Re: [PATCH RESEND v5 3/4] mm/mlock: check for THP missing the mlock in try_to_unmap_one()

From: Baolin Wang
Date: Tue May 14 2024 - 02:41:51 EST




On 2024/5/13 15:47, Lance Yang wrote:
The TTU_SPLIT_HUGE_PMD will no longer perform immediately, so we might
encounter a PMD-mapped THP missing the mlock in the VM_LOCKED range
during the pagewalk. It's likely necessary to mlock this THP to prevent
it from being picked up during page reclaim.

Suggested-by: Baolin Wang <baolin.wang@xxxxxxxxxxxxxxxxx>
Signed-off-by: Lance Yang <ioworker0@xxxxxxxxx>

IMO, this patch should be merged into patch 2, otherwise patch 2 is buggy. Quote the description in the 'submission patches.rst' document:

"When dividing your change into a series of patches, take special care to ensure that the kernel builds and runs properly after each patch in the series. Developers using ``git bisect`` to track down a problem can end up splitting your patch series at any point; they will not thank you if you introduce bugs in the middle."

---
mm/rmap.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/mm/rmap.c b/mm/rmap.c
index 4c4d14325f2e..08a93347f283 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -1671,7 +1671,8 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
if (!(flags & TTU_IGNORE_MLOCK) &&
(vma->vm_flags & VM_LOCKED)) {
/* Restore the mlock which got missed */
- if (!folio_test_large(folio))
+ if (!folio_test_large(folio) ||
+ (!pvmw.pte && (flags & TTU_SPLIT_HUGE_PMD)))
mlock_vma_folio(folio, vma);
goto walk_done_err;
}