aggressive suspend and yum...

[Paul Fox]

chris wrote:
 > Hi,
 > 
 >    > Now that we can use aggressive suspend and wifi doesn't die, it
 >    > is great. Battery lasts, it suspends opportunistically, it's
 >    > excellent.
 > 
 > Good to hear.

which platform are we talking about?  xo-1.5, i'll assume.

 > 
 >    > Until you want to run yum. Or download something from Firefox,
 >    > wget or Browse.xo.
 > 
 > We should be waking on LAN activity.  Could you file a bug with a
 > testcase that results in the laptop staying in suspend when it should
 > be in the middle of a download?  Maybe Marvell can help us find out
 > what's up.

wake-on-lan doesn't keep you from suspending -- it just brings you
back.

oh -- and wake-on-arp is missing, so if your connection is quiet for
long enough, it's not restartable.  (at least, "ethtool wol ua"
doesn't work.)

 > 
 >    > Do we have a workaround for that? The commandline tools could be
 >    > sorted with a "nosuspend" wrapper. Browse.xo and FF are a bit
 >    > harder.
 > 
 > Agreed on the "nosuspend" wrapper for yum.  Whenever we think about

powerd comes (will come) with an "olpc-nosleep" wrapper script.  (i
find myself using "olpc-nosleep sleep 3600 &" when i want some undisturbed
commandline time.

 > telling the system something manually, though, we should also think
 > about whether it's possible for it to infer it automatically; I think
 > that's possible here.  You probably don't want the laptop to suspend
 > if you're in the middle of a sustained download, so perhaps we should
 > just teach it what a sustained download looks like.
 > 
 > Does anyone have a favorite method for returning the current
 > throughput on an interface?

yes, please.  at a former job, we implemented an iptables rule
which did nothing but count specific kinds of traffic, and used
those counts.  (well, we used them to know whether to flicker
an application-specific activity light, but it's the same idea.)

paul

 > 
 >    > Long term I guess the use case warrants a periodic wakeup to soak
 >    > up the pending packets from the AP (of from the Libertas device
 >    > if it's been awake). That's probably easier said than done...
 > 
 > Again, or just inhibiting suspend in the first place if the current
 > network transfer rate is significant.  We shouldn't need this wakeup,
 > though; the module already knows how to wake us up when there are new
 > packets for us, and should already be doing so.
 > 
 > Thanks,
 > 
 > - Chris.
 > -- 
 > Chris Ball   <cjb at laptop.org>
 > One Laptop Per Child
 > _______________________________________________
 > Devel mailing list
 > Devel at lists.laptop.org
 > http://lists.laptop.org/listinfo/devel

=---------------------
 paul fox, pgf at laptop.org