[PATCH v2 32/32] drm/xe/bo: Update atomic_access attribute on madvise

[Matthew Brost]

On Mon, Apr 07, 2025 at 03:47:19PM +0530, Himal Prasad Ghimiray wrote:
> Update the bo_atomic_access based on user-provided input and determine
> the migration to smem during a CPU fault
> 
> Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray at intel.com>
> ---
>  drivers/gpu/drm/xe/xe_bo.c         | 21 ++++++++++++++---
>  drivers/gpu/drm/xe/xe_vm.c         | 11 +++++++--
>  drivers/gpu/drm/xe/xe_vm_madvise.c | 38 +++++++++++++++++++++++++++---
>  3 files changed, 62 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
> index c337790c81ae..fe78f6da7054 100644
> --- a/drivers/gpu/drm/xe/xe_bo.c
> +++ b/drivers/gpu/drm/xe/xe_bo.c
> @@ -1573,6 +1573,12 @@ static void xe_gem_object_close(struct drm_gem_object *obj,
>  	}
>  }
>  
> +static bool should_migrate_to_smem(struct xe_bo *bo)
> +{
> +	return bo->attr.atomic_access == DRM_XE_VMA_ATOMIC_GLOBAL ||
> +	       bo->attr.atomic_access == DRM_XE_VMA_ATOMIC_CPU;
> +}
> +
>  static vm_fault_t xe_gem_fault(struct vm_fault *vmf)
>  {
>  	struct ttm_buffer_object *tbo = vmf->vma->vm_private_data;
> @@ -1581,7 +1587,7 @@ static vm_fault_t xe_gem_fault(struct vm_fault *vmf)
>  	struct xe_bo *bo = ttm_to_xe_bo(tbo);
>  	bool needs_rpm = bo->flags & XE_BO_FLAG_VRAM_MASK;
>  	vm_fault_t ret;
> -	int idx;
> +	int idx, r = 0;
>  
>  	if (needs_rpm)
>  		xe_pm_runtime_get(xe);
> @@ -1593,8 +1599,17 @@ static vm_fault_t xe_gem_fault(struct vm_fault *vmf)
>  	if (drm_dev_enter(ddev, &idx)) {
>  		trace_xe_bo_cpu_fault(bo);
>  
> -		ret = ttm_bo_vm_fault_reserved(vmf, vmf->vma->vm_page_prot,
> -					       TTM_BO_VM_NUM_PREFAULT);
> +		if (should_migrate_to_smem(bo)) {
> +			r = xe_bo_migrate(bo, XE_PL_TT);
> +			if (r == -EBUSY || r == -ERESTARTSYS || r == -EINTR)
> +				ret = VM_FAULT_NOPAGE;
> +			else if (r)
> +				ret = VM_FAULT_SIGBUS;
> +		}
> +		if (!ret)
> +			ret = ttm_bo_vm_fault_reserved(vmf,
> +						       vmf->vma->vm_page_prot,
> +						       TTM_BO_VM_NUM_PREFAULT);
>  		drm_dev_exit(idx);
>  	} else {
>  		ret = ttm_bo_vm_dummy_page(vmf, vmf->vma->vm_page_prot);
> diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
> index f1d4daf90efe..189e97113dbe 100644
> --- a/drivers/gpu/drm/xe/xe_vm.c
> +++ b/drivers/gpu/drm/xe/xe_vm.c
> @@ -3104,9 +3104,16 @@ static int op_lock_and_prep(struct drm_exec *exec, struct xe_vm *vm,
>  		err = vma_lock_and_validate(exec,
>  					    gpuva_to_vma(op->base.prefetch.va),
>  					    false);
> -		if (!err && !xe_vma_has_no_bo(vma))
> -			err = xe_bo_migrate(xe_vma_bo(vma),
> +		if (!err && !xe_vma_has_no_bo(vma)) {
> +			struct xe_bo *bo = xe_vma_bo(vma);
> +
> +			if (region == 0 && !vm->xe->info.has_device_atomics_on_smem &&
> +			    bo->attr.atomic_access == DRM_XE_VMA_ATOMIC_DEVICE)
> +				region = 1;
> +

So here we disallowing migration to system if atomics don't work there?
Shouldn't we just let the GPU fault and fixup on fault?

> +			err = xe_bo_migrate(bo,
>  					    region_to_mem_type[region]);
> +		}
>  		break;
>  	}
>  	default:
> diff --git a/drivers/gpu/drm/xe/xe_vm_madvise.c b/drivers/gpu/drm/xe/xe_vm_madvise.c
> index f4e0545937b0..bbae2faee603 100644
> --- a/drivers/gpu/drm/xe/xe_vm_madvise.c
> +++ b/drivers/gpu/drm/xe/xe_vm_madvise.c
> @@ -87,16 +87,48 @@ static int madvise_atomic(struct xe_device *xe, struct xe_vm *vm,
>  			  struct xe_vma **vmas, int num_vmas,
>  			  struct drm_xe_madvise_ops ops)
>  {
> -	int i;
> +	struct xe_bo *bo;
> +	int err, i;
>  
>  	xe_assert(vm->xe, ops.type == DRM_XE_VMA_ATTR_ATOMIC);
>  	xe_assert(vm->xe, ops.atomic.val > DRM_XE_VMA_ATOMIC_UNDEFINED &&
>  		  ops.atomic.val <= DRM_XE_VMA_ATOMIC_CPU);
>  	vm_dbg(&xe->drm, "attr_value = %d", ops.atomic.val);
>  
> -	for (i = 0; i < num_vmas; i++)
> +	for (i = 0; i < num_vmas; i++) {
>  		vmas[i]->attr.atomic_access = ops.atomic.val;
> -	/*TODO: handle bo backed vmas */
> +
> +		bo = xe_vma_bo(vmas[i]);
> +		if (!bo)
> +			continue;
> +
> +		if (XE_IOCTL_DBG(xe, ops.atomic.val == DRM_XE_VMA_ATOMIC_CPU &&
> +				 !(bo->flags & XE_BO_FLAG_SYSTEM)))
> +			return -EINVAL;
> +
> +		if (XE_IOCTL_DBG(xe, ops.atomic.val == DRM_XE_VMA_ATOMIC_DEVICE &&
> +				 !(bo->flags & XE_BO_FLAG_VRAM0) &&
> +				     !(bo->flags & XE_BO_FLAG_VRAM1)))
> +			return -EINVAL;

Don't device atomics work if xe->info.has_device_atomics_on_smem is set? 

> +
> +		if (XE_IOCTL_DBG(xe, ops.atomic.val == DRM_XE_VMA_ATOMIC_GLOBAL &&
> +				 (!(bo->flags & XE_BO_FLAG_SYSTEM) ||
> +				      (!(bo->flags & XE_BO_FLAG_VRAM0) &&
> +				      !(bo->flags & XE_BO_FLAG_VRAM1)))))
> +			return -EINVAL;

One concern is all of the above are platform specific checks - e.g. if
we had a device with CXL atomics just work everywhere. I'd at least add
a comment indicating these are platform specific checks.

> +
> +		err = xe_bo_lock(bo, true);
> +		if (err)
> +			return err;
> +		bo->attr.atomic_access = ops.atomic.val;
> +
> +		/* Invalidate cpu page table, so bo can migrate to smem in next access */
> +		if (bo->attr.atomic_access == DRM_XE_VMA_ATOMIC_CPU ||
> +		    bo->attr.atomic_access == DRM_XE_VMA_ATOMIC_GLOBAL)
> +			ttm_bo_unmap_virtual(&bo->ttm);

If alreday in SMEM, you don't need to unmap do you?

Matt

> +
> +		xe_bo_unlock(bo);
> +	}
>  	return 0;
>  }
>  
> -- 
> 2.34.1
>