Re: [RFC PATCH V7 05/10] drm/xe: Extend BO purge to handle vram pages as well

["Yadav, Arvind" <arvind.yadav@intel.com>]

[-- Attachment #1: Type: text/plain, Size: 2902 bytes --]


On 30-04-2026 09:14, Matthew Brost wrote:
> On Thu, Apr 16, 2026 at 01:19:54PM +0530, Tejas Upadhyay wrote:
>> Recent driver update introduce support for purgeable buffer
>> objects (BOs), extending the API to include VRAM pages to
>> better manage memory pressure and enable memory offlining.
>>
>> Signed-off-by: Tejas Upadhyay<tejas.upadhyay@intel.com>
>> ---
>>   drivers/gpu/drm/xe/xe_bo.c | 5 +----
>>   drivers/gpu/drm/xe/xe_bo.h | 1 +
>>   2 files changed, 2 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
>> index 5ce60d161e09..04d3b25c7c8e 100644
>> --- a/drivers/gpu/drm/xe/xe_bo.c
>> +++ b/drivers/gpu/drm/xe/xe_bo.c
>> @@ -903,7 +903,7 @@ void xe_bo_set_purgeable_state(struct xe_bo *bo,
>>    *
>>    * Return: 0 on success, negative error code on failure
>>    */
>> -static int xe_ttm_bo_purge(struct ttm_buffer_object *ttm_bo, struct ttm_operation_ctx *ctx)
>> +int xe_ttm_bo_purge(struct ttm_buffer_object *ttm_bo, struct ttm_operation_ctx *ctx)
>>   {
>>   	struct xe_bo *bo = ttm_to_xe_bo(ttm_bo);
>>   	struct ttm_placement place = {};
>> @@ -911,9 +911,6 @@ static int xe_ttm_bo_purge(struct ttm_buffer_object *ttm_bo, struct ttm_operatio
>>   
>>   	xe_bo_assert_held(bo);
>>   
>> -	if (!ttm_bo->ttm)
>> -		return 0;
>> -
> I think blindly deleting this is fine but I'd run this by Arvind to make
> sure in normal purge flow this won't do anything bad but afiak in normal
> purge flows ttm_bo->ttm will never be NULL.


The purge paths (|xe_bo_move(evict=true, DONTNEED)| and |swap_notify|), 
|ttm_bo->ttm| is always non‑NULL. TTM allocates |ttm_tt| before invoking 
|->move| on a populated BO. So removing this check is safe.

~Arvind


>
> If not then adjust the existing call...
>
> @@ -978,7 +978,8 @@ static int xe_bo_move(struct ttm_buffer_object *ttm_bo, bool evict,
>           * The move_notify callback will handle invalidation asynchronously.
>           */
>          if (evict && xe_bo_madv_is_dontneed(bo)) {
> -               ret = xe_ttm_bo_purge(ttm_bo, ctx);
> +               if (ttm_bo->ttm)
> +                       ret = xe_ttm_bo_purge(ttm_bo, ctx);
>                  if (ret)
>                          return ret;
>
> Matt
>
>>   	if (!xe_bo_madv_is_dontneed(bo))
>>   		return 0;
>>   
>> diff --git a/drivers/gpu/drm/xe/xe_bo.h b/drivers/gpu/drm/xe/xe_bo.h
>> index 68dea7d25a6b..9f55b3589caf 100644
>> --- a/drivers/gpu/drm/xe/xe_bo.h
>> +++ b/drivers/gpu/drm/xe/xe_bo.h
>> @@ -500,6 +500,7 @@ struct xe_bo_shrink_flags {
>>   long xe_bo_shrink(struct ttm_operation_ctx *ctx, struct ttm_buffer_object *bo,
>>   		  const struct xe_bo_shrink_flags flags,
>>   		  unsigned long *scanned);
>> +int xe_ttm_bo_purge(struct ttm_buffer_object *ttm_bo, struct ttm_operation_ctx *ctx);
>>   
>>   /**
>>    * xe_bo_is_mem_type - Whether the bo currently resides in the given
>> -- 
>> 2.52.0
>>

[-- Attachment #2: Type: text/html, Size: 3703 bytes --]