[Fwd: [PATCH] PPC64: large INITRD causes kernel not to boot]

[Olaf Hering]

 On Mon, Aug 22, Segher Boessenkool wrote:

> >I guess that doesnt help the maple board with their PIBS 1.05 firmware.
> >I got 0x00aa7688 bytes for uncompressed kernel @ 0x00b00000,
> >which was used for the nvram msg_area (NVRAM_MAGIC did not match
> >anymore).
> >It seems SLOF isnt any better...
> 
> Hey, I told you that in private, no need to spout it on the
> mailing list.

Such bugs have to be fixed, obviously. No?

> Also, our memory map (esp. the new memory map) leaves almost
> all "low" memory to whoever wants to use it, so much less
> chance of collisions, even with broken clients (like Linux
> seems to be).

If I claim(0x0,123456,0) and OF says OK than I have to expect that its
really OK to use the mem at 0x0. In this case maple rejected everything
up to 11MB. But it has appearently some runtime data above 11MB, havent
debugged it further. At least claim(11MB,10MB,0) and claim(64MB,10MB,0)
and copying the memory from 11MB to 64MB and comparing it right away
showed differences. Pretty bad for an client owned area, no?