Re: Local loopback mounts (RFD)

Alexander Viro (viro@math.psu.edu)
Mon, 31 May 1999 09:54:50 -0400 (EDT)


On Mon, 31 May 1999, Jan Kasprzak wrote:

> Hello all,
>
> I think the Linux dentry mechanism can allow us to have
> loopback mounts faster and more effective than NFS mount from 127.0.0.1. Note

Yes, but the way you have proposed is broken.

> The loopfs_read_super() just creates the root dentry and points
> its dentry->d_inode to the inode of the target directory. This has the
> advantage that both /some/directory and /another/directory works with
> same inodes, so file locks works properly, open(O_EXCL) works properly, etc.

... and one can trivially create loops in on-disk structure that way
(think of nasty games with rename()). We do need a way to compute the
foo-is-ancestor-of-bar relation.

> I want to ask you whether the new pseudo-filesystem is the right
> interface to have. Or should we do someting completely different (e.g. extend
> ext2fs to support non-tree topology, or do the same thing at the VFS layer,
> but not using mountpoints. What do you think about it?

stackable filesystems. In VFS. And there still are some things to change
in VFS/fs interface before it will become possible.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/