2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
37 #include "opt_ipsec.h"
38 #include "opt_kdtrace.h"
39 #include "opt_mbuf_stress_test.h"
40 #include "opt_mpath.h"
41 #include "opt_route.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/malloc.h>
51 #include <sys/protosw.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sysctl.h>
56 #include <sys/ucred.h>
59 #include <net/if_llatbl.h>
60 #include <net/netisr.h>
62 #include <net/route.h>
63 #include <net/flowtable.h>
65 #include <net/radix_mpath.h>
69 #include <netinet/in.h>
70 #include <netinet/in_kdtrace.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/ip.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/in_var.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/ip_options.h>
78 #include <netinet/sctp.h>
79 #include <netinet/sctp_crc32.h>
83 #include <netinet/ip_ipsec.h>
84 #include <netipsec/ipsec.h>
87 #include <machine/in_cksum.h>
89 #include <security/mac/mac_framework.h>
91 VNET_DEFINE(u_short, ip_id);
93 #ifdef MBUF_STRESS_TEST
94 static int mbuf_frag_size = 0;
95 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
96 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
99 static void ip_mloopback
100 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
103 extern int in_mcast_loop;
104 extern struct protosw inetsw[];
107 * IP output. The packet in mbuf chain m contains a skeletal IP
108 * header (with len, off, ttl, proto, tos, src, dst).
109 * The mbuf chain containing the packet will be freed.
110 * The mbuf opt, if present, will not be freed.
111 * If route ro is present and has ro_rt initialized, route lookup would be
112 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
113 * then result of route lookup is stored in ro->ro_rt.
115 * In the IP forwarding case, the packet will arrive with options already
116 * inserted, so must have a NULL opt pointer.
119 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
120 struct ip_moptions *imo, struct inpcb *inp)
123 struct ifnet *ifp = NULL; /* keep compiler happy */
125 int hlen = sizeof (struct ip);
127 int n; /* scratchpad */
129 struct sockaddr_in *dst;
130 const struct sockaddr_in *gw;
131 struct in_ifaddr *ia;
133 uint16_t ip_len, ip_off;
134 struct route iproute;
135 struct rtentry *rte; /* cache for ro->ro_rt */
137 struct m_tag *fwd_tag = NULL;
140 int no_route_but_check_spd = 0;
145 INP_LOCK_ASSERT(inp);
146 M_SETFIB(m, inp->inp_inc.inc_fibnum);
147 if (inp->inp_flowtype != M_HASHTYPE_NONE) {
148 m->m_pkthdr.flowid = inp->inp_flowid;
149 M_HASHTYPE_SET(m, M_HASHTYPE_OPAQUE);
155 bzero(ro, sizeof (*ro));
159 if (ro->ro_rt == NULL)
160 (void )flowtable_lookup(AF_INET, m, ro);
165 m = ip_insertoptions(m, opt, &len);
167 hlen = len; /* ip->ip_hl is updated above */
169 ip = mtod(m, struct ip *);
170 ip_len = ntohs(ip->ip_len);
171 ip_off = ntohs(ip->ip_off);
174 * Fill in IP header. If we are not allowing fragmentation,
175 * then the ip_id field is meaningless, but we don't set it
176 * to zero. Doing so causes various problems when devices along
177 * the path (routers, load balancers, firewalls, etc.) illegally
178 * disable DF on our packet. Note that a 16-bit counter
179 * will wrap around in less than 10 seconds at 100 Mbit/s on a
180 * medium with MTU 1500. See Steven M. Bellovin, "A Technique
181 * for Counting NATted Hosts", Proc. IMW'02, available at
182 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
184 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
185 ip->ip_v = IPVERSION;
186 ip->ip_hl = hlen >> 2;
187 ip->ip_id = ip_newid();
188 IPSTAT_INC(ips_localout);
190 /* Header already set, fetch hlen from there */
191 hlen = ip->ip_hl << 2;
197 * dst can be rewritten but always point to &ro->ro_dst
198 * gw is readonly but can be pointed either to dst OR rt_gatewy
199 * therefore we need restore GW if we're re-doing lookup
201 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
206 * If there is a cached route,
207 * check that it is to the same destination
208 * and is still up. If not, free it and try again.
209 * The address family should also be checked in case of sharing the
213 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
214 rte->rt_ifp == NULL ||
215 !RT_LINK_IS_UP(rte->rt_ifp) ||
216 dst->sin_family != AF_INET ||
217 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
223 if (rte == NULL && fwd_tag == NULL) {
224 bzero(dst, sizeof(*dst));
225 dst->sin_family = AF_INET;
226 dst->sin_len = sizeof(*dst);
227 dst->sin_addr = ip->ip_dst;
230 * If routing to interface only, short circuit routing lookup.
231 * The use of an all-ones broadcast address implies this; an
232 * interface is specified by the broadcast address of an interface,
233 * or the destination address of a ptp interface.
235 if (flags & IP_SENDONES) {
236 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
237 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
238 IPSTAT_INC(ips_noroute);
243 ip->ip_dst.s_addr = INADDR_BROADCAST;
244 dst->sin_addr = ip->ip_dst;
248 } else if (flags & IP_ROUTETOIF) {
249 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
250 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) {
251 IPSTAT_INC(ips_noroute);
258 isbroadcast = in_broadcast(dst->sin_addr, ifp);
259 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
260 imo != NULL && imo->imo_multicast_ifp != NULL) {
262 * Bypass the normal routing lookup for multicast
263 * packets if the interface is specified.
265 ifp = imo->imo_multicast_ifp;
269 isbroadcast = 0; /* fool gcc */
272 * We want to do any cloning requested by the link layer,
273 * as this is probably required in all cases for correct
274 * operation (as it is for ARP).
278 rtalloc_mpath_fib(ro,
279 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
280 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
282 in_rtalloc_ign(ro, 0,
283 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
288 rte->rt_ifp == NULL ||
289 !RT_LINK_IS_UP(rte->rt_ifp)) {
292 * There is no route for this packet, but it is
293 * possible that a matching SPD entry exists.
295 no_route_but_check_spd = 1;
296 mtu = 0; /* Silence GCC warning. */
299 IPSTAT_INC(ips_noroute);
300 error = EHOSTUNREACH;
303 ia = ifatoia(rte->rt_ifa);
305 counter_u64_add(rte->rt_pksent, 1);
306 if (rte->rt_flags & RTF_GATEWAY)
307 gw = (struct sockaddr_in *)rte->rt_gateway;
308 if (rte->rt_flags & RTF_HOST)
309 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
311 isbroadcast = in_broadcast(gw->sin_addr, ifp);
314 * Calculate MTU. If we have a route that is up, use that,
315 * otherwise use the interface's MTU.
317 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
319 * This case can happen if the user changed the MTU
320 * of an interface after enabling IP on it. Because
321 * most netifs don't keep track of routes pointing to
322 * them, there is no way for one to update all its
323 * routes when the MTU is changed.
325 if (rte->rt_mtu > ifp->if_mtu)
326 rte->rt_mtu = ifp->if_mtu;
331 /* Catch a possible divide by zero later. */
332 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
333 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
334 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
335 m->m_flags |= M_MCAST;
337 * IP destination address is multicast. Make sure "gw"
338 * still points to the address in "ro". (It may have been
339 * changed to point to a gateway address, above.)
343 * See if the caller provided any multicast options
346 ip->ip_ttl = imo->imo_multicast_ttl;
347 if (imo->imo_multicast_vif != -1)
350 ip_mcast_src(imo->imo_multicast_vif) :
353 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
355 * Confirm that the outgoing interface supports multicast.
357 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
358 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
359 IPSTAT_INC(ips_noroute);
365 * If source address not specified yet, use address
366 * of outgoing interface.
368 if (ip->ip_src.s_addr == INADDR_ANY) {
369 /* Interface may have no addresses. */
371 ip->ip_src = IA_SIN(ia)->sin_addr;
374 if ((imo == NULL && in_mcast_loop) ||
375 (imo && imo->imo_multicast_loop)) {
377 * Loop back multicast datagram if not expressly
378 * forbidden to do so, even if we are not a member
379 * of the group; ip_input() will filter it later,
380 * thus deferring a hash lookup and mutex acquisition
381 * at the expense of a cheap copy using m_copym().
383 ip_mloopback(ifp, m, dst, hlen);
386 * If we are acting as a multicast router, perform
387 * multicast forwarding as if the packet had just
388 * arrived on the interface to which we are about
389 * to send. The multicast forwarding function
390 * recursively calls this function, using the
391 * IP_FORWARDING flag to prevent infinite recursion.
393 * Multicasts that are looped back by ip_mloopback(),
394 * above, will be forwarded by the ip_input() routine,
397 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
399 * If rsvp daemon is not running, do not
400 * set ip_moptions. This ensures that the packet
401 * is multicast and not just sent down one link
402 * as prescribed by rsvpd.
407 ip_mforward(ip, ifp, m, imo) != 0) {
415 * Multicasts with a time-to-live of zero may be looped-
416 * back, above, but must not be transmitted on a network.
417 * Also, multicasts addressed to the loopback interface
418 * are not sent -- the above call to ip_mloopback() will
419 * loop back a copy. ip_input() will drop the copy if
420 * this host does not belong to the destination group on
421 * the loopback interface.
423 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
432 * If the source address is not specified yet, use the address
433 * of the outoing interface.
435 if (ip->ip_src.s_addr == INADDR_ANY) {
436 /* Interface may have no addresses. */
438 ip->ip_src = IA_SIN(ia)->sin_addr;
443 * Verify that we have any chance at all of being able to queue the
444 * packet or packet fragments, unless ALTQ is enabled on the given
445 * interface in which case packetdrop should be done by queueing.
447 n = ip_len / mtu + 1; /* how many fragments ? */
450 (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
452 (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
454 IPSTAT_INC(ips_odropped);
455 ifp->if_snd.ifq_drops += n;
460 * Look for broadcast address and
461 * verify user is allowed to send
465 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
466 error = EADDRNOTAVAIL;
469 if ((flags & IP_ALLOWBROADCAST) == 0) {
473 /* don't allow broadcast messages to be fragmented */
478 m->m_flags |= M_BCAST;
480 m->m_flags &= ~M_BCAST;
485 switch(ip_ipsec_output(&m, inp, &flags, &error)) {
492 break; /* Continue with packet processing. */
495 * Check if there was a route for this packet; return error if not.
497 if (no_route_but_check_spd) {
498 IPSTAT_INC(ips_noroute);
499 error = EHOSTUNREACH;
502 /* Update variables that are affected by ipsec4_output(). */
503 ip = mtod(m, struct ip *);
504 hlen = ip->ip_hl << 2;
507 /* Jump over all PFIL processing if hooks are not active. */
508 if (!PFIL_HOOKED(&V_inet_pfil_hook))
511 /* Run through list of hooks for output packets. */
512 odst.s_addr = ip->ip_dst.s_addr;
513 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
514 if (error != 0 || m == NULL)
517 ip = mtod(m, struct ip *);
519 /* See if destination IP address was changed by packet filter. */
520 if (odst.s_addr != ip->ip_dst.s_addr) {
521 m->m_flags |= M_SKIP_FIREWALL;
522 /* If destination is now ourself drop to ip_input(). */
523 if (in_localip(ip->ip_dst)) {
524 m->m_flags |= M_FASTFWD_OURS;
525 if (m->m_pkthdr.rcvif == NULL)
526 m->m_pkthdr.rcvif = V_loif;
527 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
528 m->m_pkthdr.csum_flags |=
529 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
530 m->m_pkthdr.csum_data = 0xffff;
532 m->m_pkthdr.csum_flags |=
533 CSUM_IP_CHECKED | CSUM_IP_VALID;
535 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
536 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
538 error = netisr_queue(NETISR_IP, m);
542 ifa_free(&ia->ia_ifa);
543 goto again; /* Redo the routing table lookup. */
547 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
548 if (m->m_flags & M_FASTFWD_OURS) {
549 if (m->m_pkthdr.rcvif == NULL)
550 m->m_pkthdr.rcvif = V_loif;
551 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
552 m->m_pkthdr.csum_flags |=
553 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
554 m->m_pkthdr.csum_data = 0xffff;
557 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
558 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
560 m->m_pkthdr.csum_flags |=
561 CSUM_IP_CHECKED | CSUM_IP_VALID;
563 error = netisr_queue(NETISR_IP, m);
566 /* Or forward to some other address? */
567 if ((m->m_flags & M_IP_NEXTHOP) &&
568 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
569 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
570 m->m_flags |= M_SKIP_FIREWALL;
571 m->m_flags &= ~M_IP_NEXTHOP;
572 m_tag_delete(m, fwd_tag);
574 ifa_free(&ia->ia_ifa);
579 /* 127/8 must not appear on wire - RFC1122. */
580 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
581 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
582 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
583 IPSTAT_INC(ips_badaddr);
584 error = EADDRNOTAVAIL;
589 m->m_pkthdr.csum_flags |= CSUM_IP;
590 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
592 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
595 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
596 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
597 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
602 * If small enough for interface, or the interface will take
603 * care of the fragmentation for us, we can just send directly.
606 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
607 ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
609 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
610 ip->ip_sum = in_cksum(m, hlen);
611 m->m_pkthdr.csum_flags &= ~CSUM_IP;
615 * Record statistics for this interface address.
616 * With CSUM_TSO the byte/packet count will be slightly
617 * incorrect because we count the IP+TCP headers only
618 * once instead of for every generated packet.
620 if (!(flags & IP_FORWARDING) && ia) {
621 if (m->m_pkthdr.csum_flags & CSUM_TSO)
622 ia->ia_ifa.if_opackets +=
623 m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
625 ia->ia_ifa.if_opackets++;
626 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
628 #ifdef MBUF_STRESS_TEST
629 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
630 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
633 * Reset layer specific mbuf flags
634 * to avoid confusing lower layers.
637 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
638 error = (*ifp->if_output)(ifp, m,
639 (const struct sockaddr *)gw, ro);
643 /* Balk when DF bit is set or the interface didn't support TSO. */
644 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
646 IPSTAT_INC(ips_cantfrag);
651 * Too large for interface; fragment if possible. If successful,
652 * on return, m will point to a list of packets to be sent.
654 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
661 /* Record statistics for this interface address. */
663 ia->ia_ifa.if_opackets++;
664 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
667 * Reset layer specific mbuf flags
668 * to avoid confusing upper layers.
672 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
673 error = (*ifp->if_output)(ifp, m,
674 (const struct sockaddr *)gw, ro);
680 IPSTAT_INC(ips_fragmented);
686 ifa_free(&ia->ia_ifa);
694 * Create a chain of fragments which fit the given mtu. m_frag points to the
695 * mbuf to be fragmented; on return it points to the chain with the fragments.
696 * Return 0 if no error. If error, m_frag may contain a partially built
697 * chain of fragments that should be freed by the caller.
699 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
702 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
703 u_long if_hwassist_flags)
706 int hlen = ip->ip_hl << 2;
707 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
709 struct mbuf *m0 = *m_frag; /* the original packet */
713 uint16_t ip_len, ip_off;
715 ip_len = ntohs(ip->ip_len);
716 ip_off = ntohs(ip->ip_off);
718 if (ip_off & IP_DF) { /* Fragmentation not allowed */
719 IPSTAT_INC(ips_cantfrag);
724 * Must be able to put at least 8 bytes per fragment.
730 * If the interface will not calculate checksums on
731 * fragmented packets, then do it here.
733 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
734 in_delayed_cksum(m0);
735 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
738 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
739 sctp_delayed_cksum(m0, hlen);
740 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
743 if (len > PAGE_SIZE) {
745 * Fragment large datagrams such that each segment
746 * contains a multiple of PAGE_SIZE amount of data,
747 * plus headers. This enables a receiver to perform
748 * page-flipping zero-copy optimizations.
750 * XXX When does this help given that sender and receiver
751 * could have different page sizes, and also mtu could
752 * be less than the receiver's page size ?
756 off = MIN(mtu, m0->m_pkthdr.len);
759 * firstlen (off - hlen) must be aligned on an
763 goto smart_frag_failure;
764 off = ((off - hlen) & ~7) + hlen;
765 newlen = (~PAGE_MASK) & mtu;
766 if ((newlen + sizeof (struct ip)) > mtu) {
767 /* we failed, go back the default */
778 firstlen = off - hlen;
779 mnext = &m0->m_nextpkt; /* pointer to next packet */
782 * Loop through length of segment after first fragment,
783 * make new header and copy data of each part and link onto chain.
784 * Here, m0 is the original packet, m is the fragment being created.
785 * The fragments are linked off the m_nextpkt of the original
786 * packet, which after processing serves as the first fragment.
788 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
789 struct ip *mhip; /* ip header on the fragment */
791 int mhlen = sizeof (struct ip);
793 m = m_gethdr(M_NOWAIT, MT_DATA);
796 IPSTAT_INC(ips_odropped);
800 * Make sure the complete packet header gets copied
801 * from the originating mbuf to the newly created
802 * mbuf. This also ensures that existing firewall
803 * classification(s), VLAN tags and so on get copied
804 * to the resulting fragmented packet(s):
806 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
809 IPSTAT_INC(ips_odropped);
813 * In the first mbuf, leave room for the link header, then
814 * copy the original IP header including options. The payload
815 * goes into an additional mbuf chain returned by m_copym().
817 m->m_data += max_linkhdr;
818 mhip = mtod(m, struct ip *);
820 if (hlen > sizeof (struct ip)) {
821 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
822 mhip->ip_v = IPVERSION;
823 mhip->ip_hl = mhlen >> 2;
826 /* XXX do we need to add ip_off below ? */
827 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
828 if (off + len >= ip_len)
831 mhip->ip_off |= IP_MF;
832 mhip->ip_len = htons((u_short)(len + mhlen));
833 m->m_next = m_copym(m0, off, len, M_NOWAIT);
834 if (m->m_next == NULL) { /* copy failed */
836 error = ENOBUFS; /* ??? */
837 IPSTAT_INC(ips_odropped);
840 m->m_pkthdr.len = mhlen + len;
842 mac_netinet_fragment(m0, m);
844 mhip->ip_off = htons(mhip->ip_off);
846 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
847 mhip->ip_sum = in_cksum(m, mhlen);
848 m->m_pkthdr.csum_flags &= ~CSUM_IP;
851 mnext = &m->m_nextpkt;
853 IPSTAT_ADD(ips_ofragments, nfrags);
856 * Update first fragment by trimming what's been copied out
857 * and updating header.
859 m_adj(m0, hlen + firstlen - ip_len);
860 m0->m_pkthdr.len = hlen + firstlen;
861 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
862 ip->ip_off = htons(ip_off | IP_MF);
864 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
865 ip->ip_sum = in_cksum(m0, hlen);
866 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
875 in_delayed_cksum(struct mbuf *m)
878 uint16_t csum, offset, ip_len;
880 ip = mtod(m, struct ip *);
881 offset = ip->ip_hl << 2 ;
882 ip_len = ntohs(ip->ip_len);
883 csum = in_cksum_skip(m, ip_len, offset);
884 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
886 offset += m->m_pkthdr.csum_data; /* checksum offset */
888 /* find the mbuf in the chain where the checksum starts*/
889 while ((m != NULL) && (offset >= m->m_len)) {
893 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
894 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
895 *(u_short *)(m->m_data + offset) = csum;
899 * IP socket option processing.
902 ip_ctloutput(struct socket *so, struct sockopt *sopt)
904 struct inpcb *inp = sotoinpcb(so);
908 if (sopt->sopt_level != IPPROTO_IP) {
911 if (sopt->sopt_level == SOL_SOCKET &&
912 sopt->sopt_dir == SOPT_SET) {
913 switch (sopt->sopt_name) {
916 if ((so->so_options & SO_REUSEADDR) != 0)
917 inp->inp_flags2 |= INP_REUSEADDR;
919 inp->inp_flags2 &= ~INP_REUSEADDR;
925 if ((so->so_options & SO_REUSEPORT) != 0)
926 inp->inp_flags2 |= INP_REUSEPORT;
928 inp->inp_flags2 &= ~INP_REUSEPORT;
934 inp->inp_inc.inc_fibnum = so->so_fibnum;
945 switch (sopt->sopt_dir) {
947 switch (sopt->sopt_name) {
954 if (sopt->sopt_valsize > MLEN) {
958 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
963 m->m_len = sopt->sopt_valsize;
964 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
971 error = ip_pcbopts(inp, sopt->sopt_name, m);
977 if (sopt->sopt_td != NULL) {
978 error = priv_check(sopt->sopt_td,
979 PRIV_NETINET_BINDANY);
996 error = sooptcopyin(sopt, &optval, sizeof optval,
1001 switch (sopt->sopt_name) {
1003 inp->inp_ip_tos = optval;
1007 inp->inp_ip_ttl = optval;
1011 if (optval >= 0 && optval <= MAXTTL)
1012 inp->inp_ip_minttl = optval;
1017 #define OPTSET(bit) do { \
1020 inp->inp_flags |= bit; \
1022 inp->inp_flags &= ~bit; \
1027 OPTSET(INP_RECVOPTS);
1030 case IP_RECVRETOPTS:
1031 OPTSET(INP_RECVRETOPTS);
1034 case IP_RECVDSTADDR:
1035 OPTSET(INP_RECVDSTADDR);
1039 OPTSET(INP_RECVTTL);
1051 OPTSET(INP_ONESBCAST);
1054 OPTSET(INP_DONTFRAG);
1057 OPTSET(INP_BINDANY);
1060 OPTSET(INP_RECVTOS);
1067 * Multicast socket options are processed by the in_mcast
1070 case IP_MULTICAST_IF:
1071 case IP_MULTICAST_VIF:
1072 case IP_MULTICAST_TTL:
1073 case IP_MULTICAST_LOOP:
1074 case IP_ADD_MEMBERSHIP:
1075 case IP_DROP_MEMBERSHIP:
1076 case IP_ADD_SOURCE_MEMBERSHIP:
1077 case IP_DROP_SOURCE_MEMBERSHIP:
1078 case IP_BLOCK_SOURCE:
1079 case IP_UNBLOCK_SOURCE:
1081 case MCAST_JOIN_GROUP:
1082 case MCAST_LEAVE_GROUP:
1083 case MCAST_JOIN_SOURCE_GROUP:
1084 case MCAST_LEAVE_SOURCE_GROUP:
1085 case MCAST_BLOCK_SOURCE:
1086 case MCAST_UNBLOCK_SOURCE:
1087 error = inp_setmoptions(inp, sopt);
1091 error = sooptcopyin(sopt, &optval, sizeof optval,
1098 case IP_PORTRANGE_DEFAULT:
1099 inp->inp_flags &= ~(INP_LOWPORT);
1100 inp->inp_flags &= ~(INP_HIGHPORT);
1103 case IP_PORTRANGE_HIGH:
1104 inp->inp_flags &= ~(INP_LOWPORT);
1105 inp->inp_flags |= INP_HIGHPORT;
1108 case IP_PORTRANGE_LOW:
1109 inp->inp_flags &= ~(INP_HIGHPORT);
1110 inp->inp_flags |= INP_LOWPORT;
1121 case IP_IPSEC_POLICY:
1126 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1128 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1130 req = mtod(m, caddr_t);
1131 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1132 m->m_len, (sopt->sopt_td != NULL) ?
1133 sopt->sopt_td->td_ucred : NULL);
1140 error = ENOPROTOOPT;
1146 switch (sopt->sopt_name) {
1149 if (inp->inp_options)
1150 error = sooptcopyout(sopt,
1151 mtod(inp->inp_options,
1153 inp->inp_options->m_len);
1155 sopt->sopt_valsize = 0;
1162 case IP_RECVRETOPTS:
1163 case IP_RECVDSTADDR:
1172 switch (sopt->sopt_name) {
1175 optval = inp->inp_ip_tos;
1179 optval = inp->inp_ip_ttl;
1183 optval = inp->inp_ip_minttl;
1186 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1189 optval = OPTBIT(INP_RECVOPTS);
1192 case IP_RECVRETOPTS:
1193 optval = OPTBIT(INP_RECVRETOPTS);
1196 case IP_RECVDSTADDR:
1197 optval = OPTBIT(INP_RECVDSTADDR);
1201 optval = OPTBIT(INP_RECVTTL);
1205 optval = OPTBIT(INP_RECVIF);
1209 if (inp->inp_flags & INP_HIGHPORT)
1210 optval = IP_PORTRANGE_HIGH;
1211 else if (inp->inp_flags & INP_LOWPORT)
1212 optval = IP_PORTRANGE_LOW;
1218 optval = OPTBIT(INP_FAITH);
1222 optval = OPTBIT(INP_ONESBCAST);
1225 optval = OPTBIT(INP_DONTFRAG);
1228 optval = OPTBIT(INP_BINDANY);
1231 optval = OPTBIT(INP_RECVTOS);
1234 error = sooptcopyout(sopt, &optval, sizeof optval);
1238 * Multicast socket options are processed by the in_mcast
1241 case IP_MULTICAST_IF:
1242 case IP_MULTICAST_VIF:
1243 case IP_MULTICAST_TTL:
1244 case IP_MULTICAST_LOOP:
1246 error = inp_getmoptions(inp, sopt);
1250 case IP_IPSEC_POLICY:
1252 struct mbuf *m = NULL;
1257 req = mtod(m, caddr_t);
1260 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1262 error = soopt_mcopyout(sopt, m); /* XXX */
1270 error = ENOPROTOOPT;
1279 * Routine called from ip_output() to loop back a copy of an IP multicast
1280 * packet to the input queue of a specified interface. Note that this
1281 * calls the output routine of the loopback "driver", but with an interface
1282 * pointer that might NOT be a loopback interface -- evil, but easier than
1283 * replicating that code here.
1286 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1289 register struct ip *ip;
1293 * Make a deep copy of the packet because we're going to
1294 * modify the pack in order to generate checksums.
1296 copym = m_dup(m, M_NOWAIT);
1297 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1298 copym = m_pullup(copym, hlen);
1299 if (copym != NULL) {
1300 /* If needed, compute the checksum and mark it as valid. */
1301 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1302 in_delayed_cksum(copym);
1303 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1304 copym->m_pkthdr.csum_flags |=
1305 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1306 copym->m_pkthdr.csum_data = 0xffff;
1309 * We don't bother to fragment if the IP length is greater
1310 * than the interface's MTU. Can this possibly matter?
1312 ip = mtod(copym, struct ip *);
1314 ip->ip_sum = in_cksum(copym, hlen);
1316 if (dst->sin_family != AF_INET) {
1317 printf("ip_mloopback: bad address family %d\n",
1319 dst->sin_family = AF_INET;
1322 if_simloop(ifp, copym, dst->sin_family, 0);