Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20221106224841.279231-1-ebiggers@kernel.org> <Y4+ua2XftbAYd8xq@mit.edu>
In-Reply-To: <Y4+ua2XftbAYd8xq@mit.edu>
From:   harshad shirwadkar <harshadshirwadkar@gmail.com>
Date:   Fri, 9 Dec 2022 08:12:56 -0800
Message-ID: <CAD+ocbznGVej2myzU+3edpw0a_EXcVRjLAU=KS1ymXxbaEaz=Q@mail.gmail.com>
Subject: Re: [PATCH 0/7] ext4 fast-commit fixes
To:     "Theodore Ts'o" <tytso@mit.edu>
Cc:     Eric Biggers <ebiggers@kernel.org>, linux-ext4@vger.kernel.org,
        linux-fscrypt@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, 6 Dec 2022 at 13:04, Theodore Ts'o <tytso@mit.edu> wrote:
>
> On Sun, Nov 06, 2022 at 02:48:34PM -0800, Eric Biggers wrote:
> > From: Eric Biggers <ebiggers@kernel.org
> >
> > This series fixes several bugs in the fast-commit feature.
> >
> > Patch 6 may be the most controversial patch of this series, since it
> > would make old kernels unable to replay fast-commit journals created by
> > new kernels.  I'd appreciate any thoughts on whether that's okay.  I can
> > drop that patch if needed.
>
> Mumble.  Normally, it's something we would avoid, since there aren't
> that many users using fast commit, since it's not enabled by default.
> And given that the off-by-one errors are bugs, an it's a question of
> old kernels requiring a pretty buggy layout, the question is whether
> it's worth it to do an explicit version / feature flag and support
> both for some period of time.
>
> I'm inclined to say no, and just let things slide, and instead make
> sure that e2fsck can handle both the old and the new format, and let
> that handle the fast commit replay if necessary.
>
> Harshad, what do you think?
I agree. Making kernel replay backward compatible would complicate the
replay code without adding that much value (since there aren't that
many users and fast commit isn't enabled by default). So, having the
ability in e2fsck to do replays should suffice in this case.

- Harshad
>
>                                                 - Ted