APIC: Do not start zero-period timers

The APIC timer must not start when the initial count is (still) zero.

This caused occasional stalls when booting secondary CPUs of Linux SMP guests.

---
 hw/apic.c |    2 ++
 1 file changed, 2 insertions(+)

Index: b/hw/apic.c
===================================================================
@@ -566,6 +566,8 @@ static void apic_timer_update(APICState
         d = (current_time - s->initial_count_load_time) >>
             s->count_shift;
         if (s->lvt[APIC_LVT_TIMER] & APIC_LVT_TIMER_PERIODIC) {
+            if (!s->initial_count)
+                goto no_timer;
             d = ((d / ((uint64_t)s->initial_count + 1)) + 1) * ((uint64_t)s->initial_count + 1);
         } else {
             if (d >= s->initial_count)

The APIC timer must not start when the initial count is (still) zero.

This caused occasional stalls when booting secondary CPUs of Linux SMP guests.

---
 hw/apic.c |    2 ++
 1 file changed, 2 insertions(+)

Index: b/hw/apic.c
===================================================================
@@ -566,6 +566,8 @@ static void apic_timer_update(APICState
         d = (current_time - s->initial_count_load_time) >>
             s->count_shift;
         if (s->lvt[APIC_LVT_TIMER] & APIC_LVT_TIMER_PERIODIC) {
+            if (!s->initial_count)
+                goto no_timer;
             d = ((d / ((uint64_t)s->initial_count + 1)) + 1) * ((uint64_t)s->initial_count + 1);
         } else {
             if (d >= s->initial_count)

Applied. Thanks.

Besides testing, I confirmed that this is what's done in the in-kernel APIC in KVM (which is also used by Xen). It seems really unfortunate to me that there are fixes in the in-kernel APIC that aren't in QEMU (and perhaps vice-versa). I wonder if there's any way to unify the two code bases?

Regards,

Anthony Liguori

Just a thought - if the code bases have similar textual content, one could start playing games like noticing when the kernel releases have APIC changes in KVM, and then mailing those as [PATCH] automatically to this list by detecting the commit log messages for them from kernel git.

And vice versa, even?

Even just a web page that shows the two implementations side by side, highlighting the diff, could probably be useful.

-- Asheesh.

Experience is what you get when you were expecting something else.