Jump to content

Mavericks kernel testing on AMD (formerly Mountain Lion kernel testing on AMD)


theconnactic
 Share

6,414 posts in this topic

Recommended Posts

WHAT THE LITTERAL ACTUAL COW HECK?!?!?!

 

I have found the holy grail. AMD K10 has registers that allow redefining the salient information returned by a CPUID instruction. Has this been in front of us all along?

let's quote the freaking docs,

 

The following code sample shows how a VMM (or OS) can hide reporting of SSE4A instructions

on a quad-core AMD OpteronTM processor. { ..... C ..... }

AND then the actual "AuthenticAMD" string too!

 

 

ecstatic ..

 

-- Sinetek

  • Like 1
Link to comment
Share on other sites

P.S.: Sinetek, here's the patch of the sinactic kernel to be applied to the 10.7.5 XNU, just in case you want to try it yourself. I used Xcode 3.2.6 by the way.

 

sinactic.patch.zip

 

WHAT THE LITTERAL ACTUAL COW HECK?!?!?!

 

I have found the holy grail. AMD K10 has registers that allow redefining the salient information returned by a CPUID instruction. Has this been in front of us all along?

let's quote the freaking docs,

 

 

AND then the actual "AuthenticAMD" string too!

 

 

ecstatic ..

 

-- Sinetek

 

 

Congratulations for finding this, Sinetek!

 

What the extent of the benefits could it give us? What's the practical application of this to our little work here? Great expectations! :D

Link to comment
Share on other sites

Hi, folks!

 

As i promised, i recompiled the sinactic 10.7.5 Lion kernel with the ssse3 emulator using the flags recommended by Sinetek. I added them at osmkf/conf/Makefilei386.

 

Let's see if the ssse3 emulator starts without any runtime errors! Remember, this is for non-Bulldozer/Piledriver only. If possible, put vids or pics of your verbose boots. Good luck!

 

sinactic.zip

 

 

The same KP still happened, either with: or without: -legacy

 

ml_04a.gif

 

 

ml_04b.gif

Link to comment
Share on other sites

Hi theconatic,

Hi Sinetek

Hi @ll

 

As the first time I've installed Lion 10.7.4 on my AMD Phenom II X6 1100T, I was able to boot with help of cparm bootloader under the application modules stable my system. cparm and AndyVandijck have developed this boot loader for AMD CPUs. Please take a look at the site if that's helpful.

 

--->> http://osx86.co/f36/...-made-by-t7178/

 

The download link from cparm is dead, here are the standard modules and various other modules from my collection

--->> www.workupload.com/file/djUizBa

  • Like 1
Link to comment
Share on other sites

Hi, spakk!

 

By the way, for everyone's information, Andy told me a couple of days ago he also would be working on the 12.x kernel for AMD. He said he would keep me posted about his progress, so that's good news.

 

Sinetek, i still don't understand which good your discoveries would effectively do. Can you share your thoughts? :)

  • Like 1
Link to comment
Share on other sites

I can not test on Lion 10.7.5, I have only 10.7.4 on my AMD

 

I've tested the latest kernel on 10.7.4, I get the same error message as mac_carol (with the same boot flag as mac_carol). and if I boot with boot flag: -v -f

 

then comes this strange error message:

 

PCI Parity Error!

Press F1 To Continue F2 To reboot .....

PCI Parity Error!

Press F1 To Continue F2 To reboot .....

 

the F1 and F2 keys do not work, then I've switch my PC manual off and on. :worried_anim:

 

Edit:

My speaker makes a duration-sound-error during the kernel panic.

Link to comment
Share on other sites

Hi, spakk!

 

By the way, for everyone's information, Andy told me a couple of days ago he also would be working on the 12.x kernel for AMD. He said he would keep me posted about his progress, so that's good news.

 

Sinetek, i still don't understand which good your discoveries would effectively do. Can you share your thoughts? :)

 

I think he means this http://developer.amd.com/wordpress/media/2012/10/43781-3.00-PUB_Live-Virtual-Machine-Migration-on-AMD-processors.pdf. I had forgotten about the CPUID overrides that can be made by writing to MSR. Basically mask CPU features and the actually processor string from AMD to Intel. Intel allows something similar by allowing a CPUID fault to be trapped and the data replaced.

Link to comment
Share on other sites

Hi, spakk!

 

By the way, for everyone's information, Andy told me a couple of days ago he also would be working on the 12.x kernel for AMD. He said he would keep me posted about his progress, so that's good news.

 

Sinetek, i still don't understand which good your discoveries would effectively do. Can you share your thoughts? :)

 

theconatic,many thanks for this nice information and many thank Andy for your willingness to help these two young programmers

 

viannen_01.gif

 

I think he means this http://developer.amd...processors.pdf. I had forgotten about the CPUID overrides that can be made by writing to MSR. Basically mask CPU features and the actually processor string from AMD to Intel. Intel allows something similar by allowing a CPUID fault to be trapped and the data replaced.

 

the link on the AMD website is dead, here is the PDF file for download

http://www.workupload.com/file/66EBZj2

Link to comment
Share on other sites

theconatic + sinetec, also I can test on my AMD Phenom II X6, although I have only Lion 10.7.4

I'm just waiting for your new modified kernel

Link to comment
Share on other sites

AMD Athlon II X4 645

 

Also the system speaker kept beeping super fast

 

Im running 10.7.3, I cant use 10.7.4 or Ethernet wont work, I havent tried 10.7.5 as Im not sure if Ethernet will work or not.

 

Edit: Im trying to attach the picture but its not working.

post-3409-0-75385300-1355018108_thumb.jpg

Link to comment
Share on other sites

AMD Athlon II X4 645

 

Also the system speaker kept beeping super fast

 

Im running 10.7.3, I cant use 10.7.4 or Ethernet wont work, I havent tried 10.7.5 as Im not sure if Ethernet will work or not.

 

Edit: Im trying to attach the picture but its not working.

 

The kernel needs to be modified, we get the same error as you.

 

we are all waiting for the new modified kernel to test

 

please, test them with boot flag; -v-f . I want to know if you get the same error.

Link to comment
Share on other sites

The ssse3emulator isn't ready yet, so that's why all you are getting a =-16 error from the ssse3 emulator. I'm working on it, and i know Sinetek is also working on it.

 

The way i see it, our roadmap now consists of three main points:

 

1) Proper implementation of the ssse3 emulator for arch=i386 with 64-bit mode (-force64) so we finally have a trully stable 10.7.x kernel for all AMD and legacy CPUs;

 

2) Proper implementation of the ssse3 emulator in full 64-bit mode (arch=x86_64) so we can run Mountain Lion (or Lion 64-bit) in legacy AMD and Intel CPUs;

 

3) Solving the instant reboot issue with Bulldozer and Plidriver CPUs when booting 64-bit (arch=x86_64). If it wasn't for it, the first connactic patched kernels for 10.8.0 would have been sufficient for them to run either Lion 64-bit or Mountain Lion with the patched kernels. About it, i propose an experiment to the FX users (Gilles, if you're reading, you too, even if you post the results in another site): download please the latest iteration of chameleon (2.2, rev 2xxxx:Chameleon-2.2svn-r2117.pkg.zip ), one which contains the kernel patcher module and try to boot with:

 

a] The connactic kernel for 10.8.0: connatic.zip

 

b] The connactic kernel for 10.7.5, booting with archx86_64: connactic64 for 10.7.5.zip

 

As usual, post pics and vids of the results. Good luck!

  • Like 1
Link to comment
Share on other sites

The ssse3emulator isn't ready yet, so that's why all you are getting a =-16 error from the ssse3 emulator. I'm working on it, and i know Sinetek is also working on it.

 

The way i see it, our roadmap now consists of three main points:

 

1) Proper implementation of the ssse3 emulator for arch=i386 with 64-bit mode (-force64) so we finally have a trully stable 10.7.x kernel for all AMD and legacy CPUs;

 

2) Proper implementation of the ssse3 emulator in full 64-bit mode (arch=x86_64) so we can run Mountain Lion (or Lion 64-bit) in legacy AMD and Intel CPUs;

 

3) Solving the instant reboot issue with Bulldozer and Plidriver CPUs when booting 64-bit (arch=x86_64). If it wasn't for it, the first connactic patched kernels for 10.8.0 would have been sufficient for them to run either Lion 64-bit or Mountain Lion with the patched kernels. About it, i propose an experiment to the FX users (Gilles, if you're reading, you too, even if you post the results in another site): download please the latest iteration of chameleon (2.2, rev 2xxxx:Chameleon-2.2svn-r2117.pkg.zip ), one which contains the kernel patcher module and try to boot with:

 

a] The connactic kernel for 10.8.0: connatic.zip

 

b] The connactic kernel for 10.7.5, booting with archx86_64: connactic64 for 10.7.5.zip

 

As usual, post pics and vids of the results. Good luck!

 

Tried your Lion kernel on my 10.7.4 install and got stuck at Waiting For Root Device...

 

Will post pic later.

Link to comment
Share on other sites

The ssse3emulator isn't ready yet, so that's why all you are getting a =-16 error from the ssse3 emulator. I'm working on it, and i know Sinetek is also working on it.

 

The way i see it, our roadmap now consists of three main points:

 

1) Proper implementation of the ssse3 emulator for arch=i386 with 64-bit mode (-force64) so we finally have a trully stable 10.7.x kernel for all AMD and legacy CPUs;

 

2) Proper implementation of the ssse3 emulator in full 64-bit mode (arch=x86_64) so we can run Mountain Lion (or Lion 64-bit) in legacy AMD and Intel CPUs;

 

3) Solving the instant reboot issue with Bulldozer and Plidriver CPUs when booting 64-bit (arch=x86_64). If it wasn't for it, the first connactic patched kernels for 10.8.0 would have been sufficient for them to run either Lion 64-bit or Mountain Lion with the patched kernels. About it, i propose an experiment to the FX users (Gilles, if you're reading, you too, even if you post the results in another site): download please the latest iteration of chameleon (2.2, rev 2xxxx:Chameleon-2.2svn-r2117.pkg.zip ), one which contains the kernel patcher module and try to boot with:

 

a] The connactic kernel for 10.8.0: connatic.zip

 

b] The connactic kernel for 10.7.5, booting with archx86_64: connactic64 for 10.7.5.zip

 

As usual, post pics and vids of the results. Good luck!

 

I completely agree with this except for one thing - is 1) really necessary? If there's anyone who will care about running Lion on AMD if/when we've perfected Mountain Lion on AMD, I don't know who it is. Just saying. Great job everyone! :)

Link to comment
Share on other sites

I completely agree with this except for one thing - is 1) really necessary? If there's anyone who will care about running Lion on AMD if/when we've perfected Mountain Lion on AMD, I don't know who it is. Just saying. Great job everyone! :)

 

It will be helpful with the step 2 ;) And i386 is easier than x86_64, I think.

Link to comment
Share on other sites

a] The connactic kernel for 10.8.0: connatic.zip

 

sorry mate, but are you sure that kernel is your latest build?

since its build date was Oct 24, 2012

 

ml_03.gif

 

 

here are the tests:

 

1. flags: arch=x86_64 (without: fips_mode=0)

 

ml_01.gif

 

 

 

 

2. flags: arch=x86_64 fips_mode=0

 

ml_02.gif

 

 

 

 

 

 

b] The connactic kernel for 10.7.5, booting with archx86_64: connactic64 for 10.7.5.zip

 

As usual, post pics and vids of the results. Good luck!

 

 

1. flags: arch=i386 -legacy

result: instant reboot

 

 

2. flags: arch=x86_64

 

lion_01.gif

 

 

 

 

 

 

FYI, both Lion 10.7.5 & ML 10.8.0 were running within Chameleon v2.2-r2117 on AMD P-II x4 965

 

 

chameleon.gif

Link to comment
Share on other sites

sorry mate, but are you sure that kernel is your latest build?

since its build date was Oct 24, 2012

 

ml_03.gif

 

 

here are the tests:

 

1. flags: arch=x86_64 (without: fips_mode=0)

 

ml_01.gif

 

 

 

 

2. flags: arch=x86_64 fips_mode=0

 

ml_02.gif

 

 

 

 

 

 

 

 

 

1. flags: arch=i386 -legacy

result: instant reboot

 

 

2. flags: arch=x86_64

 

lion_01.gif

 

 

 

 

 

 

FYI, both Lion 10.7.5 & ML 10.8.0 were running within Chameleon v2.2-r2117 on AMD P-II x4 965

 

 

chameleon.gif

 

Wait, what? Instant reboot with arch=i386 and -legacy? :confused:

 

That's great about arch=x86_64. Though I don't know what that could mean, I do know that its not an instant reboot or even a kernel panic. :D

Link to comment
Share on other sites

Wait, what? Instant reboot with arch=i386 and -legacy? :confused:

 

 

I think the kernel was built for x64 PCs, so it won't work with x86 ones

but I don't know if it also can run on old AMD, or it's only for the Buldozer AMD

 

 

 

 

That's great about arch=x86_64. Though I don't know what that could mean, I do know that its not an instant reboot or even a kernel panic. :D

 

yup! maybe you're right ;)

 

the kernel was able to pass the BSD-root detection (no: waiting for root device error)

but then it stuck at Kernel is LP64 line

  • Like 1
Link to comment
Share on other sites

 Share

×
×
  • Create New...