n8x0 idle power problem

n8x0 idle power problem

[Felix Xiaozhu Lin]

I am trying to compile a linux-omap kernel for n810 that can do good
idle power management.

However, this seems to be harder than I expected. I've done all
following measurements with /sys/power/sleep_while_idle set to 1.
Compiler is arm-linux-gnueabi-gcc 4.2. I use n8x0_defconfig.

  * Maemo kernel (2.6.21) consumes ~0mA@3.8V when idle. This seems to
be perfect. However, it has no pm regarding serial port. Whenever
serial port is enabled, idle power is ~50mA@3.8V.

  * I've compiled 2.6.28/2.6.29 from l-o git. Serial port pm works
fine. However the idle power is ~20mA@3.8V. That is quite strange.

  * For kernel >= 2.6.30 I didn't manage to boot my N810.

I also dig out a thread discussing about N8x0 idle power:

http://www.mail-archive.com/linux-omap@vger.kernel.org/msg11954.html

where Kalle mentioned 2.6.29 idle power is OK. But I never get the same result.

I'm wondering if I did something wrong in the measurement. Anybody did
the same thing? You suggestion is highly appreciated.

Thanks,

-Felix

Re: n8x0 idle power problem

[Tony Lindgren]

* Felix Xiaozhu Lin <xzl@rice.edu> [091023 16:41]:
> I am trying to compile a linux-omap kernel for n810 that can do good
> idle power management.
> 
> However, this seems to be harder than I expected. I've done all
> following measurements with /sys/power/sleep_while_idle set to 1.
> Compiler is arm-linux-gnueabi-gcc 4.2. I use n8x0_defconfig.
> 
>   * Maemo kernel (2.6.21) consumes ~0mA@3.8V when idle. This seems to
> be perfect. However, it has no pm regarding serial port. Whenever
> serial port is enabled, idle power is ~50mA@3.8V.
> 
>   * I've compiled 2.6.28/2.6.29 from l-o git. Serial port pm works
> fine. However the idle power is ~20mA@3.8V. That is quite strange.
> 
>   * For kernel >= 2.6.30 I didn't manage to boot my N810.
> 
> I also dig out a thread discussing about N8x0 idle power:
> 
> http://www.mail-archive.com/linux-omap@vger.kernel.org/msg11954.html
> 
> where Kalle mentioned 2.6.29 idle power is OK. But I never get the same result.
> 
> I'm wondering if I did something wrong in the measurement. Anybody did
> the same thing? You suggestion is highly appreciated.

It sounds like something is blocking the deeper idle states.
The way to debug is to see what clocks and domains are on
before entering idle.

Maybe do a register dump on a working system and compare that
to the current system?

Regards,

Tony

Re: n8x0 idle power problem

[xl6]

By 'deeper-sleep state', you mean the full retention mode? I am sure  
that pm24xx.c:omap2_enter_full_retention() is called and the assembly  
code in sleep24xx.S is executed, by printing some debug info.

Thanks.

Best Regards,
--
Felix

On Tue, Nov 3, 2009 at 11:19 AM, Tony Lindgren <tony@atomide.com> wrote:

     * Felix Xiaozhu Lin <xzl@rice.edu> [091023 16:41]:
     > I am trying to compile a linux-omap kernel for n810 that can do good
     > idle power management.
     >
     > However, this seems to be harder than I expected. I've done all
     > following measurements with /sys/power/sleep_while_idle set to 1.
     > Compiler is arm-linux-gnueabi-gcc 4.2. I use n8x0_defconfig.
     >
     >   * Maemo kernel (2.6.21) consumes ~0mA@3.8V when idle. This seems to
     > be perfect. However, it has no pm regarding serial port. Whenever
     > serial port is enabled, idle power is ~50mA@3.8V.
     >
     >   * I've compiled 2.6.28/2.6.29 from l-o git. Serial port pm works
     > fine. However the idle power is ~20mA@3.8V. That is quite strange.
     >
     >   * For kernel >= 2.6.30 I didn't manage to boot my N810.
     >
     > I also dig out a thread discussing about N8x0 idle power:
     >
     > http://www.mail-archive.com/linux-omap@vger.kernel.org/msg11954.html
     >
     > where Kalle mentioned 2.6.29 idle power is OK. But I never get  
the same result.
     >
     > I'm wondering if I did something wrong in the measurement. Anybody did
     > the same thing? You suggestion is highly appreciated.

     It sounds like something is blocking the deeper idle states.
     The way to debug is to see what clocks and domains are on
     before entering idle.

     Maybe do a register dump on a working system and compare that
     to the current system?

     Regards,

     Tony