Re: [PATCH 00/14] POWER: BQ27x00: New Properties, fixes, bq27000 support

[Lars-Peter Clausen <lars@metafoo.de>]

On 02/07/2011 01:58 AM, Grazvydas Ignotas wrote:
> On Sun, Feb 6, 2011 at 2:47 AM, Lars-Peter Clausen <lars@metafoo.de> wrote:
>> This patch series contains a few updates for the bq27x00 driver:
>> * Support for additional power supply properties
>> * Support for the bq27000 battery which is identical to the bq27200 but is
>>  connected through the HDQ bus.
>> * Adds a register cache to the driver and introduces polling the batteries state
>> * Minor improvements and cleanups
>>
>> The last patch in this series is not specific to the bq27x00 driver but is
>> required for uevents to be generated properly for this driver.
>> The patch makes properties which return -ENODATA to be ignored when generating
>> uevents. Previously in such a case uevent generation would have been aborted
>> with an error. But since the bq27x00 return -ENODATA for the TIME_TO_FULL
>> property when the battery is not charging and for the TIME_TO_EMPTY property
>> when the battery is not discharging and at least one of them is always true
>> uevent generation would always fail.
>>
>> This series has so far been tested with the bq27000 and the bq27200 battery, but
>> not with the bq27500 battery, so it would be nice if somebody with a board
>> containing such a battery could test the patches to make sure that there are no
>> regressions.
> 
> Trying this on pandora board (bq27500 over i2c) and it seems there is
> something wrong with the cache, I'm always getting the old values. I'm
> reading them manually over sysfs, uptime is over 25min but values
> still match ones on boot, even after turning off the backlight.
> 
> Looking at the code delayed_work handling looks suspicious,
> bq27x00_battery_get_property() flushes it and nothing ever reschedules
> it. Other than that all new property values looks sane.

Hi

Hm, I tought that rescheduling from withing the work function would work, even when
flushing the work, at least it did here.
But the current code is broken if poll_interval is 0 anyway, cause then it wont
reschedule itself and is never run again.

Could you try:

if (time_is_before_jiffies(di->last_update + 6 * HZ)) {
	cancel_delayed_work_sync(&di->work);
	bq27x00_battery_work(&di->work);
}

- Lars