Implement interface link header precomputation API.

Add if_requestencap() interface method which is capable of calculating
  various link headers for given interface. Right now there is support
  for INET/INET6/ARP llheader calculation (IFENCAP_LL type request).
  Other types are planned to support more complex calculation
  (L2 multipath lagg nexthops, tunnel encap nexthops, etc..).

Reshape 'struct route' to be able to pass additional data (with is length)
  to prepend to mbuf.

These two changes permits routing code to pass pre-calculated nexthop data
  (like L2 header for route w/gateway) down to the stack eliminating the
  need for other lookups. It also brings us closer to more complex scenarios
  like transparently handling MPLS nexthops and tunnel interfaces.
  Last, but not least, it removes layering violation introduced by flowtable
  code (ro_lle) and simplifies handling of existing if_output consumers.

ARP/ND changes:
Make arp/ndp stack pre-calculate link header upon installing/updating lle
  record. Interface link address change are handled by re-calculating
  headers for all lles based on if_lladdr event. After these changes,
  arpresolve()/nd6_resolve() returns full pre-calculated header for
  supported interfaces thus simplifying if_output().
Move these lookups to separate ether_resolve_addr() function which ether
  returs error or fully-prepared link header. Add <arp|nd6_>resolve_addr()
  compat versions to return link addresses instead of pre-calculated data.

BPF changes:
Raw bpf writes occupied _two_ cases: AF_UNSPEC and pseudo_AF_HDRCMPLT.
Despite the naming, both of there have ther header "complete". The only
  difference is that interface source mac has to be filled by OS for
  AF_UNSPEC (controlled via BIOCGHDRCMPLT). This logic has to stay inside
  BPF and not pollute if_output() routines. Convert BPF to pass prepend data
  via new 'struct route' mechanism. Note that it does not change
  non-optimized if_output(): ro_prepend handling is purely optional.
Side note: hackish pseudo_AF_HDRCMPLT is supported for ethernet and FDDI.
  It is not needed for ethernet anymore. The only remaining FDDI user is
  dev/pdq mostly untouched since 2007. FDDI support was eliminated from
  OpenBSD in 2013 (sys/net/if_fddisubr.c rev 1.65).

Flowtable changes:
  Flowtable violates layering by saving (and not correctly managing)
  rtes/lles. Instead of passing lle pointer, pass pointer to pre-calculated
  header data from that lle.

Differential Revision:	https://reviews.freebsd.org/D4102

6 files changed, 108 insertions, 26 deletions

diff --git a/sys/netinet6/icmp6.c b/sys/netinet6/icmp6.c
index f0c5371..a7bbac0 100644
--- a/sys/netinet6/icmp6.c
+++ b/sys/netinet6/icmp6.c
@@ -2632,7 +2632,7 @@ icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
 			nd_opt->nd_opt_len = len >> 3;
 			lladdr = (char *)(nd_opt + 1);
-			bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
+			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
 			p += len;
 		}
 	}
diff --git a/sys/netinet6/in6.c b/sys/netinet6/in6.c
index aa77019..0d47d4b 100644
--- a/sys/netinet6/in6.c
+++ b/sys/netinet6/in6.c
@@ -2245,6 +2245,9 @@ in6_lltable_alloc(struct lltable *llt, u_int flags,
 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
 	struct ifnet *ifp = llt->llt_ifp;
 	struct llentry *lle;
+	char linkhdr[LLE_MAX_LINKHDR];
+	size_t linkhdrsize;
+	int lladdr_off;
 
 	KASSERT(l3addr->sa_family == AF_INET6,
 	    ("sin_family %d", l3addr->sa_family));
@@ -2265,7 +2268,12 @@ in6_lltable_alloc(struct lltable *llt, u_int flags,
 	}
 	lle->la_flags = flags;
 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
-		lltable_set_entry_addr(ifp, lle, IF_LLADDR(ifp));
+		linkhdrsize = LLE_MAX_LINKHDR;
+		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
+		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
+			return (NULL);
+		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
+		    lladdr_off);
 		lle->la_flags |= LLE_STATIC;
 	}
 
diff --git a/sys/netinet6/in6.h b/sys/netinet6/in6.h
index 8f08c8f..9bc142a 100644
--- a/sys/netinet6/in6.h
+++ b/sys/netinet6/in6.h
@@ -375,9 +375,9 @@ extern const struct in6_addr in6addr_linklocal_allv2routers;
 #if __BSD_VISIBLE
 struct route_in6 {
 	struct	rtentry *ro_rt;
-	struct	llentry *ro_lle;
-	struct	in6_addr *ro_ia6;
-	int		ro_flags;
+	char		*ro_prepend;
+	uint16_t	ro_plen;
+	uint16_t	ro_flags;
 	struct	sockaddr_in6 ro_dst;
 };
 #endif
diff --git a/sys/netinet6/nd6.c b/sys/netinet6/nd6.c
index 7ca172c..fbbf421 100644
--- a/sys/netinet6/nd6.c
+++ b/sys/netinet6/nd6.c
@@ -111,7 +111,7 @@ VNET_DEFINE(int, nd6_debug) = 1;
 VNET_DEFINE(int, nd6_debug) = 0;
 #endif
 
-static eventhandler_tag lle_event_eh;
+static eventhandler_tag lle_event_eh, iflladdr_event_eh;
 
 /* for debugging? */
 #if 0
@@ -137,7 +137,7 @@ static void nd6_llinfo_timer(void *);
 static void nd6_llinfo_settimer_locked(struct llentry *, long);
 static void clear_llinfo_pqueue(struct llentry *);
 static void nd6_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
-static int nd6_resolve_slow(struct ifnet *, struct mbuf *,
+static int nd6_resolve_slow(struct ifnet *, int, struct mbuf *,
     const struct sockaddr_in6 *, u_char *, uint32_t *);
 static int nd6_need_cache(struct ifnet *);
  
@@ -188,7 +188,7 @@ nd6_lle_event(void *arg __unused, struct llentry *lle, int evt)
 	gw.sdl_index = ifp->if_index;
 	gw.sdl_type = ifp->if_type;
 	if (evt == LLENTRY_RESOLVED)
-		bcopy(&lle->ll_addr, gw.sdl_data, ifp->if_addrlen);
+		bcopy(lle->ll_addr, gw.sdl_data, ifp->if_addrlen);
 	rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst;
 	rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw;
 	rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY;
@@ -196,6 +196,16 @@ nd6_lle_event(void *arg __unused, struct llentry *lle, int evt)
 	    type == RTM_ADD ? RTF_UP: 0), 0, RT_DEFAULT_FIB);
 }
 
+/*
+ * A handler for interface link layer address change event.
+ */
+static void
+nd6_iflladdr(void *arg __unused, struct ifnet *ifp)
+{
+
+	lltable_update_ifaddr(LLTABLE6(ifp));
+}
+
 void
 nd6_init(void)
 {
@@ -211,9 +221,12 @@ nd6_init(void)
 	    nd6_slowtimo, curvnet);
 
 	nd6_dad_init();
-	if (IS_DEFAULT_VNET(curvnet))
+	if (IS_DEFAULT_VNET(curvnet)) {
 		lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event,
 		    NULL, EVENTHANDLER_PRI_ANY);
+		iflladdr_event_eh = EVENTHANDLER_REGISTER(iflladdr_event,
+		    nd6_iflladdr, NULL, EVENTHANDLER_PRI_ANY);
+	}
 }
 
 #ifdef VIMAGE
@@ -223,8 +236,10 @@ nd6_destroy()
 
 	callout_drain(&V_nd6_slowtimo_ch);
 	callout_drain(&V_nd6_timer_ch);
-	if (IS_DEFAULT_VNET(curvnet))
+	if (IS_DEFAULT_VNET(curvnet)) {
 		EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh);
+		EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_event_eh);
+	}
 }
 #endif
 
@@ -1844,6 +1859,9 @@ nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
 	uint16_t router = 0;
 	struct sockaddr_in6 sin6;
 	struct mbuf *chain = NULL;
+	u_char linkhdr[LLE_MAX_LINKHDR];
+	size_t linkhdrsize;
+	int lladdr_off;
 
 	IF_AFDATA_UNLOCK_ASSERT(ifp);
 
@@ -1878,8 +1896,15 @@ nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
 		 * Since we already know all the data for the new entry,
 		 * fill it before insertion.
 		 */
-		if (lladdr != NULL)
-			lltable_set_entry_addr(ifp, ln, lladdr);
+		if (lladdr != NULL) {
+			linkhdrsize = sizeof(linkhdr);
+			if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
+			    linkhdr, &linkhdrsize, &lladdr_off) != 0)
+				return;
+			lltable_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
+			    lladdr_off);
+		}
+
 		IF_AFDATA_WLOCK(ifp);
 		LLE_WLOCK(ln);
 		/* Prefer any existing lle over newly-created one */
@@ -1911,7 +1936,7 @@ nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
 
 	olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
 	if (olladdr && lladdr) {
-		llchange = bcmp(lladdr, &ln->ll_addr,
+		llchange = bcmp(lladdr, ln->ll_addr,
 		    ifp->if_addrlen);
 	} else if (!olladdr && lladdr)
 		llchange = 1;
@@ -1937,7 +1962,13 @@ nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
 		 * Record source link-layer address
 		 * XXX is it dependent to ifp->if_type?
 		 */
-		if (lltable_try_set_entry_addr(ifp, ln, lladdr) == 0) {
+		linkhdrsize = sizeof(linkhdr);
+		if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
+		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
+			return;
+
+		if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
+		    lladdr_off) == 0) {
 			/* Entry was deleted */
 			return;
 		}
@@ -2093,8 +2124,8 @@ nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
 }
 
 /*
- * Do L2 address resolution for @sa_dst address. Stores found
- * address in @desten buffer. Copy of lle ln_flags can be also
+ * Lookup link headerfor @sa_dst address. Stores found
+ * data in @desten buffer. Copy of lle ln_flags can be also
  * saved in @pflags if @pflags is non-NULL.
  *
  * If destination LLE does not exists or lle state modification
@@ -2144,7 +2175,7 @@ nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
 	ln = nd6_lookup(&dst6->sin6_addr, LLE_UNLOCKED, ifp);
 	if (ln != NULL && (ln->r_flags & RLLE_VALID) != 0) {
 		/* Entry found, let's copy lle info */
-		bcopy(&ln->ll_addr, desten, ifp->if_addrlen);
+		bcopy(ln->r_linkdata, desten, ln->r_hdrlen);
 		if (pflags != NULL)
 			*pflags = LLE_VALID | (ln->r_flags & RLLE_IFADDR);
 		/* Check if we have feedback request from nd6 timer */
@@ -2159,7 +2190,7 @@ nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
 	}
 	IF_AFDATA_RUNLOCK(ifp);
 
-	return (nd6_resolve_slow(ifp, m, dst6, desten, pflags));
+	return (nd6_resolve_slow(ifp, 0, m, dst6, desten, pflags));
 }
 
 
@@ -2175,12 +2206,13 @@ nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
  * Set noinline to be dtrace-friendly
  */
 static __noinline int
-nd6_resolve_slow(struct ifnet *ifp, struct mbuf *m,
+nd6_resolve_slow(struct ifnet *ifp, int flags, struct mbuf *m,
     const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags)
 {
 	struct llentry *lle = NULL, *lle_tmp;
 	struct in6_addr *psrc, src;
-	int send_ns;
+	int send_ns, ll_len;
+	char *lladdr;
 
 	/*
 	 * Address resolution or Neighbor Unreachability Detection
@@ -2252,7 +2284,14 @@ nd6_resolve_slow(struct ifnet *ifp, struct mbuf *m,
 	 * send the packet.
 	 */
 	if (lle->ln_state > ND6_LLINFO_INCOMPLETE) {
-		bcopy(&lle->ll_addr, desten, ifp->if_addrlen);
+		if (flags & LLE_ADDRONLY) {
+			lladdr = lle->ll_addr;
+			ll_len = ifp->if_addrlen;
+		} else {
+			lladdr = lle->r_linkdata;
+			ll_len = lle->r_hdrlen;
+		}
+		bcopy(lladdr, desten, ll_len);
 		if (pflags != NULL)
 			*pflags = lle->la_flags;
 		LLE_WUNLOCK(lle);
@@ -2312,6 +2351,27 @@ nd6_resolve_slow(struct ifnet *ifp, struct mbuf *m,
 	return (EWOULDBLOCK);
 }
 
+/*
+ * Do L2 address resolution for @sa_dst address. Stores found
+ * address in @desten buffer. Copy of lle ln_flags can be also
+ * saved in @pflags if @pflags is non-NULL.
+ *
+ * Return values:
+ * - 0 on success (address copied to buffer).
+ * - EWOULDBLOCK (no local error, but address is still unresolved)
+ * - other errors (alloc failure, etc)
+ */
+int
+nd6_resolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst,
+    char *desten, uint32_t *pflags)
+{
+	int error;
+
+	flags |= LLE_ADDRONLY;
+	error = nd6_resolve_slow(ifp, flags, NULL,
+	    (const struct sockaddr_in6 *)dst, desten, pflags);
+	return (error);
+}
 
 int
 nd6_flush_holdchain(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain,
diff --git a/sys/netinet6/nd6.h b/sys/netinet6/nd6.h
index 8a0a56e..6ff4cac 100644
--- a/sys/netinet6/nd6.h
+++ b/sys/netinet6/nd6.h
@@ -410,6 +410,8 @@ void nd6_setmtu(struct ifnet *);
 void nd6_llinfo_setstate(struct llentry *lle, int newstate);
 void nd6_timer(void *);
 void nd6_purge(struct ifnet *);
+int nd6_resolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst,
+    char *desten, uint32_t *pflags);
 int nd6_resolve(struct ifnet *, int, struct mbuf *,
     const struct sockaddr *, u_char *, uint32_t *);
 int nd6_ioctl(u_long, caddr_t, struct ifnet *);
diff --git a/sys/netinet6/nd6_nbr.c b/sys/netinet6/nd6_nbr.c
index bf43fb6..a5ce2ec 100644
--- a/sys/netinet6/nd6_nbr.c
+++ b/sys/netinet6/nd6_nbr.c
@@ -643,6 +643,9 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len)
 	union nd_opts ndopts;
 	struct mbuf *chain = NULL;
 	struct sockaddr_in6 sin6;
+	u_char linkhdr[LLE_MAX_LINKHDR];
+	size_t linkhdrsize;
+	int lladdr_off;
 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
 
 	if (ip6->ip6_hlim != 255) {
@@ -765,7 +768,13 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len)
 		/*
 		 * Record link-layer address, and update the state.
 		 */
-		if (lltable_try_set_entry_addr(ifp, ln, lladdr) == 0) {
+		linkhdrsize = sizeof(linkhdr);
+		if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
+		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
+			return;
+
+		if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
+		    lladdr_off) == 0) {
 			ln = NULL;
 			goto freeit;
 		}
@@ -792,7 +801,7 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len)
 			llchange = 0;
 		else {
 			if (ln->la_flags & LLE_VALID) {
-				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
+				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
 					llchange = 1;
 				else
 					llchange = 0;
@@ -834,9 +843,12 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len)
 			 * Update link-local address, if any.
 			 */
 			if (lladdr != NULL) {
-				int ret;
-				ret = lltable_try_set_entry_addr(ifp, ln,lladdr);
-				if (ret == 0) {
+				linkhdrsize = sizeof(linkhdr);
+				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
+				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
+					goto freeit;
+				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
+				    linkhdrsize, lladdr_off) == 0) {
 					ln = NULL;
 					goto freeit;
 				}