[PATCH V2 0/2] ARM: l2c: OMAP4/AM437x: Additional register programming support.

[Tomasz Figa]

Hi Tony,

2015-01-03 9:23 GMT+09:00 Tony Lindgren <tony at atomide.com>:
> * Nishanth Menon <nm at ti.com> [150102 11:50]:
>> On 01/02/2015 12:46 PM, santosh.shilimkar at oracle.com wrote:
>> > On 1/2/15 9:43 AM, Nishanth Menon wrote:
>> >> Hi,
>> >> OMAP4 and AM437x ROM code provides services to program PL310's latency
>> >> registers and AM437x provides service for programming Address filter
>> >> registers.
>> >>
>> >> Provide support in the kernel for the same.
>> >>
>> >> V2 of the series contains documentation update and a bug fix due to a
>> >> typo introduced during patch split :(
>> >>
>> >> Nishanth Menon (2):
>> >>    ARM: l2c: OMAP4/AM437x: Introduce support for cache latency
>> >>      programming
>> >>    ARM: l2c: AM437x: Introduce support for cache filter programming
>> >>
>> > Looks fine to me ...
>> > Feel free to add my ack if you need one ...
>> >
>> > Minor: The subject looks like I2C though it is L2C ;-)
>> >
>> Yeah, the thought did occur to me, but decided instead to go with the
>> existing $subject conventions of arch/arm/mach-omap2/omap4-common.c
>> ARM: l2c: omap2+: get rid of init call
>> ARM: l2c: omap2+: get rid of redundant cache replacement policy setting
>> ARM: l2c: omap2: remove explicit non-secure access bits
>> ARM: l2c: omap2: remove cache size override
>> ARM: l2c: omap2: remove explicit SMI calls to enable L2 cache
>> ARM: l2c: omap2: implement new write_sec method
>> ARM: l2c: remove platforms/SoCs setting early BRESP
>> ARM: l2c: fix register naming
>> ARM: l2c: omap2: remove ES1.0 support
>>
>> ..
>>
>> If folks feel strongly about this, I can capitalize the same and post
>> a v3 to help confusing fonts on certain mail clients and terminals.
>> let me know if folks want me to.
>
> I guess no need to :)
>
> Looks like these still won't fix the issue we found in the
> series posted by Tomasz though. At least I'm still getting errors
> on am437x with these and the patches from Tomasz applied.

Indeed, as I figured out in the original thread about this issue,
additional patch fixing code unaffected by my series (besides changing
the condition which triggers calling it) is necessary. Namely, the
affected 4 registers need to be written using the write_sec wrapper,
instead of using writel*() directly.

Best regards,
Tomasz