This may not even be the correct location to post this thread, as I can not find anything "yet" this will have to do.
I have looked high and lo on google for an answer to this problem and as of yet have not found a "definitive" answer or a how-to.
Multiple mesh nodes using 802.11n for client connectivity with 802.11a as a backhaul.
Hardware in use (lab environment):
Gateway - standard x86 PC with an uplink and a Wistron DNMA-92 for the backhaul.
Mesh node - PC-Engines WRAP 2C SBC with an intersil NL2511 MP PLUS (I know its supposed to by 802.11n but its the only spare card I have), and a Wistron DNMA-92 (again for backhaul)
Hardware in use (production environment):
All nodes - PC-Engines ALIX 3D3 with 2 Wistron DNMA-92 for the backhaul and client connectvity.
O/S Version in use is (for the gateway PC) Debian 0.6.0a with kernel 2.6.39, all nodes use Voyage 0.7.0 (June 2, 2011 build) with kernel 2.6.39
Network class planning on using is 10.x.x.x class A.
Issues and concerns:
1. Using ad-hoc for the backhaul. Ive read that this is not a good practice for large number of connections and since we plan on having a couple of hundred at least, probably not a good idea. So I am open to any configuration. Ive even started to look at batman. At this point in time, Id be open to changing the protocol to hostapd for both wlan0 and wlan1.
2. Getting both wlan0 and wlan1 to pass traffic. When I was able to get the backhaul working, I could not ping the client interface (wlan0) from the gateway computer or ping out from the node - Possible iptables problems.
3. The network of a class A only allows for 16+ million IP Addresses. My concern is that if we outgrow this for larger networks, we will need more IP addresses. my main issue of concern with this, is that we are trying to implement a mobile network and allow end users to hop onto different nodes with out loosing their IP address. I guess, that if they need to renew their IP address then that will be a different case. And if that is the case, then instead of 16+ million network wide, it will be 16+ million per node and that will definitely not be an issue. We will just have to use NAT for everything. And for that matter we may have to use NAT for communication between wireless cards.
4. Want to only have one DHCP server in the mix if using a network wide configuration. if need to use dhcp on each of the nodes as outlined in issue/concern #3 above, that is OK as well.
I have tried OLSRD, hostapd, ad-hoc mode on backhaul (only woks for about a day and then no longer connects).
Any and I mean ANY help will be most appreciative.
Related Article:ebtables not working at all
is a Linux and Unix discussion thread by art_st that has 1 reply, was last updated 2 years ago and has been tagged with the keywords: debian, filtering, linux, networking, networking.