x86/HVM: fix processing of RTC REG_B writes

We must store the new values before calling rtc_update_irq(), and we
need to call rtc_timer_update() when PIE transitions from 0 to 1 (as we
may have previously turned off the periodic timer due to the guest not
reading REG_C, and hence may have to re-enable it in order to start
IRQs getting delivered to the guest).

Note that the timer is being kept running if PIE transitions from 1 to
0, to match the behavior of keeping it running for a brief period of
time if the guest doesn't clear PF in time (in order to avoid
permanently destroying and re-creating the periodic timer).

Signed-off-by: Jan Beulich <jbeulich@suse.com>
Tested-by: Roger Pau Monné <roger.pau@citrix.com> (FreeBSD guest)
Reviewed-by: Tim Deegan <tim@xen.org>

diff --git a/xen/arch/x86/hvm/rtc.c b/xen/arch/x86/hvm/rtc.c
index 26ab12c7f3d0f2bb28e7296440253f8399841ccb..b728d5e3db06b72a84f4d889537eeb95d6a870e3 100644 (file)
--- a/xen/arch/x86/hvm/rtc.c
+++ b/xen/arch/x86/hvm/rtc.c
@@ -468,12 +468,14 @@ static int rtc_ioport_write(void *opaque, uint32_t addr, uint32_t data)
             if ( orig & RTC_SET )
                 rtc_set_time(s);
         }
+        s->hw.cmos_data[RTC_REG_B] = data;
         /*
          * If the interrupt is already set when the interrupt becomes
          * enabled, raise an interrupt immediately.
          */
         rtc_update_irq(s);
-        s->hw.cmos_data[RTC_REG_B] = data;
+        if ( (data & RTC_PIE) && !(orig & RTC_PIE) )
+            rtc_timer_update(s);
         if ( (data ^ orig) & RTC_SET )
             check_update_timer(s);
         if ( (data ^ orig) & (RTC_24H | RTC_DM_BINARY | RTC_SET) )