On 29/05/2025 10:02, Anshuman Khandual wrote:

On 5/27/25 13:56, Dev Jain wrote:

quoted

Commit 9c006972c3fe removes the pxd_present() checks because the caller
checks pxd_present(). But, in case of vmap_try_huge_pud(), the caller only
checks pud_present(); pud_free_pmd_page() recurses on each pmd through
pmd_free_pte_page(), wherein the pmd may be none. Thus it is possible to
hit a warning in the latter, since pmd_none => !pmd_table(). Thus, add
a pmd_present() check in pud_free_pmd_page().

This problem was found by code inspection.

Fixes: 9c006972c3fe (arm64: mmu: drop pXd_present() checks from pXd_free_pYd_table())
Cc: <redacted>
Reported-by: Ryan Roberts <ryan.roberts@arm.com> 
Acked-by: David Hildenbrand <redacted>
Signed-off-by: Dev Jain <dev.jain@arm.com>

LGTM!

Reviewed-by: Ryan Roberts <ryan.roberts@arm.com>

quoted

---
This patch is based on 6.15-rc6.

v2->v3:
 - Use pmdp_get()

v1->v2:
 - Enforce check in caller

 arch/arm64/mm/mmu.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
index ea6695d53fb9..5a9bf291c649 100644
--- a/arch/arm64/mm/mmu.c
+++ b/arch/arm64/mm/mmu.c

@@ -1286,7 +1286,8 @@ int pud_free_pmd_page(pud_t *pudp, unsigned long addr)
 	next = addr;
 	end = addr + PUD_SIZE;
 	do {
-		pmd_free_pte_page(pmdp, next);
+		if (pmd_present(pmdp_get(pmdp)))

This code path is only called for the kernel mapping. Hence should
pmd_valid() be used instead of pmd_present() which also checks for
present invalid scenarios as well ? 

I think a similar question came up in a previous round, where we concluded that
it's better to be consistent with what vmalloc is already doing. So personally
I'd leave it as pmd_present():

	if (pmd_present(*pmd) && !pmd_free_pte_page(pmd, addr))
		return 0;

quoted

+			pmd_free_pte_page(pmdp, next);
 	} while (pmdp++, next += PMD_SIZE, next != end);
 
 	pud_clear(pudp);