[PATCH v3 3/4] drm/xe: Add xe_bo_tt_has_data helper

Matthew Brost matthew.brost at intel.com
Mon Jul 14 23:37:45 UTC 2025


On Fri, Jul 04, 2025 at 09:58:28AM +0100, Matthew Auld wrote:
> On 23/06/2025 16:19, Matthew Brost wrote:
> > Add xe_bo_tt_has_data helper which determines if a BO has TT data (i.e.,
> > a BO has SRAM or swap backing store).
> > 
> > Signed-off-by: Matthew Brost <matthew.brost at intel.com>
> > ---
> >   drivers/gpu/drm/xe/xe_bo.c | 21 ++++++++++++++++++---
> >   drivers/gpu/drm/xe/xe_bo.h |  2 ++
> >   2 files changed, 20 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
> > index c93381fa4858..7a412121477e 100644
> > --- a/drivers/gpu/drm/xe/xe_bo.c
> > +++ b/drivers/gpu/drm/xe/xe_bo.c
> > @@ -766,6 +766,23 @@ static int xe_bo_move_notify(struct xe_bo *bo,
> >   	return 0;
> >   }
> > +/**
> > + * xe_bo_tt_has_data() - Xe BO has TT data (i.e., Xe BO has SRAM or swap backing
> > + * store)
> > + *
> > + * Helper to determine if a BO has TT data. Useful to determine if copies or
> > + * clears are needed on BO when it is moving.
> > + *
> > + * Return: True if a BO has TT data, False otherwise
> > + */
> > +bool xe_bo_tt_has_data(struct xe_bo *bo)
> > +{
> 
> Maybe add an assert-locked for the bo? I'm pretty sure caller needs the lock
> to check such state?
> 

Yep, should have the BO's dma-resv locks here. Will add.

Matt

> Reviewed-by: Matthew Auld <matthew.auld at intel.com>
> 
> 
> > +	struct ttm_tt *ttm = bo->ttm.ttm;
> > +
> > +	return ttm && (ttm_tt_is_populated(ttm) ||
> > +		       (ttm->page_flags & TTM_TT_FLAG_SWAPPED));
> > +}
> > +
> >   static int xe_bo_move(struct ttm_buffer_object *ttm_bo, bool evict,
> >   		      struct ttm_operation_ctx *ctx,
> >   		      struct ttm_resource *new_mem,
> > @@ -801,9 +818,7 @@ static int xe_bo_move(struct ttm_buffer_object *ttm_bo, bool evict,
> >   		return ret;
> >   	}
> > -	tt_has_data = ttm && (ttm_tt_is_populated(ttm) ||
> > -			      (ttm->page_flags & TTM_TT_FLAG_SWAPPED));
> > -
> > +	tt_has_data = xe_bo_tt_has_data(bo);
> >   	move_lacks_source = !old_mem || (handle_system_ccs ? (!bo->ccs_cleared) :
> >   					 (!mem_type_is_vram(old_mem_type) && !tt_has_data));
> > diff --git a/drivers/gpu/drm/xe/xe_bo.h b/drivers/gpu/drm/xe/xe_bo.h
> > index 02ada1fb8a23..150e98ce6fac 100644
> > --- a/drivers/gpu/drm/xe/xe_bo.h
> > +++ b/drivers/gpu/drm/xe/xe_bo.h
> > @@ -374,6 +374,8 @@ void xe_bo_dev_fini(struct xe_bo_dev *bo_device);
> >   struct sg_table *xe_bo_sg(struct xe_bo *bo);
> > +bool xe_bo_tt_has_data(struct xe_bo *bo);
> > +
> >   /*
> >    * xe_sg_segment_size() - Provides upper limit for sg segment size.
> >    * @dev: device pointer
> 


More information about the Intel-xe mailing list