Re: [PATCH 4/5] drm/amdgpu: wait for gpu to complete reset

["Christian König" <ckoenig.leichtzumerken@gmail.com>]

Am 13.06.24 um 04:25 schrieb YiPeng Chai:
> Add completion to wait for gpu to complete reset.
>
> Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 12 ++++++++++++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h |  1 +
>   2 files changed, 13 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> index 7dfb2e548d70..341c9bd0d1a4 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> @@ -124,6 +124,8 @@ const char *get_ras_block_str(struct ras_common_if *ras_block)
>   
>   #define AMDGPU_RAS_RETIRE_PAGE_INTERVAL 100  //ms
>   
> +#define MAX_GPU_RESET_COMPLETION_TIME  120000 //ms
> +
>   #define RAS_POISON_FIFO_MSG_PENDING_THRESHOLD  (AMDGPU_RAS_POISON_FIFO_SIZE/4)
>   
>   enum amdgpu_ras_retire_page_reservation {
> @@ -2526,6 +2528,8 @@ static void amdgpu_ras_do_recovery(struct work_struct *work)
>   		atomic_set(&hive->ras_recovery, 0);
>   		amdgpu_put_xgmi_hive(hive);
>   	}
> +
> +	complete(&ras->gpu_reset_completion);
>   }
>   
>   /* alloc/realloc bps array */
> @@ -2946,7 +2950,14 @@ static int amdgpu_ras_poison_consumption_handler(struct amdgpu_device *adev,
>   			con->gpu_reset_flags |= reset;
>   		}
>   
> +		reinit_completion(&con->gpu_reset_completion);
> +
>   		amdgpu_ras_reset_gpu(adev);
> +
> +		if (!wait_for_completion_timeout(&con->gpu_reset_completion,
> +				msecs_to_jiffies(MAX_GPU_RESET_COMPLETION_TIME)))
> +			dev_err(adev->dev, "Waiting for GPU to complete reset timeout! reset:0x%x\n",
> +				reset);

Are there any looks taken here which the GPU reset might need as well?

Regards,
Christian.

>   	}
>   
>   	return 0;
> @@ -3072,6 +3083,7 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev)
>   		}
>   	}
>   
> +	init_completion(&con->gpu_reset_completion);
>   	mutex_init(&con->page_rsv_lock);
>   	INIT_KFIFO(con->poison_fifo);
>   	mutex_init(&con->page_retirement_lock);
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
> index 103436bb650e..d5ddd0ca5de1 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
> @@ -537,6 +537,7 @@ struct amdgpu_ras {
>   	DECLARE_KFIFO(poison_fifo, struct ras_poison_msg, AMDGPU_RAS_POISON_FIFO_SIZE);
>   	struct ras_ecc_log_info  umc_ecc_log;
>   	struct delayed_work page_retirement_dwork;
> +	struct completion gpu_reset_completion;
>   
>   	/* Fatal error detected flag */
>   	atomic_t fed;