linux-kernel - Re: [Patch v2 2/5] KVM: x86/mmu: Optimize SPTE change flow for clear-dirty-log

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: <CALzav=eEs_cy6v1r6UD0k_2MOp9RSZ33LgPy2bRrFcZ-g_MBYQ@mail.gmail.com>
Date:   Mon, 6 Feb 2023 15:34:00 -0800
From:   David Matlack <dmatlack@...gle.com>
To:     Ben Gardon <bgardon@...gle.com>
Cc:     Vipin Sharma <vipinsh@...gle.com>, seanjc@...gle.com,
        pbonzini@...hat.com, kvm@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [Patch v2 2/5] KVM: x86/mmu: Optimize SPTE change flow for clear-dirty-log

On Mon, Feb 6, 2023 at 2:06 PM Ben Gardon <bgardon@...gle.com> wrote:
>
> On Fri, Feb 3, 2023 at 11:28 AM Vipin Sharma <vipinsh@...gle.com> wrote:
> >
> > +static inline u64 kvm_tdp_mmu_clear_spte_bit(struct tdp_iter *iter, u64 mask)
> > +{
> > +       atomic64_t *sptep;
> > +
> > +       if (kvm_tdp_mmu_spte_has_volatile_bits(iter->old_spte, iter->level)) {
> > +               sptep = (atomic64_t *)rcu_dereference(iter->sptep);
> > +               return (u64)atomic64_fetch_and(~mask, sptep);
> > +       }
> > +
> > +       __kvm_tdp_mmu_write_spte(iter->sptep, iter->old_spte & ~mask);
> > +       return iter->old_spte;
> > +}
> > +
>
> If you do end up sending another version of the series and feel like
> breaking up this patch, you could probably split this part out since
> the change to how we set the SPTE and how we handle that change are
> somewhat independent. I like the switch to atomic64_fetch_and though.
> No idea if it's faster, but I would believe it could be.

The changes are actually dependent. Using the atomic-AND makes it
impossible for KVM to clear a volatile bit that it wasn't intending to
clear, and that is what enables simplifying the SPTE change
propagation.

Instead I would split out the opportunistic cleanup of
@record_dirty_log to a separate patch, since that's dead-code cleanup
and refactoring.