Apple's XNU From iOS Ported To New Platforms

[staalmannen]

XNU already runs on AArch64: iOS7 on iPhone 5s.

I think the "interesting" aspect of this work might be getting ARM-based "hackintosh phones" (XNU-kernels running non-apple ARM hardware, hosting a modified iOS)

Personally I do not see why one would aim for that, but if someone wants to make custom ROMs (and avoid Apple lawyers) I wish them good luck.

[schmidtbag]

XNU already runs on AArch64: iOS7 on iPhone 5s.

XNU also runs on 32 bit ARM, so what's your point? That's like saying "Mac already runs on x86-64" but that doesn't mean it will work on just any x86 based computer you want.

[Awesomeness]

A Phoronix reader has pointed out the recent work done by the community on porting Apple's XNU (iPhone / OS X) kernel to new platforms, namely ARMv7-A and hopefully AArch64 and ARMv6/v5 platforms...

XNU already runs on AArch64: iOS7 on iPhone 5s.

[dh04000]

While I personally have no need for XNU, I think it's pretty selfish and closed-minded to say there is "practically zero need" for it.

I recognize that. That's why I said they are free to use their free time to do what they want.

[schmidtbag]

I'm not trying to be a needless hater (we have enough of those here), but noooooo....... There is practically zero need for the XNU of iOS as a platform. Even OpenDarwin, a full desktop, has never gotten the development to take off. Why do people think a OS that is capable of an even smaller hardware set will......

But, they are 100% free to spend their free time as they feel. So, I guess we'll just have to wait and see what this turns into.

That's exactly it - with less hardware, it should be easier to port it. ARM platforms are so much easier to port something because:
* They're RISC
* Nearly all of them don't have or need an EFI
* You're pretty much limited to just USB devices
iOS in particular is easy to port because
* It has no USB ports
* It works on ARMv7 and ARMv8
* It isn't a very demanding system
* It's presumably very particular about the built-in hardware, so it shouldn't be hard for people to buy some of these parts.
The greatest challenge I see in porting iOS is getting video drivers to work, and some of the minor details like GPS or SIM cards. The fact that iOS is so locked down and simplified may limit the platforms it can be ported to, but it should make porting to those platforms much easier.

While I personally have no need for XNU, I think it's pretty selfish and closed-minded to say there is "practically zero need" for it.

[dh04000]

I'm not trying to be a needless hater (we have enough of those here), but noooooo....... There is practically zero need for the XNU of iOS as a platform. Even OpenDarwin, a full desktop, has never gotten the development to take off. Why do people think a OS that is capable of an even smaller hardware set will......

But, they are 100% free to spend their free time as they feel. So, I guess we'll just have to wait and see what this turns into.

[JX8p]

Nice work! How much does XNU actually differ from Mach? Something that might be interesting to get working for XNU would be DDELinux together with DDEKit-XNU. DDE/DDEKit has already been ported to the Hurd Mach kernel, which means that Linux drivers (device drivers, file systems ...) can be built without source code modifications for a foreign kernel. The cool thing if DDEKit was ported to XNU would be that Macs (and Hackintoshes) could get all the hardware and file system support of Linux... which would be cool. I guess they are already using userspace drivers (since Mach is a microkernel design).

For an overview of the DDEKit technology see the following links:

http://www.gnu.org/software/hurd/user/zhengda.html

http://os.inf.tu-dresden.de/ddekit/

http://os.inf.tu-dresden.de/pipermail/l4-hackers/2009/004291.html

http://genode.org/documentation/api/dde_kit_index

A similar principle is the NetBSD anykernel:

https://github.com/anttikantee/buildrump.sh

https://en.wikipedia.org/wiki/Anykernel

OS X's XNU includes a large chunk of FreeBSD's kernel, running in kernel-space and utilising Mach for threading, memory management, IPC, and so forth. However the Mach side is still quite close to Mach 3.0 from CMU.

[staalmannen]

Cool... the more the merrier (if drivers can be portable)

Nice work! How much does XNU actually differ from Mach? Something that might be interesting to get working for XNU would be DDELinux together with DDEKit-XNU. DDE/DDEKit has already been ported to the Hurd Mach kernel, which means that Linux drivers (device drivers, file systems ...) can be built without source code modifications for a foreign kernel. The cool thing if DDEKit was ported to XNU would be that Macs (and Hackintoshes) could get all the hardware and file system support of Linux... which would be cool. I guess they are already using userspace drivers (since Mach is a microkernel design).

For an overview of the DDEKit technology see the following links:

http://www.gnu.org/software/hurd/user/zhengda.html

http://os.inf.tu-dresden.de/ddekit/

http://os.inf.tu-dresden.de/pipermail/l4-hackers/2009/004291.html

http://genode.org/documentation/api/dde_kit_index

A similar principle is the NetBSD anykernel:

https://github.com/anttikantee/buildrump.sh

https://en.wikipedia.org/wiki/Anykernel

[BSDude]

That's what we needed all along, a new kernel, on ARM, from Apple.

Hey! It's an open-source implementation *sarcasm*

[mark45]

That's what we needed all along, a new kernel, on ARM, from Apple.