Re: [PATCH v3 1/5] genirq: Synchronize in-flight handlers during NMI teardown

[Thomas Gleixner <tglx@kernel.org>]

On Wed, Sep 30 2026 at 16:17, Doug Anderson wrote:
> I guess I'm still unclear on one key question: if we post up clean
> patches that allow promoting global interrupts to NMI (or pseudo-NMI)
> level, would you be willing to review and eventually land? I got a bit
> of a vibe that you just didn't like the idea in general, but maybe I
> misunderstood. I just want to make sure Mayank doesn't spend a bunch
> of time trying to solve this problem if it has zero chance of actually
> landing.

I'm pretty reluctant to export a general "promote me to NMI" interface
to drivers because that's just a recipe for disaster. Watching driver
writers creativity for several decades tought me some lessons.

Though I understand and agree that in particular cases like a watchdog
this is useful and should not require to have the watchdog built in.

One clean solution to that is to provide a function which is exported
via EXPORT_SYMBOL_FOR_MODULES(). Then driver writers who want to use it
must add their module to the export list, which provides at least some
means to catch abuse early.

Thanks,

        tglx