Luke Scharf wrote: > [0] The only exception is that the kernel NFS server module can't hop > from one filesystem to another, so if /export and /export/home are two > different partitions, then if you mount /export on a remote-client > machine, /export/home appears to be empty.
and just to bore you all further... it's not that the server can't hop from one fs to another: it can, and it does, and you can see that in that the server actually presents both /export and /export/home in the pseudo-fs "namespace" it presents to clients. The reason that /export/home appears empty is because the client refuses to traverse further once it notices that it has strayed into a different server filesystem. This is to preserve expected semantics, e.g. to guarantee file-id uniqueness per-filesystem. The correct client action, when it detects a server filesystem boundary, is to automatically mount - i.e. mirror-mount - the new server filesystem, e.g. /export/home in your example, at the corresponding place in the client directory hierarchy. cheers, c.
