Hi
Thanks for your reply
On Tue, Aug 11, 2009 at 10:38 AM, Mulyadi Santosa <mulyadi.santosa@xxxxxxxxx> wrote:
Hello
On Mon, Aug 10, 2009 at 11:07 PM, Basavaraj
Dengi<basavarajdengi@xxxxxxxxx> wrote:> I am porting 'kdb' to omap3430 platform.I have no experience about embedded development at all, including
> for command 'mdp arg' [mdp is memory dump physical and arg is any physical
> address] this code gets executed. I dint test for every address b/w
> 0-0xc0000000
> but lots of them randomnly. The other 'md' commands do work for kernel high
> memory address.[ie. > 0xc0000000] But mdp fails for all the addresses.
>
> this piece of code is part of architecture independant patch of kdb; [for
> 2.6.24 kernel]
> kdb community is very inactive and moreover this question is related to
> memory management part of the kernel, so I thought i will ask it here
Omap. But questions that cross my mind are:
1. Is Omap MMU or MMU less? If it is MMU-less, I think talking about
page frame is completely irrelevant here. Or at least MMU related
functions will try to "emulate" the MMU situation, but still not the
same as MMU based arch
OMAP is having MMU
2. Still related with above question, perhaps this Omap arch implement
kinda "memory hole"? So not every memory address range is mapped?
I am not sure about this;
3. Does Omap use memory split like x86 does? I mean 3:1 for user
space:kernel space?
Ya it has memory split.
The possibility that some "md" commands doesn't work when dealing with
address below kernel space range is it has to grab the page directory
of current running process and does the translation based on this PGD
(again, assuming Omap is using MMU). And this grabbing mechanism is
either unimplemented yet or has some bugs.
I agree; But the issue here is not about md commands failing for lower memory addresses;
Issue is that mdp fails for higher memory [>3GB] addresses. I am trying to figure out the same.
My thoughts might have flaws, so feel free to debate them.
--
