Powering down a device...

Powering down a device...

[Matthew Percival]

[-- Attachment #1: Type: text/plain, Size: 1099 bytes --]

G'Day,

	I was directed here from the linux-omap-open-source mailing list.

	I am using an OMAP5912, and have been looking at what power savings I
get when using Deep Sleep (echo -n "standby" > /sys/power/state), and
noticed that there is no change in vmain.  I am hoping that turning the
Ethernet controller off before entering deep sleep will make a
significant difference, but I do not know how to do this.

	If I look in drivers/net/smc91x.c, there is already a suspend function
written, which apparently powers the device down.  I do not believe this
is currently being used anywhere, and I am not sure how exactly it is
supposed to be used to begin with.  I was hoping that someone here would
be able to give me some information on using device_driver->suspend()
functions.

	As an aside, is there anyone here who knows how to use Big Sleep in an
OMAP environment?  I have worked out how to configure it, but not how to
enter it.  As this is not an OMAP mailing list, I am not expecting
anyone here to know the answer to that one, but it cannot hurt to ask.

	Thanks for your time,

	Matthew


[-- Attachment #2: Type: text/plain, Size: 0 bytes --]

RE: Powering down a device...

[Matthew Percival]

[-- Attachment #1: Type: text/plain, Size: 1421 bytes --]

G'Day,

	Thanks for your reply.  I have tried your suggestions out, but have not
been able to conserve any additional power.

> For device sleeps you just do an 'echo mem > /sys/power/state'.  Pm.c
> will map which sleep and wake this takes you to.

	I just tried comparing `standby' with `mem', and there was no
difference in vmain --- or anywhere else for that matter, though I was
no expecting to see any change elsewhere.

> Each individual device will also have a power/state file in the sysfs
> directory structure.  You can echo a device power state down to the
> driver and it will react how if it knows how to enter that state.

	I tried `standby', `mem' and `disc'
with /sys/devices/platform/smc91x/power/state and saw no change; there
was nothing printed on the terminal either. I guess it must not be
implemented with this driver at the moment, which I guess is why there
is no change when I put the whole system to sleep (please correct me if
my assumptions are wrong).  If it is nothing too difficult to add to the
driver, would you be able to direct me to where I may find out how to do
this?

> Also, in sleeps where you want a good effect, you will need to configure
> the external pins to the optimal sleep state.  Any floating inputs will
> consume a lot of power.... how you drive the pin states really depends
> on what it is connected to.

	That sounds like a rather good tip, thanks!

	-- Matthew


[-- Attachment #2: Type: text/plain, Size: 0 bytes --]

Re: Powering down a device...

[Geoff Levand]

[-- Attachment #1: Type: text/plain, Size: 650 bytes --]

Matthew Percival wrote:

> 	I tried `standby', `mem' and `disc'
> with /sys/devices/platform/smc91x/power/state and saw no change; there
> was nothing printed on the terminal either. I guess it must not be
> implemented with this driver at the moment, which I guess is why there
> is no change when I put the whole system to sleep (please correct me if
> my assumptions are wrong).  If it is nothing too difficult to add to the
> driver, would you be able to direct me to where I may find out how to do
> this?

H. Kaminaga from Sony is maintaining some OSK 5912 PM support:

  http://lists.osdl.org/pipermail/linux-pm/2005-July/001077.html

-Geoff


[-- Attachment #2: Type: text/plain, Size: 0 bytes --]

RE: Powering down a device...

[Patrick Mochel]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 1900 bytes --]


On Mon, 21 Nov 2005, Matthew Percival wrote:

> G'Day,
>
> 	Thanks for your reply.  I have tried your suggestions out, but have not
> been able to conserve any additional power.
>
> > For device sleeps you just do an 'echo mem > /sys/power/state'.  Pm.c
> > will map which sleep and wake this takes you to.
>
> 	I just tried comparing `standby' with `mem', and there was no
> difference in vmain --- or anywhere else for that matter, though I was
> no expecting to see any change elsewhere.

The system power states depend on what the platform port has implemented.
It may be that the OSK has not implemented those states.

> > Each individual device will also have a power/state file in the sysfs
> > directory structure.  You can echo a device power state down to the
> > driver and it will react how if it knows how to enter that state.
>
> 	I tried `standby', `mem' and `disc'
> with /sys/devices/platform/smc91x/power/state and saw no change; there
> was nothing printed on the terminal either. I guess it must not be
> implemented with this driver at the moment, which I guess is why there
> is no change when I put the whole system to sleep (please correct me if
> my assumptions are wrong).  If it is nothing too difficult to add to the
> driver, would you be able to direct me to where I may find out how to do
> this?

Unfortunately, device power management is a bit incomplete (and a bit
rough). One would not write "standby", "mem", or "disk" to a device power
file, but rather a numeric value that is equal to the power state that the
device should enter, e.g. 1, 2, 3, etc.. (The fact that this numeric value
may not correlate to a power state relevant to that device was overlooked
by the designers of the generic interface).

It could also be that the smc91x does not support power management at all,
or that the right infrastructure is not in place for it to work.

Thanks,


	Pat


[-- Attachment #2: Type: text/plain, Size: 0 bytes --]