Allwinner SoC Still Unlikely For Upstream Linux Kernel

[phoronix]

Phoronix: Allwinner SoC Still Unlikely For Upstream Linux Kernel

While Allwinner ARM SoCs are found within massive amounts of the low-cost Android tablets manufactured in China, and there is some open-source Allwinner Linux kernel support, it's still unlikely that the patches will land upstream anytime soon...

http://www.phoronix.com/vr.php?view=MTM4NDc

[madjr]

would also love for rockchip to support something other than android from the get go.

The quad core mk908 is great:

http://liliputing.com/2013/04/tronsmart-mk908-quad-core-android-tv-stick-performance-video.html

for the price, I think these things are the future and we must not let android have this market all by itself.

[hiddenmaverick]

This is most unfortunate, as mainlined Allwinner support would enjoy all the latest kernel updates and enhancements in a timely manner. I hope these issues start to get resolved in the months to come.

[tomato]

hardware manufacturers produce shitty code unless beaten with a 10 foot pole, news at 11

seriously, was there a case in which a new contributor to Linux had their patches mainlined in first iteration (or with just few minor changes, like renaming few functions, fixing spelling mistakes, etc.)?

[rvalles]

Even if Allwinner isn't doing things right on their end, a good deal of support for their SoCs has been written by third parties and is already in the Linux kernel, with even more ongoing:

http://linux-sunxi.org/Linux_mainlining_effort

The story seems to totally miss that this effort exists and how advanced it is. It of course would be great if Allwinner themselves did things properly to begin with. It'll be interesting to see how they react to this.

[droidhacker]

I've been playing with some of these chinese SoC devices. I'm finding that the price is definitely related to the quality. You pay $30, and you get the quality-proportional equivalent to something that would be about $200 from a more reputable manufacturer.

There is also another major barrier to adding support for these things upstream; although there are *some* kernel sources, the word *some* is really REALLY meaningful. You end up with some weird parts like having several binary-only kernel modules, for things like NAND. Rockchip NAND drivers are actually based strongly on MTD, but their own adjustments are BAD (crash prone, for example), and binary only.

We here WERE using allwinner and/or rockchip based tablets and STB's, but for tablets, we've found it actually CHEAPER to use Google Nexus units. We also plan to move over to a more mainstream SoC for STB's, and are waiting for these guys to move past the "pre-order" stage: http://www.inforcecomputing.com/prod...o/ifc6410.html

I really couldn't care less if the china SoC's were supported.

[rvalles]

Cost matters

Good for you. For the rest of us, price matters and these chips are cheap, nice and used in tons of cheap tablets and boards such as the cubieboard or marsboard, which btw are excellent value/money compared to a raspberry pi, as they're based on Cortex-A8 or 2x Cortex-A7 and have 1GB RAM, 4GB flash storage, SATA and Ethernet (in the SoC, not through usb).

[Kivada]

Good for you. For the rest of us, price matters and these chips are cheap, nice and used in tons of cheap tablets and boards such as the cubieboard or marsboard, which btw are excellent value/money compared to a raspberry pi, as they're based on Cortex-A8 or 2x Cortex-A7 and have 1GB RAM, 4GB flash storage, SATA and Ethernet (in the SoC, not through usb).

Maybe, but there are those of us waiting for the unicorn. A high end Cortex-A15 device with a fast GPU that has docs released for all hardware and an unlocked bootloader that lets you erase the Androids/Chrome/whatever other crappy OS it comes with and install a real Linux.

[rvalles]

Neat

Maybe, but there are those of us waiting for the unicorn. A high end Cortex-A15 device with a fast GPU that has docs released for all hardware and an unlocked bootloader that lets you erase the Androids/Chrome/whatever other crappy OS it comes with and install a real Linux.

I wouldn't mind one of those, assuming it wasn't insanely overpriced.

Those Allwinner A20 based boards are just too good value for the money.

[oliver]

I'm a little baffled at the journalistic ineptitude of this article. And I never complain about that because I find it to be unfair to Micheal.

This article should be deleted and redone. Or an errata released. That said, i haven't backread all new posts since then


First, Sun4i and Sun5i (A10, A10s, A13) HAVE MAINLINE SUPPORT TODAY.

http://linux-sunxi.org/Mainlining_Effort

This landed in 3.8. Granted it's not ueber-usefull yet (only some basic bits, but its coming). With 3.11 you can boot a kernel and have networking support. Quite neat.

Also Only Mali and CedarX (the userspace bit) is closed. NAND driver (while shit) is open source. As of this week, even the boot0/boot1 bootloaders (that are on nand flash normally) are GPLed. (U-boot works well and better, but only for mmc, no nand driver yet as that's a big mess, an mtd driver is actually in the works and does work, but needs a lot of polish). So to reitterate, while the sources are ugly, the 3.0 and 3.4 kernels from linux-sunxi (drops by allwinner with tons of work done by the community, yours truly included) is 100% Free, libre software. You can run a fully libre system (from mmc only) if you do not need Mali (you can use limare's) and without cedarx (vpu) userland stuff. The only closed bit (i hear you thinking but read on) is the Boot loader embedded in the CPU, but I think that's unavoidable (there is even a remote chance the BROM may have its source released, while unreplaceable, still nice).


Some background. Luke (lkcl) had a deadline to talk to upper management. He was uninformed and not entirely aware what has happened on linux-sunxi for the past 6-9 months. While sloppy, it happens. He went blah blah ing onto the debian mailinglist and then the lkml and lakml about how FEX is so awesome and better then devicetree. Again, luke was a little uninformed and failed to realise that devicetree does what fex does and then some. Devicetree is actually older then fex so it's a little annoying they didn't use it, but understandable.

After some posts going back/forth, luke learned that: We have mainline support today. We've had it for many months. It uses devicetree from end to end. FEX is gone.



meaning that, this post and luke's are moot and irrelevant and have been before they where even created.