uipc_socket.c revision 1.120
1 /* $NetBSD: uipc_socket.c,v 1.120 2006/06/13 21:19:56 ginsbach 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. Neither the name of the University nor the names of its contributors 52 * may be used to endorse or promote products derived from this software 53 * without specific prior written permission. 54 * 55 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 56 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 57 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 58 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 59 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 60 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 61 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 62 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 63 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 64 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 65 * SUCH DAMAGE. 66 * 67 * @(#)uipc_socket.c 8.6 (Berkeley) 5/2/95 68 */ 69 70 #include <sys/cdefs.h> 71 __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.120 2006/06/13 21:19:56 ginsbach Exp $"); 72 73 #include "opt_sock_counters.h" 74 #include "opt_sosend_loan.h" 75 #include "opt_mbuftrace.h" 76 #include "opt_somaxkva.h" 77 78 #include <sys/param.h> 79 #include <sys/systm.h> 80 #include <sys/proc.h> 81 #include <sys/file.h> 82 #include <sys/malloc.h> 83 #include <sys/mbuf.h> 84 #include <sys/domain.h> 85 #include <sys/kernel.h> 86 #include <sys/protosw.h> 87 #include <sys/socket.h> 88 #include <sys/socketvar.h> 89 #include <sys/signalvar.h> 90 #include <sys/resourcevar.h> 91 #include <sys/pool.h> 92 #include <sys/event.h> 93 #include <sys/poll.h> 94 #include <sys/kauth.h> 95 96 #include <uvm/uvm.h> 97 98 POOL_INIT(socket_pool, sizeof(struct socket), 0, 0, 0, "sockpl", NULL); 99 100 MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options"); 101 MALLOC_DEFINE(M_SONAME, "soname", "socket name"); 102 103 extern int somaxconn; /* patchable (XXX sysctl) */ 104 int somaxconn = SOMAXCONN; 105 106 #ifdef SOSEND_COUNTERS 107 #include <sys/device.h> 108 109 static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 110 NULL, "sosend", "loan big"); 111 static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 112 NULL, "sosend", "copy big"); 113 static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 114 NULL, "sosend", "copy small"); 115 static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 116 NULL, "sosend", "kva limit"); 117 118 #define SOSEND_COUNTER_INCR(ev) (ev)->ev_count++ 119 120 EVCNT_ATTACH_STATIC(sosend_loan_big); 121 EVCNT_ATTACH_STATIC(sosend_copy_big); 122 EVCNT_ATTACH_STATIC(sosend_copy_small); 123 EVCNT_ATTACH_STATIC(sosend_kvalimit); 124 #else 125 126 #define SOSEND_COUNTER_INCR(ev) /* nothing */ 127 128 #endif /* SOSEND_COUNTERS */ 129 130 static struct callback_entry sokva_reclaimerentry; 131 132 #ifdef SOSEND_NO_LOAN 133 int use_sosend_loan = 0; 134 #else 135 int use_sosend_loan = 1; 136 #endif 137 138 static struct simplelock so_pendfree_slock = SIMPLELOCK_INITIALIZER; 139 static struct mbuf *so_pendfree; 140 141 #ifndef SOMAXKVA 142 #define SOMAXKVA (16 * 1024 * 1024) 143 #endif 144 int somaxkva = SOMAXKVA; 145 static int socurkva; 146 static int sokvawaiters; 147 148 #define SOCK_LOAN_THRESH 4096 149 #define SOCK_LOAN_CHUNK 65536 150 151 static size_t sodopendfree(void); 152 static size_t sodopendfreel(void); 153 154 static vsize_t 155 sokvareserve(struct socket *so, vsize_t len) 156 { 157 int s; 158 int error; 159 160 s = splvm(); 161 simple_lock(&so_pendfree_slock); 162 while (socurkva + len > somaxkva) { 163 size_t freed; 164 165 /* 166 * try to do pendfree. 167 */ 168 169 freed = sodopendfreel(); 170 171 /* 172 * if some kva was freed, try again. 173 */ 174 175 if (freed) 176 continue; 177 178 SOSEND_COUNTER_INCR(&sosend_kvalimit); 179 sokvawaiters++; 180 error = ltsleep(&socurkva, PVM | PCATCH, "sokva", 0, 181 &so_pendfree_slock); 182 sokvawaiters--; 183 if (error) { 184 len = 0; 185 break; 186 } 187 } 188 socurkva += len; 189 simple_unlock(&so_pendfree_slock); 190 splx(s); 191 return len; 192 } 193 194 static void 195 sokvaunreserve(vsize_t len) 196 { 197 int s; 198 199 s = splvm(); 200 simple_lock(&so_pendfree_slock); 201 socurkva -= len; 202 if (sokvawaiters) 203 wakeup(&socurkva); 204 simple_unlock(&so_pendfree_slock); 205 splx(s); 206 } 207 208 /* 209 * sokvaalloc: allocate kva for loan. 210 */ 211 212 vaddr_t 213 sokvaalloc(vsize_t len, struct socket *so) 214 { 215 vaddr_t lva; 216 217 /* 218 * reserve kva. 219 */ 220 221 if (sokvareserve(so, len) == 0) 222 return 0; 223 224 /* 225 * allocate kva. 226 */ 227 228 lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA); 229 if (lva == 0) { 230 sokvaunreserve(len); 231 return (0); 232 } 233 234 return lva; 235 } 236 237 /* 238 * sokvafree: free kva for loan. 239 */ 240 241 void 242 sokvafree(vaddr_t sva, vsize_t len) 243 { 244 245 /* 246 * free kva. 247 */ 248 249 uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY); 250 251 /* 252 * unreserve kva. 253 */ 254 255 sokvaunreserve(len); 256 } 257 258 static void 259 sodoloanfree(struct vm_page **pgs, caddr_t buf, size_t size) 260 { 261 vaddr_t va, sva, eva; 262 vsize_t len; 263 paddr_t pa; 264 int i, npgs; 265 266 eva = round_page((vaddr_t) buf + size); 267 sva = trunc_page((vaddr_t) buf); 268 len = eva - sva; 269 npgs = len >> PAGE_SHIFT; 270 271 if (__predict_false(pgs == NULL)) { 272 pgs = alloca(npgs * sizeof(*pgs)); 273 274 for (i = 0, va = sva; va < eva; i++, va += PAGE_SIZE) { 275 if (pmap_extract(pmap_kernel(), va, &pa) == FALSE) 276 panic("sodoloanfree: va 0x%lx not mapped", va); 277 pgs[i] = PHYS_TO_VM_PAGE(pa); 278 } 279 } 280 281 pmap_kremove(sva, len); 282 pmap_update(pmap_kernel()); 283 uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE); 284 sokvafree(sva, len); 285 } 286 287 static size_t 288 sodopendfree() 289 { 290 int s; 291 size_t rv; 292 293 s = splvm(); 294 simple_lock(&so_pendfree_slock); 295 rv = sodopendfreel(); 296 simple_unlock(&so_pendfree_slock); 297 splx(s); 298 299 return rv; 300 } 301 302 /* 303 * sodopendfreel: free mbufs on "pendfree" list. 304 * unlock and relock so_pendfree_slock when freeing mbufs. 305 * 306 * => called with so_pendfree_slock held. 307 * => called at splvm. 308 */ 309 310 static size_t 311 sodopendfreel() 312 { 313 size_t rv = 0; 314 315 LOCK_ASSERT(simple_lock_held(&so_pendfree_slock)); 316 317 for (;;) { 318 struct mbuf *m; 319 struct mbuf *next; 320 321 m = so_pendfree; 322 if (m == NULL) 323 break; 324 so_pendfree = NULL; 325 simple_unlock(&so_pendfree_slock); 326 /* XXX splx */ 327 328 for (; m != NULL; m = next) { 329 next = m->m_next; 330 331 rv += m->m_ext.ext_size; 332 sodoloanfree((m->m_flags & M_EXT_PAGES) ? 333 m->m_ext.ext_pgs : NULL, m->m_ext.ext_buf, 334 m->m_ext.ext_size); 335 pool_cache_put(&mbpool_cache, m); 336 } 337 338 /* XXX splvm */ 339 simple_lock(&so_pendfree_slock); 340 } 341 342 return (rv); 343 } 344 345 void 346 soloanfree(struct mbuf *m, caddr_t buf, size_t size, void *arg) 347 { 348 int s; 349 350 if (m == NULL) { 351 352 /* 353 * called from MEXTREMOVE. 354 */ 355 356 sodoloanfree(NULL, buf, size); 357 return; 358 } 359 360 /* 361 * postpone freeing mbuf. 362 * 363 * we can't do it in interrupt context 364 * because we need to put kva back to kernel_map. 365 */ 366 367 s = splvm(); 368 simple_lock(&so_pendfree_slock); 369 m->m_next = so_pendfree; 370 so_pendfree = m; 371 if (sokvawaiters) 372 wakeup(&socurkva); 373 simple_unlock(&so_pendfree_slock); 374 splx(s); 375 } 376 377 static long 378 sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space) 379 { 380 struct iovec *iov = uio->uio_iov; 381 vaddr_t sva, eva; 382 vsize_t len; 383 vaddr_t lva, va; 384 int npgs, i, error; 385 386 if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace)) 387 return (0); 388 389 if (iov->iov_len < (size_t) space) 390 space = iov->iov_len; 391 if (space > SOCK_LOAN_CHUNK) 392 space = SOCK_LOAN_CHUNK; 393 394 eva = round_page((vaddr_t) iov->iov_base + space); 395 sva = trunc_page((vaddr_t) iov->iov_base); 396 len = eva - sva; 397 npgs = len >> PAGE_SHIFT; 398 399 /* XXX KDASSERT */ 400 KASSERT(npgs <= M_EXT_MAXPAGES); 401 402 lva = sokvaalloc(len, so); 403 if (lva == 0) 404 return 0; 405 406 error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len, 407 m->m_ext.ext_pgs, UVM_LOAN_TOPAGE); 408 if (error) { 409 sokvafree(lva, len); 410 return (0); 411 } 412 413 for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE) 414 pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]), 415 VM_PROT_READ); 416 pmap_update(pmap_kernel()); 417 418 lva += (vaddr_t) iov->iov_base & PAGE_MASK; 419 420 MEXTADD(m, (caddr_t) lva, space, M_MBUF, soloanfree, so); 421 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 422 423 uio->uio_resid -= space; 424 /* uio_offset not updated, not set/used for write(2) */ 425 uio->uio_iov->iov_base = (caddr_t) uio->uio_iov->iov_base + space; 426 uio->uio_iov->iov_len -= space; 427 if (uio->uio_iov->iov_len == 0) { 428 uio->uio_iov++; 429 uio->uio_iovcnt--; 430 } 431 432 return (space); 433 } 434 435 static int 436 sokva_reclaim_callback(struct callback_entry *ce, void *obj, void *arg) 437 { 438 439 KASSERT(ce == &sokva_reclaimerentry); 440 KASSERT(obj == NULL); 441 442 sodopendfree(); 443 if (!vm_map_starved_p(kernel_map)) { 444 return CALLBACK_CHAIN_ABORT; 445 } 446 return CALLBACK_CHAIN_CONTINUE; 447 } 448 449 void 450 soinit(void) 451 { 452 453 /* Set the initial adjusted socket buffer size. */ 454 if (sb_max_set(sb_max)) 455 panic("bad initial sb_max value: %lu", sb_max); 456 457 callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback, 458 &sokva_reclaimerentry, NULL, sokva_reclaim_callback); 459 } 460 461 /* 462 * Socket operation routines. 463 * These routines are called by the routines in 464 * sys_socket.c or from a system process, and 465 * implement the semantics of socket operations by 466 * switching out to the protocol specific routines. 467 */ 468 /*ARGSUSED*/ 469 int 470 socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l) 471 { 472 const struct protosw *prp; 473 struct socket *so; 474 uid_t uid; 475 int error, s; 476 477 if (proto) 478 prp = pffindproto(dom, proto, type); 479 else 480 prp = pffindtype(dom, type); 481 if (prp == 0) { 482 /* no support for domain */ 483 if (pffinddomain(dom) == 0) 484 return (EAFNOSUPPORT); 485 /* no support for socket type */ 486 if (proto == 0 && type != 0) 487 return (EPROTOTYPE); 488 return (EPROTONOSUPPORT); 489 } 490 if (prp->pr_usrreq == 0) 491 return (EPROTONOSUPPORT); 492 if (prp->pr_type != type) 493 return (EPROTOTYPE); 494 s = splsoftnet(); 495 so = pool_get(&socket_pool, PR_WAITOK); 496 memset((caddr_t)so, 0, sizeof(*so)); 497 TAILQ_INIT(&so->so_q0); 498 TAILQ_INIT(&so->so_q); 499 so->so_type = type; 500 so->so_proto = prp; 501 so->so_send = sosend; 502 so->so_receive = soreceive; 503 #ifdef MBUFTRACE 504 so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner; 505 so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner; 506 so->so_mowner = &prp->pr_domain->dom_mowner; 507 #endif 508 if (l != NULL) { 509 uid = kauth_cred_geteuid(l->l_proc->p_cred); 510 } else { 511 uid = 0; 512 } 513 so->so_uidinfo = uid_find(uid); 514 error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0, 515 (struct mbuf *)(long)proto, (struct mbuf *)0, l); 516 if (error) { 517 so->so_state |= SS_NOFDREF; 518 sofree(so); 519 splx(s); 520 return (error); 521 } 522 splx(s); 523 *aso = so; 524 return (0); 525 } 526 527 int 528 sobind(struct socket *so, struct mbuf *nam, struct lwp *l) 529 { 530 int s, error; 531 532 s = splsoftnet(); 533 error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0, 534 nam, (struct mbuf *)0, l); 535 splx(s); 536 return (error); 537 } 538 539 int 540 solisten(struct socket *so, int backlog) 541 { 542 int s, error; 543 544 s = splsoftnet(); 545 error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0, 546 (struct mbuf *)0, (struct mbuf *)0, (struct lwp *)0); 547 if (error) { 548 splx(s); 549 return (error); 550 } 551 if (TAILQ_EMPTY(&so->so_q)) 552 so->so_options |= SO_ACCEPTCONN; 553 if (backlog < 0) 554 backlog = 0; 555 so->so_qlimit = min(backlog, somaxconn); 556 splx(s); 557 return (0); 558 } 559 560 void 561 sofree(struct socket *so) 562 { 563 564 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) 565 return; 566 if (so->so_head) { 567 /* 568 * We must not decommission a socket that's on the accept(2) 569 * queue. If we do, then accept(2) may hang after select(2) 570 * indicated that the listening socket was ready. 571 */ 572 if (!soqremque(so, 0)) 573 return; 574 } 575 if (so->so_rcv.sb_hiwat) 576 (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, 577 RLIM_INFINITY); 578 if (so->so_snd.sb_hiwat) 579 (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, 580 RLIM_INFINITY); 581 sbrelease(&so->so_snd, so); 582 sorflush(so); 583 pool_put(&socket_pool, so); 584 } 585 586 /* 587 * Close a socket on last file table reference removal. 588 * Initiate disconnect if connected. 589 * Free socket when disconnect complete. 590 */ 591 int 592 soclose(struct socket *so) 593 { 594 struct socket *so2; 595 int s, error; 596 597 error = 0; 598 s = splsoftnet(); /* conservative */ 599 if (so->so_options & SO_ACCEPTCONN) { 600 while ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) { 601 (void) soqremque(so2, 0); 602 (void) soabort(so2); 603 } 604 while ((so2 = TAILQ_FIRST(&so->so_q)) != 0) { 605 (void) soqremque(so2, 1); 606 (void) soabort(so2); 607 } 608 } 609 if (so->so_pcb == 0) 610 goto discard; 611 if (so->so_state & SS_ISCONNECTED) { 612 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 613 error = sodisconnect(so); 614 if (error) 615 goto drop; 616 } 617 if (so->so_options & SO_LINGER) { 618 if ((so->so_state & SS_ISDISCONNECTING) && 619 (so->so_state & SS_NBIO)) 620 goto drop; 621 while (so->so_state & SS_ISCONNECTED) { 622 error = tsleep((caddr_t)&so->so_timeo, 623 PSOCK | PCATCH, netcls, 624 so->so_linger * hz); 625 if (error) 626 break; 627 } 628 } 629 } 630 drop: 631 if (so->so_pcb) { 632 int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH, 633 (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0, 634 (struct lwp *)0); 635 if (error == 0) 636 error = error2; 637 } 638 discard: 639 if (so->so_state & SS_NOFDREF) 640 panic("soclose: NOFDREF"); 641 so->so_state |= SS_NOFDREF; 642 sofree(so); 643 splx(s); 644 return (error); 645 } 646 647 /* 648 * Must be called at splsoftnet... 649 */ 650 int 651 soabort(struct socket *so) 652 { 653 654 return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0, 655 (struct mbuf *)0, (struct mbuf *)0, (struct lwp *)0); 656 } 657 658 int 659 soaccept(struct socket *so, struct mbuf *nam) 660 { 661 int s, error; 662 663 error = 0; 664 s = splsoftnet(); 665 if ((so->so_state & SS_NOFDREF) == 0) 666 panic("soaccept: !NOFDREF"); 667 so->so_state &= ~SS_NOFDREF; 668 if ((so->so_state & SS_ISDISCONNECTED) == 0 || 669 (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0) 670 error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT, 671 (struct mbuf *)0, nam, (struct mbuf *)0, (struct lwp *)0); 672 else 673 error = ECONNABORTED; 674 675 splx(s); 676 return (error); 677 } 678 679 int 680 soconnect(struct socket *so, struct mbuf *nam, struct lwp *l) 681 { 682 int s, error; 683 684 if (so->so_options & SO_ACCEPTCONN) 685 return (EOPNOTSUPP); 686 s = splsoftnet(); 687 /* 688 * If protocol is connection-based, can only connect once. 689 * Otherwise, if connected, try to disconnect first. 690 * This allows user to disconnect by connecting to, e.g., 691 * a null address. 692 */ 693 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 694 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 695 (error = sodisconnect(so)))) 696 error = EISCONN; 697 else 698 error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT, 699 (struct mbuf *)0, nam, (struct mbuf *)0, l); 700 splx(s); 701 return (error); 702 } 703 704 int 705 soconnect2(struct socket *so1, struct socket *so2) 706 { 707 int s, error; 708 709 s = splsoftnet(); 710 error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2, 711 (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0, 712 (struct lwp *)0); 713 splx(s); 714 return (error); 715 } 716 717 int 718 sodisconnect(struct socket *so) 719 { 720 int s, error; 721 722 s = splsoftnet(); 723 if ((so->so_state & SS_ISCONNECTED) == 0) { 724 error = ENOTCONN; 725 goto bad; 726 } 727 if (so->so_state & SS_ISDISCONNECTING) { 728 error = EALREADY; 729 goto bad; 730 } 731 error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT, 732 (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0, 733 (struct lwp *)0); 734 bad: 735 splx(s); 736 sodopendfree(); 737 return (error); 738 } 739 740 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 741 /* 742 * Send on a socket. 743 * If send must go all at once and message is larger than 744 * send buffering, then hard error. 745 * Lock against other senders. 746 * If must go all at once and not enough room now, then 747 * inform user that this would block and do nothing. 748 * Otherwise, if nonblocking, send as much as possible. 749 * The data to be sent is described by "uio" if nonzero, 750 * otherwise by the mbuf chain "top" (which must be null 751 * if uio is not). Data provided in mbuf chain must be small 752 * enough to send all at once. 753 * 754 * Returns nonzero on error, timeout or signal; callers 755 * must check for short counts if EINTR/ERESTART are returned. 756 * Data and control buffers are freed on return. 757 */ 758 int 759 sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top, 760 struct mbuf *control, int flags, struct lwp *l) 761 { 762 struct mbuf **mp, *m; 763 struct proc *p; 764 long space, len, resid, clen, mlen; 765 int error, s, dontroute, atomic; 766 767 p = l->l_proc; 768 sodopendfree(); 769 770 clen = 0; 771 atomic = sosendallatonce(so) || top; 772 if (uio) 773 resid = uio->uio_resid; 774 else 775 resid = top->m_pkthdr.len; 776 /* 777 * In theory resid should be unsigned. 778 * However, space must be signed, as it might be less than 0 779 * if we over-committed, and we must use a signed comparison 780 * of space and resid. On the other hand, a negative resid 781 * causes us to loop sending 0-length segments to the protocol. 782 */ 783 if (resid < 0) { 784 error = EINVAL; 785 goto out; 786 } 787 dontroute = 788 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 789 (so->so_proto->pr_flags & PR_ATOMIC); 790 if (p) 791 p->p_stats->p_ru.ru_msgsnd++; 792 if (control) 793 clen = control->m_len; 794 #define snderr(errno) { error = errno; splx(s); goto release; } 795 796 restart: 797 if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) 798 goto out; 799 do { 800 s = splsoftnet(); 801 if (so->so_state & SS_CANTSENDMORE) 802 snderr(EPIPE); 803 if (so->so_error) { 804 error = so->so_error; 805 so->so_error = 0; 806 splx(s); 807 goto release; 808 } 809 if ((so->so_state & SS_ISCONNECTED) == 0) { 810 if (so->so_proto->pr_flags & PR_CONNREQUIRED) { 811 if ((so->so_state & SS_ISCONFIRMING) == 0 && 812 !(resid == 0 && clen != 0)) 813 snderr(ENOTCONN); 814 } else if (addr == 0) 815 snderr(EDESTADDRREQ); 816 } 817 space = sbspace(&so->so_snd); 818 if (flags & MSG_OOB) 819 space += 1024; 820 if ((atomic && resid > so->so_snd.sb_hiwat) || 821 clen > so->so_snd.sb_hiwat) 822 snderr(EMSGSIZE); 823 if (space < resid + clen && 824 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 825 if (so->so_state & SS_NBIO) 826 snderr(EWOULDBLOCK); 827 sbunlock(&so->so_snd); 828 error = sbwait(&so->so_snd); 829 splx(s); 830 if (error) 831 goto out; 832 goto restart; 833 } 834 splx(s); 835 mp = ⊤ 836 space -= clen; 837 do { 838 if (uio == NULL) { 839 /* 840 * Data is prepackaged in "top". 841 */ 842 resid = 0; 843 if (flags & MSG_EOR) 844 top->m_flags |= M_EOR; 845 } else do { 846 if (top == 0) { 847 m = m_gethdr(M_WAIT, MT_DATA); 848 mlen = MHLEN; 849 m->m_pkthdr.len = 0; 850 m->m_pkthdr.rcvif = (struct ifnet *)0; 851 } else { 852 m = m_get(M_WAIT, MT_DATA); 853 mlen = MLEN; 854 } 855 MCLAIM(m, so->so_snd.sb_mowner); 856 if (use_sosend_loan && 857 uio->uio_iov->iov_len >= SOCK_LOAN_THRESH && 858 space >= SOCK_LOAN_THRESH && 859 (len = sosend_loan(so, uio, m, 860 space)) != 0) { 861 SOSEND_COUNTER_INCR(&sosend_loan_big); 862 space -= len; 863 goto have_data; 864 } 865 if (resid >= MINCLSIZE && space >= MCLBYTES) { 866 SOSEND_COUNTER_INCR(&sosend_copy_big); 867 m_clget(m, M_WAIT); 868 if ((m->m_flags & M_EXT) == 0) 869 goto nopages; 870 mlen = MCLBYTES; 871 if (atomic && top == 0) { 872 len = lmin(MCLBYTES - max_hdr, 873 resid); 874 m->m_data += max_hdr; 875 } else 876 len = lmin(MCLBYTES, resid); 877 space -= len; 878 } else { 879 nopages: 880 SOSEND_COUNTER_INCR(&sosend_copy_small); 881 len = lmin(lmin(mlen, resid), space); 882 space -= len; 883 /* 884 * For datagram protocols, leave room 885 * for protocol headers in first mbuf. 886 */ 887 if (atomic && top == 0 && len < mlen) 888 MH_ALIGN(m, len); 889 } 890 error = uiomove(mtod(m, caddr_t), (int)len, 891 uio); 892 have_data: 893 resid = uio->uio_resid; 894 m->m_len = len; 895 *mp = m; 896 top->m_pkthdr.len += len; 897 if (error) 898 goto release; 899 mp = &m->m_next; 900 if (resid <= 0) { 901 if (flags & MSG_EOR) 902 top->m_flags |= M_EOR; 903 break; 904 } 905 } while (space > 0 && atomic); 906 907 s = splsoftnet(); 908 909 if (so->so_state & SS_CANTSENDMORE) 910 snderr(EPIPE); 911 912 if (dontroute) 913 so->so_options |= SO_DONTROUTE; 914 if (resid > 0) 915 so->so_state |= SS_MORETOCOME; 916 error = (*so->so_proto->pr_usrreq)(so, 917 (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND, 918 top, addr, control, curlwp); /* XXX */ 919 if (dontroute) 920 so->so_options &= ~SO_DONTROUTE; 921 if (resid > 0) 922 so->so_state &= ~SS_MORETOCOME; 923 splx(s); 924 925 clen = 0; 926 control = 0; 927 top = 0; 928 mp = ⊤ 929 if (error) 930 goto release; 931 } while (resid && space > 0); 932 } while (resid); 933 934 release: 935 sbunlock(&so->so_snd); 936 out: 937 if (top) 938 m_freem(top); 939 if (control) 940 m_freem(control); 941 return (error); 942 } 943 944 /* 945 * Implement receive operations on a socket. 946 * We depend on the way that records are added to the sockbuf 947 * by sbappend*. In particular, each record (mbufs linked through m_next) 948 * must begin with an address if the protocol so specifies, 949 * followed by an optional mbuf or mbufs containing ancillary data, 950 * and then zero or more mbufs of data. 951 * In order to avoid blocking network interrupts for the entire time here, 952 * we splx() while doing the actual copy to user space. 953 * Although the sockbuf is locked, new data may still be appended, 954 * and thus we must maintain consistency of the sockbuf during that time. 955 * 956 * The caller may receive the data as a single mbuf chain by supplying 957 * an mbuf **mp0 for use in returning the chain. The uio is then used 958 * only for the count in uio_resid. 959 */ 960 int 961 soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, 962 struct mbuf **mp0, struct mbuf **controlp, int *flagsp) 963 { 964 struct lwp *l = curlwp; 965 struct mbuf *m, **mp; 966 int flags, len, error, s, offset, moff, type, orig_resid; 967 const struct protosw *pr; 968 struct mbuf *nextrecord; 969 int mbuf_removed = 0; 970 971 pr = so->so_proto; 972 mp = mp0; 973 type = 0; 974 orig_resid = uio->uio_resid; 975 976 if (paddr) 977 *paddr = 0; 978 if (controlp) 979 *controlp = 0; 980 if (flagsp) 981 flags = *flagsp &~ MSG_EOR; 982 else 983 flags = 0; 984 985 if ((flags & MSG_DONTWAIT) == 0) 986 sodopendfree(); 987 988 if (flags & MSG_OOB) { 989 m = m_get(M_WAIT, MT_DATA); 990 error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m, 991 (struct mbuf *)(long)(flags & MSG_PEEK), 992 (struct mbuf *)0, l); 993 if (error) 994 goto bad; 995 do { 996 error = uiomove(mtod(m, caddr_t), 997 (int) min(uio->uio_resid, m->m_len), uio); 998 m = m_free(m); 999 } while (uio->uio_resid && error == 0 && m); 1000 bad: 1001 if (m) 1002 m_freem(m); 1003 return (error); 1004 } 1005 if (mp) 1006 *mp = (struct mbuf *)0; 1007 if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) 1008 (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0, 1009 (struct mbuf *)0, (struct mbuf *)0, l); 1010 1011 restart: 1012 if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) 1013 return (error); 1014 s = splsoftnet(); 1015 1016 m = so->so_rcv.sb_mb; 1017 /* 1018 * If we have less data than requested, block awaiting more 1019 * (subject to any timeout) if: 1020 * 1. the current count is less than the low water mark, 1021 * 2. MSG_WAITALL is set, and it is possible to do the entire 1022 * receive operation at once if we block (resid <= hiwat), or 1023 * 3. MSG_DONTWAIT is not set. 1024 * If MSG_WAITALL is set but resid is larger than the receive buffer, 1025 * we have to do the receive in sections, and thus risk returning 1026 * a short count if a timeout or signal occurs after we start. 1027 */ 1028 if (m == 0 || (((flags & MSG_DONTWAIT) == 0 && 1029 so->so_rcv.sb_cc < uio->uio_resid) && 1030 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 1031 ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) && 1032 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) { 1033 #ifdef DIAGNOSTIC 1034 if (m == 0 && so->so_rcv.sb_cc) 1035 panic("receive 1"); 1036 #endif 1037 if (so->so_error) { 1038 if (m) 1039 goto dontblock; 1040 error = so->so_error; 1041 if ((flags & MSG_PEEK) == 0) 1042 so->so_error = 0; 1043 goto release; 1044 } 1045 if (so->so_state & SS_CANTRCVMORE) { 1046 if (m) 1047 goto dontblock; 1048 else 1049 goto release; 1050 } 1051 for (; m; m = m->m_next) 1052 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 1053 m = so->so_rcv.sb_mb; 1054 goto dontblock; 1055 } 1056 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 1057 (so->so_proto->pr_flags & PR_CONNREQUIRED)) { 1058 error = ENOTCONN; 1059 goto release; 1060 } 1061 if (uio->uio_resid == 0) 1062 goto release; 1063 if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) { 1064 error = EWOULDBLOCK; 1065 goto release; 1066 } 1067 SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1"); 1068 SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1"); 1069 sbunlock(&so->so_rcv); 1070 error = sbwait(&so->so_rcv); 1071 splx(s); 1072 if (error) 1073 return (error); 1074 goto restart; 1075 } 1076 dontblock: 1077 /* 1078 * On entry here, m points to the first record of the socket buffer. 1079 * While we process the initial mbufs containing address and control 1080 * info, we save a copy of m->m_nextpkt into nextrecord. 1081 */ 1082 if (l) 1083 l->l_proc->p_stats->p_ru.ru_msgrcv++; 1084 KASSERT(m == so->so_rcv.sb_mb); 1085 SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); 1086 SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); 1087 nextrecord = m->m_nextpkt; 1088 if (pr->pr_flags & PR_ADDR) { 1089 #ifdef DIAGNOSTIC 1090 if (m->m_type != MT_SONAME) 1091 panic("receive 1a"); 1092 #endif 1093 orig_resid = 0; 1094 if (flags & MSG_PEEK) { 1095 if (paddr) 1096 *paddr = m_copy(m, 0, m->m_len); 1097 m = m->m_next; 1098 } else { 1099 sbfree(&so->so_rcv, m); 1100 mbuf_removed = 1; 1101 if (paddr) { 1102 *paddr = m; 1103 so->so_rcv.sb_mb = m->m_next; 1104 m->m_next = 0; 1105 m = so->so_rcv.sb_mb; 1106 } else { 1107 MFREE(m, so->so_rcv.sb_mb); 1108 m = so->so_rcv.sb_mb; 1109 } 1110 } 1111 } 1112 while (m && m->m_type == MT_CONTROL && error == 0) { 1113 if (flags & MSG_PEEK) { 1114 if (controlp) 1115 *controlp = m_copy(m, 0, m->m_len); 1116 m = m->m_next; 1117 } else { 1118 sbfree(&so->so_rcv, m); 1119 mbuf_removed = 1; 1120 if (controlp) { 1121 struct domain *dom = pr->pr_domain; 1122 if (dom->dom_externalize && l && 1123 mtod(m, struct cmsghdr *)->cmsg_type == 1124 SCM_RIGHTS) 1125 error = (*dom->dom_externalize)(m, l); 1126 *controlp = m; 1127 so->so_rcv.sb_mb = m->m_next; 1128 m->m_next = 0; 1129 m = so->so_rcv.sb_mb; 1130 } else { 1131 /* 1132 * Dispose of any SCM_RIGHTS message that went 1133 * through the read path rather than recv. 1134 */ 1135 if (pr->pr_domain->dom_dispose && 1136 mtod(m, struct cmsghdr *)->cmsg_type == SCM_RIGHTS) 1137 (*pr->pr_domain->dom_dispose)(m); 1138 MFREE(m, so->so_rcv.sb_mb); 1139 m = so->so_rcv.sb_mb; 1140 } 1141 } 1142 if (controlp) { 1143 orig_resid = 0; 1144 controlp = &(*controlp)->m_next; 1145 } 1146 } 1147 1148 /* 1149 * If m is non-NULL, we have some data to read. From now on, 1150 * make sure to keep sb_lastrecord consistent when working on 1151 * the last packet on the chain (nextrecord == NULL) and we 1152 * change m->m_nextpkt. 1153 */ 1154 if (m) { 1155 if ((flags & MSG_PEEK) == 0) { 1156 m->m_nextpkt = nextrecord; 1157 /* 1158 * If nextrecord == NULL (this is a single chain), 1159 * then sb_lastrecord may not be valid here if m 1160 * was changed earlier. 1161 */ 1162 if (nextrecord == NULL) { 1163 KASSERT(so->so_rcv.sb_mb == m); 1164 so->so_rcv.sb_lastrecord = m; 1165 } 1166 } 1167 type = m->m_type; 1168 if (type == MT_OOBDATA) 1169 flags |= MSG_OOB; 1170 } else { 1171 if ((flags & MSG_PEEK) == 0) { 1172 KASSERT(so->so_rcv.sb_mb == m); 1173 so->so_rcv.sb_mb = nextrecord; 1174 SB_EMPTY_FIXUP(&so->so_rcv); 1175 } 1176 } 1177 SBLASTRECORDCHK(&so->so_rcv, "soreceive 2"); 1178 SBLASTMBUFCHK(&so->so_rcv, "soreceive 2"); 1179 1180 moff = 0; 1181 offset = 0; 1182 while (m && uio->uio_resid > 0 && error == 0) { 1183 if (m->m_type == MT_OOBDATA) { 1184 if (type != MT_OOBDATA) 1185 break; 1186 } else if (type == MT_OOBDATA) 1187 break; 1188 #ifdef DIAGNOSTIC 1189 else if (m->m_type != MT_DATA && m->m_type != MT_HEADER) 1190 panic("receive 3"); 1191 #endif 1192 so->so_state &= ~SS_RCVATMARK; 1193 len = uio->uio_resid; 1194 if (so->so_oobmark && len > so->so_oobmark - offset) 1195 len = so->so_oobmark - offset; 1196 if (len > m->m_len - moff) 1197 len = m->m_len - moff; 1198 /* 1199 * If mp is set, just pass back the mbufs. 1200 * Otherwise copy them out via the uio, then free. 1201 * Sockbuf must be consistent here (points to current mbuf, 1202 * it points to next record) when we drop priority; 1203 * we must note any additions to the sockbuf when we 1204 * block interrupts again. 1205 */ 1206 if (mp == 0) { 1207 SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); 1208 SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); 1209 splx(s); 1210 error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio); 1211 s = splsoftnet(); 1212 if (error) { 1213 /* 1214 * If any part of the record has been removed 1215 * (such as the MT_SONAME mbuf, which will 1216 * happen when PR_ADDR, and thus also 1217 * PR_ATOMIC, is set), then drop the entire 1218 * record to maintain the atomicity of the 1219 * receive operation. 1220 * 1221 * This avoids a later panic("receive 1a") 1222 * when compiled with DIAGNOSTIC. 1223 */ 1224 if (m && mbuf_removed 1225 && (pr->pr_flags & PR_ATOMIC)) 1226 (void) sbdroprecord(&so->so_rcv); 1227 1228 goto release; 1229 } 1230 } else 1231 uio->uio_resid -= len; 1232 if (len == m->m_len - moff) { 1233 if (m->m_flags & M_EOR) 1234 flags |= MSG_EOR; 1235 if (flags & MSG_PEEK) { 1236 m = m->m_next; 1237 moff = 0; 1238 } else { 1239 nextrecord = m->m_nextpkt; 1240 sbfree(&so->so_rcv, m); 1241 if (mp) { 1242 *mp = m; 1243 mp = &m->m_next; 1244 so->so_rcv.sb_mb = m = m->m_next; 1245 *mp = (struct mbuf *)0; 1246 } else { 1247 MFREE(m, so->so_rcv.sb_mb); 1248 m = so->so_rcv.sb_mb; 1249 } 1250 /* 1251 * If m != NULL, we also know that 1252 * so->so_rcv.sb_mb != NULL. 1253 */ 1254 KASSERT(so->so_rcv.sb_mb == m); 1255 if (m) { 1256 m->m_nextpkt = nextrecord; 1257 if (nextrecord == NULL) 1258 so->so_rcv.sb_lastrecord = m; 1259 } else { 1260 so->so_rcv.sb_mb = nextrecord; 1261 SB_EMPTY_FIXUP(&so->so_rcv); 1262 } 1263 SBLASTRECORDCHK(&so->so_rcv, "soreceive 3"); 1264 SBLASTMBUFCHK(&so->so_rcv, "soreceive 3"); 1265 } 1266 } else { 1267 if (flags & MSG_PEEK) 1268 moff += len; 1269 else { 1270 if (mp) 1271 *mp = m_copym(m, 0, len, M_WAIT); 1272 m->m_data += len; 1273 m->m_len -= len; 1274 so->so_rcv.sb_cc -= len; 1275 } 1276 } 1277 if (so->so_oobmark) { 1278 if ((flags & MSG_PEEK) == 0) { 1279 so->so_oobmark -= len; 1280 if (so->so_oobmark == 0) { 1281 so->so_state |= SS_RCVATMARK; 1282 break; 1283 } 1284 } else { 1285 offset += len; 1286 if (offset == so->so_oobmark) 1287 break; 1288 } 1289 } 1290 if (flags & MSG_EOR) 1291 break; 1292 /* 1293 * If the MSG_WAITALL flag is set (for non-atomic socket), 1294 * we must not quit until "uio->uio_resid == 0" or an error 1295 * termination. If a signal/timeout occurs, return 1296 * with a short count but without error. 1297 * Keep sockbuf locked against other readers. 1298 */ 1299 while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 && 1300 !sosendallatonce(so) && !nextrecord) { 1301 if (so->so_error || so->so_state & SS_CANTRCVMORE) 1302 break; 1303 /* 1304 * If we are peeking and the socket receive buffer is 1305 * full, stop since we can't get more data to peek at. 1306 */ 1307 if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0) 1308 break; 1309 /* 1310 * If we've drained the socket buffer, tell the 1311 * protocol in case it needs to do something to 1312 * get it filled again. 1313 */ 1314 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) 1315 (*pr->pr_usrreq)(so, PRU_RCVD, 1316 (struct mbuf *)0, 1317 (struct mbuf *)(long)flags, 1318 (struct mbuf *)0, l); 1319 SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2"); 1320 SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2"); 1321 error = sbwait(&so->so_rcv); 1322 if (error) { 1323 sbunlock(&so->so_rcv); 1324 splx(s); 1325 return (0); 1326 } 1327 if ((m = so->so_rcv.sb_mb) != NULL) 1328 nextrecord = m->m_nextpkt; 1329 } 1330 } 1331 1332 if (m && pr->pr_flags & PR_ATOMIC) { 1333 flags |= MSG_TRUNC; 1334 if ((flags & MSG_PEEK) == 0) 1335 (void) sbdroprecord(&so->so_rcv); 1336 } 1337 if ((flags & MSG_PEEK) == 0) { 1338 if (m == 0) { 1339 /* 1340 * First part is an inline SB_EMPTY_FIXUP(). Second 1341 * part makes sure sb_lastrecord is up-to-date if 1342 * there is still data in the socket buffer. 1343 */ 1344 so->so_rcv.sb_mb = nextrecord; 1345 if (so->so_rcv.sb_mb == NULL) { 1346 so->so_rcv.sb_mbtail = NULL; 1347 so->so_rcv.sb_lastrecord = NULL; 1348 } else if (nextrecord->m_nextpkt == NULL) 1349 so->so_rcv.sb_lastrecord = nextrecord; 1350 } 1351 SBLASTRECORDCHK(&so->so_rcv, "soreceive 4"); 1352 SBLASTMBUFCHK(&so->so_rcv, "soreceive 4"); 1353 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1354 (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0, 1355 (struct mbuf *)(long)flags, (struct mbuf *)0, l); 1356 } 1357 if (orig_resid == uio->uio_resid && orig_resid && 1358 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 1359 sbunlock(&so->so_rcv); 1360 splx(s); 1361 goto restart; 1362 } 1363 1364 if (flagsp) 1365 *flagsp |= flags; 1366 release: 1367 sbunlock(&so->so_rcv); 1368 splx(s); 1369 return (error); 1370 } 1371 1372 int 1373 soshutdown(struct socket *so, int how) 1374 { 1375 const struct protosw *pr; 1376 1377 pr = so->so_proto; 1378 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 1379 return (EINVAL); 1380 1381 if (how == SHUT_RD || how == SHUT_RDWR) 1382 sorflush(so); 1383 if (how == SHUT_WR || how == SHUT_RDWR) 1384 return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0, 1385 (struct mbuf *)0, (struct mbuf *)0, (struct lwp *)0); 1386 return (0); 1387 } 1388 1389 void 1390 sorflush(struct socket *so) 1391 { 1392 struct sockbuf *sb, asb; 1393 const struct protosw *pr; 1394 int s; 1395 1396 sb = &so->so_rcv; 1397 pr = so->so_proto; 1398 sb->sb_flags |= SB_NOINTR; 1399 (void) sblock(sb, M_WAITOK); 1400 s = splnet(); 1401 socantrcvmore(so); 1402 sbunlock(sb); 1403 asb = *sb; 1404 /* 1405 * Clear most of the sockbuf structure, but leave some of the 1406 * fields valid. 1407 */ 1408 memset(&sb->sb_startzero, 0, 1409 sizeof(*sb) - offsetof(struct sockbuf, sb_startzero)); 1410 splx(s); 1411 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) 1412 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 1413 sbrelease(&asb, so); 1414 } 1415 1416 int 1417 sosetopt(struct socket *so, int level, int optname, struct mbuf *m0) 1418 { 1419 int error; 1420 struct mbuf *m; 1421 1422 error = 0; 1423 m = m0; 1424 if (level != SOL_SOCKET) { 1425 if (so->so_proto && so->so_proto->pr_ctloutput) 1426 return ((*so->so_proto->pr_ctloutput) 1427 (PRCO_SETOPT, so, level, optname, &m0)); 1428 error = ENOPROTOOPT; 1429 } else { 1430 switch (optname) { 1431 1432 case SO_LINGER: 1433 if (m == NULL || m->m_len != sizeof(struct linger)) { 1434 error = EINVAL; 1435 goto bad; 1436 } 1437 if (mtod(m, struct linger *)->l_linger < 0 || 1438 mtod(m, struct linger *)->l_linger > (INT_MAX / hz)) { 1439 error = EDOM; 1440 goto bad; 1441 } 1442 so->so_linger = mtod(m, struct linger *)->l_linger; 1443 /* fall thru... */ 1444 1445 case SO_DEBUG: 1446 case SO_KEEPALIVE: 1447 case SO_DONTROUTE: 1448 case SO_USELOOPBACK: 1449 case SO_BROADCAST: 1450 case SO_REUSEADDR: 1451 case SO_REUSEPORT: 1452 case SO_OOBINLINE: 1453 case SO_TIMESTAMP: 1454 if (m == NULL || m->m_len < sizeof(int)) { 1455 error = EINVAL; 1456 goto bad; 1457 } 1458 if (*mtod(m, int *)) 1459 so->so_options |= optname; 1460 else 1461 so->so_options &= ~optname; 1462 break; 1463 1464 case SO_SNDBUF: 1465 case SO_RCVBUF: 1466 case SO_SNDLOWAT: 1467 case SO_RCVLOWAT: 1468 { 1469 int optval; 1470 1471 if (m == NULL || m->m_len < sizeof(int)) { 1472 error = EINVAL; 1473 goto bad; 1474 } 1475 1476 /* 1477 * Values < 1 make no sense for any of these 1478 * options, so disallow them. 1479 */ 1480 optval = *mtod(m, int *); 1481 if (optval < 1) { 1482 error = EINVAL; 1483 goto bad; 1484 } 1485 1486 switch (optname) { 1487 1488 case SO_SNDBUF: 1489 case SO_RCVBUF: 1490 if (sbreserve(optname == SO_SNDBUF ? 1491 &so->so_snd : &so->so_rcv, 1492 (u_long) optval, so) == 0) { 1493 error = ENOBUFS; 1494 goto bad; 1495 } 1496 break; 1497 1498 /* 1499 * Make sure the low-water is never greater than 1500 * the high-water. 1501 */ 1502 case SO_SNDLOWAT: 1503 so->so_snd.sb_lowat = 1504 (optval > so->so_snd.sb_hiwat) ? 1505 so->so_snd.sb_hiwat : optval; 1506 break; 1507 case SO_RCVLOWAT: 1508 so->so_rcv.sb_lowat = 1509 (optval > so->so_rcv.sb_hiwat) ? 1510 so->so_rcv.sb_hiwat : optval; 1511 break; 1512 } 1513 break; 1514 } 1515 1516 case SO_SNDTIMEO: 1517 case SO_RCVTIMEO: 1518 { 1519 struct timeval *tv; 1520 int val; 1521 1522 if (m == NULL || m->m_len < sizeof(*tv)) { 1523 error = EINVAL; 1524 goto bad; 1525 } 1526 tv = mtod(m, struct timeval *); 1527 if (tv->tv_sec > (INT_MAX - tv->tv_usec / tick) / hz) { 1528 error = EDOM; 1529 goto bad; 1530 } 1531 val = tv->tv_sec * hz + tv->tv_usec / tick; 1532 if (val == 0 && tv->tv_usec != 0) 1533 val = 1; 1534 1535 switch (optname) { 1536 1537 case SO_SNDTIMEO: 1538 so->so_snd.sb_timeo = val; 1539 break; 1540 case SO_RCVTIMEO: 1541 so->so_rcv.sb_timeo = val; 1542 break; 1543 } 1544 break; 1545 } 1546 1547 default: 1548 error = ENOPROTOOPT; 1549 break; 1550 } 1551 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) { 1552 (void) ((*so->so_proto->pr_ctloutput) 1553 (PRCO_SETOPT, so, level, optname, &m0)); 1554 m = NULL; /* freed by protocol */ 1555 } 1556 } 1557 bad: 1558 if (m) 1559 (void) m_free(m); 1560 return (error); 1561 } 1562 1563 int 1564 sogetopt(struct socket *so, int level, int optname, struct mbuf **mp) 1565 { 1566 struct mbuf *m; 1567 1568 if (level != SOL_SOCKET) { 1569 if (so->so_proto && so->so_proto->pr_ctloutput) { 1570 return ((*so->so_proto->pr_ctloutput) 1571 (PRCO_GETOPT, so, level, optname, mp)); 1572 } else 1573 return (ENOPROTOOPT); 1574 } else { 1575 m = m_get(M_WAIT, MT_SOOPTS); 1576 m->m_len = sizeof(int); 1577 1578 switch (optname) { 1579 1580 case SO_LINGER: 1581 m->m_len = sizeof(struct linger); 1582 mtod(m, struct linger *)->l_onoff = 1583 so->so_options & SO_LINGER; 1584 mtod(m, struct linger *)->l_linger = so->so_linger; 1585 break; 1586 1587 case SO_USELOOPBACK: 1588 case SO_DONTROUTE: 1589 case SO_DEBUG: 1590 case SO_KEEPALIVE: 1591 case SO_REUSEADDR: 1592 case SO_REUSEPORT: 1593 case SO_BROADCAST: 1594 case SO_OOBINLINE: 1595 case SO_TIMESTAMP: 1596 *mtod(m, int *) = so->so_options & optname; 1597 break; 1598 1599 case SO_TYPE: 1600 *mtod(m, int *) = so->so_type; 1601 break; 1602 1603 case SO_ERROR: 1604 *mtod(m, int *) = so->so_error; 1605 so->so_error = 0; 1606 break; 1607 1608 case SO_SNDBUF: 1609 *mtod(m, int *) = so->so_snd.sb_hiwat; 1610 break; 1611 1612 case SO_RCVBUF: 1613 *mtod(m, int *) = so->so_rcv.sb_hiwat; 1614 break; 1615 1616 case SO_SNDLOWAT: 1617 *mtod(m, int *) = so->so_snd.sb_lowat; 1618 break; 1619 1620 case SO_RCVLOWAT: 1621 *mtod(m, int *) = so->so_rcv.sb_lowat; 1622 break; 1623 1624 case SO_SNDTIMEO: 1625 case SO_RCVTIMEO: 1626 { 1627 int val = (optname == SO_SNDTIMEO ? 1628 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 1629 1630 m->m_len = sizeof(struct timeval); 1631 mtod(m, struct timeval *)->tv_sec = val / hz; 1632 mtod(m, struct timeval *)->tv_usec = 1633 (val % hz) * tick; 1634 break; 1635 } 1636 1637 case SO_OVERFLOWED: 1638 *mtod(m, int *) = so->so_rcv.sb_overflowed; 1639 break; 1640 1641 default: 1642 (void)m_free(m); 1643 return (ENOPROTOOPT); 1644 } 1645 *mp = m; 1646 return (0); 1647 } 1648 } 1649 1650 void 1651 sohasoutofband(struct socket *so) 1652 { 1653 fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so); 1654 selwakeup(&so->so_rcv.sb_sel); 1655 } 1656 1657 static void 1658 filt_sordetach(struct knote *kn) 1659 { 1660 struct socket *so; 1661 1662 so = (struct socket *)kn->kn_fp->f_data; 1663 SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext); 1664 if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist)) 1665 so->so_rcv.sb_flags &= ~SB_KNOTE; 1666 } 1667 1668 /*ARGSUSED*/ 1669 static int 1670 filt_soread(struct knote *kn, long hint) 1671 { 1672 struct socket *so; 1673 1674 so = (struct socket *)kn->kn_fp->f_data; 1675 kn->kn_data = so->so_rcv.sb_cc; 1676 if (so->so_state & SS_CANTRCVMORE) { 1677 kn->kn_flags |= EV_EOF; 1678 kn->kn_fflags = so->so_error; 1679 return (1); 1680 } 1681 if (so->so_error) /* temporary udp error */ 1682 return (1); 1683 if (kn->kn_sfflags & NOTE_LOWAT) 1684 return (kn->kn_data >= kn->kn_sdata); 1685 return (kn->kn_data >= so->so_rcv.sb_lowat); 1686 } 1687 1688 static void 1689 filt_sowdetach(struct knote *kn) 1690 { 1691 struct socket *so; 1692 1693 so = (struct socket *)kn->kn_fp->f_data; 1694 SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext); 1695 if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist)) 1696 so->so_snd.sb_flags &= ~SB_KNOTE; 1697 } 1698 1699 /*ARGSUSED*/ 1700 static int 1701 filt_sowrite(struct knote *kn, long hint) 1702 { 1703 struct socket *so; 1704 1705 so = (struct socket *)kn->kn_fp->f_data; 1706 kn->kn_data = sbspace(&so->so_snd); 1707 if (so->so_state & SS_CANTSENDMORE) { 1708 kn->kn_flags |= EV_EOF; 1709 kn->kn_fflags = so->so_error; 1710 return (1); 1711 } 1712 if (so->so_error) /* temporary udp error */ 1713 return (1); 1714 if (((so->so_state & SS_ISCONNECTED) == 0) && 1715 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 1716 return (0); 1717 if (kn->kn_sfflags & NOTE_LOWAT) 1718 return (kn->kn_data >= kn->kn_sdata); 1719 return (kn->kn_data >= so->so_snd.sb_lowat); 1720 } 1721 1722 /*ARGSUSED*/ 1723 static int 1724 filt_solisten(struct knote *kn, long hint) 1725 { 1726 struct socket *so; 1727 1728 so = (struct socket *)kn->kn_fp->f_data; 1729 1730 /* 1731 * Set kn_data to number of incoming connections, not 1732 * counting partial (incomplete) connections. 1733 */ 1734 kn->kn_data = so->so_qlen; 1735 return (kn->kn_data > 0); 1736 } 1737 1738 static const struct filterops solisten_filtops = 1739 { 1, NULL, filt_sordetach, filt_solisten }; 1740 static const struct filterops soread_filtops = 1741 { 1, NULL, filt_sordetach, filt_soread }; 1742 static const struct filterops sowrite_filtops = 1743 { 1, NULL, filt_sowdetach, filt_sowrite }; 1744 1745 int 1746 soo_kqfilter(struct file *fp, struct knote *kn) 1747 { 1748 struct socket *so; 1749 struct sockbuf *sb; 1750 1751 so = (struct socket *)kn->kn_fp->f_data; 1752 switch (kn->kn_filter) { 1753 case EVFILT_READ: 1754 if (so->so_options & SO_ACCEPTCONN) 1755 kn->kn_fop = &solisten_filtops; 1756 else 1757 kn->kn_fop = &soread_filtops; 1758 sb = &so->so_rcv; 1759 break; 1760 case EVFILT_WRITE: 1761 kn->kn_fop = &sowrite_filtops; 1762 sb = &so->so_snd; 1763 break; 1764 default: 1765 return (1); 1766 } 1767 SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext); 1768 sb->sb_flags |= SB_KNOTE; 1769 return (0); 1770 } 1771 1772 #include <sys/sysctl.h> 1773 1774 static int sysctl_kern_somaxkva(SYSCTLFN_PROTO); 1775 1776 /* 1777 * sysctl helper routine for kern.somaxkva. ensures that the given 1778 * value is not too small. 1779 * (XXX should we maybe make sure it's not too large as well?) 1780 */ 1781 static int 1782 sysctl_kern_somaxkva(SYSCTLFN_ARGS) 1783 { 1784 int error, new_somaxkva; 1785 struct sysctlnode node; 1786 int s; 1787 1788 new_somaxkva = somaxkva; 1789 node = *rnode; 1790 node.sysctl_data = &new_somaxkva; 1791 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1792 if (error || newp == NULL) 1793 return (error); 1794 1795 if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */ 1796 return (EINVAL); 1797 1798 s = splvm(); 1799 simple_lock(&so_pendfree_slock); 1800 somaxkva = new_somaxkva; 1801 wakeup(&socurkva); 1802 simple_unlock(&so_pendfree_slock); 1803 splx(s); 1804 1805 return (error); 1806 } 1807 1808 SYSCTL_SETUP(sysctl_kern_somaxkva_setup, "sysctl kern.somaxkva setup") 1809 { 1810 1811 sysctl_createv(clog, 0, NULL, NULL, 1812 CTLFLAG_PERMANENT, 1813 CTLTYPE_NODE, "kern", NULL, 1814 NULL, 0, NULL, 0, 1815 CTL_KERN, CTL_EOL); 1816 1817 sysctl_createv(clog, 0, NULL, NULL, 1818 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1819 CTLTYPE_INT, "somaxkva", 1820 SYSCTL_DESCR("Maximum amount of kernel memory to be " 1821 "used for socket buffers"), 1822 sysctl_kern_somaxkva, 0, NULL, 0, 1823 CTL_KERN, KERN_SOMAXKVA, CTL_EOL); 1824 } 1825
Indexes created Sat Sep 20 22:09:52 GMT 2025