Re: [PATCH] clocksource: disable irq when holding watchdog_lock.

From: Thomas Gleixner
Date: Tue Oct 24 2023 - 09:00:21 EST


On Thu, Oct 19 2023 at 23:26, Tetsuo Handa wrote:
> On 2023/10/19 21:14, Thomas Gleixner wrote:
>> Maybe because of this:
>>>>> (This environment is Oracle VM VirtualBox on Windows 11 host.)
>
> I don't think that I'm the only user running Linux on Oracle VM VirtualBox on Windows. ;-)
>
>> So you could boot with "nosmp clocksource=acpi_pm" on the command line
>> and compare that against a "nosmp" boot.
>
> This did not make much difference.

Interesting. Can you please tell what the replacement clocksource is
when the TSC is disabled?

> Just for testing, I booted several Fedora/Ubuntu kernels used by installation ISO image.
> 5.11 / 6.0 / 6.6 Fedora kernels all showed this problem. On the other hand, 5.15 Ubuntu
> kernel does not show this problem. Thus, I'd rather suspect kernel config dependent
> problem than platform dependent problem.
>
> Fedora-Everything-netinst-x86_64-Rawhide-20231018.n.0.iso
> [ 0.000000] Linux version 6.6.0-0.rc6.47.fc40.x86_64 (mockbuild@65b4f242683a45f6bd4621fe8c018a9d) (gcc (GCC) 13.2.1 20231011 (Red Hat 13.2.1-4), GNU ld version 2.41-7.fc40) #1 SMP PREEMPT_DYNAMIC Mon Oct 16 13:34:35 UTC 2023
>
> Fedora-Server-netinst-x86_64-37-1.7.iso
> [ 0.000000] Linux version 6.0.7-301.fc37.x86_64 (mockbuild@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx) (gcc (GCC) 12.2.1 20220819 (Red Hat 12.2.1-2), GNU ld version 2.38-24.fc37) #1 SMP PREEMPT_DYNAMIC Fri Nov 4 18:35:48 UTC 2022
>
> Fedora-Everything-netinst-x86_64-34-1.2.iso
> [ 0.000000] Linux version 5.11.12-300.fc34.x86_64 (mockbuild@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx) (gcc (GCC) 11.0.1 20210324 (Red Hat 11.0.1-0), GNU ld version 2.35.1-41.fc34) #1 SMP Wed Apr 7 16:31:13 UTC 2021
>
> ubuntu-22.04.3-live-server-amd64.iso
> [ 0.000001] kvm-clock: using sched offset of 33909880646 cycles
> [ 0.000004] clocksource: kvm-clock: mask: 0xffffffffffffffff max_cycles: 0x1cd42e4dffb, max_idle_ns: 881590591483 ns

Are the others using kvm-clock too? I haven't seen that in the snippets
you provided before.

Thanks,

tglx