net/packet: fix overflow in tpacket_rcv
authorOr Cohen <orcohen@paloaltonetworks.com>
Fri, 4 Sep 2020 04:05:28 +0000 (21:05 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 14 Oct 2020 07:48:13 +0000 (09:48 +0200)
commit acf69c946233259ab4d64f8869d4037a198c7f06 upstream.

Using tp_reserve to calculate netoff can overflow as
tp_reserve is unsigned int and netoff is unsigned short.

This may lead to macoff receving a smaller value then
sizeof(struct virtio_net_hdr), and if po->has_vnet_hdr
is set, an out-of-bounds write will occur when
calling virtio_net_hdr_from_skb.

The bug is fixed by converting netoff to unsigned int
and checking if it exceeds USHRT_MAX.

This addresses CVE-2020-14386

Fixes: 8913336a7e8d ("packet: add PACKET_RESERVE sockopt")
Signed-off-by: Or Cohen <orcohen@paloaltonetworks.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
[ snu: backported to pre-5.3, changed tp_drops counting/locking ]
Signed-off-by: Stefan Nuernberger <snu@amazon.com>
CC: David Woodhouse <dwmw@amazon.co.uk>
CC: Amit Shah <aams@amazon.com>
CC: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/packet/af_packet.c

index fb64394..b5b79f5 100644 (file)
@@ -2161,7 +2161,8 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev,
        int skb_len = skb->len;
        unsigned int snaplen, res;
        unsigned long status = TP_STATUS_USER;
-       unsigned short macoff, netoff, hdrlen;
+       unsigned short macoff, hdrlen;
+       unsigned int netoff;
        struct sk_buff *copy_skb = NULL;
        struct timespec ts;
        __u32 ts_status;
@@ -2223,6 +2224,12 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev,
                }
                macoff = netoff - maclen;
        }
+       if (netoff > USHRT_MAX) {
+               spin_lock(&sk->sk_receive_queue.lock);
+               po->stats.stats1.tp_drops++;
+               spin_unlock(&sk->sk_receive_queue.lock);
+               goto drop_n_restore;
+       }
        if (po->tp_version <= TPACKET_V2) {
                if (macoff + snaplen > po->rx_ring.frame_size) {
                        if (po->copy_thresh &&