[PATCH] drm/amdgpu: disable gfxoff if VCN is busy

[Chen, Jiansong (Simon)]

[AMD Official Use Only - Internal Distribution Only]

Hi James,
Thanks for your input, v2 patch is sent out.

Regards,
Jiansong
-----Original Message-----
From: Zhu, James <James.Zhu at amd.com>
Sent: Friday, October 30, 2020 9:06 PM
To: Chen, Jiansong (Simon) <Jiansong.Chen at amd.com>; amd-gfx at lists.freedesktop.org
Cc: Zhang, Hawking <Hawking.Zhang at amd.com>; Zhu, James <James.Zhu at amd.com>; Liu, Leo <Leo.Liu at amd.com>; Zhou1, Tao <Tao.Zhou1 at amd.com>
Subject: Re: [PATCH] drm/amdgpu: disable gfxoff if VCN is busy

Hi Jiansong

Pls check inline.

thanks!

James

On 2020-10-30 7:37 a.m., Jiansong Chen wrote:
> Toggle on/off gfxoff during video playback to fix gpu hang.
[JZ] It is a workaround, not a fix. Also Arcturus needn't this WA.
> Signed-off-by: Jiansong Chen <Jiansong.Chen at amd.com>
> Change-Id: I5b938c446884268c2cda0801121a53da980e603a
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 10 +++++++---
>   1 file changed, 7 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> index 277a8435dd06..444b89413232 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> @@ -358,6 +358,7 @@ static void amdgpu_vcn_idle_work_handler(struct work_struct *work)
>   }
>
>   if (!fences && !atomic_read(&adev->vcn.total_submission_cnt)) {
> +amdgpu_gfx_off_ctrl(adev, true);
>   amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
>          AMD_PG_STATE_GATE);
>   } else {
> @@ -368,13 +369,16 @@ static void amdgpu_vcn_idle_work_handler(struct work_struct *work)
>   void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring)
>   {
>   struct amdgpu_device *adev = ring->adev;
> +bool set_clocks = !cancel_delayed_work_sync(&adev->vcn.idle_work);
>
>   atomic_inc(&adev->vcn.total_submission_cnt);
> -cancel_delayed_work_sync(&adev->vcn.idle_work);
>
>   mutex_lock(&adev->vcn.vcn_pg_lock);
> -amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
> -       AMD_PG_STATE_UNGATE);
> +if (set_clocks) {
> +amdgpu_gfx_off_ctrl(adev, false);

[JZ] Move the above two lines before mutex_lock(&adev->vcn.vcn_pg_lock);

Since it may cause S3 test failure.

> +amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
> +AMD_PG_STATE_UNGATE);
> +}
>
>   if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG){
>   struct dpg_pause_state new_state;