wpa_supplicant-0.6.1 IBSS Mode

[Dan Williams]

On Tue, 2008-04-29 at 16:23 +0200, Vladimir Koutny wrote:
> Hi Dan,
> 
> [...]
> 
> >>> I found in driver_wext.c in line 970 a instruction which makes in my mind no sence.
> >>>
> >>> "if (wpa_driver_wext_set_mode(drv, 0) < 0) {"
> >>>
> >>> Why is the driver always configured in managed mode? But I'm sure this is not the reason why WPA-NONE not work. ;-)
> >> That part shouldn't really matter, because it's only set to managed when
> >> the card is initialized.  The _real_ mode you request is set in the
> >> 'associate' hook below around line 1946.  That's the only mode set that
> >> really matters.
> > 
> > I agree with you.
> 
> I don't think so.
> 
> I'm just trying to get this working with driver_wext on a mac80211-based
> driver (wireless-testing version), which won't allow you to change the
> mode once the interface is up.

There's two ways to fix this then:

1) teach wpa_supplicant about the separate mac80211 master and STA
interfaces, which could be ugly, since the mac80211 bits are obviously
linux specific (although net80211 on BSD probably works like mac80211 as
well).

2) make mac80211 STA devices allow switching between master and adhoc
mode on the fly, if they don't.  I had thought they could.

Dan

> The result is that the mode is set to managed at startup (which is being
> done with the iface down - so even if you set it to ad-hoc manually before
> starting wpa_supplicant, it is forced back) and is kept here (that
> SET_MODE: output is mine, tested on 0.6.3):
> 
> 3054.032209: State: SCANNING -> ASSOCIATING
> 3054.032335: wpa_driver_wext_set_operstate: operstate 0->0 (DORMANT)
> 3054.032453: WEXT: Operstate: linkmode=-1, operstate=5
> 3054.032654: wpa_driver_wext_associate
> 
> ***************
> SET_MODE: mode 1
> 
> ioctl[SIOCSIWMODE]: Device or resource busy
> 3054.032309: Association request to the driver failed
> 
> 
> This renders IBSS support unusable to wext drivers that won't allow
> switching mode on-the-fly.
> 
> 
> Regards,
>   Vlado
>