IFF_RUNNING and hostapd

[David Sommerseth]

On 26/10/10 15:38, Jouni Malinen wrote:
> On Tue, Oct 12, 2010 at 01:24:00PM -0700, Dave Taht wrote:
>> On Tue, Oct 12, 2010 at 10:14 AM, Jouni Malinen <j at w1.fi> wrote:
>>> On Fri, Oct 01, 2010 at 03:32:42PM -0700, Dave Taht wrote:
>>>> I recently configured an openwrt system (a nanostation M) as a pure
> 
>>> Which driver are you using for wlan0?
>>
>> ath9k with nl80211
> 
> Ah.. I finally figured out what was happening. Something else in the
> system is setting the interface to IF_OPER_DORMANT state and that
> prevents IFF_RUNNING from showing up. In normal case, this would not
> happen, but it looks like the multi-call hostapd/wpa_supplicant in
> OpenWRT triggers this. I was able to reproduce this by forcefully
> killing wpa_supplicant so that it does not have chance of cleaning up
> the operstate change (which is needed for station mode).
> 
>>> hostapd does not control this and the exact behavior is likely to depend
>>> on which driver is being used.
>>
>> What does? My thought was basically IFF_RUNNING should be set
>> somewhere between these steps on a non-bridged AP, for the first
>> client that comes online - and unset after the last client goes
>> offline.
> 
> IFF_RUNNING is set in the kernel; operstate DORMANT/UP can be set in
> user space. hostapd never set this to DORMANT, so there was normally no
> need for setting it to UP either. However, since other applications may
> leave the interface to DORMANT operstate, I now added code in hostapd to
> clear that to avoid this issue. The change in the development branch is
> here:
> http://w1.fi/gitweb/gitweb.cgi?p=hostap.git;a=commitdiff;h=e11f5a2cbc333113a3a1cc1aeea7f698c3936ca3

I've just tried to compile the updated SVN tree for OpenWRT (backfire,
r23644) and it works just as expected now.  The network interface is in
UP/RUNNING state and radvd can do it's job properly.

Thanks a lot for your promptly response!


kind regards,

David Sommerseth