Re: [PATCH 2/6] drm/amdgpu: New VM state for evicted user BOs

[Felix Kuehling <felix.kuehling@amd.com>]

Hi Christian,

I know you have objected to this patch before. I still think this is the 
best solution for what I need. I can talk you through my reasoning by 
email or offline. If I can't convince you, I will need your guidance for 
a better solution.

Thanks,
   Felix


On 2023-11-07 11:58, Felix Kuehling wrote:
> Create a new VM state to track user BOs that are in the system domain.
> In the next patch this will be used do conditionally re-validate them in
> amdgpu_vm_handle_moved.
>
> Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 17 +++++++++++++++++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h |  5 ++++-
>   2 files changed, 21 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> index 1442d97ddd0f..0d685577243c 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> @@ -232,6 +232,22 @@ static void amdgpu_vm_bo_invalidated(struct amdgpu_vm_bo_base *vm_bo)
>   	spin_unlock(&vm_bo->vm->status_lock);
>   }
>   
> +/**
> + * amdgpu_vm_bo_evicted_user - vm_bo is evicted
> + *
> + * @vm_bo: vm_bo which is evicted
> + *
> + * State for BOs used by user mode queues which are not at the location they
> + * should be.
> + */
> +static void amdgpu_vm_bo_evicted_user(struct amdgpu_vm_bo_base *vm_bo)
> +{
> +	vm_bo->moved = true;
> +	spin_lock(&vm_bo->vm->status_lock);
> +	list_move(&vm_bo->vm_status, &vm_bo->vm->evicted_user);
> +	spin_unlock(&vm_bo->vm->status_lock);
> +}
> +
>   /**
>    * amdgpu_vm_bo_relocated - vm_bo is reloacted
>    *
> @@ -2110,6 +2126,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, int32_t xcp
>   	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
>   		vm->reserved_vmid[i] = NULL;
>   	INIT_LIST_HEAD(&vm->evicted);
> +	INIT_LIST_HEAD(&vm->evicted_user);
>   	INIT_LIST_HEAD(&vm->relocated);
>   	INIT_LIST_HEAD(&vm->moved);
>   	INIT_LIST_HEAD(&vm->idle);
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
> index 12cac06fa289..939d0c2219c0 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
> @@ -286,9 +286,12 @@ struct amdgpu_vm {
>   	/* Lock to protect vm_bo add/del/move on all lists of vm */
>   	spinlock_t		status_lock;
>   
> -	/* BOs who needs a validation */
> +	/* Per VM and PT BOs who needs a validation */
>   	struct list_head	evicted;
>   
> +	/* BOs for user mode queues that need a validation */
> +	struct list_head	evicted_user;
> +
>   	/* PT BOs which relocated and their parent need an update */
>   	struct list_head	relocated;
>