On Tue, Mar 21, 2017 at 07:45:13AM -0600, Jan Beulich wrote: > >>> On 21.03.17 at 11:52, <roger....@citrix.com> wrote: > > On Tue, Mar 21, 2017 at 01:56:51AM -0600, Jan Beulich wrote: > >> >>> On 20.03.17 at 19:27, <roger....@citrix.com> wrote: > >> > On Fri, Mar 03, 2017 at 10:06:03AM -0700, Jan Beulich wrote: > >> >> >>> On 23.02.17 at 12:52, <roger....@citrix.com> wrote: > >> >> > @@ -91,13 +92,16 @@ static int pt_irq_vector(struct periodic_time > >> >> > *pt, enum hvm_intsrc src) > >> >> > + (isa_irq & 7)); > >> >> > > >> >> > ASSERT(src == hvm_intsrc_lapic); > >> >> > - return domain_vioapic(v->domain)->redirtbl[gsi].fields.vector; > >> >> > + vioapic = gsi_vioapic(v->domain, gsi, &pin); > >> >> > + > >> >> > + return vioapic->redirtbl[pin].fields.vector; > >> >> > >> >> Please don't chance de-referencing NULL here and below. > >> > > >> > Done, I've added an ASSERT. > >> > >> How about release builds then? > > > > OK, I can add a BUG_ON, but maybe it would be better to add a domain_crash > > and > > suitable printk in case this triggers. > > The latter, please, plus returning the spurious vector (as you still > need to return something).
Do we really need the domain_crash? I've added a gdprintk and returned -1. Roger. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org https://lists.xen.org/xen-devel
