Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next

Pull networking changes from David S Miller:

 1) Remove the ipv4 routing cache.  Now lookups go directly into the FIB
    trie and use prebuilt routes cached there.

    No more garbage collection, no more rDOS attacks on the routing
    cache.  Instead we now get predictable and consistent performance,
    no matter what the pattern of traffic we service.

    This has been almost 2 years in the making.  Special thanks to
    Julian Anastasov, Eric Dumazet, Steffen Klassert, and others who
    have helped along the way.

    I'm sure that with a change of this magnitude there will be some
    kind of fallout, but such things ought the be simple to fix at this
    point.  Luckily I'm not European so I'll be around all of August to
    fix things :-)

    The major stages of this work here are each fronted by a forced
    merge commit whose commit message contains a top-level description
    of the motivations and implementation issues.

 2) Pre-demux of established ipv4 TCP sockets, saves a route demux on
    input.

 3) TCP SYN/ACK performance tweaks from Eric Dumazet.

 4) Add namespace support for netfilter L4 conntrack helpers, from Gao
    Feng.

 5) Add config mechanism for Energy Efficient Ethernet to ethtool, from
    Yuval Mintz.

 6) Remove quadratic behavior from /proc/net/unix, from Eric Dumazet.

 7) Support for connection tracker helpers in userspace, from Pablo
    Neira Ayuso.

 8) Allow userspace driven TX load balancing functions in TEAM driver,
    from Jiri Pirko.

 9) Kill off NLMSG_PUT and RTA_PUT macros, more gross stuff with
    embedded gotos.

10) TCP Small Queues, essentially minimize the amount of TCP data queued
    up in the packet scheduler layer.  Whereas the existing BQL (Byte
    Queue Limits) limits the pkt_sched --> netdevice queuing levels,
    this controls the TCP --> pkt_sched queueing levels.

    From Eric Dumazet.

11) Reduce the number of get_page/put_page ops done on SKB fragments,
    from Alexander Duyck.

12) Implement protection against blind resets in TCP (RFC 5961), from
    Eric Dumazet.

13) Support the client side of TCP Fast Open, basically the ability to
    send data in the SYN exchange, from Yuchung Cheng.

    Basically, the sender queues up data with a sendmsg() call using
    MSG_FASTOPEN, then they do the connect() which emits the queued up
    fastopen data.

14) Avoid all the problems we get into in TCP when timers or PMTU events
    hit a locked socket.  The TCP Small Queues changes added a
    tcp_release_cb() that allows us to queue work up to the
    release_sock() caller, and that's what we use here too.  From Eric
    Dumazet.

15) Zero copy on TX support for TUN driver, from Michael S. Tsirkin.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1870 commits)
  genetlink: define lockdep_genl_is_held() when CONFIG_LOCKDEP
  r8169: revert "add byte queue limit support".
  ipv4: Change rt->rt_iif encoding.
  net: Make skb->skb_iif always track skb->dev
  ipv4: Prepare for change of rt->rt_iif encoding.
  ipv4: Remove all RTCF_DIRECTSRC handliing.
  ipv4: Really ignore ICMP address requests/replies.
  decnet: Don't set RTCF_DIRECTSRC.
  net/ipv4/ip_vti.c: Fix __rcu warnings detected by sparse.
  ipv4: Remove redundant assignment
  rds: set correct msg_namelen
  openvswitch: potential NULL deref in sample()
  tcp: dont drop MTU reduction indications
  bnx2x: Add new 57840 device IDs
  tcp: avoid oops in tcp_metrics and reset tcpm_stamp
  niu: Change niu_rbr_fill() to use unlikely() to check niu_rbr_add_page() return value
  niu: Fix to check for dma mapping errors.
  net: Fix references to out-of-scope variables in put_cmsg_compat()
  net: ethernet: davinci_emac: add pm_runtime support
  net: ethernet: davinci_emac: Remove unnecessary #include
  ...

1 files changed, 185 insertions, 56 deletions

diff --git a/drivers/net/can/janz-ican3.c b/drivers/net/can/janz-ican3.c
index 08c893cb7896..98ee43819911 100644
--- a/drivers/net/can/janz-ican3.c
+++ b/drivers/net/can/janz-ican3.c
@@ -116,6 +116,7 @@
 #define ICAN3_BUSERR_QUOTA_MAX	255
 
 /* Janz ICAN3 CAN Frame Conversion */
+#define ICAN3_SNGL	0x02
 #define ICAN3_ECHO	0x10
 #define ICAN3_EFF_RTR	0x40
 #define ICAN3_SFF_RTR	0x10
@@ -220,6 +221,9 @@ struct ican3_dev {
 	/* old and new style host interface */
 	unsigned int iftype;
 
+	/* queue for echo packets */
+	struct sk_buff_head echoq;
+
 	/*
 	 * Any function which changes the current DPM page must hold this
 	 * lock while it is performing data accesses. This ensures that the
@@ -235,7 +239,6 @@ struct ican3_dev {
 
 	/* fast host interface */
 	unsigned int fastrx_start;
-	unsigned int fastrx_int;
 	unsigned int fastrx_num;
 	unsigned int fasttx_start;
 	unsigned int fasttx_num;
@@ -454,7 +457,6 @@ static void __devinit ican3_init_fast_host_interface(struct ican3_dev *mod)
 	/* save the start recv page */
 	mod->fastrx_start = mod->free_page;
 	mod->fastrx_num = 0;
-	mod->fastrx_int = 0;
 
 	/* build a single fast tohost queue descriptor */
 	memset(&desc, 0, sizeof(desc));
@@ -813,10 +815,10 @@ static void ican3_to_can_frame(struct ican3_dev *mod,
 
 		cf->can_id |= desc->data[0] << 3;
 		cf->can_id |= (desc->data[1] & 0xe0) >> 5;
-		cf->can_dlc = desc->data[1] & ICAN3_CAN_DLC_MASK;
-		memcpy(cf->data, &desc->data[2], sizeof(cf->data));
+		cf->can_dlc = get_can_dlc(desc->data[1] & ICAN3_CAN_DLC_MASK);
+		memcpy(cf->data, &desc->data[2], cf->can_dlc);
 	} else {
-		cf->can_dlc = desc->data[0] & ICAN3_CAN_DLC_MASK;
+		cf->can_dlc = get_can_dlc(desc->data[0] & ICAN3_CAN_DLC_MASK);
 		if (desc->data[0] & ICAN3_EFF_RTR)
 			cf->can_id |= CAN_RTR_FLAG;
 
@@ -831,7 +833,7 @@ static void ican3_to_can_frame(struct ican3_dev *mod,
 			cf->can_id |= desc->data[3] >> 5;  /* 2-0   */
 		}
 
-		memcpy(cf->data, &desc->data[6], sizeof(cf->data));
+		memcpy(cf->data, &desc->data[6], cf->can_dlc);
 	}
 }
 
@@ -847,6 +849,10 @@ static void can_frame_to_ican3(struct ican3_dev *mod,
 	desc->data[0] |= cf->can_dlc;
 	desc->data[1] |= ICAN3_ECHO;
 
+	/* support single transmission (no retries) mode */
+	if (mod->can.ctrlmode & CAN_CTRLMODE_ONE_SHOT)
+		desc->data[1] |= ICAN3_SNGL;
+
 	if (cf->can_id & CAN_RTR_FLAG)
 		desc->data[0] |= ICAN3_EFF_RTR;
 
@@ -863,7 +869,7 @@ static void can_frame_to_ican3(struct ican3_dev *mod,
 	}
 
 	/* copy the data bits into the descriptor */
-	memcpy(&desc->data[6], cf->data, sizeof(cf->data));
+	memcpy(&desc->data[6], cf->data, cf->can_dlc);
 }
 
 /*
@@ -909,8 +915,8 @@ static void ican3_handle_msglost(struct ican3_dev *mod, struct ican3_msg *msg)
 	if (skb) {
 		cf->can_id |= CAN_ERR_CRTL;
 		cf->data[1] = CAN_ERR_CRTL_RX_OVERFLOW;
+		stats->rx_over_errors++;
 		stats->rx_errors++;
-		stats->rx_bytes += cf->can_dlc;
 		netif_rx(skb);
 	}
 }
@@ -927,7 +933,7 @@ static int ican3_handle_cevtind(struct ican3_dev *mod, struct ican3_msg *msg)
 	struct net_device *dev = mod->ndev;
 	struct net_device_stats *stats = &dev->stats;
 	enum can_state state = mod->can.state;
-	u8 status, isrc, rxerr, txerr;
+	u8 isrc, ecc, status, rxerr, txerr;
 	struct can_frame *cf;
 	struct sk_buff *skb;
 
@@ -943,15 +949,53 @@ static int ican3_handle_cevtind(struct ican3_dev *mod, struct ican3_msg *msg)
 		return -EINVAL;
 	}
 
-	skb = alloc_can_err_skb(dev, &cf);
-	if (skb == NULL)
-		return -ENOMEM;
-
 	isrc = msg->data[0];
+	ecc = msg->data[2];
 	status = msg->data[3];
 	rxerr = msg->data[4];
 	txerr = msg->data[5];
 
+	/*
+	 * This hardware lacks any support other than bus error messages to
+	 * determine if packet transmission has failed.
+	 *
+	 * When TX errors happen, one echo skb needs to be dropped from the
+	 * front of the queue.
+	 *
+	 * A small bit of code is duplicated here and below, to avoid error
+	 * skb allocation when it will just be freed immediately.
+	 */
+	if (isrc == CEVTIND_BEI) {
+		int ret;
+		dev_dbg(mod->dev, "bus error interrupt\n");
+
+		/* TX error */
+		if (!(ecc & ECC_DIR)) {
+			kfree_skb(skb_dequeue(&mod->echoq));
+			stats->tx_errors++;
+		} else {
+			stats->rx_errors++;
+		}
+
+		/*
+		 * The controller automatically disables bus-error interrupts
+		 * and therefore we must re-enable them.
+		 */
+		ret = ican3_set_buserror(mod, 1);
+		if (ret) {
+			dev_err(mod->dev, "unable to re-enable bus-error\n");
+			return ret;
+		}
+
+		/* bus error reporting is off, return immediately */
+		if (!(mod->can.ctrlmode & CAN_CTRLMODE_BERR_REPORTING))
+			return 0;
+	}
+
+	skb = alloc_can_err_skb(dev, &cf);
+	if (skb == NULL)
+		return -ENOMEM;
+
 	/* data overrun interrupt */
 	if (isrc == CEVTIND_DOI || isrc == CEVTIND_LOST) {
 		dev_dbg(mod->dev, "data overrun interrupt\n");
@@ -980,11 +1024,7 @@ static int ican3_handle_cevtind(struct ican3_dev *mod, struct ican3_msg *msg)
 
 	/* bus error interrupt */
 	if (isrc == CEVTIND_BEI) {
-		u8 ecc = msg->data[2];
-
-		dev_dbg(mod->dev, "bus error interrupt\n");
 		mod->can.can_stats.bus_error++;
-		stats->rx_errors++;
 		cf->can_id |= CAN_ERR_PROT | CAN_ERR_BUSERROR;
 
 		switch (ecc & ECC_MASK) {
@@ -1003,7 +1043,7 @@ static int ican3_handle_cevtind(struct ican3_dev *mod, struct ican3_msg *msg)
 			break;
 		}
 
-		if ((ecc & ECC_DIR) == 0)
+		if (!(ecc & ECC_DIR))
 			cf->data[2] |= CAN_ERR_PROT_TX;
 
 		cf->data[6] = txerr;
@@ -1030,8 +1070,6 @@ static int ican3_handle_cevtind(struct ican3_dev *mod, struct ican3_msg *msg)
 	}
 
 	mod->can.state = state;
-	stats->rx_errors++;
-	stats->rx_bytes += cf->can_dlc;
 	netif_rx(skb);
 	return 0;
 }
@@ -1091,6 +1129,88 @@ static void ican3_handle_message(struct ican3_dev *mod, struct ican3_msg *msg)
 }
 
 /*
+ * The ican3 needs to store all echo skbs, and therefore cannot
+ * use the generic infrastructure for this.
+ */
+static void ican3_put_echo_skb(struct ican3_dev *mod, struct sk_buff *skb)
+{
+	struct sock *srcsk = skb->sk;
+
+	if (atomic_read(&skb->users) != 1) {
+		struct sk_buff *old_skb = skb;
+
+		skb = skb_clone(old_skb, GFP_ATOMIC);
+		kfree_skb(old_skb);
+		if (!skb)
+			return;
+	} else {
+		skb_orphan(skb);
+	}
+
+	skb->sk = srcsk;
+
+	/* save this skb for tx interrupt echo handling */
+	skb_queue_tail(&mod->echoq, skb);
+}
+
+static unsigned int ican3_get_echo_skb(struct ican3_dev *mod)
+{
+	struct sk_buff *skb = skb_dequeue(&mod->echoq);
+	struct can_frame *cf;
+	u8 dlc;
+
+	/* this should never trigger unless there is a driver bug */
+	if (!skb) {
+		netdev_err(mod->ndev, "BUG: echo skb not occupied\n");
+		return 0;
+	}
+
+	cf = (struct can_frame *)skb->data;
+	dlc = cf->can_dlc;
+
+	/* check flag whether this packet has to be looped back */
+	if (skb->pkt_type != PACKET_LOOPBACK) {
+		kfree_skb(skb);
+		return dlc;
+	}
+
+	skb->protocol = htons(ETH_P_CAN);
+	skb->pkt_type = PACKET_BROADCAST;
+	skb->ip_summed = CHECKSUM_UNNECESSARY;
+	skb->dev = mod->ndev;
+	netif_receive_skb(skb);
+	return dlc;
+}
+
+/*
+ * Compare an skb with an existing echo skb
+ *
+ * This function will be used on devices which have a hardware loopback.
+ * On these devices, this function can be used to compare a received skb
+ * with the saved echo skbs so that the hardware echo skb can be dropped.
+ *
+ * Returns true if the skb's are identical, false otherwise.
+ */
+static bool ican3_echo_skb_matches(struct ican3_dev *mod, struct sk_buff *skb)
+{
+	struct can_frame *cf = (struct can_frame *)skb->data;
+	struct sk_buff *echo_skb = skb_peek(&mod->echoq);
+	struct can_frame *echo_cf;
+
+	if (!echo_skb)
+		return false;
+
+	echo_cf = (struct can_frame *)echo_skb->data;
+	if (cf->can_id != echo_cf->can_id)
+		return false;
+
+	if (cf->can_dlc != echo_cf->can_dlc)
+		return false;
+
+	return memcmp(cf->data, echo_cf->data, cf->can_dlc) == 0;
+}
+
+/*
  * Check that there is room in the TX ring to transmit another skb
  *
  * LOCKING: must hold mod->lock
@@ -1100,6 +1220,10 @@ static bool ican3_txok(struct ican3_dev *mod)
 	struct ican3_fast_desc __iomem *desc;
 	u8 control;
 
+	/* check that we have echo queue space */
+	if (skb_queue_len(&mod->echoq) >= ICAN3_TX_BUFFERS)
+		return false;
+
 	/* copy the control bits of the descriptor */
 	ican3_set_page(mod, mod->fasttx_start + (mod->fasttx_num / 16));
 	desc = mod->dpm + ((mod->fasttx_num % 16) * sizeof(*desc));
@@ -1150,10 +1274,27 @@ static int ican3_recv_skb(struct ican3_dev *mod)
 	/* convert the ICAN3 frame into Linux CAN format */
 	ican3_to_can_frame(mod, &desc, cf);
 
-	/* receive the skb, update statistics */
-	netif_receive_skb(skb);
+	/*
+	 * If this is an ECHO frame received from the hardware loopback
+	 * feature, use the skb saved in the ECHO stack instead. This allows
+	 * the Linux CAN core to support CAN_RAW_RECV_OWN_MSGS correctly.
+	 *
+	 * Since this is a confirmation of a successfully transmitted packet
+	 * sent from this host, update the transmit statistics.
+	 *
+	 * Also, the netdevice queue needs to be allowed to send packets again.
+	 */
+	if (ican3_echo_skb_matches(mod, skb)) {
+		stats->tx_packets++;
+		stats->tx_bytes += ican3_get_echo_skb(mod);
+		kfree_skb(skb);
+		goto err_noalloc;
+	}
+
+	/* update statistics, receive the skb */
 	stats->rx_packets++;
 	stats->rx_bytes += cf->can_dlc;
+	netif_receive_skb(skb);
 
 err_noalloc:
 	/* toggle the valid bit and return the descriptor to the ring */
@@ -1176,13 +1317,13 @@ err_noalloc:
 static int ican3_napi(struct napi_struct *napi, int budget)
 {
 	struct ican3_dev *mod = container_of(napi, struct ican3_dev, napi);
-	struct ican3_msg msg;
 	unsigned long flags;
 	int received = 0;
 	int ret;
 
 	/* process all communication messages */
 	while (true) {
+		struct ican3_msg msg;
 		ret = ican3_recv_msg(mod, &msg);
 		if (ret)
 			break;
@@ -1325,7 +1466,7 @@ static int __devinit ican3_startup_module(struct ican3_dev *mod)
 	}
 
 	/* default to "bus errors enabled" */
-	ret = ican3_set_buserror(mod, ICAN3_BUSERR_QUOTA_MAX);
+	ret = ican3_set_buserror(mod, 1);
 	if (ret) {
 		dev_err(mod->dev, "unable to set bus-error\n");
 		return ret;
@@ -1354,7 +1495,6 @@ static int __devinit ican3_startup_module(struct ican3_dev *mod)
 static int ican3_open(struct net_device *ndev)
 {
 	struct ican3_dev *mod = netdev_priv(ndev);
-	u8 quota;
 	int ret;
 
 	/* open the CAN layer */
@@ -1364,19 +1504,6 @@ static int ican3_open(struct net_device *ndev)
 		return ret;
 	}
 
-	/* set the bus error generation state appropriately */
-	if (mod->can.ctrlmode & CAN_CTRLMODE_BERR_REPORTING)
-		quota = ICAN3_BUSERR_QUOTA_MAX;
-	else
-		quota = 0;
-
-	ret = ican3_set_buserror(mod, quota);
-	if (ret) {
-		dev_err(mod->dev, "unable to set bus-error\n");
-		close_candev(ndev);
-		return ret;
-	}
-
 	/* bring the bus online */
 	ret = ican3_set_bus_state(mod, true);
 	if (ret) {
@@ -1408,6 +1535,9 @@ static int ican3_stop(struct net_device *ndev)
 		return ret;
 	}
 
+	/* drop all outstanding echo skbs */
+	skb_queue_purge(&mod->echoq);
+
 	/* close the CAN layer */
 	close_candev(ndev);
 	return 0;
@@ -1416,18 +1546,19 @@ static int ican3_stop(struct net_device *ndev)
 static int ican3_xmit(struct sk_buff *skb, struct net_device *ndev)
 {
 	struct ican3_dev *mod = netdev_priv(ndev);
-	struct net_device_stats *stats = &ndev->stats;
 	struct can_frame *cf = (struct can_frame *)skb->data;
 	struct ican3_fast_desc desc;
 	void __iomem *desc_addr;
 	unsigned long flags;
 
+	if (can_dropped_invalid_skb(ndev, skb))
+		return NETDEV_TX_OK;
+
 	spin_lock_irqsave(&mod->lock, flags);
 
 	/* check that we can actually transmit */
 	if (!ican3_txok(mod)) {
-		dev_err(mod->dev, "no free descriptors, stopping queue\n");
-		netif_stop_queue(ndev);
+		dev_err(mod->dev, "BUG: no free descriptors\n");
 		spin_unlock_irqrestore(&mod->lock, flags);
 		return NETDEV_TX_BUSY;
 	}
@@ -1442,6 +1573,14 @@ static int ican3_xmit(struct sk_buff *skb, struct net_device *ndev)
 	can_frame_to_ican3(mod, cf, &desc);
 
 	/*
+	 * This hardware doesn't have TX-done notifications, so we'll try and
+	 * emulate it the best we can using ECHO skbs. Add the skb to the ECHO
+	 * stack. Upon packet reception, check if the ECHO skb and received
+	 * skb match, and use that to wake the queue.
+	 */
+	ican3_put_echo_skb(mod, skb);
+
+	/*
 	 * the programming manual says that you must set the IVALID bit, then
 	 * interrupt, then set the valid bit. Quite weird, but it seems to be
 	 * required for this to work
@@ -1459,19 +1598,7 @@ static int ican3_xmit(struct sk_buff *skb, struct net_device *ndev)
 	mod->fasttx_num = (desc.control & DESC_WRAP) ? 0
 						     : (mod->fasttx_num + 1);
 
-	/* update statistics */
-	stats->tx_packets++;
-	stats->tx_bytes += cf->can_dlc;
-	kfree_skb(skb);
-
-	/*
-	 * This hardware doesn't have TX-done notifications, so we'll try and
-	 * emulate it the best we can using ECHO skbs. Get the next TX
-	 * descriptor, and see if we have room to send. If not, stop the queue.
-	 * It will be woken when the ECHO skb for the current packet is recv'd.
-	 */
-
-	/* copy the control bits of the descriptor */
+	/* if there is no free descriptor space, stop the transmit queue */
 	if (!ican3_txok(mod))
 		netif_stop_queue(ndev);
 
@@ -1490,7 +1617,7 @@ static const struct net_device_ops ican3_netdev_ops = {
  */
 
 /* This structure was stolen from drivers/net/can/sja1000/sja1000.c */
-static struct can_bittiming_const ican3_bittiming_const = {
+static const struct can_bittiming_const ican3_bittiming_const = {
 	.name = DRV_NAME,
 	.tseg1_min = 1,
 	.tseg1_max = 16,
@@ -1667,6 +1794,7 @@ static int __devinit ican3_probe(struct platform_device *pdev)
 	mod->dev = &pdev->dev;
 	mod->num = pdata->modno;
 	netif_napi_add(ndev, &mod->napi, ican3_napi, ICAN3_RX_BUFFERS);
+	skb_queue_head_init(&mod->echoq);
 	spin_lock_init(&mod->lock);
 	init_completion(&mod->termination_comp);
 	init_completion(&mod->buserror_comp);
@@ -1687,7 +1815,8 @@ static int __devinit ican3_probe(struct platform_device *pdev)
 	mod->can.do_set_mode = ican3_set_mode;
 	mod->can.do_get_berr_counter = ican3_get_berr_counter;
 	mod->can.ctrlmode_supported = CAN_CTRLMODE_3_SAMPLES
-				    | CAN_CTRLMODE_BERR_REPORTING;
+				    | CAN_CTRLMODE_BERR_REPORTING
+				    | CAN_CTRLMODE_ONE_SHOT;
 
 	/* find our IRQ number */
 	mod->irq = platform_get_irq(pdev, 0);