Re: [PATCH 1/3] autofs4 - fix device ioctl mount lookup

[Al Viro <viro@ZenIV.linux.org.uk>]

On Tue, Sep 03, 2013 at 07:18:42PM -0700, Linus Torvalds wrote:
> On Tue, Sep 3, 2013 at 7:00 PM, Al Viro <viro@zeniv.linux.org.uk> wrote:
> >
> > That aside, I'm really not happy with this kind of games; this stuff clearly
> > belongs in fs/namei.c where we can simply see the last component.  Doing that
> > on the level of "let's scan the pathname for slashes, etc." is just plain
> > wrong.  Let's step back for a minute here; what are you trying to do?
> > You have a pathname that should resolve to a mountpoint, without triggering
> > automount (or crossing into the mountpoint, for that matter) and you want
> > struct path for the bottom of that mount stack?  Or is it something
> > completely different?
> 
> Can we add a LOOKUP_NOAUTOMNT bit or something (not exposed to user
> space, only used for this particular kern_path() call). Then, if/when
> automount gets called recursively (through autofs4_lookup? Or is it
> just the autofs4_d_automount() interface?) it can just decide to not
> follow that last path.
> 
> Hmm? I don't think we pass in the lookup-flags to d_automount, but
> that could be changed. Yes?

No need, really - what we ought to do, AFAICS, is what umount(2) needs.
I've applied slightly modified variant of Jeff's "vfs: allow umount to handle
mountpoints without revalidating them" (modified by just leaving the
struct path filled with mountpoint and leaving the equivalent of follow_mount()
to caller) to the local queue and I'm pretty sure that it's what we want
here as well.

I'll push for-next tonight and send a pull request your way tomorrow, if you
are OK with that.