wifi: rtw88: process VO packets without workqueue to avoid PTK rekey failed

In the wpa_supplicant rekey flow, it sends an EAPOL packet 4/4 through
nl80211_tx_control_port() and triggers wake_tx_queue() in the driver.
Then, it sends nl80211_new_key() to configure a new key in mac80211.
However, in wake_tx_queue(), a workqueue is used to process the tx packet,
which might cause the driver to process the EAPOL packet later than
nl80211_new_key(). As a result, the EAPOL 4/4 packet is dropped by mac80211
due to the rekey configuration being finished. The EAPOL packets belongs to
VO packets that need high priority. Therefore, we process VO packets
directly without workqueue to ensure that packets can process immediately.

VO is normally used by voice application that is low traffic load and low
latency, that doesn't affect user experience.
We test iperf with VO packets(iperf3 -P4 -u -b 10000M -S 0xdf)
                          before       after
TX throughput             162M         162M
ping RTT                  3.8ms        3.7ms
Signed-off-by: Chih-Kang Chang <gary.chang@realtek.com>
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/20230616031713.16769-1-pkshih@realtek.com

drivers/net/wireless/realtek/rtw88/mac80211.c

@@ -43,7 +43,11 @@ static void rtw_ops_wake_tx_queue(struct ieee80211_hw *hw,
 		list_add_tail(&rtwtxq->list, &rtwdev->txqs);
 	spin_unlock_bh(&rtwdev->txq_lock);
 
-	queue_work(rtwdev->tx_wq, &rtwdev->tx_work);
+	/* ensure to dequeue EAPOL (4/4) at the right time */
+	if (txq->ac == IEEE80211_AC_VO)
+		__rtw_tx_work(rtwdev);
+	else
+		queue_work(rtwdev->tx_wq, &rtwdev->tx_work);
 }
 
 static int rtw_ops_start(struct ieee80211_hw *hw)

drivers/net/wireless/realtek/rtw88/tx.c

@@ -635,9 +635,8 @@ static void rtw_txq_push(struct rtw_dev *rtwdev,
 	rcu_read_unlock();
 }
 
-void rtw_tx_work(struct work_struct *w)
+void __rtw_tx_work(struct rtw_dev *rtwdev)
 {
-	struct rtw_dev *rtwdev = container_of(w, struct rtw_dev, tx_work);
 	struct rtw_txq *rtwtxq, *tmp;
 
 	spin_lock_bh(&rtwdev->txq_lock);
@@ -658,6 +657,13 @@ void rtw_tx_work(struct work_struct *w)
 	spin_unlock_bh(&rtwdev->txq_lock);
 }
 
+void rtw_tx_work(struct work_struct *w)
+{
+	struct rtw_dev *rtwdev = container_of(w, struct rtw_dev, tx_work);
+
+	__rtw_tx_work(rtwdev);
+}
+
 void rtw_txq_init(struct rtw_dev *rtwdev, struct ieee80211_txq *txq)
 {
 	struct rtw_txq *rtwtxq;

drivers/net/wireless/realtek/rtw88/tx.h

@@ -111,6 +111,7 @@ void rtw_tx(struct rtw_dev *rtwdev,
 void rtw_txq_init(struct rtw_dev *rtwdev, struct ieee80211_txq *txq);
 void rtw_txq_cleanup(struct rtw_dev *rtwdev, struct ieee80211_txq *txq);
 void rtw_tx_work(struct work_struct *w);
+void __rtw_tx_work(struct rtw_dev *rtwdev);
 void rtw_tx_pkt_info_update(struct rtw_dev *rtwdev,
 			    struct rtw_tx_pkt_info *pkt_info,
 			    struct ieee80211_sta *sta,