summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/i915/gem/i915_gem_domain.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/i915/gem/i915_gem_domain.c')
-rw-r--r--drivers/gpu/drm/i915/gem/i915_gem_domain.c59
1 files changed, 53 insertions, 6 deletions
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
index 3d435bfff764..36f54cedaaeb 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_domain.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
@@ -5,6 +5,7 @@
*/
#include "display/intel_frontbuffer.h"
+#include "gt/intel_gt.h"
#include "i915_drv.h"
#include "i915_gem_clflush.h"
@@ -15,13 +16,58 @@
#include "i915_gem_lmem.h"
#include "i915_gem_mman.h"
+static bool gpu_write_needs_clflush(struct drm_i915_gem_object *obj)
+{
+ return !(obj->cache_level == I915_CACHE_NONE ||
+ obj->cache_level == I915_CACHE_WT);
+}
+
+static void
+flush_write_domain(struct drm_i915_gem_object *obj, unsigned int flush_domains)
+{
+ struct i915_vma *vma;
+
+ assert_object_held(obj);
+
+ if (!(obj->write_domain & flush_domains))
+ return;
+
+ switch (obj->write_domain) {
+ case I915_GEM_DOMAIN_GTT:
+ spin_lock(&obj->vma.lock);
+ for_each_ggtt_vma(vma, obj) {
+ if (i915_vma_unset_ggtt_write(vma))
+ intel_gt_flush_ggtt_writes(vma->vm->gt);
+ }
+ spin_unlock(&obj->vma.lock);
+
+ i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
+ break;
+
+ case I915_GEM_DOMAIN_WC:
+ wmb();
+ break;
+
+ case I915_GEM_DOMAIN_CPU:
+ i915_gem_clflush_object(obj, I915_CLFLUSH_SYNC);
+ break;
+
+ case I915_GEM_DOMAIN_RENDER:
+ if (gpu_write_needs_clflush(obj))
+ obj->cache_dirty = true;
+ break;
+ }
+
+ obj->write_domain = 0;
+}
+
static void __i915_gem_object_flush_for_display(struct drm_i915_gem_object *obj)
{
/*
* We manually flush the CPU domain so that we can override and
* force the flush for the display, and perform it asyncrhonously.
*/
- i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+ flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
if (obj->cache_dirty)
i915_gem_clflush_object(obj, I915_CLFLUSH_FORCE);
obj->write_domain = 0;
@@ -80,7 +126,7 @@ i915_gem_object_set_to_wc_domain(struct drm_i915_gem_object *obj, bool write)
if (ret)
return ret;
- i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_WC);
+ flush_write_domain(obj, ~I915_GEM_DOMAIN_WC);
/* Serialise direct access to this object with the barriers for
* coherent writes from the GPU, by effectively invalidating the
@@ -141,7 +187,7 @@ i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write)
if (ret)
return ret;
- i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_GTT);
+ flush_write_domain(obj, ~I915_GEM_DOMAIN_GTT);
/* Serialise direct access to this object with the barriers for
* coherent writes from the GPU, by effectively invalidating the
@@ -370,6 +416,7 @@ retry:
}
vma->display_alignment = max_t(u64, vma->display_alignment, alignment);
+ i915_vma_mark_scanout(vma);
i915_gem_object_flush_if_display_locked(obj);
@@ -409,7 +456,7 @@ i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write)
if (ret)
return ret;
- i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+ flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
/* Flush the CPU cache if it's still invalid. */
if ((obj->read_domains & I915_GEM_DOMAIN_CPU) == 0) {
@@ -574,7 +621,7 @@ int i915_gem_object_prepare_read(struct drm_i915_gem_object *obj,
goto out;
}
- i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+ flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
/* If we're not in the cpu read domain, set ourself into the gtt
* read domain and manually flush cachelines (if required). This
@@ -625,7 +672,7 @@ int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj,
goto out;
}
- i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+ flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
/* If we're not in the cpu write domain, set ourself into the
* gtt write domain and manually flush cachelines (as required).