1 /* 2 * Copyright (c) 2004 Jeffrey M. Hsu. All rights reserved. 3 * Copyright (c) 2004 The DragonFly Project. All rights reserved. 4 * 5 * This code is derived from software contributed to The DragonFly Project 6 * by Jeffrey M. Hsu. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of The DragonFly Project nor the names of its 17 * contributors may be used to endorse or promote products derived 18 * from this software without specific, prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 */ 33 34 /* 35 * Copyright (c) 1982, 1986, 1988, 1990, 1993 36 * The Regents of the University of California. All rights reserved. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 3. All advertising materials mentioning features or use of this software 47 * must display the following acknowledgement: 48 * This product includes software developed by the University of 49 * California, Berkeley and its contributors. 50 * 4. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * @(#)uipc_socket.c 8.3 (Berkeley) 4/15/94 67 * $FreeBSD: src/sys/kern/uipc_socket.c,v 1.68.2.24 2003/11/11 17:18:18 silby Exp $ 68 * $DragonFly: src/sys/kern/uipc_socket.c,v 1.46 2007/11/07 18:24:06 dillon Exp $ 69 */ 70 71 #include "opt_inet.h" 72 #include "opt_sctp.h" 73 74 #include <sys/param.h> 75 #include <sys/systm.h> 76 #include <sys/fcntl.h> 77 #include <sys/malloc.h> 78 #include <sys/mbuf.h> 79 #include <sys/domain.h> 80 #include <sys/file.h> /* for struct knote */ 81 #include <sys/kernel.h> 82 #include <sys/malloc.h> 83 #include <sys/event.h> 84 #include <sys/poll.h> 85 #include <sys/proc.h> 86 #include <sys/protosw.h> 87 #include <sys/socket.h> 88 #include <sys/socketvar.h> 89 #include <sys/socketops.h> 90 #include <sys/resourcevar.h> 91 #include <sys/signalvar.h> 92 #include <sys/sysctl.h> 93 #include <sys/uio.h> 94 #include <sys/jail.h> 95 #include <vm/vm_zone.h> 96 97 #include <sys/thread2.h> 98 #include <sys/socketvar2.h> 99 100 #include <machine/limits.h> 101 102 #ifdef INET 103 static int do_setopt_accept_filter(struct socket *so, struct sockopt *sopt); 104 #endif /* INET */ 105 106 static void filt_sordetach(struct knote *kn); 107 static int filt_soread(struct knote *kn, long hint); 108 static void filt_sowdetach(struct knote *kn); 109 static int filt_sowrite(struct knote *kn, long hint); 110 static int filt_solisten(struct knote *kn, long hint); 111 112 static struct filterops solisten_filtops = 113 { 1, NULL, filt_sordetach, filt_solisten }; 114 static struct filterops soread_filtops = 115 { 1, NULL, filt_sordetach, filt_soread }; 116 static struct filterops sowrite_filtops = 117 { 1, NULL, filt_sowdetach, filt_sowrite }; 118 119 struct vm_zone *socket_zone; 120 121 MALLOC_DEFINE(M_SONAME, "soname", "socket name"); 122 MALLOC_DEFINE(M_PCB, "pcb", "protocol control block"); 123 124 125 static int somaxconn = SOMAXCONN; 126 SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW, 127 &somaxconn, 0, "Maximum pending socket connection queue size"); 128 129 /* 130 * Socket operation routines. 131 * These routines are called by the routines in 132 * sys_socket.c or from a system process, and 133 * implement the semantics of socket operations by 134 * switching out to the protocol specific routines. 135 */ 136 137 /* 138 * Get a socket structure from our zone, and initialize it. 139 * We don't implement `waitok' yet (see comments in uipc_domain.c). 140 * Note that it would probably be better to allocate socket 141 * and PCB at the same time, but I'm not convinced that all 142 * the protocols can be easily modified to do this. 143 */ 144 struct socket * 145 soalloc(int waitok) 146 { 147 struct socket *so; 148 149 so = zalloc(socket_zone); 150 if (so) { 151 /* XXX race condition for reentrant kernel */ 152 bzero(so, sizeof *so); 153 TAILQ_INIT(&so->so_aiojobq); 154 TAILQ_INIT(&so->so_rcv.ssb_sel.si_mlist); 155 TAILQ_INIT(&so->so_snd.ssb_sel.si_mlist); 156 } 157 return so; 158 } 159 160 int 161 socreate(int dom, struct socket **aso, int type, 162 int proto, struct thread *td) 163 { 164 struct proc *p = td->td_proc; 165 struct protosw *prp; 166 struct socket *so; 167 struct pru_attach_info ai; 168 int error; 169 170 if (proto) 171 prp = pffindproto(dom, proto, type); 172 else 173 prp = pffindtype(dom, type); 174 175 if (prp == 0 || prp->pr_usrreqs->pru_attach == 0) 176 return (EPROTONOSUPPORT); 177 178 if (p->p_ucred->cr_prison && jail_socket_unixiproute_only && 179 prp->pr_domain->dom_family != PF_LOCAL && 180 prp->pr_domain->dom_family != PF_INET && 181 prp->pr_domain->dom_family != PF_INET6 && 182 prp->pr_domain->dom_family != PF_ROUTE) { 183 return (EPROTONOSUPPORT); 184 } 185 186 if (prp->pr_type != type) 187 return (EPROTOTYPE); 188 so = soalloc(p != 0); 189 if (so == 0) 190 return (ENOBUFS); 191 192 TAILQ_INIT(&so->so_incomp); 193 TAILQ_INIT(&so->so_comp); 194 so->so_type = type; 195 so->so_cred = crhold(p->p_ucred); 196 so->so_proto = prp; 197 ai.sb_rlimit = &p->p_rlimit[RLIMIT_SBSIZE]; 198 ai.p_ucred = p->p_ucred; 199 ai.fd_rdir = p->p_fd->fd_rdir; 200 error = so_pru_attach(so, proto, &ai); 201 if (error) { 202 so->so_state |= SS_NOFDREF; 203 sofree(so); 204 return (error); 205 } 206 *aso = so; 207 return (0); 208 } 209 210 int 211 sobind(struct socket *so, struct sockaddr *nam, struct thread *td) 212 { 213 int error; 214 215 crit_enter(); 216 error = so_pru_bind(so, nam, td); 217 crit_exit(); 218 return (error); 219 } 220 221 void 222 sodealloc(struct socket *so) 223 { 224 if (so->so_rcv.ssb_hiwat) 225 (void)chgsbsize(so->so_cred->cr_uidinfo, 226 &so->so_rcv.ssb_hiwat, 0, RLIM_INFINITY); 227 if (so->so_snd.ssb_hiwat) 228 (void)chgsbsize(so->so_cred->cr_uidinfo, 229 &so->so_snd.ssb_hiwat, 0, RLIM_INFINITY); 230 #ifdef INET 231 /* remove accept filter if present */ 232 if (so->so_accf != NULL) 233 do_setopt_accept_filter(so, NULL); 234 #endif /* INET */ 235 crfree(so->so_cred); 236 zfree(socket_zone, so); 237 } 238 239 int 240 solisten(struct socket *so, int backlog, struct thread *td) 241 { 242 int error; 243 #ifdef SCTP 244 short oldopt, oldqlimit; 245 #endif /* SCTP */ 246 247 crit_enter(); 248 if (so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING)) { 249 crit_exit(); 250 return (EINVAL); 251 } 252 253 #ifdef SCTP 254 oldopt = so->so_options; 255 oldqlimit = so->so_qlimit; 256 #endif /* SCTP */ 257 258 if (TAILQ_EMPTY(&so->so_comp)) 259 so->so_options |= SO_ACCEPTCONN; 260 if (backlog < 0 || backlog > somaxconn) 261 backlog = somaxconn; 262 so->so_qlimit = backlog; 263 /* SCTP needs to look at tweak both the inbound backlog parameter AND 264 * the so_options (UDP model both connect's and gets inbound 265 * connections .. implicitly). 266 */ 267 error = so_pru_listen(so, td); 268 if (error) { 269 #ifdef SCTP 270 /* Restore the params */ 271 so->so_options = oldopt; 272 so->so_qlimit = oldqlimit; 273 #endif /* SCTP */ 274 crit_exit(); 275 return (error); 276 } 277 crit_exit(); 278 return (0); 279 } 280 281 void 282 sofree(struct socket *so) 283 { 284 struct socket *head = so->so_head; 285 286 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) 287 return; 288 if (head != NULL) { 289 if (so->so_state & SS_INCOMP) { 290 TAILQ_REMOVE(&head->so_incomp, so, so_list); 291 head->so_incqlen--; 292 } else if (so->so_state & SS_COMP) { 293 /* 294 * We must not decommission a socket that's 295 * on the accept(2) queue. If we do, then 296 * accept(2) may hang after select(2) indicated 297 * that the listening socket was ready. 298 */ 299 return; 300 } else { 301 panic("sofree: not queued"); 302 } 303 so->so_state &= ~SS_INCOMP; 304 so->so_head = NULL; 305 } 306 ssb_release(&so->so_snd, so); 307 sorflush(so); 308 sodealloc(so); 309 } 310 311 /* 312 * Close a socket on last file table reference removal. 313 * Initiate disconnect if connected. 314 * Free socket when disconnect complete. 315 */ 316 int 317 soclose(struct socket *so, int fflag) 318 { 319 int error = 0; 320 321 crit_enter(); 322 funsetown(so->so_sigio); 323 if (so->so_pcb == NULL) 324 goto discard; 325 if (so->so_state & SS_ISCONNECTED) { 326 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 327 error = sodisconnect(so); 328 if (error) 329 goto drop; 330 } 331 if (so->so_options & SO_LINGER) { 332 if ((so->so_state & SS_ISDISCONNECTING) && 333 (fflag & FNONBLOCK)) 334 goto drop; 335 while (so->so_state & SS_ISCONNECTED) { 336 error = tsleep((caddr_t)&so->so_timeo, 337 PCATCH, "soclos", so->so_linger * hz); 338 if (error) 339 break; 340 } 341 } 342 } 343 drop: 344 if (so->so_pcb) { 345 int error2; 346 347 error2 = so_pru_detach(so); 348 if (error == 0) 349 error = error2; 350 } 351 discard: 352 if (so->so_options & SO_ACCEPTCONN) { 353 struct socket *sp, *sonext; 354 355 sp = TAILQ_FIRST(&so->so_incomp); 356 for (; sp != NULL; sp = sonext) { 357 sonext = TAILQ_NEXT(sp, so_list); 358 (void) soabort(sp); 359 } 360 for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) { 361 sonext = TAILQ_NEXT(sp, so_list); 362 /* Dequeue from so_comp since sofree() won't do it */ 363 TAILQ_REMOVE(&so->so_comp, sp, so_list); 364 so->so_qlen--; 365 sp->so_state &= ~SS_COMP; 366 sp->so_head = NULL; 367 (void) soabort(sp); 368 } 369 } 370 if (so->so_state & SS_NOFDREF) 371 panic("soclose: NOFDREF"); 372 so->so_state |= SS_NOFDREF; 373 sofree(so); 374 crit_exit(); 375 return (error); 376 } 377 378 /* 379 * Must be called from a critical section. 380 */ 381 int 382 soabort(struct socket *so) 383 { 384 int error; 385 386 error = so_pru_abort(so); 387 if (error) { 388 sofree(so); 389 return error; 390 } 391 return (0); 392 } 393 394 int 395 soaccept(struct socket *so, struct sockaddr **nam) 396 { 397 int error; 398 399 crit_enter(); 400 if ((so->so_state & SS_NOFDREF) == 0) 401 panic("soaccept: !NOFDREF"); 402 so->so_state &= ~SS_NOFDREF; 403 error = so_pru_accept(so, nam); 404 crit_exit(); 405 return (error); 406 } 407 408 int 409 soconnect(struct socket *so, struct sockaddr *nam, struct thread *td) 410 { 411 int error; 412 413 if (so->so_options & SO_ACCEPTCONN) 414 return (EOPNOTSUPP); 415 crit_enter(); 416 /* 417 * If protocol is connection-based, can only connect once. 418 * Otherwise, if connected, try to disconnect first. 419 * This allows user to disconnect by connecting to, e.g., 420 * a null address. 421 */ 422 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 423 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 424 (error = sodisconnect(so)))) { 425 error = EISCONN; 426 } else { 427 /* 428 * Prevent accumulated error from previous connection 429 * from biting us. 430 */ 431 so->so_error = 0; 432 error = so_pru_connect(so, nam, td); 433 } 434 crit_exit(); 435 return (error); 436 } 437 438 int 439 soconnect2(struct socket *so1, struct socket *so2) 440 { 441 int error; 442 443 crit_enter(); 444 error = so_pru_connect2(so1, so2); 445 crit_exit(); 446 return (error); 447 } 448 449 int 450 sodisconnect(struct socket *so) 451 { 452 int error; 453 454 crit_enter(); 455 if ((so->so_state & SS_ISCONNECTED) == 0) { 456 error = ENOTCONN; 457 goto bad; 458 } 459 if (so->so_state & SS_ISDISCONNECTING) { 460 error = EALREADY; 461 goto bad; 462 } 463 error = so_pru_disconnect(so); 464 bad: 465 crit_exit(); 466 return (error); 467 } 468 469 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 470 /* 471 * Send on a socket. 472 * If send must go all at once and message is larger than 473 * send buffering, then hard error. 474 * Lock against other senders. 475 * If must go all at once and not enough room now, then 476 * inform user that this would block and do nothing. 477 * Otherwise, if nonblocking, send as much as possible. 478 * The data to be sent is described by "uio" if nonzero, 479 * otherwise by the mbuf chain "top" (which must be null 480 * if uio is not). Data provided in mbuf chain must be small 481 * enough to send all at once. 482 * 483 * Returns nonzero on error, timeout or signal; callers 484 * must check for short counts if EINTR/ERESTART are returned. 485 * Data and control buffers are freed on return. 486 */ 487 int 488 sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, 489 struct mbuf *top, struct mbuf *control, int flags, 490 struct thread *td) 491 { 492 struct mbuf **mp; 493 struct mbuf *m; 494 long space, len, resid; 495 int clen = 0, error, dontroute, mlen; 496 int atomic = sosendallatonce(so) || top; 497 int pru_flags; 498 499 if (uio) 500 resid = uio->uio_resid; 501 else 502 resid = top->m_pkthdr.len; 503 /* 504 * In theory resid should be unsigned. 505 * However, space must be signed, as it might be less than 0 506 * if we over-committed, and we must use a signed comparison 507 * of space and resid. On the other hand, a negative resid 508 * causes us to loop sending 0-length segments to the protocol. 509 * 510 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM 511 * type sockets since that's an error. 512 */ 513 if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) { 514 error = EINVAL; 515 goto out; 516 } 517 518 dontroute = 519 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 520 (so->so_proto->pr_flags & PR_ATOMIC); 521 if (td->td_lwp != NULL) 522 td->td_lwp->lwp_ru.ru_msgsnd++; 523 if (control) 524 clen = control->m_len; 525 #define gotoerr(errcode) { error = errcode; crit_exit(); goto release; } 526 527 restart: 528 error = ssb_lock(&so->so_snd, SBLOCKWAIT(flags)); 529 if (error) 530 goto out; 531 do { 532 crit_enter(); 533 if (so->so_state & SS_CANTSENDMORE) 534 gotoerr(EPIPE); 535 if (so->so_error) { 536 error = so->so_error; 537 so->so_error = 0; 538 crit_exit(); 539 goto release; 540 } 541 if ((so->so_state & SS_ISCONNECTED) == 0) { 542 /* 543 * `sendto' and `sendmsg' is allowed on a connection- 544 * based socket if it supports implied connect. 545 * Return ENOTCONN if not connected and no address is 546 * supplied. 547 */ 548 if ((so->so_proto->pr_flags & PR_CONNREQUIRED) && 549 (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) { 550 if ((so->so_state & SS_ISCONFIRMING) == 0 && 551 !(resid == 0 && clen != 0)) 552 gotoerr(ENOTCONN); 553 } else if (addr == 0) 554 gotoerr(so->so_proto->pr_flags & PR_CONNREQUIRED ? 555 ENOTCONN : EDESTADDRREQ); 556 } 557 space = ssb_space(&so->so_snd); 558 if (flags & MSG_OOB) 559 space += 1024; 560 if ((atomic && resid > so->so_snd.ssb_hiwat) || 561 clen > so->so_snd.ssb_hiwat) 562 gotoerr(EMSGSIZE); 563 if (space < resid + clen && uio && 564 (atomic || space < so->so_snd.ssb_lowat || space < clen)) { 565 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) 566 gotoerr(EWOULDBLOCK); 567 ssb_unlock(&so->so_snd); 568 error = ssb_wait(&so->so_snd); 569 crit_exit(); 570 if (error) 571 goto out; 572 goto restart; 573 } 574 crit_exit(); 575 mp = ⊤ 576 space -= clen; 577 do { 578 if (uio == NULL) { 579 /* 580 * Data is prepackaged in "top". 581 */ 582 resid = 0; 583 if (flags & MSG_EOR) 584 top->m_flags |= M_EOR; 585 } else do { 586 m = m_getl(resid, MB_WAIT, MT_DATA, 587 top == NULL ? M_PKTHDR : 0, &mlen); 588 if (top == NULL) { 589 m->m_pkthdr.len = 0; 590 m->m_pkthdr.rcvif = (struct ifnet *)0; 591 } 592 len = min(min(mlen, resid), space); 593 if (resid < MINCLSIZE) { 594 /* 595 * For datagram protocols, leave room 596 * for protocol headers in first mbuf. 597 */ 598 if (atomic && top == 0 && len < mlen) 599 MH_ALIGN(m, len); 600 } 601 space -= len; 602 error = uiomove(mtod(m, caddr_t), (int)len, uio); 603 resid = uio->uio_resid; 604 m->m_len = len; 605 *mp = m; 606 top->m_pkthdr.len += len; 607 if (error) 608 goto release; 609 mp = &m->m_next; 610 if (resid <= 0) { 611 if (flags & MSG_EOR) 612 top->m_flags |= M_EOR; 613 break; 614 } 615 } while (space > 0 && atomic); 616 if (dontroute) 617 so->so_options |= SO_DONTROUTE; 618 if (flags & MSG_OOB) { 619 pru_flags = PRUS_OOB; 620 } else if ((flags & MSG_EOF) && 621 (so->so_proto->pr_flags & PR_IMPLOPCL) && 622 (resid <= 0)) { 623 /* 624 * If the user set MSG_EOF, the protocol 625 * understands this flag and nothing left to 626 * send then use PRU_SEND_EOF instead of PRU_SEND. 627 */ 628 pru_flags = PRUS_EOF; 629 } else if (resid > 0 && space > 0) { 630 /* If there is more to send, set PRUS_MORETOCOME */ 631 pru_flags = PRUS_MORETOCOME; 632 } else { 633 pru_flags = 0; 634 } 635 crit_enter(); 636 /* 637 * XXX all the SS_CANTSENDMORE checks previously 638 * done could be out of date. We could have recieved 639 * a reset packet in an interrupt or maybe we slept 640 * while doing page faults in uiomove() etc. We could 641 * probably recheck again inside the splnet() protection 642 * here, but there are probably other places that this 643 * also happens. We must rethink this. 644 */ 645 error = so_pru_send(so, pru_flags, top, addr, control, td); 646 crit_exit(); 647 if (dontroute) 648 so->so_options &= ~SO_DONTROUTE; 649 clen = 0; 650 control = 0; 651 top = 0; 652 mp = ⊤ 653 if (error) 654 goto release; 655 } while (resid && space > 0); 656 } while (resid); 657 658 release: 659 ssb_unlock(&so->so_snd); 660 out: 661 if (top) 662 m_freem(top); 663 if (control) 664 m_freem(control); 665 return (error); 666 } 667 668 /* 669 * A specialization of sosend() for UDP based on protocol-specific knowledge: 670 * so->so_proto->pr_flags has the PR_ATOMIC field set. This means that 671 * sosendallatonce() returns true, 672 * the "atomic" variable is true, 673 * and sosendudp() blocks until space is available for the entire send. 674 * so->so_proto->pr_flags does not have the PR_CONNREQUIRED or 675 * PR_IMPLOPCL flags set. 676 * UDP has no out-of-band data. 677 * UDP has no control data. 678 * UDP does not support MSG_EOR. 679 */ 680 int 681 sosendudp(struct socket *so, struct sockaddr *addr, struct uio *uio, 682 struct mbuf *top, struct mbuf *control, int flags, struct thread *td) 683 { 684 int resid, error; 685 boolean_t dontroute; /* temporary SO_DONTROUTE setting */ 686 687 if (td->td_lwp != NULL) 688 td->td_lwp->lwp_ru.ru_msgsnd++; 689 if (control) 690 m_freem(control); 691 692 KASSERT((uio && !top) || (top && !uio), ("bad arguments to sosendudp")); 693 resid = uio ? uio->uio_resid : top->m_pkthdr.len; 694 695 restart: 696 error = ssb_lock(&so->so_snd, SBLOCKWAIT(flags)); 697 if (error) 698 goto out; 699 700 crit_enter(); 701 if (so->so_state & SS_CANTSENDMORE) 702 gotoerr(EPIPE); 703 if (so->so_error) { 704 error = so->so_error; 705 so->so_error = 0; 706 crit_exit(); 707 goto release; 708 } 709 if (!(so->so_state & SS_ISCONNECTED) && addr == NULL) 710 gotoerr(EDESTADDRREQ); 711 if (resid > so->so_snd.ssb_hiwat) 712 gotoerr(EMSGSIZE); 713 if (uio && ssb_space(&so->so_snd) < resid) { 714 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) 715 gotoerr(EWOULDBLOCK); 716 ssb_unlock(&so->so_snd); 717 error = ssb_wait(&so->so_snd); 718 crit_exit(); 719 if (error) 720 goto out; 721 goto restart; 722 } 723 crit_exit(); 724 725 if (uio) { 726 top = m_uiomove(uio); 727 if (top == NULL) 728 goto release; 729 } 730 731 dontroute = (flags & MSG_DONTROUTE) && !(so->so_options & SO_DONTROUTE); 732 if (dontroute) 733 so->so_options |= SO_DONTROUTE; 734 735 error = so_pru_send(so, 0, top, addr, NULL, td); 736 top = NULL; /* sent or freed in lower layer */ 737 738 if (dontroute) 739 so->so_options &= ~SO_DONTROUTE; 740 741 release: 742 ssb_unlock(&so->so_snd); 743 out: 744 if (top) 745 m_freem(top); 746 return (error); 747 } 748 749 /* 750 * Implement receive operations on a socket. 751 * We depend on the way that records are added to the signalsockbuf 752 * by sbappend*. In particular, each record (mbufs linked through m_next) 753 * must begin with an address if the protocol so specifies, 754 * followed by an optional mbuf or mbufs containing ancillary data, 755 * and then zero or more mbufs of data. 756 * In order to avoid blocking network interrupts for the entire time here, 757 * we exit the critical section while doing the actual copy to user space. 758 * Although the signalsockbuf is locked, new data may still be appended, 759 * and thus we must maintain consistency of the signalsockbuf during that time. 760 * 761 * The caller may receive the data as a single mbuf chain by supplying 762 * an mbuf **mp0 for use in returning the chain. The uio is then used 763 * only for the count in uio_resid. 764 */ 765 int 766 soreceive(struct socket *so, struct sockaddr **psa, struct uio *uio, 767 struct sockbuf *sio, struct mbuf **controlp, int *flagsp) 768 { 769 struct mbuf *m, *n; 770 struct mbuf *free_chain = NULL; 771 int flags, len, error, offset; 772 struct protosw *pr = so->so_proto; 773 int moff, type = 0; 774 int resid, orig_resid; 775 776 if (uio) 777 resid = uio->uio_resid; 778 else 779 resid = (int)(sio->sb_climit - sio->sb_cc); 780 orig_resid = resid; 781 782 if (psa) 783 *psa = NULL; 784 if (controlp) 785 *controlp = NULL; 786 if (flagsp) 787 flags = *flagsp &~ MSG_EOR; 788 else 789 flags = 0; 790 if (flags & MSG_OOB) { 791 m = m_get(MB_WAIT, MT_DATA); 792 if (m == NULL) 793 return (ENOBUFS); 794 error = so_pru_rcvoob(so, m, flags & MSG_PEEK); 795 if (error) 796 goto bad; 797 if (sio) { 798 do { 799 sbappend(sio, m); 800 resid -= m->m_len; 801 } while (resid > 0 && m); 802 } else { 803 do { 804 uio->uio_resid = resid; 805 error = uiomove(mtod(m, caddr_t), 806 (int)min(resid, m->m_len), uio); 807 resid = uio->uio_resid; 808 m = m_free(m); 809 } while (uio->uio_resid && error == 0 && m); 810 } 811 bad: 812 if (m) 813 m_freem(m); 814 return (error); 815 } 816 if (so->so_state & SS_ISCONFIRMING && resid) 817 so_pru_rcvd(so, 0); 818 819 restart: 820 crit_enter(); 821 error = ssb_lock(&so->so_rcv, SBLOCKWAIT(flags)); 822 if (error) 823 goto done; 824 825 m = so->so_rcv.ssb_mb; 826 /* 827 * If we have less data than requested, block awaiting more 828 * (subject to any timeout) if: 829 * 1. the current count is less than the low water mark, or 830 * 2. MSG_WAITALL is set, and it is possible to do the entire 831 * receive operation at once if we block (resid <= hiwat). 832 * 3. MSG_DONTWAIT is not set 833 * If MSG_WAITALL is set but resid is larger than the receive buffer, 834 * we have to do the receive in sections, and thus risk returning 835 * a short count if a timeout or signal occurs after we start. 836 */ 837 if (m == NULL || (((flags & MSG_DONTWAIT) == 0 && 838 so->so_rcv.ssb_cc < resid) && 839 (so->so_rcv.ssb_cc < so->so_rcv.ssb_lowat || 840 ((flags & MSG_WAITALL) && resid <= so->so_rcv.ssb_hiwat)) && 841 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) { 842 KASSERT(m != NULL || !so->so_rcv.ssb_cc, ("receive 1")); 843 if (so->so_error) { 844 if (m) 845 goto dontblock; 846 error = so->so_error; 847 if ((flags & MSG_PEEK) == 0) 848 so->so_error = 0; 849 goto release; 850 } 851 if (so->so_state & SS_CANTRCVMORE) { 852 if (m) 853 goto dontblock; 854 else 855 goto release; 856 } 857 for (; m; m = m->m_next) { 858 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 859 m = so->so_rcv.ssb_mb; 860 goto dontblock; 861 } 862 } 863 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 864 (pr->pr_flags & PR_CONNREQUIRED)) { 865 error = ENOTCONN; 866 goto release; 867 } 868 if (resid == 0) 869 goto release; 870 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) { 871 error = EWOULDBLOCK; 872 goto release; 873 } 874 ssb_unlock(&so->so_rcv); 875 error = ssb_wait(&so->so_rcv); 876 if (error) 877 goto done; 878 crit_exit(); 879 goto restart; 880 } 881 dontblock: 882 if (uio && uio->uio_td && uio->uio_td->td_proc) 883 uio->uio_td->td_lwp->lwp_ru.ru_msgrcv++; 884 885 /* 886 * note: m should be == sb_mb here. Cache the next record while 887 * cleaning up. Note that calling m_free*() will break out critical 888 * section. 889 */ 890 KKASSERT(m == so->so_rcv.ssb_mb); 891 892 /* 893 * Skip any address mbufs prepending the record. 894 */ 895 if (pr->pr_flags & PR_ADDR) { 896 KASSERT(m->m_type == MT_SONAME, ("receive 1a")); 897 orig_resid = 0; 898 if (psa) 899 *psa = dup_sockaddr(mtod(m, struct sockaddr *)); 900 if (flags & MSG_PEEK) 901 m = m->m_next; 902 else 903 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain); 904 } 905 906 /* 907 * Skip any control mbufs prepending the record. 908 */ 909 #ifdef SCTP 910 if (pr->pr_flags & PR_ADDR_OPT) { 911 /* 912 * For SCTP we may be getting a 913 * whole message OR a partial delivery. 914 */ 915 if (m && m->m_type == MT_SONAME) { 916 orig_resid = 0; 917 if (psa) 918 *psa = dup_sockaddr(mtod(m, struct sockaddr *)); 919 if (flags & MSG_PEEK) 920 m = m->m_next; 921 else 922 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain); 923 } 924 } 925 #endif /* SCTP */ 926 while (m && m->m_type == MT_CONTROL && error == 0) { 927 if (flags & MSG_PEEK) { 928 if (controlp) 929 *controlp = m_copy(m, 0, m->m_len); 930 m = m->m_next; /* XXX race */ 931 } else { 932 if (controlp) { 933 n = sbunlinkmbuf(&so->so_rcv.sb, m, NULL); 934 if (pr->pr_domain->dom_externalize && 935 mtod(m, struct cmsghdr *)->cmsg_type == 936 SCM_RIGHTS) 937 error = (*pr->pr_domain->dom_externalize)(m); 938 *controlp = m; 939 m = n; 940 } else { 941 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain); 942 } 943 } 944 if (controlp && *controlp) { 945 orig_resid = 0; 946 controlp = &(*controlp)->m_next; 947 } 948 } 949 950 /* 951 * flag OOB data. 952 */ 953 if (m) { 954 type = m->m_type; 955 if (type == MT_OOBDATA) 956 flags |= MSG_OOB; 957 } 958 959 /* 960 * Copy to the UIO or mbuf return chain (*mp). 961 */ 962 moff = 0; 963 offset = 0; 964 while (m && resid > 0 && error == 0) { 965 if (m->m_type == MT_OOBDATA) { 966 if (type != MT_OOBDATA) 967 break; 968 } else if (type == MT_OOBDATA) 969 break; 970 else 971 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER, 972 ("receive 3")); 973 so->so_state &= ~SS_RCVATMARK; 974 len = resid; 975 if (so->so_oobmark && len > so->so_oobmark - offset) 976 len = so->so_oobmark - offset; 977 if (len > m->m_len - moff) 978 len = m->m_len - moff; 979 980 /* 981 * Copy out to the UIO or pass the mbufs back to the SIO. 982 * The SIO is dealt with when we eat the mbuf, but deal 983 * with the resid here either way. 984 */ 985 if (uio) { 986 crit_exit(); 987 uio->uio_resid = resid; 988 error = uiomove(mtod(m, caddr_t) + moff, len, uio); 989 resid = uio->uio_resid; 990 crit_enter(); 991 if (error) 992 goto release; 993 } else { 994 resid -= len; 995 } 996 997 /* 998 * Eat the entire mbuf or just a piece of it 999 */ 1000 if (len == m->m_len - moff) { 1001 if (m->m_flags & M_EOR) 1002 flags |= MSG_EOR; 1003 #ifdef SCTP 1004 if (m->m_flags & M_NOTIFICATION) 1005 flags |= MSG_NOTIFICATION; 1006 #endif /* SCTP */ 1007 if (flags & MSG_PEEK) { 1008 m = m->m_next; 1009 moff = 0; 1010 } else { 1011 if (sio) { 1012 n = sbunlinkmbuf(&so->so_rcv.sb, m, NULL); 1013 sbappend(sio, m); 1014 m = n; 1015 } else { 1016 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain); 1017 } 1018 } 1019 } else { 1020 if (flags & MSG_PEEK) { 1021 moff += len; 1022 } else { 1023 if (sio) { 1024 n = m_copym(m, 0, len, MB_WAIT); 1025 if (n) 1026 sbappend(sio, n); 1027 } 1028 m->m_data += len; 1029 m->m_len -= len; 1030 so->so_rcv.ssb_cc -= len; 1031 } 1032 } 1033 if (so->so_oobmark) { 1034 if ((flags & MSG_PEEK) == 0) { 1035 so->so_oobmark -= len; 1036 if (so->so_oobmark == 0) { 1037 so->so_state |= SS_RCVATMARK; 1038 break; 1039 } 1040 } else { 1041 offset += len; 1042 if (offset == so->so_oobmark) 1043 break; 1044 } 1045 } 1046 if (flags & MSG_EOR) 1047 break; 1048 /* 1049 * If the MSG_WAITALL flag is set (for non-atomic socket), 1050 * we must not quit until resid == 0 or an error 1051 * termination. If a signal/timeout occurs, return 1052 * with a short count but without error. 1053 * Keep signalsockbuf locked against other readers. 1054 */ 1055 while ((flags & MSG_WAITALL) && m == NULL && 1056 resid > 0 && !sosendallatonce(so) && 1057 so->so_rcv.ssb_mb == NULL) { 1058 if (so->so_error || so->so_state & SS_CANTRCVMORE) 1059 break; 1060 /* 1061 * The window might have closed to zero, make 1062 * sure we send an ack now that we've drained 1063 * the buffer or we might end up blocking until 1064 * the idle takes over (5 seconds). 1065 */ 1066 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1067 so_pru_rcvd(so, flags); 1068 error = ssb_wait(&so->so_rcv); 1069 if (error) { 1070 ssb_unlock(&so->so_rcv); 1071 error = 0; 1072 goto done; 1073 } 1074 m = so->so_rcv.ssb_mb; 1075 } 1076 } 1077 1078 /* 1079 * If an atomic read was requested but unread data still remains 1080 * in the record, set MSG_TRUNC. 1081 */ 1082 if (m && pr->pr_flags & PR_ATOMIC) 1083 flags |= MSG_TRUNC; 1084 1085 /* 1086 * Cleanup. If an atomic read was requested drop any unread data. 1087 */ 1088 if ((flags & MSG_PEEK) == 0) { 1089 if (m && (pr->pr_flags & PR_ATOMIC)) 1090 sbdroprecord(&so->so_rcv.sb); 1091 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) 1092 so_pru_rcvd(so, flags); 1093 } 1094 1095 if (orig_resid == resid && orig_resid && 1096 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 1097 ssb_unlock(&so->so_rcv); 1098 crit_exit(); 1099 goto restart; 1100 } 1101 1102 if (flagsp) 1103 *flagsp |= flags; 1104 release: 1105 ssb_unlock(&so->so_rcv); 1106 done: 1107 crit_exit(); 1108 if (free_chain) 1109 m_freem(free_chain); 1110 return (error); 1111 } 1112 1113 int 1114 soshutdown(struct socket *so, int how) 1115 { 1116 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 1117 return (EINVAL); 1118 1119 if (how != SHUT_WR) 1120 sorflush(so); 1121 if (how != SHUT_RD) 1122 return (so_pru_shutdown(so)); 1123 return (0); 1124 } 1125 1126 void 1127 sorflush(struct socket *so) 1128 { 1129 struct signalsockbuf *ssb = &so->so_rcv; 1130 struct protosw *pr = so->so_proto; 1131 struct signalsockbuf asb; 1132 1133 ssb->ssb_flags |= SSB_NOINTR; 1134 (void) ssb_lock(ssb, M_WAITOK); 1135 1136 crit_enter(); 1137 socantrcvmore(so); 1138 ssb_unlock(ssb); 1139 asb = *ssb; 1140 bzero((caddr_t)ssb, sizeof (*ssb)); 1141 if (asb.ssb_flags & SSB_KNOTE) { 1142 ssb->ssb_sel.si_note = asb.ssb_sel.si_note; 1143 ssb->ssb_flags = SSB_KNOTE; 1144 } 1145 crit_exit(); 1146 1147 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) 1148 (*pr->pr_domain->dom_dispose)(asb.ssb_mb); 1149 ssb_release(&asb, so); 1150 } 1151 1152 #ifdef INET 1153 static int 1154 do_setopt_accept_filter(struct socket *so, struct sockopt *sopt) 1155 { 1156 struct accept_filter_arg *afap = NULL; 1157 struct accept_filter *afp; 1158 struct so_accf *af = so->so_accf; 1159 int error = 0; 1160 1161 /* do not set/remove accept filters on non listen sockets */ 1162 if ((so->so_options & SO_ACCEPTCONN) == 0) { 1163 error = EINVAL; 1164 goto out; 1165 } 1166 1167 /* removing the filter */ 1168 if (sopt == NULL) { 1169 if (af != NULL) { 1170 if (af->so_accept_filter != NULL && 1171 af->so_accept_filter->accf_destroy != NULL) { 1172 af->so_accept_filter->accf_destroy(so); 1173 } 1174 if (af->so_accept_filter_str != NULL) { 1175 FREE(af->so_accept_filter_str, M_ACCF); 1176 } 1177 FREE(af, M_ACCF); 1178 so->so_accf = NULL; 1179 } 1180 so->so_options &= ~SO_ACCEPTFILTER; 1181 return (0); 1182 } 1183 /* adding a filter */ 1184 /* must remove previous filter first */ 1185 if (af != NULL) { 1186 error = EINVAL; 1187 goto out; 1188 } 1189 /* don't put large objects on the kernel stack */ 1190 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK); 1191 error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap); 1192 afap->af_name[sizeof(afap->af_name)-1] = '\0'; 1193 afap->af_arg[sizeof(afap->af_arg)-1] = '\0'; 1194 if (error) 1195 goto out; 1196 afp = accept_filt_get(afap->af_name); 1197 if (afp == NULL) { 1198 error = ENOENT; 1199 goto out; 1200 } 1201 MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK); 1202 bzero(af, sizeof(*af)); 1203 if (afp->accf_create != NULL) { 1204 if (afap->af_name[0] != '\0') { 1205 int len = strlen(afap->af_name) + 1; 1206 1207 MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK); 1208 strcpy(af->so_accept_filter_str, afap->af_name); 1209 } 1210 af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg); 1211 if (af->so_accept_filter_arg == NULL) { 1212 FREE(af->so_accept_filter_str, M_ACCF); 1213 FREE(af, M_ACCF); 1214 so->so_accf = NULL; 1215 error = EINVAL; 1216 goto out; 1217 } 1218 } 1219 af->so_accept_filter = afp; 1220 so->so_accf = af; 1221 so->so_options |= SO_ACCEPTFILTER; 1222 out: 1223 if (afap != NULL) 1224 FREE(afap, M_TEMP); 1225 return (error); 1226 } 1227 #endif /* INET */ 1228 1229 /* 1230 * Perhaps this routine, and sooptcopyout(), below, ought to come in 1231 * an additional variant to handle the case where the option value needs 1232 * to be some kind of integer, but not a specific size. 1233 * In addition to their use here, these functions are also called by the 1234 * protocol-level pr_ctloutput() routines. 1235 */ 1236 int 1237 sooptcopyin(struct sockopt *sopt, void *buf, size_t len, size_t minlen) 1238 { 1239 size_t valsize; 1240 1241 /* 1242 * If the user gives us more than we wanted, we ignore it, 1243 * but if we don't get the minimum length the caller 1244 * wants, we return EINVAL. On success, sopt->sopt_valsize 1245 * is set to however much we actually retrieved. 1246 */ 1247 if ((valsize = sopt->sopt_valsize) < minlen) 1248 return EINVAL; 1249 if (valsize > len) 1250 sopt->sopt_valsize = valsize = len; 1251 1252 if (sopt->sopt_td != NULL) 1253 return (copyin(sopt->sopt_val, buf, valsize)); 1254 1255 bcopy(sopt->sopt_val, buf, valsize); 1256 return 0; 1257 } 1258 1259 int 1260 sosetopt(struct socket *so, struct sockopt *sopt) 1261 { 1262 int error, optval; 1263 struct linger l; 1264 struct timeval tv; 1265 u_long val; 1266 1267 error = 0; 1268 sopt->sopt_dir = SOPT_SET; 1269 if (sopt->sopt_level != SOL_SOCKET) { 1270 if (so->so_proto && so->so_proto->pr_ctloutput) { 1271 return (so_pr_ctloutput(so, sopt)); 1272 } 1273 error = ENOPROTOOPT; 1274 } else { 1275 switch (sopt->sopt_name) { 1276 #ifdef INET 1277 case SO_ACCEPTFILTER: 1278 error = do_setopt_accept_filter(so, sopt); 1279 if (error) 1280 goto bad; 1281 break; 1282 #endif /* INET */ 1283 case SO_LINGER: 1284 error = sooptcopyin(sopt, &l, sizeof l, sizeof l); 1285 if (error) 1286 goto bad; 1287 1288 so->so_linger = l.l_linger; 1289 if (l.l_onoff) 1290 so->so_options |= SO_LINGER; 1291 else 1292 so->so_options &= ~SO_LINGER; 1293 break; 1294 1295 case SO_DEBUG: 1296 case SO_KEEPALIVE: 1297 case SO_DONTROUTE: 1298 case SO_USELOOPBACK: 1299 case SO_BROADCAST: 1300 case SO_REUSEADDR: 1301 case SO_REUSEPORT: 1302 case SO_OOBINLINE: 1303 case SO_TIMESTAMP: 1304 error = sooptcopyin(sopt, &optval, sizeof optval, 1305 sizeof optval); 1306 if (error) 1307 goto bad; 1308 if (optval) 1309 so->so_options |= sopt->sopt_name; 1310 else 1311 so->so_options &= ~sopt->sopt_name; 1312 break; 1313 1314 case SO_SNDBUF: 1315 case SO_RCVBUF: 1316 case SO_SNDLOWAT: 1317 case SO_RCVLOWAT: 1318 error = sooptcopyin(sopt, &optval, sizeof optval, 1319 sizeof optval); 1320 if (error) 1321 goto bad; 1322 1323 /* 1324 * Values < 1 make no sense for any of these 1325 * options, so disallow them. 1326 */ 1327 if (optval < 1) { 1328 error = EINVAL; 1329 goto bad; 1330 } 1331 1332 switch (sopt->sopt_name) { 1333 case SO_SNDBUF: 1334 case SO_RCVBUF: 1335 if (ssb_reserve(sopt->sopt_name == SO_SNDBUF ? 1336 &so->so_snd : &so->so_rcv, (u_long)optval, 1337 so, 1338 &curproc->p_rlimit[RLIMIT_SBSIZE]) == 0) { 1339 error = ENOBUFS; 1340 goto bad; 1341 } 1342 break; 1343 1344 /* 1345 * Make sure the low-water is never greater than 1346 * the high-water. 1347 */ 1348 case SO_SNDLOWAT: 1349 so->so_snd.ssb_lowat = 1350 (optval > so->so_snd.ssb_hiwat) ? 1351 so->so_snd.ssb_hiwat : optval; 1352 break; 1353 case SO_RCVLOWAT: 1354 so->so_rcv.ssb_lowat = 1355 (optval > so->so_rcv.ssb_hiwat) ? 1356 so->so_rcv.ssb_hiwat : optval; 1357 break; 1358 } 1359 break; 1360 1361 case SO_SNDTIMEO: 1362 case SO_RCVTIMEO: 1363 error = sooptcopyin(sopt, &tv, sizeof tv, 1364 sizeof tv); 1365 if (error) 1366 goto bad; 1367 1368 /* assert(hz > 0); */ 1369 if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz || 1370 tv.tv_usec < 0 || tv.tv_usec >= 1000000) { 1371 error = EDOM; 1372 goto bad; 1373 } 1374 /* assert(tick > 0); */ 1375 /* assert(ULONG_MAX - SHRT_MAX >= 1000000); */ 1376 val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick; 1377 if (val > SHRT_MAX) { 1378 error = EDOM; 1379 goto bad; 1380 } 1381 if (val == 0 && tv.tv_usec != 0) 1382 val = 1; 1383 1384 switch (sopt->sopt_name) { 1385 case SO_SNDTIMEO: 1386 so->so_snd.ssb_timeo = val; 1387 break; 1388 case SO_RCVTIMEO: 1389 so->so_rcv.ssb_timeo = val; 1390 break; 1391 } 1392 break; 1393 default: 1394 error = ENOPROTOOPT; 1395 break; 1396 } 1397 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) { 1398 (void) so_pr_ctloutput(so, sopt); 1399 } 1400 } 1401 bad: 1402 return (error); 1403 } 1404 1405 /* Helper routine for getsockopt */ 1406 int 1407 sooptcopyout(struct sockopt *sopt, const void *buf, size_t len) 1408 { 1409 int error; 1410 size_t valsize; 1411 1412 error = 0; 1413 1414 /* 1415 * Documented get behavior is that we always return a value, 1416 * possibly truncated to fit in the user's buffer. 1417 * Traditional behavior is that we always tell the user 1418 * precisely how much we copied, rather than something useful 1419 * like the total amount we had available for her. 1420 * Note that this interface is not idempotent; the entire answer must 1421 * generated ahead of time. 1422 */ 1423 valsize = min(len, sopt->sopt_valsize); 1424 sopt->sopt_valsize = valsize; 1425 if (sopt->sopt_val != 0) { 1426 if (sopt->sopt_td != NULL) 1427 error = copyout(buf, sopt->sopt_val, valsize); 1428 else 1429 bcopy(buf, sopt->sopt_val, valsize); 1430 } 1431 return error; 1432 } 1433 1434 int 1435 sogetopt(struct socket *so, struct sockopt *sopt) 1436 { 1437 int error, optval; 1438 struct linger l; 1439 struct timeval tv; 1440 #ifdef INET 1441 struct accept_filter_arg *afap; 1442 #endif 1443 1444 error = 0; 1445 sopt->sopt_dir = SOPT_GET; 1446 if (sopt->sopt_level != SOL_SOCKET) { 1447 if (so->so_proto && so->so_proto->pr_ctloutput) { 1448 return (so_pr_ctloutput(so, sopt)); 1449 } else 1450 return (ENOPROTOOPT); 1451 } else { 1452 switch (sopt->sopt_name) { 1453 #ifdef INET 1454 case SO_ACCEPTFILTER: 1455 if ((so->so_options & SO_ACCEPTCONN) == 0) 1456 return (EINVAL); 1457 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), 1458 M_TEMP, M_WAITOK); 1459 bzero(afap, sizeof(*afap)); 1460 if ((so->so_options & SO_ACCEPTFILTER) != 0) { 1461 strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name); 1462 if (so->so_accf->so_accept_filter_str != NULL) 1463 strcpy(afap->af_arg, so->so_accf->so_accept_filter_str); 1464 } 1465 error = sooptcopyout(sopt, afap, sizeof(*afap)); 1466 FREE(afap, M_TEMP); 1467 break; 1468 #endif /* INET */ 1469 1470 case SO_LINGER: 1471 l.l_onoff = so->so_options & SO_LINGER; 1472 l.l_linger = so->so_linger; 1473 error = sooptcopyout(sopt, &l, sizeof l); 1474 break; 1475 1476 case SO_USELOOPBACK: 1477 case SO_DONTROUTE: 1478 case SO_DEBUG: 1479 case SO_KEEPALIVE: 1480 case SO_REUSEADDR: 1481 case SO_REUSEPORT: 1482 case SO_BROADCAST: 1483 case SO_OOBINLINE: 1484 case SO_TIMESTAMP: 1485 optval = so->so_options & sopt->sopt_name; 1486 integer: 1487 error = sooptcopyout(sopt, &optval, sizeof optval); 1488 break; 1489 1490 case SO_TYPE: 1491 optval = so->so_type; 1492 goto integer; 1493 1494 case SO_ERROR: 1495 optval = so->so_error; 1496 so->so_error = 0; 1497 goto integer; 1498 1499 case SO_SNDBUF: 1500 optval = so->so_snd.ssb_hiwat; 1501 goto integer; 1502 1503 case SO_RCVBUF: 1504 optval = so->so_rcv.ssb_hiwat; 1505 goto integer; 1506 1507 case SO_SNDLOWAT: 1508 optval = so->so_snd.ssb_lowat; 1509 goto integer; 1510 1511 case SO_RCVLOWAT: 1512 optval = so->so_rcv.ssb_lowat; 1513 goto integer; 1514 1515 case SO_SNDTIMEO: 1516 case SO_RCVTIMEO: 1517 optval = (sopt->sopt_name == SO_SNDTIMEO ? 1518 so->so_snd.ssb_timeo : so->so_rcv.ssb_timeo); 1519 1520 tv.tv_sec = optval / hz; 1521 tv.tv_usec = (optval % hz) * tick; 1522 error = sooptcopyout(sopt, &tv, sizeof tv); 1523 break; 1524 1525 default: 1526 error = ENOPROTOOPT; 1527 break; 1528 } 1529 return (error); 1530 } 1531 } 1532 1533 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */ 1534 int 1535 soopt_getm(struct sockopt *sopt, struct mbuf **mp) 1536 { 1537 struct mbuf *m, *m_prev; 1538 int sopt_size = sopt->sopt_valsize, msize; 1539 1540 m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_DATA, 1541 0, &msize); 1542 if (m == NULL) 1543 return (ENOBUFS); 1544 m->m_len = min(msize, sopt_size); 1545 sopt_size -= m->m_len; 1546 *mp = m; 1547 m_prev = m; 1548 1549 while (sopt_size > 0) { 1550 m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, 1551 MT_DATA, 0, &msize); 1552 if (m == NULL) { 1553 m_freem(*mp); 1554 return (ENOBUFS); 1555 } 1556 m->m_len = min(msize, sopt_size); 1557 sopt_size -= m->m_len; 1558 m_prev->m_next = m; 1559 m_prev = m; 1560 } 1561 return (0); 1562 } 1563 1564 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */ 1565 int 1566 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m) 1567 { 1568 struct mbuf *m0 = m; 1569 1570 if (sopt->sopt_val == NULL) 1571 return 0; 1572 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1573 if (sopt->sopt_td != NULL) { 1574 int error; 1575 1576 error = copyin(sopt->sopt_val, mtod(m, char *), 1577 m->m_len); 1578 if (error != 0) { 1579 m_freem(m0); 1580 return (error); 1581 } 1582 } else 1583 bcopy(sopt->sopt_val, mtod(m, char *), m->m_len); 1584 sopt->sopt_valsize -= m->m_len; 1585 sopt->sopt_val = (caddr_t)sopt->sopt_val + m->m_len; 1586 m = m->m_next; 1587 } 1588 if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */ 1589 panic("ip6_sooptmcopyin"); 1590 return 0; 1591 } 1592 1593 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */ 1594 int 1595 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m) 1596 { 1597 struct mbuf *m0 = m; 1598 size_t valsize = 0; 1599 1600 if (sopt->sopt_val == NULL) 1601 return 0; 1602 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1603 if (sopt->sopt_td != NULL) { 1604 int error; 1605 1606 error = copyout(mtod(m, char *), sopt->sopt_val, 1607 m->m_len); 1608 if (error != 0) { 1609 m_freem(m0); 1610 return (error); 1611 } 1612 } else 1613 bcopy(mtod(m, char *), sopt->sopt_val, m->m_len); 1614 sopt->sopt_valsize -= m->m_len; 1615 sopt->sopt_val = (caddr_t)sopt->sopt_val + m->m_len; 1616 valsize += m->m_len; 1617 m = m->m_next; 1618 } 1619 if (m != NULL) { 1620 /* enough soopt buffer should be given from user-land */ 1621 m_freem(m0); 1622 return (EINVAL); 1623 } 1624 sopt->sopt_valsize = valsize; 1625 return 0; 1626 } 1627 1628 void 1629 sohasoutofband(struct socket *so) 1630 { 1631 if (so->so_sigio != NULL) 1632 pgsigio(so->so_sigio, SIGURG, 0); 1633 selwakeup(&so->so_rcv.ssb_sel); 1634 } 1635 1636 int 1637 sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td) 1638 { 1639 int revents = 0; 1640 1641 crit_enter(); 1642 1643 if (events & (POLLIN | POLLRDNORM)) 1644 if (soreadable(so)) 1645 revents |= events & (POLLIN | POLLRDNORM); 1646 1647 if (events & POLLINIGNEOF) 1648 if (so->so_rcv.ssb_cc >= so->so_rcv.ssb_lowat || 1649 !TAILQ_EMPTY(&so->so_comp) || so->so_error) 1650 revents |= POLLINIGNEOF; 1651 1652 if (events & (POLLOUT | POLLWRNORM)) 1653 if (sowriteable(so)) 1654 revents |= events & (POLLOUT | POLLWRNORM); 1655 1656 if (events & (POLLPRI | POLLRDBAND)) 1657 if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) 1658 revents |= events & (POLLPRI | POLLRDBAND); 1659 1660 if (revents == 0) { 1661 if (events & 1662 (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM | 1663 POLLRDBAND)) { 1664 selrecord(td, &so->so_rcv.ssb_sel); 1665 so->so_rcv.ssb_flags |= SSB_SEL; 1666 } 1667 1668 if (events & (POLLOUT | POLLWRNORM)) { 1669 selrecord(td, &so->so_snd.ssb_sel); 1670 so->so_snd.ssb_flags |= SSB_SEL; 1671 } 1672 } 1673 1674 crit_exit(); 1675 return (revents); 1676 } 1677 1678 int 1679 sokqfilter(struct file *fp, struct knote *kn) 1680 { 1681 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1682 struct signalsockbuf *ssb; 1683 1684 switch (kn->kn_filter) { 1685 case EVFILT_READ: 1686 if (so->so_options & SO_ACCEPTCONN) 1687 kn->kn_fop = &solisten_filtops; 1688 else 1689 kn->kn_fop = &soread_filtops; 1690 ssb = &so->so_rcv; 1691 break; 1692 case EVFILT_WRITE: 1693 kn->kn_fop = &sowrite_filtops; 1694 ssb = &so->so_snd; 1695 break; 1696 default: 1697 return (1); 1698 } 1699 1700 crit_enter(); 1701 SLIST_INSERT_HEAD(&ssb->ssb_sel.si_note, kn, kn_selnext); 1702 ssb->ssb_flags |= SSB_KNOTE; 1703 crit_exit(); 1704 return (0); 1705 } 1706 1707 static void 1708 filt_sordetach(struct knote *kn) 1709 { 1710 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1711 1712 crit_enter(); 1713 SLIST_REMOVE(&so->so_rcv.ssb_sel.si_note, kn, knote, kn_selnext); 1714 if (SLIST_EMPTY(&so->so_rcv.ssb_sel.si_note)) 1715 so->so_rcv.ssb_flags &= ~SSB_KNOTE; 1716 crit_exit(); 1717 } 1718 1719 /*ARGSUSED*/ 1720 static int 1721 filt_soread(struct knote *kn, long hint) 1722 { 1723 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1724 1725 kn->kn_data = so->so_rcv.ssb_cc; 1726 if (so->so_state & SS_CANTRCVMORE) { 1727 kn->kn_flags |= EV_EOF; 1728 kn->kn_fflags = so->so_error; 1729 return (1); 1730 } 1731 if (so->so_error) /* temporary udp error */ 1732 return (1); 1733 if (kn->kn_sfflags & NOTE_LOWAT) 1734 return (kn->kn_data >= kn->kn_sdata); 1735 return (kn->kn_data >= so->so_rcv.ssb_lowat); 1736 } 1737 1738 static void 1739 filt_sowdetach(struct knote *kn) 1740 { 1741 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1742 1743 crit_enter(); 1744 SLIST_REMOVE(&so->so_snd.ssb_sel.si_note, kn, knote, kn_selnext); 1745 if (SLIST_EMPTY(&so->so_snd.ssb_sel.si_note)) 1746 so->so_snd.ssb_flags &= ~SSB_KNOTE; 1747 crit_exit(); 1748 } 1749 1750 /*ARGSUSED*/ 1751 static int 1752 filt_sowrite(struct knote *kn, long hint) 1753 { 1754 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1755 1756 kn->kn_data = ssb_space(&so->so_snd); 1757 if (so->so_state & SS_CANTSENDMORE) { 1758 kn->kn_flags |= EV_EOF; 1759 kn->kn_fflags = so->so_error; 1760 return (1); 1761 } 1762 if (so->so_error) /* temporary udp error */ 1763 return (1); 1764 if (((so->so_state & SS_ISCONNECTED) == 0) && 1765 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 1766 return (0); 1767 if (kn->kn_sfflags & NOTE_LOWAT) 1768 return (kn->kn_data >= kn->kn_sdata); 1769 return (kn->kn_data >= so->so_snd.ssb_lowat); 1770 } 1771 1772 /*ARGSUSED*/ 1773 static int 1774 filt_solisten(struct knote *kn, long hint) 1775 { 1776 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1777 1778 kn->kn_data = so->so_qlen; 1779 return (! TAILQ_EMPTY(&so->so_comp)); 1780 } 1781