[PATCH 1/3] drm/amdgpu: Don't use kiq in gpu reset

Christian König ckoenig.leichtzumerken at gmail.com
Wed Aug 22 12:23:31 UTC 2018


Am 22.08.2018 um 06:39 schrieb Emily Deng:
> When in gpu reset, don't use kiq, it will generate more TDR.
>
> Signed-off-by: Emily Deng <Emily.Deng at amd.com>

Patch #1 is Reviewed-by: Christian König <christian.koenig at amd.com>.

Patch #2 actually not necessary since we should never flush the tlb from 
interrupt context.

Patch #3: I would actually rather keep that an error message cause it 
still means that something went wrong.

Christian.

> ---
>   drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 15 ++++-----------
>   1 file changed, 4 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
> index eec991f..fcdbacb 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
> @@ -331,15 +331,8 @@ signed long  amdgpu_kiq_reg_write_reg_wait(struct amdgpu_device *adev,
>   
>   	r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
>   
> -	/* don't wait anymore for gpu reset case because this way may
> -	 * block gpu_recover() routine forever, e.g. this virt_kiq_rreg
> -	 * is triggered in TTM and ttm_bo_lock_delayed_workqueue() will
> -	 * never return if we keep waiting in virt_kiq_rreg, which cause
> -	 * gpu_recover() hang there.
> -	 *
> -	 * also don't wait anymore for IRQ context
> -	 * */
> -	if (r < 1 && (adev->in_gpu_reset || in_interrupt()))
> +	/* don't wait anymore for IRQ context */
> +	if (r < 1 && in_interrupt())
>   		goto failed_kiq;
>   
>   	might_sleep();
> @@ -387,8 +380,8 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev,
>   		u32 tmp = gmc_v9_0_get_invalidate_req(vmid);
>   
>   		if (adev->gfx.kiq.ring.ready &&
> -		    (amdgpu_sriov_runtime(adev) ||
> -		     !amdgpu_sriov_vf(adev))) {
> +		    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev)) &&
> +		    !adev->in_gpu_reset) {
>   			r = amdgpu_kiq_reg_write_reg_wait(adev, hub->vm_inv_eng0_req + eng,
>   				hub->vm_inv_eng0_ack + eng, tmp, 1 << vmid);
>   			if (!r)



More information about the amd-gfx mailing list