On Tue, Mar 21, 2017 at 08:47:50 +0100, Peter Krempa wrote:
The hyperv panic notifier reports additional data in form of 5 registers that are reported in the crash event from qemu. Log them into the VM log file and report them as a warning so that admins can see the cause of crash of their windows VMs.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1426176 --- src/qemu/qemu_driver.c | 54 ++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 52 insertions(+), 2 deletions(-)
diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index ab7c01b23..c558e0991 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -3971,10 +3971,56 @@ doCoreDumpToAutoDumpPath(virQEMUDriverPtr driver, return ret; }
+ +static char * +qemuProcessGuestPanicEventInfoFormatData(qemuMonitorEventPanicInfoPtr info)
I think this should be implemented in qemu_monitor.c as qemuMonitorEventPanicInfoFormat so that it's closer the structure is defined.
+{ + char *ret = NULL; + + switch (info->type) { + case QEMU_MONITOR_EVENT_PANIC_INFO_TYPE_HYPERV: + ignore_value(virAsprintf(&ret, "hyper-v: arg1='0x%llx', arg2='0x%llx', " + "arg3='0x%llx', arg4='0x%llx', " + "arg5='0x%llx'", + info->data.hyperv.arg1, + info->data.hyperv.arg2, + info->data.hyperv.arg3, + info->data.hyperv.arg4, + info->data.hyperv.arg5));
Wrong indentation, the additional lines should be aligned with &ret :-)
+ break; + + case QEMU_MONITOR_EVENT_PANIC_INFO_TYPE_NONE: + case QEMU_MONITOR_EVENT_PANIC_INFO_TYPE_LAST: + break; + } + + return ret; +} + + +static void +qemuProcessGuestPanicEventInfo(virQEMUDriverPtr driver, + virDomainObjPtr vm, + qemuMonitorEventPanicInfoPtr info) +{ + char *msg = qemuProcessGuestPanicEventInfoFormatData(info); + char *timestamp = virTimeStringNow(); + + if (msg && timestamp) { + qemuDomainLogAppendMessage(driver, vm, "%s: panic %s\n", timestamp, msg); + VIR_WARN("domain '%s' panic data: %s", vm->def->name, msg);
Do we really need this warning? Especially when a guest panic triggers the warning only if we have the additional data to report. We should either emit the warning in all cases or never. I think we should just drop the warning. Jirka