mac80211: add a function for running rx without passing skbs to the stack

This can be used to run mac80211 rx processing on a batch of frames in NAPI
poll before passing them to the network stack in a large batch.
This can improve icache footprint, or it can be used to pass frames via
netif_receive_skb_list.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
Link: https://lore.kernel.org/r/20200726110611.46886-1-nbd@nbd.name
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

2 files changed, 39 insertions, 23 deletions

diff --git a/net/mac80211/ieee80211_i.h b/net/mac80211/ieee80211_i.h
index 1760ce77d89f..0b1eaec6649f 100644
--- a/net/mac80211/ieee80211_i.h
+++ b/net/mac80211/ieee80211_i.h
@@ -217,7 +217,7 @@ enum ieee80211_rx_flags {
 };
 
 struct ieee80211_rx_data {
-	struct napi_struct *napi;
+	struct list_head *list;
 	struct sk_buff *skb;
 	struct ieee80211_local *local;
 	struct ieee80211_sub_if_data *sdata;
diff --git a/net/mac80211/rx.c b/net/mac80211/rx.c
index 9ac3eaa00ce6..836cde516a18 100644
--- a/net/mac80211/rx.c
+++ b/net/mac80211/rx.c
@@ -2578,8 +2578,8 @@ static void ieee80211_deliver_skb_to_local_stack(struct sk_buff *skb,
 		memset(skb->cb, 0, sizeof(skb->cb));
 
 		/* deliver to local stack */
-		if (rx->napi)
-			napi_gro_receive(rx->napi, skb);
+		if (rx->list)
+			list_add_tail(&skb->list, rx->list);
 		else
 			netif_receive_skb(skb);
 	}
@@ -3869,7 +3869,6 @@ void ieee80211_release_reorder_timeout(struct sta_info *sta, int tid)
 		/* This is OK -- must be QoS data frame */
 		.security_idx = tid,
 		.seqno_idx = tid,
-		.napi = NULL, /* must be NULL to not have races */
 	};
 	struct tid_ampdu_rx *tid_agg_rx;
 
@@ -4479,8 +4478,8 @@ static bool ieee80211_invoke_fast_rx(struct ieee80211_rx_data *rx,
 	/* deliver to local stack */
 	skb->protocol = eth_type_trans(skb, fast_rx->dev);
 	memset(skb->cb, 0, sizeof(skb->cb));
-	if (rx->napi)
-		napi_gro_receive(rx->napi, skb);
+	if (rx->list)
+		list_add_tail(&skb->list, rx->list);
 	else
 		netif_receive_skb(skb);
 
@@ -4547,7 +4546,7 @@ static bool ieee80211_prepare_and_rx_handle(struct ieee80211_rx_data *rx,
 static void __ieee80211_rx_handle_packet(struct ieee80211_hw *hw,
 					 struct ieee80211_sta *pubsta,
 					 struct sk_buff *skb,
-					 struct napi_struct *napi)
+					 struct list_head *list)
 {
 	struct ieee80211_local *local = hw_to_local(hw);
 	struct ieee80211_sub_if_data *sdata;
@@ -4562,7 +4561,7 @@ static void __ieee80211_rx_handle_packet(struct ieee80211_hw *hw,
 	memset(&rx, 0, sizeof(rx));
 	rx.skb = skb;
 	rx.local = local;
-	rx.napi = napi;
+	rx.list = list;
 
 	if (ieee80211_is_data(fc) || ieee80211_is_mgmt(fc))
 		I802_DEBUG_INC(local->dot11ReceivedFragmentCount);
@@ -4670,8 +4669,8 @@ static void __ieee80211_rx_handle_packet(struct ieee80211_hw *hw,
  * This is the receive path handler. It is called by a low level driver when an
  * 802.11 MPDU is received from the hardware.
  */
-void ieee80211_rx_napi(struct ieee80211_hw *hw, struct ieee80211_sta *pubsta,
-		       struct sk_buff *skb, struct napi_struct *napi)
+void ieee80211_rx_list(struct ieee80211_hw *hw, struct ieee80211_sta *pubsta,
+		       struct sk_buff *skb, struct list_head *list)
 {
 	struct ieee80211_local *local = hw_to_local(hw);
 	struct ieee80211_rate *rate = NULL;
@@ -4763,36 +4762,53 @@ void ieee80211_rx_napi(struct ieee80211_hw *hw, struct ieee80211_sta *pubsta,
 	status->rx_flags = 0;
 
 	/*
-	 * key references and virtual interfaces are protected using RCU
-	 * and this requires that we are in a read-side RCU section during
-	 * receive processing
-	 */
-	rcu_read_lock();
-
-	/*
 	 * Frames with failed FCS/PLCP checksum are not returned,
 	 * all other frames are returned without radiotap header
 	 * if it was previously present.
 	 * Also, frames with less than 16 bytes are dropped.
 	 */
 	skb = ieee80211_rx_monitor(local, skb, rate);
-	if (!skb) {
-		rcu_read_unlock();
+	if (!skb)
 		return;
-	}
 
 	ieee80211_tpt_led_trig_rx(local,
 			((struct ieee80211_hdr *)skb->data)->frame_control,
 			skb->len);
 
-	__ieee80211_rx_handle_packet(hw, pubsta, skb, napi);
-
-	rcu_read_unlock();
+	__ieee80211_rx_handle_packet(hw, pubsta, skb, list);
 
 	return;
  drop:
 	kfree_skb(skb);
 }
+EXPORT_SYMBOL(ieee80211_rx_list);
+
+void ieee80211_rx_napi(struct ieee80211_hw *hw, struct ieee80211_sta *pubsta,
+		       struct sk_buff *skb, struct napi_struct *napi)
+{
+	struct sk_buff *tmp;
+	LIST_HEAD(list);
+
+
+	/*
+	 * key references and virtual interfaces are protected using RCU
+	 * and this requires that we are in a read-side RCU section during
+	 * receive processing
+	 */
+	rcu_read_lock();
+	ieee80211_rx_list(hw, pubsta, skb, &list);
+	rcu_read_unlock();
+
+	if (!napi) {
+		netif_receive_skb_list(&list);
+		return;
+	}
+
+	list_for_each_entry_safe(skb, tmp, &list, list) {
+		skb_list_del_init(skb);
+		napi_gro_receive(napi, skb);
+	}
+}
 EXPORT_SYMBOL(ieee80211_rx_napi);
 
 /* This is a version of the rx handler that can be called from hard irq