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, bool stamp_tag)
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;
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) {
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;
308 struct ifnet *ifp = NULL; /* keep compiler happy */
310 int hlen = sizeof (struct ip);
313 struct sockaddr_in *dst, sin;
314 const struct sockaddr_in *gw;
315 struct in_ifaddr *ia;
318 uint16_t ip_len, ip_off;
320 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
321 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;
380 * Validate route against routing table additions;
381 * a better/more specific route might have been added.
383 if (inp != NULL && ro != NULL && ro->ro_rt != NULL)
384 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
386 * If there is a cached route,
387 * check that it is to the same destination
388 * and is still up. If not, free it and try again.
389 * The address family should also be checked in case of sharing the
391 * Also check whether routing cache needs invalidation.
393 if (ro != NULL && ro->ro_rt != NULL &&
394 ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
395 ro->ro_rt->rt_ifp == NULL || !RT_LINK_IS_UP(ro->ro_rt->rt_ifp) ||
396 dst->sin_family != AF_INET ||
397 dst->sin_addr.s_addr != ip->ip_dst.s_addr))
398 RO_INVALIDATE_CACHE(ro);
401 * If routing to interface only, short circuit routing lookup.
402 * The use of an all-ones broadcast address implies this; an
403 * interface is specified by the broadcast address of an interface,
404 * or the destination address of a ptp interface.
406 if (flags & IP_SENDONES) {
407 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
408 M_GETFIB(m)))) == NULL &&
409 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
410 M_GETFIB(m)))) == NULL) {
411 IPSTAT_INC(ips_noroute);
415 ip->ip_dst.s_addr = INADDR_BROADCAST;
416 dst->sin_addr = ip->ip_dst;
421 src = IA_SIN(ia)->sin_addr;
422 } else if (flags & IP_ROUTETOIF) {
423 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
424 M_GETFIB(m)))) == NULL &&
425 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
426 M_GETFIB(m)))) == NULL) {
427 IPSTAT_INC(ips_noroute);
434 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
435 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
436 src = IA_SIN(ia)->sin_addr;
437 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
438 imo != NULL && imo->imo_multicast_ifp != NULL) {
440 * Bypass the normal routing lookup for multicast
441 * packets if the interface is specified.
443 ifp = imo->imo_multicast_ifp;
445 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
446 isbroadcast = 0; /* fool gcc */
447 /* Interface may have no addresses. */
449 src = IA_SIN(ia)->sin_addr;
451 src.s_addr = INADDR_ANY;
452 } else if (ro != NULL) {
453 if (ro->ro_rt == NULL) {
455 * We want to do any cloning requested by the link
456 * layer, as this is probably required in all cases
457 * for correct operation (as it is for ARP).
460 rtalloc_mpath_fib(ro,
461 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
464 in_rtalloc_ign(ro, 0, fibnum);
466 if (ro->ro_rt == NULL ||
467 (ro->ro_rt->rt_flags & RTF_UP) == 0 ||
468 ro->ro_rt->rt_ifp == NULL ||
469 !RT_LINK_IS_UP(ro->ro_rt->rt_ifp)) {
470 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
472 * There is no route for this packet, but it is
473 * possible that a matching SPD entry exists.
475 no_route_but_check_spd = 1;
476 mtu = 0; /* Silence GCC warning. */
479 IPSTAT_INC(ips_noroute);
480 error = EHOSTUNREACH;
484 ia = ifatoia(ro->ro_rt->rt_ifa);
485 ifp = ro->ro_rt->rt_ifp;
486 counter_u64_add(ro->ro_rt->rt_pksent, 1);
487 rt_update_ro_flags(ro);
488 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
489 gw = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
490 if (ro->ro_rt->rt_flags & RTF_HOST)
491 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
492 else if (ifp->if_flags & IFF_BROADCAST)
493 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
496 if (ro->ro_rt->rt_flags & RTF_HOST)
497 mtu = ro->ro_rt->rt_mtu;
500 src = IA_SIN(ia)->sin_addr;
502 struct nhop4_extended nh;
504 bzero(&nh, sizeof(nh));
505 if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh) !=
507 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
509 * There is no route for this packet, but it is
510 * possible that a matching SPD entry exists.
512 no_route_but_check_spd = 1;
513 mtu = 0; /* Silence GCC warning. */
516 IPSTAT_INC(ips_noroute);
517 error = EHOSTUNREACH;
523 * We are rewriting here dst to be gw actually, contradicting
524 * comment at the beginning of the function. However, in this
525 * case we are always dealing with on stack dst.
526 * In case if pfil(9) sends us back to beginning of the
527 * function, the dst would be rewritten by ip_output_pfil().
530 dst->sin_addr = nh.nh_addr;
533 isbroadcast = (((nh.nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
534 (NHF_HOST | NHF_BROADCAST)) ||
535 ((ifp->if_flags & IFF_BROADCAST) &&
536 in_ifaddr_broadcast(dst->sin_addr, ia)));
539 /* Catch a possible divide by zero later. */
540 KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (rt_flags=0x%08x) ifp=%p",
542 (ro != NULL && ro->ro_rt != NULL) ? ro->ro_rt->rt_flags : 0, ifp));
544 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
545 m->m_flags |= M_MCAST;
547 * IP destination address is multicast. Make sure "gw"
548 * still points to the address in "ro". (It may have been
549 * changed to point to a gateway address, above.)
553 * See if the caller provided any multicast options
556 ip->ip_ttl = imo->imo_multicast_ttl;
557 if (imo->imo_multicast_vif != -1)
560 ip_mcast_src(imo->imo_multicast_vif) :
563 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
565 * Confirm that the outgoing interface supports multicast.
567 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
568 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
569 IPSTAT_INC(ips_noroute);
575 * If source address not specified yet, use address
576 * of outgoing interface.
578 if (ip->ip_src.s_addr == INADDR_ANY)
581 if ((imo == NULL && in_mcast_loop) ||
582 (imo && imo->imo_multicast_loop)) {
584 * Loop back multicast datagram if not expressly
585 * forbidden to do so, even if we are not a member
586 * of the group; ip_input() will filter it later,
587 * thus deferring a hash lookup and mutex acquisition
588 * at the expense of a cheap copy using m_copym().
590 ip_mloopback(ifp, m, hlen);
593 * If we are acting as a multicast router, perform
594 * multicast forwarding as if the packet had just
595 * arrived on the interface to which we are about
596 * to send. The multicast forwarding function
597 * recursively calls this function, using the
598 * IP_FORWARDING flag to prevent infinite recursion.
600 * Multicasts that are looped back by ip_mloopback(),
601 * above, will be forwarded by the ip_input() routine,
604 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
606 * If rsvp daemon is not running, do not
607 * set ip_moptions. This ensures that the packet
608 * is multicast and not just sent down one link
609 * as prescribed by rsvpd.
614 ip_mforward(ip, ifp, m, imo) != 0) {
622 * Multicasts with a time-to-live of zero may be looped-
623 * back, above, but must not be transmitted on a network.
624 * Also, multicasts addressed to the loopback interface
625 * are not sent -- the above call to ip_mloopback() will
626 * loop back a copy. ip_input() will drop the copy if
627 * this host does not belong to the destination group on
628 * the loopback interface.
630 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
639 * If the source address is not specified yet, use the address
640 * of the outoing interface.
642 if (ip->ip_src.s_addr == INADDR_ANY)
646 * Look for broadcast address and
647 * verify user is allowed to send
651 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
652 error = EADDRNOTAVAIL;
655 if ((flags & IP_ALLOWBROADCAST) == 0) {
659 /* don't allow broadcast messages to be fragmented */
664 m->m_flags |= M_BCAST;
666 m->m_flags &= ~M_BCAST;
670 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
671 if (IPSEC_ENABLED(ipv4)) {
672 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
673 if (error == EINPROGRESS)
679 * Check if there was a route for this packet; return error if not.
681 if (no_route_but_check_spd) {
682 IPSTAT_INC(ips_noroute);
683 error = EHOSTUNREACH;
686 /* Update variables that are affected by ipsec4_output(). */
687 ip = mtod(m, struct ip *);
688 hlen = ip->ip_hl << 2;
691 /* Jump over all PFIL processing if hooks are not active. */
692 if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
693 switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
695 case 1: /* Finished */
698 case 0: /* Continue normally */
699 ip = mtod(m, struct ip *);
702 case -1: /* Need to try again */
703 /* Reset everything for a new round */
706 ro->ro_prepend = NULL;
709 ip = mtod(m, struct ip *);
715 /* IN_LOOPBACK must not appear on the wire - RFC1122. */
716 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
717 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
718 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
719 IPSTAT_INC(ips_badaddr);
720 error = EADDRNOTAVAIL;
725 m->m_pkthdr.csum_flags |= CSUM_IP;
726 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
727 m = mb_unmapped_to_ext(m);
729 IPSTAT_INC(ips_odropped);
734 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
735 } else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
736 m = mb_unmapped_to_ext(m);
738 IPSTAT_INC(ips_odropped);
744 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
745 m = mb_unmapped_to_ext(m);
747 IPSTAT_INC(ips_odropped);
751 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
752 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
757 * If small enough for interface, or the interface will take
758 * care of the fragmentation for us, we can just send directly.
761 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_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 & CSUM_TSO)
776 counter_u64_add(ia->ia_ifa.ifa_opackets,
777 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
779 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
781 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
783 #ifdef MBUF_STRESS_TEST
784 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
785 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
788 * Reset layer specific mbuf flags
789 * to avoid confusing lower layers.
792 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
793 error = ip_output_send(inp, ifp, m, gw, ro,
794 (flags & IP_NO_SND_TAG_RL) ? false : true);
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, true);
836 IPSTAT_INC(ips_fragmented);
846 * Create a chain of fragments which fit the given mtu. m_frag points to the
847 * mbuf to be fragmented; on return it points to the chain with the fragments.
848 * Return 0 if no error. If error, m_frag may contain a partially built
849 * chain of fragments that should be freed by the caller.
851 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
854 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
855 u_long if_hwassist_flags)
858 int hlen = ip->ip_hl << 2;
859 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
861 struct mbuf *m0 = *m_frag; /* the original packet */
865 uint16_t ip_len, ip_off;
867 ip_len = ntohs(ip->ip_len);
868 ip_off = ntohs(ip->ip_off);
870 if (ip_off & IP_DF) { /* Fragmentation not allowed */
871 IPSTAT_INC(ips_cantfrag);
876 * Must be able to put at least 8 bytes per fragment.
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 m0 = mb_unmapped_to_ext(m0);
889 IPSTAT_INC(ips_odropped);
892 in_delayed_cksum(m0);
893 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
896 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
897 m0 = mb_unmapped_to_ext(m0);
900 IPSTAT_INC(ips_odropped);
903 sctp_delayed_cksum(m0, hlen);
904 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
907 if (len > PAGE_SIZE) {
909 * Fragment large datagrams such that each segment
910 * contains a multiple of PAGE_SIZE amount of data,
911 * plus headers. This enables a receiver to perform
912 * page-flipping zero-copy optimizations.
914 * XXX When does this help given that sender and receiver
915 * could have different page sizes, and also mtu could
916 * be less than the receiver's page size ?
920 off = MIN(mtu, m0->m_pkthdr.len);
923 * firstlen (off - hlen) must be aligned on an
927 goto smart_frag_failure;
928 off = ((off - hlen) & ~7) + hlen;
929 newlen = (~PAGE_MASK) & mtu;
930 if ((newlen + sizeof (struct ip)) > mtu) {
931 /* we failed, go back the default */
942 firstlen = off - hlen;
943 mnext = &m0->m_nextpkt; /* pointer to next packet */
946 * Loop through length of segment after first fragment,
947 * make new header and copy data of each part and link onto chain.
948 * Here, m0 is the original packet, m is the fragment being created.
949 * The fragments are linked off the m_nextpkt of the original
950 * packet, which after processing serves as the first fragment.
952 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
953 struct ip *mhip; /* ip header on the fragment */
955 int mhlen = sizeof (struct ip);
957 m = m_gethdr(M_NOWAIT, MT_DATA);
960 IPSTAT_INC(ips_odropped);
964 * Make sure the complete packet header gets copied
965 * from the originating mbuf to the newly created
966 * mbuf. This also ensures that existing firewall
967 * classification(s), VLAN tags and so on get copied
968 * to the resulting fragmented packet(s):
970 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
973 IPSTAT_INC(ips_odropped);
977 * In the first mbuf, leave room for the link header, then
978 * copy the original IP header including options. The payload
979 * goes into an additional mbuf chain returned by m_copym().
981 m->m_data += max_linkhdr;
982 mhip = mtod(m, struct ip *);
984 if (hlen > sizeof (struct ip)) {
985 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
986 mhip->ip_v = IPVERSION;
987 mhip->ip_hl = mhlen >> 2;
990 /* XXX do we need to add ip_off below ? */
991 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
992 if (off + len >= ip_len)
995 mhip->ip_off |= IP_MF;
996 mhip->ip_len = htons((u_short)(len + mhlen));
997 m->m_next = m_copym(m0, off, len, M_NOWAIT);
998 if (m->m_next == NULL) { /* copy failed */
1000 error = ENOBUFS; /* ??? */
1001 IPSTAT_INC(ips_odropped);
1004 m->m_pkthdr.len = mhlen + len;
1006 mac_netinet_fragment(m0, m);
1008 mhip->ip_off = htons(mhip->ip_off);
1010 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1011 mhip->ip_sum = in_cksum(m, mhlen);
1012 m->m_pkthdr.csum_flags &= ~CSUM_IP;
1015 mnext = &m->m_nextpkt;
1017 IPSTAT_ADD(ips_ofragments, nfrags);
1020 * Update first fragment by trimming what's been copied out
1021 * and updating header.
1023 m_adj(m0, hlen + firstlen - ip_len);
1024 m0->m_pkthdr.len = hlen + firstlen;
1025 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1026 ip->ip_off = htons(ip_off | IP_MF);
1028 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1029 ip->ip_sum = in_cksum(m0, hlen);
1030 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1039 in_delayed_cksum(struct mbuf *m)
1043 uint16_t cklen, csum, offset;
1045 ip = mtod(m, struct ip *);
1046 offset = ip->ip_hl << 2 ;
1048 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1049 /* if udp header is not in the first mbuf copy udplen */
1050 if (offset + sizeof(struct udphdr) > m->m_len) {
1051 m_copydata(m, offset + offsetof(struct udphdr,
1052 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1053 cklen = ntohs(cklen);
1055 uh = (struct udphdr *)mtodo(m, offset);
1056 cklen = ntohs(uh->uh_ulen);
1058 csum = in_cksum_skip(m, cklen + offset, offset);
1062 cklen = ntohs(ip->ip_len);
1063 csum = in_cksum_skip(m, cklen, offset);
1065 offset += m->m_pkthdr.csum_data; /* checksum offset */
1067 if (offset + sizeof(csum) > m->m_len)
1068 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1070 *(u_short *)mtodo(m, offset) = csum;
1074 * IP socket option processing.
1077 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1079 struct inpcb *inp = sotoinpcb(so);
1082 uint32_t rss_bucket;
1087 if (sopt->sopt_level != IPPROTO_IP) {
1090 if (sopt->sopt_level == SOL_SOCKET &&
1091 sopt->sopt_dir == SOPT_SET) {
1092 switch (sopt->sopt_name) {
1095 if ((so->so_options & SO_REUSEADDR) != 0)
1096 inp->inp_flags2 |= INP_REUSEADDR;
1098 inp->inp_flags2 &= ~INP_REUSEADDR;
1104 if ((so->so_options & SO_REUSEPORT) != 0)
1105 inp->inp_flags2 |= INP_REUSEPORT;
1107 inp->inp_flags2 &= ~INP_REUSEPORT;
1111 case SO_REUSEPORT_LB:
1113 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1114 inp->inp_flags2 |= INP_REUSEPORT_LB;
1116 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1122 inp->inp_inc.inc_fibnum = so->so_fibnum;
1126 case SO_MAX_PACING_RATE:
1129 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1143 switch (sopt->sopt_dir) {
1145 switch (sopt->sopt_name) {
1152 if (sopt->sopt_valsize > MLEN) {
1156 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1161 m->m_len = sopt->sopt_valsize;
1162 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1169 error = ip_pcbopts(inp, sopt->sopt_name, m);
1175 if (sopt->sopt_td != NULL) {
1176 error = priv_check(sopt->sopt_td,
1177 PRIV_NETINET_BINDANY);
1184 case IP_RSS_LISTEN_BUCKET:
1190 case IP_RECVRETOPTS:
1191 case IP_ORIGDSTADDR:
1192 case IP_RECVDSTADDR:
1200 case IP_RECVRSSBUCKETID:
1202 error = sooptcopyin(sopt, &optval, sizeof optval,
1207 switch (sopt->sopt_name) {
1209 inp->inp_ip_tos = optval;
1213 inp->inp_ip_ttl = optval;
1217 if (optval >= 0 && optval <= MAXTTL)
1218 inp->inp_ip_minttl = optval;
1223 #define OPTSET(bit) do { \
1226 inp->inp_flags |= bit; \
1228 inp->inp_flags &= ~bit; \
1232 #define OPTSET2(bit, val) do { \
1235 inp->inp_flags2 |= bit; \
1237 inp->inp_flags2 &= ~bit; \
1242 OPTSET(INP_RECVOPTS);
1245 case IP_RECVRETOPTS:
1246 OPTSET(INP_RECVRETOPTS);
1249 case IP_RECVDSTADDR:
1250 OPTSET(INP_RECVDSTADDR);
1253 case IP_ORIGDSTADDR:
1254 OPTSET2(INP_ORIGDSTADDR, optval);
1258 OPTSET(INP_RECVTTL);
1266 OPTSET(INP_ONESBCAST);
1269 OPTSET(INP_DONTFRAG);
1272 OPTSET(INP_BINDANY);
1275 OPTSET(INP_RECVTOS);
1278 OPTSET2(INP_BINDMULTI, optval);
1281 OPTSET2(INP_RECVFLOWID, optval);
1284 case IP_RSS_LISTEN_BUCKET:
1285 if ((optval >= 0) &&
1286 (optval < rss_getnumbuckets())) {
1287 inp->inp_rss_listen_bucket = optval;
1288 OPTSET2(INP_RSS_BUCKET_SET, 1);
1293 case IP_RECVRSSBUCKETID:
1294 OPTSET2(INP_RECVRSSBUCKETID, optval);
1303 * Multicast socket options are processed by the in_mcast
1306 case IP_MULTICAST_IF:
1307 case IP_MULTICAST_VIF:
1308 case IP_MULTICAST_TTL:
1309 case IP_MULTICAST_LOOP:
1310 case IP_ADD_MEMBERSHIP:
1311 case IP_DROP_MEMBERSHIP:
1312 case IP_ADD_SOURCE_MEMBERSHIP:
1313 case IP_DROP_SOURCE_MEMBERSHIP:
1314 case IP_BLOCK_SOURCE:
1315 case IP_UNBLOCK_SOURCE:
1317 case MCAST_JOIN_GROUP:
1318 case MCAST_LEAVE_GROUP:
1319 case MCAST_JOIN_SOURCE_GROUP:
1320 case MCAST_LEAVE_SOURCE_GROUP:
1321 case MCAST_BLOCK_SOURCE:
1322 case MCAST_UNBLOCK_SOURCE:
1323 error = inp_setmoptions(inp, sopt);
1327 error = sooptcopyin(sopt, &optval, sizeof optval,
1334 case IP_PORTRANGE_DEFAULT:
1335 inp->inp_flags &= ~(INP_LOWPORT);
1336 inp->inp_flags &= ~(INP_HIGHPORT);
1339 case IP_PORTRANGE_HIGH:
1340 inp->inp_flags &= ~(INP_LOWPORT);
1341 inp->inp_flags |= INP_HIGHPORT;
1344 case IP_PORTRANGE_LOW:
1345 inp->inp_flags &= ~(INP_HIGHPORT);
1346 inp->inp_flags |= INP_LOWPORT;
1356 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1357 case IP_IPSEC_POLICY:
1358 if (IPSEC_ENABLED(ipv4)) {
1359 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1366 error = ENOPROTOOPT;
1372 switch (sopt->sopt_name) {
1376 if (inp->inp_options) {
1377 struct mbuf *options;
1379 options = m_copym(inp->inp_options, 0,
1380 M_COPYALL, M_NOWAIT);
1382 if (options != NULL) {
1383 error = sooptcopyout(sopt,
1384 mtod(options, char *),
1391 sopt->sopt_valsize = 0;
1399 case IP_RECVRETOPTS:
1400 case IP_ORIGDSTADDR:
1401 case IP_RECVDSTADDR:
1414 case IP_RSSBUCKETID:
1415 case IP_RECVRSSBUCKETID:
1417 switch (sopt->sopt_name) {
1420 optval = inp->inp_ip_tos;
1424 optval = inp->inp_ip_ttl;
1428 optval = inp->inp_ip_minttl;
1431 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1432 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1435 optval = OPTBIT(INP_RECVOPTS);
1438 case IP_RECVRETOPTS:
1439 optval = OPTBIT(INP_RECVRETOPTS);
1442 case IP_RECVDSTADDR:
1443 optval = OPTBIT(INP_RECVDSTADDR);
1446 case IP_ORIGDSTADDR:
1447 optval = OPTBIT2(INP_ORIGDSTADDR);
1451 optval = OPTBIT(INP_RECVTTL);
1455 optval = OPTBIT(INP_RECVIF);
1459 if (inp->inp_flags & INP_HIGHPORT)
1460 optval = IP_PORTRANGE_HIGH;
1461 else if (inp->inp_flags & INP_LOWPORT)
1462 optval = IP_PORTRANGE_LOW;
1468 optval = OPTBIT(INP_ONESBCAST);
1471 optval = OPTBIT(INP_DONTFRAG);
1474 optval = OPTBIT(INP_BINDANY);
1477 optval = OPTBIT(INP_RECVTOS);
1480 optval = inp->inp_flowid;
1483 optval = inp->inp_flowtype;
1486 optval = OPTBIT2(INP_RECVFLOWID);
1489 case IP_RSSBUCKETID:
1490 retval = rss_hash2bucket(inp->inp_flowid,
1494 optval = rss_bucket;
1498 case IP_RECVRSSBUCKETID:
1499 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1503 optval = OPTBIT2(INP_BINDMULTI);
1506 error = sooptcopyout(sopt, &optval, sizeof optval);
1510 * Multicast socket options are processed by the in_mcast
1513 case IP_MULTICAST_IF:
1514 case IP_MULTICAST_VIF:
1515 case IP_MULTICAST_TTL:
1516 case IP_MULTICAST_LOOP:
1518 error = inp_getmoptions(inp, sopt);
1521 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1522 case IP_IPSEC_POLICY:
1523 if (IPSEC_ENABLED(ipv4)) {
1524 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1531 error = ENOPROTOOPT;
1540 * Routine called from ip_output() to loop back a copy of an IP multicast
1541 * packet to the input queue of a specified interface. Note that this
1542 * calls the output routine of the loopback "driver", but with an interface
1543 * pointer that might NOT be a loopback interface -- evil, but easier than
1544 * replicating that code here.
1547 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1553 * Make a deep copy of the packet because we're going to
1554 * modify the pack in order to generate checksums.
1556 copym = m_dup(m, M_NOWAIT);
1557 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1558 copym = m_pullup(copym, hlen);
1559 if (copym != NULL) {
1560 /* If needed, compute the checksum and mark it as valid. */
1561 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1562 in_delayed_cksum(copym);
1563 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1564 copym->m_pkthdr.csum_flags |=
1565 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1566 copym->m_pkthdr.csum_data = 0xffff;
1569 * We don't bother to fragment if the IP length is greater
1570 * than the interface's MTU. Can this possibly matter?
1572 ip = mtod(copym, struct ip *);
1574 ip->ip_sum = in_cksum(copym, hlen);
1575 if_simloop(ifp, copym, AF_INET, 0);