Commit | Line | Data |
---|---|---|
984263bc | 1 | /* |
66d6c637 JH |
2 | * Copyright (c) 2003, 2004 Jeffrey M. Hsu. All rights reserved. |
3 | * Copyright (c) 2003, 2004 The DragonFly Project. All rights reserved. | |
f23061d4 | 4 | * |
66d6c637 JH |
5 | * This code is derived from software contributed to The DragonFly Project |
6 | * by Jeffrey M. Hsu. | |
f23061d4 | 7 | * |
66d6c637 JH |
8 | * Redistribution and use in source and binary forms, with or without |
9 | * modification, are permitted provided that the following conditions | |
10 | * are met: | |
11 | * 1. Redistributions of source code must retain the above copyright | |
12 | * notice, this list of conditions and the following disclaimer. | |
13 | * 2. Redistributions in binary form must reproduce the above copyright | |
14 | * notice, this list of conditions and the following disclaimer in the | |
15 | * documentation and/or other materials provided with the distribution. | |
16 | * 3. Neither the name of The DragonFly Project nor the names of its | |
17 | * contributors may be used to endorse or promote products derived | |
18 | * from this software without specific, prior written permission. | |
f23061d4 | 19 | * |
66d6c637 JH |
20 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
21 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | |
22 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS | |
23 | * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE | |
24 | * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, | |
25 | * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, | |
26 | * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; | |
27 | * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED | |
28 | * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, | |
29 | * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT | |
30 | * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | |
31 | * SUCH DAMAGE. | |
32 | */ | |
33 | ||
66d6c637 | 34 | /* |
984263bc MD |
35 | * Copyright (c) 1982, 1986, 1988, 1993 |
36 | * The Regents of the University of California. All rights reserved. | |
37 | * | |
38 | * Redistribution and use in source and binary forms, with or without | |
39 | * modification, are permitted provided that the following conditions | |
40 | * are met: | |
41 | * 1. Redistributions of source code must retain the above copyright | |
42 | * notice, this list of conditions and the following disclaimer. | |
43 | * 2. Redistributions in binary form must reproduce the above copyright | |
44 | * notice, this list of conditions and the following disclaimer in the | |
45 | * documentation and/or other materials provided with the distribution. | |
46 | * 3. All advertising materials mentioning features or use of this software | |
47 | * must display the following acknowledgement: | |
48 | * This product includes software developed by the University of | |
49 | * California, Berkeley and its contributors. | |
50 | * 4. Neither the name of the University nor the names of its contributors | |
51 | * may be used to endorse or promote products derived from this software | |
52 | * without specific prior written permission. | |
53 | * | |
54 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND | |
55 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | |
56 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | |
57 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | |
58 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | |
59 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | |
60 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | |
61 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | |
62 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | |
63 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | |
64 | * SUCH DAMAGE. | |
65 | * | |
66 | * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 | |
67 | * $FreeBSD: src/sys/netinet/tcp_usrreq.c,v 1.51.2.17 2002/10/11 11:46:44 ume Exp $ | |
69891757 | 68 | * $DragonFly: src/sys/netinet/tcp_usrreq.c,v 1.51 2008/09/29 20:52:23 dillon Exp $ |
984263bc MD |
69 | */ |
70 | ||
71 | #include "opt_ipsec.h" | |
72 | #include "opt_inet6.h" | |
73 | #include "opt_tcpdebug.h" | |
74 | ||
75 | #include <sys/param.h> | |
76 | #include <sys/systm.h> | |
77 | #include <sys/kernel.h> | |
7fe56515 | 78 | #include <sys/malloc.h> |
984263bc | 79 | #include <sys/sysctl.h> |
d371a63a JH |
80 | #include <sys/globaldata.h> |
81 | #include <sys/thread.h> | |
82 | ||
984263bc MD |
83 | #include <sys/mbuf.h> |
84 | #ifdef INET6 | |
85 | #include <sys/domain.h> | |
86 | #endif /* INET6 */ | |
87 | #include <sys/socket.h> | |
88 | #include <sys/socketvar.h> | |
89 | #include <sys/protosw.h> | |
90 | ||
1cae611f | 91 | #include <sys/thread2.h> |
7fe56515 JH |
92 | #include <sys/msgport2.h> |
93 | ||
984263bc | 94 | #include <net/if.h> |
7fe56515 | 95 | #include <net/netisr.h> |
984263bc MD |
96 | #include <net/route.h> |
97 | ||
4599cf19 MD |
98 | #include <net/netmsg2.h> |
99 | ||
984263bc MD |
100 | #include <netinet/in.h> |
101 | #include <netinet/in_systm.h> | |
102 | #ifdef INET6 | |
103 | #include <netinet/ip6.h> | |
104 | #endif | |
105 | #include <netinet/in_pcb.h> | |
106 | #ifdef INET6 | |
107 | #include <netinet6/in6_pcb.h> | |
108 | #endif | |
109 | #include <netinet/in_var.h> | |
110 | #include <netinet/ip_var.h> | |
111 | #ifdef INET6 | |
112 | #include <netinet6/ip6_var.h> | |
65f3e756 | 113 | #include <netinet6/tcp6_var.h> |
984263bc MD |
114 | #endif |
115 | #include <netinet/tcp.h> | |
116 | #include <netinet/tcp_fsm.h> | |
117 | #include <netinet/tcp_seq.h> | |
118 | #include <netinet/tcp_timer.h> | |
a48c5dd5 | 119 | #include <netinet/tcp_timer2.h> |
984263bc MD |
120 | #include <netinet/tcp_var.h> |
121 | #include <netinet/tcpip.h> | |
122 | #ifdef TCPDEBUG | |
123 | #include <netinet/tcp_debug.h> | |
124 | #endif | |
125 | ||
126 | #ifdef IPSEC | |
127 | #include <netinet6/ipsec.h> | |
128 | #endif /*IPSEC*/ | |
129 | ||
130 | /* | |
131 | * TCP protocol interface to socket abstraction. | |
132 | */ | |
133 | extern char *tcpstates[]; /* XXX ??? */ | |
134 | ||
e4700d00 | 135 | static int tcp_attach (struct socket *, struct pru_attach_info *); |
48e7b118 MD |
136 | static int tcp_connect (struct tcpcb *, int flags, struct mbuf *m, |
137 | struct sockaddr *, struct thread *); | |
984263bc | 138 | #ifdef INET6 |
48e7b118 MD |
139 | static int tcp6_connect (struct tcpcb *, int flags, struct mbuf *m, |
140 | struct sockaddr *, struct thread *); | |
141 | static int tcp6_connect_oncpu(struct tcpcb *tp, int flags, struct mbuf *m, | |
142 | struct sockaddr_in6 *sin6, | |
65f3e756 | 143 | struct in6_addr *addr6); |
984263bc MD |
144 | #endif /* INET6 */ |
145 | static struct tcpcb * | |
42a7fc75 | 146 | tcp_disconnect (struct tcpcb *); |
984263bc | 147 | static struct tcpcb * |
42a7fc75 | 148 | tcp_usrclosed (struct tcpcb *); |
984263bc MD |
149 | |
150 | #ifdef TCPDEBUG | |
151 | #define TCPDEBUG0 int ostate = 0 | |
152 | #define TCPDEBUG1() ostate = tp ? tp->t_state : 0 | |
153 | #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ | |
154 | tcp_trace(TA_USER, ostate, tp, 0, 0, req) | |
155 | #else | |
156 | #define TCPDEBUG0 | |
157 | #define TCPDEBUG1() | |
158 | #define TCPDEBUG2(req) | |
159 | #endif | |
160 | ||
161 | /* | |
162 | * TCP attaches to socket via pru_attach(), reserving space, | |
163 | * and an internet control block. | |
164 | */ | |
165 | static int | |
e4700d00 | 166 | tcp_usr_attach(struct socket *so, int proto, struct pru_attach_info *ai) |
984263bc | 167 | { |
984263bc | 168 | int error; |
1cae611f | 169 | struct inpcb *inp; |
984263bc MD |
170 | struct tcpcb *tp = 0; |
171 | TCPDEBUG0; | |
172 | ||
1cae611f MD |
173 | crit_enter(); |
174 | inp = so->so_pcb; | |
984263bc MD |
175 | TCPDEBUG1(); |
176 | if (inp) { | |
177 | error = EISCONN; | |
178 | goto out; | |
179 | } | |
180 | ||
e4700d00 | 181 | error = tcp_attach(so, ai); |
984263bc MD |
182 | if (error) |
183 | goto out; | |
184 | ||
185 | if ((so->so_options & SO_LINGER) && so->so_linger == 0) | |
186 | so->so_linger = TCP_LINGERTIME; | |
187 | tp = sototcpcb(so); | |
188 | out: | |
189 | TCPDEBUG2(PRU_ATTACH); | |
1cae611f | 190 | crit_exit(); |
984263bc MD |
191 | return error; |
192 | } | |
193 | ||
194 | /* | |
195 | * pru_detach() detaches the TCP protocol from the socket. | |
196 | * If the protocol state is non-embryonic, then can't | |
197 | * do this directly: have to initiate a pru_disconnect(), | |
198 | * which may finish later; embryonic TCB's can just | |
199 | * be discarded here. | |
200 | */ | |
201 | static int | |
202 | tcp_usr_detach(struct socket *so) | |
203 | { | |
984263bc | 204 | int error = 0; |
1cae611f | 205 | struct inpcb *inp; |
984263bc MD |
206 | struct tcpcb *tp; |
207 | TCPDEBUG0; | |
208 | ||
1cae611f MD |
209 | crit_enter(); |
210 | inp = so->so_pcb; | |
69891757 MD |
211 | |
212 | /* | |
213 | * If the inp is already detached it may have been due to an async | |
214 | * close. Just return as if no error occured. | |
215 | */ | |
621989fd | 216 | if (inp == NULL) { |
1cae611f | 217 | crit_exit(); |
69891757 | 218 | return 0; |
984263bc | 219 | } |
984263bc | 220 | |
621989fd MD |
221 | /* |
222 | * It's possible for the tcpcb (tp) to disconnect from the inp due | |
223 | * to tcp_drop()->tcp_close() being called. This may occur *after* | |
224 | * the detach message has been queued so we may find a NULL tp here. | |
225 | */ | |
226 | if ((tp = intotcpcb(inp)) != NULL) { | |
227 | TCPDEBUG1(); | |
228 | tp = tcp_disconnect(tp); | |
229 | TCPDEBUG2(PRU_DETACH); | |
230 | } | |
1cae611f | 231 | crit_exit(); |
984263bc MD |
232 | return error; |
233 | } | |
234 | ||
69891757 MD |
235 | /* |
236 | * Note: ignore_error is non-zero for certain disconnection races | |
237 | * which we want to silently allow, otherwise close() may return | |
238 | * an unexpected error. | |
239 | */ | |
240 | #define COMMON_START(so, inp, ignore_error) \ | |
241 | TCPDEBUG0; \ | |
242 | \ | |
243 | crit_enter(); \ | |
244 | inp = so->so_pcb; \ | |
245 | do { \ | |
246 | if (inp == NULL) { \ | |
247 | crit_exit(); \ | |
248 | return (ignore_error ? 0 : EINVAL); \ | |
249 | } \ | |
250 | tp = intotcpcb(inp); \ | |
251 | TCPDEBUG1(); \ | |
252 | } while(0) | |
f23061d4 | 253 | |
1cae611f | 254 | #define COMMON_END(req) out: TCPDEBUG2(req); crit_exit(); return error; goto out |
984263bc MD |
255 | |
256 | ||
257 | /* | |
258 | * Give the socket an address. | |
259 | */ | |
260 | static int | |
dadab5e9 | 261 | tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) |
984263bc | 262 | { |
984263bc | 263 | int error = 0; |
1cae611f | 264 | struct inpcb *inp; |
984263bc MD |
265 | struct tcpcb *tp; |
266 | struct sockaddr_in *sinp; | |
267 | ||
69891757 | 268 | COMMON_START(so, inp, 0); |
984263bc MD |
269 | |
270 | /* | |
271 | * Must check for multicast addresses and disallow binding | |
272 | * to them. | |
273 | */ | |
274 | sinp = (struct sockaddr_in *)nam; | |
275 | if (sinp->sin_family == AF_INET && | |
276 | IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { | |
277 | error = EAFNOSUPPORT; | |
278 | goto out; | |
279 | } | |
dadab5e9 | 280 | error = in_pcbbind(inp, nam, td); |
984263bc MD |
281 | if (error) |
282 | goto out; | |
283 | COMMON_END(PRU_BIND); | |
284 | ||
285 | } | |
286 | ||
287 | #ifdef INET6 | |
288 | static int | |
dadab5e9 | 289 | tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) |
984263bc | 290 | { |
984263bc | 291 | int error = 0; |
1cae611f | 292 | struct inpcb *inp; |
984263bc MD |
293 | struct tcpcb *tp; |
294 | struct sockaddr_in6 *sin6p; | |
295 | ||
69891757 | 296 | COMMON_START(so, inp, 0); |
984263bc MD |
297 | |
298 | /* | |
299 | * Must check for multicast addresses and disallow binding | |
300 | * to them. | |
301 | */ | |
302 | sin6p = (struct sockaddr_in6 *)nam; | |
303 | if (sin6p->sin6_family == AF_INET6 && | |
304 | IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { | |
305 | error = EAFNOSUPPORT; | |
306 | goto out; | |
307 | } | |
308 | inp->inp_vflag &= ~INP_IPV4; | |
309 | inp->inp_vflag |= INP_IPV6; | |
310 | if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { | |
311 | if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) | |
312 | inp->inp_vflag |= INP_IPV4; | |
313 | else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { | |
314 | struct sockaddr_in sin; | |
315 | ||
316 | in6_sin6_2_sin(&sin, sin6p); | |
317 | inp->inp_vflag |= INP_IPV4; | |
318 | inp->inp_vflag &= ~INP_IPV6; | |
dadab5e9 | 319 | error = in_pcbbind(inp, (struct sockaddr *)&sin, td); |
984263bc MD |
320 | goto out; |
321 | } | |
322 | } | |
dadab5e9 | 323 | error = in6_pcbbind(inp, nam, td); |
984263bc MD |
324 | if (error) |
325 | goto out; | |
326 | COMMON_END(PRU_BIND); | |
327 | } | |
328 | #endif /* INET6 */ | |
329 | ||
8affadf8 JH |
330 | #ifdef SMP |
331 | struct netmsg_inswildcard { | |
4599cf19 | 332 | struct netmsg nm_netmsg; |
8affadf8 JH |
333 | struct inpcb *nm_inp; |
334 | struct inpcbinfo *nm_pcbinfo; | |
335 | }; | |
336 | ||
4599cf19 MD |
337 | static void |
338 | in_pcbinswildcardhash_handler(struct netmsg *msg0) | |
8affadf8 JH |
339 | { |
340 | struct netmsg_inswildcard *msg = (struct netmsg_inswildcard *)msg0; | |
341 | ||
342 | in_pcbinswildcardhash_oncpu(msg->nm_inp, msg->nm_pcbinfo); | |
4599cf19 | 343 | lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0); |
8affadf8 JH |
344 | } |
345 | #endif | |
346 | ||
984263bc MD |
347 | /* |
348 | * Prepare to accept connections. | |
349 | */ | |
350 | static int | |
dadab5e9 | 351 | tcp_usr_listen(struct socket *so, struct thread *td) |
984263bc | 352 | { |
984263bc | 353 | int error = 0; |
1cae611f | 354 | struct inpcb *inp; |
984263bc | 355 | struct tcpcb *tp; |
8affadf8 JH |
356 | #ifdef SMP |
357 | int cpu; | |
358 | #endif | |
984263bc | 359 | |
69891757 | 360 | COMMON_START(so, inp, 0); |
8affadf8 JH |
361 | if (inp->inp_lport == 0) { |
362 | error = in_pcbbind(inp, NULL, td); | |
363 | if (error != 0) | |
364 | goto out; | |
365 | } | |
366 | ||
367 | tp->t_state = TCPS_LISTEN; | |
0f758523 | 368 | tp->tt_msg = NULL; /* Catch any invalid timer usage */ |
8affadf8 | 369 | #ifdef SMP |
eb594563 | 370 | /* |
00638fa3 JH |
371 | * We have to set the flag because we can't have other cpus |
372 | * messing with our inp's flags. | |
eb594563 MD |
373 | */ |
374 | inp->inp_flags |= INP_WILDCARD_MP; | |
8affadf8 JH |
375 | for (cpu = 0; cpu < ncpus2; cpu++) { |
376 | struct netmsg_inswildcard *msg; | |
377 | ||
378 | if (cpu == mycpu->gd_cpuid) { | |
f3a4db39 | 379 | in_pcbinswildcardhash(inp); |
8affadf8 JH |
380 | continue; |
381 | } | |
382 | ||
efda3bd0 | 383 | msg = kmalloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG, |
4599cf19 | 384 | M_INTWAIT); |
48e7b118 MD |
385 | netmsg_init(&msg->nm_netmsg, NULL, &netisr_afree_rport, |
386 | 0, in_pcbinswildcardhash_handler); | |
8affadf8 JH |
387 | msg->nm_inp = inp; |
388 | msg->nm_pcbinfo = &tcbinfo[cpu]; | |
4599cf19 | 389 | lwkt_sendmsg(tcp_cport(cpu), &msg->nm_netmsg.nm_lmsg); |
e61cbecc | 390 | } |
8affadf8 JH |
391 | #else |
392 | in_pcbinswildcardhash(inp); | |
393 | #endif | |
984263bc MD |
394 | COMMON_END(PRU_LISTEN); |
395 | } | |
396 | ||
397 | #ifdef INET6 | |
398 | static int | |
dadab5e9 | 399 | tcp6_usr_listen(struct socket *so, struct thread *td) |
984263bc | 400 | { |
984263bc | 401 | int error = 0; |
1cae611f | 402 | struct inpcb *inp; |
984263bc | 403 | struct tcpcb *tp; |
00638fa3 JH |
404 | #ifdef SMP |
405 | int cpu; | |
406 | #endif | |
984263bc | 407 | |
69891757 | 408 | COMMON_START(so, inp, 0); |
984263bc | 409 | if (inp->inp_lport == 0) { |
00638fa3 | 410 | if (!(inp->inp_flags & IN6P_IPV6_V6ONLY)) |
984263bc | 411 | inp->inp_vflag |= INP_IPV4; |
00638fa3 JH |
412 | else |
413 | inp->inp_vflag &= ~INP_IPV4; | |
60233e58 | 414 | error = in6_pcbbind(inp, NULL, td); |
984263bc MD |
415 | } |
416 | if (error == 0) | |
417 | tp->t_state = TCPS_LISTEN; | |
00638fa3 JH |
418 | #ifdef SMP |
419 | /* | |
420 | * We have to set the flag because we can't have other cpus | |
421 | * messing with our inp's flags. | |
422 | */ | |
423 | inp->inp_flags |= INP_WILDCARD_MP; | |
424 | for (cpu = 0; cpu < ncpus2; cpu++) { | |
425 | struct netmsg_inswildcard *msg; | |
426 | ||
427 | if (cpu == mycpu->gd_cpuid) { | |
41414e90 | 428 | in_pcbinswildcardhash(inp); |
00638fa3 JH |
429 | continue; |
430 | } | |
431 | ||
efda3bd0 | 432 | msg = kmalloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG, |
4599cf19 | 433 | M_INTWAIT); |
48e7b118 MD |
434 | netmsg_init(&msg->nm_netmsg, NULL, &netisr_afree_rport, |
435 | 0, in_pcbinswildcardhash_handler); | |
00638fa3 JH |
436 | msg->nm_inp = inp; |
437 | msg->nm_pcbinfo = &tcbinfo[cpu]; | |
4599cf19 | 438 | lwkt_sendmsg(tcp_cport(cpu), &msg->nm_netmsg.nm_lmsg); |
00638fa3 JH |
439 | } |
440 | #else | |
0ca5e441 | 441 | in_pcbinswildcardhash(inp); |
00638fa3 | 442 | #endif |
984263bc MD |
443 | COMMON_END(PRU_LISTEN); |
444 | } | |
445 | #endif /* INET6 */ | |
446 | ||
447 | /* | |
448 | * Initiate connection to peer. | |
449 | * Create a template for use in transmissions on this connection. | |
450 | * Enter SYN_SENT state, and mark socket as connecting. | |
451 | * Start keep-alive timer, and seed output sequence space. | |
452 | * Send initial segment on connection. | |
453 | */ | |
454 | static int | |
dadab5e9 | 455 | tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) |
984263bc | 456 | { |
984263bc | 457 | int error = 0; |
1cae611f | 458 | struct inpcb *inp; |
984263bc MD |
459 | struct tcpcb *tp; |
460 | struct sockaddr_in *sinp; | |
461 | ||
69891757 | 462 | COMMON_START(so, inp, 0); |
984263bc MD |
463 | |
464 | /* | |
465 | * Must disallow TCP ``connections'' to multicast addresses. | |
466 | */ | |
467 | sinp = (struct sockaddr_in *)nam; | |
468 | if (sinp->sin_family == AF_INET | |
469 | && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { | |
470 | error = EAFNOSUPPORT; | |
471 | goto out; | |
472 | } | |
473 | ||
3e4150ef VBD |
474 | if (!prison_remote_ip(td, (struct sockaddr*)sinp)) { |
475 | error = EAFNOSUPPORT; /* IPv6 only jail */ | |
476 | goto out; | |
477 | } | |
984263bc | 478 | |
48e7b118 | 479 | if ((error = tcp_connect(tp, 0, NULL, nam, td)) != 0) |
984263bc | 480 | goto out; |
984263bc MD |
481 | COMMON_END(PRU_CONNECT); |
482 | } | |
483 | ||
484 | #ifdef INET6 | |
485 | static int | |
dadab5e9 | 486 | tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) |
984263bc | 487 | { |
984263bc | 488 | int error = 0; |
1cae611f | 489 | struct inpcb *inp; |
984263bc MD |
490 | struct tcpcb *tp; |
491 | struct sockaddr_in6 *sin6p; | |
492 | ||
69891757 | 493 | COMMON_START(so, inp, 0); |
984263bc MD |
494 | |
495 | /* | |
496 | * Must disallow TCP ``connections'' to multicast addresses. | |
497 | */ | |
498 | sin6p = (struct sockaddr_in6 *)nam; | |
499 | if (sin6p->sin6_family == AF_INET6 | |
500 | && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { | |
501 | error = EAFNOSUPPORT; | |
502 | goto out; | |
503 | } | |
504 | ||
3e4150ef VBD |
505 | if (!prison_remote_ip(td, nam)) { |
506 | error = EAFNOSUPPORT; /* IPv4 only jail */ | |
507 | goto out; | |
508 | } | |
509 | ||
984263bc MD |
510 | if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { |
511 | struct sockaddr_in sin; | |
512 | ||
513 | if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { | |
514 | error = EINVAL; | |
515 | goto out; | |
516 | } | |
517 | ||
518 | in6_sin6_2_sin(&sin, sin6p); | |
519 | inp->inp_vflag |= INP_IPV4; | |
520 | inp->inp_vflag &= ~INP_IPV6; | |
48e7b118 MD |
521 | error = tcp_connect(tp, 0, NULL, (struct sockaddr *)&sin, td); |
522 | if (error) | |
984263bc | 523 | goto out; |
984263bc MD |
524 | goto out; |
525 | } | |
526 | inp->inp_vflag &= ~INP_IPV4; | |
527 | inp->inp_vflag |= INP_IPV6; | |
528 | inp->inp_inc.inc_isipv6 = 1; | |
48e7b118 | 529 | if ((error = tcp6_connect(tp, 0, NULL, nam, td)) != 0) |
984263bc MD |
530 | goto out; |
531 | error = tcp_output(tp); | |
532 | COMMON_END(PRU_CONNECT); | |
533 | } | |
534 | #endif /* INET6 */ | |
535 | ||
536 | /* | |
537 | * Initiate disconnect from peer. | |
538 | * If connection never passed embryonic stage, just drop; | |
539 | * else if don't need to let data drain, then can just drop anyways, | |
540 | * else have to begin TCP shutdown process: mark socket disconnecting, | |
541 | * drain unread data, state switch to reflect user close, and | |
542 | * send segment (e.g. FIN) to peer. Socket will be really disconnected | |
543 | * when peer sends FIN and acks ours. | |
544 | * | |
545 | * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. | |
546 | */ | |
547 | static int | |
548 | tcp_usr_disconnect(struct socket *so) | |
549 | { | |
984263bc | 550 | int error = 0; |
1cae611f | 551 | struct inpcb *inp; |
984263bc MD |
552 | struct tcpcb *tp; |
553 | ||
69891757 | 554 | COMMON_START(so, inp, 1); |
984263bc MD |
555 | tp = tcp_disconnect(tp); |
556 | COMMON_END(PRU_DISCONNECT); | |
557 | } | |
558 | ||
559 | /* | |
560 | * Accept a connection. Essentially all the work is | |
561 | * done at higher levels; just return the address | |
562 | * of the peer, storing through addr. | |
563 | */ | |
564 | static int | |
565 | tcp_usr_accept(struct socket *so, struct sockaddr **nam) | |
566 | { | |
984263bc | 567 | int error = 0; |
1cae611f | 568 | struct inpcb *inp; |
984263bc MD |
569 | struct tcpcb *tp = NULL; |
570 | TCPDEBUG0; | |
571 | ||
1cae611f MD |
572 | crit_enter(); |
573 | inp = so->so_pcb; | |
984263bc MD |
574 | if (so->so_state & SS_ISDISCONNECTED) { |
575 | error = ECONNABORTED; | |
576 | goto out; | |
577 | } | |
578 | if (inp == 0) { | |
1cae611f | 579 | crit_exit(); |
984263bc MD |
580 | return (EINVAL); |
581 | } | |
582 | tp = intotcpcb(inp); | |
583 | TCPDEBUG1(); | |
584 | in_setpeeraddr(so, nam); | |
585 | COMMON_END(PRU_ACCEPT); | |
586 | } | |
587 | ||
588 | #ifdef INET6 | |
589 | static int | |
590 | tcp6_usr_accept(struct socket *so, struct sockaddr **nam) | |
591 | { | |
984263bc | 592 | int error = 0; |
1cae611f | 593 | struct inpcb *inp; |
984263bc MD |
594 | struct tcpcb *tp = NULL; |
595 | TCPDEBUG0; | |
596 | ||
1cae611f MD |
597 | crit_enter(); |
598 | inp = so->so_pcb; | |
599 | ||
984263bc MD |
600 | if (so->so_state & SS_ISDISCONNECTED) { |
601 | error = ECONNABORTED; | |
602 | goto out; | |
603 | } | |
604 | if (inp == 0) { | |
1cae611f | 605 | crit_exit(); |
984263bc MD |
606 | return (EINVAL); |
607 | } | |
608 | tp = intotcpcb(inp); | |
609 | TCPDEBUG1(); | |
610 | in6_mapped_peeraddr(so, nam); | |
611 | COMMON_END(PRU_ACCEPT); | |
612 | } | |
613 | #endif /* INET6 */ | |
614 | /* | |
615 | * Mark the connection as being incapable of further output. | |
616 | */ | |
617 | static int | |
618 | tcp_usr_shutdown(struct socket *so) | |
619 | { | |
984263bc | 620 | int error = 0; |
1cae611f | 621 | struct inpcb *inp; |
984263bc MD |
622 | struct tcpcb *tp; |
623 | ||
69891757 | 624 | COMMON_START(so, inp, 0); |
984263bc MD |
625 | socantsendmore(so); |
626 | tp = tcp_usrclosed(tp); | |
627 | if (tp) | |
628 | error = tcp_output(tp); | |
629 | COMMON_END(PRU_SHUTDOWN); | |
630 | } | |
631 | ||
632 | /* | |
633 | * After a receive, possibly send window update to peer. | |
634 | */ | |
635 | static int | |
636 | tcp_usr_rcvd(struct socket *so, int flags) | |
637 | { | |
984263bc | 638 | int error = 0; |
1cae611f | 639 | struct inpcb *inp; |
984263bc MD |
640 | struct tcpcb *tp; |
641 | ||
69891757 | 642 | COMMON_START(so, inp, 0); |
984263bc MD |
643 | tcp_output(tp); |
644 | COMMON_END(PRU_RCVD); | |
645 | } | |
646 | ||
647 | /* | |
648 | * Do a send by putting data in output queue and updating urgent | |
649 | * marker if URG set. Possibly send more data. Unlike the other | |
650 | * pru_*() routines, the mbuf chains are our responsibility. We | |
651 | * must either enqueue them or free them. The other pru_* routines | |
652 | * generally are caller-frees. | |
653 | */ | |
654 | static int | |
f23061d4 | 655 | tcp_usr_send(struct socket *so, int flags, struct mbuf *m, |
dadab5e9 | 656 | struct sockaddr *nam, struct mbuf *control, struct thread *td) |
984263bc | 657 | { |
984263bc | 658 | int error = 0; |
1cae611f | 659 | struct inpcb *inp; |
984263bc MD |
660 | struct tcpcb *tp; |
661 | #ifdef INET6 | |
662 | int isipv6; | |
663 | #endif | |
664 | TCPDEBUG0; | |
665 | ||
1cae611f MD |
666 | crit_enter(); |
667 | inp = so->so_pcb; | |
668 | ||
984263bc MD |
669 | if (inp == NULL) { |
670 | /* | |
671 | * OOPS! we lost a race, the TCP session got reset after | |
672 | * we checked SS_CANTSENDMORE, eg: while doing uiomove or a | |
1cae611f | 673 | * network interrupt in the non-critical section of sosend(). |
984263bc | 674 | */ |
48e7b118 | 675 | m_freem(m); |
984263bc MD |
676 | if (control) |
677 | m_freem(control); | |
678 | error = ECONNRESET; /* XXX EPIPE? */ | |
679 | tp = NULL; | |
680 | TCPDEBUG1(); | |
681 | goto out; | |
682 | } | |
683 | #ifdef INET6 | |
684 | isipv6 = nam && nam->sa_family == AF_INET6; | |
685 | #endif /* INET6 */ | |
686 | tp = intotcpcb(inp); | |
687 | TCPDEBUG1(); | |
688 | if (control) { | |
689 | /* TCP doesn't do control messages (rights, creds, etc) */ | |
690 | if (control->m_len) { | |
691 | m_freem(control); | |
48e7b118 | 692 | m_freem(m); |
984263bc MD |
693 | error = EINVAL; |
694 | goto out; | |
695 | } | |
696 | m_freem(control); /* empty control, just free it */ | |
697 | } | |
984263bc | 698 | |
48e7b118 MD |
699 | /* |
700 | * Don't let too much OOB data build up | |
701 | */ | |
702 | if (flags & PRUS_OOB) { | |
6d49aa6f | 703 | if (ssb_space(&so->so_snd) < -512) { |
984263bc MD |
704 | m_freem(m); |
705 | error = ENOBUFS; | |
706 | goto out; | |
707 | } | |
48e7b118 MD |
708 | } |
709 | ||
710 | /* | |
711 | * Do implied connect if not yet connected. Any data sent | |
712 | * with the connect is handled by tcp_connect() and friends. | |
713 | * | |
714 | * NOTE! PROTOCOL THREAD MAY BE CHANGED BY THE CONNECT! | |
715 | */ | |
716 | if (nam && tp->t_state < TCPS_SYN_SENT) { | |
717 | #ifdef INET6 | |
718 | if (isipv6) | |
719 | error = tcp6_connect(tp, flags, m, nam, td); | |
720 | else | |
721 | #endif /* INET6 */ | |
722 | error = tcp_connect(tp, flags, m, nam, td); | |
723 | #if 0 | |
724 | /* WTF is this doing here? */ | |
725 | tp->snd_wnd = TTCP_CLIENT_SND_WND; | |
726 | tcp_mss(tp, -1); | |
727 | #endif | |
728 | goto out; | |
729 | } | |
730 | ||
731 | /* | |
732 | * Pump the data into the socket. | |
733 | */ | |
734 | if (m) | |
735 | ssb_appendstream(&so->so_snd, m); | |
736 | if (flags & PRUS_OOB) { | |
984263bc MD |
737 | /* |
738 | * According to RFC961 (Assigned Protocols), | |
739 | * the urgent pointer points to the last octet | |
740 | * of urgent data. We continue, however, | |
741 | * to consider it to indicate the first octet | |
742 | * of data past the urgent section. | |
743 | * Otherwise, snd_up should be one lower. | |
744 | */ | |
6d49aa6f | 745 | tp->snd_up = tp->snd_una + so->so_snd.ssb_cc; |
c1d3b892 | 746 | tp->t_flags |= TF_FORCE; |
984263bc | 747 | error = tcp_output(tp); |
c1d3b892 | 748 | tp->t_flags &= ~TF_FORCE; |
48e7b118 MD |
749 | } else { |
750 | if (flags & PRUS_EOF) { | |
751 | /* | |
752 | * Close the send side of the connection after | |
753 | * the data is sent. | |
754 | */ | |
755 | socantsendmore(so); | |
756 | tp = tcp_usrclosed(tp); | |
757 | } | |
758 | if (tp != NULL) { | |
759 | if (flags & PRUS_MORETOCOME) | |
760 | tp->t_flags |= TF_MORETOCOME; | |
761 | error = tcp_output(tp); | |
762 | if (flags & PRUS_MORETOCOME) | |
763 | tp->t_flags &= ~TF_MORETOCOME; | |
764 | } | |
984263bc | 765 | } |
f23061d4 | 766 | COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : |
984263bc MD |
767 | ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); |
768 | } | |
769 | ||
770 | /* | |
771 | * Abort the TCP. | |
772 | */ | |
773 | static int | |
774 | tcp_usr_abort(struct socket *so) | |
775 | { | |
984263bc | 776 | int error = 0; |
1cae611f | 777 | struct inpcb *inp; |
984263bc MD |
778 | struct tcpcb *tp; |
779 | ||
69891757 | 780 | COMMON_START(so, inp, 1); |
984263bc MD |
781 | tp = tcp_drop(tp, ECONNABORTED); |
782 | COMMON_END(PRU_ABORT); | |
783 | } | |
784 | ||
785 | /* | |
786 | * Receive out-of-band data. | |
787 | */ | |
788 | static int | |
789 | tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) | |
790 | { | |
984263bc | 791 | int error = 0; |
1cae611f | 792 | struct inpcb *inp; |
984263bc MD |
793 | struct tcpcb *tp; |
794 | ||
69891757 | 795 | COMMON_START(so, inp, 0); |
984263bc MD |
796 | if ((so->so_oobmark == 0 && |
797 | (so->so_state & SS_RCVATMARK) == 0) || | |
798 | so->so_options & SO_OOBINLINE || | |
799 | tp->t_oobflags & TCPOOB_HADDATA) { | |
800 | error = EINVAL; | |
801 | goto out; | |
802 | } | |
803 | if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { | |
804 | error = EWOULDBLOCK; | |
805 | goto out; | |
806 | } | |
807 | m->m_len = 1; | |
808 | *mtod(m, caddr_t) = tp->t_iobc; | |
809 | if ((flags & MSG_PEEK) == 0) | |
810 | tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); | |
811 | COMMON_END(PRU_RCVOOB); | |
812 | } | |
813 | ||
814 | /* xxx - should be const */ | |
815 | struct pr_usrreqs tcp_usrreqs = { | |
fa5e758c MD |
816 | .pru_abort = tcp_usr_abort, |
817 | .pru_accept = tcp_usr_accept, | |
818 | .pru_attach = tcp_usr_attach, | |
819 | .pru_bind = tcp_usr_bind, | |
820 | .pru_connect = tcp_usr_connect, | |
821 | .pru_connect2 = pru_connect2_notsupp, | |
822 | .pru_control = in_control, | |
823 | .pru_detach = tcp_usr_detach, | |
824 | .pru_disconnect = tcp_usr_disconnect, | |
825 | .pru_listen = tcp_usr_listen, | |
826 | .pru_peeraddr = in_setpeeraddr, | |
827 | .pru_rcvd = tcp_usr_rcvd, | |
828 | .pru_rcvoob = tcp_usr_rcvoob, | |
829 | .pru_send = tcp_usr_send, | |
830 | .pru_sense = pru_sense_null, | |
831 | .pru_shutdown = tcp_usr_shutdown, | |
832 | .pru_sockaddr = in_setsockaddr, | |
833 | .pru_sosend = sosend, | |
8b5c39bb | 834 | .pru_soreceive = soreceive |
984263bc MD |
835 | }; |
836 | ||
837 | #ifdef INET6 | |
838 | struct pr_usrreqs tcp6_usrreqs = { | |
fa5e758c MD |
839 | .pru_abort = tcp_usr_abort, |
840 | .pru_accept = tcp6_usr_accept, | |
841 | .pru_attach = tcp_usr_attach, | |
842 | .pru_bind = tcp6_usr_bind, | |
843 | .pru_connect = tcp6_usr_connect, | |
844 | .pru_connect2 = pru_connect2_notsupp, | |
845 | .pru_control = in6_control, | |
846 | .pru_detach = tcp_usr_detach, | |
847 | .pru_disconnect = tcp_usr_disconnect, | |
848 | .pru_listen = tcp6_usr_listen, | |
849 | .pru_peeraddr = in6_mapped_peeraddr, | |
850 | .pru_rcvd = tcp_usr_rcvd, | |
851 | .pru_rcvoob = tcp_usr_rcvoob, | |
852 | .pru_send = tcp_usr_send, | |
853 | .pru_sense = pru_sense_null, | |
854 | .pru_shutdown = tcp_usr_shutdown, | |
855 | .pru_sockaddr = in6_mapped_sockaddr, | |
856 | .pru_sosend = sosend, | |
8b5c39bb | 857 | .pru_soreceive = soreceive |
984263bc MD |
858 | }; |
859 | #endif /* INET6 */ | |
860 | ||
984263bc | 861 | static int |
48e7b118 MD |
862 | tcp_connect_oncpu(struct tcpcb *tp, int flags, struct mbuf *m, |
863 | struct sockaddr_in *sin, struct sockaddr_in *if_sin) | |
984263bc MD |
864 | { |
865 | struct inpcb *inp = tp->t_inpcb, *oinp; | |
866 | struct socket *so = inp->inp_socket; | |
03fb1cae | 867 | struct route *ro = &inp->inp_route; |
984263bc | 868 | |
7fe56515 | 869 | oinp = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid], |
984263bc | 870 | sin->sin_addr, sin->sin_port, |
13d8907a | 871 | inp->inp_laddr.s_addr != INADDR_ANY ? |
f23061d4 | 872 | inp->inp_laddr : if_sin->sin_addr, |
7fe56515 | 873 | inp->inp_lport, 0, NULL); |
13d8907a | 874 | if (oinp != NULL) { |
27b8aee3 AE |
875 | m_freem(m); |
876 | return (EADDRINUSE); | |
984263bc MD |
877 | } |
878 | if (inp->inp_laddr.s_addr == INADDR_ANY) | |
13d8907a | 879 | inp->inp_laddr = if_sin->sin_addr; |
984263bc MD |
880 | inp->inp_faddr = sin->sin_addr; |
881 | inp->inp_fport = sin->sin_port; | |
7fe56515 | 882 | inp->inp_cpcbinfo = &tcbinfo[mycpu->gd_cpuid]; |
13d8907a | 883 | in_pcbinsconnhash(inp); |
984263bc | 884 | |
03fb1cae SZ |
885 | /* |
886 | * We are now on the inpcb's owner CPU, if the cached route was | |
887 | * freed because the rtentry's owner CPU is not the current CPU | |
888 | * (e.g. in tcp_connect()), then we try to reallocate it here with | |
889 | * the hope that a rtentry may be cloned from a RTF_PRCLONING | |
890 | * rtentry. | |
891 | */ | |
892 | if (!(inp->inp_socket->so_options & SO_DONTROUTE) && /*XXX*/ | |
893 | ro->ro_rt == NULL) { | |
894 | bzero(&ro->ro_dst, sizeof(struct sockaddr_in)); | |
895 | ro->ro_dst.sa_family = AF_INET; | |
896 | ro->ro_dst.sa_len = sizeof(struct sockaddr_in); | |
897 | ((struct sockaddr_in *)&ro->ro_dst)->sin_addr = | |
898 | sin->sin_addr; | |
899 | rtalloc(ro); | |
900 | } | |
901 | ||
3db1c8a3 | 902 | /* |
48e7b118 MD |
903 | * Now that no more errors can occur, change the protocol processing |
904 | * port to the current thread (which is the correct thread). | |
905 | * | |
3db1c8a3 SZ |
906 | * Create TCP timer message now; we are on the tcpcb's owner |
907 | * CPU/thread. | |
908 | */ | |
48e7b118 | 909 | sosetport(so, &curthread->td_msgport); |
3db1c8a3 | 910 | tcp_create_timermsg(tp, &curthread->td_msgport); |
0f758523 | 911 | |
46e92930 MD |
912 | /* |
913 | * Compute window scaling to request. Use a larger scaling then | |
914 | * needed for the initial receive buffer in case the receive buffer | |
915 | * gets expanded. | |
916 | */ | |
917 | if (tp->request_r_scale < TCP_MIN_WINSHIFT) | |
918 | tp->request_r_scale = TCP_MIN_WINSHIFT; | |
984263bc | 919 | while (tp->request_r_scale < TCP_MAX_WINSHIFT && |
46e92930 MD |
920 | (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.ssb_hiwat |
921 | ) { | |
984263bc | 922 | tp->request_r_scale++; |
46e92930 | 923 | } |
984263bc MD |
924 | |
925 | soisconnecting(so); | |
926 | tcpstat.tcps_connattempt++; | |
927 | tp->t_state = TCPS_SYN_SENT; | |
a48c5dd5 | 928 | tcp_callout_reset(tp, tp->tt_keep, tcp_keepinit, tcp_timer_keep); |
984263bc | 929 | tp->iss = tcp_new_isn(tp); |
984263bc | 930 | tcp_sendseqinit(tp); |
48e7b118 MD |
931 | if (m) { |
932 | ssb_appendstream(&so->so_snd, m); | |
933 | m = NULL; | |
934 | if (flags & PRUS_OOB) | |
935 | tp->snd_up = tp->snd_una + so->so_snd.ssb_cc; | |
936 | } | |
984263bc | 937 | |
48e7b118 MD |
938 | /* |
939 | * Close the send side of the connection after | |
940 | * the data is sent if flagged. | |
941 | */ | |
942 | if ((flags & (PRUS_OOB|PRUS_EOF)) == PRUS_EOF) { | |
943 | socantsendmore(so); | |
944 | tp = tcp_usrclosed(tp); | |
945 | } | |
946 | return (tcp_output(tp)); | |
984263bc MD |
947 | } |
948 | ||
540a0bf0 JH |
949 | #ifdef SMP |
950 | ||
951 | struct netmsg_tcp_connect { | |
4599cf19 | 952 | struct netmsg nm_netmsg; |
540a0bf0 JH |
953 | struct tcpcb *nm_tp; |
954 | struct sockaddr_in *nm_sin; | |
955 | struct sockaddr_in *nm_ifsin; | |
48e7b118 MD |
956 | int nm_flags; |
957 | struct mbuf *nm_m; | |
540a0bf0 | 958 | }; |
b44419cb | 959 | |
4599cf19 MD |
960 | static void |
961 | tcp_connect_handler(netmsg_t netmsg) | |
7fe56515 | 962 | { |
4599cf19 | 963 | struct netmsg_tcp_connect *msg = (void *)netmsg; |
7fe56515 JH |
964 | int error; |
965 | ||
48e7b118 MD |
966 | error = tcp_connect_oncpu(msg->nm_tp, msg->nm_flags, msg->nm_m, |
967 | msg->nm_sin, msg->nm_ifsin); | |
4599cf19 | 968 | lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, error); |
7fe56515 JH |
969 | } |
970 | ||
65f3e756 MD |
971 | struct netmsg_tcp6_connect { |
972 | struct netmsg nm_netmsg; | |
973 | struct tcpcb *nm_tp; | |
974 | struct sockaddr_in6 *nm_sin6; | |
975 | struct in6_addr *nm_addr6; | |
48e7b118 MD |
976 | int nm_flags; |
977 | struct mbuf *nm_m; | |
65f3e756 MD |
978 | }; |
979 | ||
2183cbe4 | 980 | #ifdef INET6 |
65f3e756 MD |
981 | static void |
982 | tcp6_connect_handler(netmsg_t netmsg) | |
983 | { | |
984 | struct netmsg_tcp6_connect *msg = (void *)netmsg; | |
985 | int error; | |
986 | ||
48e7b118 MD |
987 | error = tcp6_connect_oncpu(msg->nm_tp, msg->nm_flags, msg->nm_m, |
988 | msg->nm_sin6, msg->nm_addr6); | |
65f3e756 MD |
989 | lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, error); |
990 | } | |
b44419cb MD |
991 | #endif |
992 | ||
2183cbe4 SW |
993 | #endif /* SMP */ |
994 | ||
7fe56515 JH |
995 | /* |
996 | * Common subroutine to open a TCP connection to remote host specified | |
997 | * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local | |
998 | * port number if needed. Call in_pcbladdr to do the routing and to choose | |
27b8aee3 | 999 | * a local host address (interface). |
7fe56515 JH |
1000 | * Initialize connection parameters and enter SYN-SENT state. |
1001 | */ | |
1002 | static int | |
48e7b118 MD |
1003 | tcp_connect(struct tcpcb *tp, int flags, struct mbuf *m, |
1004 | struct sockaddr *nam, struct thread *td) | |
7fe56515 JH |
1005 | { |
1006 | struct inpcb *inp = tp->t_inpcb; | |
1007 | struct sockaddr_in *sin = (struct sockaddr_in *)nam; | |
1008 | struct sockaddr_in *if_sin; | |
1009 | int error; | |
540a0bf0 | 1010 | #ifdef SMP |
7fe56515 JH |
1011 | lwkt_port_t port; |
1012 | #endif | |
1013 | ||
48e7b118 MD |
1014 | /* |
1015 | * Bind if we have to | |
1016 | */ | |
7fe56515 | 1017 | if (inp->inp_lport == 0) { |
2038fb68 | 1018 | error = in_pcbbind(inp, NULL, td); |
48e7b118 MD |
1019 | if (error) { |
1020 | m_freem(m); | |
7fe56515 | 1021 | return (error); |
48e7b118 | 1022 | } |
7fe56515 JH |
1023 | } |
1024 | ||
1025 | /* | |
48e7b118 MD |
1026 | * Calculate the correct protocol processing thread. The connect |
1027 | * operation must run there. | |
7fe56515 | 1028 | */ |
3e4150ef | 1029 | error = in_pcbladdr(inp, nam, &if_sin, td); |
48e7b118 MD |
1030 | if (error) { |
1031 | m_freem(m); | |
7fe56515 | 1032 | return (error); |
48e7b118 | 1033 | } |
7fe56515 | 1034 | |
540a0bf0 | 1035 | #ifdef SMP |
7fe56515 JH |
1036 | port = tcp_addrport(sin->sin_addr.s_addr, sin->sin_port, |
1037 | inp->inp_laddr.s_addr ? | |
1038 | inp->inp_laddr.s_addr : if_sin->sin_addr.s_addr, | |
1039 | inp->inp_lport); | |
1040 | ||
fb0f29c4 | 1041 | if (port != &curthread->td_msgport) { |
540a0bf0 | 1042 | struct netmsg_tcp_connect msg; |
a46456f3 SZ |
1043 | struct route *ro = &inp->inp_route; |
1044 | ||
1045 | /* | |
1046 | * in_pcbladdr() may have allocated a route entry for us | |
1047 | * on the current CPU, but we need a route entry on the | |
03fb1cae | 1048 | * inpcb's owner CPU, so free it here. |
a46456f3 SZ |
1049 | */ |
1050 | if (ro->ro_rt != NULL) | |
1051 | RTFREE(ro->ro_rt); | |
1052 | bzero(ro, sizeof(*ro)); | |
540a0bf0 | 1053 | |
48e7b118 MD |
1054 | /* |
1055 | * NOTE: We haven't set so->so_port yet do not pass so | |
1056 | * to netmsg_init() or it will be improperly forwarded. | |
1057 | */ | |
1058 | netmsg_init(&msg.nm_netmsg, NULL, &curthread->td_msgport, | |
1059 | 0, tcp_connect_handler); | |
540a0bf0 JH |
1060 | msg.nm_tp = tp; |
1061 | msg.nm_sin = sin; | |
1062 | msg.nm_ifsin = if_sin; | |
48e7b118 MD |
1063 | msg.nm_flags = flags; |
1064 | msg.nm_m = m; | |
a22c590e | 1065 | error = lwkt_domsg(port, &msg.nm_netmsg.nm_lmsg, 0); |
48e7b118 MD |
1066 | } else { |
1067 | error = tcp_connect_oncpu(tp, flags, m, sin, if_sin); | |
1068 | } | |
1069 | #else | |
1070 | error = tcp_connect_oncpu(tp, flags, m, sin, if_sin); | |
7fe56515 | 1071 | #endif |
7fe56515 JH |
1072 | return (error); |
1073 | } | |
1074 | ||
984263bc | 1075 | #ifdef INET6 |
65f3e756 | 1076 | |
984263bc | 1077 | static int |
48e7b118 MD |
1078 | tcp6_connect(struct tcpcb *tp, int flags, struct mbuf *m, |
1079 | struct sockaddr *nam, struct thread *td) | |
984263bc | 1080 | { |
65f3e756 | 1081 | struct inpcb *inp = tp->t_inpcb; |
984263bc MD |
1082 | struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; |
1083 | struct in6_addr *addr6; | |
f890e67a | 1084 | #ifdef SMP |
65f3e756 | 1085 | lwkt_port_t port; |
f890e67a | 1086 | #endif |
984263bc MD |
1087 | int error; |
1088 | ||
1089 | if (inp->inp_lport == 0) { | |
60233e58 | 1090 | error = in6_pcbbind(inp, NULL, td); |
48e7b118 MD |
1091 | if (error) { |
1092 | m_freem(m); | |
1093 | return (error); | |
1094 | } | |
984263bc MD |
1095 | } |
1096 | ||
1097 | /* | |
1098 | * Cannot simply call in_pcbconnect, because there might be an | |
1099 | * earlier incarnation of this same connection still in | |
1100 | * TIME_WAIT state, creating an ADDRINUSE error. | |
1101 | */ | |
3e4150ef | 1102 | error = in6_pcbladdr(inp, nam, &addr6, td); |
48e7b118 MD |
1103 | if (error) { |
1104 | m_freem(m); | |
1105 | return (error); | |
1106 | } | |
65f3e756 MD |
1107 | |
1108 | #ifdef SMP | |
1109 | port = tcp6_addrport(); /* XXX hack for now, always cpu0 */ | |
1110 | ||
1111 | if (port != &curthread->td_msgport) { | |
1112 | struct netmsg_tcp6_connect msg; | |
1113 | struct route *ro = &inp->inp_route; | |
1114 | ||
1115 | /* | |
1116 | * in_pcbladdr() may have allocated a route entry for us | |
1117 | * on the current CPU, but we need a route entry on the | |
1118 | * inpcb's owner CPU, so free it here. | |
1119 | */ | |
1120 | if (ro->ro_rt != NULL) | |
1121 | RTFREE(ro->ro_rt); | |
1122 | bzero(ro, sizeof(*ro)); | |
1123 | ||
48e7b118 MD |
1124 | netmsg_init(&msg.nm_netmsg, NULL, &curthread->td_msgport, |
1125 | 0, tcp6_connect_handler); | |
65f3e756 MD |
1126 | msg.nm_tp = tp; |
1127 | msg.nm_sin6 = sin6; | |
1128 | msg.nm_addr6 = addr6; | |
48e7b118 MD |
1129 | msg.nm_flags = flags; |
1130 | msg.nm_m = m; | |
65f3e756 | 1131 | error = lwkt_domsg(port, &msg.nm_netmsg.nm_lmsg, 0); |
48e7b118 MD |
1132 | } else { |
1133 | error = tcp6_connect_oncpu(tp, flags, m, sin6, addr6); | |
1134 | } | |
1135 | #else | |
1136 | error = tcp6_connect_oncpu(tp, flags, m, sin6, addr6); | |
65f3e756 | 1137 | #endif |
65f3e756 MD |
1138 | return (error); |
1139 | } | |
1140 | ||
1141 | static int | |
48e7b118 MD |
1142 | tcp6_connect_oncpu(struct tcpcb *tp, int flags, struct mbuf *m, |
1143 | struct sockaddr_in6 *sin6, struct in6_addr *addr6) | |
65f3e756 MD |
1144 | { |
1145 | struct inpcb *inp = tp->t_inpcb; | |
1146 | struct socket *so = inp->inp_socket; | |
1147 | struct inpcb *oinp; | |
65f3e756 | 1148 | |
48e7b118 MD |
1149 | /* |
1150 | * Cannot simply call in_pcbconnect, because there might be an | |
1151 | * earlier incarnation of this same connection still in | |
1152 | * TIME_WAIT state, creating an ADDRINUSE error. | |
1153 | */ | |
7fe56515 | 1154 | oinp = in6_pcblookup_hash(inp->inp_cpcbinfo, |
984263bc | 1155 | &sin6->sin6_addr, sin6->sin6_port, |
13d8907a JH |
1156 | IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ? |
1157 | addr6 : &inp->in6p_laddr, | |
984263bc MD |
1158 | inp->inp_lport, 0, NULL); |
1159 | if (oinp) { | |
27b8aee3 AE |
1160 | m_freem(m); |
1161 | return (EADDRINUSE); | |
984263bc MD |
1162 | } |
1163 | if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) | |
1164 | inp->in6p_laddr = *addr6; | |
1165 | inp->in6p_faddr = sin6->sin6_addr; | |
1166 | inp->inp_fport = sin6->sin6_port; | |
3641b7ca | 1167 | if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != 0) |
984263bc | 1168 | inp->in6p_flowinfo = sin6->sin6_flowinfo; |
13d8907a | 1169 | in_pcbinsconnhash(inp); |
984263bc | 1170 | |
48e7b118 MD |
1171 | /* |
1172 | * Now that no more errors can occur, change the protocol processing | |
1173 | * port to the current thread (which is the correct thread). | |
1174 | * | |
1175 | * Create TCP timer message now; we are on the tcpcb's owner | |
1176 | * CPU/thread. | |
1177 | */ | |
1178 | sosetport(so, &curthread->td_msgport); | |
3db1c8a3 SZ |
1179 | tcp_create_timermsg(tp, &curthread->td_msgport); |
1180 | ||
984263bc | 1181 | /* Compute window scaling to request. */ |
46e92930 MD |
1182 | if (tp->request_r_scale < TCP_MIN_WINSHIFT) |
1183 | tp->request_r_scale = TCP_MIN_WINSHIFT; | |
984263bc | 1184 | while (tp->request_r_scale < TCP_MAX_WINSHIFT && |
46e92930 | 1185 | (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.ssb_hiwat) { |
984263bc | 1186 | tp->request_r_scale++; |
46e92930 | 1187 | } |
984263bc MD |
1188 | |
1189 | soisconnecting(so); | |
1190 | tcpstat.tcps_connattempt++; | |
1191 | tp->t_state = TCPS_SYN_SENT; | |
a48c5dd5 | 1192 | tcp_callout_reset(tp, tp->tt_keep, tcp_keepinit, tcp_timer_keep); |
984263bc | 1193 | tp->iss = tcp_new_isn(tp); |
984263bc | 1194 | tcp_sendseqinit(tp); |
48e7b118 MD |
1195 | if (m) { |
1196 | ssb_appendstream(&so->so_snd, m); | |
1197 | m = NULL; | |
1198 | if (flags & PRUS_OOB) | |
1199 | tp->snd_up = tp->snd_una + so->so_snd.ssb_cc; | |
1200 | } | |
984263bc | 1201 | |
48e7b118 MD |
1202 | /* |
1203 | * Close the send side of the connection after | |
1204 | * the data is sent if flagged. | |
1205 | */ | |
1206 | if ((flags & (PRUS_OOB|PRUS_EOF)) == PRUS_EOF) { | |
1207 | socantsendmore(so); | |
1208 | tp = tcp_usrclosed(tp); | |
1209 | } | |
1210 | return (tcp_output(tp)); | |
984263bc | 1211 | } |
65f3e756 | 1212 | |
984263bc MD |
1213 | #endif /* INET6 */ |
1214 | ||
1215 | /* | |
1216 | * The new sockopt interface makes it possible for us to block in the | |
1cae611f MD |
1217 | * copyin/out step (if we take a page fault). Taking a page fault while |
1218 | * in a critical section is probably a Bad Thing. (Since sockets and pcbs | |
1219 | * both now use TSM, there probably isn't any need for this function to | |
1220 | * run in a critical section any more. This needs more examination.) | |
984263bc MD |
1221 | */ |
1222 | int | |
f3f70f0d | 1223 | tcp_ctloutput(struct socket *so, struct sockopt *sopt) |
984263bc | 1224 | { |
1cae611f | 1225 | int error, opt, optval; |
984263bc MD |
1226 | struct inpcb *inp; |
1227 | struct tcpcb *tp; | |
1228 | ||
1229 | error = 0; | |
1cae611f | 1230 | crit_enter(); /* XXX */ |
ed894f8c | 1231 | inp = so->so_pcb; |
984263bc | 1232 | if (inp == NULL) { |
1cae611f | 1233 | crit_exit(); |
984263bc MD |
1234 | return (ECONNRESET); |
1235 | } | |
1236 | if (sopt->sopt_level != IPPROTO_TCP) { | |
1237 | #ifdef INET6 | |
1238 | if (INP_CHECK_SOCKAF(so, AF_INET6)) | |
1239 | error = ip6_ctloutput(so, sopt); | |
1240 | else | |
1241 | #endif /* INET6 */ | |
1242 | error = ip_ctloutput(so, sopt); | |
1cae611f | 1243 | crit_exit(); |
984263bc MD |
1244 | return (error); |
1245 | } | |
1246 | tp = intotcpcb(inp); | |
1247 | ||
1248 | switch (sopt->sopt_dir) { | |
1249 | case SOPT_SET: | |
e71a125f AE |
1250 | error = soopt_to_kbuf(sopt, &optval, sizeof optval, |
1251 | sizeof optval); | |
1252 | if (error) | |
1253 | break; | |
984263bc MD |
1254 | switch (sopt->sopt_name) { |
1255 | case TCP_NODELAY: | |
1256 | case TCP_NOOPT: | |
984263bc MD |
1257 | switch (sopt->sopt_name) { |
1258 | case TCP_NODELAY: | |
1259 | opt = TF_NODELAY; | |
1260 | break; | |
1261 | case TCP_NOOPT: | |
1262 | opt = TF_NOOPT; | |
1263 | break; | |
1264 | default: | |
1265 | opt = 0; /* dead code to fool gcc */ | |
1266 | break; | |
1267 | } | |
1268 | ||
1269 | if (optval) | |
1270 | tp->t_flags |= opt; | |
1271 | else | |
1272 | tp->t_flags &= ~opt; | |
1273 | break; | |
1274 | ||
1275 | case TCP_NOPUSH: | |
984263bc MD |
1276 | if (optval) |
1277 | tp->t_flags |= TF_NOPUSH; | |
1278 | else { | |
1279 | tp->t_flags &= ~TF_NOPUSH; | |
1280 | error = tcp_output(tp); | |
1281 | } | |
1282 | break; | |
1283 | ||
1284 | case TCP_MAXSEG: | |
153d34ac MD |
1285 | /* |
1286 | * Must be between 0 and maxseg. If the requested | |
1287 | * maxseg is too small to satisfy the desired minmss, | |
1288 | * pump it up (silently so sysctl modifications of | |
1289 | * minmss do not create unexpected program failures). | |
1290 | * Handle degenerate cases. | |
1291 | */ | |
1292 | if (optval > 0 && optval <= tp->t_maxseg) { | |
1293 | if (optval + 40 < tcp_minmss) { | |
1294 | optval = tcp_minmss - 40; | |
1295 | if (optval < 0) | |
1296 | optval = 1; | |
1297 | } | |
984263bc | 1298 | tp->t_maxseg = optval; |
153d34ac | 1299 | } else { |
984263bc | 1300 | error = EINVAL; |
153d34ac | 1301 | } |
984263bc MD |
1302 | break; |
1303 | ||
1304 | default: | |
1305 | error = ENOPROTOOPT; | |
1306 | break; | |
1307 | } | |
1308 | break; | |
1309 | ||
1310 | case SOPT_GET: | |
1311 | switch (sopt->sopt_name) { | |
1312 | case TCP_NODELAY: | |
1313 | optval = tp->t_flags & TF_NODELAY; | |
1314 | break; | |
1315 | case TCP_MAXSEG: | |
1316 | optval = tp->t_maxseg; | |
1317 | break; | |
1318 | case TCP_NOOPT: | |
1319 | optval = tp->t_flags & TF_NOOPT; | |
1320 | break; | |
1321 | case TCP_NOPUSH: | |
1322 | optval = tp->t_flags & TF_NOPUSH; | |
1323 | break; | |
1324 | default: | |
1325 | error = ENOPROTOOPT; | |
1326 | break; | |
1327 | } | |
1328 | if (error == 0) | |
e71a125f | 1329 | soopt_from_kbuf(sopt, &optval, sizeof optval); |
984263bc MD |
1330 | break; |
1331 | } | |
1cae611f | 1332 | crit_exit(); |
984263bc MD |
1333 | return (error); |
1334 | } | |
1335 | ||
1336 | /* | |
1337 | * tcp_sendspace and tcp_recvspace are the default send and receive window | |
1338 | * sizes, respectively. These are obsolescent (this information should | |
1339 | * be set by the route). | |
0560a274 MD |
1340 | * |
1341 | * Use a default that does not require tcp window scaling to be turned | |
1342 | * on. Individual programs or the administrator can increase the default. | |
984263bc | 1343 | */ |
0560a274 | 1344 | u_long tcp_sendspace = 57344; /* largest multiple of PAGE_SIZE < 64k */ |
f23061d4 | 1345 | SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, |
984263bc MD |
1346 | &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); |
1347 | u_long tcp_recvspace = 57344; /* largest multiple of PAGE_SIZE < 64k */ | |
f23061d4 | 1348 | SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, |
984263bc MD |
1349 | &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); |
1350 | ||
1351 | /* | |
1352 | * Attach TCP protocol to socket, allocating | |
1353 | * internet protocol control block, tcp control block, | |
1354 | * bufer space, and entering LISTEN state if to accept connections. | |
1355 | */ | |
1356 | static int | |
e4700d00 | 1357 | tcp_attach(struct socket *so, struct pru_attach_info *ai) |
984263bc | 1358 | { |
2256ba69 | 1359 | struct tcpcb *tp; |
984263bc MD |
1360 | struct inpcb *inp; |
1361 | int error; | |
6ca1a1cd | 1362 | int cpu; |
984263bc | 1363 | #ifdef INET6 |
3641b7ca | 1364 | int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != 0; |
984263bc MD |
1365 | #endif |
1366 | ||
6d49aa6f | 1367 | if (so->so_snd.ssb_hiwat == 0 || so->so_rcv.ssb_hiwat == 0) { |
e4700d00 JH |
1368 | error = soreserve(so, tcp_sendspace, tcp_recvspace, |
1369 | ai->sb_rlimit); | |
984263bc MD |
1370 | if (error) |
1371 | return (error); | |
1372 | } | |
5b0b9fa5 PA |
1373 | so->so_rcv.ssb_flags |= SSB_AUTOSIZE; |
1374 | so->so_snd.ssb_flags |= SSB_AUTOSIZE; | |
6ca1a1cd | 1375 | cpu = mycpu->gd_cpuid; |
6ca1a1cd | 1376 | error = in_pcballoc(so, &tcbinfo[cpu]); |
984263bc MD |
1377 | if (error) |
1378 | return (error); | |
ed894f8c | 1379 | inp = so->so_pcb; |
984263bc MD |
1380 | #ifdef INET6 |
1381 | if (isipv6) { | |
1382 | inp->inp_vflag |= INP_IPV6; | |
1383 | inp->in6p_hops = -1; /* use kernel default */ | |
1384 | } | |
1385 | else | |
1386 | #endif | |
1387 | inp->inp_vflag |= INP_IPV4; | |
1388 | tp = tcp_newtcpcb(inp); | |
1389 | if (tp == 0) { | |
1390 | int nofd = so->so_state & SS_NOFDREF; /* XXX */ | |
1391 | ||
1392 | so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ | |
1393 | #ifdef INET6 | |
1394 | if (isipv6) | |
1395 | in6_pcbdetach(inp); | |
1396 | else | |
1397 | #endif | |
1398 | in_pcbdetach(inp); | |
1399 | so->so_state |= nofd; | |
1400 | return (ENOBUFS); | |
1401 | } | |
1402 | tp->t_state = TCPS_CLOSED; | |
48e7b118 | 1403 | so->so_port = tcp_soport_attach(so); |
984263bc MD |
1404 | return (0); |
1405 | } | |
1406 | ||
1407 | /* | |
1408 | * Initiate (or continue) disconnect. | |
1409 | * If embryonic state, just send reset (once). | |
1410 | * If in ``let data drain'' option and linger null, just drop. | |
1411 | * Otherwise (hard), mark socket disconnecting and drop | |
1412 | * current input data; switch states based on user close, and | |
1413 | * send segment to peer (with FIN). | |
1414 | */ | |
1415 | static struct tcpcb * | |
f3f70f0d | 1416 | tcp_disconnect(struct tcpcb *tp) |
984263bc MD |
1417 | { |
1418 | struct socket *so = tp->t_inpcb->inp_socket; | |
1419 | ||
1420 | if (tp->t_state < TCPS_ESTABLISHED) | |
1421 | tp = tcp_close(tp); | |
1422 | else if ((so->so_options & SO_LINGER) && so->so_linger == 0) | |
1423 | tp = tcp_drop(tp, 0); | |
1424 | else { | |
1425 | soisdisconnecting(so); | |
6d49aa6f | 1426 | sbflush(&so->so_rcv.sb); |
984263bc MD |
1427 | tp = tcp_usrclosed(tp); |
1428 | if (tp) | |
f23061d4 | 1429 | tcp_output(tp); |
984263bc MD |
1430 | } |
1431 | return (tp); | |
1432 | } | |
1433 | ||
1434 | /* | |
1435 | * User issued close, and wish to trail through shutdown states: | |
1436 | * if never received SYN, just forget it. If got a SYN from peer, | |
1437 | * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. | |
1438 | * If already got a FIN from peer, then almost done; go to LAST_ACK | |
1439 | * state. In all other cases, have already sent FIN to peer (e.g. | |
1440 | * after PRU_SHUTDOWN), and just have to play tedious game waiting | |
1441 | * for peer to send FIN or not respond to keep-alives, etc. | |
1442 | * We can let the user exit from the close as soon as the FIN is acked. | |
1443 | */ | |
1444 | static struct tcpcb * | |
f3f70f0d | 1445 | tcp_usrclosed(struct tcpcb *tp) |
984263bc MD |
1446 | { |
1447 | ||
1448 | switch (tp->t_state) { | |
1449 | ||
1450 | case TCPS_CLOSED: | |
1451 | case TCPS_LISTEN: | |
1452 | tp->t_state = TCPS_CLOSED; | |
1453 | tp = tcp_close(tp); | |
1454 | break; | |
1455 | ||
1456 | case TCPS_SYN_SENT: | |
1457 | case TCPS_SYN_RECEIVED: | |
1458 | tp->t_flags |= TF_NEEDFIN; | |
1459 | break; | |
1460 | ||
1461 | case TCPS_ESTABLISHED: | |
1462 | tp->t_state = TCPS_FIN_WAIT_1; | |
1463 | break; | |
1464 | ||
1465 | case TCPS_CLOSE_WAIT: | |
1466 | tp->t_state = TCPS_LAST_ACK; | |
1467 | break; | |
1468 | } | |
1469 | if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { | |
1470 | soisdisconnected(tp->t_inpcb->inp_socket); | |
1471 | /* To prevent the connection hanging in FIN_WAIT_2 forever. */ | |
a48c5dd5 SZ |
1472 | if (tp->t_state == TCPS_FIN_WAIT_2) { |
1473 | tcp_callout_reset(tp, tp->tt_2msl, tcp_maxidle, | |
1474 | tcp_timer_2msl); | |
1475 | } | |
984263bc MD |
1476 | } |
1477 | return (tp); | |
1478 | } |