Openswan upstream is going to release security advisory for Openswan and Strongswan describing following flaw:
A vulnerability in the Dead Peer Detection (RFC-3706) code was found by
Gerd v. Egidy of Intra2net AG affecting all Openswan and all Strongswan
A malicious (or expired ISAKMP) R_U_THERE or R_U_THERE_ACK Dead Peer
Detection packet can cause the pluto IKE daemon to crash and restart. No
authentication or encryption is required to trigger this bug. One spoofed
UDP packet can cause the pluto IKE daemon to restart and be unresponsive
for a few seconds while restarting.
Openswan-2.6.20 and earlier
Strongswan-4.2.13 and earlier
Created attachment 336484 [details]
The version of openswan as shipped in Red Hat Enterprise Linux 5 differs from current upstream versions in the default value of plutorestartoncrash which controls automatic restart of pluto daemon after the crash. It seems that this was intended to default to yes, but mistakenly defaulted to no in certain versions. Upstream changelog mentions fix for this introduced in 2.6.15:
[ ... ]
* Change (back) defaults of plutorestartoncrash and uniqueids from
no to yes. The new parser mistakenly did not set these [paul]
Defaults can be checked using the following command:
ipsec addconn --configsetup
To enable automatic pluto restart on crashes, plutorestartoncrash=yes needs to be added to "config setup" section of ipsec.conf.
Public now via:
This issue has been addressed in following products:
Red Hat Enterprise Linux 5
Via RHSA-2009:0402 https://rhn.redhat.com/errata/RHSA-2009-0402.html
Current Fedora is shipping with 2.6.21 or later, so this does not affect Fedora.