linux-kernel - Re: [PATCH v3 5/9] x86/mm/tlb: Privatize 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]

Message-ID: <CALCETrXsyktwp+Wt0=LsZvv-S2UkS-pk5j2N4ud4iYAU3VFwyA@mail.gmail.com>
Date:   Sat, 20 Jul 2019 06:58:36 -0700
From:   Andy Lutomirski <luto@...nel.org>
To:     Nadav Amit <namit@...are.com>
Cc:     Dave Hansen <dave.hansen@...el.com>,
        Andy Lutomirski <luto@...nel.org>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        "the arch/x86 maintainers" <x86@...nel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>
Subject: Re: [PATCH v3 5/9] x86/mm/tlb: Privatize cpu_tlbstate

On Fri, Jul 19, 2019 at 11:54 AM Nadav Amit <namit@...are.com> wrote:
>
> > On Jul 19, 2019, at 11:48 AM, Dave Hansen <dave.hansen@...el.com> wrote:
> >
> > On 7/19/19 11:43 AM, Nadav Amit wrote:
> >> Andy said that for the lazy tlb optimizations there might soon be more
> >> shared state. If you prefer, I can move is_lazy outside of tlb_state, and
> >> not set it in any alternative struct.
> >
> > I just wanted to make sure that we capture these rules:
> >
> > 1. If the data is only ever accessed on the "owning" CPU via
> >   this_cpu_*(), put it in 'tlb_state'.
> > 2. If the data is read by other CPUs, put it in 'tlb_state_shared'.
> >
> > I actually like the idea of having two structs.
>
> Yes, that’s exactly the idea. In the (1) case, we may even be able to mark
> the struct with __thread qualifier, which IIRC would prevent memory barriers
> from causing these values being reread.

I'm okay with the patch.  If we end up changing things later, we can
rearrange as needed.