[pkg-wpa-devel] Bug#383907: Is there an easy way to check if an interface is managed by wpa_action?

[Reinhard Tartler]

Kel Modderman <kelrin at tpg.com.au> writes:

> On Monday 21 August 2006 00:02, Reinhard Tartler wrote:
>> This of course fails if the interface is controlled by wpa-roam aka
>> /sbin/wpa_action. I hardcased the check here for my interface, but it
>> would be nice if acpi-support could be improved and call `wpa_action $x
>> stop` by itself. For this we need to provide a reliable check if the
>> interface is managed by wpa_action or not.
>
> You can check for the location of the ctrl_interface domain socket (that may 
> be in a different location according to config file used) or pidfile (which 
> is what the wpa-ifupdown init script/hack uses to identify interfaces bound 
> by wpa_supplicant/wpa_cli, and it is a fixed location).

The ctrl_interface socket may be changed by the user, so this is a
nogo. The pid file is a fixed location, but might get stale (i.e. it is
still there, but the corresponding daemon has died or has been killed by
the user. Even checking if there was a PID with that number doesn't work
reliably, because there might be another process with the same PID).

How about this:

if [ -x /sbin/wpa_action && /sbin/wpa_action check $IFACE ]; 
   wpa_action ath0 down
else
   ifdown ath0
fi

> Is this a workaround for #383168? If so, my current kludge is to have the 
> suspend manager, in my case it was hibernate but now it is powersave, stop 
> the wpa-ifupdown service on suspend.

#383168 (list of optimized glibc packages in postinst script is
 outdated) seems unrelated to me. Was this a typo?

Your 'kludge' sounds interesting. How does it work with hibernate?
Perhaps we can write some hints in README.Debian (or README.suspending)

-- 
Gruesse/greetings,
Reinhard Tartler, KeyID 945348A4