Wait to receive QMP greeting before sending any monitor commands

Technically speaking we should wait until we receive the QMP
greeting message before attempting to send any QMP monitor
commands. Mostly we've got away with this, but there is a race
in some QEMU which cause it to SEGV if you sent it data too
soon after startup. Waiting for the QMP greeting avoids the
race

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>

1 files changed, 8 insertions, 1 deletions

diff --git a/src/qemu/qemu_monitor.c b/src/qemu/qemu_monitor.c
index 65694bab7..543b6cd17 100644
--- a/src/qemu/qemu_monitor.c
+++ b/src/qemu/qemu_monitor.c
@@ -79,6 +79,7 @@ struct _qemuMonitor {
 
     unsigned json: 1;
     unsigned json_hmp: 1;
+    unsigned wait_greeting: 1;
 };
 
 static virClassPtr qemuMonitorClass;
@@ -365,6 +366,9 @@ qemuMonitorIOProcess(qemuMonitorPtr mon)
     if (len < 0)
         return -1;
 
+    if (len && mon->wait_greeting)
+        mon->wait_greeting = 0;
+
     if (len < mon->bufferOffset) {
         memmove(mon->buffer, mon->buffer + len, mon->bufferOffset - len);
         mon->bufferOffset -= len;
@@ -538,7 +542,8 @@ static void qemuMonitorUpdateWatch(qemuMonitorPtr mon)
     if (mon->lastError.code == VIR_ERR_OK) {
         events |= VIR_EVENT_HANDLE_READABLE;
 
-        if (mon->msg && mon->msg->txOffset < mon->msg->txLength)
+        if ((mon->msg && mon->msg->txOffset < mon->msg->txLength) &&
+            !mon->wait_greeting)
             events |= VIR_EVENT_HANDLE_WRITABLE;
     }
 
@@ -718,6 +723,8 @@ qemuMonitorOpenInternal(virDomainObjPtr vm,
     mon->hasSendFD = hasSendFD;
     mon->vm = vm;
     mon->json = json;
+    if (json)
+        mon->wait_greeting = 1;
     mon->cb = cb;
     qemuMonitorLock(mon);