[AMD Official Use Only - AMD Internal Distribution Only]
Nack to the revert. The FLR sequence is defined as the following (host-initiated reset):
1) host sends FLR_NOTIFICATION
2) Guest gets interrupt and queues FLR work item
3) Guest sends READY_TO_RESET
4) Host sends FLR_NOTIFICATION_COMPLETION
5) Guest starts recovery
In RAS FED, guest interrupts are disabled and therefore it won't receive #1. Consequently #2 & #4 will break.
It doesn't make sense to re-use this sequence as-is in FED scenario. On the other hand,
KFD reset work item performs the guest-initiated reset:
1) Guest waits for mailbox to work (handles the FED disable mailbox)
2) Guest sends REQ_GPU_RESET_ACCESS
3) Host acks back
4) Guest starts recovery
We should keep this commit until proper guest FED reset workitem is implemented.
Thanks,
Victor
> -----Original Message-----
> From: amd-gfx <amd-gfx-bounces@xxxxxxxxxxxxxxxxxxxxx> On Behalf Of
> Yunxiang Li
> Sent: Tuesday, May 28, 2024 1:24 PM
> To: amd-gfx@xxxxxxxxxxxxxxxxxxxxx
> Cc: Deucher, Alexander <Alexander.Deucher@xxxxxxx>; Koenig, Christian
> <Christian.Koenig@xxxxxxx>; Gao, Likun <Likun.Gao@xxxxxxx>; Zhang,
> Hawking <Hawking.Zhang@xxxxxxx>; Li, Yunxiang (Teddy)
> <Yunxiang.Li@xxxxxxx>
> Subject: [PATCH v2 10/10] Revert "drm/amdgpu: Queue KFD reset workitem in
> VF FED"
>
> Caution: This message originated from an External Source. Use proper caution
> when opening attachments, clicking links, or responding.
>
>
> This reverts commit 2149ee697a7a3091a16447c647d4a30f7468553a.
>
> The issue is already fixed by
> fa5a7f2ccb7e ("drm/amdgpu: Fix two reset triggered in a row")
>
> Signed-off-by: Yunxiang Li <Yunxiang.Li@xxxxxxx>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
> index 44450507c140..4bacbf1db9e5 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
> @@ -616,7 +616,7 @@ static void
> amdgpu_virt_update_vf2pf_work_item(struct work_struct *work)
> amdgpu_sriov_runtime(adev)) {
> amdgpu_ras_set_fed(adev, true);
> if (amdgpu_reset_domain_schedule(adev->reset_domain,
> - &adev->kfd.reset_work))
> + &adev->virt.flr_work))
> return;
> else
> dev_err(adev->dev, "Failed to queue work! at %s", __func__);
> --
> 2.34.1
