On Wed, Sep 10, 2025 at 8:14 AM Prike Liang <prike.li...@amd.com> wrote:
>
> Keeping waiting the userq fence infinitely untill
> hang detection, and then suspend the hang queue and
> set the fence error.
>
> Signed-off-by: Prike Liang <prike.li...@amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 11 ++++++++---
> 1 file changed, 8 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
> index 7b7dae436e5e..2626a41a8418 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
> @@ -199,7 +199,7 @@ amdgpu_userq_map_helper(struct amdgpu_userq_mgr *uq_mgr,
> return r;
> }
>
> -static void
> +static int
> amdgpu_userq_wait_for_last_fence(struct amdgpu_userq_mgr *uq_mgr,
> struct amdgpu_usermode_queue *queue)
> {
> @@ -207,11 +207,16 @@ amdgpu_userq_wait_for_last_fence(struct
> amdgpu_userq_mgr *uq_mgr,
> int ret;
>
> if (f && !dma_fence_is_signaled(f)) {
> - ret = dma_fence_wait_timeout(f, true, msecs_to_jiffies(100));
> - if (ret <= 0)
> + ret = dma_fence_wait_timeout(f, true, MAX_SCHEDULE_TIMEOUT);
> + if (ret <= 0) {
> drm_file_err(uq_mgr->file, "Timed out waiting for
> fence=%llu:%llu\n",
> f->context, f->seqno);
> + queue->state = AMDGPU_USERQ_STATE_HUNG;
I don't think we want to set the queue state to hung here. Just
return an error. We'll detect the hang when we attempt to preempt the
queues and run detect_and_reset().
Alex
> + return -ETIME;
> + }
> }
> +
> + return ret;
> }
>
> static void
> --
> 2.34.1
>
