1 /* 2 * Copyright (c) 2004 Jeffrey M. Hsu. All rights reserved. 3 * Copyright (c) 2004 The DragonFly Project. All rights reserved. 4 * 5 * This code is derived from software contributed to The DragonFly Project 6 * by Jeffrey M. Hsu. 7 * 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. 19 * 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 34 /* 35 * Copyright (c) 1982, 1986, 1988, 1990, 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 * @(#)uipc_socket.c 8.3 (Berkeley) 4/15/94 67 * $FreeBSD: src/sys/kern/uipc_socket.c,v 1.68.2.24 2003/11/11 17:18:18 silby Exp $ 68 * $DragonFly: src/sys/kern/uipc_socket.c,v 1.45 2007/04/22 01:13:10 dillon Exp $ 69 */ 70 71 #include "opt_inet.h" 72 #include "opt_sctp.h" 73 74 #include <sys/param.h> 75 #include <sys/systm.h> 76 #include <sys/fcntl.h> 77 #include <sys/malloc.h> 78 #include <sys/mbuf.h> 79 #include <sys/domain.h> 80 #include <sys/file.h> /* for struct knote */ 81 #include <sys/kernel.h> 82 #include <sys/malloc.h> 83 #include <sys/event.h> 84 #include <sys/poll.h> 85 #include <sys/proc.h> 86 #include <sys/protosw.h> 87 #include <sys/socket.h> 88 #include <sys/socketvar.h> 89 #include <sys/socketops.h> 90 #include <sys/resourcevar.h> 91 #include <sys/signalvar.h> 92 #include <sys/sysctl.h> 93 #include <sys/uio.h> 94 #include <sys/jail.h> 95 #include <vm/vm_zone.h> 96 97 #include <sys/thread2.h> 98 99 #include <machine/limits.h> 100 101 #ifdef INET 102 static int do_setopt_accept_filter(struct socket *so, struct sockopt *sopt); 103 #endif /* INET */ 104 105 static void filt_sordetach(struct knote *kn); 106 static int filt_soread(struct knote *kn, long hint); 107 static void filt_sowdetach(struct knote *kn); 108 static int filt_sowrite(struct knote *kn, long hint); 109 static int filt_solisten(struct knote *kn, long hint); 110 111 static struct filterops solisten_filtops = 112 { 1, NULL, filt_sordetach, filt_solisten }; 113 static struct filterops soread_filtops = 114 { 1, NULL, filt_sordetach, filt_soread }; 115 static struct filterops sowrite_filtops = 116 { 1, NULL, filt_sowdetach, filt_sowrite }; 117 118 struct vm_zone *socket_zone; 119 120 MALLOC_DEFINE(M_SONAME, "soname", "socket name"); 121 MALLOC_DEFINE(M_PCB, "pcb", "protocol control block"); 122 123 124 static int somaxconn = SOMAXCONN; 125 SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW, 126 &somaxconn, 0, "Maximum pending socket connection queue size"); 127 128 /* 129 * Socket operation routines. 130 * These routines are called by the routines in 131 * sys_socket.c or from a system process, and 132 * implement the semantics of socket operations by 133 * switching out to the protocol specific routines. 134 */ 135 136 /* 137 * Get a socket structure from our zone, and initialize it. 138 * We don't implement `waitok' yet (see comments in uipc_domain.c). 139 * Note that it would probably be better to allocate socket 140 * and PCB at the same time, but I'm not convinced that all 141 * the protocols can be easily modified to do this. 142 */ 143 struct socket * 144 soalloc(int waitok) 145 { 146 struct socket *so; 147 148 so = zalloc(socket_zone); 149 if (so) { 150 /* XXX race condition for reentrant kernel */ 151 bzero(so, sizeof *so); 152 TAILQ_INIT(&so->so_aiojobq); 153 TAILQ_INIT(&so->so_rcv.ssb_sel.si_mlist); 154 TAILQ_INIT(&so->so_snd.ssb_sel.si_mlist); 155 } 156 return so; 157 } 158 159 int 160 socreate(int dom, struct socket **aso, int type, 161 int proto, struct thread *td) 162 { 163 struct proc *p = td->td_proc; 164 struct protosw *prp; 165 struct socket *so; 166 struct pru_attach_info ai; 167 int error; 168 169 if (proto) 170 prp = pffindproto(dom, proto, type); 171 else 172 prp = pffindtype(dom, type); 173 174 if (prp == 0 || prp->pr_usrreqs->pru_attach == 0) 175 return (EPROTONOSUPPORT); 176 177 if (p->p_ucred->cr_prison && jail_socket_unixiproute_only && 178 prp->pr_domain->dom_family != PF_LOCAL && 179 prp->pr_domain->dom_family != PF_INET && 180 prp->pr_domain->dom_family != PF_INET6 && 181 prp->pr_domain->dom_family != PF_ROUTE) { 182 return (EPROTONOSUPPORT); 183 } 184 185 if (prp->pr_type != type) 186 return (EPROTOTYPE); 187 so = soalloc(p != 0); 188 if (so == 0) 189 return (ENOBUFS); 190 191 TAILQ_INIT(&so->so_incomp); 192 TAILQ_INIT(&so->so_comp); 193 so->so_type = type; 194 so->so_cred = crhold(p->p_ucred); 195 so->so_proto = prp; 196 ai.sb_rlimit = &p->p_rlimit[RLIMIT_SBSIZE]; 197 ai.p_ucred = p->p_ucred; 198 ai.fd_rdir = p->p_fd->fd_rdir; 199 error = so_pru_attach(so, proto, &ai); 200 if (error) { 201 so->so_state |= SS_NOFDREF; 202 sofree(so); 203 return (error); 204 } 205 *aso = so; 206 return (0); 207 } 208 209 int 210 sobind(struct socket *so, struct sockaddr *nam, struct thread *td) 211 { 212 int error; 213 214 crit_enter(); 215 error = so_pru_bind(so, nam, td); 216 crit_exit(); 217 return (error); 218 } 219 220 void 221 sodealloc(struct socket *so) 222 { 223 if (so->so_rcv.ssb_hiwat) 224 (void)chgsbsize(so->so_cred->cr_uidinfo, 225 &so->so_rcv.ssb_hiwat, 0, RLIM_INFINITY); 226 if (so->so_snd.ssb_hiwat) 227 (void)chgsbsize(so->so_cred->cr_uidinfo, 228 &so->so_snd.ssb_hiwat, 0, RLIM_INFINITY); 229 #ifdef INET 230 /* remove accept filter if present */ 231 if (so->so_accf != NULL) 232 do_setopt_accept_filter(so, NULL); 233 #endif /* INET */ 234 crfree(so->so_cred); 235 zfree(socket_zone, so); 236 } 237 238 int 239 solisten(struct socket *so, int backlog, struct thread *td) 240 { 241 int error; 242 #ifdef SCTP 243 short oldopt, oldqlimit; 244 #endif /* SCTP */ 245 246 crit_enter(); 247 if (so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING)) { 248 crit_exit(); 249 return (EINVAL); 250 } 251 252 #ifdef SCTP 253 oldopt = so->so_options; 254 oldqlimit = so->so_qlimit; 255 #endif /* SCTP */ 256 257 if (TAILQ_EMPTY(&so->so_comp)) 258 so->so_options |= SO_ACCEPTCONN; 259 if (backlog < 0 || backlog > somaxconn) 260 backlog = somaxconn; 261 so->so_qlimit = backlog; 262 /* SCTP needs to look at tweak both the inbound backlog parameter AND 263 * the so_options (UDP model both connect's and gets inbound 264 * connections .. implicitly). 265 */ 266 error = so_pru_listen(so, td); 267 if (error) { 268 #ifdef SCTP 269 /* Restore the params */ 270 so->so_options = oldopt; 271 so->so_qlimit = oldqlimit; 272 #endif /* SCTP */ 273 crit_exit(); 274 return (error); 275 } 276 crit_exit(); 277 return (0); 278 } 279 280 void 281 sofree(struct socket *so) 282 { 283 struct socket *head = so->so_head; 284 285 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) 286 return; 287 if (head != NULL) { 288 if (so->so_state & SS_INCOMP) { 289 TAILQ_REMOVE(&head->so_incomp, so, so_list); 290 head->so_incqlen--; 291 } else if (so->so_state & SS_COMP) { 292 /* 293 * We must not decommission a socket that's 294 * on the accept(2) queue. If we do, then 295 * accept(2) may hang after select(2) indicated 296 * that the listening socket was ready. 297 */ 298 return; 299 } else { 300 panic("sofree: not queued"); 301 } 302 so->so_state &= ~SS_INCOMP; 303 so->so_head = NULL; 304 } 305 ssb_release(&so->so_snd, so); 306 sorflush(so); 307 sodealloc(so); 308 } 309 310 /* 311 * Close a socket on last file table reference removal. 312 * Initiate disconnect if connected. 313 * Free socket when disconnect complete. 314 */ 315 int 316 soclose(struct socket *so, int fflag) 317 { 318 int error = 0; 319 320 crit_enter(); 321 funsetown(so->so_sigio); 322 if (so->so_pcb == NULL) 323 goto discard; 324 if (so->so_state & SS_ISCONNECTED) { 325 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 326 error = sodisconnect(so); 327 if (error) 328 goto drop; 329 } 330 if (so->so_options & SO_LINGER) { 331 if ((so->so_state & SS_ISDISCONNECTING) && 332 (fflag & FNONBLOCK)) 333 goto drop; 334 while (so->so_state & SS_ISCONNECTED) { 335 error = tsleep((caddr_t)&so->so_timeo, 336 PCATCH, "soclos", so->so_linger * hz); 337 if (error) 338 break; 339 } 340 } 341 } 342 drop: 343 if (so->so_pcb) { 344 int error2; 345 346 error2 = so_pru_detach(so); 347 if (error == 0) 348 error = error2; 349 } 350 discard: 351 if (so->so_options & SO_ACCEPTCONN) { 352 struct socket *sp, *sonext; 353 354 sp = TAILQ_FIRST(&so->so_incomp); 355 for (; sp != NULL; sp = sonext) { 356 sonext = TAILQ_NEXT(sp, so_list); 357 (void) soabort(sp); 358 } 359 for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) { 360 sonext = TAILQ_NEXT(sp, so_list); 361 /* Dequeue from so_comp since sofree() won't do it */ 362 TAILQ_REMOVE(&so->so_comp, sp, so_list); 363 so->so_qlen--; 364 sp->so_state &= ~SS_COMP; 365 sp->so_head = NULL; 366 (void) soabort(sp); 367 } 368 } 369 if (so->so_state & SS_NOFDREF) 370 panic("soclose: NOFDREF"); 371 so->so_state |= SS_NOFDREF; 372 sofree(so); 373 crit_exit(); 374 return (error); 375 } 376 377 /* 378 * Must be called from a critical section. 379 */ 380 int 381 soabort(struct socket *so) 382 { 383 int error; 384 385 error = so_pru_abort(so); 386 if (error) { 387 sofree(so); 388 return error; 389 } 390 return (0); 391 } 392 393 int 394 soaccept(struct socket *so, struct sockaddr **nam) 395 { 396 int error; 397 398 crit_enter(); 399 if ((so->so_state & SS_NOFDREF) == 0) 400 panic("soaccept: !NOFDREF"); 401 so->so_state &= ~SS_NOFDREF; 402 error = so_pru_accept(so, nam); 403 crit_exit(); 404 return (error); 405 } 406 407 int 408 soconnect(struct socket *so, struct sockaddr *nam, struct thread *td) 409 { 410 int error; 411 412 if (so->so_options & SO_ACCEPTCONN) 413 return (EOPNOTSUPP); 414 crit_enter(); 415 /* 416 * If protocol is connection-based, can only connect once. 417 * Otherwise, if connected, try to disconnect first. 418 * This allows user to disconnect by connecting to, e.g., 419 * a null address. 420 */ 421 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 422 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 423 (error = sodisconnect(so)))) { 424 error = EISCONN; 425 } else { 426 /* 427 * Prevent accumulated error from previous connection 428 * from biting us. 429 */ 430 so->so_error = 0; 431 error = so_pru_connect(so, nam, td); 432 } 433 crit_exit(); 434 return (error); 435 } 436 437 int 438 soconnect2(struct socket *so1, struct socket *so2) 439 { 440 int error; 441 442 crit_enter(); 443 error = so_pru_connect2(so1, so2); 444 crit_exit(); 445 return (error); 446 } 447 448 int 449 sodisconnect(struct socket *so) 450 { 451 int error; 452 453 crit_enter(); 454 if ((so->so_state & SS_ISCONNECTED) == 0) { 455 error = ENOTCONN; 456 goto bad; 457 } 458 if (so->so_state & SS_ISDISCONNECTING) { 459 error = EALREADY; 460 goto bad; 461 } 462 error = so_pru_disconnect(so); 463 bad: 464 crit_exit(); 465 return (error); 466 } 467 468 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 469 /* 470 * Send on a socket. 471 * If send must go all at once and message is larger than 472 * send buffering, then hard error. 473 * Lock against other senders. 474 * If must go all at once and not enough room now, then 475 * inform user that this would block and do nothing. 476 * Otherwise, if nonblocking, send as much as possible. 477 * The data to be sent is described by "uio" if nonzero, 478 * otherwise by the mbuf chain "top" (which must be null 479 * if uio is not). Data provided in mbuf chain must be small 480 * enough to send all at once. 481 * 482 * Returns nonzero on error, timeout or signal; callers 483 * must check for short counts if EINTR/ERESTART are returned. 484 * Data and control buffers are freed on return. 485 */ 486 int 487 sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, 488 struct mbuf *top, struct mbuf *control, int flags, 489 struct thread *td) 490 { 491 struct mbuf **mp; 492 struct mbuf *m; 493 long space, len, resid; 494 int clen = 0, error, dontroute, mlen; 495 int atomic = sosendallatonce(so) || top; 496 int pru_flags; 497 498 if (uio) 499 resid = uio->uio_resid; 500 else 501 resid = top->m_pkthdr.len; 502 /* 503 * In theory resid should be unsigned. 504 * However, space must be signed, as it might be less than 0 505 * if we over-committed, and we must use a signed comparison 506 * of space and resid. On the other hand, a negative resid 507 * causes us to loop sending 0-length segments to the protocol. 508 * 509 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM 510 * type sockets since that's an error. 511 */ 512 if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) { 513 error = EINVAL; 514 goto out; 515 } 516 517 dontroute = 518 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 519 (so->so_proto->pr_flags & PR_ATOMIC); 520 if (td->td_lwp != NULL) 521 td->td_lwp->lwp_ru.ru_msgsnd++; 522 if (control) 523 clen = control->m_len; 524 #define gotoerr(errcode) { error = errcode; crit_exit(); goto release; } 525 526 restart: 527 error = ssb_lock(&so->so_snd, SBLOCKWAIT(flags)); 528 if (error) 529 goto out; 530 do { 531 crit_enter(); 532 if (so->so_state & SS_CANTSENDMORE) 533 gotoerr(EPIPE); 534 if (so->so_error) { 535 error = so->so_error; 536 so->so_error = 0; 537 crit_exit(); 538 goto release; 539 } 540 if ((so->so_state & SS_ISCONNECTED) == 0) { 541 /* 542 * `sendto' and `sendmsg' is allowed on a connection- 543 * based socket if it supports implied connect. 544 * Return ENOTCONN if not connected and no address is 545 * supplied. 546 */ 547 if ((so->so_proto->pr_flags & PR_CONNREQUIRED) && 548 (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) { 549 if ((so->so_state & SS_ISCONFIRMING) == 0 && 550 !(resid == 0 && clen != 0)) 551 gotoerr(ENOTCONN); 552 } else if (addr == 0) 553 gotoerr(so->so_proto->pr_flags & PR_CONNREQUIRED ? 554 ENOTCONN : EDESTADDRREQ); 555 } 556 space = ssb_space(&so->so_snd); 557 if (flags & MSG_OOB) 558 space += 1024; 559 if ((atomic && resid > so->so_snd.ssb_hiwat) || 560 clen > so->so_snd.ssb_hiwat) 561 gotoerr(EMSGSIZE); 562 if (space < resid + clen && uio && 563 (atomic || space < so->so_snd.ssb_lowat || space < clen)) { 564 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) 565 gotoerr(EWOULDBLOCK); 566 ssb_unlock(&so->so_snd); 567 error = ssb_wait(&so->so_snd); 568 crit_exit(); 569 if (error) 570 goto out; 571 goto restart; 572 } 573 crit_exit(); 574 mp = ⊤ 575 space -= clen; 576 do { 577 if (uio == NULL) { 578 /* 579 * Data is prepackaged in "top". 580 */ 581 resid = 0; 582 if (flags & MSG_EOR) 583 top->m_flags |= M_EOR; 584 } else do { 585 m = m_getl(resid, MB_WAIT, MT_DATA, 586 top == NULL ? M_PKTHDR : 0, &mlen); 587 if (top == NULL) { 588 m->m_pkthdr.len = 0; 589 m->m_pkthdr.rcvif = (struct ifnet *)0; 590 } 591 len = min(min(mlen, resid), space); 592 if (resid < MINCLSIZE) { 593 /* 594 * For datagram protocols, leave room 595 * for protocol headers in first mbuf. 596 */ 597 if (atomic && top == 0 && len < mlen) 598 MH_ALIGN(m, len); 599 } 600 space -= len; 601 error = uiomove(mtod(m, caddr_t), (int)len, uio); 602 resid = uio->uio_resid; 603 m->m_len = len; 604 *mp = m; 605 top->m_pkthdr.len += len; 606 if (error) 607 goto release; 608 mp = &m->m_next; 609 if (resid <= 0) { 610 if (flags & MSG_EOR) 611 top->m_flags |= M_EOR; 612 break; 613 } 614 } while (space > 0 && atomic); 615 if (dontroute) 616 so->so_options |= SO_DONTROUTE; 617 if (flags & MSG_OOB) { 618 pru_flags = PRUS_OOB; 619 } else if ((flags & MSG_EOF) && 620 (so->so_proto->pr_flags & PR_IMPLOPCL) && 621 (resid <= 0)) { 622 /* 623 * If the user set MSG_EOF, the protocol 624 * understands this flag and nothing left to 625 * send then use PRU_SEND_EOF instead of PRU_SEND. 626 */ 627 pru_flags = PRUS_EOF; 628 } else if (resid > 0 && space > 0) { 629 /* If there is more to send, set PRUS_MORETOCOME */ 630 pru_flags = PRUS_MORETOCOME; 631 } else { 632 pru_flags = 0; 633 } 634 crit_enter(); 635 /* 636 * XXX all the SS_CANTSENDMORE checks previously 637 * done could be out of date. We could have recieved 638 * a reset packet in an interrupt or maybe we slept 639 * while doing page faults in uiomove() etc. We could 640 * probably recheck again inside the splnet() protection 641 * here, but there are probably other places that this 642 * also happens. We must rethink this. 643 */ 644 error = so_pru_send(so, pru_flags, top, addr, control, td); 645 crit_exit(); 646 if (dontroute) 647 so->so_options &= ~SO_DONTROUTE; 648 clen = 0; 649 control = 0; 650 top = 0; 651 mp = ⊤ 652 if (error) 653 goto release; 654 } while (resid && space > 0); 655 } while (resid); 656 657 release: 658 ssb_unlock(&so->so_snd); 659 out: 660 if (top) 661 m_freem(top); 662 if (control) 663 m_freem(control); 664 return (error); 665 } 666 667 /* 668 * A specialization of sosend() for UDP based on protocol-specific knowledge: 669 * so->so_proto->pr_flags has the PR_ATOMIC field set. This means that 670 * sosendallatonce() returns true, 671 * the "atomic" variable is true, 672 * and sosendudp() blocks until space is available for the entire send. 673 * so->so_proto->pr_flags does not have the PR_CONNREQUIRED or 674 * PR_IMPLOPCL flags set. 675 * UDP has no out-of-band data. 676 * UDP has no control data. 677 * UDP does not support MSG_EOR. 678 */ 679 int 680 sosendudp(struct socket *so, struct sockaddr *addr, struct uio *uio, 681 struct mbuf *top, struct mbuf *control, int flags, struct thread *td) 682 { 683 int resid, error; 684 boolean_t dontroute; /* temporary SO_DONTROUTE setting */ 685 686 if (td->td_lwp != NULL) 687 td->td_lwp->lwp_ru.ru_msgsnd++; 688 if (control) 689 m_freem(control); 690 691 KASSERT((uio && !top) || (top && !uio), ("bad arguments to sosendudp")); 692 resid = uio ? uio->uio_resid : top->m_pkthdr.len; 693 694 restart: 695 error = ssb_lock(&so->so_snd, SBLOCKWAIT(flags)); 696 if (error) 697 goto out; 698 699 crit_enter(); 700 if (so->so_state & SS_CANTSENDMORE) 701 gotoerr(EPIPE); 702 if (so->so_error) { 703 error = so->so_error; 704 so->so_error = 0; 705 crit_exit(); 706 goto release; 707 } 708 if (!(so->so_state & SS_ISCONNECTED) && addr == NULL) 709 gotoerr(EDESTADDRREQ); 710 if (resid > so->so_snd.ssb_hiwat) 711 gotoerr(EMSGSIZE); 712 if (uio && ssb_space(&so->so_snd) < resid) { 713 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) 714 gotoerr(EWOULDBLOCK); 715 ssb_unlock(&so->so_snd); 716 error = ssb_wait(&so->so_snd); 717 crit_exit(); 718 if (error) 719 goto out; 720 goto restart; 721 } 722 crit_exit(); 723 724 if (uio) { 725 top = m_uiomove(uio); 726 if (top == NULL) 727 goto release; 728 } 729 730 dontroute = (flags & MSG_DONTROUTE) && !(so->so_options & SO_DONTROUTE); 731 if (dontroute) 732 so->so_options |= SO_DONTROUTE; 733 734 error = so_pru_send(so, 0, top, addr, NULL, td); 735 top = NULL; /* sent or freed in lower layer */ 736 737 if (dontroute) 738 so->so_options &= ~SO_DONTROUTE; 739 740 release: 741 ssb_unlock(&so->so_snd); 742 out: 743 if (top) 744 m_freem(top); 745 return (error); 746 } 747 748 /* 749 * Implement receive operations on a socket. 750 * We depend on the way that records are added to the signalsockbuf 751 * by sbappend*. In particular, each record (mbufs linked through m_next) 752 * must begin with an address if the protocol so specifies, 753 * followed by an optional mbuf or mbufs containing ancillary data, 754 * and then zero or more mbufs of data. 755 * In order to avoid blocking network interrupts for the entire time here, 756 * we exit the critical section while doing the actual copy to user space. 757 * Although the signalsockbuf is locked, new data may still be appended, 758 * and thus we must maintain consistency of the signalsockbuf during that time. 759 * 760 * The caller may receive the data as a single mbuf chain by supplying 761 * an mbuf **mp0 for use in returning the chain. The uio is then used 762 * only for the count in uio_resid. 763 */ 764 int 765 soreceive(struct socket *so, struct sockaddr **psa, struct uio *uio, 766 struct sockbuf *sio, struct mbuf **controlp, int *flagsp) 767 { 768 struct mbuf *m, *n; 769 struct mbuf *free_chain = NULL; 770 int flags, len, error, offset; 771 struct protosw *pr = so->so_proto; 772 int moff, type = 0; 773 int resid, orig_resid; 774 775 if (uio) 776 resid = uio->uio_resid; 777 else 778 resid = (int)(sio->sb_climit - sio->sb_cc); 779 orig_resid = resid; 780 781 if (psa) 782 *psa = NULL; 783 if (controlp) 784 *controlp = NULL; 785 if (flagsp) 786 flags = *flagsp &~ MSG_EOR; 787 else 788 flags = 0; 789 if (flags & MSG_OOB) { 790 m = m_get(MB_WAIT, MT_DATA); 791 if (m == NULL) 792 return (ENOBUFS); 793 error = so_pru_rcvoob(so, m, flags & MSG_PEEK); 794 if (error) 795 goto bad; 796 if (sio) { 797 do { 798 sbappend(sio, m); 799 resid -= m->m_len; 800 } while (resid > 0 && m); 801 } else { 802 do { 803 uio->uio_resid = resid; 804 error = uiomove(mtod(m, caddr_t), 805 (int)min(resid, m->m_len), uio); 806 resid = uio->uio_resid; 807 m = m_free(m); 808 } while (uio->uio_resid && error == 0 && m); 809 } 810 bad: 811 if (m) 812 m_freem(m); 813 return (error); 814 } 815 if (so->so_state & SS_ISCONFIRMING && resid) 816 so_pru_rcvd(so, 0); 817 818 restart: 819 crit_enter(); 820 error = ssb_lock(&so->so_rcv, SBLOCKWAIT(flags)); 821 if (error) 822 goto done; 823 824 m = so->so_rcv.ssb_mb; 825 /* 826 * If we have less data than requested, block awaiting more 827 * (subject to any timeout) if: 828 * 1. the current count is less than the low water mark, or 829 * 2. MSG_WAITALL is set, and it is possible to do the entire 830 * receive operation at once if we block (resid <= hiwat). 831 * 3. MSG_DONTWAIT is not set 832 * If MSG_WAITALL is set but resid is larger than the receive buffer, 833 * we have to do the receive in sections, and thus risk returning 834 * a short count if a timeout or signal occurs after we start. 835 */ 836 if (m == NULL || (((flags & MSG_DONTWAIT) == 0 && 837 so->so_rcv.ssb_cc < resid) && 838 (so->so_rcv.ssb_cc < so->so_rcv.ssb_lowat || 839 ((flags & MSG_WAITALL) && resid <= so->so_rcv.ssb_hiwat)) && 840 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) { 841 KASSERT(m != NULL || !so->so_rcv.ssb_cc, ("receive 1")); 842 if (so->so_error) { 843 if (m) 844 goto dontblock; 845 error = so->so_error; 846 if ((flags & MSG_PEEK) == 0) 847 so->so_error = 0; 848 goto release; 849 } 850 if (so->so_state & SS_CANTRCVMORE) { 851 if (m) 852 goto dontblock; 853 else 854 goto release; 855 } 856 for (; m; m = m->m_next) { 857 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 858 m = so->so_rcv.ssb_mb; 859 goto dontblock; 860 } 861 } 862 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 863 (pr->pr_flags & PR_CONNREQUIRED)) { 864 error = ENOTCONN; 865 goto release; 866 } 867 if (resid == 0) 868 goto release; 869 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) { 870 error = EWOULDBLOCK; 871 goto release; 872 } 873 ssb_unlock(&so->so_rcv); 874 error = ssb_wait(&so->so_rcv); 875 if (error) 876 goto done; 877 crit_exit(); 878 goto restart; 879 } 880 dontblock: 881 if (uio && uio->uio_td && uio->uio_td->td_proc) 882 uio->uio_td->td_lwp->lwp_ru.ru_msgrcv++; 883 884 /* 885 * note: m should be == sb_mb here. Cache the next record while 886 * cleaning up. Note that calling m_free*() will break out critical 887 * section. 888 */ 889 KKASSERT(m == so->so_rcv.ssb_mb); 890 891 /* 892 * Skip any address mbufs prepending the record. 893 */ 894 if (pr->pr_flags & PR_ADDR) { 895 KASSERT(m->m_type == MT_SONAME, ("receive 1a")); 896 orig_resid = 0; 897 if (psa) 898 *psa = dup_sockaddr(mtod(m, struct sockaddr *)); 899 if (flags & MSG_PEEK) 900 m = m->m_next; 901 else 902 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain); 903 } 904 905 /* 906 * Skip any control mbufs prepending the record. 907 */ 908 #ifdef SCTP 909 if (pr->pr_flags & PR_ADDR_OPT) { 910 /* 911 * For SCTP we may be getting a 912 * whole message OR a partial delivery. 913 */ 914 if (m && m->m_type == MT_SONAME) { 915 orig_resid = 0; 916 if (psa) 917 *psa = dup_sockaddr(mtod(m, struct sockaddr *)); 918 if (flags & MSG_PEEK) 919 m = m->m_next; 920 else 921 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain); 922 } 923 } 924 #endif /* SCTP */ 925 while (m && m->m_type == MT_CONTROL && error == 0) { 926 if (flags & MSG_PEEK) { 927 if (controlp) 928 *controlp = m_copy(m, 0, m->m_len); 929 m = m->m_next; /* XXX race */ 930 } else { 931 if (controlp) { 932 n = sbunlinkmbuf(&so->so_rcv.sb, m, NULL); 933 if (pr->pr_domain->dom_externalize && 934 mtod(m, struct cmsghdr *)->cmsg_type == 935 SCM_RIGHTS) 936 error = (*pr->pr_domain->dom_externalize)(m); 937 *controlp = m; 938 m = n; 939 } else { 940 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain); 941 } 942 } 943 if (controlp && *controlp) { 944 orig_resid = 0; 945 controlp = &(*controlp)->m_next; 946 } 947 } 948 949 /* 950 * flag OOB data. 951 */ 952 if (m) { 953 type = m->m_type; 954 if (type == MT_OOBDATA) 955 flags |= MSG_OOB; 956 } 957 958 /* 959 * Copy to the UIO or mbuf return chain (*mp). 960 */ 961 moff = 0; 962 offset = 0; 963 while (m && resid > 0 && error == 0) { 964 if (m->m_type == MT_OOBDATA) { 965 if (type != MT_OOBDATA) 966 break; 967 } else if (type == MT_OOBDATA) 968 break; 969 else 970 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER, 971 ("receive 3")); 972 so->so_state &= ~SS_RCVATMARK; 973 len = resid; 974 if (so->so_oobmark && len > so->so_oobmark - offset) 975 len = so->so_oobmark - offset; 976 if (len > m->m_len - moff) 977 len = m->m_len - moff; 978 979 /* 980 * Copy out to the UIO or pass the mbufs back to the SIO. 981 * The SIO is dealt with when we eat the mbuf, but deal 982 * with the resid here either way. 983 */ 984 if (uio) { 985 crit_exit(); 986 uio->uio_resid = resid; 987 error = uiomove(mtod(m, caddr_t) + moff, len, uio); 988 resid = uio->uio_resid; 989 crit_enter(); 990 if (error) 991 goto release; 992 } else { 993 resid -= len; 994 } 995 996 /* 997 * Eat the entire mbuf or just a piece of it 998 */ 999 if (len == m->m_len - moff) { 1000 if (m->m_flags & M_EOR) 1001 flags |= MSG_EOR; 1002 #ifdef SCTP 1003 if (m->m_flags & M_NOTIFICATION) 1004 flags |= MSG_NOTIFICATION; 1005 #endif /* SCTP */ 1006 if (flags & MSG_PEEK) { 1007 m = m->m_next; 1008 moff = 0; 1009 } else { 1010 if (sio) { 1011 n = sbunlinkmbuf(&so->so_rcv.sb, m, NULL); 1012 sbappend(sio, m); 1013 m = n; 1014 } else { 1015 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain); 1016 } 1017 } 1018 } else { 1019 if (flags & MSG_PEEK) { 1020 moff += len; 1021 } else { 1022 if (sio) { 1023 n = m_copym(m, 0, len, MB_WAIT); 1024 if (n) 1025 sbappend(sio, n); 1026 } 1027 m->m_data += len; 1028 m->m_len -= len; 1029 so->so_rcv.ssb_cc -= len; 1030 } 1031 } 1032 if (so->so_oobmark) { 1033 if ((flags & MSG_PEEK) == 0) { 1034 so->so_oobmark -= len; 1035 if (so->so_oobmark == 0) { 1036 so->so_state |= SS_RCVATMARK; 1037 break; 1038 } 1039 } else { 1040 offset += len; 1041 if (offset == so->so_oobmark) 1042 break; 1043 } 1044 } 1045 if (flags & MSG_EOR) 1046 break; 1047 /* 1048 * If the MSG_WAITALL flag is set (for non-atomic socket), 1049 * we must not quit until resid == 0 or an error 1050 * termination. If a signal/timeout occurs, return 1051 * with a short count but without error. 1052 * Keep signalsockbuf locked against other readers. 1053 */ 1054 while ((flags & MSG_WAITALL) && m == NULL && 1055 resid > 0 && !sosendallatonce(so) && 1056 so->so_rcv.ssb_mb == NULL) { 1057 if (so->so_error || so->so_state & SS_CANTRCVMORE) 1058 break; 1059 /* 1060 * The window might have closed to zero, make 1061 * sure we send an ack now that we've drained 1062 * the buffer or we might end up blocking until 1063 * the idle takes over (5 seconds). 1064 */ 1065 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1066 so_pru_rcvd(so, flags); 1067 error = ssb_wait(&so->so_rcv); 1068 if (error) { 1069 ssb_unlock(&so->so_rcv); 1070 error = 0; 1071 goto done; 1072 } 1073 m = so->so_rcv.ssb_mb; 1074 } 1075 } 1076 1077 /* 1078 * If an atomic read was requested but unread data still remains 1079 * in the record, set MSG_TRUNC. 1080 */ 1081 if (m && pr->pr_flags & PR_ATOMIC) 1082 flags |= MSG_TRUNC; 1083 1084 /* 1085 * Cleanup. If an atomic read was requested drop any unread data. 1086 */ 1087 if ((flags & MSG_PEEK) == 0) { 1088 if (m && (pr->pr_flags & PR_ATOMIC)) 1089 sbdroprecord(&so->so_rcv.sb); 1090 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) 1091 so_pru_rcvd(so, flags); 1092 } 1093 1094 if (orig_resid == resid && orig_resid && 1095 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 1096 ssb_unlock(&so->so_rcv); 1097 crit_exit(); 1098 goto restart; 1099 } 1100 1101 if (flagsp) 1102 *flagsp |= flags; 1103 release: 1104 ssb_unlock(&so->so_rcv); 1105 done: 1106 crit_exit(); 1107 if (free_chain) 1108 m_freem(free_chain); 1109 return (error); 1110 } 1111 1112 int 1113 soshutdown(struct socket *so, int how) 1114 { 1115 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 1116 return (EINVAL); 1117 1118 if (how != SHUT_WR) 1119 sorflush(so); 1120 if (how != SHUT_RD) 1121 return (so_pru_shutdown(so)); 1122 return (0); 1123 } 1124 1125 void 1126 sorflush(struct socket *so) 1127 { 1128 struct signalsockbuf *ssb = &so->so_rcv; 1129 struct protosw *pr = so->so_proto; 1130 struct signalsockbuf asb; 1131 1132 ssb->ssb_flags |= SSB_NOINTR; 1133 (void) ssb_lock(ssb, M_WAITOK); 1134 1135 crit_enter(); 1136 socantrcvmore(so); 1137 ssb_unlock(ssb); 1138 asb = *ssb; 1139 bzero((caddr_t)ssb, sizeof (*ssb)); 1140 if (asb.ssb_flags & SSB_KNOTE) { 1141 ssb->ssb_sel.si_note = asb.ssb_sel.si_note; 1142 ssb->ssb_flags = SSB_KNOTE; 1143 } 1144 crit_exit(); 1145 1146 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) 1147 (*pr->pr_domain->dom_dispose)(asb.ssb_mb); 1148 ssb_release(&asb, so); 1149 } 1150 1151 #ifdef INET 1152 static int 1153 do_setopt_accept_filter(struct socket *so, struct sockopt *sopt) 1154 { 1155 struct accept_filter_arg *afap = NULL; 1156 struct accept_filter *afp; 1157 struct so_accf *af = so->so_accf; 1158 int error = 0; 1159 1160 /* do not set/remove accept filters on non listen sockets */ 1161 if ((so->so_options & SO_ACCEPTCONN) == 0) { 1162 error = EINVAL; 1163 goto out; 1164 } 1165 1166 /* removing the filter */ 1167 if (sopt == NULL) { 1168 if (af != NULL) { 1169 if (af->so_accept_filter != NULL && 1170 af->so_accept_filter->accf_destroy != NULL) { 1171 af->so_accept_filter->accf_destroy(so); 1172 } 1173 if (af->so_accept_filter_str != NULL) { 1174 FREE(af->so_accept_filter_str, M_ACCF); 1175 } 1176 FREE(af, M_ACCF); 1177 so->so_accf = NULL; 1178 } 1179 so->so_options &= ~SO_ACCEPTFILTER; 1180 return (0); 1181 } 1182 /* adding a filter */ 1183 /* must remove previous filter first */ 1184 if (af != NULL) { 1185 error = EINVAL; 1186 goto out; 1187 } 1188 /* don't put large objects on the kernel stack */ 1189 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK); 1190 error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap); 1191 afap->af_name[sizeof(afap->af_name)-1] = '\0'; 1192 afap->af_arg[sizeof(afap->af_arg)-1] = '\0'; 1193 if (error) 1194 goto out; 1195 afp = accept_filt_get(afap->af_name); 1196 if (afp == NULL) { 1197 error = ENOENT; 1198 goto out; 1199 } 1200 MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK); 1201 bzero(af, sizeof(*af)); 1202 if (afp->accf_create != NULL) { 1203 if (afap->af_name[0] != '\0') { 1204 int len = strlen(afap->af_name) + 1; 1205 1206 MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK); 1207 strcpy(af->so_accept_filter_str, afap->af_name); 1208 } 1209 af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg); 1210 if (af->so_accept_filter_arg == NULL) { 1211 FREE(af->so_accept_filter_str, M_ACCF); 1212 FREE(af, M_ACCF); 1213 so->so_accf = NULL; 1214 error = EINVAL; 1215 goto out; 1216 } 1217 } 1218 af->so_accept_filter = afp; 1219 so->so_accf = af; 1220 so->so_options |= SO_ACCEPTFILTER; 1221 out: 1222 if (afap != NULL) 1223 FREE(afap, M_TEMP); 1224 return (error); 1225 } 1226 #endif /* INET */ 1227 1228 /* 1229 * Perhaps this routine, and sooptcopyout(), below, ought to come in 1230 * an additional variant to handle the case where the option value needs 1231 * to be some kind of integer, but not a specific size. 1232 * In addition to their use here, these functions are also called by the 1233 * protocol-level pr_ctloutput() routines. 1234 */ 1235 int 1236 sooptcopyin(struct sockopt *sopt, void *buf, size_t len, size_t minlen) 1237 { 1238 size_t valsize; 1239 1240 /* 1241 * If the user gives us more than we wanted, we ignore it, 1242 * but if we don't get the minimum length the caller 1243 * wants, we return EINVAL. On success, sopt->sopt_valsize 1244 * is set to however much we actually retrieved. 1245 */ 1246 if ((valsize = sopt->sopt_valsize) < minlen) 1247 return EINVAL; 1248 if (valsize > len) 1249 sopt->sopt_valsize = valsize = len; 1250 1251 if (sopt->sopt_td != NULL) 1252 return (copyin(sopt->sopt_val, buf, valsize)); 1253 1254 bcopy(sopt->sopt_val, buf, valsize); 1255 return 0; 1256 } 1257 1258 int 1259 sosetopt(struct socket *so, struct sockopt *sopt) 1260 { 1261 int error, optval; 1262 struct linger l; 1263 struct timeval tv; 1264 u_long val; 1265 1266 error = 0; 1267 sopt->sopt_dir = SOPT_SET; 1268 if (sopt->sopt_level != SOL_SOCKET) { 1269 if (so->so_proto && so->so_proto->pr_ctloutput) { 1270 return (so_pr_ctloutput(so, sopt)); 1271 } 1272 error = ENOPROTOOPT; 1273 } else { 1274 switch (sopt->sopt_name) { 1275 #ifdef INET 1276 case SO_ACCEPTFILTER: 1277 error = do_setopt_accept_filter(so, sopt); 1278 if (error) 1279 goto bad; 1280 break; 1281 #endif /* INET */ 1282 case SO_LINGER: 1283 error = sooptcopyin(sopt, &l, sizeof l, sizeof l); 1284 if (error) 1285 goto bad; 1286 1287 so->so_linger = l.l_linger; 1288 if (l.l_onoff) 1289 so->so_options |= SO_LINGER; 1290 else 1291 so->so_options &= ~SO_LINGER; 1292 break; 1293 1294 case SO_DEBUG: 1295 case SO_KEEPALIVE: 1296 case SO_DONTROUTE: 1297 case SO_USELOOPBACK: 1298 case SO_BROADCAST: 1299 case SO_REUSEADDR: 1300 case SO_REUSEPORT: 1301 case SO_OOBINLINE: 1302 case SO_TIMESTAMP: 1303 error = sooptcopyin(sopt, &optval, sizeof optval, 1304 sizeof optval); 1305 if (error) 1306 goto bad; 1307 if (optval) 1308 so->so_options |= sopt->sopt_name; 1309 else 1310 so->so_options &= ~sopt->sopt_name; 1311 break; 1312 1313 case SO_SNDBUF: 1314 case SO_RCVBUF: 1315 case SO_SNDLOWAT: 1316 case SO_RCVLOWAT: 1317 error = sooptcopyin(sopt, &optval, sizeof optval, 1318 sizeof optval); 1319 if (error) 1320 goto bad; 1321 1322 /* 1323 * Values < 1 make no sense for any of these 1324 * options, so disallow them. 1325 */ 1326 if (optval < 1) { 1327 error = EINVAL; 1328 goto bad; 1329 } 1330 1331 switch (sopt->sopt_name) { 1332 case SO_SNDBUF: 1333 case SO_RCVBUF: 1334 if (ssb_reserve(sopt->sopt_name == SO_SNDBUF ? 1335 &so->so_snd : &so->so_rcv, (u_long)optval, 1336 so, 1337 &curproc->p_rlimit[RLIMIT_SBSIZE]) == 0) { 1338 error = ENOBUFS; 1339 goto bad; 1340 } 1341 break; 1342 1343 /* 1344 * Make sure the low-water is never greater than 1345 * the high-water. 1346 */ 1347 case SO_SNDLOWAT: 1348 so->so_snd.ssb_lowat = 1349 (optval > so->so_snd.ssb_hiwat) ? 1350 so->so_snd.ssb_hiwat : optval; 1351 break; 1352 case SO_RCVLOWAT: 1353 so->so_rcv.ssb_lowat = 1354 (optval > so->so_rcv.ssb_hiwat) ? 1355 so->so_rcv.ssb_hiwat : optval; 1356 break; 1357 } 1358 break; 1359 1360 case SO_SNDTIMEO: 1361 case SO_RCVTIMEO: 1362 error = sooptcopyin(sopt, &tv, sizeof tv, 1363 sizeof tv); 1364 if (error) 1365 goto bad; 1366 1367 /* assert(hz > 0); */ 1368 if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz || 1369 tv.tv_usec < 0 || tv.tv_usec >= 1000000) { 1370 error = EDOM; 1371 goto bad; 1372 } 1373 /* assert(tick > 0); */ 1374 /* assert(ULONG_MAX - SHRT_MAX >= 1000000); */ 1375 val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick; 1376 if (val > SHRT_MAX) { 1377 error = EDOM; 1378 goto bad; 1379 } 1380 if (val == 0 && tv.tv_usec != 0) 1381 val = 1; 1382 1383 switch (sopt->sopt_name) { 1384 case SO_SNDTIMEO: 1385 so->so_snd.ssb_timeo = val; 1386 break; 1387 case SO_RCVTIMEO: 1388 so->so_rcv.ssb_timeo = val; 1389 break; 1390 } 1391 break; 1392 default: 1393 error = ENOPROTOOPT; 1394 break; 1395 } 1396 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) { 1397 (void) so_pr_ctloutput(so, sopt); 1398 } 1399 } 1400 bad: 1401 return (error); 1402 } 1403 1404 /* Helper routine for getsockopt */ 1405 int 1406 sooptcopyout(struct sockopt *sopt, const void *buf, size_t len) 1407 { 1408 int error; 1409 size_t valsize; 1410 1411 error = 0; 1412 1413 /* 1414 * Documented get behavior is that we always return a value, 1415 * possibly truncated to fit in the user's buffer. 1416 * Traditional behavior is that we always tell the user 1417 * precisely how much we copied, rather than something useful 1418 * like the total amount we had available for her. 1419 * Note that this interface is not idempotent; the entire answer must 1420 * generated ahead of time. 1421 */ 1422 valsize = min(len, sopt->sopt_valsize); 1423 sopt->sopt_valsize = valsize; 1424 if (sopt->sopt_val != 0) { 1425 if (sopt->sopt_td != NULL) 1426 error = copyout(buf, sopt->sopt_val, valsize); 1427 else 1428 bcopy(buf, sopt->sopt_val, valsize); 1429 } 1430 return error; 1431 } 1432 1433 int 1434 sogetopt(struct socket *so, struct sockopt *sopt) 1435 { 1436 int error, optval; 1437 struct linger l; 1438 struct timeval tv; 1439 #ifdef INET 1440 struct accept_filter_arg *afap; 1441 #endif 1442 1443 error = 0; 1444 sopt->sopt_dir = SOPT_GET; 1445 if (sopt->sopt_level != SOL_SOCKET) { 1446 if (so->so_proto && so->so_proto->pr_ctloutput) { 1447 return (so_pr_ctloutput(so, sopt)); 1448 } else 1449 return (ENOPROTOOPT); 1450 } else { 1451 switch (sopt->sopt_name) { 1452 #ifdef INET 1453 case SO_ACCEPTFILTER: 1454 if ((so->so_options & SO_ACCEPTCONN) == 0) 1455 return (EINVAL); 1456 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), 1457 M_TEMP, M_WAITOK); 1458 bzero(afap, sizeof(*afap)); 1459 if ((so->so_options & SO_ACCEPTFILTER) != 0) { 1460 strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name); 1461 if (so->so_accf->so_accept_filter_str != NULL) 1462 strcpy(afap->af_arg, so->so_accf->so_accept_filter_str); 1463 } 1464 error = sooptcopyout(sopt, afap, sizeof(*afap)); 1465 FREE(afap, M_TEMP); 1466 break; 1467 #endif /* INET */ 1468 1469 case SO_LINGER: 1470 l.l_onoff = so->so_options & SO_LINGER; 1471 l.l_linger = so->so_linger; 1472 error = sooptcopyout(sopt, &l, sizeof l); 1473 break; 1474 1475 case SO_USELOOPBACK: 1476 case SO_DONTROUTE: 1477 case SO_DEBUG: 1478 case SO_KEEPALIVE: 1479 case SO_REUSEADDR: 1480 case SO_REUSEPORT: 1481 case SO_BROADCAST: 1482 case SO_OOBINLINE: 1483 case SO_TIMESTAMP: 1484 optval = so->so_options & sopt->sopt_name; 1485 integer: 1486 error = sooptcopyout(sopt, &optval, sizeof optval); 1487 break; 1488 1489 case SO_TYPE: 1490 optval = so->so_type; 1491 goto integer; 1492 1493 case SO_ERROR: 1494 optval = so->so_error; 1495 so->so_error = 0; 1496 goto integer; 1497 1498 case SO_SNDBUF: 1499 optval = so->so_snd.ssb_hiwat; 1500 goto integer; 1501 1502 case SO_RCVBUF: 1503 optval = so->so_rcv.ssb_hiwat; 1504 goto integer; 1505 1506 case SO_SNDLOWAT: 1507 optval = so->so_snd.ssb_lowat; 1508 goto integer; 1509 1510 case SO_RCVLOWAT: 1511 optval = so->so_rcv.ssb_lowat; 1512 goto integer; 1513 1514 case SO_SNDTIMEO: 1515 case SO_RCVTIMEO: 1516 optval = (sopt->sopt_name == SO_SNDTIMEO ? 1517 so->so_snd.ssb_timeo : so->so_rcv.ssb_timeo); 1518 1519 tv.tv_sec = optval / hz; 1520 tv.tv_usec = (optval % hz) * tick; 1521 error = sooptcopyout(sopt, &tv, sizeof tv); 1522 break; 1523 1524 default: 1525 error = ENOPROTOOPT; 1526 break; 1527 } 1528 return (error); 1529 } 1530 } 1531 1532 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */ 1533 int 1534 soopt_getm(struct sockopt *sopt, struct mbuf **mp) 1535 { 1536 struct mbuf *m, *m_prev; 1537 int sopt_size = sopt->sopt_valsize, msize; 1538 1539 m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_DATA, 1540 0, &msize); 1541 if (m == NULL) 1542 return (ENOBUFS); 1543 m->m_len = min(msize, sopt_size); 1544 sopt_size -= m->m_len; 1545 *mp = m; 1546 m_prev = m; 1547 1548 while (sopt_size > 0) { 1549 m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, 1550 MT_DATA, 0, &msize); 1551 if (m == NULL) { 1552 m_freem(*mp); 1553 return (ENOBUFS); 1554 } 1555 m->m_len = min(msize, sopt_size); 1556 sopt_size -= m->m_len; 1557 m_prev->m_next = m; 1558 m_prev = m; 1559 } 1560 return (0); 1561 } 1562 1563 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */ 1564 int 1565 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m) 1566 { 1567 struct mbuf *m0 = m; 1568 1569 if (sopt->sopt_val == NULL) 1570 return 0; 1571 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1572 if (sopt->sopt_td != NULL) { 1573 int error; 1574 1575 error = copyin(sopt->sopt_val, mtod(m, char *), 1576 m->m_len); 1577 if (error != 0) { 1578 m_freem(m0); 1579 return (error); 1580 } 1581 } else 1582 bcopy(sopt->sopt_val, mtod(m, char *), m->m_len); 1583 sopt->sopt_valsize -= m->m_len; 1584 sopt->sopt_val = (caddr_t)sopt->sopt_val + m->m_len; 1585 m = m->m_next; 1586 } 1587 if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */ 1588 panic("ip6_sooptmcopyin"); 1589 return 0; 1590 } 1591 1592 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */ 1593 int 1594 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m) 1595 { 1596 struct mbuf *m0 = m; 1597 size_t valsize = 0; 1598 1599 if (sopt->sopt_val == NULL) 1600 return 0; 1601 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1602 if (sopt->sopt_td != NULL) { 1603 int error; 1604 1605 error = copyout(mtod(m, char *), sopt->sopt_val, 1606 m->m_len); 1607 if (error != 0) { 1608 m_freem(m0); 1609 return (error); 1610 } 1611 } else 1612 bcopy(mtod(m, char *), sopt->sopt_val, m->m_len); 1613 sopt->sopt_valsize -= m->m_len; 1614 sopt->sopt_val = (caddr_t)sopt->sopt_val + m->m_len; 1615 valsize += m->m_len; 1616 m = m->m_next; 1617 } 1618 if (m != NULL) { 1619 /* enough soopt buffer should be given from user-land */ 1620 m_freem(m0); 1621 return (EINVAL); 1622 } 1623 sopt->sopt_valsize = valsize; 1624 return 0; 1625 } 1626 1627 void 1628 sohasoutofband(struct socket *so) 1629 { 1630 if (so->so_sigio != NULL) 1631 pgsigio(so->so_sigio, SIGURG, 0); 1632 selwakeup(&so->so_rcv.ssb_sel); 1633 } 1634 1635 int 1636 sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td) 1637 { 1638 int revents = 0; 1639 1640 crit_enter(); 1641 1642 if (events & (POLLIN | POLLRDNORM)) 1643 if (soreadable(so)) 1644 revents |= events & (POLLIN | POLLRDNORM); 1645 1646 if (events & POLLINIGNEOF) 1647 if (so->so_rcv.ssb_cc >= so->so_rcv.ssb_lowat || 1648 !TAILQ_EMPTY(&so->so_comp) || so->so_error) 1649 revents |= POLLINIGNEOF; 1650 1651 if (events & (POLLOUT | POLLWRNORM)) 1652 if (sowriteable(so)) 1653 revents |= events & (POLLOUT | POLLWRNORM); 1654 1655 if (events & (POLLPRI | POLLRDBAND)) 1656 if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) 1657 revents |= events & (POLLPRI | POLLRDBAND); 1658 1659 if (revents == 0) { 1660 if (events & 1661 (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM | 1662 POLLRDBAND)) { 1663 selrecord(td, &so->so_rcv.ssb_sel); 1664 so->so_rcv.ssb_flags |= SSB_SEL; 1665 } 1666 1667 if (events & (POLLOUT | POLLWRNORM)) { 1668 selrecord(td, &so->so_snd.ssb_sel); 1669 so->so_snd.ssb_flags |= SSB_SEL; 1670 } 1671 } 1672 1673 crit_exit(); 1674 return (revents); 1675 } 1676 1677 int 1678 sokqfilter(struct file *fp, struct knote *kn) 1679 { 1680 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1681 struct signalsockbuf *ssb; 1682 1683 switch (kn->kn_filter) { 1684 case EVFILT_READ: 1685 if (so->so_options & SO_ACCEPTCONN) 1686 kn->kn_fop = &solisten_filtops; 1687 else 1688 kn->kn_fop = &soread_filtops; 1689 ssb = &so->so_rcv; 1690 break; 1691 case EVFILT_WRITE: 1692 kn->kn_fop = &sowrite_filtops; 1693 ssb = &so->so_snd; 1694 break; 1695 default: 1696 return (1); 1697 } 1698 1699 crit_enter(); 1700 SLIST_INSERT_HEAD(&ssb->ssb_sel.si_note, kn, kn_selnext); 1701 ssb->ssb_flags |= SSB_KNOTE; 1702 crit_exit(); 1703 return (0); 1704 } 1705 1706 static void 1707 filt_sordetach(struct knote *kn) 1708 { 1709 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1710 1711 crit_enter(); 1712 SLIST_REMOVE(&so->so_rcv.ssb_sel.si_note, kn, knote, kn_selnext); 1713 if (SLIST_EMPTY(&so->so_rcv.ssb_sel.si_note)) 1714 so->so_rcv.ssb_flags &= ~SSB_KNOTE; 1715 crit_exit(); 1716 } 1717 1718 /*ARGSUSED*/ 1719 static int 1720 filt_soread(struct knote *kn, long hint) 1721 { 1722 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1723 1724 kn->kn_data = so->so_rcv.ssb_cc; 1725 if (so->so_state & SS_CANTRCVMORE) { 1726 kn->kn_flags |= EV_EOF; 1727 kn->kn_fflags = so->so_error; 1728 return (1); 1729 } 1730 if (so->so_error) /* temporary udp error */ 1731 return (1); 1732 if (kn->kn_sfflags & NOTE_LOWAT) 1733 return (kn->kn_data >= kn->kn_sdata); 1734 return (kn->kn_data >= so->so_rcv.ssb_lowat); 1735 } 1736 1737 static void 1738 filt_sowdetach(struct knote *kn) 1739 { 1740 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1741 1742 crit_enter(); 1743 SLIST_REMOVE(&so->so_snd.ssb_sel.si_note, kn, knote, kn_selnext); 1744 if (SLIST_EMPTY(&so->so_snd.ssb_sel.si_note)) 1745 so->so_snd.ssb_flags &= ~SSB_KNOTE; 1746 crit_exit(); 1747 } 1748 1749 /*ARGSUSED*/ 1750 static int 1751 filt_sowrite(struct knote *kn, long hint) 1752 { 1753 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1754 1755 kn->kn_data = ssb_space(&so->so_snd); 1756 if (so->so_state & SS_CANTSENDMORE) { 1757 kn->kn_flags |= EV_EOF; 1758 kn->kn_fflags = so->so_error; 1759 return (1); 1760 } 1761 if (so->so_error) /* temporary udp error */ 1762 return (1); 1763 if (((so->so_state & SS_ISCONNECTED) == 0) && 1764 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 1765 return (0); 1766 if (kn->kn_sfflags & NOTE_LOWAT) 1767 return (kn->kn_data >= kn->kn_sdata); 1768 return (kn->kn_data >= so->so_snd.ssb_lowat); 1769 } 1770 1771 /*ARGSUSED*/ 1772 static int 1773 filt_solisten(struct knote *kn, long hint) 1774 { 1775 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1776 1777 kn->kn_data = so->so_qlen; 1778 return (! TAILQ_EMPTY(&so->so_comp)); 1779 } 1780