Re: [mm 4.15-rc8] Random oopses under memory pressure.

[Linus Torvalds]

On Tue, Jan 16, 2018 at 9:33 AM, Tetsuo Handa
<penguin-kernel@xxxxxxxxxxxxxxxxxxx> wrote:
>
> Since I got a faster reproducer, I tried full bisection between 4.11 and 4.12-rc1.
> But I have no idea why bisection arrives at c0332694903a37cf.

I don't think your reproducer is 100% reliable.

And bisection is great because it's very aggressive and optimal when
it comes to testing. But that also implies that if *any* of the
good/bad choices were incorrect, then the end result is pure garbage
and isn't even *close* to the right commit.

> It turned out that CONFIG_FLATMEM was irrelevant. I just did not hit it.

So have you actually been able to see the problem with FLATMEM, or is
this based on the bisect? Because I really think the bisect is pretty
much guaranteed to be wrong.

               Linus