Re: [PATCH v3 34/40] mm: eliminate VMA_SPECIAL_FLAGS usage when hugetlb explicitly tested

On Tue, Sep 29, 2026 at 10:42:00PM -0400, Zi Yan wrote:

On Thu Sep 17, 2026 at 12:22 PM EDT, Lorenzo Stoakes (ARM) wrote:

quoted

It is now an invariant that VMA_IO_BIT is not set except by kernel-owned
mappings, so each existing VMA_SPECIAL_FLAGS test need only test for
VMA_DONTEXPAND_BIT, VMA_PFNMAP_BIT and VMA_MIXEDMAP_BIT.

This is precisely a test for a kernel-owned or fixed mapping.

Update a number of callsites which already explicitly handle hugetlb
mappings.

vma_supports_mlock() and ksm_compatible() also explicitly bail on droppable
mappings - detecting kernel-owned, fixed or droppable mappings is handled
by vma_is_persistent(), so in these cases use this predicate.

should_skip_vma() tests for locked, kernel-owned or fixed memory (having
already excluded hugetlb mappings) so simply test for those there.

No functional change intended.

Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/internal.h | 4 +---
 mm/ksm.c      | 4 +---
 mm/vmscan.c   | 3 ++-
 3 files changed, 4 insertions(+), 7 deletions(-)

<snip>

quoted

diff --git a/mm/vmscan.c b/mm/vmscan.c
index 12ed3388ee41..09ff20c3df5b 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c

@@ -3477,7 +3477,8 @@ static int should_skip_vma(unsigned long start, unsigned long end, struct mm_wal
 	if (!vma_has_recency(vma))
 		return true;

-	if (vma->vm_flags & (VM_LOCKED | VM_SPECIAL))
+	if (vma_test(vma, VMA_LOCKED_BIT) || vma_is_kernel_owned(vma) ||

At this point, mmap_lock is held, there is no mlock WIP, so
VMA_LOCKED_BIT is enough.

Yes indeed :)

quoted

+	    vma_is_fixed_mapping(vma))
 		return true;

 	if (vma == get_gate_vma(vma->vm_mm))

LGTM.

Reviewed-by: Zi Yan <ziy@nvidia.com>

Thanks!

--
Best Regards,
Yan, Zi

--
Cheers, Lorenzo