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 1.55 2008/09/02 16:17:52 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 #include <vm/pmap.h> 97 98 #include <sys/thread2.h> 99 #include <sys/socketvar2.h> 100 101 #include <machine/limits.h> 102 103 #ifdef INET 104 static int do_setopt_accept_filter(struct socket *so, struct sockopt *sopt); 105 #endif /* INET */ 106 107 static void filt_sordetach(struct knote *kn); 108 static int filt_soread(struct knote *kn, long hint); 109 static void filt_sowdetach(struct knote *kn); 110 static int filt_sowrite(struct knote *kn, long hint); 111 static int filt_solisten(struct knote *kn, long hint); 112 113 static struct filterops solisten_filtops = 114 { 1, NULL, filt_sordetach, filt_solisten }; 115 static struct filterops soread_filtops = 116 { 1, NULL, filt_sordetach, filt_soread }; 117 static struct filterops sowrite_filtops = 118 { 1, NULL, filt_sowdetach, filt_sowrite }; 119 120 MALLOC_DEFINE(M_SOCKET, "socket", "socket struct"); 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, and initialize it. 139 * Note that it would probably be better to allocate socket 140 * and PCB at the same time, but I'm not convinced that all 141 * the protocols can be easily modified to do this. 142 */ 143 struct socket * 144 soalloc(int waitok) 145 { 146 struct socket *so; 147 unsigned waitmask; 148 149 waitmask = waitok ? M_WAITOK : M_NOWAIT; 150 so = kmalloc(sizeof(struct socket), M_SOCKET, M_ZERO|waitmask); 151 if (so) { 152 /* XXX race condition for reentrant kernel */ 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 kfree(so, M_SOCKET); 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 /* 282 * Destroy a disconnected socket. This routine is a NOP if entities 283 * still have a reference on the socket: 284 * 285 * so_pcb - The protocol stack still has a reference 286 * SS_NOFDREF - There is no longer a file pointer reference 287 * SS_ABORTING - An abort netmsg is in-flight 288 */ 289 void 290 sofree(struct socket *so) 291 { 292 struct socket *head = so->so_head; 293 294 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) 295 return; 296 if (so->so_state & SS_ABORTING) 297 return; 298 if (head != NULL) { 299 if (so->so_state & SS_INCOMP) { 300 TAILQ_REMOVE(&head->so_incomp, so, so_list); 301 head->so_incqlen--; 302 } else if (so->so_state & SS_COMP) { 303 /* 304 * We must not decommission a socket that's 305 * on the accept(2) queue. If we do, then 306 * accept(2) may hang after select(2) indicated 307 * that the listening socket was ready. 308 */ 309 return; 310 } else { 311 panic("sofree: not queued"); 312 } 313 so->so_state &= ~SS_INCOMP; 314 so->so_head = NULL; 315 } 316 ssb_release(&so->so_snd, so); 317 sorflush(so); 318 sodealloc(so); 319 } 320 321 /* 322 * Close a socket on last file table reference removal. 323 * Initiate disconnect if connected. 324 * Free socket when disconnect complete. 325 */ 326 int 327 soclose(struct socket *so, int fflag) 328 { 329 int error = 0; 330 331 crit_enter(); 332 funsetown(so->so_sigio); 333 if (so->so_pcb == NULL) 334 goto discard; 335 if (so->so_state & SS_ISCONNECTED) { 336 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 337 error = sodisconnect(so); 338 if (error) 339 goto drop; 340 } 341 if (so->so_options & SO_LINGER) { 342 if ((so->so_state & SS_ISDISCONNECTING) && 343 (fflag & FNONBLOCK)) 344 goto drop; 345 while (so->so_state & SS_ISCONNECTED) { 346 error = tsleep((caddr_t)&so->so_timeo, 347 PCATCH, "soclos", so->so_linger * hz); 348 if (error) 349 break; 350 } 351 } 352 } 353 drop: 354 if (so->so_pcb) { 355 int error2; 356 357 error2 = so_pru_detach(so); 358 if (error == 0) 359 error = error2; 360 } 361 discard: 362 if (so->so_options & SO_ACCEPTCONN) { 363 struct socket *sp; 364 365 while ((sp = TAILQ_FIRST(&so->so_incomp)) != NULL) { 366 TAILQ_REMOVE(&so->so_incomp, sp, so_list); 367 sp->so_state &= ~SS_INCOMP; 368 sp->so_head = NULL; 369 so->so_incqlen--; 370 soaborta(sp); 371 } 372 while ((sp = TAILQ_FIRST(&so->so_comp)) != NULL) { 373 TAILQ_REMOVE(&so->so_comp, sp, so_list); 374 sp->so_state &= ~SS_COMP; 375 sp->so_head = NULL; 376 so->so_qlen--; 377 soaborta(sp); 378 } 379 } 380 if (so->so_state & SS_NOFDREF) 381 panic("soclose: NOFDREF"); 382 so->so_state |= SS_NOFDREF; 383 sofree(so); 384 crit_exit(); 385 return (error); 386 } 387 388 /* 389 * Abort and destroy a socket. Only one abort can be in progress 390 * at any given moment. 391 */ 392 void 393 soabort(struct socket *so) 394 { 395 if ((so->so_state & SS_ABORTING) == 0) { 396 so->so_state |= SS_ABORTING; 397 so_pru_abort(so); 398 } 399 } 400 401 void 402 soaborta(struct socket *so) 403 { 404 if ((so->so_state & SS_ABORTING) == 0) { 405 so->so_state |= SS_ABORTING; 406 so_pru_aborta(so); 407 } 408 } 409 410 void 411 soabort_oncpu(struct socket *so) 412 { 413 if ((so->so_state & SS_ABORTING) == 0) { 414 so->so_state |= SS_ABORTING; 415 so_pru_abort_oncpu(so); 416 } 417 } 418 419 int 420 soaccept(struct socket *so, struct sockaddr **nam) 421 { 422 int error; 423 424 crit_enter(); 425 if ((so->so_state & SS_NOFDREF) == 0) 426 panic("soaccept: !NOFDREF"); 427 so->so_state &= ~SS_NOFDREF; 428 error = so_pru_accept(so, nam); 429 crit_exit(); 430 return (error); 431 } 432 433 int 434 soconnect(struct socket *so, struct sockaddr *nam, struct thread *td) 435 { 436 int error; 437 438 if (so->so_options & SO_ACCEPTCONN) 439 return (EOPNOTSUPP); 440 crit_enter(); 441 /* 442 * If protocol is connection-based, can only connect once. 443 * Otherwise, if connected, try to disconnect first. 444 * This allows user to disconnect by connecting to, e.g., 445 * a null address. 446 */ 447 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 448 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 449 (error = sodisconnect(so)))) { 450 error = EISCONN; 451 } else { 452 /* 453 * Prevent accumulated error from previous connection 454 * from biting us. 455 */ 456 so->so_error = 0; 457 error = so_pru_connect(so, nam, td); 458 } 459 crit_exit(); 460 return (error); 461 } 462 463 int 464 soconnect2(struct socket *so1, struct socket *so2) 465 { 466 int error; 467 468 crit_enter(); 469 error = so_pru_connect2(so1, so2); 470 crit_exit(); 471 return (error); 472 } 473 474 int 475 sodisconnect(struct socket *so) 476 { 477 int error; 478 479 crit_enter(); 480 if ((so->so_state & SS_ISCONNECTED) == 0) { 481 error = ENOTCONN; 482 goto bad; 483 } 484 if (so->so_state & SS_ISDISCONNECTING) { 485 error = EALREADY; 486 goto bad; 487 } 488 error = so_pru_disconnect(so); 489 bad: 490 crit_exit(); 491 return (error); 492 } 493 494 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 495 /* 496 * Send on a socket. 497 * If send must go all at once and message is larger than 498 * send buffering, then hard error. 499 * Lock against other senders. 500 * If must go all at once and not enough room now, then 501 * inform user that this would block and do nothing. 502 * Otherwise, if nonblocking, send as much as possible. 503 * The data to be sent is described by "uio" if nonzero, 504 * otherwise by the mbuf chain "top" (which must be null 505 * if uio is not). Data provided in mbuf chain must be small 506 * enough to send all at once. 507 * 508 * Returns nonzero on error, timeout or signal; callers 509 * must check for short counts if EINTR/ERESTART are returned. 510 * Data and control buffers are freed on return. 511 */ 512 int 513 sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, 514 struct mbuf *top, struct mbuf *control, int flags, 515 struct thread *td) 516 { 517 struct mbuf **mp; 518 struct mbuf *m; 519 long space, len, resid; 520 int clen = 0, error, dontroute, mlen; 521 int atomic = sosendallatonce(so) || top; 522 int pru_flags; 523 524 if (uio) 525 resid = uio->uio_resid; 526 else 527 resid = top->m_pkthdr.len; 528 /* 529 * In theory resid should be unsigned. 530 * However, space must be signed, as it might be less than 0 531 * if we over-committed, and we must use a signed comparison 532 * of space and resid. On the other hand, a negative resid 533 * causes us to loop sending 0-length segments to the protocol. 534 * 535 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM 536 * type sockets since that's an error. 537 */ 538 if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) { 539 error = EINVAL; 540 goto out; 541 } 542 543 dontroute = 544 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 545 (so->so_proto->pr_flags & PR_ATOMIC); 546 if (td->td_lwp != NULL) 547 td->td_lwp->lwp_ru.ru_msgsnd++; 548 if (control) 549 clen = control->m_len; 550 #define gotoerr(errcode) { error = errcode; crit_exit(); goto release; } 551 552 restart: 553 error = ssb_lock(&so->so_snd, SBLOCKWAIT(flags)); 554 if (error) 555 goto out; 556 do { 557 crit_enter(); 558 if (so->so_state & SS_CANTSENDMORE) 559 gotoerr(EPIPE); 560 if (so->so_error) { 561 error = so->so_error; 562 so->so_error = 0; 563 crit_exit(); 564 goto release; 565 } 566 if ((so->so_state & SS_ISCONNECTED) == 0) { 567 /* 568 * `sendto' and `sendmsg' is allowed on a connection- 569 * based socket if it supports implied connect. 570 * Return ENOTCONN if not connected and no address is 571 * supplied. 572 */ 573 if ((so->so_proto->pr_flags & PR_CONNREQUIRED) && 574 (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) { 575 if ((so->so_state & SS_ISCONFIRMING) == 0 && 576 !(resid == 0 && clen != 0)) 577 gotoerr(ENOTCONN); 578 } else if (addr == 0) 579 gotoerr(so->so_proto->pr_flags & PR_CONNREQUIRED ? 580 ENOTCONN : EDESTADDRREQ); 581 } 582 if ((atomic && resid > so->so_snd.ssb_hiwat) || 583 clen > so->so_snd.ssb_hiwat) { 584 gotoerr(EMSGSIZE); 585 } 586 space = ssb_space(&so->so_snd); 587 if (flags & MSG_OOB) 588 space += 1024; 589 if (space < resid + clen && uio && 590 (atomic || space < so->so_snd.ssb_lowat || space < clen)) { 591 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) 592 gotoerr(EWOULDBLOCK); 593 ssb_unlock(&so->so_snd); 594 error = ssb_wait(&so->so_snd); 595 crit_exit(); 596 if (error) 597 goto out; 598 goto restart; 599 } 600 crit_exit(); 601 mp = ⊤ 602 space -= clen; 603 do { 604 if (uio == NULL) { 605 /* 606 * Data is prepackaged in "top". 607 */ 608 resid = 0; 609 if (flags & MSG_EOR) 610 top->m_flags |= M_EOR; 611 } else do { 612 m = m_getl(resid, MB_WAIT, MT_DATA, 613 top == NULL ? M_PKTHDR : 0, &mlen); 614 if (top == NULL) { 615 m->m_pkthdr.len = 0; 616 m->m_pkthdr.rcvif = NULL; 617 } 618 len = min(min(mlen, resid), space); 619 if (resid < MINCLSIZE) { 620 /* 621 * For datagram protocols, leave room 622 * for protocol headers in first mbuf. 623 */ 624 if (atomic && top == 0 && len < mlen) 625 MH_ALIGN(m, len); 626 } 627 space -= len; 628 error = uiomove(mtod(m, caddr_t), (int)len, uio); 629 resid = uio->uio_resid; 630 m->m_len = len; 631 *mp = m; 632 top->m_pkthdr.len += len; 633 if (error) 634 goto release; 635 mp = &m->m_next; 636 if (resid <= 0) { 637 if (flags & MSG_EOR) 638 top->m_flags |= M_EOR; 639 break; 640 } 641 } while (space > 0 && atomic); 642 if (dontroute) 643 so->so_options |= SO_DONTROUTE; 644 if (flags & MSG_OOB) { 645 pru_flags = PRUS_OOB; 646 } else if ((flags & MSG_EOF) && 647 (so->so_proto->pr_flags & PR_IMPLOPCL) && 648 (resid <= 0)) { 649 /* 650 * If the user set MSG_EOF, the protocol 651 * understands this flag and nothing left to 652 * send then use PRU_SEND_EOF instead of PRU_SEND. 653 */ 654 pru_flags = PRUS_EOF; 655 } else if (resid > 0 && space > 0) { 656 /* If there is more to send, set PRUS_MORETOCOME */ 657 pru_flags = PRUS_MORETOCOME; 658 } else { 659 pru_flags = 0; 660 } 661 crit_enter(); 662 /* 663 * XXX all the SS_CANTSENDMORE checks previously 664 * done could be out of date. We could have recieved 665 * a reset packet in an interrupt or maybe we slept 666 * while doing page faults in uiomove() etc. We could 667 * probably recheck again inside the splnet() protection 668 * here, but there are probably other places that this 669 * also happens. We must rethink this. 670 */ 671 error = so_pru_send(so, pru_flags, top, addr, control, td); 672 crit_exit(); 673 if (dontroute) 674 so->so_options &= ~SO_DONTROUTE; 675 clen = 0; 676 control = 0; 677 top = 0; 678 mp = ⊤ 679 if (error) 680 goto release; 681 } while (resid && space > 0); 682 } while (resid); 683 684 release: 685 ssb_unlock(&so->so_snd); 686 out: 687 if (top) 688 m_freem(top); 689 if (control) 690 m_freem(control); 691 return (error); 692 } 693 694 /* 695 * A specialization of sosend() for UDP based on protocol-specific knowledge: 696 * so->so_proto->pr_flags has the PR_ATOMIC field set. This means that 697 * sosendallatonce() returns true, 698 * the "atomic" variable is true, 699 * and sosendudp() blocks until space is available for the entire send. 700 * so->so_proto->pr_flags does not have the PR_CONNREQUIRED or 701 * PR_IMPLOPCL flags set. 702 * UDP has no out-of-band data. 703 * UDP has no control data. 704 * UDP does not support MSG_EOR. 705 */ 706 int 707 sosendudp(struct socket *so, struct sockaddr *addr, struct uio *uio, 708 struct mbuf *top, struct mbuf *control, int flags, struct thread *td) 709 { 710 int resid, error; 711 boolean_t dontroute; /* temporary SO_DONTROUTE setting */ 712 713 if (td->td_lwp != NULL) 714 td->td_lwp->lwp_ru.ru_msgsnd++; 715 if (control) 716 m_freem(control); 717 718 KASSERT((uio && !top) || (top && !uio), ("bad arguments to sosendudp")); 719 resid = uio ? uio->uio_resid : top->m_pkthdr.len; 720 721 restart: 722 error = ssb_lock(&so->so_snd, SBLOCKWAIT(flags)); 723 if (error) 724 goto out; 725 726 crit_enter(); 727 if (so->so_state & SS_CANTSENDMORE) 728 gotoerr(EPIPE); 729 if (so->so_error) { 730 error = so->so_error; 731 so->so_error = 0; 732 crit_exit(); 733 goto release; 734 } 735 if (!(so->so_state & SS_ISCONNECTED) && addr == NULL) 736 gotoerr(EDESTADDRREQ); 737 if (resid > so->so_snd.ssb_hiwat) 738 gotoerr(EMSGSIZE); 739 if (uio && ssb_space(&so->so_snd) < resid) { 740 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) 741 gotoerr(EWOULDBLOCK); 742 ssb_unlock(&so->so_snd); 743 error = ssb_wait(&so->so_snd); 744 crit_exit(); 745 if (error) 746 goto out; 747 goto restart; 748 } 749 crit_exit(); 750 751 if (uio) { 752 top = m_uiomove(uio); 753 if (top == NULL) 754 goto release; 755 } 756 757 dontroute = (flags & MSG_DONTROUTE) && !(so->so_options & SO_DONTROUTE); 758 if (dontroute) 759 so->so_options |= SO_DONTROUTE; 760 761 error = so_pru_send(so, 0, top, addr, NULL, td); 762 top = NULL; /* sent or freed in lower layer */ 763 764 if (dontroute) 765 so->so_options &= ~SO_DONTROUTE; 766 767 release: 768 ssb_unlock(&so->so_snd); 769 out: 770 if (top) 771 m_freem(top); 772 return (error); 773 } 774 775 /* 776 * Implement receive operations on a socket. 777 * We depend on the way that records are added to the signalsockbuf 778 * by sbappend*. In particular, each record (mbufs linked through m_next) 779 * must begin with an address if the protocol so specifies, 780 * followed by an optional mbuf or mbufs containing ancillary data, 781 * and then zero or more mbufs of data. 782 * In order to avoid blocking network interrupts for the entire time here, 783 * we exit the critical section while doing the actual copy to user space. 784 * Although the signalsockbuf is locked, new data may still be appended, 785 * and thus we must maintain consistency of the signalsockbuf during that time. 786 * 787 * The caller may receive the data as a single mbuf chain by supplying 788 * an mbuf **mp0 for use in returning the chain. The uio is then used 789 * only for the count in uio_resid. 790 */ 791 int 792 soreceive(struct socket *so, struct sockaddr **psa, struct uio *uio, 793 struct sockbuf *sio, struct mbuf **controlp, int *flagsp) 794 { 795 struct mbuf *m, *n; 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 resid, orig_resid; 801 802 if (uio) 803 resid = uio->uio_resid; 804 else 805 resid = (int)(sio->sb_climit - sio->sb_cc); 806 orig_resid = resid; 807 808 if (psa) 809 *psa = NULL; 810 if (controlp) 811 *controlp = NULL; 812 if (flagsp) 813 flags = *flagsp &~ MSG_EOR; 814 else 815 flags = 0; 816 if (flags & MSG_OOB) { 817 m = m_get(MB_WAIT, MT_DATA); 818 if (m == NULL) 819 return (ENOBUFS); 820 error = so_pru_rcvoob(so, m, flags & MSG_PEEK); 821 if (error) 822 goto bad; 823 if (sio) { 824 do { 825 sbappend(sio, m); 826 resid -= m->m_len; 827 } while (resid > 0 && m); 828 } else { 829 do { 830 uio->uio_resid = resid; 831 error = uiomove(mtod(m, caddr_t), 832 (int)min(resid, m->m_len), uio); 833 resid = uio->uio_resid; 834 m = m_free(m); 835 } while (uio->uio_resid && error == 0 && m); 836 } 837 bad: 838 if (m) 839 m_freem(m); 840 return (error); 841 } 842 if (so->so_state & SS_ISCONFIRMING && resid) 843 so_pru_rcvd(so, 0); 844 845 restart: 846 crit_enter(); 847 error = ssb_lock(&so->so_rcv, SBLOCKWAIT(flags)); 848 if (error) 849 goto done; 850 851 m = so->so_rcv.ssb_mb; 852 /* 853 * If we have less data than requested, block awaiting more 854 * (subject to any timeout) if: 855 * 1. the current count is less than the low water mark, or 856 * 2. MSG_WAITALL is set, and it is possible to do the entire 857 * receive operation at once if we block (resid <= hiwat). 858 * 3. MSG_DONTWAIT is not set 859 * If MSG_WAITALL is set but resid is larger than the receive buffer, 860 * we have to do the receive in sections, and thus risk returning 861 * a short count if a timeout or signal occurs after we start. 862 */ 863 if (m == NULL || (((flags & MSG_DONTWAIT) == 0 && 864 so->so_rcv.ssb_cc < resid) && 865 (so->so_rcv.ssb_cc < so->so_rcv.ssb_lowat || 866 ((flags & MSG_WAITALL) && resid <= so->so_rcv.ssb_hiwat)) && 867 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) { 868 KASSERT(m != NULL || !so->so_rcv.ssb_cc, ("receive 1")); 869 if (so->so_error) { 870 if (m) 871 goto dontblock; 872 error = so->so_error; 873 if ((flags & MSG_PEEK) == 0) 874 so->so_error = 0; 875 goto release; 876 } 877 if (so->so_state & SS_CANTRCVMORE) { 878 if (m) 879 goto dontblock; 880 else 881 goto release; 882 } 883 for (; m; m = m->m_next) { 884 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 885 m = so->so_rcv.ssb_mb; 886 goto dontblock; 887 } 888 } 889 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 890 (pr->pr_flags & PR_CONNREQUIRED)) { 891 error = ENOTCONN; 892 goto release; 893 } 894 if (resid == 0) 895 goto release; 896 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) { 897 error = EWOULDBLOCK; 898 goto release; 899 } 900 ssb_unlock(&so->so_rcv); 901 error = ssb_wait(&so->so_rcv); 902 if (error) 903 goto done; 904 crit_exit(); 905 goto restart; 906 } 907 dontblock: 908 if (uio && uio->uio_td && uio->uio_td->td_proc) 909 uio->uio_td->td_lwp->lwp_ru.ru_msgrcv++; 910 911 /* 912 * note: m should be == sb_mb here. Cache the next record while 913 * cleaning up. Note that calling m_free*() will break out critical 914 * section. 915 */ 916 KKASSERT(m == so->so_rcv.ssb_mb); 917 918 /* 919 * Skip any address mbufs prepending the record. 920 */ 921 if (pr->pr_flags & PR_ADDR) { 922 KASSERT(m->m_type == MT_SONAME, ("receive 1a")); 923 orig_resid = 0; 924 if (psa) 925 *psa = dup_sockaddr(mtod(m, struct sockaddr *)); 926 if (flags & MSG_PEEK) 927 m = m->m_next; 928 else 929 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain); 930 } 931 932 /* 933 * Skip any control mbufs prepending the record. 934 */ 935 #ifdef SCTP 936 if (pr->pr_flags & PR_ADDR_OPT) { 937 /* 938 * For SCTP we may be getting a 939 * whole message OR a partial delivery. 940 */ 941 if (m && m->m_type == MT_SONAME) { 942 orig_resid = 0; 943 if (psa) 944 *psa = dup_sockaddr(mtod(m, struct sockaddr *)); 945 if (flags & MSG_PEEK) 946 m = m->m_next; 947 else 948 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain); 949 } 950 } 951 #endif /* SCTP */ 952 while (m && m->m_type == MT_CONTROL && error == 0) { 953 if (flags & MSG_PEEK) { 954 if (controlp) 955 *controlp = m_copy(m, 0, m->m_len); 956 m = m->m_next; /* XXX race */ 957 } else { 958 if (controlp) { 959 n = sbunlinkmbuf(&so->so_rcv.sb, m, NULL); 960 if (pr->pr_domain->dom_externalize && 961 mtod(m, struct cmsghdr *)->cmsg_type == 962 SCM_RIGHTS) 963 error = (*pr->pr_domain->dom_externalize)(m); 964 *controlp = m; 965 m = n; 966 } else { 967 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain); 968 } 969 } 970 if (controlp && *controlp) { 971 orig_resid = 0; 972 controlp = &(*controlp)->m_next; 973 } 974 } 975 976 /* 977 * flag OOB data. 978 */ 979 if (m) { 980 type = m->m_type; 981 if (type == MT_OOBDATA) 982 flags |= MSG_OOB; 983 } 984 985 /* 986 * Copy to the UIO or mbuf return chain (*mp). 987 */ 988 moff = 0; 989 offset = 0; 990 while (m && resid > 0 && error == 0) { 991 if (m->m_type == MT_OOBDATA) { 992 if (type != MT_OOBDATA) 993 break; 994 } else if (type == MT_OOBDATA) 995 break; 996 else 997 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER, 998 ("receive 3")); 999 so->so_state &= ~SS_RCVATMARK; 1000 len = resid; 1001 if (so->so_oobmark && len > so->so_oobmark - offset) 1002 len = so->so_oobmark - offset; 1003 if (len > m->m_len - moff) 1004 len = m->m_len - moff; 1005 1006 /* 1007 * Copy out to the UIO or pass the mbufs back to the SIO. 1008 * The SIO is dealt with when we eat the mbuf, but deal 1009 * with the resid here either way. 1010 */ 1011 if (uio) { 1012 crit_exit(); 1013 uio->uio_resid = resid; 1014 error = uiomove(mtod(m, caddr_t) + moff, len, uio); 1015 resid = uio->uio_resid; 1016 crit_enter(); 1017 if (error) 1018 goto release; 1019 } else { 1020 resid -= len; 1021 } 1022 1023 /* 1024 * Eat the entire mbuf or just a piece of it 1025 */ 1026 if (len == m->m_len - moff) { 1027 if (m->m_flags & M_EOR) 1028 flags |= MSG_EOR; 1029 #ifdef SCTP 1030 if (m->m_flags & M_NOTIFICATION) 1031 flags |= MSG_NOTIFICATION; 1032 #endif /* SCTP */ 1033 if (flags & MSG_PEEK) { 1034 m = m->m_next; 1035 moff = 0; 1036 } else { 1037 if (sio) { 1038 n = sbunlinkmbuf(&so->so_rcv.sb, m, NULL); 1039 sbappend(sio, m); 1040 m = n; 1041 } else { 1042 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain); 1043 } 1044 } 1045 } else { 1046 if (flags & MSG_PEEK) { 1047 moff += len; 1048 } else { 1049 if (sio) { 1050 n = m_copym(m, 0, len, MB_WAIT); 1051 if (n) 1052 sbappend(sio, n); 1053 } 1054 m->m_data += len; 1055 m->m_len -= len; 1056 so->so_rcv.ssb_cc -= len; 1057 } 1058 } 1059 if (so->so_oobmark) { 1060 if ((flags & MSG_PEEK) == 0) { 1061 so->so_oobmark -= len; 1062 if (so->so_oobmark == 0) { 1063 so->so_state |= SS_RCVATMARK; 1064 break; 1065 } 1066 } else { 1067 offset += len; 1068 if (offset == so->so_oobmark) 1069 break; 1070 } 1071 } 1072 if (flags & MSG_EOR) 1073 break; 1074 /* 1075 * If the MSG_WAITALL flag is set (for non-atomic socket), 1076 * we must not quit until resid == 0 or an error 1077 * termination. If a signal/timeout occurs, return 1078 * with a short count but without error. 1079 * Keep signalsockbuf locked against other readers. 1080 */ 1081 while ((flags & MSG_WAITALL) && m == NULL && 1082 resid > 0 && !sosendallatonce(so) && 1083 so->so_rcv.ssb_mb == NULL) { 1084 if (so->so_error || so->so_state & SS_CANTRCVMORE) 1085 break; 1086 /* 1087 * The window might have closed to zero, make 1088 * sure we send an ack now that we've drained 1089 * the buffer or we might end up blocking until 1090 * the idle takes over (5 seconds). 1091 */ 1092 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1093 so_pru_rcvd(so, flags); 1094 error = ssb_wait(&so->so_rcv); 1095 if (error) { 1096 ssb_unlock(&so->so_rcv); 1097 error = 0; 1098 goto done; 1099 } 1100 m = so->so_rcv.ssb_mb; 1101 } 1102 } 1103 1104 /* 1105 * If an atomic read was requested but unread data still remains 1106 * in the record, set MSG_TRUNC. 1107 */ 1108 if (m && pr->pr_flags & PR_ATOMIC) 1109 flags |= MSG_TRUNC; 1110 1111 /* 1112 * Cleanup. If an atomic read was requested drop any unread data. 1113 */ 1114 if ((flags & MSG_PEEK) == 0) { 1115 if (m && (pr->pr_flags & PR_ATOMIC)) 1116 sbdroprecord(&so->so_rcv.sb); 1117 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) 1118 so_pru_rcvd(so, flags); 1119 } 1120 1121 if (orig_resid == resid && orig_resid && 1122 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 1123 ssb_unlock(&so->so_rcv); 1124 crit_exit(); 1125 goto restart; 1126 } 1127 1128 if (flagsp) 1129 *flagsp |= flags; 1130 release: 1131 ssb_unlock(&so->so_rcv); 1132 done: 1133 crit_exit(); 1134 if (free_chain) 1135 m_freem(free_chain); 1136 return (error); 1137 } 1138 1139 int 1140 soshutdown(struct socket *so, int how) 1141 { 1142 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 1143 return (EINVAL); 1144 1145 if (how != SHUT_WR) 1146 sorflush(so); 1147 if (how != SHUT_RD) 1148 return (so_pru_shutdown(so)); 1149 return (0); 1150 } 1151 1152 void 1153 sorflush(struct socket *so) 1154 { 1155 struct signalsockbuf *ssb = &so->so_rcv; 1156 struct protosw *pr = so->so_proto; 1157 struct signalsockbuf asb; 1158 1159 ssb->ssb_flags |= SSB_NOINTR; 1160 (void) ssb_lock(ssb, M_WAITOK); 1161 1162 crit_enter(); 1163 socantrcvmore(so); 1164 ssb_unlock(ssb); 1165 asb = *ssb; 1166 bzero((caddr_t)ssb, sizeof (*ssb)); 1167 if (asb.ssb_flags & SSB_KNOTE) { 1168 ssb->ssb_sel.si_note = asb.ssb_sel.si_note; 1169 ssb->ssb_flags = SSB_KNOTE; 1170 } 1171 crit_exit(); 1172 1173 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) 1174 (*pr->pr_domain->dom_dispose)(asb.ssb_mb); 1175 ssb_release(&asb, so); 1176 } 1177 1178 #ifdef INET 1179 static int 1180 do_setopt_accept_filter(struct socket *so, struct sockopt *sopt) 1181 { 1182 struct accept_filter_arg *afap = NULL; 1183 struct accept_filter *afp; 1184 struct so_accf *af = so->so_accf; 1185 int error = 0; 1186 1187 /* do not set/remove accept filters on non listen sockets */ 1188 if ((so->so_options & SO_ACCEPTCONN) == 0) { 1189 error = EINVAL; 1190 goto out; 1191 } 1192 1193 /* removing the filter */ 1194 if (sopt == NULL) { 1195 if (af != NULL) { 1196 if (af->so_accept_filter != NULL && 1197 af->so_accept_filter->accf_destroy != NULL) { 1198 af->so_accept_filter->accf_destroy(so); 1199 } 1200 if (af->so_accept_filter_str != NULL) { 1201 FREE(af->so_accept_filter_str, M_ACCF); 1202 } 1203 FREE(af, M_ACCF); 1204 so->so_accf = NULL; 1205 } 1206 so->so_options &= ~SO_ACCEPTFILTER; 1207 return (0); 1208 } 1209 /* adding a filter */ 1210 /* must remove previous filter first */ 1211 if (af != NULL) { 1212 error = EINVAL; 1213 goto out; 1214 } 1215 /* don't put large objects on the kernel stack */ 1216 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK); 1217 error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap); 1218 afap->af_name[sizeof(afap->af_name)-1] = '\0'; 1219 afap->af_arg[sizeof(afap->af_arg)-1] = '\0'; 1220 if (error) 1221 goto out; 1222 afp = accept_filt_get(afap->af_name); 1223 if (afp == NULL) { 1224 error = ENOENT; 1225 goto out; 1226 } 1227 MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK | M_ZERO); 1228 if (afp->accf_create != NULL) { 1229 if (afap->af_name[0] != '\0') { 1230 int len = strlen(afap->af_name) + 1; 1231 1232 MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK); 1233 strcpy(af->so_accept_filter_str, afap->af_name); 1234 } 1235 af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg); 1236 if (af->so_accept_filter_arg == NULL) { 1237 FREE(af->so_accept_filter_str, M_ACCF); 1238 FREE(af, M_ACCF); 1239 so->so_accf = NULL; 1240 error = EINVAL; 1241 goto out; 1242 } 1243 } 1244 af->so_accept_filter = afp; 1245 so->so_accf = af; 1246 so->so_options |= SO_ACCEPTFILTER; 1247 out: 1248 if (afap != NULL) 1249 FREE(afap, M_TEMP); 1250 return (error); 1251 } 1252 #endif /* INET */ 1253 1254 /* 1255 * Perhaps this routine, and sooptcopyout(), below, ought to come in 1256 * an additional variant to handle the case where the option value needs 1257 * to be some kind of integer, but not a specific size. 1258 * In addition to their use here, these functions are also called by the 1259 * protocol-level pr_ctloutput() routines. 1260 */ 1261 int 1262 sooptcopyin(struct sockopt *sopt, void *buf, size_t len, size_t minlen) 1263 { 1264 return soopt_to_kbuf(sopt, buf, len, minlen); 1265 } 1266 1267 int 1268 soopt_to_kbuf(struct sockopt *sopt, void *buf, size_t len, size_t minlen) 1269 { 1270 size_t valsize; 1271 1272 KKASSERT(!sopt->sopt_val || kva_p(sopt->sopt_val)); 1273 KKASSERT(kva_p(buf)); 1274 1275 /* 1276 * If the user gives us more than we wanted, we ignore it, 1277 * but if we don't get the minimum length the caller 1278 * wants, we return EINVAL. On success, sopt->sopt_valsize 1279 * is set to however much we actually retrieved. 1280 */ 1281 if ((valsize = sopt->sopt_valsize) < minlen) 1282 return EINVAL; 1283 if (valsize > len) 1284 sopt->sopt_valsize = valsize = len; 1285 1286 bcopy(sopt->sopt_val, buf, valsize); 1287 return 0; 1288 } 1289 1290 1291 int 1292 sosetopt(struct socket *so, struct sockopt *sopt) 1293 { 1294 int error, optval; 1295 struct linger l; 1296 struct timeval tv; 1297 u_long val; 1298 1299 error = 0; 1300 sopt->sopt_dir = SOPT_SET; 1301 if (sopt->sopt_level != SOL_SOCKET) { 1302 if (so->so_proto && so->so_proto->pr_ctloutput) { 1303 return (so_pru_ctloutput(so, sopt)); 1304 } 1305 error = ENOPROTOOPT; 1306 } else { 1307 switch (sopt->sopt_name) { 1308 #ifdef INET 1309 case SO_ACCEPTFILTER: 1310 error = do_setopt_accept_filter(so, sopt); 1311 if (error) 1312 goto bad; 1313 break; 1314 #endif /* INET */ 1315 case SO_LINGER: 1316 error = sooptcopyin(sopt, &l, sizeof l, sizeof l); 1317 if (error) 1318 goto bad; 1319 1320 so->so_linger = l.l_linger; 1321 if (l.l_onoff) 1322 so->so_options |= SO_LINGER; 1323 else 1324 so->so_options &= ~SO_LINGER; 1325 break; 1326 1327 case SO_DEBUG: 1328 case SO_KEEPALIVE: 1329 case SO_DONTROUTE: 1330 case SO_USELOOPBACK: 1331 case SO_BROADCAST: 1332 case SO_REUSEADDR: 1333 case SO_REUSEPORT: 1334 case SO_OOBINLINE: 1335 case SO_TIMESTAMP: 1336 error = sooptcopyin(sopt, &optval, sizeof optval, 1337 sizeof optval); 1338 if (error) 1339 goto bad; 1340 if (optval) 1341 so->so_options |= sopt->sopt_name; 1342 else 1343 so->so_options &= ~sopt->sopt_name; 1344 break; 1345 1346 case SO_SNDBUF: 1347 case SO_RCVBUF: 1348 case SO_SNDLOWAT: 1349 case SO_RCVLOWAT: 1350 error = sooptcopyin(sopt, &optval, sizeof optval, 1351 sizeof optval); 1352 if (error) 1353 goto bad; 1354 1355 /* 1356 * Values < 1 make no sense for any of these 1357 * options, so disallow them. 1358 */ 1359 if (optval < 1) { 1360 error = EINVAL; 1361 goto bad; 1362 } 1363 1364 switch (sopt->sopt_name) { 1365 case SO_SNDBUF: 1366 case SO_RCVBUF: 1367 if (ssb_reserve(sopt->sopt_name == SO_SNDBUF ? 1368 &so->so_snd : &so->so_rcv, (u_long)optval, 1369 so, 1370 &curproc->p_rlimit[RLIMIT_SBSIZE]) == 0) { 1371 error = ENOBUFS; 1372 goto bad; 1373 } 1374 break; 1375 1376 /* 1377 * Make sure the low-water is never greater than 1378 * the high-water. 1379 */ 1380 case SO_SNDLOWAT: 1381 so->so_snd.ssb_lowat = 1382 (optval > so->so_snd.ssb_hiwat) ? 1383 so->so_snd.ssb_hiwat : optval; 1384 break; 1385 case SO_RCVLOWAT: 1386 so->so_rcv.ssb_lowat = 1387 (optval > so->so_rcv.ssb_hiwat) ? 1388 so->so_rcv.ssb_hiwat : optval; 1389 break; 1390 } 1391 break; 1392 1393 case SO_SNDTIMEO: 1394 case SO_RCVTIMEO: 1395 error = sooptcopyin(sopt, &tv, sizeof tv, 1396 sizeof tv); 1397 if (error) 1398 goto bad; 1399 1400 /* assert(hz > 0); */ 1401 if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz || 1402 tv.tv_usec < 0 || tv.tv_usec >= 1000000) { 1403 error = EDOM; 1404 goto bad; 1405 } 1406 /* assert(tick > 0); */ 1407 /* assert(ULONG_MAX - SHRT_MAX >= 1000000); */ 1408 val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick; 1409 if (val > SHRT_MAX) { 1410 error = EDOM; 1411 goto bad; 1412 } 1413 if (val == 0 && tv.tv_usec != 0) 1414 val = 1; 1415 1416 switch (sopt->sopt_name) { 1417 case SO_SNDTIMEO: 1418 so->so_snd.ssb_timeo = val; 1419 break; 1420 case SO_RCVTIMEO: 1421 so->so_rcv.ssb_timeo = val; 1422 break; 1423 } 1424 break; 1425 default: 1426 error = ENOPROTOOPT; 1427 break; 1428 } 1429 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) { 1430 (void) so_pru_ctloutput(so, sopt); 1431 } 1432 } 1433 bad: 1434 return (error); 1435 } 1436 1437 /* Helper routine for getsockopt */ 1438 int 1439 sooptcopyout(struct sockopt *sopt, const void *buf, size_t len) 1440 { 1441 soopt_from_kbuf(sopt, buf, len); 1442 return 0; 1443 } 1444 1445 void 1446 soopt_from_kbuf(struct sockopt *sopt, const void *buf, size_t len) 1447 { 1448 size_t valsize; 1449 1450 KKASSERT(!sopt->sopt_val || kva_p(sopt->sopt_val)); 1451 KKASSERT(kva_p(buf)); 1452 1453 /* 1454 * Documented get behavior is that we always return a value, 1455 * possibly truncated to fit in the user's buffer. 1456 * Traditional behavior is that we always tell the user 1457 * precisely how much we copied, rather than something useful 1458 * like the total amount we had available for her. 1459 * Note that this interface is not idempotent; the entire answer must 1460 * generated ahead of time. 1461 */ 1462 valsize = min(len, sopt->sopt_valsize); 1463 sopt->sopt_valsize = valsize; 1464 if (sopt->sopt_val != 0) { 1465 bcopy(buf, sopt->sopt_val, valsize); 1466 } 1467 } 1468 1469 int 1470 sogetopt(struct socket *so, struct sockopt *sopt) 1471 { 1472 int error, optval; 1473 struct linger l; 1474 struct timeval tv; 1475 #ifdef INET 1476 struct accept_filter_arg *afap; 1477 #endif 1478 1479 error = 0; 1480 sopt->sopt_dir = SOPT_GET; 1481 if (sopt->sopt_level != SOL_SOCKET) { 1482 if (so->so_proto && so->so_proto->pr_ctloutput) { 1483 return (so_pru_ctloutput(so, sopt)); 1484 } else 1485 return (ENOPROTOOPT); 1486 } else { 1487 switch (sopt->sopt_name) { 1488 #ifdef INET 1489 case SO_ACCEPTFILTER: 1490 if ((so->so_options & SO_ACCEPTCONN) == 0) 1491 return (EINVAL); 1492 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), 1493 M_TEMP, M_WAITOK | M_ZERO); 1494 if ((so->so_options & SO_ACCEPTFILTER) != 0) { 1495 strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name); 1496 if (so->so_accf->so_accept_filter_str != NULL) 1497 strcpy(afap->af_arg, so->so_accf->so_accept_filter_str); 1498 } 1499 error = sooptcopyout(sopt, afap, sizeof(*afap)); 1500 FREE(afap, M_TEMP); 1501 break; 1502 #endif /* INET */ 1503 1504 case SO_LINGER: 1505 l.l_onoff = so->so_options & SO_LINGER; 1506 l.l_linger = so->so_linger; 1507 error = sooptcopyout(sopt, &l, sizeof l); 1508 break; 1509 1510 case SO_USELOOPBACK: 1511 case SO_DONTROUTE: 1512 case SO_DEBUG: 1513 case SO_KEEPALIVE: 1514 case SO_REUSEADDR: 1515 case SO_REUSEPORT: 1516 case SO_BROADCAST: 1517 case SO_OOBINLINE: 1518 case SO_TIMESTAMP: 1519 optval = so->so_options & sopt->sopt_name; 1520 integer: 1521 error = sooptcopyout(sopt, &optval, sizeof optval); 1522 break; 1523 1524 case SO_TYPE: 1525 optval = so->so_type; 1526 goto integer; 1527 1528 case SO_ERROR: 1529 optval = so->so_error; 1530 so->so_error = 0; 1531 goto integer; 1532 1533 case SO_SNDBUF: 1534 optval = so->so_snd.ssb_hiwat; 1535 goto integer; 1536 1537 case SO_RCVBUF: 1538 optval = so->so_rcv.ssb_hiwat; 1539 goto integer; 1540 1541 case SO_SNDLOWAT: 1542 optval = so->so_snd.ssb_lowat; 1543 goto integer; 1544 1545 case SO_RCVLOWAT: 1546 optval = so->so_rcv.ssb_lowat; 1547 goto integer; 1548 1549 case SO_SNDTIMEO: 1550 case SO_RCVTIMEO: 1551 optval = (sopt->sopt_name == SO_SNDTIMEO ? 1552 so->so_snd.ssb_timeo : so->so_rcv.ssb_timeo); 1553 1554 tv.tv_sec = optval / hz; 1555 tv.tv_usec = (optval % hz) * tick; 1556 error = sooptcopyout(sopt, &tv, sizeof tv); 1557 break; 1558 1559 default: 1560 error = ENOPROTOOPT; 1561 break; 1562 } 1563 return (error); 1564 } 1565 } 1566 1567 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */ 1568 int 1569 soopt_getm(struct sockopt *sopt, struct mbuf **mp) 1570 { 1571 struct mbuf *m, *m_prev; 1572 int sopt_size = sopt->sopt_valsize, msize; 1573 1574 m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_DATA, 1575 0, &msize); 1576 if (m == NULL) 1577 return (ENOBUFS); 1578 m->m_len = min(msize, sopt_size); 1579 sopt_size -= m->m_len; 1580 *mp = m; 1581 m_prev = m; 1582 1583 while (sopt_size > 0) { 1584 m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, 1585 MT_DATA, 0, &msize); 1586 if (m == NULL) { 1587 m_freem(*mp); 1588 return (ENOBUFS); 1589 } 1590 m->m_len = min(msize, sopt_size); 1591 sopt_size -= m->m_len; 1592 m_prev->m_next = m; 1593 m_prev = m; 1594 } 1595 return (0); 1596 } 1597 1598 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */ 1599 int 1600 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m) 1601 { 1602 soopt_to_mbuf(sopt, m); 1603 return 0; 1604 } 1605 1606 void 1607 soopt_to_mbuf(struct sockopt *sopt, struct mbuf *m) 1608 { 1609 size_t valsize; 1610 void *val; 1611 1612 KKASSERT(!sopt->sopt_val || kva_p(sopt->sopt_val)); 1613 KKASSERT(kva_p(m)); 1614 if (sopt->sopt_val == NULL) 1615 return; 1616 val = sopt->sopt_val; 1617 valsize = sopt->sopt_valsize; 1618 while (m != NULL && valsize >= m->m_len) { 1619 bcopy(val, mtod(m, char *), m->m_len); 1620 valsize -= m->m_len; 1621 val = (caddr_t)val + m->m_len; 1622 m = m->m_next; 1623 } 1624 if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */ 1625 panic("ip6_sooptmcopyin"); 1626 } 1627 1628 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */ 1629 int 1630 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m) 1631 { 1632 return soopt_from_mbuf(sopt, m); 1633 } 1634 1635 int 1636 soopt_from_mbuf(struct sockopt *sopt, struct mbuf *m) 1637 { 1638 struct mbuf *m0 = m; 1639 size_t valsize = 0; 1640 size_t maxsize; 1641 void *val; 1642 1643 KKASSERT(!sopt->sopt_val || kva_p(sopt->sopt_val)); 1644 KKASSERT(kva_p(m)); 1645 if (sopt->sopt_val == NULL) 1646 return 0; 1647 val = sopt->sopt_val; 1648 maxsize = sopt->sopt_valsize; 1649 while (m != NULL && maxsize >= m->m_len) { 1650 bcopy(mtod(m, char *), val, m->m_len); 1651 maxsize -= m->m_len; 1652 val = (caddr_t)val + m->m_len; 1653 valsize += m->m_len; 1654 m = m->m_next; 1655 } 1656 if (m != NULL) { 1657 /* enough soopt buffer should be given from user-land */ 1658 m_freem(m0); 1659 return (EINVAL); 1660 } 1661 sopt->sopt_valsize = valsize; 1662 return 0; 1663 } 1664 1665 void 1666 sohasoutofband(struct socket *so) 1667 { 1668 if (so->so_sigio != NULL) 1669 pgsigio(so->so_sigio, SIGURG, 0); 1670 selwakeup(&so->so_rcv.ssb_sel); 1671 } 1672 1673 int 1674 sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td) 1675 { 1676 int revents = 0; 1677 1678 crit_enter(); 1679 1680 if (events & (POLLIN | POLLRDNORM)) 1681 if (soreadable(so)) 1682 revents |= events & (POLLIN | POLLRDNORM); 1683 1684 if (events & POLLINIGNEOF) 1685 if (so->so_rcv.ssb_cc >= so->so_rcv.ssb_lowat || 1686 !TAILQ_EMPTY(&so->so_comp) || so->so_error) 1687 revents |= POLLINIGNEOF; 1688 1689 if (events & (POLLOUT | POLLWRNORM)) 1690 if (sowriteable(so)) 1691 revents |= events & (POLLOUT | POLLWRNORM); 1692 1693 if (events & (POLLPRI | POLLRDBAND)) 1694 if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) 1695 revents |= events & (POLLPRI | POLLRDBAND); 1696 1697 if (revents == 0) { 1698 if (events & 1699 (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM | 1700 POLLRDBAND)) { 1701 selrecord(td, &so->so_rcv.ssb_sel); 1702 so->so_rcv.ssb_flags |= SSB_SEL; 1703 } 1704 1705 if (events & (POLLOUT | POLLWRNORM)) { 1706 selrecord(td, &so->so_snd.ssb_sel); 1707 so->so_snd.ssb_flags |= SSB_SEL; 1708 } 1709 } 1710 1711 crit_exit(); 1712 return (revents); 1713 } 1714 1715 int 1716 sokqfilter(struct file *fp, struct knote *kn) 1717 { 1718 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1719 struct signalsockbuf *ssb; 1720 1721 switch (kn->kn_filter) { 1722 case EVFILT_READ: 1723 if (so->so_options & SO_ACCEPTCONN) 1724 kn->kn_fop = &solisten_filtops; 1725 else 1726 kn->kn_fop = &soread_filtops; 1727 ssb = &so->so_rcv; 1728 break; 1729 case EVFILT_WRITE: 1730 kn->kn_fop = &sowrite_filtops; 1731 ssb = &so->so_snd; 1732 break; 1733 default: 1734 return (1); 1735 } 1736 1737 crit_enter(); 1738 SLIST_INSERT_HEAD(&ssb->ssb_sel.si_note, kn, kn_selnext); 1739 ssb->ssb_flags |= SSB_KNOTE; 1740 crit_exit(); 1741 return (0); 1742 } 1743 1744 static void 1745 filt_sordetach(struct knote *kn) 1746 { 1747 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1748 1749 crit_enter(); 1750 SLIST_REMOVE(&so->so_rcv.ssb_sel.si_note, kn, knote, kn_selnext); 1751 if (SLIST_EMPTY(&so->so_rcv.ssb_sel.si_note)) 1752 so->so_rcv.ssb_flags &= ~SSB_KNOTE; 1753 crit_exit(); 1754 } 1755 1756 /*ARGSUSED*/ 1757 static int 1758 filt_soread(struct knote *kn, long hint) 1759 { 1760 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1761 1762 kn->kn_data = so->so_rcv.ssb_cc; 1763 if (so->so_state & SS_CANTRCVMORE) { 1764 kn->kn_flags |= EV_EOF; 1765 kn->kn_fflags = so->so_error; 1766 return (1); 1767 } 1768 if (so->so_error) /* temporary udp error */ 1769 return (1); 1770 if (kn->kn_sfflags & NOTE_LOWAT) 1771 return (kn->kn_data >= kn->kn_sdata); 1772 return (kn->kn_data >= so->so_rcv.ssb_lowat); 1773 } 1774 1775 static void 1776 filt_sowdetach(struct knote *kn) 1777 { 1778 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1779 1780 crit_enter(); 1781 SLIST_REMOVE(&so->so_snd.ssb_sel.si_note, kn, knote, kn_selnext); 1782 if (SLIST_EMPTY(&so->so_snd.ssb_sel.si_note)) 1783 so->so_snd.ssb_flags &= ~SSB_KNOTE; 1784 crit_exit(); 1785 } 1786 1787 /*ARGSUSED*/ 1788 static int 1789 filt_sowrite(struct knote *kn, long hint) 1790 { 1791 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1792 1793 kn->kn_data = ssb_space(&so->so_snd); 1794 if (so->so_state & SS_CANTSENDMORE) { 1795 kn->kn_flags |= EV_EOF; 1796 kn->kn_fflags = so->so_error; 1797 return (1); 1798 } 1799 if (so->so_error) /* temporary udp error */ 1800 return (1); 1801 if (((so->so_state & SS_ISCONNECTED) == 0) && 1802 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 1803 return (0); 1804 if (kn->kn_sfflags & NOTE_LOWAT) 1805 return (kn->kn_data >= kn->kn_sdata); 1806 return (kn->kn_data >= so->so_snd.ssb_lowat); 1807 } 1808 1809 /*ARGSUSED*/ 1810 static int 1811 filt_solisten(struct knote *kn, long hint) 1812 { 1813 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1814 1815 kn->kn_data = so->so_qlen; 1816 return (! TAILQ_EMPTY(&so->so_comp)); 1817 } 1818