At 2:09 AM -0500 6/4/2000, Michel Lanners wrote:

Hi Franz,

On   3 Jun, this message from Franz Sirl echoed through cyberspace:
[hacking controlfb]

quoted

if you don't mind, would you try to replace the "dcbi" assembler
instructions
in the detection code with "dcbf" instructions? It seems Dan inherited
my bug
I had in the platinumfb driver...

Sure, no problem.

In what sense is that a bug, and what would be the symptom? I know close
to nothing about PPC assembler, and fully understanding the detection
code might help fix the detection problems we still have in some
situations...

dcbi = data cache block invalidate. If you do this on a dirty cache block,
data will be lost.

dcbf = data cache block flush. If you do this on a block that isn't in a
coherent state, data will be lost.

If you have a block that's both dirty and incoherent... well, don't do that. :)


btw, my mail server messed up and I got bounced off the list a few weeks
back, have I missed anything interesting?


Cheers - Tony :)


--
Tony Mantler       Renaissance Nerd Extraordinaire     nicoya@apia.dhs.org
Winnipeg, Manitoba, Canada                     http://nicoya.feline.pp.se/


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