linux-kernel - Re: Intel Alder Lake Question (Scheduler)

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] [day] [month] [year] [list]

Message-ID: <CALGQo4aMFOrRW4EkDyjwNxGO6jayQsxzmwh3PSmjZv+Vi3jvzQ@mail.gmail.com>
Date:   Fri, 5 Nov 2021 23:26:39 +0000
From:   Keith Cancel <admin@...th.pro>
To:     linux-kernel@...r.kernel.org
Cc:     gayatri.kammela@...el.com
Subject: Re: Intel Alder Lake Question (Scheduler)

Hello Gayatri,

So I was searching through the mailing list looking at various
patches. I don't see any for alder lake regarding what I was
mentioning. Is there something I am overlooking? I did see you were on
multiple email chains for adding alder so maybe you have a better
idea.

Thanks,
Keith Cancel



On Fri, Nov 5, 2021 at 7:51 AM Keith Cancel <admin@...th.pro> wrote:
>
> Hello,
>
> So Alder lake was supposed to have AVX-512 disabled/fused off. So it
> looks like that was not the case. For example see this news article:
> https://www.anandtech.com/show/17047/the-intel-12th-gen-core-i912900k-review-hybrid-performance-brings-hybrid-complexity/2
>
> This issue I noticed it seems to be the possibility that a bios could
> allow someone to enable both AVX-512 and and E-Cores (efficiency
> cores). How on earth would the scheduler handle this? I could imagine
> the scheduler switching a process/thread using AVX-512 to an
> efficiency core. Now suddenly the core the thread/process is running
> no longer has AVX-512 instructions... This sounds like a headache.
>
> I suppose the kernel could handle the trap and then from on only
> schedule such a process on performance cores. This issue I see here
> though is userland process could then intentionally trigger such a
> trap to then only be scheduled on a performance core. Even if it does
> not use AVX-512 otherwise.
>
> Thanks,
> Keith Cancel