On Tue, Jul 17, 2012 at 02:36:18PM -0600, Eric Blake wrote:
On 07/11/2012 07:35 AM, Daniel P. Berrange wrote:
From: "Daniel P. Berrange" <berrange@redhat.com>
Make all the virNetClient* objects use virObject APIs for reference counting
Signed-off-by: Daniel P. Berrange <berrange@redhat.com> ---
@@ -286,19 +294,22 @@ static virNetClientPtr virNetClientNew(virNetSocketPtr sock,
- if (virMutexInit(&client->lock) < 0) + if (virMutexInit(&client->lock) < 0) { + VIR_FREE(client); goto error;
Hmm - this discards the fact that client has a ref-count of 1, and thus fails to poison that memory (one of the benefits of always going through unref is that that poisoning helps detect use-after-free bugs). Thankfully, though, there is no other allocated memory owned by 'client' at this point in time, so skipping the dispose method is not technically a leak; furthermore, since the dispose method tries to call virMutexDestroy but we know the mutex failed to initialize, I think it's the best we can do. You made me think about this, but in the end I agree it is right, even though it looks fishy.
FYI, this will be temporary nastiness. I intend to actually add a virMutex to the virObjectPtr base object in a future patch series, so subclasses won't need to maintain their own mutexes. So we'll easily be able todo the right thing wrt to poisoning then. Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|