Bug report

[Josh Kelley]

About a month ago we upgraded to Samba 2.2.1a, and apparently ever since 
then we've been getting segfaults on a fairly regular basis - usually one 
or two a day, sometimes skipping a couple of days.  Apparently this doesn't 
cause any problems on the client end, otherwise we probably would have 
noticed sooner; the only indication I've received of the errors is the 
messages in the Samba log.  We're getting these errors on both of our main 
Samba servers.  Both are PCs running RedHat 6.2, one is running a linux 
2.4.9 kernel, the other is running a linux 2.2.16 kernel.  A number of 
different client machines have been causing the errors over the past month, 
so I don't think it's anything specific to one client or user.  I did check 
the OS for all of the clients that I could; one client was a Windows NT 
4.0, all of the other clients were Win2K.  I don't know if this is relevant 
or not.

Here's a sample message from the log file:
[2001/09/27 11:36:28, 3] smbd/process.c:process_smb(837)
   Transaction 89 of length 45
[2001/09/27 11:36:28, 3] smbd/process.c:switch_message(650)
   switch message SMBclose (pid 17602)
[2001/09/27 11:36:28, 3] smbd/sec_ctx.c:set_sec_ctx(317)
   setting sec ctx (1897, 100) - sec_ctx_stack_ndx = 0
[2001/09/27 11:36:28, 3] smbd/sec_ctx.c:set_sec_ctx(322)
   2 user groups:
   100 1100
[2001/09/27 11:36:28, 3] smbd/vfs.c:vfs_ChDir(643)
   vfs_ChDir to /tmp
[2001/09/27 11:36:28, 3] smbd/process.c:process_smb(837)
   Transaction 90 of length 120
[2001/09/27 11:36:28, 3] smbd/process.c:switch_message(650)
   switch message SMBtrans2 (pid 17602)
[2001/09/27 11:36:28, 3] smbd/sec_ctx.c:set_sec_ctx(317)
   setting sec ctx (1897, 100) - sec_ctx_stack_ndx = 0
[2001/09/27 11:36:28, 3] smbd/sec_ctx.c:set_sec_ctx(322)
   2 user groups:
   100 1100
[2001/09/27 11:36:28, 3] smbd/vfs.c:vfs_ChDir(643)
   vfs_ChDir to /home/myoung
[2001/09/27 11:36:28, 0] lib/fault.c:fault_report(40)
   ===============================================================
[2001/09/27 11:36:28, 0] lib/fault.c:fault_report(41)
   INTERNAL ERROR: Signal 11 in pid 17602 (2.2.1a)
   Please read the file BUGS.txt in the distribution
[2001/09/27 11:36:28, 0] lib/fault.c:fault_report(43)
   ===============================================================
[2001/09/27 11:36:28, 0] lib/util.c:smb_panic(1101)
   PANIC: internal error
[2001/09/27 11:36:28, 2] smbd/connection.c:count_fn(97)
   pid 17602 doesn't exist - deleting connections 3 [IPC$]
[2001/09/27 11:36:28, 2] smbd/connection.c:count_fn(97)
   pid 17602 doesn't exist - deleting connections 1 [homedir]
[2001/09/27 11:36:28, 2] smbd/connection.c:count_fn(97)
   pid 17602 doesn't exist - deleting connections 2 [IPC$]

I have log files (debug level 3) from two clients that triggered segfaults 
today; one client's log is around 500kB, the other is around 5MB.  If you 
want me to send all or part of these log files, I can.  If there's any 
other information (testparm output, logs at a higher debug level, etc; I'm 
still working on getting a core dump) that I can send, or anything else I 
can do to help track this problem down, please let me know.  I'm assuming 
that since we haven't noticed this problem for a month, we're not in danger 
of losing any data from the segfaults, but if you can let me know if you 
think it would be safer to go back to samba 2.0.x until this bug is fixed, 
I would greatly appreciate it.

Josh Kelley
System Administrator
Johnson Bible College