[PATCH 0/8] staging/wlan-ng: apply changes from wlan-ng-devel that get the driver working

[PATCH 0/8] staging/wlan-ng: apply changes from wlan-ng-devel that get the driver working

[Richard Kennedy]

These changes from wlan-ng-devel get the card in my laptop working. 
All the patches are needed to get it fully working.
Patches are against 2.6.28-rc2 + latest wlan-ng cleanup-patches

My laptop can now connected to wep enable access point under the control
of NetworkManager

Regards
Richard


Richard Kennedy (8):
  p80211netdev.c fix netdev alloc to prevent oops on device start
  prism2_usb.c always enable the card in probe_usb
  hfa384x_usb.c use newest version of 384x_drvr_start
  p80211netdev.c correctly enable wext handlers
  p80211wext.c add latest changes & remove extra nulls from wext_handlers
  p80211wext.c don't set default key id twice
  hfa384x_usb.c hfa384x_usbin_callback: check for hardware removed
  p80211conv.c copy code from wlan-ng-devel branch to not drop packets 

 drivers/staging/wlan-ng/hfa384x_usb.c  |   70 +++++++++++++++++++++++++------
 drivers/staging/wlan-ng/p80211conv.c   |   49 ++++++++++++++++++++++-
 drivers/staging/wlan-ng/p80211netdev.c |   18 ++------
 drivers/staging/wlan-ng/p80211wext.c   |   44 +++++++++++---------
 drivers/staging/wlan-ng/prism2_usb.c   |   14 ++++--
 5 files changed, 142 insertions(+), 53 deletions(-)

[PATCH 1/8] p80211netdev.c fix netdev alloc to prevent oops on device start

[Richard Kennedy]

a version of this patch is in the wlan-ng_devel tree

initializes netdev correctly to prevent an oops on device start.  

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
---
 drivers/staging/wlan-ng/p80211netdev.c |   16 ++++------------
 1 files changed, 4 insertions(+), 12 deletions(-)

diff --git a/drivers/staging/wlan-ng/p80211netdev.c b/drivers/staging/wlan-ng/p80211netdev.c
index d229780..cc9cd4c 100644
--- a/drivers/staging/wlan-ng/p80211netdev.c
+++ b/drivers/staging/wlan-ng/p80211netdev.c
@@ -845,13 +845,11 @@ int wlan_setup(wlandevice_t *wlandev)
 		     (unsigned long)wlandev);
 
 	/* Allocate and initialize the struct device */
-	dev = kmalloc(sizeof(netdevice_t), GFP_ATOMIC);
+	dev = alloc_netdev(0,"wlan%d",ether_setup);
 	if ( dev == NULL ) {
 		WLAN_LOG_ERROR("Failed to alloc netdev.\n");
 		result = 1;
 	} else {
-		memset( dev, 0, sizeof(netdevice_t));
-		ether_setup(dev);
 		wlandev->netdev = dev;
 		dev->priv = wlandev;
 		dev->hard_start_xmit =	p80211knetdev_hard_start_xmit;
@@ -960,15 +958,9 @@ int register_wlandev(wlandevice_t *wlandev)
 
 	DBFENTER;
 
-	i = dev_alloc_name(wlandev->netdev, "wlan%d");
-	if (i >= 0) {
-		i = register_netdev(wlandev->netdev);
-	}
-	if (i != 0) {
-		return -EIO;
-	}
-
-	strcpy(wlandev->name, dev->name);
+	i = register_netdev(wlandev->netdev);
+	if (i)
+		return i;
 
 	DBFEXIT;
 	return 0;
-- 
1.5.6.5

[PATCH 2/8] prism2_usb.c always enable the card in probe_usb

[Richard Kennedy]

always enable card in probe_usb
& update register_wlandev to match latest wlan-ng-dev tree.

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
---
 drivers/staging/wlan-ng/prism2_usb.c |   14 +++++++++-----
 1 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/drivers/staging/wlan-ng/prism2_usb.c b/drivers/staging/wlan-ng/prism2_usb.c
index f1ae371..34a36c1 100644
--- a/drivers/staging/wlan-ng/prism2_usb.c
+++ b/drivers/staging/wlan-ng/prism2_usb.c
@@ -110,11 +110,6 @@ static int prism2sta_probe_usb(
 	 * linux netdevice.
 	 */
 	SET_NETDEV_DEV(wlandev->netdev, &(interface->dev));
-        if ( register_wlandev(wlandev) != 0 ) {
-		WLAN_LOG_ERROR("%s: register_wlandev() failed.\n", dev_info);
-		result = -EIO;
-		goto failed;
-        }
 
 	/* Do a chip-level reset on the MAC */
 	if (prism2_doreset) {
@@ -138,6 +133,15 @@ static int prism2sta_probe_usb(
 
 	wlandev->msdstate = WLAN_MSD_HWPRESENT;
 
+        if ( register_wlandev(wlandev) != 0 ) {
+		WLAN_LOG_ERROR("%s: register_wlandev() failed.\n", dev_info);
+		result = -EIO;
+		goto failed;
+        }
+
+/* enable the card */
+	prism2sta_ifstate(wlandev, P80211ENUM_ifstate_enable);
+
 	goto done;
 
  failed:
-- 
1.5.6.5

Re: [PATCH 2/8] prism2_usb.c always enable the card in probe_usb

[Oliver Neukum]

Am Montag, 3. November 2008 12:13:48 schrieb Richard Kennedy:
> always enable card in probe_usb
> & update register_wlandev to match latest wlan-ng-dev tree.

This breaks both error cases.

> ---
>  drivers/staging/wlan-ng/prism2_usb.c |   14 +++++++++-----
>  1 files changed, 9 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/staging/wlan-ng/prism2_usb.c b/drivers/staging/wlan-ng/prism2_usb.c
> index f1ae371..34a36c1 100644
> --- a/drivers/staging/wlan-ng/prism2_usb.c
> +++ b/drivers/staging/wlan-ng/prism2_usb.c
> @@ -110,11 +110,6 @@ static int prism2sta_probe_usb(
>  	 * linux netdevice.
>  	 */
>  	SET_NETDEV_DEV(wlandev->netdev, &(interface->dev));
> -        if ( register_wlandev(wlandev) != 0 ) {
> -		WLAN_LOG_ERROR("%s: register_wlandev() failed.\n", dev_info);
> -		result = -EIO;
> -		goto failed;
> -        }
>  
>  	/* Do a chip-level reset on the MAC */
>  	if (prism2_doreset) {

	/* Do a chip-level reset on the MAC */
	if (prism2_doreset) {
		result = hfa384x_corereset(hw,
				prism2_reset_holdtime,
				prism2_reset_settletime, 0);
		if (result != 0) {
			unregister_wlandev(wlandev);

You must remove the unregister_wlandev here.

> @@ -138,6 +133,15 @@ static int prism2sta_probe_usb(
>  
>  	wlandev->msdstate = WLAN_MSD_HWPRESENT;
>  
> +        if ( register_wlandev(wlandev) != 0 ) {
> +		WLAN_LOG_ERROR("%s: register_wlandev() failed.\n", dev_info);
> +		result = -EIO;
> +		goto failed;

"failed" fails to undo usb_get_dev() which it must do if you jump there
from this late.

	Regards
		Oliver

Re: [PATCH 2/8] prism2_usb.c always enable the card in probe_usb

[Richard Kennedy]

On Mon, 2008-11-03 at 12:25 +0100, Oliver Neukum wrote:
> Am Montag, 3. November 2008 12:13:48 schrieb Richard Kennedy:
> > always enable card in probe_usb
> > & update register_wlandev to match latest wlan-ng-dev tree.
> 
> This breaks both error cases.
> 
> > ---
> >  drivers/staging/wlan-ng/prism2_usb.c |   14 +++++++++-----
> >  1 files changed, 9 insertions(+), 5 deletions(-)
> > 
> > diff --git a/drivers/staging/wlan-ng/prism2_usb.c b/drivers/staging/wlan-ng/prism2_usb.c
> > index f1ae371..34a36c1 100644
> > --- a/drivers/staging/wlan-ng/prism2_usb.c
> > +++ b/drivers/staging/wlan-ng/prism2_usb.c
> > @@ -110,11 +110,6 @@ static int prism2sta_probe_usb(
> >  	 * linux netdevice.
> >  	 */
> >  	SET_NETDEV_DEV(wlandev->netdev, &(interface->dev));
> > -        if ( register_wlandev(wlandev) != 0 ) {
> > -		WLAN_LOG_ERROR("%s: register_wlandev() failed.\n", dev_info);
> > -		result = -EIO;
> > -		goto failed;
> > -        }
> >  
> >  	/* Do a chip-level reset on the MAC */
> >  	if (prism2_doreset) {
> 
> 	/* Do a chip-level reset on the MAC */
> 	if (prism2_doreset) {
> 		result = hfa384x_corereset(hw,
> 				prism2_reset_holdtime,
> 				prism2_reset_settletime, 0);
> 		if (result != 0) {
> 			unregister_wlandev(wlandev);
> 
> You must remove the unregister_wlandev here.
> > @@ -138,6 +133,15 @@ static int prism2sta_probe_usb(
> >  
> >  	wlandev->msdstate = WLAN_MSD_HWPRESENT;
> >  
> > +        if ( register_wlandev(wlandev) != 0 ) {
> > +		WLAN_LOG_ERROR("%s: register_wlandev() failed.\n", dev_info);
> > +		result = -EIO;
> > +		goto failed;
> 
> "failed" fails to undo usb_get_dev() which it must do if you jump there
> from this late.
> 
> 	Regards
> 		Oliver

Oliver,
thanks for that, I will tidy it up & post a patch.

prism2_doreset is 0 by default, so that hunk of code will not usually
get called. hfa384x_corereset doesn't do much on usb, and holdtime &
settletime aren't used at all. So more targets for clean up I guess :)

regards
Richard

[PATCH 3/8] hfa384x_usb.c use newest version of 384x_drvr_start

[Richard Kennedy]

include the needed fixes from Karl Relton
<karllinuxtest.relton@ntlworld.com>

see thread on linux-wlan-devel mailing list
"Possible cause of those pesky hfa384x_usbctlx_complete_sync errors"

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
---
 drivers/staging/wlan-ng/hfa384x_usb.c |   68 ++++++++++++++++++++++++++------
 1 files changed, 55 insertions(+), 13 deletions(-)

diff --git a/drivers/staging/wlan-ng/hfa384x_usb.c b/drivers/staging/wlan-ng/hfa384x_usb.c
index 53c547d..4b1d24e 100644
--- a/drivers/staging/wlan-ng/hfa384x_usb.c
+++ b/drivers/staging/wlan-ng/hfa384x_usb.c
@@ -3140,19 +3140,38 @@ int hfa384x_drvr_setconfig(hfa384x_t *hw, UINT16 rid, void *buf, UINT16 len)
 * Call context:
 *	process
 ----------------------------------------------------------------*/
+
 int hfa384x_drvr_start(hfa384x_t *hw)
 {
-	int		result;
+	int		result, result1, result2;
+	u16		status;
 	DBFENTER;
 
 	might_sleep();
-
-	if (usb_clear_halt(hw->usb, hw->endp_in)) {
+    
+	/* Clear endpoint stalls - but only do this if the endpoint
+	 * is showing a stall status. Some prism2 cards seem to behave
+	 * badly if a clear_halt is called when the endpoint is already
+	 * ok
+	 */
+	result = usb_get_status(hw->usb, USB_RECIP_ENDPOINT, hw->endp_in, &status);
+	if (result < 0) {
+		WLAN_LOG_ERROR(
+			"Cannot get bulk in endpoint status.\n");
+		goto done;
+	}
+	if ((status == 1) && usb_clear_halt(hw->usb, hw->endp_in)) {
 		WLAN_LOG_ERROR(
 			"Failed to reset bulk in endpoint.\n");
 	}
 
-	if (usb_clear_halt(hw->usb, hw->endp_out)) {
+	result = usb_get_status(hw->usb, USB_RECIP_ENDPOINT, hw->endp_out, &status);
+	if (result < 0) {
+		WLAN_LOG_ERROR(
+			"Cannot get bulk out endpoint status.\n");
+		goto done;
+	}
+	if ((status == 1) && usb_clear_halt(hw->usb, hw->endp_out)) {
 		WLAN_LOG_ERROR(
 			"Failed to reset bulk out endpoint.\n");
 	}
@@ -3169,18 +3188,41 @@ int hfa384x_drvr_start(hfa384x_t *hw)
 		goto done;
 	}
 
-	/* call initialize */
-	result = hfa384x_cmd_initialize(hw);
-	if (result != 0) {
-		usb_kill_urb(&hw->rx_urb);
-		WLAN_LOG_ERROR(
-			"cmd_initialize() failed, result=%d\n",
-			result);
-		goto done;
+	/* Call initialize twice, with a 1 second sleep in between.
+	 * This is a nasty work-around since many prism2 cards seem to
+	 * need time to settle after an init from cold. The second
+	 * call to initialize in theory is not necessary - but we call
+	 * it anyway as a double insurance policy:
+	 * 1) If the first init should fail, the second may well succeed
+	 *    and the card can still be used
+	 * 2) It helps ensures all is well with the card after the first
+	 *    init and settle time.
+	 */
+	result1 = hfa384x_cmd_initialize(hw);
+	msleep(1000);
+	result = result2 = hfa384x_cmd_initialize(hw);
+	if (result1 != 0) {
+		if (result2 != 0) {
+			WLAN_LOG_ERROR(
+				"cmd_initialize() failed on two attempts, results %d and %d\n",
+				result1, result2);
+			usb_kill_urb(&hw->rx_urb);
+			goto done;
+		} else {
+			WLAN_LOG_DEBUG(0, "First cmd_initialize() failed (result %d),\n",
+				result1);
+			WLAN_LOG_DEBUG(0, "but second attempt succeeded. All should be ok\n");
+		}
+	} else if (result2 != 0) {
+		WLAN_LOG_WARNING(
+			"First cmd_initialize() succeeded, but second attempt failed (result=%d)\n",
+			result2);
+		WLAN_LOG_WARNING("Most likely the card will be functional\n");
+			goto done;
 	}
 
 	hw->state = HFA384x_STATE_RUNNING;
-
+	
 done:
 	DBFEXIT;
 	return result;
-- 
1.5.6.5

[PATCH 4/8] p80211netdev.c correctly enable wext handlers

[Richard Kennedy]

correctly enable wext handlers
CONFIG_NET_WIRELESS does not exist so wext_handlers did not get
defined. 

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
---
 drivers/staging/wlan-ng/p80211netdev.c |    2 --
 1 files changed, 0 insertions(+), 2 deletions(-)

diff --git a/drivers/staging/wlan-ng/p80211netdev.c b/drivers/staging/wlan-ng/p80211netdev.c
index cc9cd4c..17b9321 100644
--- a/drivers/staging/wlan-ng/p80211netdev.c
+++ b/drivers/staging/wlan-ng/p80211netdev.c
@@ -864,12 +864,10 @@ int wlan_setup(wlandevice_t *wlandev)
 		dev->open =		p80211knetdev_open;
 		dev->stop =		p80211knetdev_stop;
 
-#ifdef CONFIG_NET_WIRELESS
 #if (WIRELESS_EXT < 21)
 		dev->get_wireless_stats = p80211wext_get_wireless_stats;
 #endif
 		dev->wireless_handlers = &p80211wext_handler_def;
-#endif
 
 		netif_stop_queue(dev);
 #ifdef HAVE_CHANGE_MTU
-- 
1.5.6.5

Re: [PATCH 4/8] p80211netdev.c correctly enable wext handlers

[Greg KH]

On Mon, Nov 03, 2008 at 11:18:23AM +0000, Richard Kennedy wrote:
> correctly enable wext handlers
> CONFIG_NET_WIRELESS does not exist so wext_handlers did not get
> defined. 

This patch had already been applied to my tree from someone else.

thanks,

greg k-h

[PATCH 5/8] p80211wext.c add latest changes & remove extra nulls from wext_handlers

[Richard Kennedy]

add the latest changes from wlan-ng-devel
remove 2 stray nulls from the wext_handler table introduced
by the tidy up patch.

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
---
 drivers/staging/wlan-ng/p80211wext.c |   39
++++++++++++++++++++++-----------
 1 files changed, 26 insertions(+), 13 deletions(-)

diff --git a/drivers/staging/wlan-ng/p80211wext.c b/drivers/staging/wlan-ng/p80211wext.c
index 153e5eb..de0789a 100644
--- a/drivers/staging/wlan-ng/p80211wext.c
+++ b/drivers/staging/wlan-ng/p80211wext.c
@@ -553,10 +553,13 @@ static int p80211wext_giwencode(netdevice_t *dev,
 
 	DBFENTER;
 
+	i = (erq->flags & IW_ENCODE_INDEX) - 1;
+	erq->flags = 0;
+
 	if (wlandev->hostwep & HOSTWEP_PRIVACYINVOKED)
-		erq->flags = IW_ENCODE_ENABLED;
+		erq->flags |= IW_ENCODE_ENABLED;
 	else
-		erq->flags = IW_ENCODE_DISABLED;
+		erq->flags |= IW_ENCODE_DISABLED;
 
 	if (wlandev->hostwep & HOSTWEP_EXCLUDEUNENCRYPTED)
 		erq->flags |= IW_ENCODE_RESTRICTED;
@@ -619,17 +622,24 @@ static int p80211wext_siwencode(netdevice_t *dev,
 			err = -EFAULT;
 			goto exit;
 		}
-		else {
-			enable = 1;
+
+		/* Set current key number only if no keys are given */
+		if (erq->flags & IW_ENCODE_NOKEY) {
+			result = p80211wext_dorequest(wlandev, DIDmib_dot11smt_dot11PrivacyTable_dot11WEPDefaultKeyID, i);
+		
+			if (result) {
+				err = -EFAULT;
+				goto exit;
+			}
 		}
 
-	}
-	else {
-		// Do not thing when no Key Index
+	} else {
+		// Use defaultkey if no Key Index
+		i = wlandev->hostwep & HOSTWEP_DEFAULTKEY_MASK;
 	}
 
 	/* Check if there is no key information in the iwconfig request */
-	if((erq->flags & IW_ENCODE_NOKEY) == 0 && enable == 1) {
+	if((erq->flags & IW_ENCODE_NOKEY) == 0 ) {
 
 		/*------------------------------------------------------------
 		 * If there is WEP Key for setting, check the Key Information
@@ -684,8 +694,7 @@ static int p80211wext_siwencode(netdevice_t *dev,
 	/* Check the PrivacyInvoked flag */
 	if (erq->flags & IW_ENCODE_DISABLED) {
 		result = p80211wext_dorequest(wlandev, DIDmib_dot11smt_dot11PrivacyTable_dot11PrivacyInvoked, P80211ENUM_truth_false);
-	}
-	else if((erq->flags & IW_ENCODE_ENABLED) || enable == 1) {
+	} else {
 		result = p80211wext_dorequest(wlandev, DIDmib_dot11smt_dot11PrivacyTable_dot11PrivacyInvoked, P80211ENUM_truth_true);
 	}
 
@@ -694,7 +703,13 @@ static int p80211wext_siwencode(netdevice_t *dev,
 		goto exit;
 	}
 
-	/* Check the ExcludeUnencrypted flag */
+	/*  The  security  mode  may  be open or restricted, and its meaning
+	    depends on the card used. With  most  cards,  in  open  mode  no
+	    authentication  is  used  and  the  card  may  also  accept non-
+	    encrypted sessions, whereas in restricted  mode  only  encrypted
+	    sessions  are  accepted  and the card will use authentication if
+	    available. 
+	*/
 	if (erq->flags & IW_ENCODE_RESTRICTED) {
 		result = p80211wext_dorequest(wlandev, DIDmib_dot11smt_dot11PrivacyTable_dot11ExcludeUnencrypted, P80211ENUM_truth_true);
 	}
@@ -1760,8 +1775,6 @@ static iw_handler p80211wext_handlers[] =  {
 	(iw_handler) NULL,                  		/* SIOCGIWAPLIST */
 	(iw_handler) p80211wext_siwscan,		/* SIOCSIWSCAN */
 	(iw_handler) p80211wext_giwscan,		/* SIOCGIWSCAN */
-	(iw_handler) NULL,	/* null */		/* SIOCSIWSCAN */
-	(iw_handler) NULL,	/* null */		/* SIOCGIWSCAN */
 	(iw_handler) p80211wext_siwessid,  		/* SIOCSIWESSID */
 	(iw_handler) p80211wext_giwessid,      		/* SIOCGIWESSID */
 	(iw_handler) NULL,                 		/* SIOCSIWNICKN */
-- 
1.5.6.5

[PATCH 6/8] p80211wext don't set default key id twice

[Richard Kennedy]

p80211wext don't set default key id twice

another change from wlan-ng-devel

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
---
 drivers/staging/wlan-ng/p80211wext.c |    7 -------
 1 files changed, 0 insertions(+), 7 deletions(-)

diff --git a/drivers/staging/wlan-ng/p80211wext.c b/drivers/staging/wlan-ng/p80211wext.c
index de0789a..28c8359 100644
--- a/drivers/staging/wlan-ng/p80211wext.c
+++ b/drivers/staging/wlan-ng/p80211wext.c
@@ -616,13 +616,6 @@ static int p80211wext_siwencode(netdevice_t *dev,
 		else
 			i--;
 
-		result = p80211wext_dorequest(wlandev, DIDmib_dot11smt_dot11PrivacyTable_dot11WEPDefaultKeyID, i);
-
-		if (result) {
-			err = -EFAULT;
-			goto exit;
-		}
-
 		/* Set current key number only if no keys are given */
 		if (erq->flags & IW_ENCODE_NOKEY) {
 			result = p80211wext_dorequest(wlandev, DIDmib_dot11smt_dot11PrivacyTable_dot11WEPDefaultKeyID, i);
-- 
1.5.6.5

[PATCH 7/8] hfa384x_usbin_callback: check for hardware removed

[Richard Kennedy]

hfa384x_usbin_callback: check for hardware removed

copied from latest wlan-ng-devel version

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
---
 drivers/staging/wlan-ng/hfa384x_usb.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/staging/wlan-ng/hfa384x_usb.c b/drivers/staging/wlan-ng/hfa384x_usb.c
index 4b1d24e..9184e27 100644
--- a/drivers/staging/wlan-ng/hfa384x_usb.c
+++ b/drivers/staging/wlan-ng/hfa384x_usb.c
@@ -3788,7 +3788,7 @@ static void hfa384x_usbin_callback(struct urb *urb, struct pt_regs *regs)
 
 	if ( !wlandev ||
 	     !wlandev->netdev ||
-	     !netif_device_present(wlandev->netdev) )
+	     wlandev->hwremoved )
 		goto exit;
 
 	hw = wlandev->priv;
-- 
1.5.6.5

[PATCH 8/8] p80211conv.c copy code from wlan-ng-devel branch to not drop packets

[Richard Kennedy]

allow card to correctly receive network packets,
without this change all incoming packets are dropped.
code copied from the latest wlan-ng-devel tree.

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
---
 drivers/staging/wlan-ng/p80211conv.c |   49 +++++++++++++++++++++++++++++++++-
 1 files changed, 48 insertions(+), 1 deletions(-)

diff --git a/drivers/staging/wlan-ng/p80211conv.c b/drivers/staging/wlan-ng/p80211conv.c
index 749ea8d..b09a9ab 100644
--- a/drivers/staging/wlan-ng/p80211conv.c
+++ b/drivers/staging/wlan-ng/p80211conv.c
@@ -380,6 +380,14 @@ int skb_p80211_to_ether( wlandevice_t *wlandev, UINT32 ethconv, struct sk_buff *
 	     (memcmp(saddr, e_hdr->saddr, WLAN_ETHADDR_LEN) == 0))) {
 		WLAN_LOG_DEBUG(3, "802.3 ENCAP len: %d\n", payload_length);
 		/* 802.3 Encapsulated */
+		/* Test for an overlength frame */
+		if ( payload_length > (netdev->mtu + WLAN_ETHHDR_LEN)) {
+			/* A bogus length ethfrm has been encap'd. */
+			/* Is someone trying an oflow attack? */
+			WLAN_LOG_ERROR("ENCAP frame too large (%d > %d)\n",
+				payload_length, netdev->mtu + WLAN_ETHHDR_LEN);
+			return 1;
+		}
 
 		/* Chop off the 802.11 header.  it's already sane. */
 		skb_pull(skb, payload_offset);
@@ -399,6 +407,15 @@ int skb_p80211_to_ether( wlandevice_t *wlandev, UINT32 ethconv, struct sk_buff *
 		/* it's a SNAP + RFC1042 frame && protocol is in STT */
 		/* build 802.3 + RFC1042 */
 
+		/* Test for an overlength frame */
+		if ( payload_length > netdev->mtu ) {
+			/* A bogus length ethfrm has been sent. */
+			/* Is someone trying an oflow attack? */
+			WLAN_LOG_ERROR("SNAP frame too large (%d > %d)\n",
+				payload_length, netdev->mtu);
+			return 1;
+		}
+
 		/* chop 802.11 header from skb. */
 		skb_pull(skb, payload_offset);
 
@@ -419,6 +436,18 @@ int skb_p80211_to_ether( wlandevice_t *wlandev, UINT32 ethconv, struct sk_buff *
 		/* it's an 802.1h frame || (an RFC1042 && protocol is not in STT) */
 		/* build a DIXII + RFC894 */
 
+		/* Test for an overlength frame */
+		if ((payload_length - sizeof(wlan_llc_t) - sizeof(wlan_snap_t))
+		    > netdev->mtu) {
+			/* A bogus length ethfrm has been sent. */
+			/* Is someone trying an oflow attack? */
+			WLAN_LOG_ERROR("DIXII frame too large (%ld > %d)\n",
+					(long int) (payload_length - sizeof(wlan_llc_t) -
+						    sizeof(wlan_snap_t)),
+					netdev->mtu);
+			return 1;
+		}
+
 		/* chop 802.11 header from skb. */
 		skb_pull(skb, payload_offset);
 
@@ -443,6 +472,16 @@ int skb_p80211_to_ether( wlandevice_t *wlandev, UINT32 ethconv, struct sk_buff *
 		/*  build an 802.3 frame */
 		/* allocate space and setup hostbuf */
 
+		/* Test for an overlength frame */
+		if ( payload_length > netdev->mtu ) {
+			/* A bogus length ethfrm has been sent. */
+			/* Is someone trying an oflow attack? */
+			WLAN_LOG_ERROR("OTHER frame too large (%d > %d)\n",
+				payload_length,
+				netdev->mtu);
+			return 1;
+		}
+
 		/* Chop off the 802.11 header. */
 		skb_pull(skb, payload_offset);
 
@@ -457,8 +496,16 @@ int skb_p80211_to_ether( wlandevice_t *wlandev, UINT32 ethconv, struct sk_buff *
 
 	}
 
+        /*
+         * Note that eth_type_trans() expects an skb w/ skb->data pointing
+         * at the MAC header, it then sets the following skb members:
+         * skb->mac_header, 
+         * skb->data, and
+         * skb->pkt_type.
+         * It then _returns_ the value that _we're_ supposed to stuff in
+         * skb->protocol.  This is nuts.
+         */
 	skb->protocol = eth_type_trans(skb, netdev);
-	skb_reset_mac_header(skb);
 
         /* jkriegl: process signal and noise as set in hfa384x_int_rx() */
 	/* jkriegl: only process signal/noise if requested by iwspy */
-- 
1.5.6.5

patch staging-wlan-ng-p80211conv.c-copy-code-from-wlan-ng-devel-branch-to-not-drop-packets.patch added to gregkh-2.6 tree

[gregkh]

This is a note to let you know that I've just added the patch titled

    Subject: Staging: wlan-ng: p80211conv.c copy code from wlan-ng-devel branch to not drop packets

to my gregkh-2.6 tree.  Its filename is

    staging-wlan-ng-p80211conv.c-copy-code-from-wlan-ng-devel-branch-to-not-drop-packets.patch

This tree can be found at 
    http://www.kernel.org/pub/linux/kernel/people/gregkh/gregkh-2.6/patches/


>From richard@rsk.demon.co.uk  Wed Nov 12 14:31:45 2008
From: Richard Kennedy <richard@rsk.demon.co.uk>
Date: Mon, 03 Nov 2008 11:24:54 +0000
Subject: Staging: wlan-ng: p80211conv.c copy code from wlan-ng-devel branch to not drop packets
To: gregkh <gregkh@suse.de>
Cc: lkml <linux-kernel@vger.kernel.org>
Message-ID: <1225711494.3113.34.camel@castor.localdomain>


allow card to correctly receive network packets,
without this change all incoming packets are dropped.
code copied from the latest wlan-ng-devel tree.

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

---
 drivers/staging/wlan-ng/p80211conv.c |   49 ++++++++++++++++++++++++++++++++++-
 1 file changed, 48 insertions(+), 1 deletion(-)

--- a/drivers/staging/wlan-ng/p80211conv.c
+++ b/drivers/staging/wlan-ng/p80211conv.c
@@ -377,6 +377,14 @@ int skb_p80211_to_ether( wlandevice_t *w
 	     (memcmp(saddr, e_hdr->saddr, WLAN_ETHADDR_LEN) == 0))) {
 		WLAN_LOG_DEBUG(3, "802.3 ENCAP len: %d\n", payload_length);
 		/* 802.3 Encapsulated */
+		/* Test for an overlength frame */
+		if ( payload_length > (netdev->mtu + WLAN_ETHHDR_LEN)) {
+			/* A bogus length ethfrm has been encap'd. */
+			/* Is someone trying an oflow attack? */
+			WLAN_LOG_ERROR("ENCAP frame too large (%d > %d)\n",
+				payload_length, netdev->mtu + WLAN_ETHHDR_LEN);
+			return 1;
+		}
 
 		/* Chop off the 802.11 header.  it's already sane. */
 		skb_pull(skb, payload_offset);
@@ -396,6 +404,15 @@ int skb_p80211_to_ether( wlandevice_t *w
 		/* it's a SNAP + RFC1042 frame && protocol is in STT */
 		/* build 802.3 + RFC1042 */
 
+		/* Test for an overlength frame */
+		if ( payload_length > netdev->mtu ) {
+			/* A bogus length ethfrm has been sent. */
+			/* Is someone trying an oflow attack? */
+			WLAN_LOG_ERROR("SNAP frame too large (%d > %d)\n",
+				payload_length, netdev->mtu);
+			return 1;
+		}
+
 		/* chop 802.11 header from skb. */
 		skb_pull(skb, payload_offset);
 
@@ -416,6 +433,18 @@ int skb_p80211_to_ether( wlandevice_t *w
 		/* it's an 802.1h frame || (an RFC1042 && protocol is not in STT) */
 		/* build a DIXII + RFC894 */
 
+		/* Test for an overlength frame */
+		if ((payload_length - sizeof(wlan_llc_t) - sizeof(wlan_snap_t))
+		    > netdev->mtu) {
+			/* A bogus length ethfrm has been sent. */
+			/* Is someone trying an oflow attack? */
+			WLAN_LOG_ERROR("DIXII frame too large (%ld > %d)\n",
+					(long int) (payload_length - sizeof(wlan_llc_t) -
+						    sizeof(wlan_snap_t)),
+					netdev->mtu);
+			return 1;
+		}
+
 		/* chop 802.11 header from skb. */
 		skb_pull(skb, payload_offset);
 
@@ -440,6 +469,16 @@ int skb_p80211_to_ether( wlandevice_t *w
 		/*  build an 802.3 frame */
 		/* allocate space and setup hostbuf */
 
+		/* Test for an overlength frame */
+		if ( payload_length > netdev->mtu ) {
+			/* A bogus length ethfrm has been sent. */
+			/* Is someone trying an oflow attack? */
+			WLAN_LOG_ERROR("OTHER frame too large (%d > %d)\n",
+				payload_length,
+				netdev->mtu);
+			return 1;
+		}
+
 		/* Chop off the 802.11 header. */
 		skb_pull(skb, payload_offset);
 
@@ -454,8 +493,16 @@ int skb_p80211_to_ether( wlandevice_t *w
 
 	}
 
+        /*
+         * Note that eth_type_trans() expects an skb w/ skb->data pointing
+         * at the MAC header, it then sets the following skb members:
+         * skb->mac_header,
+         * skb->data, and
+         * skb->pkt_type.
+         * It then _returns_ the value that _we're_ supposed to stuff in
+         * skb->protocol.  This is nuts.
+         */
 	skb->protocol = eth_type_trans(skb, netdev);
-	skb_reset_mac_header(skb);
 
         /* jkriegl: process signal and noise as set in hfa384x_int_rx() */
 	/* jkriegl: only process signal/noise if requested by iwspy */


Patches currently in gregkh-2.6 which might be from richard@rsk.demon.co.uk are

staging/staging-wlan-ng-hfa384x_usb.c-use-newest-version-of-384x_drvr_start.patch
staging/staging-wlan-ng-hfa384x_usbin_callback-check-for-hardware-removed.patch
staging/staging-wlan-ng-p80211conv.c-copy-code-from-wlan-ng-devel-branch-to-not-drop-packets.patch
staging/staging-wlan-ng-p80211netdev.c-fix-netdev-alloc-to-prevent-oops-on-device-start.patch
staging/staging-wlan-ng-p80211wext.c-add-latest-changes-remove-extra-nulls-from-wext_handlers.patch
staging/staging-wlan-ng-p80211wext-don-t-set-default-key-id-twice.patch
staging/staging-wlan-ng-prism2_usb.c-always-enable-the-card-in-probe_usb.patch