ath5k: Fix frame duration oops

This patch fixes an oops which was introduced as a regression by
commit fd640775bd16e1df50c867cc547af0, on the patch titled,
"mac80211: dont use interface indices in drivers".

ieee80211_generic_frame_duration() now relies on sdata->flags which
itself gets set upon bringing the interface up. We check for the
virtual interface now before setting the rate duration registers.

After the mode changes are introduced onto mac80211 we should revisit
these changes.

This patch was tested on the following cards:

1) BG card:

Atheros AR5213A chip found (MAC: 0x59, PHY: 0x43)
RF2112A 2GHz radio found (0x46)

2) ABG card:

Atheros AR5213A chip found (MAC: 0x59,PHY: 0x43)
RF5112A multiband radio found (0x36)
Signed-off-by: Luis R. Rodriguez <mcgrof@winlab.rutgers.edu>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

drivers/net/wireless/ath5k/hw.c

@@ -734,8 +734,12 @@ int ath5k_hw_reset(struct ath5k_hw *ah, enum ieee80211_if_types op_mode,
 		if (ret)
 			return ret;
 
-		/* Write rate duration table */
-		if (ah->ah_version == AR5K_AR5212)
+		/* Write rate duration table only on AR5212 and if
+		 * virtual interface has already been brought up
+		 * XXX: rethink this after new mode changes to
+		 * mac80211 are integrated */
+		if (ah->ah_version == AR5K_AR5212 &&
+			ah->ah_sc->vif != NULL)
 			ath5k_hw_write_rate_duration(ah, driver_mode);
 
 		/*