Re: [GIT PULL] for testing: OMAP hwmod driver conversions: watchdog, UART, i2c

[Kevin Hilman <khilman@deeprootsystems.com>]

"Nayak, Rajendra" <rnayak@ti.com> writes:

>> -----Original Message-----
>> From: Kevin Hilman [mailto:khilman@deeprootsystems.com]
>> Sent: Wednesday, October 06, 2010 1:19 AM
>> To: Nayak, Rajendra
>> Cc: Paul Walmsley; Cousson, Benoit; Tony Lindgren; linux-omap@vger.kernel.org; Varadarajan, Charulatha; Raja,
>> Govindraj
>> Subject: Re: [GIT PULL] for testing: OMAP hwmod driver conversions: watchdog, UART, i2c
>> 
>> "Nayak, Rajendra" <rnayak@ti.com> writes:
>> 
>> >> -----Original Message-----
>> >> From: Kevin Hilman [mailto:khilman@deeprootsystems.com]
>> >> Sent: Tuesday, October 05, 2010 10:34 PM
>> >> To: Nayak, Rajendra
>> >> Cc: Paul Walmsley; Cousson, Benoit; Tony Lindgren; linux-omap@vger.kernel.org; Varadarajan, Charulatha; Raja,
>> >> Govindraj
>> >> Subject: Re: [GIT PULL] for testing: OMAP hwmod driver conversions: watchdog, UART, i2c
>> >>
>> >> "Nayak, Rajendra" <rnayak@ti.com> writes:
>> >>
>> >> > <snip>..
>> >> >
>> >> >> >
>> >> >> > Below is an untested patch to provide some mechanism to deal with this --
>> >> >> > I'd appreciate everyone's comments on this, particularly the comments in
>> >> >> > the patch code on how to deal with this problem.
>> >> >>
>> >> >> Hi Paul,
>> >> >>
>> >> >> I'll test this patch to see if it solves the issue seen with i2c on n800.
>> >> >
>> >> > Hi Paul, Benoit, Kevin,
>> >> >
>> >> > So looks like the issue with i2c on n800 isn't really because of additional
>> >> > delay needed after a clock enable, but related to the omap_readl/writel's
>> >> > done on the 16bit i2c registers as I was suspecting earlier.
>> >> >
>> >> > The below patch fixes the n800 issue for me. My mailer might mess up the alignment
>> >> > hence attaching it too.
>> >> >
>> >> > From af9733f2e63a16b446635d52015e05d631f0788f Mon Sep 17 00:00:00 2001
>> >> > From: Rajendra Nayak <rnayak@ti.com>
>> >> > Date: Tue, 5 Oct 2010 16:36:30 +0530
>> >> > Subject: [PATCH] OMAP: hwmod: Handle modules with 16bit registers
>> >> >
>> >> > Some modules which have 16bit registers can cause imprecise
>> >> > aborts if a __raw_readl/writel is used to read/write 32 bits.
>> >> >
>> >> > Add an additional flag to identify modules which have such
>> >> > hard requirement, and handle it in the hwmod framework.
>> >> >
>> >> > Signed-off-by: Rajendra Nayak <rnayak@ti.com>
>> >>
>> >> Can you also send a corresponding patch to the i2c hwmod data.
>> > Hi Kevin,
>> >
>> > Ok, I'll repost the original patch which adds the omap2 hwmod data with
>> > these flags added for omap2420. Does that sound fine?
>> 
>> Since I've alrady done some manual merge conflicts here, you can just
>> post an additional patch and I will fold it in.
>
> Ok, so here's the patch
> ------
> From 06b4bb7a0ce96776d6d9b344b4786b48fceeffe1 Mon Sep 17 00:00:00 2001
> From: Rajendra Nayak <rnayak@ti.com>
> Date: Wed, 6 Oct 2010 01:49:32 +0530
> Subject: [PATCH] OMAP2: hmwod: Fix impresice aborts seen on n800/n810
>
> I2C module on OMAP2420 has 16bit registers and causes imprecise
> aborts if 32bits are read/wriitten into.
> Use the HWMOD_16BIT_REG flag to notify the hmwod framework of this
> hard requirement so that __raw_writew/readw is used to read /write
> the mdoule registers.
>
> Signed-off-by: Rajendra Nayak <rnayak@ti.com>
> ---
>  arch/arm/mach-omap2/omap_hwmod_2420_data.c |    2 ++
>  1 files changed, 2 insertions(+), 0 deletions(-)

Thanks.

Shouldn't we do the same for 2430?  I guess the imprecise aborts aren't
happening on 2430, but I think we should still use the 16-bit accesses
there too.

What do you think?

Kevin