Re: [PATCH v7 2/2] drm/xe/shrinker: Take a runtime PM ref before shrinking non-system memory

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Shrinker scan aborts entirely when encountering a single unsuitable buffer object.
--

commit 90a5cf734495d4bcf94edcf044500adfd18b2a39
Author: Shuicheng Lin <shuicheng.lin@intel.com>

drm/xe/shrinker: Take a runtime PM ref before shrinking non-system memory

This commit updates the xe shrinker to take a runtime PM reference before
walking non-system memory LRUs. It reuses the existing acquire path to resume
the device directly or queue a worker, and stops the scan early once the
target is met.

> diff --git a/drivers/gpu/drm/xe/xe_shrinker.c b/drivers/gpu/drm/xe/xe_shrinker.c
> index 89445cd202380..deb4378c1ec15 100644
> --- a/drivers/gpu/drm/xe/xe_shrinker.c
> +++ b/drivers/gpu/drm/xe/xe_shrinker.c
[ ... ]
> @@ -75,23 +101,35 @@ static int __xe_shrinker_walk(struct xe_device *xe,
>  		if (!man || !man->use_tt)
>  			continue;
>  
> +		if (mem_type != XE_PL_SYSTEM && !rpm &&
> +		    xe_device_is_l2_flush_optimized(xe)) {
> +			if (!__xe_shrinker_runtime_pm_get(shrinker))
> +				break;
> +			rpm = true;
> +		}
> +
>  		ttm_bo_lru_for_each_reserved_guarded(&curs, man, &arg, ttm_bo) {
>  			if (!ttm_bo_shrink_suitable(ttm_bo, ctx))
>  				continue;
>  
>  			lret = xe_bo_shrink(ctx, ttm_bo, flags, scanned);
> -			if (lret < 0)
> -				return lret;
> +			if (lret < 0) {
> +				ret = lret;
> +				goto out;
> +			}

[Severity: High]
This is a pre-existing issue, but does this inappropriately abort the entire
LRU traversal rather than skipping to the next buffer object when a non-fatal
error code is returned?

During the __xe_shrinker_walk() LRU iteration in
drivers/gpu/drm/xe/xe_shrinker.c, if xe_bo_shrink() returns -EBUSY, the loop
treats this as a fatal error and immediately aborts the entire scan via the
goto out branch.

This can happen frequently, such as when xe_bo_eviction_valuable() evaluates
to false:

drivers/gpu/drm/xe/xe_bo.c:xe_bo_shrink() {
	...
	if (!xe_bo_eviction_valuable(bo, &place))
		return -EBUSY;
	...
}

Will the memory shrinker fail to reclaim pages if it encounters a buffer
object that cannot be evicted, potentially leading to system OOM because
reclaim progress is permanently blocked by the first unsuitable BO?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260909162102.1097006-1-shuicheng.lin@intel.com?part=2