Re: [PATCH 3/5] mac80211: remove unused status flag checks

[Juuso Oikarinen <juuso.oikarinen@nokia.com>]

On Wed, 2010-08-11 at 14:39 +0200, ext Johannes Berg wrote:
> On Wed, 2010-08-11 at 15:12 +0300, Juuso Oikarinen wrote:
> 
> > > > The decryption code verifies whether or not
> > > > a given frame was decrypted and verified by
> > > > hardware. This is unnecessary, as the crypto
> > > > RX handler already does it long before the
> > > > decryption code is even invoked, so remove
> > > > that code to avoid confusion.
> > > > 
> > > > Signed-off-by: Johannes Berg <johannes.berg@intel.com>
> > > > ---
> > > >  net/mac80211/wpa.c |   26 ++++++--------------------
> > > >  1 file changed, 6 insertions(+), 20 deletions(-)
> > > > 
> > > 
> > > This patch for some reason seems to break wl1271 WPA - association
> > > succeeds but encrypted data transfer fails.
> > > 
> > > I still don't know why, but I'm looking into it.
> > > 
> > 
> > It appears, that in function ieee80211_rx_h_decrypt we go here:
> > 
> > 	if (!is_multicast_ether_addr(hdr->addr1) && stakey) {
> > 		rx->key = stakey;
> > 		/* Skip decryption if the frame is not protected. */
> > 		if (!ieee80211_has_protected(hdr->frame_control))
> > 			return RX_CONTINUE;
> > 
> > And here, as the frame is protected, we go out of the if, and end up in
> > tkip_decrypt, which with this patch no longer checks whether the frame
> > is already decrypted.
> > 
> > The frame then ends up dropped.
> 
> Err, you're right, sorry about that. There are too many paths here. How
> about this patch?
> 
> johannes
> 
> --- wireless-testing.orig/net/mac80211/rx.c	2010-08-11 14:37:13.000000000 +0200
> +++ wireless-testing/net/mac80211/rx.c	2010-08-11 14:38:13.000000000 +0200
> @@ -873,6 +873,9 @@ ieee80211_rx_h_decrypt(struct ieee80211_
>  
>  	if (!is_multicast_ether_addr(hdr->addr1) && stakey) {
>  		rx->key = stakey;
> +		if ((status->flag & RX_FLAG_DECRYPTED) &&
> +		    (status->flag & RX_FLAG_IV_STRIPPED))
> +			return RX_CONTINUE;
>  		/* Skip decryption if the frame is not protected. */
>  		if (!ieee80211_has_protected(fc))
>  			return RX_CONTINUE;
> 
> 

This appears to work.

-Juus