[BUG] Re: [PATCH] brcmfmac: use ISO3166 country code and 0 rev as fallback

[Soeren Moch]

Hi Shawn,

On 08.09.21 03:00, Shawn Guo wrote:
> Hi Soeren,
>
> On Tue, Sep 07, 2021 at 09:22:52PM +0200, Soeren Moch wrote:
>> On 25.04.21 13:02, Shawn Guo wrote:
>>> Instead of aborting country code setup in firmware, use ISO3166 country
>>> code and 0 rev as fallback, when country_codes mapping table is not
>>> configured.  This fallback saves the country_codes table setup for recent
>>> brcmfmac chipsets/firmwares, which just use ISO3166 code and require no
>>> revision number.
>> This patch breaks wireless support on RockPro64. At least the access
>> point is not usable, station mode not tested.
>>
>> brcmfmac: brcmf_c_preinit_dcmds: Firmware: BCM4359/9 wl0: Mar  6 2017
>> 10:16:06 version 9.87.51.7 (r686312) FWID 01-4dcc75d9
>>
>> Reverting this patch makes the access point show up again with linux-5.14 .
> Sorry for breaking your device!
>
> So it sounds like you do not have country_codes configured for your
> BCM4359/9 device, while it needs particular `rev` setup for the ccode
> you are testing with.  It was "working" likely because you have a static
> `ccode` and `regrev` setting in nvram file.
It always has been a mystery to me how country codes are configured for
this device. Before I read your patch I did not even know that a
translation table is required. Is there some documentation how this is
supposed to work? Not sure if this makes a difference, BCM4359/9 is a
Cypress device I think, I added mainline support for it some time ago.

I have installed different firmware files, brcmfmac4359-sdio.clm_blob,
brcmfmac4359-sdio.bin, brcmfmac4359-sdio.txt, the latter also linked as
brcmfmac4359-sdio.pine64,rockpro64-2.1.txt. This probably is the nvram
file. ccode and regrev are set to zero, which probably means
'international save settings".
> But roaming to a different
> region will mostly get you a broken WiFi support.  Is it possible to set
> up the country_codes for your device to get it work properly?
In linux-5.13 it worked, probably with save settings (not all channels
selectable, limited tx power), with linux-5.14 it stopped working, so it
is a regression.
I personally would like to learn how all this is configured properly.
For general use I think save settings are better than no wifi at all
with this patch. This fallback to ISO CC seams to work with newer
(Synaptics?) devices only.

Soeren
>
> Shawn
>
>> Regards,
>> Soeren
>>> Signed-off-by: Shawn Guo <shawn.guo at linaro.org>
>>> ---
>>>  .../broadcom/brcm80211/brcmfmac/cfg80211.c      | 17 +++++++++++------
>>>  1 file changed, 11 insertions(+), 6 deletions(-)
>>>
>>> diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
>>> index f4405d7861b6..6cb09c7c37b6 100644
>>> --- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
>>> +++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
>>> @@ -7442,18 +7442,23 @@ static s32 brcmf_translate_country_code(struct brcmf_pub *drvr, char alpha2[2],
>>>  	s32 found_index;
>>>  	int i;
>>>
>>> -	country_codes = drvr->settings->country_codes;
>>> -	if (!country_codes) {
>>> -		brcmf_dbg(TRACE, "No country codes configured for device\n");
>>> -		return -EINVAL;
>>> -	}
>>> -
>>>  	if ((alpha2[0] == ccreq->country_abbrev[0]) &&
>>>  	    (alpha2[1] == ccreq->country_abbrev[1])) {
>>>  		brcmf_dbg(TRACE, "Country code already set\n");
>>>  		return -EAGAIN;
>>>  	}
>>>
>>> +	country_codes = drvr->settings->country_codes;
>>> +	if (!country_codes) {
>>> +		brcmf_dbg(TRACE, "No country codes configured for device, using ISO3166 code and 0 rev\n");
>>> +		memset(ccreq, 0, sizeof(*ccreq));
>>> +		ccreq->country_abbrev[0] = alpha2[0];
>>> +		ccreq->country_abbrev[1] = alpha2[1];
>>> +		ccreq->ccode[0] = alpha2[0];
>>> +		ccreq->ccode[1] = alpha2[1];
>>> +		return 0;
>>> +	}
>>> +
>>>  	found_index = -1;
>>>  	for (i = 0; i < country_codes->table_size; i++) {
>>>  		cc = &country_codes->table[i];