mt76: mt7663s: make all of packets 4-bytes aligned in sdio tx aggregation

Each packet should be padded with the additional zero to become 4-bytes
alignment in sdio tx aggregation.

Fixes: 1522ff73 ("mt76: mt7663s: introduce sdio tx aggregation")
Signed-off-by: Sean Wang <sean.wang@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

drivers/net/wireless/mediatek/mt76/mt7615/sdio_txrx.c

@@ -218,6 +218,7 @@ static int mt7663s_tx_run_queue(struct mt76_dev *dev, struct mt76_queue *q)
 	int qid, err, nframes = 0, len = 0, pse_sz = 0, ple_sz = 0;
 	bool mcu = q == dev->q_mcu[MT_MCUQ_WM];
 	struct mt76_sdio *sdio = &dev->sdio;
+	u8 pad;
 
 	qid = mcu ? ARRAY_SIZE(sdio->xmit_buf) - 1 : q->qid;
 	while (q->first != q->head) {
@@ -234,7 +235,8 @@ static int mt7663s_tx_run_queue(struct mt76_dev *dev, struct mt76_queue *q)
 			goto next;
 		}
 
-		if (len + e->skb->len + 4 > MT76S_XMIT_BUF_SZ)
+		pad = roundup(e->skb->len, 4) - e->skb->len;
+		if (len + e->skb->len + pad + 4 > MT76S_XMIT_BUF_SZ)
 			break;
 
 		if (mt7663s_tx_pick_quota(sdio, mcu, e->buf_sz, &pse_sz,
@@ -252,6 +254,11 @@ static int mt7663s_tx_run_queue(struct mt76_dev *dev, struct mt76_queue *q)
 			len += iter->len;
 			nframes++;
 		}
+
+		if (unlikely(pad)) {
+			memset(sdio->xmit_buf[qid] + len, 0, pad);
+			len += pad;
+		}
 next:
 		q->first = (q->first + 1) % q->ndesc;
 		e->done = true;