NFS client loops on mountpoint -- help

[Pascal Hoebanx]

Problem :

	when a Sun NFS server is requested by a client to lookup the ".." 
entry of a NFS mountpoint, it returns the file handle of the current entry
thereby preventing access to data outside the mounted subtree. 
The client then moves up into the part of the file tree onto which the
imported file system is mounted instead of into the imported file system 
itself, whose root had been reached anyway.

	I am currently writing a Vax NFS server which behaves in the same way
in the same condition, but the resulting effect is, surprisingly enough, that
the client Sun workstation does not leave the imported file system when 
reaching the ".." entry of the mountpoint but remains on the root entry of the
imported file system.

	I suspect that the contents of a file handle are not as opaque as Sun
claims them to be. (They contains among other data, the inode number and file
system identification).

Question :

	Could someone tell me either the correct contents of a file handle
or the way used by Sun to detect "cd .." from a mountpoint (in the equivalent
of the old namei procedure).

Please reply by mail.


Thanks in advance

					Pascal Hoebanx

					Unite d'Informatique
					Universite Catholique de Louvain
					Place Sainte Barbe, 2
					B-1348 Louvain-la-Neuve
					Belgium

e-mail : UUCP : seismo!prlb2!lln-cs!ph