linux-kernel - Re: [PATCH] bluetooth: 6lowpan: fix use after free in chan

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: <1490952907.4892.12.camel@linux.intel.com>
Date:   Fri, 31 Mar 2017 12:35:07 +0300
From:   Jukka Rissanen <jukka.rissanen@...ux.intel.com>
To:     Michael Scott <michael.scott@...aro.org>,
        Marcel Holtmann <marcel@...tmann.org>,
        Gustavo Padovan <gustavo@...ovan.org>,
        Johan Hedberg <johan.hedberg@...il.com>,
        Luiz Augusto von Dentz <luiz.dentz@...il.com>
Cc:     "David S . Miller" <davem@...emloft.net>,
        linux-bluetooth@...r.kernel.org, linux-wpan@...r.kernel.org,
        netdev@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] bluetooth: 6lowpan: fix use after free in
 chan_suspend/resume

Hi Michael,

On Tue, 2017-03-28 at 23:10 -0700, Michael Scott wrote:
> A status field in the skb_cb struct was storing a channel status
> based on channel suspend/resume events.  This stored status was
> then used to return EAGAIN if there were packet sending issues
> in snd_pkt().
> 
> The issue is that the skb has been freed by the time the callback
> to 6lowpan's suspend/resume was called.  So, this generates a
> "use after free" issue that was noticed while running kernel tests
> with KASAN debug enabled.
> 
> Let's eliminate the status field entirely as we can use the channel
> tx_credits to indicate whether we should return EAGAIN when handling
> packets.
> 
> Signed-off-by: Michael Scott <michael.scott@...aro.org>
> ---
>  net/bluetooth/6lowpan.c | 21 +++------------------
>  1 file changed, 3 insertions(+), 18 deletions(-)
> 
> diff --git a/net/bluetooth/6lowpan.c b/net/bluetooth/6lowpan.c
> index d491529332f4..e27be3ca0a0c 100644
> --- a/net/bluetooth/6lowpan.c
> +++ b/net/bluetooth/6lowpan.c
> @@ -38,7 +38,6 @@ struct skb_cb {
>  	struct in6_addr addr;
>  	struct in6_addr gw;
>  	struct l2cap_chan *chan;
> -	int status;
>  };
>  #define lowpan_cb(skb) ((struct skb_cb *)((skb)->cb))
>  
> @@ -528,7 +527,7 @@ static int send_pkt(struct l2cap_chan *chan,
> struct sk_buff *skb,
>  	}
>  
>  	if (!err)
> -		err = lowpan_cb(skb)->status;
> +		err = (!chan->tx_credits ? -EAGAIN : 0);
>  
>  	if (err < 0) {
>  		if (err == -EAGAIN)
> @@ -964,26 +963,12 @@ static struct sk_buff *chan_alloc_skb_cb(struct
> l2cap_chan *chan,
>  
>  static void chan_suspend_cb(struct l2cap_chan *chan)
>  {
> -	struct sk_buff *skb = chan->data;
> -
> -	BT_DBG("chan %p conn %p skb %p", chan, chan->conn, skb);
> -
> -	if (!skb)
> -		return;
> -
> -	lowpan_cb(skb)->status = -EAGAIN;
> +	BT_DBG("chan %p suspend", chan);
>  }
>  
>  static void chan_resume_cb(struct l2cap_chan *chan)
>  {
> -	struct sk_buff *skb = chan->data;
> -
> -	BT_DBG("chan %p conn %p skb %p", chan, chan->conn, skb);
> -
> -	if (!skb)
> -		return;
> -
> -	lowpan_cb(skb)->status = 0;
> +	BT_DBG("chan %p resume", chan);
>  }
>  
>  static long chan_get_sndtimeo_cb(struct l2cap_chan *chan)

Good catch! If we can avoid using the status variable, that is very
good. We could probably also remove the resume and suspend callbacks as
they are now empty functions (unless we need the debug info for
something).

Acked-by: Jukka Rissanen <jukka.rissanen@...ux.intel.com>


Cheers,
Jukka