Re: [fuse-devel] inode number recycle for directories

[Miklos S. <mi...@sz...>]

On Sat, 12 Jul 2008, Anand Avati wrote:
> > Generation numbers don't help.  The problem is that the dentry cache
> > for /mnt1 doesn't know about the rmdir on /mnt2.
> 
> 
> Doesn't NFS have the same issue? why dont NFS clients give the same problem?
> Is it because NFS clients chose to forget the old inodes they dont see this
> issue?

I think NFS includes the generation number in its file handle, and so
it will index uniqe inodes for the different generations.

Fuse could do this, but it doesn't...

> Just curious of how you plan to implement it. Would it be through /dev/fuse
> by extending the kernel-user protocol where requests can also originate from
> userspace? This would definitely be a very desired feature!

Yes.

Miklos