Weird problems with Samba 2.2.8 under Solaris 8 + latest kernel patch

[Pierre Belanger]

Hello all,

This weekend, we upgraded our Samba servers to 2.2.8 (pre3
according to the include/version.h -- CVS "synced" this past
Saturday afternoon, EDT). I compiled this new release for
the following Solaris/kernel :

   Solaris 6 : kernel patch 105181-33
   Solaris 7 : kernel patch 106541-23
   Solaris 8 : kernel patch 108528-19

Prior to Solaris 8 108528-19, that was installed yesterday
*not by me* , we were running 108528-12. Solaris 8 with
kernel patch 108518-19 + latest Samba is causing us troubles.
ps : nothing changed in our smb.conf file / we had no problems
before (the fcntl() bug was not an issue for us, we only have
around ~ 150 concurrent connections on that machine).

There's no problems on the other boxes (Solaris 6 & 7), note
that we have much less connections on those boxes.

[Q] Is there anyone on this list running with the latest
     Solaris 8 (108528-19) kernel patch and with Samba 2.2.8?

After receiving a few complains, I decided to dig into the log
files. Here's what I found:

1- Many dptr_close() errors, more than usually.

   log.wcanomp1775:[2003/03/17 14:04:09, 0] smbd/dir.c:dptr_close(277)
   log.wcanomp1775:  Invalid key 256 given to dptr_close

2- Many oplock_break errors, much more than we had:

   [2003/03/17 15:32:49, 0] smbd/oplock.c:oplock_break(791)
   oplock_break: end of file from client
   oplock_break failed for file New Lisp/mbold.lsp (dev = 3d8000a,
   inode = 1467387, file_id = 15).
   [2003/03/17 15:32:49, 0] smbd/oplock.c:oplock_break(879)
   oplock_break: client failure in break - shutting down this smbd.
   [2003/03/17 15:32:49, 1] smbd/service.c:close_cnum(677)
   wcanomp2081 (10.10.92.33) closed connection to service imews
   [2003/03/17 15:32:49, 1] smbd/service.c:close_cnum(677)
   wcanomp2081 (10.10.92.33) closed connection to service site_doc
   [2003/03/17 15:32:49, 1] smbd/service.c:close_cnum(677)
   wcanomp2081 (10.10.92.33) closed connection to service docoss
   [2003/03/17 15:34:24, 1] smbd/service.c:make_connection(636)
   wcanomp2081 (10.10.92.33) connect to service site_doc as user imews
   (uid=2138, gid=240) (pid 4863)
   [2003/03/17 15:35:10, 0] smbd/oplock.c:request_oplock_break(1011)
   request_oplock_break: no response received to oplock break request to
   pid 4858 on port 56392 for dev = 3d8000a, inode = 825700, file_id = 15
   [2003/03/17 15:35:10, 0] smbd/open.c:open_mode_check(652)
   open_mode_check: exlusive oplock left by process 4858 after break !
   For file C 1505A/AA1710-W.dwg, dev = 3d8000a, inode = 825700. Deleting
   it to continue...
   [2003/03/17 15:35:10, 0] smbd/open.c:open_mode_check(656)
   open_mode_check: Existent process 4858 left active oplock.
   [2003/03/17 15:36:59, 1] smbd/service.c:make_connection(636)
   wcanomp2081 (10.10.92.33) connect to service site_doc as user imews
   (uid=2138, gid=240) (pid 4883)
   [2003/03/17 15:36:59, 0] smbd/dir.c:dptr_close(277)
   Invalid key 256 given to dptr_close
   [2003/03/17 15:36:59, 0] smbd/dir.c:dptr_close(277)
   Invalid key 257 given to dptr_close
   [2003/03/17 15:37:10, 0] smbd/oplock.c:process_local_message(397)
   process_local_message: Received unsolicited break reply - dumping
   info.
   [2003/03/17 15:37:10, 0] smbd/oplock.c:process_local_message(412)
   process_local_message: unsolicited oplock break reply from pid 4863,
   port 56392, dev = 3d8000a, inode = 825700, file_id = 15
   [2003/03/17 15:38:02, 1] smbd/service.c:close_cnum(677)
   wcanomp2081 (10.10.92.33) closed connection to service site_doc
   [2003/03/17 15:38:09, 1] smbd/service.c:make_connection(636)
   wcanomp2081 (10.10.92.33) connect to service site_doc as user imews
   (uid=2138, gid=240) (pid 4904)
   [2003/03/17 15:41:22, 1] smbd/service.c:close_cnum(677)
   wcanomp2081 (10.10.92.33) closed connection to service imews
   [2003/03/17 15:41:22, 1] smbd/service.c:close_cnum(677)
   wcanomp2081 (10.10.92.33) closed connection to service docoss

I will "downgrade" tonight to the previous version that we were
running prior to the upgrade, it says "2.2.8pre1" but I remember
taken that from CVS around February the 5th, according to the
installation date!!!

I wish I would have more time for this but I don't :-( I'll "find"
time tomorrow to let you know if the downgrade helped or not.

Cheers,
Pierre B.