[Intel-xe] [PATCH] drm/xe: Use BO's GT to determine dma_offset when programming PTEs

Matthew Brost matthew.brost at intel.com
Thu Mar 23 16:13:28 UTC 2023


On Thu, Mar 23, 2023 at 10:21:12AM +0100, Thomas Hellström wrote:
> Hi.
> 
> Good catch. Some comments below.
> 
> On 3/23/23 09:15, Matthew Brost wrote:
> > Rather than using the passed in GT, use the BO's GT determine dma_offset
> > when programming PTEs as these two GT's could differ (i.e. mapping a BO
> > from a remote GT). The BO's GT is correct GT to use as this where BO
> > resides, while the passed in GT is where the mapping is created.
> > 
> > Signed-off-by: Matthew Brost <matthew.brost at intel.com>
> > ---
> >   drivers/gpu/drm/xe/xe_bo.c | 5 +++++
> >   drivers/gpu/drm/xe/xe_bo.h | 1 +
> >   drivers/gpu/drm/xe/xe_pt.c | 4 +++-
> >   3 files changed, 9 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
> > index 764b3ca9fff7..8beb35aa6f15 100644
> > --- a/drivers/gpu/drm/xe/xe_bo.c
> > +++ b/drivers/gpu/drm/xe/xe_bo.c
> > @@ -79,6 +79,11 @@ mem_type_to_gt(struct xe_device *xe, u32 mem_type)
> >   	return xe_device_get_gt(xe, mem_type == XE_PL_STOLEN ? 0 : (mem_type - XE_PL_VRAM0));
> >   }
> > +struct xe_gt *xe_bo_to_gt(struct xe_bo *bo)
> 
> Kerneldoc.
> 

Will do.

> > +{
> XE_WARN_ON_ONCE(!vram) ?

mem_type_to_gt has a BUG_ON for this.

> > +	return mem_type_to_gt(xe_bo_device(bo), bo->ttm.resource->mem_type);
> > +}
> > +
> >   static void try_add_system(struct xe_bo *bo, struct ttm_place *places,
> >   			   u32 bo_flags, u32 *c)
> >   {
> > diff --git a/drivers/gpu/drm/xe/xe_bo.h b/drivers/gpu/drm/xe/xe_bo.h
> > index f4303810f213..f9abd8036417 100644
> > --- a/drivers/gpu/drm/xe/xe_bo.h
> > +++ b/drivers/gpu/drm/xe/xe_bo.h
> > @@ -105,6 +105,7 @@ struct xe_bo *xe_bo_create_from_data(struct xe_device *xe, struct xe_gt *gt,
> >   int xe_bo_placement_for_flags(struct xe_device *xe, struct xe_bo *bo,
> >   			      u32 bo_flags);
> Newline.

Sure.

Matt

> > +struct xe_gt *xe_bo_to_gt(struct xe_bo *bo);
> >   static inline struct xe_bo *ttm_to_xe_bo(const struct ttm_buffer_object *bo)
> >   {
> > diff --git a/drivers/gpu/drm/xe/xe_pt.c b/drivers/gpu/drm/xe/xe_pt.c
> > index a2e586ebd9a8..6dc31b030687 100644
> > --- a/drivers/gpu/drm/xe/xe_pt.c
> > +++ b/drivers/gpu/drm/xe/xe_pt.c
> > @@ -792,10 +792,12 @@ xe_pt_stage_bind(struct xe_gt *gt, struct xe_vma *vma,
> >   	int ret;
> >   	if (is_vram) {
> > +		struct xe_gt *bo_gt = xe_bo_to_gt(bo);
> > +
> >   		xe_walk.default_pte = GEN12_PPGTT_PTE_LM;
> >   		if (vma && vma->gpuva.flags & XE_VMA_ATOMIC_PTE_BIT)
> >   			xe_walk.default_pte |= GEN12_USM_PPGTT_PTE_AE;
> > -		xe_walk.dma_offset = gt->mem.vram.io_start -
> > +		xe_walk.dma_offset = bo_gt->mem.vram.io_start -
> >   			gt_to_xe(gt)->mem.vram.io_start;
> >   		xe_walk.cache = XE_CACHE_WB;
> >   	} else {
> 
> Thanks,
> 
> Thomas
> 
> 


More information about the Intel-xe mailing list