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;
131 struct rtentry *rte; /* cache for ro->ro_rt */
133 #ifdef IPFIREWALL_FORWARD
134 struct m_tag *fwd_tag = NULL;
137 int no_route_but_check_spd = 0;
142 INP_LOCK_ASSERT(inp);
143 M_SETFIB(m, inp->inp_inc.inc_fibnum);
144 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
145 m->m_pkthdr.flowid = inp->inp_flowid;
146 m->m_flags |= M_FLOWID;
152 bzero(ro, sizeof (*ro));
159 * The flow table returns route entries valid for up to 30
160 * seconds; we rely on the remainder of ip_output() taking no
161 * longer than that long for the stability of ro_rt. The
162 * flow ID assignment must have happened before this point.
164 if ((fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET)) != NULL) {
165 flow_to_route(fle, ro);
174 m = ip_insertoptions(m, opt, &len);
178 ip = mtod(m, struct ip *);
181 * Fill in IP header. If we are not allowing fragmentation,
182 * then the ip_id field is meaningless, but we don't set it
183 * to zero. Doing so causes various problems when devices along
184 * the path (routers, load balancers, firewalls, etc.) illegally
185 * disable DF on our packet. Note that a 16-bit counter
186 * will wrap around in less than 10 seconds at 100 Mbit/s on a
187 * medium with MTU 1500. See Steven M. Bellovin, "A Technique
188 * for Counting NATted Hosts", Proc. IMW'02, available at
189 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
191 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
192 ip->ip_v = IPVERSION;
193 ip->ip_hl = hlen >> 2;
194 ip->ip_id = ip_newid();
195 IPSTAT_INC(ips_localout);
197 hlen = ip->ip_hl << 2;
200 dst = (struct sockaddr_in *)&ro->ro_dst;
203 * If there is a cached route,
204 * check that it is to the same destination
205 * and is still up. If not, free it and try again.
206 * The address family should also be checked in case of sharing the
210 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
211 dst->sin_family != AF_INET ||
212 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
215 rte = ro->ro_rt = (struct rtentry *)NULL;
216 ro->ro_lle = (struct llentry *)NULL;
218 #ifdef IPFIREWALL_FORWARD
219 if (rte == NULL && fwd_tag == NULL) {
223 bzero(dst, sizeof(*dst));
224 dst->sin_family = AF_INET;
225 dst->sin_len = sizeof(*dst);
226 dst->sin_addr = ip->ip_dst;
229 * If routing to interface only, short circuit routing lookup.
230 * The use of an all-ones broadcast address implies this; an
231 * interface is specified by the broadcast address of an interface,
232 * or the destination address of a ptp interface.
234 if (flags & IP_SENDONES) {
235 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
236 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
237 IPSTAT_INC(ips_noroute);
241 ip->ip_dst.s_addr = INADDR_BROADCAST;
242 dst->sin_addr = ip->ip_dst;
246 } else if (flags & IP_ROUTETOIF) {
247 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
248 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) {
249 IPSTAT_INC(ips_noroute);
255 isbroadcast = in_broadcast(dst->sin_addr, ifp);
256 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
257 imo != NULL && imo->imo_multicast_ifp != NULL) {
259 * Bypass the normal routing lookup for multicast
260 * packets if the interface is specified.
262 ifp = imo->imo_multicast_ifp;
264 isbroadcast = 0; /* fool gcc */
267 * We want to do any cloning requested by the link layer,
268 * as this is probably required in all cases for correct
269 * operation (as it is for ARP).
273 rtalloc_mpath_fib(ro,
274 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
275 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
277 in_rtalloc_ign(ro, 0,
278 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
285 * There is no route for this packet, but it is
286 * possible that a matching SPD entry exists.
288 no_route_but_check_spd = 1;
289 mtu = 0; /* Silence GCC warning. */
292 IPSTAT_INC(ips_noroute);
293 error = EHOSTUNREACH;
296 ia = ifatoia(rte->rt_ifa);
297 ifa_ref(&ia->ia_ifa);
299 rte->rt_rmx.rmx_pksent++;
300 if (rte->rt_flags & RTF_GATEWAY)
301 dst = (struct sockaddr_in *)rte->rt_gateway;
302 if (rte->rt_flags & RTF_HOST)
303 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
305 isbroadcast = in_broadcast(dst->sin_addr, ifp);
308 * Calculate MTU. If we have a route that is up, use that,
309 * otherwise use the interface's MTU.
311 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
313 * This case can happen if the user changed the MTU
314 * of an interface after enabling IP on it. Because
315 * most netifs don't keep track of routes pointing to
316 * them, there is no way for one to update all its
317 * routes when the MTU is changed.
319 if (rte->rt_rmx.rmx_mtu > ifp->if_mtu)
320 rte->rt_rmx.rmx_mtu = ifp->if_mtu;
321 mtu = rte->rt_rmx.rmx_mtu;
325 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
326 m->m_flags |= M_MCAST;
328 * IP destination address is multicast. Make sure "dst"
329 * still points to the address in "ro". (It may have been
330 * changed to point to a gateway address, above.)
332 dst = (struct sockaddr_in *)&ro->ro_dst;
334 * See if the caller provided any multicast options
337 ip->ip_ttl = imo->imo_multicast_ttl;
338 if (imo->imo_multicast_vif != -1)
341 ip_mcast_src(imo->imo_multicast_vif) :
344 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
346 * Confirm that the outgoing interface supports multicast.
348 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
349 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
350 IPSTAT_INC(ips_noroute);
356 * If source address not specified yet, use address
357 * of outgoing interface.
359 if (ip->ip_src.s_addr == INADDR_ANY) {
360 /* Interface may have no addresses. */
362 ip->ip_src = IA_SIN(ia)->sin_addr;
365 if ((imo == NULL && in_mcast_loop) ||
366 (imo && imo->imo_multicast_loop)) {
368 * Loop back multicast datagram if not expressly
369 * forbidden to do so, even if we are not a member
370 * of the group; ip_input() will filter it later,
371 * thus deferring a hash lookup and mutex acquisition
372 * at the expense of a cheap copy using m_copym().
374 ip_mloopback(ifp, m, dst, hlen);
377 * If we are acting as a multicast router, perform
378 * multicast forwarding as if the packet had just
379 * arrived on the interface to which we are about
380 * to send. The multicast forwarding function
381 * recursively calls this function, using the
382 * IP_FORWARDING flag to prevent infinite recursion.
384 * Multicasts that are looped back by ip_mloopback(),
385 * above, will be forwarded by the ip_input() routine,
388 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
390 * If rsvp daemon is not running, do not
391 * set ip_moptions. This ensures that the packet
392 * is multicast and not just sent down one link
393 * as prescribed by rsvpd.
398 ip_mforward(ip, ifp, m, imo) != 0) {
406 * Multicasts with a time-to-live of zero may be looped-
407 * back, above, but must not be transmitted on a network.
408 * Also, multicasts addressed to the loopback interface
409 * are not sent -- the above call to ip_mloopback() will
410 * loop back a copy. ip_input() will drop the copy if
411 * this host does not belong to the destination group on
412 * the loopback interface.
414 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
423 * If the source address is not specified yet, use the address
424 * of the outoing interface.
426 if (ip->ip_src.s_addr == INADDR_ANY) {
427 /* Interface may have no addresses. */
429 ip->ip_src = IA_SIN(ia)->sin_addr;
434 * Verify that we have any chance at all of being able to queue the
435 * packet or packet fragments, unless ALTQ is enabled on the given
436 * interface in which case packetdrop should be done by queueing.
439 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
440 ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >=
441 ifp->if_snd.ifq_maxlen))
443 if ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >=
444 ifp->if_snd.ifq_maxlen)
448 IPSTAT_INC(ips_odropped);
449 ifp->if_snd.ifq_drops += (ip->ip_len / ifp->if_mtu + 1);
454 * Look for broadcast address and
455 * verify user is allowed to send
459 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
460 error = EADDRNOTAVAIL;
463 if ((flags & IP_ALLOWBROADCAST) == 0) {
467 /* don't allow broadcast messages to be fragmented */
468 if (ip->ip_len > mtu) {
472 m->m_flags |= M_BCAST;
474 m->m_flags &= ~M_BCAST;
479 switch(ip_ipsec_output(&m, inp, &flags, &error, &ifp)) {
486 break; /* Continue with packet processing. */
489 * Check if there was a route for this packet; return error if not.
491 if (no_route_but_check_spd) {
492 IPSTAT_INC(ips_noroute);
493 error = EHOSTUNREACH;
496 /* Update variables that are affected by ipsec4_output(). */
497 ip = mtod(m, struct ip *);
498 hlen = ip->ip_hl << 2;
501 /* Jump over all PFIL processing if hooks are not active. */
502 if (!PFIL_HOOKED(&V_inet_pfil_hook))
505 /* Run through list of hooks for output packets. */
506 odst.s_addr = ip->ip_dst.s_addr;
507 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
508 if (error != 0 || m == NULL)
511 ip = mtod(m, struct ip *);
513 /* See if destination IP address was changed by packet filter. */
514 if (odst.s_addr != ip->ip_dst.s_addr) {
515 m->m_flags |= M_SKIP_FIREWALL;
516 /* If destination is now ourself drop to ip_input(). */
517 if (in_localip(ip->ip_dst)) {
518 m->m_flags |= M_FASTFWD_OURS;
519 if (m->m_pkthdr.rcvif == NULL)
520 m->m_pkthdr.rcvif = V_loif;
521 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
522 m->m_pkthdr.csum_flags |=
523 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
524 m->m_pkthdr.csum_data = 0xffff;
526 m->m_pkthdr.csum_flags |=
527 CSUM_IP_CHECKED | CSUM_IP_VALID;
529 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
530 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
532 error = netisr_queue(NETISR_IP, m);
535 goto again; /* Redo the routing table lookup. */
538 #ifdef IPFIREWALL_FORWARD
539 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
540 if (m->m_flags & M_FASTFWD_OURS) {
541 if (m->m_pkthdr.rcvif == NULL)
542 m->m_pkthdr.rcvif = V_loif;
543 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
544 m->m_pkthdr.csum_flags |=
545 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
546 m->m_pkthdr.csum_data = 0xffff;
549 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
550 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
552 m->m_pkthdr.csum_flags |=
553 CSUM_IP_CHECKED | CSUM_IP_VALID;
555 error = netisr_queue(NETISR_IP, m);
558 /* Or forward to some other address? */
559 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
561 dst = (struct sockaddr_in *)&ro->ro_dst;
562 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
563 m->m_flags |= M_SKIP_FIREWALL;
564 m_tag_delete(m, fwd_tag);
567 #endif /* IPFIREWALL_FORWARD */
570 /* 127/8 must not appear on wire - RFC1122. */
571 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
572 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
573 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
574 IPSTAT_INC(ips_badaddr);
575 error = EADDRNOTAVAIL;
580 m->m_pkthdr.csum_flags |= CSUM_IP;
581 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
582 if (sw_csum & CSUM_DELAY_DATA) {
584 sw_csum &= ~CSUM_DELAY_DATA;
587 if (sw_csum & CSUM_SCTP) {
588 sctp_delayed_cksum(m);
589 sw_csum &= ~CSUM_SCTP;
592 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
595 * If small enough for interface, or the interface will take
596 * care of the fragmentation for us, we can just send directly.
598 if (ip->ip_len <= mtu ||
599 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
600 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
601 ip->ip_len = htons(ip->ip_len);
602 ip->ip_off = htons(ip->ip_off);
604 if (sw_csum & CSUM_DELAY_IP)
605 ip->ip_sum = in_cksum(m, hlen);
608 * Record statistics for this interface address.
609 * With CSUM_TSO the byte/packet count will be slightly
610 * incorrect because we count the IP+TCP headers only
611 * once instead of for every generated packet.
613 if (!(flags & IP_FORWARDING) && ia) {
614 if (m->m_pkthdr.csum_flags & CSUM_TSO)
615 ia->ia_ifa.if_opackets +=
616 m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
618 ia->ia_ifa.if_opackets++;
619 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
621 #ifdef MBUF_STRESS_TEST
622 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
623 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size);
626 * Reset layer specific mbuf flags
627 * to avoid confusing lower layers.
629 m->m_flags &= ~(M_PROTOFLAGS);
630 error = (*ifp->if_output)(ifp, m,
631 (struct sockaddr *)dst, ro);
635 /* Balk when DF bit is set or the interface didn't support TSO. */
636 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
638 IPSTAT_INC(ips_cantfrag);
643 * Too large for interface; fragment if possible. If successful,
644 * on return, m will point to a list of packets to be sent.
646 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum);
653 /* Record statistics for this interface address. */
655 ia->ia_ifa.if_opackets++;
656 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
659 * Reset layer specific mbuf flags
660 * to avoid confusing upper layers.
662 m->m_flags &= ~(M_PROTOFLAGS);
664 error = (*ifp->if_output)(ifp, m,
665 (struct sockaddr *)dst, ro);
671 IPSTAT_INC(ips_fragmented);
674 if (ro == &iproute && ro->ro_rt && !nortfree) {
678 ifa_free(&ia->ia_ifa);
686 * Create a chain of fragments which fit the given mtu. m_frag points to the
687 * mbuf to be fragmented; on return it points to the chain with the fragments.
688 * Return 0 if no error. If error, m_frag may contain a partially built
689 * chain of fragments that should be freed by the caller.
691 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
692 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
695 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
696 u_long if_hwassist_flags, int sw_csum)
699 int hlen = ip->ip_hl << 2;
700 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
702 struct mbuf *m0 = *m_frag; /* the original packet */
707 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
708 IPSTAT_INC(ips_cantfrag);
713 * Must be able to put at least 8 bytes per fragment.
719 * If the interface will not calculate checksums on
720 * fragmented packets, then do it here.
722 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
723 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
724 in_delayed_cksum(m0);
725 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
728 if (m0->m_pkthdr.csum_flags & CSUM_SCTP &&
729 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
730 sctp_delayed_cksum(m0);
731 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
734 if (len > PAGE_SIZE) {
736 * Fragment large datagrams such that each segment
737 * contains a multiple of PAGE_SIZE amount of data,
738 * plus headers. This enables a receiver to perform
739 * page-flipping zero-copy optimizations.
741 * XXX When does this help given that sender and receiver
742 * could have different page sizes, and also mtu could
743 * be less than the receiver's page size ?
748 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
752 * firstlen (off - hlen) must be aligned on an
756 goto smart_frag_failure;
757 off = ((off - hlen) & ~7) + hlen;
758 newlen = (~PAGE_MASK) & mtu;
759 if ((newlen + sizeof (struct ip)) > mtu) {
760 /* we failed, go back the default */
771 firstlen = off - hlen;
772 mnext = &m0->m_nextpkt; /* pointer to next packet */
775 * Loop through length of segment after first fragment,
776 * make new header and copy data of each part and link onto chain.
777 * Here, m0 is the original packet, m is the fragment being created.
778 * The fragments are linked off the m_nextpkt of the original
779 * packet, which after processing serves as the first fragment.
781 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
782 struct ip *mhip; /* ip header on the fragment */
784 int mhlen = sizeof (struct ip);
786 MGETHDR(m, M_DONTWAIT, MT_DATA);
789 IPSTAT_INC(ips_odropped);
792 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
794 * In the first mbuf, leave room for the link header, then
795 * copy the original IP header including options. The payload
796 * goes into an additional mbuf chain returned by m_copym().
798 m->m_data += max_linkhdr;
799 mhip = mtod(m, struct ip *);
801 if (hlen > sizeof (struct ip)) {
802 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
803 mhip->ip_v = IPVERSION;
804 mhip->ip_hl = mhlen >> 2;
807 /* XXX do we need to add ip->ip_off below ? */
808 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
809 if (off + len >= ip->ip_len) { /* last fragment */
810 len = ip->ip_len - off;
811 m->m_flags |= M_LASTFRAG;
813 mhip->ip_off |= IP_MF;
814 mhip->ip_len = htons((u_short)(len + mhlen));
815 m->m_next = m_copym(m0, off, len, M_DONTWAIT);
816 if (m->m_next == NULL) { /* copy failed */
818 error = ENOBUFS; /* ??? */
819 IPSTAT_INC(ips_odropped);
822 m->m_pkthdr.len = mhlen + len;
823 m->m_pkthdr.rcvif = NULL;
825 mac_netinet_fragment(m0, m);
827 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
828 mhip->ip_off = htons(mhip->ip_off);
830 if (sw_csum & CSUM_DELAY_IP)
831 mhip->ip_sum = in_cksum(m, mhlen);
833 mnext = &m->m_nextpkt;
835 IPSTAT_ADD(ips_ofragments, nfrags);
837 /* set first marker for fragment chain */
838 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
839 m0->m_pkthdr.csum_data = nfrags;
842 * Update first fragment by trimming what's been copied out
843 * and updating header.
845 m_adj(m0, hlen + firstlen - ip->ip_len);
846 m0->m_pkthdr.len = hlen + firstlen;
847 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
849 ip->ip_off = htons(ip->ip_off);
851 if (sw_csum & CSUM_DELAY_IP)
852 ip->ip_sum = in_cksum(m0, hlen);
860 in_delayed_cksum(struct mbuf *m)
863 u_short csum, offset;
865 ip = mtod(m, struct ip *);
866 offset = ip->ip_hl << 2 ;
867 csum = in_cksum_skip(m, ip->ip_len, offset);
868 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
870 offset += m->m_pkthdr.csum_data; /* checksum offset */
872 if (offset + sizeof(u_short) > m->m_len) {
873 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
874 m->m_len, offset, ip->ip_p);
877 * this shouldn't happen, but if it does, the
878 * correct behavior may be to insert the checksum
879 * in the appropriate next mbuf in the chain.
883 *(u_short *)(m->m_data + offset) = csum;
887 * IP socket option processing.
890 ip_ctloutput(struct socket *so, struct sockopt *sopt)
892 struct inpcb *inp = sotoinpcb(so);
896 if (sopt->sopt_level != IPPROTO_IP) {
897 if ((sopt->sopt_level == SOL_SOCKET) &&
898 (sopt->sopt_name == SO_SETFIB)) {
899 inp->inp_inc.inc_fibnum = so->so_fibnum;
905 switch (sopt->sopt_dir) {
907 switch (sopt->sopt_name) {
914 if (sopt->sopt_valsize > MLEN) {
918 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
923 m->m_len = sopt->sopt_valsize;
924 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
931 error = ip_pcbopts(inp, sopt->sopt_name, m);
937 if (sopt->sopt_td != NULL) {
938 error = priv_check(sopt->sopt_td,
939 PRIV_NETINET_BINDANY);
955 error = sooptcopyin(sopt, &optval, sizeof optval,
960 switch (sopt->sopt_name) {
962 inp->inp_ip_tos = optval;
966 inp->inp_ip_ttl = optval;
970 if (optval >= 0 && optval <= MAXTTL)
971 inp->inp_ip_minttl = optval;
976 #define OPTSET(bit) do { \
979 inp->inp_flags |= bit; \
981 inp->inp_flags &= ~bit; \
986 OPTSET(INP_RECVOPTS);
990 OPTSET(INP_RECVRETOPTS);
994 OPTSET(INP_RECVDSTADDR);
1010 OPTSET(INP_ONESBCAST);
1013 OPTSET(INP_DONTFRAG);
1016 OPTSET(INP_BINDANY);
1023 * Multicast socket options are processed by the in_mcast
1026 case IP_MULTICAST_IF:
1027 case IP_MULTICAST_VIF:
1028 case IP_MULTICAST_TTL:
1029 case IP_MULTICAST_LOOP:
1030 case IP_ADD_MEMBERSHIP:
1031 case IP_DROP_MEMBERSHIP:
1032 case IP_ADD_SOURCE_MEMBERSHIP:
1033 case IP_DROP_SOURCE_MEMBERSHIP:
1034 case IP_BLOCK_SOURCE:
1035 case IP_UNBLOCK_SOURCE:
1037 case MCAST_JOIN_GROUP:
1038 case MCAST_LEAVE_GROUP:
1039 case MCAST_JOIN_SOURCE_GROUP:
1040 case MCAST_LEAVE_SOURCE_GROUP:
1041 case MCAST_BLOCK_SOURCE:
1042 case MCAST_UNBLOCK_SOURCE:
1043 error = inp_setmoptions(inp, sopt);
1047 error = sooptcopyin(sopt, &optval, sizeof optval,
1054 case IP_PORTRANGE_DEFAULT:
1055 inp->inp_flags &= ~(INP_LOWPORT);
1056 inp->inp_flags &= ~(INP_HIGHPORT);
1059 case IP_PORTRANGE_HIGH:
1060 inp->inp_flags &= ~(INP_LOWPORT);
1061 inp->inp_flags |= INP_HIGHPORT;
1064 case IP_PORTRANGE_LOW:
1065 inp->inp_flags &= ~(INP_HIGHPORT);
1066 inp->inp_flags |= INP_LOWPORT;
1077 case IP_IPSEC_POLICY:
1082 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1084 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1086 req = mtod(m, caddr_t);
1087 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1088 m->m_len, (sopt->sopt_td != NULL) ?
1089 sopt->sopt_td->td_ucred : NULL);
1096 error = ENOPROTOOPT;
1102 switch (sopt->sopt_name) {
1105 if (inp->inp_options)
1106 error = sooptcopyout(sopt,
1107 mtod(inp->inp_options,
1109 inp->inp_options->m_len);
1111 sopt->sopt_valsize = 0;
1118 case IP_RECVRETOPTS:
1119 case IP_RECVDSTADDR:
1126 switch (sopt->sopt_name) {
1129 optval = inp->inp_ip_tos;
1133 optval = inp->inp_ip_ttl;
1137 optval = inp->inp_ip_minttl;
1140 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1143 optval = OPTBIT(INP_RECVOPTS);
1146 case IP_RECVRETOPTS:
1147 optval = OPTBIT(INP_RECVRETOPTS);
1150 case IP_RECVDSTADDR:
1151 optval = OPTBIT(INP_RECVDSTADDR);
1155 optval = OPTBIT(INP_RECVTTL);
1159 optval = OPTBIT(INP_RECVIF);
1163 if (inp->inp_flags & INP_HIGHPORT)
1164 optval = IP_PORTRANGE_HIGH;
1165 else if (inp->inp_flags & INP_LOWPORT)
1166 optval = IP_PORTRANGE_LOW;
1172 optval = OPTBIT(INP_FAITH);
1176 optval = OPTBIT(INP_ONESBCAST);
1179 optval = OPTBIT(INP_DONTFRAG);
1182 error = sooptcopyout(sopt, &optval, sizeof optval);
1186 * Multicast socket options are processed by the in_mcast
1189 case IP_MULTICAST_IF:
1190 case IP_MULTICAST_VIF:
1191 case IP_MULTICAST_TTL:
1192 case IP_MULTICAST_LOOP:
1194 error = inp_getmoptions(inp, sopt);
1198 case IP_IPSEC_POLICY:
1200 struct mbuf *m = NULL;
1205 req = mtod(m, caddr_t);
1208 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1210 error = soopt_mcopyout(sopt, m); /* XXX */
1218 error = ENOPROTOOPT;
1227 * Routine called from ip_output() to loop back a copy of an IP multicast
1228 * packet to the input queue of a specified interface. Note that this
1229 * calls the output routine of the loopback "driver", but with an interface
1230 * pointer that might NOT be a loopback interface -- evil, but easier than
1231 * replicating that code here.
1234 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1237 register struct ip *ip;
1241 * Make a deep copy of the packet because we're going to
1242 * modify the pack in order to generate checksums.
1244 copym = m_dup(m, M_DONTWAIT);
1245 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1246 copym = m_pullup(copym, hlen);
1247 if (copym != NULL) {
1248 /* If needed, compute the checksum and mark it as valid. */
1249 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1250 in_delayed_cksum(copym);
1251 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1252 copym->m_pkthdr.csum_flags |=
1253 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1254 copym->m_pkthdr.csum_data = 0xffff;
1257 * We don't bother to fragment if the IP length is greater
1258 * than the interface's MTU. Can this possibly matter?
1260 ip = mtod(copym, struct ip *);
1261 ip->ip_len = htons(ip->ip_len);
1262 ip->ip_off = htons(ip->ip_off);
1264 ip->ip_sum = in_cksum(copym, hlen);
1266 if (dst->sin_family != AF_INET) {
1267 printf("ip_mloopback: bad address family %d\n",
1269 dst->sin_family = AF_INET;
1272 if_simloop(ifp, copym, dst->sin_family, 0);