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 *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, 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, const struct nhop_object *nh)
298 int nh_flags = 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;
333 const struct sockaddr *gw;
334 struct in_ifaddr *ia = NULL;
337 uint16_t ip_len, ip_off;
338 struct route iproute;
340 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
341 int no_route_but_check_spd = 0;
348 INP_LOCK_ASSERT(inp);
349 M_SETFIB(m, inp->inp_inc.inc_fibnum);
350 if ((flags & IP_NODEFAULTFLOWID) == 0) {
351 m->m_pkthdr.flowid = inp->inp_flowid;
352 M_HASHTYPE_SET(m, inp->inp_flowtype);
354 if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
355 vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
358 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
364 m = ip_insertoptions(m, opt, &len);
366 hlen = len; /* ip->ip_hl is updated above */
368 ip = mtod(m, struct ip *);
369 ip_len = ntohs(ip->ip_len);
370 ip_off = ntohs(ip->ip_off);
372 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
373 ip->ip_v = IPVERSION;
374 ip->ip_hl = hlen >> 2;
377 /* Header already set, fetch hlen from there */
378 hlen = ip->ip_hl << 2;
380 if ((flags & IP_FORWARDING) == 0)
381 IPSTAT_INC(ips_localout);
386 * gw is readonly but can point either to dst OR rt_gateway,
387 * therefore we need restore gw if we're redoing lookup.
389 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
392 bzero(ro, sizeof (*ro));
394 dst = (struct sockaddr_in *)&ro->ro_dst;
395 if (ro->ro_nh == NULL) {
396 dst->sin_family = AF_INET;
397 dst->sin_len = sizeof(*dst);
398 dst->sin_addr = ip->ip_dst;
400 gw = (const struct sockaddr *)dst;
403 * Validate route against routing table additions;
404 * a better/more specific route might have been added.
406 if (inp != NULL && ro->ro_nh != NULL)
407 NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
409 * If there is a cached route,
410 * check that it is to the same destination
411 * and is still up. If not, free it and try again.
412 * The address family should also be checked in case of sharing the
414 * Also check whether routing cache needs invalidation.
416 if (ro->ro_nh != NULL &&
417 ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
418 dst->sin_addr.s_addr != ip->ip_dst.s_addr))
419 RO_INVALIDATE_CACHE(ro);
422 * If routing to interface only, short circuit routing lookup.
423 * The use of an all-ones broadcast address implies this; an
424 * interface is specified by the broadcast address of an interface,
425 * or the destination address of a ptp interface.
427 if (flags & IP_SENDONES) {
428 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
429 M_GETFIB(m)))) == NULL &&
430 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
431 M_GETFIB(m)))) == NULL) {
432 IPSTAT_INC(ips_noroute);
436 ip->ip_dst.s_addr = INADDR_BROADCAST;
437 dst->sin_addr = ip->ip_dst;
442 src = IA_SIN(ia)->sin_addr;
443 } else if (flags & IP_ROUTETOIF) {
444 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
445 M_GETFIB(m)))) == NULL &&
446 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
447 M_GETFIB(m)))) == NULL) {
448 IPSTAT_INC(ips_noroute);
455 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
456 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
457 src = IA_SIN(ia)->sin_addr;
458 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
459 imo != NULL && imo->imo_multicast_ifp != NULL) {
461 * Bypass the normal routing lookup for multicast
462 * packets if the interface is specified.
464 ifp = imo->imo_multicast_ifp;
466 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
467 isbroadcast = 0; /* fool gcc */
468 /* Interface may have no addresses. */
470 src = IA_SIN(ia)->sin_addr;
472 src.s_addr = INADDR_ANY;
473 } else if (ro != &iproute) {
474 if (ro->ro_nh == NULL) {
476 * We want to do any cloning requested by the link
477 * layer, as this is probably required in all cases
478 * for correct operation (as it is for ARP).
481 flowid = m->m_pkthdr.flowid;
482 ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
485 if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
486 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
488 * There is no route for this packet, but it is
489 * possible that a matching SPD entry exists.
491 no_route_but_check_spd = 1;
494 IPSTAT_INC(ips_noroute);
495 error = EHOSTUNREACH;
499 struct nhop_object *nh = ro->ro_nh;
501 ia = ifatoia(nh->nh_ifa);
503 counter_u64_add(nh->nh_pksent, 1);
504 rt_update_ro_flags(ro, nh);
505 if (nh->nh_flags & NHF_GATEWAY)
507 if (nh->nh_flags & NHF_HOST)
508 isbroadcast = (nh->nh_flags & NHF_BROADCAST);
509 else if ((ifp->if_flags & IFF_BROADCAST) && (gw->sa_family == AF_INET))
510 isbroadcast = in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia);
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;
535 rt_update_ro_flags(ro, nh);
536 if (nh->nh_flags & NHF_GATEWAY)
538 ia = ifatoia(nh->nh_ifa);
539 src = IA_SIN(ia)->sin_addr;
540 isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
541 (NHF_HOST | NHF_BROADCAST)) ||
542 ((ifp->if_flags & IFF_BROADCAST) &&
543 (gw->sa_family == AF_INET) &&
544 in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia)));
547 /* Catch a possible divide by zero later. */
548 KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
550 (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
552 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
553 m->m_flags |= M_MCAST;
555 * IP destination address is multicast. Make sure "gw"
556 * still points to the address in "ro". (It may have been
557 * changed to point to a gateway address, above.)
559 gw = (const struct sockaddr *)dst;
561 * See if the caller provided any multicast options
564 ip->ip_ttl = imo->imo_multicast_ttl;
565 if (imo->imo_multicast_vif != -1)
568 ip_mcast_src(imo->imo_multicast_vif) :
571 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
573 * Confirm that the outgoing interface supports multicast.
575 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
576 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
577 IPSTAT_INC(ips_noroute);
583 * If source address not specified yet, use address
584 * of outgoing interface.
586 if (ip->ip_src.s_addr == INADDR_ANY)
589 if ((imo == NULL && in_mcast_loop) ||
590 (imo && imo->imo_multicast_loop)) {
592 * Loop back multicast datagram if not expressly
593 * forbidden to do so, even if we are not a member
594 * of the group; ip_input() will filter it later,
595 * thus deferring a hash lookup and mutex acquisition
596 * at the expense of a cheap copy using m_copym().
598 ip_mloopback(ifp, m, hlen);
601 * If we are acting as a multicast router, perform
602 * multicast forwarding as if the packet had just
603 * arrived on the interface to which we are about
604 * to send. The multicast forwarding function
605 * recursively calls this function, using the
606 * IP_FORWARDING flag to prevent infinite recursion.
608 * Multicasts that are looped back by ip_mloopback(),
609 * above, will be forwarded by the ip_input() routine,
613 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
615 * If rsvp daemon is not running, do not
616 * set ip_moptions. This ensures that the packet
617 * is multicast and not just sent down one link
618 * as prescribed by rsvpd.
623 ip_mforward(ip, ifp, m, imo) != 0) {
633 * Multicasts with a time-to-live of zero may be looped-
634 * back, above, but must not be transmitted on a network.
635 * Also, multicasts addressed to the loopback interface
636 * are not sent -- the above call to ip_mloopback() will
637 * loop back a copy. ip_input() will drop the copy if
638 * this host does not belong to the destination group on
639 * the loopback interface.
641 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
650 * If the source address is not specified yet, use the address
651 * of the outoing interface.
653 if (ip->ip_src.s_addr == INADDR_ANY)
657 * Look for broadcast address and
658 * verify user is allowed to send
662 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
663 error = EADDRNOTAVAIL;
666 if ((flags & IP_ALLOWBROADCAST) == 0) {
670 /* don't allow broadcast messages to be fragmented */
675 m->m_flags |= M_BCAST;
677 m->m_flags &= ~M_BCAST;
681 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
682 if (IPSEC_ENABLED(ipv4)) {
683 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
684 if (error == EINPROGRESS)
690 * Check if there was a route for this packet; return error if not.
692 if (no_route_but_check_spd) {
693 IPSTAT_INC(ips_noroute);
694 error = EHOSTUNREACH;
697 /* Update variables that are affected by ipsec4_output(). */
698 ip = mtod(m, struct ip *);
699 hlen = ip->ip_hl << 2;
702 /* Jump over all PFIL processing if hooks are not active. */
703 if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
704 switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
706 case 1: /* Finished */
709 case 0: /* Continue normally */
710 ip = mtod(m, struct ip *);
713 case -1: /* Need to try again */
714 /* Reset everything for a new round */
717 ro->ro_prepend = NULL;
719 gw = (const struct sockaddr *)dst;
720 ip = mtod(m, struct ip *);
726 EVL_APPLY_PRI(m, vlan_pcp);
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_MEXTPG) == 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.
772 * Note that if_vxlan could have requested TSO even though the outer
773 * frame is UDP. It is correct to not fragment such datagrams and
774 * instead just pass them on to the driver.
777 (m->m_pkthdr.csum_flags & ifp->if_hwassist &
778 (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
780 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
781 ip->ip_sum = in_cksum(m, hlen);
782 m->m_pkthdr.csum_flags &= ~CSUM_IP;
786 * Record statistics for this interface address.
787 * With CSUM_TSO the byte/packet count will be slightly
788 * incorrect because we count the IP+TCP headers only
789 * once instead of for every generated packet.
791 if (!(flags & IP_FORWARDING) && ia) {
792 if (m->m_pkthdr.csum_flags &
793 (CSUM_TSO | CSUM_INNER_TSO))
794 counter_u64_add(ia->ia_ifa.ifa_opackets,
795 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
797 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
799 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
801 #ifdef MBUF_STRESS_TEST
802 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
803 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
806 * Reset layer specific mbuf flags
807 * to avoid confusing lower layers.
810 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
811 error = ip_output_send(inp, ifp, m, gw, ro,
812 (flags & IP_NO_SND_TAG_RL) ? false : true);
816 /* Balk when DF bit is set or the interface didn't support TSO. */
817 if ((ip_off & IP_DF) ||
818 (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
820 IPSTAT_INC(ips_cantfrag);
825 * Too large for interface; fragment if possible. If successful,
826 * on return, m will point to a list of packets to be sent.
828 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
835 /* Record statistics for this interface address. */
837 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
838 counter_u64_add(ia->ia_ifa.ifa_obytes,
842 * Reset layer specific mbuf flags
843 * to avoid confusing upper layers.
847 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
848 mtod(m, struct ip *), NULL);
849 error = ip_output_send(inp, ifp, m, gw, ro, true);
855 IPSTAT_INC(ips_fragmented);
865 * Create a chain of fragments which fit the given mtu. m_frag points to the
866 * mbuf to be fragmented; on return it points to the chain with the fragments.
867 * Return 0 if no error. If error, m_frag may contain a partially built
868 * chain of fragments that should be freed by the caller.
870 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
873 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
874 u_long if_hwassist_flags)
877 int hlen = ip->ip_hl << 2;
878 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
880 struct mbuf *m0 = *m_frag; /* the original packet */
884 uint16_t ip_len, ip_off;
886 ip_len = ntohs(ip->ip_len);
887 ip_off = ntohs(ip->ip_off);
889 if (ip_off & IP_DF) { /* Fragmentation not allowed */
890 IPSTAT_INC(ips_cantfrag);
895 * Must be able to put at least 8 bytes per fragment.
901 * If the interface will not calculate checksums on
902 * fragmented packets, then do it here.
904 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
905 m0 = mb_unmapped_to_ext(m0);
908 IPSTAT_INC(ips_odropped);
911 in_delayed_cksum(m0);
912 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
914 #if defined(SCTP) || defined(SCTP_SUPPORT)
915 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
916 m0 = mb_unmapped_to_ext(m0);
919 IPSTAT_INC(ips_odropped);
922 sctp_delayed_cksum(m0, hlen);
923 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
926 if (len > PAGE_SIZE) {
928 * Fragment large datagrams such that each segment
929 * contains a multiple of PAGE_SIZE amount of data,
930 * plus headers. This enables a receiver to perform
931 * page-flipping zero-copy optimizations.
933 * XXX When does this help given that sender and receiver
934 * could have different page sizes, and also mtu could
935 * be less than the receiver's page size ?
939 off = MIN(mtu, m0->m_pkthdr.len);
942 * firstlen (off - hlen) must be aligned on an
946 goto smart_frag_failure;
947 off = ((off - hlen) & ~7) + hlen;
948 newlen = (~PAGE_MASK) & mtu;
949 if ((newlen + sizeof (struct ip)) > mtu) {
950 /* we failed, go back the default */
961 firstlen = off - hlen;
962 mnext = &m0->m_nextpkt; /* pointer to next packet */
965 * Loop through length of segment after first fragment,
966 * make new header and copy data of each part and link onto chain.
967 * Here, m0 is the original packet, m is the fragment being created.
968 * The fragments are linked off the m_nextpkt of the original
969 * packet, which after processing serves as the first fragment.
971 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
972 struct ip *mhip; /* ip header on the fragment */
974 int mhlen = sizeof (struct ip);
976 m = m_gethdr(M_NOWAIT, MT_DATA);
979 IPSTAT_INC(ips_odropped);
983 * Make sure the complete packet header gets copied
984 * from the originating mbuf to the newly created
985 * mbuf. This also ensures that existing firewall
986 * classification(s), VLAN tags and so on get copied
987 * to the resulting fragmented packet(s):
989 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
992 IPSTAT_INC(ips_odropped);
996 * In the first mbuf, leave room for the link header, then
997 * copy the original IP header including options. The payload
998 * goes into an additional mbuf chain returned by m_copym().
1000 m->m_data += max_linkhdr;
1001 mhip = mtod(m, struct ip *);
1003 if (hlen > sizeof (struct ip)) {
1004 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1005 mhip->ip_v = IPVERSION;
1006 mhip->ip_hl = mhlen >> 2;
1009 /* XXX do we need to add ip_off below ? */
1010 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1011 if (off + len >= ip_len)
1014 mhip->ip_off |= IP_MF;
1015 mhip->ip_len = htons((u_short)(len + mhlen));
1016 m->m_next = m_copym(m0, off, len, M_NOWAIT);
1017 if (m->m_next == NULL) { /* copy failed */
1019 error = ENOBUFS; /* ??? */
1020 IPSTAT_INC(ips_odropped);
1023 m->m_pkthdr.len = mhlen + len;
1025 mac_netinet_fragment(m0, m);
1027 mhip->ip_off = htons(mhip->ip_off);
1029 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1030 mhip->ip_sum = in_cksum(m, mhlen);
1031 m->m_pkthdr.csum_flags &= ~CSUM_IP;
1034 mnext = &m->m_nextpkt;
1036 IPSTAT_ADD(ips_ofragments, nfrags);
1039 * Update first fragment by trimming what's been copied out
1040 * and updating header.
1042 m_adj(m0, hlen + firstlen - ip_len);
1043 m0->m_pkthdr.len = hlen + firstlen;
1044 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1045 ip->ip_off = htons(ip_off | IP_MF);
1047 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1048 ip->ip_sum = in_cksum(m0, hlen);
1049 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1058 in_delayed_cksum(struct mbuf *m)
1062 uint16_t cklen, csum, offset;
1064 ip = mtod(m, struct ip *);
1065 offset = ip->ip_hl << 2 ;
1067 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1068 /* if udp header is not in the first mbuf copy udplen */
1069 if (offset + sizeof(struct udphdr) > m->m_len) {
1070 m_copydata(m, offset + offsetof(struct udphdr,
1071 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1072 cklen = ntohs(cklen);
1074 uh = (struct udphdr *)mtodo(m, offset);
1075 cklen = ntohs(uh->uh_ulen);
1077 csum = in_cksum_skip(m, cklen + offset, offset);
1081 cklen = ntohs(ip->ip_len);
1082 csum = in_cksum_skip(m, cklen, offset);
1084 offset += m->m_pkthdr.csum_data; /* checksum offset */
1086 if (offset + sizeof(csum) > m->m_len)
1087 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1089 *(u_short *)mtodo(m, offset) = csum;
1093 * IP socket option processing.
1096 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1098 struct inpcb *inp = sotoinpcb(so);
1101 uint32_t rss_bucket;
1106 if (sopt->sopt_level != IPPROTO_IP) {
1109 if (sopt->sopt_level == SOL_SOCKET &&
1110 sopt->sopt_dir == SOPT_SET) {
1111 switch (sopt->sopt_name) {
1114 if ((so->so_options & SO_REUSEADDR) != 0)
1115 inp->inp_flags2 |= INP_REUSEADDR;
1117 inp->inp_flags2 &= ~INP_REUSEADDR;
1123 if ((so->so_options & SO_REUSEPORT) != 0)
1124 inp->inp_flags2 |= INP_REUSEPORT;
1126 inp->inp_flags2 &= ~INP_REUSEPORT;
1130 case SO_REUSEPORT_LB:
1132 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1133 inp->inp_flags2 |= INP_REUSEPORT_LB;
1135 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1141 inp->inp_inc.inc_fibnum = so->so_fibnum;
1145 case SO_MAX_PACING_RATE:
1148 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1162 switch (sopt->sopt_dir) {
1164 switch (sopt->sopt_name) {
1171 if (sopt->sopt_valsize > MLEN) {
1175 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1180 m->m_len = sopt->sopt_valsize;
1181 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1188 error = ip_pcbopts(inp, sopt->sopt_name, m);
1194 if (sopt->sopt_td != NULL) {
1195 error = priv_check(sopt->sopt_td,
1196 PRIV_NETINET_BINDANY);
1203 case IP_RSS_LISTEN_BUCKET:
1209 case IP_RECVRETOPTS:
1210 case IP_ORIGDSTADDR:
1211 case IP_RECVDSTADDR:
1219 case IP_RECVRSSBUCKETID:
1222 error = sooptcopyin(sopt, &optval, sizeof optval,
1227 switch (sopt->sopt_name) {
1229 inp->inp_ip_tos = optval;
1233 inp->inp_ip_ttl = optval;
1237 if (optval >= 0 && optval <= MAXTTL)
1238 inp->inp_ip_minttl = optval;
1243 #define OPTSET(bit) do { \
1246 inp->inp_flags |= bit; \
1248 inp->inp_flags &= ~bit; \
1252 #define OPTSET2(bit, val) do { \
1255 inp->inp_flags2 |= bit; \
1257 inp->inp_flags2 &= ~bit; \
1262 OPTSET(INP_RECVOPTS);
1265 case IP_RECVRETOPTS:
1266 OPTSET(INP_RECVRETOPTS);
1269 case IP_RECVDSTADDR:
1270 OPTSET(INP_RECVDSTADDR);
1273 case IP_ORIGDSTADDR:
1274 OPTSET2(INP_ORIGDSTADDR, optval);
1278 OPTSET(INP_RECVTTL);
1286 OPTSET(INP_ONESBCAST);
1289 OPTSET(INP_DONTFRAG);
1292 OPTSET(INP_BINDANY);
1295 OPTSET(INP_RECVTOS);
1298 OPTSET2(INP_BINDMULTI, optval);
1301 OPTSET2(INP_RECVFLOWID, optval);
1304 case IP_RSS_LISTEN_BUCKET:
1305 if ((optval >= 0) &&
1306 (optval < rss_getnumbuckets())) {
1307 inp->inp_rss_listen_bucket = optval;
1308 OPTSET2(INP_RSS_BUCKET_SET, 1);
1313 case IP_RECVRSSBUCKETID:
1314 OPTSET2(INP_RECVRSSBUCKETID, optval);
1318 if ((optval >= -1) && (optval <=
1319 (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1333 optval << INP_2PCP_SHIFT;
1345 * Multicast socket options are processed by the in_mcast
1348 case IP_MULTICAST_IF:
1349 case IP_MULTICAST_VIF:
1350 case IP_MULTICAST_TTL:
1351 case IP_MULTICAST_LOOP:
1352 case IP_ADD_MEMBERSHIP:
1353 case IP_DROP_MEMBERSHIP:
1354 case IP_ADD_SOURCE_MEMBERSHIP:
1355 case IP_DROP_SOURCE_MEMBERSHIP:
1356 case IP_BLOCK_SOURCE:
1357 case IP_UNBLOCK_SOURCE:
1359 case MCAST_JOIN_GROUP:
1360 case MCAST_LEAVE_GROUP:
1361 case MCAST_JOIN_SOURCE_GROUP:
1362 case MCAST_LEAVE_SOURCE_GROUP:
1363 case MCAST_BLOCK_SOURCE:
1364 case MCAST_UNBLOCK_SOURCE:
1365 error = inp_setmoptions(inp, sopt);
1369 error = sooptcopyin(sopt, &optval, sizeof optval,
1376 case IP_PORTRANGE_DEFAULT:
1377 inp->inp_flags &= ~(INP_LOWPORT);
1378 inp->inp_flags &= ~(INP_HIGHPORT);
1381 case IP_PORTRANGE_HIGH:
1382 inp->inp_flags &= ~(INP_LOWPORT);
1383 inp->inp_flags |= INP_HIGHPORT;
1386 case IP_PORTRANGE_LOW:
1387 inp->inp_flags &= ~(INP_HIGHPORT);
1388 inp->inp_flags |= INP_LOWPORT;
1398 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1399 case IP_IPSEC_POLICY:
1400 if (IPSEC_ENABLED(ipv4)) {
1401 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1408 error = ENOPROTOOPT;
1414 switch (sopt->sopt_name) {
1418 if (inp->inp_options) {
1419 struct mbuf *options;
1421 options = m_copym(inp->inp_options, 0,
1422 M_COPYALL, M_NOWAIT);
1424 if (options != NULL) {
1425 error = sooptcopyout(sopt,
1426 mtod(options, char *),
1433 sopt->sopt_valsize = 0;
1441 case IP_RECVRETOPTS:
1442 case IP_ORIGDSTADDR:
1443 case IP_RECVDSTADDR:
1456 case IP_RSSBUCKETID:
1457 case IP_RECVRSSBUCKETID:
1460 switch (sopt->sopt_name) {
1462 optval = inp->inp_ip_tos;
1466 optval = inp->inp_ip_ttl;
1470 optval = inp->inp_ip_minttl;
1473 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1474 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1477 optval = OPTBIT(INP_RECVOPTS);
1480 case IP_RECVRETOPTS:
1481 optval = OPTBIT(INP_RECVRETOPTS);
1484 case IP_RECVDSTADDR:
1485 optval = OPTBIT(INP_RECVDSTADDR);
1488 case IP_ORIGDSTADDR:
1489 optval = OPTBIT2(INP_ORIGDSTADDR);
1493 optval = OPTBIT(INP_RECVTTL);
1497 optval = OPTBIT(INP_RECVIF);
1501 if (inp->inp_flags & INP_HIGHPORT)
1502 optval = IP_PORTRANGE_HIGH;
1503 else if (inp->inp_flags & INP_LOWPORT)
1504 optval = IP_PORTRANGE_LOW;
1510 optval = OPTBIT(INP_ONESBCAST);
1513 optval = OPTBIT(INP_DONTFRAG);
1516 optval = OPTBIT(INP_BINDANY);
1519 optval = OPTBIT(INP_RECVTOS);
1522 optval = inp->inp_flowid;
1525 optval = inp->inp_flowtype;
1528 optval = OPTBIT2(INP_RECVFLOWID);
1531 case IP_RSSBUCKETID:
1532 retval = rss_hash2bucket(inp->inp_flowid,
1536 optval = rss_bucket;
1540 case IP_RECVRSSBUCKETID:
1541 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1545 optval = OPTBIT2(INP_BINDMULTI);
1548 if (OPTBIT2(INP_2PCP_SET)) {
1549 optval = (inp->inp_flags2 &
1550 INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1556 error = sooptcopyout(sopt, &optval, sizeof optval);
1560 * Multicast socket options are processed by the in_mcast
1563 case IP_MULTICAST_IF:
1564 case IP_MULTICAST_VIF:
1565 case IP_MULTICAST_TTL:
1566 case IP_MULTICAST_LOOP:
1568 error = inp_getmoptions(inp, sopt);
1571 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1572 case IP_IPSEC_POLICY:
1573 if (IPSEC_ENABLED(ipv4)) {
1574 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1581 error = ENOPROTOOPT;
1590 * Routine called from ip_output() to loop back a copy of an IP multicast
1591 * packet to the input queue of a specified interface. Note that this
1592 * calls the output routine of the loopback "driver", but with an interface
1593 * pointer that might NOT be a loopback interface -- evil, but easier than
1594 * replicating that code here.
1597 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1603 * Make a deep copy of the packet because we're going to
1604 * modify the pack in order to generate checksums.
1606 copym = m_dup(m, M_NOWAIT);
1607 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1608 copym = m_pullup(copym, hlen);
1609 if (copym != NULL) {
1610 /* If needed, compute the checksum and mark it as valid. */
1611 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1612 in_delayed_cksum(copym);
1613 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1614 copym->m_pkthdr.csum_flags |=
1615 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1616 copym->m_pkthdr.csum_data = 0xffff;
1619 * We don't bother to fragment if the IP length is greater
1620 * than the interface's MTU. Can this possibly matter?
1622 ip = mtod(copym, struct ip *);
1624 ip->ip_sum = in_cksum(copym, hlen);
1625 if_simloop(ifp, copym, AF_INET, 0);