[Community] GTA04A5 / Letux 2804

Dr. H. Nikolaus Schaller hns at goldelico.com
Fri Jan 17 09:58:59 CET 2014


Am 16.01.2014 um 19:09 schrieb Dr. H. Nikolaus Schaller:

> Hi,
> 
> Am 16.01.2014 um 18:01 schrieb Rico Rommel:
> 
>> Am Donnerstag, 16. Januar 2014, 12:08:07 schrieb Dr. H. Nikolaus Schaller:
>>> Am 16.01.2014 um 11:42 schrieb Radek Polak:
>>>> On Thursday, January 16, 2014 11:21:39 AM Dr. H. Nikolaus Schaller wrote:
>>>>> Ok, please help to fix the power management (in the latest kernel
>>>>> because
>>>>> everything else makes no sense to me and is unlikely to get outside
>>>>> support).
>>>> 
>>>> I dont know how good/bad is PM in 3.12 kernel. But IIRC last time it was
>>>> much worse then 3.7. So the starting point could be to compare 3.7 and
>>>> 3.12 and get the PM to the level it is in 3.7.
>>> Hm. I think the starting point should be to check if your assumption that
>>> 3.12 is worse is still correct with the 3.12.7 kernel. Also because in the
>>> 3.12.7 kernel we have included all the 3.7 kernel patches.
>>> 
>>> There are three potential reasons:
>>> a) at the time you tested there might have been a kernel bug, that may have
>>> been fixed by now (or may be still be open) b) we may have damaged one of
>>> the 3.7 improvements while merging/rebasing c) someone (in the Linux kernel
>>> community) has rewritten a kernel driver that we use and that now draws
>>> more mA.
>>> 
>>> Most likely it is a single subsystem that does not work correctly any more.
>>> 
>>> But without testing we will never know if it is really worse.
>> 
>> I think, it's even worse than Radek says.
>> 
>> Last Weekend i played around with kernel 3.12.7 and did a trip to make some 
>> GPS-tracks. It started with approximately 200 mA and increased within 3 hours 
>> to > 400 mA. The increase is also be observed with kernel 3.7, but not in this 
>> extend.
>> 
>> Btw, after copying the modules to the phone, i forgot to build the module-
>> deps, so no modules were loaded at boot time. Power consumption was < 100 mA 
>> in idle with display on. After loading modules and restarting fso-daemons, 
>> current increased to > 200 mA.
> 
> Ok.
> 
>> It was  reproducable multiple times after removing battery before powering on.
>> 
>> While starting fsodeviced + fsousaged is powering devices on and off, so 
>> something seems to not power down correctly.
> 
> Can you better find out what the "something" is?
> 
>> Another problem is the camera driver. if it's loaded cpu doesn't enter higher 
>> C-States.
> 
> I think the 3.7 has no camera driver at all and I think nobody has analysed the
> camera driver for power management. The V4L2 is a quite complex beast...
> 
> But the camera driver chip is disabled in suspend (there are some kernel
> log messages telling about reset/initialization).

What could be is that the ISP (Image Signal Processor) in the SoC is not
suspending clocks and power.

> 
>> It was already mentioned here:
>> http://lists.goldelico.com/pipermail/gta04-owner/2013-November/005152.html
>> 
>> There was a userspace problem, but I have no application, which is accessing 
>> the camera.
>> 
>> Another difference from 3.12 to 3.7 is, that there is a new itg3200 driver that 
>> can't be turned off by setting sampling rate to 0. Maybe it keeps cpu from idle 
>> by delivering constantly data.
> 
> Ok, this all are topics that can be addressed. But please don't expect that they
> solve magically.
> 
> Patches are welcome!
> 
> BR,
> Nikolaus
> 
> _______________________________________________
> Community mailing list
> Community at openphoenux.org
> http://lists.goldelico.com/mailman/listinfo/community
> http://www.openphoenux.org




More information about the Community mailing list