Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 23.128.96.38 as permitted sender) client-ip=23.128.96.38;
Date:   Sat, 21 Oct 2023 16:36:19 -0700
From:   Kees Cook <kees@kernel.org>
To:     andy.shevchenko@gmail.com, Jan Kara <jack@suse.cz>
CC:     Andy Shevchenko <andriy.shevchenko@intel.com>,
        Baokun Li <libaokun1@huawei.com>,
        Josh Poimboeuf <jpoimboe@kernel.org>,
        Nathan Chancellor <nathan@kernel.org>,
        Nick Desaulniers <ndesaulniers@google.com>,
        Kees Cook <keescook@chromium.org>,
        Ferry Toth <ftoth@exalondelft.nl>,
        linux-fsdevel@vger.kernel.org, linux-ext4@vger.kernel.org
Subject: Re: [GIT PULL] ext2, quota, and udf fixes for 6.6-rc1
User-Agent: K-9 Mail for Android
In-Reply-To: <ZTLk1G0KCF7YNjRx@surfacebook.localdomain>
References: <ZTFh0NeYtvgcjSv8@smile.fi.intel.com> <CAHk-=wjXG52UNKCwwEU1A+QWHYfvKOieV0uFOpPkLR0NSvOjtg@mail.gmail.com> <CAHk-=whis2BJF2fv1xySAg2NTQ+C5fViNSGkLNCOqGzi-3y+8w@mail.gmail.com> <ZTFxEcjo4d6vXbo5@smile.fi.intel.com> <ZTFydEbdEYlxOxc1@smile.fi.intel.com> <CAHk-=wh_gbZE_ZsQ6+9gSPdXfoCtmuK-MFmBkO3ywMKFQEvb6g@mail.gmail.com> <ZTKUDzONVHXnWAJc@smile.fi.intel.com> <CAHk-=wipA4605yvnmjW7T9EvARPRCGLARty8UUzRGxic1SXqvg@mail.gmail.com> <ZTLHBYv6wSUVD/DW@smile.fi.intel.com> <CAHk-=wgHFSTuANT3jXsw1EtzdHQe-XQtWQACzeFxn2BEBzX-gA@mail.gmail.com> <ZTLk1G0KCF7YNjRx@surfacebook.localdomain>
Message-ID: <BF6761C0-B813-4C98-9563-8323C208F67D@kernel.org>
MIME-Version: 1.0
Content-Type: text/plain;
 charset=utf-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk


On October 20, 2023 1:36:36 PM PDT, andy=2Eshevchenko@gmail=2Ecom wrote:
>That said, if you or anyone has ideas how to debug futher, I'm all ears!

I don't think this has been tried yet:

When I've had these kind of hard-to-find glitches I've used manual built-b=
inary bisection=2E Assuming you have a source tree that works when built wi=
th Clang and not with GCC:
- build the tree with Clang with, say, O=3Dbuild-clang
- build the tree with GCC, O=3Dbuild-gcc
- make a new tree for testing: cp -a build-clang build-test
- pick a suspect =2Eo file (or files) to copy from build-gcc into build-te=
st
- perform a relink: "make O=3Dbuild-test" should DTRT since the copied-in =
=2Eo files should be newer than the =2Ea and other targets
- test for failure, repeat

Once you've isolated it to (hopefully) a single =2Eo file, then comes the =
byte-by-byte analysis or something similar=2E=2E=2E

I hope that helps! These kinds of bugs are super frustrating=2E

-Kees

--=20
Kees Cook