rtw89: pci: reclaim TX BD only if it really need

To reclaim TX BD, we need to read hardware reading index to determine if
any DMA is complete. Since this IO spends time, do this thing only if we
really need it when TX BD has no free buffer corresponding to target skb.

The experimental result shows that reading counter decreases from 26,000
to 130 per second.
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/20220421120903.73715-4-pkshih@realtek.com

drivers/net/wireless/realtek/rtw89/pci.c

@@ -412,9 +412,12 @@ static void rtw89_pci_release_txwd_skb(struct rtw89_dev *rtwdev,
 	u8 txch = tx_ring->txch;
 
 	if (!list_empty(&txwd->list)) {
-		rtw89_warn(rtwdev, "queue %d txwd %d is not idle\n",
-			   txch, seq);
-		return;
+		rtw89_pci_reclaim_txbd(rtwdev, tx_ring);
+		if (!list_empty(&txwd->list)) {
+			rtw89_warn(rtwdev, "queue %d txwd %d is not idle\n",
+				   txch, seq);
+			return;
+		}
 	}
 
 	/* currently, support for only one frame */
@@ -458,7 +461,6 @@ static void rtw89_pci_release_rpp(struct rtw89_dev *rtwdev,
 	}
 
 	tx_ring = &rtwpci->tx_rings[txch];
-	rtw89_pci_reclaim_txbd(rtwdev, tx_ring);
 	wd_ring = &tx_ring->wd_ring;
 	txwd = &wd_ring->pages[seq];
 
@@ -915,6 +917,10 @@ static u32 __rtw89_pci_check_and_reclaim_tx_resource(struct rtw89_dev *rtwdev,
 		if (!cnt)
 			goto out_unlock;
 		rtw89_pci_release_tx(rtwdev, rx_ring, cnt);
+
+		bd_cnt = rtw89_pci_get_avail_txbd_num(tx_ring);
+		if (bd_cnt == 0)
+			rtw89_pci_reclaim_txbd(rtwdev, tx_ring);
 	}
 
 	bd_cnt = rtw89_pci_get_avail_txbd_num(tx_ring);