Re: [PATCH v2] hpfs: obsolete check=none mount option

[Tetsuo Handa]

Hello, Mikulas.

Since we haven't heard back from Linus yet, I would like to move this discussion
forward with you as the maintainer to finally close these syzbot reports.

Regarding your suggestion to update the fuzzers to avoid the "check=none" option,
I am concerned that this won't be a fundamental solution. Fuzzers are incredibly
creative at bypassing restrictions. Even if we configure syzkaller not to use
"check=none" during the initial mount, it will eventually find a way to trigger
that code path - for example, by executing "mount -o remount,check=none" during
its execution.

As long as the kernel accepts "check=none" and its associated unsafe behavior,
syzbot (and other fuzzers that will appear in future) will continue to report
these crashes as valid bugs, placing a recurring burden on you to review them.

To stop this "cat-and-mouse" game and save your time in the long run, I believe
the most robust approach is to obsolete the "check=none" behavior within the
kernel itself, as proposed in this patch.

Could you please consider picking up this change or re-sending the pull request
to Linus? Closing this gap in the kernel is the only way to ensure we don't have
to deal with the same "check=none" reports repeatedly.

Best regards.