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>
68 #include <net/rss_config.h>
71 #include <netinet/in.h>
72 #include <netinet/in_kdtrace.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/ip.h>
75 #include <netinet/in_pcb.h>
76 #include <netinet/in_rss.h>
77 #include <netinet/in_var.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_options.h>
81 #include <netinet/sctp.h>
82 #include <netinet/sctp_crc32.h>
86 #include <netinet/ip_ipsec.h>
87 #include <netipsec/ipsec.h>
90 #include <machine/in_cksum.h>
92 #include <security/mac/mac_framework.h>
94 #ifdef MBUF_STRESS_TEST
95 static int mbuf_frag_size = 0;
96 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
97 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
100 static void ip_mloopback
101 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
104 extern int in_mcast_loop;
105 extern struct protosw inetsw[];
108 * IP output. The packet in mbuf chain m contains a skeletal IP
109 * header (with len, off, ttl, proto, tos, src, dst).
110 * The mbuf chain containing the packet will be freed.
111 * The mbuf opt, if present, will not be freed.
112 * If route ro is present and has ro_rt initialized, route lookup would be
113 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
114 * then result of route lookup is stored in ro->ro_rt.
116 * In the IP forwarding case, the packet will arrive with options already
117 * inserted, so must have a NULL opt pointer.
120 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
121 struct ip_moptions *imo, struct inpcb *inp)
124 struct ifnet *ifp = NULL; /* keep compiler happy */
126 int hlen = sizeof (struct ip);
129 struct sockaddr_in *dst;
130 const struct sockaddr_in *gw;
131 struct in_ifaddr *ia;
133 uint16_t ip_len, ip_off;
134 struct route iproute;
135 struct rtentry *rte; /* cache for ro->ro_rt */
137 struct m_tag *fwd_tag = NULL;
142 int no_route_but_check_spd = 0;
147 INP_LOCK_ASSERT(inp);
148 M_SETFIB(m, inp->inp_inc.inc_fibnum);
149 if ((flags & IP_NODEFAULTFLOWID) == 0) {
150 m->m_pkthdr.flowid = inp->inp_flowid;
151 M_HASHTYPE_SET(m, inp->inp_flowtype);
157 bzero(ro, sizeof (*ro));
161 if (ro->ro_rt == NULL)
162 (void )flowtable_lookup(AF_INET, m, ro);
167 m = ip_insertoptions(m, opt, &len);
169 hlen = len; /* ip->ip_hl is updated above */
171 ip = mtod(m, struct ip *);
172 ip_len = ntohs(ip->ip_len);
173 ip_off = ntohs(ip->ip_off);
175 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
176 ip->ip_v = IPVERSION;
177 ip->ip_hl = hlen >> 2;
179 IPSTAT_INC(ips_localout);
181 /* Header already set, fetch hlen from there */
182 hlen = ip->ip_hl << 2;
188 * dst can be rewritten but always points to &ro->ro_dst.
189 * gw is readonly but can point either to dst OR rt_gateway,
190 * therefore we need restore gw if we're redoing lookup.
192 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
193 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
198 * If there is a cached route, check that it is to the same
199 * destination and is still up. If not, free it and try again.
200 * The address family should also be checked in case of sharing
201 * the cache with IPv6.
204 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
205 rte->rt_ifp == NULL ||
206 !RT_LINK_IS_UP(rte->rt_ifp) ||
207 dst->sin_family != AF_INET ||
208 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
214 if (rte == NULL && fwd_tag == NULL) {
215 bzero(dst, sizeof(*dst));
216 dst->sin_family = AF_INET;
217 dst->sin_len = sizeof(*dst);
218 dst->sin_addr = ip->ip_dst;
221 * If routing to interface only, short circuit routing lookup.
222 * The use of an all-ones broadcast address implies this; an
223 * interface is specified by the broadcast address of an interface,
224 * or the destination address of a ptp interface.
226 if (flags & IP_SENDONES) {
227 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
228 M_GETFIB(m)))) == NULL &&
229 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
230 M_GETFIB(m)))) == NULL) {
231 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),
243 M_GETFIB(m)))) == NULL &&
244 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
245 M_GETFIB(m)))) == NULL) {
246 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;
264 isbroadcast = 0; /* fool gcc */
267 * We want to do any cloning requested by the link layer,
268 * as this is probably required in all cases for correct
269 * operation (as it is for ARP).
273 rtalloc_mpath_fib(ro,
274 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
277 in_rtalloc_ign(ro, 0, fibnum);
282 rte->rt_ifp == NULL ||
283 !RT_LINK_IS_UP(rte->rt_ifp)) {
286 * There is no route for this packet, but it is
287 * possible that a matching SPD entry exists.
289 no_route_but_check_spd = 1;
290 mtu = 0; /* Silence GCC warning. */
293 IPSTAT_INC(ips_noroute);
294 error = EHOSTUNREACH;
297 ia = ifatoia(rte->rt_ifa);
299 counter_u64_add(rte->rt_pksent, 1);
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)))
315 /* Catch a possible divide by zero later. */
316 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
317 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
318 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
319 m->m_flags |= M_MCAST;
321 * IP destination address is multicast. Make sure "gw"
322 * still points to the address in "ro". (It may have been
323 * changed to point to a gateway address, above.)
327 * See if the caller provided any multicast options
330 ip->ip_ttl = imo->imo_multicast_ttl;
331 if (imo->imo_multicast_vif != -1)
334 ip_mcast_src(imo->imo_multicast_vif) :
337 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
339 * Confirm that the outgoing interface supports multicast.
341 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
342 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
343 IPSTAT_INC(ips_noroute);
349 * If source address not specified yet, use address
350 * of outgoing interface.
352 if (ip->ip_src.s_addr == INADDR_ANY) {
353 /* Interface may have no addresses. */
355 ip->ip_src = IA_SIN(ia)->sin_addr;
358 if ((imo == NULL && in_mcast_loop) ||
359 (imo && imo->imo_multicast_loop)) {
361 * Loop back multicast datagram if not expressly
362 * forbidden to do so, even if we are not a member
363 * of the group; ip_input() will filter it later,
364 * thus deferring a hash lookup and mutex acquisition
365 * at the expense of a cheap copy using m_copym().
367 ip_mloopback(ifp, m, dst, hlen);
370 * If we are acting as a multicast router, perform
371 * multicast forwarding as if the packet had just
372 * arrived on the interface to which we are about
373 * to send. The multicast forwarding function
374 * recursively calls this function, using the
375 * IP_FORWARDING flag to prevent infinite recursion.
377 * Multicasts that are looped back by ip_mloopback(),
378 * above, will be forwarded by the ip_input() routine,
381 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
383 * If rsvp daemon is not running, do not
384 * set ip_moptions. This ensures that the packet
385 * is multicast and not just sent down one link
386 * as prescribed by rsvpd.
391 ip_mforward(ip, ifp, m, imo) != 0) {
399 * Multicasts with a time-to-live of zero may be looped-
400 * back, above, but must not be transmitted on a network.
401 * Also, multicasts addressed to the loopback interface
402 * are not sent -- the above call to ip_mloopback() will
403 * loop back a copy. ip_input() will drop the copy if
404 * this host does not belong to the destination group on
405 * the loopback interface.
407 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
416 * If the source address is not specified yet, use the address
417 * of the outoing interface.
419 if (ip->ip_src.s_addr == INADDR_ANY) {
420 /* Interface may have no addresses. */
422 ip->ip_src = IA_SIN(ia)->sin_addr;
427 * Look for broadcast address and
428 * verify user is allowed to send
432 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
433 error = EADDRNOTAVAIL;
436 if ((flags & IP_ALLOWBROADCAST) == 0) {
440 /* don't allow broadcast messages to be fragmented */
445 m->m_flags |= M_BCAST;
447 m->m_flags &= ~M_BCAST;
452 switch(ip_ipsec_output(&m, inp, &error)) {
459 break; /* Continue with packet processing. */
462 * Check if there was a route for this packet; return error if not.
464 if (no_route_but_check_spd) {
465 IPSTAT_INC(ips_noroute);
466 error = EHOSTUNREACH;
469 /* Update variables that are affected by ipsec4_output(). */
470 ip = mtod(m, struct ip *);
471 hlen = ip->ip_hl << 2;
474 /* Jump over all PFIL processing if hooks are not active. */
475 if (!PFIL_HOOKED(&V_inet_pfil_hook))
478 /* Run through list of hooks for output packets. */
479 odst.s_addr = ip->ip_dst.s_addr;
480 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
481 if (error != 0 || m == NULL)
484 ip = mtod(m, struct ip *);
487 /* See if destination IP address was changed by packet filter. */
488 if (odst.s_addr != ip->ip_dst.s_addr) {
489 m->m_flags |= M_SKIP_FIREWALL;
490 /* If destination is now ourself drop to ip_input(). */
491 if (in_localip(ip->ip_dst)) {
492 m->m_flags |= M_FASTFWD_OURS;
493 if (m->m_pkthdr.rcvif == NULL)
494 m->m_pkthdr.rcvif = V_loif;
495 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
496 m->m_pkthdr.csum_flags |=
497 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
498 m->m_pkthdr.csum_data = 0xffff;
500 m->m_pkthdr.csum_flags |=
501 CSUM_IP_CHECKED | CSUM_IP_VALID;
503 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
504 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
506 error = netisr_queue(NETISR_IP, m);
510 ifa_free(&ia->ia_ifa);
511 needfiblookup = 1; /* Redo the routing table lookup. */
514 /* See if fib was changed by packet filter. */
515 if (fibnum != M_GETFIB(m)) {
516 m->m_flags |= M_SKIP_FIREWALL;
517 fibnum = M_GETFIB(m);
524 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
525 if (m->m_flags & M_FASTFWD_OURS) {
526 if (m->m_pkthdr.rcvif == NULL)
527 m->m_pkthdr.rcvif = V_loif;
528 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
529 m->m_pkthdr.csum_flags |=
530 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
531 m->m_pkthdr.csum_data = 0xffff;
534 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
535 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
537 m->m_pkthdr.csum_flags |=
538 CSUM_IP_CHECKED | CSUM_IP_VALID;
540 error = netisr_queue(NETISR_IP, m);
543 /* Or forward to some other address? */
544 if ((m->m_flags & M_IP_NEXTHOP) &&
545 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
546 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
547 m->m_flags |= M_SKIP_FIREWALL;
548 m->m_flags &= ~M_IP_NEXTHOP;
549 m_tag_delete(m, fwd_tag);
551 ifa_free(&ia->ia_ifa);
556 /* 127/8 must not appear on wire - RFC1122. */
557 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
558 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
559 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
560 IPSTAT_INC(ips_badaddr);
561 error = EADDRNOTAVAIL;
566 m->m_pkthdr.csum_flags |= CSUM_IP;
567 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
569 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
572 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
573 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
574 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
579 * If small enough for interface, or the interface will take
580 * care of the fragmentation for us, we can just send directly.
583 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
585 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
586 ip->ip_sum = in_cksum(m, hlen);
587 m->m_pkthdr.csum_flags &= ~CSUM_IP;
591 * Record statistics for this interface address.
592 * With CSUM_TSO the byte/packet count will be slightly
593 * incorrect because we count the IP+TCP headers only
594 * once instead of for every generated packet.
596 if (!(flags & IP_FORWARDING) && ia) {
597 if (m->m_pkthdr.csum_flags & CSUM_TSO)
598 counter_u64_add(ia->ia_ifa.ifa_opackets,
599 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
601 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
603 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
605 #ifdef MBUF_STRESS_TEST
606 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
607 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
610 * Reset layer specific mbuf flags
611 * to avoid confusing lower layers.
614 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
615 error = (*ifp->if_output)(ifp, m,
616 (const struct sockaddr *)gw, ro);
620 /* Balk when DF bit is set or the interface didn't support TSO. */
621 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
623 IPSTAT_INC(ips_cantfrag);
628 * Too large for interface; fragment if possible. If successful,
629 * on return, m will point to a list of packets to be sent.
631 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
638 /* Record statistics for this interface address. */
640 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
641 counter_u64_add(ia->ia_ifa.ifa_obytes,
645 * Reset layer specific mbuf flags
646 * to avoid confusing upper layers.
650 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
651 error = (*ifp->if_output)(ifp, m,
652 (const struct sockaddr *)gw, ro);
658 IPSTAT_INC(ips_fragmented);
664 ifa_free(&ia->ia_ifa);
672 * Create a chain of fragments which fit the given mtu. m_frag points to the
673 * mbuf to be fragmented; on return it points to the chain with the fragments.
674 * Return 0 if no error. If error, m_frag may contain a partially built
675 * chain of fragments that should be freed by the caller.
677 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
680 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
681 u_long if_hwassist_flags)
684 int hlen = ip->ip_hl << 2;
685 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
687 struct mbuf *m0 = *m_frag; /* the original packet */
691 uint16_t ip_len, ip_off;
693 ip_len = ntohs(ip->ip_len);
694 ip_off = ntohs(ip->ip_off);
696 if (ip_off & IP_DF) { /* Fragmentation not allowed */
697 IPSTAT_INC(ips_cantfrag);
702 * Must be able to put at least 8 bytes per fragment.
708 * If the interface will not calculate checksums on
709 * fragmented packets, then do it here.
711 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
712 in_delayed_cksum(m0);
713 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
716 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
717 sctp_delayed_cksum(m0, hlen);
718 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
721 if (len > PAGE_SIZE) {
723 * Fragment large datagrams such that each segment
724 * contains a multiple of PAGE_SIZE amount of data,
725 * plus headers. This enables a receiver to perform
726 * page-flipping zero-copy optimizations.
728 * XXX When does this help given that sender and receiver
729 * could have different page sizes, and also mtu could
730 * be less than the receiver's page size ?
734 off = MIN(mtu, m0->m_pkthdr.len);
737 * firstlen (off - hlen) must be aligned on an
741 goto smart_frag_failure;
742 off = ((off - hlen) & ~7) + hlen;
743 newlen = (~PAGE_MASK) & mtu;
744 if ((newlen + sizeof (struct ip)) > mtu) {
745 /* we failed, go back the default */
756 firstlen = off - hlen;
757 mnext = &m0->m_nextpkt; /* pointer to next packet */
760 * Loop through length of segment after first fragment,
761 * make new header and copy data of each part and link onto chain.
762 * Here, m0 is the original packet, m is the fragment being created.
763 * The fragments are linked off the m_nextpkt of the original
764 * packet, which after processing serves as the first fragment.
766 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
767 struct ip *mhip; /* ip header on the fragment */
769 int mhlen = sizeof (struct ip);
771 m = m_gethdr(M_NOWAIT, MT_DATA);
774 IPSTAT_INC(ips_odropped);
778 * Make sure the complete packet header gets copied
779 * from the originating mbuf to the newly created
780 * mbuf. This also ensures that existing firewall
781 * classification(s), VLAN tags and so on get copied
782 * to the resulting fragmented packet(s):
784 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
787 IPSTAT_INC(ips_odropped);
791 * In the first mbuf, leave room for the link header, then
792 * copy the original IP header including options. The payload
793 * goes into an additional mbuf chain returned by m_copym().
795 m->m_data += max_linkhdr;
796 mhip = mtod(m, struct ip *);
798 if (hlen > sizeof (struct ip)) {
799 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
800 mhip->ip_v = IPVERSION;
801 mhip->ip_hl = mhlen >> 2;
804 /* XXX do we need to add ip_off below ? */
805 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
806 if (off + len >= ip_len)
809 mhip->ip_off |= IP_MF;
810 mhip->ip_len = htons((u_short)(len + mhlen));
811 m->m_next = m_copym(m0, off, len, M_NOWAIT);
812 if (m->m_next == NULL) { /* copy failed */
814 error = ENOBUFS; /* ??? */
815 IPSTAT_INC(ips_odropped);
818 m->m_pkthdr.len = mhlen + len;
820 mac_netinet_fragment(m0, m);
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 /* find the mbuf in the chain where the checksum starts*/
867 while ((m != NULL) && (offset >= m->m_len)) {
871 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
872 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
873 *(u_short *)(m->m_data + offset) = csum;
877 * IP socket option processing.
880 ip_ctloutput(struct socket *so, struct sockopt *sopt)
882 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);
968 case IP_RSS_LISTEN_BUCKET:
983 case IP_RECVRSSBUCKETID:
985 error = sooptcopyin(sopt, &optval, sizeof optval,
990 switch (sopt->sopt_name) {
992 inp->inp_ip_tos = optval;
996 inp->inp_ip_ttl = optval;
1000 if (optval >= 0 && optval <= MAXTTL)
1001 inp->inp_ip_minttl = optval;
1006 #define OPTSET(bit) do { \
1009 inp->inp_flags |= bit; \
1011 inp->inp_flags &= ~bit; \
1015 #define OPTSET2(bit, val) do { \
1018 inp->inp_flags2 |= bit; \
1020 inp->inp_flags2 &= ~bit; \
1025 OPTSET(INP_RECVOPTS);
1028 case IP_RECVRETOPTS:
1029 OPTSET(INP_RECVRETOPTS);
1032 case IP_RECVDSTADDR:
1033 OPTSET(INP_RECVDSTADDR);
1037 OPTSET(INP_RECVTTL);
1045 OPTSET(INP_ONESBCAST);
1048 OPTSET(INP_DONTFRAG);
1051 OPTSET(INP_BINDANY);
1054 OPTSET(INP_RECVTOS);
1057 OPTSET2(INP_BINDMULTI, optval);
1060 OPTSET2(INP_RECVFLOWID, optval);
1063 case IP_RSS_LISTEN_BUCKET:
1064 if ((optval >= 0) &&
1065 (optval < rss_getnumbuckets())) {
1066 inp->inp_rss_listen_bucket = optval;
1067 OPTSET2(INP_RSS_BUCKET_SET, 1);
1072 case IP_RECVRSSBUCKETID:
1073 OPTSET2(INP_RECVRSSBUCKETID, optval);
1082 * Multicast socket options are processed by the in_mcast
1085 case IP_MULTICAST_IF:
1086 case IP_MULTICAST_VIF:
1087 case IP_MULTICAST_TTL:
1088 case IP_MULTICAST_LOOP:
1089 case IP_ADD_MEMBERSHIP:
1090 case IP_DROP_MEMBERSHIP:
1091 case IP_ADD_SOURCE_MEMBERSHIP:
1092 case IP_DROP_SOURCE_MEMBERSHIP:
1093 case IP_BLOCK_SOURCE:
1094 case IP_UNBLOCK_SOURCE:
1096 case MCAST_JOIN_GROUP:
1097 case MCAST_LEAVE_GROUP:
1098 case MCAST_JOIN_SOURCE_GROUP:
1099 case MCAST_LEAVE_SOURCE_GROUP:
1100 case MCAST_BLOCK_SOURCE:
1101 case MCAST_UNBLOCK_SOURCE:
1102 error = inp_setmoptions(inp, sopt);
1106 error = sooptcopyin(sopt, &optval, sizeof optval,
1113 case IP_PORTRANGE_DEFAULT:
1114 inp->inp_flags &= ~(INP_LOWPORT);
1115 inp->inp_flags &= ~(INP_HIGHPORT);
1118 case IP_PORTRANGE_HIGH:
1119 inp->inp_flags &= ~(INP_LOWPORT);
1120 inp->inp_flags |= INP_HIGHPORT;
1123 case IP_PORTRANGE_LOW:
1124 inp->inp_flags &= ~(INP_HIGHPORT);
1125 inp->inp_flags |= INP_LOWPORT;
1136 case IP_IPSEC_POLICY:
1141 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1143 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1145 req = mtod(m, caddr_t);
1146 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1147 m->m_len, (sopt->sopt_td != NULL) ?
1148 sopt->sopt_td->td_ucred : NULL);
1155 error = ENOPROTOOPT;
1161 switch (sopt->sopt_name) {
1164 if (inp->inp_options)
1165 error = sooptcopyout(sopt,
1166 mtod(inp->inp_options,
1168 inp->inp_options->m_len);
1170 sopt->sopt_valsize = 0;
1177 case IP_RECVRETOPTS:
1178 case IP_RECVDSTADDR:
1191 case IP_RSSBUCKETID:
1192 case IP_RECVRSSBUCKETID:
1194 switch (sopt->sopt_name) {
1197 optval = inp->inp_ip_tos;
1201 optval = inp->inp_ip_ttl;
1205 optval = inp->inp_ip_minttl;
1208 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1209 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1212 optval = OPTBIT(INP_RECVOPTS);
1215 case IP_RECVRETOPTS:
1216 optval = OPTBIT(INP_RECVRETOPTS);
1219 case IP_RECVDSTADDR:
1220 optval = OPTBIT(INP_RECVDSTADDR);
1224 optval = OPTBIT(INP_RECVTTL);
1228 optval = OPTBIT(INP_RECVIF);
1232 if (inp->inp_flags & INP_HIGHPORT)
1233 optval = IP_PORTRANGE_HIGH;
1234 else if (inp->inp_flags & INP_LOWPORT)
1235 optval = IP_PORTRANGE_LOW;
1241 optval = OPTBIT(INP_ONESBCAST);
1244 optval = OPTBIT(INP_DONTFRAG);
1247 optval = OPTBIT(INP_BINDANY);
1250 optval = OPTBIT(INP_RECVTOS);
1253 optval = inp->inp_flowid;
1256 optval = inp->inp_flowtype;
1259 optval = OPTBIT2(INP_RECVFLOWID);
1262 case IP_RSSBUCKETID:
1263 retval = rss_hash2bucket(inp->inp_flowid,
1267 optval = rss_bucket;
1271 case IP_RECVRSSBUCKETID:
1272 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1276 optval = OPTBIT2(INP_BINDMULTI);
1279 error = sooptcopyout(sopt, &optval, sizeof optval);
1283 * Multicast socket options are processed by the in_mcast
1286 case IP_MULTICAST_IF:
1287 case IP_MULTICAST_VIF:
1288 case IP_MULTICAST_TTL:
1289 case IP_MULTICAST_LOOP:
1291 error = inp_getmoptions(inp, sopt);
1295 case IP_IPSEC_POLICY:
1297 struct mbuf *m = NULL;
1302 req = mtod(m, caddr_t);
1305 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1307 error = soopt_mcopyout(sopt, m); /* XXX */
1315 error = ENOPROTOOPT;
1324 * Routine called from ip_output() to loop back a copy of an IP multicast
1325 * packet to the input queue of a specified interface. Note that this
1326 * calls the output routine of the loopback "driver", but with an interface
1327 * pointer that might NOT be a loopback interface -- evil, but easier than
1328 * replicating that code here.
1331 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1334 register struct ip *ip;
1338 * Make a deep copy of the packet because we're going to
1339 * modify the pack in order to generate checksums.
1341 copym = m_dup(m, M_NOWAIT);
1342 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1343 copym = m_pullup(copym, hlen);
1344 if (copym != NULL) {
1345 /* If needed, compute the checksum and mark it as valid. */
1346 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1347 in_delayed_cksum(copym);
1348 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1349 copym->m_pkthdr.csum_flags |=
1350 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1351 copym->m_pkthdr.csum_data = 0xffff;
1354 * We don't bother to fragment if the IP length is greater
1355 * than the interface's MTU. Can this possibly matter?
1357 ip = mtod(copym, struct ip *);
1359 ip->ip_sum = in_cksum(copym, hlen);
1361 if (dst->sin_family != AF_INET) {
1362 printf("ip_mloopback: bad address family %d\n",
1364 dst->sin_family = AF_INET;
1367 if_simloop(ifp, copym, dst->sin_family, 0);