Re: [PATCH 3/6] xen/softirq: Rework arch_skip_send_event_check() into arch_pend_softirq()

["Roger Pau Monné" <roger.pau@citrix.com>]

On Fri, Jul 04, 2025 at 09:23:29AM +0200, Jan Beulich wrote:
> On 03.07.2025 18:21, Roger Pau Monné wrote:
> > On Wed, Jul 02, 2025 at 03:41:18PM +0100, Andrew Cooper wrote:
> >> --- a/xen/include/xen/softirq.h
> >> +++ b/xen/include/xen/softirq.h
> >> @@ -23,6 +23,22 @@ enum {
> >>  
> >>  #define NR_SOFTIRQS (NR_COMMON_SOFTIRQS + NR_ARCH_SOFTIRQS)
> >>  
> >> +/*
> >> + * Ensure softirq @nr is pending on @cpu.  Return true if an IPI can be
> >> + * skipped, false if the IPI cannot be skipped.
> >> + */
> >> +#ifndef arch_pend_softirq
> >> +static always_inline bool arch_pend_softirq(unsigned int nr, unsigned int cpu)
> > 
> > Nit: I would maybe it arch_set_softirq(), I find `pend` not that clear
> > (I would rather fully spell `pending` instead).
> 
> I, too, did wonder about the naming here. But using "pending" as you suggest
> has the effect of giving the function a name we would normally associate with
> a predicate ("Is it pending?"), whereas here the function is used to _mark_ a
> softirq as pending. Hence in the end I didn't comment at all; I'd be fine
> with "set", but I'm also okay with "pend".

It's a set and check kind of function, so I don't care much.  Just
found the pend a bit too short, I don't think we usually abbreviate
pending to pend.  In fact we use pend in more than one variable name
to store the end of a physical memory region.

Thanks, Roger.