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_mpath.h"
42 #include "opt_ratelimit.h"
43 #include "opt_route.h"
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/kernel.h>
52 #include <sys/malloc.h>
56 #include <sys/protosw.h>
57 #include <sys/rmlock.h>
59 #include <sys/socket.h>
60 #include <sys/socketvar.h>
61 #include <sys/sysctl.h>
62 #include <sys/ucred.h>
65 #include <net/if_var.h>
66 #include <net/if_llatbl.h>
67 #include <net/netisr.h>
69 #include <net/route.h>
71 #include <net/radix_mpath.h>
73 #include <net/rss_config.h>
76 #include <netinet/in.h>
77 #include <netinet/in_fib.h>
78 #include <netinet/in_kdtrace.h>
79 #include <netinet/in_systm.h>
80 #include <netinet/ip.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>
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);
110 extern int in_mcast_loop;
111 extern struct protosw inetsw[];
114 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
115 struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
117 struct m_tag *fwd_tag = NULL;
121 int pflags = PFIL_OUT;
123 if (flags & IP_FORWARDING)
127 ip = mtod(m, struct ip *);
129 /* Run through list of hooks for output packets. */
130 odst.s_addr = ip->ip_dst.s_addr;
131 switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
136 return 1; /* Finished */
141 ip = mtod(m, struct ip *);
143 /* See if destination IP address was changed by packet filter. */
144 if (odst.s_addr != ip->ip_dst.s_addr) {
145 m->m_flags |= M_SKIP_FIREWALL;
146 /* If destination is now ourself drop to ip_input(). */
147 if (in_localip(ip->ip_dst)) {
148 m->m_flags |= M_FASTFWD_OURS;
149 if (m->m_pkthdr.rcvif == NULL)
150 m->m_pkthdr.rcvif = V_loif;
151 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
152 m->m_pkthdr.csum_flags |=
153 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
154 m->m_pkthdr.csum_data = 0xffff;
156 m->m_pkthdr.csum_flags |=
157 CSUM_IP_CHECKED | CSUM_IP_VALID;
159 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
160 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
162 *error = netisr_queue(NETISR_IP, m);
163 return 1; /* Finished */
166 bzero(dst, sizeof(*dst));
167 dst->sin_family = AF_INET;
168 dst->sin_len = sizeof(*dst);
169 dst->sin_addr = ip->ip_dst;
171 return -1; /* Reloop */
173 /* See if fib was changed by packet filter. */
174 if ((*fibnum) != M_GETFIB(m)) {
175 m->m_flags |= M_SKIP_FIREWALL;
176 *fibnum = M_GETFIB(m);
177 return -1; /* Reloop for FIB change */
180 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
181 if (m->m_flags & M_FASTFWD_OURS) {
182 if (m->m_pkthdr.rcvif == NULL)
183 m->m_pkthdr.rcvif = V_loif;
184 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
185 m->m_pkthdr.csum_flags |=
186 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
187 m->m_pkthdr.csum_data = 0xffff;
190 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
191 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
193 m->m_pkthdr.csum_flags |=
194 CSUM_IP_CHECKED | CSUM_IP_VALID;
196 *error = netisr_queue(NETISR_IP, m);
197 return 1; /* Finished */
199 /* Or forward to some other address? */
200 if ((m->m_flags & M_IP_NEXTHOP) &&
201 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
202 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
203 m->m_flags |= M_SKIP_FIREWALL;
204 m->m_flags &= ~M_IP_NEXTHOP;
205 m_tag_delete(m, fwd_tag);
207 return -1; /* Reloop for CHANGE of dst */
214 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
215 const struct sockaddr_in *gw, struct route *ro)
218 struct ktls_session *tls = NULL;
220 struct m_snd_tag *mst;
223 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
228 * If this is an unencrypted TLS record, save a reference to
229 * the record. This local reference is used to call
230 * ktls_output_eagain after the mbuf has been freed (thus
231 * dropping the mbuf's reference) in if_output.
233 if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
234 tls = ktls_hold(m->m_next->m_ext.ext_pgs->tls);
238 * If a TLS session doesn't have a valid tag, it must
239 * have had an earlier ifp mismatch, so drop this
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;
260 KASSERT(m->m_pkthdr.rcvif == NULL,
261 ("trying to add a send tag to a forwarded packet"));
262 if (mst->ifp != ifp) {
267 /* stamp send tag on mbuf */
268 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
269 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
272 error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
275 /* Check for route change invalidating send tags. */
279 error = ktls_output_eagain(inp, tls);
285 in_pcboutput_eagain(inp);
291 * IP output. The packet in mbuf chain m contains a skeletal IP
292 * header (with len, off, ttl, proto, tos, src, dst).
293 * The mbuf chain containing the packet will be freed.
294 * The mbuf opt, if present, will not be freed.
295 * If route ro is present and has ro_rt initialized, route lookup would be
296 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
297 * then result of route lookup is stored in ro->ro_rt.
299 * In the IP forwarding case, the packet will arrive with options already
300 * inserted, so must have a NULL opt pointer.
303 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
304 struct ip_moptions *imo, struct inpcb *inp)
306 struct rm_priotracker in_ifa_tracker;
307 struct epoch_tracker et;
309 struct ifnet *ifp = NULL; /* keep compiler happy */
311 int hlen = sizeof (struct ip);
314 struct sockaddr_in *dst, sin;
315 const struct sockaddr_in *gw;
316 struct in_ifaddr *ia;
319 uint16_t ip_len, ip_off;
321 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
322 int no_route_but_check_spd = 0;
328 INP_LOCK_ASSERT(inp);
329 M_SETFIB(m, inp->inp_inc.inc_fibnum);
330 if ((flags & IP_NODEFAULTFLOWID) == 0) {
331 m->m_pkthdr.flowid = inp->inp_flowid;
332 M_HASHTYPE_SET(m, inp->inp_flowtype);
335 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
341 m = ip_insertoptions(m, opt, &len);
343 hlen = len; /* ip->ip_hl is updated above */
345 ip = mtod(m, struct ip *);
346 ip_len = ntohs(ip->ip_len);
347 ip_off = ntohs(ip->ip_off);
349 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
350 ip->ip_v = IPVERSION;
351 ip->ip_hl = hlen >> 2;
354 /* Header already set, fetch hlen from there */
355 hlen = ip->ip_hl << 2;
357 if ((flags & IP_FORWARDING) == 0)
358 IPSTAT_INC(ips_localout);
363 * gw is readonly but can point either to dst OR rt_gateway,
364 * therefore we need restore gw if we're redoing lookup.
366 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
368 dst = (struct sockaddr_in *)&ro->ro_dst;
371 if (ro == NULL || ro->ro_rt == NULL) {
372 bzero(dst, sizeof(*dst));
373 dst->sin_family = AF_INET;
374 dst->sin_len = sizeof(*dst);
375 dst->sin_addr = ip->ip_dst;
381 * Validate route against routing table additions;
382 * a better/more specific route might have been added.
384 if (inp != NULL && ro != NULL && ro->ro_rt != NULL)
385 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
387 * If there is a cached route,
388 * check that it is to the same destination
389 * and is still up. If not, free it and try again.
390 * The address family should also be checked in case of sharing the
392 * Also check whether routing cache needs invalidation.
394 if (ro != NULL && ro->ro_rt != NULL &&
395 ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
396 ro->ro_rt->rt_ifp == NULL || !RT_LINK_IS_UP(ro->ro_rt->rt_ifp) ||
397 dst->sin_family != AF_INET ||
398 dst->sin_addr.s_addr != ip->ip_dst.s_addr))
399 RO_INVALIDATE_CACHE(ro);
402 * If routing to interface only, short circuit routing lookup.
403 * The use of an all-ones broadcast address implies this; an
404 * interface is specified by the broadcast address of an interface,
405 * or the destination address of a ptp interface.
407 if (flags & IP_SENDONES) {
408 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
409 M_GETFIB(m)))) == NULL &&
410 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
411 M_GETFIB(m)))) == NULL) {
412 IPSTAT_INC(ips_noroute);
416 ip->ip_dst.s_addr = INADDR_BROADCAST;
417 dst->sin_addr = ip->ip_dst;
422 src = IA_SIN(ia)->sin_addr;
423 } else if (flags & IP_ROUTETOIF) {
424 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
425 M_GETFIB(m)))) == NULL &&
426 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
427 M_GETFIB(m)))) == NULL) {
428 IPSTAT_INC(ips_noroute);
435 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
436 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
437 src = IA_SIN(ia)->sin_addr;
438 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
439 imo != NULL && imo->imo_multicast_ifp != NULL) {
441 * Bypass the normal routing lookup for multicast
442 * packets if the interface is specified.
444 ifp = imo->imo_multicast_ifp;
446 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
447 isbroadcast = 0; /* fool gcc */
448 /* Interface may have no addresses. */
450 src = IA_SIN(ia)->sin_addr;
452 src.s_addr = INADDR_ANY;
453 } else if (ro != NULL) {
454 if (ro->ro_rt == NULL) {
456 * We want to do any cloning requested by the link
457 * layer, as this is probably required in all cases
458 * for correct operation (as it is for ARP).
461 rtalloc_mpath_fib(ro,
462 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
465 in_rtalloc_ign(ro, 0, fibnum);
467 if (ro->ro_rt == NULL ||
468 (ro->ro_rt->rt_flags & RTF_UP) == 0 ||
469 ro->ro_rt->rt_ifp == NULL ||
470 !RT_LINK_IS_UP(ro->ro_rt->rt_ifp)) {
471 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
473 * There is no route for this packet, but it is
474 * possible that a matching SPD entry exists.
476 no_route_but_check_spd = 1;
477 mtu = 0; /* Silence GCC warning. */
480 IPSTAT_INC(ips_noroute);
481 error = EHOSTUNREACH;
485 ia = ifatoia(ro->ro_rt->rt_ifa);
486 ifp = ro->ro_rt->rt_ifp;
487 counter_u64_add(ro->ro_rt->rt_pksent, 1);
488 rt_update_ro_flags(ro);
489 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
490 gw = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
491 if (ro->ro_rt->rt_flags & RTF_HOST)
492 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
493 else if (ifp->if_flags & IFF_BROADCAST)
494 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
497 if (ro->ro_rt->rt_flags & RTF_HOST)
498 mtu = ro->ro_rt->rt_mtu;
501 src = IA_SIN(ia)->sin_addr;
503 struct nhop4_extended nh;
505 bzero(&nh, sizeof(nh));
506 if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh) !=
508 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
510 * There is no route for this packet, but it is
511 * possible that a matching SPD entry exists.
513 no_route_but_check_spd = 1;
514 mtu = 0; /* Silence GCC warning. */
517 IPSTAT_INC(ips_noroute);
518 error = EHOSTUNREACH;
524 * We are rewriting here dst to be gw actually, contradicting
525 * comment at the beginning of the function. However, in this
526 * case we are always dealing with on stack dst.
527 * In case if pfil(9) sends us back to beginning of the
528 * function, the dst would be rewritten by ip_output_pfil().
531 dst->sin_addr = nh.nh_addr;
534 isbroadcast = (((nh.nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
535 (NHF_HOST | NHF_BROADCAST)) ||
536 ((ifp->if_flags & IFF_BROADCAST) &&
537 in_ifaddr_broadcast(dst->sin_addr, ia)));
540 /* Catch a possible divide by zero later. */
541 KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (rt_flags=0x%08x) ifp=%p",
543 (ro != NULL && ro->ro_rt != NULL) ? ro->ro_rt->rt_flags : 0, ifp));
545 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
546 m->m_flags |= M_MCAST;
548 * IP destination address is multicast. Make sure "gw"
549 * still points to the address in "ro". (It may have been
550 * changed to point to a gateway address, above.)
554 * See if the caller provided any multicast options
557 ip->ip_ttl = imo->imo_multicast_ttl;
558 if (imo->imo_multicast_vif != -1)
561 ip_mcast_src(imo->imo_multicast_vif) :
564 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
566 * Confirm that the outgoing interface supports multicast.
568 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
569 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
570 IPSTAT_INC(ips_noroute);
576 * If source address not specified yet, use address
577 * of outgoing interface.
579 if (ip->ip_src.s_addr == INADDR_ANY)
582 if ((imo == NULL && in_mcast_loop) ||
583 (imo && imo->imo_multicast_loop)) {
585 * Loop back multicast datagram if not expressly
586 * forbidden to do so, even if we are not a member
587 * of the group; ip_input() will filter it later,
588 * thus deferring a hash lookup and mutex acquisition
589 * at the expense of a cheap copy using m_copym().
591 ip_mloopback(ifp, m, hlen);
594 * If we are acting as a multicast router, perform
595 * multicast forwarding as if the packet had just
596 * arrived on the interface to which we are about
597 * to send. The multicast forwarding function
598 * recursively calls this function, using the
599 * IP_FORWARDING flag to prevent infinite recursion.
601 * Multicasts that are looped back by ip_mloopback(),
602 * above, will be forwarded by the ip_input() routine,
605 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
607 * If rsvp daemon is not running, do not
608 * set ip_moptions. This ensures that the packet
609 * is multicast and not just sent down one link
610 * as prescribed by rsvpd.
615 ip_mforward(ip, ifp, m, imo) != 0) {
623 * Multicasts with a time-to-live of zero may be looped-
624 * back, above, but must not be transmitted on a network.
625 * Also, multicasts addressed to the loopback interface
626 * are not sent -- the above call to ip_mloopback() will
627 * loop back a copy. ip_input() will drop the copy if
628 * this host does not belong to the destination group on
629 * the loopback interface.
631 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
640 * If the source address is not specified yet, use the address
641 * of the outoing interface.
643 if (ip->ip_src.s_addr == INADDR_ANY)
647 * Look for broadcast address and
648 * verify user is allowed to send
652 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
653 error = EADDRNOTAVAIL;
656 if ((flags & IP_ALLOWBROADCAST) == 0) {
660 /* don't allow broadcast messages to be fragmented */
665 m->m_flags |= M_BCAST;
667 m->m_flags &= ~M_BCAST;
671 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
672 if (IPSEC_ENABLED(ipv4)) {
673 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
674 if (error == EINPROGRESS)
680 * Check if there was a route for this packet; return error if not.
682 if (no_route_but_check_spd) {
683 IPSTAT_INC(ips_noroute);
684 error = EHOSTUNREACH;
687 /* Update variables that are affected by ipsec4_output(). */
688 ip = mtod(m, struct ip *);
689 hlen = ip->ip_hl << 2;
692 /* Jump over all PFIL processing if hooks are not active. */
693 if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
694 switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
696 case 1: /* Finished */
699 case 0: /* Continue normally */
700 ip = mtod(m, struct ip *);
703 case -1: /* Need to try again */
704 /* Reset everything for a new round */
707 ro->ro_prepend = NULL;
710 ip = mtod(m, struct ip *);
716 /* IN_LOOPBACK must not appear on the wire - RFC1122. */
717 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
718 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
719 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
720 IPSTAT_INC(ips_badaddr);
721 error = EADDRNOTAVAIL;
726 m->m_pkthdr.csum_flags |= CSUM_IP;
727 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
728 m = mb_unmapped_to_ext(m);
730 IPSTAT_INC(ips_odropped);
735 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
736 } else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
737 m = mb_unmapped_to_ext(m);
739 IPSTAT_INC(ips_odropped);
745 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
746 m = mb_unmapped_to_ext(m);
748 IPSTAT_INC(ips_odropped);
752 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
753 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
758 * If small enough for interface, or the interface will take
759 * care of the fragmentation for us, we can just send directly.
762 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
764 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
765 ip->ip_sum = in_cksum(m, hlen);
766 m->m_pkthdr.csum_flags &= ~CSUM_IP;
770 * Record statistics for this interface address.
771 * With CSUM_TSO the byte/packet count will be slightly
772 * incorrect because we count the IP+TCP headers only
773 * once instead of for every generated packet.
775 if (!(flags & IP_FORWARDING) && ia) {
776 if (m->m_pkthdr.csum_flags & CSUM_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);
798 /* Balk when DF bit is set or the interface didn't support TSO. */
799 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
801 IPSTAT_INC(ips_cantfrag);
806 * Too large for interface; fragment if possible. If successful,
807 * on return, m will point to a list of packets to be sent.
809 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
816 /* Record statistics for this interface address. */
818 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
819 counter_u64_add(ia->ia_ifa.ifa_obytes,
823 * Reset layer specific mbuf flags
824 * to avoid confusing upper layers.
828 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
829 mtod(m, struct ip *), NULL);
830 error = ip_output_send(inp, ifp, m, gw, ro);
836 IPSTAT_INC(ips_fragmented);
847 * Create a chain of fragments which fit the given mtu. m_frag points to the
848 * mbuf to be fragmented; on return it points to the chain with the fragments.
849 * Return 0 if no error. If error, m_frag may contain a partially built
850 * chain of fragments that should be freed by the caller.
852 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
855 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
856 u_long if_hwassist_flags)
859 int hlen = ip->ip_hl << 2;
860 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
862 struct mbuf *m0 = *m_frag; /* the original packet */
866 uint16_t ip_len, ip_off;
868 ip_len = ntohs(ip->ip_len);
869 ip_off = ntohs(ip->ip_off);
871 if (ip_off & IP_DF) { /* Fragmentation not allowed */
872 IPSTAT_INC(ips_cantfrag);
877 * Must be able to put at least 8 bytes per fragment.
883 * If the interface will not calculate checksums on
884 * fragmented packets, then do it here.
886 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
887 m0 = mb_unmapped_to_ext(m0);
890 IPSTAT_INC(ips_odropped);
893 in_delayed_cksum(m0);
894 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
897 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
898 m0 = mb_unmapped_to_ext(m0);
901 IPSTAT_INC(ips_odropped);
904 sctp_delayed_cksum(m0, hlen);
905 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
908 if (len > PAGE_SIZE) {
910 * Fragment large datagrams such that each segment
911 * contains a multiple of PAGE_SIZE amount of data,
912 * plus headers. This enables a receiver to perform
913 * page-flipping zero-copy optimizations.
915 * XXX When does this help given that sender and receiver
916 * could have different page sizes, and also mtu could
917 * be less than the receiver's page size ?
921 off = MIN(mtu, m0->m_pkthdr.len);
924 * firstlen (off - hlen) must be aligned on an
928 goto smart_frag_failure;
929 off = ((off - hlen) & ~7) + hlen;
930 newlen = (~PAGE_MASK) & mtu;
931 if ((newlen + sizeof (struct ip)) > mtu) {
932 /* we failed, go back the default */
943 firstlen = off - hlen;
944 mnext = &m0->m_nextpkt; /* pointer to next packet */
947 * Loop through length of segment after first fragment,
948 * make new header and copy data of each part and link onto chain.
949 * Here, m0 is the original packet, m is the fragment being created.
950 * The fragments are linked off the m_nextpkt of the original
951 * packet, which after processing serves as the first fragment.
953 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
954 struct ip *mhip; /* ip header on the fragment */
956 int mhlen = sizeof (struct ip);
958 m = m_gethdr(M_NOWAIT, MT_DATA);
961 IPSTAT_INC(ips_odropped);
965 * Make sure the complete packet header gets copied
966 * from the originating mbuf to the newly created
967 * mbuf. This also ensures that existing firewall
968 * classification(s), VLAN tags and so on get copied
969 * to the resulting fragmented packet(s):
971 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
974 IPSTAT_INC(ips_odropped);
978 * In the first mbuf, leave room for the link header, then
979 * copy the original IP header including options. The payload
980 * goes into an additional mbuf chain returned by m_copym().
982 m->m_data += max_linkhdr;
983 mhip = mtod(m, struct ip *);
985 if (hlen > sizeof (struct ip)) {
986 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
987 mhip->ip_v = IPVERSION;
988 mhip->ip_hl = mhlen >> 2;
991 /* XXX do we need to add ip_off below ? */
992 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
993 if (off + len >= ip_len)
996 mhip->ip_off |= IP_MF;
997 mhip->ip_len = htons((u_short)(len + mhlen));
998 m->m_next = m_copym(m0, off, len, M_NOWAIT);
999 if (m->m_next == NULL) { /* copy failed */
1001 error = ENOBUFS; /* ??? */
1002 IPSTAT_INC(ips_odropped);
1005 m->m_pkthdr.len = mhlen + len;
1007 mac_netinet_fragment(m0, m);
1009 mhip->ip_off = htons(mhip->ip_off);
1011 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1012 mhip->ip_sum = in_cksum(m, mhlen);
1013 m->m_pkthdr.csum_flags &= ~CSUM_IP;
1016 mnext = &m->m_nextpkt;
1018 IPSTAT_ADD(ips_ofragments, nfrags);
1021 * Update first fragment by trimming what's been copied out
1022 * and updating header.
1024 m_adj(m0, hlen + firstlen - ip_len);
1025 m0->m_pkthdr.len = hlen + firstlen;
1026 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1027 ip->ip_off = htons(ip_off | IP_MF);
1029 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1030 ip->ip_sum = in_cksum(m0, hlen);
1031 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1040 in_delayed_cksum(struct mbuf *m)
1044 uint16_t cklen, csum, offset;
1046 ip = mtod(m, struct ip *);
1047 offset = ip->ip_hl << 2 ;
1049 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1050 /* if udp header is not in the first mbuf copy udplen */
1051 if (offset + sizeof(struct udphdr) > m->m_len) {
1052 m_copydata(m, offset + offsetof(struct udphdr,
1053 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1054 cklen = ntohs(cklen);
1056 uh = (struct udphdr *)mtodo(m, offset);
1057 cklen = ntohs(uh->uh_ulen);
1059 csum = in_cksum_skip(m, cklen + offset, offset);
1063 cklen = ntohs(ip->ip_len);
1064 csum = in_cksum_skip(m, cklen, offset);
1066 offset += m->m_pkthdr.csum_data; /* checksum offset */
1068 if (offset + sizeof(csum) > m->m_len)
1069 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1071 *(u_short *)mtodo(m, offset) = csum;
1075 * IP socket option processing.
1078 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1080 struct inpcb *inp = sotoinpcb(so);
1083 uint32_t rss_bucket;
1088 if (sopt->sopt_level != IPPROTO_IP) {
1091 if (sopt->sopt_level == SOL_SOCKET &&
1092 sopt->sopt_dir == SOPT_SET) {
1093 switch (sopt->sopt_name) {
1096 if ((so->so_options & SO_REUSEADDR) != 0)
1097 inp->inp_flags2 |= INP_REUSEADDR;
1099 inp->inp_flags2 &= ~INP_REUSEADDR;
1105 if ((so->so_options & SO_REUSEPORT) != 0)
1106 inp->inp_flags2 |= INP_REUSEPORT;
1108 inp->inp_flags2 &= ~INP_REUSEPORT;
1112 case SO_REUSEPORT_LB:
1114 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1115 inp->inp_flags2 |= INP_REUSEPORT_LB;
1117 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1123 inp->inp_inc.inc_fibnum = so->so_fibnum;
1127 case SO_MAX_PACING_RATE:
1130 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1144 switch (sopt->sopt_dir) {
1146 switch (sopt->sopt_name) {
1153 if (sopt->sopt_valsize > MLEN) {
1157 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1162 m->m_len = sopt->sopt_valsize;
1163 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1170 error = ip_pcbopts(inp, sopt->sopt_name, m);
1176 if (sopt->sopt_td != NULL) {
1177 error = priv_check(sopt->sopt_td,
1178 PRIV_NETINET_BINDANY);
1185 case IP_RSS_LISTEN_BUCKET:
1191 case IP_RECVRETOPTS:
1192 case IP_ORIGDSTADDR:
1193 case IP_RECVDSTADDR:
1201 case IP_RECVRSSBUCKETID:
1203 error = sooptcopyin(sopt, &optval, sizeof optval,
1208 switch (sopt->sopt_name) {
1210 inp->inp_ip_tos = optval;
1214 inp->inp_ip_ttl = optval;
1218 if (optval >= 0 && optval <= MAXTTL)
1219 inp->inp_ip_minttl = optval;
1224 #define OPTSET(bit) do { \
1227 inp->inp_flags |= bit; \
1229 inp->inp_flags &= ~bit; \
1233 #define OPTSET2(bit, val) do { \
1236 inp->inp_flags2 |= bit; \
1238 inp->inp_flags2 &= ~bit; \
1243 OPTSET(INP_RECVOPTS);
1246 case IP_RECVRETOPTS:
1247 OPTSET(INP_RECVRETOPTS);
1250 case IP_RECVDSTADDR:
1251 OPTSET(INP_RECVDSTADDR);
1254 case IP_ORIGDSTADDR:
1255 OPTSET2(INP_ORIGDSTADDR, optval);
1259 OPTSET(INP_RECVTTL);
1267 OPTSET(INP_ONESBCAST);
1270 OPTSET(INP_DONTFRAG);
1273 OPTSET(INP_BINDANY);
1276 OPTSET(INP_RECVTOS);
1279 OPTSET2(INP_BINDMULTI, optval);
1282 OPTSET2(INP_RECVFLOWID, optval);
1285 case IP_RSS_LISTEN_BUCKET:
1286 if ((optval >= 0) &&
1287 (optval < rss_getnumbuckets())) {
1288 inp->inp_rss_listen_bucket = optval;
1289 OPTSET2(INP_RSS_BUCKET_SET, 1);
1294 case IP_RECVRSSBUCKETID:
1295 OPTSET2(INP_RECVRSSBUCKETID, optval);
1304 * Multicast socket options are processed by the in_mcast
1307 case IP_MULTICAST_IF:
1308 case IP_MULTICAST_VIF:
1309 case IP_MULTICAST_TTL:
1310 case IP_MULTICAST_LOOP:
1311 case IP_ADD_MEMBERSHIP:
1312 case IP_DROP_MEMBERSHIP:
1313 case IP_ADD_SOURCE_MEMBERSHIP:
1314 case IP_DROP_SOURCE_MEMBERSHIP:
1315 case IP_BLOCK_SOURCE:
1316 case IP_UNBLOCK_SOURCE:
1318 case MCAST_JOIN_GROUP:
1319 case MCAST_LEAVE_GROUP:
1320 case MCAST_JOIN_SOURCE_GROUP:
1321 case MCAST_LEAVE_SOURCE_GROUP:
1322 case MCAST_BLOCK_SOURCE:
1323 case MCAST_UNBLOCK_SOURCE:
1324 error = inp_setmoptions(inp, sopt);
1328 error = sooptcopyin(sopt, &optval, sizeof optval,
1335 case IP_PORTRANGE_DEFAULT:
1336 inp->inp_flags &= ~(INP_LOWPORT);
1337 inp->inp_flags &= ~(INP_HIGHPORT);
1340 case IP_PORTRANGE_HIGH:
1341 inp->inp_flags &= ~(INP_LOWPORT);
1342 inp->inp_flags |= INP_HIGHPORT;
1345 case IP_PORTRANGE_LOW:
1346 inp->inp_flags &= ~(INP_HIGHPORT);
1347 inp->inp_flags |= INP_LOWPORT;
1357 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1358 case IP_IPSEC_POLICY:
1359 if (IPSEC_ENABLED(ipv4)) {
1360 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1367 error = ENOPROTOOPT;
1373 switch (sopt->sopt_name) {
1377 if (inp->inp_options) {
1378 struct mbuf *options;
1380 options = m_copym(inp->inp_options, 0,
1381 M_COPYALL, M_NOWAIT);
1383 if (options != NULL) {
1384 error = sooptcopyout(sopt,
1385 mtod(options, char *),
1392 sopt->sopt_valsize = 0;
1400 case IP_RECVRETOPTS:
1401 case IP_ORIGDSTADDR:
1402 case IP_RECVDSTADDR:
1415 case IP_RSSBUCKETID:
1416 case IP_RECVRSSBUCKETID:
1418 switch (sopt->sopt_name) {
1421 optval = inp->inp_ip_tos;
1425 optval = inp->inp_ip_ttl;
1429 optval = inp->inp_ip_minttl;
1432 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1433 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1436 optval = OPTBIT(INP_RECVOPTS);
1439 case IP_RECVRETOPTS:
1440 optval = OPTBIT(INP_RECVRETOPTS);
1443 case IP_RECVDSTADDR:
1444 optval = OPTBIT(INP_RECVDSTADDR);
1447 case IP_ORIGDSTADDR:
1448 optval = OPTBIT2(INP_ORIGDSTADDR);
1452 optval = OPTBIT(INP_RECVTTL);
1456 optval = OPTBIT(INP_RECVIF);
1460 if (inp->inp_flags & INP_HIGHPORT)
1461 optval = IP_PORTRANGE_HIGH;
1462 else if (inp->inp_flags & INP_LOWPORT)
1463 optval = IP_PORTRANGE_LOW;
1469 optval = OPTBIT(INP_ONESBCAST);
1472 optval = OPTBIT(INP_DONTFRAG);
1475 optval = OPTBIT(INP_BINDANY);
1478 optval = OPTBIT(INP_RECVTOS);
1481 optval = inp->inp_flowid;
1484 optval = inp->inp_flowtype;
1487 optval = OPTBIT2(INP_RECVFLOWID);
1490 case IP_RSSBUCKETID:
1491 retval = rss_hash2bucket(inp->inp_flowid,
1495 optval = rss_bucket;
1499 case IP_RECVRSSBUCKETID:
1500 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1504 optval = OPTBIT2(INP_BINDMULTI);
1507 error = sooptcopyout(sopt, &optval, sizeof optval);
1511 * Multicast socket options are processed by the in_mcast
1514 case IP_MULTICAST_IF:
1515 case IP_MULTICAST_VIF:
1516 case IP_MULTICAST_TTL:
1517 case IP_MULTICAST_LOOP:
1519 error = inp_getmoptions(inp, sopt);
1522 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1523 case IP_IPSEC_POLICY:
1524 if (IPSEC_ENABLED(ipv4)) {
1525 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1532 error = ENOPROTOOPT;
1541 * Routine called from ip_output() to loop back a copy of an IP multicast
1542 * packet to the input queue of a specified interface. Note that this
1543 * calls the output routine of the loopback "driver", but with an interface
1544 * pointer that might NOT be a loopback interface -- evil, but easier than
1545 * replicating that code here.
1548 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1554 * Make a deep copy of the packet because we're going to
1555 * modify the pack in order to generate checksums.
1557 copym = m_dup(m, M_NOWAIT);
1558 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1559 copym = m_pullup(copym, hlen);
1560 if (copym != NULL) {
1561 /* If needed, compute the checksum and mark it as valid. */
1562 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1563 in_delayed_cksum(copym);
1564 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1565 copym->m_pkthdr.csum_flags |=
1566 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1567 copym->m_pkthdr.csum_data = 0xffff;
1570 * We don't bother to fragment if the IP length is greater
1571 * than the interface's MTU. Can this possibly matter?
1573 ip = mtod(copym, struct ip *);
1575 ip->ip_sum = in_cksum(copym, hlen);
1576 if_simloop(ifp, copym, AF_INET, 0);