Dissappearing directories

[Trond Myklebust]

On Thu, 2007-01-25 at 14:23 +0100, Gabriel Barazer wrote:
> On 01/24/2007 18:18:16 +0100, "J. Bruce Fields" <bfields at fieldses.org> 
> wrote:
> 
> > On Wed, Jan 24, 2007 at 11:16:31AM -0500, Talpey, Thomas wrote:
> >> At 10:59 AM 1/24/2007, J. Bruce Fields wrote:
> >>>> Is "amazon:/" a typo? You exported "/NFS" but you're mounting "/"...
> >>> That's the result of our dumb hand-constructed pseudofilesystem
> >>> namespace.  See
> >>>
> >>> 	http://wiki.linux-nfs.org/index.php/Pseudofilesystem_improvements
> >>>
> >>> for some commentary and ideas about how to fix it....
> >> Oh, right. Any chance his server's nfs-utils-1.0.6 is doing the wrong thing
> >> and not setting up /NFS correctly then?
> > 
> > Maybe there was a nohide bug in that version?  Could you try also adding
> > the "crossmnt" export option to the export entries for /NFS?
> > 
> 
> BTW, the "exports" man page (dated from March 2005), states that 
> "nohide" and "crossmnt" options don't operate well when there is 2 files 
> with the same inode number on different mounted filesystem. Is this 
> still true (maybe there is somewhere a list of known bugs) ? Is it now 
> possible to use "nohide" or "crossmnt" on subnet exports ?
> 

It should just work with kernels >= 2.6.18. They will automatically
generate a new mountpoint for you, ensuring that the 2 filesystems stay
separate on the client. This again should ensure that the inode number
uniqueness rule applies (assuming that it applies on the server).

  Trond