#include <linux/log2.h>
#include <linux/slab.h>
-int
-nouveau_bo_sync_gpu(struct nouveau_bo *nvbo, struct nouveau_channel *chan)
-{
- struct nouveau_fence *prev_fence = nvbo->bo.sync_obj;
- int ret;
-
- if (!prev_fence || nouveau_fence_channel(prev_fence) == chan)
- return 0;
-
- spin_lock(&nvbo->bo.lock);
- ret = ttm_bo_wait(&nvbo->bo, false, false, false);
- spin_unlock(&nvbo->bo.lock);
- return ret;
-}
-
static void
nouveau_bo_del_ttm(struct ttm_buffer_object *bo)
{
struct drm_device *dev = dev_priv->dev;
struct nouveau_bo *nvbo = nouveau_bo(bo);
- ttm_bo_kunmap(&nvbo->kmap);
-
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);
-
+ nv10_mem_put_tile_region(dev, nvbo->tile, NULL);
kfree(nvbo);
}
nvbo->no_vm = no_vm;
nvbo->tile_mode = tile_mode;
nvbo->tile_flags = tile_flags;
+ nvbo->bo.bdev = &dev_priv->ttm.bdev;
- nouveau_bo_fixup_align(dev, tile_mode, tile_flags, &align, &size);
+ nouveau_bo_fixup_align(dev, tile_mode, nouveau_bo_tile_layout(nvbo),
+ &align, &size);
align >>= PAGE_SHIFT;
- nvbo->placement.fpfn = 0;
- nvbo->placement.lpfn = mappable ? dev_priv->fb_mappable_pages : 0;
nouveau_bo_placement_set(nvbo, flags, 0);
nvbo->channel = chan;
pl[(*n)++] = TTM_PL_FLAG_SYSTEM | flags;
}
+static void
+set_placement_range(struct nouveau_bo *nvbo, uint32_t type)
+{
+ struct drm_nouveau_private *dev_priv = nouveau_bdev(nvbo->bo.bdev);
+
+ if (dev_priv->card_type == NV_10 &&
+ nvbo->tile_mode && (type & TTM_PL_FLAG_VRAM)) {
+ /*
+ * Make sure that the color and depth buffers are handled
+ * by independent memory controller units. Up to a 9x
+ * speed up when alpha-blending and depth-test are enabled
+ * at the same time.
+ */
+ int vram_pages = dev_priv->vram_size >> PAGE_SHIFT;
+
+ if (nvbo->tile_flags & NOUVEAU_GEM_TILE_ZETA) {
+ nvbo->placement.fpfn = vram_pages / 2;
+ nvbo->placement.lpfn = ~0;
+ } else {
+ nvbo->placement.fpfn = 0;
+ nvbo->placement.lpfn = vram_pages / 2;
+ }
+ }
+}
+
void
nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t type, uint32_t busy)
{
pl->busy_placement = nvbo->busy_placements;
set_placement_list(nvbo->busy_placements, &pl->num_busy_placement,
type | busy, flags);
+
+ set_placement_range(nvbo, type);
}
int
void
nouveau_bo_unmap(struct nouveau_bo *nvbo)
{
- ttm_bo_kunmap(&nvbo->kmap);
+ if (nvbo)
+ ttm_bo_kunmap(&nvbo->kmap);
}
u16
man->default_caching = TTM_PL_FLAG_CACHED;
break;
case TTM_PL_VRAM:
+ man->func = &ttm_bo_manager_func;
man->flags = TTM_MEMTYPE_FLAG_FIXED |
TTM_MEMTYPE_FLAG_MAPPABLE;
man->available_caching = TTM_PL_FLAG_UNCACHED |
TTM_PL_FLAG_WC;
man->default_caching = TTM_PL_FLAG_WC;
- man->gpu_offset = dev_priv->vm_vram_base;
+ if (dev_priv->card_type == NV_50)
+ man->gpu_offset = 0x40000000;
+ else
+ man->gpu_offset = 0;
break;
case TTM_PL_TT:
+ man->func = &ttm_bo_manager_func;
switch (dev_priv->gart_info.type) {
case NOUVEAU_GART_AGP:
man->flags = TTM_MEMTYPE_FLAG_MAPPABLE;
- man->available_caching = TTM_PL_FLAG_UNCACHED;
- man->default_caching = TTM_PL_FLAG_UNCACHED;
+ man->available_caching = TTM_PL_FLAG_UNCACHED |
+ TTM_PL_FLAG_WC;
+ man->default_caching = TTM_PL_FLAG_WC;
break;
case NOUVEAU_GART_SGDMA:
man->flags = TTM_MEMTYPE_FLAG_MAPPABLE |
if (ret)
return ret;
- ret = ttm_bo_move_accel_cleanup(&nvbo->bo, fence, NULL,
- evict || (nvbo->channel &&
- nvbo->channel != chan),
+ ret = ttm_bo_move_accel_cleanup(&nvbo->bo, fence, NULL, evict,
no_wait_reserve, no_wait_gpu, new_mem);
- nouveau_fence_unref((void *)&fence);
+ nouveau_fence_unref(&fence);
return ret;
}
static inline uint32_t
-nouveau_bo_mem_ctxdma(struct nouveau_bo *nvbo, struct nouveau_channel *chan,
- struct ttm_mem_reg *mem)
+nouveau_bo_mem_ctxdma(struct ttm_buffer_object *bo,
+ struct nouveau_channel *chan, struct ttm_mem_reg *mem)
{
- if (chan == nouveau_bdev(nvbo->bo.bdev)->channel) {
+ struct nouveau_bo *nvbo = nouveau_bo(bo);
+
+ if (nvbo->no_vm) {
if (mem->mem_type == TTM_PL_TT)
return NvDmaGART;
return NvDmaVRAM;
}
static int
-nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict, bool intr,
- bool no_wait_reserve, bool no_wait_gpu,
- struct ttm_mem_reg *new_mem)
+nv50_bo_move_m2mf(struct nouveau_channel *chan, struct ttm_buffer_object *bo,
+ struct ttm_mem_reg *old_mem, 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;
+ struct nouveau_bo *nvbo = nouveau_bo(bo);
+ u64 length = (new_mem->num_pages << PAGE_SHIFT);
+ u64 src_offset, dst_offset;
int ret;
- chan = nvbo->channel;
- if (!chan || nvbo->tile_flags || nvbo->no_vm)
- chan = dev_priv->channel;
-
- src_offset = old_mem->mm_node->start << PAGE_SHIFT;
- dst_offset = new_mem->mm_node->start << PAGE_SHIFT;
- if (chan != dev_priv->channel) {
- if (old_mem->mem_type == TTM_PL_TT)
- src_offset += dev_priv->vm_gart_base;
- else
+ src_offset = old_mem->start << PAGE_SHIFT;
+ dst_offset = new_mem->start << PAGE_SHIFT;
+ if (!nvbo->no_vm) {
+ if (old_mem->mem_type == TTM_PL_VRAM)
src_offset += dev_priv->vm_vram_base;
-
- if (new_mem->mem_type == TTM_PL_TT)
- dst_offset += dev_priv->vm_gart_base;
else
+ src_offset += dev_priv->vm_gart_base;
+
+ if (new_mem->mem_type == TTM_PL_VRAM)
dst_offset += dev_priv->vm_vram_base;
+ else
+ dst_offset += dev_priv->vm_gart_base;
}
ret = RING_SPACE(chan, 3);
if (ret)
return ret;
- BEGIN_RING(chan, NvSubM2MF, NV_MEMORY_TO_MEMORY_FORMAT_DMA_SOURCE, 2);
- OUT_RING(chan, nouveau_bo_mem_ctxdma(nvbo, chan, old_mem));
- OUT_RING(chan, nouveau_bo_mem_ctxdma(nvbo, chan, new_mem));
- if (dev_priv->card_type >= NV_50) {
- ret = RING_SPACE(chan, 4);
+ BEGIN_RING(chan, NvSubM2MF, 0x0184, 2);
+ OUT_RING (chan, nouveau_bo_mem_ctxdma(bo, chan, old_mem));
+ OUT_RING (chan, nouveau_bo_mem_ctxdma(bo, chan, new_mem));
+
+ while (length) {
+ u32 amount, stride, height;
+
+ amount = min(length, (u64)(4 * 1024 * 1024));
+ stride = 16 * 4;
+ height = amount / stride;
+
+ if (new_mem->mem_type == TTM_PL_VRAM &&
+ nouveau_bo_tile_layout(nvbo)) {
+ ret = RING_SPACE(chan, 8);
+ if (ret)
+ return ret;
+
+ BEGIN_RING(chan, NvSubM2MF, 0x0200, 7);
+ OUT_RING (chan, 0);
+ OUT_RING (chan, 0);
+ OUT_RING (chan, stride);
+ OUT_RING (chan, height);
+ OUT_RING (chan, 1);
+ OUT_RING (chan, 0);
+ OUT_RING (chan, 0);
+ } else {
+ ret = RING_SPACE(chan, 2);
+ if (ret)
+ return ret;
+
+ BEGIN_RING(chan, NvSubM2MF, 0x0200, 1);
+ OUT_RING (chan, 1);
+ }
+ if (old_mem->mem_type == TTM_PL_VRAM &&
+ nouveau_bo_tile_layout(nvbo)) {
+ ret = RING_SPACE(chan, 8);
+ if (ret)
+ return ret;
+
+ BEGIN_RING(chan, NvSubM2MF, 0x021c, 7);
+ OUT_RING (chan, 0);
+ OUT_RING (chan, 0);
+ OUT_RING (chan, stride);
+ OUT_RING (chan, height);
+ OUT_RING (chan, 1);
+ OUT_RING (chan, 0);
+ OUT_RING (chan, 0);
+ } else {
+ ret = RING_SPACE(chan, 2);
+ if (ret)
+ return ret;
+
+ BEGIN_RING(chan, NvSubM2MF, 0x021c, 1);
+ OUT_RING (chan, 1);
+ }
+
+ ret = RING_SPACE(chan, 14);
if (ret)
return ret;
- BEGIN_RING(chan, NvSubM2MF, 0x0200, 1);
- OUT_RING(chan, 1);
- BEGIN_RING(chan, NvSubM2MF, 0x021c, 1);
- OUT_RING(chan, 1);
+
+ BEGIN_RING(chan, NvSubM2MF, 0x0238, 2);
+ OUT_RING (chan, upper_32_bits(src_offset));
+ OUT_RING (chan, upper_32_bits(dst_offset));
+ BEGIN_RING(chan, NvSubM2MF, 0x030c, 8);
+ OUT_RING (chan, lower_32_bits(src_offset));
+ OUT_RING (chan, lower_32_bits(dst_offset));
+ OUT_RING (chan, stride);
+ OUT_RING (chan, stride);
+ OUT_RING (chan, stride);
+ OUT_RING (chan, height);
+ OUT_RING (chan, 0x00000101);
+ OUT_RING (chan, 0x00000000);
+ BEGIN_RING(chan, NvSubM2MF, NV_MEMORY_TO_MEMORY_FORMAT_NOP, 1);
+ OUT_RING (chan, 0);
+
+ length -= amount;
+ src_offset += amount;
+ dst_offset += amount;
}
+ return 0;
+}
+
+static int
+nv04_bo_move_m2mf(struct nouveau_channel *chan, struct ttm_buffer_object *bo,
+ struct ttm_mem_reg *old_mem, struct ttm_mem_reg *new_mem)
+{
+ u32 src_offset = old_mem->start << PAGE_SHIFT;
+ u32 dst_offset = new_mem->start << PAGE_SHIFT;
+ u32 page_count = new_mem->num_pages;
+ int ret;
+
+ ret = RING_SPACE(chan, 3);
+ if (ret)
+ return ret;
+
+ BEGIN_RING(chan, NvSubM2MF, NV_MEMORY_TO_MEMORY_FORMAT_DMA_SOURCE, 2);
+ OUT_RING (chan, nouveau_bo_mem_ctxdma(bo, chan, old_mem));
+ OUT_RING (chan, nouveau_bo_mem_ctxdma(bo, chan, new_mem));
+
page_count = new_mem->num_pages;
while (page_count) {
int line_count = (page_count > 2047) ? 2047 : page_count;
- if (dev_priv->card_type >= NV_50) {
- ret = RING_SPACE(chan, 3);
- if (ret)
- return ret;
- BEGIN_RING(chan, NvSubM2MF, 0x0238, 2);
- OUT_RING(chan, upper_32_bits(src_offset));
- OUT_RING(chan, upper_32_bits(dst_offset));
- }
ret = RING_SPACE(chan, 11);
if (ret)
return ret;
+
BEGIN_RING(chan, NvSubM2MF,
NV_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 8);
- OUT_RING(chan, lower_32_bits(src_offset));
- OUT_RING(chan, lower_32_bits(dst_offset));
- OUT_RING(chan, PAGE_SIZE); /* src_pitch */
- OUT_RING(chan, PAGE_SIZE); /* dst_pitch */
- OUT_RING(chan, PAGE_SIZE); /* line_length */
- OUT_RING(chan, line_count);
- OUT_RING(chan, (1<<8)|(1<<0));
- OUT_RING(chan, 0);
+ OUT_RING (chan, src_offset);
+ OUT_RING (chan, dst_offset);
+ OUT_RING (chan, PAGE_SIZE); /* src_pitch */
+ OUT_RING (chan, PAGE_SIZE); /* dst_pitch */
+ OUT_RING (chan, PAGE_SIZE); /* line_length */
+ OUT_RING (chan, line_count);
+ OUT_RING (chan, 0x00000101);
+ OUT_RING (chan, 0x00000000);
BEGIN_RING(chan, NvSubM2MF, NV_MEMORY_TO_MEMORY_FORMAT_NOP, 1);
- OUT_RING(chan, 0);
+ OUT_RING (chan, 0);
page_count -= line_count;
src_offset += (PAGE_SIZE * line_count);
dst_offset += (PAGE_SIZE * line_count);
}
- return nouveau_bo_move_accel_cleanup(chan, nvbo, evict, no_wait_reserve, no_wait_gpu, new_mem);
+ return 0;
+}
+
+static int
+nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict, bool intr,
+ bool no_wait_reserve, bool no_wait_gpu,
+ struct ttm_mem_reg *new_mem)
+{
+ struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
+ struct nouveau_bo *nvbo = nouveau_bo(bo);
+ struct nouveau_channel *chan;
+ int ret;
+
+ chan = nvbo->channel;
+ if (!chan || nvbo->no_vm) {
+ chan = dev_priv->channel;
+ mutex_lock_nested(&chan->mutex, NOUVEAU_KCHANNEL_MUTEX);
+ }
+
+ if (dev_priv->card_type < NV_50)
+ ret = nv04_bo_move_m2mf(chan, bo, &bo->mem, new_mem);
+ else
+ ret = nv50_bo_move_m2mf(chan, bo, &bo->mem, new_mem);
+ if (ret == 0) {
+ ret = nouveau_bo_move_accel_cleanup(chan, nvbo, evict,
+ no_wait_reserve,
+ no_wait_gpu, new_mem);
+ }
+
+ if (chan == dev_priv->channel)
+ mutex_unlock(&chan->mutex);
+ return ret;
}
static int
ret = ttm_bo_move_ttm(bo, evict, no_wait_reserve, no_wait_gpu, new_mem);
out:
- if (tmp_mem.mm_node) {
- spin_lock(&bo->bdev->glob->lru_lock);
- drm_mm_put_block(tmp_mem.mm_node);
- spin_unlock(&bo->bdev->glob->lru_lock);
- }
-
+ ttm_bo_mem_put(bo, &tmp_mem);
return ret;
}
goto out;
out:
- if (tmp_mem.mm_node) {
- spin_lock(&bo->bdev->glob->lru_lock);
- drm_mm_put_block(tmp_mem.mm_node);
- spin_unlock(&bo->bdev->glob->lru_lock);
- }
-
+ ttm_bo_mem_put(bo, &tmp_mem);
return ret;
}
return 0;
}
- offset = new_mem->mm_node->start << PAGE_SHIFT;
+ offset = new_mem->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,
+ new_mem->size,
+ nouveau_bo_tile_layout(nvbo),
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);
+ nvbo->tile_mode,
+ nvbo->tile_flags);
}
return 0;
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);
-
+ nv10_mem_put_tile_region(dev, *old_tile, bo->sync_obj);
*old_tile = new_tile;
}
}
if (ret)
return ret;
- /* Software copy if the card isn't up and running yet. */
- if (!dev_priv->channel) {
- ret = ttm_bo_move_memcpy(bo, evict, no_wait_reserve, no_wait_gpu, new_mem);
- goto out;
- }
-
/* Fake bo copy. */
if (old_mem->mem_type == TTM_PL_SYSTEM && !bo->ttm) {
BUG_ON(bo->mem.mm_node != NULL);
goto out;
}
+ /* Software copy if the card isn't up and running yet. */
+ if (!dev_priv->channel) {
+ ret = ttm_bo_move_memcpy(bo, evict, no_wait_reserve, no_wait_gpu, new_mem);
+ goto out;
+ }
+
/* Hardware assisted copy. */
if (new_mem->mem_type == TTM_PL_SYSTEM)
ret = nouveau_bo_move_flipd(bo, evict, intr, no_wait_reserve, no_wait_gpu, new_mem);
case TTM_PL_TT:
#if __OS_HAS_AGP
if (dev_priv->gart_info.type == NOUVEAU_GART_AGP) {
- mem->bus.offset = mem->mm_node->start << PAGE_SHIFT;
+ mem->bus.offset = mem->start << PAGE_SHIFT;
mem->bus.base = dev_priv->gart_info.aper_base;
mem->bus.is_iomem = true;
}
#endif
break;
case TTM_PL_VRAM:
- mem->bus.offset = mem->mm_node->start << PAGE_SHIFT;
+ mem->bus.offset = mem->start << PAGE_SHIFT;
mem->bus.base = pci_resource_start(dev->pdev, 1);
mem->bus.is_iomem = true;
break;
static int
nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
{
- return 0;
+ struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
+ struct nouveau_bo *nvbo = nouveau_bo(bo);
+
+ /* as long as the bo isn't in vram, and isn't tiled, we've got
+ * nothing to do here.
+ */
+ if (bo->mem.mem_type != TTM_PL_VRAM) {
+ if (dev_priv->card_type < NV_50 ||
+ !nouveau_bo_tile_layout(nvbo))
+ return 0;
+ }
+
+ /* make sure bo is in mappable vram */
+ if (bo->mem.start + bo->mem.num_pages < dev_priv->fb_mappable_pages)
+ return 0;
+
+
+ nvbo->placement.fpfn = 0;
+ nvbo->placement.lpfn = dev_priv->fb_mappable_pages;
+ nouveau_bo_placement_set(nvbo, TTM_PL_VRAM, 0);
+ return ttm_bo_validate(bo, &nvbo->placement, false, true, false);
+}
+
+void
+nouveau_bo_fence(struct nouveau_bo *nvbo, struct nouveau_fence *fence)
+{
+ struct nouveau_fence *old_fence;
+
+ if (likely(fence))
+ nouveau_fence_ref(fence);
+
+ spin_lock(&nvbo->bo.bdev->fence_lock);
+ old_fence = nvbo->bo.sync_obj;
+ nvbo->bo.sync_obj = fence;
+ spin_unlock(&nvbo->bo.bdev->fence_lock);
+
+ nouveau_fence_unref(&old_fence);
}
struct ttm_bo_driver nouveau_bo_driver = {
.evict_flags = nouveau_bo_evict_flags,
.move = nouveau_bo_move,
.verify_access = nouveau_bo_verify_access,
- .sync_obj_signaled = nouveau_fence_signalled,
- .sync_obj_wait = nouveau_fence_wait,
- .sync_obj_flush = nouveau_fence_flush,
- .sync_obj_unref = nouveau_fence_unref,
- .sync_obj_ref = nouveau_fence_ref,
+ .sync_obj_signaled = __nouveau_fence_signalled,
+ .sync_obj_wait = __nouveau_fence_wait,
+ .sync_obj_flush = __nouveau_fence_flush,
+ .sync_obj_unref = __nouveau_fence_unref,
+ .sync_obj_ref = __nouveau_fence_ref,
.fault_reserve_notify = &nouveau_ttm_fault_reserve_notify,
.io_mem_reserve = &nouveau_ttm_io_mem_reserve,
.io_mem_free = &nouveau_ttm_io_mem_free,
projects / firefly-linux-kernel-4.4.55.git / blobdiff