]> git.ipfire.org Git - thirdparty/kernel/stable.git/commitdiff
mm/memremap: fix spurious large folio warning for FS-DAX
authorJohn Groves <John@Groves.net>
Fri, 19 Dec 2025 12:37:17 +0000 (06:37 -0600)
committerAndrew Morton <akpm@linux-foundation.org>
Tue, 23 Dec 2025 19:23:16 +0000 (11:23 -0800)
This patch addresses a warning that I discovered while working on famfs,
which is an fs-dax file system that virtually always does PMD faults (next
famfs patch series coming after the holidays).

However, XFS also does PMD faults in fs-dax mode, and it also triggers the
warning.  It takes some effort to get XFS to do a PMD fault, but
instructions to reproduce it are below.

The VM_WARN_ON_ONCE(folio_test_large(folio)) check in
free_zone_device_folio() incorrectly triggers for MEMORY_DEVICE_FS_DAX
when PMD (2MB) mappings are used.

FS-DAX legitimately creates large file-backed folios when handling PMD
faults.  This is a core feature of FS-DAX that provides significant
performance benefits by mapping 2MB regions directly to persistent memory.
When these mappings are unmapped, the large folios are freed through
free_zone_device_folio(), which triggers the spurious warning.

The warning was introduced by commit that added support for large zone
device private folios.  However, that commit did not account for FS-DAX
file-backed folios, which have always supported large (PMD-sized)
mappings.

The check distinguishes between anonymous folios (which clear
AnonExclusive flags for each sub-page) and file-backed folios.  For
file-backed folios, it assumes large folios are unexpected - but this
assumption is incorrect for FS-DAX.

The fix is to exempt MEMORY_DEVICE_FS_DAX from the large folio warning,
allowing FS-DAX to continue using PMD mappings without triggering false
warnings.

Link: https://lkml.kernel.org/r/20251219123717.39330-1-john@groves.net
Fixes: d245f9b4ab80 ("mm/zone_device: support large zone device private folios")
Signed-off-by: John Groves <john@groves.net>
Acked-by: David Hildenbrand (Red Hat) <david@kernel.org>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Tested-by: Alison Schofield <alison.schofield@intel.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: "Darrick J. Wong" <djwong@kernel.org>
Cc: Gregory Price <gourry@gourry.net>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
mm/memremap.c

index 4c2e0d68eb279871b782f5aa60f2aba8c42eab8a..63c6ab4fdf082c95c2a0c057afb0b6f0f06b5a14 100644 (file)
@@ -427,8 +427,6 @@ void free_zone_device_folio(struct folio *folio)
        if (folio_test_anon(folio)) {
                for (i = 0; i < nr; i++)
                        __ClearPageAnonExclusive(folio_page(folio, i));
-       } else {
-               VM_WARN_ON_ONCE(folio_test_large(folio));
        }
 
        /*