Re: reconnect after sleeping

[Dan Williams <dcbw redhat com>]

On Wed, 2009-08-19 at 20:58 +0100, Rick Jones wrote:
> --On Wednesday, August 19, 2009 14:43:36 -0500 Dan Williams
> <dcbw redhat com> wrote:
> 
> > On Wed, 2009-08-19 at 20:21 +0100, Rick Jones wrote:
> > > 
> > > It seems to me this would be a good opportunity to handle WOL
> > > properly. I've never been able to get WOL to work, and messing
> around
> > > with ethtool shouldn't be necessary anyway. It seems to be
> something
> > > that just isn't easy on Linux. If NM were to take control of
> > > suspending the network devices then it should be easy for it to
> set up
> > > the conditions for WOL to work.
> > > 
> > > Another good reason to press ahead with this?
> > 
> > I wouldn't expect WOL to work at the driver/hardware level before
> 2.6.32
> > anyway given that there's only recently been a lot of work on it
> > upstream.  I'm not sure of the specifics though, since it seems like
> > we'd need some sort of communication between the card/driver and
> > userspace to alert NM or the supplicant to *why* the machine was
> woken
> > up, and potentially what to do in response to that.
> > 
> > Dan
> 
> I wouldn't have thought reacting to a WOL wakeup would need to be any
> different from the normal power-button wakeup. The problem at the
> moment is that it doesn't seem possble to suspend the machine in such
> a way that it will wakeup at all in response to network activity -
> either magic packets or other. Is this what the 2.6.32 work is
> intended to solve?

Not really sure; the hardware needs to be put into special states and
IRQ lines set up and whatnot so that when the card does see the magic
packet it can raise the interrupt and then the machine wakes up.  That
requires a lot of driver and stack support that is only starting to land
in 2.6.32.  I'm not sure how far it is now, but I wouldn't expect
anything before 2.6.32 to work for anything WOL-related.

Dan