Re: [RFC] dynamic device power management proposal

[David Brownell <david-b@pacbell.net>]

On Wednesday 21 March 2007 2:39 pm, Pavel Machek wrote:

> > > USB gets the dependencies right, just copy that.
> > Does USB include all kinds of dependence, eg non parent-children
> > dependence?
> 
> No, probably not, as USB was designed properly.

It was designed to work with as few wires as possible:  "minimalist"
design.  That's not the same as "proper"; not everything works with
the same design constraints.


> How common are those cross dependencies?

Common.  SOC audio setups are relatively simple examples, which
I know have been making trouble on Linux for some time.  One hopes
that the new ALSA SOC stuff helps sort this out:

  - SOC serial controller (I2S, AC97, McBSP, SII, etc) manages
    encoded data bitsteams (e.g. N channels, PCM or uLaw, etc)
    for input, output, or both.

  - External codec interprets those bitstreams.

  - That codec is controlled using I2C, SPI, or some other bus
    for control, e.g. "set volume", "enter stereo mode".

  - The codec and serial controllers have separate power controls.

Think of this as separate control, data, and power channels;
which are not multiplexed like USB.

At the hardware level it probably looks like two different driver
model devices (I2S/etc, and codec) that could have semi-generic
drivers (I think this is what the ALSA SOC stuff is aiming for),
plus board-specific "power codec on/off" interfaces.  Then there
are the /dev interfaces ALSA creates for each functionality.


> Do we really want to solve them generically? 

What we need to be generic is:  PM not preventing good solutions.
I'm not sure we're there yet.

- Dave