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/8.3/sys/netinet/ip_output.c 220828 2011-04-19 08:56:26Z bz $");
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 #define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\
88 x, (ntohl(a.s_addr)>>24)&0xFF,\
89 (ntohl(a.s_addr)>>16)&0xFF,\
90 (ntohl(a.s_addr)>>8)&0xFF,\
91 (ntohl(a.s_addr))&0xFF, y);
92
93 VNET_DEFINE(u_short, ip_id);
94
95 #ifdef MBUF_STRESS_TEST
96 static int mbuf_frag_size = 0;
97 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
98 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
99 #endif
100
101 static void ip_mloopback
102 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
103
104
105 extern int in_mcast_loop;
106 extern struct protosw inetsw[];
107
108 /*
109 * IP output. The packet in mbuf chain m contains a skeletal IP
110 * header (with len, off, ttl, proto, tos, src, dst).
111 * The mbuf chain containing the packet will be freed.
112 * The mbuf opt, if present, will not be freed.
113 * In the IP forwarding case, the packet will arrive with options already
114 * inserted, so must have a NULL opt pointer.
115 */
116 int
117 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
118 struct ip_moptions *imo, struct inpcb *inp)
119 {
120 struct ip *ip;
121 struct ifnet *ifp = NULL; /* keep compiler happy */
122 struct mbuf *m0;
123 int hlen = sizeof (struct ip);
124 int mtu;
125 int len, error = 0;
126 int nortfree = 0;
127 struct sockaddr_in *dst = NULL; /* keep compiler happy */
128 struct in_ifaddr *ia = NULL;
129 int isbroadcast, sw_csum;
130 struct route iproute;
131 struct rtentry *rte; /* cache for ro->ro_rt */
132 struct in_addr odst;
133 #ifdef IPFIREWALL_FORWARD
134 struct m_tag *fwd_tag = NULL;
135 #endif
136 #ifdef IPSEC
137 int no_route_but_check_spd = 0;
138 #endif
139 M_ASSERTPKTHDR(m);
140
141 if (inp != NULL) {
142 INP_LOCK_ASSERT(inp);
143 M_SETFIB(m, inp->inp_inc.inc_fibnum);
144 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
145 m->m_pkthdr.flowid = inp->inp_flowid;
146 m->m_flags |= M_FLOWID;
147 }
148 }
149
150 if (ro == NULL) {
151 ro = &iproute;
152 bzero(ro, sizeof (*ro));
153
154 #ifdef FLOWTABLE
155 {
156 struct flentry *fle;
157
158 /*
159 * The flow table returns route entries valid for up to 30
160 * seconds; we rely on the remainder of ip_output() taking no
161 * longer than that long for the stability of ro_rt. The
162 * flow ID assignment must have happened before this point.
163 */
164 if ((fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET)) != NULL) {
165 flow_to_route(fle, ro);
166 nortfree = 1;
167 }
168 }
169 #endif
170 }
171
172 if (opt) {
173 len = 0;
174 m = ip_insertoptions(m, opt, &len);
175 if (len != 0)
176 hlen = len;
177 }
178 ip = mtod(m, struct ip *);
179
180 /*
181 * Fill in IP header. If we are not allowing fragmentation,
182 * then the ip_id field is meaningless, but we don't set it
183 * to zero. Doing so causes various problems when devices along
184 * the path (routers, load balancers, firewalls, etc.) illegally
185 * disable DF on our packet. Note that a 16-bit counter
186 * will wrap around in less than 10 seconds at 100 Mbit/s on a
187 * medium with MTU 1500. See Steven M. Bellovin, "A Technique
188 * for Counting NATted Hosts", Proc. IMW'02, available at
189 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
190 */
191 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
192 ip->ip_v = IPVERSION;
193 ip->ip_hl = hlen >> 2;
194 ip->ip_id = ip_newid();
195 IPSTAT_INC(ips_localout);
196 } else {
197 hlen = ip->ip_hl << 2;
198 }
199
200 dst = (struct sockaddr_in *)&ro->ro_dst;
201 again:
202 /*
203 * If there is a cached route,
204 * check that it is to the same destination
205 * and is still up. If not, free it and try again.
206 * The address family should also be checked in case of sharing the
207 * cache with IPv6.
208 */
209 rte = ro->ro_rt;
210 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
211 rte->rt_ifp == NULL ||
212 !RT_LINK_IS_UP(rte->rt_ifp) ||
213 dst->sin_family != AF_INET ||
214 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
215 if (!nortfree)
216 RTFREE(rte);
217 rte = ro->ro_rt = (struct rtentry *)NULL;
218 ro->ro_lle = (struct llentry *)NULL;
219 }
220 #ifdef IPFIREWALL_FORWARD
221 if (rte == NULL && fwd_tag == NULL) {
222 #else
223 if (rte == NULL) {
224 #endif
225 bzero(dst, sizeof(*dst));
226 dst->sin_family = AF_INET;
227 dst->sin_len = sizeof(*dst);
228 dst->sin_addr = ip->ip_dst;
229 }
230 /*
231 * If routing to interface only, short circuit routing lookup.
232 * The use of an all-ones broadcast address implies this; an
233 * interface is specified by the broadcast address of an interface,
234 * or the destination address of a ptp interface.
235 */
236 if (flags & IP_SENDONES) {
237 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
238 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
239 IPSTAT_INC(ips_noroute);
240 error = ENETUNREACH;
241 goto bad;
242 }
243 ip->ip_dst.s_addr = INADDR_BROADCAST;
244 dst->sin_addr = ip->ip_dst;
245 ifp = ia->ia_ifp;
246 ip->ip_ttl = 1;
247 isbroadcast = 1;
248 } else if (flags & IP_ROUTETOIF) {
249 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
250 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) {
251 IPSTAT_INC(ips_noroute);
252 error = ENETUNREACH;
253 goto bad;
254 }
255 ifp = ia->ia_ifp;
256 ip->ip_ttl = 1;
257 isbroadcast = in_broadcast(dst->sin_addr, ifp);
258 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
259 imo != NULL && imo->imo_multicast_ifp != NULL) {
260 /*
261 * Bypass the normal routing lookup for multicast
262 * packets if the interface is specified.
263 */
264 ifp = imo->imo_multicast_ifp;
265 IFP_TO_IA(ifp, ia);
266 isbroadcast = 0; /* fool gcc */
267 } else {
268 /*
269 * We want to do any cloning requested by the link layer,
270 * as this is probably required in all cases for correct
271 * operation (as it is for ARP).
272 */
273 if (rte == NULL) {
274 #ifdef RADIX_MPATH
275 rtalloc_mpath_fib(ro,
276 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
277 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
278 #else
279 in_rtalloc_ign(ro, 0,
280 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
281 #endif
282 rte = ro->ro_rt;
283 }
284 if (rte == NULL ||
285 rte->rt_ifp == NULL ||
286 !RT_LINK_IS_UP(rte->rt_ifp)) {
287 #ifdef IPSEC
288 /*
289 * There is no route for this packet, but it is
290 * possible that a matching SPD entry exists.
291 */
292 no_route_but_check_spd = 1;
293 mtu = 0; /* Silence GCC warning. */
294 goto sendit;
295 #endif
296 IPSTAT_INC(ips_noroute);
297 error = EHOSTUNREACH;
298 goto bad;
299 }
300 ia = ifatoia(rte->rt_ifa);
301 ifa_ref(&ia->ia_ifa);
302 ifp = rte->rt_ifp;
303 rte->rt_rmx.rmx_pksent++;
304 if (rte->rt_flags & RTF_GATEWAY)
305 dst = (struct sockaddr_in *)rte->rt_gateway;
306 if (rte->rt_flags & RTF_HOST)
307 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
308 else
309 isbroadcast = in_broadcast(dst->sin_addr, ifp);
310 }
311 /*
312 * Calculate MTU. If we have a route that is up, use that,
313 * otherwise use the interface's MTU.
314 */
315 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
316 /*
317 * This case can happen if the user changed the MTU
318 * of an interface after enabling IP on it. Because
319 * most netifs don't keep track of routes pointing to
320 * them, there is no way for one to update all its
321 * routes when the MTU is changed.
322 */
323 if (rte->rt_rmx.rmx_mtu > ifp->if_mtu)
324 rte->rt_rmx.rmx_mtu = ifp->if_mtu;
325 mtu = rte->rt_rmx.rmx_mtu;
326 } else {
327 mtu = ifp->if_mtu;
328 }
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 #ifdef ALTQ
443 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
444 ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >=
445 ifp->if_snd.ifq_maxlen))
446 #else
447 if ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >=
448 ifp->if_snd.ifq_maxlen)
449 #endif /* ALTQ */
450 {
451 error = ENOBUFS;
452 IPSTAT_INC(ips_odropped);
453 ifp->if_snd.ifq_drops += (ip->ip_len / ifp->if_mtu + 1);
454 goto bad;
455 }
456
457 /*
458 * Look for broadcast address and
459 * verify user is allowed to send
460 * such a packet.
461 */
462 if (isbroadcast) {
463 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
464 error = EADDRNOTAVAIL;
465 goto bad;
466 }
467 if ((flags & IP_ALLOWBROADCAST) == 0) {
468 error = EACCES;
469 goto bad;
470 }
471 /* don't allow broadcast messages to be fragmented */
472 if (ip->ip_len > mtu) {
473 error = EMSGSIZE;
474 goto bad;
475 }
476 m->m_flags |= M_BCAST;
477 } else {
478 m->m_flags &= ~M_BCAST;
479 }
480
481 sendit:
482 #ifdef IPSEC
483 switch(ip_ipsec_output(&m, inp, &flags, &error, &ifp)) {
484 case 1:
485 goto bad;
486 case -1:
487 goto done;
488 case 0:
489 default:
490 break; /* Continue with packet processing. */
491 }
492 /*
493 * Check if there was a route for this packet; return error if not.
494 */
495 if (no_route_but_check_spd) {
496 IPSTAT_INC(ips_noroute);
497 error = EHOSTUNREACH;
498 goto bad;
499 }
500 /* Update variables that are affected by ipsec4_output(). */
501 ip = mtod(m, struct ip *);
502 hlen = ip->ip_hl << 2;
503 #endif /* IPSEC */
504
505 /* Jump over all PFIL processing if hooks are not active. */
506 if (!PFIL_HOOKED(&V_inet_pfil_hook))
507 goto passout;
508
509 /* Run through list of hooks for output packets. */
510 odst.s_addr = ip->ip_dst.s_addr;
511 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
512 if (error != 0 || m == NULL)
513 goto done;
514
515 ip = mtod(m, struct ip *);
516
517 /* See if destination IP address was changed by packet filter. */
518 if (odst.s_addr != ip->ip_dst.s_addr) {
519 m->m_flags |= M_SKIP_FIREWALL;
520 /* If destination is now ourself drop to ip_input(). */
521 if (in_localip(ip->ip_dst)) {
522 m->m_flags |= M_FASTFWD_OURS;
523 if (m->m_pkthdr.rcvif == NULL)
524 m->m_pkthdr.rcvif = V_loif;
525 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
526 m->m_pkthdr.csum_flags |=
527 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
528 m->m_pkthdr.csum_data = 0xffff;
529 }
530 m->m_pkthdr.csum_flags |=
531 CSUM_IP_CHECKED | CSUM_IP_VALID;
532 #ifdef SCTP
533 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
534 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
535 #endif
536 error = netisr_queue(NETISR_IP, m);
537 goto done;
538 } else
539 goto again; /* Redo the routing table lookup. */
540 }
541
542 #ifdef IPFIREWALL_FORWARD
543 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
544 if (m->m_flags & M_FASTFWD_OURS) {
545 if (m->m_pkthdr.rcvif == NULL)
546 m->m_pkthdr.rcvif = V_loif;
547 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
548 m->m_pkthdr.csum_flags |=
549 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
550 m->m_pkthdr.csum_data = 0xffff;
551 }
552 #ifdef SCTP
553 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
554 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
555 #endif
556 m->m_pkthdr.csum_flags |=
557 CSUM_IP_CHECKED | CSUM_IP_VALID;
558
559 error = netisr_queue(NETISR_IP, m);
560 goto done;
561 }
562 /* Or forward to some other address? */
563 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
564 if (fwd_tag) {
565 dst = (struct sockaddr_in *)&ro->ro_dst;
566 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
567 m->m_flags |= M_SKIP_FIREWALL;
568 m_tag_delete(m, fwd_tag);
569 goto again;
570 }
571 #endif /* IPFIREWALL_FORWARD */
572
573 passout:
574 /* 127/8 must not appear on wire - RFC1122. */
575 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
576 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
577 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
578 IPSTAT_INC(ips_badaddr);
579 error = EADDRNOTAVAIL;
580 goto bad;
581 }
582 }
583
584 m->m_pkthdr.csum_flags |= CSUM_IP;
585 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
586 if (sw_csum & CSUM_DELAY_DATA) {
587 in_delayed_cksum(m);
588 sw_csum &= ~CSUM_DELAY_DATA;
589 }
590 #ifdef SCTP
591 if (sw_csum & CSUM_SCTP) {
592 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
593 sw_csum &= ~CSUM_SCTP;
594 }
595 #endif
596 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
597
598 /*
599 * If small enough for interface, or the interface will take
600 * care of the fragmentation for us, we can just send directly.
601 */
602 if (ip->ip_len <= mtu ||
603 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
604 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
605 ip->ip_len = htons(ip->ip_len);
606 ip->ip_off = htons(ip->ip_off);
607 ip->ip_sum = 0;
608 if (sw_csum & CSUM_DELAY_IP)
609 ip->ip_sum = in_cksum(m, hlen);
610
611 /*
612 * Record statistics for this interface address.
613 * With CSUM_TSO the byte/packet count will be slightly
614 * incorrect because we count the IP+TCP headers only
615 * once instead of for every generated packet.
616 */
617 if (!(flags & IP_FORWARDING) && ia) {
618 if (m->m_pkthdr.csum_flags & CSUM_TSO)
619 ia->ia_ifa.if_opackets +=
620 m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
621 else
622 ia->ia_ifa.if_opackets++;
623 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
624 }
625 #ifdef MBUF_STRESS_TEST
626 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
627 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size);
628 #endif
629 /*
630 * Reset layer specific mbuf flags
631 * to avoid confusing lower layers.
632 */
633 m->m_flags &= ~(M_PROTOFLAGS);
634 error = (*ifp->if_output)(ifp, m,
635 (struct sockaddr *)dst, ro);
636 goto done;
637 }
638
639 /* Balk when DF bit is set or the interface didn't support TSO. */
640 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
641 error = EMSGSIZE;
642 IPSTAT_INC(ips_cantfrag);
643 goto bad;
644 }
645
646 /*
647 * Too large for interface; fragment if possible. If successful,
648 * on return, m will point to a list of packets to be sent.
649 */
650 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum);
651 if (error)
652 goto bad;
653 for (; m; m = m0) {
654 m0 = m->m_nextpkt;
655 m->m_nextpkt = 0;
656 if (error == 0) {
657 /* Record statistics for this interface address. */
658 if (ia != NULL) {
659 ia->ia_ifa.if_opackets++;
660 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
661 }
662 /*
663 * Reset layer specific mbuf flags
664 * to avoid confusing upper layers.
665 */
666 m->m_flags &= ~(M_PROTOFLAGS);
667
668 error = (*ifp->if_output)(ifp, m,
669 (struct sockaddr *)dst, ro);
670 } else
671 m_freem(m);
672 }
673
674 if (error == 0)
675 IPSTAT_INC(ips_fragmented);
676
677 done:
678 if (ro == &iproute && ro->ro_rt && !nortfree) {
679 RTFREE(ro->ro_rt);
680 }
681 if (ia != NULL)
682 ifa_free(&ia->ia_ifa);
683 return (error);
684 bad:
685 m_freem(m);
686 goto done;
687 }
688
689 /*
690 * Create a chain of fragments which fit the given mtu. m_frag points to the
691 * mbuf to be fragmented; on return it points to the chain with the fragments.
692 * Return 0 if no error. If error, m_frag may contain a partially built
693 * chain of fragments that should be freed by the caller.
694 *
695 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
696 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
697 */
698 int
699 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
700 u_long if_hwassist_flags, int sw_csum)
701 {
702 int error = 0;
703 int hlen = ip->ip_hl << 2;
704 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
705 int off;
706 struct mbuf *m0 = *m_frag; /* the original packet */
707 int firstlen;
708 struct mbuf **mnext;
709 int nfrags;
710
711 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
712 IPSTAT_INC(ips_cantfrag);
713 return EMSGSIZE;
714 }
715
716 /*
717 * Must be able to put at least 8 bytes per fragment.
718 */
719 if (len < 8)
720 return EMSGSIZE;
721
722 /*
723 * If the interface will not calculate checksums on
724 * fragmented packets, then do it here.
725 */
726 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
727 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
728 in_delayed_cksum(m0);
729 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
730 }
731 #ifdef SCTP
732 if (m0->m_pkthdr.csum_flags & CSUM_SCTP &&
733 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
734 sctp_delayed_cksum(m0, hlen);
735 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
736 }
737 #endif
738 if (len > PAGE_SIZE) {
739 /*
740 * Fragment large datagrams such that each segment
741 * contains a multiple of PAGE_SIZE amount of data,
742 * plus headers. This enables a receiver to perform
743 * page-flipping zero-copy optimizations.
744 *
745 * XXX When does this help given that sender and receiver
746 * could have different page sizes, and also mtu could
747 * be less than the receiver's page size ?
748 */
749 int newlen;
750 struct mbuf *m;
751
752 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
753 off += m->m_len;
754
755 /*
756 * firstlen (off - hlen) must be aligned on an
757 * 8-byte boundary
758 */
759 if (off < hlen)
760 goto smart_frag_failure;
761 off = ((off - hlen) & ~7) + hlen;
762 newlen = (~PAGE_MASK) & mtu;
763 if ((newlen + sizeof (struct ip)) > mtu) {
764 /* we failed, go back the default */
765 smart_frag_failure:
766 newlen = len;
767 off = hlen + len;
768 }
769 len = newlen;
770
771 } else {
772 off = hlen + len;
773 }
774
775 firstlen = off - hlen;
776 mnext = &m0->m_nextpkt; /* pointer to next packet */
777
778 /*
779 * Loop through length of segment after first fragment,
780 * make new header and copy data of each part and link onto chain.
781 * Here, m0 is the original packet, m is the fragment being created.
782 * The fragments are linked off the m_nextpkt of the original
783 * packet, which after processing serves as the first fragment.
784 */
785 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
786 struct ip *mhip; /* ip header on the fragment */
787 struct mbuf *m;
788 int mhlen = sizeof (struct ip);
789
790 MGETHDR(m, M_DONTWAIT, MT_DATA);
791 if (m == NULL) {
792 error = ENOBUFS;
793 IPSTAT_INC(ips_odropped);
794 goto done;
795 }
796 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
797 /*
798 * In the first mbuf, leave room for the link header, then
799 * copy the original IP header including options. The payload
800 * goes into an additional mbuf chain returned by m_copym().
801 */
802 m->m_data += max_linkhdr;
803 mhip = mtod(m, struct ip *);
804 *mhip = *ip;
805 if (hlen > sizeof (struct ip)) {
806 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
807 mhip->ip_v = IPVERSION;
808 mhip->ip_hl = mhlen >> 2;
809 }
810 m->m_len = mhlen;
811 /* XXX do we need to add ip->ip_off below ? */
812 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
813 if (off + len >= ip->ip_len) { /* last fragment */
814 len = ip->ip_len - off;
815 m->m_flags |= M_LASTFRAG;
816 } else
817 mhip->ip_off |= IP_MF;
818 mhip->ip_len = htons((u_short)(len + mhlen));
819 m->m_next = m_copym(m0, off, len, M_DONTWAIT);
820 if (m->m_next == NULL) { /* copy failed */
821 m_free(m);
822 error = ENOBUFS; /* ??? */
823 IPSTAT_INC(ips_odropped);
824 goto done;
825 }
826 m->m_pkthdr.len = mhlen + len;
827 m->m_pkthdr.rcvif = NULL;
828 #ifdef MAC
829 mac_netinet_fragment(m0, m);
830 #endif
831 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
832 mhip->ip_off = htons(mhip->ip_off);
833 mhip->ip_sum = 0;
834 if (sw_csum & CSUM_DELAY_IP)
835 mhip->ip_sum = in_cksum(m, mhlen);
836 *mnext = m;
837 mnext = &m->m_nextpkt;
838 }
839 IPSTAT_ADD(ips_ofragments, nfrags);
840
841 /* set first marker for fragment chain */
842 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
843 m0->m_pkthdr.csum_data = nfrags;
844
845 /*
846 * Update first fragment by trimming what's been copied out
847 * and updating header.
848 */
849 m_adj(m0, hlen + firstlen - ip->ip_len);
850 m0->m_pkthdr.len = hlen + firstlen;
851 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
852 ip->ip_off |= IP_MF;
853 ip->ip_off = htons(ip->ip_off);
854 ip->ip_sum = 0;
855 if (sw_csum & CSUM_DELAY_IP)
856 ip->ip_sum = in_cksum(m0, hlen);
857
858 done:
859 *m_frag = m0;
860 return error;
861 }
862
863 void
864 in_delayed_cksum(struct mbuf *m)
865 {
866 struct ip *ip;
867 u_short csum, offset;
868
869 ip = mtod(m, struct ip *);
870 offset = ip->ip_hl << 2 ;
871 csum = in_cksum_skip(m, ip->ip_len, offset);
872 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
873 csum = 0xffff;
874 offset += m->m_pkthdr.csum_data; /* checksum offset */
875
876 if (offset + sizeof(u_short) > m->m_len) {
877 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
878 m->m_len, offset, ip->ip_p);
879 /*
880 * XXX
881 * this shouldn't happen, but if it does, the
882 * correct behavior may be to insert the checksum
883 * in the appropriate next mbuf in the chain.
884 */
885 return;
886 }
887 *(u_short *)(m->m_data + offset) = csum;
888 }
889
890 /*
891 * IP socket option processing.
892 */
893 int
894 ip_ctloutput(struct socket *so, struct sockopt *sopt)
895 {
896 struct inpcb *inp = sotoinpcb(so);
897 int error, optval;
898
899 error = optval = 0;
900 if (sopt->sopt_level != IPPROTO_IP) {
901 if ((sopt->sopt_level == SOL_SOCKET) &&
902 (sopt->sopt_name == SO_SETFIB)) {
903 inp->inp_inc.inc_fibnum = so->so_fibnum;
904 return (0);
905 }
906 return (EINVAL);
907 }
908
909 switch (sopt->sopt_dir) {
910 case SOPT_SET:
911 switch (sopt->sopt_name) {
912 case IP_OPTIONS:
913 #ifdef notyet
914 case IP_RETOPTS:
915 #endif
916 {
917 struct mbuf *m;
918 if (sopt->sopt_valsize > MLEN) {
919 error = EMSGSIZE;
920 break;
921 }
922 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
923 if (m == NULL) {
924 error = ENOBUFS;
925 break;
926 }
927 m->m_len = sopt->sopt_valsize;
928 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
929 m->m_len);
930 if (error) {
931 m_free(m);
932 break;
933 }
934 INP_WLOCK(inp);
935 error = ip_pcbopts(inp, sopt->sopt_name, m);
936 INP_WUNLOCK(inp);
937 return (error);
938 }
939
940 case IP_BINDANY:
941 if (sopt->sopt_td != NULL) {
942 error = priv_check(sopt->sopt_td,
943 PRIV_NETINET_BINDANY);
944 if (error)
945 break;
946 }
947 /* FALLTHROUGH */
948 case IP_TOS:
949 case IP_TTL:
950 case IP_MINTTL:
951 case IP_RECVOPTS:
952 case IP_RECVRETOPTS:
953 case IP_RECVDSTADDR:
954 case IP_RECVTTL:
955 case IP_RECVIF:
956 case IP_FAITH:
957 case IP_ONESBCAST:
958 case IP_DONTFRAG:
959 error = sooptcopyin(sopt, &optval, sizeof optval,
960 sizeof optval);
961 if (error)
962 break;
963
964 switch (sopt->sopt_name) {
965 case IP_TOS:
966 inp->inp_ip_tos = optval;
967 break;
968
969 case IP_TTL:
970 inp->inp_ip_ttl = optval;
971 break;
972
973 case IP_MINTTL:
974 if (optval >= 0 && optval <= MAXTTL)
975 inp->inp_ip_minttl = optval;
976 else
977 error = EINVAL;
978 break;
979
980 #define OPTSET(bit) do { \
981 INP_WLOCK(inp); \
982 if (optval) \
983 inp->inp_flags |= bit; \
984 else \
985 inp->inp_flags &= ~bit; \
986 INP_WUNLOCK(inp); \
987 } while (0)
988
989 case IP_RECVOPTS:
990 OPTSET(INP_RECVOPTS);
991 break;
992
993 case IP_RECVRETOPTS:
994 OPTSET(INP_RECVRETOPTS);
995 break;
996
997 case IP_RECVDSTADDR:
998 OPTSET(INP_RECVDSTADDR);
999 break;
1000
1001 case IP_RECVTTL:
1002 OPTSET(INP_RECVTTL);
1003 break;
1004
1005 case IP_RECVIF:
1006 OPTSET(INP_RECVIF);
1007 break;
1008
1009 case IP_FAITH:
1010 OPTSET(INP_FAITH);
1011 break;
1012
1013 case IP_ONESBCAST:
1014 OPTSET(INP_ONESBCAST);
1015 break;
1016 case IP_DONTFRAG:
1017 OPTSET(INP_DONTFRAG);
1018 break;
1019 case IP_BINDANY:
1020 OPTSET(INP_BINDANY);
1021 break;
1022 }
1023 break;
1024 #undef OPTSET
1025
1026 /*
1027 * Multicast socket options are processed by the in_mcast
1028 * module.
1029 */
1030 case IP_MULTICAST_IF:
1031 case IP_MULTICAST_VIF:
1032 case IP_MULTICAST_TTL:
1033 case IP_MULTICAST_LOOP:
1034 case IP_ADD_MEMBERSHIP:
1035 case IP_DROP_MEMBERSHIP:
1036 case IP_ADD_SOURCE_MEMBERSHIP:
1037 case IP_DROP_SOURCE_MEMBERSHIP:
1038 case IP_BLOCK_SOURCE:
1039 case IP_UNBLOCK_SOURCE:
1040 case IP_MSFILTER:
1041 case MCAST_JOIN_GROUP:
1042 case MCAST_LEAVE_GROUP:
1043 case MCAST_JOIN_SOURCE_GROUP:
1044 case MCAST_LEAVE_SOURCE_GROUP:
1045 case MCAST_BLOCK_SOURCE:
1046 case MCAST_UNBLOCK_SOURCE:
1047 error = inp_setmoptions(inp, sopt);
1048 break;
1049
1050 case IP_PORTRANGE:
1051 error = sooptcopyin(sopt, &optval, sizeof optval,
1052 sizeof optval);
1053 if (error)
1054 break;
1055
1056 INP_WLOCK(inp);
1057 switch (optval) {
1058 case IP_PORTRANGE_DEFAULT:
1059 inp->inp_flags &= ~(INP_LOWPORT);
1060 inp->inp_flags &= ~(INP_HIGHPORT);
1061 break;
1062
1063 case IP_PORTRANGE_HIGH:
1064 inp->inp_flags &= ~(INP_LOWPORT);
1065 inp->inp_flags |= INP_HIGHPORT;
1066 break;
1067
1068 case IP_PORTRANGE_LOW:
1069 inp->inp_flags &= ~(INP_HIGHPORT);
1070 inp->inp_flags |= INP_LOWPORT;
1071 break;
1072
1073 default:
1074 error = EINVAL;
1075 break;
1076 }
1077 INP_WUNLOCK(inp);
1078 break;
1079
1080 #ifdef IPSEC
1081 case IP_IPSEC_POLICY:
1082 {
1083 caddr_t req;
1084 struct mbuf *m;
1085
1086 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1087 break;
1088 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1089 break;
1090 req = mtod(m, caddr_t);
1091 error = ipsec_set_policy(inp, sopt->sopt_name, req,
1092 m->m_len, (sopt->sopt_td != NULL) ?
1093 sopt->sopt_td->td_ucred : NULL);
1094 m_freem(m);
1095 break;
1096 }
1097 #endif /* IPSEC */
1098
1099 default:
1100 error = ENOPROTOOPT;
1101 break;
1102 }
1103 break;
1104
1105 case SOPT_GET:
1106 switch (sopt->sopt_name) {
1107 case IP_OPTIONS:
1108 case IP_RETOPTS:
1109 if (inp->inp_options)
1110 error = sooptcopyout(sopt,
1111 mtod(inp->inp_options,
1112 char *),
1113 inp->inp_options->m_len);
1114 else
1115 sopt->sopt_valsize = 0;
1116 break;
1117
1118 case IP_TOS:
1119 case IP_TTL:
1120 case IP_MINTTL:
1121 case IP_RECVOPTS:
1122 case IP_RECVRETOPTS:
1123 case IP_RECVDSTADDR:
1124 case IP_RECVTTL:
1125 case IP_RECVIF:
1126 case IP_PORTRANGE:
1127 case IP_FAITH:
1128 case IP_ONESBCAST:
1129 case IP_DONTFRAG:
1130 case IP_BINDANY:
1131 switch (sopt->sopt_name) {
1132
1133 case IP_TOS:
1134 optval = inp->inp_ip_tos;
1135 break;
1136
1137 case IP_TTL:
1138 optval = inp->inp_ip_ttl;
1139 break;
1140
1141 case IP_MINTTL:
1142 optval = inp->inp_ip_minttl;
1143 break;
1144
1145 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1146
1147 case IP_RECVOPTS:
1148 optval = OPTBIT(INP_RECVOPTS);
1149 break;
1150
1151 case IP_RECVRETOPTS:
1152 optval = OPTBIT(INP_RECVRETOPTS);
1153 break;
1154
1155 case IP_RECVDSTADDR:
1156 optval = OPTBIT(INP_RECVDSTADDR);
1157 break;
1158
1159 case IP_RECVTTL:
1160 optval = OPTBIT(INP_RECVTTL);
1161 break;
1162
1163 case IP_RECVIF:
1164 optval = OPTBIT(INP_RECVIF);
1165 break;
1166
1167 case IP_PORTRANGE:
1168 if (inp->inp_flags & INP_HIGHPORT)
1169 optval = IP_PORTRANGE_HIGH;
1170 else if (inp->inp_flags & INP_LOWPORT)
1171 optval = IP_PORTRANGE_LOW;
1172 else
1173 optval = 0;
1174 break;
1175
1176 case IP_FAITH:
1177 optval = OPTBIT(INP_FAITH);
1178 break;
1179
1180 case IP_ONESBCAST:
1181 optval = OPTBIT(INP_ONESBCAST);
1182 break;
1183 case IP_DONTFRAG:
1184 optval = OPTBIT(INP_DONTFRAG);
1185 break;
1186 case IP_BINDANY:
1187 optval = OPTBIT(INP_BINDANY);
1188 break;
1189 }
1190 error = sooptcopyout(sopt, &optval, sizeof optval);
1191 break;
1192
1193 /*
1194 * Multicast socket options are processed by the in_mcast
1195 * module.
1196 */
1197 case IP_MULTICAST_IF:
1198 case IP_MULTICAST_VIF:
1199 case IP_MULTICAST_TTL:
1200 case IP_MULTICAST_LOOP:
1201 case IP_MSFILTER:
1202 error = inp_getmoptions(inp, sopt);
1203 break;
1204
1205 #ifdef IPSEC
1206 case IP_IPSEC_POLICY:
1207 {
1208 struct mbuf *m = NULL;
1209 caddr_t req = NULL;
1210 size_t len = 0;
1211
1212 if (m != 0) {
1213 req = mtod(m, caddr_t);
1214 len = m->m_len;
1215 }
1216 error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1217 if (error == 0)
1218 error = soopt_mcopyout(sopt, m); /* XXX */
1219 if (error == 0)
1220 m_freem(m);
1221 break;
1222 }
1223 #endif /* IPSEC */
1224
1225 default:
1226 error = ENOPROTOOPT;
1227 break;
1228 }
1229 break;
1230 }
1231 return (error);
1232 }
1233
1234 /*
1235 * Routine called from ip_output() to loop back a copy of an IP multicast
1236 * packet to the input queue of a specified interface. Note that this
1237 * calls the output routine of the loopback "driver", but with an interface
1238 * pointer that might NOT be a loopback interface -- evil, but easier than
1239 * replicating that code here.
1240 */
1241 static void
1242 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1243 int hlen)
1244 {
1245 register struct ip *ip;
1246 struct mbuf *copym;
1247
1248 /*
1249 * Make a deep copy of the packet because we're going to
1250 * modify the pack in order to generate checksums.
1251 */
1252 copym = m_dup(m, M_DONTWAIT);
1253 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1254 copym = m_pullup(copym, hlen);
1255 if (copym != NULL) {
1256 /* If needed, compute the checksum and mark it as valid. */
1257 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1258 in_delayed_cksum(copym);
1259 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1260 copym->m_pkthdr.csum_flags |=
1261 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1262 copym->m_pkthdr.csum_data = 0xffff;
1263 }
1264 /*
1265 * We don't bother to fragment if the IP length is greater
1266 * than the interface's MTU. Can this possibly matter?
1267 */
1268 ip = mtod(copym, struct ip *);
1269 ip->ip_len = htons(ip->ip_len);
1270 ip->ip_off = htons(ip->ip_off);
1271 ip->ip_sum = 0;
1272 ip->ip_sum = in_cksum(copym, hlen);
1273 #if 1 /* XXX */
1274 if (dst->sin_family != AF_INET) {
1275 printf("ip_mloopback: bad address family %d\n",
1276 dst->sin_family);
1277 dst->sin_family = AF_INET;
1278 }
1279 #endif
1280 if_simloop(ifp, copym, dst->sin_family, 0);
1281 }
1282 }
Cache object: f369fb7804c7c93ce7c40d083eb06fa6
|