Received-SPF: pass (google.com: domain of linux-kernel+bounces-108965-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240320125635.1444-1-abelova@astralinux.ru> <Zfrmv4u0tVcYGS5n@nanopsycho>
In-Reply-To: <Zfrmv4u0tVcYGS5n@nanopsycho>
From: Eric Dumazet <edumazet@google.com>
Date: Wed, 20 Mar 2024 14:43:22 +0100
Message-ID: <CANn89iLz4ZesK23DQJmMdn5EA2akh_z+8ZLU-oEuRKy3JDEbAw@mail.gmail.com>
Subject: Re: [PATCH] flow_dissector: prevent NULL pointer dereference in __skb_flow_dissect
To: Jiri Pirko <jiri@resnulli.us>
Cc: Anastasia Belova <abelova@astralinux.ru>, "David S. Miller" <davem@davemloft.net>, 
	Jakub Kicinski <kuba@kernel.org>, netdev@vger.kernel.org, linux-kernel@vger.kernel.org, 
	lvc-project@linuxtesting.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Wed, Mar 20, 2024 at 2:38=E2=80=AFPM Jiri Pirko <jiri@resnulli.us> wrote=
:
>
> Wed, Mar 20, 2024 at 01:56:35PM CET, abelova@astralinux.ru wrote:
> >skb is an optional parameter, so it may be NULL.
> >Add check defore dereference in eth_hdr.
> >
> >Found by Linux Verification Center (linuxtesting.org) with SVACE.
>
> Either drop this line which provides no value, or attach a link to the
> actual report.
>
>
> >
> >Fixes: 67a900cc0436 ("flow_dissector: introduce support for Ethernet add=
resses")
>
> This looks incorrect. I believe that this is the offending commit:
> commit 690e36e726d00d2528bc569809048adf61550d80
> Author: David S. Miller <davem@davemloft.net>
> Date:   Sat Aug 23 12:13:41 2014 -0700
>
>     net: Allow raw buffers to be passed into the flow dissector.
>
>
>
> >Signed-off-by: Anastasia Belova <abelova@astralinux.ru>
> >---
> > net/core/flow_dissector.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> >diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
> >index 272f09251343..05db3a8aa771 100644
> >--- a/net/core/flow_dissector.c
> >+++ b/net/core/flow_dissector.c
> >@@ -1137,7 +1137,7 @@ bool __skb_flow_dissect(const struct net *net,
> >               rcu_read_unlock();
> >       }
> >
> >-      if (dissector_uses_key(flow_dissector,
> >+      if (skb && dissector_uses_key(flow_dissector,
> >                              FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
> >               struct ethhdr *eth =3D eth_hdr(skb);
> >               struct flow_dissector_key_eth_addrs *key_eth_addrs;
>
> Looks like FLOW_DISSECT_RET_OUT_BAD should be returned in case the
> FLOW_DISSECTOR_KEY_ETH_ADDRS are selected and there is no skb, no?
>

It would be nice knowing in which context we could have a NULL skb and
FLOW_DISSECTOR_KEY_ETH_ADDRS
at the same time.

It seems this fix is based on some kind of static analysis, but no real bug=
.