iiab-clone-wifi needs template from hostapd - record wifi_up_down [& doc re: Ansible facts like ansible_ap0, ansible_eth0, ansible_wlan0] [& 5GHz versus 2.4GHz test]

[jvonau]

Fixes bug:

This issue should show up on a clean install with wifi_up_down: True, slight hole in testing

Description of changes proposed in this pull request:

iiab-clone-wifi needs template from hostapd.yml that moved so reposition the restart to be after hostapd.yml

Smoke-tested on which OS or OS's:

Not as of yet, but if installs start failing at the noted restarting of iiab-clone-wifi, the service file is not installed yet.

Mention a team member @username e.g. to help with code review:

[holta]

@EMG70 if you have time to test this in the coming week, that'd be incredible! Using either approach:

1. Fresh install of PR iiab-clone-wifi needs template from hostapd - record wifi_up_down [& doc re: Ansible facts like ansible_ap0, ansible_eth0, ansible_wlan0] [& 5GHz versus 2.4GHz test] #3555 with IIAB on a new OS:

   curl iiab.io/install.txt | bash -s 3555
   

2. Or on an existing IIAB:

   sudo rm /etc/iiab/install-flags/iiab-complete
   sudo iiab 3555
   

CLARIF / TEST PLAN: (DRAFT)

• Make sure wifi_up_down: True remains in /etc/iiab/local_vars.yml during all testing of this PR (3555).
• If IIAB's hotspot is working well out-of-the-gate...
  • Run: sudo iiab-hotspot-off
  • Try connecting IIAB to your home's WiFi (actual Internet) by clicking on the widget (for WiFi configuration) in the top-right of the IIAB's Graphical Desktop. Or use sudo raspi-config or similar, if your IIAB's OS has no graphical desktop.
  • Run: sudo iiab-hotspot-on
• Test:
  • Confirm IIAB's still online, after removing its Ethernet cable as an experiment ?
  • Confirm whether WiFi client devices can still connect to IIAB and view the IIAB home page ?

Tangentially related:

• PR RFC 2210 2304 fixed https://bugs.launchpad.net/bugs/1951586 [netplan WiFi country code / regulatory domain] #3537
• PR Unified PR for WiFi hotspot testing of channel-changing, country code, etc #3549

[jvonau]

No this issue will only show up on a fresh install with wifi_up_down: True as that is what puts the service file in place.

[jvonau]

Better yet just do a plain install with unittest to confirm there is a bug, then apply the fix.

[jvonau]

Confirmed bug's presents and this is the fix http://sprunge.us/K6rCCu

[holta]

What sets ansible_ap0 to help make this work?

(Is there a better search than...

https://github.com/iiab/iiab/search?q=ansible_ap0

...to help shed light on these kinds of questions ?)

[holta]

I don't know where Ansible best documents this. But Ansible evidently sets up a variable like ansible_eth0 for each network interface, e.g.:

root@box:~# ansible -m setup localhost | grep -A85 ansible_eth0 | pastebinit -b sprunge.us
[WARNING]: No inventory was parsed, only implicit localhost is available
( http://sprunge.us/G0YIUr?en )

Which I assume means that variable ansible_ap0 does not exist until... the next Ansible run after ap0 [*] is created?

[*] ap0 is the downstream hotspot interface side of... IIAB's bidirectional WiFi... i.e. when wifi_up_down: True is set.

[jvonau]

Funny that search point right to the answer. Guess you forgot about #3389 (comment) where that discussion is documented.

[holta]

Ok.

(Ansible docs are improving, but they're still not very searchable in cases like this.)

[jvonau]

Ref: https://docs.ansible.com/ansible/latest/playbook_guide/playbooks_vars_facts.html#vars-and-facts

[holta]

@EMG70 worked hard to test this with a new install of SMALL-sized IIAB on 64-bit RasPiOS with Desktop, on RPi 400 — and asks if everything looks Ok below?

Does this mean 5GHz is partially working?
(i.e. is 2.4GHz perhaps no longer mandatory in some cases?)

1165 COMMAND: /usr/sbin/iw dev    # List wireless interfaces
1166 
1167 phy#0
1168 	Unnamed/non-netdev interface
1169 		wdev 0x3
1170 		addr e6:5f:01:0d:8f:7f
1171 		type P2P-device
1172 		txpower 31.00 dBm
1173 	Interface ap0
1174 		ifindex 4
1175 		wdev 0x2
1176 		addr 02:64:ce:eb:c5:6d
1177 		ssid Internet in a Box
1178 		type AP
1179 		channel 36 (5180 MHz), width: 80 MHz, center1: 5210 MHz
1180 		txpower 31.00 dBm
1181 	Interface wlan0
1182 		ifindex 3
1183 		wdev 0x1
1184 		addr e4:5f:01:0d:8f:7f
1185 		ssid SKY3C17E
1186 		type managed
1187 		channel 36 (5180 MHz), width: 80 MHz, center1: 5210 MHz
1188 		txpower 31.00 dBm

His iiab-diagnostics: http://sprunge.us/BtTyun?en

He writes:

Ethernet cable was removed after installation and connected RPI400 to SKY3C17E (home wifi)
Here are results.
I can connect to internet in a box hotspot and view iiab homepage OK

At first iiab-test-wifi had hostapd on different channel and message saying editing channel..
It changed on next display without doing any changes

If I hook up to Sky router, I can also connect to box directly via my dhcp enabled router

His screenshot:

[holta]

@EMG70 writes:

Just disabled 5GHz in router advanced menu

There is not really a function for disabling 5GHz but just disable syncing on the two channels.
Soon as I did that the 5GHz disappeared on my CLI tests

In short, he disabled both checkboxes, by the red below:

Then he ran iw dev (on the RPi 400) whose output is here:

And he ran iiab-test-wifi whose output is here:

[jvonau]

I feel I keep repeating myself but the message gets lost somewhere. If you note in the picture of the router that there are different tabs at the top for 2.4/5GHz. I'll bet one can have a different ssid between 2.4/5GHz, change one to be able to tell the bands apart. I don't have time to get into details as I'm moving and may not have internet for awhile.

[jvonau]

Carefully reread #3537 but be warned that RasPiOS doesn't offer the same band limiting feature as netplan does. The band blocking is done on netplan (I should of posted the generated /run/netplan/* file) by adding a list of frequencies (freq_list:) to the generated wpa_supplicant.conf file. See https://w1.fi/cgit/hostap/plain/wpa_supplicant/wpa_supplicant.conf

[holta]

I'll bet one can have a different ssid between 2.4/5GHz, change one to be able to tell the bands apart.

Indeed @EMG70 ran exactly that experiment over the past hour, renaming his home's 5GHz hotspot to "SKY3C17E-5G" for now.

And it worked, for some reason even despite the fact that he forgot to run iiab-hotspot-off before clicking on RasPiOS's WiFi-SSID-selecting wizard (risky, but the OS appears not to be damaged in this case!)

Finally, it seems he can choose either 2.4 GHz or 5 GHz (clicking on the graphical wizard to select one Upstream SSID or the other) as shown below, in this case with 5 GHz across both of the RPi 400's WiFi interfaces:

OBSERVATION: Channel 11 (as set in /etc/iiab/local_vars/yml) is entirely ignored above, but at least it works on Channel 36 as mandated by the home's router.

I'm moving

Good Luck with the move.

[jvonau]

And it worked, for some reason even despite the fact that he forgot to run iiab-hotspot-off before clicking on RasPiOS's WiFi-SSID-selecting wizard (risky, but the OS appears not to be damaged in this case!)

That would be because of the 'hooks' that are installed within dhcpcd or networkd-dispatcher. Does more or less the same function as iiab-test-wifi but on the fly not just at boot.

Finally, it seems he can choose either 2.4 GHz or 5 GHz (by selecting one Upstream SSID or the other) as shown here with 5 GHz across both of the RPi 400's WiFi interfaces:

Good, now you can document how your home AP can (should?) be configured.

OBSERVATION: Channel 11 (as set in /etc/iiab/local_vars/yml) is entirely ignored above, but at least it works on Channel 36 as mandated by the home's router.

Same would hold true for country code.

[jvonau]

Same would hold true for country code.

That would only apply for iiab-test-wifi run during boot, the hooks just deal with channel.