[netdrvr 8139cp] support NAPI on RX path; Ditch RX frag handling.

NAPI is turned on unconditionally for the RX path.  The hardware
supports interrupt mitigation, so that should be investigated too.

RX fragment handling removed.  We simply ensure that we alloc
buffers large enough to hold incoming packets.  Any stray RX
frags that occur (shouldn't be any) will be dropped.

drivers/net/8139cp.c

@@ -290,12 +290,11 @@ enum {
 	UWF             = (1 << 4),  /* Accept Unicast wakeup frame */
 	LANWake         = (1 << 1),  /* Enable LANWake signal */
 	PMEStatus	= (1 << 0),  /* PME status can be reset by PCI RST# */
-};
 
-static const unsigned int cp_intr_mask =
-	PciErr | LinkChg |
-	RxOK | RxErr | RxEmpty | RxFIFOOvr |
-	TxOK | TxErr | TxEmpty;
+	cp_norx_intr_mask = PciErr | LinkChg | TxOK | TxErr | TxEmpty,
+	cp_rx_intr_mask = RxOK | RxErr | RxEmpty | RxFIFOOvr,
+	cp_intr_mask = cp_rx_intr_mask | cp_norx_intr_mask,
+};
 
 static const unsigned int cp_rx_config =
 	  (RX_FIFO_THRESH << RxCfgFIFOShift) |
@@ -469,7 +468,7 @@ static inline void cp_rx_skb (struct cp_private *cp, struct sk_buff *skb,
 		vlan_hwaccel_rx(skb, cp->vlgrp, be16_to_cpu(desc->opts2 & 0xffff));
 	} else
 #endif
-		netif_rx(skb);
+		netif_receive_skb(skb);
 }
 
 static void cp_rx_err_acct (struct cp_private *cp, unsigned rx_tail,
@@ -484,81 +483,14 @@ static void cp_rx_err_acct (struct cp_private *cp, unsigned rx_tail,
 		cp->net_stats.rx_frame_errors++;
 	if (status & RxErrCRC)
 		cp->net_stats.rx_crc_errors++;
-	if (status & RxErrRunt)
+	if ((status & RxErrRunt) || (status & RxErrLong))
 		cp->net_stats.rx_length_errors++;
-	if (status & RxErrLong)
+	if ((status & (FirstFrag | LastFrag)) != (FirstFrag | LastFrag))
 		cp->net_stats.rx_length_errors++;
 	if (status & RxErrFIFO)
 		cp->net_stats.rx_fifo_errors++;
 }
 
-static void cp_rx_frag (struct cp_private *cp, unsigned rx_tail,
-			struct sk_buff *skb, u32 status, u32 len)
-{
-	struct sk_buff *copy_skb, *frag_skb = cp->frag_skb;
-	unsigned orig_len = frag_skb ? frag_skb->len : 0;
-	unsigned target_len = orig_len + len;
-	unsigned first_frag = status & FirstFrag;
-	unsigned last_frag = status & LastFrag;
-
-	if (netif_msg_rx_status (cp))
-		printk (KERN_DEBUG "%s: rx %s%sfrag, slot %d status 0x%x len %d\n",
-			cp->dev->name,
-			cp->dropping_frag ? "dropping " : "",
-			first_frag ? "first " :
-			last_frag ? "last " : "",
-			rx_tail, status, len);
-
-	cp->cp_stats.rx_frags++;
-
-	if (!frag_skb && !first_frag)
-		cp->dropping_frag = 1;
-	if (cp->dropping_frag)
-		goto drop_frag;
-
-	copy_skb = dev_alloc_skb (target_len + RX_OFFSET);
-	if (!copy_skb) {
-		printk(KERN_WARNING "%s: rx slot %d alloc failed\n",
-		       cp->dev->name, rx_tail);
-
-		cp->dropping_frag = 1;
-drop_frag:
-		if (frag_skb) {
-			dev_kfree_skb_irq(frag_skb);
-			cp->frag_skb = NULL;
-		}
-		if (last_frag) {
-			cp->net_stats.rx_dropped++;
-			cp->dropping_frag = 0;
-		}
-		return;
-	}
-
-	copy_skb->dev = cp->dev;
-	skb_reserve(copy_skb, RX_OFFSET);
-	skb_put(copy_skb, target_len);
-	if (frag_skb) {
-		memcpy(copy_skb->data, frag_skb->data, orig_len);
-		dev_kfree_skb_irq(frag_skb);
-	}
-	pci_dma_sync_single(cp->pdev, cp->rx_skb[rx_tail].mapping,
-			    len, PCI_DMA_FROMDEVICE);
-	memcpy(copy_skb->data + orig_len, skb->data, len);
-
-	copy_skb->ip_summed = CHECKSUM_NONE;
-
-	if (last_frag) {
-		if (status & (RxError | RxErrFIFO)) {
-			cp_rx_err_acct(cp, rx_tail, status, len);
-			dev_kfree_skb_irq(copy_skb);
-		} else
-			cp_rx_skb(cp, copy_skb, &cp->rx_ring[rx_tail]);
-		cp->frag_skb = NULL;
-	} else {
-		cp->frag_skb = copy_skb;
-	}
-}
-
 static inline unsigned int cp_rx_csum_ok (u32 status)
 {
 	unsigned int protocol = (status >> 16) & 0x3;
@@ -572,12 +504,17 @@ static inline unsigned int cp_rx_csum_ok (u32 status)
 	return 0;
 }
 
-static void cp_rx (struct cp_private *cp)
+static int cp_rx_poll (struct net_device *dev, int *budget)
 {
+	struct cp_private *cp = dev->priv;
 	unsigned rx_tail = cp->rx_tail;
-	unsigned rx_work = 100;
+	unsigned rx_work = dev->quota;
+	unsigned rx = 0;
+
+rx_status_loop:
+	cpw16(IntrStatus, cp_rx_intr_mask);
 
-	while (rx_work--) {
+	while (1) {
 		u32 status, len;
 		dma_addr_t mapping;
 		struct sk_buff *skb, *new_skb;
@@ -597,7 +534,13 @@ static void cp_rx (struct cp_private *cp)
 		mapping = cp->rx_skb[rx_tail].mapping;
 
 		if ((status & (FirstFrag | LastFrag)) != (FirstFrag | LastFrag)) {
-			cp_rx_frag(cp, rx_tail, skb, status, len);
+			/* we don't support incoming fragmented frames.
+			 * instead, we attempt to ensure that the
+			 * pre-allocated RX skbs are properly sized such
+			 * that RX fragments are never encountered
+			 */
+			cp_rx_err_acct(cp, rx_tail, status, len);
+			cp->net_stats.rx_dropped++;
 			goto rx_next;
 		}
 
@@ -638,6 +581,7 @@ static void cp_rx (struct cp_private *cp)
 		cp->rx_skb[rx_tail].skb = new_skb;
 
 		cp_rx_skb(cp, skb, desc);
+		rx++;
 
 rx_next:
 		cp->rx_ring[rx_tail].opts2 = 0;
@@ -648,12 +592,30 @@ static void cp_rx (struct cp_private *cp)
 		else
 			desc->opts1 = cpu_to_le32(DescOwn | cp->rx_buf_sz);
 		rx_tail = NEXT_RX(rx_tail);
-	}
 
-	if (!rx_work)
-		printk(KERN_WARNING "%s: rx work limit reached\n", cp->dev->name);
+		if (!rx_work--)
+			break;
+	}
 
 	cp->rx_tail = rx_tail;
+
+	dev->quota -= rx;
+	*budget -= rx;
+
+	/* if we did not reach work limit, then we're done with
+	 * this round of polling
+	 */
+	if (rx_work) {
+		if (cpr16(IntrStatus) & cp_rx_intr_mask)
+			goto rx_status_loop;
+
+		cpw16_f(IntrMask, cp_intr_mask);
+		netif_rx_complete(dev);
+
+		return 0;	/* done */
+	}
+
+	return 1;		/* not done */
 }
 
 static irqreturn_t
@@ -671,12 +633,16 @@ cp_interrupt (int irq, void *dev_instance, struct pt_regs *regs)
 		printk(KERN_DEBUG "%s: intr, status %04x cmd %02x cpcmd %04x\n",
 		        dev->name, status, cpr8(Cmd), cpr16(CpCmd));
 
-	cpw16_f(IntrStatus, status);
+	cpw16(IntrStatus, status & ~cp_rx_intr_mask);
 
 	spin_lock(&cp->lock);
 
-	if (status & (RxOK | RxErr | RxEmpty | RxFIFOOvr))
-		cp_rx(cp);
+	if (status & (RxOK | RxErr | RxEmpty | RxFIFOOvr)) {
+		if (netif_rx_schedule_prep(dev)) {
+			cpw16_f(IntrMask, cp_norx_intr_mask);
+			__netif_rx_schedule(dev);
+		}
+	}
 	if (status & (TxOK | TxErr | TxEmpty | SWInt))
 		cp_tx(cp);
 	if (status & LinkChg)
@@ -689,6 +655,8 @@ cp_interrupt (int irq, void *dev_instance, struct pt_regs *regs)
 		pci_write_config_word(cp->pdev, PCI_STATUS, pci_status);
 		printk(KERN_ERR "%s: PCI bus error, status=%04x, PCI status=%04x\n",
 		       dev->name, status, pci_status);
+
+		/* TODO: reset hardware */
 	}
 
 	spin_unlock(&cp->lock);
@@ -986,11 +954,10 @@ static void cp_stop_hw (struct cp_private *cp)
 {
 	struct net_device *dev = cp->dev;
 
-	cpw16(IntrMask, 0);
-	cpr16(IntrMask);
+	cpw16(IntrStatus, ~(cpr16(IntrStatus)));
+	cpw16_f(IntrMask, 0);
 	cpw8(Cmd, 0);
-	cpw16(CpCmd, 0);
-	cpr16(CpCmd);
+	cpw16_f(CpCmd, 0);
 	cpw16(IntrStatus, ~(cpr16(IntrStatus)));
 	synchronize_irq(dev->irq);
 	udelay(10);
@@ -1779,6 +1746,8 @@ static int __devinit cp_init_one (struct pci_dev *pdev,
 	dev->hard_start_xmit = cp_start_xmit;
 	dev->get_stats = cp_get_stats;
 	dev->do_ioctl = cp_ioctl;
+	dev->poll = cp_rx_poll;
+	dev->weight = 16;	/* arbitrary? from NAPI_HOWTO.txt. */
 #ifdef BROKEN
 	dev->change_mtu = cp_change_mtu;
 #endif