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) 2004 Jeffrey M. Hsu. All rights reserved. 36 * 37 * License terms: all terms for the DragonFly license above plus the following: 38 * 39 * 4. All advertising materials mentioning features or use of this software 40 * must display the following acknowledgement: 41 * 42 * This product includes software developed by Jeffrey M. Hsu 43 * for the DragonFly Project. 44 * 45 * This requirement may be waived with permission from Jeffrey Hsu. 46 * This requirement will sunset and may be removed on July 8 2005, 47 * after which the standard DragonFly license (as shown above) will 48 * apply. 49 */ 50 51 /* 52 * Copyright (c) 1982, 1986, 1988, 1990, 1993 53 * The Regents of the University of California. All rights reserved. 54 * 55 * Redistribution and use in source and binary forms, with or without 56 * modification, are permitted provided that the following conditions 57 * are met: 58 * 1. Redistributions of source code must retain the above copyright 59 * notice, this list of conditions and the following disclaimer. 60 * 2. Redistributions in binary form must reproduce the above copyright 61 * notice, this list of conditions and the following disclaimer in the 62 * documentation and/or other materials provided with the distribution. 63 * 3. All advertising materials mentioning features or use of this software 64 * must display the following acknowledgement: 65 * This product includes software developed by the University of 66 * California, Berkeley and its contributors. 67 * 4. Neither the name of the University nor the names of its contributors 68 * may be used to endorse or promote products derived from this software 69 * without specific prior written permission. 70 * 71 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 72 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 73 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 74 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 75 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 76 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 77 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 78 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 79 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 80 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 81 * SUCH DAMAGE. 82 * 83 * @(#)uipc_socket.c 8.3 (Berkeley) 4/15/94 84 * $FreeBSD: src/sys/kern/uipc_socket.c,v 1.68.2.24 2003/11/11 17:18:18 silby Exp $ 85 * $DragonFly: src/sys/kern/uipc_socket.c,v 1.42 2007/01/01 22:51:17 corecode Exp $ 86 */ 87 88 #include "opt_inet.h" 89 #include "opt_sctp.h" 90 91 #include <sys/param.h> 92 #include <sys/systm.h> 93 #include <sys/fcntl.h> 94 #include <sys/malloc.h> 95 #include <sys/mbuf.h> 96 #include <sys/domain.h> 97 #include <sys/file.h> /* for struct knote */ 98 #include <sys/kernel.h> 99 #include <sys/malloc.h> 100 #include <sys/event.h> 101 #include <sys/poll.h> 102 #include <sys/proc.h> 103 #include <sys/protosw.h> 104 #include <sys/socket.h> 105 #include <sys/socketvar.h> 106 #include <sys/socketops.h> 107 #include <sys/resourcevar.h> 108 #include <sys/signalvar.h> 109 #include <sys/sysctl.h> 110 #include <sys/uio.h> 111 #include <sys/jail.h> 112 #include <vm/vm_zone.h> 113 114 #include <sys/thread2.h> 115 116 #include <machine/limits.h> 117 118 #ifdef INET 119 static int do_setopt_accept_filter(struct socket *so, struct sockopt *sopt); 120 #endif /* INET */ 121 122 static void filt_sordetach(struct knote *kn); 123 static int filt_soread(struct knote *kn, long hint); 124 static void filt_sowdetach(struct knote *kn); 125 static int filt_sowrite(struct knote *kn, long hint); 126 static int filt_solisten(struct knote *kn, long hint); 127 128 static struct filterops solisten_filtops = 129 { 1, NULL, filt_sordetach, filt_solisten }; 130 static struct filterops soread_filtops = 131 { 1, NULL, filt_sordetach, filt_soread }; 132 static struct filterops sowrite_filtops = 133 { 1, NULL, filt_sowdetach, filt_sowrite }; 134 135 struct vm_zone *socket_zone; 136 137 MALLOC_DEFINE(M_SONAME, "soname", "socket name"); 138 MALLOC_DEFINE(M_PCB, "pcb", "protocol control block"); 139 140 141 static int somaxconn = SOMAXCONN; 142 SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW, 143 &somaxconn, 0, "Maximum pending socket connection queue size"); 144 145 /* 146 * Socket operation routines. 147 * These routines are called by the routines in 148 * sys_socket.c or from a system process, and 149 * implement the semantics of socket operations by 150 * switching out to the protocol specific routines. 151 */ 152 153 /* 154 * Get a socket structure from our zone, and initialize it. 155 * We don't implement `waitok' yet (see comments in uipc_domain.c). 156 * Note that it would probably be better to allocate socket 157 * and PCB at the same time, but I'm not convinced that all 158 * the protocols can be easily modified to do this. 159 */ 160 struct socket * 161 soalloc(int waitok) 162 { 163 struct socket *so; 164 165 so = zalloc(socket_zone); 166 if (so) { 167 /* XXX race condition for reentrant kernel */ 168 bzero(so, sizeof *so); 169 TAILQ_INIT(&so->so_aiojobq); 170 TAILQ_INIT(&so->so_rcv.sb_sel.si_mlist); 171 TAILQ_INIT(&so->so_snd.sb_sel.si_mlist); 172 } 173 return so; 174 } 175 176 int 177 socreate(int dom, struct socket **aso, int type, 178 int proto, struct thread *td) 179 { 180 struct proc *p = td->td_proc; 181 struct protosw *prp; 182 struct socket *so; 183 struct pru_attach_info ai; 184 int error; 185 186 if (proto) 187 prp = pffindproto(dom, proto, type); 188 else 189 prp = pffindtype(dom, type); 190 191 if (prp == 0 || prp->pr_usrreqs->pru_attach == 0) 192 return (EPROTONOSUPPORT); 193 194 if (p->p_ucred->cr_prison && jail_socket_unixiproute_only && 195 prp->pr_domain->dom_family != PF_LOCAL && 196 prp->pr_domain->dom_family != PF_INET && 197 prp->pr_domain->dom_family != PF_INET6 && 198 prp->pr_domain->dom_family != PF_ROUTE) { 199 return (EPROTONOSUPPORT); 200 } 201 202 if (prp->pr_type != type) 203 return (EPROTOTYPE); 204 so = soalloc(p != 0); 205 if (so == 0) 206 return (ENOBUFS); 207 208 TAILQ_INIT(&so->so_incomp); 209 TAILQ_INIT(&so->so_comp); 210 so->so_type = type; 211 so->so_cred = crhold(p->p_ucred); 212 so->so_proto = prp; 213 ai.sb_rlimit = &p->p_rlimit[RLIMIT_SBSIZE]; 214 ai.p_ucred = p->p_ucred; 215 ai.fd_rdir = p->p_fd->fd_rdir; 216 error = so_pru_attach(so, proto, &ai); 217 if (error) { 218 so->so_state |= SS_NOFDREF; 219 sofree(so); 220 return (error); 221 } 222 *aso = so; 223 return (0); 224 } 225 226 int 227 sobind(struct socket *so, struct sockaddr *nam, struct thread *td) 228 { 229 int error; 230 231 crit_enter(); 232 error = so_pru_bind(so, nam, td); 233 crit_exit(); 234 return (error); 235 } 236 237 void 238 sodealloc(struct socket *so) 239 { 240 if (so->so_rcv.sb_hiwat) 241 (void)chgsbsize(so->so_cred->cr_uidinfo, 242 &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY); 243 if (so->so_snd.sb_hiwat) 244 (void)chgsbsize(so->so_cred->cr_uidinfo, 245 &so->so_snd.sb_hiwat, 0, RLIM_INFINITY); 246 #ifdef INET 247 /* remove accept filter if present */ 248 if (so->so_accf != NULL) 249 do_setopt_accept_filter(so, NULL); 250 #endif /* INET */ 251 crfree(so->so_cred); 252 zfree(socket_zone, so); 253 } 254 255 int 256 solisten(struct socket *so, int backlog, struct thread *td) 257 { 258 int error; 259 #ifdef SCTP 260 short oldopt, oldqlimit; 261 #endif /* SCTP */ 262 263 crit_enter(); 264 if (so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING)) { 265 crit_exit(); 266 return (EINVAL); 267 } 268 269 #ifdef SCTP 270 oldopt = so->so_options; 271 oldqlimit = so->so_qlimit; 272 #endif /* SCTP */ 273 274 if (TAILQ_EMPTY(&so->so_comp)) 275 so->so_options |= SO_ACCEPTCONN; 276 if (backlog < 0 || backlog > somaxconn) 277 backlog = somaxconn; 278 so->so_qlimit = backlog; 279 /* SCTP needs to look at tweak both the inbound backlog parameter AND 280 * the so_options (UDP model both connect's and gets inbound 281 * connections .. implicitly). 282 */ 283 error = so_pru_listen(so, td); 284 if (error) { 285 #ifdef SCTP 286 /* Restore the params */ 287 so->so_options = oldopt; 288 so->so_qlimit = oldqlimit; 289 #endif /* SCTP */ 290 crit_exit(); 291 return (error); 292 } 293 crit_exit(); 294 return (0); 295 } 296 297 void 298 sofree(struct socket *so) 299 { 300 struct socket *head = so->so_head; 301 302 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) 303 return; 304 if (head != NULL) { 305 if (so->so_state & SS_INCOMP) { 306 TAILQ_REMOVE(&head->so_incomp, so, so_list); 307 head->so_incqlen--; 308 } else if (so->so_state & SS_COMP) { 309 /* 310 * We must not decommission a socket that's 311 * on the accept(2) queue. If we do, then 312 * accept(2) may hang after select(2) indicated 313 * that the listening socket was ready. 314 */ 315 return; 316 } else { 317 panic("sofree: not queued"); 318 } 319 so->so_state &= ~SS_INCOMP; 320 so->so_head = NULL; 321 } 322 sbrelease(&so->so_snd, so); 323 sorflush(so); 324 sodealloc(so); 325 } 326 327 /* 328 * Close a socket on last file table reference removal. 329 * Initiate disconnect if connected. 330 * Free socket when disconnect complete. 331 */ 332 int 333 soclose(struct socket *so, int fflag) 334 { 335 int error = 0; 336 337 crit_enter(); 338 funsetown(so->so_sigio); 339 if (so->so_pcb == NULL) 340 goto discard; 341 if (so->so_state & SS_ISCONNECTED) { 342 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 343 error = sodisconnect(so); 344 if (error) 345 goto drop; 346 } 347 if (so->so_options & SO_LINGER) { 348 if ((so->so_state & SS_ISDISCONNECTING) && 349 (fflag & FNONBLOCK)) 350 goto drop; 351 while (so->so_state & SS_ISCONNECTED) { 352 error = tsleep((caddr_t)&so->so_timeo, 353 PCATCH, "soclos", so->so_linger * hz); 354 if (error) 355 break; 356 } 357 } 358 } 359 drop: 360 if (so->so_pcb) { 361 int error2; 362 363 error2 = so_pru_detach(so); 364 if (error == 0) 365 error = error2; 366 } 367 discard: 368 if (so->so_options & SO_ACCEPTCONN) { 369 struct socket *sp, *sonext; 370 371 sp = TAILQ_FIRST(&so->so_incomp); 372 for (; sp != NULL; sp = sonext) { 373 sonext = TAILQ_NEXT(sp, so_list); 374 (void) soabort(sp); 375 } 376 for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) { 377 sonext = TAILQ_NEXT(sp, so_list); 378 /* Dequeue from so_comp since sofree() won't do it */ 379 TAILQ_REMOVE(&so->so_comp, sp, so_list); 380 so->so_qlen--; 381 sp->so_state &= ~SS_COMP; 382 sp->so_head = NULL; 383 (void) soabort(sp); 384 } 385 } 386 if (so->so_state & SS_NOFDREF) 387 panic("soclose: NOFDREF"); 388 so->so_state |= SS_NOFDREF; 389 sofree(so); 390 crit_exit(); 391 return (error); 392 } 393 394 /* 395 * Must be called from a critical section. 396 */ 397 int 398 soabort(struct socket *so) 399 { 400 int error; 401 402 error = so_pru_abort(so); 403 if (error) { 404 sofree(so); 405 return error; 406 } 407 return (0); 408 } 409 410 int 411 soaccept(struct socket *so, struct sockaddr **nam) 412 { 413 int error; 414 415 crit_enter(); 416 if ((so->so_state & SS_NOFDREF) == 0) 417 panic("soaccept: !NOFDREF"); 418 so->so_state &= ~SS_NOFDREF; 419 error = so_pru_accept(so, nam); 420 crit_exit(); 421 return (error); 422 } 423 424 int 425 soconnect(struct socket *so, struct sockaddr *nam, struct thread *td) 426 { 427 int error; 428 429 if (so->so_options & SO_ACCEPTCONN) 430 return (EOPNOTSUPP); 431 crit_enter(); 432 /* 433 * If protocol is connection-based, can only connect once. 434 * Otherwise, if connected, try to disconnect first. 435 * This allows user to disconnect by connecting to, e.g., 436 * a null address. 437 */ 438 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 439 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 440 (error = sodisconnect(so)))) { 441 error = EISCONN; 442 } else { 443 /* 444 * Prevent accumulated error from previous connection 445 * from biting us. 446 */ 447 so->so_error = 0; 448 error = so_pru_connect(so, nam, td); 449 } 450 crit_exit(); 451 return (error); 452 } 453 454 int 455 soconnect2(struct socket *so1, struct socket *so2) 456 { 457 int error; 458 459 crit_enter(); 460 error = so_pru_connect2(so1, so2); 461 crit_exit(); 462 return (error); 463 } 464 465 int 466 sodisconnect(struct socket *so) 467 { 468 int error; 469 470 crit_enter(); 471 if ((so->so_state & SS_ISCONNECTED) == 0) { 472 error = ENOTCONN; 473 goto bad; 474 } 475 if (so->so_state & SS_ISDISCONNECTING) { 476 error = EALREADY; 477 goto bad; 478 } 479 error = so_pru_disconnect(so); 480 bad: 481 crit_exit(); 482 return (error); 483 } 484 485 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 486 /* 487 * Send on a socket. 488 * If send must go all at once and message is larger than 489 * send buffering, then hard error. 490 * Lock against other senders. 491 * If must go all at once and not enough room now, then 492 * inform user that this would block and do nothing. 493 * Otherwise, if nonblocking, send as much as possible. 494 * The data to be sent is described by "uio" if nonzero, 495 * otherwise by the mbuf chain "top" (which must be null 496 * if uio is not). Data provided in mbuf chain must be small 497 * enough to send all at once. 498 * 499 * Returns nonzero on error, timeout or signal; callers 500 * must check for short counts if EINTR/ERESTART are returned. 501 * Data and control buffers are freed on return. 502 */ 503 int 504 sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, 505 struct mbuf *top, struct mbuf *control, int flags, 506 struct thread *td) 507 { 508 struct mbuf **mp; 509 struct mbuf *m; 510 long space, len, resid; 511 int clen = 0, error, dontroute, mlen; 512 int atomic = sosendallatonce(so) || top; 513 int pru_flags; 514 515 if (uio) 516 resid = uio->uio_resid; 517 else 518 resid = top->m_pkthdr.len; 519 /* 520 * In theory resid should be unsigned. 521 * However, space must be signed, as it might be less than 0 522 * if we over-committed, and we must use a signed comparison 523 * of space and resid. On the other hand, a negative resid 524 * causes us to loop sending 0-length segments to the protocol. 525 * 526 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM 527 * type sockets since that's an error. 528 */ 529 if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) { 530 error = EINVAL; 531 goto out; 532 } 533 534 dontroute = 535 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 536 (so->so_proto->pr_flags & PR_ATOMIC); 537 if (td->td_lwp != NULL) 538 td->td_lwp->lwp_ru.ru_msgsnd++; 539 if (control) 540 clen = control->m_len; 541 #define gotoerr(errcode) { error = errcode; crit_exit(); goto release; } 542 543 restart: 544 error = sblock(&so->so_snd, SBLOCKWAIT(flags)); 545 if (error) 546 goto out; 547 do { 548 crit_enter(); 549 if (so->so_state & SS_CANTSENDMORE) 550 gotoerr(EPIPE); 551 if (so->so_error) { 552 error = so->so_error; 553 so->so_error = 0; 554 crit_exit(); 555 goto release; 556 } 557 if ((so->so_state & SS_ISCONNECTED) == 0) { 558 /* 559 * `sendto' and `sendmsg' is allowed on a connection- 560 * based socket if it supports implied connect. 561 * Return ENOTCONN if not connected and no address is 562 * supplied. 563 */ 564 if ((so->so_proto->pr_flags & PR_CONNREQUIRED) && 565 (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) { 566 if ((so->so_state & SS_ISCONFIRMING) == 0 && 567 !(resid == 0 && clen != 0)) 568 gotoerr(ENOTCONN); 569 } else if (addr == 0) 570 gotoerr(so->so_proto->pr_flags & PR_CONNREQUIRED ? 571 ENOTCONN : EDESTADDRREQ); 572 } 573 space = sbspace(&so->so_snd); 574 if (flags & MSG_OOB) 575 space += 1024; 576 if ((atomic && resid > so->so_snd.sb_hiwat) || 577 clen > so->so_snd.sb_hiwat) 578 gotoerr(EMSGSIZE); 579 if (space < resid + clen && uio && 580 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 581 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) 582 gotoerr(EWOULDBLOCK); 583 sbunlock(&so->so_snd); 584 error = sbwait(&so->so_snd); 585 crit_exit(); 586 if (error) 587 goto out; 588 goto restart; 589 } 590 crit_exit(); 591 mp = ⊤ 592 space -= clen; 593 do { 594 if (uio == NULL) { 595 /* 596 * Data is prepackaged in "top". 597 */ 598 resid = 0; 599 if (flags & MSG_EOR) 600 top->m_flags |= M_EOR; 601 } else do { 602 m = m_getl(resid, MB_WAIT, MT_DATA, 603 top == NULL ? M_PKTHDR : 0, &mlen); 604 if (top == NULL) { 605 m->m_pkthdr.len = 0; 606 m->m_pkthdr.rcvif = (struct ifnet *)0; 607 } 608 len = min(min(mlen, resid), space); 609 if (resid < MINCLSIZE) { 610 /* 611 * For datagram protocols, leave room 612 * for protocol headers in first mbuf. 613 */ 614 if (atomic && top == 0 && len < mlen) 615 MH_ALIGN(m, len); 616 } 617 space -= len; 618 error = uiomove(mtod(m, caddr_t), (int)len, uio); 619 resid = uio->uio_resid; 620 m->m_len = len; 621 *mp = m; 622 top->m_pkthdr.len += len; 623 if (error) 624 goto release; 625 mp = &m->m_next; 626 if (resid <= 0) { 627 if (flags & MSG_EOR) 628 top->m_flags |= M_EOR; 629 break; 630 } 631 } while (space > 0 && atomic); 632 if (dontroute) 633 so->so_options |= SO_DONTROUTE; 634 if (flags & MSG_OOB) { 635 pru_flags = PRUS_OOB; 636 } else if ((flags & MSG_EOF) && 637 (so->so_proto->pr_flags & PR_IMPLOPCL) && 638 (resid <= 0)) { 639 /* 640 * If the user set MSG_EOF, the protocol 641 * understands this flag and nothing left to 642 * send then use PRU_SEND_EOF instead of PRU_SEND. 643 */ 644 pru_flags = PRUS_EOF; 645 } else if (resid > 0 && space > 0) { 646 /* If there is more to send, set PRUS_MORETOCOME */ 647 pru_flags = PRUS_MORETOCOME; 648 } else { 649 pru_flags = 0; 650 } 651 crit_enter(); 652 /* 653 * XXX all the SS_CANTSENDMORE checks previously 654 * done could be out of date. We could have recieved 655 * a reset packet in an interrupt or maybe we slept 656 * while doing page faults in uiomove() etc. We could 657 * probably recheck again inside the splnet() protection 658 * here, but there are probably other places that this 659 * also happens. We must rethink this. 660 */ 661 error = so_pru_send(so, pru_flags, top, addr, control, td); 662 crit_exit(); 663 if (dontroute) 664 so->so_options &= ~SO_DONTROUTE; 665 clen = 0; 666 control = 0; 667 top = 0; 668 mp = ⊤ 669 if (error) 670 goto release; 671 } while (resid && space > 0); 672 } while (resid); 673 674 release: 675 sbunlock(&so->so_snd); 676 out: 677 if (top) 678 m_freem(top); 679 if (control) 680 m_freem(control); 681 return (error); 682 } 683 684 /* 685 * A specialization of sosend() for UDP based on protocol-specific knowledge: 686 * so->so_proto->pr_flags has the PR_ATOMIC field set. This means that 687 * sosendallatonce() returns true, 688 * the "atomic" variable is true, 689 * and sosendudp() blocks until space is available for the entire send. 690 * so->so_proto->pr_flags does not have the PR_CONNREQUIRED or 691 * PR_IMPLOPCL flags set. 692 * UDP has no out-of-band data. 693 * UDP has no control data. 694 * UDP does not support MSG_EOR. 695 */ 696 int 697 sosendudp(struct socket *so, struct sockaddr *addr, struct uio *uio, 698 struct mbuf *top, struct mbuf *control, int flags, struct thread *td) 699 { 700 int resid, error; 701 boolean_t dontroute; /* temporary SO_DONTROUTE setting */ 702 703 if (td->td_lwp != NULL) 704 td->td_lwp->lwp_ru.ru_msgsnd++; 705 if (control) 706 m_freem(control); 707 708 KASSERT((uio && !top) || (top && !uio), ("bad arguments to sosendudp")); 709 resid = uio ? uio->uio_resid : top->m_pkthdr.len; 710 711 restart: 712 error = sblock(&so->so_snd, SBLOCKWAIT(flags)); 713 if (error) 714 goto out; 715 716 crit_enter(); 717 if (so->so_state & SS_CANTSENDMORE) 718 gotoerr(EPIPE); 719 if (so->so_error) { 720 error = so->so_error; 721 so->so_error = 0; 722 crit_exit(); 723 goto release; 724 } 725 if (!(so->so_state & SS_ISCONNECTED) && addr == NULL) 726 gotoerr(EDESTADDRREQ); 727 if (resid > so->so_snd.sb_hiwat) 728 gotoerr(EMSGSIZE); 729 if (uio && sbspace(&so->so_snd) < resid) { 730 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) 731 gotoerr(EWOULDBLOCK); 732 sbunlock(&so->so_snd); 733 error = sbwait(&so->so_snd); 734 crit_exit(); 735 if (error) 736 goto out; 737 goto restart; 738 } 739 crit_exit(); 740 741 if (uio) { 742 top = m_uiomove(uio); 743 if (top == NULL) 744 goto release; 745 } 746 747 dontroute = (flags & MSG_DONTROUTE) && !(so->so_options & SO_DONTROUTE); 748 if (dontroute) 749 so->so_options |= SO_DONTROUTE; 750 751 error = so_pru_send(so, 0, top, addr, NULL, td); 752 top = NULL; /* sent or freed in lower layer */ 753 754 if (dontroute) 755 so->so_options &= ~SO_DONTROUTE; 756 757 release: 758 sbunlock(&so->so_snd); 759 out: 760 if (top) 761 m_freem(top); 762 return (error); 763 } 764 765 /* 766 * Implement receive operations on a socket. 767 * We depend on the way that records are added to the sockbuf 768 * by sbappend*. In particular, each record (mbufs linked through m_next) 769 * must begin with an address if the protocol so specifies, 770 * followed by an optional mbuf or mbufs containing ancillary data, 771 * and then zero or more mbufs of data. 772 * In order to avoid blocking network interrupts for the entire time here, 773 * we exit the critical section while doing the actual copy to user space. 774 * Although the sockbuf is locked, new data may still be appended, 775 * and thus we must maintain consistency of the sockbuf during that time. 776 * 777 * The caller may receive the data as a single mbuf chain by supplying 778 * an mbuf **mp0 for use in returning the chain. The uio is then used 779 * only for the count in uio_resid. 780 */ 781 int 782 soreceive(struct socket *so, struct sockaddr **psa, struct uio *uio, 783 struct mbuf **mp0, struct mbuf **controlp, int *flagsp) 784 { 785 struct mbuf *m, *n, **mp; 786 struct mbuf *free_chain = NULL; 787 int flags, len, error, offset; 788 struct protosw *pr = so->so_proto; 789 int moff, type = 0; 790 int orig_resid = uio->uio_resid; 791 792 mp = mp0; 793 if (psa) 794 *psa = NULL; 795 if (controlp) 796 *controlp = NULL; 797 if (flagsp) 798 flags = *flagsp &~ MSG_EOR; 799 else 800 flags = 0; 801 if (flags & MSG_OOB) { 802 m = m_get(MB_WAIT, MT_DATA); 803 if (m == NULL) 804 return (ENOBUFS); 805 error = so_pru_rcvoob(so, m, flags & MSG_PEEK); 806 if (error) 807 goto bad; 808 do { 809 error = uiomove(mtod(m, caddr_t), 810 (int) min(uio->uio_resid, m->m_len), uio); 811 m = m_free(m); 812 } while (uio->uio_resid && error == 0 && m); 813 bad: 814 if (m) 815 m_freem(m); 816 return (error); 817 } 818 if (mp) 819 *mp = NULL; 820 if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) 821 so_pru_rcvd(so, 0); 822 823 restart: 824 crit_enter(); 825 error = sblock(&so->so_rcv, SBLOCKWAIT(flags)); 826 if (error) 827 goto done; 828 829 m = so->so_rcv.sb_mb; 830 /* 831 * If we have less data than requested, block awaiting more 832 * (subject to any timeout) if: 833 * 1. the current count is less than the low water mark, or 834 * 2. MSG_WAITALL is set, and it is possible to do the entire 835 * receive operation at once if we block (resid <= hiwat). 836 * 3. MSG_DONTWAIT is not set 837 * If MSG_WAITALL is set but resid is larger than the receive buffer, 838 * we have to do the receive in sections, and thus risk returning 839 * a short count if a timeout or signal occurs after we start. 840 */ 841 if (m == NULL || (((flags & MSG_DONTWAIT) == 0 && 842 so->so_rcv.sb_cc < uio->uio_resid) && 843 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 844 ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) && 845 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) { 846 KASSERT(m != NULL || !so->so_rcv.sb_cc, ("receive 1")); 847 if (so->so_error) { 848 if (m) 849 goto dontblock; 850 error = so->so_error; 851 if ((flags & MSG_PEEK) == 0) 852 so->so_error = 0; 853 goto release; 854 } 855 if (so->so_state & SS_CANTRCVMORE) { 856 if (m) 857 goto dontblock; 858 else 859 goto release; 860 } 861 for (; m; m = m->m_next) { 862 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 863 m = so->so_rcv.sb_mb; 864 goto dontblock; 865 } 866 } 867 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 868 (pr->pr_flags & PR_CONNREQUIRED)) { 869 error = ENOTCONN; 870 goto release; 871 } 872 if (uio->uio_resid == 0) 873 goto release; 874 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) { 875 error = EWOULDBLOCK; 876 goto release; 877 } 878 sbunlock(&so->so_rcv); 879 error = sbwait(&so->so_rcv); 880 if (error) 881 goto done; 882 crit_exit(); 883 goto restart; 884 } 885 dontblock: 886 if (uio->uio_td && uio->uio_td->td_proc) 887 uio->uio_td->td_lwp->lwp_ru.ru_msgrcv++; 888 889 /* 890 * note: m should be == sb_mb here. Cache the next record while 891 * cleaning up. Note that calling m_free*() will break out critical 892 * section. 893 */ 894 KKASSERT(m == so->so_rcv.sb_mb); 895 896 /* 897 * Skip any address mbufs prepending the record. 898 */ 899 if (pr->pr_flags & PR_ADDR) { 900 KASSERT(m->m_type == MT_SONAME, ("receive 1a")); 901 orig_resid = 0; 902 if (psa) 903 *psa = dup_sockaddr(mtod(m, struct sockaddr *)); 904 if (flags & MSG_PEEK) 905 m = m->m_next; 906 else 907 m = sbunlinkmbuf(&so->so_rcv, m, &free_chain); 908 } 909 910 /* 911 * Skip any control mbufs prepending the record. 912 */ 913 #ifdef SCTP 914 if (pr->pr_flags & PR_ADDR_OPT) { 915 /* 916 * For SCTP we may be getting a 917 * whole message OR a partial delivery. 918 */ 919 if (m && m->m_type == MT_SONAME) { 920 orig_resid = 0; 921 if (psa) 922 *psa = dup_sockaddr(mtod(m, struct sockaddr *)); 923 if (flags & MSG_PEEK) 924 m = m->m_next; 925 else 926 m = sbunlinkmbuf(&so->so_rcv, m, &free_chain); 927 } 928 } 929 #endif /* SCTP */ 930 while (m && m->m_type == MT_CONTROL && error == 0) { 931 if (flags & MSG_PEEK) { 932 if (controlp) 933 *controlp = m_copy(m, 0, m->m_len); 934 m = m->m_next; /* XXX race */ 935 } else { 936 if (controlp) { 937 n = sbunlinkmbuf(&so->so_rcv, m, NULL); 938 if (pr->pr_domain->dom_externalize && 939 mtod(m, struct cmsghdr *)->cmsg_type == 940 SCM_RIGHTS) 941 error = (*pr->pr_domain->dom_externalize)(m); 942 *controlp = m; 943 m = n; 944 } else { 945 m = sbunlinkmbuf(&so->so_rcv, m, &free_chain); 946 } 947 } 948 if (controlp && *controlp) { 949 orig_resid = 0; 950 controlp = &(*controlp)->m_next; 951 } 952 } 953 954 /* 955 * flag OOB data. 956 */ 957 if (m) { 958 type = m->m_type; 959 if (type == MT_OOBDATA) 960 flags |= MSG_OOB; 961 } 962 963 /* 964 * Copy to the UIO or mbuf return chain (*mp). 965 */ 966 moff = 0; 967 offset = 0; 968 while (m && uio->uio_resid > 0 && error == 0) { 969 if (m->m_type == MT_OOBDATA) { 970 if (type != MT_OOBDATA) 971 break; 972 } else if (type == MT_OOBDATA) 973 break; 974 else 975 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER, 976 ("receive 3")); 977 so->so_state &= ~SS_RCVATMARK; 978 len = uio->uio_resid; 979 if (so->so_oobmark && len > so->so_oobmark - offset) 980 len = so->so_oobmark - offset; 981 if (len > m->m_len - moff) 982 len = m->m_len - moff; 983 /* 984 * If mp is set, just pass back the mbufs. 985 * Otherwise copy them out via the uio, then free. 986 * Sockbuf must be consistent here (points to current mbuf, 987 * it points to next record) when we drop priority; 988 * we must note any additions to the sockbuf when we 989 * block interrupts again. 990 */ 991 if (mp == NULL) { 992 crit_exit(); 993 error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio); 994 crit_enter(); 995 if (error) 996 goto release; 997 } else { 998 uio->uio_resid -= len; 999 } 1000 1001 /* 1002 * Eat the entire mbuf or just a piece of it 1003 */ 1004 if (len == m->m_len - moff) { 1005 if (m->m_flags & M_EOR) 1006 flags |= MSG_EOR; 1007 #ifdef SCTP 1008 if (m->m_flags & M_NOTIFICATION) 1009 flags |= MSG_NOTIFICATION; 1010 #endif /* SCTP */ 1011 if (flags & MSG_PEEK) { 1012 m = m->m_next; 1013 moff = 0; 1014 } else { 1015 if (mp) { 1016 n = sbunlinkmbuf(&so->so_rcv, m, NULL); 1017 *mp = m; 1018 mp = &m->m_next; 1019 m = n; 1020 } else { 1021 m = sbunlinkmbuf(&so->so_rcv, m, &free_chain); 1022 } 1023 } 1024 } else { 1025 if (flags & MSG_PEEK) { 1026 moff += len; 1027 } else { 1028 if (mp) 1029 *mp = m_copym(m, 0, len, MB_WAIT); 1030 m->m_data += len; 1031 m->m_len -= len; 1032 so->so_rcv.sb_cc -= len; 1033 } 1034 } 1035 if (so->so_oobmark) { 1036 if ((flags & MSG_PEEK) == 0) { 1037 so->so_oobmark -= len; 1038 if (so->so_oobmark == 0) { 1039 so->so_state |= SS_RCVATMARK; 1040 break; 1041 } 1042 } else { 1043 offset += len; 1044 if (offset == so->so_oobmark) 1045 break; 1046 } 1047 } 1048 if (flags & MSG_EOR) 1049 break; 1050 /* 1051 * If the MSG_WAITALL flag is set (for non-atomic socket), 1052 * we must not quit until "uio->uio_resid == 0" or an error 1053 * termination. If a signal/timeout occurs, return 1054 * with a short count but without error. 1055 * Keep sockbuf locked against other readers. 1056 */ 1057 while (flags & MSG_WAITALL && m == NULL && 1058 uio->uio_resid > 0 && !sosendallatonce(so) && 1059 so->so_rcv.sb_mb == NULL) { 1060 if (so->so_error || so->so_state & SS_CANTRCVMORE) 1061 break; 1062 /* 1063 * The window might have closed to zero, make 1064 * sure we send an ack now that we've drained 1065 * the buffer or we might end up blocking until 1066 * the idle takes over (5 seconds). 1067 */ 1068 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1069 so_pru_rcvd(so, flags); 1070 error = sbwait(&so->so_rcv); 1071 if (error) { 1072 sbunlock(&so->so_rcv); 1073 error = 0; 1074 goto done; 1075 } 1076 m = so->so_rcv.sb_mb; 1077 } 1078 } 1079 1080 /* 1081 * If an atomic read was requested but unread data still remains 1082 * in the record, set MSG_TRUNC. 1083 */ 1084 if (m && pr->pr_flags & PR_ATOMIC) 1085 flags |= MSG_TRUNC; 1086 1087 /* 1088 * Cleanup. If an atomic read was requested drop any unread data. 1089 */ 1090 if ((flags & MSG_PEEK) == 0) { 1091 if (m && (pr->pr_flags & PR_ATOMIC)) 1092 sbdroprecord(&so->so_rcv); 1093 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) 1094 so_pru_rcvd(so, flags); 1095 } 1096 1097 if (orig_resid == uio->uio_resid && orig_resid && 1098 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 1099 sbunlock(&so->so_rcv); 1100 crit_exit(); 1101 goto restart; 1102 } 1103 1104 if (flagsp) 1105 *flagsp |= flags; 1106 release: 1107 sbunlock(&so->so_rcv); 1108 done: 1109 crit_exit(); 1110 if (free_chain) 1111 m_freem(free_chain); 1112 return (error); 1113 } 1114 1115 int 1116 soshutdown(struct socket *so, int how) 1117 { 1118 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 1119 return (EINVAL); 1120 1121 if (how != SHUT_WR) 1122 sorflush(so); 1123 if (how != SHUT_RD) 1124 return (so_pru_shutdown(so)); 1125 return (0); 1126 } 1127 1128 void 1129 sorflush(struct socket *so) 1130 { 1131 struct sockbuf *sb = &so->so_rcv; 1132 struct protosw *pr = so->so_proto; 1133 struct sockbuf asb; 1134 1135 sb->sb_flags |= SB_NOINTR; 1136 (void) sblock(sb, M_WAITOK); 1137 1138 crit_enter(); 1139 socantrcvmore(so); 1140 sbunlock(sb); 1141 asb = *sb; 1142 bzero((caddr_t)sb, sizeof (*sb)); 1143 if (asb.sb_flags & SB_KNOTE) { 1144 sb->sb_sel.si_note = asb.sb_sel.si_note; 1145 sb->sb_flags = SB_KNOTE; 1146 } 1147 crit_exit(); 1148 1149 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) 1150 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 1151 sbrelease(&asb, so); 1152 } 1153 1154 #ifdef INET 1155 static int 1156 do_setopt_accept_filter(struct socket *so, struct sockopt *sopt) 1157 { 1158 struct accept_filter_arg *afap = NULL; 1159 struct accept_filter *afp; 1160 struct so_accf *af = so->so_accf; 1161 int error = 0; 1162 1163 /* do not set/remove accept filters on non listen sockets */ 1164 if ((so->so_options & SO_ACCEPTCONN) == 0) { 1165 error = EINVAL; 1166 goto out; 1167 } 1168 1169 /* removing the filter */ 1170 if (sopt == NULL) { 1171 if (af != NULL) { 1172 if (af->so_accept_filter != NULL && 1173 af->so_accept_filter->accf_destroy != NULL) { 1174 af->so_accept_filter->accf_destroy(so); 1175 } 1176 if (af->so_accept_filter_str != NULL) { 1177 FREE(af->so_accept_filter_str, M_ACCF); 1178 } 1179 FREE(af, M_ACCF); 1180 so->so_accf = NULL; 1181 } 1182 so->so_options &= ~SO_ACCEPTFILTER; 1183 return (0); 1184 } 1185 /* adding a filter */ 1186 /* must remove previous filter first */ 1187 if (af != NULL) { 1188 error = EINVAL; 1189 goto out; 1190 } 1191 /* don't put large objects on the kernel stack */ 1192 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK); 1193 error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap); 1194 afap->af_name[sizeof(afap->af_name)-1] = '\0'; 1195 afap->af_arg[sizeof(afap->af_arg)-1] = '\0'; 1196 if (error) 1197 goto out; 1198 afp = accept_filt_get(afap->af_name); 1199 if (afp == NULL) { 1200 error = ENOENT; 1201 goto out; 1202 } 1203 MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK); 1204 bzero(af, sizeof(*af)); 1205 if (afp->accf_create != NULL) { 1206 if (afap->af_name[0] != '\0') { 1207 int len = strlen(afap->af_name) + 1; 1208 1209 MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK); 1210 strcpy(af->so_accept_filter_str, afap->af_name); 1211 } 1212 af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg); 1213 if (af->so_accept_filter_arg == NULL) { 1214 FREE(af->so_accept_filter_str, M_ACCF); 1215 FREE(af, M_ACCF); 1216 so->so_accf = NULL; 1217 error = EINVAL; 1218 goto out; 1219 } 1220 } 1221 af->so_accept_filter = afp; 1222 so->so_accf = af; 1223 so->so_options |= SO_ACCEPTFILTER; 1224 out: 1225 if (afap != NULL) 1226 FREE(afap, M_TEMP); 1227 return (error); 1228 } 1229 #endif /* INET */ 1230 1231 /* 1232 * Perhaps this routine, and sooptcopyout(), below, ought to come in 1233 * an additional variant to handle the case where the option value needs 1234 * to be some kind of integer, but not a specific size. 1235 * In addition to their use here, these functions are also called by the 1236 * protocol-level pr_ctloutput() routines. 1237 */ 1238 int 1239 sooptcopyin(struct sockopt *sopt, void *buf, size_t len, size_t minlen) 1240 { 1241 size_t valsize; 1242 1243 /* 1244 * If the user gives us more than we wanted, we ignore it, 1245 * but if we don't get the minimum length the caller 1246 * wants, we return EINVAL. On success, sopt->sopt_valsize 1247 * is set to however much we actually retrieved. 1248 */ 1249 if ((valsize = sopt->sopt_valsize) < minlen) 1250 return EINVAL; 1251 if (valsize > len) 1252 sopt->sopt_valsize = valsize = len; 1253 1254 if (sopt->sopt_td != NULL) 1255 return (copyin(sopt->sopt_val, buf, valsize)); 1256 1257 bcopy(sopt->sopt_val, buf, valsize); 1258 return 0; 1259 } 1260 1261 int 1262 sosetopt(struct socket *so, struct sockopt *sopt) 1263 { 1264 int error, optval; 1265 struct linger l; 1266 struct timeval tv; 1267 u_long val; 1268 1269 error = 0; 1270 sopt->sopt_dir = SOPT_SET; 1271 if (sopt->sopt_level != SOL_SOCKET) { 1272 if (so->so_proto && so->so_proto->pr_ctloutput) { 1273 return (so_pr_ctloutput(so, sopt)); 1274 } 1275 error = ENOPROTOOPT; 1276 } else { 1277 switch (sopt->sopt_name) { 1278 #ifdef INET 1279 case SO_ACCEPTFILTER: 1280 error = do_setopt_accept_filter(so, sopt); 1281 if (error) 1282 goto bad; 1283 break; 1284 #endif /* INET */ 1285 case SO_LINGER: 1286 error = sooptcopyin(sopt, &l, sizeof l, sizeof l); 1287 if (error) 1288 goto bad; 1289 1290 so->so_linger = l.l_linger; 1291 if (l.l_onoff) 1292 so->so_options |= SO_LINGER; 1293 else 1294 so->so_options &= ~SO_LINGER; 1295 break; 1296 1297 case SO_DEBUG: 1298 case SO_KEEPALIVE: 1299 case SO_DONTROUTE: 1300 case SO_USELOOPBACK: 1301 case SO_BROADCAST: 1302 case SO_REUSEADDR: 1303 case SO_REUSEPORT: 1304 case SO_OOBINLINE: 1305 case SO_TIMESTAMP: 1306 error = sooptcopyin(sopt, &optval, sizeof optval, 1307 sizeof optval); 1308 if (error) 1309 goto bad; 1310 if (optval) 1311 so->so_options |= sopt->sopt_name; 1312 else 1313 so->so_options &= ~sopt->sopt_name; 1314 break; 1315 1316 case SO_SNDBUF: 1317 case SO_RCVBUF: 1318 case SO_SNDLOWAT: 1319 case SO_RCVLOWAT: 1320 error = sooptcopyin(sopt, &optval, sizeof optval, 1321 sizeof optval); 1322 if (error) 1323 goto bad; 1324 1325 /* 1326 * Values < 1 make no sense for any of these 1327 * options, so disallow them. 1328 */ 1329 if (optval < 1) { 1330 error = EINVAL; 1331 goto bad; 1332 } 1333 1334 switch (sopt->sopt_name) { 1335 case SO_SNDBUF: 1336 case SO_RCVBUF: 1337 if (sbreserve(sopt->sopt_name == SO_SNDBUF ? 1338 &so->so_snd : &so->so_rcv, (u_long)optval, 1339 so, 1340 &curproc->p_rlimit[RLIMIT_SBSIZE]) == 0) { 1341 error = ENOBUFS; 1342 goto bad; 1343 } 1344 break; 1345 1346 /* 1347 * Make sure the low-water is never greater than 1348 * the high-water. 1349 */ 1350 case SO_SNDLOWAT: 1351 so->so_snd.sb_lowat = 1352 (optval > so->so_snd.sb_hiwat) ? 1353 so->so_snd.sb_hiwat : optval; 1354 break; 1355 case SO_RCVLOWAT: 1356 so->so_rcv.sb_lowat = 1357 (optval > so->so_rcv.sb_hiwat) ? 1358 so->so_rcv.sb_hiwat : optval; 1359 break; 1360 } 1361 break; 1362 1363 case SO_SNDTIMEO: 1364 case SO_RCVTIMEO: 1365 error = sooptcopyin(sopt, &tv, sizeof tv, 1366 sizeof tv); 1367 if (error) 1368 goto bad; 1369 1370 /* assert(hz > 0); */ 1371 if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz || 1372 tv.tv_usec < 0 || tv.tv_usec >= 1000000) { 1373 error = EDOM; 1374 goto bad; 1375 } 1376 /* assert(tick > 0); */ 1377 /* assert(ULONG_MAX - SHRT_MAX >= 1000000); */ 1378 val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick; 1379 if (val > SHRT_MAX) { 1380 error = EDOM; 1381 goto bad; 1382 } 1383 if (val == 0 && tv.tv_usec != 0) 1384 val = 1; 1385 1386 switch (sopt->sopt_name) { 1387 case SO_SNDTIMEO: 1388 so->so_snd.sb_timeo = val; 1389 break; 1390 case SO_RCVTIMEO: 1391 so->so_rcv.sb_timeo = val; 1392 break; 1393 } 1394 break; 1395 default: 1396 error = ENOPROTOOPT; 1397 break; 1398 } 1399 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) { 1400 (void) so_pr_ctloutput(so, sopt); 1401 } 1402 } 1403 bad: 1404 return (error); 1405 } 1406 1407 /* Helper routine for getsockopt */ 1408 int 1409 sooptcopyout(struct sockopt *sopt, const void *buf, size_t len) 1410 { 1411 int error; 1412 size_t valsize; 1413 1414 error = 0; 1415 1416 /* 1417 * Documented get behavior is that we always return a value, 1418 * possibly truncated to fit in the user's buffer. 1419 * Traditional behavior is that we always tell the user 1420 * precisely how much we copied, rather than something useful 1421 * like the total amount we had available for her. 1422 * Note that this interface is not idempotent; the entire answer must 1423 * generated ahead of time. 1424 */ 1425 valsize = min(len, sopt->sopt_valsize); 1426 sopt->sopt_valsize = valsize; 1427 if (sopt->sopt_val != 0) { 1428 if (sopt->sopt_td != NULL) 1429 error = copyout(buf, sopt->sopt_val, valsize); 1430 else 1431 bcopy(buf, sopt->sopt_val, valsize); 1432 } 1433 return error; 1434 } 1435 1436 int 1437 sogetopt(struct socket *so, struct sockopt *sopt) 1438 { 1439 int error, optval; 1440 struct linger l; 1441 struct timeval tv; 1442 #ifdef INET 1443 struct accept_filter_arg *afap; 1444 #endif 1445 1446 error = 0; 1447 sopt->sopt_dir = SOPT_GET; 1448 if (sopt->sopt_level != SOL_SOCKET) { 1449 if (so->so_proto && so->so_proto->pr_ctloutput) { 1450 return (so_pr_ctloutput(so, sopt)); 1451 } else 1452 return (ENOPROTOOPT); 1453 } else { 1454 switch (sopt->sopt_name) { 1455 #ifdef INET 1456 case SO_ACCEPTFILTER: 1457 if ((so->so_options & SO_ACCEPTCONN) == 0) 1458 return (EINVAL); 1459 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), 1460 M_TEMP, M_WAITOK); 1461 bzero(afap, sizeof(*afap)); 1462 if ((so->so_options & SO_ACCEPTFILTER) != 0) { 1463 strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name); 1464 if (so->so_accf->so_accept_filter_str != NULL) 1465 strcpy(afap->af_arg, so->so_accf->so_accept_filter_str); 1466 } 1467 error = sooptcopyout(sopt, afap, sizeof(*afap)); 1468 FREE(afap, M_TEMP); 1469 break; 1470 #endif /* INET */ 1471 1472 case SO_LINGER: 1473 l.l_onoff = so->so_options & SO_LINGER; 1474 l.l_linger = so->so_linger; 1475 error = sooptcopyout(sopt, &l, sizeof l); 1476 break; 1477 1478 case SO_USELOOPBACK: 1479 case SO_DONTROUTE: 1480 case SO_DEBUG: 1481 case SO_KEEPALIVE: 1482 case SO_REUSEADDR: 1483 case SO_REUSEPORT: 1484 case SO_BROADCAST: 1485 case SO_OOBINLINE: 1486 case SO_TIMESTAMP: 1487 optval = so->so_options & sopt->sopt_name; 1488 integer: 1489 error = sooptcopyout(sopt, &optval, sizeof optval); 1490 break; 1491 1492 case SO_TYPE: 1493 optval = so->so_type; 1494 goto integer; 1495 1496 case SO_ERROR: 1497 optval = so->so_error; 1498 so->so_error = 0; 1499 goto integer; 1500 1501 case SO_SNDBUF: 1502 optval = so->so_snd.sb_hiwat; 1503 goto integer; 1504 1505 case SO_RCVBUF: 1506 optval = so->so_rcv.sb_hiwat; 1507 goto integer; 1508 1509 case SO_SNDLOWAT: 1510 optval = so->so_snd.sb_lowat; 1511 goto integer; 1512 1513 case SO_RCVLOWAT: 1514 optval = so->so_rcv.sb_lowat; 1515 goto integer; 1516 1517 case SO_SNDTIMEO: 1518 case SO_RCVTIMEO: 1519 optval = (sopt->sopt_name == SO_SNDTIMEO ? 1520 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 1521 1522 tv.tv_sec = optval / hz; 1523 tv.tv_usec = (optval % hz) * tick; 1524 error = sooptcopyout(sopt, &tv, sizeof tv); 1525 break; 1526 1527 default: 1528 error = ENOPROTOOPT; 1529 break; 1530 } 1531 return (error); 1532 } 1533 } 1534 1535 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */ 1536 int 1537 soopt_getm(struct sockopt *sopt, struct mbuf **mp) 1538 { 1539 struct mbuf *m, *m_prev; 1540 int sopt_size = sopt->sopt_valsize, msize; 1541 1542 m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_DATA, 1543 0, &msize); 1544 if (m == NULL) 1545 return (ENOBUFS); 1546 m->m_len = min(msize, sopt_size); 1547 sopt_size -= m->m_len; 1548 *mp = m; 1549 m_prev = m; 1550 1551 while (sopt_size > 0) { 1552 m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, 1553 MT_DATA, 0, &msize); 1554 if (m == NULL) { 1555 m_freem(*mp); 1556 return (ENOBUFS); 1557 } 1558 m->m_len = min(msize, sopt_size); 1559 sopt_size -= m->m_len; 1560 m_prev->m_next = m; 1561 m_prev = m; 1562 } 1563 return (0); 1564 } 1565 1566 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */ 1567 int 1568 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m) 1569 { 1570 struct mbuf *m0 = m; 1571 1572 if (sopt->sopt_val == NULL) 1573 return 0; 1574 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1575 if (sopt->sopt_td != NULL) { 1576 int error; 1577 1578 error = copyin(sopt->sopt_val, mtod(m, char *), 1579 m->m_len); 1580 if (error != 0) { 1581 m_freem(m0); 1582 return (error); 1583 } 1584 } else 1585 bcopy(sopt->sopt_val, mtod(m, char *), m->m_len); 1586 sopt->sopt_valsize -= m->m_len; 1587 sopt->sopt_val = (caddr_t)sopt->sopt_val + m->m_len; 1588 m = m->m_next; 1589 } 1590 if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */ 1591 panic("ip6_sooptmcopyin"); 1592 return 0; 1593 } 1594 1595 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */ 1596 int 1597 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m) 1598 { 1599 struct mbuf *m0 = m; 1600 size_t valsize = 0; 1601 1602 if (sopt->sopt_val == NULL) 1603 return 0; 1604 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1605 if (sopt->sopt_td != NULL) { 1606 int error; 1607 1608 error = copyout(mtod(m, char *), sopt->sopt_val, 1609 m->m_len); 1610 if (error != 0) { 1611 m_freem(m0); 1612 return (error); 1613 } 1614 } else 1615 bcopy(mtod(m, char *), sopt->sopt_val, m->m_len); 1616 sopt->sopt_valsize -= m->m_len; 1617 sopt->sopt_val = (caddr_t)sopt->sopt_val + m->m_len; 1618 valsize += m->m_len; 1619 m = m->m_next; 1620 } 1621 if (m != NULL) { 1622 /* enough soopt buffer should be given from user-land */ 1623 m_freem(m0); 1624 return (EINVAL); 1625 } 1626 sopt->sopt_valsize = valsize; 1627 return 0; 1628 } 1629 1630 void 1631 sohasoutofband(struct socket *so) 1632 { 1633 if (so->so_sigio != NULL) 1634 pgsigio(so->so_sigio, SIGURG, 0); 1635 selwakeup(&so->so_rcv.sb_sel); 1636 } 1637 1638 int 1639 sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td) 1640 { 1641 int revents = 0; 1642 1643 crit_enter(); 1644 1645 if (events & (POLLIN | POLLRDNORM)) 1646 if (soreadable(so)) 1647 revents |= events & (POLLIN | POLLRDNORM); 1648 1649 if (events & POLLINIGNEOF) 1650 if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat || 1651 !TAILQ_EMPTY(&so->so_comp) || so->so_error) 1652 revents |= POLLINIGNEOF; 1653 1654 if (events & (POLLOUT | POLLWRNORM)) 1655 if (sowriteable(so)) 1656 revents |= events & (POLLOUT | POLLWRNORM); 1657 1658 if (events & (POLLPRI | POLLRDBAND)) 1659 if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) 1660 revents |= events & (POLLPRI | POLLRDBAND); 1661 1662 if (revents == 0) { 1663 if (events & 1664 (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM | 1665 POLLRDBAND)) { 1666 selrecord(td, &so->so_rcv.sb_sel); 1667 so->so_rcv.sb_flags |= SB_SEL; 1668 } 1669 1670 if (events & (POLLOUT | POLLWRNORM)) { 1671 selrecord(td, &so->so_snd.sb_sel); 1672 so->so_snd.sb_flags |= SB_SEL; 1673 } 1674 } 1675 1676 crit_exit(); 1677 return (revents); 1678 } 1679 1680 int 1681 sokqfilter(struct file *fp, struct knote *kn) 1682 { 1683 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1684 struct sockbuf *sb; 1685 1686 switch (kn->kn_filter) { 1687 case EVFILT_READ: 1688 if (so->so_options & SO_ACCEPTCONN) 1689 kn->kn_fop = &solisten_filtops; 1690 else 1691 kn->kn_fop = &soread_filtops; 1692 sb = &so->so_rcv; 1693 break; 1694 case EVFILT_WRITE: 1695 kn->kn_fop = &sowrite_filtops; 1696 sb = &so->so_snd; 1697 break; 1698 default: 1699 return (1); 1700 } 1701 1702 crit_enter(); 1703 SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext); 1704 sb->sb_flags |= SB_KNOTE; 1705 crit_exit(); 1706 return (0); 1707 } 1708 1709 static void 1710 filt_sordetach(struct knote *kn) 1711 { 1712 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1713 1714 crit_enter(); 1715 SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext); 1716 if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note)) 1717 so->so_rcv.sb_flags &= ~SB_KNOTE; 1718 crit_exit(); 1719 } 1720 1721 /*ARGSUSED*/ 1722 static int 1723 filt_soread(struct knote *kn, long hint) 1724 { 1725 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1726 1727 kn->kn_data = so->so_rcv.sb_cc; 1728 if (so->so_state & SS_CANTRCVMORE) { 1729 kn->kn_flags |= EV_EOF; 1730 kn->kn_fflags = so->so_error; 1731 return (1); 1732 } 1733 if (so->so_error) /* temporary udp error */ 1734 return (1); 1735 if (kn->kn_sfflags & NOTE_LOWAT) 1736 return (kn->kn_data >= kn->kn_sdata); 1737 return (kn->kn_data >= so->so_rcv.sb_lowat); 1738 } 1739 1740 static void 1741 filt_sowdetach(struct knote *kn) 1742 { 1743 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1744 1745 crit_enter(); 1746 SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext); 1747 if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note)) 1748 so->so_snd.sb_flags &= ~SB_KNOTE; 1749 crit_exit(); 1750 } 1751 1752 /*ARGSUSED*/ 1753 static int 1754 filt_sowrite(struct knote *kn, long hint) 1755 { 1756 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1757 1758 kn->kn_data = sbspace(&so->so_snd); 1759 if (so->so_state & SS_CANTSENDMORE) { 1760 kn->kn_flags |= EV_EOF; 1761 kn->kn_fflags = so->so_error; 1762 return (1); 1763 } 1764 if (so->so_error) /* temporary udp error */ 1765 return (1); 1766 if (((so->so_state & SS_ISCONNECTED) == 0) && 1767 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 1768 return (0); 1769 if (kn->kn_sfflags & NOTE_LOWAT) 1770 return (kn->kn_data >= kn->kn_sdata); 1771 return (kn->kn_data >= so->so_snd.sb_lowat); 1772 } 1773 1774 /*ARGSUSED*/ 1775 static int 1776 filt_solisten(struct knote *kn, long hint) 1777 { 1778 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1779 1780 kn->kn_data = so->so_qlen; 1781 return (! TAILQ_EMPTY(&so->so_comp)); 1782 } 1783