When migrate_vma_insert_huge_pmd_page() jumps to unlock_abort due
to a PMD check failure, the pgtable allocated earlier via
pte_alloc_one() is never freed, causing a memory leak.

Add a pte_free() call in the unlock_abort error path to release
the pgtable before returning.Also included before goto abort in the
folio check path.

Signed-off-by: Sunny Patel <nueralspacetech@gmail.com>
---

Changes in v2:
  - Added pte_free() before goto abort in the folio_is_zone_device()
    check path. The lock is not taken at this point so goto unlock_abort would be incorrect here.
  - v1 only fixed the unlock_abort path, this version fixes both
    leak locations.

 mm/migrate_device.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/mm/migrate_device.c b/mm/migrate_device.c
index fbfe5715f635..7e132196856b 100644
--- a/mm/migrate_device.c
+++ b/mm/migrate_device.c
@@ -840,6 +840,7 @@ static int migrate_vma_insert_huge_pmd_page(struct migrate_vma *migrate,
 	} else {
 		if (folio_is_zone_device(folio) &&
 		    !folio_is_device_coherent(folio)) {
+			pte_free(vma->vm_mm, pgtable);
 			goto abort;
 		}
 		entry = folio_mk_pmd(folio, vma->vm_page_prot);
@@ -893,6 +894,7 @@ static int migrate_vma_insert_huge_pmd_page(struct migrate_vma *migrate,
 
 unlock_abort:
 	spin_unlock(ptl);
+	pte_free(vma->vm_mm, pgtable);
 abort:
 	for (i = 0; i < HPAGE_PMD_NR; i++)
 		src[i] &= ~MIGRATE_PFN_MIGRATE;
-- 
2.43.0

…
> Add a pte_free() call in the unlock_abort error path to release
> the pgtable before returning.Also included before goto abort in the
> folio check path.

See also:
* https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst?h=v7.0#n145
* https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/stable-kernel-rules.rst?h=v7.0#n34

Regards,
Markus

Sunny Patel <nueralspacetech@gmail.com> writes:

> When migrate_vma_insert_huge_pmd_page() jumps to unlock_abort due
> to a PMD check failure, the pgtable allocated earlier via
> pte_alloc_one() is never freed, causing a memory leak.
>
> Add a pte_free() call in the unlock_abort error path to release
> the pgtable before returning.Also included before goto abort in the
> folio check path.
>
> Signed-off-by: Sunny Patel <nueralspacetech@gmail.com>
> ---
>
> Changes in v2:
>   - Added pte_free() before goto abort in the folio_is_zone_device()
>     check path. The lock is not taken at this point so goto unlock_abort would be incorrect here.
>   - v1 only fixed the unlock_abort path, this version fixes both
>     leak locations.
>
>  mm/migrate_device.c | 2 ++
>  1 file changed, 2 insertions(+)
>
> diff --git a/mm/migrate_device.c b/mm/migrate_device.c
> index fbfe5715f635..7e132196856b 100644
> --- a/mm/migrate_device.c
> +++ b/mm/migrate_device.c
> @@ -840,6 +840,7 @@ static int migrate_vma_insert_huge_pmd_page(struct migrate_vma *migrate,
>  	} else {
>  		if (folio_is_zone_device(folio) &&
>  		    !folio_is_device_coherent(folio)) {
> +			pte_free(vma->vm_mm, pgtable);
>  			goto abort;

goto free_abort?

>  		}
>  		entry = folio_mk_pmd(folio, vma->vm_page_prot);
> @@ -893,6 +894,7 @@ static int migrate_vma_insert_huge_pmd_page(struct migrate_vma *migrate,
>  
>  unlock_abort:
>  	spin_unlock(ptl);

free_abort:

> +	pte_free(vma->vm_mm, pgtable);
>  abort:
>  	for (i = 0; i < HPAGE_PMD_NR; i++)
>  		src[i] &= ~MIGRATE_PFN_MIGRATE;

Although I prefer guard based memory freeing, if we stick with the
original error handling, I suggest following existing style more
closely.

---
Best Regards,
Huang, Ying

On 4/27/26 08:37, Sunny Patel wrote:
> When migrate_vma_insert_huge_pmd_page() jumps to unlock_abort due
> to a PMD check failure, the pgtable allocated earlier via
> pte_alloc_one() is never freed, causing a memory leak.
> 
> Add a pte_free() call in the unlock_abort error path to release
> the pgtable before returning.Also included before goto abort in the
> folio check path.
> 
> Signed-off-by: Sunny Patel <nueralspacetech@gmail.com>
> ---

Fixes: a30b48bf1b24 ("mm/migrate_device: implement THP migration of zone device
pages")
Cc: <stable@vger.kernel.org>

Sashiko moans about a lot of unrelated stuff, one of them fixed by your other
fix, and one of them to be cleaned up by your other patch.

There is another thing about migrate_vma_split_folio() and temporary
split_folio() failures which I think we can ignore.

Acked-by: David Hildenbrand (Arm) <david@kernel.org>

-- 
Cheers,

David