Enable of Data Cache on mpc8240 and mpc8245

[Till Straumann]

Jennifer Averett wrote:

>I don't know if anyone has already resolved this,
>
I had. (my message from 2005/11/18 attached).
I committed everything *but* the actual L1_caches_enables() call
(I got delayed and now I don't have the loaner 2100 board anymore).
However, I am sure I tested it and it worked. But since I cannot test,
I was reluctant to commit this.

> but I was able to enable
>data cache on an mpc8245 processor.  I had to move the L1_caches_enables
>call to the bottom of the pretasking hook.
>  
>
why would you have to do that? For me, it worked the 'normal' way:

start.S: cache off, MMU off
            setup bat to map memory
            (should set up basic/empty page table but that is currently 
not done)
            MMU on

bspstart.c:
            setup MMU/bat mappings for I/O *prior* to doing *any* I/O
            enable L1 cache

Two states that must be avoided:
   MMU off, cache on   [all operations would be write-back cached,  
including  I/O
                                     which is bad]

   MMU on, insufficient mappings [ I/O fails w/o a way to print to console]

HTH Till

>Jennifer
>  
>

-------------- next part --------------
An embedded message was scrubbed...
From: Till Straumann <strauman at slac.stanford.edu>
Subject: Re: cache on mvme2100
Date: Fri, 18 Nov 2005 14:17:23 -0800
Size: 2729
URL: <http://lists.rtems.org/pipermail/users/attachments/20060222/d90ee008/attachment-0001.mht>