Re: next-20241031: kernel/time/clockevents.c:455 clockevents_register_device

[Thomas Gleixner <tglx@linutronix.de>]

On Thu, Oct 31 2024 at 14:10, Naresh Kamboju wrote:
> The QEMU-ARM64 boot has failed with the Linux next-20241031 tag.
> The boot log shows warnings at clockevents_register_device and followed
> by rcu_preempt detected stalls.
>
> However, the system did not proceed far enough to reach the login prompt.
> The fvp-aemva, Qemu-arm64, Qemu-armv7 and Qemu-riscv64 boot failed.
>
> Please find the incomplete boot log links below for your reference.
> The Qemu version is 9.0.2.
> <4>[ 0.220657] WARNING: CPU: 1 PID: 0 at kernel/time/clockevents.c:455
> clockevents_register_device (kernel/time/clockevents.c:455
> <4>[ 0.225218] clockevents_register_device+0x170/0x188 P
> <4>[ 0.225367] clockevents_config_and_register+0x34/0x50 L
> <4>[ 0.225487] clockevents_config_and_register (kernel/time/clockevents.c:523)
> <4>[ 0.225553] arch_timer_starting_cpu
> (drivers/clocksource/arm_arch_timer.c:1034)
> <4>[ 0.225602] cpuhp_invoke_callback (kernel/cpu.c:194)
> <4>[ 0.225649] __cpuhp_invoke_callback_range (kernel/cpu.c:965)
> <4>[ 0.225691] notify_cpu_starting (kernel/cpu.c:1604)

That's obvious what happens here. notify_cpu_starting() is invoked
before the CPU is marked online, which triggers the new check in
clockevents_register_device().

I removed the warning and force pushed the fixed up branch, so that
should be gone by tomorrow.

Thanks,

        tglx