When resources are allocated dynamically during an IOCTL we need to make sure that a fence slot is reserved so that the resulting fence can be added in the end.
Signed-off-by: Christian König christian.koenig@amd.com Fixes: c8d4c18bfbc4 ("dma-buf/drivers: make reserving a shared slot mandatory v4") --- drivers/gpu/drm/vmwgfx/vmwgfx_resource.c | 4 ++++ 1 file changed, 4 insertions(+)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c index a7d62a4eb47b..edce228f737c 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c @@ -399,6 +399,10 @@ static int vmw_resource_do_validate(struct vmw_resource *res, ret = func->create(res); if (unlikely(ret != 0)) return ret; + + ret = dma_resv_reserve_fences(res->backup->base.base.resv, 1); + if (unlikely(ret)) + goto out_bind_failed; }
if (func->bind &&
Am 22.04.22 um 11:20 schrieb Christian König:
When resources are allocated dynamically during an IOCTL we need to make sure that a fence slot is reserved so that the resulting fence can be added in the end.
I should probably add that this is only compile tested.
Zack you should probably give it a try for your issue.
Thanks for the help and sorry for the noise, Christian.
Signed-off-by: Christian König christian.koenig@amd.com Fixes: c8d4c18bfbc4 ("dma-buf/drivers: make reserving a shared slot mandatory v4")
drivers/gpu/drm/vmwgfx/vmwgfx_resource.c | 4 ++++ 1 file changed, 4 insertions(+)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c index a7d62a4eb47b..edce228f737c 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c @@ -399,6 +399,10 @@ static int vmw_resource_do_validate(struct vmw_resource *res, ret = func->create(res); if (unlikely(ret != 0)) return ret;
ret = dma_resv_reserve_fences(res->backup->base.base.resv, 1);if (unlikely(ret))goto out_bind_failed;}
if (func->bind &&
On Fri, 2022-04-22 at 11:21 +0200, Christian König wrote:
Am 22.04.22 um 11:20 schrieb Christian König:
When resources are allocated dynamically during an IOCTL we need to make sure that a fence slot is reserved so that the resulting fence can be added in the end.
I should probably add that this is only compile tested.
Zack you should probably give it a try for your issue.
I think we're on the right track but cotables are the only objects that create a bo in the create callback so we endup double reserving all the other backup objects. If you don't mind I'd prefer to move the code to localize the reservation to the problematic spot. I'll send it in a sec (6d0fdf27e98a ("drm/vmwgfx: Reserve fence slots on buffer objects in cotables") let me know if I can add a Co-developed-by: Christian König christian.koenig@amd.com tag to it.
z
Am 22.04.22 um 18:13 schrieb Zack Rusin:
On Fri, 2022-04-22 at 11:21 +0200, Christian König wrote:
Am 22.04.22 um 11:20 schrieb Christian König:
When resources are allocated dynamically during an IOCTL we need to make sure that a fence slot is reserved so that the resulting fence can be added in the end.
I should probably add that this is only compile tested.
Zack you should probably give it a try for your issue.
I think we're on the right track but cotables are the only objects that create a bo in the create callback so we endup double reserving all the other backup objects. If you don't mind I'd prefer to move the code to localize the reservation to the problematic spot. I'll send it in a sec (6d0fdf27e98a ("drm/vmwgfx: Reserve fence slots on buffer objects in cotables") let me know if I can add a Co-developed-by: Christian König christian.koenig@amd.com tag to it.
Sure, please go ahead. You know the code much better than I do anyway.
Thanks, Christian.
z
dri-devel@lists.freedesktop.org
-
Christian König -
Christian König -
Zack Rusin