netdev_ieee802154_submac: unify legacy and radio HAL bootstrap

[jia200x]

Contribution description

This PR unifies the init procedure for the Radio HAL and legacy netdev_<device> interface. This allocates a netdev_ieee802154_submac_t descriptor if module netdev_ieee802154_submac is present. If it's not present, it allocates the legacy netdev_ieee802154_t only if the Radio HAL is not present. This ensures that no netdev dependencies are pulled if only the Radio HAL is used (e.g OpenWSN).

This whole step allows to run all radios that implement the Radio HAL with netdev_ieee802154_submac without changing the bootstrap code. For instance, #15132 should work with this PR.

TL;DR: USEMODULE+=netdev_ieee802154_submac should work out of the box for all radios that implement the Radio HAL (e.g LWIP, OpenThread can bootstrap a radio in the same way they used to do with netdev)

Testing procedure

• Murdock should do its work compiling tests/ieee802154_hal and tests/ieee802154_submac.
• Try testing gnrc_networking with and without netdev_ieee802154_submac.
• Try LWIP with nrf802154 (the cc2538_rf is not yet in the bootstrap list :/)
• Test that addresses don't change if netdev_ieee802154_submac is selected.

Issues/PRs references

#15132

[jia200x]

done! it works now with netdev_register

[benpicco]

Please squash.

[jia200x]

done! I also dropped the luid dependency and squashed directly

[jia200x]

shall I squash?

[benpicco]

nrf52840dk now gets stuck in a boot loop with examples/gnrc_networking.

Works fine when using netdev_ieee802154_submac.

[benpicco]

The above assertion is triggered, channel = 0

[jia200x]

fixed. I squashes and amended directly (it fails because I removed the initialization). Also, nobody was calling netdev_ieee802154_reset there o.O

[benpicco]

Yup, that fixed it.
btw: when squashing please don't do a git rebase master, instead only do a git rebase HEAD~n where n is the number of commits you want to go back, or git rebase <hash of commit before the PR> so when clicking the force-pushed link only the changes from this PR show up.

I'll give it a try on the openmote-b with cc2538_rf just to be sure 😉

[benpicco]

cc2538_rf is also still working, both with netdev_ieee802154_submac and without.

[jia200x]

btw: when squashing please don't do a git rebase master, instead only do a git rebase HEAD~n where n is the number of commits you want to go back, or git rebase so when clicking the force-pushed link only the changes from this PR show up.

Sure!

[benpicco]

Strange observation (possibly unrelated):

With netdev_ieee802154_submac on nrf52840dk I don't get a public address from the border router in tests/riotboot_flashwrite (and the border router can't ping it) - but without it I always do.

netdev_ieee802154_submac

2020-10-19 18:31:44,300 # Iface  5  HWaddr: 3F:3A  Channel: 26  NID: 0x23 
2020-10-19 18:31:44,304 #           Long HWaddr: DE:07:24:D6:6E:0B:3F:3A 
2020-10-19 18:31:44,306 #            State: IDLE 
2020-10-19 18:31:44,311 #           ACK_REQ  L2-PDU:102  MTU:1280  HL:64  RTR  
2020-10-19 18:31:44,313 #           6LO  IPHC  
2020-10-19 18:31:44,316 #           Source address length: 8
2020-10-19 18:31:44,319 #           Link type: wireless
2020-10-19 18:31:44,325 #           inet6 addr: fe80::dc07:24d6:6e0b:3f3a  scope: link  VAL
2020-10-19 18:31:44,328 #           inet6 group: ff02::2
2020-10-19 18:31:44,330 #           inet6 group: ff02::1
2020-10-19 18:31:44,334 #           inet6 group: ff02::1:ff0b:3f3a

classic netdev

2020-10-19 18:32:12,203 # Configured network interfaces:
2020-10-19 18:32:12,208 # Iface  5  HWaddr: 3F:3A  Channel: 26  NID: 0x23 
2020-10-19 18:32:12,212 #           Long HWaddr: DE:07:24:D6:6E:0B:3F:3A 
2020-10-19 18:32:12,218 #            TX-Power: 0dBm L2-PDU:102  MTU:1280  HL:64  RTR  
2020-10-19 18:32:12,221 #           RTR_ADV  6LO  IPHC  
2020-10-19 18:32:12,224 #           Source address length: 8
2020-10-19 18:32:12,227 #           Link type: wireless
2020-10-19 18:32:12,232 #           inet6 addr: fe80::dc07:24d6:6e0b:3f3a  scope: link  VAL
2020-10-19 18:32:12,238 #           inet6 addr: 2001:db8::dc07:24d6:6e0b:3f3a  scope: global  VAL
2020-10-19 18:32:12,241 #           inet6 group: ff02::2
2020-10-19 18:32:12,244 #           inet6 group: ff02::1
2020-10-19 18:32:12,247 #           inet6 group: ff02::1:ff0b:3f3a

examples/gnrc_networking works either way.

[jia200x]

hmmm strange. I will check what's going on there

[jia200x]

@benpicco did #15237 by chance fix the issue?

[benpicco]

Unfortunately not, the issue is still present on current master.