Thoughts about Mountain Lion with Atom (yes, Atom)

[theconnactic]

1) Is it somehow possible to "tame" (sorry for the awkward metaphor, but it's perhaps adequate) the vanilla Mountain Lion kernel (with boot flags, removal of some kexts etc) to at least boot when running on an atom CPU (which, besides its low performance, is Intel-made, is as new as core2 CPUs, is 64-bit ssse3-enabled and was indeed supported by Apple, even not being anymore)?

 

I won't endeavour this on my almost perfectly working lion netbook, but i do have an older disassembled atom netbook from which i removed the 2'5''HD to use with a Windows machine. I could format this HD and reassembly the netbook, but i won't effort this if the more experienced folks here say that's a waste of time;

 

 

2) Is it somehow possible, even without direct access to the kernels's source code, to take the previous patches made for atom in past kernels and install them over the new one, like any other kext or driver, so Atom be supported?

 

Thank you in advance and sorry if the questions are inept everyway.

[eep357]

Not likely to be happening anytime soon, not that I know how to patch a kernel, but I know it's not that easy, or I'd know how to patch a kernel. Really with all these AMD and atom systems you keep coming up with, maybe trade a couple for an i5 or something and get everything working the way you want, instead of hours months of frustration that will still leave you "almost" having everything working. Then repeat the process again after every OS update. Friendly advice so don't take it the wrong way.

[mtek]

I see no reason why ML should be that different to make it work on a Atom processor than on Lion. My Asus 1201n netbook runs flawlessly Lion 10.7.4 and hopefully will be able to run ML in the future.

[fantomas]

if you have an atom which runs the 64-bit, it should pass on ML. But need a patched kernel, like on Lion, no vanilla kernel be possible since SL 10.6.2 !

[meklort]

1) Is it somehow possible to "tame" (sorry for the awkward metaphor, but it's perhaps adequate) the vanilla Mountain Lion kernel (with boot flags, removal of some kexts etc) to at least boot when running on an atom CPU (which, besides its low performance, is Intel-made, is as new as core2 CPUs, is 64-bit ssse3-enabled and was indeed supported by Apple, even not being anymore)?

As long as you have a 64bit capable cpu, the ML kernel will run on it. You will not, however, get the the gpu working (which is used quite a bit now in os x).

2) Is it somehow possible, even without direct access to the kernels's source code, to take the previous patches made for atom in past kernels and install them over the new one, like any other kext or driver, so Atom be supported?

The binary patch to the kernel is pretty strait forward, plus it's better than the source patch...

 

 

Anyway, the real issue is the gpu, it's not supported. Unless if you are using a something like the nvidia ion gpu (like the 1201n), you'll be out of luck.

[theconnactic]

Thank you, meklort! Very much indeed!

 

 

As long as you have a 64bit capable cpu, the ML kernel will run on it.

 

Does this mean that the vanilla ML kernel wil run on my Atom CPU without any patches? Or do you mean run with some necessary patches, like (hopefully) it would on an AMD 64-bit (and ssse3 capable) processor?

 

 

The binary patch to the kernel is pretty strait forward, plus it's better than the source patch...

 

Then could i apply this patch directly to the ML kernel, as i said before, like it was another driver/kext install? How do i do this, and where do i find the patch binaries?

 

 

 

You will not, however, get the the gpu working (which is used quite a bit now in os x).

 

What if we use older graphics kexts to do the trick? As long as these are 64-bit kexts, won't they (again, hopefully) work in ML?

 

 

Least but not last, could it be possible to build an i386 kernel from ML's source code, once it goes open source?

 

I think the answer to this question, and to the one related to graphics, would be perhaps relevant to "real" Mac users too, since some of them were left out of the new OS because of graphics and firmware architeture related issues, wouldn't?

[meklort]

Any kernel from 10.6.2 newer needs either a source patch or a binary patch. As of 10.7.x 32bit only cpus require a source patch, however 64bit capable cpus can run with only a binary patch.

 

If you want a binary patch, I can make one, but only if it's really going to be useful (like for the 1201N).

 

The only cpu that has kexts that *might* work is the N270 (would work with the GMA950 kext from 10.6.2), however it's a 32bit cpu, so you cannot run ML on it.

 

Yes, you can build a 32bit ML kernel, but you still *need* a 64bit cpu, the N270 cannot be used if you want to use Finder. Any other atom cpu that's 64bit doesn't have a supported gpu (again, unless if it's something like the ion).

 

As far as real macs, Yes, the core 2 duos with GMA X3100 or GMA950 *might* work with a 32bit kernel and the kexts from 10.7.x. The big issue is the graphics drivers (aka, did apple change the interface?) I may have to try this out sometime...

[theconnactic]

First of all, meklort, thank you for the offer! The patch will be useful, as i have access to an ASUS 1201 with nVidia ion. It wil be useful to mtek too, and a lot of others who have that wonderful netbook. Thank you very much!! Can you upload the patch here, as well as instructions for install it?

[buoo]

If you want a binary patch, I can make one, but only if it's really going to be useful (like for the 1201N).

 

 

My Atom D525 has always worked well with this binpatch http://tdev.me/2011/07/mach-kernel-atom-10-8-0-xnu-1504-15-31-mac-os-x-10-6-8/?lang=en in Snow Leopard. Unfortunately I have not been able to make a binpatch for Lion's kernel.

 

It would be great if you share the right binpatch for the ML's kernel, you could make a lot of people happy

 

Thanks

[edmondwck]

Yes, atom kernel for ML please.

My CPU is celeron 420 and it require a patched kernel.

Thanks!

[meklort]

Alright, it'll take me a few days to get everything ready.

[theconnactic]

Thank you, meklort!

 

I already installed ML on a partition of an ASUS 1201n. Did it manually, running all the packages inside essentials from my lion system but targeting the partition reserved for ML, and adding the other files (libraries etc), then i ran the latest build of chameleon, and deleted mach_kernel.

 

Waiting for your patched one!

[theconnactic]

Hi everyone!

 

Well, it looks like the source code of Mountain Lion's kernel was released earlier than expected: http://opensource.ap...e/mac-os-x-108/

 

And someone else managed to build an Atom kernel from the source using meklort's patched 10.7.1: http://www.osx86.net...ernel_atom.html

 

I dowloaded this kernel and i'm here to share my experience: i simply put it on my prepared disk (see last post) and got it to run. Instant reboot. Since it was kindda clumsy install (again, see my last post), i keep calm and carry on another installation, this time following this instructions: http://www.insanelym...howtopic=280756 - no luck again. Trying booting from the USB stick, instant reboot. Tried the metapackage method, as good old Lion. Funny thing happened: when i ran the mpkg, the installation failed because (sic) "no software was found", like the mpkg wasn't managing to point correctly to each pkg. Blaming some possible mistake of mine, i simply done the whole thing from the start again. Same result. A bit frustrated, i opened the mpkg using pacifist. One hour later, everything finished (not without a lot of error reports and so), i applied chameleon, reboot and got it! A brand new instant reboot! ACPI error, unfortunately.

 

Therefore, maybe i'm missing something. Maybe the problem is that i tried to install it in an external drive (i have one with lion which works like a charm). But, anyway, i'm still waiting for meklort's new binary.

[mtek]

Alright, it'll take me a few days to get everything ready.

 

Thank you very much for your hard work.

My 1201n waits anxiously to run Mountain Lion with you patched kernel!

[mtek]

Meanwhile I ventured using the patched ML kernel posted here to try to make my Asus 1201n (Atom N330 dual-core 64bit) install.

 

The patched kernel worked fine with the ML installer and I was able to make the full installation on my 8GB pen USB partition without any issues.

 

After that I copied the same patched kernel to the root of the final ML partition and it booted correctly up to a point where it stuck/stopped.

This is my verbose screen output - any ideas/suggestions on what might have gone wrong?

[theconnactic]

I think you're almost there. Did i tel you that i experienced a lot of checksum errors when i tried to install it with pacifist? (with installer it was a no-go). What version did you use, retail or GM? Maybe my dmg is somehow corrupted.

 

As for your question: boot with npci=0x2000 or npci=0x3000. I think is simple like that.

[mtek]

I think you're almost there. Did i tel you that i experienced a lot of checksum errors when i tried to install it with pacifist? (with installer it was a no-go). What version did you use, retail or GM? Maybe my dmg is somehow corrupted.

 

As for your question: boot with npci=0x2000 or npci=0x3000. I think is simple like that.

 

Hi,

 

I'm booting with npci=0x2000 already but for some weird reason had to boot ML (Retail) for the first time with -X flag. After that I rebooted and everything was just fine.

 

So I've got my 1201n running with ML right now, but with some kext issues. Checkout this post: http://www.insanelymac.com/forum/index.php?showtopic=264740&st=540

 

Good luck!

[meklort]

Well, unfortunately I'm currently on ML, I need to downgrade to 10.6 (preferably) to work on this properly. I should have everything setup today, after that it should be by next work (I'll probably end up putting a kernel patcher module in forge instead of a patched kernel, one the module is create I can always generate the patched kernel from it if you need that).

[theconnactic]

Glad to know, meklort! Glad to know also that mtek got it working with the osx.net patched kernel (which didn't work for me).

[mtek]

Glad to know, meklort! Glad to know also that mtek got it working with the osx.net patched kernel (which didn't work for me).

 

Yep this patched kernel worked fine for my dual-core native 64bit Atom N330 processor. No KP issues at all so far in ML either.

 

[buoo]

The patched kernel works good in my case, but I've a problem with my ION2.

It has always worked well in SL and Lion with GraphicsEnabler or with a DSDT injection, but now in ML I get black screen. Deleting NV50HAL.kext and using the -x flag I can boot but without QE/CI.

 

Any idea?

[splonk]

Try NVEnabler

[meklort]

If anyone here hangs out in the my hack irc, message me there if you'd like to do a bit of testing.

[buoo]

Hi meklort!

 

Glad to help you

 

 

 

 

Unfortunately I have to use the NullCPU otherwise I get kernel panics differently from SL and L.

In Snow Leopard using the Chameleon's functions (Generate and Pstates Cstates) I'm able to load the AppleIntelCPUPowerManagement, so I'm not forced to use the SleepEnabler.

 

I hope it can be fixed

[mtek]

I also use nullcpu next on my Asus 1201n install both in Lion and Mountain Lion - it seems to run quite happily with it...