[PATCH] Cache messaging dgm connections

[Jeremy Allison]

On Sun, Sep 18, 2016 at 12:23:24PM +0200, Andreas Schneider wrote:
> On Saturday, 17 September 2016 22:17:00 CEST Ralph Böhme wrote:
> > On Sat, Sep 17, 2016 at 01:17:55PM +0200, Andreas Schneider wrote:
> > > On Friday, 16 September 2016 14:03:42 CEST Jeremy Allison wrote:
> > > > Still got the crash with this in autobuild, now with the backtrace
> > > > below. So we're still not catching everything (or something else
> > > > is freeing ev behind our backs :-).
> > > > 
> > > > As I know we're going to be co-located on Monday, let's
> > > > schedule a time to look at this directly at the SNIA
> > > > conf.
> > > 
> > > This looks like a bug report in Fedora with Samba 4.5.0. smbd directly
> > > segfaults at startup!
> > > 
> > > Backtrace and core file is attached to the bug. We have this since rc1.
> > > 
> > > https://bugzilla.redhat.com/show_bug.cgi?id=1375973
> > 
> > this looks like a different issue.
> 
> Are you sure? The code looks like this:
> 
> 1 void tevent_debug(struct tevent_context *ev, enum tevent_debug_level level,
> 2                   const char *fmt, ...)
> 3 {
> 4        va_list ap;
> 5        if (!ev) {
> 6                return;
> 7        }
> 8        if (ev->debug_ops.debug == NULL) {
> 
> In line 5 the ev was a valid pointer and in line 8 we fail to derefence it and 
> segfault. This either means the pointer ev was freed berween line 5 and 8 or 
> the pointer has been freed before but not set to NULL.
> 
> 
> smbd_notifyd_init() segfaults at smbd startup!

Simple test - apply the patch in this thread and see if it fixes it :-).