Bug 2426 - nmbd crashes with INTERNAL ERROR (signal 6)
nmbd crashes with INTERNAL ERROR (signal 6)
Status: RESOLVED LATER
Product: Samba 3.0
Classification: Unclassified
Component: nmbd
3.0.11
All Linux
: P3 major
: none
Assigned To: Samba Bugzilla Account
Samba QA Contact
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2005-03-09 06:07 UTC by Simon Detheridge
Modified: 2005-03-10 07:02 UTC (History)
0 users

See Also:


Attachments
log.nmbd at level 10 (310.97 KB, text/plain)
2005-03-09 06:09 UTC, Simon Detheridge
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Simon Detheridge 2005-03-09 06:07:10 UTC
nmbd is crashing with both samba 3.0.11, and samba 3.0.12pre1. I'm not sure what
I'm doing to cause this, as it seems to crash after a period of time... Normally
a few minutes after being started.

Here is that the logfile spits out when it crashes: (I will attach the full
310kb level-10 logfile shortly)

-----------------------------------------------
[2005/03/09 13:01:06, 4] nmbd/nmbd_workgroupdb.c:dump_workgroups(284)
  dump_workgroups()
   dump workgroup on subnet  UNICAST_SUBNET: netmask=        0.0.0.0:
  	WIDGITSOFTWARE(3) current master browser = FILESERVER
  	LONDON(2) current master browser = LEPRECHAUN
  	SIMONS-HOUSE(1) current master browser = UNKNOWN
  		RUMPLESTILTSKIN 40099a03 (What's my name 3.0.12pre1)
[2005/03/09 13:01:06, 4] nmbd/nmbd_workgroupdb.c:find_workgroup_on_subnet(173)
  find_workgroup_on_subnet: workgroup search for SIMONS-HOUSE on subnet
UNICAST_SUBNET: found.
[2005/03/09 13:01:06, 4]
nmbd/nmbd_packets.c:retransmit_or_expire_response_records(1615)
  retransmit_or_expire_response_records: timeout for packet id 30638 to IP
10.0.0.200 on subnet UNICAST_SUBNET
*** glibc detected *** free(): invalid next size (fast): 0x801a5d20 ***
[2005/03/09 13:01:06, 0] lib/fault.c:fault_report(36)
  ===============================================================
[2005/03/09 13:01:06, 0] lib/fault.c:fault_report(37)
  INTERNAL ERROR: Signal 6 in pid 10268 (3.0.12pre1)
  Please read the appendix Bugs of the Samba HOWTO collection
[2005/03/09 13:01:06, 0] lib/fault.c:fault_report(39)
  ===============================================================
[2005/03/09 13:01:06, 0] lib/util.c:smb_panic2(1485)
  smb_panic(): calling panic action [/bin/sleep 900000]
-----------------------------------------------

I tried to generate a backtrace, but it doesn't look very useful. I did try and
compile with '-g' but I can only see one function name about 680 levels deep. I
know Gentoo strips binaries by default, so that could have something to do with
it... I did try and make it not strip samba...

Here is the end of it:


-----------------------------------------------
#668 0x00000000 in ?? ()
#669 0x00000000 in ?? ()
#670 0x00000000 in ?? ()
#671 0x00000000 in ?? ()
#672 0x00000000 in ?? ()
#673 0x40010f11 in ?? ()
#674 0x4000167b in ?? ()
#675 0x40016fb4 in ?? ()
#676 0x0000000a in ?? ()
#677 0xbffff810 in ?? ()
#678 0xbffff9f0 in ?? ()
#679 0x4000fd56 in ?? ()
#680 0x80000034 in ?? ()
#681 0x00000140 in ?? ()
#682 0xbffff810 in ?? ()
#683 0x00000003 in ?? ()
#684 0x0002041c in become_local_master_browser (subrec=0xbffff0ac, work=0x2)
    at nmbd_become_lmb.c:530
    #685 0x00000000 in ?? ()
(gdb) 
-----------------------------------------------

Here is my smb.conf:

-----------------------------------------------
[global]
   workgroup = SIMONS-HOUSE
   netbios name = RUMPLESTILTSKIN
   server string = What's my name %v
   log file = /var/log/samba3/log.%m
   max log size = 500
   panic action = /bin/sleep 900000
   log level = 10
   hosts allow = 192.168.1. 192.168.2. 192.168.3. 127. 10.0.0.
   security = user
   encrypt passwords = yes
   smb passwd file = /etc/samba/private/smbpasswd
   winbind use default domain = yes
   socket options = TCP_NODELAY SO_RCVBUF=8192 SO_SNDBUF=8192
   interfaces = eth1
   remote browse sync = 10.0.0.200 192.168.1.3
   remote announce = 192.168.1.3/SIMONS-HOUSE 10.0.0.200/SIMONS-HOUSE
   local master = yes
   os level = 99
   domain master = yes
   preferred master = yes
   name resolve order = wins lmhosts bcast
   wins support = no
   wins server = 10.0.0.200
   wins proxy = yes
   dns proxy = no 

[homes]
   comment = Home Directories
   browseable = yes
   writable = yes
   print command = lpr-cups -P %p -o raw %s -r

[movies]
	comment = Some films
	path = /home/movies
	read only = no
	browseable = yes
	public = no
	writeable = yes
	create mode = 0664
	directory mode = 0775

[goodies]
	comment = The large random stuff collection
	path = /home/goodies
	read only = no
	browseable = yes
	public = no
	writeable = yes
	create mode = 0664
	directory mode = 0775

[software]
	comment = Some installable software
	path = /home/software
	read only = no
	browseable = yes
	public = no
	writeable = yes
	create mode = 0664
	directory mode = 0775
-----------------------------------------------

The operating system is Gentoo linux, but I haven't used any custom cflags
(except for -g) before you ask. ;-)
Comment 1 Simon Detheridge 2005-03-09 06:09:15 UTC
Created attachment 1020 [details]
log.nmbd at level 10

This is the nmbd log file from the session that crashed, created at log level
10.
Comment 2 Jeremy Allison 2005-03-09 13:12:27 UTC
We really need a good backtrace generated with gcc -g. Can you fix gentoo
to do that ?
Jeremy.
Comment 3 Gerald (Jerry) Carter 2005-03-09 13:34:01 UTC
any chance this is related to the nmblob.c bug reported in bug 2147 ?
Comment 4 Simon Detheridge 2005-03-10 02:47:46 UTC
My apologies. In my attempts to get a proper backtrace, I recompiled with a
different GCC version. (GCC 3.3.5 instead of GCC 3.4.3)

The problem appears to have gone away. Looks like a GCC issue.
Comment 5 Gerald (Jerry) Carter 2005-03-10 07:02:30 UTC
Thanks for the feedback.  we'll do some additional testing 
under valgrind.