2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
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>
88 #include <netinet/ip_ipsec.h>
89 #include <netipsec/ipsec.h>
92 #include <machine/in_cksum.h>
94 #include <security/mac/mac_framework.h>
96 #ifdef MBUF_STRESS_TEST
97 static int mbuf_frag_size = 0;
98 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
99 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
102 static void ip_mloopback(struct ifnet *, const struct mbuf *, int);
105 extern int in_mcast_loop;
106 extern struct protosw inetsw[];
109 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
110 struct sockaddr_in *dst, int *fibnum, int *error)
112 struct m_tag *fwd_tag = NULL;
118 ip = mtod(m, struct ip *);
120 /* Run through list of hooks for output packets. */
121 odst.s_addr = ip->ip_dst.s_addr;
122 *error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, inp);
124 if ((*error) != 0 || m == NULL)
125 return 1; /* Finished */
127 ip = mtod(m, struct ip *);
129 /* See if destination IP address was changed by packet filter. */
130 if (odst.s_addr != ip->ip_dst.s_addr) {
131 m->m_flags |= M_SKIP_FIREWALL;
132 /* If destination is now ourself drop to ip_input(). */
133 if (in_localip(ip->ip_dst)) {
134 m->m_flags |= M_FASTFWD_OURS;
135 if (m->m_pkthdr.rcvif == NULL)
136 m->m_pkthdr.rcvif = V_loif;
137 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
138 m->m_pkthdr.csum_flags |=
139 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
140 m->m_pkthdr.csum_data = 0xffff;
142 m->m_pkthdr.csum_flags |=
143 CSUM_IP_CHECKED | CSUM_IP_VALID;
145 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
146 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
148 *error = netisr_queue(NETISR_IP, m);
149 return 1; /* Finished */
152 bzero(dst, sizeof(*dst));
153 dst->sin_family = AF_INET;
154 dst->sin_len = sizeof(*dst);
155 dst->sin_addr = ip->ip_dst;
157 return -1; /* Reloop */
159 /* See if fib was changed by packet filter. */
160 if ((*fibnum) != M_GETFIB(m)) {
161 m->m_flags |= M_SKIP_FIREWALL;
162 *fibnum = M_GETFIB(m);
163 return -1; /* Reloop for FIB change */
166 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
167 if (m->m_flags & M_FASTFWD_OURS) {
168 if (m->m_pkthdr.rcvif == NULL)
169 m->m_pkthdr.rcvif = V_loif;
170 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
171 m->m_pkthdr.csum_flags |=
172 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
173 m->m_pkthdr.csum_data = 0xffff;
176 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
177 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
179 m->m_pkthdr.csum_flags |=
180 CSUM_IP_CHECKED | CSUM_IP_VALID;
182 *error = netisr_queue(NETISR_IP, m);
183 return 1; /* Finished */
185 /* Or forward to some other address? */
186 if ((m->m_flags & M_IP_NEXTHOP) &&
187 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
188 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
189 m->m_flags |= M_SKIP_FIREWALL;
190 m->m_flags &= ~M_IP_NEXTHOP;
191 m_tag_delete(m, fwd_tag);
193 return -1; /* Reloop for CHANGE of dst */
200 * IP output. The packet in mbuf chain m contains a skeletal IP
201 * header (with len, off, ttl, proto, tos, src, dst).
202 * The mbuf chain containing the packet will be freed.
203 * The mbuf opt, if present, will not be freed.
204 * If route ro is present and has ro_rt initialized, route lookup would be
205 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
206 * then result of route lookup is stored in ro->ro_rt.
208 * In the IP forwarding case, the packet will arrive with options already
209 * inserted, so must have a NULL opt pointer.
212 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
213 struct ip_moptions *imo, struct inpcb *inp)
215 struct rm_priotracker in_ifa_tracker;
217 struct ifnet *ifp = NULL; /* keep compiler happy */
219 int hlen = sizeof (struct ip);
222 struct sockaddr_in *dst;
223 const struct sockaddr_in *gw;
224 struct in_ifaddr *ia;
226 uint16_t ip_len, ip_off;
227 struct route iproute;
228 struct rtentry *rte; /* cache for ro->ro_rt */
232 int no_route_but_check_spd = 0;
237 INP_LOCK_ASSERT(inp);
238 M_SETFIB(m, inp->inp_inc.inc_fibnum);
239 if ((flags & IP_NODEFAULTFLOWID) == 0) {
240 m->m_pkthdr.flowid = inp->inp_flowid;
241 M_HASHTYPE_SET(m, inp->inp_flowtype);
247 bzero(ro, sizeof (*ro));
249 ro->ro_flags |= RT_LLE_CACHE;
252 if (ro->ro_rt == NULL)
253 (void )flowtable_lookup(AF_INET, m, ro);
258 m = ip_insertoptions(m, opt, &len);
260 hlen = len; /* ip->ip_hl is updated above */
262 ip = mtod(m, struct ip *);
263 ip_len = ntohs(ip->ip_len);
264 ip_off = ntohs(ip->ip_off);
266 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
267 ip->ip_v = IPVERSION;
268 ip->ip_hl = hlen >> 2;
270 IPSTAT_INC(ips_localout);
272 /* Header already set, fetch hlen from there */
273 hlen = ip->ip_hl << 2;
279 * dst can be rewritten but always points to &ro->ro_dst.
280 * gw is readonly but can point either to dst OR rt_gateway,
281 * therefore we need restore gw if we're redoing lookup.
283 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
284 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
287 bzero(dst, sizeof(*dst));
288 dst->sin_family = AF_INET;
289 dst->sin_len = sizeof(*dst);
290 dst->sin_addr = ip->ip_dst;
294 * Validate route against routing table additions;
295 * a better/more specific route might have been added.
298 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
300 * If there is a cached route,
301 * check that it is to the same destination
302 * and is still up. If not, free it and try again.
303 * The address family should also be checked in case of sharing the
305 * Also check whether routing cache needs invalidation.
308 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
309 rte->rt_ifp == NULL ||
310 !RT_LINK_IS_UP(rte->rt_ifp) ||
311 dst->sin_family != AF_INET ||
312 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
314 rte = ro->ro_rt = (struct rtentry *)NULL;
316 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
317 ro->ro_lle = (struct llentry *)NULL;
322 * If routing to interface only, short circuit routing lookup.
323 * The use of an all-ones broadcast address implies this; an
324 * interface is specified by the broadcast address of an interface,
325 * or the destination address of a ptp interface.
327 if (flags & IP_SENDONES) {
328 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
329 M_GETFIB(m)))) == NULL &&
330 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
331 M_GETFIB(m)))) == NULL) {
332 IPSTAT_INC(ips_noroute);
337 ip->ip_dst.s_addr = INADDR_BROADCAST;
338 dst->sin_addr = ip->ip_dst;
342 } else if (flags & IP_ROUTETOIF) {
343 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
344 M_GETFIB(m)))) == NULL &&
345 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
346 M_GETFIB(m)))) == NULL) {
347 IPSTAT_INC(ips_noroute);
354 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
355 in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
356 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
357 imo != NULL && imo->imo_multicast_ifp != NULL) {
359 * Bypass the normal routing lookup for multicast
360 * packets if the interface is specified.
362 ifp = imo->imo_multicast_ifp;
363 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
366 isbroadcast = 0; /* fool gcc */
369 * We want to do any cloning requested by the link layer,
370 * as this is probably required in all cases for correct
371 * operation (as it is for ARP).
375 rtalloc_mpath_fib(ro,
376 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
379 in_rtalloc_ign(ro, 0, fibnum);
384 (rte->rt_flags & RTF_UP) == 0 ||
385 rte->rt_ifp == NULL ||
386 !RT_LINK_IS_UP(rte->rt_ifp)) {
389 * There is no route for this packet, but it is
390 * possible that a matching SPD entry exists.
392 no_route_but_check_spd = 1;
393 mtu = 0; /* Silence GCC warning. */
396 IPSTAT_INC(ips_noroute);
397 error = EHOSTUNREACH;
400 ia = ifatoia(rte->rt_ifa);
402 counter_u64_add(rte->rt_pksent, 1);
403 rt_update_ro_flags(ro);
404 if (rte->rt_flags & RTF_GATEWAY)
405 gw = (struct sockaddr_in *)rte->rt_gateway;
406 if (rte->rt_flags & RTF_HOST)
407 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
408 else if (ifp->if_flags & IFF_BROADCAST)
409 isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
415 * Calculate MTU. If we have a route that is up, use that,
416 * otherwise use the interface's MTU.
418 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
422 /* Catch a possible divide by zero later. */
423 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
424 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
426 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
427 m->m_flags |= M_MCAST;
429 * IP destination address is multicast. Make sure "gw"
430 * still points to the address in "ro". (It may have been
431 * changed to point to a gateway address, above.)
435 * See if the caller provided any multicast options
438 ip->ip_ttl = imo->imo_multicast_ttl;
439 if (imo->imo_multicast_vif != -1)
442 ip_mcast_src(imo->imo_multicast_vif) :
445 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
447 * Confirm that the outgoing interface supports multicast.
449 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
450 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
451 IPSTAT_INC(ips_noroute);
457 * If source address not specified yet, use address
458 * of outgoing interface.
460 if (ip->ip_src.s_addr == INADDR_ANY) {
461 /* Interface may have no addresses. */
463 ip->ip_src = IA_SIN(ia)->sin_addr;
466 if ((imo == NULL && in_mcast_loop) ||
467 (imo && imo->imo_multicast_loop)) {
469 * Loop back multicast datagram if not expressly
470 * forbidden to do so, even if we are not a member
471 * of the group; ip_input() will filter it later,
472 * thus deferring a hash lookup and mutex acquisition
473 * at the expense of a cheap copy using m_copym().
475 ip_mloopback(ifp, m, hlen);
478 * If we are acting as a multicast router, perform
479 * multicast forwarding as if the packet had just
480 * arrived on the interface to which we are about
481 * to send. The multicast forwarding function
482 * recursively calls this function, using the
483 * IP_FORWARDING flag to prevent infinite recursion.
485 * Multicasts that are looped back by ip_mloopback(),
486 * above, will be forwarded by the ip_input() routine,
489 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
491 * If rsvp daemon is not running, do not
492 * set ip_moptions. This ensures that the packet
493 * is multicast and not just sent down one link
494 * as prescribed by rsvpd.
499 ip_mforward(ip, ifp, m, imo) != 0) {
507 * Multicasts with a time-to-live of zero may be looped-
508 * back, above, but must not be transmitted on a network.
509 * Also, multicasts addressed to the loopback interface
510 * are not sent -- the above call to ip_mloopback() will
511 * loop back a copy. ip_input() will drop the copy if
512 * this host does not belong to the destination group on
513 * the loopback interface.
515 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
524 * If the source address is not specified yet, use the address
525 * of the outoing interface.
527 if (ip->ip_src.s_addr == INADDR_ANY) {
528 /* Interface may have no addresses. */
530 ip->ip_src = IA_SIN(ia)->sin_addr;
535 * Look for broadcast address and
536 * verify user is allowed to send
540 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
541 error = EADDRNOTAVAIL;
544 if ((flags & IP_ALLOWBROADCAST) == 0) {
548 /* don't allow broadcast messages to be fragmented */
553 m->m_flags |= M_BCAST;
555 m->m_flags &= ~M_BCAST;
560 switch(ip_ipsec_output(&m, inp, &error)) {
567 break; /* Continue with packet processing. */
570 * Check if there was a route for this packet; return error if not.
572 if (no_route_but_check_spd) {
573 IPSTAT_INC(ips_noroute);
574 error = EHOSTUNREACH;
577 /* Update variables that are affected by ipsec4_output(). */
578 ip = mtod(m, struct ip *);
579 hlen = ip->ip_hl << 2;
582 /* Jump over all PFIL processing if hooks are not active. */
583 if (PFIL_HOOKED(&V_inet_pfil_hook)) {
584 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
585 case 1: /* Finished */
588 case 0: /* Continue normally */
589 ip = mtod(m, struct ip *);
592 case -1: /* Need to try again */
593 /* Reset everything for a new round */
596 ifa_free(&ia->ia_ifa);
597 ro->ro_prepend = NULL;
600 ip = mtod(m, struct ip *);
606 /* 127/8 must not appear on wire - RFC1122. */
607 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
608 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
609 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
610 IPSTAT_INC(ips_badaddr);
611 error = EADDRNOTAVAIL;
616 m->m_pkthdr.csum_flags |= CSUM_IP;
617 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
619 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
622 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
623 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
624 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
629 * If small enough for interface, or the interface will take
630 * care of the fragmentation for us, we can just send directly.
633 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
635 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
636 ip->ip_sum = in_cksum(m, hlen);
637 m->m_pkthdr.csum_flags &= ~CSUM_IP;
641 * Record statistics for this interface address.
642 * With CSUM_TSO the byte/packet count will be slightly
643 * incorrect because we count the IP+TCP headers only
644 * once instead of for every generated packet.
646 if (!(flags & IP_FORWARDING) && ia) {
647 if (m->m_pkthdr.csum_flags & CSUM_TSO)
648 counter_u64_add(ia->ia_ifa.ifa_opackets,
649 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
651 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
653 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
655 #ifdef MBUF_STRESS_TEST
656 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
657 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
660 * Reset layer specific mbuf flags
661 * to avoid confusing lower layers.
664 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
667 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
668 in_pcboutput_txrtlmt(inp, ifp, m);
669 /* stamp send tag on mbuf */
670 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
672 m->m_pkthdr.snd_tag = NULL;
675 error = (*ifp->if_output)(ifp, m,
676 (const struct sockaddr *)gw, ro);
678 /* check for route change */
680 in_pcboutput_eagain(inp);
685 /* Balk when DF bit is set or the interface didn't support TSO. */
686 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
688 IPSTAT_INC(ips_cantfrag);
693 * Too large for interface; fragment if possible. If successful,
694 * on return, m will point to a list of packets to be sent.
696 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
703 /* Record statistics for this interface address. */
705 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
706 counter_u64_add(ia->ia_ifa.ifa_obytes,
710 * Reset layer specific mbuf flags
711 * to avoid confusing upper layers.
715 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
716 mtod(m, struct ip *), NULL);
719 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
720 in_pcboutput_txrtlmt(inp, ifp, m);
721 /* stamp send tag on mbuf */
722 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
724 m->m_pkthdr.snd_tag = NULL;
727 error = (*ifp->if_output)(ifp, m,
728 (const struct sockaddr *)gw, ro);
730 /* check for route change */
732 in_pcboutput_eagain(inp);
739 IPSTAT_INC(ips_fragmented);
744 else if (rte == NULL)
746 * If the caller supplied a route but somehow the reference
747 * to it has been released need to prevent the caller
748 * calling RTFREE on it again.
752 ifa_free(&ia->ia_ifa);
760 * Create a chain of fragments which fit the given mtu. m_frag points to the
761 * mbuf to be fragmented; on return it points to the chain with the fragments.
762 * Return 0 if no error. If error, m_frag may contain a partially built
763 * chain of fragments that should be freed by the caller.
765 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
768 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
769 u_long if_hwassist_flags)
772 int hlen = ip->ip_hl << 2;
773 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
775 struct mbuf *m0 = *m_frag; /* the original packet */
779 uint16_t ip_len, ip_off;
781 ip_len = ntohs(ip->ip_len);
782 ip_off = ntohs(ip->ip_off);
784 if (ip_off & IP_DF) { /* Fragmentation not allowed */
785 IPSTAT_INC(ips_cantfrag);
790 * Must be able to put at least 8 bytes per fragment.
796 * If the interface will not calculate checksums on
797 * fragmented packets, then do it here.
799 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
800 in_delayed_cksum(m0);
801 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
804 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
805 sctp_delayed_cksum(m0, hlen);
806 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
809 if (len > PAGE_SIZE) {
811 * Fragment large datagrams such that each segment
812 * contains a multiple of PAGE_SIZE amount of data,
813 * plus headers. This enables a receiver to perform
814 * page-flipping zero-copy optimizations.
816 * XXX When does this help given that sender and receiver
817 * could have different page sizes, and also mtu could
818 * be less than the receiver's page size ?
822 off = MIN(mtu, m0->m_pkthdr.len);
825 * firstlen (off - hlen) must be aligned on an
829 goto smart_frag_failure;
830 off = ((off - hlen) & ~7) + hlen;
831 newlen = (~PAGE_MASK) & mtu;
832 if ((newlen + sizeof (struct ip)) > mtu) {
833 /* we failed, go back the default */
844 firstlen = off - hlen;
845 mnext = &m0->m_nextpkt; /* pointer to next packet */
848 * Loop through length of segment after first fragment,
849 * make new header and copy data of each part and link onto chain.
850 * Here, m0 is the original packet, m is the fragment being created.
851 * The fragments are linked off the m_nextpkt of the original
852 * packet, which after processing serves as the first fragment.
854 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
855 struct ip *mhip; /* ip header on the fragment */
857 int mhlen = sizeof (struct ip);
859 m = m_gethdr(M_NOWAIT, MT_DATA);
862 IPSTAT_INC(ips_odropped);
866 * Make sure the complete packet header gets copied
867 * from the originating mbuf to the newly created
868 * mbuf. This also ensures that existing firewall
869 * classification(s), VLAN tags and so on get copied
870 * to the resulting fragmented packet(s):
872 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
875 IPSTAT_INC(ips_odropped);
879 * In the first mbuf, leave room for the link header, then
880 * copy the original IP header including options. The payload
881 * goes into an additional mbuf chain returned by m_copym().
883 m->m_data += max_linkhdr;
884 mhip = mtod(m, struct ip *);
886 if (hlen > sizeof (struct ip)) {
887 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
888 mhip->ip_v = IPVERSION;
889 mhip->ip_hl = mhlen >> 2;
892 /* XXX do we need to add ip_off below ? */
893 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
894 if (off + len >= ip_len)
897 mhip->ip_off |= IP_MF;
898 mhip->ip_len = htons((u_short)(len + mhlen));
899 m->m_next = m_copym(m0, off, len, M_NOWAIT);
900 if (m->m_next == NULL) { /* copy failed */
902 error = ENOBUFS; /* ??? */
903 IPSTAT_INC(ips_odropped);
906 m->m_pkthdr.len = mhlen + len;
908 mac_netinet_fragment(m0, m);
910 mhip->ip_off = htons(mhip->ip_off);
912 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
913 mhip->ip_sum = in_cksum(m, mhlen);
914 m->m_pkthdr.csum_flags &= ~CSUM_IP;
917 mnext = &m->m_nextpkt;
919 IPSTAT_ADD(ips_ofragments, nfrags);
922 * Update first fragment by trimming what's been copied out
923 * and updating header.
925 m_adj(m0, hlen + firstlen - ip_len);
926 m0->m_pkthdr.len = hlen + firstlen;
927 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
928 ip->ip_off = htons(ip_off | IP_MF);
930 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
931 ip->ip_sum = in_cksum(m0, hlen);
932 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
941 in_delayed_cksum(struct mbuf *m)
944 uint16_t csum, offset, ip_len;
946 ip = mtod(m, struct ip *);
947 offset = ip->ip_hl << 2 ;
948 ip_len = ntohs(ip->ip_len);
949 csum = in_cksum_skip(m, ip_len, offset);
950 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
952 offset += m->m_pkthdr.csum_data; /* checksum offset */
954 /* find the mbuf in the chain where the checksum starts*/
955 while ((m != NULL) && (offset >= m->m_len)) {
959 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
960 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
961 *(u_short *)(m->m_data + offset) = csum;
965 * IP socket option processing.
968 ip_ctloutput(struct socket *so, struct sockopt *sopt)
970 struct inpcb *inp = sotoinpcb(so);
978 if (sopt->sopt_level != IPPROTO_IP) {
981 if (sopt->sopt_level == SOL_SOCKET &&
982 sopt->sopt_dir == SOPT_SET) {
983 switch (sopt->sopt_name) {
986 if ((so->so_options & SO_REUSEADDR) != 0)
987 inp->inp_flags2 |= INP_REUSEADDR;
989 inp->inp_flags2 &= ~INP_REUSEADDR;
995 if ((so->so_options & SO_REUSEPORT) != 0)
996 inp->inp_flags2 |= INP_REUSEPORT;
998 inp->inp_flags2 &= ~INP_REUSEPORT;
1004 inp->inp_inc.inc_fibnum = so->so_fibnum;
1008 case SO_MAX_PACING_RATE:
1011 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1025 switch (sopt->sopt_dir) {
1027 switch (sopt->sopt_name) {
1034 if (sopt->sopt_valsize > MLEN) {
1038 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1043 m->m_len = sopt->sopt_valsize;
1044 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1051 error = ip_pcbopts(inp, sopt->sopt_name, m);
1057 if (sopt->sopt_td != NULL) {
1058 error = priv_check(sopt->sopt_td,
1059 PRIV_NETINET_BINDANY);
1066 case IP_RSS_LISTEN_BUCKET:
1072 case IP_RECVRETOPTS:
1073 case IP_RECVDSTADDR:
1081 case IP_RECVRSSBUCKETID:
1083 error = sooptcopyin(sopt, &optval, sizeof optval,
1088 switch (sopt->sopt_name) {
1090 inp->inp_ip_tos = optval;
1094 inp->inp_ip_ttl = optval;
1098 if (optval >= 0 && optval <= MAXTTL)
1099 inp->inp_ip_minttl = optval;
1104 #define OPTSET(bit) do { \
1107 inp->inp_flags |= bit; \
1109 inp->inp_flags &= ~bit; \
1113 #define OPTSET2(bit, val) do { \
1116 inp->inp_flags2 |= bit; \
1118 inp->inp_flags2 &= ~bit; \
1123 OPTSET(INP_RECVOPTS);
1126 case IP_RECVRETOPTS:
1127 OPTSET(INP_RECVRETOPTS);
1130 case IP_RECVDSTADDR:
1131 OPTSET(INP_RECVDSTADDR);
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;
1234 case IP_IPSEC_POLICY:
1239 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1241 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1243 req = mtod(m, caddr_t);
1244 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1245 m->m_len, (sopt->sopt_td != NULL) ?
1246 sopt->sopt_td->td_ucred : NULL);
1253 error = ENOPROTOOPT;
1259 switch (sopt->sopt_name) {
1262 if (inp->inp_options)
1263 error = sooptcopyout(sopt,
1264 mtod(inp->inp_options,
1266 inp->inp_options->m_len);
1268 sopt->sopt_valsize = 0;
1275 case IP_RECVRETOPTS:
1276 case IP_RECVDSTADDR:
1289 case IP_RSSBUCKETID:
1290 case IP_RECVRSSBUCKETID:
1292 switch (sopt->sopt_name) {
1295 optval = inp->inp_ip_tos;
1299 optval = inp->inp_ip_ttl;
1303 optval = inp->inp_ip_minttl;
1306 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1307 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1310 optval = OPTBIT(INP_RECVOPTS);
1313 case IP_RECVRETOPTS:
1314 optval = OPTBIT(INP_RECVRETOPTS);
1317 case IP_RECVDSTADDR:
1318 optval = OPTBIT(INP_RECVDSTADDR);
1322 optval = OPTBIT(INP_RECVTTL);
1326 optval = OPTBIT(INP_RECVIF);
1330 if (inp->inp_flags & INP_HIGHPORT)
1331 optval = IP_PORTRANGE_HIGH;
1332 else if (inp->inp_flags & INP_LOWPORT)
1333 optval = IP_PORTRANGE_LOW;
1339 optval = OPTBIT(INP_ONESBCAST);
1342 optval = OPTBIT(INP_DONTFRAG);
1345 optval = OPTBIT(INP_BINDANY);
1348 optval = OPTBIT(INP_RECVTOS);
1351 optval = inp->inp_flowid;
1354 optval = inp->inp_flowtype;
1357 optval = OPTBIT2(INP_RECVFLOWID);
1360 case IP_RSSBUCKETID:
1361 retval = rss_hash2bucket(inp->inp_flowid,
1365 optval = rss_bucket;
1369 case IP_RECVRSSBUCKETID:
1370 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1374 optval = OPTBIT2(INP_BINDMULTI);
1377 error = sooptcopyout(sopt, &optval, sizeof optval);
1381 * Multicast socket options are processed by the in_mcast
1384 case IP_MULTICAST_IF:
1385 case IP_MULTICAST_VIF:
1386 case IP_MULTICAST_TTL:
1387 case IP_MULTICAST_LOOP:
1389 error = inp_getmoptions(inp, sopt);
1393 case IP_IPSEC_POLICY:
1395 struct mbuf *m = NULL;
1400 req = mtod(m, caddr_t);
1403 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1405 error = soopt_mcopyout(sopt, m); /* XXX */
1413 error = ENOPROTOOPT;
1422 * Routine called from ip_output() to loop back a copy of an IP multicast
1423 * packet to the input queue of a specified interface. Note that this
1424 * calls the output routine of the loopback "driver", but with an interface
1425 * pointer that might NOT be a loopback interface -- evil, but easier than
1426 * replicating that code here.
1429 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1435 * Make a deep copy of the packet because we're going to
1436 * modify the pack in order to generate checksums.
1438 copym = m_dup(m, M_NOWAIT);
1439 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1440 copym = m_pullup(copym, hlen);
1441 if (copym != NULL) {
1442 /* If needed, compute the checksum and mark it as valid. */
1443 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1444 in_delayed_cksum(copym);
1445 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1446 copym->m_pkthdr.csum_flags |=
1447 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1448 copym->m_pkthdr.csum_data = 0xffff;
1451 * We don't bother to fragment if the IP length is greater
1452 * than the interface's MTU. Can this possibly matter?
1454 ip = mtod(copym, struct ip *);
1456 ip->ip_sum = in_cksum(copym, hlen);
1457 if_simloop(ifp, copym, AF_INET, 0);