[PATCH] amd/amdgpu: perform soft reset for sdma fed error

[Alex Deucher]

On Fri, May 12, 2023 at 3:49 AM YiPeng Chai <YiPeng.Chai at amd.com> wrote:
>
> When testing sdma ib ring fails to detect sdma
> hang for sdma fed error, force to perform soft
> reset.
>
> Signed-off-by: YiPeng Chai <YiPeng.Chai at amd.com>

Acked-by: Alex Deucher <alexander.deucher at amd.com>

> ---
>  drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c | 16 ++++++++++++++++
>  1 file changed, 16 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
> index 3d9a80511a45..1d463e1fd3ae 100644
> --- a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
> @@ -53,6 +53,12 @@ MODULE_FIRMWARE("amdgpu/sdma_6_0_3.bin");
>  #define SDMA0_HYP_DEC_REG_START 0x5880
>  #define SDMA0_HYP_DEC_REG_END 0x589a
>  #define SDMA1_HYP_DEC_REG_OFFSET 0x20
> +#define regRLC_RLCS_FED_STATUS_0                     0x4eff
> +#define regRLC_RLCS_FED_STATUS_0_BASE_IDX            1
> +#define RLC_RLCS_FED_STATUS_0__SDMA0_FED_ERR_MASK    0x00000040L
> +#define RLC_RLCS_FED_STATUS_0__SDMA1_FED_ERR_MASK    0x00000080L
> +#define RLC_RLCS_FED_STATUS_0__SDMA0_FED_ERR__SHIFT  0x6
> +#define RLC_RLCS_FED_STATUS_0__SDMA1_FED_ERR__SHIFT  0x7
>
>  static void sdma_v6_0_set_ring_funcs(struct amdgpu_device *adev);
>  static void sdma_v6_0_set_buffer_funcs(struct amdgpu_device *adev);
> @@ -760,6 +766,7 @@ static bool sdma_v6_0_check_soft_reset(void *handle)
>         struct amdgpu_ring *ring;
>         int i, r;
>         long tmo = msecs_to_jiffies(1000);
> +       uint32_t rlc_status0 = 0;
>
>         for (i = 0; i < adev->sdma.num_instances; i++) {
>                 ring = &adev->sdma.instance[i].ring;
> @@ -768,6 +775,15 @@ static bool sdma_v6_0_check_soft_reset(void *handle)
>                         return true;
>         }
>
> +       /* Soft reset should be performed for sdma fed error to
> +        * recover sdma engine, so when testing sdma ib ring fails
> +        * to detect sdma hang, force to perform soft reset.
> +        */
> +       rlc_status0 = RREG32_SOC15(GC, 0, regRLC_RLCS_FED_STATUS_0);
> +       if (REG_GET_FIELD(rlc_status0, RLC_RLCS_FED_STATUS_0, SDMA0_FED_ERR) ||
> +           REG_GET_FIELD(rlc_status0, RLC_RLCS_FED_STATUS_0, SDMA1_FED_ERR))
> +               return true;
> +
>         return false;
>  }
>
> --
> 2.34.1
>