Re: [Madwifi-devel] sta association problem with latest revision (r1518)

[Daniel W. <dy...@gm...>]

here's a new patch, to try, instead of allowing the setparam to not run int=
o
the state machine (i think that's whats happening previously), we should
make it scan for the AP again, and associate that way.

Care to try and test it ?


Index: net80211/ieee80211_wireless.c
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
--- net80211/ieee80211_wireless.c    (revision 1521)
+++ net80211/ieee80211_wireless.c    (working copy)
@@ -371,7 +371,7 @@

     retv =3D ifmedia_ioctl(vap->iv_dev, &ifr, &vap->iv_media, SIOCSIFMEDIA=
);
     if (retv =3D=3D ENETRESET)
-        retv =3D IS_UP_AUTO(vap) ? ieee80211_open(vap->iv_dev) : 0;
+        retv =3D IS_UP_AUTO(vap) ? -ieee80211_init(vap->iv_dev, 0) : 0;
     return -retv;
 }

@@ -2320,12 +2321,12 @@
      */
     if (!vap->iv_xrvap || (vap->iv_xrvap && !(vap->iv_flags &
IEEE80211_F_XR))) {
         if (retv =3D=3D ENETRESET)
-            retv =3D IS_UP_AUTO(vap) ? ieee80211_open(vap->iv_dev) : 0;
+            retv =3D IS_UP_AUTO(vap) ? -ieee80211_init(vap->iv_dev, RESCAN=
) :
0;
     }
 #else
     /* XXX should any of these cause a rescan? */
     if (retv =3D=3D ENETRESET)
-        retv =3D IS_UP_AUTO(vap) ? ieee80211_open(vap->iv_dev) : 0;
+        retv =3D IS_UP_AUTO(vap) ? -ieee80211_init(vap->iv_dev, RESCAN) : =
0;
 #endif
     return -retv;
 }
Index: net80211/ieee80211_proto.c
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
--- net80211/ieee80211_proto.c    (revision 1521)
+++ net80211/ieee80211_proto.c    (working copy)
@@ -967,9 +967,9 @@
              * real soon now, so we don't muck with the state
              * machine.
              */
-            if (forcescan)
+            if (vap->iv_state !=3D IEEE80211_S_RUN || forcescan)
                 ieee80211_new_state(vap, IEEE80211_S_SCAN, 0);
-            else if (vap->iv_state =3D=3D IEEE80211_S_RUN)
+            else
                 ieee80211_new_state(vap, IEEE80211_S_ASSOC, 1);
         } else {
             /*




On 4/20/06, Brian Eaton <eat...@gm...> wrote:
>
> On 4/20/06, Daniel Wu <dy...@gm...> wrote:
> > Here's a patch to try, I'm not sure if it'll work, but i think it
> should.
> >  Try two scenarios.
> >  One with ifconfig down first then set the setparam stuff.
> >
> >  The other is to have ifconfig up then set the setparam stuff.
>
> I gave that patch a spin.  It worked OK for the manual configuration
> case, but when I tried to use it with wpa_supplicant my machine hung.
> The problem didn't look like a kernel panic, there were no blinking
> caps lock lights.  I think the driver and wpa_supplicant may have
> gotten themselves into a nice tight loop somehow.
>
> I tried r1518 without the patch, and that worked fine for the
> wpa_supplicant case, but was broken for the manual configuration case.
> So the good news is that you aren't the only one who can see this
> problem.  I'll meditate on the mysteries of the state machine a bit
> this weekend, to see if I can understand what is going on here.
>
> At this point I'm leaning towards breaking ieee80211_ioctl_setparam up
> into some smaller pieces and avoiding the call to ieee80211_open.  If
> nothing else I think it will make the state machine transitions easier
> to debug.
>
> Regards,
> Brian
>