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"
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/kernel.h>
43 #include <sys/malloc.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/sysctl.h>
51 #include <sys/ucred.h>
54 #include <net/netisr.h>
56 #include <net/route.h>
58 #include <netinet/in.h>
59 #include <netinet/in_systm.h>
60 #include <netinet/ip.h>
61 #include <netinet/in_pcb.h>
62 #include <netinet/in_var.h>
63 #include <netinet/ip_var.h>
64 #include <netinet/ip_options.h>
67 #include <netinet/ip_ipsec.h>
68 #include <netipsec/ipsec.h>
71 #include <machine/in_cksum.h>
73 #include <security/mac/mac_framework.h>
75 #define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\
76 x, (ntohl(a.s_addr)>>24)&0xFF,\
77 (ntohl(a.s_addr)>>16)&0xFF,\
78 (ntohl(a.s_addr)>>8)&0xFF,\
79 (ntohl(a.s_addr))&0xFF, y);
83 #ifdef MBUF_STRESS_TEST
84 int mbuf_frag_size = 0;
85 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
86 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
89 static void ip_mloopback
90 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
93 extern struct protosw inetsw[];
96 * IP output. The packet in mbuf chain m contains a skeletal IP
97 * header (with len, off, ttl, proto, tos, src, dst).
98 * The mbuf chain containing the packet will be freed.
99 * The mbuf opt, if present, will not be freed.
100 * In the IP forwarding case, the packet will arrive with options already
101 * inserted, so must have a NULL opt pointer.
104 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
105 struct ip_moptions *imo, struct inpcb *inp)
108 struct ifnet *ifp = NULL; /* keep compiler happy */
110 int hlen = sizeof (struct ip);
113 struct sockaddr_in *dst = NULL; /* keep compiler happy */
114 struct in_ifaddr *ia = NULL;
115 int isbroadcast, sw_csum;
116 struct route iproute;
118 #ifdef IPFIREWALL_FORWARD
119 struct m_tag *fwd_tag = NULL;
125 bzero(ro, sizeof (*ro));
129 INP_LOCK_ASSERT(inp);
133 m = ip_insertoptions(m, opt, &len);
137 ip = mtod(m, struct ip *);
140 * Fill in IP header. If we are not allowing fragmentation,
141 * then the ip_id field is meaningless, but we don't set it
142 * to zero. Doing so causes various problems when devices along
143 * the path (routers, load balancers, firewalls, etc.) illegally
144 * disable DF on our packet. Note that a 16-bit counter
145 * will wrap around in less than 10 seconds at 100 Mbit/s on a
146 * medium with MTU 1500. See Steven M. Bellovin, "A Technique
147 * for Counting NATted Hosts", Proc. IMW'02, available at
148 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
150 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
151 ip->ip_v = IPVERSION;
152 ip->ip_hl = hlen >> 2;
153 ip->ip_id = ip_newid();
154 ipstat.ips_localout++;
156 hlen = ip->ip_hl << 2;
159 dst = (struct sockaddr_in *)&ro->ro_dst;
162 * If there is a cached route,
163 * check that it is to the same destination
164 * and is still up. If not, free it and try again.
165 * The address family should also be checked in case of sharing the
168 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
169 dst->sin_family != AF_INET ||
170 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
172 ro->ro_rt = (struct rtentry *)NULL;
174 #ifdef IPFIREWALL_FORWARD
175 if (ro->ro_rt == NULL && fwd_tag == NULL) {
177 if (ro->ro_rt == NULL) {
179 bzero(dst, sizeof(*dst));
180 dst->sin_family = AF_INET;
181 dst->sin_len = sizeof(*dst);
182 dst->sin_addr = ip->ip_dst;
185 * If routing to interface only, short circuit routing lookup.
186 * The use of an all-ones broadcast address implies this; an
187 * interface is specified by the broadcast address of an interface,
188 * or the destination address of a ptp interface.
190 if (flags & IP_SENDONES) {
191 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
192 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
193 ipstat.ips_noroute++;
197 ip->ip_dst.s_addr = INADDR_BROADCAST;
198 dst->sin_addr = ip->ip_dst;
202 } else if (flags & IP_ROUTETOIF) {
203 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
204 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) {
205 ipstat.ips_noroute++;
211 isbroadcast = in_broadcast(dst->sin_addr, ifp);
212 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
213 imo != NULL && imo->imo_multicast_ifp != NULL) {
215 * Bypass the normal routing lookup for multicast
216 * packets if the interface is specified.
218 ifp = imo->imo_multicast_ifp;
220 isbroadcast = 0; /* fool gcc */
223 * We want to do any cloning requested by the link layer,
224 * as this is probably required in all cases for correct
225 * operation (as it is for ARP).
227 if (ro->ro_rt == NULL)
229 if (ro->ro_rt == NULL) {
230 ipstat.ips_noroute++;
231 error = EHOSTUNREACH;
234 ia = ifatoia(ro->ro_rt->rt_ifa);
235 ifp = ro->ro_rt->rt_ifp;
236 ro->ro_rt->rt_rmx.rmx_pksent++;
237 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
238 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
239 if (ro->ro_rt->rt_flags & RTF_HOST)
240 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
242 isbroadcast = in_broadcast(dst->sin_addr, ifp);
245 * Calculate MTU. If we have a route that is up, use that,
246 * otherwise use the interface's MTU.
248 if (ro->ro_rt != NULL && (ro->ro_rt->rt_flags & (RTF_UP|RTF_HOST))) {
250 * This case can happen if the user changed the MTU
251 * of an interface after enabling IP on it. Because
252 * most netifs don't keep track of routes pointing to
253 * them, there is no way for one to update all its
254 * routes when the MTU is changed.
256 if (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)
257 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
258 mtu = ro->ro_rt->rt_rmx.rmx_mtu;
262 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
263 struct in_multi *inm;
265 m->m_flags |= M_MCAST;
267 * IP destination address is multicast. Make sure "dst"
268 * still points to the address in "ro". (It may have been
269 * changed to point to a gateway address, above.)
271 dst = (struct sockaddr_in *)&ro->ro_dst;
273 * See if the caller provided any multicast options
276 ip->ip_ttl = imo->imo_multicast_ttl;
277 if (imo->imo_multicast_vif != -1)
280 ip_mcast_src(imo->imo_multicast_vif) :
283 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
285 * Confirm that the outgoing interface supports multicast.
287 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
288 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
289 ipstat.ips_noroute++;
295 * If source address not specified yet, use address
296 * of outgoing interface.
298 if (ip->ip_src.s_addr == INADDR_ANY) {
299 /* Interface may have no addresses. */
301 ip->ip_src = IA_SIN(ia)->sin_addr;
305 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
307 (imo == NULL || imo->imo_multicast_loop)) {
310 * If we belong to the destination multicast group
311 * on the outgoing interface, and the caller did not
312 * forbid loopback, loop back a copy.
314 ip_mloopback(ifp, m, dst, hlen);
319 * If we are acting as a multicast router, perform
320 * multicast forwarding as if the packet had just
321 * arrived on the interface to which we are about
322 * to send. The multicast forwarding function
323 * recursively calls this function, using the
324 * IP_FORWARDING flag to prevent infinite recursion.
326 * Multicasts that are looped back by ip_mloopback(),
327 * above, will be forwarded by the ip_input() routine,
330 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
332 * If rsvp daemon is not running, do not
333 * set ip_moptions. This ensures that the packet
334 * is multicast and not just sent down one link
335 * as prescribed by rsvpd.
340 ip_mforward(ip, ifp, m, imo) != 0) {
348 * Multicasts with a time-to-live of zero may be looped-
349 * back, above, but must not be transmitted on a network.
350 * Also, multicasts addressed to the loopback interface
351 * are not sent -- the above call to ip_mloopback() will
352 * loop back a copy if this host actually belongs to the
353 * destination group on the loopback interface.
355 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
364 * If the source address is not specified yet, use the address
365 * of the outoing interface.
367 if (ip->ip_src.s_addr == INADDR_ANY) {
368 /* Interface may have no addresses. */
370 ip->ip_src = IA_SIN(ia)->sin_addr;
375 * Verify that we have any chance at all of being able to queue the
376 * packet or packet fragments, unless ALTQ is enabled on the given
377 * interface in which case packetdrop should be done by queueing.
380 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
381 ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >=
382 ifp->if_snd.ifq_maxlen))
384 if ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >=
385 ifp->if_snd.ifq_maxlen)
389 ipstat.ips_odropped++;
390 ifp->if_snd.ifq_drops += (ip->ip_len / ifp->if_mtu + 1);
395 * Look for broadcast address and
396 * verify user is allowed to send
400 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
401 error = EADDRNOTAVAIL;
404 if ((flags & IP_ALLOWBROADCAST) == 0) {
408 /* don't allow broadcast messages to be fragmented */
409 if (ip->ip_len > mtu) {
413 m->m_flags |= M_BCAST;
415 m->m_flags &= ~M_BCAST;
420 switch(ip_ipsec_output(&m, inp, &flags, &error, &ro, &iproute, &dst, &ia, &ifp)) {
427 break; /* Continue with packet processing. */
429 /* Update variables that are affected by ipsec4_output(). */
430 ip = mtod(m, struct ip *);
431 hlen = ip->ip_hl << 2;
434 /* Jump over all PFIL processing if hooks are not active. */
435 if (!PFIL_HOOKED(&inet_pfil_hook))
438 /* Run through list of hooks for output packets. */
439 odst.s_addr = ip->ip_dst.s_addr;
440 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
441 if (error != 0 || m == NULL)
444 ip = mtod(m, struct ip *);
446 /* See if destination IP address was changed by packet filter. */
447 if (odst.s_addr != ip->ip_dst.s_addr) {
448 m->m_flags |= M_SKIP_FIREWALL;
449 /* If destination is now ourself drop to ip_input(). */
450 if (in_localip(ip->ip_dst)) {
451 m->m_flags |= M_FASTFWD_OURS;
452 if (m->m_pkthdr.rcvif == NULL)
453 m->m_pkthdr.rcvif = loif;
454 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
455 m->m_pkthdr.csum_flags |=
456 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
457 m->m_pkthdr.csum_data = 0xffff;
459 m->m_pkthdr.csum_flags |=
460 CSUM_IP_CHECKED | CSUM_IP_VALID;
462 error = netisr_queue(NETISR_IP, m);
465 goto again; /* Redo the routing table lookup. */
468 #ifdef IPFIREWALL_FORWARD
469 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
470 if (m->m_flags & M_FASTFWD_OURS) {
471 if (m->m_pkthdr.rcvif == NULL)
472 m->m_pkthdr.rcvif = loif;
473 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
474 m->m_pkthdr.csum_flags |=
475 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
476 m->m_pkthdr.csum_data = 0xffff;
478 m->m_pkthdr.csum_flags |=
479 CSUM_IP_CHECKED | CSUM_IP_VALID;
481 error = netisr_queue(NETISR_IP, m);
484 /* Or forward to some other address? */
485 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
487 dst = (struct sockaddr_in *)&ro->ro_dst;
488 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
489 m->m_flags |= M_SKIP_FIREWALL;
490 m_tag_delete(m, fwd_tag);
493 #endif /* IPFIREWALL_FORWARD */
496 /* 127/8 must not appear on wire - RFC1122. */
497 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
498 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
499 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
500 ipstat.ips_badaddr++;
501 error = EADDRNOTAVAIL;
506 m->m_pkthdr.csum_flags |= CSUM_IP;
507 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
508 if (sw_csum & CSUM_DELAY_DATA) {
510 sw_csum &= ~CSUM_DELAY_DATA;
512 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
515 * If small enough for interface, or the interface will take
516 * care of the fragmentation for us, we can just send directly.
518 if (ip->ip_len <= mtu ||
519 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
520 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
521 ip->ip_len = htons(ip->ip_len);
522 ip->ip_off = htons(ip->ip_off);
524 if (sw_csum & CSUM_DELAY_IP)
525 ip->ip_sum = in_cksum(m, hlen);
528 * Record statistics for this interface address.
529 * With CSUM_TSO the byte/packet count will be slightly
530 * incorrect because we count the IP+TCP headers only
531 * once instead of for every generated packet.
533 if (!(flags & IP_FORWARDING) && ia) {
534 if (m->m_pkthdr.csum_flags & CSUM_TSO)
535 ia->ia_ifa.if_opackets +=
536 m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
538 ia->ia_ifa.if_opackets++;
539 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
541 #ifdef MBUF_STRESS_TEST
542 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
543 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size);
546 * Reset layer specific mbuf flags
547 * to avoid confusing lower layers.
549 m->m_flags &= ~(M_PROTOFLAGS);
551 error = (*ifp->if_output)(ifp, m,
552 (struct sockaddr *)dst, ro->ro_rt);
556 /* Balk when DF bit is set or the interface didn't support TSO. */
557 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
559 ipstat.ips_cantfrag++;
564 * Too large for interface; fragment if possible. If successful,
565 * on return, m will point to a list of packets to be sent.
567 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum);
574 /* Record statistics for this interface address. */
576 ia->ia_ifa.if_opackets++;
577 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
580 * Reset layer specific mbuf flags
581 * to avoid confusing upper layers.
583 m->m_flags &= ~(M_PROTOFLAGS);
585 error = (*ifp->if_output)(ifp, m,
586 (struct sockaddr *)dst, ro->ro_rt);
592 ipstat.ips_fragmented++;
595 if (ro == &iproute && ro->ro_rt) {
605 * Create a chain of fragments which fit the given mtu. m_frag points to the
606 * mbuf to be fragmented; on return it points to the chain with the fragments.
607 * Return 0 if no error. If error, m_frag may contain a partially built
608 * chain of fragments that should be freed by the caller.
610 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
611 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
614 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
615 u_long if_hwassist_flags, int sw_csum)
618 int hlen = ip->ip_hl << 2;
619 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
621 struct mbuf *m0 = *m_frag; /* the original packet */
626 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
627 ipstat.ips_cantfrag++;
632 * Must be able to put at least 8 bytes per fragment.
638 * If the interface will not calculate checksums on
639 * fragmented packets, then do it here.
641 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
642 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
643 in_delayed_cksum(m0);
644 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
647 if (len > PAGE_SIZE) {
649 * Fragment large datagrams such that each segment
650 * contains a multiple of PAGE_SIZE amount of data,
651 * plus headers. This enables a receiver to perform
652 * page-flipping zero-copy optimizations.
654 * XXX When does this help given that sender and receiver
655 * could have different page sizes, and also mtu could
656 * be less than the receiver's page size ?
661 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
665 * firstlen (off - hlen) must be aligned on an
669 goto smart_frag_failure;
670 off = ((off - hlen) & ~7) + hlen;
671 newlen = (~PAGE_MASK) & mtu;
672 if ((newlen + sizeof (struct ip)) > mtu) {
673 /* we failed, go back the default */
684 firstlen = off - hlen;
685 mnext = &m0->m_nextpkt; /* pointer to next packet */
688 * Loop through length of segment after first fragment,
689 * make new header and copy data of each part and link onto chain.
690 * Here, m0 is the original packet, m is the fragment being created.
691 * The fragments are linked off the m_nextpkt of the original
692 * packet, which after processing serves as the first fragment.
694 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
695 struct ip *mhip; /* ip header on the fragment */
697 int mhlen = sizeof (struct ip);
699 MGETHDR(m, M_DONTWAIT, MT_DATA);
702 ipstat.ips_odropped++;
705 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
707 * In the first mbuf, leave room for the link header, then
708 * copy the original IP header including options. The payload
709 * goes into an additional mbuf chain returned by m_copy().
711 m->m_data += max_linkhdr;
712 mhip = mtod(m, struct ip *);
714 if (hlen > sizeof (struct ip)) {
715 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
716 mhip->ip_v = IPVERSION;
717 mhip->ip_hl = mhlen >> 2;
720 /* XXX do we need to add ip->ip_off below ? */
721 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
722 if (off + len >= ip->ip_len) { /* last fragment */
723 len = ip->ip_len - off;
724 m->m_flags |= M_LASTFRAG;
726 mhip->ip_off |= IP_MF;
727 mhip->ip_len = htons((u_short)(len + mhlen));
728 m->m_next = m_copy(m0, off, len);
729 if (m->m_next == NULL) { /* copy failed */
731 error = ENOBUFS; /* ??? */
732 ipstat.ips_odropped++;
735 m->m_pkthdr.len = mhlen + len;
736 m->m_pkthdr.rcvif = NULL;
738 mac_netinet_fragment(m0, m);
740 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
741 mhip->ip_off = htons(mhip->ip_off);
743 if (sw_csum & CSUM_DELAY_IP)
744 mhip->ip_sum = in_cksum(m, mhlen);
746 mnext = &m->m_nextpkt;
748 ipstat.ips_ofragments += nfrags;
750 /* set first marker for fragment chain */
751 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
752 m0->m_pkthdr.csum_data = nfrags;
755 * Update first fragment by trimming what's been copied out
756 * and updating header.
758 m_adj(m0, hlen + firstlen - ip->ip_len);
759 m0->m_pkthdr.len = hlen + firstlen;
760 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
762 ip->ip_off = htons(ip->ip_off);
764 if (sw_csum & CSUM_DELAY_IP)
765 ip->ip_sum = in_cksum(m0, hlen);
773 in_delayed_cksum(struct mbuf *m)
776 u_short csum, offset;
778 ip = mtod(m, struct ip *);
779 offset = ip->ip_hl << 2 ;
780 csum = in_cksum_skip(m, ip->ip_len, offset);
781 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
783 offset += m->m_pkthdr.csum_data; /* checksum offset */
785 if (offset + sizeof(u_short) > m->m_len) {
786 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
787 m->m_len, offset, ip->ip_p);
790 * this shouldn't happen, but if it does, the
791 * correct behavior may be to insert the checksum
792 * in the appropriate next mbuf in the chain.
796 *(u_short *)(m->m_data + offset) = csum;
800 * IP socket option processing.
803 ip_ctloutput(struct socket *so, struct sockopt *sopt)
805 struct inpcb *inp = sotoinpcb(so);
809 if (sopt->sopt_level != IPPROTO_IP) {
813 switch (sopt->sopt_dir) {
815 switch (sopt->sopt_name) {
822 if (sopt->sopt_valsize > MLEN) {
826 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA);
831 m->m_len = sopt->sopt_valsize;
832 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
839 error = ip_pcbopts(inp, sopt->sopt_name, m);
855 error = sooptcopyin(sopt, &optval, sizeof optval,
860 switch (sopt->sopt_name) {
862 inp->inp_ip_tos = optval;
866 inp->inp_ip_ttl = optval;
870 if (optval > 0 && optval <= MAXTTL)
871 inp->inp_ip_minttl = optval;
876 #define OPTSET(bit) do { \
879 inp->inp_flags |= bit; \
881 inp->inp_flags &= ~bit; \
886 OPTSET(INP_RECVOPTS);
890 OPTSET(INP_RECVRETOPTS);
894 OPTSET(INP_RECVDSTADDR);
910 OPTSET(INP_ONESBCAST);
913 OPTSET(INP_DONTFRAG);
920 * Multicast socket options are processed by the in_mcast
923 case IP_MULTICAST_IF:
924 case IP_MULTICAST_VIF:
925 case IP_MULTICAST_TTL:
926 case IP_MULTICAST_LOOP:
927 case IP_ADD_MEMBERSHIP:
928 case IP_DROP_MEMBERSHIP:
929 case IP_ADD_SOURCE_MEMBERSHIP:
930 case IP_DROP_SOURCE_MEMBERSHIP:
931 case IP_BLOCK_SOURCE:
932 case IP_UNBLOCK_SOURCE:
934 case MCAST_JOIN_GROUP:
935 case MCAST_LEAVE_GROUP:
936 case MCAST_JOIN_SOURCE_GROUP:
937 case MCAST_LEAVE_SOURCE_GROUP:
938 case MCAST_BLOCK_SOURCE:
939 case MCAST_UNBLOCK_SOURCE:
940 error = inp_setmoptions(inp, sopt);
944 error = sooptcopyin(sopt, &optval, sizeof optval,
951 case IP_PORTRANGE_DEFAULT:
952 inp->inp_flags &= ~(INP_LOWPORT);
953 inp->inp_flags &= ~(INP_HIGHPORT);
956 case IP_PORTRANGE_HIGH:
957 inp->inp_flags &= ~(INP_LOWPORT);
958 inp->inp_flags |= INP_HIGHPORT;
961 case IP_PORTRANGE_LOW:
962 inp->inp_flags &= ~(INP_HIGHPORT);
963 inp->inp_flags |= INP_LOWPORT;
974 case IP_IPSEC_POLICY:
979 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
981 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
983 req = mtod(m, caddr_t);
984 error = ipsec4_set_policy(inp, sopt->sopt_name, req,
985 m->m_len, (sopt->sopt_td != NULL) ?
986 sopt->sopt_td->td_ucred : NULL);
999 switch (sopt->sopt_name) {
1002 if (inp->inp_options)
1003 error = sooptcopyout(sopt,
1004 mtod(inp->inp_options,
1006 inp->inp_options->m_len);
1008 sopt->sopt_valsize = 0;
1015 case IP_RECVRETOPTS:
1016 case IP_RECVDSTADDR:
1023 switch (sopt->sopt_name) {
1026 optval = inp->inp_ip_tos;
1030 optval = inp->inp_ip_ttl;
1034 optval = inp->inp_ip_minttl;
1037 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1040 optval = OPTBIT(INP_RECVOPTS);
1043 case IP_RECVRETOPTS:
1044 optval = OPTBIT(INP_RECVRETOPTS);
1047 case IP_RECVDSTADDR:
1048 optval = OPTBIT(INP_RECVDSTADDR);
1052 optval = OPTBIT(INP_RECVTTL);
1056 optval = OPTBIT(INP_RECVIF);
1060 if (inp->inp_flags & INP_HIGHPORT)
1061 optval = IP_PORTRANGE_HIGH;
1062 else if (inp->inp_flags & INP_LOWPORT)
1063 optval = IP_PORTRANGE_LOW;
1069 optval = OPTBIT(INP_FAITH);
1073 optval = OPTBIT(INP_ONESBCAST);
1076 optval = OPTBIT(INP_DONTFRAG);
1079 error = sooptcopyout(sopt, &optval, sizeof optval);
1083 * Multicast socket options are processed by the in_mcast
1086 case IP_MULTICAST_IF:
1087 case IP_MULTICAST_VIF:
1088 case IP_MULTICAST_TTL:
1089 case IP_MULTICAST_LOOP:
1091 error = inp_getmoptions(inp, sopt);
1095 case IP_IPSEC_POLICY:
1097 struct mbuf *m = NULL;
1102 req = mtod(m, caddr_t);
1105 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m);
1107 error = soopt_mcopyout(sopt, m); /* XXX */
1115 error = ENOPROTOOPT;
1124 * Routine called from ip_output() to loop back a copy of an IP multicast
1125 * packet to the input queue of a specified interface. Note that this
1126 * calls the output routine of the loopback "driver", but with an interface
1127 * pointer that might NOT be a loopback interface -- evil, but easier than
1128 * replicating that code here.
1131 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1134 register struct ip *ip;
1137 copym = m_copy(m, 0, M_COPYALL);
1138 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1139 copym = m_pullup(copym, hlen);
1140 if (copym != NULL) {
1141 /* If needed, compute the checksum and mark it as valid. */
1142 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1143 in_delayed_cksum(copym);
1144 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1145 copym->m_pkthdr.csum_flags |=
1146 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1147 copym->m_pkthdr.csum_data = 0xffff;
1150 * We don't bother to fragment if the IP length is greater
1151 * than the interface's MTU. Can this possibly matter?
1153 ip = mtod(copym, struct ip *);
1154 ip->ip_len = htons(ip->ip_len);
1155 ip->ip_off = htons(ip->ip_off);
1157 ip->ip_sum = in_cksum(copym, hlen);
1160 * It's not clear whether there are any lingering
1161 * reentrancy problems in other areas which might
1162 * be exposed by using ip_input directly (in
1163 * particular, everything which modifies the packet
1164 * in-place). Yet another option is using the
1165 * protosw directly to deliver the looped back
1166 * packet. For the moment, we'll err on the side
1167 * of safety by using if_simloop().
1170 if (dst->sin_family != AF_INET) {
1171 printf("ip_mloopback: bad address family %d\n",
1173 dst->sin_family = AF_INET;
1178 copym->m_pkthdr.rcvif = ifp;
1181 if_simloop(ifp, copym, dst->sin_family, 0);