open80211s HOWTO

This page explains how to set up a mesh network using open80211s and a supported driver.

Note: at this time, this HOWTO assumes that you are not running a WiFi manager such as NetworkManager. Please kill or otherwise stop NetworkManager or similar software before setting up your mesh interfaces or tell it to ignore the device that you plan to use for mesh.

Note: open80211s is disabled in the 2.6.31 kernel, however it is enabled in subsequent kernels. Please use 2.6.32 or above or consider using the wireless-testing kernel or the compat-wireless package.

Getting all the pieces

Each mesh node needs to be configured as follows:

The kernel

Get the wireless-testing kernel: 

$ git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-testing.git

Configure the kernel including the mac80211 stack with mesh capabilities (CONFIG_MAC80211, CONFIG_MAC80211_MESH) and enable the driver(s) that you need, for example the zd1211-rw driver (CONFIG_ZD1211RW). Compile and install on your target. Here are instructions for Ubuntu.

You also have the option of using compat-wireless instead of building the kernel or you may also use the 2.6.26 or later mainline kernel. Using wireless-testing is strongly recommended as it represents the latest development and driver support.

iw

Follow the instructions at iw documentation page to build the latest version of iw from source or use a recent released version of iw.

Testing

  • Boot several nodes with the compiled kernel and with iw properly installed.
  • Choose a mesh id. The mesh id is a string that will be used in the mesh node beacons. Only mesh points with the same mesh id will be able to communicate with each other. The mesh id is a maximum of 32 bytes long. In this document, we refer to the mesh id as $MESH_ID.
  • Add a new mesh interface for the wlanN device on each node. You can choose the name of the mesh interface. "mesh" is a popular choice. In this document, we refer to the mesh interface as $MESH_IFACE. 
# iw dev wlan0 interface add $MESH_IFACE type mp mesh_id $MESH_ID

A new interface will be created: 

# ifconfig -a | grep mesh
mesh      Link encap:Ethernet  HWaddr 00:18:39:14:48:B5

Note: depending on your Linux distribution, you may see a slightly different name for the mesh interface.

Bring up your $MESH_IFACE interface: 

# ifconfig $MESH_IFACE up
  • All mesh interfaces start by default in channel 1. The channel can be changed with 
# iw dev $MESH_IFACE set channel <channel>

or 

# iwconfig $MESH_IFACE channel <channel>
  • Assign a same-subnet IP to the mesh interface in each node. 
# ifconfig $MESH_IFACE <ip_addr>

For this tutorial, let's assume that our mesh nodes are assigned addresses in the 192.168.3.80/28 subnet. 

# ifconfig $MESH_IFACE 192.168.3.80
  • In the default configuration, mesh points will automatically attempt to create peer links will all other mesh nodes with the same mesh ID. You can examine which peer links have been established with the iw dev $MESH_IFACE station dump command: 
# iw dev $MESH_IFACE station dump
Station 00:15:6d:84:14:10 (on mesh)
	inactive time:	1320 ms
	rx bytes:	352
	rx packets:	4
	tx bytes:	174
	tx packets:	2
	signal:  	-61 dBm
	tx bitrate:	1.0 MBit/s
	mesh llid:	32577
	mesh plid:	15969
	mesh plink:	ESTAB
Station 00:15:6d:84:14:09 (on mesh)
	inactive time:	3370 ms
	rx bytes:	1064
	rx packets:	12
	tx bytes:	545
	tx packets:	7
	signal:  	-53 dBm
	tx bitrate:	1.0 MBit/s
	mesh llid:	41036
	mesh plid:	24435
	mesh plink:	ESTAB

Each Station stanza in the station dump output represents a mesh node with whom we have a peer link. The first line shows the MAC address of the mesh node. The mesh plink field reports the state of the peer link with this mesh node. When this value is ESTAB, direct communication with this mesh node is possible.

  • You should now be able to ping any node in the mesh. After doing that you can use arp and iw to inspect the arp table and path table. This will reveal where the packet was actually sent in order to get to its destination. 
192-168-3-80:~# ping -c 2 192.168.3.83
PING 192.168.3.83 (192.168.3.83) 56(84) bytes of data.
64 bytes from 192.168.3.83: icmp_seq=1 ttl=64 time=8.85 ms
64 bytes from 192.168.3.83: icmp_seq=2 ttl=64 time=3.42 ms

192-168-3-80:~# arp
? (192.168.3.83) at 00:15:6d:84:14:0e [ether]  on mesh

192-168-3-80:~# iw dev $MESH_IFACE mpath dump
00:15:6d:84:14:10 00:15:6d:84:14:10 mesh	0	228 	0	2196731904	0	0	0x10
00:15:6d:84:14:0e 00:15:6d:84:14:10 mesh	2	456 	0	2196731904	360	0	0x4
00:15:6d:84:14:09 00:15:6d:84:14:09 mesh	1	8193	0	2196731904	0	0	0x10

The arp table shows the hardware address of the destination node. This is useful for interpreting the output of the mpath dump command. The mpath dump shows the table that the mesh stack uses to route packets. This table is constructed as the mesh stack successfully establishes paths to destinations in the mesh. Each output line shows the destination address in the first column and the next hop address in the second column. Look up the destination address of our ping packet (00:15:6d:84:14:0e) in this table. You can see that the next hop is 00:15:6d:84:14:10, revealing that the packet will not be sent directly to the destination. Instead, it will be sent to 00:15:6d:84:14:10 who will presumably forward it on appropriately. In contrast, if the destination of the packet were 00:15:6d:84:14:09, the packet would be sent directly to that node. This is because the path table reports the same address for the destination and next hop.

Each entry in the mpath dump has several trailing fields after the destination and next hop addresses. For more information on the meaning of these fields, see the mpath page. Also, be sure to check the the driver status page for any known issues related to your driver.

Et voilà! But keep reading for...

Advanced Tinkering

  • You can modify the mesh path table with these commands:
    • iw dev $MESH_IFACE mpath del $DST_ADDR: to delete a mesh path entry.
    • iw dev $MESH_IFACE mpath new $DST_ADDR next_hop $NEXTHOP_ADDR to force a specific network topology.
  • You can modify the mesh peer links table with these commands:
    • iw dev $MESH_IFACE station dump: list all the plinks.
    • iw dev $MESH_IFACE station $HW_ADDR set plink_action [open|block]: trigers a specific peer link stablishment process or blocks a specific peer link.
    • iw dev $MESH_IFACE station del $HW_ADDR: deletes a peer link from the peer table.
  • You can get and set mesh parameters using the "mesh_param" command in iw. For example, to change the mesh TTL value (which defaults to 5), one would: 
    # iw dev $MESH_IFACE get mesh_param mesh_ttl
5
# iw dev $MESH_IFACE set mesh_param mesh_ttl 3

The following mesh parameters can be inspected and set:

  • mesh_retry_timeout
  • mesh_confirm_timeout
  • mesh_holding_timeout
  • mesh_max_peer_links
  • mesh_max_retries
  • mesh_ttl
  • mesh_auto_open_plinks
  • mesh_hwmp_max_preq_retries
  • mesh_path_refresh_time
  • mesh_min_discovery_timeout
  • mesh_hwmp_active_path_timeout
  • mesh_hwmp_preq_min_interval
  • mesh_hwmp_net_diameter_traversal_time

Wireshark

Patch wireshark and start analyzing traffic. Thanks to Steve Brown for porting this patch from an older version of wireshark

Mesh Portal (MPP)

A Mesh Portal (MPP) provides the ability to bridge networks over a mesh. Although the implementation of the full MPP specification detailed in the 802.11s draft has not yet been finished, an interesting intermediate step has been contributed that will be outlined here.

Note: At this time, mesh points will not be able to ping outside the mesh and vice versa. This is because support for proxied entities in the forwarding table have not been implemented. However, this implementation will allow an MPP to use the 6 address frame format and therefore be able to propagate traffic though the mesh successfully.Thanks to YanBo? for submitting this functionality.

To bring up an MPP we need to setup up a bridge between a mesh interface and, for example, an Ethernet interface.

1. Bring up a mesh interface as described in the "Testing" section above. 

mpp$ iw dev wlan0 interface add $MESH_IFACE type mp mesh_id ${MESH_ID}
mpp$ ifconfig $MESH_IFACE up

2. Add this new interface and your ethN interface to a new bridge: 

mpp$ brctl addbr br0
mpp$ brctl stp br0 off
mpp$ brctl addif br0 eth1
mpp$ brctl addif br0 mesh
mpp$ ifconfig $MESH_IFACE down
mpp$ ifconfig eth1 down
mpp$ ifconfig $MESH_IFACE 0.0.0.0 up
mpp$ ifconfig eth1 0.0.0.0 up
mpp$ ifconfig br0 ${MESH_IP}

If you are having problem setting up the bridge, here is a great howto: Ethernet Bridge + netfilter HOWTO

Comments? Questions? Praise? Flames?

Attachments