Re: [PATCH v12 13/15] arm64/delay: enable testing smp_cond_load_relaxed_timeout()

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] The new generic smp_cond_load_relaxed_timeout() and atomic timeout APIs fail to link in loadable modules on ARM64 when KUnit is disabled due to conditionally exported backend functions.
--

commit b5a10ff859d6e282c17d5ed23e24d2c5703beb7a
Author: Ankur Arora <ankur.a.arora@oracle.com>

arm64/delay: enable testing smp_cond_load_relaxed_timeout()

This commit enables the barrier tests to be built as a module by exporting
__delay_cycles and arch_timer_evtstrm_available specifically for KUnit
environments.

> diff --git a/arch/arm64/lib/delay.c b/arch/arm64/lib/delay.c
> --- a/arch/arm64/lib/delay.c
> +++ b/arch/arm64/lib/delay.c
[ ... ]
> @@ -30,6 +31,7 @@ u64 notrace __delay_cycles(void)
>  	guard(preempt_notrace)();
>  	return __arch_counter_get_cntvct_stable();
>  }
> +EXPORT_SYMBOL_IF_KUNIT(__delay_cycles);

[Severity: Medium]
Does exporting this only for KUnit prevent normal kernel modules from using
the generic smp_cond_load_relaxed_timeout() macro?

Since smp_cond_load_relaxed_timeout() expands into inline code that calls
__delay_cycles(), a loadable module using this generic API will fail to link
due to an undefined symbol in production kernels where KUnit is disabled.
Should this be unconditionally exported, perhaps via EXPORT_SYMBOL_GPL?

[ ... ]
> diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
> --- a/drivers/clocksource/arm_arch_timer.c
> +++ b/drivers/clocksource/arm_arch_timer.c
[ ... ]
> @@ -896,6 +897,7 @@ bool arch_timer_evtstrm_available(void)
>  	 */
>  	return cpumask_test_cpu(raw_smp_processor_id(), &evtstrm_available);
>  }
> +EXPORT_SYMBOL_IF_KUNIT(arch_timer_evtstrm_available);

[Severity: Medium]
Similarly, does conditionally exporting this feature-check break the generic
barrier timeout primitives for modules?

Like __delay_cycles(), arch_timer_evtstrm_available() is called by the
inline implementations of the atomic timeout APIs. Leaving this unexported
for standard kernel configurations could cause link errors for any driver
that uses atomic_cond_read_acquire_timeout() or similar macros.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260608080440.127491-1-ankur.a.arora@oracle.com?part=13