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

Pull networking fixes from David Miller:

 1) The value choosen for the new SO_MAX_PACING_RATE socket option on
    parisc was very poorly choosen, let's fix it while we still can.
    From Eric Dumazet.

 2) Our generic reciprocal divide was found to handle some edge cases
    incorrectly, part of this is encoded into the BPF as deep as the JIT
    engines themselves.  Just use a real divide throughout for now.
    From Eric Dumazet.

 3) Because the initial lookup is lockless, the TCP metrics engine can
    end up creating two entries for the same lookup key.  Fix this by
    doing a second lookup under the lock before we actually create the
    new entry.  From Christoph Paasch.

 4) Fix scatter-gather list init in usbnet driver, from Bjørn Mork.

 5) Fix unintended 32-bit truncation in cxgb4 driver's bit shifting.
    From Dan Carpenter.

 6) Netlink socket dumping uses the wrong socket state for timewait
    sockets.  Fix from Neal Cardwell.

 7) Fix netlink memory leak in ieee802154_add_iface(), from Christian
    Engelmayer.

 8) Multicast forwarding in ipv4 can overflow the per-rule reference
    counts, causing all multicast traffic to cease.  Fix from Hannes
    Frederic Sowa.

 9) via-rhine needs to stop all TX queues when it resets the device,
    from Richard Weinberger.

10) Fix RDS per-cpu accesses broken by the this_cpu_* conversions.  From
    Gerald Schaefer.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  s390/bpf,jit: fix 32 bit divisions, use unsigned divide instructions
  parisc: fix SO_MAX_PACING_RATE typo
  ipv6: simplify detection of first operational link-local address on interface
  tcp: metrics: Avoid duplicate entries with the same destination-IP
  net: rds: fix per-cpu helper usage
  e1000e: Fix compilation warning when !CONFIG_PM_SLEEP
  bpf: do not use reciprocal divide
  be2net: add dma_mapping_error() check for dma_map_page()
  bnx2x: Don't release PCI bars on shutdown
  net,via-rhine: Fix tx_timeout handling
  batman-adv: fix batman-adv header overhead calculation
  qlge: Fix vlan netdev features.
  net: avoid reference counter overflows on fib_rules in multicast forwarding
  dm9601: add USB IDs for new dm96xx variants
  MAINTAINERS: add virtio-dev ML for virtio
  ieee802154: Fix memory leak in ieee802154_add_iface()
  net: usbnet: fix SG initialisation
  inet_diag: fix inet_diag_dump_icsk() to use correct state for timewait sockets
  cxgb4: silence shift wrapping static checker warning

1 files changed, 32 insertions, 19 deletions

diff --git a/net/ipv4/tcp_metrics.c b/net/ipv4/tcp_metrics.c
index 06493736fbc8..098b3a29f6f3 100644
--- a/net/ipv4/tcp_metrics.c
+++ b/net/ipv4/tcp_metrics.c
@@ -22,6 +22,9 @@
 
 int sysctl_tcp_nometrics_save __read_mostly;
 
+static struct tcp_metrics_block *__tcp_get_metrics(const struct inetpeer_addr *addr,
+						   struct net *net, unsigned int hash);
+
 struct tcp_fastopen_metrics {
 	u16	mss;
 	u16	syn_loss:10;		/* Recurring Fast Open SYN losses */
@@ -130,16 +133,41 @@ static void tcpm_suck_dst(struct tcp_metrics_block *tm, struct dst_entry *dst,
 	}
 }
 
+#define TCP_METRICS_TIMEOUT		(60 * 60 * HZ)
+
+static void tcpm_check_stamp(struct tcp_metrics_block *tm, struct dst_entry *dst)
+{
+	if (tm && unlikely(time_after(jiffies, tm->tcpm_stamp + TCP_METRICS_TIMEOUT)))
+		tcpm_suck_dst(tm, dst, false);
+}
+
+#define TCP_METRICS_RECLAIM_DEPTH	5
+#define TCP_METRICS_RECLAIM_PTR		(struct tcp_metrics_block *) 0x1UL
+
 static struct tcp_metrics_block *tcpm_new(struct dst_entry *dst,
 					  struct inetpeer_addr *addr,
-					  unsigned int hash,
-					  bool reclaim)
+					  unsigned int hash)
 {
 	struct tcp_metrics_block *tm;
 	struct net *net;
+	bool reclaim = false;
 
 	spin_lock_bh(&tcp_metrics_lock);
 	net = dev_net(dst->dev);
+
+	/* While waiting for the spin-lock the cache might have been populated
+	 * with this entry and so we have to check again.
+	 */
+	tm = __tcp_get_metrics(addr, net, hash);
+	if (tm == TCP_METRICS_RECLAIM_PTR) {
+		reclaim = true;
+		tm = NULL;
+	}
+	if (tm) {
+		tcpm_check_stamp(tm, dst);
+		goto out_unlock;
+	}
+
 	if (unlikely(reclaim)) {
 		struct tcp_metrics_block *oldest;
 
@@ -169,17 +197,6 @@ out_unlock:
 	return tm;
 }
 
-#define TCP_METRICS_TIMEOUT		(60 * 60 * HZ)
-
-static void tcpm_check_stamp(struct tcp_metrics_block *tm, struct dst_entry *dst)
-{
-	if (tm && unlikely(time_after(jiffies, tm->tcpm_stamp + TCP_METRICS_TIMEOUT)))
-		tcpm_suck_dst(tm, dst, false);
-}
-
-#define TCP_METRICS_RECLAIM_DEPTH	5
-#define TCP_METRICS_RECLAIM_PTR		(struct tcp_metrics_block *) 0x1UL
-
 static struct tcp_metrics_block *tcp_get_encode(struct tcp_metrics_block *tm, int depth)
 {
 	if (tm)
@@ -282,7 +299,6 @@ static struct tcp_metrics_block *tcp_get_metrics(struct sock *sk,
 	struct inetpeer_addr addr;
 	unsigned int hash;
 	struct net *net;
-	bool reclaim;
 
 	addr.family = sk->sk_family;
 	switch (addr.family) {
@@ -304,13 +320,10 @@ static struct tcp_metrics_block *tcp_get_metrics(struct sock *sk,
 	hash = hash_32(hash, net->ipv4.tcp_metrics_hash_log);
 
 	tm = __tcp_get_metrics(&addr, net, hash);
-	reclaim = false;
-	if (tm == TCP_METRICS_RECLAIM_PTR) {
-		reclaim = true;
+	if (tm == TCP_METRICS_RECLAIM_PTR)
 		tm = NULL;
-	}
 	if (!tm && create)
-		tm = tcpm_new(dst, &addr, hash, reclaim);
+		tm = tcpm_new(dst, &addr, hash);
 	else
 		tcpm_check_stamp(tm, dst);