[PATCH Review 1/1] drm/amdgpu: Workaround to skip kiq ring test during ras gpu recovery

[Christian König]

Am 17.10.23 um 16:36 schrieb Stanley.Yang:
> This is workaround, kiq ring test failed in suspend stage when do ras
> recovery for gfx v9_4_3.

Any idea why that failed? Problems like this usually point to an 
incorrect init or in this case re-init procedure and are actually what 
the ring test should uncover.

Christian.

>
> Change-Id: I8de9900aa76706f59bc029d4e9e8438c6e1db8e0
> Signed-off-by: Stanley.Yang <Stanley.Yang at amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 21 +++++++++++++++++++++
>   1 file changed, 21 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> index 9a158018ae16..902e60203809 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> @@ -29,6 +29,7 @@
>   #include "amdgpu_rlc.h"
>   #include "amdgpu_ras.h"
>   #include "amdgpu_xcp.h"
> +#include "amdgpu_xgmi.h"
>   
>   /* delay 0.1 second to enable gfx off feature */
>   #define GFX_OFF_DELAY_ENABLE         msecs_to_jiffies(100)
> @@ -501,6 +502,9 @@ int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev, int xcc_id)
>   {
>   	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
>   	struct amdgpu_ring *kiq_ring = &kiq->ring;
> +	struct amdgpu_hive_info *hive;
> +	struct amdgpu_ras *ras;
> +	int hive_ras_recovery;
>   	int i, r = 0;
>   	int j;
>   
> @@ -521,6 +525,23 @@ int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev, int xcc_id)
>   					   RESET_QUEUES, 0, 0);
>   	}
>   
> +	/**
> +	 * This is workaround: only skip kiq_ring test
> +	 * during ras recovery in suspend stage for gfx v9_4_3
> +	 */
> +	hive = amdgpu_get_xgmi_hive(adev);
> +	if (hive) {
> +		hive_ras_recovery = atomic_read(&hive->ras_recovery);
> +		amdgpu_put_xgmi_hive(hive);
> +	}
> +
> +	ras = amdgpu_ras_get_context(adev);
> +	if ((amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3)) &&
> +		ras && (atomic_read(&ras->in_recovery) || hive_ras_recovery)) {
> +		spin_unlock(&kiq->ring_lock);
> +		return 0;
> +	}
> +
>   	if (kiq_ring->sched.ready && !adev->job_hang)
>   		r = amdgpu_ring_test_helper(kiq_ring);
>   	spin_unlock(&kiq->ring_lock);