Re: SmartFirmware /memory (was: Pegasos dmesg)

[Frank Wille <frank%phoenix.owl.de@localhost>]

Tim Rightnour wrote:

>> The battable[] already contains a valid VA==PA entry for 0x10000000. This
>> is done automatically by oea_batinit(), which calls mem_regions() and
>> takes it from the "reg" property of "/memory".
>> 
>> So it doesn't help. System crash on the first OF-call, when adding the
>> region to "available".
> 
> What about getting it added to the available set, and then running the map
> command in OF to create a translation for it.

No difference.

Unless somebody (like Matt Sealey) states the opposite, I would guess that
most methods, like "map" and "unmap" in /cpus, are dummies. They do nothing.
I can play with any stupid arguments I like, but it has no effect.

Then I had a look into the SmartFirmware source at openbios.org. The map
function in cpu-ppc.c does nothing (same for unmap):
---8<---
C(f_mmu_map)        /* map (phys.lo ... phys.hi virt size mode --) */
{
    Int cells = get_address_cells(e->currpkg->parent);
    Byte *virt;
    Int size, mode;
    Int lo, hi;

    IFCKSP(e, cells + 3, 0);
    POP(e, mode);
    POP(e, size);
    POPT(e, virt, Byte*);

    /* hack to get only the hi and lo cells - machine-dependent */
    POP(e, lo);
    DROPN(e, cells - 2);
    POP(e, hi);

    /* perform machine-dependent mapping here */

    return NO_ERROR;
}
---8<---

I don't think that bplan/Genesi changed that... :|


>>> That, or run through it with pmap_enter to
>>> add it all to the page table?
>> 
>> Honestly I don't know where to do that. Would it make any difference?
> 
> It might. You can try it real quick by fiddling with the if 0'd code at
> the bottom of pmap.c inside the PMAP_NEED_MAPKERN

Ok. I tried this:
---8<---
        for (va = 0x10000000, pa = 0x10000000; va < 0x20000000; 
             pa += PAGE_SIZE, va += PAGE_SIZE) {
            ptegidx = va_to_pteg(pm, va);
            pmap_pte_create(&pt, pm, va, pa | PTE_M|PTE_BW);
            pmap_pte_insert(ptegidx, &pt);
        }
---8<---

But again, same crash.

-- 
    _  Frank Wille (frank%phoenix.owl.de@localhost)
 _ //  http://sun.hasenbraten.de/~frank/
 \X/   Phx @ #AmigaGer