Re: [PATCH V3 5/5] rtc: max77686: move initialisation of rtc regmap, irq chip locally

[Krzysztof Kozlowski]

On 04.02.2016 11:03, Laxman Dewangan wrote:
> 
> On Thursday 04 February 2016 05:40 AM, Krzysztof Kozlowski wrote:
>> On 03.02.2016 22:47, Laxman Dewangan wrote:
>>>
>>> Changes from V2:
>>> - Taken care of missed sequence for removing the resource.
>>> - Fix the crash when doing unbind by using requested_threaded_irq()
>>>    instead of demv_requested_threaded_irq().
>>>
>>>   drivers/mfd/max77686.c               |  85 +-------------------
>>>   drivers/rtc/rtc-max77686.c           | 148
>>> ++++++++++++++++++++++++++++++-----
>>>   include/linux/mfd/max77686-private.h |   3 -
>>>   3 files changed, 130 insertions(+), 106 deletions(-)
>>>
>> Looks good and works properly (RTC, unbind/rebind, wakealarm again)
>> but... after applying this patch the device cannot suspend to RAM after
>> unbind/rebind.
>>
>> Reproduce path, after booting:
>> $ echo max77686-rtc > /sys/bus/platform/drivers/max77686-rtc/unbind
>> $ echo max77686-rtc > /sys/bus/platform/drivers/max77686-rtc/bind
>> $ rtcwake -d rtc2 -m mem -s 5 -v
>> (or instead of rtcwake: echo mem > /sys/power/state)
> 
> 
> My device does not support suspend and hence I can nto test this part
> and need your help to debug it.
> 
> Few question to narrow down:
> 1. Have you try to bind/unbind mfd driver to make sure that there is no
> inherent issue in the framework driver?

That might be tricky... there are also regulator and clock drivers.  I
would have to disable them... It might take me some time to prepare an
environment for this.

> 
> 2. Does suspend entry and wake alarm work if you dont do unbind and just
> bind it and enter into suspend?
If by "just bind" you mean bind during system boot, then answer is yes.
Without the unbind+bind, the system suspends correctly.

Without your patch, system suspends correctly always (no rebind, rebind
etc.).

Actually I don't have to re-bind the RTC. Just:
$ echo max77686-rtc > /sys/bus/platform/drivers/max77686-rtc/unbind
$ echo mem > /sys/power/state
triggers the issue.

> 
> 3. Can you try locally to not use devm_regmap_init_i2c() and just use
> the regmap_init_i2c() and proper removal of this from error path and
> remove callback?

I'll try to find some time for that. Maybe tomorrow.

BR,
Krzysztof

> Three things we are doing new on this movement, i2c client creating,
> regmap_init and adding irq chip.
> And probably some issue on this calls?
> Let's try with not using devm_ * for this calls?
> 
> 
> 
>