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_route.h"
38 #include "opt_mbuf_stress_test.h"
39 #include "opt_mpath.h"
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/malloc.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sysctl.h>
53 #include <sys/ucred.h>
56 #include <net/if_llatbl.h>
57 #include <net/netisr.h>
59 #include <net/route.h>
60 #include <net/flowtable.h>
62 #include <net/radix_mpath.h>
66 #include <netinet/in.h>
67 #include <netinet/in_systm.h>
68 #include <netinet/ip.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/in_var.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/ip_options.h>
74 #include <netinet/sctp.h>
75 #include <netinet/sctp_crc32.h>
79 #include <netinet/ip_ipsec.h>
80 #include <netipsec/ipsec.h>
83 #include <machine/in_cksum.h>
85 #include <security/mac/mac_framework.h>
87 #define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\
88 x, (ntohl(a.s_addr)>>24)&0xFF,\
89 (ntohl(a.s_addr)>>16)&0xFF,\
90 (ntohl(a.s_addr)>>8)&0xFF,\
91 (ntohl(a.s_addr))&0xFF, y);
93 VNET_DEFINE(u_short, ip_id);
95 #ifdef MBUF_STRESS_TEST
96 int mbuf_frag_size = 0;
97 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
98 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
101 static void ip_mloopback
102 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
105 extern int in_mcast_loop;
106 extern struct protosw inetsw[];
109 * IP output. The packet in mbuf chain m contains a skeletal IP
110 * header (with len, off, ttl, proto, tos, src, dst).
111 * The mbuf chain containing the packet will be freed.
112 * The mbuf opt, if present, will not be freed.
113 * In the IP forwarding case, the packet will arrive with options already
114 * inserted, so must have a NULL opt pointer.
117 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
118 struct ip_moptions *imo, struct inpcb *inp)
121 struct ifnet *ifp = NULL; /* keep compiler happy */
123 int hlen = sizeof (struct ip);
127 struct sockaddr_in *dst = NULL; /* keep compiler happy */
128 struct in_ifaddr *ia = NULL;
129 int isbroadcast, sw_csum;
130 struct route iproute;
132 #ifdef IPFIREWALL_FORWARD
133 struct m_tag *fwd_tag = NULL;
136 int no_route_but_check_spd = 0;
141 INP_LOCK_ASSERT(inp);
142 M_SETFIB(m, inp->inp_inc.inc_fibnum);
143 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
144 m->m_pkthdr.flowid = inp->inp_flowid;
145 m->m_flags |= M_FLOWID;
151 bzero(ro, sizeof (*ro));
155 * The flow table returns route entries valid for up to 30
156 * seconds; we rely on the remainder of ip_output() taking no
157 * longer than that long for the stability of ro_rt. The
158 * flow ID assignment must have happened before this point.
160 if (flowtable_lookup(V_ip_ft, m, ro, M_GETFIB(m)) == 0)
167 m = ip_insertoptions(m, opt, &len);
171 ip = mtod(m, struct ip *);
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 hlen = ip->ip_hl << 2;
193 dst = (struct sockaddr_in *)&ro->ro_dst;
196 * If there is a cached route,
197 * check that it is to the same destination
198 * and is still up. If not, free it and try again.
199 * The address family should also be checked in case of sharing the
202 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
203 dst->sin_family != AF_INET ||
204 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
207 ro->ro_rt = (struct rtentry *)NULL;
208 ro->ro_lle = (struct llentry *)NULL;
210 #ifdef IPFIREWALL_FORWARD
211 if (ro->ro_rt == NULL && fwd_tag == NULL) {
213 if (ro->ro_rt == NULL) {
215 bzero(dst, sizeof(*dst));
216 dst->sin_family = AF_INET;
217 dst->sin_len = sizeof(*dst);
218 dst->sin_addr = ip->ip_dst;
221 * If routing to interface only, short circuit routing lookup.
222 * The use of an all-ones broadcast address implies this; an
223 * interface is specified by the broadcast address of an interface,
224 * or the destination address of a ptp interface.
226 if (flags & IP_SENDONES) {
227 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
228 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
229 IPSTAT_INC(ips_noroute);
233 ip->ip_dst.s_addr = INADDR_BROADCAST;
234 dst->sin_addr = ip->ip_dst;
238 } else if (flags & IP_ROUTETOIF) {
239 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
240 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) {
241 IPSTAT_INC(ips_noroute);
247 isbroadcast = in_broadcast(dst->sin_addr, ifp);
248 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
249 imo != NULL && imo->imo_multicast_ifp != NULL) {
251 * Bypass the normal routing lookup for multicast
252 * packets if the interface is specified.
254 ifp = imo->imo_multicast_ifp;
256 isbroadcast = 0; /* fool gcc */
259 * We want to do any cloning requested by the link layer,
260 * as this is probably required in all cases for correct
261 * operation (as it is for ARP).
263 if (ro->ro_rt == NULL)
265 rtalloc_mpath_fib(ro,
266 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
267 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
269 in_rtalloc_ign(ro, 0,
270 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
272 if (ro->ro_rt == NULL) {
275 * There is no route for this packet, but it is
276 * possible that a matching SPD entry exists.
278 no_route_but_check_spd = 1;
279 mtu = 0; /* Silence GCC warning. */
282 IPSTAT_INC(ips_noroute);
283 error = EHOSTUNREACH;
286 ia = ifatoia(ro->ro_rt->rt_ifa);
287 ifa_ref(&ia->ia_ifa);
288 ifp = ro->ro_rt->rt_ifp;
289 ro->ro_rt->rt_rmx.rmx_pksent++;
290 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
291 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
292 if (ro->ro_rt->rt_flags & RTF_HOST)
293 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
295 isbroadcast = in_broadcast(dst->sin_addr, ifp);
298 * Calculate MTU. If we have a route that is up, use that,
299 * otherwise use the interface's MTU.
301 if (ro->ro_rt != NULL && (ro->ro_rt->rt_flags & (RTF_UP|RTF_HOST))) {
303 * This case can happen if the user changed the MTU
304 * of an interface after enabling IP on it. Because
305 * most netifs don't keep track of routes pointing to
306 * them, there is no way for one to update all its
307 * routes when the MTU is changed.
309 if (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)
310 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
311 mtu = ro->ro_rt->rt_rmx.rmx_mtu;
315 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
316 m->m_flags |= M_MCAST;
318 * IP destination address is multicast. Make sure "dst"
319 * still points to the address in "ro". (It may have been
320 * changed to point to a gateway address, above.)
322 dst = (struct sockaddr_in *)&ro->ro_dst;
324 * See if the caller provided any multicast options
327 ip->ip_ttl = imo->imo_multicast_ttl;
328 if (imo->imo_multicast_vif != -1)
331 ip_mcast_src(imo->imo_multicast_vif) :
334 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
336 * Confirm that the outgoing interface supports multicast.
338 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
339 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
340 IPSTAT_INC(ips_noroute);
346 * If source address not specified yet, use address
347 * of outgoing interface.
349 if (ip->ip_src.s_addr == INADDR_ANY) {
350 /* Interface may have no addresses. */
352 ip->ip_src = IA_SIN(ia)->sin_addr;
355 if ((imo == NULL && in_mcast_loop) ||
356 (imo && imo->imo_multicast_loop)) {
358 * Loop back multicast datagram if not expressly
359 * forbidden to do so, even if we are not a member
360 * of the group; ip_input() will filter it later,
361 * thus deferring a hash lookup and mutex acquisition
362 * at the expense of a cheap copy using m_copym().
364 ip_mloopback(ifp, m, dst, hlen);
367 * If we are acting as a multicast router, perform
368 * multicast forwarding as if the packet had just
369 * arrived on the interface to which we are about
370 * to send. The multicast forwarding function
371 * recursively calls this function, using the
372 * IP_FORWARDING flag to prevent infinite recursion.
374 * Multicasts that are looped back by ip_mloopback(),
375 * above, will be forwarded by the ip_input() routine,
378 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
380 * If rsvp daemon is not running, do not
381 * set ip_moptions. This ensures that the packet
382 * is multicast and not just sent down one link
383 * as prescribed by rsvpd.
388 ip_mforward(ip, ifp, m, imo) != 0) {
396 * Multicasts with a time-to-live of zero may be looped-
397 * back, above, but must not be transmitted on a network.
398 * Also, multicasts addressed to the loopback interface
399 * are not sent -- the above call to ip_mloopback() will
400 * loop back a copy. ip_input() will drop the copy if
401 * this host does not belong to the destination group on
402 * the loopback interface.
404 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
413 * If the source address is not specified yet, use the address
414 * of the outoing interface.
416 if (ip->ip_src.s_addr == INADDR_ANY) {
417 /* Interface may have no addresses. */
419 ip->ip_src = IA_SIN(ia)->sin_addr;
424 * Verify that we have any chance at all of being able to queue the
425 * packet or packet fragments, unless ALTQ is enabled on the given
426 * interface in which case packetdrop should be done by queueing.
429 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
430 ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >=
431 ifp->if_snd.ifq_maxlen))
433 if ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >=
434 ifp->if_snd.ifq_maxlen)
438 IPSTAT_INC(ips_odropped);
439 ifp->if_snd.ifq_drops += (ip->ip_len / ifp->if_mtu + 1);
444 * Look for broadcast address and
445 * verify user is allowed to send
449 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
450 error = EADDRNOTAVAIL;
453 if ((flags & IP_ALLOWBROADCAST) == 0) {
457 /* don't allow broadcast messages to be fragmented */
458 if (ip->ip_len > mtu) {
462 m->m_flags |= M_BCAST;
464 m->m_flags &= ~M_BCAST;
469 switch(ip_ipsec_output(&m, inp, &flags, &error, &ifp)) {
476 break; /* Continue with packet processing. */
479 * Check if there was a route for this packet; return error if not.
481 if (no_route_but_check_spd) {
482 IPSTAT_INC(ips_noroute);
483 error = EHOSTUNREACH;
486 /* Update variables that are affected by ipsec4_output(). */
487 ip = mtod(m, struct ip *);
488 hlen = ip->ip_hl << 2;
491 /* Jump over all PFIL processing if hooks are not active. */
492 if (!PFIL_HOOKED(&inet_pfil_hook))
495 /* Run through list of hooks for output packets. */
496 odst.s_addr = ip->ip_dst.s_addr;
497 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
498 if (error != 0 || m == NULL)
501 ip = mtod(m, struct ip *);
503 /* See if destination IP address was changed by packet filter. */
504 if (odst.s_addr != ip->ip_dst.s_addr) {
505 m->m_flags |= M_SKIP_FIREWALL;
506 /* If destination is now ourself drop to ip_input(). */
507 if (in_localip(ip->ip_dst)) {
508 m->m_flags |= M_FASTFWD_OURS;
509 if (m->m_pkthdr.rcvif == NULL)
510 m->m_pkthdr.rcvif = V_loif;
511 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
512 m->m_pkthdr.csum_flags |=
513 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
514 m->m_pkthdr.csum_data = 0xffff;
516 m->m_pkthdr.csum_flags |=
517 CSUM_IP_CHECKED | CSUM_IP_VALID;
519 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
520 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
522 error = netisr_queue(NETISR_IP, m);
525 goto again; /* Redo the routing table lookup. */
528 #ifdef IPFIREWALL_FORWARD
529 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
530 if (m->m_flags & M_FASTFWD_OURS) {
531 if (m->m_pkthdr.rcvif == NULL)
532 m->m_pkthdr.rcvif = V_loif;
533 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
534 m->m_pkthdr.csum_flags |=
535 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
536 m->m_pkthdr.csum_data = 0xffff;
539 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
540 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
542 m->m_pkthdr.csum_flags |=
543 CSUM_IP_CHECKED | CSUM_IP_VALID;
545 error = netisr_queue(NETISR_IP, m);
548 /* Or forward to some other address? */
549 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
551 dst = (struct sockaddr_in *)&ro->ro_dst;
552 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
553 m->m_flags |= M_SKIP_FIREWALL;
554 m_tag_delete(m, fwd_tag);
557 #endif /* IPFIREWALL_FORWARD */
560 /* 127/8 must not appear on wire - RFC1122. */
561 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
562 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
563 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
564 IPSTAT_INC(ips_badaddr);
565 error = EADDRNOTAVAIL;
570 m->m_pkthdr.csum_flags |= CSUM_IP;
571 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
572 if (sw_csum & CSUM_DELAY_DATA) {
574 sw_csum &= ~CSUM_DELAY_DATA;
577 if (sw_csum & CSUM_SCTP) {
578 sctp_delayed_cksum(m);
579 sw_csum &= ~CSUM_SCTP;
582 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
585 * If small enough for interface, or the interface will take
586 * care of the fragmentation for us, we can just send directly.
588 if (ip->ip_len <= mtu ||
589 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
590 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
591 ip->ip_len = htons(ip->ip_len);
592 ip->ip_off = htons(ip->ip_off);
594 if (sw_csum & CSUM_DELAY_IP)
595 ip->ip_sum = in_cksum(m, hlen);
598 * Record statistics for this interface address.
599 * With CSUM_TSO the byte/packet count will be slightly
600 * incorrect because we count the IP+TCP headers only
601 * once instead of for every generated packet.
603 if (!(flags & IP_FORWARDING) && ia) {
604 if (m->m_pkthdr.csum_flags & CSUM_TSO)
605 ia->ia_ifa.if_opackets +=
606 m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
608 ia->ia_ifa.if_opackets++;
609 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
611 #ifdef MBUF_STRESS_TEST
612 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
613 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size);
616 * Reset layer specific mbuf flags
617 * to avoid confusing lower layers.
619 m->m_flags &= ~(M_PROTOFLAGS);
620 error = (*ifp->if_output)(ifp, m,
621 (struct sockaddr *)dst, ro);
625 /* Balk when DF bit is set or the interface didn't support TSO. */
626 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
628 IPSTAT_INC(ips_cantfrag);
633 * Too large for interface; fragment if possible. If successful,
634 * on return, m will point to a list of packets to be sent.
636 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum);
643 /* Record statistics for this interface address. */
645 ia->ia_ifa.if_opackets++;
646 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
649 * Reset layer specific mbuf flags
650 * to avoid confusing upper layers.
652 m->m_flags &= ~(M_PROTOFLAGS);
654 error = (*ifp->if_output)(ifp, m,
655 (struct sockaddr *)dst, ro);
661 IPSTAT_INC(ips_fragmented);
664 if (ro == &iproute && ro->ro_rt && !nortfree) {
668 ifa_free(&ia->ia_ifa);
676 * Create a chain of fragments which fit the given mtu. m_frag points to the
677 * mbuf to be fragmented; on return it points to the chain with the fragments.
678 * Return 0 if no error. If error, m_frag may contain a partially built
679 * chain of fragments that should be freed by the caller.
681 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
682 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
685 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
686 u_long if_hwassist_flags, int sw_csum)
689 int hlen = ip->ip_hl << 2;
690 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
692 struct mbuf *m0 = *m_frag; /* the original packet */
697 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
698 IPSTAT_INC(ips_cantfrag);
703 * Must be able to put at least 8 bytes per fragment.
709 * If the interface will not calculate checksums on
710 * fragmented packets, then do it here.
712 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
713 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
714 in_delayed_cksum(m0);
715 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
718 if (m0->m_pkthdr.csum_flags & CSUM_SCTP &&
719 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
720 sctp_delayed_cksum(m0);
721 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
724 if (len > PAGE_SIZE) {
726 * Fragment large datagrams such that each segment
727 * contains a multiple of PAGE_SIZE amount of data,
728 * plus headers. This enables a receiver to perform
729 * page-flipping zero-copy optimizations.
731 * XXX When does this help given that sender and receiver
732 * could have different page sizes, and also mtu could
733 * be less than the receiver's page size ?
738 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
742 * firstlen (off - hlen) must be aligned on an
746 goto smart_frag_failure;
747 off = ((off - hlen) & ~7) + hlen;
748 newlen = (~PAGE_MASK) & mtu;
749 if ((newlen + sizeof (struct ip)) > mtu) {
750 /* we failed, go back the default */
761 firstlen = off - hlen;
762 mnext = &m0->m_nextpkt; /* pointer to next packet */
765 * Loop through length of segment after first fragment,
766 * make new header and copy data of each part and link onto chain.
767 * Here, m0 is the original packet, m is the fragment being created.
768 * The fragments are linked off the m_nextpkt of the original
769 * packet, which after processing serves as the first fragment.
771 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
772 struct ip *mhip; /* ip header on the fragment */
774 int mhlen = sizeof (struct ip);
776 MGETHDR(m, M_DONTWAIT, MT_DATA);
779 IPSTAT_INC(ips_odropped);
782 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
784 * In the first mbuf, leave room for the link header, then
785 * copy the original IP header including options. The payload
786 * goes into an additional mbuf chain returned by m_copym().
788 m->m_data += max_linkhdr;
789 mhip = mtod(m, struct ip *);
791 if (hlen > sizeof (struct ip)) {
792 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
793 mhip->ip_v = IPVERSION;
794 mhip->ip_hl = mhlen >> 2;
797 /* XXX do we need to add ip->ip_off below ? */
798 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
799 if (off + len >= ip->ip_len) { /* last fragment */
800 len = ip->ip_len - off;
801 m->m_flags |= M_LASTFRAG;
803 mhip->ip_off |= IP_MF;
804 mhip->ip_len = htons((u_short)(len + mhlen));
805 m->m_next = m_copym(m0, off, len, M_DONTWAIT);
806 if (m->m_next == NULL) { /* copy failed */
808 error = ENOBUFS; /* ??? */
809 IPSTAT_INC(ips_odropped);
812 m->m_pkthdr.len = mhlen + len;
813 m->m_pkthdr.rcvif = NULL;
815 mac_netinet_fragment(m0, m);
817 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
818 mhip->ip_off = htons(mhip->ip_off);
820 if (sw_csum & CSUM_DELAY_IP)
821 mhip->ip_sum = in_cksum(m, mhlen);
823 mnext = &m->m_nextpkt;
825 IPSTAT_ADD(ips_ofragments, nfrags);
827 /* set first marker for fragment chain */
828 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
829 m0->m_pkthdr.csum_data = nfrags;
832 * Update first fragment by trimming what's been copied out
833 * and updating header.
835 m_adj(m0, hlen + firstlen - ip->ip_len);
836 m0->m_pkthdr.len = hlen + firstlen;
837 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
839 ip->ip_off = htons(ip->ip_off);
841 if (sw_csum & CSUM_DELAY_IP)
842 ip->ip_sum = in_cksum(m0, hlen);
850 in_delayed_cksum(struct mbuf *m)
853 u_short csum, offset;
855 ip = mtod(m, struct ip *);
856 offset = ip->ip_hl << 2 ;
857 csum = in_cksum_skip(m, ip->ip_len, offset);
858 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
860 offset += m->m_pkthdr.csum_data; /* checksum offset */
862 if (offset + sizeof(u_short) > m->m_len) {
863 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
864 m->m_len, offset, ip->ip_p);
867 * this shouldn't happen, but if it does, the
868 * correct behavior may be to insert the checksum
869 * in the appropriate next mbuf in the chain.
873 *(u_short *)(m->m_data + offset) = csum;
877 * IP socket option processing.
880 ip_ctloutput(struct socket *so, struct sockopt *sopt)
882 struct inpcb *inp = sotoinpcb(so);
886 if (sopt->sopt_level != IPPROTO_IP) {
887 if ((sopt->sopt_level == SOL_SOCKET) &&
888 (sopt->sopt_name == SO_SETFIB)) {
889 inp->inp_inc.inc_fibnum = so->so_fibnum;
895 switch (sopt->sopt_dir) {
897 switch (sopt->sopt_name) {
904 if (sopt->sopt_valsize > MLEN) {
908 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
913 m->m_len = sopt->sopt_valsize;
914 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
921 error = ip_pcbopts(inp, sopt->sopt_name, m);
927 if (sopt->sopt_td != NULL) {
928 error = priv_check(sopt->sopt_td,
929 PRIV_NETINET_BINDANY);
945 error = sooptcopyin(sopt, &optval, sizeof optval,
950 switch (sopt->sopt_name) {
952 inp->inp_ip_tos = optval;
956 inp->inp_ip_ttl = optval;
960 if (optval >= 0 && optval <= MAXTTL)
961 inp->inp_ip_minttl = optval;
966 #define OPTSET(bit) do { \
969 inp->inp_flags |= bit; \
971 inp->inp_flags &= ~bit; \
976 OPTSET(INP_RECVOPTS);
980 OPTSET(INP_RECVRETOPTS);
984 OPTSET(INP_RECVDSTADDR);
1000 OPTSET(INP_ONESBCAST);
1003 OPTSET(INP_DONTFRAG);
1006 OPTSET(INP_BINDANY);
1013 * Multicast socket options are processed by the in_mcast
1016 case IP_MULTICAST_IF:
1017 case IP_MULTICAST_VIF:
1018 case IP_MULTICAST_TTL:
1019 case IP_MULTICAST_LOOP:
1020 case IP_ADD_MEMBERSHIP:
1021 case IP_DROP_MEMBERSHIP:
1022 case IP_ADD_SOURCE_MEMBERSHIP:
1023 case IP_DROP_SOURCE_MEMBERSHIP:
1024 case IP_BLOCK_SOURCE:
1025 case IP_UNBLOCK_SOURCE:
1027 case MCAST_JOIN_GROUP:
1028 case MCAST_LEAVE_GROUP:
1029 case MCAST_JOIN_SOURCE_GROUP:
1030 case MCAST_LEAVE_SOURCE_GROUP:
1031 case MCAST_BLOCK_SOURCE:
1032 case MCAST_UNBLOCK_SOURCE:
1033 error = inp_setmoptions(inp, sopt);
1037 error = sooptcopyin(sopt, &optval, sizeof optval,
1044 case IP_PORTRANGE_DEFAULT:
1045 inp->inp_flags &= ~(INP_LOWPORT);
1046 inp->inp_flags &= ~(INP_HIGHPORT);
1049 case IP_PORTRANGE_HIGH:
1050 inp->inp_flags &= ~(INP_LOWPORT);
1051 inp->inp_flags |= INP_HIGHPORT;
1054 case IP_PORTRANGE_LOW:
1055 inp->inp_flags &= ~(INP_HIGHPORT);
1056 inp->inp_flags |= INP_LOWPORT;
1067 case IP_IPSEC_POLICY:
1072 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1074 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1076 req = mtod(m, caddr_t);
1077 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1078 m->m_len, (sopt->sopt_td != NULL) ?
1079 sopt->sopt_td->td_ucred : NULL);
1086 error = ENOPROTOOPT;
1092 switch (sopt->sopt_name) {
1095 if (inp->inp_options)
1096 error = sooptcopyout(sopt,
1097 mtod(inp->inp_options,
1099 inp->inp_options->m_len);
1101 sopt->sopt_valsize = 0;
1108 case IP_RECVRETOPTS:
1109 case IP_RECVDSTADDR:
1116 switch (sopt->sopt_name) {
1119 optval = inp->inp_ip_tos;
1123 optval = inp->inp_ip_ttl;
1127 optval = inp->inp_ip_minttl;
1130 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1133 optval = OPTBIT(INP_RECVOPTS);
1136 case IP_RECVRETOPTS:
1137 optval = OPTBIT(INP_RECVRETOPTS);
1140 case IP_RECVDSTADDR:
1141 optval = OPTBIT(INP_RECVDSTADDR);
1145 optval = OPTBIT(INP_RECVTTL);
1149 optval = OPTBIT(INP_RECVIF);
1153 if (inp->inp_flags & INP_HIGHPORT)
1154 optval = IP_PORTRANGE_HIGH;
1155 else if (inp->inp_flags & INP_LOWPORT)
1156 optval = IP_PORTRANGE_LOW;
1162 optval = OPTBIT(INP_FAITH);
1166 optval = OPTBIT(INP_ONESBCAST);
1169 optval = OPTBIT(INP_DONTFRAG);
1172 error = sooptcopyout(sopt, &optval, sizeof optval);
1176 * Multicast socket options are processed by the in_mcast
1179 case IP_MULTICAST_IF:
1180 case IP_MULTICAST_VIF:
1181 case IP_MULTICAST_TTL:
1182 case IP_MULTICAST_LOOP:
1184 error = inp_getmoptions(inp, sopt);
1188 case IP_IPSEC_POLICY:
1190 struct mbuf *m = NULL;
1195 req = mtod(m, caddr_t);
1198 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1200 error = soopt_mcopyout(sopt, m); /* XXX */
1208 error = ENOPROTOOPT;
1217 * Routine called from ip_output() to loop back a copy of an IP multicast
1218 * packet to the input queue of a specified interface. Note that this
1219 * calls the output routine of the loopback "driver", but with an interface
1220 * pointer that might NOT be a loopback interface -- evil, but easier than
1221 * replicating that code here.
1224 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1227 register struct ip *ip;
1231 * Make a deep copy of the packet because we're going to
1232 * modify the pack in order to generate checksums.
1234 copym = m_dup(m, M_DONTWAIT);
1235 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1236 copym = m_pullup(copym, hlen);
1237 if (copym != NULL) {
1238 /* If needed, compute the checksum and mark it as valid. */
1239 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1240 in_delayed_cksum(copym);
1241 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1242 copym->m_pkthdr.csum_flags |=
1243 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1244 copym->m_pkthdr.csum_data = 0xffff;
1247 * We don't bother to fragment if the IP length is greater
1248 * than the interface's MTU. Can this possibly matter?
1250 ip = mtod(copym, struct ip *);
1251 ip->ip_len = htons(ip->ip_len);
1252 ip->ip_off = htons(ip->ip_off);
1254 ip->ip_sum = in_cksum(copym, hlen);
1256 if (dst->sin_family != AF_INET) {
1257 printf("ip_mloopback: bad address family %d\n",
1259 dst->sin_family = AF_INET;
1262 if_simloop(ifp, copym, dst->sin_family, 0);