Bug#403045: [pkg-wpa-devel] Bug#403045: wpasupplicant: race condition in wpa_action disconnect between ifdown and if_post_down_up

[Kel Modderman]

On Friday 15 December 2006 18:19, Henning Glawe wrote:
> On Fri, Dec 15, 2006 at 01:33:12PM +1000, Kel Modderman wrote:
> > > the problem is, at least in my case, that the link is _down_
> > > afterwards, so wpasupplicant is not able to scan further. This seems to
> > > be caused by the interface not being completely downed yet when ifdown
> > > finishes; putting a 'sleep 1' into functions.sh::ifdown immediately
> > > after the call to /sbin/ifdown "solves" this problem.
> >
> > Yeah, I'm not a fan of unconditional sleeps though.
>
> me neither, thats why I set the word solves in doublequotes ;)
>
> > Do you have iproute installed? That slighly changes behaviour of
> > if_post_down_up, causing the interface to be flushed before 'upping' it
> > again.
>
> iproute was installed when I discovered the race.
>
> > In any case, ifdown should not exit until it has fully completed what it
> > had to do with the interface, at least in my opinion.
>
> I fully agree; I was also quite surprized about this fact when I did the
> research for this report.

What is output of `wpa_cli -i $iface status' when this condition results?

(you may have to provide -p /path/to/ctrl/interface if you use a custom 
ctrl_interface)

Thanks, Kel.