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("$FreeBSD$");
36 #include "opt_ipsec.h"
37 #include "opt_kdtrace.h"
38 #include "opt_mbuf_stress_test.h"
39 #include "opt_mpath.h"
40 #include "opt_route.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
46 #include <sys/malloc.h>
50 #include <sys/protosw.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/sysctl.h>
55 #include <sys/ucred.h>
58 #include <net/if_llatbl.h>
59 #include <net/netisr.h>
61 #include <net/route.h>
62 #include <net/flowtable.h>
64 #include <net/radix_mpath.h>
68 #include <netinet/in.h>
69 #include <netinet/in_kdtrace.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/ip.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/in_var.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_options.h>
77 #include <netinet/sctp.h>
78 #include <netinet/sctp_crc32.h>
82 #include <netinet/ip_ipsec.h>
83 #include <netipsec/ipsec.h>
86 #include <machine/in_cksum.h>
88 #include <security/mac/mac_framework.h>
90 VNET_DEFINE(u_short, ip_id);
92 #ifdef MBUF_STRESS_TEST
93 static int mbuf_frag_size = 0;
94 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
95 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
98 static void ip_mloopback
99 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
102 extern int in_mcast_loop;
103 extern struct protosw inetsw[];
106 * IP output. The packet in mbuf chain m contains a skeletal IP
107 * header (with len, off, ttl, proto, tos, src, dst).
108 * The mbuf chain containing the packet will be freed.
109 * The mbuf opt, if present, will not be freed.
110 * If route ro is present and has ro_rt initialized, route lookup would be
111 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
112 * then result of route lookup is stored in ro->ro_rt.
114 * In the IP forwarding case, the packet will arrive with options already
115 * inserted, so must have a NULL opt pointer.
118 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
119 struct ip_moptions *imo, struct inpcb *inp)
122 struct ifnet *ifp = NULL; /* keep compiler happy */
124 int hlen = sizeof (struct ip);
126 int n; /* scratchpad */
128 struct sockaddr_in *dst;
129 const struct sockaddr_in *gw;
130 struct in_ifaddr *ia;
132 uint16_t ip_len, ip_off;
133 struct route iproute;
134 struct rtentry *rte; /* cache for ro->ro_rt */
136 struct m_tag *fwd_tag = NULL;
138 int no_route_but_check_spd = 0;
143 INP_LOCK_ASSERT(inp);
144 M_SETFIB(m, inp->inp_inc.inc_fibnum);
145 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
146 m->m_pkthdr.flowid = inp->inp_flowid;
147 m->m_flags |= M_FLOWID;
153 bzero(ro, sizeof (*ro));
157 if (ro->ro_rt == NULL) {
161 * The flow table returns route entries valid for up to 30
162 * seconds; we rely on the remainder of ip_output() taking no
163 * longer than that long for the stability of ro_rt. The
164 * flow ID assignment must have happened before this point.
166 fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET);
168 flow_to_route(fle, ro);
174 m = ip_insertoptions(m, opt, &len);
176 hlen = len; /* ip->ip_hl is updated above */
178 ip = mtod(m, struct ip *);
179 ip_len = ntohs(ip->ip_len);
180 ip_off = ntohs(ip->ip_off);
183 * Fill in IP header. If we are not allowing fragmentation,
184 * then the ip_id field is meaningless, but we don't set it
185 * to zero. Doing so causes various problems when devices along
186 * the path (routers, load balancers, firewalls, etc.) illegally
187 * disable DF on our packet. Note that a 16-bit counter
188 * will wrap around in less than 10 seconds at 100 Mbit/s on a
189 * medium with MTU 1500. See Steven M. Bellovin, "A Technique
190 * for Counting NATted Hosts", Proc. IMW'02, available at
191 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
193 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
194 ip->ip_v = IPVERSION;
195 ip->ip_hl = hlen >> 2;
196 ip->ip_id = ip_newid();
197 IPSTAT_INC(ips_localout);
199 /* Header already set, fetch hlen from there */
200 hlen = ip->ip_hl << 2;
206 * dst can be rewritten but always point to &ro->ro_dst
207 * gw is readonly but can be pointed either to dst OR rt_gatewy
208 * therefore we need restore GW if we're re-doing lookup
210 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
214 * If there is a cached route,
215 * check that it is to the same destination
216 * and is still up. If not, free it and try again.
217 * The address family should also be checked in case of sharing the
221 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
222 rte->rt_ifp == NULL ||
223 !RT_LINK_IS_UP(rte->rt_ifp) ||
224 dst->sin_family != AF_INET ||
225 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
231 if (rte == NULL && fwd_tag == NULL) {
232 bzero(dst, sizeof(*dst));
233 dst->sin_family = AF_INET;
234 dst->sin_len = sizeof(*dst);
235 dst->sin_addr = ip->ip_dst;
238 * If routing to interface only, short circuit routing lookup.
239 * The use of an all-ones broadcast address implies this; an
240 * interface is specified by the broadcast address of an interface,
241 * or the destination address of a ptp interface.
243 if (flags & IP_SENDONES) {
244 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
245 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
246 IPSTAT_INC(ips_noroute);
250 ip->ip_dst.s_addr = INADDR_BROADCAST;
251 dst->sin_addr = ip->ip_dst;
255 } else if (flags & IP_ROUTETOIF) {
256 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
257 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) {
258 IPSTAT_INC(ips_noroute);
264 isbroadcast = in_broadcast(dst->sin_addr, ifp);
265 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
266 imo != NULL && imo->imo_multicast_ifp != NULL) {
268 * Bypass the normal routing lookup for multicast
269 * packets if the interface is specified.
271 ifp = imo->imo_multicast_ifp;
273 isbroadcast = 0; /* fool gcc */
276 * We want to do any cloning requested by the link layer,
277 * as this is probably required in all cases for correct
278 * operation (as it is for ARP).
282 rtalloc_mpath_fib(ro,
283 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
284 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
286 in_rtalloc_ign(ro, 0,
287 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
292 rte->rt_ifp == NULL ||
293 !RT_LINK_IS_UP(rte->rt_ifp)) {
296 * There is no route for this packet, but it is
297 * possible that a matching SPD entry exists.
299 no_route_but_check_spd = 1;
300 mtu = 0; /* Silence GCC warning. */
303 IPSTAT_INC(ips_noroute);
304 error = EHOSTUNREACH;
307 ia = ifatoia(rte->rt_ifa);
308 ifa_ref(&ia->ia_ifa);
310 rte->rt_rmx.rmx_pksent++;
311 if (rte->rt_flags & RTF_GATEWAY)
312 gw = (struct sockaddr_in *)rte->rt_gateway;
313 if (rte->rt_flags & RTF_HOST)
314 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
316 isbroadcast = in_broadcast(gw->sin_addr, ifp);
319 * Calculate MTU. If we have a route that is up, use that,
320 * otherwise use the interface's MTU.
322 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
324 * This case can happen if the user changed the MTU
325 * of an interface after enabling IP on it. Because
326 * most netifs don't keep track of routes pointing to
327 * them, there is no way for one to update all its
328 * routes when the MTU is changed.
330 if (rte->rt_rmx.rmx_mtu > ifp->if_mtu)
331 rte->rt_rmx.rmx_mtu = ifp->if_mtu;
332 mtu = rte->rt_rmx.rmx_mtu;
336 /* Catch a possible divide by zero later. */
337 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
338 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
339 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
340 m->m_flags |= M_MCAST;
342 * IP destination address is multicast. Make sure "gw"
343 * still points to the address in "ro". (It may have been
344 * changed to point to a gateway address, above.)
348 * See if the caller provided any multicast options
351 ip->ip_ttl = imo->imo_multicast_ttl;
352 if (imo->imo_multicast_vif != -1)
355 ip_mcast_src(imo->imo_multicast_vif) :
358 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
360 * Confirm that the outgoing interface supports multicast.
362 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
363 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
364 IPSTAT_INC(ips_noroute);
370 * If source address not specified yet, use address
371 * of outgoing interface.
373 if (ip->ip_src.s_addr == INADDR_ANY) {
374 /* Interface may have no addresses. */
376 ip->ip_src = IA_SIN(ia)->sin_addr;
379 if ((imo == NULL && in_mcast_loop) ||
380 (imo && imo->imo_multicast_loop)) {
382 * Loop back multicast datagram if not expressly
383 * forbidden to do so, even if we are not a member
384 * of the group; ip_input() will filter it later,
385 * thus deferring a hash lookup and mutex acquisition
386 * at the expense of a cheap copy using m_copym().
388 ip_mloopback(ifp, m, dst, hlen);
391 * If we are acting as a multicast router, perform
392 * multicast forwarding as if the packet had just
393 * arrived on the interface to which we are about
394 * to send. The multicast forwarding function
395 * recursively calls this function, using the
396 * IP_FORWARDING flag to prevent infinite recursion.
398 * Multicasts that are looped back by ip_mloopback(),
399 * above, will be forwarded by the ip_input() routine,
402 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
404 * If rsvp daemon is not running, do not
405 * set ip_moptions. This ensures that the packet
406 * is multicast and not just sent down one link
407 * as prescribed by rsvpd.
412 ip_mforward(ip, ifp, m, imo) != 0) {
420 * Multicasts with a time-to-live of zero may be looped-
421 * back, above, but must not be transmitted on a network.
422 * Also, multicasts addressed to the loopback interface
423 * are not sent -- the above call to ip_mloopback() will
424 * loop back a copy. ip_input() will drop the copy if
425 * this host does not belong to the destination group on
426 * the loopback interface.
428 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
437 * If the source address is not specified yet, use the address
438 * of the outoing interface.
440 if (ip->ip_src.s_addr == INADDR_ANY) {
441 /* Interface may have no addresses. */
443 ip->ip_src = IA_SIN(ia)->sin_addr;
448 * Verify that we have any chance at all of being able to queue the
449 * packet or packet fragments, unless ALTQ is enabled on the given
450 * interface in which case packetdrop should be done by queueing.
452 n = ip_len / mtu + 1; /* how many fragments ? */
455 (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
457 (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
459 IPSTAT_INC(ips_odropped);
460 ifp->if_snd.ifq_drops += n;
465 * Look for broadcast address and
466 * verify user is allowed to send
470 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
471 error = EADDRNOTAVAIL;
474 if ((flags & IP_ALLOWBROADCAST) == 0) {
478 /* don't allow broadcast messages to be fragmented */
483 m->m_flags |= M_BCAST;
485 m->m_flags &= ~M_BCAST;
490 switch(ip_ipsec_output(&m, inp, &flags, &error)) {
497 break; /* Continue with packet processing. */
500 * Check if there was a route for this packet; return error if not.
502 if (no_route_but_check_spd) {
503 IPSTAT_INC(ips_noroute);
504 error = EHOSTUNREACH;
507 /* Update variables that are affected by ipsec4_output(). */
508 ip = mtod(m, struct ip *);
509 hlen = ip->ip_hl << 2;
512 /* Jump over all PFIL processing if hooks are not active. */
513 if (!PFIL_HOOKED(&V_inet_pfil_hook))
516 /* Run through list of hooks for output packets. */
517 odst.s_addr = ip->ip_dst.s_addr;
518 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
519 if (error != 0 || m == NULL)
522 ip = mtod(m, struct ip *);
524 /* See if destination IP address was changed by packet filter. */
525 if (odst.s_addr != ip->ip_dst.s_addr) {
526 m->m_flags |= M_SKIP_FIREWALL;
527 /* If destination is now ourself drop to ip_input(). */
528 if (in_localip(ip->ip_dst)) {
529 m->m_flags |= M_FASTFWD_OURS;
530 if (m->m_pkthdr.rcvif == NULL)
531 m->m_pkthdr.rcvif = V_loif;
532 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
533 m->m_pkthdr.csum_flags |=
534 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
535 m->m_pkthdr.csum_data = 0xffff;
537 m->m_pkthdr.csum_flags |=
538 CSUM_IP_CHECKED | CSUM_IP_VALID;
540 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
541 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
543 error = netisr_queue(NETISR_IP, m);
547 ifa_free(&ia->ia_ifa);
548 goto again; /* Redo the routing table lookup. */
552 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
553 if (m->m_flags & M_FASTFWD_OURS) {
554 if (m->m_pkthdr.rcvif == NULL)
555 m->m_pkthdr.rcvif = V_loif;
556 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
557 m->m_pkthdr.csum_flags |=
558 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
559 m->m_pkthdr.csum_data = 0xffff;
562 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
563 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
565 m->m_pkthdr.csum_flags |=
566 CSUM_IP_CHECKED | CSUM_IP_VALID;
568 error = netisr_queue(NETISR_IP, m);
571 /* Or forward to some other address? */
572 if ((m->m_flags & M_IP_NEXTHOP) &&
573 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
574 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
575 m->m_flags |= M_SKIP_FIREWALL;
576 m->m_flags &= ~M_IP_NEXTHOP;
577 m_tag_delete(m, fwd_tag);
579 ifa_free(&ia->ia_ifa);
584 /* 127/8 must not appear on wire - RFC1122. */
585 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
586 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
587 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
588 IPSTAT_INC(ips_badaddr);
589 error = EADDRNOTAVAIL;
594 m->m_pkthdr.csum_flags |= CSUM_IP;
595 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
597 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
600 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
601 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
602 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
607 * If small enough for interface, or the interface will take
608 * care of the fragmentation for us, we can just send directly.
611 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
612 ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
614 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
615 ip->ip_sum = in_cksum(m, hlen);
616 m->m_pkthdr.csum_flags &= ~CSUM_IP;
620 * Record statistics for this interface address.
621 * With CSUM_TSO the byte/packet count will be slightly
622 * incorrect because we count the IP+TCP headers only
623 * once instead of for every generated packet.
625 if (!(flags & IP_FORWARDING) && ia) {
626 if (m->m_pkthdr.csum_flags & CSUM_TSO)
627 ia->ia_ifa.if_opackets +=
628 m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
630 ia->ia_ifa.if_opackets++;
631 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
633 #ifdef MBUF_STRESS_TEST
634 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
635 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
638 * Reset layer specific mbuf flags
639 * to avoid confusing lower layers.
642 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
643 error = (*ifp->if_output)(ifp, m,
644 (const struct sockaddr *)gw, ro);
648 /* Balk when DF bit is set or the interface didn't support TSO. */
649 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
651 IPSTAT_INC(ips_cantfrag);
656 * Too large for interface; fragment if possible. If successful,
657 * on return, m will point to a list of packets to be sent.
659 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
666 /* Record statistics for this interface address. */
668 ia->ia_ifa.if_opackets++;
669 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
672 * Reset layer specific mbuf flags
673 * to avoid confusing upper layers.
677 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
678 error = (*ifp->if_output)(ifp, m,
679 (const struct sockaddr *)gw, ro);
685 IPSTAT_INC(ips_fragmented);
691 ifa_free(&ia->ia_ifa);
699 * Create a chain of fragments which fit the given mtu. m_frag points to the
700 * mbuf to be fragmented; on return it points to the chain with the fragments.
701 * Return 0 if no error. If error, m_frag may contain a partially built
702 * chain of fragments that should be freed by the caller.
704 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
707 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
708 u_long if_hwassist_flags)
711 int hlen = ip->ip_hl << 2;
712 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
714 struct mbuf *m0 = *m_frag; /* the original packet */
718 uint16_t ip_len, ip_off;
720 ip_len = ntohs(ip->ip_len);
721 ip_off = ntohs(ip->ip_off);
723 if (ip_off & IP_DF) { /* Fragmentation not allowed */
724 IPSTAT_INC(ips_cantfrag);
729 * Must be able to put at least 8 bytes per fragment.
735 * If the interface will not calculate checksums on
736 * fragmented packets, then do it here.
738 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
739 in_delayed_cksum(m0);
740 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
743 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
744 sctp_delayed_cksum(m0, hlen);
745 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
748 if (len > PAGE_SIZE) {
750 * Fragment large datagrams such that each segment
751 * contains a multiple of PAGE_SIZE amount of data,
752 * plus headers. This enables a receiver to perform
753 * page-flipping zero-copy optimizations.
755 * XXX When does this help given that sender and receiver
756 * could have different page sizes, and also mtu could
757 * be less than the receiver's page size ?
762 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
766 * firstlen (off - hlen) must be aligned on an
770 goto smart_frag_failure;
771 off = ((off - hlen) & ~7) + hlen;
772 newlen = (~PAGE_MASK) & mtu;
773 if ((newlen + sizeof (struct ip)) > mtu) {
774 /* we failed, go back the default */
785 firstlen = off - hlen;
786 mnext = &m0->m_nextpkt; /* pointer to next packet */
789 * Loop through length of segment after first fragment,
790 * make new header and copy data of each part and link onto chain.
791 * Here, m0 is the original packet, m is the fragment being created.
792 * The fragments are linked off the m_nextpkt of the original
793 * packet, which after processing serves as the first fragment.
795 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
796 struct ip *mhip; /* ip header on the fragment */
798 int mhlen = sizeof (struct ip);
800 m = m_gethdr(M_NOWAIT, MT_DATA);
803 IPSTAT_INC(ips_odropped);
806 m->m_flags |= (m0->m_flags & M_MCAST);
808 * In the first mbuf, leave room for the link header, then
809 * copy the original IP header including options. The payload
810 * goes into an additional mbuf chain returned by m_copym().
812 m->m_data += max_linkhdr;
813 mhip = mtod(m, struct ip *);
815 if (hlen > sizeof (struct ip)) {
816 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
817 mhip->ip_v = IPVERSION;
818 mhip->ip_hl = mhlen >> 2;
821 /* XXX do we need to add ip_off below ? */
822 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
823 if (off + len >= ip_len)
826 mhip->ip_off |= IP_MF;
827 mhip->ip_len = htons((u_short)(len + mhlen));
828 m->m_next = m_copym(m0, off, len, M_NOWAIT);
829 if (m->m_next == NULL) { /* copy failed */
831 error = ENOBUFS; /* ??? */
832 IPSTAT_INC(ips_odropped);
835 m->m_pkthdr.len = mhlen + len;
836 m->m_pkthdr.rcvif = NULL;
838 mac_netinet_fragment(m0, m);
840 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
841 mhip->ip_off = htons(mhip->ip_off);
843 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
844 mhip->ip_sum = in_cksum(m, mhlen);
845 m->m_pkthdr.csum_flags &= ~CSUM_IP;
848 mnext = &m->m_nextpkt;
850 IPSTAT_ADD(ips_ofragments, nfrags);
853 * Update first fragment by trimming what's been copied out
854 * and updating header.
856 m_adj(m0, hlen + firstlen - ip_len);
857 m0->m_pkthdr.len = hlen + firstlen;
858 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
859 ip->ip_off = htons(ip_off | IP_MF);
861 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
862 ip->ip_sum = in_cksum(m0, hlen);
863 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
872 in_delayed_cksum(struct mbuf *m)
875 uint16_t csum, offset, ip_len;
877 ip = mtod(m, struct ip *);
878 offset = ip->ip_hl << 2 ;
879 ip_len = ntohs(ip->ip_len);
880 csum = in_cksum_skip(m, ip_len, offset);
881 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
883 offset += m->m_pkthdr.csum_data; /* checksum offset */
885 if (offset + sizeof(u_short) > m->m_len) {
886 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
887 m->m_len, offset, ip->ip_p);
890 * this shouldn't happen, but if it does, the
891 * correct behavior may be to insert the checksum
892 * in the appropriate next mbuf in the chain.
896 *(u_short *)(m->m_data + offset) = csum;
900 * IP socket option processing.
903 ip_ctloutput(struct socket *so, struct sockopt *sopt)
905 struct inpcb *inp = sotoinpcb(so);
909 if (sopt->sopt_level != IPPROTO_IP) {
912 if (sopt->sopt_level == SOL_SOCKET &&
913 sopt->sopt_dir == SOPT_SET) {
914 switch (sopt->sopt_name) {
917 if ((so->so_options & SO_REUSEADDR) != 0)
918 inp->inp_flags2 |= INP_REUSEADDR;
920 inp->inp_flags2 &= ~INP_REUSEADDR;
926 if ((so->so_options & SO_REUSEPORT) != 0)
927 inp->inp_flags2 |= INP_REUSEPORT;
929 inp->inp_flags2 &= ~INP_REUSEPORT;
935 inp->inp_inc.inc_fibnum = so->so_fibnum;
946 switch (sopt->sopt_dir) {
948 switch (sopt->sopt_name) {
955 if (sopt->sopt_valsize > MLEN) {
959 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
964 m->m_len = sopt->sopt_valsize;
965 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
972 error = ip_pcbopts(inp, sopt->sopt_name, m);
978 if (sopt->sopt_td != NULL) {
979 error = priv_check(sopt->sopt_td,
980 PRIV_NETINET_BINDANY);
997 error = sooptcopyin(sopt, &optval, sizeof optval,
1002 switch (sopt->sopt_name) {
1004 inp->inp_ip_tos = optval;
1008 inp->inp_ip_ttl = optval;
1012 if (optval >= 0 && optval <= MAXTTL)
1013 inp->inp_ip_minttl = optval;
1018 #define OPTSET(bit) do { \
1021 inp->inp_flags |= bit; \
1023 inp->inp_flags &= ~bit; \
1028 OPTSET(INP_RECVOPTS);
1031 case IP_RECVRETOPTS:
1032 OPTSET(INP_RECVRETOPTS);
1035 case IP_RECVDSTADDR:
1036 OPTSET(INP_RECVDSTADDR);
1040 OPTSET(INP_RECVTTL);
1052 OPTSET(INP_ONESBCAST);
1055 OPTSET(INP_DONTFRAG);
1058 OPTSET(INP_BINDANY);
1061 OPTSET(INP_RECVTOS);
1068 * Multicast socket options are processed by the in_mcast
1071 case IP_MULTICAST_IF:
1072 case IP_MULTICAST_VIF:
1073 case IP_MULTICAST_TTL:
1074 case IP_MULTICAST_LOOP:
1075 case IP_ADD_MEMBERSHIP:
1076 case IP_DROP_MEMBERSHIP:
1077 case IP_ADD_SOURCE_MEMBERSHIP:
1078 case IP_DROP_SOURCE_MEMBERSHIP:
1079 case IP_BLOCK_SOURCE:
1080 case IP_UNBLOCK_SOURCE:
1082 case MCAST_JOIN_GROUP:
1083 case MCAST_LEAVE_GROUP:
1084 case MCAST_JOIN_SOURCE_GROUP:
1085 case MCAST_LEAVE_SOURCE_GROUP:
1086 case MCAST_BLOCK_SOURCE:
1087 case MCAST_UNBLOCK_SOURCE:
1088 error = inp_setmoptions(inp, sopt);
1092 error = sooptcopyin(sopt, &optval, sizeof optval,
1099 case IP_PORTRANGE_DEFAULT:
1100 inp->inp_flags &= ~(INP_LOWPORT);
1101 inp->inp_flags &= ~(INP_HIGHPORT);
1104 case IP_PORTRANGE_HIGH:
1105 inp->inp_flags &= ~(INP_LOWPORT);
1106 inp->inp_flags |= INP_HIGHPORT;
1109 case IP_PORTRANGE_LOW:
1110 inp->inp_flags &= ~(INP_HIGHPORT);
1111 inp->inp_flags |= INP_LOWPORT;
1122 case IP_IPSEC_POLICY:
1127 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1129 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1131 req = mtod(m, caddr_t);
1132 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1133 m->m_len, (sopt->sopt_td != NULL) ?
1134 sopt->sopt_td->td_ucred : NULL);
1141 error = ENOPROTOOPT;
1147 switch (sopt->sopt_name) {
1150 if (inp->inp_options)
1151 error = sooptcopyout(sopt,
1152 mtod(inp->inp_options,
1154 inp->inp_options->m_len);
1156 sopt->sopt_valsize = 0;
1163 case IP_RECVRETOPTS:
1164 case IP_RECVDSTADDR:
1173 switch (sopt->sopt_name) {
1176 optval = inp->inp_ip_tos;
1180 optval = inp->inp_ip_ttl;
1184 optval = inp->inp_ip_minttl;
1187 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1190 optval = OPTBIT(INP_RECVOPTS);
1193 case IP_RECVRETOPTS:
1194 optval = OPTBIT(INP_RECVRETOPTS);
1197 case IP_RECVDSTADDR:
1198 optval = OPTBIT(INP_RECVDSTADDR);
1202 optval = OPTBIT(INP_RECVTTL);
1206 optval = OPTBIT(INP_RECVIF);
1210 if (inp->inp_flags & INP_HIGHPORT)
1211 optval = IP_PORTRANGE_HIGH;
1212 else if (inp->inp_flags & INP_LOWPORT)
1213 optval = IP_PORTRANGE_LOW;
1219 optval = OPTBIT(INP_FAITH);
1223 optval = OPTBIT(INP_ONESBCAST);
1226 optval = OPTBIT(INP_DONTFRAG);
1229 optval = OPTBIT(INP_BINDANY);
1232 optval = OPTBIT(INP_RECVTOS);
1235 error = sooptcopyout(sopt, &optval, sizeof optval);
1239 * Multicast socket options are processed by the in_mcast
1242 case IP_MULTICAST_IF:
1243 case IP_MULTICAST_VIF:
1244 case IP_MULTICAST_TTL:
1245 case IP_MULTICAST_LOOP:
1247 error = inp_getmoptions(inp, sopt);
1251 case IP_IPSEC_POLICY:
1253 struct mbuf *m = NULL;
1258 req = mtod(m, caddr_t);
1261 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1263 error = soopt_mcopyout(sopt, m); /* XXX */
1271 error = ENOPROTOOPT;
1280 * Routine called from ip_output() to loop back a copy of an IP multicast
1281 * packet to the input queue of a specified interface. Note that this
1282 * calls the output routine of the loopback "driver", but with an interface
1283 * pointer that might NOT be a loopback interface -- evil, but easier than
1284 * replicating that code here.
1287 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1290 register struct ip *ip;
1294 * Make a deep copy of the packet because we're going to
1295 * modify the pack in order to generate checksums.
1297 copym = m_dup(m, M_NOWAIT);
1298 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1299 copym = m_pullup(copym, hlen);
1300 if (copym != NULL) {
1301 /* If needed, compute the checksum and mark it as valid. */
1302 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1303 in_delayed_cksum(copym);
1304 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1305 copym->m_pkthdr.csum_flags |=
1306 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1307 copym->m_pkthdr.csum_data = 0xffff;
1310 * We don't bother to fragment if the IP length is greater
1311 * than the interface's MTU. Can this possibly matter?
1313 ip = mtod(copym, struct ip *);
1315 ip->ip_sum = in_cksum(copym, hlen);
1317 if (dst->sin_family != AF_INET) {
1318 printf("ip_mloopback: bad address family %d\n",
1320 dst->sin_family = AF_INET;
1323 if_simloop(ifp, copym, dst->sin_family, 0);