[PATCH 1/6] rtc: rtc-s3c: Fix access unit from byte to word on RTCCON

[MyungJoo Ham]

On Wed, Oct 27, 2010 at 4:58 PM, Kukjin Kim <kgene.kim at samsung.com> wrote:
> MyungJoo Ham wrote:
>>
>> On Fri, Oct 8, 2010 at 8:41 AM, Kukjin Kim <kgene.kim at samsung.com> wrote:
>> > From: Changhwan Youn <chaos.youn at samsung.com>
>> >
>> > S3C2410_RTCCON of TYPE_S3C64XX RTC should be read/written by
>> > readw and writew, because TYPE_S3C64XX RTC uses bit 8 and 9.
>> > And TYPE_S3C2410 RTC also can access it by readw and writew.
>> >
>> > Signed-off-by: Changhwan Youn <chaos.youn at samsung.com>
>> > [atul.dahiya at samsung.com: tested on smdk2416]
>> > Tested-by: Atul Dahiya <atul.dahiya at samsung.com>
>> > Signed-off-by: Kukjin Kim <kgene.kim at samsung.com>
>> > Cc: Ben Dooks <ben-linux at fluff.org>
>> > ---
>> >  drivers/rtc/rtc-s3c.c |   36 ++++++++++++++++++------------------
>> >  1 files changed, 18 insertions(+), 18 deletions(-)
>>
>> Hello,
>>
> Hi,
>>
>> Sorry for a late reply...
>>
> Yeah, too late :-(
>
>> Anyway, I have a small question in this rtc-s3c.c driver.
>>
>> Is there any reason to use read/write b/w to access registers of rtc-s3c?
>>
> See the git comment.
>
>> Why don't we use readl/writel when accessing registers in this drivers
>
> I don't know why we should use readl/writel for all case...
> even though we can use just word or byte access.
>
>> and just forget which registers require at least 8 or 16 or 32 bits?
>>
>> In fact, it appears that readw/writew accesses 16bits, not 32bits in
>
> Yes...
>
>> ARM machines, which may incur problems with TICCNT(S3C2410_TICNT in the code)/CURTICCNT
>> registers.
>>
> I can't get your comment...

It is because TICCNT and CURTICCNT RTC registers of S5PC210 require
32bit access according to the manual while this patch still uses 16bit
accesses for them.

Besides, ALMYEAR/BCDYEAR(S3C2410_RTCYEAR) requires at least 16bit access.


This issue was found today while testing suspend-to-mem with s5pc210
and rtc-wakeup loop. My guess is that saving and restoring only 8 bits
of TICNT at suspend/resume function incurred the instable
suspend-to-mem (kernel hangs with about 1/100 probability) Using
readl/writel for TICNT solved the kernel hang issue.

>
> Thanks.
>
> Best regards,
> Kgene.
> --
> Kukjin Kim <kgene.kim at samsung.com>, Senior Engineer,
> SW Solution Development Team, Samsung Electronics Co., Ltd.
>
>



-- 
MyungJoo Ham, Ph.D.
Mobile Software Platform Lab,
Digital Media and Communications (DMC) Business
Samsung Electronics
cell: 82-10-6714-2858