Merge branch 'for-airlied' of /ssd/git/drm-nouveau-next into drm-linus

* 'for-airlied' of /ssd/git/drm-nouveau-next: (28 commits)
  drm/nv04: Fix set_operation software method.
  drm/nouveau: initialise DMA tracking parameters earlier
  drm/nouveau: use dma.max rather than pushbuf size for checking GET validity
  drm/nv04: differentiate between nv04/nv05
  drm/nouveau: Fix null deref in nouveau_fence_emit due to deleted fence
  drm/nv50: prevent a possible ctxprog hang
  drm/nouveau: have ttm's fault handler called directly
  drm/nv50: restore correct cache1 get/put address on fifoctx load
  drm/nouveau: create function for "dealing" with gpu lockup
  drm/nouveau: remove unused nouveau_channel_idle() function
  drm/nouveau: fix handling of fbcon colours in 8bpp
  drm/nv04: Context switching fixes.
  drm/nouveau: Use the software object for fencing.
  drm/nouveau: Allocate a per-channel instance of NV_SW.
  drm/nv50: make the blocksize depend on vram size
  drm/nouveau: better alignment of bo sizes and use roundup instead of ALIGN
  drm/nouveau: Don't skip card take down on nv0x.
  drm/nouveau: Implement nv42-nv43 TV load detection.
  drm/nouveau: Clean up the nv17-nv4x load detection code a bit.
  drm/nv50: fix fillrect color
  ...

29 files changed, 853 insertions, 496 deletions

diff --git a/drivers/gpu/drm/nouveau/Kconfig b/drivers/gpu/drm/nouveau/Kconfig
index b1bc1ea182b8..1175429da102 100644
--- a/drivers/gpu/drm/nouveau/Kconfig
+++ b/drivers/gpu/drm/nouveau/Kconfig
@@ -30,12 +30,11 @@ config DRM_NOUVEAU_DEBUG
 	  via debugfs.
 
 menu "I2C encoder or helper chips"
-     depends on DRM && I2C
+     depends on DRM && DRM_KMS_HELPER && I2C
 
 config DRM_I2C_CH7006
 	tristate "Chrontel ch7006 TV encoder"
-	depends on DRM_NOUVEAU
-	default m
+	default m if DRM_NOUVEAU
 	help
 	  Support for Chrontel ch7006 and similar TV encoders, found
 	  on some nVidia video cards.
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 0cad6d834eb2..e342a418d434 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -33,10 +33,13 @@
 #include "nouveau_drv.h"
 #include "nouveau_dma.h"
 
+#include <linux/log2.h>
+
 static void
 nouveau_bo_del_ttm(struct ttm_buffer_object *bo)
 {
 	struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
+	struct drm_device *dev = dev_priv->dev;
 	struct nouveau_bo *nvbo = nouveau_bo(bo);
 
 	ttm_bo_kunmap(&nvbo->kmap);
@@ -44,12 +47,87 @@ nouveau_bo_del_ttm(struct ttm_buffer_object *bo)
 	if (unlikely(nvbo->gem))
 		DRM_ERROR("bo %p still attached to GEM object\n", bo);
 
+	if (nvbo->tile)
+		nv10_mem_expire_tiling(dev, nvbo->tile, NULL);
+
 	spin_lock(&dev_priv->ttm.bo_list_lock);
 	list_del(&nvbo->head);
 	spin_unlock(&dev_priv->ttm.bo_list_lock);
 	kfree(nvbo);
 }
 
+static void
+nouveau_bo_fixup_align(struct drm_device *dev,
+		       uint32_t tile_mode, uint32_t tile_flags,
+		       int *align, int *size)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	/*
+	 * Some of the tile_flags have a periodic structure of N*4096 bytes,
+	 * align to to that as well as the page size. Overallocate memory to
+	 * avoid corruption of other buffer objects.
+	 */
+	if (dev_priv->card_type == NV_50) {
+		uint32_t block_size = nouveau_mem_fb_amount(dev) >> 15;
+		int i;
+
+		switch (tile_flags) {
+		case 0x1800:
+		case 0x2800:
+		case 0x4800:
+		case 0x7a00:
+			*size = roundup(*size, block_size);
+			if (is_power_of_2(block_size)) {
+				*size += 3 * block_size;
+				for (i = 1; i < 10; i++) {
+					*align = 12 * i * block_size;
+					if (!(*align % 65536))
+						break;
+				}
+			} else {
+				*size += 6 * block_size;
+				for (i = 1; i < 10; i++) {
+					*align = 8 * i * block_size;
+					if (!(*align % 65536))
+						break;
+				}
+			}
+			break;
+		default:
+			break;
+		}
+
+	} else {
+		if (tile_mode) {
+			if (dev_priv->chipset >= 0x40) {
+				*align = 65536;
+				*size = roundup(*size, 64 * tile_mode);
+
+			} else if (dev_priv->chipset >= 0x30) {
+				*align = 32768;
+				*size = roundup(*size, 64 * tile_mode);
+
+			} else if (dev_priv->chipset >= 0x20) {
+				*align = 16384;
+				*size = roundup(*size, 64 * tile_mode);
+
+			} else if (dev_priv->chipset >= 0x10) {
+				*align = 16384;
+				*size = roundup(*size, 32 * tile_mode);
+			}
+		}
+	}
+
+	/* ALIGN works only on powers of two. */
+	*size = roundup(*size, PAGE_SIZE);
+
+	if (dev_priv->card_type == NV_50) {
+		*size = roundup(*size, 65536);
+		*align = max(65536, *align);
+	}
+}
+
 int
 nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan,
 	       int size, int align, uint32_t flags, uint32_t tile_mode,
@@ -58,7 +136,7 @@ nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan,
 {
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_bo *nvbo;
-	int ret, n = 0;
+	int ret = 0;
 
 	nvbo = kzalloc(sizeof(struct nouveau_bo), GFP_KERNEL);
 	if (!nvbo)
@@ -70,59 +148,14 @@ nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan,
 	nvbo->tile_mode = tile_mode;
 	nvbo->tile_flags = tile_flags;
 
-	/*
-	 * Some of the tile_flags have a periodic structure of N*4096 bytes,
-	 * align to to that as well as the page size. Overallocate memory to
-	 * avoid corruption of other buffer objects.
-	 */
-	switch (tile_flags) {
-	case 0x1800:
-	case 0x2800:
-	case 0x4800:
-	case 0x7a00:
-		if (dev_priv->chipset >= 0xA0) {
-			/* This is based on high end cards with 448 bits
-			 * memory bus, could be different elsewhere.*/
-			size += 6 * 28672;
-			/* 8 * 28672 is the actual alignment requirement,
-			 * but we must also align to page size. */
-			align = 2 * 8 * 28672;
-		} else if (dev_priv->chipset >= 0x90) {
-			size += 3 * 16384;
-			align = 12 * 16834;
-		} else {
-			size += 3 * 8192;
-			/* 12 * 8192 is the actual alignment requirement,
-			 * but we must also align to page size. */
-			align = 2 * 12 * 8192;
-		}
-		break;
-	default:
-		break;
-	}
-
+	nouveau_bo_fixup_align(dev, tile_mode, tile_flags, &align, &size);
 	align >>= PAGE_SHIFT;
 
-	size = (size + (PAGE_SIZE - 1)) & ~(PAGE_SIZE - 1);
-	if (dev_priv->card_type == NV_50) {
-		size = (size + 65535) & ~65535;
-		if (align < (65536 / PAGE_SIZE))
-			align = (65536 / PAGE_SIZE);
-	}
-
-	if (flags & TTM_PL_FLAG_VRAM)
-		nvbo->placements[n++] = TTM_PL_FLAG_VRAM | TTM_PL_MASK_CACHING;
-	if (flags & TTM_PL_FLAG_TT)
-		nvbo->placements[n++] = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING;
 	nvbo->placement.fpfn = 0;
 	nvbo->placement.lpfn = mappable ? dev_priv->fb_mappable_pages : 0;
-	nvbo->placement.placement = nvbo->placements;
-	nvbo->placement.busy_placement = nvbo->placements;
-	nvbo->placement.num_placement = n;
-	nvbo->placement.num_busy_placement = n;
+	nouveau_bo_placement_set(nvbo, flags);
 
 	nvbo->channel = chan;
-	nouveau_bo_placement_set(nvbo, flags);
 	ret = ttm_bo_init(&dev_priv->ttm.bdev, &nvbo->bo, size,
 			  ttm_bo_type_device, &nvbo->placement, align, 0,
 			  false, NULL, size, nouveau_bo_del_ttm);
@@ -421,6 +454,7 @@ nouveau_bo_evict_flags(struct ttm_buffer_object *bo, struct ttm_placement *pl)
 /* GPU-assisted copy using NV_MEMORY_TO_MEMORY_FORMAT, can access
  * TTM_PL_{VRAM,TT} directly.
  */
+
 static int
 nouveau_bo_move_accel_cleanup(struct nouveau_channel *chan,
 			      struct nouveau_bo *nvbo, bool evict, bool no_wait,
@@ -455,11 +489,12 @@ nouveau_bo_mem_ctxdma(struct nouveau_bo *nvbo, struct nouveau_channel *chan,
 }
 
 static int
-nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict, int no_wait,
-		     struct ttm_mem_reg *old_mem, struct ttm_mem_reg *new_mem)
+nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict, bool intr,
+		     int no_wait, struct ttm_mem_reg *new_mem)
 {
 	struct nouveau_bo *nvbo = nouveau_bo(bo);
 	struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
+	struct ttm_mem_reg *old_mem = &bo->mem;
 	struct nouveau_channel *chan;
 	uint64_t src_offset, dst_offset;
 	uint32_t page_count;
@@ -547,7 +582,7 @@ nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict, bool intr,
 
 	placement.fpfn = placement.lpfn = 0;
 	placement.num_placement = placement.num_busy_placement = 1;
-	placement.placement = &placement_memtype;
+	placement.placement = placement.busy_placement = &placement_memtype;
 
 	tmp_mem = *new_mem;
 	tmp_mem.mm_node = NULL;
@@ -559,7 +594,7 @@ nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict, bool intr,
 	if (ret)
 		goto out;
 
-	ret = nouveau_bo_move_m2mf(bo, true, no_wait, &bo->mem, &tmp_mem);
+	ret = nouveau_bo_move_m2mf(bo, true, intr, no_wait, &tmp_mem);
 	if (ret)
 		goto out;
 
@@ -585,7 +620,7 @@ nouveau_bo_move_flips(struct ttm_buffer_object *bo, bool evict, bool intr,
 
 	placement.fpfn = placement.lpfn = 0;
 	placement.num_placement = placement.num_busy_placement = 1;
-	placement.placement = &placement_memtype;
+	placement.placement = placement.busy_placement = &placement_memtype;
 
 	tmp_mem = *new_mem;
 	tmp_mem.mm_node = NULL;
@@ -597,7 +632,7 @@ nouveau_bo_move_flips(struct ttm_buffer_object *bo, bool evict, bool intr,
 	if (ret)
 		goto out;
 
-	ret = nouveau_bo_move_m2mf(bo, true, no_wait, &bo->mem, new_mem);
+	ret = nouveau_bo_move_m2mf(bo, evict, intr, no_wait, new_mem);
 	if (ret)
 		goto out;
 
@@ -612,52 +647,106 @@ out:
 }
 
 static int
-nouveau_bo_move(struct ttm_buffer_object *bo, bool evict, bool intr,
-		bool no_wait, struct ttm_mem_reg *new_mem)
+nouveau_bo_vm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *new_mem,
+		   struct nouveau_tile_reg **new_tile)
 {
 	struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
-	struct nouveau_bo *nvbo = nouveau_bo(bo);
 	struct drm_device *dev = dev_priv->dev;
-	struct ttm_mem_reg *old_mem = &bo->mem;
+	struct nouveau_bo *nvbo = nouveau_bo(bo);
+	uint64_t offset;
 	int ret;
 
-	if (dev_priv->card_type == NV_50 && new_mem->mem_type == TTM_PL_VRAM &&
-	    !nvbo->no_vm) {
-		uint64_t offset = new_mem->mm_node->start << PAGE_SHIFT;
+	if (nvbo->no_vm || new_mem->mem_type != TTM_PL_VRAM) {
+		/* Nothing to do. */
+		*new_tile = NULL;
+		return 0;
+	}
+
+	offset = new_mem->mm_node->start << PAGE_SHIFT;
 
+	if (dev_priv->card_type == NV_50) {
 		ret = nv50_mem_vm_bind_linear(dev,
 					      offset + dev_priv->vm_vram_base,
 					      new_mem->size, nvbo->tile_flags,
 					      offset);
 		if (ret)
 			return ret;
+
+	} else if (dev_priv->card_type >= NV_10) {
+		*new_tile = nv10_mem_set_tiling(dev, offset, new_mem->size,
+						nvbo->tile_mode);
 	}
 
+	return 0;
+}
+
+static void
+nouveau_bo_vm_cleanup(struct ttm_buffer_object *bo,
+		      struct nouveau_tile_reg *new_tile,
+		      struct nouveau_tile_reg **old_tile)
+{
+	struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
+	struct drm_device *dev = dev_priv->dev;
+
+	if (dev_priv->card_type >= NV_10 &&
+	    dev_priv->card_type < NV_50) {
+		if (*old_tile)
+			nv10_mem_expire_tiling(dev, *old_tile, bo->sync_obj);
+
+		*old_tile = new_tile;
+	}
+}
+
+static int
+nouveau_bo_move(struct ttm_buffer_object *bo, bool evict, bool intr,
+		bool no_wait, struct ttm_mem_reg *new_mem)
+{
+	struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
+	struct nouveau_bo *nvbo = nouveau_bo(bo);
+	struct ttm_mem_reg *old_mem = &bo->mem;
+	struct nouveau_tile_reg *new_tile = NULL;
+	int ret = 0;
+
+	ret = nouveau_bo_vm_bind(bo, new_mem, &new_tile);
+	if (ret)
+		return ret;
+
+	/* Software copy if the card isn't up and running yet. */
 	if (dev_priv->init_state != NOUVEAU_CARD_INIT_DONE ||
-	    !dev_priv->channel)
-		return ttm_bo_move_memcpy(bo, evict, no_wait, new_mem);
+	    !dev_priv->channel) {
+		ret = ttm_bo_move_memcpy(bo, evict, no_wait, new_mem);
+		goto out;
+	}
 
+	/* Fake bo copy. */
 	if (old_mem->mem_type == TTM_PL_SYSTEM && !bo->ttm) {
 		BUG_ON(bo->mem.mm_node != NULL);
 		bo->mem = *new_mem;
 		new_mem->mm_node = NULL;
-		return 0;
+		goto out;
 	}
 
-	if (new_mem->mem_type == TTM_PL_SYSTEM) {
-		if (old_mem->mem_type == TTM_PL_SYSTEM)
-			return ttm_bo_move_memcpy(bo, evict, no_wait, new_mem);
-		if (nouveau_bo_move_flipd(bo, evict, intr, no_wait, new_mem))
-			return ttm_bo_move_memcpy(bo, evict, no_wait, new_mem);
-	} else if (old_mem->mem_type == TTM_PL_SYSTEM) {
-		if (nouveau_bo_move_flips(bo, evict, intr, no_wait, new_mem))
-			return ttm_bo_move_memcpy(bo, evict, no_wait, new_mem);
-	} else {
-		if (nouveau_bo_move_m2mf(bo, evict, no_wait, old_mem, new_mem))
-			return ttm_bo_move_memcpy(bo, evict, no_wait, new_mem);
-	}
+	/* Hardware assisted copy. */
+	if (new_mem->mem_type == TTM_PL_SYSTEM)
+		ret = nouveau_bo_move_flipd(bo, evict, intr, no_wait, new_mem);
+	else if (old_mem->mem_type == TTM_PL_SYSTEM)
+		ret = nouveau_bo_move_flips(bo, evict, intr, no_wait, new_mem);
+	else
+		ret = nouveau_bo_move_m2mf(bo, evict, intr, no_wait, new_mem);
 
-	return 0;
+	if (!ret)
+		goto out;
+
+	/* Fallback to software copy. */
+	ret = ttm_bo_move_memcpy(bo, evict, no_wait, new_mem);
+
+out:
+	if (ret)
+		nouveau_bo_vm_cleanup(bo, NULL, &new_tile);
+	else
+		nouveau_bo_vm_cleanup(bo, new_tile, &nvbo->tile);
+
+	return ret;
 }
 
 static int
diff --git a/drivers/gpu/drm/nouveau/nouveau_channel.c b/drivers/gpu/drm/nouveau/nouveau_channel.c
index 9aaa972f8822..343d718a9667 100644
--- a/drivers/gpu/drm/nouveau/nouveau_channel.c
+++ b/drivers/gpu/drm/nouveau/nouveau_channel.c
@@ -158,6 +158,8 @@ nouveau_channel_alloc(struct drm_device *dev, struct nouveau_channel **chan_ret,
 		return ret;
 	}
 
+	nouveau_dma_pre_init(chan);
+
 	/* Locate channel's user control regs */
 	if (dev_priv->card_type < NV_40)
 		user = NV03_USER(channel);
@@ -235,47 +237,6 @@ nouveau_channel_alloc(struct drm_device *dev, struct nouveau_channel **chan_ret,
 	return 0;
 }
 
-int
-nouveau_channel_idle(struct nouveau_channel *chan)
-{
-	struct drm_device *dev = chan->dev;
-	struct drm_nouveau_private *dev_priv = dev->dev_private;
-	struct nouveau_engine *engine = &dev_priv->engine;
-	uint32_t caches;
-	int idle;
-
-	if (!chan) {
-		NV_ERROR(dev, "no channel...\n");
-		return 1;
-	}
-
-	caches = nv_rd32(dev, NV03_PFIFO_CACHES);
-	nv_wr32(dev, NV03_PFIFO_CACHES, caches & ~1);
-
-	if (engine->fifo.channel_id(dev) != chan->id) {
-		struct nouveau_gpuobj *ramfc =
-			chan->ramfc ? chan->ramfc->gpuobj : NULL;
-
-		if (!ramfc) {
-			NV_ERROR(dev, "No RAMFC for channel %d\n", chan->id);
-			return 1;
-		}
-
-		engine->instmem.prepare_access(dev, false);
-		if (nv_ro32(dev, ramfc, 0) != nv_ro32(dev, ramfc, 1))
-			idle = 0;
-		else
-			idle = 1;
-		engine->instmem.finish_access(dev);
-	} else {
-		idle = (nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_GET) ==
-			nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_PUT));
-	}
-
-	nv_wr32(dev, NV03_PFIFO_CACHES, caches);
-	return idle;
-}
-
 /* stops a fifo */
 void
 nouveau_channel_free(struct nouveau_channel *chan)
@@ -414,7 +375,9 @@ nouveau_ioctl_fifo_alloc(struct drm_device *dev, void *data,
 		init->subchan[0].grclass = 0x0039;
 	else
 		init->subchan[0].grclass = 0x5039;
-	init->nr_subchan = 1;
+	init->subchan[1].handle = NvSw;
+	init->subchan[1].grclass = NV_SW;
+	init->nr_subchan = 2;
 
 	/* Named memory object area */
 	ret = drm_gem_handle_create(file_priv, chan->notifier_bo->gem,
diff --git a/drivers/gpu/drm/nouveau/nouveau_dma.c b/drivers/gpu/drm/nouveau/nouveau_dma.c
index 703553687b20..7afbe8b40d51 100644
--- a/drivers/gpu/drm/nouveau/nouveau_dma.c
+++ b/drivers/gpu/drm/nouveau/nouveau_dma.c
@@ -29,12 +29,22 @@
 #include "nouveau_drv.h"
 #include "nouveau_dma.h"
 
+void
+nouveau_dma_pre_init(struct nouveau_channel *chan)
+{
+	chan->dma.max  = (chan->pushbuf_bo->bo.mem.size >> 2) - 2;
+	chan->dma.put  = 0;
+	chan->dma.cur  = chan->dma.put;
+	chan->dma.free = chan->dma.max - chan->dma.cur;
+}
+
 int
 nouveau_dma_init(struct nouveau_channel *chan)
 {
 	struct drm_device *dev = chan->dev;
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_gpuobj *m2mf = NULL;
+	struct nouveau_gpuobj *nvsw = NULL;
 	int ret, i;
 
 	/* Create NV_MEMORY_TO_MEMORY_FORMAT for buffer moves */
@@ -47,6 +57,15 @@ nouveau_dma_init(struct nouveau_channel *chan)
 	if (ret)
 		return ret;
 
+	/* Create an NV_SW object for various sync purposes */
+	ret = nouveau_gpuobj_sw_new(chan, NV_SW, &nvsw);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_ref_add(dev, chan, NvSw, nvsw, NULL);
+	if (ret)
+		return ret;
+
 	/* NV_MEMORY_TO_MEMORY_FORMAT requires a notifier object */
 	ret = nouveau_notifier_alloc(chan, NvNotify0, 32, &chan->m2mf_ntfy);
 	if (ret)
@@ -64,12 +83,6 @@ nouveau_dma_init(struct nouveau_channel *chan)
 			return ret;
 	}
 
-	/* Initialise DMA vars */
-	chan->dma.max  = (chan->pushbuf_bo->bo.mem.size >> 2) - 2;
-	chan->dma.put  = 0;
-	chan->dma.cur  = chan->dma.put;
-	chan->dma.free = chan->dma.max - chan->dma.cur;
-
 	/* Insert NOPS for NOUVEAU_DMA_SKIPS */
 	ret = RING_SPACE(chan, NOUVEAU_DMA_SKIPS);
 	if (ret)
@@ -87,6 +100,13 @@ nouveau_dma_init(struct nouveau_channel *chan)
 	BEGIN_RING(chan, NvSubM2MF, NV_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY, 1);
 	OUT_RING(chan, NvNotify0);
 
+	/* Initialise NV_SW */
+	ret = RING_SPACE(chan, 2);
+	if (ret)
+		return ret;
+	BEGIN_RING(chan, NvSubSw, 0, 1);
+	OUT_RING(chan, NvSw);
+
 	/* Sit back and pray the channel works.. */
 	FIRE_RING(chan);
 
@@ -113,7 +133,7 @@ READ_GET(struct nouveau_channel *chan, uint32_t *get)
 
 	val = nvchan_rd32(chan, chan->user_get);
 	if (val < chan->pushbuf_base ||
-	    val >= chan->pushbuf_base + chan->pushbuf_bo->bo.mem.size) {
+	    val > chan->pushbuf_base + (chan->dma.max << 2)) {
 		/* meaningless to dma_wait() except to know whether the
 		 * GPU has stalled or not
 		 */
diff --git a/drivers/gpu/drm/nouveau/nouveau_dma.h b/drivers/gpu/drm/nouveau/nouveau_dma.h
index 04e85d8f757e..dabfd655f93e 100644
--- a/drivers/gpu/drm/nouveau/nouveau_dma.h
+++ b/drivers/gpu/drm/nouveau/nouveau_dma.h
@@ -46,10 +46,11 @@
 /* Hardcoded object assignments to subchannels (subchannel id). */
 enum {
 	NvSubM2MF	= 0,
-	NvSub2D		= 1,
-	NvSubCtxSurf2D  = 1,
-	NvSubGdiRect    = 2,
-	NvSubImageBlit  = 3
+	NvSubSw		= 1,
+	NvSub2D		= 2,
+	NvSubCtxSurf2D  = 2,
+	NvSubGdiRect    = 3,
+	NvSubImageBlit  = 4
 };
 
 /* Object handles. */
@@ -67,6 +68,7 @@ enum {
 	NvClipRect	= 0x8000000b,
 	NvGdiRect	= 0x8000000c,
 	NvImageBlit	= 0x8000000d,
+	NvSw		= 0x8000000e,
 
 	/* G80+ display objects */
 	NvEvoVRAM	= 0x01000000,
diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h
index 5f8cbb79c499..026419fe8791 100644
--- a/drivers/gpu/drm/nouveau/nouveau_drv.h
+++ b/drivers/gpu/drm/nouveau/nouveau_drv.h
@@ -59,11 +59,19 @@ struct nouveau_grctx;
 #define MAX_NUM_DCB_ENTRIES 16
 
 #define NOUVEAU_MAX_CHANNEL_NR 128
+#define NOUVEAU_MAX_TILE_NR 15
 
 #define NV50_VM_MAX_VRAM (2*1024*1024*1024ULL)
 #define NV50_VM_BLOCK    (512*1024*1024ULL)
 #define NV50_VM_VRAM_NR  (NV50_VM_MAX_VRAM / NV50_VM_BLOCK)
 
+struct nouveau_tile_reg {
+	struct nouveau_fence *fence;
+	uint32_t addr;
+	uint32_t size;
+	bool used;
+};
+
 struct nouveau_bo {
 	struct ttm_buffer_object bo;
 	struct ttm_placement placement;
@@ -83,6 +91,7 @@ struct nouveau_bo {
 
 	uint32_t tile_mode;
 	uint32_t tile_flags;
+	struct nouveau_tile_reg *tile;
 
 	struct drm_gem_object *gem;
 	struct drm_file *cpu_filp;
@@ -277,8 +286,13 @@ struct nouveau_timer_engine {
 };
 
 struct nouveau_fb_engine {
+	int num_tiles;
+
 	int  (*init)(struct drm_device *dev);
 	void (*takedown)(struct drm_device *dev);
+
+	void (*set_region_tiling)(struct drm_device *dev, int i, uint32_t addr,
+				 uint32_t size, uint32_t pitch);
 };
 
 struct nouveau_fifo_engine {
@@ -292,6 +306,8 @@ struct nouveau_fifo_engine {
 	void (*disable)(struct drm_device *);
 	void (*enable)(struct drm_device *);
 	bool (*reassign)(struct drm_device *, bool enable);
+	bool (*cache_flush)(struct drm_device *dev);
+	bool (*cache_pull)(struct drm_device *dev, bool enable);
 
 	int  (*channel_id)(struct drm_device *);
 
@@ -330,6 +346,9 @@ struct nouveau_pgraph_engine {
 	void (*destroy_context)(struct nouveau_channel *);
 	int  (*load_context)(struct nouveau_channel *);
 	int  (*unload_context)(struct drm_device *);
+
+	void (*set_region_tiling)(struct drm_device *dev, int i, uint32_t addr,
+				  uint32_t size, uint32_t pitch);
 };
 
 struct nouveau_engine {
@@ -548,6 +567,12 @@ struct drm_nouveau_private {
 		unsigned long sg_handle;
 	} gart_info;
 
+	/* nv10-nv40 tiling regions */
+	struct {
+		struct nouveau_tile_reg reg[NOUVEAU_MAX_TILE_NR];
+		spinlock_t lock;
+	} tile;
+
 	/* G8x/G9x virtual address space */
 	uint64_t vm_gart_base;
 	uint64_t vm_gart_size;
@@ -685,6 +710,13 @@ 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 *);
 extern void nouveau_mem_close(struct drm_device *);
+extern struct nouveau_tile_reg *nv10_mem_set_tiling(struct drm_device *dev,
+						    uint32_t addr,
+						    uint32_t size,
+						    uint32_t pitch);
+extern void nv10_mem_expire_tiling(struct drm_device *dev,
+				   struct nouveau_tile_reg *tile,
+				   struct nouveau_fence *fence);
 extern int  nv50_mem_vm_bind_linear(struct drm_device *, uint64_t virt,
 				    uint32_t size, uint32_t flags,
 				    uint64_t phys);
@@ -713,7 +745,6 @@ extern int  nouveau_channel_alloc(struct drm_device *dev,
 				  struct drm_file *file_priv,
 				  uint32_t fb_ctxdma, uint32_t tt_ctxdma);
 extern void nouveau_channel_free(struct nouveau_channel *);
-extern int  nouveau_channel_idle(struct nouveau_channel *chan);
 
 /* nouveau_object.c */
 extern int  nouveau_gpuobj_early_init(struct drm_device *);
@@ -756,6 +787,8 @@ extern int nouveau_gpuobj_gart_dma_new(struct nouveau_channel *,
 				       uint32_t *o_ret);
 extern int nouveau_gpuobj_gr_new(struct nouveau_channel *, int class,
 				 struct nouveau_gpuobj **);
+extern int nouveau_gpuobj_sw_new(struct nouveau_channel *, int class,
+				 struct nouveau_gpuobj **);
 extern int nouveau_ioctl_grobj_alloc(struct drm_device *, void *data,
 				     struct drm_file *);
 extern int nouveau_ioctl_gpuobj_free(struct drm_device *, void *data,
@@ -804,6 +837,7 @@ nouveau_debugfs_channel_fini(struct nouveau_channel *chan)
 #endif
 
 /* nouveau_dma.c */
+extern void nouveau_dma_pre_init(struct nouveau_channel *);
 extern int  nouveau_dma_init(struct nouveau_channel *);
 extern int  nouveau_dma_wait(struct nouveau_channel *, int size);
 
@@ -879,16 +913,22 @@ extern void nv04_fb_takedown(struct drm_device *);
 /* nv10_fb.c */
 extern int  nv10_fb_init(struct drm_device *);
 extern void nv10_fb_takedown(struct drm_device *);
+extern void nv10_fb_set_region_tiling(struct drm_device *, int, uint32_t,
+				      uint32_t, uint32_t);
 
 /* nv40_fb.c */
 extern int  nv40_fb_init(struct drm_device *);
 extern void nv40_fb_takedown(struct drm_device *);
+extern void nv40_fb_set_region_tiling(struct drm_device *, int, uint32_t,
+				      uint32_t, uint32_t);
 
 /* nv04_fifo.c */
 extern int  nv04_fifo_init(struct drm_device *);
 extern void nv04_fifo_disable(struct drm_device *);
 extern void nv04_fifo_enable(struct drm_device *);
 extern bool nv04_fifo_reassign(struct drm_device *, bool);
+extern bool nv04_fifo_cache_flush(struct drm_device *);
+extern bool nv04_fifo_cache_pull(struct drm_device *, bool);
 extern int  nv04_fifo_channel_id(struct drm_device *);
 extern int  nv04_fifo_create_context(struct nouveau_channel *);
 extern void nv04_fifo_destroy_context(struct nouveau_channel *);
@@ -941,6 +981,8 @@ extern void nv10_graph_destroy_context(struct nouveau_channel *);
 extern int  nv10_graph_load_context(struct nouveau_channel *);
 extern int  nv10_graph_unload_context(struct drm_device *);
 extern void nv10_graph_context_switch(struct drm_device *);
+extern void nv10_graph_set_region_tiling(struct drm_device *, int, uint32_t,
+					 uint32_t, uint32_t);
 
 /* nv20_graph.c */
 extern struct nouveau_pgraph_object_class nv20_graph_grclass[];
@@ -952,6 +994,8 @@ extern int  nv20_graph_unload_context(struct drm_device *);
 extern int  nv20_graph_init(struct drm_device *);
 extern void nv20_graph_takedown(struct drm_device *);
 extern int  nv30_graph_init(struct drm_device *);
+extern void nv20_graph_set_region_tiling(struct drm_device *, int, uint32_t,
+					 uint32_t, uint32_t);
 
 /* nv40_graph.c */
 extern struct nouveau_pgraph_object_class nv40_graph_grclass[];
@@ -963,6 +1007,8 @@ extern void nv40_graph_destroy_context(struct nouveau_channel *);
 extern int  nv40_graph_load_context(struct nouveau_channel *);
 extern int  nv40_graph_unload_context(struct drm_device *);
 extern void nv40_grctx_init(struct nouveau_grctx *);
+extern void nv40_graph_set_region_tiling(struct drm_device *, int, uint32_t,
+					 uint32_t, uint32_t);
 
 /* nv50_graph.c */
 extern struct nouveau_pgraph_object_class nv50_graph_grclass[];
@@ -1030,8 +1076,7 @@ extern long nouveau_compat_ioctl(struct file *file, unsigned int cmd,
 
 /* nv04_dac.c */
 extern int nv04_dac_create(struct drm_device *dev, struct dcb_entry *entry);
-extern enum drm_connector_status nv17_dac_detect(struct drm_encoder *encoder,
-						 struct drm_connector *connector);
+extern uint32_t nv17_dac_sample_load(struct drm_encoder *encoder);
 extern int nv04_dac_output_offset(struct drm_encoder *encoder);
 extern void nv04_dac_update_dacclk(struct drm_encoder *encoder, bool enable);
 
@@ -1049,9 +1094,6 @@ extern int nv04_tv_create(struct drm_device *dev, struct dcb_entry *entry);
 
 /* nv17_tv.c */
 extern int nv17_tv_create(struct drm_device *dev, struct dcb_entry *entry);
-extern enum drm_connector_status nv17_tv_detect(struct drm_encoder *encoder,
-						struct drm_connector *connector,
-						uint32_t pin_mask);
 
 /* nv04_display.c */
 extern int nv04_display_create(struct drm_device *);
@@ -1290,14 +1332,14 @@ nv_two_reg_pll(struct drm_device *dev)
 	return false;
 }
 
-#define NV50_NVSW                                                    0x0000506e
-#define NV50_NVSW_DMA_SEMAPHORE                                      0x00000060
-#define NV50_NVSW_SEMAPHORE_OFFSET                                   0x00000064
-#define NV50_NVSW_SEMAPHORE_ACQUIRE                                  0x00000068
-#define NV50_NVSW_SEMAPHORE_RELEASE                                  0x0000006c
-#define NV50_NVSW_DMA_VBLSEM                                         0x0000018c
-#define NV50_NVSW_VBLSEM_OFFSET                                      0x00000400
-#define NV50_NVSW_VBLSEM_RELEASE_VALUE                               0x00000404
-#define NV50_NVSW_VBLSEM_RELEASE                                     0x00000408
+#define NV_SW                                                        0x0000506e
+#define NV_SW_DMA_SEMAPHORE                                          0x00000060
+#define NV_SW_SEMAPHORE_OFFSET                                       0x00000064
+#define NV_SW_SEMAPHORE_ACQUIRE                                      0x00000068
+#define NV_SW_SEMAPHORE_RELEASE                                      0x0000006c
+#define NV_SW_DMA_VBLSEM                                             0x0000018c
+#define NV_SW_VBLSEM_OFFSET                                          0x00000400
+#define NV_SW_VBLSEM_RELEASE_VALUE                                   0x00000404
+#define NV_SW_VBLSEM_RELEASE                                         0x00000408
 
 #endif /* __NOUVEAU_DRV_H__ */
diff --git a/drivers/gpu/drm/nouveau/nouveau_fbcon.c b/drivers/gpu/drm/nouveau/nouveau_fbcon.c
index 84af25c238b6..0b05c869e0e7 100644
--- a/drivers/gpu/drm/nouveau/nouveau_fbcon.c
+++ b/drivers/gpu/drm/nouveau/nouveau_fbcon.c
@@ -64,8 +64,7 @@ nouveau_fbcon_sync(struct fb_info *info)
 		return 0;
 
 	if (RING_SPACE(chan, 4)) {
-		NV_ERROR(dev, "GPU lockup - switching to software fbcon\n");
-		info->flags |= FBINFO_HWACCEL