hw/rtc/goldfish: keep time offset when resetting

Currently resetting the leads to resynchronizing the Goldfish RTC
with the system clock of the host. In real hardware an RTC reset
would not change the wall time. Other RTCs like pl031 do not show
this behavior.

Move the synchronization of the RTC with the system clock to the
instance realization.

Cc: qemu-stable@nongnu.org
Reported-by: Frederik Du Toit Lotter <fred.lotter@canonical.com>
Fixes: 9a5b40b8427 ("hw: rtc: Add Goldfish RTC device")
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20250321221248.17764-1-heinrich.schuchardt@canonical.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

hw/rtc/goldfish_rtc.c

@@ -239,15 +239,8 @@ static const VMStateDescription goldfish_rtc_vmstate = {
 static void goldfish_rtc_reset(DeviceState *dev)
 {
     GoldfishRTCState *s = GOLDFISH_RTC(dev);
-    struct tm tm;
 
     timer_del(s->timer);
-
-    qemu_get_timedate(&tm, 0);
-    s->tick_offset = mktimegm(&tm);
-    s->tick_offset *= NANOSECONDS_PER_SECOND;
-    s->tick_offset -= qemu_clock_get_ns(rtc_clock);
-    s->tick_offset_vmstate = 0;
     s->alarm_next = 0;
     s->alarm_running = 0;
     s->irq_pending = 0;
@@ -258,6 +251,7 @@ static void goldfish_rtc_realize(DeviceState *d, Error **errp)
 {
     SysBusDevice *dev = SYS_BUS_DEVICE(d);
     GoldfishRTCState *s = GOLDFISH_RTC(d);
+    struct tm tm;
 
     memory_region_init_io(&s->iomem, OBJECT(s),
                           &goldfish_rtc_ops[s->big_endian], s,
@@ -267,6 +261,11 @@ static void goldfish_rtc_realize(DeviceState *d, Error **errp)
     sysbus_init_irq(dev, &s->irq);
 
     s->timer = timer_new_ns(rtc_clock, goldfish_rtc_interrupt, s);
+
+    qemu_get_timedate(&tm, 0);
+    s->tick_offset = mktimegm(&tm);
+    s->tick_offset *= NANOSECONDS_PER_SECOND;
+    s->tick_offset -= qemu_clock_get_ns(rtc_clock);
 }
 
 static const Property goldfish_rtc_properties[] = {