On 11/22/2010 11:09 AM, Daniel P. Berrange wrote:
Perform a handshake during QEMU startup to allow file relabelling to be performed after QEMU has been forked, but before it has been exec'd. This is to allow the lock manager to acquire locks against the QEMU pid, before relabelling takes place
* src/qemu/qemu_driver.c: Handshake during QEMU startup --- src/qemu/qemu_driver.c | 23 ++++++++++++++--------- 1 files changed, 14 insertions(+), 9 deletions(-)
diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 0b0b6fe..bf6cd0e 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -3683,9 +3683,6 @@ static int qemudSecurityHook(void *data) { if (virSecurityManagerSetProcessLabel(h->driver->securityManager, h->vm) < 0) goto cleanup;
- /* XXX temp hack to let disk labelling complete */ - sleep(10); - ret = 0;
cleanup: @@ -4097,28 +4094,36 @@ static int qemudStartVMDaemon(virConnectPtr conn, virCommandSetErrorFD(cmd, &logfile); virCommandSetPidFile(cmd, pidfile); virCommandDaemonize(cmd); + virCommandRequireHandshake(cmd);
ret = virCommandRun(cmd, NULL);
Oh, I see - you want to daemonize, require handshake, AND do a blocking virCommandRun. I see now - __virExec does two forks when daemonize is requested, and only runs the callback hook in the second (daemon) child, rather than in the (short-lived) intermediary. That answers some of my questions in the previous patch.
- VIR_WARN("Executing done %s", vm->def->emulator); VIR_FREE(pidfile);
- /* XXX this is bollocks. Need a sync point */ - sleep(5); + VIR_WARN0("Waiting for handshake from child"); + if (virCommandHandshakeWait(cmd) < 0) {
But it still probably requires virCommand to do some sanity checking that HandshakeWait is called after the child has been spawned.
+ VIR_WARN0("Labelling done, completing hanshake to child");
s/hanshake/handshake/ Definitely an improvement. -- Eric Blake eblake@redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org