linux-kernel - Re: [PATCH] RAS/CEC: Add debugfs switch to disable at run time

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: <CAM_iQpVpnBoGbjUP4Ek0TDEN8aimq3ty=0EUmco9ywWDZa37Vw@mail.gmail.com>
Date:   Sat, 20 Apr 2019 12:08:08 -0700
From:   Cong Wang <xiyou.wangcong@...il.com>
To:     Borislav Petkov <bp@...en8.de>
Cc:     Tony Luck <tony.luck@...el.com>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] RAS/CEC: Add debugfs switch to disable at run time

On Sat, Apr 20, 2019 at 11:47 AM Borislav Petkov <bp@...en8.de> wrote:
> IOW, when you have the CEC enabled, you don't need to log memory errors
> with a userspace agent. The CEC collects them and discards them if they
> don't repeat.

So, you mean breaking mcelog is intentionally, if so, why not break it
loudly?

That is, for example, preventing mcelog from starting by disabling
CONFIG_X86_MCELOG_LEGACY in Kconfig _automatically_ when
CONFIG_RAS is enabled? (Like what I showed in my PoC change.)

Or, for another example, print a kernel warning and let users know this
behavior is intentional?

>
> If they do repeat, then it offlines the page.
>
> Without user intervention and interference.
>
> Now, if you still want to know how many errors and where they happened
> and when they happened and yadda yadda, you *disable* the CEC.

Well, I believe rasdaemon has the counters too, it is not hard to count
the trace events at all. I don't worry about this at all. What I worry is how
we treat mcelog when having CONFIG_RAS=y.

>
> I hope this makes more sense now.

Yes, thanks for the information. It is kinda what I expected, as I keep saying,
I believe we can improve this situation to avoid users' confusion, rather than
just saying CONFIG_RAS=n is the answer.

Thanks.