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 VNET_DEFINE(u_short, ip_id);
89 #ifdef MBUF_STRESS_TEST
90 static int mbuf_frag_size = 0;
91 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
92 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
95 static void ip_mloopback
96 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
99 extern int in_mcast_loop;
100 extern struct protosw inetsw[];
103 * IP output. The packet in mbuf chain m contains a skeletal IP
104 * header (with len, off, ttl, proto, tos, src, dst).
105 * ip_len and ip_off are in host format.
106 * The mbuf chain containing the packet will be freed.
107 * The mbuf opt, if present, will not be freed.
108 * If route ro is present and has ro_rt initialized, route lookup would be
109 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
110 * then result of route lookup is stored in ro->ro_rt.
112 * In the IP forwarding case, the packet will arrive with options already
113 * inserted, so must have a NULL opt pointer.
116 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
117 struct ip_moptions *imo, struct inpcb *inp)
120 struct ifnet *ifp = NULL; /* keep compiler happy */
122 int hlen = sizeof (struct ip);
124 int n; /* scratchpad */
126 struct sockaddr_in *dst;
127 struct in_ifaddr *ia;
128 int isbroadcast, sw_csum;
129 struct route iproute;
130 struct rtentry *rte; /* cache for ro->ro_rt */
132 struct m_tag *fwd_tag = NULL;
134 int no_route_but_check_spd = 0;
139 INP_LOCK_ASSERT(inp);
140 M_SETFIB(m, inp->inp_inc.inc_fibnum);
141 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
142 m->m_pkthdr.flowid = inp->inp_flowid;
143 m->m_flags |= M_FLOWID;
149 bzero(ro, sizeof (*ro));
153 if (ro->ro_rt == NULL) {
157 * The flow table returns route entries valid for up to 30
158 * seconds; we rely on the remainder of ip_output() taking no
159 * longer than that long for the stability of ro_rt. The
160 * flow ID assignment must have happened before this point.
162 fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET);
164 flow_to_route(fle, ro);
170 m = ip_insertoptions(m, opt, &len);
172 hlen = len; /* ip->ip_hl is updated above */
174 ip = mtod(m, struct ip *);
177 * Fill in IP header. If we are not allowing fragmentation,
178 * then the ip_id field is meaningless, but we don't set it
179 * to zero. Doing so causes various problems when devices along
180 * the path (routers, load balancers, firewalls, etc.) illegally
181 * disable DF on our packet. Note that a 16-bit counter
182 * will wrap around in less than 10 seconds at 100 Mbit/s on a
183 * medium with MTU 1500. See Steven M. Bellovin, "A Technique
184 * for Counting NATted Hosts", Proc. IMW'02, available at
185 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
187 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
188 ip->ip_v = IPVERSION;
189 ip->ip_hl = hlen >> 2;
190 ip->ip_id = ip_newid();
191 IPSTAT_INC(ips_localout);
193 /* Header already set, fetch hlen from there */
194 hlen = ip->ip_hl << 2;
198 dst = (struct sockaddr_in *)&ro->ro_dst;
201 * If there is a cached route,
202 * check that it is to the same destination
203 * and is still up. If not, free it and try again.
204 * The address family should also be checked in case of sharing the
208 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
209 rte->rt_ifp == NULL ||
210 !RT_LINK_IS_UP(rte->rt_ifp) ||
211 dst->sin_family != AF_INET ||
212 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
217 if (rte == NULL && fwd_tag == NULL) {
218 bzero(dst, sizeof(*dst));
219 dst->sin_family = AF_INET;
220 dst->sin_len = sizeof(*dst);
221 dst->sin_addr = ip->ip_dst;
224 * If routing to interface only, short circuit routing lookup.
225 * The use of an all-ones broadcast address implies this; an
226 * interface is specified by the broadcast address of an interface,
227 * or the destination address of a ptp interface.
229 if (flags & IP_SENDONES) {
230 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
231 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
232 IPSTAT_INC(ips_noroute);
236 ip->ip_dst.s_addr = INADDR_BROADCAST;
237 dst->sin_addr = ip->ip_dst;
241 } else if (flags & IP_ROUTETOIF) {
242 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
243 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) {
244 IPSTAT_INC(ips_noroute);
250 isbroadcast = in_broadcast(dst->sin_addr, ifp);
251 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
252 imo != NULL && imo->imo_multicast_ifp != NULL) {
254 * Bypass the normal routing lookup for multicast
255 * packets if the interface is specified.
257 ifp = imo->imo_multicast_ifp;
259 isbroadcast = 0; /* fool gcc */
262 * We want to do any cloning requested by the link layer,
263 * as this is probably required in all cases for correct
264 * operation (as it is for ARP).
268 rtalloc_mpath_fib(ro,
269 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
270 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
272 in_rtalloc_ign(ro, 0,
273 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
278 rte->rt_ifp == NULL ||
279 !RT_LINK_IS_UP(rte->rt_ifp)) {
282 * There is no route for this packet, but it is
283 * possible that a matching SPD entry exists.
285 no_route_but_check_spd = 1;
286 mtu = 0; /* Silence GCC warning. */
289 IPSTAT_INC(ips_noroute);
290 error = EHOSTUNREACH;
293 ia = ifatoia(rte->rt_ifa);
294 ifa_ref(&ia->ia_ifa);
296 rte->rt_rmx.rmx_pksent++;
297 if (rte->rt_flags & RTF_GATEWAY)
298 dst = (struct sockaddr_in *)rte->rt_gateway;
299 if (rte->rt_flags & RTF_HOST)
300 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
302 isbroadcast = in_broadcast(dst->sin_addr, ifp);
305 * Calculate MTU. If we have a route that is up, use that,
306 * otherwise use the interface's MTU.
308 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
310 * This case can happen if the user changed the MTU
311 * of an interface after enabling IP on it. Because
312 * most netifs don't keep track of routes pointing to
313 * them, there is no way for one to update all its
314 * routes when the MTU is changed.
316 if (rte->rt_rmx.rmx_mtu > ifp->if_mtu)
317 rte->rt_rmx.rmx_mtu = ifp->if_mtu;
318 mtu = rte->rt_rmx.rmx_mtu;
322 /* Catch a possible divide by zero later. */
323 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
324 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
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.
438 n = ip->ip_len / mtu + 1; /* how many fragments ? */
441 (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
443 (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
445 IPSTAT_INC(ips_odropped);
446 ifp->if_snd.ifq_drops += n;
451 * Look for broadcast address and
452 * verify user is allowed to send
456 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
457 error = EADDRNOTAVAIL;
460 if ((flags & IP_ALLOWBROADCAST) == 0) {
464 /* don't allow broadcast messages to be fragmented */
465 if (ip->ip_len > mtu) {
469 m->m_flags |= M_BCAST;
471 m->m_flags &= ~M_BCAST;
476 switch(ip_ipsec_output(&m, inp, &flags, &error)) {
483 break; /* Continue with packet processing. */
486 * Check if there was a route for this packet; return error if not.
488 if (no_route_but_check_spd) {
489 IPSTAT_INC(ips_noroute);
490 error = EHOSTUNREACH;
493 /* Update variables that are affected by ipsec4_output(). */
494 ip = mtod(m, struct ip *);
495 hlen = ip->ip_hl << 2;
498 /* Jump over all PFIL processing if hooks are not active. */
499 if (!PFIL_HOOKED(&V_inet_pfil_hook))
502 /* Run through list of hooks for output packets. */
503 odst.s_addr = ip->ip_dst.s_addr;
504 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
505 if (error != 0 || m == NULL)
508 ip = mtod(m, struct ip *);
510 /* See if destination IP address was changed by packet filter. */
511 if (odst.s_addr != ip->ip_dst.s_addr) {
512 m->m_flags |= M_SKIP_FIREWALL;
513 /* If destination is now ourself drop to ip_input(). */
514 if (in_localip(ip->ip_dst)) {
515 m->m_flags |= M_FASTFWD_OURS;
516 if (m->m_pkthdr.rcvif == NULL)
517 m->m_pkthdr.rcvif = V_loif;
518 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
519 m->m_pkthdr.csum_flags |=
520 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
521 m->m_pkthdr.csum_data = 0xffff;
523 m->m_pkthdr.csum_flags |=
524 CSUM_IP_CHECKED | CSUM_IP_VALID;
526 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
527 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
529 error = netisr_queue(NETISR_IP, m);
533 ifa_free(&ia->ia_ifa);
534 goto again; /* Redo the routing table lookup. */
538 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
539 if (m->m_flags & M_FASTFWD_OURS) {
540 if (m->m_pkthdr.rcvif == NULL)
541 m->m_pkthdr.rcvif = V_loif;
542 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
543 m->m_pkthdr.csum_flags |=
544 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
545 m->m_pkthdr.csum_data = 0xffff;
548 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
549 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
551 m->m_pkthdr.csum_flags |=
552 CSUM_IP_CHECKED | CSUM_IP_VALID;
554 error = netisr_queue(NETISR_IP, m);
557 /* Or forward to some other address? */
558 if ((m->m_flags & M_IP_NEXTHOP) &&
559 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
560 dst = (struct sockaddr_in *)&ro->ro_dst;
561 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
562 m->m_flags |= M_SKIP_FIREWALL;
563 m->m_flags &= ~M_IP_NEXTHOP;
564 m_tag_delete(m, fwd_tag);
566 ifa_free(&ia->ia_ifa);
571 /* 127/8 must not appear on wire - RFC1122. */
572 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
573 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
574 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
575 IPSTAT_INC(ips_badaddr);
576 error = EADDRNOTAVAIL;
581 m->m_pkthdr.csum_flags |= CSUM_IP;
582 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
583 if (sw_csum & CSUM_DELAY_DATA) {
585 sw_csum &= ~CSUM_DELAY_DATA;
588 if (sw_csum & CSUM_SCTP) {
589 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
590 sw_csum &= ~CSUM_SCTP;
593 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
596 * If small enough for interface, or the interface will take
597 * care of the fragmentation for us, we can just send directly.
599 if (ip->ip_len <= mtu ||
600 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
601 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
602 ip->ip_len = htons(ip->ip_len);
603 ip->ip_off = htons(ip->ip_off);
605 if (sw_csum & CSUM_DELAY_IP)
606 ip->ip_sum = in_cksum(m, hlen);
609 * Record statistics for this interface address.
610 * With CSUM_TSO the byte/packet count will be slightly
611 * incorrect because we count the IP+TCP headers only
612 * once instead of for every generated packet.
614 if (!(flags & IP_FORWARDING) && ia) {
615 if (m->m_pkthdr.csum_flags & CSUM_TSO)
616 ia->ia_ifa.if_opackets +=
617 m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
619 ia->ia_ifa.if_opackets++;
620 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
622 #ifdef MBUF_STRESS_TEST
623 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
624 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size);
627 * Reset layer specific mbuf flags
628 * to avoid confusing lower layers.
630 m->m_flags &= ~(M_PROTOFLAGS);
631 error = (*ifp->if_output)(ifp, m,
632 (struct sockaddr *)dst, ro);
636 /* Balk when DF bit is set or the interface didn't support TSO. */
637 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
639 IPSTAT_INC(ips_cantfrag);
644 * Too large for interface; fragment if possible. If successful,
645 * on return, m will point to a list of packets to be sent.
647 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum);
654 /* Record statistics for this interface address. */
656 ia->ia_ifa.if_opackets++;
657 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
660 * Reset layer specific mbuf flags
661 * to avoid confusing upper layers.
663 m->m_flags &= ~(M_PROTOFLAGS);
665 error = (*ifp->if_output)(ifp, m,
666 (struct sockaddr *)dst, ro);
672 IPSTAT_INC(ips_fragmented);
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 in_delayed_cksum(m0);
724 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
727 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
728 sctp_delayed_cksum(m0, hlen);
729 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
732 if (len > PAGE_SIZE) {
734 * Fragment large datagrams such that each segment
735 * contains a multiple of PAGE_SIZE amount of data,
736 * plus headers. This enables a receiver to perform
737 * page-flipping zero-copy optimizations.
739 * XXX When does this help given that sender and receiver
740 * could have different page sizes, and also mtu could
741 * be less than the receiver's page size ?
745 off = MIN(mtu, m0->m_pkthdr.len);
748 * firstlen (off - hlen) must be aligned on an
752 goto smart_frag_failure;
753 off = ((off - hlen) & ~7) + hlen;
754 newlen = (~PAGE_MASK) & mtu;
755 if ((newlen + sizeof (struct ip)) > mtu) {
756 /* we failed, go back the default */
767 firstlen = off - hlen;
768 mnext = &m0->m_nextpkt; /* pointer to next packet */
771 * Loop through length of segment after first fragment,
772 * make new header and copy data of each part and link onto chain.
773 * Here, m0 is the original packet, m is the fragment being created.
774 * The fragments are linked off the m_nextpkt of the original
775 * packet, which after processing serves as the first fragment.
777 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
778 struct ip *mhip; /* ip header on the fragment */
780 int mhlen = sizeof (struct ip);
782 MGETHDR(m, M_DONTWAIT, MT_DATA);
785 IPSTAT_INC(ips_odropped);
789 * Make sure the complete packet header gets copied
790 * from the originating mbuf to the newly created
791 * mbuf. This also ensures that existing firewall
792 * classification(s), VLAN tags and so on get copied
793 * to the resulting fragmented packet(s):
795 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
798 IPSTAT_INC(ips_odropped);
801 m->m_flags |= M_FRAG;
803 * In the first mbuf, leave room for the link header, then
804 * copy the original IP header including options. The payload
805 * goes into an additional mbuf chain returned by m_copym().
807 m->m_data += max_linkhdr;
808 mhip = mtod(m, struct ip *);
810 if (hlen > sizeof (struct ip)) {
811 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
812 mhip->ip_v = IPVERSION;
813 mhip->ip_hl = mhlen >> 2;
816 /* XXX do we need to add ip->ip_off below ? */
817 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
818 if (off + len >= ip->ip_len) { /* last fragment */
819 len = ip->ip_len - off;
820 m->m_flags |= M_LASTFRAG;
822 mhip->ip_off |= IP_MF;
823 mhip->ip_len = htons((u_short)(len + mhlen));
824 m->m_next = m_copym(m0, off, len, M_DONTWAIT);
825 if (m->m_next == NULL) { /* copy failed */
827 error = ENOBUFS; /* ??? */
828 IPSTAT_INC(ips_odropped);
831 m->m_pkthdr.len = mhlen + len;
833 mac_netinet_fragment(m0, m);
835 mhip->ip_off = htons(mhip->ip_off);
837 if (sw_csum & CSUM_DELAY_IP)
838 mhip->ip_sum = in_cksum(m, mhlen);
840 mnext = &m->m_nextpkt;
842 IPSTAT_ADD(ips_ofragments, nfrags);
844 /* set first marker for fragment chain */
845 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
846 m0->m_pkthdr.csum_data = nfrags;
849 * Update first fragment by trimming what's been copied out
850 * and updating header.
852 m_adj(m0, hlen + firstlen - ip->ip_len);
853 m0->m_pkthdr.len = hlen + firstlen;
854 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
856 ip->ip_off = htons(ip->ip_off);
858 if (sw_csum & CSUM_DELAY_IP)
859 ip->ip_sum = in_cksum(m0, hlen);
867 in_delayed_cksum(struct mbuf *m)
870 u_short csum, offset;
872 ip = mtod(m, struct ip *);
873 offset = ip->ip_hl << 2 ;
874 csum = in_cksum_skip(m, ip->ip_len, offset);
875 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
877 offset += m->m_pkthdr.csum_data; /* checksum offset */
879 /* find the mbuf in the chain where the checksum starts*/
880 while ((m != NULL) && (offset >= m->m_len)) {
884 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
885 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
886 *(u_short *)(m->m_data + offset) = csum;
890 * IP socket option processing.
893 ip_ctloutput(struct socket *so, struct sockopt *sopt)
895 struct inpcb *inp = sotoinpcb(so);
899 if (sopt->sopt_level != IPPROTO_IP) {
902 if (sopt->sopt_level == SOL_SOCKET &&
903 sopt->sopt_dir == SOPT_SET) {
904 switch (sopt->sopt_name) {
907 if ((so->so_options & SO_REUSEADDR) != 0)
908 inp->inp_flags2 |= INP_REUSEADDR;
910 inp->inp_flags2 &= ~INP_REUSEADDR;
916 if ((so->so_options & SO_REUSEPORT) != 0)
917 inp->inp_flags2 |= INP_REUSEPORT;
919 inp->inp_flags2 &= ~INP_REUSEPORT;
925 inp->inp_inc.inc_fibnum = so->so_fibnum;
936 switch (sopt->sopt_dir) {
938 switch (sopt->sopt_name) {
945 if (sopt->sopt_valsize > MLEN) {
949 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
954 m->m_len = sopt->sopt_valsize;
955 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
962 error = ip_pcbopts(inp, sopt->sopt_name, m);
968 if (sopt->sopt_td != NULL) {
969 error = priv_check(sopt->sopt_td,
970 PRIV_NETINET_BINDANY);
987 error = sooptcopyin(sopt, &optval, sizeof optval,
992 switch (sopt->sopt_name) {
994 inp->inp_ip_tos = optval;
998 inp->inp_ip_ttl = optval;
1002 if (optval >= 0 && optval <= MAXTTL)
1003 inp->inp_ip_minttl = optval;
1008 #define OPTSET(bit) do { \
1011 inp->inp_flags |= bit; \
1013 inp->inp_flags &= ~bit; \
1018 OPTSET(INP_RECVOPTS);
1021 case IP_RECVRETOPTS:
1022 OPTSET(INP_RECVRETOPTS);
1025 case IP_RECVDSTADDR:
1026 OPTSET(INP_RECVDSTADDR);
1030 OPTSET(INP_RECVTTL);
1042 OPTSET(INP_ONESBCAST);
1045 OPTSET(INP_DONTFRAG);
1048 OPTSET(INP_BINDANY);
1051 OPTSET(INP_RECVTOS);
1058 * Multicast socket options are processed by the in_mcast
1061 case IP_MULTICAST_IF:
1062 case IP_MULTICAST_VIF:
1063 case IP_MULTICAST_TTL:
1064 case IP_MULTICAST_LOOP:
1065 case IP_ADD_MEMBERSHIP:
1066 case IP_DROP_MEMBERSHIP:
1067 case IP_ADD_SOURCE_MEMBERSHIP:
1068 case IP_DROP_SOURCE_MEMBERSHIP:
1069 case IP_BLOCK_SOURCE:
1070 case IP_UNBLOCK_SOURCE:
1072 case MCAST_JOIN_GROUP:
1073 case MCAST_LEAVE_GROUP:
1074 case MCAST_JOIN_SOURCE_GROUP:
1075 case MCAST_LEAVE_SOURCE_GROUP:
1076 case MCAST_BLOCK_SOURCE:
1077 case MCAST_UNBLOCK_SOURCE:
1078 error = inp_setmoptions(inp, sopt);
1082 error = sooptcopyin(sopt, &optval, sizeof optval,
1089 case IP_PORTRANGE_DEFAULT:
1090 inp->inp_flags &= ~(INP_LOWPORT);
1091 inp->inp_flags &= ~(INP_HIGHPORT);
1094 case IP_PORTRANGE_HIGH:
1095 inp->inp_flags &= ~(INP_LOWPORT);
1096 inp->inp_flags |= INP_HIGHPORT;
1099 case IP_PORTRANGE_LOW:
1100 inp->inp_flags &= ~(INP_HIGHPORT);
1101 inp->inp_flags |= INP_LOWPORT;
1112 case IP_IPSEC_POLICY:
1117 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1119 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1121 req = mtod(m, caddr_t);
1122 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1123 m->m_len, (sopt->sopt_td != NULL) ?
1124 sopt->sopt_td->td_ucred : NULL);
1131 error = ENOPROTOOPT;
1137 switch (sopt->sopt_name) {
1140 if (inp->inp_options)
1141 error = sooptcopyout(sopt,
1142 mtod(inp->inp_options,
1144 inp->inp_options->m_len);
1146 sopt->sopt_valsize = 0;
1153 case IP_RECVRETOPTS:
1154 case IP_RECVDSTADDR:
1163 switch (sopt->sopt_name) {
1166 optval = inp->inp_ip_tos;
1170 optval = inp->inp_ip_ttl;
1174 optval = inp->inp_ip_minttl;
1177 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1180 optval = OPTBIT(INP_RECVOPTS);
1183 case IP_RECVRETOPTS:
1184 optval = OPTBIT(INP_RECVRETOPTS);
1187 case IP_RECVDSTADDR:
1188 optval = OPTBIT(INP_RECVDSTADDR);
1192 optval = OPTBIT(INP_RECVTTL);
1196 optval = OPTBIT(INP_RECVIF);
1200 if (inp->inp_flags & INP_HIGHPORT)
1201 optval = IP_PORTRANGE_HIGH;
1202 else if (inp->inp_flags & INP_LOWPORT)
1203 optval = IP_PORTRANGE_LOW;
1209 optval = OPTBIT(INP_FAITH);
1213 optval = OPTBIT(INP_ONESBCAST);
1216 optval = OPTBIT(INP_DONTFRAG);
1219 optval = OPTBIT(INP_BINDANY);
1222 optval = OPTBIT(INP_RECVTOS);
1225 error = sooptcopyout(sopt, &optval, sizeof optval);
1229 * Multicast socket options are processed by the in_mcast
1232 case IP_MULTICAST_IF:
1233 case IP_MULTICAST_VIF:
1234 case IP_MULTICAST_TTL:
1235 case IP_MULTICAST_LOOP:
1237 error = inp_getmoptions(inp, sopt);
1241 case IP_IPSEC_POLICY:
1243 struct mbuf *m = NULL;
1248 req = mtod(m, caddr_t);
1251 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1253 error = soopt_mcopyout(sopt, m); /* XXX */
1261 error = ENOPROTOOPT;
1270 * Routine called from ip_output() to loop back a copy of an IP multicast
1271 * packet to the input queue of a specified interface. Note that this
1272 * calls the output routine of the loopback "driver", but with an interface
1273 * pointer that might NOT be a loopback interface -- evil, but easier than
1274 * replicating that code here.
1277 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1280 register struct ip *ip;
1284 * Make a deep copy of the packet because we're going to
1285 * modify the pack in order to generate checksums.
1287 copym = m_dup(m, M_DONTWAIT);
1288 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1289 copym = m_pullup(copym, hlen);
1290 if (copym != NULL) {
1291 /* If needed, compute the checksum and mark it as valid. */
1292 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1293 in_delayed_cksum(copym);
1294 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1295 copym->m_pkthdr.csum_flags |=
1296 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1297 copym->m_pkthdr.csum_data = 0xffff;
1300 * We don't bother to fragment if the IP length is greater
1301 * than the interface's MTU. Can this possibly matter?
1303 ip = mtod(copym, struct ip *);
1304 ip->ip_len = htons(ip->ip_len);
1305 ip->ip_off = htons(ip->ip_off);
1307 ip->ip_sum = in_cksum(copym, hlen);
1309 if (dst->sin_family != AF_INET) {
1310 printf("ip_mloopback: bad address family %d\n",
1312 dst->sin_family = AF_INET;
1315 if_simloop(ifp, copym, dst->sin_family, 0);