uipc_socket.c revision 1.56.2.7
1 /* $NetBSD: uipc_socket.c,v 1.56.2.7 2002/09/06 08:48:16 jdolecek Exp $ */ 2 3 /*- 4 * Copyright (c) 2002 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe of Wasabi Systems, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the NetBSD 21 * Foundation, Inc. and its contributors. 22 * 4. Neither the name of The NetBSD Foundation nor the names of its 23 * contributors may be used to endorse or promote products derived 24 * from this software without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 36 * POSSIBILITY OF SUCH DAMAGE. 37 */ 38 39 /* 40 * Copyright (c) 1982, 1986, 1988, 1990, 1993 41 * The Regents of the University of California. All rights reserved. 42 * 43 * Redistribution and use in source and binary forms, with or without 44 * modification, are permitted provided that the following conditions 45 * are met: 46 * 1. Redistributions of source code must retain the above copyright 47 * notice, this list of conditions and the following disclaimer. 48 * 2. Redistributions in binary form must reproduce the above copyright 49 * notice, this list of conditions and the following disclaimer in the 50 * documentation and/or other materials provided with the distribution. 51 * 3. All advertising materials mentioning features or use of this software 52 * must display the following acknowledgement: 53 * This product includes software developed by the University of 54 * California, Berkeley and its contributors. 55 * 4. Neither the name of the University nor the names of its contributors 56 * may be used to endorse or promote products derived from this software 57 * without specific prior written permission. 58 * 59 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 60 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 61 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 62 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 63 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 64 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 65 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 66 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 67 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 68 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 69 * SUCH DAMAGE. 70 * 71 * @(#)uipc_socket.c 8.6 (Berkeley) 5/2/95 72 */ 73 74 #include <sys/cdefs.h> 75 __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.56.2.7 2002/09/06 08:48:16 jdolecek Exp $"); 76 77 #include "opt_sock_counters.h" 78 #include "opt_sosend_loan.h" 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/proc.h> 83 #include <sys/file.h> 84 #include <sys/malloc.h> 85 #include <sys/mbuf.h> 86 #include <sys/domain.h> 87 #include <sys/kernel.h> 88 #include <sys/protosw.h> 89 #include <sys/socket.h> 90 #include <sys/socketvar.h> 91 #include <sys/signalvar.h> 92 #include <sys/resourcevar.h> 93 #include <sys/pool.h> 94 #include <sys/event.h> 95 96 #include <uvm/uvm.h> 97 98 static void filt_sordetach(struct knote *kn); 99 static int filt_soread(struct knote *kn, long hint); 100 static void filt_sowdetach(struct knote *kn); 101 static int filt_sowrite(struct knote *kn, long hint); 102 static int filt_solisten(struct knote *kn, long hint); 103 104 static const struct filterops solisten_filtops = 105 { 1, NULL, filt_sordetach, filt_solisten }; 106 const struct filterops soread_filtops = 107 { 1, NULL, filt_sordetach, filt_soread }; 108 const struct filterops sowrite_filtops = 109 { 1, NULL, filt_sowdetach, filt_sowrite }; 110 111 struct pool socket_pool; 112 113 extern int somaxconn; /* patchable (XXX sysctl) */ 114 int somaxconn = SOMAXCONN; 115 116 #ifdef SOSEND_COUNTERS 117 #include <sys/device.h> 118 119 struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 120 NULL, "sosend", "loan big"); 121 struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 122 NULL, "sosend", "copy big"); 123 struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 124 NULL, "sosend", "copy small"); 125 struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 126 NULL, "sosend", "kva limit"); 127 128 #define SOSEND_COUNTER_INCR(ev) (ev)->ev_count++ 129 130 #else 131 132 #define SOSEND_COUNTER_INCR(ev) /* nothing */ 133 134 #endif /* SOSEND_COUNTERS */ 135 136 void 137 soinit(void) 138 { 139 140 pool_init(&socket_pool, sizeof(struct socket), 0, 0, 0, 141 "sockpl", NULL); 142 143 #ifdef SOSEND_COUNTERS 144 evcnt_attach_static(&sosend_loan_big); 145 evcnt_attach_static(&sosend_copy_big); 146 evcnt_attach_static(&sosend_copy_small); 147 evcnt_attach_static(&sosend_kvalimit); 148 #endif /* SOSEND_COUNTERS */ 149 } 150 151 #ifdef SOSEND_NO_LOAN 152 int use_sosend_loan = 0; 153 #else 154 int use_sosend_loan = 1; 155 #endif 156 157 struct mbuf *so_pendfree; 158 159 int somaxkva = 16 * 1024 * 1024; 160 int socurkva; 161 int sokvawaiters; 162 163 #define SOCK_LOAN_THRESH 4096 164 #define SOCK_LOAN_CHUNK 65536 165 166 static void 167 sodoloanfree(caddr_t buf, u_int size) 168 { 169 struct vm_page **pgs; 170 vaddr_t va, sva, eva; 171 vsize_t len; 172 paddr_t pa; 173 int i, npgs; 174 175 eva = round_page((vaddr_t) buf + size); 176 sva = trunc_page((vaddr_t) buf); 177 len = eva - sva; 178 npgs = len >> PAGE_SHIFT; 179 180 pgs = alloca(npgs * sizeof(*pgs)); 181 182 for (i = 0, va = sva; va < eva; i++, va += PAGE_SIZE) { 183 if (pmap_extract(pmap_kernel(), va, &pa) == FALSE) 184 panic("sodoloanfree: va 0x%lx not mapped", va); 185 pgs[i] = PHYS_TO_VM_PAGE(pa); 186 } 187 188 pmap_kremove(sva, len); 189 pmap_update(pmap_kernel()); 190 uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE); 191 uvm_km_free(kernel_map, sva, len); 192 socurkva -= len; 193 if (sokvawaiters) 194 wakeup(&socurkva); 195 } 196 197 static size_t 198 sodopendfree(struct socket *so) 199 { 200 struct mbuf *m; 201 size_t rv = 0; 202 int s; 203 204 s = splvm(); 205 206 for (;;) { 207 m = so_pendfree; 208 if (m == NULL) 209 break; 210 so_pendfree = m->m_next; 211 splx(s); 212 213 rv += m->m_ext.ext_size; 214 sodoloanfree(m->m_ext.ext_buf, m->m_ext.ext_size); 215 s = splvm(); 216 pool_cache_put(&mbpool_cache, m); 217 } 218 219 for (;;) { 220 m = so->so_pendfree; 221 if (m == NULL) 222 break; 223 so->so_pendfree = m->m_next; 224 splx(s); 225 226 rv += m->m_ext.ext_size; 227 sodoloanfree(m->m_ext.ext_buf, m->m_ext.ext_size); 228 s = splvm(); 229 pool_cache_put(&mbpool_cache, m); 230 } 231 232 splx(s); 233 return (rv); 234 } 235 236 static void 237 soloanfree(struct mbuf *m, caddr_t buf, u_int size, void *arg) 238 { 239 struct socket *so = arg; 240 int s; 241 242 if (m == NULL) { 243 sodoloanfree(buf, size); 244 return; 245 } 246 247 s = splvm(); 248 m->m_next = so->so_pendfree; 249 so->so_pendfree = m; 250 splx(s); 251 if (sokvawaiters) 252 wakeup(&socurkva); 253 } 254 255 static long 256 sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space) 257 { 258 struct iovec *iov = uio->uio_iov; 259 vaddr_t sva, eva; 260 vsize_t len; 261 struct vm_page **pgs; 262 vaddr_t lva, va; 263 int npgs, s, i, error; 264 265 if (uio->uio_segflg != UIO_USERSPACE) 266 return (0); 267 268 if (iov->iov_len < (size_t) space) 269 space = iov->iov_len; 270 if (space > SOCK_LOAN_CHUNK) 271 space = SOCK_LOAN_CHUNK; 272 273 eva = round_page((vaddr_t) iov->iov_base + space); 274 sva = trunc_page((vaddr_t) iov->iov_base); 275 len = eva - sva; 276 npgs = len >> PAGE_SHIFT; 277 278 while (socurkva + len > somaxkva) { 279 if (sodopendfree(so)) 280 continue; 281 SOSEND_COUNTER_INCR(&sosend_kvalimit); 282 s = splvm(); 283 sokvawaiters++; 284 (void) tsleep(&socurkva, PVM, "sokva", 0); 285 sokvawaiters--; 286 splx(s); 287 } 288 289 lva = uvm_km_valloc_wait(kernel_map, len); 290 if (lva == 0) 291 return (0); 292 socurkva += len; 293 294 pgs = alloca(npgs * sizeof(*pgs)); 295 296 error = uvm_loan(&uio->uio_procp->p_vmspace->vm_map, sva, len, 297 pgs, UVM_LOAN_TOPAGE); 298 if (error) { 299 uvm_km_free(kernel_map, lva, len); 300 socurkva -= len; 301 return (0); 302 } 303 304 for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE) 305 pmap_kenter_pa(va, VM_PAGE_TO_PHYS(pgs[i]), VM_PROT_READ); 306 pmap_update(pmap_kernel()); 307 308 lva += (vaddr_t) iov->iov_base & PAGE_MASK; 309 310 MEXTADD(m, (caddr_t) lva, space, M_MBUF, soloanfree, so); 311 312 uio->uio_resid -= space; 313 /* uio_offset not updated, not set/used for write(2) */ 314 uio->uio_iov->iov_base = (caddr_t) uio->uio_iov->iov_base + space; 315 uio->uio_iov->iov_len -= space; 316 if (uio->uio_iov->iov_len == 0) { 317 uio->uio_iov++; 318 uio->uio_iovcnt--; 319 } 320 321 return (space); 322 } 323 324 /* 325 * Socket operation routines. 326 * These routines are called by the routines in 327 * sys_socket.c or from a system process, and 328 * implement the semantics of socket operations by 329 * switching out to the protocol specific routines. 330 */ 331 /*ARGSUSED*/ 332 int 333 socreate(int dom, struct socket **aso, int type, int proto) 334 { 335 struct proc *p; 336 struct protosw *prp; 337 struct socket *so; 338 int error, s; 339 340 p = curproc; /* XXX */ 341 if (proto) 342 prp = pffindproto(dom, proto, type); 343 else 344 prp = pffindtype(dom, type); 345 if (prp == 0 || prp->pr_usrreq == 0) 346 return (EPROTONOSUPPORT); 347 if (prp->pr_type != type) 348 return (EPROTOTYPE); 349 s = splsoftnet(); 350 so = pool_get(&socket_pool, PR_WAITOK); 351 memset((caddr_t)so, 0, sizeof(*so)); 352 TAILQ_INIT(&so->so_q0); 353 TAILQ_INIT(&so->so_q); 354 so->so_type = type; 355 so->so_proto = prp; 356 so->so_send = sosend; 357 so->so_receive = soreceive; 358 if (p != 0) 359 so->so_uid = p->p_ucred->cr_uid; 360 error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0, 361 (struct mbuf *)(long)proto, (struct mbuf *)0, p); 362 if (error) { 363 so->so_state |= SS_NOFDREF; 364 sofree(so); 365 splx(s); 366 return (error); 367 } 368 splx(s); 369 *aso = so; 370 return (0); 371 } 372 373 int 374 sobind(struct socket *so, struct mbuf *nam, struct proc *p) 375 { 376 int s, error; 377 378 s = splsoftnet(); 379 error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0, 380 nam, (struct mbuf *)0, p); 381 splx(s); 382 return (error); 383 } 384 385 int 386 solisten(struct socket *so, int backlog) 387 { 388 int s, error; 389 390 s = splsoftnet(); 391 error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0, 392 (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0); 393 if (error) { 394 splx(s); 395 return (error); 396 } 397 if (TAILQ_EMPTY(&so->so_q)) 398 so->so_options |= SO_ACCEPTCONN; 399 if (backlog < 0) 400 backlog = 0; 401 so->so_qlimit = min(backlog, somaxconn); 402 splx(s); 403 return (0); 404 } 405 406 void 407 sofree(struct socket *so) 408 { 409 struct mbuf *m; 410 411 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) 412 return; 413 if (so->so_head) { 414 /* 415 * We must not decommission a socket that's on the accept(2) 416 * queue. If we do, then accept(2) may hang after select(2) 417 * indicated that the listening socket was ready. 418 */ 419 if (!soqremque(so, 0)) 420 return; 421 } 422 sbrelease(&so->so_snd); 423 sorflush(so); 424 while ((m = so->so_pendfree) != NULL) { 425 so->so_pendfree = m->m_next; 426 m->m_next = so_pendfree; 427 so_pendfree = m; 428 } 429 pool_put(&socket_pool, so); 430 } 431 432 /* 433 * Close a socket on last file table reference removal. 434 * Initiate disconnect if connected. 435 * Free socket when disconnect complete. 436 */ 437 int 438 soclose(struct socket *so) 439 { 440 struct socket *so2; 441 int s, error; 442 443 error = 0; 444 s = splsoftnet(); /* conservative */ 445 if (so->so_options & SO_ACCEPTCONN) { 446 while ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) { 447 (void) soqremque(so2, 0); 448 (void) soabort(so2); 449 } 450 while ((so2 = TAILQ_FIRST(&so->so_q)) != 0) { 451 (void) soqremque(so2, 1); 452 (void) soabort(so2); 453 } 454 } 455 if (so->so_pcb == 0) 456 goto discard; 457 if (so->so_state & SS_ISCONNECTED) { 458 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 459 error = sodisconnect(so); 460 if (error) 461 goto drop; 462 } 463 if (so->so_options & SO_LINGER) { 464 if ((so->so_state & SS_ISDISCONNECTING) && 465 (so->so_state & SS_NBIO)) 466 goto drop; 467 while (so->so_state & SS_ISCONNECTED) { 468 error = tsleep((caddr_t)&so->so_timeo, 469 PSOCK | PCATCH, netcls, 470 so->so_linger * hz); 471 if (error) 472 break; 473 } 474 } 475 } 476 drop: 477 if (so->so_pcb) { 478 int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH, 479 (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0, 480 (struct proc *)0); 481 if (error == 0) 482 error = error2; 483 } 484 discard: 485 if (so->so_state & SS_NOFDREF) 486 panic("soclose: NOFDREF"); 487 so->so_state |= SS_NOFDREF; 488 sofree(so); 489 splx(s); 490 return (error); 491 } 492 493 /* 494 * Must be called at splsoftnet... 495 */ 496 int 497 soabort(struct socket *so) 498 { 499 500 return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0, 501 (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0); 502 } 503 504 int 505 soaccept(struct socket *so, struct mbuf *nam) 506 { 507 int s, error; 508 509 error = 0; 510 s = splsoftnet(); 511 if ((so->so_state & SS_NOFDREF) == 0) 512 panic("soaccept: !NOFDREF"); 513 so->so_state &= ~SS_NOFDREF; 514 if ((so->so_state & SS_ISDISCONNECTED) == 0 || 515 (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0) 516 error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT, 517 (struct mbuf *)0, nam, (struct mbuf *)0, (struct proc *)0); 518 else 519 error = ECONNABORTED; 520 521 splx(s); 522 return (error); 523 } 524 525 int 526 soconnect(struct socket *so, struct mbuf *nam) 527 { 528 struct proc *p; 529 int s, error; 530 531 p = curproc; /* XXX */ 532 if (so->so_options & SO_ACCEPTCONN) 533 return (EOPNOTSUPP); 534 s = splsoftnet(); 535 /* 536 * If protocol is connection-based, can only connect once. 537 * Otherwise, if connected, try to disconnect first. 538 * This allows user to disconnect by connecting to, e.g., 539 * a null address. 540 */ 541 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 542 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 543 (error = sodisconnect(so)))) 544 error = EISCONN; 545 else 546 error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT, 547 (struct mbuf *)0, nam, (struct mbuf *)0, p); 548 splx(s); 549 return (error); 550 } 551 552 int 553 soconnect2(struct socket *so1, struct socket *so2) 554 { 555 int s, error; 556 557 s = splsoftnet(); 558 error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2, 559 (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0, 560 (struct proc *)0); 561 splx(s); 562 return (error); 563 } 564 565 int 566 sodisconnect(struct socket *so) 567 { 568 int s, error; 569 570 s = splsoftnet(); 571 if ((so->so_state & SS_ISCONNECTED) == 0) { 572 error = ENOTCONN; 573 goto bad; 574 } 575 if (so->so_state & SS_ISDISCONNECTING) { 576 error = EALREADY; 577 goto bad; 578 } 579 error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT, 580 (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0, 581 (struct proc *)0); 582 bad: 583 splx(s); 584 sodopendfree(so); 585 return (error); 586 } 587 588 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 589 /* 590 * Send on a socket. 591 * If send must go all at once and message is larger than 592 * send buffering, then hard error. 593 * Lock against other senders. 594 * If must go all at once and not enough room now, then 595 * inform user that this would block and do nothing. 596 * Otherwise, if nonblocking, send as much as possible. 597 * The data to be sent is described by "uio" if nonzero, 598 * otherwise by the mbuf chain "top" (which must be null 599 * if uio is not). Data provided in mbuf chain must be small 600 * enough to send all at once. 601 * 602 * Returns nonzero on error, timeout or signal; callers 603 * must check for short counts if EINTR/ERESTART are returned. 604 * Data and control buffers are freed on return. 605 */ 606 int 607 sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top, 608 struct mbuf *control, int flags) 609 { 610 struct proc *p; 611 struct mbuf **mp, *m; 612 long space, len, resid, clen, mlen; 613 int error, s, dontroute, atomic; 614 615 sodopendfree(so); 616 617 p = curproc; /* XXX */ 618 clen = 0; 619 atomic = sosendallatonce(so) || top; 620 if (uio) 621 resid = uio->uio_resid; 622 else 623 resid = top->m_pkthdr.len; 624 /* 625 * In theory resid should be unsigned. 626 * However, space must be signed, as it might be less than 0 627 * if we over-committed, and we must use a signed comparison 628 * of space and resid. On the other hand, a negative resid 629 * causes us to loop sending 0-length segments to the protocol. 630 */ 631 if (resid < 0) { 632 error = EINVAL; 633 goto out; 634 } 635 dontroute = 636 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 637 (so->so_proto->pr_flags & PR_ATOMIC); 638 p->p_stats->p_ru.ru_msgsnd++; 639 if (control) 640 clen = control->m_len; 641 #define snderr(errno) { error = errno; splx(s); goto release; } 642 643 restart: 644 if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) 645 goto out; 646 do { 647 s = splsoftnet(); 648 if (so->so_state & SS_CANTSENDMORE) 649 snderr(EPIPE); 650 if (so->so_error) { 651 error = so->so_error; 652 so->so_error = 0; 653 splx(s); 654 goto release; 655 } 656 if ((so->so_state & SS_ISCONNECTED) == 0) { 657 if (so->so_proto->pr_flags & PR_CONNREQUIRED) { 658 if ((so->so_state & SS_ISCONFIRMING) == 0 && 659 !(resid == 0 && clen != 0)) 660 snderr(ENOTCONN); 661 } else if (addr == 0) 662 snderr(EDESTADDRREQ); 663 } 664 space = sbspace(&so->so_snd); 665 if (flags & MSG_OOB) 666 space += 1024; 667 if ((atomic && resid > so->so_snd.sb_hiwat) || 668 clen > so->so_snd.sb_hiwat) 669 snderr(EMSGSIZE); 670 if (space < resid + clen && uio && 671 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 672 if (so->so_state & SS_NBIO) 673 snderr(EWOULDBLOCK); 674 sbunlock(&so->so_snd); 675 error = sbwait(&so->so_snd); 676 splx(s); 677 if (error) 678 goto out; 679 goto restart; 680 } 681 splx(s); 682 mp = ⊤ 683 space -= clen; 684 do { 685 if (uio == NULL) { 686 /* 687 * Data is prepackaged in "top". 688 */ 689 resid = 0; 690 if (flags & MSG_EOR) 691 top->m_flags |= M_EOR; 692 } else do { 693 if (top == 0) { 694 MGETHDR(m, M_WAIT, MT_DATA); 695 mlen = MHLEN; 696 m->m_pkthdr.len = 0; 697 m->m_pkthdr.rcvif = (struct ifnet *)0; 698 } else { 699 MGET(m, M_WAIT, MT_DATA); 700 mlen = MLEN; 701 } 702 if (use_sosend_loan && 703 uio->uio_iov->iov_len >= SOCK_LOAN_THRESH && 704 space >= SOCK_LOAN_THRESH && 705 (len = sosend_loan(so, uio, m, 706 space)) != 0) { 707 SOSEND_COUNTER_INCR(&sosend_loan_big); 708 space -= len; 709 goto have_data; 710 } 711 if (resid >= MINCLSIZE && space >= MCLBYTES) { 712 SOSEND_COUNTER_INCR(&sosend_copy_big); 713 MCLGET(m, M_WAIT); 714 if ((m->m_flags & M_EXT) == 0) 715 goto nopages; 716 mlen = MCLBYTES; 717 if (atomic && top == 0) { 718 len = lmin(MCLBYTES - max_hdr, 719 resid); 720 m->m_data += max_hdr; 721 } else 722 len = lmin(MCLBYTES, resid); 723 space -= len; 724 } else { 725 nopages: 726 SOSEND_COUNTER_INCR(&sosend_copy_small); 727 len = lmin(lmin(mlen, resid), space); 728 space -= len; 729 /* 730 * For datagram protocols, leave room 731 * for protocol headers in first mbuf. 732 */ 733 if (atomic && top == 0 && len < mlen) 734 MH_ALIGN(m, len); 735 } 736 error = uiomove(mtod(m, caddr_t), (int)len, 737 uio); 738 have_data: 739 resid = uio->uio_resid; 740 m->m_len = len; 741 *mp = m; 742 top->m_pkthdr.len += len; 743 if (error) 744 goto release; 745 mp = &m->m_next; 746 if (resid <= 0) { 747 if (flags & MSG_EOR) 748 top->m_flags |= M_EOR; 749 break; 750 } 751 } while (space > 0 && atomic); 752 753 s = splsoftnet(); 754 755 if (so->so_state & SS_CANTSENDMORE) 756 snderr(EPIPE); 757 758 if (dontroute) 759 so->so_options |= SO_DONTROUTE; 760 if (resid > 0) 761 so->so_state |= SS_MORETOCOME; 762 error = (*so->so_proto->pr_usrreq)(so, 763 (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND, 764 top, addr, control, p); 765 if (dontroute) 766 so->so_options &= ~SO_DONTROUTE; 767 if (resid > 0) 768 so->so_state &= ~SS_MORETOCOME; 769 splx(s); 770 771 clen = 0; 772 control = 0; 773 top = 0; 774 mp = ⊤ 775 if (error) 776 goto release; 777 } while (resid && space > 0); 778 } while (resid); 779 780 release: 781 sbunlock(&so->so_snd); 782 out: 783 if (top) 784 m_freem(top); 785 if (control) 786 m_freem(control); 787 return (error); 788 } 789 790 /* 791 * Implement receive operations on a socket. 792 * We depend on the way that records are added to the sockbuf 793 * by sbappend*. In particular, each record (mbufs linked through m_next) 794 * must begin with an address if the protocol so specifies, 795 * followed by an optional mbuf or mbufs containing ancillary data, 796 * and then zero or more mbufs of data. 797 * In order to avoid blocking network interrupts for the entire time here, 798 * we splx() while doing the actual copy to user space. 799 * Although the sockbuf is locked, new data may still be appended, 800 * and thus we must maintain consistency of the sockbuf during that time. 801 * 802 * The caller may receive the data as a single mbuf chain by supplying 803 * an mbuf **mp0 for use in returning the chain. The uio is then used 804 * only for the count in uio_resid. 805 */ 806 int 807 soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, 808 struct mbuf **mp0, struct mbuf **controlp, int *flagsp) 809 { 810 struct mbuf *m, **mp; 811 int flags, len, error, s, offset, moff, type, orig_resid; 812 struct protosw *pr; 813 struct mbuf *nextrecord; 814 int mbuf_removed = 0; 815 816 pr = so->so_proto; 817 mp = mp0; 818 type = 0; 819 orig_resid = uio->uio_resid; 820 if (paddr) 821 *paddr = 0; 822 if (controlp) 823 *controlp = 0; 824 if (flagsp) 825 flags = *flagsp &~ MSG_EOR; 826 else 827 flags = 0; 828 829 if ((flags & MSG_DONTWAIT) == 0) 830 sodopendfree(so); 831 832 if (flags & MSG_OOB) { 833 m = m_get(M_WAIT, MT_DATA); 834 error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m, 835 (struct mbuf *)(long)(flags & MSG_PEEK), (struct mbuf *)0, 836 (struct proc *)0); 837 if (error) 838 goto bad; 839 do { 840 error = uiomove(mtod(m, caddr_t), 841 (int) min(uio->uio_resid, m->m_len), uio); 842 m = m_free(m); 843 } while (uio->uio_resid && error == 0 && m); 844 bad: 845 if (m) 846 m_freem(m); 847 return (error); 848 } 849 if (mp) 850 *mp = (struct mbuf *)0; 851 if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) 852 (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0, 853 (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0); 854 855 restart: 856 if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) 857 return (error); 858 s = splsoftnet(); 859 860 m = so->so_rcv.sb_mb; 861 /* 862 * If we have less data than requested, block awaiting more 863 * (subject to any timeout) if: 864 * 1. the current count is less than the low water mark, 865 * 2. MSG_WAITALL is set, and it is possible to do the entire 866 * receive operation at once if we block (resid <= hiwat), or 867 * 3. MSG_DONTWAIT is not set. 868 * If MSG_WAITALL is set but resid is larger than the receive buffer, 869 * we have to do the receive in sections, and thus risk returning 870 * a short count if a timeout or signal occurs after we start. 871 */ 872 if (m == 0 || (((flags & MSG_DONTWAIT) == 0 && 873 so->so_rcv.sb_cc < uio->uio_resid) && 874 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 875 ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) && 876 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) { 877 #ifdef DIAGNOSTIC 878 if (m == 0 && so->so_rcv.sb_cc) 879 panic("receive 1"); 880 #endif 881 if (so->so_error) { 882 if (m) 883 goto dontblock; 884 error = so->so_error; 885 if ((flags & MSG_PEEK) == 0) 886 so->so_error = 0; 887 goto release; 888 } 889 if (so->so_state & SS_CANTRCVMORE) { 890 if (m) 891 goto dontblock; 892 else 893 goto release; 894 } 895 for (; m; m = m->m_next) 896 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 897 m = so->so_rcv.sb_mb; 898 goto dontblock; 899 } 900 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 901 (so->so_proto->pr_flags & PR_CONNREQUIRED)) { 902 error = ENOTCONN; 903 goto release; 904 } 905 if (uio->uio_resid == 0) 906 goto release; 907 if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) { 908 error = EWOULDBLOCK; 909 goto release; 910 } 911 SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1"); 912 SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1"); 913 sbunlock(&so->so_rcv); 914 error = sbwait(&so->so_rcv); 915 splx(s); 916 if (error) 917 return (error); 918 goto restart; 919 } 920 dontblock: 921 /* 922 * On entry here, m points to the first record of the socket buffer. 923 * While we process the initial mbufs containing address and control 924 * info, we save a copy of m->m_nextpkt into nextrecord. 925 */ 926 #ifdef notyet /* XXXX */ 927 if (uio->uio_procp) 928 uio->uio_procp->p_stats->p_ru.ru_msgrcv++; 929 #endif 930 KASSERT(m == so->so_rcv.sb_mb); 931 SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); 932 SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); 933 nextrecord = m->m_nextpkt; 934 if (pr->pr_flags & PR_ADDR) { 935 #ifdef DIAGNOSTIC 936 if (m->m_type != MT_SONAME) 937 panic("receive 1a"); 938 #endif 939 orig_resid = 0; 940 if (flags & MSG_PEEK) { 941 if (paddr) 942 *paddr = m_copy(m, 0, m->m_len); 943 m = m->m_next; 944 } else { 945 sbfree(&so->so_rcv, m); 946 mbuf_removed = 1; 947 if (paddr) { 948 *paddr = m; 949 so->so_rcv.sb_mb = m->m_next; 950 m->m_next = 0; 951 m = so->so_rcv.sb_mb; 952 } else { 953 MFREE(m, so->so_rcv.sb_mb); 954 m = so->so_rcv.sb_mb; 955 } 956 } 957 } 958 while (m && m->m_type == MT_CONTROL && error == 0) { 959 if (flags & MSG_PEEK) { 960 if (controlp) 961 *controlp = m_copy(m, 0, m->m_len); 962 m = m->m_next; 963 } else { 964 sbfree(&so->so_rcv, m); 965 mbuf_removed = 1; 966 if (controlp) { 967 if (pr->pr_domain->dom_externalize && 968 mtod(m, struct cmsghdr *)->cmsg_type == 969 SCM_RIGHTS) 970 error = (*pr->pr_domain->dom_externalize)(m); 971 *controlp = m; 972 so->so_rcv.sb_mb = m->m_next; 973 m->m_next = 0; 974 m = so->so_rcv.sb_mb; 975 } else { 976 MFREE(m, so->so_rcv.sb_mb); 977 m = so->so_rcv.sb_mb; 978 } 979 } 980 if (controlp) { 981 orig_resid = 0; 982 controlp = &(*controlp)->m_next; 983 } 984 } 985 986 /* 987 * If m is non-NULL, we have some data to read. From now on, 988 * make sure to keep sb_lastrecord consistent when working on 989 * the last packet on the chain (nextrecord == NULL) and we 990 * change m->m_nextpkt. 991 */ 992 if (m) { 993 if ((flags & MSG_PEEK) == 0) { 994 m->m_nextpkt = nextrecord; 995 /* 996 * If nextrecord == NULL (this is a single chain), 997 * then sb_lastrecord may not be valid here if m 998 * was changed earlier. 999 */ 1000 if (nextrecord == NULL) { 1001 KASSERT(so->so_rcv.sb_mb == m); 1002 so->so_rcv.sb_lastrecord = m; 1003 } 1004 } 1005 type = m->m_type; 1006 if (type == MT_OOBDATA) 1007 flags |= MSG_OOB; 1008 } else { 1009 if ((flags & MSG_PEEK) == 0) { 1010 KASSERT(so->so_rcv.sb_mb == m); 1011 so->so_rcv.sb_mb = nextrecord; 1012 SB_EMPTY_FIXUP(&so->so_rcv); 1013 } 1014 } 1015 SBLASTRECORDCHK(&so->so_rcv, "soreceive 2"); 1016 SBLASTMBUFCHK(&so->so_rcv, "soreceive 2"); 1017 1018 moff = 0; 1019 offset = 0; 1020 while (m && uio->uio_resid > 0 && error == 0) { 1021 if (m->m_type == MT_OOBDATA) { 1022 if (type != MT_OOBDATA) 1023 break; 1024 } else if (type == MT_OOBDATA) 1025 break; 1026 #ifdef DIAGNOSTIC 1027 else if (m->m_type != MT_DATA && m->m_type != MT_HEADER) 1028 panic("receive 3"); 1029 #endif 1030 so->so_state &= ~SS_RCVATMARK; 1031 len = uio->uio_resid; 1032 if (so->so_oobmark && len > so->so_oobmark - offset) 1033 len = so->so_oobmark - offset; 1034 if (len > m->m_len - moff) 1035 len = m->m_len - moff; 1036 /* 1037 * If mp is set, just pass back the mbufs. 1038 * Otherwise copy them out via the uio, then free. 1039 * Sockbuf must be consistent here (points to current mbuf, 1040 * it points to next record) when we drop priority; 1041 * we must note any additions to the sockbuf when we 1042 * block interrupts again. 1043 */ 1044 if (mp == 0) { 1045 SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); 1046 SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); 1047 splx(s); 1048 error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio); 1049 s = splsoftnet(); 1050 if (error) { 1051 /* 1052 * If any part of the record has been removed 1053 * (such as the MT_SONAME mbuf, which will 1054 * happen when PR_ADDR, and thus also 1055 * PR_ATOMIC, is set), then drop the entire 1056 * record to maintain the atomicity of the 1057 * receive operation. 1058 * 1059 * This avoids a later panic("receive 1a") 1060 * when compiled with DIAGNOSTIC. 1061 */ 1062 if (m && mbuf_removed 1063 && (pr->pr_flags & PR_ATOMIC)) 1064 (void) sbdroprecord(&so->so_rcv); 1065 1066 goto release; 1067 } 1068 } else 1069 uio->uio_resid -= len; 1070 if (len == m->m_len - moff) { 1071 if (m->m_flags & M_EOR) 1072 flags |= MSG_EOR; 1073 if (flags & MSG_PEEK) { 1074 m = m->m_next; 1075 moff = 0; 1076 } else { 1077 nextrecord = m->m_nextpkt; 1078 sbfree(&so->so_rcv, m); 1079 if (mp) { 1080 *mp = m; 1081 mp = &m->m_next; 1082 so->so_rcv.sb_mb = m = m->m_next; 1083 *mp = (struct mbuf *)0; 1084 } else { 1085 MFREE(m, so->so_rcv.sb_mb); 1086 m = so->so_rcv.sb_mb; 1087 } 1088 /* 1089 * If m != NULL, we also know that 1090 * so->so_rcv.sb_mb != NULL. 1091 */ 1092 KASSERT(so->so_rcv.sb_mb == m); 1093 if (m) { 1094 m->m_nextpkt = nextrecord; 1095 if (nextrecord == NULL) 1096 so->so_rcv.sb_lastrecord = m; 1097 } else { 1098 so->so_rcv.sb_mb = nextrecord; 1099 SB_EMPTY_FIXUP(&so->so_rcv); 1100 } 1101 SBLASTRECORDCHK(&so->so_rcv, "soreceive 3"); 1102 SBLASTMBUFCHK(&so->so_rcv, "soreceive 3"); 1103 } 1104 } else { 1105 if (flags & MSG_PEEK) 1106 moff += len; 1107 else { 1108 if (mp) 1109 *mp = m_copym(m, 0, len, M_WAIT); 1110 m->m_data += len; 1111 m->m_len -= len; 1112 so->so_rcv.sb_cc -= len; 1113 } 1114 } 1115 if (so->so_oobmark) { 1116 if ((flags & MSG_PEEK) == 0) { 1117 so->so_oobmark -= len; 1118 if (so->so_oobmark == 0) { 1119 so->so_state |= SS_RCVATMARK; 1120 break; 1121 } 1122 } else { 1123 offset += len; 1124 if (offset == so->so_oobmark) 1125 break; 1126 } 1127 } 1128 if (flags & MSG_EOR) 1129 break; 1130 /* 1131 * If the MSG_WAITALL flag is set (for non-atomic socket), 1132 * we must not quit until "uio->uio_resid == 0" or an error 1133 * termination. If a signal/timeout occurs, return 1134 * with a short count but without error. 1135 * Keep sockbuf locked against other readers. 1136 */ 1137 while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 && 1138 !sosendallatonce(so) && !nextrecord) { 1139 if (so->so_error || so->so_state & SS_CANTRCVMORE) 1140 break; 1141 /* 1142 * If we are peeking and the socket receive buffer is 1143 * full, stop since we can't get more data to peek at. 1144 */ 1145 if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0) 1146 break; 1147 /* 1148 * If we've drained the socket buffer, tell the 1149 * protocol in case it needs to do something to 1150 * get it filled again. 1151 */ 1152 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) 1153 (*pr->pr_usrreq)(so, PRU_RCVD, 1154 (struct mbuf *)0, 1155 (struct mbuf *)(long)flags, 1156 (struct mbuf *)0, 1157 (struct proc *)0); 1158 SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2"); 1159 SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2"); 1160 error = sbwait(&so->so_rcv); 1161 if (error) { 1162 sbunlock(&so->so_rcv); 1163 splx(s); 1164 return (0); 1165 } 1166 if ((m = so->so_rcv.sb_mb) != NULL) 1167 nextrecord = m->m_nextpkt; 1168 } 1169 } 1170 1171 if (m && pr->pr_flags & PR_ATOMIC) { 1172 flags |= MSG_TRUNC; 1173 if ((flags & MSG_PEEK) == 0) 1174 (void) sbdroprecord(&so->so_rcv); 1175 } 1176 if ((flags & MSG_PEEK) == 0) { 1177 if (m == 0) { 1178 /* 1179 * First part is an inline SB_EMPTY_FIXUP(). Second 1180 * part makes sure sb_lastrecord is up-to-date if 1181 * there is still data in the socket buffer. 1182 */ 1183 so->so_rcv.sb_mb = nextrecord; 1184 if (so->so_rcv.sb_mb == NULL) { 1185 so->so_rcv.sb_mbtail = NULL; 1186 so->so_rcv.sb_lastrecord = NULL; 1187 } else if (nextrecord->m_nextpkt == NULL) 1188 so->so_rcv.sb_lastrecord = nextrecord; 1189 } 1190 SBLASTRECORDCHK(&so->so_rcv, "soreceive 4"); 1191 SBLASTMBUFCHK(&so->so_rcv, "soreceive 4"); 1192 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1193 (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0, 1194 (struct mbuf *)(long)flags, (struct mbuf *)0, 1195 (struct proc *)0); 1196 } 1197 if (orig_resid == uio->uio_resid && orig_resid && 1198 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 1199 sbunlock(&so->so_rcv); 1200 splx(s); 1201 goto restart; 1202 } 1203 1204 if (flagsp) 1205 *flagsp |= flags; 1206 release: 1207 sbunlock(&so->so_rcv); 1208 splx(s); 1209 return (error); 1210 } 1211 1212 int 1213 soshutdown(struct socket *so, int how) 1214 { 1215 struct protosw *pr; 1216 1217 pr = so->so_proto; 1218 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 1219 return (EINVAL); 1220 1221 if (how == SHUT_RD || how == SHUT_RDWR) 1222 sorflush(so); 1223 if (how == SHUT_WR || how == SHUT_RDWR) 1224 return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0, 1225 (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0); 1226 return (0); 1227 } 1228 1229 void 1230 sorflush(struct socket *so) 1231 { 1232 struct sockbuf *sb, asb; 1233 struct protosw *pr; 1234 int s; 1235 1236 sb = &so->so_rcv; 1237 pr = so->so_proto; 1238 sb->sb_flags |= SB_NOINTR; 1239 (void) sblock(sb, M_WAITOK); 1240 s = splnet(); 1241 socantrcvmore(so); 1242 sbunlock(sb); 1243 asb = *sb; 1244 memset((caddr_t)sb, 0, sizeof(*sb)); 1245 splx(s); 1246 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) 1247 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 1248 sbrelease(&asb); 1249 } 1250 1251 int 1252 sosetopt(struct socket *so, int level, int optname, struct mbuf *m0) 1253 { 1254 int error; 1255 struct mbuf *m; 1256 1257 error = 0; 1258 m = m0; 1259 if (level != SOL_SOCKET) { 1260 if (so->so_proto && so->so_proto->pr_ctloutput) 1261 return ((*so->so_proto->pr_ctloutput) 1262 (PRCO_SETOPT, so, level, optname, &m0)); 1263 error = ENOPROTOOPT; 1264 } else { 1265 switch (optname) { 1266 1267 case SO_LINGER: 1268 if (m == NULL || m->m_len != sizeof(struct linger)) { 1269 error = EINVAL; 1270 goto bad; 1271 } 1272 so->so_linger = mtod(m, struct linger *)->l_linger; 1273 /* fall thru... */ 1274 1275 case SO_DEBUG: 1276 case SO_KEEPALIVE: 1277 case SO_DONTROUTE: 1278 case SO_USELOOPBACK: 1279 case SO_BROADCAST: 1280 case SO_REUSEADDR: 1281 case SO_REUSEPORT: 1282 case SO_OOBINLINE: 1283 case SO_TIMESTAMP: 1284 if (m == NULL || m->m_len < sizeof(int)) { 1285 error = EINVAL; 1286 goto bad; 1287 } 1288 if (*mtod(m, int *)) 1289 so->so_options |= optname; 1290 else 1291 so->so_options &= ~optname; 1292 break; 1293 1294 case SO_SNDBUF: 1295 case SO_RCVBUF: 1296 case SO_SNDLOWAT: 1297 case SO_RCVLOWAT: 1298 { 1299 int optval; 1300 1301 if (m == NULL || m->m_len < sizeof(int)) { 1302 error = EINVAL; 1303 goto bad; 1304 } 1305 1306 /* 1307 * Values < 1 make no sense for any of these 1308 * options, so disallow them. 1309 */ 1310 optval = *mtod(m, int *); 1311 if (optval < 1) { 1312 error = EINVAL; 1313 goto bad; 1314 } 1315 1316 switch (optname) { 1317 1318 case SO_SNDBUF: 1319 case SO_RCVBUF: 1320 if (sbreserve(optname == SO_SNDBUF ? 1321 &so->so_snd : &so->so_rcv, 1322 (u_long) optval) == 0) { 1323 error = ENOBUFS; 1324 goto bad; 1325 } 1326 break; 1327 1328 /* 1329 * Make sure the low-water is never greater than 1330 * the high-water. 1331 */ 1332 case SO_SNDLOWAT: 1333 so->so_snd.sb_lowat = 1334 (optval > so->so_snd.sb_hiwat) ? 1335 so->so_snd.sb_hiwat : optval; 1336 break; 1337 case SO_RCVLOWAT: 1338 so->so_rcv.sb_lowat = 1339 (optval > so->so_rcv.sb_hiwat) ? 1340 so->so_rcv.sb_hiwat : optval; 1341 break; 1342 } 1343 break; 1344 } 1345 1346 case SO_SNDTIMEO: 1347 case SO_RCVTIMEO: 1348 { 1349 struct timeval *tv; 1350 short val; 1351 1352 if (m == NULL || m->m_len < sizeof(*tv)) { 1353 error = EINVAL; 1354 goto bad; 1355 } 1356 tv = mtod(m, struct timeval *); 1357 if (tv->tv_sec * hz + tv->tv_usec / tick > SHRT_MAX) { 1358 error = EDOM; 1359 goto bad; 1360 } 1361 val = tv->tv_sec * hz + tv->tv_usec / tick; 1362 1363 switch (optname) { 1364 1365 case SO_SNDTIMEO: 1366 so->so_snd.sb_timeo = val; 1367 break; 1368 case SO_RCVTIMEO: 1369 so->so_rcv.sb_timeo = val; 1370 break; 1371 } 1372 break; 1373 } 1374 1375 default: 1376 error = ENOPROTOOPT; 1377 break; 1378 } 1379 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) { 1380 (void) ((*so->so_proto->pr_ctloutput) 1381 (PRCO_SETOPT, so, level, optname, &m0)); 1382 m = NULL; /* freed by protocol */ 1383 } 1384 } 1385 bad: 1386 if (m) 1387 (void) m_free(m); 1388 return (error); 1389 } 1390 1391 int 1392 sogetopt(struct socket *so, int level, int optname, struct mbuf **mp) 1393 { 1394 struct mbuf *m; 1395 1396 if (level != SOL_SOCKET) { 1397 if (so->so_proto && so->so_proto->pr_ctloutput) { 1398 return ((*so->so_proto->pr_ctloutput) 1399 (PRCO_GETOPT, so, level, optname, mp)); 1400 } else 1401 return (ENOPROTOOPT); 1402 } else { 1403 m = m_get(M_WAIT, MT_SOOPTS); 1404 m->m_len = sizeof(int); 1405 1406 switch (optname) { 1407 1408 case SO_LINGER: 1409 m->m_len = sizeof(struct linger); 1410 mtod(m, struct linger *)->l_onoff = 1411 so->so_options & SO_LINGER; 1412 mtod(m, struct linger *)->l_linger = so->so_linger; 1413 break; 1414 1415 case SO_USELOOPBACK: 1416 case SO_DONTROUTE: 1417 case SO_DEBUG: 1418 case SO_KEEPALIVE: 1419 case SO_REUSEADDR: 1420 case SO_REUSEPORT: 1421 case SO_BROADCAST: 1422 case SO_OOBINLINE: 1423 case SO_TIMESTAMP: 1424 *mtod(m, int *) = so->so_options & optname; 1425 break; 1426 1427 case SO_TYPE: 1428 *mtod(m, int *) = so->so_type; 1429 break; 1430 1431 case SO_ERROR: 1432 *mtod(m, int *) = so->so_error; 1433 so->so_error = 0; 1434 break; 1435 1436 case SO_SNDBUF: 1437 *mtod(m, int *) = so->so_snd.sb_hiwat; 1438 break; 1439 1440 case SO_RCVBUF: 1441 *mtod(m, int *) = so->so_rcv.sb_hiwat; 1442 break; 1443 1444 case SO_SNDLOWAT: 1445 *mtod(m, int *) = so->so_snd.sb_lowat; 1446 break; 1447 1448 case SO_RCVLOWAT: 1449 *mtod(m, int *) = so->so_rcv.sb_lowat; 1450 break; 1451 1452 case SO_SNDTIMEO: 1453 case SO_RCVTIMEO: 1454 { 1455 int val = (optname == SO_SNDTIMEO ? 1456 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 1457 1458 m->m_len = sizeof(struct timeval); 1459 mtod(m, struct timeval *)->tv_sec = val / hz; 1460 mtod(m, struct timeval *)->tv_usec = 1461 (val % hz) * tick; 1462 break; 1463 } 1464 1465 default: 1466 (void)m_free(m); 1467 return (ENOPROTOOPT); 1468 } 1469 *mp = m; 1470 return (0); 1471 } 1472 } 1473 1474 void 1475 sohasoutofband(struct socket *so) 1476 { 1477 struct proc *p; 1478 1479 if (so->so_pgid < 0) 1480 gsignal(-so->so_pgid, SIGURG); 1481 else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0) 1482 psignal(p, SIGURG); 1483 selwakeup(&so->so_rcv.sb_sel); 1484 } 1485 1486 1487 int 1488 soo_kqfilter(struct file *fp, struct knote *kn) 1489 { 1490 struct socket *so; 1491 struct sockbuf *sb; 1492 1493 so = (struct socket *)kn->kn_fp->f_data; 1494 switch (kn->kn_filter) { 1495 case EVFILT_READ: 1496 if (so->so_options & SO_ACCEPTCONN) 1497 kn->kn_fop = &solisten_filtops; 1498 else 1499 kn->kn_fop = &soread_filtops; 1500 sb = &so->so_rcv; 1501 break; 1502 case EVFILT_WRITE: 1503 kn->kn_fop = &sowrite_filtops; 1504 sb = &so->so_snd; 1505 break; 1506 default: 1507 return (1); 1508 } 1509 SLIST_INSERT_HEAD(&sb->sb_sel.si_klist, kn, kn_selnext); 1510 sb->sb_flags |= SB_KNOTE; 1511 return (0); 1512 } 1513 1514 static void 1515 filt_sordetach(struct knote *kn) 1516 { 1517 struct socket *so; 1518 1519 so = (struct socket *)kn->kn_fp->f_data; 1520 SLIST_REMOVE(&so->so_rcv.sb_sel.si_klist, kn, knote, kn_selnext); 1521 if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_klist)) 1522 so->so_rcv.sb_flags &= ~SB_KNOTE; 1523 } 1524 1525 /*ARGSUSED*/ 1526 static int 1527 filt_soread(struct knote *kn, long hint) 1528 { 1529 struct socket *so; 1530 1531 so = (struct socket *)kn->kn_fp->f_data; 1532 kn->kn_data = so->so_rcv.sb_cc; 1533 if (so->so_state & SS_CANTRCVMORE) { 1534 kn->kn_flags |= EV_EOF; 1535 kn->kn_fflags = so->so_error; 1536 return (1); 1537 } 1538 if (so->so_error) /* temporary udp error */ 1539 return (1); 1540 if (kn->kn_sfflags & NOTE_LOWAT) 1541 return (kn->kn_data >= kn->kn_sdata); 1542 return (kn->kn_data >= so->so_rcv.sb_lowat); 1543 } 1544 1545 static void 1546 filt_sowdetach(struct knote *kn) 1547 { 1548 struct socket *so; 1549 1550 so = (struct socket *)kn->kn_fp->f_data; 1551 SLIST_REMOVE(&so->so_snd.sb_sel.si_klist, kn, knote, kn_selnext); 1552 if (SLIST_EMPTY(&so->so_snd.sb_sel.si_klist)) 1553 so->so_snd.sb_flags &= ~SB_KNOTE; 1554 } 1555 1556 /*ARGSUSED*/ 1557 static int 1558 filt_sowrite(struct knote *kn, long hint) 1559 { 1560 struct socket *so; 1561 1562 so = (struct socket *)kn->kn_fp->f_data; 1563 kn->kn_data = sbspace(&so->so_snd); 1564 if (so->so_state & SS_CANTSENDMORE) { 1565 kn->kn_flags |= EV_EOF; 1566 kn->kn_fflags = so->so_error; 1567 return (1); 1568 } 1569 if (so->so_error) /* temporary udp error */ 1570 return (1); 1571 if (((so->so_state & SS_ISCONNECTED) == 0) && 1572 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 1573 return (0); 1574 if (kn->kn_sfflags & NOTE_LOWAT) 1575 return (kn->kn_data >= kn->kn_sdata); 1576 return (kn->kn_data >= so->so_snd.sb_lowat); 1577 } 1578 1579 /*ARGSUSED*/ 1580 static int 1581 filt_solisten(struct knote *kn, long hint) 1582 { 1583 struct socket *so; 1584 1585 so = (struct socket *)kn->kn_fp->f_data; 1586 1587 /* 1588 * Set kn_data to number of incoming connections, not 1589 * counting partial (incomplete) connections. 1590 */ 1591 kn->kn_data = so->so_qlen; 1592 return (kn->kn_data > 0); 1593 } 1594
Indexes created Sat Sep 20 22:09:52 GMT 2025