Re: [libvirt] [Xen-devel] [PATCH RFC] libxl: use libxl_event_wait to process libxl events

On Fri, 2015-11-13 at 14:36 -0700, Jim Fehlig wrote: > Prior to this patch, libxl events were delivered to libvirt via > the libxlDomainEventHandler callback registered with libxl. > Documenation in $xensrc/tools/libxl/libxl_event.h states that the > callback "may occur on any thread in which the application calls > libxl". This can result in deadlock since many of the libvirt > callees of libxl hold a lock on the virDomainObj they are working > on. When the callback is invoked, it attempts to find a virDomainObj > corresponding to the domain ID provided by libxl. Searching the > domain obj list results in locking each obj before checking if it is > active, and its ID equals the requested ID. Deadlock is possible > when attempting to lock an obj that is already locked further up > the call stack. Indeed, Max Ustermann recently reported an instance > of this deadlock > > https://www.redhat.com/archives/libvir-list/2015-November/msg00130.html > > This patch moves processing of libxl events to a thread, where > libxl_event_wait() is used to collect events. This allows processing > libxl events asynchronously in libvirt, avoiding the deadlock. > > Reported-by: max ustermann <ustermann78(a)web.de&gt; > Signed-off-by: Jim Fehlig <jfehlig(a)suse.com&gt; > --- > > The only reservations I have about this patch come from comments > about libxl_event_wait in libxl_event.h > > Like libxl_event_check but blocks if no suitable events are > available, until some are. Uses libxl_osevent_beforepoll/ > _afterpoll so may be inefficient if very many domains are being > handled by a single program. > > libvirt does handle "very many domains" :-). But thus far, I haven't > noticed any problems in my testing. The reporter expressed interest > in testing the patch. Positive results from that testing would improve > my confidence, as would an ACK from Ian Jackson. Would it be possible/allowable to have your cake and eat it by using the callbacks but having them simply enqueue the events on the libvirt side and kick the dedicated thread for further processing?