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;
Date:   Tue, 2 May 2023 14:49:34 -0300
From:   Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
To:     Kuniyuki Iwashima <kuniyu@amazon.com>
Cc:     ilia.gavrilov@infotecs.ru, davem@davemloft.net,
        edumazet@google.com, kuba@kernel.org, linux-kernel@vger.kernel.org,
        linux-sctp@vger.kernel.org, lucien.xin@gmail.com,
        lvc-project@linuxtesting.org, netdev@vger.kernel.org,
        nhorman@tuxdriver.com, pabeni@redhat.com, simon.horman@corigine.com
Subject: Re: [PATCH net v2] sctp: fix a potential buffer overflow in
 sctp_sched_set_sched()
Message-ID: <ZFFNLtBYepvBzoPr@t14s.localdomain>
References: <20230502130316.2680585-1-Ilia.Gavrilov@infotecs.ru>
 <20230502170516.39760-1-kuniyu@amazon.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230502170516.39760-1-kuniyu@amazon.com>
Precedence: bulk

On Tue, May 02, 2023 at 10:05:16AM -0700, Kuniyuki Iwashima wrote:
> From:   Gavrilov Ilia <Ilia.Gavrilov@infotecs.ru>
> Date:   Tue, 2 May 2023 13:03:24 +0000
> > The 'sched' index value must be checked before accessing an element
> > of the 'sctp_sched_ops' array. Otherwise, it can lead to buffer overflow.
> 
> OOB access ?

My thought as well.

> But it's not true because it does not happen in the first place.
> 
> > 
> > Note that it's harmless since the 'sched' parameter is checked before
> > calling 'sctp_sched_set_sched'.
> > 
> > Found by InfoTeCS on behalf of Linux Verification Center
> > (linuxtesting.org) with SVACE.
> > 
> > Fixes: 5bbbbe32a431 ("sctp: introduce stream scheduler foundations")
> > Reviewed-by: Simon Horman <simon.horman@corigine.com>
> > Signed-off-by: Ilia.Gavrilov <Ilia.Gavrilov@infotecs.ru>
> > ---
> > V2:
> >  - Change the order of local variables 
> >  - Specify the target tree in the subject
> >  net/sctp/stream_sched.c | 9 +++++----
> >  1 file changed, 5 insertions(+), 4 deletions(-)
> > 
> > diff --git a/net/sctp/stream_sched.c b/net/sctp/stream_sched.c
> > index 330067002deb..4d076a9b8592 100644
> > --- a/net/sctp/stream_sched.c
> > +++ b/net/sctp/stream_sched.c
> > @@ -146,18 +146,19 @@ static void sctp_sched_free_sched(struct sctp_stream *stream)
> >  int sctp_sched_set_sched(struct sctp_association *asoc,
> >  			 enum sctp_sched_type sched)
> >  {
> > -	struct sctp_sched_ops *n = sctp_sched_ops[sched];
> >  	struct sctp_sched_ops *old = asoc->outqueue.sched;
> >  	struct sctp_datamsg *msg = NULL;
> > +	struct sctp_sched_ops *n;
> >  	struct sctp_chunk *ch;
> >  	int i, ret = 0;
> >  
> > -	if (old == n)
> > -		return ret;
> > -
> >  	if (sched > SCTP_SS_MAX)
> >  		return -EINVAL;
> 
> I'd just remove this check instead because the same test is done
> in the caller side, sctp_setsockopt_scheduler(), and this errno
> is never returned.
> 
> This unnecessary test confuses a reader like sched could be over
> SCTP_SS_MAX here.

It's actualy better to keep the test here and remove it from the
callers: they don't need to know the specifics, and further new calls
will be protected already.

> 
> Since the OOB access does not happen, I think this patch should
> go to net-next without the Fixes tag after the merge window.

Yup.

> 
> Thanks,
> Kuniyuki
> 
> 
> >  
> > +	n = sctp_sched_ops[sched];
> > +	if (old == n)
> > +		return ret;
> > +
> >  	if (old)
> >  		sctp_sched_free_sched(&asoc->stream);
> >  
> > -- 
> > 2.30.2