Hi All!

I've just spent two afternoons figuring out why the hell the third
installation of Samba on my home network
does not work, but all the others do. Hopefully someone (besides me) will
benefit from hours of tcpdump-ing and
going insane only to write it of as a windows bug. "I guess no news is good
news"

LAN Setup:
----------------------------------------------------
1. Compaq notebook Win 2K (the company's PC)
(Work LAN)
IP: 10.99.16.38 / 255.255.224.0
DNS: 10.99.6.18
GW: 10.99.1.1
DOMAIN: xxx.local

(Home LAN)
IP2: 192.168.0.100 / 255.255.0.0
DNS2: 192.168.0.2

---------------------------
2. Server #1 - "kelly" RH Linux 7.3
IP: 192.168.0.2 / 255.255.0.0

IP2: 10.99.1.1 / 255.255.224.0 (This takes care of the packets going to
default gateway - Fake work LAN GW)
IP3: 10.99.6.18 / 255.255.224.0 (This takes care of DNS requests for work
LAN)

running services:
named (DNS - configured to listen to IP & IP3. Both work & home zones /
reverses defined)
smbd (SAMBA - configured only to allow Unix:NT user aliasing)
telnet, ftp, ssh, httpd

---------------------------
3. Server #2 - "charlie" RH Linux 7.3
IP: 192.168.0.1 / 255.255.0.0

running services:
smbd (configured same as above)
telnet, ftp, ssh, httpd


FACTS:
1. All hosts are able to ping any other by IP and/or hostname.
2. all nslookups use the same DNS and are 100% correct.
3. All hosts have firewalls switched OFF.
4. Network services such as ftp, telnet work perfectly.
5. Win2K is able to connect only to '\\kelly' (192.168.0.2)
6. Win2K responds "network path could not be found" when connecting to
'\\charlie'
7. both sambas are in the same workgroup
8. 'smbclient -L charlie' from CHARLIE works!
9. 'smbclient -L charlie' from KELLY works! Funny, ha?

NOTES:
Facts 8 and 9 were really confusing since the Win2k was able to connect only
to one of them.
After a while I've noticed in the TCPDUMP that the IP 10.99.16.38 keeps
popping up now and than.

Here is the command I've used:
tcpdump -ln -vv 'host 192.168.0.1 and port ! 23' | tee tcpdump.samba

Very cool for doing TCP dumps over telnet.

At first I figured that maybe the DNS reverse lookup zones somehow made the
notebook revert to the work IP, so I stopped DNS service and relied only on
/etc/hosts for the hostname resolution. All for nothing! Still no dice.
Finally I've deleted the work IP from the notebook and taaaa-daaaa both
Sambas are working perfectly???
Naturally, I got the same result when giving the notebook back it's work IP
(must have that for Monday) and adding yet another IP to CHARLIE
(10.99.16.50).

Here is the tcp dump from above:
----------------------------------
more tcpdump.samba
22:43:06.868022 192.168.0.100.1274 > 192.168.0.1.microsoft-ds: S [tcp sum
ok] 3054026780:3054026780(0) win 64240 (DF) (ttl
128, id 13588, len 48)

22:43:06.868299 192.168.0.1.microsoft-ds > 192.168.0.100.1274: R [tcp sum
ok] 0:0(0) ack 3054026781 win 0 (DF) (ttl 255, id 0, len 40)

22:43:06.868215 10.99.16.38.1275 > 192.168.0.1.netbios-ssn: S [tcp sum ok]
3054076997:3054076997(0) win 64240 (DF) (ttl 128,
id 13589, len 48)

22:43:07.362325 192.168.0.100.1274 > 192.168.0.1.microsoft-ds: S [tcp sum
ok] 3054026780:3054026780(0) win 64240 (DF) (ttl
128, id 13591, len 48)

22:43:07.362453 192.168.0.1.microsoft-ds > 192.168.0.100.1274: R [tcp sum
ok] 0:0(0) ack 1 win 0 (DF) (ttl 255, id 0, len 40)

22:43:07.863077 192.168.0.100.1274 > 192.168.0.1.microsoft-ds: S [tcp sum
ok] 3054026780:3054026780(0) win 64240 (DF) (ttl
128, id 13593, len 48)

22:43:07.863197 192.168.0.1.microsoft-ds > 192.168.0.100.1274: R [tcp sum
ok] 0:0(0) ack 1 win 0 (DF) (ttl 255, id 0, len 40)

22:43:09.866172 10.99.16.38.1275 > 192.168.0.1.netbios-ssn: S [tcp sum ok]
3054076997:3054076997(0) win 64240 (DF) (ttl 128,
id 13595, len 48)

22:43:15.875378 10.99.16.38.1275 > 192.168.0.1.netbios-ssn: S [tcp sum ok]
3054076997:3054076997(0) win 64240 (DF) (ttl 128,
id 13597, len 48)
-----------------------------

What gives with the last two packets?
I admit my IP knowledge aint that good, but what the hell is that IP even
doing communicating on a wrong subnet?
Pinging 192.168.0.1 from 10.99.16.38 (ONLY) results in an Network
Unreachable???

BOTTOM LINE:
I'll be happy if anyone can explain this to me. All this works fine now
using several more IPs than
would otherwise be necessary.

My guess here is that Win2K has a problem deciding which IP to use when
given multiple choice.

regards

Nitz