]> xenbits.xensource.com Git - people/julieng/freebsd.git/commitdiff
Comments, mostly to remind myself of what's going on and why.
authoradrian <adrian@FreeBSD.org>
Mon, 28 Sep 2015 00:51:24 +0000 (00:51 +0000)
committeradrian <adrian@FreeBSD.org>
Mon, 28 Sep 2015 00:51:24 +0000 (00:51 +0000)
sys/net80211/ieee80211_output.c

index 4e061d80ffb0b26b91619fd67980b9e81c1841cf..27fe929ac28870484dc9d23acd4c1a0f7c048f03 100644 (file)
@@ -213,6 +213,22 @@ ieee80211_vap_pkt_send_dest(struct ieee80211vap *vap, struct mbuf *m,
                }
        }
 
+       /*
+        * XXX If we aren't doing AMPDU TX then we /could/ do
+        * fast-frames encapsulation, however right now this
+        * output logic doesn't handle that case.
+        *
+        * So we'll be limited to "fast-frames" xmit for non-11n STA
+        * and "no fast frames" xmit for 11n STAs.
+        * It'd be nice to eventually test fast-frames out by
+        * gracefully falling from failing A-MPDU transmission
+        * (driver says no, fail to negotiate it with peer) to
+        * using fast-frames.
+        *
+        * Note: we can actually put A-MSDU's inside an A-MPDU,
+        * so hopefully we can figure out how to make that particular
+        * combination work right.
+        */
 #ifdef IEEE80211_SUPPORT_SUPERG
        else if (IEEE80211_ATH_CAP(vap, ni, IEEE80211_NODE_FF)) {
                m = ieee80211_ff_check(ni, m);
@@ -230,6 +246,11 @@ ieee80211_vap_pkt_send_dest(struct ieee80211vap *vap, struct mbuf *m,
         */
        IEEE80211_TX_LOCK(ic);
 
+       /*
+        * XXX make the encap and transmit code a separate function
+        * so things like the FF (and later A-MSDU) path can just call
+        * it for flushed frames.
+        */
        if (__predict_true((vap->iv_caps & IEEE80211_C_8023ENCAP) == 0)) {
                /*
                 * Encapsulate the packet in prep for transmission.