[ltp] Re: Getting ipw3945 kill switch to bring eth1 down?

[Theodore Tso]

On Mon, Nov 19, 2007 at 10:51:06AM -0500, Stefan Monnier wrote:
> > That said, I usually down the interfaces and remove the device driver
> > modules for stuff I am not using when on battery, because the kernel powers
> > down or deactivates a lot of stuff when we do that, and that saves a
> > noticeable ammount of power.
> 
> What it's worth: my T60's ipw3945 chip seems to consume 0.7W *more* when
> the module is unloaded than when the interface is up and running in
> power_save mode 7.

Yeah, apparently some userspace programs expect that iwlist eth1 scan
be able to return accesspoint lists (seems like a broken assumption to
me) while the interface is down, and the iwl3945/iwl4965 drivers are
catering to that assumption.  (As far as I know this is not universal;
the Atheros drivers do not do this, for example.)  

In any case, it's actually running with the radios listening to every
single packet while the interface is down.  I'm guessing this actually
probably violates some FCC/FAA regulations since even if the radios
are in transmit-only mode, they are still generating potential RF
interference due to the Intermediate Frequency (IF) oscillators being
run.  I normally completely unload the iwl4965 driver before I get on
a plane, just for battery life reasons.  

I haven't had the time to go try to convince the
iwl4965/iwl3945/ipw3945 device driver maintainers of this, or to hack
in the correct fix myself, but the workaround (rmmod the driver) is
easy enough so it isn't a major tragedy.

  	 					- Ted


P.S.  For this and other details I found when I did an intesive, "how
low can you go" effort on my x61s, please see:

http://thunk.org/tytso/blog/2007/10/29/tip-o-the-hat-wag-o-the-finger-linux-power-savings-for-laptop-users/

There are a number of helpful comments made by some blog readers as well.