Re: [PATCH v3] Add kernel config option for tweaking kernel behavior.

[Linus Torvalds]

On Tue, Apr 14, 2020 at 5:06 PM Tetsuo Handa
<penguin-kernel@xxxxxxxxxxxxxxxxxxx> wrote:
>
> I can't judge whether this WARN_ON() makes sense outside of syzkaller.
> But I'd like to suppress this WARN_ON() under testing by syzkaller so that
> syzkaller can spend resource for seeking for different bugs (assuming that
> suppressing this WARN_ON() is harmless).

That's crazy talk.

Either that WARN_ON() is valid, or it's not. It doesn't matter if it's
syzcaller, and we would never change the behavior of it depending on a
flag - compile-time or dynamic.

There's no way we'd accept that as some "#ifndef CONFIG_FUZZER" thing either.

If that WARN_ON() is a problem, then the people behind it should be
appraised of it, and it should probably be removed. I'm assuming it
was some kind of "I don't think this can happen, so if it does, I want
to see how it happened" WARN_ON.

                Linus