Merge remote branch 'nouveau/for-airlied' of ../drm-nouveau-next into drm-linus

* 'nouveau/for-airlied' of ../drm-nouveau-next: (21 commits)
  drm/nouveau: bail out of auxch transaction if we repeatedly recieve defers
  drm/nv50: implement gpio set/get routines
  drm/nv50: parse/use some more de-magiced parts of gpio table entries
  drm/nouveau: store raw gpio table entry in bios gpio structs
  drm/nv40: Init some tiling-related PGRAPH state.
  drm/nv50: Add NVA3 support in ctxprog/ctxvals generator.
  drm/nv50: another dodgy DP hack
  drm/nv50: punt hotplug irq handling out to workqueue
  drm/nv50: preserve an unknown SOR_MODECTRL value for DP encoders
  drm/nv50: Allow using the NVA3 new compute class.
  drm/nv50: cleanup properly if PDISPLAY init fails
  drm/nouveau: fixup the init failure paths some more
  drm/nv50: fix instmem init on IGPs if stolen mem crosses 4GiB mark
  drm/nv40: add LVDS table quirk for Dell Latitude D620
  drm/nv40: rework lvds table parsing
  drm/nouveau: detect vram amount once, and save the value
  drm/nouveau: remove some unused members from drm_nouveau_private
  drm/nouveau: Make use of TTM busy_placements.
  drm/nv50: add more 0x100c80 flushy magic
  drm/nv50: fix fbcon when framebuffer above 4GiB mark
  ...

24 files changed, 429 insertions, 241 deletions

diff --git a/drivers/gpu/drm/nouveau/Makefile b/drivers/gpu/drm/nouveau/Makefile
index 7f0d807..453df3f 100644
--- a/drivers/gpu/drm/nouveau/Makefile
+++ b/drivers/gpu/drm/nouveau/Makefile
@@ -22,7 +22,7 @@ nouveau-y := nouveau_drv.o nouveau_state.o nouveau_channel.o nouveau_mem.o \
              nv50_cursor.o nv50_display.o nv50_fbcon.o \
              nv04_dac.o nv04_dfp.o nv04_tv.o nv17_tv.o nv17_tv_modes.o \
              nv04_crtc.o nv04_display.o nv04_cursor.o nv04_fbcon.o \
-             nv17_gpio.o
+             nv17_gpio.o nv50_gpio.o
 
 nouveau-$(CONFIG_DRM_NOUVEAU_DEBUG) += nouveau_debugfs.o
 nouveau-$(CONFIG_COMPAT) += nouveau_ioc32.o
diff --git a/drivers/gpu/drm/nouveau/nouveau_bios.c b/drivers/gpu/drm/nouveau/nouveau_bios.c
index b5a9336..abc382a 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bios.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bios.c
@@ -2573,48 +2573,34 @@ init_gpio(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
 	 * each GPIO according to various values listed in each entry
 	 */
 
-	const uint32_t nv50_gpio_reg[4] = { 0xe104, 0xe108, 0xe280, 0xe284 };
+	struct drm_nouveau_private *dev_priv = bios->dev->dev_private;
 	const uint32_t nv50_gpio_ctl[2] = { 0xe100, 0xe28c };
-	const uint8_t *gpio_table = &bios->data[bios->dcb.gpio_table_ptr];
-	const uint8_t *gpio_entry;
 	int i;
 
-	if (!iexec->execute)
-		return 1;
-
-	if (bios->dcb.version != 0x40) {
-		NV_ERROR(bios->dev, "DCB table not version 4.0\n");
-		return 0;
-	}
-
-	if (!bios->dcb.gpio_table_ptr) {
-		NV_WARN(bios->dev, "Invalid pointer to INIT_8E table\n");
-		return 0;
+	if (dev_priv->card_type != NV_50) {
+		NV_ERROR(bios->dev, "INIT_GPIO on unsupported chipset\n");
+		return -ENODEV;
 	}
 
-	gpio_entry = gpio_table + gpio_table[1];
-	for (i = 0; i < gpio_table[2]; i++, gpio_entry += gpio_table[3]) {
-		uint32_t entry = ROM32(gpio_entry[0]), r, s, v;
-		int line = (entry & 0x0000001f);
+	if (!iexec->execute)
+		return 1;
 
-		BIOSLOG(bios, "0x%04X: Entry: 0x%08X\n", offset, entry);
+	for (i = 0; i < bios->dcb.gpio.entries; i++) {
+		struct dcb_gpio_entry *gpio = &bios->dcb.gpio.entry[i];
+		uint32_t r, s, v;
 
-		if ((entry & 0x0000ff00) == 0x0000ff00)
-			continue;
+		BIOSLOG(bios, "0x%04X: Entry: 0x%08X\n", offset, gpio->entry);
 
-		r = nv50_gpio_reg[line >> 3];
-		s = (line & 0x07) << 2;
-		v = bios_rd32(bios, r) & ~(0x00000003 << s);
-		if (entry & 0x01000000)
-			v |= (((entry & 0x60000000) >> 29) ^ 2) << s;
-		else
-			v |= (((entry & 0x18000000) >> 27) ^ 2) << s;
-		bios_wr32(bios, r, v);
+		nv50_gpio_set(bios->dev, gpio->tag, gpio->state_default);
 
-		r = nv50_gpio_ctl[line >> 4];
-		s = (line & 0x0f);
+		/* The NVIDIA binary driver doesn't appear to actually do
+		 * any of this, my VBIOS does however.
+		 */
+		/* Not a clue, needs de-magicing */
+		r = nv50_gpio_ctl[gpio->line >> 4];
+		s = (gpio->line & 0x0f);
 		v = bios_rd32(bios, r) & ~(0x00010001 << s);
-		switch ((entry & 0x06000000) >> 25) {
+		switch ((gpio->entry & 0x06000000) >> 25) {
 		case 1:
 			v |= (0x00000001 << s);
 			break;
@@ -3198,7 +3184,6 @@ static int run_lvds_table(struct drm_device *dev, struct dcb_entry *dcbent, int
 	struct nvbios *bios = &dev_priv->vbios;
 	unsigned int outputset = (dcbent->or == 4) ? 1 : 0;
 	uint16_t scriptptr = 0, clktable;
-	uint8_t clktableptr = 0;
 
 	/*
 	 * For now we assume version 3.0 table - g80 support will need some
@@ -3217,26 +3202,29 @@ static int run_lvds_table(struct drm_device *dev, struct dcb_entry *dcbent, int
 		scriptptr = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 11 + outputset * 2]);
 		break;
 	case LVDS_RESET:
+		clktable = bios->fp.lvdsmanufacturerpointer + 15;
+		if (dcbent->or == 4)
+			clktable += 8;
+
 		if (dcbent->lvdsconf.use_straps_for_mode) {
 			if (bios->fp.dual_link)
-				clktableptr += 2;
-			if (bios->fp.BITbit1)
-				clktableptr++;
+				clktable += 4;
+			if (bios->fp.if_is_24bit)
+				clktable += 2;
 		} else {
 			/* using EDID */
-			uint8_t fallback = bios->data[bios->fp.lvdsmanufacturerpointer + 4];
-			int fallbackcmpval = (dcbent->or == 4) ? 4 : 1;
+			int cmpval_24bit = (dcbent->or == 4) ? 4 : 1;
 
 			if (bios->fp.dual_link) {
-				clktableptr += 2;
-				fallbackcmpval *= 2;
+				clktable += 4;
+				cmpval_24bit <<= 1;
 			}
-			if (fallbackcmpval & fallback)
-				clktableptr++;
+
+			if (bios->fp.strapless_is_24bit & cmpval_24bit)
+				clktable += 2;
 		}
 
-		/* adding outputset * 8 may not be correct */
-		clktable = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 15 + clktableptr * 2 + outputset * 8]);
+		clktable = ROM16(bios->data[clktable]);
 		if (!clktable) {
 			NV_ERROR(dev, "Pixel clock comparison table not found\n");
 			return -ENOENT;
@@ -3638,37 +3626,40 @@ int nouveau_bios_parse_lvds_table(struct drm_device *dev, int pxclk, bool *dl, b
 		*if_is_24bit = bios->data[lvdsofs] & 16;
 		break;
 	case 0x30:
-		/*
-		 * My money would be on there being a 24 bit interface bit in
-		 * this table, but I have no example of a laptop bios with a
-		 * 24 bit panel to confirm that. Hence we shout loudly if any
-		 * bit other than bit 0 is set (I've not even seen bit 1)
-		 */
-		if (bios->data[lvdsofs] > 1)
-			NV_ERROR(dev,
-				 "You have a very unusual laptop display; please report it\n");
+	case 0x40:
 		/*
 		 * No sign of the "power off for reset" or "reset for panel
 		 * on" bits, but it's safer to assume we should
 		 */
 		bios->fp.power_off_for_reset = true;
 		bios->fp.reset_after_pclk_change = true;
+
 		/*
 		 * It's ok lvdsofs is wrong for nv4x edid case; dual_link is
-		 * over-written, and BITbit1 isn't used
+		 * over-written, and if_is_24bit isn't used
 		 */
 		bios->fp.dual_link = bios->data[lvdsofs] & 1;
-		bios->fp.BITbit1 = bios->data[lvdsofs] & 2;
-		bios->fp.duallink_transition_clk = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 5]) * 10;
-		break;
-	case 0x40:
-		bios->fp.dual_link = bios->data[lvdsofs] & 1;
 		bios->fp.if_is_24bit = bios->data[lvdsofs] & 2;
 		bios->fp.strapless_is_24bit = bios->data[bios->fp.lvdsmanufacturerpointer + 4];
 		bios->fp.duallink_transition_clk = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 5]) * 10;
 		break;
 	}
 
+	/* Dell Latitude D620 reports a too-high value for the dual-link
+	 * transition freq, causing us to program the panel incorrectly.
+	 *
+	 * It doesn't appear the VBIOS actually uses its transition freq
+	 * (90000kHz), instead it uses the "Number of LVDS channels" field
+	 * out of the panel ID structure (http://www.spwg.org/).
+	 *
+	 * For the moment, a quirk will do :)
+	 */
+	if ((dev->pdev->device == 0x01d7) &&
+	    (dev->pdev->subsystem_vendor == 0x1028) &&
+	    (dev->pdev->subsystem_device == 0x01c2)) {
+		bios->fp.duallink_transition_clk = 80000;
+	}
+
 	/* set dual_link flag for EDID case */
 	if (pxclk && (chip_version < 0x25 || chip_version > 0x28))
 		bios->fp.dual_link = (pxclk >= bios->fp.duallink_transition_clk);
@@ -5077,25 +5068,25 @@ parse_dcb30_gpio_entry(struct nvbios *bios, uint16_t offset)
 	gpio->tag = tag;
 	gpio->line = line;
 	gpio->invert = flags != 4;
+	gpio->entry = ent;
 }
 
 static void
 parse_dcb40_gpio_entry(struct nvbios *bios, uint16_t offset)
 {
+	uint32_t entry = ROM32(bios->data[offset]);
 	struct dcb_gpio_entry *gpio;
-	uint32_t ent = ROM32(bios->data[offset]);
-	uint8_t line = ent & 0x1f,
-		tag = ent >> 8 & 0xff;
 
-	if (tag == 0xff)
+	if ((entry & 0x0000ff00) == 0x0000ff00)
 		return;
 
 	gpio = new_gpio_entry(bios);
-
-	/* Currently unused, we may need more fields parsed at some
-	 * point. */
-	gpio->tag = tag;
-	gpio->line = line;
+	gpio->tag = (entry & 0x0000ff00) >> 8;
+	gpio->line = (entry & 0x0000001f) >> 0;
+	gpio->state_default = (entry & 0x01000000) >> 24;
+	gpio->state[0] = (entry & 0x18000000) >> 27;
+	gpio->state[1] = (entry & 0x60000000) >> 29;
+	gpio->entry = entry;
 }
 
 static void
diff --git a/drivers/gpu/drm/nouveau/nouveau_bios.h b/drivers/gpu/drm/nouveau/nouveau_bios.h
index 4f88e69..c0d7b0a 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bios.h
+++ b/drivers/gpu/drm/nouveau/nouveau_bios.h
@@ -49,6 +49,9 @@ struct dcb_gpio_entry {
 	enum dcb_gpio_tag tag;
 	int line;
 	bool invert;
+	uint32_t entry;
+	uint8_t state_default;
+	uint8_t state[2];
 };
 
 struct dcb_gpio_table {
@@ -267,7 +270,6 @@ struct nvbios {
 		bool reset_after_pclk_change;
 		bool dual_link;
 		bool link_c_increment;
-		bool BITbit1;
 		bool if_is_24bit;
 		int duallink_transition_clk;
 		uint8_t strapless_is_24bit;
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 0266124..eeab3fb 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -71,7 +71,7 @@ nouveau_bo_fixup_align(struct drm_device *dev,
 	 * many small buffers.
 	 */
 	if (dev_priv->card_type == NV_50) {
-		uint32_t block_size = nouveau_mem_fb_amount(dev) >> 15;
+		uint32_t block_size = dev_priv->vram_size >> 15;
 		int i;
 
 		switch (tile_flags) {
@@ -153,7 +153,7 @@ nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan,
 
 	nvbo->placement.fpfn = 0;
 	nvbo->placement.lpfn = mappable ? dev_priv->fb_mappable_pages : 0;
-	nouveau_bo_placement_set(nvbo, flags);
+	nouveau_bo_placement_set(nvbo, flags, 0);
 
 	nvbo->channel = chan;
 	ret = ttm_bo_init(&dev_priv->ttm.bdev, &nvbo->bo, size,
@@ -172,26 +172,33 @@ nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan,
 	return 0;
 }
 
+static void
+set_placement_list(uint32_t *pl, unsigned *n, uint32_t type, uint32_t flags)
+{
+	*n = 0;
+
+	if (type & TTM_PL_FLAG_VRAM)
+		pl[(*n)++] = TTM_PL_FLAG_VRAM | flags;
+	if (type & TTM_PL_FLAG_TT)
+		pl[(*n)++] = TTM_PL_FLAG_TT | flags;
+	if (type & TTM_PL_FLAG_SYSTEM)
+		pl[(*n)++] = TTM_PL_FLAG_SYSTEM | flags;
+}
+
 void
-nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t memtype)
+nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t type, uint32_t busy)
 {
-	int n = 0;
-
-	if (memtype & TTM_PL_FLAG_VRAM)
-		nvbo->placements[n++] = TTM_PL_FLAG_VRAM | TTM_PL_MASK_CACHING;
-	if (memtype & TTM_PL_FLAG_TT)
-		nvbo->placements[n++] = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING;
-	if (memtype & TTM_PL_FLAG_SYSTEM)
-		nvbo->placements[n++] = TTM_PL_FLAG_SYSTEM | TTM_PL_MASK_CACHING;
-	nvbo->placement.placement = nvbo->placements;
-	nvbo->placement.busy_placement = nvbo->placements;
-	nvbo->placement.num_placement = n;
-	nvbo->placement.num_busy_placement = n;
-
-	if (nvbo->pin_refcnt) {
-		while (n--)
-			nvbo->placements[n] |= TTM_PL_FLAG_NO_EVICT;
-	}
+	struct ttm_placement *pl = &nvbo->placement;
+	uint32_t flags = TTM_PL_MASK_CACHING |
+		(nvbo->pin_refcnt ? TTM_PL_FLAG_NO_EVICT : 0);
+
+	pl->placement = nvbo->placements;
+	set_placement_list(nvbo->placements, &pl->num_placement,
+			   type, flags);
+
+	pl->busy_placement = nvbo->busy_placements;
+	set_placement_list(nvbo->busy_placements, &pl->num_busy_placement,
+			   type | busy, flags);
 }
 
 int
@@ -199,7 +206,7 @@ nouveau_bo_pin(struct nouveau_bo *nvbo, uint32_t memtype)
 {
 	struct drm_nouveau_private *dev_priv = nouveau_bdev(nvbo->bo.bdev);
 	struct ttm_buffer_object *bo = &nvbo->bo;
-	int ret, i;
+	int ret;
 
 	if (nvbo->pin_refcnt && !(memtype & (1 << bo->mem.mem_type))) {
 		NV_ERROR(nouveau_bdev(bo->bdev)->dev,
@@ -215,9 +222,7 @@ nouveau_bo_pin(struct nouveau_bo *nvbo, uint32_t memtype)
 	if (ret)
 		goto out;
 
-	nouveau_bo_placement_set(nvbo, memtype);
-	for (i = 0; i < nvbo->placement.num_placement; i++)
-		nvbo->placements[i] |= TTM_PL_FLAG_NO_EVICT;
+	nouveau_bo_placement_set(nvbo, memtype, 0);
 
 	ret = ttm_bo_validate(bo, &nvbo->placement, false, false);
 	if (ret == 0) {
@@ -244,7 +249,7 @@ nouveau_bo_unpin(struct nouveau_bo *nvbo)
 {
 	struct drm_nouveau_private *dev_priv = nouveau_bdev(nvbo->bo.bdev);
 	struct ttm_buffer_object *bo = &nvbo->bo;
-	int ret, i;
+	int ret;
 
 	if (--nvbo->pin_refcnt)
 		return 0;
@@ -253,8 +258,7 @@ nouveau_bo_unpin(struct nouveau_bo *nvbo)
 	if (ret)
 		return ret;
 
-	for (i = 0; i < nvbo->placement.num_placement; i++)
-		nvbo->placements[i] &= ~TTM_PL_FLAG_NO_EVICT;
+	nouveau_bo_placement_set(nvbo, bo->mem.placement, 0);
 
 	ret = ttm_bo_validate(bo, &nvbo->placement, false, false);
 	if (ret == 0) {
@@ -395,8 +399,8 @@ nouveau_bo_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
 		man->io_addr = NULL;
 		man->io_offset = drm_get_resource_start(dev, 1);
 		man->io_size = drm_get_resource_len(dev, 1);
-		if (man->io_size > nouveau_mem_fb_amount(dev))
-			man->io_size = nouveau_mem_fb_amount(dev);
+		if (man->io_size > dev_priv->vram_size)
+			man->io_size = dev_priv->vram_size;
 
 		man->gpu_offset = dev_priv->vm_vram_base;
 		break;
@@ -439,10 +443,11 @@ nouveau_bo_evict_flags(struct ttm_buffer_object *bo, struct ttm_placement *pl)
 
 	switch (bo->mem.mem_type) {
 	case TTM_PL_VRAM:
-		nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_TT);
+		nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_TT,
+					 TTM_PL_FLAG_SYSTEM);
 		break;
 	default:
-		nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_SYSTEM);
+		nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_SYSTEM, 0);
 		break;
 	}
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_channel.c b/drivers/gpu/drm/nouveau/nouveau_channel.c
index 6dfb425..1fc57ef 100644
--- a/drivers/gpu/drm/nouveau/nouveau_channel.c
+++ b/drivers/gpu/drm/nouveau/nouveau_channel.c
@@ -142,7 +142,6 @@ nouveau_channel_alloc(struct drm_device *dev, struct nouveau_channel **chan_ret,
 					   GFP_KERNEL);
 	if (!dev_priv->fifos[channel])
 		return -ENOMEM;
-	dev_priv->fifo_alloc_count++;
 	chan = dev_priv->fifos[channel];
 	INIT_LIST_HEAD(&chan->nvsw.vbl_wait);
 	INIT_LIST_HEAD(&chan->fence.pending);
@@ -321,7 +320,6 @@ nouveau_channel_free(struct nouveau_channel *chan)
 		iounmap(chan->user);
 
 	dev_priv->fifos[chan->id] = NULL;
-	dev_priv->fifo_alloc_count--;
 	kfree(chan);
 }
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_debugfs.c b/drivers/gpu/drm/nouveau/nouveau_debugfs.c
index 8ff9ef5..a251886 100644
--- a/drivers/gpu/drm/nouveau/nouveau_debugfs.c
+++ b/drivers/gpu/drm/nouveau/nouveau_debugfs.c
@@ -137,10 +137,9 @@ nouveau_debugfs_memory_info(struct seq_file *m, void *data)
 {
 	struct drm_info_node *node = (struct drm_info_node *) m->private;
 	struct drm_minor *minor = node->minor;
-	struct drm_device *dev = minor->dev;
+	struct drm_nouveau_private *dev_priv = minor->dev->dev_private;
 
-	seq_printf(m, "VRAM total: %dKiB\n",
-		   (int)(nouveau_mem_fb_amount(dev) >> 10));
+	seq_printf(m, "VRAM total: %dKiB\n", (int)(dev_priv->vram_size >> 10));
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_dp.c b/drivers/gpu/drm/nouveau/nouveau_dp.c
index f954ad9..deeb21c 100644
--- a/drivers/gpu/drm/nouveau/nouveau_dp.c
+++ b/drivers/gpu/drm/nouveau/nouveau_dp.c
@@ -483,7 +483,7 @@ nouveau_dp_auxch(struct nouveau_i2c_chan *auxch, int cmd, int addr,
 	ctrl |= (cmd << NV50_AUXCH_CTRL_CMD_SHIFT);
 	ctrl |= ((data_nr - 1) << NV50_AUXCH_CTRL_LEN_SHIFT);
 
-	for (;;) {
+	for (i = 0; i < 16; i++) {
 		nv_wr32(dev, NV50_AUXCH_CTRL(index), ctrl | 0x80000000);
 		nv_wr32(dev, NV50_AUXCH_CTRL(index), ctrl);
 		nv_wr32(dev, NV50_AUXCH_CTRL(index), ctrl | 0x00010000);
@@ -502,6 +502,12 @@ nouveau_dp_auxch(struct nouveau_i2c_chan *auxch, int cmd, int addr,
 			break;
 	}
 
+	if (i == 16) {
+		NV_ERROR(dev, "auxch DEFER too many times, bailing\n");
+		ret = -EREMOTEIO;
+		goto out;
+	}
+
 	if (cmd & 1) {
 		if ((stat & NV50_AUXCH_STAT_COUNT) != data_nr) {
 			ret = -EREMOTEIO;
diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h
index d8b5590..ace630a 100644
--- a/drivers/gpu/drm/nouveau/nouveau_drv.h
+++ b/drivers/gpu/drm/nouveau/nouveau_drv.h
@@ -76,6 +76,7 @@ struct nouveau_bo {
 	struct ttm_buffer_object bo;
 	struct ttm_placement placement;
 	u32 placements[3];
+	u32 busy_placements[3];
 	struct ttm_bo_kmap_obj kmap;
 	struct list_head head;
 
@@ -519,6 +520,7 @@ struct drm_nouveau_private {
 
 	struct workqueue_struct *wq;
 	struct work_struct irq_work;
+	struct work_struct hpd_work;
 
 	struct list_head vbl_waiting;
 
@@ -533,7 +535,6 @@ struct drm_nouveau_private {
 
 	struct fb_info *fbdev_info;
 
-	int fifo_alloc_count;
 	struct nouveau_channel *fifos[NOUVEAU_MAX_CHANNEL_NR];
 
 	struct nouveau_engine engine;
@@ -553,12 +554,6 @@ struct drm_nouveau_private {
 	uint32_t ramro_offset;
 	uint32_t ramro_size;
 
-	/* base physical addresses */
-	uint64_t fb_phys;
-	uint64_t fb_available_size;
-	uint64_t fb_mappable_pages;
-	uint64_t fb_aper_free;
-
 	struct {
 		enum {
 			NOUVEAU_GART_NONE = 0,
@@ -572,10 +567,6 @@ struct drm_nouveau_private {
 		struct nouveau_gpuobj *sg_ctxdma;
 		struct page *sg_dummy_page;
 		dma_addr_t sg_dummy_bus;
-
-		/* nottm hack */
-		struct drm_ttm_backend *sg_be;
-		unsigned long sg_handle;
 	} gart_info;
 
 	/* nv10-nv40 tiling regions */
@@ -584,6 +575,16 @@ struct drm_nouveau_private {
 		spinlock_t lock;
 	} tile;
 
+	/* VRAM/fb configuration */
+	uint64_t vram_size;
+	uint64_t vram_sys_base;
+
+	uint64_t fb_phys;
+	uint64_t fb_available_size;
+	uint64_t fb_mappable_pages;
+	uint64_t fb_aper_free;
+	int fb_mtrr;
+
 	/* G8x/G9x virtual address space */
 	uint64_t vm_gart_base;
 	uint64_t vm_gart_size;
@@ -592,10 +593,6 @@ struct drm_nouveau_private {
 	uint64_t vm_end;
 	struct nouveau_gpuobj *vm_vram_pt[NV50_VM_VRAM_NR];
 	int vm_vram_pt_nr;
-	uint64_t vram_sys_base;
-
-	/* the mtrr covering the FB */
-	int fb_mtrr;
 
 	struct mem_block *ramin_heap;
 
@@ -614,11 +611,7 @@ struct drm_nouveau_private {
 	uint32_t dac_users[4];
 
 	struct nouveau_suspend_resume {
-		uint32_t fifo_mode;
-		uint32_t graph_ctx_control;
-		uint32_t graph_state;
 		uint32_t *ramin_copy;
-		uint64_t ramin_size;
 	} susres;
 
 	struct backlight_device *backlight;
@@ -717,7 +710,7 @@ extern struct mem_block *nouveau_mem_alloc_block(struct mem_block *,
 						 struct drm_file *, int tail);
 extern void nouveau_mem_takedown(struct mem_block **heap);
 extern void nouveau_mem_free_block(struct mem_block *);
-extern uint64_t nouveau_mem_fb_amount(struct drm_device *);
+extern int  nouveau_mem_detect(struct drm_device *dev);
 extern void nouveau_mem_release(struct drm_file *, struct mem_block *heap);
 extern int  nouveau_mem_init(struct drm_device *);
 extern int  nouveau_mem_init_agp(struct drm_device *);
@@ -1124,7 +1117,8 @@ extern int nouveau_bo_pin(struct nouveau_bo *, uint32_t flags);
 extern int nouveau_bo_unpin(struct nouveau_bo *);
 extern int nouveau_bo_map(struct nouveau_bo *);
 extern void nouveau_bo_unmap(struct nouveau_bo *);
-extern void nouveau_bo_placement_set(struct nouveau_bo *, uint32_t memtype);
+extern void nouveau_bo_placement_set(struct nouveau_bo *, uint32_t type,
+				     uint32_t busy);
 extern u16 nouveau_bo_rd16(struct nouveau_bo *nvbo, unsigned index);
 extern void nouveau_bo_wr16(struct nouveau_bo *nvbo, unsigned index, u16 val);
 extern u32 nouveau_bo_rd32(struct nouveau_bo *nvbo, unsigned index);
@@ -1168,6 +1162,10 @@ extern int nouveau_gem_ioctl_info(struct drm_device *, void *,
 int nv17_gpio_get(struct drm_device *dev, enum dcb_gpio_tag tag);
 int nv17_gpio_set(struct drm_device *dev, enum dcb_gpio_tag tag, int state);
 
+/* nv50_gpio.c */
+int nv50_gpio_get(struct drm_device *dev, enum dcb_gpio_tag tag);
+int nv50_gpio_set(struct drm_device *dev, enum dcb_gpio_tag tag, int state);
+
 #ifndef ioread32_native
 #ifdef __BIG_ENDIAN
 #define ioread16_native ioread16be
diff --git a/drivers/gpu/drm/nouveau/nouveau_encoder.h b/drivers/gpu/drm/nouveau/nouveau_encoder.h
index bc4a240..9f28b94 100644
--- a/drivers/gpu/drm/nouveau/nouveau_encoder.h
+++ b/drivers/gpu/drm/nouveau/nouveau_encoder.h
@@ -47,6 +47,7 @@ struct nouveau_encoder {
 
 	union {
 		struct {
+			int mc_unknown;
 			int dpcd_version;
 			int link_nr;
 			int link_bw;
diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c
index 0d22f66..1bc0b38 100644
--- a/drivers/gpu/drm/nouveau/nouveau_gem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_gem.c
@@ -180,40 +180,35 @@ nouveau_gem_set_domain(struct drm_gem_object *gem, uint32_t read_domains,
 {
 	struct nouveau_bo *nvbo = gem->driver_private;
 	struct ttm_buffer_object *bo = &nvbo->bo;
-	uint64_t flags;
+	uint32_t domains = valid_domains &
+		(write_domains ? write_domains : read_domains);
+	uint32_t pref_flags = 0, valid_flags = 0;
 
-	if (!valid_domains || (!read_domains && !write_domains))
+	if (!domains)
 		return -EINVAL;
 
-	if (write_domains) {
-		if ((valid_domains & NOUVEAU_GEM_DOMAIN_VRAM) &&
-		    (write_domains & NOUVEAU_GEM_DOMAIN_VRAM))
-			flags = TTM_PL_FLAG_VRAM;
-		else
-		if ((valid_domains & NOUVEAU_GEM_DOMAIN_GART) &&
-		    (write_domains & NOUVEAU_GEM_DOMAIN_GART))
-			flags = TTM_PL_FLAG_TT;
-		else
-			return -EINVAL;
-	} else {
-		if ((valid_domains & NOUVEAU_GEM_DOMAIN_VRAM) &&
-		    (read_domains & NOUVEAU_GEM_DOMAIN_VRAM) &&
-		    bo->mem.mem_type == TTM_PL_VRAM)
-			flags = TTM_PL_FLAG_VRAM;
-		else
-		if ((valid_domains & NOUVEAU_GEM_DOMAIN_GART) &&
-		    (read_domains & NOUVEAU_GEM_DOMAIN_GART) &&
-		    bo->mem.mem_type == TTM_PL_TT)
-			flags = TTM_PL_FLAG_TT;
-		else
-		if ((valid_domains & NOUVEAU_GEM_DOMAIN_VRAM) &&
-		    (read_domains & NOUVEAU_GEM_DOMAIN_VRAM))
-			flags = TTM_PL_FLAG_VRAM;
-		else
-			flags = TTM_PL_FLAG_TT;
-	}
+	if (valid_domains & NOUVEAU_GEM_DOMAIN_VRAM)
+		valid_flags |= TTM_PL_FLAG_VRAM;
+
+	if (valid_domains & NOUVEAU_GEM_DOMAIN_GART)
+		valid_flags |= TTM_PL_FLAG_TT;
+
+	if ((domains & NOUVEAU_GEM_DOMAIN_VRAM) &&
+	    bo->mem.mem_type == TTM_PL_VRAM)
+		pref_flags |= TTM_PL_FLAG_VRAM;
+
+	else if ((domains & NOUVEAU_GEM_DOMAIN_GART) &&
+		 bo->mem.mem_type == TTM_PL_TT)
+		pref_flags |= TTM_PL_FLAG_TT;
+
+	else if (domains & NOUVEAU_GEM_DOMAIN_VRAM)
+		pref_flags |= TTM_PL_FLAG_VRAM;
+
+	else
+		pref_flags |= TTM_PL_FLAG_TT;
+
+	nouveau_bo_placement_set(nvbo, pref_flags, valid_flags);
 
-	nouveau_bo_placement_set(nvbo, flags);
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_irq.c b/drivers/gpu/drm/nouveau/nouveau_irq.c
index 2bd59a9..13e73ce 100644
--- a/drivers/gpu/drm/nouveau/nouveau_irq.c
+++ b/drivers/gpu/drm/nouveau/nouveau_irq.c
@@ -51,6 +51,7 @@ nouveau_irq_preinstall(struct drm_device *dev)
 
 	if (dev_priv->card_type == NV_50) {
 		INIT_WORK(&dev_priv->irq_work, nv50_display_irq_handler_bh);
+		INIT_WORK(&dev_priv->hpd_work, nv50_display_irq_hotplug_bh);
 		INIT_LIST_HEAD(&dev_priv->vbl_waiting);
 	}
 }
diff --git a/drivers/gpu/drm/nouveau/nouveau_mem.c b/drivers/gpu/drm/nouveau/nouveau_mem.c
index 2dc09db..775a701 100644
--- a/drivers/gpu/drm/nouveau/nouveau_mem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_mem.c
@@ -347,6 +347,20 @@ nv50_mem_vm_bind_linear(struct drm_device *dev, uint64_t virt, uint32_t size,
 		return -EBUSY;
 	}
 
+	nv_wr32(dev, 0x100c80, 0x00040001);
+	if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+		NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+		NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80));
+		return -EBUSY;
+	}
+
+	nv_wr32(dev, 0x100c80, 0x00060001);
+	if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+		NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+		NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80));
+		return -EBUSY;
+	}
+
 	return 0;
 }
 
@@ -387,6 +401,20 @@ nv50_mem_vm_unbind(struct drm_device *dev, uint64_t virt, uint32_t size)
 	if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
 		NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
 		NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80));
+		return;
+	}
+
+	nv_wr32(dev, 0x100c80, 0x00040001);
+	if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+		NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+		NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80));
+		return;
+	}
+
+	nv_wr32(dev, 0x100c80, 0x00060001);
+	if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+		NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+		NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80));
 	}
 }
 
@@ -449,9 +477,30 @@ void nouveau_mem_close(struct drm_device *dev)
 	}
 }
 
-/*XXX won't work on BSD because of pci_read_config_dword */
 static uint32_t
-nouveau_mem_fb_amount_igp(struct drm_device *dev)
+nouveau_mem_detect_nv04(struct drm_device *dev)
+{
+	uint32_t boot0 = nv_rd32(dev, NV03_BOOT_0);
+
+	if (boot0 & 0x00000100)
+		return (((boot0 >> 12) & 0xf) * 2 + 2) * 1024 * 1024;
+
+	switch (boot0 & NV03_BOOT_0_RAM_AMOUNT) {
+	case NV04_BOOT_0_RAM_AMOUNT_32MB:
+		return 32 * 1024 * 1024;
+	case NV04_BOOT_0_RAM_AMOUNT_16MB:
+		return 16 * 1024 * 1024;
+	case NV04_BOOT_0_RAM_AMOUNT_8MB:
+		return 8 * 1024 * 1024;
+	case NV04_BOOT_0_RAM_AMOUNT_4MB:
+		return 4 * 1024 * 1024;
+	}
+
+	return 0;
+}
+
+static uint32_t
+nouveau_mem_detect_nforce(struct drm_device *dev)
 {
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct pci_dev *bridge;
@@ -463,11 +512,11 @@ nouveau_mem_fb_amount_igp(struct drm_device *dev)
 		return 0;
 	}
 
-	if (dev_priv->flags&NV_NFORCE) {
+	if (dev_priv->flags & NV_NFORCE) {
 		pci_read_config_dword(bridge, 0x7C, &mem);
 		return (uint64_t)(((mem >> 6) & 31) + 1)*1024*1024;
 	} else
-	if (dev_priv->flags&NV_NFORCE2) {
+	if (dev_priv->flags & NV_NFORCE2) {
 		pci_read_config_dword(bridge, 0x84, &mem);
 		return (uint64_t)(((mem >> 4) & 127) + 1)*1024*1024;
 	}
@@ -477,50 +526,32 @@ nouveau_mem_fb_amount_igp(struct drm_device *dev)
 }
 
 /* returns the amount of FB ram in bytes */
-uint64_t nouveau_mem_fb_amount(struct drm_device *dev)
+int
+nouveau_mem_detect(struct drm_device *dev)
 {
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
-	uint32_t boot0;
-
-	switch (dev_priv->card_type) {
-	case NV_04:
-		boot0 = nv_rd32(dev, NV03_BOOT_0);
-		if (boot0 & 0x00000100)
-			return (((boot0 >> 12) & 0xf) * 2 + 2) * 1024 * 1024;
-
-		switch (boot0 & NV03_BOOT_0_RAM_AMOUNT) {
-		case NV04_BOOT_0_RAM_AMOUNT_32MB:
-			return 32 * 1024 * 1024;
-		case NV04_BOOT_0_RAM_AMOUNT_16MB:
-			return 16 * 1024 * 1024;
-		case NV04_BOOT_0_RAM_AMOUNT_8MB:
-			return 8 * 1024 * 1024;
-		case NV04_BOOT_0_RAM_AMOUNT_4MB:
-			return 4 * 1024 * 1024;
-		}
-		break;
-	case NV_10:
-	case NV_20:
-	case NV_30:
-	case NV_40:
-	case NV_50:
-	default:
-		if (dev_priv->flags & (NV_NFORCE | NV_NFORCE2)) {
-			return nouveau_mem_fb_amount_igp(dev);
-		} else {
-			uint64_t mem;
-			mem = (nv_rd32(dev, NV04_FIFO_DATA) &
-					NV10_FIFO_DATA_RAM_AMOUNT_MB_MASK) >>
-					NV10_FIFO_DATA_RAM_AMOUNT_MB_SHIFT;
-			return mem * 1024 * 1024;
-		}
-		break;
+
+	if (dev_priv->card_type == NV_04) {
+		dev_priv->vram_size = nouveau_mem_detect_nv04(dev);
+	} else
+	if (dev_priv->flags & (NV_NFORCE | NV_NFORCE2)) {
+		dev_priv->vram_size = nouveau_mem_detect_nforce(dev);
+	} else {
+		dev_priv->vram_size  = nv_rd32(dev, NV04_FIFO_DATA);
+		dev_priv->vram_size &= NV10_FIFO_DATA_RAM_AMOUNT_MB_MASK;
+		if (dev_priv->chipset == 0xaa || dev_priv->chipset == 0xac)
+			dev_priv->vram_sys_base = nv_rd32(dev, 0x100e10) << 12;
 	}
 
-	NV_ERROR(dev,
-		"Unable to detect video ram size. Please report your setup to "
-							DRIVER_EMAIL "\n");
-	return 0;
+	NV_INFO(dev, "Detected %dMiB VRAM\n", (int)(dev_priv->vram_size >> 20));
+	if (dev_priv->vram_sys_base) {
+		NV_INFO(dev, "Stolen system memory at: 0x%010llx\n",
+			dev_priv->vram_sys_base);
+	}
+
+	if (dev_priv->vram_size)
+		return 0;
+	return -ENOMEM;
 }
 
 #if __OS_HAS_AGP
@@ -631,15 +662,12 @@ nouveau_mem_init(struct drm_device *dev)
 	spin_lock_init(&dev_priv->ttm.bo_list_lock);
 	spin_lock_init(&dev_priv->tile.lock);
 
-	dev_priv->fb_available_size = nouveau_mem_fb_amount(dev);
-
+	dev_priv->fb_available_size = dev_priv->vram_size;
 	dev_priv->fb_mappable_pages = dev_priv->fb_available_size;
 	if (dev_priv->fb_mappable_pages > drm_get_resource_len(dev, 1))
 		dev_priv->fb_mappable_pages = drm_get_resource_len(dev, 1);
 	dev_priv->fb_mappable_pages >>= PAGE_SHIFT;
 
-	NV_INFO(dev, "%d MiB VRAM\n", (int)(dev_priv->fb_available_size >> 20));
-
 	/* remove reserved space at end of vram from available amount */
 	dev_priv->fb_available_size -= dev_priv->ramin_rsvd_vram;
 	dev_priv->fb_aper_free = dev_priv->fb_available_size;
diff --git a/drivers/gpu/drm/nouveau/nouveau_sgdma.c b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
index ed15905..554fb45 100644
--- a/drivers/gpu/drm/nouveau/nouveau_sgdma.c
+++ b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
@@ -171,6 +171,24 @@ nouveau_sgdma_unbind(struct ttm_backend *be)
 	}
 	dev_priv->engine.instmem.finish_access(nvbe->dev);
 
+	if (dev_priv->card_type == NV_50) {
+		nv_wr32(dev, 0x100c80, 0x00050001);
+		if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+			NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+			NV_ERROR(dev, "0x100c80 = 0x%08x\n",
+						nv_rd32(dev, 0x100c80));
+			return -EBUSY;
+		}
+
+		nv_wr32(dev, 0x100c80, 0x00000001);
+		if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+			NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+			NV_ERROR(dev, "0x100c80 = 0x%08x\n",
+						nv_rd32(dev, 0x100c80));
+			return -EBUSY;
+		}
+	}
+
 	nvbe->bound = false;
 	return 0;
 }
diff --git a/drivers/gpu/drm/nouveau/nouveau_state.c b/drivers/gpu/drm/nouveau/nouveau_state.c
index 58b4680..a9e9cf3 100644
--- a/drivers/gpu/drm/nouveau/nouveau_state.c
+++ b/drivers/gpu/drm/nouveau/nouveau_state.c
@@ -340,7 +340,7 @@ nouveau_card_init_channel(struct drm_device *dev)
 
 	gpuobj = NULL;
 	ret = nouveau_gpuobj_dma_new(dev_priv->channel, NV_CLASS_DMA_IN_MEMORY,
-				     0, nouveau_mem_fb_amount(dev),
+				     0, dev_priv->vram_size,
 				     NV_DMA_ACCESS_RW, NV_DMA_TARGET_VIDMEM,
 				     &gpuobj);
 	if (ret)
@@ -426,6 +426,10 @@ nouveau_card_init(struct drm_device *dev)
 			goto out;
 	}
 
+	ret = nouveau_mem_detect(dev);
+	if (ret)
+		goto out_bios;
+
 	ret = nouveau_gpuobj_early_init(dev);
 	if (ret)
 		goto out_bios;
@@ -501,7 +505,7 @@ nouveau_card_init(struct drm_device *dev)
 		else
 			ret = nv04_display_create(dev);
 		if (ret)
-			goto out_irq;
+			goto out_channel;
 	}
 
 	ret = nouveau_backlight_init(dev);
@@ -515,6 +519,11 @@ nouveau_card_init(struct drm_device *dev)
 
 	return 0;
 
+out_channel:
+	if (dev_priv->channel) {
+		nouveau_channel_free(dev_priv->channel);
+		dev_priv->channel = NULL;
+	}
 out_irq:
 	drm_irq_uninstall(dev);
 out_fifo:
@@ -532,6 +541,7 @@ out_mc:
 out_gpuobj:
 	nouveau_gpuobj_takedown(dev);
 out_mem:
+	nouveau_sgdma_takedown(dev);
 	nouveau_mem_close(dev);
 out_instmem:
 	engine->instmem.takedown(dev);
diff --git a/drivers/gpu/drm/nouveau/nv40_fifo.c b/drivers/gpu/drm/nouveau/nv40_fifo.c
index 6b2ef4a..500ccfd 100644
--- a/drivers/gpu/drm/nouveau/nv40_fifo.c
+++ b/drivers/gpu/drm/nouveau/nv40_fifo.c
@@ -278,7 +278,7 @@ nv40_fifo_init_ramxx(struct drm_device *dev)
 	default:
 		nv_wr32(dev, 0x2230, 0);
 		nv_wr32(dev, NV40_PFIFO_RAMFC,
-			((nouveau_mem_fb_amount(dev) - 512 * 1024 +
+			((dev_priv->vram_size - 512 * 1024 +
 			  dev_priv->ramfc_offset) >> 16) | (3 << 16));
 		break;
 	}
diff --git a/drivers/gpu/drm/nouveau/nv40_graph.c b/drivers/gpu/drm/nouveau/nv40_graph.c
index 53e8afe..0616c96 100644
--- a/drivers/gpu/drm/nouveau/nv40_graph.c
+++ b/drivers/gpu/drm/nouveau/nv40_graph.c
@@ -335,6 +335,27 @@ nv40_graph_init(struct drm_device *dev)
 	nv_wr32(dev, 0x400b38, 0x2ffff800);
 	nv_wr32(dev, 0x400b3c, 0x00006000);
 
+	/* Tiling related stuff. */
+	switch (dev_priv->chipset) {
+	case 0x44:
+	case 0x4a:
+		nv_wr32(dev, 0x400bc4, 0x1003d888);
+		nv_wr32(dev, 0x400bbc, 0xb7a7b500);
+		break;
+	case 0x46:
+		nv_wr32(dev, 0x400bc4, 0x0000e024);
+		nv_wr32(dev, 0x400bbc, 0xb7a7b520);
+		break;
+	case 0x4c:
+	case 0x4e:
+	case 0x67:
+		nv_wr32(dev, 0x400bc4, 0x1003d888);
+		nv_wr32(dev, 0x400bbc, 0xb7a7b540);
+		break;
+	default:
+		break;
+	}
+
 	/* Turn all the tiling regions off. */
 	for (i = 0; i < pfb->num_tiles; i++)
 		nv40_graph_set_region_tiling(dev, i, 0, 0, 0);
diff --git a/drivers/gpu/drm/nouveau/nv50_display.c b/drivers/gpu/drm/nouveau/nv50_display.c
index fac6c88..649db4c 100644
--- a/drivers/gpu/drm/nouveau/nv50_display.c
+++ b/drivers/gpu/drm/nouveau/nv50_display.c
@@ -143,7 +143,7 @@ nv50_evo_channel_new(struct drm_device *dev, struct nouveau_channel **pchan)
 	}
 
 	ret = nv50_evo_dmaobj_new(chan, 0x3d, NvEvoVRAM, 0, 0x19,
-				  0, nouveau_mem_fb_amount(dev));
+				  0, dev_priv->vram_size);
 	if (ret) {
 		nv50_evo_channel_del(pchan);
 		return ret;
@@ -231,7 +231,7 @@ nv50_display_init(struct drm_device *dev)
 	/* This used to be in crtc unblank, but seems out of place there. */
 	nv_wr32(dev, NV50_PDISPLAY_UNK_380, 0);
 	/* RAM is clamped to 256 MiB. */
-	ram_amount = nouveau_mem_fb_amount(dev);
+	ram_amount = dev_priv->vram_size;
 	NV_DEBUG_KMS(dev, "ram_amount %d\n", ram_amount);
 	if (ram_amount > 256*1024*1024)
 		ram_amount = 256*1024*1024;
@@ -529,8 +529,10 @@ int nv50_display_create(struct drm_device *dev)
 	}
 
 	ret = nv50_display_init(dev);
-	if (ret)
+	if (ret) {
+		nv50_display_destroy(dev);
 		return ret;
+	}
 
 	return 0;
 }
@@ -885,10 +887,12 @@ nv50_display_error_handler(struct drm_device *dev)
 	nv_wr32(dev, NV50_PDISPLAY_TRAPPED_ADDR, 0x90000000);
 }
 
-static void
-nv50_display_irq_hotplug(struct drm_device *dev)
+void
+nv50_display_irq_hotplug_bh(struct work_struct *work)
 {
-	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_nouveau_private *dev_priv =
+		container_of(work, struct drm_nouveau_private, hpd_work);
+	struct drm_device *dev = dev_priv->dev;
 	struct drm_connector *connector;
 	const uint32_t gpio_reg[4] = { 0xe104, 0xe108, 0xe280, 0xe284 };
 	uint32_t unplug_mask, plug_mask, change_mask;
@@ -949,8 +953,10 @@ nv50_display_irq_handler(struct drm_device *dev)
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	uint32_t delayed = 0;
 
-	while (nv_rd32(dev, NV50_PMC_INTR_0) & NV50_PMC_INTR_0_HOTPLUG)
-		nv50_display_irq_hotplug(dev);
+	if (nv_rd32(dev, NV50_PMC_INTR_0) & NV50_PMC_INTR_0_HOTPLUG) {
+		if (!work_pending(&dev_priv->hpd_work))
+			queue_work(dev_priv->wq, &dev_priv->hpd_work);
+	}
 
 	while (nv_rd32(dev, NV50_PMC_INTR_0) & NV50_PMC_INTR_0_DISPLAY) {
 		uint32_t intr0 = nv_rd32(dev, NV50_PDISPLAY_INTR_0);
diff --git a/drivers/gpu/drm/nouveau/nv50_display.h b/drivers/gpu/drm/nouveau/nv50_display.h
index 3ae8d07..581d405 100644
--- a/drivers/gpu/drm/nouveau/nv50_display.h
+++ b/drivers/gpu/drm/nouveau/nv50_display.h
@@ -37,6 +37,7 @@
 
 void nv50_display_irq_handler(struct drm_device *dev);
 void nv50_display_irq_handler_bh(struct work_struct *work);
+void nv50_display_irq_hotplug_bh(struct work_struct *work);
 int nv50_display_init(struct drm_device *dev);
 int nv50_display_create(struct drm_device *dev);
 int nv50_display_destroy(struct drm_device *dev);
diff --git a/drivers/gpu/drm/nouveau/nv50_fbcon.c b/drivers/gpu/drm/nouveau/nv50_fbcon.c
index 25a3cd8..a8c70e7 100644
--- a/drivers/gpu/drm/nouveau/nv50_fbcon.c
+++ b/drivers/gpu/drm/nouveau/nv50_fbcon.c
@@ -157,8 +157,11 @@ nv50_fbcon_accel_init(struct fb_info *info)
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_channel *chan = dev_priv->channel;
 	struct nouveau_gpuobj *eng2d = NULL;
+	uint64_t fb;
 	int ret, format;
 
+	fb = info->fix.smem_start - dev_priv->fb_phys + dev_priv->vm_vram_base;
+
 	switch (info->var.bits_per_pixel) {
 	case 8:
 		format = 0xf3;
@@ -248,9 +251,8 @@ nv50_fbcon_accel_init(struct fb_info *info)
 	OUT_RING(chan, info->fix.line_length);
 	OUT_RING(chan, info->var.xres_virtual);
 	OUT_RING(chan, info->var.yres_virtual);
-	OUT_RING(chan, 0);
-	OUT_RING(chan, info->fix.smem_start - dev_priv->fb_phys +
-			 dev_priv->vm_vram_base);
+	OUT_RING(chan, upper_32_bits(fb));
+	OUT_RING(chan, lower_32_bits(fb));
 	BEGIN_RING(chan, NvSub2D, 0x0230, 2);
 	OUT_RING(chan, format);
 	OUT_RING(chan, 1);
@@ -258,9 +260,8 @@ nv50_fbcon_accel_init(struct fb_info *info)
 	OUT_RING(chan, info->fix.line_length);
 	OUT_RING(chan, info->var.xres_virtual);
 	OUT_RING(chan, info->var.yres_virtual);
-	OUT_RING(chan, 0);
-	OUT_RING(chan, info->fix.smem_start - dev_priv->fb_phys +
-			 dev_priv->vm_vram_base);
+	OUT_RING(chan, upper_32_bits(fb));
+	OUT_RING(chan, lower_32_bits(fb));
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/nouveau/nv50_gpio.c b/drivers/gpu/drm/nouveau/nv50_gpio.c
new file mode 100644
index 0000000..c61782b
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_gpio.c
@@ -0,0 +1,76 @@
+/*
+ * 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 "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_hw.h"
+
+static int
+nv50_gpio_location(struct dcb_gpio_entry *gpio, uint32_t *reg, uint32_t *shift)
+{
+	const uint32_t nv50_gpio_reg[4] = { 0xe104, 0xe108, 0xe280, 0xe284 };
+
+	if (gpio->line > 32)
+		return -EINVAL;
+
+	*reg = nv50_gpio_reg[gpio->line >> 3];
+	*shift = (gpio->line & 7) << 2;
+	return 0;
+}
+
+int
+nv50_gpio_get(struct drm_device *dev, enum dcb_gpio_tag tag)
+{
+	struct dcb_gpio_entry *gpio;
+	uint32_t r, s, v;
+
+	gpio = nouveau_bios_gpio_entry(dev, tag);
+	if (!gpio)
+		return -ENOENT;
+
+	if (nv50_gpio_location(gpio, &r, &s))
+		return -EINVAL;
+
+	v = nv_rd32(dev, r) >> (s + 2);
+	return ((v & 1) == (gpio->state[1] & 1));
+}
+
+int
+nv50_gpio_set(struct drm_device *dev, enum dcb_gpio_tag tag, int state)
+{
+	struct dcb_gpio_entry *gpio;
+	uint32_t r, s, v;
+
+	gpio = nouveau_bios_gpio_entry(dev, tag);
+	if (!gpio)
+		return -ENOENT;
+
+	if (nv50_gpio_location(gpio, &r, &s))
+		return -EINVAL;
+
+	v  = nv_rd32(dev, r) & ~(0x3 << s);
+	v |= (gpio->state[state] ^ 2) << s;
+	nv_wr32(dev, r, v);
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nv50_graph.c b/drivers/gpu/drm/nouveau/nv50_graph.c
index c62b33a..b203d06 100644
--- a/drivers/gpu/drm/nouveau/nv50_graph.c
+++ b/drivers/gpu/drm/nouveau/nv50_graph.c
@@ -410,9 +410,10 @@ struct nouveau_pgraph_object_class nv50_graph_grclass[] = {
 	{ 0x5039, false, NULL }, /* m2mf */
 	{ 0x502d, false, NULL }, /* 2d */
 	{ 0x50c0, false, NULL }, /* compute */
+	{ 0x85c0, false, NULL }, /* compute (nva3, nva5, nva8) */
 	{ 0x5097, false, NULL }, /* tesla (nv50) */
-	{ 0x8297, false, NULL }, /* tesla (nv80/nv90) */
-	{ 0x8397, false, NULL }, /* tesla (nva0) */
-	{ 0x8597, false, NULL }, /* tesla (nva8) */
+	{ 0x8297, false, NULL }, /* tesla (nv8x/nv9x) */
+	{ 0x8397, false, NULL }, /* tesla (nva0, nvaa, nvac) */
+	{ 0x8597, false, NULL }, /* tesla (nva3, nva5, nva8) */
 	{}
 };
diff --git a/drivers/gpu/drm/nouveau/nv50_grctx.c b/drivers/gpu/drm/nouveau/nv50_grctx.c
index 546b319..42a8fb2 100644
--- a/drivers/gpu/drm/nouveau/nv50_grctx.c
+++ b/drivers/gpu/drm/nouveau/nv50_grctx.c
@@ -55,12 +55,12 @@
 #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_NEWCTX                ((2 * 32) + 12)
-#define CP_FLAG_NEWCTX_BUSY           0
-#define CP_FLAG_NEWCTX_DONE           1
 #define CP_FLAG_ALWAYS                ((2 * 32) + 13)
 #define CP_FLAG_ALWAYS_FALSE          0
 #define CP_FLAG_ALWAYS_TRUE           1
@@ -177,6 +177,7 @@ nv50_grctx_init(struct nouveau_grctx *ctx)
 	case 0x96:
 	case 0x98:
 	case 0xa0:
+	case 0xa3:
 	case 0xa5:
 	case 0xa8:
 	case 0xaa:
@@ -364,6 +365,7 @@ nv50_graph_construct_mmio(struct nouveau_grctx *ctx)
 	case 0xac:
 		gr_def(ctx, 0x401c00, 0x042500df);
 		break;
+	case 0xa3:
 	case 0xa5:
 	case 0xa8:
 		gr_def(ctx, 0x401c00, 0x142500df);
@@ -418,6 +420,7 @@ nv50_graph_construct_mmio(struct nouveau_grctx *ctx)
 		break;
 	case 0x84:
 	case 0xa0:
+	case 0xa3:
 	case 0xa5:
 	case 0xa8:
 	case 0xaa:
@@ -792,6 +795,7 @@ nv50_graph_construct_mmio(struct nouveau_grctx *ctx)
 				case 0xa5:
 					gr_def(ctx, offset + 0x1c, 0x310c0000);
 					break;
+				case 0xa3:
 				case 0xa8:
 				case 0xaa:
 				case 0xac:
@@ -859,6 +863,8 @@ nv50_graph_construct_mmio(struct nouveau_grctx *ctx)
 			else
 				gr_def(ctx, offset + 0x8, 0x05010202);
 			gr_def(ctx, offset + 0xc, 0x00030201);
+			if (dev_priv->chipset == 0xa3)
+				cp_ctx(ctx, base + 0x36c, 1);
 
 			cp_ctx(ctx, base + 0x400, 2);
 			gr_def(ctx, base + 0x404, 0x00000040);
@@ -1159,7 +1165,9 @@ nv50_graph_construct_xfer1(struct nouveau_grctx *ctx)
 		nv50_graph_construct_gene_unk8(ctx);
 		if (dev_priv->chipset == 0xa0)
 			xf_emit(ctx, 0x189, 0);
-		else if (dev_priv->chipset < 0xa8)
+		else if (dev_priv->chipset == 0xa3)
+			xf_emit(ctx, 0xd5, 0);
+		else if (dev_priv->chipset == 0xa5)
 			xf_emit(ctx, 0x99, 0);
 		else if (dev_priv->chipset == 0xaa)
 			xf_emit(ctx, 0x65, 0);
@@ -1197,6 +1205,8 @@ nv50_graph_construct_xfer1(struct nouveau_grctx *ctx)
 		ctx->ctxvals_pos = offset + 4;
 		if (dev_priv->chipset == 0xa0)
 			xf_emit(ctx, 0xa80, 0);
+		else if (dev_priv->chipset == 0xa3)
+			xf_emit(ctx, 0xa7c, 0);
 		else
 			xf_emit(ctx, 0xa7a, 0);
 		xf_emit(ctx, 1, 0x3fffff);
@@ -1341,6 +1351,7 @@ nv50_graph_construct_gene_unk1(struct nouveau_grctx *ctx)
 		xf_emit(ctx, 0x942, 0);
 		break;
 	case 0xa0:
+	case 0xa3:
 		xf_emit(ctx, 0x2042, 0);
 		break;
 	case 0xa5:
diff --git a/drivers/gpu/drm/nouveau/nv50_instmem.c b/drivers/gpu/drm/nouveau/nv50_instmem.c
index de1f5b0..5f21df3 100644
--- a/drivers/gpu/drm/nouveau/nv50_instmem.c
+++ b/drivers/gpu/drm/nouveau/nv50_instmem.c
@@ -63,9 +63,10 @@ nv50_instmem_init(struct drm_device *dev)
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_channel *chan;
 	uint32_t c_offset, c_size, c_ramfc, c_vmpd, c_base, pt_size;
+	uint32_t save_nv001700;
+	uint64_t v;
 	struct nv50_instmem_priv *priv;
 	int ret, i;
-	uint32_t v, save_nv001700;
 
 	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
 	if (!priv)
@@ -76,17 +77,12 @@ nv50_instmem_init(struct drm_device *dev)
 	for (i = 0x1700; i <= 0x1710; i += 4)
 		priv->save1700[(i-0x1700)/4] = nv_rd32(dev, i);
 
-	if (dev_priv->chipset == 0xaa || dev_priv->chipset == 0xac)
-		dev_priv->vram_sys_base = nv_rd32(dev, 0x100e10) << 12;
-	else
-		dev_priv->vram_sys_base = 0;
-
 	/* Reserve the last MiB of VRAM, we should probably try to avoid
 	 * setting up the below tables over the top of the VBIOS image at
 	 * some point.
 	 */
 	dev_priv->ramin_rsvd_vram = 1 << 20;
-	c_offset = nouveau_mem_fb_amount(dev) - dev_priv->ramin_rsvd_vram;
+	c_offset = dev_priv->vram_size - dev_priv->ramin_rsvd_vram;
 	c_size   = 128 << 10;
 	c_vmpd   = ((dev_priv->chipset & 0xf0) == 0x50) ? 0x1400 : 0x200;
 	c_ramfc  = ((dev_priv->chipset & 0xf0) == 0x50) ? 0x0 : 0x20;
@@ -106,7 +102,7 @@ nv50_instmem_init(struct drm_device *dev)
 	dev_priv->vm_gart_size = NV50_VM_BLOCK;
 
 	dev_priv->vm_vram_base = dev_priv->vm_gart_base + dev_priv->vm_gart_size;
-	dev_priv->vm_vram_size = nouveau_mem_fb_amount(dev);
+	dev_priv->vm_vram_size = dev_priv->vram_size;
 	if (dev_priv->vm_vram_size > NV50_VM_MAX_VRAM)
 		dev_priv->vm_vram_size = NV50_VM_MAX_VRAM;
 	dev_priv->vm_vram_size = roundup(dev_priv->vm_vram_size, NV50_VM_BLOCK);
@@ -189,8 +185,8 @@ nv50_instmem_init(struct drm_device *dev)
 
 	i = 0;
 	while (v < dev_priv->vram_sys_base + c_offset + c_size) {
-		BAR0_WI32(priv->pramin_pt->gpuobj, i + 0, v);
-		BAR0_WI32(priv->pramin_pt->gpuobj, i + 4, 0x00000000);
+		BAR0_WI32(priv->pramin_pt->gpuobj, i + 0, lower_32_bits(v));
+		BAR0_WI32(priv->pramin_pt->gpuobj, i + 4, upper_32_bits(v));
 		v += 0x1000;
 		i += 8;
 	}
diff --git a/drivers/gpu/drm/nouveau/nv50_sor.c b/drivers/gpu/drm/nouveau/nv50_sor.c
index c2fff54..0c68698 100644
--- a/drivers/gpu/drm/nouveau/nv50_sor.c
+++ b/drivers/gpu/drm/nouveau/nv50_sor.c
@@ -211,7 +211,7 @@ nv50_sor_mode_set(struct drm_encoder *encoder, struct drm_display_mode *mode,
 			mode_ctl = 0x0200;
 		break;
 	case OUTPUT_DP:
-		mode_ctl |= 0x00050000;
+		mode_ctl |= (nv_encoder->dp.mc_unknown << 16);
 		if (nv_encoder->dcb->sorconf.link & 1)
 			mode_ctl |= 0x00000800;
 		else
@@ -274,6 +274,7 @@ static const struct drm_encoder_funcs nv50_sor_encoder_funcs = {
 int
 nv50_sor_create(struct drm_device *dev, struct dcb_entry *entry)
 {
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_encoder *nv_encoder = NULL;
 	struct drm_encoder *encoder;
 	bool dum;
@@ -319,5 +320,27 @@ nv50_sor_create(struct drm_device *dev, struct dcb_entry *entry)
 	encoder->possible_crtcs = entry->heads;
 	encoder->possible_clones = 0;
 
+	if (nv_encoder->dcb->type == OUTPUT_DP) {
+		uint32_t mc, or = nv_encoder->or;
+
+		if (dev_priv->chipset < 0x90 ||
+		    dev_priv->chipset == 0x92 || dev_priv->chipset == 0xa0)
+			mc = nv_rd32(dev, NV50_PDISPLAY_SOR_MODE_CTRL_C(or));
+		else
+			mc = nv_rd32(dev, NV90_PDISPLAY_SOR_MODE_CTRL_C(or));
+
+		switch ((mc & 0x00000f00) >> 8) {
+		case 8:
+		case 9:
+			nv_encoder->dp.mc_unknown = (mc & 0x000f0000) >> 16;
+			break;
+		default:
+			break;
+		}
+
+		if (!nv_encoder->dp.mc_unknown)
+			nv_encoder->dp.mc_unknown = 5;
+	}
+
 	return 0;
 }