On Thu, Nov 19, 2015 at 2:18 PM, Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote:
> Hi,
>
> On Thu, Nov 19, 2015 at 7:24 PM, Dmitry Torokhov
> <dmitry.torokhov@xxxxxxxxx> wrote:
>> On Thu, Nov 19, 2015 at 02:26:37PM +0200, Irina Tirdea wrote:
>>> Implement suspend/resume for goodix driver.
>>>
>
> [cut]
>
>>>
>>> +static int __maybe_unused goodix_suspend(struct device *dev)
>>> +{
>>> + struct i2c_client *client = to_i2c_client(dev);
>>> + struct goodix_ts_data *ts = i2c_get_clientdata(client);
>>> + int error;
>>> +
>>> + /* We need gpio pins to suspend/resume */
>>> + if (!ts->gpiod_int || !ts->gpiod_rst)
>>> + return 0;
>>> +
>>> + wait_for_completion(&ts->firmware_loading_complete);
>>
>> This is not that nice as it may lead to angry splats from the PM core if
>> firmware loading takes too long and we start suspending before it
>> completes.
>
> What exactly do you mean? The suspend watchdog or something else?
I was thinking about dpm watchdog that will panic the system if
suspend takes too long. Hmm, this is debug facility and the default
timeout is 60 seconds, so I guess we can ignore my concern.
>
>> Rafael, if we issue pm_stay_awake() before requesting firmware and
>> pm_relax() once it is done, will this prevent the current suspend
>> timeouts?
>
> pm_stay_awake() only works if the checking of wakeup sources is
> enabled which generally depends on what user space does.
I see.
Thanks.
--
Dmitry
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
