2020-06-13 12:36:22

by Xiyu Yang

Subject: [PATCH] drm/ttm: Fix dma_fence refcnt leak when adding move fence

ttm_bo_add_move_fence() invokes dma_fence_get(), which returns a
reference of the specified dma_fence object to "fence" with increased
refcnt.

When ttm_bo_add_move_fence() returns, local variable "fence" becomes
invalid, so the refcount should be decreased to keep refcount balanced.

The reference counting issue happens in one exception handling path of
ttm_bo_add_move_fence(). When no_wait_gpu flag is equals to true, the
function forgets to decrease the refcnt increased by dma_fence_get(),
causing a refcnt leak.

Fix this issue by calling dma_fence_put() when no_wait_gpu flag is
equals to true.

Signed-off-by: Xiyu Yang <[email protected]>
Signed-off-by: Xin Tan <[email protected]>
---
drivers/gpu/drm/ttm/ttm_bo.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index f73b81c2576e..0f20e14a4cfd 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -883,8 +883,10 @@ static int ttm_bo_add_move_fence(struct ttm_buffer_object *bo,
if (!fence)
return 0;

- if (no_wait_gpu)
+ if (no_wait_gpu) {
+ dma_fence_put(fence);
return -EBUSY;
+ }

dma_resv_add_shared_fence(bo->base.resv, fence);

--
2.7.4

2020-06-15 08:58:35

by Christian König

Subject: Re: [PATCH] drm/ttm: Fix dma_fence refcnt leak when adding move fence

Am 13.06.20 um 14:30 schrieb Xiyu Yang:
> ttm_bo_add_move_fence() invokes dma_fence_get(), which returns a
> reference of the specified dma_fence object to "fence" with increased
> refcnt.
>
> When ttm_bo_add_move_fence() returns, local variable "fence" becomes
> invalid, so the refcount should be decreased to keep refcount balanced.
>
> The reference counting issue happens in one exception handling path of
> ttm_bo_add_move_fence(). When no_wait_gpu flag is equals to true, the
> function forgets to decrease the refcnt increased by dma_fence_get(),
> causing a refcnt leak.
>
> Fix this issue by calling dma_fence_put() when no_wait_gpu flag is
> equals to true.
>
> Signed-off-by: Xiyu Yang <[email protected]>
> Signed-off-by: Xin Tan <[email protected]>

Reviewed and pushed this one as well as the other ttm fix to drm-misc-fixes.

That should show up in Linus tree rather soon.

Thanks for the help,
Christian.

PS: Are you working on some new automated scripts to catch that stuff or
how did you stumbled over it?

> ---
> drivers/gpu/drm/ttm/ttm_bo.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
> index f73b81c2576e..0f20e14a4cfd 100644
> --- a/drivers/gpu/drm/ttm/ttm_bo.c
> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
> @@ -883,8 +883,10 @@ static int ttm_bo_add_move_fence(struct ttm_buffer_object *bo,
> if (!fence)
> return 0;
>
> - if (no_wait_gpu)
> + if (no_wait_gpu) {
> + dma_fence_put(fence);
> return -EBUSY;
> + }
>
> dma_resv_add_shared_fence(bo->base.resv, fence);
>