On Tue, 13 May 2014 11:00:28 +1000
Dave Chinner <david@xxxxxxxxxxxxx> wrote:
> On Mon, May 12, 2014 at 08:35:05PM -0400, Jeff Layton wrote:
> > On Tue, 13 May 2014 08:58:45 +1000
> > Dave Chinner <david@xxxxxxxxxxxxx> wrote:
> > > On Mon, May 12, 2014 at 11:06:29AM -0400, Jeff Layton wrote:
> > > > This patchset does some general cleanup of the locktest binary, adds
> > > > some infrastructure to allow testing F_GETLK requests, and adds a new
> > > > F_GETLK test to the pile.
> > > >
> > > > The main impetus here is a regression that I caused in F_GETLK handling
> > > > for v3.15. The patch is making its way to Linus now, but I want to be
> > > > sure that it doesn't regress in the future.
> > >
> > > So do these changes cause locktest to fail on older kernels? i.e.
> > > does changing the test cause the locktest tests to fail where
> > > previously they passed? If so, we're going to have to make this a
> > > little more complex...
> > >
> > I haven't tested on much in the way of older kernels, but I wouldn't
> > expect it to cause any problems. The only behavior change that should
> > be introduced is the F_GETLK test, and older kernels should pass that
> > just fine (modulo v3.15 which has a regression that should be patched
> > soon). The rest of the changes are just cleanups, and shouldn't
> > introduce any behavioral changes.
> Is this the regression in question?
> +Server failure in 29:Verify that F_GETLK for F_WRLCK doesn't
> require that file be opened for write
Yes, that's the test I'm adding for the kernel regression. The patch
for the bug is:
[PATCH v2] locks: only validate the lock vs. f_mode in F_SETLK codepaths
...and I sent a pull request to Linus for it today. It should go in
fairly soon, I hope.
Jeff Layton <jlayton@xxxxxxxxxxxxxxx>