#10363 NORM 10.1.3: Auto-Suspend gets in the way when sharing over Salut

[Zarro Boogs per Child]

#10363: Auto-Suspend gets in the way when sharing over Salut
---------------------------------------+------------------------------------
           Reporter:  erikos           |       Owner:  erikos                           
               Type:  defect           |      Status:  new                              
           Priority:  normal           |   Milestone:  10.1.3                           
          Component:  telepathy-salut  |     Version:  Development build as of this date
         Resolution:                   |    Keywords:                                   
        Next_action:  diagnose         |    Verified:  0                                
Deployment_affected:                   |   Blockedby:                                   
           Blocking:                   |  
---------------------------------------+------------------------------------

Comment(by erikos):

 Replying to [comment:8 pgf]:
 >
 > i thought i had mentioned this earlier on this ticket, but it must have
 been in a different forum:  powerd uses iptables rules to do crude
 monitoring of type and frequency of traffic on the network in order to
 inhibit suspend.  (see the function init_netactivity_tracking() in
 /usr/sbin/powerd.)  (and remember, this is about inhibiting suspend, not
 about causing wakeups.)
 >
 > if certain types of traffic occur within 5 seconds of the anticipated
 suspend time, it won't suspend. in particular, powerd currently looks for
 incoming ICMP pings, incoming TCP traffic for already connected sessions,
 and for any outgoing TCP or UDP traffic at all, _except_ for MDNS traffic.
 these choices were partly tuned to letting the laptop go to sleep.  since
 the current problem is that it goes to sleep too much, perhaps some
 tweaking is in order.
 >
 > i don't have a suggestion for changing the behavior, but this might give
 those of you who better understand our sharing protocols an idea of what
 might be possible.

 Thanks Paul for those pointers. It is true that one can adjust those rules
 to get a similar behavior like inhibiting suspend from within Sugar when
 in sharing mode (like in my patch from above). I changed
 "init_netactivity_tracking" to monitor as well incoming and outgoing udp
 mdns packets which works well for inhibiting suspend during the sharing.

 I verified as well if the chatter of two machines affects another one that
 is in the same network. The third machine was able to suspend fine even
 though the other machines were talking.

 This is one part we need to fix, either way (not yet sure which one I
 prefer). The other blocker what worries me more at the moment is that we
 need to sync the state when we come out from resume. As we do not wakeup
 on multicast packets directed to us this is essential to provide a solid
 sharing experience. As the chances to solve this low level are rather
 small I think it involves
 [http://lists.freedesktop.org/archives/avahi/2010-October/001939.html
 pinging Avahi when we resume].

-- 
Ticket URL: <http://dev.laptop.org/ticket/10363#comment:9>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system