cpu/nrf52: initial kconfig modeling (no netif)

[fjmolinas]

Contribution description

This PR provides Kconfig dependencies modeling for nrf52 and two nrf52 BOARDs.

Testing procedure

• check the modeling
• murdock should do
• look at menuconfig

Issues/PRs references

Split out of #16780

[MrKevinWeiss]

Is guess this depends on #16836

[jia200x]

I propose here a couple of changes:

• This should probably be in cpu/nrf52
• I think we should define an aux symbol for this radio (e.g NRF52_RADIO) that declares a feature (HAVE_NRF52RF_RADIO). This one can be set to y if MODULE_NETDEV_DEFAULT is present. Then, the backend selection depends on HAVE_NRF52_RADIO.

[fjmolinas]

* This should probably be in `cpu/nrf52`

Ups, yes of course.

* I think we should define an aux symbol for this radio (e.g NRF52_RADIO) that declares a feature (HAVE_NRF52RF_RADIO). This one can be set to `y` if MODULE_NETDEV_DEFAULT is present. Then, the backend selection depends on HAVE_NRF52_RADIO.

I don't understand why the auxiliary symbol needs to be used seems like a weird dependency. I don't see why HAVE_NRF52RF_RADIO would depend on NETDEV_DEFAULT either, if it's a feature it should always be selected. Maybe if you can add a code snipped I can understand better your suggestion.

[MrKevinWeiss]

ping @leandrolanzieri I think this was the thing we were talking about.

[MrKevinWeiss]

I started #16845 with the HAVE_* but think maybe that is not the best way to move forward. I think the stage of discussion we are at is:

The boards/cpu can select things based on other modules enabled (in my case STDIO, in your case NETDEV_DEFAULT) if selecting is simple and don't have many dependencies or so many options.
The HAVE_* is like FEATURES_PROVIDED for the features that are not provided in the makefile.

I am still pretty murky of the issues that can arise so it may be good to get together and work through the corner cases.

[leandrolanzieri]

According to what I see in the Makefile.dep, the selection of these modules depends on CPU_MODEL, I think it might have to do with having HAS_RADIO_NRF802154 but I'm not sure, as this is done by model name so far.

I think it makes sense to have a symbol like jose proposes, which selects a module that will enable this choice. I'm thinking of something in the direction of:

config HAVE_NRF52_RADIO
    bool
    select MODULE_NRF52_RADIO if MODULE_NETDEV_DEFAULT
    help
        Indicates that an NRF52 radio is present.

menuconfig MODULE_NRF52_RADIO
    bool "NRF52 radio driver"
    depends on HAVE_NRF52_RADIO
    depends on TEST_KCONFIG

if MODULE_NRF52_RADIO

choice
    bool "NRF52 radio backend"

config NRF802154
    bool "nrf802154"
    select MODULE_NRF802154

config NRFBLE
    bool "nrfble"
    select MODULE_NRFBLE

config NRFMIN
    bool "nrfmin"
    select MODULE_NRFMIN

endchoice

endif

[leandrolanzieri]

Thanks for this! Some initial thoughts

[leandrolanzieri]

MODULE_SAUL_NRF_TEMPERATURE is already defined in drivers/saul/Kconfig.
Maybe we can follow here the 'feature selects module` approach?

config HAVE_SAUL_NRF_TEMPERATURE
    bool
    select MODULE_SAUL_NRF_TEMPERATURE if MODULE_SAUL_DEFAULT && HAS_PERIPH_TEMPERATURE
    help
        Indicates that a SAUL wrapper to the temperature peripheral is present.

[fjmolinas]

addressed

[leandrolanzieri]

Enabling MODULE_SAUL_GPIO when MODULE_SAUL_DEFAULT is done already in drivers/saul/Kconfig if HAVE_SAUL_GPIO is provided.

[leandrolanzieri]

I still think this is not needed.

[leandrolanzieri]

I think this feature should be placed in drivers/lisdh12/Kconfig and selected by the boards that have the device.

[leandrolanzieri]

According to what I see in the Makefile.dep, the selection of these modules depends on CPU_MODEL, I think it might have to do with having HAS_RADIO_NRF802154 but I'm not sure, as this is done by model name so far.

I think it makes sense to have a symbol like jose proposes, which selects a module that will enable this choice. I'm thinking of something in the direction of:

config HAVE_NRF52_RADIO
    bool
    select MODULE_NRF52_RADIO if MODULE_NETDEV_DEFAULT
    help
        Indicates that an NRF52 radio is present.

menuconfig MODULE_NRF52_RADIO
    bool "NRF52 radio driver"
    depends on HAVE_NRF52_RADIO
    depends on TEST_KCONFIG

if MODULE_NRF52_RADIO

choice
    bool "NRF52 radio backend"

config NRF802154
    bool "nrf802154"
    select MODULE_NRF802154

config NRFBLE
    bool "nrfble"
    select MODULE_NRFBLE

config NRFMIN
    bool "nrfmin"
    select MODULE_NRFMIN

endchoice

endif

[leandrolanzieri]

I think this needs rebasing now

[fjmolinas]

Rebased and I think I implemented your suggestions @jia200x and @leandrolanzieri.

[fjmolinas]

Since you suggested radio dependencies to be in CPU I moved the Makefile.dep as well...

[leandrolanzieri]

I still think this is not needed.

[leandrolanzieri]

The static tests don't like this.

[fjmolinas]

lets see if I addressed it

[leandrolanzieri]

Let's give it a try with the CI

[fjmolinas]

I'm seeing some unrelated failures, but lets let it run, if those are the only failures I could squash afterwards...

[fjmolinas]

Finally green! Ok to squash @leandrolanzieri?

[leandrolanzieri]

Finally green! Ok to squash @leandrolanzieri?

Please go ahead

[leandrolanzieri]

Thanks again @fjmolinas for tackling this one. Modelling looks good and the CI agrees -> ACK!