7 files changed, 205 insertions, 2 deletions

diff --git a/sys/net/ieee8023ad_lacp.c b/sys/net/ieee8023ad_lacp.c
index 4863ac98035b..bf73a5c75e2b 100644
--- a/sys/net/ieee8023ad_lacp.c
+++ b/sys/net/ieee8023ad_lacp.c
@@ -30,6 +30,8 @@
 #include <sys/cdefs.h>
 __FBSDID("$FreeBSD$");
 
+#include "opt_ratelimit.h"
+
 #include <sys/param.h>
 #include <sys/callout.h>
 #include <sys/eventhandler.h>
@@ -853,6 +855,35 @@ lacp_select_tx_port(struct lagg_softc *sc, struct mbuf *m)
 
 	return (lp->lp_lagg);
 }
+
+#ifdef RATELIMIT
+struct lagg_port *
+lacp_select_tx_port_by_hash(struct lagg_softc *sc, uint32_t flowid)
+{
+	struct lacp_softc *lsc = LACP_SOFTC(sc);
+	struct lacp_portmap *pm;
+	struct lacp_port *lp;
+	uint32_t hash;
+
+	if (__predict_false(lsc->lsc_suppress_distributing)) {
+		LACP_DPRINTF((NULL, "%s: waiting transit\n", __func__));
+		return (NULL);
+	}
+
+	pm = &lsc->lsc_pmap[lsc->lsc_activemap];
+	if (pm->pm_count == 0) {
+		LACP_DPRINTF((NULL, "%s: no active aggregator\n", __func__));
+		return (NULL);
+	}
+
+	hash = flowid >> sc->flowid_shift;
+	hash %= pm->pm_count;
+	lp = pm->pm_map[hash];
+
+	return (lp->lp_lagg);
+}
+#endif
+
 /*
  * lacp_suppress_distributing: drop transmit packets for a while
  * to preserve packet ordering.
diff --git a/sys/net/ieee8023ad_lacp.h b/sys/net/ieee8023ad_lacp.h
index 8f0f51a71c48..b26e2c9251f2 100644
--- a/sys/net/ieee8023ad_lacp.h
+++ b/sys/net/ieee8023ad_lacp.h
@@ -284,6 +284,9 @@ struct lacp_softc {
 
 struct mbuf	*lacp_input(struct lagg_port *, struct mbuf *);
 struct lagg_port *lacp_select_tx_port(struct lagg_softc *, struct mbuf *);
+#ifdef RATELIMIT
+struct lagg_port *lacp_select_tx_port_by_hash(struct lagg_softc *, uint32_t);
+#endif
 void		lacp_attach(struct lagg_softc *);
 void		lacp_detach(void *);
 void		lacp_init(struct lagg_softc *);
diff --git a/sys/net/if.h b/sys/net/if.h
index 98ae0a823cd3..26f078700100 100644
--- a/sys/net/if.h
+++ b/sys/net/if.h
@@ -239,6 +239,7 @@ struct if_data {
 #define	IFCAP_RXCSUM_IPV6	0x200000  /* can offload checksum on IPv6 RX */
 #define	IFCAP_TXCSUM_IPV6	0x400000  /* can offload checksum on IPv6 TX */
 #define	IFCAP_HWSTATS		0x800000 /* manages counters internally */
+#define	IFCAP_TXRTLMT		0x1000000 /* hardware supports TX rate limiting */
 
 #define IFCAP_HWCSUM_IPV6	(IFCAP_RXCSUM_IPV6 | IFCAP_TXCSUM_IPV6)
 
diff --git a/sys/net/if_dead.c b/sys/net/if_dead.c
index 8b36ab348628..0e32dcf29116 100644
--- a/sys/net/if_dead.c
+++ b/sys/net/if_dead.c
@@ -100,6 +100,30 @@ ifdead_get_counter(struct ifnet *ifp, ift_counter cnt)
 	return (0);
 }
 
+static int
+ifdead_snd_tag_alloc(struct ifnet *ifp, union if_snd_tag_alloc_params *params,
+    struct m_snd_tag **ppmt)
+{
+	return (EOPNOTSUPP);
+}
+
+static int
+ifdead_snd_tag_modify(struct m_snd_tag *pmt, union if_snd_tag_modify_params *params)
+{
+	return (EOPNOTSUPP);
+}
+
+static int
+ifdead_snd_tag_query(struct m_snd_tag *pmt, union if_snd_tag_query_params *params)
+{
+	return (EOPNOTSUPP);
+}
+
+static void
+ifdead_snd_tag_free(struct m_snd_tag *pmt)
+{
+}
+
 void
 if_dead(struct ifnet *ifp)
 {
@@ -112,4 +136,8 @@ if_dead(struct ifnet *ifp)
 	ifp->if_qflush = ifdead_qflush;
 	ifp->if_transmit = ifdead_transmit;
 	ifp->if_get_counter = ifdead_get_counter;
+	ifp->if_snd_tag_alloc = ifdead_snd_tag_alloc;
+	ifp->if_snd_tag_modify = ifdead_snd_tag_modify;
+	ifp->if_snd_tag_query = ifdead_snd_tag_query;
+	ifp->if_snd_tag_free = ifdead_snd_tag_free;
 }
diff --git a/sys/net/if_lagg.c b/sys/net/if_lagg.c
index 0a5fc159e57b..b44fdc6e6c9c 100644
--- a/sys/net/if_lagg.c
+++ b/sys/net/if_lagg.c
@@ -23,6 +23,7 @@ __FBSDID("$FreeBSD$");
 
 #include "opt_inet.h"
 #include "opt_inet6.h"
+#include "opt_ratelimit.h"
 
 #include <sys/param.h>
 #include <sys/kernel.h>
@@ -118,6 +119,11 @@ static void	lagg_port2req(struct lagg_port *, struct lagg_reqport *);
 static void	lagg_init(void *);
 static void	lagg_stop(struct lagg_softc *);
 static int	lagg_ioctl(struct ifnet *, u_long, caddr_t);
+#ifdef RATELIMIT
+static int	lagg_snd_tag_alloc(struct ifnet *,
+		    union if_snd_tag_alloc_params *,
+		    struct m_snd_tag **);
+#endif
 static int	lagg_ether_setmulti(struct lagg_softc *);
 static int	lagg_ether_cmdmulti(struct lagg_port *, int);
 static	int	lagg_setflag(struct lagg_port *, int, int,
@@ -503,7 +509,12 @@ lagg_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 	ifp->if_ioctl = lagg_ioctl;
 	ifp->if_get_counter = lagg_get_counter;
 	ifp->if_flags = IFF_SIMPLEX | IFF_BROADCAST | IFF_MULTICAST;
+#ifdef RATELIMIT
+	ifp->if_snd_tag_alloc = lagg_snd_tag_alloc;
+	ifp->if_capenable = ifp->if_capabilities = IFCAP_HWSTATS | IFCAP_TXRTLMT;
+#else
 	ifp->if_capenable = ifp->if_capabilities = IFCAP_HWSTATS;
+#endif
 
 	/*
 	 * Attach as an ordinary ethernet device, children will be attached
@@ -1549,6 +1560,52 @@ lagg_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 	return (error);
 }
 
+#ifdef RATELIMIT
+static int
+lagg_snd_tag_alloc(struct ifnet *ifp,
+    union if_snd_tag_alloc_params *params,
+    struct m_snd_tag **ppmt)
+{
+	struct lagg_softc *sc = (struct lagg_softc *)ifp->if_softc;
+	struct lagg_port *lp;
+	struct lagg_lb *lb;
+	uint32_t p;
+
+	switch (sc->sc_proto) {
+	case LAGG_PROTO_FAILOVER:
+		lp = lagg_link_active(sc, sc->sc_primary);
+		break;
+	case LAGG_PROTO_LOADBALANCE:
+		if ((sc->sc_opts & LAGG_OPT_USE_FLOWID) == 0 ||
+		    params->hdr.flowtype == M_HASHTYPE_NONE)
+			return (EOPNOTSUPP);
+		p = params->hdr.flowid >> sc->flowid_shift;
+		p %= sc->sc_count;
+		lb = (struct lagg_lb *)sc->sc_psc;
+		lp = lb->lb_ports[p];
+		lp = lagg_link_active(sc, lp);
+		break;
+	case LAGG_PROTO_LACP:
+		if ((sc->sc_opts & LAGG_OPT_USE_FLOWID) == 0 ||
+		    params->hdr.flowtype == M_HASHTYPE_NONE)
+			return (EOPNOTSUPP);
+		lp = lacp_select_tx_port_by_hash(sc, params->hdr.flowid);
+		break;
+	default:
+		return (EOPNOTSUPP);
+	}
+	if (lp == NULL)
+		return (EOPNOTSUPP);
+	ifp = lp->lp_ifp;
+	if (ifp == NULL || ifp->if_snd_tag_alloc == NULL ||
+	    (ifp->if_capenable & IFCAP_TXRTLMT) == 0)
+		return (EOPNOTSUPP);
+
+	/* forward allocation request */
+	return (ifp->if_snd_tag_alloc(ifp, params, ppmt));
+}
+#endif
+
 static int
 lagg_ether_setmulti(struct lagg_softc *sc)
 {
diff --git a/sys/net/if_var.h b/sys/net/if_var.h
index 62c4bfdd33fb..7c8fd7ec80a4 100644
--- a/sys/net/if_var.h
+++ b/sys/net/if_var.h
@@ -175,6 +175,49 @@ struct if_encap_req {
 
 #define	IFENCAP_FLAG_BROADCAST	0x02	/* Destination is broadcast */
 
+/*
+ * Network interface send tag support. The storage of "struct
+ * m_snd_tag" comes from the network driver and it is free to allocate
+ * as much additional space as it wants for its own use.
+ */
+struct m_snd_tag;
+
+#define	IF_SND_TAG_TYPE_RATE_LIMIT 0
+#define	IF_SND_TAG_TYPE_MAX 1
+
+struct if_snd_tag_alloc_header {
+	uint32_t type;		/* send tag type, see IF_SND_TAG_XXX */
+	uint32_t flowid;	/* mbuf hash value */
+	uint32_t flowtype;	/* mbuf hash type */
+};
+
+struct if_snd_tag_alloc_rate_limit {
+	struct if_snd_tag_alloc_header hdr;
+	uint64_t max_rate;	/* in bytes/s */
+};
+
+struct if_snd_tag_rate_limit_params {
+	uint64_t max_rate;	/* in bytes/s */
+};
+
+union if_snd_tag_alloc_params {
+	struct if_snd_tag_alloc_header hdr;
+	struct if_snd_tag_alloc_rate_limit rate_limit;
+};
+
+union if_snd_tag_modify_params {
+	struct if_snd_tag_rate_limit_params rate_limit;
+};
+
+union if_snd_tag_query_params {
+	struct if_snd_tag_rate_limit_params rate_limit;
+};
+
+typedef int (if_snd_tag_alloc_t)(struct ifnet *, union if_snd_tag_alloc_params *,
+    struct m_snd_tag **);
+typedef int (if_snd_tag_modify_t)(struct m_snd_tag *, union if_snd_tag_modify_params *);
+typedef int (if_snd_tag_query_t)(struct m_snd_tag *, union if_snd_tag_query_params *);
+typedef void (if_snd_tag_free_t)(struct m_snd_tag *);
 
 /*
  * Structure defining a network interface.
@@ -304,12 +347,19 @@ struct ifnet {
 	u_int	if_hw_tsomaxsegsize;	/* TSO maximum segment size in bytes */
 
 	/*
+	 * Network adapter send tag support:
+	 */
+	if_snd_tag_alloc_t *if_snd_tag_alloc;
+	if_snd_tag_modify_t *if_snd_tag_modify;
+	if_snd_tag_query_t *if_snd_tag_query;
+	if_snd_tag_free_t *if_snd_tag_free;
+
+	/*
 	 * Spare fields to be added before branching a stable branch, so
 	 * that structure can be enhanced without changing the kernel
 	 * binary interface.
 	 */
-	void	*if_pspare[4];		/* packet pacing / general use */
-	int	if_ispare[4];		/* packet pacing / general use */
+	int	if_ispare[4];		/* general use */
 };
 
 /* for compatibility with other BSDs */
diff --git a/sys/net/if_vlan.c b/sys/net/if_vlan.c
index 73470dc80e3f..608826a6dc92 100644
--- a/sys/net/if_vlan.c
+++ b/sys/net/if_vlan.c
@@ -46,6 +46,7 @@ __FBSDID("$FreeBSD$");
 
 #include "opt_inet.h"
 #include "opt_vlan.h"
+#include "opt_ratelimit.h"
 
 #include <sys/param.h>
 #include <sys/eventhandler.h>
@@ -212,6 +213,10 @@ static	void trunk_destroy(struct ifvlantrunk *trunk);
 static	void vlan_init(void *foo);
 static	void vlan_input(struct ifnet *ifp, struct mbuf *m);
 static	int vlan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t addr);
+#ifdef RATELIMIT
+static	int vlan_snd_tag_alloc(struct ifnet *,
+    union if_snd_tag_alloc_params *, struct m_snd_tag **);
+#endif
 static	void vlan_qflush(struct ifnet *ifp);
 static	int vlan_setflag(struct ifnet *ifp, int flag, int status,
     int (*func)(struct ifnet *, int));
@@ -971,6 +976,9 @@ vlan_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
 	ifp->if_transmit = vlan_transmit;
 	ifp->if_qflush = vlan_qflush;
 	ifp->if_ioctl = vlan_ioctl;
+#ifdef RATELIMIT
+	ifp->if_snd_tag_alloc = vlan_snd_tag_alloc;
+#endif
 	ifp->if_flags = VLAN_IFFLAGS;
 	ether_ifattach(ifp, eaddr);
 	/* Now undo some of the damage... */
@@ -1591,6 +1599,15 @@ vlan_capabilities(struct ifvlan *ifv)
 		TOEDEV(ifp) = TOEDEV(p);
 		ifp->if_capenable |= p->if_capenable & IFCAP_TOE;
 	}
+
+#ifdef RATELIMIT
+	/*
+	 * If the parent interface supports ratelimiting, so does the
+	 * VLAN interface.
+	 */
+	ifp->if_capabilities |= (p->if_capabilities & IFCAP_TXRTLMT);
+	ifp->if_capenable |= (p->if_capenable & IFCAP_TXRTLMT);
+#endif
 }
 
 static void
@@ -1801,3 +1818,19 @@ vlan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 
 	return (error);
 }
+
+#ifdef RATELIMIT
+static int
+vlan_snd_tag_alloc(struct ifnet *ifp,
+    union if_snd_tag_alloc_params *params,
+    struct m_snd_tag **ppmt)
+{
+
+	/* get trunk device */
+	ifp = vlan_trunkdev(ifp);
+	if (ifp == NULL || (ifp->if_capenable & IFCAP_TXRTLMT) == 0)
+		return (EOPNOTSUPP);
+	/* forward allocation request */
+	return (ifp->if_snd_tag_alloc(ifp, params, ppmt));
+}
+#endif