[PATCH net-next 0/2] mpls: Packet stats

This patchset records per-interface packet stats in the MPLS
forwarding path and exports them using a nest of attributes root at a
new IFLA_STATS_AF_SPEC attribute as part of RTM_GETSTATS messages:

[IFLA_STATS_AF_SPEC]
 -> [AF_MPLS]
  -> [MPLS_STATS_LINK]
   -> struct mpls_link_stats

The first patch adds the rtnl infrastructure for this, including a new
callbacks to per-AF ops of fill_stats_af and get_stats_af_size. The
second patch records MPLS stats and makes use of the infrastructure to
export them. The rtnl infrastructure could also be used to export IPv6
stats in the future.

Robert Shearman (2):
  net: AF-specific RTM_GETSTATS attributes
  mpls: Packet stats

 include/net/rtnetlink.h      |   4 +
 include/uapi/linux/if_link.h |   1 +
 include/uapi/linux/mpls.h    |  30 ++++++++
 net/core/rtnetlink.c         |  50 ++++++++++++
 net/mpls/af_mpls.c           | 179 +++++++++++++++++++++++++++++++++++++------
 net/mpls/internal.h          |  58 +++++++++++++-
 net/mpls/mpls_iptunnel.c     |  11 ++-
 7 files changed, 305 insertions(+), 28 deletions(-)

-- 
2.1.4

Add the functionality for including address-family-specific per-link
stats in RTM_GETSTATS messages. This is done through adding a new
IFLA_STATS_AF_SPEC attribute under which address family attributes are
nested and then the AF-specific attributes can be further nested. This
follows the model of IFLA_AF_SPEC on RTM_*LINK messages and it has the
advantage of presenting an easily extended hierarchy. The rtnl_af_ops
structure is extended to provide AFs with the opportunity to fill and
provide the size of their stats attributes.

One alternative would have been to provide AFs with the ability to add
attributes directly into the RTM_GETSTATS message without a nested
hierarchy. I discounted this approach as it increases the rate at
which the 32 attribute number space is used up and it makes
implementation a little more tricky for stats dump resuming (at the
moment the order in which attributes are added to the message has to
match the numeric order of the attributes).

Another alternative would have been to register per-AF RTM_GETSTATS
handlers. I discounted this approach as I perceived a common use-case
to be getting all the stats for an interface and this approach would
necessitate multiple requests/dumps to retrieve them all.

Signed-off-by: Robert Shearman <redacted>
---
 include/net/rtnetlink.h      |  4 ++++
 include/uapi/linux/if_link.h |  1 +
 net/core/rtnetlink.c         | 50 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 55 insertions(+)

diff --git a/include/net/rtnetlink.h b/include/net/rtnetlink.h
index 4113916cc1bb..106de5f7bf06 100644
--- a/include/net/rtnetlink.h
+++ b/include/net/rtnetlink.h

@@ -139,6 +139,10 @@ struct rtnl_af_ops {
 						    const struct nlattr *attr);
 	int			(*set_link_af)(struct net_device *dev,
 					       const struct nlattr *attr);
+
+	int			(*fill_stats_af)(struct sk_buff *skb,
+						 const struct net_device *dev);
+	size_t			(*get_stats_af_size)(const struct net_device *dev);
 };
 
 void __rtnl_af_unregister(struct rtnl_af_ops *ops);

diff --git a/include/uapi/linux/if_link.h b/include/uapi/linux/if_link.h
index 6b13e591abc9..184b16ed2b84 100644
--- a/include/uapi/linux/if_link.h
+++ b/include/uapi/linux/if_link.h

@@ -847,6 +847,7 @@ enum {
 	IFLA_STATS_LINK_XSTATS,
 	IFLA_STATS_LINK_XSTATS_SLAVE,
 	IFLA_STATS_LINK_OFFLOAD_XSTATS,
+	IFLA_STATS_AF_SPEC,
 	__IFLA_STATS_MAX,
 };
 

diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index 18b5aae99bec..4edc1bd7a735 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c

@@ -3829,6 +3829,39 @@ static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
 		*idxattr = 0;
 	}
 
+	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
+		struct rtnl_af_ops *af_ops;
+
+		*idxattr = IFLA_STATS_AF_SPEC;
+		attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
+		if (!attr)
+			goto nla_put_failure;
+
+		list_for_each_entry(af_ops, &rtnl_af_ops, list) {
+			if (af_ops->fill_stats_af) {
+				struct nlattr *af;
+				int err;
+
+				af = nla_nest_start(skb, af_ops->family);
+				if (!af)
+					goto nla_put_failure;
+
+				err = af_ops->fill_stats_af(skb, dev);
+
+				if (err == -ENODATA)
+					nla_nest_cancel(skb, af);
+				else if (err < 0)
+					goto nla_put_failure;
+
+				nla_nest_end(skb, af);
+			}
+		}
+
+		nla_nest_end(skb, attr);
+
+		*idxattr = 0;
+	}
+
 	nlmsg_end(skb, nlh);
 
 	return 0;

@@ -3885,6 +3918,23 @@ static size_t if_nlmsg_stats_size(const struct net_device *dev,
 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
 		size += rtnl_get_offload_stats_size(dev);
 
+	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
+		struct rtnl_af_ops *af_ops;
+
+		/* for IFLA_STATS_AF_SPEC */
+		size += nla_total_size(0);
+
+		list_for_each_entry(af_ops, &rtnl_af_ops, list) {
+			if (af_ops->get_stats_af_size) {
+				size += nla_total_size(
+					af_ops->get_stats_af_size(dev));
+
+				/* for AF_* */
+				size += nla_total_size(0);
+			}
+		}
+	}
+
 	return size;
 }
 

-- 
2.1.4

Having MPLS packet stats is useful for observing network operation and
for diagnosing network problems. In the absence of anything better,
RFC2863 and RFC3813 are used for guidance for which stats to expose
and the semantics of them. In particular rx_noroutes maps to in
unknown protos in RFC2863. The stats are exposed to userspace via
AF_MPLS attributes embedded in the IFLA_STATS_AF_SPEC attribute of
RTM_GETSTATS messages.

All the introduced fields are 64-bit, even error ones, to ensure no
overflow with long uptimes. Per-CPU counters are used to avoid
cache-line contention on the commonly used fields. The other fields
have also been made per-CPU for code to avoid performance problems in
error conditions on the assumption that on some platforms the cost of
atomic operations could be more expensive than sending the packet
(which is what would be done in the success case). If that's not the
case, we could instead not use per-CPU counters for these fields.

Only unicast and non-fragment are exposed at the moment, but other
counters can be exposed in the future either by adding to the end of
struct mpls_link_stats or by additional netlink attributes in the
AF_MPLS IFLA_STATS_AF_SPEC nested attribute.

Signed-off-by: Robert Shearman <redacted>
---
 include/uapi/linux/mpls.h |  30 ++++++++
 net/mpls/af_mpls.c        | 179 ++++++++++++++++++++++++++++++++++++++++------
 net/mpls/internal.h       |  58 ++++++++++++++-
 net/mpls/mpls_iptunnel.c  |  11 ++-
 4 files changed, 250 insertions(+), 28 deletions(-)

diff --git a/include/uapi/linux/mpls.h b/include/uapi/linux/mpls.h
index 24a6cb1aec86..77a19dfe3990 100644
--- a/include/uapi/linux/mpls.h
+++ b/include/uapi/linux/mpls.h

@@ -43,4 +43,34 @@ struct mpls_label {
 
 #define MPLS_LABEL_FIRST_UNRESERVED	16 /* RFC3032 */
 
+/* These are embedded into IFLA_STATS_AF_SPEC:
+ * [IFLA_STATS_AF_SPEC]
+ * -> [AF_MPLS]
+ *    -> [MPLS_STATS_xxx]
+ *
+ * Attributes:
+ * [MPLS_STATS_LINK] = {
+ *     struct mpls_link_stats
+ * }
+ */
+enum {
+	MPLS_STATS_UNSPEC, /* also used as 64bit pad attribute */
+	MPLS_STATS_LINK,
+	__MPLS_STATS_MAX,
+};
+
+#define MPLS_STATS_MAX (__MPLS_STATS_MAX - 1)
+
+struct mpls_link_stats {
+	__u64	rx_packets;		/* total packets received	*/
+	__u64	tx_packets;		/* total packets transmitted	*/
+	__u64	rx_bytes;		/* total bytes received		*/
+	__u64	tx_bytes;		/* total bytes transmitted	*/
+	__u64	rx_errors;		/* bad packets received		*/
+	__u64	tx_errors;		/* packet transmit problems	*/
+	__u64	rx_dropped;		/* packet dropped on receive	*/
+	__u64	tx_dropped;		/* packet dropped on transmit	*/
+	__u64	rx_noroute;		/* no route for packet dest	*/
+};
+
 #endif /* _UAPI_MPLS_H */

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index 15fe97644ffe..fb20941cdda2 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c

@@ -8,6 +8,7 @@
 #include <linux/ipv6.h>
 #include <linux/mpls.h>
 #include <linux/vmalloc.h>
+#include <linux/percpu.h>
 #include <net/ip.h>
 #include <net/dst.h>
 #include <net/sock.h>

@@ -17,8 +18,8 @@
 #include <net/netns/generic.h>
 #if IS_ENABLED(CONFIG_IPV6)
 #include <net/ipv6.h>
-#include <net/addrconf.h>
 #endif
+#include <net/addrconf.h>
 #include <net/nexthop.h>
 #include "internal.h"
 

@@ -48,11 +49,6 @@ static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned index)
 	return rt;
 }
 
-static inline struct mpls_dev *mpls_dev_get(const struct net_device *dev)
-{
-	return rcu_dereference_rtnl(dev->mpls_ptr);
-}
-
 bool mpls_output_possible(const struct net_device *dev)
 {
 	return dev && (dev->flags & IFF_UP) && netif_carrier_ok(dev);

@@ -98,6 +94,29 @@ bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
 }
 EXPORT_SYMBOL_GPL(mpls_pkt_too_big);
 
+void mpls_stats_inc_outucastpkts(struct net_device *dev,
+				 const struct sk_buff *skb)
+{
+	struct mpls_dev *mdev;
+	struct inet6_dev *in6dev;
+
+	if (skb->protocol == htons(ETH_P_MPLS_UC)) {
+		mdev = mpls_dev_get(dev);
+		if (mdev)
+			MPLS_INC_STATS_LEN(mdev, skb->len,
+					   tx_packets,
+					   tx_bytes);
+	} else if (skb->protocol == htons(ETH_P_IP)) {
+		IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len);
+	} else if (skb->protocol == htons(ETH_P_IPV6)) {
+		in6dev = __in6_dev_get(dev);
+		if (in6dev)
+			IP6_UPD_PO_STATS(dev_net(dev), in6dev,
+					 IPSTATS_MIB_OUT, skb->len);
+	}
+}
+EXPORT_SYMBOL_GPL(mpls_stats_inc_outucastpkts);
+
 static u32 mpls_multipath_hash(struct mpls_route *rt,
 			       struct sk_buff *skb, bool bos)
 {

@@ -253,6 +272,7 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 	struct mpls_nh *nh;
 	struct mpls_entry_decoded dec;
 	struct net_device *out_dev;
+	struct mpls_dev *out_mdev;
 	struct mpls_dev *mdev;
 	unsigned int hh_len;
 	unsigned int new_header_size;

@@ -262,17 +282,25 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 	/* Careful this entire function runs inside of an rcu critical section */
 
 	mdev = mpls_dev_get(dev);
-	if (!mdev || !mdev->input_enabled)
+	if (!mdev)
 		goto drop;
 
-	if (skb->pkt_type != PACKET_HOST)
+	MPLS_INC_STATS_LEN(mdev, skb->len, rx_packets,
+			   rx_bytes);
+
+	if (!mdev->input_enabled) {
+		MPLS_INC_STATS(mdev, rx_dropped);
 		goto drop;
+	}
+
+	if (skb->pkt_type != PACKET_HOST)
+		goto err;
 
 	if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL)
-		goto drop;
+		goto err;
 
 	if (!pskb_may_pull(skb, sizeof(*hdr)))
-		goto drop;
+		goto err;
 
 	/* Read and decode the label */
 	hdr = mpls_hdr(skb);

@@ -285,33 +313,35 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 	skb_orphan(skb);
 
 	rt = mpls_route_input_rcu(net, dec.label);
-	if (!rt)
+	if (!rt) {
+		MPLS_INC_STATS(mdev, rx_noroute);
 		goto drop;
+	}
 
 	nh = mpls_select_multipath(rt, skb, dec.bos);
 	if (!nh)
-		goto drop;
-
-	/* Find the output device */
-	out_dev = rcu_dereference(nh->nh_dev);
-	if (!mpls_output_possible(out_dev))
-		goto drop;
+		goto err;
 
 	if (skb_warn_if_lro(skb))
-		goto drop;
+		goto err;
 
 	skb_forward_csum(skb);
 
 	/* Verify ttl is valid */
 	if (dec.ttl <= 1)
-		goto drop;
+		goto err;
 	dec.ttl -= 1;
 
+	/* Find the output device */
+	out_dev = rcu_dereference(nh->nh_dev);
+	if (!mpls_output_possible(out_dev))
+		goto tx_err;
+
 	/* Verify the destination can hold the packet */
 	new_header_size = mpls_nh_header_size(nh);
 	mtu = mpls_dev_mtu(out_dev);
 	if (mpls_pkt_too_big(skb, mtu - new_header_size))
-		goto drop;
+		goto tx_err;
 
 	hh_len = LL_RESERVED_SPACE(out_dev);
 	if (!out_dev->header_ops)

@@ -319,7 +349,7 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 
 	/* Ensure there is enough space for the headers in the skb */
 	if (skb_cow(skb, hh_len + new_header_size))
-		goto drop;
+		goto tx_err;
 
 	skb->dev = out_dev;
 	skb->protocol = htons(ETH_P_MPLS_UC);

@@ -327,7 +357,7 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 	if (unlikely(!new_header_size && dec.bos)) {
 		/* Penultimate hop popping */
 		if (!mpls_egress(rt, skb, dec))
-			goto drop;
+			goto err;
 	} else {
 		bool bos;
 		int i;

@@ -343,6 +373,8 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 		}
 	}
 
+	mpls_stats_inc_outucastpkts(out_dev, skb);
+
 	/* If via wasn't specified then send out using device address */
 	if (nh->nh_via_table == MPLS_NEIGH_TABLE_UNSPEC)
 		err = neigh_xmit(NEIGH_LINK_TABLE, out_dev,

@@ -355,6 +387,13 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 				    __func__, err);
 	return 0;
 
+tx_err:
+	out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
+	if (out_mdev)
+		MPLS_INC_STATS(out_mdev, tx_errors);
+	goto drop;
+err:
+	MPLS_INC_STATS(mdev, rx_errors);
 drop:
 	kfree_skb(skb);
 	return NET_RX_DROP;

@@ -853,6 +892,70 @@ static int mpls_route_del(struct mpls_route_config *cfg)
 	return err;
 }
 
+static void mpls_get_stats(struct mpls_dev *mdev,
+			   struct mpls_link_stats *stats)
+{
+	struct mpls_pcpu_stats *p;
+	int i;
+
+	memset(stats, 0, sizeof(*stats));
+
+	for_each_possible_cpu(i) {
+		struct mpls_link_stats local;
+		unsigned int start;
+
+		p = per_cpu_ptr(mdev->stats, i);
+		do {
+			start = u64_stats_fetch_begin(&p->syncp);
+			local = p->stats;
+		} while (u64_stats_fetch_retry(&p->syncp, start));
+
+		stats->rx_packets	+= local.rx_packets;
+		stats->rx_bytes		+= local.rx_bytes;
+		stats->tx_packets	+= local.tx_packets;
+		stats->tx_bytes		+= local.tx_bytes;
+		stats->rx_errors	+= local.rx_errors;
+		stats->tx_errors	+= local.tx_errors;
+		stats->rx_dropped	+= local.rx_dropped;
+		stats->tx_dropped	+= local.tx_dropped;
+		stats->rx_noroute	+= local.rx_noroute;
+	}
+}
+
+static int mpls_fill_stats_af(struct sk_buff *skb,
+			      const struct net_device *dev)
+{
+	struct mpls_link_stats *stats;
+	struct mpls_dev *mdev;
+	struct nlattr *nla;
+
+	mdev = mpls_dev_get(dev);
+	if (!mdev)
+		return -ENODATA;
+
+	nla = nla_reserve_64bit(skb, MPLS_STATS_LINK,
+				sizeof(struct mpls_link_stats),
+				MPLS_STATS_UNSPEC);
+	if (!nla)
+		return -EMSGSIZE;
+
+	stats = nla_data(nla);
+	mpls_get_stats(mdev, stats);
+
+	return 0;
+}
+
+static size_t mpls_get_stats_af_size(const struct net_device *dev)
+{
+	struct mpls_dev *mdev;
+
+	mdev = mpls_dev_get(dev);
+	if (!mdev)
+		return 0;
+
+	return nla_total_size_64bit(sizeof(struct mpls_link_stats));
+}
+
 #define MPLS_PERDEV_SYSCTL_OFFSET(field)	\
 	(&((struct mpls_dev *)0)->field)
 

@@ -911,6 +1014,7 @@ static struct mpls_dev *mpls_add_dev(struct net_device *dev)
 {
 	struct mpls_dev *mdev;
 	int err = -ENOMEM;
+	int i;
 
 	ASSERT_RTNL();
 

@@ -918,6 +1022,17 @@ static struct mpls_dev *mpls_add_dev(struct net_device *dev)
 	if (!mdev)
 		return ERR_PTR(err);
 
+	mdev->stats = alloc_percpu(struct mpls_pcpu_stats);
+	if (!mdev->stats)
+		goto free;
+
+	for_each_possible_cpu(i) {
+		struct mpls_pcpu_stats *mpls_stats;
+
+		mpls_stats = per_cpu_ptr(mdev->stats, i);
+		u64_stats_init(&mpls_stats->syncp);
+	}
+
 	err = mpls_dev_sysctl_register(dev, mdev);
 	if (err)
 		goto free;

@@ -927,10 +1042,19 @@ static struct mpls_dev *mpls_add_dev(struct net_device *dev)
 	return mdev;
 
 free:
+	free_percpu(mdev->stats);
 	kfree(mdev);
 	return ERR_PTR(err);
 }
 
+static void mpls_dev_destroy_rcu(struct rcu_head *head)
+{
+	struct mpls_dev *mdev = container_of(head, struct mpls_dev, rcu);
+
+	free_percpu(mdev->stats);
+	kfree(mdev);
+}
+
 static void mpls_ifdown(struct net_device *dev, int event)
 {
 	struct mpls_route __rcu **platform_label;

@@ -1045,7 +1169,7 @@ static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
 		if (mdev) {
 			mpls_dev_sysctl_unregister(mdev);
 			RCU_INIT_POINTER(dev->mpls_ptr, NULL);
-			kfree_rcu(mdev, rcu);
+			call_rcu(&mdev->rcu, mpls_dev_destroy_rcu);
 		}
 		break;
 	case NETDEV_CHANGENAME:

@@ -1706,6 +1830,12 @@ static struct pernet_operations mpls_net_ops = {
 	.exit = mpls_net_exit,
 };
 
+static struct rtnl_af_ops mpls_af_ops __read_mostly = {
+	.family		   = AF_MPLS,
+	.fill_stats_af	   = mpls_fill_stats_af,
+	.get_stats_af_size = mpls_get_stats_af_size,
+};
+
 static int __init mpls_init(void)
 {
 	int err;

@@ -1722,6 +1852,8 @@ static int __init mpls_init(void)
 
 	dev_add_pack(&mpls_packet_type);
 
+	rtnl_af_register(&mpls_af_ops);
+
 	rtnl_register(PF_MPLS, RTM_NEWROUTE, mpls_rtm_newroute, NULL, NULL);
 	rtnl_register(PF_MPLS, RTM_DELROUTE, mpls_rtm_delroute, NULL, NULL);
 	rtnl_register(PF_MPLS, RTM_GETROUTE, NULL, mpls_dump_routes, NULL);

@@ -1738,6 +1870,7 @@ module_init(mpls_init);
 static void __exit mpls_exit(void)
 {
 	rtnl_unregister_all(PF_MPLS);
+	rtnl_af_unregister(&mpls_af_ops);
 	dev_remove_pack(&mpls_packet_type);
 	unregister_netdevice_notifier(&mpls_dev_notifier);
 	unregister_pernet_subsys(&mpls_net_ops);

diff --git a/net/mpls/internal.h b/net/mpls/internal.h
index bdfef6c3271a..d97243034605 100644
--- a/net/mpls/internal.h
+++ b/net/mpls/internal.h

@@ -9,13 +9,58 @@ struct mpls_entry_decoded {
 	u8 bos;
 };
 
+struct mpls_pcpu_stats {
+	struct mpls_link_stats	stats;
+	struct u64_stats_sync	syncp;
+};
+
 struct mpls_dev {
-	int			input_enabled;
+	int				input_enabled;
 
-	struct ctl_table_header *sysctl;
-	struct rcu_head		rcu;
+	struct mpls_pcpu_stats __percpu	*stats;
+
+	struct ctl_table_header		*sysctl;
+	struct rcu_head			rcu;
 };
 
+#if BITS_PER_LONG == 32
+
+#define MPLS_INC_STATS_LEN(mdev, len, pkts_field, bytes_field)		\
+	do {								\
+		__typeof__(*(mdev)->stats) *ptr =			\
+			raw_cpu_ptr((mdev)->stats);			\
+		local_bh_disable();					\
+		u64_stats_update_begin(&ptr->syncp);			\
+		ptr->stats.pkts_field++;				\
+		ptr->stats.bytes_field += (len);			\
+		u64_stats_update_end(&ptr->syncp);			\
+		local_bh_enable();					\
+	} while (0)
+
+#define MPLS_INC_STATS(mdev, field)					\
+	do {								\
+		__typeof__(*(mdev)->stats) *ptr =			\
+			raw_cpu_ptr((mdev)->stats);			\
+		local_bh_disable();					\
+		u64_stats_update_begin(&ptr->syncp);			\
+		ptr->stats.field++;					\
+		u64_stats_update_end(&ptr->syncp);			\
+		local_bh_enable();					\
+	} while (0)
+
+#else
+
+#define MPLS_INC_STATS_LEN(mdev, len, pkts_field, bytes_field)		\
+	do {								\
+		this_cpu_inc((mdev)->stats->stats.pkts_field);		\
+		this_cpu_add((mdev)->stats->stats.bytes_field, (len));	\
+	} while (0)
+
+#define MPLS_INC_STATS(mdev, field)			\
+	this_cpu_inc((mdev)->stats->stats.field)
+
+#endif
+
 struct sk_buff;
 
 #define LABEL_NOT_SPECIFIED (1 << 20)

@@ -114,6 +159,11 @@ static inline struct mpls_entry_decoded mpls_entry_decode(struct mpls_shim_hdr *
 	return result;
 }
 
+static inline struct mpls_dev *mpls_dev_get(const struct net_device *dev)
+{
+	return rcu_dereference_rtnl(dev->mpls_ptr);
+}
+
 int nla_put_labels(struct sk_buff *skb, int attrtype,  u8 labels,
 		   const u32 label[]);
 int nla_get_labels(const struct nlattr *nla, u32 max_labels, u8 *labels,

@@ -123,5 +173,7 @@ int nla_get_via(const struct nlattr *nla, u8 *via_alen, u8 *via_table,
 bool mpls_output_possible(const struct net_device *dev);
 unsigned int mpls_dev_mtu(const struct net_device *dev);
 bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu);
+void mpls_stats_inc_outucastpkts(struct net_device *dev,
+				 const struct sk_buff *skb);
 
 #endif /* MPLS_INTERNAL_H */

diff --git a/net/mpls/mpls_iptunnel.c b/net/mpls/mpls_iptunnel.c
index 2f7ccd934416..02531284bc49 100644
--- a/net/mpls/mpls_iptunnel.c
+++ b/net/mpls/mpls_iptunnel.c

@@ -48,11 +48,15 @@ static int mpls_xmit(struct sk_buff *skb)
 	struct dst_entry *dst = skb_dst(skb);
 	struct rtable *rt = NULL;
 	struct rt6_info *rt6 = NULL;
+	struct mpls_dev *out_mdev;
 	int err = 0;
 	bool bos;
 	int i;
 	unsigned int ttl;
 
+	/* Find the output device */
+	out_dev = dst->dev;
+
 	/* Obtain the ttl */
 	if (dst->ops->family == AF_INET) {
 		ttl = ip_hdr(skb)->ttl;

@@ -66,8 +70,6 @@ static int mpls_xmit(struct sk_buff *skb)
 
 	skb_orphan(skb);
 
-	/* Find the output device */
-	out_dev = dst->dev;
 	if (!mpls_output_possible(out_dev) ||
 	    !dst->lwtstate || skb_warn_if_lro(skb))
 		goto drop;

@@ -109,6 +111,8 @@ static int mpls_xmit(struct sk_buff *skb)
 		bos = false;
 	}
 
+	mpls_stats_inc_outucastpkts(out_dev, skb);
+
 	if (rt)
 		err = neigh_xmit(NEIGH_ARP_TABLE, out_dev, &rt->rt_gateway,
 				 skb);

@@ -122,6 +126,9 @@ static int mpls_xmit(struct sk_buff *skb)
 	return LWTUNNEL_XMIT_DONE;
 
 drop:
+	out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
+	if (out_mdev)
+		MPLS_INC_STATS(out_mdev, tx_errors);
 	kfree_skb(skb);
 	return -EINVAL;
 }

-- 
2.1.4

On 1/13/17, 10:14 AM, Robert Shearman wrote:

This patchset records per-interface packet stats in the MPLS
forwarding path and exports them using a nest of attributes root at a
new IFLA_STATS_AF_SPEC attribute as part of RTM_GETSTATS messages:

[IFLA_STATS_AF_SPEC]
 -> [AF_MPLS]
  -> [MPLS_STATS_LINK]
   -> struct mpls_link_stats

The first patch adds the rtnl infrastructure for this, including a new
callbacks to per-AF ops of fill_stats_af and get_stats_af_size. The
second patch records MPLS stats and makes use of the infrastructure to
export them. The rtnl infrastructure could also be used to export IPv6
stats in the future.

Robert Shearman (2):
  net: AF-specific RTM_GETSTATS attributes
  mpls: Packet stats

LGTM, thanks Robert

On 1/13/17, 10:14 AM, Robert Shearman wrote:

Add the functionality for including address-family-specific per-link
stats in RTM_GETSTATS messages. This is done through adding a new
IFLA_STATS_AF_SPEC attribute under which address family attributes are
nested and then the AF-specific attributes can be further nested. This
follows the model of IFLA_AF_SPEC on RTM_*LINK messages and it has the
advantage of presenting an easily extended hierarchy. The rtnl_af_ops
structure is extended to provide AFs with the opportunity to fill and
provide the size of their stats attributes.

One alternative would have been to provide AFs with the ability to add
attributes directly into the RTM_GETSTATS message without a nested
hierarchy. I discounted this approach as it increases the rate at
which the 32 attribute number space is used up and it makes
implementation a little more tricky for stats dump resuming (at the
moment the order in which attributes are added to the message has to
match the numeric order of the attributes).

Another alternative would have been to register per-AF RTM_GETSTATS
handlers. I discounted this approach as I perceived a common use-case
to be getting all the stats for an interface and this approach would
necessitate multiple requests/dumps to retrieve them all.

Signed-off-by: Robert Shearman <redacted>

Acked-by: Roopa Prabhu <redacted>

On 1/13/17, 10:14 AM, Robert Shearman wrote:

Having MPLS packet stats is useful for observing network operation and
for diagnosing network problems. In the absence of anything better,
RFC2863 and RFC3813 are used for guidance for which stats to expose
and the semantics of them. In particular rx_noroutes maps to in
unknown protos in RFC2863. The stats are exposed to userspace via
AF_MPLS attributes embedded in the IFLA_STATS_AF_SPEC attribute of
RTM_GETSTATS messages.

All the introduced fields are 64-bit, even error ones, to ensure no
overflow with long uptimes. Per-CPU counters are used to avoid
cache-line contention on the commonly used fields. The other fields
have also been made per-CPU for code to avoid performance problems in
error conditions on the assumption that on some platforms the cost of
atomic operations could be more expensive than sending the packet
(which is what would be done in the success case). If that's not the
case, we could instead not use per-CPU counters for these fields.

Only unicast and non-fragment are exposed at the moment, but other
counters can be exposed in the future either by adding to the end of
struct mpls_link_stats or by additional netlink attributes in the
AF_MPLS IFLA_STATS_AF_SPEC nested attribute.

Signed-off-by: Robert Shearman <redacted>
---

Acked-by: Roopa Prabhu <redacted>

Hi Robert,

[auto build test ERROR on net-next/master]

url:    https://github.com/0day-ci/linux/commits/Robert-Shearman/net-AF-specific-RTM_GETSTATS-attributes/20170114-095819
config: x86_64-randconfig-u0-01141334 (attached as .config)
compiler: gcc-6 (Debian 6.2.0-3) 6.2.0 20160901
reproduce:
        # save the attached .config to linux build tree
        make ARCH=x86_64 

All error/warnings (new ones prefixed by >>):

   In file included from include/net/netns/mib.h:4:0,
                    from include/net/net_namespace.h:14,
                    from include/linux/netdevice.h:43,
                    from include/uapi/linux/if_arp.h:26,
                    from include/linux/if_arp.h:27,
                    from net/mpls/af_mpls.c:7:
   net/mpls/af_mpls.c: In function 'mpls_stats_inc_outucastpkts':

quoted

include/net/ipv6.h:163:35: error: 'struct netns_mib' has no member named 'ipv6_statistics'; did you mean 'ip_statistics'?

     mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
                                      ^
   include/net/snmp.h:145:15: note: in definition of macro 'SNMP_UPD_PO_STATS'
      __typeof__((mib->mibs) + 0) ptr = mib->mibs; \
                  ^~~

quoted

include/net/ipv6.h:177:3: note: in expansion of macro '_DEVUPD'

      _DEVUPD(net, ipv6, , idev, field, val)
      ^~~~~~~

quoted

net/mpls/af_mpls.c:114:4: note: in expansion of macro 'IP6_UPD_PO_STATS'

       IP6_UPD_PO_STATS(dev_net(dev), in6dev,
       ^~~~~~~~~~~~~~~~

quoted

include/net/ipv6.h:163:35: error: 'struct netns_mib' has no member named 'ipv6_statistics'; did you mean 'ip_statistics'?

     mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
                                      ^
   include/net/snmp.h:145:37: note: in definition of macro 'SNMP_UPD_PO_STATS'
      __typeof__((mib->mibs) + 0) ptr = mib->mibs; \
                                        ^~~

quoted

include/net/ipv6.h:177:3: note: in expansion of macro '_DEVUPD'

      _DEVUPD(net, ipv6, , idev, field, val)
      ^~~~~~~

quoted

net/mpls/af_mpls.c:114:4: note: in expansion of macro 'IP6_UPD_PO_STATS'

       IP6_UPD_PO_STATS(dev_net(dev), in6dev,
       ^~~~~~~~~~~~~~~~
   In file included from include/asm-generic/percpu.h:6:0,
                    from arch/x86/include/asm/percpu.h:542,
                    from arch/x86/include/asm/preempt.h:5,
                    from include/linux/preempt.h:59,
                    from include/linux/spinlock.h:50,
                    from include/linux/mm_types.h:8,
                    from include/linux/kmemcheck.h:4,
                    from include/linux/skbuff.h:18,
                    from net/mpls/af_mpls.c:2:

quoted

include/net/snmp.h:146:19: error: subscripted value is neither array nor pointer nor vector

      this_cpu_inc(ptr[basefield##PKTS]);  \
                      ^
   include/linux/percpu-defs.h:206:47: note: in definition of macro '__verify_pcpu_ptr'
     const void __percpu *__vpp_verify = (typeof((ptr) + 0))NULL; \
                                                  ^~~
   include/linux/percpu-defs.h:496:33: note: in expansion of macro '__pcpu_size_call'
    #define this_cpu_add(pcp, val)  __pcpu_size_call(this_cpu_add_, pcp, val)
                                    ^~~~~~~~~~~~~~~~
   include/linux/percpu-defs.h:507:28: note: in expansion of macro 'this_cpu_add'
    #define this_cpu_inc(pcp)  this_cpu_add(pcp, 1)
                               ^~~~~~~~~~~~
   include/net/snmp.h:146:3: note: in expansion of macro 'this_cpu_inc'
      this_cpu_inc(ptr[basefield##PKTS]);  \
      ^~~~~~~~~~~~

quoted

include/net/ipv6.h:163:7: note: in expansion of macro 'SNMP_UPD_PO_STATS'

     mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
          ^~~~~~~~~~~~~~~~~

quoted

include/net/ipv6.h:177:3: note: in expansion of macro '_DEVUPD'

      _DEVUPD(net, ipv6, , idev, field, val)
      ^~~~~~~

quoted

net/mpls/af_mpls.c:114:4: note: in expansion of macro 'IP6_UPD_PO_STATS'

       IP6_UPD_PO_STATS(dev_net(dev), in6dev,
       ^~~~~~~~~~~~~~~~

quoted

include/net/snmp.h:146:19: error: subscripted value is neither array nor pointer nor vector

      this_cpu_inc(ptr[basefield##PKTS]);  \
                      ^
   include/linux/percpu-defs.h:363:16: note: in definition of macro '__pcpu_size_call'
     switch(sizeof(variable)) {     \
                   ^~~~~~~~
   include/linux/percpu-defs.h:507:28: note: in expansion of macro 'this_cpu_add'
    #define this_cpu_inc(pcp)  this_cpu_add(pcp, 1)
                               ^~~~~~~~~~~~
   include/net/snmp.h:146:3: note: in expansion of macro 'this_cpu_inc'
      this_cpu_inc(ptr[basefield##PKTS]);  \
      ^~~~~~~~~~~~

quoted

include/net/ipv6.h:163:7: note: in expansion of macro 'SNMP_UPD_PO_STATS'

     mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
          ^~~~~~~~~~~~~~~~~

quoted

include/net/ipv6.h:177:3: note: in expansion of macro '_DEVUPD'

      _DEVUPD(net, ipv6, , idev, field, val)
      ^~~~~~~

quoted

net/mpls/af_mpls.c:114:4: note: in expansion of macro 'IP6_UPD_PO_STATS'

       IP6_UPD_PO_STATS(dev_net(dev), in6dev,
       ^~~~~~~~~~~~~~~~
   In file included from arch/x86/include/asm/preempt.h:5:0,
                    from include/linux/preempt.h:59,
                    from include/linux/spinlock.h:50,
                    from include/linux/mm_types.h:8,
                    from include/linux/kmemcheck.h:4,
                    from include/linux/skbuff.h:18,
                    from net/mpls/af_mpls.c:2:

quoted

include/net/snmp.h:146:19: error: subscripted value is neither array nor pointer nor vector

      this_cpu_inc(ptr[basefield##PKTS]);  \
                      ^
   arch/x86/include/asm/percpu.h:128:17: note: in definition of macro 'percpu_add_op'
     typedef typeof(var) pao_T__;     \
                    ^~~
   include/linux/percpu-defs.h:364:11: note: in expansion of macro 'this_cpu_add_1'
      case 1: stem##1(variable, __VA_ARGS__);break;  \
              ^~~~
   include/linux/percpu-defs.h:496:33: note: in expansion of macro '__pcpu_size_call'
    #define this_cpu_add(pcp, val)  __pcpu_size_call(this_cpu_add_, pcp, val)
                                    ^~~~~~~~~~~~~~~~
   include/linux/percpu-defs.h:507:28: note: in expansion of macro 'this_cpu_add'
    #define this_cpu_inc(pcp)  this_cpu_add(pcp, 1)
                               ^~~~~~~~~~~~
   include/net/snmp.h:146:3: note: in expansion of macro 'this_cpu_inc'
      this_cpu_inc(ptr[basefield##PKTS]);  \
      ^~~~~~~~~~~~

quoted

include/net/ipv6.h:163:7: note: in expansion of macro 'SNMP_UPD_PO_STATS'

     mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
          ^~~~~~~~~~~~~~~~~

quoted

include/net/ipv6.h:177:3: note: in expansion of macro '_DEVUPD'

      _DEVUPD(net, ipv6, , idev, field, val)
      ^~~~~~~

quoted

net/mpls/af_mpls.c:114:4: note: in expansion of macro 'IP6_UPD_PO_STATS'

       IP6_UPD_PO_STATS(dev_net(dev), in6dev,
       ^~~~~~~~~~~~~~~~

quoted

include/net/snmp.h:146:19: error: subscripted value is neither array nor pointer nor vector

      this_cpu_inc(ptr[basefield##PKTS]);  \
                      ^
   arch/x86/include/asm/percpu.h:137:17: note: in definition of macro 'percpu_add_op'
     switch (sizeof(var)) {      \
                    ^~~
   include/linux/percpu-defs.h:364:11: note: in expansion of macro 'this_cpu_add_1'
      case 1: stem##1(variable, __VA_ARGS__);break;  \
              ^~~~
   include/linux/percpu-defs.h:496:33: note: in expansion of macro '__pcpu_size_call'
    #define this_cpu_add(pcp, val)  __pcpu_size_call(this_cpu_add_, pcp, val)
                                    ^~~~~~~~~~~~~~~~
   include/linux/percpu-defs.h:507:28: note: in expansion of macro 'this_cpu_add'
    #define this_cpu_inc(pcp)  this_cpu_add(pcp, 1)
                               ^~~~~~~~~~~~
   include/net/snmp.h:146:3: note: in expansion of macro 'this_cpu_inc'
      this_cpu_inc(ptr[basefield##PKTS]);  \
      ^~~~~~~~~~~~

quoted

include/net/ipv6.h:163:7: note: in expansion of macro 'SNMP_UPD_PO_STATS'

     mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
          ^~~~~~~~~~~~~~~~~

vim +163 include/net/ipv6.h

8e7999c4 Pavel Emelyanov 2007-10-15  157  
13415e46 Eric Dumazet    2016-04-27  158  #define _DEVUPD(net, statname, mod, idev, field, val)			\
edf391ff Neil Horman     2009-04-27  159  ({									\
edf391ff Neil Horman     2009-04-27  160  	struct inet6_dev *_idev = (idev);				\
edf391ff Neil Horman     2009-04-27  161  	if (likely(_idev != NULL))					\
13415e46 Eric Dumazet    2016-04-27  162  		mod##SNMP_UPD_PO_STATS((_idev)->stats.statname, field, (val)); \
13415e46 Eric Dumazet    2016-04-27 @163  	mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
edf391ff Neil Horman     2009-04-27  164  })
edf391ff Neil Horman     2009-04-27  165  
14878f75 David L Stevens 2007-09-16  166  /* MIBs */
14878f75 David L Stevens 2007-09-16  167  
087fe240 Denis V. Lunev  2008-10-08  168  #define IP6_INC_STATS(net, idev,field)		\
13415e46 Eric Dumazet    2016-04-27  169  		_DEVINC(net, ipv6, , idev, field)
1d015503 Eric Dumazet    2016-04-27  170  #define __IP6_INC_STATS(net, idev,field)	\
13415e46 Eric Dumazet    2016-04-27  171  		_DEVINC(net, ipv6, __, idev, field)
edf391ff Neil Horman     2009-04-27  172  #define IP6_ADD_STATS(net, idev,field,val)	\
13415e46 Eric Dumazet    2016-04-27  173  		_DEVADD(net, ipv6, , idev, field, val)
1d015503 Eric Dumazet    2016-04-27  174  #define __IP6_ADD_STATS(net, idev,field,val)	\
13415e46 Eric Dumazet    2016-04-27  175  		_DEVADD(net, ipv6, __, idev, field, val)
edf391ff Neil Horman     2009-04-27  176  #define IP6_UPD_PO_STATS(net, idev,field,val)   \
13415e46 Eric Dumazet    2016-04-27 @177  		_DEVUPD(net, ipv6, , idev, field, val)
c2005eb0 Eric Dumazet    2016-04-27  178  #define __IP6_UPD_PO_STATS(net, idev,field,val)   \
13415e46 Eric Dumazet    2016-04-27  179  		_DEVUPD(net, ipv6, __, idev, field, val)
087fe240 Denis V. Lunev  2008-10-08  180  #define ICMP6_INC_STATS(net, idev, field)	\

:::::: The code at line 163 was first introduced by commit
:::::: 13415e46c5915e2dac089de516369005fbc045f9 net: snmp: kill STATS_BH macros

:::::: TO: Eric Dumazet [off-list ref]
:::::: CC: David S. Miller [off-list ref]

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

Hi Robert,

[auto build test ERROR on net-next/master]

url:    https://github.com/0day-ci/linux/commits/Robert-Shearman/net-AF-specific-RTM_GETSTATS-attributes/20170114-095819
config: i386-randconfig-sb0-01141243 (attached as .config)
compiler: gcc-5 (Debian 5.4.1-2) 5.4.1 20160904
reproduce:
        # save the attached .config to linux build tree
        make ARCH=i386 

All errors (new ones prefixed by >>):

   In file included from include/net/netns/mib.h:4:0,
                    from include/net/net_namespace.h:14,
                    from include/linux/netdevice.h:43,
                    from include/uapi/linux/if_arp.h:26,
                    from include/linux/if_arp.h:27,
                    from net/mpls/af_mpls.c:7:
   net/mpls/af_mpls.c: In function 'mpls_stats_inc_outucastpkts':

quoted

include/net/ipv6.h:163:35: error: 'struct netns_mib' has no member named 'ipv6_statistics'

     mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
                                      ^
   include/net/snmp.h:145:15: note: in definition of macro 'SNMP_UPD_PO_STATS'
      __typeof__((mib->mibs) + 0) ptr = mib->mibs; \
                  ^
   include/net/ipv6.h:177:3: note: in expansion of macro '_DEVUPD'
      _DEVUPD(net, ipv6, , idev, field, val)
      ^
   net/mpls/af_mpls.c:114:4: note: in expansion of macro 'IP6_UPD_PO_STATS'
       IP6_UPD_PO_STATS(dev_net(dev), in6dev,
       ^

quoted

include/net/ipv6.h:163:35: error: 'struct netns_mib' has no member named 'ipv6_statistics'

     mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
                                      ^
   include/net/snmp.h:145:37: note: in definition of macro 'SNMP_UPD_PO_STATS'
      __typeof__((mib->mibs) + 0) ptr = mib->mibs; \
                                        ^
   include/net/ipv6.h:177:3: note: in expansion of macro '_DEVUPD'
      _DEVUPD(net, ipv6, , idev, field, val)
      ^
   net/mpls/af_mpls.c:114:4: note: in expansion of macro 'IP6_UPD_PO_STATS'
       IP6_UPD_PO_STATS(dev_net(dev), in6dev,
       ^
   In file included from include/asm-generic/percpu.h:6:0,
                    from arch/x86/include/asm/percpu.h:542,
                    from arch/x86/include/asm/preempt.h:5,
                    from include/linux/preempt.h:59,
                    from include/linux/spinlock.h:50,
                    from include/linux/mm_types.h:8,
                    from include/linux/kmemcheck.h:4,
                    from include/linux/skbuff.h:18,
                    from net/mpls/af_mpls.c:2:
   include/net/snmp.h:146:19: error: subscripted value is neither array nor pointer nor vector
      this_cpu_inc(ptr[basefield##PKTS]);  \
                      ^
   include/linux/percpu-defs.h:206:47: note: in definition of macro '__verify_pcpu_ptr'
     const void __percpu *__vpp_verify = (typeof((ptr) + 0))NULL; \
                                                  ^
   include/linux/percpu-defs.h:496:33: note: in expansion of macro '__pcpu_size_call'
    #define this_cpu_add(pcp, val)  __pcpu_size_call(this_cpu_add_, pcp, val)
                                    ^
   include/linux/percpu-defs.h:507:28: note: in expansion of macro 'this_cpu_add'
    #define this_cpu_inc(pcp)  this_cpu_add(pcp, 1)
                               ^
   include/net/snmp.h:146:3: note: in expansion of macro 'this_cpu_inc'
      this_cpu_inc(ptr[basefield##PKTS]);  \
      ^
   include/net/ipv6.h:163:7: note: in expansion of macro 'SNMP_UPD_PO_STATS'
     mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
          ^
   include/net/ipv6.h:177:3: note: in expansion of macro '_DEVUPD'
      _DEVUPD(net, ipv6, , idev, field, val)
      ^
   net/mpls/af_mpls.c:114:4: note: in expansion of macro 'IP6_UPD_PO_STATS'
       IP6_UPD_PO_STATS(dev_net(dev), in6dev,
       ^
   include/net/snmp.h:146:19: error: subscripted value is neither array nor pointer nor vector
      this_cpu_inc(ptr[basefield##PKTS]);  \
                      ^
   include/linux/percpu-defs.h:363:16: note: in definition of macro '__pcpu_size_call'
     switch(sizeof(variable)) {     \
                   ^
   include/linux/percpu-defs.h:507:28: note: in expansion of macro 'this_cpu_add'
    #define this_cpu_inc(pcp)  this_cpu_add(pcp, 1)
                               ^
   include/net/snmp.h:146:3: note: in expansion of macro 'this_cpu_inc'
      this_cpu_inc(ptr[basefield##PKTS]);  \
      ^
   include/net/ipv6.h:163:7: note: in expansion of macro 'SNMP_UPD_PO_STATS'
     mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
          ^
   include/net/ipv6.h:177:3: note: in expansion of macro '_DEVUPD'
      _DEVUPD(net, ipv6, , idev, field, val)
      ^
   net/mpls/af_mpls.c:114:4: note: in expansion of macro 'IP6_UPD_PO_STATS'
       IP6_UPD_PO_STATS(dev_net(dev), in6dev,
       ^
   In file included from arch/x86/include/asm/preempt.h:5:0,
                    from include/linux/preempt.h:59,
                    from include/linux/spinlock.h:50,
                    from include/linux/mm_types.h:8,
                    from include/linux/kmemcheck.h:4,
                    from include/linux/skbuff.h:18,
                    from net/mpls/af_mpls.c:2:
   include/net/snmp.h:146:19: error: subscripted value is neither array nor pointer nor vector
      this_cpu_inc(ptr[basefield##PKTS]);  \
                      ^
   arch/x86/include/asm/percpu.h:128:17: note: in definition of macro 'percpu_add_op'
     typedef typeof(var) pao_T__;     \
                    ^
   include/linux/percpu-defs.h:364:11: note: in expansion of macro 'this_cpu_add_1'
      case 1: stem##1(variable, __VA_ARGS__);break;  \
              ^
   include/linux/percpu-defs.h:496:33: note: in expansion of macro '__pcpu_size_call'
    #define this_cpu_add(pcp, val)  __pcpu_size_call(this_cpu_add_, pcp, val)
                                    ^
   include/linux/percpu-defs.h:507:28: note: in expansion of macro 'this_cpu_add'
    #define this_cpu_inc(pcp)  this_cpu_add(pcp, 1)
                               ^
   include/net/snmp.h:146:3: note: in expansion of macro 'this_cpu_inc'
      this_cpu_inc(ptr[basefield##PKTS]);  \
      ^
   include/net/ipv6.h:163:7: note: in expansion of macro 'SNMP_UPD_PO_STATS'
     mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
          ^
   include/net/ipv6.h:177:3: note: in expansion of macro '_DEVUPD'
      _DEVUPD(net, ipv6, , idev, field, val)
      ^
   net/mpls/af_mpls.c:114:4: note: in expansion of macro 'IP6_UPD_PO_STATS'
       IP6_UPD_PO_STATS(dev_net(dev), in6dev,
       ^
   include/net/snmp.h:146:19: error: subscripted value is neither array nor pointer nor vector

vim +163 include/net/ipv6.h

8e7999c4 Pavel Emelyanov 2007-10-15  157  
13415e46 Eric Dumazet    2016-04-27  158  #define _DEVUPD(net, statname, mod, idev, field, val)			\
edf391ff Neil Horman     2009-04-27  159  ({									\
edf391ff Neil Horman     2009-04-27  160  	struct inet6_dev *_idev = (idev);				\
edf391ff Neil Horman     2009-04-27  161  	if (likely(_idev != NULL))					\
13415e46 Eric Dumazet    2016-04-27  162  		mod##SNMP_UPD_PO_STATS((_idev)->stats.statname, field, (val)); \
13415e46 Eric Dumazet    2016-04-27 @163  	mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
edf391ff Neil Horman     2009-04-27  164  })
edf391ff Neil Horman     2009-04-27  165  
14878f75 David L Stevens 2007-09-16  166  /* MIBs */

:::::: The code at line 163 was first introduced by commit
:::::: 13415e46c5915e2dac089de516369005fbc045f9 net: snmp: kill STATS_BH macros

:::::: TO: Eric Dumazet [off-list ref]
:::::: CC: David S. Miller [off-list ref]

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

On 14/01/17 06:41, kbuild test robot wrote:

Hi Robert,

[auto build test ERROR on net-next/master]

url:    https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_0day-2Dci_linux_commits_Robert-2DShearman_net-2DAF-2Dspecific-2DRTM-5FGETSTATS-2Dattributes_20170114-2D095819&d=DwIBAg&c=IL_XqQWOjubgfqINi2jTzg&r=HbzsGgI7IidV3zRVGbZOYKAd0w1ch0IRBV2pqme4k9w&m=-FIZAjvyj_cLBHBTW14fNKp5IiJpkr1laQrxgxKnZ3g&s=_d4vequnI_QXJkUb0X3X8AUHdi6qD4R86g-mTBsEJ7g&e=
config: x86_64-randconfig-u0-01141334 (attached as .config)
compiler: gcc-6 (Debian 6.2.0-3) 6.2.0 20160901
reproduce:
        # save the attached .config to linux build tree
        make ARCH=x86_64

All error/warnings (new ones prefixed by >>):

   In file included from include/net/netns/mib.h:4:0,
                    from include/net/net_namespace.h:14,
                    from include/linux/netdevice.h:43,
                    from include/uapi/linux/if_arp.h:26,
                    from include/linux/if_arp.h:27,
                    from net/mpls/af_mpls.c:7:
   net/mpls/af_mpls.c: In function 'mpls_stats_inc_outucastpkts':

quoted

quoted

include/net/ipv6.h:163:35: error: 'struct netns_mib' has no member named 'ipv6_statistics'; did you mean 'ip_statistics'?

     mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\

Good catch kbuild test robot.

Expect to see a v2 shortly with this CONFIG_IPV6=n build failure fixed.

Thanks,
Rob

This patchset records per-interface packet stats in the MPLS
forwarding path and exports them using a nest of attributes root at a
new IFLA_STATS_AF_SPEC attribute as part of RTM_GETSTATS messages:

[IFLA_STATS_AF_SPEC]
 -> [AF_MPLS]
  -> [MPLS_STATS_LINK]
   -> struct mpls_link_stats

The first patch adds the rtnl infrastructure for this, including a new
callbacks to per-AF ops of fill_stats_af and get_stats_af_size. The
second patch records MPLS stats and makes use of the infrastructure to
export them. The rtnl infrastructure could also be used to export IPv6
stats in the future.

Changes in v2:
 - make incrementing IPv6 stats in mpls_stats_inc_outucastpkts
   conditional on CONFIG_IPV6 to fix build with CONFIG_IPV6=n

Robert Shearman (2):
  net: AF-specific RTM_GETSTATS attributes
  mpls: Packet stats

 include/net/rtnetlink.h      |   4 +
 include/uapi/linux/if_link.h |   1 +
 include/uapi/linux/mpls.h    |  30 +++++++
 net/core/rtnetlink.c         |  50 ++++++++++++
 net/mpls/af_mpls.c           | 181 +++++++++++++++++++++++++++++++++++++------
 net/mpls/internal.h          |  58 +++++++++++++-
 net/mpls/mpls_iptunnel.c     |  11 ++-
 7 files changed, 307 insertions(+), 28 deletions(-)

-- 
2.1.4

Add the functionality for including address-family-specific per-link
stats in RTM_GETSTATS messages. This is done through adding a new
IFLA_STATS_AF_SPEC attribute under which address family attributes are
nested and then the AF-specific attributes can be further nested. This
follows the model of IFLA_AF_SPEC on RTM_*LINK messages and it has the
advantage of presenting an easily extended hierarchy. The rtnl_af_ops
structure is extended to provide AFs with the opportunity to fill and
provide the size of their stats attributes.

One alternative would have been to provide AFs with the ability to add
attributes directly into the RTM_GETSTATS message without a nested
hierarchy. I discounted this approach as it increases the rate at
which the 32 attribute number space is used up and it makes
implementation a little more tricky for stats dump resuming (at the
moment the order in which attributes are added to the message has to
match the numeric order of the attributes).

Another alternative would have been to register per-AF RTM_GETSTATS
handlers. I discounted this approach as I perceived a common use-case
to be getting all the stats for an interface and this approach would
necessitate multiple requests/dumps to retrieve them all.

Signed-off-by: Robert Shearman <redacted>
Acked-by: Roopa Prabhu <redacted>
---
 include/net/rtnetlink.h      |  4 ++++
 include/uapi/linux/if_link.h |  1 +
 net/core/rtnetlink.c         | 50 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 55 insertions(+)

diff --git a/include/net/rtnetlink.h b/include/net/rtnetlink.h
index 4113916cc1bb..106de5f7bf06 100644
--- a/include/net/rtnetlink.h
+++ b/include/net/rtnetlink.h

@@ -139,6 +139,10 @@ struct rtnl_af_ops {
 						    const struct nlattr *attr);
 	int			(*set_link_af)(struct net_device *dev,
 					       const struct nlattr *attr);
+
+	int			(*fill_stats_af)(struct sk_buff *skb,
+						 const struct net_device *dev);
+	size_t			(*get_stats_af_size)(const struct net_device *dev);
 };
 
 void __rtnl_af_unregister(struct rtnl_af_ops *ops);

diff --git a/include/uapi/linux/if_link.h b/include/uapi/linux/if_link.h
index 6b13e591abc9..184b16ed2b84 100644
--- a/include/uapi/linux/if_link.h
+++ b/include/uapi/linux/if_link.h

@@ -847,6 +847,7 @@ enum {
 	IFLA_STATS_LINK_XSTATS,
 	IFLA_STATS_LINK_XSTATS_SLAVE,
 	IFLA_STATS_LINK_OFFLOAD_XSTATS,
+	IFLA_STATS_AF_SPEC,
 	__IFLA_STATS_MAX,
 };
 

diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index 18b5aae99bec..4edc1bd7a735 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c

@@ -3829,6 +3829,39 @@ static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
 		*idxattr = 0;
 	}
 
+	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
+		struct rtnl_af_ops *af_ops;
+
+		*idxattr = IFLA_STATS_AF_SPEC;
+		attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
+		if (!attr)
+			goto nla_put_failure;
+
+		list_for_each_entry(af_ops, &rtnl_af_ops, list) {
+			if (af_ops->fill_stats_af) {
+				struct nlattr *af;
+				int err;
+
+				af = nla_nest_start(skb, af_ops->family);
+				if (!af)
+					goto nla_put_failure;
+
+				err = af_ops->fill_stats_af(skb, dev);
+
+				if (err == -ENODATA)
+					nla_nest_cancel(skb, af);
+				else if (err < 0)
+					goto nla_put_failure;
+
+				nla_nest_end(skb, af);
+			}
+		}
+
+		nla_nest_end(skb, attr);
+
+		*idxattr = 0;
+	}
+
 	nlmsg_end(skb, nlh);
 
 	return 0;

@@ -3885,6 +3918,23 @@ static size_t if_nlmsg_stats_size(const struct net_device *dev,
 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
 		size += rtnl_get_offload_stats_size(dev);
 
+	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
+		struct rtnl_af_ops *af_ops;
+
+		/* for IFLA_STATS_AF_SPEC */
+		size += nla_total_size(0);
+
+		list_for_each_entry(af_ops, &rtnl_af_ops, list) {
+			if (af_ops->get_stats_af_size) {
+				size += nla_total_size(
+					af_ops->get_stats_af_size(dev));
+
+				/* for AF_* */
+				size += nla_total_size(0);
+			}
+		}
+	}
+
 	return size;
 }
 

-- 
2.1.4

Having MPLS packet stats is useful for observing network operation and
for diagnosing network problems. In the absence of anything better,
RFC2863 and RFC3813 are used for guidance for which stats to expose
and the semantics of them. In particular rx_noroutes maps to in
unknown protos in RFC2863. The stats are exposed to userspace via
AF_MPLS attributes embedded in the IFLA_STATS_AF_SPEC attribute of
RTM_GETSTATS messages.

All the introduced fields are 64-bit, even error ones, to ensure no
overflow with long uptimes. Per-CPU counters are used to avoid
cache-line contention on the commonly used fields. The other fields
have also been made per-CPU for code to avoid performance problems in
error conditions on the assumption that on some platforms the cost of
atomic operations could be more expensive than sending the packet
(which is what would be done in the success case). If that's not the
case, we could instead not use per-CPU counters for these fields.

Only unicast and non-fragment are exposed at the moment, but other
counters can be exposed in the future either by adding to the end of
struct mpls_link_stats or by additional netlink attributes in the
AF_MPLS IFLA_STATS_AF_SPEC nested attribute.

Signed-off-by: Robert Shearman <redacted>
---
 include/uapi/linux/mpls.h |  30 ++++++++
 net/mpls/af_mpls.c        | 181 ++++++++++++++++++++++++++++++++++++++++------
 net/mpls/internal.h       |  58 ++++++++++++++-
 net/mpls/mpls_iptunnel.c  |  11 ++-
 4 files changed, 252 insertions(+), 28 deletions(-)

diff --git a/include/uapi/linux/mpls.h b/include/uapi/linux/mpls.h
index 24a6cb1aec86..77a19dfe3990 100644
--- a/include/uapi/linux/mpls.h
+++ b/include/uapi/linux/mpls.h

@@ -43,4 +43,34 @@ struct mpls_label {
 
 #define MPLS_LABEL_FIRST_UNRESERVED	16 /* RFC3032 */
 
+/* These are embedded into IFLA_STATS_AF_SPEC:
+ * [IFLA_STATS_AF_SPEC]
+ * -> [AF_MPLS]
+ *    -> [MPLS_STATS_xxx]
+ *
+ * Attributes:
+ * [MPLS_STATS_LINK] = {
+ *     struct mpls_link_stats
+ * }
+ */
+enum {
+	MPLS_STATS_UNSPEC, /* also used as 64bit pad attribute */
+	MPLS_STATS_LINK,
+	__MPLS_STATS_MAX,
+};
+
+#define MPLS_STATS_MAX (__MPLS_STATS_MAX - 1)
+
+struct mpls_link_stats {
+	__u64	rx_packets;		/* total packets received	*/
+	__u64	tx_packets;		/* total packets transmitted	*/
+	__u64	rx_bytes;		/* total bytes received		*/
+	__u64	tx_bytes;		/* total bytes transmitted	*/
+	__u64	rx_errors;		/* bad packets received		*/
+	__u64	tx_errors;		/* packet transmit problems	*/
+	__u64	rx_dropped;		/* packet dropped on receive	*/
+	__u64	tx_dropped;		/* packet dropped on transmit	*/
+	__u64	rx_noroute;		/* no route for packet dest	*/
+};
+
 #endif /* _UAPI_MPLS_H */

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index 15fe97644ffe..4dc81963af8f 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c

@@ -8,6 +8,7 @@
 #include <linux/ipv6.h>
 #include <linux/mpls.h>
 #include <linux/vmalloc.h>
+#include <linux/percpu.h>
 #include <net/ip.h>
 #include <net/dst.h>
 #include <net/sock.h>

@@ -17,8 +18,8 @@
 #include <net/netns/generic.h>
 #if IS_ENABLED(CONFIG_IPV6)
 #include <net/ipv6.h>
-#include <net/addrconf.h>
 #endif
+#include <net/addrconf.h>
 #include <net/nexthop.h>
 #include "internal.h"
 

@@ -48,11 +49,6 @@ static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned index)
 	return rt;
 }
 
-static inline struct mpls_dev *mpls_dev_get(const struct net_device *dev)
-{
-	return rcu_dereference_rtnl(dev->mpls_ptr);
-}
-
 bool mpls_output_possible(const struct net_device *dev)
 {
 	return dev && (dev->flags & IFF_UP) && netif_carrier_ok(dev);

@@ -98,6 +94,31 @@ bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
 }
 EXPORT_SYMBOL_GPL(mpls_pkt_too_big);
 
+void mpls_stats_inc_outucastpkts(struct net_device *dev,
+				 const struct sk_buff *skb)
+{
+	struct mpls_dev *mdev;
+
+	if (skb->protocol == htons(ETH_P_MPLS_UC)) {
+		mdev = mpls_dev_get(dev);
+		if (mdev)
+			MPLS_INC_STATS_LEN(mdev, skb->len,
+					   tx_packets,
+					   tx_bytes);
+	} else if (skb->protocol == htons(ETH_P_IP)) {
+		IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len);
+#if IS_ENABLED(CONFIG_IPV6)
+	} else if (skb->protocol == htons(ETH_P_IPV6)) {
+		struct inet6_dev *in6dev = __in6_dev_get(dev);
+
+		if (in6dev)
+			IP6_UPD_PO_STATS(dev_net(dev), in6dev,
+					 IPSTATS_MIB_OUT, skb->len);
+#endif
+	}
+}
+EXPORT_SYMBOL_GPL(mpls_stats_inc_outucastpkts);
+
 static u32 mpls_multipath_hash(struct mpls_route *rt,
 			       struct sk_buff *skb, bool bos)
 {

@@ -253,6 +274,7 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 	struct mpls_nh *nh;
 	struct mpls_entry_decoded dec;
 	struct net_device *out_dev;
+	struct mpls_dev *out_mdev;
 	struct mpls_dev *mdev;
 	unsigned int hh_len;
 	unsigned int new_header_size;

@@ -262,17 +284,25 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 	/* Careful this entire function runs inside of an rcu critical section */
 
 	mdev = mpls_dev_get(dev);
-	if (!mdev || !mdev->input_enabled)
+	if (!mdev)
 		goto drop;
 
-	if (skb->pkt_type != PACKET_HOST)
+	MPLS_INC_STATS_LEN(mdev, skb->len, rx_packets,
+			   rx_bytes);
+
+	if (!mdev->input_enabled) {
+		MPLS_INC_STATS(mdev, rx_dropped);
 		goto drop;
+	}
+
+	if (skb->pkt_type != PACKET_HOST)
+		goto err;
 
 	if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL)
-		goto drop;
+		goto err;
 
 	if (!pskb_may_pull(skb, sizeof(*hdr)))
-		goto drop;
+		goto err;
 
 	/* Read and decode the label */
 	hdr = mpls_hdr(skb);

@@ -285,33 +315,35 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 	skb_orphan(skb);
 
 	rt = mpls_route_input_rcu(net, dec.label);
-	if (!rt)
+	if (!rt) {
+		MPLS_INC_STATS(mdev, rx_noroute);
 		goto drop;
+	}
 
 	nh = mpls_select_multipath(rt, skb, dec.bos);
 	if (!nh)
-		goto drop;
-
-	/* Find the output device */
-	out_dev = rcu_dereference(nh->nh_dev);
-	if (!mpls_output_possible(out_dev))
-		goto drop;
+		goto err;
 
 	if (skb_warn_if_lro(skb))
-		goto drop;
+		goto err;
 
 	skb_forward_csum(skb);
 
 	/* Verify ttl is valid */
 	if (dec.ttl <= 1)
-		goto drop;
+		goto err;
 	dec.ttl -= 1;
 
+	/* Find the output device */
+	out_dev = rcu_dereference(nh->nh_dev);
+	if (!mpls_output_possible(out_dev))
+		goto tx_err;
+
 	/* Verify the destination can hold the packet */
 	new_header_size = mpls_nh_header_size(nh);
 	mtu = mpls_dev_mtu(out_dev);
 	if (mpls_pkt_too_big(skb, mtu - new_header_size))
-		goto drop;
+		goto tx_err;
 
 	hh_len = LL_RESERVED_SPACE(out_dev);
 	if (!out_dev->header_ops)

@@ -319,7 +351,7 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 
 	/* Ensure there is enough space for the headers in the skb */
 	if (skb_cow(skb, hh_len + new_header_size))
-		goto drop;
+		goto tx_err;
 
 	skb->dev = out_dev;
 	skb->protocol = htons(ETH_P_MPLS_UC);

@@ -327,7 +359,7 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 	if (unlikely(!new_header_size && dec.bos)) {
 		/* Penultimate hop popping */
 		if (!mpls_egress(rt, skb, dec))
-			goto drop;
+			goto err;
 	} else {
 		bool bos;
 		int i;

@@ -343,6 +375,8 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 		}
 	}
 
+	mpls_stats_inc_outucastpkts(out_dev, skb);
+
 	/* If via wasn't specified then send out using device address */
 	if (nh->nh_via_table == MPLS_NEIGH_TABLE_UNSPEC)
 		err = neigh_xmit(NEIGH_LINK_TABLE, out_dev,

@@ -355,6 +389,13 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 				    __func__, err);
 	return 0;
 
+tx_err:
+	out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
+	if (out_mdev)
+		MPLS_INC_STATS(out_mdev, tx_errors);
+	goto drop;
+err:
+	MPLS_INC_STATS(mdev, rx_errors);
 drop:
 	kfree_skb(skb);
 	return NET_RX_DROP;

@@ -853,6 +894,70 @@ static int mpls_route_del(struct mpls_route_config *cfg)
 	return err;
 }
 
+static void mpls_get_stats(struct mpls_dev *mdev,
+			   struct mpls_link_stats *stats)
+{
+	struct mpls_pcpu_stats *p;
+	int i;
+
+	memset(stats, 0, sizeof(*stats));
+
+	for_each_possible_cpu(i) {
+		struct mpls_link_stats local;
+		unsigned int start;
+
+		p = per_cpu_ptr(mdev->stats, i);
+		do {
+			start = u64_stats_fetch_begin(&p->syncp);
+			local = p->stats;
+		} while (u64_stats_fetch_retry(&p->syncp, start));
+
+		stats->rx_packets	+= local.rx_packets;
+		stats->rx_bytes		+= local.rx_bytes;
+		stats->tx_packets	+= local.tx_packets;
+		stats->tx_bytes		+= local.tx_bytes;
+		stats->rx_errors	+= local.rx_errors;
+		stats->tx_errors	+= local.tx_errors;
+		stats->rx_dropped	+= local.rx_dropped;
+		stats->tx_dropped	+= local.tx_dropped;
+		stats->rx_noroute	+= local.rx_noroute;
+	}
+}
+
+static int mpls_fill_stats_af(struct sk_buff *skb,
+			      const struct net_device *dev)
+{
+	struct mpls_link_stats *stats;
+	struct mpls_dev *mdev;
+	struct nlattr *nla;
+
+	mdev = mpls_dev_get(dev);
+	if (!mdev)
+		return -ENODATA;
+
+	nla = nla_reserve_64bit(skb, MPLS_STATS_LINK,
+				sizeof(struct mpls_link_stats),
+				MPLS_STATS_UNSPEC);
+	if (!nla)
+		return -EMSGSIZE;
+
+	stats = nla_data(nla);
+	mpls_get_stats(mdev, stats);
+
+	return 0;
+}
+
+static size_t mpls_get_stats_af_size(const struct net_device *dev)
+{
+	struct mpls_dev *mdev;
+
+	mdev = mpls_dev_get(dev);
+	if (!mdev)
+		return 0;
+
+	return nla_total_size_64bit(sizeof(struct mpls_link_stats));
+}
+
 #define MPLS_PERDEV_SYSCTL_OFFSET(field)	\
 	(&((struct mpls_dev *)0)->field)
 

@@ -911,6 +1016,7 @@ static struct mpls_dev *mpls_add_dev(struct net_device *dev)
 {
 	struct mpls_dev *mdev;
 	int err = -ENOMEM;
+	int i;
 
 	ASSERT_RTNL();
 

@@ -918,6 +1024,17 @@ static struct mpls_dev *mpls_add_dev(struct net_device *dev)
 	if (!mdev)
 		return ERR_PTR(err);
 
+	mdev->stats = alloc_percpu(struct mpls_pcpu_stats);
+	if (!mdev->stats)
+		goto free;
+
+	for_each_possible_cpu(i) {
+		struct mpls_pcpu_stats *mpls_stats;
+
+		mpls_stats = per_cpu_ptr(mdev->stats, i);
+		u64_stats_init(&mpls_stats->syncp);
+	}
+
 	err = mpls_dev_sysctl_register(dev, mdev);
 	if (err)
 		goto free;

@@ -927,10 +1044,19 @@ static struct mpls_dev *mpls_add_dev(struct net_device *dev)
 	return mdev;
 
 free:
+	free_percpu(mdev->stats);
 	kfree(mdev);
 	return ERR_PTR(err);
 }
 
+static void mpls_dev_destroy_rcu(struct rcu_head *head)
+{
+	struct mpls_dev *mdev = container_of(head, struct mpls_dev, rcu);
+
+	free_percpu(mdev->stats);
+	kfree(mdev);
+}
+
 static void mpls_ifdown(struct net_device *dev, int event)
 {
 	struct mpls_route __rcu **platform_label;

@@ -1045,7 +1171,7 @@ static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
 		if (mdev) {
 			mpls_dev_sysctl_unregister(mdev);
 			RCU_INIT_POINTER(dev->mpls_ptr, NULL);
-			kfree_rcu(mdev, rcu);
+			call_rcu(&mdev->rcu, mpls_dev_destroy_rcu);
 		}
 		break;
 	case NETDEV_CHANGENAME:

@@ -1706,6 +1832,12 @@ static struct pernet_operations mpls_net_ops = {
 	.exit = mpls_net_exit,
 };
 
+static struct rtnl_af_ops mpls_af_ops __read_mostly = {
+	.family		   = AF_MPLS,
+	.fill_stats_af	   = mpls_fill_stats_af,
+	.get_stats_af_size = mpls_get_stats_af_size,
+};
+
 static int __init mpls_init(void)
 {
 	int err;

@@ -1722,6 +1854,8 @@ static int __init mpls_init(void)
 
 	dev_add_pack(&mpls_packet_type);
 
+	rtnl_af_register(&mpls_af_ops);
+
 	rtnl_register(PF_MPLS, RTM_NEWROUTE, mpls_rtm_newroute, NULL, NULL);
 	rtnl_register(PF_MPLS, RTM_DELROUTE, mpls_rtm_delroute, NULL, NULL);
 	rtnl_register(PF_MPLS, RTM_GETROUTE, NULL, mpls_dump_routes, NULL);

@@ -1738,6 +1872,7 @@ module_init(mpls_init);
 static void __exit mpls_exit(void)
 {
 	rtnl_unregister_all(PF_MPLS);
+	rtnl_af_unregister(&mpls_af_ops);
 	dev_remove_pack(&mpls_packet_type);
 	unregister_netdevice_notifier(&mpls_dev_notifier);
 	unregister_pernet_subsys(&mpls_net_ops);

diff --git a/net/mpls/internal.h b/net/mpls/internal.h
index bdfef6c3271a..d97243034605 100644
--- a/net/mpls/internal.h
+++ b/net/mpls/internal.h

@@ -9,13 +9,58 @@ struct mpls_entry_decoded {
 	u8 bos;
 };
 
+struct mpls_pcpu_stats {
+	struct mpls_link_stats	stats;
+	struct u64_stats_sync	syncp;
+};
+
 struct mpls_dev {
-	int			input_enabled;
+	int				input_enabled;
 
-	struct ctl_table_header *sysctl;
-	struct rcu_head		rcu;
+	struct mpls_pcpu_stats __percpu	*stats;
+
+	struct ctl_table_header		*sysctl;
+	struct rcu_head			rcu;
 };
 
+#if BITS_PER_LONG == 32
+
+#define MPLS_INC_STATS_LEN(mdev, len, pkts_field, bytes_field)		\
+	do {								\
+		__typeof__(*(mdev)->stats) *ptr =			\
+			raw_cpu_ptr((mdev)->stats);			\
+		local_bh_disable();					\
+		u64_stats_update_begin(&ptr->syncp);			\
+		ptr->stats.pkts_field++;				\
+		ptr->stats.bytes_field += (len);			\
+		u64_stats_update_end(&ptr->syncp);			\
+		local_bh_enable();					\
+	} while (0)
+
+#define MPLS_INC_STATS(mdev, field)					\
+	do {								\
+		__typeof__(*(mdev)->stats) *ptr =			\
+			raw_cpu_ptr((mdev)->stats);			\
+		local_bh_disable();					\
+		u64_stats_update_begin(&ptr->syncp);			\
+		ptr->stats.field++;					\
+		u64_stats_update_end(&ptr->syncp);			\
+		local_bh_enable();					\
+	} while (0)
+
+#else
+
+#define MPLS_INC_STATS_LEN(mdev, len, pkts_field, bytes_field)		\
+	do {								\
+		this_cpu_inc((mdev)->stats->stats.pkts_field);		\
+		this_cpu_add((mdev)->stats->stats.bytes_field, (len));	\
+	} while (0)
+
+#define MPLS_INC_STATS(mdev, field)			\
+	this_cpu_inc((mdev)->stats->stats.field)
+
+#endif
+
 struct sk_buff;
 
 #define LABEL_NOT_SPECIFIED (1 << 20)