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 LLE_FREE(ro->ro_lle);
209 ro->ro_rt = (struct rtentry *)NULL;
210 ro->ro_lle = (struct llentry *)NULL;
212 #ifdef IPFIREWALL_FORWARD
213 if (ro->ro_rt == NULL && fwd_tag == NULL) {
215 if (ro->ro_rt == NULL) {
217 bzero(dst, sizeof(*dst));
218 dst->sin_family = AF_INET;
219 dst->sin_len = sizeof(*dst);
220 dst->sin_addr = ip->ip_dst;
223 * If routing to interface only, short circuit routing lookup.
224 * The use of an all-ones broadcast address implies this; an
225 * interface is specified by the broadcast address of an interface,
226 * or the destination address of a ptp interface.
228 if (flags & IP_SENDONES) {
229 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
230 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
231 IPSTAT_INC(ips_noroute);
235 ip->ip_dst.s_addr = INADDR_BROADCAST;
236 dst->sin_addr = ip->ip_dst;
240 } else if (flags & IP_ROUTETOIF) {
241 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
242 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) {
243 IPSTAT_INC(ips_noroute);
249 isbroadcast = in_broadcast(dst->sin_addr, ifp);
250 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
251 imo != NULL && imo->imo_multicast_ifp != NULL) {
253 * Bypass the normal routing lookup for multicast
254 * packets if the interface is specified.
256 ifp = imo->imo_multicast_ifp;
258 isbroadcast = 0; /* fool gcc */
261 * We want to do any cloning requested by the link layer,
262 * as this is probably required in all cases for correct
263 * operation (as it is for ARP).
265 if (ro->ro_rt == NULL)
267 rtalloc_mpath_fib(ro,
268 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
269 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
271 in_rtalloc_ign(ro, 0,
272 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
274 if (ro->ro_rt == NULL) {
277 * There is no route for this packet, but it is
278 * possible that a matching SPD entry exists.
280 no_route_but_check_spd = 1;
281 mtu = 0; /* Silence GCC warning. */
284 IPSTAT_INC(ips_noroute);
285 error = EHOSTUNREACH;
288 ia = ifatoia(ro->ro_rt->rt_ifa);
289 ifa_ref(&ia->ia_ifa);
290 ifp = ro->ro_rt->rt_ifp;
291 ro->ro_rt->rt_rmx.rmx_pksent++;
292 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
293 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
294 if (ro->ro_rt->rt_flags & RTF_HOST)
295 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
297 isbroadcast = in_broadcast(dst->sin_addr, ifp);
300 * Calculate MTU. If we have a route that is up, use that,
301 * otherwise use the interface's MTU.
303 if (ro->ro_rt != NULL && (ro->ro_rt->rt_flags & (RTF_UP|RTF_HOST))) {
305 * This case can happen if the user changed the MTU
306 * of an interface after enabling IP on it. Because
307 * most netifs don't keep track of routes pointing to
308 * them, there is no way for one to update all its
309 * routes when the MTU is changed.
311 if (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)
312 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
313 mtu = ro->ro_rt->rt_rmx.rmx_mtu;
317 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
318 m->m_flags |= M_MCAST;
320 * IP destination address is multicast. Make sure "dst"
321 * still points to the address in "ro". (It may have been
322 * changed to point to a gateway address, above.)
324 dst = (struct sockaddr_in *)&ro->ro_dst;
326 * See if the caller provided any multicast options
329 ip->ip_ttl = imo->imo_multicast_ttl;
330 if (imo->imo_multicast_vif != -1)
333 ip_mcast_src(imo->imo_multicast_vif) :
336 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
338 * Confirm that the outgoing interface supports multicast.
340 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
341 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
342 IPSTAT_INC(ips_noroute);
348 * If source address not specified yet, use address
349 * of outgoing interface.
351 if (ip->ip_src.s_addr == INADDR_ANY) {
352 /* Interface may have no addresses. */
354 ip->ip_src = IA_SIN(ia)->sin_addr;
357 if ((imo == NULL && in_mcast_loop) ||
358 (imo && imo->imo_multicast_loop)) {
360 * Loop back multicast datagram if not expressly
361 * forbidden to do so, even if we are not a member
362 * of the group; ip_input() will filter it later,
363 * thus deferring a hash lookup and mutex acquisition
364 * at the expense of a cheap copy using m_copym().
366 ip_mloopback(ifp, m, dst, hlen);
369 * If we are acting as a multicast router, perform
370 * multicast forwarding as if the packet had just
371 * arrived on the interface to which we are about
372 * to send. The multicast forwarding function
373 * recursively calls this function, using the
374 * IP_FORWARDING flag to prevent infinite recursion.
376 * Multicasts that are looped back by ip_mloopback(),
377 * above, will be forwarded by the ip_input() routine,
380 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
382 * If rsvp daemon is not running, do not
383 * set ip_moptions. This ensures that the packet
384 * is multicast and not just sent down one link
385 * as prescribed by rsvpd.
390 ip_mforward(ip, ifp, m, imo) != 0) {
398 * Multicasts with a time-to-live of zero may be looped-
399 * back, above, but must not be transmitted on a network.
400 * Also, multicasts addressed to the loopback interface
401 * are not sent -- the above call to ip_mloopback() will
402 * loop back a copy. ip_input() will drop the copy if
403 * this host does not belong to the destination group on
404 * the loopback interface.
406 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
415 * If the source address is not specified yet, use the address
416 * of the outoing interface.
418 if (ip->ip_src.s_addr == INADDR_ANY) {
419 /* Interface may have no addresses. */
421 ip->ip_src = IA_SIN(ia)->sin_addr;
426 * Verify that we have any chance at all of being able to queue the
427 * packet or packet fragments, unless ALTQ is enabled on the given
428 * interface in which case packetdrop should be done by queueing.
431 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
432 ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >=
433 ifp->if_snd.ifq_maxlen))
435 if ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >=
436 ifp->if_snd.ifq_maxlen)
440 IPSTAT_INC(ips_odropped);
441 ifp->if_snd.ifq_drops += (ip->ip_len / ifp->if_mtu + 1);
446 * Look for broadcast address and
447 * verify user is allowed to send
451 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
452 error = EADDRNOTAVAIL;
455 if ((flags & IP_ALLOWBROADCAST) == 0) {
459 /* don't allow broadcast messages to be fragmented */
460 if (ip->ip_len > mtu) {
464 m->m_flags |= M_BCAST;
466 m->m_flags &= ~M_BCAST;
471 switch(ip_ipsec_output(&m, inp, &flags, &error, &ifp)) {
478 break; /* Continue with packet processing. */
481 * Check if there was a route for this packet; return error if not.
483 if (no_route_but_check_spd) {
484 IPSTAT_INC(ips_noroute);
485 error = EHOSTUNREACH;
488 /* Update variables that are affected by ipsec4_output(). */
489 ip = mtod(m, struct ip *);
490 hlen = ip->ip_hl << 2;
493 /* Jump over all PFIL processing if hooks are not active. */
494 if (!PFIL_HOOKED(&inet_pfil_hook))
497 /* Run through list of hooks for output packets. */
498 odst.s_addr = ip->ip_dst.s_addr;
499 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
500 if (error != 0 || m == NULL)
503 ip = mtod(m, struct ip *);
505 /* See if destination IP address was changed by packet filter. */
506 if (odst.s_addr != ip->ip_dst.s_addr) {
507 m->m_flags |= M_SKIP_FIREWALL;
508 /* If destination is now ourself drop to ip_input(). */
509 if (in_localip(ip->ip_dst)) {
510 m->m_flags |= M_FASTFWD_OURS;
511 if (m->m_pkthdr.rcvif == NULL)
512 m->m_pkthdr.rcvif = V_loif;
513 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
514 m->m_pkthdr.csum_flags |=
515 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
516 m->m_pkthdr.csum_data = 0xffff;
518 m->m_pkthdr.csum_flags |=
519 CSUM_IP_CHECKED | CSUM_IP_VALID;
521 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
522 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
524 error = netisr_queue(NETISR_IP, m);
527 goto again; /* Redo the routing table lookup. */
530 #ifdef IPFIREWALL_FORWARD
531 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
532 if (m->m_flags & M_FASTFWD_OURS) {
533 if (m->m_pkthdr.rcvif == NULL)
534 m->m_pkthdr.rcvif = V_loif;
535 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
536 m->m_pkthdr.csum_flags |=
537 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
538 m->m_pkthdr.csum_data = 0xffff;
541 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
542 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
544 m->m_pkthdr.csum_flags |=
545 CSUM_IP_CHECKED | CSUM_IP_VALID;
547 error = netisr_queue(NETISR_IP, m);
550 /* Or forward to some other address? */
551 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
553 dst = (struct sockaddr_in *)&ro->ro_dst;
554 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
555 m->m_flags |= M_SKIP_FIREWALL;
556 m_tag_delete(m, fwd_tag);
559 #endif /* IPFIREWALL_FORWARD */
562 /* 127/8 must not appear on wire - RFC1122. */
563 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
564 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
565 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
566 IPSTAT_INC(ips_badaddr);
567 error = EADDRNOTAVAIL;
572 m->m_pkthdr.csum_flags |= CSUM_IP;
573 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
574 if (sw_csum & CSUM_DELAY_DATA) {
576 sw_csum &= ~CSUM_DELAY_DATA;
579 if (sw_csum & CSUM_SCTP) {
580 sctp_delayed_cksum(m);
581 sw_csum &= ~CSUM_SCTP;
584 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
587 * If small enough for interface, or the interface will take
588 * care of the fragmentation for us, we can just send directly.
590 if (ip->ip_len <= mtu ||
591 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
592 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
593 ip->ip_len = htons(ip->ip_len);
594 ip->ip_off = htons(ip->ip_off);
596 if (sw_csum & CSUM_DELAY_IP)
597 ip->ip_sum = in_cksum(m, hlen);
600 * Record statistics for this interface address.
601 * With CSUM_TSO the byte/packet count will be slightly
602 * incorrect because we count the IP+TCP headers only
603 * once instead of for every generated packet.
605 if (!(flags & IP_FORWARDING) && ia) {
606 if (m->m_pkthdr.csum_flags & CSUM_TSO)
607 ia->ia_ifa.if_opackets +=
608 m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
610 ia->ia_ifa.if_opackets++;
611 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
613 #ifdef MBUF_STRESS_TEST
614 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
615 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size);
618 * Reset layer specific mbuf flags
619 * to avoid confusing lower layers.
621 m->m_flags &= ~(M_PROTOFLAGS);
622 error = (*ifp->if_output)(ifp, m,
623 (struct sockaddr *)dst, ro);
627 /* Balk when DF bit is set or the interface didn't support TSO. */
628 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
630 IPSTAT_INC(ips_cantfrag);
635 * Too large for interface; fragment if possible. If successful,
636 * on return, m will point to a list of packets to be sent.
638 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum);
645 /* Record statistics for this interface address. */
647 ia->ia_ifa.if_opackets++;
648 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
651 * Reset layer specific mbuf flags
652 * to avoid confusing upper layers.
654 m->m_flags &= ~(M_PROTOFLAGS);
656 error = (*ifp->if_output)(ifp, m,
657 (struct sockaddr *)dst, ro);
663 IPSTAT_INC(ips_fragmented);
666 if (ro == &iproute && ro->ro_rt && !nortfree) {
670 ifa_free(&ia->ia_ifa);
678 * Create a chain of fragments which fit the given mtu. m_frag points to the
679 * mbuf to be fragmented; on return it points to the chain with the fragments.
680 * Return 0 if no error. If error, m_frag may contain a partially built
681 * chain of fragments that should be freed by the caller.
683 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
684 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
687 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
688 u_long if_hwassist_flags, int sw_csum)
691 int hlen = ip->ip_hl << 2;
692 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
694 struct mbuf *m0 = *m_frag; /* the original packet */
699 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
700 IPSTAT_INC(ips_cantfrag);
705 * Must be able to put at least 8 bytes per fragment.
711 * If the interface will not calculate checksums on
712 * fragmented packets, then do it here.
714 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
715 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
716 in_delayed_cksum(m0);
717 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
720 if (m0->m_pkthdr.csum_flags & CSUM_SCTP &&
721 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
722 sctp_delayed_cksum(m0);
723 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
726 if (len > PAGE_SIZE) {
728 * Fragment large datagrams such that each segment
729 * contains a multiple of PAGE_SIZE amount of data,
730 * plus headers. This enables a receiver to perform
731 * page-flipping zero-copy optimizations.
733 * XXX When does this help given that sender and receiver
734 * could have different page sizes, and also mtu could
735 * be less than the receiver's page size ?
740 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
744 * firstlen (off - hlen) must be aligned on an
748 goto smart_frag_failure;
749 off = ((off - hlen) & ~7) + hlen;
750 newlen = (~PAGE_MASK) & mtu;
751 if ((newlen + sizeof (struct ip)) > mtu) {
752 /* we failed, go back the default */
763 firstlen = off - hlen;
764 mnext = &m0->m_nextpkt; /* pointer to next packet */
767 * Loop through length of segment after first fragment,
768 * make new header and copy data of each part and link onto chain.
769 * Here, m0 is the original packet, m is the fragment being created.
770 * The fragments are linked off the m_nextpkt of the original
771 * packet, which after processing serves as the first fragment.
773 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
774 struct ip *mhip; /* ip header on the fragment */
776 int mhlen = sizeof (struct ip);
778 MGETHDR(m, M_DONTWAIT, MT_DATA);
781 IPSTAT_INC(ips_odropped);
784 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
786 * In the first mbuf, leave room for the link header, then
787 * copy the original IP header including options. The payload
788 * goes into an additional mbuf chain returned by m_copym().
790 m->m_data += max_linkhdr;
791 mhip = mtod(m, struct ip *);
793 if (hlen > sizeof (struct ip)) {
794 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
795 mhip->ip_v = IPVERSION;
796 mhip->ip_hl = mhlen >> 2;
799 /* XXX do we need to add ip->ip_off below ? */
800 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
801 if (off + len >= ip->ip_len) { /* last fragment */
802 len = ip->ip_len - off;
803 m->m_flags |= M_LASTFRAG;
805 mhip->ip_off |= IP_MF;
806 mhip->ip_len = htons((u_short)(len + mhlen));
807 m->m_next = m_copym(m0, off, len, M_DONTWAIT);
808 if (m->m_next == NULL) { /* copy failed */
810 error = ENOBUFS; /* ??? */
811 IPSTAT_INC(ips_odropped);
814 m->m_pkthdr.len = mhlen + len;
815 m->m_pkthdr.rcvif = NULL;
817 mac_netinet_fragment(m0, m);
819 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
820 mhip->ip_off = htons(mhip->ip_off);
822 if (sw_csum & CSUM_DELAY_IP)
823 mhip->ip_sum = in_cksum(m, mhlen);
825 mnext = &m->m_nextpkt;
827 IPSTAT_ADD(ips_ofragments, nfrags);
829 /* set first marker for fragment chain */
830 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
831 m0->m_pkthdr.csum_data = nfrags;
834 * Update first fragment by trimming what's been copied out
835 * and updating header.
837 m_adj(m0, hlen + firstlen - ip->ip_len);
838 m0->m_pkthdr.len = hlen + firstlen;
839 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
841 ip->ip_off = htons(ip->ip_off);
843 if (sw_csum & CSUM_DELAY_IP)
844 ip->ip_sum = in_cksum(m0, hlen);
852 in_delayed_cksum(struct mbuf *m)
855 u_short csum, offset;
857 ip = mtod(m, struct ip *);
858 offset = ip->ip_hl << 2 ;
859 csum = in_cksum_skip(m, ip->ip_len, offset);
860 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
862 offset += m->m_pkthdr.csum_data; /* checksum offset */
864 if (offset + sizeof(u_short) > m->m_len) {
865 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
866 m->m_len, offset, ip->ip_p);
869 * this shouldn't happen, but if it does, the
870 * correct behavior may be to insert the checksum
871 * in the appropriate next mbuf in the chain.
875 *(u_short *)(m->m_data + offset) = csum;
879 * IP socket option processing.
882 ip_ctloutput(struct socket *so, struct sockopt *sopt)
884 struct inpcb *inp = sotoinpcb(so);
888 if (sopt->sopt_level != IPPROTO_IP) {
889 if ((sopt->sopt_level == SOL_SOCKET) &&
890 (sopt->sopt_name == SO_SETFIB)) {
891 inp->inp_inc.inc_fibnum = so->so_fibnum;
897 switch (sopt->sopt_dir) {
899 switch (sopt->sopt_name) {
906 if (sopt->sopt_valsize > MLEN) {
910 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
915 m->m_len = sopt->sopt_valsize;
916 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
923 error = ip_pcbopts(inp, sopt->sopt_name, m);
929 if (sopt->sopt_td != NULL) {
930 error = priv_check(sopt->sopt_td,
931 PRIV_NETINET_BINDANY);
947 error = sooptcopyin(sopt, &optval, sizeof optval,
952 switch (sopt->sopt_name) {
954 inp->inp_ip_tos = optval;
958 inp->inp_ip_ttl = optval;
962 if (optval >= 0 && optval <= MAXTTL)
963 inp->inp_ip_minttl = optval;
968 #define OPTSET(bit) do { \
971 inp->inp_flags |= bit; \
973 inp->inp_flags &= ~bit; \
978 OPTSET(INP_RECVOPTS);
982 OPTSET(INP_RECVRETOPTS);
986 OPTSET(INP_RECVDSTADDR);
1002 OPTSET(INP_ONESBCAST);
1005 OPTSET(INP_DONTFRAG);
1008 OPTSET(INP_BINDANY);
1015 * Multicast socket options are processed by the in_mcast
1018 case IP_MULTICAST_IF:
1019 case IP_MULTICAST_VIF:
1020 case IP_MULTICAST_TTL:
1021 case IP_MULTICAST_LOOP:
1022 case IP_ADD_MEMBERSHIP:
1023 case IP_DROP_MEMBERSHIP:
1024 case IP_ADD_SOURCE_MEMBERSHIP:
1025 case IP_DROP_SOURCE_MEMBERSHIP:
1026 case IP_BLOCK_SOURCE:
1027 case IP_UNBLOCK_SOURCE:
1029 case MCAST_JOIN_GROUP:
1030 case MCAST_LEAVE_GROUP:
1031 case MCAST_JOIN_SOURCE_GROUP:
1032 case MCAST_LEAVE_SOURCE_GROUP:
1033 case MCAST_BLOCK_SOURCE:
1034 case MCAST_UNBLOCK_SOURCE:
1035 error = inp_setmoptions(inp, sopt);
1039 error = sooptcopyin(sopt, &optval, sizeof optval,
1046 case IP_PORTRANGE_DEFAULT:
1047 inp->inp_flags &= ~(INP_LOWPORT);
1048 inp->inp_flags &= ~(INP_HIGHPORT);
1051 case IP_PORTRANGE_HIGH:
1052 inp->inp_flags &= ~(INP_LOWPORT);
1053 inp->inp_flags |= INP_HIGHPORT;
1056 case IP_PORTRANGE_LOW:
1057 inp->inp_flags &= ~(INP_HIGHPORT);
1058 inp->inp_flags |= INP_LOWPORT;
1069 case IP_IPSEC_POLICY:
1074 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1076 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1078 req = mtod(m, caddr_t);
1079 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1080 m->m_len, (sopt->sopt_td != NULL) ?
1081 sopt->sopt_td->td_ucred : NULL);
1088 error = ENOPROTOOPT;
1094 switch (sopt->sopt_name) {
1097 if (inp->inp_options)
1098 error = sooptcopyout(sopt,
1099 mtod(inp->inp_options,
1101 inp->inp_options->m_len);
1103 sopt->sopt_valsize = 0;
1110 case IP_RECVRETOPTS:
1111 case IP_RECVDSTADDR:
1118 switch (sopt->sopt_name) {
1121 optval = inp->inp_ip_tos;
1125 optval = inp->inp_ip_ttl;
1129 optval = inp->inp_ip_minttl;
1132 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1135 optval = OPTBIT(INP_RECVOPTS);
1138 case IP_RECVRETOPTS:
1139 optval = OPTBIT(INP_RECVRETOPTS);
1142 case IP_RECVDSTADDR:
1143 optval = OPTBIT(INP_RECVDSTADDR);
1147 optval = OPTBIT(INP_RECVTTL);
1151 optval = OPTBIT(INP_RECVIF);
1155 if (inp->inp_flags & INP_HIGHPORT)
1156 optval = IP_PORTRANGE_HIGH;
1157 else if (inp->inp_flags & INP_LOWPORT)
1158 optval = IP_PORTRANGE_LOW;
1164 optval = OPTBIT(INP_FAITH);
1168 optval = OPTBIT(INP_ONESBCAST);
1171 optval = OPTBIT(INP_DONTFRAG);
1174 error = sooptcopyout(sopt, &optval, sizeof 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:
1186 error = inp_getmoptions(inp, sopt);
1190 case IP_IPSEC_POLICY:
1192 struct mbuf *m = NULL;
1197 req = mtod(m, caddr_t);
1200 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1202 error = soopt_mcopyout(sopt, m); /* XXX */
1210 error = ENOPROTOOPT;
1219 * Routine called from ip_output() to loop back a copy of an IP multicast
1220 * packet to the input queue of a specified interface. Note that this
1221 * calls the output routine of the loopback "driver", but with an interface
1222 * pointer that might NOT be a loopback interface -- evil, but easier than
1223 * replicating that code here.
1226 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1229 register struct ip *ip;
1233 * Make a deep copy of the packet because we're going to
1234 * modify the pack in order to generate checksums.
1236 copym = m_dup(m, M_DONTWAIT);
1237 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1238 copym = m_pullup(copym, hlen);
1239 if (copym != NULL) {
1240 /* If needed, compute the checksum and mark it as valid. */
1241 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1242 in_delayed_cksum(copym);
1243 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1244 copym->m_pkthdr.csum_flags |=
1245 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1246 copym->m_pkthdr.csum_data = 0xffff;
1249 * We don't bother to fragment if the IP length is greater
1250 * than the interface's MTU. Can this possibly matter?
1252 ip = mtod(copym, struct ip *);
1253 ip->ip_len = htons(ip->ip_len);
1254 ip->ip_off = htons(ip->ip_off);
1256 ip->ip_sum = in_cksum(copym, hlen);
1258 if (dst->sin_family != AF_INET) {
1259 printf("ip_mloopback: bad address family %d\n",
1261 dst->sin_family = AF_INET;
1264 if_simloop(ifp, copym, dst->sin_family, 0);