uipc_socket.c revision 72968
1/* 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)uipc_socket.c 8.3 (Berkeley) 4/15/94 34 * $FreeBSD: head/sys/kern/uipc_socket.c 72968 2001-02-24 01:41:31Z jlemon $ 35 */ 36 37#include "opt_inet.h" 38 39#include <sys/param.h> 40#include <sys/systm.h> 41#include <sys/fcntl.h> 42#include <sys/malloc.h> 43#include <sys/mbuf.h> 44#include <sys/domain.h> 45#include <sys/file.h> /* for struct knote */ 46#include <sys/kernel.h> 47#include <sys/malloc.h> 48#include <sys/event.h> 49#include <sys/poll.h> 50#include <sys/proc.h> 51#include <sys/protosw.h> 52#include <sys/socket.h> 53#include <sys/socketvar.h> 54#include <sys/resourcevar.h> 55#include <sys/signalvar.h> 56#include <sys/sysctl.h> 57#include <sys/uio.h> 58#include <sys/jail.h> 59#include <vm/vm_zone.h> 60 61#include <machine/limits.h> 62 63#ifdef INET 64static int do_setopt_accept_filter(struct socket *so, struct sockopt *sopt); 65#endif 66 67static void filt_sordetach(struct knote *kn); 68static int filt_soread(struct knote *kn, long hint); 69static void filt_sowdetach(struct knote *kn); 70static int filt_sowrite(struct knote *kn, long hint); 71static int filt_solisten(struct knote *kn, long hint); 72 73static struct filterops solisten_filtops = 74 { 1, NULL, filt_sordetach, filt_solisten }; 75static struct filterops soread_filtops = 76 { 1, NULL, filt_sordetach, filt_soread }; 77static struct filterops sowrite_filtops = 78 { 1, NULL, filt_sowdetach, filt_sowrite }; 79 80struct vm_zone *socket_zone; 81so_gen_t so_gencnt; /* generation count for sockets */ 82 83MALLOC_DEFINE(M_SONAME, "soname", "socket name"); 84MALLOC_DEFINE(M_PCB, "pcb", "protocol control block"); 85 86SYSCTL_DECL(_kern_ipc); 87 88static int somaxconn = SOMAXCONN; 89SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW, 90 &somaxconn, 0, "Maximum pending socket connection queue size"); 91 92/* 93 * Socket operation routines. 94 * These routines are called by the routines in 95 * sys_socket.c or from a system process, and 96 * implement the semantics of socket operations by 97 * switching out to the protocol specific routines. 98 */ 99 100/* 101 * Get a socket structure from our zone, and initialize it. 102 * We don't implement `waitok' yet (see comments in uipc_domain.c). 103 * Note that it would probably be better to allocate socket 104 * and PCB at the same time, but I'm not convinced that all 105 * the protocols can be easily modified to do this. 106 */ 107struct socket * 108soalloc(waitok) 109 int waitok; 110{ 111 struct socket *so; 112 113 so = zalloc(socket_zone); 114 if (so) { 115 /* XXX race condition for reentrant kernel */ 116 bzero(so, sizeof *so); 117 so->so_gencnt = ++so_gencnt; 118 so->so_zone = socket_zone; 119 TAILQ_INIT(&so->so_aiojobq); 120 } 121 return so; 122} 123 124int 125socreate(dom, aso, type, proto, p) 126 int dom; 127 struct socket **aso; 128 register int type; 129 int proto; 130 struct proc *p; 131{ 132 register struct protosw *prp; 133 register struct socket *so; 134 register int error; 135 136 if (proto) 137 prp = pffindproto(dom, proto, type); 138 else 139 prp = pffindtype(dom, type); 140 141 if (prp == 0 || prp->pr_usrreqs->pru_attach == 0) 142 return (EPROTONOSUPPORT); 143 144 if (jailed(p->p_ucred) && jail_socket_unixiproute_only && 145 prp->pr_domain->dom_family != PF_LOCAL && 146 prp->pr_domain->dom_family != PF_INET && 147 prp->pr_domain->dom_family != PF_ROUTE) { 148 return (EPROTONOSUPPORT); 149 } 150 151 if (prp->pr_type != type) 152 return (EPROTOTYPE); 153 so = soalloc(p != 0); 154 if (so == 0) 155 return (ENOBUFS); 156 157 TAILQ_INIT(&so->so_incomp); 158 TAILQ_INIT(&so->so_comp); 159 so->so_type = type; 160 so->so_cred = p->p_ucred; 161 crhold(so->so_cred); 162 so->so_proto = prp; 163 error = (*prp->pr_usrreqs->pru_attach)(so, proto, p); 164 if (error) { 165 so->so_state |= SS_NOFDREF; 166 sofree(so); 167 return (error); 168 } 169 *aso = so; 170 return (0); 171} 172 173int 174sobind(so, nam, p) 175 struct socket *so; 176 struct sockaddr *nam; 177 struct proc *p; 178{ 179 int s = splnet(); 180 int error; 181 182 error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam, p); 183 splx(s); 184 return (error); 185} 186 187void 188sodealloc(so) 189 struct socket *so; 190{ 191 192 so->so_gencnt = ++so_gencnt; 193 if (so->so_rcv.sb_hiwat) 194 (void)chgsbsize(so->so_cred->cr_uidinfo, 195 &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY); 196 if (so->so_snd.sb_hiwat) 197 (void)chgsbsize(so->so_cred->cr_uidinfo, 198 &so->so_snd.sb_hiwat, 0, RLIM_INFINITY); 199#ifdef INET 200 if (so->so_accf != NULL) { 201 if (so->so_accf->so_accept_filter != NULL && 202 so->so_accf->so_accept_filter->accf_destroy != NULL) { 203 so->so_accf->so_accept_filter->accf_destroy(so); 204 } 205 if (so->so_accf->so_accept_filter_str != NULL) 206 FREE(so->so_accf->so_accept_filter_str, M_ACCF); 207 FREE(so->so_accf, M_ACCF); 208 } 209#endif 210 crfree(so->so_cred); 211 zfree(so->so_zone, so); 212} 213 214int 215solisten(so, backlog, p) 216 register struct socket *so; 217 int backlog; 218 struct proc *p; 219{ 220 int s, error; 221 222 s = splnet(); 223 error = (*so->so_proto->pr_usrreqs->pru_listen)(so, p); 224 if (error) { 225 splx(s); 226 return (error); 227 } 228 if (TAILQ_EMPTY(&so->so_comp)) 229 so->so_options |= SO_ACCEPTCONN; 230 if (backlog < 0 || backlog > somaxconn) 231 backlog = somaxconn; 232 so->so_qlimit = backlog; 233 splx(s); 234 return (0); 235} 236 237void 238sofree(so) 239 register struct socket *so; 240{ 241 struct socket *head = so->so_head; 242 243 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) 244 return; 245 if (head != NULL) { 246 if (so->so_state & SS_INCOMP) { 247 TAILQ_REMOVE(&head->so_incomp, so, so_list); 248 head->so_incqlen--; 249 } else if (so->so_state & SS_COMP) { 250 /* 251 * We must not decommission a socket that's 252 * on the accept(2) queue. If we do, then 253 * accept(2) may hang after select(2) indicated 254 * that the listening socket was ready. 255 */ 256 return; 257 } else { 258 panic("sofree: not queued"); 259 } 260 head->so_qlen--; 261 so->so_state &= ~SS_INCOMP; 262 so->so_head = NULL; 263 } 264 sbrelease(&so->so_snd, so); 265 sorflush(so); 266 sodealloc(so); 267} 268 269/* 270 * Close a socket on last file table reference removal. 271 * Initiate disconnect if connected. 272 * Free socket when disconnect complete. 273 */ 274int 275soclose(so) 276 register struct socket *so; 277{ 278 int s = splnet(); /* conservative */ 279 int error = 0; 280 281 funsetown(so->so_sigio); 282 if (so->so_options & SO_ACCEPTCONN) { 283 struct socket *sp, *sonext; 284 285 sp = TAILQ_FIRST(&so->so_incomp); 286 for (; sp != NULL; sp = sonext) { 287 sonext = TAILQ_NEXT(sp, so_list); 288 (void) soabort(sp); 289 } 290 for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) { 291 sonext = TAILQ_NEXT(sp, so_list); 292 /* Dequeue from so_comp since sofree() won't do it */ 293 TAILQ_REMOVE(&so->so_comp, sp, so_list); 294 so->so_qlen--; 295 sp->so_state &= ~SS_COMP; 296 sp->so_head = NULL; 297 (void) soabort(sp); 298 } 299 } 300 if (so->so_pcb == 0) 301 goto discard; 302 if (so->so_state & SS_ISCONNECTED) { 303 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 304 error = sodisconnect(so); 305 if (error) 306 goto drop; 307 } 308 if (so->so_options & SO_LINGER) { 309 if ((so->so_state & SS_ISDISCONNECTING) && 310 (so->so_state & SS_NBIO)) 311 goto drop; 312 while (so->so_state & SS_ISCONNECTED) { 313 error = tsleep((caddr_t)&so->so_timeo, 314 PSOCK | PCATCH, "soclos", so->so_linger * hz); 315 if (error) 316 break; 317 } 318 } 319 } 320drop: 321 if (so->so_pcb) { 322 int error2 = (*so->so_proto->pr_usrreqs->pru_detach)(so); 323 if (error == 0) 324 error = error2; 325 } 326discard: 327 if (so->so_state & SS_NOFDREF) 328 panic("soclose: NOFDREF"); 329 so->so_state |= SS_NOFDREF; 330 sofree(so); 331 splx(s); 332 return (error); 333} 334 335/* 336 * Must be called at splnet... 337 */ 338int 339soabort(so) 340 struct socket *so; 341{ 342 int error; 343 344 error = (*so->so_proto->pr_usrreqs->pru_abort)(so); 345 if (error) { 346 sofree(so); 347 return error; 348 } 349 return (0); 350} 351 352int 353soaccept(so, nam) 354 register struct socket *so; 355 struct sockaddr **nam; 356{ 357 int s = splnet(); 358 int error; 359 360 if ((so->so_state & SS_NOFDREF) == 0) 361 panic("soaccept: !NOFDREF"); 362 so->so_state &= ~SS_NOFDREF; 363 if ((so->so_state & SS_ISDISCONNECTED) == 0) 364 error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam); 365 else 366 error = ECONNABORTED; 367 splx(s); 368 return (error); 369} 370 371int 372soconnect(so, nam, p) 373 register struct socket *so; 374 struct sockaddr *nam; 375 struct proc *p; 376{ 377 int s; 378 int error; 379 380 if (so->so_options & SO_ACCEPTCONN) 381 return (EOPNOTSUPP); 382 s = splnet(); 383 /* 384 * If protocol is connection-based, can only connect once. 385 * Otherwise, if connected, try to disconnect first. 386 * This allows user to disconnect by connecting to, e.g., 387 * a null address. 388 */ 389 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 390 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 391 (error = sodisconnect(so)))) 392 error = EISCONN; 393 else 394 error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, p); 395 splx(s); 396 return (error); 397} 398 399int 400soconnect2(so1, so2) 401 register struct socket *so1; 402 struct socket *so2; 403{ 404 int s = splnet(); 405 int error; 406 407 error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2); 408 splx(s); 409 return (error); 410} 411 412int 413sodisconnect(so) 414 register struct socket *so; 415{ 416 int s = splnet(); 417 int error; 418 419 if ((so->so_state & SS_ISCONNECTED) == 0) { 420 error = ENOTCONN; 421 goto bad; 422 } 423 if (so->so_state & SS_ISDISCONNECTING) { 424 error = EALREADY; 425 goto bad; 426 } 427 error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so); 428bad: 429 splx(s); 430 return (error); 431} 432 433#define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 434/* 435 * Send on a socket. 436 * If send must go all at once and message is larger than 437 * send buffering, then hard error. 438 * Lock against other senders. 439 * If must go all at once and not enough room now, then 440 * inform user that this would block and do nothing. 441 * Otherwise, if nonblocking, send as much as possible. 442 * The data to be sent is described by "uio" if nonzero, 443 * otherwise by the mbuf chain "top" (which must be null 444 * if uio is not). Data provided in mbuf chain must be small 445 * enough to send all at once. 446 * 447 * Returns nonzero on error, timeout or signal; callers 448 * must check for short counts if EINTR/ERESTART are returned. 449 * Data and control buffers are freed on return. 450 */ 451int 452sosend(so, addr, uio, top, control, flags, p) 453 register struct socket *so; 454 struct sockaddr *addr; 455 struct uio *uio; 456 struct mbuf *top; 457 struct mbuf *control; 458 int flags; 459 struct proc *p; 460{ 461 struct mbuf **mp; 462 register struct mbuf *m; 463 register long space, len, resid; 464 int clen = 0, error, s, dontroute, mlen; 465 int atomic = sosendallatonce(so) || top; 466 467 if (uio) 468 resid = uio->uio_resid; 469 else 470 resid = top->m_pkthdr.len; 471 /* 472 * In theory resid should be unsigned. 473 * However, space must be signed, as it might be less than 0 474 * if we over-committed, and we must use a signed comparison 475 * of space and resid. On the other hand, a negative resid 476 * causes us to loop sending 0-length segments to the protocol. 477 * 478 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM 479 * type sockets since that's an error. 480 */ 481 if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) { 482 error = EINVAL; 483 goto out; 484 } 485 486 dontroute = 487 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 488 (so->so_proto->pr_flags & PR_ATOMIC); 489 if (p) 490 p->p_stats->p_ru.ru_msgsnd++; 491 if (control) 492 clen = control->m_len; 493#define snderr(errno) { error = errno; splx(s); goto release; } 494 495restart: 496 error = sblock(&so->so_snd, SBLOCKWAIT(flags)); 497 if (error) 498 goto out; 499 do { 500 s = splnet(); 501 if (so->so_state & SS_CANTSENDMORE) 502 snderr(EPIPE); 503 if (so->so_error) { 504 error = so->so_error; 505 so->so_error = 0; 506 splx(s); 507 goto release; 508 } 509 if ((so->so_state & SS_ISCONNECTED) == 0) { 510 /* 511 * `sendto' and `sendmsg' is allowed on a connection- 512 * based socket if it supports implied connect. 513 * Return ENOTCONN if not connected and no address is 514 * supplied. 515 */ 516 if ((so->so_proto->pr_flags & PR_CONNREQUIRED) && 517 (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) { 518 if ((so->so_state & SS_ISCONFIRMING) == 0 && 519 !(resid == 0 && clen != 0)) 520 snderr(ENOTCONN); 521 } else if (addr == 0) 522 snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ? 523 ENOTCONN : EDESTADDRREQ); 524 } 525 space = sbspace(&so->so_snd); 526 if (flags & MSG_OOB) 527 space += 1024; 528 if ((atomic && resid > so->so_snd.sb_hiwat) || 529 clen > so->so_snd.sb_hiwat) 530 snderr(EMSGSIZE); 531 if (space < resid + clen && uio && 532 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 533 if (so->so_state & SS_NBIO) 534 snderr(EWOULDBLOCK); 535 sbunlock(&so->so_snd); 536 error = sbwait(&so->so_snd); 537 splx(s); 538 if (error) 539 goto out; 540 goto restart; 541 } 542 splx(s); 543 mp = ⊤ 544 space -= clen; 545 do { 546 if (uio == NULL) { 547 /* 548 * Data is prepackaged in "top". 549 */ 550 resid = 0; 551 if (flags & MSG_EOR) 552 top->m_flags |= M_EOR; 553 } else do { 554 if (top == 0) { 555 MGETHDR(m, M_TRYWAIT, MT_DATA); 556 if (m == NULL) { 557 error = ENOBUFS; 558 goto release; 559 } 560 mlen = MHLEN; 561 m->m_pkthdr.len = 0; 562 m->m_pkthdr.rcvif = (struct ifnet *)0; 563 } else { 564 MGET(m, M_TRYWAIT, MT_DATA); 565 if (m == NULL) { 566 error = ENOBUFS; 567 goto release; 568 } 569 mlen = MLEN; 570 } 571 if (resid >= MINCLSIZE) { 572 MCLGET(m, M_TRYWAIT); 573 if ((m->m_flags & M_EXT) == 0) 574 goto nopages; 575 mlen = MCLBYTES; 576 len = min(min(mlen, resid), space); 577 } else { 578nopages: 579 len = min(min(mlen, resid), space); 580 /* 581 * For datagram protocols, leave room 582 * for protocol headers in first mbuf. 583 */ 584 if (atomic && top == 0 && len < mlen) 585 MH_ALIGN(m, len); 586 } 587 space -= len; 588 error = uiomove(mtod(m, caddr_t), (int)len, uio); 589 resid = uio->uio_resid; 590 m->m_len = len; 591 *mp = m; 592 top->m_pkthdr.len += len; 593 if (error) 594 goto release; 595 mp = &m->m_next; 596 if (resid <= 0) { 597 if (flags & MSG_EOR) 598 top->m_flags |= M_EOR; 599 break; 600 } 601 } while (space > 0 && atomic); 602 if (dontroute) 603 so->so_options |= SO_DONTROUTE; 604 s = splnet(); /* XXX */ 605 /* 606 * XXX all the SS_CANTSENDMORE checks previously 607 * done could be out of date. We could have recieved 608 * a reset packet in an interrupt or maybe we slept 609 * while doing page faults in uiomove() etc. We could 610 * probably recheck again inside the splnet() protection 611 * here, but there are probably other places that this 612 * also happens. We must rethink this. 613 */ 614 error = (*so->so_proto->pr_usrreqs->pru_send)(so, 615 (flags & MSG_OOB) ? PRUS_OOB : 616 /* 617 * If the user set MSG_EOF, the protocol 618 * understands this flag and nothing left to 619 * send then use PRU_SEND_EOF instead of PRU_SEND. 620 */ 621 ((flags & MSG_EOF) && 622 (so->so_proto->pr_flags & PR_IMPLOPCL) && 623 (resid <= 0)) ? 624 PRUS_EOF : 625 /* If there is more to send set PRUS_MORETOCOME */ 626 (resid > 0 && space > 0) ? PRUS_MORETOCOME : 0, 627 top, addr, control, p); 628 splx(s); 629 if (dontroute) 630 so->so_options &= ~SO_DONTROUTE; 631 clen = 0; 632 control = 0; 633 top = 0; 634 mp = ⊤ 635 if (error) 636 goto release; 637 } while (resid && space > 0); 638 } while (resid); 639 640release: 641 sbunlock(&so->so_snd); 642out: 643 if (top) 644 m_freem(top); 645 if (control) 646 m_freem(control); 647 return (error); 648} 649 650/* 651 * Implement receive operations on a socket. 652 * We depend on the way that records are added to the sockbuf 653 * by sbappend*. In particular, each record (mbufs linked through m_next) 654 * must begin with an address if the protocol so specifies, 655 * followed by an optional mbuf or mbufs containing ancillary data, 656 * and then zero or more mbufs of data. 657 * In order to avoid blocking network interrupts for the entire time here, 658 * we splx() while doing the actual copy to user space. 659 * Although the sockbuf is locked, new data may still be appended, 660 * and thus we must maintain consistency of the sockbuf during that time. 661 * 662 * The caller may receive the data as a single mbuf chain by supplying 663 * an mbuf **mp0 for use in returning the chain. The uio is then used 664 * only for the count in uio_resid. 665 */ 666int 667soreceive(so, psa, uio, mp0, controlp, flagsp) 668 register struct socket *so; 669 struct sockaddr **psa; 670 struct uio *uio; 671 struct mbuf **mp0; 672 struct mbuf **controlp; 673 int *flagsp; 674{ 675 register struct mbuf *m, **mp; 676 register int flags, len, error, s, offset; 677 struct protosw *pr = so->so_proto; 678 struct mbuf *nextrecord; 679 int moff, type = 0; 680 int orig_resid = uio->uio_resid; 681 682 mp = mp0; 683 if (psa) 684 *psa = 0; 685 if (controlp) 686 *controlp = 0; 687 if (flagsp) 688 flags = *flagsp &~ MSG_EOR; 689 else 690 flags = 0; 691 if (flags & MSG_OOB) { 692 m = m_get(M_TRYWAIT, MT_DATA); 693 if (m == NULL) 694 return (ENOBUFS); 695 error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK); 696 if (error) 697 goto bad; 698 do { 699 error = uiomove(mtod(m, caddr_t), 700 (int) min(uio->uio_resid, m->m_len), uio); 701 m = m_free(m); 702 } while (uio->uio_resid && error == 0 && m); 703bad: 704 if (m) 705 m_freem(m); 706 return (error); 707 } 708 if (mp) 709 *mp = (struct mbuf *)0; 710 if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) 711 (*pr->pr_usrreqs->pru_rcvd)(so, 0); 712 713restart: 714 error = sblock(&so->so_rcv, SBLOCKWAIT(flags)); 715 if (error) 716 return (error); 717 s = splnet(); 718 719 m = so->so_rcv.sb_mb; 720 /* 721 * If we have less data than requested, block awaiting more 722 * (subject to any timeout) if: 723 * 1. the current count is less than the low water mark, or 724 * 2. MSG_WAITALL is set, and it is possible to do the entire 725 * receive operation at once if we block (resid <= hiwat). 726 * 3. MSG_DONTWAIT is not set 727 * If MSG_WAITALL is set but resid is larger than the receive buffer, 728 * we have to do the receive in sections, and thus risk returning 729 * a short count if a timeout or signal occurs after we start. 730 */ 731 if (m == 0 || (((flags & MSG_DONTWAIT) == 0 && 732 so->so_rcv.sb_cc < uio->uio_resid) && 733 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 734 ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) && 735 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) { 736 KASSERT(m != 0 || !so->so_rcv.sb_cc, ("receive 1")); 737 if (so->so_error) { 738 if (m) 739 goto dontblock; 740 error = so->so_error; 741 if ((flags & MSG_PEEK) == 0) 742 so->so_error = 0; 743 goto release; 744 } 745 if (so->so_state & SS_CANTRCVMORE) { 746 if (m) 747 goto dontblock; 748 else 749 goto release; 750 } 751 for (; m; m = m->m_next) 752 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 753 m = so->so_rcv.sb_mb; 754 goto dontblock; 755 } 756 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 757 (so->so_proto->pr_flags & PR_CONNREQUIRED)) { 758 error = ENOTCONN; 759 goto release; 760 } 761 if (uio->uio_resid == 0) 762 goto release; 763 if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) { 764 error = EWOULDBLOCK; 765 goto release; 766 } 767 sbunlock(&so->so_rcv); 768 error = sbwait(&so->so_rcv); 769 splx(s); 770 if (error) 771 return (error); 772 goto restart; 773 } 774dontblock: 775 if (uio->uio_procp) 776 uio->uio_procp->p_stats->p_ru.ru_msgrcv++; 777 nextrecord = m->m_nextpkt; 778 if (pr->pr_flags & PR_ADDR) { 779 KASSERT(m->m_type == MT_SONAME, ("receive 1a")); 780 orig_resid = 0; 781 if (psa) 782 *psa = dup_sockaddr(mtod(m, struct sockaddr *), 783 mp0 == 0); 784 if (flags & MSG_PEEK) { 785 m = m->m_next; 786 } else { 787 sbfree(&so->so_rcv, m); 788 MFREE(m, so->so_rcv.sb_mb); 789 m = so->so_rcv.sb_mb; 790 } 791 } 792 while (m && m->m_type == MT_CONTROL && error == 0) { 793 if (flags & MSG_PEEK) { 794 if (controlp) 795 *controlp = m_copy(m, 0, m->m_len); 796 m = m->m_next; 797 } else { 798 sbfree(&so->so_rcv, m); 799 if (controlp) { 800 if (pr->pr_domain->dom_externalize && 801 mtod(m, struct cmsghdr *)->cmsg_type == 802 SCM_RIGHTS) 803 error = (*pr->pr_domain->dom_externalize)(m); 804 *controlp = m; 805 so->so_rcv.sb_mb = m->m_next; 806 m->m_next = 0; 807 m = so->so_rcv.sb_mb; 808 } else { 809 MFREE(m, so->so_rcv.sb_mb); 810 m = so->so_rcv.sb_mb; 811 } 812 } 813 if (controlp) { 814 orig_resid = 0; 815 controlp = &(*controlp)->m_next; 816 } 817 } 818 if (m) { 819 if ((flags & MSG_PEEK) == 0) 820 m->m_nextpkt = nextrecord; 821 type = m->m_type; 822 if (type == MT_OOBDATA) 823 flags |= MSG_OOB; 824 } 825 moff = 0; 826 offset = 0; 827 while (m && uio->uio_resid > 0 && error == 0) { 828 if (m->m_type == MT_OOBDATA) { 829 if (type != MT_OOBDATA) 830 break; 831 } else if (type == MT_OOBDATA) 832 break; 833 else 834 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER, 835 ("receive 3")); 836 so->so_state &= ~SS_RCVATMARK; 837 len = uio->uio_resid; 838 if (so->so_oobmark && len > so->so_oobmark - offset) 839 len = so->so_oobmark - offset; 840 if (len > m->m_len - moff) 841 len = m->m_len - moff; 842 /* 843 * If mp is set, just pass back the mbufs. 844 * Otherwise copy them out via the uio, then free. 845 * Sockbuf must be consistent here (points to current mbuf, 846 * it points to next record) when we drop priority; 847 * we must note any additions to the sockbuf when we 848 * block interrupts again. 849 */ 850 if (mp == 0) { 851 splx(s); 852 error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio); 853 s = splnet(); 854 if (error) 855 goto release; 856 } else 857 uio->uio_resid -= len; 858 if (len == m->m_len - moff) { 859 if (m->m_flags & M_EOR) 860 flags |= MSG_EOR; 861 if (flags & MSG_PEEK) { 862 m = m->m_next; 863 moff = 0; 864 } else { 865 nextrecord = m->m_nextpkt; 866 sbfree(&so->so_rcv, m); 867 if (mp) { 868 *mp = m; 869 mp = &m->m_next; 870 so->so_rcv.sb_mb = m = m->m_next; 871 *mp = (struct mbuf *)0; 872 } else { 873 MFREE(m, so->so_rcv.sb_mb); 874 m = so->so_rcv.sb_mb; 875 } 876 if (m) 877 m->m_nextpkt = nextrecord; 878 } 879 } else { 880 if (flags & MSG_PEEK) 881 moff += len; 882 else { 883 if (mp) 884 *mp = m_copym(m, 0, len, M_TRYWAIT); 885 m->m_data += len; 886 m->m_len -= len; 887 so->so_rcv.sb_cc -= len; 888 } 889 } 890 if (so->so_oobmark) { 891 if ((flags & MSG_PEEK) == 0) { 892 so->so_oobmark -= len; 893 if (so->so_oobmark == 0) { 894 so->so_state |= SS_RCVATMARK; 895 break; 896 } 897 } else { 898 offset += len; 899 if (offset == so->so_oobmark) 900 break; 901 } 902 } 903 if (flags & MSG_EOR) 904 break; 905 /* 906 * If the MSG_WAITALL flag is set (for non-atomic socket), 907 * we must not quit until "uio->uio_resid == 0" or an error 908 * termination. If a signal/timeout occurs, return 909 * with a short count but without error. 910 * Keep sockbuf locked against other readers. 911 */ 912 while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 && 913 !sosendallatonce(so) && !nextrecord) { 914 if (so->so_error || so->so_state & SS_CANTRCVMORE) 915 break; 916 error = sbwait(&so->so_rcv); 917 if (error) { 918 sbunlock(&so->so_rcv); 919 splx(s); 920 return (0); 921 } 922 m = so->so_rcv.sb_mb; 923 if (m) 924 nextrecord = m->m_nextpkt; 925 } 926 } 927 928 if (m && pr->pr_flags & PR_ATOMIC) { 929 flags |= MSG_TRUNC; 930 if ((flags & MSG_PEEK) == 0) 931 (void) sbdroprecord(&so->so_rcv); 932 } 933 if ((flags & MSG_PEEK) == 0) { 934 if (m == 0) 935 so->so_rcv.sb_mb = nextrecord; 936 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 937 (*pr->pr_usrreqs->pru_rcvd)(so, flags); 938 } 939 if (orig_resid == uio->uio_resid && orig_resid && 940 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 941 sbunlock(&so->so_rcv); 942 splx(s); 943 goto restart; 944 } 945 946 if (flagsp) 947 *flagsp |= flags; 948release: 949 sbunlock(&so->so_rcv); 950 splx(s); 951 return (error); 952} 953 954int 955soshutdown(so, how) 956 register struct socket *so; 957 register int how; 958{ 959 register struct protosw *pr = so->so_proto; 960 961 how++; 962 if (how & FREAD) 963 sorflush(so); 964 if (how & FWRITE) 965 return ((*pr->pr_usrreqs->pru_shutdown)(so)); 966 return (0); 967} 968 969void 970sorflush(so) 971 register struct socket *so; 972{ 973 register struct sockbuf *sb = &so->so_rcv; 974 register struct protosw *pr = so->so_proto; 975 register int s; 976 struct sockbuf asb; 977 978 sb->sb_flags |= SB_NOINTR; 979 (void) sblock(sb, M_WAITOK); 980 s = splimp(); 981 socantrcvmore(so); 982 sbunlock(sb); 983 asb = *sb; 984 bzero((caddr_t)sb, sizeof (*sb)); 985 splx(s); 986 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) 987 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 988 sbrelease(&asb, so); 989} 990 991#ifdef INET 992static int 993do_setopt_accept_filter(so, sopt) 994 struct socket *so; 995 struct sockopt *sopt; 996{ 997 struct accept_filter_arg *afap = NULL; 998 struct accept_filter *afp; 999 struct so_accf *af = so->so_accf; 1000 int error = 0; 1001 1002 /* do not set/remove accept filters on non listen sockets */ 1003 if ((so->so_options & SO_ACCEPTCONN) == 0) { 1004 error = EINVAL; 1005 goto out; 1006 } 1007 1008 /* removing the filter */ 1009 if (sopt == NULL) { 1010 if (af != NULL) { 1011 if (af->so_accept_filter != NULL && 1012 af->so_accept_filter->accf_destroy != NULL) { 1013 af->so_accept_filter->accf_destroy(so); 1014 } 1015 if (af->so_accept_filter_str != NULL) { 1016 FREE(af->so_accept_filter_str, M_ACCF); 1017 } 1018 FREE(af, M_ACCF); 1019 so->so_accf = NULL; 1020 } 1021 so->so_options &= ~SO_ACCEPTFILTER; 1022 return (0); 1023 } 1024 /* adding a filter */ 1025 /* must remove previous filter first */ 1026 if (af != NULL) { 1027 error = EINVAL; 1028 goto out; 1029 } 1030 /* don't put large objects on the kernel stack */ 1031 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK); 1032 error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap); 1033 afap->af_name[sizeof(afap->af_name)-1] = '\0'; 1034 afap->af_arg[sizeof(afap->af_arg)-1] = '\0'; 1035 if (error) 1036 goto out; 1037 afp = accept_filt_get(afap->af_name); 1038 if (afp == NULL) { 1039 error = ENOENT; 1040 goto out; 1041 } 1042 MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK | M_ZERO); 1043 if (afp->accf_create != NULL) { 1044 if (afap->af_name[0] != '\0') { 1045 int len = strlen(afap->af_name) + 1; 1046 1047 MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK); 1048 strcpy(af->so_accept_filter_str, afap->af_name); 1049 } 1050 af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg); 1051 if (af->so_accept_filter_arg == NULL) { 1052 FREE(af->so_accept_filter_str, M_ACCF); 1053 FREE(af, M_ACCF); 1054 so->so_accf = NULL; 1055 error = EINVAL; 1056 goto out; 1057 } 1058 } 1059 af->so_accept_filter = afp; 1060 so->so_accf = af; 1061 so->so_options |= SO_ACCEPTFILTER; 1062out: 1063 if (afap != NULL) 1064 FREE(afap, M_TEMP); 1065 return (error); 1066} 1067#endif /* INET */ 1068 1069/* 1070 * Perhaps this routine, and sooptcopyout(), below, ought to come in 1071 * an additional variant to handle the case where the option value needs 1072 * to be some kind of integer, but not a specific size. 1073 * In addition to their use here, these functions are also called by the 1074 * protocol-level pr_ctloutput() routines. 1075 */ 1076int 1077sooptcopyin(sopt, buf, len, minlen) 1078 struct sockopt *sopt; 1079 void *buf; 1080 size_t len; 1081 size_t minlen; 1082{ 1083 size_t valsize; 1084 1085 /* 1086 * If the user gives us more than we wanted, we ignore it, 1087 * but if we don't get the minimum length the caller 1088 * wants, we return EINVAL. On success, sopt->sopt_valsize 1089 * is set to however much we actually retrieved. 1090 */ 1091 if ((valsize = sopt->sopt_valsize) < minlen) 1092 return EINVAL; 1093 if (valsize > len) 1094 sopt->sopt_valsize = valsize = len; 1095 1096 if (sopt->sopt_p != 0) 1097 return (copyin(sopt->sopt_val, buf, valsize)); 1098 1099 bcopy(sopt->sopt_val, buf, valsize); 1100 return 0; 1101} 1102 1103int 1104sosetopt(so, sopt) 1105 struct socket *so; 1106 struct sockopt *sopt; 1107{ 1108 int error, optval; 1109 struct linger l; 1110 struct timeval tv; 1111 u_long val; 1112 1113 error = 0; 1114 if (sopt->sopt_level != SOL_SOCKET) { 1115 if (so->so_proto && so->so_proto->pr_ctloutput) 1116 return ((*so->so_proto->pr_ctloutput) 1117 (so, sopt)); 1118 error = ENOPROTOOPT; 1119 } else { 1120 switch (sopt->sopt_name) { 1121#ifdef INET 1122 case SO_ACCEPTFILTER: 1123 error = do_setopt_accept_filter(so, sopt); 1124 if (error) 1125 goto bad; 1126 break; 1127#endif 1128 case SO_LINGER: 1129 error = sooptcopyin(sopt, &l, sizeof l, sizeof l); 1130 if (error) 1131 goto bad; 1132 1133 so->so_linger = l.l_linger; 1134 if (l.l_onoff) 1135 so->so_options |= SO_LINGER; 1136 else 1137 so->so_options &= ~SO_LINGER; 1138 break; 1139 1140 case SO_DEBUG: 1141 case SO_KEEPALIVE: 1142 case SO_DONTROUTE: 1143 case SO_USELOOPBACK: 1144 case SO_BROADCAST: 1145 case SO_REUSEADDR: 1146 case SO_REUSEPORT: 1147 case SO_OOBINLINE: 1148 case SO_TIMESTAMP: 1149 error = sooptcopyin(sopt, &optval, sizeof optval, 1150 sizeof optval); 1151 if (error) 1152 goto bad; 1153 if (optval) 1154 so->so_options |= sopt->sopt_name; 1155 else 1156 so->so_options &= ~sopt->sopt_name; 1157 break; 1158 1159 case SO_SNDBUF: 1160 case SO_RCVBUF: 1161 case SO_SNDLOWAT: 1162 case SO_RCVLOWAT: 1163 error = sooptcopyin(sopt, &optval, sizeof optval, 1164 sizeof optval); 1165 if (error) 1166 goto bad; 1167 1168 /* 1169 * Values < 1 make no sense for any of these 1170 * options, so disallow them. 1171 */ 1172 if (optval < 1) { 1173 error = EINVAL; 1174 goto bad; 1175 } 1176 1177 switch (sopt->sopt_name) { 1178 case SO_SNDBUF: 1179 case SO_RCVBUF: 1180 if (sbreserve(sopt->sopt_name == SO_SNDBUF ? 1181 &so->so_snd : &so->so_rcv, (u_long)optval, 1182 so, curproc) == 0) { 1183 error = ENOBUFS; 1184 goto bad; 1185 } 1186 break; 1187 1188 /* 1189 * Make sure the low-water is never greater than 1190 * the high-water. 1191 */ 1192 case SO_SNDLOWAT: 1193 so->so_snd.sb_lowat = 1194 (optval > so->so_snd.sb_hiwat) ? 1195 so->so_snd.sb_hiwat : optval; 1196 break; 1197 case SO_RCVLOWAT: 1198 so->so_rcv.sb_lowat = 1199 (optval > so->so_rcv.sb_hiwat) ? 1200 so->so_rcv.sb_hiwat : optval; 1201 break; 1202 } 1203 break; 1204 1205 case SO_SNDTIMEO: 1206 case SO_RCVTIMEO: 1207 error = sooptcopyin(sopt, &tv, sizeof tv, 1208 sizeof tv); 1209 if (error) 1210 goto bad; 1211 1212 /* assert(hz > 0); */ 1213 if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz || 1214 tv.tv_usec < 0 || tv.tv_usec >= 1000000) { 1215 error = EDOM; 1216 goto bad; 1217 } 1218 /* assert(tick > 0); */ 1219 /* assert(ULONG_MAX - SHRT_MAX >= 1000000); */ 1220 val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick; 1221 if (val > SHRT_MAX) { 1222 error = EDOM; 1223 goto bad; 1224 } 1225 1226 switch (sopt->sopt_name) { 1227 case SO_SNDTIMEO: 1228 so->so_snd.sb_timeo = val; 1229 break; 1230 case SO_RCVTIMEO: 1231 so->so_rcv.sb_timeo = val; 1232 break; 1233 } 1234 break; 1235 default: 1236 error = ENOPROTOOPT; 1237 break; 1238 } 1239 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) { 1240 (void) ((*so->so_proto->pr_ctloutput) 1241 (so, sopt)); 1242 } 1243 } 1244bad: 1245 return (error); 1246} 1247 1248/* Helper routine for getsockopt */ 1249int 1250sooptcopyout(sopt, buf, len) 1251 struct sockopt *sopt; 1252 void *buf; 1253 size_t len; 1254{ 1255 int error; 1256 size_t valsize; 1257 1258 error = 0; 1259 1260 /* 1261 * Documented get behavior is that we always return a value, 1262 * possibly truncated to fit in the user's buffer. 1263 * Traditional behavior is that we always tell the user 1264 * precisely how much we copied, rather than something useful 1265 * like the total amount we had available for her. 1266 * Note that this interface is not idempotent; the entire answer must 1267 * generated ahead of time. 1268 */ 1269 valsize = min(len, sopt->sopt_valsize); 1270 sopt->sopt_valsize = valsize; 1271 if (sopt->sopt_val != 0) { 1272 if (sopt->sopt_p != 0) 1273 error = copyout(buf, sopt->sopt_val, valsize); 1274 else 1275 bcopy(buf, sopt->sopt_val, valsize); 1276 } 1277 return error; 1278} 1279 1280int 1281sogetopt(so, sopt) 1282 struct socket *so; 1283 struct sockopt *sopt; 1284{ 1285 int error, optval; 1286 struct linger l; 1287 struct timeval tv; 1288#ifdef INET 1289 struct accept_filter_arg *afap; 1290#endif 1291 1292 error = 0; 1293 if (sopt->sopt_level != SOL_SOCKET) { 1294 if (so->so_proto && so->so_proto->pr_ctloutput) { 1295 return ((*so->so_proto->pr_ctloutput) 1296 (so, sopt)); 1297 } else 1298 return (ENOPROTOOPT); 1299 } else { 1300 switch (sopt->sopt_name) { 1301#ifdef INET 1302 case SO_ACCEPTFILTER: 1303 if ((so->so_options & SO_ACCEPTCONN) == 0) 1304 return (EINVAL); 1305 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), 1306 M_TEMP, M_WAITOK | M_ZERO); 1307 if ((so->so_options & SO_ACCEPTFILTER) != 0) { 1308 strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name); 1309 if (so->so_accf->so_accept_filter_str != NULL) 1310 strcpy(afap->af_arg, so->so_accf->so_accept_filter_str); 1311 } 1312 error = sooptcopyout(sopt, afap, sizeof(*afap)); 1313 FREE(afap, M_TEMP); 1314 break; 1315#endif 1316 1317 case SO_LINGER: 1318 l.l_onoff = so->so_options & SO_LINGER; 1319 l.l_linger = so->so_linger; 1320 error = sooptcopyout(sopt, &l, sizeof l); 1321 break; 1322 1323 case SO_USELOOPBACK: 1324 case SO_DONTROUTE: 1325 case SO_DEBUG: 1326 case SO_KEEPALIVE: 1327 case SO_REUSEADDR: 1328 case SO_REUSEPORT: 1329 case SO_BROADCAST: 1330 case SO_OOBINLINE: 1331 case SO_TIMESTAMP: 1332 optval = so->so_options & sopt->sopt_name; 1333integer: 1334 error = sooptcopyout(sopt, &optval, sizeof optval); 1335 break; 1336 1337 case SO_TYPE: 1338 optval = so->so_type; 1339 goto integer; 1340 1341 case SO_ERROR: 1342 optval = so->so_error; 1343 so->so_error = 0; 1344 goto integer; 1345 1346 case SO_SNDBUF: 1347 optval = so->so_snd.sb_hiwat; 1348 goto integer; 1349 1350 case SO_RCVBUF: 1351 optval = so->so_rcv.sb_hiwat; 1352 goto integer; 1353 1354 case SO_SNDLOWAT: 1355 optval = so->so_snd.sb_lowat; 1356 goto integer; 1357 1358 case SO_RCVLOWAT: 1359 optval = so->so_rcv.sb_lowat; 1360 goto integer; 1361 1362 case SO_SNDTIMEO: 1363 case SO_RCVTIMEO: 1364 optval = (sopt->sopt_name == SO_SNDTIMEO ? 1365 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 1366 1367 tv.tv_sec = optval / hz; 1368 tv.tv_usec = (optval % hz) * tick; 1369 error = sooptcopyout(sopt, &tv, sizeof tv); 1370 break; 1371 1372 default: 1373 error = ENOPROTOOPT; 1374 break; 1375 } 1376 return (error); 1377 } 1378} 1379 1380/* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */ 1381int 1382soopt_getm(struct sockopt *sopt, struct mbuf **mp) 1383{ 1384 struct mbuf *m, *m_prev; 1385 int sopt_size = sopt->sopt_valsize; 1386 1387 MGET(m, sopt->sopt_p ? M_TRYWAIT : M_DONTWAIT, MT_DATA); 1388 if (m == 0) 1389 return ENOBUFS; 1390 if (sopt_size > MLEN) { 1391 MCLGET(m, sopt->sopt_p ? M_TRYWAIT : M_DONTWAIT); 1392 if ((m->m_flags & M_EXT) == 0) { 1393 m_free(m); 1394 return ENOBUFS; 1395 } 1396 m->m_len = min(MCLBYTES, sopt_size); 1397 } else { 1398 m->m_len = min(MLEN, sopt_size); 1399 } 1400 sopt_size -= m->m_len; 1401 *mp = m; 1402 m_prev = m; 1403 1404 while (sopt_size) { 1405 MGET(m, sopt->sopt_p ? M_TRYWAIT : M_DONTWAIT, MT_DATA); 1406 if (m == 0) { 1407 m_freem(*mp); 1408 return ENOBUFS; 1409 } 1410 if (sopt_size > MLEN) { 1411 MCLGET(m, sopt->sopt_p ? M_TRYWAIT : M_DONTWAIT); 1412 if ((m->m_flags & M_EXT) == 0) { 1413 m_freem(*mp); 1414 return ENOBUFS; 1415 } 1416 m->m_len = min(MCLBYTES, sopt_size); 1417 } else { 1418 m->m_len = min(MLEN, sopt_size); 1419 } 1420 sopt_size -= m->m_len; 1421 m_prev->m_next = m; 1422 m_prev = m; 1423 } 1424 return 0; 1425} 1426 1427/* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */ 1428int 1429soopt_mcopyin(struct sockopt *sopt, struct mbuf *m) 1430{ 1431 struct mbuf *m0 = m; 1432 1433 if (sopt->sopt_val == NULL) 1434 return 0; 1435 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1436 if (sopt->sopt_p != NULL) { 1437 int error; 1438 1439 error = copyin(sopt->sopt_val, mtod(m, char *), 1440 m->m_len); 1441 if (error != 0) { 1442 m_freem(m0); 1443 return(error); 1444 } 1445 } else 1446 bcopy(sopt->sopt_val, mtod(m, char *), m->m_len); 1447 sopt->sopt_valsize -= m->m_len; 1448 (caddr_t)sopt->sopt_val += m->m_len; 1449 m = m->m_next; 1450 } 1451 if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */ 1452 panic("ip6_sooptmcopyin"); 1453 return 0; 1454} 1455 1456/* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */ 1457int 1458soopt_mcopyout(struct sockopt *sopt, struct mbuf *m) 1459{ 1460 struct mbuf *m0 = m; 1461 size_t valsize = 0; 1462 1463 if (sopt->sopt_val == NULL) 1464 return 0; 1465 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1466 if (sopt->sopt_p != NULL) { 1467 int error; 1468 1469 error = copyout(mtod(m, char *), sopt->sopt_val, 1470 m->m_len); 1471 if (error != 0) { 1472 m_freem(m0); 1473 return(error); 1474 } 1475 } else 1476 bcopy(mtod(m, char *), sopt->sopt_val, m->m_len); 1477 sopt->sopt_valsize -= m->m_len; 1478 (caddr_t)sopt->sopt_val += m->m_len; 1479 valsize += m->m_len; 1480 m = m->m_next; 1481 } 1482 if (m != NULL) { 1483 /* enough soopt buffer should be given from user-land */ 1484 m_freem(m0); 1485 return(EINVAL); 1486 } 1487 sopt->sopt_valsize = valsize; 1488 return 0; 1489} 1490 1491void 1492sohasoutofband(so) 1493 register struct socket *so; 1494{ 1495 if (so->so_sigio != NULL) 1496 pgsigio(so->so_sigio, SIGURG, 0); 1497 selwakeup(&so->so_rcv.sb_sel); 1498} 1499 1500int 1501sopoll(struct socket *so, int events, struct ucred *cred, struct proc *p) 1502{ 1503 int revents = 0; 1504 int s = splnet(); 1505 1506 if (events & (POLLIN | POLLRDNORM)) 1507 if (soreadable(so)) 1508 revents |= events & (POLLIN | POLLRDNORM); 1509 1510 if (events & (POLLOUT | POLLWRNORM)) 1511 if (sowriteable(so)) 1512 revents |= events & (POLLOUT | POLLWRNORM); 1513 1514 if (events & (POLLPRI | POLLRDBAND)) 1515 if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) 1516 revents |= events & (POLLPRI | POLLRDBAND); 1517 1518 if (revents == 0) { 1519 if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) { 1520 selrecord(p, &so->so_rcv.sb_sel); 1521 so->so_rcv.sb_flags |= SB_SEL; 1522 } 1523 1524 if (events & (POLLOUT | POLLWRNORM)) { 1525 selrecord(p, &so->so_snd.sb_sel); 1526 so->so_snd.sb_flags |= SB_SEL; 1527 } 1528 } 1529 1530 splx(s); 1531 return (revents); 1532} 1533 1534int 1535sokqfilter(struct file *fp, struct knote *kn) 1536{ 1537 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1538 struct sockbuf *sb; 1539 int s; 1540 1541 switch (kn->kn_filter) { 1542 case EVFILT_READ: 1543 if (so->so_options & SO_ACCEPTCONN) 1544 kn->kn_fop = &solisten_filtops; 1545 else 1546 kn->kn_fop = &soread_filtops; 1547 sb = &so->so_rcv; 1548 break; 1549 case EVFILT_WRITE: 1550 kn->kn_fop = &sowrite_filtops; 1551 sb = &so->so_snd; 1552 break; 1553 default: 1554 return (1); 1555 } 1556 1557 s = splnet(); 1558 SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext); 1559 sb->sb_flags |= SB_KNOTE; 1560 splx(s); 1561 return (0); 1562} 1563 1564static void 1565filt_sordetach(struct knote *kn) 1566{ 1567 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1568 int s = splnet(); 1569 1570 SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext); 1571 if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note)) 1572 so->so_rcv.sb_flags &= ~SB_KNOTE; 1573 splx(s); 1574} 1575 1576/*ARGSUSED*/ 1577static int 1578filt_soread(struct knote *kn, long hint) 1579{ 1580 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1581 1582 kn->kn_data = so->so_rcv.sb_cc; 1583 if (so->so_state & SS_CANTRCVMORE) { 1584 kn->kn_flags |= EV_EOF; 1585 kn->kn_fflags = so->so_error; 1586 return (1); 1587 } 1588 if (so->so_error) /* temporary udp error */ 1589 return (1); 1590 if (kn->kn_sfflags & NOTE_LOWAT) 1591 return (kn->kn_data >= kn->kn_sdata); 1592 return (kn->kn_data >= so->so_rcv.sb_lowat); 1593} 1594 1595static void 1596filt_sowdetach(struct knote *kn) 1597{ 1598 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1599 int s = splnet(); 1600 1601 SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext); 1602 if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note)) 1603 so->so_snd.sb_flags &= ~SB_KNOTE; 1604 splx(s); 1605} 1606 1607/*ARGSUSED*/ 1608static int 1609filt_sowrite(struct knote *kn, long hint) 1610{ 1611 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1612 1613 kn->kn_data = sbspace(&so->so_snd); 1614 if (so->so_state & SS_CANTSENDMORE) { 1615 kn->kn_flags |= EV_EOF; 1616 kn->kn_fflags = so->so_error; 1617 return (1); 1618 } 1619 if (so->so_error) /* temporary udp error */ 1620 return (1); 1621 if (((so->so_state & SS_ISCONNECTED) == 0) && 1622 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 1623 return (0); 1624 if (kn->kn_sfflags & NOTE_LOWAT) 1625 return (kn->kn_data >= kn->kn_sdata); 1626 return (kn->kn_data >= so->so_snd.sb_lowat); 1627} 1628 1629/*ARGSUSED*/ 1630static int 1631filt_solisten(struct knote *kn, long hint) 1632{ 1633 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1634 1635 kn->kn_data = so->so_qlen - so->so_incqlen; 1636 return (! TAILQ_EMPTY(&so->so_comp)); 1637} 1638