User space is currently sending a OVS_FLOW_ATTR_PROBE for both flow
and packet messages. This leads to an out-of-bounds access in
ovs_packet_cmd_execute() because OVS_FLOW_ATTR_PROBE >
OVS_PACKET_ATTR_MAX.
Introduce a new OVS_PACKET_ATTR_PROBE with the same numeric value
as OVS_FLOW_ATTR_PROBE to grow the range of accepted packet attributes
while maintaining to be binary compatible with existing OVS binaries.
Fixes: 05da589 ("openvswitch: Add support for OVS_FLOW_ATTR_PROBE.")
Reported-by: Sander Eikelenboom <linux@eikelenboom.it>
Tracked-down-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Reviewed-by: Jesse Gross <jesse@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
        OVS_PACKET_ATTR_USERDATA,    /* OVS_ACTION_ATTR_USERSPACE arg. */
        OVS_PACKET_ATTR_EGRESS_TUN_KEY,  /* Nested OVS_TUNNEL_KEY_ATTR_*
                                            attributes. */
+       OVS_PACKET_ATTR_UNUSED1,
+       OVS_PACKET_ATTR_UNUSED2,
+       OVS_PACKET_ATTR_PROBE,      /* Packet operation is a feature probe,
+                                      error logging should be suppressed. */
        __OVS_PACKET_ATTR_MAX
 };
 
 
        struct vport *input_vport;
        int len;
        int err;
-       bool log = !a[OVS_FLOW_ATTR_PROBE];
+       bool log = !a[OVS_PACKET_ATTR_PROBE];
 
        err = -EINVAL;
        if (!a[OVS_PACKET_ATTR_PACKET] || !a[OVS_PACKET_ATTR_KEY] ||
        [OVS_PACKET_ATTR_PACKET] = { .len = ETH_HLEN },
        [OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED },
        [OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED },
+       [OVS_PACKET_ATTR_PROBE] = { .type = NLA_FLAG },
 };
 
 static const struct genl_ops dp_packet_genl_ops[] = {