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 * 3. 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_ratelimit.h"
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>
48 #include <sys/malloc.h>
52 #include <sys/protosw.h>
53 #include <sys/rmlock.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sysctl.h>
58 #include <sys/ucred.h>
61 #include <net/if_var.h>
62 #include <net/if_llatbl.h>
63 #include <net/netisr.h>
65 #include <net/route.h>
66 #include <net/flowtable.h>
68 #include <net/radix_mpath.h>
70 #include <net/rss_config.h>
73 #include <netinet/in.h>
74 #include <netinet/in_kdtrace.h>
75 #include <netinet/in_systm.h>
76 #include <netinet/ip.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/in_rss.h>
79 #include <netinet/in_var.h>
80 #include <netinet/ip_var.h>
81 #include <netinet/ip_options.h>
83 #include <netinet/sctp.h>
84 #include <netinet/sctp_crc32.h>
87 #include <netipsec/ipsec_support.h>
89 #include <machine/in_cksum.h>
91 #include <security/mac/mac_framework.h>
93 #ifdef MBUF_STRESS_TEST
94 static int mbuf_frag_size = 0;
95 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
96 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
99 static void ip_mloopback(struct ifnet *, const struct mbuf *, int);
102 extern int in_mcast_loop;
103 extern struct protosw inetsw[];
106 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
107 struct sockaddr_in *dst, int *fibnum, int *error)
109 struct m_tag *fwd_tag = NULL;
115 ip = mtod(m, struct ip *);
117 /* Run through list of hooks for output packets. */
118 odst.s_addr = ip->ip_dst.s_addr;
119 *error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, inp);
121 if ((*error) != 0 || m == NULL)
122 return 1; /* Finished */
124 ip = mtod(m, struct ip *);
126 /* See if destination IP address was changed by packet filter. */
127 if (odst.s_addr != ip->ip_dst.s_addr) {
128 m->m_flags |= M_SKIP_FIREWALL;
129 /* If destination is now ourself drop to ip_input(). */
130 if (in_localip(ip->ip_dst)) {
131 m->m_flags |= M_FASTFWD_OURS;
132 if (m->m_pkthdr.rcvif == NULL)
133 m->m_pkthdr.rcvif = V_loif;
134 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
135 m->m_pkthdr.csum_flags |=
136 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
137 m->m_pkthdr.csum_data = 0xffff;
139 m->m_pkthdr.csum_flags |=
140 CSUM_IP_CHECKED | CSUM_IP_VALID;
142 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
143 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
145 *error = netisr_queue(NETISR_IP, m);
146 return 1; /* Finished */
149 bzero(dst, sizeof(*dst));
150 dst->sin_family = AF_INET;
151 dst->sin_len = sizeof(*dst);
152 dst->sin_addr = ip->ip_dst;
154 return -1; /* Reloop */
156 /* See if fib was changed by packet filter. */
157 if ((*fibnum) != M_GETFIB(m)) {
158 m->m_flags |= M_SKIP_FIREWALL;
159 *fibnum = M_GETFIB(m);
160 return -1; /* Reloop for FIB change */
163 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
164 if (m->m_flags & M_FASTFWD_OURS) {
165 if (m->m_pkthdr.rcvif == NULL)
166 m->m_pkthdr.rcvif = V_loif;
167 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
168 m->m_pkthdr.csum_flags |=
169 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
170 m->m_pkthdr.csum_data = 0xffff;
173 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
174 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
176 m->m_pkthdr.csum_flags |=
177 CSUM_IP_CHECKED | CSUM_IP_VALID;
179 *error = netisr_queue(NETISR_IP, m);
180 return 1; /* Finished */
182 /* Or forward to some other address? */
183 if ((m->m_flags & M_IP_NEXTHOP) &&
184 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
185 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
186 m->m_flags |= M_SKIP_FIREWALL;
187 m->m_flags &= ~M_IP_NEXTHOP;
188 m_tag_delete(m, fwd_tag);
190 return -1; /* Reloop for CHANGE of dst */
197 * IP output. The packet in mbuf chain m contains a skeletal IP
198 * header (with len, off, ttl, proto, tos, src, dst).
199 * The mbuf chain containing the packet will be freed.
200 * The mbuf opt, if present, will not be freed.
201 * If route ro is present and has ro_rt initialized, route lookup would be
202 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
203 * then result of route lookup is stored in ro->ro_rt.
205 * In the IP forwarding case, the packet will arrive with options already
206 * inserted, so must have a NULL opt pointer.
209 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
210 struct ip_moptions *imo, struct inpcb *inp)
212 struct rm_priotracker in_ifa_tracker;
214 struct ifnet *ifp = NULL; /* keep compiler happy */
216 int hlen = sizeof (struct ip);
219 struct sockaddr_in *dst;
220 const struct sockaddr_in *gw;
221 struct in_ifaddr *ia;
223 uint16_t ip_len, ip_off;
224 struct route iproute;
225 struct rtentry *rte; /* cache for ro->ro_rt */
228 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
229 int no_route_but_check_spd = 0;
234 INP_LOCK_ASSERT(inp);
235 M_SETFIB(m, inp->inp_inc.inc_fibnum);
236 if ((flags & IP_NODEFAULTFLOWID) == 0) {
237 m->m_pkthdr.flowid = inp->inp_flowid;
238 M_HASHTYPE_SET(m, inp->inp_flowtype);
244 bzero(ro, sizeof (*ro));
248 if (ro->ro_rt == NULL)
249 (void )flowtable_lookup(AF_INET, m, ro);
254 m = ip_insertoptions(m, opt, &len);
256 hlen = len; /* ip->ip_hl is updated above */
258 ip = mtod(m, struct ip *);
259 ip_len = ntohs(ip->ip_len);
260 ip_off = ntohs(ip->ip_off);
262 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
263 ip->ip_v = IPVERSION;
264 ip->ip_hl = hlen >> 2;
266 IPSTAT_INC(ips_localout);
268 /* Header already set, fetch hlen from there */
269 hlen = ip->ip_hl << 2;
275 * dst can be rewritten but always points to &ro->ro_dst.
276 * gw is readonly but can point either to dst OR rt_gateway,
277 * therefore we need restore gw if we're redoing lookup.
279 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
280 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
283 bzero(dst, sizeof(*dst));
284 dst->sin_family = AF_INET;
285 dst->sin_len = sizeof(*dst);
286 dst->sin_addr = ip->ip_dst;
290 * Validate route against routing table additions;
291 * a better/more specific route might have been added.
294 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
296 * If there is a cached route,
297 * check that it is to the same destination
298 * and is still up. If not, free it and try again.
299 * The address family should also be checked in case of sharing the
301 * Also check whether routing cache needs invalidation.
304 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
305 rte->rt_ifp == NULL ||
306 !RT_LINK_IS_UP(rte->rt_ifp) ||
307 dst->sin_family != AF_INET ||
308 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
310 rte = ro->ro_rt = (struct rtentry *)NULL;
312 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
313 ro->ro_lle = (struct llentry *)NULL;
318 * If routing to interface only, short circuit routing lookup.
319 * The use of an all-ones broadcast address implies this; an
320 * interface is specified by the broadcast address of an interface,
321 * or the destination address of a ptp interface.
323 if (flags & IP_SENDONES) {
324 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
325 M_GETFIB(m)))) == NULL &&
326 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
327 M_GETFIB(m)))) == NULL) {
328 IPSTAT_INC(ips_noroute);
333 ip->ip_dst.s_addr = INADDR_BROADCAST;
334 dst->sin_addr = ip->ip_dst;
338 } else if (flags & IP_ROUTETOIF) {
339 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
340 M_GETFIB(m)))) == NULL &&
341 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
342 M_GETFIB(m)))) == NULL) {
343 IPSTAT_INC(ips_noroute);
350 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
351 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
352 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
353 imo != NULL && imo->imo_multicast_ifp != NULL) {
355 * Bypass the normal routing lookup for multicast
356 * packets if the interface is specified.
358 ifp = imo->imo_multicast_ifp;
359 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
362 isbroadcast = 0; /* fool gcc */
365 * We want to do any cloning requested by the link layer,
366 * as this is probably required in all cases for correct
367 * operation (as it is for ARP).
371 rtalloc_mpath_fib(ro,
372 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
375 in_rtalloc_ign(ro, 0, fibnum);
380 (rte->rt_flags & RTF_UP) == 0 ||
381 rte->rt_ifp == NULL ||
382 !RT_LINK_IS_UP(rte->rt_ifp)) {
383 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
385 * There is no route for this packet, but it is
386 * possible that a matching SPD entry exists.
388 no_route_but_check_spd = 1;
389 mtu = 0; /* Silence GCC warning. */
392 IPSTAT_INC(ips_noroute);
393 error = EHOSTUNREACH;
396 ia = ifatoia(rte->rt_ifa);
398 counter_u64_add(rte->rt_pksent, 1);
399 rt_update_ro_flags(ro);
400 if (rte->rt_flags & RTF_GATEWAY)
401 gw = (struct sockaddr_in *)rte->rt_gateway;
402 if (rte->rt_flags & RTF_HOST)
403 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
404 else if (ifp->if_flags & IFF_BROADCAST)
405 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
411 * Calculate MTU. If we have a route that is up, use that,
412 * otherwise use the interface's MTU.
414 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
418 /* Catch a possible divide by zero later. */
419 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
420 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
422 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
423 m->m_flags |= M_MCAST;
425 * IP destination address is multicast. Make sure "gw"
426 * still points to the address in "ro". (It may have been
427 * changed to point to a gateway address, above.)
431 * See if the caller provided any multicast options
434 ip->ip_ttl = imo->imo_multicast_ttl;
435 if (imo->imo_multicast_vif != -1)
438 ip_mcast_src(imo->imo_multicast_vif) :
441 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
443 * Confirm that the outgoing interface supports multicast.
445 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
446 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
447 IPSTAT_INC(ips_noroute);
453 * If source address not specified yet, use address
454 * of outgoing interface.
456 if (ip->ip_src.s_addr == INADDR_ANY) {
457 /* Interface may have no addresses. */
459 ip->ip_src = IA_SIN(ia)->sin_addr;
462 if ((imo == NULL && in_mcast_loop) ||
463 (imo && imo->imo_multicast_loop)) {
465 * Loop back multicast datagram if not expressly
466 * forbidden to do so, even if we are not a member
467 * of the group; ip_input() will filter it later,
468 * thus deferring a hash lookup and mutex acquisition
469 * at the expense of a cheap copy using m_copym().
471 ip_mloopback(ifp, m, hlen);
474 * If we are acting as a multicast router, perform
475 * multicast forwarding as if the packet had just
476 * arrived on the interface to which we are about
477 * to send. The multicast forwarding function
478 * recursively calls this function, using the
479 * IP_FORWARDING flag to prevent infinite recursion.
481 * Multicasts that are looped back by ip_mloopback(),
482 * above, will be forwarded by the ip_input() routine,
485 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
487 * If rsvp daemon is not running, do not
488 * set ip_moptions. This ensures that the packet
489 * is multicast and not just sent down one link
490 * as prescribed by rsvpd.
495 ip_mforward(ip, ifp, m, imo) != 0) {
503 * Multicasts with a time-to-live of zero may be looped-
504 * back, above, but must not be transmitted on a network.
505 * Also, multicasts addressed to the loopback interface
506 * are not sent -- the above call to ip_mloopback() will
507 * loop back a copy. ip_input() will drop the copy if
508 * this host does not belong to the destination group on
509 * the loopback interface.
511 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
520 * If the source address is not specified yet, use the address
521 * of the outoing interface.
523 if (ip->ip_src.s_addr == INADDR_ANY) {
524 /* Interface may have no addresses. */
526 ip->ip_src = IA_SIN(ia)->sin_addr;
531 * Look for broadcast address and
532 * verify user is allowed to send
536 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
537 error = EADDRNOTAVAIL;
540 if ((flags & IP_ALLOWBROADCAST) == 0) {
544 /* don't allow broadcast messages to be fragmented */
549 m->m_flags |= M_BCAST;
551 m->m_flags &= ~M_BCAST;
555 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
556 if (IPSEC_ENABLED(ipv4)) {
557 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
558 if (error == EINPROGRESS)
564 * Check if there was a route for this packet; return error if not.
566 if (no_route_but_check_spd) {
567 IPSTAT_INC(ips_noroute);
568 error = EHOSTUNREACH;
571 /* Update variables that are affected by ipsec4_output(). */
572 ip = mtod(m, struct ip *);
573 hlen = ip->ip_hl << 2;
576 /* Jump over all PFIL processing if hooks are not active. */
577 if (PFIL_HOOKED(&V_inet_pfil_hook)) {
578 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
579 case 1: /* Finished */
582 case 0: /* Continue normally */
583 ip = mtod(m, struct ip *);
586 case -1: /* Need to try again */
587 /* Reset everything for a new round */
590 ifa_free(&ia->ia_ifa);
591 ro->ro_prepend = NULL;
594 ip = mtod(m, struct ip *);
600 /* 127/8 must not appear on wire - RFC1122. */
601 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
602 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
603 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
604 IPSTAT_INC(ips_badaddr);
605 error = EADDRNOTAVAIL;
610 m->m_pkthdr.csum_flags |= CSUM_IP;
611 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
613 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
616 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
617 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
618 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
623 * If small enough for interface, or the interface will take
624 * care of the fragmentation for us, we can just send directly.
627 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
629 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
630 ip->ip_sum = in_cksum(m, hlen);
631 m->m_pkthdr.csum_flags &= ~CSUM_IP;
635 * Record statistics for this interface address.
636 * With CSUM_TSO the byte/packet count will be slightly
637 * incorrect because we count the IP+TCP headers only
638 * once instead of for every generated packet.
640 if (!(flags & IP_FORWARDING) && ia) {
641 if (m->m_pkthdr.csum_flags & CSUM_TSO)
642 counter_u64_add(ia->ia_ifa.ifa_opackets,
643 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
645 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
647 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
649 #ifdef MBUF_STRESS_TEST
650 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
651 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
654 * Reset layer specific mbuf flags
655 * to avoid confusing lower layers.
658 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
661 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
662 in_pcboutput_txrtlmt(inp, ifp, m);
663 /* stamp send tag on mbuf */
664 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
666 m->m_pkthdr.snd_tag = NULL;
669 error = (*ifp->if_output)(ifp, m,
670 (const struct sockaddr *)gw, ro);
672 /* check for route change */
674 in_pcboutput_eagain(inp);
679 /* Balk when DF bit is set or the interface didn't support TSO. */
680 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
682 IPSTAT_INC(ips_cantfrag);
687 * Too large for interface; fragment if possible. If successful,
688 * on return, m will point to a list of packets to be sent.
690 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
697 /* Record statistics for this interface address. */
699 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
700 counter_u64_add(ia->ia_ifa.ifa_obytes,
704 * Reset layer specific mbuf flags
705 * to avoid confusing upper layers.
709 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
710 mtod(m, struct ip *), NULL);
713 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
714 in_pcboutput_txrtlmt(inp, ifp, m);
715 /* stamp send tag on mbuf */
716 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
718 m->m_pkthdr.snd_tag = NULL;
721 error = (*ifp->if_output)(ifp, m,
722 (const struct sockaddr *)gw, ro);
724 /* check for route change */
726 in_pcboutput_eagain(inp);
733 IPSTAT_INC(ips_fragmented);
738 else if (rte == NULL)
740 * If the caller supplied a route but somehow the reference
741 * to it has been released need to prevent the caller
742 * calling RTFREE on it again.
746 ifa_free(&ia->ia_ifa);
754 * Create a chain of fragments which fit the given mtu. m_frag points to the
755 * mbuf to be fragmented; on return it points to the chain with the fragments.
756 * Return 0 if no error. If error, m_frag may contain a partially built
757 * chain of fragments that should be freed by the caller.
759 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
762 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
763 u_long if_hwassist_flags)
766 int hlen = ip->ip_hl << 2;
767 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
769 struct mbuf *m0 = *m_frag; /* the original packet */
773 uint16_t ip_len, ip_off;
775 ip_len = ntohs(ip->ip_len);
776 ip_off = ntohs(ip->ip_off);
778 if (ip_off & IP_DF) { /* Fragmentation not allowed */
779 IPSTAT_INC(ips_cantfrag);
784 * Must be able to put at least 8 bytes per fragment.
790 * If the interface will not calculate checksums on
791 * fragmented packets, then do it here.
793 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
794 in_delayed_cksum(m0);
795 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
798 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
799 sctp_delayed_cksum(m0, hlen);
800 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
803 if (len > PAGE_SIZE) {
805 * Fragment large datagrams such that each segment
806 * contains a multiple of PAGE_SIZE amount of data,
807 * plus headers. This enables a receiver to perform
808 * page-flipping zero-copy optimizations.
810 * XXX When does this help given that sender and receiver
811 * could have different page sizes, and also mtu could
812 * be less than the receiver's page size ?
816 off = MIN(mtu, m0->m_pkthdr.len);
819 * firstlen (off - hlen) must be aligned on an
823 goto smart_frag_failure;
824 off = ((off - hlen) & ~7) + hlen;
825 newlen = (~PAGE_MASK) & mtu;
826 if ((newlen + sizeof (struct ip)) > mtu) {
827 /* we failed, go back the default */
838 firstlen = off - hlen;
839 mnext = &m0->m_nextpkt; /* pointer to next packet */
842 * Loop through length of segment after first fragment,
843 * make new header and copy data of each part and link onto chain.
844 * Here, m0 is the original packet, m is the fragment being created.
845 * The fragments are linked off the m_nextpkt of the original
846 * packet, which after processing serves as the first fragment.
848 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
849 struct ip *mhip; /* ip header on the fragment */
851 int mhlen = sizeof (struct ip);
853 m = m_gethdr(M_NOWAIT, MT_DATA);
856 IPSTAT_INC(ips_odropped);
860 * Make sure the complete packet header gets copied
861 * from the originating mbuf to the newly created
862 * mbuf. This also ensures that existing firewall
863 * classification(s), VLAN tags and so on get copied
864 * to the resulting fragmented packet(s):
866 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
869 IPSTAT_INC(ips_odropped);
873 * In the first mbuf, leave room for the link header, then
874 * copy the original IP header including options. The payload
875 * goes into an additional mbuf chain returned by m_copym().
877 m->m_data += max_linkhdr;
878 mhip = mtod(m, struct ip *);
880 if (hlen > sizeof (struct ip)) {
881 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
882 mhip->ip_v = IPVERSION;
883 mhip->ip_hl = mhlen >> 2;
886 /* XXX do we need to add ip_off below ? */
887 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
888 if (off + len >= ip_len)
891 mhip->ip_off |= IP_MF;
892 mhip->ip_len = htons((u_short)(len + mhlen));
893 m->m_next = m_copym(m0, off, len, M_NOWAIT);
894 if (m->m_next == NULL) { /* copy failed */
896 error = ENOBUFS; /* ??? */
897 IPSTAT_INC(ips_odropped);
900 m->m_pkthdr.len = mhlen + len;
902 mac_netinet_fragment(m0, m);
904 mhip->ip_off = htons(mhip->ip_off);
906 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
907 mhip->ip_sum = in_cksum(m, mhlen);
908 m->m_pkthdr.csum_flags &= ~CSUM_IP;
911 mnext = &m->m_nextpkt;
913 IPSTAT_ADD(ips_ofragments, nfrags);
916 * Update first fragment by trimming what's been copied out
917 * and updating header.
919 m_adj(m0, hlen + firstlen - ip_len);
920 m0->m_pkthdr.len = hlen + firstlen;
921 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
922 ip->ip_off = htons(ip_off | IP_MF);
924 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
925 ip->ip_sum = in_cksum(m0, hlen);
926 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
935 in_delayed_cksum(struct mbuf *m)
938 uint16_t csum, offset, ip_len;
940 ip = mtod(m, struct ip *);
941 offset = ip->ip_hl << 2 ;
942 ip_len = ntohs(ip->ip_len);
943 csum = in_cksum_skip(m, ip_len, offset);
944 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
946 offset += m->m_pkthdr.csum_data; /* checksum offset */
948 /* find the mbuf in the chain where the checksum starts*/
949 while ((m != NULL) && (offset >= m->m_len)) {
953 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
954 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
955 *(u_short *)(m->m_data + offset) = csum;
959 * IP socket option processing.
962 ip_ctloutput(struct socket *so, struct sockopt *sopt)
964 struct inpcb *inp = sotoinpcb(so);
972 if (sopt->sopt_level != IPPROTO_IP) {
975 if (sopt->sopt_level == SOL_SOCKET &&
976 sopt->sopt_dir == SOPT_SET) {
977 switch (sopt->sopt_name) {
980 if ((so->so_options & SO_REUSEADDR) != 0)
981 inp->inp_flags2 |= INP_REUSEADDR;
983 inp->inp_flags2 &= ~INP_REUSEADDR;
989 if ((so->so_options & SO_REUSEPORT) != 0)
990 inp->inp_flags2 |= INP_REUSEPORT;
992 inp->inp_flags2 &= ~INP_REUSEPORT;
998 inp->inp_inc.inc_fibnum = so->so_fibnum;
1002 case SO_MAX_PACING_RATE:
1005 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1019 switch (sopt->sopt_dir) {
1021 switch (sopt->sopt_name) {
1028 if (sopt->sopt_valsize > MLEN) {
1032 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1037 m->m_len = sopt->sopt_valsize;
1038 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1045 error = ip_pcbopts(inp, sopt->sopt_name, m);
1051 if (sopt->sopt_td != NULL) {
1052 error = priv_check(sopt->sopt_td,
1053 PRIV_NETINET_BINDANY);
1060 case IP_RSS_LISTEN_BUCKET:
1066 case IP_RECVRETOPTS:
1067 case IP_ORIGDSTADDR:
1068 case IP_RECVDSTADDR:
1076 case IP_RECVRSSBUCKETID:
1078 error = sooptcopyin(sopt, &optval, sizeof optval,
1083 switch (sopt->sopt_name) {
1085 inp->inp_ip_tos = optval;
1089 inp->inp_ip_ttl = optval;
1093 if (optval >= 0 && optval <= MAXTTL)
1094 inp->inp_ip_minttl = optval;
1099 #define OPTSET(bit) do { \
1102 inp->inp_flags |= bit; \
1104 inp->inp_flags &= ~bit; \
1108 #define OPTSET2(bit, val) do { \
1111 inp->inp_flags2 |= bit; \
1113 inp->inp_flags2 &= ~bit; \
1118 OPTSET(INP_RECVOPTS);
1121 case IP_RECVRETOPTS:
1122 OPTSET(INP_RECVRETOPTS);
1125 case IP_RECVDSTADDR:
1126 OPTSET(INP_RECVDSTADDR);
1129 case IP_ORIGDSTADDR:
1130 OPTSET2(INP_ORIGDSTADDR, optval);
1134 OPTSET(INP_RECVTTL);
1142 OPTSET(INP_ONESBCAST);
1145 OPTSET(INP_DONTFRAG);
1148 OPTSET(INP_BINDANY);
1151 OPTSET(INP_RECVTOS);
1154 OPTSET2(INP_BINDMULTI, optval);
1157 OPTSET2(INP_RECVFLOWID, optval);
1160 case IP_RSS_LISTEN_BUCKET:
1161 if ((optval >= 0) &&
1162 (optval < rss_getnumbuckets())) {
1163 inp->inp_rss_listen_bucket = optval;
1164 OPTSET2(INP_RSS_BUCKET_SET, 1);
1169 case IP_RECVRSSBUCKETID:
1170 OPTSET2(INP_RECVRSSBUCKETID, optval);
1179 * Multicast socket options are processed by the in_mcast
1182 case IP_MULTICAST_IF:
1183 case IP_MULTICAST_VIF:
1184 case IP_MULTICAST_TTL:
1185 case IP_MULTICAST_LOOP:
1186 case IP_ADD_MEMBERSHIP:
1187 case IP_DROP_MEMBERSHIP:
1188 case IP_ADD_SOURCE_MEMBERSHIP:
1189 case IP_DROP_SOURCE_MEMBERSHIP:
1190 case IP_BLOCK_SOURCE:
1191 case IP_UNBLOCK_SOURCE:
1193 case MCAST_JOIN_GROUP:
1194 case MCAST_LEAVE_GROUP:
1195 case MCAST_JOIN_SOURCE_GROUP:
1196 case MCAST_LEAVE_SOURCE_GROUP:
1197 case MCAST_BLOCK_SOURCE:
1198 case MCAST_UNBLOCK_SOURCE:
1199 error = inp_setmoptions(inp, sopt);
1203 error = sooptcopyin(sopt, &optval, sizeof optval,
1210 case IP_PORTRANGE_DEFAULT:
1211 inp->inp_flags &= ~(INP_LOWPORT);
1212 inp->inp_flags &= ~(INP_HIGHPORT);
1215 case IP_PORTRANGE_HIGH:
1216 inp->inp_flags &= ~(INP_LOWPORT);
1217 inp->inp_flags |= INP_HIGHPORT;
1220 case IP_PORTRANGE_LOW:
1221 inp->inp_flags &= ~(INP_HIGHPORT);
1222 inp->inp_flags |= INP_LOWPORT;
1232 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1233 case IP_IPSEC_POLICY:
1234 if (IPSEC_ENABLED(ipv4)) {
1235 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1242 error = ENOPROTOOPT;
1248 switch (sopt->sopt_name) {
1251 if (inp->inp_options)
1252 error = sooptcopyout(sopt,
1253 mtod(inp->inp_options,
1255 inp->inp_options->m_len);
1257 sopt->sopt_valsize = 0;
1264 case IP_RECVRETOPTS:
1265 case IP_ORIGDSTADDR:
1266 case IP_RECVDSTADDR:
1279 case IP_RSSBUCKETID:
1280 case IP_RECVRSSBUCKETID:
1282 switch (sopt->sopt_name) {
1285 optval = inp->inp_ip_tos;
1289 optval = inp->inp_ip_ttl;
1293 optval = inp->inp_ip_minttl;
1296 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1297 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1300 optval = OPTBIT(INP_RECVOPTS);
1303 case IP_RECVRETOPTS:
1304 optval = OPTBIT(INP_RECVRETOPTS);
1307 case IP_RECVDSTADDR:
1308 optval = OPTBIT(INP_RECVDSTADDR);
1311 case IP_ORIGDSTADDR:
1312 optval = OPTBIT2(INP_ORIGDSTADDR);
1316 optval = OPTBIT(INP_RECVTTL);
1320 optval = OPTBIT(INP_RECVIF);
1324 if (inp->inp_flags & INP_HIGHPORT)
1325 optval = IP_PORTRANGE_HIGH;
1326 else if (inp->inp_flags & INP_LOWPORT)
1327 optval = IP_PORTRANGE_LOW;
1333 optval = OPTBIT(INP_ONESBCAST);
1336 optval = OPTBIT(INP_DONTFRAG);
1339 optval = OPTBIT(INP_BINDANY);
1342 optval = OPTBIT(INP_RECVTOS);
1345 optval = inp->inp_flowid;
1348 optval = inp->inp_flowtype;
1351 optval = OPTBIT2(INP_RECVFLOWID);
1354 case IP_RSSBUCKETID:
1355 retval = rss_hash2bucket(inp->inp_flowid,
1359 optval = rss_bucket;
1363 case IP_RECVRSSBUCKETID:
1364 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1368 optval = OPTBIT2(INP_BINDMULTI);
1371 error = sooptcopyout(sopt, &optval, sizeof optval);
1375 * Multicast socket options are processed by the in_mcast
1378 case IP_MULTICAST_IF:
1379 case IP_MULTICAST_VIF:
1380 case IP_MULTICAST_TTL:
1381 case IP_MULTICAST_LOOP:
1383 error = inp_getmoptions(inp, sopt);
1386 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1387 case IP_IPSEC_POLICY:
1388 if (IPSEC_ENABLED(ipv4)) {
1389 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1396 error = ENOPROTOOPT;
1405 * Routine called from ip_output() to loop back a copy of an IP multicast
1406 * packet to the input queue of a specified interface. Note that this
1407 * calls the output routine of the loopback "driver", but with an interface
1408 * pointer that might NOT be a loopback interface -- evil, but easier than
1409 * replicating that code here.
1412 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1418 * Make a deep copy of the packet because we're going to
1419 * modify the pack in order to generate checksums.
1421 copym = m_dup(m, M_NOWAIT);
1422 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1423 copym = m_pullup(copym, hlen);
1424 if (copym != NULL) {
1425 /* If needed, compute the checksum and mark it as valid. */
1426 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1427 in_delayed_cksum(copym);
1428 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1429 copym->m_pkthdr.csum_flags |=
1430 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1431 copym->m_pkthdr.csum_data = 0xffff;
1434 * We don't bother to fragment if the IP length is greater
1435 * than the interface's MTU. Can this possibly matter?
1437 ip = mtod(copym, struct ip *);
1439 ip->ip_sum = in_cksum(copym, hlen);
1440 if_simloop(ifp, copym, AF_INET, 0);