Re: [libvirt] [PATCH] fix deadlock when qemu cannot start

On Fri, Mar 30, 2012 at 08:45:35AM -0500, Serge Hallyn wrote: > Quoting Daniel Veillard (veillard(a)redhat.com): > > On Fri, Mar 16, 2012 at 02:37:42PM +0800, Wen Congyang wrote: > > > When qemu cannot start, we may call qemuProcessStop() twice. > > > We have check whether the vm is running at the beginning of > > > qemuProcessStop() to avoid libvirt deadlock. We call > > > qemuProcessStop() with driver and vm locked. It seems that > > > we can avoid libvirt deadlock. But unfortunately we may > > > unlock driver and vm in the function qemuProcessKill() while > > > vm->def->id is not -1. So qemuProcessStop() will be run twice, > > > and monitor will be freed unexpectedly. So we should set > > > vm->def->id to -1 at the beginning of qemuProcessStop(). > > Oh, uh, Huh. This seems like it could be responsible for what I was > reporting a few days ago (*1). I spent most of yesterday trying to > track it down, only to finally realize that the QemuProcessStart would > silently die at various places. So i was getting ready to send an email > postulating that what's happening is that a virsh list starts, then > a virsh start starts, and when the virsh list ends it somehow causes > the virsh start to be killed. > > I'll test and see if this patch fixes it. I guess I need to resurect my patch checker, tune it and make it send warning on patches not ACK'ed or reviewed automatically, or even better make it a web page to avoid boring people. http://libvirt.org/git/?p=patchchecker.git;a=summary I had that patch on my radar and though I took an awful long time to review it it's there. BTW I'm late for rc2, I will make it tomorrow I guess, sorry about it this is due to events out of my control :-\ Daniel