1 /*-
2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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.
16 *
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
27 * SUCH DAMAGE.
28 *
29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
30 */
31
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD: releng/9.0/sys/netinet/ip_output.c 227428 2011-11-10 20:28:30Z trociny $");
34
35 #include "opt_ipfw.h"
36 #include "opt_ipsec.h"
37 #include "opt_route.h"
38 #include "opt_mbuf_stress_test.h"
39 #include "opt_mpath.h"
40 #include "opt_sctp.h"
41
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/malloc.h>
46 #include <sys/mbuf.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sysctl.h>
53 #include <sys/ucred.h>
54
55 #include <net/if.h>
56 #include <net/if_llatbl.h>
57 #include <net/netisr.h>
58 #include <net/pfil.h>
59 #include <net/route.h>
60 #include <net/flowtable.h>
61 #ifdef RADIX_MPATH
62 #include <net/radix_mpath.h>
63 #endif
64 #include <net/vnet.h>
65
66 #include <netinet/in.h>
67 #include <netinet/in_systm.h>
68 #include <netinet/ip.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/in_var.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/ip_options.h>
73 #ifdef SCTP
74 #include <netinet/sctp.h>
75 #include <netinet/sctp_crc32.h>
76 #endif
77
78 #ifdef IPSEC
79 #include <netinet/ip_ipsec.h>
80 #include <netipsec/ipsec.h>
81 #endif /* IPSEC*/
82
83 #include <machine/in_cksum.h>
84
85 #include <security/mac/mac_framework.h>
86
87 VNET_DEFINE(u_short, ip_id);
88
89 #ifdef MBUF_STRESS_TEST
90 static int mbuf_frag_size = 0;
91 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
92 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
93 #endif
94
95 static void ip_mloopback
96 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
97
98
99 extern int in_mcast_loop;
100 extern struct protosw inetsw[];
101
102 /*
103 * IP output. The packet in mbuf chain m contains a skeletal IP
104 * header (with len, off, ttl, proto, tos, src, dst).
105 * ip_len and ip_off are in host format.
106 * The mbuf chain containing the packet will be freed.
107 * The mbuf opt, if present, will not be freed.
108 * In the IP forwarding case, the packet will arrive with options already
109 * inserted, so must have a NULL opt pointer.
110 */
111 int
112 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
113 struct ip_moptions *imo, struct inpcb *inp)
114 {
115 struct ip *ip;
116 struct ifnet *ifp = NULL; /* keep compiler happy */
117 struct mbuf *m0;
118 int hlen = sizeof (struct ip);
119 int mtu;
120 int n; /* scratchpad */
121 int error = 0;
122 int nortfree = 0;
123 struct sockaddr_in *dst;
124 struct in_ifaddr *ia = NULL;
125 int isbroadcast, sw_csum;
126 struct route iproute;
127 struct rtentry *rte; /* cache for ro->ro_rt */
128 struct in_addr odst;
129 #ifdef IPFIREWALL_FORWARD
130 struct m_tag *fwd_tag = NULL;
131 #endif
132 #ifdef IPSEC
133 int no_route_but_check_spd = 0;
134 #endif
135 M_ASSERTPKTHDR(m);
136
137 if (inp != NULL) {
138 INP_LOCK_ASSERT(inp);
139 M_SETFIB(m, inp->inp_inc.inc_fibnum);
140 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
141 m->m_pkthdr.flowid = inp->inp_flowid;
142 m->m_flags |= M_FLOWID;
143 }
144 }
145
146 if (ro == NULL) {
147 ro = &iproute;
148 bzero(ro, sizeof (*ro));
149
150 #ifdef FLOWTABLE
151 {
152 struct flentry *fle;
153
154 /*
155 * The flow table returns route entries valid for up to 30
156 * seconds; we rely on the remainder of ip_output() taking no
157 * longer than that long for the stability of ro_rt. The
158 * flow ID assignment must have happened before this point.
159 */
160 if ((fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET)) != NULL) {
161 flow_to_route(fle, ro);
162 nortfree = 1;
163 }
164 }
165 #endif
166 }
167
168 if (opt) {
169 int len = 0;
170 m = ip_insertoptions(m, opt, &len);
171 if (len != 0)
172 hlen = len; /* ip->ip_hl is updated above */
173 }
174 ip = mtod(m, struct ip *);
175
176 /*
177 * Fill in IP header. If we are not allowing fragmentation,
178 * then the ip_id field is meaningless, but we don't set it
179 * to zero. Doing so causes various problems when devices along
180 * the path (routers, load balancers, firewalls, etc.) illegally
181 * disable DF on our packet. Note that a 16-bit counter
182 * will wrap around in less than 10 seconds at 100 Mbit/s on a
183 * medium with MTU 1500. See Steven M. Bellovin, "A Technique
184 * for Counting NATted Hosts", Proc. IMW'02, available at
185 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
186 */
187 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
188 ip->ip_v = IPVERSION;
189 ip->ip_hl = hlen >> 2;
190 ip->ip_id = ip_newid();
191 IPSTAT_INC(ips_localout);
192 } else {
193 /* Header already set, fetch hlen from there */
194 hlen = ip->ip_hl << 2;
195 }
196
197 dst = (struct sockaddr_in *)&ro->ro_dst;
198 again:
199 /*
200 * If there is a cached route,
201 * check that it is to the same destination
202 * and is still up. If not, free it and try again.
203 * The address family should also be checked in case of sharing the
204 * cache with IPv6.
205 */
206 rte = ro->ro_rt;
207 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
208 rte->rt_ifp == NULL ||
209 !RT_LINK_IS_UP(rte->rt_ifp) ||
210 dst->sin_family != AF_INET ||
211 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
212 if (!nortfree)
213 RTFREE(rte);
214 rte = ro->ro_rt = (struct rtentry *)NULL;
215 ro->ro_lle = (struct llentry *)NULL;
216 }
217 #ifdef IPFIREWALL_FORWARD
218 if (rte == NULL && fwd_tag == NULL) {
219 #else
220 if (rte == NULL) {
221 #endif
222 bzero(dst, sizeof(*dst));
223 dst->sin_family = AF_INET;
224 dst->sin_len = sizeof(*dst);
225 dst->sin_addr = ip->ip_dst;
226 }
227 /*
228 * If routing to interface only, short circuit routing lookup.
229 * The use of an all-ones broadcast address implies this; an
230 * interface is specified by the broadcast address of an interface,
231 * or the destination address of a ptp interface.
232 */
233 if (flags & IP_SENDONES) {
234 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
235 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
236 IPSTAT_INC(ips_noroute);
237 error = ENETUNREACH;
238 goto bad;
239 }
240 ip->ip_dst.s_addr = INADDR_BROADCAST;
241 dst->sin_addr = ip->ip_dst;
242 ifp = ia->ia_ifp;
243 ip->ip_ttl = 1;
244 isbroadcast = 1;
245 } else if (flags & IP_ROUTETOIF) {
246 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
247 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) {
248 IPSTAT_INC(ips_noroute);
249 error = ENETUNREACH;
250 goto bad;
251 }
252 ifp = ia->ia_ifp;
253 ip->ip_ttl = 1;
254 isbroadcast = in_broadcast(dst->sin_addr, ifp);
255 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
256 imo != NULL && imo->imo_multicast_ifp != NULL) {
257 /*
258 * Bypass the normal routing lookup for multicast
259 * packets if the interface is specified.
260 */
261 ifp = imo->imo_multicast_ifp;
262 IFP_TO_IA(ifp, ia);
263 isbroadcast = 0; /* fool gcc */
264 } else {
265 /*
266 * We want to do any cloning requested by the link layer,
267 * as this is probably required in all cases for correct
268 * operation (as it is for ARP).
269 */
270 if (rte == NULL) {
271 #ifdef RADIX_MPATH
272 rtalloc_mpath_fib(ro,
273 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
274 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
275 #else
276 in_rtalloc_ign(ro, 0,
277 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
278 #endif
279 rte = ro->ro_rt;
280 }
281 if (rte == NULL ||
282 rte->rt_ifp == NULL ||
283 !RT_LINK_IS_UP(rte->rt_ifp)) {
284 #ifdef IPSEC
285 /*
286 * There is no route for this packet, but it is
287 * possible that a matching SPD entry exists.
288 */
289 no_route_but_check_spd = 1;
290 mtu = 0; /* Silence GCC warning. */
291 goto sendit;
292 #endif
293 IPSTAT_INC(ips_noroute);
294 error = EHOSTUNREACH;
295 goto bad;
296 }
297 ia = ifatoia(rte->rt_ifa);
298 ifa_ref(&ia->ia_ifa);
299 ifp = rte->rt_ifp;
300 rte->rt_rmx.rmx_pksent++;
301 if (rte->rt_flags & RTF_GATEWAY)
302 dst = (struct sockaddr_in *)rte->rt_gateway;
303 if (rte->rt_flags & RTF_HOST)
304 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
305 else
306 isbroadcast = in_broadcast(dst->sin_addr, ifp);
307 }
308 /*
309 * Calculate MTU. If we have a route that is up, use that,
310 * otherwise use the interface's MTU.
311 */
312 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
313 /*
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.
319 */
320 if (rte->rt_rmx.rmx_mtu > ifp->if_mtu)
321 rte->rt_rmx.rmx_mtu = ifp->if_mtu;
322 mtu = rte->rt_rmx.rmx_mtu;
323 } else {
324 mtu = ifp->if_mtu;
325 }
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;
331 /*
332 * IP destination address is multicast. Make sure "dst"
333 * still points to the address in "ro". (It may have been
334 * changed to point to a gateway address, above.)
335 */
336 dst = (struct sockaddr_in *)&ro->ro_dst;
337 /*
338 * See if the caller provided any multicast options
339 */
340 if (imo != NULL) {
341 ip->ip_ttl = imo->imo_multicast_ttl;
342 if (imo->imo_multicast_vif != -1)
343 ip->ip_src.s_addr =
344 ip_mcast_src ?
345 ip_mcast_src(imo->imo_multicast_vif) :
346 INADDR_ANY;
347 } else
348 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
349 /*
350 * Confirm that the outgoing interface supports multicast.
351 */
352 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
353 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
354 IPSTAT_INC(ips_noroute);
355 error = ENETUNREACH;
356 goto bad;
357 }
358 }
359 /*
360 * If source address not specified yet, use address
361 * of outgoing interface.
362 */
363 if (ip->ip_src.s_addr == INADDR_ANY) {
364 /* Interface may have no addresses. */
365 if (ia != NULL)
366 ip->ip_src = IA_SIN(ia)->sin_addr;
367 }
368
369 if ((imo == NULL && in_mcast_loop) ||
370 (imo && imo->imo_multicast_loop)) {
371 /*
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().
377 */
378 ip_mloopback(ifp, m, dst, hlen);
379 } else {
380 /*
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.
387 *
388 * Multicasts that are looped back by ip_mloopback(),
389 * above, will be forwarded by the ip_input() routine,
390 * if necessary.
391 */
392 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
393 /*
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.
398 */
399 if (!V_rsvp_on)
400 imo = NULL;
401 if (ip_mforward &&
402 ip_mforward(ip, ifp, m, imo) != 0) {
403 m_freem(m);
404 goto done;
405 }
406 }
407 }
408
409 /*
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.
417 */
418 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
419 m_freem(m);
420 goto done;
421 }
422
423 goto sendit;
424 }
425
426 /*
427 * If the source address is not specified yet, use the address
428 * of the outoing interface.
429 */
430 if (ip->ip_src.s_addr == INADDR_ANY) {
431 /* Interface may have no addresses. */
432 if (ia != NULL) {
433 ip->ip_src = IA_SIN(ia)->sin_addr;
434 }
435 }
436
437 /*
438 * Verify that we have any chance at all of being able to queue the
439 * packet or packet fragments, unless ALTQ is enabled on the given
440 * interface in which case packetdrop should be done by queueing.
441 */
442 n = ip->ip_len / mtu + 1; /* how many fragments ? */
443 if (
444 #ifdef ALTQ
445 (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
446 #endif /* ALTQ */
447 (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
448 error = ENOBUFS;
449 IPSTAT_INC(ips_odropped);
450 ifp->if_snd.ifq_drops += n;
451 goto bad;
452 }
453
454 /*
455 * Look for broadcast address and
456 * verify user is allowed to send
457 * such a packet.
458 */
459 if (isbroadcast) {
460 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
461 error = EADDRNOTAVAIL;
462 goto bad;
463 }
464 if ((flags & IP_ALLOWBROADCAST) == 0) {
465 error = EACCES;
466 goto bad;
467 }
468 /* don't allow broadcast messages to be fragmented */
469 if (ip->ip_len > mtu) {
470 error = EMSGSIZE;
471 goto bad;
472 }
473 m->m_flags |= M_BCAST;
474 } else {
475 m->m_flags &= ~M_BCAST;
476 }
477
478 sendit:
479 #ifdef IPSEC
480 switch(ip_ipsec_output(&m, inp, &flags, &error)) {
481 case 1:
482 goto bad;
483 case -1:
484 goto done;
485 case 0:
486 default:
487 break; /* Continue with packet processing. */
488 }
489 /*
490 * Check if there was a route for this packet; return error if not.
491 */
492 if (no_route_but_check_spd) {
493 IPSTAT_INC(ips_noroute);
494 error = EHOSTUNREACH;
495 goto bad;
496 }
497 /* Update variables that are affected by ipsec4_output(). */
498 ip = mtod(m, struct ip *);
499 hlen = ip->ip_hl << 2;
500 #endif /* IPSEC */
501
502 /* Jump over all PFIL processing if hooks are not active. */
503 if (!PFIL_HOOKED(&V_inet_pfil_hook))
504 goto passout;
505
506 /* Run through list of hooks for output packets. */
507 odst.s_addr = ip->ip_dst.s_addr;
508 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
509 if (error != 0 || m == NULL)
510 goto done;
511
512 ip = mtod(m, struct ip *);
513
514 /* See if destination IP address was changed by packet filter. */
515 if (odst.s_addr != ip->ip_dst.s_addr) {
516 m->m_flags |= M_SKIP_FIREWALL;
517 /* If destination is now ourself drop to ip_input(). */
518 if (in_localip(ip->ip_dst)) {
519 m->m_flags |= M_FASTFWD_OURS;
520 if (m->m_pkthdr.rcvif == NULL)
521 m->m_pkthdr.rcvif = V_loif;
522 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
523 m->m_pkthdr.csum_flags |=
524 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
525 m->m_pkthdr.csum_data = 0xffff;
526 }
527 m->m_pkthdr.csum_flags |=
528 CSUM_IP_CHECKED | CSUM_IP_VALID;
529 #ifdef SCTP
530 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
531 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
532 #endif
533 error = netisr_queue(NETISR_IP, m);
534 goto done;
535 } else
536 goto again; /* Redo the routing table lookup. */
537 }
538
539 #ifdef IPFIREWALL_FORWARD
540 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
541 if (m->m_flags & M_FASTFWD_OURS) {
542 if (m->m_pkthdr.rcvif == NULL)
543 m->m_pkthdr.rcvif = V_loif;
544 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
545 m->m_pkthdr.csum_flags |=
546 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
547 m->m_pkthdr.csum_data = 0xffff;
548 }
549 #ifdef SCTP
550 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
551 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
552 #endif
553 m->m_pkthdr.csum_flags |=
554 CSUM_IP_CHECKED | CSUM_IP_VALID;
555
556 error = netisr_queue(NETISR_IP, m);
557 goto done;
558 }
559 /* Or forward to some other address? */
560 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
561 if (fwd_tag) {
562 dst = (struct sockaddr_in *)&ro->ro_dst;
563 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
564 m->m_flags |= M_SKIP_FIREWALL;
565 m_tag_delete(m, fwd_tag);
566 goto again;
567 }
568 #endif /* IPFIREWALL_FORWARD */
569
570 passout:
571 /* 127/8 must not appear on wire - RFC1122. */
572 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
573 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
574 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
575 IPSTAT_INC(ips_badaddr);
576 error = EADDRNOTAVAIL;
577 goto bad;
578 }
579 }
580
581 m->m_pkthdr.csum_flags |= CSUM_IP;
582 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
583 if (sw_csum & CSUM_DELAY_DATA) {
584 in_delayed_cksum(m);
585 sw_csum &= ~CSUM_DELAY_DATA;
586 }
587 #ifdef SCTP
588 if (sw_csum & CSUM_SCTP) {
589 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
590 sw_csum &= ~CSUM_SCTP;
591 }
592 #endif
593 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
594
595 /*
596 * If small enough for interface, or the interface will take
597 * care of the fragmentation for us, we can just send directly.
598 */
599 if (ip->ip_len <= mtu ||
600 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
601 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
602 ip->ip_len = htons(ip->ip_len);
603 ip->ip_off = htons(ip->ip_off);
604 ip->ip_sum = 0;
605 if (sw_csum & CSUM_DELAY_IP)
606 ip->ip_sum = in_cksum(m, hlen);
607
608 /*
609 * Record statistics for this interface address.
610 * With CSUM_TSO the byte/packet count will be slightly
611 * incorrect because we count the IP+TCP headers only
612 * once instead of for every generated packet.
613 */
614 if (!(flags & IP_FORWARDING) && ia) {
615 if (m->m_pkthdr.csum_flags & CSUM_TSO)
616 ia->ia_ifa.if_opackets +=
617 m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
618 else
619 ia->ia_ifa.if_opackets++;
620 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
621 }
622 #ifdef MBUF_STRESS_TEST
623 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
624 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size);
625 #endif
626 /*
627 * Reset layer specific mbuf flags
628 * to avoid confusing lower layers.
629 */
630 m->m_flags &= ~(M_PROTOFLAGS);
631 error = (*ifp->if_output)(ifp, m,
632 (struct sockaddr *)dst, ro);
633 goto done;
634 }
635
636 /* Balk when DF bit is set or the interface didn't support TSO. */
637 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
638 error = EMSGSIZE;
639 IPSTAT_INC(ips_cantfrag);
640 goto bad;
641 }
642
643 /*
644 * Too large for interface; fragment if possible. If successful,
645 * on return, m will point to a list of packets to be sent.
646 */
647 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum);
648 if (error)
649 goto bad;
650 for (; m; m = m0) {
651 m0 = m->m_nextpkt;
652 m->m_nextpkt = 0;
653 if (error == 0) {
654 /* Record statistics for this interface address. */
655 if (ia != NULL) {
656 ia->ia_ifa.if_opackets++;
657 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
658 }
659 /*
660 * Reset layer specific mbuf flags
661 * to avoid confusing upper layers.
662 */
663 m->m_flags &= ~(M_PROTOFLAGS);
664
665 error = (*ifp->if_output)(ifp, m,
666 (struct sockaddr *)dst, ro);
667 } else
668 m_freem(m);
669 }
670
671 if (error == 0)
672 IPSTAT_INC(ips_fragmented);
673
674 done:
675 if (ro == &iproute && ro->ro_rt && !nortfree) {
676 RTFREE(ro->ro_rt);
677 }
678 if (ia != NULL)
679 ifa_free(&ia->ia_ifa);
680 return (error);
681 bad:
682 m_freem(m);
683 goto done;
684 }
685
686 /*
687 * Create a chain of fragments which fit the given mtu. m_frag points to the
688 * mbuf to be fragmented; on return it points to the chain with the fragments.
689 * Return 0 if no error. If error, m_frag may contain a partially built
690 * chain of fragments that should be freed by the caller.
691 *
692 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
693 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
694 */
695 int
696 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
697 u_long if_hwassist_flags, int sw_csum)
698 {
699 int error = 0;
700 int hlen = ip->ip_hl << 2;
701 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
702 int off;
703 struct mbuf *m0 = *m_frag; /* the original packet */
704 int firstlen;
705 struct mbuf **mnext;
706 int nfrags;
707
708 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
709 IPSTAT_INC(ips_cantfrag);
710 return EMSGSIZE;
711 }
712
713 /*
714 * Must be able to put at least 8 bytes per fragment.
715 */
716 if (len < 8)
717 return EMSGSIZE;
718
719 /*
720 * If the interface will not calculate checksums on
721 * fragmented packets, then do it here.
722 */
723 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
724 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
725 in_delayed_cksum(m0);
726 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
727 }
728 #ifdef SCTP
729 if (m0->m_pkthdr.csum_flags & CSUM_SCTP &&
730 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
731 sctp_delayed_cksum(m0, hlen);
732 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
733 }
734 #endif
735 if (len > PAGE_SIZE) {
736 /*
737 * Fragment large datagrams such that each segment
738 * contains a multiple of PAGE_SIZE amount of data,
739 * plus headers. This enables a receiver to perform
740 * page-flipping zero-copy optimizations.
741 *
742 * XXX When does this help given that sender and receiver
743 * could have different page sizes, and also mtu could
744 * be less than the receiver's page size ?
745 */
746 int newlen;
747 struct mbuf *m;
748
749 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
750 off += m->m_len;
751
752 /*
753 * firstlen (off - hlen) must be aligned on an
754 * 8-byte boundary
755 */
756 if (off < hlen)
757 goto smart_frag_failure;
758 off = ((off - hlen) & ~7) + hlen;
759 newlen = (~PAGE_MASK) & mtu;
760 if ((newlen + sizeof (struct ip)) > mtu) {
761 /* we failed, go back the default */
762 smart_frag_failure:
763 newlen = len;
764 off = hlen + len;
765 }
766 len = newlen;
767
768 } else {
769 off = hlen + len;
770 }
771
772 firstlen = off - hlen;
773 mnext = &m0->m_nextpkt; /* pointer to next packet */
774
775 /*
776 * Loop through length of segment after first fragment,
777 * make new header and copy data of each part and link onto chain.
778 * Here, m0 is the original packet, m is the fragment being created.
779 * The fragments are linked off the m_nextpkt of the original
780 * packet, which after processing serves as the first fragment.
781 */
782 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
783 struct ip *mhip; /* ip header on the fragment */
784 struct mbuf *m;
785 int mhlen = sizeof (struct ip);
786
787 MGETHDR(m, M_DONTWAIT, MT_DATA);
788 if (m == NULL) {
789 error = ENOBUFS;
790 IPSTAT_INC(ips_odropped);
791 goto done;
792 }
793 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
794 /*
795 * In the first mbuf, leave room for the link header, then
796 * copy the original IP header including options. The payload
797 * goes into an additional mbuf chain returned by m_copym().
798 */
799 m->m_data += max_linkhdr;
800 mhip = mtod(m, struct ip *);
801 *mhip = *ip;
802 if (hlen > sizeof (struct ip)) {
803 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
804 mhip->ip_v = IPVERSION;
805 mhip->ip_hl = mhlen >> 2;
806 }
807 m->m_len = mhlen;
808 /* XXX do we need to add ip->ip_off below ? */
809 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
810 if (off + len >= ip->ip_len) { /* last fragment */
811 len = ip->ip_len - off;
812 m->m_flags |= M_LASTFRAG;
813 } else
814 mhip->ip_off |= IP_MF;
815 mhip->ip_len = htons((u_short)(len + mhlen));
816 m->m_next = m_copym(m0, off, len, M_DONTWAIT);
817 if (m->m_next == NULL) { /* copy failed */
818 m_free(m);
819 error = ENOBUFS; /* ??? */
820 IPSTAT_INC(ips_odropped);
821 goto done;
822 }
823 m->m_pkthdr.len = mhlen + len;
824 m->m_pkthdr.rcvif = NULL;
825 #ifdef MAC
826 mac_netinet_fragment(m0, m);
827 #endif
828 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
829 mhip->ip_off = htons(mhip->ip_off);
830 mhip->ip_sum = 0;
831 if (sw_csum & CSUM_DELAY_IP)
832 mhip->ip_sum = in_cksum(m, mhlen);
833 *mnext = m;
834 mnext = &m->m_nextpkt;
835 }
836 IPSTAT_ADD(ips_ofragments, nfrags);
837
838 /* set first marker for fragment chain */
839 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
840 m0->m_pkthdr.csum_data = nfrags;
841
842 /*
843 * Update first fragment by trimming what's been copied out
844 * and updating header.
845 */
846 m_adj(m0, hlen + firstlen - ip->ip_len);
847 m0->m_pkthdr.len = hlen + firstlen;
848 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
849 ip->ip_off |= IP_MF;
850 ip->ip_off = htons(ip->ip_off);
851 ip->ip_sum = 0;
852 if (sw_csum & CSUM_DELAY_IP)
853 ip->ip_sum = in_cksum(m0, hlen);
854
855 done:
856 *m_frag = m0;
857 return error;
858 }
859
860 void
861 in_delayed_cksum(struct mbuf *m)
862 {
863 struct ip *ip;
864 u_short csum, offset;
865
866 ip = mtod(m, struct ip *);
867 offset = ip->ip_hl << 2 ;
868 csum = in_cksum_skip(m, ip->ip_len, offset);
869 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
870 csum = 0xffff;
871 offset += m->m_pkthdr.csum_data; /* checksum offset */
872
873 if (offset + sizeof(u_short) > m->m_len) {
874 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
875 m->m_len, offset, ip->ip_p);
876 /*
877 * XXX
878 * this shouldn't happen, but if it does, the
879 * correct behavior may be to insert the checksum
880 * in the appropriate next mbuf in the chain.
881 */
882 return;
883 }
884 *(u_short *)(m->m_data + offset) = csum;
885 }
886
887 /*
888 * IP socket option processing.
889 */
890 int
891 ip_ctloutput(struct socket *so, struct sockopt *sopt)
892 {
893 struct inpcb *inp = sotoinpcb(so);
894 int error, optval;
895
896 error = optval = 0;
897 if (sopt->sopt_level != IPPROTO_IP) {
898 error = EINVAL;
899
900 if (sopt->sopt_level == SOL_SOCKET &&
901 sopt->sopt_dir == SOPT_SET) {
902 switch (sopt->sopt_name) {
903 case SO_REUSEADDR:
904 INP_WLOCK(inp);
905 if (IN_MULTICAST(ntohl(inp->inp_laddr.s_addr))) {
906 if ((so->so_options &
907 (SO_REUSEADDR | SO_REUSEPORT)) != 0)
908 inp->inp_flags2 |= INP_REUSEPORT;
909 else
910 inp->inp_flags2 &= ~INP_REUSEPORT;
911 }
912 INP_WUNLOCK(inp);
913 error = 0;
914 break;
915 case SO_REUSEPORT:
916 INP_WLOCK(inp);
917 if ((so->so_options & SO_REUSEPORT) != 0)
918 inp->inp_flags2 |= INP_REUSEPORT;
919 else
920 inp->inp_flags2 &= ~INP_REUSEPORT;
921 INP_WUNLOCK(inp);
922 error = 0;
923 break;
924 case SO_SETFIB:
925 INP_WLOCK(inp);
926 inp->inp_inc.inc_fibnum = so->so_fibnum;
927 INP_WUNLOCK(inp);
928 error = 0;
929 break;
930 default:
931 break;
932 }
933 }
934 return (error);
935 }
936
937 switch (sopt->sopt_dir) {
938 case SOPT_SET:
939 switch (sopt->sopt_name) {
940 case IP_OPTIONS:
941 #ifdef notyet
942 case IP_RETOPTS:
943 #endif
944 {
945 struct mbuf *m;
946 if (sopt->sopt_valsize > MLEN) {
947 error = EMSGSIZE;
948 break;
949 }
950 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
951 if (m == NULL) {
952 error = ENOBUFS;
953 break;
954 }
955 m->m_len = sopt->sopt_valsize;
956 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
957 m->m_len);
958 if (error) {
959 m_free(m);
960 break;
961 }
962 INP_WLOCK(inp);
963 error = ip_pcbopts(inp, sopt->sopt_name, m);
964 INP_WUNLOCK(inp);
965 return (error);
966 }
967
968 case IP_BINDANY:
969 if (sopt->sopt_td != NULL) {
970 error = priv_check(sopt->sopt_td,
971 PRIV_NETINET_BINDANY);
972 if (error)
973 break;
974 }
975 /* FALLTHROUGH */
976 case IP_TOS:
977 case IP_TTL:
978 case IP_MINTTL:
979 case IP_RECVOPTS:
980 case IP_RECVRETOPTS:
981 case IP_RECVDSTADDR:
982 case IP_RECVTTL:
983 case IP_RECVIF:
984 case IP_FAITH:
985 case IP_ONESBCAST:
986 case IP_DONTFRAG:
987 error = sooptcopyin(sopt, &optval, sizeof optval,
988 sizeof optval);
989 if (error)
990 break;
991
992 switch (sopt->sopt_name) {
993 case IP_TOS:
994 inp->inp_ip_tos = optval;
995 break;
996
997 case IP_TTL:
998 inp->inp_ip_ttl = optval;
999 break;
1000
1001 case IP_MINTTL:
1002 if (optval >= 0 && optval <= MAXTTL)
1003 inp->inp_ip_minttl = optval;
1004 else
1005 error = EINVAL;
1006 break;
1007
1008 #define OPTSET(bit) do { \
1009 INP_WLOCK(inp); \
1010 if (optval) \
1011 inp->inp_flags |= bit; \
1012 else \
1013 inp->inp_flags &= ~bit; \
1014 INP_WUNLOCK(inp); \
1015 } while (0)
1016
1017 case IP_RECVOPTS:
1018 OPTSET(INP_RECVOPTS);
1019 break;
1020
1021 case IP_RECVRETOPTS:
1022 OPTSET(INP_RECVRETOPTS);
1023 break;
1024
1025 case IP_RECVDSTADDR:
1026 OPTSET(INP_RECVDSTADDR);
1027 break;
1028
1029 case IP_RECVTTL:
1030 OPTSET(INP_RECVTTL);
1031 break;
1032
1033 case IP_RECVIF:
1034 OPTSET(INP_RECVIF);
1035 break;
1036
1037 case IP_FAITH:
1038 OPTSET(INP_FAITH);
1039 break;
1040
1041 case IP_ONESBCAST:
1042 OPTSET(INP_ONESBCAST);
1043 break;
1044 case IP_DONTFRAG:
1045 OPTSET(INP_DONTFRAG);
1046 break;
1047 case IP_BINDANY:
1048 OPTSET(INP_BINDANY);
1049 break;
1050 }
1051 break;
1052 #undef OPTSET
1053
1054 /*
1055 * Multicast socket options are processed by the in_mcast
1056 * module.
1057 */
1058 case IP_MULTICAST_IF:
1059 case IP_MULTICAST_VIF:
1060 case IP_MULTICAST_TTL:
1061 case IP_MULTICAST_LOOP:
1062 case IP_ADD_MEMBERSHIP:
1063 case IP_DROP_MEMBERSHIP:
1064 case IP_ADD_SOURCE_MEMBERSHIP:
1065 case IP_DROP_SOURCE_MEMBERSHIP:
1066 case IP_BLOCK_SOURCE:
1067 case IP_UNBLOCK_SOURCE:
1068 case IP_MSFILTER:
1069 case MCAST_JOIN_GROUP:
1070 case MCAST_LEAVE_GROUP:
1071 case MCAST_JOIN_SOURCE_GROUP:
1072 case MCAST_LEAVE_SOURCE_GROUP:
1073 case MCAST_BLOCK_SOURCE:
1074 case MCAST_UNBLOCK_SOURCE:
1075 error = inp_setmoptions(inp, sopt);
1076 break;
1077
1078 case IP_PORTRANGE:
1079 error = sooptcopyin(sopt, &optval, sizeof optval,
1080 sizeof optval);
1081 if (error)
1082 break;
1083
1084 INP_WLOCK(inp);
1085 switch (optval) {
1086 case IP_PORTRANGE_DEFAULT:
1087 inp->inp_flags &= ~(INP_LOWPORT);
1088 inp->inp_flags &= ~(INP_HIGHPORT);
1089 break;
1090
1091 case IP_PORTRANGE_HIGH:
1092 inp->inp_flags &= ~(INP_LOWPORT);
1093 inp->inp_flags |= INP_HIGHPORT;
1094 break;
1095
1096 case IP_PORTRANGE_LOW:
1097 inp->inp_flags &= ~(INP_HIGHPORT);
1098 inp->inp_flags |= INP_LOWPORT;
1099 break;
1100
1101 default:
1102 error = EINVAL;
1103 break;
1104 }
1105 INP_WUNLOCK(inp);
1106 break;
1107
1108 #ifdef IPSEC
1109 case IP_IPSEC_POLICY:
1110 {
1111 caddr_t req;
1112 struct mbuf *m;
1113
1114 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1115 break;
1116 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1117 break;
1118 req = mtod(m, caddr_t);
1119 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1120 m->m_len, (sopt->sopt_td != NULL) ?
1121 sopt->sopt_td->td_ucred : NULL);
1122 m_freem(m);
1123 break;
1124 }
1125 #endif /* IPSEC */
1126
1127 default:
1128 error = ENOPROTOOPT;
1129 break;
1130 }
1131 break;
1132
1133 case SOPT_GET:
1134 switch (sopt->sopt_name) {
1135 case IP_OPTIONS:
1136 case IP_RETOPTS:
1137 if (inp->inp_options)
1138 error = sooptcopyout(sopt,
1139 mtod(inp->inp_options,
1140 char *),
1141 inp->inp_options->m_len);
1142 else
1143 sopt->sopt_valsize = 0;
1144 break;
1145
1146 case IP_TOS:
1147 case IP_TTL:
1148 case IP_MINTTL:
1149 case IP_RECVOPTS:
1150 case IP_RECVRETOPTS:
1151 case IP_RECVDSTADDR:
1152 case IP_RECVTTL:
1153 case IP_RECVIF:
1154 case IP_PORTRANGE:
1155 case IP_FAITH:
1156 case IP_ONESBCAST:
1157 case IP_DONTFRAG:
1158 case IP_BINDANY:
1159 switch (sopt->sopt_name) {
1160
1161 case IP_TOS:
1162 optval = inp->inp_ip_tos;
1163 break;
1164
1165 case IP_TTL:
1166 optval = inp->inp_ip_ttl;
1167 break;
1168
1169 case IP_MINTTL:
1170 optval = inp->inp_ip_minttl;
1171 break;
1172
1173 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1174
1175 case IP_RECVOPTS:
1176 optval = OPTBIT(INP_RECVOPTS);
1177 break;
1178
1179 case IP_RECVRETOPTS:
1180 optval = OPTBIT(INP_RECVRETOPTS);
1181 break;
1182
1183 case IP_RECVDSTADDR:
1184 optval = OPTBIT(INP_RECVDSTADDR);
1185 break;
1186
1187 case IP_RECVTTL:
1188 optval = OPTBIT(INP_RECVTTL);
1189 break;
1190
1191 case IP_RECVIF:
1192 optval = OPTBIT(INP_RECVIF);
1193 break;
1194
1195 case IP_PORTRANGE:
1196 if (inp->inp_flags & INP_HIGHPORT)
1197 optval = IP_PORTRANGE_HIGH;
1198 else if (inp->inp_flags & INP_LOWPORT)
1199 optval = IP_PORTRANGE_LOW;
1200 else
1201 optval = 0;
1202 break;
1203
1204 case IP_FAITH:
1205 optval = OPTBIT(INP_FAITH);
1206 break;
1207
1208 case IP_ONESBCAST:
1209 optval = OPTBIT(INP_ONESBCAST);
1210 break;
1211 case IP_DONTFRAG:
1212 optval = OPTBIT(INP_DONTFRAG);
1213 break;
1214 case IP_BINDANY:
1215 optval = OPTBIT(INP_BINDANY);
1216 break;
1217 }
1218 error = sooptcopyout(sopt, &optval, sizeof optval);
1219 break;
1220
1221 /*
1222 * Multicast socket options are processed by the in_mcast
1223 * module.
1224 */
1225 case IP_MULTICAST_IF:
1226 case IP_MULTICAST_VIF:
1227 case IP_MULTICAST_TTL:
1228 case IP_MULTICAST_LOOP:
1229 case IP_MSFILTER:
1230 error = inp_getmoptions(inp, sopt);
1231 break;
1232
1233 #ifdef IPSEC
1234 case IP_IPSEC_POLICY:
1235 {
1236 struct mbuf *m = NULL;
1237 caddr_t req = NULL;
1238 size_t len = 0;
1239
1240 if (m != 0) {
1241 req = mtod(m, caddr_t);
1242 len = m->m_len;
1243 }
1244 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1245 if (error == 0)
1246 error = soopt_mcopyout(sopt, m); /* XXX */
1247 if (error == 0)
1248 m_freem(m);
1249 break;
1250 }
1251 #endif /* IPSEC */
1252
1253 default:
1254 error = ENOPROTOOPT;
1255 break;
1256 }
1257 break;
1258 }
1259 return (error);
1260 }
1261
1262 /*
1263 * Routine called from ip_output() to loop back a copy of an IP multicast
1264 * packet to the input queue of a specified interface. Note that this
1265 * calls the output routine of the loopback "driver", but with an interface
1266 * pointer that might NOT be a loopback interface -- evil, but easier than
1267 * replicating that code here.
1268 */
1269 static void
1270 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1271 int hlen)
1272 {
1273 register struct ip *ip;
1274 struct mbuf *copym;
1275
1276 /*
1277 * Make a deep copy of the packet because we're going to
1278 * modify the pack in order to generate checksums.
1279 */
1280 copym = m_dup(m, M_DONTWAIT);
1281 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1282 copym = m_pullup(copym, hlen);
1283 if (copym != NULL) {
1284 /* If needed, compute the checksum and mark it as valid. */
1285 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1286 in_delayed_cksum(copym);
1287 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1288 copym->m_pkthdr.csum_flags |=
1289 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1290 copym->m_pkthdr.csum_data = 0xffff;
1291 }
1292 /*
1293 * We don't bother to fragment if the IP length is greater
1294 * than the interface's MTU. Can this possibly matter?
1295 */
1296 ip = mtod(copym, struct ip *);
1297 ip->ip_len = htons(ip->ip_len);
1298 ip->ip_off = htons(ip->ip_off);
1299 ip->ip_sum = 0;
1300 ip->ip_sum = in_cksum(copym, hlen);
1301 #if 1 /* XXX */
1302 if (dst->sin_family != AF_INET) {
1303 printf("ip_mloopback: bad address family %d\n",
1304 dst->sin_family);
1305 dst->sin_family = AF_INET;
1306 }
1307 #endif
1308 if_simloop(ifp, copym, dst->sin_family, 0);
1309 }
1310 }
Cache object: 35bcc754fed9fd24e1c0c8a9f94a2445
|