samba – using non-standard host names
This article shows you how to use non-standard host names. By default, samba
uses the hostname of the machine on which it is running. In some situations this can
cause a problem.
The problem
Some time ago, I installed samba and recently I
found it wasn’t working. I could not see any services from my NT box. So I
went back to my original article and found a reference to the samba diagnostics.
It is frequently difficult to find the DIAGNOSIS.TXT file referred to by so many SAMBA
documents. It seems to move around the website. Hopefully this will settle down. I suggest
trying the following in the order provided:
- http://www.cise.ufl.edu/help/software/doc/samba/DIAGNOSIS.txt
- A copy of the above file stored on my server
- this Google search
Good luck
The diagnostics
I started going through that diagnostic document step by step. Things started
failing on step 3:
# smbclient -L fred
No interface found for address 210.55.152.36
Added interface ip=210.55.152.36 bcast=210.55.152.255
nmask=255.255.255.0
startlmhosts: Can't open lmhosts file /usr/local/etc/lmhosts.
Error was No such file or directory
Server time is Sat Jun 12 21:03:45 1999
Timezone is UTC+12.0
Domain=[FAB] OS=[Unix] Server=[Samba 1.9.18p8]
security=share
Server=[FRED] User=[root] Workgroup=[FAB] Domain=[FAB]
Sharename Type Comment
--------- ---- -------
backup Disk Backup directory
ftp Disk ftp server file area
IPC$ IPC IPC Service (Samba 1.9.18p8)
lp Printer local line printer
public Disk public access area
root Disk Home Directories
The first step was to create /usr/local/etc/lmhosts. So I did this:
# touch /usr/local/etc/lmhosts
So onto step 4 of the diagnostics.
# nmblookup -B ducky __SAMBA__ Sending queries to 192.168.0.20 name_query failed to find name __SAMBA__
This tells me that nmbd is incorrectly installed. So I tried starting
it:
# /usr/local/sbin/nmbd -D
Checking the log files I found this:
# tail /var/log/log.nmb
1999/06/12 21:44:59 netbios nameserver version 1.9.18p8 started
Copyright Andrew Tridgell 1994-1997
No interface found for address 210.55.152.36
bind failed on port 137 socket_addr=210.55.152.36
(Can't assign requested address)
make_subnet: Failed to open nmb socket on interface
210.55.152.36 for port 137.
Error was Can't assign requested address
ERROR: Failed when creating subnet lists. Exiting.
I didn’t understand that. That IP address didn’t belong to my box. It
belonged to my router. So I started composing a message to the questions mailing
list. Then I had an idea. I checked the man pages for nmbd(8).
That’s when it struck me: samba is using hostname. DOH! I’ll
explain later why this is wrong for me later.
Outdated information
There was another problem with samba. It was holding outdated
information. Here’s what I found in another log file:
# tail /var/log/samba/log.smb
Get_Hostbyname(wocker.dvl-software.co.nz):
lookup failureMatchname failed on
wocker.dvl-software.co.nz 192.168.0.99
Get_Hostbyname(wocker.dvl-software.co.nz):
lookup failureMatchname failed on
wocker.dvl-software.co.nz 192.168.0.99
Get_Hostbyname(wocker.dvl-software.co.nz):
lookup failureMatchname failed on
wocker.dvl-software.co.nz 192.168.0.99
Get_Hostbyname(wocker.dvl-software.co.nz):
lookup failureMatchname failed on
wocker.dvl-software.co.nz 192.168.0.99
Get_Hostbyname(wocker.dvl-software.co.nz):
lookup failureMatchname failed on
wocker.dvl-software.co.nz 192.168.0.99
Gethostbyaddr failed for 192.168.0.99
No interface found for address 210.55.210.87
First, this hostname is long out of date. I’ve not used that hostname for months.
And the IP address 210.55.210.87 was very out of date. My box has a DHCP
connection. The IP address was quite different now. So I killed and restarted samba:
# ps -auwx | grep smb root 58 0.0 0.6 134 72 ?? Is 10Mar99 0:01.69 /usr/local/sbin/smbd -D root 663 0.0 1.4 38 180 p0 R+ 9:17PM 0:00.15 grep smb # kill -quit 358 # /usr/local/sbin/smbd -D
Then a the log file contained:
Got SIGHUP No path in service printers - using /tmp No interface found for address 210.55.152.36
Good! Now we’re getting somewhere. Slowly.
Why change the hostname?
My current topology is rather unusual. If you
look up the hostname of my box, it doesn’t match any of the network interfaces connected
to my box. Here’s what I mean:
# ifconfig ed0
ed0: flags=c843<up,broadcast,running,simplex,link2,multicast> mtu 1500
inet 192.168.1.1 netmask 0xffffff00 broadcast 192.168.1.255
ether 00:80:ad:df:f5:db
# ifconfig ed1
ed1: flags=8843<up,broadcast,running,simplex,multicast> mtu 1500
inet 192.168.0.20 netmask 0xffffff00 broadcast 255.255.255.0
ether 00:80:ad:df:f5:d6
# hostname
ns.freebsddiary.cx
# nslookup ns.freebsddiary.cx
Server: localhost.freebsddiary.cx
Address: 127.0.0.1
Name: ns.freebsddiary.cx
Address: 210.55.152.36
As you can see, neither ed0 nor ed1 matches up to the hostname of
my computer. That is because 210.55.152.36 is my ADSL modem/router. That is
why nmbd cannot open a socket on that address. It’s on another device.
[insert big grin here]
What I tried first
Initially, I thought we could tell samba to use a name other than that
returned by hostname. Specifically, I thought that setting the following
in /usr/local/etc/smb.conf would suffice:
netbios name = fred
But I couldn’t get that to work. So I took another approach.
Changing the hostname
It’s time to change the hostname of my FreeBSD box.
# hostname ns.freebsddiary.cx # hostname fred.freebsddiary.cx # hostname fred.freebsddiary.cx
All done. Now when I start nmbd, I get this in the logs:
$ tail log.nmb 1999/06/12 23:26:34 netbios nameserver version 1.9.18p8 started Copyright Andrew Tridgell 1994-1997 standard input is not a socket, assuming -D option
And a quick check shows that both smbd and nmbd are running.
# ps -auwx | grep mbd root 228 19.0 4.0 72 58 p2 R+ 11:26PM 0:00.20 grep mbd root 282 0.0 0.6 104 8 ?? Is 10:53PM 0:00.10 /usr/local/sbin/smbd -D root 225 0.2 6.0 144 86 ?? Ss 11:26PM 0:00.30 /usr/local/sbin/nmbd
Continuing with the diagnostics
And now step 4 gives this:
# nmblookup -B fred __SAMBA__ Sending queries to 192.168.0.20 192.168.0.20 __SAMBA__<00>
That’s much better! [insert sigh here]
The rest of the diagnostics proceeded without incident. samba is now
running fine!
Thanks
My thanks to Andrew Tridgell of the Samba team for creating the diagnostics checklist.