Re: [PATCH v2] drm: Fix drm_pending_vblank_event leak in error path for out_fence_ptr

[Melissa Wen]

On 29/07/2026 08:10, Thadeu Lima de Souza Cascardo wrote:
> When an out_fence_ptr is provided but DRM_MODE_PAGE_FLIP_EVENT is not
> set, a drm_pending_vblank_event will be allocated. If later, there is an
> allocation failure or another failure at setup_out_fence(), that event
> will not have base.fence set and it will not be released at
> complete_signaling().
>
> Release the event and set crtc_state->event to NULL just like in the
> DRM_MODE_PAGE_FLIP_EVENT case when there is a failure at
> drm_event_reserve_init(). That is, prepare_signaling() releases the
> event and there is nothing to be done at complete_signaling(). Use
> drm_event_cancel_free() as that will also undo drm_event_reserve_init()
> in case it has been called.
Hey,

Nice catch!
> Reported-by: sashiko-bot@xxxxxxxxxx

Take this with a grain of salt: I think checkpatch will complain about 
reporter
as just an email. I usually use:
`Reported-by: Sashiko <sashiko-bot@xxxxxxxxxx>` but not 100% sure.
> Closes: https://sashiko.dev/#/patchset/20260727-drm_crtc_atomic_commit_leak-v1-1-23d9948a9d7c@xxxxxxxxxx?part=1
> Fixes: 92c715fca907 ("drm/atomic: Fix double free in drm_atomic_state_default_clear")
> Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@xxxxxxxxxx>
> ---
> Changes in v2:
> - Fix UAF when DRM_MODE_PAGE_FLIP_EVENT is used.
> - Link to v1: https://patch.msgid.link/20260728-drm_pending_vblank_event_leak-v1-1-08429b920b16@xxxxxxxxxx
> ---
>   drivers/gpu/drm/drm_atomic_uapi.c | 13 +++++++++++--
>   1 file changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/drm_atomic_uapi.c b/drivers/gpu/drm/drm_atomic_uapi.c
> index 1050dddadb17..aed6040182eb 100644
> --- a/drivers/gpu/drm/drm_atomic_uapi.c
> +++ b/drivers/gpu/drm/drm_atomic_uapi.c
> @@ -1464,11 +1464,15 @@ static int prepare_signaling(struct drm_device *dev,
>   		if (fence_ptr) {
>   			struct dma_fence *fence;
>   			struct drm_out_fence_state *f;
> +			struct drm_pending_vblank_event *e = crtc_state->event;
>   
>   			f = krealloc(*fence_state, sizeof(**fence_state) *
>   				     (*num_fences + 1), GFP_KERNEL);
> -			if (!f)
> +			if (!f) {
> +				drm_event_cancel_free(dev, &e->base);
> +				crtc_state->event = NULL;
>   				return -ENOMEM;
> +			}
>   
>   			memset(&f[*num_fences], 0, sizeof(*f));
>   
> @@ -1476,12 +1480,17 @@ static int prepare_signaling(struct drm_device *dev,
>   			*fence_state = f;
>   
>   			fence = drm_crtc_create_fence(crtc);
> -			if (!fence)
> +			if (!fence) {
> +				drm_event_cancel_free(dev, &e->base);
> +				crtc_state->event = NULL;
>   				return -ENOMEM;
> +			}
>   
>   			ret = setup_out_fence(&f[(*num_fences)++], fence);
>   			if (ret) {
>   				dma_fence_put(fence);
> +				drm_event_cancel_free(dev, &e->base);
> +				crtc_state->event = NULL;

How about to goto a single error handle instead of repeating those three 
lines
three times? I think it'd be better for maintenance, but the fix looks 
correct as is:

Reviewed-by: Melissa Wen <mwen@xxxxxxxxxx>

>   				return ret;
>   			}
>   
>
> ---
> base-commit: df0311845915442ab75d03ee89fb443b444c1370
> change-id: 20260728-drm_pending_vblank_event_leak-a36cedb296ba
>
> Best regards,
> --
> Thadeu Lima de Souza Cascardo <cascardo@xxxxxxxxxx>