linux-kernel - Re: [PATCH] arch_topology: Support SMT control on arm64

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: <2247ced8-c4f9-80a7-a0a2-0953abed135f@arm.com>
Date:   Fri, 22 Sep 2023 13:13:35 +0200
From:   Dietmar Eggemann <dietmar.eggemann@....com>
To:     Yicong Yang <yangyicong@...wei.com>, catalin.marinas@....com,
        will@...nel.org, sudeep.holla@....com,
        linux-arm-kernel@...ts.infradead.org
Cc:     yangyicong@...ilicon.com, gregkh@...uxfoundation.org,
        rafael@...nel.org, jonathan.cameron@...wei.com,
        prime.zeng@...ilicon.com, linuxarm@...wei.com,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] arch_topology: Support SMT control on arm64

On 21/09/2023 10:56, Yicong Yang wrote:
> On 2023/9/21 1:08, Dietmar Eggemann wrote:
>> On 19/09/2023 14:33, Yicong Yang wrote:
>>> From: Yicong Yang <yangyicong@...ilicon.com>

[...]

> If you manually disable SMT by offline each CPUs the cpu_smt_control will
> not be updated. It'll updated when using the interface like
> `/sys/devices/system/cpu/smt/control` or cmdline. By these means,
> the framework will use topology_is_primary_thread() to decide which CPU
> in the SMT will keep online:
> 
> // e.g. echo off > /sys/devices/system/cpu/smt/control
> [ kernel/cpu.c ]
> control_store()
>   __store_smt_control()
>     cpuhp_smt_disable()
>       for_each_online_cpu(cpu)
>         if (topology_is_primary_thread(cpu))
>           continue;            <---------- will skip the primary thread
>       [...]
>       cpu_smt_control = CPU_SMT_DISABLED;
> 
> topology_is_primary_thread() checking only applies to the SMT control but
> not to the CPU offline.

I see, make sense. Retested on my SMT4 Arm64 server with 256 CPUs.