Re: [PATCH v6 04/13] xen/arm: Don't release IRQs on suspend

[Mykola Kvach <xakep.amatop@xxxxxxxxx>]

Hi Julien,

Thank you for the review.

On Sat, Sep 13, 2025 at 2:45 AM Julien Grall <julien@xxxxxxx> wrote:
>
> Hi Mykola,
>
> On 01/09/2025 23:10, Mykola Kvach wrote:
> > From: Mykola Kvach <mykola_kvach@xxxxxxxx>
> >
> > If we call disable_nonboot_cpus on ARM64 with system_state set
> > to SYS_STATE_suspend, the following assertion will be triggered:
>
> Looking at the stack trace, I don't understand why this error would not
> happen when offlining a CPU. Can you clarify?
>
> Anyway, I am not very happy to special case suspend/resume in the IRQ
> code. So I would strongly prefer if we follow a different approach.
>
> The one that come to my mind is to switch from request_irq() to
> setup_irq() and allocate the action in a per-cpu variable. With that,
> there should be no free happening with the stop_machine helper.

Yes, this should help in my case and it also looks like a cleaner
solution, thank you.

Interestingly, my teammate Mykyta Poturai came up with the same idea a
few days ago when he faced a similar problem during CPU hotplug
implementation.

So I will just reuse his commits this is the one of the commits:

https://github.com/Deedone/xen/commit/3817601c2f437453035839f29e94c069a770817d

>
> Cheers,
>
> --
> Julien Grall
>

Best regards,
Mykola