[AMD Official Use Only - General]
Is it better to handle CE and UE list separately?
Anyway Reviewed-by: Stanley.Yang <Stanley.Yang@xxxxxxx>
Regards,
Stanley
> -----Original Message-----
> From: amd-gfx <amd-gfx-bounces@xxxxxxxxxxxxxxxxxxxxx> On Behalf Of Tao
> Zhou
> Sent: Tuesday, October 31, 2023 3:09 PM
> To: amd-gfx@xxxxxxxxxxxxxxxxxxxxx
> Cc: Chai, Thomas <YiPeng.Chai@xxxxxxx>; Zhou1, Tao
> <Tao.Zhou1@xxxxxxx>
> Subject: [PATCH] drm/amdgpu: handle extra UE register entries for gfx v9_4_3
>
> The UE registe list is larger than CE list.
>
> Reported-by: yipeng.chai@xxxxxxx
> Signed-off-by: Tao Zhou <tao.zhou1@xxxxxxx>
> ---
> drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 38
> +++++++++++++++++++++++++
> 1 file changed, 38 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
> b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
> index 41bbabd9ad4d..046ae95b366a 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
> @@ -3799,6 +3799,27 @@ static void
> gfx_v9_4_3_inst_query_ras_err_count(struct amdgpu_device *adev,
> }
> }
>
> + /* handle extra register entries of UE */
> + for (; i < ARRAY_SIZE(gfx_v9_4_3_ue_reg_list); i++) {
> + for (j = 0; j < gfx_v9_4_3_ue_reg_list[i].se_num; j++) {
> + for (k = 0; k <
> gfx_v9_4_3_ue_reg_list[i].reg_entry.reg_inst; k++) {
> + /* no need to select if instance number is 1 */
> + if (gfx_v9_4_3_ue_reg_list[i].se_num > 1 ||
> +
> gfx_v9_4_3_ue_reg_list[i].reg_entry.reg_inst > 1)
> + gfx_v9_4_3_xcc_select_se_sh(adev, j,
> 0, k, xcc_id);
> +
> +
> amdgpu_ras_inst_query_ras_error_count(adev,
> +
> &(gfx_v9_4_3_ue_reg_list[i].reg_entry),
> + 1,
> +
> gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ue_reg_list[i].mem_id_t
> ype].mem_id_ent,
> +
> gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ue_reg_list[i].mem_id_t
> ype].size,
> + GET_INST(GC, xcc_id),
> +
> AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
> + &ue_count);
> + }
> + }
> + }
> +
> gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
> xcc_id);
> mutex_unlock(&adev->grbm_idx_mutex);
> @@ -3838,6 +3859,23 @@ static void
> gfx_v9_4_3_inst_reset_ras_err_count(struct amdgpu_device *adev,
> }
> }
>
> + /* handle extra register entries of UE */
> + for (; i < ARRAY_SIZE(gfx_v9_4_3_ue_reg_list); i++) {
> + for (j = 0; j < gfx_v9_4_3_ue_reg_list[i].se_num; j++) {
> + for (k = 0; k <
> gfx_v9_4_3_ue_reg_list[i].reg_entry.reg_inst; k++) {
> + /* no need to select if instance number is 1 */
> + if (gfx_v9_4_3_ue_reg_list[i].se_num > 1 ||
> +
> gfx_v9_4_3_ue_reg_list[i].reg_entry.reg_inst > 1)
> + gfx_v9_4_3_xcc_select_se_sh(adev, j,
> 0, k, xcc_id);
> +
> +
> amdgpu_ras_inst_reset_ras_error_count(adev,
> +
> &(gfx_v9_4_3_ue_reg_list[i].reg_entry),
> + 1,
> + GET_INST(GC, xcc_id));
> + }
> + }
> + }
> +
> gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
> xcc_id);
> mutex_unlock(&adev->grbm_idx_mutex);
> --
> 2.35.1
