Re: powersaving when interface is up but not associated

From: Kalle Valo <kalle.valo@iki.fi>
To: Johannes Berg <johannes@sipsolutions.net>
Cc: linux-wireless <linux-wireless@vger.kernel.org>,
	Marcel Holtmann <marcel@holtmann.org>, "Guy\,
	Wey-Yi W" <wey-yi.w.guy@intel.com>,
	Arjan van de Ven <arjan@linux.intel.com>
Subject: Re: powersaving when interface is up but not associated
Date: Thu, 23 Apr 2009 11:00:44 +0300	[thread overview]
Message-ID: <87vdovbxzn.fsf@litku.valot.fi> (raw)
In-Reply-To: <1240431356.30082.18.camel@johannes.local> (Johannes Berg's message of "Wed\, 22 Apr 2009 22\:15\:56 +0200")

Johannes Berg <johannes@sipsolutions.net> writes:

> Hi!

Hello,

> Just had this idea -- when our networking interfaces are UP, but not
> actually doing anything, the RF chains consume power. 

Yes, and also firmware consumes power. For example stlc4550 has a
hibernate feature to put the firmware sleep and wake it up quickly.

> This is, often, the default state, since you'll have your wifi on due
> to NM wanting to find APs, but not associated, for example when on the
> road. Sure, you can manually turn it off, but...

That's what we do in Nokia Internet tablets. wlancond it putting
interface up and down all the time and it's a hack.

> However, it doesn't seem necessary. When we are not associated to an AP,
> and there are no extra interfaces in mac80211 active like monitor
> interfaces, it seems that we could very well turn off the radio while
> not scanning. It would be entirely trivial for mac80211 to do this,
> probably a 20 line patch or so (need some bookkeeping about why radio is
> disabled so it's not just four lines).
>
> Thoughts?

I think it's a good idea. Sometimes I have thought myself that we would
need something like this.

-- 
Kalle Valo