Re: Trouble waking from sleep

[Dan Williams <dcbw redhat com>]

On Thu, 2009-05-14 at 15:01 -0400, Dan Williams wrote:
> On Thu, 2009-05-14 at 16:25 +0200, Christopher Lang wrote:
> > Hi,
> > 
> > I am seeing a similar behavior with NM (or dbus) in general, regardless of the 
> > type of network being used:
> > 
> > Under Ubuntu 8.10:
> > 
> > In the file:
> > /usr/lib/pm-utils/sleep.d/10NetworkManager
> > 
> > a signal is sent to NM via dbus when the PC wants to go to sleep. I can see, 
> > that this signal is properly delivered to NM, when going to suspend.
> > 
> > However, when coming back from suspend, although the 
> > log in /var/log/pm-suspend.log clearly shows that the 10NM... script was 
> > executed properly no "wake" signal is delivered to NM via dbus (at least not 
> > always).
> > 
> > This is strange, because one would assume that after comming back from suspend 
> > dbus is up and operational immediately; here in this case it looks that the 
> > dbus-send in 10NM... is not always going through to NM in the case of resume.
> > 
> > I belive in these cases when NM is not waking up properly it is necessary to 
> > debug signal delivery to NM from the 10NetworkManager script via dbus-send.
> 
> Guys, do some searching on the list.  This problem has been covered over
> and over again.

As in:

"Wireless is disabled message"  (May 11 2009)
"nm-applet loosing state" (May 8 2009)
"Network disabled after suspend/resume - sometimes" (Jan 7 2009)

https://bugzilla.redhat.com/show_bug.cgi?id=477964


> The problem is a bug in dbus where the bus ignores messages from very
> short-lived processes.  If the process (liek dbus-send) exits before
> dbus has had a chance to process the message, there's a chance the
> message gets dropped.
> 
> That's why --print-reply works; it forces the dbus-send to stick around
> longer, and the sleep/wake to NM doesn't get dropped.
> 
> NM prints out messages when it gets sleep/wake calls, and it will dump
> them to syslog.  If you don't see "waking up..." from NM, then NM didn't
> get the signal.
> 
> Dan
> 
> > Does NM provide some log where we can see the delivery of "sleep" and "wake" 
> > (or Sleep(s) with new API) as debug output from NM? That would come in handy.
> > 
> > thanks
> > 
> > Chris
> > 
> > http://www.acurana.de/
> > 
> > 
> > 
> > On Thursday 14 May 2009 15:28:01 Neal Becker wrote:
> > > On Monday 11 May 2009, Dan Williams wrote:
> > > > On Fri, 2009-05-08 at 18:47 -0400, Neal Becker wrote:
> > > > > NetworkManager-0.7.0.99-5.git20090326.fc10.x86_64
> > > > >
> > > > > After being connected to wired enet, going to sleep, and waking without
> > > > > wired enet, results are random.  Sometimes wlan is connected
> > > > > seemlessly, other times not.
> > > >
> > > > What wifi?  After resume, are there APs in the menu?
> > > >
> > > > Dan
> > > >
> > > > > In those cases restarting NetworkManager does not fix anything.
> > > > >
> > > > > Logging out/in always fixes it.
> > >
> > > Actually, it doesn't work correctly even on wired lan.  I just tried with
> > > NetworkManager-0.7.1-1.fc10.x86_64
> > > kde-plasma-networkmanagement-0.1-0.11.20090504svn.fc10.x86_64
> > >
> > > When it woke, only loopback IF was configured.  In this case, restarting NM
> > > did fix it.
> > >
> > > Looking at log, it is clear what the problem is.  There are no messages
> > > from NM from the time of wakeup until I restarted it!
> > > _______________________________________________
> > > NetworkManager-list mailing list
> > > NetworkManager-list gnome org
> > > http://mail.gnome.org/mailman/listinfo/networkmanager-list
> > 
> > _______________________________________________
> > NetworkManager-list mailing list
> > NetworkManager-list gnome org
> > http://mail.gnome.org/mailman/listinfo/networkmanager-list
> 
> _______________________________________________
> NetworkManager-list mailing list
> NetworkManager-list gnome org
> http://mail.gnome.org/mailman/listinfo/networkmanager-list