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>
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));
251 if (ro->ro_rt == NULL)
252 (void )flowtable_lookup(AF_INET, m, ro);
257 m = ip_insertoptions(m, opt, &len);
259 hlen = len; /* ip->ip_hl is updated above */
261 ip = mtod(m, struct ip *);
262 ip_len = ntohs(ip->ip_len);
263 ip_off = ntohs(ip->ip_off);
265 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
266 ip->ip_v = IPVERSION;
267 ip->ip_hl = hlen >> 2;
269 IPSTAT_INC(ips_localout);
271 /* Header already set, fetch hlen from there */
272 hlen = ip->ip_hl << 2;
278 * dst can be rewritten but always points to &ro->ro_dst.
279 * gw is readonly but can point either to dst OR rt_gateway,
280 * therefore we need restore gw if we're redoing lookup.
282 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
283 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
286 * The address family should also be checked in case of sharing
287 * the cache with IPv6.
289 if (rte == NULL || dst->sin_family != AF_INET) {
290 bzero(dst, sizeof(*dst));
291 dst->sin_family = AF_INET;
292 dst->sin_len = sizeof(*dst);
293 dst->sin_addr = ip->ip_dst;
299 * If routing to interface only, short circuit routing lookup.
300 * The use of an all-ones broadcast address implies this; an
301 * interface is specified by the broadcast address of an interface,
302 * or the destination address of a ptp interface.
304 if (flags & IP_SENDONES) {
305 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
306 M_GETFIB(m)))) == NULL &&
307 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
308 M_GETFIB(m)))) == NULL) {
309 IPSTAT_INC(ips_noroute);
314 ip->ip_dst.s_addr = INADDR_BROADCAST;
315 dst->sin_addr = ip->ip_dst;
319 } else if (flags & IP_ROUTETOIF) {
320 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
321 M_GETFIB(m)))) == NULL &&
322 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
323 M_GETFIB(m)))) == NULL) {
324 IPSTAT_INC(ips_noroute);
331 isbroadcast = in_broadcast(dst->sin_addr, ifp);
332 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
333 imo != NULL && imo->imo_multicast_ifp != NULL) {
335 * Bypass the normal routing lookup for multicast
336 * packets if the interface is specified.
338 ifp = imo->imo_multicast_ifp;
339 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
342 isbroadcast = 0; /* fool gcc */
345 * We want to do any cloning requested by the link layer,
346 * as this is probably required in all cases for correct
347 * operation (as it is for ARP).
351 rtalloc_mpath_fib(ro,
352 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
355 in_rtalloc_ign(ro, 0, fibnum);
360 (rte->rt_flags & RTF_UP) == 0 ||
361 rte->rt_ifp == NULL ||
362 !RT_LINK_IS_UP(rte->rt_ifp)) {
365 * There is no route for this packet, but it is
366 * possible that a matching SPD entry exists.
368 no_route_but_check_spd = 1;
369 mtu = 0; /* Silence GCC warning. */
372 IPSTAT_INC(ips_noroute);
373 error = EHOSTUNREACH;
376 ia = ifatoia(rte->rt_ifa);
378 counter_u64_add(rte->rt_pksent, 1);
379 rt_update_ro_flags(ro);
380 if (rte->rt_flags & RTF_GATEWAY)
381 gw = (struct sockaddr_in *)rte->rt_gateway;
382 if (rte->rt_flags & RTF_HOST)
383 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
385 isbroadcast = in_broadcast(gw->sin_addr, ifp);
389 * Calculate MTU. If we have a route that is up, use that,
390 * otherwise use the interface's MTU.
392 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
396 /* Catch a possible divide by zero later. */
397 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
398 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
400 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
401 m->m_flags |= M_MCAST;
403 * IP destination address is multicast. Make sure "gw"
404 * still points to the address in "ro". (It may have been
405 * changed to point to a gateway address, above.)
409 * See if the caller provided any multicast options
412 ip->ip_ttl = imo->imo_multicast_ttl;
413 if (imo->imo_multicast_vif != -1)
416 ip_mcast_src(imo->imo_multicast_vif) :
419 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
421 * Confirm that the outgoing interface supports multicast.
423 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
424 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
425 IPSTAT_INC(ips_noroute);
431 * If source address not specified yet, use address
432 * of outgoing interface.
434 if (ip->ip_src.s_addr == INADDR_ANY) {
435 /* Interface may have no addresses. */
437 ip->ip_src = IA_SIN(ia)->sin_addr;
440 if ((imo == NULL && in_mcast_loop) ||
441 (imo && imo->imo_multicast_loop)) {
443 * Loop back multicast datagram if not expressly
444 * forbidden to do so, even if we are not a member
445 * of the group; ip_input() will filter it later,
446 * thus deferring a hash lookup and mutex acquisition
447 * at the expense of a cheap copy using m_copym().
449 ip_mloopback(ifp, m, hlen);
452 * If we are acting as a multicast router, perform
453 * multicast forwarding as if the packet had just
454 * arrived on the interface to which we are about
455 * to send. The multicast forwarding function
456 * recursively calls this function, using the
457 * IP_FORWARDING flag to prevent infinite recursion.
459 * Multicasts that are looped back by ip_mloopback(),
460 * above, will be forwarded by the ip_input() routine,
463 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
465 * If rsvp daemon is not running, do not
466 * set ip_moptions. This ensures that the packet
467 * is multicast and not just sent down one link
468 * as prescribed by rsvpd.
473 ip_mforward(ip, ifp, m, imo) != 0) {
481 * Multicasts with a time-to-live of zero may be looped-
482 * back, above, but must not be transmitted on a network.
483 * Also, multicasts addressed to the loopback interface
484 * are not sent -- the above call to ip_mloopback() will
485 * loop back a copy. ip_input() will drop the copy if
486 * this host does not belong to the destination group on
487 * the loopback interface.
489 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
498 * If the source address is not specified yet, use the address
499 * of the outoing interface.
501 if (ip->ip_src.s_addr == INADDR_ANY) {
502 /* Interface may have no addresses. */
504 ip->ip_src = IA_SIN(ia)->sin_addr;
509 * Look for broadcast address and
510 * verify user is allowed to send
514 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
515 error = EADDRNOTAVAIL;
518 if ((flags & IP_ALLOWBROADCAST) == 0) {
522 /* don't allow broadcast messages to be fragmented */
527 m->m_flags |= M_BCAST;
529 m->m_flags &= ~M_BCAST;
534 switch(ip_ipsec_output(&m, inp, &error)) {
541 break; /* Continue with packet processing. */
544 * Check if there was a route for this packet; return error if not.
546 if (no_route_but_check_spd) {
547 IPSTAT_INC(ips_noroute);
548 error = EHOSTUNREACH;
551 /* Update variables that are affected by ipsec4_output(). */
552 ip = mtod(m, struct ip *);
553 hlen = ip->ip_hl << 2;
556 /* Jump over all PFIL processing if hooks are not active. */
557 if (PFIL_HOOKED(&V_inet_pfil_hook)) {
558 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
559 case 1: /* Finished */
562 case 0: /* Continue normally */
563 ip = mtod(m, struct ip *);
566 case -1: /* Need to try again */
567 /* Reset everything for a new round */
570 ifa_free(&ia->ia_ifa);
571 ro->ro_prepend = NULL;
574 ip = mtod(m, struct ip *);
580 /* 127/8 must not appear on wire - RFC1122. */
581 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
582 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
583 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
584 IPSTAT_INC(ips_badaddr);
585 error = EADDRNOTAVAIL;
590 m->m_pkthdr.csum_flags |= CSUM_IP;
591 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
593 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
596 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
597 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
598 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
603 * If small enough for interface, or the interface will take
604 * care of the fragmentation for us, we can just send directly.
607 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
609 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
610 ip->ip_sum = in_cksum(m, hlen);
611 m->m_pkthdr.csum_flags &= ~CSUM_IP;
615 * Record statistics for this interface address.
616 * With CSUM_TSO the byte/packet count will be slightly
617 * incorrect because we count the IP+TCP headers only
618 * once instead of for every generated packet.
620 if (!(flags & IP_FORWARDING) && ia) {
621 if (m->m_pkthdr.csum_flags & CSUM_TSO)
622 counter_u64_add(ia->ia_ifa.ifa_opackets,
623 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
625 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
627 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
629 #ifdef MBUF_STRESS_TEST
630 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
631 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
634 * Reset layer specific mbuf flags
635 * to avoid confusing lower layers.
638 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
639 error = (*ifp->if_output)(ifp, m,
640 (const struct sockaddr *)gw, ro);
644 /* Balk when DF bit is set or the interface didn't support TSO. */
645 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
647 IPSTAT_INC(ips_cantfrag);
652 * Too large for interface; fragment if possible. If successful,
653 * on return, m will point to a list of packets to be sent.
655 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
662 /* Record statistics for this interface address. */
664 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
665 counter_u64_add(ia->ia_ifa.ifa_obytes,
669 * Reset layer specific mbuf flags
670 * to avoid confusing upper layers.
674 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
675 error = (*ifp->if_output)(ifp, m,
676 (const struct sockaddr *)gw, ro);
682 IPSTAT_INC(ips_fragmented);
687 else if (rte == NULL)
689 * If the caller supplied a route but somehow the reference
690 * to it has been released need to prevent the caller
691 * calling RTFREE on it again.
695 ifa_free(&ia->ia_ifa);
703 * Create a chain of fragments which fit the given mtu. m_frag points to the
704 * mbuf to be fragmented; on return it points to the chain with the fragments.
705 * Return 0 if no error. If error, m_frag may contain a partially built
706 * chain of fragments that should be freed by the caller.
708 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
711 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
712 u_long if_hwassist_flags)
715 int hlen = ip->ip_hl << 2;
716 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
718 struct mbuf *m0 = *m_frag; /* the original packet */
722 uint16_t ip_len, ip_off;
724 ip_len = ntohs(ip->ip_len);
725 ip_off = ntohs(ip->ip_off);
727 if (ip_off & IP_DF) { /* Fragmentation not allowed */
728 IPSTAT_INC(ips_cantfrag);
733 * Must be able to put at least 8 bytes per fragment.
739 * If the interface will not calculate checksums on
740 * fragmented packets, then do it here.
742 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
743 in_delayed_cksum(m0);
744 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
747 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
748 sctp_delayed_cksum(m0, hlen);
749 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
752 if (len > PAGE_SIZE) {
754 * Fragment large datagrams such that each segment
755 * contains a multiple of PAGE_SIZE amount of data,
756 * plus headers. This enables a receiver to perform
757 * page-flipping zero-copy optimizations.
759 * XXX When does this help given that sender and receiver
760 * could have different page sizes, and also mtu could
761 * be less than the receiver's page size ?
765 off = MIN(mtu, m0->m_pkthdr.len);
768 * firstlen (off - hlen) must be aligned on an
772 goto smart_frag_failure;
773 off = ((off - hlen) & ~7) + hlen;
774 newlen = (~PAGE_MASK) & mtu;
775 if ((newlen + sizeof (struct ip)) > mtu) {
776 /* we failed, go back the default */
787 firstlen = off - hlen;
788 mnext = &m0->m_nextpkt; /* pointer to next packet */
791 * Loop through length of segment after first fragment,
792 * make new header and copy data of each part and link onto chain.
793 * Here, m0 is the original packet, m is the fragment being created.
794 * The fragments are linked off the m_nextpkt of the original
795 * packet, which after processing serves as the first fragment.
797 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
798 struct ip *mhip; /* ip header on the fragment */
800 int mhlen = sizeof (struct ip);
802 m = m_gethdr(M_NOWAIT, MT_DATA);
805 IPSTAT_INC(ips_odropped);
809 * Make sure the complete packet header gets copied
810 * from the originating mbuf to the newly created
811 * mbuf. This also ensures that existing firewall
812 * classification(s), VLAN tags and so on get copied
813 * to the resulting fragmented packet(s):
815 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
818 IPSTAT_INC(ips_odropped);
822 * In the first mbuf, leave room for the link header, then
823 * copy the original IP header including options. The payload
824 * goes into an additional mbuf chain returned by m_copym().
826 m->m_data += max_linkhdr;
827 mhip = mtod(m, struct ip *);
829 if (hlen > sizeof (struct ip)) {
830 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
831 mhip->ip_v = IPVERSION;
832 mhip->ip_hl = mhlen >> 2;
835 /* XXX do we need to add ip_off below ? */
836 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
837 if (off + len >= ip_len)
840 mhip->ip_off |= IP_MF;
841 mhip->ip_len = htons((u_short)(len + mhlen));
842 m->m_next = m_copym(m0, off, len, M_NOWAIT);
843 if (m->m_next == NULL) { /* copy failed */
845 error = ENOBUFS; /* ??? */
846 IPSTAT_INC(ips_odropped);
849 m->m_pkthdr.len = mhlen + len;
851 mac_netinet_fragment(m0, m);
853 mhip->ip_off = htons(mhip->ip_off);
855 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
856 mhip->ip_sum = in_cksum(m, mhlen);
857 m->m_pkthdr.csum_flags &= ~CSUM_IP;
860 mnext = &m->m_nextpkt;
862 IPSTAT_ADD(ips_ofragments, nfrags);
865 * Update first fragment by trimming what's been copied out
866 * and updating header.
868 m_adj(m0, hlen + firstlen - ip_len);
869 m0->m_pkthdr.len = hlen + firstlen;
870 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
871 ip->ip_off = htons(ip_off | IP_MF);
873 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
874 ip->ip_sum = in_cksum(m0, hlen);
875 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
884 in_delayed_cksum(struct mbuf *m)
887 uint16_t csum, offset, ip_len;
889 ip = mtod(m, struct ip *);
890 offset = ip->ip_hl << 2 ;
891 ip_len = ntohs(ip->ip_len);
892 csum = in_cksum_skip(m, ip_len, offset);
893 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
895 offset += m->m_pkthdr.csum_data; /* checksum offset */
897 /* find the mbuf in the chain where the checksum starts*/
898 while ((m != NULL) && (offset >= m->m_len)) {
902 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
903 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
904 *(u_short *)(m->m_data + offset) = csum;
908 * IP socket option processing.
911 ip_ctloutput(struct socket *so, struct sockopt *sopt)
913 struct inpcb *inp = sotoinpcb(so);
921 if (sopt->sopt_level != IPPROTO_IP) {
924 if (sopt->sopt_level == SOL_SOCKET &&
925 sopt->sopt_dir == SOPT_SET) {
926 switch (sopt->sopt_name) {
929 if ((so->so_options & SO_REUSEADDR) != 0)
930 inp->inp_flags2 |= INP_REUSEADDR;
932 inp->inp_flags2 &= ~INP_REUSEADDR;
938 if ((so->so_options & SO_REUSEPORT) != 0)
939 inp->inp_flags2 |= INP_REUSEPORT;
941 inp->inp_flags2 &= ~INP_REUSEPORT;
947 inp->inp_inc.inc_fibnum = so->so_fibnum;
958 switch (sopt->sopt_dir) {
960 switch (sopt->sopt_name) {
967 if (sopt->sopt_valsize > MLEN) {
971 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
976 m->m_len = sopt->sopt_valsize;
977 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
984 error = ip_pcbopts(inp, sopt->sopt_name, m);
990 if (sopt->sopt_td != NULL) {
991 error = priv_check(sopt->sopt_td,
992 PRIV_NETINET_BINDANY);
999 case IP_RSS_LISTEN_BUCKET:
1005 case IP_RECVRETOPTS:
1006 case IP_RECVDSTADDR:
1014 case IP_RECVRSSBUCKETID:
1016 error = sooptcopyin(sopt, &optval, sizeof optval,
1021 switch (sopt->sopt_name) {
1023 inp->inp_ip_tos = optval;
1027 inp->inp_ip_ttl = optval;
1031 if (optval >= 0 && optval <= MAXTTL)
1032 inp->inp_ip_minttl = optval;
1037 #define OPTSET(bit) do { \
1040 inp->inp_flags |= bit; \
1042 inp->inp_flags &= ~bit; \
1046 #define OPTSET2(bit, val) do { \
1049 inp->inp_flags2 |= bit; \
1051 inp->inp_flags2 &= ~bit; \
1056 OPTSET(INP_RECVOPTS);
1059 case IP_RECVRETOPTS:
1060 OPTSET(INP_RECVRETOPTS);
1063 case IP_RECVDSTADDR:
1064 OPTSET(INP_RECVDSTADDR);
1068 OPTSET(INP_RECVTTL);
1076 OPTSET(INP_ONESBCAST);
1079 OPTSET(INP_DONTFRAG);
1082 OPTSET(INP_BINDANY);
1085 OPTSET(INP_RECVTOS);
1088 OPTSET2(INP_BINDMULTI, optval);
1091 OPTSET2(INP_RECVFLOWID, optval);
1094 case IP_RSS_LISTEN_BUCKET:
1095 if ((optval >= 0) &&
1096 (optval < rss_getnumbuckets())) {
1097 inp->inp_rss_listen_bucket = optval;
1098 OPTSET2(INP_RSS_BUCKET_SET, 1);
1103 case IP_RECVRSSBUCKETID:
1104 OPTSET2(INP_RECVRSSBUCKETID, optval);
1113 * Multicast socket options are processed by the in_mcast
1116 case IP_MULTICAST_IF:
1117 case IP_MULTICAST_VIF:
1118 case IP_MULTICAST_TTL:
1119 case IP_MULTICAST_LOOP:
1120 case IP_ADD_MEMBERSHIP:
1121 case IP_DROP_MEMBERSHIP:
1122 case IP_ADD_SOURCE_MEMBERSHIP:
1123 case IP_DROP_SOURCE_MEMBERSHIP:
1124 case IP_BLOCK_SOURCE:
1125 case IP_UNBLOCK_SOURCE:
1127 case MCAST_JOIN_GROUP:
1128 case MCAST_LEAVE_GROUP:
1129 case MCAST_JOIN_SOURCE_GROUP:
1130 case MCAST_LEAVE_SOURCE_GROUP:
1131 case MCAST_BLOCK_SOURCE:
1132 case MCAST_UNBLOCK_SOURCE:
1133 error = inp_setmoptions(inp, sopt);
1137 error = sooptcopyin(sopt, &optval, sizeof optval,
1144 case IP_PORTRANGE_DEFAULT:
1145 inp->inp_flags &= ~(INP_LOWPORT);
1146 inp->inp_flags &= ~(INP_HIGHPORT);
1149 case IP_PORTRANGE_HIGH:
1150 inp->inp_flags &= ~(INP_LOWPORT);
1151 inp->inp_flags |= INP_HIGHPORT;
1154 case IP_PORTRANGE_LOW:
1155 inp->inp_flags &= ~(INP_HIGHPORT);
1156 inp->inp_flags |= INP_LOWPORT;
1167 case IP_IPSEC_POLICY:
1172 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1174 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1176 req = mtod(m, caddr_t);
1177 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1178 m->m_len, (sopt->sopt_td != NULL) ?
1179 sopt->sopt_td->td_ucred : NULL);
1186 error = ENOPROTOOPT;
1192 switch (sopt->sopt_name) {
1195 if (inp->inp_options)
1196 error = sooptcopyout(sopt,
1197 mtod(inp->inp_options,
1199 inp->inp_options->m_len);
1201 sopt->sopt_valsize = 0;
1208 case IP_RECVRETOPTS:
1209 case IP_RECVDSTADDR:
1222 case IP_RSSBUCKETID:
1223 case IP_RECVRSSBUCKETID:
1225 switch (sopt->sopt_name) {
1228 optval = inp->inp_ip_tos;
1232 optval = inp->inp_ip_ttl;
1236 optval = inp->inp_ip_minttl;
1239 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1240 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1243 optval = OPTBIT(INP_RECVOPTS);
1246 case IP_RECVRETOPTS:
1247 optval = OPTBIT(INP_RECVRETOPTS);
1250 case IP_RECVDSTADDR:
1251 optval = OPTBIT(INP_RECVDSTADDR);
1255 optval = OPTBIT(INP_RECVTTL);
1259 optval = OPTBIT(INP_RECVIF);
1263 if (inp->inp_flags & INP_HIGHPORT)
1264 optval = IP_PORTRANGE_HIGH;
1265 else if (inp->inp_flags & INP_LOWPORT)
1266 optval = IP_PORTRANGE_LOW;
1272 optval = OPTBIT(INP_ONESBCAST);
1275 optval = OPTBIT(INP_DONTFRAG);
1278 optval = OPTBIT(INP_BINDANY);
1281 optval = OPTBIT(INP_RECVTOS);
1284 optval = inp->inp_flowid;
1287 optval = inp->inp_flowtype;
1290 optval = OPTBIT2(INP_RECVFLOWID);
1293 case IP_RSSBUCKETID:
1294 retval = rss_hash2bucket(inp->inp_flowid,
1298 optval = rss_bucket;
1302 case IP_RECVRSSBUCKETID:
1303 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1307 optval = OPTBIT2(INP_BINDMULTI);
1310 error = sooptcopyout(sopt, &optval, sizeof optval);
1314 * Multicast socket options are processed by the in_mcast
1317 case IP_MULTICAST_IF:
1318 case IP_MULTICAST_VIF:
1319 case IP_MULTICAST_TTL:
1320 case IP_MULTICAST_LOOP:
1322 error = inp_getmoptions(inp, sopt);
1326 case IP_IPSEC_POLICY:
1328 struct mbuf *m = NULL;
1333 req = mtod(m, caddr_t);
1336 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1338 error = soopt_mcopyout(sopt, m); /* XXX */
1346 error = ENOPROTOOPT;
1355 * Routine called from ip_output() to loop back a copy of an IP multicast
1356 * packet to the input queue of a specified interface. Note that this
1357 * calls the output routine of the loopback "driver", but with an interface
1358 * pointer that might NOT be a loopback interface -- evil, but easier than
1359 * replicating that code here.
1362 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1368 * Make a deep copy of the packet because we're going to
1369 * modify the pack in order to generate checksums.
1371 copym = m_dup(m, M_NOWAIT);
1372 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1373 copym = m_pullup(copym, hlen);
1374 if (copym != NULL) {
1375 /* If needed, compute the checksum and mark it as valid. */
1376 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1377 in_delayed_cksum(copym);
1378 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1379 copym->m_pkthdr.csum_flags |=
1380 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1381 copym->m_pkthdr.csum_data = 0xffff;
1384 * We don't bother to fragment if the IP length is greater
1385 * than the interface's MTU. Can this possibly matter?
1387 ip = mtod(copym, struct ip *);
1389 ip->ip_sum = in_cksum(copym, hlen);
1390 if_simloop(ifp, copym, AF_INET, 0);