Re: [PATCH 3/3] KVM: arm64: selftests: Test a timer set past the counter's wrap

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] selftests: flaky test due to scheduling delay in test_timers_past_counter_wrap
--

commit 8788b1ebd8d97a89915e8766ef7a7c75ad83b3c9
Author: Fuad Tabba <fuad.tabba@linux.dev>

KVM: arm64: selftests: Test a timer set past the counter's wrap

This commit adds a KVM selftest to verify timer behavior when programmed
with a cval past the counter's wrap point. It programs an unmasked timer
and tests that no unexpected interrupts fire since the guest's counter
cannot realistically reach the deadline.

> diff --git a/tools/testing/selftests/kvm/arm64/arch_timer_edge_cases.c b/tools/testing/selftests/kvm/arm64/arch_timer_edge_cases.c
> index d9c9377a63256..69986a7f0d37d 100644
> --- a/tools/testing/selftests/kvm/arm64/arch_timer_edge_cases.c
> +++ b/tools/testing/selftests/kvm/arm64/arch_timer_edge_cases.c

[ ... ]

> +/*
> + * With the counter at wait_ms, a tval of -(wait_ms + long_wait_ms) puts cval
> + * within long_wait_ms of the counter's wrap: a deadline further away than the
> + * counter can reach. Unmasked, so an early ISTATUS shows up as an IRQ.
> + */
> +static void test_timers_past_counter_wrap(enum arch_timer timer)
> +{
> +	s32 tval = -(s32)msec_to_cycles(test_args.wait_ms + test_args.long_wait_ms);
> +	int i;
> +
> +	for (i = 0; i < ARRAY_SIZE(sleep_method); i++) {
> +		sleep_method_t sm = sleep_method[i];
> +
> +		local_irq_disable();
> +		set_counter(timer, msec_to_cycles(test_args.wait_ms));
> +		set_tval_irq(timer, tval, CTL_ENABLE);

[Severity: Medium]
Could this sequence lead to a flaky test on a loaded host system?

If a host scheduling delay exceeds long_wait_ms between setting the counter
and writing the TVAL, the guest's CNT can advance such that:
CNT > wait_ms + long_wait_ms

When set_tval_irq() then executes, the hardware calculates the absolute
deadline as CVAL = CNT + TVAL. Because TVAL is a negative value of
-(wait_ms + long_wait_ms), CVAL evaluates to a small positive number
instead of wrapping around to near U64_MAX.

Since CVAL is now less than CNT, the timer condition (CNT >= CVAL) is
met immediately, and the hardware legitimately fires the unmasked timer.
This would cause the subsequent assert_irqs_handled(0) check to fail.

> +		sm(timer, msecs_to_usecs(test_args.wait_ms) + TIMEOUT_NO_IRQ_US);
> +		local_irq_enable();
> +		isb();
> +		assert_irqs_handled(0);
> +		timer_set_ctl(timer, CTL_IMASK);
> +	}
> +}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260921140427.2211373-1-fuad.tabba@linux.dev?part=3