Re: [RFC 0/4] orinoco: use cfg80211 for key manipulation

[Dan Williams <dcbw@redhat.com>]

On Fri, 2009-08-07 at 22:38 +0100, Dave wrote:
> Dan Williams wrote:
> > On Fri, 2009-08-07 at 21:42 +0100, Dave wrote:
> >> David Kilroy wrote:
> >>> First attempt at connect/disconnect, join_ibss/leave_ibss, and add_key
> >>> etc. These patches rely on the series I just sent, but conflict with
> >>> Holgers set_channel patch. Simple to resolve, and I'll do that in my
> >>> next update.
> >>>
> >>> This doesn't work yet*, but I've a number of questions.
> >> FYI, I've got this kinda working with wpa_supplicant in wext mode.
> >>
> >> First: on it's own wpa_supplicant keeps doing the 4 way handshake, while
> >>  cfg80211 pumps out the following:
> >>
> >> ------------[ cut here ]------------
> >> WARNING: at net/wireless/sme.c:451 __cfg80211_roamed+0x9a/0x1a5 [cfg80211]()
> >> Hardware name: Latitude C600
> >> Modules linked in: orinoco_cs orinoco cfg80211 michael_mic snd_maestro3
> >> aty128fb [last unloaded: cfg80211]
> >> Pid: 5, comm: events/0 Tainted: G   M    W  2.6.31-rc4-wl #87
> >> Call Trace:
> >>  [<c101ca51>] warn_slowpath_common+0x60/0x90
> >>  [<c101ca8e>] warn_slowpath_null+0xd/0x10
> >>  [<d1521659>] __cfg80211_roamed+0x9a/0x1a5 [cfg80211]
> >>  [<d15139a9>] cfg80211_event_work+0x104/0x1bc [cfg80211]
> >>  [<c102b39c>] worker_thread+0x15e/0x208
> >>  [<d15138a5>] ? cfg80211_event_work+0x0/0x1bc [cfg80211]
> >>  [<c102e5d5>] ? autoremove_wake_function+0x0/0x33
> >>  [<c102b23e>] ? worker_thread+0x0/0x208
> >>  [<c102e261>] kthread+0x66/0x6b
> >>  [<c102e1fb>] ? kthread+0x0/0x6b
> >>  [<c1003053>] kernel_thread_helper+0x7/0x10
> >> ---[ end trace 12e5dbe2024c37ae ]---
> >>
> >> Cfg80211 is trying to select the bss we've authenticated against.
> >> Unfortunately the BSS list is empty.
> >>
> >> If before starting wpa_supplicant I run:
> >>
> >> iw dev eth1 scan ssid <SSID>
> >>
> >> Then wpa_supplicant connects fine. The AP in question is configured with
> >> a hidden SSID.
> >>
> >> Any suggestions as to how I need to address this?
> > 
> > Do a scan before you try to associate?  Most fullmac cards do one
> > internally in the firmware before they try to associate anyway, but they
> > may not emit the event to the driver.  If you're not doing a scan when
> > you get the associate request, you probably should do one, let cfg80211
> > cache the bss list, and then you're fine, right?
> 
> That should work. I just realised why it's doing this. I have
> wpa_supplicant configured with ap_scan=2, so it doesn't scan before
> connecting.

Yeah, Don't Do That :)

You shouldn't really ever use ap_scan=2 unless you want IW_MODE_MASTER
or IW_MODE_IBSS.  Think of the children, and let the supplicant do the
work.

Dan

> Just tried wpa_supplicant in nl80211 mode. In that case, it's stuck in
> scanning (presumably ignoring my ap_scan setting). It's not seeing my AP
> (because it's hidden?), and so doesn't proceed. And I can't seem to get
> it to do an active scan for my AP.
> 
> 
> Dave.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html