Samba and NIS+ (lengthy desperate plea)

[Jason Tackaberry]

Hi everyone,

Let me briefly describe my setup.  I work at a small university and we are
setting up 3 labs.  2 are filled with Ultra Sparc 5's, and the 3rd will be
filled with PCs running NT4 workstation.  We have an Enterprise Server
3500 running NIS+ to authenticate the Unix labs (or rather we will, once
I get it working), and I'd like to set up Samba to authenticate the NT
labs using the same login/passwd that they have on the Unix server.  The
server is Solaris 2.6, and that's where samba will run.

Now I assume I'll have to do some magicalness to handle synchronizing
between the NIS+ passwd table and samba, but I haven't gotten that far
yet.  At this point I just want to set up Samba to use encrypted
passwords, but I am having major troubles with smbpasswd.  I am using the
samba 2.0.3 binary package from the samba website.

I'm fairly new to Solaris and NIS+ so it's entirely possible I've botched
up things which as a result are breaking smbpasswd.  However, I'm able to
use the nis* suite of commands and there's no indication to me that
anything's broken.  I just thought I should warn you that you're dealing
with a potentially clueless NIS+ admin here. :)

When I run smbpasswd -a to add a new user, it complains about a
non-existent database and then core dumps.  The output of smbpasswd and a
gdb backtrace of the core dump is appended to the end of this message.

Desperate, and thinking the binary package for Solaris was broken, I
retrieved the latest CVS sources.  Unfortunately, when compiling them
--with-nisplus I got a linker error (with a slew of nisplus_initialize_*
functions) and quickly gave up.  Then I noticed in the CVS source's script
directory there was a mknissmbpwdtbl.sh.  After running it, I now have an
empty smbpasswd.org_dir table.  I thought this might fix my smbpasswd
problem, but it didn't.  I also ran the mknissmbpasswd.sh script, but it
referenced some field called gcos that the table creation script didn't
make.  Since these sources were from CVS I didn't give too much thought --
CVS is broken a lot, I realize that. 

I would _so much_ appreciate any direction on this.  Eventually I would
like to set up the server as a PDC to authenticate the NT lab and provide
printing and home directory shares, so any tips and pointers on that would
also be appreciated.

Thanks,
Jason.

---- snip ----

(gdb) run -a tack
Starting program: /opt/samba.package/bin/smbpasswd -a tack
pm_process() returned Yes
load_client_codepage: loading codepage 850.
Derived broadcast address 199.212.55.255
Added interface ip=199.212.55.220 bcast=199.212.55.255 nmask=255.255.255.0
New SMB password:
Retype new SMB password:
make_smb_from_nisp: NIS+ lookup failure: Database for table does not exist
add_nisppwd_entry: nis_list failure:
[name=tack],/opt/samba/private/smbpasswd: Database for table does not
exist

Program received signal SIGSEGV, Segmentation fault.
0xef5a4614 in strlen ()
(gdb) bt
#0  0xef5a4614 in strlen ()
#1  0xef695164 in xdr_string ()
#2  0xef69b130 in xdr_nis_name ()
#3  0xef69eeac in xdr_nis_object ()
#4  0xef69b2d8 in xdr_array ()
#5  0xef69f528 in xdr_nis_result ()
#6  0xef69bb5c in xdr_free ()
#7  0xef69f854 in nis_freeresult ()
#8  0x38a48 in add_nisp21pwd_entry ()
#9  0x390ec in add_nisppwd_entry ()
#10 0x35900 in add_smbpwd_entry ()
#11 0x39258 in add_new_user ()
#12 0x393d8 in local_password_change ()
#13 0x2322c in password_change ()
#14 0x236cc in process_root ()
#15 0x23b70 in main ()