[linux-cifs-client] [PATCH 0/2] posix locks behaviour on Windows server

[Pavel Shilovsky]

On Tuesday 24 February 2009 18:21:49 you wrote:
> On Tue, 2009-02-24 at 18:00 +0300, Pavel Shilovsky wrote:
> > simo wrote:
> > > On Tue, 2009-02-24 at 17:36 +0300, Pavel Shilovsky wrote:
> > >> Hello!
> > >>
> > >> I had a problem with cifs connected with posix locks behaviour during
> > >> closing files when I mounted Windows share. When any descriptor of
> > >> file is  being closed, it deletes only locks that connected with it,
> > >> but it isn't posix. This patch provide posix locks behaviour on
> > >> windows server - deletes all locks that connected with inode.
> > >>
> > >> Summary of changes:
> > >> [1/2] patch adds cifs_lock_storage module
> > >> [1/2] patch adds use cifs_lock_storage module operations in file.c
> > >
> > > Do we really want to "fix" one of the most horrible and harmful posix
> > > semantics ? :-)
> >
> > Yes, it in congruence with posix standard and noncompliance of it can
> > break work of some applications.
>
> I know, it was just a pun to this horrible semantic.
> I think it is a good thing to be standard compliant (maybe with an
> option not to? :-)

Mount option? May be, but i think posix behaviour should be by default.

>
> Btw, I wonder if there really is any app that *depend* on this
> semantic ?

Of course, Wine! We have a problem with it, because we expect posix behaviour 
on locks, but on Windows share it isn't so. It adds many troubles in 
implementation.

> I think it is more probable that there are applications that misbehave,
> very few realize that all locks are dropped when any file descriptor
> connected with the same file is closed, but posix, is posix is posix :/
>
> Simo.

-- 
Best regards,
Pavel Shilovsky.