SA-14:19 (Denial of Service in TCP packet processing) and jails issue ?

[Mike Tancsa]

While running a Qualys scan against one of our RELENG_9 boxes 
(9.3-STABLE #16 r281494), it came up complaining about the SA below as 
an issue still ?

https://www.freebsd.org/security/advisories/FreeBSD-SA-14:19.tcp.asc

The server should have this fix in place no ?

# svn info /usr/src
Path: /usr/src
Working Copy Root Path: /usr/src
URL: svn://svn.freebsd.org/base/stable/9
Relative URL: ^/stable/9
Repository Root: svn://svn.freebsd.org/base
Repository UUID: ccf9f872-aa2e-dd11-9fc8-001c23d0bc1f
Revision: 281494
Node Kind: directory
Schedule: normal
Last Changed Author: jkim
Last Changed Rev: 281389
Last Changed Date: 2015-04-10 21:17:19 -0400 (Fri, 10 Apr 2015)

Also, we have pf in place with the first rule being

#block everything by default coming in
block in log all

and then explicitly letting in the ports we want, with keep state

The IP being scanned is in a jail.  If I run the scan to an IP not 
associated with the jail, the scan does not complain. Its only on the 
jailed IP that the scan flags as problematic for this vulnerability.


Looking at the src of tcp_input.c, I see

      todrop = tp->rcv_nxt - th->th_seq;
         if (todrop > 0) {
                 if (thflags & TH_SYN) {
                         thflags &= ~TH_SYN;
                         th->th_seq++;
                         if (th->th_urp > 1)
                                 th->th_urp--;
                         else
                                 thflags &= ~TH_URG;
                         todrop--;
                 }

so the code path looks right based on the commit in 
https://lists.freebsd.org/pipermail/svn-src-stable-9/2014-September/007198.html

If this is a false positive, how can I be sure thats the case ? I have 
pcaps of the scan both against the jailed IP (with the scan saying its 
vulnerable) and against an IP not associated with the jail, saying its 
not an issue.




QID:
82054
Category:
TCP/IP
CVE ID:
CVE-2004-0230
Vendor Reference
-
Bugtraq ID:
10183
Service Modified:
02/02/2010
User Modified:
-
Edited:
No
PCI Vuln:
No
Ticket State:

TCP provides stateful communications between hosts on a network. TCP 
sessions are established by a three-way handshake and use random 32-bit 
sequence and acknowledgement numbers to ensure the validity of traffic. 
A vulnerability was reported that may permit TCP sequence numbers to be 
more easily approximated by remote attackers. This issue affects 
products released by multiple vendors.
The cause of the vulnerability is that affected implementations will 
accept TCP sequence numbers within a certain range, known as the 
acknowledgement range, of the expected sequence number for a packet in 
the session. This is determined by the TCP window size, which is 
negotiated during the three-way handshake for the session. Larger TCP 
window sizes may be set to allow for more throughput, but the larger the 
TCP window size, the more probable it is to guess a TCP sequence number 
that falls within an acceptable range. It was initially thought that 
guessing an acceptable sequence number was relatively difficult for most 
implementations given random distribution, making this type of attack 
impractical. However, some implementations may make it easier to 
successfully approximate an acceptable TCP sequence number, making these 
attacks possible with a number of protocols and implementations.
This is further compounded by the fact that some implementations may 
support the use of the TCP Window Scale Option, as described in RFC 
1323, to extend the TCP window size to a maximum value of 1 billion.
This vulnerability will permit a remote attacker to inject a SYN or RST 
packet into the session, causing it to be reset and effectively allowing 
for denial of service attacks. An attacker would exploit this issue by 
sending a packet to a receiving implementation with an approximated 
sequence number and a forged source IP address and TCP port.
There are a few factors that may present viable target implementations, 
such as those which depend on long-lived TCP connections, those that 
have known or easily guessed IP address endpoints and those 
implementations with easily guessed TCP source ports. It has been noted 
that Border Gateway Protocol (BGP) is reported to be particularly 
vulnerable to this type of attack, due to the use of long-lived TCP 
sessions and the possibility that some implementations may use the TCP 
Window Scale Option. As a result, this issue is likely to affect a 
number of routing platforms.
Another factor to consider is the relative difficulty of injecting 
packets into TCP sessions, as a number of receiving implementations will 
reassemble packets in order, dropping any duplicates. This may make some 
implementations more resistant to attacks than others.
It should be noted that while a number of vendors have confirmed this 
issue in various products, investigations are ongoing and it is likely 
that many other vendors and products will turn out to be vulnerable as 
the issue is investigated further.


-- 
-------------------
Mike Tancsa, tel +1 519 651 3400
Sentex Communications, mike at sentex.net
Providing Internet services since 1994 www.sentex.net
Cambridge, Ontario Canada   http://www.tancsa.com/