Lion kernel testing on AMD (don't ask help here: use the Help Topic)

[instant idiot]

Hold on, testing soon!

 

anv_32_64 -F -v arch=x86_64 npci=0x3000

Kernel is LP64!!!!1!111!

 

anv_32_64 -F -v arch=i386 -legacy npci=0x3000

One moment please...

Boots straight to desktop. I would be posting from it, but Lnx2Mac's Ethernet driver isn't playing nice for some reason.

[Deltac0]

Hold on, testing soon!

 

anv_32_64 -F -v arch=x86_64 npci=0x3000

Kernel is LP64!!!!1!111!

 

anv_32_64 -F -v arch=i386 -legacy npci=0x3000

One moment please...

Boots straight to desktop. I would be posting from it, but Lnx2Mac's Ethernet driver isn't playing nice for some reason.

 

I think the kernel thinks we are stupid or something... Why can't he understand that we already know that he is LP64...

No idea what causes this, probably the launchd again.

[spakk]

Test with Lion 10-75 with:

 

anv_32_64 -v -f hangs on: Kernel is LP64

anv_32_64 -s hangs on: Kernel is LP64

anv_32_64 -enable_ssse3emu busratio=17 cpus=1 usekernelcache=no -v -f hangs on: Kernel is LP64

 

anv_lion_32 -v -f hangs on: BSD root: disk0s3, major 14, minor 3

jnl: unknown-dev: journal start / end pointers reset! (jnl 0x7792e60; s 0x1e3a00 0x1e3a00)

 

anv_lion_32 -s hangs on: BSD root: disk0s3, major 14, minor 3

jnl: unknown-dev: journal start / end pointers reset! (jnl ...................................)

 

anv_lion_32 -enable_ssse3emu busratio=17 cpus=1 usekernelcache=no -v -f hangs on: BSD root: disk0s3, major 14, minor 3

jnl: unknown-dev: journal start / end pointers reset! (jnl ..............)

 

anv_lion_64 -v -f hangs on: Kernel is LP64

anv_lion_64 -s hangs on: Kernel is LP64

anv_lion_64 -enable_ssse3emu busratio=17 cpus=1 usekernelcache=no -v -f hangs on: Kernel is LP64

 

 

I'm testing now with Lion 10.7.2, follow results.

[spakk]

Test with Lion 10-72 with:

 

anv_32_64 -v -f Kernel Panic /sbin/launchd failed. errno .......:3550

anv_32_64 -s Kernel Panic /sbin/launchd failed. errno .......:3550

anv_32_64 -enable_ssse3emu busratio=17 arch=x86_64 cpus=1 usekernelcache=no -v -f Kernel-Panic: BlackScreen goes blank and loud beep sound!!

anv_lion_32 -v -f Kernel Panic /sbin/launchd failed. errno .......:3550

anv_lion_32 arch=x86_64 -s Kernel Panic /sbin/launchd failed. errno .......:3550

anv_lion_32 -enable_ssse3emu arch=i386 busratio=17 cpus=1 usekernelcache=no -v -f

Kernel Panic /sbin/launchd failed. errno .......:3550

 

anv_lion_64 -v -f Kernel Panic /sbin/launchd failed. errno .......:3550

anv_lion_64 -s Kernel Panic /sbin/launchd failed. errno .......:3550

anv_lion_64 -enable_ssse3emu cpus=1 arch=x86_64 -v -f Kernel-Panic:Screen goes blank many fragments on the screen, similar to pixel errors, a loud beep sound!!!

 

 

 

PS: I think that the basic idea of the kernel must be rebuilt. It is indeed good to know that the kernel has created the Bronzovka for Lion has sufficient service with its 32-bit, but we come as no step further. We're always in the same trap !!!!

[Andy Vandijck]

Gonna check something later.

Apparently it needs to be completely rebuilt for x64...

Will be for tomorrow though...

[Deltac0]

Gonna check something later.

Apparently it needs to be completely rebuilt for x64...

Will be for tomorrow though...

 

What needs to be rebuilt, the kernel or launchd?

Good luck with testing!

[Andy Vandijck]

I think the kernel blocks launchd for some reason...

The other possibility is dyld checks it code signature (believe they added that piece of code since 10.7) so it might need patching...

Hopefully not launchd as it isn't easy to build one...

[spakk]

Here's an other Lion kernel from our friends chriss1111 from Canada, please test this kernel and deposit them here your resultative.

 

BOOT USB MAC OS X LION (AMD-PC) CHRIS1111 with Lion legacy_kernel

 

or only Kernel:

 

chris1111 Lion legacy_kernel

 

 

 

we want to know from you how this test kernel runs with your hardware. Please write a short statement to compare with the last Andy's test kernel.

 

Thanks for your help.

Edited December 25, 2012 by spakk

[Andy Vandijck]

Hold on, testing soon!

 

anv_32_64 -F -v arch=x86_64 npci=0x3000

Kernel is LP64!!!!1!111!

 

anv_32_64 -F -v arch=i386 -legacy npci=0x3000

One moment please...

Boots straight to desktop. I would be posting from it, but Lnx2Mac's Ethernet driver isn't playing nice for some reason.

What if we make the Lion launchd 32-bit only?

ditto -arch i386 /sbin/launchd <tmp_path>/launchd

Would do it (in creating a 32-bit only launchd)...

If it can boot in 32-bit only it may give us a way to do it this way...

note: backup launchd first...

Test with both arch=i386 -force64 and arch=x86_64 (-force64)

[Deltac0]

What if we make the Lion launchd 32-bit only?

ditto -arch i386 /sbin/launchd <tmp_path>/launchd

Would do it (in creating a 32-bit only launchd)...

If it can boot in 32-bit only it may give us a way to do it this way...

note: backup launchd first...

Test with both arch=i386 -force64 and arch=x86_64 (-force64)

 

So the problem is that kernel can't launch 64-bit processes?

Clever solution, even if it doesn't work. I must admit that I admire your wisdom.

 

EDIT: Lion 10.7 myHack usb with AnV kernel: Kernel is LP64.

32-bit only launchd is impossible to make, only x86_64 in the vanilla launchd.

I will try to patch the dyld and try i386 launchd from SL.

[Andy Vandijck]

 

 

So the problem is that kernel can't launch 64-bit processes?

Clever solution, even if it doesn't work. I must admit that I admire your wisdom.

 

EDIT: Lion 10.7 myHack usb with AnV kernel: Kernel is LP64.

32-bit only launchd is impossible to make, only x86_64 in the vanilla launchd.

I will try to patch the dyld and try i386 launchd from SL.

No i386 launchd?

How does it boot with -legacy then?

 

I think we may have something here...

If it doesn't load launchd and works we may be able to get it to go...

[Deltac0]

No i386 launchd?

How does it boot with -legacy then?

 

Good point. Now I'm totally confused... I'll try to look again.

NOTE: This is just myHack USB, it may have something different.

Or it might be my dmg or something.

 

EDIT: I'll install Lion retail to another partition. This should give us i386 launchd.

[Andy Vandijck]

 

 

Good point. Now I'm totally confused... I'll try to look again.

NOTE: This is just myHack USB, it may have something different.

Or it might be my dmg or something.

Please check...

If launchd isn't needed we could mod the kernel...

[Andy Vandijck]

 

 

Good point. Now I'm totally confused... I'll try to look again.

NOTE: This is just myHack USB, it may have something different.

Or it might be my dmg or something.

Please check...

If launchd isn't needed we could mod the kernel...

 

 

 

Good point. Now I'm totally confused... I'll try to look again.

NOTE: This is just myHack USB, it may have something different.

Or it might be my dmg or something.

Please check...

If launchd isn't needed we could mod the kernel...

[Deltac0]

Okay... Now my mind is finally blown.

Mounted the InstallESD.dmg.

 

sh-3.2# cd /Vol*/Mac*
sh-3.2# cd sbin
sh-3.2# file launchd
launchd: Mach-O 64-bit executable x86_64

 

I gotta try this with my other Lion dmg...

[Andy Vandijck]

Lets hope it is the same!

I could then check for AMD cpu and bypass launchd...

[Deltac0]

Lets hope it is the same!

I could then check for AMD cpu and bypass launchd...

 

This is pretty damn weird... But... If you say it is good...

 

sh-3.2# cd "/Volumes/Mac OS X Base System/sbin"
sh-3.2# ls | grep launchd
launchd
sh-3.2# file launchd
launchd: Mach-O 64-bit executable x86_64

 

And I think this is 10.7.5 BaseSystem.dmg.

 

EDIT: I think we may have wrong alert... Asked @ IRC #LegacyKernel if someone could do the file /sbin/launchd command... Yousseflion uses 32-bit Lion (on AMD I think) and he got this:

sh-3.2# sudo file /sbin/launchd
/sbin/launchd: Mach-O universal binary with 2 architectures
/sbin/launchd (for architecture x86_64): Mach-O 64-bit executable x86_64
/sbin/launchd (for architecture i386): Mach-O executable i386

 

 

 

EDIT2: I'll make myHack USB with 10.7.5 now. Launchd is x86_64 only. I'll try to boot it with BronzovkaAnv kernel, i386 and -legacy. If it boots... Then I'm just like WTF.

[spakk]

This is pretty damn weird... But... If you say it is good...

 

sh-3.2# cd "/Volumes/Mac OS X Base System/sbin"
sh-3.2# ls | grep launchd
launchd
sh-3.2# file launchd
launchd: Mach-O 64-bit executable x86_64

 

And I think this is 10.7.5 BaseSystem.dmg.

 

EDIT: I think we may have wrong alert... Asked @ IRC #LegacyKernel if someone could do the file /sbin/launchd command... Yousseflion uses 32-bit Lion (on AMD I think) and he got this:

sh-3.2# sudo file /sbin/launchd
/sbin/launchd: Mach-O universal binary with 2 architectures
/sbin/launchd (for architecture x86_64): Mach-O 64-bit executable x86_64
/sbin/launchd (for architecture i386): Mach-O executable i386

 

 

 

EDIT2: I'll make myHack USB with 10.7.5 now. Launchd is x86_64 only. I'll try to boot it with BronzovkaAnv kernel, i386 and -legacy. If it boots... Then I'm just like WTF.

 

I always get this message, big {censored}

 

bash-3.2# Mach-O executable i386

bash: Mach-O: command not found

bash-3.2#

 

what am I doing wrong?

[Andy Vandijck]

In the meantime I did some more work on ML kernel (kern_exec.c)

I made launchd disable with -nolaunchd boot arg

Also it doesn't panic on launchd anymore.

I also put the cs flags in there to always valid.

It also doesn't check exec permissions anymore...

We will see the result later...

[Deltac0]

I always get this message, big {censored}

 

bash-3.2# Mach-O executable i386

bash: Mach-O: command not found

bash-3.2#

 

what am I doing wrong?

 

Sorry I'm laughing for you... Mach-O isn't a command.

 

sh-3.2# file /sbin/launchd

 

is the only command I've ran, and the

 

launchd: Mach-O 64-bit executable x86_64

 

is the result of it.

 

 

 

 

In the meantime I did some more work on ML kernel (kern_exec.c)

I made launchd disable with -nolaunchd boot arg

Also it doesn't panic on launchd anymore.

I also put the cs flags in there to always valid.

It also doesn't check exec permissions anymore...

We will see the result later...

 

Nice to see ML progress too!

[spakk]

from Lion 10.7.2 launchd i386

launchd orig

 

I hope I did not make s-hit

[Andy Vandijck]

Guess what is in the exec_check_permissions function I removed... return(EACCESS) also known as errno=13:D

[Deltac0]

from Lion 10.7.2 launchd i386

 

 

Nice to see you could do it.

Have you tried to boot with it yet?

 

I looked @ kern_exec.c and couldn't fail to notice this:

 

* Notes: The process that is passed in is the first manufactured
* process on the system, and gets here via bsd_ast() firing
* for the first time. [b]This is done to ensure that bsd_init()
* has run to completion.[/b]

 

It sounds like the load_init_program is just a test to see if bsd_init() has ran?

So it could it just be like a verify process in the bsd_init()?

 

Guess what is in the exec_check_permissions function I removed... return(EACCESS) also known as errno=13:D

 

xD You can't be serious? So... Correct permissions would work?

LOL... I'm really rolling on the floor laughing.... You really are a wizard!

[Andy Vandijck]

I just posted the ML version...

The mods in kern_exec.c can be applied to the Lion diff too...

 

Booting now...

 

-nolaunchd is a no-go...

It just stops after Kernel is LP64

Trying stock (without -nolaunchd)

[Deltac0]

Here is a Lion kernel with few changes to kern_exec.c. Also, it should launchd fsck_hfs instead of launchd... It's just a test.

X86_64 only, no need to even try with i386 -legacy.

 

Download

 

I will post diff later if someone needs it, but basicly it's just AnV lion with launchd -> fsck_hfs and few debug messages.

 

Mm'kay, I'm experimenting with different things in kern_exec.c... I'll post my test kernels on this post, try and tell me if you get any further than Kernel is LP64 or launchd errno.

 

 

 

 

lion-test-0 - Should stuck @ errno 13 or Kernel is LP64. Arch x86_64 only.

http://www.solidfile...m/d/d6b64345d2/

 

lion-test-1 - Some debug messages, should still get stuck...

http://www.solidfile...m/d/9ffd0f459d/

 

lion-test-2 - Okay... Now I'm getting frustrated. Added over 20 debug messages to see where this damn kernel stucks.

http://www.solidfile...m/d/4fbb6a1db4/

 

lion-test-3 - More debugging, now with bsd_init.c

http://www.solidfile...m/d/d5c3b4c098/

 

lion-test-4 - Same as lion-test-3, but with parts of SL code

http://www.solidfile...m/d/209880df2a/

 

lion-test-5 - More debugging... execve()

http://www.solidfile...m/d/2b4f29c452/

 

lion-test-6 - A little more... debugging? I may have a little clue why it stucks on Kernel is LP64...

http://www.solidfile...m/d/4af3e357c8/

 

lion-test-7 - Just to verify what I'm thinking... Something related to check_for_signature()

http://www.solidfile...m/d/6ac85e795b/

 

lion-test-8 - Skips some EACCES parts...

http://www.solidfile...m/d/f05e9bda2e/

 

lion-test-9 - More debug messages, special surprise... (If it can even run to that point...)

http://www.solidfile...m/d/72c8723623/

 

lion-test-10 - This may get past Kernel is LP64...

http://www.solidfile...m/d/415cc6f220/

 

lion-test-11 - More f u c k i n g debugging...

http://www.solidfile...m/d/f683b58def/

 

lion-test-12 - This is weird...

http://www.solidfiles.com/d/5870b0838b/