#10058 NORM 1.5-sof: Wake-on-LAN does not work during downloads

[Zarro Boogs per Child]

#10058: Wake-on-LAN does not work during downloads
---------------------------------------+------------------------------------
           Reporter:  martin.langhoff  |       Owner:                     
               Type:  defect           |      Status:  new                
           Priority:  normal           |   Milestone:  1.5-software-update
          Component:  not assigned     |     Version:  not specified      
         Resolution:                   |    Keywords:                     
        Next_action:  never set        |    Verified:  0                  
Deployment_affected:                   |   Blockedby:                     
           Blocking:                   |  
---------------------------------------+------------------------------------

Comment(by pgf):

 i've been experimenting with a set of iptables rules, to allow us to
 prevent suspend based on counts of specific packets.  i think this
 approach would help work around the tcp stream cases we've seen (using
 wget, or ssh sessions, as in this bug), by usually preventing the
 problematic suspend in the first place.

 current rules i've been trying look like:

 {{{
 iface=eth0

 iptables -P INPUT ACCEPT
 iptables -P OUTPUT ACCEPT
 iptables -P FORWARD ACCEPT
 iptables -F

 iptables -N netactivity         # create the chain
 # iptables -I INPUT -p icmp -i $iface -j netactivity
 iptables -I INPUT -p tcp ! --syn -i $iface -j netactivity
 iptables -I OUTPUT -p tcp ! --syn -o $iface -j netactivity
 iptables -I netactivity
 }}}

 powerd then examines the resulting counts periodically, with something
 like:
 {{{
 iptables --list netactivity --zero --verbose --numeric
 }}}

 a non-zero count in the period leading up to wanting to suspend prevents
 the suspend.
 as i say, this is very experimental right now.

 i'm not sure how this might be applied to the collaboration traffic which
 is the subject of #9535, and which, of course, is the more interesting
 feature than tcp streams.

-- 
Ticket URL: <http://dev.laptop.org/ticket/10058#comment:5>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system