Re: success 2.4.0-test8 with latest bk

I got curious, and I tracked down the commented out
__clear_user(Hash, hash_size)

to version 1.10 of init.c

arch/ppc/mm/init.c
  1.10 00/02/04 13:20:23 cort@medea.fsmlabs.com +15 -2
  Changes to fix PPC64 MM (short term fix) and protect the system
  map memory from being written over.

Cort, any chance you happen to recall why that was needed

On Thu, 14 Sep 2000, Paul Mackerras wrote:

quoted

Takashi Oe writes:

quoted

Could you try the following change for arch/ppc/mm/init.c?

[snip]

quoted

		Hash = mem_pieces_find(Hash_size, Hash_size);
		clear_hash_table();

Huh?  There is a __clear_user(Hash, Hash_size) call in there but it's
commented out, which I don't understand at all.  Certainly the hash
table needs to be cleared.

Yeah, that's what I thought when I first saw the code a few months ago.

quoted

                            Using cacheable_memzero would probably be
the fastest way to do it.

__clear_user() works but I can't use cacheable_memzero() since frame
buffer memory resides on the same memory block.


Takashi Oe

--
--------------------------------------------------------------------------
Troy Benjegerdes                'da hozer'                hozer@drgw.net


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/