Re: [PATCH v4 9/9] mm, sparc: batch arch_unmap_one()

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Medium] sparc: leak of ADI tag_users and tag storage in
  adi_save_tags_range()
--

commit ccbef0704b88d6e8aff2c2f0c82f32d9131dcc3f
Author: Dev Jain <dev.jain@arm.com>

mm, sparc: batch arch_unmap_one()

This commit extends arch_unmap_one() to accept a page count so SPARC can
save ADI tags across a PTE batch during anonymous large-folio unmapping.

> diff --git a/arch/sparc/kernel/adi_64.c b/arch/sparc/kernel/adi_64.c
> index 18036a43cf568..ffa378accefe3 100644
> --- a/arch/sparc/kernel/adi_64.c
> +++ b/arch/sparc/kernel/adi_64.c
[ ... ]
> @@ -394,3 +395,23 @@ int adi_save_tags(struct mm_struct *mm, struct vm_area_struct *vma,
>  
>  	return 0;
>  }
> +
> +int adi_save_tags_range(struct mm_struct *mm, struct vm_area_struct *vma,
> +			unsigned long addr, pte_t oldpte, unsigned long nr)
> +{
> +	unsigned long i;
> +
> +	for (i = 0; i < nr; i++, addr += PAGE_SIZE,
> +	     oldpte = pte_next_pfn(oldpte)) {
> +		int ret;
> +
> +		if (!(pte_val(oldpte) & _PAGE_MCD_4V))
> +			continue;
> +
> +		ret = adi_save_tags(mm, vma, addr, oldpte);
> +		if (ret)
> +			return ret;
> +	}
> +
> +	return 0;
> +}

[Severity: Medium]
Does this leak tag_desc->tag_users references and tag storage when
adi_save_tags() fails partway through a batch?

Each successful adi_save_tags() call invokes alloc_tag_store(), which
allocates a tag_storage_desc_t slot (and its tag_desc->tags buffer) or
increments tag_desc->tag_users on an existing descriptor:

arch/sparc/kernel/adi_64.c:alloc_tag_store() {
    ...
				if ((addr >= tag_desc->start) &&
				    (tag_desc->end >= (addr + PAGE_SIZE - 1))) {
					tag_desc->tag_users++;
					goto out;
				}
    ...
}

If a batch crosses tag_desc->end at page i > 0 and alloc_tag_store() fails
its GFP_NOWAIT allocation for the new range, adi_save_tags_range() returns
an error without unwinding the first i pages.  In mm/rmap.c,
__ttu_anon_swapbacked_folio() then fails and ttu_anon_swapbacked_folio()
restores the present PTEs for the entire sub-batch via set_ptes().  Because
those first i pages are not swapped out, adi_restore_tags() ->
del_tag_store() never runs for them, leaking their tag_users references and
preventing del_tag_store() from freeing tag_desc->tags or releasing the
descriptor slot until process exit.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261115143905.3045084-1-dev.jain@arm.com?part=9