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));
246 ro->ro_flags |= RT_LLE_CACHE;
249 if (ro->ro_rt == NULL)
250 (void )flowtable_lookup(AF_INET, m, ro);
255 m = ip_insertoptions(m, opt, &len);
257 hlen = len; /* ip->ip_hl is updated above */
259 ip = mtod(m, struct ip *);
260 ip_len = ntohs(ip->ip_len);
261 ip_off = ntohs(ip->ip_off);
263 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
264 ip->ip_v = IPVERSION;
265 ip->ip_hl = hlen >> 2;
267 IPSTAT_INC(ips_localout);
269 /* Header already set, fetch hlen from there */
270 hlen = ip->ip_hl << 2;
276 * dst can be rewritten but always points to &ro->ro_dst.
277 * gw is readonly but can point either to dst OR rt_gateway,
278 * therefore we need restore gw if we're redoing lookup.
280 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
281 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
284 bzero(dst, sizeof(*dst));
285 dst->sin_family = AF_INET;
286 dst->sin_len = sizeof(*dst);
287 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)) {
311 rte = ro->ro_rt = (struct rtentry *)NULL;
313 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
314 ro->ro_lle = (struct llentry *)NULL;
319 * If routing to interface only, short circuit routing lookup.
320 * The use of an all-ones broadcast address implies this; an
321 * interface is specified by the broadcast address of an interface,
322 * or the destination address of a ptp interface.
324 if (flags & IP_SENDONES) {
325 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
326 M_GETFIB(m)))) == NULL &&
327 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
328 M_GETFIB(m)))) == NULL) {
329 IPSTAT_INC(ips_noroute);
334 ip->ip_dst.s_addr = INADDR_BROADCAST;
335 dst->sin_addr = ip->ip_dst;
339 } else if (flags & IP_ROUTETOIF) {
340 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
341 M_GETFIB(m)))) == NULL &&
342 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
343 M_GETFIB(m)))) == NULL) {
344 IPSTAT_INC(ips_noroute);
351 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
352 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
353 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
354 imo != NULL && imo->imo_multicast_ifp != NULL) {
356 * Bypass the normal routing lookup for multicast
357 * packets if the interface is specified.
359 ifp = imo->imo_multicast_ifp;
360 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
363 isbroadcast = 0; /* fool gcc */
366 * We want to do any cloning requested by the link layer,
367 * as this is probably required in all cases for correct
368 * operation (as it is for ARP).
372 rtalloc_mpath_fib(ro,
373 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
376 in_rtalloc_ign(ro, 0, fibnum);
381 (rte->rt_flags & RTF_UP) == 0 ||
382 rte->rt_ifp == NULL ||
383 !RT_LINK_IS_UP(rte->rt_ifp)) {
384 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
386 * There is no route for this packet, but it is
387 * possible that a matching SPD entry exists.
389 no_route_but_check_spd = 1;
390 mtu = 0; /* Silence GCC warning. */
393 IPSTAT_INC(ips_noroute);
394 error = EHOSTUNREACH;
397 ia = ifatoia(rte->rt_ifa);
399 counter_u64_add(rte->rt_pksent, 1);
400 rt_update_ro_flags(ro);
401 if (rte->rt_flags & RTF_GATEWAY)
402 gw = (struct sockaddr_in *)rte->rt_gateway;
403 if (rte->rt_flags & RTF_HOST)
404 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
405 else if (ifp->if_flags & IFF_BROADCAST)
406 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
412 * Calculate MTU. If we have a route that is up, use that,
413 * otherwise use the interface's MTU.
415 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
419 /* Catch a possible divide by zero later. */
420 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
421 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
423 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
424 m->m_flags |= M_MCAST;
426 * IP destination address is multicast. Make sure "gw"
427 * still points to the address in "ro". (It may have been
428 * changed to point to a gateway address, above.)
432 * See if the caller provided any multicast options
435 ip->ip_ttl = imo->imo_multicast_ttl;
436 if (imo->imo_multicast_vif != -1)
439 ip_mcast_src(imo->imo_multicast_vif) :
442 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
444 * Confirm that the outgoing interface supports multicast.
446 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
447 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
448 IPSTAT_INC(ips_noroute);
454 * If source address not specified yet, use address
455 * of outgoing interface.
457 if (ip->ip_src.s_addr == INADDR_ANY) {
458 /* Interface may have no addresses. */
460 ip->ip_src = IA_SIN(ia)->sin_addr;
463 if ((imo == NULL && in_mcast_loop) ||
464 (imo && imo->imo_multicast_loop)) {
466 * Loop back multicast datagram if not expressly
467 * forbidden to do so, even if we are not a member
468 * of the group; ip_input() will filter it later,
469 * thus deferring a hash lookup and mutex acquisition
470 * at the expense of a cheap copy using m_copym().
472 ip_mloopback(ifp, m, hlen);
475 * If we are acting as a multicast router, perform
476 * multicast forwarding as if the packet had just
477 * arrived on the interface to which we are about
478 * to send. The multicast forwarding function
479 * recursively calls this function, using the
480 * IP_FORWARDING flag to prevent infinite recursion.
482 * Multicasts that are looped back by ip_mloopback(),
483 * above, will be forwarded by the ip_input() routine,
486 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
488 * If rsvp daemon is not running, do not
489 * set ip_moptions. This ensures that the packet
490 * is multicast and not just sent down one link
491 * as prescribed by rsvpd.
496 ip_mforward(ip, ifp, m, imo) != 0) {
504 * Multicasts with a time-to-live of zero may be looped-
505 * back, above, but must not be transmitted on a network.
506 * Also, multicasts addressed to the loopback interface
507 * are not sent -- the above call to ip_mloopback() will
508 * loop back a copy. ip_input() will drop the copy if
509 * this host does not belong to the destination group on
510 * the loopback interface.
512 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
521 * If the source address is not specified yet, use the address
522 * of the outoing interface.
524 if (ip->ip_src.s_addr == INADDR_ANY) {
525 /* Interface may have no addresses. */
527 ip->ip_src = IA_SIN(ia)->sin_addr;
532 * Look for broadcast address and
533 * verify user is allowed to send
537 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
538 error = EADDRNOTAVAIL;
541 if ((flags & IP_ALLOWBROADCAST) == 0) {
545 /* don't allow broadcast messages to be fragmented */
550 m->m_flags |= M_BCAST;
552 m->m_flags &= ~M_BCAST;
556 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
557 if (IPSEC_ENABLED(ipv4)) {
558 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
559 if (error == EINPROGRESS)
565 * Check if there was a route for this packet; return error if not.
567 if (no_route_but_check_spd) {
568 IPSTAT_INC(ips_noroute);
569 error = EHOSTUNREACH;
572 /* Update variables that are affected by ipsec4_output(). */
573 ip = mtod(m, struct ip *);
574 hlen = ip->ip_hl << 2;
577 /* Jump over all PFIL processing if hooks are not active. */
578 if (PFIL_HOOKED(&V_inet_pfil_hook)) {
579 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
580 case 1: /* Finished */
583 case 0: /* Continue normally */
584 ip = mtod(m, struct ip *);
587 case -1: /* Need to try again */
588 /* Reset everything for a new round */
591 ifa_free(&ia->ia_ifa);
592 ro->ro_prepend = NULL;
595 ip = mtod(m, struct ip *);
601 /* 127/8 must not appear on wire - RFC1122. */
602 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
603 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
604 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
605 IPSTAT_INC(ips_badaddr);
606 error = EADDRNOTAVAIL;
611 m->m_pkthdr.csum_flags |= CSUM_IP;
612 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
614 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
617 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
618 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
619 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
624 * If small enough for interface, or the interface will take
625 * care of the fragmentation for us, we can just send directly.
628 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
630 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
631 ip->ip_sum = in_cksum(m, hlen);
632 m->m_pkthdr.csum_flags &= ~CSUM_IP;
636 * Record statistics for this interface address.
637 * With CSUM_TSO the byte/packet count will be slightly
638 * incorrect because we count the IP+TCP headers only
639 * once instead of for every generated packet.
641 if (!(flags & IP_FORWARDING) && ia) {
642 if (m->m_pkthdr.csum_flags & CSUM_TSO)
643 counter_u64_add(ia->ia_ifa.ifa_opackets,
644 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
646 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
648 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
650 #ifdef MBUF_STRESS_TEST
651 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
652 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
655 * Reset layer specific mbuf flags
656 * to avoid confusing lower layers.
659 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
662 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
663 in_pcboutput_txrtlmt(inp, ifp, m);
664 /* stamp send tag on mbuf */
665 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
667 m->m_pkthdr.snd_tag = NULL;
670 error = (*ifp->if_output)(ifp, m,
671 (const struct sockaddr *)gw, ro);
673 /* check for route change */
675 in_pcboutput_eagain(inp);
680 /* Balk when DF bit is set or the interface didn't support TSO. */
681 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
683 IPSTAT_INC(ips_cantfrag);
688 * Too large for interface; fragment if possible. If successful,
689 * on return, m will point to a list of packets to be sent.
691 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
698 /* Record statistics for this interface address. */
700 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
701 counter_u64_add(ia->ia_ifa.ifa_obytes,
705 * Reset layer specific mbuf flags
706 * to avoid confusing upper layers.
710 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
711 mtod(m, struct ip *), NULL);
714 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
715 in_pcboutput_txrtlmt(inp, ifp, m);
716 /* stamp send tag on mbuf */
717 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
719 m->m_pkthdr.snd_tag = NULL;
722 error = (*ifp->if_output)(ifp, m,
723 (const struct sockaddr *)gw, ro);
725 /* check for route change */
727 in_pcboutput_eagain(inp);
734 IPSTAT_INC(ips_fragmented);
739 else if (rte == NULL)
741 * If the caller supplied a route but somehow the reference
742 * to it has been released need to prevent the caller
743 * calling RTFREE on it again.
747 ifa_free(&ia->ia_ifa);
755 * Create a chain of fragments which fit the given mtu. m_frag points to the
756 * mbuf to be fragmented; on return it points to the chain with the fragments.
757 * Return 0 if no error. If error, m_frag may contain a partially built
758 * chain of fragments that should be freed by the caller.
760 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
763 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
764 u_long if_hwassist_flags)
767 int hlen = ip->ip_hl << 2;
768 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
770 struct mbuf *m0 = *m_frag; /* the original packet */
774 uint16_t ip_len, ip_off;
776 ip_len = ntohs(ip->ip_len);
777 ip_off = ntohs(ip->ip_off);
779 if (ip_off & IP_DF) { /* Fragmentation not allowed */
780 IPSTAT_INC(ips_cantfrag);
785 * Must be able to put at least 8 bytes per fragment.
791 * If the interface will not calculate checksums on
792 * fragmented packets, then do it here.
794 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
795 in_delayed_cksum(m0);
796 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
799 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
800 sctp_delayed_cksum(m0, hlen);
801 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
804 if (len > PAGE_SIZE) {
806 * Fragment large datagrams such that each segment
807 * contains a multiple of PAGE_SIZE amount of data,
808 * plus headers. This enables a receiver to perform
809 * page-flipping zero-copy optimizations.
811 * XXX When does this help given that sender and receiver
812 * could have different page sizes, and also mtu could
813 * be less than the receiver's page size ?
817 off = MIN(mtu, m0->m_pkthdr.len);
820 * firstlen (off - hlen) must be aligned on an
824 goto smart_frag_failure;
825 off = ((off - hlen) & ~7) + hlen;
826 newlen = (~PAGE_MASK) & mtu;
827 if ((newlen + sizeof (struct ip)) > mtu) {
828 /* we failed, go back the default */
839 firstlen = off - hlen;
840 mnext = &m0->m_nextpkt; /* pointer to next packet */
843 * Loop through length of segment after first fragment,
844 * make new header and copy data of each part and link onto chain.
845 * Here, m0 is the original packet, m is the fragment being created.
846 * The fragments are linked off the m_nextpkt of the original
847 * packet, which after processing serves as the first fragment.
849 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
850 struct ip *mhip; /* ip header on the fragment */
852 int mhlen = sizeof (struct ip);
854 m = m_gethdr(M_NOWAIT, MT_DATA);
857 IPSTAT_INC(ips_odropped);
861 * Make sure the complete packet header gets copied
862 * from the originating mbuf to the newly created
863 * mbuf. This also ensures that existing firewall
864 * classification(s), VLAN tags and so on get copied
865 * to the resulting fragmented packet(s):
867 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
870 IPSTAT_INC(ips_odropped);
874 * In the first mbuf, leave room for the link header, then
875 * copy the original IP header including options. The payload
876 * goes into an additional mbuf chain returned by m_copym().
878 m->m_data += max_linkhdr;
879 mhip = mtod(m, struct ip *);
881 if (hlen > sizeof (struct ip)) {
882 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
883 mhip->ip_v = IPVERSION;
884 mhip->ip_hl = mhlen >> 2;
887 /* XXX do we need to add ip_off below ? */
888 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
889 if (off + len >= ip_len)
892 mhip->ip_off |= IP_MF;
893 mhip->ip_len = htons((u_short)(len + mhlen));
894 m->m_next = m_copym(m0, off, len, M_NOWAIT);
895 if (m->m_next == NULL) { /* copy failed */
897 error = ENOBUFS; /* ??? */
898 IPSTAT_INC(ips_odropped);
901 m->m_pkthdr.len = mhlen + len;
903 mac_netinet_fragment(m0, m);
905 mhip->ip_off = htons(mhip->ip_off);
907 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
908 mhip->ip_sum = in_cksum(m, mhlen);
909 m->m_pkthdr.csum_flags &= ~CSUM_IP;
912 mnext = &m->m_nextpkt;
914 IPSTAT_ADD(ips_ofragments, nfrags);
917 * Update first fragment by trimming what's been copied out
918 * and updating header.
920 m_adj(m0, hlen + firstlen - ip_len);
921 m0->m_pkthdr.len = hlen + firstlen;
922 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
923 ip->ip_off = htons(ip_off | IP_MF);
925 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
926 ip->ip_sum = in_cksum(m0, hlen);
927 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
936 in_delayed_cksum(struct mbuf *m)
939 uint16_t csum, offset, ip_len;
941 ip = mtod(m, struct ip *);
942 offset = ip->ip_hl << 2 ;
943 ip_len = ntohs(ip->ip_len);
944 csum = in_cksum_skip(m, ip_len, offset);
945 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
947 offset += m->m_pkthdr.csum_data; /* checksum offset */
949 /* find the mbuf in the chain where the checksum starts*/
950 while ((m != NULL) && (offset >= m->m_len)) {
954 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
955 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
956 *(u_short *)(m->m_data + offset) = csum;
960 * IP socket option processing.
963 ip_ctloutput(struct socket *so, struct sockopt *sopt)
965 struct inpcb *inp = sotoinpcb(so);
973 if (sopt->sopt_level != IPPROTO_IP) {
976 if (sopt->sopt_level == SOL_SOCKET &&
977 sopt->sopt_dir == SOPT_SET) {
978 switch (sopt->sopt_name) {
981 if ((so->so_options & SO_REUSEADDR) != 0)
982 inp->inp_flags2 |= INP_REUSEADDR;
984 inp->inp_flags2 &= ~INP_REUSEADDR;
990 if ((so->so_options & SO_REUSEPORT) != 0)
991 inp->inp_flags2 |= INP_REUSEPORT;
993 inp->inp_flags2 &= ~INP_REUSEPORT;
999 inp->inp_inc.inc_fibnum = so->so_fibnum;
1003 case SO_MAX_PACING_RATE:
1006 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1020 switch (sopt->sopt_dir) {
1022 switch (sopt->sopt_name) {
1029 if (sopt->sopt_valsize > MLEN) {
1033 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1038 m->m_len = sopt->sopt_valsize;
1039 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1046 error = ip_pcbopts(inp, sopt->sopt_name, m);
1052 if (sopt->sopt_td != NULL) {
1053 error = priv_check(sopt->sopt_td,
1054 PRIV_NETINET_BINDANY);
1061 case IP_RSS_LISTEN_BUCKET:
1067 case IP_RECVRETOPTS:
1068 case IP_ORIGDSTADDR:
1069 case IP_RECVDSTADDR:
1077 case IP_RECVRSSBUCKETID:
1079 error = sooptcopyin(sopt, &optval, sizeof optval,
1084 switch (sopt->sopt_name) {
1086 inp->inp_ip_tos = optval;
1090 inp->inp_ip_ttl = optval;
1094 if (optval >= 0 && optval <= MAXTTL)
1095 inp->inp_ip_minttl = optval;
1100 #define OPTSET(bit) do { \
1103 inp->inp_flags |= bit; \
1105 inp->inp_flags &= ~bit; \
1109 #define OPTSET2(bit, val) do { \
1112 inp->inp_flags2 |= bit; \
1114 inp->inp_flags2 &= ~bit; \
1119 OPTSET(INP_RECVOPTS);
1122 case IP_RECVRETOPTS:
1123 OPTSET(INP_RECVRETOPTS);
1126 case IP_RECVDSTADDR:
1127 OPTSET(INP_RECVDSTADDR);
1130 case IP_ORIGDSTADDR:
1131 OPTSET2(INP_ORIGDSTADDR, optval);
1135 OPTSET(INP_RECVTTL);
1143 OPTSET(INP_ONESBCAST);
1146 OPTSET(INP_DONTFRAG);
1149 OPTSET(INP_BINDANY);
1152 OPTSET(INP_RECVTOS);
1155 OPTSET2(INP_BINDMULTI, optval);
1158 OPTSET2(INP_RECVFLOWID, optval);
1161 case IP_RSS_LISTEN_BUCKET:
1162 if ((optval >= 0) &&
1163 (optval < rss_getnumbuckets())) {
1164 inp->inp_rss_listen_bucket = optval;
1165 OPTSET2(INP_RSS_BUCKET_SET, 1);
1170 case IP_RECVRSSBUCKETID:
1171 OPTSET2(INP_RECVRSSBUCKETID, optval);
1180 * Multicast socket options are processed by the in_mcast
1183 case IP_MULTICAST_IF:
1184 case IP_MULTICAST_VIF:
1185 case IP_MULTICAST_TTL:
1186 case IP_MULTICAST_LOOP:
1187 case IP_ADD_MEMBERSHIP:
1188 case IP_DROP_MEMBERSHIP:
1189 case IP_ADD_SOURCE_MEMBERSHIP:
1190 case IP_DROP_SOURCE_MEMBERSHIP:
1191 case IP_BLOCK_SOURCE:
1192 case IP_UNBLOCK_SOURCE:
1194 case MCAST_JOIN_GROUP:
1195 case MCAST_LEAVE_GROUP:
1196 case MCAST_JOIN_SOURCE_GROUP:
1197 case MCAST_LEAVE_SOURCE_GROUP:
1198 case MCAST_BLOCK_SOURCE:
1199 case MCAST_UNBLOCK_SOURCE:
1200 error = inp_setmoptions(inp, sopt);
1204 error = sooptcopyin(sopt, &optval, sizeof optval,
1211 case IP_PORTRANGE_DEFAULT:
1212 inp->inp_flags &= ~(INP_LOWPORT);
1213 inp->inp_flags &= ~(INP_HIGHPORT);
1216 case IP_PORTRANGE_HIGH:
1217 inp->inp_flags &= ~(INP_LOWPORT);
1218 inp->inp_flags |= INP_HIGHPORT;
1221 case IP_PORTRANGE_LOW:
1222 inp->inp_flags &= ~(INP_HIGHPORT);
1223 inp->inp_flags |= INP_LOWPORT;
1233 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1234 case IP_IPSEC_POLICY:
1235 if (IPSEC_ENABLED(ipv4)) {
1236 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1243 error = ENOPROTOOPT;
1249 switch (sopt->sopt_name) {
1252 if (inp->inp_options)
1253 error = sooptcopyout(sopt,
1254 mtod(inp->inp_options,
1256 inp->inp_options->m_len);
1258 sopt->sopt_valsize = 0;
1265 case IP_RECVRETOPTS:
1266 case IP_ORIGDSTADDR:
1267 case IP_RECVDSTADDR:
1280 case IP_RSSBUCKETID:
1281 case IP_RECVRSSBUCKETID:
1283 switch (sopt->sopt_name) {
1286 optval = inp->inp_ip_tos;
1290 optval = inp->inp_ip_ttl;
1294 optval = inp->inp_ip_minttl;
1297 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1298 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1301 optval = OPTBIT(INP_RECVOPTS);
1304 case IP_RECVRETOPTS:
1305 optval = OPTBIT(INP_RECVRETOPTS);
1308 case IP_RECVDSTADDR:
1309 optval = OPTBIT(INP_RECVDSTADDR);
1312 case IP_ORIGDSTADDR:
1313 optval = OPTBIT2(INP_ORIGDSTADDR);
1317 optval = OPTBIT(INP_RECVTTL);
1321 optval = OPTBIT(INP_RECVIF);
1325 if (inp->inp_flags & INP_HIGHPORT)
1326 optval = IP_PORTRANGE_HIGH;
1327 else if (inp->inp_flags & INP_LOWPORT)
1328 optval = IP_PORTRANGE_LOW;
1334 optval = OPTBIT(INP_ONESBCAST);
1337 optval = OPTBIT(INP_DONTFRAG);
1340 optval = OPTBIT(INP_BINDANY);
1343 optval = OPTBIT(INP_RECVTOS);
1346 optval = inp->inp_flowid;
1349 optval = inp->inp_flowtype;
1352 optval = OPTBIT2(INP_RECVFLOWID);
1355 case IP_RSSBUCKETID:
1356 retval = rss_hash2bucket(inp->inp_flowid,
1360 optval = rss_bucket;
1364 case IP_RECVRSSBUCKETID:
1365 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1369 optval = OPTBIT2(INP_BINDMULTI);
1372 error = sooptcopyout(sopt, &optval, sizeof optval);
1376 * Multicast socket options are processed by the in_mcast
1379 case IP_MULTICAST_IF:
1380 case IP_MULTICAST_VIF:
1381 case IP_MULTICAST_TTL:
1382 case IP_MULTICAST_LOOP:
1384 error = inp_getmoptions(inp, sopt);
1387 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1388 case IP_IPSEC_POLICY:
1389 if (IPSEC_ENABLED(ipv4)) {
1390 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1397 error = ENOPROTOOPT;
1406 * Routine called from ip_output() to loop back a copy of an IP multicast
1407 * packet to the input queue of a specified interface. Note that this
1408 * calls the output routine of the loopback "driver", but with an interface
1409 * pointer that might NOT be a loopback interface -- evil, but easier than
1410 * replicating that code here.
1413 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1419 * Make a deep copy of the packet because we're going to
1420 * modify the pack in order to generate checksums.
1422 copym = m_dup(m, M_NOWAIT);
1423 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1424 copym = m_pullup(copym, hlen);
1425 if (copym != NULL) {
1426 /* If needed, compute the checksum and mark it as valid. */
1427 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1428 in_delayed_cksum(copym);
1429 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1430 copym->m_pkthdr.csum_flags |=
1431 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1432 copym->m_pkthdr.csum_data = 0xffff;
1435 * We don't bother to fragment if the IP length is greater
1436 * than the interface's MTU. Can this possibly matter?
1438 ip = mtod(copym, struct ip *);
1440 ip->ip_sum = in_cksum(copym, hlen);
1441 if_simloop(ifp, copym, AF_INET, 0);