Re: Howto test OFF Mode Support

[Kevin Hilman <khilman@deeprootsystems.com>]

"Sriram V" <vshrirama@gmail.com> writes:

> Hi,
>    I am using I am using the pm-20081106 on omap3evm.
>    I am trying to test if various pwr domains are set to OFF State.
>    Basically i want to test if various power domains (mpu/core) are being
>    set to different states
>     i see that core/mpu domain does not enter OFF/RET state.
>
>     $ /debug/pm_debug # cat count
>       usbhost_pwrdm (OFF),OFF:1,RET:1,INA:0,ON:1
>      sgx_pwrdm (OFF),OFF:1,RET:1,INA:0,ON:1
>      per_pwrdm (ON),OFF:0,RET:0,INA:0,ON:1
>      dss_pwrdm (OFF),OFF:1,RET:1,INA:0,ON:1
>      cam_pwrdm (OFF),OFF:1,RET:1,INA:0,ON:1
>      core_pwrdm (ON),OFF:0,RET:0,INA:0,ON:1
>      neon_pwrdm (ON),OFF:0,RET:0,INA:7646,ON:7647
>     mpu_pwrdm (ON),OFF:0,RET:0,INA:7646,ON:7647
>     iva2_pwrdm (OFF),OFF:1,RET:1,INA:0,ON:1
>
>    I notice that with pm support enable. The entire system is somewhat slow.
>    uart responses are very slow. has anyone observed this before?

Yes, UART response will be slow.  In order to hit retention or OFF,
UART clocks must be disabled.  So the PM branch has a set of patches
which disable the UART clocks on idle.

A side effect of this is that using the UART as a wakeup source, the
first character (which caused the wakeup) is lost.

>    Are these states depended on clocks on/off? ie if i put the
>    states to OFF/RET should i take care of the clocks too?

Clocks must all be off in order for you to hit these states.

For starters, I suggest you test with an extremely minimal defconfig,
where no drivers are enabled.  Any driver that does not disable its
clocks will prevent the system from hitting retention or off.

>     On running powertop, I see states upto C4 only.
>    i can see changes across C0, C2, C4 states.
>
>     PowerTOP version 1.9       (C) 2007 Intel Corporation
>
>       Cn                Avg residency       P-states (frequencies)
>       C0 (cpu running)        ( 0.2%)
>       C1                0.0ms ( 0.0%)
>       C2                4.8ms ( 0.3%)
>       C3                7.3ms ( 0.5%)
>       C4                0.0ms ( 0.0%)
>
>      Wakeups-from-idle per second :  3.0     interval: 3.0s
>
>    I tried to set sysfs entry enable_off_mode = 1, I can see that some
> of the domains
>    are set to OFF mode through debug print. But not core/mpu domains.

I would guess because you have some clocks still active.  Please
try my suggestion of a minimal defconfig.

>    Is there any reason why off mode can be set manually and not
> through cpu idle?

The 'enable_off_mode' flag does not set the powerdomain to off.  It
just allows the various power domains to go to OFF.  The rest of the
conditions (module idles, clocks disabled etc.)  must also be true for
a powerdomain to enter off.

>    has anyone tried this with pm-20081106.

We have this working on both omap3430SDP and custom hardware.

Kevin

>    Are there any utilities/methods to test cpu idle/pm in general other than
>    analyzing these debug prints.
>
>    am looking at power tester.
>
> Regards,
> sriram
> --
> To unsubscribe from this list: send the line "unsubscribe linux-omap" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html