[PATCH v2] mm/slub: Genericize x86-specific cmpxchg16b comment
From: Tiezhu Yang
Date: Sun Nov 08 2026 - 23:01:39 EST
The kmalloc_nolock() infrastructure relies on slab caches with double-word
atomic cmpxchg capabilities (__CMPXCHG_DOUBLE). The current inline comment
in __kmalloc_nolock_noprof() uses 'cmpxchg16b' as the prerequisite hardware
capability.
However, "cmpxchg16b" is an x86-specific instruction mnemonic (CX16). On
the modern architectures supported by the Linux kernel, this double-width
atomic primitive is satisfied by different hardware implementations:
(1) arm64: satisfies it via LSE atomics ('casp' instruction)
(2) s390: satisfies it via native 'cdsg' instruction
(3) riscv: satisfies it via ZACAS extension ('amocas.q' instruction)
(4) loongarch: satisfies it via SCQ extension (using an 'll.d'/'sc.q' loop)
Update the comment to "double-width atomic cmpxchg", which provides the
following advantages:
(1) Accurately describes the cross-platform hardware invariant.
(2) Covers both 32-bit and 64-bit platforms.
(3) Aligns well with the __CMPXCHG_DOUBLE flag.
(4) Avoids misleading developers of non-x86 architectures.
Signed-off-by: Tiezhu Yang <yangtiezhu@xxxxxxxxxxx>
---
v2:
- Update "128-bit atomic cmpxchg" to "double-width atomic cmpxchg" to
properly cover both 32-bit and 64-bit platforms, suggested by Harry.
- Refine the commit message to format the architecture implementations
and advantages into structured lists for better readability.
mm/slub.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/mm/slub.c b/mm/slub.c
index 45e077c7526c..090302f1daeb 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -5589,8 +5589,8 @@ static void *__kmalloc_nolock_noprof(DECL_TOKEN_PARAMS(size, token), gfp_t gfp_f
if (!(s->flags & __CMPXCHG_DOUBLE) && !kmem_cache_debug(s))
/*
* kmalloc_nolock() is not supported on architectures that
- * don't implement cmpxchg16b and thus need slab_lock()
- * which could be preempted by a nmi.
+ * don't implement double-width atomic cmpxchg and thus need
+ * slab_lock() which could be preempted by a nmi.
* But debug caches don't use that and only rely on
* kmem_cache_node->list_lock, so kmalloc_nolock() can attempt
* to allocate from debug caches by
--
2.42.0