mm/memremap: fix spurious large folio warning for FS-DAX
Subsystems (3)
PATCH v2 1/1 mm/memremap: fix spurious large folio warning for FS-DAX
From: John Groves <[email protected]>
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.
---
Change since V1: Deleted the warning altogether, rather than exempting fs-dax.
=== How to reproduce ===
A reproducer is available at:
git clone https://github.com/jagalactic/dax-pmd-test.git
cd xfs-dax-test
make
sudo make test
This will set up XFS on pmem with 2MB stripe alignment and run a test that triggers the warning.
Alternatively, follow the manual steps below.
Prerequisites:
- Linux kernel with FS-DAX support and CONFIG_DEBUG_VM=y
- A pmem device (real or emulated)
- An fsdax namespace configured via ndctl as /dev/pmem0
Manual steps:
1. Create an fsdax namespace (if not already present):
# ndctl create-namespace -m fsdax -e namespace0.0
2. Create XFS with 2MB stripe alignment:
# mkfs.xfs -f -d su=2m,sw=1 /dev/pmem0 # mount -o dax /dev/pmem0 /mnt/pmem
3. Compile and run the reproducer:
# gcc -Wall -O2 -o dax_pmd_test dax_pmd_test.c # ./dax_pmd_test /mnt/pmem/testfile
4. Check dmesg for the warning:
WARNING: mm/memremap.c:431 at free_zone_device_folio+0x.../0x...
Note: The 2MB stripe alignment (-d su=2m,sw=1) is critical. XFS normally allocates blocks at arbitrary offsets, causing PMD faults to fall back to PTE faults. The stripe alignment forces 2MB-aligned allocations, allowing PMD faults to succeed and exposing this bug.
mm/memremap.c | 2 -- 1 file changed, 2 deletions(-)
mm/memremap.c +0 -2
@@ -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)); }/*
base-commit: 8f0b4cce4481fb22653697cced8d0d04027cb1e8
Signature · 1 line
2.49.0
Quoted 90 lines · John Groves
From: John Groves <[email protected]>
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 forMEMORY_DEVICE_FS_DAXwhen 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_DAXfrom the large folio warning, allowing FS-DAX to continue using PMD mappings without triggering false warnings.Fixes d245f9b4ab80 ("mm/zone_device: support large zone device private folios")Signed-off-by John Groves <[email protected]>---
Change since V1: Deleted the warning altogether, rather than exempting fs-dax.
=== How to reproduce ===
A reproducer is available at:
git clone https://github.com/jagalactic/dax-pmd-test.git
cd xfs-dax-test
make
sudo make testThis will set up XFS on pmem with 2MB stripe alignment and run a test that triggers the warning.
Alternatively, follow the manual steps below.
Prerequisites:
- Linux kernel with FS-DAX support andCONFIG_DEBUG_VM=y
- A pmem device (real or emulated)
- An fsdax namespace configured via ndctl as/dev/pmem0Manual steps:
1. Create an fsdax namespace (if not already present):
# ndctl create-namespace -m fsdax -e namespace0.02. Create XFS with 2MB stripe alignment:
# mkfs.xfs -f -d su=2m,sw=1 /dev/pmem0 # mount -o dax /dev/pmem0 /mnt/pmem3. Compile and run the reproducer:
# gcc -Wall -O2 -o dax_pmd_test dax_pmd_test.c # ./dax_pmd_test /mnt/pmem/testfile4. Check dmesg for the warning:
WARNING: mm/memremap.c:431 at free_zone_device_folio+0x.../0x...Note: The 2MB stripe alignment (-d su=2m,sw=1) is critical. XFS normally allocates blocks at arbitrary offsets, causing PMD faults to fall back to PTE faults. The stripe alignment forces 2MB-aligned allocations, allowing PMD faults to succeed and exposing this bug.
mm/memremap.c | 2 -- 1 file changed, 2 deletions(-)mm/memremap.c +0
-0
mm/memremap.c void free_zone_device_folio(struct folio *folio) · line 431
Quoted 7 lines
mm/memremap.c +0
-2@@ -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)); }
LGTM
Signature · 3 lines
Cheers David
John Groves wrote:
[..]
Quoted 3 lines
The fix is to exempt
MEMORY_DEVICE_FS_DAXfrom the large folio warning, allowing FS-DAX to continue using PMD mappings without triggering false warnings.
As you note, this patch no longer exempts MEMORY_DEVICE_FS_DAX explicitly, it just removes the bogus warning, so maybe Andrew can adjust this note on applying?
Quoted 7 lines
Fixes d245f9b4ab80 ("mm/zone_device: support large zone device private folios")Signed-off-by John Groves <[email protected]>---
Change since V1: Deleted the warning altogether, rather than exempting fs-dax.
Quoted 8 lines
=== How to reproduce ===
A reproducer is available at:
git clone https://github.com/jagalactic/dax-pmd-test.git
cd xfs-dax-test
make
sudo make test
Thanks John, outside of the fixup above, this looks good to me.
Now, my first thoughts when seeing this were:
"ooh, I want that test in the regression suite"
...then:
"wait, that sounds exactly like the existing dax.sh test [1]"
[1]: https://github.com/pmem/ndctl/blob/main/test/dax.sh
Alison reports that indeed that existing test triggers the problem which indicates some process problems to solve.
- Folks touching mm/memremap.c (and anything dax related) do not know about / run the regression tests.
- The bespoke nature of the dax testing environment needs some work to get it into a kselftest amenable flow, or otherwise need more automation to run those tests automatically upon seeing those files touched in linux-next so folks see breakage like this earlier.
Quoted 30 lines · John Groves
From: John Groves <[email protected]>
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 forMEMORY_DEVICE_FS_DAXwhen 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.
Quoted 7 lines
The fix is to exempt
MEMORY_DEVICE_FS_DAXfrom the large folio warning, allowing FS-DAX to continue using PMD mappings without triggering false warnings.Fixes d245f9b4ab80 ("mm/zone_device: support large zone device private folios")Signed-off-by John Groves <[email protected]>---