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>
70 #include <net/route/nhop.h>
71 #include <net/rss_config.h>
74 #include <netinet/in.h>
75 #include <netinet/in_fib.h>
76 #include <netinet/in_kdtrace.h>
77 #include <netinet/in_systm.h>
78 #include <netinet/ip.h>
79 #include <netinet/in_fib.h>
80 #include <netinet/in_pcb.h>
81 #include <netinet/in_rss.h>
82 #include <netinet/in_var.h>
83 #include <netinet/ip_var.h>
84 #include <netinet/ip_options.h>
86 #include <netinet/udp.h>
87 #include <netinet/udp_var.h>
89 #if defined(SCTP) || defined(SCTP_SUPPORT)
90 #include <netinet/sctp.h>
91 #include <netinet/sctp_crc32.h>
94 #include <netipsec/ipsec_support.h>
96 #include <machine/in_cksum.h>
98 #include <security/mac/mac_framework.h>
100 #ifdef MBUF_STRESS_TEST
101 static int mbuf_frag_size = 0;
102 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
103 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
106 static void ip_mloopback(struct ifnet *, const struct mbuf *, int);
109 extern int in_mcast_loop;
110 extern struct protosw inetsw[];
113 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
114 struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
116 struct m_tag *fwd_tag = NULL;
120 int pflags = PFIL_OUT;
122 if (flags & IP_FORWARDING)
126 ip = mtod(m, struct ip *);
128 /* Run through list of hooks for output packets. */
129 odst.s_addr = ip->ip_dst.s_addr;
130 switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
135 return 1; /* Finished */
140 ip = mtod(m, struct ip *);
142 /* See if destination IP address was changed by packet filter. */
143 if (odst.s_addr != ip->ip_dst.s_addr) {
144 m->m_flags |= M_SKIP_FIREWALL;
145 /* If destination is now ourself drop to ip_input(). */
146 if (in_localip(ip->ip_dst)) {
147 m->m_flags |= M_FASTFWD_OURS;
148 if (m->m_pkthdr.rcvif == NULL)
149 m->m_pkthdr.rcvif = V_loif;
150 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
151 m->m_pkthdr.csum_flags |=
152 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
153 m->m_pkthdr.csum_data = 0xffff;
155 m->m_pkthdr.csum_flags |=
156 CSUM_IP_CHECKED | CSUM_IP_VALID;
157 #if defined(SCTP) || defined(SCTP_SUPPORT)
158 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
159 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
161 *error = netisr_queue(NETISR_IP, m);
162 return 1; /* Finished */
165 bzero(dst, sizeof(*dst));
166 dst->sin_family = AF_INET;
167 dst->sin_len = sizeof(*dst);
168 dst->sin_addr = ip->ip_dst;
170 return -1; /* Reloop */
172 /* See if fib was changed by packet filter. */
173 if ((*fibnum) != M_GETFIB(m)) {
174 m->m_flags |= M_SKIP_FIREWALL;
175 *fibnum = M_GETFIB(m);
176 return -1; /* Reloop for FIB change */
179 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
180 if (m->m_flags & M_FASTFWD_OURS) {
181 if (m->m_pkthdr.rcvif == NULL)
182 m->m_pkthdr.rcvif = V_loif;
183 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
184 m->m_pkthdr.csum_flags |=
185 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
186 m->m_pkthdr.csum_data = 0xffff;
188 #if defined(SCTP) || defined(SCTP_SUPPORT)
189 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
190 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
192 m->m_pkthdr.csum_flags |=
193 CSUM_IP_CHECKED | CSUM_IP_VALID;
195 *error = netisr_queue(NETISR_IP, m);
196 return 1; /* Finished */
198 /* Or forward to some other address? */
199 if ((m->m_flags & M_IP_NEXTHOP) &&
200 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
201 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
202 m->m_flags |= M_SKIP_FIREWALL;
203 m->m_flags &= ~M_IP_NEXTHOP;
204 m_tag_delete(m, fwd_tag);
206 return -1; /* Reloop for CHANGE of dst */
213 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
214 const struct sockaddr_in *gw, struct route *ro, bool stamp_tag)
217 struct ktls_session *tls = NULL;
219 struct m_snd_tag *mst;
222 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
227 * If this is an unencrypted TLS record, save a reference to
228 * the record. This local reference is used to call
229 * ktls_output_eagain after the mbuf has been freed (thus
230 * dropping the mbuf's reference) in if_output.
232 if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
233 tls = ktls_hold(m->m_next->m_epg_tls);
237 * If a TLS session doesn't have a valid tag, it must
238 * have had an earlier ifp mismatch, so drop this
246 * Always stamp tags that include NIC ktls.
252 if (inp != NULL && mst == NULL) {
253 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
254 (inp->inp_snd_tag != NULL &&
255 inp->inp_snd_tag->ifp != ifp))
256 in_pcboutput_txrtlmt(inp, ifp, m);
258 if (inp->inp_snd_tag != NULL)
259 mst = inp->inp_snd_tag;
262 if (stamp_tag && mst != NULL) {
263 KASSERT(m->m_pkthdr.rcvif == NULL,
264 ("trying to add a send tag to a forwarded packet"));
265 if (mst->ifp != ifp) {
270 /* stamp send tag on mbuf */
271 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
272 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
275 error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
278 /* Check for route change invalidating send tags. */
282 error = ktls_output_eagain(inp, tls);
288 in_pcboutput_eagain(inp);
293 /* rte<>ro_flags translation */
295 rt_update_ro_flags(struct route *ro)
297 int nh_flags = ro->ro_nh->nh_flags;
299 ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
301 ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
302 ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
303 ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
307 * IP output. The packet in mbuf chain m contains a skeletal IP
308 * header (with len, off, ttl, proto, tos, src, dst).
309 * The mbuf chain containing the packet will be freed.
310 * The mbuf opt, if present, will not be freed.
311 * If route ro is present and has ro_rt initialized, route lookup would be
312 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
313 * then result of route lookup is stored in ro->ro_rt.
315 * In the IP forwarding case, the packet will arrive with options already
316 * inserted, so must have a NULL opt pointer.
319 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
320 struct ip_moptions *imo, struct inpcb *inp)
322 struct rm_priotracker in_ifa_tracker;
324 struct ifnet *ifp = NULL; /* keep compiler happy */
326 int hlen = sizeof (struct ip);
329 struct sockaddr_in *dst, sin;
330 const struct sockaddr_in *gw;
331 struct in_ifaddr *ia;
334 uint16_t ip_len, ip_off;
336 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
337 int no_route_but_check_spd = 0;
344 INP_LOCK_ASSERT(inp);
345 M_SETFIB(m, inp->inp_inc.inc_fibnum);
346 if ((flags & IP_NODEFAULTFLOWID) == 0) {
347 m->m_pkthdr.flowid = inp->inp_flowid;
348 M_HASHTYPE_SET(m, inp->inp_flowtype);
351 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
357 m = ip_insertoptions(m, opt, &len);
359 hlen = len; /* ip->ip_hl is updated above */
361 ip = mtod(m, struct ip *);
362 ip_len = ntohs(ip->ip_len);
363 ip_off = ntohs(ip->ip_off);
365 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
366 ip->ip_v = IPVERSION;
367 ip->ip_hl = hlen >> 2;
370 /* Header already set, fetch hlen from there */
371 hlen = ip->ip_hl << 2;
373 if ((flags & IP_FORWARDING) == 0)
374 IPSTAT_INC(ips_localout);
379 * gw is readonly but can point either to dst OR rt_gateway,
380 * therefore we need restore gw if we're redoing lookup.
382 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
384 dst = (struct sockaddr_in *)&ro->ro_dst;
387 if (ro == NULL || ro->ro_nh == NULL) {
388 bzero(dst, sizeof(*dst));
389 dst->sin_family = AF_INET;
390 dst->sin_len = sizeof(*dst);
391 dst->sin_addr = ip->ip_dst;
396 * Validate route against routing table additions;
397 * a better/more specific route might have been added.
399 if (inp != NULL && ro != NULL && ro->ro_nh != NULL)
400 NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
402 * If there is a cached route,
403 * check that it is to the same destination
404 * and is still up. If not, free it and try again.
405 * The address family should also be checked in case of sharing the
407 * Also check whether routing cache needs invalidation.
409 if (ro != NULL && ro->ro_nh != NULL &&
410 ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
411 dst->sin_addr.s_addr != ip->ip_dst.s_addr))
412 RO_INVALIDATE_CACHE(ro);
415 * If routing to interface only, short circuit routing lookup.
416 * The use of an all-ones broadcast address implies this; an
417 * interface is specified by the broadcast address of an interface,
418 * or the destination address of a ptp interface.
420 if (flags & IP_SENDONES) {
421 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
422 M_GETFIB(m)))) == NULL &&
423 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
424 M_GETFIB(m)))) == NULL) {
425 IPSTAT_INC(ips_noroute);
429 ip->ip_dst.s_addr = INADDR_BROADCAST;
430 dst->sin_addr = ip->ip_dst;
435 src = IA_SIN(ia)->sin_addr;
436 } else if (flags & IP_ROUTETOIF) {
437 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
438 M_GETFIB(m)))) == NULL &&
439 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
440 M_GETFIB(m)))) == NULL) {
441 IPSTAT_INC(ips_noroute);
448 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
449 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
450 src = IA_SIN(ia)->sin_addr;
451 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
452 imo != NULL && imo->imo_multicast_ifp != NULL) {
454 * Bypass the normal routing lookup for multicast
455 * packets if the interface is specified.
457 ifp = imo->imo_multicast_ifp;
459 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
460 isbroadcast = 0; /* fool gcc */
461 /* Interface may have no addresses. */
463 src = IA_SIN(ia)->sin_addr;
465 src.s_addr = INADDR_ANY;
466 } else if (ro != NULL) {
467 if (ro->ro_nh == NULL) {
469 * We want to do any cloning requested by the link
470 * layer, as this is probably required in all cases
471 * for correct operation (as it is for ARP).
475 flowid = ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr);
477 flowid = m->m_pkthdr.flowid;
479 ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
482 if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
483 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
485 * There is no route for this packet, but it is
486 * possible that a matching SPD entry exists.
488 no_route_but_check_spd = 1;
489 mtu = 0; /* Silence GCC warning. */
492 IPSTAT_INC(ips_noroute);
493 error = EHOSTUNREACH;
497 ia = ifatoia(ro->ro_nh->nh_ifa);
498 ifp = ro->ro_nh->nh_ifp;
499 counter_u64_add(ro->ro_nh->nh_pksent, 1);
500 rt_update_ro_flags(ro);
501 if (ro->ro_nh->nh_flags & NHF_GATEWAY)
502 gw = &ro->ro_nh->gw4_sa;
503 if (ro->ro_nh->nh_flags & NHF_HOST)
504 isbroadcast = (ro->ro_nh->nh_flags & NHF_BROADCAST);
505 else if (ifp->if_flags & IFF_BROADCAST)
506 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
509 if (ro->ro_nh->nh_flags & NHF_HOST)
510 mtu = ro->ro_nh->nh_mtu;
513 src = IA_SIN(ia)->sin_addr;
515 struct nhop_object *nh;
517 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0);
519 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
521 * There is no route for this packet, but it is
522 * possible that a matching SPD entry exists.
524 no_route_but_check_spd = 1;
525 mtu = 0; /* Silence GCC warning. */
528 IPSTAT_INC(ips_noroute);
529 error = EHOSTUNREACH;
535 * We are rewriting here dst to be gw actually, contradicting
536 * comment at the beginning of the function. However, in this
537 * case we are always dealing with on stack dst.
538 * In case if pfil(9) sends us back to beginning of the
539 * function, the dst would be rewritten by ip_output_pfil().
542 if (nh->nh_flags & NHF_GATEWAY)
543 dst->sin_addr = nh->gw4_sa.sin_addr;
544 ia = ifatoia(nh->nh_ifa);
545 src = IA_SIN(ia)->sin_addr;
546 isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
547 (NHF_HOST | NHF_BROADCAST)) ||
548 ((ifp->if_flags & IFF_BROADCAST) &&
549 in_ifaddr_broadcast(dst->sin_addr, ia)));
552 /* Catch a possible divide by zero later. */
553 KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
555 (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
557 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
558 m->m_flags |= M_MCAST;
560 * IP destination address is multicast. Make sure "gw"
561 * still points to the address in "ro". (It may have been
562 * changed to point to a gateway address, above.)
566 * See if the caller provided any multicast options
569 ip->ip_ttl = imo->imo_multicast_ttl;
570 if (imo->imo_multicast_vif != -1)
573 ip_mcast_src(imo->imo_multicast_vif) :
576 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
578 * Confirm that the outgoing interface supports multicast.
580 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
581 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
582 IPSTAT_INC(ips_noroute);
588 * If source address not specified yet, use address
589 * of outgoing interface.
591 if (ip->ip_src.s_addr == INADDR_ANY)
594 if ((imo == NULL && in_mcast_loop) ||
595 (imo && imo->imo_multicast_loop)) {
597 * Loop back multicast datagram if not expressly
598 * forbidden to do so, even if we are not a member
599 * of the group; ip_input() will filter it later,
600 * thus deferring a hash lookup and mutex acquisition
601 * at the expense of a cheap copy using m_copym().
603 ip_mloopback(ifp, m, hlen);
606 * If we are acting as a multicast router, perform
607 * multicast forwarding as if the packet had just
608 * arrived on the interface to which we are about
609 * to send. The multicast forwarding function
610 * recursively calls this function, using the
611 * IP_FORWARDING flag to prevent infinite recursion.
613 * Multicasts that are looped back by ip_mloopback(),
614 * above, will be forwarded by the ip_input() routine,
617 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
619 * If rsvp daemon is not running, do not
620 * set ip_moptions. This ensures that the packet
621 * is multicast and not just sent down one link
622 * as prescribed by rsvpd.
627 ip_mforward(ip, ifp, m, imo) != 0) {
635 * Multicasts with a time-to-live of zero may be looped-
636 * back, above, but must not be transmitted on a network.
637 * Also, multicasts addressed to the loopback interface
638 * are not sent -- the above call to ip_mloopback() will
639 * loop back a copy. ip_input() will drop the copy if
640 * this host does not belong to the destination group on
641 * the loopback interface.
643 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
652 * If the source address is not specified yet, use the address
653 * of the outoing interface.
655 if (ip->ip_src.s_addr == INADDR_ANY)
659 * Look for broadcast address and
660 * verify user is allowed to send
664 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
665 error = EADDRNOTAVAIL;
668 if ((flags & IP_ALLOWBROADCAST) == 0) {
672 /* don't allow broadcast messages to be fragmented */
677 m->m_flags |= M_BCAST;
679 m->m_flags &= ~M_BCAST;
683 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
684 if (IPSEC_ENABLED(ipv4)) {
685 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
686 if (error == EINPROGRESS)
692 * Check if there was a route for this packet; return error if not.
694 if (no_route_but_check_spd) {
695 IPSTAT_INC(ips_noroute);
696 error = EHOSTUNREACH;
699 /* Update variables that are affected by ipsec4_output(). */
700 ip = mtod(m, struct ip *);
701 hlen = ip->ip_hl << 2;
704 /* Jump over all PFIL processing if hooks are not active. */
705 if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
706 switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
708 case 1: /* Finished */
711 case 0: /* Continue normally */
712 ip = mtod(m, struct ip *);
715 case -1: /* Need to try again */
716 /* Reset everything for a new round */
719 ro->ro_prepend = NULL;
722 ip = mtod(m, struct ip *);
728 /* IN_LOOPBACK must not appear on the wire - RFC1122. */
729 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
730 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
731 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
732 IPSTAT_INC(ips_badaddr);
733 error = EADDRNOTAVAIL;
738 m->m_pkthdr.csum_flags |= CSUM_IP;
739 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
740 m = mb_unmapped_to_ext(m);
742 IPSTAT_INC(ips_odropped);
747 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
748 } else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
749 m = mb_unmapped_to_ext(m);
751 IPSTAT_INC(ips_odropped);
756 #if defined(SCTP) || defined(SCTP_SUPPORT)
757 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
758 m = mb_unmapped_to_ext(m);
760 IPSTAT_INC(ips_odropped);
764 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
765 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
770 * If small enough for interface, or the interface will take
771 * care of the fragmentation for us, we can just send directly.
774 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
776 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
777 ip->ip_sum = in_cksum(m, hlen);
778 m->m_pkthdr.csum_flags &= ~CSUM_IP;
782 * Record statistics for this interface address.
783 * With CSUM_TSO the byte/packet count will be slightly
784 * incorrect because we count the IP+TCP headers only
785 * once instead of for every generated packet.
787 if (!(flags & IP_FORWARDING) && ia) {
788 if (m->m_pkthdr.csum_flags & CSUM_TSO)
789 counter_u64_add(ia->ia_ifa.ifa_opackets,
790 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
792 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
794 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
796 #ifdef MBUF_STRESS_TEST
797 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
798 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
801 * Reset layer specific mbuf flags
802 * to avoid confusing lower layers.
805 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
806 error = ip_output_send(inp, ifp, m, gw, ro,
807 (flags & IP_NO_SND_TAG_RL) ? false : true);
811 /* Balk when DF bit is set or the interface didn't support TSO. */
812 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
814 IPSTAT_INC(ips_cantfrag);
819 * Too large for interface; fragment if possible. If successful,
820 * on return, m will point to a list of packets to be sent.
822 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
829 /* Record statistics for this interface address. */
831 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
832 counter_u64_add(ia->ia_ifa.ifa_obytes,
836 * Reset layer specific mbuf flags
837 * to avoid confusing upper layers.
841 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
842 mtod(m, struct ip *), NULL);
843 error = ip_output_send(inp, ifp, m, gw, ro, true);
849 IPSTAT_INC(ips_fragmented);
859 * Create a chain of fragments which fit the given mtu. m_frag points to the
860 * mbuf to be fragmented; on return it points to the chain with the fragments.
861 * Return 0 if no error. If error, m_frag may contain a partially built
862 * chain of fragments that should be freed by the caller.
864 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
867 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
868 u_long if_hwassist_flags)
871 int hlen = ip->ip_hl << 2;
872 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
874 struct mbuf *m0 = *m_frag; /* the original packet */
878 uint16_t ip_len, ip_off;
880 ip_len = ntohs(ip->ip_len);
881 ip_off = ntohs(ip->ip_off);
883 if (ip_off & IP_DF) { /* Fragmentation not allowed */
884 IPSTAT_INC(ips_cantfrag);
889 * Must be able to put at least 8 bytes per fragment.
895 * If the interface will not calculate checksums on
896 * fragmented packets, then do it here.
898 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
899 m0 = mb_unmapped_to_ext(m0);
902 IPSTAT_INC(ips_odropped);
905 in_delayed_cksum(m0);
906 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
908 #if defined(SCTP) || defined(SCTP_SUPPORT)
909 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
910 m0 = mb_unmapped_to_ext(m0);
913 IPSTAT_INC(ips_odropped);
916 sctp_delayed_cksum(m0, hlen);
917 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
920 if (len > PAGE_SIZE) {
922 * Fragment large datagrams such that each segment
923 * contains a multiple of PAGE_SIZE amount of data,
924 * plus headers. This enables a receiver to perform
925 * page-flipping zero-copy optimizations.
927 * XXX When does this help given that sender and receiver
928 * could have different page sizes, and also mtu could
929 * be less than the receiver's page size ?
933 off = MIN(mtu, m0->m_pkthdr.len);
936 * firstlen (off - hlen) must be aligned on an
940 goto smart_frag_failure;
941 off = ((off - hlen) & ~7) + hlen;
942 newlen = (~PAGE_MASK) & mtu;
943 if ((newlen + sizeof (struct ip)) > mtu) {
944 /* we failed, go back the default */
955 firstlen = off - hlen;
956 mnext = &m0->m_nextpkt; /* pointer to next packet */
959 * Loop through length of segment after first fragment,
960 * make new header and copy data of each part and link onto chain.
961 * Here, m0 is the original packet, m is the fragment being created.
962 * The fragments are linked off the m_nextpkt of the original
963 * packet, which after processing serves as the first fragment.
965 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
966 struct ip *mhip; /* ip header on the fragment */
968 int mhlen = sizeof (struct ip);
970 m = m_gethdr(M_NOWAIT, MT_DATA);
973 IPSTAT_INC(ips_odropped);
977 * Make sure the complete packet header gets copied
978 * from the originating mbuf to the newly created
979 * mbuf. This also ensures that existing firewall
980 * classification(s), VLAN tags and so on get copied
981 * to the resulting fragmented packet(s):
983 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
986 IPSTAT_INC(ips_odropped);
990 * In the first mbuf, leave room for the link header, then
991 * copy the original IP header including options. The payload
992 * goes into an additional mbuf chain returned by m_copym().
994 m->m_data += max_linkhdr;
995 mhip = mtod(m, struct ip *);
997 if (hlen > sizeof (struct ip)) {
998 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
999 mhip->ip_v = IPVERSION;
1000 mhip->ip_hl = mhlen >> 2;
1003 /* XXX do we need to add ip_off below ? */
1004 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1005 if (off + len >= ip_len)
1008 mhip->ip_off |= IP_MF;
1009 mhip->ip_len = htons((u_short)(len + mhlen));
1010 m->m_next = m_copym(m0, off, len, M_NOWAIT);
1011 if (m->m_next == NULL) { /* copy failed */
1013 error = ENOBUFS; /* ??? */
1014 IPSTAT_INC(ips_odropped);
1017 m->m_pkthdr.len = mhlen + len;
1019 mac_netinet_fragment(m0, m);
1021 mhip->ip_off = htons(mhip->ip_off);
1023 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1024 mhip->ip_sum = in_cksum(m, mhlen);
1025 m->m_pkthdr.csum_flags &= ~CSUM_IP;
1028 mnext = &m->m_nextpkt;
1030 IPSTAT_ADD(ips_ofragments, nfrags);
1033 * Update first fragment by trimming what's been copied out
1034 * and updating header.
1036 m_adj(m0, hlen + firstlen - ip_len);
1037 m0->m_pkthdr.len = hlen + firstlen;
1038 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1039 ip->ip_off = htons(ip_off | IP_MF);
1041 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1042 ip->ip_sum = in_cksum(m0, hlen);
1043 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1052 in_delayed_cksum(struct mbuf *m)
1056 uint16_t cklen, csum, offset;
1058 ip = mtod(m, struct ip *);
1059 offset = ip->ip_hl << 2 ;
1061 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1062 /* if udp header is not in the first mbuf copy udplen */
1063 if (offset + sizeof(struct udphdr) > m->m_len) {
1064 m_copydata(m, offset + offsetof(struct udphdr,
1065 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1066 cklen = ntohs(cklen);
1068 uh = (struct udphdr *)mtodo(m, offset);
1069 cklen = ntohs(uh->uh_ulen);
1071 csum = in_cksum_skip(m, cklen + offset, offset);
1075 cklen = ntohs(ip->ip_len);
1076 csum = in_cksum_skip(m, cklen, offset);
1078 offset += m->m_pkthdr.csum_data; /* checksum offset */
1080 if (offset + sizeof(csum) > m->m_len)
1081 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1083 *(u_short *)mtodo(m, offset) = csum;
1087 * IP socket option processing.
1090 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1092 struct inpcb *inp = sotoinpcb(so);
1095 uint32_t rss_bucket;
1100 if (sopt->sopt_level != IPPROTO_IP) {
1103 if (sopt->sopt_level == SOL_SOCKET &&
1104 sopt->sopt_dir == SOPT_SET) {
1105 switch (sopt->sopt_name) {
1108 if ((so->so_options & SO_REUSEADDR) != 0)
1109 inp->inp_flags2 |= INP_REUSEADDR;
1111 inp->inp_flags2 &= ~INP_REUSEADDR;
1117 if ((so->so_options & SO_REUSEPORT) != 0)
1118 inp->inp_flags2 |= INP_REUSEPORT;
1120 inp->inp_flags2 &= ~INP_REUSEPORT;
1124 case SO_REUSEPORT_LB:
1126 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1127 inp->inp_flags2 |= INP_REUSEPORT_LB;
1129 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1135 inp->inp_inc.inc_fibnum = so->so_fibnum;
1139 case SO_MAX_PACING_RATE:
1142 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1156 switch (sopt->sopt_dir) {
1158 switch (sopt->sopt_name) {
1165 if (sopt->sopt_valsize > MLEN) {
1169 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1174 m->m_len = sopt->sopt_valsize;
1175 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1182 error = ip_pcbopts(inp, sopt->sopt_name, m);
1188 if (sopt->sopt_td != NULL) {
1189 error = priv_check(sopt->sopt_td,
1190 PRIV_NETINET_BINDANY);
1197 case IP_RSS_LISTEN_BUCKET:
1203 case IP_RECVRETOPTS:
1204 case IP_ORIGDSTADDR:
1205 case IP_RECVDSTADDR:
1213 case IP_RECVRSSBUCKETID:
1215 error = sooptcopyin(sopt, &optval, sizeof optval,
1220 switch (sopt->sopt_name) {
1222 inp->inp_ip_tos = optval;
1226 inp->inp_ip_ttl = optval;
1230 if (optval >= 0 && optval <= MAXTTL)
1231 inp->inp_ip_minttl = optval;
1236 #define OPTSET(bit) do { \
1239 inp->inp_flags |= bit; \
1241 inp->inp_flags &= ~bit; \
1245 #define OPTSET2(bit, val) do { \
1248 inp->inp_flags2 |= bit; \
1250 inp->inp_flags2 &= ~bit; \
1255 OPTSET(INP_RECVOPTS);
1258 case IP_RECVRETOPTS:
1259 OPTSET(INP_RECVRETOPTS);
1262 case IP_RECVDSTADDR:
1263 OPTSET(INP_RECVDSTADDR);
1266 case IP_ORIGDSTADDR:
1267 OPTSET2(INP_ORIGDSTADDR, optval);
1271 OPTSET(INP_RECVTTL);
1279 OPTSET(INP_ONESBCAST);
1282 OPTSET(INP_DONTFRAG);
1285 OPTSET(INP_BINDANY);
1288 OPTSET(INP_RECVTOS);
1291 OPTSET2(INP_BINDMULTI, optval);
1294 OPTSET2(INP_RECVFLOWID, optval);
1297 case IP_RSS_LISTEN_BUCKET:
1298 if ((optval >= 0) &&
1299 (optval < rss_getnumbuckets())) {
1300 inp->inp_rss_listen_bucket = optval;
1301 OPTSET2(INP_RSS_BUCKET_SET, 1);
1306 case IP_RECVRSSBUCKETID:
1307 OPTSET2(INP_RECVRSSBUCKETID, optval);
1316 * Multicast socket options are processed by the in_mcast
1319 case IP_MULTICAST_IF:
1320 case IP_MULTICAST_VIF:
1321 case IP_MULTICAST_TTL:
1322 case IP_MULTICAST_LOOP:
1323 case IP_ADD_MEMBERSHIP:
1324 case IP_DROP_MEMBERSHIP:
1325 case IP_ADD_SOURCE_MEMBERSHIP:
1326 case IP_DROP_SOURCE_MEMBERSHIP:
1327 case IP_BLOCK_SOURCE:
1328 case IP_UNBLOCK_SOURCE:
1330 case MCAST_JOIN_GROUP:
1331 case MCAST_LEAVE_GROUP:
1332 case MCAST_JOIN_SOURCE_GROUP:
1333 case MCAST_LEAVE_SOURCE_GROUP:
1334 case MCAST_BLOCK_SOURCE:
1335 case MCAST_UNBLOCK_SOURCE:
1336 error = inp_setmoptions(inp, sopt);
1340 error = sooptcopyin(sopt, &optval, sizeof optval,
1347 case IP_PORTRANGE_DEFAULT:
1348 inp->inp_flags &= ~(INP_LOWPORT);
1349 inp->inp_flags &= ~(INP_HIGHPORT);
1352 case IP_PORTRANGE_HIGH:
1353 inp->inp_flags &= ~(INP_LOWPORT);
1354 inp->inp_flags |= INP_HIGHPORT;
1357 case IP_PORTRANGE_LOW:
1358 inp->inp_flags &= ~(INP_HIGHPORT);
1359 inp->inp_flags |= INP_LOWPORT;
1369 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1370 case IP_IPSEC_POLICY:
1371 if (IPSEC_ENABLED(ipv4)) {
1372 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1379 error = ENOPROTOOPT;
1385 switch (sopt->sopt_name) {
1389 if (inp->inp_options) {
1390 struct mbuf *options;
1392 options = m_copym(inp->inp_options, 0,
1393 M_COPYALL, M_NOWAIT);
1395 if (options != NULL) {
1396 error = sooptcopyout(sopt,
1397 mtod(options, char *),
1404 sopt->sopt_valsize = 0;
1412 case IP_RECVRETOPTS:
1413 case IP_ORIGDSTADDR:
1414 case IP_RECVDSTADDR:
1427 case IP_RSSBUCKETID:
1428 case IP_RECVRSSBUCKETID:
1430 switch (sopt->sopt_name) {
1433 optval = inp->inp_ip_tos;
1437 optval = inp->inp_ip_ttl;
1441 optval = inp->inp_ip_minttl;
1444 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1445 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1448 optval = OPTBIT(INP_RECVOPTS);
1451 case IP_RECVRETOPTS:
1452 optval = OPTBIT(INP_RECVRETOPTS);
1455 case IP_RECVDSTADDR:
1456 optval = OPTBIT(INP_RECVDSTADDR);
1459 case IP_ORIGDSTADDR:
1460 optval = OPTBIT2(INP_ORIGDSTADDR);
1464 optval = OPTBIT(INP_RECVTTL);
1468 optval = OPTBIT(INP_RECVIF);
1472 if (inp->inp_flags & INP_HIGHPORT)
1473 optval = IP_PORTRANGE_HIGH;
1474 else if (inp->inp_flags & INP_LOWPORT)
1475 optval = IP_PORTRANGE_LOW;
1481 optval = OPTBIT(INP_ONESBCAST);
1484 optval = OPTBIT(INP_DONTFRAG);
1487 optval = OPTBIT(INP_BINDANY);
1490 optval = OPTBIT(INP_RECVTOS);
1493 optval = inp->inp_flowid;
1496 optval = inp->inp_flowtype;
1499 optval = OPTBIT2(INP_RECVFLOWID);
1502 case IP_RSSBUCKETID:
1503 retval = rss_hash2bucket(inp->inp_flowid,
1507 optval = rss_bucket;
1511 case IP_RECVRSSBUCKETID:
1512 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1516 optval = OPTBIT2(INP_BINDMULTI);
1519 error = sooptcopyout(sopt, &optval, sizeof optval);
1523 * Multicast socket options are processed by the in_mcast
1526 case IP_MULTICAST_IF:
1527 case IP_MULTICAST_VIF:
1528 case IP_MULTICAST_TTL:
1529 case IP_MULTICAST_LOOP:
1531 error = inp_getmoptions(inp, sopt);
1534 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1535 case IP_IPSEC_POLICY:
1536 if (IPSEC_ENABLED(ipv4)) {
1537 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1544 error = ENOPROTOOPT;
1553 * Routine called from ip_output() to loop back a copy of an IP multicast
1554 * packet to the input queue of a specified interface. Note that this
1555 * calls the output routine of the loopback "driver", but with an interface
1556 * pointer that might NOT be a loopback interface -- evil, but easier than
1557 * replicating that code here.
1560 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1566 * Make a deep copy of the packet because we're going to
1567 * modify the pack in order to generate checksums.
1569 copym = m_dup(m, M_NOWAIT);
1570 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1571 copym = m_pullup(copym, hlen);
1572 if (copym != NULL) {
1573 /* If needed, compute the checksum and mark it as valid. */
1574 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1575 in_delayed_cksum(copym);
1576 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1577 copym->m_pkthdr.csum_flags |=
1578 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1579 copym->m_pkthdr.csum_data = 0xffff;
1582 * We don't bother to fragment if the IP length is greater
1583 * than the interface's MTU. Can this possibly matter?
1585 ip = mtod(copym, struct ip *);
1587 ip->ip_sum = in_cksum(copym, hlen);
1588 if_simloop(ifp, copym, AF_INET, 0);