segfaults testing winbindd in trunk

John P Janosik jpjanosi at
Tue May 24 21:34:55 GMT 2005

I've been testing the version of winbindd in trunk since I'm interested in
improvements in this area for Samba file servers that are members of large
NT domains.  It seems so far to function properly in our domain but under
heavy nsswitch activity(200 name lookups/minute) I can get winbindd to
segfault after a few minutes.  I tested this last week and again this
morning with revison 6955.  Note that winbindd in samba 3.0.14a can't
handle this number of lookups per minute in my environment (see bugzilla
2736) but it does not segfault.

Here is a stack trace from gdb:

(gdb) bt
#0  0x002fa248 in wait () from /lib/tls/
#1  0x08072934 in sigchld_handler (signum=17) at nsswitch/winbindd.c:210
#2  <signal handler called>
#3  0x00278f79 in sigprocmask () from /lib/tls/
#4  0x0028f201 in do_system () from /lib/tls/
#5  0x080e5514 in smb_panic2 (why=0x81d1932 "internal error",
decrement_pid_count=1) at lib/util.c:1539
#6  0x080e5420 in smb_panic (why=0x81d1932 "internal error") at
#7  0x080d0402 in fault_report (sig=6) at lib/fault.c:41
#8  0x080d0469 in sig_fault (sig=6) at lib/fault.c:64
#9  <signal handler called>
#10 0x00278cef in raise () from /lib/tls/
#11 0x0027a4f5 in abort () from /lib/tls/
#12 0x080eb5a0 in talloc_chunk_from_ptr (ptr=0x944bcc0) at lib/talloc.c:111
#13 0x080ebcb1 in talloc_get_name (ptr=0x944bcc0) at lib/talloc.c:432
#14 0x080ebd09 in talloc_check_name (ptr=0x944bcc0, name=0x81c16cc "struct
    at lib/talloc.c:451
#15 0x080e7afe in talloc_check_name_abort (ptr=0x944bcc0, name=0x81c16cc
"struct winbindd_async_request")
    at lib/util.c:2682
#16 0x0808f42b in async_request_sent (private=0x944bcc0, success=1) at
#17 0x08072de9 in rw_callback (event=0x9310270, flags=2) at
#18 0x0807399d in process_loop () at nsswitch/winbindd.c:801
#19 0x0807424b in main (argc=2, argv=0xbfffbe84) at

John Janosik
jpjanosi at

More information about the samba-technical mailing list