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>
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_vlan_var.h>
65 #include <net/if_llatbl.h>
66 #include <net/ethernet.h>
67 #include <net/netisr.h>
69 #include <net/route.h>
70 #include <net/route/nhop.h>
71 #include <net/rss_config.h>
74 #include <netinet/in.h>
75 #include <netinet/in_fib.h>
76 #include <netinet/in_kdtrace.h>
77 #include <netinet/in_systm.h>
78 #include <netinet/ip.h>
79 #include <netinet/in_fib.h>
80 #include <netinet/in_pcb.h>
81 #include <netinet/in_rss.h>
82 #include <netinet/in_var.h>
83 #include <netinet/ip_var.h>
84 #include <netinet/ip_options.h>
86 #include <netinet/udp.h>
87 #include <netinet/udp_var.h>
89 #if defined(SCTP) || defined(SCTP_SUPPORT)
90 #include <netinet/sctp.h>
91 #include <netinet/sctp_crc32.h>
94 #include <netipsec/ipsec_support.h>
96 #include <machine/in_cksum.h>
98 #include <security/mac/mac_framework.h>
100 #ifdef MBUF_STRESS_TEST
101 static int mbuf_frag_size = 0;
102 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
103 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
106 static void ip_mloopback(struct ifnet *, const struct mbuf *, int);
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;
156 #if defined(SCTP) || defined(SCTP_SUPPORT)
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;
187 #if defined(SCTP) || defined(SCTP_SUPPORT)
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 *gw, struct route *ro, bool stamp_tag)
216 struct ktls_session *tls = NULL;
218 struct m_snd_tag *mst;
221 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
226 * If this is an unencrypted TLS record, save a reference to
227 * the record. This local reference is used to call
228 * ktls_output_eagain after the mbuf has been freed (thus
229 * dropping the mbuf's reference) in if_output.
231 if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
232 tls = ktls_hold(m->m_next->m_epg_tls);
236 * If a TLS session doesn't have a valid tag, it must
237 * have had an earlier ifp mismatch, so drop this
246 * Always stamp tags that include NIC ktls.
252 if (inp != NULL && mst == NULL) {
253 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
254 (inp->inp_snd_tag != NULL &&
255 inp->inp_snd_tag->ifp != ifp))
256 in_pcboutput_txrtlmt(inp, ifp, m);
258 if (inp->inp_snd_tag != NULL)
259 mst = inp->inp_snd_tag;
262 if (stamp_tag && mst != NULL) {
263 KASSERT(m->m_pkthdr.rcvif == NULL,
264 ("trying to add a send tag to a forwarded packet"));
265 if (mst->ifp != ifp) {
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)
324 struct ifnet *ifp = NULL; /* keep compiler happy */
326 int hlen = sizeof (struct ip);
330 struct sockaddr_in *dst;
331 const struct sockaddr *gw;
332 struct in_ifaddr *ia = NULL;
335 uint16_t ip_len, ip_off;
336 struct route iproute;
338 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
339 int no_route_but_check_spd = 0;
346 INP_LOCK_ASSERT(inp);
347 M_SETFIB(m, inp->inp_inc.inc_fibnum);
348 if ((flags & IP_NODEFAULTFLOWID) == 0) {
349 m->m_pkthdr.flowid = inp->inp_flowid;
350 M_HASHTYPE_SET(m, inp->inp_flowtype);
352 if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
353 vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
356 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
362 m = ip_insertoptions(m, opt, &len);
364 hlen = len; /* ip->ip_hl is updated above */
366 ip = mtod(m, struct ip *);
367 ip_len = ntohs(ip->ip_len);
368 ip_off = ntohs(ip->ip_off);
370 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
371 ip->ip_v = IPVERSION;
372 ip->ip_hl = hlen >> 2;
375 /* Header already set, fetch hlen from there */
376 hlen = ip->ip_hl << 2;
378 if ((flags & IP_FORWARDING) == 0)
379 IPSTAT_INC(ips_localout);
384 * gw is readonly but can point either to dst OR rt_gateway,
385 * therefore we need restore gw if we're redoing lookup.
387 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
390 bzero(ro, sizeof (*ro));
392 dst = (struct sockaddr_in *)&ro->ro_dst;
393 if (ro->ro_nh == NULL) {
394 dst->sin_family = AF_INET;
395 dst->sin_len = sizeof(*dst);
396 dst->sin_addr = ip->ip_dst;
398 gw = (const struct sockaddr *)dst;
401 * Validate route against routing table additions;
402 * a better/more specific route might have been added.
404 if (inp != NULL && ro->ro_nh != NULL)
405 NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
407 * If there is a cached route,
408 * check that it is to the same destination
409 * and is still up. If not, free it and try again.
410 * The address family should also be checked in case of sharing the
412 * Also check whether routing cache needs invalidation.
414 if (ro->ro_nh != NULL &&
415 ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
416 dst->sin_addr.s_addr != ip->ip_dst.s_addr))
417 RO_INVALIDATE_CACHE(ro);
420 * If routing to interface only, short circuit routing lookup.
421 * The use of an all-ones broadcast address implies this; an
422 * interface is specified by the broadcast address of an interface,
423 * or the destination address of a ptp interface.
425 if (flags & IP_SENDONES) {
426 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
427 M_GETFIB(m)))) == NULL &&
428 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
429 M_GETFIB(m)))) == NULL) {
430 IPSTAT_INC(ips_noroute);
434 ip->ip_dst.s_addr = INADDR_BROADCAST;
435 dst->sin_addr = ip->ip_dst;
440 src = IA_SIN(ia)->sin_addr;
441 } else if (flags & IP_ROUTETOIF) {
442 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
443 M_GETFIB(m)))) == NULL &&
444 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
445 M_GETFIB(m)))) == NULL) {
446 IPSTAT_INC(ips_noroute);
453 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
454 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
455 src = IA_SIN(ia)->sin_addr;
456 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
457 imo != NULL && imo->imo_multicast_ifp != NULL) {
459 * Bypass the normal routing lookup for multicast
460 * packets if the interface is specified.
462 ifp = imo->imo_multicast_ifp;
465 isbroadcast = 0; /* fool gcc */
466 /* Interface may have no addresses. */
468 src = IA_SIN(ia)->sin_addr;
470 src.s_addr = INADDR_ANY;
471 } else if (ro != &iproute) {
472 if (ro->ro_nh == NULL) {
474 * We want to do any cloning requested by the link
475 * layer, as this is probably required in all cases
476 * for correct operation (as it is for ARP).
479 flowid = m->m_pkthdr.flowid;
480 ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
483 if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
484 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
486 * There is no route for this packet, but it is
487 * possible that a matching SPD entry exists.
489 no_route_but_check_spd = 1;
492 IPSTAT_INC(ips_noroute);
493 error = EHOSTUNREACH;
497 struct nhop_object *nh = ro->ro_nh;
499 ia = ifatoia(nh->nh_ifa);
501 counter_u64_add(nh->nh_pksent, 1);
502 rt_update_ro_flags(ro, nh);
503 if (nh->nh_flags & NHF_GATEWAY)
505 if (nh->nh_flags & NHF_HOST)
506 isbroadcast = (nh->nh_flags & NHF_BROADCAST);
507 else if ((ifp->if_flags & IFF_BROADCAST) && (gw->sa_family == AF_INET))
508 isbroadcast = in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia);
512 src = IA_SIN(ia)->sin_addr;
514 struct nhop_object *nh;
516 nh = fib4_lookup(M_GETFIB(m), dst->sin_addr, 0, NHR_NONE,
519 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
521 * There is no route for this packet, but it is
522 * possible that a matching SPD entry exists.
524 no_route_but_check_spd = 1;
527 IPSTAT_INC(ips_noroute);
528 error = EHOSTUNREACH;
533 rt_update_ro_flags(ro, nh);
534 if (nh->nh_flags & NHF_GATEWAY)
536 ia = ifatoia(nh->nh_ifa);
537 src = IA_SIN(ia)->sin_addr;
538 isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
539 (NHF_HOST | NHF_BROADCAST)) ||
540 ((ifp->if_flags & IFF_BROADCAST) &&
541 (gw->sa_family == AF_INET) &&
542 in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->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.)
557 gw = (const struct sockaddr *)dst;
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;
714 gw = (const struct sockaddr *)dst;
715 ip = mtod(m, struct ip *);
721 EVL_APPLY_PRI(m, vlan_pcp);
723 /* IN_LOOPBACK must not appear on the wire - RFC1122. */
724 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
725 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
726 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
727 IPSTAT_INC(ips_badaddr);
728 error = EADDRNOTAVAIL;
733 /* Ensure the packet data is mapped if the interface requires it. */
734 if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
735 m = mb_unmapped_to_ext(m);
737 IPSTAT_INC(ips_odropped);
743 m->m_pkthdr.csum_flags |= CSUM_IP;
744 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
746 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
748 #if defined(SCTP) || defined(SCTP_SUPPORT)
749 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
750 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
751 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
756 * If small enough for interface, or the interface will take
757 * care of the fragmentation for us, we can just send directly.
758 * Note that if_vxlan could have requested TSO even though the outer
759 * frame is UDP. It is correct to not fragment such datagrams and
760 * instead just pass them on to the driver.
763 (m->m_pkthdr.csum_flags & ifp->if_hwassist &
764 (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
766 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
767 ip->ip_sum = in_cksum(m, hlen);
768 m->m_pkthdr.csum_flags &= ~CSUM_IP;
772 * Record statistics for this interface address.
773 * With CSUM_TSO the byte/packet count will be slightly
774 * incorrect because we count the IP+TCP headers only
775 * once instead of for every generated packet.
777 if (!(flags & IP_FORWARDING) && ia) {
778 if (m->m_pkthdr.csum_flags &
779 (CSUM_TSO | CSUM_INNER_TSO))
780 counter_u64_add(ia->ia_ifa.ifa_opackets,
781 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
783 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
785 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
787 #ifdef MBUF_STRESS_TEST
788 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
789 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
792 * Reset layer specific mbuf flags
793 * to avoid confusing lower layers.
796 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
797 error = ip_output_send(inp, ifp, m, gw, ro,
798 (flags & IP_NO_SND_TAG_RL) ? false : true);
802 /* Balk when DF bit is set or the interface didn't support TSO. */
803 if ((ip_off & IP_DF) ||
804 (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
806 IPSTAT_INC(ips_cantfrag);
811 * Too large for interface; fragment if possible. If successful,
812 * on return, m will point to a list of packets to be sent.
814 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
821 /* Record statistics for this interface address. */
823 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
824 counter_u64_add(ia->ia_ifa.ifa_obytes,
828 * Reset layer specific mbuf flags
829 * to avoid confusing upper layers.
833 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
834 mtod(m, struct ip *), NULL);
835 error = ip_output_send(inp, ifp, m, gw, ro, true);
841 IPSTAT_INC(ips_fragmented);
851 * Create a chain of fragments which fit the given mtu. m_frag points to the
852 * mbuf to be fragmented; on return it points to the chain with the fragments.
853 * Return 0 if no error. If error, m_frag may contain a partially built
854 * chain of fragments that should be freed by the caller.
856 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
859 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
860 u_long if_hwassist_flags)
863 int hlen = ip->ip_hl << 2;
864 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
866 struct mbuf *m0 = *m_frag; /* the original packet */
870 uint16_t ip_len, ip_off;
872 ip_len = ntohs(ip->ip_len);
873 ip_off = ntohs(ip->ip_off);
875 if (ip_off & IP_DF) { /* Fragmentation not allowed */
876 IPSTAT_INC(ips_cantfrag);
881 * Must be able to put at least 8 bytes per fragment.
887 * If the interface will not calculate checksums on
888 * fragmented packets, then do it here.
890 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
891 in_delayed_cksum(m0);
892 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
894 #if defined(SCTP) || defined(SCTP_SUPPORT)
895 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
896 sctp_delayed_cksum(m0, hlen);
897 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
900 if (len > PAGE_SIZE) {
902 * Fragment large datagrams such that each segment
903 * contains a multiple of PAGE_SIZE amount of data,
904 * plus headers. This enables a receiver to perform
905 * page-flipping zero-copy optimizations.
907 * XXX When does this help given that sender and receiver
908 * could have different page sizes, and also mtu could
909 * be less than the receiver's page size ?
913 off = MIN(mtu, m0->m_pkthdr.len);
916 * firstlen (off - hlen) must be aligned on an
920 goto smart_frag_failure;
921 off = ((off - hlen) & ~7) + hlen;
922 newlen = (~PAGE_MASK) & mtu;
923 if ((newlen + sizeof (struct ip)) > mtu) {
924 /* we failed, go back the default */
935 firstlen = off - hlen;
936 mnext = &m0->m_nextpkt; /* pointer to next packet */
939 * Loop through length of segment after first fragment,
940 * make new header and copy data of each part and link onto chain.
941 * Here, m0 is the original packet, m is the fragment being created.
942 * The fragments are linked off the m_nextpkt of the original
943 * packet, which after processing serves as the first fragment.
945 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
946 struct ip *mhip; /* ip header on the fragment */
948 int mhlen = sizeof (struct ip);
950 m = m_gethdr(M_NOWAIT, MT_DATA);
953 IPSTAT_INC(ips_odropped);
957 * Make sure the complete packet header gets copied
958 * from the originating mbuf to the newly created
959 * mbuf. This also ensures that existing firewall
960 * classification(s), VLAN tags and so on get copied
961 * to the resulting fragmented packet(s):
963 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
966 IPSTAT_INC(ips_odropped);
970 * In the first mbuf, leave room for the link header, then
971 * copy the original IP header including options. The payload
972 * goes into an additional mbuf chain returned by m_copym().
974 m->m_data += max_linkhdr;
975 mhip = mtod(m, struct ip *);
977 if (hlen > sizeof (struct ip)) {
978 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
979 mhip->ip_v = IPVERSION;
980 mhip->ip_hl = mhlen >> 2;
983 /* XXX do we need to add ip_off below ? */
984 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
985 if (off + len >= ip_len)
988 mhip->ip_off |= IP_MF;
989 mhip->ip_len = htons((u_short)(len + mhlen));
990 m->m_next = m_copym(m0, off, len, M_NOWAIT);
991 if (m->m_next == NULL) { /* copy failed */
993 error = ENOBUFS; /* ??? */
994 IPSTAT_INC(ips_odropped);
997 m->m_pkthdr.len = mhlen + len;
999 mac_netinet_fragment(m0, m);
1001 mhip->ip_off = htons(mhip->ip_off);
1003 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1004 mhip->ip_sum = in_cksum(m, mhlen);
1005 m->m_pkthdr.csum_flags &= ~CSUM_IP;
1008 mnext = &m->m_nextpkt;
1010 IPSTAT_ADD(ips_ofragments, nfrags);
1013 * Update first fragment by trimming what's been copied out
1014 * and updating header.
1016 m_adj(m0, hlen + firstlen - ip_len);
1017 m0->m_pkthdr.len = hlen + firstlen;
1018 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1019 ip->ip_off = htons(ip_off | IP_MF);
1021 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1022 ip->ip_sum = in_cksum(m0, hlen);
1023 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1032 in_delayed_cksum(struct mbuf *m)
1036 uint16_t cklen, csum, offset;
1038 ip = mtod(m, struct ip *);
1039 offset = ip->ip_hl << 2 ;
1041 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1042 /* if udp header is not in the first mbuf copy udplen */
1043 if (offset + sizeof(struct udphdr) > m->m_len) {
1044 m_copydata(m, offset + offsetof(struct udphdr,
1045 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1046 cklen = ntohs(cklen);
1048 uh = (struct udphdr *)mtodo(m, offset);
1049 cklen = ntohs(uh->uh_ulen);
1051 csum = in_cksum_skip(m, cklen + offset, offset);
1055 cklen = ntohs(ip->ip_len);
1056 csum = in_cksum_skip(m, cklen, offset);
1058 offset += m->m_pkthdr.csum_data; /* checksum offset */
1060 if (offset + sizeof(csum) > m->m_len)
1061 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1063 *(u_short *)mtodo(m, offset) = csum;
1067 * IP socket option processing.
1070 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1072 struct inpcb *inp = sotoinpcb(so);
1075 uint32_t rss_bucket;
1080 if (sopt->sopt_level != IPPROTO_IP) {
1083 if (sopt->sopt_level == SOL_SOCKET &&
1084 sopt->sopt_dir == SOPT_SET) {
1085 switch (sopt->sopt_name) {
1088 if ((so->so_options & SO_REUSEADDR) != 0)
1089 inp->inp_flags2 |= INP_REUSEADDR;
1091 inp->inp_flags2 &= ~INP_REUSEADDR;
1097 if ((so->so_options & SO_REUSEPORT) != 0)
1098 inp->inp_flags2 |= INP_REUSEPORT;
1100 inp->inp_flags2 &= ~INP_REUSEPORT;
1104 case SO_REUSEPORT_LB:
1106 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1107 inp->inp_flags2 |= INP_REUSEPORT_LB;
1109 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1115 inp->inp_inc.inc_fibnum = so->so_fibnum;
1119 case SO_MAX_PACING_RATE:
1122 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1136 switch (sopt->sopt_dir) {
1138 switch (sopt->sopt_name) {
1145 if (sopt->sopt_valsize > MLEN) {
1149 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1154 m->m_len = sopt->sopt_valsize;
1155 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1162 error = ip_pcbopts(inp, sopt->sopt_name, m);
1168 if (sopt->sopt_td != NULL) {
1169 error = priv_check(sopt->sopt_td,
1170 PRIV_NETINET_BINDANY);
1177 case IP_RSS_LISTEN_BUCKET:
1183 case IP_RECVRETOPTS:
1184 case IP_ORIGDSTADDR:
1185 case IP_RECVDSTADDR:
1193 case IP_RECVRSSBUCKETID:
1196 error = sooptcopyin(sopt, &optval, sizeof optval,
1201 switch (sopt->sopt_name) {
1203 inp->inp_ip_tos = optval;
1207 inp->inp_ip_ttl = optval;
1211 if (optval >= 0 && optval <= MAXTTL)
1212 inp->inp_ip_minttl = optval;
1217 #define OPTSET(bit) do { \
1220 inp->inp_flags |= bit; \
1222 inp->inp_flags &= ~bit; \
1226 #define OPTSET2(bit, val) do { \
1229 inp->inp_flags2 |= bit; \
1231 inp->inp_flags2 &= ~bit; \
1236 OPTSET(INP_RECVOPTS);
1239 case IP_RECVRETOPTS:
1240 OPTSET(INP_RECVRETOPTS);
1243 case IP_RECVDSTADDR:
1244 OPTSET(INP_RECVDSTADDR);
1247 case IP_ORIGDSTADDR:
1248 OPTSET2(INP_ORIGDSTADDR, optval);
1252 OPTSET(INP_RECVTTL);
1260 OPTSET(INP_ONESBCAST);
1263 OPTSET(INP_DONTFRAG);
1266 OPTSET(INP_BINDANY);
1269 OPTSET(INP_RECVTOS);
1272 OPTSET2(INP_BINDMULTI, optval);
1275 OPTSET2(INP_RECVFLOWID, optval);
1278 case IP_RSS_LISTEN_BUCKET:
1279 if ((optval >= 0) &&
1280 (optval < rss_getnumbuckets())) {
1281 inp->inp_rss_listen_bucket = optval;
1282 OPTSET2(INP_RSS_BUCKET_SET, 1);
1287 case IP_RECVRSSBUCKETID:
1288 OPTSET2(INP_RECVRSSBUCKETID, optval);
1292 if ((optval >= -1) && (optval <=
1293 (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1307 optval << INP_2PCP_SHIFT;
1319 * Multicast socket options are processed by the in_mcast
1322 case IP_MULTICAST_IF:
1323 case IP_MULTICAST_VIF:
1324 case IP_MULTICAST_TTL:
1325 case IP_MULTICAST_LOOP:
1326 case IP_ADD_MEMBERSHIP:
1327 case IP_DROP_MEMBERSHIP:
1328 case IP_ADD_SOURCE_MEMBERSHIP:
1329 case IP_DROP_SOURCE_MEMBERSHIP:
1330 case IP_BLOCK_SOURCE:
1331 case IP_UNBLOCK_SOURCE:
1333 case MCAST_JOIN_GROUP:
1334 case MCAST_LEAVE_GROUP:
1335 case MCAST_JOIN_SOURCE_GROUP:
1336 case MCAST_LEAVE_SOURCE_GROUP:
1337 case MCAST_BLOCK_SOURCE:
1338 case MCAST_UNBLOCK_SOURCE:
1339 error = inp_setmoptions(inp, sopt);
1343 error = sooptcopyin(sopt, &optval, sizeof optval,
1350 case IP_PORTRANGE_DEFAULT:
1351 inp->inp_flags &= ~(INP_LOWPORT);
1352 inp->inp_flags &= ~(INP_HIGHPORT);
1355 case IP_PORTRANGE_HIGH:
1356 inp->inp_flags &= ~(INP_LOWPORT);
1357 inp->inp_flags |= INP_HIGHPORT;
1360 case IP_PORTRANGE_LOW:
1361 inp->inp_flags &= ~(INP_HIGHPORT);
1362 inp->inp_flags |= INP_LOWPORT;
1372 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1373 case IP_IPSEC_POLICY:
1374 if (IPSEC_ENABLED(ipv4)) {
1375 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1382 error = ENOPROTOOPT;
1388 switch (sopt->sopt_name) {
1392 if (inp->inp_options) {
1393 struct mbuf *options;
1395 options = m_copym(inp->inp_options, 0,
1396 M_COPYALL, M_NOWAIT);
1398 if (options != NULL) {
1399 error = sooptcopyout(sopt,
1400 mtod(options, char *),
1407 sopt->sopt_valsize = 0;
1415 case IP_RECVRETOPTS:
1416 case IP_ORIGDSTADDR:
1417 case IP_RECVDSTADDR:
1430 case IP_RSSBUCKETID:
1431 case IP_RECVRSSBUCKETID:
1434 switch (sopt->sopt_name) {
1436 optval = inp->inp_ip_tos;
1440 optval = inp->inp_ip_ttl;
1444 optval = inp->inp_ip_minttl;
1447 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1448 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1451 optval = OPTBIT(INP_RECVOPTS);
1454 case IP_RECVRETOPTS:
1455 optval = OPTBIT(INP_RECVRETOPTS);
1458 case IP_RECVDSTADDR:
1459 optval = OPTBIT(INP_RECVDSTADDR);
1462 case IP_ORIGDSTADDR:
1463 optval = OPTBIT2(INP_ORIGDSTADDR);
1467 optval = OPTBIT(INP_RECVTTL);
1471 optval = OPTBIT(INP_RECVIF);
1475 if (inp->inp_flags & INP_HIGHPORT)
1476 optval = IP_PORTRANGE_HIGH;
1477 else if (inp->inp_flags & INP_LOWPORT)
1478 optval = IP_PORTRANGE_LOW;
1484 optval = OPTBIT(INP_ONESBCAST);
1487 optval = OPTBIT(INP_DONTFRAG);
1490 optval = OPTBIT(INP_BINDANY);
1493 optval = OPTBIT(INP_RECVTOS);
1496 optval = inp->inp_flowid;
1499 optval = inp->inp_flowtype;
1502 optval = OPTBIT2(INP_RECVFLOWID);
1505 case IP_RSSBUCKETID:
1506 retval = rss_hash2bucket(inp->inp_flowid,
1510 optval = rss_bucket;
1514 case IP_RECVRSSBUCKETID:
1515 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1519 optval = OPTBIT2(INP_BINDMULTI);
1522 if (OPTBIT2(INP_2PCP_SET)) {
1523 optval = (inp->inp_flags2 &
1524 INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1530 error = sooptcopyout(sopt, &optval, sizeof optval);
1534 * Multicast socket options are processed by the in_mcast
1537 case IP_MULTICAST_IF:
1538 case IP_MULTICAST_VIF:
1539 case IP_MULTICAST_TTL:
1540 case IP_MULTICAST_LOOP:
1542 error = inp_getmoptions(inp, sopt);
1545 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1546 case IP_IPSEC_POLICY:
1547 if (IPSEC_ENABLED(ipv4)) {
1548 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1555 error = ENOPROTOOPT;
1564 * Routine called from ip_output() to loop back a copy of an IP multicast
1565 * packet to the input queue of a specified interface. Note that this
1566 * calls the output routine of the loopback "driver", but with an interface
1567 * pointer that might NOT be a loopback interface -- evil, but easier than
1568 * replicating that code here.
1571 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1577 * Make a deep copy of the packet because we're going to
1578 * modify the pack in order to generate checksums.
1580 copym = m_dup(m, M_NOWAIT);
1581 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1582 copym = m_pullup(copym, hlen);
1583 if (copym != NULL) {
1584 /* If needed, compute the checksum and mark it as valid. */
1585 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1586 in_delayed_cksum(copym);
1587 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1588 copym->m_pkthdr.csum_flags |=
1589 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1590 copym->m_pkthdr.csum_data = 0xffff;
1593 * We don't bother to fragment if the IP length is greater
1594 * than the interface's MTU. Can this possibly matter?
1596 ip = mtod(copym, struct ip *);
1598 ip->ip_sum = in_cksum(copym, hlen);
1599 if_simloop(ifp, copym, AF_INET, 0);