Re: Linux 6.3-rc2

[Linus Torvalds]

On Mon, Mar 13, 2023 at 4:11 PM Frederic Weisbecker <frederic@xxxxxxxxxx> wrote:
>
> So the hardirq interrupted some code that has softirqs disabled (or
> servicing) from the preempt mask POV but not from lockdep POV.

.. but those are two different counters, so isn't that *always* racy?
IOW, one counter being updated before the other...

IOW, isn't this *why* that check isn't supposed to ever trigger at all
from hard-irq context?

               Linus