diff options
author | Eric Anholt <eric@anholt.net> | 2008-11-14 16:27:47 -0800 |
---|---|---|
committer | Dave Airlie <airlied@redhat.com> | 2008-12-04 11:21:48 +1000 |
commit | b670d8158283c35842ae1c650f75c375d8710607 (patch) | |
tree | 5acd986f51f15b50f401826e5ff53345a86194a8 /drivers | |
parent | ce44b0ea3dc04236d852d78a06f850d1f7b03f3d (diff) |
drm/i915: If interrupted while setting object domains, still emit the flush.
Otherwise, we would leave the objects in an inconsistent state, such as
write_domain == 0 but on the flushing list.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Diffstat (limited to 'drivers')
-rw-r--r-- | drivers/gpu/drm/i915/i915_gem.c | 14 |
1 files changed, 13 insertions, 1 deletions
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 9fd28ebe0aa3..3fe108b7e2fa 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1880,8 +1880,20 @@ i915_gem_execbuffer(struct drm_device *dev, void *data, ret = i915_gem_object_set_domain(obj, obj->pending_read_domains, obj->pending_write_domain); - if (ret) + if (ret) { + /* As we've partially updated domains on our buffers, + * we have to emit the flush we've accumulated + * before exiting, or we'll have broken the + * active/flushing/inactive invariants. + * + * We'll potentially have some things marked as + * being in write domains that they actually aren't, + * but that should be merely a minor performance loss. + */ + flush_domains = i915_gem_dev_set_domain(dev); + (void)i915_add_request(dev, flush_domains); goto err; + } } i915_verify_inactive(dev, __FILE__, __LINE__); |