MSDFS problem with hostname

dpk dpk at
Tue Jun 8 21:10:08 GMT 2004

I have a similar problem to the one found here:

I have two Debian GNU/Linux systems running Sarge with Samba 3.0.2a 
with a Foundry ServerIron load balancer using a "switch back" 
configuration, meaning, connections are not received on an Ethernet 
interface, but instead on a loopback interface with the IP address of 
the load balancer.

If I connect to \\server\dfs, clicking a share name "works".  All DFS 
pointers reference the FQDN, i.e., which works. 
However, if I map to \\\dfs, a Windows XP client 
receives the "... refers to a location that is unavailable..." error 
message.  A log level of 10 produces:

[2004/06/08 16:17:35, 10] msdfs/msdfs.c:parse_dfs_path(44)
   temp in parse_dfs_path:\dfs. after trimming \'s
[2004/06/08 16:17:35, 10] msdfs/msdfs.c:parse_dfs_path(53)
[2004/06/08 16:17:35, 3] msdfs/msdfs.c:get_referred_path(383)
   get_referred_path: Invalid hostname in path 
[2004/06/08 16:17:35, 3] smbd/error.c:error_packet(118)
   error packet at smbd/trans2.c(3301) cmd=50 (SMBtrans2) 

The FQDN resolves to the IP of the load balancer and that defined on the 
loopback interface, lo:0.  I can "host" and "ping" the FQDN on each 
Samba server.  I don't know Samba source internals that well, but is the 
check performed in msdfs.c only a "sanity" check to avoid admins 
creating invalid DFS pointers?  Could it be removed?  I removed nisplus 
hostname resolution and receive the same results, so I'm I don't think 
it is a problem there.  Could it be related to using lo:0?  Any thoughts 
or suggestions are welcome.


More information about the samba-technical mailing list