LPAE for outer caches functiopns (was Re: [RFC 3/6] arm: cache-l2x0: add support for Aurora L2 cache ctrl)

[Gregory CLEMENT]

On 08/10/2012 04:47 PM, Will Deacon wrote:
> On Thu, Aug 09, 2012 at 05:48:44PM +0100, Gregory CLEMENT wrote:
>> Hi Will,
> 
> Hi Gregory,
> 
>> You will find an updated version of this patch with LPAE support. I've
>> tested with and without LPAE selected.

Just a rectification, in fact I managed to build in both case, and to run
only without LPAE support. I didn't have the support for LPAE yet. I was
surprised that it worked out of the box, but in fact I tested the wrong
kernel! I realized this this morning.

> 
> Thanks for the patches.
> 
>> Now I get some warning during compilation: "initialization from
>> incompatible pointer type [enabled by default]"
>>
>> It is because the outer_cache_fns struct still embed functions with
>> unsigned long for address instead of phys_addr_t. You are aware of
>> it, as you started to work on it with your patch "ARM: 6671/1: LPAE:
>> use phys_addr_t instead of unsigned long in outercache functions".
> 
> Correct, I just fixed up the wrapper functions in that patch since no outer
> cache implementations required >32 bits of physical address. You're the
> lucky guy with the first implementation of such a controller :)
> 
>> So a first step would be to update the definitions in struct
>> outer_cache_fns and also in the files using this prototype, I found
>> only 4 files:
>>
>> git grep -w  outer_.*_range arch/arm | grep = | cut -f 1| uniq
>> arch/arm/mm/cache-feroceon-l2.c:
>> arch/arm/mm/cache-l2x0.c:
>> arch/arm/mm/cache-tauros2.c:
>> arch/arm/mm/cache-xsc3l2.c:
>>
>> But it is not enough we also fixed the call to theses functions:
>> git grep -w  outer_.*_range arch/arm | grep -v = | cut -f 1 -d: | uniq
>> arch/arm/include/asm/outercache.h
>> arch/arm/kernel/smp.c
>> arch/arm/kernel/suspend.c
>> arch/arm/mach-exynos/platsmp.c
>> arch/arm/mach-highbank/highbank.c
>> arch/arm/mach-msm/platsmp.c
>> arch/arm/mach-omap2/omap-secure.c
>> arch/arm/mach-ux500/platsmp.c
>> arch/arm/mm/dma-mapping.c
>> arch/arm/mm/fault-armv.c
>> arch/arm/plat-versatile/platsmp.c
>>
>> Most of them use __pa or directly __virt_to_phys, so once the patch
>> "[PATCH 03/22] ARM: LPAE: use phys_addr_t on virt <--> phys
>> conversion" will be merged the correct type will be used.
> 
> Which patch is this? part of the keystone series?

yes it is!

> 
>> Finally the last file which need some change will be
>> arch/arm/mm/dma-mapping.c.
>>
>> Does it sound correct?
>>
>> If it does, then I can prepare a patch for it.
> 
> Yes please, that sounds like the right direction for this. We should use
> phys_addr_t wherever we're dealing with physical addresses.
> 
> Cheers,
> 
> Will
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
> 


-- 
Gregory Clement, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com