Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20200907134055.2878499-1-elver@google.com> <20200907134055.2878499-2-elver@google.com>
 <CACT4Y+bfp2ch2KbSMkUd3142aA4p2CiMOmdXrr0-muu6bQ5xXg@mail.gmail.com> <CAG_fn=W4es7jaTotDORt2SwspE4A804mdwAY1j4gcaSEKtRjiw@mail.gmail.com>
In-Reply-To: <CAG_fn=W4es7jaTotDORt2SwspE4A804mdwAY1j4gcaSEKtRjiw@mail.gmail.com>
From:   Dmitry Vyukov <dvyukov@google.com>
Date:   Thu, 10 Sep 2020 19:11:41 +0200
Message-ID: <CACT4Y+awrz-j8y5Qc8OS9qkov4doMnw1V=obwp3MB_LTvaUFXw@mail.gmail.com>
Subject: Re: [PATCH RFC 01/10] mm: add Kernel Electric-Fence infrastructure
To:     Alexander Potapenko <glider@google.com>
Cc:     Marco Elver <elver@google.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Catalin Marinas <catalin.marinas@arm.com>,
        Christoph Lameter <cl@linux.com>,
        David Rientjes <rientjes@google.com>,
        Joonsoo Kim <iamjoonsoo.kim@lge.com>,
        Mark Rutland <mark.rutland@arm.com>,
        Pekka Enberg <penberg@kernel.org>,
        "H. Peter Anvin" <hpa@zytor.com>,
        "Paul E. McKenney" <paulmck@kernel.org>,
        Andrey Konovalov <andreyknvl@google.com>,
        Andrey Ryabinin <aryabinin@virtuozzo.com>,
        Andy Lutomirski <luto@kernel.org>,
        Borislav Petkov <bp@alien8.de>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        Eric Dumazet <edumazet@google.com>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Ingo Molnar <mingo@redhat.com>, Jann Horn <jannh@google.com>,
        Jonathan Corbet <corbet@lwn.net>,
        Kees Cook <keescook@chromium.org>,
        Peter Zijlstra <peterz@infradead.org>, Qian Cai <cai@lca.pw>,
        Thomas Gleixner <tglx@linutronix.de>,
        Will Deacon <will@kernel.org>,
        "the arch/x86 maintainers" <x86@kernel.org>,
        "open list:DOCUMENTATION" <linux-doc@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>,
        kasan-dev <kasan-dev@googlegroups.com>,
        Linux ARM <linux-arm-kernel@lists.infradead.org>,
        Linux-MM <linux-mm@kvack.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, Sep 10, 2020 at 6:19 PM Alexander Potapenko <glider@google.com> wrote:
>
> On Thu, Sep 10, 2020 at 5:43 PM Dmitry Vyukov <dvyukov@google.com> wrote:
>
>
> > > +       /* Calculate address for this allocation. */
> > > +       if (right)
> > > +               meta->addr += PAGE_SIZE - size;
> > > +       meta->addr = ALIGN_DOWN(meta->addr, cache->align);
> >
> > I would move this ALIGN_DOWN under the (right) if.
> > Do I understand it correctly that it will work, but we expect it to do
> > nothing for !right? If cache align is >PAGE_SIZE, nothing good will
> > happen anyway, right?
> > The previous 2 lines look like part of the same calculation -- "figure
> > out the addr for the right case".
>
> Yes, makes sense.
>
> > > +
> > > +       schedule_delayed_work(&kfence_timer, 0);
> > > +       WRITE_ONCE(kfence_enabled, true);
> >
> > Can toggle_allocation_gate run before we set kfence_enabled? If yes,
> > it can break. If not, it's still somewhat confusing.
>
> Correct, it should go after we enable KFENCE. We'll fix that in v2.
>
> > > +void __kfence_free(void *addr)
> > > +{
> > > +       struct kfence_metadata *meta = addr_to_metadata((unsigned long)addr);
> > > +
> > > +       if (unlikely(meta->cache->flags & SLAB_TYPESAFE_BY_RCU))
> >
> > This may deserve a comment as to why we apply rcu on object level
> > whereas SLAB_TYPESAFE_BY_RCU means slab level only.
>
> Sorry, what do you mean by "slab level"?
> SLAB_TYPESAFE_BY_RCU means we have to wait for possible RCU accesses
> in flight before freeing objects from that slab - that's basically
> what we are doing here below:

Exactly! You see it is confusing :)
SLAB_TYPESAFE_BY_RCU does not mean that. rcu-freeing only applies to
whole pages, that's what I mean by "slab level" (whole slabs are freed
by rcu).

> > > +               call_rcu(&meta->rcu_head, rcu_guarded_free);
> > > +       else
> > > +               kfence_guarded_free(addr, meta);
> > > +}