mm: assert exclusive nid/zonenum bits at the page/folio access sites
Subsystems (3)
PATCH v6 1/1 mm: assert exclusive nid/zonenum bits at the page/folio access sites
From: Hui Zhu <[email protected]>
KCSAN reports a data race between page_to_nid()/folio_pgdat() reading page->flags and folio_trylock()/folio_lock() concurrently doing test_and_set_bit_lock(PG_locked, ...) on the same word, e.g.:
BUG: KCSAN: data-race in __lruvec_stat_mod_folio / shmem_get_folio_gfp
The node id and zone id occupy fixed bit-ranges of page->flags that are set once at page init and never modified afterwards, so they can never overlap with the low PG_locked/PG_waiters bits touched by the folio lock path.
ASSERT_EXCLUSIVE_BITS(mdf.f, ...) inside memdesc_nid()/memdesc_zonenum() used to check a by-value copy of the flags word, not the actual shared page->flags/folio->flags being modified concurrently, so it didn't reliably assert anything about the real race.
For zonenum, move the assertion out of memdesc_zonenum() into page_zonenum() and folio_zonenum(), where flags is dereferenced directly from the page/folio.
For nid, turn memdesc_nid() into a macro instead, so the mdf argument is expanded as the caller's own flags expression (PF_POISONED_CHECK(page)->flags or folio->flags) rather than copied into a function parameter, letting ASSERT_EXCLUSIVE_BITS() check the real page->flags/folio->flags directly.
On CONFIG_NUMA=n, NODES_MASK is 0 and the old memdesc_nid() body folded to a constant, so page->flags/folio->flags was never actually read. ASSERT_EXCLUSIVE_BITS() is a real runtime check that can't be folded away, so doing it unconditionally would add a pointless read of page->flags/folio->flags and a check that can never fire. Keep page_to_nid()/folio_nid() as plain "return 0" static inline stubs under CONFIG_NUMA=n instead.
---
Changelog:
v6:
According to the comments of David, turn memdesc_nid() from a static inline function into a macro so ASSERT_EXCLUSIVE_BITS() can check the caller's page->flags/folio->flags directly.
v5:
According to the comments of Sashiko, guard the ASSERT_EXCLUSIVE_BITS() calls with #ifndef NODE_NOT_IN_PAGE_FLAGS (for nid) and #if ZONES_WIDTH != 0 (for zonenum).
According to the comments of David, avoid calling PF_POISONED_CHECK(page) twice in page_to_nid().
According to the warning of lkp, switch the CONFIG_NUMA=n page_to_nid()/folio_nid() stubs from macros to static inline functions. v4:
According to the comments of Andrew and Sashiko, set page_to_nid()/folio_nid() as static inline stubs returning 0 under CONFIG_NUMA=n.
v3:
According to the comments of Andrew and Sashiko, move ASSERT_EXCLUSIVE_BITS out of memdesc_nid()/memdesc_zonenum() into the page/folio call sites.
v2:
According to the comments of David, remove useless comments and use ASSERT_EXCLUSIVE_BITS() in memdesc_nid() instead of data_race() in page_to_nid().
include/linux/mm.h | 21 +++++++++++++++++---- include/linux/mmzone.h | 7 ++++++- 2 files changed, 23 insertions(+), 5 deletions(-)
include/linux/mm.h +17 -4
@@ -2288,12 +2288,14 @@ static inline int page_zone_id(struct page *page)#ifdef NODE_NOT_IN_PAGE_FLAGS int memdesc_nid (memdesc_flags_t mdf);#else -static inline int memdesc_nid (memdesc_flags_t mdf) -{ -return (mdf.f >>NODES_PGSHIFT ) &NODES_MASK ; -} +#define memdesc_nid (mdf) \ +({ \ + ASSERT_EXCLUSIVE_BITS(mdf.f, NODES_MASK << NODES_PGSHIFT); \ + (int)((mdf.f >> NODES_PGSHIFT) & NODES_MASK); \ +})#endif +#ifdef CONFIG_NUMA static inline int page_to_nid (const struct page *page) {return memdesc_nid (PF_POISONED_CHECK (page)->flags);
@@ -2303,6 +2305,17 @@ static inline int folio_nid(const struct folio *folio) {return memdesc_nid (folio->flags); } +#else +static inline int page_to_nid (const struct page *page) +{ +return 0 ; +} + +static inline int folio_nid (const struct folio *folio) +{ +return 0 ; +} +#endif #ifdef CONFIG_NUMA_BALANCING /* page access time bits needs to hold at least 4 seconds */
include/linux/mmzone.h +6 -1
@@ -1274,17 +1274,22 @@ static inline bool zone_is_empty(const struct zone *zone)static inline enum zone_type memdesc_zonenum (memdesc_flags_t flags) { -ASSERT_EXCLUSIVE_BITS (flags.f,ZONES_MASK <<ZONES_PGSHIFT );return (flags.f >>ZONES_PGSHIFT ) &ZONES_MASK ; }static inline enum zone_type page_zonenum (const struct page *page) { +#if ZONES_WIDTH !=0 +ASSERT_EXCLUSIVE_BITS (page->flags,ZONES_MASK <<ZONES_PGSHIFT ); +#endif return memdesc_zonenum (page->flags); }static inline enum zone_type folio_zonenum (const struct folio *folio) { +#if ZONES_WIDTH !=0 +ASSERT_EXCLUSIVE_BITS (folio->flags,ZONES_MASK <<ZONES_PGSHIFT ); +#endif return memdesc_zonenum (folio->flags); }
Signature · 1 line
2.43.0
Andrew Morton
maintainerQuoted 28 lines · On Fri, 26 Jun 2026 10:06:29 +0800 Hui Zhu <[email protected]> wrote:
From: Hui Zhu <[email protected]>
KCSAN reports a data race between page_to_nid()/folio_pgdat() reading
page->flagsand folio_trylock()/folio_lock() concurrently doingtest_and_set_bit_lock(PG_locked, ...)on the same word, e.g.:BUG: KCSAN: data-race in __lruvec_stat_mod_folio / shmem_get_folio_gfpThe node id and zone id occupy fixed bit-ranges of
page->flagsthat are set once at page init and never modified afterwards, so they can never overlap with the low PG_locked/PG_waiters bits touched by the folio lock path.
ASSERT_EXCLUSIVE_BITS(mdf.f, ...)inside memdesc_nid()/memdesc_zonenum() used to check a by-value copy of the flags word, not the actual shared page->flags/folio->flags being modified concurrently, so it didn't reliably assert anything about the real race.For zonenum, move the assertion out of
memdesc_zonenum()intopage_zonenum()andfolio_zonenum(), where flags is dereferenced directly from the page/folio.For nid, turn
memdesc_nid()into a macro instead, so the mdf argument is expanded as the caller's own flags expression (PF_POISONED_CHECK(page)->flags orfolio->flags) rather than copied into a function parameter, lettingASSERT_EXCLUSIVE_BITS()check the real page->flags/folio->flags directly.
Quoted 8 lines
On
CONFIG_NUMA=n,NODES_MASKis 0 and the oldmemdesc_nid()body folded to a constant, so page->flags/folio->flags was never actually read.ASSERT_EXCLUSIVE_BITS()is a real runtime check that can't be folded away, so doing it unconditionally would add a pointless read of page->flags/folio->flags and a check that can never fire. Keep page_to_nid()/folio_nid() as plain "return 0" static inline stubs underCONFIG_NUMA=ninstead.
Thanks for persisting with this.
ASSERT_EXCLUSIVE_BITS is defined in kcsan-checks.h and I'm not immediately seeing a reliable route by which mm.h and mmzone.h get to include that file. I don't think kasan.h gives that to us, surprisingly.
Unless I've missed it, we're presently getting kcsan-checks.h thanks to good luck. Let's not rely upon good luck ;)
Also, Sashiko is saying stuff again. Does it look legitimate?
https://sashiko.dev/#/patchset/[email protected]
David Hildenbrand (Arm)
maintainerQuoted 46 lines · Andrew Morton
Quoted 28 lines · On Fri, 26 Jun 2026 10:06:29 +0800 Hui Zhu <[email protected]> wrote:
From: Hui Zhu <[email protected]>
KCSAN reports a data race between page_to_nid()/folio_pgdat() reading
page->flagsand folio_trylock()/folio_lock() concurrently doingtest_and_set_bit_lock(PG_locked, ...)on the same word, e.g.:BUG: KCSAN: data-race in __lruvec_stat_mod_folio / shmem_get_folio_gfpThe node id and zone id occupy fixed bit-ranges of
page->flagsthat are set once at page init and never modified afterwards, so they can never overlap with the low PG_locked/PG_waiters bits touched by the folio lock path.
ASSERT_EXCLUSIVE_BITS(mdf.f, ...)inside memdesc_nid()/memdesc_zonenum() used to check a by-value copy of the flags word, not the actual shared page->flags/folio->flags being modified concurrently, so it didn't reliably assert anything about the real race.For zonenum, move the assertion out of
memdesc_zonenum()intopage_zonenum()andfolio_zonenum(), where flags is dereferenced directly from the page/folio.For nid, turn
memdesc_nid()into a macro instead, so the mdf argument is expanded as the caller's own flags expression (PF_POISONED_CHECK(page)->flags orfolio->flags) rather than copied into a function parameter, lettingASSERT_EXCLUSIVE_BITS()check the real page->flags/folio->flags directly.Quoted 8 lines
On
CONFIG_NUMA=n,NODES_MASKis 0 and the oldmemdesc_nid()body folded to a constant, so page->flags/folio->flags was never actually read.ASSERT_EXCLUSIVE_BITS()is a real runtime check that can't be folded away, so doing it unconditionally would add a pointless read of page->flags/folio->flags and a check that can never fire. Keep page_to_nid()/folio_nid() as plain "return 0" static inline stubs underCONFIG_NUMA=ninstead.Thanks for persisting with this.
ASSERT_EXCLUSIVE_BITSis defined inkcsan-checks.hand I'm not immediately seeing a reliable route by whichmm.handmmzone.hget to include that file. I don't thinkkasan.hgives that to us, surprisingly.
Quoted 7 lines
Unless I've missed it, we're presently getting
kcsan-checks.hthanks to good luck. Let's not rely upon good luck ;)Also, Sashiko is saying stuff again. Does it look legitimate?
https://sashiko.dev/#/patchset/[email protected]
Let's see whether pass-by-reference as suggested in reply to v7 sorts this out.
Signature · 3 lines
Cheers, David