[rtems commit] Shared MMU initialization for ARM BSPs and RaspberryPi MMU support

[Sebastian Huber]

On 2013-10-04 11:07, Hesham Moustafa wrote:
> On Fri, Oct 4, 2013 at 9:50 AM, Sebastian Huber
> <sebastian.huber at embedded-brains.de
> <mailto:sebastian.huber at embedded-brains.de>> wrote:
>
>     On 2013-10-03 14:49, Gedare Bloom wrote:
>
>         @@ -87,9 +90,9 @@
>         arm_cp15_start_setup___translation_table_and_enable___mmu_and_cache(
>              arm_cp15_set_domain_access___control(dac);
>              arm_cp15_set_translation___table_base(ttb);
>
>         -  /* Initialize translation table with invalid entries */
>         +  /* Initialize translation table with fixed-map read-write entries */
>              for (i = 0; i < ARM_MMU_TRANSLATION_TABLE___ENTRY_COUNT; ++i) {
>         -    ttb [i] = 0;
>         +    ttb [i] = (i << ARM_MMU_SECT_BASE_SHIFT) | ARMV7_MMU_DATA_READ_WRITE;
>              }
>
>
>     What is the reason for this change?  I don't think that unconditional
>     read-write access is a good idea.
>
> I made this change when I was trying to get RaspberryPi MMU initialization to
> work.

Then it is probably better to fix the memory map description of this particular 
BSP.

> Also I thought
> the default behavior for RTEMS is the unprotected mode, and the new protection
> (limitation) attributes
> are new and should be applied when needed (by configuration table at startup or
> dynamically).

Allow everything with explicit deny is very bad design.  It should be always 
deny everything with explicit allow.

-- 
Sebastian Huber, embedded brains GmbH

Address : Dornierstr. 4, D-82178 Puchheim, Germany
Phone   : +49 89 189 47 41-16
Fax     : +49 89 189 47 41-09
E-Mail  : sebastian.huber at embedded-brains.de
PGP     : Public key available on request.

Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG.