On Sat, 2022-11-19 at 11:38 +0000, Durrant, Paul wrote:
> > + size_t user_len, user_len1, user_len2;
>
> I think it might make the code marginally neater to use two-element arrays... but only marginally.
Not sure; I kind of prefer it like this. At least overall, this version
is somewhat nicer than some of the interim attempts that you saw... :)
> > + * Attempt to obtain the GPC lock on *both* (if there are two)
> > + * gfn_to_pfn caches that cover the region.
> > + */
> > + read_lock_irqsave(&gpc1->lock, flags);
> > + while (!kvm_gfn_to_pfn_cache_check(v->kvm, gpc1, gpc1->gpa, user_len1)) {
> > + read_unlock_irqrestore(&gpc1->lock, flags);
> >
> > - vx->runstate_times[RUNSTATE_runnable] += steal_ns;
> > + /* When invoked from kvm_sched_out() we cannot sleep */
> > + if (atomic)
> > + return;
> > +
> > + if (kvm_gfn_to_pfn_cache_refresh(v->kvm, gpc1, gpc1->gpa, user_len1))
> > + return;
> > +
> > + read_lock_irqsave(&gpc1->lock, flags);
>
> This is a repeated pattern now, so how about a helper function?
I've been tempted by that a few times but although it's a repeating
pattern, there are differences every time around the conditions for
bailing out early — the if (atomic) part above.
Attachment:
smime.p7s
Description: S/MIME cryptographic signature
