[PATCH] sky2: incremental transmit completion

Since a transmit can take several control blocks, the old code waited
until the last control block was marked as done. This code processes
the return values incrementally. This makes slots in the tx ring available
and less chance of getting stuck.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

1 files changed, 47 insertions, 48 deletions

diff --git a/drivers/net/sky2.c b/drivers/net/sky2.c
index b908596..64af764 100644
--- a/drivers/net/sky2.c
+++ b/drivers/net/sky2.c
@@ -769,9 +769,16 @@ static inline struct sky2_tx_le *get_tx_le(struct sky2_port *sky2)
 	struct sky2_tx_le *le = sky2->tx_le + sky2->tx_prod;
 
 	sky2->tx_prod = RING_NEXT(sky2->tx_prod, TX_RING_SIZE);
+	le->ctrl = 0;
 	return le;
 }
 
+static inline struct tx_ring_info *tx_le_re(struct sky2_port *sky2,
+					    struct sky2_tx_le *le)
+{
+	return sky2->tx_ring + (le - sky2->tx_le);
+}
+
 /* Update chip's next pointer */
 static inline void sky2_put_idx(struct sky2_hw *hw, unsigned q, u16 idx)
 {
@@ -786,6 +793,7 @@ static inline struct sky2_rx_le *sky2_next_rx(struct sky2_port *sky2)
 {
 	struct sky2_rx_le *le = sky2->rx_le + sky2->rx_put;
 	sky2->rx_put = RING_NEXT(sky2->rx_put, RX_LE_SIZE);
+	le->ctrl = 0;
 	return le;
 }
 
@@ -805,7 +813,6 @@ static void sky2_rx_add(struct sky2_port *sky2, dma_addr_t map)
 	if (sky2->rx_addr64 != hi) {
 		le = sky2_next_rx(sky2);
 		le->addr = cpu_to_le32(hi);
-		le->ctrl = 0;
 		le->opcode = OP_ADDR64 | HW_OWNER;
 		sky2->rx_addr64 = high32(map + len);
 	}
@@ -813,7 +820,6 @@ static void sky2_rx_add(struct sky2_port *sky2, dma_addr_t map)
 	le = sky2_next_rx(sky2);
 	le->addr = cpu_to_le32((u32) map);
 	le->length = cpu_to_le16(len);
-	le->ctrl = 0;
 	le->opcode = OP_PACKET | HW_OWNER;
 }
 
@@ -877,7 +883,7 @@ static void sky2_rx_clean(struct sky2_port *sky2)
 
 	memset(sky2->rx_le, 0, RX_LE_BYTES);
 	for (i = 0; i < sky2->rx_pending; i++) {
-		struct ring_info *re = sky2->rx_ring + i;
+		struct rx_ring_info *re = sky2->rx_ring + i;
 
 		if (re->skb) {
 			pci_unmap_single(sky2->hw->pdev,
@@ -1008,7 +1014,7 @@ static int sky2_rx_start(struct sky2_port *sky2)
 
 	rx_set_checksum(sky2);
 	for (i = 0; i < sky2->rx_pending; i++) {
-		struct ring_info *re = sky2->rx_ring + i;
+		struct rx_ring_info *re = sky2->rx_ring + i;
 
 		re->skb = sky2_alloc_skb(sky2->netdev, sky2->rx_bufsize,
 					 GFP_KERNEL);
@@ -1094,7 +1100,7 @@ static int sky2_up(struct net_device *dev)
 		goto err_out;
 	memset(sky2->rx_le, 0, RX_LE_BYTES);
 
-	sky2->rx_ring = kcalloc(sky2->rx_pending, sizeof(struct ring_info),
+	sky2->rx_ring = kcalloc(sky2->rx_pending, sizeof(struct rx_ring_info),
 				GFP_KERNEL);
 	if (!sky2->rx_ring)
 		goto err_out;
@@ -1241,13 +1247,10 @@ static int sky2_xmit_frame(struct sk_buff *skb, struct net_device *dev)
 	mapping = pci_map_single(hw->pdev, skb->data, len, PCI_DMA_TODEVICE);
 	addr64 = high32(mapping);
 
-	re = sky2->tx_ring + sky2->tx_prod;
-
 	/* Send high bits if changed or crosses boundary */
 	if (addr64 != sky2->tx_addr64 || high32(mapping + len) != sky2->tx_addr64) {
 		le = get_tx_le(sky2);
 		le->addr = cpu_to_le32(addr64);
-		le->ctrl = 0;
 		le->opcode = OP_ADDR64 | HW_OWNER;
 		sky2->tx_addr64 = high32(mapping + len);
 	}
@@ -1263,7 +1266,6 @@ static int sky2_xmit_frame(struct sk_buff *skb, struct net_device *dev)
 			le = get_tx_le(sky2);
 			le->addr = cpu_to_le32(mss);
 			le->opcode = OP_LRGLEN | HW_OWNER;
-			le->ctrl = 0;
 			sky2->tx_last_mss = mss;
 		}
 	}
@@ -1276,7 +1278,6 @@ static int sky2_xmit_frame(struct sk_buff *skb, struct net_device *dev)
 			le = get_tx_le(sky2);
 			le->addr = 0;
 			le->opcode = OP_VLAN|HW_OWNER;
-			le->ctrl = 0;
 		} else
 			le->opcode |= OP_VLAN;
 		le->length = cpu_to_be16(vlan_tx_tag_get(skb));
@@ -1313,13 +1314,13 @@ static int sky2_xmit_frame(struct sk_buff *skb, struct net_device *dev)
 	le->ctrl = ctrl;
 	le->opcode = mss ? (OP_LARGESEND | HW_OWNER) : (OP_PACKET | HW_OWNER);
 
-	/* Record the transmit mapping info */
+	re = tx_le_re(sky2, le);
 	re->skb = skb;
 	pci_unmap_addr_set(re, mapaddr, mapping);
+	pci_unmap_len_set(re, maplen, len);
 
 	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
-		skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
-		struct tx_ring_info *fre;
+		const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
 
 		mapping = pci_map_page(hw->pdev, frag->page, frag->page_offset,
 				       frag->size, PCI_DMA_TODEVICE);
@@ -1338,12 +1339,12 @@ static int sky2_xmit_frame(struct sk_buff *skb, struct net_device *dev)
 		le->ctrl = ctrl;
 		le->opcode = OP_BUFFER | HW_OWNER;
 
-		fre = sky2->tx_ring
-			+ RING_NEXT((re - sky2->tx_ring) + i, TX_RING_SIZE);
-		pci_unmap_addr_set(fre, mapaddr, mapping);
+		re = tx_le_re(sky2, le);
+		re->skb = skb;
+		pci_unmap_addr_set(re, mapaddr, mapping);
+		pci_unmap_len_set(re, maplen, frag->size);
 	}
 
-	re->idx = sky2->tx_prod;
 	le->ctrl |= EOP;
 
 	if (tx_avail(sky2) <= MAX_SKB_TX_LE)
@@ -1361,49 +1362,47 @@ static int sky2_xmit_frame(struct sk_buff *skb, struct net_device *dev)
  * Free ring elements from starting at tx_cons until "done"
  *
  * NB: the hardware will tell us about partial completion of multi-part
- *     buffers; these are deferred until completion.
+ *     buffers so make sure not to free skb to early.
  */
 static void sky2_tx_complete(struct sky2_port *sky2, u16 done)
 {
 	struct net_device *dev = sky2->netdev;
 	struct pci_dev *pdev = sky2->hw->pdev;
-	u16 nxt, put;
-	unsigned i;
+	unsigned idx;
 
 	BUG_ON(done >= TX_RING_SIZE);
 
-	if (unlikely(netif_msg_tx_done(sky2)))
-		printk(KERN_DEBUG "%s: tx done, up to %u\n",
-		       dev->name, done);
-
-	for (put = sky2->tx_cons; put != done; put = nxt) {
-		struct tx_ring_info *re = sky2->tx_ring + put;
-		struct sk_buff *skb = re->skb;
-
-		nxt = re->idx;
-		BUG_ON(nxt >= TX_RING_SIZE);
-		prefetch(sky2->tx_ring + nxt);
-
-		/* Check for partial status */
-		if (tx_dist(put, done) < tx_dist(put, nxt))
+	for (idx = sky2->tx_cons; idx != done;
+	     idx = RING_NEXT(idx, TX_RING_SIZE)) {
+		struct sky2_tx_le *le = sky2->tx_le + idx;
+		struct tx_ring_info *re = sky2->tx_ring + idx;
+
+		switch(le->opcode & ~HW_OWNER) {
+		case OP_LARGESEND:
+		case OP_PACKET:
+			pci_unmap_single(pdev,
+					 pci_unmap_addr(re, mapaddr),
+					 pci_unmap_len(re, maplen),
+					 PCI_DMA_TODEVICE);
 			break;
-
-		skb = re->skb;
-		pci_unmap_single(pdev, pci_unmap_addr(re, mapaddr),
-				 skb_headlen(skb), PCI_DMA_TODEVICE);
-
-		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
-			struct tx_ring_info *fre;
-			fre = sky2->tx_ring + RING_NEXT(put + i, TX_RING_SIZE);
-			pci_unmap_page(pdev, pci_unmap_addr(fre, mapaddr),
-				       skb_shinfo(skb)->frags[i].size,
+		case OP_BUFFER:
+			pci_unmap_page(pdev, pci_unmap_addr(re, mapaddr),
+				       pci_unmap_len(re, maplen),
 				       PCI_DMA_TODEVICE);
+			break;
+		}
+
+		if (le->ctrl & EOP) {
+			if (unlikely(netif_msg_tx_done(sky2)))
+				printk(KERN_DEBUG "%s: tx done %u\n",
+				       dev->name, idx);
+			dev_kfree_skb(re->skb);
 		}
 
-		dev_kfree_skb(skb);
+		le->opcode = 0;	/* paranoia */
 	}
 
-	sky2->tx_cons = put;
+	sky2->tx_cons = idx;
 	if (tx_avail(sky2) > MAX_SKB_TX_LE + 4)
 		netif_wake_queue(dev);
 }
@@ -1843,7 +1842,7 @@ static struct sk_buff *sky2_receive(struct net_device *dev,
 				    u16 length, u32 status)
 {
  	struct sky2_port *sky2 = netdev_priv(dev);
-	struct ring_info *re = sky2->rx_ring + sky2->rx_next;
+	struct rx_ring_info *re = sky2->rx_ring + sky2->rx_next;
 	struct sk_buff *skb = NULL;
 
 	if (unlikely(netif_msg_rx_status(sky2)))
@@ -1889,7 +1888,7 @@ static struct sk_buff *sky2_receive(struct net_device *dev,
 		prefetch(skb->data);
 
 		re->mapaddr = pci_map_single(sky2->hw->pdev, nskb->data,
-					     sky2->rx_bufsize, PCI_DMA_FROMDEVICE);
+					 sky2->rx_bufsize, PCI_DMA_FROMDEVICE);
 	}
 
 	skb_put(skb, length);