[PATCH v2 5/5] drm/xe/uapi, drm/xe: Make the PT code handle placement per PTE rather than per vma / range

Thomas Hellström thomas.hellstrom at linux.intel.com
Tue Mar 25 14:23:14 UTC 2025


On Fri, 2025-03-21 at 10:16 -0700, Matthew Brost wrote:
> On Fri, Mar 21, 2025 at 05:34:16PM +0100, Thomas Hellström wrote:
> > With SVM, ranges forwarded to the PT code for binding can, mostly
> > due to races when migrating, point to both VRAM and system /
> > foreign
> > device memory. Make the PT code able to handle that by checking,
> > for each PTE set up, whether it points to local VRAM or to system
> > memory.
> > 
> > The UAPI is changed implicitly in that before this patch,
> > global atomics required a bo with VRAM/System placements. With
> > this patch that is changed to requiring LR mode, and
> > if the required placement is not available upon GPU atomic access
> > pagefault, an error will be generated and the VM banned.
> 
> I think the queue will get banned if I'm reading the code correctly.
> 
> > 
> > v2:
> > - Fix system memory GPU atomic access.
> > 
> > Signed-off-by: Thomas Hellström <thomas.hellstrom at linux.intel.com>
> > ---
> >  drivers/gpu/drm/xe/xe_bo.c |  12 +++--
> >  drivers/gpu/drm/xe/xe_pt.c | 106 ++++++++++++++++-----------------
> > ----
> >  2 files changed, 56 insertions(+), 62 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/xe/xe_bo.c
> > b/drivers/gpu/drm/xe/xe_bo.c
> > index 9d043d2c30fd..3c7c2353d3c8 100644
> > --- a/drivers/gpu/drm/xe/xe_bo.c
> > +++ b/drivers/gpu/drm/xe/xe_bo.c
> > @@ -2116,10 +2116,16 @@ uint64_t vram_region_gpu_offset(struct
> > ttm_resource *res)
> >  {
> >  	struct xe_device *xe = ttm_to_xe_device(res->bo->bdev);
> >  
> > -	if (res->mem_type == XE_PL_STOLEN)
> > +	switch (res->mem_type) {
> > +	case XE_PL_STOLEN:
> >  		return xe_ttm_stolen_gpu_offset(xe);
> > -
> > -	return res_to_mem_region(res)->dpa_base;
> > +	case XE_PL_TT:
> > +	case XE_PL_SYSTEM:
> > +		return 0;
> > +	default:
> > +		return res_to_mem_region(res)->dpa_base;
> > +	}
> > +	return 0;
> >  }
> >  
> >  /**
> > diff --git a/drivers/gpu/drm/xe/xe_pt.c
> > b/drivers/gpu/drm/xe/xe_pt.c
> > index 9e719535a3bb..3ffe135c27f1 100644
> > --- a/drivers/gpu/drm/xe/xe_pt.c
> > +++ b/drivers/gpu/drm/xe/xe_pt.c
> > @@ -278,13 +278,15 @@ struct xe_pt_stage_bind_walk {
> >  	struct xe_vm *vm;
> >  	/** @tile: The tile we're building for. */
> >  	struct xe_tile *tile;
> > -	/** @default_pte: PTE flag only template. No address is
> > associated */
> > -	u64 default_pte;
> > +	/** @default_pte: PTE flag only template for VRAM. No
> > address is associated */
> > +	u64 default_vram_pte;
> > +	/** @default_pte: PTE flag only template for VRAM. No
> > address is associated */
> > +	u64 default_system_pte;
> >  	/** @dma_offset: DMA offset to add to the PTE. */
> >  	u64 dma_offset;
> >  	/**
> >  	 * @needs_64k: This address range enforces 64K alignment
> > and
> > -	 * granularity.
> > +	 * granularity on VRAM.
> >  	 */
> >  	bool needs_64K;
> >  	/**
> > @@ -515,13 +517,16 @@ xe_pt_stage_bind_entry(struct xe_ptw *parent,
> > pgoff_t offset,
> >  	if (level == 0 || xe_pt_hugepte_possible(addr, next,
> > level, xe_walk)) {
> >  		struct xe_res_cursor *curs = xe_walk->curs;
> >  		bool is_null = xe_vma_is_null(xe_walk->vma);
> > +		bool is_vram = is_null ? false :
> > xe_res_is_vram(curs);
> 
> So xe_res_is_vram is still per range / VMA. I assume a follow on will
> change this to be per page? 

For VMAs it's per VMA, but for ranges it's per mapped page AFAICT (at
least that's the intention).

> 
> Patch LGTM though. With that:
> Reviewed-by: Matthew Brost <matthew.brost at intel.com>

Thanks. Does this hold for v3 as well? (I dropped the atomic UAPI
change).

/Thomas


> 
> >  
> >  		XE_WARN_ON(xe_walk->va_curs_start != addr);
> >  
> >  		pte = vm->pt_ops->pte_encode_vma(is_null ? 0 :
> >  						 xe_res_dma(curs)
> > + xe_walk->dma_offset,
> >  						 xe_walk->vma,
> > pat_index, level);
> > -		pte |= xe_walk->default_pte;
> > +		if (!is_null)
> > +			pte |= is_vram ? xe_walk->default_vram_pte
> > :
> > +				xe_walk->default_system_pte;
> >  
> >  		/*
> >  		 * Set the XE_PTE_PS64 hint if possible, otherwise
> > if
> > @@ -531,7 +536,7 @@ xe_pt_stage_bind_entry(struct xe_ptw *parent,
> > pgoff_t offset,
> >  			if (xe_pt_is_pte_ps64K(addr, next,
> > xe_walk)) {
> >  				xe_walk->vma->gpuva.flags |=
> > XE_VMA_PTE_64K;
> >  				pte |= XE_PTE_PS64;
> > -			} else if (XE_WARN_ON(xe_walk->needs_64K))
> > {
> > +			} else if (XE_WARN_ON(xe_walk->needs_64K
> > && is_vram)) {
> >  				return -EINVAL;
> >  			}
> >  		}
> > @@ -603,6 +608,31 @@ static const struct xe_pt_walk_ops
> > xe_pt_stage_bind_ops = {
> >  	.pt_entry = xe_pt_stage_bind_entry,
> >  };
> >  
> > +/* The GPU can always to atomics in VRAM */
> > +static bool xe_atomic_for_vram(struct xe_vm *vm)
> > +{
> > +	return true;
> > +}
> > +
> > +/*
> > + * iGFX always expect to be able to do atomics in system.
> > + *
> > + * For DGFX, 3D clients want to do atomics in system that is
> > + * not coherent with CPU atomics. Compute clients want
> > + * atomics that look coherent with CPU atomics. We
> > + * distinguish the two by checking for lr mode. For
> > + * compute we then disallow atomics in system.
> > + * Compute attempts to perform atomics in system memory would
> > + * then cause an unrecoverable page-fault in preempt-fence
> > + * mode, but in fault mode the data would be migrated to VRAM
> > + * for GPU atomics and to system for CPU atomics.
> > + */
> > +static bool xe_atomic_for_system(struct xe_vm *vm)
> > +{
> > +	return (!IS_DGFX(vm->xe) || !xe_vm_in_lr_mode(vm)) &&
> > +		vm->xe->info.has_device_atomics_on_smem;
> > +}
> > +
> >  /**
> >   * xe_pt_stage_bind() - Build a disconnected page-table tree for a
> > given address
> >   * range.
> > @@ -629,9 +659,8 @@ xe_pt_stage_bind(struct xe_tile *tile, struct
> > xe_vma *vma,
> >  {
> >  	struct xe_device *xe = tile_to_xe(tile);
> >  	struct xe_bo *bo = xe_vma_bo(vma);
> > -	bool is_devmem = !xe_vma_is_userptr(vma) && bo &&
> > -		(xe_bo_is_vram(bo) || xe_bo_is_stolen_devmem(bo));
> >  	struct xe_res_cursor curs;
> > +	struct xe_vm *vm = xe_vma_vm(vma);
> >  	struct xe_pt_stage_bind_walk xe_walk = {
> >  		.base = {
> >  			.ops = &xe_pt_stage_bind_ops,
> > @@ -639,7 +668,7 @@ xe_pt_stage_bind(struct xe_tile *tile, struct
> > xe_vma *vma,
> >  			.max_level = XE_PT_HIGHEST_LEVEL,
> >  			.staging = true,
> >  		},
> > -		.vm = xe_vma_vm(vma),
> > +		.vm = vm,
> >  		.tile = tile,
> >  		.curs = &curs,
> >  		.va_curs_start = range ? range->base.itree.start :
> > @@ -647,26 +676,22 @@ xe_pt_stage_bind(struct xe_tile *tile, struct
> > xe_vma *vma,
> >  		.vma = vma,
> >  		.wupd.entries = entries,
> >  	};
> > -	struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
> > +	struct xe_pt *pt = vm->pt_root[tile->id];
> >  	int ret;
> >  
> >  	if (range) {
> >  		/* Move this entire thing to xe_svm.c? */
> > -		xe_svm_notifier_lock(xe_vma_vm(vma));
> > +		xe_svm_notifier_lock(vm);
> >  		if (!xe_svm_range_pages_valid(range)) {
> >  			xe_svm_range_debug(range, "BIND PREPARE -
> > RETRY");
> > -			xe_svm_notifier_unlock(xe_vma_vm(vma));
> > +			xe_svm_notifier_unlock(vm);
> >  			return -EAGAIN;
> >  		}
> >  		if (xe_svm_range_has_dma_mapping(range)) {
> >  			xe_res_first_dma(range->base.dma_addr, 0,
> >  					 range->base.itree.last +
> > 1 - range->base.itree.start,
> >  					 &curs);
> > -			is_devmem = xe_res_is_vram(&curs);
> > -			if (is_devmem)
> > -				xe_svm_range_debug(range, "BIND
> > PREPARE - DMA VRAM");
> > -			else
> > -				xe_svm_range_debug(range, "BIND
> > PREPARE - DMA");
> > +			xe_svm_range_debug(range, "BIND PREPARE -
> > MIXED");
> >  		} else {
> >  			xe_assert(xe, false);
> >  		}
> > @@ -674,54 +699,17 @@ xe_pt_stage_bind(struct xe_tile *tile, struct
> > xe_vma *vma,
> >  		 * Note, when unlocking the resource cursor dma
> > addresses may become
> >  		 * stale, but the bind will be aborted anyway at
> > commit time.
> >  		 */
> > -		xe_svm_notifier_unlock(xe_vma_vm(vma));
> > +		xe_svm_notifier_unlock(vm);
> >  	}
> >  
> > -	xe_walk.needs_64K = (xe_vma_vm(vma)->flags &
> > XE_VM_FLAG_64K) && is_devmem;
> > -
> > -	/**
> > -	 * Default atomic expectations for different allocation
> > scenarios are as follows:
> > -	 *
> > -	 * 1. Traditional API: When the VM is not in LR mode:
> > -	 *    - Device atomics are expected to function with all
> > allocations.
> > -	 *
> > -	 * 2. Compute/SVM API: When the VM is in LR mode:
> > -	 *    - Device atomics are the default behavior when the
> > bo is placed in a single region.
> > -	 *    - In all other cases device atomics will be disabled
> > with AE=0 until an application
> > -	 *      request differently using a ioctl like madvise.
> > -	 */
> > +	xe_walk.needs_64K = (vm->flags & XE_VM_FLAG_64K);
> >  	if (vma->gpuva.flags & XE_VMA_ATOMIC_PTE_BIT) {
> > -		if (xe_vm_in_lr_mode(xe_vma_vm(vma))) {
> > -			if (bo && xe_bo_has_single_placement(bo))
> > -				xe_walk.default_pte |=
> > XE_USM_PPGTT_PTE_AE;
> > -			/**
> > -			 * If a SMEM+LMEM allocation is backed by
> > SMEM, a device
> > -			 * atomics will cause a gpu page fault and
> > which then
> > -			 * gets migrated to LMEM, bind such
> > allocations with
> > -			 * device atomics enabled.
> > -			 */
> > -			else if (is_devmem)
> > -				xe_walk.default_pte |=
> > XE_USM_PPGTT_PTE_AE;
> > -		} else {
> > -			xe_walk.default_pte |=
> > XE_USM_PPGTT_PTE_AE;
> > -		}
> > -
> > -		/**
> > -		 * Unset AE if the platform(PVC) doesn't support
> > it on an
> > -		 * allocation
> > -		 */
> > -		if (!xe->info.has_device_atomics_on_smem &&
> > !is_devmem)
> > -			xe_walk.default_pte &=
> > ~XE_USM_PPGTT_PTE_AE;
> > +		xe_walk.default_vram_pte = xe_atomic_for_vram(vm)
> > ? XE_USM_PPGTT_PTE_AE : 0;
> > +		xe_walk.default_system_pte =
> > xe_atomic_for_system(vm) ? XE_USM_PPGTT_PTE_AE : 0;
> >  	}
> >  
> > -	if (is_devmem) {
> > -		xe_walk.default_pte |= XE_PPGTT_PTE_DM;
> > -		xe_walk.dma_offset = bo ?
> > vram_region_gpu_offset(bo->ttm.resource) : 0;
> > -	}
> > -
> > -	if (!xe_vma_has_no_bo(vma) && xe_bo_is_stolen(bo))
> > -		xe_walk.dma_offset =
> > xe_ttm_stolen_gpu_offset(xe_bo_device(bo));
> > -
> > +	xe_walk.default_vram_pte |= XE_PPGTT_PTE_DM;
> > +	xe_walk.dma_offset = bo ? vram_region_gpu_offset(bo-
> > >ttm.resource) : 0;
> >  	if (!range)
> >  		xe_bo_assert_held(bo);
> >  
> > -- 
> > 2.48.1
> > 



More information about the Intel-xe mailing list