[PATCH] drm/amdgpu: Fix KIQ hang on bare metal for device unbind/bind back.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Mar 28, 2018 at 8:37 AM, Andrey Grodzovsky
<andrey.grodzovsky at amd.com> wrote:
> Problem: When unbind and then bind back the device KIQ hangs on Vega
> after mapping KCQs request.
>
> Fix: Adding deinitialzie code from CAIL during HW fini solves the
> hang.
>
> Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky at amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 52 ++++++++++++++++++++++++++++++++++-
>  1 file changed, 51 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> index 1ae3de1..fd13065 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> @@ -2757,6 +2757,45 @@ static int gfx_v9_0_kiq_init_register(struct amdgpu_ring *ring)
>         return 0;
>  }
>
> +static int gfx_v9_0_kiq_fini_register(struct amdgpu_ring *ring)
> +{
> +       struct amdgpu_device *adev = ring->adev;
> +       int j;
> +
> +       /* disable the queue if it's active */
> +       if (RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1) {
> +
> +               WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST, 1);
> +
> +               for (j = 0; j < adev->usec_timeout; j++) {
> +                       if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1))
> +                               break;
> +                       udelay(1);
> +               }
> +
> +               if (adev->usec_timeout == AMDGPU_MAX_USEC_TIMEOUT) {
> +                       DRM_DEBUG("KIQ dequeue request failed.\n");
> +
> +                       WREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE, 0);
> +               }
> +
> +               /* Manual disable if dequeue request times out */
> +               WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST,
> +                     0);
> +       }
> +
> +       WREG32_SOC15(GC, 0, mmCP_HQD_IQ_TIMER, 0);
> +       WREG32_SOC15(GC, 0, mmCP_HQD_IB_CONTROL, 0);
> +       WREG32_SOC15(GC, 0, mmCP_HQD_PERSISTENT_STATE, 0);
> +       WREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL, 0x40000000);
> +       WREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL, 0);
> +       WREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR, 0);
> +       WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_HI, 0);
> +       WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_LO, 0);
> +
> +       return 0;
> +}
> +
>  static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
>  {
>         struct amdgpu_device *adev = ring->adev;
> @@ -3010,7 +3049,6 @@ static int gfx_v9_0_kcq_disable(struct amdgpu_ring *kiq_ring,struct amdgpu_ring
>         return r;
>  }
>
> -
>  static int gfx_v9_0_hw_fini(void *handle)
>  {
>         struct amdgpu_device *adev = (struct amdgpu_device *)handle;
> @@ -3033,6 +3071,18 @@ static int gfx_v9_0_hw_fini(void *handle)
>                 WREG32_FIELD15(GC, 0, CP_PQ_WPTR_POLL_CNTL, EN, 0);
>                 return 0;
>         }
> +
> +       /* Use deinitialize sequence from CAIL when unbinding device from driver,
> +        * otherwise KIQ is hanging when binding back
> +        */
> +       if (!adev->in_gpu_reset && !adev->gfx.in_suspend) {

You need to take the srbm_mutex before using soc15_grbm_select().
WIth that fixed:
Reviewed-by: Alex Deucher <alexander.deucher at amd.com>

> +               soc15_grbm_select(adev, adev->gfx.kiq.ring.me,
> +                               adev->gfx.kiq.ring.pipe,
> +                               adev->gfx.kiq.ring.queue, 0);
> +               gfx_v9_0_kiq_fini_register(&adev->gfx.kiq.ring);
> +               soc15_grbm_select(adev, 0, 0, 0, 0);
> +       }
> +
>         gfx_v9_0_cp_enable(adev, false);
>         gfx_v9_0_rlc_stop(adev);
>
> --
> 2.7.4
>
> _______________________________________________
> amd-gfx mailing list
> amd-gfx at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx


[Index of Archives]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux