uipc_socket.c revision 108708
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 108708 2003-01-05 11:14:04Z alfred $ 35 */ 36 37#include "opt_inet.h" 38#include "opt_mac.h" 39#include "opt_zero.h" 40 41#include <sys/param.h> 42#include <sys/systm.h> 43#include <sys/fcntl.h> 44#include <sys/lock.h> 45#include <sys/mac.h> 46#include <sys/malloc.h> 47#include <sys/mbuf.h> 48#include <sys/mutex.h> 49#include <sys/domain.h> 50#include <sys/file.h> /* for struct knote */ 51#include <sys/kernel.h> 52#include <sys/malloc.h> 53#include <sys/event.h> 54#include <sys/poll.h> 55#include <sys/proc.h> 56#include <sys/protosw.h> 57#include <sys/socket.h> 58#include <sys/socketvar.h> 59#include <sys/resourcevar.h> 60#include <sys/signalvar.h> 61#include <sys/sysctl.h> 62#include <sys/uio.h> 63#include <sys/jail.h> 64 65#include <vm/uma.h> 66 67#include <machine/limits.h> 68 69#ifdef INET 70static int do_setopt_accept_filter(struct socket *so, struct sockopt *sopt); 71#endif 72 73static void filt_sordetach(struct knote *kn); 74static int filt_soread(struct knote *kn, long hint); 75static void filt_sowdetach(struct knote *kn); 76static int filt_sowrite(struct knote *kn, long hint); 77static int filt_solisten(struct knote *kn, long hint); 78 79static struct filterops solisten_filtops = 80 { 1, NULL, filt_sordetach, filt_solisten }; 81static struct filterops soread_filtops = 82 { 1, NULL, filt_sordetach, filt_soread }; 83static struct filterops sowrite_filtops = 84 { 1, NULL, filt_sowdetach, filt_sowrite }; 85 86uma_zone_t socket_zone; 87so_gen_t so_gencnt; /* generation count for sockets */ 88 89MALLOC_DEFINE(M_SONAME, "soname", "socket name"); 90MALLOC_DEFINE(M_PCB, "pcb", "protocol control block"); 91 92SYSCTL_DECL(_kern_ipc); 93 94static int somaxconn = SOMAXCONN; 95SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW, 96 &somaxconn, 0, "Maximum pending socket connection queue size"); 97static int numopensockets; 98SYSCTL_INT(_kern_ipc, OID_AUTO, numopensockets, CTLFLAG_RD, 99 &numopensockets, 0, "Number of open sockets"); 100#ifdef ZERO_COPY_SOCKETS 101/* These aren't static because they're used in other files. */ 102int so_zero_copy_send = 1; 103int so_zero_copy_receive = 1; 104SYSCTL_NODE(_kern_ipc, OID_AUTO, zero_copy, CTLFLAG_RD, 0, 105 "Zero copy controls"); 106SYSCTL_INT(_kern_ipc_zero_copy, OID_AUTO, receive, CTLFLAG_RW, 107 &so_zero_copy_receive, 0, "Enable zero copy receive"); 108SYSCTL_INT(_kern_ipc_zero_copy, OID_AUTO, send, CTLFLAG_RW, 109 &so_zero_copy_send, 0, "Enable zero copy send"); 110#endif /* ZERO_COPY_SOCKETS */ 111 112 113/* 114 * Socket operation routines. 115 * These routines are called by the routines in 116 * sys_socket.c or from a system process, and 117 * implement the semantics of socket operations by 118 * switching out to the protocol specific routines. 119 */ 120 121/* 122 * Get a socket structure from our zone, and initialize it. 123 * Note that it would probably be better to allocate socket 124 * and PCB at the same time, but I'm not convinced that all 125 * the protocols can be easily modified to do this. 126 * 127 * soalloc() returns a socket with a ref count of 0. 128 */ 129struct socket * 130soalloc(waitok) 131 int waitok; 132{ 133 struct socket *so; 134#ifdef MAC 135 int error; 136#endif 137 int flag; 138 139 if (waitok == 1) 140 flag = M_WAITOK; 141 else 142 flag = M_NOWAIT; 143 flag |= M_ZERO; 144 so = uma_zalloc(socket_zone, flag); 145 if (so) { 146#ifdef MAC 147 error = mac_init_socket(so, flag); 148 if (error != 0) { 149 uma_zfree(socket_zone, so); 150 so = NULL; 151 return so; 152 } 153#endif 154 /* XXX race condition for reentrant kernel */ 155 so->so_gencnt = ++so_gencnt; 156 /* sx_init(&so->so_sxlock, "socket sxlock"); */ 157 TAILQ_INIT(&so->so_aiojobq); 158 ++numopensockets; 159 } 160 return so; 161} 162 163/* 164 * socreate returns a socket with a ref count of 1. The socket should be 165 * closed with soclose(). 166 */ 167int 168socreate(dom, aso, type, proto, cred, td) 169 int dom; 170 struct socket **aso; 171 register int type; 172 int proto; 173 struct ucred *cred; 174 struct thread *td; 175{ 176 register struct protosw *prp; 177 register struct socket *so; 178 register int error; 179 180 if (proto) 181 prp = pffindproto(dom, proto, type); 182 else 183 prp = pffindtype(dom, type); 184 185 if (prp == 0 || prp->pr_usrreqs->pru_attach == 0) 186 return (EPROTONOSUPPORT); 187 188 if (jailed(cred) && jail_socket_unixiproute_only && 189 prp->pr_domain->dom_family != PF_LOCAL && 190 prp->pr_domain->dom_family != PF_INET && 191 prp->pr_domain->dom_family != PF_ROUTE) { 192 return (EPROTONOSUPPORT); 193 } 194 195 if (prp->pr_type != type) 196 return (EPROTOTYPE); 197 so = soalloc(M_NOWAIT); 198 if (so == NULL) 199 return (ENOBUFS); 200 201 TAILQ_INIT(&so->so_incomp); 202 TAILQ_INIT(&so->so_comp); 203 so->so_type = type; 204 so->so_cred = crhold(cred); 205 so->so_proto = prp; 206#ifdef MAC 207 mac_create_socket(cred, so); 208#endif 209 soref(so); 210 error = (*prp->pr_usrreqs->pru_attach)(so, proto, td); 211 if (error) { 212 so->so_state |= SS_NOFDREF; 213 sorele(so); 214 return (error); 215 } 216 *aso = so; 217 return (0); 218} 219 220int 221sobind(so, nam, td) 222 struct socket *so; 223 struct sockaddr *nam; 224 struct thread *td; 225{ 226 int s = splnet(); 227 int error; 228 229 error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam, td); 230 splx(s); 231 return (error); 232} 233 234void 235sodealloc(struct socket *so) 236{ 237 238 KASSERT(so->so_count == 0, ("sodealloc(): so_count %d", so->so_count)); 239 so->so_gencnt = ++so_gencnt; 240 if (so->so_rcv.sb_hiwat) 241 (void)chgsbsize(so->so_cred->cr_uidinfo, 242 &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY); 243 if (so->so_snd.sb_hiwat) 244 (void)chgsbsize(so->so_cred->cr_uidinfo, 245 &so->so_snd.sb_hiwat, 0, RLIM_INFINITY); 246#ifdef INET 247 /* remove acccept filter if one is present. */ 248 if (so->so_accf != NULL) 249 do_setopt_accept_filter(so, NULL); 250#endif 251#ifdef MAC 252 mac_destroy_socket(so); 253#endif 254 crfree(so->so_cred); 255 /* sx_destroy(&so->so_sxlock); */ 256 uma_zfree(socket_zone, so); 257 --numopensockets; 258} 259 260int 261solisten(so, backlog, td) 262 register struct socket *so; 263 int backlog; 264 struct thread *td; 265{ 266 int s, error; 267 268 s = splnet(); 269 error = (*so->so_proto->pr_usrreqs->pru_listen)(so, td); 270 if (error) { 271 splx(s); 272 return (error); 273 } 274 if (TAILQ_EMPTY(&so->so_comp)) 275 so->so_options |= SO_ACCEPTCONN; 276 if (backlog < 0 || backlog > somaxconn) 277 backlog = somaxconn; 278 so->so_qlimit = backlog; 279 splx(s); 280 return (0); 281} 282 283void 284sofree(so) 285 register struct socket *so; 286{ 287 struct socket *head = so->so_head; 288 289 KASSERT(so->so_count == 0, ("socket %p so_count not 0", so)); 290 291 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) 292 return; 293 if (head != NULL) { 294 if (so->so_state & SS_INCOMP) { 295 TAILQ_REMOVE(&head->so_incomp, so, so_list); 296 head->so_incqlen--; 297 } else if (so->so_state & SS_COMP) { 298 /* 299 * We must not decommission a socket that's 300 * on the accept(2) queue. If we do, then 301 * accept(2) may hang after select(2) indicated 302 * that the listening socket was ready. 303 */ 304 return; 305 } else { 306 panic("sofree: not queued"); 307 } 308 so->so_state &= ~SS_INCOMP; 309 so->so_head = NULL; 310 } 311 sbrelease(&so->so_snd, so); 312 sorflush(so); 313 sodealloc(so); 314} 315 316/* 317 * Close a socket on last file table reference removal. 318 * Initiate disconnect if connected. 319 * Free socket when disconnect complete. 320 * 321 * This function will sorele() the socket. Note that soclose() may be 322 * called prior to the ref count reaching zero. The actual socket 323 * structure will not be freed until the ref count reaches zero. 324 */ 325int 326soclose(so) 327 register struct socket *so; 328{ 329 int s = splnet(); /* conservative */ 330 int error = 0; 331 332 funsetown(&so->so_sigio); 333 if (so->so_options & SO_ACCEPTCONN) { 334 struct socket *sp, *sonext; 335 336 sp = TAILQ_FIRST(&so->so_incomp); 337 for (; sp != NULL; sp = sonext) { 338 sonext = TAILQ_NEXT(sp, so_list); 339 (void) soabort(sp); 340 } 341 for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) { 342 sonext = TAILQ_NEXT(sp, so_list); 343 /* Dequeue from so_comp since sofree() won't do it */ 344 TAILQ_REMOVE(&so->so_comp, sp, so_list); 345 so->so_qlen--; 346 sp->so_state &= ~SS_COMP; 347 sp->so_head = NULL; 348 (void) soabort(sp); 349 } 350 } 351 if (so->so_pcb == 0) 352 goto discard; 353 if (so->so_state & SS_ISCONNECTED) { 354 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 355 error = sodisconnect(so); 356 if (error) 357 goto drop; 358 } 359 if (so->so_options & SO_LINGER) { 360 if ((so->so_state & SS_ISDISCONNECTING) && 361 (so->so_state & SS_NBIO)) 362 goto drop; 363 while (so->so_state & SS_ISCONNECTED) { 364 error = tsleep(&so->so_timeo, 365 PSOCK | PCATCH, "soclos", so->so_linger * hz); 366 if (error) 367 break; 368 } 369 } 370 } 371drop: 372 if (so->so_pcb) { 373 int error2 = (*so->so_proto->pr_usrreqs->pru_detach)(so); 374 if (error == 0) 375 error = error2; 376 } 377discard: 378 if (so->so_state & SS_NOFDREF) 379 panic("soclose: NOFDREF"); 380 so->so_state |= SS_NOFDREF; 381 sorele(so); 382 splx(s); 383 return (error); 384} 385 386/* 387 * Must be called at splnet... 388 */ 389int 390soabort(so) 391 struct socket *so; 392{ 393 int error; 394 395 error = (*so->so_proto->pr_usrreqs->pru_abort)(so); 396 if (error) { 397 sotryfree(so); /* note: does not decrement the ref count */ 398 return error; 399 } 400 return (0); 401} 402 403int 404soaccept(so, nam) 405 register struct socket *so; 406 struct sockaddr **nam; 407{ 408 int s = splnet(); 409 int error; 410 411 if ((so->so_state & SS_NOFDREF) == 0) 412 panic("soaccept: !NOFDREF"); 413 so->so_state &= ~SS_NOFDREF; 414 error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam); 415 splx(s); 416 return (error); 417} 418 419int 420soconnect(so, nam, td) 421 register struct socket *so; 422 struct sockaddr *nam; 423 struct thread *td; 424{ 425 int s; 426 int error; 427 428 if (so->so_options & SO_ACCEPTCONN) 429 return (EOPNOTSUPP); 430 s = splnet(); 431 /* 432 * If protocol is connection-based, can only connect once. 433 * Otherwise, if connected, try to disconnect first. 434 * This allows user to disconnect by connecting to, e.g., 435 * a null address. 436 */ 437 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 438 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 439 (error = sodisconnect(so)))) 440 error = EISCONN; 441 else 442 error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, td); 443 splx(s); 444 return (error); 445} 446 447int 448soconnect2(so1, so2) 449 register struct socket *so1; 450 struct socket *so2; 451{ 452 int s = splnet(); 453 int error; 454 455 error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2); 456 splx(s); 457 return (error); 458} 459 460int 461sodisconnect(so) 462 register struct socket *so; 463{ 464 int s = splnet(); 465 int error; 466 467 if ((so->so_state & SS_ISCONNECTED) == 0) { 468 error = ENOTCONN; 469 goto bad; 470 } 471 if (so->so_state & SS_ISDISCONNECTING) { 472 error = EALREADY; 473 goto bad; 474 } 475 error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so); 476bad: 477 splx(s); 478 return (error); 479} 480 481#define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 482/* 483 * Send on a socket. 484 * If send must go all at once and message is larger than 485 * send buffering, then hard error. 486 * Lock against other senders. 487 * If must go all at once and not enough room now, then 488 * inform user that this would block and do nothing. 489 * Otherwise, if nonblocking, send as much as possible. 490 * The data to be sent is described by "uio" if nonzero, 491 * otherwise by the mbuf chain "top" (which must be null 492 * if uio is not). Data provided in mbuf chain must be small 493 * enough to send all at once. 494 * 495 * Returns nonzero on error, timeout or signal; callers 496 * must check for short counts if EINTR/ERESTART are returned. 497 * Data and control buffers are freed on return. 498 */ 499 500#ifdef ZERO_COPY_SOCKETS 501struct so_zerocopy_stats{ 502 int size_ok; 503 int align_ok; 504 int found_ifp; 505}; 506struct so_zerocopy_stats so_zerocp_stats = {0,0,0}; 507#include <netinet/in.h> 508#include <net/route.h> 509#include <netinet/in_pcb.h> 510#include <vm/vm.h> 511#include <vm/vm_page.h> 512#include <vm/vm_object.h> 513#endif /*ZERO_COPY_SOCKETS*/ 514 515int 516sosend(so, addr, uio, top, control, flags, td) 517 register struct socket *so; 518 struct sockaddr *addr; 519 struct uio *uio; 520 struct mbuf *top; 521 struct mbuf *control; 522 int flags; 523 struct thread *td; 524{ 525 struct mbuf **mp; 526 register struct mbuf *m; 527 register long space, len, resid; 528 int clen = 0, error, s, dontroute, mlen; 529 int atomic = sosendallatonce(so) || top; 530#ifdef ZERO_COPY_SOCKETS 531 int cow_send; 532#endif /* ZERO_COPY_SOCKETS */ 533 534 if (uio) 535 resid = uio->uio_resid; 536 else 537 resid = top->m_pkthdr.len; 538 /* 539 * In theory resid should be unsigned. 540 * However, space must be signed, as it might be less than 0 541 * if we over-committed, and we must use a signed comparison 542 * of space and resid. On the other hand, a negative resid 543 * causes us to loop sending 0-length segments to the protocol. 544 * 545 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM 546 * type sockets since that's an error. 547 */ 548 if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) { 549 error = EINVAL; 550 goto out; 551 } 552 553 dontroute = 554 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 555 (so->so_proto->pr_flags & PR_ATOMIC); 556 if (td) 557 td->td_proc->p_stats->p_ru.ru_msgsnd++; 558 if (control) 559 clen = control->m_len; 560#define snderr(errno) { error = (errno); splx(s); goto release; } 561 562restart: 563 error = sblock(&so->so_snd, SBLOCKWAIT(flags)); 564 if (error) 565 goto out; 566 do { 567 s = splnet(); 568 if (so->so_state & SS_CANTSENDMORE) 569 snderr(EPIPE); 570 if (so->so_error) { 571 error = so->so_error; 572 so->so_error = 0; 573 splx(s); 574 goto release; 575 } 576 if ((so->so_state & SS_ISCONNECTED) == 0) { 577 /* 578 * `sendto' and `sendmsg' is allowed on a connection- 579 * based socket if it supports implied connect. 580 * Return ENOTCONN if not connected and no address is 581 * supplied. 582 */ 583 if ((so->so_proto->pr_flags & PR_CONNREQUIRED) && 584 (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) { 585 if ((so->so_state & SS_ISCONFIRMING) == 0 && 586 !(resid == 0 && clen != 0)) 587 snderr(ENOTCONN); 588 } else if (addr == 0) 589 snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ? 590 ENOTCONN : EDESTADDRREQ); 591 } 592 space = sbspace(&so->so_snd); 593 if (flags & MSG_OOB) 594 space += 1024; 595 if ((atomic && resid > so->so_snd.sb_hiwat) || 596 clen > so->so_snd.sb_hiwat) 597 snderr(EMSGSIZE); 598 if (space < resid + clen && 599 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 600 if (so->so_state & SS_NBIO) 601 snderr(EWOULDBLOCK); 602 sbunlock(&so->so_snd); 603 error = sbwait(&so->so_snd); 604 splx(s); 605 if (error) 606 goto out; 607 goto restart; 608 } 609 splx(s); 610 mp = ⊤ 611 space -= clen; 612 do { 613 if (uio == NULL) { 614 /* 615 * Data is prepackaged in "top". 616 */ 617 resid = 0; 618 if (flags & MSG_EOR) 619 top->m_flags |= M_EOR; 620 } else do { 621#ifdef ZERO_COPY_SOCKETS 622 cow_send = 0; 623#endif /* ZERO_COPY_SOCKETS */ 624 if (top == 0) { 625 MGETHDR(m, M_TRYWAIT, MT_DATA); 626 if (m == NULL) { 627 error = ENOBUFS; 628 goto release; 629 } 630 mlen = MHLEN; 631 m->m_pkthdr.len = 0; 632 m->m_pkthdr.rcvif = (struct ifnet *)0; 633 } else { 634 MGET(m, M_TRYWAIT, MT_DATA); 635 if (m == NULL) { 636 error = ENOBUFS; 637 goto release; 638 } 639 mlen = MLEN; 640 } 641 if (resid >= MINCLSIZE) { 642#ifdef ZERO_COPY_SOCKETS 643 if (so_zero_copy_send && 644 resid>=PAGE_SIZE && 645 space>=PAGE_SIZE && 646 uio->uio_iov->iov_len>=PAGE_SIZE) { 647 so_zerocp_stats.size_ok++; 648 if (!((vm_offset_t) 649 uio->uio_iov->iov_base & PAGE_MASK)){ 650 so_zerocp_stats.align_ok++; 651 cow_send = socow_setup(m, uio); 652 } 653 } 654 if (!cow_send){ 655#endif /* ZERO_COPY_SOCKETS */ 656 MCLGET(m, M_TRYWAIT); 657 if ((m->m_flags & M_EXT) == 0) 658 goto nopages; 659 mlen = MCLBYTES; 660 len = min(min(mlen, resid), space); 661 } else { 662#ifdef ZERO_COPY_SOCKETS 663 len = PAGE_SIZE; 664 } 665 666 } else { 667#endif /* ZERO_COPY_SOCKETS */ 668nopages: 669 len = min(min(mlen, resid), space); 670 /* 671 * For datagram protocols, leave room 672 * for protocol headers in first mbuf. 673 */ 674 if (atomic && top == 0 && len < mlen) 675 MH_ALIGN(m, len); 676 } 677 space -= len; 678#ifdef ZERO_COPY_SOCKETS 679 if (cow_send) 680 error = 0; 681 else 682#endif /* ZERO_COPY_SOCKETS */ 683 error = uiomove(mtod(m, caddr_t), (int)len, uio); 684 resid = uio->uio_resid; 685 m->m_len = len; 686 *mp = m; 687 top->m_pkthdr.len += len; 688 if (error) 689 goto release; 690 mp = &m->m_next; 691 if (resid <= 0) { 692 if (flags & MSG_EOR) 693 top->m_flags |= M_EOR; 694 break; 695 } 696 } while (space > 0 && atomic); 697 if (dontroute) 698 so->so_options |= SO_DONTROUTE; 699 s = splnet(); /* XXX */ 700 /* 701 * XXX all the SS_CANTSENDMORE checks previously 702 * done could be out of date. We could have recieved 703 * a reset packet in an interrupt or maybe we slept 704 * while doing page faults in uiomove() etc. We could 705 * probably recheck again inside the splnet() protection 706 * here, but there are probably other places that this 707 * also happens. We must rethink this. 708 */ 709 error = (*so->so_proto->pr_usrreqs->pru_send)(so, 710 (flags & MSG_OOB) ? PRUS_OOB : 711 /* 712 * If the user set MSG_EOF, the protocol 713 * understands this flag and nothing left to 714 * send then use PRU_SEND_EOF instead of PRU_SEND. 715 */ 716 ((flags & MSG_EOF) && 717 (so->so_proto->pr_flags & PR_IMPLOPCL) && 718 (resid <= 0)) ? 719 PRUS_EOF : 720 /* If there is more to send set PRUS_MORETOCOME */ 721 (resid > 0 && space > 0) ? PRUS_MORETOCOME : 0, 722 top, addr, control, td); 723 splx(s); 724 if (dontroute) 725 so->so_options &= ~SO_DONTROUTE; 726 clen = 0; 727 control = 0; 728 top = 0; 729 mp = ⊤ 730 if (error) 731 goto release; 732 } while (resid && space > 0); 733 } while (resid); 734 735release: 736 sbunlock(&so->so_snd); 737out: 738 if (top) 739 m_freem(top); 740 if (control) 741 m_freem(control); 742 return (error); 743} 744 745/* 746 * Implement receive operations on a socket. 747 * We depend on the way that records are added to the sockbuf 748 * by sbappend*. In particular, each record (mbufs linked through m_next) 749 * must begin with an address if the protocol so specifies, 750 * followed by an optional mbuf or mbufs containing ancillary data, 751 * and then zero or more mbufs of data. 752 * In order to avoid blocking network interrupts for the entire time here, 753 * we splx() while doing the actual copy to user space. 754 * Although the sockbuf is locked, new data may still be appended, 755 * and thus we must maintain consistency of the sockbuf during that time. 756 * 757 * The caller may receive the data as a single mbuf chain by supplying 758 * an mbuf **mp0 for use in returning the chain. The uio is then used 759 * only for the count in uio_resid. 760 */ 761int 762soreceive(so, psa, uio, mp0, controlp, flagsp) 763 register struct socket *so; 764 struct sockaddr **psa; 765 struct uio *uio; 766 struct mbuf **mp0; 767 struct mbuf **controlp; 768 int *flagsp; 769{ 770 struct mbuf *m, **mp; 771 register int flags, len, error, s, offset; 772 struct protosw *pr = so->so_proto; 773 struct mbuf *nextrecord; 774 int moff, type = 0; 775 int orig_resid = uio->uio_resid; 776 777 mp = mp0; 778 if (psa) 779 *psa = 0; 780 if (controlp) 781 *controlp = 0; 782 if (flagsp) 783 flags = *flagsp &~ MSG_EOR; 784 else 785 flags = 0; 786 if (flags & MSG_OOB) { 787 m = m_get(M_TRYWAIT, MT_DATA); 788 if (m == NULL) 789 return (ENOBUFS); 790 error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK); 791 if (error) 792 goto bad; 793 do { 794#ifdef ZERO_COPY_SOCKETS 795 if (so_zero_copy_receive) { 796 vm_page_t pg; 797 int disposable; 798 799 if ((m->m_flags & M_EXT) 800 && (m->m_ext.ext_type == EXT_DISPOSABLE)) 801 disposable = 1; 802 else 803 disposable = 0; 804 805 pg = PHYS_TO_VM_PAGE(vtophys(mtod(m, caddr_t))); 806 if (uio->uio_offset == -1) 807 uio->uio_offset =IDX_TO_OFF(pg->pindex); 808 809 error = uiomoveco(mtod(m, caddr_t), 810 min(uio->uio_resid, m->m_len), 811 uio, pg->object, 812 disposable); 813 } else 814#endif /* ZERO_COPY_SOCKETS */ 815 error = uiomove(mtod(m, caddr_t), 816 (int) min(uio->uio_resid, m->m_len), uio); 817 m = m_free(m); 818 } while (uio->uio_resid && error == 0 && m); 819bad: 820 if (m) 821 m_freem(m); 822 return (error); 823 } 824 if (mp) 825 *mp = (struct mbuf *)0; 826 if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) 827 (*pr->pr_usrreqs->pru_rcvd)(so, 0); 828 829restart: 830 error = sblock(&so->so_rcv, SBLOCKWAIT(flags)); 831 if (error) 832 return (error); 833 s = splnet(); 834 835 m = so->so_rcv.sb_mb; 836 /* 837 * If we have less data than requested, block awaiting more 838 * (subject to any timeout) if: 839 * 1. the current count is less than the low water mark, or 840 * 2. MSG_WAITALL is set, and it is possible to do the entire 841 * receive operation at once if we block (resid <= hiwat). 842 * 3. MSG_DONTWAIT is not set 843 * If MSG_WAITALL is set but resid is larger than the receive buffer, 844 * we have to do the receive in sections, and thus risk returning 845 * a short count if a timeout or signal occurs after we start. 846 */ 847 if (m == 0 || (((flags & MSG_DONTWAIT) == 0 && 848 so->so_rcv.sb_cc < uio->uio_resid) && 849 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 850 ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) && 851 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) { 852 KASSERT(m != 0 || !so->so_rcv.sb_cc, 853 ("receive: m == %p so->so_rcv.sb_cc == %u", 854 m, so->so_rcv.sb_cc)); 855 if (so->so_error) { 856 if (m) 857 goto dontblock; 858 error = so->so_error; 859 if ((flags & MSG_PEEK) == 0) 860 so->so_error = 0; 861 goto release; 862 } 863 if (so->so_state & SS_CANTRCVMORE) { 864 if (m) 865 goto dontblock; 866 else 867 goto release; 868 } 869 for (; m; m = m->m_next) 870 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 871 m = so->so_rcv.sb_mb; 872 goto dontblock; 873 } 874 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 875 (so->so_proto->pr_flags & PR_CONNREQUIRED)) { 876 error = ENOTCONN; 877 goto release; 878 } 879 if (uio->uio_resid == 0) 880 goto release; 881 if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) { 882 error = EWOULDBLOCK; 883 goto release; 884 } 885 sbunlock(&so->so_rcv); 886 error = sbwait(&so->so_rcv); 887 splx(s); 888 if (error) 889 return (error); 890 goto restart; 891 } 892dontblock: 893 if (uio->uio_td) 894 uio->uio_td->td_proc->p_stats->p_ru.ru_msgrcv++; 895 nextrecord = m->m_nextpkt; 896 if (pr->pr_flags & PR_ADDR) { 897 KASSERT(m->m_type == MT_SONAME, 898 ("m->m_type == %d", m->m_type)); 899 orig_resid = 0; 900 if (psa) 901 *psa = dup_sockaddr(mtod(m, struct sockaddr *), 902 mp0 == 0); 903 if (flags & MSG_PEEK) { 904 m = m->m_next; 905 } else { 906 sbfree(&so->so_rcv, m); 907 so->so_rcv.sb_mb = m_free(m); 908 m = so->so_rcv.sb_mb; 909 } 910 } 911 while (m && m->m_type == MT_CONTROL && error == 0) { 912 if (flags & MSG_PEEK) { 913 if (controlp) 914 *controlp = m_copy(m, 0, m->m_len); 915 m = m->m_next; 916 } else { 917 sbfree(&so->so_rcv, m); 918 so->so_rcv.sb_mb = m->m_next; 919 m->m_next = NULL; 920 if (pr->pr_domain->dom_externalize) 921 error = 922 (*pr->pr_domain->dom_externalize)(m, controlp); 923 else if (controlp) 924 *controlp = m; 925 else 926 m_freem(m); 927 m = so->so_rcv.sb_mb; 928 } 929 if (controlp) { 930 orig_resid = 0; 931 do 932 controlp = &(*controlp)->m_next; 933 while (*controlp != NULL); 934 } 935 } 936 if (m) { 937 if ((flags & MSG_PEEK) == 0) 938 m->m_nextpkt = nextrecord; 939 type = m->m_type; 940 if (type == MT_OOBDATA) 941 flags |= MSG_OOB; 942 } 943 moff = 0; 944 offset = 0; 945 while (m && uio->uio_resid > 0 && error == 0) { 946 if (m->m_type == MT_OOBDATA) { 947 if (type != MT_OOBDATA) 948 break; 949 } else if (type == MT_OOBDATA) 950 break; 951 else 952 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER, 953 ("m->m_type == %d", m->m_type)); 954 so->so_state &= ~SS_RCVATMARK; 955 len = uio->uio_resid; 956 if (so->so_oobmark && len > so->so_oobmark - offset) 957 len = so->so_oobmark - offset; 958 if (len > m->m_len - moff) 959 len = m->m_len - moff; 960 /* 961 * If mp is set, just pass back the mbufs. 962 * Otherwise copy them out via the uio, then free. 963 * Sockbuf must be consistent here (points to current mbuf, 964 * it points to next record) when we drop priority; 965 * we must note any additions to the sockbuf when we 966 * block interrupts again. 967 */ 968 if (mp == 0) { 969 splx(s); 970#ifdef ZERO_COPY_SOCKETS 971 if (so_zero_copy_receive) { 972 vm_page_t pg; 973 int disposable; 974 975 if ((m->m_flags & M_EXT) 976 && (m->m_ext.ext_type == EXT_DISPOSABLE)) 977 disposable = 1; 978 else 979 disposable = 0; 980 981 pg = PHYS_TO_VM_PAGE(vtophys(mtod(m, caddr_t) + 982 moff)); 983 984 if (uio->uio_offset == -1) 985 uio->uio_offset =IDX_TO_OFF(pg->pindex); 986 987 error = uiomoveco(mtod(m, caddr_t) + moff, 988 (int)len, uio,pg->object, 989 disposable); 990 } else 991#endif /* ZERO_COPY_SOCKETS */ 992 error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio); 993 s = splnet(); 994 if (error) 995 goto release; 996 } else 997 uio->uio_resid -= len; 998 if (len == m->m_len - moff) { 999 if (m->m_flags & M_EOR) 1000 flags |= MSG_EOR; 1001 if (flags & MSG_PEEK) { 1002 m = m->m_next; 1003 moff = 0; 1004 } else { 1005 nextrecord = m->m_nextpkt; 1006 sbfree(&so->so_rcv, m); 1007 if (mp) { 1008 *mp = m; 1009 mp = &m->m_next; 1010 so->so_rcv.sb_mb = m = m->m_next; 1011 *mp = (struct mbuf *)0; 1012 } else { 1013 so->so_rcv.sb_mb = m_free(m); 1014 m = so->so_rcv.sb_mb; 1015 } 1016 if (m) 1017 m->m_nextpkt = nextrecord; 1018 } 1019 } else { 1020 if (flags & MSG_PEEK) 1021 moff += len; 1022 else { 1023 if (mp) 1024 *mp = m_copym(m, 0, len, M_TRYWAIT); 1025 m->m_data += len; 1026 m->m_len -= len; 1027 so->so_rcv.sb_cc -= len; 1028 } 1029 } 1030 if (so->so_oobmark) { 1031 if ((flags & MSG_PEEK) == 0) { 1032 so->so_oobmark -= len; 1033 if (so->so_oobmark == 0) { 1034 so->so_state |= SS_RCVATMARK; 1035 break; 1036 } 1037 } else { 1038 offset += len; 1039 if (offset == so->so_oobmark) 1040 break; 1041 } 1042 } 1043 if (flags & MSG_EOR) 1044 break; 1045 /* 1046 * If the MSG_WAITALL flag is set (for non-atomic socket), 1047 * we must not quit until "uio->uio_resid == 0" or an error 1048 * termination. If a signal/timeout occurs, return 1049 * with a short count but without error. 1050 * Keep sockbuf locked against other readers. 1051 */ 1052 while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 && 1053 !sosendallatonce(so) && !nextrecord) { 1054 if (so->so_error || so->so_state & SS_CANTRCVMORE) 1055 break; 1056 /* 1057 * Notify the protocol that some data has been 1058 * drained before blocking. 1059 */ 1060 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1061 (*pr->pr_usrreqs->pru_rcvd)(so, flags); 1062 error = sbwait(&so->so_rcv); 1063 if (error) { 1064 sbunlock(&so->so_rcv); 1065 splx(s); 1066 return (0); 1067 } 1068 m = so->so_rcv.sb_mb; 1069 if (m) 1070 nextrecord = m->m_nextpkt; 1071 } 1072 } 1073 1074 if (m && pr->pr_flags & PR_ATOMIC) { 1075 flags |= MSG_TRUNC; 1076 if ((flags & MSG_PEEK) == 0) 1077 (void) sbdroprecord(&so->so_rcv); 1078 } 1079 if ((flags & MSG_PEEK) == 0) { 1080 if (m == 0) 1081 so->so_rcv.sb_mb = nextrecord; 1082 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1083 (*pr->pr_usrreqs->pru_rcvd)(so, flags); 1084 } 1085 if (orig_resid == uio->uio_resid && orig_resid && 1086 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 1087 sbunlock(&so->so_rcv); 1088 splx(s); 1089 goto restart; 1090 } 1091 1092 if (flagsp) 1093 *flagsp |= flags; 1094release: 1095 sbunlock(&so->so_rcv); 1096 splx(s); 1097 return (error); 1098} 1099 1100int 1101soshutdown(so, how) 1102 register struct socket *so; 1103 register int how; 1104{ 1105 register struct protosw *pr = so->so_proto; 1106 1107 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 1108 return (EINVAL); 1109 1110 if (how != SHUT_WR) 1111 sorflush(so); 1112 if (how != SHUT_RD) 1113 return ((*pr->pr_usrreqs->pru_shutdown)(so)); 1114 return (0); 1115} 1116 1117void 1118sorflush(so) 1119 register struct socket *so; 1120{ 1121 register struct sockbuf *sb = &so->so_rcv; 1122 register struct protosw *pr = so->so_proto; 1123 register int s; 1124 struct sockbuf asb; 1125 1126 sb->sb_flags |= SB_NOINTR; 1127 (void) sblock(sb, M_WAITOK); 1128 s = splimp(); 1129 socantrcvmore(so); 1130 sbunlock(sb); 1131 asb = *sb; 1132 bzero(sb, sizeof (*sb)); 1133 splx(s); 1134 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) 1135 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 1136 sbrelease(&asb, so); 1137} 1138 1139#ifdef INET 1140static int 1141do_setopt_accept_filter(so, sopt) 1142 struct socket *so; 1143 struct sockopt *sopt; 1144{ 1145 struct accept_filter_arg *afap = NULL; 1146 struct accept_filter *afp; 1147 struct so_accf *af = so->so_accf; 1148 int error = 0; 1149 1150 /* do not set/remove accept filters on non listen sockets */ 1151 if ((so->so_options & SO_ACCEPTCONN) == 0) { 1152 error = EINVAL; 1153 goto out; 1154 } 1155 1156 /* removing the filter */ 1157 if (sopt == NULL) { 1158 if (af != NULL) { 1159 if (af->so_accept_filter != NULL && 1160 af->so_accept_filter->accf_destroy != NULL) { 1161 af->so_accept_filter->accf_destroy(so); 1162 } 1163 if (af->so_accept_filter_str != NULL) { 1164 FREE(af->so_accept_filter_str, M_ACCF); 1165 } 1166 FREE(af, M_ACCF); 1167 so->so_accf = NULL; 1168 } 1169 so->so_options &= ~SO_ACCEPTFILTER; 1170 return (0); 1171 } 1172 /* adding a filter */ 1173 /* must remove previous filter first */ 1174 if (af != NULL) { 1175 error = EINVAL; 1176 goto out; 1177 } 1178 /* don't put large objects on the kernel stack */ 1179 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK); 1180 error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap); 1181 afap->af_name[sizeof(afap->af_name)-1] = '\0'; 1182 afap->af_arg[sizeof(afap->af_arg)-1] = '\0'; 1183 if (error) 1184 goto out; 1185 afp = accept_filt_get(afap->af_name); 1186 if (afp == NULL) { 1187 error = ENOENT; 1188 goto out; 1189 } 1190 MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK | M_ZERO); 1191 if (afp->accf_create != NULL) { 1192 if (afap->af_name[0] != '\0') { 1193 int len = strlen(afap->af_name) + 1; 1194 1195 MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK); 1196 strcpy(af->so_accept_filter_str, afap->af_name); 1197 } 1198 af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg); 1199 if (af->so_accept_filter_arg == NULL) { 1200 FREE(af->so_accept_filter_str, M_ACCF); 1201 FREE(af, M_ACCF); 1202 so->so_accf = NULL; 1203 error = EINVAL; 1204 goto out; 1205 } 1206 } 1207 af->so_accept_filter = afp; 1208 so->so_accf = af; 1209 so->so_options |= SO_ACCEPTFILTER; 1210out: 1211 if (afap != NULL) 1212 FREE(afap, M_TEMP); 1213 return (error); 1214} 1215#endif /* INET */ 1216 1217/* 1218 * Perhaps this routine, and sooptcopyout(), below, ought to come in 1219 * an additional variant to handle the case where the option value needs 1220 * to be some kind of integer, but not a specific size. 1221 * In addition to their use here, these functions are also called by the 1222 * protocol-level pr_ctloutput() routines. 1223 */ 1224int 1225sooptcopyin(sopt, buf, len, minlen) 1226 struct sockopt *sopt; 1227 void *buf; 1228 size_t len; 1229 size_t minlen; 1230{ 1231 size_t valsize; 1232 1233 /* 1234 * If the user gives us more than we wanted, we ignore it, 1235 * but if we don't get the minimum length the caller 1236 * wants, we return EINVAL. On success, sopt->sopt_valsize 1237 * is set to however much we actually retrieved. 1238 */ 1239 if ((valsize = sopt->sopt_valsize) < minlen) 1240 return EINVAL; 1241 if (valsize > len) 1242 sopt->sopt_valsize = valsize = len; 1243 1244 if (sopt->sopt_td != 0) 1245 return (copyin(sopt->sopt_val, buf, valsize)); 1246 1247 bcopy(sopt->sopt_val, buf, valsize); 1248 return 0; 1249} 1250 1251int 1252sosetopt(so, sopt) 1253 struct socket *so; 1254 struct sockopt *sopt; 1255{ 1256 int error, optval; 1257 struct linger l; 1258 struct timeval tv; 1259 u_long val; 1260#ifdef MAC 1261 struct mac extmac; 1262#endif 1263 1264 error = 0; 1265 if (sopt->sopt_level != SOL_SOCKET) { 1266 if (so->so_proto && so->so_proto->pr_ctloutput) 1267 return ((*so->so_proto->pr_ctloutput) 1268 (so, sopt)); 1269 error = ENOPROTOOPT; 1270 } else { 1271 switch (sopt->sopt_name) { 1272#ifdef INET 1273 case SO_ACCEPTFILTER: 1274 error = do_setopt_accept_filter(so, sopt); 1275 if (error) 1276 goto bad; 1277 break; 1278#endif 1279 case SO_LINGER: 1280 error = sooptcopyin(sopt, &l, sizeof l, sizeof l); 1281 if (error) 1282 goto bad; 1283 1284 so->so_linger = l.l_linger; 1285 if (l.l_onoff) 1286 so->so_options |= SO_LINGER; 1287 else 1288 so->so_options &= ~SO_LINGER; 1289 break; 1290 1291 case SO_DEBUG: 1292 case SO_KEEPALIVE: 1293 case SO_DONTROUTE: 1294 case SO_USELOOPBACK: 1295 case SO_BROADCAST: 1296 case SO_REUSEADDR: 1297 case SO_REUSEPORT: 1298 case SO_OOBINLINE: 1299 case SO_TIMESTAMP: 1300 case SO_NOSIGPIPE: 1301 error = sooptcopyin(sopt, &optval, sizeof optval, 1302 sizeof optval); 1303 if (error) 1304 goto bad; 1305 if (optval) 1306 so->so_options |= sopt->sopt_name; 1307 else 1308 so->so_options &= ~sopt->sopt_name; 1309 break; 1310 1311 case SO_SNDBUF: 1312 case SO_RCVBUF: 1313 case SO_SNDLOWAT: 1314 case SO_RCVLOWAT: 1315 error = sooptcopyin(sopt, &optval, sizeof optval, 1316 sizeof optval); 1317 if (error) 1318 goto bad; 1319 1320 /* 1321 * Values < 1 make no sense for any of these 1322 * options, so disallow them. 1323 */ 1324 if (optval < 1) { 1325 error = EINVAL; 1326 goto bad; 1327 } 1328 1329 switch (sopt->sopt_name) { 1330 case SO_SNDBUF: 1331 case SO_RCVBUF: 1332 if (sbreserve(sopt->sopt_name == SO_SNDBUF ? 1333 &so->so_snd : &so->so_rcv, (u_long)optval, 1334 so, curthread) == 0) { 1335 error = ENOBUFS; 1336 goto bad; 1337 } 1338 break; 1339 1340 /* 1341 * Make sure the low-water is never greater than 1342 * the high-water. 1343 */ 1344 case SO_SNDLOWAT: 1345 so->so_snd.sb_lowat = 1346 (optval > so->so_snd.sb_hiwat) ? 1347 so->so_snd.sb_hiwat : optval; 1348 break; 1349 case SO_RCVLOWAT: 1350 so->so_rcv.sb_lowat = 1351 (optval > so->so_rcv.sb_hiwat) ? 1352 so->so_rcv.sb_hiwat : optval; 1353 break; 1354 } 1355 break; 1356 1357 case SO_SNDTIMEO: 1358 case SO_RCVTIMEO: 1359 error = sooptcopyin(sopt, &tv, sizeof tv, 1360 sizeof tv); 1361 if (error) 1362 goto bad; 1363 1364 /* assert(hz > 0); */ 1365 if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz || 1366 tv.tv_usec < 0 || tv.tv_usec >= 1000000) { 1367 error = EDOM; 1368 goto bad; 1369 } 1370 /* assert(tick > 0); */ 1371 /* assert(ULONG_MAX - SHRT_MAX >= 1000000); */ 1372 val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick; 1373 if (val > SHRT_MAX) { 1374 error = EDOM; 1375 goto bad; 1376 } 1377 if (val == 0 && tv.tv_usec != 0) 1378 val = 1; 1379 1380 switch (sopt->sopt_name) { 1381 case SO_SNDTIMEO: 1382 so->so_snd.sb_timeo = val; 1383 break; 1384 case SO_RCVTIMEO: 1385 so->so_rcv.sb_timeo = val; 1386 break; 1387 } 1388 break; 1389 case SO_LABEL: 1390#ifdef MAC 1391 error = sooptcopyin(sopt, &extmac, sizeof extmac, 1392 sizeof extmac); 1393 if (error) 1394 goto bad; 1395 1396 error = mac_setsockopt_label_set( 1397 sopt->sopt_td->td_ucred, so, &extmac); 1398 1399#else 1400 error = EOPNOTSUPP; 1401#endif 1402 break; 1403 default: 1404 error = ENOPROTOOPT; 1405 break; 1406 } 1407 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) { 1408 (void) ((*so->so_proto->pr_ctloutput) 1409 (so, sopt)); 1410 } 1411 } 1412bad: 1413 return (error); 1414} 1415 1416/* Helper routine for getsockopt */ 1417int 1418sooptcopyout(sopt, buf, len) 1419 struct sockopt *sopt; 1420 void *buf; 1421 size_t len; 1422{ 1423 int error; 1424 size_t valsize; 1425 1426 error = 0; 1427 1428 /* 1429 * Documented get behavior is that we always return a value, 1430 * possibly truncated to fit in the user's buffer. 1431 * Traditional behavior is that we always tell the user 1432 * precisely how much we copied, rather than something useful 1433 * like the total amount we had available for her. 1434 * Note that this interface is not idempotent; the entire answer must 1435 * generated ahead of time. 1436 */ 1437 valsize = min(len, sopt->sopt_valsize); 1438 sopt->sopt_valsize = valsize; 1439 if (sopt->sopt_val != 0) { 1440 if (sopt->sopt_td != 0) 1441 error = copyout(buf, sopt->sopt_val, valsize); 1442 else 1443 bcopy(buf, sopt->sopt_val, valsize); 1444 } 1445 return error; 1446} 1447 1448int 1449sogetopt(so, sopt) 1450 struct socket *so; 1451 struct sockopt *sopt; 1452{ 1453 int error, optval; 1454 struct linger l; 1455 struct timeval tv; 1456#ifdef INET 1457 struct accept_filter_arg *afap; 1458#endif 1459#ifdef MAC 1460 struct mac extmac; 1461#endif 1462 1463 error = 0; 1464 if (sopt->sopt_level != SOL_SOCKET) { 1465 if (so->so_proto && so->so_proto->pr_ctloutput) { 1466 return ((*so->so_proto->pr_ctloutput) 1467 (so, sopt)); 1468 } else 1469 return (ENOPROTOOPT); 1470 } else { 1471 switch (sopt->sopt_name) { 1472#ifdef INET 1473 case SO_ACCEPTFILTER: 1474 if ((so->so_options & SO_ACCEPTCONN) == 0) 1475 return (EINVAL); 1476 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), 1477 M_TEMP, M_WAITOK | M_ZERO); 1478 if ((so->so_options & SO_ACCEPTFILTER) != 0) { 1479 strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name); 1480 if (so->so_accf->so_accept_filter_str != NULL) 1481 strcpy(afap->af_arg, so->so_accf->so_accept_filter_str); 1482 } 1483 error = sooptcopyout(sopt, afap, sizeof(*afap)); 1484 FREE(afap, M_TEMP); 1485 break; 1486#endif 1487 1488 case SO_LINGER: 1489 l.l_onoff = so->so_options & SO_LINGER; 1490 l.l_linger = so->so_linger; 1491 error = sooptcopyout(sopt, &l, sizeof l); 1492 break; 1493 1494 case SO_USELOOPBACK: 1495 case SO_DONTROUTE: 1496 case SO_DEBUG: 1497 case SO_KEEPALIVE: 1498 case SO_REUSEADDR: 1499 case SO_REUSEPORT: 1500 case SO_BROADCAST: 1501 case SO_OOBINLINE: 1502 case SO_TIMESTAMP: 1503 case SO_NOSIGPIPE: 1504 optval = so->so_options & sopt->sopt_name; 1505integer: 1506 error = sooptcopyout(sopt, &optval, sizeof optval); 1507 break; 1508 1509 case SO_TYPE: 1510 optval = so->so_type; 1511 goto integer; 1512 1513 case SO_ERROR: 1514 optval = so->so_error; 1515 so->so_error = 0; 1516 goto integer; 1517 1518 case SO_SNDBUF: 1519 optval = so->so_snd.sb_hiwat; 1520 goto integer; 1521 1522 case SO_RCVBUF: 1523 optval = so->so_rcv.sb_hiwat; 1524 goto integer; 1525 1526 case SO_SNDLOWAT: 1527 optval = so->so_snd.sb_lowat; 1528 goto integer; 1529 1530 case SO_RCVLOWAT: 1531 optval = so->so_rcv.sb_lowat; 1532 goto integer; 1533 1534 case SO_SNDTIMEO: 1535 case SO_RCVTIMEO: 1536 optval = (sopt->sopt_name == SO_SNDTIMEO ? 1537 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 1538 1539 tv.tv_sec = optval / hz; 1540 tv.tv_usec = (optval % hz) * tick; 1541 error = sooptcopyout(sopt, &tv, sizeof tv); 1542 break; 1543 case SO_LABEL: 1544#ifdef MAC 1545 error = mac_getsockopt_label_get( 1546 sopt->sopt_td->td_ucred, so, &extmac); 1547 if (error) 1548 return (error); 1549 error = sooptcopyout(sopt, &extmac, sizeof extmac); 1550#else 1551 error = EOPNOTSUPP; 1552#endif 1553 break; 1554 case SO_PEERLABEL: 1555#ifdef MAC 1556 error = mac_getsockopt_peerlabel_get( 1557 sopt->sopt_td->td_ucred, so, &extmac); 1558 if (error) 1559 return (error); 1560 error = sooptcopyout(sopt, &extmac, sizeof extmac); 1561#else 1562 error = EOPNOTSUPP; 1563#endif 1564 break; 1565 default: 1566 error = ENOPROTOOPT; 1567 break; 1568 } 1569 return (error); 1570 } 1571} 1572 1573/* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */ 1574int 1575soopt_getm(struct sockopt *sopt, struct mbuf **mp) 1576{ 1577 struct mbuf *m, *m_prev; 1578 int sopt_size = sopt->sopt_valsize; 1579 1580 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA); 1581 if (m == 0) 1582 return ENOBUFS; 1583 if (sopt_size > MLEN) { 1584 MCLGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT); 1585 if ((m->m_flags & M_EXT) == 0) { 1586 m_free(m); 1587 return ENOBUFS; 1588 } 1589 m->m_len = min(MCLBYTES, sopt_size); 1590 } else { 1591 m->m_len = min(MLEN, sopt_size); 1592 } 1593 sopt_size -= m->m_len; 1594 *mp = m; 1595 m_prev = m; 1596 1597 while (sopt_size) { 1598 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA); 1599 if (m == 0) { 1600 m_freem(*mp); 1601 return ENOBUFS; 1602 } 1603 if (sopt_size > MLEN) { 1604 MCLGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT); 1605 if ((m->m_flags & M_EXT) == 0) { 1606 m_freem(*mp); 1607 return ENOBUFS; 1608 } 1609 m->m_len = min(MCLBYTES, sopt_size); 1610 } else { 1611 m->m_len = min(MLEN, sopt_size); 1612 } 1613 sopt_size -= m->m_len; 1614 m_prev->m_next = m; 1615 m_prev = m; 1616 } 1617 return 0; 1618} 1619 1620/* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */ 1621int 1622soopt_mcopyin(struct sockopt *sopt, struct mbuf *m) 1623{ 1624 struct mbuf *m0 = m; 1625 1626 if (sopt->sopt_val == NULL) 1627 return 0; 1628 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1629 if (sopt->sopt_td != NULL) { 1630 int error; 1631 1632 error = copyin(sopt->sopt_val, mtod(m, char *), 1633 m->m_len); 1634 if (error != 0) { 1635 m_freem(m0); 1636 return(error); 1637 } 1638 } else 1639 bcopy(sopt->sopt_val, mtod(m, char *), m->m_len); 1640 sopt->sopt_valsize -= m->m_len; 1641 (caddr_t)sopt->sopt_val += m->m_len; 1642 m = m->m_next; 1643 } 1644 if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */ 1645 panic("ip6_sooptmcopyin"); 1646 return 0; 1647} 1648 1649/* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */ 1650int 1651soopt_mcopyout(struct sockopt *sopt, struct mbuf *m) 1652{ 1653 struct mbuf *m0 = m; 1654 size_t valsize = 0; 1655 1656 if (sopt->sopt_val == NULL) 1657 return 0; 1658 while (m != NULL && sopt->sopt_valsize >= m->m_len) { 1659 if (sopt->sopt_td != NULL) { 1660 int error; 1661 1662 error = copyout(mtod(m, char *), sopt->sopt_val, 1663 m->m_len); 1664 if (error != 0) { 1665 m_freem(m0); 1666 return(error); 1667 } 1668 } else 1669 bcopy(mtod(m, char *), sopt->sopt_val, m->m_len); 1670 sopt->sopt_valsize -= m->m_len; 1671 (caddr_t)sopt->sopt_val += m->m_len; 1672 valsize += m->m_len; 1673 m = m->m_next; 1674 } 1675 if (m != NULL) { 1676 /* enough soopt buffer should be given from user-land */ 1677 m_freem(m0); 1678 return(EINVAL); 1679 } 1680 sopt->sopt_valsize = valsize; 1681 return 0; 1682} 1683 1684void 1685sohasoutofband(so) 1686 register struct socket *so; 1687{ 1688 if (so->so_sigio != NULL) 1689 pgsigio(&so->so_sigio, SIGURG, 0); 1690 selwakeup(&so->so_rcv.sb_sel); 1691} 1692 1693int 1694sopoll(struct socket *so, int events, struct ucred *active_cred, 1695 struct thread *td) 1696{ 1697 int revents = 0; 1698 int s = splnet(); 1699 1700 if (events & (POLLIN | POLLRDNORM)) 1701 if (soreadable(so)) 1702 revents |= events & (POLLIN | POLLRDNORM); 1703 1704 if (events & POLLINIGNEOF) 1705 if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat || 1706 !TAILQ_EMPTY(&so->so_comp) || so->so_error) 1707 revents |= POLLINIGNEOF; 1708 1709 if (events & (POLLOUT | POLLWRNORM)) 1710 if (sowriteable(so)) 1711 revents |= events & (POLLOUT | POLLWRNORM); 1712 1713 if (events & (POLLPRI | POLLRDBAND)) 1714 if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) 1715 revents |= events & (POLLPRI | POLLRDBAND); 1716 1717 if (revents == 0) { 1718 if (events & 1719 (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM | 1720 POLLRDBAND)) { 1721 selrecord(td, &so->so_rcv.sb_sel); 1722 so->so_rcv.sb_flags |= SB_SEL; 1723 } 1724 1725 if (events & (POLLOUT | POLLWRNORM)) { 1726 selrecord(td, &so->so_snd.sb_sel); 1727 so->so_snd.sb_flags |= SB_SEL; 1728 } 1729 } 1730 1731 splx(s); 1732 return (revents); 1733} 1734 1735int 1736soo_kqfilter(struct file *fp, struct knote *kn) 1737{ 1738 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1739 struct sockbuf *sb; 1740 int s; 1741 1742 switch (kn->kn_filter) { 1743 case EVFILT_READ: 1744 if (so->so_options & SO_ACCEPTCONN) 1745 kn->kn_fop = &solisten_filtops; 1746 else 1747 kn->kn_fop = &soread_filtops; 1748 sb = &so->so_rcv; 1749 break; 1750 case EVFILT_WRITE: 1751 kn->kn_fop = &sowrite_filtops; 1752 sb = &so->so_snd; 1753 break; 1754 default: 1755 return (1); 1756 } 1757 1758 s = splnet(); 1759 SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext); 1760 sb->sb_flags |= SB_KNOTE; 1761 splx(s); 1762 return (0); 1763} 1764 1765static void 1766filt_sordetach(struct knote *kn) 1767{ 1768 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1769 int s = splnet(); 1770 1771 SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext); 1772 if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note)) 1773 so->so_rcv.sb_flags &= ~SB_KNOTE; 1774 splx(s); 1775} 1776 1777/*ARGSUSED*/ 1778static int 1779filt_soread(struct knote *kn, long hint) 1780{ 1781 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1782 1783 kn->kn_data = so->so_rcv.sb_cc - so->so_rcv.sb_ctl; 1784 if (so->so_state & SS_CANTRCVMORE) { 1785 kn->kn_flags |= EV_EOF; 1786 kn->kn_fflags = so->so_error; 1787 return (1); 1788 } 1789 if (so->so_error) /* temporary udp error */ 1790 return (1); 1791 if (kn->kn_sfflags & NOTE_LOWAT) 1792 return (kn->kn_data >= kn->kn_sdata); 1793 return (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat); 1794} 1795 1796static void 1797filt_sowdetach(struct knote *kn) 1798{ 1799 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1800 int s = splnet(); 1801 1802 SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext); 1803 if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note)) 1804 so->so_snd.sb_flags &= ~SB_KNOTE; 1805 splx(s); 1806} 1807 1808/*ARGSUSED*/ 1809static int 1810filt_sowrite(struct knote *kn, long hint) 1811{ 1812 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1813 1814 kn->kn_data = sbspace(&so->so_snd); 1815 if (so->so_state & SS_CANTSENDMORE) { 1816 kn->kn_flags |= EV_EOF; 1817 kn->kn_fflags = so->so_error; 1818 return (1); 1819 } 1820 if (so->so_error) /* temporary udp error */ 1821 return (1); 1822 if (((so->so_state & SS_ISCONNECTED) == 0) && 1823 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 1824 return (0); 1825 if (kn->kn_sfflags & NOTE_LOWAT) 1826 return (kn->kn_data >= kn->kn_sdata); 1827 return (kn->kn_data >= so->so_snd.sb_lowat); 1828} 1829 1830/*ARGSUSED*/ 1831static int 1832filt_solisten(struct knote *kn, long hint) 1833{ 1834 struct socket *so = (struct socket *)kn->kn_fp->f_data; 1835 1836 kn->kn_data = so->so_qlen; 1837 return (! TAILQ_EMPTY(&so->so_comp)); 1838} 1839 1840int 1841socheckuid(struct socket *so, uid_t uid) 1842{ 1843 1844 if (so == NULL) 1845 return (EPERM); 1846 if (so->so_cred->cr_uid == uid) 1847 return (0); 1848 return (EPERM); 1849} 1850