Skip to contentSkip to conversation

mm: avoid KCSAN false positive in memdesc_nid()

PATCH In review Hui Zhu 3 participants 5 messages Jun 23 – Jun 25

Subsystems (2)
lore
Share & tools
lore

Hui Zhu

author
#0 lore raw

PATCH v2 1/1 mm: avoid KCSAN false positive in memdesc_nid()

From: Hui Zhu <[email protected]>

KCSAN reports a data race between page_to_nid()/folio_nid() 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 occupies a fixed bit-range of page->flags that is set once at page init and never modified afterwards, so it can never overlap with the low PG_locked/PG_waiters bits touched by the folio lock path.

Use ASSERT_EXCLUSIVE_BITS() in memdesc_nid() to scope the exemption to just the node-id bits, consistent with how memdesc_zonenum() already handles the same class of race for the zone-id bits.

---
Changelog:
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 | 1 +
 1 file changed, 1 insertion(+)
include/linux/mm.h +1 -0
index 485df9c2dbdd..7518d6364a00 100644
@@ -2290,6 +2290,7 @@ int memdesc_nid(memdesc_flags_t mdf);
 #else
 static inline int memdesc_nid(memdesc_flags_t mdf)
 {
+	ASSERT_EXCLUSIVE_BITS(mdf.f, NODES_MASK << NODES_PGSHIFT);
 	return (mdf.f >> NODES_PGSHIFT) & NODES_MASK;
 }
 #endif
Signature · 1 line
2.43.0
maintainer Acked-by on v2 1/1
#1 lore raw
Quoted 13 lines · Hui Zhu

From: Hui Zhu <[email protected]>

KCSAN reports a data race between page_to_nid()/folio_nid() 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 occupies a fixed bit-range of page->flags that is set once at page init and never modified afterwards, so it can never overlap with the low PG_locked/PG_waiters bits touched by the folio lock path.

Quoted 6 lines

Use ASSERT_EXCLUSIVE_BITS() in memdesc_nid() to scope the exemption to just the node-id bits, consistent with how memdesc_zonenum() already handles the same class of race for the zone-id bits.

---

Signature · 3 lines
Cheers,

David
maintainer
#2 lore raw
In reply to Hui Zhu
Quoted 22 lines · On Tue, 23 Jun 2026 16:44:32 +0800 Hui Zhu <[email protected]> wrote:

From: Hui Zhu <[email protected]>

KCSAN reports a data race between page_to_nid()/folio_nid() 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 occupies a fixed bit-range of page->flags that is set once at page init and never modified afterwards, so it can never overlap with the low PG_locked/PG_waiters bits touched by the folio lock path.

Use ASSERT_EXCLUSIVE_BITS() in memdesc_nid() to scope the exemption to just the node-id bits, consistent with how memdesc_zonenum() already handles the same class of race for the zone-id bits.

...

include/linux/mm.h int memdesc_nid(memdesc_flags_t mdf); · line 2296

Quoted 7 lines
include/linux/mm.h +1 -0
@@ -2290,6 +2290,7 @@ int memdesc_nid(memdesc_flags_t mdf);
 #else
 static inline int memdesc_nid(memdesc_flags_t mdf)
 {
+	ASSERT_EXCLUSIVE_BITS(mdf.f, NODES_MASK << NODES_PGSHIFT);
 	return (mdf.f >> NODES_PGSHIFT) & NODES_MASK;
 }
 #endif

It seems weird to be doing this against a local variable within a random function, seemingly unrelated to the problematic functions which you've identified.

Seems that it fooled Sashiko:

	https://sashiko.dev/#/patchset/[email protected]

I'm wondering what the heck is going on here?

author
#3 lore raw
Quoted 34 lines · On Tue, 23 Jun 2026 16:44:32 +0800 Hui Zhu <[email protected]> wrote:
Quoted 21 lines

From: Hui Zhu <[email protected]>

KCSAN reports a data race between page_to_nid()/folio_nid() 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 occupies a fixed bit-range of page->flags that is set once at page init and never modified afterwards, so it can never overlap with the low PG_locked/PG_waiters bits touched by the folio lock path.

Use ASSERT_EXCLUSIVE_BITS() in memdesc_nid() to scope the exemption to just the node-id bits, consistent with how memdesc_zonenum() already handles the same class of race for the zone-id bits.

...

Quoted 11 lines

--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2290,6 +2290,7 @@ int memdesc_nid(memdesc_flags_t mdf); #else
static inline int memdesc_nid(memdesc_flags_t mdf) {
+ ASSERT_EXCLUSIVE_BITS(mdf.f, NODES_MASK << NODES_PGSHIFT); return (mdf.f >> NODES_PGSHIFT) & NODES_MASK;
}
#endif

Quoted 9 lines

It seems weird to be doing this against a local variable within a random function, seemingly unrelated to the problematic functions which you've identified.

Seems that it fooled Sashiko:
https://sashiko.dev/#/patchset/[email protected]

I'm wondering what the heck is going on here?

Hi Andrew,

Good catch. ASSERT_EXCLUSIVE_BITS(mdf.f, ...) is checking a by-value copy of the flags word inside memdesc_nid(), not the actual shared page->flags/folio->flags being modified by folio_trylock(). Whatever made it appear to suppress the KCSAN report is likely an artifact of inlining/codegen (kcsan_atomic_next() happening to land on the real load after inlining), not a principled fix - so Sashiko's pass is not reassuring here.

I'll move the assertion to where the real dereference happens (at the page_to_nid()/folio_nid() call sites) instead of inside the by-value helper. This probably also applies to the existing memdesc_zonenum() pattern - is that one actually verified to work, or does it have the same issue?

Best,
Hui

maintainer
#4 lore raw
Quoted 7 lines · On Thu, 25 Jun 2026 01:32:49 +0000 "Hui Zhu" <[email protected]> wrote:

Good catch. ASSERT_EXCLUSIVE_BITS(mdf.f, ...) is checking a by-value copy of the flags word inside memdesc_nid(), not the actual shared page->flags/folio->flags being modified by folio_trylock(). Whatever made it appear to suppress the KCSAN report is likely an artifact of inlining/codegen (kcsan_atomic_next() happening to land on the real load after inlining), not a principled fix - so Sashiko's pass is not reassuring here.

Yeah, I was wondering if the inlining accidentally gave the macro the correct thing. Which seems wrong - an inlined function should treat an incoming arg purely as a local thing. Maybe we fooled the compiler.

Quoted 5 lines

I'll move the assertion to where the real dereference happens (at the page_to_nid()/folio_nid() call sites) instead of inside the by-value helper. This probably also applies to the existing memdesc_zonenum() pattern - is that one actually verified to work, or does it have the same issue?

I assume the memdesc_zonenum() code worked, for the same (poorly understood) reason as did your patch.

Yes, moving this into the sites where we officially have access to the shared storage seems the right approach.