}
static int
-i915_gem_wait_for_error(struct drm_device *dev)
+i915_gem_wait_for_error(struct i915_gpu_error *error)
{
- struct drm_i915_private *dev_priv = dev->dev_private;
- struct completion *x = &dev_priv->error_completion;
+ struct completion *x = &error->completion;
unsigned long flags;
int ret;
- if (!atomic_read(&dev_priv->mm.wedged))
+ if (!atomic_read(&error->wedged))
return 0;
/*
return ret;
}
- if (atomic_read(&dev_priv->mm.wedged)) {
+ if (atomic_read(&error->wedged)) {
/* GPU is hung, bump the completion count to account for
* the token we just consumed so that we never hit zero and
* end up waiting upon a subsequent completion event that
int i915_mutex_lock_interruptible(struct drm_device *dev)
{
+ struct drm_i915_private *dev_priv = dev->dev_private;
int ret;
- ret = i915_gem_wait_for_error(dev);
+ ret = i915_gem_wait_for_error(&dev_priv->gpu_error);
if (ret)
return ret;
i915_gem_init_ioctl(struct drm_device *dev, void *data,
struct drm_file *file)
{
+ struct drm_i915_private *dev_priv = dev->dev_private;
struct drm_i915_gem_init *args = data;
if (drm_core_check_feature(dev, DRIVER_MODESET))
return -ENODEV;
mutex_lock(&dev->struct_mutex);
- i915_gem_init_global_gtt(dev, args->gtt_start,
- args->gtt_end, args->gtt_end);
+ i915_gem_setup_global_gtt(dev, args->gtt_start, args->gtt_end,
+ args->gtt_end);
+ dev_priv->gtt.mappable_end = args->gtt_end;
mutex_unlock(&dev->struct_mutex);
return 0;
pinned += obj->gtt_space->size;
mutex_unlock(&dev->struct_mutex);
- args->aper_size = dev_priv->mm.gtt_total;
+ args->aper_size = dev_priv->gtt.total;
args->aper_available_size = args->aper_size - pinned;
return 0;
* source page isn't available. Return the error and we'll
* retry in the slow path.
*/
- if (fast_user_write(dev_priv->mm.gtt_mapping, page_base,
+ if (fast_user_write(dev_priv->gtt.mappable, page_base,
page_offset, user_data, page_length)) {
ret = -EFAULT;
goto out_unpin;
}
int
-i915_gem_check_wedge(struct drm_i915_private *dev_priv,
+i915_gem_check_wedge(struct i915_gpu_error *error,
bool interruptible)
{
- if (atomic_read(&dev_priv->mm.wedged)) {
- struct completion *x = &dev_priv->error_completion;
+ if (atomic_read(&error->wedged)) {
+ struct completion *x = &error->completion;
bool recovery_complete;
unsigned long flags;
#define EXIT_COND \
(i915_seqno_passed(ring->get_seqno(ring, false), seqno) || \
- atomic_read(&dev_priv->mm.wedged))
+ atomic_read(&dev_priv->gpu_error.wedged))
do {
if (interruptible)
end = wait_event_interruptible_timeout(ring->irq_queue,
end = wait_event_timeout(ring->irq_queue, EXIT_COND,
timeout_jiffies);
- ret = i915_gem_check_wedge(dev_priv, interruptible);
+ ret = i915_gem_check_wedge(&dev_priv->gpu_error, interruptible);
if (ret)
end = ret;
} while (end == 0 && wait_forever);
BUG_ON(!mutex_is_locked(&dev->struct_mutex));
BUG_ON(seqno == 0);
- ret = i915_gem_check_wedge(dev_priv, interruptible);
+ ret = i915_gem_check_wedge(&dev_priv->gpu_error, interruptible);
if (ret)
return ret;
if (seqno == 0)
return 0;
- ret = i915_gem_check_wedge(dev_priv, true);
+ ret = i915_gem_check_wedge(&dev_priv->gpu_error, true);
if (ret)
return ret;
obj->fault_mappable = true;
- pfn = ((dev_priv->mm.gtt_base_addr + obj->gtt_offset) >> PAGE_SHIFT) +
+ pfn = ((dev_priv->gtt.mappable_base + obj->gtt_offset) >> PAGE_SHIFT) +
page_offset;
/* Finally, remap it using the new GTT offset */
/* If this -EIO is due to a gpu hang, give the reset code a
* chance to clean up the mess. Otherwise return the proper
* SIGBUS. */
- if (!atomic_read(&dev_priv->mm.wedged))
+ if (!atomic_read(&dev_priv->gpu_error.wedged))
return VM_FAULT_SIGBUS;
case -EAGAIN:
/* Give the error handler a chance to run and move the
obj->fault_mappable = false;
}
-static uint32_t
+uint32_t
i915_gem_get_gtt_size(struct drm_device *dev, uint32_t size, int tiling_mode)
{
uint32_t gtt_size;
* Return the required GTT alignment for an object, taking into account
* potential fence register mapping.
*/
-static uint32_t
-i915_gem_get_gtt_alignment(struct drm_device *dev,
- uint32_t size,
- int tiling_mode)
+uint32_t
+i915_gem_get_gtt_alignment(struct drm_device *dev, uint32_t size,
+ int tiling_mode, bool fenced)
{
/*
* Minimum alignment is 4k (GTT page size), but might be greater
* if a fence register is needed for the object.
*/
- if (INTEL_INFO(dev)->gen >= 4 ||
+ if (INTEL_INFO(dev)->gen >= 4 || (!fenced && IS_G33(dev)) ||
tiling_mode == I915_TILING_NONE)
return 4096;
return i915_gem_get_gtt_size(dev, size, tiling_mode);
}
-/**
- * i915_gem_get_unfenced_gtt_alignment - return required GTT alignment for an
- * unfenced object
- * @dev: the device
- * @size: size of the object
- * @tiling_mode: tiling mode of the object
- *
- * Return the required GTT alignment for an object, only taking into account
- * unfenced tiled surface requirements.
- */
-uint32_t
-i915_gem_get_unfenced_gtt_alignment(struct drm_device *dev,
- uint32_t size,
- int tiling_mode)
-{
- /*
- * Minimum alignment is 4k (GTT page size) for sane hw.
- */
- if (INTEL_INFO(dev)->gen >= 4 || IS_G33(dev) ||
- tiling_mode == I915_TILING_NONE)
- return 4096;
-
- /* Previous hardware however needs to be aligned to a power-of-two
- * tile height. The simplest method for determining this is to reuse
- * the power-of-tile object size.
- */
- return i915_gem_get_gtt_size(dev, size, tiling_mode);
-}
-
static int i915_gem_object_create_mmap_offset(struct drm_i915_gem_object *obj)
{
struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
if (obj->base.map_list.map)
return 0;
+ dev_priv->mm.shrinker_no_lock_stealing = true;
+
ret = drm_gem_create_mmap_offset(&obj->base);
if (ret != -ENOSPC)
- return ret;
+ goto out;
/* Badly fragmented mmap space? The only way we can recover
* space is by destroying unwanted objects. We can't randomly release
i915_gem_purge(dev_priv, obj->base.size >> PAGE_SHIFT);
ret = drm_gem_create_mmap_offset(&obj->base);
if (ret != -ENOSPC)
- return ret;
+ goto out;
i915_gem_shrink_all(dev_priv);
- return drm_gem_create_mmap_offset(&obj->base);
+ ret = drm_gem_create_mmap_offset(&obj->base);
+out:
+ dev_priv->mm.shrinker_no_lock_stealing = false;
+
+ return ret;
}
static void i915_gem_object_free_mmap_offset(struct drm_i915_gem_object *obj)
goto unlock;
}
- if (obj->base.size > dev_priv->mm.gtt_mappable_end) {
+ if (obj->base.size > dev_priv->gtt.mappable_end) {
ret = -E2BIG;
goto out;
}
kfree(obj->pages);
}
-static int
+int
i915_gem_object_put_pages(struct drm_i915_gem_object *obj)
{
const struct drm_i915_gem_object_ops *ops = obj->ops;
if (obj->pages_pin_count)
return -EBUSY;
+ /* ->put_pages might need to allocate memory for the bit17 swizzle
+ * array, hence protect them from being reaped by removing them from gtt
+ * lists early. */
+ list_del(&obj->gtt_list);
+
ops->put_pages(obj);
obj->pages = NULL;
- list_del(&obj->gtt_list);
if (i915_gem_object_is_purgeable(obj))
i915_gem_object_truncate(obj);
*/
mapping = obj->base.filp->f_path.dentry->d_inode->i_mapping;
gfp = mapping_gfp_mask(mapping);
- gfp |= __GFP_NORETRY | __GFP_NOWARN;
+ gfp |= __GFP_NORETRY | __GFP_NOWARN | __GFP_NO_KSWAPD;
gfp &= ~(__GFP_IO | __GFP_WAIT);
for_each_sg(st->sgl, sg, page_count, i) {
page = shmem_read_mapping_page_gfp(mapping, i, gfp);
* our own buffer, now let the real VM do its job and
* go down in flames if truly OOM.
*/
- gfp &= ~(__GFP_NORETRY | __GFP_NOWARN);
+ gfp &= ~(__GFP_NORETRY | __GFP_NOWARN | __GFP_NO_KSWAPD);
gfp |= __GFP_IO | __GFP_WAIT;
i915_gem_shrink_all(dev_priv);
if (IS_ERR(page))
goto err_pages;
- gfp |= __GFP_NORETRY | __GFP_NOWARN;
+ gfp |= __GFP_NORETRY | __GFP_NOWARN | __GFP_NO_KSWAPD;
gfp &= ~(__GFP_IO | __GFP_WAIT);
}
if (obj->pages)
return 0;
+ if (obj->madv != I915_MADV_WILLNEED) {
+ DRM_ERROR("Attempting to obtain a purgeable object\n");
+ return -EINVAL;
+ }
+
BUG_ON(obj->pages_pin_count);
ret = ops->get_pages(obj);
if (!dev_priv->mm.suspended) {
if (i915_enable_hangcheck) {
- mod_timer(&dev_priv->hangcheck_timer,
+ mod_timer(&dev_priv->gpu_error.hangcheck_timer,
round_jiffies_up(jiffies + DRM_I915_HANGCHECK_JIFFIES));
}
if (was_empty) {
i915_gem_object_unbind(struct drm_i915_gem_object *obj)
{
drm_i915_private_t *dev_priv = obj->base.dev->dev_private;
- int ret = 0;
+ int ret;
if (obj->gtt_space == NULL)
return 0;
return 0;
}
-static void sandybridge_write_fence_reg(struct drm_device *dev, int reg,
- struct drm_i915_gem_object *obj)
-{
- drm_i915_private_t *dev_priv = dev->dev_private;
- uint64_t val;
-
- if (obj) {
- u32 size = obj->gtt_space->size;
-
- val = (uint64_t)((obj->gtt_offset + size - 4096) &
- 0xfffff000) << 32;
- val |= obj->gtt_offset & 0xfffff000;
- val |= (uint64_t)((obj->stride / 128) - 1) <<
- SANDYBRIDGE_FENCE_PITCH_SHIFT;
-
- if (obj->tiling_mode == I915_TILING_Y)
- val |= 1 << I965_FENCE_TILING_Y_SHIFT;
- val |= I965_FENCE_REG_VALID;
- } else
- val = 0;
-
- I915_WRITE64(FENCE_REG_SANDYBRIDGE_0 + reg * 8, val);
- POSTING_READ(FENCE_REG_SANDYBRIDGE_0 + reg * 8);
-}
-
static void i965_write_fence_reg(struct drm_device *dev, int reg,
struct drm_i915_gem_object *obj)
{
drm_i915_private_t *dev_priv = dev->dev_private;
+ int fence_reg;
+ int fence_pitch_shift;
uint64_t val;
+ if (INTEL_INFO(dev)->gen >= 6) {
+ fence_reg = FENCE_REG_SANDYBRIDGE_0;
+ fence_pitch_shift = SANDYBRIDGE_FENCE_PITCH_SHIFT;
+ } else {
+ fence_reg = FENCE_REG_965_0;
+ fence_pitch_shift = I965_FENCE_PITCH_SHIFT;
+ }
+
if (obj) {
u32 size = obj->gtt_space->size;
val = (uint64_t)((obj->gtt_offset + size - 4096) &
0xfffff000) << 32;
val |= obj->gtt_offset & 0xfffff000;
- val |= ((obj->stride / 128) - 1) << I965_FENCE_PITCH_SHIFT;
+ val |= (uint64_t)((obj->stride / 128) - 1) << fence_pitch_shift;
if (obj->tiling_mode == I915_TILING_Y)
val |= 1 << I965_FENCE_TILING_Y_SHIFT;
val |= I965_FENCE_REG_VALID;
} else
val = 0;
- I915_WRITE64(FENCE_REG_965_0 + reg * 8, val);
- POSTING_READ(FENCE_REG_965_0 + reg * 8);
+ fence_reg += reg * 8;
+ I915_WRITE64(fence_reg, val);
+ POSTING_READ(fence_reg);
}
static void i915_write_fence_reg(struct drm_device *dev, int reg,
{
switch (INTEL_INFO(dev)->gen) {
case 7:
- case 6: sandybridge_write_fence_reg(dev, reg, obj); break;
+ case 6:
case 5:
case 4: i965_write_fence_reg(dev, reg, obj); break;
case 3: i915_write_fence_reg(dev, reg, obj); break;
{
struct drm_device *dev = obj->base.dev;
drm_i915_private_t *dev_priv = dev->dev_private;
- struct drm_mm_node *free_space;
+ struct drm_mm_node *node;
u32 size, fence_size, fence_alignment, unfenced_alignment;
bool mappable, fenceable;
int ret;
- if (obj->madv != I915_MADV_WILLNEED) {
- DRM_ERROR("Attempting to bind a purgeable object\n");
- return -EINVAL;
- }
-
fence_size = i915_gem_get_gtt_size(dev,
obj->base.size,
obj->tiling_mode);
fence_alignment = i915_gem_get_gtt_alignment(dev,
obj->base.size,
- obj->tiling_mode);
+ obj->tiling_mode, true);
unfenced_alignment =
- i915_gem_get_unfenced_gtt_alignment(dev,
+ i915_gem_get_gtt_alignment(dev,
obj->base.size,
- obj->tiling_mode);
+ obj->tiling_mode, false);
if (alignment == 0)
alignment = map_and_fenceable ? fence_alignment :
* before evicting everything in a vain attempt to find space.
*/
if (obj->base.size >
- (map_and_fenceable ? dev_priv->mm.gtt_mappable_end : dev_priv->mm.gtt_total)) {
+ (map_and_fenceable ? dev_priv->gtt.mappable_end : dev_priv->gtt.total)) {
DRM_ERROR("Attempting to bind an object larger than the aperture\n");
return -E2BIG;
}
i915_gem_object_pin_pages(obj);
+ node = kzalloc(sizeof(*node), GFP_KERNEL);
+ if (node == NULL) {
+ i915_gem_object_unpin_pages(obj);
+ return -ENOMEM;
+ }
+
search_free:
if (map_and_fenceable)
- free_space = drm_mm_search_free_in_range_color(&dev_priv->mm.gtt_space,
- size, alignment, obj->cache_level,
- 0, dev_priv->mm.gtt_mappable_end,
- false);
+ ret = drm_mm_insert_node_in_range_generic(&dev_priv->mm.gtt_space, node,
+ size, alignment, obj->cache_level,
+ 0, dev_priv->gtt.mappable_end);
else
- free_space = drm_mm_search_free_color(&dev_priv->mm.gtt_space,
- size, alignment, obj->cache_level,
- false);
-
- if (free_space != NULL) {
- if (map_and_fenceable)
- free_space =
- drm_mm_get_block_range_generic(free_space,
- size, alignment, obj->cache_level,
- 0, dev_priv->mm.gtt_mappable_end,
- false);
- else
- free_space =
- drm_mm_get_block_generic(free_space,
- size, alignment, obj->cache_level,
- false);
- }
- if (free_space == NULL) {
+ ret = drm_mm_insert_node_generic(&dev_priv->mm.gtt_space, node,
+ size, alignment, obj->cache_level);
+ if (ret) {
ret = i915_gem_evict_something(dev, size, alignment,
obj->cache_level,
map_and_fenceable,
nonblocking);
- if (ret) {
- i915_gem_object_unpin_pages(obj);
- return ret;
- }
+ if (ret == 0)
+ goto search_free;
- goto search_free;
+ i915_gem_object_unpin_pages(obj);
+ kfree(node);
+ return ret;
}
- if (WARN_ON(!i915_gem_valid_gtt_space(dev,
- free_space,
- obj->cache_level))) {
+ if (WARN_ON(!i915_gem_valid_gtt_space(dev, node, obj->cache_level))) {
i915_gem_object_unpin_pages(obj);
- drm_mm_put_block(free_space);
+ drm_mm_put_block(node);
return -EINVAL;
}
ret = i915_gem_gtt_prepare_object(obj);
if (ret) {
i915_gem_object_unpin_pages(obj);
- drm_mm_put_block(free_space);
+ drm_mm_put_block(node);
return ret;
}
list_move_tail(&obj->gtt_list, &dev_priv->mm.bound_list);
list_add_tail(&obj->mm_list, &dev_priv->mm.inactive_list);
- obj->gtt_space = free_space;
- obj->gtt_offset = free_space->start;
+ obj->gtt_space = node;
+ obj->gtt_offset = node->start;
fenceable =
- free_space->size == fence_size &&
- (free_space->start & (fence_alignment - 1)) == 0;
+ node->size == fence_size &&
+ (node->start & (fence_alignment - 1)) == 0;
mappable =
- obj->gtt_offset + obj->base.size <= dev_priv->mm.gtt_mappable_end;
+ obj->gtt_offset + obj->base.size <= dev_priv->gtt.mappable_end;
obj->map_and_fenceable = mappable && fenceable;
u32 seqno = 0;
int ret;
- if (atomic_read(&dev_priv->mm.wedged))
+ if (atomic_read(&dev_priv->gpu_error.wedged))
return -EIO;
spin_lock(&file_priv->mm.lock);
* And not confound mm.suspended!
*/
dev_priv->mm.suspended = 1;
- del_timer_sync(&dev_priv->hangcheck_timer);
+ del_timer_sync(&dev_priv->gpu_error.hangcheck_timer);
i915_kernel_lost_context(dev);
i915_gem_cleanup_ringbuffer(dev);
return ret;
}
-static bool
-intel_enable_ppgtt(struct drm_device *dev)
-{
- if (i915_enable_ppgtt >= 0)
- return i915_enable_ppgtt;
-
-#ifdef CONFIG_INTEL_IOMMU
- /* Disable ppgtt on SNB if VT-d is on. */
- if (INTEL_INFO(dev)->gen == 6 && intel_iommu_gfx_mapped)
- return false;
-#endif
-
- return true;
-}
-
int i915_gem_init(struct drm_device *dev)
{
struct drm_i915_private *dev_priv = dev->dev_private;
- unsigned long gtt_size, mappable_size;
int ret;
- gtt_size = dev_priv->mm.gtt->gtt_total_entries << PAGE_SHIFT;
- mappable_size = dev_priv->mm.gtt->gtt_mappable_entries << PAGE_SHIFT;
-
mutex_lock(&dev->struct_mutex);
- if (intel_enable_ppgtt(dev) && HAS_ALIASING_PPGTT(dev)) {
- /* PPGTT pdes are stolen from global gtt ptes, so shrink the
- * aperture accordingly when using aliasing ppgtt. */
- gtt_size -= I915_PPGTT_PD_ENTRIES*PAGE_SIZE;
-
- i915_gem_init_global_gtt(dev, 0, mappable_size, gtt_size);
-
- ret = i915_gem_init_aliasing_ppgtt(dev);
- if (ret) {
- mutex_unlock(&dev->struct_mutex);
- return ret;
- }
- } else {
- /* Let GEM Manage all of the aperture.
- *
- * However, leave one page at the end still bound to the scratch
- * page. There are a number of places where the hardware
- * apparently prefetches past the end of the object, and we've
- * seen multiple hangs with the GPU head pointer stuck in a
- * batchbuffer bound at the last page of the aperture. One page
- * should be enough to keep any prefetching inside of the
- * aperture.
- */
- i915_gem_init_global_gtt(dev, 0, mappable_size,
- gtt_size);
- }
-
+ i915_gem_init_global_gtt(dev);
ret = i915_gem_init_hw(dev);
mutex_unlock(&dev->struct_mutex);
if (ret) {
if (drm_core_check_feature(dev, DRIVER_MODESET))
return 0;
- if (atomic_read(&dev_priv->mm.wedged)) {
+ if (atomic_read(&dev_priv->gpu_error.wedged)) {
DRM_ERROR("Reenabling wedged hardware, good luck\n");
- atomic_set(&dev_priv->mm.wedged, 0);
+ atomic_set(&dev_priv->gpu_error.wedged, 0);
}
mutex_lock(&dev->struct_mutex);
INIT_LIST_HEAD(&dev_priv->fence_regs[i].lru_list);
INIT_DELAYED_WORK(&dev_priv->mm.retire_work,
i915_gem_retire_work_handler);
- init_completion(&dev_priv->error_completion);
+ init_completion(&dev_priv->gpu_error.completion);
/* On GEN3 we really need to make sure the ARB C3 LP bit is set */
if (IS_GEN3(dev)) {
if (!mutex_is_locked_by(&dev->struct_mutex, current))
return 0;
+ if (dev_priv->mm.shrinker_no_lock_stealing)
+ return 0;
+
unlock = false;
}