On Wed, 2009-11-25 at 19:10 +0000, srinidhi kasagar wrote:

On Tue, 2009-11-24 at 14:55 +0100, Catalin Marinas wrote:

quoted

On Tue, 2009-11-24 at 13:35 +0000, Surinder P Singh wrote:

quoted

On Tue, Nov 24, 2009 at 6:05 AM, srinidhi kasagar
[off-list ref] wrote:

quoted

quoted

2 points:

1. Since this code is also valid for devices based on pre-ARMv6, maybe
making this code conditional for >=ARMv6 would be cleaner ?

rather it depends on l2 controller being used. L210 controllers
do not have such restrictions whereas l220/pl310 have such kind
of secure/non-secure restrictions. So would it be better to keep
condition based on l2 controller being used?

Thats probably better. You can read the L2 cache ID register to figure
out if its a L210/220 or PL310 and so on. You can couple this check
with the cpu secure/non-secure state before deciding to write to the
registers.

I don't think it's worth the hassle. Just always check whether it is
already enabled without additional ifdefs. IIRC, L210 is used on
RealView PB1176 (not entirely sure).

So, does the patch which I have sent still valid which just checks
whether l2x0 is already enabled?

OK with me as well, just minor issues with the coding style (the
multi-line comment).

-- 
Catalin