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("$FreeBSD$");
37 #include "opt_ipsec.h"
38 #include "opt_mbuf_stress_test.h"
39 #include "opt_mpath.h"
40 #include "opt_route.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/malloc.h>
51 #include <sys/protosw.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sysctl.h>
56 #include <sys/ucred.h>
59 #include <net/if_var.h>
60 #include <net/if_llatbl.h>
61 #include <net/netisr.h>
63 #include <net/route.h>
64 #include <net/flowtable.h>
66 #include <net/radix_mpath.h>
70 #include <netinet/in.h>
71 #include <netinet/in_kdtrace.h>
72 #include <netinet/in_systm.h>
73 #include <netinet/ip.h>
74 #include <netinet/in_pcb.h>
75 #include <netinet/in_rss.h>
76 #include <netinet/in_var.h>
77 #include <netinet/ip_var.h>
78 #include <netinet/ip_options.h>
80 #include <netinet/sctp.h>
81 #include <netinet/sctp_crc32.h>
85 #include <netinet/ip_ipsec.h>
86 #include <netipsec/ipsec.h>
89 #include <machine/in_cksum.h>
91 #include <security/mac/mac_framework.h>
93 VNET_DEFINE(u_short, ip_id);
95 #ifdef MBUF_STRESS_TEST
96 static int mbuf_frag_size = 0;
97 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
98 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
101 static void ip_mloopback
102 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
105 extern int in_mcast_loop;
106 extern struct protosw inetsw[];
109 * IP output. The packet in mbuf chain m contains a skeletal IP
110 * header (with len, off, ttl, proto, tos, src, dst).
111 * The mbuf chain containing the packet will be freed.
112 * The mbuf opt, if present, will not be freed.
113 * If route ro is present and has ro_rt initialized, route lookup would be
114 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
115 * then result of route lookup is stored in ro->ro_rt.
117 * In the IP forwarding case, the packet will arrive with options already
118 * inserted, so must have a NULL opt pointer.
121 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
122 struct ip_moptions *imo, struct inpcb *inp)
125 struct ifnet *ifp = NULL; /* keep compiler happy */
127 int hlen = sizeof (struct ip);
130 struct sockaddr_in *dst;
131 const struct sockaddr_in *gw;
132 struct in_ifaddr *ia;
134 uint16_t ip_len, ip_off;
135 struct route iproute;
136 struct rtentry *rte; /* cache for ro->ro_rt */
138 struct m_tag *fwd_tag = NULL;
141 int no_route_but_check_spd = 0;
146 INP_LOCK_ASSERT(inp);
147 M_SETFIB(m, inp->inp_inc.inc_fibnum);
148 if (((flags & IP_NODEFAULTFLOWID) == 0) &&
149 inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
150 m->m_pkthdr.flowid = inp->inp_flowid;
151 M_HASHTYPE_SET(m, inp->inp_flowtype);
152 m->m_flags |= M_FLOWID;
158 bzero(ro, sizeof (*ro));
162 if (ro->ro_rt == NULL)
163 (void )flowtable_lookup(AF_INET, m, ro);
168 m = ip_insertoptions(m, opt, &len);
170 hlen = len; /* ip->ip_hl is updated above */
172 ip = mtod(m, struct ip *);
173 ip_len = ntohs(ip->ip_len);
174 ip_off = ntohs(ip->ip_off);
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;
200 * dst can be rewritten but always points to &ro->ro_dst.
201 * gw is readonly but can point either to dst OR rt_gateway,
202 * therefore we need restore gw if we're redoing lookup.
204 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
209 * If there is a cached route, check that it is to the same
210 * destination and is still up. If not, free it and try again.
211 * The address family should also be checked in case of sharing
212 * the cache with IPv6.
215 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
216 rte->rt_ifp == NULL ||
217 !RT_LINK_IS_UP(rte->rt_ifp) ||
218 dst->sin_family != AF_INET ||
219 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
225 if (rte == NULL && fwd_tag == NULL) {
226 bzero(dst, sizeof(*dst));
227 dst->sin_family = AF_INET;
228 dst->sin_len = sizeof(*dst);
229 dst->sin_addr = ip->ip_dst;
232 * If routing to interface only, short circuit routing lookup.
233 * The use of an all-ones broadcast address implies this; an
234 * interface is specified by the broadcast address of an interface,
235 * or the destination address of a ptp interface.
237 if (flags & IP_SENDONES) {
238 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
239 M_GETFIB(m)))) == NULL &&
240 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
241 M_GETFIB(m)))) == NULL) {
242 IPSTAT_INC(ips_noroute);
247 ip->ip_dst.s_addr = INADDR_BROADCAST;
248 dst->sin_addr = ip->ip_dst;
252 } else if (flags & IP_ROUTETOIF) {
253 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
254 M_GETFIB(m)))) == NULL &&
255 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
256 M_GETFIB(m)))) == NULL) {
257 IPSTAT_INC(ips_noroute);
264 isbroadcast = in_broadcast(dst->sin_addr, ifp);
265 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
266 imo != NULL && imo->imo_multicast_ifp != NULL) {
268 * Bypass the normal routing lookup for multicast
269 * packets if the interface is specified.
271 ifp = imo->imo_multicast_ifp;
275 isbroadcast = 0; /* fool gcc */
278 * We want to do any cloning requested by the link layer,
279 * as this is probably required in all cases for correct
280 * operation (as it is for ARP).
284 rtalloc_mpath_fib(ro,
285 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
286 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
288 in_rtalloc_ign(ro, 0,
289 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
294 rte->rt_ifp == NULL ||
295 !RT_LINK_IS_UP(rte->rt_ifp)) {
298 * There is no route for this packet, but it is
299 * possible that a matching SPD entry exists.
301 no_route_but_check_spd = 1;
302 mtu = 0; /* Silence GCC warning. */
305 IPSTAT_INC(ips_noroute);
306 error = EHOSTUNREACH;
309 ia = ifatoia(rte->rt_ifa);
311 counter_u64_add(rte->rt_pksent, 1);
312 if (rte->rt_flags & RTF_GATEWAY)
313 gw = (struct sockaddr_in *)rte->rt_gateway;
314 if (rte->rt_flags & RTF_HOST)
315 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
317 isbroadcast = in_broadcast(gw->sin_addr, ifp);
320 * Calculate MTU. If we have a route that is up, use that,
321 * otherwise use the interface's MTU.
323 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
325 * This case can happen if the user changed the MTU
326 * of an interface after enabling IP on it. Because
327 * most netifs don't keep track of routes pointing to
328 * them, there is no way for one to update all its
329 * routes when the MTU is changed.
331 if (rte->rt_mtu > ifp->if_mtu)
332 rte->rt_mtu = ifp->if_mtu;
337 /* Catch a possible divide by zero later. */
338 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
339 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
340 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
341 m->m_flags |= M_MCAST;
343 * IP destination address is multicast. Make sure "gw"
344 * still points to the address in "ro". (It may have been
345 * changed to point to a gateway address, above.)
349 * See if the caller provided any multicast options
352 ip->ip_ttl = imo->imo_multicast_ttl;
353 if (imo->imo_multicast_vif != -1)
356 ip_mcast_src(imo->imo_multicast_vif) :
359 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
361 * Confirm that the outgoing interface supports multicast.
363 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
364 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
365 IPSTAT_INC(ips_noroute);
371 * If source address not specified yet, use address
372 * of outgoing interface.
374 if (ip->ip_src.s_addr == INADDR_ANY) {
375 /* Interface may have no addresses. */
377 ip->ip_src = IA_SIN(ia)->sin_addr;
380 if ((imo == NULL && in_mcast_loop) ||
381 (imo && imo->imo_multicast_loop)) {
383 * Loop back multicast datagram if not expressly
384 * forbidden to do so, even if we are not a member
385 * of the group; ip_input() will filter it later,
386 * thus deferring a hash lookup and mutex acquisition
387 * at the expense of a cheap copy using m_copym().
389 ip_mloopback(ifp, m, dst, hlen);
392 * If we are acting as a multicast router, perform
393 * multicast forwarding as if the packet had just
394 * arrived on the interface to which we are about
395 * to send. The multicast forwarding function
396 * recursively calls this function, using the
397 * IP_FORWARDING flag to prevent infinite recursion.
399 * Multicasts that are looped back by ip_mloopback(),
400 * above, will be forwarded by the ip_input() routine,
403 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
405 * If rsvp daemon is not running, do not
406 * set ip_moptions. This ensures that the packet
407 * is multicast and not just sent down one link
408 * as prescribed by rsvpd.
413 ip_mforward(ip, ifp, m, imo) != 0) {
421 * Multicasts with a time-to-live of zero may be looped-
422 * back, above, but must not be transmitted on a network.
423 * Also, multicasts addressed to the loopback interface
424 * are not sent -- the above call to ip_mloopback() will
425 * loop back a copy. ip_input() will drop the copy if
426 * this host does not belong to the destination group on
427 * the loopback interface.
429 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
438 * If the source address is not specified yet, use the address
439 * of the outoing interface.
441 if (ip->ip_src.s_addr == INADDR_ANY) {
442 /* Interface may have no addresses. */
444 ip->ip_src = IA_SIN(ia)->sin_addr;
449 * Look for broadcast address and
450 * verify user is allowed to send
454 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
455 error = EADDRNOTAVAIL;
458 if ((flags & IP_ALLOWBROADCAST) == 0) {
462 /* don't allow broadcast messages to be fragmented */
467 m->m_flags |= M_BCAST;
469 m->m_flags &= ~M_BCAST;
474 switch(ip_ipsec_output(&m, inp, &flags, &error)) {
481 break; /* Continue with packet processing. */
484 * Check if there was a route for this packet; return error if not.
486 if (no_route_but_check_spd) {
487 IPSTAT_INC(ips_noroute);
488 error = EHOSTUNREACH;
491 /* Update variables that are affected by ipsec4_output(). */
492 ip = mtod(m, struct ip *);
493 hlen = ip->ip_hl << 2;
496 /* Jump over all PFIL processing if hooks are not active. */
497 if (!PFIL_HOOKED(&V_inet_pfil_hook))
500 /* Run through list of hooks for output packets. */
501 odst.s_addr = ip->ip_dst.s_addr;
502 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
503 if (error != 0 || m == NULL)
506 ip = mtod(m, struct ip *);
508 /* See if destination IP address was changed by packet filter. */
509 if (odst.s_addr != ip->ip_dst.s_addr) {
510 m->m_flags |= M_SKIP_FIREWALL;
511 /* If destination is now ourself drop to ip_input(). */
512 if (in_localip(ip->ip_dst)) {
513 m->m_flags |= M_FASTFWD_OURS;
514 if (m->m_pkthdr.rcvif == NULL)
515 m->m_pkthdr.rcvif = V_loif;
516 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
517 m->m_pkthdr.csum_flags |=
518 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
519 m->m_pkthdr.csum_data = 0xffff;
521 m->m_pkthdr.csum_flags |=
522 CSUM_IP_CHECKED | CSUM_IP_VALID;
524 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
525 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
527 error = netisr_queue(NETISR_IP, m);
531 ifa_free(&ia->ia_ifa);
532 goto again; /* Redo the routing table lookup. */
536 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
537 if (m->m_flags & M_FASTFWD_OURS) {
538 if (m->m_pkthdr.rcvif == NULL)
539 m->m_pkthdr.rcvif = V_loif;
540 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
541 m->m_pkthdr.csum_flags |=
542 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
543 m->m_pkthdr.csum_data = 0xffff;
546 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
547 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
549 m->m_pkthdr.csum_flags |=
550 CSUM_IP_CHECKED | CSUM_IP_VALID;
552 error = netisr_queue(NETISR_IP, m);
555 /* Or forward to some other address? */
556 if ((m->m_flags & M_IP_NEXTHOP) &&
557 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
558 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
559 m->m_flags |= M_SKIP_FIREWALL;
560 m->m_flags &= ~M_IP_NEXTHOP;
561 m_tag_delete(m, fwd_tag);
563 ifa_free(&ia->ia_ifa);
568 /* 127/8 must not appear on wire - RFC1122. */
569 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
570 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
571 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
572 IPSTAT_INC(ips_badaddr);
573 error = EADDRNOTAVAIL;
578 m->m_pkthdr.csum_flags |= CSUM_IP;
579 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
581 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
584 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
585 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
586 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
591 * If small enough for interface, or the interface will take
592 * care of the fragmentation for us, we can just send directly.
595 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
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 counter_u64_add(ia->ia_ifa.ifa_opackets,
611 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
613 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
615 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
617 #ifdef MBUF_STRESS_TEST
618 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
619 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
622 * Reset layer specific mbuf flags
623 * to avoid confusing lower layers.
626 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
627 error = (*ifp->if_output)(ifp, m,
628 (const struct sockaddr *)gw, ro);
632 /* Balk when DF bit is set or the interface didn't support TSO. */
633 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
635 IPSTAT_INC(ips_cantfrag);
640 * Too large for interface; fragment if possible. If successful,
641 * on return, m will point to a list of packets to be sent.
643 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
650 /* Record statistics for this interface address. */
652 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
653 counter_u64_add(ia->ia_ifa.ifa_obytes,
657 * Reset layer specific mbuf flags
658 * to avoid confusing upper layers.
662 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
663 error = (*ifp->if_output)(ifp, m,
664 (const struct sockaddr *)gw, ro);
670 IPSTAT_INC(ips_fragmented);
676 ifa_free(&ia->ia_ifa);
684 * Create a chain of fragments which fit the given mtu. m_frag points to the
685 * mbuf to be fragmented; on return it points to the chain with the fragments.
686 * Return 0 if no error. If error, m_frag may contain a partially built
687 * chain of fragments that should be freed by the caller.
689 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
692 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
693 u_long if_hwassist_flags)
696 int hlen = ip->ip_hl << 2;
697 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
699 struct mbuf *m0 = *m_frag; /* the original packet */
703 uint16_t ip_len, ip_off;
705 ip_len = ntohs(ip->ip_len);
706 ip_off = ntohs(ip->ip_off);
708 if (ip_off & IP_DF) { /* Fragmentation not allowed */
709 IPSTAT_INC(ips_cantfrag);
714 * Must be able to put at least 8 bytes per fragment.
720 * If the interface will not calculate checksums on
721 * fragmented packets, then do it here.
723 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
724 in_delayed_cksum(m0);
725 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
728 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
729 sctp_delayed_cksum(m0, hlen);
730 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
733 if (len > PAGE_SIZE) {
735 * Fragment large datagrams such that each segment
736 * contains a multiple of PAGE_SIZE amount of data,
737 * plus headers. This enables a receiver to perform
738 * page-flipping zero-copy optimizations.
740 * XXX When does this help given that sender and receiver
741 * could have different page sizes, and also mtu could
742 * be less than the receiver's page size ?
747 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
751 * firstlen (off - hlen) must be aligned on an
755 goto smart_frag_failure;
756 off = ((off - hlen) & ~7) + hlen;
757 newlen = (~PAGE_MASK) & mtu;
758 if ((newlen + sizeof (struct ip)) > mtu) {
759 /* we failed, go back the default */
770 firstlen = off - hlen;
771 mnext = &m0->m_nextpkt; /* pointer to next packet */
774 * Loop through length of segment after first fragment,
775 * make new header and copy data of each part and link onto chain.
776 * Here, m0 is the original packet, m is the fragment being created.
777 * The fragments are linked off the m_nextpkt of the original
778 * packet, which after processing serves as the first fragment.
780 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
781 struct ip *mhip; /* ip header on the fragment */
783 int mhlen = sizeof (struct ip);
785 m = m_gethdr(M_NOWAIT, MT_DATA);
788 IPSTAT_INC(ips_odropped);
791 m->m_flags |= (m0->m_flags & M_MCAST);
793 * In the first mbuf, leave room for the link header, then
794 * copy the original IP header including options. The payload
795 * goes into an additional mbuf chain returned by m_copym().
797 m->m_data += max_linkhdr;
798 mhip = mtod(m, struct ip *);
800 if (hlen > sizeof (struct ip)) {
801 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
802 mhip->ip_v = IPVERSION;
803 mhip->ip_hl = mhlen >> 2;
806 /* XXX do we need to add ip_off below ? */
807 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
808 if (off + len >= ip_len)
811 mhip->ip_off |= IP_MF;
812 mhip->ip_len = htons((u_short)(len + mhlen));
813 m->m_next = m_copym(m0, off, len, M_NOWAIT);
814 if (m->m_next == NULL) { /* copy failed */
816 error = ENOBUFS; /* ??? */
817 IPSTAT_INC(ips_odropped);
820 m->m_pkthdr.len = mhlen + len;
821 m->m_pkthdr.rcvif = NULL;
823 mac_netinet_fragment(m0, m);
825 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
826 mhip->ip_off = htons(mhip->ip_off);
828 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
829 mhip->ip_sum = in_cksum(m, mhlen);
830 m->m_pkthdr.csum_flags &= ~CSUM_IP;
833 mnext = &m->m_nextpkt;
835 IPSTAT_ADD(ips_ofragments, nfrags);
838 * Update first fragment by trimming what's been copied out
839 * and updating header.
841 m_adj(m0, hlen + firstlen - ip_len);
842 m0->m_pkthdr.len = hlen + firstlen;
843 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
844 ip->ip_off = htons(ip_off | IP_MF);
846 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
847 ip->ip_sum = in_cksum(m0, hlen);
848 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
857 in_delayed_cksum(struct mbuf *m)
860 uint16_t csum, offset, ip_len;
862 ip = mtod(m, struct ip *);
863 offset = ip->ip_hl << 2 ;
864 ip_len = ntohs(ip->ip_len);
865 csum = in_cksum_skip(m, ip_len, offset);
866 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
868 offset += m->m_pkthdr.csum_data; /* checksum offset */
870 /* find the mbuf in the chain where the checksum starts*/
871 while ((m != NULL) && (offset >= m->m_len)) {
875 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
876 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
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);
894 if (sopt->sopt_level != IPPROTO_IP) {
897 if (sopt->sopt_level == SOL_SOCKET &&
898 sopt->sopt_dir == SOPT_SET) {
899 switch (sopt->sopt_name) {
902 if ((so->so_options & SO_REUSEADDR) != 0)
903 inp->inp_flags2 |= INP_REUSEADDR;
905 inp->inp_flags2 &= ~INP_REUSEADDR;
911 if ((so->so_options & SO_REUSEPORT) != 0)
912 inp->inp_flags2 |= INP_REUSEPORT;
914 inp->inp_flags2 &= ~INP_REUSEPORT;
920 inp->inp_inc.inc_fibnum = so->so_fibnum;
931 switch (sopt->sopt_dir) {
933 switch (sopt->sopt_name) {
940 if (sopt->sopt_valsize > MLEN) {
944 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
949 m->m_len = sopt->sopt_valsize;
950 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
957 error = ip_pcbopts(inp, sopt->sopt_name, m);
963 if (sopt->sopt_td != NULL) {
964 error = priv_check(sopt->sopt_td,
965 PRIV_NETINET_BINDANY);
972 case IP_RSS_LISTEN_BUCKET:
988 case IP_RECVRSSBUCKETID:
990 error = sooptcopyin(sopt, &optval, sizeof optval,
995 switch (sopt->sopt_name) {
997 inp->inp_ip_tos = optval;
1001 inp->inp_ip_ttl = optval;
1005 if (optval >= 0 && optval <= MAXTTL)
1006 inp->inp_ip_minttl = optval;
1011 #define OPTSET(bit) do { \
1014 inp->inp_flags |= bit; \
1016 inp->inp_flags &= ~bit; \
1020 #define OPTSET2(bit, val) do { \
1023 inp->inp_flags2 |= bit; \
1025 inp->inp_flags2 &= ~bit; \
1030 OPTSET(INP_RECVOPTS);
1033 case IP_RECVRETOPTS:
1034 OPTSET(INP_RECVRETOPTS);
1037 case IP_RECVDSTADDR:
1038 OPTSET(INP_RECVDSTADDR);
1042 OPTSET(INP_RECVTTL);
1054 OPTSET(INP_ONESBCAST);
1057 OPTSET(INP_DONTFRAG);
1060 OPTSET(INP_BINDANY);
1063 OPTSET(INP_RECVTOS);
1066 OPTSET2(INP_BINDMULTI, optval);
1069 OPTSET2(INP_RECVFLOWID, optval);
1072 case IP_RSS_LISTEN_BUCKET:
1073 if ((optval >= 0) &&
1074 (optval < rss_getnumbuckets())) {
1075 inp->inp_rss_listen_bucket = optval;
1076 OPTSET2(INP_RSS_BUCKET_SET, 1);
1081 case IP_RECVRSSBUCKETID:
1082 OPTSET2(INP_RECVRSSBUCKETID, optval);
1091 * Multicast socket options are processed by the in_mcast
1094 case IP_MULTICAST_IF:
1095 case IP_MULTICAST_VIF:
1096 case IP_MULTICAST_TTL:
1097 case IP_MULTICAST_LOOP:
1098 case IP_ADD_MEMBERSHIP:
1099 case IP_DROP_MEMBERSHIP:
1100 case IP_ADD_SOURCE_MEMBERSHIP:
1101 case IP_DROP_SOURCE_MEMBERSHIP:
1102 case IP_BLOCK_SOURCE:
1103 case IP_UNBLOCK_SOURCE:
1105 case MCAST_JOIN_GROUP:
1106 case MCAST_LEAVE_GROUP:
1107 case MCAST_JOIN_SOURCE_GROUP:
1108 case MCAST_LEAVE_SOURCE_GROUP:
1109 case MCAST_BLOCK_SOURCE:
1110 case MCAST_UNBLOCK_SOURCE:
1111 error = inp_setmoptions(inp, sopt);
1115 error = sooptcopyin(sopt, &optval, sizeof optval,
1122 case IP_PORTRANGE_DEFAULT:
1123 inp->inp_flags &= ~(INP_LOWPORT);
1124 inp->inp_flags &= ~(INP_HIGHPORT);
1127 case IP_PORTRANGE_HIGH:
1128 inp->inp_flags &= ~(INP_LOWPORT);
1129 inp->inp_flags |= INP_HIGHPORT;
1132 case IP_PORTRANGE_LOW:
1133 inp->inp_flags &= ~(INP_HIGHPORT);
1134 inp->inp_flags |= INP_LOWPORT;
1145 case IP_IPSEC_POLICY:
1150 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1152 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1154 req = mtod(m, caddr_t);
1155 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1156 m->m_len, (sopt->sopt_td != NULL) ?
1157 sopt->sopt_td->td_ucred : NULL);
1164 error = ENOPROTOOPT;
1170 switch (sopt->sopt_name) {
1173 if (inp->inp_options)
1174 error = sooptcopyout(sopt,
1175 mtod(inp->inp_options,
1177 inp->inp_options->m_len);
1179 sopt->sopt_valsize = 0;
1186 case IP_RECVRETOPTS:
1187 case IP_RECVDSTADDR:
1201 case IP_RSSBUCKETID:
1202 case IP_RECVRSSBUCKETID:
1204 switch (sopt->sopt_name) {
1207 optval = inp->inp_ip_tos;
1211 optval = inp->inp_ip_ttl;
1215 optval = inp->inp_ip_minttl;
1218 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1219 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1222 optval = OPTBIT(INP_RECVOPTS);
1225 case IP_RECVRETOPTS:
1226 optval = OPTBIT(INP_RECVRETOPTS);
1229 case IP_RECVDSTADDR:
1230 optval = OPTBIT(INP_RECVDSTADDR);
1234 optval = OPTBIT(INP_RECVTTL);
1238 optval = OPTBIT(INP_RECVIF);
1242 if (inp->inp_flags & INP_HIGHPORT)
1243 optval = IP_PORTRANGE_HIGH;
1244 else if (inp->inp_flags & INP_LOWPORT)
1245 optval = IP_PORTRANGE_LOW;
1251 optval = OPTBIT(INP_FAITH);
1255 optval = OPTBIT(INP_ONESBCAST);
1258 optval = OPTBIT(INP_DONTFRAG);
1261 optval = OPTBIT(INP_BINDANY);
1264 optval = OPTBIT(INP_RECVTOS);
1267 optval = inp->inp_flowid;
1270 optval = inp->inp_flowtype;
1273 optval = OPTBIT2(INP_RECVFLOWID);
1276 case IP_RSSBUCKETID:
1277 retval = rss_hash2bucket(inp->inp_flowid,
1281 optval = rss_bucket;
1285 case IP_RECVRSSBUCKETID:
1286 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1290 optval = OPTBIT2(INP_BINDMULTI);
1293 error = sooptcopyout(sopt, &optval, sizeof optval);
1297 * Multicast socket options are processed by the in_mcast
1300 case IP_MULTICAST_IF:
1301 case IP_MULTICAST_VIF:
1302 case IP_MULTICAST_TTL:
1303 case IP_MULTICAST_LOOP:
1305 error = inp_getmoptions(inp, sopt);
1309 case IP_IPSEC_POLICY:
1311 struct mbuf *m = NULL;
1316 req = mtod(m, caddr_t);
1319 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1321 error = soopt_mcopyout(sopt, m); /* XXX */
1329 error = ENOPROTOOPT;
1338 * Routine called from ip_output() to loop back a copy of an IP multicast
1339 * packet to the input queue of a specified interface. Note that this
1340 * calls the output routine of the loopback "driver", but with an interface
1341 * pointer that might NOT be a loopback interface -- evil, but easier than
1342 * replicating that code here.
1345 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1348 register struct ip *ip;
1352 * Make a deep copy of the packet because we're going to
1353 * modify the pack in order to generate checksums.
1355 copym = m_dup(m, M_NOWAIT);
1356 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1357 copym = m_pullup(copym, hlen);
1358 if (copym != NULL) {
1359 /* If needed, compute the checksum and mark it as valid. */
1360 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1361 in_delayed_cksum(copym);
1362 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1363 copym->m_pkthdr.csum_flags |=
1364 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1365 copym->m_pkthdr.csum_data = 0xffff;
1368 * We don't bother to fragment if the IP length is greater
1369 * than the interface's MTU. Can this possibly matter?
1371 ip = mtod(copym, struct ip *);
1373 ip->ip_sum = in_cksum(copym, hlen);
1375 if (dst->sin_family != AF_INET) {
1376 printf("ip_mloopback: bad address family %d\n",
1378 dst->sin_family = AF_INET;
1381 if_simloop(ifp, copym, dst->sin_family, 0);