linux-kernel - Re: BUG 5.5.6, from patch "cpu/SMT: Fix x86 link error without CONFIG

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

Message-ID: <875zdmttsr.fsf@nanos.tec.linutronix.de>
Date:   Sun, 26 Apr 2020 12:39:32 +0200
From:   Thomas Gleixner <tglx@...utronix.de>
To:     Holger Schurig <holger.schurig@...antech.de>
Cc:     "linux-kernel\@vger.kernel.org" <linux-kernel@...r.kernel.org>,
        x86@...nel.org, Arnd Bergmann <arnd@...db.de>
Subject: Re: BUG 5.5.6, from patch "cpu/SMT: Fix x86 link error without CONFIG_SYSFS"

Holger,

Holger Schurig <holger.schurig@...antech.de> writes:

> if one compiled kernel 5.6.6 with CONFIG_HOTPLUG_CPU=n, then the
> compilation will fail. The reason is your patch, which introduces
> this:

> +int cpuhp_smt_disable(enum cpuhp_smt_control ctrlval)

It would have been more helpful to provide the actual commit SHA instead
of copying tons of patch code.

> It happens that cpu_down_maps_locked is only defined when
> CONFIG_HOTPLUG_CPU=y, but not otherwise.

Right.

> On my embedded target I so far had the CPU hotplugging off. After all,
> the CPU is physically soldered, there will never be a hotplugging
> happen.

CPU hotplug is used for far more than physical hotplug.

SMT control depends on it in order to apply speculation mitigations,
hibernation requires working CPU hotplug as well.

As the speculation mess requires SMT control we more or less have to
make CPU hotplug mandatory on x86...

Thanks,

        tglx