Re: [PATCH V1 09/16] arm/ioreq: Introduce arch specific bits for IOREQ/DM features

[Oleksandr <olekstysh@xxxxxxxxx>]

On 16.09.20 10:51, Jan Beulich wrote:

Hi Jan

> On 10.09.2020 22:22, Oleksandr Tyshchenko wrote:
> > @@ -2277,8 +2299,10 @@ void leave_hypervisor_to_guest(void)
> >   {
> >       local_irq_disable();
> >   
> > -    check_for_vcpu_work();
> > -    check_for_pcpu_work();
> > +    do
> > +    {
> > +        check_for_pcpu_work();
> > +    } while ( check_for_vcpu_work() );
> Looking at patch 11 I've stumbled across changes done to code
> related to this, and now I wonder: There's no mention in the
> description of why this safe (i.e. not a potentially unbounded
> loop).
Indeed there was a discussion regarding that. Probably I should have 
added an explanation.
Please see the thoughts about that and a reason why it was decided to 
let the vCPU to spin forever if I/O never completes
(check_for_vcpu_work() never returns false) and why it was considered as 
safe action:
https://patchwork.kernel.org/patch/11698549/#23540209


> As a nit - the opening brace does not belong on its own line in
> this specific case.
ok

--
Regards,

Oleksandr Tyshchenko