Jump to content

kernel panic while booting installer on Atom 330 board [SOLVED]

povverty

By povverty,
in OSx86 10.6 (Snow Leopard)

 Share
16 posts in this topic

Recommended Posts

povverty
povverty
Posted
Posted

When I try to boot the prepared 10.6 installer (booting with latest chameleon, patched dsdt in place) on my POV-ION-330 board (same as PEGATRON), I get a kernel panic and "should have 2 cores, but only found 1".

 

I had the same problem with 10.5, but it could be fixed with cpus=2. With 10.6 it seems to have no effect.

 

any ideas

 

EDIT: Kernel panic fixed by decompiling the dsdt.aml, removing P003, P004 and the CPU aliases and recompiling

dsdt_pov330.zip

 

EDIT2: fixed Ethernet connection problem by installing the driver from this thread

http://www.insanelymac.com/forum/index.php?showtopic=125569

(remove nvenet.kext from /System/Library/Extensions/IONetworkingFamily.kext/Contents/PlugIns/ and put nForceLAN.kext there - could not get it running via /Extra/)

shoarthing
shoarthing
Posted
Posted
When I try to boot the prepared 10.6 installer (booting with latest chameleon, patched dsdt in place) on my POV-ION-330 board (same as PEGATRON), I get a kernel panic and "should have 2 cores, but only found 1".

 

I had the same problem with 10.5, but it could be fixed with cpus=2. With 10.6 it seems to have no effect.

 

any ideas

This error comes from AppleACPICPU - yes? . . if so, you need both a patched dsdt.aml [with CPU alias' removed] & to have disabled HTT in your BIOS.

 

Here's a link to a report to the Voodoo kernel team [you get this message with all kernels after 10.5.6] . . . in there are further links to other reports from Atom 330 users.

povverty
povverty
Posted
  • Author
Posted

The kernel trace looks similar to the photo in the link you provided - so yes it seems to be the same problem.

 

 

What puzzles me ist that I am able to boot a stock 10.5.8 kernel with cpus=2 and my patched dsdt but found no way to boot the 10.6 kernel with a similar trick.

 

My problem is that my BIOS has no option to disable hyperthreading - so I cannot test that :-(

shoarthing
shoarthing
Posted
Posted
The kernel trace looks similar to the photo in the link you provided - so yes it seems to be the same problem.

 

 

What puzzles me ist that I am able to boot a stock 10.5.8 kernel with cpus=2 and my patched dsdt but found no way to boot the 10.6 kernel with a similar trick.

 

My problem is that my BIOS has no option to disable hyperthreading - so I cannot test that :-(

. . . suggest you edit the dsdt to remove the HTT "CPU's" - these may be numbers 1 & 3 but you'll have to use trial & error.

 

Failing that . . well, I have read reports [take pinch of salt to taste] that flags such as: 

cpus=3

might work . . . me I'd nag POV support to provide a customer BIOS with HTT Enable/Disable feature - say you are using "Darwin" as your OS & they may well understand.

povverty
povverty
Posted
  • Author
Posted

I decompiled the dstl.aml, removed P002 P004 and the Alias and compiled again. The result is that I get no error but hang at "using 16384 buffer headers and 4096 cluster IO buffer headers" (I think the same behaviour as when providing no dsdt.aml at all).

 

I might have to wait for PEGATRON or POV to provide a new BIOS ;)

 

	{
	Processor (P001, 0x01, 0x00004010, 0x06) {}
	Processor (P002, 0x02, 0x00000000, 0x00) {}
	Alias (P002, CPU2)
	Processor (P003, 0x03, 0x00000000, 0x00) {}
	Processor (P004, 0x04, 0x00000000, 0x00) {}
	Alias (P004, CPU4)
}

shoarthing
shoarthing
Posted
Posted

. . what happens if you remove all CPU instances bar the first one?

 

 

. . & don't wait for Pegatron/POV - nag 'em [politely] - you've paid for a mobo which doesn't work with your OS of choice & it is their responsibility to fix this: making a custom AMI BIOS is pretty trivial if you have the toolkit.

povverty
povverty
Posted
  • Author
Posted
. . what happens if you remove all CPU instances bar the first one?

 

then... it works! ;)

 

I got to the install screen - and even booting with usb 2.0 enabled seems to work.

 

OFFTOPIC:

I didnt know however that a GUID partition table is a requirement. so I cannot get further now (or is it possible to create a GUID partition table without loosing data? I have a 300G linux partition that I dont want to loose....

shoarthing
shoarthing
Posted
Posted
then... it works! ;)

 

I got to the install screen - and even booting with usb 2.0 enabled seems to work.

 

OFFTOPIC:

I didnt know however that a GUID partition table is a requirement. so I cannot get further now (or is it possible to create a GUID partition table without loosing data? I have a 300G linux partition that I dont want to loose....

. . you can make [or search for & d/l] a hacked 'osinstall.mpkg' which'll allow installation onto MBR-partitioned disks.

 

Glad the dsdt hack worked; but we'll need to see whether this in fact gives you 2x real cores, each w/ 512KB of L2 cache, or just the first Atom core plus its HTT echo.

 

Please report on this ;)

povverty
povverty
Posted
  • Author
Posted
. . you can make [or search for & d/l] a hacked 'osinstall.mpkg' which'll allow installation onto MBR-partitioned disks.

Ok, that did the trick - thanks! :(

Installation took really long, but it worked.

 

Glad the dsdt hack worked; but we'll need to see whether this in fact gives you 2x real cores, each w/ 512KB of L2 cache, or just the first Atom core plus its HTT echo.

 

Please report on this ;)

system profiler reports one cpu with one core. :(

I dont know about better ways to check it like cat /proc/cpuinfo in linux.

 

What works:

- nvidia graphics with dvi out

- sleep

- usb 2.0 (sometimes)

 

What does not:

- second cpu core

- hyperthreading

- audio

- ethernet (detected but no link)

- usb 2.0 (sometimes)

 

(i didnt try to fix anything yet - just glad that it works now - at least to some degree)

shoarthing
shoarthing
Posted
Posted

Are you disabling AppleIntelCPUPowerManagement - using, say, NullCPUPowerManagement.kext from/Extra/Extensions?

 

. . if so, you *should* be able to edit out from your dsdt the 2 HTT 'cores' - leaving the 2 actual cores from your Atom 330

 

ugokind [re: inserting power-states] this may well be an avenue worth looking at; to your knowledge has anyone got injected powerstates to work with this CPU?

povverty
povverty
Posted
  • Author
Posted
Are you disabling AppleIntelCPUPowerManagement - using, say, NullCPUPowerManagement.kext from/Extra/Extensions?

 

I used the Disabler from Chameleon 2.0-RC3

 

. . if so, you *should* be able to edit out from your dsdt the 2 HTT 'cores' - leaving the 2 actual cores from your Atom 330

 

At least it did not work to remove cpu 2 and 4 while leaving 1 and 3 intact.

I could try to remove 3 and 4 and leave 1 and 2 intact....

shoarthing
shoarthing
Posted
Posted
I used the Disabler from Chameleon 2.0-RC3

 

 

 

At least it did not work to remove cpu 2 and 4 while leaving 1 and 3 intact.

I could try to remove 3 and 4 and leave 1 and 2 intact....

. . . check in its plist that this 'disabler.kext' is not the upstreamclient disabler [you probably need this as well].

 

There's a useful repository of SL kexts in Stellarols's blog.

povverty
povverty
Posted
  • Author
Posted
. . . check in its plist that this 'disabler.kext' is not the upstreamclient disabler [you probably need this as well].

 

Yes, the disabler I am using disables AppleIntelCPUPowerManagement

 

removing cpu 3 and 4 from the dsdt worked. Now SL reports 1 cpu with 2 cores. I hope the second is a real core, not the hyperthreading fake core.

shoarthing
shoarthing
Posted
Posted
Yes, the disabler I am using disables AppleIntelCPUPowerManagement

 

removing cpu 3 and 4 from the dsdt worked. Now SL reports 1 cpu with 2 cores. I hope the second is a real core, not the hyperthreading fake core.

. . . what I'm looking at in System Profiler in 10.6.1 w/ HTT disabled says: "L2 cache (per core) 512KB"
 Share
Go to topic listing
×
×
  • Create New...