From: Felix Fietkau <nbd@nbd.name>
Date: Thu, 16 Feb 2023 11:07:30 +0100
Subject: [PATCH] wifi: mac80211: use mesh header cache to speed up mesh
 forwarding

Significantly reduces mesh forwarding path CPU usage and enables the
direct use of iTXQ.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
---

--- a/net/mac80211/rx.c
+++ b/net/mac80211/rx.c
@@ -2720,6 +2720,65 @@ ieee80211_deliver_skb(struct ieee80211_r
 	}
 }
 
+#ifdef CPTCFG_MAC80211_MESH
+static bool
+ieee80211_rx_mesh_fast_forward(struct ieee80211_sub_if_data *sdata,
+			       struct sk_buff *skb, int hdrlen)
+{
+	struct ieee80211_if_mesh *ifmsh = &sdata->u.mesh;
+	struct ieee80211_mesh_fast_tx *entry = NULL;
+	struct ieee80211s_hdr *mesh_hdr;
+	struct tid_ampdu_tx *tid_tx;
+	struct sta_info *sta;
+	struct ethhdr eth;
+	u8 tid;
+
+	mesh_hdr = (struct ieee80211s_hdr *)(skb->data + sizeof(eth));
+	if ((mesh_hdr->flags & MESH_FLAGS_AE) == MESH_FLAGS_AE_A5_A6)
+		entry = mesh_fast_tx_get(sdata, mesh_hdr->eaddr1);
+	else if (!(mesh_hdr->flags & MESH_FLAGS_AE))
+		entry = mesh_fast_tx_get(sdata, skb->data);
+	if (!entry)
+		return false;
+
+	sta = rcu_dereference(entry->mpath->next_hop);
+	if (!sta)
+		return false;
+
+	if (skb_linearize(skb))
+		return false;
+
+	tid = skb->priority & IEEE80211_QOS_CTL_TAG1D_MASK;
+	tid_tx = rcu_dereference(sta->ampdu_mlme.tid_tx[tid]);
+	if (tid_tx) {
+		if (!test_bit(HT_AGG_STATE_OPERATIONAL, &tid_tx->state))
+			return false;
+
+		if (tid_tx->timeout)
+			tid_tx->last_tx = jiffies;
+	}
+
+	ieee80211_aggr_check(sdata, sta, skb);
+
+	if (ieee80211_get_8023_tunnel_proto(skb->data + hdrlen,
+					    &skb->protocol))
+		hdrlen += ETH_ALEN;
+	else
+		skb->protocol = htons(skb->len - hdrlen);
+	skb_set_network_header(skb, hdrlen + 2);
+
+	skb->dev = sdata->dev;
+	memcpy(&eth, skb->data, ETH_HLEN - 2);
+	skb_pull(skb, sizeof(eth));
+	__ieee80211_xmit_fast(sdata, sta, &entry->fast_tx, skb, tid_tx,
+			      eth.h_dest, eth.h_source);
+	IEEE80211_IFSTA_MESH_CTR_INC(ifmsh, fwded_unicast);
+	IEEE80211_IFSTA_MESH_CTR_INC(ifmsh, fwded_frames);
+
+	return true;
+}
+#endif
+
 static ieee80211_rx_result
 ieee80211_rx_mesh_data(struct ieee80211_sub_if_data *sdata, struct sta_info *sta,
 		       struct sk_buff *skb)
@@ -2824,6 +2883,10 @@ ieee80211_rx_mesh_data(struct ieee80211_
 
 	skb_set_queue_mapping(skb, ieee802_1d_to_ac[skb->priority]);
 
+	if (!multicast &&
+	    ieee80211_rx_mesh_fast_forward(sdata, skb, mesh_hdrlen))
+		return RX_QUEUED;
+
 	ieee80211_fill_mesh_addresses(&hdr, &hdr.frame_control,
 				      eth->h_dest, eth->h_source);
 	hdrlen = ieee80211_hdrlen(hdr.frame_control);
@@ -2862,6 +2925,7 @@ ieee80211_rx_mesh_data(struct ieee80211_
 	info->control.flags |= IEEE80211_TX_INTCFL_NEED_TXPROCESSING;
 	info->control.vif = &sdata->vif;
 	info->control.jiffies = jiffies;
+	fwd_skb->dev = sdata->dev;
 	if (multicast) {
 		IEEE80211_IFSTA_MESH_CTR_INC(ifmsh, fwded_mcast);
 		memcpy(fwd_hdr->addr2, sdata->vif.addr, ETH_ALEN);
@@ -2883,7 +2947,6 @@ ieee80211_rx_mesh_data(struct ieee80211_
 	}
 
 	IEEE80211_IFSTA_MESH_CTR_INC(ifmsh, fwded_frames);
-	fwd_skb->dev = sdata->dev;
 	ieee80211_add_pending_skb(local, fwd_skb);
 
 rx_accept:
--- a/net/mac80211/ieee80211_i.h
+++ b/net/mac80211/ieee80211_i.h
@@ -2018,6 +2018,8 @@ void __ieee80211_xmit_fast(struct ieee80
 			   struct ieee80211_fast_tx *fast_tx,
 			   struct sk_buff *skb, bool ampdu,
 			   const u8 *da, const u8 *sa);
+void ieee80211_aggr_check(struct ieee80211_sub_if_data *sdata,
+			  struct sta_info *sta, struct sk_buff *skb);
 
 /* HT */
 void ieee80211_apply_htcap_overrides(struct ieee80211_sub_if_data *sdata,
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -1191,10 +1191,8 @@ static bool ieee80211_tx_prep_agg(struct
 	return queued;
 }
 
-static void
-ieee80211_aggr_check(struct ieee80211_sub_if_data *sdata,
-		     struct sta_info *sta,
-		     struct sk_buff *skb)
+void ieee80211_aggr_check(struct ieee80211_sub_if_data *sdata,
+			  struct sta_info *sta, struct sk_buff *skb)
 {
 	struct rate_control_ref *ref = sdata->local->rate_ctrl;
 	u16 tid;