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 1.36 2005/07/23 07:28:34 dillon 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 so_gen_t so_gencnt; /* generation count for sockets */ 137 138 MALLOC_DEFINE(M_SONAME, "soname", "socket name"); 139 MALLOC_DEFINE(M_PCB, "pcb", "protocol control block"); 140 141 142 static int somaxconn = SOMAXCONN; 143 SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW, 144 &somaxconn, 0, "Maximum pending socket connection queue size"); 145 146 /* 147 * Socket operation routines. 148 * These routines are called by the routines in 149 * sys_socket.c or from a system process, and 150 * implement the semantics of socket operations by 151 * switching out to the protocol specific routines. 152 */ 153 154 /* 155 * Get a socket structure from our zone, and initialize it. 156 * We don't implement `waitok' yet (see comments in uipc_domain.c). 157 * Note that it would probably be better to allocate socket 158 * and PCB at the same time, but I'm not convinced that all 159 * the protocols can be easily modified to do this. 160 */ 161 struct socket * 162 soalloc(waitok) 163 int waitok; 164 { 165 struct socket *so; 166 167 so = zalloc(socket_zone); 168 if (so) { 169 /* XXX race condition for reentrant kernel */ 170 bzero(so, sizeof *so); 171 so->so_gencnt = ++so_gencnt; 172 TAILQ_INIT(&so->so_aiojobq); 173 TAILQ_INIT(&so->so_rcv.sb_sel.si_mlist); 174 TAILQ_INIT(&so->so_snd.sb_sel.si_mlist); 175 } 176 return so; 177 } 178 179 int 180 socreate(int dom, struct socket **aso, int type, 181 int proto, struct thread *td) 182 { 183 struct proc *p = td->td_proc; 184 struct protosw *prp; 185 struct socket *so; 186 struct pru_attach_info ai; 187 int error; 188 189 if (proto) 190 prp = pffindproto(dom, proto, type); 191 else 192 prp = pffindtype(dom, type); 193 194 if (prp == 0 || prp->pr_usrreqs->pru_attach == 0) 195 return (EPROTONOSUPPORT); 196 197 if (p->p_ucred->cr_prison && jail_socket_unixiproute_only && 198 prp->pr_domain->dom_family != PF_LOCAL && 199 prp->pr_domain->dom_family != PF_INET && 200 prp->pr_domain->dom_family != PF_ROUTE) { 201 return (EPROTONOSUPPORT); 202 } 203 204 if (prp->pr_type != type) 205 return (EPROTOTYPE); 206 so = soalloc(p != 0); 207 if (so == 0) 208 return (ENOBUFS); 209 210 TAILQ_INIT(&so->so_incomp); 211 TAILQ_INIT(&so->so_comp); 212 so->so_type = type; 213 so->so_cred = crhold(p->p_ucred); 214 so->so_proto = prp; 215 ai.sb_rlimit = &p->p_rlimit[RLIMIT_SBSIZE]; 216 ai.p_ucred = p->p_ucred; 217 ai.fd_rdir = p->p_fd->fd_rdir; 218 error = so_pru_attach(so, proto, &ai); 219 if (error) { 220 so->so_state |= SS_NOFDREF; 221 sofree(so); 222 return (error); 223 } 224 *aso = so; 225 return (0); 226 } 227 228 int 229 sobind(struct socket *so, struct sockaddr *nam, struct thread *td) 230 { 231 int error; 232 233 crit_enter(); 234 error = so_pru_bind(so, nam, td); 235 crit_exit(); 236 return (error); 237 } 238 239 void 240 sodealloc(struct socket *so) 241 { 242 243 so->so_gencnt = ++so_gencnt; 244 if (so->so_rcv.sb_hiwat) 245 (void)chgsbsize(so->so_cred->cr_uidinfo, 246 &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY); 247 if (so->so_snd.sb_hiwat) 248 (void)chgsbsize(so->so_cred->cr_uidinfo, 249 &so->so_snd.sb_hiwat, 0, RLIM_INFINITY); 250 #ifdef INET 251 /* remove accept filter if present */ 252 if (so->so_accf != NULL) 253 do_setopt_accept_filter(so, NULL); 254 #endif /* INET */ 255 crfree(so->so_cred); 256 zfree(socket_zone, so); 257 } 258 259 int 260 solisten(struct socket *so, int backlog, struct thread *td) 261 { 262 int error; 263 #ifdef SCTP 264 short oldopt, oldqlimit; 265 #endif /* SCTP */ 266 267 crit_enter(); 268 if (so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING)) { 269 crit_exit(); 270 return (EINVAL); 271 } 272 273 #ifdef SCTP 274 oldopt = so->so_options; 275 oldqlimit = so->so_qlimit; 276 #endif /* SCTP */ 277 278 if (TAILQ_EMPTY(&so->so_comp)) 279 so->so_options |= SO_ACCEPTCONN; 280 if (backlog < 0 || backlog > somaxconn) 281 backlog = somaxconn; 282 so->so_qlimit = backlog; 283 /* SCTP needs to look at tweak both the inbound backlog parameter AND 284 * the so_options (UDP model both connect's and gets inbound 285 * connections .. implicitly). 286 */ 287 error = so_pru_listen(so, td); 288 if (error) { 289 #ifdef SCTP 290 /* Restore the params */ 291 so->so_options = oldopt; 292 so->so_qlimit = oldqlimit; 293 #endif /* SCTP */ 294 crit_exit(); 295 return (error); 296 } 297 crit_exit(); 298 return (0); 299 } 300 301 void 302 sofree(struct socket *so) 303 { 304 struct socket *head = so->so_head; 305 306 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) 307 return; 308 if (head != NULL) { 309 if (so->so_state & SS_INCOMP) { 310 TAILQ_REMOVE(&head->so_incomp, so, so_list); 311 head->so_incqlen--; 312 } else if (so->so_state & SS_COMP) { 313 /* 314 * We must not decommission a socket that's 315 * on the accept(2) queue. If we do, then 316 * accept(2) may hang after select(2) indicated 317 * that the listening socket was ready. 318 */ 319 return; 320 } else { 321 panic("sofree: not queued"); 322 } 323 so->so_state &= ~SS_INCOMP; 324 so->so_head = NULL; 325 } 326 sbrelease(&so->so_snd, so); 327 sorflush(so); 328 sodealloc(so); 329 } 330 331 /* 332 * Close a socket on last file table reference removal. 333 * Initiate disconnect if connected. 334 * Free socket when disconnect complete. 335 */ 336 int 337 soclose(struct socket *so) 338 { 339 int error = 0; 340 341 crit_enter(); 342 funsetown(so->so_sigio); 343 if (so->so_pcb == NULL) 344 goto discard; 345 if (so->so_state & SS_ISCONNECTED) { 346 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 347 error = sodisconnect(so); 348 if (error) 349 goto drop; 350 } 351 if (so->so_options & SO_LINGER) { 352 if ((so->so_state & SS_ISDISCONNECTING) && 353 (so->so_state & SS_NBIO)) 354 goto drop; 355 while (so->so_state & SS_ISCONNECTED) { 356 error = tsleep((caddr_t)&so->so_timeo, 357 PCATCH, "soclos", so->so_linger * hz); 358 if (error) 359 break; 360 } 361 } 362 } 363 drop: 364 if (so->so_pcb) { 365 int error2; 366 367 error2 = so_pru_detach(so); 368 if (error == 0) 369 error = error2; 370 } 371 discard: 372 if (so->so_options & SO_ACCEPTCONN) { 373 struct socket *sp, *sonext; 374 375 sp = TAILQ_FIRST(&so->so_incomp); 376 for (; sp != NULL; sp = sonext) { 377 sonext = TAILQ_NEXT(sp, so_list); 378 (void) soabort(sp); 379 } 380 for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) { 381 sonext = TAILQ_NEXT(sp, so_list); 382 /* Dequeue from so_comp since sofree() won't do it */ 383 TAILQ_REMOVE(&so->so_comp, sp, so_list); 384 so->so_qlen--; 385 sp->so_state &= ~SS_COMP; 386 sp->so_head = NULL; 387 (void) soabort(sp); 388 } 389 } 390 if (so->so_state & SS_NOFDREF) 391 panic("soclose: NOFDREF"); 392 so->so_state |= SS_NOFDREF; 393 sofree(so); 394 crit_exit(); 395 return (error); 396 } 397 398 /* 399 * Must be called from a critical section. 400 */ 401 int 402 soabort(so) 403 struct socket *so; 404 { 405 int error; 406 407 error = so_pru_abort(so); 408 if (error) { 409 sofree(so); 410 return error; 411 } 412 return (0); 413 } 414 415 int 416 soaccept(struct socket *so, struct sockaddr **nam) 417 { 418 int error; 419 420 crit_enter(); 421 if ((so->so_state & SS_NOFDREF) == 0) 422 panic("soaccept: !NOFDREF"); 423 so->so_state &= ~SS_NOFDREF; 424 error = so_pru_accept(so, nam); 425 crit_exit(); 426 return (error); 427 } 428 429 int 430 soconnect(struct socket *so, struct sockaddr *nam, struct thread *td) 431 { 432 int error; 433 434 if (so->so_options & SO_ACCEPTCONN) 435 return (EOPNOTSUPP); 436 crit_enter(); 437 /* 438 * If protocol is connection-based, can only connect once. 439 * Otherwise, if connected, try to disconnect first. 440 * This allows user to disconnect by connecting to, e.g., 441 * a null address. 442 */ 443 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 444 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 445 (error = sodisconnect(so)))) { 446 error = EISCONN; 447 } else { 448 /* 449 * Prevent accumulated error from previous connection 450 * from biting us. 451 */ 452 so->so_error = 0; 453 error = so_pru_connect(so, nam, td); 454 } 455 crit_exit(); 456 return (error); 457 } 458 459 int 460 soconnect2(struct socket *so1, struct socket *so2) 461 { 462 int error; 463 464 crit_enter(); 465 error = so_pru_connect2(so1, so2); 466 crit_exit(); 467 return (error); 468 } 469 470 int 471 sodisconnect(struct socket *so) 472 { 473 int error; 474 475 crit_enter(); 476 if ((so->so_state & SS_ISCONNECTED) == 0) { 477 error = ENOTCONN; 478 goto bad; 479 } 480 if (so->so_state & SS_ISDISCONNECTING) { 481 error = EALREADY; 482 goto bad; 483 } 484 error = so_pru_disconnect(so); 485 bad: 486 crit_exit(); 487 return (error); 488 } 489 490 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 491 /* 492 * Send on a socket. 493 * If send must go all at once and message is larger than 494 * send buffering, then hard error. 495 * Lock against other senders. 496 * If must go all at once and not enough room now, then 497 * inform user that this would block and do nothing. 498 * Otherwise, if nonblocking, send as much as possible. 499 * The data to be sent is described by "uio" if nonzero, 500 * otherwise by the mbuf chain "top" (which must be null 501 * if uio is not). Data provided in mbuf chain must be small 502 * enough to send all at once. 503 * 504 * Returns nonzero on error, timeout or signal; callers 505 * must check for short counts if EINTR/ERESTART are returned. 506 * Data and control buffers are freed on return. 507 */ 508 int 509 sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, 510 struct mbuf *top, struct mbuf *control, int flags, 511 struct thread *td) 512 { 513 struct mbuf **mp; 514 struct mbuf *m; 515 long space, len, resid; 516 int clen = 0, error, dontroute, mlen; 517 int atomic = sosendallatonce(so) || top; 518 int pru_flags; 519 520 if (uio) 521 resid = uio->uio_resid; 522 else 523 resid = top->m_pkthdr.len; 524 /* 525 * In theory resid should be unsigned. 526 * However, space must be signed, as it might be less than 0 527 * if we over-committed, and we must use a signed comparison 528 * of space and resid. On the other hand, a negative resid 529 * causes us to loop sending 0-length segments to the protocol. 530 * 531 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM 532 * type sockets since that's an error. 533 */ 534 if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) { 535 error = EINVAL; 536 goto out; 537 } 538 539 dontroute = 540 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 541 (so->so_proto->pr_flags & PR_ATOMIC); 542 if (td->td_proc && td->td_proc->p_stats) 543 td->td_proc->p_stats->p_ru.ru_msgsnd++; 544 if (control) 545 clen = control->m_len; 546 #define gotoerr(errno) { error = errno; crit_exit(); goto release; } 547 548 restart: 549 error = sblock(&so->so_snd, SBLOCKWAIT(flags)); 550 if (error) 551 goto out; 552 do { 553 crit_enter(); 554 if (so->so_state & SS_CANTSENDMORE) 555 gotoerr(EPIPE); 556 if (so->so_error) { 557 error = so->so_error; 558 so->so_error = 0; 559 crit_exit(); 560 goto release; 561 } 562 if ((so->so_state & SS_ISCONNECTED) == 0) { 563 /* 564 * `sendto' and `sendmsg' is allowed on a connection- 565 * based socket if it supports implied connect. 566 * Return ENOTCONN if not connected and no address is 567 * supplied. 568 */ 569 if ((so->so_proto->pr_flags & PR_CONNREQUIRED) && 570 (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) { 571 if ((so->so_state & SS_ISCONFIRMING) == 0 && 572 !(resid == 0 && clen != 0)) 573 gotoerr(ENOTCONN); 574 } else if (addr == 0) 575 gotoerr(so->so_proto->pr_flags & PR_CONNREQUIRED ? 576 ENOTCONN : EDESTADDRREQ); 577 } 578 space = sbspace(&so->so_snd); 579 if (flags & MSG_OOB) 580 space += 1024; 581 if ((atomic && resid > so->so_snd.sb_hiwat) || 582 clen > so->so_snd.sb_hiwat) 583 gotoerr(EMSGSIZE); 584 if (space < resid + clen && uio && 585 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 586 if (so->so_state & SS_NBIO) 587 gotoerr(EWOULDBLOCK); 588 sbunlock(&so->so_snd); 589 error = sbwait(&so->so_snd); 590 crit_exit(); 591 if (error) 592 goto out; 593 goto restart; 594 } 595 crit_exit(); 596 mp = ⊤ 597 space -= clen; 598 do { 599 if (uio == NULL) { 600 /* 601 * Data is prepackaged in "top". 602 */ 603 resid = 0; 604 if (flags & MSG_EOR) 605 top->m_flags |= M_EOR; 606 } else do { 607 m = m_getl(resid, MB_WAIT, MT_DATA, 608 top == NULL ? M_PKTHDR : 0, &mlen); 609 if (top == NULL) { 610 m->m_pkthdr.len = 0; 611 m->m_pkthdr.rcvif = (struct ifnet *)0; 612 } 613 len = min(min(mlen, resid), space); 614 if (resid < MINCLSIZE) { 615 /* 616 * For datagram protocols, leave room 617 * for protocol headers in first mbuf. 618 */ 619 if (atomic && top == 0 && len < mlen) 620 MH_ALIGN(m, len); 621 } 622 space -= len; 623 error = uiomove(mtod(m, caddr_t), (int)len, uio); 624 resid = uio->uio_resid; 625 m->m_len = len; 626 *mp = m; 627 top->m_pkthdr.len += len; 628 if (error) 629 goto release; 630 mp = &m->m_next; 631 if (resid <= 0) { 632 if (flags & MSG_EOR) 633 top->m_flags |= M_EOR; 634 break; 635 } 636 } while (space > 0 && atomic); 637 if (dontroute) 638 so->so_options |= SO_DONTROUTE; 639 if (flags & MSG_OOB) { 640 pru_flags = PRUS_OOB; 641 } else if ((flags & MSG_EOF) && 642 (so->so_proto->pr_flags & PR_IMPLOPCL) && 643 (resid <= 0)) { 644 /* 645 * If the user set MSG_EOF, the protocol 646 * understands this flag and nothing left to 647 * send then use PRU_SEND_EOF instead of PRU_SEND. 648 */ 649 pru_flags = PRUS_EOF; 650 } else if (resid > 0 && space > 0) { 651 /* If there is more to send, set PRUS_MORETOCOME */ 652 pru_flags = PRUS_MORETOCOME; 653 } else { 654 pru_flags = 0; 655 } 656 crit_enter(); 657 /* 658 * XXX all the SS_CANTSENDMORE checks previously 659 * done could be out of date. We could have recieved 660 * a reset packet in an interrupt or maybe we slept 661 * while doing page faults in uiomove() etc. We could 662 * probably recheck again inside the splnet() protection 663 * here, but there are probably other places that this 664 * also happens. We must rethink this. 665 */ 666 error = so_pru_send(so, pru_flags, top, addr, control, td); 667 crit_exit(); 668 if (dontroute) 669 so->so_options &= ~SO_DONTROUTE; 670 clen = 0; 671 control = 0; 672 top = 0; 673 mp = ⊤ 674 if (error) 675 goto release; 676 } while (resid && space > 0); 677 } while (resid); 678 679 release: 680 sbunlock(&so->so_snd); 681 out: 682 if (top) 683 m_freem(top); 684 if (control) 685 m_freem(control); 686 return (error); 687 } 688 689 /* 690 * A specialization of sosend() for UDP based on protocol-specific knowledge: 691 * so->so_proto->pr_flags has the PR_ATOMIC field set. This means that 692 * sosendallatonce() returns true, 693 * the "atomic" variable is true, 694 * and sosendudp() blocks until space is available for the entire send. 695 * so->so_proto->pr_flags does not have the PR_CONNREQUIRED or 696 * PR_IMPLOPCL flags set. 697 * UDP has no out-of-band data. 698 * UDP has no control data. 699 * UDP does not support MSG_EOR. 700 */ 701 int 702 sosendudp(struct socket *so, struct sockaddr *addr, struct uio *uio, 703 struct mbuf *top, struct mbuf *control, int flags, struct thread *td) 704 { 705 int resid, error; 706 boolean_t dontroute; /* temporary SO_DONTROUTE setting */ 707 708 if (td->td_proc && td->td_proc->p_stats) 709 td->td_proc->p_stats->p_ru.ru_msgsnd++; 710 if (control) 711 m_freem(control); 712 713 KASSERT((uio && !top) || (top && !uio), ("bad arguments to sosendudp")); 714 resid = uio ? uio->uio_resid : top->m_pkthdr.len; 715 716 restart: 717 error = sblock(&so->so_snd, SBLOCKWAIT(flags)); 718 if (error) 719 goto out; 720 721 crit_enter(); 722 if (so->so_state & SS_CANTSENDMORE) 723 gotoerr(EPIPE); 724 if (so->so_error) { 725 error = so->so_error; 726 so->so_error = 0; 727 crit_exit(); 728 goto release; 729 } 730 if (!(so->so_state & SS_ISCONNECTED) && addr == NULL) 731 gotoerr(EDESTADDRREQ); 732 if (resid > so->so_snd.sb_hiwat) 733 gotoerr(EMSGSIZE); 734 if (uio && sbspace(&so->so_snd) < resid) { 735 if (so->so_state & SS_NBIO) 736 gotoerr(EWOULDBLOCK); 737 sbunlock(&so->so_snd); 738 error = sbwait(&so->so_snd); 739 crit_exit(); 740 if (error) 741 goto out; 742 goto restart; 743 } 744 crit_exit(); 745 746 if (uio) { 747 top = m_uiomove(uio); 748 if (top == NULL) 749 goto release; 750 } 751 752 dontroute = (flags & MSG_DONTROUTE) && !(so->so_options & SO_DONTROUTE); 753 if (dontroute) 754 so->so_options |= SO_DONTROUTE; 755 756 error = so_pru_send(so, 0, top, addr, NULL, td); 757 top = NULL; /* sent or freed in lower layer */ 758 759 if (dontroute) 760 so->so_options &= ~SO_DONTROUTE; 761 762 release: 763 sbunlock(&so->so_snd); 764 out: 765 if (top) 766 m_freem(top); 767 return (error); 768 } 769 770 /* 771 * Implement receive operations on a socket. 772 * We depend on the way that records are added to the sockbuf 773 * by sbappend*. In particular, each record (mbufs linked through m_next) 774 * must begin with an address if the protocol so specifies, 775 * followed by an optional mbuf or mbufs containing ancillary data, 776 * and then zero or more mbufs of data. 777 * In order to avoid blocking network interrupts for the entire time here, 778 * we exit the critical section while doing the actual copy to user space. 779 * Although the sockbuf is locked, new data may still be appended, 780 * and thus we must maintain consistency of the sockbuf during that time. 781 * 782 * The caller may receive the data as a single mbuf chain by supplying 783 * an mbuf **mp0 for use in returning the chain. The uio is then used 784 * only for the count in uio_resid. 785 */ 786 int 787 soreceive(so, psa, uio, mp0, controlp, flagsp) 788 struct socket *so; 789 struct sockaddr **psa; 790 struct uio *uio; 791 struct mbuf **mp0; 792 struct mbuf **controlp; 793 int *flagsp; 794 { 795 struct mbuf *m, *n, **mp; 796 struct mbuf *free_chain = NULL; 797 int flags, len, error, offset; 798 struct protosw *pr = so->so_proto; 799 int moff, type = 0; 800 int orig_resid = uio->uio_resid; 801 802 mp = mp0; 803 if (psa) 804 *psa = NULL; 805 if (controlp) 806 *controlp = NULL; 807 if (flagsp) 808 flags = *flagsp &~ MSG_EOR; 809 else 810 flags = 0; 811 if (flags & MSG_OOB) { 812 m = m_get(MB_WAIT, MT_DATA); 813 if (m == NULL) 814 return (ENOBUFS); 815 error = so_pru_rcvoob(so, m, flags & MSG_PEEK); 816 if (error) 817 goto bad; 818 do { 819 error = uiomove(mtod(m, caddr_t), 820 (int) min(uio->uio_resid, m->m_len), uio); 821 m = m_free(m); 822 } while (uio->uio_resid && error == 0 && m); 823 bad: 824 if (m) 825 m_freem(m); 826 return (error); 827 } 828 if (mp) 829 *mp = NULL; 830 if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) 831 so_pru_rcvd(so, 0); 832 833 restart: 834 crit_enter(); 835 error = sblock(&so->so_rcv, SBLOCKWAIT(flags)); 836 if (error) 837 goto done; 838 839 m = so->so_rcv.sb_mb; 840 /* 841 * If we have less data than requested, block awaiting more 842 * (subject to any timeout) if: 843 * 1. the current count is less than the low water mark, or 844 * 2. MSG_WAITALL is set, and it is possible to do the entire 845 * receive operation at once if we block (resid <= hiwat). 846 * 3. MSG_DONTWAIT is not set 847 * If MSG_WAITALL is set but resid is larger than the receive buffer, 848 * we have to do the receive in sections, and thus risk returning 849 * a short count if a timeout or signal occurs after we start. 850 */ 851 if (m == NULL || (((flags & MSG_DONTWAIT) == 0 && 852 so->so_rcv.sb_cc < uio->uio_resid) && 853 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 854 ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) && 855 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) { 856 KASSERT(m != NULL || !so->so_rcv.sb_cc, ("receive 1")); 857 if (so->so_error) { 858 if (m) 859 goto dontblock; 860 error = so->so_error; 861 if ((flags & MSG_PEEK) == 0) 862 so->so_error = 0; 863 goto release; 864 } 865 if (so->so_state & SS_CANTRCVMORE) { 866 if (m) 867 goto dontblock; 868 else 869 goto release; 870 } 871 for (; m; m = m->m_next) { 872 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 873 m = so->so_rcv.sb_mb; 874 goto dontblock; 875 } 876 } 877 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 878 (pr->pr_flags & PR_CONNREQUIRED)) { 879 error = ENOTCONN; 880 goto release; 881 } 882 if (uio->uio_resid == 0) 883 goto release; 884 if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) { 885 error = EWOULDBLOCK; 886 goto release; 887 } 888 sbunlock(&so->so_rcv); 889 error = sbwait(&so->so_rcv); 890 if (error) 891 goto done; 892 crit_exit(); 893 goto restart; 894 } 895 dontblock: 896 if (uio->uio_td && uio->uio_td->td_proc) 897 uio->uio_td->td_proc->p_stats->p_ru.ru_msgrcv++; 898 899 /* 900 * note: m should be == sb_mb here. Cache the next record while 901 * cleaning up. Note that calling m_free*() will break out critical 902 * section. 903 */ 904 KKASSERT(m == so->so_rcv.sb_mb); 905 906 /* 907 * Skip any address mbufs prepending the record. 908 */ 909 if (pr->pr_flags & PR_ADDR) { 910 KASSERT(m->m_type == MT_SONAME, ("receive 1a")); 911 orig_resid = 0; 912 if (psa) 913 *psa = dup_sockaddr(mtod(m, struct sockaddr *)); 914 if (flags & MSG_PEEK) 915 m = m->m_next; 916 else 917 m = sbunlinkmbuf(&so->so_rcv, m, &free_chain); 918 } 919 920 /* 921 * Skip any control mbufs prepending the record. 922 */ 923 #ifdef SCTP 924 if (pr->pr_flags & PR_ADDR_OPT) { 925 /* 926 * For SCTP we may be getting a 927 * whole message OR a partial delivery. 928 */ 929 if (m && m->m_type == MT_SONAME) { 930 orig_resid = 0; 931 if (psa) 932 *psa = dup_sockaddr(mtod(m, struct sockaddr *)); 933 if (flags & MSG_PEEK) 934 m = m->m_next; 935 else 936 m = sbunlinkmbuf(&so->so_rcv, m, &free_chain); 937 } 938 } 939 #endif /* SCTP */ 940 while (m && m->m_type == MT_CONTROL && error == 0) { 941 if (flags & MSG_PEEK) { 942 if (controlp) 943 *controlp = m_copy(m, 0, m->m_len); 944 m = m->m_next; /* XXX race */ 945 } else { 946 if (controlp) { 947 n = sbunlinkmbuf(&so->so_rcv, m, NULL); 948 if (pr->pr_domain->dom_externalize && 949 mtod(m, struct cmsghdr *)->cmsg_type == 950 SCM_RIGHTS) 951 error = (*pr->pr_domain->dom_externalize)(m); 952 *controlp = m; 953 m = n; 954 } else { 955 m = sbunlinkmbuf(&so->so_rcv, m, &free_chain); 956 } 957 } 958 if (controlp && *controlp) { 959 orig_resid = 0; 960 controlp = &(*controlp)->m_next; 961 } 962 } 963 964 /* 965 * flag OOB data. 966 */ 967 if (m) { 968 type = m->m_type; 969 if (type == MT_OOBDATA) 970 flags |= MSG_OOB; 971 } 972 973 /* 974 * Copy to the UIO or mbuf return chain (*mp). 975 */ 976 moff = 0; 977 offset = 0; 978 while (m && uio->uio_resid > 0 && error == 0) { 979 if (m->m_type == MT_OOBDATA) { 980 if (type != MT_OOBDATA) 981 break; 982 } else if (type == MT_OOBDATA) 983 break; 984 else 985 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER, 986 ("receive 3")); 987 so->so_state &= ~SS_RCVATMARK; 988 len = uio->uio_resid; 989 if (so->so_oobmark && len > so->so_oobmark - offset) 990 len = so->so_oobmark - offset; 991 if (len > m->m_len - moff) 992 len = m->m_len - moff; 993 /* 994 * If mp is set, just pass back the mbufs. 995 * Otherwise copy them out via the uio, then free. 996 * Sockbuf must be consistent here (points to current mbuf, 997 * it points to next record) when we drop priority; 998 * we must note any additions to the sockbuf when we 999 * block interrupts again. 1000 */ 1001 if (mp == NULL) { 1002 crit_exit(); 1003 error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio); 1004 crit_enter(); 1005 if (error) 1006 goto release; 1007 } else { 1008 uio->uio_resid -= len; 1009 } 1010 1011 /* 1012 * Eat the entire mbuf or just a piece of it 1013 */ 1014 if (len == m->m_len - moff) { 1015 if (m->m_flags & M_EOR) 1016 flags |= MSG_EOR; 1017 #ifdef SCTP 1018 if (m->m_flags & M_NOTIFICATION) 1019 flags |= MSG_NOTIFICATION; 1020 #endif /* SCTP */ 1021 if (flags & MSG_PEEK) { 1022 m = m->m_next; 1023 moff = 0; 1024 } else { 1025 if (mp) { 1026 n = sbunlinkmbuf(&so->so_rcv, m, NULL); 1027 *mp = m; 1028 mp = &m->m_next; 1029 m = n; 1030 } else { 1031 m = sbunlinkmbuf(&so->so_rcv, m, &free_chain); 1032 } 1033 } 1034 } else { 1035 if (flags & MSG_PEEK) { 1036 moff += len; 1037 } else { 1038 if (mp) 1039 *mp = m_copym(m, 0, len, MB_WAIT); 1040 m->m_data += len; 1041 m->m_len -= len; 1042 so->so_rcv.sb_cc -= len; 1043 } 1044 } 1045 if (so->so_oobmark) { 1046 if ((flags & MSG_PEEK) == 0) { 1047 so->so_oobmark -= len; 1048 if (so->so_oobmark == 0) { 1049 so->so_state |= SS_RCVATMARK; 1050 break; 1051 } 1052 } else { 1053 offset += len; 1054 if (offset == so->so_oobmark) 1055 break; 1056 } 1057 } 1058 if (flags & MSG_EOR) 1059 break; 1060 /* 1061 * If the MSG_WAITALL flag is set (for non-atomic socket), 1062 * we must not quit until "uio->uio_resid == 0" or an error 1063 * termination. If a signal/timeout occurs, return 1064 * with a short count but without error. 1065 * Keep sockbuf locked against other readers. 1066 */ 1067 while (flags & MSG_WAITALL && m == NULL && 1068 uio->uio_resid > 0 && !sosendallatonce(so) && 1069 so->so_rcv.sb_mb == NULL) { 1070 if (so->so_error || so->so_state & SS_CANTRCVMORE) 1071 break; 1072 /* 1073 * The window might have closed to zero, make 1074 * sure we send an ack now that we've drained 1075 * the buffer or we might end up blocking until 1076 * the idle takes over (5 seconds). 1077 */ 1078 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1079 so_pru_rcvd(so, flags); 1080 error = sbwait(&so->so_rcv); 1081 if (error) { 1082 sbunlock(&so->so_rcv); 1083 error = 0; 1084 goto done; 1085 } 1086 m = so->so_rcv.sb_mb; 1087 } 1088 } 1089 1090 /* 1091 * If an atomic read was requested but unread data still remains 1092 * in the record, set MSG_TRUNC. 1093 */ 1094 if (m && pr->pr_flags & PR_ATOMIC) 1095 flags |= MSG_TRUNC; 1096 1097 /* 1098 * Cleanup. If an atomic read was requested drop any unread data. 1099 */ 1100 if ((flags & MSG_PEEK) == 0) { 1101 if (m && (pr->pr_flags & PR_ATOMIC)) 1102 sbdroprecord(&so->so_rcv); 1103 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) 1104 so_pru_rcvd(so, flags); 1105 } 1106 1107 if (orig_resid == uio->uio_resid && orig_resid && 1108 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 1109 sbunlock(&so->so_rcv); 1110 crit_exit(); 1111 goto restart; 1112 } 1113 1114 if (flagsp) 1115 *flagsp |= flags; 1116 release: 1117 sbunlock(&so->so_rcv); 1118 done: 1119 crit_exit(); 1120 if (free_chain) 1121 m_freem(free_chain); 1122 return (error); 1123 } 1124 1125 int 1126 soshutdown(so, how) 1127 struct socket *so; 1128 int how; 1129 { 1130 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 1131 return (EINVAL); 1132 1133 if (how != SHUT_WR) 1134 sorflush(so); 1135 if (how != SHUT_RD) 1136 return (so_pru_shutdown(so)); 1137 return (0); 1138 } 1139 1140 void 1141 sorflush(so) 1142 struct socket *so; 1143 { 1144 struct sockbuf *sb = &so->so_rcv; 1145 struct protosw *pr = so->so_proto; 1146 struct sockbuf asb; 1147 1148 sb->sb_flags |= SB_NOINTR; 1149 (void) sblock(sb, M_WAITOK); 1150 1151 crit_enter(); 1152 socantrcvmore(so); 1153 sbunlock(sb); 1154 asb = *sb; 1155 bzero((caddr_t)sb, sizeof (*sb)); 1156 if (asb.sb_flags & SB_KNOTE) { 1157 sb->sb_sel.si_note = asb.sb_sel.si_note; 1158 sb->sb_flags = SB_KNOTE; 1159 } 1160 crit_exit(); 1161 1162 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) 1163 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 1164 sbrelease(&asb, so); 1165 } 1166 1167 #ifdef INET 1168 static int 1169 do_setopt_accept_filter(so, sopt) 1170 struct socket *so; 1171 struct sockopt *sopt; 1172 { 1173 struct accept_filter_arg *afap = NULL; 1174 struct accept_filter *afp; 1175 struct so_accf *af = so->so_accf; 1176 int error = 0; 1177 1178 /* do not set/remove accept filters on non listen sockets */ 1179 if ((so->so_options & SO_ACCEPTCONN) == 0) { 1180 error = EINVAL; 1181 goto out; 1182 } 1183 1184 /* removing the filter */ 1185 if (sopt == NULL) { 1186 if (af != NULL) { 1187 if (af->so_accept_filter != NULL && 1188 af->so_accept_filter->accf_destroy != NULL) { 1189 af->so_accept_filter->accf_destroy(so); 1190 } 1191 if (af->so_accept_filter_str != NULL) { 1192 FREE(af->so_accept_filter_str, M_ACCF); 1193 } 1194 FREE(af, M_ACCF); 1195 so->so_accf = NULL; 1196 } 1197 so->so_options &= ~SO_ACCEPTFILTER; 1198 return (0); 1199 } 1200 /* adding a filter */ 1201 /* must remove previous filter first */ 1202 if (af != NULL) { 1203 error = EINVAL; 1204 goto out; 1205 } 1206 /* don't put large objects on the kernel stack */ 1207 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK); 1208 error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap); 1209 afap->af_name[sizeof(afap->af_name)-1] = '\0'; 1210 afap->af_arg[sizeof(afap->af_arg)-1] = '\0'; 1211 if (error) 1212 goto out; 1213 afp = accept_filt_get(afap->af_name); 1214 if (afp == NULL) { 1215 error = ENOENT; 1216 goto out; 1217 } 1218 MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK); 1219 bzero(af, sizeof(*af)); 1220 if (afp->accf_create != NULL) { 1221 if (afap->af_name[0] != '\0') { 1222 int len = strlen(afap->af_name) + 1; 1223 1224 MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK); 1225 strcpy(af->so_accept_filter_str, afap->af_name); 1226 } 1227 af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg); 1228 if (af->so_accept_filter_arg == NULL) { 1229 FREE(af->so_accept_filter_str, M_ACCF); 1230 FREE(af, M_ACCF); 1231 so->so_accf = NULL; 1232 error = EINVAL; 1233 goto out; 1234 } 1235 } 1236 af->so_accept_filter = afp; 1237 so->so_accf = af; 1238 so->so_options |= SO_ACCEPTFILTER; 1239 out: 1240 if (afap != NULL) 1241 FREE(afap, M_TEMP); 1242 return (error); 1243 } 1244 #endif /* INET */ 1245 1246 /* 1247 * Perhaps this routine, and sooptcopyout(), below, ought to come in 1248 * an additional variant to handle the case where the option value needs 1249 * to be some kind of integer, but not a specific size. 1250 * In addition to their use here, these functions are also called by the 1251 * protocol-level pr_ctloutput() routines. 1252 */ 1253 int 1254 sooptcopyin(sopt, buf, len, minlen) 1255 struct sockopt *sopt; 1256 void *buf; 1257 size_t len; 1258 size_t minlen; 1259 { 1260 size_t valsize; 1261 1262 /* 1263 * If the user gives us more than we wanted, we ignore it, 1264 * but if we don't get the minimum length the caller 1265 * wants, we return EINVAL. On success, sopt->sopt_valsize 1266 * is set to however much we actually retrieved. 1267 */ 1268 if ((valsize = sopt->sopt_valsize) < minlen) 1269 return EINVAL; 1270 if (valsize > len) 1271 sopt->sopt_valsize = valsize = len; 1272 1273 if (sopt->sopt_td != NULL) 1274 return (copyin(sopt->sopt_val, buf, valsize)); 1275 1276 bcopy(sopt->sopt_val, buf, valsize); 1277 return 0; 1278 } 1279 1280 int 1281 sosetopt(so, sopt) 1282 struct socket *so; 1283 struct sockopt *sopt; 1284 { 1285 int error, optval; 1286 struct linger l; 1287 struct timeval tv; 1288 u_long val; 1289 1290 error = 0; 1291 sopt->sopt_dir = SOPT_SET; 1292 if (sopt->sopt_level != SOL_SOCKET) { 1293 if (so->so_proto && so->so_proto->pr_ctloutput) { 1294 return (so_pr_ctloutput(so, sopt)); 1295 } 1296 error = ENOPROTOOPT; 1297 } else { 1298 switch (sopt->sopt_name) { 1299 #ifdef INET 1300 case SO_ACCEPTFILTER: 1301 error = do_setopt_accept_filter(so, sopt); 1302 if (error) 1303 goto bad; 1304 break; 1305 #endif /* INET */ 1306 case SO_LINGER: 1307 error = sooptcopyin(sopt, &l, sizeof l, sizeof l); 1308 if (error) 1309 goto bad; 1310 1311 so->so_linger = l.l_linger; 1312 if (l.l_onoff) 1313 so->so_options |= SO_LINGER; 1314 else 1315 so->so_options &= ~SO_LINGER; 1316 break; 1317 1318 case SO_DEBUG: 1319 case SO_KEEPALIVE: 1320 case SO_DONTROUTE: 1321 case SO_USELOOPBACK: 1322 case SO_BROADCAST: 1323 case SO_REUSEADDR: 1324 case SO_REUSEPORT: 1325 case SO_OOBINLINE: 1326 case SO_TIMESTAMP: 1327 error = sooptcopyin(sopt, &optval, sizeof optval, 1328 sizeof optval); 1329 if (error) 1330 goto bad; 1331 if (optval) 1332 so->so_options |= sopt->sopt_name; 1333 else 1334 so->so_options &= ~sopt->sopt_name; 1335 break; 1336 1337 case SO_SNDBUF: 1338 case SO_RCVBUF: 1339 case SO_SNDLOWAT: 1340 case SO_RCVLOWAT: 1341 error = sooptcopyin(sopt, &optval, sizeof optval, 1342 sizeof optval); 1343 if (error) 1344 goto bad; 1345 1346 /* 1347 * Values < 1 make no sense for any of these 1348 * options, so disallow them. 1349 */ 1350 if (optval < 1) { 1351 error = EINVAL; 1352 goto bad; 1353 } 1354 1355 switch (sopt->sopt_name) { 1356 case SO_SNDBUF: 1357 case SO_RCVBUF: 1358 if (sbreserve(sopt->sopt_name == SO_SNDBUF ? 1359 &so->so_snd : &so->so_rcv, (u_long)optval, 1360 so, 1361 &curproc->p_rlimit[RLIMIT_SBSIZE]) == 0) { 1362 error = ENOBUFS; 1363 goto bad; 1364 } 1365 break; 1366 1367 /* 1368 * Make sure the low-water is never greater than 1369 * the high-water. 1370 */ 1371 case SO_SNDLOWAT: 1372 so->so_snd.sb_lowat = 1373 (optval > so->so_snd.sb_hiwat) ? 1374 so->so_snd.sb_hiwat : optval; 1375 break; 1376 case SO_RCVLOWAT: 1377 so->so_rcv.sb_lowat = 1378 (optval > so->so_rcv.sb_hiwat) ? 1379 so->so_rcv.sb_hiwat : optval; 1380 break; 1381 } 1382 break; 1383 1384 case SO_SNDTIMEO: 1385 case SO_RCVTIMEO: 1386 error = sooptcopyin(sopt, &tv, sizeof tv, 1387 sizeof tv); 1388 if (error) 1389 goto bad; 1390 1391 /* assert(hz > 0); */ 1392 if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz || 1393 tv.tv_usec < 0 || tv.tv_usec >= 1000000) { 1394 error = EDOM; 1395 goto bad; 1396 } 1397 /* assert(tick > 0); */ 1398 /* assert(ULONG_MAX - SHRT_MAX >= 1000000); */ 1399 val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick; 1400 if (val > SHRT_MAX) { 1401 error = EDOM; 1402 goto bad; 1403 } 1404 if (val == 0 && tv.tv_usec != 0) 1405 val = 1; 1406 1407 switch (sopt->sopt_name) { 1408 case SO_SNDTIMEO: 1409 so->so_snd.sb_timeo = val; 1410 break; 1411 case SO_RCVTIMEO: 1412 so->so_rcv.sb_timeo = val; 1413 break; 1414 } 1415 break; 1416 default: 1417 error = ENOPROTOOPT; 1418 break; 1419 } 1420 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) { 1421 (void) so_pr_ctloutput(so, sopt); 1422 } 1423 } 1424 bad: 1425 return (error); 1426 } 1427 1428 /* Helper routine for getsockopt */ 1429 int 1430 sooptcopyout(struct sockopt *sopt, const void *buf, size_t len) 1431 { 1432 int error; 1433 size_t valsize; 1434 1435 error = 0; 1436 1437 /* 1438 * Documented get behavior is that we always return a value, 1439 * possibly truncated to fit in the user's buffer. 1440 * Traditional behavior is that we always tell the user 1441 * precisely how much we copied, rather than something useful 1442 * like the total amount we had available for her. 1443 * Note that this interface is not idempotent; the entire answer must 1444 * generated ahead of time. 1445 */ 1446 valsize = min(len, sopt->sopt_valsize); 1447 sopt->sopt_valsize = valsize; 1448 if (sopt->sopt_val != 0) { 1449 if (sopt->sopt_td != NULL) 1450 error = copyout(buf, sopt->sopt_val, valsize); 1451 else 1452 bcopy(buf, sopt->sopt_val, valsize); 1453 } 1454 return error; 1455 } 1456 1457 int 1458 sogetopt(so, sopt) 1459 struct socket *so; 1460 struct sockopt *sopt; 1461 { 1462 int error, optval; 1463 struct linger l; 1464 struct timeval tv; 1465 #ifdef INET 1466 struct accept_filter_arg *afap; 1467 #endif 1468 1469 error = 0; 1470 sopt->sopt_dir = SOPT_GET; 1471 if (sopt->sopt_level != SOL_SOCKET) { 1472 if (so->so_proto && so->so_proto->pr_ctloutput) { 1473 return (so_pr_ctloutput(so, sopt)); 1474 } else 1475 return (ENOPROTOOPT); 1476 } else { 1477 switch (sopt->sopt_name) { 1478 #ifdef INET 1479 case SO_ACCEPTFILTER: 1480 if ((so->so_options & SO_ACCEPTCONN) == 0) 1481 return (EINVAL); 1482 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), 1483 M_TEMP, M_WAITOK); 1484 bzero(afap, sizeof(*afap)); 1485 if ((so->so_options & SO_ACCEPTFILTER) != 0) { 1486 strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name); 1487 if (so->so_accf->so_accept_filter_str != NULL) 1488 strcpy(afap->af_arg, so->so_accf->so_accept_filter_str); 1489 } 1490 error = sooptcopyout(sopt, afap, sizeof(*afap)); 1491 FREE(afap, M_TEMP); 1492 break; 1493 #endif /* INET */ 1494 1495 case SO_LINGER: 1496 l.l_onoff = so->so_options & SO_LINGER; 1497 l.l_linger = so->so_linger; 1498 error = sooptcopyout(sopt, &l, sizeof l); 1499 break; 1500 1501 case SO_USELOOPBACK: 1502 case SO_DONTROUTE: 1503 case SO_DEBUG: 1504 case SO_KEEPALIVE: 1505 case SO_REUSEADDR: 1506 case SO_REUSEPORT: 1507 case SO_BROADCAST: 1508 case SO_OOBINLINE: 1509 case SO_TIMESTAMP: 1510 optval = so->so_options & sopt->sopt_name; 1511 integer: 1512 error = sooptcopyout(sopt, &optval, sizeof optval); 1513 break; 1514 1515 case SO_TYPE: 1516 optval = so->so_type; 1517 goto integer; 1518 1519 case SO_ERROR: 1520 optval = so->so_error; 1521 so->so_error = 0; 1522 goto integer; 1523 1524 case SO_SNDBUF: 1525 optval = so->so_snd.sb_hiwat; 1526 goto integer; 1527 1528 case SO_RCVBUF: 1529 optval = so->so_rcv.sb_hiwat; 1530 goto integer; 1531 1532 case SO_SNDLOWAT: 1533 optval = so->so_snd.sb_lowat; 1534 goto integer; 1535 1536 case SO_RCVLOWAT: 1537 optval = so->so_rcv.sb_lowat; 1538 goto integer; 1539 1540 case SO_SNDTIMEO: 1541 case SO_RCVTIMEO: 1542 optval = (sopt->sopt_name == SO_SNDTIMEO ? 1543 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 1544 1545 tv.tv_sec = optval / hz; 1546 tv.tv_usec = (optval % hz) * tick; 1547 error = sooptcopyout(sopt, &tv, sizeof tv); 1548 break; 1549 1550 default: 1551 error = ENOPROTOOPT; 1552 break; 1553 } 1554 return (error); 1555 } 1556 } 1557 1558 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */ 1559 int 1560 soopt_getm(struct sockopt *sopt, struct mbuf **mp) 1561 { 1562 struct mbuf *m, *m_prev; 1563 int sopt_size = sopt->sopt_valsize, msize; 1564 1565 m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_DATA, 1566 0, &msize); 1567 if (m == NULL) 1568 return (ENOBUFS); 1569 m->m_len = min(msize, sopt_size); 1570 sopt_size -= m->m_len; 1571 *mp = m; 1572 m_prev = m; 1573 1574 while (sopt_size > 0) { 1575 m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, 1576 MT_DATA, 0, &msize); 1577 if (m == NULL) { 1578 m_freem(*mp); 1579 return (ENOBUFS); 1580 } 1581 m->m_len = min(msize, sopt_size); 1582 sopt_size -= m->m_len; 1583 m_prev->m_next = m; 1584 m_prev = m; 1585 } 1586 return (0); 1587 } 1588 1589 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */ 1590 int 1591 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m) 1592 { 1593 struct mbuf *m0 = m; 1594 1595 if (sopt->sopt_val == NULL) 1596 return 0; 1597 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1598 if (sopt->sopt_td != NULL) { 1599 int error; 1600 1601 error = copyin(sopt->sopt_val, mtod(m, char *), 1602 m->m_len); 1603 if (error != 0) { 1604 m_freem(m0); 1605 return (error); 1606 } 1607 } else 1608 bcopy(sopt->sopt_val, mtod(m, char *), m->m_len); 1609 sopt->sopt_valsize -= m->m_len; 1610 sopt->sopt_val = (caddr_t)sopt->sopt_val + m->m_len; 1611 m = m->m_next; 1612 } 1613 if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */ 1614 panic("ip6_sooptmcopyin"); 1615 return 0; 1616 } 1617 1618 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */ 1619 int 1620 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m) 1621 { 1622 struct mbuf *m0 = m; 1623 size_t valsize = 0; 1624 1625 if (sopt->sopt_val == NULL) 1626 return 0; 1627 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1628 if (sopt->sopt_td != NULL) { 1629 int error; 1630 1631 error = copyout(mtod(m, char *), sopt->sopt_val, 1632 m->m_len); 1633 if (error != 0) { 1634 m_freem(m0); 1635 return (error); 1636 } 1637 } else 1638 bcopy(mtod(m, char *), sopt->sopt_val, m->m_len); 1639 sopt->sopt_valsize -= m->m_len; 1640 sopt->sopt_val = (caddr_t)sopt->sopt_val + m->m_len; 1641 valsize += m->m_len; 1642 m = m->m_next; 1643 } 1644 if (m != NULL) { 1645 /* enough soopt buffer should be given from user-land */ 1646 m_freem(m0); 1647 return (EINVAL); 1648 } 1649 sopt->sopt_valsize = valsize; 1650 return 0; 1651 } 1652 1653 void 1654 sohasoutofband(so) 1655 struct socket *so; 1656 { 1657 if (so->so_sigio != NULL) 1658 pgsigio(so->so_sigio, SIGURG, 0); 1659 selwakeup(&so->so_rcv.sb_sel); 1660 } 1661 1662 int 1663 sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td) 1664 { 1665 int revents = 0; 1666 1667 crit_enter(); 1668 1669 if (events & (POLLIN | POLLRDNORM)) 1670 if (soreadable(so)) 1671 revents |= events & (POLLIN | POLLRDNORM); 1672 1673 if (events & POLLINIGNEOF) 1674 if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat || 1675 !TAILQ_EMPTY(&so->so_comp) || so->so_error) 1676 revents |= POLLINIGNEOF; 1677 1678 if (events & (POLLOUT | POLLWRNORM)) 1679 if (sowriteable(so)) 1680 revents |= events & (POLLOUT | POLLWRNORM); 1681 1682 if (events & (POLLPRI | POLLRDBAND)) 1683 if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) 1684 revents |= events & (POLLPRI | POLLRDBAND); 1685 1686 if (revents == 0) { 1687 if (events & 1688 (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM | 1689 POLLRDBAND)) { 1690 selrecord(td, &so->so_rcv.sb_sel); 1691 so->so_rcv.sb_flags |= SB_SEL; 1692 } 1693 1694 if (events & (POLLOUT | POLLWRNORM)) { 1695 selrecord(td, &so->so_snd.sb_sel); 1696 so->so_snd.sb_flags |= SB_SEL; 1697 } 1698 } 1699 1700 crit_exit(); 1701 return (revents); 1702 } 1703 1704 int 1705 sokqfilter(struct file *fp, struct knote *kn) 1706 { 1707 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1708 struct sockbuf *sb; 1709 1710 switch (kn->kn_filter) { 1711 case EVFILT_READ: 1712 if (so->so_options & SO_ACCEPTCONN) 1713 kn->kn_fop = &solisten_filtops; 1714 else 1715 kn->kn_fop = &soread_filtops; 1716 sb = &so->so_rcv; 1717 break; 1718 case EVFILT_WRITE: 1719 kn->kn_fop = &sowrite_filtops; 1720 sb = &so->so_snd; 1721 break; 1722 default: 1723 return (1); 1724 } 1725 1726 crit_enter(); 1727 SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext); 1728 sb->sb_flags |= SB_KNOTE; 1729 crit_exit(); 1730 return (0); 1731 } 1732 1733 static void 1734 filt_sordetach(struct knote *kn) 1735 { 1736 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1737 1738 crit_enter(); 1739 SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext); 1740 if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note)) 1741 so->so_rcv.sb_flags &= ~SB_KNOTE; 1742 crit_exit(); 1743 } 1744 1745 /*ARGSUSED*/ 1746 static int 1747 filt_soread(struct knote *kn, long hint) 1748 { 1749 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1750 1751 kn->kn_data = so->so_rcv.sb_cc; 1752 if (so->so_state & SS_CANTRCVMORE) { 1753 kn->kn_flags |= EV_EOF; 1754 kn->kn_fflags = so->so_error; 1755 return (1); 1756 } 1757 if (so->so_error) /* temporary udp error */ 1758 return (1); 1759 if (kn->kn_sfflags & NOTE_LOWAT) 1760 return (kn->kn_data >= kn->kn_sdata); 1761 return (kn->kn_data >= so->so_rcv.sb_lowat); 1762 } 1763 1764 static void 1765 filt_sowdetach(struct knote *kn) 1766 { 1767 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1768 1769 crit_enter(); 1770 SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext); 1771 if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note)) 1772 so->so_snd.sb_flags &= ~SB_KNOTE; 1773 crit_exit(); 1774 } 1775 1776 /*ARGSUSED*/ 1777 static int 1778 filt_sowrite(struct knote *kn, long hint) 1779 { 1780 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1781 1782 kn->kn_data = sbspace(&so->so_snd); 1783 if (so->so_state & SS_CANTSENDMORE) { 1784 kn->kn_flags |= EV_EOF; 1785 kn->kn_fflags = so->so_error; 1786 return (1); 1787 } 1788 if (so->so_error) /* temporary udp error */ 1789 return (1); 1790 if (((so->so_state & SS_ISCONNECTED) == 0) && 1791 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 1792 return (0); 1793 if (kn->kn_sfflags & NOTE_LOWAT) 1794 return (kn->kn_data >= kn->kn_sdata); 1795 return (kn->kn_data >= so->so_snd.sb_lowat); 1796 } 1797 1798 /*ARGSUSED*/ 1799 static int 1800 filt_solisten(struct knote *kn, long hint) 1801 { 1802 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1803 1804 kn->kn_data = so->so_qlen; 1805 return (! TAILQ_EMPTY(&so->so_comp)); 1806 } 1807