[linux-cifs-client] [PATCH] cifs: tighten up default file_mode/dir_mode

[Jeff Layton]

On Sun, 05 Apr 2009 14:19:08 +0000
simo <idra at samba.org> wrote:

> On Sun, 2009-04-05 at 07:33 -0400, Jeff Layton wrote:
> > 
> > Note that this patch also removes the mandatory locking flags from the
> > default file_mode. After having looked at how these flags are used by
> > the kernel, I don't think that keeping them as the default offers any
> > real benefit. That flag combination makes it so that the kernel
> > enforces
> > mandatory locking.
> > 
> > Since the server is going to do that for us anyway, I don't think we
> > want the client to enforce this by default on applications that just
> > want advisory locks. Anyone that does want this behavior can always
> > enable it by setting the file_mode appropriately.
> 
> If we don't enforce it, what happen on the client side?
> Do we do a synchronous write to the server to find out if the app was
> actually able to write? Or are we going to tell the app the write was
> fine to find out only later that we were not able to write back to the
> server ?

This really only changes the behavior if there is a lock set by another
task on the client itself. If there's a lock set on the server, then
the client isn't really aware of it anyway and can't enforce it.

So it really comes down to what the expected behavior is when a process
on the box is holding a lock on the file and another process wants to
do I/O to it. IMO, it makes sense not to have the kernel attempt to
enforce mandatory locking unless someone specifically wants it. They
can still get it by setting the mode appropriately.

-- 
Jeff Layton <jlayton at redhat.com>