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("$FreeBSD$");
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>
47 #include <sys/malloc.h>
51 #include <sys/protosw.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sysctl.h>
56 #include <sys/ucred.h>
59 #include <net/if_var.h>
60 #include <net/if_llatbl.h>
61 #include <net/netisr.h>
63 #include <net/route.h>
64 #include <net/flowtable.h>
66 #include <net/radix_mpath.h>
70 #include <netinet/in.h>
71 #include <netinet/in_kdtrace.h>
72 #include <netinet/in_systm.h>
73 #include <netinet/ip.h>
74 #include <netinet/in_pcb.h>
75 #include <netinet/in_rss.h>
76 #include <netinet/in_var.h>
77 #include <netinet/ip_var.h>
78 #include <netinet/ip_options.h>
80 #include <netinet/sctp.h>
81 #include <netinet/sctp_crc32.h>
85 #include <netinet/ip_ipsec.h>
86 #include <netipsec/ipsec.h>
89 #include <machine/in_cksum.h>
91 #include <security/mac/mac_framework.h>
93 VNET_DEFINE(u_short, ip_id);
95 #ifdef MBUF_STRESS_TEST
96 static int mbuf_frag_size = 0;
97 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
98 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
101 static void ip_mloopback
102 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
105 extern int in_mcast_loop;
106 extern struct protosw inetsw[];
109 * IP output. The packet in mbuf chain m contains a skeletal IP
110 * header (with len, off, ttl, proto, tos, src, dst).
111 * The mbuf chain containing the packet will be freed.
112 * The mbuf opt, if present, will not be freed.
113 * If route ro is present and has ro_rt initialized, route lookup would be
114 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
115 * then result of route lookup is stored in ro->ro_rt.
117 * In the IP forwarding case, the packet will arrive with options already
118 * inserted, so must have a NULL opt pointer.
121 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
122 struct ip_moptions *imo, struct inpcb *inp)
125 struct ifnet *ifp = NULL; /* keep compiler happy */
127 int hlen = sizeof (struct ip);
130 struct sockaddr_in *dst;
131 const struct sockaddr_in *gw;
132 struct in_ifaddr *ia;
134 uint16_t ip_len, ip_off;
135 struct route iproute;
136 struct rtentry *rte; /* cache for ro->ro_rt */
138 struct m_tag *fwd_tag = NULL;
140 int no_route_but_check_spd = 0;
145 INP_LOCK_ASSERT(inp);
146 M_SETFIB(m, inp->inp_inc.inc_fibnum);
147 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
148 m->m_pkthdr.flowid = inp->inp_flowid;
149 M_HASHTYPE_SET(m, inp->inp_flowtype);
150 m->m_flags |= M_FLOWID;
156 bzero(ro, sizeof (*ro));
160 if (ro->ro_rt == NULL)
161 (void )flowtable_lookup(AF_INET, m, ro);
166 m = ip_insertoptions(m, opt, &len);
168 hlen = len; /* ip->ip_hl is updated above */
170 ip = mtod(m, struct ip *);
171 ip_len = ntohs(ip->ip_len);
172 ip_off = ntohs(ip->ip_off);
175 * Fill in IP header. If we are not allowing fragmentation,
176 * then the ip_id field is meaningless, but we don't set it
177 * to zero. Doing so causes various problems when devices along
178 * the path (routers, load balancers, firewalls, etc.) illegally
179 * disable DF on our packet. Note that a 16-bit counter
180 * will wrap around in less than 10 seconds at 100 Mbit/s on a
181 * medium with MTU 1500. See Steven M. Bellovin, "A Technique
182 * for Counting NATted Hosts", Proc. IMW'02, available at
183 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
185 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
186 ip->ip_v = IPVERSION;
187 ip->ip_hl = hlen >> 2;
188 ip->ip_id = ip_newid();
189 IPSTAT_INC(ips_localout);
191 /* Header already set, fetch hlen from there */
192 hlen = ip->ip_hl << 2;
198 * dst can be rewritten but always points to &ro->ro_dst.
199 * gw is readonly but can point either to dst OR rt_gateway,
200 * therefore we need restore gw if we're redoing lookup.
202 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
206 * If there is a cached route, check that it is to the same
207 * destination and is still up. If not, free it and try again.
208 * The address family should also be checked in case of sharing
209 * the cache with IPv6.
212 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
213 rte->rt_ifp == NULL ||
214 !RT_LINK_IS_UP(rte->rt_ifp) ||
215 dst->sin_family != AF_INET ||
216 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
222 if (rte == NULL && fwd_tag == NULL) {
223 bzero(dst, sizeof(*dst));
224 dst->sin_family = AF_INET;
225 dst->sin_len = sizeof(*dst);
226 dst->sin_addr = ip->ip_dst;
229 * If routing to interface only, short circuit routing lookup.
230 * The use of an all-ones broadcast address implies this; an
231 * interface is specified by the broadcast address of an interface,
232 * or the destination address of a ptp interface.
234 if (flags & IP_SENDONES) {
235 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
236 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
237 IPSTAT_INC(ips_noroute);
241 ip->ip_dst.s_addr = INADDR_BROADCAST;
242 dst->sin_addr = ip->ip_dst;
246 } else if (flags & IP_ROUTETOIF) {
247 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
248 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) {
249 IPSTAT_INC(ips_noroute);
255 isbroadcast = in_broadcast(dst->sin_addr, ifp);
256 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
257 imo != NULL && imo->imo_multicast_ifp != NULL) {
259 * Bypass the normal routing lookup for multicast
260 * packets if the interface is specified.
262 ifp = imo->imo_multicast_ifp;
264 isbroadcast = 0; /* fool gcc */
267 * We want to do any cloning requested by the link layer,
268 * as this is probably required in all cases for correct
269 * operation (as it is for ARP).
273 rtalloc_mpath_fib(ro,
274 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
275 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
277 in_rtalloc_ign(ro, 0,
278 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
283 rte->rt_ifp == NULL ||
284 !RT_LINK_IS_UP(rte->rt_ifp)) {
287 * There is no route for this packet, but it is
288 * possible that a matching SPD entry exists.
290 no_route_but_check_spd = 1;
291 mtu = 0; /* Silence GCC warning. */
294 IPSTAT_INC(ips_noroute);
295 error = EHOSTUNREACH;
298 ia = ifatoia(rte->rt_ifa);
300 counter_u64_add(rte->rt_pksent, 1);
301 if (rte->rt_flags & RTF_GATEWAY)
302 gw = (struct sockaddr_in *)rte->rt_gateway;
303 if (rte->rt_flags & RTF_HOST)
304 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
306 isbroadcast = in_broadcast(gw->sin_addr, ifp);
309 * Calculate MTU. If we have a route that is up, use that,
310 * otherwise use the interface's MTU.
312 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
314 * This case can happen if the user changed the MTU
315 * of an interface after enabling IP on it. Because
316 * most netifs don't keep track of routes pointing to
317 * them, there is no way for one to update all its
318 * routes when the MTU is changed.
320 if (rte->rt_mtu > ifp->if_mtu)
321 rte->rt_mtu = ifp->if_mtu;
326 /* Catch a possible divide by zero later. */
327 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
328 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
329 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
330 m->m_flags |= M_MCAST;
332 * IP destination address is multicast. Make sure "gw"
333 * still points to the address in "ro". (It may have been
334 * changed to point to a gateway address, above.)
338 * See if the caller provided any multicast options
341 ip->ip_ttl = imo->imo_multicast_ttl;
342 if (imo->imo_multicast_vif != -1)
345 ip_mcast_src(imo->imo_multicast_vif) :
348 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
350 * Confirm that the outgoing interface supports multicast.
352 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
353 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
354 IPSTAT_INC(ips_noroute);
360 * If source address not specified yet, use address
361 * of outgoing interface.
363 if (ip->ip_src.s_addr == INADDR_ANY) {
364 /* Interface may have no addresses. */
366 ip->ip_src = IA_SIN(ia)->sin_addr;
369 if ((imo == NULL && in_mcast_loop) ||
370 (imo && imo->imo_multicast_loop)) {
372 * Loop back multicast datagram if not expressly
373 * forbidden to do so, even if we are not a member
374 * of the group; ip_input() will filter it later,
375 * thus deferring a hash lookup and mutex acquisition
376 * at the expense of a cheap copy using m_copym().
378 ip_mloopback(ifp, m, dst, hlen);
381 * If we are acting as a multicast router, perform
382 * multicast forwarding as if the packet had just
383 * arrived on the interface to which we are about
384 * to send. The multicast forwarding function
385 * recursively calls this function, using the
386 * IP_FORWARDING flag to prevent infinite recursion.
388 * Multicasts that are looped back by ip_mloopback(),
389 * above, will be forwarded by the ip_input() routine,
392 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
394 * If rsvp daemon is not running, do not
395 * set ip_moptions. This ensures that the packet
396 * is multicast and not just sent down one link
397 * as prescribed by rsvpd.
402 ip_mforward(ip, ifp, m, imo) != 0) {
410 * Multicasts with a time-to-live of zero may be looped-
411 * back, above, but must not be transmitted on a network.
412 * Also, multicasts addressed to the loopback interface
413 * are not sent -- the above call to ip_mloopback() will
414 * loop back a copy. ip_input() will drop the copy if
415 * this host does not belong to the destination group on
416 * the loopback interface.
418 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
427 * If the source address is not specified yet, use the address
428 * of the outoing interface.
430 if (ip->ip_src.s_addr == INADDR_ANY) {
431 /* Interface may have no addresses. */
433 ip->ip_src = IA_SIN(ia)->sin_addr;
438 * Both in the SMP world, pre-emption world if_transmit() world,
439 * the following code doesn't really function as intended any further.
441 * + There can and will be multiple CPUs running this code path
442 * in parallel, and we do no lock holding when checking the
444 * + And since other threads can be running concurrently, even if
445 * we do pass this check, another thread may queue some frames
446 * before this thread does and it will end up partially or fully
447 * failing to send anyway;
448 * + if_transmit() based drivers don't necessarily set ifq_len
451 * This should be replaced with a method of pushing an entire list
452 * of fragment frames to the driver and have the driver decide
453 * whether it can queue or not queue the entire set.
457 * Verify that we have any chance at all of being able to queue the
458 * packet or packet fragments, unless ALTQ is enabled on the given
459 * interface in which case packetdrop should be done by queueing.
461 n = ip_len / mtu + 1; /* how many fragments ? */
464 (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
466 (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
468 IPSTAT_INC(ips_odropped);
469 ifp->if_snd.ifq_drops += n;
475 * Look for broadcast address and
476 * verify user is allowed to send
480 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
481 error = EADDRNOTAVAIL;
484 if ((flags & IP_ALLOWBROADCAST) == 0) {
488 /* don't allow broadcast messages to be fragmented */
493 m->m_flags |= M_BCAST;
495 m->m_flags &= ~M_BCAST;
500 switch(ip_ipsec_output(&m, inp, &flags, &error)) {
507 break; /* Continue with packet processing. */
510 * Check if there was a route for this packet; return error if not.
512 if (no_route_but_check_spd) {
513 IPSTAT_INC(ips_noroute);
514 error = EHOSTUNREACH;
517 /* Update variables that are affected by ipsec4_output(). */
518 ip = mtod(m, struct ip *);
519 hlen = ip->ip_hl << 2;
522 /* Jump over all PFIL processing if hooks are not active. */
523 if (!PFIL_HOOKED(&V_inet_pfil_hook))
526 /* Run through list of hooks for output packets. */
527 odst.s_addr = ip->ip_dst.s_addr;
528 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
529 if (error != 0 || m == NULL)
532 ip = mtod(m, struct ip *);
534 /* See if destination IP address was changed by packet filter. */
535 if (odst.s_addr != ip->ip_dst.s_addr) {
536 m->m_flags |= M_SKIP_FIREWALL;
537 /* If destination is now ourself drop to ip_input(). */
538 if (in_localip(ip->ip_dst)) {
539 m->m_flags |= M_FASTFWD_OURS;
540 if (m->m_pkthdr.rcvif == NULL)
541 m->m_pkthdr.rcvif = V_loif;
542 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
543 m->m_pkthdr.csum_flags |=
544 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
545 m->m_pkthdr.csum_data = 0xffff;
547 m->m_pkthdr.csum_flags |=
548 CSUM_IP_CHECKED | CSUM_IP_VALID;
550 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
551 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
553 error = netisr_queue(NETISR_IP, m);
556 goto again; /* Redo the routing table lookup. */
559 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
560 if (m->m_flags & M_FASTFWD_OURS) {
561 if (m->m_pkthdr.rcvif == NULL)
562 m->m_pkthdr.rcvif = V_loif;
563 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
564 m->m_pkthdr.csum_flags |=
565 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
566 m->m_pkthdr.csum_data = 0xffff;
569 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
570 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
572 m->m_pkthdr.csum_flags |=
573 CSUM_IP_CHECKED | CSUM_IP_VALID;
575 error = netisr_queue(NETISR_IP, m);
578 /* Or forward to some other address? */
579 if ((m->m_flags & M_IP_NEXTHOP) &&
580 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
581 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
582 m->m_flags |= M_SKIP_FIREWALL;
583 m->m_flags &= ~M_IP_NEXTHOP;
584 m_tag_delete(m, fwd_tag);
589 /* 127/8 must not appear on wire - RFC1122. */
590 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
591 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
592 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
593 IPSTAT_INC(ips_badaddr);
594 error = EADDRNOTAVAIL;
599 m->m_pkthdr.csum_flags |= CSUM_IP;
600 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
602 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
605 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
606 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
607 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
612 * If small enough for interface, or the interface will take
613 * care of the fragmentation for us, we can just send directly.
616 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
617 ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
619 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
620 ip->ip_sum = in_cksum(m, hlen);
621 m->m_pkthdr.csum_flags &= ~CSUM_IP;
625 * Record statistics for this interface address.
626 * With CSUM_TSO the byte/packet count will be slightly
627 * incorrect because we count the IP+TCP headers only
628 * once instead of for every generated packet.
630 if (!(flags & IP_FORWARDING) && ia) {
631 if (m->m_pkthdr.csum_flags & CSUM_TSO)
632 counter_u64_add(ia->ia_ifa.ifa_opackets,
633 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
635 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
637 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
639 #ifdef MBUF_STRESS_TEST
640 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
641 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
644 * Reset layer specific mbuf flags
645 * to avoid confusing lower layers.
648 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
649 error = (*ifp->if_output)(ifp, m,
650 (const struct sockaddr *)gw, ro);
654 /* Balk when DF bit is set or the interface didn't support TSO. */
655 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
657 IPSTAT_INC(ips_cantfrag);
662 * Too large for interface; fragment if possible. If successful,
663 * on return, m will point to a list of packets to be sent.
665 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
672 /* Record statistics for this interface address. */
674 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
675 counter_u64_add(ia->ia_ifa.ifa_obytes,
679 * Reset layer specific mbuf flags
680 * to avoid confusing upper layers.
684 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
685 error = (*ifp->if_output)(ifp, m,
686 (const struct sockaddr *)gw, ro);
692 IPSTAT_INC(ips_fragmented);
704 * Create a chain of fragments which fit the given mtu. m_frag points to the
705 * mbuf to be fragmented; on return it points to the chain with the fragments.
706 * Return 0 if no error. If error, m_frag may contain a partially built
707 * chain of fragments that should be freed by the caller.
709 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
712 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
713 u_long if_hwassist_flags)
716 int hlen = ip->ip_hl << 2;
717 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
719 struct mbuf *m0 = *m_frag; /* the original packet */
723 uint16_t ip_len, ip_off;
725 ip_len = ntohs(ip->ip_len);
726 ip_off = ntohs(ip->ip_off);
728 if (ip_off & IP_DF) { /* Fragmentation not allowed */
729 IPSTAT_INC(ips_cantfrag);
734 * Must be able to put at least 8 bytes per fragment.
740 * If the interface will not calculate checksums on
741 * fragmented packets, then do it here.
743 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
744 in_delayed_cksum(m0);
745 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
748 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
749 sctp_delayed_cksum(m0, hlen);
750 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
753 if (len > PAGE_SIZE) {
755 * Fragment large datagrams such that each segment
756 * contains a multiple of PAGE_SIZE amount of data,
757 * plus headers. This enables a receiver to perform
758 * page-flipping zero-copy optimizations.
760 * XXX When does this help given that sender and receiver
761 * could have different page sizes, and also mtu could
762 * be less than the receiver's page size ?
767 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
771 * firstlen (off - hlen) must be aligned on an
775 goto smart_frag_failure;
776 off = ((off - hlen) & ~7) + hlen;
777 newlen = (~PAGE_MASK) & mtu;
778 if ((newlen + sizeof (struct ip)) > mtu) {
779 /* we failed, go back the default */
790 firstlen = off - hlen;
791 mnext = &m0->m_nextpkt; /* pointer to next packet */
794 * Loop through length of segment after first fragment,
795 * make new header and copy data of each part and link onto chain.
796 * Here, m0 is the original packet, m is the fragment being created.
797 * The fragments are linked off the m_nextpkt of the original
798 * packet, which after processing serves as the first fragment.
800 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
801 struct ip *mhip; /* ip header on the fragment */
803 int mhlen = sizeof (struct ip);
805 m = m_gethdr(M_NOWAIT, MT_DATA);
808 IPSTAT_INC(ips_odropped);
811 m->m_flags |= (m0->m_flags & M_MCAST);
813 * In the first mbuf, leave room for the link header, then
814 * copy the original IP header including options. The payload
815 * goes into an additional mbuf chain returned by m_copym().
817 m->m_data += max_linkhdr;
818 mhip = mtod(m, struct ip *);
820 if (hlen > sizeof (struct ip)) {
821 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
822 mhip->ip_v = IPVERSION;
823 mhip->ip_hl = mhlen >> 2;
826 /* XXX do we need to add ip_off below ? */
827 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
828 if (off + len >= ip_len)
831 mhip->ip_off |= IP_MF;
832 mhip->ip_len = htons((u_short)(len + mhlen));
833 m->m_next = m_copym(m0, off, len, M_NOWAIT);
834 if (m->m_next == NULL) { /* copy failed */
836 error = ENOBUFS; /* ??? */
837 IPSTAT_INC(ips_odropped);
840 m->m_pkthdr.len = mhlen + len;
841 m->m_pkthdr.rcvif = NULL;
843 mac_netinet_fragment(m0, m);
845 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
846 mhip->ip_off = htons(mhip->ip_off);
848 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
849 mhip->ip_sum = in_cksum(m, mhlen);
850 m->m_pkthdr.csum_flags &= ~CSUM_IP;
853 mnext = &m->m_nextpkt;
855 IPSTAT_ADD(ips_ofragments, nfrags);
858 * Update first fragment by trimming what's been copied out
859 * and updating header.
861 m_adj(m0, hlen + firstlen - ip_len);
862 m0->m_pkthdr.len = hlen + firstlen;
863 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
864 ip->ip_off = htons(ip_off | IP_MF);
866 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
867 ip->ip_sum = in_cksum(m0, hlen);
868 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
877 in_delayed_cksum(struct mbuf *m)
880 uint16_t csum, offset, ip_len;
882 ip = mtod(m, struct ip *);
883 offset = ip->ip_hl << 2 ;
884 ip_len = ntohs(ip->ip_len);
885 csum = in_cksum_skip(m, ip_len, offset);
886 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
888 offset += m->m_pkthdr.csum_data; /* checksum offset */
890 /* find the mbuf in the chain where the checksum starts*/
891 while ((m != NULL) && (offset >= m->m_len)) {
895 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
896 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
897 *(u_short *)(m->m_data + offset) = csum;
901 * IP socket option processing.
904 ip_ctloutput(struct socket *so, struct sockopt *sopt)
906 struct inpcb *inp = sotoinpcb(so);
910 if (sopt->sopt_level != IPPROTO_IP) {
913 if (sopt->sopt_level == SOL_SOCKET &&
914 sopt->sopt_dir == SOPT_SET) {
915 switch (sopt->sopt_name) {
918 if ((so->so_options & SO_REUSEADDR) != 0)
919 inp->inp_flags2 |= INP_REUSEADDR;
921 inp->inp_flags2 &= ~INP_REUSEADDR;
927 if ((so->so_options & SO_REUSEPORT) != 0)
928 inp->inp_flags2 |= INP_REUSEPORT;
930 inp->inp_flags2 &= ~INP_REUSEPORT;
936 inp->inp_inc.inc_fibnum = so->so_fibnum;
947 switch (sopt->sopt_dir) {
949 switch (sopt->sopt_name) {
956 if (sopt->sopt_valsize > MLEN) {
960 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
965 m->m_len = sopt->sopt_valsize;
966 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
973 error = ip_pcbopts(inp, sopt->sopt_name, m);
979 if (sopt->sopt_td != NULL) {
980 error = priv_check(sopt->sopt_td,
981 PRIV_NETINET_BINDANY);
998 error = sooptcopyin(sopt, &optval, sizeof optval,
1003 switch (sopt->sopt_name) {
1005 inp->inp_ip_tos = optval;
1009 inp->inp_ip_ttl = optval;
1013 if (optval >= 0 && optval <= MAXTTL)
1014 inp->inp_ip_minttl = optval;
1019 #define OPTSET(bit) do { \
1022 inp->inp_flags |= bit; \
1024 inp->inp_flags &= ~bit; \
1029 OPTSET(INP_RECVOPTS);
1032 case IP_RECVRETOPTS:
1033 OPTSET(INP_RECVRETOPTS);
1036 case IP_RECVDSTADDR:
1037 OPTSET(INP_RECVDSTADDR);
1041 OPTSET(INP_RECVTTL);
1053 OPTSET(INP_ONESBCAST);
1056 OPTSET(INP_DONTFRAG);
1059 OPTSET(INP_BINDANY);
1062 OPTSET(INP_RECVTOS);
1069 * Multicast socket options are processed by the in_mcast
1072 case IP_MULTICAST_IF:
1073 case IP_MULTICAST_VIF:
1074 case IP_MULTICAST_TTL:
1075 case IP_MULTICAST_LOOP:
1076 case IP_ADD_MEMBERSHIP:
1077 case IP_DROP_MEMBERSHIP:
1078 case IP_ADD_SOURCE_MEMBERSHIP:
1079 case IP_DROP_SOURCE_MEMBERSHIP:
1080 case IP_BLOCK_SOURCE:
1081 case IP_UNBLOCK_SOURCE:
1083 case MCAST_JOIN_GROUP:
1084 case MCAST_LEAVE_GROUP:
1085 case MCAST_JOIN_SOURCE_GROUP:
1086 case MCAST_LEAVE_SOURCE_GROUP:
1087 case MCAST_BLOCK_SOURCE:
1088 case MCAST_UNBLOCK_SOURCE:
1089 error = inp_setmoptions(inp, sopt);
1093 error = sooptcopyin(sopt, &optval, sizeof optval,
1100 case IP_PORTRANGE_DEFAULT:
1101 inp->inp_flags &= ~(INP_LOWPORT);
1102 inp->inp_flags &= ~(INP_HIGHPORT);
1105 case IP_PORTRANGE_HIGH:
1106 inp->inp_flags &= ~(INP_LOWPORT);
1107 inp->inp_flags |= INP_HIGHPORT;
1110 case IP_PORTRANGE_LOW:
1111 inp->inp_flags &= ~(INP_HIGHPORT);
1112 inp->inp_flags |= INP_LOWPORT;
1123 case IP_IPSEC_POLICY:
1128 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1130 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1132 req = mtod(m, caddr_t);
1133 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1134 m->m_len, (sopt->sopt_td != NULL) ?
1135 sopt->sopt_td->td_ucred : NULL);
1142 error = ENOPROTOOPT;
1148 switch (sopt->sopt_name) {
1151 if (inp->inp_options)
1152 error = sooptcopyout(sopt,
1153 mtod(inp->inp_options,
1155 inp->inp_options->m_len);
1157 sopt->sopt_valsize = 0;
1164 case IP_RECVRETOPTS:
1165 case IP_RECVDSTADDR:
1179 switch (sopt->sopt_name) {
1182 optval = inp->inp_ip_tos;
1186 optval = inp->inp_ip_ttl;
1190 optval = inp->inp_ip_minttl;
1193 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1196 optval = OPTBIT(INP_RECVOPTS);
1199 case IP_RECVRETOPTS:
1200 optval = OPTBIT(INP_RECVRETOPTS);
1203 case IP_RECVDSTADDR:
1204 optval = OPTBIT(INP_RECVDSTADDR);
1208 optval = OPTBIT(INP_RECVTTL);
1212 optval = OPTBIT(INP_RECVIF);
1216 if (inp->inp_flags & INP_HIGHPORT)
1217 optval = IP_PORTRANGE_HIGH;
1218 else if (inp->inp_flags & INP_LOWPORT)
1219 optval = IP_PORTRANGE_LOW;
1225 optval = OPTBIT(INP_FAITH);
1229 optval = OPTBIT(INP_ONESBCAST);
1232 optval = OPTBIT(INP_DONTFRAG);
1235 optval = OPTBIT(INP_BINDANY);
1238 optval = OPTBIT(INP_RECVTOS);
1241 optval = inp->inp_flowid;
1244 optval = inp->inp_flowtype;
1247 case IP_RSSBUCKETID:
1248 retval = rss_hash2bucket(inp->inp_flowid,
1252 optval = rss_bucket;
1258 error = sooptcopyout(sopt, &optval, sizeof optval);
1262 * Multicast socket options are processed by the in_mcast
1265 case IP_MULTICAST_IF:
1266 case IP_MULTICAST_VIF:
1267 case IP_MULTICAST_TTL:
1268 case IP_MULTICAST_LOOP:
1270 error = inp_getmoptions(inp, sopt);
1274 case IP_IPSEC_POLICY:
1276 struct mbuf *m = NULL;
1281 req = mtod(m, caddr_t);
1284 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1286 error = soopt_mcopyout(sopt, m); /* XXX */
1294 error = ENOPROTOOPT;
1303 * Routine called from ip_output() to loop back a copy of an IP multicast
1304 * packet to the input queue of a specified interface. Note that this
1305 * calls the output routine of the loopback "driver", but with an interface
1306 * pointer that might NOT be a loopback interface -- evil, but easier than
1307 * replicating that code here.
1310 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1313 register struct ip *ip;
1317 * Make a deep copy of the packet because we're going to
1318 * modify the pack in order to generate checksums.
1320 copym = m_dup(m, M_NOWAIT);
1321 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1322 copym = m_pullup(copym, hlen);
1323 if (copym != NULL) {
1324 /* If needed, compute the checksum and mark it as valid. */
1325 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1326 in_delayed_cksum(copym);
1327 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1328 copym->m_pkthdr.csum_flags |=
1329 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1330 copym->m_pkthdr.csum_data = 0xffff;
1333 * We don't bother to fragment if the IP length is greater
1334 * than the interface's MTU. Can this possibly matter?
1336 ip = mtod(copym, struct ip *);
1338 ip->ip_sum = in_cksum(copym, hlen);
1340 if (dst->sin_family != AF_INET) {
1341 printf("ip_mloopback: bad address family %d\n",
1343 dst->sin_family = AF_INET;
1346 if_simloop(ifp, copym, dst->sin_family, 0);