linux-kernel - Re: [mm 4.15-rc7] Random oopses under memory pressure.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CA+55aFwvgm+KKkRLaFsuAjTdfQooS=UaMScC0CbZQ9WnX_AF=g@mail.gmail.com>
Date:   Mon, 15 Jan 2018 15:05:20 -0800
From:   Linus Torvalds <torvalds@...ux-foundation.org>
To:     Tetsuo Handa <penguin-kernel@...ove.sakura.ne.jp>
Cc:     Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        linux-mm <linux-mm@...ck.org>,
        "the arch/x86 maintainers" <x86@...nel.org>,
        linux-fsdevel <linux-fsdevel@...r.kernel.org>,
        Michal Hocko <mhocko@...nel.org>
Subject: Re: [mm 4.15-rc7] Random oopses under memory pressure.

On Sun, Jan 14, 2018 at 3:54 AM, Tetsuo Handa
<penguin-kernel@...ove.sakura.ne.jp> wrote:
> This memory corruption bug occurs even on CONFIG_SMP=n CONFIG_PREEMPT_NONE=y
> kernel. This bug highly depends on timing and thus too difficult to bisect.
> This bug seems to exist at least since Linux 4.8 (judging from the traces, though
> the cause might be different). None of debugging configuration gives me a clue.
> So far only CONFIG_HIGHMEM=y CONFIG_DEBUG_PAGEALLOC=y kernel (with RAM enough to
> use HighMem: zone) seems to hit this bug, but it might be just by chance caused
> by timings. Thus, there is no evidence that 64bit kernels are not affected by
> this bug. But I can't narrow down any more. Thus, I call for developers who can
> narrow down / identify where the memory corruption bug is.

Hmm.

I guess I'm still hung up on the "it does not look like a valid
'struct page *'" thing.

Can you reproduce this with CONFIG_FLATMEM=y instead of CONFIG_SPARSEMEM?

Because if you can, I think we can easily add a few more pfn and
'struct page' validation debug statements. With SPARSEMEM, it gets
pretty complicated because the whole struct page setup is much more
complex.

              Linus