FreeBSD/Linux Kernel Cross Reference
sys/netinet/raw_ip.c
1 /*-
2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California.
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 4. Neither the name of the University nor the names of its contributors
15 * may be used to endorse or promote products derived from this software
16 * without specific prior written permission.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 * SUCH DAMAGE.
29 *
30 * @(#)raw_ip.c 8.7 (Berkeley) 5/15/95
31 */
32
33 #include <sys/cdefs.h>
34 __FBSDID("$FreeBSD$");
35
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38 #include "opt_mac.h"
39
40 #include <sys/param.h>
41 #include <sys/jail.h>
42 #include <sys/kernel.h>
43 #include <sys/lock.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/priv.h>
47 #include <sys/proc.h>
48 #include <sys/protosw.h>
49 #include <sys/signalvar.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sx.h>
53 #include <sys/sysctl.h>
54 #include <sys/systm.h>
55
56 #include <vm/uma.h>
57
58 #include <net/if.h>
59 #include <net/route.h>
60
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/in_pcb.h>
64 #include <netinet/in_var.h>
65 #include <netinet/ip.h>
66 #include <netinet/ip_var.h>
67 #include <netinet/ip_mroute.h>
68
69 #include <netinet/ip_fw.h>
70 #include <netinet/ip_dummynet.h>
71
72 #ifdef IPSEC
73 #include <netipsec/ipsec.h>
74 #endif /*IPSEC*/
75
76 #include <security/mac/mac_framework.h>
77
78 struct inpcbhead ripcb;
79 struct inpcbinfo ripcbinfo;
80
81 /* control hooks for ipfw and dummynet */
82 ip_fw_ctl_t *ip_fw_ctl_ptr = NULL;
83 ip_dn_ctl_t *ip_dn_ctl_ptr = NULL;
84
85 /*
86 * Hooks for multicast routing. They all default to NULL, so leave them not
87 * initialized and rely on BSS being set to 0.
88 */
89
90 /*
91 * The socket used to communicate with the multicast routing daemon.
92 */
93 struct socket *ip_mrouter;
94
95 /*
96 * The various mrouter and rsvp functions.
97 */
98 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
99 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
100 int (*ip_mrouter_done)(void);
101 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
102 struct ip_moptions *);
103 int (*mrt_ioctl)(int, caddr_t, int);
104 int (*legal_vif_num)(int);
105 u_long (*ip_mcast_src)(int);
106
107 void (*rsvp_input_p)(struct mbuf *m, int off);
108 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
109 void (*ip_rsvp_force_done)(struct socket *);
110
111 /*
112 * Hash functions
113 */
114
115 #define INP_PCBHASH_RAW_SIZE 256
116 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \
117 (((proto) + (laddr) + (faddr)) % (mask) + 1)
118
119 static void
120 rip_inshash(struct inpcb *inp)
121 {
122 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
123 struct inpcbhead *pcbhash;
124 int hash;
125
126 INP_INFO_WLOCK_ASSERT(pcbinfo);
127 INP_WLOCK_ASSERT(inp);
128
129 if (inp->inp_ip_p != 0 &&
130 inp->inp_laddr.s_addr != INADDR_ANY &&
131 inp->inp_faddr.s_addr != INADDR_ANY) {
132 hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr,
133 inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask);
134 } else
135 hash = 0;
136 pcbhash = &pcbinfo->ipi_hashbase[hash];
137 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
138 }
139
140 static void
141 rip_delhash(struct inpcb *inp)
142 {
143
144 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
145 INP_WLOCK_ASSERT(inp);
146
147 LIST_REMOVE(inp, inp_hash);
148 }
149
150 /*
151 * Raw interface to IP protocol.
152 */
153
154 /*
155 * Initialize raw connection block q.
156 */
157 static void
158 rip_zone_change(void *tag)
159 {
160
161 uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets);
162 }
163
164 static int
165 rip_inpcb_init(void *mem, int size, int flags)
166 {
167 struct inpcb *inp = mem;
168
169 INP_LOCK_INIT(inp, "inp", "rawinp");
170 return (0);
171 }
172
173 void
174 rip_init(void)
175 {
176
177 INP_INFO_LOCK_INIT(&ripcbinfo, "rip");
178 LIST_INIT(&ripcb);
179 ripcbinfo.ipi_listhead = &ripcb;
180 ripcbinfo.ipi_hashbase = hashinit(INP_PCBHASH_RAW_SIZE, M_PCB,
181 &ripcbinfo.ipi_hashmask);
182 ripcbinfo.ipi_porthashbase = hashinit(1, M_PCB,
183 &ripcbinfo.ipi_porthashmask);
184 ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb),
185 NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
186 uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets);
187 EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL,
188 EVENTHANDLER_PRI_ANY);
189 }
190
191 static int
192 rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
193 struct sockaddr_in *ripsrc)
194 {
195 int policyfail = 0;
196
197 INP_RLOCK_ASSERT(last);
198
199 #ifdef IPSEC
200 /* check AH/ESP integrity. */
201 if (ipsec4_in_reject(n, last)) {
202 policyfail = 1;
203 }
204 #endif /* IPSEC */
205 #ifdef MAC
206 if (!policyfail && mac_check_inpcb_deliver(last, n) != 0)
207 policyfail = 1;
208 #endif
209 /* Check the minimum TTL for socket. */
210 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
211 policyfail = 1;
212 if (!policyfail) {
213 struct mbuf *opts = NULL;
214 struct socket *so;
215
216 so = last->inp_socket;
217 if ((last->inp_flags & INP_CONTROLOPTS) ||
218 (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
219 ip_savecontrol(last, &opts, ip, n);
220 SOCKBUF_LOCK(&so->so_rcv);
221 if (sbappendaddr_locked(&so->so_rcv,
222 (struct sockaddr *)ripsrc, n, opts) == 0) {
223 /* should notify about lost packet */
224 m_freem(n);
225 if (opts)
226 m_freem(opts);
227 SOCKBUF_UNLOCK(&so->so_rcv);
228 } else
229 sorwakeup_locked(so);
230 } else
231 m_freem(n);
232 return (policyfail);
233 }
234
235 /*
236 * Setup generic address and protocol structures for raw_input routine, then
237 * pass them along with mbuf chain.
238 */
239 void
240 rip_input(struct mbuf *m, int off)
241 {
242 struct ip *ip = mtod(m, struct ip *);
243 int proto = ip->ip_p;
244 struct inpcb *inp, *last;
245 struct sockaddr_in ripsrc;
246 int hash;
247
248 bzero(&ripsrc, sizeof(ripsrc));
249 ripsrc.sin_len = sizeof(ripsrc);
250 ripsrc.sin_family = AF_INET;
251 ripsrc.sin_addr = ip->ip_src;
252 last = NULL;
253 hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr,
254 ip->ip_dst.s_addr, ripcbinfo.ipi_hashmask);
255 INP_INFO_RLOCK(&ripcbinfo);
256 LIST_FOREACH(inp, &ripcbinfo.ipi_hashbase[hash], inp_hash) {
257 if (inp->inp_ip_p != proto)
258 continue;
259 #ifdef INET6
260 /* XXX inp locking */
261 if ((inp->inp_vflag & INP_IPV4) == 0)
262 continue;
263 #endif
264 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
265 continue;
266 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr)
267 continue;
268 if (jailed(inp->inp_cred) &&
269 (htonl(prison_getip(inp->inp_cred)) !=
270 ip->ip_dst.s_addr)) {
271 continue;
272 }
273 if (last) {
274 struct mbuf *n;
275
276 n = m_copy(m, 0, (int)M_COPYALL);
277 if (n != NULL)
278 (void) rip_append(last, ip, n, &ripsrc);
279 /* XXX count dropped packet */
280 INP_RUNLOCK(last);
281 }
282 INP_RLOCK(inp);
283 last = inp;
284 }
285 LIST_FOREACH(inp, &ripcbinfo.ipi_hashbase[0], inp_hash) {
286 if (inp->inp_ip_p && inp->inp_ip_p != proto)
287 continue;
288 #ifdef INET6
289 /* XXX inp locking */
290 if ((inp->inp_vflag & INP_IPV4) == 0)
291 continue;
292 #endif
293 if (inp->inp_laddr.s_addr &&
294 inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
295 continue;
296 if (inp->inp_faddr.s_addr &&
297 inp->inp_faddr.s_addr != ip->ip_src.s_addr)
298 continue;
299 if (jailed(inp->inp_cred) &&
300 (htonl(prison_getip(inp->inp_cred)) !=
301 ip->ip_dst.s_addr)) {
302 continue;
303 }
304 if (last) {
305 struct mbuf *n;
306
307 n = m_copy(m, 0, (int)M_COPYALL);
308 if (n != NULL)
309 (void) rip_append(last, ip, n, &ripsrc);
310 /* XXX count dropped packet */
311 INP_RUNLOCK(last);
312 }
313 INP_RLOCK(inp);
314 last = inp;
315 }
316 INP_INFO_RUNLOCK(&ripcbinfo);
317 if (last != NULL) {
318 if (rip_append(last, ip, m, &ripsrc) != 0)
319 ipstat.ips_delivered--;
320 INP_RUNLOCK(last);
321 } else {
322 m_freem(m);
323 ipstat.ips_noproto++;
324 ipstat.ips_delivered--;
325 }
326 }
327
328 /*
329 * Generate IP header and pass packet to ip_output. Tack on options user may
330 * have setup with control call.
331 */
332 int
333 rip_output(struct mbuf *m, struct socket *so, u_long dst)
334 {
335 struct ip *ip;
336 int error;
337 struct inpcb *inp = sotoinpcb(so);
338 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
339 IP_ALLOWBROADCAST;
340
341 /*
342 * If the user handed us a complete IP packet, use it. Otherwise,
343 * allocate an mbuf for a header and fill it in.
344 */
345 if ((inp->inp_flags & INP_HDRINCL) == 0) {
346 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
347 m_freem(m);
348 return(EMSGSIZE);
349 }
350 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
351 if (m == NULL)
352 return(ENOBUFS);
353
354 INP_RLOCK(inp);
355 ip = mtod(m, struct ip *);
356 ip->ip_tos = inp->inp_ip_tos;
357 if (inp->inp_flags & INP_DONTFRAG)
358 ip->ip_off = IP_DF;
359 else
360 ip->ip_off = 0;
361 ip->ip_p = inp->inp_ip_p;
362 ip->ip_len = m->m_pkthdr.len;
363 if (jailed(inp->inp_cred))
364 ip->ip_src.s_addr =
365 htonl(prison_getip(inp->inp_cred));
366 else
367 ip->ip_src = inp->inp_laddr;
368 ip->ip_dst.s_addr = dst;
369 ip->ip_ttl = inp->inp_ip_ttl;
370 } else {
371 if (m->m_pkthdr.len > IP_MAXPACKET) {
372 m_freem(m);
373 return(EMSGSIZE);
374 }
375 INP_RLOCK(inp);
376 ip = mtod(m, struct ip *);
377 if (jailed(inp->inp_cred)) {
378 if (ip->ip_src.s_addr !=
379 htonl(prison_getip(inp->inp_cred))) {
380 INP_RUNLOCK(inp);
381 m_freem(m);
382 return (EPERM);
383 }
384 }
385
386 /*
387 * Don't allow both user specified and setsockopt options,
388 * and don't allow packet length sizes that will crash.
389 */
390 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options)
391 || (ip->ip_len > m->m_pkthdr.len)
392 || (ip->ip_len < (ip->ip_hl << 2))) {
393 INP_RUNLOCK(inp);
394 m_freem(m);
395 return (EINVAL);
396 }
397 if (ip->ip_id == 0)
398 ip->ip_id = ip_newid();
399
400 /*
401 * XXX prevent ip_output from overwriting header fields.
402 */
403 flags |= IP_RAWOUTPUT;
404 ipstat.ips_rawout++;
405 }
406
407 if (inp->inp_flags & INP_ONESBCAST)
408 flags |= IP_SENDONES;
409
410 #ifdef MAC
411 mac_create_mbuf_from_inpcb(inp, m);
412 #endif
413
414 error = ip_output(m, inp->inp_options, NULL, flags,
415 inp->inp_moptions, inp);
416 INP_RUNLOCK(inp);
417 return (error);
418 }
419
420 /*
421 * Raw IP socket option processing.
422 *
423 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
424 * only be created by a privileged process, and as such, socket option
425 * operations to manage system properties on any raw socket were allowed to
426 * take place without explicit additional access control checks. However,
427 * raw sockets can now also be created in jail(), and therefore explicit
428 * checks are now required. Likewise, raw sockets can be used by a process
429 * after it gives up privilege, so some caution is required. For options
430 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
431 * performed in ip_ctloutput() and therefore no check occurs here.
432 * Unilaterally checking priv_check() here breaks normal IP socket option
433 * operations on raw sockets.
434 *
435 * When adding new socket options here, make sure to add access control
436 * checks here as necessary.
437 */
438 int
439 rip_ctloutput(struct socket *so, struct sockopt *sopt)
440 {
441 struct inpcb *inp = sotoinpcb(so);
442 int error, optval;
443
444 if (sopt->sopt_level != IPPROTO_IP) {
445 if ((sopt->sopt_level == SOL_SOCKET) &&
446 (sopt->sopt_name == SO_SETFIB)) {
447 inp->inp_inc.inc_fibnum = so->so_fibnum;
448 return (0);
449 }
450 return (EINVAL);
451 }
452
453 error = 0;
454 switch (sopt->sopt_dir) {
455 case SOPT_GET:
456 switch (sopt->sopt_name) {
457 case IP_HDRINCL:
458 optval = inp->inp_flags & INP_HDRINCL;
459 error = sooptcopyout(sopt, &optval, sizeof optval);
460 break;
461
462 case IP_FW_ADD: /* ADD actually returns the body... */
463 case IP_FW_GET:
464 case IP_FW_TABLE_GETSIZE:
465 case IP_FW_TABLE_LIST:
466 case IP_FW_NAT_GET_CONFIG:
467 case IP_FW_NAT_GET_LOG:
468 if (ip_fw_ctl_ptr != NULL)
469 error = ip_fw_ctl_ptr(sopt);
470 else
471 error = ENOPROTOOPT;
472 break;
473
474 case IP_DUMMYNET_GET:
475 if (ip_dn_ctl_ptr != NULL)
476 error = ip_dn_ctl_ptr(sopt);
477 else
478 error = ENOPROTOOPT;
479 break ;
480
481 case MRT_INIT:
482 case MRT_DONE:
483 case MRT_ADD_VIF:
484 case MRT_DEL_VIF:
485 case MRT_ADD_MFC:
486 case MRT_DEL_MFC:
487 case MRT_VERSION:
488 case MRT_ASSERT:
489 case MRT_API_SUPPORT:
490 case MRT_API_CONFIG:
491 case MRT_ADD_BW_UPCALL:
492 case MRT_DEL_BW_UPCALL:
493 error = priv_check(curthread, PRIV_NETINET_MROUTE);
494 if (error != 0)
495 return (error);
496 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
497 EOPNOTSUPP;
498 break;
499
500 default:
501 error = ip_ctloutput(so, sopt);
502 break;
503 }
504 break;
505
506 case SOPT_SET:
507 switch (sopt->sopt_name) {
508 case IP_HDRINCL:
509 error = sooptcopyin(sopt, &optval, sizeof optval,
510 sizeof optval);
511 if (error)
512 break;
513 if (optval)
514 inp->inp_flags |= INP_HDRINCL;
515 else
516 inp->inp_flags &= ~INP_HDRINCL;
517 break;
518
519 case IP_FW_ADD:
520 case IP_FW_DEL:
521 case IP_FW_FLUSH:
522 case IP_FW_ZERO:
523 case IP_FW_RESETLOG:
524 case IP_FW_TABLE_ADD:
525 case IP_FW_TABLE_DEL:
526 case IP_FW_TABLE_FLUSH:
527 case IP_FW_NAT_CFG:
528 case IP_FW_NAT_DEL:
529 if (ip_fw_ctl_ptr != NULL)
530 error = ip_fw_ctl_ptr(sopt);
531 else
532 error = ENOPROTOOPT;
533 break;
534
535 case IP_DUMMYNET_CONFIGURE:
536 case IP_DUMMYNET_DEL:
537 case IP_DUMMYNET_FLUSH:
538 if (ip_dn_ctl_ptr != NULL)
539 error = ip_dn_ctl_ptr(sopt);
540 else
541 error = ENOPROTOOPT ;
542 break ;
543
544 case IP_RSVP_ON:
545 error = priv_check(curthread, PRIV_NETINET_MROUTE);
546 if (error != 0)
547 return (error);
548 error = ip_rsvp_init(so);
549 break;
550
551 case IP_RSVP_OFF:
552 error = priv_check(curthread, PRIV_NETINET_MROUTE);
553 if (error != 0)
554 return (error);
555 error = ip_rsvp_done();
556 break;
557
558 case IP_RSVP_VIF_ON:
559 case IP_RSVP_VIF_OFF:
560 error = priv_check(curthread, PRIV_NETINET_MROUTE);
561 if (error != 0)
562 return (error);
563 error = ip_rsvp_vif ?
564 ip_rsvp_vif(so, sopt) : EINVAL;
565 break;
566
567 case MRT_INIT:
568 case MRT_DONE:
569 case MRT_ADD_VIF:
570 case MRT_DEL_VIF:
571 case MRT_ADD_MFC:
572 case MRT_DEL_MFC:
573 case MRT_VERSION:
574 case MRT_ASSERT:
575 case MRT_API_SUPPORT:
576 case MRT_API_CONFIG:
577 case MRT_ADD_BW_UPCALL:
578 case MRT_DEL_BW_UPCALL:
579 error = priv_check(curthread, PRIV_NETINET_MROUTE);
580 if (error != 0)
581 return (error);
582 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
583 EOPNOTSUPP;
584 break;
585
586 default:
587 error = ip_ctloutput(so, sopt);
588 break;
589 }
590 break;
591 }
592
593 return (error);
594 }
595
596 /*
597 * This function exists solely to receive the PRC_IFDOWN messages which are
598 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls
599 * in_ifadown() to remove all routes corresponding to that address. It also
600 * receives the PRC_IFUP messages from if_up() and reinstalls the interface
601 * routes.
602 */
603 void
604 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
605 {
606 struct in_ifaddr *ia;
607 struct ifnet *ifp;
608 int err;
609 int flags;
610
611 switch (cmd) {
612 case PRC_IFDOWN:
613 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
614 if (ia->ia_ifa.ifa_addr == sa
615 && (ia->ia_flags & IFA_ROUTE)) {
616 /*
617 * in_ifscrub kills the interface route.
618 */
619 in_ifscrub(ia->ia_ifp, ia);
620 /*
621 * in_ifadown gets rid of all the rest of the
622 * routes. This is not quite the right thing
623 * to do, but at least if we are running a
624 * routing process they will come back.
625 */
626 in_ifadown(&ia->ia_ifa, 0);
627 break;
628 }
629 }
630 break;
631
632 case PRC_IFUP:
633 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
634 if (ia->ia_ifa.ifa_addr == sa)
635 break;
636 }
637 if (ia == 0 || (ia->ia_flags & IFA_ROUTE))
638 return;
639 flags = RTF_UP;
640 ifp = ia->ia_ifa.ifa_ifp;
641
642 if ((ifp->if_flags & IFF_LOOPBACK)
643 || (ifp->if_flags & IFF_POINTOPOINT))
644 flags |= RTF_HOST;
645
646 err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
647 if (err == 0)
648 ia->ia_flags |= IFA_ROUTE;
649 break;
650 }
651 }
652
653 u_long rip_sendspace = 9216;
654 u_long rip_recvspace = 9216;
655
656 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
657 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
658 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
659 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
660
661 static int
662 rip_attach(struct socket *so, int proto, struct thread *td)
663 {
664 struct inpcb *inp;
665 int error;
666
667 inp = sotoinpcb(so);
668 KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
669
670 error = priv_check(td, PRIV_NETINET_RAW);
671 if (error)
672 return (error);
673 if (proto >= IPPROTO_MAX || proto < 0)
674 return EPROTONOSUPPORT;
675 error = soreserve(so, rip_sendspace, rip_recvspace);
676 if (error)
677 return (error);
678 INP_INFO_WLOCK(&ripcbinfo);
679 error = in_pcballoc(so, &ripcbinfo);
680 if (error) {
681 INP_INFO_WUNLOCK(&ripcbinfo);
682 return (error);
683 }
684 inp = (struct inpcb *)so->so_pcb;
685 inp->inp_vflag |= INP_IPV4;
686 inp->inp_ip_p = proto;
687 inp->inp_ip_ttl = ip_defttl;
688 rip_inshash(inp);
689 INP_INFO_WUNLOCK(&ripcbinfo);
690 INP_WUNLOCK(inp);
691 return (0);
692 }
693
694 static void
695 rip_detach(struct socket *so)
696 {
697 struct inpcb *inp;
698
699 inp = sotoinpcb(so);
700 KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
701 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
702 ("rip_detach: not closed"));
703
704 INP_INFO_WLOCK(&ripcbinfo);
705 INP_WLOCK(inp);
706 rip_delhash(inp);
707 if (so == ip_mrouter && ip_mrouter_done)
708 ip_mrouter_done();
709 if (ip_rsvp_force_done)
710 ip_rsvp_force_done(so);
711 if (so == ip_rsvpd)
712 ip_rsvp_done();
713 in_pcbdetach(inp);
714 in_pcbfree(inp);
715 INP_INFO_WUNLOCK(&ripcbinfo);
716 }
717
718 static void
719 rip_dodisconnect(struct socket *so, struct inpcb *inp)
720 {
721
722 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
723 INP_WLOCK_ASSERT(inp);
724
725 rip_delhash(inp);
726 inp->inp_faddr.s_addr = INADDR_ANY;
727 rip_inshash(inp);
728 SOCK_LOCK(so);
729 so->so_state &= ~SS_ISCONNECTED;
730 SOCK_UNLOCK(so);
731 }
732
733 static void
734 rip_abort(struct socket *so)
735 {
736 struct inpcb *inp;
737
738 inp = sotoinpcb(so);
739 KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
740
741 INP_INFO_WLOCK(&ripcbinfo);
742 INP_WLOCK(inp);
743 rip_dodisconnect(so, inp);
744 INP_WUNLOCK(inp);
745 INP_INFO_WUNLOCK(&ripcbinfo);
746 }
747
748 static void
749 rip_close(struct socket *so)
750 {
751 struct inpcb *inp;
752
753 inp = sotoinpcb(so);
754 KASSERT(inp != NULL, ("rip_close: inp == NULL"));
755
756 INP_INFO_WLOCK(&ripcbinfo);
757 INP_WLOCK(inp);
758 rip_dodisconnect(so, inp);
759 INP_WUNLOCK(inp);
760 INP_INFO_WUNLOCK(&ripcbinfo);
761 }
762
763 static int
764 rip_disconnect(struct socket *so)
765 {
766 struct inpcb *inp;
767
768 if ((so->so_state & SS_ISCONNECTED) == 0)
769 return (ENOTCONN);
770
771 inp = sotoinpcb(so);
772 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
773
774 INP_INFO_WLOCK(&ripcbinfo);
775 INP_WLOCK(inp);
776 rip_dodisconnect(so, inp);
777 INP_WUNLOCK(inp);
778 INP_INFO_WUNLOCK(&ripcbinfo);
779 return (0);
780 }
781
782 static int
783 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
784 {
785 struct sockaddr_in *addr = (struct sockaddr_in *)nam;
786 struct inpcb *inp;
787
788 if (nam->sa_len != sizeof(*addr))
789 return (EINVAL);
790
791 if (jailed(td->td_ucred)) {
792 if (addr->sin_addr.s_addr == INADDR_ANY)
793 addr->sin_addr.s_addr =
794 htonl(prison_getip(td->td_ucred));
795 if (htonl(prison_getip(td->td_ucred)) != addr->sin_addr.s_addr)
796 return (EADDRNOTAVAIL);
797 }
798
799 if (TAILQ_EMPTY(&ifnet) ||
800 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
801 (addr->sin_addr.s_addr &&
802 ifa_ifwithaddr((struct sockaddr *)addr) == 0))
803 return (EADDRNOTAVAIL);
804
805 inp = sotoinpcb(so);
806 KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
807
808 INP_INFO_WLOCK(&ripcbinfo);
809 INP_WLOCK(inp);
810 rip_delhash(inp);
811 inp->inp_laddr = addr->sin_addr;
812 rip_inshash(inp);
813 INP_WUNLOCK(inp);
814 INP_INFO_WUNLOCK(&ripcbinfo);
815 return (0);
816 }
817
818 static int
819 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
820 {
821 struct sockaddr_in *addr = (struct sockaddr_in *)nam;
822 struct inpcb *inp;
823
824 if (nam->sa_len != sizeof(*addr))
825 return (EINVAL);
826 if (TAILQ_EMPTY(&ifnet))
827 return (EADDRNOTAVAIL);
828 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
829 return (EAFNOSUPPORT);
830
831 inp = sotoinpcb(so);
832 KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
833
834 INP_INFO_WLOCK(&ripcbinfo);
835 INP_WLOCK(inp);
836 rip_delhash(inp);
837 inp->inp_faddr = addr->sin_addr;
838 rip_inshash(inp);
839 soisconnected(so);
840 INP_WUNLOCK(inp);
841 INP_INFO_WUNLOCK(&ripcbinfo);
842 return (0);
843 }
844
845 static int
846 rip_shutdown(struct socket *so)
847 {
848 struct inpcb *inp;
849
850 inp = sotoinpcb(so);
851 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
852
853 INP_WLOCK(inp);
854 socantsendmore(so);
855 INP_WUNLOCK(inp);
856 return (0);
857 }
858
859 static int
860 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
861 struct mbuf *control, struct thread *td)
862 {
863 struct inpcb *inp;
864 u_long dst;
865
866 inp = sotoinpcb(so);
867 KASSERT(inp != NULL, ("rip_send: inp == NULL"));
868
869 /*
870 * Note: 'dst' reads below are unlocked.
871 */
872 if (so->so_state & SS_ISCONNECTED) {
873 if (nam) {
874 m_freem(m);
875 return (EISCONN);
876 }
877 dst = inp->inp_faddr.s_addr; /* Unlocked read. */
878 } else {
879 if (nam == NULL) {
880 m_freem(m);
881 return (ENOTCONN);
882 }
883 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
884 }
885 return (rip_output(m, so, dst));
886 }
887
888 static int
889 rip_pcblist(SYSCTL_HANDLER_ARGS)
890 {
891 int error, i, n;
892 struct inpcb *inp, **inp_list;
893 inp_gen_t gencnt;
894 struct xinpgen xig;
895
896 /*
897 * The process of preparing the TCB list is too time-consuming and
898 * resource-intensive to repeat twice on every request.
899 */
900 if (req->oldptr == 0) {
901 n = ripcbinfo.ipi_count;
902 req->oldidx = 2 * (sizeof xig)
903 + (n + n/8) * sizeof(struct xinpcb);
904 return (0);
905 }
906
907 if (req->newptr != 0)
908 return (EPERM);
909
910 /*
911 * OK, now we're committed to doing something.
912 */
913 INP_INFO_RLOCK(&ripcbinfo);
914 gencnt = ripcbinfo.ipi_gencnt;
915 n = ripcbinfo.ipi_count;
916 INP_INFO_RUNLOCK(&ripcbinfo);
917
918 xig.xig_len = sizeof xig;
919 xig.xig_count = n;
920 xig.xig_gen = gencnt;
921 xig.xig_sogen = so_gencnt;
922 error = SYSCTL_OUT(req, &xig, sizeof xig);
923 if (error)
924 return (error);
925
926 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
927 if (inp_list == 0)
928 return (ENOMEM);
929
930 INP_INFO_RLOCK(&ripcbinfo);
931 for (inp = LIST_FIRST(ripcbinfo.ipi_listhead), i = 0; inp && i < n;
932 inp = LIST_NEXT(inp, inp_list)) {
933 INP_RLOCK(inp);
934 if (inp->inp_gencnt <= gencnt &&
935 cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
936 /* XXX held references? */
937 inp_list[i++] = inp;
938 }
939 INP_RUNLOCK(inp);
940 }
941 INP_INFO_RUNLOCK(&ripcbinfo);
942 n = i;
943
944 error = 0;
945 for (i = 0; i < n; i++) {
946 inp = inp_list[i];
947 INP_RLOCK(inp);
948 if (inp->inp_gencnt <= gencnt) {
949 struct xinpcb xi;
950 bzero(&xi, sizeof(xi));
951 xi.xi_len = sizeof xi;
952 /* XXX should avoid extra copy */
953 bcopy(inp, &xi.xi_inp, sizeof *inp);
954 if (inp->inp_socket)
955 sotoxsocket(inp->inp_socket, &xi.xi_socket);
956 INP_RUNLOCK(inp);
957 error = SYSCTL_OUT(req, &xi, sizeof xi);
958 } else
959 INP_RUNLOCK(inp);
960 }
961 if (!error) {
962 /*
963 * Give the user an updated idea of our state. If the
964 * generation differs from what we told her before, she knows
965 * that something happened while we were processing this
966 * request, and it might be necessary to retry.
967 */
968 INP_INFO_RLOCK(&ripcbinfo);
969 xig.xig_gen = ripcbinfo.ipi_gencnt;
970 xig.xig_sogen = so_gencnt;
971 xig.xig_count = ripcbinfo.ipi_count;
972 INP_INFO_RUNLOCK(&ripcbinfo);
973 error = SYSCTL_OUT(req, &xig, sizeof xig);
974 }
975 free(inp_list, M_TEMP);
976 return (error);
977 }
978
979 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0,
980 rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
981
982 struct pr_usrreqs rip_usrreqs = {
983 .pru_abort = rip_abort,
984 .pru_attach = rip_attach,
985 .pru_bind = rip_bind,
986 .pru_connect = rip_connect,
987 .pru_control = in_control,
988 .pru_detach = rip_detach,
989 .pru_disconnect = rip_disconnect,
990 .pru_peeraddr = in_getpeeraddr,
991 .pru_send = rip_send,
992 .pru_shutdown = rip_shutdown,
993 .pru_sockaddr = in_getsockaddr,
994 .pru_sosetlabel = in_pcbsosetlabel,
995 .pru_close = rip_close,
996 };
Cache object: 38437977e184a1c3bad1bc0d02e92419
|