802.11k Neighbor report daemon

kissadamfkut · August 4, 2021, 5:51pm

The 802.11k standard specifies how the neighboring BSS' could report each other to a STA. OpenWRT's hostapd supports this feature, but the synchronization has to be done in a separate daemon such as this one.

It is more lightweight than DAWN (many thanks to Polynomdivision), basically it is based on this script How does rrm work? - #60 by ParanoidZoid although it is improved a bit so the neighbor report would include the other band interface for the same AP and some more fixes.

Tested on 20.02rc2 with 22 different AC1750. Using 802.11r configured roam over the air, and disabled the bss transition bit on the network it works like a charm. Walking with my 2,4GHz only phone wifiman show roaming well. My Arch based dual band laptop works even better. Cannot even notice that walking along the corridor my BSS changes nearly every second step.

I've plans to make a package from it and make it published as an ipk. I'll also add a sample config file for /etc/config/wireless so the whole test setup could be reconstructed.

Update: I've missed the github link: https://github.com/simonyiszk/openwrt-rrm-nr-distributor

joe_internet · September 3, 2021, 10:14pm

This is pretty awesome, thanks!

I have it running across 4 AP's and it seems to handle the transition pretty smoothly. I noticed your repo lacks a sample wireless config, so I'm pasting the relevant bits from mine should anyone else want to give this a go.

For starters, each AP is configured as a "Dumb AP" according to the docs. I have also installed the following:

wpad-wolfssl
umdns
libseccomp
scmp_sys_resolver
iwinfo

To get umdns working, I configured /etc/config/umdns for my network:

config umdns
  option jail 1
  list network lan

I also had to comment out the seccomp line in /etc/init.d/umdns as described here.

For /etc/config/wireless, I set up 802.11r by adding the below to the wifi-iface section for my SSID:

option ieee80311r ‘1’
option ft_over_ds ‘0'
option ft_psk_generate_local ‘1’

I then set up 802.11k and 802.11v (steering) as follows:

option ieee80211k '1'
option ieee80211v '1'
option wnm_sleep_mode '1'
option bss_transition '0'
option time_advertisement '2'
option time_zone 'CST+6CDT,M3.2.0/2,M11.1.0/2'

My time zone is explicitly specified so as to appear in the beacons although I've read in some threads this isn't necessary and would be handled automatically by hostapd. For that matter, I've read that 802.11v isn't required anymore either, but I've left it in anyway.

I've also disabled bss_transition and set FT to over-the-air as you indicated in your README.

Finally, I copied your rrm_nr scripts and started them up. This was running the latest rc4 of the 21.02 release.

This appears to be working, but aside from seeing FT skipping the handshake in the logs and my client roaming from one AP to the other as I walk about the space, I haven't done much else to prove it's working as designed.

kissadamfkut · September 4, 2021, 1:45pm

Your seems to be right. I've updated the repo with my config file. (I've just recognized that the nasid is the same for the 2.4 GHz and the 5 GHz radio. I'm a bit uncertain if its good this way.)

github.com

simonyiszk/openwrt-rrm-nr-distributor/blob/0cc0d7d697f5ce0b595dfd4101a93441a5afc19e/example_wireless_config


config wifi-device 'radio0'
        option type 'mac80211'
        option hwmode '11a'
        option path 'pci0000:00/0000:00:00.0'
        option htmode 'VHT20'
        option txpower '18'
        option cell_density '0'
        option beacon_int '50'
        option distance '30'
        option country 'HU'
        option channel '60'
        option log_level '1'

config wifi-iface 'default_radio0'
        option device 'radio0'
        option mode 'ap'
        option key 'intranetintranetintranet'
        option ssid 'intranet'
        option encryption 'psk2'

This file has been truncated. show original

Fixing umdns startap's seccomp section is also an issue in my devices.

As far as I know bss_transition should be disabled, since this package doesn't actively suggest roaming for the STAs.

For testing you can try ubus call hostapd.wlan0 rrm_nr_list. If it shows your other BSSes, than the daemon is working. You could try wifiman from Ubiquiti for eg. to test this on your android STA. Just connect to your AP at 2.4 GHz, and if the 5 GHz shows up immediately in the BSS list for the SSID, everything is okay. Getting the traffic with a laptop in monitoring mode and wireshark could also show what is in the air. The edge of this project is hostapd's nr_list. Capturing management traffic from the AP itself should also work.

eR2022 · February 19, 2023, 4:22pm

It's a pretty old post, but have to say it's working great on my latest SNAPSHOT !
Thanks

_FailSafe · April 21, 2023, 4:05pm

@kissadamfkut This is a really slick tool! Thanks for putting it out there for others

I gave it a try yesterday and unfortunately hit an issue with the TXT record length given the number of SSIDs in my environment (details in the PR). So, I put together a PR with tested code that resolves this issue for me. Hopefully it can help others, too:

github.com/simonyiszk/openwrt-rrm-nr-distributor

Switch to TXT record per neighbor

simonyiszk:main ← Fail-Safe:main

opened 04:02PM - 21 Apr 23 UTC

Fail-Safe

+33 -11

While testing this, I found that due to the number of SSIDs in my environment, I… was exceeding the 255 char limit for a single TXT record. Therefore, my neighbor TXT record was being truncated and failed to work properly. This PR is a tested conversion to creating an array of TXT records whereby neighbors now have a 1:1 TXT record. Each TXT record format is: `<SSID>=[ "<MAC>", "<SSID>", "<NR Information Element>" ]` For example (w/sensitive bits redacted): ```json root@AP:~# ubus call umdns browse '{ "service": "_rrm_nr._udp", "array": true }' { "_rrm_nr._udp": { "AP-LivingRoom": { "iface": "br-lan.1", "ipv4": [ "192.168.xx.xx" ], "ipv6": [ "fe80::ea9f:80ff:fef0:6886", "2600:1700:994::x" ], "port": 5247, "txt": [ "Z9_example=[ \"e8:9f:80:xx:yy:zz\", \"Z9_example\", \"e89f80xxyyzze.......................\" ]", "Z99_example=[ \"ea:9f:80:xx:yy:zz\", \"Z99_example\", \"ea9f80xxyyzze.......................\" ]", "Z7_3_example=[ \"ee:9f:80:xx:yy:zz\", \"Z7_3_example\", \"ee9f80xxyyzze.......................\" ]", "Z9_3_example=[ \"e2:9f:80:xx:yy:zz\", \"Z9_3_example\", \"e29f80xxyyzze.......................\" ]", "Z9_example=[ \"e8:9f:80:xx:yy:zz\", \"Z9_example\", \"e89f80xxyyzze.......................\" ]", "Z7_3_example=[ \"ea:9f:80:xx:yy:zz\", \"Z7_3_example\", \"ea9f80xxyyzze.......................\" ]", "Z9_3_example=[ \"ee:9f:80:xx:yy:zz\", \"Z9_3_example\", \"ee9f80xxyyzze.......................\" ]" ] }, "AP-Office": { "iface": "br-lan.1", "ipv4": [ "192.168.xx.xx" ], "ipv6": [ "2600:1700:994::x", "fe80::ea9f:80ff:fe50:5ca8" ], "port": 5247, "txt": [ "Z7_example=[ \"e8:9f:80:xx:yy:zz\", \"Z7_example\", \"e89f80xxyyzze.......................\" ]", "Z9_example=[ \"ea:9f:80:xx:yy:zz\", \"Z9_example\", \"ea9f80xxyyzze.......................\" ]", "Z99_example=[ \"ee:9f:80:xx:yy:zz\", \"Z99_example\", \"ee9f80xxyyzze.......................\" ]", "Z9_3_example=[ \"e2:9f:80:xx:yy:zz\", \"Z9_3_example\", \"e29f80xxyyzze.......................\" ]", "Z7_example=[ \"e8:9f:80:xx:yy:zz\", \"Z7_example\", \"e89f80xxyyzze.......................\" ]", "Z9_example=[ \"ea:9f:80:xx:yy:zz\", \"Z9_example\", \"ea9f80xxyyzze.......................\" ]", "Z9_3_example=[ \"ee:9f:80:xx:yy:zz\", \"Z9_3_example\", \"ee9f80xxyyzze.......................\" ]" ] } } } ``` Unfortunately, the `package/network/services/umdns/Makefile` is not synced to the latest umDNS commit, which includes a number of improvements including the ability to pass a service name by which to filter the `browse` call, as well as a boolean flag to return the browse result as an array. For this reason, I am including a patch file that will update the umDNS makefile to pull these latest updates. I realize this is not an ideal situation, but perhaps John Crispin could be convinced to push an official update to the umDNS makefile.

The updated format I'm proposing looks like this:

root@AP:~# ubus call umdns browse '{ "service": "_rrm_nr._udp", "array": true }'
{
	"_rrm_nr._udp": {
		"AP-LivingRoom": {
			"iface": "br-lan.1",
			"ipv4": [
				"192.168.xx.xx"
			],
			"ipv6": [
				"fe80::ea9f:80ff:fef0:6886",
				"2600:1700:994::x"
			],
			"port": 5247,
			"txt": [
				"Z9_example=[ \"e8:9f:80:xx:yy:zz\", \"Z9_example\", \"e89f80xxyyzze.......................\" ]",
				"Z99_example=[ \"ea:9f:80:xx:yy:zz\", \"Z99_example\", \"ea9f80xxyyzze.......................\" ]",
				"Z7_3_example=[ \"ee:9f:80:xx:yy:zz\", \"Z7_3_example\", \"ee9f80xxyyzze.......................\" ]",
				"Z9_3_example=[ \"e2:9f:80:xx:yy:zz\", \"Z9_3_example\", \"e29f80xxyyzze.......................\" ]",
				"Z9_example=[ \"e8:9f:80:xx:yy:zz\", \"Z9_example\", \"e89f80xxyyzze.......................\" ]",
				"Z7_3_example=[ \"ea:9f:80:xx:yy:zz\", \"Z7_3_example\", \"ea9f80xxyyzze.......................\" ]",
				"Z9_3_example=[ \"ee:9f:80:xx:yy:zz\", \"Z9_3_example\", \"ee9f80xxyyzze.......................\" ]"
			]
		},
		"AP-Office": {
			"iface": "br-lan.1",
			"ipv4": [
				"192.168.xx.xx"
			],
			"ipv6": [
				"2600:1700:994::x",
				"fe80::ea9f:80ff:fe50:5ca8"
			],
			"port": 5247,
			"txt": [
				"Z7_example=[ \"e8:9f:80:xx:yy:zz\", \"Z7_example\", \"e89f80xxyyzze.......................\" ]",
				"Z9_example=[ \"ea:9f:80:xx:yy:zz\", \"Z9_example\", \"ea9f80xxyyzze.......................\" ]",
				"Z99_example=[ \"ee:9f:80:xx:yy:zz\", \"Z99_example\", \"ee9f80xxyyzze.......................\" ]",
				"Z9_3_example=[ \"e2:9f:80:xx:yy:zz\", \"Z9_3_example\", \"e29f80xxyyzze.......................\" ]",
				"Z7_example=[ \"e8:9f:80:xx:yy:zz\", \"Z7_example\", \"e89f80xxyyzze.......................\" ]",
				"Z9_example=[ \"ea:9f:80:xx:yy:zz\", \"Z9_example\", \"ea9f80xxyyzze.......................\" ]",
				"Z9_3_example=[ \"ee:9f:80:xx:yy:zz\", \"Z9_3_example\", \"ee9f80xxyyzze.......................\" ]"
			]
		}
	}
}

A secondary benefit to the proposed updates is that I switched to using a '+' as the delimiter instead of '|'. Therefore the previous known issue around not being able to use '|' in SSID names is resolved.

kissadamfkut · April 21, 2023, 9:02pm

Thank you for the PR, just accepted it! Hope this daemon will serve well.

_FailSafe · April 22, 2023, 12:53am

Sure thing! I'm excited to have stumbled on this thread and your project because I have been trying to find a good way to dynamically populate my WAP neighbors without using Dawn or Usteer. Very nice work!

FYI, I started another thread around umDNS and the hope that the latest commits in the mdnsd project get pulled into the OpenWrt branches soon. For anybody interested: Query, Fetch, and Browse with umDNS

If the latest commits get pulled in soon, the patch I included in the PR can be dropped.

_FailSafe · April 24, 2023, 10:05pm

I created another PR and welcome your review and testing of it prior to acceptance. It's a concept that came to mind as I was reviewing the code originally. Feel free to decline it if you feel it's not the right direction for your tool

github.com/simonyiszk/openwrt-rrm-nr-distributor

Draft: Refactor to implement hash map functionality

simonyiszk:main ← Fail-Safe:main

opened 10:03PM - 24 Apr 23 UTC

Fail-Safe

+93 -10

This is a proposed refactor that makes use of a hash map that gets stored in a n…ew directory in `/tmp`. As `/tmp` resides on tmpfs (volatile memory), this should not be a costly write for limited write storage devices. The goal of the hash map is to do a one-time mapping of hostapd devices --> SSID names. In this way, the script can lookup against that hash map instead of having to call expensive `ubus list hostapd.*` and `iwinfo <iface> info` each `do_updates` iteration. Also added a trigger for `network` config file changes to cause a service restart.

UPDATE I am rethinking parts of this PR, so I have placed it in Draft for now.

_FailSafe · April 25, 2023, 9:23pm

This is a new PR that incorporates a few of the ready-to-go updates from PR#6. While I am mulling over that PR, I wanted to go ahead and bring forward some painless fixes:

github.com/simonyiszk/openwrt-rrm-nr-distributor

Fix TXT record duplication and fix for unclean exit (Redo)

simonyiszk:main ← Fail-Safe:bug/fix-txt-record-duplication

opened 09:59PM - 25 Apr 23 UTC

Fail-Safe

+23 -5

These should be safe changes: 1. Per mDNS standard, avoid duplicate TXT key n…ames by switching to `SSID<incrementing #>=[ "<MAC>", "<SSID>", "<NR Information Element>" ]` 2. In order to facilitate a clean `rrm_nr` service exit, the bin script now has an alternate approach to the 60 second refresh whereby instead of sleeping for 60 seconds, it only sleeps for 3 seconds * 20 iterations. This results in a 60 second delay between update checks, but allows the service to end the bin script cleanly without timing out and having to issue a SIGKILL.

rossini · April 28, 2023, 7:55pm

So if i am correct, the current version does not work as umdns is not up to date in current snapshot?

_FailSafe · April 28, 2023, 9:13pm

Correct, you can either pull this commit and use the present version of umDNS or pull the latest commit and use the patch file to update umDNS to HEAD for a snapshot build.

Spacebar · October 28, 2023, 7:20am

Thanks for all the work Will try this soon on my AP's.

This is very useful information in the Readme when mentioning umdns. For new people reading this thread, this is the doc for umdns https://openwrt.org/docs/guide-developer/mdns

Is this still something we need to do in 2023?

And I saw that they have updated umdns to latest version in Main. So I believe we just need to copy bin and initscript, install and configure umdns?