drm/i915: framebuffer compression for pre-GM45

This patch adds framebuffer compression (good for about ~0.5W power
savings in the best case) support for pre-GM45 chips.  GM45+ have a new,
more flexible FBC scheme that will be added in a separate patch.

FBC can't always be enabled: the compressed buffer must be physically
contiguous and reside in stolen space.  So if you have a large display
and a small amount of stolen memory, you may not be able to take
advantage of FBC.  In some cases, a BIOS setting controls how much
stolen space is available.  Increasing this to 8 or 16M can help.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Eric Anholt <eric@anholt.net>

5 files changed, 382 insertions, 19 deletions

diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
index 9909505..5a6b731 100644
--- a/drivers/gpu/drm/i915/i915_dma.c
+++ b/drivers/gpu/drm/i915/i915_dma.c
@@ -921,7 +921,8 @@ static int i915_get_bridge_dev(struct drm_device *dev)
  * how much was set aside so we can use it for our own purposes.
  */
 static int i915_probe_agp(struct drm_device *dev, uint32_t *aperture_size,
-			  uint32_t *preallocated_size)
+			  uint32_t *preallocated_size,
+			  uint32_t *start)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u16 tmp = 0;
@@ -1008,11 +1009,148 @@ static int i915_probe_agp(struct drm_device *dev, uint32_t *aperture_size,
 		return -1;
 	}
 	*preallocated_size = stolen - overhead;
+	*start = overhead;
 
 	return 0;
 }
 
+#define PTE_ADDRESS_MASK		0xfffff000
+#define PTE_ADDRESS_MASK_HIGH		0x000000f0 /* i915+ */
+#define PTE_MAPPING_TYPE_UNCACHED	(0 << 1)
+#define PTE_MAPPING_TYPE_DCACHE		(1 << 1) /* i830 only */
+#define PTE_MAPPING_TYPE_CACHED		(3 << 1)
+#define PTE_MAPPING_TYPE_MASK		(3 << 1)
+#define PTE_VALID			(1 << 0)
+
+/**
+ * i915_gtt_to_phys - take a GTT address and turn it into a physical one
+ * @dev: drm device
+ * @gtt_addr: address to translate
+ *
+ * Some chip functions require allocations from stolen space but need the
+ * physical address of the memory in question.  We use this routine
+ * to get a physical address suitable for register programming from a given
+ * GTT address.
+ */
+static unsigned long i915_gtt_to_phys(struct drm_device *dev,
+				      unsigned long gtt_addr)
+{
+	unsigned long *gtt;
+	unsigned long entry, phys;
+	int gtt_bar = IS_I9XX(dev) ? 0 : 1;
+	int gtt_offset, gtt_size;
+
+	if (IS_I965G(dev)) {
+		if (IS_G4X(dev) || IS_IGDNG(dev)) {
+			gtt_offset = 2*1024*1024;
+			gtt_size = 2*1024*1024;
+		} else {
+			gtt_offset = 512*1024;
+			gtt_size = 512*1024;
+		}
+	} else {
+		gtt_bar = 3;
+		gtt_offset = 0;
+		gtt_size = pci_resource_len(dev->pdev, gtt_bar);
+	}
+
+	gtt = ioremap_wc(pci_resource_start(dev->pdev, gtt_bar) + gtt_offset,
+			 gtt_size);
+	if (!gtt) {
+		DRM_ERROR("ioremap of GTT failed\n");
+		return 0;
+	}
+
+	entry = *(volatile u32 *)(gtt + (gtt_addr / 1024));
+
+	DRM_DEBUG("GTT addr: 0x%08lx, PTE: 0x%08lx\n", gtt_addr, entry);
+
+	/* Mask out these reserved bits on this hardware. */
+	if (!IS_I9XX(dev) || IS_I915G(dev) || IS_I915GM(dev) ||
+	    IS_I945G(dev) || IS_I945GM(dev)) {
+		entry &= ~PTE_ADDRESS_MASK_HIGH;
+	}
+
+	/* If it's not a mapping type we know, then bail. */
+	if ((entry & PTE_MAPPING_TYPE_MASK) != PTE_MAPPING_TYPE_UNCACHED &&
+	    (entry & PTE_MAPPING_TYPE_MASK) != PTE_MAPPING_TYPE_CACHED)	{
+		iounmap(gtt);
+		return 0;
+	}
+
+	if (!(entry & PTE_VALID)) {
+		DRM_ERROR("bad GTT entry in stolen space\n");
+		iounmap(gtt);
+		return 0;
+	}
+
+	iounmap(gtt);
+
+	phys =(entry & PTE_ADDRESS_MASK) |
+		((uint64_t)(entry & PTE_ADDRESS_MASK_HIGH) << (32 - 4));
+
+	DRM_DEBUG("GTT addr: 0x%08lx, phys addr: 0x%08lx\n", gtt_addr, phys);
+
+	return phys;
+}
+
+static void i915_warn_stolen(struct drm_device *dev)
+{
+	DRM_ERROR("not enough stolen space for compressed buffer, disabling\n");
+	DRM_ERROR("hint: you may be able to increase stolen memory size in the BIOS to avoid this\n");
+}
+
+static void i915_setup_compression(struct drm_device *dev, int size)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_mm_node *compressed_fb, *compressed_llb;
+	unsigned long cfb_base, ll_base;
+
+	/* Leave 1M for line length buffer & misc. */
+	compressed_fb = drm_mm_search_free(&dev_priv->vram, size, 4096, 0);
+	if (!compressed_fb) {
+		i915_warn_stolen(dev);
+		return;
+	}
+
+	compressed_fb = drm_mm_get_block(compressed_fb, size, 4096);
+	if (!compressed_fb) {
+		i915_warn_stolen(dev);
+		return;
+	}
+
+	compressed_llb = drm_mm_search_free(&dev_priv->vram, 4096, 4096, 0);
+	if (!compressed_llb) {
+		i915_warn_stolen(dev);
+		return;
+	}
+
+	compressed_llb = drm_mm_get_block(compressed_fb, 4096, 4096);
+	if (!compressed_llb) {
+		i915_warn_stolen(dev);
+		return;
+	}
+
+	dev_priv->cfb_size = size;
+
+	cfb_base = i915_gtt_to_phys(dev, compressed_fb->start);
+	ll_base = i915_gtt_to_phys(dev, compressed_llb->start);
+	if (!cfb_base || !ll_base) {
+		DRM_ERROR("failed to get stolen phys addr, disabling FBC\n");
+		drm_mm_put_block(compressed_fb);
+		drm_mm_put_block(compressed_llb);
+	}
+
+	i8xx_disable_fbc(dev);
+
+	DRM_DEBUG("FBC base 0x%08lx, ll base 0x%08lx, size %dM\n", cfb_base,
+		  ll_base, size >> 20);
+	I915_WRITE(FBC_CFB_BASE, cfb_base);
+	I915_WRITE(FBC_LL_BASE, ll_base);
+}
+
 static int i915_load_modeset_init(struct drm_device *dev,
+				  unsigned long prealloc_start,
 				  unsigned long prealloc_size,
 				  unsigned long agp_size)
 {
@@ -1033,6 +1171,7 @@ static int i915_load_modeset_init(struct drm_device *dev,
 
 	/* Basic memrange allocator for stolen space (aka vram) */
 	drm_mm_init(&dev_priv->vram, 0, prealloc_size);
+	DRM_INFO("set up %ldM of stolen space\n", prealloc_size / (1024*1024));
 
 	/* Let GEM Manage from end of prealloc space to end of aperture.
 	 *
@@ -1049,6 +1188,19 @@ static int i915_load_modeset_init(struct drm_device *dev,
 	if (ret)
 		goto out;
 
+	/* Try to set up FBC with a reasonable compressed buffer size */
+	if (IS_MOBILE(dev) && (IS_I9XX(dev) || IS_I965G(dev)) &&
+	    i915_powersave) {
+		int cfb_size;
+
+		/* Try to get an 8M buffer... */
+		if (prealloc_size > (9*1024*1024))
+			cfb_size = 8*1024*1024;
+		else /* fall back to 7/8 of the stolen space */
+			cfb_size = prealloc_size * 7 / 8;
+		i915_setup_compression(dev, cfb_size);
+	}
+
 	/* Allow hardware batchbuffers unless told otherwise.
 	 */
 	dev_priv->allow_batchbuffer = 1;
@@ -1161,7 +1313,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	resource_size_t base, size;
 	int ret = 0, mmio_bar = IS_I9XX(dev) ? 0 : 1;
-	uint32_t agp_size, prealloc_size;
+	uint32_t agp_size, prealloc_size, prealloc_start;
 
 	/* i915 has 4 more counters */
 	dev->counters += 4;
@@ -1215,7 +1367,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 			 "performance may suffer.\n");
 	}
 
-	ret = i915_probe_agp(dev, &agp_size, &prealloc_size);
+	ret = i915_probe_agp(dev, &agp_size, &prealloc_size, &prealloc_start);
 	if (ret)
 		goto out_iomapfree;
 
@@ -1282,7 +1434,8 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 	}
 
 	if (drm_core_check_feature(dev, DRIVER_MODESET)) {
-		ret = i915_load_modeset_init(dev, prealloc_size, agp_size);
+		ret = i915_load_modeset_init(dev, prealloc_start,
+					     prealloc_size, agp_size);
 		if (ret < 0) {
 			DRM_ERROR("failed to init modeset\n");
 			goto out_workqueue_free;
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index f2b8d27..0344afd 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -48,6 +48,11 @@ enum pipe {
 	PIPE_B,
 };
 
+enum plane {
+	PLANE_A = 0,
+	PLANE_B,
+};
+
 #define I915_NUM_PIPE	2
 
 /* Interface history:
@@ -202,6 +207,11 @@ typedef struct drm_i915_private {
 
 	struct drm_mm vram;
 
+	unsigned long cfb_size;
+	unsigned long cfb_pitch;
+	int cfb_fence;
+	int cfb_plane;
+
 	int irq_enabled;
 
 	struct intel_opregion opregion;
@@ -768,6 +778,7 @@ static inline void opregion_enable_asle(struct drm_device *dev) { return; }
 /* modesetting */
 extern void intel_modeset_init(struct drm_device *dev);
 extern void intel_modeset_cleanup(struct drm_device *dev);
+extern void i8xx_disable_fbc(struct drm_device *dev);
 
 /**
  * Lock test for when it's just for synchronization of ring access.
@@ -918,6 +929,7 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller);
 
 #define HAS_FW_BLC(dev) (IS_I9XX(dev) || IS_G4X(dev) || IS_IGDNG(dev))
 #define HAS_PIPE_CXSR(dev) (IS_G4X(dev) || IS_IGDNG(dev))
+#define I915_HAS_FBC(dev) (IS_I9XX(dev) || IS_I965G(dev))
 
 #define PRIMARY_RINGBUFFER_SIZE         (128*1024)
 
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index e38cd21..f7c2de8 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -343,6 +343,7 @@
 #define   FBC_CTL_PLANEA	(0<<0)
 #define   FBC_CTL_PLANEB	(1<<0)
 #define FBC_FENCE_OFF		0x0321b
+#define FBC_TAG			0x03300
 
 #define FBC_LL_SIZE		(1536)
 
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 92e0aae..cadb9ef 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -954,6 +954,174 @@ intel_wait_for_vblank(struct drm_device *dev)
 	mdelay(20);
 }
 
+/* Parameters have changed, update FBC info */
+static void i8xx_enable_fbc(struct drm_crtc *crtc, unsigned long interval)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_framebuffer *fb = crtc->fb;
+	struct intel_framebuffer *intel_fb = to_intel_framebuffer(fb);
+	struct drm_i915_gem_object *obj_priv = intel_fb->obj->driver_private;
+	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
+	int plane, i;
+	u32 fbc_ctl, fbc_ctl2;
+
+	dev_priv->cfb_pitch = dev_priv->cfb_size / FBC_LL_SIZE;
+
+	if (fb->pitch < dev_priv->cfb_pitch)
+		dev_priv->cfb_pitch = fb->pitch;
+
+	/* FBC_CTL wants 64B units */
+	dev_priv->cfb_pitch = (dev_priv->cfb_pitch / 64) - 1;
+	dev_priv->cfb_fence = obj_priv->fence_reg;
+	dev_priv->cfb_plane = intel_crtc->plane;
+	plane = dev_priv->cfb_plane == 0 ? FBC_CTL_PLANEA : FBC_CTL_PLANEB;
+
+	/* Clear old tags */
+	for (i = 0; i < (FBC_LL_SIZE / 32) + 1; i++)
+		I915_WRITE(FBC_TAG + (i * 4), 0);
+
+	/* Set it up... */
+	fbc_ctl2 = FBC_CTL_FENCE_DBL | FBC_CTL_IDLE_IMM | plane;
+	if (obj_priv->tiling_mode != I915_TILING_NONE)
+		fbc_ctl2 |= FBC_CTL_CPU_FENCE;
+	I915_WRITE(FBC_CONTROL2, fbc_ctl2);
+	I915_WRITE(FBC_FENCE_OFF, crtc->y);
+
+	/* enable it... */
+	fbc_ctl = FBC_CTL_EN | FBC_CTL_PERIODIC;
+	fbc_ctl |= (dev_priv->cfb_pitch & 0xff) << FBC_CTL_STRIDE_SHIFT;
+	fbc_ctl |= (interval & 0x2fff) << FBC_CTL_INTERVAL_SHIFT;
+	if (obj_priv->tiling_mode != I915_TILING_NONE)
+		fbc_ctl |= dev_priv->cfb_fence;
+	I915_WRITE(FBC_CONTROL, fbc_ctl);
+
+	DRM_DEBUG("enabled FBC, pitch %ld, yoff %d, plane %d, ",
+		  dev_priv->cfb_pitch, crtc->y, dev_priv->cfb_plane);
+}
+
+void i8xx_disable_fbc(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	u32 fbc_ctl;
+
+	/* Disable compression */
+	fbc_ctl = I915_READ(FBC_CONTROL);
+	fbc_ctl &= ~FBC_CTL_EN;
+	I915_WRITE(FBC_CONTROL, fbc_ctl);
+
+	/* Wait for compressing bit to clear */
+	while (I915_READ(FBC_STATUS) & FBC_STAT_COMPRESSING)
+		; /* nothing */
+
+	intel_wait_for_vblank(dev);
+
+	DRM_DEBUG("disabled FBC\n");
+}
+
+static bool i8xx_fbc_enabled(struct drm_crtc *crtc)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	return I915_READ(FBC_CONTROL) & FBC_CTL_EN;
+}
+
+/**
+ * intel_update_fbc - enable/disable FBC as needed
+ * @crtc: CRTC to point the compressor at
+ * @mode: mode in use
+ *
+ * Set up the framebuffer compression hardware at mode set time.  We
+ * enable it if possible:
+ *   - plane A only (on pre-965)
+ *   - no pixel mulitply/line duplication
+ *   - no alpha buffer discard
+ *   - no dual wide
+ *   - framebuffer <= 2048 in width, 1536 in height
+ *
+ * We can't assume that any compression will take place (worst case),
+ * so the compressed buffer has to be the same size as the uncompressed
+ * one.  It also must reside (along with the line length buffer) in
+ * stolen memory.
+ *
+ * We need to enable/disable FBC on a global basis.
+ */
+static void intel_update_fbc(struct drm_crtc *crtc,
+			     struct drm_display_mode *mode)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_framebuffer *fb = crtc->fb;
+	struct intel_framebuffer *intel_fb;
+	struct drm_i915_gem_object *obj_priv;
+	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
+	int plane = intel_crtc->plane;
+
+	if (!i915_powersave)
+		return;
+
+	if (!crtc->fb)
+		return;
+
+	intel_fb = to_intel_framebuffer(fb);
+	obj_priv = intel_fb->obj->driver_private;
+
+	/*
+	 * If FBC is already on, we just have to verify that we can
+	 * keep it that way...
+	 * Need to disable if:
+	 *   - changing FBC params (stride, fence, mode)
+	 *   - new fb is too large to fit in compressed buffer
+	 *   - going to an unsupported config (interlace, pixel multiply, etc.)
+	 */
+	if (intel_fb->obj->size > dev_priv->cfb_size) {
+		DRM_DEBUG("framebuffer too large, disabling compression\n");
+		goto out_disable;
+	}
+	if ((mode->flags & DRM_MODE_FLAG_INTERLACE) ||
+	    (mode->flags & DRM_MODE_FLAG_DBLSCAN)) {
+		DRM_DEBUG("mode incompatible with compression, disabling\n");
+		goto out_disable;
+	}
+	if ((mode->hdisplay > 2048) ||
+	    (mode->vdisplay > 1536)) {
+		DRM_DEBUG("mode too large for compression, disabling\n");
+		goto out_disable;
+	}
+	if (IS_I9XX(dev) && plane != 0) {
+		DRM_DEBUG("plane not 0, disabling compression\n");
+		goto out_disable;
+	}
+	if (obj_priv->tiling_mode != I915_TILING_X) {
+		DRM_DEBUG("framebuffer not tiled, disabling compression\n");
+		goto out_disable;
+	}
+
+	if (i8xx_fbc_enabled(crtc)) {
+		/* We can re-enable it in this case, but need to update pitch */
+		if (fb->pitch > dev_priv->cfb_pitch)
+			i8xx_disable_fbc(dev);
+		if (obj_priv->fence_reg != dev_priv->cfb_fence)
+			i8xx_disable_fbc(dev);
+		if (plane != dev_priv->cfb_plane)
+			i8xx_disable_fbc(dev);
+	}
+
+	if (!i8xx_fbc_enabled(crtc)) {
+		/* Now try to turn it back on if possible */
+		i8xx_enable_fbc(crtc, 500);
+	}
+
+	return;
+
+out_disable:
+	DRM_DEBUG("unsupported config, disabling FBC\n");
+	/* Multiple disables should be harmless */
+	if (i8xx_fbc_enabled(crtc))
+		i8xx_disable_fbc(dev);
+}
+
 static int
 intel_pipe_set_base(struct drm_crtc *crtc, int x, int y,
 		    struct drm_framebuffer *old_fb)
@@ -966,12 +1134,13 @@ intel_pipe_set_base(struct drm_crtc *crtc, int x, int y,
 	struct drm_i915_gem_object *obj_priv;
 	struct drm_gem_object *obj;
 	int pipe = intel_crtc->pipe;
+	int plane = intel_crtc->plane;
 	unsigned long Start, Offset;
-	int dspbase = (pipe == 0 ? DSPAADDR : DSPBADDR);
-	int dspsurf = (pipe == 0 ? DSPASURF : DSPBSURF);
-	int dspstride = (pipe == 0) ? DSPASTRIDE : DSPBSTRIDE;
-	int dsptileoff = (pipe == 0 ? DSPATILEOFF : DSPBTILEOFF);
-	int dspcntr_reg = (pipe == 0) ? DSPACNTR : DSPBCNTR;
+	int dspbase = (plane == 0 ? DSPAADDR : DSPBADDR);
+	int dspsurf = (plane == 0 ? DSPASURF : DSPBSURF);
+	int dspstride = (plane == 0) ? DSPASTRIDE : DSPBSTRIDE;
+	int dsptileoff = (plane == 0 ? DSPATILEOFF : DSPBTILEOFF);
+	int dspcntr_reg = (plane == 0) ? DSPACNTR : DSPBCNTR;
 	u32 dspcntr, alignment;
 	int ret;
 
@@ -981,12 +1150,12 @@ intel_pipe_set_base(struct drm_crtc *crtc, int x, int y,
 		return 0;
 	}
 
-	switch (pipe) {
+	switch (plane) {
 	case 0:
 	case 1:
 		break;
 	default:
-		DRM_ERROR("Can't update pipe %d in SAREA\n", pipe);
+		DRM_ERROR("Can't update plane %d in SAREA\n", plane);
 		return -EINVAL;
 	}
 
@@ -1114,6 +1283,9 @@ intel_pipe_set_base(struct drm_crtc *crtc, int x, int y,
 		master_priv->sarea_priv->pipeA_y = y;
 	}
 
+	if (I915_HAS_FBC(dev) && (IS_I965G(dev) || plane == 0))
+		intel_update_fbc(crtc, &crtc->mode);
+
 	return 0;
 }
 
@@ -1534,9 +1706,10 @@ static void i9xx_crtc_dpms(struct drm_crtc *crtc, int mode)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
 	int pipe = intel_crtc->pipe;
+	int plane = intel_crtc->plane;
 	int dpll_reg = (pipe == 0) ? DPLL_A : DPLL_B;
-	int dspcntr_reg = (pipe == 0) ? DSPACNTR : DSPBCNTR;
-	int dspbase_reg = (pipe == 0) ? DSPAADDR : DSPBADDR;
+	int dspcntr_reg = (plane == 0) ? DSPACNTR : DSPBCNTR;
+	int dspbase_reg = (plane == 0) ? DSPAADDR : DSPBADDR;
 	int pipeconf_reg = (pipe == 0) ? PIPEACONF : PIPEBCONF;
 	u32 temp;
 
@@ -1579,6 +1752,9 @@ static void i9xx_crtc_dpms(struct drm_crtc *crtc, int mode)
 
 		intel_crtc_load_lut(crtc);
 
+		if (I915_HAS_FBC(dev) && (IS_I965G(dev) || plane == 0))
+			intel_update_fbc(crtc, &crtc->mode);
+
 		/* Give the overlay scaler a chance to enable if it's on this pipe */
 		//intel_crtc_dpms_video(crtc, true); TODO
 		intel_update_watermarks(dev);
@@ -1588,6 +1764,9 @@ static void i9xx_crtc_dpms(struct drm_crtc *crtc, int mode)
 		/* Give the overlay scaler a chance to disable if it's on this pipe */
 		//intel_crtc_dpms_video(crtc, FALSE); TODO
 
+		if (dev_priv->cfb_plane == plane)
+			i8xx_disable_fbc(dev);
+
 		/* Disable the VGA plane that we never use */
 		i915_disable_vga(dev);
 
@@ -2325,10 +2504,11 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
 	int pipe = intel_crtc->pipe;
+	int plane = intel_crtc->plane;
 	int fp_reg = (pipe == 0) ? FPA0 : FPB0;
 	int dpll_reg = (pipe == 0) ? DPLL_A : DPLL_B;
 	int dpll_md_reg = (intel_crtc->pipe == 0) ? DPLL_A_MD : DPLL_B_MD;
-	int dspcntr_reg = (pipe == 0) ? DSPACNTR : DSPBCNTR;
+	int dspcntr_reg = (plane == 0) ? DSPACNTR : DSPBCNTR;
 	int pipeconf_reg = (pipe == 0) ? PIPEACONF : PIPEBCONF;
 	int htot_reg = (pipe == 0) ? HTOTAL_A : HTOTAL_B;
 	int hblank_reg = (pipe == 0) ? HBLANK_A : HBLANK_B;
@@ -2336,8 +2516,8 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 	int vtot_reg = (pipe == 0) ? VTOTAL_A : VTOTAL_B;
 	int vblank_reg = (pipe == 0) ? VBLANK_A : VBLANK_B;
 	int vsync_reg = (pipe == 0) ? VSYNC_A : VSYNC_B;
-	int dspsize_reg = (pipe == 0) ? DSPASIZE : DSPBSIZE;
-	int dsppos_reg = (pipe == 0) ? DSPAPOS : DSPBPOS;
+	int dspsize_reg = (plane == 0) ? DSPASIZE : DSPBSIZE;
+	int dsppos_reg = (plane == 0) ? DSPAPOS : DSPBPOS;
 	int pipesrc_reg = (pipe == 0) ? PIPEASRC : PIPEBSRC;
 	int refclk, num_outputs = 0;
 	intel_clock_t clock, reduced_clock;
@@ -2570,7 +2750,7 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 	   enable color space conversion */
 	if (!IS_IGDNG(dev)) {
 		if (pipe == 0)
-			dspcntr |= DISPPLANE_SEL_PIPE_A;
+			dspcntr &= ~DISPPLANE_SEL_PIPE_MASK;
 		else
 			dspcntr |= DISPPLANE_SEL_PIPE_B;
 	}
@@ -2739,6 +2919,8 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 	/* Flush the plane changes */
 	ret = intel_pipe_set_base(crtc, x, y, old_fb);
 
+	if (I915_HAS_FBC(dev) && (IS_I965G(dev) || plane == 0))
+		intel_update_fbc(crtc, &crtc->mode);
 	intel_update_watermarks(dev);
 
 	drm_vblank_post_modeset(dev, pipe);
@@ -2783,6 +2965,7 @@ static int intel_crtc_cursor_set(struct drm_crtc *crtc,
 	struct drm_gem_object *bo;
 	struct drm_i915_gem_object *obj_priv;
 	int pipe = intel_crtc->pipe;
+	int plane = intel_crtc->plane;
 	uint32_t control = (pipe == 0) ? CURACNTR : CURBCNTR;
 	uint32_t base = (pipe == 0) ? CURABASE : CURBBASE;
 	uint32_t temp = I915_READ(control);
@@ -2868,6 +3051,10 @@ static int intel_crtc_cursor_set(struct drm_crtc *crtc,
 			i915_gem_object_unpin(intel_crtc->cursor_bo);
 		drm_gem_object_unreference(intel_crtc->cursor_bo);
 	}
+
+	if (I915_HAS_FBC(dev) && (IS_I965G(dev) || plane == 0))
+		intel_update_fbc(crtc, &crtc->mode);
+
 	mutex_unlock(&dev->struct_mutex);
 
 	intel_crtc->cursor_addr = addr;
@@ -3549,6 +3736,14 @@ static void intel_crtc_init(struct drm_device *dev, int pipe)
 		intel_crtc->lut_b[i] = i;
 	}
 
+	/* Swap pipes & planes for FBC on pre-965 */
+	intel_crtc->pipe = pipe;
+	intel_crtc->plane = pipe;
+	if (IS_MOBILE(dev) && (IS_I9XX(dev) && !IS_I965G(dev))) {
+		DRM_DEBUG("swapping pipes & planes for FBC\n");
+		intel_crtc->plane = ((pipe == 0) ? 1 : 0);
+	}
+
 	intel_crtc->cursor_addr = 0;
 	intel_crtc->dpms_mode = DRM_MODE_DPMS_OFF;
 	drm_crtc_helper_add(&intel_crtc->base, &intel_helper_funcs);
@@ -3909,6 +4104,7 @@ void intel_modeset_cleanup(struct drm_device *dev)
 
 	mutex_unlock(&dev->struct_mutex);
 
+	i8xx_disable_fbc(dev);
 	drm_mode_config_cleanup(dev);
 }
 
diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
index b9e47f1..7478196 100644
--- a/drivers/gpu/drm/i915/intel_drv.h
+++ b/drivers/gpu/drm/i915/intel_drv.h
@@ -28,6 +28,7 @@
 #include <linux/i2c.h>
 #include <linux/i2c-id.h>
 #include <linux/i2c-algo-bit.h>
+#include "i915_drv.h"
 #include "drm_crtc.h"
 
 #include "drm_crtc_helper.h"
@@ -110,8 +111,8 @@ struct intel_output {
 
 struct intel_crtc {
 	struct drm_crtc base;
-	int pipe;
-	int plane;
+	enum pipe pipe;
+	enum plane plane;
 	struct drm_gem_object *cursor_bo;
 	uint32_t cursor_addr;
 	u8 lut_r[256], lut_g[256], lut_b[256];