Signal 11 in samba 2.0.7 and other "bugs"

[Rudolf Kollien]

Hello all,

since a couple of days samba reports an internal error with signal 11 to the
log. As i found in the mailing list, this error occured in versions prior to
2.0.7 and should be fixed in 2.0.7. The error is logged as:

[2000/06/20 17:10:10, 1] smbd/service.c:close_cnum(583)
  pc015 (a.a.a.a) closed connection to service fibu
[2000/06/20 18:29:36, 0] lib/fault.c:fault_report(40)
  ===============================================================
[2000/06/20 18:29:36, 0] lib/fault.c:fault_report(41)
  INTERNAL ERROR: Signal 11 in pid 7126 (2.0.7)
  Please read the file BUGS.txt in the distribution
[2000/06/20 18:29:36, 0] lib/fault.c:fault_report(43)
  ===============================================================
[2000/06/20 18:29:36, 0] lib/util.c:smb_panic(2381)
  PANIC: internal error

As i can see on the timestamp, the error occurs not directly above the last
logged action. The error occurs on different clients (all win95a/b/c) all
over the day. On previous versions of samba there were reports on panics
like this when the netbios name is set to a dns name (which is true in my
case) and should be fixed in 2.0.7. In all older versions we were running in
the past, the setup was equal to the current and no errors occured. As on a
pre version of 2.0.7 this panic was reported on changing smb.conf while smb
is running. This is not the cause on my site. Changing smb.conf on the
running system produces no errors. Restarting samba solves the problem only
for some hours.

My configuration is:
samba 2.0.7 (compiled on the target system)
suse linux 6.3, kernel 2.2.13 with SMP
running on a
Dual PIII-550 with 256MB RAM, 2 x 18GB HD with ICP RAID-Controller
samba config:

[global]
  debuglevel = 1
  os level = 65
  time server = yes
  domain master = yes
  local master = yes
  preferred master = yes
  wins support = yes
  security = USER
  encrypt passwords = yes
  printing = LPRNG
  workgroup = public
  domain logons = yes
  name resolve order = host wins
  logon script = startup.bat
  logon path = \\%N\%U\profile
  logon home = \\%N\%U\profile
  keepalive = 300
  utmp = true
  utmp consolidate = true

Two other machines with samba 2.0.7 which use the "panic samba" as the
password and wins server have similar configs (no domain logon/master, no
wins support of course) and never panic!

Another strange behaving is that my utmp entries are not cleared. Even if i
stop samba i get a list of "ghost" users when executing the "who" command.
This entries seem to stay for ever in the system.

And last but not least there seems to be a bug in the log rotation. The old
log "log.smb.old" start e.g. at 2000/05/19 09:51:51 and ends up on
2000/06/20 18:29:48. The actual log then start as 2000/06/14 07:34:06 (!!!)
and ends up at 2000/06/20 19:28:58. The two log files are "overlayed". Maybe
this error is caused by the panic.

Did anyone experienced similar problems? Has anyone a explanation or
solution for this problems? Any hints are welcome.

Best regards

Rudolf Kollien
email: Rudolf.Kollien at medas.de
       Rudolf.Kollien at kollien.de
*********************************************************
Never trust a operating system you have no sources for
*********************************************************
Buying an operating system without source is like buying
a self-assembly Space Shuttle with no instructions.
*********************************************************