SMCRoute is a static multicast routing daemon providing fine grained control over the multicast forwarding cache (MFC) in the UNIX kernel. Both IPv4 and IPv6 are fully supported.
SMCRoute can be used as an alternative to dynamic multicast routers like mrouted, pimd, or pim6sd in setups where static multicast routes should be maintained and/or no proper IGMP or MLD signaling exists.
Multicast routes exist in the UNIX kernel as long as a multicast routing daemon runs. On Linux, multiple multicast routers can run simultaneously using different multicast routing tables.
The full documentation of SMCRoute is available in the manual pages, see smcrouted(8), smcroutectl(8), and smcroute.conf(5).
All features, except mrdisc, are supported for both IPv4 and IPv6. Please note, some features may not be available on systems other than Linux. E.g., FreeBSD does not have SSM group join support.
- Configuration file support,
/etc/smcroute.conf
- Configuration snippet include support,
/etc/smcroute.d/*.conf
- Daemon startup options support,
/etc/default/smcroute
- Support for seamless reloading of the configuration on
SIGHUP
- Source-less on-demand routing, a.k.a. wildcard
(*,G)
based static routing, including support for(*,G/LEN)
and(S/LEN,G/LEN)
- Optional built-in mrdisc support for IPv4, RFC4286
- Support for multiple routing tables on Linux
- Client to add/remove routes, join/leave groups, and built-in support to show both routes and joined groups
- Interface wildcard matching,
eth+
matcheseth0, eth15
Note:
smcroutectl
can be used to freely modify the runtime state ofsmcrouted
, but any changes made (routes/groups) are lost when the configuration is reloaded. This is by design.
smcrouted [-nNhsv] [-c SEC] [-d SEC] [-e CMD] [-f FILE] [-i NAME]
[-l LVL] [-p USER:GROUP] [-P FILE] [-t ID] [-u FILE]
smcroutectl [-dptv] [-i NAME] [-u FILE] [COMMAND]
smcroutectl ⟨kill | reload⟩
smcroutectl ⟨add | rem⟩ ⟨ROUTE⟩
smcroutectl ⟨join | leave⟩ ⟨GROUP⟩
smcroutectl show [ routes | groups]
To set multicast routes and join groups you must first start the daemon,
which needs root privileges, or CAP_NET_ADMIN
. Use smcrouted -n
to run the daemon in the foreground, as required by modern init daemons
like systemd and Finit.
When started from systemd, smcrouted
runs with the -n -s
options,
i.e. supervised in the foreground and uses syslog for logging output.
The default log level is INFO
, this can be adjusted using the file
/etc/default/smcroute
:
SMCROUTED_OPTS=-l debug
When configured with --sysconfdir=/etc
, like most Linux distributions
do, smcrouted
reads /etc/smcroute.conf
, which can look something
like this:
mgroup from eth0 group 225.1.2.3
mgroup from eth0 group 225.1.2.3 source 192.168.1.42
mroute from eth0 group 225.1.2.3 source 192.168.1.42 to eth1 eth2
The first line means "Join multicast group 225.1.2.3 on interface eth0".
Useful if eth0
is not directly connected to the source, but to a LAN
with switches with IGMP snooping. Joining the group opens up multicast
for that group towards eth0
. See below Caveat for limitations.
The second mgroup
is for source specific group join, i.e. the host
specifies that it wants packets from 192.168.1.42 and no other source.
The third mroute
line is the actual layer-3 routing entry. Here we
say that multicast data originating from 192.168.1.42 on eth0
to the
multicast group 225.1.2.3 should be forwarded to interfaces eth1
and
eth2
.
Note: To test the above you can use ping from another device. The multicast should be visible as long as your IP# matches the source above and you ping 225.1.2.3 -- REMEMBER TO SET THE TTL >1
ping -I eth0 -t 2 225.1.2.3
The TTL is what usually bites people first trying out multicast routing.
Most TCP/IP stacks default to a TTL of 1 for multicast frames, e.g. ping
above requires -t 2
, or greater. This limitation is intentional and
reduces the risk of someone accidentally flooding multicast. Remember,
multicast behaves like broadcast unless limited.
The TTL should preferably be set on the sender side, e.g. the camera,
but can also be modified in the firewall on a router. Be careful though
because the TTL is the only thing that helps prevent routing loops! On
Linux the following iptables
command can be used to change the TTL:
iptables -t mangle -A PREROUTING -i eth0 -d 225.1.2.3 -j TTL --ttl-inc 1
Some commands, like this one, must usually be run with root privileges or the correct set of capabilities.
On some platforms there is a limit of 20 groups per socket. This stems
from a limit in BSD UNIX, which also affects Linux. The setting that
controls this is IP_MAX_MEMBERSHIPTS
, defined in the system header
file netinet/in.h
. Linux users can tweak this with the following
/proc
setting:
echo 40 > /proc/sys/net/ipv4/igmp_max_memberships
smcrouted
probes this at runtime by attempting to join as many groups
as possible (as have been requested), when the kernel accepts no further
joins on a socket, smcrouted
opens a new one.
For large setups it is recommended to investigate enabling multicast router ports in the switches, either statically or by enabling support for multicast router discovery, RFC 4286, or possibly use a dynamic multicast routing protocol.
smcrouted -e /path/to/script
With -e CMD
a user script or command can be called when smcrouted
receives a SIGHUP
or installs a multicast route to the kernel. This
is useful if you, for instance, also run a NAT firewall and need to
flush connection tracking after installing a multicast route.
smcrouted -N
With the -N
command line option SMCRoute does not prepare all system
interfaces for multicast routing. Very useful if your system has a lot
of interfaces but only a select few are required for multicast routing.
Use the following in /etc/smcroute.conf
to enable interfaces:
phyint eth0 enable
phyint eth1 enable
phyint eth2 enable
It is possible to use any interface that supports the MULTICAST
flag.
Note, however, that depending on the UNIX kernel in use, you may have to
have an interface address set, in the relevant address family, and the
interface may likely also have to be UP
.
On Linux it is possible to run multiple multicast routing daemons due to its support for multiple multicast routing tables. In such setups it may be useful to change the default identity of SMCRoute:
smcrouted -i mrt1 -t 1
smcrouted -i mrt2 -t 2
The -i NAME
option alters the default syslog name, config file, PID
file, and client socket file name used. In the first instance above,
smcrouted
will use:
/etc/mrt1.conf
/var/run/mrt1.pid
/var/run/mrt1.sock
and syslog messages will use the mrt1
identity as well. Remember to
use the same -i NAME
also to smcroutectl
.
SMCRoute also has a client interface to interact with the daemon:
smcroutectl join eth0 225.1.2.3
smcroutectl add eth0 192.168.1.42 225.1.2.3 eth1 eth2
If the daemon runs with a different identity the client needs to be called using the same identity:
smcrouted -i mrt
smcroutectl -i mrt show
There are more commands. See the man page or the online help for details:
smcroutectl help
Note: Root privileges are required by default for
smcroutectl
due to the IPC socket permissions.
Multicast often originates from different sources but usually not at the
same time. For a more generic setup, and to reduce the number of rules
required, it is possible to set (*,G)
multicast routes for both IPv4
and IPv6. Variants include (*,G/LEN)
and (S/LEN,G/LEN
. These
wildcard routes are used as "templates" to match against and install
proper (S,G)
routes when the kernel informs smcrouted
of inbound
multicast from new sources.
Example smcroute.conf
:
phyint eth0 enable mrdisc
phyint eth1 enable
phyint eth2 enable
mgroup from eth0 group 225.1.2.3
mroute from eth0 group 225.1.2.3 to eth1 eth2
or, from the command line:
# smcroutectl join eth0 225.1.2.3
# smcroutectl add eth0 225.1.2.3 eth1 eth2
Also, see the smcrouted -c SEC
option for periodic flushing of learned
(*,G)
rules, including the automatic blocking of unknown multicast, and
the smcroutectl flush
command.
Another interesting feature is multicast router discovery, mrdisc,
described in RFC4286. This feature is disabled by default, enable
with configure --enable-mrdisc
. When enabled it periodically sends
out an IGMP message on inbound interfaces¹ to alert switches to open up
multicast in that direction. Not many managed switches have support for
this yet.
Note: mrdisc only works on Linux due to
SO_BINDTODEVICE
.
¹ Notice the mrdisc
flag to the above phyint eth0
directive, which
is missing for eth1
and eth2
.
SMCRoute should in theory work on any UNIX like operating system which supports the BSD MROUTING API. Both Linux and FreeBSD are tested on a regular basis.
On Linux the following kernel config is required:
CONFIG_IP_MROUTE=y
CONFIG_IP_PIMSM_V1=y
CONFIG_IP_PIMSM_V2=y
CONFIG_IP_MROUTE_MULTIPLE_TABLES=y # For multiple routing tables
CONFIG_IPV6_MROUTE_MULTIPLE_TABLES=y # For multiple routing tables
On *BSD the following kernel config is required:
options MROUTING # Multicast routing
options PIM # pimd extensions used for (*,G) support
FreeBSD support module loading, kldload(8)
, edit /boot/loader.conf
:
ip_mroute_load="yes"
ip_mroute6_load="yes"
Check the list of multicast capable interfaces:
cat /proc/net/dev_mcast
or look for interfaces with the MULTICAST
flag in the output from:
ifconfig
Some interfaces have the MULTICAST
flag disabled by default, like lo
and greN
. Usually this flag can be enabled administratively.
The GNU Configure & Build system use /usr/local
as the default install
prefix. In many cases this is useful, but this means the configuration
files, cache, and PID files will also use that prefix. Most users have
come to expect those files in /etc/
and /var/
and configure has a
few useful options that are recommended to use. For SMCRoute you may
want to use something like this:
./configure --prefix=/usr --sysconfdir=/etc --runstatedir=/var/run
make -j5
sudo make install-strip
Usually your system reserves /usr
for native pacakges, so most users
drop --prefix
, installing to /usr/local
, or use --prefix=/opt
.
Note: On some systems --runstatedir
may not be available in the
configure script, try --localstatedir=/var
instead.
As of SMCRoute v2.2 support for privilege separation using the libcap
library was added. It is used to drop full root privileges at startup,
retaining only CAP_NET_ADMIN
for managing the multicast routes.
The build system searches for the libcap
library and header file(s).
Both libcap-dev
and pkg-config
are required.
Note: Although support is automatically detected, the build system
will issue a warning if libcap
is missing. This can be
silenced with configure --without-libcap
For systemd integration libsystemd-dev
and pkg-config
are required.
When the unit file is installed, systemctl
can be used to enable and
start smcrouted
:
$ sudo systemctl enable smcroute.service
$ sudo systemctl start smcroute.service
Check that it started properly by inspecting the system log, or:
$ sudo systemctl status smcroute.service
Some people want to build statically, to do this with autoconf
add the
following LDFLAGS=
after the configure script. You may also need to
add LIBS=...
, which will depend on your particular system:
./configure LDFLAGS="-static" ...
The configure
script and the Makefile.in
files are generated and not
stored in GIT. So if you checkout the sources from GitHub you first
need to generated these files using ./autogen.sh
.
SMCRoute is maintained collaboratively at GitHub. Bug reports, feature requests, patches/pull requests, and documentation fixes are most welcome. The project was previously hosted and maintained by Debian at Alioth and before that by Carsten Schill, the original author.