[PATCH v3] drm/i915: Fix page cleanup on DMA remap failure

[Ville Syrjälä]

On Mon, Jan 27, 2025 at 03:43:32PM -0500, Brian Geffon wrote:
> When converting to folios the cleanup path of shmem_get_pages() was
> missed. When a DMA remap fails and the max segment size is greater than
> PAGE_SIZE it will attempt to retry the remap with a PAGE_SIZEd segment
> size. The cleanup code isn't properly using the folio apis and as a
> result isn't handling compound pages correctly.
> 
> v2 -> v3:
> (Ville) Just use shmem_sg_free_table() as-is in the failure path of
> shmem_get_pages(). shmem_sg_free_table() will clear mapping unevictable
> but it will be reset when it retries in shmem_sg_alloc_table().
> 
> v1 -> v2:
> (Ville) Fixed locations where we were not clearing mapping unevictable.
> 
> Cc: stable at vger.kernel.org
> Cc: Ville Syrjala <ville.syrjala at linux.intel.com>
> Cc: Vidya Srinivas <vidya.srinivas at intel.com>
> Link: https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/13487
> Link: https://lore.kernel.org/lkml/20250116135636.410164-1-bgeffon@google.com/
> Fixes: 0b62af28f249 ("i915: convert shmem_sg_free_table() to use a folio_batch")
> Signed-off-by: Brian Geffon <bgeffon at google.com>
> Suggested-by: Tomasz Figa <tfiga at google.com>

Thanks. Pushed to drm-intel-gt-next.

> ---
>  drivers/gpu/drm/i915/gem/i915_gem_shmem.c | 6 +-----
>  1 file changed, 1 insertion(+), 5 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
> index fe69f2c8527d..ae3343c81a64 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
> @@ -209,8 +209,6 @@ static int shmem_get_pages(struct drm_i915_gem_object *obj)
>  	struct address_space *mapping = obj->base.filp->f_mapping;
>  	unsigned int max_segment = i915_sg_segment_size(i915->drm.dev);
>  	struct sg_table *st;
> -	struct sgt_iter sgt_iter;
> -	struct page *page;
>  	int ret;
>  
>  	/*
> @@ -239,9 +237,7 @@ static int shmem_get_pages(struct drm_i915_gem_object *obj)
>  		 * for PAGE_SIZE chunks instead may be helpful.
>  		 */
>  		if (max_segment > PAGE_SIZE) {
> -			for_each_sgt_page(page, sgt_iter, st)
> -				put_page(page);
> -			sg_free_table(st);
> +			shmem_sg_free_table(st, mapping, false, false);
>  			kfree(st);
>  
>  			max_segment = PAGE_SIZE;
> -- 
> 2.48.1.262.g85cc9f2d1e-goog

-- 
Ville Syrjälä
Intel