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"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
46 #include <sys/malloc.h>
50 #include <sys/protosw.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/sysctl.h>
55 #include <sys/ucred.h>
58 #include <net/if_var.h>
59 #include <net/if_llatbl.h>
60 #include <net/netisr.h>
62 #include <net/route.h>
63 #include <net/flowtable.h>
65 #include <net/radix_mpath.h>
69 #include <netinet/in.h>
70 #include <netinet/in_kdtrace.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/ip.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/in_var.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/ip_options.h>
78 #include <netinet/sctp.h>
79 #include <netinet/sctp_crc32.h>
83 #include <netinet/ip_ipsec.h>
84 #include <netipsec/ipsec.h>
87 #include <machine/in_cksum.h>
89 #include <security/mac/mac_framework.h>
91 VNET_DEFINE(u_short, ip_id);
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
100 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
103 extern int in_mcast_loop;
104 extern struct protosw inetsw[];
107 * IP output. The packet in mbuf chain m contains a skeletal IP
108 * header (with len, off, ttl, proto, tos, src, dst).
109 * The mbuf chain containing the packet will be freed.
110 * The mbuf opt, if present, will not be freed.
111 * If route ro is present and has ro_rt initialized, route lookup would be
112 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
113 * then result of route lookup is stored in ro->ro_rt.
115 * In the IP forwarding case, the packet will arrive with options already
116 * inserted, so must have a NULL opt pointer.
119 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
120 struct ip_moptions *imo, struct inpcb *inp)
123 struct ifnet *ifp = NULL; /* keep compiler happy */
125 int hlen = sizeof (struct ip);
128 struct sockaddr_in *dst;
129 const struct sockaddr_in *gw;
130 struct in_ifaddr *ia;
132 uint16_t ip_len, ip_off;
133 struct route iproute;
134 struct rtentry *rte; /* cache for ro->ro_rt */
136 struct m_tag *fwd_tag = NULL;
138 int no_route_but_check_spd = 0;
143 INP_LOCK_ASSERT(inp);
144 M_SETFIB(m, inp->inp_inc.inc_fibnum);
145 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
146 m->m_pkthdr.flowid = inp->inp_flowid;
147 m->m_flags |= M_FLOWID;
153 bzero(ro, sizeof (*ro));
157 if (ro->ro_rt == NULL) {
161 * The flow table returns route entries valid for up to 30
162 * seconds; we rely on the remainder of ip_output() taking no
163 * longer than that long for the stability of ro_rt. The
164 * flow ID assignment must have happened before this point.
166 fle = flowtable_lookup(AF_INET, m);
168 flow_to_route(fle, ro);
174 m = ip_insertoptions(m, opt, &len);
176 hlen = len; /* ip->ip_hl is updated above */
178 ip = mtod(m, struct ip *);
179 ip_len = ntohs(ip->ip_len);
180 ip_off = ntohs(ip->ip_off);
183 * Fill in IP header. If we are not allowing fragmentation,
184 * then the ip_id field is meaningless, but we don't set it
185 * to zero. Doing so causes various problems when devices along
186 * the path (routers, load balancers, firewalls, etc.) illegally
187 * disable DF on our packet. Note that a 16-bit counter
188 * will wrap around in less than 10 seconds at 100 Mbit/s on a
189 * medium with MTU 1500. See Steven M. Bellovin, "A Technique
190 * for Counting NATted Hosts", Proc. IMW'02, available at
191 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
193 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
194 ip->ip_v = IPVERSION;
195 ip->ip_hl = hlen >> 2;
196 ip->ip_id = ip_newid();
197 IPSTAT_INC(ips_localout);
199 /* Header already set, fetch hlen from there */
200 hlen = ip->ip_hl << 2;
206 * dst can be rewritten but always points to &ro->ro_dst.
207 * gw is readonly but can point either to dst OR rt_gateway,
208 * therefore we need restore gw if we're redoing lookup.
210 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
214 * If there is a cached route, check that it is to the same
215 * destination and is still up. If not, free it and try again.
216 * The address family should also be checked in case of sharing
217 * the cache with IPv6.
220 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
221 rte->rt_ifp == NULL ||
222 !RT_LINK_IS_UP(rte->rt_ifp) ||
223 dst->sin_family != AF_INET ||
224 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
230 if (rte == NULL && fwd_tag == NULL) {
231 bzero(dst, sizeof(*dst));
232 dst->sin_family = AF_INET;
233 dst->sin_len = sizeof(*dst);
234 dst->sin_addr = ip->ip_dst;
237 * If routing to interface only, short circuit routing lookup.
238 * The use of an all-ones broadcast address implies this; an
239 * interface is specified by the broadcast address of an interface,
240 * or the destination address of a ptp interface.
242 if (flags & IP_SENDONES) {
243 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
244 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
245 IPSTAT_INC(ips_noroute);
249 ip->ip_dst.s_addr = INADDR_BROADCAST;
250 dst->sin_addr = ip->ip_dst;
254 } else if (flags & IP_ROUTETOIF) {
255 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
256 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) {
257 IPSTAT_INC(ips_noroute);
263 isbroadcast = in_broadcast(dst->sin_addr, ifp);
264 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
265 imo != NULL && imo->imo_multicast_ifp != NULL) {
267 * Bypass the normal routing lookup for multicast
268 * packets if the interface is specified.
270 ifp = imo->imo_multicast_ifp;
272 isbroadcast = 0; /* fool gcc */
275 * We want to do any cloning requested by the link layer,
276 * as this is probably required in all cases for correct
277 * operation (as it is for ARP).
281 rtalloc_mpath_fib(ro,
282 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
283 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
285 in_rtalloc_ign(ro, 0,
286 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
291 rte->rt_ifp == NULL ||
292 !RT_LINK_IS_UP(rte->rt_ifp)) {
295 * There is no route for this packet, but it is
296 * possible that a matching SPD entry exists.
298 no_route_but_check_spd = 1;
299 mtu = 0; /* Silence GCC warning. */
302 IPSTAT_INC(ips_noroute);
303 error = EHOSTUNREACH;
306 ia = ifatoia(rte->rt_ifa);
307 ifa_ref(&ia->ia_ifa);
309 rte->rt_rmx.rmx_pksent++;
310 if (rte->rt_flags & RTF_GATEWAY)
311 gw = (struct sockaddr_in *)rte->rt_gateway;
312 if (rte->rt_flags & RTF_HOST)
313 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
315 isbroadcast = in_broadcast(gw->sin_addr, ifp);
318 * Calculate MTU. If we have a route that is up, use that,
319 * otherwise use the interface's MTU.
321 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
323 * This case can happen if the user changed the MTU
324 * of an interface after enabling IP on it. Because
325 * most netifs don't keep track of routes pointing to
326 * them, there is no way for one to update all its
327 * routes when the MTU is changed.
329 if (rte->rt_rmx.rmx_mtu > ifp->if_mtu)
330 rte->rt_rmx.rmx_mtu = ifp->if_mtu;
331 mtu = rte->rt_rmx.rmx_mtu;
335 /* Catch a possible divide by zero later. */
336 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
337 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
338 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
339 m->m_flags |= M_MCAST;
341 * IP destination address is multicast. Make sure "gw"
342 * still points to the address in "ro". (It may have been
343 * changed to point to a gateway address, above.)
347 * See if the caller provided any multicast options
350 ip->ip_ttl = imo->imo_multicast_ttl;
351 if (imo->imo_multicast_vif != -1)
354 ip_mcast_src(imo->imo_multicast_vif) :
357 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
359 * Confirm that the outgoing interface supports multicast.
361 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
362 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
363 IPSTAT_INC(ips_noroute);
369 * If source address not specified yet, use address
370 * of outgoing interface.
372 if (ip->ip_src.s_addr == INADDR_ANY) {
373 /* Interface may have no addresses. */
375 ip->ip_src = IA_SIN(ia)->sin_addr;
378 if ((imo == NULL && in_mcast_loop) ||
379 (imo && imo->imo_multicast_loop)) {
381 * Loop back multicast datagram if not expressly
382 * forbidden to do so, even if we are not a member
383 * of the group; ip_input() will filter it later,
384 * thus deferring a hash lookup and mutex acquisition
385 * at the expense of a cheap copy using m_copym().
387 ip_mloopback(ifp, m, dst, hlen);
390 * If we are acting as a multicast router, perform
391 * multicast forwarding as if the packet had just
392 * arrived on the interface to which we are about
393 * to send. The multicast forwarding function
394 * recursively calls this function, using the
395 * IP_FORWARDING flag to prevent infinite recursion.
397 * Multicasts that are looped back by ip_mloopback(),
398 * above, will be forwarded by the ip_input() routine,
401 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
403 * If rsvp daemon is not running, do not
404 * set ip_moptions. This ensures that the packet
405 * is multicast and not just sent down one link
406 * as prescribed by rsvpd.
411 ip_mforward(ip, ifp, m, imo) != 0) {
419 * Multicasts with a time-to-live of zero may be looped-
420 * back, above, but must not be transmitted on a network.
421 * Also, multicasts addressed to the loopback interface
422 * are not sent -- the above call to ip_mloopback() will
423 * loop back a copy. ip_input() will drop the copy if
424 * this host does not belong to the destination group on
425 * the loopback interface.
427 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
436 * If the source address is not specified yet, use the address
437 * of the outoing interface.
439 if (ip->ip_src.s_addr == INADDR_ANY) {
440 /* Interface may have no addresses. */
442 ip->ip_src = IA_SIN(ia)->sin_addr;
447 * Both in the SMP world, pre-emption world if_transmit() world,
448 * the following code doesn't really function as intended any further.
450 * + There can and will be multiple CPUs running this code path
451 * in parallel, and we do no lock holding when checking the
453 * + And since other threads can be running concurrently, even if
454 * we do pass this check, another thread may queue some frames
455 * before this thread does and it will end up partially or fully
456 * failing to send anyway;
457 * + if_transmit() based drivers don't necessarily set ifq_len
460 * This should be replaced with a method of pushing an entire list
461 * of fragment frames to the driver and have the driver decide
462 * whether it can queue or not queue the entire set.
466 * Verify that we have any chance at all of being able to queue the
467 * packet or packet fragments, unless ALTQ is enabled on the given
468 * interface in which case packetdrop should be done by queueing.
470 n = ip_len / mtu + 1; /* how many fragments ? */
473 (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
475 (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
477 IPSTAT_INC(ips_odropped);
478 ifp->if_snd.ifq_drops += n;
484 * Look for broadcast address and
485 * verify user is allowed to send
489 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
490 error = EADDRNOTAVAIL;
493 if ((flags & IP_ALLOWBROADCAST) == 0) {
497 /* don't allow broadcast messages to be fragmented */
502 m->m_flags |= M_BCAST;
504 m->m_flags &= ~M_BCAST;
509 switch(ip_ipsec_output(&m, inp, &flags, &error)) {
516 break; /* Continue with packet processing. */
519 * Check if there was a route for this packet; return error if not.
521 if (no_route_but_check_spd) {
522 IPSTAT_INC(ips_noroute);
523 error = EHOSTUNREACH;
526 /* Update variables that are affected by ipsec4_output(). */
527 ip = mtod(m, struct ip *);
528 hlen = ip->ip_hl << 2;
531 /* Jump over all PFIL processing if hooks are not active. */
532 if (!PFIL_HOOKED(&V_inet_pfil_hook))
535 /* Run through list of hooks for output packets. */
536 odst.s_addr = ip->ip_dst.s_addr;
537 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
538 if (error != 0 || m == NULL)
541 ip = mtod(m, struct ip *);
543 /* See if destination IP address was changed by packet filter. */
544 if (odst.s_addr != ip->ip_dst.s_addr) {
545 m->m_flags |= M_SKIP_FIREWALL;
546 /* If destination is now ourself drop to ip_input(). */
547 if (in_localip(ip->ip_dst)) {
548 m->m_flags |= M_FASTFWD_OURS;
549 if (m->m_pkthdr.rcvif == NULL)
550 m->m_pkthdr.rcvif = V_loif;
551 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
552 m->m_pkthdr.csum_flags |=
553 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
554 m->m_pkthdr.csum_data = 0xffff;
556 m->m_pkthdr.csum_flags |=
557 CSUM_IP_CHECKED | CSUM_IP_VALID;
559 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
560 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
562 error = netisr_queue(NETISR_IP, m);
566 ifa_free(&ia->ia_ifa);
567 goto again; /* Redo the routing table lookup. */
571 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
572 if (m->m_flags & M_FASTFWD_OURS) {
573 if (m->m_pkthdr.rcvif == NULL)
574 m->m_pkthdr.rcvif = V_loif;
575 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
576 m->m_pkthdr.csum_flags |=
577 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
578 m->m_pkthdr.csum_data = 0xffff;
581 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
582 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
584 m->m_pkthdr.csum_flags |=
585 CSUM_IP_CHECKED | CSUM_IP_VALID;
587 error = netisr_queue(NETISR_IP, m);
590 /* Or forward to some other address? */
591 if ((m->m_flags & M_IP_NEXTHOP) &&
592 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
593 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
594 m->m_flags |= M_SKIP_FIREWALL;
595 m->m_flags &= ~M_IP_NEXTHOP;
596 m_tag_delete(m, fwd_tag);
598 ifa_free(&ia->ia_ifa);
603 /* 127/8 must not appear on wire - RFC1122. */
604 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
605 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
606 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
607 IPSTAT_INC(ips_badaddr);
608 error = EADDRNOTAVAIL;
613 m->m_pkthdr.csum_flags |= CSUM_IP;
614 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
616 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
619 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
620 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
621 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
626 * If small enough for interface, or the interface will take
627 * care of the fragmentation for us, we can just send directly.
630 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
631 ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
633 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
634 ip->ip_sum = in_cksum(m, hlen);
635 m->m_pkthdr.csum_flags &= ~CSUM_IP;
639 * Record statistics for this interface address.
640 * With CSUM_TSO the byte/packet count will be slightly
641 * incorrect because we count the IP+TCP headers only
642 * once instead of for every generated packet.
644 if (!(flags & IP_FORWARDING) && ia) {
645 if (m->m_pkthdr.csum_flags & CSUM_TSO)
646 counter_u64_add(ia->ia_ifa.ifa_opackets,
647 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
649 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
651 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
653 #ifdef MBUF_STRESS_TEST
654 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
655 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
658 * Reset layer specific mbuf flags
659 * to avoid confusing lower layers.
662 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
663 error = (*ifp->if_output)(ifp, m,
664 (const struct sockaddr *)gw, ro);
668 /* Balk when DF bit is set or the interface didn't support TSO. */
669 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
671 IPSTAT_INC(ips_cantfrag);
676 * Too large for interface; fragment if possible. If successful,
677 * on return, m will point to a list of packets to be sent.
679 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
686 /* Record statistics for this interface address. */
688 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
689 counter_u64_add(ia->ia_ifa.ifa_obytes,
693 * Reset layer specific mbuf flags
694 * to avoid confusing upper layers.
698 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
699 error = (*ifp->if_output)(ifp, m,
700 (const struct sockaddr *)gw, ro);
706 IPSTAT_INC(ips_fragmented);
712 ifa_free(&ia->ia_ifa);
720 * Create a chain of fragments which fit the given mtu. m_frag points to the
721 * mbuf to be fragmented; on return it points to the chain with the fragments.
722 * Return 0 if no error. If error, m_frag may contain a partially built
723 * chain of fragments that should be freed by the caller.
725 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
728 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
729 u_long if_hwassist_flags)
732 int hlen = ip->ip_hl << 2;
733 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
735 struct mbuf *m0 = *m_frag; /* the original packet */
739 uint16_t ip_len, ip_off;
741 ip_len = ntohs(ip->ip_len);
742 ip_off = ntohs(ip->ip_off);
744 if (ip_off & IP_DF) { /* Fragmentation not allowed */
745 IPSTAT_INC(ips_cantfrag);
750 * Must be able to put at least 8 bytes per fragment.
756 * If the interface will not calculate checksums on
757 * fragmented packets, then do it here.
759 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
760 in_delayed_cksum(m0);
761 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
764 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
765 sctp_delayed_cksum(m0, hlen);
766 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
769 if (len > PAGE_SIZE) {
771 * Fragment large datagrams such that each segment
772 * contains a multiple of PAGE_SIZE amount of data,
773 * plus headers. This enables a receiver to perform
774 * page-flipping zero-copy optimizations.
776 * XXX When does this help given that sender and receiver
777 * could have different page sizes, and also mtu could
778 * be less than the receiver's page size ?
783 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
787 * firstlen (off - hlen) must be aligned on an
791 goto smart_frag_failure;
792 off = ((off - hlen) & ~7) + hlen;
793 newlen = (~PAGE_MASK) & mtu;
794 if ((newlen + sizeof (struct ip)) > mtu) {
795 /* we failed, go back the default */
806 firstlen = off - hlen;
807 mnext = &m0->m_nextpkt; /* pointer to next packet */
810 * Loop through length of segment after first fragment,
811 * make new header and copy data of each part and link onto chain.
812 * Here, m0 is the original packet, m is the fragment being created.
813 * The fragments are linked off the m_nextpkt of the original
814 * packet, which after processing serves as the first fragment.
816 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
817 struct ip *mhip; /* ip header on the fragment */
819 int mhlen = sizeof (struct ip);
821 m = m_gethdr(M_NOWAIT, MT_DATA);
824 IPSTAT_INC(ips_odropped);
827 m->m_flags |= (m0->m_flags & M_MCAST);
829 * In the first mbuf, leave room for the link header, then
830 * copy the original IP header including options. The payload
831 * goes into an additional mbuf chain returned by m_copym().
833 m->m_data += max_linkhdr;
834 mhip = mtod(m, struct ip *);
836 if (hlen > sizeof (struct ip)) {
837 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
838 mhip->ip_v = IPVERSION;
839 mhip->ip_hl = mhlen >> 2;
842 /* XXX do we need to add ip_off below ? */
843 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
844 if (off + len >= ip_len)
847 mhip->ip_off |= IP_MF;
848 mhip->ip_len = htons((u_short)(len + mhlen));
849 m->m_next = m_copym(m0, off, len, M_NOWAIT);
850 if (m->m_next == NULL) { /* copy failed */
852 error = ENOBUFS; /* ??? */
853 IPSTAT_INC(ips_odropped);
856 m->m_pkthdr.len = mhlen + len;
857 m->m_pkthdr.rcvif = NULL;
859 mac_netinet_fragment(m0, m);
861 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
862 mhip->ip_off = htons(mhip->ip_off);
864 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
865 mhip->ip_sum = in_cksum(m, mhlen);
866 m->m_pkthdr.csum_flags &= ~CSUM_IP;
869 mnext = &m->m_nextpkt;
871 IPSTAT_ADD(ips_ofragments, nfrags);
874 * Update first fragment by trimming what's been copied out
875 * and updating header.
877 m_adj(m0, hlen + firstlen - ip_len);
878 m0->m_pkthdr.len = hlen + firstlen;
879 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
880 ip->ip_off = htons(ip_off | IP_MF);
882 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
883 ip->ip_sum = in_cksum(m0, hlen);
884 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
893 in_delayed_cksum(struct mbuf *m)
896 uint16_t csum, offset, ip_len;
898 ip = mtod(m, struct ip *);
899 offset = ip->ip_hl << 2 ;
900 ip_len = ntohs(ip->ip_len);
901 csum = in_cksum_skip(m, ip_len, offset);
902 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
904 offset += m->m_pkthdr.csum_data; /* checksum offset */
906 if (offset + sizeof(u_short) > m->m_len) {
907 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
908 m->m_len, offset, ip->ip_p);
911 * this shouldn't happen, but if it does, the
912 * correct behavior may be to insert the checksum
913 * in the appropriate next mbuf in the chain.
917 *(u_short *)(m->m_data + offset) = csum;
921 * IP socket option processing.
924 ip_ctloutput(struct socket *so, struct sockopt *sopt)
926 struct inpcb *inp = sotoinpcb(so);
930 if (sopt->sopt_level != IPPROTO_IP) {
933 if (sopt->sopt_level == SOL_SOCKET &&
934 sopt->sopt_dir == SOPT_SET) {
935 switch (sopt->sopt_name) {
938 if ((so->so_options & SO_REUSEADDR) != 0)
939 inp->inp_flags2 |= INP_REUSEADDR;
941 inp->inp_flags2 &= ~INP_REUSEADDR;
947 if ((so->so_options & SO_REUSEPORT) != 0)
948 inp->inp_flags2 |= INP_REUSEPORT;
950 inp->inp_flags2 &= ~INP_REUSEPORT;
956 inp->inp_inc.inc_fibnum = so->so_fibnum;
967 switch (sopt->sopt_dir) {
969 switch (sopt->sopt_name) {
976 if (sopt->sopt_valsize > MLEN) {
980 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
985 m->m_len = sopt->sopt_valsize;
986 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
993 error = ip_pcbopts(inp, sopt->sopt_name, m);
999 if (sopt->sopt_td != NULL) {
1000 error = priv_check(sopt->sopt_td,
1001 PRIV_NETINET_BINDANY);
1010 case IP_RECVRETOPTS:
1011 case IP_RECVDSTADDR:
1018 error = sooptcopyin(sopt, &optval, sizeof optval,
1023 switch (sopt->sopt_name) {
1025 inp->inp_ip_tos = optval;
1029 inp->inp_ip_ttl = optval;
1033 if (optval >= 0 && optval <= MAXTTL)
1034 inp->inp_ip_minttl = optval;
1039 #define OPTSET(bit) do { \
1042 inp->inp_flags |= bit; \
1044 inp->inp_flags &= ~bit; \
1049 OPTSET(INP_RECVOPTS);
1052 case IP_RECVRETOPTS:
1053 OPTSET(INP_RECVRETOPTS);
1056 case IP_RECVDSTADDR:
1057 OPTSET(INP_RECVDSTADDR);
1061 OPTSET(INP_RECVTTL);
1073 OPTSET(INP_ONESBCAST);
1076 OPTSET(INP_DONTFRAG);
1079 OPTSET(INP_BINDANY);
1082 OPTSET(INP_RECVTOS);
1089 * Multicast socket options are processed by the in_mcast
1092 case IP_MULTICAST_IF:
1093 case IP_MULTICAST_VIF:
1094 case IP_MULTICAST_TTL:
1095 case IP_MULTICAST_LOOP:
1096 case IP_ADD_MEMBERSHIP:
1097 case IP_DROP_MEMBERSHIP:
1098 case IP_ADD_SOURCE_MEMBERSHIP:
1099 case IP_DROP_SOURCE_MEMBERSHIP:
1100 case IP_BLOCK_SOURCE:
1101 case IP_UNBLOCK_SOURCE:
1103 case MCAST_JOIN_GROUP:
1104 case MCAST_LEAVE_GROUP:
1105 case MCAST_JOIN_SOURCE_GROUP:
1106 case MCAST_LEAVE_SOURCE_GROUP:
1107 case MCAST_BLOCK_SOURCE:
1108 case MCAST_UNBLOCK_SOURCE:
1109 error = inp_setmoptions(inp, sopt);
1113 error = sooptcopyin(sopt, &optval, sizeof optval,
1120 case IP_PORTRANGE_DEFAULT:
1121 inp->inp_flags &= ~(INP_LOWPORT);
1122 inp->inp_flags &= ~(INP_HIGHPORT);
1125 case IP_PORTRANGE_HIGH:
1126 inp->inp_flags &= ~(INP_LOWPORT);
1127 inp->inp_flags |= INP_HIGHPORT;
1130 case IP_PORTRANGE_LOW:
1131 inp->inp_flags &= ~(INP_HIGHPORT);
1132 inp->inp_flags |= INP_LOWPORT;
1143 case IP_IPSEC_POLICY:
1148 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1150 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1152 req = mtod(m, caddr_t);
1153 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1154 m->m_len, (sopt->sopt_td != NULL) ?
1155 sopt->sopt_td->td_ucred : NULL);
1162 error = ENOPROTOOPT;
1168 switch (sopt->sopt_name) {
1171 if (inp->inp_options)
1172 error = sooptcopyout(sopt,
1173 mtod(inp->inp_options,
1175 inp->inp_options->m_len);
1177 sopt->sopt_valsize = 0;
1184 case IP_RECVRETOPTS:
1185 case IP_RECVDSTADDR:
1194 switch (sopt->sopt_name) {
1197 optval = inp->inp_ip_tos;
1201 optval = inp->inp_ip_ttl;
1205 optval = inp->inp_ip_minttl;
1208 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1211 optval = OPTBIT(INP_RECVOPTS);
1214 case IP_RECVRETOPTS:
1215 optval = OPTBIT(INP_RECVRETOPTS);
1218 case IP_RECVDSTADDR:
1219 optval = OPTBIT(INP_RECVDSTADDR);
1223 optval = OPTBIT(INP_RECVTTL);
1227 optval = OPTBIT(INP_RECVIF);
1231 if (inp->inp_flags & INP_HIGHPORT)
1232 optval = IP_PORTRANGE_HIGH;
1233 else if (inp->inp_flags & INP_LOWPORT)
1234 optval = IP_PORTRANGE_LOW;
1240 optval = OPTBIT(INP_FAITH);
1244 optval = OPTBIT(INP_ONESBCAST);
1247 optval = OPTBIT(INP_DONTFRAG);
1250 optval = OPTBIT(INP_BINDANY);
1253 optval = OPTBIT(INP_RECVTOS);
1256 error = sooptcopyout(sopt, &optval, sizeof optval);
1260 * Multicast socket options are processed by the in_mcast
1263 case IP_MULTICAST_IF:
1264 case IP_MULTICAST_VIF:
1265 case IP_MULTICAST_TTL:
1266 case IP_MULTICAST_LOOP:
1268 error = inp_getmoptions(inp, sopt);
1272 case IP_IPSEC_POLICY:
1274 struct mbuf *m = NULL;
1279 req = mtod(m, caddr_t);
1282 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1284 error = soopt_mcopyout(sopt, m); /* XXX */
1292 error = ENOPROTOOPT;
1301 * Routine called from ip_output() to loop back a copy of an IP multicast
1302 * packet to the input queue of a specified interface. Note that this
1303 * calls the output routine of the loopback "driver", but with an interface
1304 * pointer that might NOT be a loopback interface -- evil, but easier than
1305 * replicating that code here.
1308 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1311 register struct ip *ip;
1315 * Make a deep copy of the packet because we're going to
1316 * modify the pack in order to generate checksums.
1318 copym = m_dup(m, M_NOWAIT);
1319 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1320 copym = m_pullup(copym, hlen);
1321 if (copym != NULL) {
1322 /* If needed, compute the checksum and mark it as valid. */
1323 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1324 in_delayed_cksum(copym);
1325 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1326 copym->m_pkthdr.csum_flags |=
1327 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1328 copym->m_pkthdr.csum_data = 0xffff;
1331 * We don't bother to fragment if the IP length is greater
1332 * than the interface's MTU. Can this possibly matter?
1334 ip = mtod(copym, struct ip *);
1336 ip->ip_sum = in_cksum(copym, hlen);
1338 if (dst->sin_family != AF_INET) {
1339 printf("ip_mloopback: bad address family %d\n",
1341 dst->sin_family = AF_INET;
1344 if_simloop(ifp, copym, dst->sin_family, 0);