mt7601u: fix tx status reporting contexts

mac80211 requires that rx path does not run concurrently with
tx status reporting.  Since rx path is run in driver tasklet,
tx status cannot be reported directly from interrupt context
(there would be no way to lock it out).

Add tasklet for tx and move all possible code from irq handler
there.

Note: tx tasklet is needed because workqueue is queued very
      rarely and that kills TCP performance.
Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

drivers/net/wireless/mediatek/mt7601u/dma.c

@@ -236,23 +236,42 @@ static void mt7601u_complete_tx(struct urb *urb)
 	skb = q->e[q->start].skb;
 	trace_mt_tx_dma_done(dev, skb);
 
-	mt7601u_tx_status(dev, skb);
+	__skb_queue_tail(&dev->tx_skb_done, skb);
+	tasklet_schedule(&dev->tx_tasklet);
 
 	if (q->used == q->entries - q->entries / 8)
 		ieee80211_wake_queue(dev->hw, skb_get_queue_mapping(skb));
 
 	q->start = (q->start + 1) % q->entries;
 	q->used--;
+out:
+	spin_unlock_irqrestore(&dev->tx_lock, flags);
+}
 
-	if (urb->status)
-		goto out;
+static void mt7601u_tx_tasklet(unsigned long data)
+{
+	struct mt7601u_dev *dev = (struct mt7601u_dev *) data;
+	struct sk_buff_head skbs;
+	unsigned long flags;
+
+	__skb_queue_head_init(&skbs);
+
+	spin_lock_irqsave(&dev->tx_lock, flags);
 
 	set_bit(MT7601U_STATE_MORE_STATS, &dev->state);
 	if (!test_and_set_bit(MT7601U_STATE_READING_STATS, &dev->state))
 		queue_delayed_work(dev->stat_wq, &dev->stat_work,
 				   msecs_to_jiffies(10));
-out:
+
+	skb_queue_splice_init(&dev->tx_skb_done, &skbs);
+
 	spin_unlock_irqrestore(&dev->tx_lock, flags);
+
+	while (!skb_queue_empty(&skbs)) {
+		struct sk_buff *skb = __skb_dequeue(&skbs);
+
+		mt7601u_tx_status(dev, skb);
+	}
 }
 
 static int mt7601u_dma_submit_tx(struct mt7601u_dev *dev,
@@ -475,6 +494,7 @@ int mt7601u_dma_init(struct mt7601u_dev *dev)
 {
 	int ret = -ENOMEM;
 
+	tasklet_init(&dev->tx_tasklet, mt7601u_tx_tasklet, (unsigned long) dev);
 	tasklet_init(&dev->rx_tasklet, mt7601u_rx_tasklet, (unsigned long) dev);
 
 	ret = mt7601u_alloc_tx(dev);
@@ -502,4 +522,6 @@ void mt7601u_dma_cleanup(struct mt7601u_dev *dev)
 
 	mt7601u_free_rx(dev);
 	mt7601u_free_tx(dev);
+
+	tasklet_kill(&dev->tx_tasklet);
 }

drivers/net/wireless/mediatek/mt7601u/init.c

@@ -456,6 +456,7 @@ struct mt7601u_dev *mt7601u_alloc_device(struct device *pdev)
 	spin_lock_init(&dev->lock);
 	spin_lock_init(&dev->con_mon_lock);
 	atomic_set(&dev->avg_ampdu_len, 1);
+	skb_queue_head_init(&dev->tx_skb_done);
 
 	dev->stat_wq = alloc_workqueue("mt7601u", WQ_UNBOUND, 0);
 	if (!dev->stat_wq) {

drivers/net/wireless/mediatek/mt7601u/mac.c

@@ -181,7 +181,11 @@ void mt76_send_tx_status(struct mt7601u_dev *dev, struct mt76_tx_status *stat)
 	}
 
 	mt76_mac_fill_tx_status(dev, &info, stat);
+
+	local_bh_disable();
 	ieee80211_tx_status_noskb(dev->hw, sta, &info);
+	local_bh_enable();
+
 	rcu_read_unlock();
 }
 

drivers/net/wireless/mediatek/mt7601u/mt7601u.h

@@ -199,7 +199,9 @@ struct mt7601u_dev {
 
 	/* TX */
 	spinlock_t tx_lock;
+	struct tasklet_struct tx_tasklet;
 	struct mt7601u_tx_queue *tx_q;
+	struct sk_buff_head tx_skb_done;
 
 	atomic_t avg_ampdu_len;