linux-kernel - Re: [PATCH RFC 3/7] x86: kernel: cpu: bugs.c: modify static_cpu_has to boot_cpu

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:   Mon, 06 Jul 2020 16:46:33 +0200
From:   Thomas Gleixner <tglx@...utronix.de>
To:     Mihai Carabas <mihai.carabas@...cle.com>,
        linux-kernel@...r.kernel.org
Cc:     mingo@...hat.com, bp@...en8.de, x86@...nel.org,
        boris.ostrovsky@...cle.com, konrad.wilk@...cle.com,
        Mihai Carabas <mihai.carabas@...cle.com>
Subject: Re: [PATCH RFC 3/7] x86: kernel: cpu: bugs.c: modify static_cpu_has to boot_cpu_has

Mihai Carabas <mihai.carabas@...cle.com> writes:

> The usage of static_cpu_has in bugs.c file is counter-productive
> since the code is executed once but there is extra effort to patch
> it and keep alternatives in a special section --- so there is both
> space and time cost.
>
> Quote from _static_cpu_has definition:
> /*
>  * Static testing of CPU features. Used the same as boot_cpu_has(). It
>  * statically patches the target code for additional performance. Use
>  * static_cpu_has() only in fast paths, where every cycle counts. Which
>  * means that the boot_cpu_has() variant is already fast enough for the
>  * majority of cases and you should stick to using it as it is generally
>  * only two instructions: a RIP-relative MOV and a TEST.
>  */
>
> There are two other places where static_cpu_has is used and might be
> considered critical paths: __speculation_ctrl_update() and vmx_l1d_flush().
>
> Given these facts, changing static_cpu_has to boot_cpu_has is done in
> order to bypass alternative instructions which cannot be updated at runtime
> for now.

Not going to happen. We are not adding 4 conditionals to context switch
just to support that late loading horrors. There are better ways to do
that.

Thanks,

        tglx