uipc_socket.c revision 74018
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 74018 2001-03-09 08:16:40Z 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 error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam); 364 splx(s); 365 return (error); 366} 367 368int 369soconnect(so, nam, p) 370 register struct socket *so; 371 struct sockaddr *nam; 372 struct proc *p; 373{ 374 int s; 375 int error; 376 377 if (so->so_options & SO_ACCEPTCONN) 378 return (EOPNOTSUPP); 379 s = splnet(); 380 /* 381 * If protocol is connection-based, can only connect once. 382 * Otherwise, if connected, try to disconnect first. 383 * This allows user to disconnect by connecting to, e.g., 384 * a null address. 385 */ 386 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 387 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 388 (error = sodisconnect(so)))) 389 error = EISCONN; 390 else 391 error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, p); 392 splx(s); 393 return (error); 394} 395 396int 397soconnect2(so1, so2) 398 register struct socket *so1; 399 struct socket *so2; 400{ 401 int s = splnet(); 402 int error; 403 404 error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2); 405 splx(s); 406 return (error); 407} 408 409int 410sodisconnect(so) 411 register struct socket *so; 412{ 413 int s = splnet(); 414 int error; 415 416 if ((so->so_state & SS_ISCONNECTED) == 0) { 417 error = ENOTCONN; 418 goto bad; 419 } 420 if (so->so_state & SS_ISDISCONNECTING) { 421 error = EALREADY; 422 goto bad; 423 } 424 error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so); 425bad: 426 splx(s); 427 return (error); 428} 429 430#define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 431/* 432 * Send on a socket. 433 * If send must go all at once and message is larger than 434 * send buffering, then hard error. 435 * Lock against other senders. 436 * If must go all at once and not enough room now, then 437 * inform user that this would block and do nothing. 438 * Otherwise, if nonblocking, send as much as possible. 439 * The data to be sent is described by "uio" if nonzero, 440 * otherwise by the mbuf chain "top" (which must be null 441 * if uio is not). Data provided in mbuf chain must be small 442 * enough to send all at once. 443 * 444 * Returns nonzero on error, timeout or signal; callers 445 * must check for short counts if EINTR/ERESTART are returned. 446 * Data and control buffers are freed on return. 447 */ 448int 449sosend(so, addr, uio, top, control, flags, p) 450 register struct socket *so; 451 struct sockaddr *addr; 452 struct uio *uio; 453 struct mbuf *top; 454 struct mbuf *control; 455 int flags; 456 struct proc *p; 457{ 458 struct mbuf **mp; 459 register struct mbuf *m; 460 register long space, len, resid; 461 int clen = 0, error, s, dontroute, mlen; 462 int atomic = sosendallatonce(so) || top; 463 464 if (uio) 465 resid = uio->uio_resid; 466 else 467 resid = top->m_pkthdr.len; 468 /* 469 * In theory resid should be unsigned. 470 * However, space must be signed, as it might be less than 0 471 * if we over-committed, and we must use a signed comparison 472 * of space and resid. On the other hand, a negative resid 473 * causes us to loop sending 0-length segments to the protocol. 474 * 475 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM 476 * type sockets since that's an error. 477 */ 478 if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) { 479 error = EINVAL; 480 goto out; 481 } 482 483 dontroute = 484 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 485 (so->so_proto->pr_flags & PR_ATOMIC); 486 if (p) 487 p->p_stats->p_ru.ru_msgsnd++; 488 if (control) 489 clen = control->m_len; 490#define snderr(errno) { error = errno; splx(s); goto release; } 491 492restart: 493 error = sblock(&so->so_snd, SBLOCKWAIT(flags)); 494 if (error) 495 goto out; 496 do { 497 s = splnet(); 498 if (so->so_state & SS_CANTSENDMORE) 499 snderr(EPIPE); 500 if (so->so_error) { 501 error = so->so_error; 502 so->so_error = 0; 503 splx(s); 504 goto release; 505 } 506 if ((so->so_state & SS_ISCONNECTED) == 0) { 507 /* 508 * `sendto' and `sendmsg' is allowed on a connection- 509 * based socket if it supports implied connect. 510 * Return ENOTCONN if not connected and no address is 511 * supplied. 512 */ 513 if ((so->so_proto->pr_flags & PR_CONNREQUIRED) && 514 (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) { 515 if ((so->so_state & SS_ISCONFIRMING) == 0 && 516 !(resid == 0 && clen != 0)) 517 snderr(ENOTCONN); 518 } else if (addr == 0) 519 snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ? 520 ENOTCONN : EDESTADDRREQ); 521 } 522 space = sbspace(&so->so_snd); 523 if (flags & MSG_OOB) 524 space += 1024; 525 if ((atomic && resid > so->so_snd.sb_hiwat) || 526 clen > so->so_snd.sb_hiwat) 527 snderr(EMSGSIZE); 528 if (space < resid + clen && uio && 529 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 530 if (so->so_state & SS_NBIO) 531 snderr(EWOULDBLOCK); 532 sbunlock(&so->so_snd); 533 error = sbwait(&so->so_snd); 534 splx(s); 535 if (error) 536 goto out; 537 goto restart; 538 } 539 splx(s); 540 mp = ⊤ 541 space -= clen; 542 do { 543 if (uio == NULL) { 544 /* 545 * Data is prepackaged in "top". 546 */ 547 resid = 0; 548 if (flags & MSG_EOR) 549 top->m_flags |= M_EOR; 550 } else do { 551 if (top == 0) { 552 MGETHDR(m, M_TRYWAIT, MT_DATA); 553 if (m == NULL) { 554 error = ENOBUFS; 555 goto release; 556 } 557 mlen = MHLEN; 558 m->m_pkthdr.len = 0; 559 m->m_pkthdr.rcvif = (struct ifnet *)0; 560 } else { 561 MGET(m, M_TRYWAIT, MT_DATA); 562 if (m == NULL) { 563 error = ENOBUFS; 564 goto release; 565 } 566 mlen = MLEN; 567 } 568 if (resid >= MINCLSIZE) { 569 MCLGET(m, M_TRYWAIT); 570 if ((m->m_flags & M_EXT) == 0) 571 goto nopages; 572 mlen = MCLBYTES; 573 len = min(min(mlen, resid), space); 574 } else { 575nopages: 576 len = min(min(mlen, resid), space); 577 /* 578 * For datagram protocols, leave room 579 * for protocol headers in first mbuf. 580 */ 581 if (atomic && top == 0 && len < mlen) 582 MH_ALIGN(m, len); 583 } 584 space -= len; 585 error = uiomove(mtod(m, caddr_t), (int)len, uio); 586 resid = uio->uio_resid; 587 m->m_len = len; 588 *mp = m; 589 top->m_pkthdr.len += len; 590 if (error) 591 goto release; 592 mp = &m->m_next; 593 if (resid <= 0) { 594 if (flags & MSG_EOR) 595 top->m_flags |= M_EOR; 596 break; 597 } 598 } while (space > 0 && atomic); 599 if (dontroute) 600 so->so_options |= SO_DONTROUTE; 601 s = splnet(); /* XXX */ 602 /* 603 * XXX all the SS_CANTSENDMORE checks previously 604 * done could be out of date. We could have recieved 605 * a reset packet in an interrupt or maybe we slept 606 * while doing page faults in uiomove() etc. We could 607 * probably recheck again inside the splnet() protection 608 * here, but there are probably other places that this 609 * also happens. We must rethink this. 610 */ 611 error = (*so->so_proto->pr_usrreqs->pru_send)(so, 612 (flags & MSG_OOB) ? PRUS_OOB : 613 /* 614 * If the user set MSG_EOF, the protocol 615 * understands this flag and nothing left to 616 * send then use PRU_SEND_EOF instead of PRU_SEND. 617 */ 618 ((flags & MSG_EOF) && 619 (so->so_proto->pr_flags & PR_IMPLOPCL) && 620 (resid <= 0)) ? 621 PRUS_EOF : 622 /* If there is more to send set PRUS_MORETOCOME */ 623 (resid > 0 && space > 0) ? PRUS_MORETOCOME : 0, 624 top, addr, control, p); 625 splx(s); 626 if (dontroute) 627 so->so_options &= ~SO_DONTROUTE; 628 clen = 0; 629 control = 0; 630 top = 0; 631 mp = ⊤ 632 if (error) 633 goto release; 634 } while (resid && space > 0); 635 } while (resid); 636 637release: 638 sbunlock(&so->so_snd); 639out: 640 if (top) 641 m_freem(top); 642 if (control) 643 m_freem(control); 644 return (error); 645} 646 647/* 648 * Implement receive operations on a socket. 649 * We depend on the way that records are added to the sockbuf 650 * by sbappend*. In particular, each record (mbufs linked through m_next) 651 * must begin with an address if the protocol so specifies, 652 * followed by an optional mbuf or mbufs containing ancillary data, 653 * and then zero or more mbufs of data. 654 * In order to avoid blocking network interrupts for the entire time here, 655 * we splx() while doing the actual copy to user space. 656 * Although the sockbuf is locked, new data may still be appended, 657 * and thus we must maintain consistency of the sockbuf during that time. 658 * 659 * The caller may receive the data as a single mbuf chain by supplying 660 * an mbuf **mp0 for use in returning the chain. The uio is then used 661 * only for the count in uio_resid. 662 */ 663int 664soreceive(so, psa, uio, mp0, controlp, flagsp) 665 register struct socket *so; 666 struct sockaddr **psa; 667 struct uio *uio; 668 struct mbuf **mp0; 669 struct mbuf **controlp; 670 int *flagsp; 671{ 672 register struct mbuf *m, **mp; 673 register int flags, len, error, s, offset; 674 struct protosw *pr = so->so_proto; 675 struct mbuf *nextrecord; 676 int moff, type = 0; 677 int orig_resid = uio->uio_resid; 678 679 mp = mp0; 680 if (psa) 681 *psa = 0; 682 if (controlp) 683 *controlp = 0; 684 if (flagsp) 685 flags = *flagsp &~ MSG_EOR; 686 else 687 flags = 0; 688 if (flags & MSG_OOB) { 689 m = m_get(M_TRYWAIT, MT_DATA); 690 if (m == NULL) 691 return (ENOBUFS); 692 error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK); 693 if (error) 694 goto bad; 695 do { 696 error = uiomove(mtod(m, caddr_t), 697 (int) min(uio->uio_resid, m->m_len), uio); 698 m = m_free(m); 699 } while (uio->uio_resid && error == 0 && m); 700bad: 701 if (m) 702 m_freem(m); 703 return (error); 704 } 705 if (mp) 706 *mp = (struct mbuf *)0; 707 if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) 708 (*pr->pr_usrreqs->pru_rcvd)(so, 0); 709 710restart: 711 error = sblock(&so->so_rcv, SBLOCKWAIT(flags)); 712 if (error) 713 return (error); 714 s = splnet(); 715 716 m = so->so_rcv.sb_mb; 717 /* 718 * If we have less data than requested, block awaiting more 719 * (subject to any timeout) if: 720 * 1. the current count is less than the low water mark, or 721 * 2. MSG_WAITALL is set, and it is possible to do the entire 722 * receive operation at once if we block (resid <= hiwat). 723 * 3. MSG_DONTWAIT is not set 724 * If MSG_WAITALL is set but resid is larger than the receive buffer, 725 * we have to do the receive in sections, and thus risk returning 726 * a short count if a timeout or signal occurs after we start. 727 */ 728 if (m == 0 || (((flags & MSG_DONTWAIT) == 0 && 729 so->so_rcv.sb_cc < uio->uio_resid) && 730 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 731 ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) && 732 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) { 733 KASSERT(m != 0 || !so->so_rcv.sb_cc, ("receive 1")); 734 if (so->so_error) { 735 if (m) 736 goto dontblock; 737 error = so->so_error; 738 if ((flags & MSG_PEEK) == 0) 739 so->so_error = 0; 740 goto release; 741 } 742 if (so->so_state & SS_CANTRCVMORE) { 743 if (m) 744 goto dontblock; 745 else 746 goto release; 747 } 748 for (; m; m = m->m_next) 749 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 750 m = so->so_rcv.sb_mb; 751 goto dontblock; 752 } 753 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 754 (so->so_proto->pr_flags & PR_CONNREQUIRED)) { 755 error = ENOTCONN; 756 goto release; 757 } 758 if (uio->uio_resid == 0) 759 goto release; 760 if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) { 761 error = EWOULDBLOCK; 762 goto release; 763 } 764 sbunlock(&so->so_rcv); 765 error = sbwait(&so->so_rcv); 766 splx(s); 767 if (error) 768 return (error); 769 goto restart; 770 } 771dontblock: 772 if (uio->uio_procp) 773 uio->uio_procp->p_stats->p_ru.ru_msgrcv++; 774 nextrecord = m->m_nextpkt; 775 if (pr->pr_flags & PR_ADDR) { 776 KASSERT(m->m_type == MT_SONAME, ("receive 1a")); 777 orig_resid = 0; 778 if (psa) 779 *psa = dup_sockaddr(mtod(m, struct sockaddr *), 780 mp0 == 0); 781 if (flags & MSG_PEEK) { 782 m = m->m_next; 783 } else { 784 sbfree(&so->so_rcv, m); 785 MFREE(m, so->so_rcv.sb_mb); 786 m = so->so_rcv.sb_mb; 787 } 788 } 789 while (m && m->m_type == MT_CONTROL && error == 0) { 790 if (flags & MSG_PEEK) { 791 if (controlp) 792 *controlp = m_copy(m, 0, m->m_len); 793 m = m->m_next; 794 } else { 795 sbfree(&so->so_rcv, m); 796 if (controlp) { 797 if (pr->pr_domain->dom_externalize && 798 mtod(m, struct cmsghdr *)->cmsg_type == 799 SCM_RIGHTS) 800 error = (*pr->pr_domain->dom_externalize)(m); 801 *controlp = m; 802 so->so_rcv.sb_mb = m->m_next; 803 m->m_next = 0; 804 m = so->so_rcv.sb_mb; 805 } else { 806 MFREE(m, so->so_rcv.sb_mb); 807 m = so->so_rcv.sb_mb; 808 } 809 } 810 if (controlp) { 811 orig_resid = 0; 812 controlp = &(*controlp)->m_next; 813 } 814 } 815 if (m) { 816 if ((flags & MSG_PEEK) == 0) 817 m->m_nextpkt = nextrecord; 818 type = m->m_type; 819 if (type == MT_OOBDATA) 820 flags |= MSG_OOB; 821 } 822 moff = 0; 823 offset = 0; 824 while (m && uio->uio_resid > 0 && error == 0) { 825 if (m->m_type == MT_OOBDATA) { 826 if (type != MT_OOBDATA) 827 break; 828 } else if (type == MT_OOBDATA) 829 break; 830 else 831 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER, 832 ("receive 3")); 833 so->so_state &= ~SS_RCVATMARK; 834 len = uio->uio_resid; 835 if (so->so_oobmark && len > so->so_oobmark - offset) 836 len = so->so_oobmark - offset; 837 if (len > m->m_len - moff) 838 len = m->m_len - moff; 839 /* 840 * If mp is set, just pass back the mbufs. 841 * Otherwise copy them out via the uio, then free. 842 * Sockbuf must be consistent here (points to current mbuf, 843 * it points to next record) when we drop priority; 844 * we must note any additions to the sockbuf when we 845 * block interrupts again. 846 */ 847 if (mp == 0) { 848 splx(s); 849 error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio); 850 s = splnet(); 851 if (error) 852 goto release; 853 } else 854 uio->uio_resid -= len; 855 if (len == m->m_len - moff) { 856 if (m->m_flags & M_EOR) 857 flags |= MSG_EOR; 858 if (flags & MSG_PEEK) { 859 m = m->m_next; 860 moff = 0; 861 } else { 862 nextrecord = m->m_nextpkt; 863 sbfree(&so->so_rcv, m); 864 if (mp) { 865 *mp = m; 866 mp = &m->m_next; 867 so->so_rcv.sb_mb = m = m->m_next; 868 *mp = (struct mbuf *)0; 869 } else { 870 MFREE(m, so->so_rcv.sb_mb); 871 m = so->so_rcv.sb_mb; 872 } 873 if (m) 874 m->m_nextpkt = nextrecord; 875 } 876 } else { 877 if (flags & MSG_PEEK) 878 moff += len; 879 else { 880 if (mp) 881 *mp = m_copym(m, 0, len, M_TRYWAIT); 882 m->m_data += len; 883 m->m_len -= len; 884 so->so_rcv.sb_cc -= len; 885 } 886 } 887 if (so->so_oobmark) { 888 if ((flags & MSG_PEEK) == 0) { 889 so->so_oobmark -= len; 890 if (so->so_oobmark == 0) { 891 so->so_state |= SS_RCVATMARK; 892 break; 893 } 894 } else { 895 offset += len; 896 if (offset == so->so_oobmark) 897 break; 898 } 899 } 900 if (flags & MSG_EOR) 901 break; 902 /* 903 * If the MSG_WAITALL flag is set (for non-atomic socket), 904 * we must not quit until "uio->uio_resid == 0" or an error 905 * termination. If a signal/timeout occurs, return 906 * with a short count but without error. 907 * Keep sockbuf locked against other readers. 908 */ 909 while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 && 910 !sosendallatonce(so) && !nextrecord) { 911 if (so->so_error || so->so_state & SS_CANTRCVMORE) 912 break; 913 error = sbwait(&so->so_rcv); 914 if (error) { 915 sbunlock(&so->so_rcv); 916 splx(s); 917 return (0); 918 } 919 m = so->so_rcv.sb_mb; 920 if (m) 921 nextrecord = m->m_nextpkt; 922 } 923 } 924 925 if (m && pr->pr_flags & PR_ATOMIC) { 926 flags |= MSG_TRUNC; 927 if ((flags & MSG_PEEK) == 0) 928 (void) sbdroprecord(&so->so_rcv); 929 } 930 if ((flags & MSG_PEEK) == 0) { 931 if (m == 0) 932 so->so_rcv.sb_mb = nextrecord; 933 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 934 (*pr->pr_usrreqs->pru_rcvd)(so, flags); 935 } 936 if (orig_resid == uio->uio_resid && orig_resid && 937 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 938 sbunlock(&so->so_rcv); 939 splx(s); 940 goto restart; 941 } 942 943 if (flagsp) 944 *flagsp |= flags; 945release: 946 sbunlock(&so->so_rcv); 947 splx(s); 948 return (error); 949} 950 951int 952soshutdown(so, how) 953 register struct socket *so; 954 register int how; 955{ 956 register struct protosw *pr = so->so_proto; 957 958 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 959 return (EINVAL); 960 961 if (how != SHUT_WR) 962 sorflush(so); 963 if (how != SHUT_RD) 964 return ((*pr->pr_usrreqs->pru_shutdown)(so)); 965 return (0); 966} 967 968void 969sorflush(so) 970 register struct socket *so; 971{ 972 register struct sockbuf *sb = &so->so_rcv; 973 register struct protosw *pr = so->so_proto; 974 register int s; 975 struct sockbuf asb; 976 977 sb->sb_flags |= SB_NOINTR; 978 (void) sblock(sb, M_WAITOK); 979 s = splimp(); 980 socantrcvmore(so); 981 sbunlock(sb); 982 asb = *sb; 983 bzero((caddr_t)sb, sizeof (*sb)); 984 splx(s); 985 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) 986 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 987 sbrelease(&asb, so); 988} 989 990#ifdef INET 991static int 992do_setopt_accept_filter(so, sopt) 993 struct socket *so; 994 struct sockopt *sopt; 995{ 996 struct accept_filter_arg *afap = NULL; 997 struct accept_filter *afp; 998 struct so_accf *af = so->so_accf; 999 int error = 0; 1000 1001 /* do not set/remove accept filters on non listen sockets */ 1002 if ((so->so_options & SO_ACCEPTCONN) == 0) { 1003 error = EINVAL; 1004 goto out; 1005 } 1006 1007 /* removing the filter */ 1008 if (sopt == NULL) { 1009 if (af != NULL) { 1010 if (af->so_accept_filter != NULL && 1011 af->so_accept_filter->accf_destroy != NULL) { 1012 af->so_accept_filter->accf_destroy(so); 1013 } 1014 if (af->so_accept_filter_str != NULL) { 1015 FREE(af->so_accept_filter_str, M_ACCF); 1016 } 1017 FREE(af, M_ACCF); 1018 so->so_accf = NULL; 1019 } 1020 so->so_options &= ~SO_ACCEPTFILTER; 1021 return (0); 1022 } 1023 /* adding a filter */ 1024 /* must remove previous filter first */ 1025 if (af != NULL) { 1026 error = EINVAL; 1027 goto out; 1028 } 1029 /* don't put large objects on the kernel stack */ 1030 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK); 1031 error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap); 1032 afap->af_name[sizeof(afap->af_name)-1] = '\0'; 1033 afap->af_arg[sizeof(afap->af_arg)-1] = '\0'; 1034 if (error) 1035 goto out; 1036 afp = accept_filt_get(afap->af_name); 1037 if (afp == NULL) { 1038 error = ENOENT; 1039 goto out; 1040 } 1041 MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK | M_ZERO); 1042 if (afp->accf_create != NULL) { 1043 if (afap->af_name[0] != '\0') { 1044 int len = strlen(afap->af_name) + 1; 1045 1046 MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK); 1047 strcpy(af->so_accept_filter_str, afap->af_name); 1048 } 1049 af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg); 1050 if (af->so_accept_filter_arg == NULL) { 1051 FREE(af->so_accept_filter_str, M_ACCF); 1052 FREE(af, M_ACCF); 1053 so->so_accf = NULL; 1054 error = EINVAL; 1055 goto out; 1056 } 1057 } 1058 af->so_accept_filter = afp; 1059 so->so_accf = af; 1060 so->so_options |= SO_ACCEPTFILTER; 1061out: 1062 if (afap != NULL) 1063 FREE(afap, M_TEMP); 1064 return (error); 1065} 1066#endif /* INET */ 1067 1068/* 1069 * Perhaps this routine, and sooptcopyout(), below, ought to come in 1070 * an additional variant to handle the case where the option value needs 1071 * to be some kind of integer, but not a specific size. 1072 * In addition to their use here, these functions are also called by the 1073 * protocol-level pr_ctloutput() routines. 1074 */ 1075int 1076sooptcopyin(sopt, buf, len, minlen) 1077 struct sockopt *sopt; 1078 void *buf; 1079 size_t len; 1080 size_t minlen; 1081{ 1082 size_t valsize; 1083 1084 /* 1085 * If the user gives us more than we wanted, we ignore it, 1086 * but if we don't get the minimum length the caller 1087 * wants, we return EINVAL. On success, sopt->sopt_valsize 1088 * is set to however much we actually retrieved. 1089 */ 1090 if ((valsize = sopt->sopt_valsize) < minlen) 1091 return EINVAL; 1092 if (valsize > len) 1093 sopt->sopt_valsize = valsize = len; 1094 1095 if (sopt->sopt_p != 0) 1096 return (copyin(sopt->sopt_val, buf, valsize)); 1097 1098 bcopy(sopt->sopt_val, buf, valsize); 1099 return 0; 1100} 1101 1102int 1103sosetopt(so, sopt) 1104 struct socket *so; 1105 struct sockopt *sopt; 1106{ 1107 int error, optval; 1108 struct linger l; 1109 struct timeval tv; 1110 u_long val; 1111 1112 error = 0; 1113 if (sopt->sopt_level != SOL_SOCKET) { 1114 if (so->so_proto && so->so_proto->pr_ctloutput) 1115 return ((*so->so_proto->pr_ctloutput) 1116 (so, sopt)); 1117 error = ENOPROTOOPT; 1118 } else { 1119 switch (sopt->sopt_name) { 1120#ifdef INET 1121 case SO_ACCEPTFILTER: 1122 error = do_setopt_accept_filter(so, sopt); 1123 if (error) 1124 goto bad; 1125 break; 1126#endif 1127 case SO_LINGER: 1128 error = sooptcopyin(sopt, &l, sizeof l, sizeof l); 1129 if (error) 1130 goto bad; 1131 1132 so->so_linger = l.l_linger; 1133 if (l.l_onoff) 1134 so->so_options |= SO_LINGER; 1135 else 1136 so->so_options &= ~SO_LINGER; 1137 break; 1138 1139 case SO_DEBUG: 1140 case SO_KEEPALIVE: 1141 case SO_DONTROUTE: 1142 case SO_USELOOPBACK: 1143 case SO_BROADCAST: 1144 case SO_REUSEADDR: 1145 case SO_REUSEPORT: 1146 case SO_OOBINLINE: 1147 case SO_TIMESTAMP: 1148 error = sooptcopyin(sopt, &optval, sizeof optval, 1149 sizeof optval); 1150 if (error) 1151 goto bad; 1152 if (optval) 1153 so->so_options |= sopt->sopt_name; 1154 else 1155 so->so_options &= ~sopt->sopt_name; 1156 break; 1157 1158 case SO_SNDBUF: 1159 case SO_RCVBUF: 1160 case SO_SNDLOWAT: 1161 case SO_RCVLOWAT: 1162 error = sooptcopyin(sopt, &optval, sizeof optval, 1163 sizeof optval); 1164 if (error) 1165 goto bad; 1166 1167 /* 1168 * Values < 1 make no sense for any of these 1169 * options, so disallow them. 1170 */ 1171 if (optval < 1) { 1172 error = EINVAL; 1173 goto bad; 1174 } 1175 1176 switch (sopt->sopt_name) { 1177 case SO_SNDBUF: 1178 case SO_RCVBUF: 1179 if (sbreserve(sopt->sopt_name == SO_SNDBUF ? 1180 &so->so_snd : &so->so_rcv, (u_long)optval, 1181 so, curproc) == 0) { 1182 error = ENOBUFS; 1183 goto bad; 1184 } 1185 break; 1186 1187 /* 1188 * Make sure the low-water is never greater than 1189 * the high-water. 1190 */ 1191 case SO_SNDLOWAT: 1192 so->so_snd.sb_lowat = 1193 (optval > so->so_snd.sb_hiwat) ? 1194 so->so_snd.sb_hiwat : optval; 1195 break; 1196 case SO_RCVLOWAT: 1197 so->so_rcv.sb_lowat = 1198 (optval > so->so_rcv.sb_hiwat) ? 1199 so->so_rcv.sb_hiwat : optval; 1200 break; 1201 } 1202 break; 1203 1204 case SO_SNDTIMEO: 1205 case SO_RCVTIMEO: 1206 error = sooptcopyin(sopt, &tv, sizeof tv, 1207 sizeof tv); 1208 if (error) 1209 goto bad; 1210 1211 /* assert(hz > 0); */ 1212 if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz || 1213 tv.tv_usec < 0 || tv.tv_usec >= 1000000) { 1214 error = EDOM; 1215 goto bad; 1216 } 1217 /* assert(tick > 0); */ 1218 /* assert(ULONG_MAX - SHRT_MAX >= 1000000); */ 1219 val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick; 1220 if (val > SHRT_MAX) { 1221 error = EDOM; 1222 goto bad; 1223 } 1224 1225 switch (sopt->sopt_name) { 1226 case SO_SNDTIMEO: 1227 so->so_snd.sb_timeo = val; 1228 break; 1229 case SO_RCVTIMEO: 1230 so->so_rcv.sb_timeo = val; 1231 break; 1232 } 1233 break; 1234 default: 1235 error = ENOPROTOOPT; 1236 break; 1237 } 1238 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) { 1239 (void) ((*so->so_proto->pr_ctloutput) 1240 (so, sopt)); 1241 } 1242 } 1243bad: 1244 return (error); 1245} 1246 1247/* Helper routine for getsockopt */ 1248int 1249sooptcopyout(sopt, buf, len) 1250 struct sockopt *sopt; 1251 void *buf; 1252 size_t len; 1253{ 1254 int error; 1255 size_t valsize; 1256 1257 error = 0; 1258 1259 /* 1260 * Documented get behavior is that we always return a value, 1261 * possibly truncated to fit in the user's buffer. 1262 * Traditional behavior is that we always tell the user 1263 * precisely how much we copied, rather than something useful 1264 * like the total amount we had available for her. 1265 * Note that this interface is not idempotent; the entire answer must 1266 * generated ahead of time. 1267 */ 1268 valsize = min(len, sopt->sopt_valsize); 1269 sopt->sopt_valsize = valsize; 1270 if (sopt->sopt_val != 0) { 1271 if (sopt->sopt_p != 0) 1272 error = copyout(buf, sopt->sopt_val, valsize); 1273 else 1274 bcopy(buf, sopt->sopt_val, valsize); 1275 } 1276 return error; 1277} 1278 1279int 1280sogetopt(so, sopt) 1281 struct socket *so; 1282 struct sockopt *sopt; 1283{ 1284 int error, optval; 1285 struct linger l; 1286 struct timeval tv; 1287#ifdef INET 1288 struct accept_filter_arg *afap; 1289#endif 1290 1291 error = 0; 1292 if (sopt->sopt_level != SOL_SOCKET) { 1293 if (so->so_proto && so->so_proto->pr_ctloutput) { 1294 return ((*so->so_proto->pr_ctloutput) 1295 (so, sopt)); 1296 } else 1297 return (ENOPROTOOPT); 1298 } else { 1299 switch (sopt->sopt_name) { 1300#ifdef INET 1301 case SO_ACCEPTFILTER: 1302 if ((so->so_options & SO_ACCEPTCONN) == 0) 1303 return (EINVAL); 1304 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), 1305 M_TEMP, M_WAITOK | M_ZERO); 1306 if ((so->so_options & SO_ACCEPTFILTER) != 0) { 1307 strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name); 1308 if (so->so_accf->so_accept_filter_str != NULL) 1309 strcpy(afap->af_arg, so->so_accf->so_accept_filter_str); 1310 } 1311 error = sooptcopyout(sopt, afap, sizeof(*afap)); 1312 FREE(afap, M_TEMP); 1313 break; 1314#endif 1315 1316 case SO_LINGER: 1317 l.l_onoff = so->so_options & SO_LINGER; 1318 l.l_linger = so->so_linger; 1319 error = sooptcopyout(sopt, &l, sizeof l); 1320 break; 1321 1322 case SO_USELOOPBACK: 1323 case SO_DONTROUTE: 1324 case SO_DEBUG: 1325 case SO_KEEPALIVE: 1326 case SO_REUSEADDR: 1327 case SO_REUSEPORT: 1328 case SO_BROADCAST: 1329 case SO_OOBINLINE: 1330 case SO_TIMESTAMP: 1331 optval = so->so_options & sopt->sopt_name; 1332integer: 1333 error = sooptcopyout(sopt, &optval, sizeof optval); 1334 break; 1335 1336 case SO_TYPE: 1337 optval = so->so_type; 1338 goto integer; 1339 1340 case SO_ERROR: 1341 optval = so->so_error; 1342 so->so_error = 0; 1343 goto integer; 1344 1345 case SO_SNDBUF: 1346 optval = so->so_snd.sb_hiwat; 1347 goto integer; 1348 1349 case SO_RCVBUF: 1350 optval = so->so_rcv.sb_hiwat; 1351 goto integer; 1352 1353 case SO_SNDLOWAT: 1354 optval = so->so_snd.sb_lowat; 1355 goto integer; 1356 1357 case SO_RCVLOWAT: 1358 optval = so->so_rcv.sb_lowat; 1359 goto integer; 1360 1361 case SO_SNDTIMEO: 1362 case SO_RCVTIMEO: 1363 optval = (sopt->sopt_name == SO_SNDTIMEO ? 1364 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 1365 1366 tv.tv_sec = optval / hz; 1367 tv.tv_usec = (optval % hz) * tick; 1368 error = sooptcopyout(sopt, &tv, sizeof tv); 1369 break; 1370 1371 default: 1372 error = ENOPROTOOPT; 1373 break; 1374 } 1375 return (error); 1376 } 1377} 1378 1379/* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */ 1380int 1381soopt_getm(struct sockopt *sopt, struct mbuf **mp) 1382{ 1383 struct mbuf *m, *m_prev; 1384 int sopt_size = sopt->sopt_valsize; 1385 1386 MGET(m, sopt->sopt_p ? M_TRYWAIT : M_DONTWAIT, MT_DATA); 1387 if (m == 0) 1388 return ENOBUFS; 1389 if (sopt_size > MLEN) { 1390 MCLGET(m, sopt->sopt_p ? M_TRYWAIT : M_DONTWAIT); 1391 if ((m->m_flags & M_EXT) == 0) { 1392 m_free(m); 1393 return ENOBUFS; 1394 } 1395 m->m_len = min(MCLBYTES, sopt_size); 1396 } else { 1397 m->m_len = min(MLEN, sopt_size); 1398 } 1399 sopt_size -= m->m_len; 1400 *mp = m; 1401 m_prev = m; 1402 1403 while (sopt_size) { 1404 MGET(m, sopt->sopt_p ? M_TRYWAIT : M_DONTWAIT, MT_DATA); 1405 if (m == 0) { 1406 m_freem(*mp); 1407 return ENOBUFS; 1408 } 1409 if (sopt_size > MLEN) { 1410 MCLGET(m, sopt->sopt_p ? M_TRYWAIT : M_DONTWAIT); 1411 if ((m->m_flags & M_EXT) == 0) { 1412 m_freem(*mp); 1413 return ENOBUFS; 1414 } 1415 m->m_len = min(MCLBYTES, sopt_size); 1416 } else { 1417 m->m_len = min(MLEN, sopt_size); 1418 } 1419 sopt_size -= m->m_len; 1420 m_prev->m_next = m; 1421 m_prev = m; 1422 } 1423 return 0; 1424} 1425 1426/* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */ 1427int 1428soopt_mcopyin(struct sockopt *sopt, struct mbuf *m) 1429{ 1430 struct mbuf *m0 = m; 1431 1432 if (sopt->sopt_val == NULL) 1433 return 0; 1434 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1435 if (sopt->sopt_p != NULL) { 1436 int error; 1437 1438 error = copyin(sopt->sopt_val, mtod(m, char *), 1439 m->m_len); 1440 if (error != 0) { 1441 m_freem(m0); 1442 return(error); 1443 } 1444 } else 1445 bcopy(sopt->sopt_val, mtod(m, char *), m->m_len); 1446 sopt->sopt_valsize -= m->m_len; 1447 (caddr_t)sopt->sopt_val += m->m_len; 1448 m = m->m_next; 1449 } 1450 if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */ 1451 panic("ip6_sooptmcopyin"); 1452 return 0; 1453} 1454 1455/* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */ 1456int 1457soopt_mcopyout(struct sockopt *sopt, struct mbuf *m) 1458{ 1459 struct mbuf *m0 = m; 1460 size_t valsize = 0; 1461 1462 if (sopt->sopt_val == NULL) 1463 return 0; 1464 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1465 if (sopt->sopt_p != NULL) { 1466 int error; 1467 1468 error = copyout(mtod(m, char *), sopt->sopt_val, 1469 m->m_len); 1470 if (error != 0) { 1471 m_freem(m0); 1472 return(error); 1473 } 1474 } else 1475 bcopy(mtod(m, char *), sopt->sopt_val, m->m_len); 1476 sopt->sopt_valsize -= m->m_len; 1477 (caddr_t)sopt->sopt_val += m->m_len; 1478 valsize += m->m_len; 1479 m = m->m_next; 1480 } 1481 if (m != NULL) { 1482 /* enough soopt buffer should be given from user-land */ 1483 m_freem(m0); 1484 return(EINVAL); 1485 } 1486 sopt->sopt_valsize = valsize; 1487 return 0; 1488} 1489 1490void 1491sohasoutofband(so) 1492 register struct socket *so; 1493{ 1494 if (so->so_sigio != NULL) 1495 pgsigio(so->so_sigio, SIGURG, 0); 1496 selwakeup(&so->so_rcv.sb_sel); 1497} 1498 1499int 1500sopoll(struct socket *so, int events, struct ucred *cred, struct proc *p) 1501{ 1502 int revents = 0; 1503 int s = splnet(); 1504 1505 if (events & (POLLIN | POLLRDNORM)) 1506 if (soreadable(so)) 1507 revents |= events & (POLLIN | POLLRDNORM); 1508 1509 if (events & (POLLOUT | POLLWRNORM)) 1510 if (sowriteable(so)) 1511 revents |= events & (POLLOUT | POLLWRNORM); 1512 1513 if (events & (POLLPRI | POLLRDBAND)) 1514 if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) 1515 revents |= events & (POLLPRI | POLLRDBAND); 1516 1517 if (revents == 0) { 1518 if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) { 1519 selrecord(p, &so->so_rcv.sb_sel); 1520 so->so_rcv.sb_flags |= SB_SEL; 1521 } 1522 1523 if (events & (POLLOUT | POLLWRNORM)) { 1524 selrecord(p, &so->so_snd.sb_sel); 1525 so->so_snd.sb_flags |= SB_SEL; 1526 } 1527 } 1528 1529 splx(s); 1530 return (revents); 1531} 1532 1533int 1534sokqfilter(struct file *fp, struct knote *kn) 1535{ 1536 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1537 struct sockbuf *sb; 1538 int s; 1539 1540 switch (kn->kn_filter) { 1541 case EVFILT_READ: 1542 if (so->so_options & SO_ACCEPTCONN) 1543 kn->kn_fop = &solisten_filtops; 1544 else 1545 kn->kn_fop = &soread_filtops; 1546 sb = &so->so_rcv; 1547 break; 1548 case EVFILT_WRITE: 1549 kn->kn_fop = &sowrite_filtops; 1550 sb = &so->so_snd; 1551 break; 1552 default: 1553 return (1); 1554 } 1555 1556 s = splnet(); 1557 SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext); 1558 sb->sb_flags |= SB_KNOTE; 1559 splx(s); 1560 return (0); 1561} 1562 1563static void 1564filt_sordetach(struct knote *kn) 1565{ 1566 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1567 int s = splnet(); 1568 1569 SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext); 1570 if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note)) 1571 so->so_rcv.sb_flags &= ~SB_KNOTE; 1572 splx(s); 1573} 1574 1575/*ARGSUSED*/ 1576static int 1577filt_soread(struct knote *kn, long hint) 1578{ 1579 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1580 1581 kn->kn_data = so->so_rcv.sb_cc; 1582 if (so->so_state & SS_CANTRCVMORE) { 1583 kn->kn_flags |= EV_EOF; 1584 kn->kn_fflags = so->so_error; 1585 return (1); 1586 } 1587 if (so->so_error) /* temporary udp error */ 1588 return (1); 1589 if (kn->kn_sfflags & NOTE_LOWAT) 1590 return (kn->kn_data >= kn->kn_sdata); 1591 return (kn->kn_data >= so->so_rcv.sb_lowat); 1592} 1593 1594static void 1595filt_sowdetach(struct knote *kn) 1596{ 1597 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1598 int s = splnet(); 1599 1600 SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext); 1601 if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note)) 1602 so->so_snd.sb_flags &= ~SB_KNOTE; 1603 splx(s); 1604} 1605 1606/*ARGSUSED*/ 1607static int 1608filt_sowrite(struct knote *kn, long hint) 1609{ 1610 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1611 1612 kn->kn_data = sbspace(&so->so_snd); 1613 if (so->so_state & SS_CANTSENDMORE) { 1614 kn->kn_flags |= EV_EOF; 1615 kn->kn_fflags = so->so_error; 1616 return (1); 1617 } 1618 if (so->so_error) /* temporary udp error */ 1619 return (1); 1620 if (((so->so_state & SS_ISCONNECTED) == 0) && 1621 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 1622 return (0); 1623 if (kn->kn_sfflags & NOTE_LOWAT) 1624 return (kn->kn_data >= kn->kn_sdata); 1625 return (kn->kn_data >= so->so_snd.sb_lowat); 1626} 1627 1628/*ARGSUSED*/ 1629static int 1630filt_solisten(struct knote *kn, long hint) 1631{ 1632 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1633 1634 kn->kn_data = so->so_qlen - so->so_incqlen; 1635 return (! TAILQ_EMPTY(&so->so_comp)); 1636} 1637