Re: [PATCH] VFS: Suppress automount on [l]stat, [l]getxattr, etc.

[Trond Myklebust <Trond.Myklebust@netapp.com>]

On Mon, 2011-09-26 at 15:56 -0700, Linus Torvalds wrote: 
> On Mon, Sep 26, 2011 at 3:33 PM, Trond Myklebust
> <Trond.Myklebust@netapp.com> wrote:
> >
> > Lookup permission checks would be replaced with open permission checks
> > on the server.
> >
> > IOW: the operation could potentially fail due to a completely unrelated
> > issue.
> 
> Quite frankly, that still sounds like a "I'm trying to make a problem
> out of something that isn't actually a problem". And you still seem to
> be unwilling to admit that LOOKUP_FOLLOW is a problem that has
> actually been reported in real life, so you just cut out that part of
> my question about how this would be a bigger issue.

I never questioned your argument that we need to address the problem of
regressions. That's why I cut out the LOOKUP_FOLLOW part.

I just questioned your proposed solution because it seemed to introduce
new problems. We're in the 3rd iteration of a bugfix (the first being
the introduction of ->d_automount() and the second being Miklos's
patch), and I'm assuming nobody is interested in seeing a 4th iteration.

> But whatever. I can't really care, since it's a two-liner to add a new
> flag, and then it falls down to "if you want to follow automounts, you
> can set that flag instead".
> 
> Almost nobody is ever going to bother setting the flag anyway, since
> LOOKUP_OPEN and LOOKUP_DIRECTORY are going to take care of all the
> common cases.
> 
> So here. You can set LOOKUP_AUTOMOUNT to force an automount traversal. Ok?

I assume that means that we can get rid of LOOKUP_NO_AUTOMOUNT, and just
replace the convoluted logic in follow_automount() with a test for
LOOKUP_AUTOMOUNT? If we have agreed on a default behaviour, then that
would seem cleaner than enumerating all these exceptions.

Cheers
  Trond

-- 
Trond Myklebust
Linux NFS client maintainer

NetApp
Trond.Myklebust@netapp.com
www.netapp.com