uipc_socket.c revision 1.56.2.1
1 /* $NetBSD: uipc_socket.c,v 1.56.2.1 2001/07/10 13:48:47 lukem Exp $ */ 2 3 /* 4 * Copyright (c) 1982, 1986, 1988, 1990, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed by the University of 18 * California, Berkeley and its contributors. 19 * 4. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * @(#)uipc_socket.c 8.6 (Berkeley) 5/2/95 36 */ 37 38 #include "opt_compat_sunos.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/proc.h> 43 #include <sys/file.h> 44 #include <sys/malloc.h> 45 #include <sys/mbuf.h> 46 #include <sys/domain.h> 47 #include <sys/kernel.h> 48 #include <sys/protosw.h> 49 #include <sys/socket.h> 50 #include <sys/socketvar.h> 51 #include <sys/signalvar.h> 52 #include <sys/resourcevar.h> 53 #include <sys/pool.h> 54 #include <sys/event.h> 55 56 static void filt_sordetach(struct knote *kn); 57 static int filt_soread(struct knote *kn, long hint); 58 static void filt_sowdetach(struct knote *kn); 59 static int filt_sowrite(struct knote *kn, long hint); 60 static int filt_solisten(struct knote *kn, long hint); 61 62 static struct filterops solisten_filtops = 63 { 1, NULL, filt_sordetach, filt_solisten }; 64 struct filterops soread_filtops = 65 { 1, NULL, filt_sordetach, filt_soread }; 66 struct filterops sowrite_filtops = 67 { 1, NULL, filt_sowdetach, filt_sowrite }; 68 69 struct pool socket_pool; 70 71 extern int somaxconn; /* patchable (XXX sysctl) */ 72 int somaxconn = SOMAXCONN; 73 74 void 75 soinit(void) 76 { 77 78 pool_init(&socket_pool, sizeof(struct socket), 0, 0, 0, 79 "sockpl", 0, NULL, NULL, M_SOCKET); 80 } 81 82 /* 83 * Socket operation routines. 84 * These routines are called by the routines in 85 * sys_socket.c or from a system process, and 86 * implement the semantics of socket operations by 87 * switching out to the protocol specific routines. 88 */ 89 /*ARGSUSED*/ 90 int 91 socreate(int dom, struct socket **aso, int type, int proto) 92 { 93 struct proc *p; 94 struct protosw *prp; 95 struct socket *so; 96 int error, s; 97 98 p = curproc; /* XXX */ 99 if (proto) 100 prp = pffindproto(dom, proto, type); 101 else 102 prp = pffindtype(dom, type); 103 if (prp == 0 || prp->pr_usrreq == 0) 104 return (EPROTONOSUPPORT); 105 if (prp->pr_type != type) 106 return (EPROTOTYPE); 107 s = splsoftnet(); 108 so = pool_get(&socket_pool, PR_WAITOK); 109 memset((caddr_t)so, 0, sizeof(*so)); 110 TAILQ_INIT(&so->so_q0); 111 TAILQ_INIT(&so->so_q); 112 so->so_type = type; 113 so->so_proto = prp; 114 so->so_send = sosend; 115 so->so_receive = soreceive; 116 if (p != 0) 117 so->so_uid = p->p_ucred->cr_uid; 118 error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0, 119 (struct mbuf *)(long)proto, (struct mbuf *)0, p); 120 if (error) { 121 so->so_state |= SS_NOFDREF; 122 sofree(so); 123 splx(s); 124 return (error); 125 } 126 #ifdef COMPAT_SUNOS 127 { 128 extern struct emul emul_sunos; 129 if (p->p_emul == &emul_sunos && type == SOCK_DGRAM) 130 so->so_options |= SO_BROADCAST; 131 } 132 #endif 133 splx(s); 134 *aso = so; 135 return (0); 136 } 137 138 int 139 sobind(struct socket *so, struct mbuf *nam, struct proc *p) 140 { 141 int s, error; 142 143 s = splsoftnet(); 144 error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0, 145 nam, (struct mbuf *)0, p); 146 splx(s); 147 return (error); 148 } 149 150 int 151 solisten(struct socket *so, int backlog) 152 { 153 int s, error; 154 155 s = splsoftnet(); 156 error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0, 157 (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0); 158 if (error) { 159 splx(s); 160 return (error); 161 } 162 if (so->so_q.tqh_first == NULL) 163 so->so_options |= SO_ACCEPTCONN; 164 if (backlog < 0) 165 backlog = 0; 166 so->so_qlimit = min(backlog, somaxconn); 167 splx(s); 168 return (0); 169 } 170 171 void 172 sofree(struct socket *so) 173 { 174 175 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) 176 return; 177 if (so->so_head) { 178 /* 179 * We must not decommission a socket that's on the accept(2) 180 * queue. If we do, then accept(2) may hang after select(2) 181 * indicated that the listening socket was ready. 182 */ 183 if (!soqremque(so, 0)) 184 return; 185 } 186 sbrelease(&so->so_snd); 187 sorflush(so); 188 pool_put(&socket_pool, so); 189 } 190 191 /* 192 * Close a socket on last file table reference removal. 193 * Initiate disconnect if connected. 194 * Free socket when disconnect complete. 195 */ 196 int 197 soclose(struct socket *so) 198 { 199 struct socket *so2; 200 int s, error; 201 202 error = 0; 203 s = splsoftnet(); /* conservative */ 204 if (so->so_options & SO_ACCEPTCONN) { 205 while ((so2 = so->so_q0.tqh_first) != 0) { 206 (void) soqremque(so2, 0); 207 (void) soabort(so2); 208 } 209 while ((so2 = so->so_q.tqh_first) != 0) { 210 (void) soqremque(so2, 1); 211 (void) soabort(so2); 212 } 213 } 214 if (so->so_pcb == 0) 215 goto discard; 216 if (so->so_state & SS_ISCONNECTED) { 217 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 218 error = sodisconnect(so); 219 if (error) 220 goto drop; 221 } 222 if (so->so_options & SO_LINGER) { 223 if ((so->so_state & SS_ISDISCONNECTING) && 224 (so->so_state & SS_NBIO)) 225 goto drop; 226 while (so->so_state & SS_ISCONNECTED) { 227 error = tsleep((caddr_t)&so->so_timeo, 228 PSOCK | PCATCH, netcls, 229 so->so_linger * hz); 230 if (error) 231 break; 232 } 233 } 234 } 235 drop: 236 if (so->so_pcb) { 237 int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH, 238 (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0, 239 (struct proc *)0); 240 if (error == 0) 241 error = error2; 242 } 243 discard: 244 if (so->so_state & SS_NOFDREF) 245 panic("soclose: NOFDREF"); 246 so->so_state |= SS_NOFDREF; 247 sofree(so); 248 splx(s); 249 return (error); 250 } 251 252 /* 253 * Must be called at splsoftnet... 254 */ 255 int 256 soabort(struct socket *so) 257 { 258 259 return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0, 260 (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0); 261 } 262 263 int 264 soaccept(struct socket *so, struct mbuf *nam) 265 { 266 int s, error; 267 268 error = 0; 269 s = splsoftnet(); 270 if ((so->so_state & SS_NOFDREF) == 0) 271 panic("soaccept: !NOFDREF"); 272 so->so_state &= ~SS_NOFDREF; 273 if ((so->so_state & SS_ISDISCONNECTED) == 0 || 274 (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0) 275 error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT, 276 (struct mbuf *)0, nam, (struct mbuf *)0, (struct proc *)0); 277 else 278 error = ECONNABORTED; 279 280 splx(s); 281 return (error); 282 } 283 284 int 285 soconnect(struct socket *so, struct mbuf *nam) 286 { 287 struct proc *p; 288 int s, error; 289 290 p = curproc; /* XXX */ 291 if (so->so_options & SO_ACCEPTCONN) 292 return (EOPNOTSUPP); 293 s = splsoftnet(); 294 /* 295 * If protocol is connection-based, can only connect once. 296 * Otherwise, if connected, try to disconnect first. 297 * This allows user to disconnect by connecting to, e.g., 298 * a null address. 299 */ 300 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 301 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 302 (error = sodisconnect(so)))) 303 error = EISCONN; 304 else 305 error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT, 306 (struct mbuf *)0, nam, (struct mbuf *)0, p); 307 splx(s); 308 return (error); 309 } 310 311 int 312 soconnect2(struct socket *so1, struct socket *so2) 313 { 314 int s, error; 315 316 s = splsoftnet(); 317 error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2, 318 (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0, 319 (struct proc *)0); 320 splx(s); 321 return (error); 322 } 323 324 int 325 sodisconnect(struct socket *so) 326 { 327 int s, error; 328 329 s = splsoftnet(); 330 if ((so->so_state & SS_ISCONNECTED) == 0) { 331 error = ENOTCONN; 332 goto bad; 333 } 334 if (so->so_state & SS_ISDISCONNECTING) { 335 error = EALREADY; 336 goto bad; 337 } 338 error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT, 339 (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0, 340 (struct proc *)0); 341 bad: 342 splx(s); 343 return (error); 344 } 345 346 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 347 /* 348 * Send on a socket. 349 * If send must go all at once and message is larger than 350 * send buffering, then hard error. 351 * Lock against other senders. 352 * If must go all at once and not enough room now, then 353 * inform user that this would block and do nothing. 354 * Otherwise, if nonblocking, send as much as possible. 355 * The data to be sent is described by "uio" if nonzero, 356 * otherwise by the mbuf chain "top" (which must be null 357 * if uio is not). Data provided in mbuf chain must be small 358 * enough to send all at once. 359 * 360 * Returns nonzero on error, timeout or signal; callers 361 * must check for short counts if EINTR/ERESTART are returned. 362 * Data and control buffers are freed on return. 363 */ 364 int 365 sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top, 366 struct mbuf *control, int flags) 367 { 368 struct proc *p; 369 struct mbuf **mp, *m; 370 long space, len, resid; 371 int clen, error, s, dontroute, mlen, atomic; 372 373 p = curproc; /* XXX */ 374 clen = 0; 375 atomic = sosendallatonce(so) || top; 376 if (uio) 377 resid = uio->uio_resid; 378 else 379 resid = top->m_pkthdr.len; 380 /* 381 * In theory resid should be unsigned. 382 * However, space must be signed, as it might be less than 0 383 * if we over-committed, and we must use a signed comparison 384 * of space and resid. On the other hand, a negative resid 385 * causes us to loop sending 0-length segments to the protocol. 386 */ 387 if (resid < 0) { 388 error = EINVAL; 389 goto out; 390 } 391 dontroute = 392 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 393 (so->so_proto->pr_flags & PR_ATOMIC); 394 p->p_stats->p_ru.ru_msgsnd++; 395 if (control) 396 clen = control->m_len; 397 #define snderr(errno) { error = errno; splx(s); goto release; } 398 399 restart: 400 if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) 401 goto out; 402 do { 403 s = splsoftnet(); 404 if (so->so_state & SS_CANTSENDMORE) 405 snderr(EPIPE); 406 if (so->so_error) { 407 error = so->so_error; 408 so->so_error = 0; 409 splx(s); 410 goto release; 411 } 412 if ((so->so_state & SS_ISCONNECTED) == 0) { 413 if (so->so_proto->pr_flags & PR_CONNREQUIRED) { 414 if ((so->so_state & SS_ISCONFIRMING) == 0 && 415 !(resid == 0 && clen != 0)) 416 snderr(ENOTCONN); 417 } else if (addr == 0) 418 snderr(EDESTADDRREQ); 419 } 420 space = sbspace(&so->so_snd); 421 if (flags & MSG_OOB) 422 space += 1024; 423 if ((atomic && resid > so->so_snd.sb_hiwat) || 424 clen > so->so_snd.sb_hiwat) 425 snderr(EMSGSIZE); 426 if (space < resid + clen && uio && 427 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 428 if (so->so_state & SS_NBIO) 429 snderr(EWOULDBLOCK); 430 sbunlock(&so->so_snd); 431 error = sbwait(&so->so_snd); 432 splx(s); 433 if (error) 434 goto out; 435 goto restart; 436 } 437 splx(s); 438 mp = ⊤ 439 space -= clen; 440 do { 441 if (uio == NULL) { 442 /* 443 * Data is prepackaged in "top". 444 */ 445 resid = 0; 446 if (flags & MSG_EOR) 447 top->m_flags |= M_EOR; 448 } else do { 449 if (top == 0) { 450 MGETHDR(m, M_WAIT, MT_DATA); 451 mlen = MHLEN; 452 m->m_pkthdr.len = 0; 453 m->m_pkthdr.rcvif = (struct ifnet *)0; 454 } else { 455 MGET(m, M_WAIT, MT_DATA); 456 mlen = MLEN; 457 } 458 if (resid >= MINCLSIZE && space >= MCLBYTES) { 459 MCLGET(m, M_WAIT); 460 if ((m->m_flags & M_EXT) == 0) 461 goto nopages; 462 mlen = MCLBYTES; 463 #ifdef MAPPED_MBUFS 464 len = min(MCLBYTES, resid); 465 #else 466 if (atomic && top == 0) { 467 len = min(MCLBYTES - max_hdr, 468 resid); 469 m->m_data += max_hdr; 470 } else 471 len = min(MCLBYTES, resid); 472 #endif 473 space -= len; 474 } else { 475 nopages: 476 len = min(min(mlen, resid), space); 477 space -= len; 478 /* 479 * For datagram protocols, leave room 480 * for protocol headers in first mbuf. 481 */ 482 if (atomic && top == 0 && len < mlen) 483 MH_ALIGN(m, len); 484 } 485 error = uiomove(mtod(m, caddr_t), (int)len, 486 uio); 487 resid = uio->uio_resid; 488 m->m_len = len; 489 *mp = m; 490 top->m_pkthdr.len += len; 491 if (error) 492 goto release; 493 mp = &m->m_next; 494 if (resid <= 0) { 495 if (flags & MSG_EOR) 496 top->m_flags |= M_EOR; 497 break; 498 } 499 } while (space > 0 && atomic); 500 501 s = splsoftnet(); 502 503 if (so->so_state & SS_CANTSENDMORE) 504 snderr(EPIPE); 505 506 if (dontroute) 507 so->so_options |= SO_DONTROUTE; 508 if (resid > 0) 509 so->so_state |= SS_MORETOCOME; 510 error = (*so->so_proto->pr_usrreq)(so, 511 (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND, 512 top, addr, control, p); 513 if (dontroute) 514 so->so_options &= ~SO_DONTROUTE; 515 if (resid > 0) 516 so->so_state &= ~SS_MORETOCOME; 517 splx(s); 518 519 clen = 0; 520 control = 0; 521 top = 0; 522 mp = ⊤ 523 if (error) 524 goto release; 525 } while (resid && space > 0); 526 } while (resid); 527 528 release: 529 sbunlock(&so->so_snd); 530 out: 531 if (top) 532 m_freem(top); 533 if (control) 534 m_freem(control); 535 return (error); 536 } 537 538 /* 539 * Implement receive operations on a socket. 540 * We depend on the way that records are added to the sockbuf 541 * by sbappend*. In particular, each record (mbufs linked through m_next) 542 * must begin with an address if the protocol so specifies, 543 * followed by an optional mbuf or mbufs containing ancillary data, 544 * and then zero or more mbufs of data. 545 * In order to avoid blocking network interrupts for the entire time here, 546 * we splx() while doing the actual copy to user space. 547 * Although the sockbuf is locked, new data may still be appended, 548 * and thus we must maintain consistency of the sockbuf during that time. 549 * 550 * The caller may receive the data as a single mbuf chain by supplying 551 * an mbuf **mp0 for use in returning the chain. The uio is then used 552 * only for the count in uio_resid. 553 */ 554 int 555 soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, 556 struct mbuf **mp0, struct mbuf **controlp, int *flagsp) 557 { 558 struct mbuf *m, **mp; 559 int flags, len, error, s, offset, moff, type, orig_resid; 560 struct protosw *pr; 561 struct mbuf *nextrecord; 562 563 pr = so->so_proto; 564 mp = mp0; 565 type = 0; 566 orig_resid = uio->uio_resid; 567 if (paddr) 568 *paddr = 0; 569 if (controlp) 570 *controlp = 0; 571 if (flagsp) 572 flags = *flagsp &~ MSG_EOR; 573 else 574 flags = 0; 575 if (flags & MSG_OOB) { 576 m = m_get(M_WAIT, MT_DATA); 577 error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m, 578 (struct mbuf *)(long)(flags & MSG_PEEK), (struct mbuf *)0, 579 (struct proc *)0); 580 if (error) 581 goto bad; 582 do { 583 error = uiomove(mtod(m, caddr_t), 584 (int) min(uio->uio_resid, m->m_len), uio); 585 m = m_free(m); 586 } while (uio->uio_resid && error == 0 && m); 587 bad: 588 if (m) 589 m_freem(m); 590 return (error); 591 } 592 if (mp) 593 *mp = (struct mbuf *)0; 594 if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) 595 (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0, 596 (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0); 597 598 restart: 599 if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) 600 return (error); 601 s = splsoftnet(); 602 603 m = so->so_rcv.sb_mb; 604 /* 605 * If we have less data than requested, block awaiting more 606 * (subject to any timeout) if: 607 * 1. the current count is less than the low water mark, 608 * 2. MSG_WAITALL is set, and it is possible to do the entire 609 * receive operation at once if we block (resid <= hiwat), or 610 * 3. MSG_DONTWAIT is not set. 611 * If MSG_WAITALL is set but resid is larger than the receive buffer, 612 * we have to do the receive in sections, and thus risk returning 613 * a short count if a timeout or signal occurs after we start. 614 */ 615 if (m == 0 || (((flags & MSG_DONTWAIT) == 0 && 616 so->so_rcv.sb_cc < uio->uio_resid) && 617 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 618 ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) && 619 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) { 620 #ifdef DIAGNOSTIC 621 if (m == 0 && so->so_rcv.sb_cc) 622 panic("receive 1"); 623 #endif 624 if (so->so_error) { 625 if (m) 626 goto dontblock; 627 error = so->so_error; 628 if ((flags & MSG_PEEK) == 0) 629 so->so_error = 0; 630 goto release; 631 } 632 if (so->so_state & SS_CANTRCVMORE) { 633 if (m) 634 goto dontblock; 635 else 636 goto release; 637 } 638 for (; m; m = m->m_next) 639 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 640 m = so->so_rcv.sb_mb; 641 goto dontblock; 642 } 643 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 644 (so->so_proto->pr_flags & PR_CONNREQUIRED)) { 645 error = ENOTCONN; 646 goto release; 647 } 648 if (uio->uio_resid == 0) 649 goto release; 650 if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) { 651 error = EWOULDBLOCK; 652 goto release; 653 } 654 sbunlock(&so->so_rcv); 655 error = sbwait(&so->so_rcv); 656 splx(s); 657 if (error) 658 return (error); 659 goto restart; 660 } 661 dontblock: 662 #ifdef notyet /* XXXX */ 663 if (uio->uio_procp) 664 uio->uio_procp->p_stats->p_ru.ru_msgrcv++; 665 #endif 666 nextrecord = m->m_nextpkt; 667 if (pr->pr_flags & PR_ADDR) { 668 #ifdef DIAGNOSTIC 669 if (m->m_type != MT_SONAME) 670 panic("receive 1a"); 671 #endif 672 orig_resid = 0; 673 if (flags & MSG_PEEK) { 674 if (paddr) 675 *paddr = m_copy(m, 0, m->m_len); 676 m = m->m_next; 677 } else { 678 sbfree(&so->so_rcv, m); 679 if (paddr) { 680 *paddr = m; 681 so->so_rcv.sb_mb = m->m_next; 682 m->m_next = 0; 683 m = so->so_rcv.sb_mb; 684 } else { 685 MFREE(m, so->so_rcv.sb_mb); 686 m = so->so_rcv.sb_mb; 687 } 688 } 689 } 690 while (m && m->m_type == MT_CONTROL && error == 0) { 691 if (flags & MSG_PEEK) { 692 if (controlp) 693 *controlp = m_copy(m, 0, m->m_len); 694 m = m->m_next; 695 } else { 696 sbfree(&so->so_rcv, m); 697 if (controlp) { 698 if (pr->pr_domain->dom_externalize && 699 mtod(m, struct cmsghdr *)->cmsg_type == 700 SCM_RIGHTS) 701 error = (*pr->pr_domain->dom_externalize)(m); 702 *controlp = m; 703 so->so_rcv.sb_mb = m->m_next; 704 m->m_next = 0; 705 m = so->so_rcv.sb_mb; 706 } else { 707 MFREE(m, so->so_rcv.sb_mb); 708 m = so->so_rcv.sb_mb; 709 } 710 } 711 if (controlp) { 712 orig_resid = 0; 713 controlp = &(*controlp)->m_next; 714 } 715 } 716 if (m) { 717 if ((flags & MSG_PEEK) == 0) 718 m->m_nextpkt = nextrecord; 719 type = m->m_type; 720 if (type == MT_OOBDATA) 721 flags |= MSG_OOB; 722 } 723 moff = 0; 724 offset = 0; 725 while (m && uio->uio_resid > 0 && error == 0) { 726 if (m->m_type == MT_OOBDATA) { 727 if (type != MT_OOBDATA) 728 break; 729 } else if (type == MT_OOBDATA) 730 break; 731 #ifdef DIAGNOSTIC 732 else if (m->m_type != MT_DATA && m->m_type != MT_HEADER) 733 panic("receive 3"); 734 #endif 735 so->so_state &= ~SS_RCVATMARK; 736 len = uio->uio_resid; 737 if (so->so_oobmark && len > so->so_oobmark - offset) 738 len = so->so_oobmark - offset; 739 if (len > m->m_len - moff) 740 len = m->m_len - moff; 741 /* 742 * If mp is set, just pass back the mbufs. 743 * Otherwise copy them out via the uio, then free. 744 * Sockbuf must be consistent here (points to current mbuf, 745 * it points to next record) when we drop priority; 746 * we must note any additions to the sockbuf when we 747 * block interrupts again. 748 */ 749 if (mp == 0) { 750 splx(s); 751 error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio); 752 s = splsoftnet(); 753 } else 754 uio->uio_resid -= len; 755 if (len == m->m_len - moff) { 756 if (m->m_flags & M_EOR) 757 flags |= MSG_EOR; 758 if (flags & MSG_PEEK) { 759 m = m->m_next; 760 moff = 0; 761 } else { 762 nextrecord = m->m_nextpkt; 763 sbfree(&so->so_rcv, m); 764 if (mp) { 765 *mp = m; 766 mp = &m->m_next; 767 so->so_rcv.sb_mb = m = m->m_next; 768 *mp = (struct mbuf *)0; 769 } else { 770 MFREE(m, so->so_rcv.sb_mb); 771 m = so->so_rcv.sb_mb; 772 } 773 if (m) 774 m->m_nextpkt = nextrecord; 775 } 776 } else { 777 if (flags & MSG_PEEK) 778 moff += len; 779 else { 780 if (mp) 781 *mp = m_copym(m, 0, len, M_WAIT); 782 m->m_data += len; 783 m->m_len -= len; 784 so->so_rcv.sb_cc -= len; 785 } 786 } 787 if (so->so_oobmark) { 788 if ((flags & MSG_PEEK) == 0) { 789 so->so_oobmark -= len; 790 if (so->so_oobmark == 0) { 791 so->so_state |= SS_RCVATMARK; 792 break; 793 } 794 } else { 795 offset += len; 796 if (offset == so->so_oobmark) 797 break; 798 } 799 } 800 if (flags & MSG_EOR) 801 break; 802 /* 803 * If the MSG_WAITALL flag is set (for non-atomic socket), 804 * we must not quit until "uio->uio_resid == 0" or an error 805 * termination. If a signal/timeout occurs, return 806 * with a short count but without error. 807 * Keep sockbuf locked against other readers. 808 */ 809 while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 && 810 !sosendallatonce(so) && !nextrecord) { 811 if (so->so_error || so->so_state & SS_CANTRCVMORE) 812 break; 813 error = sbwait(&so->so_rcv); 814 if (error) { 815 sbunlock(&so->so_rcv); 816 splx(s); 817 return (0); 818 } 819 if ((m = so->so_rcv.sb_mb) != NULL) 820 nextrecord = m->m_nextpkt; 821 } 822 } 823 824 if (m && pr->pr_flags & PR_ATOMIC) { 825 flags |= MSG_TRUNC; 826 if ((flags & MSG_PEEK) == 0) 827 (void) sbdroprecord(&so->so_rcv); 828 } 829 if ((flags & MSG_PEEK) == 0) { 830 if (m == 0) 831 so->so_rcv.sb_mb = nextrecord; 832 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 833 (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0, 834 (struct mbuf *)(long)flags, (struct mbuf *)0, 835 (struct proc *)0); 836 } 837 if (orig_resid == uio->uio_resid && orig_resid && 838 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 839 sbunlock(&so->so_rcv); 840 splx(s); 841 goto restart; 842 } 843 844 if (flagsp) 845 *flagsp |= flags; 846 release: 847 sbunlock(&so->so_rcv); 848 splx(s); 849 return (error); 850 } 851 852 int 853 soshutdown(struct socket *so, int how) 854 { 855 struct protosw *pr; 856 857 pr = so->so_proto; 858 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 859 return (EINVAL); 860 861 if (how == SHUT_RD || how == SHUT_RDWR) 862 sorflush(so); 863 if (how == SHUT_WR || how == SHUT_RDWR) 864 return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0, 865 (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0); 866 return (0); 867 } 868 869 void 870 sorflush(struct socket *so) 871 { 872 struct sockbuf *sb, asb; 873 struct protosw *pr; 874 int s; 875 876 sb = &so->so_rcv; 877 pr = so->so_proto; 878 sb->sb_flags |= SB_NOINTR; 879 (void) sblock(sb, M_WAITOK); 880 s = splnet(); 881 socantrcvmore(so); 882 sbunlock(sb); 883 asb = *sb; 884 memset((caddr_t)sb, 0, sizeof(*sb)); 885 splx(s); 886 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) 887 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 888 sbrelease(&asb); 889 } 890 891 int 892 sosetopt(struct socket *so, int level, int optname, struct mbuf *m0) 893 { 894 int error; 895 struct mbuf *m; 896 897 error = 0; 898 m = m0; 899 if (level != SOL_SOCKET) { 900 if (so->so_proto && so->so_proto->pr_ctloutput) 901 return ((*so->so_proto->pr_ctloutput) 902 (PRCO_SETOPT, so, level, optname, &m0)); 903 error = ENOPROTOOPT; 904 } else { 905 switch (optname) { 906 907 case SO_LINGER: 908 if (m == NULL || m->m_len != sizeof(struct linger)) { 909 error = EINVAL; 910 goto bad; 911 } 912 so->so_linger = mtod(m, struct linger *)->l_linger; 913 /* fall thru... */ 914 915 case SO_DEBUG: 916 case SO_KEEPALIVE: 917 case SO_DONTROUTE: 918 case SO_USELOOPBACK: 919 case SO_BROADCAST: 920 case SO_REUSEADDR: 921 case SO_REUSEPORT: 922 case SO_OOBINLINE: 923 case SO_TIMESTAMP: 924 if (m == NULL || m->m_len < sizeof(int)) { 925 error = EINVAL; 926 goto bad; 927 } 928 if (*mtod(m, int *)) 929 so->so_options |= optname; 930 else 931 so->so_options &= ~optname; 932 break; 933 934 case SO_SNDBUF: 935 case SO_RCVBUF: 936 case SO_SNDLOWAT: 937 case SO_RCVLOWAT: 938 { 939 int optval; 940 941 if (m == NULL || m->m_len < sizeof(int)) { 942 error = EINVAL; 943 goto bad; 944 } 945 946 /* 947 * Values < 1 make no sense for any of these 948 * options, so disallow them. 949 */ 950 optval = *mtod(m, int *); 951 if (optval < 1) { 952 error = EINVAL; 953 goto bad; 954 } 955 956 switch (optname) { 957 958 case SO_SNDBUF: 959 case SO_RCVBUF: 960 if (sbreserve(optname == SO_SNDBUF ? 961 &so->so_snd : &so->so_rcv, 962 (u_long) optval) == 0) { 963 error = ENOBUFS; 964 goto bad; 965 } 966 break; 967 968 /* 969 * Make sure the low-water is never greater than 970 * the high-water. 971 */ 972 case SO_SNDLOWAT: 973 so->so_snd.sb_lowat = 974 (optval > so->so_snd.sb_hiwat) ? 975 so->so_snd.sb_hiwat : optval; 976 break; 977 case SO_RCVLOWAT: 978 so->so_rcv.sb_lowat = 979 (optval > so->so_rcv.sb_hiwat) ? 980 so->so_rcv.sb_hiwat : optval; 981 break; 982 } 983 break; 984 } 985 986 case SO_SNDTIMEO: 987 case SO_RCVTIMEO: 988 { 989 struct timeval *tv; 990 short val; 991 992 if (m == NULL || m->m_len < sizeof(*tv)) { 993 error = EINVAL; 994 goto bad; 995 } 996 tv = mtod(m, struct timeval *); 997 if (tv->tv_sec * hz + tv->tv_usec / tick > SHRT_MAX) { 998 error = EDOM; 999 goto bad; 1000 } 1001 val = tv->tv_sec * hz + tv->tv_usec / tick; 1002 1003 switch (optname) { 1004 1005 case SO_SNDTIMEO: 1006 so->so_snd.sb_timeo = val; 1007 break; 1008 case SO_RCVTIMEO: 1009 so->so_rcv.sb_timeo = val; 1010 break; 1011 } 1012 break; 1013 } 1014 1015 default: 1016 error = ENOPROTOOPT; 1017 break; 1018 } 1019 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) { 1020 (void) ((*so->so_proto->pr_ctloutput) 1021 (PRCO_SETOPT, so, level, optname, &m0)); 1022 m = NULL; /* freed by protocol */ 1023 } 1024 } 1025 bad: 1026 if (m) 1027 (void) m_free(m); 1028 return (error); 1029 } 1030 1031 int 1032 sogetopt(struct socket *so, int level, int optname, struct mbuf **mp) 1033 { 1034 struct mbuf *m; 1035 1036 if (level != SOL_SOCKET) { 1037 if (so->so_proto && so->so_proto->pr_ctloutput) { 1038 return ((*so->so_proto->pr_ctloutput) 1039 (PRCO_GETOPT, so, level, optname, mp)); 1040 } else 1041 return (ENOPROTOOPT); 1042 } else { 1043 m = m_get(M_WAIT, MT_SOOPTS); 1044 m->m_len = sizeof(int); 1045 1046 switch (optname) { 1047 1048 case SO_LINGER: 1049 m->m_len = sizeof(struct linger); 1050 mtod(m, struct linger *)->l_onoff = 1051 so->so_options & SO_LINGER; 1052 mtod(m, struct linger *)->l_linger = so->so_linger; 1053 break; 1054 1055 case SO_USELOOPBACK: 1056 case SO_DONTROUTE: 1057 case SO_DEBUG: 1058 case SO_KEEPALIVE: 1059 case SO_REUSEADDR: 1060 case SO_REUSEPORT: 1061 case SO_BROADCAST: 1062 case SO_OOBINLINE: 1063 case SO_TIMESTAMP: 1064 *mtod(m, int *) = so->so_options & optname; 1065 break; 1066 1067 case SO_TYPE: 1068 *mtod(m, int *) = so->so_type; 1069 break; 1070 1071 case SO_ERROR: 1072 *mtod(m, int *) = so->so_error; 1073 so->so_error = 0; 1074 break; 1075 1076 case SO_SNDBUF: 1077 *mtod(m, int *) = so->so_snd.sb_hiwat; 1078 break; 1079 1080 case SO_RCVBUF: 1081 *mtod(m, int *) = so->so_rcv.sb_hiwat; 1082 break; 1083 1084 case SO_SNDLOWAT: 1085 *mtod(m, int *) = so->so_snd.sb_lowat; 1086 break; 1087 1088 case SO_RCVLOWAT: 1089 *mtod(m, int *) = so->so_rcv.sb_lowat; 1090 break; 1091 1092 case SO_SNDTIMEO: 1093 case SO_RCVTIMEO: 1094 { 1095 int val = (optname == SO_SNDTIMEO ? 1096 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 1097 1098 m->m_len = sizeof(struct timeval); 1099 mtod(m, struct timeval *)->tv_sec = val / hz; 1100 mtod(m, struct timeval *)->tv_usec = 1101 (val % hz) * tick; 1102 break; 1103 } 1104 1105 default: 1106 (void)m_free(m); 1107 return (ENOPROTOOPT); 1108 } 1109 *mp = m; 1110 return (0); 1111 } 1112 } 1113 1114 void 1115 sohasoutofband(struct socket *so) 1116 { 1117 struct proc *p; 1118 1119 if (so->so_pgid < 0) 1120 gsignal(-so->so_pgid, SIGURG); 1121 else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0) 1122 psignal(p, SIGURG); 1123 selwakeup(&so->so_rcv.sb_sel); 1124 } 1125 1126 1127 int 1128 soo_kqfilter(struct file *fp, struct knote *kn) 1129 { 1130 struct socket *so; 1131 struct sockbuf *sb; 1132 int s; 1133 1134 so = (struct socket *)kn->kn_fp->f_data; 1135 switch (kn->kn_filter) { 1136 case EVFILT_READ: 1137 if (so->so_options & SO_ACCEPTCONN) 1138 kn->kn_fop = &solisten_filtops; 1139 else 1140 kn->kn_fop = &soread_filtops; 1141 sb = &so->so_rcv; 1142 break; 1143 case EVFILT_WRITE: 1144 kn->kn_fop = &sowrite_filtops; 1145 sb = &so->so_snd; 1146 break; 1147 default: 1148 return (1); 1149 } 1150 s = splnet(); /* XXXLUKEM: maybe splsoftnet() ? */ 1151 SLIST_INSERT_HEAD(&sb->sb_sel.si_klist, kn, kn_selnext); 1152 sb->sb_flags |= SB_KNOTE; 1153 splx(s); 1154 return (0); 1155 } 1156 1157 static void 1158 filt_sordetach(struct knote *kn) 1159 { 1160 struct socket *so; 1161 int s; 1162 1163 so = (struct socket *)kn->kn_fp->f_data; 1164 s = splnet(); /* XXXLUKEM: maybe splsoftnet() ? */ 1165 SLIST_REMOVE(&so->so_rcv.sb_sel.si_klist, kn, knote, kn_selnext); 1166 if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_klist)) 1167 so->so_rcv.sb_flags &= ~SB_KNOTE; 1168 splx(s); 1169 } 1170 1171 /*ARGSUSED*/ 1172 static int 1173 filt_soread(struct knote *kn, long hint) 1174 { 1175 struct socket *so; 1176 1177 so = (struct socket *)kn->kn_fp->f_data; 1178 kn->kn_data = so->so_rcv.sb_cc; 1179 if (so->so_state & SS_CANTRCVMORE) { 1180 kn->kn_flags |= EV_EOF; 1181 kn->kn_fflags = so->so_error; 1182 return (1); 1183 } 1184 if (so->so_error) /* temporary udp error */ 1185 return (1); 1186 if (kn->kn_sfflags & NOTE_LOWAT) 1187 return (kn->kn_data >= kn->kn_sdata); 1188 return (kn->kn_data >= so->so_rcv.sb_lowat); 1189 } 1190 1191 static void 1192 filt_sowdetach(struct knote *kn) 1193 { 1194 struct socket *so; 1195 int s; 1196 1197 so = (struct socket *)kn->kn_fp->f_data; 1198 s = splnet(); /* XXXLUKEM: maybe splsoftnet() ? */ 1199 SLIST_REMOVE(&so->so_snd.sb_sel.si_klist, kn, knote, kn_selnext); 1200 if (SLIST_EMPTY(&so->so_snd.sb_sel.si_klist)) 1201 so->so_snd.sb_flags &= ~SB_KNOTE; 1202 splx(s); 1203 } 1204 1205 /*ARGSUSED*/ 1206 static int 1207 filt_sowrite(struct knote *kn, long hint) 1208 { 1209 struct socket *so; 1210 1211 so = (struct socket *)kn->kn_fp->f_data; 1212 kn->kn_data = sbspace(&so->so_snd); 1213 if (so->so_state & SS_CANTSENDMORE) { 1214 kn->kn_flags |= EV_EOF; 1215 kn->kn_fflags = so->so_error; 1216 return (1); 1217 } 1218 if (so->so_error) /* temporary udp error */ 1219 return (1); 1220 if (((so->so_state & SS_ISCONNECTED) == 0) && 1221 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 1222 return (0); 1223 if (kn->kn_sfflags & NOTE_LOWAT) 1224 return (kn->kn_data >= kn->kn_sdata); 1225 return (kn->kn_data >= so->so_snd.sb_lowat); 1226 } 1227 1228 /*ARGSUSED*/ 1229 static int 1230 filt_solisten(struct knote *kn, long hint) 1231 { 1232 struct socket *so; 1233 1234 so = (struct socket *)kn->kn_fp->f_data; 1235 #if 0 1236 /* 1237 * XXXLUKEM: this was freebsd's code. it appears that they 1238 * XXXLUKEM: modified the socket code to store the count 1239 * XXXLUKEM: of all connections in so_qlen, and separately 1240 * XXXLUKEM: track the number of incompletes in so_incqlen. 1241 * XXXLUKEM: as we appear to keep only completed connections 1242 * XXXLUKEM: on so_qlen we can just return that. 1243 * XXXLUKEM: that said, a socket guru should double check for me :) 1244 */ 1245 kn->kn_data = so->so_qlen - so->so_incqlen; 1246 return (! TAILQ_EMPTY(&so->so_comp)); 1247 #else 1248 kn->kn_data = so->so_qlen; 1249 return (kn->kn_data > 0); 1250 #endif 1251 } 1252
Indexes created Sat Sep 20 22:09:52 GMT 2025