Milton Miller wrote:

On Sat Jun 16 08:06:19 EST 2007, Geoff Levand wrote:

quoted

This just removes the dependency __mmu_off has on the symbol
__after_prom_start.  I found the current code inconvenient when I
wanted to put some debugging code between the call to __mmu_off
and the branch to __after_prom_start.

You are entering this code with the mmu on?

No, mmu is off on PS3.

My first thought was that is unexpected, but aparently Apple machines 
do that, so its supported for the master -- not the slaves.

Yes, Apple enters with the mmu on.

quoted

Signed-off-by: Geoff Levand <geoffrey.levand at am.sony.com>
---
 arch/powerpc/kernel/head_64.S |    4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

--- a/arch/powerpc/kernel/head_64.S
+++ b/arch/powerpc/kernel/head_64.S

@@ -1695,9 +1695,11 @@ _GLOBAL(__start_initialization_multiplat
 2:

        /* Switch off MMU if not already */
-       LOAD_REG_IMMEDIATE(r4, .__after_prom_start - KERNELBASE)
+       LOAD_REG_IMMEDIATE(r4, __mmu_off_return - KERNELBASE)
        add     r4,r4,r30
        bl      .__mmu_off
+__mmu_off_return:
+
        b       .__after_prom_start

 _STATIC(__boot_from_prom)

-- 

That seems like a long label.

I could stand a 1: there (ie a local numeric label).

Hmm... how about always going directly there:

-       bl      .__mmu_off
-       b       .__after_prom_start
+       mtlr    r4
+       b       .__mmu_off


So that there is no space to add code that is skipped if the mmu was on?

This was intended to be a simple update.  I think the existing code is OK,
so I would prefer to just drop this patch for 2.6.23.  Feel free to continue
the effort.

-Geoff