Cheat Engine

[Tag-X]

Ok, more progress as it seems!

Boot with 6 cores and HT disabled:
https:/ /paste.ee/p/1Axgc

Debug build seems now to get as far as the release build.
After the last output, it halted with no reaction on key press or hard drive access (at line 1317 it tries to execute DB 0x66, wtf?!)

Offloading with 6 cores and HT enable:
https:/ /paste.ee/p/210L1

Seems to work now (yay!):
http:/ /pl.vc/5nj0z

However, I tried it again after reboot (which also worked: https:/ /paste.ee/p/xiHuh ) and then put my PC into energy saving mode (I read that it disables the VM and it seem to have worked). After that I tried the offloading again and got this:
https:/ /paste.ee/p/LgULL

[Dark Byte]

db 0x66 means to run the next command in the reverse addressing mode than it's currently in (32-bit code executes as 16-bit, 16-bit code executes as 32-bit)
the disassembler library used does it a bit weird with some instructions (like lgdt)
this is a valid part of initializing an CPU

As for the crash, it looks like it's happening after windows has awakened the other cpu's (I have a feeling that if you boot with just 1 cpu without hyperthreading it'll work)

Note that during the initialization of the secondary cpu's you have to be quick. (e.g linux only gives you 1 seconds before deciding a cpu has failed)

The issue here is that it's going into a unhandled doublefault, which would normally cause a reboot.
perhaps some of the realmode emulation goes wrong, it took too long and windows zeroed out the entrypoint for the cpu.

as for the reason, I can't say right now
apparently at 2699 an pagefault occurs, trying to dispatch interrupt 8 but the IDT hasn't been setup.
which was caused by 10:fffff8000311b2e6->0:2699

There's a reason I prefer runtime offloading (but perhaps when I get time I'll check if my latest system has hardware realmode VM support so I don't have to emulate that and implement that. (i7 920 does not) )


After energy saving mode, did you close and unload the driver first ? (kernelmoduleunloader)
else the driver might have remembered it's still loaded, and might instead load the first cpu assuming it's already initialized, which can explain the pagefault interrupts

[Tag-X]