2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1990, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
38 #include "opt_ipsec.h"
39 #include "opt_kern_tls.h"
40 #include "opt_mbuf_stress_test.h"
41 #include "opt_ratelimit.h"
42 #include "opt_route.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
51 #include <sys/malloc.h>
55 #include <sys/protosw.h>
56 #include <sys/rmlock.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
60 #include <sys/sysctl.h>
61 #include <sys/ucred.h>
64 #include <net/if_var.h>
65 #include <net/if_vlan_var.h>
66 #include <net/if_llatbl.h>
67 #include <net/ethernet.h>
68 #include <net/netisr.h>
70 #include <net/route.h>
71 #include <net/route/nhop.h>
72 #include <net/rss_config.h>
75 #include <netinet/in.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_kdtrace.h>
78 #include <netinet/in_systm.h>
79 #include <netinet/ip.h>
80 #include <netinet/in_fib.h>
81 #include <netinet/in_pcb.h>
82 #include <netinet/in_rss.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip_var.h>
85 #include <netinet/ip_options.h>
86 #include <netinet/ip_mroute.h>
88 #include <netinet/udp.h>
89 #include <netinet/udp_var.h>
91 #if defined(SCTP) || defined(SCTP_SUPPORT)
92 #include <netinet/sctp.h>
93 #include <netinet/sctp_crc32.h>
96 #include <netipsec/ipsec_support.h>
98 #include <machine/in_cksum.h>
100 #include <security/mac/mac_framework.h>
102 #ifdef MBUF_STRESS_TEST
103 static int mbuf_frag_size = 0;
104 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
105 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
108 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;
158 #if defined(SCTP) || defined(SCTP_SUPPORT)
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;
189 #if defined(SCTP) || defined(SCTP_SUPPORT)
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_epg_tls);
238 * If a TLS session doesn't have a valid tag, it must
239 * have had an earlier ifp mismatch, so drop this
247 * Always stamp tags that include NIC ktls.
253 if (inp != NULL && mst == NULL) {
254 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
255 (inp->inp_snd_tag != NULL &&
256 inp->inp_snd_tag->ifp != ifp))
257 in_pcboutput_txrtlmt(inp, ifp, m);
259 if (inp->inp_snd_tag != NULL)
260 mst = inp->inp_snd_tag;
263 if (stamp_tag && mst != NULL) {
264 KASSERT(m->m_pkthdr.rcvif == NULL,
265 ("trying to add a send tag to a forwarded packet"));
266 if (mst->ifp != ifp) {
271 /* stamp send tag on mbuf */
272 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
273 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
276 error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
279 /* Check for route change invalidating send tags. */
283 error = ktls_output_eagain(inp, tls);
289 in_pcboutput_eagain(inp);
294 /* rte<>ro_flags translation */
296 rt_update_ro_flags(struct route *ro)
298 int nh_flags = ro->ro_nh->nh_flags;
300 ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
302 ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
303 ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
304 ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
308 * IP output. The packet in mbuf chain m contains a skeletal IP
309 * header (with len, off, ttl, proto, tos, src, dst).
310 * The mbuf chain containing the packet will be freed.
311 * The mbuf opt, if present, will not be freed.
312 * If route ro is present and has ro_rt initialized, route lookup would be
313 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
314 * then result of route lookup is stored in ro->ro_rt.
316 * In the IP forwarding case, the packet will arrive with options already
317 * inserted, so must have a NULL opt pointer.
320 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
321 struct ip_moptions *imo, struct inpcb *inp)
323 MROUTER_RLOCK_TRACKER;
324 struct rm_priotracker in_ifa_tracker;
326 struct ifnet *ifp = NULL; /* keep compiler happy */
328 int hlen = sizeof (struct ip);
332 struct sockaddr_in *dst, sin;
333 const struct sockaddr_in *gw;
334 struct in_ifaddr *ia = NULL;
337 uint16_t ip_len, ip_off;
339 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
340 int no_route_but_check_spd = 0;
347 INP_LOCK_ASSERT(inp);
348 M_SETFIB(m, inp->inp_inc.inc_fibnum);
349 if ((flags & IP_NODEFAULTFLOWID) == 0) {
350 m->m_pkthdr.flowid = inp->inp_flowid;
351 M_HASHTYPE_SET(m, inp->inp_flowtype);
353 if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
354 vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
357 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
363 m = ip_insertoptions(m, opt, &len);
365 hlen = len; /* ip->ip_hl is updated above */
367 ip = mtod(m, struct ip *);
368 ip_len = ntohs(ip->ip_len);
369 ip_off = ntohs(ip->ip_off);
371 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
372 ip->ip_v = IPVERSION;
373 ip->ip_hl = hlen >> 2;
376 /* Header already set, fetch hlen from there */
377 hlen = ip->ip_hl << 2;
379 if ((flags & IP_FORWARDING) == 0)
380 IPSTAT_INC(ips_localout);
385 * gw is readonly but can point either to dst OR rt_gateway,
386 * therefore we need restore gw if we're redoing lookup.
388 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
390 dst = (struct sockaddr_in *)&ro->ro_dst;
393 if (ro == NULL || ro->ro_nh == NULL) {
394 bzero(dst, sizeof(*dst));
395 dst->sin_family = AF_INET;
396 dst->sin_len = sizeof(*dst);
397 dst->sin_addr = ip->ip_dst;
402 * Validate route against routing table additions;
403 * a better/more specific route might have been added.
405 if (inp != NULL && ro != NULL && ro->ro_nh != NULL)
406 NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
408 * If there is a cached route,
409 * check that it is to the same destination
410 * and is still up. If not, free it and try again.
411 * The address family should also be checked in case of sharing the
413 * Also check whether routing cache needs invalidation.
415 if (ro != NULL && ro->ro_nh != NULL &&
416 ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
417 dst->sin_addr.s_addr != ip->ip_dst.s_addr))
418 RO_INVALIDATE_CACHE(ro);
421 * If routing to interface only, short circuit routing lookup.
422 * The use of an all-ones broadcast address implies this; an
423 * interface is specified by the broadcast address of an interface,
424 * or the destination address of a ptp interface.
426 if (flags & IP_SENDONES) {
427 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
428 M_GETFIB(m)))) == NULL &&
429 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
430 M_GETFIB(m)))) == NULL) {
431 IPSTAT_INC(ips_noroute);
435 ip->ip_dst.s_addr = INADDR_BROADCAST;
436 dst->sin_addr = ip->ip_dst;
441 src = IA_SIN(ia)->sin_addr;
442 } else if (flags & IP_ROUTETOIF) {
443 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
444 M_GETFIB(m)))) == NULL &&
445 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
446 M_GETFIB(m)))) == NULL) {
447 IPSTAT_INC(ips_noroute);
454 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
455 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
456 src = IA_SIN(ia)->sin_addr;
457 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
458 imo != NULL && imo->imo_multicast_ifp != NULL) {
460 * Bypass the normal routing lookup for multicast
461 * packets if the interface is specified.
463 ifp = imo->imo_multicast_ifp;
465 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
466 isbroadcast = 0; /* fool gcc */
467 /* Interface may have no addresses. */
469 src = IA_SIN(ia)->sin_addr;
471 src.s_addr = INADDR_ANY;
472 } else if (ro != NULL) {
473 if (ro->ro_nh == NULL) {
475 * We want to do any cloning requested by the link
476 * layer, as this is probably required in all cases
477 * for correct operation (as it is for ARP).
480 flowid = m->m_pkthdr.flowid;
481 ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
484 if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
485 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
487 * There is no route for this packet, but it is
488 * possible that a matching SPD entry exists.
490 no_route_but_check_spd = 1;
493 IPSTAT_INC(ips_noroute);
494 error = EHOSTUNREACH;
498 ia = ifatoia(ro->ro_nh->nh_ifa);
499 ifp = ro->ro_nh->nh_ifp;
500 counter_u64_add(ro->ro_nh->nh_pksent, 1);
501 rt_update_ro_flags(ro);
502 if (ro->ro_nh->nh_flags & NHF_GATEWAY)
503 gw = &ro->ro_nh->gw4_sa;
504 if (ro->ro_nh->nh_flags & NHF_HOST)
505 isbroadcast = (ro->ro_nh->nh_flags & NHF_BROADCAST);
506 else if (ifp->if_flags & IFF_BROADCAST)
507 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
510 if (ro->ro_nh->nh_flags & NHF_HOST)
511 mtu = ro->ro_nh->nh_mtu;
514 src = IA_SIN(ia)->sin_addr;
516 struct nhop_object *nh;
518 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE,
521 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
523 * There is no route for this packet, but it is
524 * possible that a matching SPD entry exists.
526 no_route_but_check_spd = 1;
529 IPSTAT_INC(ips_noroute);
530 error = EHOSTUNREACH;
536 * We are rewriting here dst to be gw actually, contradicting
537 * comment at the beginning of the function. However, in this
538 * case we are always dealing with on stack dst.
539 * In case if pfil(9) sends us back to beginning of the
540 * function, the dst would be rewritten by ip_output_pfil().
543 if (nh->nh_flags & NHF_GATEWAY)
544 dst->sin_addr = nh->gw4_sa.sin_addr;
545 ia = ifatoia(nh->nh_ifa);
546 src = IA_SIN(ia)->sin_addr;
547 isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
548 (NHF_HOST | NHF_BROADCAST)) ||
549 ((ifp->if_flags & IFF_BROADCAST) &&
550 in_ifaddr_broadcast(dst->sin_addr, ia)));
553 /* Catch a possible divide by zero later. */
554 KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
556 (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
558 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
559 m->m_flags |= M_MCAST;
561 * IP destination address is multicast. Make sure "gw"
562 * still points to the address in "ro". (It may have been
563 * changed to point to a gateway address, above.)
567 * See if the caller provided any multicast options
570 ip->ip_ttl = imo->imo_multicast_ttl;
571 if (imo->imo_multicast_vif != -1)
574 ip_mcast_src(imo->imo_multicast_vif) :
577 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
579 * Confirm that the outgoing interface supports multicast.
581 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
582 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
583 IPSTAT_INC(ips_noroute);
589 * If source address not specified yet, use address
590 * of outgoing interface.
592 if (ip->ip_src.s_addr == INADDR_ANY)
595 if ((imo == NULL && in_mcast_loop) ||
596 (imo && imo->imo_multicast_loop)) {
598 * Loop back multicast datagram if not expressly
599 * forbidden to do so, even if we are not a member
600 * of the group; ip_input() will filter it later,
601 * thus deferring a hash lookup and mutex acquisition
602 * at the expense of a cheap copy using m_copym().
604 ip_mloopback(ifp, m, hlen);
607 * If we are acting as a multicast router, perform
608 * multicast forwarding as if the packet had just
609 * arrived on the interface to which we are about
610 * to send. The multicast forwarding function
611 * recursively calls this function, using the
612 * IP_FORWARDING flag to prevent infinite recursion.
614 * Multicasts that are looped back by ip_mloopback(),
615 * above, will be forwarded by the ip_input() routine,
619 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
621 * If rsvp daemon is not running, do not
622 * set ip_moptions. This ensures that the packet
623 * is multicast and not just sent down one link
624 * as prescribed by rsvpd.
629 ip_mforward(ip, ifp, m, imo) != 0) {
639 * Multicasts with a time-to-live of zero may be looped-
640 * back, above, but must not be transmitted on a network.
641 * Also, multicasts addressed to the loopback interface
642 * are not sent -- the above call to ip_mloopback() will
643 * loop back a copy. ip_input() will drop the copy if
644 * this host does not belong to the destination group on
645 * the loopback interface.
647 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
656 * If the source address is not specified yet, use the address
657 * of the outoing interface.
659 if (ip->ip_src.s_addr == INADDR_ANY)
663 * Look for broadcast address and
664 * verify user is allowed to send
668 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
669 error = EADDRNOTAVAIL;
672 if ((flags & IP_ALLOWBROADCAST) == 0) {
676 /* don't allow broadcast messages to be fragmented */
681 m->m_flags |= M_BCAST;
683 m->m_flags &= ~M_BCAST;
687 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
688 if (IPSEC_ENABLED(ipv4)) {
689 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
690 if (error == EINPROGRESS)
696 * Check if there was a route for this packet; return error if not.
698 if (no_route_but_check_spd) {
699 IPSTAT_INC(ips_noroute);
700 error = EHOSTUNREACH;
703 /* Update variables that are affected by ipsec4_output(). */
704 ip = mtod(m, struct ip *);
705 hlen = ip->ip_hl << 2;
708 /* Jump over all PFIL processing if hooks are not active. */
709 if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
710 switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
712 case 1: /* Finished */
715 case 0: /* Continue normally */
716 ip = mtod(m, struct ip *);
719 case -1: /* Need to try again */
720 /* Reset everything for a new round */
723 ro->ro_prepend = NULL;
726 ip = mtod(m, struct ip *);
732 EVL_APPLY_PRI(m, vlan_pcp);
734 /* IN_LOOPBACK must not appear on the wire - RFC1122. */
735 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
736 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
737 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
738 IPSTAT_INC(ips_badaddr);
739 error = EADDRNOTAVAIL;
744 m->m_pkthdr.csum_flags |= CSUM_IP;
745 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
746 m = mb_unmapped_to_ext(m);
748 IPSTAT_INC(ips_odropped);
753 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
754 } else if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
755 m = mb_unmapped_to_ext(m);
757 IPSTAT_INC(ips_odropped);
762 #if defined(SCTP) || defined(SCTP_SUPPORT)
763 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
764 m = mb_unmapped_to_ext(m);
766 IPSTAT_INC(ips_odropped);
770 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
771 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
776 * If small enough for interface, or the interface will take
777 * care of the fragmentation for us, we can just send directly.
778 * Note that if_vxlan could have requested TSO even though the outer
779 * frame is UDP. It is correct to not fragment such datagrams and
780 * instead just pass them on to the driver.
783 (m->m_pkthdr.csum_flags & ifp->if_hwassist &
784 (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
786 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
787 ip->ip_sum = in_cksum(m, hlen);
788 m->m_pkthdr.csum_flags &= ~CSUM_IP;
792 * Record statistics for this interface address.
793 * With CSUM_TSO the byte/packet count will be slightly
794 * incorrect because we count the IP+TCP headers only
795 * once instead of for every generated packet.
797 if (!(flags & IP_FORWARDING) && ia) {
798 if (m->m_pkthdr.csum_flags &
799 (CSUM_TSO | CSUM_INNER_TSO))
800 counter_u64_add(ia->ia_ifa.ifa_opackets,
801 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
803 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
805 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
807 #ifdef MBUF_STRESS_TEST
808 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
809 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
812 * Reset layer specific mbuf flags
813 * to avoid confusing lower layers.
816 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
817 error = ip_output_send(inp, ifp, m, gw, ro,
818 (flags & IP_NO_SND_TAG_RL) ? false : true);
822 /* Balk when DF bit is set or the interface didn't support TSO. */
823 if ((ip_off & IP_DF) ||
824 (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
826 IPSTAT_INC(ips_cantfrag);
831 * Too large for interface; fragment if possible. If successful,
832 * on return, m will point to a list of packets to be sent.
834 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
841 /* Record statistics for this interface address. */
843 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
844 counter_u64_add(ia->ia_ifa.ifa_obytes,
848 * Reset layer specific mbuf flags
849 * to avoid confusing upper layers.
853 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
854 mtod(m, struct ip *), NULL);
855 error = ip_output_send(inp, ifp, m, gw, ro, true);
861 IPSTAT_INC(ips_fragmented);
871 * Create a chain of fragments which fit the given mtu. m_frag points to the
872 * mbuf to be fragmented; on return it points to the chain with the fragments.
873 * Return 0 if no error. If error, m_frag may contain a partially built
874 * chain of fragments that should be freed by the caller.
876 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
879 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
880 u_long if_hwassist_flags)
883 int hlen = ip->ip_hl << 2;
884 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
886 struct mbuf *m0 = *m_frag; /* the original packet */
890 uint16_t ip_len, ip_off;
892 ip_len = ntohs(ip->ip_len);
893 ip_off = ntohs(ip->ip_off);
895 if (ip_off & IP_DF) { /* Fragmentation not allowed */
896 IPSTAT_INC(ips_cantfrag);
901 * Must be able to put at least 8 bytes per fragment.
907 * If the interface will not calculate checksums on
908 * fragmented packets, then do it here.
910 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
911 m0 = mb_unmapped_to_ext(m0);
914 IPSTAT_INC(ips_odropped);
917 in_delayed_cksum(m0);
918 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
920 #if defined(SCTP) || defined(SCTP_SUPPORT)
921 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
922 m0 = mb_unmapped_to_ext(m0);
925 IPSTAT_INC(ips_odropped);
928 sctp_delayed_cksum(m0, hlen);
929 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
932 if (len > PAGE_SIZE) {
934 * Fragment large datagrams such that each segment
935 * contains a multiple of PAGE_SIZE amount of data,
936 * plus headers. This enables a receiver to perform
937 * page-flipping zero-copy optimizations.
939 * XXX When does this help given that sender and receiver
940 * could have different page sizes, and also mtu could
941 * be less than the receiver's page size ?
945 off = MIN(mtu, m0->m_pkthdr.len);
948 * firstlen (off - hlen) must be aligned on an
952 goto smart_frag_failure;
953 off = ((off - hlen) & ~7) + hlen;
954 newlen = (~PAGE_MASK) & mtu;
955 if ((newlen + sizeof (struct ip)) > mtu) {
956 /* we failed, go back the default */
967 firstlen = off - hlen;
968 mnext = &m0->m_nextpkt; /* pointer to next packet */
971 * Loop through length of segment after first fragment,
972 * make new header and copy data of each part and link onto chain.
973 * Here, m0 is the original packet, m is the fragment being created.
974 * The fragments are linked off the m_nextpkt of the original
975 * packet, which after processing serves as the first fragment.
977 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
978 struct ip *mhip; /* ip header on the fragment */
980 int mhlen = sizeof (struct ip);
982 m = m_gethdr(M_NOWAIT, MT_DATA);
985 IPSTAT_INC(ips_odropped);
989 * Make sure the complete packet header gets copied
990 * from the originating mbuf to the newly created
991 * mbuf. This also ensures that existing firewall
992 * classification(s), VLAN tags and so on get copied
993 * to the resulting fragmented packet(s):
995 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
998 IPSTAT_INC(ips_odropped);
1002 * In the first mbuf, leave room for the link header, then
1003 * copy the original IP header including options. The payload
1004 * goes into an additional mbuf chain returned by m_copym().
1006 m->m_data += max_linkhdr;
1007 mhip = mtod(m, struct ip *);
1009 if (hlen > sizeof (struct ip)) {
1010 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1011 mhip->ip_v = IPVERSION;
1012 mhip->ip_hl = mhlen >> 2;
1015 /* XXX do we need to add ip_off below ? */
1016 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1017 if (off + len >= ip_len)
1020 mhip->ip_off |= IP_MF;
1021 mhip->ip_len = htons((u_short)(len + mhlen));
1022 m->m_next = m_copym(m0, off, len, M_NOWAIT);
1023 if (m->m_next == NULL) { /* copy failed */
1025 error = ENOBUFS; /* ??? */
1026 IPSTAT_INC(ips_odropped);
1029 m->m_pkthdr.len = mhlen + len;
1031 mac_netinet_fragment(m0, m);
1033 mhip->ip_off = htons(mhip->ip_off);
1035 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1036 mhip->ip_sum = in_cksum(m, mhlen);
1037 m->m_pkthdr.csum_flags &= ~CSUM_IP;
1040 mnext = &m->m_nextpkt;
1042 IPSTAT_ADD(ips_ofragments, nfrags);
1045 * Update first fragment by trimming what's been copied out
1046 * and updating header.
1048 m_adj(m0, hlen + firstlen - ip_len);
1049 m0->m_pkthdr.len = hlen + firstlen;
1050 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1051 ip->ip_off = htons(ip_off | IP_MF);
1053 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1054 ip->ip_sum = in_cksum(m0, hlen);
1055 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1064 in_delayed_cksum(struct mbuf *m)
1068 uint16_t cklen, csum, offset;
1070 ip = mtod(m, struct ip *);
1071 offset = ip->ip_hl << 2 ;
1073 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1074 /* if udp header is not in the first mbuf copy udplen */
1075 if (offset + sizeof(struct udphdr) > m->m_len) {
1076 m_copydata(m, offset + offsetof(struct udphdr,
1077 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1078 cklen = ntohs(cklen);
1080 uh = (struct udphdr *)mtodo(m, offset);
1081 cklen = ntohs(uh->uh_ulen);
1083 csum = in_cksum_skip(m, cklen + offset, offset);
1087 cklen = ntohs(ip->ip_len);
1088 csum = in_cksum_skip(m, cklen, offset);
1090 offset += m->m_pkthdr.csum_data; /* checksum offset */
1092 if (offset + sizeof(csum) > m->m_len)
1093 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1095 *(u_short *)mtodo(m, offset) = csum;
1099 * IP socket option processing.
1102 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1104 struct inpcb *inp = sotoinpcb(so);
1107 uint32_t rss_bucket;
1112 if (sopt->sopt_level != IPPROTO_IP) {
1115 if (sopt->sopt_level == SOL_SOCKET &&
1116 sopt->sopt_dir == SOPT_SET) {
1117 switch (sopt->sopt_name) {
1120 if ((so->so_options & SO_REUSEADDR) != 0)
1121 inp->inp_flags2 |= INP_REUSEADDR;
1123 inp->inp_flags2 &= ~INP_REUSEADDR;
1129 if ((so->so_options & SO_REUSEPORT) != 0)
1130 inp->inp_flags2 |= INP_REUSEPORT;
1132 inp->inp_flags2 &= ~INP_REUSEPORT;
1136 case SO_REUSEPORT_LB:
1138 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1139 inp->inp_flags2 |= INP_REUSEPORT_LB;
1141 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1147 inp->inp_inc.inc_fibnum = so->so_fibnum;
1151 case SO_MAX_PACING_RATE:
1154 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1168 switch (sopt->sopt_dir) {
1170 switch (sopt->sopt_name) {
1177 if (sopt->sopt_valsize > MLEN) {
1181 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1186 m->m_len = sopt->sopt_valsize;
1187 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1194 error = ip_pcbopts(inp, sopt->sopt_name, m);
1200 if (sopt->sopt_td != NULL) {
1201 error = priv_check(sopt->sopt_td,
1202 PRIV_NETINET_BINDANY);
1209 case IP_RSS_LISTEN_BUCKET:
1215 case IP_RECVRETOPTS:
1216 case IP_ORIGDSTADDR:
1217 case IP_RECVDSTADDR:
1225 case IP_RECVRSSBUCKETID:
1228 error = sooptcopyin(sopt, &optval, sizeof optval,
1233 switch (sopt->sopt_name) {
1235 inp->inp_ip_tos = optval;
1239 inp->inp_ip_ttl = optval;
1243 if (optval >= 0 && optval <= MAXTTL)
1244 inp->inp_ip_minttl = optval;
1249 #define OPTSET(bit) do { \
1252 inp->inp_flags |= bit; \
1254 inp->inp_flags &= ~bit; \
1258 #define OPTSET2(bit, val) do { \
1261 inp->inp_flags2 |= bit; \
1263 inp->inp_flags2 &= ~bit; \
1268 OPTSET(INP_RECVOPTS);
1271 case IP_RECVRETOPTS:
1272 OPTSET(INP_RECVRETOPTS);
1275 case IP_RECVDSTADDR:
1276 OPTSET(INP_RECVDSTADDR);
1279 case IP_ORIGDSTADDR:
1280 OPTSET2(INP_ORIGDSTADDR, optval);
1284 OPTSET(INP_RECVTTL);
1292 OPTSET(INP_ONESBCAST);
1295 OPTSET(INP_DONTFRAG);
1298 OPTSET(INP_BINDANY);
1301 OPTSET(INP_RECVTOS);
1304 OPTSET2(INP_BINDMULTI, optval);
1307 OPTSET2(INP_RECVFLOWID, optval);
1310 case IP_RSS_LISTEN_BUCKET:
1311 if ((optval >= 0) &&
1312 (optval < rss_getnumbuckets())) {
1313 inp->inp_rss_listen_bucket = optval;
1314 OPTSET2(INP_RSS_BUCKET_SET, 1);
1319 case IP_RECVRSSBUCKETID:
1320 OPTSET2(INP_RECVRSSBUCKETID, optval);
1324 if ((optval >= -1) && (optval <=
1325 (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1339 optval << INP_2PCP_SHIFT;
1351 * Multicast socket options are processed by the in_mcast
1354 case IP_MULTICAST_IF:
1355 case IP_MULTICAST_VIF:
1356 case IP_MULTICAST_TTL:
1357 case IP_MULTICAST_LOOP:
1358 case IP_ADD_MEMBERSHIP:
1359 case IP_DROP_MEMBERSHIP:
1360 case IP_ADD_SOURCE_MEMBERSHIP:
1361 case IP_DROP_SOURCE_MEMBERSHIP:
1362 case IP_BLOCK_SOURCE:
1363 case IP_UNBLOCK_SOURCE:
1365 case MCAST_JOIN_GROUP:
1366 case MCAST_LEAVE_GROUP:
1367 case MCAST_JOIN_SOURCE_GROUP:
1368 case MCAST_LEAVE_SOURCE_GROUP:
1369 case MCAST_BLOCK_SOURCE:
1370 case MCAST_UNBLOCK_SOURCE:
1371 error = inp_setmoptions(inp, sopt);
1375 error = sooptcopyin(sopt, &optval, sizeof optval,
1382 case IP_PORTRANGE_DEFAULT:
1383 inp->inp_flags &= ~(INP_LOWPORT);
1384 inp->inp_flags &= ~(INP_HIGHPORT);
1387 case IP_PORTRANGE_HIGH:
1388 inp->inp_flags &= ~(INP_LOWPORT);
1389 inp->inp_flags |= INP_HIGHPORT;
1392 case IP_PORTRANGE_LOW:
1393 inp->inp_flags &= ~(INP_HIGHPORT);
1394 inp->inp_flags |= INP_LOWPORT;
1404 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1405 case IP_IPSEC_POLICY:
1406 if (IPSEC_ENABLED(ipv4)) {
1407 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1414 error = ENOPROTOOPT;
1420 switch (sopt->sopt_name) {
1424 if (inp->inp_options) {
1425 struct mbuf *options;
1427 options = m_copym(inp->inp_options, 0,
1428 M_COPYALL, M_NOWAIT);
1430 if (options != NULL) {
1431 error = sooptcopyout(sopt,
1432 mtod(options, char *),
1439 sopt->sopt_valsize = 0;
1447 case IP_RECVRETOPTS:
1448 case IP_ORIGDSTADDR:
1449 case IP_RECVDSTADDR:
1462 case IP_RSSBUCKETID:
1463 case IP_RECVRSSBUCKETID:
1466 switch (sopt->sopt_name) {
1468 optval = inp->inp_ip_tos;
1472 optval = inp->inp_ip_ttl;
1476 optval = inp->inp_ip_minttl;
1479 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1480 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1483 optval = OPTBIT(INP_RECVOPTS);
1486 case IP_RECVRETOPTS:
1487 optval = OPTBIT(INP_RECVRETOPTS);
1490 case IP_RECVDSTADDR:
1491 optval = OPTBIT(INP_RECVDSTADDR);
1494 case IP_ORIGDSTADDR:
1495 optval = OPTBIT2(INP_ORIGDSTADDR);
1499 optval = OPTBIT(INP_RECVTTL);
1503 optval = OPTBIT(INP_RECVIF);
1507 if (inp->inp_flags & INP_HIGHPORT)
1508 optval = IP_PORTRANGE_HIGH;
1509 else if (inp->inp_flags & INP_LOWPORT)
1510 optval = IP_PORTRANGE_LOW;
1516 optval = OPTBIT(INP_ONESBCAST);
1519 optval = OPTBIT(INP_DONTFRAG);
1522 optval = OPTBIT(INP_BINDANY);
1525 optval = OPTBIT(INP_RECVTOS);
1528 optval = inp->inp_flowid;
1531 optval = inp->inp_flowtype;
1534 optval = OPTBIT2(INP_RECVFLOWID);
1537 case IP_RSSBUCKETID:
1538 retval = rss_hash2bucket(inp->inp_flowid,
1542 optval = rss_bucket;
1546 case IP_RECVRSSBUCKETID:
1547 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1551 optval = OPTBIT2(INP_BINDMULTI);
1554 if (OPTBIT2(INP_2PCP_SET)) {
1555 optval = (inp->inp_flags2 &
1556 INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1562 error = sooptcopyout(sopt, &optval, sizeof optval);
1566 * Multicast socket options are processed by the in_mcast
1569 case IP_MULTICAST_IF:
1570 case IP_MULTICAST_VIF:
1571 case IP_MULTICAST_TTL:
1572 case IP_MULTICAST_LOOP:
1574 error = inp_getmoptions(inp, sopt);
1577 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1578 case IP_IPSEC_POLICY:
1579 if (IPSEC_ENABLED(ipv4)) {
1580 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1587 error = ENOPROTOOPT;
1596 * Routine called from ip_output() to loop back a copy of an IP multicast
1597 * packet to the input queue of a specified interface. Note that this
1598 * calls the output routine of the loopback "driver", but with an interface
1599 * pointer that might NOT be a loopback interface -- evil, but easier than
1600 * replicating that code here.
1603 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1609 * Make a deep copy of the packet because we're going to
1610 * modify the pack in order to generate checksums.
1612 copym = m_dup(m, M_NOWAIT);
1613 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1614 copym = m_pullup(copym, hlen);
1615 if (copym != NULL) {
1616 /* If needed, compute the checksum and mark it as valid. */
1617 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1618 in_delayed_cksum(copym);
1619 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1620 copym->m_pkthdr.csum_flags |=
1621 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1622 copym->m_pkthdr.csum_data = 0xffff;
1625 * We don't bother to fragment if the IP length is greater
1626 * than the interface's MTU. Can this possibly matter?
1628 ip = mtod(copym, struct ip *);
1630 ip->ip_sum = in_cksum(copym, hlen);
1631 if_simloop(ifp, copym, AF_INET, 0);