diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c | 13 |
1 files changed, 10 insertions, 3 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c index fe31cbeccfe9..5fd67e1cc2a0 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c @@ -238,16 +238,20 @@ static void xgpu_ai_mailbox_flr_work(struct work_struct *work) struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, flr_work); struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt); int timeout = AI_MAILBOX_POLL_FLR_TIMEDOUT; + int locked; /* block amdgpu_gpu_recover till msg FLR COMPLETE received, * otherwise the mailbox msg will be ruined/reseted by * the VF FLR. * - * we can unlock the reset_sem to allow "amdgpu_job_timedout" + * we can unlock the lock_reset to allow "amdgpu_job_timedout" * to run gpu_recover() after FLR_NOTIFICATION_CMPL received * which means host side had finished this VF's FLR. */ - down_read(&adev->reset_sem); + locked = mutex_trylock(&adev->lock_reset); + if (locked) + adev->in_gpu_reset = true; + do { if (xgpu_ai_mailbox_peek_msg(adev) == IDH_FLR_NOTIFICATION_CMPL) goto flr_done; @@ -257,7 +261,10 @@ static void xgpu_ai_mailbox_flr_work(struct work_struct *work) } while (timeout > 1); flr_done: - up_read(&adev->reset_sem); + if (locked) { + adev->in_gpu_reset = false; + mutex_unlock(&adev->lock_reset); + } /* Trigger recovery for world switch failure if no TDR */ if (amdgpu_device_should_recover_gpu(adev) |