rtw88: replace tx tasklet with work queue

Replace tasklet so we can do tx scheduling in parallel. Since throughput
is delay-sensitive in most cases, we allocate a dedicated, high priority
wq for our needs.
Signed-off-by: Po-Hao Huang <phhuang@realtek.com>
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Reviewed-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20210209070755.23019-5-pkshih@realtek.com

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

@@ -42,7 +42,7 @@ static void rtw_ops_wake_tx_queue(struct ieee80211_hw *hw,
 		list_add_tail(&rtwtxq->list, &rtwdev->txqs);
 	spin_unlock_bh(&rtwdev->txq_lock);
 
-	tasklet_schedule(&rtwdev->tx_tasklet);
+	queue_work(rtwdev->tx_wq, &rtwdev->tx_work);
 }
 
 static int rtw_ops_start(struct ieee80211_hw *hw)

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

@@ -1658,7 +1658,7 @@ int rtw_core_init(struct rtw_dev *rtwdev)
 
 	timer_setup(&rtwdev->tx_report.purge_timer,
 		    rtw_tx_report_purge_timer, 0);
-	tasklet_setup(&rtwdev->tx_tasklet, rtw_tx_tasklet);
+	rtwdev->tx_wq = alloc_workqueue("rtw_tx_wq", WQ_UNBOUND | WQ_HIGHPRI, 0);
 
 	INIT_DELAYED_WORK(&rtwdev->watch_dog_work, rtw_watch_dog_work);
 	INIT_DELAYED_WORK(&coex->bt_relink_work, rtw_coex_bt_relink_work);
@@ -1670,6 +1670,7 @@ int rtw_core_init(struct rtw_dev *rtwdev)
 	INIT_DELAYED_WORK(&coex->bt_multi_link_remain_work,
 			  rtw_coex_bt_multi_link_remain_work);
 	INIT_DELAYED_WORK(&coex->wl_ccklock_work, rtw_coex_wl_ccklock_work);
+	INIT_WORK(&rtwdev->tx_work, rtw_tx_work);
 	INIT_WORK(&rtwdev->c2h_work, rtw_c2h_work);
 	INIT_WORK(&rtwdev->fw_recovery_work, rtw_fw_recovery_work);
 	INIT_WORK(&rtwdev->ba_work, rtw_txq_ba_work);
@@ -1736,7 +1737,7 @@ void rtw_core_deinit(struct rtw_dev *rtwdev)
 	if (wow_fw->firmware)
 		release_firmware(wow_fw->firmware);
 
-	tasklet_kill(&rtwdev->tx_tasklet);
+	destroy_workqueue(rtwdev->tx_wq);
 	spin_lock_irqsave(&rtwdev->tx_report.q_lock, flags);
 	skb_queue_purge(&rtwdev->tx_report.queue);
 	spin_unlock_irqrestore(&rtwdev->tx_report.q_lock, flags);

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

@@ -13,6 +13,7 @@
 #include <linux/bitfield.h>
 #include <linux/iopoll.h>
 #include <linux/interrupt.h>
+#include <linux/workqueue.h>
 
 #include "util.h"
 
@@ -1772,7 +1773,8 @@ struct rtw_dev {
 	/* used to protect txqs list */
 	spinlock_t txq_lock;
 	struct list_head txqs;
-	struct tasklet_struct tx_tasklet;
+	struct workqueue_struct *tx_wq;
+	struct work_struct tx_work;
 	struct work_struct ba_work;
 
 	struct rtw_tx_report tx_report;

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

@@ -592,9 +592,9 @@ static void rtw_txq_push(struct rtw_dev *rtwdev,
 	rcu_read_unlock();
 }
 
-void rtw_tx_tasklet(struct tasklet_struct *t)
+void rtw_tx_work(struct work_struct *w)
 {
-	struct rtw_dev *rtwdev = from_tasklet(rtwdev, t, tx_tasklet);
+	struct rtw_dev *rtwdev = container_of(w, struct rtw_dev, tx_work);
 	struct rtw_txq *rtwtxq, *tmp;
 
 	spin_lock_bh(&rtwdev->txq_lock);

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

@@ -98,7 +98,7 @@ void rtw_tx(struct rtw_dev *rtwdev,
 	    struct sk_buff *skb);
 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_tasklet(struct tasklet_struct *t);
+void rtw_tx_work(struct work_struct *w);
 void rtw_tx_pkt_info_update(struct rtw_dev *rtwdev,
 			    struct rtw_tx_pkt_info *pkt_info,
 			    struct ieee80211_sta *sta,