samba - Nov 2001 - prevent infinite loop with recurse option enabled?

I am creating a web-based gui to provide searching of files shared via 
SMB on a local network.  It goes through and finds each computer, its 
shares, and its files using smbclient.  Currently, a nasty (and 
incredibly slow) set of loops in php recurses the directories on each 
share using a smbclient call for each directory up to a set depth.

I know for a fact of several computers on my LAN with links/shortcuts 
that would lead to infinite recursion if no depth limit or other 
measure was used to prevent this.  I've tried smbclient's recurse 
option, which is of course *much* faster, but it unfortunately follows 
the links for as long as I've allowed it to run.

Is there any current way to set a depth limit, or even better, just 
simply prevent visiting a directory twice (I've heard that wget hashes 
the directory contents and then compares them upon each new directory 
to prevent recursion at all...) ??  Any help would be *greatly* 
appreciated.  I am not opposed to coding something into smbclient 
myself, I actually prefer doing that instead of doing it outside of 
smbclient.  If nothing is currently available, pointers/suggestions 
would be great & I will try to put something together for all to 
benefit...

Thanks
Justin

I'm not sure how helpful this will be, but here goes:

AFAIK SMB (and thus anything using it) is totally unaware of the idea of
circularly linked filesystems because they don't happen in Windows (unless
your FS is screwed. Incidentally this is also the reason I blithely assume
that you're talking about Samba servers rather than Windows boxes. Complain
if I'm wrong). I don't know of any way you could deal with this at the
client side, short of checksumming the directory contents, and even that
would be potentially harmful (no way to tell for sure that it is a repeat
directory and not just a very similar one).

What I know of that can be done is rather less elegant than a clean
client-side solution would be, but if you don't really depend on following
any symbolic links in the search process, you can work around the problem
by duplicating every share in your smb.conf file (changing the name in a
consistent way, like adding "search_" at the beginning) and then in
the new
share definitions set 
browsable = no
writable = no
guest only = yes
follow symlinks = no

Respectively these will prevent the share from showing up in browse lists,
force it to be read only, not require authentication for read access
(meaning you don't need to have your script spit real passwords around over
the network) and prevent the Samba server from following symbolic links for
the client. Unless you've been hard linking directories this should solve
the recursion depth problem. Of course if you've got large numbers of
servers/shares the cure may be worse than the disease.
I think it should be possible using the fancy machine-name-specific include
features in smb.conf to have appropriate options come on just for specific
clients (i.e. the machine that runs the search program), but I've never
tried this. I'm sure some of the guru's can say if and how it could be
done.

At 05:49 PM 11/5/01 -0500, Justin Yackoski wrote:
>Is there any current way to set a depth limit, or even better, just 
>simply prevent visiting a directory twice (I've heard that wget hashes 
-- 

Clinical Psychologist: Someone who tries to figure out whether an 
infant has more fun in infancy than an adult has in adultery.

Chris Watt wrote:
> At 09:09 AM 11/6/01 -0500, Justin Yackoski wrote:
> 
>> I may be wrong, but it 
>>seems to me that the best solution is to checksum/hash the directory 
>>contents and name, and I find it hard to believe that two directories 
>>would be exactly identical very often...
>>
> 
> Do you get the same effect if you actually smbmount the filesystem and do a
> ls -R on the mountpoint?
> I guess you probably would actually. . .

Yes, actually I've determined that whether I use smbmount or 
smbclient, I can recurse until I get 40 directories deep in the share. 
  This is the case no matter how I do the recursion, so I assume 40 is 
hardcoded someplace as the max limit.

I noticed on page 1 of the wget manpage, 2nd P under Description: 
"Infinite recursion loops are always avoided by hashing the retrieved 
data."  Is this for some reason not applicable to smbclient, or would 
it be useful to add this, or should I be posting this to the samba 
developer list??
> Ok, the best solution I can suggest is that you use a contents checksum or
> depth limit 

Do you mean in smbclient?  So currently thats not possible, you're 
suggesting it is useful to add such a feature/option?
> It may take you a fair chunk of time to update the index, but as long as
> you don't write to the same file your PHP script is trying to read
from,
> you can do this in parallel with processing queries (if you need to update
> the index during "office hours" when the Windows boxes happen to
be turned
> on). If you want to do complex queries (using info other than the filename)
> it might be worth sticking your index in mySQL (or something similar),
> which happens to play nicely with PHP.
I actually am using MySQL at the moment.  Indexing 100,000+ files with 
a text file (and doing searches!), well, yes that would be bad.

Thanks again for your help

Justin