Re: HTC: touchscreen driver

["Arve Hjønnevåg" <arve@android.com>]

2009/7/14 Trilok Soni <soni.trilok@gmail.com>:
> Hi Arve,
>
> 2009/7/15 Arve Hjønnevåg <arve@android.com>:
>> On Tue, Jul 14, 2009 at 3:16 PM, Pavel Machek<pavel@ucw.cz> wrote:
>> ...
>>>> >> This driver only supports a subset of Synaptics' devices so a more
>>>> >> generic driver will eventually be needed. The patch below adds support
>>>> >> for a more recent but similar panel.
>>>> >
>>>> > Ok, that should be simple enough to apply, but lets do improvements
>>>> > when we cleaned the code enough for the mainline...?
>>>>
>>>> We need this change now. Your cleanup will cause conflicts for anyone
>>>> using our driver, so it would be better if it includes all our fixes.
>>>
>>> Well, I don't expect you to use the cleaned-up driver as-is: I had to
>>> remove the wakelocks functionality as infrastructure is not in
>>> mainline.
>>
>> OK. The driver does not use wakelocks tough, only earlysuspend, so
>> your cleaned up version should be fully functional. The early suspend
>> hooks are only used to save power in idle when the screen is off.
>>
>>> (Plus, I do not have hardware to test your latest version).
>>
>> True, but the protocol documentation is available:
>> http://www.synaptics.com/sites/default/files/511_000099_01F.pdf
>>
>>>
>>> Applying small patch on top of cleaned up driver should be better than
>>> carrying whole driver itself, right?
>>
>> Perhaps, but the only problem I encountered with this driver so far
>> was when calling disable_irq from the interrupt handler stopped
>> working on non-smp systems in 2.6.30 (which is another patch you
>> should probably apply first).
>
> disable_irq review comment is already given in the patch posted on
> linux-input by Pavel. We should convert this driver to
> request_threaded_irq infrastructure instead. I have added linux-input
> ML in CC.
>

I looked into using request_threaded_irq a while ago, but decided it
was not worth it since I would still need to create the workqueue if
the interrupt was not available. However, if we drop support for
running without an irq then using request_threaded_irq should be
simpler.

-- 
Arve Hjønnevåg