uipc_socket.c revision 1.168.2.2
1 /* $NetBSD: uipc_socket.c,v 1.168.2.2 2008/12/13 01:15:09 haad Exp $ */ 2 3 /*- 4 * Copyright (c) 2002, 2007, 2008 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 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 2004 The FreeBSD Foundation 34 * Copyright (c) 2004 Robert Watson 35 * Copyright (c) 1982, 1986, 1988, 1990, 1993 36 * The Regents of the University of California. All rights reserved. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 3. Neither the name of the University nor the names of its contributors 47 * may be used to endorse or promote products derived from this software 48 * without specific prior written permission. 49 * 50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 60 * SUCH DAMAGE. 61 * 62 * @(#)uipc_socket.c 8.6 (Berkeley) 5/2/95 63 */ 64 65 #include <sys/cdefs.h> 66 __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.168.2.2 2008/12/13 01:15:09 haad Exp $"); 67 68 #include "opt_sock_counters.h" 69 #include "opt_sosend_loan.h" 70 #include "opt_mbuftrace.h" 71 #include "opt_somaxkva.h" 72 #include "opt_multiprocessor.h" /* XXX */ 73 74 #include <sys/param.h> 75 #include <sys/systm.h> 76 #include <sys/proc.h> 77 #include <sys/file.h> 78 #include <sys/filedesc.h> 79 #include <sys/kmem.h> 80 #include <sys/mbuf.h> 81 #include <sys/domain.h> 82 #include <sys/kernel.h> 83 #include <sys/protosw.h> 84 #include <sys/socket.h> 85 #include <sys/socketvar.h> 86 #include <sys/signalvar.h> 87 #include <sys/resourcevar.h> 88 #include <sys/uidinfo.h> 89 #include <sys/event.h> 90 #include <sys/poll.h> 91 #include <sys/kauth.h> 92 #include <sys/mutex.h> 93 #include <sys/condvar.h> 94 95 #include <uvm/uvm.h> 96 97 MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options"); 98 MALLOC_DEFINE(M_SONAME, "soname", "socket name"); 99 100 extern const struct fileops socketops; 101 102 extern int somaxconn; /* patchable (XXX sysctl) */ 103 int somaxconn = SOMAXCONN; 104 kmutex_t *softnet_lock; 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 #if defined(SOSEND_NO_LOAN) || defined(MULTIPROCESSOR) 133 int sock_loan_thresh = -1; 134 #else 135 int sock_loan_thresh = 4096; 136 #endif 137 138 static kmutex_t so_pendfree_lock; 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 kcondvar_t socurkva_cv; 147 148 #define SOCK_LOAN_CHUNK 65536 149 150 static size_t sodopendfree(void); 151 static size_t sodopendfreel(void); 152 153 static void sysctl_kern_somaxkva_setup(void); 154 static struct sysctllog *socket_sysctllog; 155 156 static vsize_t 157 sokvareserve(struct socket *so, vsize_t len) 158 { 159 int error; 160 161 mutex_enter(&so_pendfree_lock); 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 error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock); 180 if (error) { 181 len = 0; 182 break; 183 } 184 } 185 socurkva += len; 186 mutex_exit(&so_pendfree_lock); 187 return len; 188 } 189 190 static void 191 sokvaunreserve(vsize_t len) 192 { 193 194 mutex_enter(&so_pendfree_lock); 195 socurkva -= len; 196 cv_broadcast(&socurkva_cv); 197 mutex_exit(&so_pendfree_lock); 198 } 199 200 /* 201 * sokvaalloc: allocate kva for loan. 202 */ 203 204 vaddr_t 205 sokvaalloc(vsize_t len, struct socket *so) 206 { 207 vaddr_t lva; 208 209 /* 210 * reserve kva. 211 */ 212 213 if (sokvareserve(so, len) == 0) 214 return 0; 215 216 /* 217 * allocate kva. 218 */ 219 220 lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA); 221 if (lva == 0) { 222 sokvaunreserve(len); 223 return (0); 224 } 225 226 return lva; 227 } 228 229 /* 230 * sokvafree: free kva for loan. 231 */ 232 233 void 234 sokvafree(vaddr_t sva, vsize_t len) 235 { 236 237 /* 238 * free kva. 239 */ 240 241 uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY); 242 243 /* 244 * unreserve kva. 245 */ 246 247 sokvaunreserve(len); 248 } 249 250 static void 251 sodoloanfree(struct vm_page **pgs, void *buf, size_t size) 252 { 253 vaddr_t sva, eva; 254 vsize_t len; 255 int npgs; 256 257 KASSERT(pgs != NULL); 258 259 eva = round_page((vaddr_t) buf + size); 260 sva = trunc_page((vaddr_t) buf); 261 len = eva - sva; 262 npgs = len >> PAGE_SHIFT; 263 264 pmap_kremove(sva, len); 265 pmap_update(pmap_kernel()); 266 uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE); 267 sokvafree(sva, len); 268 } 269 270 static size_t 271 sodopendfree(void) 272 { 273 size_t rv; 274 275 if (__predict_true(so_pendfree == NULL)) 276 return 0; 277 278 mutex_enter(&so_pendfree_lock); 279 rv = sodopendfreel(); 280 mutex_exit(&so_pendfree_lock); 281 282 return rv; 283 } 284 285 /* 286 * sodopendfreel: free mbufs on "pendfree" list. 287 * unlock and relock so_pendfree_lock when freeing mbufs. 288 * 289 * => called with so_pendfree_lock held. 290 */ 291 292 static size_t 293 sodopendfreel(void) 294 { 295 struct mbuf *m, *next; 296 size_t rv = 0; 297 298 KASSERT(mutex_owned(&so_pendfree_lock)); 299 300 while (so_pendfree != NULL) { 301 m = so_pendfree; 302 so_pendfree = NULL; 303 mutex_exit(&so_pendfree_lock); 304 305 for (; m != NULL; m = next) { 306 next = m->m_next; 307 KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0); 308 KASSERT(m->m_ext.ext_refcnt == 0); 309 310 rv += m->m_ext.ext_size; 311 sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf, 312 m->m_ext.ext_size); 313 pool_cache_put(mb_cache, m); 314 } 315 316 mutex_enter(&so_pendfree_lock); 317 } 318 319 return (rv); 320 } 321 322 void 323 soloanfree(struct mbuf *m, void *buf, size_t size, void *arg) 324 { 325 326 KASSERT(m != NULL); 327 328 /* 329 * postpone freeing mbuf. 330 * 331 * we can't do it in interrupt context 332 * because we need to put kva back to kernel_map. 333 */ 334 335 mutex_enter(&so_pendfree_lock); 336 m->m_next = so_pendfree; 337 so_pendfree = m; 338 cv_broadcast(&socurkva_cv); 339 mutex_exit(&so_pendfree_lock); 340 } 341 342 static long 343 sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space) 344 { 345 struct iovec *iov = uio->uio_iov; 346 vaddr_t sva, eva; 347 vsize_t len; 348 vaddr_t lva; 349 int npgs, error; 350 vaddr_t va; 351 int i; 352 353 if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace)) 354 return (0); 355 356 if (iov->iov_len < (size_t) space) 357 space = iov->iov_len; 358 if (space > SOCK_LOAN_CHUNK) 359 space = SOCK_LOAN_CHUNK; 360 361 eva = round_page((vaddr_t) iov->iov_base + space); 362 sva = trunc_page((vaddr_t) iov->iov_base); 363 len = eva - sva; 364 npgs = len >> PAGE_SHIFT; 365 366 KASSERT(npgs <= M_EXT_MAXPAGES); 367 368 lva = sokvaalloc(len, so); 369 if (lva == 0) 370 return 0; 371 372 error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len, 373 m->m_ext.ext_pgs, UVM_LOAN_TOPAGE); 374 if (error) { 375 sokvafree(lva, len); 376 return (0); 377 } 378 379 for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE) 380 pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]), 381 VM_PROT_READ); 382 pmap_update(pmap_kernel()); 383 384 lva += (vaddr_t) iov->iov_base & PAGE_MASK; 385 386 MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so); 387 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 388 389 uio->uio_resid -= space; 390 /* uio_offset not updated, not set/used for write(2) */ 391 uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space; 392 uio->uio_iov->iov_len -= space; 393 if (uio->uio_iov->iov_len == 0) { 394 uio->uio_iov++; 395 uio->uio_iovcnt--; 396 } 397 398 return (space); 399 } 400 401 static int 402 sokva_reclaim_callback(struct callback_entry *ce, void *obj, void *arg) 403 { 404 405 KASSERT(ce == &sokva_reclaimerentry); 406 KASSERT(obj == NULL); 407 408 sodopendfree(); 409 if (!vm_map_starved_p(kernel_map)) { 410 return CALLBACK_CHAIN_ABORT; 411 } 412 return CALLBACK_CHAIN_CONTINUE; 413 } 414 415 struct mbuf * 416 getsombuf(struct socket *so, int type) 417 { 418 struct mbuf *m; 419 420 m = m_get(M_WAIT, type); 421 MCLAIM(m, so->so_mowner); 422 return m; 423 } 424 425 void 426 soinit(void) 427 { 428 429 sysctl_kern_somaxkva_setup(); 430 431 mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM); 432 softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 433 cv_init(&socurkva_cv, "sokva"); 434 soinit2(); 435 436 /* Set the initial adjusted socket buffer size. */ 437 if (sb_max_set(sb_max)) 438 panic("bad initial sb_max value: %lu", sb_max); 439 440 callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback, 441 &sokva_reclaimerentry, NULL, sokva_reclaim_callback); 442 } 443 444 /* 445 * Socket operation routines. 446 * These routines are called by the routines in 447 * sys_socket.c or from a system process, and 448 * implement the semantics of socket operations by 449 * switching out to the protocol specific routines. 450 */ 451 /*ARGSUSED*/ 452 int 453 socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l, 454 struct socket *lockso) 455 { 456 const struct protosw *prp; 457 struct socket *so; 458 uid_t uid; 459 int error; 460 kmutex_t *lock; 461 462 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 463 KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type), 464 KAUTH_ARG(proto)); 465 if (error != 0) 466 return error; 467 468 if (proto) 469 prp = pffindproto(dom, proto, type); 470 else 471 prp = pffindtype(dom, type); 472 if (prp == NULL) { 473 /* no support for domain */ 474 if (pffinddomain(dom) == 0) 475 return EAFNOSUPPORT; 476 /* no support for socket type */ 477 if (proto == 0 && type != 0) 478 return EPROTOTYPE; 479 return EPROTONOSUPPORT; 480 } 481 if (prp->pr_usrreq == NULL) 482 return EPROTONOSUPPORT; 483 if (prp->pr_type != type) 484 return EPROTOTYPE; 485 486 so = soget(true); 487 so->so_type = type; 488 so->so_proto = prp; 489 so->so_send = sosend; 490 so->so_receive = soreceive; 491 #ifdef MBUFTRACE 492 so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner; 493 so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner; 494 so->so_mowner = &prp->pr_domain->dom_mowner; 495 #endif 496 uid = kauth_cred_geteuid(l->l_cred); 497 so->so_uidinfo = uid_find(uid); 498 so->so_egid = kauth_cred_getegid(l->l_cred); 499 so->so_cpid = l->l_proc->p_pid; 500 if (lockso != NULL) { 501 /* Caller wants us to share a lock. */ 502 lock = lockso->so_lock; 503 so->so_lock = lock; 504 mutex_obj_hold(lock); 505 mutex_enter(lock); 506 } else { 507 /* Lock assigned and taken during PRU_ATTACH. */ 508 } 509 error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL, 510 (struct mbuf *)(long)proto, NULL, l); 511 KASSERT(solocked(so)); 512 if (error != 0) { 513 so->so_state |= SS_NOFDREF; 514 sofree(so); 515 return error; 516 } 517 sounlock(so); 518 *aso = so; 519 return 0; 520 } 521 522 /* On success, write file descriptor to fdout and return zero. On 523 * failure, return non-zero; *fdout will be undefined. 524 */ 525 int 526 fsocreate(int domain, struct socket **sop, int type, int protocol, 527 struct lwp *l, int *fdout) 528 { 529 struct socket *so; 530 struct file *fp; 531 int fd, error; 532 533 if ((error = fd_allocfile(&fp, &fd)) != 0) 534 return (error); 535 fp->f_flag = FREAD|FWRITE; 536 fp->f_type = DTYPE_SOCKET; 537 fp->f_ops = &socketops; 538 error = socreate(domain, &so, type, protocol, l, NULL); 539 if (error != 0) { 540 fd_abort(curproc, fp, fd); 541 } else { 542 if (sop != NULL) 543 *sop = so; 544 fp->f_data = so; 545 fd_affix(curproc, fp, fd); 546 *fdout = fd; 547 } 548 return error; 549 } 550 551 int 552 sobind(struct socket *so, struct mbuf *nam, struct lwp *l) 553 { 554 int error; 555 556 solock(so); 557 error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l); 558 sounlock(so); 559 return error; 560 } 561 562 int 563 solisten(struct socket *so, int backlog, struct lwp *l) 564 { 565 int error; 566 567 solock(so); 568 if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | 569 SS_ISDISCONNECTING)) != 0) { 570 sounlock(so); 571 return (EOPNOTSUPP); 572 } 573 error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL, 574 NULL, NULL, l); 575 if (error != 0) { 576 sounlock(so); 577 return error; 578 } 579 if (TAILQ_EMPTY(&so->so_q)) 580 so->so_options |= SO_ACCEPTCONN; 581 if (backlog < 0) 582 backlog = 0; 583 so->so_qlimit = min(backlog, somaxconn); 584 sounlock(so); 585 return 0; 586 } 587 588 void 589 sofree(struct socket *so) 590 { 591 u_int refs; 592 593 KASSERT(solocked(so)); 594 595 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) { 596 sounlock(so); 597 return; 598 } 599 if (so->so_head) { 600 /* 601 * We must not decommission a socket that's on the accept(2) 602 * queue. If we do, then accept(2) may hang after select(2) 603 * indicated that the listening socket was ready. 604 */ 605 if (!soqremque(so, 0)) { 606 sounlock(so); 607 return; 608 } 609 } 610 if (so->so_rcv.sb_hiwat) 611 (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, 612 RLIM_INFINITY); 613 if (so->so_snd.sb_hiwat) 614 (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, 615 RLIM_INFINITY); 616 sbrelease(&so->so_snd, so); 617 KASSERT(!cv_has_waiters(&so->so_cv)); 618 KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv)); 619 KASSERT(!cv_has_waiters(&so->so_snd.sb_cv)); 620 sorflush(so); 621 refs = so->so_aborting; /* XXX */ 622 /* Remove acccept filter if one is present. */ 623 if (so->so_accf != NULL) 624 (void)accept_filt_clear(so); 625 sounlock(so); 626 if (refs == 0) /* XXX */ 627 soput(so); 628 } 629 630 /* 631 * Close a socket on last file table reference removal. 632 * Initiate disconnect if connected. 633 * Free socket when disconnect complete. 634 */ 635 int 636 soclose(struct socket *so) 637 { 638 struct socket *so2; 639 int error; 640 int error2; 641 642 error = 0; 643 solock(so); 644 if (so->so_options & SO_ACCEPTCONN) { 645 for (;;) { 646 if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) { 647 KASSERT(solocked2(so, so2)); 648 (void) soqremque(so2, 0); 649 /* soabort drops the lock. */ 650 (void) soabort(so2); 651 solock(so); 652 continue; 653 } 654 if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) { 655 KASSERT(solocked2(so, so2)); 656 (void) soqremque(so2, 1); 657 /* soabort drops the lock. */ 658 (void) soabort(so2); 659 solock(so); 660 continue; 661 } 662 break; 663 } 664 } 665 if (so->so_pcb == 0) 666 goto discard; 667 if (so->so_state & SS_ISCONNECTED) { 668 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 669 error = sodisconnect(so); 670 if (error) 671 goto drop; 672 } 673 if (so->so_options & SO_LINGER) { 674 if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio) 675 goto drop; 676 while (so->so_state & SS_ISCONNECTED) { 677 error = sowait(so, so->so_linger * hz); 678 if (error) 679 break; 680 } 681 } 682 } 683 drop: 684 if (so->so_pcb) { 685 error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH, 686 NULL, NULL, NULL, NULL); 687 if (error == 0) 688 error = error2; 689 } 690 discard: 691 if (so->so_state & SS_NOFDREF) 692 panic("soclose: NOFDREF"); 693 so->so_state |= SS_NOFDREF; 694 sofree(so); 695 return (error); 696 } 697 698 /* 699 * Must be called with the socket locked.. Will return with it unlocked. 700 */ 701 int 702 soabort(struct socket *so) 703 { 704 u_int refs; 705 int error; 706 707 KASSERT(solocked(so)); 708 KASSERT(so->so_head == NULL); 709 710 so->so_aborting++; /* XXX */ 711 error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL, 712 NULL, NULL, NULL); 713 refs = --so->so_aborting; /* XXX */ 714 if (error || (refs == 0)) { 715 sofree(so); 716 } else { 717 sounlock(so); 718 } 719 return error; 720 } 721 722 int 723 soaccept(struct socket *so, struct mbuf *nam) 724 { 725 int error; 726 727 KASSERT(solocked(so)); 728 729 error = 0; 730 if ((so->so_state & SS_NOFDREF) == 0) 731 panic("soaccept: !NOFDREF"); 732 so->so_state &= ~SS_NOFDREF; 733 if ((so->so_state & SS_ISDISCONNECTED) == 0 || 734 (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0) 735 error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT, 736 NULL, nam, NULL, NULL); 737 else 738 error = ECONNABORTED; 739 740 return (error); 741 } 742 743 int 744 soconnect(struct socket *so, struct mbuf *nam, struct lwp *l) 745 { 746 int error; 747 748 KASSERT(solocked(so)); 749 750 if (so->so_options & SO_ACCEPTCONN) 751 return (EOPNOTSUPP); 752 /* 753 * If protocol is connection-based, can only connect once. 754 * Otherwise, if connected, try to disconnect first. 755 * This allows user to disconnect by connecting to, e.g., 756 * a null address. 757 */ 758 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 759 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 760 (error = sodisconnect(so)))) 761 error = EISCONN; 762 else 763 error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT, 764 NULL, nam, NULL, l); 765 return (error); 766 } 767 768 int 769 soconnect2(struct socket *so1, struct socket *so2) 770 { 771 int error; 772 773 KASSERT(solocked2(so1, so2)); 774 775 error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2, 776 NULL, (struct mbuf *)so2, NULL, NULL); 777 return (error); 778 } 779 780 int 781 sodisconnect(struct socket *so) 782 { 783 int error; 784 785 KASSERT(solocked(so)); 786 787 if ((so->so_state & SS_ISCONNECTED) == 0) { 788 error = ENOTCONN; 789 } else if (so->so_state & SS_ISDISCONNECTING) { 790 error = EALREADY; 791 } else { 792 error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT, 793 NULL, NULL, NULL, NULL); 794 } 795 sodopendfree(); 796 return (error); 797 } 798 799 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 800 /* 801 * Send on a socket. 802 * If send must go all at once and message is larger than 803 * send buffering, then hard error. 804 * Lock against other senders. 805 * If must go all at once and not enough room now, then 806 * inform user that this would block and do nothing. 807 * Otherwise, if nonblocking, send as much as possible. 808 * The data to be sent is described by "uio" if nonzero, 809 * otherwise by the mbuf chain "top" (which must be null 810 * if uio is not). Data provided in mbuf chain must be small 811 * enough to send all at once. 812 * 813 * Returns nonzero on error, timeout or signal; callers 814 * must check for short counts if EINTR/ERESTART are returned. 815 * Data and control buffers are freed on return. 816 */ 817 int 818 sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top, 819 struct mbuf *control, int flags, struct lwp *l) 820 { 821 struct mbuf **mp, *m; 822 struct proc *p; 823 long space, len, resid, clen, mlen; 824 int error, s, dontroute, atomic; 825 826 p = l->l_proc; 827 sodopendfree(); 828 clen = 0; 829 830 /* 831 * solock() provides atomicity of access. splsoftnet() prevents 832 * protocol processing soft interrupts from interrupting us and 833 * blocking (expensive). 834 */ 835 s = splsoftnet(); 836 solock(so); 837 atomic = sosendallatonce(so) || top; 838 if (uio) 839 resid = uio->uio_resid; 840 else 841 resid = top->m_pkthdr.len; 842 /* 843 * In theory resid should be unsigned. 844 * However, space must be signed, as it might be less than 0 845 * if we over-committed, and we must use a signed comparison 846 * of space and resid. On the other hand, a negative resid 847 * causes us to loop sending 0-length segments to the protocol. 848 */ 849 if (resid < 0) { 850 error = EINVAL; 851 goto out; 852 } 853 dontroute = 854 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 855 (so->so_proto->pr_flags & PR_ATOMIC); 856 l->l_ru.ru_msgsnd++; 857 if (control) 858 clen = control->m_len; 859 restart: 860 if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) 861 goto out; 862 do { 863 if (so->so_state & SS_CANTSENDMORE) { 864 error = EPIPE; 865 goto release; 866 } 867 if (so->so_error) { 868 error = so->so_error; 869 so->so_error = 0; 870 goto release; 871 } 872 if ((so->so_state & SS_ISCONNECTED) == 0) { 873 if (so->so_proto->pr_flags & PR_CONNREQUIRED) { 874 if ((so->so_state & SS_ISCONFIRMING) == 0 && 875 !(resid == 0 && clen != 0)) { 876 error = ENOTCONN; 877 goto release; 878 } 879 } else if (addr == 0) { 880 error = EDESTADDRREQ; 881 goto release; 882 } 883 } 884 space = sbspace(&so->so_snd); 885 if (flags & MSG_OOB) 886 space += 1024; 887 if ((atomic && resid > so->so_snd.sb_hiwat) || 888 clen > so->so_snd.sb_hiwat) { 889 error = EMSGSIZE; 890 goto release; 891 } 892 if (space < resid + clen && 893 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 894 if (so->so_nbio) { 895 error = EWOULDBLOCK; 896 goto release; 897 } 898 sbunlock(&so->so_snd); 899 error = sbwait(&so->so_snd); 900 if (error) 901 goto out; 902 goto restart; 903 } 904 mp = ⊤ 905 space -= clen; 906 do { 907 if (uio == NULL) { 908 /* 909 * Data is prepackaged in "top". 910 */ 911 resid = 0; 912 if (flags & MSG_EOR) 913 top->m_flags |= M_EOR; 914 } else do { 915 sounlock(so); 916 splx(s); 917 if (top == NULL) { 918 m = m_gethdr(M_WAIT, MT_DATA); 919 mlen = MHLEN; 920 m->m_pkthdr.len = 0; 921 m->m_pkthdr.rcvif = NULL; 922 } else { 923 m = m_get(M_WAIT, MT_DATA); 924 mlen = MLEN; 925 } 926 MCLAIM(m, so->so_snd.sb_mowner); 927 if (sock_loan_thresh >= 0 && 928 uio->uio_iov->iov_len >= sock_loan_thresh && 929 space >= sock_loan_thresh && 930 (len = sosend_loan(so, uio, m, 931 space)) != 0) { 932 SOSEND_COUNTER_INCR(&sosend_loan_big); 933 space -= len; 934 goto have_data; 935 } 936 if (resid >= MINCLSIZE && space >= MCLBYTES) { 937 SOSEND_COUNTER_INCR(&sosend_copy_big); 938 m_clget(m, M_WAIT); 939 if ((m->m_flags & M_EXT) == 0) 940 goto nopages; 941 mlen = MCLBYTES; 942 if (atomic && top == 0) { 943 len = lmin(MCLBYTES - max_hdr, 944 resid); 945 m->m_data += max_hdr; 946 } else 947 len = lmin(MCLBYTES, resid); 948 space -= len; 949 } else { 950 nopages: 951 SOSEND_COUNTER_INCR(&sosend_copy_small); 952 len = lmin(lmin(mlen, resid), space); 953 space -= len; 954 /* 955 * For datagram protocols, leave room 956 * for protocol headers in first mbuf. 957 */ 958 if (atomic && top == 0 && len < mlen) 959 MH_ALIGN(m, len); 960 } 961 error = uiomove(mtod(m, void *), (int)len, uio); 962 have_data: 963 resid = uio->uio_resid; 964 m->m_len = len; 965 *mp = m; 966 top->m_pkthdr.len += len; 967 s = splsoftnet(); 968 solock(so); 969 if (error != 0) 970 goto release; 971 mp = &m->m_next; 972 if (resid <= 0) { 973 if (flags & MSG_EOR) 974 top->m_flags |= M_EOR; 975 break; 976 } 977 } while (space > 0 && atomic); 978 979 if (so->so_state & SS_CANTSENDMORE) { 980 error = EPIPE; 981 goto release; 982 } 983 if (dontroute) 984 so->so_options |= SO_DONTROUTE; 985 if (resid > 0) 986 so->so_state |= SS_MORETOCOME; 987 error = (*so->so_proto->pr_usrreq)(so, 988 (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND, 989 top, addr, control, curlwp); 990 if (dontroute) 991 so->so_options &= ~SO_DONTROUTE; 992 if (resid > 0) 993 so->so_state &= ~SS_MORETOCOME; 994 clen = 0; 995 control = NULL; 996 top = NULL; 997 mp = ⊤ 998 if (error != 0) 999 goto release; 1000 } while (resid && space > 0); 1001 } while (resid); 1002 1003 release: 1004 sbunlock(&so->so_snd); 1005 out: 1006 sounlock(so); 1007 splx(s); 1008 if (top) 1009 m_freem(top); 1010 if (control) 1011 m_freem(control); 1012 return (error); 1013 } 1014 1015 /* 1016 * Following replacement or removal of the first mbuf on the first 1017 * mbuf chain of a socket buffer, push necessary state changes back 1018 * into the socket buffer so that other consumers see the values 1019 * consistently. 'nextrecord' is the callers locally stored value of 1020 * the original value of sb->sb_mb->m_nextpkt which must be restored 1021 * when the lead mbuf changes. NOTE: 'nextrecord' may be NULL. 1022 */ 1023 static void 1024 sbsync(struct sockbuf *sb, struct mbuf *nextrecord) 1025 { 1026 1027 KASSERT(solocked(sb->sb_so)); 1028 1029 /* 1030 * First, update for the new value of nextrecord. If necessary, 1031 * make it the first record. 1032 */ 1033 if (sb->sb_mb != NULL) 1034 sb->sb_mb->m_nextpkt = nextrecord; 1035 else 1036 sb->sb_mb = nextrecord; 1037 1038 /* 1039 * Now update any dependent socket buffer fields to reflect 1040 * the new state. This is an inline of SB_EMPTY_FIXUP, with 1041 * the addition of a second clause that takes care of the 1042 * case where sb_mb has been updated, but remains the last 1043 * record. 1044 */ 1045 if (sb->sb_mb == NULL) { 1046 sb->sb_mbtail = NULL; 1047 sb->sb_lastrecord = NULL; 1048 } else if (sb->sb_mb->m_nextpkt == NULL) 1049 sb->sb_lastrecord = sb->sb_mb; 1050 } 1051 1052 /* 1053 * Implement receive operations on a socket. 1054 * We depend on the way that records are added to the sockbuf 1055 * by sbappend*. In particular, each record (mbufs linked through m_next) 1056 * must begin with an address if the protocol so specifies, 1057 * followed by an optional mbuf or mbufs containing ancillary data, 1058 * and then zero or more mbufs of data. 1059 * In order to avoid blocking network interrupts for the entire time here, 1060 * we splx() while doing the actual copy to user space. 1061 * Although the sockbuf is locked, new data may still be appended, 1062 * and thus we must maintain consistency of the sockbuf during that time. 1063 * 1064 * The caller may receive the data as a single mbuf chain by supplying 1065 * an mbuf **mp0 for use in returning the chain. The uio is then used 1066 * only for the count in uio_resid. 1067 */ 1068 int 1069 soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, 1070 struct mbuf **mp0, struct mbuf **controlp, int *flagsp) 1071 { 1072 struct lwp *l = curlwp; 1073 struct mbuf *m, **mp, *mt; 1074 int atomic, flags, len, error, s, offset, moff, type, orig_resid; 1075 const struct protosw *pr; 1076 struct mbuf *nextrecord; 1077 int mbuf_removed = 0; 1078 const struct domain *dom; 1079 1080 pr = so->so_proto; 1081 atomic = pr->pr_flags & PR_ATOMIC; 1082 dom = pr->pr_domain; 1083 mp = mp0; 1084 type = 0; 1085 orig_resid = uio->uio_resid; 1086 1087 if (paddr != NULL) 1088 *paddr = NULL; 1089 if (controlp != NULL) 1090 *controlp = NULL; 1091 if (flagsp != NULL) 1092 flags = *flagsp &~ MSG_EOR; 1093 else 1094 flags = 0; 1095 1096 if ((flags & MSG_DONTWAIT) == 0) 1097 sodopendfree(); 1098 1099 if (flags & MSG_OOB) { 1100 m = m_get(M_WAIT, MT_DATA); 1101 solock(so); 1102 error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m, 1103 (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l); 1104 sounlock(so); 1105 if (error) 1106 goto bad; 1107 do { 1108 error = uiomove(mtod(m, void *), 1109 (int) min(uio->uio_resid, m->m_len), uio); 1110 m = m_free(m); 1111 } while (uio->uio_resid > 0 && error == 0 && m); 1112 bad: 1113 if (m != NULL) 1114 m_freem(m); 1115 return error; 1116 } 1117 if (mp != NULL) 1118 *mp = NULL; 1119 1120 /* 1121 * solock() provides atomicity of access. splsoftnet() prevents 1122 * protocol processing soft interrupts from interrupting us and 1123 * blocking (expensive). 1124 */ 1125 s = splsoftnet(); 1126 solock(so); 1127 if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) 1128 (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l); 1129 1130 restart: 1131 if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) { 1132 sounlock(so); 1133 splx(s); 1134 return error; 1135 } 1136 1137 m = so->so_rcv.sb_mb; 1138 /* 1139 * If we have less data than requested, block awaiting more 1140 * (subject to any timeout) if: 1141 * 1. the current count is less than the low water mark, 1142 * 2. MSG_WAITALL is set, and it is possible to do the entire 1143 * receive operation at once if we block (resid <= hiwat), or 1144 * 3. MSG_DONTWAIT is not set. 1145 * If MSG_WAITALL is set but resid is larger than the receive buffer, 1146 * we have to do the receive in sections, and thus risk returning 1147 * a short count if a timeout or signal occurs after we start. 1148 */ 1149 if (m == NULL || 1150 ((flags & MSG_DONTWAIT) == 0 && 1151 so->so_rcv.sb_cc < uio->uio_resid && 1152 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 1153 ((flags & MSG_WAITALL) && 1154 uio->uio_resid <= so->so_rcv.sb_hiwat)) && 1155 m->m_nextpkt == NULL && !atomic)) { 1156 #ifdef DIAGNOSTIC 1157 if (m == NULL && so->so_rcv.sb_cc) 1158 panic("receive 1"); 1159 #endif 1160 if (so->so_error) { 1161 if (m != NULL) 1162 goto dontblock; 1163 error = so->so_error; 1164 if ((flags & MSG_PEEK) == 0) 1165 so->so_error = 0; 1166 goto release; 1167 } 1168 if (so->so_state & SS_CANTRCVMORE) { 1169 if (m != NULL) 1170 goto dontblock; 1171 else 1172 goto release; 1173 } 1174 for (; m != NULL; m = m->m_next) 1175 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 1176 m = so->so_rcv.sb_mb; 1177 goto dontblock; 1178 } 1179 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 1180 (so->so_proto->pr_flags & PR_CONNREQUIRED)) { 1181 error = ENOTCONN; 1182 goto release; 1183 } 1184 if (uio->uio_resid == 0) 1185 goto release; 1186 if (so->so_nbio || (flags & MSG_DONTWAIT)) { 1187 error = EWOULDBLOCK; 1188 goto release; 1189 } 1190 SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1"); 1191 SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1"); 1192 sbunlock(&so->so_rcv); 1193 error = sbwait(&so->so_rcv); 1194 if (error != 0) { 1195 sounlock(so); 1196 splx(s); 1197 return error; 1198 } 1199 goto restart; 1200 } 1201 dontblock: 1202 /* 1203 * On entry here, m points to the first record of the socket buffer. 1204 * From this point onward, we maintain 'nextrecord' as a cache of the 1205 * pointer to the next record in the socket buffer. We must keep the 1206 * various socket buffer pointers and local stack versions of the 1207 * pointers in sync, pushing out modifications before dropping the 1208 * socket lock, and re-reading them when picking it up. 1209 * 1210 * Otherwise, we will race with the network stack appending new data 1211 * or records onto the socket buffer by using inconsistent/stale 1212 * versions of the field, possibly resulting in socket buffer 1213 * corruption. 1214 * 1215 * By holding the high-level sblock(), we prevent simultaneous 1216 * readers from pulling off the front of the socket buffer. 1217 */ 1218 if (l != NULL) 1219 l->l_ru.ru_msgrcv++; 1220 KASSERT(m == so->so_rcv.sb_mb); 1221 SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); 1222 SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); 1223 nextrecord = m->m_nextpkt; 1224 if (pr->pr_flags & PR_ADDR) { 1225 #ifdef DIAGNOSTIC 1226 if (m->m_type != MT_SONAME) 1227 panic("receive 1a"); 1228 #endif 1229 orig_resid = 0; 1230 if (flags & MSG_PEEK) { 1231 if (paddr) 1232 *paddr = m_copy(m, 0, m->m_len); 1233 m = m->m_next; 1234 } else { 1235 sbfree(&so->so_rcv, m); 1236 mbuf_removed = 1; 1237 if (paddr != NULL) { 1238 *paddr = m; 1239 so->so_rcv.sb_mb = m->m_next; 1240 m->m_next = NULL; 1241 m = so->so_rcv.sb_mb; 1242 } else { 1243 MFREE(m, so->so_rcv.sb_mb); 1244 m = so->so_rcv.sb_mb; 1245 } 1246 sbsync(&so->so_rcv, nextrecord); 1247 } 1248 } 1249 1250 /* 1251 * Process one or more MT_CONTROL mbufs present before any data mbufs 1252 * in the first mbuf chain on the socket buffer. If MSG_PEEK, we 1253 * just copy the data; if !MSG_PEEK, we call into the protocol to 1254 * perform externalization (or freeing if controlp == NULL). 1255 */ 1256 if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) { 1257 struct mbuf *cm = NULL, *cmn; 1258 struct mbuf **cme = &cm; 1259 1260 do { 1261 if (flags & MSG_PEEK) { 1262 if (controlp != NULL) { 1263 *controlp = m_copy(m, 0, m->m_len); 1264 controlp = &(*controlp)->m_next; 1265 } 1266 m = m->m_next; 1267 } else { 1268 sbfree(&so->so_rcv, m); 1269 so->so_rcv.sb_mb = m->m_next; 1270 m->m_next = NULL; 1271 *cme = m; 1272 cme = &(*cme)->m_next; 1273 m = so->so_rcv.sb_mb; 1274 } 1275 } while (m != NULL && m->m_type == MT_CONTROL); 1276 if ((flags & MSG_PEEK) == 0) 1277 sbsync(&so->so_rcv, nextrecord); 1278 for (; cm != NULL; cm = cmn) { 1279 cmn = cm->m_next; 1280 cm->m_next = NULL; 1281 type = mtod(cm, struct cmsghdr *)->cmsg_type; 1282 if (controlp != NULL) { 1283 if (dom->dom_externalize != NULL && 1284 type == SCM_RIGHTS) { 1285 sounlock(so); 1286 splx(s); 1287 error = (*dom->dom_externalize)(cm, l); 1288 s = splsoftnet(); 1289 solock(so); 1290 } 1291 *controlp = cm; 1292 while (*controlp != NULL) 1293 controlp = &(*controlp)->m_next; 1294 } else { 1295 /* 1296 * Dispose of any SCM_RIGHTS message that went 1297 * through the read path rather than recv. 1298 */ 1299 if (dom->dom_dispose != NULL && 1300 type == SCM_RIGHTS) { 1301 sounlock(so); 1302 (*dom->dom_dispose)(cm); 1303 solock(so); 1304 } 1305 m_freem(cm); 1306 } 1307 } 1308 if (m != NULL) 1309 nextrecord = so->so_rcv.sb_mb->m_nextpkt; 1310 else 1311 nextrecord = so->so_rcv.sb_mb; 1312 orig_resid = 0; 1313 } 1314 1315 /* If m is non-NULL, we have some data to read. */ 1316 if (__predict_true(m != NULL)) { 1317 type = m->m_type; 1318 if (type == MT_OOBDATA) 1319 flags |= MSG_OOB; 1320 } 1321 SBLASTRECORDCHK(&so->so_rcv, "soreceive 2"); 1322 SBLASTMBUFCHK(&so->so_rcv, "soreceive 2"); 1323 1324 moff = 0; 1325 offset = 0; 1326 while (m != NULL && uio->uio_resid > 0 && error == 0) { 1327 if (m->m_type == MT_OOBDATA) { 1328 if (type != MT_OOBDATA) 1329 break; 1330 } else if (type == MT_OOBDATA) 1331 break; 1332 #ifdef DIAGNOSTIC 1333 else if (m->m_type != MT_DATA && m->m_type != MT_HEADER) 1334 panic("receive 3"); 1335 #endif 1336 so->so_state &= ~SS_RCVATMARK; 1337 len = uio->uio_resid; 1338 if (so->so_oobmark && len > so->so_oobmark - offset) 1339 len = so->so_oobmark - offset; 1340 if (len > m->m_len - moff) 1341 len = m->m_len - moff; 1342 /* 1343 * If mp is set, just pass back the mbufs. 1344 * Otherwise copy them out via the uio, then free. 1345 * Sockbuf must be consistent here (points to current mbuf, 1346 * it points to next record) when we drop priority; 1347 * we must note any additions to the sockbuf when we 1348 * block interrupts again. 1349 */ 1350 if (mp == NULL) { 1351 SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); 1352 SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); 1353 sounlock(so); 1354 splx(s); 1355 error = uiomove(mtod(m, char *) + moff, (int)len, uio); 1356 s = splsoftnet(); 1357 solock(so); 1358 if (error != 0) { 1359 /* 1360 * If any part of the record has been removed 1361 * (such as the MT_SONAME mbuf, which will 1362 * happen when PR_ADDR, and thus also 1363 * PR_ATOMIC, is set), then drop the entire 1364 * record to maintain the atomicity of the 1365 * receive operation. 1366 * 1367 * This avoids a later panic("receive 1a") 1368 * when compiled with DIAGNOSTIC. 1369 */ 1370 if (m && mbuf_removed && atomic) 1371 (void) sbdroprecord(&so->so_rcv); 1372 1373 goto release; 1374 } 1375 } else 1376 uio->uio_resid -= len; 1377 if (len == m->m_len - moff) { 1378 if (m->m_flags & M_EOR) 1379 flags |= MSG_EOR; 1380 if (flags & MSG_PEEK) { 1381 m = m->m_next; 1382 moff = 0; 1383 } else { 1384 nextrecord = m->m_nextpkt; 1385 sbfree(&so->so_rcv, m); 1386 if (mp) { 1387 *mp = m; 1388 mp = &m->m_next; 1389 so->so_rcv.sb_mb = m = m->m_next; 1390 *mp = NULL; 1391 } else { 1392 MFREE(m, so->so_rcv.sb_mb); 1393 m = so->so_rcv.sb_mb; 1394 } 1395 /* 1396 * If m != NULL, we also know that 1397 * so->so_rcv.sb_mb != NULL. 1398 */ 1399 KASSERT(so->so_rcv.sb_mb == m); 1400 if (m) { 1401 m->m_nextpkt = nextrecord; 1402 if (nextrecord == NULL) 1403 so->so_rcv.sb_lastrecord = m; 1404 } else { 1405 so->so_rcv.sb_mb = nextrecord; 1406 SB_EMPTY_FIXUP(&so->so_rcv); 1407 } 1408 SBLASTRECORDCHK(&so->so_rcv, "soreceive 3"); 1409 SBLASTMBUFCHK(&so->so_rcv, "soreceive 3"); 1410 } 1411 } else if (flags & MSG_PEEK) 1412 moff += len; 1413 else { 1414 if (mp != NULL) { 1415 mt = m_copym(m, 0, len, M_NOWAIT); 1416 if (__predict_false(mt == NULL)) { 1417 sounlock(so); 1418 mt = m_copym(m, 0, len, M_WAIT); 1419 solock(so); 1420 } 1421 *mp = mt; 1422 } 1423 m->m_data += len; 1424 m->m_len -= len; 1425 so->so_rcv.sb_cc -= len; 1426 } 1427 if (so->so_oobmark) { 1428 if ((flags & MSG_PEEK) == 0) { 1429 so->so_oobmark -= len; 1430 if (so->so_oobmark == 0) { 1431 so->so_state |= SS_RCVATMARK; 1432 break; 1433 } 1434 } else { 1435 offset += len; 1436 if (offset == so->so_oobmark) 1437 break; 1438 } 1439 } 1440 if (flags & MSG_EOR) 1441 break; 1442 /* 1443 * If the MSG_WAITALL flag is set (for non-atomic socket), 1444 * we must not quit until "uio->uio_resid == 0" or an error 1445 * termination. If a signal/timeout occurs, return 1446 * with a short count but without error. 1447 * Keep sockbuf locked against other readers. 1448 */ 1449 while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 && 1450 !sosendallatonce(so) && !nextrecord) { 1451 if (so->so_error || so->so_state & SS_CANTRCVMORE) 1452 break; 1453 /* 1454 * If we are peeking and the socket receive buffer is 1455 * full, stop since we can't get more data to peek at. 1456 */ 1457 if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0) 1458 break; 1459 /* 1460 * If we've drained the socket buffer, tell the 1461 * protocol in case it needs to do something to 1462 * get it filled again. 1463 */ 1464 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) 1465 (*pr->pr_usrreq)(so, PRU_RCVD, 1466 NULL, (struct mbuf *)(long)flags, NULL, l); 1467 SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2"); 1468 SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2"); 1469 error = sbwait(&so->so_rcv); 1470 if (error != 0) { 1471 sbunlock(&so->so_rcv); 1472 sounlock(so); 1473 splx(s); 1474 return 0; 1475 } 1476 if ((m = so->so_rcv.sb_mb) != NULL) 1477 nextrecord = m->m_nextpkt; 1478 } 1479 } 1480 1481 if (m && atomic) { 1482 flags |= MSG_TRUNC; 1483 if ((flags & MSG_PEEK) == 0) 1484 (void) sbdroprecord(&so->so_rcv); 1485 } 1486 if ((flags & MSG_PEEK) == 0) { 1487 if (m == NULL) { 1488 /* 1489 * First part is an inline SB_EMPTY_FIXUP(). Second 1490 * part makes sure sb_lastrecord is up-to-date if 1491 * there is still data in the socket buffer. 1492 */ 1493 so->so_rcv.sb_mb = nextrecord; 1494 if (so->so_rcv.sb_mb == NULL) { 1495 so->so_rcv.sb_mbtail = NULL; 1496 so->so_rcv.sb_lastrecord = NULL; 1497 } else if (nextrecord->m_nextpkt == NULL) 1498 so->so_rcv.sb_lastrecord = nextrecord; 1499 } 1500 SBLASTRECORDCHK(&so->so_rcv, "soreceive 4"); 1501 SBLASTMBUFCHK(&so->so_rcv, "soreceive 4"); 1502 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1503 (*pr->pr_usrreq)(so, PRU_RCVD, NULL, 1504 (struct mbuf *)(long)flags, NULL, l); 1505 } 1506 if (orig_resid == uio->uio_resid && orig_resid && 1507 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 1508 sbunlock(&so->so_rcv); 1509 goto restart; 1510 } 1511 1512 if (flagsp != NULL) 1513 *flagsp |= flags; 1514 release: 1515 sbunlock(&so->so_rcv); 1516 sounlock(so); 1517 splx(s); 1518 return error; 1519 } 1520 1521 int 1522 soshutdown(struct socket *so, int how) 1523 { 1524 const struct protosw *pr; 1525 int error; 1526 1527 KASSERT(solocked(so)); 1528 1529 pr = so->so_proto; 1530 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 1531 return (EINVAL); 1532 1533 if (how == SHUT_RD || how == SHUT_RDWR) { 1534 sorflush(so); 1535 error = 0; 1536 } 1537 if (how == SHUT_WR || how == SHUT_RDWR) 1538 error = (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL, 1539 NULL, NULL, NULL); 1540 1541 return error; 1542 } 1543 1544 void 1545 sorflush(struct socket *so) 1546 { 1547 struct sockbuf *sb, asb; 1548 const struct protosw *pr; 1549 1550 KASSERT(solocked(so)); 1551 1552 sb = &so->so_rcv; 1553 pr = so->so_proto; 1554 socantrcvmore(so); 1555 sb->sb_flags |= SB_NOINTR; 1556 (void )sblock(sb, M_WAITOK); 1557 sbunlock(sb); 1558 asb = *sb; 1559 /* 1560 * Clear most of the sockbuf structure, but leave some of the 1561 * fields valid. 1562 */ 1563 memset(&sb->sb_startzero, 0, 1564 sizeof(*sb) - offsetof(struct sockbuf, sb_startzero)); 1565 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) { 1566 sounlock(so); 1567 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 1568 solock(so); 1569 } 1570 sbrelease(&asb, so); 1571 } 1572 1573 /* 1574 * internal set SOL_SOCKET options 1575 */ 1576 static int 1577 sosetopt1(struct socket *so, const struct sockopt *sopt) 1578 { 1579 int error, optval; 1580 struct linger l; 1581 struct timeval tv; 1582 1583 switch (sopt->sopt_name) { 1584 1585 case SO_ACCEPTFILTER: 1586 error = accept_filt_setopt(so, sopt); 1587 KASSERT(solocked(so)); 1588 break; 1589 1590 case SO_LINGER: 1591 error = sockopt_get(sopt, &l, sizeof(l)); 1592 solock(so); 1593 if (error) 1594 break; 1595 if (l.l_linger < 0 || l.l_linger > USHRT_MAX || 1596 l.l_linger > (INT_MAX / hz)) { 1597 error = EDOM; 1598 break; 1599 } 1600 so->so_linger = l.l_linger; 1601 if (l.l_onoff) 1602 so->so_options |= SO_LINGER; 1603 else 1604 so->so_options &= ~SO_LINGER; 1605 break; 1606 1607 case SO_DEBUG: 1608 case SO_KEEPALIVE: 1609 case SO_DONTROUTE: 1610 case SO_USELOOPBACK: 1611 case SO_BROADCAST: 1612 case SO_REUSEADDR: 1613 case SO_REUSEPORT: 1614 case SO_OOBINLINE: 1615 case SO_TIMESTAMP: 1616 error = sockopt_getint(sopt, &optval); 1617 solock(so); 1618 if (error) 1619 break; 1620 if (optval) 1621 so->so_options |= sopt->sopt_name; 1622 else 1623 so->so_options &= ~sopt->sopt_name; 1624 break; 1625 1626 case SO_SNDBUF: 1627 case SO_RCVBUF: 1628 case SO_SNDLOWAT: 1629 case SO_RCVLOWAT: 1630 error = sockopt_getint(sopt, &optval); 1631 solock(so); 1632 if (error) 1633 break; 1634 1635 /* 1636 * Values < 1 make no sense for any of these 1637 * options, so disallow them. 1638 */ 1639 if (optval < 1) { 1640 error = EINVAL; 1641 break; 1642 } 1643 1644 switch (sopt->sopt_name) { 1645 case SO_SNDBUF: 1646 if (sbreserve(&so->so_snd, (u_long)optval, so) == 0) { 1647 error = ENOBUFS; 1648 break; 1649 } 1650 so->so_snd.sb_flags &= ~SB_AUTOSIZE; 1651 break; 1652 1653 case SO_RCVBUF: 1654 if (sbreserve(&so->so_rcv, (u_long)optval, so) == 0) { 1655 error = ENOBUFS; 1656 break; 1657 } 1658 so->so_rcv.sb_flags &= ~SB_AUTOSIZE; 1659 break; 1660 1661 /* 1662 * Make sure the low-water is never greater than 1663 * the high-water. 1664 */ 1665 case SO_SNDLOWAT: 1666 if (optval > so->so_snd.sb_hiwat) 1667 optval = so->so_snd.sb_hiwat; 1668 1669 so->so_snd.sb_lowat = optval; 1670 break; 1671 1672 case SO_RCVLOWAT: 1673 if (optval > so->so_rcv.sb_hiwat) 1674 optval = so->so_rcv.sb_hiwat; 1675 1676 so->so_rcv.sb_lowat = optval; 1677 break; 1678 } 1679 break; 1680 1681 case SO_SNDTIMEO: 1682 case SO_RCVTIMEO: 1683 error = sockopt_get(sopt, &tv, sizeof(tv)); 1684 solock(so); 1685 if (error) 1686 break; 1687 1688 if (tv.tv_sec > (INT_MAX - tv.tv_usec / tick) / hz) { 1689 error = EDOM; 1690 break; 1691 } 1692 1693 optval = tv.tv_sec * hz + tv.tv_usec / tick; 1694 if (optval == 0 && tv.tv_usec != 0) 1695 optval = 1; 1696 1697 switch (sopt->sopt_name) { 1698 case SO_SNDTIMEO: 1699 so->so_snd.sb_timeo = optval; 1700 break; 1701 case SO_RCVTIMEO: 1702 so->so_rcv.sb_timeo = optval; 1703 break; 1704 } 1705 break; 1706 1707 default: 1708 solock(so); 1709 error = ENOPROTOOPT; 1710 break; 1711 } 1712 KASSERT(solocked(so)); 1713 return error; 1714 } 1715 1716 int 1717 sosetopt(struct socket *so, struct sockopt *sopt) 1718 { 1719 int error, prerr; 1720 1721 if (sopt->sopt_level == SOL_SOCKET) { 1722 error = sosetopt1(so, sopt); 1723 KASSERT(solocked(so)); 1724 } else { 1725 error = ENOPROTOOPT; 1726 solock(so); 1727 } 1728 1729 if ((error == 0 || error == ENOPROTOOPT) && 1730 so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) { 1731 /* give the protocol stack a shot */ 1732 prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, sopt); 1733 if (prerr == 0) 1734 error = 0; 1735 else if (prerr != ENOPROTOOPT) 1736 error = prerr; 1737 } 1738 sounlock(so); 1739 return error; 1740 } 1741 1742 /* 1743 * so_setsockopt() is a wrapper providing a sockopt structure for sosetopt() 1744 */ 1745 int 1746 so_setsockopt(struct lwp *l, struct socket *so, int level, int name, 1747 const void *val, size_t valsize) 1748 { 1749 struct sockopt sopt; 1750 int error; 1751 1752 KASSERT(valsize == 0 || val != NULL); 1753 1754 sockopt_init(&sopt, level, name, valsize); 1755 sockopt_set(&sopt, val, valsize); 1756 1757 error = sosetopt(so, &sopt); 1758 1759 sockopt_destroy(&sopt); 1760 1761 return error; 1762 } 1763 1764 /* 1765 * internal get SOL_SOCKET options 1766 */ 1767 static int 1768 sogetopt1(struct socket *so, struct sockopt *sopt) 1769 { 1770 int error, optval; 1771 struct linger l; 1772 struct timeval tv; 1773 1774 switch (sopt->sopt_name) { 1775 1776 case SO_ACCEPTFILTER: 1777 error = accept_filt_getopt(so, sopt); 1778 break; 1779 1780 case SO_LINGER: 1781 l.l_onoff = (so->so_options & SO_LINGER) ? 1 : 0; 1782 l.l_linger = so->so_linger; 1783 1784 error = sockopt_set(sopt, &l, sizeof(l)); 1785 break; 1786 1787 case SO_USELOOPBACK: 1788 case SO_DONTROUTE: 1789 case SO_DEBUG: 1790 case SO_KEEPALIVE: 1791 case SO_REUSEADDR: 1792 case SO_REUSEPORT: 1793 case SO_BROADCAST: 1794 case SO_OOBINLINE: 1795 case SO_TIMESTAMP: 1796 error = sockopt_setint(sopt, 1797 (so->so_options & sopt->sopt_name) ? 1 : 0); 1798 break; 1799 1800 case SO_TYPE: 1801 error = sockopt_setint(sopt, so->so_type); 1802 break; 1803 1804 case SO_ERROR: 1805 error = sockopt_setint(sopt, so->so_error); 1806 so->so_error = 0; 1807 break; 1808 1809 case SO_SNDBUF: 1810 error = sockopt_setint(sopt, so->so_snd.sb_hiwat); 1811 break; 1812 1813 case SO_RCVBUF: 1814 error = sockopt_setint(sopt, so->so_rcv.sb_hiwat); 1815 break; 1816 1817 case SO_SNDLOWAT: 1818 error = sockopt_setint(sopt, so->so_snd.sb_lowat); 1819 break; 1820 1821 case SO_RCVLOWAT: 1822 error = sockopt_setint(sopt, so->so_rcv.sb_lowat); 1823 break; 1824 1825 case SO_SNDTIMEO: 1826 case SO_RCVTIMEO: 1827 optval = (sopt->sopt_name == SO_SNDTIMEO ? 1828 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 1829 1830 tv.tv_sec = optval / hz; 1831 tv.tv_usec = (optval % hz) * tick; 1832 1833 error = sockopt_set(sopt, &tv, sizeof(tv)); 1834 break; 1835 1836 case SO_OVERFLOWED: 1837 error = sockopt_setint(sopt, so->so_rcv.sb_overflowed); 1838 break; 1839 1840 default: 1841 error = ENOPROTOOPT; 1842 break; 1843 } 1844 1845 return (error); 1846 } 1847 1848 int 1849 sogetopt(struct socket *so, struct sockopt *sopt) 1850 { 1851 int error; 1852 1853 solock(so); 1854 if (sopt->sopt_level != SOL_SOCKET) { 1855 if (so->so_proto && so->so_proto->pr_ctloutput) { 1856 error = ((*so->so_proto->pr_ctloutput) 1857 (PRCO_GETOPT, so, sopt)); 1858 } else 1859 error = (ENOPROTOOPT); 1860 } else { 1861 error = sogetopt1(so, sopt); 1862 } 1863 sounlock(so); 1864 return (error); 1865 } 1866 1867 /* 1868 * alloc sockopt data buffer buffer 1869 * - will be released at destroy 1870 */ 1871 static int 1872 sockopt_alloc(struct sockopt *sopt, size_t len, km_flag_t kmflag) 1873 { 1874 1875 KASSERT(sopt->sopt_size == 0); 1876 1877 if (len > sizeof(sopt->sopt_buf)) { 1878 sopt->sopt_data = kmem_zalloc(len, kmflag); 1879 if (sopt->sopt_data == NULL) 1880 return ENOMEM; 1881 } else 1882 sopt->sopt_data = sopt->sopt_buf; 1883 1884 sopt->sopt_size = len; 1885 return 0; 1886 } 1887 1888 /* 1889 * initialise sockopt storage 1890 * - MAY sleep during allocation 1891 */ 1892 void 1893 sockopt_init(struct sockopt *sopt, int level, int name, size_t size) 1894 { 1895 1896 memset(sopt, 0, sizeof(*sopt)); 1897 1898 sopt->sopt_level = level; 1899 sopt->sopt_name = name; 1900 (void)sockopt_alloc(sopt, size, KM_SLEEP); 1901 } 1902 1903 /* 1904 * destroy sockopt storage 1905 * - will release any held memory references 1906 */ 1907 void 1908 sockopt_destroy(struct sockopt *sopt) 1909 { 1910 1911 if (sopt->sopt_data != sopt->sopt_buf) 1912 kmem_free(sopt->sopt_data, sopt->sopt_size); 1913 1914 memset(sopt, 0, sizeof(*sopt)); 1915 } 1916 1917 /* 1918 * set sockopt value 1919 * - value is copied into sockopt 1920 * - memory is allocated when necessary, will not sleep 1921 */ 1922 int 1923 sockopt_set(struct sockopt *sopt, const void *buf, size_t len) 1924 { 1925 int error; 1926 1927 if (sopt->sopt_size == 0) { 1928 error = sockopt_alloc(sopt, len, KM_NOSLEEP); 1929 if (error) 1930 return error; 1931 } 1932 1933 KASSERT(sopt->sopt_size == len); 1934 memcpy(sopt->sopt_data, buf, len); 1935 return 0; 1936 } 1937 1938 /* 1939 * common case of set sockopt integer value 1940 */ 1941 int 1942 sockopt_setint(struct sockopt *sopt, int val) 1943 { 1944 1945 return sockopt_set(sopt, &val, sizeof(int)); 1946 } 1947 1948 /* 1949 * get sockopt value 1950 * - correct size must be given 1951 */ 1952 int 1953 sockopt_get(const struct sockopt *sopt, void *buf, size_t len) 1954 { 1955 1956 if (sopt->sopt_size != len) 1957 return EINVAL; 1958 1959 memcpy(buf, sopt->sopt_data, len); 1960 return 0; 1961 } 1962 1963 /* 1964 * common case of get sockopt integer value 1965 */ 1966 int 1967 sockopt_getint(const struct sockopt *sopt, int *valp) 1968 { 1969 1970 return sockopt_get(sopt, valp, sizeof(int)); 1971 } 1972 1973 /* 1974 * set sockopt value from mbuf 1975 * - ONLY for legacy code 1976 * - mbuf is released by sockopt 1977 * - will not sleep 1978 */ 1979 int 1980 sockopt_setmbuf(struct sockopt *sopt, struct mbuf *m) 1981 { 1982 size_t len; 1983 int error; 1984 1985 len = m_length(m); 1986 1987 if (sopt->sopt_size == 0) { 1988 error = sockopt_alloc(sopt, len, KM_NOSLEEP); 1989 if (error) 1990 return error; 1991 } 1992 1993 KASSERT(sopt->sopt_size == len); 1994 m_copydata(m, 0, len, sopt->sopt_data); 1995 m_freem(m); 1996 1997 return 0; 1998 } 1999 2000 /* 2001 * get sockopt value into mbuf 2002 * - ONLY for legacy code 2003 * - mbuf to be released by the caller 2004 * - will not sleep 2005 */ 2006 struct mbuf * 2007 sockopt_getmbuf(const struct sockopt *sopt) 2008 { 2009 struct mbuf *m; 2010 2011 if (sopt->sopt_size > MCLBYTES) 2012 return NULL; 2013 2014 m = m_get(M_DONTWAIT, MT_SOOPTS); 2015 if (m == NULL) 2016 return NULL; 2017 2018 if (sopt->sopt_size > MLEN) { 2019 MCLGET(m, M_DONTWAIT); 2020 if ((m->m_flags & M_EXT) == 0) { 2021 m_free(m); 2022 return NULL; 2023 } 2024 } 2025 2026 memcpy(mtod(m, void *), sopt->sopt_data, sopt->sopt_size); 2027 m->m_len = sopt->sopt_size; 2028 2029 return m; 2030 } 2031 2032 void 2033 sohasoutofband(struct socket *so) 2034 { 2035 2036 fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so); 2037 selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, 0); 2038 } 2039 2040 static void 2041 filt_sordetach(struct knote *kn) 2042 { 2043 struct socket *so; 2044 2045 so = ((file_t *)kn->kn_obj)->f_data; 2046 solock(so); 2047 SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext); 2048 if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist)) 2049 so->so_rcv.sb_flags &= ~SB_KNOTE; 2050 sounlock(so); 2051 } 2052 2053 /*ARGSUSED*/ 2054 static int 2055 filt_soread(struct knote *kn, long hint) 2056 { 2057 struct socket *so; 2058 int rv; 2059 2060 so = ((file_t *)kn->kn_obj)->f_data; 2061 if (hint != NOTE_SUBMIT) 2062 solock(so); 2063 kn->kn_data = so->so_rcv.sb_cc; 2064 if (so->so_state & SS_CANTRCVMORE) { 2065 kn->kn_flags |= EV_EOF; 2066 kn->kn_fflags = so->so_error; 2067 rv = 1; 2068 } else if (so->so_error) /* temporary udp error */ 2069 rv = 1; 2070 else if (kn->kn_sfflags & NOTE_LOWAT) 2071 rv = (kn->kn_data >= kn->kn_sdata); 2072 else 2073 rv = (kn->kn_data >= so->so_rcv.sb_lowat); 2074 if (hint != NOTE_SUBMIT) 2075 sounlock(so); 2076 return rv; 2077 } 2078 2079 static void 2080 filt_sowdetach(struct knote *kn) 2081 { 2082 struct socket *so; 2083 2084 so = ((file_t *)kn->kn_obj)->f_data; 2085 solock(so); 2086 SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext); 2087 if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist)) 2088 so->so_snd.sb_flags &= ~SB_KNOTE; 2089 sounlock(so); 2090 } 2091 2092 /*ARGSUSED*/ 2093 static int 2094 filt_sowrite(struct knote *kn, long hint) 2095 { 2096 struct socket *so; 2097 int rv; 2098 2099 so = ((file_t *)kn->kn_obj)->f_data; 2100 if (hint != NOTE_SUBMIT) 2101 solock(so); 2102 kn->kn_data = sbspace(&so->so_snd); 2103 if (so->so_state & SS_CANTSENDMORE) { 2104 kn->kn_flags |= EV_EOF; 2105 kn->kn_fflags = so->so_error; 2106 rv = 1; 2107 } else if (so->so_error) /* temporary udp error */ 2108 rv = 1; 2109 else if (((so->so_state & SS_ISCONNECTED) == 0) && 2110 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 2111 rv = 0; 2112 else if (kn->kn_sfflags & NOTE_LOWAT) 2113 rv = (kn->kn_data >= kn->kn_sdata); 2114 else 2115 rv = (kn->kn_data >= so->so_snd.sb_lowat); 2116 if (hint != NOTE_SUBMIT) 2117 sounlock(so); 2118 return rv; 2119 } 2120 2121 /*ARGSUSED*/ 2122 static int 2123 filt_solisten(struct knote *kn, long hint) 2124 { 2125 struct socket *so; 2126 int rv; 2127 2128 so = ((file_t *)kn->kn_obj)->f_data; 2129 2130 /* 2131 * Set kn_data to number of incoming connections, not 2132 * counting partial (incomplete) connections. 2133 */ 2134 if (hint != NOTE_SUBMIT) 2135 solock(so); 2136 kn->kn_data = so->so_qlen; 2137 rv = (kn->kn_data > 0); 2138 if (hint != NOTE_SUBMIT) 2139 sounlock(so); 2140 return rv; 2141 } 2142 2143 static const struct filterops solisten_filtops = 2144 { 1, NULL, filt_sordetach, filt_solisten }; 2145 static const struct filterops soread_filtops = 2146 { 1, NULL, filt_sordetach, filt_soread }; 2147 static const struct filterops sowrite_filtops = 2148 { 1, NULL, filt_sowdetach, filt_sowrite }; 2149 2150 int 2151 soo_kqfilter(struct file *fp, struct knote *kn) 2152 { 2153 struct socket *so; 2154 struct sockbuf *sb; 2155 2156 so = ((file_t *)kn->kn_obj)->f_data; 2157 solock(so); 2158 switch (kn->kn_filter) { 2159 case EVFILT_READ: 2160 if (so->so_options & SO_ACCEPTCONN) 2161 kn->kn_fop = &solisten_filtops; 2162 else 2163 kn->kn_fop = &soread_filtops; 2164 sb = &so->so_rcv; 2165 break; 2166 case EVFILT_WRITE: 2167 kn->kn_fop = &sowrite_filtops; 2168 sb = &so->so_snd; 2169 break; 2170 default: 2171 sounlock(so); 2172 return (EINVAL); 2173 } 2174 SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext); 2175 sb->sb_flags |= SB_KNOTE; 2176 sounlock(so); 2177 return (0); 2178 } 2179 2180 static int 2181 sodopoll(struct socket *so, int events) 2182 { 2183 int revents; 2184 2185 revents = 0; 2186 2187 if (events & (POLLIN | POLLRDNORM)) 2188 if (soreadable(so)) 2189 revents |= events & (POLLIN | POLLRDNORM); 2190 2191 if (events & (POLLOUT | POLLWRNORM)) 2192 if (sowritable(so)) 2193 revents |= events & (POLLOUT | POLLWRNORM); 2194 2195 if (events & (POLLPRI | POLLRDBAND)) 2196 if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) 2197 revents |= events & (POLLPRI | POLLRDBAND); 2198 2199 return revents; 2200 } 2201 2202 int 2203 sopoll(struct socket *so, int events) 2204 { 2205 int revents = 0; 2206 2207 #ifndef DIAGNOSTIC 2208 /* 2209 * Do a quick, unlocked check in expectation that the socket 2210 * will be ready for I/O. Don't do this check if DIAGNOSTIC, 2211 * as the solocked() assertions will fail. 2212 */ 2213 if ((revents = sodopoll(so, events)) != 0) 2214 return revents; 2215 #endif 2216 2217 solock(so); 2218 if ((revents = sodopoll(so, events)) == 0) { 2219 if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) { 2220 selrecord(curlwp, &so->so_rcv.sb_sel); 2221 so->so_rcv.sb_flags |= SB_NOTIFY; 2222 } 2223 2224 if (events & (POLLOUT | POLLWRNORM)) { 2225 selrecord(curlwp, &so->so_snd.sb_sel); 2226 so->so_snd.sb_flags |= SB_NOTIFY; 2227 } 2228 } 2229 sounlock(so); 2230 2231 return revents; 2232 } 2233 2234 2235 #include <sys/sysctl.h> 2236 2237 static int sysctl_kern_somaxkva(SYSCTLFN_PROTO); 2238 2239 /* 2240 * sysctl helper routine for kern.somaxkva. ensures that the given 2241 * value is not too small. 2242 * (XXX should we maybe make sure it's not too large as well?) 2243 */ 2244 static int 2245 sysctl_kern_somaxkva(SYSCTLFN_ARGS) 2246 { 2247 int error, new_somaxkva; 2248 struct sysctlnode node; 2249 2250 new_somaxkva = somaxkva; 2251 node = *rnode; 2252 node.sysctl_data = &new_somaxkva; 2253 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2254 if (error || newp == NULL) 2255 return (error); 2256 2257 if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */ 2258 return (EINVAL); 2259 2260 mutex_enter(&so_pendfree_lock); 2261 somaxkva = new_somaxkva; 2262 cv_broadcast(&socurkva_cv); 2263 mutex_exit(&so_pendfree_lock); 2264 2265 return (error); 2266 } 2267 2268 static void 2269 sysctl_kern_somaxkva_setup() 2270 { 2271 2272 KASSERT(socket_sysctllog == NULL); 2273 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2274 CTLFLAG_PERMANENT, 2275 CTLTYPE_NODE, "kern", NULL, 2276 NULL, 0, NULL, 0, 2277 CTL_KERN, CTL_EOL); 2278 2279 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2280 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2281 CTLTYPE_INT, "somaxkva", 2282 SYSCTL_DESCR("Maximum amount of kernel memory to be " 2283 "used for socket buffers"), 2284 sysctl_kern_somaxkva, 0, NULL, 0, 2285 CTL_KERN, KERN_SOMAXKVA, CTL_EOL); 2286 } 2287
Indexes created Sat Sep 20 22:09:52 GMT 2025