b43: Putting out feelers on adding support for BCM4360

b43: Putting out feelers on adding support for BCM4360

[Helga K]

Hi,

I'm a C dev that works in network hardware testing and bringup.
I'm extremely tired of the myriad issues of the broadcom-wl driver, and I'd like to add 
support for the BCM4360 card to b43 instead. The last few times this card was mentioned
by developers were:

* Arend van Spriel in 2015 saying support would be unlikely because Broadcom didn't
  clear it and because it'd require adding VHT support to brcmsmac (that hasn't changed in 
  10 years - brcmsmac still doesn't support VHT+ hardware)
* b43 developers adding a stub saying its PHY wasn't supported (but could be in the future)

AFAIK support for this card in b43 is feasible - it's only AC PHY operations that are missing.

My questions are:

* Is it possible to merge new functionality into the driver given its orphaned state? I'm
  not capable of taking over maintaining b43.
* Is anyone in the community broadly familiar with reverse-engineering Broadcom hardware
  and can bring me up to speed with the state of the art?

Thanks in advance,
Helga

Re: b43: Putting out feelers on adding support for BCM4360

[Jonas Gorski]

Hi Helga,

there is a b43-dev ML that may be more on topic, so I added it to CC:

On Sun, Mar 23, 2025 at 11:21 AM Helga K <Strike0347@proton.me> wrote:
>
> Hi,
>
> I'm a C dev that works in network hardware testing and bringup.
> I'm extremely tired of the myriad issues of the broadcom-wl driver, and I'd like to add
> support for the BCM4360 card to b43 instead. The last few times this card was mentioned
> by developers were:
>
> * Arend van Spriel in 2015 saying support would be unlikely because Broadcom didn't
>   clear it and because it'd require adding VHT support to brcmsmac (that hasn't changed in
>   10 years - brcmsmac still doesn't support VHT+ hardware)
> * b43 developers adding a stub saying its PHY wasn't supported (but could be in the future)
>
> AFAIK support for this card in b43 is feasible - it's only AC PHY operations that are missing.
>
> My questions are:
>
> * Is it possible to merge new functionality into the driver given its orphaned state? I'm
>   not capable of taking over maintaining b43.

Technically it would be possible, the issue would be finding someone
capable of (properly) reviewing and/or testing the code changes.

You may be able to lift some parts from brcmsmac, but the appropriate
handling for the 11AC core and PHY will need to come from somewhere
else. You will also need to add support for BCM2069 radio, which isn't
in b43 yet either, as the appropriate AC tables are also missing.

And there are likely checks missing in the common code for the AC PHY.
So you may find out that it is more work than expected. Not to
discourage you, just to make you aware of what else is missing.

> * Is anyone in the community broadly familiar with reverse-engineering Broadcom hardware
>   and can bring me up to speed with the state of the art?

For b43 and its reverse engineering support there was a wiki [1],
though I don't know how far the effort went. Additionally, you may
find code fragments and part of the code in the original commit for
brcmsmac [2] before it was cleaned up.

Be aware that you will still be limited to 11g/a with b43. There is no
11n support in b43 yet, so no 11n rates, no 40 MHz, no aggregation, no
MIMO, even though 11n capable cores are supported.

If you are daring, you can try lifting the 11n support from brcmsmac
and port it to b43, but you will probably need a supported card for
that for testing before attempting to make it work for 11ac cores.

Best regards,
Jonas

[1] https://bcm-v4.sipsolutions.net/
[2] https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=a9533e7ea3c410fed2f4cd8b3e1e213e48529b75

>
> Thanks in advance,
> Helga