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;
Subject: Re: [PATCH] netlink: add buffer boundary checking
To:     Mark Salyzyn <salyzyn@android.com>, linux-kernel@vger.kernel.org
Cc:     kernel-team@android.com, netdev@vger.kernel.org,
        "David S. Miller" <davem@davemloft.net>,
        Jakub Kicinski <kuba@kernel.org>, Thomas Graf <tgraf@suug.ch>
References: <20200723182136.2550163-1-salyzyn@android.com>
From:   Eric Dumazet <eric.dumazet@gmail.com>
Message-ID: <09cd1829-8e41-bef5-ba5e-1c446c166778@gmail.com>
Date:   Thu, 23 Jul 2020 12:35:24 -0700
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.9.0
MIME-Version: 1.0
In-Reply-To: <20200723182136.2550163-1-salyzyn@android.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On 7/23/20 11:21 AM, Mark Salyzyn wrote:
> Many of the nla_get_* inlines fail to check attribute's length before
> copying the content resulting in possible out-of-boundary accesses.
> Adjust the inlines to perform nla_len checking, for the most part
> using the nla_memcpy function to faciliate since these are not
> necessarily performance critical and do not need a likely fast path.
> 
> Signed-off-by: Mark Salyzyn <salyzyn@android.com>
> Cc: netdev@vger.kernel.org
> Cc: linux-kernel@vger.kernel.org
> Cc: kernel-team@android.com
> Cc: "David S. Miller" <davem@davemloft.net>
> Cc: Jakub Kicinski <kuba@kernel.org>
> Cc: Thomas Graf <tgraf@suug.ch>
> Fixes: bfa83a9e03cf ("[NETLINK]: Type-safe netlink messages/attributes interface")
> ---
>  include/net/netlink.h | 66 +++++++++++++++++++++++++++++++++++--------
>  1 file changed, 54 insertions(+), 12 deletions(-)
> 
> diff --git a/include/net/netlink.h b/include/net/netlink.h
> index c0411f14fb53..11c0f153be7c 100644
> --- a/include/net/netlink.h
> +++ b/include/net/netlink.h
> @@ -1538,7 +1538,11 @@ static inline int nla_put_bitfield32(struct sk_buff *skb, int attrtype,
>   */
>  static inline u32 nla_get_u32(const struct nlattr *nla)
>  {
> -	return *(u32 *) nla_data(nla);
> +	u32 tmp;
> +
> +	nla_memcpy(&tmp, nla, sizeof(tmp));
> +
> +	return tmp;

I believe this will hide bugs, that syzbot was able to catch.

Instead, you could perhaps introduce a CONFIG_DEBUG_NETLINK option,
and add a WARN_ON_ONCE(nla_len(nla) < sizeof(u32)) so that we can detect bugs in callers.