph_priv *priv;
+	int ret;
+
+	ret = nouveau_graph_create(parent, engine, oclass, true, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 32 * 4, 16,
+				 NVOBJ_FLAG_ZERO_ALLOC, &priv->ctxtab);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00001000;
+	nv_subdev(priv)->intr = nv20_graph_intr;
+	nv_engine(priv)->cclass = &nv30_graph_cclass;
+	nv_engine(priv)->sclass = nv30_graph_sclass;
+	nv_engine(priv)->tile_prog = nv20_graph_tile_prog;
+	return 0;
+}
+
+int
+nv30_graph_init(struct nouveau_object *object)
+{
+	struct nouveau_engine *engine = nv_engine(object);
+	struct nv20_graph_priv *priv = (void *)engine;
+	struct nouveau_fb *pfb = nouveau_fb(object);
+	int ret, i;
+
+	ret = nouveau_graph_init(&priv->base);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, NV20_PGRAPH_CHANNEL_CTX_TABLE, priv->ctxtab->addr >> 4);
+
+	nv_wr32(priv, NV03_PGRAPH_INTR   , 0xFFFFFFFF);
+	nv_wr32(priv, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF);
+
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x00000000);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_1, 0x401287c0);
+	nv_wr32(priv, 0x400890, 0x01b463ff);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0xf2de0475);
+	nv_wr32(priv, NV10_PGRAPH_DEBUG_4, 0x00008000);
+	nv_wr32(priv, NV04_PGRAPH_LIMIT_VIOL_PIX, 0xf04bdff6);
+	nv_wr32(priv, 0x400B80, 0x1003d888);
+	nv_wr32(priv, 0x400B84, 0x0c000000);
+	nv_wr32(priv, 0x400098, 0x00000000);
+	nv_wr32(priv, 0x40009C, 0x0005ad00);
+	nv_wr32(priv, 0x400B88, 0x62ff00ff); /* suspiciously like PGRAPH_DEBUG_2 */
+	nv_wr32(priv, 0x4000a0, 0x00000000);
+	nv_wr32(priv, 0x4000a4, 0x00000008);
+	nv_wr32(priv, 0x4008a8, 0xb784a400);
+	nv_wr32(priv, 0x400ba0, 0x002f8685);
+	nv_wr32(priv, 0x400ba4, 0x00231f3f);
+	nv_wr32(priv, 0x4008a4, 0x40000020);
+
+	if (nv_device(priv)->chipset == 0x34) {
+		nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0004);
+		nv_wr32(priv, NV10_PGRAPH_RDI_DATA , 0x00200201);
+		nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0008);
+		nv_wr32(priv, NV10_PGRAPH_RDI_DATA , 0x00000008);
+		nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0000);
+		nv_wr32(priv, NV10_PGRAPH_RDI_DATA , 0x00000032);
+		nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00E00004);
+		nv_wr32(priv, NV10_PGRAPH_RDI_DATA , 0x00000002);
+	}
+
+	nv_wr32(priv, 0x4000c0, 0x00000016);
+
+	/* Turn all the tiling regions off. */
+	for (i = 0; i < pfb->tile.regions; i++)
+		engine->tile_prog(engine, i);
+
+	nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10000100);
+	nv_wr32(priv, NV10_PGRAPH_STATE      , 0xFFFFFFFF);
+	nv_wr32(priv, 0x0040075c             , 0x00000001);
+
+	/* begin RAM config */
+	/* vramsz = pci_resource_len(priv->dev->pdev, 0) - 1; */
+	nv_wr32(priv, 0x4009A4, nv_rd32(priv, 0x100200));
+	nv_wr32(priv, 0x4009A8, nv_rd32(priv, 0x100204));
+	if (nv_device(priv)->chipset != 0x34) {
+		nv_wr32(priv, 0x400750, 0x00EA0000);
+		nv_wr32(priv, 0x400754, nv_rd32(priv, 0x100200));
+		nv_wr32(priv, 0x400750, 0x00EA0004);
+		nv_wr32(priv, 0x400754, nv_rd32(priv, 0x100204));
+	}
+	return 0;
+}
+
+struct nouveau_oclass
+nv30_graph_oclass = {
+	.handle = NV_ENGINE(GR, 0x30),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv30_graph_ctor,
+		.dtor = nv20_graph_dtor,
+		.init = nv30_graph_init,
+		.fini = _nouveau_graph_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv34.c b/drivers/gpu/drm/nouveau/core/engine/graph/nv34.c
new file mode 100644
index 0000000..273f632
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv34.c
@@ -0,0 +1,168 @@
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/enum.h>
+
+#include <subdev/timer.h>
+#include <subdev/fb.h>
+
+#include <engine/graph.h>
+
+#include "nv20.h"
+#include "regs.h"
+
+/*******************************************************************************
+ * Graphics object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv34_graph_sclass[] = {
+	{ 0x0012, &nv04_graph_ofuncs, NULL }, /* beta1 */
+	{ 0x0019, &nv04_graph_ofuncs, NULL }, /* clip */
+	{ 0x0030, &nv04_graph_ofuncs, NULL }, /* null */
+	{ 0x0039, &nv04_graph_ofuncs, NULL }, /* m2mf */
+	{ 0x0043, &nv04_graph_ofuncs, NULL }, /* rop */
+	{ 0x0044, &nv04_graph_ofuncs, NULL }, /* patt */
+	{ 0x004a, &nv04_graph_ofuncs, NULL }, /* gdi */
+	{ 0x0062, &nv04_graph_ofuncs, NULL }, /* surf2d */
+	{ 0x0072, &nv04_graph_ofuncs, NULL }, /* beta4 */
+	{ 0x0089, &nv04_graph_ofuncs, NULL }, /* sifm */
+	{ 0x008a, &nv04_graph_ofuncs, NULL }, /* ifc */
+	{ 0x009f, &nv04_graph_ofuncs, NULL }, /* imageblit */
+	{ 0x0362, &nv04_graph_ofuncs, NULL }, /* surf2d (nv30) */
+	{ 0x0389, &nv04_graph_ofuncs, NULL }, /* sifm (nv30) */
+	{ 0x038a, &nv04_graph_ofuncs, NULL }, /* ifc (nv30) */
+	{ 0x039e, &nv04_graph_ofuncs, NULL }, /* swzsurf (nv30) */
+	{ 0x0697, &nv04_graph_ofuncs, NULL }, /* rankine */
+	{},
+};
+
+/*******************************************************************************
+ * PGRAPH context
+ ******************************************************************************/
+
+static int
+nv34_graph_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nv20_graph_chan *chan;
+	int ret, i;
+
+	ret = nouveau_graph_context_create(parent, engine, oclass, NULL, 0x46dc,
+					   16, NVOBJ_FLAG_ZERO_ALLOC, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	chan->chid = nouveau_fifo_chan(parent)->chid;
+
+	nv_wo32(chan, 0x0028, 0x00000001 | (chan->chid << 24));
+	nv_wo32(chan, 0x040c, 0x01000101);
+	nv_wo32(chan, 0x0420, 0x00000111);
+	nv_wo32(chan, 0x0424, 0x00000060);
+	nv_wo32(chan, 0x0440, 0x00000080);
+	nv_wo32(chan, 0x0444, 0xffff0000);
+	nv_wo32(chan, 0x0448, 0x00000001);
+	nv_wo32(chan, 0x045c, 0x44400000);
+	nv_wo32(chan, 0x0480, 0xffff0000);
+	for (i = 0x04d4; i < 0x04dc; i += 4)
+		nv_wo32(chan, i, 0x0fff0000);
+	nv_wo32(chan, 0x04e0, 0x00011100);
+	for (i = 0x04fc; i < 0x053c; i += 4)
+		nv_wo32(chan, i, 0x07ff0000);
+	nv_wo32(chan, 0x0544, 0x4b7fffff);
+	nv_wo32(chan, 0x057c, 0x00000080);
+	nv_wo32(chan, 0x0580, 0x30201000);
+	nv_wo32(chan, 0x0584, 0x70605040);
+	nv_wo32(chan, 0x0588, 0xb8a89888);
+	nv_wo32(chan, 0x058c, 0xf8e8d8c8);
+	nv_wo32(chan, 0x05a0, 0xb0000000);
+	for (i = 0x05f0; i < 0x0630; i += 4)
+		nv_wo32(chan, i, 0x00010588);
+	for (i = 0x0630; i < 0x0670; i += 4)
+		nv_wo32(chan, i, 0x00030303);
+	for (i = 0x06b0; i < 0x06f0; i += 4)
+		nv_wo32(chan, i, 0x0008aae4);
+	for (i = 0x06f0; i < 0x0730; i += 4)
+		nv_wo32(chan, i, 0x01012000);
+	for (i = 0x0730; i < 0x0770; i += 4)
+		nv_wo32(chan, i, 0x00080008);
+	nv_wo32(chan, 0x0850, 0x00040000);
+	nv_wo32(chan, 0x0854, 0x00010000);
+	for (i = 0x0858; i < 0x0868; i += 4)
+		nv_wo32(chan, i, 0x00040004);
+	for (i = 0x15ac; i <= 0x271c ; i += 16) {
+		nv_wo32(chan, i + 0, 0x10700ff9);
+		nv_wo32(chan, i + 1, 0x0436086c);
+		nv_wo32(chan, i + 2, 0x000c001b);
+	}
+	for (i = 0x274c; i < 0x275c; i += 4)
+		nv_wo32(chan, i, 0x0000ffff);
+	nv_wo32(chan, 0x2ae0, 0x3f800000);
+	nv_wo32(chan, 0x2e9c, 0x3f800000);
+	nv_wo32(chan, 0x2eb0, 0x3f800000);
+	nv_wo32(chan, 0x2edc, 0x40000000);
+	nv_wo32(chan, 0x2ee0, 0x3f800000);
+	nv_wo32(chan, 0x2ee4, 0x3f000000);
+	nv_wo32(chan, 0x2eec, 0x40000000);
+	nv_wo32(chan, 0x2ef0, 0x3f800000);
+	nv_wo32(chan, 0x2ef8, 0xbf800000);
+	nv_wo32(chan, 0x2f00, 0xbf800000);
+	return 0;
+}
+
+static struct nouveau_oclass
+nv34_graph_cclass = {
+	.handle = NV_ENGCTX(GR, 0x34),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv34_graph_context_ctor,
+		.dtor = _nouveau_graph_context_dtor,
+		.init = nv20_graph_context_init,
+		.fini = nv20_graph_context_fini,
+		.rd32 = _nouveau_graph_context_rd32,
+		.wr32 = _nouveau_graph_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PGRAPH engine/subdev functions
+ ******************************************************************************/
+
+static int
+nv34_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv20_graph_priv *priv;
+	int ret;
+
+	ret = nouveau_graph_create(parent, engine, oclass, true, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 32 * 4, 16,
+				 NVOBJ_FLAG_ZERO_ALLOC, &priv->ctxtab);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00001000;
+	nv_subdev(priv)->intr = nv20_graph_intr;
+	nv_engine(priv)->cclass = &nv34_graph_cclass;
+	nv_engine(priv)->sclass = nv34_graph_sclass;
+	nv_engine(priv)->tile_prog = nv20_graph_tile_prog;
+	return 0;
+}
+
+struct nouveau_oclass
+nv34_graph_oclass = {
+	.handle = NV_ENGINE(GR, 0x34),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv34_graph_ctor,
+		.dtor = nv20_graph_dtor,
+		.init = nv30_graph_init,
+		.fini = _nouveau_graph_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv35.c b/drivers/gpu/drm/nouveau/core/engine/graph/nv35.c
new file mode 100644
index 0000000..f40ee21
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv35.c
@@ -0,0 +1,166 @@
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/enum.h>
+
+#include <subdev/timer.h>
+#include <subdev/fb.h>
+
+#include "nv20.h"
+#include "regs.h"
+
+/*******************************************************************************
+ * Graphics object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv35_graph_sclass[] = {
+	{ 0x0012, &nv04_graph_ofuncs, NULL }, /* beta1 */
+	{ 0x0019, &nv04_graph_ofuncs, NULL }, /* clip */
+	{ 0x0030, &nv04_graph_ofuncs, NULL }, /* null */
+	{ 0x0039, &nv04_graph_ofuncs, NULL }, /* m2mf */
+	{ 0x0043, &nv04_graph_ofuncs, NULL }, /* rop */
+	{ 0x0044, &nv04_graph_ofuncs, NULL }, /* patt */
+	{ 0x004a, &nv04_graph_ofuncs, NULL }, /* gdi */
+	{ 0x0062, &nv04_graph_ofuncs, NULL }, /* surf2d */
+	{ 0x0072, &nv04_graph_ofuncs, NULL }, /* beta4 */
+	{ 0x0089, &nv04_graph_ofuncs, NULL }, /* sifm */
+	{ 0x008a, &nv04_graph_ofuncs, NULL }, /* ifc */
+	{ 0x009f, &nv04_graph_ofuncs, NULL }, /* imageblit */
+	{ 0x0362, &nv04_graph_ofuncs, NULL }, /* surf2d (nv30) */
+	{ 0x0389, &nv04_graph_ofuncs, NULL }, /* sifm (nv30) */
+	{ 0x038a, &nv04_graph_ofuncs, NULL }, /* ifc (nv30) */
+	{ 0x039e, &nv04_graph_ofuncs, NULL }, /* swzsurf (nv30) */
+	{ 0x0497, &nv04_graph_ofuncs, NULL }, /* rankine */
+	{},
+};
+
+/*******************************************************************************
+ * PGRAPH context
+ ******************************************************************************/
+
+static int
+nv35_graph_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nv20_graph_chan *chan;
+	int ret, i;
+
+	ret = nouveau_graph_context_create(parent, engine, oclass, NULL, 0x577c,
+					   16, NVOBJ_FLAG_ZERO_ALLOC, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	chan->chid = nouveau_fifo_chan(parent)->chid;
+
+	nv_wo32(chan, 0x0028, 0x00000001 | (chan->chid << 24));
+	nv_wo32(chan, 0x040c, 0x00000101);
+	nv_wo32(chan, 0x0420, 0x00000111);
+	nv_wo32(chan, 0x0424, 0x00000060);
+	nv_wo32(chan, 0x0440, 0x00000080);
+	nv_wo32(chan, 0x0444, 0xffff0000);
+	nv_wo32(chan, 0x0448, 0x00000001);
+	nv_wo32(chan, 0x045c, 0x44400000);
+	nv_wo32(chan, 0x0488, 0xffff0000);
+	for (i = 0x04dc; i < 0x04e4; i += 4)
+		nv_wo32(chan, i, 0x0fff0000);
+	nv_wo32(chan, 0x04e8, 0x00011100);
+	for (i = 0x0504; i < 0x0544; i += 4)
+		nv_wo32(chan, i, 0x07ff0000);
+	nv_wo32(chan, 0x054c, 0x4b7fffff);
+	nv_wo32(chan, 0x0588, 0x00000080);
+	nv_wo32(chan, 0x058c, 0x30201000);
+	nv_wo32(chan, 0x0590, 0x70605040);
+	nv_wo32(chan, 0x0594, 0xb8a89888);
+	nv_wo32(chan, 0x0598, 0xf8e8d8c8);
+	nv_wo32(chan, 0x05ac, 0xb0000000);
+	for (i = 0x0604; i < 0x0644; i += 4)
+		nv_wo32(chan, i, 0x00010588);
+	for (i = 0x0644; i < 0x0684; i += 4)
+		nv_wo32(chan, i, 0x00030303);
+	for (i = 0x06c4; i < 0x0704; i += 4)
+		nv_wo32(chan, i, 0x0008aae4);
+	for (i = 0x0704; i < 0x0744; i += 4)
+		nv_wo32(chan, i, 0x01012000);
+	for (i = 0x0744; i < 0x0784; i += 4)
+		nv_wo32(chan, i, 0x00080008);
+	nv_wo32(chan, 0x0860, 0x00040000);
+	nv_wo32(chan, 0x0864, 0x00010000);
+	for (i = 0x0868; i < 0x0878; i += 4)
+		nv_wo32(chan, i, 0x00040004);
+	for (i = 0x1f1c; i <= 0x308c ; i += 16) {
+		nv_wo32(chan, i + 0, 0x10700ff9);
+		nv_wo32(chan, i + 4, 0x0436086c);
+		nv_wo32(chan, i + 8, 0x000c001b);
+	}
+	for (i = 0x30bc; i < 0x30cc; i += 4)
+		nv_wo32(chan, i, 0x0000ffff);
+	nv_wo32(chan, 0x3450, 0x3f800000);
+	nv_wo32(chan, 0x380c, 0x3f800000);
+	nv_wo32(chan, 0x3820, 0x3f800000);
+	nv_wo32(chan, 0x384c, 0x40000000);
+	nv_wo32(chan, 0x3850, 0x3f800000);
+	nv_wo32(chan, 0x3854, 0x3f000000);
+	nv_wo32(chan, 0x385c, 0x40000000);
+	nv_wo32(chan, 0x3860, 0x3f800000);
+	nv_wo32(chan, 0x3868, 0xbf800000);
+	nv_wo32(chan, 0x3870, 0xbf800000);
+	return 0;
+}
+
+static struct nouveau_oclass
+nv35_graph_cclass = {
+	.handle = NV_ENGCTX(GR, 0x35),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv35_graph_context_ctor,
+		.dtor = _nouveau_graph_context_dtor,
+		.init = nv20_graph_context_init,
+		.fini = nv20_graph_context_fini,
+		.rd32 = _nouveau_graph_context_rd32,
+		.wr32 = _nouveau_graph_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PGRAPH engine/subdev functions
+ ******************************************************************************/
+
+static int
+nv35_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv20_graph_priv *priv;
+	int ret;
+
+	ret = nouveau_graph_create(parent, engine, oclass, true, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 32 * 4, 16,
+				 NVOBJ_FLAG_ZERO_ALLOC, &priv->ctxtab);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00001000;
+	nv_subdev(priv)->intr = nv20_graph_intr;
+	nv_engine(priv)->cclass = &nv35_graph_cclass;
+	nv_engine(priv)->sclass = nv35_graph_sclass;
+	nv_engine(priv)->tile_prog = nv20_graph_tile_prog;
+	return 0;
+}
+
+struct nouveau_oclass
+nv35_graph_oclass = {
+	.handle = NV_ENGINE(GR, 0x35),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv35_graph_ctor,
+		.dtor = nv20_graph_dtor,
+		.init = nv30_graph_init,
+		.fini = _nouveau_graph_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv40.c b/drivers/gpu/drm/nouveau/core/engine/graph/nv40.c
new file mode 100644
index 0000000..8d00210
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv40.c
@@ -0,0 +1,495 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/handle.h>
+#include <core/engctx.h>
+
+#include <subdev/fb.h>
+#include <subdev/timer.h>
+
+#include <engine/graph.h>
+#include <engine/fifo.h>
+
+#include "nv40.h"
+#include "regs.h"
+
+struct nv40_graph_priv {
+	struct nouveau_graph base;
+	u32 size;
+};
+
+struct nv40_graph_chan {
+	struct nouveau_graph_chan base;
+};
+
+/*******************************************************************************
+ * Graphics object classes
+ ******************************************************************************/
+
+static int
+nv40_graph_object_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nouveau_gpuobj *obj;
+	int ret;
+
+	ret = nouveau_gpuobj_create(parent, engine, oclass, 0, parent,
+				    20, 16, 0, &obj);
+	*pobject = nv_object(obj);
+	if (ret)
+		return ret;
+
+	nv_wo32(obj, 0x00, nv_mclass(obj));
+	nv_wo32(obj, 0x04, 0x00000000);
+	nv_wo32(obj, 0x08, 0x00000000);
+#ifdef __BIG_ENDIAN
+	nv_mo32(obj, 0x08, 0x01000000, 0x01000000);
+#endif
+	nv_wo32(obj, 0x0c, 0x00000000);
+	nv_wo32(obj, 0x10, 0x00000000);
+	return 0;
+}
+
+static struct nouveau_ofuncs
+nv40_graph_ofuncs = {
+	.ctor = nv40_graph_object_ctor,
+	.dtor = _nouveau_gpuobj_dtor,
+	.init = _nouveau_gpuobj_init,
+	.fini = _nouveau_gpuobj_fini,
+	.rd32 = _nouveau_gpuobj_rd32,
+	.wr32 = _nouveau_gpuobj_wr32,
+};
+
+static struct nouveau_oclass
+nv40_graph_sclass[] = {
+	{ 0x0012, &nv40_graph_ofuncs, NULL }, /* beta1 */
+	{ 0x0019, &nv40_graph_ofuncs, NULL }, /* clip */
+	{ 0x0030, &nv40_graph_ofuncs, NULL }, /* null */
+	{ 0x0039, &nv40_graph_ofuncs, NULL }, /* m2mf */
+	{ 0x0043, &nv40_graph_ofuncs, NULL }, /* rop */
+	{ 0x0044, &nv40_graph_ofuncs, NULL }, /* patt */
+	{ 0x004a, &nv40_graph_ofuncs, NULL }, /* gdi */
+	{ 0x0062, &nv40_graph_ofuncs, NULL }, /* surf2d */
+	{ 0x0072, &nv40_graph_ofuncs, NULL }, /* beta4 */
+	{ 0x0089, &nv40_graph_ofuncs, NULL }, /* sifm */
+	{ 0x008a, &nv40_graph_ofuncs, NULL }, /* ifc */
+	{ 0x009f, &nv40_graph_ofuncs, NULL }, /* imageblit */
+	{ 0x3062, &nv40_graph_ofuncs, NULL }, /* surf2d (nv40) */
+	{ 0x3089, &nv40_graph_ofuncs, NULL }, /* sifm (nv40) */
+	{ 0x309e, &nv40_graph_ofuncs, NULL }, /* swzsurf (nv40) */
+	{ 0x4097, &nv40_graph_ofuncs, NULL }, /* curie */
+	{},
+};
+
+static struct nouveau_oclass
+nv44_graph_sclass[] = {
+	{ 0x0012, &nv40_graph_ofuncs, NULL }, /* beta1 */
+	{ 0x0019, &nv40_graph_ofuncs, NULL }, /* clip */
+	{ 0x0030, &nv40_graph_ofuncs, NULL }, /* null */
+	{ 0x0039, &nv40_graph_ofuncs, NULL }, /* m2mf */
+	{ 0x0043, &nv40_graph_ofuncs, NULL }, /* rop */
+	{ 0x0044, &nv40_graph_ofuncs, NULL }, /* patt */
+	{ 0x004a, &nv40_graph_ofuncs, NULL }, /* gdi */
+	{ 0x0062, &nv40_graph_ofuncs, NULL }, /* surf2d */
+	{ 0x0072, &nv40_graph_ofuncs, NULL }, /* beta4 */
+	{ 0x0089, &nv40_graph_ofuncs, NULL }, /* sifm */
+	{ 0x008a, &nv40_graph_ofuncs, NULL }, /* ifc */
+	{ 0x009f, &nv40_graph_ofuncs, NULL }, /* imageblit */
+	{ 0x3062, &nv40_graph_ofuncs, NULL }, /* surf2d (nv40) */
+	{ 0x3089, &nv40_graph_ofuncs, NULL }, /* sifm (nv40) */
+	{ 0x309e, &nv40_graph_ofuncs, NULL }, /* swzsurf (nv40) */
+	{ 0x4497, &nv40_graph_ofuncs, NULL }, /* curie */
+	{},
+};
+
+/*******************************************************************************
+ * PGRAPH context
+ ******************************************************************************/
+
+static int
+nv40_graph_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nv40_graph_priv *priv = (void *)engine;
+	struct nv40_graph_chan *chan;
+	int ret;
+
+	ret = nouveau_graph_context_create(parent, engine, oclass, NULL,
+					   priv->size, 16,
+					   NVOBJ_FLAG_ZERO_ALLOC, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	nv40_grctx_fill(nv_device(priv), nv_gpuobj(chan));
+	nv_wo32(chan, 0x00000, nv_gpuobj(chan)->addr >> 4);
+	return 0;
+}
+
+static int
+nv40_graph_context_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv04_graph_priv *priv = (void *)object->engine;
+	struct nv04_graph_chan *chan = (void *)object;
+	u32 inst = 0x01000000 | nv_gpuobj(chan)->addr >> 4;
+	int ret = 0;
+
+	nv_mask(priv, 0x400720, 0x00000001, 0x00000000);
+
+	if (nv_rd32(priv, 0x40032c) == inst) {
+		if (suspend) {
+			nv_wr32(priv, 0x400720, 0x00000000);
+			nv_wr32(priv, 0x400784, inst);
+			nv_mask(priv, 0x400310, 0x00000020, 0x00000020);
+			nv_mask(priv, 0x400304, 0x00000001, 0x00000001);
+			if (!nv_wait(priv, 0x400300, 0x00000001, 0x00000000)) {
+				u32 insn = nv_rd32(priv, 0x400308);
+				nv_warn(priv, "ctxprog timeout 0x%08x\n", insn);
+				ret = -EBUSY;
+			}
+		}
+
+		nv_mask(priv, 0x40032c, 0x01000000, 0x00000000);
+	}
+
+	if (nv_rd32(priv, 0x400330) == inst)
+		nv_mask(priv, 0x400330, 0x01000000, 0x00000000);
+
+	nv_mask(priv, 0x400720, 0x00000001, 0x00000001);
+	return ret;
+}
+
+static struct nouveau_oclass
+nv40_graph_cclass = {
+	.handle = NV_ENGCTX(GR, 0x40),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv40_graph_context_ctor,
+		.dtor = _nouveau_graph_context_dtor,
+		.init = _nouveau_graph_context_init,
+		.fini = nv40_graph_context_fini,
+		.rd32 = _nouveau_graph_context_rd32,
+		.wr32 = _nouveau_graph_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PGRAPH engine/subdev functions
+ ******************************************************************************/
+
+static void
+nv40_graph_tile_prog(struct nouveau_engine *engine, int i)
+{
+	struct nouveau_fb_tile *tile = &nouveau_fb(engine)->tile.region[i];
+	struct nouveau_fifo *pfifo = nouveau_fifo(engine);
+	struct nv40_graph_priv *priv = (void *)engine;
+	unsigned long flags;
+
+	pfifo->pause(pfifo, &flags);
+	nv04_graph_idle(priv);
+
+	switch (nv_device(priv)->chipset) {
+	case 0x40:
+	case 0x41: /* guess */
+	case 0x42:
+	case 0x43:
+	case 0x45: /* guess */
+	case 0x4e:
+		nv_wr32(priv, NV20_PGRAPH_TSIZE(i), tile->pitch);
+		nv_wr32(priv, NV20_PGRAPH_TLIMIT(i), tile->limit);
+		nv_wr32(priv, NV20_PGRAPH_TILE(i), tile->addr);
+		nv_wr32(priv, NV40_PGRAPH_TSIZE1(i), tile->pitch);
+		nv_wr32(priv, NV40_PGRAPH_TLIMIT1(i), tile->limit);
+		nv_wr32(priv, NV40_PGRAPH_TILE1(i), tile->addr);
+		break;
+	case 0x44:
+	case 0x4a:
+		nv_wr32(priv, NV20_PGRAPH_TSIZE(i), tile->pitch);
+		nv_wr32(priv, NV20_PGRAPH_TLIMIT(i), tile->limit);
+		nv_wr32(priv, NV20_PGRAPH_TILE(i), tile->addr);
+		break;
+	case 0x46:
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+	case 0x4c:
+	case 0x67:
+	default:
+		nv_wr32(priv, NV47_PGRAPH_TSIZE(i), tile->pitch);
+		nv_wr32(priv, NV47_PGRAPH_TLIMIT(i), tile->limit);
+		nv_wr32(priv, NV47_PGRAPH_TILE(i), tile->addr);
+		nv_wr32(priv, NV40_PGRAPH_TSIZE1(i), tile->pitch);
+		nv_wr32(priv, NV40_PGRAPH_TLIMIT1(i), tile->limit);
+		nv_wr32(priv, NV40_PGRAPH_TILE1(i), tile->addr);
+		break;
+	}
+
+	pfifo->start(pfifo, &flags);
+}
+
+static void
+nv40_graph_intr(struct nouveau_subdev *subdev)
+{
+	struct nouveau_fifo *pfifo = nouveau_fifo(subdev);
+	struct nouveau_engine *engine = nv_engine(subdev);
+	struct nouveau_object *engctx;
+	struct nouveau_handle *handle = NULL;
+	struct nv40_graph_priv *priv = (void *)subdev;
+	u32 stat = nv_rd32(priv, NV03_PGRAPH_INTR);
+	u32 nsource = nv_rd32(priv, NV03_PGRAPH_NSOURCE);
+	u32 nstatus = nv_rd32(priv, NV03_PGRAPH_NSTATUS);
+	u32 inst = nv_rd32(priv, 0x40032c) & 0x000fffff;
+	u32 addr = nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR);
+	u32 subc = (addr & 0x00070000) >> 16;
+	u32 mthd = (addr & 0x00001ffc);
+	u32 data = nv_rd32(priv, NV04_PGRAPH_TRAPPED_DATA);
+	u32 class = nv_rd32(priv, 0x400160 + subc * 4) & 0xffff;
+	u32 show = stat;
+	int chid;
+
+	engctx = nouveau_engctx_get(engine, inst);
+	chid   = pfifo->chid(pfifo, engctx);
+
+	if (stat & NV_PGRAPH_INTR_ERROR) {
+		if (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD) {
+			handle = nouveau_handle_get_class(engctx, class);
+			if (handle && !nv_call(handle->object, mthd, data))
+				show &= ~NV_PGRAPH_INTR_ERROR;
+			nouveau_handle_put(handle);
+		}
+
+		if (nsource & NV03_PGRAPH_NSOURCE_DMA_VTX_PROTECTION) {
+			nv_mask(priv, 0x402000, 0, 0);
+		}
+	}
+
+	nv_wr32(priv, NV03_PGRAPH_INTR, stat);
+	nv_wr32(priv, NV04_PGRAPH_FIFO, 0x00000001);
+
+	if (show) {
+		nv_info(priv, "");
+		nouveau_bitfield_print(nv10_graph_intr_name, show);
+		printk(" nsource:");
+		nouveau_bitfield_print(nv04_graph_nsource, nsource);
+		printk(" nstatus:");
+		nouveau_bitfield_print(nv10_graph_nstatus, nstatus);
+		printk("\n");
+		nv_error(priv, "ch %d [0x%08x] subc %d class 0x%04x "
+			       "mthd 0x%04x data 0x%08x\n",
+			 chid, inst << 4, subc, class, mthd, data);
+	}
+
+	nouveau_engctx_put(engctx);
+}
+
+static int
+nv40_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv40_graph_priv *priv;
+	int ret;
+
+	ret = nouveau_graph_create(parent, engine, oclass, true, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00001000;
+	nv_subdev(priv)->intr = nv40_graph_intr;
+	nv_engine(priv)->cclass = &nv40_graph_cclass;
+	if (nv44_graph_class(priv))
+		nv_engine(priv)->sclass = nv44_graph_sclass;
+	else
+		nv_engine(priv)->sclass = nv40_graph_sclass;
+	nv_engine(priv)->tile_prog = nv40_graph_tile_prog;
+	return 0;
+}
+
+static int
+nv40_graph_init(struct nouveau_object *object)
+{
+	struct nouveau_engine *engine = nv_engine(object);
+	struct nouveau_fb *pfb = nouveau_fb(object);
+	struct nv40_graph_priv *priv = (void *)engine;
+	int ret, i, j;
+	u32 vramsz;
+
+	ret = nouveau_graph_init(&priv->base);
+	if (ret)
+		return ret;
+
+	/* generate and upload context program */
+	nv40_grctx_init(nv_device(priv), &priv->size);
+
+	/* No context present currently */
+	nv_wr32(priv, NV40_PGRAPH_CTXCTL_CUR, 0x00000000);
+
+	nv_wr32(priv, NV03_PGRAPH_INTR   , 0xFFFFFFFF);
+	nv_wr32(priv, NV40_PGRAPH_INTR_EN, 0xFFFFFFFF);
+
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x00000000);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_1, 0x401287c0);
+	nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0xe0de8055);
+	nv_wr32(priv, NV10_PGRAPH_DEBUG_4, 0x00008000);
+	nv_wr32(priv, NV04_PGRAPH_LIMIT_VIOL_PIX, 0x00be3c5f);
+
+	nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10010100);
+	nv_wr32(priv, NV10_PGRAPH_STATE      , 0xFFFFFFFF);
+
+	j = nv_rd32(priv, 0x1540) & 0xff;
+	if (j) {
+		for (i = 0; !(j & 1); j >>= 1, i++)
+			;
+		nv_wr32(priv, 0x405000, i);
+	}
+
+	if (nv_device(priv)->chipset == 0x40) {
+		nv_wr32(priv, 0x4009b0, 0x83280fff);
+		nv_wr32(priv, 0x4009b4, 0x000000a0);
+	} else {
+		nv_wr32(priv, 0x400820, 0x83280eff);
+		nv_wr32(priv, 0x400824, 0x000000a0);
+	}
+
+	switch (nv_device(priv)->chipset) {
+	case 0x40:
+	case 0x45:
+		nv_wr32(priv, 0x4009b8, 0x0078e366);
+		nv_wr32(priv, 0x4009bc, 0x0000014c);
+		break;
+	case 0x41:
+	case 0x42: /* pciid also 0x00Cx */
+	/* case 0x0120: XXX (pciid) */
+		nv_wr32(priv, 0x400828, 0x007596ff);
+		nv_wr32(priv, 0x40082c, 0x00000108);
+		break;
+	case 0x43:
+		nv_wr32(priv, 0x400828, 0x0072cb77);
+		nv_wr32(priv, 0x40082c, 0x00000108);
+		break;
+	case 0x44:
+	case 0x46: /* G72 */
+	case 0x4a:
+	case 0x4c: /* G7x-based C51 */
+	case 0x4e:
+		nv_wr32(priv, 0x400860, 0);
+		nv_wr32(priv, 0x400864, 0);
+		break;
+	case 0x47: /* G70 */
+	case 0x49: /* G71 */
+	case 0x4b: /* G73 */
+		nv_wr32(priv, 0x400828, 0x07830610);
+		nv_wr32(priv, 0x40082c, 0x0000016A);
+		break;
+	default:
+		break;
+	}
+
+	nv_wr32(priv, 0x400b38, 0x2ffff800);
+	nv_wr32(priv, 0x400b3c, 0x00006000);
+
+	/* Tiling related stuff. */
+	switch (nv_device(priv)->chipset) {
+	case 0x44:
+	case 0x4a:
+		nv_wr32(priv, 0x400bc4, 0x1003d888);
+		nv_wr32(priv, 0x400bbc, 0xb7a7b500);
+		break;
+	case 0x46:
+		nv_wr32(priv, 0x400bc4, 0x0000e024);
+		nv_wr32(priv, 0x400bbc, 0xb7a7b520);
+		break;
+	case 0x4c:
+	case 0x4e:
+	case 0x67:
+		nv_wr32(priv, 0x400bc4, 0x1003d888);
+		nv_wr32(priv, 0x400bbc, 0xb7a7b540);
+		break;
+	default:
+		break;
+	}
+
+	/* Turn all the tiling regions off. */
+	for (i = 0; i < pfb->tile.regions; i++)
+		engine->tile_prog(engine, i);
+
+	/* begin RAM config */
+	vramsz = pci_resource_len(nv_device(priv)->pdev, 0) - 1;
+	switch (nv_device(priv)->chipset) {
+	case 0x40:
+		nv_wr32(priv, 0x4009A4, nv_rd32(priv, 0x100200));
+		nv_wr32(priv, 0x4009A8, nv_rd32(priv, 0x100204));
+		nv_wr32(priv, 0x4069A4, nv_rd32(priv, 0x100200));
+		nv_wr32(priv, 0x4069A8, nv_rd32(priv, 0x100204));
+		nv_wr32(priv, 0x400820, 0);
+		nv_wr32(priv, 0x400824, 0);
+		nv_wr32(priv, 0x400864, vramsz);
+		nv_wr32(priv, 0x400868, vramsz);
+		break;
+	default:
+		switch (nv_device(priv)->chipset) {
+		case 0x41:
+		case 0x42:
+		case 0x43:
+		case 0x45:
+		case 0x4e:
+		case 0x44:
+		case 0x4a:
+			nv_wr32(priv, 0x4009F0, nv_rd32(priv, 0x100200));
+			nv_wr32(priv, 0x4009F4, nv_rd32(priv, 0x100204));
+			break;
+		default:
+			nv_wr32(priv, 0x400DF0, nv_rd32(priv, 0x100200));
+			nv_wr32(priv, 0x400DF4, nv_rd32(priv, 0x100204));
+			break;
+		}
+		nv_wr32(priv, 0x4069F0, nv_rd32(priv, 0x100200));
+		nv_wr32(priv, 0x4069F4, nv_rd32(priv, 0x100204));
+		nv_wr32(priv, 0x400840, 0);
+		nv_wr32(priv, 0x400844, 0);
+		nv_wr32(priv, 0x4008A0, vramsz);
+		nv_wr32(priv, 0x4008A4, vramsz);
+		break;
+	}
+
+	return 0;
+}
+
+struct nouveau_oclass
+nv40_graph_oclass = {
+	.handle = NV_ENGINE(GR, 0x40),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv40_graph_ctor,
+		.dtor = _nouveau_graph_dtor,
+		.init = nv40_graph_init,
+		.fini = _nouveau_graph_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv40.h b/drivers/gpu/drm/nouveau/core/engine/graph/nv40.h
new file mode 100644
index 0000000..d2ac975
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv40.h
@@ -0,0 +1,21 @@
+#ifndef __NV40_GRAPH_H__
+#define __NV40_GRAPH_H__
+
+/* returns 1 if device is one of the nv4x using the 0x4497 object class,
+ * helpful to determine a number of other hardware features
+ */
+static inline int
+nv44_graph_class(void *priv)
+{
+	struct nouveau_device *device = nv_device(priv);
+
+	if ((device->chipset & 0xf0) == 0x60)
+		return 1;
+
+	return !(0x0baf & (1 << (device->chipset & 0x0f)));
+}
+
+void nv40_grctx_init(struct nouveau_device *, u32 *size);
+void nv40_grctx_fill(struct nouveau_device *, struct nouveau_gpuobj *);
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv50.c b/drivers/gpu/drm/nouveau/core/engine/graph/nv50.c
new file mode 100644
index 0000000..ab3b9dc
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv50.c
@@ -0,0 +1,888 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/handle.h>
+#include <core/engctx.h>
+#include <core/enum.h>
+
+#include <subdev/fb.h>
+#include <subdev/vm.h>
+#include <subdev/timer.h>
+
+#include <engine/fifo.h>
+#include <engine/graph.h>
+
+#include "nv50.h"
+
+struct nv50_graph_priv {
+	struct nouveau_graph base;
+	spinlock_t lock;
+	u32 size;
+};
+
+struct nv50_graph_chan {
+	struct nouveau_graph_chan base;
+};
+
+/*******************************************************************************
+ * Graphics object classes
+ ******************************************************************************/
+
+static int
+nv50_graph_object_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nouveau_gpuobj *obj;
+	int ret;
+
+	ret = nouveau_gpuobj_create(parent, engine, oclass, 0, parent,
+				    16, 16, 0, &obj);
+	*pobject = nv_object(obj);
+	if (ret)
+		return ret;
+
+	nv_wo32(obj, 0x00, nv_mclass(obj));
+	nv_wo32(obj, 0x04, 0x00000000);
+	nv_wo32(obj, 0x08, 0x00000000);
+	nv_wo32(obj, 0x0c, 0x00000000);
+	return 0;
+}
+
+static struct nouveau_ofuncs
+nv50_graph_ofuncs = {
+	.ctor = nv50_graph_object_ctor,
+	.dtor = _nouveau_gpuobj_dtor,
+	.init = _nouveau_gpuobj_init,
+	.fini = _nouveau_gpuobj_fini,
+	.rd32 = _nouveau_gpuobj_rd32,
+	.wr32 = _nouveau_gpuobj_wr32,
+};
+
+static struct nouveau_oclass
+nv50_graph_sclass[] = {
+	{ 0x0030, &nv50_graph_ofuncs },
+	{ 0x502d, &nv50_graph_ofuncs },
+	{ 0x5039, &nv50_graph_ofuncs },
+	{ 0x5097, &nv50_graph_ofuncs },
+	{ 0x50c0, &nv50_graph_ofuncs },
+	{}
+};
+
+static struct nouveau_oclass
+nv84_graph_sclass[] = {
+	{ 0x0030, &nv50_graph_ofuncs },
+	{ 0x502d, &nv50_graph_ofuncs },
+	{ 0x5039, &nv50_graph_ofuncs },
+	{ 0x50c0, &nv50_graph_ofuncs },
+	{ 0x8297, &nv50_graph_ofuncs },
+	{}
+};
+
+static struct nouveau_oclass
+nva0_graph_sclass[] = {
+	{ 0x0030, &nv50_graph_ofuncs },
+	{ 0x502d, &nv50_graph_ofuncs },
+	{ 0x5039, &nv50_graph_ofuncs },
+	{ 0x50c0, &nv50_graph_ofuncs },
+	{ 0x8397, &nv50_graph_ofuncs },
+	{}
+};
+
+static struct nouveau_oclass
+nva3_graph_sclass[] = {
+	{ 0x0030, &nv50_graph_ofuncs },
+	{ 0x502d, &nv50_graph_ofuncs },
+	{ 0x5039, &nv50_graph_ofuncs },
+	{ 0x50c0, &nv50_graph_ofuncs },
+	{ 0x8597, &nv50_graph_ofuncs },
+	{ 0x85c0, &nv50_graph_ofuncs },
+	{}
+};
+
+static struct nouveau_oclass
+nvaf_graph_sclass[] = {
+	{ 0x0030, &nv50_graph_ofuncs },
+	{ 0x502d, &nv50_graph_ofuncs },
+	{ 0x5039, &nv50_graph_ofuncs },
+	{ 0x50c0, &nv50_graph_ofuncs },
+	{ 0x85c0, &nv50_graph_ofuncs },
+	{ 0x8697, &nv50_graph_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * PGRAPH context
+ ******************************************************************************/
+
+static int
+nv50_graph_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *data, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nv50_graph_priv *priv = (void *)engine;
+	struct nv50_graph_chan *chan;
+	int ret;
+
+	ret = nouveau_graph_context_create(parent, engine, oclass, NULL,
+					   priv->size, 0,
+					   NVOBJ_FLAG_ZERO_ALLOC, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	nv50_grctx_fill(nv_device(priv), nv_gpuobj(chan));
+	return 0;
+}
+
+static struct nouveau_oclass
+nv50_graph_cclass = {
+	.handle = NV_ENGCTX(GR, 0x50),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv50_graph_context_ctor,
+		.dtor = _nouveau_graph_context_dtor,
+		.init = _nouveau_graph_context_init,
+		.fini = _nouveau_graph_context_fini,
+		.rd32 = _nouveau_graph_context_rd32,
+		.wr32 = _nouveau_graph_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PGRAPH engine/subdev functions
+ ******************************************************************************/
+
+static int
+nv50_graph_tlb_flush(struct nouveau_engine *engine)
+{
+	nv50_vm_flush_engine(&engine->base, 0x00);
+	return 0;
+}
+
+static int
+nv84_graph_tlb_flush(struct nouveau_engine *engine)
+{
+	struct nouveau_timer *ptimer = nouveau_timer(engine);
+	struct nv50_graph_priv *priv = (void *)engine;
+	bool idle, timeout = false;
+	unsigned long flags;
+	u64 start;
+	u32 tmp;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	nv_mask(priv, 0x400500, 0x00000001, 0x00000000);
+
+	start = ptimer->read(ptimer);
+	do {
+		idle = true;
+
+		for (tmp = nv_rd32(priv, 0x400380); tmp && idle; tmp >>= 3) {
+			if ((tmp & 7) == 1)
+				idle = false;
+		}
+
+		for (tmp = nv_rd32(priv, 0x400384); tmp && idle; tmp >>= 3) {
+			if ((tmp & 7) == 1)
+				idle = false;
+		}
+
+		for (tmp = nv_rd32(priv, 0x400388); tmp && idle; tmp >>= 3) {
+			if ((tmp & 7) == 1)
+				idle = false;
+		}
+	} while (!idle &&
+		 !(timeout = ptimer->read(ptimer) - start > 2000000000));
+
+	if (timeout) {
+		nv_error(priv, "PGRAPH TLB flush idle timeout fail: "
+			      "0x%08x 0x%08x 0x%08x 0x%08x\n",
+			 nv_rd32(priv, 0x400700), nv_rd32(priv, 0x400380),
+			 nv_rd32(priv, 0x400384), nv_rd32(priv, 0x400388));
+	}
+
+	nv50_vm_flush_engine(&engine->base, 0x00);
+
+	nv_mask(priv, 0x400500, 0x00000001, 0x00000001);
+	spin_unlock_irqrestore(&priv->lock, flags);
+	return timeout ? -EBUSY : 0;
+}
+
+static const struct nouveau_enum nv50_mp_exec_error_names[] = {
+	{ 3, "STACK_UNDERFLOW", NULL },
+	{ 4, "QUADON_ACTIVE", NULL },
+	{ 8, "TIMEOUT", NULL },
+	{ 0x10, "INVALID_OPCODE", NULL },
+	{ 0x40, "BREAKPOINT", NULL },
+	{}
+};
+
+static const struct nouveau_bitfield nv50_graph_trap_m2mf[] = {
+	{ 0x00000001, "NOTIFY" },
+	{ 0x00000002, "IN" },
+	{ 0x00000004, "OUT" },
+	{}
+};
+
+static const struct nouveau_bitfield nv50_graph_trap_vfetch[] = {
+	{ 0x00000001, "FAULT" },
+	{}
+};
+
+static const struct nouveau_bitfield nv50_graph_trap_strmout[] = {
+	{ 0x00000001, "FAULT" },
+	{}
+};
+
+static const struct nouveau_bitfield nv50_graph_trap_ccache[] = {
+	{ 0x00000001, "FAULT" },
+	{}
+};
+
+/* There must be a *lot* of these. Will take some time to gather them up. */
+const struct nouveau_enum nv50_data_error_names[] = {
+	{ 0x00000003, "INVALID_OPERATION", NULL },
+	{ 0x00000004, "INVALID_VALUE", NULL },
+	{ 0x00000005, "INVALID_ENUM", NULL },
+	{ 0x00000008, "INVALID_OBJECT", NULL },
+	{ 0x00000009, "READ_ONLY_OBJECT", NULL },
+	{ 0x0000000a, "SUPERVISOR_OBJECT", NULL },
+	{ 0x0000000b, "INVALID_ADDRESS_ALIGNMENT", NULL },
+	{ 0x0000000c, "INVALID_BITFIELD", NULL },
+	{ 0x0000000d, "BEGIN_END_ACTIVE", NULL },
+	{ 0x0000000e, "SEMANTIC_COLOR_BACK_OVER_LIMIT", NULL },
+	{ 0x0000000f, "VIEWPORT_ID_NEEDS_GP", NULL },
+	{ 0x00000010, "RT_DOUBLE_BIND", NULL },
+	{ 0x00000011, "RT_TYPES_MISMATCH", NULL },
+	{ 0x00000012, "RT_LINEAR_WITH_ZETA", NULL },
+	{ 0x00000015, "FP_TOO_FEW_REGS", NULL },
+	{ 0x00000016, "ZETA_FORMAT_CSAA_MISMATCH", NULL },
+	{ 0x00000017, "RT_LINEAR_WITH_MSAA", NULL },
+	{ 0x00000018, "FP_INTERPOLANT_START_OVER_LIMIT", NULL },
+	{ 0x00000019, "SEMANTIC_LAYER_OVER_LIMIT", NULL },
+	{ 0x0000001a, "RT_INVALID_ALIGNMENT", NULL },
+	{ 0x0000001b, "SAMPLER_OVER_LIMIT", NULL },
+	{ 0x0000001c, "TEXTURE_OVER_LIMIT", NULL },
+	{ 0x0000001e, "GP_TOO_MANY_OUTPUTS", NULL },
+	{ 0x0000001f, "RT_BPP128_WITH_MS8", NULL },
+	{ 0x00000021, "Z_OUT_OF_BOUNDS", NULL },
+	{ 0x00000023, "XY_OUT_OF_BOUNDS", NULL },
+	{ 0x00000024, "VP_ZERO_INPUTS", NULL },
+	{ 0x00000027, "CP_MORE_PARAMS_THAN_SHARED", NULL },
+	{ 0x00000028, "CP_NO_REG_SPACE_STRIPED", NULL },
+	{ 0x00000029, "CP_NO_REG_SPACE_PACKED", NULL },
+	{ 0x0000002a, "CP_NOT_ENOUGH_WARPS", NULL },
+	{ 0x0000002b, "CP_BLOCK_SIZE_MISMATCH", NULL },
+	{ 0x0000002c, "CP_NOT_ENOUGH_LOCAL_WARPS", NULL },
+	{ 0x0000002d, "CP_NOT_ENOUGH_STACK_WARPS", NULL },
+	{ 0x0000002e, "CP_NO_BLOCKDIM_LATCH", NULL },
+	{ 0x00000031, "ENG2D_FORMAT_MISMATCH", NULL },
+	{ 0x0000003f, "PRIMITIVE_ID_NEEDS_GP", NULL },
+	{ 0x00000044, "SEMANTIC_VIEWPORT_OVER_LIMIT", NULL },
+	{ 0x00000045, "SEMANTIC_COLOR_FRONT_OVER_LIMIT", NULL },
+	{ 0x00000046, "LAYER_ID_NEEDS_GP", NULL },
+	{ 0x00000047, "SEMANTIC_CLIP_OVER_LIMIT", NULL },
+	{ 0x00000048, "SEMANTIC_PTSZ_OVER_LIMIT", NULL },
+	{}
+};
+
+static const struct nouveau_bitfield nv50_graph_intr_name[] = {
+	{ 0x00000001, "NOTIFY" },
+	{ 0x00000002, "COMPUTE_QUERY" },
+	{ 0x00000010, "ILLEGAL_MTHD" },
+	{ 0x00000020, "ILLEGAL_CLASS" },
+	{ 0x00000040, "DOUBLE_NOTIFY" },
+	{ 0x00001000, "CONTEXT_SWITCH" },
+	{ 0x00010000, "BUFFER_NOTIFY" },
+	{ 0x00100000, "DATA_ERROR" },
+	{ 0x00200000, "TRAP" },
+	{ 0x01000000, "SINGLE_STEP" },
+	{}
+};
+
+static void
+nv50_priv_mp_trap(struct nv50_graph_priv *priv, int tpid, int display)
+{
+	u32 units = nv_rd32(priv, 0x1540);
+	u32 addr, mp10, status, pc, oplow, ophigh;
+	int i;
+	int mps = 0;
+	for (i = 0; i < 4; i++) {
+		if (!(units & 1 << (i+24)))
+			continue;
+		if (nv_device(priv)->chipset < 0xa0)
+			addr = 0x408200 + (tpid << 12) + (i << 7);
+		else
+			addr = 0x408100 + (tpid << 11) + (i << 7);
+		mp10 = nv_rd32(priv, addr + 0x10);
+		status = nv_rd32(priv, addr + 0x14);
+		if (!status)
+			continue;
+		if (display) {
+			nv_rd32(priv, addr + 0x20);
+			pc = nv_rd32(priv, addr + 0x24);
+			oplow = nv_rd32(priv, addr + 0x70);
+			ophigh = nv_rd32(priv, addr + 0x74);
+			nv_error(priv, "TRAP_MP_EXEC - "
+					"TP %d MP %d: ", tpid, i);
+			nouveau_enum_print(nv50_mp_exec_error_names, status);
+			printk(" at %06x warp %d, opcode %08x %08x\n",
+					pc&0xffffff, pc >> 24,
+					oplow, ophigh);
+		}
+		nv_wr32(priv, addr + 0x10, mp10);
+		nv_wr32(priv, addr + 0x14, 0);
+		mps++;
+	}
+	if (!mps && display)
+		nv_error(priv, "TRAP_MP_EXEC - TP %d: "
+				"No MPs claiming errors?\n", tpid);
+}
+
+static void
+nv50_priv_tp_trap(struct nv50_graph_priv *priv, int type, u32 ustatus_old,
+		u32 ustatus_new, int display, const char *name)
+{
+	int tps = 0;
+	u32 units = nv_rd32(priv, 0x1540);
+	int i, r;
+	u32 ustatus_addr, ustatus;
+	for (i = 0; i < 16; i++) {
+		if (!(units & (1 << i)))
+			continue;
+		if (nv_device(priv)->chipset < 0xa0)
+			ustatus_addr = ustatus_old + (i << 12);
+		else
+			ustatus_addr = ustatus_new + (i << 11);
+		ustatus = nv_rd32(priv, ustatus_addr) & 0x7fffffff;
+		if (!ustatus)
+			continue;
+		tps++;
+		switch (type) {
+		case 6: /* texture error... unknown for now */
+			if (display) {
+				nv_error(priv, "magic set %d:\n", i);
+				for (r = ustatus_addr + 4; r <= ustatus_addr + 0x10; r += 4)
+					nv_error(priv, "\t0x%08x: 0x%08x\n", r,
+						nv_rd32(priv, r));
+			}
+			break;
+		case 7: /* MP error */
+			if (ustatus & 0x04030000) {
+				nv50_priv_mp_trap(priv, i, display);
+				ustatus &= ~0x04030000;
+			}
+			break;
+		case 8: /* TPDMA error */
+			{
+			u32 e0c = nv_rd32(priv, ustatus_addr + 4);
+			u32 e10 = nv_rd32(priv, ustatus_addr + 8);
+			u32 e14 = nv_rd32(priv, ustatus_addr + 0xc);
+			u32 e18 = nv_rd32(priv, ustatus_addr + 0x10);
+			u32 e1c = nv_rd32(priv, ustatus_addr + 0x14);
+			u32 e20 = nv_rd32(priv, ustatus_addr + 0x18);
+			u32 e24 = nv_rd32(priv, ustatus_addr + 0x1c);
+			/* 2d engine destination */
+			if (ustatus & 0x00000010) {
+				if (display) {
+					nv_error(priv, "TRAP_TPDMA_2D - TP %d - Unknown fault at address %02x%08x\n",
+							i, e14, e10);
+					nv_error(priv, "TRAP_TPDMA_2D - TP %d - e0c: %08x, e18: %08x, e1c: %08x, e20: %08x, e24: %08x\n",
+							i, e0c, e18, e1c, e20, e24);
+				}
+				ustatus &= ~0x00000010;
+			}
+			/* Render target */
+			if (ustatus & 0x00000040) {
+				if (display) {
+					nv_error(priv, "TRAP_TPDMA_RT - TP %d - Unknown fault at address %02x%08x\n",
+							i, e14, e10);
+					nv_error(priv, "TRAP_TPDMA_RT - TP %d - e0c: %08x, e18: %08x, e1c: %08x, e20: %08x, e24: %08x\n",
+							i, e0c, e18, e1c, e20, e24);
+				}
+				ustatus &= ~0x00000040;
+			}
+			/* CUDA memory: l[], g[] or stack. */
+			if (ustatus & 0x00000080) {
+				if (display) {
+					if (e18 & 0x80000000) {
+						/* g[] read fault? */
+						nv_error(priv, "TRAP_TPDMA - TP %d - Global read fault at address %02x%08x\n",
+								i, e14, e10 | ((e18 >> 24) & 0x1f));
+						e18 &= ~0x1f000000;
+					} else if (e18 & 0xc) {
+						/* g[] write fault? */
+						nv_error(priv, "TRAP_TPDMA - TP %d - Global write fault at address %02x%08x\n",
+								i, e14, e10 | ((e18 >> 7) & 0x1f));
+						e18 &= ~0x00000f80;
+					} else {
+						nv_error(priv, "TRAP_TPDMA - TP %d - Unknown CUDA fault at address %02x%08x\n",
+								i, e14, e10);
+					}
+					nv_error(priv, "TRAP_TPDMA - TP %d - e0c: %08x, e18: %08x, e1c: %08x, e20: %08x, e24: %08x\n",
+							i, e0c, e18, e1c, e20, e24);
+				}
+				ustatus &= ~0x00000080;
+			}
+			}
+			break;
+		}
+		if (ustatus) {
+			if (display)
+				nv_info(priv, "%s - TP%d: Unhandled ustatus 0x%08x\n", name, i, ustatus);
+		}
+		nv_wr32(priv, ustatus_addr, 0xc0000000);
+	}
+
+	if (!tps && display)
+		nv_info(priv, "%s - No TPs claiming errors?\n", name);
+}
+
+static int
+nv50_graph_trap_handler(struct nv50_graph_priv *priv, u32 display,
+			int chid, u64 inst)
+{
+	u32 status = nv_rd32(priv, 0x400108);
+	u32 ustatus;
+
+	if (!status && display) {
+		nv_error(priv, "TRAP: no units reporting traps?\n");
+		return 1;
+	}
+
+	/* DISPATCH: Relays commands to other units and handles NOTIFY,
+	 * COND, QUERY. If you get a trap from it, the command is still stuck
+	 * in DISPATCH and you need to do something about it. */
+	if (status & 0x001) {
+		ustatus = nv_rd32(priv, 0x400804) & 0x7fffffff;
+		if (!ustatus && display) {
+			nv_error(priv, "TRAP_DISPATCH - no ustatus?\n");
+		}
+
+		nv_wr32(priv, 0x400500, 0x00000000);
+
+		/* Known to be triggered by screwed up NOTIFY and COND... */
+		if (ustatus & 0x00000001) {
+			u32 addr = nv_rd32(priv, 0x400808);
+			u32 subc = (addr & 0x00070000) >> 16;
+			u32 mthd = (addr & 0x00001ffc);
+			u32 datal = nv_rd32(priv, 0x40080c);
+			u32 datah = nv_rd32(priv, 0x400810);
+			u32 class = nv_rd32(priv, 0x400814);
+			u32 r848 = nv_rd32(priv, 0x400848);
+
+			nv_error(priv, "TRAP DISPATCH_FAULT\n");
+			if (display && (addr & 0x80000000)) {
+				nv_error(priv, "ch %d [0x%010llx] "
+					     "subc %d class 0x%04x mthd 0x%04x "
+					     "data 0x%08x%08x "
+					     "400808 0x%08x 400848 0x%08x\n",
+					chid, inst, subc, class, mthd, datah,
+					datal, addr, r848);
+			} else
+			if (display) {
+				nv_error(priv, "no stuck command?\n");
+			}
+
+			nv_wr32(priv, 0x400808, 0);
+			nv_wr32(priv, 0x4008e8, nv_rd32(priv, 0x4008e8) & 3);
+			nv_wr32(priv, 0x400848, 0);
+			ustatus &= ~0x00000001;
+		}
+
+		if (ustatus & 0x00000002) {
+			u32 addr = nv_rd32(priv, 0x40084c);
+			u32 subc = (addr & 0x00070000) >> 16;
+			u32 mthd = (addr & 0x00001ffc);
+			u32 data = nv_rd32(priv, 0x40085c);
+			u32 class = nv_rd32(priv, 0x400814);
+
+			nv_error(priv, "TRAP DISPATCH_QUERY\n");
+			if (display && (addr & 0x80000000)) {
+				nv_error(priv, "ch %d [0x%010llx] "
+					     "subc %d class 0x%04x mthd 0x%04x "
+					     "data 0x%08x 40084c 0x%08x\n",
+					chid, inst, subc, class, mthd,
+					data, addr);
+			} else
+			if (display) {
+				nv_error(priv, "no stuck command?\n");
+			}
+
+			nv_wr32(priv, 0x40084c, 0);
+			ustatus &= ~0x00000002;
+		}
+
+		if (ustatus && display) {
+			nv_error(priv, "TRAP_DISPATCH (unknown "
+				      "0x%08x)\n", ustatus);
+		}
+
+		nv_wr32(priv, 0x400804, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x001);
+		status &= ~0x001;
+		if (!status)
+			return 0;
+	}
+
+	/* M2MF: Memory to memory copy engine. */
+	if (status & 0x002) {
+		u32 ustatus = nv_rd32(priv, 0x406800) & 0x7fffffff;
+		if (display) {
+			nv_error(priv, "TRAP_M2MF");
+			nouveau_bitfield_print(nv50_graph_trap_m2mf, ustatus);
+			printk("\n");
+			nv_error(priv, "TRAP_M2MF %08x %08x %08x %08x\n",
+				nv_rd32(priv, 0x406804), nv_rd32(priv, 0x406808),
+				nv_rd32(priv, 0x40680c), nv_rd32(priv, 0x406810));
+
+		}
+
+		/* No sane way found yet -- just reset the bugger. */
+		nv_wr32(priv, 0x400040, 2);
+		nv_wr32(priv, 0x400040, 0);
+		nv_wr32(priv, 0x406800, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x002);
+		status &= ~0x002;
+	}
+
+	/* VFETCH: Fetches data from vertex buffers. */
+	if (status & 0x004) {
+		u32 ustatus = nv_rd32(priv, 0x400c04) & 0x7fffffff;
+		if (display) {
+			nv_error(priv, "TRAP_VFETCH");
+			nouveau_bitfield_print(nv50_graph_trap_vfetch, ustatus);
+			printk("\n");
+			nv_error(priv, "TRAP_VFETCH %08x %08x %08x %08x\n",
+				nv_rd32(priv, 0x400c00), nv_rd32(priv, 0x400c08),
+				nv_rd32(priv, 0x400c0c), nv_rd32(priv, 0x400c10));
+		}
+
+		nv_wr32(priv, 0x400c04, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x004);
+		status &= ~0x004;
+	}
+
+	/* STRMOUT: DirectX streamout / OpenGL transform feedback. */
+	if (status & 0x008) {
+		ustatus = nv_rd32(priv, 0x401800) & 0x7fffffff;
+		if (display) {
+			nv_error(priv, "TRAP_STRMOUT");
+			nouveau_bitfield_print(nv50_graph_trap_strmout, ustatus);
+			printk("\n");
+			nv_error(priv, "TRAP_STRMOUT %08x %08x %08x %08x\n",
+				nv_rd32(priv, 0x401804), nv_rd32(priv, 0x401808),
+				nv_rd32(priv, 0x40180c), nv_rd32(priv, 0x401810));
+
+		}
+
+		/* No sane way found yet -- just reset the bugger. */
+		nv_wr32(priv, 0x400040, 0x80);
+		nv_wr32(priv, 0x400040, 0);
+		nv_wr32(priv, 0x401800, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x008);
+		status &= ~0x008;
+	}
+
+	/* CCACHE: Handles code and c[] caches and fills them. */
+	if (status & 0x010) {
+		ustatus = nv_rd32(priv, 0x405018) & 0x7fffffff;
+		if (display) {
+			nv_error(priv, "TRAP_CCACHE");
+			nouveau_bitfield_print(nv50_graph_trap_ccache, ustatus);
+			printk("\n");
+			nv_error(priv, "TRAP_CCACHE %08x %08x %08x %08x"
+				     " %08x %08x %08x\n",
+				nv_rd32(priv, 0x405000), nv_rd32(priv, 0x405004),
+				nv_rd32(priv, 0x405008), nv_rd32(priv, 0x40500c),
+				nv_rd32(priv, 0x405010), nv_rd32(priv, 0x405014),
+				nv_rd32(priv, 0x40501c));
+
+		}
+
+		nv_wr32(priv, 0x405018, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x010);
+		status &= ~0x010;
+	}
+
+	/* Unknown, not seen yet... 0x402000 is the only trap status reg
+	 * remaining, so try to handle it anyway. Perhaps related to that
+	 * unknown DMA slot on tesla? */
+	if (status & 0x20) {
+		ustatus = nv_rd32(priv, 0x402000) & 0x7fffffff;
+		if (display)
+			nv_error(priv, "TRAP_UNKC04 0x%08x\n", ustatus);
+		nv_wr32(priv, 0x402000, 0xc0000000);
+		/* no status modifiction on purpose */
+	}
+
+	/* TEXTURE: CUDA texturing units */
+	if (status & 0x040) {
+		nv50_priv_tp_trap(priv, 6, 0x408900, 0x408600, display,
+				    "TRAP_TEXTURE");
+		nv_wr32(priv, 0x400108, 0x040);
+		status &= ~0x040;
+	}
+
+	/* MP: CUDA execution engines. */
+	if (status & 0x080) {
+		nv50_priv_tp_trap(priv, 7, 0x408314, 0x40831c, display,
+				    "TRAP_MP");
+		nv_wr32(priv, 0x400108, 0x080);
+		status &= ~0x080;
+	}
+
+	/* TPDMA:  Handles TP-initiated uncached memory accesses:
+	 * l[], g[], stack, 2d surfaces, render targets. */
+	if (status & 0x100) {
+		nv50_priv_tp_trap(priv, 8, 0x408e08, 0x408708, display,
+				    "TRAP_TPDMA");
+		nv_wr32(priv, 0x400108, 0x100);
+		status &= ~0x100;
+	}
+
+	if (status) {
+		if (display)
+			nv_error(priv, "TRAP: unknown 0x%08x\n", status);
+		nv_wr32(priv, 0x400108, status);
+	}
+
+	return 1;
+}
+
+static void
+nv50_graph_intr(struct nouveau_subdev *subdev)
+{
+	struct nouveau_fifo *pfifo = nouveau_fifo(subdev);
+	struct nouveau_engine *engine = nv_engine(subdev);
+	struct nouveau_object *engctx;
+	struct nouveau_handle *handle = NULL;
+	struct nv50_graph_priv *priv = (void *)subdev;
+	u32 stat = nv_rd32(priv, 0x400100);
+	u32 inst = nv_rd32(priv, 0x40032c) & 0x0fffffff;
+	u32 addr = nv_rd32(priv, 0x400704);
+	u32 subc = (addr & 0x00070000) >> 16;
+	u32 mthd = (addr & 0x00001ffc);
+	u32 data = nv_rd32(priv, 0x400708);
+	u32 class = nv_rd32(priv, 0x400814);
+	u32 show = stat;
+	int chid;
+
+	engctx = nouveau_engctx_get(engine, inst);
+	chid   = pfifo->chid(pfifo, engctx);
+
+	if (stat & 0x00000010) {
+		handle = nouveau_handle_get_class(engctx, class);
+		if (handle && !nv_call(handle->object, mthd, data))
+			show &= ~0x00000010;
+		nouveau_handle_put(handle);
+	}
+
+	if (show & 0x00100000) {
+		u32 ecode = nv_rd32(priv, 0x400110);
+		nv_error(priv, "DATA_ERROR ");
+		nouveau_enum_print(nv50_data_error_names, ecode);
+		printk("\n");
+	}
+
+	if (stat & 0x00200000) {
+		if (!nv50_graph_trap_handler(priv, show, chid, (u64)inst << 12))
+			show &= ~0x00200000;
+	}
+
+	nv_wr32(priv, 0x400100, stat);
+	nv_wr32(priv, 0x400500, 0x00010001);
+
+	if (show) {
+		nv_info(priv, "");
+		nouveau_bitfield_print(nv50_graph_intr_name, show);
+		printk("\n");
+		nv_error(priv, "ch %d [0x%010llx] subc %d class 0x%04x "
+			       "mthd 0x%04x data 0x%08x\n",
+			 chid, (u64)inst << 12, subc, class, mthd, data);
+		nv50_fb_trap(nouveau_fb(priv), 1);
+	}
+
+	if (nv_rd32(priv, 0x400824) & (1 << 31))
+		nv_wr32(priv, 0x400824, nv_rd32(priv, 0x400824) & ~(1 << 31));
+
+	nouveau_engctx_put(engctx);
+}
+
+static int
+nv50_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv50_graph_priv *priv;
+	int ret;
+
+	ret = nouveau_graph_create(parent, engine, oclass, true, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00201000;
+	nv_subdev(priv)->intr = nv50_graph_intr;
+	nv_engine(priv)->cclass = &nv50_graph_cclass;
+
+	switch (nv_device(priv)->chipset) {
+	case 0x50:
+		nv_engine(priv)->sclass = nv50_graph_sclass;
+		break;
+	case 0x84:
+	case 0x86:
+	case 0x92:
+	case 0x94:
+	case 0x96:
+	case 0x98:
+		nv_engine(priv)->sclass = nv84_graph_sclass;
+		break;
+	case 0xa0:
+	case 0xaa:
+	case 0xac:
+		nv_engine(priv)->sclass = nva0_graph_sclass;
+		break;
+	case 0xa3:
+	case 0xa5:
+	case 0xa8:
+		nv_engine(priv)->sclass = nva3_graph_sclass;
+		break;
+	case 0xaf:
+		nv_engine(priv)->sclass = nvaf_graph_sclass;
+		break;
+
+	};
+
+	if (nv_device(priv)->chipset == 0x50 ||
+	    nv_device(priv)->chipset == 0xac)
+		nv_engine(priv)->tlb_flush = nv50_graph_tlb_flush;
+	else
+		nv_engine(priv)->tlb_flush = nv84_graph_tlb_flush;
+
+	spin_lock_init(&priv->lock);
+	return 0;
+}
+
+static int
+nv50_graph_init(struct nouveau_object *object)
+{
+	struct nv50_graph_priv *priv = (void *)object;
+	int ret, units, i;
+
+	ret = nouveau_graph_init(&priv->base);
+	if (ret)
+		return ret;
+
+	/* NV_PGRAPH_DEBUG_3_HW_CTX_SWITCH_ENABLED */
+	nv_wr32(priv, 0x40008c, 0x00000004);
+
+	/* reset/enable traps and interrupts */
+	nv_wr32(priv, 0x400804, 0xc0000000);
+	nv_wr32(priv, 0x406800, 0xc0000000);
+	nv_wr32(priv, 0x400c04, 0xc0000000);
+	nv_wr32(priv, 0x401800, 0xc0000000);
+	nv_wr32(priv, 0x405018, 0xc0000000);
+	nv_wr32(priv, 0x402000, 0xc0000000);
+
+	units = nv_rd32(priv, 0x001540);
+	for (i = 0; i < 16; i++) {
+		if (!(units & (1 << i)))
+			continue;
+
+		if (nv_device(priv)->chipset < 0xa0) {
+			nv_wr32(priv, 0x408900 + (i << 12), 0xc0000000);
+			nv_wr32(priv, 0x408e08 + (i << 12), 0xc0000000);
+			nv_wr32(priv, 0x408314 + (i << 12), 0xc0000000);
+		} else {
+			nv_wr32(priv, 0x408600 + (i << 11), 0xc0000000);
+			nv_wr32(priv, 0x408708 + (i << 11), 0xc0000000);
+			nv_wr32(priv, 0x40831c + (i << 11), 0xc0000000);
+		}
+	}
+
+	nv_wr32(priv, 0x400108, 0xffffffff);
+	nv_wr32(priv, 0x400138, 0xffffffff);
+	nv_wr32(priv, 0x400100, 0xffffffff);
+	nv_wr32(priv, 0x40013c, 0xffffffff);
+	nv_wr32(priv, 0x400500, 0x00010001);
+
+	/* upload context program, initialise ctxctl defaults */
+	ret = nv50_grctx_init(nv_device(priv), &priv->size);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, 0x400824, 0x00000000);
+	nv_wr32(priv, 0x400828, 0x00000000);
+	nv_wr32(priv, 0x40082c, 0x00000000);
+	nv_wr32(priv, 0x400830, 0x00000000);
+	nv_wr32(priv, 0x400724, 0x00000000);
+	nv_wr32(priv, 0x40032c, 0x00000000);
+	nv_wr32(priv, 0x400320, 4);	/* CTXCTL_CMD = NEWCTXDMA */
+
+	/* some unknown zcull magic */
+	switch (nv_device(priv)->chipset & 0xf0) {
+	case 0x50:
+	case 0x80:
+	case 0x90:
+		nv_wr32(priv, 0x402ca8, 0x00000800);
+		break;
+	case 0xa0:
+	default:
+		nv_wr32(priv, 0x402cc0, 0x00000000);
+		if (nv_device(priv)->chipset == 0xa0 ||
+		    nv_device(priv)->chipset == 0xaa ||
+		    nv_device(priv)->chipset == 0xac) {
+			nv_wr32(priv, 0x402ca8, 0x00000802);
+		} else {
+			nv_wr32(priv, 0x402cc0, 0x00000000);
+			nv_wr32(priv, 0x402ca8, 0x00000002);
+		}
+
+		break;
+	}
+
+	/* zero out zcull regions */
+	for (i = 0; i < 8; i++) {
+		nv_wr32(priv, 0x402c20 + (i * 8), 0x00000000);
+		nv_wr32(priv, 0x402c24 + (i * 8), 0x00000000);
+		nv_wr32(priv, 0x402c28 + (i * 8), 0x00000000);
+		nv_wr32(priv, 0x402c2c + (i * 8), 0x00000000);
+	}
+	return 0;
+}
+
+struct nouveau_oclass
+nv50_graph_oclass = {
+	.handle = NV_ENGINE(GR, 0x50),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv50_graph_ctor,
+		.dtor = _nouveau_graph_dtor,
+		.init = nv50_graph_init,
+		.fini = _nouveau_graph_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nv50.h b/drivers/gpu/drm/nouveau/core/engine/graph/nv50.h
new file mode 100644
index 0000000..0505fb4
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nv50.h
@@ -0,0 +1,7 @@
+#ifndef __NV50_GRAPH_H__
+#define __NV50_GRAPH_H__
+
+int  nv50_grctx_init(struct nouveau_device *, u32 *size);
+void nv50_grctx_fill(struct nouveau_device *, struct nouveau_gpuobj *);
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.c b/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.c
new file mode 100644
index 0000000..c62f2d0
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.c
@@ -0,0 +1,955 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include "nvc0.h"
+#include "fuc/hubnvc0.fuc.h"
+#include "fuc/gpcnvc0.fuc.h"
+
+/*******************************************************************************
+ * Graphics object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nvc0_graph_sclass[] = {
+	{ 0x902d, &nouveau_object_ofuncs },
+	{ 0x9039, &nouveau_object_ofuncs },
+	{ 0x9097, &nouveau_object_ofuncs },
+	{ 0x90c0, &nouveau_object_ofuncs },
+	{}
+};
+
+static struct nouveau_oclass
+nvc1_graph_sclass[] = {
+	{ 0x902d, &nouveau_object_ofuncs },
+	{ 0x9039, &nouveau_object_ofuncs },
+	{ 0x9097, &nouveau_object_ofuncs },
+	{ 0x90c0, &nouveau_object_ofuncs },
+	{ 0x9197, &nouveau_object_ofuncs },
+	{}
+};
+
+static struct nouveau_oclass
+nvc8_graph_sclass[] = {
+	{ 0x902d, &nouveau_object_ofuncs },
+	{ 0x9039, &nouveau_object_ofuncs },
+	{ 0x9097, &nouveau_object_ofuncs },
+	{ 0x90c0, &nouveau_object_ofuncs },
+	{ 0x9197, &nouveau_object_ofuncs },
+	{ 0x9297, &nouveau_object_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * PGRAPH context
+ ******************************************************************************/
+
+int
+nvc0_graph_context_ctor(struct nouveau_object *parent,
+			struct nouveau_object *engine,
+			struct nouveau_oclass *oclass, void *args, u32 size,
+			struct nouveau_object **pobject)
+{
+	struct nouveau_vm *vm = nouveau_client(parent)->vm;
+	struct nvc0_graph_priv *priv = (void *)engine;
+	struct nvc0_graph_data *data = priv->mmio_data;
+	struct nvc0_graph_mmio *mmio = priv->mmio_list;
+	struct nvc0_graph_chan *chan;
+	int ret, i;
+
+	/* allocate memory for context, and fill with default values */
+	ret = nouveau_graph_context_create(parent, engine, oclass, NULL,
+					   priv->size, 0x100,
+					   NVOBJ_FLAG_ZERO_ALLOC, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	/* allocate memory for a "mmio list" buffer that's used by the HUB
+	 * fuc to modify some per-context register settings on first load
+	 * of the context.
+	 */
+	ret = nouveau_gpuobj_new(parent, NULL, 0x1000, 0x100, 0, &chan->mmio);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_map_vm(nv_gpuobj(chan->mmio), vm,
+				    NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS,
+				    &chan->mmio_vma);
+	if (ret)
+		return ret;
+
+	/* allocate buffers referenced by mmio list */
+	for (i = 0; data->size && i < ARRAY_SIZE(priv->mmio_data); i++) {
+		ret = nouveau_gpuobj_new(parent, NULL, data->size, data->align,
+					 0, &chan->data[i].mem);
+		if (ret)
+			return ret;
+
+		ret = nouveau_gpuobj_map_vm(chan->data[i].mem, vm, data->access,
+					   &chan->data[i].vma);
+		if (ret)
+			return ret;
+
+		data++;
+	}
+
+	/* finally, fill in the mmio list and point the context at it */
+	for (i = 0; mmio->addr && i < ARRAY_SIZE(priv->mmio_list); i++) {
+		u32 addr = mmio->addr;
+		u32 data = mmio->data;
+
+		if (mmio->shift) {
+			u64 info = chan->data[mmio->buffer].vma.offset;
+			data |= info >> mmio->shift;
+		}
+
+		nv_wo32(chan->mmio, chan->mmio_nr++ * 4, addr);
+		nv_wo32(chan->mmio, chan->mmio_nr++ * 4, data);
+		mmio++;
+	}
+
+	for (i = 0; i < priv->size; i += 4)
+		nv_wo32(chan, i, priv->data[i / 4]);
+
+	if (!priv->firmware) {
+		nv_wo32(chan, 0x00, chan->mmio_nr / 2);
+		nv_wo32(chan, 0x04, chan->mmio_vma.offset >> 8);
+	} else {
+		nv_wo32(chan, 0xf4, 0);
+		nv_wo32(chan, 0xf8, 0);
+		nv_wo32(chan, 0x10, chan->mmio_nr / 2);
+		nv_wo32(chan, 0x14, lower_32_bits(chan->mmio_vma.offset));
+		nv_wo32(chan, 0x18, upper_32_bits(chan->mmio_vma.offset));
+		nv_wo32(chan, 0x1c, 1);
+		nv_wo32(chan, 0x20, 0);
+		nv_wo32(chan, 0x28, 0);
+		nv_wo32(chan, 0x2c, 0);
+	}
+
+	return 0;
+}
+
+void
+nvc0_graph_context_dtor(struct nouveau_object *object)
+{
+	struct nvc0_graph_chan *chan = (void *)object;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(chan->data); i++) {
+		nouveau_gpuobj_unmap(&chan->data[i].vma);
+		nouveau_gpuobj_ref(NULL, &chan->data[i].mem);
+	}
+
+	nouveau_gpuobj_unmap(&chan->mmio_vma);
+	nouveau_gpuobj_ref(NULL, &chan->mmio);
+
+	nouveau_graph_context_destroy(&chan->base);
+}
+
+static struct nouveau_oclass
+nvc0_graph_cclass = {
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvc0_graph_context_ctor,
+		.dtor = nvc0_graph_context_dtor,
+		.init = _nouveau_graph_context_init,
+		.fini = _nouveau_graph_context_fini,
+		.rd32 = _nouveau_graph_context_rd32,
+		.wr32 = _nouveau_graph_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PGRAPH engine/subdev functions
+ ******************************************************************************/
+
+static void
+nvc0_graph_ctxctl_debug_unit(struct nvc0_graph_priv *priv, u32 base)
+{
+	nv_error(priv, "%06x - done 0x%08x\n", base,
+		 nv_rd32(priv, base + 0x400));
+	nv_error(priv, "%06x - stat 0x%08x 0x%08x 0x%08x 0x%08x\n", base,
+		 nv_rd32(priv, base + 0x800), nv_rd32(priv, base + 0x804),
+		 nv_rd32(priv, base + 0x808), nv_rd32(priv, base + 0x80c));
+	nv_error(priv, "%06x - stat 0x%08x 0x%08x 0x%08x 0x%08x\n", base,
+		 nv_rd32(priv, base + 0x810), nv_rd32(priv, base + 0x814),
+		 nv_rd32(priv, base + 0x818), nv_rd32(priv, base + 0x81c));
+}
+
+void
+nvc0_graph_ctxctl_debug(struct nvc0_graph_priv *priv)
+{
+	u32 gpcnr = nv_rd32(priv, 0x409604) & 0xffff;
+	u32 gpc;
+
+	nvc0_graph_ctxctl_debug_unit(priv, 0x409000);
+	for (gpc = 0; gpc < gpcnr; gpc++)
+		nvc0_graph_ctxctl_debug_unit(priv, 0x502000 + (gpc * 0x8000));
+}
+
+static void
+nvc0_graph_ctxctl_isr(struct nvc0_graph_priv *priv)
+{
+	u32 ustat = nv_rd32(priv, 0x409c18);
+
+	if (ustat & 0x00000001)
+		nv_error(priv, "CTXCTRL ucode error\n");
+	if (ustat & 0x00080000)
+		nv_error(priv, "CTXCTRL watchdog timeout\n");
+	if (ustat & ~0x00080001)
+		nv_error(priv, "CTXCTRL 0x%08x\n", ustat);
+
+	nvc0_graph_ctxctl_debug(priv);
+	nv_wr32(priv, 0x409c20, ustat);
+}
+
+static void
+nvc0_graph_trap_tpc(struct nvc0_graph_priv *priv, int gpc, int tpc)
+{
+	u32 stat = nv_rd32(priv, TPC_UNIT(gpc, tpc, 0x0508));
+
+	if (stat & 0x00000001) {
+		u32 trap = nv_rd32(priv, TPC_UNIT(gpc, tpc, 0x0224));
+		nv_error(priv, "GPC%d/TPC%d/TEX: 0x%08x\n", gpc, tpc, trap);
+		nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x0224), 0xc0000000);
+		nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x0508), 0x00000001);
+		stat &= ~0x00000001;
+	}
+
+	if (stat & 0x00000002) {
+		u32 trap0 = nv_rd32(priv, TPC_UNIT(gpc, tpc, 0x0644));
+		u32 trap1 = nv_rd32(priv, TPC_UNIT(gpc, tpc, 0x064c));
+		nv_error(priv, "GPC%d/TPC%d/MP: 0x%08x 0x%08x\n",
+			       gpc, tpc, trap0, trap1);
+		nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x0644), 0x001ffffe);
+		nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x064c), 0x0000000f);
+		nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x0508), 0x00000002);
+		stat &= ~0x00000002;
+	}
+
+	if (stat & 0x00000004) {
+		u32 trap = nv_rd32(priv, TPC_UNIT(gpc, tpc, 0x0084));
+		nv_error(priv, "GPC%d/TPC%d/POLY: 0x%08x\n", gpc, tpc, trap);
+		nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x0084), 0xc0000000);
+		nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x0508), 0x00000004);
+		stat &= ~0x00000004;
+	}
+
+	if (stat & 0x00000008) {
+		u32 trap = nv_rd32(priv, TPC_UNIT(gpc, tpc, 0x048c));
+		nv_error(priv, "GPC%d/TPC%d/L1C: 0x%08x\n", gpc, tpc, trap);
+		nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x048c), 0xc0000000);
+		nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x0508), 0x00000008);
+		stat &= ~0x00000008;
+	}
+
+	if (stat) {
+		nv_error(priv, "GPC%d/TPC%d/0x%08x: unknown\n", gpc, tpc, stat);
+		nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x0508), stat);
+	}
+}
+
+static void
+nvc0_graph_trap_gpc(struct nvc0_graph_priv *priv, int gpc)
+{
+	u32 stat = nv_rd32(priv, GPC_UNIT(gpc, 0x2c90));
+	int tpc;
+
+	if (stat & 0x00000001) {
+		u32 trap = nv_rd32(priv, GPC_UNIT(gpc, 0x0420));
+		nv_error(priv, "GPC%d/PROP: 0x%08x\n", gpc, trap);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0420), 0xc0000000);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), 0x00000001);
+		stat &= ~0x00000001;
+	}
+
+	if (stat & 0x00000002) {
+		u32 trap = nv_rd32(priv, GPC_UNIT(gpc, 0x0900));
+		nv_error(priv, "GPC%d/ZCULL: 0x%08x\n", gpc, trap);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0900), 0xc0000000);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), 0x00000002);
+		stat &= ~0x00000002;
+	}
+
+	if (stat & 0x00000004) {
+		u32 trap = nv_rd32(priv, GPC_UNIT(gpc, 0x1028));
+		nv_error(priv, "GPC%d/CCACHE: 0x%08x\n", gpc, trap);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x1028), 0xc0000000);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), 0x00000004);
+		stat &= ~0x00000004;
+	}
+
+	if (stat & 0x00000008) {
+		u32 trap = nv_rd32(priv, GPC_UNIT(gpc, 0x0824));
+		nv_error(priv, "GPC%d/ESETUP: 0x%08x\n", gpc, trap);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0824), 0xc0000000);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), 0x00000008);
+		stat &= ~0x00000009;
+	}
+
+	for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) {
+		u32 mask = 0x00010000 << tpc;
+		if (stat & mask) {
+			nvc0_graph_trap_tpc(priv, gpc, tpc);
+			nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), mask);
+			stat &= ~mask;
+		}
+	}
+
+	if (stat) {
+		nv_error(priv, "GPC%d/0x%08x: unknown\n", gpc, stat);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), stat);
+	}
+}
+
+static void
+nvc0_graph_trap_intr(struct nvc0_graph_priv *priv)
+{
+	u32 trap = nv_rd32(priv, 0x400108);
+	int rop, gpc;
+
+	if (trap & 0x00000001) {
+		u32 stat = nv_rd32(priv, 0x404000);
+		nv_error(priv, "DISPATCH 0x%08x\n", stat);
+		nv_wr32(priv, 0x404000, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x00000001);
+		trap &= ~0x00000001;
+	}
+
+	if (trap & 0x00000002) {
+		u32 stat = nv_rd32(priv, 0x404600);
+		nv_error(priv, "M2MF 0x%08x\n", stat);
+		nv_wr32(priv, 0x404600, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x00000002);
+		trap &= ~0x00000002;
+	}
+
+	if (trap & 0x00000008) {
+		u32 stat = nv_rd32(priv, 0x408030);
+		nv_error(priv, "CCACHE 0x%08x\n", stat);
+		nv_wr32(priv, 0x408030, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x00000008);
+		trap &= ~0x00000008;
+	}
+
+	if (trap & 0x00000010) {
+		u32 stat = nv_rd32(priv, 0x405840);
+		nv_error(priv, "SHADER 0x%08x\n", stat);
+		nv_wr32(priv, 0x405840, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x00000010);
+		trap &= ~0x00000010;
+	}
+
+	if (trap & 0x00000040) {
+		u32 stat = nv_rd32(priv, 0x40601c);
+		nv_error(priv, "UNK6 0x%08x\n", stat);
+		nv_wr32(priv, 0x40601c, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x00000040);
+		trap &= ~0x00000040;
+	}
+
+	if (trap & 0x00000080) {
+		u32 stat = nv_rd32(priv, 0x404490);
+		nv_error(priv, "MACRO 0x%08x\n", stat);
+		nv_wr32(priv, 0x404490, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x00000080);
+		trap &= ~0x00000080;
+	}
+
+	if (trap & 0x01000000) {
+		u32 stat = nv_rd32(priv, 0x400118);
+		for (gpc = 0; stat && gpc < priv->gpc_nr; gpc++) {
+			u32 mask = 0x00000001 << gpc;
+			if (stat & mask) {
+				nvc0_graph_trap_gpc(priv, gpc);
+				nv_wr32(priv, 0x400118, mask);
+				stat &= ~mask;
+			}
+		}
+		nv_wr32(priv, 0x400108, 0x01000000);
+		trap &= ~0x01000000;
+	}
+
+	if (trap & 0x02000000) {
+		for (rop = 0; rop < priv->rop_nr; rop++) {
+			u32 statz = nv_rd32(priv, ROP_UNIT(rop, 0x070));
+			u32 statc = nv_rd32(priv, ROP_UNIT(rop, 0x144));
+			nv_error(priv, "ROP%d 0x%08x 0x%08x\n",
+				 rop, statz, statc);
+			nv_wr32(priv, ROP_UNIT(rop, 0x070), 0xc0000000);
+			nv_wr32(priv, ROP_UNIT(rop, 0x144), 0xc0000000);
+		}
+		nv_wr32(priv, 0x400108, 0x02000000);
+		trap &= ~0x02000000;
+	}
+
+	if (trap) {
+		nv_error(priv, "TRAP UNHANDLED 0x%08x\n", trap);
+		nv_wr32(priv, 0x400108, trap);
+	}
+}
+
+static void
+nvc0_graph_intr(struct nouveau_subdev *subdev)
+{
+	struct nouveau_fifo *pfifo = nouveau_fifo(subdev);
+	struct nouveau_engine *engine = nv_engine(subdev);
+	struct nouveau_object *engctx;
+	struct nouveau_handle *handle;
+	struct nvc0_graph_priv *priv = (void *)subdev;
+	u64 inst = nv_rd32(priv, 0x409b00) & 0x0fffffff;
+	u32 stat = nv_rd32(priv, 0x400100);
+	u32 addr = nv_rd32(priv, 0x400704);
+	u32 mthd = (addr & 0x00003ffc);
+	u32 subc = (addr & 0x00070000) >> 16;
+	u32 data = nv_rd32(priv, 0x400708);
+	u32 code = nv_rd32(priv, 0x400110);
+	u32 class = nv_rd32(priv, 0x404200 + (subc * 4));
+	int chid;
+
+	engctx = nouveau_engctx_get(engine, inst);
+	chid   = pfifo->chid(pfifo, engctx);
+
+	if (stat & 0x00000010) {
+		handle = nouveau_handle_get_class(engctx, class);
+		if (!handle || nv_call(handle->object, mthd, data)) {
+			nv_error(priv, "ILLEGAL_MTHD ch %d [0x%010llx] "
+				     "subc %d class 0x%04x mthd 0x%04x "
+				     "data 0x%08x\n",
+				 chid, inst << 12, subc, class, mthd, data);
+		}
+		nouveau_handle_put(handle);
+		nv_wr32(priv, 0x400100, 0x00000010);
+		stat &= ~0x00000010;
+	}
+
+	if (stat & 0x00000020) {
+		nv_error(priv, "ILLEGAL_CLASS ch %d [0x%010llx] subc %d "
+			     "class 0x%04x mthd 0x%04x data 0x%08x\n",
+			chid, inst << 12, subc, class, mthd, data);
+		nv_wr32(priv, 0x400100, 0x00000020);
+		stat &= ~0x00000020;
+	}
+
+	if (stat & 0x00100000) {
+		nv_error(priv, "DATA_ERROR [");
+		nouveau_enum_print(nv50_data_error_names, code);
+		printk("] ch %d [0x%010llx] subc %d class 0x%04x "
+		       "mthd 0x%04x data 0x%08x\n",
+		       chid, inst << 12, subc, class, mthd, data);
+		nv_wr32(priv, 0x400100, 0x00100000);
+		stat &= ~0x00100000;
+	}
+
+	if (stat & 0x00200000) {
+		nv_error(priv, "TRAP ch %d [0x%010llx]\n", chid, inst << 12);
+		nvc0_graph_trap_intr(priv);
+		nv_wr32(priv, 0x400100, 0x00200000);
+		stat &= ~0x00200000;
+	}
+
+	if (stat & 0x00080000) {
+		nvc0_graph_ctxctl_isr(priv);
+		nv_wr32(priv, 0x400100, 0x00080000);
+		stat &= ~0x00080000;
+	}
+
+	if (stat) {
+		nv_error(priv, "unknown stat 0x%08x\n", stat);
+		nv_wr32(priv, 0x400100, stat);
+	}
+
+	nv_wr32(priv, 0x400500, 0x00010001);
+	nouveau_engctx_put(engctx);
+}
+
+int
+nvc0_graph_ctor_fw(struct nvc0_graph_priv *priv, const char *fwname,
+		   struct nvc0_graph_fuc *fuc)
+{
+	struct nouveau_device *device = nv_device(priv);
+	const struct firmware *fw;
+	char f[32];
+	int ret;
+
+	snprintf(f, sizeof(f), "nouveau/nv%02x_%s", device->chipset, fwname);
+	ret = request_firmware(&fw, f, &device->pdev->dev);
+	if (ret) {
+		snprintf(f, sizeof(f), "nouveau/%s", fwname);
+		ret = request_firmware(&fw, f, &device->pdev->dev);
+		if (ret) {
+			nv_error(priv, "failed to load %s\n", fwname);
+			return ret;
+		}
+	}
+
+	fuc->size = fw->size;
+	fuc->data = kmemdup(fw->data, fuc->size, GFP_KERNEL);
+	release_firmware(fw);
+	return (fuc->data != NULL) ? 0 : -ENOMEM;
+}
+
+static int
+nvc0_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		struct nouveau_oclass *oclass, void *data, u32 size,
+		struct nouveau_object **pobject)
+{
+	struct nouveau_device *device = nv_device(parent);
+	struct nvc0_graph_priv *priv;
+	bool enable = true;
+	int ret, i;
+
+	switch (device->chipset) {
+	case 0xd9: /* known broken without binary driver firmware */
+		enable = false;
+		break;
+	default:
+		break;
+	}
+
+	ret = nouveau_graph_create(parent, engine, oclass, enable, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x18001000;
+	nv_subdev(priv)->intr = nvc0_graph_intr;
+	nv_engine(priv)->cclass = &nvc0_graph_cclass;
+
+	if (nouveau_boolopt(device->cfgopt, "NvGrUseFW", false)) {
+		nv_info(priv, "using external firmware\n");
+		if (nvc0_graph_ctor_fw(priv, "fuc409c", &priv->fuc409c) ||
+		    nvc0_graph_ctor_fw(priv, "fuc409d", &priv->fuc409d) ||
+		    nvc0_graph_ctor_fw(priv, "fuc41ac", &priv->fuc41ac) ||
+		    nvc0_graph_ctor_fw(priv, "fuc41ad", &priv->fuc41ad))
+			return -EINVAL;
+		priv->firmware = true;
+	}
+
+	switch (nvc0_graph_class(priv)) {
+	case 0x9097:
+		nv_engine(priv)->sclass = nvc0_graph_sclass;
+		break;
+	case 0x9197:
+		nv_engine(priv)->sclass = nvc1_graph_sclass;
+		break;
+	case 0x9297:
+		nv_engine(priv)->sclass = nvc8_graph_sclass;
+		break;
+	}
+
+	ret = nouveau_gpuobj_new(parent, NULL, 0x1000, 256, 0, &priv->unk4188b4);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 0x1000, 256, 0, &priv->unk4188b8);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < 0x1000; i += 4) {
+		nv_wo32(priv->unk4188b4, i, 0x00000010);
+		nv_wo32(priv->unk4188b8, i, 0x00000010);
+	}
+
+	priv->rop_nr = (nv_rd32(priv, 0x409604) & 0x001f0000) >> 16;
+	priv->gpc_nr =  nv_rd32(priv, 0x409604) & 0x0000001f;
+	for (i = 0; i < priv->gpc_nr; i++) {
+		priv->tpc_nr[i]  = nv_rd32(priv, GPC_UNIT(i, 0x2608));
+		priv->tpc_total += priv->tpc_nr[i];
+	}
+
+	/*XXX: these need figuring out... though it might not even matter */
+	switch (nv_device(priv)->chipset) {
+	case 0xc0:
+		if (priv->tpc_total == 11) { /* 465, 3/4/4/0, 4 */
+			priv->magic_not_rop_nr = 0x07;
+		} else
+		if (priv->tpc_total == 14) { /* 470, 3/3/4/4, 5 */
+			priv->magic_not_rop_nr = 0x05;
+		} else
+		if (priv->tpc_total == 15) { /* 480, 3/4/4/4, 6 */
+			priv->magic_not_rop_nr = 0x06;
+		}
+		break;
+	case 0xc3: /* 450, 4/0/0/0, 2 */
+		priv->magic_not_rop_nr = 0x03;
+		break;
+	case 0xc4: /* 460, 3/4/0/0, 4 */
+		priv->magic_not_rop_nr = 0x01;
+		break;
+	case 0xc1: /* 2/0/0/0, 1 */
+		priv->magic_not_rop_nr = 0x01;
+		break;
+	case 0xc8: /* 4/4/3/4, 5 */
+		priv->magic_not_rop_nr = 0x06;
+		break;
+	case 0xce: /* 4/4/0/0, 4 */
+		priv->magic_not_rop_nr = 0x03;
+		break;
+	case 0xcf: /* 4/0/0/0, 3 */
+		priv->magic_not_rop_nr = 0x03;
+		break;
+	case 0xd9: /* 1/0/0/0, 1 */
+		priv->magic_not_rop_nr = 0x01;
+		break;
+	}
+
+	return 0;
+}
+
+static void
+nvc0_graph_dtor_fw(struct nvc0_graph_fuc *fuc)
+{
+	if (fuc->data) {
+		kfree(fuc->data);
+		fuc->data = NULL;
+	}
+}
+
+void
+nvc0_graph_dtor(struct nouveau_object *object)
+{
+	struct nvc0_graph_priv *priv = (void *)object;
+
+	if (priv->data)
+		kfree(priv->data);
+
+	nvc0_graph_dtor_fw(&priv->fuc409c);
+	nvc0_graph_dtor_fw(&priv->fuc409d);
+	nvc0_graph_dtor_fw(&priv->fuc41ac);
+	nvc0_graph_dtor_fw(&priv->fuc41ad);
+
+	nouveau_gpuobj_ref(NULL, &priv->unk4188b8);
+	nouveau_gpuobj_ref(NULL, &priv->unk4188b4);
+
+	nouveau_graph_destroy(&priv->base);
+}
+
+static void
+nvc0_graph_init_obj418880(struct nvc0_graph_priv *priv)
+{
+	int i;
+
+	nv_wr32(priv, GPC_BCAST(0x0880), 0x00000000);
+	nv_wr32(priv, GPC_BCAST(0x08a4), 0x00000000);
+	for (i = 0; i < 4; i++)
+		nv_wr32(priv, GPC_BCAST(0x0888) + (i * 4), 0x00000000);
+	nv_wr32(priv, GPC_BCAST(0x08b4), priv->unk4188b4->addr >> 8);
+	nv_wr32(priv, GPC_BCAST(0x08b8), priv->unk4188b8->addr >> 8);
+}
+
+static void
+nvc0_graph_init_regs(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x400080, 0x003083c2);
+	nv_wr32(priv, 0x400088, 0x00006fe7);
+	nv_wr32(priv, 0x40008c, 0x00000000);
+	nv_wr32(priv, 0x400090, 0x00000030);
+	nv_wr32(priv, 0x40013c, 0x013901f7);
+	nv_wr32(priv, 0x400140, 0x00000100);
+	nv_wr32(priv, 0x400144, 0x00000000);
+	nv_wr32(priv, 0x400148, 0x00000110);
+	nv_wr32(priv, 0x400138, 0x00000000);
+	nv_wr32(priv, 0x400130, 0x00000000);
+	nv_wr32(priv, 0x400134, 0x00000000);
+	nv_wr32(priv, 0x400124, 0x00000002);
+}
+
+static void
+nvc0_graph_init_gpc_0(struct nvc0_graph_priv *priv)
+{
+	const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, priv->tpc_total);
+	u32 data[TPC_MAX / 8];
+	u8  tpcnr[GPC_MAX];
+	int i, gpc, tpc;
+
+	nv_wr32(priv, TPC_UNIT(0, 0, 0x5c), 1); /* affects TFB offset queries */
+
+	/*
+	 *      TP      ROP UNKVAL(magic_not_rop_nr)
+	 * 450: 4/0/0/0 2        3
+	 * 460: 3/4/0/0 4        1
+	 * 465: 3/4/4/0 4        7
+	 * 470: 3/3/4/4 5        5
+	 * 480: 3/4/4/4 6        6
+	 */
+
+	memset(data, 0x00, sizeof(data));
+	memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr));
+	for (i = 0, gpc = -1; i < priv->tpc_total; i++) {
+		do {
+			gpc = (gpc + 1) % priv->gpc_nr;
+		} while (!tpcnr[gpc]);
+		tpc = priv->tpc_nr[gpc] - tpcnr[gpc]--;
+
+		data[i / 8] |= tpc << ((i % 8) * 4);
+	}
+
+	nv_wr32(priv, GPC_BCAST(0x0980), data[0]);
+	nv_wr32(priv, GPC_BCAST(0x0984), data[1]);
+	nv_wr32(priv, GPC_BCAST(0x0988), data[2]);
+	nv_wr32(priv, GPC_BCAST(0x098c), data[3]);
+
+	for (gpc = 0; gpc < priv->gpc_nr; gpc++) {
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0914), priv->magic_not_rop_nr << 8 |
+						  priv->tpc_nr[gpc]);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0910), 0x00040000 | priv->tpc_total);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0918), magicgpc918);
+	}
+
+	nv_wr32(priv, GPC_BCAST(0x1bd4), magicgpc918);
+	nv_wr32(priv, GPC_BCAST(0x08ac), nv_rd32(priv, 0x100800));
+}
+
+static void
+nvc0_graph_init_units(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x409c24, 0x000f0000);
+	nv_wr32(priv, 0x404000, 0xc0000000); /* DISPATCH */
+	nv_wr32(priv, 0x404600, 0xc0000000); /* M2MF */
+	nv_wr32(priv, 0x408030, 0xc0000000);
+	nv_wr32(priv, 0x40601c, 0xc0000000);
+	nv_wr32(priv, 0x404490, 0xc0000000); /* MACRO */
+	nv_wr32(priv, 0x406018, 0xc0000000);
+	nv_wr32(priv, 0x405840, 0xc0000000);
+	nv_wr32(priv, 0x405844, 0x00ffffff);
+	nv_mask(priv, 0x419cc0, 0x00000008, 0x00000008);
+	nv_mask(priv, 0x419eb4, 0x00001000, 0x00001000);
+}
+
+static void
+nvc0_graph_init_gpc_1(struct nvc0_graph_priv *priv)
+{
+	int gpc, tpc;
+
+	for (gpc = 0; gpc < priv->gpc_nr; gpc++) {
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0420), 0xc0000000);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0900), 0xc0000000);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x1028), 0xc0000000);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0824), 0xc0000000);
+		for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) {
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff);
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff);
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000);
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000);
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000);
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x644), 0x001ffffe);
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x64c), 0x0000000f);
+		}
+		nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), 0xffffffff);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x2c94), 0xffffffff);
+	}
+}
+
+static void
+nvc0_graph_init_rop(struct nvc0_graph_priv *priv)
+{
+	int rop;
+
+	for (rop = 0; rop < priv->rop_nr; rop++) {
+		nv_wr32(priv, ROP_UNIT(rop, 0x144), 0xc0000000);
+		nv_wr32(priv, ROP_UNIT(rop, 0x070), 0xc0000000);
+		nv_wr32(priv, ROP_UNIT(rop, 0x204), 0xffffffff);
+		nv_wr32(priv, ROP_UNIT(rop, 0x208), 0xffffffff);
+	}
+}
+
+void
+nvc0_graph_init_fw(struct nvc0_graph_priv *priv, u32 fuc_base,
+		   struct nvc0_graph_fuc *code, struct nvc0_graph_fuc *data)
+{
+	int i;
+
+	nv_wr32(priv, fuc_base + 0x01c0, 0x01000000);
+	for (i = 0; i < data->size / 4; i++)
+		nv_wr32(priv, fuc_base + 0x01c4, data->data[i]);
+
+	nv_wr32(priv, fuc_base + 0x0180, 0x01000000);
+	for (i = 0; i < code->size / 4; i++) {
+		if ((i & 0x3f) == 0)
+			nv_wr32(priv, fuc_base + 0x0188, i >> 6);
+		nv_wr32(priv, fuc_base + 0x0184, code->data[i]);
+	}
+}
+
+static int
+nvc0_graph_init_ctxctl(struct nvc0_graph_priv *priv)
+{
+	u32 r000260;
+	int i;
+
+	if (priv->firmware) {
+		/* load fuc microcode */
+		r000260 = nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
+		nvc0_graph_init_fw(priv, 0x409000, &priv->fuc409c,
+						   &priv->fuc409d);
+		nvc0_graph_init_fw(priv, 0x41a000, &priv->fuc41ac,
+						   &priv->fuc41ad);
+		nv_wr32(priv, 0x000260, r000260);
+
+		/* start both of them running */
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x41a10c, 0x00000000);
+		nv_wr32(priv, 0x40910c, 0x00000000);
+		nv_wr32(priv, 0x41a100, 0x00000002);
+		nv_wr32(priv, 0x409100, 0x00000002);
+		if (!nv_wait(priv, 0x409800, 0x00000001, 0x00000001))
+			nv_info(priv, "0x409800 wait failed\n");
+
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x409500, 0x7fffffff);
+		nv_wr32(priv, 0x409504, 0x00000021);
+
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x409500, 0x00000000);
+		nv_wr32(priv, 0x409504, 0x00000010);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x10 timeout\n");
+			return -EBUSY;
+		}
+		priv->size = nv_rd32(priv, 0x409800);
+
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x409500, 0x00000000);
+		nv_wr32(priv, 0x409504, 0x00000016);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x16 timeout\n");
+			return -EBUSY;
+		}
+
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x409500, 0x00000000);
+		nv_wr32(priv, 0x409504, 0x00000025);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x25 timeout\n");
+			return -EBUSY;
+		}
+
+		if (priv->data == NULL) {
+			int ret = nvc0_grctx_generate(priv);
+			if (ret) {
+				nv_error(priv, "failed to construct context\n");
+				return ret;
+			}
+		}
+
+		return 0;
+	}
+
+	/* load HUB microcode */
+	r000260 = nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
+	nv_wr32(priv, 0x4091c0, 0x01000000);
+	for (i = 0; i < sizeof(nvc0_grhub_data) / 4; i++)
+		nv_wr32(priv, 0x4091c4, nvc0_grhub_data[i]);
+
+	nv_wr32(priv, 0x409180, 0x01000000);
+	for (i = 0; i < sizeof(nvc0_grhub_code) / 4; i++) {
+		if ((i & 0x3f) == 0)
+			nv_wr32(priv, 0x409188, i >> 6);
+		nv_wr32(priv, 0x409184, nvc0_grhub_code[i]);
+	}
+
+	/* load GPC microcode */
+	nv_wr32(priv, 0x41a1c0, 0x01000000);
+	for (i = 0; i < sizeof(nvc0_grgpc_data) / 4; i++)
+		nv_wr32(priv, 0x41a1c4, nvc0_grgpc_data[i]);
+
+	nv_wr32(priv, 0x41a180, 0x01000000);
+	for (i = 0; i < sizeof(nvc0_grgpc_code) / 4; i++) {
+		if ((i & 0x3f) == 0)
+			nv_wr32(priv, 0x41a188, i >> 6);
+		nv_wr32(priv, 0x41a184, nvc0_grgpc_code[i]);
+	}
+	nv_wr32(priv, 0x000260, r000260);
+
+	/* start HUB ucode running, it'll init the GPCs */
+	nv_wr32(priv, 0x409800, nv_device(priv)->chipset);
+	nv_wr32(priv, 0x40910c, 0x00000000);
+	nv_wr32(priv, 0x409100, 0x00000002);
+	if (!nv_wait(priv, 0x409800, 0x80000000, 0x80000000)) {
+		nv_error(priv, "HUB_INIT timed out\n");
+		nvc0_graph_ctxctl_debug(priv);
+		return -EBUSY;
+	}
+
+	priv->size = nv_rd32(priv, 0x409804);
+	if (priv->data == NULL) {
+		int ret = nvc0_grctx_generate(priv);
+		if (ret) {
+			nv_error(priv, "failed to construct context\n");
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
+static int
+nvc0_graph_init(struct nouveau_object *object)
+{
+	struct nvc0_graph_priv *priv = (void *)object;
+	int ret;
+
+	ret = nouveau_graph_init(&priv->base);
+	if (ret)
+		return ret;
+
+	nvc0_graph_init_obj418880(priv);
+	nvc0_graph_init_regs(priv);
+	/*nvc0_graph_init_unitplemented_magics(priv);*/
+	nvc0_graph_init_gpc_0(priv);
+	/*nvc0_graph_init_unitplemented_c242(priv);*/
+
+	nv_wr32(priv, 0x400500, 0x00010001);
+	nv_wr32(priv, 0x400100, 0xffffffff);
+	nv_wr32(priv, 0x40013c, 0xffffffff);
+
+	nvc0_graph_init_units(priv);
+	nvc0_graph_init_gpc_1(priv);
+	nvc0_graph_init_rop(priv);
+
+	nv_wr32(priv, 0x400108, 0xffffffff);
+	nv_wr32(priv, 0x400138, 0xffffffff);
+	nv_wr32(priv, 0x400118, 0xffffffff);
+	nv_wr32(priv, 0x400130, 0xffffffff);
+	nv_wr32(priv, 0x40011c, 0xffffffff);
+	nv_wr32(priv, 0x400134, 0xffffffff);
+	nv_wr32(priv, 0x400054, 0x34ce3464);
+
+	ret = nvc0_graph_init_ctxctl(priv);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+struct nouveau_oclass
+nvc0_graph_oclass = {
+	.handle = NV_ENGINE(GR, 0xc0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvc0_graph_ctor,
+		.dtor = nvc0_graph_dtor,
+		.init = nvc0_graph_init,
+		.fini = _nouveau_graph_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.h b/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.h
new file mode 100644
index 0000000..18d2210
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.h
@@ -0,0 +1,171 @@
+/*
+ * Copyright 2010 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#ifndef __NVC0_GRAPH_H__
+#define __NVC0_GRAPH_H__
+
+#include <core/client.h>
+#include <core/handle.h>
+#include <core/gpuobj.h>
+#include <core/option.h>
+
+#include <subdev/fb.h>
+#include <subdev/vm.h>
+#include <subdev/bar.h>
+#include <subdev/timer.h>
+
+#include <engine/fifo.h>
+#include <engine/graph.h>
+
+#define GPC_MAX 4
+#define TPC_MAX 32
+
+#define ROP_BCAST(r)      (0x408800 + (r))
+#define ROP_UNIT(u, r)    (0x410000 + (u) * 0x400 + (r))
+#define GPC_BCAST(r)      (0x418000 + (r))
+#define GPC_UNIT(t, r)    (0x500000 + (t) * 0x8000 + (r))
+#define TPC_UNIT(t, m, r) (0x504000 + (t) * 0x8000 + (m) * 0x800 + (r))
+
+struct nvc0_graph_data {
+	u32 size;
+	u32 align;
+	u32 access;
+};
+
+struct nvc0_graph_mmio {
+	u32 addr;
+	u32 data;
+	u32 shift;
+	u32 buffer;
+};
+
+struct nvc0_graph_fuc {
+	u32 *data;
+	u32  size;
+};
+
+struct nvc0_graph_priv {
+	struct nouveau_graph base;
+
+	struct nvc0_graph_fuc fuc409c;
+	struct nvc0_graph_fuc fuc409d;
+	struct nvc0_graph_fuc fuc41ac;
+	struct nvc0_graph_fuc fuc41ad;
+	bool firmware;
+
+	u8 rop_nr;
+	u8 gpc_nr;
+	u8 tpc_nr[GPC_MAX];
+	u8 tpc_total;
+
+	struct nouveau_gpuobj *unk4188b4;
+	struct nouveau_gpuobj *unk4188b8;
+
+	struct nvc0_graph_data mmio_data[4];
+	struct nvc0_graph_mmio mmio_list[4096/8];
+	u32  size;
+	u32 *data;
+
+	u8 magic_not_rop_nr;
+};
+
+struct nvc0_graph_chan {
+	struct nouveau_graph_chan base;
+
+	struct nouveau_gpuobj *mmio;
+	struct nouveau_vma mmio_vma;
+	int mmio_nr;
+	struct {
+		struct nouveau_gpuobj *mem;
+		struct nouveau_vma vma;
+	} data[4];
+};
+
+static inline u32
+nvc0_graph_class(void *obj)
+{
+	struct nouveau_device *device = nv_device(obj);
+
+	switch (device->chipset) {
+	case 0xc0:
+	case 0xc3:
+	case 0xc4:
+	case 0xce: /* guess, mmio trace shows only 0x9097 state */
+	case 0xcf: /* guess, mmio trace shows only 0x9097 state */
+		return 0x9097;
+	case 0xc1:
+		return 0x9197;
+	case 0xc8:
+	case 0xd9:
+		return 0x9297;
+	case 0xe4:
+	case 0xe7:
+		return 0xa097;
+	default:
+		return 0;
+	}
+}
+
+void nv_icmd(struct nvc0_graph_priv *priv, u32 icmd, u32 data);
+
+static inline void
+nv_mthd(struct nvc0_graph_priv *priv, u32 class, u32 mthd, u32 data)
+{
+	nv_wr32(priv, 0x40448c, data);
+	nv_wr32(priv, 0x404488, 0x80000000 | (mthd << 14) | class);
+}
+
+struct nvc0_grctx {
+	struct nvc0_graph_priv *priv;
+	struct nvc0_graph_data *data;
+	struct nvc0_graph_mmio *mmio;
+	struct nouveau_gpuobj *chan;
+	int buffer_nr;
+	u64 buffer[4];
+	u64 addr;
+};
+
+int  nvc0_grctx_generate(struct nvc0_graph_priv *);
+int  nvc0_grctx_init(struct nvc0_graph_priv *, struct nvc0_grctx *);
+void nvc0_grctx_data(struct nvc0_grctx *, u32, u32, u32);
+void nvc0_grctx_mmio(struct nvc0_grctx *, u32, u32, u32, u32);
+int  nvc0_grctx_fini(struct nvc0_grctx *);
+
+int  nve0_grctx_generate(struct nvc0_graph_priv *);
+
+#define mmio_data(s,a,p) nvc0_grctx_data(&info, (s), (a), (p))
+#define mmio_list(r,d,s,b) nvc0_grctx_mmio(&info, (r), (d), (s), (b))
+
+void nvc0_graph_ctxctl_debug(struct nvc0_graph_priv *);
+int  nvc0_graph_ctor_fw(struct nvc0_graph_priv *, const char *,
+			struct nvc0_graph_fuc *);
+void nvc0_graph_dtor(struct nouveau_object *);
+void nvc0_graph_init_fw(struct nvc0_graph_priv *, u32 base,
+			struct nvc0_graph_fuc *, struct nvc0_graph_fuc *);
+int  nvc0_graph_context_ctor(struct nouveau_object *, struct nouveau_object *,
+			     struct nouveau_oclass *, void *, u32,
+			     struct nouveau_object **);
+void nvc0_graph_context_dtor(struct nouveau_object *);
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nve0.c b/drivers/gpu/drm/nouveau/core/engine/graph/nve0.c
new file mode 100644
index 0000000..539d4c7
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nve0.c
@@ -0,0 +1,576 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include "nvc0.h"
+#include "fuc/hubnve0.fuc.h"
+#include "fuc/gpcnve0.fuc.h"
+
+/*******************************************************************************
+ * Graphics object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nve0_graph_sclass[] = {
+	{ 0x902d, &nouveau_object_ofuncs },
+	{ 0xa040, &nouveau_object_ofuncs },
+	{ 0xa097, &nouveau_object_ofuncs },
+	{ 0xa0c0, &nouveau_object_ofuncs },
+	{ 0xa0b5, &nouveau_object_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * PGRAPH context
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nve0_graph_cclass = {
+	.handle = NV_ENGCTX(GR, 0xe0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvc0_graph_context_ctor,
+		.dtor = nvc0_graph_context_dtor,
+		.init = _nouveau_graph_context_init,
+		.fini = _nouveau_graph_context_fini,
+		.rd32 = _nouveau_graph_context_rd32,
+		.wr32 = _nouveau_graph_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PGRAPH engine/subdev functions
+ ******************************************************************************/
+
+static void
+nve0_graph_ctxctl_isr(struct nvc0_graph_priv *priv)
+{
+	u32 ustat = nv_rd32(priv, 0x409c18);
+
+	if (ustat & 0x00000001)
+		nv_error(priv, "CTXCTRL ucode error\n");
+	if (ustat & 0x00080000)
+		nv_error(priv, "CTXCTRL watchdog timeout\n");
+	if (ustat & ~0x00080001)
+		nv_error(priv, "CTXCTRL 0x%08x\n", ustat);
+
+	nvc0_graph_ctxctl_debug(priv);
+	nv_wr32(priv, 0x409c20, ustat);
+}
+
+static void
+nve0_graph_trap_isr(struct nvc0_graph_priv *priv, int chid, u64 inst)
+{
+	u32 trap = nv_rd32(priv, 0x400108);
+	int rop;
+
+	if (trap & 0x00000001) {
+		u32 stat = nv_rd32(priv, 0x404000);
+		nv_error(priv, "DISPATCH ch %d [0x%010llx] 0x%08x\n",
+			 chid, inst, stat);
+		nv_wr32(priv, 0x404000, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x00000001);
+		trap &= ~0x00000001;
+	}
+
+	if (trap & 0x00000010) {
+		u32 stat = nv_rd32(priv, 0x405840);
+		nv_error(priv, "SHADER ch %d [0x%010llx] 0x%08x\n",
+			 chid, inst, stat);
+		nv_wr32(priv, 0x405840, 0xc0000000);
+		nv_wr32(priv, 0x400108, 0x00000010);
+		trap &= ~0x00000010;
+	}
+
+	if (trap & 0x02000000) {
+		for (rop = 0; rop < priv->rop_nr; rop++) {
+			u32 statz = nv_rd32(priv, ROP_UNIT(rop, 0x070));
+			u32 statc = nv_rd32(priv, ROP_UNIT(rop, 0x144));
+			nv_error(priv, "ROP%d ch %d [0x%010llx] 0x%08x 0x%08x\n",
+				 rop, chid, inst, statz, statc);
+			nv_wr32(priv, ROP_UNIT(rop, 0x070), 0xc0000000);
+			nv_wr32(priv, ROP_UNIT(rop, 0x144), 0xc0000000);
+		}
+		nv_wr32(priv, 0x400108, 0x02000000);
+		trap &= ~0x02000000;
+	}
+
+	if (trap) {
+		nv_error(priv, "TRAP ch %d [0x%010llx] 0x%08x\n",
+			 chid, inst, trap);
+		nv_wr32(priv, 0x400108, trap);
+	}
+}
+
+static void
+nve0_graph_intr(struct nouveau_subdev *subdev)
+{
+	struct nouveau_fifo *pfifo = nouveau_fifo(subdev);
+	struct nouveau_engine *engine = nv_engine(subdev);
+	struct nouveau_object *engctx;
+	struct nouveau_handle *handle;
+	struct nvc0_graph_priv *priv = (void *)subdev;
+	u64 inst = nv_rd32(priv, 0x409b00) & 0x0fffffff;
+	u32 stat = nv_rd32(priv, 0x400100);
+	u32 addr = nv_rd32(priv, 0x400704);
+	u32 mthd = (addr & 0x00003ffc);
+	u32 subc = (addr & 0x00070000) >> 16;
+	u32 data = nv_rd32(priv, 0x400708);
+	u32 code = nv_rd32(priv, 0x400110);
+	u32 class = nv_rd32(priv, 0x404200 + (subc * 4));
+	int chid;
+
+	engctx = nouveau_engctx_get(engine, inst);
+	chid   = pfifo->chid(pfifo, engctx);
+
+	if (stat & 0x00000010) {
+		handle = nouveau_handle_get_class(engctx, class);
+		if (!handle || nv_call(handle->object, mthd, data)) {
+			nv_error(priv, "ILLEGAL_MTHD ch %d [0x%010llx] "
+				     "subc %d class 0x%04x mthd 0x%04x "
+				     "data 0x%08x\n",
+				 chid, inst, subc, class, mthd, data);
+		}
+		nouveau_handle_put(handle);
+		nv_wr32(priv, 0x400100, 0x00000010);
+		stat &= ~0x00000010;
+	}
+
+	if (stat & 0x00000020) {
+		nv_error(priv, "ILLEGAL_CLASS ch %d [0x%010llx] subc %d "
+			     "class 0x%04x mthd 0x%04x data 0x%08x\n",
+			 chid, inst, subc, class, mthd, data);
+		nv_wr32(priv, 0x400100, 0x00000020);
+		stat &= ~0x00000020;
+	}
+
+	if (stat & 0x00100000) {
+		nv_error(priv, "DATA_ERROR [");
+		nouveau_enum_print(nv50_data_error_names, code);
+		printk("] ch %d [0x%010llx] subc %d class 0x%04x "
+		       "mthd 0x%04x data 0x%08x\n",
+		       chid, inst, subc, class, mthd, data);
+		nv_wr32(priv, 0x400100, 0x00100000);
+		stat &= ~0x00100000;
+	}
+
+	if (stat & 0x00200000) {
+		nve0_graph_trap_isr(priv, chid, inst);
+		nv_wr32(priv, 0x400100, 0x00200000);
+		stat &= ~0x00200000;
+	}
+
+	if (stat & 0x00080000) {
+		nve0_graph_ctxctl_isr(priv);
+		nv_wr32(priv, 0x400100, 0x00080000);
+		stat &= ~0x00080000;
+	}
+
+	if (stat) {
+		nv_error(priv, "unknown stat 0x%08x\n", stat);
+		nv_wr32(priv, 0x400100, stat);
+	}
+
+	nv_wr32(priv, 0x400500, 0x00010001);
+	nouveau_engctx_put(engctx);
+}
+
+static int
+nve0_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nouveau_device *device = nv_device(parent);
+	struct nvc0_graph_priv *priv;
+	int ret, i;
+
+	ret = nouveau_graph_create(parent, engine, oclass, false, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x18001000;
+	nv_subdev(priv)->intr = nve0_graph_intr;
+	nv_engine(priv)->cclass = &nve0_graph_cclass;
+	nv_engine(priv)->sclass = nve0_graph_sclass;
+
+	if (nouveau_boolopt(device->cfgopt, "NvGrUseFW", false)) {
+		nv_info(priv, "using external firmware\n");
+		if (nvc0_graph_ctor_fw(priv, "fuc409c", &priv->fuc409c) ||
+		    nvc0_graph_ctor_fw(priv, "fuc409d", &priv->fuc409d) ||
+		    nvc0_graph_ctor_fw(priv, "fuc41ac", &priv->fuc41ac) ||
+		    nvc0_graph_ctor_fw(priv, "fuc41ad", &priv->fuc41ad))
+			return -EINVAL;
+		priv->firmware = true;
+	}
+
+	ret = nouveau_gpuobj_new(parent, NULL, 0x1000, 256, 0, &priv->unk4188b4);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_new(parent, NULL, 0x1000, 256, 0, &priv->unk4188b8);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < 0x1000; i += 4) {
+		nv_wo32(priv->unk4188b4, i, 0x00000010);
+		nv_wo32(priv->unk4188b8, i, 0x00000010);
+	}
+
+	priv->gpc_nr =  nv_rd32(priv, 0x409604) & 0x0000001f;
+	priv->rop_nr = (nv_rd32(priv, 0x409604) & 0x001f0000) >> 16;
+	for (i = 0; i < priv->gpc_nr; i++) {
+		priv->tpc_nr[i] = nv_rd32(priv, GPC_UNIT(i, 0x2608));
+		priv->tpc_total += priv->tpc_nr[i];
+	}
+
+	switch (nv_device(priv)->chipset) {
+	case 0xe4:
+		if (priv->tpc_total == 8)
+			priv->magic_not_rop_nr = 3;
+		else
+		if (priv->tpc_total == 7)
+			priv->magic_not_rop_nr = 1;
+		break;
+	case 0xe7:
+		priv->magic_not_rop_nr = 1;
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static void
+nve0_graph_init_obj418880(struct nvc0_graph_priv *priv)
+{
+	int i;
+
+	nv_wr32(priv, GPC_BCAST(0x0880), 0x00000000);
+	nv_wr32(priv, GPC_BCAST(0x08a4), 0x00000000);
+	for (i = 0; i < 4; i++)
+		nv_wr32(priv, GPC_BCAST(0x0888) + (i * 4), 0x00000000);
+	nv_wr32(priv, GPC_BCAST(0x08b4), priv->unk4188b4->addr >> 8);
+	nv_wr32(priv, GPC_BCAST(0x08b8), priv->unk4188b8->addr >> 8);
+}
+
+static void
+nve0_graph_init_regs(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x400080, 0x003083c2);
+	nv_wr32(priv, 0x400088, 0x0001ffe7);
+	nv_wr32(priv, 0x40008c, 0x00000000);
+	nv_wr32(priv, 0x400090, 0x00000030);
+	nv_wr32(priv, 0x40013c, 0x003901f7);
+	nv_wr32(priv, 0x400140, 0x00000100);
+	nv_wr32(priv, 0x400144, 0x00000000);
+	nv_wr32(priv, 0x400148, 0x00000110);
+	nv_wr32(priv, 0x400138, 0x00000000);
+	nv_wr32(priv, 0x400130, 0x00000000);
+	nv_wr32(priv, 0x400134, 0x00000000);
+	nv_wr32(priv, 0x400124, 0x00000002);
+}
+
+static void
+nve0_graph_init_units(struct nvc0_graph_priv *priv)
+{
+	nv_wr32(priv, 0x409ffc, 0x00000000);
+	nv_wr32(priv, 0x409c14, 0x00003e3e);
+	nv_wr32(priv, 0x409c24, 0x000f0000);
+
+	nv_wr32(priv, 0x404000, 0xc0000000);
+	nv_wr32(priv, 0x404600, 0xc0000000);
+	nv_wr32(priv, 0x408030, 0xc0000000);
+	nv_wr32(priv, 0x404490, 0xc0000000);
+	nv_wr32(priv, 0x406018, 0xc0000000);
+	nv_wr32(priv, 0x407020, 0xc0000000);
+	nv_wr32(priv, 0x405840, 0xc0000000);
+	nv_wr32(priv, 0x405844, 0x00ffffff);
+
+	nv_mask(priv, 0x419cc0, 0x00000008, 0x00000008);
+	nv_mask(priv, 0x419eb4, 0x00001000, 0x00001000);
+
+}
+
+static void
+nve0_graph_init_gpc_0(struct nvc0_graph_priv *priv)
+{
+	const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, priv->tpc_total);
+	u32 data[TPC_MAX / 8];
+	u8  tpcnr[GPC_MAX];
+	int i, gpc, tpc;
+
+	nv_wr32(priv, GPC_UNIT(0, 0x3018), 0x00000001);
+
+	memset(data, 0x00, sizeof(data));
+	memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr));
+	for (i = 0, gpc = -1; i < priv->tpc_total; i++) {
+		do {
+			gpc = (gpc + 1) % priv->gpc_nr;
+		} while (!tpcnr[gpc]);
+		tpc = priv->tpc_nr[gpc] - tpcnr[gpc]--;
+
+		data[i / 8] |= tpc << ((i % 8) * 4);
+	}
+
+	nv_wr32(priv, GPC_BCAST(0x0980), data[0]);
+	nv_wr32(priv, GPC_BCAST(0x0984), data[1]);
+	nv_wr32(priv, GPC_BCAST(0x0988), data[2]);
+	nv_wr32(priv, GPC_BCAST(0x098c), data[3]);
+
+	for (gpc = 0; gpc < priv->gpc_nr; gpc++) {
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0914), priv->magic_not_rop_nr << 8 |
+						  priv->tpc_nr[gpc]);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0910), 0x00040000 | priv->tpc_total);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0918), magicgpc918);
+	}
+
+	nv_wr32(priv, GPC_BCAST(0x1bd4), magicgpc918);
+	nv_wr32(priv, GPC_BCAST(0x08ac), nv_rd32(priv, 0x100800));
+}
+
+static void
+nve0_graph_init_gpc_1(struct nvc0_graph_priv *priv)
+{
+	int gpc, tpc;
+
+	for (gpc = 0; gpc < priv->gpc_nr; gpc++) {
+		nv_wr32(priv, GPC_UNIT(gpc, 0x3038), 0xc0000000);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0420), 0xc0000000);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0900), 0xc0000000);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x1028), 0xc0000000);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x0824), 0xc0000000);
+		for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) {
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff);
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff);
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000);
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000);
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000);
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x644), 0x001ffffe);
+			nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x64c), 0x0000000f);
+		}
+		nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), 0xffffffff);
+		nv_wr32(priv, GPC_UNIT(gpc, 0x2c94), 0xffffffff);
+	}
+}
+
+static void
+nve0_graph_init_rop(struct nvc0_graph_priv *priv)
+{
+	int rop;
+
+	for (rop = 0; rop < priv->rop_nr; rop++) {
+		nv_wr32(priv, ROP_UNIT(rop, 0x144), 0xc0000000);
+		nv_wr32(priv, ROP_UNIT(rop, 0x070), 0xc0000000);
+		nv_wr32(priv, ROP_UNIT(rop, 0x204), 0xffffffff);
+		nv_wr32(priv, ROP_UNIT(rop, 0x208), 0xffffffff);
+	}
+}
+
+static int
+nve0_graph_init_ctxctl(struct nvc0_graph_priv *priv)
+{
+	u32 r000260;
+	int i;
+
+	if (priv->firmware) {
+		/* load fuc microcode */
+		r000260 = nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
+		nvc0_graph_init_fw(priv, 0x409000, &priv->fuc409c, &priv->fuc409d);
+		nvc0_graph_init_fw(priv, 0x41a000, &priv->fuc41ac, &priv->fuc41ad);
+		nv_wr32(priv, 0x000260, r000260);
+
+		/* start both of them running */
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x41a10c, 0x00000000);
+		nv_wr32(priv, 0x40910c, 0x00000000);
+		nv_wr32(priv, 0x41a100, 0x00000002);
+		nv_wr32(priv, 0x409100, 0x00000002);
+		if (!nv_wait(priv, 0x409800, 0x00000001, 0x00000001))
+			nv_error(priv, "0x409800 wait failed\n");
+
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x409500, 0x7fffffff);
+		nv_wr32(priv, 0x409504, 0x00000021);
+
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x409500, 0x00000000);
+		nv_wr32(priv, 0x409504, 0x00000010);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x10 timeout\n");
+			return -EBUSY;
+		}
+		priv->size = nv_rd32(priv, 0x409800);
+
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x409500, 0x00000000);
+		nv_wr32(priv, 0x409504, 0x00000016);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x16 timeout\n");
+			return -EBUSY;
+		}
+
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x409500, 0x00000000);
+		nv_wr32(priv, 0x409504, 0x00000025);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x25 timeout\n");
+			return -EBUSY;
+		}
+
+		nv_wr32(priv, 0x409800, 0x00000000);
+		nv_wr32(priv, 0x409500, 0x00000001);
+		nv_wr32(priv, 0x409504, 0x00000030);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x30 timeout\n");
+			return -EBUSY;
+		}
+
+		nv_wr32(priv, 0x409810, 0xb00095c8);
+		nv_wr32(priv, 0x409800, 0x00000000);
+		nv_wr32(priv, 0x409500, 0x00000001);
+		nv_wr32(priv, 0x409504, 0x00000031);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x31 timeout\n");
+			return -EBUSY;
+		}
+
+		nv_wr32(priv, 0x409810, 0x00080420);
+		nv_wr32(priv, 0x409800, 0x00000000);
+		nv_wr32(priv, 0x409500, 0x00000001);
+		nv_wr32(priv, 0x409504, 0x00000032);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x32 timeout\n");
+			return -EBUSY;
+		}
+
+		nv_wr32(priv, 0x409614, 0x00000070);
+		nv_wr32(priv, 0x409614, 0x00000770);
+		nv_wr32(priv, 0x40802c, 0x00000001);
+
+		if (priv->data == NULL) {
+			int ret = nve0_grctx_generate(priv);
+			if (ret) {
+				nv_error(priv, "failed to construct context\n");
+				return ret;
+			}
+		}
+
+		return 0;
+	}
+
+	/* load HUB microcode */
+	r000260 = nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
+	nv_wr32(priv, 0x4091c0, 0x01000000);
+	for (i = 0; i < sizeof(nve0_grhub_data) / 4; i++)
+		nv_wr32(priv, 0x4091c4, nve0_grhub_data[i]);
+
+	nv_wr32(priv, 0x409180, 0x01000000);
+	for (i = 0; i < sizeof(nve0_grhub_code) / 4; i++) {
+		if ((i & 0x3f) == 0)
+			nv_wr32(priv, 0x409188, i >> 6);
+		nv_wr32(priv, 0x409184, nve0_grhub_code[i]);
+	}
+
+	/* load GPC microcode */
+	nv_wr32(priv, 0x41a1c0, 0x01000000);
+	for (i = 0; i < sizeof(nve0_grgpc_data) / 4; i++)
+		nv_wr32(priv, 0x41a1c4, nve0_grgpc_data[i]);
+
+	nv_wr32(priv, 0x41a180, 0x01000000);
+	for (i = 0; i < sizeof(nve0_grgpc_code) / 4; i++) {
+		if ((i & 0x3f) == 0)
+			nv_wr32(priv, 0x41a188, i >> 6);
+		nv_wr32(priv, 0x41a184, nve0_grgpc_code[i]);
+	}
+	nv_wr32(priv, 0x000260, r000260);
+
+	/* start HUB ucode running, it'll init the GPCs */
+	nv_wr32(priv, 0x409800, nv_device(priv)->chipset);
+	nv_wr32(priv, 0x40910c, 0x00000000);
+	nv_wr32(priv, 0x409100, 0x00000002);
+	if (!nv_wait(priv, 0x409800, 0x80000000, 0x80000000)) {
+		nv_error(priv, "HUB_INIT timed out\n");
+		nvc0_graph_ctxctl_debug(priv);
+		return -EBUSY;
+	}
+
+	priv->size = nv_rd32(priv, 0x409804);
+	if (priv->data == NULL) {
+		int ret = nve0_grctx_generate(priv);
+		if (ret) {
+			nv_error(priv, "failed to construct context\n");
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
+static int
+nve0_graph_init(struct nouveau_object *object)
+{
+	struct nvc0_graph_priv *priv = (void *)object;
+	int ret;
+
+	ret = nouveau_graph_init(&priv->base);
+	if (ret)
+		return ret;
+
+	nve0_graph_init_obj418880(priv);
+	nve0_graph_init_regs(priv);
+	nve0_graph_init_gpc_0(priv);
+
+	nv_wr32(priv, 0x400500, 0x00010001);
+	nv_wr32(priv, 0x400100, 0xffffffff);
+	nv_wr32(priv, 0x40013c, 0xffffffff);
+
+	nve0_graph_init_units(priv);
+	nve0_graph_init_gpc_1(priv);
+	nve0_graph_init_rop(priv);
+
+	nv_wr32(priv, 0x400108, 0xffffffff);
+	nv_wr32(priv, 0x400138, 0xffffffff);
+	nv_wr32(priv, 0x400118, 0xffffffff);
+	nv_wr32(priv, 0x400130, 0xffffffff);
+	nv_wr32(priv, 0x40011c, 0xffffffff);
+	nv_wr32(priv, 0x400134, 0xffffffff);
+	nv_wr32(priv, 0x400054, 0x34ce3464);
+
+	ret = nve0_graph_init_ctxctl(priv);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+struct nouveau_oclass
+nve0_graph_oclass = {
+	.handle = NV_ENGINE(GR, 0xe0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nve0_graph_ctor,
+		.dtor = nvc0_graph_dtor,
+		.init = nve0_graph_init,
+		.fini = _nouveau_graph_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/regs.h b/drivers/gpu/drm/nouveau/core/engine/graph/regs.h
new file mode 100644
index 0000000..9c715a2
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/regs.h
@@ -0,0 +1,269 @@
+#ifndef __NOUVEAU_GRAPH_REGS_H__
+#define __NOUVEAU_GRAPH_REGS_H__
+
+#define NV04_PGRAPH_DEBUG_0                                0x00400080
+#define NV04_PGRAPH_DEBUG_1                                0x00400084
+#define NV04_PGRAPH_DEBUG_2                                0x00400088
+#define NV04_PGRAPH_DEBUG_3                                0x0040008c
+#define NV10_PGRAPH_DEBUG_4                                0x00400090
+#define NV03_PGRAPH_INTR                                   0x00400100
+#define NV03_PGRAPH_NSTATUS                                0x00400104
+#    define NV04_PGRAPH_NSTATUS_STATE_IN_USE                  (1<<11)
+#    define NV04_PGRAPH_NSTATUS_INVALID_STATE                 (1<<12)
+#    define NV04_PGRAPH_NSTATUS_BAD_ARGUMENT                  (1<<13)
+#    define NV04_PGRAPH_NSTATUS_PROTECTION_FAULT              (1<<14)
+#    define NV10_PGRAPH_NSTATUS_STATE_IN_USE                  (1<<23)
+#    define NV10_PGRAPH_NSTATUS_INVALID_STATE                 (1<<24)
+#    define NV10_PGRAPH_NSTATUS_BAD_ARGUMENT                  (1<<25)
+#    define NV10_PGRAPH_NSTATUS_PROTECTION_FAULT              (1<<26)
+#define NV03_PGRAPH_NSOURCE                                0x00400108
+#    define NV03_PGRAPH_NSOURCE_NOTIFICATION                   (1<<0)
+#    define NV03_PGRAPH_NSOURCE_DATA_ERROR                     (1<<1)
+#    define NV03_PGRAPH_NSOURCE_PROTECTION_ERROR               (1<<2)
+#    define NV03_PGRAPH_NSOURCE_RANGE_EXCEPTION                (1<<3)
+#    define NV03_PGRAPH_NSOURCE_LIMIT_COLOR                    (1<<4)
+#    define NV03_PGRAPH_NSOURCE_LIMIT_ZETA                     (1<<5)
+#    define NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD                   (1<<6)
+#    define NV03_PGRAPH_NSOURCE_DMA_R_PROTECTION               (1<<7)
+#    define NV03_PGRAPH_NSOURCE_DMA_W_PROTECTION               (1<<8)
+#    define NV03_PGRAPH_NSOURCE_FORMAT_EXCEPTION               (1<<9)
+#    define NV03_PGRAPH_NSOURCE_PATCH_EXCEPTION               (1<<10)
+#    define NV03_PGRAPH_NSOURCE_STATE_INVALID                 (1<<11)
+#    define NV03_PGRAPH_NSOURCE_DOUBLE_NOTIFY                 (1<<12)
+#    define NV03_PGRAPH_NSOURCE_NOTIFY_IN_USE                 (1<<13)
+#    define NV03_PGRAPH_NSOURCE_METHOD_CNT                    (1<<14)
+#    define NV03_PGRAPH_NSOURCE_BFR_NOTIFICATION              (1<<15)
+#    define NV03_PGRAPH_NSOURCE_DMA_VTX_PROTECTION            (1<<16)
+#    define NV03_PGRAPH_NSOURCE_DMA_WIDTH_A                   (1<<17)
+#    define NV03_PGRAPH_NSOURCE_DMA_WIDTH_B                   (1<<18)
+#define NV03_PGRAPH_INTR_EN                                0x00400140
+#define NV40_PGRAPH_INTR_EN                                0x0040013C
+#    define NV_PGRAPH_INTR_NOTIFY                              (1<<0)
+#    define NV_PGRAPH_INTR_MISSING_HW                          (1<<4)
+#    define NV_PGRAPH_INTR_CONTEXT_SWITCH                     (1<<12)
+#    define NV_PGRAPH_INTR_BUFFER_NOTIFY                      (1<<16)
+#    define NV_PGRAPH_INTR_ERROR                              (1<<20)
+#define NV10_PGRAPH_CTX_CONTROL                            0x00400144
+#define NV10_PGRAPH_CTX_USER                               0x00400148
+#define NV10_PGRAPH_CTX_SWITCH(i)                         (0x0040014C + 0x4*(i))
+#define NV04_PGRAPH_CTX_SWITCH1                            0x00400160
+#define NV10_PGRAPH_CTX_CACHE(i, j)                       (0x00400160	\
+							   + 0x4*(i) + 0x20*(j))
+#define NV04_PGRAPH_CTX_SWITCH2                            0x00400164
+#define NV04_PGRAPH_CTX_SWITCH3                            0x00400168
+#define NV04_PGRAPH_CTX_SWITCH4                            0x0040016C
+#define NV04_PGRAPH_CTX_CONTROL                            0x00400170
+#define NV04_PGRAPH_CTX_USER                               0x00400174
+#define NV04_PGRAPH_CTX_CACHE1                             0x00400180
+#define NV03_PGRAPH_CTX_CONTROL                            0x00400190
+#define NV03_PGRAPH_CTX_USER                               0x00400194
+#define NV04_PGRAPH_CTX_CACHE2                             0x004001A0
+#define NV04_PGRAPH_CTX_CACHE3                             0x004001C0
+#define NV04_PGRAPH_CTX_CACHE4                             0x004001E0
+#define NV40_PGRAPH_CTXCTL_0304                            0x00400304
+#define NV40_PGRAPH_CTXCTL_0304_XFER_CTX                   0x00000001
+#define NV40_PGRAPH_CTXCTL_UCODE_STAT                      0x00400308
+#define NV40_PGRAPH_CTXCTL_UCODE_STAT_IP_MASK              0xff000000
+#define NV40_PGRAPH_CTXCTL_UCODE_STAT_IP_SHIFT                     24
+#define NV40_PGRAPH_CTXCTL_UCODE_STAT_OP_MASK              0x00ffffff
+#define NV40_PGRAPH_CTXCTL_0310                            0x00400310
+#define NV40_PGRAPH_CTXCTL_0310_XFER_SAVE                  0x00000020
+#define NV40_PGRAPH_CTXCTL_0310_XFER_LOAD                  0x00000040
+#define NV40_PGRAPH_CTXCTL_030C                            0x0040030c
+#define NV40_PGRAPH_CTXCTL_UCODE_INDEX                     0x00400324
+#define NV40_PGRAPH_CTXCTL_UCODE_DATA                      0x00400328
+#define NV40_PGRAPH_CTXCTL_CUR                             0x0040032c
+#define NV40_PGRAPH_CTXCTL_CUR_LOADED                      0x01000000
+#define NV40_PGRAPH_CTXCTL_CUR_INSTANCE                    0x000FFFFF
+#define NV40_PGRAPH_CTXCTL_NEXT                            0x00400330
+#define NV40_PGRAPH_CTXCTL_NEXT_INSTANCE                   0x000fffff
+#define NV50_PGRAPH_CTXCTL_CUR                             0x0040032c
+#define NV50_PGRAPH_CTXCTL_CUR_LOADED                      0x80000000
+#define NV50_PGRAPH_CTXCTL_CUR_INSTANCE                    0x00ffffff
+#define NV50_PGRAPH_CTXCTL_NEXT                            0x00400330
+#define NV50_PGRAPH_CTXCTL_NEXT_INSTANCE                   0x00ffffff
+#define NV03_PGRAPH_ABS_X_RAM                              0x00400400
+#define NV03_PGRAPH_ABS_Y_RAM                              0x00400480
+#define NV03_PGRAPH_X_MISC                                 0x00400500
+#define NV03_PGRAPH_Y_MISC                                 0x00400504
+#define NV04_PGRAPH_VALID1                                 0x00400508
+#define NV04_PGRAPH_SOURCE_COLOR                           0x0040050C
+#define NV04_PGRAPH_MISC24_0                               0x00400510
+#define NV03_PGRAPH_XY_LOGIC_MISC0                         0x00400514
+#define NV03_PGRAPH_XY_LOGIC_MISC1                         0x00400518
+#define NV03_PGRAPH_XY_LOGIC_MISC2                         0x0040051C
+#define NV03_PGRAPH_XY_LOGIC_MISC3                         0x00400520
+#define NV03_PGRAPH_CLIPX_0                                0x00400524
+#define NV03_PGRAPH_CLIPX_1                                0x00400528
+#define NV03_PGRAPH_CLIPY_0                                0x0040052C
+#define NV03_PGRAPH_CLIPY_1                                0x00400530
+#define NV03_PGRAPH_ABS_ICLIP_XMAX                         0x00400534
+#define NV03_PGRAPH_ABS_ICLIP_YMAX                         0x00400538
+#define NV03_PGRAPH_ABS_UCLIP_XMIN                         0x0040053C
+#define NV03_PGRAPH_ABS_UCLIP_YMIN                         0x00400540
+#define NV03_PGRAPH_ABS_UCLIP_XMAX                         0x00400544
+#define NV03_PGRAPH_ABS_UCLIP_YMAX                         0x00400548
+#define NV03_PGRAPH_ABS_UCLIPA_XMIN                        0x00400560
+#define NV03_PGRAPH_ABS_UCLIPA_YMIN                        0x00400564
+#define NV03_PGRAPH_ABS_UCLIPA_XMAX                        0x00400568
+#define NV03_PGRAPH_ABS_UCLIPA_YMAX                        0x0040056C
+#define NV04_PGRAPH_MISC24_1                               0x00400570
+#define NV04_PGRAPH_MISC24_2                               0x00400574
+#define NV04_PGRAPH_VALID2                                 0x00400578
+#define NV04_PGRAPH_PASSTHRU_0                             0x0040057C
+#define NV04_PGRAPH_PASSTHRU_1                             0x00400580
+#define NV04_PGRAPH_PASSTHRU_2                             0x00400584
+#define NV10_PGRAPH_DIMX_TEXTURE                           0x00400588
+#define NV10_PGRAPH_WDIMX_TEXTURE                          0x0040058C
+#define NV04_PGRAPH_COMBINE_0_ALPHA                        0x00400590
+#define NV04_PGRAPH_COMBINE_0_COLOR                        0x00400594
+#define NV04_PGRAPH_COMBINE_1_ALPHA                        0x00400598
+#define NV04_PGRAPH_COMBINE_1_COLOR                        0x0040059C
+#define NV04_PGRAPH_FORMAT_0                               0x004005A8
+#define NV04_PGRAPH_FORMAT_1                               0x004005AC
+#define NV04_PGRAPH_FILTER_0                               0x004005B0
+#define NV04_PGRAPH_FILTER_1                               0x004005B4
+#define NV03_PGRAPH_MONO_COLOR0                            0x00400600
+#define NV04_PGRAPH_ROP3                                   0x00400604
+#define NV04_PGRAPH_BETA_AND                               0x00400608
+#define NV04_PGRAPH_BETA_PREMULT                           0x0040060C
+#define NV04_PGRAPH_LIMIT_VIOL_PIX                         0x00400610
+#define NV04_PGRAPH_FORMATS                                0x00400618
+#define NV10_PGRAPH_DEBUG_2                                0x00400620
+#define NV04_PGRAPH_BOFFSET0                               0x00400640
+#define NV04_PGRAPH_BOFFSET1                               0x00400644
+#define NV04_PGRAPH_BOFFSET2                               0x00400648
+#define NV04_PGRAPH_BOFFSET3                               0x0040064C
+#define NV04_PGRAPH_BOFFSET4                               0x00400650
+#define NV04_PGRAPH_BOFFSET5                               0x00400654
+#define NV04_PGRAPH_BBASE0                                 0x00400658
+#define NV04_PGRAPH_BBASE1                                 0x0040065C
+#define NV04_PGRAPH_BBASE2                                 0x00400660
+#define NV04_PGRAPH_BBASE3                                 0x00400664
+#define NV04_PGRAPH_BBASE4                                 0x00400668
+#define NV04_PGRAPH_BBASE5                                 0x0040066C
+#define NV04_PGRAPH_BPITCH0                                0x00400670
+#define NV04_PGRAPH_BPITCH1                                0x00400674
+#define NV04_PGRAPH_BPITCH2                                0x00400678
+#define NV04_PGRAPH_BPITCH3                                0x0040067C
+#define NV04_PGRAPH_BPITCH4                                0x00400680
+#define NV04_PGRAPH_BLIMIT0                                0x00400684
+#define NV04_PGRAPH_BLIMIT1                                0x00400688
+#define NV04_PGRAPH_BLIMIT2                                0x0040068C
+#define NV04_PGRAPH_BLIMIT3                                0x00400690
+#define NV04_PGRAPH_BLIMIT4                                0x00400694
+#define NV04_PGRAPH_BLIMIT5                                0x00400698
+#define NV04_PGRAPH_BSWIZZLE2                              0x0040069C
+#define NV04_PGRAPH_BSWIZZLE5                              0x004006A0
+#define NV03_PGRAPH_STATUS                                 0x004006B0
+#define NV04_PGRAPH_STATUS                                 0x00400700
+#    define NV40_PGRAPH_STATUS_SYNC_STALL                  0x00004000
+#define NV04_PGRAPH_TRAPPED_ADDR                           0x00400704
+#define NV04_PGRAPH_TRAPPED_DATA                           0x00400708
+#define NV04_PGRAPH_SURFACE                                0x0040070C
+#define NV10_PGRAPH_TRAPPED_DATA_HIGH                      0x0040070C
+#define NV04_PGRAPH_STATE                                  0x00400710
+#define NV10_PGRAPH_SURFACE                                0x00400710
+#define NV04_PGRAPH_NOTIFY                                 0x00400714
+#define NV10_PGRAPH_STATE                                  0x00400714
+#define NV10_PGRAPH_NOTIFY                                 0x00400718
+
+#define NV04_PGRAPH_FIFO                                   0x00400720
+
+#define NV04_PGRAPH_BPIXEL                                 0x00400724
+#define NV10_PGRAPH_RDI_INDEX                              0x00400750
+#define NV04_PGRAPH_FFINTFC_ST2                            0x00400754
+#define NV10_PGRAPH_RDI_DATA                               0x00400754
+#define NV04_PGRAPH_DMA_PITCH                              0x00400760
+#define NV10_PGRAPH_FFINTFC_FIFO_PTR                       0x00400760
+#define NV04_PGRAPH_DVD_COLORFMT                           0x00400764
+#define NV10_PGRAPH_FFINTFC_ST2                            0x00400764
+#define NV04_PGRAPH_SCALED_FORMAT                          0x00400768
+#define NV10_PGRAPH_FFINTFC_ST2_DL                         0x00400768
+#define NV10_PGRAPH_FFINTFC_ST2_DH                         0x0040076c
+#define NV10_PGRAPH_DMA_PITCH                              0x00400770
+#define NV10_PGRAPH_DVD_COLORFMT                           0x00400774
+#define NV10_PGRAPH_SCALED_FORMAT                          0x00400778
+#define NV20_PGRAPH_CHANNEL_CTX_TABLE                      0x00400780
+#define NV20_PGRAPH_CHANNEL_CTX_POINTER                    0x00400784
+#define NV20_PGRAPH_CHANNEL_CTX_XFER                       0x00400788
+#define NV20_PGRAPH_CHANNEL_CTX_XFER_LOAD                  0x00000001
+#define NV20_PGRAPH_CHANNEL_CTX_XFER_SAVE                  0x00000002
+#define NV04_PGRAPH_PATT_COLOR0                            0x00400800
+#define NV04_PGRAPH_PATT_COLOR1                            0x00400804
+#define NV04_PGRAPH_PATTERN                                0x00400808
+#define NV04_PGRAPH_PATTERN_SHAPE                          0x00400810
+#define NV04_PGRAPH_CHROMA                                 0x00400814
+#define NV04_PGRAPH_CONTROL0                               0x00400818
+#define NV04_PGRAPH_CONTROL1                               0x0040081C
+#define NV04_PGRAPH_CONTROL2                               0x00400820
+#define NV04_PGRAPH_BLEND                                  0x00400824
+#define NV04_PGRAPH_STORED_FMT                             0x00400830
+#define NV04_PGRAPH_PATT_COLORRAM                          0x00400900
+#define NV20_PGRAPH_TILE(i)                                (0x00400900 + (i*16))
+#define NV20_PGRAPH_TLIMIT(i)                              (0x00400904 + (i*16))
+#define NV20_PGRAPH_TSIZE(i)                               (0x00400908 + (i*16))
+#define NV20_PGRAPH_TSTATUS(i)                             (0x0040090C + (i*16))
+#define NV20_PGRAPH_ZCOMP(i)                               (0x00400980 + 4*(i))
+#define NV10_PGRAPH_TILE(i)                                (0x00400B00 + (i*16))
+#define NV10_PGRAPH_TLIMIT(i)                              (0x00400B04 + (i*16))
+#define NV10_PGRAPH_TSIZE(i)                               (0x00400B08 + (i*16))
+#define NV10_PGRAPH_TSTATUS(i)                             (0x00400B0C + (i*16))
+#define NV04_PGRAPH_U_RAM                                  0x00400D00
+#define NV47_PGRAPH_TILE(i)                                (0x00400D00 + (i*16))
+#define NV47_PGRAPH_TLIMIT(i)                              (0x00400D04 + (i*16))
+#define NV47_PGRAPH_TSIZE(i)                               (0x00400D08 + (i*16))
+#define NV47_PGRAPH_TSTATUS(i)                             (0x00400D0C + (i*16))
+#define NV04_PGRAPH_V_RAM                                  0x00400D40
+#define NV04_PGRAPH_W_RAM                                  0x00400D80
+#define NV10_PGRAPH_COMBINER0_IN_ALPHA                     0x00400E40
+#define NV10_PGRAPH_COMBINER1_IN_ALPHA                     0x00400E44
+#define NV10_PGRAPH_COMBINER0_IN_RGB                       0x00400E48
+#define NV10_PGRAPH_COMBINER1_IN_RGB                       0x00400E4C
+#define NV10_PGRAPH_COMBINER_COLOR0                        0x00400E50
+#define NV10_PGRAPH_COMBINER_COLOR1                        0x00400E54
+#define NV10_PGRAPH_COMBINER0_OUT_ALPHA                    0x00400E58
+#define NV10_PGRAPH_COMBINER1_OUT_ALPHA                    0x00400E5C
+#define NV10_PGRAPH_COMBINER0_OUT_RGB                      0x00400E60
+#define NV10_PGRAPH_COMBINER1_OUT_RGB                      0x00400E64
+#define NV10_PGRAPH_COMBINER_FINAL0                        0x00400E68
+#define NV10_PGRAPH_COMBINER_FINAL1                        0x00400E6C
+#define NV10_PGRAPH_WINDOWCLIP_HORIZONTAL                  0x00400F00
+#define NV10_PGRAPH_WINDOWCLIP_VERTICAL                    0x00400F20
+#define NV10_PGRAPH_XFMODE0                                0x00400F40
+#define NV10_PGRAPH_XFMODE1                                0x00400F44
+#define NV10_PGRAPH_GLOBALSTATE0                           0x00400F48
+#define NV10_PGRAPH_GLOBALSTATE1                           0x00400F4C
+#define NV10_PGRAPH_PIPE_ADDRESS                           0x00400F50
+#define NV10_PGRAPH_PIPE_DATA                              0x00400F54
+#define NV04_PGRAPH_DMA_START_0                            0x00401000
+#define NV04_PGRAPH_DMA_START_1                            0x00401004
+#define NV04_PGRAPH_DMA_LENGTH                             0x00401008
+#define NV04_PGRAPH_DMA_MISC                               0x0040100C
+#define NV04_PGRAPH_DMA_DATA_0                             0x00401020
+#define NV04_PGRAPH_DMA_DATA_1                             0x00401024
+#define NV04_PGRAPH_DMA_RM                                 0x00401030
+#define NV04_PGRAPH_DMA_A_XLATE_INST                       0x00401040
+#define NV04_PGRAPH_DMA_A_CONTROL                          0x00401044
+#define NV04_PGRAPH_DMA_A_LIMIT                            0x00401048
+#define NV04_PGRAPH_DMA_A_TLB_PTE                          0x0040104C
+#define NV04_PGRAPH_DMA_A_TLB_TAG                          0x00401050
+#define NV04_PGRAPH_DMA_A_ADJ_OFFSET                       0x00401054
+#define NV04_PGRAPH_DMA_A_OFFSET                           0x00401058
+#define NV04_PGRAPH_DMA_A_SIZE                             0x0040105C
+#define NV04_PGRAPH_DMA_A_Y_SIZE                           0x00401060
+#define NV04_PGRAPH_DMA_B_XLATE_INST                       0x00401080
+#define NV04_PGRAPH_DMA_B_CONTROL                          0x00401084
+#define NV04_PGRAPH_DMA_B_LIMIT                            0x00401088
+#define NV04_PGRAPH_DMA_B_TLB_PTE                          0x0040108C
+#define NV04_PGRAPH_DMA_B_TLB_TAG                          0x00401090
+#define NV04_PGRAPH_DMA_B_ADJ_OFFSET                       0x00401094
+#define NV04_PGRAPH_DMA_B_OFFSET                           0x00401098
+#define NV04_PGRAPH_DMA_B_SIZE                             0x0040109C
+#define NV04_PGRAPH_DMA_B_Y_SIZE                           0x004010A0
+#define NV40_PGRAPH_TILE1(i)                               (0x00406900 + (i*16))
+#define NV40_PGRAPH_TLIMIT1(i)                             (0x00406904 + (i*16))
+#define NV40_PGRAPH_TSIZE1(i)                              (0x00406908 + (i*16))
+#define NV40_PGRAPH_TSTATUS1(i)                            (0x0040690C + (i*16))
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/core/engine/mpeg/nv31.c b/drivers/gpu/drm/nouveau/core/engine/mpeg/nv31.c
new file mode 100644
index 0000000..1f394a2
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/mpeg/nv31.c
@@ -0,0 +1,308 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/handle.h>
+
+#include <subdev/fb.h>
+#include <subdev/timer.h>
+#include <subdev/instmem.h>
+
+#include <engine/fifo.h>
+#include <engine/mpeg.h>
+#include <engine/graph/nv40.h>
+
+struct nv31_mpeg_priv {
+	struct nouveau_mpeg base;
+	atomic_t refcount;
+};
+
+struct nv31_mpeg_chan {
+	struct nouveau_object base;
+};
+
+/*******************************************************************************
+ * MPEG object classes
+ ******************************************************************************/
+
+static int
+nv31_mpeg_object_ctor(struct nouveau_object *parent,
+		      struct nouveau_object *engine,
+		      struct nouveau_oclass *oclass, void *data, u32 size,
+		      struct nouveau_object **pobject)
+{
+	struct nouveau_gpuobj *obj;
+	int ret;
+
+	ret = nouveau_gpuobj_create(parent, engine, oclass, 0, parent,
+				    20, 16, 0, &obj);
+	*pobject = nv_object(obj);
+	if (ret)
+		return ret;
+
+	nv_wo32(obj, 0x00, nv_mclass(obj));
+	nv_wo32(obj, 0x04, 0x00000000);
+	nv_wo32(obj, 0x08, 0x00000000);
+	nv_wo32(obj, 0x0c, 0x00000000);
+	return 0;
+}
+
+static int
+nv31_mpeg_mthd_dma(struct nouveau_object *object, u32 mthd, void *arg, u32 len)
+{
+	struct nouveau_instmem *imem = nouveau_instmem(object);
+	struct nv31_mpeg_priv *priv = (void *)object->engine;
+	u32 inst = *(u32 *)arg << 4;
+	u32 dma0 = nv_ro32(imem, inst + 0);
+	u32 dma1 = nv_ro32(imem, inst + 4);
+	u32 dma2 = nv_ro32(imem, inst + 8);
+	u32 base = (dma2 & 0xfffff000) | (dma0 >> 20);
+	u32 size = dma1 + 1;
+
+	/* only allow linear DMA objects */
+	if (!(dma0 & 0x00002000))
+		return -EINVAL;
+
+	if (mthd == 0x0190) {
+		/* DMA_CMD */
+		nv_mask(priv, 0x00b300, 0x00030000, (dma0 & 0x00030000));
+		nv_wr32(priv, 0x00b334, base);
+		nv_wr32(priv, 0x00b324, size);
+	} else
+	if (mthd == 0x01a0) {
+		/* DMA_DATA */
+		nv_mask(priv, 0x00b300, 0x000c0000, (dma0 & 0x00030000) << 2);
+		nv_wr32(priv, 0x00b360, base);
+		nv_wr32(priv, 0x00b364, size);
+	} else {
+		/* DMA_IMAGE, VRAM only */
+		if (dma0 & 0x000c0000)
+			return -EINVAL;
+
+		nv_wr32(priv, 0x00b370, base);
+		nv_wr32(priv, 0x00b374, size);
+	}
+
+	return 0;
+}
+
+static struct nouveau_ofuncs
+nv31_mpeg_ofuncs = {
+	.ctor = nv31_mpeg_object_ctor,
+	.dtor = _nouveau_gpuobj_dtor,
+	.init = _nouveau_gpuobj_init,
+	.fini = _nouveau_gpuobj_fini,
+	.rd32 = _nouveau_gpuobj_rd32,
+	.wr32 = _nouveau_gpuobj_wr32,
+};
+
+static struct nouveau_omthds
+nv31_mpeg_omthds[] = {
+	{ 0x0190, nv31_mpeg_mthd_dma },
+	{ 0x01a0, nv31_mpeg_mthd_dma },
+	{ 0x01b0, nv31_mpeg_mthd_dma },
+	{}
+};
+
+struct nouveau_oclass
+nv31_mpeg_sclass[] = {
+	{ 0x3174, &nv31_mpeg_ofuncs, nv31_mpeg_omthds },
+	{}
+};
+
+/*******************************************************************************
+ * PMPEG context
+ ******************************************************************************/
+
+static int
+nv31_mpeg_context_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nv31_mpeg_priv *priv = (void *)engine;
+	struct nv31_mpeg_chan *chan;
+	int ret;
+
+	if (!atomic_add_unless(&priv->refcount, 1, 1))
+		return -EBUSY;
+
+	ret = nouveau_object_create(parent, engine, oclass, 0, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static void
+nv31_mpeg_context_dtor(struct nouveau_object *object)
+{
+	struct nv31_mpeg_priv *priv = (void *)object->engine;
+	struct nv31_mpeg_chan *chan = (void *)object;
+	atomic_dec(&priv->refcount);
+	nouveau_object_destroy(&chan->base);
+}
+
+static struct nouveau_oclass
+nv31_mpeg_cclass = {
+	.handle = NV_ENGCTX(MPEG, 0x31),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv31_mpeg_context_ctor,
+		.dtor = nv31_mpeg_context_dtor,
+		.init = nouveau_object_init,
+		.fini = nouveau_object_fini,
+	},
+};
+
+/*******************************************************************************
+ * PMPEG engine/subdev functions
+ ******************************************************************************/
+
+void
+nv31_mpeg_tile_prog(struct nouveau_engine *engine, int i)
+{
+	struct nouveau_fb_tile *tile = &nouveau_fb(engine)->tile.region[i];
+	struct nv31_mpeg_priv *priv = (void *)engine;
+
+	nv_wr32(priv, 0x00b008 + (i * 0x10), tile->pitch);
+	nv_wr32(priv, 0x00b004 + (i * 0x10), tile->limit);
+	nv_wr32(priv, 0x00b000 + (i * 0x10), tile->addr);
+}
+
+void
+nv31_mpeg_intr(struct nouveau_subdev *subdev)
+{
+	struct nouveau_fifo *pfifo = nouveau_fifo(subdev);
+	struct nouveau_engine *engine = nv_engine(subdev);
+	struct nouveau_object *engctx;
+	struct nouveau_handle *handle;
+	struct nv31_mpeg_priv *priv = (void *)subdev;
+	u32 inst = nv_rd32(priv, 0x00b318) & 0x000fffff;
+	u32 stat = nv_rd32(priv, 0x00b100);
+	u32 type = nv_rd32(priv, 0x00b230);
+	u32 mthd = nv_rd32(priv, 0x00b234);
+	u32 data = nv_rd32(priv, 0x00b238);
+	u32 show = stat;
+	int chid;
+
+	engctx = nouveau_engctx_get(engine, inst);
+	chid   = pfifo->chid(pfifo, engctx);
+
+	if (stat & 0x01000000) {
+		/* happens on initial binding of the object */
+		if (type == 0x00000020 && mthd == 0x0000) {
+			nv_mask(priv, 0x00b308, 0x00000000, 0x00000000);
+			show &= ~0x01000000;
+		}
+
+		if (type == 0x00000010) {
+			handle = nouveau_handle_get_class(engctx, 0x3174);
+			if (handle && !nv_call(handle->object, mthd, data))
+				show &= ~0x01000000;
+			nouveau_handle_put(handle);
+		}
+	}
+
+	nv_wr32(priv, 0x00b100, stat);
+	nv_wr32(priv, 0x00b230, 0x00000001);
+
+	if (show) {
+		nv_error(priv, "ch %d [0x%08x] 0x%08x 0x%08x 0x%08x 0x%08x\n",
+			 chid, inst << 4, stat, type, mthd, data);
+	}
+
+	nouveau_engctx_put(engctx);
+}
+
+static int
+nv31_mpeg_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv31_mpeg_priv *priv;
+	int ret;
+
+	ret = nouveau_mpeg_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00000002;
+	nv_subdev(priv)->intr = nv31_mpeg_intr;
+	nv_engine(priv)->cclass = &nv31_mpeg_cclass;
+	nv_engine(priv)->sclass = nv31_mpeg_sclass;
+	nv_engine(priv)->tile_prog = nv31_mpeg_tile_prog;
+	return 0;
+}
+
+int
+nv31_mpeg_init(struct nouveau_object *object)
+{
+	struct nouveau_engine *engine = nv_engine(object->engine);
+	struct nv31_mpeg_priv *priv = (void *)engine;
+	struct nouveau_fb *pfb = nouveau_fb(object);
+	int ret, i;
+
+	ret = nouveau_mpeg_init(&priv->base);
+	if (ret)
+		return ret;
+
+	/* VPE init */
+	nv_wr32(priv, 0x00b0e0, 0x00000020); /* nvidia: rd 0x01, wr 0x20 */
+	nv_wr32(priv, 0x00b0e8, 0x00000020); /* nvidia: rd 0x01, wr 0x20 */
+
+	for (i = 0; i < pfb->tile.regions; i++)
+		engine->tile_prog(engine, i);
+
+	/* PMPEG init */
+	nv_wr32(priv, 0x00b32c, 0x00000000);
+	nv_wr32(priv, 0x00b314, 0x00000100);
+	nv_wr32(priv, 0x00b220, nv44_graph_class(priv) ? 0x00000044 : 0x00000031);
+	nv_wr32(priv, 0x00b300, 0x02001ec1);
+	nv_mask(priv, 0x00b32c, 0x00000001, 0x00000001);
+
+	nv_wr32(priv, 0x00b100, 0xffffffff);
+	nv_wr32(priv, 0x00b140, 0xffffffff);
+
+	if (!nv_wait(priv, 0x00b200, 0x00000001, 0x00000000)) {
+		nv_error(priv, "timeout 0x%08x\n", nv_rd32(priv, 0x00b200));
+		return -EBUSY;
+	}
+
+	return 0;
+}
+
+struct nouveau_oclass
+nv31_mpeg_oclass = {
+	.handle = NV_ENGINE(MPEG, 0x31),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv31_mpeg_ctor,
+		.dtor = _nouveau_mpeg_dtor,
+		.init = nv31_mpeg_init,
+		.fini = _nouveau_mpeg_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/mpeg/nv40.c b/drivers/gpu/drm/nouveau/core/engine/mpeg/nv40.c
new file mode 100644
index 0000000..1241857
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/mpeg/nv40.c
@@ -0,0 +1,144 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <subdev/fb.h>
+#include <subdev/timer.h>
+#include <subdev/instmem.h>
+
+#include <engine/mpeg.h>
+#include <engine/graph/nv40.h>
+
+struct nv40_mpeg_priv {
+	struct nouveau_mpeg base;
+};
+
+struct nv40_mpeg_chan {
+	struct nouveau_mpeg base;
+};
+
+/*******************************************************************************
+ * PMPEG context
+ ******************************************************************************/
+
+static int
+nv40_mpeg_context_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nv40_mpeg_chan *chan;
+	int ret;
+
+	ret = nouveau_mpeg_context_create(parent, engine, oclass, NULL,
+					  264 * 4, 16,
+					  NVOBJ_FLAG_ZERO_ALLOC, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int
+nv40_mpeg_context_fini(struct nouveau_object *object, bool suspend)
+{
+
+	struct nv40_mpeg_priv *priv = (void *)object->engine;
+	struct nv40_mpeg_chan *chan = (void *)object;
+	u32 inst = 0x80000000 | nv_gpuobj(chan)->addr >> 4;
+
+	nv_mask(priv, 0x00b32c, 0x00000001, 0x00000000);
+	if (nv_rd32(priv, 0x00b318) == inst)
+		nv_mask(priv, 0x00b318, 0x80000000, 0x00000000);
+	nv_mask(priv, 0x00b32c, 0x00000001, 0x00000001);
+	return 0;
+}
+
+static struct nouveau_oclass
+nv40_mpeg_cclass = {
+	.handle = NV_ENGCTX(MPEG, 0x40),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv40_mpeg_context_ctor,
+		.dtor = _nouveau_mpeg_context_dtor,
+		.init = _nouveau_mpeg_context_init,
+		.fini = nv40_mpeg_context_fini,
+		.rd32 = _nouveau_mpeg_context_rd32,
+		.wr32 = _nouveau_mpeg_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PMPEG engine/subdev functions
+ ******************************************************************************/
+
+static void
+nv40_mpeg_intr(struct nouveau_subdev *subdev)
+{
+	struct nv40_mpeg_priv *priv = (void *)subdev;
+	u32 stat;
+
+	if ((stat = nv_rd32(priv, 0x00b100)))
+		nv31_mpeg_intr(subdev);
+
+	if ((stat = nv_rd32(priv, 0x00b800))) {
+		nv_error(priv, "PMSRCH 0x%08x\n", stat);
+		nv_wr32(priv, 0x00b800, stat);
+	}
+}
+
+static int
+nv40_mpeg_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv40_mpeg_priv *priv;
+	int ret;
+
+	ret = nouveau_mpeg_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00000002;
+	nv_subdev(priv)->intr = nv40_mpeg_intr;
+	nv_engine(priv)->cclass = &nv40_mpeg_cclass;
+	nv_engine(priv)->sclass = nv31_mpeg_sclass;
+	nv_engine(priv)->tile_prog = nv31_mpeg_tile_prog;
+	return 0;
+}
+
+struct nouveau_oclass
+nv40_mpeg_oclass = {
+	.handle = NV_ENGINE(MPEG, 0x40),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv40_mpeg_ctor,
+		.dtor = _nouveau_mpeg_dtor,
+		.init = nv31_mpeg_init,
+		.fini = _nouveau_mpeg_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/mpeg/nv50.c b/drivers/gpu/drm/nouveau/core/engine/mpeg/nv50.c
new file mode 100644
index 0000000..8678a99
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/mpeg/nv50.c
@@ -0,0 +1,240 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <subdev/vm.h>
+#include <subdev/bar.h>
+#include <subdev/timer.h>
+
+#include <engine/mpeg.h>
+
+struct nv50_mpeg_priv {
+	struct nouveau_mpeg base;
+};
+
+struct nv50_mpeg_chan {
+	struct nouveau_mpeg_chan base;
+};
+
+/*******************************************************************************
+ * MPEG object classes
+ ******************************************************************************/
+
+static int
+nv50_mpeg_object_ctor(struct nouveau_object *parent,
+		      struct nouveau_object *engine,
+		      struct nouveau_oclass *oclass, void *data, u32 size,
+		      struct nouveau_object **pobject)
+{
+	struct nouveau_gpuobj *obj;
+	int ret;
+
+	ret = nouveau_gpuobj_create(parent, engine, oclass, 0, parent,
+				    16, 16, 0, &obj);
+	*pobject = nv_object(obj);
+	if (ret)
+		return ret;
+
+	nv_wo32(obj, 0x00, nv_mclass(obj));
+	nv_wo32(obj, 0x04, 0x00000000);
+	nv_wo32(obj, 0x08, 0x00000000);
+	nv_wo32(obj, 0x0c, 0x00000000);
+	return 0;
+}
+
+struct nouveau_ofuncs
+nv50_mpeg_ofuncs = {
+	.ctor = nv50_mpeg_object_ctor,
+	.dtor = _nouveau_gpuobj_dtor,
+	.init = _nouveau_gpuobj_init,
+	.fini = _nouveau_gpuobj_fini,
+	.rd32 = _nouveau_gpuobj_rd32,
+	.wr32 = _nouveau_gpuobj_wr32,
+};
+
+static struct nouveau_oclass
+nv50_mpeg_sclass[] = {
+	{ 0x3174, &nv50_mpeg_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * PMPEG context
+ ******************************************************************************/
+
+int
+nv50_mpeg_context_ctor(struct nouveau_object *parent,
+		       struct nouveau_object *engine,
+		       struct nouveau_oclass *oclass, void *data, u32 size,
+		       struct nouveau_object **pobject)
+{
+	struct nouveau_bar *bar = nouveau_bar(parent);
+	struct nv50_mpeg_chan *chan;
+	int ret;
+
+	ret = nouveau_mpeg_context_create(parent, engine, oclass, NULL, 128 * 4,
+					  0, NVOBJ_FLAG_ZERO_ALLOC, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	nv_wo32(chan, 0x0070, 0x00801ec1);
+	nv_wo32(chan, 0x007c, 0x0000037c);
+	bar->flush(bar);
+	return 0;
+}
+
+static struct nouveau_oclass
+nv50_mpeg_cclass = {
+	.handle = NV_ENGCTX(MPEG, 0x50),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv50_mpeg_context_ctor,
+		.dtor = _nouveau_mpeg_context_dtor,
+		.init = _nouveau_mpeg_context_init,
+		.fini = _nouveau_mpeg_context_fini,
+		.rd32 = _nouveau_mpeg_context_rd32,
+		.wr32 = _nouveau_mpeg_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PMPEG engine/subdev functions
+ ******************************************************************************/
+
+int
+nv50_mpeg_tlb_flush(struct nouveau_engine *engine)
+{
+	nv50_vm_flush_engine(&engine->base, 0x08);
+	return 0;
+}
+
+void
+nv50_mpeg_intr(struct nouveau_subdev *subdev)
+{
+	struct nv50_mpeg_priv *priv = (void *)subdev;
+	u32 stat = nv_rd32(priv, 0x00b100);
+	u32 type = nv_rd32(priv, 0x00b230);
+	u32 mthd = nv_rd32(priv, 0x00b234);
+	u32 data = nv_rd32(priv, 0x00b238);
+	u32 show = stat;
+
+	if (stat & 0x01000000) {
+		/* happens on initial binding of the object */
+		if (type == 0x00000020 && mthd == 0x0000) {
+			nv_wr32(priv, 0x00b308, 0x00000100);
+			show &= ~0x01000000;
+		}
+	}
+
+	if (show) {
+		nv_info(priv, "0x%08x 0x%08x 0x%08x 0x%08x\n",
+			stat, type, mthd, data);
+	}
+
+	nv_wr32(priv, 0x00b100, stat);
+	nv_wr32(priv, 0x00b230, 0x00000001);
+	nv50_fb_trap(nouveau_fb(priv), 1);
+}
+
+static void
+nv50_vpe_intr(struct nouveau_subdev *subdev)
+{
+	struct nv50_mpeg_priv *priv = (void *)subdev;
+
+	if (nv_rd32(priv, 0x00b100))
+		nv50_mpeg_intr(subdev);
+
+	if (nv_rd32(priv, 0x00b800)) {
+		u32 stat = nv_rd32(priv, 0x00b800);
+		nv_info(priv, "PMSRCH: 0x%08x\n", stat);
+		nv_wr32(priv, 0xb800, stat);
+	}
+}
+
+static int
+nv50_mpeg_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv50_mpeg_priv *priv;
+	int ret;
+
+	ret = nouveau_mpeg_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00400002;
+	nv_subdev(priv)->intr = nv50_vpe_intr;
+	nv_engine(priv)->cclass = &nv50_mpeg_cclass;
+	nv_engine(priv)->sclass = nv50_mpeg_sclass;
+	nv_engine(priv)->tlb_flush = nv50_mpeg_tlb_flush;
+	return 0;
+}
+
+int
+nv50_mpeg_init(struct nouveau_object *object)
+{
+	struct nv50_mpeg_priv *priv = (void *)object;
+	int ret;
+
+	ret = nouveau_mpeg_init(&priv->base);
+	if (ret)
+		return ret;
+
+	nv_wr32(priv, 0x00b32c, 0x00000000);
+	nv_wr32(priv, 0x00b314, 0x00000100);
+	nv_wr32(priv, 0x00b0e0, 0x0000001a);
+
+	nv_wr32(priv, 0x00b220, 0x00000044);
+	nv_wr32(priv, 0x00b300, 0x00801ec1);
+	nv_wr32(priv, 0x00b390, 0x00000000);
+	nv_wr32(priv, 0x00b394, 0x00000000);
+	nv_wr32(priv, 0x00b398, 0x00000000);
+	nv_mask(priv, 0x00b32c, 0x00000001, 0x00000001);
+
+	nv_wr32(priv, 0x00b100, 0xffffffff);
+	nv_wr32(priv, 0x00b140, 0xffffffff);
+
+	if (!nv_wait(priv, 0x00b200, 0x00000001, 0x00000000)) {
+		nv_error(priv, "timeout 0x%08x\n", nv_rd32(priv, 0x00b200));
+		return -EBUSY;
+	}
+
+	return 0;
+}
+
+struct nouveau_oclass
+nv50_mpeg_oclass = {
+	.handle = NV_ENGINE(MPEG, 0x50),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv50_mpeg_ctor,
+		.dtor = _nouveau_mpeg_dtor,
+		.init = nv50_mpeg_init,
+		.fini = _nouveau_mpeg_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/mpeg/nv84.c b/drivers/gpu/drm/nouveau/core/engine/mpeg/nv84.c
new file mode 100644
index 0000000..8f805b4
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/mpeg/nv84.c
@@ -0,0 +1,104 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <subdev/vm.h>
+#include <subdev/bar.h>
+#include <subdev/timer.h>
+
+#include <engine/mpeg.h>
+
+struct nv84_mpeg_priv {
+	struct nouveau_mpeg base;
+};
+
+struct nv84_mpeg_chan {
+	struct nouveau_mpeg_chan base;
+};
+
+/*******************************************************************************
+ * MPEG object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv84_mpeg_sclass[] = {
+	{ 0x8274, &nv50_mpeg_ofuncs },
+	{}
+};
+
+/*******************************************************************************
+ * PMPEG context
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv84_mpeg_cclass = {
+	.handle = NV_ENGCTX(MPEG, 0x84),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv50_mpeg_context_ctor,
+		.dtor = _nouveau_mpeg_context_dtor,
+		.init = _nouveau_mpeg_context_init,
+		.fini = _nouveau_mpeg_context_fini,
+		.rd32 = _nouveau_mpeg_context_rd32,
+		.wr32 = _nouveau_mpeg_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PMPEG engine/subdev functions
+ ******************************************************************************/
+
+static int
+nv84_mpeg_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct nv84_mpeg_priv *priv;
+	int ret;
+
+	ret = nouveau_mpeg_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00000002;
+	nv_subdev(priv)->intr = nv50_mpeg_intr;
+	nv_engine(priv)->cclass = &nv84_mpeg_cclass;
+	nv_engine(priv)->sclass = nv84_mpeg_sclass;
+	nv_engine(priv)->tlb_flush = nv50_mpeg_tlb_flush;
+	return 0;
+}
+
+struct nouveau_oclass
+nv84_mpeg_oclass = {
+	.handle = NV_ENGINE(MPEG, 0x84),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv84_mpeg_ctor,
+		.dtor = _nouveau_mpeg_dtor,
+		.init = nv50_mpeg_init,
+		.fini = _nouveau_mpeg_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/ppp/nv98.c b/drivers/gpu/drm/nouveau/core/engine/ppp/nv98.c
new file mode 100644
index 0000000..50e7e0d
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/ppp/nv98.c
@@ -0,0 +1,175 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <engine/ppp.h>
+
+struct nv98_ppp_priv {
+	struct nouveau_ppp base;
+};
+
+struct nv98_ppp_chan {
+	struct nouveau_ppp_chan base;
+};
+
+/*******************************************************************************
+ * PPP object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv98_ppp_sclass[] = {
+	{},
+};
+
+/*******************************************************************************
+ * PPPP context
+ ******************************************************************************/
+
+static int
+nv98_ppp_context_ctor(struct nouveau_object *parent,
+		      struct nouveau_object *engine,
+		      struct nouveau_oclass *oclass, void *data, u32 size,
+		      struct nouveau_object **pobject)
+{
+	struct nv98_ppp_chan *priv;
+	int ret;
+
+	ret = nouveau_ppp_context_create(parent, engine, oclass, NULL,
+					 0, 0, 0, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static void
+nv98_ppp_context_dtor(struct nouveau_object *object)
+{
+	struct nv98_ppp_chan *priv = (void *)object;
+	nouveau_ppp_context_destroy(&priv->base);
+}
+
+static int
+nv98_ppp_context_init(struct nouveau_object *object)
+{
+	struct nv98_ppp_chan *priv = (void *)object;
+	int ret;
+
+	ret = nouveau_ppp_context_init(&priv->base);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int
+nv98_ppp_context_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv98_ppp_chan *priv = (void *)object;
+	return nouveau_ppp_context_fini(&priv->base, suspend);
+}
+
+static struct nouveau_oclass
+nv98_ppp_cclass = {
+	.handle = NV_ENGCTX(PPP, 0x98),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv98_ppp_context_ctor,
+		.dtor = nv98_ppp_context_dtor,
+		.init = nv98_ppp_context_init,
+		.fini = nv98_ppp_context_fini,
+		.rd32 = _nouveau_ppp_context_rd32,
+		.wr32 = _nouveau_ppp_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PPPP engine/subdev functions
+ ******************************************************************************/
+
+static void
+nv98_ppp_intr(struct nouveau_subdev *subdev)
+{
+}
+
+static int
+nv98_ppp_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	      struct nouveau_oclass *oclass, void *data, u32 size,
+	      struct nouveau_object **pobject)
+{
+	struct nv98_ppp_priv *priv;
+	int ret;
+
+	ret = nouveau_ppp_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x00400002;
+	nv_subdev(priv)->intr = nv98_ppp_intr;
+	nv_engine(priv)->cclass = &nv98_ppp_cclass;
+	nv_engine(priv)->sclass = nv98_ppp_sclass;
+	return 0;
+}
+
+static void
+nv98_ppp_dtor(struct nouveau_object *object)
+{
+	struct nv98_ppp_priv *priv = (void *)object;
+	nouveau_ppp_destroy(&priv->base);
+}
+
+static int
+nv98_ppp_init(struct nouveau_object *object)
+{
+	struct nv98_ppp_priv *priv = (void *)object;
+	int ret;
+
+	ret = nouveau_ppp_init(&priv->base);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int
+nv98_ppp_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv98_ppp_priv *priv = (void *)object;
+	return nouveau_ppp_fini(&priv->base, suspend);
+}
+
+struct nouveau_oclass
+nv98_ppp_oclass = {
+	.handle = NV_ENGINE(PPP, 0x98),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv98_ppp_ctor,
+		.dtor = nv98_ppp_dtor,
+		.init = nv98_ppp_init,
+		.fini = nv98_ppp_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/software/nv04.c b/drivers/gpu/drm/nouveau/core/engine/software/nv04.c
new file mode 100644
index 0000000..3ca4c3a
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/software/nv04.c
@@ -0,0 +1,147 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <engine/software.h>
+#include <engine/fifo.h>
+
+struct nv04_software_priv {
+	struct nouveau_software base;
+};
+
+struct nv04_software_chan {
+	struct nouveau_software_chan base;
+};
+
+/*******************************************************************************
+ * software object classes
+ ******************************************************************************/
+
+static int
+nv04_software_set_ref(struct nouveau_object *object, u32 mthd,
+		      void *data, u32 size)
+{
+	struct nouveau_object *channel = (void *)nv_engctx(object->parent);
+	struct nouveau_fifo_chan *fifo = (void *)channel->parent;
+	atomic_set(&fifo->refcnt, *(u32*)data);
+	return 0;
+}
+
+static int
+nv04_software_flip(struct nouveau_object *object, u32 mthd,
+		   void *args, u32 size)
+{
+	struct nv04_software_chan *chan = (void *)nv_engctx(object->parent);
+	if (chan->base.flip)
+		return chan->base.flip(chan->base.flip_data);
+	return -EINVAL;
+}
+
+static struct nouveau_omthds
+nv04_software_omthds[] = {
+	{ 0x0150, nv04_software_set_ref },
+	{ 0x0500, nv04_software_flip },
+	{}
+};
+
+static struct nouveau_oclass
+nv04_software_sclass[] = {
+	{ 0x006e, &nouveau_object_ofuncs, nv04_software_omthds },
+	{}
+};
+
+/*******************************************************************************
+ * software context
+ ******************************************************************************/
+
+static int
+nv04_software_context_ctor(struct nouveau_object *parent,
+		      struct nouveau_object *engine,
+		      struct nouveau_oclass *oclass, void *data, u32 size,
+		      struct nouveau_object **pobject)
+{
+	struct nv04_software_chan *chan;
+	int ret;
+
+	ret = nouveau_software_context_create(parent, engine, oclass, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static struct nouveau_oclass
+nv04_software_cclass = {
+	.handle = NV_ENGCTX(SW, 0x04),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv04_software_context_ctor,
+		.dtor = _nouveau_software_context_dtor,
+		.init = _nouveau_software_context_init,
+		.fini = _nouveau_software_context_fini,
+	},
+};
+
+/*******************************************************************************
+ * software engine/subdev functions
+ ******************************************************************************/
+
+void
+nv04_software_intr(struct nouveau_subdev *subdev)
+{
+	nv_mask(subdev, 0x000100, 0x80000000, 0x00000000);
+}
+
+static int
+nv04_software_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	      struct nouveau_oclass *oclass, void *data, u32 size,
+	      struct nouveau_object **pobject)
+{
+	struct nv04_software_priv *priv;
+	int ret;
+
+	ret = nouveau_software_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_engine(priv)->cclass = &nv04_software_cclass;
+	nv_engine(priv)->sclass = nv04_software_sclass;
+	nv_subdev(priv)->intr = nv04_software_intr;
+	return 0;
+}
+
+struct nouveau_oclass
+nv04_software_oclass = {
+	.handle = NV_ENGINE(SW, 0x04),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv04_software_ctor,
+		.dtor = _nouveau_software_dtor,
+		.init = _nouveau_software_init,
+		.fini = _nouveau_software_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/software/nv10.c b/drivers/gpu/drm/nouveau/core/engine/software/nv10.c
new file mode 100644
index 0000000..6e699af
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/software/nv10.c
@@ -0,0 +1,129 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <engine/software.h>
+
+struct nv10_software_priv {
+	struct nouveau_software base;
+};
+
+struct nv10_software_chan {
+	struct nouveau_software_chan base;
+};
+
+/*******************************************************************************
+ * software object classes
+ ******************************************************************************/
+
+static int
+nv10_software_flip(struct nouveau_object *object, u32 mthd,
+		   void *args, u32 size)
+{
+	struct nv10_software_chan *chan = (void *)nv_engctx(object->parent);
+	if (chan->base.flip)
+		return chan->base.flip(chan->base.flip_data);
+	return -EINVAL;
+}
+
+static struct nouveau_omthds
+nv10_software_omthds[] = {
+	{ 0x0500, nv10_software_flip },
+	{}
+};
+
+static struct nouveau_oclass
+nv10_software_sclass[] = {
+	{ 0x016e, &nouveau_object_ofuncs, nv10_software_omthds },
+	{}
+};
+
+/*******************************************************************************
+ * software context
+ ******************************************************************************/
+
+static int
+nv10_software_context_ctor(struct nouveau_object *parent,
+		      struct nouveau_object *engine,
+		      struct nouveau_oclass *oclass, void *data, u32 size,
+		      struct nouveau_object **pobject)
+{
+	struct nv10_software_chan *chan;
+	int ret;
+
+	ret = nouveau_software_context_create(parent, engine, oclass, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static struct nouveau_oclass
+nv10_software_cclass = {
+	.handle = NV_ENGCTX(SW, 0x04),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv10_software_context_ctor,
+		.dtor = _nouveau_software_context_dtor,
+		.init = _nouveau_software_context_init,
+		.fini = _nouveau_software_context_fini,
+	},
+};
+
+/*******************************************************************************
+ * software engine/subdev functions
+ ******************************************************************************/
+
+static int
+nv10_software_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	      struct nouveau_oclass *oclass, void *data, u32 size,
+	      struct nouveau_object **pobject)
+{
+	struct nv10_software_priv *priv;
+	int ret;
+
+	ret = nouveau_software_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_engine(priv)->cclass = &nv10_software_cclass;
+	nv_engine(priv)->sclass = nv10_software_sclass;
+	nv_subdev(priv)->intr = nv04_software_intr;
+	return 0;
+}
+
+struct nouveau_oclass
+nv10_software_oclass = {
+	.handle = NV_ENGINE(SW, 0x10),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv10_software_ctor,
+		.dtor = _nouveau_software_dtor,
+		.init = _nouveau_software_init,
+		.fini = _nouveau_software_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/software/nv50.c b/drivers/gpu/drm/nouveau/core/engine/software/nv50.c
new file mode 100644
index 0000000..a2edcd3
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/software/nv50.c
@@ -0,0 +1,199 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+#include <core/namedb.h>
+#include <core/handle.h>
+#include <core/gpuobj.h>
+
+#include <engine/software.h>
+#include <engine/disp.h>
+
+struct nv50_software_priv {
+	struct nouveau_software base;
+};
+
+struct nv50_software_chan {
+	struct nouveau_software_chan base;
+};
+
+/*******************************************************************************
+ * software object classes
+ ******************************************************************************/
+
+static int
+nv50_software_mthd_dma_vblsem(struct nouveau_object *object, u32 mthd,
+			      void *args, u32 size)
+{
+	struct nv50_software_chan *chan = (void *)nv_engctx(object->parent);
+	struct nouveau_fifo_chan *fifo = (void *)nv_object(chan)->parent;
+	struct nouveau_handle *handle;
+	int ret = -EINVAL;
+
+	handle = nouveau_namedb_get(nv_namedb(fifo), *(u32 *)args);
+	if (!handle)
+		return -ENOENT;
+
+	if (nv_iclass(handle->object, NV_GPUOBJ_CLASS)) {
+		struct nouveau_gpuobj *gpuobj = nv_gpuobj(handle->object);
+		chan->base.vblank.ctxdma = gpuobj->node->offset >> 4;
+		ret = 0;
+	}
+	nouveau_namedb_put(handle);
+	return ret;
+}
+
+static int
+nv50_software_mthd_vblsem_offset(struct nouveau_object *object, u32 mthd,
+				 void *args, u32 size)
+{
+	struct nv50_software_chan *chan = (void *)nv_engctx(object->parent);
+	chan->base.vblank.offset = *(u32 *)args;
+	return 0;
+}
+
+static int
+nv50_software_mthd_vblsem_value(struct nouveau_object *object, u32 mthd,
+				void *args, u32 size)
+{
+	struct nv50_software_chan *chan = (void *)nv_engctx(object->parent);
+	chan->base.vblank.value = *(u32 *)args;
+	return 0;
+}
+
+static int
+nv50_software_mthd_vblsem_release(struct nouveau_object *object, u32 mthd,
+				  void *args, u32 size)
+{
+	struct nv50_software_chan *chan = (void *)nv_engctx(object->parent);
+	struct nouveau_disp *disp = nouveau_disp(object);
+	unsigned long flags;
+	u32 crtc = *(u32 *)args;
+
+	if (crtc > 1)
+		return -EINVAL;
+
+	disp->vblank.get(disp->vblank.data, crtc);
+
+	spin_lock_irqsave(&disp->vblank.lock, flags);
+	list_add(&chan->base.vblank.head, &disp->vblank.list);
+	chan->base.vblank.crtc = crtc;
+	spin_unlock_irqrestore(&disp->vblank.lock, flags);
+	return 0;
+}
+
+static int
+nv50_software_mthd_flip(struct nouveau_object *object, u32 mthd,
+			void *args, u32 size)
+{
+	struct nv50_software_chan *chan = (void *)nv_engctx(object->parent);
+	if (chan->base.flip)
+		return chan->base.flip(chan->base.flip_data);
+	return -EINVAL;
+}
+
+static struct nouveau_omthds
+nv50_software_omthds[] = {
+	{ 0x018c, nv50_software_mthd_dma_vblsem },
+	{ 0x0400, nv50_software_mthd_vblsem_offset },
+	{ 0x0404, nv50_software_mthd_vblsem_value },
+	{ 0x0408, nv50_software_mthd_vblsem_release },
+	{ 0x0500, nv50_software_mthd_flip },
+	{}
+};
+
+static struct nouveau_oclass
+nv50_software_sclass[] = {
+	{ 0x506e, &nouveau_object_ofuncs, nv50_software_omthds },
+	{}
+};
+
+/*******************************************************************************
+ * software context
+ ******************************************************************************/
+
+static int
+nv50_software_context_ctor(struct nouveau_object *parent,
+			   struct nouveau_object *engine,
+			   struct nouveau_oclass *oclass, void *data, u32 size,
+			   struct nouveau_object **pobject)
+{
+	struct nv50_software_chan *chan;
+	int ret;
+
+	ret = nouveau_software_context_create(parent, engine, oclass, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	chan->base.vblank.channel = nv_gpuobj(parent->parent)->addr >> 12;
+	return 0;
+}
+
+static struct nouveau_oclass
+nv50_software_cclass = {
+	.handle = NV_ENGCTX(SW, 0x50),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv50_software_context_ctor,
+		.dtor = _nouveau_software_context_dtor,
+		.init = _nouveau_software_context_init,
+		.fini = _nouveau_software_context_fini,
+	},
+};
+
+/*******************************************************************************
+ * software engine/subdev functions
+ ******************************************************************************/
+
+static int
+nv50_software_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	      struct nouveau_oclass *oclass, void *data, u32 size,
+	      struct nouveau_object **pobject)
+{
+	struct nv50_software_priv *priv;
+	int ret;
+
+	ret = nouveau_software_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_engine(priv)->cclass = &nv50_software_cclass;
+	nv_engine(priv)->sclass = nv50_software_sclass;
+	nv_subdev(priv)->intr = nv04_software_intr;
+	return 0;
+}
+
+struct nouveau_oclass
+nv50_software_oclass = {
+	.handle = NV_ENGINE(SW, 0x50),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv50_software_ctor,
+		.dtor = _nouveau_software_dtor,
+		.init = _nouveau_software_init,
+		.fini = _nouveau_software_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/software/nvc0.c b/drivers/gpu/drm/nouveau/core/engine/software/nvc0.c
new file mode 100644
index 0000000..b7b0d7e
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/software/nvc0.c
@@ -0,0 +1,181 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <engine/software.h>
+#include <engine/disp.h>
+
+struct nvc0_software_priv {
+	struct nouveau_software base;
+};
+
+struct nvc0_software_chan {
+	struct nouveau_software_chan base;
+};
+
+/*******************************************************************************
+ * software object classes
+ ******************************************************************************/
+
+static int
+nvc0_software_mthd_vblsem_offset(struct nouveau_object *object, u32 mthd,
+				 void *args, u32 size)
+{
+	struct nvc0_software_chan *chan = (void *)nv_engctx(object->parent);
+	u64 data = *(u32 *)args;
+	if (mthd == 0x0400) {
+		chan->base.vblank.offset &= 0x00ffffffffULL;
+		chan->base.vblank.offset |= data << 32;
+	} else {
+		chan->base.vblank.offset &= 0xff00000000ULL;
+		chan->base.vblank.offset |= data;
+	}
+	return 0;
+}
+
+static int
+nvc0_software_mthd_vblsem_value(struct nouveau_object *object, u32 mthd,
+				void *args, u32 size)
+{
+	struct nvc0_software_chan *chan = (void *)nv_engctx(object->parent);
+	chan->base.vblank.value = *(u32 *)args;
+	return 0;
+}
+
+static int
+nvc0_software_mthd_vblsem_release(struct nouveau_object *object, u32 mthd,
+				  void *args, u32 size)
+{
+	struct nvc0_software_chan *chan = (void *)nv_engctx(object->parent);
+	struct nouveau_disp *disp = nouveau_disp(object);
+	unsigned long flags;
+	u32 crtc = *(u32 *)args;
+
+	if ((nv_device(object)->card_type < NV_E0 && crtc > 1) || crtc > 3)
+		return -EINVAL;
+
+	disp->vblank.get(disp->vblank.data, crtc);
+
+	spin_lock_irqsave(&disp->vblank.lock, flags);
+	list_add(&chan->base.vblank.head, &disp->vblank.list);
+	chan->base.vblank.crtc = crtc;
+	spin_unlock_irqrestore(&disp->vblank.lock, flags);
+	return 0;
+}
+
+static int
+nvc0_software_mthd_flip(struct nouveau_object *object, u32 mthd,
+			void *args, u32 size)
+{
+	struct nvc0_software_chan *chan = (void *)nv_engctx(object->parent);
+	if (chan->base.flip)
+		return chan->base.flip(chan->base.flip_data);
+	return -EINVAL;
+}
+
+static struct nouveau_omthds
+nvc0_software_omthds[] = {
+	{ 0x0400, nvc0_software_mthd_vblsem_offset },
+	{ 0x0404, nvc0_software_mthd_vblsem_offset },
+	{ 0x0408, nvc0_software_mthd_vblsem_value },
+	{ 0x040c, nvc0_software_mthd_vblsem_release },
+	{ 0x0500, nvc0_software_mthd_flip },
+	{}
+};
+
+static struct nouveau_oclass
+nvc0_software_sclass[] = {
+	{ 0x906e, &nouveau_object_ofuncs, nvc0_software_omthds },
+	{}
+};
+
+/*******************************************************************************
+ * software context
+ ******************************************************************************/
+
+static int
+nvc0_software_context_ctor(struct nouveau_object *parent,
+			   struct nouveau_object *engine,
+			   struct nouveau_oclass *oclass, void *data, u32 size,
+			   struct nouveau_object **pobject)
+{
+	struct nvc0_software_chan *chan;
+	int ret;
+
+	ret = nouveau_software_context_create(parent, engine, oclass, &chan);
+	*pobject = nv_object(chan);
+	if (ret)
+		return ret;
+
+	chan->base.vblank.channel = nv_gpuobj(parent->parent)->addr >> 12;
+	return 0;
+}
+
+static struct nouveau_oclass
+nvc0_software_cclass = {
+	.handle = NV_ENGCTX(SW, 0xc0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvc0_software_context_ctor,
+		.dtor = _nouveau_software_context_dtor,
+		.init = _nouveau_software_context_init,
+		.fini = _nouveau_software_context_fini,
+	},
+};
+
+/*******************************************************************************
+ * software engine/subdev functions
+ ******************************************************************************/
+
+static int
+nvc0_software_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		   struct nouveau_oclass *oclass, void *data, u32 size,
+		   struct nouveau_object **pobject)
+{
+	struct nvc0_software_priv *priv;
+	int ret;
+
+	ret = nouveau_software_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_engine(priv)->cclass = &nvc0_software_cclass;
+	nv_engine(priv)->sclass = nvc0_software_sclass;
+	nv_subdev(priv)->intr = nv04_software_intr;
+	return 0;
+}
+
+struct nouveau_oclass
+nvc0_software_oclass = {
+	.handle = NV_ENGINE(SW, 0xc0),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvc0_software_ctor,
+		.dtor = _nouveau_software_dtor,
+		.init = _nouveau_software_init,
+		.fini = _nouveau_software_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/vp/nv84.c b/drivers/gpu/drm/nouveau/core/engine/vp/nv84.c
new file mode 100644
index 0000000..dd23c80
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/vp/nv84.c
@@ -0,0 +1,175 @@
+/*
+ * Copyright 2012 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <core/os.h>
+#include <core/class.h>
+#include <core/engctx.h>
+
+#include <engine/vp.h>
+
+struct nv84_vp_priv {
+	struct nouveau_vp base;
+};
+
+struct nv84_vp_chan {
+	struct nouveau_vp_chan base;
+};
+
+/*******************************************************************************
+ * VP object classes
+ ******************************************************************************/
+
+static struct nouveau_oclass
+nv84_vp_sclass[] = {
+	{},
+};
+
+/*******************************************************************************
+ * PVP context
+ ******************************************************************************/
+
+static int
+nv84_vp_context_ctor(struct nouveau_object *parent,
+		     struct nouveau_object *engine,
+		     struct nouveau_oclass *oclass, void *data, u32 size,
+		     struct nouveau_object **pobject)
+{
+	struct nv84_vp_chan *priv;
+	int ret;
+
+	ret = nouveau_vp_context_create(parent, engine, oclass, NULL,
+					0, 0, 0, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static void
+nv84_vp_context_dtor(struct nouveau_object *object)
+{
+	struct nv84_vp_chan *priv = (void *)object;
+	nouveau_vp_context_destroy(&priv->base);
+}
+
+static int
+nv84_vp_context_init(struct nouveau_object *object)
+{
+	struct nv84_vp_chan *priv = (void *)object;
+	int ret;
+
+	ret = nouveau_vp_context_init(&priv->base);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int
+nv84_vp_context_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv84_vp_chan *priv = (void *)object;
+	return nouveau_vp_context_fini(&priv->base, suspend);
+}
+
+static struct nouveau_oclass
+nv84_vp_cclass = {
+	.handle = NV_ENGCTX(VP, 0x84),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv84_vp_context_ctor,
+		.dtor = nv84_vp_context_dtor,
+		.init = nv84_vp_context_init,
+		.fini = nv84_vp_context_fini,
+		.rd32 = _nouveau_vp_context_rd32,
+		.wr32 = _nouveau_vp_context_wr32,
+	},
+};
+
+/*******************************************************************************
+ * PVP engine/subdev functions
+ ******************************************************************************/
+
+static void
+nv84_vp_intr(struct nouveau_subdev *subdev)
+{
+}
+
+static int
+nv84_vp_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	     struct nouveau_oclass *oclass, void *data, u32 size,
+	     struct nouveau_object **pobject)
+{
+	struct nv84_vp_priv *priv;
+	int ret;
+
+	ret = nouveau_vp_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->unit = 0x01020000;
+	nv_subdev(priv)->intr = nv84_vp_intr;
+	nv_engine(priv)->cclass = &nv84_vp_cclass;
+	nv_engine(priv)->sclass = nv84_vp_sclass;
+	return 0;
+}
+
+static void
+nv84_vp_dtor(struct nouveau_object *object)
+{
+	struct nv84_vp_priv *priv = (void *)object;
+	nouveau_vp_destroy(&priv->base);
+}
+
+static int
+nv84_vp_init(struct nouveau_object *object)
+{
+	struct nv84_vp_priv *priv = (void *)object;
+	int ret;
+
+	ret = nouveau_vp_init(&priv->base);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int
+nv84_vp_fini(struct nouveau_object *object, bool suspend)
+{
+	struct nv84_vp_priv *priv = (void *)object;
+	return nouveau_vp_fini(&priv->base, suspend);
+}
+
+struct nouveau_oclass
+nv84_vp_oclass = {
+	.handle = NV_ENGINE(VP, 0x84),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nv84_vp_ctor,
+		.dtor = nv84_vp_dtor,
+		.init = nv84_vp_init,
+		.fini = nv84_vp_fini,
+	},
+};