Return-path: Received: from mx1.redhat.com ([66.187.233.31]:56807 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752090AbYIOHZc (ORCPT ); Mon, 15 Sep 2008 03:25:32 -0400 Subject: Re: [PATCH] libertas: Reduce the WPA key installation time. From: Dan Williams To: Javier Cardona Cc: linux-wireless@vger.kernel.org, libertas-dev@lists.infradead.org In-Reply-To: <1221461196.27102.77.camel@localhost.localdomain> References: <48c9b342.081b600a.61cf.fffffdb5@mx.google.com> <1221461196.27102.77.camel@localhost.localdomain> Content-Type: text/plain Date: Mon, 15 Sep 2008 03:26:11 -0400 Message-Id: <1221463571.27102.94.camel@localhost.localdomain> (sfid-20080915_092536_039864_C5F80B2D) Mime-Version: 1.0 Sender: linux-wireless-owner@vger.kernel.org List-ID: On Mon, 2008-09-15 at 02:46 -0400, Dan Williams wrote: > On Thu, 2008-09-11 at 15:32 -0700, Javier Cardona wrote: > > WPA requires that the PTK is installed immediately after the 4-way handshake > > in order to properly decrypt the subsequent incoming EAPOL-GTK frame. If the > > PTK is not enabled by the time the EAPOL-GTK frame arrives, the frame is > > dropped and the supplicant does not receive the group key. > > > > This will happen with fast Access Points that send the EAPOL-GTK frame before > > the suplicant has successfully installed and enabled the PTK. To mitigate > > this situation, this patch simplifies and accelerates the SIOCSIWENCODEEXT > > execution. > > Good catch though this would break ordering of iwconfig commands like: > > iwconfig eth0 essid foobar key 235232363632 mode managed channel 11 > > which would punt the mode + channel arguments to another association > request, thus causing the first one to either fail (because the mode > wouldn't be switched yet) or to get the wrong AP on the wrong channel > (because the channel lock request wouldn't come until after it had tried > to associate on channels 1 - 10, and if it found a matching AP not on > channel 11 it would use that first). > > The problem is of course that the patch unconditionally sets the key > immediately, but that was also the solution to the OLPC bug. > Unfortunately this is something we can't really handle well with WEXT, > but we can try to work around it by testing the following: > > 1) in infrastructure mode AND one or both of (2) or (3) > 2) WPA is on > 3) key management (via SIWAUTH) is one of > a) IW_AUTH_KEY_MGMT_802_1X > b) IW_AUTH_KEY_MGMT_PSK > > and if this is true, set keys immediately, otherwise punt to the > association handler. > > Does something like the following (based off your patch) still fix the > bug? > > Dan > > diff --git a/drivers/net/wireless/libertas/dev.h b/drivers/net/wireless/libertas/dev.h > index fd59e18..6d2bf5c 100644 > --- a/drivers/net/wireless/libertas/dev.h > +++ b/drivers/net/wireless/libertas/dev.h > @@ -58,6 +58,7 @@ struct lbs_802_11_security { > u8 WPA2enabled; > u8 wep_enabled; > u8 auth_mode; > + u32 key_mgmt; > }; > > /** Current Basic Service Set State Structure */ > diff --git a/drivers/net/wireless/libertas/wext.c b/drivers/net/wireless/libertas/wext.c > index 426f1fe..43ba998 100644 > --- a/drivers/net/wireless/libertas/wext.c > +++ b/drivers/net/wireless/libertas/wext.c > @@ -30,6 +30,14 @@ static inline void lbs_postpone_association_work(struct lbs_private *priv) > queue_delayed_work(priv->work_thread, &priv->assoc_work, HZ / 2); > } > > +static inline void lbs_do_association_work(struct lbs_private *priv) > +{ > + if (priv->surpriseremoved) > + return; > + cancel_delayed_work(&priv->assoc_work); > + queue_delayed_work(priv->work_thread, &priv->assoc_work, 0); > +} > + > static inline void lbs_cancel_association_work(struct lbs_private *priv) > { > cancel_delayed_work(&priv->assoc_work); > @@ -1578,12 +1586,27 @@ static int lbs_set_encodeext(struct net_device *dev, > set_bit(ASSOC_FLAG_SECINFO, &assoc_req->flags); > } > > + /* Only disable wep if necessary: can't waste time here. */ > + if (priv->mac_control & CMD_ACT_MAC_WEP_ENABLE) > + disable_wep(assoc_req); > disable_wep (assoc_req); It's late; this second disable_wep() slipped through. Remove it of course :) > } > > out: > if (ret == 0) { > - lbs_postpone_association_work(priv); > + /* 802.1x and WPA rekeying must happen as quickly as possible, > + * especially during the 4-way handshake; thus if in > + * infrastructure mode, and either (a) 802.1x is enabled or > + * (b) WPA is being used, set the key right away. > + */ > + if (assoc_req->mode == IW_MODE_INFRA && > + ((assoc_req->secinfo.key_mgmt & IW_AUTH_KEY_MGMT_802_1X) || > + (assoc_req->secinfo.key_mgmt & IW_AUTH_KEY_MGMT_PSK) || > + assoc_req->secinfo.WPAenabled || > + assoc_req->secinfo.WPA2enabled)) { > + lbs_do_association_work(priv); > + } else > + lbs_postpone_association_work(priv); > } else { > lbs_cancel_association_work(priv); > } > @@ -1691,13 +1714,17 @@ static int lbs_set_auth(struct net_device *dev, > case IW_AUTH_TKIP_COUNTERMEASURES: > case IW_AUTH_CIPHER_PAIRWISE: > case IW_AUTH_CIPHER_GROUP: > - case IW_AUTH_KEY_MGMT: > case IW_AUTH_DROP_UNENCRYPTED: > /* > * libertas does not use these parameters > */ > break; > > + case IW_AUTH_KEY_MGMT: > + assoc_req->secinfo.key_mgmt = dwrq->value; > + updated = 1; > + break; > + > case IW_AUTH_WPA_VERSION: > if (dwrq->value & IW_AUTH_WPA_VERSION_DISABLED) { > assoc_req->secinfo.WPAenabled = 0; > @@ -1777,6 +1804,10 @@ static int lbs_get_auth(struct net_device *dev, > lbs_deb_enter(LBS_DEB_WEXT); > > switch (dwrq->flags & IW_AUTH_INDEX) { > + case IW_AUTH_KEY_MGMT: > + dwrq->value = priv->secinfo.key_mgmt; > + break; > + > case IW_AUTH_WPA_VERSION: > dwrq->value = 0; > if (priv->secinfo.WPAenabled) > > > -- > 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