Re: [PATCH 1/4] drm/amdgpu/vcn: fix race condition issue for vcn start

[Christian König <ckoenig.leichtzumerken@xxxxxxxxx>]

Am 03.03.20 um 19:16 schrieb James Zhu:
> Fix race condition issue when multiple vcn starts are called.
>
> Signed-off-by: James Zhu <James.Zhu@xxxxxxx>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 4 ++++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h | 1 +
>   2 files changed, 5 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> index f96464e..aa7663f 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> @@ -63,6 +63,7 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
>   	int i, r;
>   
>   	INIT_DELAYED_WORK(&adev->vcn.idle_work, amdgpu_vcn_idle_work_handler);
> +	mutex_init(&adev->vcn.vcn_pg_lock);
>   
>   	switch (adev->asic_type) {
>   	case CHIP_RAVEN:
> @@ -210,6 +211,7 @@ int amdgpu_vcn_sw_fini(struct amdgpu_device *adev)
>   	}
>   
>   	release_firmware(adev->vcn.fw);
> +	mutex_destroy(&adev->vcn.vcn_pg_lock);
>   
>   	return 0;
>   }
> @@ -321,6 +323,7 @@ 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);
>   
> +	mutex_lock(&adev->vcn.vcn_pg_lock);

That still won't work correctly here.

The whole idea of the cancel_delayed_work_sync() and 
schedule_delayed_work() dance is that you have exactly one user of that. 
If you have multiple rings that whole thing won't work correctly.

To fix this you need to call mutex_lock() before 
cancel_delayed_work_sync() and schedule_delayed_work() before 
mutex_unlock().

Regards,
Christian.

>   	if (set_clocks) {
>   		amdgpu_gfx_off_ctrl(adev, false);
>   		amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
> @@ -345,6 +348,7 @@ void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring)
>   
>   		adev->vcn.pause_dpg_mode(adev, ring->me, &new_state);
>   	}
> +	mutex_unlock(&adev->vcn.vcn_pg_lock);
>   }
>   
>   void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring)
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
> index 6fe0573..2ae110d 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
> @@ -200,6 +200,7 @@ struct amdgpu_vcn {
>   	struct drm_gpu_scheduler *vcn_dec_sched[AMDGPU_MAX_VCN_INSTANCES];
>   	uint32_t		 num_vcn_enc_sched;
>   	uint32_t		 num_vcn_dec_sched;
> +	struct mutex		 vcn_pg_lock;
>   
>   	unsigned	harvest_config;
>   	int (*pause_dpg_mode)(struct amdgpu_device *adev,

_______________________________________________
amd-gfx mailing list
amd-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/amd-gfx