Re: "Repair" function

[David Zeuthen <davidz redhat com>]

On Mon, 2005-06-27 at 11:52 -0400, Derek Atkins wrote:
> *shrugs*  I can't defend their choice, but it makes sense if the standard usage
> is a quazi-static location.  MIT doesn't support mobility, per se, so to move
> from one location to another you need to re-run DHCP.  Most OSes rerun DHCP on
> "resume", but NM does seem to have a particular problem recovering from
> suspend/resume (is there some synchronous signal you can run to force NM into
> suspend mode?  dbus-send'ing the sleep call isn't synchronous).
> 
> But you're right -- just re-clicking the network would work.. BUT it would be
> nice to have better 'scripting' support for suspend/resume to force a new
> network discovery and re-running DHCP even if the lease is still valid and the
> SSID hasn't changed.

Actually the suspend scripts I'm looking to put in Fedora Core does
exactly this (RH #155613). Here's a snippet from suspend

    # Tell NetworkManager to shut down networking
    dbus-send --system                         \
        --print-reply                          \
        --reply-timeout=2000                   \
        --dest=org.freedesktop.NetworkManager  \
        /org/freedesktop/NetworkManager        \
        org.freedesktop.NetworkManager.sleep

and on resume

    # Wake up NetworkManager and make it do a new connection
    dbus-send --system                        \
        --dest=org.freedesktop.NetworkManager \
        /org/freedesktop/NetworkManager       \
        org.freedesktop.NetworkManager.wake

I did test this extensively at GUADEC (suspending and resuming all the
time) and it worked surprisingly well. There may be some bugs with the
current Rawhide tries now however, I have a recollection of this
failing. I could be wrong though.

    David