Use percpu counters with etheripstat.

Input and OK jca@, OK florian@

5 files changed, 109 insertions, 67 deletions

diff --git a/sys/net/if_etherip.c b/sys/net/if_etherip.c
index fecbd2f9d9c..b392bb69b92 100644
--- a/sys/net/if_etherip.c
+++ b/sys/net/if_etherip.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: if_etherip.c,v 1.21 2017/10/25 09:24:09 mpi Exp $	*/
+/*	$OpenBSD: if_etherip.c,v 1.22 2017/11/10 02:37:14 visa Exp $	*/
 /*
  * Copyright (c) 2015 Kazuya GODA <goda@openbsd.org>
  *
@@ -78,8 +78,6 @@ LIST_HEAD(, etherip_softc) etherip_softc_list;
  * net.inet.etherip.allow value. Zero means drop them, all else is acceptance.
  */
 int etherip_allow = 0;
-
-struct etheripstat etheripstat;
 #endif
 
 void etheripattach(int);
@@ -371,7 +369,7 @@ ip_etherip_output(struct ifnet *ifp, struct mbuf *m)
 
 	M_PREPEND(m, sizeof(struct etherip_header), M_DONTWAIT);
 	if (m == NULL) {
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		return ENOBUFS;
 	}
 	eip = mtod(m, struct etherip_header *);
@@ -381,7 +379,7 @@ ip_etherip_output(struct ifnet *ifp, struct mbuf *m)
 
 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
 	if (m == NULL) {
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		return ENOBUFS;
 	}
 	ip = mtod(m, struct ip *);
@@ -402,8 +400,7 @@ ip_etherip_output(struct ifnet *ifp, struct mbuf *m)
 #if NPF > 0
 	pf_pkt_addr_changed(m);
 #endif
-	etheripstat.etherips_opackets++;
-	etheripstat.etherips_obytes += (m->m_pkthdr.len -
+	etheripstat_pkt(etherips_opackets, etherips_obytes, m->m_pkthdr.len -
 	    (sizeof(struct ip) + sizeof(struct etherip_header)));
 
 	return ip_output(m, NULL, NULL, IP_RAWOUTPUT, NULL, NULL, 0);
@@ -430,7 +427,7 @@ ip_etherip_input(struct mbuf **mp, int *offp, int proto, int af)
 
 	if (!etherip_allow && (m->m_flags & (M_AUTH|M_CONF)) == 0) {
 		m_freem(m);
-		etheripstat.etherips_pdrops++;
+		etheripstat_inc(etherips_pdrops);
 		return IPPROTO_DONE;
 	}
 
@@ -460,7 +457,7 @@ ip_etherip_input(struct mbuf **mp, int *offp, int proto, int af)
 		 */
 		return etherip_input(mp, offp, proto, af);
 #else
-		etheripstat.etherips_noifdrops++;
+		etheripstat_inc(etherips_noifdrops);
 		m_freem(m);
 		return IPPROTO_DONE;
 #endif /* NGIF */
@@ -469,25 +466,24 @@ ip_etherip_input(struct mbuf **mp, int *offp, int proto, int af)
 	m_adj(m, *offp);
 	m = *mp = m_pullup(m, sizeof(struct etherip_header));
 	if (m == NULL) {
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		return IPPROTO_DONE;
 	}
 
 	eip = mtod(m, struct etherip_header *);
 	if (eip->eip_ver != ETHERIP_VERSION || eip->eip_pad) {
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		m_freem(m);
 		return IPPROTO_DONE;
 	}
 
-	etheripstat.etherips_ipackets++;
-	etheripstat.etherips_ibytes += (m->m_pkthdr.len -
+	etheripstat_pkt(etherips_ipackets, etherips_ibytes, m->m_pkthdr.len -
 	    sizeof(struct etherip_header));
 
 	m_adj(m, sizeof(struct etherip_header));
 	m = *mp = m_pullup(m, sizeof(struct ether_header));
 	if (m == NULL) {
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		return IPPROTO_DONE;
 	}
 	m->m_flags &= ~(M_BCAST|M_MCAST);
@@ -532,7 +528,7 @@ ip6_etherip_output(struct ifnet *ifp, struct mbuf *m)
 
 	M_PREPEND(m, sizeof(struct etherip_header), M_DONTWAIT);
 	if (m == NULL) {
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		return ENOBUFS;
 	}
 	eip = mtod(m, struct etherip_header *);
@@ -542,7 +538,7 @@ ip6_etherip_output(struct ifnet *ifp, struct mbuf *m)
 
 	M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
 	if (m == NULL) {
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		return ENOBUFS;
 	}
 	ip6 = mtod(m, struct ip6_hdr *);
@@ -564,8 +560,7 @@ ip6_etherip_output(struct ifnet *ifp, struct mbuf *m)
 #if NPF > 0
 	pf_pkt_addr_changed(m);
 #endif
-	etheripstat.etherips_opackets++;
-	etheripstat.etherips_obytes += (m->m_pkthdr.len -
+	etheripstat_pkt(etherips_opackets, etherips_obytes, m->m_pkthdr.len -
 	    (sizeof(struct ip6_hdr) + sizeof(struct etherip_header)));
 
 	return ip6_output(m, 0, NULL, IPV6_MINMTU, 0, NULL);
@@ -590,7 +585,7 @@ ip6_etherip_input(struct mbuf **mp, int *offp, int proto, int af)
 
 	if (!etherip_allow && (m->m_flags & (M_AUTH|M_CONF)) == 0) {
 		m_freem(m);
-		etheripstat.etherips_pdrops++;
+		etheripstat_inc(etherips_pdrops);
 		return IPPROTO_NONE;
 	}
 
@@ -624,7 +619,7 @@ ip6_etherip_input(struct mbuf **mp, int *offp, int proto, int af)
 		 */
 		return etherip_input(mp, offp, proto, af);
 #else
-		etheripstat.etherips_noifdrops++;
+		etheripstat_inc(etherips_noifdrops);
 		m_freem(m);
 		return IPPROTO_DONE;
 #endif /* NGIF */
@@ -633,24 +628,23 @@ ip6_etherip_input(struct mbuf **mp, int *offp, int proto, int af)
 	m_adj(m, *offp);
 	m = *mp = m_pullup(m, sizeof(struct etherip_header));
 	if (m == NULL) {
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		return IPPROTO_DONE;
 	}
 
 	eip = mtod(m, struct etherip_header *);
 	if ((eip->eip_ver != ETHERIP_VERSION) || eip->eip_pad) {
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		m_freem(m);
 		return IPPROTO_DONE;
 	}
-	etheripstat.etherips_ipackets++;
-	etheripstat.etherips_ibytes += (m->m_pkthdr.len -
+	etheripstat_pkt(etherips_ipackets, etherips_ibytes, m->m_pkthdr.len -
 	    sizeof(struct etherip_header));
 
 	m_adj(m, sizeof(struct etherip_header));
 	m = *mp = m_pullup(m, sizeof(struct ether_header));
 	if (m == NULL) {
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		return IPPROTO_DONE;
 	}
 
@@ -683,13 +677,7 @@ ip_etherip_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
 		NET_UNLOCK();
 		return (error);
 	case ETHERIPCTL_STATS:
-		if (newp != NULL)
-			return EPERM;
-		NET_LOCK();
-		error = sysctl_struct(oldp, oldlenp, newp, newlen,
-		    &etheripstat, sizeof(etheripstat));
-		NET_UNLOCK();
-		return (error);
+		return (etherip_sysctl_etheripstat(oldp, oldlenp, newp));
 	default:
 		break;
 	}
diff --git a/sys/net/if_etherip.h b/sys/net/if_etherip.h
index f4c205632ed..6349e671850 100644
--- a/sys/net/if_etherip.h
+++ b/sys/net/if_etherip.h
@@ -27,7 +27,6 @@
  */
 
 extern int etherip_allow;
-extern struct etheripstat etheripstat;
 
 struct etheripstat {
 	u_int64_t	etherips_hdrops;	/* packet shorter than header shows */
diff --git a/sys/netinet/in_proto.c b/sys/netinet/in_proto.c
index 3104a8ff6f3..56a81dee3d8 100644
--- a/sys/netinet/in_proto.c
+++ b/sys/netinet/in_proto.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: in_proto.c,v 1.81 2017/11/05 13:19:59 florian Exp $	*/
+/*	$OpenBSD: in_proto.c,v 1.82 2017/11/10 02:37:14 visa Exp $	*/
 /*	$NetBSD: in_proto.c,v 1.14 1996/02/18 18:58:32 christos Exp $	*/
 
 /*
@@ -279,6 +279,7 @@ struct protosw inetsw[] = {
   .pr_usrreq	= rip_usrreq,
   .pr_attach	= rip_attach,
   .pr_detach	= rip_detach,
+  .pr_init	= etherip_init,
   .pr_sysctl	= etherip_sysctl
 },
 #endif /* NGIF */
diff --git a/sys/netinet/ip_ether.c b/sys/netinet/ip_ether.c
index 1ff526cb0dd..7bc9267949f 100644
--- a/sys/netinet/ip_ether.c
+++ b/sys/netinet/ip_ether.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ip_ether.c,v 1.88 2017/11/06 15:12:43 mpi Exp $  */
+/*	$OpenBSD: ip_ether.c,v 1.89 2017/11/10 02:37:14 visa Exp $  */
 /*
  * The author of this code is Angelos D. Keromytis (kermit@adk.gr)
  *
@@ -82,7 +82,13 @@ struct gif_softc	*etherip_getgif(struct mbuf *);
  */
 int etherip_allow = 0;
 
-struct etheripstat etheripstat;
+struct cpumem *etheripcounters;
+
+void
+etherip_init(void)
+{
+	etheripcounters = counters_alloc(etherips_ncounters);
+}
 
 /*
  * etherip_input gets called when we receive an encapsulated packet.
@@ -96,7 +102,7 @@ etherip_input(struct mbuf **mp, int *offp, int proto, int af)
 		/* If we do not accept EtherIP explicitly, drop. */
 		if (!etherip_allow && ((*mp)->m_flags & (M_AUTH|M_CONF)) == 0) {
 			DPRINTF(("%s: dropped due to policy\n", __func__));
-			etheripstat.etherips_pdrops++;
+			etheripstat_inc(etherips_pdrops);
 			m_freemp(mp);
 			return IPPROTO_DONE;
 		}
@@ -110,7 +116,7 @@ etherip_input(struct mbuf **mp, int *offp, int proto, int af)
 #endif
 	default:
 		DPRINTF(("%s: dropped, unhandled protocol\n", __func__));
-		etheripstat.etherips_pdrops++;
+		etheripstat_inc(etherips_pdrops);
 		m_freemp(mp);
 		return IPPROTO_DONE;
 	}
@@ -124,7 +130,7 @@ etherip_decap(struct mbuf *m, int iphlen)
 	struct gif_softc *sc;
 	struct mbuf_list ml = MBUF_LIST_INITIALIZER();
 
-	etheripstat.etherips_ipackets++;
+	etheripstat_inc(etherips_ipackets);
 
 	/*
 	 * Make sure there's at least an ethernet header's and an EtherIP
@@ -133,7 +139,7 @@ etherip_decap(struct mbuf *m, int iphlen)
 	if (m->m_pkthdr.len < iphlen + sizeof(struct ether_header) +
 	    sizeof(struct etherip_header)) {
 		DPRINTF(("%s: encapsulated packet too short\n", __func__));
-		etheripstat.etherips_hdrops++;
+		etheripstat_inc(etherips_hdrops);
 		m_freem(m);
 		return;
 	}
@@ -145,7 +151,7 @@ etherip_decap(struct mbuf *m, int iphlen)
 	} else {
 		DPRINTF(("%s: received EtherIP version number %d not "
 		    "supported\n", __func__, eip.eip_ver));
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		m_freem(m);
 		return;
 	}
@@ -153,7 +159,7 @@ etherip_decap(struct mbuf *m, int iphlen)
 	/* Finally, the pad value must be zero. */
 	if (eip.eip_pad) {
 		DPRINTF(("%s: received EtherIP invalid pad value\n", __func__));
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		m_freem(m);
 		return;
 	}
@@ -164,7 +170,7 @@ etherip_decap(struct mbuf *m, int iphlen)
 		if ((m = m_pullup(m, iphlen + sizeof(struct ether_header) +
 		    sizeof(struct etherip_header))) == NULL) {
 			DPRINTF(("%s: m_pullup() failed\n", __func__));
-			etheripstat.etherips_adrops++;
+			etheripstat_inc(etherips_adrops);
 			return;
 		}
 	}
@@ -174,7 +180,7 @@ etherip_decap(struct mbuf *m, int iphlen)
 		return;
 	if (sc->gif_if.if_bridgeport == NULL) {
 		DPRINTF(("%s: interface not part of bridge\n", __func__));
-		etheripstat.etherips_noifdrops++;
+		etheripstat_inc(etherips_noifdrops);
 		m_freem(m);
 		return;
 	}
@@ -183,7 +189,7 @@ etherip_decap(struct mbuf *m, int iphlen)
 	m_adj(m, iphlen + sizeof(struct etherip_header));
 
 	/* Statistics */
-	etheripstat.etherips_ibytes += m->m_pkthdr.len;
+	etheripstat_add(etherips_ibytes, m->m_pkthdr.len);
 
 	/* Reset the flags based on the inner packet */
 	m->m_flags &= ~(M_BCAST|M_MCAST|M_AUTH|M_CONF|M_PROTO1);
@@ -203,7 +209,7 @@ mplsip_decap(struct mbuf *m, int iphlen)
 {
 	struct gif_softc *sc;
 
-	etheripstat.etherips_ipackets++;
+	etheripstat_inc(etherips_ipackets);
 
 	/*
 	 * Make sure there's at least one MPLS label worth of data after
@@ -211,7 +217,7 @@ mplsip_decap(struct mbuf *m, int iphlen)
 	 */
 	if (m->m_pkthdr.len < iphlen + sizeof(struct shim_hdr)) {
 		DPRINTF(("%s: encapsulated packet too short\n", __func__));
-		etheripstat.etherips_hdrops++;
+		etheripstat_inc(etherips_hdrops);
 		m_freem(m);
 		return;
 	}
@@ -221,7 +227,7 @@ mplsip_decap(struct mbuf *m, int iphlen)
 		if ((m = m_pullup(m, iphlen + sizeof(struct shim_hdr))) ==
 		    NULL) {
 			DPRINTF(("%s: m_pullup() failed\n", __func__));
-			etheripstat.etherips_adrops++;
+			etheripstat_inc(etherips_adrops);
 			return;
 		}
 	}
@@ -234,7 +240,7 @@ mplsip_decap(struct mbuf *m, int iphlen)
 	m_adj(m, iphlen);
 
 	/* Statistics */
-	etheripstat.etherips_ibytes += m->m_pkthdr.len;
+	etheripstat_add(etherips_ibytes, m->m_pkthdr.len);
 
 	/* Reset the flags based */
 	m->m_flags &= ~(M_BCAST|M_MCAST);
@@ -292,7 +298,7 @@ etherip_getgif(struct mbuf *m)
 	default:
 		DPRINTF(("%s: invalid protocol %d\n", __func__, v));
 		m_freem(m);
-		etheripstat.etherips_hdrops++;
+		etheripstat_inc(etherips_hdrops);
 		return NULL;
 	}
 
@@ -311,7 +317,7 @@ etherip_getgif(struct mbuf *m)
 	/* None found. */
 	if (sc == NULL) {
 		DPRINTF(("%s: no interface found\n", __func__));
-		etheripstat.etherips_noifdrops++;
+		etheripstat_inc(etherips_noifdrops);
 		m_freem(m);
 		return NULL;
 	}
@@ -335,7 +341,7 @@ etherip_output(struct mbuf *m, struct tdb *tdb, struct mbuf **mp, int proto)
 	    (tdb->tdb_src.sa.sa_family != AF_INET6)) {
 		DPRINTF(("%s: IP in protocol-family <%d> attempted, aborting",
 		    __func__, tdb->tdb_src.sa.sa_family));
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		m_freem(m);
 		return EINVAL;
 	}
@@ -344,7 +350,7 @@ etherip_output(struct mbuf *m, struct tdb *tdb, struct mbuf **mp, int proto)
 	    (tdb->tdb_dst.sa.sa_family != AF_INET6)) {
 		DPRINTF(("%s: IP in protocol-family <%d> attempted, aborting",
 		    __func__, tdb->tdb_dst.sa.sa_family));
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		m_freem(m);
 		return EINVAL;
 	}
@@ -353,7 +359,7 @@ etherip_output(struct mbuf *m, struct tdb *tdb, struct mbuf **mp, int proto)
 		DPRINTF(("%s: mismatch in tunnel source and destination address"
 		    " protocol families (%d/%d), aborting", __func__,
 		    tdb->tdb_src.sa.sa_family, tdb->tdb_dst.sa.sa_family));
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		m_freem(m);
 		return EINVAL;
 	}
@@ -370,7 +376,7 @@ etherip_output(struct mbuf *m, struct tdb *tdb, struct mbuf **mp, int proto)
 	default:
 		DPRINTF(("%s: unsupported tunnel protocol family <%d>, "
 		    "aborting", __func__, tdb->tdb_dst.sa.sa_family));
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		m_freem(m);
 		return EINVAL;
 	}
@@ -382,7 +388,7 @@ etherip_output(struct mbuf *m, struct tdb *tdb, struct mbuf **mp, int proto)
 	M_PREPEND(m, hlen, M_DONTWAIT);
 	if (m == NULL) {
 		DPRINTF(("%s: M_PREPEND failed\n", __func__));
-		etheripstat.etherips_adrops++;
+		etheripstat_inc(etherips_adrops);
 		return ENOBUFS;
 	}
 
@@ -399,8 +405,8 @@ etherip_output(struct mbuf *m, struct tdb *tdb, struct mbuf **mp, int proto)
 	}
 
 	/* Statistics */
-	etheripstat.etherips_opackets++;
-	etheripstat.etherips_obytes += m->m_pkthdr.len - hlen;
+	etheripstat_pkt(etherips_opackets, etherips_obytes,
+	    m->m_pkthdr.len - hlen);
 
 	switch (tdb->tdb_dst.sa.sa_family) {
 	case AF_INET:
@@ -481,15 +487,23 @@ etherip_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
 		NET_UNLOCK();
 		return (error);
 	case ETHERIPCTL_STATS:
-		if (newp != NULL)
-			return (EPERM);
-		NET_LOCK();
-		error = sysctl_struct(oldp, oldlenp, newp, newlen,
-		    &etheripstat, sizeof(etheripstat));
-		NET_UNLOCK();
-		return (error);
+		return (etherip_sysctl_etheripstat(oldp, oldlenp, newp));
 	default:
 		return (ENOPROTOOPT);
 	}
 	/* NOTREACHED */
 }
+
+int
+etherip_sysctl_etheripstat(void *oldp, size_t *oldlenp, void *newp)
+{
+	struct etheripstat etheripstat;
+
+	CTASSERT(sizeof(etheripstat) == (etherips_ncounters *
+	    sizeof(uint64_t)));
+	memset(&etheripstat, 0, sizeof etheripstat);
+	counters_read(etheripcounters, (uint64_t *)&etheripstat,
+	    etherips_ncounters);
+	return (sysctl_rdstruct(oldp, oldlenp, newp, &etheripstat,
+	    sizeof(etheripstat)));
+}
diff --git a/sys/netinet/ip_ether.h b/sys/netinet/ip_ether.h
index 88f6875eaa1..d48ca952a46 100644
--- a/sys/netinet/ip_ether.h
+++ b/sys/netinet/ip_ether.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ip_ether.h,v 1.22 2017/04/14 20:46:31 bluhm Exp $ */
+/*	$OpenBSD: ip_ether.h,v 1.23 2017/11/10 02:37:14 visa Exp $ */
 /*
  * The author of this code is Angelos D. Keromytis (angelos@adk.gr)
  *
@@ -69,13 +69,53 @@ struct etherip_header {
 }
 
 #ifdef _KERNEL
+
+#include <sys/percpu.h>
+
+enum etheripstat_counters {
+	etherips_hdrops,		/* packet shorter than header shows */
+	etherips_qfull,			/* bridge queue full, packet dropped */
+	etherips_noifdrops,		/* no interface/bridge information */
+	etherips_pdrops,		/* packet dropped due to policy */
+	etherips_adrops,		/* all other drops */
+	etherips_ipackets,		/* total input packets */
+	etherips_opackets,		/* total output packets */
+	etherips_ibytes,		/* input bytes */
+	etherips_obytes,		/* output bytes */
+
+	etherips_ncounters
+};
+
+extern struct cpumem *etheripcounters;
+
+static inline void
+etheripstat_inc(enum etheripstat_counters c)
+{
+	counters_inc(etheripcounters, c);
+}
+
+static inline void
+etheripstat_add(enum etheripstat_counters c, uint64_t v)
+{
+	counters_add(etheripcounters, c, v);
+}
+
+static inline void
+etheripstat_pkt(enum etheripstat_counters pcounter,
+    enum etheripstat_counters bcounter, uint64_t v)
+{
+	counters_pkt(etheripcounters, pcounter, bcounter, v);
+}
+
 struct tdb;
 
+void	etherip_init(void);
 int	etherip_output(struct mbuf *, struct tdb *, struct mbuf **, int);
 int	etherip_input(struct mbuf **, int *, int, int);
 int	etherip_sysctl(int *, u_int, void *, size_t *, void *, size_t);
+int	etherip_sysctl_etheripstat(void *, size_t *, void *);
 
 extern int etherip_allow;
-extern struct etheripstat etheripstat;
+
 #endif /* _KERNEL */
 #endif /* _NETINET_IP_ETHER_H_ */