Re: [patch 0/5] seqlock consolidation

[Thomas Gleixner <tglx@linutronix.de>]

On Thu, 15 Mar 2012, Al Viro wrote:

> On Thu, Mar 15, 2012 at 08:17:12PM +0100, Thomas Gleixner wrote:
> > > +			hits += replace_path(&fs->root, old_root, new_root);
> > > +			hits += replace_path(&fs->pwd, old_root, new_root);
> > 
> > Wouldn't it be simpler to just do: 
> > 
> > +			count += replace_path(&fs->root, old_root, new_root);
> > +			count += replace_path(&fs->pwd, old_root, new_root);
> > 
> > > +			write_seqcount_end(&fs->seq);
> > > +			while (hits--) {
> > >  				count++;
> > 
> > Instead of that loop ?
> 
> This loop also contains path_get_longterm() and we need to do it before
> dropping fs->lock.  We are holding a reference to new_root, all right,
> but once it's place into ->fs->{root,pwd} of another task and ->fs->lock
> is dropped, there's nothing to stop that task of doing chdir() and dropping
> its reference.  Which could outweight the single reference we are holding
> pretty soon...

Oops, sorry I missed that path_get_longterm() in the loop due to the
+/- forrest around it.