obj_priv->dirty = 0;
for (i = 0; i < page_count; i++) {
- if (obj_priv->pages[i] == NULL)
- break;
-
if (obj_priv->dirty)
set_page_dirty(obj_priv->pages[i]);
else
list_move_tail(&obj_priv->list, &dev_priv->mm.inactive_list);
+ BUG_ON(!list_empty(&obj_priv->gpu_write_list));
+
obj_priv->last_rendering_seqno = 0;
if (obj_priv->active) {
obj_priv->active = 0;
i915_verify_inactive(dev, __FILE__, __LINE__);
}
+#define PIPE_CONTROL_FLUSH(addr) \
+ OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE | \
+ PIPE_CONTROL_DEPTH_STALL); \
+ OUT_RING(addr | PIPE_CONTROL_GLOBAL_GTT); \
+ OUT_RING(0); \
+ OUT_RING(0); \
+
/**
* Creates a new sequence number, emitting a write of it to the status page
* plus an interrupt, which will trigger i915_user_interrupt_handler.
if (dev_priv->mm.next_gem_seqno == 0)
dev_priv->mm.next_gem_seqno++;
- BEGIN_LP_RING(4);
- OUT_RING(MI_STORE_DWORD_INDEX);
- OUT_RING(I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
- OUT_RING(seqno);
+ if (HAS_PIPE_CONTROL(dev)) {
+ u32 scratch_addr = dev_priv->seqno_gfx_addr + 128;
- OUT_RING(MI_USER_INTERRUPT);
- ADVANCE_LP_RING();
+ /*
+ * Workaround qword write incoherence by flushing the
+ * PIPE_NOTIFY buffers out to memory before requesting
+ * an interrupt.
+ */
+ BEGIN_LP_RING(32);
+ OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE |
+ PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_TC_FLUSH);
+ OUT_RING(dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT);
+ OUT_RING(seqno);
+ OUT_RING(0);
+ PIPE_CONTROL_FLUSH(scratch_addr);
+ scratch_addr += 128; /* write to separate cachelines */
+ PIPE_CONTROL_FLUSH(scratch_addr);
+ scratch_addr += 128;
+ PIPE_CONTROL_FLUSH(scratch_addr);
+ scratch_addr += 128;
+ PIPE_CONTROL_FLUSH(scratch_addr);
+ scratch_addr += 128;
+ PIPE_CONTROL_FLUSH(scratch_addr);
+ scratch_addr += 128;
+ PIPE_CONTROL_FLUSH(scratch_addr);
+ OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE |
+ PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_TC_FLUSH |
+ PIPE_CONTROL_NOTIFY);
+ OUT_RING(dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT);
+ OUT_RING(seqno);
+ OUT_RING(0);
+ ADVANCE_LP_RING();
+ } else {
+ BEGIN_LP_RING(4);
+ OUT_RING(MI_STORE_DWORD_INDEX);
+ OUT_RING(I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
+ OUT_RING(seqno);
+
+ OUT_RING(MI_USER_INTERRUPT);
+ ADVANCE_LP_RING();
+ }
DRM_DEBUG_DRIVER("%d\n", seqno);
struct drm_i915_gem_object *obj_priv, *next;
list_for_each_entry_safe(obj_priv, next,
- &dev_priv->mm.flushing_list, list) {
+ &dev_priv->mm.gpu_write_list,
+ gpu_write_list) {
struct drm_gem_object *obj = obj_priv->obj;
if ((obj->write_domain & flush_domains) ==
uint32_t old_write_domain = obj->write_domain;
obj->write_domain = 0;
+ list_del_init(&obj_priv->gpu_write_list);
i915_gem_object_move_to_active(obj, seqno);
trace_i915_gem_object_change_domain(obj,
{
drm_i915_private_t *dev_priv = dev->dev_private;
- return READ_HWSP(dev_priv, I915_GEM_HWS_INDEX);
+ if (HAS_PIPE_CONTROL(dev))
+ return ((volatile u32 *)(dev_priv->seqno_page))[0];
+ else
+ return READ_HWSP(dev_priv, I915_GEM_HWS_INDEX);
}
/**
return -EIO;
if (!i915_seqno_passed(i915_get_gem_seqno(dev), seqno)) {
- if (IS_IRONLAKE(dev))
+ if (HAS_PCH_SPLIT(dev))
ier = I915_READ(DEIER) | I915_READ(GTIER);
else
ier = I915_READ(IER);
i915_gem_evict_everything(struct drm_device *dev)
{
drm_i915_private_t *dev_priv = dev->dev_private;
- uint32_t seqno;
int ret;
+ uint32_t seqno;
bool lists_empty;
spin_lock(&dev_priv->mm.active_list_lock);
if (ret)
return ret;
+ BUG_ON(!list_empty(&dev_priv->mm.flushing_list));
+
ret = i915_gem_evict_from_inactive_list(dev);
if (ret)
return ret;
struct address_space *mapping;
struct inode *inode;
struct page *page;
- int ret;
if (obj_priv->pages_refcount++ != 0)
return 0;
mapping_gfp_mask (mapping) |
__GFP_COLD |
gfpmask);
- if (IS_ERR(page)) {
- ret = PTR_ERR(page);
- i915_gem_object_put_pages(obj);
- return ret;
- }
+ if (IS_ERR(page))
+ goto err_pages;
+
obj_priv->pages[i] = page;
}
i915_gem_object_do_bit_17_swizzle(obj);
return 0;
+
+err_pages:
+ while (i--)
+ page_cache_release(obj_priv->pages[i]);
+
+ drm_free_large(obj_priv->pages);
+ obj_priv->pages = NULL;
+ obj_priv->pages_refcount--;
+ return PTR_ERR(page);
}
static void i965_write_fence_reg(struct drm_i915_fence_reg *reg)
pitch_val = obj_priv->stride / tile_width;
pitch_val = ffs(pitch_val) - 1;
+ if (obj_priv->tiling_mode == I915_TILING_Y &&
+ HAS_128_BYTE_Y_TILING(dev))
+ WARN_ON(pitch_val > I830_FENCE_MAX_PITCH_VAL);
+ else
+ WARN_ON(pitch_val > I915_FENCE_MAX_PITCH_VAL);
+
val = obj_priv->gtt_offset;
if (obj_priv->tiling_mode == I915_TILING_Y)
val |= 1 << I830_FENCE_TILING_Y_SHIFT;
return -EINVAL;
}
+ /* If the object is bigger than the entire aperture, reject it early
+ * before evicting everything in a vain attempt to find space.
+ */
+ if (obj->size > dev->gtt_total) {
+ DRM_ERROR("Attempting to bind an object larger than the aperture\n");
+ return -E2BIG;
+ }
+
search_free:
free_space = drm_mm_search_free(&dev_priv->mm.gtt_space,
obj->size, alignment, 0);
old_write_domain = obj->write_domain;
i915_gem_flush(dev, 0, obj->write_domain);
seqno = i915_add_request(dev, NULL, obj->write_domain);
- obj->write_domain = 0;
+ BUG_ON(obj->write_domain);
i915_gem_object_move_to_active(obj, seqno);
trace_i915_gem_object_change_domain(obj,
uint32_t reloc_count = 0, i;
int ret = 0;
+ if (relocs == NULL)
+ return 0;
+
for (i = 0; i < buffer_count; i++) {
struct drm_i915_gem_relocation_entry __user *user_relocs;
int unwritten;
struct drm_gem_object *batch_obj;
struct drm_i915_gem_object *obj_priv;
struct drm_clip_rect *cliprects = NULL;
- struct drm_i915_gem_relocation_entry *relocs;
+ struct drm_i915_gem_relocation_entry *relocs = NULL;
int ret = 0, ret2, i, pinned = 0;
uint64_t exec_offset;
uint32_t seqno, flush_domains, reloc_index;
if (args->num_cliprects != 0) {
cliprects = kcalloc(args->num_cliprects, sizeof(*cliprects),
GFP_KERNEL);
- if (cliprects == NULL)
+ if (cliprects == NULL) {
+ ret = -ENOMEM;
goto pre_mutex_err;
+ }
ret = copy_from_user(cliprects,
(struct drm_clip_rect __user *)
if (object_list[i] == NULL) {
DRM_ERROR("Invalid object handle %d at index %d\n",
exec_list[i].handle, i);
+ /* prevent error path from reading uninitialized data */
+ args->buffer_count = i + 1;
ret = -EBADF;
goto err;
}
if (obj_priv->in_execbuffer) {
DRM_ERROR("Object %p appears more than once in object list\n",
object_list[i]);
+ /* prevent error path from reading uninitialized data */
+ args->buffer_count = i + 1;
ret = -EBADF;
goto err;
}
i915_gem_flush(dev,
dev->invalidate_domains,
dev->flush_domains);
- if (dev->flush_domains)
+ if (dev->flush_domains & I915_GEM_GPU_DOMAINS)
(void)i915_add_request(dev, file_priv,
dev->flush_domains);
}
for (i = 0; i < args->buffer_count; i++) {
struct drm_gem_object *obj = object_list[i];
+ struct drm_i915_gem_object *obj_priv = obj->driver_private;
uint32_t old_write_domain = obj->write_domain;
obj->write_domain = obj->pending_write_domain;
+ if (obj->write_domain)
+ list_move_tail(&obj_priv->gpu_write_list,
+ &dev_priv->mm.gpu_write_list);
+ else
+ list_del_init(&obj_priv->gpu_write_list);
+
trace_i915_gem_object_change_domain(obj,
obj->read_domains,
old_write_domain);
mutex_unlock(&dev->struct_mutex);
+pre_mutex_err:
/* Copy the updated relocations out regardless of current error
* state. Failure to update the relocs would mean that the next
* time userland calls execbuf, it would do so with presumed offset
ret = ret2;
}
-pre_mutex_err:
drm_free_large(object_list);
kfree(cliprects);
obj_priv->obj = obj;
obj_priv->fence_reg = I915_FENCE_REG_NONE;
INIT_LIST_HEAD(&obj_priv->list);
+ INIT_LIST_HEAD(&obj_priv->gpu_write_list);
INIT_LIST_HEAD(&obj_priv->fence_list);
obj_priv->madv = I915_MADV_WILLNEED;
return 0;
}
+/*
+ * 965+ support PIPE_CONTROL commands, which provide finer grained control
+ * over cache flushing.
+ */
+static int
+i915_gem_init_pipe_control(struct drm_device *dev)
+{
+ drm_i915_private_t *dev_priv = dev->dev_private;
+ struct drm_gem_object *obj;
+ struct drm_i915_gem_object *obj_priv;
+ int ret;
+
+ obj = drm_gem_object_alloc(dev, 4096);
+ if (obj == NULL) {
+ DRM_ERROR("Failed to allocate seqno page\n");
+ ret = -ENOMEM;
+ goto err;
+ }
+ obj_priv = obj->driver_private;
+ obj_priv->agp_type = AGP_USER_CACHED_MEMORY;
+
+ ret = i915_gem_object_pin(obj, 4096);
+ if (ret)
+ goto err_unref;
+
+ dev_priv->seqno_gfx_addr = obj_priv->gtt_offset;
+ dev_priv->seqno_page = kmap(obj_priv->pages[0]);
+ if (dev_priv->seqno_page == NULL)
+ goto err_unpin;
+
+ dev_priv->seqno_obj = obj;
+ memset(dev_priv->seqno_page, 0, PAGE_SIZE);
+
+ return 0;
+
+err_unpin:
+ i915_gem_object_unpin(obj);
+err_unref:
+ drm_gem_object_unreference(obj);
+err:
+ return ret;
+}
+
static int
i915_gem_init_hws(struct drm_device *dev)
{
obj = drm_gem_object_alloc(dev, 4096);
if (obj == NULL) {
DRM_ERROR("Failed to allocate status page\n");
- return -ENOMEM;
+ ret = -ENOMEM;
+ goto err;
}
obj_priv = obj->driver_private;
obj_priv->agp_type = AGP_USER_CACHED_MEMORY;
ret = i915_gem_object_pin(obj, 4096);
if (ret != 0) {
drm_gem_object_unreference(obj);
- return ret;
+ goto err_unref;
}
dev_priv->status_gfx_addr = obj_priv->gtt_offset;
if (dev_priv->hw_status_page == NULL) {
DRM_ERROR("Failed to map status page.\n");
memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map));
- i915_gem_object_unpin(obj);
- drm_gem_object_unreference(obj);
- return -EINVAL;
+ ret = -EINVAL;
+ goto err_unpin;
}
+
+ if (HAS_PIPE_CONTROL(dev)) {
+ ret = i915_gem_init_pipe_control(dev);
+ if (ret)
+ goto err_unpin;
+ }
+
dev_priv->hws_obj = obj;
memset(dev_priv->hw_status_page, 0, PAGE_SIZE);
I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr);
DRM_DEBUG_DRIVER("hws offset: 0x%08x\n", dev_priv->status_gfx_addr);
return 0;
+
+err_unpin:
+ i915_gem_object_unpin(obj);
+err_unref:
+ drm_gem_object_unreference(obj);
+err:
+ return 0;
+}
+
+static void
+i915_gem_cleanup_pipe_control(struct drm_device *dev)
+{
+ drm_i915_private_t *dev_priv = dev->dev_private;
+ struct drm_gem_object *obj;
+ struct drm_i915_gem_object *obj_priv;
+
+ obj = dev_priv->seqno_obj;
+ obj_priv = obj->driver_private;
+ kunmap(obj_priv->pages[0]);
+ i915_gem_object_unpin(obj);
+ drm_gem_object_unreference(obj);
+ dev_priv->seqno_obj = NULL;
+
+ dev_priv->seqno_page = NULL;
}
static void
memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map));
dev_priv->hw_status_page = NULL;
+ if (HAS_PIPE_CONTROL(dev))
+ i915_gem_cleanup_pipe_control(dev);
+
/* Write high address into HWS_PGA when disabling. */
I915_WRITE(HWS_PGA, 0x1ffff000);
}
spin_lock_init(&dev_priv->mm.active_list_lock);
INIT_LIST_HEAD(&dev_priv->mm.active_list);
INIT_LIST_HEAD(&dev_priv->mm.flushing_list);
+ INIT_LIST_HEAD(&dev_priv->mm.gpu_write_list);
INIT_LIST_HEAD(&dev_priv->mm.inactive_list);
INIT_LIST_HEAD(&dev_priv->mm.request_list);
INIT_LIST_HEAD(&dev_priv->mm.fence_list);