By moving map-and-fenceable tracking from the object to the VMA, we gain
fine-grained tracking and the ability to track individual fences on the VMA
(subsequent patch).
Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
---
drivers/gpu/drm/i915/i915_debugfs.c | 46 +++++++++++++-----------------
drivers/gpu/drm/i915/i915_drv.h | 6 ----
drivers/gpu/drm/i915/i915_gem.c | 31 +++++++++-----------
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 3 +-
drivers/gpu/drm/i915/i915_gem_fence.c | 5 +---
drivers/gpu/drm/i915/i915_gem_gtt.h | 6 ++++
drivers/gpu/drm/i915/i915_gem_tiling.c | 4 +--
drivers/gpu/drm/i915/intel_display.c | 6 ++--
8 files changed, 46 insertions(+), 61 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index e923dc192f54..418b80de5246 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -112,19 +112,6 @@ static inline const char *get_global_flag(struct drm_i915_gem_object *obj)
return i915_gem_object_to_ggtt(obj, NULL) ? "g" : " ";
}
-static u64 i915_gem_obj_total_ggtt_size(struct drm_i915_gem_object *obj)
-{
- u64 size = 0;
- struct i915_vma *vma;
-
- list_for_each_entry(vma, &obj->vma_list, obj_link) {
- if (vma->is_ggtt && drm_mm_node_allocated(&vma->node))
- size += vma->node.size;
- }
-
- return size;
-}
-
static void
describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
{
@@ -309,17 +296,6 @@ static int i915_gem_stolen_list_info(struct seq_file *m, void *data)
return 0;
}
-#define count_objects(list, member) do { \
- list_for_each_entry(obj, list, member) { \
- size += i915_gem_obj_total_ggtt_size(obj); \
- ++count; \
- if (obj->map_and_fenceable) { \
- mappable_size += obj->base.size; \
- ++mappable_count; \
- } \
- } \
-} while (0)
-
struct file_stats {
struct drm_i915_file_private *file_priv;
unsigned long count;
@@ -404,7 +380,7 @@ static void print_batch_pool_stats(struct seq_file *m,
list_for_each_entry(vma, list, member) { \
size += vma->size; \
++count; \
- if (vma->obj->map_and_fenceable) { \
+ if (vma->map_and_fenceable) { \
mappable_size += vma->size; \
++mappable_count; \
} \
@@ -433,7 +409,25 @@ static int i915_gem_object_info(struct seq_file *m, void* data)
dev_priv->mm.object_memory);
size = count = mappable_size = mappable_count = 0;
- count_objects(&dev_priv->mm.bound_list, global_list);
+ list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) {
+ bool allocated = false, mappable = false;
+
+ list_for_each_entry(vma, &obj->vma_list, obj_link) {
+ if (!vma->is_ggtt)
+ continue;
+
+ allocated = true;
+ size += vma->node.size;
+
+ if (vma->map_and_fenceable) {
+ mappable = true;
+ mappable_size += vma->node.size;
+ }
+ }
+
+ count += allocated;
+ mappable_count += mappable;
+ }
seq_printf(m, "%u [%u] objects, %llu [%llu] bytes in gtt\n",
count, mappable_count, size, mappable_size);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 45a0da0947cd..cfc4430d3b50 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2078,12 +2078,6 @@ struct drm_i915_gem_object {
unsigned int fence_dirty:1;
/**
- * Is the object at the current location in the gtt mappable and
- * fenceable? Used to avoid costly recalculations.
- */
- unsigned int map_and_fenceable:1;
-
- /**
* Whether the current gtt mapping needs to be mappable (and isn't just
* mappable by accident). Track pin and fault separate for a more
* accurate mappable working set.
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index d81821c6f9a1..0c4e8e1aeeff 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2691,13 +2691,15 @@ int i915_vma_unbind(struct i915_vma *vma)
GEM_BUG_ON(obj->bind_count == 0);
GEM_BUG_ON(obj->pages == NULL);
- if (vma->is_ggtt && vma->ggtt_view.type == I915_GGTT_VIEW_NORMAL) {
+ if (vma->map_and_fenceable) {
i915_gem_object_finish_gtt(obj);
/* release the fence reg _after_ flushing */
ret = i915_gem_object_put_fence(obj);
if (ret)
return ret;
+
+ vma->map_and_fenceable = false;
}
if (likely(!vma->vm->closed)) {
@@ -2709,10 +2711,8 @@ int i915_vma_unbind(struct i915_vma *vma)
drm_mm_remove_node(&vma->node);
list_move_tail(&vma->vm_link, &vma->vm->unbound_list);
- if (vma->is_ggtt) {
- if (vma->ggtt_view.type == I915_GGTT_VIEW_NORMAL) {
- obj->map_and_fenceable = false;
- } else if (vma->ggtt_view.pages) {
+ if (vma->ggtt_view.pages) {
+ if (vma->ggtt_view.type != I915_GGTT_VIEW_NORMAL) {
sg_free_table(vma->ggtt_view.pages);
kfree(vma->ggtt_view.pages);
}
@@ -3480,8 +3480,6 @@ i915_vma_misplaced(struct i915_vma *vma,
uint64_t alignment,
uint64_t flags)
{
- struct drm_i915_gem_object *obj = vma->obj;
-
if (!drm_mm_node_allocated(&vma->node))
return false;
@@ -3491,7 +3489,7 @@ i915_vma_misplaced(struct i915_vma *vma,
if (alignment && vma->node.start & (alignment - 1))
return true;
- if (flags & PIN_MAPPABLE && !obj->map_and_fenceable)
+ if (flags & PIN_MAPPABLE && !vma->map_and_fenceable)
return true;
if (flags & PIN_OFFSET_BIAS &&
@@ -3511,13 +3509,10 @@ void __i915_vma_set_map_and_fenceable(struct i915_vma *vma)
bool mappable, fenceable;
u32 fence_size, fence_alignment;
- fence_size = i915_gem_get_gtt_size(obj->base.dev,
- obj->base.size,
+ fence_size = i915_gem_get_gtt_size(obj->base.dev, vma->size,
obj->tiling_mode);
- fence_alignment = i915_gem_get_gtt_alignment(obj->base.dev,
- obj->base.size,
- obj->tiling_mode,
- true);
+ fence_alignment = i915_gem_get_gtt_alignment(obj->base.dev, vma->size,
+ obj->tiling_mode, true);
fenceable = (vma->node.size == fence_size &&
(vma->node.start & (fence_alignment - 1)) == 0);
@@ -3525,7 +3520,7 @@ void __i915_vma_set_map_and_fenceable(struct i915_vma *vma)
mappable = (vma->node.start + fence_size <=
to_i915(obj->base.dev)->gtt.mappable_end);
- obj->map_and_fenceable = mappable && fenceable;
+ vma->map_and_fenceable = mappable && fenceable;
}
int
@@ -3593,13 +3588,13 @@ i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj,
WARN(vma->pin_count,
"bo is already pinned in ggtt with incorrect alignment:"
- " offset=%08x %08x, req.alignment=%llx, req.map_and_fenceable=%d,"
- " obj->map_and_fenceable=%d\n",
+ " offset=%08x %08x, req.alignment=%llx,"
+ " req.map_and_fenceable=%d, vma->map_and_fenceable=%d\n",
upper_32_bits(vma->node.start),
lower_32_bits(vma->node.start),
(long long)alignment,
!!(flags & PIN_MAPPABLE),
- obj->map_and_fenceable);
+ vma->map_and_fenceable);
ret = i915_vma_unbind(vma);
if (ret)
return ERR_PTR(ret);
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 569be409c049..d13b7e507b3d 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -792,7 +792,6 @@ static bool
eb_vma_misplaced(struct i915_vma *vma)
{
struct drm_i915_gem_exec_object2 *entry = vma->exec_entry;
- struct drm_i915_gem_object *obj = vma->obj;
WARN_ON(entry->flags & __EXEC_OBJECT_NEEDS_MAP && !vma->is_ggtt);
@@ -812,7 +811,7 @@ eb_vma_misplaced(struct i915_vma *vma)
return true;
/* avoid costly ping-pong once a batch bo ended up non-mappable */
- if (entry->flags & __EXEC_OBJECT_NEEDS_MAP && !obj->map_and_fenceable)
+ if (entry->flags & __EXEC_OBJECT_NEEDS_MAP && !vma->map_and_fenceable)
return !only_mappable_for_reloc(entry->flags);
if ((entry->flags & EXEC_OBJECT_SUPPORTS_48B_ADDRESS) == 0 &&
diff --git a/drivers/gpu/drm/i915/i915_gem_fence.c b/drivers/gpu/drm/i915/i915_gem_fence.c
index 8ba05a0f15d2..e0f5fba22931 100644
--- a/drivers/gpu/drm/i915/i915_gem_fence.c
+++ b/drivers/gpu/drm/i915/i915_gem_fence.c
@@ -124,7 +124,7 @@ static void i915_write_fence_reg(struct drm_device *dev, int reg,
(vma->node.start & (vma->node.size - 1)),
"object 0x%08lx [fenceable? %d] not 1M or pot-size (0x%08lx) aligned\n",
(long)vma->node.start,
- obj->map_and_fenceable,
+ vma->map_and_fenceable,
(long)vma->node.size);
if (obj->tiling_mode == I915_TILING_Y && HAS_128_BYTE_Y_TILING(dev))
@@ -378,9 +378,6 @@ i915_gem_object_get_fence(struct drm_i915_gem_object *obj)
return 0;
}
} else if (enable) {
- if (WARN_ON(!obj->map_and_fenceable))
- return -EINVAL;
-
reg = i915_find_fence_reg(dev);
if (IS_ERR(reg))
return PTR_ERR(reg);
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 7f57dea246d8..6b0f557982d5 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -205,6 +205,12 @@ struct i915_vma {
unsigned int bound : 4;
unsigned int active : I915_NUM_RINGS;
bool is_ggtt : 1;
+ /**
+ * Is the vma/object at the current location in the gtt
+ * mappable and fenceable? Used to avoid costly
+ * recalculations.
+ */
+ bool map_and_fenceable : 1;
bool closed : 1;
};
unsigned int flags;
diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c
index f83cb4329c8d..7c2da8060757 100644
--- a/drivers/gpu/drm/i915/i915_gem_tiling.c
+++ b/drivers/gpu/drm/i915/i915_gem_tiling.c
@@ -130,7 +130,7 @@ i915_gem_object_fence_ok(struct drm_i915_gem_object *obj, int tiling_mode)
if (vma == NULL)
return 0;
- if (!obj->map_and_fenceable)
+ if (!vma->map_and_fenceable)
return 0;
if (INTEL_INFO(obj->base.dev)->gen == 3) {
@@ -141,7 +141,7 @@ i915_gem_object_fence_ok(struct drm_i915_gem_object *obj, int tiling_mode)
goto bad;
}
- size = i915_gem_get_gtt_size(obj->base.dev, obj->base.size, tiling_mode);
+ size = i915_gem_get_gtt_size(obj->base.dev, vma->size, tiling_mode);
if (vma->node.size < size)
goto bad;
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 313f1fb144b9..218bfd3c99fc 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -2391,7 +2391,7 @@ intel_pin_and_fence_fb_obj(struct drm_plane *plane,
* framebuffer compression. For simplicity, we always install
* a fence as the cost is not that onerous.
*/
- if (view.type == I915_GGTT_VIEW_NORMAL) {
+ if (vma->map_and_fenceable) {
ret = i915_gem_object_get_fence(obj);
if (ret == -EDEADLK) {
/*
@@ -2430,11 +2430,11 @@ static void intel_unpin_fb_obj(struct drm_framebuffer *fb,
WARN_ON(!mutex_is_locked(&obj->base.dev->struct_mutex));
intel_fill_fb_ggtt_view(&view, fb, state);
+ vma = i915_gem_object_to_ggtt(obj, &view);
- if (view.type == I915_GGTT_VIEW_NORMAL)
+ if (vma->map_and_fenceable)
i915_gem_object_unpin_fence(obj);
- vma = i915_gem_object_to_ggtt(obj, &view);
i915_gem_object_unpin_from_display_plane(vma);
}
--
2.7.0.rc3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
