Samba 3.0 CVS from 2003-07-25 smbd crashes when it tries to change his machine trust password. The domain is controlled by samba 3.0 DCs (3x, with LDAP backend),same CVS date. As the clients try it again and again I assume they're failing changing it. This happend on four machines nearly at the same time, which had been joined exactly one week ago. Linux distro is debian woody. There was no traffic on samba at that time. regards Dariush Forouher Aug 5 10:53:49 cassiopeia smbd[25975]: [2003/08/05 10:53:49, 0] lib/fault.c:fault_report(36) Aug 5 10:53:49 cassiopeia smbd[25975]: =============================================================== Aug 5 10:53:49 cassiopeia smbd[25975]: [2003/08/05 10:53:49, 0] lib/fault.c:fault_report(37) Aug 5 10:53:49 cassiopeia smbd[25975]: INTERNAL ERROR: Signal 11 in pid 25975 (3.0.0rc1) Aug 5 10:53:49 cassiopeia smbd[25975]: Please read the appendix Bugs of the Samba HOWTO collection Aug 5 10:53:49 cassiopeia smbd[25975]: [2003/08/05 10:53:49, 0] lib/fault.c:fault_report(39) Aug 5 10:53:49 cassiopeia smbd[25975]: =============================================================== Aug 5 10:53:49 cassiopeia smbd[25975]: [2003/08/05 10:53:49, 0] lib/util.c:smb_panic(1462) Aug 5 10:53:49 cassiopeia smbd[25975]: PANIC: internal error Aug 5 10:53:49 cassiopeia smbd[25975]: [2003/08/05 10:53:49, 0] lib/util.c:smb_panic(1469) Aug 5 10:53:49 cassiopeia smbd[25975]: BACKTRACE: 12 stack frames: Aug 5 10:53:49 cassiopeia smbd[25975]: #0 /usr/sbin/smbd(smb_panic+0x16b) [0x8179c43] Aug 5 10:53:49 cassiopeia smbd[25975]: #1 /usr/sbin/smbd [0x816a962] Aug 5 10:53:49 cassiopeia smbd[25975]: #2 /usr/sbin/smbd [0x816a9b5] Aug 5 10:53:49 cassiopeia smbd[25975]: #3 /lib/libc.so.6 [0x400e46b8] Aug 5 10:53:49 cassiopeia smbd[25975]: #4 /usr/sbin/smbd(make_nmb_name+0x2b) [0x80dc8a3] Aug 5 10:53:49 cassiopeia smbd[25975]: #5 /usr/sbin/smbd(cli_full_connection+0x78) [0x80c0348] Aug 5 10:53:49 cassiopeia smbd[25975]: #6 /usr/sbin/smbd(change_trust_account_password+0x105) [0x80b8705] Aug 5 10:53:49 cassiopeia smbd[25975]: #7 /usr/sbin/smbd [0x80b4fd4] Aug 5 10:53:49 cassiopeia smbd[25975]: #8 /usr/sbin/smbd(smbd_process+0x152) [0x80b5192] Aug 5 10:53:49 cassiopeia smbd[25975]: #9 /usr/sbin/smbd(main+0x68c) [0x81c8b50] Aug 5 10:53:49 cassiopeia smbd[25975]: #10 /lib/libc.so.6(__libc_start_main+0xbb) [0x400d414f] Aug 5 10:53:49 cassiopeia smbd[25975]: #11 /usr/sbin/smbd(ldap_msgfree+0x71) [0x8072ce1] Aug 5 10:53:49 cassiopeia smbd[25975]: # smb.conf [global] workgroup = BRGS security = domain password server = * syslog only = yes wins server = wins1 dos charset = CP850 unix charset = iso8859-15
we've had a lot of character set fixes in the past week or so. Any chance you could run 'cvs update' and see if the problem still occurs? I'm referring to the make_nmb_name() call.
Marking it as fixed for RC2 (due to all the multibyte character set work). Pleas reopen if that is not the case.
Sorry for this late answer, I wasn't at home last days. Unfortunatly the problem still persists, even after upgrading samba to current CVS on the member server and on the DCs. I've also tried and deleted secrets.tdb and rejoined the machine (and set 'machine password timeout' to a low value) without any success. regards Dariush
Created attachment 122 [details] log snapshot this is a level 5 log snapshot from the time when the panic happened. Hope it helps.
What we really need is a good stack backtrace, created when smbd is compiled with -g, and a panic action = /bin/sleep 999999 placed in the [global] section of smb.conf. Allow the crash to take place, then attach to the crashed process with gdb and send in the output of the "bt" command. This will give use the line numbers of the problem. Jeremy.
Created attachment 125 [details] smbd backtrace Here it goes. regards Dariush
Created attachment 132 [details] Proposed patch for m/c account password change crash.
Setting as fixed - patch was pretty obvious. Re-open if this doesn't cure it. Jeremy.
Yep, this solves the problem. thanks Dariush
originally reported against 3.0.0beta3. CLeaning out non-production release versions.
sorry for the same, cleaning up the database to prevent unecessary reopens of bugs.
database cleanup