1 files changed, 215 insertions, 63 deletions

diff --git a/sys/netpfil/ipfw/ip_fw2.c b/sys/netpfil/ipfw/ip_fw2.c
index 1a5b699..c07546b 100644
--- a/sys/netpfil/ipfw/ip_fw2.c
+++ b/sys/netpfil/ipfw/ip_fw2.c
@@ -140,8 +140,7 @@ VNET_DEFINE(int, fw_verbose);
 VNET_DEFINE(u_int64_t, norule_counter);
 VNET_DEFINE(int, verbose_limit);
 
-/* layer3_chain contains the list of rules for layer 3 */
-VNET_DEFINE(struct ip_fw_chain, layer3_chain);
+VNET_DEFINE(struct ip_fw_contextes, ip_fw_contexts);
 
 VNET_DEFINE(int, ipfw_nat_ready) = 0;
 
@@ -182,9 +181,6 @@ SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
     "Make the default rule accept all packets.");
 TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept);
 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
-SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count,
-    CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
-    "Number of static rules");
 
 #ifdef INET6
 SYSCTL_DECL(_net_inet6_ip6);
@@ -358,8 +354,8 @@ iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, uin
 	/* Check by name or by IP address */
 	if (cmd->name[0] != '\0') { /* match by name */
 		if (cmd->name[0] == '\1') /* use tablearg to match */
-			return ipfw_lookup_table_extended(chain, cmd->p.glob,
-				ifp->if_xname, tablearg, IPFW_TABLE_INTERFACE);
+			return (ipfw_lookup_table_extended(chain, cmd->p.glob,
+				ifp->if_xname, tablearg, IPFW_TABLE_INTERFACE, NULL) != NULL);
 		/* Check name */
 		if (cmd->p.glob) {
 			if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
@@ -904,6 +900,9 @@ ipfw_chk(struct ip_fw_args *args)
 	 */
 	struct ifnet *oif = args->oif;
 
+	if (V_ip_fw_contexts.chain[oif->if_ispare[0]] == NULL)
+		return (IP_FW_PASS);
+
 	int f_pos = 0;		/* index of current rule in the array */
 	int retval = 0;
 
@@ -954,7 +953,14 @@ ipfw_chk(struct ip_fw_args *args)
 	 */
 	int dyn_dir = MATCH_UNKNOWN;
 	ipfw_dyn_rule *q = NULL;
-	struct ip_fw_chain *chain = &V_layer3_chain;
+	void *tblent = NULL, *tblent2 = NULL;
+
+	/* XXX: WARNING - The chain is accessed unlocked here.
+	 * There is a potential race here with context handling.
+	 * The chain pointer will get destroyed and a NULL
+	 * pointer dereference can happen!
+	 */
+	struct ip_fw_chain *chain = V_ip_fw_contexts.chain[oif->if_ispare[0]];
 
 	/*
 	 * We store in ulp a pointer to the upper layer protocol header.
@@ -1288,6 +1294,8 @@ do {								\
 			continue;
 
 		skip_or = 0;
+		tblent = NULL;
+		tblent2 = NULL;
 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
 		    l -= cmdlen, cmd += cmdlen) {
 			int match;
@@ -1402,7 +1410,7 @@ do {								\
 				break;
 
 			case O_IN:	/* "out" is "not in" */
-				match = (oif == NULL);
+				match = (args->dir == DIR_IN);
 				break;
 
 			case O_LAYER2:
@@ -1438,11 +1446,18 @@ do {								\
 			case O_IP_SRC_LOOKUP:
 			case O_IP_DST_LOOKUP:
 				if (is_ipv4) {
+					struct ether_addr *ea = NULL;
+
 				    uint32_t key =
 					(cmd->opcode == O_IP_DST_LOOKUP) ?
 					    dst_ip.s_addr : src_ip.s_addr;
 				    uint32_t v = 0;
 
+					if (args->eh) {
+						ea = (struct ether_addr*)((cmd->opcode == O_IP_DST_LOOKUP) ?
+							args->eh->ether_dhost :
+							args->eh->ether_shost);
+					}
 				    if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
 					/* generic lookup. The key must be
 					 * in 32bit big-endian format.
@@ -1484,22 +1499,37 @@ do {								\
 					} else
 					    break;
 				    }
-				    match = ipfw_lookup_table(chain,
-					cmd->arg1, key, &v);
-				    if (!match)
+				    tblent2 = ipfw_lookup_table(chain,
+					cmd->arg1, key, &v, ea);
+				    if (tblent2 == NULL) {
+					match = 0;
 					break;
+				    } else
+					match = 1;
 				    if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
 					match =
 					    ((ipfw_insn_u32 *)cmd)->d[0] == v;
-				    else
+				    if (match)
 					tablearg = v;
 				} else if (is_ipv6) {
+					struct ether_addr *ea = NULL;
 					uint32_t v = 0;
+
+					if (args->eh) {
+						ea = (struct ether_addr*)((cmd->opcode == O_IP_DST_LOOKUP) ?
+							args->eh->ether_dhost :
+							args->eh->ether_shost);
+					}
 					void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
 						&args->f_id.dst_ip6: &args->f_id.src_ip6;
-					match = ipfw_lookup_table_extended(chain,
+					tblent = ipfw_lookup_table_extended(chain,
 							cmd->arg1, pkey, &v,
-							IPFW_TABLE_CIDR);
+							IPFW_TABLE_CIDR, ea);
+				    if (tblent == NULL) {
+					match = 0;
+					break;
+				    } else
+					match = 1;
 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
 						match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
 					if (match)
@@ -2314,8 +2344,7 @@ do {								\
 				break;
 
 			case O_FORWARD_IP:
-				if (args->eh)	/* not valid on layer2 pkts */
-					break;
+				if (!args->eh)	{/* not valid on layer2 pkts */
 				if (q == NULL || q->rule != f ||
 				    dyn_dir == MATCH_FORWARD) {
 				    struct sockaddr_in *sa;
@@ -2330,6 +2359,48 @@ do {								\
 					args->next_hop = sa;
 				    }
 				}
+				} else if (args->eh) {
+					struct m_tag *fwd_tag;
+					struct sockaddr_in *sa;
+					u_short sum;
+
+					/*
+					* Checksum correct? (from ip_fastfwd.c)
+					*/
+					if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED)
+						sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
+					else {
+						if (hlen == sizeof(struct ip))
+							sum = in_cksum_hdr(ip);
+						else
+							sum = in_cksum(m, hlen);
+					}
+					if (sum) {
+						IPSTAT_INC(ips_badsum);
+						retval = IP_FW_DENY;
+						break;
+					}
+
+					/*
+					* Remember that we have checked the IP header and found it valid.
+					*/
+					m->m_pkthdr.csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
+
+					sa = &(((ipfw_insn_sa *)cmd)->sa);
+					fwd_tag = m_tag_get(PACKET_TAG_IPFORWARD,
+						sizeof(struct sockaddr_in), M_NOWAIT);
+					if (fwd_tag == NULL)
+						retval = IP_FW_DENY;
+					else {
+						bcopy(sa, (fwd_tag+1), sizeof(struct sockaddr_in));
+						m_tag_prepend(m, fwd_tag);
+
+						if (in_localip(sa->sin_addr))
+							m->m_flags |= M_FASTFWD_OURS;
+						m->m_flags |= M_IP_NEXTHOP;
+					}
+				}
+
 				retval = IP_FW_PASS;
 				l = 0;          /* exit inner loop */
 				done = 1;       /* exit outer loop */
@@ -2337,8 +2408,7 @@ do {								\
 
 #ifdef INET6
 			case O_FORWARD_IP6:
-				if (args->eh)	/* not valid on layer2 pkts */
-					break;
+				if (!args->eh) {/* not valid on layer2 pkts */
 				if (q == NULL || q->rule != f ||
 				    dyn_dir == MATCH_FORWARD) {
 					struct sockaddr_in6 *sin6;
@@ -2346,6 +2416,24 @@ do {								\
 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
 					args->next_hop6 = sin6;
 				}
+				} else if (args->eh) {
+					struct m_tag *fwd_tag;
+					struct sockaddr_in6 *sin6;
+
+					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
+					fwd_tag = m_tag_get(PACKET_TAG_IPFORWARD,
+						sizeof(struct sockaddr_in6), M_NOWAIT);
+					if (fwd_tag == NULL)
+						retval = IP_FW_DENY;
+					else {
+						bcopy(sin6, (fwd_tag+1), sizeof(struct sockaddr_in6));
+						m_tag_prepend(m, fwd_tag);
+
+						if (in6_localip(&sin6->sin6_addr))
+							m->m_flags |= M_FASTFWD_OURS;
+						m->m_flags |= M_IP6_NEXTHOP;
+					}
+				}
 				retval = IP_FW_PASS;
 				l = 0;		/* exit inner loop */
 				done = 1;	/* exit outer loop */
@@ -2417,7 +2505,7 @@ do {								\
 				set_match(args, f_pos, chain);
 				/* Check if this is 'global' nat rule */
 				if (cmd->arg1 == 0) {
-					retval = ipfw_nat_ptr(args, NULL, m);
+					retval = ipfw_nat_ptr(args, NULL, m, chain);
 					break;
 				}
 				t = ((ipfw_insn_nat *)cmd)->nat;
@@ -2432,7 +2520,7 @@ do {								\
 					if (cmd->arg1 != IP_FW_TABLEARG)
 					    ((ipfw_insn_nat *)cmd)->nat = t;
 				}
-				retval = ipfw_nat_ptr(args, t, m);
+				retval = ipfw_nat_ptr(args, t, m, chain);
 				break;
 
 			case O_REASS: {
@@ -2502,6 +2590,10 @@ do {								\
 		struct ip_fw *rule = chain->map[f_pos];
 		/* Update statistics */
 		IPFW_INC_RULE_COUNTER(rule, pktlen);
+		if (tblent != NULL)
+			ipfw_count_table_xentry_stats(tblent, pktlen);
+		if (tblent2 != NULL)
+			ipfw_count_table_entry_stats(tblent2, pktlen);
 	} else {
 		retval = IP_FW_DENY;
 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
@@ -2536,7 +2628,9 @@ sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
 	if ((error != 0) || (req->newptr == NULL))
 		return (error);
 
-	return (ipfw_resize_tables(&V_layer3_chain, ntables));
+	for (int i = 1; i < IP_FW_MAXCTX; i++)
+		error += ipfw_resize_tables(V_ip_fw_contexts.chain[i], ntables);
+	return (error);
 }
 #endif
 /*
@@ -2614,11 +2708,6 @@ ipfw_destroy(void)
 static int
 vnet_ipfw_init(const void *unused)
 {
-	int error;
-	struct ip_fw *rule = NULL;
-	struct ip_fw_chain *chain;
-
-	chain = &V_layer3_chain;
 
 	/* First set up some values that are compile time options */
 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
@@ -2629,10 +2718,55 @@ vnet_ipfw_init(const void *unused)
 #ifdef IPFIREWALL_VERBOSE_LIMIT
 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
 #endif
+
+	for (int i = 0; i < IP_FW_MAXCTX; i++)
+		V_ip_fw_contexts.chain[i] = NULL;
+
+	IPFW_CTX_LOCK_INIT();
+
+	V_ip_fw_contexts.ifnet_arrival = EVENTHANDLER_REGISTER(ifnet_arrival_event,
+		ipfw_attach_ifnet_event, NULL, EVENTHANDLER_PRI_ANY);
+
+	ipfw_dyn_init();
+
+	/* First set up some values that are compile time options */
+	V_ipfw_vnet_ready = 1;		/* Open for business */
+
+	/*
+	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
+	 * Even if the latter two fail we still keep the module alive
+	 * because the sockopt and layer2 paths are still useful.
+	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
+	 * so we can ignore the exact return value and just set a flag.
+	 *
+	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
+	 * changes in the underlying (per-vnet) variables trigger
+	 * immediate hook()/unhook() calls.
+	 * In layer2 we have the same behaviour, except that V_ether_ipfw
+	 * is checked on each packet because there are no pfil hooks.
+	 */
+	V_ip_fw_ctl_ptr = ipfw_ctl;
+	return ipfw_attach_hooks(1);
+}
+
+int
+ipfw_context_init(int index)
+{
+	struct ip_fw_chain *chain;
+	struct ip_fw *rule = NULL;
+
+	if (index >= IP_FW_MAXCTX)
+		return (-1);
+
+	TAILQ_INIT(&V_ip_fw_contexts.iflist[index]);
+
+	chain = V_ip_fw_contexts.chain[index];
+
+	IPFW_LOCK_INIT(chain);
+
 #ifdef IPFIREWALL_NAT
 	LIST_INIT(&chain->nat);
 #endif
-
 	/* insert the default rule and create the initial map */
 	chain->n_rules = 1;
 	chain->static_len = sizeof(struct ip_fw);
@@ -2642,13 +2776,7 @@ vnet_ipfw_init(const void *unused)
 
 	/* Set initial number of tables */
 	V_fw_tables_max = default_fw_tables;
-	error = ipfw_init_tables(chain);
-	if (error) {
-		printf("ipfw2: setting up tables failed\n");
-		free(chain->map, M_IPFW);
-		free(rule, M_IPFW);
-		return (ENOSPC);
-	}
+	ipfw_init_tables(chain);
 
 	/* fill and insert the default rule */
 	rule->act_ofs = 0;
@@ -2660,28 +2788,13 @@ vnet_ipfw_init(const void *unused)
 	chain->default_rule = chain->map[0] = rule;
 	chain->id = rule->id = 1;
 
-	IPFW_LOCK_INIT(chain);
-	ipfw_dyn_init(chain);
-
-	/* First set up some values that are compile time options */
-	V_ipfw_vnet_ready = 1;		/* Open for business */
+        /*
+         * This can potentially be done on first dynamic rule
+         * being added to chain.
+         */
+        resize_dynamic_table(chain, V_curr_dyn_buckets);
 
-	/*
-	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
-	 * Even if the latter two fail we still keep the module alive
-	 * because the sockopt and layer2 paths are still useful.
-	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
-	 * so we can ignore the exact return value and just set a flag.
-	 *
-	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
-	 * changes in the underlying (per-vnet) variables trigger
-	 * immediate hook()/unhook() calls.
-	 * In layer2 we have the same behaviour, except that V_ether_ipfw
-	 * is checked on each packet because there are no pfil hooks.
-	 */
-	V_ip_fw_ctl_ptr = ipfw_ctl;
-	error = ipfw_attach_hooks(1);
-	return (error);
+	return (0);
 }
 
 /*
@@ -2690,11 +2803,9 @@ vnet_ipfw_init(const void *unused)
 static int
 vnet_ipfw_uninit(const void *unused)
 {
-	struct ip_fw *reap, *rule;
-	struct ip_fw_chain *chain = &V_layer3_chain;
-	int i;
 
 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
+
 	/*
 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
 	 * Then grab, release and grab again the WLOCK so we make
@@ -2702,12 +2813,51 @@ vnet_ipfw_uninit(const void *unused)
 	 */
 	(void)ipfw_attach_hooks(0 /* detach */);
 	V_ip_fw_ctl_ptr = NULL;
-	IPFW_UH_WLOCK(chain);
-	IPFW_UH_WUNLOCK(chain);
 
 	ipfw_dyn_uninit(0);	/* run the callout_drain */
 
+	IPFW_CTX_WLOCK();
+	EVENTHANDLER_DEREGISTER(ifnet_arrival_event, V_ip_fw_contexts.ifnet_arrival);
+	for (int i = 0; i < IP_FW_MAXCTX; i++) {
+		ipfw_context_uninit(i);
+	}
+	IPFW_CTX_WUNLOCK();
+	IPFW_CTX_LOCK_DESTROY();
+
+	ipfw_dyn_uninit(1);	/* free the remaining parts */
+
+	return (0);
+}
+
+int
+ipfw_context_uninit(int index)
+{
+	struct ip_fw_chain *chain;
+	struct ip_fw_ctx_iflist *ifl;
+	struct ip_fw *reap, *rule;
+	struct ifnet *ifp;
+	int i;
+
+	if (index >= IP_FW_MAXCTX)
+		return (-1);
+
+	chain = V_ip_fw_contexts.chain[index];
+	if (chain == NULL)
+		return (0);
+
+	while (!TAILQ_EMPTY(&V_ip_fw_contexts.iflist[index])) {
+		ifl = TAILQ_FIRST(&V_ip_fw_contexts.iflist[index]);
+		TAILQ_REMOVE(&V_ip_fw_contexts.iflist[index], ifl, entry);
+		ifp = ifunit(ifl->ifname);
+		if (ifp != NULL)
+			ifp->if_ispare[0] = 0;
+		free(ifl, M_IPFW);
+	}
+
 	IPFW_UH_WLOCK(chain);
+	IPFW_UH_WUNLOCK(chain);
+	IPFW_UH_WLOCK(chain);
+
 	ipfw_destroy_tables(chain);
 	reap = NULL;
 	IPFW_WLOCK(chain);
@@ -2723,8 +2873,10 @@ vnet_ipfw_uninit(const void *unused)
 	if (reap != NULL)
 		ipfw_reap_rules(reap);
 	IPFW_LOCK_DESTROY(chain);
-	ipfw_dyn_uninit(1);	/* free the remaining parts */
-	return 0;
+
+	free(chain, M_IPFW);
+
+	return (0);
 }
 
 /*