Re: More fun with unmounting ESTALE directories.

[NeilBrown <neilb@suse.de>]

[-- Attachment #1: Type: text/plain, Size: 3604 bytes --]

On Mon, 18 Feb 2013 23:17:42 +0000 "Myklebust, Trond"
<Trond.Myklebust@netapp.com> wrote:

> On Tue, 2013-02-19 at 10:10 +1100, NeilBrown wrote:
> > On Mon, 18 Feb 2013 18:46:09 +0000 Al Viro <viro@ZenIV.linux.org.uk> wrote:
> > 
> > > On Mon, Feb 18, 2013 at 01:25:09PM +1100, NeilBrown wrote:
> > > 
> > > > I would be really nice if sys_unmount used a LOOKUP_MOUNTPOINT flag that
> > > > works a bit like LOOKUP_PARENT and LOOKUP_NOFOLLOW in that it skips the very
> > > > last step and returns the mounted-on directory, not the mountpoint that is
> > > > mounted there - or at least makes sure not revalidate happens on that final
> > > > mounted directory.
> > > 
> > > I don't think LOOKUP_MOUNTPOINT is a good idea.  For one thing, we have
> > > fairly few places that might want it, all of them in core VFS.  Might as
> > > well provide a separate function for them, a-la path_lookupat() vs.
> > > path_openat().
> > > 
> > > For another, we need to decide what to do with a really nasty corner case:
> > > 	a/b is a mountpoint, with c/d bound on it.
> > > 	c/d is a symlink to c/e
> > > 	c/e is a mountpoint
> > > What should umount("a/b", 0) do?  There are two possibilities - removing
> > > vfsmount on top of a/b or one on top of c/e...
> > > 
> > > We have the latter semantics; _that_ is what this GETATTR is about.  It's
> > > a fairly obscure corner case - the question is not even whether to follow
> > > symlinks, it's whether to follow _mounts_ on the last component.  Hell
> > > knows; I'm seriously tempted to change it "don't follow mounts" and see if
> > > anyone complains.  The only case when behaviour would change would be
> > > a symlink mounted somewhere (note that this is _not_ something that can easily
> > > happen; e.g. mount --bind does follow symlinks) and umount(2) given the
> > > path resolving to the mountpoint of that symlink.
> > 
> > Thinking about this some more, I now realise that my LOOKUP_MOUNTPOINT idea
> > was too simplistic and missed the real point.
> > 
> > The real point is that for unmount, we want to resolve the the path without
> > any reference to any filesystem at all - the lookup should be handled
> > entirely by the dcache.
> > Any mountpoint is pinned in the dcache, and consequently any ancestor of any
> > mount point also is.  So the dcache will lead us to the dentry that we want.
> > 
> > And the dentry is *all* we want.  It doesn't really matter what the inode is
> > like, or whether the filesystem thinks that the inode or name still exist.
> > All we need to do is find a dentry that must be  in the cache, and detach the
> > mount that is there.
> > 
> > Whether that is achieved by a LOOKUP_ flag or a separate lookup function
> > doesn't matter much to me.
> > 
> > I suspect we need to allow for passing a symlink to unmount, and the symlink
> > might not be in cache, so we cannot insist uniformly on only using the dcache.
> > However if a name is in the cache, and the cached data suggests that it is a
> > directory, then we should trust that and avoid referring to the filesystem.
> > 
> > umount is really quite unique in this.  All other times we lookup a path we
> > want to use the thing we found.  With umount, we want to stop using it.
>
> > ???
> 
> Add a umountat() syscall so that you can supply a file descriptor? :-)
> 

If I could get that file descriptor by opening some magic file in /proc which
led immediately to the mount point, then I'd say "yes please!".
Otherwise, I don't think it helps, and so support your ":-)".

NeilBrown

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]