linux-kernel - Re: [PATCH v4 1/2] x86/mce: Check for writes ignored in MCA

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]

Date:   Thu, 5 May 2022 12:03:36 -0700
From:   Smita Koralahalli <Smita.KoralahalliChannabasappa@....com>
To:     Borislav Petkov <bp@...en8.de>
Cc:     x86@...nel.org, linux-edac@...r.kernel.org,
        linux-kernel@...r.kernel.org, Tony Luck <tony.luck@...el.com>,
        "H . Peter Anvin" <hpa@...or.com>,
        Yazen Ghannam <yazen.ghannam@....com>
Subject: Re: [PATCH v4 1/2] x86/mce: Check for writes ignored in MCA_STATUS
 register

On 5/3/2022 2:14 PM, Borislav Petkov wrote:
> On Mon, May 02, 2022 at 08:28:47PM -0700, Smita Koralahalli wrote:
>> I'm bit more inclined towards your previous approach of
>> hw_injection_possible
>> check in do_inject(). This seems better than doing it in flags_write().
> If you don't do it in flags_write() then the user would do
>
>     echo "hw" > flags
>
> the command will succeed and the user will think that hw injection is
> possible and then wonder why it fails later.
That's right!
>
> I even actually think that in the first run, when hw_injection_possible
> is not determined yet, you should try to poke at MCi_STATUS of some
> non-reserved bank - and we enumerate which those are at boot in
> __mcheck_cpu_check_banks(), so you can pick a random, non-RAZ bank, save
> its MCi_STATUS, try to write it and if it succeeds, restore it.
>
> This way you'll determine whether hw injection is possible, store it
> in the static hw_injection_possible and then query only that variable.
> I.e., you'll have to poke that MCi_STATUS only once on driver init.

Okay I agree too. I will work on this.

Thanks,
Smita
>   And
> this way it'll be the most optimal, methinks.
>
> Thx.
>