[jcifs] no name with type 0x00 with no scope for host

[Allen, Michael B (RSCH)]

> -----Original Message-----
> From:	Christopher R. Hertel [SMTP:crh at ubiqx.org]
> Sent:	Tuesday, February 19, 2002 8:53 PM
> To:	Allen, Michael B (RSCH)
> Cc:	'crh at ubiqx.mn.org'; jcifs at lists.samba.org
> Subject:	Re: [jcifs] no name with type 0x00 with no scope for host
> 
> "Allen, Michael B (RSCH)" wrote:
> :
> > > Use "*\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0" for NODE STATUS REQUESTS.
> > >
> >         Same thing. The 16th byte is <00>.
> 
> But you are not sending a name.  You're sending the wildcard.  If I do:
> 
> $ nmblookup -S foo
> 
> Then the first thing nmblookup does is send a name query for
> 
>   FOO\20\20\20\20\20\20\20\20\20\20\20\20\0
> 
> If it gets a response, the -S option will cause it to send a follow-up node
> query for the same name.  Yes, it ends with a <00>, but it still starts with
> "FOO".  We know this will work, however, because the positive name query
> response tells us that the name in question exists on the other node.
> 
> If you don't know what names are on the other node, then use a wildcard
> query.  Yes, it ends with a <00> but then entire name is different.
> 
> Note that Microsoft doesn't even bother to decode the names when doing
> comparisons.  Windows compares the level-1 encoded name itself.  So, it's
> looking for CKAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA.  That's a wildcard name to
> Windows.
> 
	Well, I'm not sure I understand what the significance of using the name verses the
	wildcard is but jCIFS *always* uses the wildcard because all methods that provoke the
	Node Status are doing so to determine the names available. If you look at the
	getAllByAddress it builds a NodeStatusRequest with the
	NbtAddress.ALL_HOSTS_NAME which is the "*\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0" name and
	with a  type of <00>.

> > >   It's a
> > > wierd name, I know, but it's actually in the RFCs.  It's an asterisk
> > > followed by 15 nul bytes, then encoded as usual.  It's the wildcard name.
> > > It is illegal to register it, but all NBT nodes are supposed to respond to
> > > it (within the proper scope ID, of course).  It is also the precursor to the
> > > "*SMBSERVER" name.
> > >
> >         What is, node status?
> 
> No, the wildcard name (just '*' followed by 15 nuls) is the precursor to
> *SMBSERVER.
> 
	I don't understand. The NodeStatus *is* the wildcard name with 15 nuls, a hex code of
	<00>, and a type of NBSTAT. The *SMBSERVER is used in the Session Establishment
	message and has nothing to do with Node Status. You might use Node Status to help
	figure out a good calling name though.

> > Actually at this point sending the node status *after*
> > SMBSERVER is a little better I think. The node status is a last resort.
> 
> Depending upon what you're doing.  I you are trying to find a called name in
> order to set up an SMB session, then it all depends upon whether you think
> there are more Samba, Win/NT4, W2K, and W/XP systems out there than W/9x and
> W/Me.  Probably a good guess, so *SMBSERVER should come before adapter
> status.
> 
	What's different about these servers? We always do the same thing when trying to do
	session est.:

	Try to use whatever name was provided
	    or use the first label if it looks like a dns name
	    or use the ip if that's all that was provided
	If that fails try SMBSERVER
	If that fails use ip if it was not already.
	Finally, try to figure out the name using Node Status.
	else throw UnknownHostException

	I think that's how it goes. The firstCalledName and nextCalledName code in UniAddress
	and NbtAddress shows clearly what happends. Now, I don't think I've ever seen this fail
	where another client succeeded so I'm interested to know what the OPs problem is
	with XP... :~(

> All I'm saying about Adapter Status is that you should use the wildcard name
> unless you *know* that a specific name (eg., FOO<00>) exists on the remote
> node.  From the original problem report, it looked as though the
> <machine><00> name was being used.
> 
	Ahh. No. The debugging message just printed it that way. jCIFS *always* uses the
	wildcard name for node status b/c it's not very useful to print out the "*\0\0\0\0\.." name. It
	prints the IP instead.

>   On XP, which doesn't register
> <machine><00> by default, you really do need to use the wildcard name.
> 
	We always do....

	Mike