Re: exited on signal 11 (no core dump - other error)

On Fri, Jul 12, 2024 at 10:45:31AM +0000, Dave Cottlehuber wrote:
> On Fri, 12 Jul 2024, at 03:39, Zhenlei Huang wrote:
> > Hi
> >
> > I observed something weird on Release 14.1.
> >
> > When rebooting my dev machine, I got
> > ...
> > IIUC all processes will get signal to quit on system reboot. But what does the
> > signal 11 mean ? Is it EDEADLK in sys/sys/errno.h ?
> >
> > If yes, then why they get dead locked ?
> 
> I see the same on 15.0-CURRENT too here. In my case this is just after syslog-ng is stopped.
> 
> <6>[1920] pid 6090 (wezterm-gui), jid 0, uid 1002: exited on signal 11 (no core dump - other error)
> <6>[1920] pid 6039 (polkitd), jid 0, uid 565: exited on signal 11 (no core dump - bad address)
> <6>[1920] pid 4306 (dbus-daemon), jid 0, uid 556: exited on signal 11 (no core dump - bad address)

Most natural cause for SIGSEGV during shutdown is because root is unmounted
while the processes are still handling signals (SIGTERM) from init.  The
text vnodes for the process binary and shared libraries are force-reclaimed,
and any page-in request results in the unhandled fault.

I regularly see these SIGSEGVs on nfs-booted crash boxes.