[LEDE-DEV] Backporting stuff to lede-17.01

[Álvaro Fernández Rojas]

After doing some more tests it looks like these problems only come out when bridging WiFi & ethernet on the Raspberry Pi.
https://github.com/raspberrypi/firmware/issues/673

I'm currently running wifi and eth on separate interfaces and so far so good :).

~ Álvaro.

El 07/02/2017 a las 10:53, Álvaro Fernández Rojas escribió:
> Hi Rafał,
> 
> El 02/02/2017 a las 12:32, Rafał Miłecki escribió:
>> Hi,
>>
>> There are few changes that recently went into master and I'd like to
>> see in lede-17.01. Let me what do you think, if backporting any of
>> these is a bad idea.
>>
>> 1) bgmac support for external PHYs
>> 6a853776a502 ("kernel: backport bgmac support for external PHYs")
>> This adds support for some AP devices that don't have external switch
>> or just don't use internal one. Driver is used by brcm47xx/mips74k and
>> bcm53xx, I can test it (and I do) with both targets.
>>
>> 2) hostapd support for complete logging
>> 37b489fe0417 ("hostapd: backport support for sending debug messages to
>> the syslog")
>> This allows reading all error messages (and debugging ones if you
>> enable them) from the syslog (logread command). I find it very
>> important for debugging wireless/hostapd issues. This is a pretty
>> simple single hostapd patch.
>>
>> 3) brcmfmac update
>> There were quite a few improvements/fixes for brcmfmac. That includes
>> small fixes & debugging improvements. There are still some people
>> reporting problems with brcmfmac so getting useful info even with
>> 17.01 release would be desired.
>> We use brcmfmac for BCM43602 & BCM4366 PCIe devices and we support 4
>> routers with these. I own 3 of them so my testing should have a good
>> coverage.
>> One minor risk is I can't test brcmfmac with Raspberry Pi. The good
>> thing is that all patches were backported separately (and in few
>> commits) so any regression should be trivial to track.
> 
> I finally had some time to test this on my Raspberry Pi 3.
> BTW, thanks for these improvements ;D.
> 
> For what I've seen on my tests, performance has really improved since I added support for the Raspberry Pi 3.
> There were some issues causing packet loss, but now it seems that those problems are gone.
> However, I'm still getting these spam messages "brcmfmac: brcmf_proto_bcdc_hdrpull: wlan0: non-BCDC packet received, flags 0x??", and don't know where they come from...
> http://pastebin.com/1iEVLYQk
> Should we keep that message from being printed for brcm2708?
> 
> P.S: I will keep my Raspberry Pi 3 running for a while in order to check that performance doesn't degrade over time.
> 
> Regards,
> Álvaro.
>