[Gta04-owner] Now that 3.2-final is out...

[NeilBrown]

On Sun, 8 Jan 2012 16:56:03 +0000 Radek Polak <psonek2 at seznam.cz> wrote:

> On Friday 06 January 2012 13:01:49 NeilBrown wrote:
> 
> > As 3.3-rc1 approaches, the 'merge' branch may get unstable.  If you don't
> > want to help test that, feel free to use '3.2-gta04'.
> 
> I tested 3.2-gta04 quickly. I will do more tests probably tomorrow. For now 
> suspend seems to work fine, but after resume the display starts blinking and 
> the image shifts up and down. I can take a video if needed. I think i had the 
> same problem with previous rc too.

Hmmm... that is unfortunate, though it doesn't surprise me a whole lot.

I spent a while examining 'cpuidle' which is supposed to put the CPU in
low-power states when idle.  But it doesn't.

You can confirm by something like:

   grep . /sys/devices/system/cpu/cpu0/cpuidle/state?/usage

That will report how often it was in each state (I don't know what the units
are).  All states > 0 have a usage of 0.

The reason is that the UART loses characters when you drop to a lower power
state so it refuses to do it when the UART is active - which is always.

You can adjust this with the "sleep_timeout" setting - 1 per uart.  I don't
remember exactly where in sysfs this is:
   find /sys -name sleep_timeout


If you set all these to (say) 10 (seconds), then after 10 seconds of no uart
activity it will drop to lower power states.  I've measure an idle current
below 100mA.

However other things than the UART stop working.

One it the battery monitor (bq27000) - it just fails to report meaningful
values.
The other (and this is where you get to see the relevance of this diversion)
is the DSS - display sub system.  It reports something about losing sync and
gets into a loop where it keeps disabling and re-enabling the display.

(I only managed to measure current by getting the display to blank first).

Do maybe if you make sure to blank the display before suspending it might
wake up again properly after resume??


>                                     New regression is touchscreen - it discards 
> like 50% of touch events (in qtmoko). I will check in console is there is some 
> data on touchscreen's /dev/input/event.

Cannot think of anything that was changed there..

> 
> I also plan to measure current with ammeter - it's cool that it now eats < 
> 200mA, because i can now use 200mA range on my ammeter.

Remember that the 200mA is only when idle.  It needs more to boot or do
anything useful.

Thanks,
NeilBrown
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 828 bytes
Desc: not available
URL: <http://lists.goldelico.com/pipermail/gta04-owner/attachments/20120109/e5c2f64b/attachment.bin>