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/sctp.h>
85 #include <netinet/sctp_crc32.h>
88 #include <netipsec/ipsec_support.h>
90 #include <machine/in_cksum.h>
92 #include <security/mac/mac_framework.h>
94 #ifdef MBUF_STRESS_TEST
95 static int mbuf_frag_size = 0;
96 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
97 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
100 static void ip_mloopback(struct ifnet *, const struct mbuf *, int);
103 extern int in_mcast_loop;
104 extern struct protosw inetsw[];
107 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
108 struct sockaddr_in *dst, int *fibnum, int *error)
110 struct m_tag *fwd_tag = NULL;
116 ip = mtod(m, struct ip *);
118 /* Run through list of hooks for output packets. */
119 odst.s_addr = ip->ip_dst.s_addr;
120 *error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, 0, inp);
122 if ((*error) != 0 || m == NULL)
123 return 1; /* Finished */
125 ip = mtod(m, struct ip *);
127 /* See if destination IP address was changed by packet filter. */
128 if (odst.s_addr != ip->ip_dst.s_addr) {
129 m->m_flags |= M_SKIP_FIREWALL;
130 /* If destination is now ourself drop to ip_input(). */
131 if (in_localip(ip->ip_dst)) {
132 m->m_flags |= M_FASTFWD_OURS;
133 if (m->m_pkthdr.rcvif == NULL)
134 m->m_pkthdr.rcvif = V_loif;
135 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
136 m->m_pkthdr.csum_flags |=
137 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
138 m->m_pkthdr.csum_data = 0xffff;
140 m->m_pkthdr.csum_flags |=
141 CSUM_IP_CHECKED | CSUM_IP_VALID;
143 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
144 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
146 *error = netisr_queue(NETISR_IP, m);
147 return 1; /* Finished */
150 bzero(dst, sizeof(*dst));
151 dst->sin_family = AF_INET;
152 dst->sin_len = sizeof(*dst);
153 dst->sin_addr = ip->ip_dst;
155 return -1; /* Reloop */
157 /* See if fib was changed by packet filter. */
158 if ((*fibnum) != M_GETFIB(m)) {
159 m->m_flags |= M_SKIP_FIREWALL;
160 *fibnum = M_GETFIB(m);
161 return -1; /* Reloop for FIB change */
164 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
165 if (m->m_flags & M_FASTFWD_OURS) {
166 if (m->m_pkthdr.rcvif == NULL)
167 m->m_pkthdr.rcvif = V_loif;
168 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
169 m->m_pkthdr.csum_flags |=
170 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
171 m->m_pkthdr.csum_data = 0xffff;
174 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
175 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
177 m->m_pkthdr.csum_flags |=
178 CSUM_IP_CHECKED | CSUM_IP_VALID;
180 *error = netisr_queue(NETISR_IP, m);
181 return 1; /* Finished */
183 /* Or forward to some other address? */
184 if ((m->m_flags & M_IP_NEXTHOP) &&
185 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
186 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
187 m->m_flags |= M_SKIP_FIREWALL;
188 m->m_flags &= ~M_IP_NEXTHOP;
189 m_tag_delete(m, fwd_tag);
191 return -1; /* Reloop for CHANGE of dst */
198 * IP output. The packet in mbuf chain m contains a skeletal IP
199 * header (with len, off, ttl, proto, tos, src, dst).
200 * The mbuf chain containing the packet will be freed.
201 * The mbuf opt, if present, will not be freed.
202 * If route ro is present and has ro_rt initialized, route lookup would be
203 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
204 * then result of route lookup is stored in ro->ro_rt.
206 * In the IP forwarding case, the packet will arrive with options already
207 * inserted, so must have a NULL opt pointer.
210 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
211 struct ip_moptions *imo, struct inpcb *inp)
213 struct rm_priotracker in_ifa_tracker;
215 struct ifnet *ifp = NULL; /* keep compiler happy */
217 int hlen = sizeof (struct ip);
220 struct sockaddr_in *dst;
221 const struct sockaddr_in *gw;
222 struct in_ifaddr *ia;
224 uint16_t ip_len, ip_off;
225 struct route iproute;
226 struct rtentry *rte; /* cache for ro->ro_rt */
228 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
229 int no_route_but_check_spd = 0;
234 INP_LOCK_ASSERT(inp);
235 M_SETFIB(m, inp->inp_inc.inc_fibnum);
236 if ((flags & IP_NODEFAULTFLOWID) == 0) {
237 m->m_pkthdr.flowid = inp->inp_flowid;
238 M_HASHTYPE_SET(m, inp->inp_flowtype);
244 bzero(ro, sizeof (*ro));
249 m = ip_insertoptions(m, opt, &len);
251 hlen = len; /* ip->ip_hl is updated above */
253 ip = mtod(m, struct ip *);
254 ip_len = ntohs(ip->ip_len);
255 ip_off = ntohs(ip->ip_off);
257 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
258 ip->ip_v = IPVERSION;
259 ip->ip_hl = hlen >> 2;
261 IPSTAT_INC(ips_localout);
263 /* Header already set, fetch hlen from there */
264 hlen = ip->ip_hl << 2;
270 * dst can be rewritten but always points to &ro->ro_dst.
271 * gw is readonly but can point either to dst OR rt_gateway,
272 * therefore we need restore gw if we're redoing lookup.
274 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
275 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
278 bzero(dst, sizeof(*dst));
279 dst->sin_family = AF_INET;
280 dst->sin_len = sizeof(*dst);
281 dst->sin_addr = ip->ip_dst;
286 * Validate route against routing table additions;
287 * a better/more specific route might have been added.
290 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
292 * If there is a cached route,
293 * check that it is to the same destination
294 * and is still up. If not, free it and try again.
295 * The address family should also be checked in case of sharing the
297 * Also check whether routing cache needs invalidation.
300 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
301 rte->rt_ifp == NULL ||
302 !RT_LINK_IS_UP(rte->rt_ifp) ||
303 dst->sin_family != AF_INET ||
304 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
305 RO_INVALIDATE_CACHE(ro);
310 * If routing to interface only, short circuit routing lookup.
311 * The use of an all-ones broadcast address implies this; an
312 * interface is specified by the broadcast address of an interface,
313 * or the destination address of a ptp interface.
315 if (flags & IP_SENDONES) {
316 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
317 M_GETFIB(m)))) == NULL &&
318 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
319 M_GETFIB(m)))) == NULL) {
320 IPSTAT_INC(ips_noroute);
324 ip->ip_dst.s_addr = INADDR_BROADCAST;
325 dst->sin_addr = ip->ip_dst;
329 } else if (flags & IP_ROUTETOIF) {
330 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
331 M_GETFIB(m)))) == NULL &&
332 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
333 M_GETFIB(m)))) == NULL) {
334 IPSTAT_INC(ips_noroute);
340 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
341 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
342 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
343 imo != NULL && imo->imo_multicast_ifp != NULL) {
345 * Bypass the normal routing lookup for multicast
346 * packets if the interface is specified.
348 ifp = imo->imo_multicast_ifp;
349 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
350 isbroadcast = 0; /* fool gcc */
353 * We want to do any cloning requested by the link layer,
354 * as this is probably required in all cases for correct
355 * operation (as it is for ARP).
359 rtalloc_mpath_fib(ro,
360 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
363 in_rtalloc_ign(ro, 0, fibnum);
368 (rte->rt_flags & RTF_UP) == 0 ||
369 rte->rt_ifp == NULL ||
370 !RT_LINK_IS_UP(rte->rt_ifp)) {
371 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
373 * There is no route for this packet, but it is
374 * possible that a matching SPD entry exists.
376 no_route_but_check_spd = 1;
377 mtu = 0; /* Silence GCC warning. */
380 IPSTAT_INC(ips_noroute);
381 error = EHOSTUNREACH;
384 ia = ifatoia(rte->rt_ifa);
386 counter_u64_add(rte->rt_pksent, 1);
387 rt_update_ro_flags(ro);
388 if (rte->rt_flags & RTF_GATEWAY)
389 gw = (struct sockaddr_in *)rte->rt_gateway;
390 if (rte->rt_flags & RTF_HOST)
391 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
392 else if (ifp->if_flags & IFF_BROADCAST)
393 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
399 * Calculate MTU. If we have a route that is up, use that,
400 * otherwise use the interface's MTU.
402 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
406 /* Catch a possible divide by zero later. */
407 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
408 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
410 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
411 m->m_flags |= M_MCAST;
413 * IP destination address is multicast. Make sure "gw"
414 * still points to the address in "ro". (It may have been
415 * changed to point to a gateway address, above.)
419 * See if the caller provided any multicast options
422 ip->ip_ttl = imo->imo_multicast_ttl;
423 if (imo->imo_multicast_vif != -1)
426 ip_mcast_src(imo->imo_multicast_vif) :
429 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
431 * Confirm that the outgoing interface supports multicast.
433 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
434 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
435 IPSTAT_INC(ips_noroute);
441 * If source address not specified yet, use address
442 * of outgoing interface.
444 if (ip->ip_src.s_addr == INADDR_ANY) {
445 /* Interface may have no addresses. */
447 ip->ip_src = IA_SIN(ia)->sin_addr;
450 if ((imo == NULL && in_mcast_loop) ||
451 (imo && imo->imo_multicast_loop)) {
453 * Loop back multicast datagram if not expressly
454 * forbidden to do so, even if we are not a member
455 * of the group; ip_input() will filter it later,
456 * thus deferring a hash lookup and mutex acquisition
457 * at the expense of a cheap copy using m_copym().
459 ip_mloopback(ifp, m, hlen);
462 * If we are acting as a multicast router, perform
463 * multicast forwarding as if the packet had just
464 * arrived on the interface to which we are about
465 * to send. The multicast forwarding function
466 * recursively calls this function, using the
467 * IP_FORWARDING flag to prevent infinite recursion.
469 * Multicasts that are looped back by ip_mloopback(),
470 * above, will be forwarded by the ip_input() routine,
473 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
475 * If rsvp daemon is not running, do not
476 * set ip_moptions. This ensures that the packet
477 * is multicast and not just sent down one link
478 * as prescribed by rsvpd.
483 ip_mforward(ip, ifp, m, imo) != 0) {
491 * Multicasts with a time-to-live of zero may be looped-
492 * back, above, but must not be transmitted on a network.
493 * Also, multicasts addressed to the loopback interface
494 * are not sent -- the above call to ip_mloopback() will
495 * loop back a copy. ip_input() will drop the copy if
496 * this host does not belong to the destination group on
497 * the loopback interface.
499 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
508 * If the source address is not specified yet, use the address
509 * of the outoing interface.
511 if (ip->ip_src.s_addr == INADDR_ANY) {
512 /* Interface may have no addresses. */
514 ip->ip_src = IA_SIN(ia)->sin_addr;
519 * Look for broadcast address and
520 * verify user is allowed to send
524 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
525 error = EADDRNOTAVAIL;
528 if ((flags & IP_ALLOWBROADCAST) == 0) {
532 /* don't allow broadcast messages to be fragmented */
537 m->m_flags |= M_BCAST;
539 m->m_flags &= ~M_BCAST;
543 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
544 if (IPSEC_ENABLED(ipv4)) {
545 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
546 if (error == EINPROGRESS)
552 * Check if there was a route for this packet; return error if not.
554 if (no_route_but_check_spd) {
555 IPSTAT_INC(ips_noroute);
556 error = EHOSTUNREACH;
559 /* Update variables that are affected by ipsec4_output(). */
560 ip = mtod(m, struct ip *);
561 hlen = ip->ip_hl << 2;
564 /* Jump over all PFIL processing if hooks are not active. */
565 if (PFIL_HOOKED(&V_inet_pfil_hook)) {
566 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
567 case 1: /* Finished */
570 case 0: /* Continue normally */
571 ip = mtod(m, struct ip *);
574 case -1: /* Need to try again */
575 /* Reset everything for a new round */
577 ro->ro_prepend = NULL;
580 ip = mtod(m, struct ip *);
586 /* 127/8 must not appear on wire - RFC1122. */
587 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
588 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
589 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
590 IPSTAT_INC(ips_badaddr);
591 error = EADDRNOTAVAIL;
596 m->m_pkthdr.csum_flags |= CSUM_IP;
597 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
599 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
602 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
603 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
604 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
609 * If small enough for interface, or the interface will take
610 * care of the fragmentation for us, we can just send directly.
613 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
615 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
616 ip->ip_sum = in_cksum(m, hlen);
617 m->m_pkthdr.csum_flags &= ~CSUM_IP;
621 * Record statistics for this interface address.
622 * With CSUM_TSO the byte/packet count will be slightly
623 * incorrect because we count the IP+TCP headers only
624 * once instead of for every generated packet.
626 if (!(flags & IP_FORWARDING) && ia) {
627 if (m->m_pkthdr.csum_flags & CSUM_TSO)
628 counter_u64_add(ia->ia_ifa.ifa_opackets,
629 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
631 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
633 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
635 #ifdef MBUF_STRESS_TEST
636 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
637 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
640 * Reset layer specific mbuf flags
641 * to avoid confusing lower layers.
644 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
647 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
648 in_pcboutput_txrtlmt(inp, ifp, m);
649 /* stamp send tag on mbuf */
650 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
652 m->m_pkthdr.snd_tag = NULL;
655 error = (*ifp->if_output)(ifp, m,
656 (const struct sockaddr *)gw, ro);
658 /* check for route change */
660 in_pcboutput_eagain(inp);
665 /* Balk when DF bit is set or the interface didn't support TSO. */
666 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
668 IPSTAT_INC(ips_cantfrag);
673 * Too large for interface; fragment if possible. If successful,
674 * on return, m will point to a list of packets to be sent.
676 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
683 /* Record statistics for this interface address. */
685 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
686 counter_u64_add(ia->ia_ifa.ifa_obytes,
690 * Reset layer specific mbuf flags
691 * to avoid confusing upper layers.
695 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
696 mtod(m, struct ip *), NULL);
699 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
700 in_pcboutput_txrtlmt(inp, ifp, m);
701 /* stamp send tag on mbuf */
702 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
704 m->m_pkthdr.snd_tag = NULL;
707 error = (*ifp->if_output)(ifp, m,
708 (const struct sockaddr *)gw, ro);
710 /* check for route change */
712 in_pcboutput_eagain(inp);
719 IPSTAT_INC(ips_fragmented);
724 else if (rte == NULL)
726 * If the caller supplied a route but somehow the reference
727 * to it has been released need to prevent the caller
728 * calling RTFREE on it again.
739 * Create a chain of fragments which fit the given mtu. m_frag points to the
740 * mbuf to be fragmented; on return it points to the chain with the fragments.
741 * Return 0 if no error. If error, m_frag may contain a partially built
742 * chain of fragments that should be freed by the caller.
744 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
747 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
748 u_long if_hwassist_flags)
751 int hlen = ip->ip_hl << 2;
752 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
754 struct mbuf *m0 = *m_frag; /* the original packet */
758 uint16_t ip_len, ip_off;
760 ip_len = ntohs(ip->ip_len);
761 ip_off = ntohs(ip->ip_off);
763 if (ip_off & IP_DF) { /* Fragmentation not allowed */
764 IPSTAT_INC(ips_cantfrag);
769 * Must be able to put at least 8 bytes per fragment.
775 * If the interface will not calculate checksums on
776 * fragmented packets, then do it here.
778 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
779 in_delayed_cksum(m0);
780 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
783 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
784 sctp_delayed_cksum(m0, hlen);
785 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
788 if (len > PAGE_SIZE) {
790 * Fragment large datagrams such that each segment
791 * contains a multiple of PAGE_SIZE amount of data,
792 * plus headers. This enables a receiver to perform
793 * page-flipping zero-copy optimizations.
795 * XXX When does this help given that sender and receiver
796 * could have different page sizes, and also mtu could
797 * be less than the receiver's page size ?
801 off = MIN(mtu, m0->m_pkthdr.len);
804 * firstlen (off - hlen) must be aligned on an
808 goto smart_frag_failure;
809 off = ((off - hlen) & ~7) + hlen;
810 newlen = (~PAGE_MASK) & mtu;
811 if ((newlen + sizeof (struct ip)) > mtu) {
812 /* we failed, go back the default */
823 firstlen = off - hlen;
824 mnext = &m0->m_nextpkt; /* pointer to next packet */
827 * Loop through length of segment after first fragment,
828 * make new header and copy data of each part and link onto chain.
829 * Here, m0 is the original packet, m is the fragment being created.
830 * The fragments are linked off the m_nextpkt of the original
831 * packet, which after processing serves as the first fragment.
833 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
834 struct ip *mhip; /* ip header on the fragment */
836 int mhlen = sizeof (struct ip);
838 m = m_gethdr(M_NOWAIT, MT_DATA);
841 IPSTAT_INC(ips_odropped);
845 * Make sure the complete packet header gets copied
846 * from the originating mbuf to the newly created
847 * mbuf. This also ensures that existing firewall
848 * classification(s), VLAN tags and so on get copied
849 * to the resulting fragmented packet(s):
851 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
854 IPSTAT_INC(ips_odropped);
858 * In the first mbuf, leave room for the link header, then
859 * copy the original IP header including options. The payload
860 * goes into an additional mbuf chain returned by m_copym().
862 m->m_data += max_linkhdr;
863 mhip = mtod(m, struct ip *);
865 if (hlen > sizeof (struct ip)) {
866 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
867 mhip->ip_v = IPVERSION;
868 mhip->ip_hl = mhlen >> 2;
871 /* XXX do we need to add ip_off below ? */
872 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
873 if (off + len >= ip_len)
876 mhip->ip_off |= IP_MF;
877 mhip->ip_len = htons((u_short)(len + mhlen));
878 m->m_next = m_copym(m0, off, len, M_NOWAIT);
879 if (m->m_next == NULL) { /* copy failed */
881 error = ENOBUFS; /* ??? */
882 IPSTAT_INC(ips_odropped);
885 m->m_pkthdr.len = mhlen + len;
887 mac_netinet_fragment(m0, m);
889 mhip->ip_off = htons(mhip->ip_off);
891 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
892 mhip->ip_sum = in_cksum(m, mhlen);
893 m->m_pkthdr.csum_flags &= ~CSUM_IP;
896 mnext = &m->m_nextpkt;
898 IPSTAT_ADD(ips_ofragments, nfrags);
901 * Update first fragment by trimming what's been copied out
902 * and updating header.
904 m_adj(m0, hlen + firstlen - ip_len);
905 m0->m_pkthdr.len = hlen + firstlen;
906 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
907 ip->ip_off = htons(ip_off | IP_MF);
909 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
910 ip->ip_sum = in_cksum(m0, hlen);
911 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
920 in_delayed_cksum(struct mbuf *m)
923 uint16_t csum, offset, ip_len;
925 ip = mtod(m, struct ip *);
926 offset = ip->ip_hl << 2 ;
927 ip_len = ntohs(ip->ip_len);
928 csum = in_cksum_skip(m, ip_len, offset);
929 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
931 offset += m->m_pkthdr.csum_data; /* checksum offset */
933 /* find the mbuf in the chain where the checksum starts*/
934 while ((m != NULL) && (offset >= m->m_len)) {
938 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
939 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
940 *(u_short *)(m->m_data + offset) = csum;
944 * IP socket option processing.
947 ip_ctloutput(struct socket *so, struct sockopt *sopt)
949 struct inpcb *inp = sotoinpcb(so);
957 if (sopt->sopt_level != IPPROTO_IP) {
960 if (sopt->sopt_level == SOL_SOCKET &&
961 sopt->sopt_dir == SOPT_SET) {
962 switch (sopt->sopt_name) {
965 if ((so->so_options & SO_REUSEADDR) != 0)
966 inp->inp_flags2 |= INP_REUSEADDR;
968 inp->inp_flags2 &= ~INP_REUSEADDR;
974 if ((so->so_options & SO_REUSEPORT) != 0)
975 inp->inp_flags2 |= INP_REUSEPORT;
977 inp->inp_flags2 &= ~INP_REUSEPORT;
983 inp->inp_inc.inc_fibnum = so->so_fibnum;
987 case SO_MAX_PACING_RATE:
990 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1004 switch (sopt->sopt_dir) {
1006 switch (sopt->sopt_name) {
1013 if (sopt->sopt_valsize > MLEN) {
1017 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1022 m->m_len = sopt->sopt_valsize;
1023 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1030 error = ip_pcbopts(inp, sopt->sopt_name, m);
1036 if (sopt->sopt_td != NULL) {
1037 error = priv_check(sopt->sopt_td,
1038 PRIV_NETINET_BINDANY);
1045 case IP_RSS_LISTEN_BUCKET:
1051 case IP_RECVRETOPTS:
1052 case IP_ORIGDSTADDR:
1053 case IP_RECVDSTADDR:
1061 case IP_RECVRSSBUCKETID:
1063 error = sooptcopyin(sopt, &optval, sizeof optval,
1068 switch (sopt->sopt_name) {
1070 inp->inp_ip_tos = optval;
1074 inp->inp_ip_ttl = optval;
1078 if (optval >= 0 && optval <= MAXTTL)
1079 inp->inp_ip_minttl = optval;
1084 #define OPTSET(bit) do { \
1087 inp->inp_flags |= bit; \
1089 inp->inp_flags &= ~bit; \
1093 #define OPTSET2(bit, val) do { \
1096 inp->inp_flags2 |= bit; \
1098 inp->inp_flags2 &= ~bit; \
1103 OPTSET(INP_RECVOPTS);
1106 case IP_RECVRETOPTS:
1107 OPTSET(INP_RECVRETOPTS);
1110 case IP_RECVDSTADDR:
1111 OPTSET(INP_RECVDSTADDR);
1114 case IP_ORIGDSTADDR:
1115 OPTSET2(INP_ORIGDSTADDR, optval);
1119 OPTSET(INP_RECVTTL);
1127 OPTSET(INP_ONESBCAST);
1130 OPTSET(INP_DONTFRAG);
1133 OPTSET(INP_BINDANY);
1136 OPTSET(INP_RECVTOS);
1139 OPTSET2(INP_BINDMULTI, optval);
1142 OPTSET2(INP_RECVFLOWID, optval);
1145 case IP_RSS_LISTEN_BUCKET:
1146 if ((optval >= 0) &&
1147 (optval < rss_getnumbuckets())) {
1148 inp->inp_rss_listen_bucket = optval;
1149 OPTSET2(INP_RSS_BUCKET_SET, 1);
1154 case IP_RECVRSSBUCKETID:
1155 OPTSET2(INP_RECVRSSBUCKETID, optval);
1164 * Multicast socket options are processed by the in_mcast
1167 case IP_MULTICAST_IF:
1168 case IP_MULTICAST_VIF:
1169 case IP_MULTICAST_TTL:
1170 case IP_MULTICAST_LOOP:
1171 case IP_ADD_MEMBERSHIP:
1172 case IP_DROP_MEMBERSHIP:
1173 case IP_ADD_SOURCE_MEMBERSHIP:
1174 case IP_DROP_SOURCE_MEMBERSHIP:
1175 case IP_BLOCK_SOURCE:
1176 case IP_UNBLOCK_SOURCE:
1178 case MCAST_JOIN_GROUP:
1179 case MCAST_LEAVE_GROUP:
1180 case MCAST_JOIN_SOURCE_GROUP:
1181 case MCAST_LEAVE_SOURCE_GROUP:
1182 case MCAST_BLOCK_SOURCE:
1183 case MCAST_UNBLOCK_SOURCE:
1184 error = inp_setmoptions(inp, sopt);
1188 error = sooptcopyin(sopt, &optval, sizeof optval,
1195 case IP_PORTRANGE_DEFAULT:
1196 inp->inp_flags &= ~(INP_LOWPORT);
1197 inp->inp_flags &= ~(INP_HIGHPORT);
1200 case IP_PORTRANGE_HIGH:
1201 inp->inp_flags &= ~(INP_LOWPORT);
1202 inp->inp_flags |= INP_HIGHPORT;
1205 case IP_PORTRANGE_LOW:
1206 inp->inp_flags &= ~(INP_HIGHPORT);
1207 inp->inp_flags |= INP_LOWPORT;
1217 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1218 case IP_IPSEC_POLICY:
1219 if (IPSEC_ENABLED(ipv4)) {
1220 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1227 error = ENOPROTOOPT;
1233 switch (sopt->sopt_name) {
1236 if (inp->inp_options)
1237 error = sooptcopyout(sopt,
1238 mtod(inp->inp_options,
1240 inp->inp_options->m_len);
1242 sopt->sopt_valsize = 0;
1249 case IP_RECVRETOPTS:
1250 case IP_ORIGDSTADDR:
1251 case IP_RECVDSTADDR:
1264 case IP_RSSBUCKETID:
1265 case IP_RECVRSSBUCKETID:
1267 switch (sopt->sopt_name) {
1270 optval = inp->inp_ip_tos;
1274 optval = inp->inp_ip_ttl;
1278 optval = inp->inp_ip_minttl;
1281 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1282 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1285 optval = OPTBIT(INP_RECVOPTS);
1288 case IP_RECVRETOPTS:
1289 optval = OPTBIT(INP_RECVRETOPTS);
1292 case IP_RECVDSTADDR:
1293 optval = OPTBIT(INP_RECVDSTADDR);
1296 case IP_ORIGDSTADDR:
1297 optval = OPTBIT2(INP_ORIGDSTADDR);
1301 optval = OPTBIT(INP_RECVTTL);
1305 optval = OPTBIT(INP_RECVIF);
1309 if (inp->inp_flags & INP_HIGHPORT)
1310 optval = IP_PORTRANGE_HIGH;
1311 else if (inp->inp_flags & INP_LOWPORT)
1312 optval = IP_PORTRANGE_LOW;
1318 optval = OPTBIT(INP_ONESBCAST);
1321 optval = OPTBIT(INP_DONTFRAG);
1324 optval = OPTBIT(INP_BINDANY);
1327 optval = OPTBIT(INP_RECVTOS);
1330 optval = inp->inp_flowid;
1333 optval = inp->inp_flowtype;
1336 optval = OPTBIT2(INP_RECVFLOWID);
1339 case IP_RSSBUCKETID:
1340 retval = rss_hash2bucket(inp->inp_flowid,
1344 optval = rss_bucket;
1348 case IP_RECVRSSBUCKETID:
1349 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1353 optval = OPTBIT2(INP_BINDMULTI);
1356 error = sooptcopyout(sopt, &optval, sizeof optval);
1360 * Multicast socket options are processed by the in_mcast
1363 case IP_MULTICAST_IF:
1364 case IP_MULTICAST_VIF:
1365 case IP_MULTICAST_TTL:
1366 case IP_MULTICAST_LOOP:
1368 error = inp_getmoptions(inp, sopt);
1371 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1372 case IP_IPSEC_POLICY:
1373 if (IPSEC_ENABLED(ipv4)) {
1374 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1381 error = ENOPROTOOPT;
1390 * Routine called from ip_output() to loop back a copy of an IP multicast
1391 * packet to the input queue of a specified interface. Note that this
1392 * calls the output routine of the loopback "driver", but with an interface
1393 * pointer that might NOT be a loopback interface -- evil, but easier than
1394 * replicating that code here.
1397 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1403 * Make a deep copy of the packet because we're going to
1404 * modify the pack in order to generate checksums.
1406 copym = m_dup(m, M_NOWAIT);
1407 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1408 copym = m_pullup(copym, hlen);
1409 if (copym != NULL) {
1410 /* If needed, compute the checksum and mark it as valid. */
1411 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1412 in_delayed_cksum(copym);
1413 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1414 copym->m_pkthdr.csum_flags |=
1415 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1416 copym->m_pkthdr.csum_data = 0xffff;
1419 * We don't bother to fragment if the IP length is greater
1420 * than the interface's MTU. Can this possibly matter?
1422 ip = mtod(copym, struct ip *);
1424 ip->ip_sum = in_cksum(copym, hlen);
1425 if_simloop(ifp, copym, AF_INET, 0);