| PFSYNC(4) | Kernel Interfaces Manual | PFSYNC(4) |
By default, all local changes to the state table are exposed via pfsync. However, state changes from packets received by pfsync over the network are not rebroadcast. States created by a rule marked with the no-sync keyword are omitted from the pfsync interface (see pf.conf(5) for details).
The pfsync interface will attempt to collapse multiple updates of the same state into one message where possible. The maximum number of times this can be done before the update is sent out is controlled by the maxupd parameter to ifconfig (see ifconfig(8) and the example below for more details).
Each packet retrieved on this interface has a header associated with it of length PFSYNC_HDRLEN. The header indicates the version of the protocol, address family, action taken on the following states, and the number of state table entries attached in this packet. This structure is defined in <net/if_pfsync.h> as:
struct pfsync_header {
u_int8_t version;
u_int8_t af;
u_int8_t action;
u_int8_t count;
};
# ifconfig pfsync0 syncdev fxp0
By default, state change messages are sent out on the synchronisation interface using IP multicast packets. The protocol is IP protocol 240, PFSYNC, and the multicast group used is 224.0.0.240. When a peer address is specified using the syncpeer keyword, the peer address is used as a destination for the pfsync traffic.
It is important that the pfsync traffic be well secured as there is no authentication on the protocol and it would be trivial to spoof packets which create states, bypassing the pf ruleset. Either run the pfsync protocol on a trusted network - ideally a network dedicated to pfsync messages such as a crossover cable between two firewalls, or specify a peer address and protect the traffic with ipsec(4) (it is not supported at the moment on NetBSD due to the lack of any encapsulation pseudo-device).
There is a one-to-one correspondence between packets seen by bpf(4) on the pfsync interface, and packets sent out on the synchronisation interface, i.e. a packet with 4 state deletion messages on pfsync means that the same 4 deletions were sent out on the synchronisation interface. However, the actual packet contents may differ as the messages sent over the network are "compressed" where possible, containing only the necessary information.
Both firewalls in this example have three sis(4) interfaces. sis0 is the external interface, on the 10.0.0.0/24 subnet; sis1 is the internal interface, on the 192.168.0.0/24 subnet; and sis2 is the pfsync interface, using the 192.168.254.0/24 subnet. A crossover cable connects the two firewalls via their sis2 interfaces. On all three interfaces, firewall A uses the .254 address, while firewall B uses .253. The interfaces are configured as follows (firewall A unless otherwise indicated):
/etc/ifconfig.sis0:
inet 10.0.0.254 255.255.255.0 NONE
/etc/ifconfig.sis1:
inet 192.168.0.254 255.255.255.0 NONE
/etc/ifconfig.sis2:
inet 192.168.254.254 255.255.255.0 NONE
/etc/ifconfig.carp0:
inet 10.0.0.1 255.255.255.0 10.0.0.255 vhid 1 pass foo
/etc/ifconfig.carp1:
inet 192.168.0.1 255.255.255.0 192.168.0.255 vhid 2 pass bar
/etc/ifconfig.pfsync0:
up syncdev sis2
pf(4) must also be configured to allow pfsync and carp(4) traffic through. The following should be added to the top of /etc/pf.conf:
pass quick on { sis2 } proto pfsync
pass on { sis0 sis1 } proto carp
If it is preferable that one firewall handle the traffic, the advskew on the backup firewall's carp(4) interfaces should be set to something higher than the primary's. For example, if firewall B is the backup, its /etc/ifconfig.carp1 would look like this:
inet 192.168.0.1 255.255.255.0 192.168.0.255 vhid 2 pass bar \ advskew 100
The following must also be added to /etc/sysctl.conf:
net.inet.carp.preempt=1
| April 12, 2010 | NetBSD 6.1 |