On Thu, Jan 22, 2026 at 11:35:06AM +0100, Jan Beulich wrote:
> On 22.01.2026 11:23, Roger Pau Monné wrote:
> > On Mon, Nov 17, 2025 at 03:40:08PM +0100, Jan Beulich wrote:
> >> @@ -162,10 +161,15 @@ static int reprogram_hpet_evt_channel(
> >>
> >> ch->next_event = expire;
> >>
> >> - delta = min_t(int64_t, delta, MAX_DELTA_NS);
> >> delta = max_t(int64_t, delta, MIN_DELTA_NS);
> >> delta = ns2ticks(delta, ch->shift, ch->mult);
> >>
> >> + if ( delta > UINT32_MAX )
> >> + {
> >> + hpet_write32(hpet_read32(HPET_COUNTER), HPET_Tn_CMP(ch->idx));
> >
> > Should Xen disable interrupts around this call to avoid unexpected
> > latency between the counter read and the comparator write?
>
> Such latency could then still arise, due NMI or SMI. What's your underlying
> concern here?
For NMI or SMI there isn't much we can do. I guess this is much less
of a concern here than it is in hpet_next_event(), given the next
event is expected to be after a full HPET counter period. One of
those events taking a full HPET counter period overlap would make a
lot of others things explode.
Reviewed-by: Roger Pau Monné <[email protected]>
Thanks, Roger.
