[OpenBIOS] Haiku/ppc regression?

[Igor Kovalenko]

On Fri, May 21, 2010 at 12:00 PM, Mark Cave-Ayland
<mark.cave-ayland at siriusit.co.uk> wrote:
> Andreas Färber wrote:
>
>> I've dug some more and noticed that the data parsed from
>> /chosen/mmu/translations seems wrong:
>>
>> 0: map: 0x00000000, length 16384 -> physical: 0x00000000, mode 132972544
>> 1: map: 0x00030000, length 0 -> physical: 0x07f00000, mode 1048576
>> 2: map: 0x00000002, length -2147483648 -> physical: 0x000eb000, mode 106
>>
>> Note the weird mode and subsequent values. It looks as if Haiku expects
>> (and gets on real OpenFirmware):
>> void *virtual_address
>> int length
>> void *physical_address
>> int mode
>> whereas OpenBIOS writes in
>> libopenbios/ofmem_common.c:ofmem_update_mmu_translations:
>> props[ncells++] = t->virt
>> props[ncells++] = t->size
>> props[ncells++] = t->mode
>>
>> Should this be changed as follows, or is this platform-dependent?
>
> Hmmm that's interesting. According to the OpenSolaris kernel source the
> translations property is mapped to the following struct:
>
> http://cvs.opensolaris.org/source/xref/onnv/onnv-gate/usr/src/uts/sun4v/sys/prom_plat.h
>
> struct translation {
>        uint32_t virt_hi;       /* upper 32 bits of vaddr */
>        uint32_t virt_lo;       /* lower 32 bits of vaddr */
>        uint32_t size_hi;       /* upper 32 bits of size in bytes */
>        uint32_t size_lo;       /* lower 32 bits of size in bytes */
>        uint32_t tte_hi;        /* higher 32 bites of tte */
>        uint32_t tte_lo;        /* lower 32 bits of tte */
> };
>
> And in Haiku:
>
> http://dev.haiku-os.org/browser/haiku/trunk/src/system/boot/platform/openfirmware/arch/ppc/mmu.cpp
>
> struct translation_map {
>        void    *virtual_address;
>        int     length;
>        void    *physical_address;
>        int     mode;
> } translations[64];
>
>
> So in short, it does indeed look as if there are some platform differences
> here. I guess this extra field is based upon the Mac OF implementation so it
> would be good for someone to verify this on a real Mac before we go and hack
> the source for PPC-only hosts. Note that I can't see any references to the
> translations property in the OF spec so I guess it must be an extension.

For arm and ppc we have {virt,size,phys,mode} documented in proposed
platform bindings.
http://playground.sun.com/1275/bindings/arm/arm0_3d.html#34579
http://playground.sun.com/1275/bindings/ppc/release/ppc-2_1.html#REF34579

For sparc there seems to be no such proposal, and linux uses {virt,size,mode}

-- 
Kind regards,
Igor V. Kovalenko