New ARM Platforms Supported By Linux 3.14 Kernel

[phoronix]

Phoronix: New ARM Platforms Supported By Linux 3.14 Kernel

There's support for several new ARM SoC platforms with the upcoming Linux 3.14 kernel...

http://www.phoronix.com/vr.php?view=MTU4MTE

[uid313]

It would be nice to see the ODROID/ and DragonBoard single-board computers supported. They seem pretty nice. Exynos and Snapdragon 800 based.

[schmidtbag]

It would be nice to see the ODROID/ and DragonBoard single-board computers supported. They seem pretty nice. Exynos and Snapdragon 800 based.

Which odroid's aren't supported? I have the U2 and that works fine.

[uid313]

Which odroid's aren't supported? I have the U2 and that works fine.

The U3 I believe.
But I don't think the U2 is supported either, is it?
I think you need a patched kernel for that.
Don't think there is support upstreams in mainline.

[schmidtbag]

The U3 I believe.
But I don't think the U2 is supported either, is it?
I think you need a patched kernel for that.
Don't think there is support upstreams in mainline.

The GPU drivers require some work but the platform as a whole runs great. The CPU is fast enough that depending on what you do, you don't need the GPU. Don't expect to do games or HD videos without the GPU drivers though.

[diriel]

Arm and the mainstream...

Oddly enough, I it would be cool to see the newest 64bit arm chips benched against an old Intel Q6600 ...

[LiquidAcid]

The U3 I believe.
But I don't think the U2 is supported either, is it?
I think you need a patched kernel for that.
Don't think there is support upstreams in mainline.

Correct, there is no upstream support for the ODROID-X2,U2,U3 (Exynos4-based) or the XU (Exynos5-based). And Hardkernel doesn't seem to be interested in upstreaming any of their code.

[schmidtbag]

Correct, there is no upstream support for the ODROID-X2,U2,U3 (Exynos4-based) or the XU (Exynos5-based). And Hardkernel doesn't seem to be interested in upstreaming any of their code.

Unless hardkernel is a division of Samsung, it isn't up to them to upstream their code because their efforts to get the Mali drivers working in linux is based on pre-built closed-source patches. On another note, considering their products are named "ODROID", they clearly don't prioritize linux, which is unfortunate and a little confusing, since android doesn't really take advantage of static devices.

[LiquidAcid]

Unless hardkernel is a division of Samsung, it isn't up to them to upstream their code because their efforts to get the Mali drivers working in linux is based on pre-built closed-source patches.

Wrong. The Mali kernelspace driver is open-source and provided by ARM itself. Also, when I speak about upstreaming code, I don't refer to Mali but:
- providing a DTS (like the ODROID-X has in mainline) to give the user at least a chance to boot the board
- showing a least a remote interest in pushing bugfixes to subsystems (like USB, clocks, sound) into mainline

The Mali kernelspace code will never go upstream, at least not while there is any major userspace application that uses it. And while Lima is this userspace application, it simply isn't "important" enough atm. This would change if the code were to go into mesa, at least from my perspective. But even then, the Mali kernelspace code would need a major redesign so that it gets accepted by the DRM maintainers.

And concerning Samsung. They're doing a lot of upstream work. But just because the kernel supports the corresponding Exynos SoC, doesn't mean you can also boot the board with it.

[schmidtbag]

Wrong. The Mali kernelspace driver is open-source and provided by ARM itself. Also, when I speak about upstreaming code, I don't refer to Mali but:
- providing a DTS (like the ODROID-X has in mainline) to give the user at least a chance to boot the board
- showing a least a remote interest in pushing bugfixes to subsystems (like USB, clocks, sound) into mainline

The Mali kernelspace code will never go upstream, at least not while there is any major userspace application that uses it. And while Lima is this userspace application, it simply isn't "important" enough atm. This would change if the code were to go into mesa, at least from my perspective. But even then, the Mali kernelspace code would need a major redesign so that it gets accepted by the DRM maintainers.

And concerning Samsung. They're doing a lot of upstream work. But just because the kernel supports the corresponding Exynos SoC, doesn't mean you can also boot the board with it.

Hmm I must have been misinformed then. I could have sworn Mali was closed source, unless it used to be at one point. I own another Mali based device that also has GPU driver issues.

I personally have installed linux on 4 different ARM platforms, so I'm well aware of the burdens that come with ARM. What I don't get is why does ARM have the burden of needing custom kernels per platform but x86 doesn't? I realize that MAYBE it would make sense between manufacturers (for example, a Freescale kernel wouldn't work on a Samsung kernel). It also makes sense if one platform requires a different bootloader. But it really doesn't make sense to me how you might not be able to use the ODROID-X kernel on the ODROID-U2 system. Normally, I wouldn't care that much but GPU drivers are so picky and only work on specific kernels, so I might have to use some outdated crappy kernel which is missing drivers and has problems fixed in newer kernels. I also wouldn't care so much if ARM (in the platforms I own) had virtualization. If I'm not going to get useful/working GPU drivers, I would LOVE to just use whatever kernel I want through a VM.