uipc_socket.c revision 94160
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 94160 2002-04-08 03:04:22Z hsu $ 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/lock.h> 43#include <sys/malloc.h> 44#include <sys/mbuf.h> 45#include <sys/mutex.h> 46#include <sys/domain.h> 47#include <sys/file.h> /* for struct knote */ 48#include <sys/kernel.h> 49#include <sys/malloc.h> 50#include <sys/event.h> 51#include <sys/poll.h> 52#include <sys/proc.h> 53#include <sys/protosw.h> 54#include <sys/socket.h> 55#include <sys/socketvar.h> 56#include <sys/resourcevar.h> 57#include <sys/signalvar.h> 58#include <sys/sysctl.h> 59#include <sys/uio.h> 60#include <sys/jail.h> 61 62#include <vm/uma.h> 63 64#include <machine/limits.h> 65 66#ifdef INET 67static int do_setopt_accept_filter(struct socket *so, struct sockopt *sopt); 68#endif 69 70static void filt_sordetach(struct knote *kn); 71static int filt_soread(struct knote *kn, long hint); 72static void filt_sowdetach(struct knote *kn); 73static int filt_sowrite(struct knote *kn, long hint); 74static int filt_solisten(struct knote *kn, long hint); 75 76static struct filterops solisten_filtops = 77 { 1, NULL, filt_sordetach, filt_solisten }; 78static struct filterops soread_filtops = 79 { 1, NULL, filt_sordetach, filt_soread }; 80static struct filterops sowrite_filtops = 81 { 1, NULL, filt_sowdetach, filt_sowrite }; 82 83uma_zone_t socket_zone; 84so_gen_t so_gencnt; /* generation count for sockets */ 85 86MALLOC_DEFINE(M_SONAME, "soname", "socket name"); 87MALLOC_DEFINE(M_PCB, "pcb", "protocol control block"); 88 89SYSCTL_DECL(_kern_ipc); 90 91static int somaxconn = SOMAXCONN; 92SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW, 93 &somaxconn, 0, "Maximum pending socket connection queue size"); 94static int numopensockets; 95SYSCTL_INT(_kern_ipc, OID_AUTO, numopensockets, CTLFLAG_RD, 96 &numopensockets, 0, "Number of open sockets"); 97 98 99/* 100 * Socket operation routines. 101 * These routines are called by the routines in 102 * sys_socket.c or from a system process, and 103 * implement the semantics of socket operations by 104 * switching out to the protocol specific routines. 105 */ 106 107/* 108 * Get a socket structure from our zone, and initialize it. 109 * Note that it would probably be better to allocate socket 110 * and PCB at the same time, but I'm not convinced that all 111 * the protocols can be easily modified to do this. 112 * 113 * soalloc() returns a socket with a ref count of 0. 114 */ 115struct socket * 116soalloc(waitok) 117 int waitok; 118{ 119 struct socket *so; 120 int flag; 121 122 if (waitok == 1) 123 flag = M_WAITOK; 124 else 125 flag = M_NOWAIT; 126 127 so = uma_zalloc(socket_zone, flag); 128 if (so) { 129 /* XXX race condition for reentrant kernel */ 130 bzero(so, sizeof *so); 131 so->so_gencnt = ++so_gencnt; 132 /* sx_init(&so->so_sxlock, "socket sxlock"); */ 133 TAILQ_INIT(&so->so_aiojobq); 134 ++numopensockets; 135 } 136 return so; 137} 138 139/* 140 * socreate returns a socket with a ref count of 1. The socket should be 141 * closed with soclose(). 142 */ 143int 144socreate(dom, aso, type, proto, cred, td) 145 int dom; 146 struct socket **aso; 147 register int type; 148 int proto; 149 struct ucred *cred; 150 struct thread *td; 151{ 152 register struct protosw *prp; 153 register struct socket *so; 154 register int error; 155 156 if (proto) 157 prp = pffindproto(dom, proto, type); 158 else 159 prp = pffindtype(dom, type); 160 161 if (prp == 0 || prp->pr_usrreqs->pru_attach == 0) 162 return (EPROTONOSUPPORT); 163 164 if (jailed(td->td_ucred) && jail_socket_unixiproute_only && 165 prp->pr_domain->dom_family != PF_LOCAL && 166 prp->pr_domain->dom_family != PF_INET && 167 prp->pr_domain->dom_family != PF_ROUTE) { 168 return (EPROTONOSUPPORT); 169 } 170 171 if (prp->pr_type != type) 172 return (EPROTOTYPE); 173 so = soalloc(td != 0); 174 if (so == 0) 175 return (ENOBUFS); 176 177 TAILQ_INIT(&so->so_incomp); 178 TAILQ_INIT(&so->so_comp); 179 so->so_type = type; 180 so->so_cred = crhold(cred); 181 so->so_proto = prp; 182 soref(so); 183 error = (*prp->pr_usrreqs->pru_attach)(so, proto, td); 184 if (error) { 185 so->so_state |= SS_NOFDREF; 186 sorele(so); 187 return (error); 188 } 189 *aso = so; 190 return (0); 191} 192 193int 194sobind(so, nam, td) 195 struct socket *so; 196 struct sockaddr *nam; 197 struct thread *td; 198{ 199 int s = splnet(); 200 int error; 201 202 error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam, td); 203 splx(s); 204 return (error); 205} 206 207static void 208sodealloc(struct socket *so) 209{ 210 211 KASSERT(so->so_count == 0, ("sodealloc(): so_count %d", so->so_count)); 212 so->so_gencnt = ++so_gencnt; 213 if (so->so_rcv.sb_hiwat) 214 (void)chgsbsize(so->so_cred->cr_uidinfo, 215 &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY); 216 if (so->so_snd.sb_hiwat) 217 (void)chgsbsize(so->so_cred->cr_uidinfo, 218 &so->so_snd.sb_hiwat, 0, RLIM_INFINITY); 219#ifdef INET 220 if (so->so_accf != NULL) { 221 if (so->so_accf->so_accept_filter != NULL && 222 so->so_accf->so_accept_filter->accf_destroy != NULL) { 223 so->so_accf->so_accept_filter->accf_destroy(so); 224 } 225 if (so->so_accf->so_accept_filter_str != NULL) 226 FREE(so->so_accf->so_accept_filter_str, M_ACCF); 227 FREE(so->so_accf, M_ACCF); 228 } 229#endif 230 crfree(so->so_cred); 231 /* sx_destroy(&so->so_sxlock); */ 232 uma_zfree(socket_zone, so); 233 --numopensockets; 234} 235 236int 237solisten(so, backlog, td) 238 register struct socket *so; 239 int backlog; 240 struct thread *td; 241{ 242 int s, error; 243 244 s = splnet(); 245 error = (*so->so_proto->pr_usrreqs->pru_listen)(so, td); 246 if (error) { 247 splx(s); 248 return (error); 249 } 250 if (TAILQ_EMPTY(&so->so_comp)) 251 so->so_options |= SO_ACCEPTCONN; 252 if (backlog < 0 || backlog > somaxconn) 253 backlog = somaxconn; 254 so->so_qlimit = backlog; 255 splx(s); 256 return (0); 257} 258 259void 260sofree(so) 261 register struct socket *so; 262{ 263 struct socket *head = so->so_head; 264 265 KASSERT(so->so_count == 0, ("socket %p so_count not 0", so)); 266 267 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) 268 return; 269 if (head != NULL) { 270 if (so->so_state & SS_INCOMP) { 271 TAILQ_REMOVE(&head->so_incomp, so, so_list); 272 head->so_incqlen--; 273 } else if (so->so_state & SS_COMP) { 274 /* 275 * We must not decommission a socket that's 276 * on the accept(2) queue. If we do, then 277 * accept(2) may hang after select(2) indicated 278 * that the listening socket was ready. 279 */ 280 return; 281 } else { 282 panic("sofree: not queued"); 283 } 284 head->so_qlen--; 285 so->so_state &= ~SS_INCOMP; 286 so->so_head = NULL; 287 } 288 sbrelease(&so->so_snd, so); 289 sorflush(so); 290 sodealloc(so); 291} 292 293/* 294 * Close a socket on last file table reference removal. 295 * Initiate disconnect if connected. 296 * Free socket when disconnect complete. 297 * 298 * This function will sorele() the socket. Note that soclose() may be 299 * called prior to the ref count reaching zero. The actual socket 300 * structure will not be freed until the ref count reaches zero. 301 */ 302int 303soclose(so) 304 register struct socket *so; 305{ 306 int s = splnet(); /* conservative */ 307 int error = 0; 308 309 funsetown(so->so_sigio); 310 if (so->so_options & SO_ACCEPTCONN) { 311 struct socket *sp, *sonext; 312 313 sp = TAILQ_FIRST(&so->so_incomp); 314 for (; sp != NULL; sp = sonext) { 315 sonext = TAILQ_NEXT(sp, so_list); 316 (void) soabort(sp); 317 } 318 for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) { 319 sonext = TAILQ_NEXT(sp, so_list); 320 /* Dequeue from so_comp since sofree() won't do it */ 321 TAILQ_REMOVE(&so->so_comp, sp, so_list); 322 so->so_qlen--; 323 sp->so_state &= ~SS_COMP; 324 sp->so_head = NULL; 325 (void) soabort(sp); 326 } 327 } 328 if (so->so_pcb == 0) 329 goto discard; 330 if (so->so_state & SS_ISCONNECTED) { 331 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 332 error = sodisconnect(so); 333 if (error) 334 goto drop; 335 } 336 if (so->so_options & SO_LINGER) { 337 if ((so->so_state & SS_ISDISCONNECTING) && 338 (so->so_state & SS_NBIO)) 339 goto drop; 340 while (so->so_state & SS_ISCONNECTED) { 341 error = tsleep((caddr_t)&so->so_timeo, 342 PSOCK | PCATCH, "soclos", so->so_linger * hz); 343 if (error) 344 break; 345 } 346 } 347 } 348drop: 349 if (so->so_pcb) { 350 int error2 = (*so->so_proto->pr_usrreqs->pru_detach)(so); 351 if (error == 0) 352 error = error2; 353 } 354discard: 355 if (so->so_state & SS_NOFDREF) 356 panic("soclose: NOFDREF"); 357 so->so_state |= SS_NOFDREF; 358 sorele(so); 359 splx(s); 360 return (error); 361} 362 363/* 364 * Must be called at splnet... 365 */ 366int 367soabort(so) 368 struct socket *so; 369{ 370 int error; 371 372 error = (*so->so_proto->pr_usrreqs->pru_abort)(so); 373 if (error) { 374 sotryfree(so); /* note: does not decrement the ref count */ 375 return error; 376 } 377 return (0); 378} 379 380int 381soaccept(so, nam) 382 register struct socket *so; 383 struct sockaddr **nam; 384{ 385 int s = splnet(); 386 int error; 387 388 if ((so->so_state & SS_NOFDREF) == 0) 389 panic("soaccept: !NOFDREF"); 390 so->so_state &= ~SS_NOFDREF; 391 error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam); 392 splx(s); 393 return (error); 394} 395 396int 397soconnect(so, nam, td) 398 register struct socket *so; 399 struct sockaddr *nam; 400 struct thread *td; 401{ 402 int s; 403 int error; 404 405 if (so->so_options & SO_ACCEPTCONN) 406 return (EOPNOTSUPP); 407 s = splnet(); 408 /* 409 * If protocol is connection-based, can only connect once. 410 * Otherwise, if connected, try to disconnect first. 411 * This allows user to disconnect by connecting to, e.g., 412 * a null address. 413 */ 414 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 415 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 416 (error = sodisconnect(so)))) 417 error = EISCONN; 418 else 419 error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, td); 420 splx(s); 421 return (error); 422} 423 424int 425soconnect2(so1, so2) 426 register struct socket *so1; 427 struct socket *so2; 428{ 429 int s = splnet(); 430 int error; 431 432 error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2); 433 splx(s); 434 return (error); 435} 436 437int 438sodisconnect(so) 439 register struct socket *so; 440{ 441 int s = splnet(); 442 int error; 443 444 if ((so->so_state & SS_ISCONNECTED) == 0) { 445 error = ENOTCONN; 446 goto bad; 447 } 448 if (so->so_state & SS_ISDISCONNECTING) { 449 error = EALREADY; 450 goto bad; 451 } 452 error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so); 453bad: 454 splx(s); 455 return (error); 456} 457 458#define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 459/* 460 * Send on a socket. 461 * If send must go all at once and message is larger than 462 * send buffering, then hard error. 463 * Lock against other senders. 464 * If must go all at once and not enough room now, then 465 * inform user that this would block and do nothing. 466 * Otherwise, if nonblocking, send as much as possible. 467 * The data to be sent is described by "uio" if nonzero, 468 * otherwise by the mbuf chain "top" (which must be null 469 * if uio is not). Data provided in mbuf chain must be small 470 * enough to send all at once. 471 * 472 * Returns nonzero on error, timeout or signal; callers 473 * must check for short counts if EINTR/ERESTART are returned. 474 * Data and control buffers are freed on return. 475 */ 476int 477sosend(so, addr, uio, top, control, flags, td) 478 register struct socket *so; 479 struct sockaddr *addr; 480 struct uio *uio; 481 struct mbuf *top; 482 struct mbuf *control; 483 int flags; 484 struct thread *td; 485{ 486 struct mbuf **mp; 487 register struct mbuf *m; 488 register long space, len, resid; 489 int clen = 0, error, s, dontroute, mlen; 490 int atomic = sosendallatonce(so) || top; 491 492 if (uio) 493 resid = uio->uio_resid; 494 else 495 resid = top->m_pkthdr.len; 496 /* 497 * In theory resid should be unsigned. 498 * However, space must be signed, as it might be less than 0 499 * if we over-committed, and we must use a signed comparison 500 * of space and resid. On the other hand, a negative resid 501 * causes us to loop sending 0-length segments to the protocol. 502 * 503 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM 504 * type sockets since that's an error. 505 */ 506 if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) { 507 error = EINVAL; 508 goto out; 509 } 510 511 dontroute = 512 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 513 (so->so_proto->pr_flags & PR_ATOMIC); 514 if (td) 515 td->td_proc->p_stats->p_ru.ru_msgsnd++; 516 if (control) 517 clen = control->m_len; 518#define snderr(errno) { error = errno; splx(s); goto release; } 519 520restart: 521 error = sblock(&so->so_snd, SBLOCKWAIT(flags)); 522 if (error) 523 goto out; 524 do { 525 s = splnet(); 526 if (so->so_state & SS_CANTSENDMORE) 527 snderr(EPIPE); 528 if (so->so_error) { 529 error = so->so_error; 530 so->so_error = 0; 531 splx(s); 532 goto release; 533 } 534 if ((so->so_state & SS_ISCONNECTED) == 0) { 535 /* 536 * `sendto' and `sendmsg' is allowed on a connection- 537 * based socket if it supports implied connect. 538 * Return ENOTCONN if not connected and no address is 539 * supplied. 540 */ 541 if ((so->so_proto->pr_flags & PR_CONNREQUIRED) && 542 (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) { 543 if ((so->so_state & SS_ISCONFIRMING) == 0 && 544 !(resid == 0 && clen != 0)) 545 snderr(ENOTCONN); 546 } else if (addr == 0) 547 snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ? 548 ENOTCONN : EDESTADDRREQ); 549 } 550 space = sbspace(&so->so_snd); 551 if (flags & MSG_OOB) 552 space += 1024; 553 if ((atomic && resid > so->so_snd.sb_hiwat) || 554 clen > so->so_snd.sb_hiwat) 555 snderr(EMSGSIZE); 556 if (space < resid + clen && 557 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 558 if (so->so_state & SS_NBIO) 559 snderr(EWOULDBLOCK); 560 sbunlock(&so->so_snd); 561 error = sbwait(&so->so_snd); 562 splx(s); 563 if (error) 564 goto out; 565 goto restart; 566 } 567 splx(s); 568 mp = ⊤ 569 space -= clen; 570 do { 571 if (uio == NULL) { 572 /* 573 * Data is prepackaged in "top". 574 */ 575 resid = 0; 576 if (flags & MSG_EOR) 577 top->m_flags |= M_EOR; 578 } else do { 579 if (top == 0) { 580 MGETHDR(m, M_TRYWAIT, MT_DATA); 581 if (m == NULL) { 582 error = ENOBUFS; 583 goto release; 584 } 585 mlen = MHLEN; 586 m->m_pkthdr.len = 0; 587 m->m_pkthdr.rcvif = (struct ifnet *)0; 588 } else { 589 MGET(m, M_TRYWAIT, MT_DATA); 590 if (m == NULL) { 591 error = ENOBUFS; 592 goto release; 593 } 594 mlen = MLEN; 595 } 596 if (resid >= MINCLSIZE) { 597 MCLGET(m, M_TRYWAIT); 598 if ((m->m_flags & M_EXT) == 0) 599 goto nopages; 600 mlen = MCLBYTES; 601 len = min(min(mlen, resid), space); 602 } else { 603nopages: 604 len = min(min(mlen, resid), space); 605 /* 606 * For datagram protocols, leave room 607 * for protocol headers in first mbuf. 608 */ 609 if (atomic && top == 0 && len < mlen) 610 MH_ALIGN(m, len); 611 } 612 space -= len; 613 error = uiomove(mtod(m, caddr_t), (int)len, uio); 614 resid = uio->uio_resid; 615 m->m_len = len; 616 *mp = m; 617 top->m_pkthdr.len += len; 618 if (error) 619 goto release; 620 mp = &m->m_next; 621 if (resid <= 0) { 622 if (flags & MSG_EOR) 623 top->m_flags |= M_EOR; 624 break; 625 } 626 } while (space > 0 && atomic); 627 if (dontroute) 628 so->so_options |= SO_DONTROUTE; 629 s = splnet(); /* XXX */ 630 /* 631 * XXX all the SS_CANTSENDMORE checks previously 632 * done could be out of date. We could have recieved 633 * a reset packet in an interrupt or maybe we slept 634 * while doing page faults in uiomove() etc. We could 635 * probably recheck again inside the splnet() protection 636 * here, but there are probably other places that this 637 * also happens. We must rethink this. 638 */ 639 error = (*so->so_proto->pr_usrreqs->pru_send)(so, 640 (flags & MSG_OOB) ? PRUS_OOB : 641 /* 642 * If the user set MSG_EOF, the protocol 643 * understands this flag and nothing left to 644 * send then use PRU_SEND_EOF instead of PRU_SEND. 645 */ 646 ((flags & MSG_EOF) && 647 (so->so_proto->pr_flags & PR_IMPLOPCL) && 648 (resid <= 0)) ? 649 PRUS_EOF : 650 /* If there is more to send set PRUS_MORETOCOME */ 651 (resid > 0 && space > 0) ? PRUS_MORETOCOME : 0, 652 top, addr, control, td); 653 splx(s); 654 if (dontroute) 655 so->so_options &= ~SO_DONTROUTE; 656 clen = 0; 657 control = 0; 658 top = 0; 659 mp = ⊤ 660 if (error) 661 goto release; 662 } while (resid && space > 0); 663 } while (resid); 664 665release: 666 sbunlock(&so->so_snd); 667out: 668 if (top) 669 m_freem(top); 670 if (control) 671 m_freem(control); 672 return (error); 673} 674 675/* 676 * Implement receive operations on a socket. 677 * We depend on the way that records are added to the sockbuf 678 * by sbappend*. In particular, each record (mbufs linked through m_next) 679 * must begin with an address if the protocol so specifies, 680 * followed by an optional mbuf or mbufs containing ancillary data, 681 * and then zero or more mbufs of data. 682 * In order to avoid blocking network interrupts for the entire time here, 683 * we splx() while doing the actual copy to user space. 684 * Although the sockbuf is locked, new data may still be appended, 685 * and thus we must maintain consistency of the sockbuf during that time. 686 * 687 * The caller may receive the data as a single mbuf chain by supplying 688 * an mbuf **mp0 for use in returning the chain. The uio is then used 689 * only for the count in uio_resid. 690 */ 691int 692soreceive(so, psa, uio, mp0, controlp, flagsp) 693 register struct socket *so; 694 struct sockaddr **psa; 695 struct uio *uio; 696 struct mbuf **mp0; 697 struct mbuf **controlp; 698 int *flagsp; 699{ 700 struct mbuf *m, **mp; 701 register int flags, len, error, s, offset; 702 struct protosw *pr = so->so_proto; 703 struct mbuf *nextrecord; 704 int moff, type = 0; 705 int orig_resid = uio->uio_resid; 706 707 mp = mp0; 708 if (psa) 709 *psa = 0; 710 if (controlp) 711 *controlp = 0; 712 if (flagsp) 713 flags = *flagsp &~ MSG_EOR; 714 else 715 flags = 0; 716 if (flags & MSG_OOB) { 717 m = m_get(M_TRYWAIT, MT_DATA); 718 if (m == NULL) 719 return (ENOBUFS); 720 error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK); 721 if (error) 722 goto bad; 723 do { 724 error = uiomove(mtod(m, caddr_t), 725 (int) min(uio->uio_resid, m->m_len), uio); 726 m = m_free(m); 727 } while (uio->uio_resid && error == 0 && m); 728bad: 729 if (m) 730 m_freem(m); 731 return (error); 732 } 733 if (mp) 734 *mp = (struct mbuf *)0; 735 if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) 736 (*pr->pr_usrreqs->pru_rcvd)(so, 0); 737 738restart: 739 error = sblock(&so->so_rcv, SBLOCKWAIT(flags)); 740 if (error) 741 return (error); 742 s = splnet(); 743 744 m = so->so_rcv.sb_mb; 745 /* 746 * If we have less data than requested, block awaiting more 747 * (subject to any timeout) if: 748 * 1. the current count is less than the low water mark, or 749 * 2. MSG_WAITALL is set, and it is possible to do the entire 750 * receive operation at once if we block (resid <= hiwat). 751 * 3. MSG_DONTWAIT is not set 752 * If MSG_WAITALL is set but resid is larger than the receive buffer, 753 * we have to do the receive in sections, and thus risk returning 754 * a short count if a timeout or signal occurs after we start. 755 */ 756 if (m == 0 || (((flags & MSG_DONTWAIT) == 0 && 757 so->so_rcv.sb_cc < uio->uio_resid) && 758 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 759 ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) && 760 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) { 761 KASSERT(m != 0 || !so->so_rcv.sb_cc, 762 ("receive: m == %p so->so_rcv.sb_cc == %lu", 763 m, so->so_rcv.sb_cc)); 764 if (so->so_error) { 765 if (m) 766 goto dontblock; 767 error = so->so_error; 768 if ((flags & MSG_PEEK) == 0) 769 so->so_error = 0; 770 goto release; 771 } 772 if (so->so_state & SS_CANTRCVMORE) { 773 if (m) 774 goto dontblock; 775 else 776 goto release; 777 } 778 for (; m; m = m->m_next) 779 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 780 m = so->so_rcv.sb_mb; 781 goto dontblock; 782 } 783 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 784 (so->so_proto->pr_flags & PR_CONNREQUIRED)) { 785 error = ENOTCONN; 786 goto release; 787 } 788 if (uio->uio_resid == 0) 789 goto release; 790 if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) { 791 error = EWOULDBLOCK; 792 goto release; 793 } 794 sbunlock(&so->so_rcv); 795 error = sbwait(&so->so_rcv); 796 splx(s); 797 if (error) 798 return (error); 799 goto restart; 800 } 801dontblock: 802 if (uio->uio_td) 803 uio->uio_td->td_proc->p_stats->p_ru.ru_msgrcv++; 804 nextrecord = m->m_nextpkt; 805 if (pr->pr_flags & PR_ADDR) { 806 KASSERT(m->m_type == MT_SONAME, 807 ("m->m_type == %d", m->m_type)); 808 orig_resid = 0; 809 if (psa) 810 *psa = dup_sockaddr(mtod(m, struct sockaddr *), 811 mp0 == 0); 812 if (flags & MSG_PEEK) { 813 m = m->m_next; 814 } else { 815 sbfree(&so->so_rcv, m); 816 so->so_rcv.sb_mb = m_free(m); 817 m = so->so_rcv.sb_mb; 818 } 819 } 820 while (m && m->m_type == MT_CONTROL && error == 0) { 821 if (flags & MSG_PEEK) { 822 if (controlp) 823 *controlp = m_copy(m, 0, m->m_len); 824 m = m->m_next; 825 } else { 826 sbfree(&so->so_rcv, m); 827 so->so_rcv.sb_mb = m->m_next; 828 m->m_next = NULL; 829 if (pr->pr_domain->dom_externalize) 830 error = 831 (*pr->pr_domain->dom_externalize)(m, controlp); 832 else if (controlp) 833 *controlp = m; 834 else 835 m_freem(m); 836 m = so->so_rcv.sb_mb; 837 } 838 if (controlp) { 839 orig_resid = 0; 840 do 841 controlp = &(*controlp)->m_next; 842 while (*controlp != NULL); 843 } 844 } 845 if (m) { 846 if ((flags & MSG_PEEK) == 0) 847 m->m_nextpkt = nextrecord; 848 type = m->m_type; 849 if (type == MT_OOBDATA) 850 flags |= MSG_OOB; 851 } 852 moff = 0; 853 offset = 0; 854 while (m && uio->uio_resid > 0 && error == 0) { 855 if (m->m_type == MT_OOBDATA) { 856 if (type != MT_OOBDATA) 857 break; 858 } else if (type == MT_OOBDATA) 859 break; 860 else 861 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER, 862 ("m->m_type == %d", m->m_type)); 863 so->so_state &= ~SS_RCVATMARK; 864 len = uio->uio_resid; 865 if (so->so_oobmark && len > so->so_oobmark - offset) 866 len = so->so_oobmark - offset; 867 if (len > m->m_len - moff) 868 len = m->m_len - moff; 869 /* 870 * If mp is set, just pass back the mbufs. 871 * Otherwise copy them out via the uio, then free. 872 * Sockbuf must be consistent here (points to current mbuf, 873 * it points to next record) when we drop priority; 874 * we must note any additions to the sockbuf when we 875 * block interrupts again. 876 */ 877 if (mp == 0) { 878 splx(s); 879 error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio); 880 s = splnet(); 881 if (error) 882 goto release; 883 } else 884 uio->uio_resid -= len; 885 if (len == m->m_len - moff) { 886 if (m->m_flags & M_EOR) 887 flags |= MSG_EOR; 888 if (flags & MSG_PEEK) { 889 m = m->m_next; 890 moff = 0; 891 } else { 892 nextrecord = m->m_nextpkt; 893 sbfree(&so->so_rcv, m); 894 if (mp) { 895 *mp = m; 896 mp = &m->m_next; 897 so->so_rcv.sb_mb = m = m->m_next; 898 *mp = (struct mbuf *)0; 899 } else { 900 so->so_rcv.sb_mb = m_free(m); 901 m = so->so_rcv.sb_mb; 902 } 903 if (m) 904 m->m_nextpkt = nextrecord; 905 } 906 } else { 907 if (flags & MSG_PEEK) 908 moff += len; 909 else { 910 if (mp) 911 *mp = m_copym(m, 0, len, M_TRYWAIT); 912 m->m_data += len; 913 m->m_len -= len; 914 so->so_rcv.sb_cc -= len; 915 } 916 } 917 if (so->so_oobmark) { 918 if ((flags & MSG_PEEK) == 0) { 919 so->so_oobmark -= len; 920 if (so->so_oobmark == 0) { 921 so->so_state |= SS_RCVATMARK; 922 break; 923 } 924 } else { 925 offset += len; 926 if (offset == so->so_oobmark) 927 break; 928 } 929 } 930 if (flags & MSG_EOR) 931 break; 932 /* 933 * If the MSG_WAITALL flag is set (for non-atomic socket), 934 * we must not quit until "uio->uio_resid == 0" or an error 935 * termination. If a signal/timeout occurs, return 936 * with a short count but without error. 937 * Keep sockbuf locked against other readers. 938 */ 939 while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 && 940 !sosendallatonce(so) && !nextrecord) { 941 if (so->so_error || so->so_state & SS_CANTRCVMORE) 942 break; 943 /* 944 * Notify the protocol that some data has been 945 * drained before blocking. 946 */ 947 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 948 (*pr->pr_usrreqs->pru_rcvd)(so, flags); 949 error = sbwait(&so->so_rcv); 950 if (error) { 951 sbunlock(&so->so_rcv); 952 splx(s); 953 return (0); 954 } 955 m = so->so_rcv.sb_mb; 956 if (m) 957 nextrecord = m->m_nextpkt; 958 } 959 } 960 961 if (m && pr->pr_flags & PR_ATOMIC) { 962 flags |= MSG_TRUNC; 963 if ((flags & MSG_PEEK) == 0) 964 (void) sbdroprecord(&so->so_rcv); 965 } 966 if ((flags & MSG_PEEK) == 0) { 967 if (m == 0) 968 so->so_rcv.sb_mb = nextrecord; 969 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 970 (*pr->pr_usrreqs->pru_rcvd)(so, flags); 971 } 972 if (orig_resid == uio->uio_resid && orig_resid && 973 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 974 sbunlock(&so->so_rcv); 975 splx(s); 976 goto restart; 977 } 978 979 if (flagsp) 980 *flagsp |= flags; 981release: 982 sbunlock(&so->so_rcv); 983 splx(s); 984 return (error); 985} 986 987int 988soshutdown(so, how) 989 register struct socket *so; 990 register int how; 991{ 992 register struct protosw *pr = so->so_proto; 993 994 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 995 return (EINVAL); 996 997 if (how != SHUT_WR) 998 sorflush(so); 999 if (how != SHUT_RD) 1000 return ((*pr->pr_usrreqs->pru_shutdown)(so)); 1001 return (0); 1002} 1003 1004void 1005sorflush(so) 1006 register struct socket *so; 1007{ 1008 register struct sockbuf *sb = &so->so_rcv; 1009 register struct protosw *pr = so->so_proto; 1010 register int s; 1011 struct sockbuf asb; 1012 1013 sb->sb_flags |= SB_NOINTR; 1014 (void) sblock(sb, M_WAITOK); 1015 s = splimp(); 1016 socantrcvmore(so); 1017 sbunlock(sb); 1018 asb = *sb; 1019 bzero((caddr_t)sb, sizeof (*sb)); 1020 splx(s); 1021 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) 1022 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 1023 sbrelease(&asb, so); 1024} 1025 1026#ifdef INET 1027static int 1028do_setopt_accept_filter(so, sopt) 1029 struct socket *so; 1030 struct sockopt *sopt; 1031{ 1032 struct accept_filter_arg *afap = NULL; 1033 struct accept_filter *afp; 1034 struct so_accf *af = so->so_accf; 1035 int error = 0; 1036 1037 /* do not set/remove accept filters on non listen sockets */ 1038 if ((so->so_options & SO_ACCEPTCONN) == 0) { 1039 error = EINVAL; 1040 goto out; 1041 } 1042 1043 /* removing the filter */ 1044 if (sopt == NULL) { 1045 if (af != NULL) { 1046 if (af->so_accept_filter != NULL && 1047 af->so_accept_filter->accf_destroy != NULL) { 1048 af->so_accept_filter->accf_destroy(so); 1049 } 1050 if (af->so_accept_filter_str != NULL) { 1051 FREE(af->so_accept_filter_str, M_ACCF); 1052 } 1053 FREE(af, M_ACCF); 1054 so->so_accf = NULL; 1055 } 1056 so->so_options &= ~SO_ACCEPTFILTER; 1057 return (0); 1058 } 1059 /* adding a filter */ 1060 /* must remove previous filter first */ 1061 if (af != NULL) { 1062 error = EINVAL; 1063 goto out; 1064 } 1065 /* don't put large objects on the kernel stack */ 1066 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK); 1067 error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap); 1068 afap->af_name[sizeof(afap->af_name)-1] = '\0'; 1069 afap->af_arg[sizeof(afap->af_arg)-1] = '\0'; 1070 if (error) 1071 goto out; 1072 afp = accept_filt_get(afap->af_name); 1073 if (afp == NULL) { 1074 error = ENOENT; 1075 goto out; 1076 } 1077 MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK | M_ZERO); 1078 if (afp->accf_create != NULL) { 1079 if (afap->af_name[0] != '\0') { 1080 int len = strlen(afap->af_name) + 1; 1081 1082 MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK); 1083 strcpy(af->so_accept_filter_str, afap->af_name); 1084 } 1085 af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg); 1086 if (af->so_accept_filter_arg == NULL) { 1087 FREE(af->so_accept_filter_str, M_ACCF); 1088 FREE(af, M_ACCF); 1089 so->so_accf = NULL; 1090 error = EINVAL; 1091 goto out; 1092 } 1093 } 1094 af->so_accept_filter = afp; 1095 so->so_accf = af; 1096 so->so_options |= SO_ACCEPTFILTER; 1097out: 1098 if (afap != NULL) 1099 FREE(afap, M_TEMP); 1100 return (error); 1101} 1102#endif /* INET */ 1103 1104/* 1105 * Perhaps this routine, and sooptcopyout(), below, ought to come in 1106 * an additional variant to handle the case where the option value needs 1107 * to be some kind of integer, but not a specific size. 1108 * In addition to their use here, these functions are also called by the 1109 * protocol-level pr_ctloutput() routines. 1110 */ 1111int 1112sooptcopyin(sopt, buf, len, minlen) 1113 struct sockopt *sopt; 1114 void *buf; 1115 size_t len; 1116 size_t minlen; 1117{ 1118 size_t valsize; 1119 1120 /* 1121 * If the user gives us more than we wanted, we ignore it, 1122 * but if we don't get the minimum length the caller 1123 * wants, we return EINVAL. On success, sopt->sopt_valsize 1124 * is set to however much we actually retrieved. 1125 */ 1126 if ((valsize = sopt->sopt_valsize) < minlen) 1127 return EINVAL; 1128 if (valsize > len) 1129 sopt->sopt_valsize = valsize = len; 1130 1131 if (sopt->sopt_td != 0) 1132 return (copyin(sopt->sopt_val, buf, valsize)); 1133 1134 bcopy(sopt->sopt_val, buf, valsize); 1135 return 0; 1136} 1137 1138int 1139sosetopt(so, sopt) 1140 struct socket *so; 1141 struct sockopt *sopt; 1142{ 1143 int error, optval; 1144 struct linger l; 1145 struct timeval tv; 1146 u_long val; 1147 1148 error = 0; 1149 if (sopt->sopt_level != SOL_SOCKET) { 1150 if (so->so_proto && so->so_proto->pr_ctloutput) 1151 return ((*so->so_proto->pr_ctloutput) 1152 (so, sopt)); 1153 error = ENOPROTOOPT; 1154 } else { 1155 switch (sopt->sopt_name) { 1156#ifdef INET 1157 case SO_ACCEPTFILTER: 1158 error = do_setopt_accept_filter(so, sopt); 1159 if (error) 1160 goto bad; 1161 break; 1162#endif 1163 case SO_LINGER: 1164 error = sooptcopyin(sopt, &l, sizeof l, sizeof l); 1165 if (error) 1166 goto bad; 1167 1168 so->so_linger = l.l_linger; 1169 if (l.l_onoff) 1170 so->so_options |= SO_LINGER; 1171 else 1172 so->so_options &= ~SO_LINGER; 1173 break; 1174 1175 case SO_DEBUG: 1176 case SO_KEEPALIVE: 1177 case SO_DONTROUTE: 1178 case SO_USELOOPBACK: 1179 case SO_BROADCAST: 1180 case SO_REUSEADDR: 1181 case SO_REUSEPORT: 1182 case SO_OOBINLINE: 1183 case SO_TIMESTAMP: 1184 error = sooptcopyin(sopt, &optval, sizeof optval, 1185 sizeof optval); 1186 if (error) 1187 goto bad; 1188 if (optval) 1189 so->so_options |= sopt->sopt_name; 1190 else 1191 so->so_options &= ~sopt->sopt_name; 1192 break; 1193 1194 case SO_SNDBUF: 1195 case SO_RCVBUF: 1196 case SO_SNDLOWAT: 1197 case SO_RCVLOWAT: 1198 error = sooptcopyin(sopt, &optval, sizeof optval, 1199 sizeof optval); 1200 if (error) 1201 goto bad; 1202 1203 /* 1204 * Values < 1 make no sense for any of these 1205 * options, so disallow them. 1206 */ 1207 if (optval < 1) { 1208 error = EINVAL; 1209 goto bad; 1210 } 1211 1212 switch (sopt->sopt_name) { 1213 case SO_SNDBUF: 1214 case SO_RCVBUF: 1215 if (sbreserve(sopt->sopt_name == SO_SNDBUF ? 1216 &so->so_snd : &so->so_rcv, (u_long)optval, 1217 so, curthread) == 0) { 1218 error = ENOBUFS; 1219 goto bad; 1220 } 1221 break; 1222 1223 /* 1224 * Make sure the low-water is never greater than 1225 * the high-water. 1226 */ 1227 case SO_SNDLOWAT: 1228 so->so_snd.sb_lowat = 1229 (optval > so->so_snd.sb_hiwat) ? 1230 so->so_snd.sb_hiwat : optval; 1231 break; 1232 case SO_RCVLOWAT: 1233 so->so_rcv.sb_lowat = 1234 (optval > so->so_rcv.sb_hiwat) ? 1235 so->so_rcv.sb_hiwat : optval; 1236 break; 1237 } 1238 break; 1239 1240 case SO_SNDTIMEO: 1241 case SO_RCVTIMEO: 1242 error = sooptcopyin(sopt, &tv, sizeof tv, 1243 sizeof tv); 1244 if (error) 1245 goto bad; 1246 1247 /* assert(hz > 0); */ 1248 if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz || 1249 tv.tv_usec < 0 || tv.tv_usec >= 1000000) { 1250 error = EDOM; 1251 goto bad; 1252 } 1253 /* assert(tick > 0); */ 1254 /* assert(ULONG_MAX - SHRT_MAX >= 1000000); */ 1255 val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick; 1256 if (val > SHRT_MAX) { 1257 error = EDOM; 1258 goto bad; 1259 } 1260 1261 switch (sopt->sopt_name) { 1262 case SO_SNDTIMEO: 1263 so->so_snd.sb_timeo = val; 1264 break; 1265 case SO_RCVTIMEO: 1266 so->so_rcv.sb_timeo = val; 1267 break; 1268 } 1269 break; 1270 default: 1271 error = ENOPROTOOPT; 1272 break; 1273 } 1274 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) { 1275 (void) ((*so->so_proto->pr_ctloutput) 1276 (so, sopt)); 1277 } 1278 } 1279bad: 1280 return (error); 1281} 1282 1283/* Helper routine for getsockopt */ 1284int 1285sooptcopyout(sopt, buf, len) 1286 struct sockopt *sopt; 1287 void *buf; 1288 size_t len; 1289{ 1290 int error; 1291 size_t valsize; 1292 1293 error = 0; 1294 1295 /* 1296 * Documented get behavior is that we always return a value, 1297 * possibly truncated to fit in the user's buffer. 1298 * Traditional behavior is that we always tell the user 1299 * precisely how much we copied, rather than something useful 1300 * like the total amount we had available for her. 1301 * Note that this interface is not idempotent; the entire answer must 1302 * generated ahead of time. 1303 */ 1304 valsize = min(len, sopt->sopt_valsize); 1305 sopt->sopt_valsize = valsize; 1306 if (sopt->sopt_val != 0) { 1307 if (sopt->sopt_td != 0) 1308 error = copyout(buf, sopt->sopt_val, valsize); 1309 else 1310 bcopy(buf, sopt->sopt_val, valsize); 1311 } 1312 return error; 1313} 1314 1315int 1316sogetopt(so, sopt) 1317 struct socket *so; 1318 struct sockopt *sopt; 1319{ 1320 int error, optval; 1321 struct linger l; 1322 struct timeval tv; 1323#ifdef INET 1324 struct accept_filter_arg *afap; 1325#endif 1326 1327 error = 0; 1328 if (sopt->sopt_level != SOL_SOCKET) { 1329 if (so->so_proto && so->so_proto->pr_ctloutput) { 1330 return ((*so->so_proto->pr_ctloutput) 1331 (so, sopt)); 1332 } else 1333 return (ENOPROTOOPT); 1334 } else { 1335 switch (sopt->sopt_name) { 1336#ifdef INET 1337 case SO_ACCEPTFILTER: 1338 if ((so->so_options & SO_ACCEPTCONN) == 0) 1339 return (EINVAL); 1340 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), 1341 M_TEMP, M_WAITOK | M_ZERO); 1342 if ((so->so_options & SO_ACCEPTFILTER) != 0) { 1343 strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name); 1344 if (so->so_accf->so_accept_filter_str != NULL) 1345 strcpy(afap->af_arg, so->so_accf->so_accept_filter_str); 1346 } 1347 error = sooptcopyout(sopt, afap, sizeof(*afap)); 1348 FREE(afap, M_TEMP); 1349 break; 1350#endif 1351 1352 case SO_LINGER: 1353 l.l_onoff = so->so_options & SO_LINGER; 1354 l.l_linger = so->so_linger; 1355 error = sooptcopyout(sopt, &l, sizeof l); 1356 break; 1357 1358 case SO_USELOOPBACK: 1359 case SO_DONTROUTE: 1360 case SO_DEBUG: 1361 case SO_KEEPALIVE: 1362 case SO_REUSEADDR: 1363 case SO_REUSEPORT: 1364 case SO_BROADCAST: 1365 case SO_OOBINLINE: 1366 case SO_TIMESTAMP: 1367 optval = so->so_options & sopt->sopt_name; 1368integer: 1369 error = sooptcopyout(sopt, &optval, sizeof optval); 1370 break; 1371 1372 case SO_TYPE: 1373 optval = so->so_type; 1374 goto integer; 1375 1376 case SO_ERROR: 1377 optval = so->so_error; 1378 so->so_error = 0; 1379 goto integer; 1380 1381 case SO_SNDBUF: 1382 optval = so->so_snd.sb_hiwat; 1383 goto integer; 1384 1385 case SO_RCVBUF: 1386 optval = so->so_rcv.sb_hiwat; 1387 goto integer; 1388 1389 case SO_SNDLOWAT: 1390 optval = so->so_snd.sb_lowat; 1391 goto integer; 1392 1393 case SO_RCVLOWAT: 1394 optval = so->so_rcv.sb_lowat; 1395 goto integer; 1396 1397 case SO_SNDTIMEO: 1398 case SO_RCVTIMEO: 1399 optval = (sopt->sopt_name == SO_SNDTIMEO ? 1400 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 1401 1402 tv.tv_sec = optval / hz; 1403 tv.tv_usec = (optval % hz) * tick; 1404 error = sooptcopyout(sopt, &tv, sizeof tv); 1405 break; 1406 1407 default: 1408 error = ENOPROTOOPT; 1409 break; 1410 } 1411 return (error); 1412 } 1413} 1414 1415/* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */ 1416int 1417soopt_getm(struct sockopt *sopt, struct mbuf **mp) 1418{ 1419 struct mbuf *m, *m_prev; 1420 int sopt_size = sopt->sopt_valsize; 1421 1422 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA); 1423 if (m == 0) 1424 return ENOBUFS; 1425 if (sopt_size > MLEN) { 1426 MCLGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT); 1427 if ((m->m_flags & M_EXT) == 0) { 1428 m_free(m); 1429 return ENOBUFS; 1430 } 1431 m->m_len = min(MCLBYTES, sopt_size); 1432 } else { 1433 m->m_len = min(MLEN, sopt_size); 1434 } 1435 sopt_size -= m->m_len; 1436 *mp = m; 1437 m_prev = m; 1438 1439 while (sopt_size) { 1440 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA); 1441 if (m == 0) { 1442 m_freem(*mp); 1443 return ENOBUFS; 1444 } 1445 if (sopt_size > MLEN) { 1446 MCLGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT); 1447 if ((m->m_flags & M_EXT) == 0) { 1448 m_freem(*mp); 1449 return ENOBUFS; 1450 } 1451 m->m_len = min(MCLBYTES, sopt_size); 1452 } else { 1453 m->m_len = min(MLEN, sopt_size); 1454 } 1455 sopt_size -= m->m_len; 1456 m_prev->m_next = m; 1457 m_prev = m; 1458 } 1459 return 0; 1460} 1461 1462/* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */ 1463int 1464soopt_mcopyin(struct sockopt *sopt, struct mbuf *m) 1465{ 1466 struct mbuf *m0 = m; 1467 1468 if (sopt->sopt_val == NULL) 1469 return 0; 1470 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1471 if (sopt->sopt_td != NULL) { 1472 int error; 1473 1474 error = copyin(sopt->sopt_val, mtod(m, char *), 1475 m->m_len); 1476 if (error != 0) { 1477 m_freem(m0); 1478 return(error); 1479 } 1480 } else 1481 bcopy(sopt->sopt_val, mtod(m, char *), m->m_len); 1482 sopt->sopt_valsize -= m->m_len; 1483 (caddr_t)sopt->sopt_val += m->m_len; 1484 m = m->m_next; 1485 } 1486 if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */ 1487 panic("ip6_sooptmcopyin"); 1488 return 0; 1489} 1490 1491/* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */ 1492int 1493soopt_mcopyout(struct sockopt *sopt, struct mbuf *m) 1494{ 1495 struct mbuf *m0 = m; 1496 size_t valsize = 0; 1497 1498 if (sopt->sopt_val == NULL) 1499 return 0; 1500 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1501 if (sopt->sopt_td != NULL) { 1502 int error; 1503 1504 error = copyout(mtod(m, char *), sopt->sopt_val, 1505 m->m_len); 1506 if (error != 0) { 1507 m_freem(m0); 1508 return(error); 1509 } 1510 } else 1511 bcopy(mtod(m, char *), sopt->sopt_val, m->m_len); 1512 sopt->sopt_valsize -= m->m_len; 1513 (caddr_t)sopt->sopt_val += m->m_len; 1514 valsize += m->m_len; 1515 m = m->m_next; 1516 } 1517 if (m != NULL) { 1518 /* enough soopt buffer should be given from user-land */ 1519 m_freem(m0); 1520 return(EINVAL); 1521 } 1522 sopt->sopt_valsize = valsize; 1523 return 0; 1524} 1525 1526void 1527sohasoutofband(so) 1528 register struct socket *so; 1529{ 1530 if (so->so_sigio != NULL) 1531 pgsigio(so->so_sigio, SIGURG, 0); 1532 selwakeup(&so->so_rcv.sb_sel); 1533} 1534 1535int 1536sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td) 1537{ 1538 int revents = 0; 1539 int s = splnet(); 1540 1541 if (events & (POLLIN | POLLRDNORM)) 1542 if (soreadable(so)) 1543 revents |= events & (POLLIN | POLLRDNORM); 1544 1545 if (events & POLLINIGNEOF) 1546 if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat || 1547 !TAILQ_EMPTY(&so->so_comp) || so->so_error) 1548 revents |= POLLINIGNEOF; 1549 1550 if (events & (POLLOUT | POLLWRNORM)) 1551 if (sowriteable(so)) 1552 revents |= events & (POLLOUT | POLLWRNORM); 1553 1554 if (events & (POLLPRI | POLLRDBAND)) 1555 if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) 1556 revents |= events & (POLLPRI | POLLRDBAND); 1557 1558 if (revents == 0) { 1559 if (events & 1560 (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM | 1561 POLLRDBAND)) { 1562 selrecord(td, &so->so_rcv.sb_sel); 1563 so->so_rcv.sb_flags |= SB_SEL; 1564 } 1565 1566 if (events & (POLLOUT | POLLWRNORM)) { 1567 selrecord(td, &so->so_snd.sb_sel); 1568 so->so_snd.sb_flags |= SB_SEL; 1569 } 1570 } 1571 1572 splx(s); 1573 return (revents); 1574} 1575 1576int 1577sokqfilter(struct file *fp, struct knote *kn) 1578{ 1579 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1580 struct sockbuf *sb; 1581 int s; 1582 1583 switch (kn->kn_filter) { 1584 case EVFILT_READ: 1585 if (so->so_options & SO_ACCEPTCONN) 1586 kn->kn_fop = &solisten_filtops; 1587 else 1588 kn->kn_fop = &soread_filtops; 1589 sb = &so->so_rcv; 1590 break; 1591 case EVFILT_WRITE: 1592 kn->kn_fop = &sowrite_filtops; 1593 sb = &so->so_snd; 1594 break; 1595 default: 1596 return (1); 1597 } 1598 1599 s = splnet(); 1600 SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext); 1601 sb->sb_flags |= SB_KNOTE; 1602 splx(s); 1603 return (0); 1604} 1605 1606static void 1607filt_sordetach(struct knote *kn) 1608{ 1609 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1610 int s = splnet(); 1611 1612 SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext); 1613 if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note)) 1614 so->so_rcv.sb_flags &= ~SB_KNOTE; 1615 splx(s); 1616} 1617 1618/*ARGSUSED*/ 1619static int 1620filt_soread(struct knote *kn, long hint) 1621{ 1622 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1623 1624 kn->kn_data = so->so_rcv.sb_cc; 1625 if (so->so_state & SS_CANTRCVMORE) { 1626 kn->kn_flags |= EV_EOF; 1627 kn->kn_fflags = so->so_error; 1628 return (1); 1629 } 1630 if (so->so_error) /* temporary udp error */ 1631 return (1); 1632 if (kn->kn_sfflags & NOTE_LOWAT) 1633 return (kn->kn_data >= kn->kn_sdata); 1634 return (kn->kn_data >= so->so_rcv.sb_lowat); 1635} 1636 1637static void 1638filt_sowdetach(struct knote *kn) 1639{ 1640 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1641 int s = splnet(); 1642 1643 SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext); 1644 if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note)) 1645 so->so_snd.sb_flags &= ~SB_KNOTE; 1646 splx(s); 1647} 1648 1649/*ARGSUSED*/ 1650static int 1651filt_sowrite(struct knote *kn, long hint) 1652{ 1653 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1654 1655 kn->kn_data = sbspace(&so->so_snd); 1656 if (so->so_state & SS_CANTSENDMORE) { 1657 kn->kn_flags |= EV_EOF; 1658 kn->kn_fflags = so->so_error; 1659 return (1); 1660 } 1661 if (so->so_error) /* temporary udp error */ 1662 return (1); 1663 if (((so->so_state & SS_ISCONNECTED) == 0) && 1664 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 1665 return (0); 1666 if (kn->kn_sfflags & NOTE_LOWAT) 1667 return (kn->kn_data >= kn->kn_sdata); 1668 return (kn->kn_data >= so->so_snd.sb_lowat); 1669} 1670 1671/*ARGSUSED*/ 1672static int 1673filt_solisten(struct knote *kn, long hint) 1674{ 1675 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1676 1677 kn->kn_data = so->so_qlen - so->so_incqlen; 1678 return (! TAILQ_EMPTY(&so->so_comp)); 1679} 1680 1681int 1682socheckuid(struct socket *so, uid_t uid) 1683{ 1684 1685 if (so == NULL) 1686 return (EPERM); 1687 if (so->so_cred->cr_uid == uid) 1688 return (0); 1689 return (EPERM); 1690} 1691