Date:   Mon, 23 Jan 2023 11:05:56 +0100
From:   Jan Kara <jack@suse.cz>
To:     Linus Torvalds <torvalds@linux-foundation.org>
Cc:     Andreas Gruenbacher <agruenba@redhat.com>,
        David Howells <dhowells@redhat.com>,
        Steve French <smfrench@gmail.com>,
        Theodore Ts'o <tytso@mit.edu>, Jan Kara <jack@suse.cz>,
        Trond Myklebust <trondmy@hammerspace.com>,
        Christoph Hellwig <hch@lst.de>,
        "Darrick J. Wong" <djwong@kernel.org>, cluster-devel@redhat.com,
        linux-kernel@vger.kernel.org,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>
Subject: Re: [GIT PULL] gfs2 writepage fix
Message-ID: <20230123100556.qxdjdmcms5ven52v@quack3>
References: <20230122090115.1563753-1-agruenba@redhat.com>
 <CAHk-=wgjMNbNG0FMatHtmzEZPj0ZmQpNRsnRvH47igJoC9TBww@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAHk-=wgjMNbNG0FMatHtmzEZPj0ZmQpNRsnRvH47igJoC9TBww@mail.gmail.com>
Precedence: bulk

On Sun 22-01-23 12:05:53, Linus Torvalds wrote:
> On Sun, Jan 22, 2023 at 1:01 AM Andreas Gruenbacher <agruenba@redhat.com> wrote:
> >
> > gfs2 writepage fix
> >
> > - Fix a regression introduced by commit "gfs2: stop using
> >   generic_writepages in gfs2_ail1_start_one".
> 
> Hmm. I'm adding a few more people and linux-fsdevel to the reply,
> because we had a number of filesystems remove writepages use lately,
> including some that did it as a fix after the merge window.
> 
> Everybody involved seemed to claim it was just a no-brainer
> switch-over, and I just took that on faith. Now it looks like that
> wasn't true at least for gfs2 due to different semantics.
> 
> Maybe the gfs2 issue is purely because of how gfs2 did the conversion
> (generic_writepages -> filemap_fdatawrite_wbc), but let's make people
> look at their own cases.

Thanks for the heads up. So we had kind of a similar issue for ext4 but it
got caught by Ted during his regression runs so we've basically done what
GFS2 is doing for the merge window and now there's patchset pending to
convert the data=journal path as well because as Andreas states in his
changelog of the revert that's a bit more tricky. But at least for ext4
the conversion of data=journal path resulted in much cleaner and shorter
code.

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR