Re: [PATCH v2 2/7] drm/amdgpu: Switch to delayed work from work_struct.

[Christian König <christian.koenig@xxxxxxx>]

Am 17.05.22 um 21:20 schrieb Andrey Grodzovsky:
> We need to be able to non blocking cancel pending reset works
> from within GPU reset. Currently kernel API allows this only
> for delayed_work and not for work_struct. Switch to delayed
> work and queue it with delay 0 which is equal to queueing work
> struct.

Ok well that needs further explanation. Why should this be used exactly?

Christian.

> Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@xxxxxxx>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 8 ++++----
>   drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h  | 4 ++--
>   drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h   | 2 +-
>   drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c      | 4 ++--
>   drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c      | 4 ++--
>   drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c      | 4 ++--
>   6 files changed, 13 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> index 3948e7db6ad7..ea41edf52a6f 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> @@ -5309,7 +5309,7 @@ int amdgpu_device_gpu_recover_imp(struct amdgpu_device *adev,
>   }
>   
>   struct amdgpu_recover_work_struct {
> -	struct work_struct base;
> +	struct delayed_work base;
>   	struct amdgpu_device *adev;
>   	struct amdgpu_job *job;
>   	int ret;
> @@ -5317,7 +5317,7 @@ struct amdgpu_recover_work_struct {
>   
>   static void amdgpu_device_queue_gpu_recover_work(struct work_struct *work)
>   {
> -	struct amdgpu_recover_work_struct *recover_work = container_of(work, struct amdgpu_recover_work_struct, base);
> +	struct amdgpu_recover_work_struct *recover_work = container_of(work, struct amdgpu_recover_work_struct, base.work);
>   
>   	amdgpu_device_gpu_recover_imp(recover_work->adev, recover_work->job);
>   }
> @@ -5329,12 +5329,12 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
>   {
>   	struct amdgpu_recover_work_struct work = {.adev = adev, .job = job};
>   
> -	INIT_WORK(&work.base, amdgpu_device_queue_gpu_recover_work);
> +	INIT_DELAYED_WORK(&work.base, amdgpu_device_queue_gpu_recover_work);
>   
>   	if (!amdgpu_reset_domain_schedule(adev->reset_domain, &work.base))
>   		return -EAGAIN;
>   
> -	flush_work(&work.base);
> +	flush_delayed_work(&work.base);
>   
>   	return atomic_read(&adev->reset_domain->reset_res);
>   }
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
> index 9e55a5d7a825..fee9376d495a 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
> @@ -114,9 +114,9 @@ static inline void amdgpu_reset_put_reset_domain(struct amdgpu_reset_domain *dom
>   }
>   
>   static inline bool amdgpu_reset_domain_schedule(struct amdgpu_reset_domain *domain,
> -						struct work_struct *work)
> +						struct delayed_work *work)
>   {
> -	return queue_work(domain->wq, work);
> +	return queue_delayed_work(domain->wq, work, 0);
>   }
>   
>   void amdgpu_device_lock_reset_domain(struct amdgpu_reset_domain *reset_domain);
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
> index 239f232f9c02..b53b7a794459 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
> @@ -230,7 +230,7 @@ struct amdgpu_virt {
>   	uint32_t			reg_val_offs;
>   	struct amdgpu_irq_src		ack_irq;
>   	struct amdgpu_irq_src		rcv_irq;
> -	struct work_struct		flr_work;
> +	struct delayed_work		flr_work;
>   	struct amdgpu_mm_table		mm_table;
>   	const struct amdgpu_virt_ops	*ops;
>   	struct amdgpu_vf_error_buffer	vf_errors;
> diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c
> index b81acf59870c..aa5f6d6ea1e3 100644
> --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c
> +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c
> @@ -251,7 +251,7 @@ static int xgpu_ai_set_mailbox_ack_irq(struct amdgpu_device *adev,
>   
>   static void xgpu_ai_mailbox_flr_work(struct work_struct *work)
>   {
> -	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, flr_work);
> +	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, flr_work.work);
>   	struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt);
>   	int timeout = AI_MAILBOX_POLL_FLR_TIMEDOUT;
>   
> @@ -380,7 +380,7 @@ int xgpu_ai_mailbox_get_irq(struct amdgpu_device *adev)
>   		return r;
>   	}
>   
> -	INIT_WORK(&adev->virt.flr_work, xgpu_ai_mailbox_flr_work);
> +	INIT_DELAYED_WORK(&adev->virt.flr_work, xgpu_ai_mailbox_flr_work);
>   
>   	return 0;
>   }
> diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c
> index 22c10b97ea81..dd9f6b6f62f5 100644
> --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c
> +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c
> @@ -275,7 +275,7 @@ static int xgpu_nv_set_mailbox_ack_irq(struct amdgpu_device *adev,
>   
>   static void xgpu_nv_mailbox_flr_work(struct work_struct *work)
>   {
> -	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, flr_work);
> +	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, flr_work.work);
>   	struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt);
>   	int timeout = NV_MAILBOX_POLL_FLR_TIMEDOUT;
>   
> @@ -407,7 +407,7 @@ int xgpu_nv_mailbox_get_irq(struct amdgpu_device *adev)
>   		return r;
>   	}
>   
> -	INIT_WORK(&adev->virt.flr_work, xgpu_nv_mailbox_flr_work);
> +	INIT_DELAYED_WORK(&adev->virt.flr_work, xgpu_nv_mailbox_flr_work);
>   
>   	return 0;
>   }
> diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c
> index 7b63d30b9b79..b11831da1b13 100644
> --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c
> +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c
> @@ -512,7 +512,7 @@ static int xgpu_vi_set_mailbox_ack_irq(struct amdgpu_device *adev,
>   
>   static void xgpu_vi_mailbox_flr_work(struct work_struct *work)
>   {
> -	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, flr_work);
> +	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, flr_work.work);
>   	struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt);
>   
>   	/* wait until RCV_MSG become 3 */
> @@ -610,7 +610,7 @@ int xgpu_vi_mailbox_get_irq(struct amdgpu_device *adev)
>   		return r;
>   	}
>   
> -	INIT_WORK(&adev->virt.flr_work, xgpu_vi_mailbox_flr_work);
> +	INIT_DELAYED_WORK(&adev->virt.flr_work, xgpu_vi_mailbox_flr_work);
>   
>   	return 0;
>   }