Description of problem:

After recovery I found the following in logs:

systemd[1]: Caught <BUS>, dumped core as pid 16169.
systemd[1]: Freezing execution.

Soon after this entries like that started to show up in logs:

systemd-cgroups-agent[16355]: Failed to get D-Bus connection: Failed to connect to socket /org/freedesktop/systemd1/private: Connection refused

and a machine was effectively crippled.  As this was a remote machine then for all practical purposes it was gone, as sshd was rejecting attempts to connect, and with no path to a recovery.  Even a local reboot was difficult and required few retries before this was acomplished.

abrtd did not catch anything bug systemd left a core as /core.16169.  With
systemd-debuginfo loaded gdb reports the following:


Using host libthread_db library "/lib64/libthread_db.so.1".
Core was generated by `/bin/systemd --log-level info --log-target syslog-or-kmsg --system --dump-core'.
Program terminated with signal 7, Bus error.
#0  0x00000033b1c0f3cb in raise () from /lib64/libpthread.so.0

and a backtrace is:

#0  0x00000033b1c0f3cb in raise () from /lib64/libpthread.so.0
#1  0x000000000040b42c in crash (sig=7) at src/main.c:123
#2  <signal handler called>
#3  hashmap_iterate (h=0x7f495a4510c0, i=0x7fff5b69cc20, key=0x0)
    at src/hashmap.c:474
#4  0x0000000000412eb8 in transaction_add_job_and_dependencies (tr=0x15ac8f0, 
    type=<optimized out>, unit=<optimized out>, by=<optimized out>, 
    matters=false, override=false, conflicts=false, ignore_requirements=false, 
    ignore_order=false, e=0x7fff5b69ce30) at src/transaction.c:904
#5  0x0000000000412a2a in transaction_add_job_and_dependencies (tr=0x15ac8f0, 
    type=JOB_START, unit=<optimized out>, by=<optimized out>, matters=false, 
    override=false, conflicts=false, ignore_requirements=false, 
    ignore_order=false, e=0x7fff5b69ce30) at src/transaction.c:872
#6  0x0000000000412a2a in transaction_add_job_and_dependencies (tr=0x15ac8f0, 
    type=JOB_RESTART, unit=<optimized out>, by=<optimized out>, matters=true, 
    override=false, conflicts=false, ignore_requirements=false, 
    ignore_order=false, e=0x7fff5b69ce30) at src/transaction.c:872
#7  0x000000000040cb26 in manager_add_job (m=0x159d5b0, type=JOB_RESTART, 
    unit=0x160bca0, mode=JOB_FAIL, override=false, e=0x7fff5b69ce30, _ret=0x0)
    at src/manager.c:654
#8  0x0000000000420dd2 in service_enter_restart (s=0x160bca0)
    at src/service.c:2134
#9  service_timer_event (u=0x160bca0, elapsed=<optimized out>, 
    w=<optimized out>) at src/service.c:2947
#10 0x0000000000410d72 in process_event (ev=0x7fff5b69cea0, m=0x159d5b0)
    at src/manager.c:1330
#11 manager_loop (m=0x159d5b0) at src/manager.c:1427
#12 0x0000000000409207 in main (argc=<optimized out>, argv=<optimized out>)
    at src/main.c:1466

This is a loop which bombed out:

1465	        for (;;) {
1466	                if ((r = manager_loop(m)) < 0) {
1467	                        log_error("Failed to run mainloop: %s", strerror(-r));
1468	                        goto finish;
1469	                }

A value for 'm' is '(Manager *) 0x159d5b0' and this is a rather big structure.

In 'manager_loop' (frame #11) this is:

1427	                if ((r = process_event(m, &event)) < 0)
1428	                        return r;

and 'event' has a value {events = 1, data = {ptr = 0x160c110, fd = 23118096, u32 = 23118096, u64 = 23118096}}

Version-Release number of selected component (if applicable):
systemd-37-25.fc16.x86_64

How reproducible:
No idea.  I did not see that before.
  
Actual results:
A mess.

Expected results:
A machine does not fall apart because systemd has hiccups.