Re: [boqun:rust-rcu 14/27] rust/helpers/interrupt.c:17:9: warning: static function 'arch_local_save_flags' is used in an inline function with external linkage

On Wed, Jul 15, 2026 at 12:12:53PM -0700, Boqun Feng wrote:
> [Cc s390]
> >    In file included from rust/helpers/helpers.c:68:
> > >> rust/helpers/interrupt.c:17:9: warning: static function 'arch_local_save_flags' is used in an inline function with external linkage [-Wstatic-in-inline]
> >    17 |         return irqs_disabled();
> >    |                ^
> >    include/linux/irqflags.h:256:3: note: expanded from macro 'irqs_disabled'
> >    256 |                 raw_local_save_flags(_flags);           \
> >    |                 ^
> >    include/linux/irqflags.h:184:11: note: expanded from macro 'raw_local_save_flags'
> >    184 |                 flags = arch_local_save_flags();        \
> >    |                         ^
> >    rust/helpers/interrupt.c:15:1: note: use 'static' to give inline function 'rust_helper_irqs_disabled' internal linkage
> >    15 | __rust_helper bool rust_helper_irqs_disabled(void)
> >    | ^
> >    | static
> >    rust/helpers/helpers.c:38:23: note: expanded from macro '__rust_helper'
> >    38 | #define __rust_helper __always_inline
> >    |                       ^
> >    include/linux/compiler_attributes.h:55:41: note: expanded from macro '__always_inline'
> >    55 | #define __always_inline                 inline __attribute__((__always_inline__))
> >    |                                         ^
> >    include/linux/compiler_types.h:235:16: note: expanded from macro 'inline'
> >    235 | #define inline inline __gnu_inline __inline_maybe_unused notrace
> >    |                ^
> >    arch/s390/include/asm/irqflags.h:46:48: note: 'arch_local_save_flags' declared here
> >    46 | static arch_local_irq_attributes unsigned long arch_local_save_flags(void)
> >    |                                                ^
> >    1 warning generated.
> >    warning: 1 warning emitted
> > 
> 
> This seems not a Rust-only issue to me, basically you got the warning
> whenever you have an inline function calling irqs_disabled(). I'm a bit
> curious why 1b301f5f28ba ("s390/irqflags: do not instrument
> arch_local_irq_*() with KMSAN") wanted to add noinline when KMSAN=y, all
> you need is noinstr, no?

While the combination of __always_inline and noinstr _seems_ to work
with clang, this looks just wrong to me. I guess we really would need
a (single) out-of-line variant of the affected functions in question
for KMSAN.

Ilya, since that was your commit, does that sound sane?