I found on more sites that 4.6 kernel supports A83T. Is possible to boot an image on BPI-M3 with this kernel and use an usb wifi adapter (rt2870 chipset) ?
Nope
http://linux-sunxi.org/Linux_mainlining_effort#Merged_into_4.6
Basic support for A83T means: Boots a single CPU core with a fixed clockspeed. Pretty much nothing else is working. If you’re a kernel developer your best starting point is wens’ repo and the USB/EMAC branches for A83T: https://github.com/wens/linux/tree/a83-emac
If you’re a normal user don’t even think about switching to another kernel. Put your device in the drawer and check situation in 2017 again. Unlike other Allwinner SoCs A83T receives only limited attention so work does not progress that far compared to other SoCs like eg A20 or H3 (better choices)
1 Like
Just to clarify. Adding support is doable - it’s just a substantial chunk of work. FreeBSD already supports the BPI M3 thanks to the work of Jared McNeill:
https://wiki.freebsd.org/FreeBSD/arm/Allwinner
So far, they implemented support only for the 1st CPU cluster (which equals to 4 CPUs), Gbit Ethernet and thermal throttling - plus other things that are common to all Allwinner SoCs.
https://reviews.freebsd.org/rS299871 https://reviews.freebsd.org/rS299084
It is not full support but it’s already very good, compared to what we have on Linux. I am studying his sources at the moment. The real problem is that Allwinner/Sinovoip only provides one single kernel, which is an Android kernel, and to make the matter worse, it’s also heavily customized - eg. you can’t easily reconcile its patches with a mainline 3.4.39 Android kernel. See it for yourself here:
https://android.googlesource.com/kernel/common/+/d80428a98bd7f840ceae5c527baad4086381cbf5
For FreeBSD, it seems that Jared wrote everything from scratch - kudos to him.
Unless you do that - writing from scratch - there is no “easy” way to port the drivers to mainline - the only way is to do it to get a real understanding of how the hardware works - like Jared did on FBSD - and then perhaps get some sort of “inspiration” from the existing Android sources. That requires time - or better - someone working on it fulltime.
I am currently focusing on the “easy” solution of reconciling patches to 3.4.112 - but ATM I can only dedicate few hours per week to it. For various reasons - mainlining is at least one order of magnitude harder so - tkaiser is right, I can’t see that happening anytime soon.
I hope it helps to understand what is the status ATM. Regards, DA
2 Likes
Thx for the update. The progress made by Jared is really impressive. So there’s some hope to enjoy mainline kernel next year on H8 and R58 devices 
I came from the original Raspberry Pi first generation (used for NAS an TV Server). Not I thought it would be a good idea to switch to Banana Pi (better for NAS). But now I get stuck because of the old kernel version (not supported by DvbSky) 
So thanks for the Informations, I will follow the advice of tkaiser (“don’t even think about switching to another kernel. Put your device in the drawer and check situation in 2017”).