Re: NWFS rename() problem

From: Andi Kleen (
Date: Thu Aug 24 2000 - 17:12:26 EST

On Thu, Aug 24, 2000 at 03:59:27PM -0600, Jeff V. Merkey wrote:
> Andi Kleen wrote:
> >
> > On Thu, Aug 24, 2000 at 01:49:20PM -0600, Jeff V. Merkey wrote:
> > > It sounds like inode numbers are arbitrary, and so long as I have cached
> > > a hash structure in the inode (which I do), I can just make up inode
> > > numbers for files (provided they are unique), and use the
> > > inode->generic_ip pointer to access the actual directory number.
> > >
> > > Is this correct?
> >
> > Just note that you'll break the NFS server when you don't have stable
> > inode numbers in your file system (at least in 2.2, 2.4 or 2.5 may get an
> > better "file key" interface so that you could manage nfs file handles
> > yourself)

I forgot to mention: you always need a unique inode number because the
NFS clients require it. This inode number needs to be valid over
server reboots.

> > Also there are applications that make assumption about the inode numbers --
> > like the popular maildir format. They should be better unique over the
> > file system, not just over the referenced inodes.
> So if I make them unique across all volumes for the fs driver (i.e. a
> counter in the driver that generates unique numbers across all volumes)
> this should work? Yes?

A simple counter will not work, unless you store the counter result into
the "inode" and search the volumes to avoid collisions (would be kinda like
a pid)

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

This archive was generated by hypermail 2b29 : Thu Aug 31 2000 - 21:00:14 EST