Re: [PATCH net v2] net: stmmac: unwind the WoL IRQ after a safety IRQ request failure

[James Hilliard]

On Wed, Oct 7, 2026 at 6:46 AM Andrew Lunn <andrew@xxxxxxx> wrote:
>
> On Tue, Oct 06, 2026 at 07:27:05PM -0600, James Hilliard wrote:
> > On Tue, Oct 6, 2026 at 6:57 PM Andrew Lunn <andrew@xxxxxxx> wrote:
> > >
> > > On Tue, Oct 06, 2026 at 05:17:20PM -0600, James Hilliard wrote:
> > > > On Tue, Oct 6, 2026 at 4:52 PM Jakub Kicinski <kuba@xxxxxxxxxx> wrote:
> > > > >
> > > > > On Tue, 6 Oct 2026 01:14:52 -0600 James Hilliard wrote:
> > > > > > > Please do not repost the series just to address the above. Instead,
> > > > > > > reply to this email with the missing information, so that reviewers
> > > > > > > can take it into account. If the series needs another revision for
> > > > > > > other reasons, please include the information in the commit messages
> > > > > > > then.
> > > > >
> > > > > Please make sure you read this paragraph, to the end.
> > > >
> > > > How much detail should I be including?
> > >
> > > We want to decide should the patch go to net, because it is a real
> > > problem which bothers somebody. Or is it a theoretical problem which
> > > will never happen, so we might want net-next, or maybe /dev/null.
> >
> > I mean, it's in an error handling pathway that could be hit if there's
> > an -ENOMEM error on an IRQ-action allocation.
>
> ENOMEM does not happen for small allocations.
>
> https://lwn.net/Articles/627419/
>
> So at first glance, this seems totally theoretical, it will never
> happen. So net-next, no need for a Fixes tag, etc.

It looks like there might be a few other ways to hit this cleanup pathway,
not sure how plausible this is in practice however.

The unwind also handles other request_irq() failures. With forced IRQ
threading, including PREEMPT_RT, creating the safety IRQ thread can
return -EINTR if the task opening the interface receives a fatal signal
while waiting in kthread_create(). That reaches the same incorrect
cleanup without an allocation failure.

>
>         Andrew