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 *error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, 0, inp);
126 if ((*error) != 0 || m == NULL)
127 return 1; /* Finished */
129 ip = mtod(m, struct ip *);
131 /* See if destination IP address was changed by packet filter. */
132 if (odst.s_addr != ip->ip_dst.s_addr) {
133 m->m_flags |= M_SKIP_FIREWALL;
134 /* If destination is now ourself drop to ip_input(). */
135 if (in_localip(ip->ip_dst)) {
136 m->m_flags |= M_FASTFWD_OURS;
137 if (m->m_pkthdr.rcvif == NULL)
138 m->m_pkthdr.rcvif = V_loif;
139 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
140 m->m_pkthdr.csum_flags |=
141 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
142 m->m_pkthdr.csum_data = 0xffff;
144 m->m_pkthdr.csum_flags |=
145 CSUM_IP_CHECKED | CSUM_IP_VALID;
147 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
148 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
150 *error = netisr_queue(NETISR_IP, m);
151 return 1; /* Finished */
154 bzero(dst, sizeof(*dst));
155 dst->sin_family = AF_INET;
156 dst->sin_len = sizeof(*dst);
157 dst->sin_addr = ip->ip_dst;
159 return -1; /* Reloop */
161 /* See if fib was changed by packet filter. */
162 if ((*fibnum) != M_GETFIB(m)) {
163 m->m_flags |= M_SKIP_FIREWALL;
164 *fibnum = M_GETFIB(m);
165 return -1; /* Reloop for FIB change */
168 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
169 if (m->m_flags & M_FASTFWD_OURS) {
170 if (m->m_pkthdr.rcvif == NULL)
171 m->m_pkthdr.rcvif = V_loif;
172 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
173 m->m_pkthdr.csum_flags |=
174 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
175 m->m_pkthdr.csum_data = 0xffff;
178 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
179 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
181 m->m_pkthdr.csum_flags |=
182 CSUM_IP_CHECKED | CSUM_IP_VALID;
184 *error = netisr_queue(NETISR_IP, m);
185 return 1; /* Finished */
187 /* Or forward to some other address? */
188 if ((m->m_flags & M_IP_NEXTHOP) &&
189 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
190 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
191 m->m_flags |= M_SKIP_FIREWALL;
192 m->m_flags &= ~M_IP_NEXTHOP;
193 m_tag_delete(m, fwd_tag);
195 return -1; /* Reloop for CHANGE of dst */
202 * IP output. The packet in mbuf chain m contains a skeletal IP
203 * header (with len, off, ttl, proto, tos, src, dst).
204 * The mbuf chain containing the packet will be freed.
205 * The mbuf opt, if present, will not be freed.
206 * If route ro is present and has ro_rt initialized, route lookup would be
207 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
208 * then result of route lookup is stored in ro->ro_rt.
210 * In the IP forwarding case, the packet will arrive with options already
211 * inserted, so must have a NULL opt pointer.
214 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
215 struct ip_moptions *imo, struct inpcb *inp)
217 struct rm_priotracker in_ifa_tracker;
219 struct ifnet *ifp = NULL; /* keep compiler happy */
221 int hlen = sizeof (struct ip);
224 struct sockaddr_in *dst;
225 const struct sockaddr_in *gw;
226 struct in_ifaddr *ia;
228 uint16_t ip_len, ip_off;
229 struct route iproute;
230 struct rtentry *rte; /* cache for ro->ro_rt */
232 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
233 int no_route_but_check_spd = 0;
238 INP_LOCK_ASSERT(inp);
239 M_SETFIB(m, inp->inp_inc.inc_fibnum);
240 if ((flags & IP_NODEFAULTFLOWID) == 0) {
241 m->m_pkthdr.flowid = inp->inp_flowid;
242 M_HASHTYPE_SET(m, inp->inp_flowtype);
248 bzero(ro, sizeof (*ro));
253 m = ip_insertoptions(m, opt, &len);
255 hlen = len; /* ip->ip_hl is updated above */
257 ip = mtod(m, struct ip *);
258 ip_len = ntohs(ip->ip_len);
259 ip_off = ntohs(ip->ip_off);
261 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
262 ip->ip_v = IPVERSION;
263 ip->ip_hl = hlen >> 2;
265 IPSTAT_INC(ips_localout);
267 /* Header already set, fetch hlen from there */
268 hlen = ip->ip_hl << 2;
274 * dst can be rewritten but always points to &ro->ro_dst.
275 * gw is readonly but can point either to dst OR rt_gateway,
276 * therefore we need restore gw if we're redoing lookup.
278 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
279 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
282 bzero(dst, sizeof(*dst));
283 dst->sin_family = AF_INET;
284 dst->sin_len = sizeof(*dst);
285 dst->sin_addr = ip->ip_dst;
290 * Validate route against routing table additions;
291 * a better/more specific route might have been added.
294 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
296 * If there is a cached route,
297 * check that it is to the same destination
298 * and is still up. If not, free it and try again.
299 * The address family should also be checked in case of sharing the
301 * Also check whether routing cache needs invalidation.
304 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
305 rte->rt_ifp == NULL ||
306 !RT_LINK_IS_UP(rte->rt_ifp) ||
307 dst->sin_family != AF_INET ||
308 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
309 RO_INVALIDATE_CACHE(ro);
314 * If routing to interface only, short circuit routing lookup.
315 * The use of an all-ones broadcast address implies this; an
316 * interface is specified by the broadcast address of an interface,
317 * or the destination address of a ptp interface.
319 if (flags & IP_SENDONES) {
320 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
321 M_GETFIB(m)))) == NULL &&
322 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
323 M_GETFIB(m)))) == NULL) {
324 IPSTAT_INC(ips_noroute);
328 ip->ip_dst.s_addr = INADDR_BROADCAST;
329 dst->sin_addr = ip->ip_dst;
333 } else if (flags & IP_ROUTETOIF) {
334 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
335 M_GETFIB(m)))) == NULL &&
336 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
337 M_GETFIB(m)))) == NULL) {
338 IPSTAT_INC(ips_noroute);
344 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
345 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
346 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
347 imo != NULL && imo->imo_multicast_ifp != NULL) {
349 * Bypass the normal routing lookup for multicast
350 * packets if the interface is specified.
352 ifp = imo->imo_multicast_ifp;
353 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
354 isbroadcast = 0; /* fool gcc */
357 * We want to do any cloning requested by the link layer,
358 * as this is probably required in all cases for correct
359 * operation (as it is for ARP).
363 rtalloc_mpath_fib(ro,
364 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
367 in_rtalloc_ign(ro, 0, fibnum);
372 (rte->rt_flags & RTF_UP) == 0 ||
373 rte->rt_ifp == NULL ||
374 !RT_LINK_IS_UP(rte->rt_ifp)) {
375 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
377 * There is no route for this packet, but it is
378 * possible that a matching SPD entry exists.
380 no_route_but_check_spd = 1;
381 mtu = 0; /* Silence GCC warning. */
384 IPSTAT_INC(ips_noroute);
385 error = EHOSTUNREACH;
388 ia = ifatoia(rte->rt_ifa);
390 counter_u64_add(rte->rt_pksent, 1);
391 rt_update_ro_flags(ro);
392 if (rte->rt_flags & RTF_GATEWAY)
393 gw = (struct sockaddr_in *)rte->rt_gateway;
394 if (rte->rt_flags & RTF_HOST)
395 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
396 else if (ifp->if_flags & IFF_BROADCAST)
397 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
403 * Calculate MTU. If we have a route that is up, use that,
404 * otherwise use the interface's MTU.
406 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
410 /* Catch a possible divide by zero later. */
411 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
412 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
414 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
415 m->m_flags |= M_MCAST;
417 * IP destination address is multicast. Make sure "gw"
418 * still points to the address in "ro". (It may have been
419 * changed to point to a gateway address, above.)
423 * See if the caller provided any multicast options
426 ip->ip_ttl = imo->imo_multicast_ttl;
427 if (imo->imo_multicast_vif != -1)
430 ip_mcast_src(imo->imo_multicast_vif) :
433 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
435 * Confirm that the outgoing interface supports multicast.
437 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
438 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
439 IPSTAT_INC(ips_noroute);
445 * If source address not specified yet, use address
446 * of outgoing interface.
448 if (ip->ip_src.s_addr == INADDR_ANY) {
449 /* Interface may have no addresses. */
451 ip->ip_src = IA_SIN(ia)->sin_addr;
454 if ((imo == NULL && in_mcast_loop) ||
455 (imo && imo->imo_multicast_loop)) {
457 * Loop back multicast datagram if not expressly
458 * forbidden to do so, even if we are not a member
459 * of the group; ip_input() will filter it later,
460 * thus deferring a hash lookup and mutex acquisition
461 * at the expense of a cheap copy using m_copym().
463 ip_mloopback(ifp, m, hlen);
466 * If we are acting as a multicast router, perform
467 * multicast forwarding as if the packet had just
468 * arrived on the interface to which we are about
469 * to send. The multicast forwarding function
470 * recursively calls this function, using the
471 * IP_FORWARDING flag to prevent infinite recursion.
473 * Multicasts that are looped back by ip_mloopback(),
474 * above, will be forwarded by the ip_input() routine,
477 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
479 * If rsvp daemon is not running, do not
480 * set ip_moptions. This ensures that the packet
481 * is multicast and not just sent down one link
482 * as prescribed by rsvpd.
487 ip_mforward(ip, ifp, m, imo) != 0) {
495 * Multicasts with a time-to-live of zero may be looped-
496 * back, above, but must not be transmitted on a network.
497 * Also, multicasts addressed to the loopback interface
498 * are not sent -- the above call to ip_mloopback() will
499 * loop back a copy. ip_input() will drop the copy if
500 * this host does not belong to the destination group on
501 * the loopback interface.
503 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
512 * If the source address is not specified yet, use the address
513 * of the outoing interface.
515 if (ip->ip_src.s_addr == INADDR_ANY) {
516 /* Interface may have no addresses. */
518 ip->ip_src = IA_SIN(ia)->sin_addr;
523 * Look for broadcast address and
524 * verify user is allowed to send
528 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
529 error = EADDRNOTAVAIL;
532 if ((flags & IP_ALLOWBROADCAST) == 0) {
536 /* don't allow broadcast messages to be fragmented */
541 m->m_flags |= M_BCAST;
543 m->m_flags &= ~M_BCAST;
547 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
548 if (IPSEC_ENABLED(ipv4)) {
549 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
550 if (error == EINPROGRESS)
556 * Check if there was a route for this packet; return error if not.
558 if (no_route_but_check_spd) {
559 IPSTAT_INC(ips_noroute);
560 error = EHOSTUNREACH;
563 /* Update variables that are affected by ipsec4_output(). */
564 ip = mtod(m, struct ip *);
565 hlen = ip->ip_hl << 2;
568 /* Jump over all PFIL processing if hooks are not active. */
569 if (PFIL_HOOKED(&V_inet_pfil_hook)) {
570 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
571 case 1: /* Finished */
574 case 0: /* Continue normally */
575 ip = mtod(m, struct ip *);
578 case -1: /* Need to try again */
579 /* Reset everything for a new round */
581 ro->ro_prepend = NULL;
584 ip = mtod(m, struct ip *);
590 /* 127/8 must not appear on wire - RFC1122. */
591 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
592 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
593 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
594 IPSTAT_INC(ips_badaddr);
595 error = EADDRNOTAVAIL;
600 m->m_pkthdr.csum_flags |= CSUM_IP;
601 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
603 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
606 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
607 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
608 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
613 * If small enough for interface, or the interface will take
614 * care of the fragmentation for us, we can just send directly.
617 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
619 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
620 ip->ip_sum = in_cksum(m, hlen);
621 m->m_pkthdr.csum_flags &= ~CSUM_IP;
625 * Record statistics for this interface address.
626 * With CSUM_TSO the byte/packet count will be slightly
627 * incorrect because we count the IP+TCP headers only
628 * once instead of for every generated packet.
630 if (!(flags & IP_FORWARDING) && ia) {
631 if (m->m_pkthdr.csum_flags & CSUM_TSO)
632 counter_u64_add(ia->ia_ifa.ifa_opackets,
633 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
635 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
637 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
639 #ifdef MBUF_STRESS_TEST
640 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
641 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
644 * Reset layer specific mbuf flags
645 * to avoid confusing lower layers.
648 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
651 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
652 in_pcboutput_txrtlmt(inp, ifp, m);
653 /* stamp send tag on mbuf */
654 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
656 m->m_pkthdr.snd_tag = NULL;
659 error = (*ifp->if_output)(ifp, m,
660 (const struct sockaddr *)gw, ro);
662 /* check for route change */
664 in_pcboutput_eagain(inp);
669 /* Balk when DF bit is set or the interface didn't support TSO. */
670 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
672 IPSTAT_INC(ips_cantfrag);
677 * Too large for interface; fragment if possible. If successful,
678 * on return, m will point to a list of packets to be sent.
680 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
687 /* Record statistics for this interface address. */
689 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
690 counter_u64_add(ia->ia_ifa.ifa_obytes,
694 * Reset layer specific mbuf flags
695 * to avoid confusing upper layers.
699 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
700 mtod(m, struct ip *), NULL);
703 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
704 in_pcboutput_txrtlmt(inp, ifp, m);
705 /* stamp send tag on mbuf */
706 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
708 m->m_pkthdr.snd_tag = NULL;
711 error = (*ifp->if_output)(ifp, m,
712 (const struct sockaddr *)gw, ro);
714 /* check for route change */
716 in_pcboutput_eagain(inp);
723 IPSTAT_INC(ips_fragmented);
728 else if (rte == NULL)
730 * If the caller supplied a route but somehow the reference
731 * to it has been released need to prevent the caller
732 * calling RTFREE on it again.
743 * Create a chain of fragments which fit the given mtu. m_frag points to the
744 * mbuf to be fragmented; on return it points to the chain with the fragments.
745 * Return 0 if no error. If error, m_frag may contain a partially built
746 * chain of fragments that should be freed by the caller.
748 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
751 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
752 u_long if_hwassist_flags)
755 int hlen = ip->ip_hl << 2;
756 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
758 struct mbuf *m0 = *m_frag; /* the original packet */
762 uint16_t ip_len, ip_off;
764 ip_len = ntohs(ip->ip_len);
765 ip_off = ntohs(ip->ip_off);
767 if (ip_off & IP_DF) { /* Fragmentation not allowed */
768 IPSTAT_INC(ips_cantfrag);
773 * Must be able to put at least 8 bytes per fragment.
779 * If the interface will not calculate checksums on
780 * fragmented packets, then do it here.
782 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
783 in_delayed_cksum(m0);
784 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
787 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
788 sctp_delayed_cksum(m0, hlen);
789 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
792 if (len > PAGE_SIZE) {
794 * Fragment large datagrams such that each segment
795 * contains a multiple of PAGE_SIZE amount of data,
796 * plus headers. This enables a receiver to perform
797 * page-flipping zero-copy optimizations.
799 * XXX When does this help given that sender and receiver
800 * could have different page sizes, and also mtu could
801 * be less than the receiver's page size ?
805 off = MIN(mtu, m0->m_pkthdr.len);
808 * firstlen (off - hlen) must be aligned on an
812 goto smart_frag_failure;
813 off = ((off - hlen) & ~7) + hlen;
814 newlen = (~PAGE_MASK) & mtu;
815 if ((newlen + sizeof (struct ip)) > mtu) {
816 /* we failed, go back the default */
827 firstlen = off - hlen;
828 mnext = &m0->m_nextpkt; /* pointer to next packet */
831 * Loop through length of segment after first fragment,
832 * make new header and copy data of each part and link onto chain.
833 * Here, m0 is the original packet, m is the fragment being created.
834 * The fragments are linked off the m_nextpkt of the original
835 * packet, which after processing serves as the first fragment.
837 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
838 struct ip *mhip; /* ip header on the fragment */
840 int mhlen = sizeof (struct ip);
842 m = m_gethdr(M_NOWAIT, MT_DATA);
845 IPSTAT_INC(ips_odropped);
849 * Make sure the complete packet header gets copied
850 * from the originating mbuf to the newly created
851 * mbuf. This also ensures that existing firewall
852 * classification(s), VLAN tags and so on get copied
853 * to the resulting fragmented packet(s):
855 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
858 IPSTAT_INC(ips_odropped);
862 * In the first mbuf, leave room for the link header, then
863 * copy the original IP header including options. The payload
864 * goes into an additional mbuf chain returned by m_copym().
866 m->m_data += max_linkhdr;
867 mhip = mtod(m, struct ip *);
869 if (hlen > sizeof (struct ip)) {
870 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
871 mhip->ip_v = IPVERSION;
872 mhip->ip_hl = mhlen >> 2;
875 /* XXX do we need to add ip_off below ? */
876 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
877 if (off + len >= ip_len)
880 mhip->ip_off |= IP_MF;
881 mhip->ip_len = htons((u_short)(len + mhlen));
882 m->m_next = m_copym(m0, off, len, M_NOWAIT);
883 if (m->m_next == NULL) { /* copy failed */
885 error = ENOBUFS; /* ??? */
886 IPSTAT_INC(ips_odropped);
889 m->m_pkthdr.len = mhlen + len;
891 mac_netinet_fragment(m0, m);
893 mhip->ip_off = htons(mhip->ip_off);
895 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
896 mhip->ip_sum = in_cksum(m, mhlen);
897 m->m_pkthdr.csum_flags &= ~CSUM_IP;
900 mnext = &m->m_nextpkt;
902 IPSTAT_ADD(ips_ofragments, nfrags);
905 * Update first fragment by trimming what's been copied out
906 * and updating header.
908 m_adj(m0, hlen + firstlen - ip_len);
909 m0->m_pkthdr.len = hlen + firstlen;
910 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
911 ip->ip_off = htons(ip_off | IP_MF);
913 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
914 ip->ip_sum = in_cksum(m0, hlen);
915 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
924 in_delayed_cksum(struct mbuf *m)
928 uint16_t cklen, csum, offset;
930 ip = mtod(m, struct ip *);
931 offset = ip->ip_hl << 2 ;
933 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
934 /* if udp header is not in the first mbuf copy udplen */
935 if (offset + sizeof(struct udphdr) > m->m_len)
936 m_copydata(m, offset + offsetof(struct udphdr,
937 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
939 uh = (struct udphdr *)mtodo(m, offset);
940 cklen = ntohs(uh->uh_ulen);
942 csum = in_cksum_skip(m, cklen + offset, offset);
946 cklen = ntohs(ip->ip_len);
947 csum = in_cksum_skip(m, cklen, offset);
949 offset += m->m_pkthdr.csum_data; /* checksum offset */
951 if (offset + sizeof(csum) > m->m_len)
952 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
954 *(u_short *)mtodo(m, offset) = csum;
958 * IP socket option processing.
961 ip_ctloutput(struct socket *so, struct sockopt *sopt)
963 struct inpcb *inp = sotoinpcb(so);
971 if (sopt->sopt_level != IPPROTO_IP) {
974 if (sopt->sopt_level == SOL_SOCKET &&
975 sopt->sopt_dir == SOPT_SET) {
976 switch (sopt->sopt_name) {
979 if ((so->so_options & SO_REUSEADDR) != 0)
980 inp->inp_flags2 |= INP_REUSEADDR;
982 inp->inp_flags2 &= ~INP_REUSEADDR;
988 if ((so->so_options & SO_REUSEPORT) != 0)
989 inp->inp_flags2 |= INP_REUSEPORT;
991 inp->inp_flags2 &= ~INP_REUSEPORT;
997 inp->inp_inc.inc_fibnum = so->so_fibnum;
1001 case SO_MAX_PACING_RATE:
1004 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1018 switch (sopt->sopt_dir) {
1020 switch (sopt->sopt_name) {
1027 if (sopt->sopt_valsize > MLEN) {
1031 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1036 m->m_len = sopt->sopt_valsize;
1037 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1044 error = ip_pcbopts(inp, sopt->sopt_name, m);
1050 if (sopt->sopt_td != NULL) {
1051 error = priv_check(sopt->sopt_td,
1052 PRIV_NETINET_BINDANY);
1059 case IP_RSS_LISTEN_BUCKET:
1065 case IP_RECVRETOPTS:
1066 case IP_ORIGDSTADDR:
1067 case IP_RECVDSTADDR:
1075 case IP_RECVRSSBUCKETID:
1077 error = sooptcopyin(sopt, &optval, sizeof optval,
1082 switch (sopt->sopt_name) {
1084 inp->inp_ip_tos = optval;
1088 inp->inp_ip_ttl = optval;
1092 if (optval >= 0 && optval <= MAXTTL)
1093 inp->inp_ip_minttl = optval;
1098 #define OPTSET(bit) do { \
1101 inp->inp_flags |= bit; \
1103 inp->inp_flags &= ~bit; \
1107 #define OPTSET2(bit, val) do { \
1110 inp->inp_flags2 |= bit; \
1112 inp->inp_flags2 &= ~bit; \
1117 OPTSET(INP_RECVOPTS);
1120 case IP_RECVRETOPTS:
1121 OPTSET(INP_RECVRETOPTS);
1124 case IP_RECVDSTADDR:
1125 OPTSET(INP_RECVDSTADDR);
1128 case IP_ORIGDSTADDR:
1129 OPTSET2(INP_ORIGDSTADDR, optval);
1133 OPTSET(INP_RECVTTL);
1141 OPTSET(INP_ONESBCAST);
1144 OPTSET(INP_DONTFRAG);
1147 OPTSET(INP_BINDANY);
1150 OPTSET(INP_RECVTOS);
1153 OPTSET2(INP_BINDMULTI, optval);
1156 OPTSET2(INP_RECVFLOWID, optval);
1159 case IP_RSS_LISTEN_BUCKET:
1160 if ((optval >= 0) &&
1161 (optval < rss_getnumbuckets())) {
1162 inp->inp_rss_listen_bucket = optval;
1163 OPTSET2(INP_RSS_BUCKET_SET, 1);
1168 case IP_RECVRSSBUCKETID:
1169 OPTSET2(INP_RECVRSSBUCKETID, optval);
1178 * Multicast socket options are processed by the in_mcast
1181 case IP_MULTICAST_IF:
1182 case IP_MULTICAST_VIF:
1183 case IP_MULTICAST_TTL:
1184 case IP_MULTICAST_LOOP:
1185 case IP_ADD_MEMBERSHIP:
1186 case IP_DROP_MEMBERSHIP:
1187 case IP_ADD_SOURCE_MEMBERSHIP:
1188 case IP_DROP_SOURCE_MEMBERSHIP:
1189 case IP_BLOCK_SOURCE:
1190 case IP_UNBLOCK_SOURCE:
1192 case MCAST_JOIN_GROUP:
1193 case MCAST_LEAVE_GROUP:
1194 case MCAST_JOIN_SOURCE_GROUP:
1195 case MCAST_LEAVE_SOURCE_GROUP:
1196 case MCAST_BLOCK_SOURCE:
1197 case MCAST_UNBLOCK_SOURCE:
1198 error = inp_setmoptions(inp, sopt);
1202 error = sooptcopyin(sopt, &optval, sizeof optval,
1209 case IP_PORTRANGE_DEFAULT:
1210 inp->inp_flags &= ~(INP_LOWPORT);
1211 inp->inp_flags &= ~(INP_HIGHPORT);
1214 case IP_PORTRANGE_HIGH:
1215 inp->inp_flags &= ~(INP_LOWPORT);
1216 inp->inp_flags |= INP_HIGHPORT;
1219 case IP_PORTRANGE_LOW:
1220 inp->inp_flags &= ~(INP_HIGHPORT);
1221 inp->inp_flags |= INP_LOWPORT;
1231 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1232 case IP_IPSEC_POLICY:
1233 if (IPSEC_ENABLED(ipv4)) {
1234 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1241 error = ENOPROTOOPT;
1247 switch (sopt->sopt_name) {
1250 if (inp->inp_options)
1251 error = sooptcopyout(sopt,
1252 mtod(inp->inp_options,
1254 inp->inp_options->m_len);
1256 sopt->sopt_valsize = 0;
1263 case IP_RECVRETOPTS:
1264 case IP_ORIGDSTADDR:
1265 case IP_RECVDSTADDR:
1278 case IP_RSSBUCKETID:
1279 case IP_RECVRSSBUCKETID:
1281 switch (sopt->sopt_name) {
1284 optval = inp->inp_ip_tos;
1288 optval = inp->inp_ip_ttl;
1292 optval = inp->inp_ip_minttl;
1295 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1296 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1299 optval = OPTBIT(INP_RECVOPTS);
1302 case IP_RECVRETOPTS:
1303 optval = OPTBIT(INP_RECVRETOPTS);
1306 case IP_RECVDSTADDR:
1307 optval = OPTBIT(INP_RECVDSTADDR);
1310 case IP_ORIGDSTADDR:
1311 optval = OPTBIT2(INP_ORIGDSTADDR);
1315 optval = OPTBIT(INP_RECVTTL);
1319 optval = OPTBIT(INP_RECVIF);
1323 if (inp->inp_flags & INP_HIGHPORT)
1324 optval = IP_PORTRANGE_HIGH;
1325 else if (inp->inp_flags & INP_LOWPORT)
1326 optval = IP_PORTRANGE_LOW;
1332 optval = OPTBIT(INP_ONESBCAST);
1335 optval = OPTBIT(INP_DONTFRAG);
1338 optval = OPTBIT(INP_BINDANY);
1341 optval = OPTBIT(INP_RECVTOS);
1344 optval = inp->inp_flowid;
1347 optval = inp->inp_flowtype;
1350 optval = OPTBIT2(INP_RECVFLOWID);
1353 case IP_RSSBUCKETID:
1354 retval = rss_hash2bucket(inp->inp_flowid,
1358 optval = rss_bucket;
1362 case IP_RECVRSSBUCKETID:
1363 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1367 optval = OPTBIT2(INP_BINDMULTI);
1370 error = sooptcopyout(sopt, &optval, sizeof optval);
1374 * Multicast socket options are processed by the in_mcast
1377 case IP_MULTICAST_IF:
1378 case IP_MULTICAST_VIF:
1379 case IP_MULTICAST_TTL:
1380 case IP_MULTICAST_LOOP:
1382 error = inp_getmoptions(inp, sopt);
1385 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1386 case IP_IPSEC_POLICY:
1387 if (IPSEC_ENABLED(ipv4)) {
1388 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1395 error = ENOPROTOOPT;
1404 * Routine called from ip_output() to loop back a copy of an IP multicast
1405 * packet to the input queue of a specified interface. Note that this
1406 * calls the output routine of the loopback "driver", but with an interface
1407 * pointer that might NOT be a loopback interface -- evil, but easier than
1408 * replicating that code here.
1411 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1417 * Make a deep copy of the packet because we're going to
1418 * modify the pack in order to generate checksums.
1420 copym = m_dup(m, M_NOWAIT);
1421 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1422 copym = m_pullup(copym, hlen);
1423 if (copym != NULL) {
1424 /* If needed, compute the checksum and mark it as valid. */
1425 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1426 in_delayed_cksum(copym);
1427 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1428 copym->m_pkthdr.csum_flags |=
1429 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1430 copym->m_pkthdr.csum_data = 0xffff;
1433 * We don't bother to fragment if the IP length is greater
1434 * than the interface's MTU. Can this possibly matter?
1436 ip = mtod(copym, struct ip *);
1438 ip->ip_sum = in_cksum(copym, hlen);
1439 if_simloop(ifp, copym, AF_INET, 0);