Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220901030610.1121299-1-keescook@chromium.org>
 <20220901030610.1121299-2-keescook@chromium.org> <20220831201825.378d748d@kernel.org>
 <202208312324.F2F8B28CA@keescook> <20220901124915.24ebc067@kernel.org>
In-Reply-To: <20220901124915.24ebc067@kernel.org>
From:   Eric Dumazet <edumazet@google.com>
Date:   Thu, 1 Sep 2022 13:54:22 -0700
Message-ID: <CANn89iLPLu=cP0zu5r+hsXudSjMzieWke=nh0YwccH1QJUnzog@mail.gmail.com>
Subject: Re: [PATCH 1/2] netlink: Bounds-check nlmsg_len()
To:     Jakub Kicinski <kuba@kernel.org>
Cc:     Kees Cook <keescook@chromium.org>,
        "David S. Miller" <davem@davemloft.net>,
        Paolo Abeni <pabeni@redhat.com>,
        Pablo Neira Ayuso <pablo@netfilter.org>,
        Jozsef Kadlecsik <kadlec@netfilter.org>,
        Florian Westphal <fw@strlen.de>,
        syzbot <syzkaller@googlegroups.com>,
        Yajun Deng <yajun.deng@linux.dev>,
        netdev <netdev@vger.kernel.org>, netfilter-devel@vger.kernel.org,
        coreteam@netfilter.org, Oliver Hartkopp <socketcan@hartkopp.net>,
        Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>,
        LKML <linux-kernel@vger.kernel.org>,
        linux-hardening@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, Sep 1, 2022 at 12:49 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Wed, 31 Aug 2022 23:27:08 -0700 Kees Cook wrote:
> > This would catch corrupted values...
> >
> > Is the concern the growth in image size? The check_sub_overflow() isn't
> > large at all -- it's just adding a single overflow bit test. The WARNs
> > are heavier, but they're all out-of-line.
>
> It turns the most obvious function into a noodle bar :(
>
> Looking at this function in particular is quite useful, because
> it clearly indicates that the nlmsg_len includes the header.
>
> How about we throw in a
>
>         WARN_ON_ONCE(nlh->nlmsg_len < NLMSG_HDRLEN ||
>                      nlh->nlmsg_len > INT_MAX);
>
> but leave the actual calculation human readable C?

This is inlined, and will add a lot of extra code. We are making the
kernel slower at each release.

What about letting fuzzers like syzbot find the potential issues ?

DEBUG_NET_WARN_ON_ONCE(...);