MMU Page table support for RTEMS
VALETTE Eric
valette at crf.canon.fr
Fri Jun 8 08:23:37 UTC 2001
tom_armistead wrote:
> I am working with RTEMs on PowerPC boards. RTEMs for PowerPC uses
> the PPC BATs for mapping of address space but unfortunately, BATs cannot
> address all of the address space needed for the boards that I am working
> with (1 Gig+ of RAM).
What a amount of memory!!! A high-end project I suppose :-)
> First, has anyone already done this? I'd rather not reinvent the wheel
> if it already exists.
For powerpc no. For Intel yes because of lack of decent cache flush
instruction for PCI devices registers handling...
NB : the MCP750 boot loader (written not be me but by Gabriel Paubert)
does enable pagination if I remember correctly. The MCP750 code (I
wrote) use only the BAT as you mentionned eralier.
> Second, would it be worthwhile to add hooks at the highest CPU level so
> that other CPU architectures could add their own implementation of the
> functions (similar to the way the cache functions work in libcpu)? I
> don't know enough about other CPU architectures to actually implement it
> for them but I could put in hooks to allow others to easily do so. Or
> would it be better to keep this code at the PowerPC specific CPU level?
Depends what you want to do with the MMU mapping. If you keep the 1-1
mapping and do not implement U/S proctection, we could probably define
something to
1) establish one/one mapping for chunk of memory,
2) manipulate page attribute like setting protection (RWX),
3) and caching atttribute (ON/OFF/Writethrough/Copyback) for chunk of
memory aligned on pages boundary.
I think joel could elaborate more...
My 0.02$
--
__
/ ` Eric Valette - Canon CRF
/-- __ o _. Canon Development Europe Team Leader
(___, / (_(_(__ Rue de la touche lambert
35517 Cesson-Sevigne Cedex
FRANCE
Tel: +33 (0)2 99 87 68 91 Fax: +33 (0)2 99 84 11 30
E-mail: valette at crf.canon.fr http://www.crf.canon.fr
More information about the users
mailing list