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"
38 #include "opt_mbuf_stress_test.h"
39 #include "opt_mpath.h"
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/malloc.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/sysctl.h>
52 #include <sys/ucred.h>
55 #include <net/netisr.h>
57 #include <net/route.h>
59 #include <net/radix_mpath.h>
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/ip.h>
65 #include <netinet/in_pcb.h>
66 #include <netinet/in_var.h>
67 #include <netinet/ip_var.h>
68 #include <netinet/ip_options.h>
71 #include <netinet/ip_ipsec.h>
72 #include <netipsec/ipsec.h>
75 #include <machine/in_cksum.h>
77 #include <security/mac/mac_framework.h>
79 #define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\
80 x, (ntohl(a.s_addr)>>24)&0xFF,\
81 (ntohl(a.s_addr)>>16)&0xFF,\
82 (ntohl(a.s_addr)>>8)&0xFF,\
83 (ntohl(a.s_addr))&0xFF, y);
87 #ifdef MBUF_STRESS_TEST
88 int mbuf_frag_size = 0;
89 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
90 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
93 static void ip_mloopback
94 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
97 extern struct protosw inetsw[];
100 * IP output. The packet in mbuf chain m contains a skeletal IP
101 * header (with len, off, ttl, proto, tos, src, dst).
102 * The mbuf chain containing the packet will be freed.
103 * The mbuf opt, if present, will not be freed.
104 * In the IP forwarding case, the packet will arrive with options already
105 * inserted, so must have a NULL opt pointer.
108 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
109 struct ip_moptions *imo, struct inpcb *inp)
112 struct ifnet *ifp = NULL; /* keep compiler happy */
114 int hlen = sizeof (struct ip);
117 struct sockaddr_in *dst = NULL; /* keep compiler happy */
118 struct in_ifaddr *ia = NULL;
119 int isbroadcast, sw_csum;
120 struct route iproute;
122 #ifdef IPFIREWALL_FORWARD
123 struct m_tag *fwd_tag = NULL;
129 bzero(ro, sizeof (*ro));
133 INP_LOCK_ASSERT(inp);
137 m = ip_insertoptions(m, opt, &len);
141 ip = mtod(m, struct ip *);
144 * Fill in IP header. If we are not allowing fragmentation,
145 * then the ip_id field is meaningless, but we don't set it
146 * to zero. Doing so causes various problems when devices along
147 * the path (routers, load balancers, firewalls, etc.) illegally
148 * disable DF on our packet. Note that a 16-bit counter
149 * will wrap around in less than 10 seconds at 100 Mbit/s on a
150 * medium with MTU 1500. See Steven M. Bellovin, "A Technique
151 * for Counting NATted Hosts", Proc. IMW'02, available at
152 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
154 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
155 ip->ip_v = IPVERSION;
156 ip->ip_hl = hlen >> 2;
157 ip->ip_id = ip_newid();
158 ipstat.ips_localout++;
160 hlen = ip->ip_hl << 2;
163 dst = (struct sockaddr_in *)&ro->ro_dst;
166 * If there is a cached route,
167 * check that it is to the same destination
168 * and is still up. If not, free it and try again.
169 * The address family should also be checked in case of sharing the
172 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
173 dst->sin_family != AF_INET ||
174 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
176 ro->ro_rt = (struct rtentry *)NULL;
178 #ifdef IPFIREWALL_FORWARD
179 if (ro->ro_rt == NULL && fwd_tag == NULL) {
181 if (ro->ro_rt == NULL) {
183 bzero(dst, sizeof(*dst));
184 dst->sin_family = AF_INET;
185 dst->sin_len = sizeof(*dst);
186 dst->sin_addr = ip->ip_dst;
189 * If routing to interface only, short circuit routing lookup.
190 * The use of an all-ones broadcast address implies this; an
191 * interface is specified by the broadcast address of an interface,
192 * or the destination address of a ptp interface.
194 if (flags & IP_SENDONES) {
195 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
196 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
197 ipstat.ips_noroute++;
201 ip->ip_dst.s_addr = INADDR_BROADCAST;
202 dst->sin_addr = ip->ip_dst;
206 } else if (flags & IP_ROUTETOIF) {
207 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
208 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) {
209 ipstat.ips_noroute++;
215 isbroadcast = in_broadcast(dst->sin_addr, ifp);
216 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
217 imo != NULL && imo->imo_multicast_ifp != NULL) {
219 * Bypass the normal routing lookup for multicast
220 * packets if the interface is specified.
222 ifp = imo->imo_multicast_ifp;
224 isbroadcast = 0; /* fool gcc */
227 * We want to do any cloning requested by the link layer,
228 * as this is probably required in all cases for correct
229 * operation (as it is for ARP).
231 if (ro->ro_rt == NULL)
233 rtalloc_mpath_fib(ro,
234 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
235 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
237 in_rtalloc_ign(ro, 0,
238 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
240 if (ro->ro_rt == NULL) {
241 ipstat.ips_noroute++;
242 error = EHOSTUNREACH;
245 ia = ifatoia(ro->ro_rt->rt_ifa);
246 ifp = ro->ro_rt->rt_ifp;
247 ro->ro_rt->rt_rmx.rmx_pksent++;
248 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
249 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
250 if (ro->ro_rt->rt_flags & RTF_HOST)
251 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
253 isbroadcast = in_broadcast(dst->sin_addr, ifp);
256 * Calculate MTU. If we have a route that is up, use that,
257 * otherwise use the interface's MTU.
259 if (ro->ro_rt != NULL && (ro->ro_rt->rt_flags & (RTF_UP|RTF_HOST))) {
261 * This case can happen if the user changed the MTU
262 * of an interface after enabling IP on it. Because
263 * most netifs don't keep track of routes pointing to
264 * them, there is no way for one to update all its
265 * routes when the MTU is changed.
267 if (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)
268 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
269 mtu = ro->ro_rt->rt_rmx.rmx_mtu;
273 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
274 struct in_multi *inm;
276 m->m_flags |= M_MCAST;
278 * IP destination address is multicast. Make sure "dst"
279 * still points to the address in "ro". (It may have been
280 * changed to point to a gateway address, above.)
282 dst = (struct sockaddr_in *)&ro->ro_dst;
284 * See if the caller provided any multicast options
287 ip->ip_ttl = imo->imo_multicast_ttl;
288 if (imo->imo_multicast_vif != -1)
291 ip_mcast_src(imo->imo_multicast_vif) :
294 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
296 * Confirm that the outgoing interface supports multicast.
298 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
299 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
300 ipstat.ips_noroute++;
306 * If source address not specified yet, use address
307 * of outgoing interface.
309 if (ip->ip_src.s_addr == INADDR_ANY) {
310 /* Interface may have no addresses. */
312 ip->ip_src = IA_SIN(ia)->sin_addr;
316 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
318 (imo == NULL || imo->imo_multicast_loop)) {
321 * If we belong to the destination multicast group
322 * on the outgoing interface, and the caller did not
323 * forbid loopback, loop back a copy.
325 ip_mloopback(ifp, m, dst, hlen);
330 * If we are acting as a multicast router, perform
331 * multicast forwarding as if the packet had just
332 * arrived on the interface to which we are about
333 * to send. The multicast forwarding function
334 * recursively calls this function, using the
335 * IP_FORWARDING flag to prevent infinite recursion.
337 * Multicasts that are looped back by ip_mloopback(),
338 * above, will be forwarded by the ip_input() routine,
341 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
343 * If rsvp daemon is not running, do not
344 * set ip_moptions. This ensures that the packet
345 * is multicast and not just sent down one link
346 * as prescribed by rsvpd.
351 ip_mforward(ip, ifp, m, imo) != 0) {
359 * Multicasts with a time-to-live of zero may be looped-
360 * back, above, but must not be transmitted on a network.
361 * Also, multicasts addressed to the loopback interface
362 * are not sent -- the above call to ip_mloopback() will
363 * loop back a copy if this host actually belongs to the
364 * destination group on the loopback interface.
366 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
375 * If the source address is not specified yet, use the address
376 * of the outoing interface.
378 if (ip->ip_src.s_addr == INADDR_ANY) {
379 /* Interface may have no addresses. */
381 ip->ip_src = IA_SIN(ia)->sin_addr;
386 * Verify that we have any chance at all of being able to queue the
387 * packet or packet fragments, unless ALTQ is enabled on the given
388 * interface in which case packetdrop should be done by queueing.
391 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
392 ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >=
393 ifp->if_snd.ifq_maxlen))
395 if ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >=
396 ifp->if_snd.ifq_maxlen)
400 ipstat.ips_odropped++;
401 ifp->if_snd.ifq_drops += (ip->ip_len / ifp->if_mtu + 1);
406 * Look for broadcast address and
407 * verify user is allowed to send
411 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
412 error = EADDRNOTAVAIL;
415 if ((flags & IP_ALLOWBROADCAST) == 0) {
419 /* don't allow broadcast messages to be fragmented */
420 if (ip->ip_len > mtu) {
424 m->m_flags |= M_BCAST;
426 m->m_flags &= ~M_BCAST;
431 switch(ip_ipsec_output(&m, inp, &flags, &error, &ro, &iproute, &dst, &ia, &ifp)) {
438 break; /* Continue with packet processing. */
440 /* Update variables that are affected by ipsec4_output(). */
441 ip = mtod(m, struct ip *);
442 hlen = ip->ip_hl << 2;
445 /* Jump over all PFIL processing if hooks are not active. */
446 if (!PFIL_HOOKED(&inet_pfil_hook))
449 /* Run through list of hooks for output packets. */
450 odst.s_addr = ip->ip_dst.s_addr;
451 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
452 if (error != 0 || m == NULL)
455 ip = mtod(m, struct ip *);
457 /* See if destination IP address was changed by packet filter. */
458 if (odst.s_addr != ip->ip_dst.s_addr) {
459 m->m_flags |= M_SKIP_FIREWALL;
460 /* If destination is now ourself drop to ip_input(). */
461 if (in_localip(ip->ip_dst)) {
462 m->m_flags |= M_FASTFWD_OURS;
463 if (m->m_pkthdr.rcvif == NULL)
464 m->m_pkthdr.rcvif = loif;
465 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
466 m->m_pkthdr.csum_flags |=
467 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
468 m->m_pkthdr.csum_data = 0xffff;
470 m->m_pkthdr.csum_flags |=
471 CSUM_IP_CHECKED | CSUM_IP_VALID;
473 error = netisr_queue(NETISR_IP, m);
476 goto again; /* Redo the routing table lookup. */
479 #ifdef IPFIREWALL_FORWARD
480 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
481 if (m->m_flags & M_FASTFWD_OURS) {
482 if (m->m_pkthdr.rcvif == NULL)
483 m->m_pkthdr.rcvif = loif;
484 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
485 m->m_pkthdr.csum_flags |=
486 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
487 m->m_pkthdr.csum_data = 0xffff;
489 m->m_pkthdr.csum_flags |=
490 CSUM_IP_CHECKED | CSUM_IP_VALID;
492 error = netisr_queue(NETISR_IP, m);
495 /* Or forward to some other address? */
496 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
498 dst = (struct sockaddr_in *)&ro->ro_dst;
499 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
500 m->m_flags |= M_SKIP_FIREWALL;
501 m_tag_delete(m, fwd_tag);
504 #endif /* IPFIREWALL_FORWARD */
507 /* 127/8 must not appear on wire - RFC1122. */
508 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
509 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
510 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
511 ipstat.ips_badaddr++;
512 error = EADDRNOTAVAIL;
517 m->m_pkthdr.csum_flags |= CSUM_IP;
518 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
519 if (sw_csum & CSUM_DELAY_DATA) {
521 sw_csum &= ~CSUM_DELAY_DATA;
523 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
526 * If small enough for interface, or the interface will take
527 * care of the fragmentation for us, we can just send directly.
529 if (ip->ip_len <= mtu ||
530 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
531 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
532 ip->ip_len = htons(ip->ip_len);
533 ip->ip_off = htons(ip->ip_off);
535 if (sw_csum & CSUM_DELAY_IP)
536 ip->ip_sum = in_cksum(m, hlen);
539 * Record statistics for this interface address.
540 * With CSUM_TSO the byte/packet count will be slightly
541 * incorrect because we count the IP+TCP headers only
542 * once instead of for every generated packet.
544 if (!(flags & IP_FORWARDING) && ia) {
545 if (m->m_pkthdr.csum_flags & CSUM_TSO)
546 ia->ia_ifa.if_opackets +=
547 m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
549 ia->ia_ifa.if_opackets++;
550 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
552 #ifdef MBUF_STRESS_TEST
553 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
554 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size);
557 * Reset layer specific mbuf flags
558 * to avoid confusing lower layers.
560 m->m_flags &= ~(M_PROTOFLAGS);
562 error = (*ifp->if_output)(ifp, m,
563 (struct sockaddr *)dst, ro->ro_rt);
567 /* Balk when DF bit is set or the interface didn't support TSO. */
568 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
570 ipstat.ips_cantfrag++;
575 * Too large for interface; fragment if possible. If successful,
576 * on return, m will point to a list of packets to be sent.
578 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum);
585 /* Record statistics for this interface address. */
587 ia->ia_ifa.if_opackets++;
588 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
591 * Reset layer specific mbuf flags
592 * to avoid confusing upper layers.
594 m->m_flags &= ~(M_PROTOFLAGS);
596 error = (*ifp->if_output)(ifp, m,
597 (struct sockaddr *)dst, ro->ro_rt);
603 ipstat.ips_fragmented++;
606 if (ro == &iproute && ro->ro_rt) {
616 * Create a chain of fragments which fit the given mtu. m_frag points to the
617 * mbuf to be fragmented; on return it points to the chain with the fragments.
618 * Return 0 if no error. If error, m_frag may contain a partially built
619 * chain of fragments that should be freed by the caller.
621 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
622 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
625 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
626 u_long if_hwassist_flags, int sw_csum)
629 int hlen = ip->ip_hl << 2;
630 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
632 struct mbuf *m0 = *m_frag; /* the original packet */
637 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
638 ipstat.ips_cantfrag++;
643 * Must be able to put at least 8 bytes per fragment.
649 * If the interface will not calculate checksums on
650 * fragmented packets, then do it here.
652 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
653 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
654 in_delayed_cksum(m0);
655 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
658 if (len > PAGE_SIZE) {
660 * Fragment large datagrams such that each segment
661 * contains a multiple of PAGE_SIZE amount of data,
662 * plus headers. This enables a receiver to perform
663 * page-flipping zero-copy optimizations.
665 * XXX When does this help given that sender and receiver
666 * could have different page sizes, and also mtu could
667 * be less than the receiver's page size ?
672 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
676 * firstlen (off - hlen) must be aligned on an
680 goto smart_frag_failure;
681 off = ((off - hlen) & ~7) + hlen;
682 newlen = (~PAGE_MASK) & mtu;
683 if ((newlen + sizeof (struct ip)) > mtu) {
684 /* we failed, go back the default */
695 firstlen = off - hlen;
696 mnext = &m0->m_nextpkt; /* pointer to next packet */
699 * Loop through length of segment after first fragment,
700 * make new header and copy data of each part and link onto chain.
701 * Here, m0 is the original packet, m is the fragment being created.
702 * The fragments are linked off the m_nextpkt of the original
703 * packet, which after processing serves as the first fragment.
705 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
706 struct ip *mhip; /* ip header on the fragment */
708 int mhlen = sizeof (struct ip);
710 MGETHDR(m, M_DONTWAIT, MT_DATA);
713 ipstat.ips_odropped++;
716 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
718 * In the first mbuf, leave room for the link header, then
719 * copy the original IP header including options. The payload
720 * goes into an additional mbuf chain returned by m_copy().
722 m->m_data += max_linkhdr;
723 mhip = mtod(m, struct ip *);
725 if (hlen > sizeof (struct ip)) {
726 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
727 mhip->ip_v = IPVERSION;
728 mhip->ip_hl = mhlen >> 2;
731 /* XXX do we need to add ip->ip_off below ? */
732 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
733 if (off + len >= ip->ip_len) { /* last fragment */
734 len = ip->ip_len - off;
735 m->m_flags |= M_LASTFRAG;
737 mhip->ip_off |= IP_MF;
738 mhip->ip_len = htons((u_short)(len + mhlen));
739 m->m_next = m_copy(m0, off, len);
740 if (m->m_next == NULL) { /* copy failed */
742 error = ENOBUFS; /* ??? */
743 ipstat.ips_odropped++;
746 m->m_pkthdr.len = mhlen + len;
747 m->m_pkthdr.rcvif = NULL;
749 mac_netinet_fragment(m0, m);
751 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
752 mhip->ip_off = htons(mhip->ip_off);
754 if (sw_csum & CSUM_DELAY_IP)
755 mhip->ip_sum = in_cksum(m, mhlen);
757 mnext = &m->m_nextpkt;
759 ipstat.ips_ofragments += nfrags;
761 /* set first marker for fragment chain */
762 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
763 m0->m_pkthdr.csum_data = nfrags;
766 * Update first fragment by trimming what's been copied out
767 * and updating header.
769 m_adj(m0, hlen + firstlen - ip->ip_len);
770 m0->m_pkthdr.len = hlen + firstlen;
771 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
773 ip->ip_off = htons(ip->ip_off);
775 if (sw_csum & CSUM_DELAY_IP)
776 ip->ip_sum = in_cksum(m0, hlen);
784 in_delayed_cksum(struct mbuf *m)
787 u_short csum, offset;
789 ip = mtod(m, struct ip *);
790 offset = ip->ip_hl << 2 ;
791 csum = in_cksum_skip(m, ip->ip_len, offset);
792 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
794 offset += m->m_pkthdr.csum_data; /* checksum offset */
796 if (offset + sizeof(u_short) > m->m_len) {
797 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
798 m->m_len, offset, ip->ip_p);
801 * this shouldn't happen, but if it does, the
802 * correct behavior may be to insert the checksum
803 * in the appropriate next mbuf in the chain.
807 *(u_short *)(m->m_data + offset) = csum;
811 * IP socket option processing.
814 ip_ctloutput(struct socket *so, struct sockopt *sopt)
816 struct inpcb *inp = sotoinpcb(so);
820 if (sopt->sopt_level != IPPROTO_IP) {
824 switch (sopt->sopt_dir) {
826 switch (sopt->sopt_name) {
833 if (sopt->sopt_valsize > MLEN) {
837 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
842 m->m_len = sopt->sopt_valsize;
843 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
850 error = ip_pcbopts(inp, sopt->sopt_name, m);
866 error = sooptcopyin(sopt, &optval, sizeof optval,
871 switch (sopt->sopt_name) {
873 inp->inp_ip_tos = optval;
877 inp->inp_ip_ttl = optval;
881 if (optval > 0 && optval <= MAXTTL)
882 inp->inp_ip_minttl = optval;
887 #define OPTSET(bit) do { \
890 inp->inp_flags |= bit; \
892 inp->inp_flags &= ~bit; \
897 OPTSET(INP_RECVOPTS);
901 OPTSET(INP_RECVRETOPTS);
905 OPTSET(INP_RECVDSTADDR);
921 OPTSET(INP_ONESBCAST);
924 OPTSET(INP_DONTFRAG);
931 * Multicast socket options are processed by the in_mcast
934 case IP_MULTICAST_IF:
935 case IP_MULTICAST_VIF:
936 case IP_MULTICAST_TTL:
937 case IP_MULTICAST_LOOP:
938 case IP_ADD_MEMBERSHIP:
939 case IP_DROP_MEMBERSHIP:
940 case IP_ADD_SOURCE_MEMBERSHIP:
941 case IP_DROP_SOURCE_MEMBERSHIP:
942 case IP_BLOCK_SOURCE:
943 case IP_UNBLOCK_SOURCE:
945 case MCAST_JOIN_GROUP:
946 case MCAST_LEAVE_GROUP:
947 case MCAST_JOIN_SOURCE_GROUP:
948 case MCAST_LEAVE_SOURCE_GROUP:
949 case MCAST_BLOCK_SOURCE:
950 case MCAST_UNBLOCK_SOURCE:
951 error = inp_setmoptions(inp, sopt);
955 error = sooptcopyin(sopt, &optval, sizeof optval,
962 case IP_PORTRANGE_DEFAULT:
963 inp->inp_flags &= ~(INP_LOWPORT);
964 inp->inp_flags &= ~(INP_HIGHPORT);
967 case IP_PORTRANGE_HIGH:
968 inp->inp_flags &= ~(INP_LOWPORT);
969 inp->inp_flags |= INP_HIGHPORT;
972 case IP_PORTRANGE_LOW:
973 inp->inp_flags &= ~(INP_HIGHPORT);
974 inp->inp_flags |= INP_LOWPORT;
985 case IP_IPSEC_POLICY:
990 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
992 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
994 req = mtod(m, caddr_t);
995 error = ipsec4_set_policy(inp, sopt->sopt_name, req,
996 m->m_len, (sopt->sopt_td != NULL) ?
997 sopt->sopt_td->td_ucred : NULL);
1004 error = ENOPROTOOPT;
1010 switch (sopt->sopt_name) {
1013 if (inp->inp_options)
1014 error = sooptcopyout(sopt,
1015 mtod(inp->inp_options,
1017 inp->inp_options->m_len);
1019 sopt->sopt_valsize = 0;
1026 case IP_RECVRETOPTS:
1027 case IP_RECVDSTADDR:
1034 switch (sopt->sopt_name) {
1037 optval = inp->inp_ip_tos;
1041 optval = inp->inp_ip_ttl;
1045 optval = inp->inp_ip_minttl;
1048 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1051 optval = OPTBIT(INP_RECVOPTS);
1054 case IP_RECVRETOPTS:
1055 optval = OPTBIT(INP_RECVRETOPTS);
1058 case IP_RECVDSTADDR:
1059 optval = OPTBIT(INP_RECVDSTADDR);
1063 optval = OPTBIT(INP_RECVTTL);
1067 optval = OPTBIT(INP_RECVIF);
1071 if (inp->inp_flags & INP_HIGHPORT)
1072 optval = IP_PORTRANGE_HIGH;
1073 else if (inp->inp_flags & INP_LOWPORT)
1074 optval = IP_PORTRANGE_LOW;
1080 optval = OPTBIT(INP_FAITH);
1084 optval = OPTBIT(INP_ONESBCAST);
1087 optval = OPTBIT(INP_DONTFRAG);
1090 error = sooptcopyout(sopt, &optval, sizeof optval);
1094 * Multicast socket options are processed by the in_mcast
1097 case IP_MULTICAST_IF:
1098 case IP_MULTICAST_VIF:
1099 case IP_MULTICAST_TTL:
1100 case IP_MULTICAST_LOOP:
1102 error = inp_getmoptions(inp, sopt);
1106 case IP_IPSEC_POLICY:
1108 struct mbuf *m = NULL;
1113 req = mtod(m, caddr_t);
1116 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m);
1118 error = soopt_mcopyout(sopt, m); /* XXX */
1126 error = ENOPROTOOPT;
1135 * Routine called from ip_output() to loop back a copy of an IP multicast
1136 * packet to the input queue of a specified interface. Note that this
1137 * calls the output routine of the loopback "driver", but with an interface
1138 * pointer that might NOT be a loopback interface -- evil, but easier than
1139 * replicating that code here.
1142 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1145 register struct ip *ip;
1148 copym = m_copy(m, 0, M_COPYALL);
1149 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1150 copym = m_pullup(copym, hlen);
1151 if (copym != NULL) {
1152 /* If needed, compute the checksum and mark it as valid. */
1153 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1154 in_delayed_cksum(copym);
1155 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1156 copym->m_pkthdr.csum_flags |=
1157 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1158 copym->m_pkthdr.csum_data = 0xffff;
1161 * We don't bother to fragment if the IP length is greater
1162 * than the interface's MTU. Can this possibly matter?
1164 ip = mtod(copym, struct ip *);
1165 ip->ip_len = htons(ip->ip_len);
1166 ip->ip_off = htons(ip->ip_off);
1168 ip->ip_sum = in_cksum(copym, hlen);
1171 * It's not clear whether there are any lingering
1172 * reentrancy problems in other areas which might
1173 * be exposed by using ip_input directly (in
1174 * particular, everything which modifies the packet
1175 * in-place). Yet another option is using the
1176 * protosw directly to deliver the looped back
1177 * packet. For the moment, we'll err on the side
1178 * of safety by using if_simloop().
1181 if (dst->sin_family != AF_INET) {
1182 printf("ip_mloopback: bad address family %d\n",
1184 dst->sin_family = AF_INET;
1189 copym->m_pkthdr.rcvif = ifp;
1192 if_simloop(ifp, copym, dst->sin_family, 0);