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 * The mbuf chain containing the packet will be freed.
106 * The mbuf opt, if present, will not be freed.
107 * If route ro is present and has ro_rt initialized, route lookup would be
108 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
109 * then result of route lookup is stored in ro->ro_rt.
111 * In the IP forwarding case, the packet will arrive with options already
112 * inserted, so must have a NULL opt pointer.
115 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
116 struct ip_moptions *imo, struct inpcb *inp)
119 struct ifnet *ifp = NULL; /* keep compiler happy */
121 int hlen = sizeof (struct ip);
123 int n; /* scratchpad */
125 struct sockaddr_in *dst;
126 const struct sockaddr_in *gw;
127 struct in_ifaddr *ia;
129 uint16_t ip_len, ip_off;
130 struct route iproute;
131 struct rtentry *rte; /* cache for ro->ro_rt */
133 struct m_tag *fwd_tag = NULL;
135 int no_route_but_check_spd = 0;
140 INP_LOCK_ASSERT(inp);
141 M_SETFIB(m, inp->inp_inc.inc_fibnum);
142 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
143 m->m_pkthdr.flowid = inp->inp_flowid;
144 m->m_flags |= M_FLOWID;
150 bzero(ro, sizeof (*ro));
154 if (ro->ro_rt == NULL) {
158 * The flow table returns route entries valid for up to 30
159 * seconds; we rely on the remainder of ip_output() taking no
160 * longer than that long for the stability of ro_rt. The
161 * flow ID assignment must have happened before this point.
163 fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET);
165 flow_to_route(fle, ro);
171 m = ip_insertoptions(m, opt, &len);
173 hlen = len; /* ip->ip_hl is updated above */
175 ip = mtod(m, struct ip *);
176 ip_len = ntohs(ip->ip_len);
177 ip_off = ntohs(ip->ip_off);
180 * Fill in IP header. If we are not allowing fragmentation,
181 * then the ip_id field is meaningless, but we don't set it
182 * to zero. Doing so causes various problems when devices along
183 * the path (routers, load balancers, firewalls, etc.) illegally
184 * disable DF on our packet. Note that a 16-bit counter
185 * will wrap around in less than 10 seconds at 100 Mbit/s on a
186 * medium with MTU 1500. See Steven M. Bellovin, "A Technique
187 * for Counting NATted Hosts", Proc. IMW'02, available at
188 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
190 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
191 ip->ip_v = IPVERSION;
192 ip->ip_hl = hlen >> 2;
193 ip->ip_id = ip_newid();
194 IPSTAT_INC(ips_localout);
196 /* Header already set, fetch hlen from there */
197 hlen = ip->ip_hl << 2;
200 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
204 * If there is a cached route,
205 * check that it is to the same destination
206 * and is still up. If not, free it and try again.
207 * The address family should also be checked in case of sharing the
211 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
212 rte->rt_ifp == NULL ||
213 !RT_LINK_IS_UP(rte->rt_ifp) ||
214 dst->sin_family != AF_INET ||
215 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
220 if (rte == NULL && fwd_tag == NULL) {
221 bzero(dst, sizeof(*dst));
222 dst->sin_family = AF_INET;
223 dst->sin_len = sizeof(*dst);
224 dst->sin_addr = ip->ip_dst;
227 * If routing to interface only, short circuit routing lookup.
228 * The use of an all-ones broadcast address implies this; an
229 * interface is specified by the broadcast address of an interface,
230 * or the destination address of a ptp interface.
232 if (flags & IP_SENDONES) {
233 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
234 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
235 IPSTAT_INC(ips_noroute);
239 ip->ip_dst.s_addr = INADDR_BROADCAST;
240 dst->sin_addr = ip->ip_dst;
244 } else if (flags & IP_ROUTETOIF) {
245 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
246 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) {
247 IPSTAT_INC(ips_noroute);
253 isbroadcast = in_broadcast(dst->sin_addr, ifp);
254 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
255 imo != NULL && imo->imo_multicast_ifp != NULL) {
257 * Bypass the normal routing lookup for multicast
258 * packets if the interface is specified.
260 ifp = imo->imo_multicast_ifp;
262 isbroadcast = 0; /* fool gcc */
265 * We want to do any cloning requested by the link layer,
266 * as this is probably required in all cases for correct
267 * operation (as it is for ARP).
271 rtalloc_mpath_fib(ro,
272 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
273 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
275 in_rtalloc_ign(ro, 0,
276 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
281 rte->rt_ifp == NULL ||
282 !RT_LINK_IS_UP(rte->rt_ifp)) {
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 gw = (struct sockaddr_in *)rte->rt_gateway;
302 if (rte->rt_flags & RTF_HOST)
303 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
305 isbroadcast = in_broadcast(gw->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 /* Catch a possible divide by zero later. */
326 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
327 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
328 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
329 m->m_flags |= M_MCAST;
331 * See if the caller provided any multicast options
334 ip->ip_ttl = imo->imo_multicast_ttl;
335 if (imo->imo_multicast_vif != -1)
338 ip_mcast_src(imo->imo_multicast_vif) :
341 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
343 * Confirm that the outgoing interface supports multicast.
345 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
346 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
347 IPSTAT_INC(ips_noroute);
353 * If source address not specified yet, use address
354 * of outgoing interface.
356 if (ip->ip_src.s_addr == INADDR_ANY) {
357 /* Interface may have no addresses. */
359 ip->ip_src = IA_SIN(ia)->sin_addr;
362 if ((imo == NULL && in_mcast_loop) ||
363 (imo && imo->imo_multicast_loop)) {
365 * Loop back multicast datagram if not expressly
366 * forbidden to do so, even if we are not a member
367 * of the group; ip_input() will filter it later,
368 * thus deferring a hash lookup and mutex acquisition
369 * at the expense of a cheap copy using m_copym().
371 ip_mloopback(ifp, m, dst, hlen);
374 * If we are acting as a multicast router, perform
375 * multicast forwarding as if the packet had just
376 * arrived on the interface to which we are about
377 * to send. The multicast forwarding function
378 * recursively calls this function, using the
379 * IP_FORWARDING flag to prevent infinite recursion.
381 * Multicasts that are looped back by ip_mloopback(),
382 * above, will be forwarded by the ip_input() routine,
385 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
387 * If rsvp daemon is not running, do not
388 * set ip_moptions. This ensures that the packet
389 * is multicast and not just sent down one link
390 * as prescribed by rsvpd.
395 ip_mforward(ip, ifp, m, imo) != 0) {
403 * Multicasts with a time-to-live of zero may be looped-
404 * back, above, but must not be transmitted on a network.
405 * Also, multicasts addressed to the loopback interface
406 * are not sent -- the above call to ip_mloopback() will
407 * loop back a copy. ip_input() will drop the copy if
408 * this host does not belong to the destination group on
409 * the loopback interface.
411 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
420 * If the source address is not specified yet, use the address
421 * of the outoing interface.
423 if (ip->ip_src.s_addr == INADDR_ANY) {
424 /* Interface may have no addresses. */
426 ip->ip_src = IA_SIN(ia)->sin_addr;
431 * Verify that we have any chance at all of being able to queue the
432 * packet or packet fragments, unless ALTQ is enabled on the given
433 * interface in which case packetdrop should be done by queueing.
435 n = ip_len / mtu + 1; /* how many fragments ? */
438 (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
440 (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
442 IPSTAT_INC(ips_odropped);
443 ifp->if_snd.ifq_drops += n;
448 * Look for broadcast address and
449 * verify user is allowed to send
453 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
454 error = EADDRNOTAVAIL;
457 if ((flags & IP_ALLOWBROADCAST) == 0) {
461 /* don't allow broadcast messages to be fragmented */
466 m->m_flags |= M_BCAST;
468 m->m_flags &= ~M_BCAST;
473 switch(ip_ipsec_output(&m, inp, &flags, &error)) {
480 break; /* Continue with packet processing. */
483 * Check if there was a route for this packet; return error if not.
485 if (no_route_but_check_spd) {
486 IPSTAT_INC(ips_noroute);
487 error = EHOSTUNREACH;
490 /* Update variables that are affected by ipsec4_output(). */
491 ip = mtod(m, struct ip *);
492 hlen = ip->ip_hl << 2;
495 /* Jump over all PFIL processing if hooks are not active. */
496 if (!PFIL_HOOKED(&V_inet_pfil_hook))
499 /* Run through list of hooks for output packets. */
500 odst.s_addr = ip->ip_dst.s_addr;
501 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
502 if (error != 0 || m == NULL)
505 ip = mtod(m, struct ip *);
507 /* See if destination IP address was changed by packet filter. */
508 if (odst.s_addr != ip->ip_dst.s_addr) {
509 m->m_flags |= M_SKIP_FIREWALL;
510 /* If destination is now ourself drop to ip_input(). */
511 if (in_localip(ip->ip_dst)) {
512 m->m_flags |= M_FASTFWD_OURS;
513 if (m->m_pkthdr.rcvif == NULL)
514 m->m_pkthdr.rcvif = V_loif;
515 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
516 m->m_pkthdr.csum_flags |=
517 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
518 m->m_pkthdr.csum_data = 0xffff;
520 m->m_pkthdr.csum_flags |=
521 CSUM_IP_CHECKED | CSUM_IP_VALID;
523 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
524 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
526 error = netisr_queue(NETISR_IP, m);
530 ifa_free(&ia->ia_ifa);
531 goto again; /* Redo the routing table lookup. */
535 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
536 if (m->m_flags & M_FASTFWD_OURS) {
537 if (m->m_pkthdr.rcvif == NULL)
538 m->m_pkthdr.rcvif = V_loif;
539 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
540 m->m_pkthdr.csum_flags |=
541 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
542 m->m_pkthdr.csum_data = 0xffff;
545 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
546 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
548 m->m_pkthdr.csum_flags |=
549 CSUM_IP_CHECKED | CSUM_IP_VALID;
551 error = netisr_queue(NETISR_IP, m);
554 /* Or forward to some other address? */
555 if ((m->m_flags & M_IP_NEXTHOP) &&
556 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
557 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
558 m->m_flags |= M_SKIP_FIREWALL;
559 m->m_flags &= ~M_IP_NEXTHOP;
560 m_tag_delete(m, fwd_tag);
562 ifa_free(&ia->ia_ifa);
567 /* 127/8 must not appear on wire - RFC1122. */
568 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
569 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
570 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
571 IPSTAT_INC(ips_badaddr);
572 error = EADDRNOTAVAIL;
577 m->m_pkthdr.csum_flags |= CSUM_IP;
578 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
580 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
583 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
584 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
585 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
590 * If small enough for interface, or the interface will take
591 * care of the fragmentation for us, we can just send directly.
594 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
595 ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
597 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
598 ip->ip_sum = in_cksum(m, hlen);
599 m->m_pkthdr.csum_flags &= ~CSUM_IP;
603 * Record statistics for this interface address.
604 * With CSUM_TSO the byte/packet count will be slightly
605 * incorrect because we count the IP+TCP headers only
606 * once instead of for every generated packet.
608 if (!(flags & IP_FORWARDING) && ia) {
609 if (m->m_pkthdr.csum_flags & CSUM_TSO)
610 ia->ia_ifa.if_opackets +=
611 m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
613 ia->ia_ifa.if_opackets++;
614 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
616 #ifdef MBUF_STRESS_TEST
617 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
618 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
621 * Reset layer specific mbuf flags
622 * to avoid confusing lower layers.
625 error = (*ifp->if_output)(ifp, m,
626 (const struct sockaddr *)gw, ro);
630 /* Balk when DF bit is set or the interface didn't support TSO. */
631 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
633 IPSTAT_INC(ips_cantfrag);
638 * Too large for interface; fragment if possible. If successful,
639 * on return, m will point to a list of packets to be sent.
641 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
648 /* Record statistics for this interface address. */
650 ia->ia_ifa.if_opackets++;
651 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
654 * Reset layer specific mbuf flags
655 * to avoid confusing upper layers.
659 error = (*ifp->if_output)(ifp, m,
660 (const struct sockaddr *)gw, ro);
666 IPSTAT_INC(ips_fragmented);
672 ifa_free(&ia->ia_ifa);
680 * Create a chain of fragments which fit the given mtu. m_frag points to the
681 * mbuf to be fragmented; on return it points to the chain with the fragments.
682 * Return 0 if no error. If error, m_frag may contain a partially built
683 * chain of fragments that should be freed by the caller.
685 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
688 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
689 u_long if_hwassist_flags)
692 int hlen = ip->ip_hl << 2;
693 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
695 struct mbuf *m0 = *m_frag; /* the original packet */
699 uint16_t ip_len, ip_off;
701 ip_len = ntohs(ip->ip_len);
702 ip_off = ntohs(ip->ip_off);
704 if (ip_off & IP_DF) { /* Fragmentation not allowed */
705 IPSTAT_INC(ips_cantfrag);
710 * Must be able to put at least 8 bytes per fragment.
716 * If the interface will not calculate checksums on
717 * fragmented packets, then do it here.
719 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
720 in_delayed_cksum(m0);
721 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
724 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
725 sctp_delayed_cksum(m0, hlen);
726 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
729 if (len > PAGE_SIZE) {
731 * Fragment large datagrams such that each segment
732 * contains a multiple of PAGE_SIZE amount of data,
733 * plus headers. This enables a receiver to perform
734 * page-flipping zero-copy optimizations.
736 * XXX When does this help given that sender and receiver
737 * could have different page sizes, and also mtu could
738 * be less than the receiver's page size ?
743 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
747 * firstlen (off - hlen) must be aligned on an
751 goto smart_frag_failure;
752 off = ((off - hlen) & ~7) + hlen;
753 newlen = (~PAGE_MASK) & mtu;
754 if ((newlen + sizeof (struct ip)) > mtu) {
755 /* we failed, go back the default */
766 firstlen = off - hlen;
767 mnext = &m0->m_nextpkt; /* pointer to next packet */
770 * Loop through length of segment after first fragment,
771 * make new header and copy data of each part and link onto chain.
772 * Here, m0 is the original packet, m is the fragment being created.
773 * The fragments are linked off the m_nextpkt of the original
774 * packet, which after processing serves as the first fragment.
776 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
777 struct ip *mhip; /* ip header on the fragment */
779 int mhlen = sizeof (struct ip);
781 m = m_gethdr(M_NOWAIT, MT_DATA);
784 IPSTAT_INC(ips_odropped);
787 m->m_flags |= (m0->m_flags & M_MCAST);
789 * In the first mbuf, leave room for the link header, then
790 * copy the original IP header including options. The payload
791 * goes into an additional mbuf chain returned by m_copym().
793 m->m_data += max_linkhdr;
794 mhip = mtod(m, struct ip *);
796 if (hlen > sizeof (struct ip)) {
797 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
798 mhip->ip_v = IPVERSION;
799 mhip->ip_hl = mhlen >> 2;
802 /* XXX do we need to add ip_off below ? */
803 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
804 if (off + len >= ip_len)
807 mhip->ip_off |= IP_MF;
808 mhip->ip_len = htons((u_short)(len + mhlen));
809 m->m_next = m_copym(m0, off, len, M_NOWAIT);
810 if (m->m_next == NULL) { /* copy failed */
812 error = ENOBUFS; /* ??? */
813 IPSTAT_INC(ips_odropped);
816 m->m_pkthdr.len = mhlen + len;
817 m->m_pkthdr.rcvif = NULL;
819 mac_netinet_fragment(m0, m);
821 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
822 mhip->ip_off = htons(mhip->ip_off);
824 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
825 mhip->ip_sum = in_cksum(m, mhlen);
826 m->m_pkthdr.csum_flags &= ~CSUM_IP;
829 mnext = &m->m_nextpkt;
831 IPSTAT_ADD(ips_ofragments, nfrags);
834 * Update first fragment by trimming what's been copied out
835 * and updating header.
837 m_adj(m0, hlen + firstlen - ip_len);
838 m0->m_pkthdr.len = hlen + firstlen;
839 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
840 ip->ip_off = htons(ip_off | IP_MF);
842 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
843 ip->ip_sum = in_cksum(m0, hlen);
844 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
853 in_delayed_cksum(struct mbuf *m)
856 uint16_t csum, offset, ip_len;
858 ip = mtod(m, struct ip *);
859 offset = ip->ip_hl << 2 ;
860 ip_len = ntohs(ip->ip_len);
861 csum = in_cksum_skip(m, ip_len, offset);
862 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
864 offset += m->m_pkthdr.csum_data; /* checksum offset */
866 if (offset + sizeof(u_short) > m->m_len) {
867 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
868 m->m_len, offset, ip->ip_p);
871 * this shouldn't happen, but if it does, the
872 * correct behavior may be to insert the checksum
873 * in the appropriate next mbuf in the chain.
877 *(u_short *)(m->m_data + offset) = csum;
881 * IP socket option processing.
884 ip_ctloutput(struct socket *so, struct sockopt *sopt)
886 struct inpcb *inp = sotoinpcb(so);
890 if (sopt->sopt_level != IPPROTO_IP) {
893 if (sopt->sopt_level == SOL_SOCKET &&
894 sopt->sopt_dir == SOPT_SET) {
895 switch (sopt->sopt_name) {
898 if ((so->so_options & SO_REUSEADDR) != 0)
899 inp->inp_flags2 |= INP_REUSEADDR;
901 inp->inp_flags2 &= ~INP_REUSEADDR;
907 if ((so->so_options & SO_REUSEPORT) != 0)
908 inp->inp_flags2 |= INP_REUSEPORT;
910 inp->inp_flags2 &= ~INP_REUSEPORT;
916 inp->inp_inc.inc_fibnum = so->so_fibnum;
927 switch (sopt->sopt_dir) {
929 switch (sopt->sopt_name) {
936 if (sopt->sopt_valsize > MLEN) {
940 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
945 m->m_len = sopt->sopt_valsize;
946 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
953 error = ip_pcbopts(inp, sopt->sopt_name, m);
959 if (sopt->sopt_td != NULL) {
960 error = priv_check(sopt->sopt_td,
961 PRIV_NETINET_BINDANY);
978 error = sooptcopyin(sopt, &optval, sizeof optval,
983 switch (sopt->sopt_name) {
985 inp->inp_ip_tos = optval;
989 inp->inp_ip_ttl = optval;
993 if (optval >= 0 && optval <= MAXTTL)
994 inp->inp_ip_minttl = optval;
999 #define OPTSET(bit) do { \
1002 inp->inp_flags |= bit; \
1004 inp->inp_flags &= ~bit; \
1009 OPTSET(INP_RECVOPTS);
1012 case IP_RECVRETOPTS:
1013 OPTSET(INP_RECVRETOPTS);
1016 case IP_RECVDSTADDR:
1017 OPTSET(INP_RECVDSTADDR);
1021 OPTSET(INP_RECVTTL);
1033 OPTSET(INP_ONESBCAST);
1036 OPTSET(INP_DONTFRAG);
1039 OPTSET(INP_BINDANY);
1042 OPTSET(INP_RECVTOS);
1049 * Multicast socket options are processed by the in_mcast
1052 case IP_MULTICAST_IF:
1053 case IP_MULTICAST_VIF:
1054 case IP_MULTICAST_TTL:
1055 case IP_MULTICAST_LOOP:
1056 case IP_ADD_MEMBERSHIP:
1057 case IP_DROP_MEMBERSHIP:
1058 case IP_ADD_SOURCE_MEMBERSHIP:
1059 case IP_DROP_SOURCE_MEMBERSHIP:
1060 case IP_BLOCK_SOURCE:
1061 case IP_UNBLOCK_SOURCE:
1063 case MCAST_JOIN_GROUP:
1064 case MCAST_LEAVE_GROUP:
1065 case MCAST_JOIN_SOURCE_GROUP:
1066 case MCAST_LEAVE_SOURCE_GROUP:
1067 case MCAST_BLOCK_SOURCE:
1068 case MCAST_UNBLOCK_SOURCE:
1069 error = inp_setmoptions(inp, sopt);
1073 error = sooptcopyin(sopt, &optval, sizeof optval,
1080 case IP_PORTRANGE_DEFAULT:
1081 inp->inp_flags &= ~(INP_LOWPORT);
1082 inp->inp_flags &= ~(INP_HIGHPORT);
1085 case IP_PORTRANGE_HIGH:
1086 inp->inp_flags &= ~(INP_LOWPORT);
1087 inp->inp_flags |= INP_HIGHPORT;
1090 case IP_PORTRANGE_LOW:
1091 inp->inp_flags &= ~(INP_HIGHPORT);
1092 inp->inp_flags |= INP_LOWPORT;
1103 case IP_IPSEC_POLICY:
1108 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1110 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1112 req = mtod(m, caddr_t);
1113 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1114 m->m_len, (sopt->sopt_td != NULL) ?
1115 sopt->sopt_td->td_ucred : NULL);
1122 error = ENOPROTOOPT;
1128 switch (sopt->sopt_name) {
1131 if (inp->inp_options)
1132 error = sooptcopyout(sopt,
1133 mtod(inp->inp_options,
1135 inp->inp_options->m_len);
1137 sopt->sopt_valsize = 0;
1144 case IP_RECVRETOPTS:
1145 case IP_RECVDSTADDR:
1154 switch (sopt->sopt_name) {
1157 optval = inp->inp_ip_tos;
1161 optval = inp->inp_ip_ttl;
1165 optval = inp->inp_ip_minttl;
1168 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1171 optval = OPTBIT(INP_RECVOPTS);
1174 case IP_RECVRETOPTS:
1175 optval = OPTBIT(INP_RECVRETOPTS);
1178 case IP_RECVDSTADDR:
1179 optval = OPTBIT(INP_RECVDSTADDR);
1183 optval = OPTBIT(INP_RECVTTL);
1187 optval = OPTBIT(INP_RECVIF);
1191 if (inp->inp_flags & INP_HIGHPORT)
1192 optval = IP_PORTRANGE_HIGH;
1193 else if (inp->inp_flags & INP_LOWPORT)
1194 optval = IP_PORTRANGE_LOW;
1200 optval = OPTBIT(INP_FAITH);
1204 optval = OPTBIT(INP_ONESBCAST);
1207 optval = OPTBIT(INP_DONTFRAG);
1210 optval = OPTBIT(INP_BINDANY);
1213 optval = OPTBIT(INP_RECVTOS);
1216 error = sooptcopyout(sopt, &optval, sizeof optval);
1220 * Multicast socket options are processed by the in_mcast
1223 case IP_MULTICAST_IF:
1224 case IP_MULTICAST_VIF:
1225 case IP_MULTICAST_TTL:
1226 case IP_MULTICAST_LOOP:
1228 error = inp_getmoptions(inp, sopt);
1232 case IP_IPSEC_POLICY:
1234 struct mbuf *m = NULL;
1239 req = mtod(m, caddr_t);
1242 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1244 error = soopt_mcopyout(sopt, m); /* XXX */
1252 error = ENOPROTOOPT;
1261 * Routine called from ip_output() to loop back a copy of an IP multicast
1262 * packet to the input queue of a specified interface. Note that this
1263 * calls the output routine of the loopback "driver", but with an interface
1264 * pointer that might NOT be a loopback interface -- evil, but easier than
1265 * replicating that code here.
1268 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1271 register struct ip *ip;
1275 * Make a deep copy of the packet because we're going to
1276 * modify the pack in order to generate checksums.
1278 copym = m_dup(m, M_NOWAIT);
1279 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1280 copym = m_pullup(copym, hlen);
1281 if (copym != NULL) {
1282 /* If needed, compute the checksum and mark it as valid. */
1283 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1284 in_delayed_cksum(copym);
1285 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1286 copym->m_pkthdr.csum_flags |=
1287 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1288 copym->m_pkthdr.csum_data = 0xffff;
1291 * We don't bother to fragment if the IP length is greater
1292 * than the interface's MTU. Can this possibly matter?
1294 ip = mtod(copym, struct ip *);
1296 ip->ip_sum = in_cksum(copym, hlen);
1298 if (dst->sin_family != AF_INET) {
1299 printf("ip_mloopback: bad address family %d\n",
1301 dst->sin_family = AF_INET;
1304 if_simloop(ifp, copym, dst->sin_family, 0);