linux-kernel - Re: [PATCH] mm/rmap: Convert from atomic_t to refcount_t on anon

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <202108200017.9F1744F76@keescook>
Date:   Fri, 20 Aug 2021 00:33:59 -0700
From:   Kees Cook <keescook@...omium.org>
To:     Peter Zijlstra <peterz@...radead.org>
Cc:     Linus Torvalds <torvalds@...ux-foundation.org>,
        Will Deacon <will@...nel.org>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Borislav Petkov <bp@...en8.de>,
        Xiyu Yang <xiyuyang19@...an.edu.cn>,
        Alistair Popple <apopple@...dia.com>,
        Yang Shi <shy828301@...il.com>,
        Shakeel Butt <shakeelb@...gle.com>,
        Hugh Dickins <hughd@...gle.com>,
        Miaohe Lin <linmiaohe@...wei.com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Linux-MM <linux-mm@...ck.org>, yuanxzhang@...an.edu.cn,
        Xin Tan <tanxin.ctf@...il.com>,
        Will Deacon <will.deacon@....com>,
        David Howells <dhowells@...hat.com>
Subject: Re: [PATCH] mm/rmap: Convert from atomic_t to refcount_t on
 anon_vma->refcount

On Fri, Aug 20, 2021 at 08:43:40AM +0200, Peter Zijlstra wrote:
> On Thu, Aug 19, 2021 at 12:09:37PM -0700, Linus Torvalds wrote:
> > On Thu, Aug 19, 2021 at 8:21 AM Peter Zijlstra <peterz@...radead.org> wrote:
> > >
> > > If we can skip the OF... we can do something like this:
> > 
> > Honestly, I think a lot of the refcount code is questionable. It was
> > absolutely written with no care for performance AT ALL.
> 
> That's a bit unfair I feel. Will's last rewrite of the stuff was
> specifically to address performance issues.

Well, to address performance issues with the "full" version. The default
x86-specific code was already as fast as atomic_t. Will got it to nearly
match while making it catch all conditions, not just the exploitable
ones. (i.e. it didn't bother trying to catch underflow; there's no way
to mitigate it).

Will's version gave us three properties: correctness (it catches all the
pathological conditions), speed (it was very nearly the same speed as
regular atomic_t), and arch-agnosticism, which expanded this protection
to things beyond just x86 and arm64.

> > But see above: maybe just make this a separate "careful atomic_t",
> > with the option to panic-on-overflow. So then we could get rid of
> > refcount_warn_saturate() enmtirely above, and instead just have a
> > (compile-time option) BUG() case, with the non-careful version just
> > being our existing atomic_dec_and_test.

This is nearly what we had before. But refcount_t should always saturate
on overflow -- that's specifically the mitigation needed to defang the
traditional atomic_t overflow exploits (of which we had several a year
before refcount_t and now we've seen zero since).

> We used to have that option; the argument was made that everybody cares
> about security and as long as this doesn't show up on benchmarks we
> good.
> 
> Also, I don't think most people want the overflow to go BUG, WARN is
> mostly the right thing and only the super paranoid use panic-on-warn or
> something.

Saturating on overflow stops exploitability. WARNing is informational.
BUG kills the system for no good reason: the saturation is the defense
against attack, and the WARN is the "oh, I found a bug" details needed
to fix it.

I prefer the arch-agnostic, fully checked, very fast version of this
(i.e. what we have right now). :P I appreciate it's larger, but in my
opinion size isn't as important as correctness and speed. If it's just
as fast as a small version but has greater coverage, that seems worth
the size.

-- 
Kees Cook