drm/i915: fix regression due to ba3d8d749b01548b9
We don't track gpu flush request in any special way. So even with
obj->write_domain == 0, a gpu flush might be outstanding but no
yet executed. Even worse, the latest request might use the object
only for reading. So and unconditional call to object_wait_rendering
is needed for !pipelined.
Hence revert that patch fully and untangle the flushing from the
synchronization again.
Reported-by: Keith Packard <keithp@keithp.com>
Tested-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 92b097d..5e54821 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -38,8 +38,7 @@
static uint32_t i915_gem_get_gtt_alignment(struct drm_gem_object *obj);
-static int i915_gem_object_flush_gpu_write_domain(struct drm_gem_object *obj,
- bool pipelined);
+static int i915_gem_object_flush_gpu_write_domain(struct drm_gem_object *obj);
static void i915_gem_object_flush_gtt_write_domain(struct drm_gem_object *obj);
static void i915_gem_object_flush_cpu_write_domain(struct drm_gem_object *obj);
static int i915_gem_object_set_to_cpu_domain(struct drm_gem_object *obj,
@@ -2594,7 +2593,7 @@
if (reg->gpu) {
int ret;
- ret = i915_gem_object_flush_gpu_write_domain(obj, true);
+ ret = i915_gem_object_flush_gpu_write_domain(obj);
if (ret)
return ret;
@@ -2742,8 +2741,7 @@
/** Flushes any GPU write domain for the object if it's dirty. */
static int
-i915_gem_object_flush_gpu_write_domain(struct drm_gem_object *obj,
- bool pipelined)
+i915_gem_object_flush_gpu_write_domain(struct drm_gem_object *obj)
{
struct drm_device *dev = obj->dev;
uint32_t old_write_domain;
@@ -2762,10 +2760,7 @@
obj->read_domains,
old_write_domain);
- if (pipelined)
- return 0;
-
- return i915_gem_object_wait_rendering(obj, true);
+ return 0;
}
/** Flushes the GTT write domain for the object if it's dirty. */
@@ -2826,18 +2821,15 @@
if (obj_priv->gtt_space == NULL)
return -EINVAL;
- ret = i915_gem_object_flush_gpu_write_domain(obj, false);
+ ret = i915_gem_object_flush_gpu_write_domain(obj);
if (ret != 0)
return ret;
+ ret = i915_gem_object_wait_rendering(obj, true);
+ if (ret)
+ return ret;
i915_gem_object_flush_cpu_write_domain(obj);
- if (write) {
- ret = i915_gem_object_wait_rendering(obj, true);
- if (ret)
- return ret;
- }
-
old_write_domain = obj->write_domain;
old_read_domains = obj->read_domains;
@@ -2875,7 +2867,7 @@
if (obj_priv->gtt_space == NULL)
return -EINVAL;
- ret = i915_gem_object_flush_gpu_write_domain(obj, true);
+ ret = i915_gem_object_flush_gpu_write_domain(obj);
if (ret)
return ret;
@@ -2924,9 +2916,12 @@
uint32_t old_write_domain, old_read_domains;
int ret;
- ret = i915_gem_object_flush_gpu_write_domain(obj, false);
+ ret = i915_gem_object_flush_gpu_write_domain(obj);
if (ret != 0)
return ret;
+ ret = i915_gem_object_wait_rendering(obj, true);
+ if (ret)
+ return ret;
i915_gem_object_flush_gtt_write_domain(obj);
@@ -2935,12 +2930,6 @@
*/
i915_gem_object_set_to_full_cpu_read_domain(obj);
- if (write) {
- ret = i915_gem_object_wait_rendering(obj, true);
- if (ret)
- return ret;
- }
-
old_write_domain = obj->write_domain;
old_read_domains = obj->read_domains;
@@ -3205,9 +3194,13 @@
if (offset == 0 && size == obj->size)
return i915_gem_object_set_to_cpu_domain(obj, 0);
- ret = i915_gem_object_flush_gpu_write_domain(obj, false);
+ ret = i915_gem_object_flush_gpu_write_domain(obj);
if (ret != 0)
return ret;
+ ret = i915_gem_object_wait_rendering(obj, true);
+ if (ret)
+ return ret;
+
i915_gem_object_flush_gtt_write_domain(obj);
/* If we're already fully in the CPU read domain, we're done. */