[External] Re: [PATCH 1/2] watchdog: refactor watchdog_hld functionality

[Doug Anderson]

Hi,

On Tue, Sep 23, 2025 at 7:41 PM yunhui cui <cuiyunhui at bytedance.com> wrote:
>
> Hi Doug,
>
> On Sat, Aug 30, 2025 at 5:34 AM Doug Anderson <dianders at chromium.org> wrote:
> >
> > Hi,
> >
> > On Wed, Aug 27, 2025 at 3:10 AM Yunhui Cui <cuiyunhui at bytedance.com> wrote:
> > >
> > > Move watchdog_hld.c to kernel/, and rename arm_pmu_irq_is_nmi()
> > > to arch_pmu_irq_is_nmi() for cross-arch reusability.
> > >
> > > Signed-off-by: Yunhui Cui <cuiyunhui at bytedance.com>
> > > ---
> > >  arch/arm64/kernel/Makefile                   | 1 -
> > >  drivers/perf/arm_pmu.c                       | 2 +-
> > >  include/linux/nmi.h                          | 1 +
> > >  include/linux/perf/arm_pmu.h                 | 2 --
> > >  kernel/Makefile                              | 2 +-
> > >  {arch/arm64/kernel => kernel}/watchdog_hld.c | 8 ++++++--
> > >  6 files changed, 9 insertions(+), 7 deletions(-)
> > >  rename {arch/arm64/kernel => kernel}/watchdog_hld.c (97%)
> >
> > I'm not a huge fan of the perf hardlockup detector and IMO we should
> > maybe just delete it. Thus spreading it to support a new architecture
> > isn't my favorite thing to do. Can't you use the buddy hardlockup
> > detector?
> >
> > That being said, I did a quick look at your patch. I'm pretty sure you
> > can't just move the arm64 "watchdog_hld.c" to be generic. Won't
> > hw_nmi_get_sample_period() conflict with everyone else's (x86 and
> > powerpc)?
>
> After discussing whether to remove watchdog perf, it still seems
> necessary to keep advancing with it. For the code, we just need to
> decorate hw_nmi_get_sample_period() with __weak, right?

That would probably work, but IMO you should make sure you can figure
out how to at least compile the x86/powerpc kernels to confirm.

-Doug