2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1990, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
38 #include "opt_ipsec.h"
39 #include "opt_kern_tls.h"
40 #include "opt_mbuf_stress_test.h"
41 #include "opt_ratelimit.h"
42 #include "opt_route.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
51 #include <sys/malloc.h>
55 #include <sys/protosw.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/sysctl.h>
60 #include <sys/ucred.h>
63 #include <net/if_var.h>
64 #include <net/if_private.h>
65 #include <net/if_vlan_var.h>
66 #include <net/if_llatbl.h>
67 #include <net/ethernet.h>
68 #include <net/netisr.h>
70 #include <net/route.h>
71 #include <net/route/nhop.h>
72 #include <net/rss_config.h>
75 #include <netinet/in.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_kdtrace.h>
78 #include <netinet/in_systm.h>
79 #include <netinet/ip.h>
80 #include <netinet/in_fib.h>
81 #include <netinet/in_pcb.h>
82 #include <netinet/in_rss.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip_var.h>
85 #include <netinet/ip_options.h>
87 #include <netinet/udp.h>
88 #include <netinet/udp_var.h>
90 #if defined(SCTP) || defined(SCTP_SUPPORT)
91 #include <netinet/sctp.h>
92 #include <netinet/sctp_crc32.h>
95 #include <netipsec/ipsec_support.h>
97 #include <machine/in_cksum.h>
99 #include <security/mac/mac_framework.h>
101 #ifdef MBUF_STRESS_TEST
102 static int mbuf_frag_size = 0;
103 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
104 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
107 static void ip_mloopback(struct ifnet *, const struct mbuf *, int);
109 extern int in_mcast_loop;
112 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
113 struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
115 struct m_tag *fwd_tag = NULL;
119 int pflags = PFIL_OUT;
122 ip = mtod(m, struct ip *);
124 /* Run through list of hooks for output packets. */
125 odst.s_addr = ip->ip_dst.s_addr;
126 switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
131 return 1; /* Finished */
136 ip = mtod(m, struct ip *);
138 /* See if destination IP address was changed by packet filter. */
139 if (odst.s_addr != ip->ip_dst.s_addr) {
140 m->m_flags |= M_SKIP_FIREWALL;
141 /* If destination is now ourself drop to ip_input(). */
142 if (in_localip(ip->ip_dst)) {
143 m->m_flags |= M_FASTFWD_OURS;
144 if (m->m_pkthdr.rcvif == NULL)
145 m->m_pkthdr.rcvif = V_loif;
146 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
147 m->m_pkthdr.csum_flags |=
148 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
149 m->m_pkthdr.csum_data = 0xffff;
151 m->m_pkthdr.csum_flags |=
152 CSUM_IP_CHECKED | CSUM_IP_VALID;
153 #if defined(SCTP) || defined(SCTP_SUPPORT)
154 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
155 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
157 *error = netisr_queue(NETISR_IP, m);
158 return 1; /* Finished */
161 bzero(dst, sizeof(*dst));
162 dst->sin_family = AF_INET;
163 dst->sin_len = sizeof(*dst);
164 dst->sin_addr = ip->ip_dst;
166 return -1; /* Reloop */
168 /* See if fib was changed by packet filter. */
169 if ((*fibnum) != M_GETFIB(m)) {
170 m->m_flags |= M_SKIP_FIREWALL;
171 *fibnum = M_GETFIB(m);
172 return -1; /* Reloop for FIB change */
175 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
176 if (m->m_flags & M_FASTFWD_OURS) {
177 if (m->m_pkthdr.rcvif == NULL)
178 m->m_pkthdr.rcvif = V_loif;
179 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
180 m->m_pkthdr.csum_flags |=
181 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
182 m->m_pkthdr.csum_data = 0xffff;
184 #if defined(SCTP) || defined(SCTP_SUPPORT)
185 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
186 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
188 m->m_pkthdr.csum_flags |=
189 CSUM_IP_CHECKED | CSUM_IP_VALID;
191 *error = netisr_queue(NETISR_IP, m);
192 return 1; /* Finished */
194 /* Or forward to some other address? */
195 if ((m->m_flags & M_IP_NEXTHOP) &&
196 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
197 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
198 m->m_flags |= M_SKIP_FIREWALL;
199 m->m_flags &= ~M_IP_NEXTHOP;
200 m_tag_delete(m, fwd_tag);
202 return -1; /* Reloop for CHANGE of dst */
209 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
210 const struct sockaddr *gw, struct route *ro, bool stamp_tag)
213 struct ktls_session *tls = NULL;
215 struct m_snd_tag *mst;
218 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
223 * If this is an unencrypted TLS record, save a reference to
224 * the record. This local reference is used to call
225 * ktls_output_eagain after the mbuf has been freed (thus
226 * dropping the mbuf's reference) in if_output.
228 if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
229 tls = ktls_hold(m->m_next->m_epg_tls);
233 * If a TLS session doesn't have a valid tag, it must
234 * have had an earlier ifp mismatch, so drop this
243 * Always stamp tags that include NIC ktls.
249 if (inp != NULL && mst == NULL) {
250 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
251 (inp->inp_snd_tag != NULL &&
252 inp->inp_snd_tag->ifp != ifp))
253 in_pcboutput_txrtlmt(inp, ifp, m);
255 if (inp->inp_snd_tag != NULL)
256 mst = inp->inp_snd_tag;
259 if (stamp_tag && mst != NULL) {
260 KASSERT(m->m_pkthdr.rcvif == NULL,
261 ("trying to add a send tag to a forwarded packet"));
262 if (mst->ifp != ifp) {
268 /* stamp send tag on mbuf */
269 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
270 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
273 error = (*ifp->if_output)(ifp, m, gw, ro);
276 /* Check for route change invalidating send tags. */
280 error = ktls_output_eagain(inp, tls);
286 in_pcboutput_eagain(inp);
291 /* rte<>ro_flags translation */
293 rt_update_ro_flags(struct route *ro, const struct nhop_object *nh)
295 int nh_flags = nh->nh_flags;
297 ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
299 ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
300 ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
301 ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
305 * IP output. The packet in mbuf chain m contains a skeletal IP
306 * header (with len, off, ttl, proto, tos, src, dst).
307 * The mbuf chain containing the packet will be freed.
308 * The mbuf opt, if present, will not be freed.
309 * If route ro is present and has ro_rt initialized, route lookup would be
310 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
311 * then result of route lookup is stored in ro->ro_rt.
313 * In the IP forwarding case, the packet will arrive with options already
314 * inserted, so must have a NULL opt pointer.
317 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
318 struct ip_moptions *imo, struct inpcb *inp)
321 struct ifnet *ifp = NULL; /* keep compiler happy */
323 int hlen = sizeof (struct ip);
327 struct sockaddr_in *dst;
328 const struct sockaddr *gw;
329 struct in_ifaddr *ia = NULL;
332 uint16_t ip_len, ip_off;
333 struct route iproute;
335 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
336 int no_route_but_check_spd = 0;
343 INP_LOCK_ASSERT(inp);
344 M_SETFIB(m, inp->inp_inc.inc_fibnum);
345 if ((flags & IP_NODEFAULTFLOWID) == 0) {
346 m->m_pkthdr.flowid = inp->inp_flowid;
347 M_HASHTYPE_SET(m, inp->inp_flowtype);
349 if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
350 vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
353 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
359 m = ip_insertoptions(m, opt, &len);
361 hlen = len; /* ip->ip_hl is updated above */
363 ip = mtod(m, struct ip *);
364 ip_len = ntohs(ip->ip_len);
365 ip_off = ntohs(ip->ip_off);
367 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
368 ip->ip_v = IPVERSION;
369 ip->ip_hl = hlen >> 2;
372 /* Header already set, fetch hlen from there */
373 hlen = ip->ip_hl << 2;
375 if ((flags & IP_FORWARDING) == 0)
376 IPSTAT_INC(ips_localout);
381 * gw is readonly but can point either to dst OR rt_gateway,
382 * therefore we need restore gw if we're redoing lookup.
384 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
387 bzero(ro, sizeof (*ro));
389 dst = (struct sockaddr_in *)&ro->ro_dst;
390 if (ro->ro_nh == NULL) {
391 dst->sin_family = AF_INET;
392 dst->sin_len = sizeof(*dst);
393 dst->sin_addr = ip->ip_dst;
395 gw = (const struct sockaddr *)dst;
398 * Validate route against routing table additions;
399 * a better/more specific route might have been added.
401 if (inp != NULL && ro->ro_nh != NULL)
402 NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
404 * If there is a cached route,
405 * check that it is to the same destination
406 * and is still up. If not, free it and try again.
407 * The address family should also be checked in case of sharing the
409 * Also check whether routing cache needs invalidation.
411 if (ro->ro_nh != NULL &&
412 ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
413 dst->sin_addr.s_addr != ip->ip_dst.s_addr))
414 RO_INVALIDATE_CACHE(ro);
417 * If routing to interface only, short circuit routing lookup.
418 * The use of an all-ones broadcast address implies this; an
419 * interface is specified by the broadcast address of an interface,
420 * or the destination address of a ptp interface.
422 if (flags & IP_SENDONES) {
423 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
424 M_GETFIB(m)))) == NULL &&
425 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
426 M_GETFIB(m)))) == NULL) {
427 IPSTAT_INC(ips_noroute);
431 ip->ip_dst.s_addr = INADDR_BROADCAST;
432 dst->sin_addr = ip->ip_dst;
437 src = IA_SIN(ia)->sin_addr;
438 } else if (flags & IP_ROUTETOIF) {
439 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
440 M_GETFIB(m)))) == NULL &&
441 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
442 M_GETFIB(m)))) == NULL) {
443 IPSTAT_INC(ips_noroute);
450 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
451 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
452 src = IA_SIN(ia)->sin_addr;
453 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
454 imo != NULL && imo->imo_multicast_ifp != NULL) {
456 * Bypass the normal routing lookup for multicast
457 * packets if the interface is specified.
459 ifp = imo->imo_multicast_ifp;
462 isbroadcast = 0; /* fool gcc */
463 /* Interface may have no addresses. */
465 src = IA_SIN(ia)->sin_addr;
467 src.s_addr = INADDR_ANY;
468 } else if (ro != &iproute) {
469 if (ro->ro_nh == NULL) {
471 * We want to do any cloning requested by the link
472 * layer, as this is probably required in all cases
473 * for correct operation (as it is for ARP).
476 flowid = m->m_pkthdr.flowid;
477 ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
480 if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
481 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
483 * There is no route for this packet, but it is
484 * possible that a matching SPD entry exists.
486 no_route_but_check_spd = 1;
489 IPSTAT_INC(ips_noroute);
490 error = EHOSTUNREACH;
494 struct nhop_object *nh = ro->ro_nh;
496 ia = ifatoia(nh->nh_ifa);
498 counter_u64_add(nh->nh_pksent, 1);
499 rt_update_ro_flags(ro, nh);
500 if (nh->nh_flags & NHF_GATEWAY)
502 if (nh->nh_flags & NHF_HOST)
503 isbroadcast = (nh->nh_flags & NHF_BROADCAST);
504 else if ((ifp->if_flags & IFF_BROADCAST) && (gw->sa_family == AF_INET))
505 isbroadcast = in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia);
509 src = IA_SIN(ia)->sin_addr;
511 struct nhop_object *nh;
513 nh = fib4_lookup(M_GETFIB(m), dst->sin_addr, 0, NHR_NONE,
516 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
518 * There is no route for this packet, but it is
519 * possible that a matching SPD entry exists.
521 no_route_but_check_spd = 1;
524 IPSTAT_INC(ips_noroute);
525 error = EHOSTUNREACH;
530 rt_update_ro_flags(ro, nh);
531 if (nh->nh_flags & NHF_GATEWAY)
533 ia = ifatoia(nh->nh_ifa);
534 src = IA_SIN(ia)->sin_addr;
535 isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
536 (NHF_HOST | NHF_BROADCAST)) ||
537 ((ifp->if_flags & IFF_BROADCAST) &&
538 (gw->sa_family == AF_INET) &&
539 in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia)));
542 /* Catch a possible divide by zero later. */
543 KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
545 (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
547 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
548 m->m_flags |= M_MCAST;
550 * IP destination address is multicast. Make sure "gw"
551 * still points to the address in "ro". (It may have been
552 * changed to point to a gateway address, above.)
554 gw = (const struct sockaddr *)dst;
556 * See if the caller provided any multicast options
559 ip->ip_ttl = imo->imo_multicast_ttl;
560 if (imo->imo_multicast_vif != -1)
563 ip_mcast_src(imo->imo_multicast_vif) :
566 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
568 * Confirm that the outgoing interface supports multicast.
570 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
571 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
572 IPSTAT_INC(ips_noroute);
578 * If source address not specified yet, use address
579 * of outgoing interface.
581 if (ip->ip_src.s_addr == INADDR_ANY)
584 if ((imo == NULL && in_mcast_loop) ||
585 (imo && imo->imo_multicast_loop)) {
587 * Loop back multicast datagram if not expressly
588 * forbidden to do so, even if we are not a member
589 * of the group; ip_input() will filter it later,
590 * thus deferring a hash lookup and mutex acquisition
591 * at the expense of a cheap copy using m_copym().
593 ip_mloopback(ifp, m, hlen);
596 * If we are acting as a multicast router, perform
597 * multicast forwarding as if the packet had just
598 * arrived on the interface to which we are about
599 * to send. The multicast forwarding function
600 * recursively calls this function, using the
601 * IP_FORWARDING flag to prevent infinite recursion.
603 * Multicasts that are looped back by ip_mloopback(),
604 * above, will be forwarded by the ip_input() routine,
607 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
609 * If rsvp daemon is not running, do not
610 * set ip_moptions. This ensures that the packet
611 * is multicast and not just sent down one link
612 * as prescribed by rsvpd.
617 ip_mforward(ip, ifp, m, imo) != 0) {
625 * Multicasts with a time-to-live of zero may be looped-
626 * back, above, but must not be transmitted on a network.
627 * Also, multicasts addressed to the loopback interface
628 * are not sent -- the above call to ip_mloopback() will
629 * loop back a copy. ip_input() will drop the copy if
630 * this host does not belong to the destination group on
631 * the loopback interface.
633 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
642 * If the source address is not specified yet, use the address
643 * of the outoing interface.
645 if (ip->ip_src.s_addr == INADDR_ANY)
649 * Look for broadcast address and
650 * verify user is allowed to send
654 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
655 error = EADDRNOTAVAIL;
658 if ((flags & IP_ALLOWBROADCAST) == 0) {
662 /* don't allow broadcast messages to be fragmented */
667 m->m_flags |= M_BCAST;
669 m->m_flags &= ~M_BCAST;
673 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
674 if (IPSEC_ENABLED(ipv4)) {
675 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
676 if (error == EINPROGRESS)
682 * Check if there was a route for this packet; return error if not.
684 if (no_route_but_check_spd) {
685 IPSTAT_INC(ips_noroute);
686 error = EHOSTUNREACH;
689 /* Update variables that are affected by ipsec4_output(). */
690 ip = mtod(m, struct ip *);
691 hlen = ip->ip_hl << 2;
694 /* Jump over all PFIL processing if hooks are not active. */
695 if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
696 switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
698 case 1: /* Finished */
701 case 0: /* Continue normally */
702 ip = mtod(m, struct ip *);
705 case -1: /* Need to try again */
706 /* Reset everything for a new round */
709 ro->ro_prepend = NULL;
711 gw = (const struct sockaddr *)dst;
712 ip = mtod(m, struct ip *);
718 EVL_APPLY_PRI(m, vlan_pcp);
720 /* IN_LOOPBACK must not appear on the wire - RFC1122. */
721 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
722 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
723 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
724 IPSTAT_INC(ips_badaddr);
725 error = EADDRNOTAVAIL;
730 /* Ensure the packet data is mapped if the interface requires it. */
731 if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
732 m = mb_unmapped_to_ext(m);
734 IPSTAT_INC(ips_odropped);
740 m->m_pkthdr.csum_flags |= CSUM_IP;
741 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
743 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
745 #if defined(SCTP) || defined(SCTP_SUPPORT)
746 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
747 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
748 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
753 * If small enough for interface, or the interface will take
754 * care of the fragmentation for us, we can just send directly.
755 * Note that if_vxlan could have requested TSO even though the outer
756 * frame is UDP. It is correct to not fragment such datagrams and
757 * instead just pass them on to the driver.
760 (m->m_pkthdr.csum_flags & ifp->if_hwassist &
761 (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
763 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
764 ip->ip_sum = in_cksum(m, hlen);
765 m->m_pkthdr.csum_flags &= ~CSUM_IP;
769 * Record statistics for this interface address.
770 * With CSUM_TSO the byte/packet count will be slightly
771 * incorrect because we count the IP+TCP headers only
772 * once instead of for every generated packet.
774 if (!(flags & IP_FORWARDING) && ia) {
775 if (m->m_pkthdr.csum_flags &
776 (CSUM_TSO | CSUM_INNER_TSO))
777 counter_u64_add(ia->ia_ifa.ifa_opackets,
778 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
780 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
782 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
784 #ifdef MBUF_STRESS_TEST
785 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
786 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
789 * Reset layer specific mbuf flags
790 * to avoid confusing lower layers.
793 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
794 error = ip_output_send(inp, ifp, m, gw, ro,
795 (flags & IP_NO_SND_TAG_RL) ? false : true);
799 /* Balk when DF bit is set or the interface didn't support TSO. */
800 if ((ip_off & IP_DF) ||
801 (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
803 IPSTAT_INC(ips_cantfrag);
808 * Too large for interface; fragment if possible. If successful,
809 * on return, m will point to a list of packets to be sent.
811 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
818 /* Record statistics for this interface address. */
820 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
821 counter_u64_add(ia->ia_ifa.ifa_obytes,
825 * Reset layer specific mbuf flags
826 * to avoid confusing upper layers.
830 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
831 mtod(m, struct ip *), NULL);
832 error = ip_output_send(inp, ifp, m, gw, ro, true);
838 IPSTAT_INC(ips_fragmented);
848 * Create a chain of fragments which fit the given mtu. m_frag points to the
849 * mbuf to be fragmented; on return it points to the chain with the fragments.
850 * Return 0 if no error. If error, m_frag may contain a partially built
851 * chain of fragments that should be freed by the caller.
853 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
856 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
857 u_long if_hwassist_flags)
860 int hlen = ip->ip_hl << 2;
861 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
863 struct mbuf *m0 = *m_frag; /* the original packet */
867 uint16_t ip_len, ip_off;
869 ip_len = ntohs(ip->ip_len);
870 ip_off = ntohs(ip->ip_off);
873 * Packet shall not have "Don't Fragment" flag and have at least 8
876 if (__predict_false((ip_off & IP_DF) || len < 8)) {
877 IPSTAT_INC(ips_cantfrag);
882 * If the interface will not calculate checksums on
883 * fragmented packets, then do it here.
885 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
886 in_delayed_cksum(m0);
887 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
889 #if defined(SCTP) || defined(SCTP_SUPPORT)
890 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
891 sctp_delayed_cksum(m0, hlen);
892 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
895 if (len > PAGE_SIZE) {
897 * Fragment large datagrams such that each segment
898 * contains a multiple of PAGE_SIZE amount of data,
899 * plus headers. This enables a receiver to perform
900 * page-flipping zero-copy optimizations.
902 * XXX When does this help given that sender and receiver
903 * could have different page sizes, and also mtu could
904 * be less than the receiver's page size ?
908 off = MIN(mtu, m0->m_pkthdr.len);
911 * firstlen (off - hlen) must be aligned on an
915 goto smart_frag_failure;
916 off = ((off - hlen) & ~7) + hlen;
917 newlen = (~PAGE_MASK) & mtu;
918 if ((newlen + sizeof (struct ip)) > mtu) {
919 /* we failed, go back the default */
930 firstlen = off - hlen;
931 mnext = &m0->m_nextpkt; /* pointer to next packet */
934 * Loop through length of segment after first fragment,
935 * make new header and copy data of each part and link onto chain.
936 * Here, m0 is the original packet, m is the fragment being created.
937 * The fragments are linked off the m_nextpkt of the original
938 * packet, which after processing serves as the first fragment.
940 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
941 struct ip *mhip; /* ip header on the fragment */
943 int mhlen = sizeof (struct ip);
945 m = m_gethdr(M_NOWAIT, MT_DATA);
948 IPSTAT_INC(ips_odropped);
952 * Make sure the complete packet header gets copied
953 * from the originating mbuf to the newly created
954 * mbuf. This also ensures that existing firewall
955 * classification(s), VLAN tags and so on get copied
956 * to the resulting fragmented packet(s):
958 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
961 IPSTAT_INC(ips_odropped);
965 * In the first mbuf, leave room for the link header, then
966 * copy the original IP header including options. The payload
967 * goes into an additional mbuf chain returned by m_copym().
969 m->m_data += max_linkhdr;
970 mhip = mtod(m, struct ip *);
972 if (hlen > sizeof (struct ip)) {
973 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
974 mhip->ip_v = IPVERSION;
975 mhip->ip_hl = mhlen >> 2;
978 /* XXX do we need to add ip_off below ? */
979 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
980 if (off + len >= ip_len)
983 mhip->ip_off |= IP_MF;
984 mhip->ip_len = htons((u_short)(len + mhlen));
985 m->m_next = m_copym(m0, off, len, M_NOWAIT);
986 if (m->m_next == NULL) { /* copy failed */
988 error = ENOBUFS; /* ??? */
989 IPSTAT_INC(ips_odropped);
992 m->m_pkthdr.len = mhlen + len;
994 mac_netinet_fragment(m0, m);
996 mhip->ip_off = htons(mhip->ip_off);
998 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
999 mhip->ip_sum = in_cksum(m, mhlen);
1000 m->m_pkthdr.csum_flags &= ~CSUM_IP;
1003 mnext = &m->m_nextpkt;
1005 IPSTAT_ADD(ips_ofragments, nfrags);
1008 * Update first fragment by trimming what's been copied out
1009 * and updating header.
1011 m_adj(m0, hlen + firstlen - ip_len);
1012 m0->m_pkthdr.len = hlen + firstlen;
1013 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1014 ip->ip_off = htons(ip_off | IP_MF);
1016 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1017 ip->ip_sum = in_cksum(m0, hlen);
1018 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1027 in_delayed_cksum(struct mbuf *m)
1031 uint16_t cklen, csum, offset;
1033 ip = mtod(m, struct ip *);
1034 offset = ip->ip_hl << 2 ;
1036 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1037 /* if udp header is not in the first mbuf copy udplen */
1038 if (offset + sizeof(struct udphdr) > m->m_len) {
1039 m_copydata(m, offset + offsetof(struct udphdr,
1040 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1041 cklen = ntohs(cklen);
1043 uh = (struct udphdr *)mtodo(m, offset);
1044 cklen = ntohs(uh->uh_ulen);
1046 csum = in_cksum_skip(m, cklen + offset, offset);
1050 cklen = ntohs(ip->ip_len);
1051 csum = in_cksum_skip(m, cklen, offset);
1053 offset += m->m_pkthdr.csum_data; /* checksum offset */
1055 if (offset + sizeof(csum) > m->m_len)
1056 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1058 *(u_short *)mtodo(m, offset) = csum;
1062 * IP socket option processing.
1065 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1067 struct inpcb *inp = sotoinpcb(so);
1070 uint32_t rss_bucket;
1075 if (sopt->sopt_level != IPPROTO_IP) {
1078 if (sopt->sopt_level == SOL_SOCKET &&
1079 sopt->sopt_dir == SOPT_SET) {
1080 switch (sopt->sopt_name) {
1083 if ((so->so_options & SO_REUSEADDR) != 0)
1084 inp->inp_flags2 |= INP_REUSEADDR;
1086 inp->inp_flags2 &= ~INP_REUSEADDR;
1092 if ((so->so_options & SO_REUSEPORT) != 0)
1093 inp->inp_flags2 |= INP_REUSEPORT;
1095 inp->inp_flags2 &= ~INP_REUSEPORT;
1099 case SO_REUSEPORT_LB:
1101 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1102 inp->inp_flags2 |= INP_REUSEPORT_LB;
1104 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1110 inp->inp_inc.inc_fibnum = so->so_fibnum;
1114 case SO_MAX_PACING_RATE:
1117 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1131 switch (sopt->sopt_dir) {
1133 switch (sopt->sopt_name) {
1140 if (sopt->sopt_valsize > MLEN) {
1144 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1149 m->m_len = sopt->sopt_valsize;
1150 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1157 error = ip_pcbopts(inp, sopt->sopt_name, m);
1163 if (sopt->sopt_td != NULL) {
1164 error = priv_check(sopt->sopt_td,
1165 PRIV_NETINET_BINDANY);
1172 case IP_RSS_LISTEN_BUCKET:
1178 case IP_RECVRETOPTS:
1179 case IP_ORIGDSTADDR:
1180 case IP_RECVDSTADDR:
1188 case IP_RECVRSSBUCKETID:
1191 error = sooptcopyin(sopt, &optval, sizeof optval,
1196 switch (sopt->sopt_name) {
1198 inp->inp_ip_tos = optval;
1202 inp->inp_ip_ttl = optval;
1206 if (optval >= 0 && optval <= MAXTTL)
1207 inp->inp_ip_minttl = optval;
1212 #define OPTSET(bit) do { \
1215 inp->inp_flags |= bit; \
1217 inp->inp_flags &= ~bit; \
1221 #define OPTSET2(bit, val) do { \
1224 inp->inp_flags2 |= bit; \
1226 inp->inp_flags2 &= ~bit; \
1231 OPTSET(INP_RECVOPTS);
1234 case IP_RECVRETOPTS:
1235 OPTSET(INP_RECVRETOPTS);
1238 case IP_RECVDSTADDR:
1239 OPTSET(INP_RECVDSTADDR);
1242 case IP_ORIGDSTADDR:
1243 OPTSET2(INP_ORIGDSTADDR, optval);
1247 OPTSET(INP_RECVTTL);
1255 OPTSET(INP_ONESBCAST);
1258 OPTSET(INP_DONTFRAG);
1261 OPTSET(INP_BINDANY);
1264 OPTSET(INP_RECVTOS);
1267 OPTSET2(INP_BINDMULTI, optval);
1270 OPTSET2(INP_RECVFLOWID, optval);
1273 case IP_RSS_LISTEN_BUCKET:
1274 if ((optval >= 0) &&
1275 (optval < rss_getnumbuckets())) {
1276 inp->inp_rss_listen_bucket = optval;
1277 OPTSET2(INP_RSS_BUCKET_SET, 1);
1282 case IP_RECVRSSBUCKETID:
1283 OPTSET2(INP_RECVRSSBUCKETID, optval);
1287 if ((optval >= -1) && (optval <=
1288 (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1302 optval << INP_2PCP_SHIFT;
1314 * Multicast socket options are processed by the in_mcast
1317 case IP_MULTICAST_IF:
1318 case IP_MULTICAST_VIF:
1319 case IP_MULTICAST_TTL:
1320 case IP_MULTICAST_LOOP:
1321 case IP_ADD_MEMBERSHIP:
1322 case IP_DROP_MEMBERSHIP:
1323 case IP_ADD_SOURCE_MEMBERSHIP:
1324 case IP_DROP_SOURCE_MEMBERSHIP:
1325 case IP_BLOCK_SOURCE:
1326 case IP_UNBLOCK_SOURCE:
1328 case MCAST_JOIN_GROUP:
1329 case MCAST_LEAVE_GROUP:
1330 case MCAST_JOIN_SOURCE_GROUP:
1331 case MCAST_LEAVE_SOURCE_GROUP:
1332 case MCAST_BLOCK_SOURCE:
1333 case MCAST_UNBLOCK_SOURCE:
1334 error = inp_setmoptions(inp, sopt);
1338 error = sooptcopyin(sopt, &optval, sizeof optval,
1345 case IP_PORTRANGE_DEFAULT:
1346 inp->inp_flags &= ~(INP_LOWPORT);
1347 inp->inp_flags &= ~(INP_HIGHPORT);
1350 case IP_PORTRANGE_HIGH:
1351 inp->inp_flags &= ~(INP_LOWPORT);
1352 inp->inp_flags |= INP_HIGHPORT;
1355 case IP_PORTRANGE_LOW:
1356 inp->inp_flags &= ~(INP_HIGHPORT);
1357 inp->inp_flags |= INP_LOWPORT;
1367 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1368 case IP_IPSEC_POLICY:
1369 if (IPSEC_ENABLED(ipv4)) {
1370 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1377 error = ENOPROTOOPT;
1383 switch (sopt->sopt_name) {
1387 if (inp->inp_options) {
1388 struct mbuf *options;
1390 options = m_copym(inp->inp_options, 0,
1391 M_COPYALL, M_NOWAIT);
1393 if (options != NULL) {
1394 error = sooptcopyout(sopt,
1395 mtod(options, char *),
1402 sopt->sopt_valsize = 0;
1410 case IP_RECVRETOPTS:
1411 case IP_ORIGDSTADDR:
1412 case IP_RECVDSTADDR:
1425 case IP_RSSBUCKETID:
1426 case IP_RECVRSSBUCKETID:
1429 switch (sopt->sopt_name) {
1431 optval = inp->inp_ip_tos;
1435 optval = inp->inp_ip_ttl;
1439 optval = inp->inp_ip_minttl;
1442 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1443 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1446 optval = OPTBIT(INP_RECVOPTS);
1449 case IP_RECVRETOPTS:
1450 optval = OPTBIT(INP_RECVRETOPTS);
1453 case IP_RECVDSTADDR:
1454 optval = OPTBIT(INP_RECVDSTADDR);
1457 case IP_ORIGDSTADDR:
1458 optval = OPTBIT2(INP_ORIGDSTADDR);
1462 optval = OPTBIT(INP_RECVTTL);
1466 optval = OPTBIT(INP_RECVIF);
1470 if (inp->inp_flags & INP_HIGHPORT)
1471 optval = IP_PORTRANGE_HIGH;
1472 else if (inp->inp_flags & INP_LOWPORT)
1473 optval = IP_PORTRANGE_LOW;
1479 optval = OPTBIT(INP_ONESBCAST);
1482 optval = OPTBIT(INP_DONTFRAG);
1485 optval = OPTBIT(INP_BINDANY);
1488 optval = OPTBIT(INP_RECVTOS);
1491 optval = inp->inp_flowid;
1494 optval = inp->inp_flowtype;
1497 optval = OPTBIT2(INP_RECVFLOWID);
1500 case IP_RSSBUCKETID:
1501 retval = rss_hash2bucket(inp->inp_flowid,
1505 optval = rss_bucket;
1509 case IP_RECVRSSBUCKETID:
1510 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1514 optval = OPTBIT2(INP_BINDMULTI);
1517 if (OPTBIT2(INP_2PCP_SET)) {
1518 optval = (inp->inp_flags2 &
1519 INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1525 error = sooptcopyout(sopt, &optval, sizeof optval);
1529 * Multicast socket options are processed by the in_mcast
1532 case IP_MULTICAST_IF:
1533 case IP_MULTICAST_VIF:
1534 case IP_MULTICAST_TTL:
1535 case IP_MULTICAST_LOOP:
1537 error = inp_getmoptions(inp, sopt);
1540 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1541 case IP_IPSEC_POLICY:
1542 if (IPSEC_ENABLED(ipv4)) {
1543 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1550 error = ENOPROTOOPT;
1559 * Routine called from ip_output() to loop back a copy of an IP multicast
1560 * packet to the input queue of a specified interface. Note that this
1561 * calls the output routine of the loopback "driver", but with an interface
1562 * pointer that might NOT be a loopback interface -- evil, but easier than
1563 * replicating that code here.
1566 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1572 * Make a deep copy of the packet because we're going to
1573 * modify the pack in order to generate checksums.
1575 copym = m_dup(m, M_NOWAIT);
1576 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1577 copym = m_pullup(copym, hlen);
1578 if (copym != NULL) {
1579 /* If needed, compute the checksum and mark it as valid. */
1580 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1581 in_delayed_cksum(copym);
1582 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1583 copym->m_pkthdr.csum_flags |=
1584 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1585 copym->m_pkthdr.csum_data = 0xffff;
1588 * We don't bother to fragment if the IP length is greater
1589 * than the interface's MTU. Can this possibly matter?
1591 ip = mtod(copym, struct ip *);
1593 ip->ip_sum = in_cksum(copym, hlen);
1594 if_simloop(ifp, copym, AF_INET, 0);