Re: [GIT PULL] please pull file-locking related changes for v3.20

["J. Bruce Fields" <bfields@fieldses.org>]

On Thu, Feb 26, 2015 at 09:45:00AM -0500, J. Bruce Fields wrote:
> On Thu, Feb 26, 2015 at 11:00:46AM +0000, One Thousand Gnomes wrote:
> > On Tue, 17 Feb 2015 11:13:39 -0800
> > Linus Torvalds <torvalds@linux-foundation.org> wrote:
> > 
> > > On Tue, Feb 17, 2015 at 11:08 AM, J. Bruce Fields <bfields@fieldses.org> wrote:
> > > >
> > > > I agree that it's weird, but I think it's what we're stuck with.
> > > 
> > > And if by "weird" you mean "flock is really not a well-defined or sane
> > > interface", I'll agree with you.
> > > 
> > > That said, I'm not at all sure about the "we're stuck with it". We can
> > > improve the semantics without anybody noticing, because it's not like
> > > anybody could *depend* on the weaker semantics - they needed
> > > particular races and timings to hit anyway.
> > 
> > The BSD implementation does not documented such a race, or indeed appear
> > to have one. That implies that nothing using flock should have this
> > problem.
> 
> Which race are you talking about exactly, and what evidence are you
> working from?

To clarify: I previously conflated two issues:

	- the temporary drop of the spinlock in flock_lock_file().
	  Agreed that that's pointless, and has been fixed.

	- non-atomic flock upgrades: that's definitely documented
	  behavior on BSD.

--b.