[PATCH v4 0/2] arm64: Fixes for __READ_ONCE() with CONFIG_LTO=y

Marco Elver posted 2 patches 1 month, 2 weeks ago
arch/arm64/include/asm/rwonce.h | 24 +++++++++++++++++++-----
1 file changed, 19 insertions(+), 5 deletions(-)
[PATCH v4 0/2] arm64: Fixes for __READ_ONCE() with CONFIG_LTO=y
Posted by Marco Elver 1 month, 2 weeks ago
While investigating a Clang Context Analysis [1] false positive [2], I
started to dig deeper into arm64's __READ_ONCE() implementation with
LTO. That rabbit hole led me to find one critical bug with the current
implementation (fixed already in [3]), and subtle improvements that then
enabled me to fix the original false positive.

Patch 1 refactors the macro to use a different way of getting an
unqualified type and eliminates the ternary conditional.

Building on the refactor, patch 2 fixes the context analysis false
positive, by helping its alias analysis "see through" the __READ_ONCE
despite the inline asm.

## Note on Alternative for Patch 2

An alternative considered for the Context Analysis fix was introducing a
helper function to redirect the pointer alias; specifically passing a
pointer to const-pointer does not invalidate an alias either (casting
away the const is a deliberate escape hatch, albeit somewhat unusual
looking). This approach was slightly more verbose, so the simpler
approach was chosen for now. It is preserved here for future reference
in case we need it for something else:

	static __always_inline void __set_pointer_opaque(void *const *dst, const void *val)
	{
	    *(void **)dst = (void *)val;
	}

	...
	__set_pointer_opaque((void *const *)&__ret, &__u.__val);
	...

[1] https://docs.kernel.org/next/dev-tools/context-analysis.html
[2] https://lore.kernel.org/all/202601221040.TeM0ihff-lkp@intel.com/
[3] https://git.kernel.org/torvalds/c/bb0c99e08ab9aa6d04b

---
v4:
* Use the return-value-of-function-drops-qualifiers hack.
* Rebase (original patch 1 already merged).

v3: https://lore.kernel.org/all/20260130132951.2714396-1-elver@google.com/
* Comments-smithing.
* Use 'typeof(*__ret) __val'

v2:
* Add __rwonce_typeof_unqual() as fallback for old compilers.

Marco Elver (2):
  arm64: Optimize __READ_ONCE() with CONFIG_LTO=y
  arm64, compiler-context-analysis: Permit alias analysis through
    __READ_ONCE() with CONFIG_LTO=y

 arch/arm64/include/asm/rwonce.h | 24 +++++++++++++++++++-----
 1 file changed, 19 insertions(+), 5 deletions(-)

-- 
2.53.0.335.g19a08e0c02-goog
Re: [PATCH v4 0/2] arm64: Fixes for __READ_ONCE() with CONFIG_LTO=y
Posted by Will Deacon 1 month ago
On Mon, 16 Feb 2026 15:16:21 +0100, Marco Elver wrote:
> While investigating a Clang Context Analysis [1] false positive [2], I
> started to dig deeper into arm64's __READ_ONCE() implementation with
> LTO. That rabbit hole led me to find one critical bug with the current
> implementation (fixed already in [3]), and subtle improvements that then
> enabled me to fix the original false positive.
> 
> Patch 1 refactors the macro to use a different way of getting an
> unqualified type and eliminates the ternary conditional.
> 
> [...]

Applied to arm64 (for-next/read-once), thanks!

[1/2] arm64: Optimize __READ_ONCE() with CONFIG_LTO=y
      https://git.kernel.org/arm64/c/abf1be684dc2
[2/2] arm64, compiler-context-analysis: Permit alias analysis through __READ_ONCE() with CONFIG_LTO=y
      https://git.kernel.org/arm64/c/773b24bcedc1

Peter -- please don't pull this just yet. I'd like it to get some
exposure to the various build bots and CI systems in case of any nasty
toolchain interactions.

Cheers,
-- 
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev