[PATCH v7 00/26] gpio/omap: driver cleanup and fixes

[DebBarma, Tarun Kanti]

On Tue, Sep 27, 2011 at 9:52 PM, Kevin Hilman <khilman at ti.com> wrote:
> "DebBarma, Tarun Kanti" <tarun.kanti at ti.com> writes:
>
>> On Tue, Sep 27, 2011 at 4:40 AM, Kevin Hilman <khilman at ti.com> wrote:
>>> "DebBarma, Tarun Kanti" <tarun.kanti at ti.com> writes:
>>>
>>> [...]
>>>
>>>> As pointed out by Kevin, debounce clock was not getting disabled.
>>>> In my testing I was somehow grepping CORE power domain instead
>>>> of PER power domain and hence missed it. The fix for the debounce
>>>> clock issue is at the end of the email.
>>>>
>>>> - Have re-based the for_3.2/gpio-cleanup branch against 3.1-rc6.
>>>> - Dropped [PATCH 26/26] gpio/omap: add dbclk aliases for all gpio modules
>>>> as suggested by Kevin since it's already taken care by hwmod.
>>>> - Added the debounce clock fix in the end.
>>>
>>> That debounce fix definitely makes things look better, but it's not
>>> solving the problem...
>>>
>>>> With above, PER is hitting low power state in Suspend and Idle path.
>>>>
>>>> Have pushed a branch at below URL with mentioned changes.
>>>> git://gitorious.org/omap-sw-develoment/linux-omap-dev.git
>>>> for_3.2/kevin/gpio-cleanup
>>>
>>> I tested your branch on my 3430/n900 and PER is still not hitting
>>> retention.  Setting all debounce values in the board file to zero using
>>> the patch below[1] makes PER hit retention again.
>>>
>>> Assuming you don't have an n900 to test with, I suggest you just copy
>>> the GPIO keys init from board-rx51-peripherals.c (or some of it) into
>>> the board file you are testing with.
>>>
>>> The problem is most likely be related to having more than one GPIO in a
>>> bank with debounce enabled, or more than one bank with GPIOs enabled and
>>> your current test is not be catching it.
>>>
>> As per commit c8c9fda506945 {OMAP: PM: disable idle on suspend for
>> GPIO and UART}, the gpio code needs to be fixed once GPIO driver is
>> run-time adapted.
>
> Great, good catch.
>
>> So I did below change as per the commit and now suspend is working
>> fine even with board files change for debounce functionality. So the
>> last series + below one line change is whats needed for suspend to
>> work. Can you please see if this does help on your board ?
>
> Yeah, with your patch, PER is hitting retention in suspend on my
> 3430/n900.
>
>> I am not finished my idle testing yet but just reporting the suspend
>> results.
>
> For me, PER is not hitting retention on idle.
I am working on this right now.

>
> Also, your repost of v7 doesn't included any of the comments I made on
> it yesterday.
That's right. By the time I received the comments the series was already posted.
Anyways, I will include them along with the PER retention issue in the
Idle path.
--
Tarun
>
> Kevin
>