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_llatbl.h>
66 #include <net/netisr.h>
68 #include <net/route.h>
69 #include <net/route/nhop.h>
70 #include <net/rss_config.h>
73 #include <netinet/in.h>
74 #include <netinet/in_fib.h>
75 #include <netinet/in_kdtrace.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/ip.h>
78 #include <netinet/in_fib.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>
88 #if defined(SCTP) || defined(SCTP_SUPPORT)
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);
107 extern int in_mcast_loop;
108 extern struct protosw inetsw[];
111 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
112 struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
114 struct m_tag *fwd_tag = NULL;
118 int pflags = PFIL_OUT;
120 if (flags & IP_FORWARDING)
124 ip = mtod(m, struct ip *);
126 /* Run through list of hooks for output packets. */
127 odst.s_addr = ip->ip_dst.s_addr;
128 switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
133 return 1; /* Finished */
138 ip = mtod(m, struct ip *);
140 /* See if destination IP address was changed by packet filter. */
141 if (odst.s_addr != ip->ip_dst.s_addr) {
142 m->m_flags |= M_SKIP_FIREWALL;
143 /* If destination is now ourself drop to ip_input(). */
144 if (in_localip(ip->ip_dst)) {
145 m->m_flags |= M_FASTFWD_OURS;
146 if (m->m_pkthdr.rcvif == NULL)
147 m->m_pkthdr.rcvif = V_loif;
148 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
149 m->m_pkthdr.csum_flags |=
150 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
151 m->m_pkthdr.csum_data = 0xffff;
153 m->m_pkthdr.csum_flags |=
154 CSUM_IP_CHECKED | CSUM_IP_VALID;
155 #if defined(SCTP) || defined(SCTP_SUPPORT)
156 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
157 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
159 *error = netisr_queue(NETISR_IP, m);
160 return 1; /* Finished */
163 bzero(dst, sizeof(*dst));
164 dst->sin_family = AF_INET;
165 dst->sin_len = sizeof(*dst);
166 dst->sin_addr = ip->ip_dst;
168 return -1; /* Reloop */
170 /* See if fib was changed by packet filter. */
171 if ((*fibnum) != M_GETFIB(m)) {
172 m->m_flags |= M_SKIP_FIREWALL;
173 *fibnum = M_GETFIB(m);
174 return -1; /* Reloop for FIB change */
177 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
178 if (m->m_flags & M_FASTFWD_OURS) {
179 if (m->m_pkthdr.rcvif == NULL)
180 m->m_pkthdr.rcvif = V_loif;
181 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
182 m->m_pkthdr.csum_flags |=
183 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
184 m->m_pkthdr.csum_data = 0xffff;
186 #if defined(SCTP) || defined(SCTP_SUPPORT)
187 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
188 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
190 m->m_pkthdr.csum_flags |=
191 CSUM_IP_CHECKED | CSUM_IP_VALID;
193 *error = netisr_queue(NETISR_IP, m);
194 return 1; /* Finished */
196 /* Or forward to some other address? */
197 if ((m->m_flags & M_IP_NEXTHOP) &&
198 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
199 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
200 m->m_flags |= M_SKIP_FIREWALL;
201 m->m_flags &= ~M_IP_NEXTHOP;
202 m_tag_delete(m, fwd_tag);
204 return -1; /* Reloop for CHANGE of dst */
211 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
212 const struct sockaddr_in *gw, struct route *ro, bool stamp_tag)
215 struct ktls_session *tls = NULL;
217 struct m_snd_tag *mst;
220 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
225 * If this is an unencrypted TLS record, save a reference to
226 * the record. This local reference is used to call
227 * ktls_output_eagain after the mbuf has been freed (thus
228 * dropping the mbuf's reference) in if_output.
230 if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
231 tls = ktls_hold(m->m_next->m_epg_tls);
235 * If a TLS session doesn't have a valid tag, it must
236 * have had an earlier ifp mismatch, so drop this
244 * Always stamp tags that include NIC ktls.
250 if (inp != NULL && mst == NULL) {
251 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
252 (inp->inp_snd_tag != NULL &&
253 inp->inp_snd_tag->ifp != ifp))
254 in_pcboutput_txrtlmt(inp, ifp, m);
256 if (inp->inp_snd_tag != NULL)
257 mst = inp->inp_snd_tag;
260 if (stamp_tag && mst != NULL) {
261 KASSERT(m->m_pkthdr.rcvif == NULL,
262 ("trying to add a send tag to a forwarded packet"));
263 if (mst->ifp != ifp) {
268 /* stamp send tag on mbuf */
269 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
270 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
273 error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
276 /* Check for route change invalidating send tags. */
280 error = ktls_output_eagain(inp, tls);
286 in_pcboutput_eagain(inp);
291 /* rte<>ro_flags translation */
293 rt_update_ro_flags(struct route *ro)
295 int nh_flags = ro->ro_nh->nh_flags;
297 ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
299 ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
300 ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
301 ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
305 * IP output. The packet in mbuf chain m contains a skeletal IP
306 * header (with len, off, ttl, proto, tos, src, dst).
307 * The mbuf chain containing the packet will be freed.
308 * The mbuf opt, if present, will not be freed.
309 * If route ro is present and has ro_rt initialized, route lookup would be
310 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
311 * then result of route lookup is stored in ro->ro_rt.
313 * In the IP forwarding case, the packet will arrive with options already
314 * inserted, so must have a NULL opt pointer.
317 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
318 struct ip_moptions *imo, struct inpcb *inp)
320 struct rm_priotracker in_ifa_tracker;
322 struct ifnet *ifp = NULL; /* keep compiler happy */
324 int hlen = sizeof (struct ip);
327 struct sockaddr_in *dst, sin;
328 const struct sockaddr_in *gw;
329 struct in_ifaddr *ia = NULL;
332 uint16_t ip_len, ip_off;
334 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
335 int no_route_but_check_spd = 0;
342 INP_LOCK_ASSERT(inp);
343 M_SETFIB(m, inp->inp_inc.inc_fibnum);
344 if ((flags & IP_NODEFAULTFLOWID) == 0) {
345 m->m_pkthdr.flowid = inp->inp_flowid;
346 M_HASHTYPE_SET(m, inp->inp_flowtype);
349 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
355 m = ip_insertoptions(m, opt, &len);
357 hlen = len; /* ip->ip_hl is updated above */
359 ip = mtod(m, struct ip *);
360 ip_len = ntohs(ip->ip_len);
361 ip_off = ntohs(ip->ip_off);
363 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
364 ip->ip_v = IPVERSION;
365 ip->ip_hl = hlen >> 2;
368 /* Header already set, fetch hlen from there */
369 hlen = ip->ip_hl << 2;
371 if ((flags & IP_FORWARDING) == 0)
372 IPSTAT_INC(ips_localout);
377 * gw is readonly but can point either to dst OR rt_gateway,
378 * therefore we need restore gw if we're redoing lookup.
380 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
382 dst = (struct sockaddr_in *)&ro->ro_dst;
385 if (ro == NULL || ro->ro_nh == NULL) {
386 bzero(dst, sizeof(*dst));
387 dst->sin_family = AF_INET;
388 dst->sin_len = sizeof(*dst);
389 dst->sin_addr = ip->ip_dst;
394 * Validate route against routing table additions;
395 * a better/more specific route might have been added.
397 if (inp != NULL && ro != NULL && ro->ro_nh != NULL)
398 NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
400 * If there is a cached route,
401 * check that it is to the same destination
402 * and is still up. If not, free it and try again.
403 * The address family should also be checked in case of sharing the
405 * Also check whether routing cache needs invalidation.
407 if (ro != NULL && ro->ro_nh != NULL &&
408 ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
409 dst->sin_addr.s_addr != ip->ip_dst.s_addr))
410 RO_INVALIDATE_CACHE(ro);
413 * If routing to interface only, short circuit routing lookup.
414 * The use of an all-ones broadcast address implies this; an
415 * interface is specified by the broadcast address of an interface,
416 * or the destination address of a ptp interface.
418 if (flags & IP_SENDONES) {
419 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
420 M_GETFIB(m)))) == NULL &&
421 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
422 M_GETFIB(m)))) == NULL) {
423 IPSTAT_INC(ips_noroute);
427 ip->ip_dst.s_addr = INADDR_BROADCAST;
428 dst->sin_addr = ip->ip_dst;
433 src = IA_SIN(ia)->sin_addr;
434 } else if (flags & IP_ROUTETOIF) {
435 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
436 M_GETFIB(m)))) == NULL &&
437 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
438 M_GETFIB(m)))) == NULL) {
439 IPSTAT_INC(ips_noroute);
446 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
447 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
448 src = IA_SIN(ia)->sin_addr;
449 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
450 imo != NULL && imo->imo_multicast_ifp != NULL) {
452 * Bypass the normal routing lookup for multicast
453 * packets if the interface is specified.
455 ifp = imo->imo_multicast_ifp;
457 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
458 isbroadcast = 0; /* fool gcc */
459 /* Interface may have no addresses. */
461 src = IA_SIN(ia)->sin_addr;
463 src.s_addr = INADDR_ANY;
464 } else if (ro != NULL) {
465 if (ro->ro_nh == NULL) {
467 * We want to do any cloning requested by the link
468 * layer, as this is probably required in all cases
469 * for correct operation (as it is for ARP).
472 flowid = m->m_pkthdr.flowid;
473 ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
476 if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
477 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
479 * There is no route for this packet, but it is
480 * possible that a matching SPD entry exists.
482 no_route_but_check_spd = 1;
485 IPSTAT_INC(ips_noroute);
486 error = EHOSTUNREACH;
490 ia = ifatoia(ro->ro_nh->nh_ifa);
491 ifp = ro->ro_nh->nh_ifp;
492 counter_u64_add(ro->ro_nh->nh_pksent, 1);
493 rt_update_ro_flags(ro);
494 if (ro->ro_nh->nh_flags & NHF_GATEWAY)
495 gw = &ro->ro_nh->gw4_sa;
496 if (ro->ro_nh->nh_flags & NHF_HOST)
497 isbroadcast = (ro->ro_nh->nh_flags & NHF_BROADCAST);
498 else if (ifp->if_flags & IFF_BROADCAST)
499 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
502 if (ro->ro_nh->nh_flags & NHF_HOST)
503 mtu = ro->ro_nh->nh_mtu;
506 src = IA_SIN(ia)->sin_addr;
508 struct nhop_object *nh;
510 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE,
513 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
515 * There is no route for this packet, but it is
516 * possible that a matching SPD entry exists.
518 no_route_but_check_spd = 1;
521 IPSTAT_INC(ips_noroute);
522 error = EHOSTUNREACH;
528 * We are rewriting here dst to be gw actually, contradicting
529 * comment at the beginning of the function. However, in this
530 * case we are always dealing with on stack dst.
531 * In case if pfil(9) sends us back to beginning of the
532 * function, the dst would be rewritten by ip_output_pfil().
535 if (nh->nh_flags & NHF_GATEWAY)
536 dst->sin_addr = nh->gw4_sa.sin_addr;
537 ia = ifatoia(nh->nh_ifa);
538 src = IA_SIN(ia)->sin_addr;
539 isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
540 (NHF_HOST | NHF_BROADCAST)) ||
541 ((ifp->if_flags & IFF_BROADCAST) &&
542 in_ifaddr_broadcast(dst->sin_addr, ia)));
545 /* Catch a possible divide by zero later. */
546 KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
548 (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
550 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
551 m->m_flags |= M_MCAST;
553 * IP destination address is multicast. Make sure "gw"
554 * still points to the address in "ro". (It may have been
555 * changed to point to a gateway address, above.)
559 * See if the caller provided any multicast options
562 ip->ip_ttl = imo->imo_multicast_ttl;
563 if (imo->imo_multicast_vif != -1)
566 ip_mcast_src(imo->imo_multicast_vif) :
569 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
571 * Confirm that the outgoing interface supports multicast.
573 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
574 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
575 IPSTAT_INC(ips_noroute);
581 * If source address not specified yet, use address
582 * of outgoing interface.
584 if (ip->ip_src.s_addr == INADDR_ANY)
587 if ((imo == NULL && in_mcast_loop) ||
588 (imo && imo->imo_multicast_loop)) {
590 * Loop back multicast datagram if not expressly
591 * forbidden to do so, even if we are not a member
592 * of the group; ip_input() will filter it later,
593 * thus deferring a hash lookup and mutex acquisition
594 * at the expense of a cheap copy using m_copym().
596 ip_mloopback(ifp, m, hlen);
599 * If we are acting as a multicast router, perform
600 * multicast forwarding as if the packet had just
601 * arrived on the interface to which we are about
602 * to send. The multicast forwarding function
603 * recursively calls this function, using the
604 * IP_FORWARDING flag to prevent infinite recursion.
606 * Multicasts that are looped back by ip_mloopback(),
607 * above, will be forwarded by the ip_input() routine,
610 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
612 * If rsvp daemon is not running, do not
613 * set ip_moptions. This ensures that the packet
614 * is multicast and not just sent down one link
615 * as prescribed by rsvpd.
620 ip_mforward(ip, ifp, m, imo) != 0) {
628 * Multicasts with a time-to-live of zero may be looped-
629 * back, above, but must not be transmitted on a network.
630 * Also, multicasts addressed to the loopback interface
631 * are not sent -- the above call to ip_mloopback() will
632 * loop back a copy. ip_input() will drop the copy if
633 * this host does not belong to the destination group on
634 * the loopback interface.
636 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
645 * If the source address is not specified yet, use the address
646 * of the outoing interface.
648 if (ip->ip_src.s_addr == INADDR_ANY)
652 * Look for broadcast address and
653 * verify user is allowed to send
657 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
658 error = EADDRNOTAVAIL;
661 if ((flags & IP_ALLOWBROADCAST) == 0) {
665 /* don't allow broadcast messages to be fragmented */
670 m->m_flags |= M_BCAST;
672 m->m_flags &= ~M_BCAST;
676 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
677 if (IPSEC_ENABLED(ipv4)) {
678 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
679 if (error == EINPROGRESS)
685 * Check if there was a route for this packet; return error if not.
687 if (no_route_but_check_spd) {
688 IPSTAT_INC(ips_noroute);
689 error = EHOSTUNREACH;
692 /* Update variables that are affected by ipsec4_output(). */
693 ip = mtod(m, struct ip *);
694 hlen = ip->ip_hl << 2;
697 /* Jump over all PFIL processing if hooks are not active. */
698 if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
699 switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
701 case 1: /* Finished */
704 case 0: /* Continue normally */
705 ip = mtod(m, struct ip *);
708 case -1: /* Need to try again */
709 /* Reset everything for a new round */
712 ro->ro_prepend = NULL;
715 ip = mtod(m, struct ip *);
720 /* IN_LOOPBACK must not appear on the wire - RFC1122. */
721 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
722 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
723 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
724 IPSTAT_INC(ips_badaddr);
725 error = EADDRNOTAVAIL;
730 m->m_pkthdr.csum_flags |= CSUM_IP;
731 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
732 m = mb_unmapped_to_ext(m);
734 IPSTAT_INC(ips_odropped);
739 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
740 } else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
741 m = mb_unmapped_to_ext(m);
743 IPSTAT_INC(ips_odropped);
748 #if defined(SCTP) || defined(SCTP_SUPPORT)
749 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
750 m = mb_unmapped_to_ext(m);
752 IPSTAT_INC(ips_odropped);
756 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
757 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
762 * If small enough for interface, or the interface will take
763 * care of the fragmentation for us, we can just send directly.
764 * Note that if_vxlan could have requested TSO even though the outer
765 * frame is UDP. It is correct to not fragment such datagrams and
766 * instead just pass them on to the driver.
769 (m->m_pkthdr.csum_flags & ifp->if_hwassist &
770 (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
772 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
773 ip->ip_sum = in_cksum(m, hlen);
774 m->m_pkthdr.csum_flags &= ~CSUM_IP;
778 * Record statistics for this interface address.
779 * With CSUM_TSO the byte/packet count will be slightly
780 * incorrect because we count the IP+TCP headers only
781 * once instead of for every generated packet.
783 if (!(flags & IP_FORWARDING) && ia) {
784 if (m->m_pkthdr.csum_flags &
785 (CSUM_TSO | CSUM_INNER_TSO))
786 counter_u64_add(ia->ia_ifa.ifa_opackets,
787 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
789 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
791 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
793 #ifdef MBUF_STRESS_TEST
794 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
795 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
798 * Reset layer specific mbuf flags
799 * to avoid confusing lower layers.
802 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
803 error = ip_output_send(inp, ifp, m, gw, ro,
804 (flags & IP_NO_SND_TAG_RL) ? false : true);
808 /* Balk when DF bit is set or the interface didn't support TSO. */
809 if ((ip_off & IP_DF) ||
810 (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
812 IPSTAT_INC(ips_cantfrag);
817 * Too large for interface; fragment if possible. If successful,
818 * on return, m will point to a list of packets to be sent.
820 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
827 /* Record statistics for this interface address. */
829 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
830 counter_u64_add(ia->ia_ifa.ifa_obytes,
834 * Reset layer specific mbuf flags
835 * to avoid confusing upper layers.
839 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
840 mtod(m, struct ip *), NULL);
841 error = ip_output_send(inp, ifp, m, gw, ro, true);
847 IPSTAT_INC(ips_fragmented);
857 * Create a chain of fragments which fit the given mtu. m_frag points to the
858 * mbuf to be fragmented; on return it points to the chain with the fragments.
859 * Return 0 if no error. If error, m_frag may contain a partially built
860 * chain of fragments that should be freed by the caller.
862 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
865 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
866 u_long if_hwassist_flags)
869 int hlen = ip->ip_hl << 2;
870 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
872 struct mbuf *m0 = *m_frag; /* the original packet */
876 uint16_t ip_len, ip_off;
878 ip_len = ntohs(ip->ip_len);
879 ip_off = ntohs(ip->ip_off);
881 if (ip_off & IP_DF) { /* Fragmentation not allowed */
882 IPSTAT_INC(ips_cantfrag);
887 * Must be able to put at least 8 bytes per fragment.
893 * If the interface will not calculate checksums on
894 * fragmented packets, then do it here.
896 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
897 m0 = mb_unmapped_to_ext(m0);
900 IPSTAT_INC(ips_odropped);
903 in_delayed_cksum(m0);
904 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
906 #if defined(SCTP) || defined(SCTP_SUPPORT)
907 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
908 m0 = mb_unmapped_to_ext(m0);
911 IPSTAT_INC(ips_odropped);
914 sctp_delayed_cksum(m0, hlen);
915 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
918 if (len > PAGE_SIZE) {
920 * Fragment large datagrams such that each segment
921 * contains a multiple of PAGE_SIZE amount of data,
922 * plus headers. This enables a receiver to perform
923 * page-flipping zero-copy optimizations.
925 * XXX When does this help given that sender and receiver
926 * could have different page sizes, and also mtu could
927 * be less than the receiver's page size ?
931 off = MIN(mtu, m0->m_pkthdr.len);
934 * firstlen (off - hlen) must be aligned on an
938 goto smart_frag_failure;
939 off = ((off - hlen) & ~7) + hlen;
940 newlen = (~PAGE_MASK) & mtu;
941 if ((newlen + sizeof (struct ip)) > mtu) {
942 /* we failed, go back the default */
953 firstlen = off - hlen;
954 mnext = &m0->m_nextpkt; /* pointer to next packet */
957 * Loop through length of segment after first fragment,
958 * make new header and copy data of each part and link onto chain.
959 * Here, m0 is the original packet, m is the fragment being created.
960 * The fragments are linked off the m_nextpkt of the original
961 * packet, which after processing serves as the first fragment.
963 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
964 struct ip *mhip; /* ip header on the fragment */
966 int mhlen = sizeof (struct ip);
968 m = m_gethdr(M_NOWAIT, MT_DATA);
971 IPSTAT_INC(ips_odropped);
975 * Make sure the complete packet header gets copied
976 * from the originating mbuf to the newly created
977 * mbuf. This also ensures that existing firewall
978 * classification(s), VLAN tags and so on get copied
979 * to the resulting fragmented packet(s):
981 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
984 IPSTAT_INC(ips_odropped);
988 * In the first mbuf, leave room for the link header, then
989 * copy the original IP header including options. The payload
990 * goes into an additional mbuf chain returned by m_copym().
992 m->m_data += max_linkhdr;
993 mhip = mtod(m, struct ip *);
995 if (hlen > sizeof (struct ip)) {
996 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
997 mhip->ip_v = IPVERSION;
998 mhip->ip_hl = mhlen >> 2;
1001 /* XXX do we need to add ip_off below ? */
1002 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1003 if (off + len >= ip_len)
1006 mhip->ip_off |= IP_MF;
1007 mhip->ip_len = htons((u_short)(len + mhlen));
1008 m->m_next = m_copym(m0, off, len, M_NOWAIT);
1009 if (m->m_next == NULL) { /* copy failed */
1011 error = ENOBUFS; /* ??? */
1012 IPSTAT_INC(ips_odropped);
1015 m->m_pkthdr.len = mhlen + len;
1017 mac_netinet_fragment(m0, m);
1019 mhip->ip_off = htons(mhip->ip_off);
1021 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1022 mhip->ip_sum = in_cksum(m, mhlen);
1023 m->m_pkthdr.csum_flags &= ~CSUM_IP;
1026 mnext = &m->m_nextpkt;
1028 IPSTAT_ADD(ips_ofragments, nfrags);
1031 * Update first fragment by trimming what's been copied out
1032 * and updating header.
1034 m_adj(m0, hlen + firstlen - ip_len);
1035 m0->m_pkthdr.len = hlen + firstlen;
1036 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1037 ip->ip_off = htons(ip_off | IP_MF);
1039 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1040 ip->ip_sum = in_cksum(m0, hlen);
1041 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1050 in_delayed_cksum(struct mbuf *m)
1054 uint16_t cklen, csum, offset;
1056 ip = mtod(m, struct ip *);
1057 offset = ip->ip_hl << 2 ;
1059 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1060 /* if udp header is not in the first mbuf copy udplen */
1061 if (offset + sizeof(struct udphdr) > m->m_len) {
1062 m_copydata(m, offset + offsetof(struct udphdr,
1063 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1064 cklen = ntohs(cklen);
1066 uh = (struct udphdr *)mtodo(m, offset);
1067 cklen = ntohs(uh->uh_ulen);
1069 csum = in_cksum_skip(m, cklen + offset, offset);
1073 cklen = ntohs(ip->ip_len);
1074 csum = in_cksum_skip(m, cklen, offset);
1076 offset += m->m_pkthdr.csum_data; /* checksum offset */
1078 if (offset + sizeof(csum) > m->m_len)
1079 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1081 *(u_short *)mtodo(m, offset) = csum;
1085 * IP socket option processing.
1088 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1090 struct inpcb *inp = sotoinpcb(so);
1093 uint32_t rss_bucket;
1098 if (sopt->sopt_level != IPPROTO_IP) {
1101 if (sopt->sopt_level == SOL_SOCKET &&
1102 sopt->sopt_dir == SOPT_SET) {
1103 switch (sopt->sopt_name) {
1106 if ((so->so_options & SO_REUSEADDR) != 0)
1107 inp->inp_flags2 |= INP_REUSEADDR;
1109 inp->inp_flags2 &= ~INP_REUSEADDR;
1115 if ((so->so_options & SO_REUSEPORT) != 0)
1116 inp->inp_flags2 |= INP_REUSEPORT;
1118 inp->inp_flags2 &= ~INP_REUSEPORT;
1122 case SO_REUSEPORT_LB:
1124 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1125 inp->inp_flags2 |= INP_REUSEPORT_LB;
1127 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1133 inp->inp_inc.inc_fibnum = so->so_fibnum;
1137 case SO_MAX_PACING_RATE:
1140 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1154 switch (sopt->sopt_dir) {
1156 switch (sopt->sopt_name) {
1163 if (sopt->sopt_valsize > MLEN) {
1167 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1172 m->m_len = sopt->sopt_valsize;
1173 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1180 error = ip_pcbopts(inp, sopt->sopt_name, m);
1186 if (sopt->sopt_td != NULL) {
1187 error = priv_check(sopt->sopt_td,
1188 PRIV_NETINET_BINDANY);
1195 case IP_RSS_LISTEN_BUCKET:
1201 case IP_RECVRETOPTS:
1202 case IP_ORIGDSTADDR:
1203 case IP_RECVDSTADDR:
1211 case IP_RECVRSSBUCKETID:
1213 error = sooptcopyin(sopt, &optval, sizeof optval,
1218 switch (sopt->sopt_name) {
1220 inp->inp_ip_tos = optval;
1224 inp->inp_ip_ttl = optval;
1228 if (optval >= 0 && optval <= MAXTTL)
1229 inp->inp_ip_minttl = optval;
1234 #define OPTSET(bit) do { \
1237 inp->inp_flags |= bit; \
1239 inp->inp_flags &= ~bit; \
1243 #define OPTSET2(bit, val) do { \
1246 inp->inp_flags2 |= bit; \
1248 inp->inp_flags2 &= ~bit; \
1253 OPTSET(INP_RECVOPTS);
1256 case IP_RECVRETOPTS:
1257 OPTSET(INP_RECVRETOPTS);
1260 case IP_RECVDSTADDR:
1261 OPTSET(INP_RECVDSTADDR);
1264 case IP_ORIGDSTADDR:
1265 OPTSET2(INP_ORIGDSTADDR, optval);
1269 OPTSET(INP_RECVTTL);
1277 OPTSET(INP_ONESBCAST);
1280 OPTSET(INP_DONTFRAG);
1283 OPTSET(INP_BINDANY);
1286 OPTSET(INP_RECVTOS);
1289 OPTSET2(INP_BINDMULTI, optval);
1292 OPTSET2(INP_RECVFLOWID, optval);
1295 case IP_RSS_LISTEN_BUCKET:
1296 if ((optval >= 0) &&
1297 (optval < rss_getnumbuckets())) {
1298 inp->inp_rss_listen_bucket = optval;
1299 OPTSET2(INP_RSS_BUCKET_SET, 1);
1304 case IP_RECVRSSBUCKETID:
1305 OPTSET2(INP_RECVRSSBUCKETID, optval);
1314 * Multicast socket options are processed by the in_mcast
1317 case IP_MULTICAST_IF:
1318 case IP_MULTICAST_VIF:
1319 case IP_MULTICAST_TTL:
1320 case IP_MULTICAST_LOOP:
1321 case IP_ADD_MEMBERSHIP:
1322 case IP_DROP_MEMBERSHIP:
1323 case IP_ADD_SOURCE_MEMBERSHIP:
1324 case IP_DROP_SOURCE_MEMBERSHIP:
1325 case IP_BLOCK_SOURCE:
1326 case IP_UNBLOCK_SOURCE:
1328 case MCAST_JOIN_GROUP:
1329 case MCAST_LEAVE_GROUP:
1330 case MCAST_JOIN_SOURCE_GROUP:
1331 case MCAST_LEAVE_SOURCE_GROUP:
1332 case MCAST_BLOCK_SOURCE:
1333 case MCAST_UNBLOCK_SOURCE:
1334 error = inp_setmoptions(inp, sopt);
1338 error = sooptcopyin(sopt, &optval, sizeof optval,
1345 case IP_PORTRANGE_DEFAULT:
1346 inp->inp_flags &= ~(INP_LOWPORT);
1347 inp->inp_flags &= ~(INP_HIGHPORT);
1350 case IP_PORTRANGE_HIGH:
1351 inp->inp_flags &= ~(INP_LOWPORT);
1352 inp->inp_flags |= INP_HIGHPORT;
1355 case IP_PORTRANGE_LOW:
1356 inp->inp_flags &= ~(INP_HIGHPORT);
1357 inp->inp_flags |= INP_LOWPORT;
1367 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1368 case IP_IPSEC_POLICY:
1369 if (IPSEC_ENABLED(ipv4)) {
1370 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1377 error = ENOPROTOOPT;
1383 switch (sopt->sopt_name) {
1387 if (inp->inp_options) {
1388 struct mbuf *options;
1390 options = m_copym(inp->inp_options, 0,
1391 M_COPYALL, M_NOWAIT);
1393 if (options != NULL) {
1394 error = sooptcopyout(sopt,
1395 mtod(options, char *),
1402 sopt->sopt_valsize = 0;
1410 case IP_RECVRETOPTS:
1411 case IP_ORIGDSTADDR:
1412 case IP_RECVDSTADDR:
1425 case IP_RSSBUCKETID:
1426 case IP_RECVRSSBUCKETID:
1428 switch (sopt->sopt_name) {
1430 optval = inp->inp_ip_tos;
1434 optval = inp->inp_ip_ttl;
1438 optval = inp->inp_ip_minttl;
1441 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1442 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1445 optval = OPTBIT(INP_RECVOPTS);
1448 case IP_RECVRETOPTS:
1449 optval = OPTBIT(INP_RECVRETOPTS);
1452 case IP_RECVDSTADDR:
1453 optval = OPTBIT(INP_RECVDSTADDR);
1456 case IP_ORIGDSTADDR:
1457 optval = OPTBIT2(INP_ORIGDSTADDR);
1461 optval = OPTBIT(INP_RECVTTL);
1465 optval = OPTBIT(INP_RECVIF);
1469 if (inp->inp_flags & INP_HIGHPORT)
1470 optval = IP_PORTRANGE_HIGH;
1471 else if (inp->inp_flags & INP_LOWPORT)
1472 optval = IP_PORTRANGE_LOW;
1478 optval = OPTBIT(INP_ONESBCAST);
1481 optval = OPTBIT(INP_DONTFRAG);
1484 optval = OPTBIT(INP_BINDANY);
1487 optval = OPTBIT(INP_RECVTOS);
1490 optval = inp->inp_flowid;
1493 optval = inp->inp_flowtype;
1496 optval = OPTBIT2(INP_RECVFLOWID);
1499 case IP_RSSBUCKETID:
1500 retval = rss_hash2bucket(inp->inp_flowid,
1504 optval = rss_bucket;
1508 case IP_RECVRSSBUCKETID:
1509 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1513 optval = OPTBIT2(INP_BINDMULTI);
1516 error = sooptcopyout(sopt, &optval, sizeof optval);
1520 * Multicast socket options are processed by the in_mcast
1523 case IP_MULTICAST_IF:
1524 case IP_MULTICAST_VIF:
1525 case IP_MULTICAST_TTL:
1526 case IP_MULTICAST_LOOP:
1528 error = inp_getmoptions(inp, sopt);
1531 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1532 case IP_IPSEC_POLICY:
1533 if (IPSEC_ENABLED(ipv4)) {
1534 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1541 error = ENOPROTOOPT;
1550 * Routine called from ip_output() to loop back a copy of an IP multicast
1551 * packet to the input queue of a specified interface. Note that this
1552 * calls the output routine of the loopback "driver", but with an interface
1553 * pointer that might NOT be a loopback interface -- evil, but easier than
1554 * replicating that code here.
1557 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1563 * Make a deep copy of the packet because we're going to
1564 * modify the pack in order to generate checksums.
1566 copym = m_dup(m, M_NOWAIT);
1567 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1568 copym = m_pullup(copym, hlen);
1569 if (copym != NULL) {
1570 /* If needed, compute the checksum and mark it as valid. */
1571 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1572 in_delayed_cksum(copym);
1573 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1574 copym->m_pkthdr.csum_flags |=
1575 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1576 copym->m_pkthdr.csum_data = 0xffff;
1579 * We don't bother to fragment if the IP length is greater
1580 * than the interface's MTU. Can this possibly matter?
1582 ip = mtod(copym, struct ip *);
1584 ip->ip_sum = in_cksum(copym, hlen);
1585 if_simloop(ifp, copym, AF_INET, 0);