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>
67 #include <net/radix_mpath.h>
69 #include <net/rss_config.h>
72 #include <netinet/in.h>
73 #include <netinet/in_kdtrace.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/ip.h>
76 #include <netinet/in_pcb.h>
77 #include <netinet/in_rss.h>
78 #include <netinet/in_var.h>
79 #include <netinet/ip_var.h>
80 #include <netinet/ip_options.h>
82 #include <netinet/sctp.h>
83 #include <netinet/sctp_crc32.h>
86 #include <netipsec/ipsec_support.h>
88 #include <machine/in_cksum.h>
90 #include <security/mac/mac_framework.h>
92 #ifdef MBUF_STRESS_TEST
93 static int mbuf_frag_size = 0;
94 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
95 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
98 static void ip_mloopback(struct ifnet *, const struct mbuf *, int);
101 extern int in_mcast_loop;
102 extern struct protosw inetsw[];
105 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
106 struct sockaddr_in *dst, int *fibnum, int *error)
108 struct m_tag *fwd_tag = NULL;
114 ip = mtod(m, struct ip *);
116 /* Run through list of hooks for output packets. */
117 odst.s_addr = ip->ip_dst.s_addr;
118 *error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, inp);
120 if ((*error) != 0 || m == NULL)
121 return 1; /* Finished */
123 ip = mtod(m, struct ip *);
125 /* See if destination IP address was changed by packet filter. */
126 if (odst.s_addr != ip->ip_dst.s_addr) {
127 m->m_flags |= M_SKIP_FIREWALL;
128 /* If destination is now ourself drop to ip_input(). */
129 if (in_localip(ip->ip_dst)) {
130 m->m_flags |= M_FASTFWD_OURS;
131 if (m->m_pkthdr.rcvif == NULL)
132 m->m_pkthdr.rcvif = V_loif;
133 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
134 m->m_pkthdr.csum_flags |=
135 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
136 m->m_pkthdr.csum_data = 0xffff;
138 m->m_pkthdr.csum_flags |=
139 CSUM_IP_CHECKED | CSUM_IP_VALID;
141 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
142 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
144 *error = netisr_queue(NETISR_IP, m);
145 return 1; /* Finished */
148 bzero(dst, sizeof(*dst));
149 dst->sin_family = AF_INET;
150 dst->sin_len = sizeof(*dst);
151 dst->sin_addr = ip->ip_dst;
153 return -1; /* Reloop */
155 /* See if fib was changed by packet filter. */
156 if ((*fibnum) != M_GETFIB(m)) {
157 m->m_flags |= M_SKIP_FIREWALL;
158 *fibnum = M_GETFIB(m);
159 return -1; /* Reloop for FIB change */
162 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
163 if (m->m_flags & M_FASTFWD_OURS) {
164 if (m->m_pkthdr.rcvif == NULL)
165 m->m_pkthdr.rcvif = V_loif;
166 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
167 m->m_pkthdr.csum_flags |=
168 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
169 m->m_pkthdr.csum_data = 0xffff;
172 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
173 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
175 m->m_pkthdr.csum_flags |=
176 CSUM_IP_CHECKED | CSUM_IP_VALID;
178 *error = netisr_queue(NETISR_IP, m);
179 return 1; /* Finished */
181 /* Or forward to some other address? */
182 if ((m->m_flags & M_IP_NEXTHOP) &&
183 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
184 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
185 m->m_flags |= M_SKIP_FIREWALL;
186 m->m_flags &= ~M_IP_NEXTHOP;
187 m_tag_delete(m, fwd_tag);
189 return -1; /* Reloop for CHANGE of dst */
196 * IP output. The packet in mbuf chain m contains a skeletal IP
197 * header (with len, off, ttl, proto, tos, src, dst).
198 * The mbuf chain containing the packet will be freed.
199 * The mbuf opt, if present, will not be freed.
200 * If route ro is present and has ro_rt initialized, route lookup would be
201 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
202 * then result of route lookup is stored in ro->ro_rt.
204 * In the IP forwarding case, the packet will arrive with options already
205 * inserted, so must have a NULL opt pointer.
208 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
209 struct ip_moptions *imo, struct inpcb *inp)
211 struct rm_priotracker in_ifa_tracker;
213 struct ifnet *ifp = NULL; /* keep compiler happy */
215 int hlen = sizeof (struct ip);
218 struct sockaddr_in *dst;
219 const struct sockaddr_in *gw;
220 struct in_ifaddr *ia;
222 uint16_t ip_len, ip_off;
223 struct route iproute;
224 struct rtentry *rte; /* cache for ro->ro_rt */
227 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
228 int no_route_but_check_spd = 0;
233 INP_LOCK_ASSERT(inp);
234 M_SETFIB(m, inp->inp_inc.inc_fibnum);
235 if ((flags & IP_NODEFAULTFLOWID) == 0) {
236 m->m_pkthdr.flowid = inp->inp_flowid;
237 M_HASHTYPE_SET(m, inp->inp_flowtype);
243 bzero(ro, sizeof (*ro));
248 m = ip_insertoptions(m, opt, &len);
250 hlen = len; /* ip->ip_hl is updated above */
252 ip = mtod(m, struct ip *);
253 ip_len = ntohs(ip->ip_len);
254 ip_off = ntohs(ip->ip_off);
256 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
257 ip->ip_v = IPVERSION;
258 ip->ip_hl = hlen >> 2;
260 IPSTAT_INC(ips_localout);
262 /* Header already set, fetch hlen from there */
263 hlen = ip->ip_hl << 2;
269 * dst can be rewritten but always points to &ro->ro_dst.
270 * gw is readonly but can point either to dst OR rt_gateway,
271 * therefore we need restore gw if we're redoing lookup.
273 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
274 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
277 bzero(dst, sizeof(*dst));
278 dst->sin_family = AF_INET;
279 dst->sin_len = sizeof(*dst);
280 dst->sin_addr = ip->ip_dst;
284 * Validate route against routing table additions;
285 * a better/more specific route might have been added.
288 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
290 * If there is a cached route,
291 * check that it is to the same destination
292 * and is still up. If not, free it and try again.
293 * The address family should also be checked in case of sharing the
295 * Also check whether routing cache needs invalidation.
298 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
299 rte->rt_ifp == NULL ||
300 !RT_LINK_IS_UP(rte->rt_ifp) ||
301 dst->sin_family != AF_INET ||
302 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
304 rte = ro->ro_rt = (struct rtentry *)NULL;
306 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
307 ro->ro_lle = (struct llentry *)NULL;
312 * If routing to interface only, short circuit routing lookup.
313 * The use of an all-ones broadcast address implies this; an
314 * interface is specified by the broadcast address of an interface,
315 * or the destination address of a ptp interface.
317 if (flags & IP_SENDONES) {
318 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
319 M_GETFIB(m)))) == NULL &&
320 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
321 M_GETFIB(m)))) == NULL) {
322 IPSTAT_INC(ips_noroute);
327 ip->ip_dst.s_addr = INADDR_BROADCAST;
328 dst->sin_addr = ip->ip_dst;
332 } else if (flags & IP_ROUTETOIF) {
333 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
334 M_GETFIB(m)))) == NULL &&
335 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
336 M_GETFIB(m)))) == NULL) {
337 IPSTAT_INC(ips_noroute);
344 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
345 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
346 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
347 imo != NULL && imo->imo_multicast_ifp != NULL) {
349 * Bypass the normal routing lookup for multicast
350 * packets if the interface is specified.
352 ifp = imo->imo_multicast_ifp;
353 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
356 isbroadcast = 0; /* fool gcc */
359 * We want to do any cloning requested by the link layer,
360 * as this is probably required in all cases for correct
361 * operation (as it is for ARP).
365 rtalloc_mpath_fib(ro,
366 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
369 in_rtalloc_ign(ro, 0, fibnum);
374 (rte->rt_flags & RTF_UP) == 0 ||
375 rte->rt_ifp == NULL ||
376 !RT_LINK_IS_UP(rte->rt_ifp)) {
377 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
379 * There is no route for this packet, but it is
380 * possible that a matching SPD entry exists.
382 no_route_but_check_spd = 1;
383 mtu = 0; /* Silence GCC warning. */
386 IPSTAT_INC(ips_noroute);
387 error = EHOSTUNREACH;
390 ia = ifatoia(rte->rt_ifa);
392 counter_u64_add(rte->rt_pksent, 1);
393 rt_update_ro_flags(ro);
394 if (rte->rt_flags & RTF_GATEWAY)
395 gw = (struct sockaddr_in *)rte->rt_gateway;
396 if (rte->rt_flags & RTF_HOST)
397 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
398 else if (ifp->if_flags & IFF_BROADCAST)
399 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
405 * Calculate MTU. If we have a route that is up, use that,
406 * otherwise use the interface's MTU.
408 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
412 /* Catch a possible divide by zero later. */
413 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
414 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
416 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
417 m->m_flags |= M_MCAST;
419 * IP destination address is multicast. Make sure "gw"
420 * still points to the address in "ro". (It may have been
421 * changed to point to a gateway address, above.)
425 * See if the caller provided any multicast options
428 ip->ip_ttl = imo->imo_multicast_ttl;
429 if (imo->imo_multicast_vif != -1)
432 ip_mcast_src(imo->imo_multicast_vif) :
435 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
437 * Confirm that the outgoing interface supports multicast.
439 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
440 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
441 IPSTAT_INC(ips_noroute);
447 * If source address not specified yet, use address
448 * of outgoing interface.
450 if (ip->ip_src.s_addr == INADDR_ANY) {
451 /* Interface may have no addresses. */
453 ip->ip_src = IA_SIN(ia)->sin_addr;
456 if ((imo == NULL && in_mcast_loop) ||
457 (imo && imo->imo_multicast_loop)) {
459 * Loop back multicast datagram if not expressly
460 * forbidden to do so, even if we are not a member
461 * of the group; ip_input() will filter it later,
462 * thus deferring a hash lookup and mutex acquisition
463 * at the expense of a cheap copy using m_copym().
465 ip_mloopback(ifp, m, hlen);
468 * If we are acting as a multicast router, perform
469 * multicast forwarding as if the packet had just
470 * arrived on the interface to which we are about
471 * to send. The multicast forwarding function
472 * recursively calls this function, using the
473 * IP_FORWARDING flag to prevent infinite recursion.
475 * Multicasts that are looped back by ip_mloopback(),
476 * above, will be forwarded by the ip_input() routine,
479 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
481 * If rsvp daemon is not running, do not
482 * set ip_moptions. This ensures that the packet
483 * is multicast and not just sent down one link
484 * as prescribed by rsvpd.
489 ip_mforward(ip, ifp, m, imo) != 0) {
497 * Multicasts with a time-to-live of zero may be looped-
498 * back, above, but must not be transmitted on a network.
499 * Also, multicasts addressed to the loopback interface
500 * are not sent -- the above call to ip_mloopback() will
501 * loop back a copy. ip_input() will drop the copy if
502 * this host does not belong to the destination group on
503 * the loopback interface.
505 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
514 * If the source address is not specified yet, use the address
515 * of the outoing interface.
517 if (ip->ip_src.s_addr == INADDR_ANY) {
518 /* Interface may have no addresses. */
520 ip->ip_src = IA_SIN(ia)->sin_addr;
525 * Look for broadcast address and
526 * verify user is allowed to send
530 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
531 error = EADDRNOTAVAIL;
534 if ((flags & IP_ALLOWBROADCAST) == 0) {
538 /* don't allow broadcast messages to be fragmented */
543 m->m_flags |= M_BCAST;
545 m->m_flags &= ~M_BCAST;
549 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
550 if (IPSEC_ENABLED(ipv4)) {
551 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
552 if (error == EINPROGRESS)
558 * Check if there was a route for this packet; return error if not.
560 if (no_route_but_check_spd) {
561 IPSTAT_INC(ips_noroute);
562 error = EHOSTUNREACH;
565 /* Update variables that are affected by ipsec4_output(). */
566 ip = mtod(m, struct ip *);
567 hlen = ip->ip_hl << 2;
570 /* Jump over all PFIL processing if hooks are not active. */
571 if (PFIL_HOOKED(&V_inet_pfil_hook)) {
572 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
573 case 1: /* Finished */
576 case 0: /* Continue normally */
577 ip = mtod(m, struct ip *);
580 case -1: /* Need to try again */
581 /* Reset everything for a new round */
584 ifa_free(&ia->ia_ifa);
585 ro->ro_prepend = NULL;
588 ip = mtod(m, struct ip *);
594 /* 127/8 must not appear on wire - RFC1122. */
595 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
596 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
597 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
598 IPSTAT_INC(ips_badaddr);
599 error = EADDRNOTAVAIL;
604 m->m_pkthdr.csum_flags |= CSUM_IP;
605 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
607 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
610 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
611 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
612 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
617 * If small enough for interface, or the interface will take
618 * care of the fragmentation for us, we can just send directly.
621 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
623 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
624 ip->ip_sum = in_cksum(m, hlen);
625 m->m_pkthdr.csum_flags &= ~CSUM_IP;
629 * Record statistics for this interface address.
630 * With CSUM_TSO the byte/packet count will be slightly
631 * incorrect because we count the IP+TCP headers only
632 * once instead of for every generated packet.
634 if (!(flags & IP_FORWARDING) && ia) {
635 if (m->m_pkthdr.csum_flags & CSUM_TSO)
636 counter_u64_add(ia->ia_ifa.ifa_opackets,
637 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
639 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
641 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
643 #ifdef MBUF_STRESS_TEST
644 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
645 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
648 * Reset layer specific mbuf flags
649 * to avoid confusing lower layers.
652 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
655 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
656 in_pcboutput_txrtlmt(inp, ifp, m);
657 /* stamp send tag on mbuf */
658 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
660 m->m_pkthdr.snd_tag = NULL;
663 error = (*ifp->if_output)(ifp, m,
664 (const struct sockaddr *)gw, ro);
666 /* check for route change */
668 in_pcboutput_eagain(inp);
673 /* Balk when DF bit is set or the interface didn't support TSO. */
674 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
676 IPSTAT_INC(ips_cantfrag);
681 * Too large for interface; fragment if possible. If successful,
682 * on return, m will point to a list of packets to be sent.
684 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
691 /* Record statistics for this interface address. */
693 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
694 counter_u64_add(ia->ia_ifa.ifa_obytes,
698 * Reset layer specific mbuf flags
699 * to avoid confusing upper layers.
703 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
704 mtod(m, struct ip *), NULL);
707 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
708 in_pcboutput_txrtlmt(inp, ifp, m);
709 /* stamp send tag on mbuf */
710 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
712 m->m_pkthdr.snd_tag = NULL;
715 error = (*ifp->if_output)(ifp, m,
716 (const struct sockaddr *)gw, ro);
718 /* check for route change */
720 in_pcboutput_eagain(inp);
727 IPSTAT_INC(ips_fragmented);
732 else if (rte == NULL)
734 * If the caller supplied a route but somehow the reference
735 * to it has been released need to prevent the caller
736 * calling RTFREE on it again.
740 ifa_free(&ia->ia_ifa);
748 * Create a chain of fragments which fit the given mtu. m_frag points to the
749 * mbuf to be fragmented; on return it points to the chain with the fragments.
750 * Return 0 if no error. If error, m_frag may contain a partially built
751 * chain of fragments that should be freed by the caller.
753 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
756 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
757 u_long if_hwassist_flags)
760 int hlen = ip->ip_hl << 2;
761 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
763 struct mbuf *m0 = *m_frag; /* the original packet */
767 uint16_t ip_len, ip_off;
769 ip_len = ntohs(ip->ip_len);
770 ip_off = ntohs(ip->ip_off);
772 if (ip_off & IP_DF) { /* Fragmentation not allowed */
773 IPSTAT_INC(ips_cantfrag);
778 * Must be able to put at least 8 bytes per fragment.
784 * If the interface will not calculate checksums on
785 * fragmented packets, then do it here.
787 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
788 in_delayed_cksum(m0);
789 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
792 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
793 sctp_delayed_cksum(m0, hlen);
794 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
797 if (len > PAGE_SIZE) {
799 * Fragment large datagrams such that each segment
800 * contains a multiple of PAGE_SIZE amount of data,
801 * plus headers. This enables a receiver to perform
802 * page-flipping zero-copy optimizations.
804 * XXX When does this help given that sender and receiver
805 * could have different page sizes, and also mtu could
806 * be less than the receiver's page size ?
810 off = MIN(mtu, m0->m_pkthdr.len);
813 * firstlen (off - hlen) must be aligned on an
817 goto smart_frag_failure;
818 off = ((off - hlen) & ~7) + hlen;
819 newlen = (~PAGE_MASK) & mtu;
820 if ((newlen + sizeof (struct ip)) > mtu) {
821 /* we failed, go back the default */
832 firstlen = off - hlen;
833 mnext = &m0->m_nextpkt; /* pointer to next packet */
836 * Loop through length of segment after first fragment,
837 * make new header and copy data of each part and link onto chain.
838 * Here, m0 is the original packet, m is the fragment being created.
839 * The fragments are linked off the m_nextpkt of the original
840 * packet, which after processing serves as the first fragment.
842 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
843 struct ip *mhip; /* ip header on the fragment */
845 int mhlen = sizeof (struct ip);
847 m = m_gethdr(M_NOWAIT, MT_DATA);
850 IPSTAT_INC(ips_odropped);
854 * Make sure the complete packet header gets copied
855 * from the originating mbuf to the newly created
856 * mbuf. This also ensures that existing firewall
857 * classification(s), VLAN tags and so on get copied
858 * to the resulting fragmented packet(s):
860 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
863 IPSTAT_INC(ips_odropped);
867 * In the first mbuf, leave room for the link header, then
868 * copy the original IP header including options. The payload
869 * goes into an additional mbuf chain returned by m_copym().
871 m->m_data += max_linkhdr;
872 mhip = mtod(m, struct ip *);
874 if (hlen > sizeof (struct ip)) {
875 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
876 mhip->ip_v = IPVERSION;
877 mhip->ip_hl = mhlen >> 2;
880 /* XXX do we need to add ip_off below ? */
881 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
882 if (off + len >= ip_len)
885 mhip->ip_off |= IP_MF;
886 mhip->ip_len = htons((u_short)(len + mhlen));
887 m->m_next = m_copym(m0, off, len, M_NOWAIT);
888 if (m->m_next == NULL) { /* copy failed */
890 error = ENOBUFS; /* ??? */
891 IPSTAT_INC(ips_odropped);
894 m->m_pkthdr.len = mhlen + len;
896 mac_netinet_fragment(m0, m);
898 mhip->ip_off = htons(mhip->ip_off);
900 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
901 mhip->ip_sum = in_cksum(m, mhlen);
902 m->m_pkthdr.csum_flags &= ~CSUM_IP;
905 mnext = &m->m_nextpkt;
907 IPSTAT_ADD(ips_ofragments, nfrags);
910 * Update first fragment by trimming what's been copied out
911 * and updating header.
913 m_adj(m0, hlen + firstlen - ip_len);
914 m0->m_pkthdr.len = hlen + firstlen;
915 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
916 ip->ip_off = htons(ip_off | IP_MF);
918 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
919 ip->ip_sum = in_cksum(m0, hlen);
920 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
929 in_delayed_cksum(struct mbuf *m)
932 uint16_t csum, offset, ip_len;
934 ip = mtod(m, struct ip *);
935 offset = ip->ip_hl << 2 ;
936 ip_len = ntohs(ip->ip_len);
937 csum = in_cksum_skip(m, ip_len, offset);
938 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
940 offset += m->m_pkthdr.csum_data; /* checksum offset */
942 /* find the mbuf in the chain where the checksum starts*/
943 while ((m != NULL) && (offset >= m->m_len)) {
947 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
948 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
949 *(u_short *)(m->m_data + offset) = csum;
953 * IP socket option processing.
956 ip_ctloutput(struct socket *so, struct sockopt *sopt)
958 struct inpcb *inp = sotoinpcb(so);
966 if (sopt->sopt_level != IPPROTO_IP) {
969 if (sopt->sopt_level == SOL_SOCKET &&
970 sopt->sopt_dir == SOPT_SET) {
971 switch (sopt->sopt_name) {
974 if ((so->so_options & SO_REUSEADDR) != 0)
975 inp->inp_flags2 |= INP_REUSEADDR;
977 inp->inp_flags2 &= ~INP_REUSEADDR;
983 if ((so->so_options & SO_REUSEPORT) != 0)
984 inp->inp_flags2 |= INP_REUSEPORT;
986 inp->inp_flags2 &= ~INP_REUSEPORT;
992 inp->inp_inc.inc_fibnum = so->so_fibnum;
996 case SO_MAX_PACING_RATE:
999 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1013 switch (sopt->sopt_dir) {
1015 switch (sopt->sopt_name) {
1022 if (sopt->sopt_valsize > MLEN) {
1026 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1031 m->m_len = sopt->sopt_valsize;
1032 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1039 error = ip_pcbopts(inp, sopt->sopt_name, m);
1045 if (sopt->sopt_td != NULL) {
1046 error = priv_check(sopt->sopt_td,
1047 PRIV_NETINET_BINDANY);
1054 case IP_RSS_LISTEN_BUCKET:
1060 case IP_RECVRETOPTS:
1061 case IP_ORIGDSTADDR:
1062 case IP_RECVDSTADDR:
1070 case IP_RECVRSSBUCKETID:
1072 error = sooptcopyin(sopt, &optval, sizeof optval,
1077 switch (sopt->sopt_name) {
1079 inp->inp_ip_tos = optval;
1083 inp->inp_ip_ttl = optval;
1087 if (optval >= 0 && optval <= MAXTTL)
1088 inp->inp_ip_minttl = optval;
1093 #define OPTSET(bit) do { \
1096 inp->inp_flags |= bit; \
1098 inp->inp_flags &= ~bit; \
1102 #define OPTSET2(bit, val) do { \
1105 inp->inp_flags2 |= bit; \
1107 inp->inp_flags2 &= ~bit; \
1112 OPTSET(INP_RECVOPTS);
1115 case IP_RECVRETOPTS:
1116 OPTSET(INP_RECVRETOPTS);
1119 case IP_RECVDSTADDR:
1120 OPTSET(INP_RECVDSTADDR);
1123 case IP_ORIGDSTADDR:
1124 OPTSET2(INP_ORIGDSTADDR, optval);
1128 OPTSET(INP_RECVTTL);
1136 OPTSET(INP_ONESBCAST);
1139 OPTSET(INP_DONTFRAG);
1142 OPTSET(INP_BINDANY);
1145 OPTSET(INP_RECVTOS);
1148 OPTSET2(INP_BINDMULTI, optval);
1151 OPTSET2(INP_RECVFLOWID, optval);
1154 case IP_RSS_LISTEN_BUCKET:
1155 if ((optval >= 0) &&
1156 (optval < rss_getnumbuckets())) {
1157 inp->inp_rss_listen_bucket = optval;
1158 OPTSET2(INP_RSS_BUCKET_SET, 1);
1163 case IP_RECVRSSBUCKETID:
1164 OPTSET2(INP_RECVRSSBUCKETID, optval);
1173 * Multicast socket options are processed by the in_mcast
1176 case IP_MULTICAST_IF:
1177 case IP_MULTICAST_VIF:
1178 case IP_MULTICAST_TTL:
1179 case IP_MULTICAST_LOOP:
1180 case IP_ADD_MEMBERSHIP:
1181 case IP_DROP_MEMBERSHIP:
1182 case IP_ADD_SOURCE_MEMBERSHIP:
1183 case IP_DROP_SOURCE_MEMBERSHIP:
1184 case IP_BLOCK_SOURCE:
1185 case IP_UNBLOCK_SOURCE:
1187 case MCAST_JOIN_GROUP:
1188 case MCAST_LEAVE_GROUP:
1189 case MCAST_JOIN_SOURCE_GROUP:
1190 case MCAST_LEAVE_SOURCE_GROUP:
1191 case MCAST_BLOCK_SOURCE:
1192 case MCAST_UNBLOCK_SOURCE:
1193 error = inp_setmoptions(inp, sopt);
1197 error = sooptcopyin(sopt, &optval, sizeof optval,
1204 case IP_PORTRANGE_DEFAULT:
1205 inp->inp_flags &= ~(INP_LOWPORT);
1206 inp->inp_flags &= ~(INP_HIGHPORT);
1209 case IP_PORTRANGE_HIGH:
1210 inp->inp_flags &= ~(INP_LOWPORT);
1211 inp->inp_flags |= INP_HIGHPORT;
1214 case IP_PORTRANGE_LOW:
1215 inp->inp_flags &= ~(INP_HIGHPORT);
1216 inp->inp_flags |= INP_LOWPORT;
1226 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1227 case IP_IPSEC_POLICY:
1228 if (IPSEC_ENABLED(ipv4)) {
1229 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1236 error = ENOPROTOOPT;
1242 switch (sopt->sopt_name) {
1245 if (inp->inp_options)
1246 error = sooptcopyout(sopt,
1247 mtod(inp->inp_options,
1249 inp->inp_options->m_len);
1251 sopt->sopt_valsize = 0;
1258 case IP_RECVRETOPTS:
1259 case IP_ORIGDSTADDR:
1260 case IP_RECVDSTADDR:
1273 case IP_RSSBUCKETID:
1274 case IP_RECVRSSBUCKETID:
1276 switch (sopt->sopt_name) {
1279 optval = inp->inp_ip_tos;
1283 optval = inp->inp_ip_ttl;
1287 optval = inp->inp_ip_minttl;
1290 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1291 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1294 optval = OPTBIT(INP_RECVOPTS);
1297 case IP_RECVRETOPTS:
1298 optval = OPTBIT(INP_RECVRETOPTS);
1301 case IP_RECVDSTADDR:
1302 optval = OPTBIT(INP_RECVDSTADDR);
1305 case IP_ORIGDSTADDR:
1306 optval = OPTBIT2(INP_ORIGDSTADDR);
1310 optval = OPTBIT(INP_RECVTTL);
1314 optval = OPTBIT(INP_RECVIF);
1318 if (inp->inp_flags & INP_HIGHPORT)
1319 optval = IP_PORTRANGE_HIGH;
1320 else if (inp->inp_flags & INP_LOWPORT)
1321 optval = IP_PORTRANGE_LOW;
1327 optval = OPTBIT(INP_ONESBCAST);
1330 optval = OPTBIT(INP_DONTFRAG);
1333 optval = OPTBIT(INP_BINDANY);
1336 optval = OPTBIT(INP_RECVTOS);
1339 optval = inp->inp_flowid;
1342 optval = inp->inp_flowtype;
1345 optval = OPTBIT2(INP_RECVFLOWID);
1348 case IP_RSSBUCKETID:
1349 retval = rss_hash2bucket(inp->inp_flowid,
1353 optval = rss_bucket;
1357 case IP_RECVRSSBUCKETID:
1358 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1362 optval = OPTBIT2(INP_BINDMULTI);
1365 error = sooptcopyout(sopt, &optval, sizeof optval);
1369 * Multicast socket options are processed by the in_mcast
1372 case IP_MULTICAST_IF:
1373 case IP_MULTICAST_VIF:
1374 case IP_MULTICAST_TTL:
1375 case IP_MULTICAST_LOOP:
1377 error = inp_getmoptions(inp, sopt);
1380 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1381 case IP_IPSEC_POLICY:
1382 if (IPSEC_ENABLED(ipv4)) {
1383 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1390 error = ENOPROTOOPT;
1399 * Routine called from ip_output() to loop back a copy of an IP multicast
1400 * packet to the input queue of a specified interface. Note that this
1401 * calls the output routine of the loopback "driver", but with an interface
1402 * pointer that might NOT be a loopback interface -- evil, but easier than
1403 * replicating that code here.
1406 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1412 * Make a deep copy of the packet because we're going to
1413 * modify the pack in order to generate checksums.
1415 copym = m_dup(m, M_NOWAIT);
1416 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1417 copym = m_pullup(copym, hlen);
1418 if (copym != NULL) {
1419 /* If needed, compute the checksum and mark it as valid. */
1420 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1421 in_delayed_cksum(copym);
1422 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1423 copym->m_pkthdr.csum_flags |=
1424 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1425 copym->m_pkthdr.csum_data = 0xffff;
1428 * We don't bother to fragment if the IP length is greater
1429 * than the interface's MTU. Can this possibly matter?
1431 ip = mtod(copym, struct ip *);
1433 ip->ip_sum = in_cksum(copym, hlen);
1434 if_simloop(ifp, copym, AF_INET, 0);