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_kdtrace.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/ip.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/in_rss.h>
80 #include <netinet/in_var.h>
81 #include <netinet/ip_var.h>
82 #include <netinet/ip_options.h>
84 #include <netinet/udp.h>
85 #include <netinet/udp_var.h>
88 #include <netinet/sctp.h>
89 #include <netinet/sctp_crc32.h>
92 #include <netipsec/ipsec_support.h>
94 #include <machine/in_cksum.h>
96 #include <security/mac/mac_framework.h>
98 #ifdef MBUF_STRESS_TEST
99 static int mbuf_frag_size = 0;
100 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
101 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
104 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, struct inpcb *inp,
112 struct sockaddr_in *dst, int *fibnum, int *error)
114 struct m_tag *fwd_tag = NULL;
120 ip = mtod(m, struct ip *);
122 /* Run through list of hooks for output packets. */
123 odst.s_addr = ip->ip_dst.s_addr;
124 switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, PFIL_OUT, inp)) {
129 return 1; /* Finished */
134 ip = mtod(m, struct ip *);
136 /* See if destination IP address was changed by packet filter. */
137 if (odst.s_addr != ip->ip_dst.s_addr) {
138 m->m_flags |= M_SKIP_FIREWALL;
139 /* If destination is now ourself drop to ip_input(). */
140 if (in_localip(ip->ip_dst)) {
141 m->m_flags |= M_FASTFWD_OURS;
142 if (m->m_pkthdr.rcvif == NULL)
143 m->m_pkthdr.rcvif = V_loif;
144 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
145 m->m_pkthdr.csum_flags |=
146 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
147 m->m_pkthdr.csum_data = 0xffff;
149 m->m_pkthdr.csum_flags |=
150 CSUM_IP_CHECKED | CSUM_IP_VALID;
152 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
153 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
155 *error = netisr_queue(NETISR_IP, m);
156 return 1; /* Finished */
159 bzero(dst, sizeof(*dst));
160 dst->sin_family = AF_INET;
161 dst->sin_len = sizeof(*dst);
162 dst->sin_addr = ip->ip_dst;
164 return -1; /* Reloop */
166 /* See if fib was changed by packet filter. */
167 if ((*fibnum) != M_GETFIB(m)) {
168 m->m_flags |= M_SKIP_FIREWALL;
169 *fibnum = M_GETFIB(m);
170 return -1; /* Reloop for FIB change */
173 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
174 if (m->m_flags & M_FASTFWD_OURS) {
175 if (m->m_pkthdr.rcvif == NULL)
176 m->m_pkthdr.rcvif = V_loif;
177 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
178 m->m_pkthdr.csum_flags |=
179 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
180 m->m_pkthdr.csum_data = 0xffff;
183 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
184 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
186 m->m_pkthdr.csum_flags |=
187 CSUM_IP_CHECKED | CSUM_IP_VALID;
189 *error = netisr_queue(NETISR_IP, m);
190 return 1; /* Finished */
192 /* Or forward to some other address? */
193 if ((m->m_flags & M_IP_NEXTHOP) &&
194 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
195 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
196 m->m_flags |= M_SKIP_FIREWALL;
197 m->m_flags &= ~M_IP_NEXTHOP;
198 m_tag_delete(m, fwd_tag);
200 return -1; /* Reloop for CHANGE of dst */
207 * IP output. The packet in mbuf chain m contains a skeletal IP
208 * header (with len, off, ttl, proto, tos, src, dst).
209 * The mbuf chain containing the packet will be freed.
210 * The mbuf opt, if present, will not be freed.
211 * If route ro is present and has ro_rt initialized, route lookup would be
212 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
213 * then result of route lookup is stored in ro->ro_rt.
215 * In the IP forwarding case, the packet will arrive with options already
216 * inserted, so must have a NULL opt pointer.
219 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
220 struct ip_moptions *imo, struct inpcb *inp)
222 struct rm_priotracker in_ifa_tracker;
223 struct epoch_tracker et;
225 struct ifnet *ifp = NULL; /* keep compiler happy */
227 int hlen = sizeof (struct ip);
230 struct sockaddr_in *dst;
231 const struct sockaddr_in *gw;
232 struct in_ifaddr *ia;
234 uint16_t ip_len, ip_off;
235 struct route iproute;
236 struct rtentry *rte; /* cache for ro->ro_rt */
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);
254 bzero(ro, sizeof (*ro));
259 m = ip_insertoptions(m, opt, &len);
261 hlen = len; /* ip->ip_hl is updated above */
263 ip = mtod(m, struct ip *);
264 ip_len = ntohs(ip->ip_len);
265 ip_off = ntohs(ip->ip_off);
267 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
268 ip->ip_v = IPVERSION;
269 ip->ip_hl = hlen >> 2;
272 /* Header already set, fetch hlen from there */
273 hlen = ip->ip_hl << 2;
275 if ((flags & IP_FORWARDING) == 0)
276 IPSTAT_INC(ips_localout);
281 * dst can be rewritten but always points to &ro->ro_dst.
282 * gw is readonly but can point either to dst OR rt_gateway,
283 * therefore we need restore gw if we're redoing lookup.
285 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
286 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
289 bzero(dst, sizeof(*dst));
290 dst->sin_family = AF_INET;
291 dst->sin_len = sizeof(*dst);
292 dst->sin_addr = ip->ip_dst;
297 * Validate route against routing table additions;
298 * a better/more specific route might have been added.
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.
311 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
312 rte->rt_ifp == NULL ||
313 !RT_LINK_IS_UP(rte->rt_ifp) ||
314 dst->sin_family != AF_INET ||
315 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
316 RO_INVALIDATE_CACHE(ro);
321 * If routing to interface only, short circuit routing lookup.
322 * The use of an all-ones broadcast address implies this; an
323 * interface is specified by the broadcast address of an interface,
324 * or the destination address of a ptp interface.
326 if (flags & IP_SENDONES) {
327 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
328 M_GETFIB(m)))) == NULL &&
329 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
330 M_GETFIB(m)))) == NULL) {
331 IPSTAT_INC(ips_noroute);
335 ip->ip_dst.s_addr = INADDR_BROADCAST;
336 dst->sin_addr = ip->ip_dst;
340 } else if (flags & IP_ROUTETOIF) {
341 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
342 M_GETFIB(m)))) == NULL &&
343 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
344 M_GETFIB(m)))) == NULL) {
345 IPSTAT_INC(ips_noroute);
351 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
352 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
353 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
354 imo != NULL && imo->imo_multicast_ifp != NULL) {
356 * Bypass the normal routing lookup for multicast
357 * packets if the interface is specified.
359 ifp = imo->imo_multicast_ifp;
360 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
361 isbroadcast = 0; /* fool gcc */
364 * We want to do any cloning requested by the link layer,
365 * as this is probably required in all cases for correct
366 * operation (as it is for ARP).
370 rtalloc_mpath_fib(ro,
371 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
374 in_rtalloc_ign(ro, 0, fibnum);
379 (rte->rt_flags & RTF_UP) == 0 ||
380 rte->rt_ifp == NULL ||
381 !RT_LINK_IS_UP(rte->rt_ifp)) {
382 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
384 * There is no route for this packet, but it is
385 * possible that a matching SPD entry exists.
387 no_route_but_check_spd = 1;
388 mtu = 0; /* Silence GCC warning. */
391 IPSTAT_INC(ips_noroute);
392 error = EHOSTUNREACH;
395 ia = ifatoia(rte->rt_ifa);
397 counter_u64_add(rte->rt_pksent, 1);
398 rt_update_ro_flags(ro);
399 if (rte->rt_flags & RTF_GATEWAY)
400 gw = (struct sockaddr_in *)rte->rt_gateway;
401 if (rte->rt_flags & RTF_HOST)
402 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
403 else if (ifp->if_flags & IFF_BROADCAST)
404 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
410 * Calculate MTU. If we have a route that is up, use that,
411 * otherwise use the interface's MTU.
413 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
417 /* Catch a possible divide by zero later. */
418 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
419 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
421 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
422 m->m_flags |= M_MCAST;
424 * IP destination address is multicast. Make sure "gw"
425 * still points to the address in "ro". (It may have been
426 * changed to point to a gateway address, above.)
430 * See if the caller provided any multicast options
433 ip->ip_ttl = imo->imo_multicast_ttl;
434 if (imo->imo_multicast_vif != -1)
437 ip_mcast_src(imo->imo_multicast_vif) :
440 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
442 * Confirm that the outgoing interface supports multicast.
444 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
445 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
446 IPSTAT_INC(ips_noroute);
452 * If source address not specified yet, use address
453 * of outgoing interface.
455 if (ip->ip_src.s_addr == INADDR_ANY) {
456 /* Interface may have no addresses. */
458 ip->ip_src = IA_SIN(ia)->sin_addr;
461 if ((imo == NULL && in_mcast_loop) ||
462 (imo && imo->imo_multicast_loop)) {
464 * Loop back multicast datagram if not expressly
465 * forbidden to do so, even if we are not a member
466 * of the group; ip_input() will filter it later,
467 * thus deferring a hash lookup and mutex acquisition
468 * at the expense of a cheap copy using m_copym().
470 ip_mloopback(ifp, m, hlen);
473 * If we are acting as a multicast router, perform
474 * multicast forwarding as if the packet had just
475 * arrived on the interface to which we are about
476 * to send. The multicast forwarding function
477 * recursively calls this function, using the
478 * IP_FORWARDING flag to prevent infinite recursion.
480 * Multicasts that are looped back by ip_mloopback(),
481 * above, will be forwarded by the ip_input() routine,
484 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
486 * If rsvp daemon is not running, do not
487 * set ip_moptions. This ensures that the packet
488 * is multicast and not just sent down one link
489 * as prescribed by rsvpd.
494 ip_mforward(ip, ifp, m, imo) != 0) {
502 * Multicasts with a time-to-live of zero may be looped-
503 * back, above, but must not be transmitted on a network.
504 * Also, multicasts addressed to the loopback interface
505 * are not sent -- the above call to ip_mloopback() will
506 * loop back a copy. ip_input() will drop the copy if
507 * this host does not belong to the destination group on
508 * the loopback interface.
510 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
519 * If the source address is not specified yet, use the address
520 * of the outoing interface.
522 if (ip->ip_src.s_addr == INADDR_ANY) {
523 /* Interface may have no addresses. */
525 ip->ip_src = IA_SIN(ia)->sin_addr;
530 * Look for broadcast address and
531 * verify user is allowed to send
535 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
536 error = EADDRNOTAVAIL;
539 if ((flags & IP_ALLOWBROADCAST) == 0) {
543 /* don't allow broadcast messages to be fragmented */
548 m->m_flags |= M_BCAST;
550 m->m_flags &= ~M_BCAST;
554 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
555 if (IPSEC_ENABLED(ipv4)) {
556 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
557 if (error == EINPROGRESS)
563 * Check if there was a route for this packet; return error if not.
565 if (no_route_but_check_spd) {
566 IPSTAT_INC(ips_noroute);
567 error = EHOSTUNREACH;
570 /* Update variables that are affected by ipsec4_output(). */
571 ip = mtod(m, struct ip *);
572 hlen = ip->ip_hl << 2;
575 /* Jump over all PFIL processing if hooks are not active. */
576 if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
577 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
578 case 1: /* Finished */
581 case 0: /* Continue normally */
582 ip = mtod(m, struct ip *);
585 case -1: /* Need to try again */
586 /* Reset everything for a new round */
588 ro->ro_prepend = NULL;
591 ip = mtod(m, struct ip *);
597 /* 127/8 must not appear on wire - RFC1122. */
598 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
599 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
600 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
601 IPSTAT_INC(ips_badaddr);
602 error = EADDRNOTAVAIL;
607 m->m_pkthdr.csum_flags |= CSUM_IP;
608 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
610 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
613 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
614 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
615 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
620 * If small enough for interface, or the interface will take
621 * care of the fragmentation for us, we can just send directly.
624 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
626 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
627 ip->ip_sum = in_cksum(m, hlen);
628 m->m_pkthdr.csum_flags &= ~CSUM_IP;
632 * Record statistics for this interface address.
633 * With CSUM_TSO the byte/packet count will be slightly
634 * incorrect because we count the IP+TCP headers only
635 * once instead of for every generated packet.
637 if (!(flags & IP_FORWARDING) && ia) {
638 if (m->m_pkthdr.csum_flags & CSUM_TSO)
639 counter_u64_add(ia->ia_ifa.ifa_opackets,
640 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
642 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
644 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
646 #ifdef MBUF_STRESS_TEST
647 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
648 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
651 * Reset layer specific mbuf flags
652 * to avoid confusing lower layers.
655 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
658 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
659 in_pcboutput_txrtlmt(inp, ifp, m);
660 /* stamp send tag on mbuf */
661 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
663 m->m_pkthdr.snd_tag = NULL;
666 error = (*ifp->if_output)(ifp, m,
667 (const struct sockaddr *)gw, ro);
669 /* check for route change */
671 in_pcboutput_eagain(inp);
676 /* Balk when DF bit is set or the interface didn't support TSO. */
677 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
679 IPSTAT_INC(ips_cantfrag);
684 * Too large for interface; fragment if possible. If successful,
685 * on return, m will point to a list of packets to be sent.
687 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
694 /* Record statistics for this interface address. */
696 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
697 counter_u64_add(ia->ia_ifa.ifa_obytes,
701 * Reset layer specific mbuf flags
702 * to avoid confusing upper layers.
706 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
707 mtod(m, struct ip *), NULL);
710 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
711 in_pcboutput_txrtlmt(inp, ifp, m);
712 /* stamp send tag on mbuf */
713 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
715 m->m_pkthdr.snd_tag = NULL;
718 error = (*ifp->if_output)(ifp, m,
719 (const struct sockaddr *)gw, ro);
721 /* check for route change */
723 in_pcboutput_eagain(inp);
730 IPSTAT_INC(ips_fragmented);
735 else if (rte == NULL)
737 * If the caller supplied a route but somehow the reference
738 * to it has been released need to prevent the caller
739 * calling RTFREE on it again.
750 * Create a chain of fragments which fit the given mtu. m_frag points to the
751 * mbuf to be fragmented; on return it points to the chain with the fragments.
752 * Return 0 if no error. If error, m_frag may contain a partially built
753 * chain of fragments that should be freed by the caller.
755 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
758 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
759 u_long if_hwassist_flags)
762 int hlen = ip->ip_hl << 2;
763 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
765 struct mbuf *m0 = *m_frag; /* the original packet */
769 uint16_t ip_len, ip_off;
771 ip_len = ntohs(ip->ip_len);
772 ip_off = ntohs(ip->ip_off);
774 if (ip_off & IP_DF) { /* Fragmentation not allowed */
775 IPSTAT_INC(ips_cantfrag);
780 * Must be able to put at least 8 bytes per fragment.
786 * If the interface will not calculate checksums on
787 * fragmented packets, then do it here.
789 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
790 in_delayed_cksum(m0);
791 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
794 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
795 sctp_delayed_cksum(m0, hlen);
796 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
799 if (len > PAGE_SIZE) {
801 * Fragment large datagrams such that each segment
802 * contains a multiple of PAGE_SIZE amount of data,
803 * plus headers. This enables a receiver to perform
804 * page-flipping zero-copy optimizations.
806 * XXX When does this help given that sender and receiver
807 * could have different page sizes, and also mtu could
808 * be less than the receiver's page size ?
812 off = MIN(mtu, m0->m_pkthdr.len);
815 * firstlen (off - hlen) must be aligned on an
819 goto smart_frag_failure;
820 off = ((off - hlen) & ~7) + hlen;
821 newlen = (~PAGE_MASK) & mtu;
822 if ((newlen + sizeof (struct ip)) > mtu) {
823 /* we failed, go back the default */
834 firstlen = off - hlen;
835 mnext = &m0->m_nextpkt; /* pointer to next packet */
838 * Loop through length of segment after first fragment,
839 * make new header and copy data of each part and link onto chain.
840 * Here, m0 is the original packet, m is the fragment being created.
841 * The fragments are linked off the m_nextpkt of the original
842 * packet, which after processing serves as the first fragment.
844 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
845 struct ip *mhip; /* ip header on the fragment */
847 int mhlen = sizeof (struct ip);
849 m = m_gethdr(M_NOWAIT, MT_DATA);
852 IPSTAT_INC(ips_odropped);
856 * Make sure the complete packet header gets copied
857 * from the originating mbuf to the newly created
858 * mbuf. This also ensures that existing firewall
859 * classification(s), VLAN tags and so on get copied
860 * to the resulting fragmented packet(s):
862 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
865 IPSTAT_INC(ips_odropped);
869 * In the first mbuf, leave room for the link header, then
870 * copy the original IP header including options. The payload
871 * goes into an additional mbuf chain returned by m_copym().
873 m->m_data += max_linkhdr;
874 mhip = mtod(m, struct ip *);
876 if (hlen > sizeof (struct ip)) {
877 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
878 mhip->ip_v = IPVERSION;
879 mhip->ip_hl = mhlen >> 2;
882 /* XXX do we need to add ip_off below ? */
883 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
884 if (off + len >= ip_len)
887 mhip->ip_off |= IP_MF;
888 mhip->ip_len = htons((u_short)(len + mhlen));
889 m->m_next = m_copym(m0, off, len, M_NOWAIT);
890 if (m->m_next == NULL) { /* copy failed */
892 error = ENOBUFS; /* ??? */
893 IPSTAT_INC(ips_odropped);
896 m->m_pkthdr.len = mhlen + len;
898 mac_netinet_fragment(m0, m);
900 mhip->ip_off = htons(mhip->ip_off);
902 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
903 mhip->ip_sum = in_cksum(m, mhlen);
904 m->m_pkthdr.csum_flags &= ~CSUM_IP;
907 mnext = &m->m_nextpkt;
909 IPSTAT_ADD(ips_ofragments, nfrags);
912 * Update first fragment by trimming what's been copied out
913 * and updating header.
915 m_adj(m0, hlen + firstlen - ip_len);
916 m0->m_pkthdr.len = hlen + firstlen;
917 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
918 ip->ip_off = htons(ip_off | IP_MF);
920 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
921 ip->ip_sum = in_cksum(m0, hlen);
922 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
931 in_delayed_cksum(struct mbuf *m)
935 uint16_t cklen, csum, offset;
937 ip = mtod(m, struct ip *);
938 offset = ip->ip_hl << 2 ;
940 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
941 /* if udp header is not in the first mbuf copy udplen */
942 if (offset + sizeof(struct udphdr) > m->m_len) {
943 m_copydata(m, offset + offsetof(struct udphdr,
944 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
945 cklen = ntohs(cklen);
947 uh = (struct udphdr *)mtodo(m, offset);
948 cklen = ntohs(uh->uh_ulen);
950 csum = in_cksum_skip(m, cklen + offset, offset);
954 cklen = ntohs(ip->ip_len);
955 csum = in_cksum_skip(m, cklen, offset);
957 offset += m->m_pkthdr.csum_data; /* checksum offset */
959 if (offset + sizeof(csum) > m->m_len)
960 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
962 *(u_short *)mtodo(m, offset) = csum;
966 * IP socket option processing.
969 ip_ctloutput(struct socket *so, struct sockopt *sopt)
971 struct inpcb *inp = sotoinpcb(so);
979 if (sopt->sopt_level != IPPROTO_IP) {
982 if (sopt->sopt_level == SOL_SOCKET &&
983 sopt->sopt_dir == SOPT_SET) {
984 switch (sopt->sopt_name) {
987 if ((so->so_options & SO_REUSEADDR) != 0)
988 inp->inp_flags2 |= INP_REUSEADDR;
990 inp->inp_flags2 &= ~INP_REUSEADDR;
996 if ((so->so_options & SO_REUSEPORT) != 0)
997 inp->inp_flags2 |= INP_REUSEPORT;
999 inp->inp_flags2 &= ~INP_REUSEPORT;
1003 case SO_REUSEPORT_LB:
1005 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1006 inp->inp_flags2 |= INP_REUSEPORT_LB;
1008 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1014 inp->inp_inc.inc_fibnum = so->so_fibnum;
1018 case SO_MAX_PACING_RATE:
1021 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1035 switch (sopt->sopt_dir) {
1037 switch (sopt->sopt_name) {
1044 if (sopt->sopt_valsize > MLEN) {
1048 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1053 m->m_len = sopt->sopt_valsize;
1054 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1061 error = ip_pcbopts(inp, sopt->sopt_name, m);
1067 if (sopt->sopt_td != NULL) {
1068 error = priv_check(sopt->sopt_td,
1069 PRIV_NETINET_BINDANY);
1076 case IP_RSS_LISTEN_BUCKET:
1082 case IP_RECVRETOPTS:
1083 case IP_ORIGDSTADDR:
1084 case IP_RECVDSTADDR:
1092 case IP_RECVRSSBUCKETID:
1094 error = sooptcopyin(sopt, &optval, sizeof optval,
1099 switch (sopt->sopt_name) {
1101 inp->inp_ip_tos = optval;
1105 inp->inp_ip_ttl = optval;
1109 if (optval >= 0 && optval <= MAXTTL)
1110 inp->inp_ip_minttl = optval;
1115 #define OPTSET(bit) do { \
1118 inp->inp_flags |= bit; \
1120 inp->inp_flags &= ~bit; \
1124 #define OPTSET2(bit, val) do { \
1127 inp->inp_flags2 |= bit; \
1129 inp->inp_flags2 &= ~bit; \
1134 OPTSET(INP_RECVOPTS);
1137 case IP_RECVRETOPTS:
1138 OPTSET(INP_RECVRETOPTS);
1141 case IP_RECVDSTADDR:
1142 OPTSET(INP_RECVDSTADDR);
1145 case IP_ORIGDSTADDR:
1146 OPTSET2(INP_ORIGDSTADDR, optval);
1150 OPTSET(INP_RECVTTL);
1158 OPTSET(INP_ONESBCAST);
1161 OPTSET(INP_DONTFRAG);
1164 OPTSET(INP_BINDANY);
1167 OPTSET(INP_RECVTOS);
1170 OPTSET2(INP_BINDMULTI, optval);
1173 OPTSET2(INP_RECVFLOWID, optval);
1176 case IP_RSS_LISTEN_BUCKET:
1177 if ((optval >= 0) &&
1178 (optval < rss_getnumbuckets())) {
1179 inp->inp_rss_listen_bucket = optval;
1180 OPTSET2(INP_RSS_BUCKET_SET, 1);
1185 case IP_RECVRSSBUCKETID:
1186 OPTSET2(INP_RECVRSSBUCKETID, optval);
1195 * Multicast socket options are processed by the in_mcast
1198 case IP_MULTICAST_IF:
1199 case IP_MULTICAST_VIF:
1200 case IP_MULTICAST_TTL:
1201 case IP_MULTICAST_LOOP:
1202 case IP_ADD_MEMBERSHIP:
1203 case IP_DROP_MEMBERSHIP:
1204 case IP_ADD_SOURCE_MEMBERSHIP:
1205 case IP_DROP_SOURCE_MEMBERSHIP:
1206 case IP_BLOCK_SOURCE:
1207 case IP_UNBLOCK_SOURCE:
1209 case MCAST_JOIN_GROUP:
1210 case MCAST_LEAVE_GROUP:
1211 case MCAST_JOIN_SOURCE_GROUP:
1212 case MCAST_LEAVE_SOURCE_GROUP:
1213 case MCAST_BLOCK_SOURCE:
1214 case MCAST_UNBLOCK_SOURCE:
1215 error = inp_setmoptions(inp, sopt);
1219 error = sooptcopyin(sopt, &optval, sizeof optval,
1226 case IP_PORTRANGE_DEFAULT:
1227 inp->inp_flags &= ~(INP_LOWPORT);
1228 inp->inp_flags &= ~(INP_HIGHPORT);
1231 case IP_PORTRANGE_HIGH:
1232 inp->inp_flags &= ~(INP_LOWPORT);
1233 inp->inp_flags |= INP_HIGHPORT;
1236 case IP_PORTRANGE_LOW:
1237 inp->inp_flags &= ~(INP_HIGHPORT);
1238 inp->inp_flags |= INP_LOWPORT;
1248 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1249 case IP_IPSEC_POLICY:
1250 if (IPSEC_ENABLED(ipv4)) {
1251 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1258 error = ENOPROTOOPT;
1264 switch (sopt->sopt_name) {
1268 if (inp->inp_options) {
1269 struct mbuf *options;
1271 options = m_copym(inp->inp_options, 0,
1272 M_COPYALL, M_NOWAIT);
1274 if (options != NULL) {
1275 error = sooptcopyout(sopt,
1276 mtod(options, char *),
1283 sopt->sopt_valsize = 0;
1291 case IP_RECVRETOPTS:
1292 case IP_ORIGDSTADDR:
1293 case IP_RECVDSTADDR:
1306 case IP_RSSBUCKETID:
1307 case IP_RECVRSSBUCKETID:
1309 switch (sopt->sopt_name) {
1312 optval = inp->inp_ip_tos;
1316 optval = inp->inp_ip_ttl;
1320 optval = inp->inp_ip_minttl;
1323 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1324 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1327 optval = OPTBIT(INP_RECVOPTS);
1330 case IP_RECVRETOPTS:
1331 optval = OPTBIT(INP_RECVRETOPTS);
1334 case IP_RECVDSTADDR:
1335 optval = OPTBIT(INP_RECVDSTADDR);
1338 case IP_ORIGDSTADDR:
1339 optval = OPTBIT2(INP_ORIGDSTADDR);
1343 optval = OPTBIT(INP_RECVTTL);
1347 optval = OPTBIT(INP_RECVIF);
1351 if (inp->inp_flags & INP_HIGHPORT)
1352 optval = IP_PORTRANGE_HIGH;
1353 else if (inp->inp_flags & INP_LOWPORT)
1354 optval = IP_PORTRANGE_LOW;
1360 optval = OPTBIT(INP_ONESBCAST);
1363 optval = OPTBIT(INP_DONTFRAG);
1366 optval = OPTBIT(INP_BINDANY);
1369 optval = OPTBIT(INP_RECVTOS);
1372 optval = inp->inp_flowid;
1375 optval = inp->inp_flowtype;
1378 optval = OPTBIT2(INP_RECVFLOWID);
1381 case IP_RSSBUCKETID:
1382 retval = rss_hash2bucket(inp->inp_flowid,
1386 optval = rss_bucket;
1390 case IP_RECVRSSBUCKETID:
1391 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1395 optval = OPTBIT2(INP_BINDMULTI);
1398 error = sooptcopyout(sopt, &optval, sizeof optval);
1402 * Multicast socket options are processed by the in_mcast
1405 case IP_MULTICAST_IF:
1406 case IP_MULTICAST_VIF:
1407 case IP_MULTICAST_TTL:
1408 case IP_MULTICAST_LOOP:
1410 error = inp_getmoptions(inp, sopt);
1413 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1414 case IP_IPSEC_POLICY:
1415 if (IPSEC_ENABLED(ipv4)) {
1416 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1423 error = ENOPROTOOPT;
1432 * Routine called from ip_output() to loop back a copy of an IP multicast
1433 * packet to the input queue of a specified interface. Note that this
1434 * calls the output routine of the loopback "driver", but with an interface
1435 * pointer that might NOT be a loopback interface -- evil, but easier than
1436 * replicating that code here.
1439 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1445 * Make a deep copy of the packet because we're going to
1446 * modify the pack in order to generate checksums.
1448 copym = m_dup(m, M_NOWAIT);
1449 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1450 copym = m_pullup(copym, hlen);
1451 if (copym != NULL) {
1452 /* If needed, compute the checksum and mark it as valid. */
1453 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1454 in_delayed_cksum(copym);
1455 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1456 copym->m_pkthdr.csum_flags |=
1457 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1458 copym->m_pkthdr.csum_data = 0xffff;
1461 * We don't bother to fragment if the IP length is greater
1462 * than the interface's MTU. Can this possibly matter?
1464 ip = mtod(copym, struct ip *);
1466 ip->ip_sum = in_cksum(copym, hlen);
1467 if_simloop(ifp, copym, AF_INET, 0);