netdev - Re: [PATCH net v2] net: zero-initialize tc skb extension on allocation

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: <CAM_iQpUtGw5MO0DAWkVuHP7PU-iSkmEsBWa+SkCqiZtB3eeSoQ@mail.gmail.com>
Date:   Tue, 25 May 2021 14:48:55 -0700
From:   Cong Wang <xiyou.wangcong@...il.com>
To:     Vlad Buslov <vladbu@...dia.com>
Cc:     David Miller <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>,
        Jamal Hadi Salim <jhs@...atatu.com>,
        Jiri Pirko <jiri@...nulli.us>,
        Saeed Mahameed <saeedm@...lanox.com>,
        Florian Westphal <fw@...len.de>, wenxu <wenxu@...oud.cn>,
        Linux Kernel Network Developers <netdev@...r.kernel.org>
Subject: Re: [PATCH net v2] net: zero-initialize tc skb extension on allocation

On Tue, May 25, 2021 at 6:22 AM Vlad Buslov <vladbu@...dia.com> wrote:
>
> Function skb_ext_add() doesn't initialize created skb extension with any
> value and leaves it up to the user. However, since extension of type
> TC_SKB_EXT originally contained only single value tc_skb_ext->chain its
> users used to just assign the chain value without setting whole extension
> memory to zero first. This assumption changed when TC_SKB_EXT extension was
> extended with additional fields but not all users were updated to
> initialize the new fields which leads to use of uninitialized memory
> afterwards. UBSAN log:

Hm, I thought the memset() in __skb_ext_alloc() does the job, clearly
I was wrong.

[...]
>
> Fix the issue by providing new function tc_skb_ext_alloc() that allocates
> tc skb extension and initializes its memory to 0 before returning it to the
> caller. Change all existing users to use new API instead of calling
> skb_ext_add() directly.

Just a note: struct tc_skb_ext is currently only 8-byte long, so memset()
it should not be a problem for performance.

>
> Fixes: 038ebb1a713d ("net/sched: act_ct: fix miss set mru for ovs after defrag in act_ct")
> Fixes: d29334c15d33 ("net/sched: act_api: fix miss set post_ct for ovs after do conntrack in act_ct")
> Signed-off-by: Vlad Buslov <vladbu@...dia.com>

Acked-by: Cong Wang <cong.wang@...edance.com>

Thanks.