Jump to content

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

theconnactic

By theconnactic,
in OS X 10.9 (Mavericks)

 Share
6,414 posts in this topic

Recommended Posts

wjywbs
wjywbs
Posted
Posted

I will give some updates. I added trace features in sinetek's kernel opemu.c to print what exactly sse instructions are used. The only two sse instructions that mac system 10.9.0 used in boot phase are palignr and pshufb. They are used a great amount of times in both kernel space and user space. rdmsr unknown location was caught once when booting. After mac os boots, I didn't see any other sse instructions was used (caught by opemu.c). I tried to use chrome, vlc playback, Geek3D GPU Test and basic softwares (dock, finder, etc), and they all didn't use sse3+ features. I couldn't use iTunes to play movies, but iTunes didn't use sse3+ either. I didn't track the usage of syscall or sysexit yet.

 

So my graphics card driver issues (white lines and squares on screen, hd 6850) shouldn't relate to sse errors. I don't know whether palignr and pshufb code have any bugs yet. And I disabled logging of palignr and pshufb because there're too many. But they should be fine because the system calls them so many times and boots. To make sure the other sse3+ instructions are trapped, I put one ssse3 assembly instruction in a simple c code, compiled and ran. I saw the instruction was logged by the kernel.

 

@sinetek, @Andy Please make sure there's no bug in palignr and pshufb.

 

I found a very interesting behavior: when I have intensive IO to hard disk, my screen starts to have white lines and squares. I believe AppleVIAATA or AppleATIATA driver has bugs that cause GPU to hang (I tried both). They are 9 years old on Apple's site. Can someone try this: Find a big zip file and unzip it in terminal. In the mean time, run GPU benchmark or open chrome browser, and check whether there're white lines on screen. I will start to look at AppleVIAATA driver. If someone can develop an up-to-date AppleVIAATA driver, that will be much helpful.

 

So the GPU video corruption bug could be AppleVIAATA problem (need comfimation).

Link to comment
Share on other sites
spakk
spakk
Posted
Posted (edited)

I will give some updates. I added trace features in sinetek's kernel opemu.c to print what exactly sse instructions are used. The only two sse instructions that mac system 10.9.0 used in boot phase are palignr and pshufb. They are used a great amount of times in both kernel space and user space. rdmsr unknown location was caught once when booting. After mac os boots, I didn't see any other sse instructions was used (caught by opemu.c). I tried to use chrome, vlc playback, Geek3D GPU Test and basic softwares (dock, finder, etc), and they all didn't use sse3+ features. I couldn't use iTunes to play movies, but iTunes didn't use sse3+ either. I didn't track the usage of syscall or sysexit yet.

 

So my graphics card driver issues (white lines and squares on screen, hd 6850) shouldn't relate to sse errors. I don't know whether palignr and pshufb code have any bugs yet. And I disabled logging of palignr and pshufb because there're too many. But they should be fine because the system calls them so many times and boots. To make sure the other sse3+ instructions are trapped, I put one ssse3 assembly instruction in a simple c code, compiled and ran. I saw the instruction was logged by the kernel.

 

@sinetek, @Andy Please make sure there's no bug in palignr and pshufb.

 

I found a very interesting behavior: when I have intensive IO to hard disk, my screen starts to have white lines and squares. I believe AppleVIAATA or AppleATIATA driver has bugs that cause GPU to hang (I tried both). They are 9 years old on Apple's site. Can someone try this: Find a big zip file and unzip it in terminal. In the mean time, run GPU benchmark or open chrome browser, and check whether there're white lines on screen. I will start to look at AppleVIAATA driver. If someone can develop an up-to-date AppleVIAATA driver, that will be much helpful.

 

So the GPU video corruption bug could be AppleVIAATA problem (need comfimation).

 

are you using a current version of ANV or an older version? this is my consider the last revision that Andy has created for Mountain Lion. from 2012/2013

http://www.insanelymac.com/forum/files/file/66-superviaata-103-anv-for-mountain-lion/

Edited by spakk
Link to comment
Share on other sites
wjywbs
wjywbs
Posted
Posted

are you using a current version of ANV or an older version? this is my consider the last revision that Andy has created for Mountain Lion. from 2012/2013

http://www.insanelymac.com/forum/files/file/66-superviaata-103-anv-for-mountain-lion/

I added log sse feature to the latest sinetek's github version. I tried sinetek and latest Савва's kernel and they both had the graphics problem. The chipset is ATI SB450 without AHCI.

 

Sorry, spakk, I didn't try the one in your link. I'll try and report it here.

 

I tried Andy's SuperVIAATA just now and the problem persisted. The screen freeze when there's a little more hard disk IO. Maybe it is caused at the kernel side in iokit.

Link to comment
Share on other sites
spakk
spakk
Posted
Posted

copy AMD6000Controller.kext to your desktop
open it with the right mouse button: AMD6000Controller.kext/Contents/Info.plist open the plist with the editor and edit the "IOPCIMatch" wear here your VEN ID and DEV ID , then save the file, then install the driver using by Kext Wizard to S / L / E
then restart your system

Link to comment
Share on other sites
spakk
spakk
Posted
Posted

Hi spakk, Actually there already has my graphics card ID in the plist, and I can see the translucent menu bar. I'm sure the driver is loaded.

then it might actually be the kernel I do not know unfortunately. Our programmers do not register to take position.

Link to comment
Share on other sites
spakk
spakk
Posted
Posted (edited)

please be patient a little while .......

 

I'm going to install the update on 13C48, if it should behave the same as the previous update then it would be wonderful

 

Soory, unfortunately no new results. the first update was successful. and the last update I got the  OPEMU error message. I had to the changes the system drivers in S / L / E  to get into the system  :unsure:

Edited by spakk
Link to comment
Share on other sites
NycoFR
NycoFR
Posted
Posted

Hello,

I use a pentium D 940 processor (which do not have ssse3 bit) on a ASUS P5LD2-MQ motherboard. I use the kernel linked in the first post, but i get this :

07/02/14 17:00:18,000 kernel[0]: SMC::smcReadKeyAction ERROR TM7P kSMCBadArgumentError(0x89) fKeyHashTable=0x0xffffff800ae46a00
 
and this :
07/02/14 17:12:59,286 com.apple.kextcache[15597]: kxld[com.apple.driver.AppleIntelCPUPowerManagementClient]: In interface com.apple.kpi.private of __kernel__, couldn't find symbol _acpi_idle_kernel
 
flooding my logs...
 
It started when I updated from 10.8.2 to 10.8.5.
 
Have I to change my kernel ? Or update to mavericks ?

 

Sorry if I am in the wrong thread,

Nyco

Link to comment
Share on other sites
spakk
spakk
Posted
Posted

Hello,

I use a pentium D 940 processor (which do not have ssse3 bit) on a ASUS P5LD2-MQ motherboard. I use the kernel linked in the first post, but i get this :

07/02/14 17:00:18,000 kernel[0]: SMC::smcReadKeyAction ERROR TM7P kSMCBadArgumentError(0x89) fKeyHashTable=0x0xffffff800ae46a00
 
and this :
07/02/14 17:12:59,286 com.apple.kextcache[15597]: kxld[com.apple.driver.AppleIntelCPUPowerManagementClient]: In interface com.apple.kpi.private of __kernel__, couldn't find symbol _acpi_idle_kernel
 
flooding my logs...
 
It started when I updated from 10.8.2 to 10.8.5.
 
Have I to change my kernel ? Or update to mavericks ?

 

Sorry if I am in the wrong thread,

Nyco

you can try to remove the AppleIntelCPUPowerManagement from S / L / Extensions and install Nullpowermanagement.kext, or install a patched ACPM

 

and a new FakeSMC with plugins

 

good luck!

Edit:

 

what kernel do you use currently?

for Mountain lion 10.8.5 use this last kernel from bronzovka http://www.insanelymac.com/forum/files/file/179-amd-108x/

Link to comment
Share on other sites
NycoFR
NycoFR
Posted
Posted

Currently I use the one link in the first post of this thread. I'll try, thanks.

For FakeSMC, I use the only one which gives me the correct temps of cpu & others captors.

I forgot to say that sleep doesn't work for me..

Link to comment
Share on other sites
spakk
spakk
Posted
Posted

Currently I use the one link in the first post of this thread. I'll try, thanks.

For FakeSMC, I use the only one which gives me the correct temps of cpu & others captors.

I forgot to say that sleep doesn't work for me..

here

install in the Extra / Extensions or in S / L / E with kext wizard, Repair permissions then restart your system

Link to comment
Share on other sites
spakk
spakk
Posted
Posted

@spakk.. no problem here with build 13C53 .. but unfortunately still get OPEMU error without rollback kext from 10.9.1..

ok , pico joe, many thanks for your response

Link to comment
Share on other sites
NycoFR
NycoFR
Posted
Posted

I tried rc5 kernel and got stuck at : IOATAcontroler device blocking bus.

Moreover, I'd already booted successfully with this kernel before, and there was a slight bug (I think with RTC) : my computer's time elapsed faster than real time : 15 real sec for about 1 min on my pc... I'm doing a clean install (from Iatkos ML), then updating to 10.8.5 ; then see for kext.

Link to comment
Share on other sites
syscl
syscl
Posted
Posted

Two issues:

 

1.incorrect cpu information:

 

L3 cache : 8KB 

HT Bus    : 800MHz

 

2.The animation of lots of applications run in an abnormal speed(so fast).In which case, when playing music, the sound goes discontinuously.

 

Details:

replace kexts: System.kext

installed kexts : voodoohda.kext(2.8.4)  

 

kern.version: Darwin Kernel Version 13.0.0: Mo  6 Jan 2014 02:09:31 CET; root:xnu-sinetek/BUILD/obj/RELEASE_X86_64

2014-02-09 21:34:14 -0800 kernel[0]: .

2014-02-09 21:34:14 -0800 kernel[0]: Error unserializing info dictionary for device tree entry Driver-20829000: OSUnserializeXML: syntax error near line 1

2014-02-09 21:34:14 -0800 kernel[0]: .

 

hw.l1icachesize = 65536

hw.l1dcachesize = 65536

hw.l2settings = 1

hw.l2cachesize = 524288

hw.l3settings = 1

hw.l3cachesize = 4096

hw.logicalcpu: 2

hw.logicalcpu_max: 2

hw.l1icachesize: 65536

hw.l1dcachesize: 65536

hw.l2cachesize: 524288

hw.l3cachesize: 4096

machdep.pmap.hashwalks: 443683

 

 

Link to comment
Share on other sites
wjywbs
wjywbs
Posted
Posted

Hi spakk, it shouldn't matter. This file is at BUILD/obj/SETUP/config/lexer.yy.c and is generated at build time from SETUP/config/lexer.l  It says the function input is not used and it's just a warning. You should still get a mach_kernel.

 

The OPEMU UD2 message in your kernel should be caused by incompatible drivers, not sse errors. Try to panic on these OPEMU message (post #5301 as reference) and see which kext or part cause that message. ud2 is not a sse instruction. The kernel traps all failed instructions and passes to OPEMU (including sse instructions).

Link to comment
Share on other sites
spakk
spakk
Posted
Posted

Hi spakk, it shouldn't matter. This file is at BUILD/obj/SETUP/config/lexer.yy.c and is generated at build time from SETUP/config/lexer.l  It says the function input is not used and it's just a warning. You should still get a mach_kernel.

 

The OPEMU UD2 message in your kernel should be caused by incompatible drivers, not sse errors. Try to panic on these OPEMU message (post #5301 as reference) and see which kext or part cause that message. ud2 is not a sse instruction. The kernel traps all failed instructions and passes to OPEMU (including sse instructions).

sorry, I was too focused on the build of the kernel and failed to delete this entry. But thank you wjywbs for your information

I do not know where to start, since I'm not a programmer, there is much to read on Ud2 bug fix in the field linux.

Link to comment
Share on other sites
spakk
spakk
Posted
Posted (edited)

We have a massive problem with OPCode since the Beta version from 10.9.1 to 10.9.2.
Although we have found a way around to boot without get OPEMU: Ud2 Error message (Replace the System.kext and sandbox.kext from 10.9), but we do not know what lasting impact this will have on the stability of the system?

As we all know, sinetek and Andy has worked mainly on the OPCODE for AMD Kernel, unfortunately, we have now come to a point where the further development has come to a standstill. The reasons are not known to me?? :(

to all programmers in the situation are to experiment with Sinetek's OPEMU, can take as base around the xnu-sinetek  to see construction of the opcode.

Please give your suggestions for further development and testing proposals for a bugfix :)

 

 

I understand that so right? that our Linux friends have created a bug fix? http://valgrind.org/docs/manual/dist.news.html
if that is true,  is there possibility to use this bug fix for OS X ?

Edited by spakk
Link to comment
Share on other sites
gils83
gils83
Posted
Posted

:)

 

hello spakk, 
 
why develop a kernel 13.2 beta? it would be wiser to wait for the official release 10.9.2 Maveriks to continue, beta kernel changes can still happen so the best solution may be to wait for the official version that the coders working on it.
 
the Maveriks Version 10.9.0/1 works perfectly on AMD and has nothing to envy to the vanilla version for Intel. The Haswell processors (Intel) encounter more problems than AMD CPU is a little history of the biter !  :) 
  • Like 1
Link to comment
Share on other sites
 Share
Go to topic listing
×
×
  • Create New...