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_mbuf_stress_test.h"
40 #include "opt_mpath.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>
50 #include <sys/malloc.h>
54 #include <sys/protosw.h>
55 #include <sys/rmlock.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/sysctl.h>
60 #include <sys/ucred.h>
63 #include <net/if_var.h>
64 #include <net/if_llatbl.h>
65 #include <net/netisr.h>
67 #include <net/route.h>
69 #include <net/radix_mpath.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_pcb.h>
80 #include <netinet/in_rss.h>
81 #include <netinet/in_var.h>
82 #include <netinet/ip_var.h>
83 #include <netinet/ip_options.h>
85 #include <netinet/udp.h>
86 #include <netinet/udp_var.h>
89 #include <netinet/sctp.h>
90 #include <netinet/sctp_crc32.h>
93 #include <netipsec/ipsec_support.h>
95 #include <machine/in_cksum.h>
97 #include <security/mac/mac_framework.h>
99 #ifdef MBUF_STRESS_TEST
100 static int mbuf_frag_size = 0;
101 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
102 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
105 static void ip_mloopback(struct ifnet *, const struct mbuf *, int);
108 extern int in_mcast_loop;
109 extern struct protosw inetsw[];
112 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
113 struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
115 struct m_tag *fwd_tag = NULL;
119 int pflags = PFIL_OUT;
121 if (flags & IP_FORWARDING)
125 ip = mtod(m, struct ip *);
127 /* Run through list of hooks for output packets. */
128 odst.s_addr = ip->ip_dst.s_addr;
129 switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
134 return 1; /* Finished */
139 ip = mtod(m, struct ip *);
141 /* See if destination IP address was changed by packet filter. */
142 if (odst.s_addr != ip->ip_dst.s_addr) {
143 m->m_flags |= M_SKIP_FIREWALL;
144 /* If destination is now ourself drop to ip_input(). */
145 if (in_localip(ip->ip_dst)) {
146 m->m_flags |= M_FASTFWD_OURS;
147 if (m->m_pkthdr.rcvif == NULL)
148 m->m_pkthdr.rcvif = V_loif;
149 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
150 m->m_pkthdr.csum_flags |=
151 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
152 m->m_pkthdr.csum_data = 0xffff;
154 m->m_pkthdr.csum_flags |=
155 CSUM_IP_CHECKED | CSUM_IP_VALID;
157 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
158 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
160 *error = netisr_queue(NETISR_IP, m);
161 return 1; /* Finished */
164 bzero(dst, sizeof(*dst));
165 dst->sin_family = AF_INET;
166 dst->sin_len = sizeof(*dst);
167 dst->sin_addr = ip->ip_dst;
169 return -1; /* Reloop */
171 /* See if fib was changed by packet filter. */
172 if ((*fibnum) != M_GETFIB(m)) {
173 m->m_flags |= M_SKIP_FIREWALL;
174 *fibnum = M_GETFIB(m);
175 return -1; /* Reloop for FIB change */
178 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
179 if (m->m_flags & M_FASTFWD_OURS) {
180 if (m->m_pkthdr.rcvif == NULL)
181 m->m_pkthdr.rcvif = V_loif;
182 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
183 m->m_pkthdr.csum_flags |=
184 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
185 m->m_pkthdr.csum_data = 0xffff;
188 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
189 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
191 m->m_pkthdr.csum_flags |=
192 CSUM_IP_CHECKED | CSUM_IP_VALID;
194 *error = netisr_queue(NETISR_IP, m);
195 return 1; /* Finished */
197 /* Or forward to some other address? */
198 if ((m->m_flags & M_IP_NEXTHOP) &&
199 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
200 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
201 m->m_flags |= M_SKIP_FIREWALL;
202 m->m_flags &= ~M_IP_NEXTHOP;
203 m_tag_delete(m, fwd_tag);
205 return -1; /* Reloop for CHANGE of dst */
212 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
213 const struct sockaddr_in *gw, struct route *ro)
215 struct m_snd_tag *mst;
218 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
223 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
224 (inp->inp_snd_tag != NULL &&
225 inp->inp_snd_tag->ifp != ifp))
226 in_pcboutput_txrtlmt(inp, ifp, m);
228 if (inp->inp_snd_tag != NULL)
229 mst = inp->inp_snd_tag;
233 KASSERT(m->m_pkthdr.rcvif == NULL,
234 ("trying to add a send tag to a forwarded packet"));
235 if (mst->ifp != ifp) {
240 /* stamp send tag on mbuf */
241 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
242 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
245 error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
248 /* Check for route change invalidating send tags. */
251 in_pcboutput_eagain(inp);
257 * IP output. The packet in mbuf chain m contains a skeletal IP
258 * header (with len, off, ttl, proto, tos, src, dst).
259 * The mbuf chain containing the packet will be freed.
260 * The mbuf opt, if present, will not be freed.
261 * If route ro is present and has ro_rt initialized, route lookup would be
262 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
263 * then result of route lookup is stored in ro->ro_rt.
265 * In the IP forwarding case, the packet will arrive with options already
266 * inserted, so must have a NULL opt pointer.
269 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
270 struct ip_moptions *imo, struct inpcb *inp)
272 struct rm_priotracker in_ifa_tracker;
273 struct epoch_tracker et;
275 struct ifnet *ifp = NULL; /* keep compiler happy */
277 int hlen = sizeof (struct ip);
280 struct sockaddr_in *dst, sin;
281 const struct sockaddr_in *gw;
282 struct in_ifaddr *ia;
285 uint16_t ip_len, ip_off;
287 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
288 int no_route_but_check_spd = 0;
294 INP_LOCK_ASSERT(inp);
295 M_SETFIB(m, inp->inp_inc.inc_fibnum);
296 if ((flags & IP_NODEFAULTFLOWID) == 0) {
297 m->m_pkthdr.flowid = inp->inp_flowid;
298 M_HASHTYPE_SET(m, inp->inp_flowtype);
301 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
307 m = ip_insertoptions(m, opt, &len);
309 hlen = len; /* ip->ip_hl is updated above */
311 ip = mtod(m, struct ip *);
312 ip_len = ntohs(ip->ip_len);
313 ip_off = ntohs(ip->ip_off);
315 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
316 ip->ip_v = IPVERSION;
317 ip->ip_hl = hlen >> 2;
320 /* Header already set, fetch hlen from there */
321 hlen = ip->ip_hl << 2;
323 if ((flags & IP_FORWARDING) == 0)
324 IPSTAT_INC(ips_localout);
329 * gw is readonly but can point either to dst OR rt_gateway,
330 * therefore we need restore gw if we're redoing lookup.
332 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
334 dst = (struct sockaddr_in *)&ro->ro_dst;
337 if (ro == NULL || ro->ro_rt == NULL) {
338 bzero(dst, sizeof(*dst));
339 dst->sin_family = AF_INET;
340 dst->sin_len = sizeof(*dst);
341 dst->sin_addr = ip->ip_dst;
347 * Validate route against routing table additions;
348 * a better/more specific route might have been added.
350 if (inp != NULL && ro != NULL && ro->ro_rt != NULL)
351 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
353 * If there is a cached route,
354 * check that it is to the same destination
355 * and is still up. If not, free it and try again.
356 * The address family should also be checked in case of sharing the
358 * Also check whether routing cache needs invalidation.
360 if (ro != NULL && ro->ro_rt != NULL &&
361 ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
362 ro->ro_rt->rt_ifp == NULL || !RT_LINK_IS_UP(ro->ro_rt->rt_ifp) ||
363 dst->sin_family != AF_INET ||
364 dst->sin_addr.s_addr != ip->ip_dst.s_addr))
365 RO_INVALIDATE_CACHE(ro);
368 * If routing to interface only, short circuit routing lookup.
369 * The use of an all-ones broadcast address implies this; an
370 * interface is specified by the broadcast address of an interface,
371 * or the destination address of a ptp interface.
373 if (flags & IP_SENDONES) {
374 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
375 M_GETFIB(m)))) == NULL &&
376 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
377 M_GETFIB(m)))) == NULL) {
378 IPSTAT_INC(ips_noroute);
382 ip->ip_dst.s_addr = INADDR_BROADCAST;
383 dst->sin_addr = ip->ip_dst;
388 src = IA_SIN(ia)->sin_addr;
389 } else if (flags & IP_ROUTETOIF) {
390 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
391 M_GETFIB(m)))) == NULL &&
392 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
393 M_GETFIB(m)))) == NULL) {
394 IPSTAT_INC(ips_noroute);
401 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
402 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
403 src = IA_SIN(ia)->sin_addr;
404 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
405 imo != NULL && imo->imo_multicast_ifp != NULL) {
407 * Bypass the normal routing lookup for multicast
408 * packets if the interface is specified.
410 ifp = imo->imo_multicast_ifp;
412 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
413 isbroadcast = 0; /* fool gcc */
414 /* Interface may have no addresses. */
416 src = IA_SIN(ia)->sin_addr;
418 src.s_addr = INADDR_ANY;
419 } else if (ro != NULL) {
420 if (ro->ro_rt == NULL) {
422 * We want to do any cloning requested by the link
423 * layer, as this is probably required in all cases
424 * for correct operation (as it is for ARP).
427 rtalloc_mpath_fib(ro,
428 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
431 in_rtalloc_ign(ro, 0, fibnum);
433 if (ro->ro_rt == NULL ||
434 (ro->ro_rt->rt_flags & RTF_UP) == 0 ||
435 ro->ro_rt->rt_ifp == NULL ||
436 !RT_LINK_IS_UP(ro->ro_rt->rt_ifp)) {
437 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
439 * There is no route for this packet, but it is
440 * possible that a matching SPD entry exists.
442 no_route_but_check_spd = 1;
443 mtu = 0; /* Silence GCC warning. */
446 IPSTAT_INC(ips_noroute);
447 error = EHOSTUNREACH;
451 ia = ifatoia(ro->ro_rt->rt_ifa);
452 ifp = ro->ro_rt->rt_ifp;
453 counter_u64_add(ro->ro_rt->rt_pksent, 1);
454 rt_update_ro_flags(ro);
455 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
456 gw = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
457 if (ro->ro_rt->rt_flags & RTF_HOST)
458 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
459 else if (ifp->if_flags & IFF_BROADCAST)
460 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
463 if (ro->ro_rt->rt_flags & RTF_HOST)
464 mtu = ro->ro_rt->rt_mtu;
467 src = IA_SIN(ia)->sin_addr;
469 struct nhop4_extended nh;
471 bzero(&nh, sizeof(nh));
472 if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh) !=
474 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
476 * There is no route for this packet, but it is
477 * possible that a matching SPD entry exists.
479 no_route_but_check_spd = 1;
480 mtu = 0; /* Silence GCC warning. */
483 IPSTAT_INC(ips_noroute);
484 error = EHOSTUNREACH;
490 * We are rewriting here dst to be gw actually, contradicting
491 * comment at the beginning of the function. However, in this
492 * case we are always dealing with on stack dst.
493 * In case if pfil(9) sends us back to beginning of the
494 * function, the dst would be rewritten by ip_output_pfil().
497 dst->sin_addr = nh.nh_addr;
500 isbroadcast = (((nh.nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
501 (NHF_HOST | NHF_BROADCAST)) ||
502 ((ifp->if_flags & IFF_BROADCAST) &&
503 in_ifaddr_broadcast(dst->sin_addr, ia)));
506 /* Catch a possible divide by zero later. */
507 KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (rt_flags=0x%08x) ifp=%p",
509 (ro != NULL && ro->ro_rt != NULL) ? ro->ro_rt->rt_flags : 0, ifp));
511 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
512 m->m_flags |= M_MCAST;
514 * IP destination address is multicast. Make sure "gw"
515 * still points to the address in "ro". (It may have been
516 * changed to point to a gateway address, above.)
520 * See if the caller provided any multicast options
523 ip->ip_ttl = imo->imo_multicast_ttl;
524 if (imo->imo_multicast_vif != -1)
527 ip_mcast_src(imo->imo_multicast_vif) :
530 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
532 * Confirm that the outgoing interface supports multicast.
534 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
535 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
536 IPSTAT_INC(ips_noroute);
542 * If source address not specified yet, use address
543 * of outgoing interface.
545 if (ip->ip_src.s_addr == INADDR_ANY)
548 if ((imo == NULL && in_mcast_loop) ||
549 (imo && imo->imo_multicast_loop)) {
551 * Loop back multicast datagram if not expressly
552 * forbidden to do so, even if we are not a member
553 * of the group; ip_input() will filter it later,
554 * thus deferring a hash lookup and mutex acquisition
555 * at the expense of a cheap copy using m_copym().
557 ip_mloopback(ifp, m, hlen);
560 * If we are acting as a multicast router, perform
561 * multicast forwarding as if the packet had just
562 * arrived on the interface to which we are about
563 * to send. The multicast forwarding function
564 * recursively calls this function, using the
565 * IP_FORWARDING flag to prevent infinite recursion.
567 * Multicasts that are looped back by ip_mloopback(),
568 * above, will be forwarded by the ip_input() routine,
571 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
573 * If rsvp daemon is not running, do not
574 * set ip_moptions. This ensures that the packet
575 * is multicast and not just sent down one link
576 * as prescribed by rsvpd.
581 ip_mforward(ip, ifp, m, imo) != 0) {
589 * Multicasts with a time-to-live of zero may be looped-
590 * back, above, but must not be transmitted on a network.
591 * Also, multicasts addressed to the loopback interface
592 * are not sent -- the above call to ip_mloopback() will
593 * loop back a copy. ip_input() will drop the copy if
594 * this host does not belong to the destination group on
595 * the loopback interface.
597 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
606 * If the source address is not specified yet, use the address
607 * of the outoing interface.
609 if (ip->ip_src.s_addr == INADDR_ANY)
613 * Look for broadcast address and
614 * verify user is allowed to send
618 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
619 error = EADDRNOTAVAIL;
622 if ((flags & IP_ALLOWBROADCAST) == 0) {
626 /* don't allow broadcast messages to be fragmented */
631 m->m_flags |= M_BCAST;
633 m->m_flags &= ~M_BCAST;
637 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
638 if (IPSEC_ENABLED(ipv4)) {
639 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
640 if (error == EINPROGRESS)
646 * Check if there was a route for this packet; return error if not.
648 if (no_route_but_check_spd) {
649 IPSTAT_INC(ips_noroute);
650 error = EHOSTUNREACH;
653 /* Update variables that are affected by ipsec4_output(). */
654 ip = mtod(m, struct ip *);
655 hlen = ip->ip_hl << 2;
658 /* Jump over all PFIL processing if hooks are not active. */
659 if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
660 switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
662 case 1: /* Finished */
665 case 0: /* Continue normally */
666 ip = mtod(m, struct ip *);
669 case -1: /* Need to try again */
670 /* Reset everything for a new round */
673 ro->ro_prepend = NULL;
676 ip = mtod(m, struct ip *);
682 /* IN_LOOPBACK must not appear on the wire - RFC1122. */
683 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
684 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
685 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
686 IPSTAT_INC(ips_badaddr);
687 error = EADDRNOTAVAIL;
692 m->m_pkthdr.csum_flags |= CSUM_IP;
693 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
694 m = mb_unmapped_to_ext(m);
696 IPSTAT_INC(ips_odropped);
701 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
702 } else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
703 m = mb_unmapped_to_ext(m);
705 IPSTAT_INC(ips_odropped);
711 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
712 m = mb_unmapped_to_ext(m);
714 IPSTAT_INC(ips_odropped);
718 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
719 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
724 * If small enough for interface, or the interface will take
725 * care of the fragmentation for us, we can just send directly.
728 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
730 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
731 ip->ip_sum = in_cksum(m, hlen);
732 m->m_pkthdr.csum_flags &= ~CSUM_IP;
736 * Record statistics for this interface address.
737 * With CSUM_TSO the byte/packet count will be slightly
738 * incorrect because we count the IP+TCP headers only
739 * once instead of for every generated packet.
741 if (!(flags & IP_FORWARDING) && ia) {
742 if (m->m_pkthdr.csum_flags & CSUM_TSO)
743 counter_u64_add(ia->ia_ifa.ifa_opackets,
744 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
746 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
748 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
750 #ifdef MBUF_STRESS_TEST
751 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
752 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
755 * Reset layer specific mbuf flags
756 * to avoid confusing lower layers.
759 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
760 error = ip_output_send(inp, ifp, m, gw, ro);
764 /* Balk when DF bit is set or the interface didn't support TSO. */
765 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
767 IPSTAT_INC(ips_cantfrag);
772 * Too large for interface; fragment if possible. If successful,
773 * on return, m will point to a list of packets to be sent.
775 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
782 /* Record statistics for this interface address. */
784 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
785 counter_u64_add(ia->ia_ifa.ifa_obytes,
789 * Reset layer specific mbuf flags
790 * to avoid confusing upper layers.
794 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
795 mtod(m, struct ip *), NULL);
796 error = ip_output_send(inp, ifp, m, gw, ro);
802 IPSTAT_INC(ips_fragmented);
813 * Create a chain of fragments which fit the given mtu. m_frag points to the
814 * mbuf to be fragmented; on return it points to the chain with the fragments.
815 * Return 0 if no error. If error, m_frag may contain a partially built
816 * chain of fragments that should be freed by the caller.
818 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
821 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
822 u_long if_hwassist_flags)
825 int hlen = ip->ip_hl << 2;
826 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
828 struct mbuf *m0 = *m_frag; /* the original packet */
832 uint16_t ip_len, ip_off;
834 ip_len = ntohs(ip->ip_len);
835 ip_off = ntohs(ip->ip_off);
837 if (ip_off & IP_DF) { /* Fragmentation not allowed */
838 IPSTAT_INC(ips_cantfrag);
843 * Must be able to put at least 8 bytes per fragment.
849 * If the interface will not calculate checksums on
850 * fragmented packets, then do it here.
852 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
853 m0 = mb_unmapped_to_ext(m0);
856 IPSTAT_INC(ips_odropped);
859 in_delayed_cksum(m0);
860 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
863 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
864 m0 = mb_unmapped_to_ext(m0);
867 IPSTAT_INC(ips_odropped);
870 sctp_delayed_cksum(m0, hlen);
871 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
874 if (len > PAGE_SIZE) {
876 * Fragment large datagrams such that each segment
877 * contains a multiple of PAGE_SIZE amount of data,
878 * plus headers. This enables a receiver to perform
879 * page-flipping zero-copy optimizations.
881 * XXX When does this help given that sender and receiver
882 * could have different page sizes, and also mtu could
883 * be less than the receiver's page size ?
887 off = MIN(mtu, m0->m_pkthdr.len);
890 * firstlen (off - hlen) must be aligned on an
894 goto smart_frag_failure;
895 off = ((off - hlen) & ~7) + hlen;
896 newlen = (~PAGE_MASK) & mtu;
897 if ((newlen + sizeof (struct ip)) > mtu) {
898 /* we failed, go back the default */
909 firstlen = off - hlen;
910 mnext = &m0->m_nextpkt; /* pointer to next packet */
913 * Loop through length of segment after first fragment,
914 * make new header and copy data of each part and link onto chain.
915 * Here, m0 is the original packet, m is the fragment being created.
916 * The fragments are linked off the m_nextpkt of the original
917 * packet, which after processing serves as the first fragment.
919 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
920 struct ip *mhip; /* ip header on the fragment */
922 int mhlen = sizeof (struct ip);
924 m = m_gethdr(M_NOWAIT, MT_DATA);
927 IPSTAT_INC(ips_odropped);
931 * Make sure the complete packet header gets copied
932 * from the originating mbuf to the newly created
933 * mbuf. This also ensures that existing firewall
934 * classification(s), VLAN tags and so on get copied
935 * to the resulting fragmented packet(s):
937 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
940 IPSTAT_INC(ips_odropped);
944 * In the first mbuf, leave room for the link header, then
945 * copy the original IP header including options. The payload
946 * goes into an additional mbuf chain returned by m_copym().
948 m->m_data += max_linkhdr;
949 mhip = mtod(m, struct ip *);
951 if (hlen > sizeof (struct ip)) {
952 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
953 mhip->ip_v = IPVERSION;
954 mhip->ip_hl = mhlen >> 2;
957 /* XXX do we need to add ip_off below ? */
958 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
959 if (off + len >= ip_len)
962 mhip->ip_off |= IP_MF;
963 mhip->ip_len = htons((u_short)(len + mhlen));
964 m->m_next = m_copym(m0, off, len, M_NOWAIT);
965 if (m->m_next == NULL) { /* copy failed */
967 error = ENOBUFS; /* ??? */
968 IPSTAT_INC(ips_odropped);
971 m->m_pkthdr.len = mhlen + len;
973 mac_netinet_fragment(m0, m);
975 mhip->ip_off = htons(mhip->ip_off);
977 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
978 mhip->ip_sum = in_cksum(m, mhlen);
979 m->m_pkthdr.csum_flags &= ~CSUM_IP;
982 mnext = &m->m_nextpkt;
984 IPSTAT_ADD(ips_ofragments, nfrags);
987 * Update first fragment by trimming what's been copied out
988 * and updating header.
990 m_adj(m0, hlen + firstlen - ip_len);
991 m0->m_pkthdr.len = hlen + firstlen;
992 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
993 ip->ip_off = htons(ip_off | IP_MF);
995 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
996 ip->ip_sum = in_cksum(m0, hlen);
997 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1006 in_delayed_cksum(struct mbuf *m)
1010 uint16_t cklen, csum, offset;
1012 ip = mtod(m, struct ip *);
1013 offset = ip->ip_hl << 2 ;
1015 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1016 /* if udp header is not in the first mbuf copy udplen */
1017 if (offset + sizeof(struct udphdr) > m->m_len) {
1018 m_copydata(m, offset + offsetof(struct udphdr,
1019 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1020 cklen = ntohs(cklen);
1022 uh = (struct udphdr *)mtodo(m, offset);
1023 cklen = ntohs(uh->uh_ulen);
1025 csum = in_cksum_skip(m, cklen + offset, offset);
1029 cklen = ntohs(ip->ip_len);
1030 csum = in_cksum_skip(m, cklen, offset);
1032 offset += m->m_pkthdr.csum_data; /* checksum offset */
1034 if (offset + sizeof(csum) > m->m_len)
1035 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1037 *(u_short *)mtodo(m, offset) = csum;
1041 * IP socket option processing.
1044 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1046 struct inpcb *inp = sotoinpcb(so);
1049 uint32_t rss_bucket;
1054 if (sopt->sopt_level != IPPROTO_IP) {
1057 if (sopt->sopt_level == SOL_SOCKET &&
1058 sopt->sopt_dir == SOPT_SET) {
1059 switch (sopt->sopt_name) {
1062 if ((so->so_options & SO_REUSEADDR) != 0)
1063 inp->inp_flags2 |= INP_REUSEADDR;
1065 inp->inp_flags2 &= ~INP_REUSEADDR;
1071 if ((so->so_options & SO_REUSEPORT) != 0)
1072 inp->inp_flags2 |= INP_REUSEPORT;
1074 inp->inp_flags2 &= ~INP_REUSEPORT;
1078 case SO_REUSEPORT_LB:
1080 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1081 inp->inp_flags2 |= INP_REUSEPORT_LB;
1083 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1089 inp->inp_inc.inc_fibnum = so->so_fibnum;
1093 case SO_MAX_PACING_RATE:
1096 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1110 switch (sopt->sopt_dir) {
1112 switch (sopt->sopt_name) {
1119 if (sopt->sopt_valsize > MLEN) {
1123 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1128 m->m_len = sopt->sopt_valsize;
1129 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1136 error = ip_pcbopts(inp, sopt->sopt_name, m);
1142 if (sopt->sopt_td != NULL) {
1143 error = priv_check(sopt->sopt_td,
1144 PRIV_NETINET_BINDANY);
1151 case IP_RSS_LISTEN_BUCKET:
1157 case IP_RECVRETOPTS:
1158 case IP_ORIGDSTADDR:
1159 case IP_RECVDSTADDR:
1167 case IP_RECVRSSBUCKETID:
1169 error = sooptcopyin(sopt, &optval, sizeof optval,
1174 switch (sopt->sopt_name) {
1176 inp->inp_ip_tos = optval;
1180 inp->inp_ip_ttl = optval;
1184 if (optval >= 0 && optval <= MAXTTL)
1185 inp->inp_ip_minttl = optval;
1190 #define OPTSET(bit) do { \
1193 inp->inp_flags |= bit; \
1195 inp->inp_flags &= ~bit; \
1199 #define OPTSET2(bit, val) do { \
1202 inp->inp_flags2 |= bit; \
1204 inp->inp_flags2 &= ~bit; \
1209 OPTSET(INP_RECVOPTS);
1212 case IP_RECVRETOPTS:
1213 OPTSET(INP_RECVRETOPTS);
1216 case IP_RECVDSTADDR:
1217 OPTSET(INP_RECVDSTADDR);
1220 case IP_ORIGDSTADDR:
1221 OPTSET2(INP_ORIGDSTADDR, optval);
1225 OPTSET(INP_RECVTTL);
1233 OPTSET(INP_ONESBCAST);
1236 OPTSET(INP_DONTFRAG);
1239 OPTSET(INP_BINDANY);
1242 OPTSET(INP_RECVTOS);
1245 OPTSET2(INP_BINDMULTI, optval);
1248 OPTSET2(INP_RECVFLOWID, optval);
1251 case IP_RSS_LISTEN_BUCKET:
1252 if ((optval >= 0) &&
1253 (optval < rss_getnumbuckets())) {
1254 inp->inp_rss_listen_bucket = optval;
1255 OPTSET2(INP_RSS_BUCKET_SET, 1);
1260 case IP_RECVRSSBUCKETID:
1261 OPTSET2(INP_RECVRSSBUCKETID, optval);
1270 * Multicast socket options are processed by the in_mcast
1273 case IP_MULTICAST_IF:
1274 case IP_MULTICAST_VIF:
1275 case IP_MULTICAST_TTL:
1276 case IP_MULTICAST_LOOP:
1277 case IP_ADD_MEMBERSHIP:
1278 case IP_DROP_MEMBERSHIP:
1279 case IP_ADD_SOURCE_MEMBERSHIP:
1280 case IP_DROP_SOURCE_MEMBERSHIP:
1281 case IP_BLOCK_SOURCE:
1282 case IP_UNBLOCK_SOURCE:
1284 case MCAST_JOIN_GROUP:
1285 case MCAST_LEAVE_GROUP:
1286 case MCAST_JOIN_SOURCE_GROUP:
1287 case MCAST_LEAVE_SOURCE_GROUP:
1288 case MCAST_BLOCK_SOURCE:
1289 case MCAST_UNBLOCK_SOURCE:
1290 error = inp_setmoptions(inp, sopt);
1294 error = sooptcopyin(sopt, &optval, sizeof optval,
1301 case IP_PORTRANGE_DEFAULT:
1302 inp->inp_flags &= ~(INP_LOWPORT);
1303 inp->inp_flags &= ~(INP_HIGHPORT);
1306 case IP_PORTRANGE_HIGH:
1307 inp->inp_flags &= ~(INP_LOWPORT);
1308 inp->inp_flags |= INP_HIGHPORT;
1311 case IP_PORTRANGE_LOW:
1312 inp->inp_flags &= ~(INP_HIGHPORT);
1313 inp->inp_flags |= INP_LOWPORT;
1323 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1324 case IP_IPSEC_POLICY:
1325 if (IPSEC_ENABLED(ipv4)) {
1326 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1333 error = ENOPROTOOPT;
1339 switch (sopt->sopt_name) {
1343 if (inp->inp_options) {
1344 struct mbuf *options;
1346 options = m_copym(inp->inp_options, 0,
1347 M_COPYALL, M_NOWAIT);
1349 if (options != NULL) {
1350 error = sooptcopyout(sopt,
1351 mtod(options, char *),
1358 sopt->sopt_valsize = 0;
1366 case IP_RECVRETOPTS:
1367 case IP_ORIGDSTADDR:
1368 case IP_RECVDSTADDR:
1381 case IP_RSSBUCKETID:
1382 case IP_RECVRSSBUCKETID:
1384 switch (sopt->sopt_name) {
1387 optval = inp->inp_ip_tos;
1391 optval = inp->inp_ip_ttl;
1395 optval = inp->inp_ip_minttl;
1398 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1399 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1402 optval = OPTBIT(INP_RECVOPTS);
1405 case IP_RECVRETOPTS:
1406 optval = OPTBIT(INP_RECVRETOPTS);
1409 case IP_RECVDSTADDR:
1410 optval = OPTBIT(INP_RECVDSTADDR);
1413 case IP_ORIGDSTADDR:
1414 optval = OPTBIT2(INP_ORIGDSTADDR);
1418 optval = OPTBIT(INP_RECVTTL);
1422 optval = OPTBIT(INP_RECVIF);
1426 if (inp->inp_flags & INP_HIGHPORT)
1427 optval = IP_PORTRANGE_HIGH;
1428 else if (inp->inp_flags & INP_LOWPORT)
1429 optval = IP_PORTRANGE_LOW;
1435 optval = OPTBIT(INP_ONESBCAST);
1438 optval = OPTBIT(INP_DONTFRAG);
1441 optval = OPTBIT(INP_BINDANY);
1444 optval = OPTBIT(INP_RECVTOS);
1447 optval = inp->inp_flowid;
1450 optval = inp->inp_flowtype;
1453 optval = OPTBIT2(INP_RECVFLOWID);
1456 case IP_RSSBUCKETID:
1457 retval = rss_hash2bucket(inp->inp_flowid,
1461 optval = rss_bucket;
1465 case IP_RECVRSSBUCKETID:
1466 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1470 optval = OPTBIT2(INP_BINDMULTI);
1473 error = sooptcopyout(sopt, &optval, sizeof optval);
1477 * Multicast socket options are processed by the in_mcast
1480 case IP_MULTICAST_IF:
1481 case IP_MULTICAST_VIF:
1482 case IP_MULTICAST_TTL:
1483 case IP_MULTICAST_LOOP:
1485 error = inp_getmoptions(inp, sopt);
1488 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1489 case IP_IPSEC_POLICY:
1490 if (IPSEC_ENABLED(ipv4)) {
1491 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1498 error = ENOPROTOOPT;
1507 * Routine called from ip_output() to loop back a copy of an IP multicast
1508 * packet to the input queue of a specified interface. Note that this
1509 * calls the output routine of the loopback "driver", but with an interface
1510 * pointer that might NOT be a loopback interface -- evil, but easier than
1511 * replicating that code here.
1514 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1520 * Make a deep copy of the packet because we're going to
1521 * modify the pack in order to generate checksums.
1523 copym = m_dup(m, M_NOWAIT);
1524 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1525 copym = m_pullup(copym, hlen);
1526 if (copym != NULL) {
1527 /* If needed, compute the checksum and mark it as valid. */
1528 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1529 in_delayed_cksum(copym);
1530 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1531 copym->m_pkthdr.csum_flags |=
1532 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1533 copym->m_pkthdr.csum_data = 0xffff;
1536 * We don't bother to fragment if the IP length is greater
1537 * than the interface's MTU. Can this possibly matter?
1539 ip = mtod(copym, struct ip *);
1541 ip->ip_sum = in_cksum(copym, hlen);
1542 if_simloop(ifp, copym, AF_INET, 0);