Re: [Qemu-devel] Mac OS X PPC host support (was: [4932] Preliminary PPC64/Linux host support)

[malc <av1474@comtv.ru>]

On Tue, 29 Jul 2008, Andreas F?rber wrote:

>
> Am 27.07.2008 um 20:14 schrieb malc:
>
>> On Sat, 26 Jul 2008, Andreas F?rber wrote:
>> 
>>> 
>>> Am 26.07.2008 um 20:54 schrieb malc:
>>> 
>>>>> there are currently numbers hardcoded for Linux.
>>>> It's not just about numbers, last time i looked (cursory) the ABIs where
>>>> a lot different (PPC32 case) just changing the numbers wont get one very
>>>> far i think.
>>> 
>>> Of course not only, but it's one step that could be applied. The stack 
>>> alignment is 16 Bytes for both, and the prolog in my ppc(32) branch seems 
>>> to match the referenced Apple pseudocode now.
>>> 
>>> Where else do you have differences in mind? Which additional functions may 
>>> need to be adjusted?
>> 
>> I'd have to check Darwins ABI to be sure, but here's the short list:
>
> http://developer.apple.com/documentation/DeveloperTools/Conceptual/LowLevelABI/Articles/32bitPowerPC.html
>
>> a. Alignement of 64 bit arguments (in the register file)
>>   (This was the thing that confused Sparcs ASI helpers when i first
>>    tried to boot Sparc on, then new, PPC32 tcg)
>
> What register file? Please be more detailed, I'm not intimate with TCG.

I meant this http://en.wikipedia.org/wiki/Register_file
But anyhow it is not aligned hence the ASI failure You saw earlier.

>> b. List of reserved registers
>
> According to the Apple ppc ABI document, r2 is not system-reserved compared 
> to SVR4 ppc ABI. The other general-purpose registers look similar (assuming 
> we don't have nested functions), although Apple seems to have less strict 
> prescriptions what to store where.
>
>> c. The approach to TOC (or however Apple calls the substitute)
>
> Not sure what you mean. Pointers to concrete files and functions might help.

I have checked and MacOS X on PPC32 doesn't use those.

>
>> In any case, PPC32 + MacOS X i can help with. PPC64 + MacOS X not so much.
>
> Restoring Mac OS X ppc(32) to work again would be just great.
> PPC64 didn't work pre-TCG. We'll probably know where to poke ourselves once 
> ppc is running.

I have made a stab at making it work and tried it on MacOS X (Tiger) that
i have here, for Sparc it progress a bit further but still meets it's doom
shortly:

qemu: fatal: Trap 0x29 while interrupts disabled, Error state
pc: ffd04020  npc: ffd04024
General Registers:
%g0: 00000000   %g1: 00000000   %g2: ffd908e0   %g3: ffd00000
%g4: ffd83038   %g5: 00000001   %g6: 00000000   %g7: 00000000
Current Register Window:
%o0: 00000000   %o1: 00000000   %o2: 00000000   %o3: 00000000
%o4: 00000000   %o5: 00000000   %o6: ffdd3fa0   %o7: 00000000
%l0: 00000000   %l1: 00000000   %l2: 00000000   %l3: 00000000
%l4: 00000000   %l5: 00000000   %l6: 00000000   %l7: 00000000
%i0: 00000000   %i1: 00000000   %i2: 00000000   %i3: 00000000
%i4: 00000000   %i5: 00000000   %i6: 00000000   %i7: 00000000

Floating Point Registers:
[all bit zero FPRs snipped]
psr: 0x04001fc0 -> ---- SP- wim: 0x00000001
fsr: 0x00080000

BlueSwirl, rings any bells?

I tried my best to hack it while in Mac OSX, but, honestly, i can not
stand that system and have no spare (quiet enough) machine to SSH into
it and do things remotely, sorry.

The code is available at:

http://repo.or.cz/w/qemu/malc.git?a=shortlog;h=refs/heads/tcgppc32macosx

-- 
mailto:av1474@comtv.ru