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_ratelimit.h"
39 #include "opt_ipsec.h"
40 #include "opt_mbuf_stress_test.h"
41 #include "opt_mpath.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, struct inpcb *inp,
113 struct sockaddr_in *dst, int *fibnum, int *error)
115 struct m_tag *fwd_tag = NULL;
121 ip = mtod(m, struct ip *);
123 /* Run through list of hooks for output packets. */
124 odst.s_addr = ip->ip_dst.s_addr;
125 switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, PFIL_OUT, inp)) {
130 return 1; /* Finished */
135 ip = mtod(m, struct ip *);
137 /* See if destination IP address was changed by packet filter. */
138 if (odst.s_addr != ip->ip_dst.s_addr) {
139 m->m_flags |= M_SKIP_FIREWALL;
140 /* If destination is now ourself drop to ip_input(). */
141 if (in_localip(ip->ip_dst)) {
142 m->m_flags |= M_FASTFWD_OURS;
143 if (m->m_pkthdr.rcvif == NULL)
144 m->m_pkthdr.rcvif = V_loif;
145 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
146 m->m_pkthdr.csum_flags |=
147 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
148 m->m_pkthdr.csum_data = 0xffff;
150 m->m_pkthdr.csum_flags |=
151 CSUM_IP_CHECKED | CSUM_IP_VALID;
153 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
154 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
156 *error = netisr_queue(NETISR_IP, m);
157 return 1; /* Finished */
160 bzero(dst, sizeof(*dst));
161 dst->sin_family = AF_INET;
162 dst->sin_len = sizeof(*dst);
163 dst->sin_addr = ip->ip_dst;
165 return -1; /* Reloop */
167 /* See if fib was changed by packet filter. */
168 if ((*fibnum) != M_GETFIB(m)) {
169 m->m_flags |= M_SKIP_FIREWALL;
170 *fibnum = M_GETFIB(m);
171 return -1; /* Reloop for FIB change */
174 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
175 if (m->m_flags & M_FASTFWD_OURS) {
176 if (m->m_pkthdr.rcvif == NULL)
177 m->m_pkthdr.rcvif = V_loif;
178 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
179 m->m_pkthdr.csum_flags |=
180 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
181 m->m_pkthdr.csum_data = 0xffff;
184 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
185 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
187 m->m_pkthdr.csum_flags |=
188 CSUM_IP_CHECKED | CSUM_IP_VALID;
190 *error = netisr_queue(NETISR_IP, m);
191 return 1; /* Finished */
193 /* Or forward to some other address? */
194 if ((m->m_flags & M_IP_NEXTHOP) &&
195 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
196 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
197 m->m_flags |= M_SKIP_FIREWALL;
198 m->m_flags &= ~M_IP_NEXTHOP;
199 m_tag_delete(m, fwd_tag);
201 return -1; /* Reloop for CHANGE of dst */
208 * IP output. The packet in mbuf chain m contains a skeletal IP
209 * header (with len, off, ttl, proto, tos, src, dst).
210 * The mbuf chain containing the packet will be freed.
211 * The mbuf opt, if present, will not be freed.
212 * If route ro is present and has ro_rt initialized, route lookup would be
213 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
214 * then result of route lookup is stored in ro->ro_rt.
216 * In the IP forwarding case, the packet will arrive with options already
217 * inserted, so must have a NULL opt pointer.
220 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
221 struct ip_moptions *imo, struct inpcb *inp)
223 struct rm_priotracker in_ifa_tracker;
224 struct epoch_tracker et;
226 struct ifnet *ifp = NULL; /* keep compiler happy */
228 int hlen = sizeof (struct ip);
231 struct sockaddr_in *dst, sin;
232 const struct sockaddr_in *gw;
233 struct in_ifaddr *ia;
236 uint16_t ip_len, ip_off;
238 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
239 int no_route_but_check_spd = 0;
244 INP_LOCK_ASSERT(inp);
245 M_SETFIB(m, inp->inp_inc.inc_fibnum);
246 if ((flags & IP_NODEFAULTFLOWID) == 0) {
247 m->m_pkthdr.flowid = inp->inp_flowid;
248 M_HASHTYPE_SET(m, inp->inp_flowtype);
251 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
257 m = ip_insertoptions(m, opt, &len);
259 hlen = len; /* ip->ip_hl is updated above */
261 ip = mtod(m, struct ip *);
262 ip_len = ntohs(ip->ip_len);
263 ip_off = ntohs(ip->ip_off);
265 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
266 ip->ip_v = IPVERSION;
267 ip->ip_hl = hlen >> 2;
270 /* Header already set, fetch hlen from there */
271 hlen = ip->ip_hl << 2;
273 if ((flags & IP_FORWARDING) == 0)
274 IPSTAT_INC(ips_localout);
279 * gw is readonly but can point either to dst OR rt_gateway,
280 * therefore we need restore gw if we're redoing lookup.
282 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
284 dst = (struct sockaddr_in *)&ro->ro_dst;
287 if (ro == NULL || ro->ro_rt == NULL) {
288 bzero(dst, sizeof(*dst));
289 dst->sin_family = AF_INET;
290 dst->sin_len = sizeof(*dst);
291 dst->sin_addr = ip->ip_dst;
297 * Validate route against routing table additions;
298 * a better/more specific route might have been added.
300 if (inp != NULL && ro != NULL && ro->ro_rt != NULL)
301 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
303 * If there is a cached route,
304 * check that it is to the same destination
305 * and is still up. If not, free it and try again.
306 * The address family should also be checked in case of sharing the
308 * Also check whether routing cache needs invalidation.
310 if (ro != NULL && ro->ro_rt != NULL &&
311 ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
312 ro->ro_rt->rt_ifp == NULL || !RT_LINK_IS_UP(ro->ro_rt->rt_ifp) ||
313 dst->sin_family != AF_INET ||
314 dst->sin_addr.s_addr != ip->ip_dst.s_addr))
315 RO_INVALIDATE_CACHE(ro);
318 * If routing to interface only, short circuit routing lookup.
319 * The use of an all-ones broadcast address implies this; an
320 * interface is specified by the broadcast address of an interface,
321 * or the destination address of a ptp interface.
323 if (flags & IP_SENDONES) {
324 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
325 M_GETFIB(m)))) == NULL &&
326 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
327 M_GETFIB(m)))) == NULL) {
328 IPSTAT_INC(ips_noroute);
332 ip->ip_dst.s_addr = INADDR_BROADCAST;
333 dst->sin_addr = ip->ip_dst;
338 src = IA_SIN(ia)->sin_addr;
339 } else if (flags & IP_ROUTETOIF) {
340 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
341 M_GETFIB(m)))) == NULL &&
342 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
343 M_GETFIB(m)))) == NULL) {
344 IPSTAT_INC(ips_noroute);
351 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
352 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
353 src = IA_SIN(ia)->sin_addr;
354 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
355 imo != NULL && imo->imo_multicast_ifp != NULL) {
357 * Bypass the normal routing lookup for multicast
358 * packets if the interface is specified.
360 ifp = imo->imo_multicast_ifp;
362 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
363 isbroadcast = 0; /* fool gcc */
364 /* Interface may have no addresses. */
366 src = IA_SIN(ia)->sin_addr;
368 src.s_addr = INADDR_ANY;
369 } else if (ro != NULL) {
370 if (ro->ro_rt == NULL) {
372 * We want to do any cloning requested by the link
373 * layer, as this is probably required in all cases
374 * for correct operation (as it is for ARP).
377 rtalloc_mpath_fib(ro,
378 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
381 in_rtalloc_ign(ro, 0, fibnum);
383 if (ro->ro_rt == NULL ||
384 (ro->ro_rt->rt_flags & RTF_UP) == 0 ||
385 ro->ro_rt->rt_ifp == NULL ||
386 !RT_LINK_IS_UP(ro->ro_rt->rt_ifp)) {
387 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
389 * There is no route for this packet, but it is
390 * possible that a matching SPD entry exists.
392 no_route_but_check_spd = 1;
393 mtu = 0; /* Silence GCC warning. */
396 IPSTAT_INC(ips_noroute);
397 error = EHOSTUNREACH;
401 ia = ifatoia(ro->ro_rt->rt_ifa);
402 ifp = ro->ro_rt->rt_ifp;
403 counter_u64_add(ro->ro_rt->rt_pksent, 1);
404 rt_update_ro_flags(ro);
405 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
406 gw = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
407 if (ro->ro_rt->rt_flags & RTF_HOST)
408 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
409 else if (ifp->if_flags & IFF_BROADCAST)
410 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
413 if (ro->ro_rt->rt_flags & RTF_HOST)
414 mtu = ro->ro_rt->rt_mtu;
417 src = IA_SIN(ia)->sin_addr;
419 struct nhop4_extended nh;
421 bzero(&nh, sizeof(nh));
422 if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh) !=
424 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
426 * There is no route for this packet, but it is
427 * possible that a matching SPD entry exists.
429 no_route_but_check_spd = 1;
430 mtu = 0; /* Silence GCC warning. */
433 IPSTAT_INC(ips_noroute);
434 error = EHOSTUNREACH;
440 * We are rewriting here dst to be gw actually, contradicting
441 * comment at the beginning of the function. However, in this
442 * case we are always dealing with on stack dst.
443 * In case if pfil(9) sends us back to beginning of the
444 * function, the dst would be rewritten by ip_output_pfil().
447 dst->sin_addr = nh.nh_addr;
450 isbroadcast = (((nh.nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
451 (NHF_HOST | NHF_BROADCAST)) ||
452 ((ifp->if_flags & IFF_BROADCAST) &&
453 in_ifaddr_broadcast(dst->sin_addr, ia)));
456 /* Catch a possible divide by zero later. */
457 KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (rt_flags=0x%08x) ifp=%p",
459 (ro != NULL && ro->ro_rt != NULL) ? ro->ro_rt->rt_flags : 0, ifp));
461 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
462 m->m_flags |= M_MCAST;
464 * IP destination address is multicast. Make sure "gw"
465 * still points to the address in "ro". (It may have been
466 * changed to point to a gateway address, above.)
470 * See if the caller provided any multicast options
473 ip->ip_ttl = imo->imo_multicast_ttl;
474 if (imo->imo_multicast_vif != -1)
477 ip_mcast_src(imo->imo_multicast_vif) :
480 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
482 * Confirm that the outgoing interface supports multicast.
484 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
485 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
486 IPSTAT_INC(ips_noroute);
492 * If source address not specified yet, use address
493 * of outgoing interface.
495 if (ip->ip_src.s_addr == INADDR_ANY)
498 if ((imo == NULL && in_mcast_loop) ||
499 (imo && imo->imo_multicast_loop)) {
501 * Loop back multicast datagram if not expressly
502 * forbidden to do so, even if we are not a member
503 * of the group; ip_input() will filter it later,
504 * thus deferring a hash lookup and mutex acquisition
505 * at the expense of a cheap copy using m_copym().
507 ip_mloopback(ifp, m, hlen);
510 * If we are acting as a multicast router, perform
511 * multicast forwarding as if the packet had just
512 * arrived on the interface to which we are about
513 * to send. The multicast forwarding function
514 * recursively calls this function, using the
515 * IP_FORWARDING flag to prevent infinite recursion.
517 * Multicasts that are looped back by ip_mloopback(),
518 * above, will be forwarded by the ip_input() routine,
521 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
523 * If rsvp daemon is not running, do not
524 * set ip_moptions. This ensures that the packet
525 * is multicast and not just sent down one link
526 * as prescribed by rsvpd.
531 ip_mforward(ip, ifp, m, imo) != 0) {
539 * Multicasts with a time-to-live of zero may be looped-
540 * back, above, but must not be transmitted on a network.
541 * Also, multicasts addressed to the loopback interface
542 * are not sent -- the above call to ip_mloopback() will
543 * loop back a copy. ip_input() will drop the copy if
544 * this host does not belong to the destination group on
545 * the loopback interface.
547 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
556 * If the source address is not specified yet, use the address
557 * of the outoing interface.
559 if (ip->ip_src.s_addr == INADDR_ANY)
563 * Look for broadcast address and
564 * verify user is allowed to send
568 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
569 error = EADDRNOTAVAIL;
572 if ((flags & IP_ALLOWBROADCAST) == 0) {
576 /* don't allow broadcast messages to be fragmented */
581 m->m_flags |= M_BCAST;
583 m->m_flags &= ~M_BCAST;
587 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
588 if (IPSEC_ENABLED(ipv4)) {
589 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
590 if (error == EINPROGRESS)
596 * Check if there was a route for this packet; return error if not.
598 if (no_route_but_check_spd) {
599 IPSTAT_INC(ips_noroute);
600 error = EHOSTUNREACH;
603 /* Update variables that are affected by ipsec4_output(). */
604 ip = mtod(m, struct ip *);
605 hlen = ip->ip_hl << 2;
608 /* Jump over all PFIL processing if hooks are not active. */
609 if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
610 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
611 case 1: /* Finished */
614 case 0: /* Continue normally */
615 ip = mtod(m, struct ip *);
618 case -1: /* Need to try again */
619 /* Reset everything for a new round */
622 ro->ro_prepend = NULL;
625 ip = mtod(m, struct ip *);
631 /* IN_LOOPBACK must not appear on the wire - RFC1122. */
632 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
633 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
634 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
635 IPSTAT_INC(ips_badaddr);
636 error = EADDRNOTAVAIL;
641 m->m_pkthdr.csum_flags |= CSUM_IP;
642 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
644 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
647 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
648 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
649 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
654 * If small enough for interface, or the interface will take
655 * care of the fragmentation for us, we can just send directly.
658 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
660 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
661 ip->ip_sum = in_cksum(m, hlen);
662 m->m_pkthdr.csum_flags &= ~CSUM_IP;
666 * Record statistics for this interface address.
667 * With CSUM_TSO the byte/packet count will be slightly
668 * incorrect because we count the IP+TCP headers only
669 * once instead of for every generated packet.
671 if (!(flags & IP_FORWARDING) && ia) {
672 if (m->m_pkthdr.csum_flags & CSUM_TSO)
673 counter_u64_add(ia->ia_ifa.ifa_opackets,
674 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
676 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
678 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
680 #ifdef MBUF_STRESS_TEST
681 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
682 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
685 * Reset layer specific mbuf flags
686 * to avoid confusing lower layers.
689 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
692 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
693 in_pcboutput_txrtlmt(inp, ifp, m);
694 /* stamp send tag on mbuf */
695 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
697 m->m_pkthdr.snd_tag = NULL;
700 error = (*ifp->if_output)(ifp, m,
701 (const struct sockaddr *)gw, ro);
703 /* check for route change */
705 in_pcboutput_eagain(inp);
710 /* Balk when DF bit is set or the interface didn't support TSO. */
711 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
713 IPSTAT_INC(ips_cantfrag);
718 * Too large for interface; fragment if possible. If successful,
719 * on return, m will point to a list of packets to be sent.
721 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
728 /* Record statistics for this interface address. */
730 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
731 counter_u64_add(ia->ia_ifa.ifa_obytes,
735 * Reset layer specific mbuf flags
736 * to avoid confusing upper layers.
740 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
741 mtod(m, struct ip *), NULL);
744 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
745 in_pcboutput_txrtlmt(inp, ifp, m);
746 /* stamp send tag on mbuf */
747 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
749 m->m_pkthdr.snd_tag = NULL;
752 error = (*ifp->if_output)(ifp, m,
753 (const struct sockaddr *)gw, ro);
755 /* check for route change */
757 in_pcboutput_eagain(inp);
764 IPSTAT_INC(ips_fragmented);
775 * Create a chain of fragments which fit the given mtu. m_frag points to the
776 * mbuf to be fragmented; on return it points to the chain with the fragments.
777 * Return 0 if no error. If error, m_frag may contain a partially built
778 * chain of fragments that should be freed by the caller.
780 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
783 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
784 u_long if_hwassist_flags)
787 int hlen = ip->ip_hl << 2;
788 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
790 struct mbuf *m0 = *m_frag; /* the original packet */
794 uint16_t ip_len, ip_off;
796 ip_len = ntohs(ip->ip_len);
797 ip_off = ntohs(ip->ip_off);
799 if (ip_off & IP_DF) { /* Fragmentation not allowed */
800 IPSTAT_INC(ips_cantfrag);
805 * Must be able to put at least 8 bytes per fragment.
811 * If the interface will not calculate checksums on
812 * fragmented packets, then do it here.
814 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
815 in_delayed_cksum(m0);
816 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
819 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
820 sctp_delayed_cksum(m0, hlen);
821 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
824 if (len > PAGE_SIZE) {
826 * Fragment large datagrams such that each segment
827 * contains a multiple of PAGE_SIZE amount of data,
828 * plus headers. This enables a receiver to perform
829 * page-flipping zero-copy optimizations.
831 * XXX When does this help given that sender and receiver
832 * could have different page sizes, and also mtu could
833 * be less than the receiver's page size ?
837 off = MIN(mtu, m0->m_pkthdr.len);
840 * firstlen (off - hlen) must be aligned on an
844 goto smart_frag_failure;
845 off = ((off - hlen) & ~7) + hlen;
846 newlen = (~PAGE_MASK) & mtu;
847 if ((newlen + sizeof (struct ip)) > mtu) {
848 /* we failed, go back the default */
859 firstlen = off - hlen;
860 mnext = &m0->m_nextpkt; /* pointer to next packet */
863 * Loop through length of segment after first fragment,
864 * make new header and copy data of each part and link onto chain.
865 * Here, m0 is the original packet, m is the fragment being created.
866 * The fragments are linked off the m_nextpkt of the original
867 * packet, which after processing serves as the first fragment.
869 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
870 struct ip *mhip; /* ip header on the fragment */
872 int mhlen = sizeof (struct ip);
874 m = m_gethdr(M_NOWAIT, MT_DATA);
877 IPSTAT_INC(ips_odropped);
881 * Make sure the complete packet header gets copied
882 * from the originating mbuf to the newly created
883 * mbuf. This also ensures that existing firewall
884 * classification(s), VLAN tags and so on get copied
885 * to the resulting fragmented packet(s):
887 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
890 IPSTAT_INC(ips_odropped);
894 * In the first mbuf, leave room for the link header, then
895 * copy the original IP header including options. The payload
896 * goes into an additional mbuf chain returned by m_copym().
898 m->m_data += max_linkhdr;
899 mhip = mtod(m, struct ip *);
901 if (hlen > sizeof (struct ip)) {
902 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
903 mhip->ip_v = IPVERSION;
904 mhip->ip_hl = mhlen >> 2;
907 /* XXX do we need to add ip_off below ? */
908 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
909 if (off + len >= ip_len)
912 mhip->ip_off |= IP_MF;
913 mhip->ip_len = htons((u_short)(len + mhlen));
914 m->m_next = m_copym(m0, off, len, M_NOWAIT);
915 if (m->m_next == NULL) { /* copy failed */
917 error = ENOBUFS; /* ??? */
918 IPSTAT_INC(ips_odropped);
921 m->m_pkthdr.len = mhlen + len;
923 mac_netinet_fragment(m0, m);
925 mhip->ip_off = htons(mhip->ip_off);
927 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
928 mhip->ip_sum = in_cksum(m, mhlen);
929 m->m_pkthdr.csum_flags &= ~CSUM_IP;
932 mnext = &m->m_nextpkt;
934 IPSTAT_ADD(ips_ofragments, nfrags);
937 * Update first fragment by trimming what's been copied out
938 * and updating header.
940 m_adj(m0, hlen + firstlen - ip_len);
941 m0->m_pkthdr.len = hlen + firstlen;
942 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
943 ip->ip_off = htons(ip_off | IP_MF);
945 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
946 ip->ip_sum = in_cksum(m0, hlen);
947 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
956 in_delayed_cksum(struct mbuf *m)
960 uint16_t cklen, csum, offset;
962 ip = mtod(m, struct ip *);
963 offset = ip->ip_hl << 2 ;
965 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
966 /* if udp header is not in the first mbuf copy udplen */
967 if (offset + sizeof(struct udphdr) > m->m_len) {
968 m_copydata(m, offset + offsetof(struct udphdr,
969 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
970 cklen = ntohs(cklen);
972 uh = (struct udphdr *)mtodo(m, offset);
973 cklen = ntohs(uh->uh_ulen);
975 csum = in_cksum_skip(m, cklen + offset, offset);
979 cklen = ntohs(ip->ip_len);
980 csum = in_cksum_skip(m, cklen, offset);
982 offset += m->m_pkthdr.csum_data; /* checksum offset */
984 if (offset + sizeof(csum) > m->m_len)
985 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
987 *(u_short *)mtodo(m, offset) = csum;
991 * IP socket option processing.
994 ip_ctloutput(struct socket *so, struct sockopt *sopt)
996 struct inpcb *inp = sotoinpcb(so);
1004 if (sopt->sopt_level != IPPROTO_IP) {
1007 if (sopt->sopt_level == SOL_SOCKET &&
1008 sopt->sopt_dir == SOPT_SET) {
1009 switch (sopt->sopt_name) {
1012 if ((so->so_options & SO_REUSEADDR) != 0)
1013 inp->inp_flags2 |= INP_REUSEADDR;
1015 inp->inp_flags2 &= ~INP_REUSEADDR;
1021 if ((so->so_options & SO_REUSEPORT) != 0)
1022 inp->inp_flags2 |= INP_REUSEPORT;
1024 inp->inp_flags2 &= ~INP_REUSEPORT;
1028 case SO_REUSEPORT_LB:
1030 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1031 inp->inp_flags2 |= INP_REUSEPORT_LB;
1033 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1039 inp->inp_inc.inc_fibnum = so->so_fibnum;
1043 case SO_MAX_PACING_RATE:
1046 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1060 switch (sopt->sopt_dir) {
1062 switch (sopt->sopt_name) {
1069 if (sopt->sopt_valsize > MLEN) {
1073 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1078 m->m_len = sopt->sopt_valsize;
1079 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1086 error = ip_pcbopts(inp, sopt->sopt_name, m);
1092 if (sopt->sopt_td != NULL) {
1093 error = priv_check(sopt->sopt_td,
1094 PRIV_NETINET_BINDANY);
1101 case IP_RSS_LISTEN_BUCKET:
1107 case IP_RECVRETOPTS:
1108 case IP_ORIGDSTADDR:
1109 case IP_RECVDSTADDR:
1117 case IP_RECVRSSBUCKETID:
1119 error = sooptcopyin(sopt, &optval, sizeof optval,
1124 switch (sopt->sopt_name) {
1126 inp->inp_ip_tos = optval;
1130 inp->inp_ip_ttl = optval;
1134 if (optval >= 0 && optval <= MAXTTL)
1135 inp->inp_ip_minttl = optval;
1140 #define OPTSET(bit) do { \
1143 inp->inp_flags |= bit; \
1145 inp->inp_flags &= ~bit; \
1149 #define OPTSET2(bit, val) do { \
1152 inp->inp_flags2 |= bit; \
1154 inp->inp_flags2 &= ~bit; \
1159 OPTSET(INP_RECVOPTS);
1162 case IP_RECVRETOPTS:
1163 OPTSET(INP_RECVRETOPTS);
1166 case IP_RECVDSTADDR:
1167 OPTSET(INP_RECVDSTADDR);
1170 case IP_ORIGDSTADDR:
1171 OPTSET2(INP_ORIGDSTADDR, optval);
1175 OPTSET(INP_RECVTTL);
1183 OPTSET(INP_ONESBCAST);
1186 OPTSET(INP_DONTFRAG);
1189 OPTSET(INP_BINDANY);
1192 OPTSET(INP_RECVTOS);
1195 OPTSET2(INP_BINDMULTI, optval);
1198 OPTSET2(INP_RECVFLOWID, optval);
1201 case IP_RSS_LISTEN_BUCKET:
1202 if ((optval >= 0) &&
1203 (optval < rss_getnumbuckets())) {
1204 inp->inp_rss_listen_bucket = optval;
1205 OPTSET2(INP_RSS_BUCKET_SET, 1);
1210 case IP_RECVRSSBUCKETID:
1211 OPTSET2(INP_RECVRSSBUCKETID, optval);
1220 * Multicast socket options are processed by the in_mcast
1223 case IP_MULTICAST_IF:
1224 case IP_MULTICAST_VIF:
1225 case IP_MULTICAST_TTL:
1226 case IP_MULTICAST_LOOP:
1227 case IP_ADD_MEMBERSHIP:
1228 case IP_DROP_MEMBERSHIP:
1229 case IP_ADD_SOURCE_MEMBERSHIP:
1230 case IP_DROP_SOURCE_MEMBERSHIP:
1231 case IP_BLOCK_SOURCE:
1232 case IP_UNBLOCK_SOURCE:
1234 case MCAST_JOIN_GROUP:
1235 case MCAST_LEAVE_GROUP:
1236 case MCAST_JOIN_SOURCE_GROUP:
1237 case MCAST_LEAVE_SOURCE_GROUP:
1238 case MCAST_BLOCK_SOURCE:
1239 case MCAST_UNBLOCK_SOURCE:
1240 error = inp_setmoptions(inp, sopt);
1244 error = sooptcopyin(sopt, &optval, sizeof optval,
1251 case IP_PORTRANGE_DEFAULT:
1252 inp->inp_flags &= ~(INP_LOWPORT);
1253 inp->inp_flags &= ~(INP_HIGHPORT);
1256 case IP_PORTRANGE_HIGH:
1257 inp->inp_flags &= ~(INP_LOWPORT);
1258 inp->inp_flags |= INP_HIGHPORT;
1261 case IP_PORTRANGE_LOW:
1262 inp->inp_flags &= ~(INP_HIGHPORT);
1263 inp->inp_flags |= INP_LOWPORT;
1273 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1274 case IP_IPSEC_POLICY:
1275 if (IPSEC_ENABLED(ipv4)) {
1276 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1283 error = ENOPROTOOPT;
1289 switch (sopt->sopt_name) {
1293 if (inp->inp_options) {
1294 struct mbuf *options;
1296 options = m_copym(inp->inp_options, 0,
1297 M_COPYALL, M_NOWAIT);
1299 if (options != NULL) {
1300 error = sooptcopyout(sopt,
1301 mtod(options, char *),
1308 sopt->sopt_valsize = 0;
1316 case IP_RECVRETOPTS:
1317 case IP_ORIGDSTADDR:
1318 case IP_RECVDSTADDR:
1331 case IP_RSSBUCKETID:
1332 case IP_RECVRSSBUCKETID:
1334 switch (sopt->sopt_name) {
1337 optval = inp->inp_ip_tos;
1341 optval = inp->inp_ip_ttl;
1345 optval = inp->inp_ip_minttl;
1348 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1349 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1352 optval = OPTBIT(INP_RECVOPTS);
1355 case IP_RECVRETOPTS:
1356 optval = OPTBIT(INP_RECVRETOPTS);
1359 case IP_RECVDSTADDR:
1360 optval = OPTBIT(INP_RECVDSTADDR);
1363 case IP_ORIGDSTADDR:
1364 optval = OPTBIT2(INP_ORIGDSTADDR);
1368 optval = OPTBIT(INP_RECVTTL);
1372 optval = OPTBIT(INP_RECVIF);
1376 if (inp->inp_flags & INP_HIGHPORT)
1377 optval = IP_PORTRANGE_HIGH;
1378 else if (inp->inp_flags & INP_LOWPORT)
1379 optval = IP_PORTRANGE_LOW;
1385 optval = OPTBIT(INP_ONESBCAST);
1388 optval = OPTBIT(INP_DONTFRAG);
1391 optval = OPTBIT(INP_BINDANY);
1394 optval = OPTBIT(INP_RECVTOS);
1397 optval = inp->inp_flowid;
1400 optval = inp->inp_flowtype;
1403 optval = OPTBIT2(INP_RECVFLOWID);
1406 case IP_RSSBUCKETID:
1407 retval = rss_hash2bucket(inp->inp_flowid,
1411 optval = rss_bucket;
1415 case IP_RECVRSSBUCKETID:
1416 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1420 optval = OPTBIT2(INP_BINDMULTI);
1423 error = sooptcopyout(sopt, &optval, sizeof optval);
1427 * Multicast socket options are processed by the in_mcast
1430 case IP_MULTICAST_IF:
1431 case IP_MULTICAST_VIF:
1432 case IP_MULTICAST_TTL:
1433 case IP_MULTICAST_LOOP:
1435 error = inp_getmoptions(inp, sopt);
1438 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1439 case IP_IPSEC_POLICY:
1440 if (IPSEC_ENABLED(ipv4)) {
1441 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1448 error = ENOPROTOOPT;
1457 * Routine called from ip_output() to loop back a copy of an IP multicast
1458 * packet to the input queue of a specified interface. Note that this
1459 * calls the output routine of the loopback "driver", but with an interface
1460 * pointer that might NOT be a loopback interface -- evil, but easier than
1461 * replicating that code here.
1464 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1470 * Make a deep copy of the packet because we're going to
1471 * modify the pack in order to generate checksums.
1473 copym = m_dup(m, M_NOWAIT);
1474 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1475 copym = m_pullup(copym, hlen);
1476 if (copym != NULL) {
1477 /* If needed, compute the checksum and mark it as valid. */
1478 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1479 in_delayed_cksum(copym);
1480 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1481 copym->m_pkthdr.csum_flags |=
1482 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1483 copym->m_pkthdr.csum_data = 0xffff;
1486 * We don't bother to fragment if the IP length is greater
1487 * than the interface's MTU. Can this possibly matter?
1489 ip = mtod(copym, struct ip *);
1491 ip->ip_sum = in_cksum(copym, hlen);
1492 if_simloop(ifp, copym, AF_INET, 0);