2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1990, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
38 #include "opt_ratelimit.h"
39 #include "opt_ipsec.h"
40 #include "opt_mbuf_stress_test.h"
41 #include "opt_mpath.h"
42 #include "opt_route.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
50 #include <sys/malloc.h>
54 #include <sys/protosw.h>
55 #include <sys/rmlock.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/sysctl.h>
60 #include <sys/ucred.h>
63 #include <net/if_var.h>
64 #include <net/if_llatbl.h>
65 #include <net/netisr.h>
67 #include <net/route.h>
69 #include <net/radix_mpath.h>
71 #include <net/rss_config.h>
74 #include <netinet/in.h>
75 #include <netinet/in_kdtrace.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/ip.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/in_rss.h>
80 #include <netinet/in_var.h>
81 #include <netinet/ip_var.h>
82 #include <netinet/ip_options.h>
84 #include <netinet/udp.h>
85 #include <netinet/udp_var.h>
88 #include <netinet/sctp.h>
89 #include <netinet/sctp_crc32.h>
92 #include <netipsec/ipsec_support.h>
94 #include <machine/in_cksum.h>
96 #include <security/mac/mac_framework.h>
98 #ifdef MBUF_STRESS_TEST
99 static int mbuf_frag_size = 0;
100 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
101 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
104 static void ip_mloopback(struct ifnet *, const struct mbuf *, int);
107 extern int in_mcast_loop;
108 extern struct protosw inetsw[];
111 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
112 struct sockaddr_in *dst, int *fibnum, int *error)
114 struct m_tag *fwd_tag = NULL;
120 ip = mtod(m, struct ip *);
122 /* Run through list of hooks for output packets. */
123 odst.s_addr = ip->ip_dst.s_addr;
124 *error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, 0, inp);
126 if ((*error) != 0 || m == NULL)
127 return 1; /* Finished */
129 ip = mtod(m, struct ip *);
131 /* See if destination IP address was changed by packet filter. */
132 if (odst.s_addr != ip->ip_dst.s_addr) {
133 m->m_flags |= M_SKIP_FIREWALL;
134 /* If destination is now ourself drop to ip_input(). */
135 if (in_localip(ip->ip_dst)) {
136 m->m_flags |= M_FASTFWD_OURS;
137 if (m->m_pkthdr.rcvif == NULL)
138 m->m_pkthdr.rcvif = V_loif;
139 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
140 m->m_pkthdr.csum_flags |=
141 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
142 m->m_pkthdr.csum_data = 0xffff;
144 m->m_pkthdr.csum_flags |=
145 CSUM_IP_CHECKED | CSUM_IP_VALID;
147 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
148 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
150 *error = netisr_queue(NETISR_IP, m);
151 return 1; /* Finished */
154 bzero(dst, sizeof(*dst));
155 dst->sin_family = AF_INET;
156 dst->sin_len = sizeof(*dst);
157 dst->sin_addr = ip->ip_dst;
159 return -1; /* Reloop */
161 /* See if fib was changed by packet filter. */
162 if ((*fibnum) != M_GETFIB(m)) {
163 m->m_flags |= M_SKIP_FIREWALL;
164 *fibnum = M_GETFIB(m);
165 return -1; /* Reloop for FIB change */
168 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
169 if (m->m_flags & M_FASTFWD_OURS) {
170 if (m->m_pkthdr.rcvif == NULL)
171 m->m_pkthdr.rcvif = V_loif;
172 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
173 m->m_pkthdr.csum_flags |=
174 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
175 m->m_pkthdr.csum_data = 0xffff;
178 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
179 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
181 m->m_pkthdr.csum_flags |=
182 CSUM_IP_CHECKED | CSUM_IP_VALID;
184 *error = netisr_queue(NETISR_IP, m);
185 return 1; /* Finished */
187 /* Or forward to some other address? */
188 if ((m->m_flags & M_IP_NEXTHOP) &&
189 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
190 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
191 m->m_flags |= M_SKIP_FIREWALL;
192 m->m_flags &= ~M_IP_NEXTHOP;
193 m_tag_delete(m, fwd_tag);
195 return -1; /* Reloop for CHANGE of dst */
202 * IP output. The packet in mbuf chain m contains a skeletal IP
203 * header (with len, off, ttl, proto, tos, src, dst).
204 * The mbuf chain containing the packet will be freed.
205 * The mbuf opt, if present, will not be freed.
206 * If route ro is present and has ro_rt initialized, route lookup would be
207 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
208 * then result of route lookup is stored in ro->ro_rt.
210 * In the IP forwarding case, the packet will arrive with options already
211 * inserted, so must have a NULL opt pointer.
214 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
215 struct ip_moptions *imo, struct inpcb *inp)
217 struct rm_priotracker in_ifa_tracker;
219 struct ifnet *ifp = NULL; /* keep compiler happy */
221 int hlen = sizeof (struct ip);
224 struct sockaddr_in *dst;
225 const struct sockaddr_in *gw;
226 struct in_ifaddr *ia;
228 uint16_t ip_len, ip_off;
229 struct route iproute;
230 struct rtentry *rte; /* cache for ro->ro_rt */
232 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
233 int no_route_but_check_spd = 0;
238 INP_LOCK_ASSERT(inp);
239 M_SETFIB(m, inp->inp_inc.inc_fibnum);
240 if ((flags & IP_NODEFAULTFLOWID) == 0) {
241 m->m_pkthdr.flowid = inp->inp_flowid;
242 M_HASHTYPE_SET(m, inp->inp_flowtype);
248 bzero(ro, sizeof (*ro));
253 m = ip_insertoptions(m, opt, &len);
255 hlen = len; /* ip->ip_hl is updated above */
257 ip = mtod(m, struct ip *);
258 ip_len = ntohs(ip->ip_len);
259 ip_off = ntohs(ip->ip_off);
261 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
262 ip->ip_v = IPVERSION;
263 ip->ip_hl = hlen >> 2;
266 /* Header already set, fetch hlen from there */
267 hlen = ip->ip_hl << 2;
269 if ((flags & IP_FORWARDING) == 0)
270 IPSTAT_INC(ips_localout);
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;
291 * Validate route against routing table additions;
292 * a better/more specific route might have been added.
295 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
297 * If there is a cached route,
298 * check that it is to the same destination
299 * and is still up. If not, free it and try again.
300 * The address family should also be checked in case of sharing the
302 * Also check whether routing cache needs invalidation.
305 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
306 rte->rt_ifp == NULL ||
307 !RT_LINK_IS_UP(rte->rt_ifp) ||
308 dst->sin_family != AF_INET ||
309 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
310 RO_INVALIDATE_CACHE(ro);
315 * If routing to interface only, short circuit routing lookup.
316 * The use of an all-ones broadcast address implies this; an
317 * interface is specified by the broadcast address of an interface,
318 * or the destination address of a ptp interface.
320 if (flags & IP_SENDONES) {
321 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
322 M_GETFIB(m)))) == NULL &&
323 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
324 M_GETFIB(m)))) == NULL) {
325 IPSTAT_INC(ips_noroute);
329 ip->ip_dst.s_addr = INADDR_BROADCAST;
330 dst->sin_addr = ip->ip_dst;
334 } else if (flags & IP_ROUTETOIF) {
335 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
336 M_GETFIB(m)))) == NULL &&
337 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
338 M_GETFIB(m)))) == NULL) {
339 IPSTAT_INC(ips_noroute);
345 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
346 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
347 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
348 imo != NULL && imo->imo_multicast_ifp != NULL) {
350 * Bypass the normal routing lookup for multicast
351 * packets if the interface is specified.
353 ifp = imo->imo_multicast_ifp;
354 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
355 isbroadcast = 0; /* fool gcc */
358 * We want to do any cloning requested by the link layer,
359 * as this is probably required in all cases for correct
360 * operation (as it is for ARP).
364 rtalloc_mpath_fib(ro,
365 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
368 in_rtalloc_ign(ro, 0, fibnum);
373 (rte->rt_flags & RTF_UP) == 0 ||
374 rte->rt_ifp == NULL ||
375 !RT_LINK_IS_UP(rte->rt_ifp)) {
376 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
378 * There is no route for this packet, but it is
379 * possible that a matching SPD entry exists.
381 no_route_but_check_spd = 1;
382 mtu = 0; /* Silence GCC warning. */
385 IPSTAT_INC(ips_noroute);
386 error = EHOSTUNREACH;
389 ia = ifatoia(rte->rt_ifa);
391 counter_u64_add(rte->rt_pksent, 1);
392 rt_update_ro_flags(ro);
393 if (rte->rt_flags & RTF_GATEWAY)
394 gw = (struct sockaddr_in *)rte->rt_gateway;
395 if (rte->rt_flags & RTF_HOST)
396 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
397 else if (ifp->if_flags & IFF_BROADCAST)
398 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
404 * Calculate MTU. If we have a route that is up, use that,
405 * otherwise use the interface's MTU.
407 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
411 /* Catch a possible divide by zero later. */
412 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
413 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
415 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
416 m->m_flags |= M_MCAST;
418 * IP destination address is multicast. Make sure "gw"
419 * still points to the address in "ro". (It may have been
420 * changed to point to a gateway address, above.)
424 * See if the caller provided any multicast options
427 ip->ip_ttl = imo->imo_multicast_ttl;
428 if (imo->imo_multicast_vif != -1)
431 ip_mcast_src(imo->imo_multicast_vif) :
434 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
436 * Confirm that the outgoing interface supports multicast.
438 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
439 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
440 IPSTAT_INC(ips_noroute);
446 * If source address not specified yet, use address
447 * of outgoing interface.
449 if (ip->ip_src.s_addr == INADDR_ANY) {
450 /* Interface may have no addresses. */
452 ip->ip_src = IA_SIN(ia)->sin_addr;
455 if ((imo == NULL && in_mcast_loop) ||
456 (imo && imo->imo_multicast_loop)) {
458 * Loop back multicast datagram if not expressly
459 * forbidden to do so, even if we are not a member
460 * of the group; ip_input() will filter it later,
461 * thus deferring a hash lookup and mutex acquisition
462 * at the expense of a cheap copy using m_copym().
464 ip_mloopback(ifp, m, hlen);
467 * If we are acting as a multicast router, perform
468 * multicast forwarding as if the packet had just
469 * arrived on the interface to which we are about
470 * to send. The multicast forwarding function
471 * recursively calls this function, using the
472 * IP_FORWARDING flag to prevent infinite recursion.
474 * Multicasts that are looped back by ip_mloopback(),
475 * above, will be forwarded by the ip_input() routine,
478 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
480 * If rsvp daemon is not running, do not
481 * set ip_moptions. This ensures that the packet
482 * is multicast and not just sent down one link
483 * as prescribed by rsvpd.
488 ip_mforward(ip, ifp, m, imo) != 0) {
496 * Multicasts with a time-to-live of zero may be looped-
497 * back, above, but must not be transmitted on a network.
498 * Also, multicasts addressed to the loopback interface
499 * are not sent -- the above call to ip_mloopback() will
500 * loop back a copy. ip_input() will drop the copy if
501 * this host does not belong to the destination group on
502 * the loopback interface.
504 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
513 * If the source address is not specified yet, use the address
514 * of the outoing interface.
516 if (ip->ip_src.s_addr == INADDR_ANY) {
517 /* Interface may have no addresses. */
519 ip->ip_src = IA_SIN(ia)->sin_addr;
524 * Look for broadcast address and
525 * verify user is allowed to send
529 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
530 error = EADDRNOTAVAIL;
533 if ((flags & IP_ALLOWBROADCAST) == 0) {
537 /* don't allow broadcast messages to be fragmented */
542 m->m_flags |= M_BCAST;
544 m->m_flags &= ~M_BCAST;
548 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
549 if (IPSEC_ENABLED(ipv4)) {
550 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
551 if (error == EINPROGRESS)
557 * Check if there was a route for this packet; return error if not.
559 if (no_route_but_check_spd) {
560 IPSTAT_INC(ips_noroute);
561 error = EHOSTUNREACH;
564 /* Update variables that are affected by ipsec4_output(). */
565 ip = mtod(m, struct ip *);
566 hlen = ip->ip_hl << 2;
569 /* Jump over all PFIL processing if hooks are not active. */
570 if (PFIL_HOOKED(&V_inet_pfil_hook)) {
571 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
572 case 1: /* Finished */
575 case 0: /* Continue normally */
576 ip = mtod(m, struct ip *);
579 case -1: /* Need to try again */
580 /* Reset everything for a new round */
582 ro->ro_prepend = NULL;
585 ip = mtod(m, struct ip *);
591 /* IN_LOOPBACK must not appear on the wire - RFC1122. */
592 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
593 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
594 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
595 IPSTAT_INC(ips_badaddr);
596 error = EADDRNOTAVAIL;
601 m->m_pkthdr.csum_flags |= CSUM_IP;
602 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
604 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
607 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
608 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
609 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
614 * If small enough for interface, or the interface will take
615 * care of the fragmentation for us, we can just send directly.
618 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
620 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
621 ip->ip_sum = in_cksum(m, hlen);
622 m->m_pkthdr.csum_flags &= ~CSUM_IP;
626 * Record statistics for this interface address.
627 * With CSUM_TSO the byte/packet count will be slightly
628 * incorrect because we count the IP+TCP headers only
629 * once instead of for every generated packet.
631 if (!(flags & IP_FORWARDING) && ia) {
632 if (m->m_pkthdr.csum_flags & CSUM_TSO)
633 counter_u64_add(ia->ia_ifa.ifa_opackets,
634 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
636 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
638 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
640 #ifdef MBUF_STRESS_TEST
641 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
642 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
645 * Reset layer specific mbuf flags
646 * to avoid confusing lower layers.
649 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
652 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
653 in_pcboutput_txrtlmt(inp, ifp, m);
654 /* stamp send tag on mbuf */
655 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
656 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
658 m->m_pkthdr.snd_tag = NULL;
661 error = (*ifp->if_output)(ifp, m,
662 (const struct sockaddr *)gw, ro);
664 /* check for route change */
666 in_pcboutput_eagain(inp);
671 /* Balk when DF bit is set or the interface didn't support TSO. */
672 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
674 IPSTAT_INC(ips_cantfrag);
679 * Too large for interface; fragment if possible. If successful,
680 * on return, m will point to a list of packets to be sent.
682 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
689 /* Record statistics for this interface address. */
691 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
692 counter_u64_add(ia->ia_ifa.ifa_obytes,
696 * Reset layer specific mbuf flags
697 * to avoid confusing upper layers.
701 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
702 mtod(m, struct ip *), NULL);
705 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
706 in_pcboutput_txrtlmt(inp, ifp, m);
707 /* stamp send tag on mbuf */
708 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
709 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
711 m->m_pkthdr.snd_tag = NULL;
714 error = (*ifp->if_output)(ifp, m,
715 (const struct sockaddr *)gw, ro);
717 /* check for route change */
719 in_pcboutput_eagain(inp);
726 IPSTAT_INC(ips_fragmented);
731 else if (rte == NULL)
733 * If the caller supplied a route but somehow the reference
734 * to it has been released need to prevent the caller
735 * calling RTFREE on it again.
746 * Create a chain of fragments which fit the given mtu. m_frag points to the
747 * mbuf to be fragmented; on return it points to the chain with the fragments.
748 * Return 0 if no error. If error, m_frag may contain a partially built
749 * chain of fragments that should be freed by the caller.
751 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
754 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
755 u_long if_hwassist_flags)
758 int hlen = ip->ip_hl << 2;
759 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
761 struct mbuf *m0 = *m_frag; /* the original packet */
765 uint16_t ip_len, ip_off;
767 ip_len = ntohs(ip->ip_len);
768 ip_off = ntohs(ip->ip_off);
770 if (ip_off & IP_DF) { /* Fragmentation not allowed */
771 IPSTAT_INC(ips_cantfrag);
776 * Must be able to put at least 8 bytes per fragment.
782 * If the interface will not calculate checksums on
783 * fragmented packets, then do it here.
785 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
786 in_delayed_cksum(m0);
787 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
790 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
791 sctp_delayed_cksum(m0, hlen);
792 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
795 if (len > PAGE_SIZE) {
797 * Fragment large datagrams such that each segment
798 * contains a multiple of PAGE_SIZE amount of data,
799 * plus headers. This enables a receiver to perform
800 * page-flipping zero-copy optimizations.
802 * XXX When does this help given that sender and receiver
803 * could have different page sizes, and also mtu could
804 * be less than the receiver's page size ?
808 off = MIN(mtu, m0->m_pkthdr.len);
811 * firstlen (off - hlen) must be aligned on an
815 goto smart_frag_failure;
816 off = ((off - hlen) & ~7) + hlen;
817 newlen = (~PAGE_MASK) & mtu;
818 if ((newlen + sizeof (struct ip)) > mtu) {
819 /* we failed, go back the default */
830 firstlen = off - hlen;
831 mnext = &m0->m_nextpkt; /* pointer to next packet */
834 * Loop through length of segment after first fragment,
835 * make new header and copy data of each part and link onto chain.
836 * Here, m0 is the original packet, m is the fragment being created.
837 * The fragments are linked off the m_nextpkt of the original
838 * packet, which after processing serves as the first fragment.
840 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
841 struct ip *mhip; /* ip header on the fragment */
843 int mhlen = sizeof (struct ip);
845 m = m_gethdr(M_NOWAIT, MT_DATA);
848 IPSTAT_INC(ips_odropped);
852 * Make sure the complete packet header gets copied
853 * from the originating mbuf to the newly created
854 * mbuf. This also ensures that existing firewall
855 * classification(s), VLAN tags and so on get copied
856 * to the resulting fragmented packet(s):
858 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
861 IPSTAT_INC(ips_odropped);
865 * In the first mbuf, leave room for the link header, then
866 * copy the original IP header including options. The payload
867 * goes into an additional mbuf chain returned by m_copym().
869 m->m_data += max_linkhdr;
870 mhip = mtod(m, struct ip *);
872 if (hlen > sizeof (struct ip)) {
873 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
874 mhip->ip_v = IPVERSION;
875 mhip->ip_hl = mhlen >> 2;
878 /* XXX do we need to add ip_off below ? */
879 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
880 if (off + len >= ip_len)
883 mhip->ip_off |= IP_MF;
884 mhip->ip_len = htons((u_short)(len + mhlen));
885 m->m_next = m_copym(m0, off, len, M_NOWAIT);
886 if (m->m_next == NULL) { /* copy failed */
888 error = ENOBUFS; /* ??? */
889 IPSTAT_INC(ips_odropped);
892 m->m_pkthdr.len = mhlen + len;
894 mac_netinet_fragment(m0, m);
896 mhip->ip_off = htons(mhip->ip_off);
898 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
899 mhip->ip_sum = in_cksum(m, mhlen);
900 m->m_pkthdr.csum_flags &= ~CSUM_IP;
903 mnext = &m->m_nextpkt;
905 IPSTAT_ADD(ips_ofragments, nfrags);
908 * Update first fragment by trimming what's been copied out
909 * and updating header.
911 m_adj(m0, hlen + firstlen - ip_len);
912 m0->m_pkthdr.len = hlen + firstlen;
913 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
914 ip->ip_off = htons(ip_off | IP_MF);
916 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
917 ip->ip_sum = in_cksum(m0, hlen);
918 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
927 in_delayed_cksum(struct mbuf *m)
931 uint16_t cklen, csum, offset;
933 ip = mtod(m, struct ip *);
934 offset = ip->ip_hl << 2 ;
936 if (m->m_pkthdr.csum_flags & CSUM_UDP) {
937 /* if udp header is not in the first mbuf copy udplen */
938 if (offset + sizeof(struct udphdr) > m->m_len) {
939 m_copydata(m, offset + offsetof(struct udphdr,
940 uh_ulen), sizeof(cklen), (caddr_t)&cklen);
941 cklen = ntohs(cklen);
943 uh = (struct udphdr *)mtodo(m, offset);
944 cklen = ntohs(uh->uh_ulen);
946 csum = in_cksum_skip(m, cklen + offset, offset);
950 cklen = ntohs(ip->ip_len);
951 csum = in_cksum_skip(m, cklen, offset);
953 offset += m->m_pkthdr.csum_data; /* checksum offset */
955 if (offset + sizeof(csum) > m->m_len)
956 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
958 *(u_short *)mtodo(m, offset) = csum;
962 * IP socket option processing.
965 ip_ctloutput(struct socket *so, struct sockopt *sopt)
967 struct inpcb *inp = sotoinpcb(so);
975 if (sopt->sopt_level != IPPROTO_IP) {
978 if (sopt->sopt_level == SOL_SOCKET &&
979 sopt->sopt_dir == SOPT_SET) {
980 switch (sopt->sopt_name) {
983 if ((so->so_options & SO_REUSEADDR) != 0)
984 inp->inp_flags2 |= INP_REUSEADDR;
986 inp->inp_flags2 &= ~INP_REUSEADDR;
992 if ((so->so_options & SO_REUSEPORT) != 0)
993 inp->inp_flags2 |= INP_REUSEPORT;
995 inp->inp_flags2 &= ~INP_REUSEPORT;
999 case SO_REUSEPORT_LB:
1001 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1002 inp->inp_flags2 |= INP_REUSEPORT_LB;
1004 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1010 inp->inp_inc.inc_fibnum = so->so_fibnum;
1014 case SO_MAX_PACING_RATE:
1017 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1031 switch (sopt->sopt_dir) {
1033 switch (sopt->sopt_name) {
1040 if (sopt->sopt_valsize > MLEN) {
1044 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1049 m->m_len = sopt->sopt_valsize;
1050 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1057 error = ip_pcbopts(inp, sopt->sopt_name, m);
1063 if (sopt->sopt_td != NULL) {
1064 error = priv_check(sopt->sopt_td,
1065 PRIV_NETINET_BINDANY);
1072 case IP_RSS_LISTEN_BUCKET:
1078 case IP_RECVRETOPTS:
1079 case IP_ORIGDSTADDR:
1080 case IP_RECVDSTADDR:
1088 case IP_RECVRSSBUCKETID:
1090 error = sooptcopyin(sopt, &optval, sizeof optval,
1095 switch (sopt->sopt_name) {
1097 inp->inp_ip_tos = optval;
1101 inp->inp_ip_ttl = optval;
1105 if (optval >= 0 && optval <= MAXTTL)
1106 inp->inp_ip_minttl = optval;
1111 #define OPTSET(bit) do { \
1114 inp->inp_flags |= bit; \
1116 inp->inp_flags &= ~bit; \
1120 #define OPTSET2(bit, val) do { \
1123 inp->inp_flags2 |= bit; \
1125 inp->inp_flags2 &= ~bit; \
1130 OPTSET(INP_RECVOPTS);
1133 case IP_RECVRETOPTS:
1134 OPTSET(INP_RECVRETOPTS);
1137 case IP_RECVDSTADDR:
1138 OPTSET(INP_RECVDSTADDR);
1141 case IP_ORIGDSTADDR:
1142 OPTSET2(INP_ORIGDSTADDR, optval);
1146 OPTSET(INP_RECVTTL);
1154 OPTSET(INP_ONESBCAST);
1157 OPTSET(INP_DONTFRAG);
1160 OPTSET(INP_BINDANY);
1163 OPTSET(INP_RECVTOS);
1166 OPTSET2(INP_BINDMULTI, optval);
1169 OPTSET2(INP_RECVFLOWID, optval);
1172 case IP_RSS_LISTEN_BUCKET:
1173 if ((optval >= 0) &&
1174 (optval < rss_getnumbuckets())) {
1175 inp->inp_rss_listen_bucket = optval;
1176 OPTSET2(INP_RSS_BUCKET_SET, 1);
1181 case IP_RECVRSSBUCKETID:
1182 OPTSET2(INP_RECVRSSBUCKETID, optval);
1191 * Multicast socket options are processed by the in_mcast
1194 case IP_MULTICAST_IF:
1195 case IP_MULTICAST_VIF:
1196 case IP_MULTICAST_TTL:
1197 case IP_MULTICAST_LOOP:
1198 case IP_ADD_MEMBERSHIP:
1199 case IP_DROP_MEMBERSHIP:
1200 case IP_ADD_SOURCE_MEMBERSHIP:
1201 case IP_DROP_SOURCE_MEMBERSHIP:
1202 case IP_BLOCK_SOURCE:
1203 case IP_UNBLOCK_SOURCE:
1205 case MCAST_JOIN_GROUP:
1206 case MCAST_LEAVE_GROUP:
1207 case MCAST_JOIN_SOURCE_GROUP:
1208 case MCAST_LEAVE_SOURCE_GROUP:
1209 case MCAST_BLOCK_SOURCE:
1210 case MCAST_UNBLOCK_SOURCE:
1211 error = inp_setmoptions(inp, sopt);
1215 error = sooptcopyin(sopt, &optval, sizeof optval,
1222 case IP_PORTRANGE_DEFAULT:
1223 inp->inp_flags &= ~(INP_LOWPORT);
1224 inp->inp_flags &= ~(INP_HIGHPORT);
1227 case IP_PORTRANGE_HIGH:
1228 inp->inp_flags &= ~(INP_LOWPORT);
1229 inp->inp_flags |= INP_HIGHPORT;
1232 case IP_PORTRANGE_LOW:
1233 inp->inp_flags &= ~(INP_HIGHPORT);
1234 inp->inp_flags |= INP_LOWPORT;
1244 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1245 case IP_IPSEC_POLICY:
1246 if (IPSEC_ENABLED(ipv4)) {
1247 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1254 error = ENOPROTOOPT;
1260 switch (sopt->sopt_name) {
1264 if (inp->inp_options) {
1265 struct mbuf *options;
1267 options = m_copym(inp->inp_options, 0,
1268 M_COPYALL, M_NOWAIT);
1270 if (options != NULL) {
1271 error = sooptcopyout(sopt,
1272 mtod(options, char *),
1279 sopt->sopt_valsize = 0;
1287 case IP_RECVRETOPTS:
1288 case IP_ORIGDSTADDR:
1289 case IP_RECVDSTADDR:
1302 case IP_RSSBUCKETID:
1303 case IP_RECVRSSBUCKETID:
1305 switch (sopt->sopt_name) {
1308 optval = inp->inp_ip_tos;
1312 optval = inp->inp_ip_ttl;
1316 optval = inp->inp_ip_minttl;
1319 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1320 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1323 optval = OPTBIT(INP_RECVOPTS);
1326 case IP_RECVRETOPTS:
1327 optval = OPTBIT(INP_RECVRETOPTS);
1330 case IP_RECVDSTADDR:
1331 optval = OPTBIT(INP_RECVDSTADDR);
1334 case IP_ORIGDSTADDR:
1335 optval = OPTBIT2(INP_ORIGDSTADDR);
1339 optval = OPTBIT(INP_RECVTTL);
1343 optval = OPTBIT(INP_RECVIF);
1347 if (inp->inp_flags & INP_HIGHPORT)
1348 optval = IP_PORTRANGE_HIGH;
1349 else if (inp->inp_flags & INP_LOWPORT)
1350 optval = IP_PORTRANGE_LOW;
1356 optval = OPTBIT(INP_ONESBCAST);
1359 optval = OPTBIT(INP_DONTFRAG);
1362 optval = OPTBIT(INP_BINDANY);
1365 optval = OPTBIT(INP_RECVTOS);
1368 optval = inp->inp_flowid;
1371 optval = inp->inp_flowtype;
1374 optval = OPTBIT2(INP_RECVFLOWID);
1377 case IP_RSSBUCKETID:
1378 retval = rss_hash2bucket(inp->inp_flowid,
1382 optval = rss_bucket;
1386 case IP_RECVRSSBUCKETID:
1387 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1391 optval = OPTBIT2(INP_BINDMULTI);
1394 error = sooptcopyout(sopt, &optval, sizeof optval);
1398 * Multicast socket options are processed by the in_mcast
1401 case IP_MULTICAST_IF:
1402 case IP_MULTICAST_VIF:
1403 case IP_MULTICAST_TTL:
1404 case IP_MULTICAST_LOOP:
1406 error = inp_getmoptions(inp, sopt);
1409 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1410 case IP_IPSEC_POLICY:
1411 if (IPSEC_ENABLED(ipv4)) {
1412 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1419 error = ENOPROTOOPT;
1428 * Routine called from ip_output() to loop back a copy of an IP multicast
1429 * packet to the input queue of a specified interface. Note that this
1430 * calls the output routine of the loopback "driver", but with an interface
1431 * pointer that might NOT be a loopback interface -- evil, but easier than
1432 * replicating that code here.
1435 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1441 * Make a deep copy of the packet because we're going to
1442 * modify the pack in order to generate checksums.
1444 copym = m_dup(m, M_NOWAIT);
1445 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1446 copym = m_pullup(copym, hlen);
1447 if (copym != NULL) {
1448 /* If needed, compute the checksum and mark it as valid. */
1449 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1450 in_delayed_cksum(copym);
1451 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1452 copym->m_pkthdr.csum_flags |=
1453 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1454 copym->m_pkthdr.csum_data = 0xffff;
1457 * We don't bother to fragment if the IP length is greater
1458 * than the interface's MTU. Can this possibly matter?
1460 ip = mtod(copym, struct ip *);
1462 ip->ip_sum = in_cksum(copym, hlen);
1463 if_simloop(ifp, copym, AF_INET, 0);