uipc_socket.c revision 1.297
1 /* $NetBSD: uipc_socket.c,v 1.297 2021/09/29 02:47:22 thorpej Exp $ */ 2 3 /* 4 * Copyright (c) 2002, 2007, 2008, 2009 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, and by Andrew Doran. 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 /* 66 * Socket operation routines. 67 * 68 * These routines are called by the routines in sys_socket.c or from a 69 * system process, and implement the semantics of socket operations by 70 * switching out to the protocol specific routines. 71 */ 72 73 #include <sys/cdefs.h> 74 __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.297 2021/09/29 02:47:22 thorpej Exp $"); 75 76 #ifdef _KERNEL_OPT 77 #include "opt_compat_netbsd.h" 78 #include "opt_sock_counters.h" 79 #include "opt_sosend_loan.h" 80 #include "opt_mbuftrace.h" 81 #include "opt_somaxkva.h" 82 #include "opt_multiprocessor.h" /* XXX */ 83 #include "opt_sctp.h" 84 #endif 85 86 #include <sys/param.h> 87 #include <sys/systm.h> 88 #include <sys/proc.h> 89 #include <sys/file.h> 90 #include <sys/filedesc.h> 91 #include <sys/kmem.h> 92 #include <sys/mbuf.h> 93 #include <sys/domain.h> 94 #include <sys/kernel.h> 95 #include <sys/protosw.h> 96 #include <sys/socket.h> 97 #include <sys/socketvar.h> 98 #include <sys/signalvar.h> 99 #include <sys/resourcevar.h> 100 #include <sys/uidinfo.h> 101 #include <sys/event.h> 102 #include <sys/poll.h> 103 #include <sys/kauth.h> 104 #include <sys/mutex.h> 105 #include <sys/condvar.h> 106 #include <sys/kthread.h> 107 #include <sys/compat_stub.h> 108 109 #include <compat/sys/time.h> 110 #include <compat/sys/socket.h> 111 112 #include <uvm/uvm_extern.h> 113 #include <uvm/uvm_loan.h> 114 #include <uvm/uvm_page.h> 115 116 #ifdef SCTP 117 #include <netinet/sctp_route.h> 118 #endif 119 120 MALLOC_DEFINE(M_SONAME, "soname", "socket name"); 121 122 extern const struct fileops socketops; 123 124 static int sooptions; 125 extern int somaxconn; /* patchable (XXX sysctl) */ 126 int somaxconn = SOMAXCONN; 127 kmutex_t *softnet_lock; 128 129 #ifdef SOSEND_COUNTERS 130 #include <sys/device.h> 131 132 static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 133 NULL, "sosend", "loan big"); 134 static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 135 NULL, "sosend", "copy big"); 136 static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 137 NULL, "sosend", "copy small"); 138 static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 139 NULL, "sosend", "kva limit"); 140 141 #define SOSEND_COUNTER_INCR(ev) (ev)->ev_count++ 142 143 EVCNT_ATTACH_STATIC(sosend_loan_big); 144 EVCNT_ATTACH_STATIC(sosend_copy_big); 145 EVCNT_ATTACH_STATIC(sosend_copy_small); 146 EVCNT_ATTACH_STATIC(sosend_kvalimit); 147 #else 148 149 #define SOSEND_COUNTER_INCR(ev) /* nothing */ 150 151 #endif /* SOSEND_COUNTERS */ 152 153 #if defined(SOSEND_NO_LOAN) || defined(MULTIPROCESSOR) 154 int sock_loan_thresh = -1; 155 #else 156 int sock_loan_thresh = 4096; 157 #endif 158 159 static kmutex_t so_pendfree_lock; 160 static struct mbuf *so_pendfree = NULL; 161 162 #ifndef SOMAXKVA 163 #define SOMAXKVA (16 * 1024 * 1024) 164 #endif 165 int somaxkva = SOMAXKVA; 166 static int socurkva; 167 static kcondvar_t socurkva_cv; 168 169 #ifndef SOFIXEDBUF 170 #define SOFIXEDBUF true 171 #endif 172 bool sofixedbuf = SOFIXEDBUF; 173 174 static kauth_listener_t socket_listener; 175 176 #define SOCK_LOAN_CHUNK 65536 177 178 static void sopendfree_thread(void *); 179 static kcondvar_t pendfree_thread_cv; 180 static lwp_t *sopendfree_lwp; 181 182 static void sysctl_kern_socket_setup(void); 183 static struct sysctllog *socket_sysctllog; 184 185 static vsize_t 186 sokvareserve(struct socket *so, vsize_t len) 187 { 188 int error; 189 190 mutex_enter(&so_pendfree_lock); 191 while (socurkva + len > somaxkva) { 192 SOSEND_COUNTER_INCR(&sosend_kvalimit); 193 error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock); 194 if (error) { 195 len = 0; 196 break; 197 } 198 } 199 socurkva += len; 200 mutex_exit(&so_pendfree_lock); 201 return len; 202 } 203 204 static void 205 sokvaunreserve(vsize_t len) 206 { 207 208 mutex_enter(&so_pendfree_lock); 209 socurkva -= len; 210 cv_broadcast(&socurkva_cv); 211 mutex_exit(&so_pendfree_lock); 212 } 213 214 /* 215 * sokvaalloc: allocate kva for loan. 216 */ 217 vaddr_t 218 sokvaalloc(vaddr_t sva, vsize_t len, struct socket *so) 219 { 220 vaddr_t lva; 221 222 if (sokvareserve(so, len) == 0) 223 return 0; 224 225 lva = uvm_km_alloc(kernel_map, len, atop(sva) & uvmexp.colormask, 226 UVM_KMF_COLORMATCH | UVM_KMF_VAONLY | UVM_KMF_WAITVA); 227 if (lva == 0) { 228 sokvaunreserve(len); 229 return 0; 230 } 231 232 return lva; 233 } 234 235 /* 236 * sokvafree: free kva for loan. 237 */ 238 void 239 sokvafree(vaddr_t sva, vsize_t len) 240 { 241 242 uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY); 243 sokvaunreserve(len); 244 } 245 246 static void 247 sodoloanfree(struct vm_page **pgs, void *buf, size_t size) 248 { 249 vaddr_t sva, eva; 250 vsize_t len; 251 int npgs; 252 253 KASSERT(pgs != NULL); 254 255 eva = round_page((vaddr_t) buf + size); 256 sva = trunc_page((vaddr_t) buf); 257 len = eva - sva; 258 npgs = len >> PAGE_SHIFT; 259 260 pmap_kremove(sva, len); 261 pmap_update(pmap_kernel()); 262 uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE); 263 sokvafree(sva, len); 264 } 265 266 /* 267 * sopendfree_thread: free mbufs on "pendfree" list. Unlock and relock 268 * so_pendfree_lock when freeing mbufs. 269 */ 270 static void 271 sopendfree_thread(void *v) 272 { 273 struct mbuf *m, *next; 274 size_t rv; 275 276 mutex_enter(&so_pendfree_lock); 277 278 for (;;) { 279 rv = 0; 280 while (so_pendfree != NULL) { 281 m = so_pendfree; 282 so_pendfree = NULL; 283 mutex_exit(&so_pendfree_lock); 284 285 for (; m != NULL; m = next) { 286 next = m->m_next; 287 KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 288 0); 289 KASSERT(m->m_ext.ext_refcnt == 0); 290 291 rv += m->m_ext.ext_size; 292 sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf, 293 m->m_ext.ext_size); 294 pool_cache_put(mb_cache, m); 295 } 296 297 mutex_enter(&so_pendfree_lock); 298 } 299 if (rv) 300 cv_broadcast(&socurkva_cv); 301 cv_wait(&pendfree_thread_cv, &so_pendfree_lock); 302 } 303 panic("sopendfree_thread"); 304 /* NOTREACHED */ 305 } 306 307 void 308 soloanfree(struct mbuf *m, void *buf, size_t size, void *arg) 309 { 310 311 KASSERT(m != NULL); 312 313 /* 314 * postpone freeing mbuf. 315 * 316 * we can't do it in interrupt context 317 * because we need to put kva back to kernel_map. 318 */ 319 320 mutex_enter(&so_pendfree_lock); 321 m->m_next = so_pendfree; 322 so_pendfree = m; 323 cv_signal(&pendfree_thread_cv); 324 mutex_exit(&so_pendfree_lock); 325 } 326 327 static long 328 sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space) 329 { 330 struct iovec *iov = uio->uio_iov; 331 vaddr_t sva, eva; 332 vsize_t len; 333 vaddr_t lva; 334 int npgs, error; 335 vaddr_t va; 336 int i; 337 338 if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace)) 339 return 0; 340 341 if (iov->iov_len < (size_t) space) 342 space = iov->iov_len; 343 if (space > SOCK_LOAN_CHUNK) 344 space = SOCK_LOAN_CHUNK; 345 346 eva = round_page((vaddr_t) iov->iov_base + space); 347 sva = trunc_page((vaddr_t) iov->iov_base); 348 len = eva - sva; 349 npgs = len >> PAGE_SHIFT; 350 351 KASSERT(npgs <= M_EXT_MAXPAGES); 352 353 lva = sokvaalloc(sva, len, so); 354 if (lva == 0) 355 return 0; 356 357 error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len, 358 m->m_ext.ext_pgs, UVM_LOAN_TOPAGE); 359 if (error) { 360 sokvafree(lva, len); 361 return 0; 362 } 363 364 for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE) 365 pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]), 366 VM_PROT_READ, 0); 367 pmap_update(pmap_kernel()); 368 369 lva += (vaddr_t) iov->iov_base & PAGE_MASK; 370 371 MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so); 372 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 373 374 uio->uio_resid -= space; 375 /* uio_offset not updated, not set/used for write(2) */ 376 uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space; 377 uio->uio_iov->iov_len -= space; 378 if (uio->uio_iov->iov_len == 0) { 379 uio->uio_iov++; 380 uio->uio_iovcnt--; 381 } 382 383 return space; 384 } 385 386 static int 387 socket_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 388 void *arg0, void *arg1, void *arg2, void *arg3) 389 { 390 int result; 391 enum kauth_network_req req; 392 393 result = KAUTH_RESULT_DEFER; 394 req = (enum kauth_network_req)(uintptr_t)arg0; 395 396 if ((action != KAUTH_NETWORK_SOCKET) && 397 (action != KAUTH_NETWORK_BIND)) 398 return result; 399 400 switch (req) { 401 case KAUTH_REQ_NETWORK_BIND_PORT: 402 result = KAUTH_RESULT_ALLOW; 403 break; 404 405 case KAUTH_REQ_NETWORK_SOCKET_DROP: { 406 /* Normal users can only drop their own connections. */ 407 struct socket *so = (struct socket *)arg1; 408 409 if (so->so_cred && proc_uidmatch(cred, so->so_cred) == 0) 410 result = KAUTH_RESULT_ALLOW; 411 412 break; 413 } 414 415 case KAUTH_REQ_NETWORK_SOCKET_OPEN: 416 /* We allow "raw" routing/bluetooth sockets to anyone. */ 417 switch ((u_long)arg1) { 418 case PF_ROUTE: 419 case PF_OROUTE: 420 case PF_BLUETOOTH: 421 case PF_CAN: 422 result = KAUTH_RESULT_ALLOW; 423 break; 424 default: 425 /* Privileged, let secmodel handle this. */ 426 if ((u_long)arg2 == SOCK_RAW) 427 break; 428 result = KAUTH_RESULT_ALLOW; 429 break; 430 } 431 break; 432 433 case KAUTH_REQ_NETWORK_SOCKET_CANSEE: 434 result = KAUTH_RESULT_ALLOW; 435 436 break; 437 438 default: 439 break; 440 } 441 442 return result; 443 } 444 445 void 446 soinit(void) 447 { 448 449 sysctl_kern_socket_setup(); 450 451 #ifdef SCTP 452 /* Update the SCTP function hooks if necessary*/ 453 454 vec_sctp_add_ip_address = sctp_add_ip_address; 455 vec_sctp_delete_ip_address = sctp_delete_ip_address; 456 #endif 457 458 mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM); 459 softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 460 cv_init(&socurkva_cv, "sokva"); 461 cv_init(&pendfree_thread_cv, "sopendfr"); 462 soinit2(); 463 464 /* Set the initial adjusted socket buffer size. */ 465 if (sb_max_set(sb_max)) 466 panic("bad initial sb_max value: %lu", sb_max); 467 468 socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 469 socket_listener_cb, NULL); 470 } 471 472 void 473 soinit1(void) 474 { 475 int error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL, 476 sopendfree_thread, NULL, &sopendfree_lwp, "sopendfree"); 477 if (error) 478 panic("soinit1 %d", error); 479 } 480 481 /* 482 * socreate: create a new socket of the specified type and the protocol. 483 * 484 * => Caller may specify another socket for lock sharing (must not be held). 485 * => Returns the new socket without lock held. 486 */ 487 int 488 socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l, 489 struct socket *lockso) 490 { 491 const struct protosw *prp; 492 struct socket *so; 493 uid_t uid; 494 int error; 495 kmutex_t *lock; 496 497 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 498 KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type), 499 KAUTH_ARG(proto)); 500 if (error != 0) 501 return error; 502 503 if (proto) 504 prp = pffindproto(dom, proto, type); 505 else 506 prp = pffindtype(dom, type); 507 if (prp == NULL) { 508 /* no support for domain */ 509 if (pffinddomain(dom) == 0) 510 return EAFNOSUPPORT; 511 /* no support for socket type */ 512 if (proto == 0 && type != 0) 513 return EPROTOTYPE; 514 return EPROTONOSUPPORT; 515 } 516 if (prp->pr_usrreqs == NULL) 517 return EPROTONOSUPPORT; 518 if (prp->pr_type != type) 519 return EPROTOTYPE; 520 521 so = soget(true); 522 so->so_type = type; 523 so->so_proto = prp; 524 so->so_send = sosend; 525 so->so_receive = soreceive; 526 so->so_options = sooptions; 527 #ifdef MBUFTRACE 528 so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner; 529 so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner; 530 so->so_mowner = &prp->pr_domain->dom_mowner; 531 #endif 532 uid = kauth_cred_geteuid(l->l_cred); 533 so->so_uidinfo = uid_find(uid); 534 so->so_egid = kauth_cred_getegid(l->l_cred); 535 so->so_cpid = l->l_proc->p_pid; 536 537 /* 538 * Lock assigned and taken during PCB attach, unless we share 539 * the lock with another socket, e.g. socketpair(2) case. 540 */ 541 if (lockso) { 542 lock = lockso->so_lock; 543 so->so_lock = lock; 544 mutex_obj_hold(lock); 545 mutex_enter(lock); 546 } 547 548 /* Attach the PCB (returns with the socket lock held). */ 549 error = (*prp->pr_usrreqs->pr_attach)(so, proto); 550 KASSERT(solocked(so)); 551 552 if (error) { 553 KASSERT(so->so_pcb == NULL); 554 so->so_state |= SS_NOFDREF; 555 sofree(so); 556 return error; 557 } 558 so->so_cred = kauth_cred_dup(l->l_cred); 559 sounlock(so); 560 561 *aso = so; 562 return 0; 563 } 564 565 /* 566 * fsocreate: create a socket and a file descriptor associated with it. 567 * 568 * => On success, write file descriptor to fdout and return zero. 569 * => On failure, return non-zero; *fdout will be undefined. 570 */ 571 int 572 fsocreate(int domain, struct socket **sop, int type, int proto, int *fdout) 573 { 574 lwp_t *l = curlwp; 575 int error, fd, flags; 576 struct socket *so; 577 struct file *fp; 578 579 if ((error = fd_allocfile(&fp, &fd)) != 0) { 580 return error; 581 } 582 flags = type & SOCK_FLAGS_MASK; 583 fd_set_exclose(l, fd, (flags & SOCK_CLOEXEC) != 0); 584 fp->f_flag = FREAD|FWRITE|((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)| 585 ((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0); 586 fp->f_type = DTYPE_SOCKET; 587 fp->f_ops = &socketops; 588 589 type &= ~SOCK_FLAGS_MASK; 590 error = socreate(domain, &so, type, proto, l, NULL); 591 if (error) { 592 fd_abort(curproc, fp, fd); 593 return error; 594 } 595 if (flags & SOCK_NONBLOCK) { 596 so->so_state |= SS_NBIO; 597 } 598 fp->f_socket = so; 599 fd_affix(curproc, fp, fd); 600 601 if (sop != NULL) { 602 *sop = so; 603 } 604 *fdout = fd; 605 return error; 606 } 607 608 int 609 sofamily(const struct socket *so) 610 { 611 const struct protosw *pr; 612 const struct domain *dom; 613 614 if ((pr = so->so_proto) == NULL) 615 return AF_UNSPEC; 616 if ((dom = pr->pr_domain) == NULL) 617 return AF_UNSPEC; 618 return dom->dom_family; 619 } 620 621 int 622 sobind(struct socket *so, struct sockaddr *nam, struct lwp *l) 623 { 624 int error; 625 626 solock(so); 627 if (nam->sa_family != so->so_proto->pr_domain->dom_family) { 628 sounlock(so); 629 return EAFNOSUPPORT; 630 } 631 error = (*so->so_proto->pr_usrreqs->pr_bind)(so, nam, l); 632 sounlock(so); 633 return error; 634 } 635 636 int 637 solisten(struct socket *so, int backlog, struct lwp *l) 638 { 639 int error; 640 short oldopt, oldqlimit; 641 642 solock(so); 643 if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | 644 SS_ISDISCONNECTING)) != 0) { 645 sounlock(so); 646 return EINVAL; 647 } 648 oldopt = so->so_options; 649 oldqlimit = so->so_qlimit; 650 if (TAILQ_EMPTY(&so->so_q)) 651 so->so_options |= SO_ACCEPTCONN; 652 if (backlog < 0) 653 backlog = 0; 654 so->so_qlimit = uimin(backlog, somaxconn); 655 656 error = (*so->so_proto->pr_usrreqs->pr_listen)(so, l); 657 if (error != 0) { 658 so->so_options = oldopt; 659 so->so_qlimit = oldqlimit; 660 sounlock(so); 661 return error; 662 } 663 sounlock(so); 664 return 0; 665 } 666 667 void 668 sofree(struct socket *so) 669 { 670 u_int refs; 671 672 KASSERT(solocked(so)); 673 674 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) { 675 sounlock(so); 676 return; 677 } 678 if (so->so_head) { 679 /* 680 * We must not decommission a socket that's on the accept(2) 681 * queue. If we do, then accept(2) may hang after select(2) 682 * indicated that the listening socket was ready. 683 */ 684 if (!soqremque(so, 0)) { 685 sounlock(so); 686 return; 687 } 688 } 689 if (so->so_rcv.sb_hiwat) 690 (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, 691 RLIM_INFINITY); 692 if (so->so_snd.sb_hiwat) 693 (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, 694 RLIM_INFINITY); 695 sbrelease(&so->so_snd, so); 696 KASSERT(!cv_has_waiters(&so->so_cv)); 697 KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv)); 698 KASSERT(!cv_has_waiters(&so->so_snd.sb_cv)); 699 sorflush(so); 700 refs = so->so_aborting; /* XXX */ 701 /* Remove acccept filter if one is present. */ 702 if (so->so_accf != NULL) 703 (void)accept_filt_clear(so); 704 sounlock(so); 705 if (refs == 0) /* XXX */ 706 soput(so); 707 } 708 709 /* 710 * soclose: close a socket on last file table reference removal. 711 * Initiate disconnect if connected. Free socket when disconnect complete. 712 */ 713 int 714 soclose(struct socket *so) 715 { 716 struct socket *so2; 717 int error = 0; 718 719 solock(so); 720 if (so->so_options & SO_ACCEPTCONN) { 721 for (;;) { 722 if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) { 723 KASSERT(solocked2(so, so2)); 724 (void) soqremque(so2, 0); 725 /* soabort drops the lock. */ 726 (void) soabort(so2); 727 solock(so); 728 continue; 729 } 730 if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) { 731 KASSERT(solocked2(so, so2)); 732 (void) soqremque(so2, 1); 733 /* soabort drops the lock. */ 734 (void) soabort(so2); 735 solock(so); 736 continue; 737 } 738 break; 739 } 740 } 741 if (so->so_pcb == NULL) 742 goto discard; 743 if (so->so_state & SS_ISCONNECTED) { 744 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 745 error = sodisconnect(so); 746 if (error) 747 goto drop; 748 } 749 if (so->so_options & SO_LINGER) { 750 if ((so->so_state & (SS_ISDISCONNECTING|SS_NBIO)) == 751 (SS_ISDISCONNECTING|SS_NBIO)) 752 goto drop; 753 while (so->so_state & SS_ISCONNECTED) { 754 error = sowait(so, true, so->so_linger * hz); 755 if (error) 756 break; 757 } 758 } 759 } 760 drop: 761 if (so->so_pcb) { 762 KASSERT(solocked(so)); 763 (*so->so_proto->pr_usrreqs->pr_detach)(so); 764 } 765 discard: 766 KASSERT((so->so_state & SS_NOFDREF) == 0); 767 kauth_cred_free(so->so_cred); 768 so->so_cred = NULL; 769 so->so_state |= SS_NOFDREF; 770 sofree(so); 771 return error; 772 } 773 774 /* 775 * Must be called with the socket locked.. Will return with it unlocked. 776 */ 777 int 778 soabort(struct socket *so) 779 { 780 u_int refs; 781 int error; 782 783 KASSERT(solocked(so)); 784 KASSERT(so->so_head == NULL); 785 786 so->so_aborting++; /* XXX */ 787 error = (*so->so_proto->pr_usrreqs->pr_abort)(so); 788 refs = --so->so_aborting; /* XXX */ 789 if (error || (refs == 0)) { 790 sofree(so); 791 } else { 792 sounlock(so); 793 } 794 return error; 795 } 796 797 int 798 soaccept(struct socket *so, struct sockaddr *nam) 799 { 800 int error; 801 802 KASSERT(solocked(so)); 803 KASSERT((so->so_state & SS_NOFDREF) != 0); 804 805 so->so_state &= ~SS_NOFDREF; 806 if ((so->so_state & SS_ISDISCONNECTED) == 0 || 807 (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0) 808 error = (*so->so_proto->pr_usrreqs->pr_accept)(so, nam); 809 else 810 error = ECONNABORTED; 811 812 return error; 813 } 814 815 int 816 soconnect(struct socket *so, struct sockaddr *nam, struct lwp *l) 817 { 818 int error; 819 820 KASSERT(solocked(so)); 821 822 if (so->so_options & SO_ACCEPTCONN) 823 return EOPNOTSUPP; 824 /* 825 * If protocol is connection-based, can only connect once. 826 * Otherwise, if connected, try to disconnect first. 827 * This allows user to disconnect by connecting to, e.g., 828 * a null address. 829 */ 830 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 831 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 832 (error = sodisconnect(so)))) { 833 error = EISCONN; 834 } else { 835 if (nam->sa_family != so->so_proto->pr_domain->dom_family) { 836 return EAFNOSUPPORT; 837 } 838 error = (*so->so_proto->pr_usrreqs->pr_connect)(so, nam, l); 839 } 840 841 return error; 842 } 843 844 int 845 soconnect2(struct socket *so1, struct socket *so2) 846 { 847 KASSERT(solocked2(so1, so2)); 848 849 return (*so1->so_proto->pr_usrreqs->pr_connect2)(so1, so2); 850 } 851 852 int 853 sodisconnect(struct socket *so) 854 { 855 int error; 856 857 KASSERT(solocked(so)); 858 859 if ((so->so_state & SS_ISCONNECTED) == 0) { 860 error = ENOTCONN; 861 } else if (so->so_state & SS_ISDISCONNECTING) { 862 error = EALREADY; 863 } else { 864 error = (*so->so_proto->pr_usrreqs->pr_disconnect)(so); 865 } 866 return error; 867 } 868 869 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 870 /* 871 * Send on a socket. 872 * If send must go all at once and message is larger than 873 * send buffering, then hard error. 874 * Lock against other senders. 875 * If must go all at once and not enough room now, then 876 * inform user that this would block and do nothing. 877 * Otherwise, if nonblocking, send as much as possible. 878 * The data to be sent is described by "uio" if nonzero, 879 * otherwise by the mbuf chain "top" (which must be null 880 * if uio is not). Data provided in mbuf chain must be small 881 * enough to send all at once. 882 * 883 * Returns nonzero on error, timeout or signal; callers 884 * must check for short counts if EINTR/ERESTART are returned. 885 * Data and control buffers are freed on return. 886 */ 887 int 888 sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, 889 struct mbuf *top, struct mbuf *control, int flags, struct lwp *l) 890 { 891 struct mbuf **mp, *m; 892 long space, len, resid, clen, mlen; 893 int error, s, dontroute, atomic; 894 short wakeup_state = 0; 895 896 clen = 0; 897 898 /* 899 * solock() provides atomicity of access. splsoftnet() prevents 900 * protocol processing soft interrupts from interrupting us and 901 * blocking (expensive). 902 */ 903 s = splsoftnet(); 904 solock(so); 905 atomic = sosendallatonce(so) || top; 906 if (uio) 907 resid = uio->uio_resid; 908 else 909 resid = top->m_pkthdr.len; 910 /* 911 * In theory resid should be unsigned. 912 * However, space must be signed, as it might be less than 0 913 * if we over-committed, and we must use a signed comparison 914 * of space and resid. On the other hand, a negative resid 915 * causes us to loop sending 0-length segments to the protocol. 916 */ 917 if (resid < 0) { 918 error = EINVAL; 919 goto out; 920 } 921 dontroute = 922 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 923 (so->so_proto->pr_flags & PR_ATOMIC); 924 l->l_ru.ru_msgsnd++; 925 if (control) 926 clen = control->m_len; 927 restart: 928 if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) 929 goto out; 930 do { 931 if (so->so_state & SS_CANTSENDMORE) { 932 error = EPIPE; 933 goto release; 934 } 935 if (so->so_error) { 936 error = so->so_error; 937 if ((flags & MSG_PEEK) == 0) 938 so->so_error = 0; 939 goto release; 940 } 941 if ((so->so_state & SS_ISCONNECTED) == 0) { 942 if (so->so_proto->pr_flags & PR_CONNREQUIRED) { 943 if (resid || clen == 0) { 944 error = ENOTCONN; 945 goto release; 946 } 947 } else if (addr == NULL) { 948 error = EDESTADDRREQ; 949 goto release; 950 } 951 } 952 space = sbspace(&so->so_snd); 953 if (flags & MSG_OOB) 954 space += 1024; 955 if ((atomic && resid > so->so_snd.sb_hiwat) || 956 clen > so->so_snd.sb_hiwat) { 957 error = EMSGSIZE; 958 goto release; 959 } 960 if (space < resid + clen && 961 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 962 if ((so->so_state & SS_NBIO) || (flags & MSG_NBIO)) { 963 error = EWOULDBLOCK; 964 goto release; 965 } 966 sbunlock(&so->so_snd); 967 if (wakeup_state & SS_RESTARTSYS) { 968 error = ERESTART; 969 goto out; 970 } 971 error = sbwait(&so->so_snd); 972 if (error) 973 goto out; 974 wakeup_state = so->so_state; 975 goto restart; 976 } 977 wakeup_state = 0; 978 mp = ⊤ 979 space -= clen; 980 do { 981 if (uio == NULL) { 982 /* 983 * Data is prepackaged in "top". 984 */ 985 resid = 0; 986 if (flags & MSG_EOR) 987 top->m_flags |= M_EOR; 988 } else do { 989 sounlock(so); 990 splx(s); 991 if (top == NULL) { 992 m = m_gethdr(M_WAIT, MT_DATA); 993 mlen = MHLEN; 994 m->m_pkthdr.len = 0; 995 m_reset_rcvif(m); 996 } else { 997 m = m_get(M_WAIT, MT_DATA); 998 mlen = MLEN; 999 } 1000 MCLAIM(m, so->so_snd.sb_mowner); 1001 if (sock_loan_thresh >= 0 && 1002 uio->uio_iov->iov_len >= sock_loan_thresh && 1003 space >= sock_loan_thresh && 1004 (len = sosend_loan(so, uio, m, 1005 space)) != 0) { 1006 SOSEND_COUNTER_INCR(&sosend_loan_big); 1007 space -= len; 1008 goto have_data; 1009 } 1010 if (resid >= MINCLSIZE && space >= MCLBYTES) { 1011 SOSEND_COUNTER_INCR(&sosend_copy_big); 1012 m_clget(m, M_DONTWAIT); 1013 if ((m->m_flags & M_EXT) == 0) 1014 goto nopages; 1015 mlen = MCLBYTES; 1016 if (atomic && top == 0) { 1017 len = lmin(MCLBYTES - max_hdr, 1018 resid); 1019 m->m_data += max_hdr; 1020 } else 1021 len = lmin(MCLBYTES, resid); 1022 space -= len; 1023 } else { 1024 nopages: 1025 SOSEND_COUNTER_INCR(&sosend_copy_small); 1026 len = lmin(lmin(mlen, resid), space); 1027 space -= len; 1028 /* 1029 * For datagram protocols, leave room 1030 * for protocol headers in first mbuf. 1031 */ 1032 if (atomic && top == 0 && len < mlen) 1033 m_align(m, len); 1034 } 1035 error = uiomove(mtod(m, void *), (int)len, uio); 1036 have_data: 1037 resid = uio->uio_resid; 1038 m->m_len = len; 1039 *mp = m; 1040 top->m_pkthdr.len += len; 1041 s = splsoftnet(); 1042 solock(so); 1043 if (error != 0) 1044 goto release; 1045 mp = &m->m_next; 1046 if (resid <= 0) { 1047 if (flags & MSG_EOR) 1048 top->m_flags |= M_EOR; 1049 break; 1050 } 1051 } while (space > 0 && atomic); 1052 1053 if (so->so_state & SS_CANTSENDMORE) { 1054 error = EPIPE; 1055 goto release; 1056 } 1057 if (dontroute) 1058 so->so_options |= SO_DONTROUTE; 1059 if (resid > 0) 1060 so->so_state |= SS_MORETOCOME; 1061 if (flags & MSG_OOB) { 1062 error = (*so->so_proto->pr_usrreqs->pr_sendoob)( 1063 so, top, control); 1064 } else { 1065 error = (*so->so_proto->pr_usrreqs->pr_send)(so, 1066 top, addr, control, l); 1067 } 1068 if (dontroute) 1069 so->so_options &= ~SO_DONTROUTE; 1070 if (resid > 0) 1071 so->so_state &= ~SS_MORETOCOME; 1072 clen = 0; 1073 control = NULL; 1074 top = NULL; 1075 mp = ⊤ 1076 if (error != 0) 1077 goto release; 1078 } while (resid && space > 0); 1079 } while (resid); 1080 1081 release: 1082 sbunlock(&so->so_snd); 1083 out: 1084 sounlock(so); 1085 splx(s); 1086 if (top) 1087 m_freem(top); 1088 if (control) 1089 m_freem(control); 1090 return error; 1091 } 1092 1093 /* 1094 * Following replacement or removal of the first mbuf on the first 1095 * mbuf chain of a socket buffer, push necessary state changes back 1096 * into the socket buffer so that other consumers see the values 1097 * consistently. 'nextrecord' is the caller's locally stored value of 1098 * the original value of sb->sb_mb->m_nextpkt which must be restored 1099 * when the lead mbuf changes. NOTE: 'nextrecord' may be NULL. 1100 */ 1101 static void 1102 sbsync(struct sockbuf *sb, struct mbuf *nextrecord) 1103 { 1104 1105 KASSERT(solocked(sb->sb_so)); 1106 1107 /* 1108 * First, update for the new value of nextrecord. If necessary, 1109 * make it the first record. 1110 */ 1111 if (sb->sb_mb != NULL) 1112 sb->sb_mb->m_nextpkt = nextrecord; 1113 else 1114 sb->sb_mb = nextrecord; 1115 1116 /* 1117 * Now update any dependent socket buffer fields to reflect 1118 * the new state. This is an inline of SB_EMPTY_FIXUP, with 1119 * the addition of a second clause that takes care of the 1120 * case where sb_mb has been updated, but remains the last 1121 * record. 1122 */ 1123 if (sb->sb_mb == NULL) { 1124 sb->sb_mbtail = NULL; 1125 sb->sb_lastrecord = NULL; 1126 } else if (sb->sb_mb->m_nextpkt == NULL) 1127 sb->sb_lastrecord = sb->sb_mb; 1128 } 1129 1130 /* 1131 * Implement receive operations on a socket. 1132 * 1133 * We depend on the way that records are added to the sockbuf by sbappend*. In 1134 * particular, each record (mbufs linked through m_next) must begin with an 1135 * address if the protocol so specifies, followed by an optional mbuf or mbufs 1136 * containing ancillary data, and then zero or more mbufs of data. 1137 * 1138 * In order to avoid blocking network interrupts for the entire time here, we 1139 * splx() while doing the actual copy to user space. Although the sockbuf is 1140 * locked, new data may still be appended, and thus we must maintain 1141 * consistency of the sockbuf during that time. 1142 * 1143 * The caller may receive the data as a single mbuf chain by supplying an mbuf 1144 * **mp0 for use in returning the chain. The uio is then used only for the 1145 * count in uio_resid. 1146 */ 1147 int 1148 soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, 1149 struct mbuf **mp0, struct mbuf **controlp, int *flagsp) 1150 { 1151 struct lwp *l = curlwp; 1152 struct mbuf *m, **mp, *mt; 1153 size_t len, offset, moff, orig_resid; 1154 int atomic, flags, error, s, type; 1155 const struct protosw *pr; 1156 struct mbuf *nextrecord; 1157 int mbuf_removed = 0; 1158 const struct domain *dom; 1159 short wakeup_state = 0; 1160 1161 pr = so->so_proto; 1162 atomic = pr->pr_flags & PR_ATOMIC; 1163 dom = pr->pr_domain; 1164 mp = mp0; 1165 type = 0; 1166 orig_resid = uio->uio_resid; 1167 1168 if (paddr != NULL) 1169 *paddr = NULL; 1170 if (controlp != NULL) 1171 *controlp = NULL; 1172 if (flagsp != NULL) 1173 flags = *flagsp &~ MSG_EOR; 1174 else 1175 flags = 0; 1176 1177 if (flags & MSG_OOB) { 1178 m = m_get(M_WAIT, MT_DATA); 1179 solock(so); 1180 error = (*pr->pr_usrreqs->pr_recvoob)(so, m, flags & MSG_PEEK); 1181 sounlock(so); 1182 if (error) 1183 goto bad; 1184 do { 1185 error = uiomove(mtod(m, void *), 1186 MIN(uio->uio_resid, m->m_len), uio); 1187 m = m_free(m); 1188 } while (uio->uio_resid > 0 && error == 0 && m); 1189 bad: 1190 if (m != NULL) 1191 m_freem(m); 1192 return error; 1193 } 1194 if (mp != NULL) 1195 *mp = NULL; 1196 1197 /* 1198 * solock() provides atomicity of access. splsoftnet() prevents 1199 * protocol processing soft interrupts from interrupting us and 1200 * blocking (expensive). 1201 */ 1202 s = splsoftnet(); 1203 solock(so); 1204 restart: 1205 if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) { 1206 sounlock(so); 1207 splx(s); 1208 return error; 1209 } 1210 m = so->so_rcv.sb_mb; 1211 1212 /* 1213 * If we have less data than requested, block awaiting more 1214 * (subject to any timeout) if: 1215 * 1. the current count is less than the low water mark, 1216 * 2. MSG_WAITALL is set, and it is possible to do the entire 1217 * receive operation at once if we block (resid <= hiwat), or 1218 * 3. MSG_DONTWAIT is not set. 1219 * If MSG_WAITALL is set but resid is larger than the receive buffer, 1220 * we have to do the receive in sections, and thus risk returning 1221 * a short count if a timeout or signal occurs after we start. 1222 */ 1223 if (m == NULL || 1224 ((flags & MSG_DONTWAIT) == 0 && 1225 so->so_rcv.sb_cc < uio->uio_resid && 1226 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 1227 ((flags & MSG_WAITALL) && 1228 uio->uio_resid <= so->so_rcv.sb_hiwat)) && 1229 m->m_nextpkt == NULL && !atomic)) { 1230 #ifdef DIAGNOSTIC 1231 if (m == NULL && so->so_rcv.sb_cc) 1232 panic("receive 1"); 1233 #endif 1234 if (so->so_error || so->so_rerror) { 1235 u_short *e; 1236 if (m != NULL) 1237 goto dontblock; 1238 e = so->so_error ? &so->so_error : &so->so_rerror; 1239 error = *e; 1240 if ((flags & MSG_PEEK) == 0) 1241 *e = 0; 1242 goto release; 1243 } 1244 if (so->so_state & SS_CANTRCVMORE) { 1245 if (m != NULL) 1246 goto dontblock; 1247 else 1248 goto release; 1249 } 1250 for (; m != NULL; m = m->m_next) 1251 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 1252 m = so->so_rcv.sb_mb; 1253 goto dontblock; 1254 } 1255 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 1256 (so->so_proto->pr_flags & PR_CONNREQUIRED)) { 1257 error = ENOTCONN; 1258 goto release; 1259 } 1260 if (uio->uio_resid == 0) 1261 goto release; 1262 if ((so->so_state & SS_NBIO) || 1263 (flags & (MSG_DONTWAIT|MSG_NBIO))) { 1264 error = EWOULDBLOCK; 1265 goto release; 1266 } 1267 SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1"); 1268 SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1"); 1269 sbunlock(&so->so_rcv); 1270 if (wakeup_state & SS_RESTARTSYS) 1271 error = ERESTART; 1272 else 1273 error = sbwait(&so->so_rcv); 1274 if (error != 0) { 1275 sounlock(so); 1276 splx(s); 1277 return error; 1278 } 1279 wakeup_state = so->so_state; 1280 goto restart; 1281 } 1282 1283 dontblock: 1284 /* 1285 * On entry here, m points to the first record of the socket buffer. 1286 * From this point onward, we maintain 'nextrecord' as a cache of the 1287 * pointer to the next record in the socket buffer. We must keep the 1288 * various socket buffer pointers and local stack versions of the 1289 * pointers in sync, pushing out modifications before dropping the 1290 * socket lock, and re-reading them when picking it up. 1291 * 1292 * Otherwise, we will race with the network stack appending new data 1293 * or records onto the socket buffer by using inconsistent/stale 1294 * versions of the field, possibly resulting in socket buffer 1295 * corruption. 1296 * 1297 * By holding the high-level sblock(), we prevent simultaneous 1298 * readers from pulling off the front of the socket buffer. 1299 */ 1300 if (l != NULL) 1301 l->l_ru.ru_msgrcv++; 1302 KASSERT(m == so->so_rcv.sb_mb); 1303 SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); 1304 SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); 1305 nextrecord = m->m_nextpkt; 1306 1307 if (pr->pr_flags & PR_ADDR) { 1308 KASSERT(m->m_type == MT_SONAME); 1309 orig_resid = 0; 1310 if (flags & MSG_PEEK) { 1311 if (paddr) 1312 *paddr = m_copym(m, 0, m->m_len, M_DONTWAIT); 1313 m = m->m_next; 1314 } else { 1315 sbfree(&so->so_rcv, m); 1316 mbuf_removed = 1; 1317 if (paddr != NULL) { 1318 *paddr = m; 1319 so->so_rcv.sb_mb = m->m_next; 1320 m->m_next = NULL; 1321 m = so->so_rcv.sb_mb; 1322 } else { 1323 m = so->so_rcv.sb_mb = m_free(m); 1324 } 1325 sbsync(&so->so_rcv, nextrecord); 1326 } 1327 } 1328 1329 if (pr->pr_flags & PR_ADDR_OPT) { 1330 /* 1331 * For SCTP we may be getting a whole message OR a partial 1332 * delivery. 1333 */ 1334 if (m->m_type == MT_SONAME) { 1335 orig_resid = 0; 1336 if (flags & MSG_PEEK) { 1337 if (paddr) 1338 *paddr = m_copym(m, 0, m->m_len, M_DONTWAIT); 1339 m = m->m_next; 1340 } else { 1341 sbfree(&so->so_rcv, m); 1342 mbuf_removed = 1; 1343 if (paddr) { 1344 *paddr = m; 1345 so->so_rcv.sb_mb = m->m_next; 1346 m->m_next = 0; 1347 m = so->so_rcv.sb_mb; 1348 } else { 1349 m = so->so_rcv.sb_mb = m_free(m); 1350 } 1351 sbsync(&so->so_rcv, nextrecord); 1352 } 1353 } 1354 } 1355 1356 /* 1357 * Process one or more MT_CONTROL mbufs present before any data mbufs 1358 * in the first mbuf chain on the socket buffer. If MSG_PEEK, we 1359 * just copy the data; if !MSG_PEEK, we call into the protocol to 1360 * perform externalization (or freeing if controlp == NULL). 1361 */ 1362 if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) { 1363 struct mbuf *cm = NULL, *cmn; 1364 struct mbuf **cme = &cm; 1365 1366 do { 1367 if (flags & MSG_PEEK) { 1368 if (controlp != NULL) { 1369 *controlp = m_copym(m, 0, m->m_len, M_DONTWAIT); 1370 controlp = &(*controlp)->m_next; 1371 } 1372 m = m->m_next; 1373 } else { 1374 sbfree(&so->so_rcv, m); 1375 so->so_rcv.sb_mb = m->m_next; 1376 m->m_next = NULL; 1377 *cme = m; 1378 cme = &(*cme)->m_next; 1379 m = so->so_rcv.sb_mb; 1380 } 1381 } while (m != NULL && m->m_type == MT_CONTROL); 1382 if ((flags & MSG_PEEK) == 0) 1383 sbsync(&so->so_rcv, nextrecord); 1384 1385 for (; cm != NULL; cm = cmn) { 1386 cmn = cm->m_next; 1387 cm->m_next = NULL; 1388 type = mtod(cm, struct cmsghdr *)->cmsg_type; 1389 if (controlp != NULL) { 1390 if (dom->dom_externalize != NULL && 1391 type == SCM_RIGHTS) { 1392 sounlock(so); 1393 splx(s); 1394 error = (*dom->dom_externalize)(cm, l, 1395 (flags & MSG_CMSG_CLOEXEC) ? 1396 O_CLOEXEC : 0); 1397 s = splsoftnet(); 1398 solock(so); 1399 } 1400 *controlp = cm; 1401 while (*controlp != NULL) 1402 controlp = &(*controlp)->m_next; 1403 } else { 1404 /* 1405 * Dispose of any SCM_RIGHTS message that went 1406 * through the read path rather than recv. 1407 */ 1408 if (dom->dom_dispose != NULL && 1409 type == SCM_RIGHTS) { 1410 sounlock(so); 1411 (*dom->dom_dispose)(cm); 1412 solock(so); 1413 } 1414 m_freem(cm); 1415 } 1416 } 1417 if (m != NULL) 1418 nextrecord = so->so_rcv.sb_mb->m_nextpkt; 1419 else 1420 nextrecord = so->so_rcv.sb_mb; 1421 orig_resid = 0; 1422 } 1423 1424 /* If m is non-NULL, we have some data to read. */ 1425 if (__predict_true(m != NULL)) { 1426 type = m->m_type; 1427 if (type == MT_OOBDATA) 1428 flags |= MSG_OOB; 1429 } 1430 SBLASTRECORDCHK(&so->so_rcv, "soreceive 2"); 1431 SBLASTMBUFCHK(&so->so_rcv, "soreceive 2"); 1432 1433 moff = 0; 1434 offset = 0; 1435 while (m != NULL && uio->uio_resid > 0 && error == 0) { 1436 /* 1437 * If the type of mbuf has changed, end the receive 1438 * operation and do a short read. 1439 */ 1440 if (m->m_type == MT_OOBDATA) { 1441 if (type != MT_OOBDATA) 1442 break; 1443 } else if (type == MT_OOBDATA) { 1444 break; 1445 } else if (m->m_type == MT_CONTROL) { 1446 break; 1447 } 1448 #ifdef DIAGNOSTIC 1449 else if (m->m_type != MT_DATA && m->m_type != MT_HEADER) { 1450 panic("%s: m_type=%d", __func__, m->m_type); 1451 } 1452 #endif 1453 1454 so->so_state &= ~SS_RCVATMARK; 1455 wakeup_state = 0; 1456 len = uio->uio_resid; 1457 if (so->so_oobmark && len > so->so_oobmark - offset) 1458 len = so->so_oobmark - offset; 1459 if (len > m->m_len - moff) 1460 len = m->m_len - moff; 1461 1462 /* 1463 * If mp is set, just pass back the mbufs. 1464 * Otherwise copy them out via the uio, then free. 1465 * Sockbuf must be consistent here (points to current mbuf, 1466 * it points to next record) when we drop priority; 1467 * we must note any additions to the sockbuf when we 1468 * block interrupts again. 1469 */ 1470 if (mp == NULL) { 1471 SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); 1472 SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); 1473 sounlock(so); 1474 splx(s); 1475 error = uiomove(mtod(m, char *) + moff, len, uio); 1476 s = splsoftnet(); 1477 solock(so); 1478 if (error != 0) { 1479 /* 1480 * If any part of the record has been removed 1481 * (such as the MT_SONAME mbuf, which will 1482 * happen when PR_ADDR, and thus also 1483 * PR_ATOMIC, is set), then drop the entire 1484 * record to maintain the atomicity of the 1485 * receive operation. 1486 * 1487 * This avoids a later panic("receive 1a") 1488 * when compiled with DIAGNOSTIC. 1489 */ 1490 if (m && mbuf_removed && atomic) 1491 (void) sbdroprecord(&so->so_rcv); 1492 1493 goto release; 1494 } 1495 } else { 1496 uio->uio_resid -= len; 1497 } 1498 1499 if (len == m->m_len - moff) { 1500 if (m->m_flags & M_EOR) 1501 flags |= MSG_EOR; 1502 #ifdef SCTP 1503 if (m->m_flags & M_NOTIFICATION) 1504 flags |= MSG_NOTIFICATION; 1505 #endif 1506 if (flags & MSG_PEEK) { 1507 m = m->m_next; 1508 moff = 0; 1509 } else { 1510 nextrecord = m->m_nextpkt; 1511 sbfree(&so->so_rcv, m); 1512 if (mp) { 1513 *mp = m; 1514 mp = &m->m_next; 1515 so->so_rcv.sb_mb = m = m->m_next; 1516 *mp = NULL; 1517 } else { 1518 m = so->so_rcv.sb_mb = m_free(m); 1519 } 1520 /* 1521 * If m != NULL, we also know that 1522 * so->so_rcv.sb_mb != NULL. 1523 */ 1524 KASSERT(so->so_rcv.sb_mb == m); 1525 if (m) { 1526 m->m_nextpkt = nextrecord; 1527 if (nextrecord == NULL) 1528 so->so_rcv.sb_lastrecord = m; 1529 } else { 1530 so->so_rcv.sb_mb = nextrecord; 1531 SB_EMPTY_FIXUP(&so->so_rcv); 1532 } 1533 SBLASTRECORDCHK(&so->so_rcv, "soreceive 3"); 1534 SBLASTMBUFCHK(&so->so_rcv, "soreceive 3"); 1535 } 1536 } else if (flags & MSG_PEEK) { 1537 moff += len; 1538 } else { 1539 if (mp != NULL) { 1540 mt = m_copym(m, 0, len, M_NOWAIT); 1541 if (__predict_false(mt == NULL)) { 1542 sounlock(so); 1543 mt = m_copym(m, 0, len, M_WAIT); 1544 solock(so); 1545 } 1546 *mp = mt; 1547 } 1548 m->m_data += len; 1549 m->m_len -= len; 1550 so->so_rcv.sb_cc -= len; 1551 } 1552 1553 if (so->so_oobmark) { 1554 if ((flags & MSG_PEEK) == 0) { 1555 so->so_oobmark -= len; 1556 if (so->so_oobmark == 0) { 1557 so->so_state |= SS_RCVATMARK; 1558 break; 1559 } 1560 } else { 1561 offset += len; 1562 if (offset == so->so_oobmark) 1563 break; 1564 } 1565 } else { 1566 so->so_state &= ~SS_POLLRDBAND; 1567 } 1568 if (flags & MSG_EOR) 1569 break; 1570 1571 /* 1572 * If the MSG_WAITALL flag is set (for non-atomic socket), 1573 * we must not quit until "uio->uio_resid == 0" or an error 1574 * termination. If a signal/timeout occurs, return 1575 * with a short count but without error. 1576 * Keep sockbuf locked against other readers. 1577 */ 1578 while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 && 1579 !sosendallatonce(so) && !nextrecord) { 1580 if (so->so_error || so->so_rerror || 1581 so->so_state & SS_CANTRCVMORE) 1582 break; 1583 /* 1584 * If we are peeking and the socket receive buffer is 1585 * full, stop since we can't get more data to peek at. 1586 */ 1587 if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0) 1588 break; 1589 /* 1590 * If we've drained the socket buffer, tell the 1591 * protocol in case it needs to do something to 1592 * get it filled again. 1593 */ 1594 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) 1595 (*pr->pr_usrreqs->pr_rcvd)(so, flags, l); 1596 SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2"); 1597 SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2"); 1598 if (wakeup_state & SS_RESTARTSYS) 1599 error = ERESTART; 1600 else 1601 error = sbwait(&so->so_rcv); 1602 if (error != 0) { 1603 sbunlock(&so->so_rcv); 1604 sounlock(so); 1605 splx(s); 1606 return 0; 1607 } 1608 if ((m = so->so_rcv.sb_mb) != NULL) 1609 nextrecord = m->m_nextpkt; 1610 wakeup_state = so->so_state; 1611 } 1612 } 1613 1614 if (m && atomic) { 1615 flags |= MSG_TRUNC; 1616 if ((flags & MSG_PEEK) == 0) 1617 (void) sbdroprecord(&so->so_rcv); 1618 } 1619 if ((flags & MSG_PEEK) == 0) { 1620 if (m == NULL) { 1621 /* 1622 * First part is an inline SB_EMPTY_FIXUP(). Second 1623 * part makes sure sb_lastrecord is up-to-date if 1624 * there is still data in the socket buffer. 1625 */ 1626 so->so_rcv.sb_mb = nextrecord; 1627 if (so->so_rcv.sb_mb == NULL) { 1628 so->so_rcv.sb_mbtail = NULL; 1629 so->so_rcv.sb_lastrecord = NULL; 1630 } else if (nextrecord->m_nextpkt == NULL) 1631 so->so_rcv.sb_lastrecord = nextrecord; 1632 } 1633 SBLASTRECORDCHK(&so->so_rcv, "soreceive 4"); 1634 SBLASTMBUFCHK(&so->so_rcv, "soreceive 4"); 1635 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1636 (*pr->pr_usrreqs->pr_rcvd)(so, flags, l); 1637 } 1638 if (orig_resid == uio->uio_resid && orig_resid && 1639 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 1640 sbunlock(&so->so_rcv); 1641 goto restart; 1642 } 1643 1644 if (flagsp != NULL) 1645 *flagsp |= flags; 1646 release: 1647 sbunlock(&so->so_rcv); 1648 sounlock(so); 1649 splx(s); 1650 return error; 1651 } 1652 1653 int 1654 soshutdown(struct socket *so, int how) 1655 { 1656 const struct protosw *pr; 1657 int error; 1658 1659 KASSERT(solocked(so)); 1660 1661 pr = so->so_proto; 1662 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 1663 return EINVAL; 1664 1665 if (how == SHUT_RD || how == SHUT_RDWR) { 1666 sorflush(so); 1667 error = 0; 1668 } 1669 if (how == SHUT_WR || how == SHUT_RDWR) 1670 error = (*pr->pr_usrreqs->pr_shutdown)(so); 1671 1672 return error; 1673 } 1674 1675 void 1676 sorestart(struct socket *so) 1677 { 1678 /* 1679 * An application has called close() on an fd on which another 1680 * of its threads has called a socket system call. 1681 * Mark this and wake everyone up, and code that would block again 1682 * instead returns ERESTART. 1683 * On system call re-entry the fd is validated and EBADF returned. 1684 * Any other fd will block again on the 2nd syscall. 1685 */ 1686 solock(so); 1687 so->so_state |= SS_RESTARTSYS; 1688 cv_broadcast(&so->so_cv); 1689 cv_broadcast(&so->so_snd.sb_cv); 1690 cv_broadcast(&so->so_rcv.sb_cv); 1691 sounlock(so); 1692 } 1693 1694 void 1695 sorflush(struct socket *so) 1696 { 1697 struct sockbuf *sb, asb; 1698 const struct protosw *pr; 1699 1700 KASSERT(solocked(so)); 1701 1702 sb = &so->so_rcv; 1703 pr = so->so_proto; 1704 socantrcvmore(so); 1705 sb->sb_flags |= SB_NOINTR; 1706 (void )sblock(sb, M_WAITOK); 1707 sbunlock(sb); 1708 asb = *sb; 1709 /* 1710 * Clear most of the sockbuf structure, but leave some of the 1711 * fields valid. 1712 */ 1713 memset(&sb->sb_startzero, 0, 1714 sizeof(*sb) - offsetof(struct sockbuf, sb_startzero)); 1715 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) { 1716 sounlock(so); 1717 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 1718 solock(so); 1719 } 1720 sbrelease(&asb, so); 1721 } 1722 1723 /* 1724 * internal set SOL_SOCKET options 1725 */ 1726 static int 1727 sosetopt1(struct socket *so, const struct sockopt *sopt) 1728 { 1729 int error, opt; 1730 int optval = 0; /* XXX: gcc */ 1731 struct linger l; 1732 struct timeval tv; 1733 1734 opt = sopt->sopt_name; 1735 1736 switch (opt) { 1737 1738 case SO_ACCEPTFILTER: 1739 error = accept_filt_setopt(so, sopt); 1740 KASSERT(solocked(so)); 1741 break; 1742 1743 case SO_LINGER: 1744 error = sockopt_get(sopt, &l, sizeof(l)); 1745 solock(so); 1746 if (error) 1747 break; 1748 if (l.l_linger < 0 || l.l_linger > USHRT_MAX || 1749 l.l_linger > (INT_MAX / hz)) { 1750 error = EDOM; 1751 break; 1752 } 1753 so->so_linger = l.l_linger; 1754 if (l.l_onoff) 1755 so->so_options |= SO_LINGER; 1756 else 1757 so->so_options &= ~SO_LINGER; 1758 break; 1759 1760 case SO_DEBUG: 1761 case SO_KEEPALIVE: 1762 case SO_DONTROUTE: 1763 case SO_USELOOPBACK: 1764 case SO_BROADCAST: 1765 case SO_REUSEADDR: 1766 case SO_REUSEPORT: 1767 case SO_OOBINLINE: 1768 case SO_TIMESTAMP: 1769 case SO_NOSIGPIPE: 1770 case SO_RERROR: 1771 error = sockopt_getint(sopt, &optval); 1772 solock(so); 1773 if (error) 1774 break; 1775 if (optval) 1776 so->so_options |= opt; 1777 else 1778 so->so_options &= ~opt; 1779 break; 1780 1781 case SO_SNDBUF: 1782 case SO_RCVBUF: 1783 case SO_SNDLOWAT: 1784 case SO_RCVLOWAT: 1785 error = sockopt_getint(sopt, &optval); 1786 solock(so); 1787 if (error) 1788 break; 1789 1790 /* 1791 * Values < 1 make no sense for any of these 1792 * options, so disallow them. 1793 */ 1794 if (optval < 1) { 1795 error = EINVAL; 1796 break; 1797 } 1798 1799 switch (opt) { 1800 case SO_SNDBUF: 1801 if (sbreserve(&so->so_snd, (u_long)optval, so) == 0) { 1802 error = ENOBUFS; 1803 break; 1804 } 1805 if (sofixedbuf) 1806 so->so_snd.sb_flags &= ~SB_AUTOSIZE; 1807 break; 1808 1809 case SO_RCVBUF: 1810 if (sbreserve(&so->so_rcv, (u_long)optval, so) == 0) { 1811 error = ENOBUFS; 1812 break; 1813 } 1814 if (sofixedbuf) 1815 so->so_rcv.sb_flags &= ~SB_AUTOSIZE; 1816 break; 1817 1818 /* 1819 * Make sure the low-water is never greater than 1820 * the high-water. 1821 */ 1822 case SO_SNDLOWAT: 1823 if (optval > so->so_snd.sb_hiwat) 1824 optval = so->so_snd.sb_hiwat; 1825 1826 so->so_snd.sb_lowat = optval; 1827 break; 1828 1829 case SO_RCVLOWAT: 1830 if (optval > so->so_rcv.sb_hiwat) 1831 optval = so->so_rcv.sb_hiwat; 1832 1833 so->so_rcv.sb_lowat = optval; 1834 break; 1835 } 1836 break; 1837 1838 case SO_SNDTIMEO: 1839 case SO_RCVTIMEO: 1840 solock(so); 1841 error = sockopt_get(sopt, &tv, sizeof(tv)); 1842 if (error) 1843 break; 1844 1845 if (tv.tv_sec < 0 || tv.tv_usec < 0 || tv.tv_usec >= 1000000) { 1846 error = EDOM; 1847 break; 1848 } 1849 if (tv.tv_sec > (INT_MAX - tv.tv_usec / tick) / hz) { 1850 error = EDOM; 1851 break; 1852 } 1853 1854 optval = tv.tv_sec * hz + tv.tv_usec / tick; 1855 if (optval == 0 && tv.tv_usec != 0) 1856 optval = 1; 1857 1858 switch (opt) { 1859 case SO_SNDTIMEO: 1860 so->so_snd.sb_timeo = optval; 1861 break; 1862 case SO_RCVTIMEO: 1863 so->so_rcv.sb_timeo = optval; 1864 break; 1865 } 1866 break; 1867 1868 default: 1869 MODULE_HOOK_CALL(uipc_socket_50_setopt1_hook, 1870 (opt, so, sopt), enosys(), error); 1871 if (error == ENOSYS || error == EPASSTHROUGH) { 1872 solock(so); 1873 error = ENOPROTOOPT; 1874 } 1875 break; 1876 } 1877 KASSERT(solocked(so)); 1878 return error; 1879 } 1880 1881 int 1882 sosetopt(struct socket *so, struct sockopt *sopt) 1883 { 1884 int error, prerr; 1885 1886 if (sopt->sopt_level == SOL_SOCKET) { 1887 error = sosetopt1(so, sopt); 1888 KASSERT(solocked(so)); 1889 } else { 1890 error = ENOPROTOOPT; 1891 solock(so); 1892 } 1893 1894 if ((error == 0 || error == ENOPROTOOPT) && 1895 so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) { 1896 /* give the protocol stack a shot */ 1897 prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, sopt); 1898 if (prerr == 0) 1899 error = 0; 1900 else if (prerr != ENOPROTOOPT) 1901 error = prerr; 1902 } 1903 sounlock(so); 1904 return error; 1905 } 1906 1907 /* 1908 * so_setsockopt() is a wrapper providing a sockopt structure for sosetopt() 1909 */ 1910 int 1911 so_setsockopt(struct lwp *l, struct socket *so, int level, int name, 1912 const void *val, size_t valsize) 1913 { 1914 struct sockopt sopt; 1915 int error; 1916 1917 KASSERT(valsize == 0 || val != NULL); 1918 1919 sockopt_init(&sopt, level, name, valsize); 1920 sockopt_set(&sopt, val, valsize); 1921 1922 error = sosetopt(so, &sopt); 1923 1924 sockopt_destroy(&sopt); 1925 1926 return error; 1927 } 1928 1929 /* 1930 * internal get SOL_SOCKET options 1931 */ 1932 static int 1933 sogetopt1(struct socket *so, struct sockopt *sopt) 1934 { 1935 int error, optval, opt; 1936 struct linger l; 1937 struct timeval tv; 1938 1939 switch ((opt = sopt->sopt_name)) { 1940 1941 case SO_ACCEPTFILTER: 1942 error = accept_filt_getopt(so, sopt); 1943 break; 1944 1945 case SO_LINGER: 1946 l.l_onoff = (so->so_options & SO_LINGER) ? 1 : 0; 1947 l.l_linger = so->so_linger; 1948 1949 error = sockopt_set(sopt, &l, sizeof(l)); 1950 break; 1951 1952 case SO_USELOOPBACK: 1953 case SO_DONTROUTE: 1954 case SO_DEBUG: 1955 case SO_KEEPALIVE: 1956 case SO_REUSEADDR: 1957 case SO_REUSEPORT: 1958 case SO_BROADCAST: 1959 case SO_OOBINLINE: 1960 case SO_TIMESTAMP: 1961 case SO_NOSIGPIPE: 1962 case SO_RERROR: 1963 case SO_ACCEPTCONN: 1964 error = sockopt_setint(sopt, (so->so_options & opt) ? 1 : 0); 1965 break; 1966 1967 case SO_TYPE: 1968 error = sockopt_setint(sopt, so->so_type); 1969 break; 1970 1971 case SO_ERROR: 1972 if (so->so_error == 0) { 1973 so->so_error = so->so_rerror; 1974 so->so_rerror = 0; 1975 } 1976 error = sockopt_setint(sopt, so->so_error); 1977 so->so_error = 0; 1978 break; 1979 1980 case SO_SNDBUF: 1981 error = sockopt_setint(sopt, so->so_snd.sb_hiwat); 1982 break; 1983 1984 case SO_RCVBUF: 1985 error = sockopt_setint(sopt, so->so_rcv.sb_hiwat); 1986 break; 1987 1988 case SO_SNDLOWAT: 1989 error = sockopt_setint(sopt, so->so_snd.sb_lowat); 1990 break; 1991 1992 case SO_RCVLOWAT: 1993 error = sockopt_setint(sopt, so->so_rcv.sb_lowat); 1994 break; 1995 1996 case SO_SNDTIMEO: 1997 case SO_RCVTIMEO: 1998 optval = (opt == SO_SNDTIMEO ? 1999 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 2000 2001 memset(&tv, 0, sizeof(tv)); 2002 tv.tv_sec = optval / hz; 2003 tv.tv_usec = (optval % hz) * tick; 2004 2005 error = sockopt_set(sopt, &tv, sizeof(tv)); 2006 break; 2007 2008 case SO_OVERFLOWED: 2009 error = sockopt_setint(sopt, so->so_rcv.sb_overflowed); 2010 break; 2011 2012 default: 2013 MODULE_HOOK_CALL(uipc_socket_50_getopt1_hook, 2014 (opt, so, sopt), enosys(), error); 2015 if (error) 2016 error = ENOPROTOOPT; 2017 break; 2018 } 2019 2020 return error; 2021 } 2022 2023 int 2024 sogetopt(struct socket *so, struct sockopt *sopt) 2025 { 2026 int error; 2027 2028 solock(so); 2029 if (sopt->sopt_level != SOL_SOCKET) { 2030 if (so->so_proto && so->so_proto->pr_ctloutput) { 2031 error = ((*so->so_proto->pr_ctloutput) 2032 (PRCO_GETOPT, so, sopt)); 2033 } else 2034 error = (ENOPROTOOPT); 2035 } else { 2036 error = sogetopt1(so, sopt); 2037 } 2038 sounlock(so); 2039 return error; 2040 } 2041 2042 /* 2043 * alloc sockopt data buffer buffer 2044 * - will be released at destroy 2045 */ 2046 static int 2047 sockopt_alloc(struct sockopt *sopt, size_t len, km_flag_t kmflag) 2048 { 2049 void *data; 2050 2051 KASSERT(sopt->sopt_size == 0); 2052 2053 if (len > sizeof(sopt->sopt_buf)) { 2054 data = kmem_zalloc(len, kmflag); 2055 if (data == NULL) 2056 return ENOMEM; 2057 sopt->sopt_data = data; 2058 } else 2059 sopt->sopt_data = sopt->sopt_buf; 2060 2061 sopt->sopt_size = len; 2062 return 0; 2063 } 2064 2065 /* 2066 * initialise sockopt storage 2067 * - MAY sleep during allocation 2068 */ 2069 void 2070 sockopt_init(struct sockopt *sopt, int level, int name, size_t size) 2071 { 2072 2073 memset(sopt, 0, sizeof(*sopt)); 2074 2075 sopt->sopt_level = level; 2076 sopt->sopt_name = name; 2077 (void)sockopt_alloc(sopt, size, KM_SLEEP); 2078 } 2079 2080 /* 2081 * destroy sockopt storage 2082 * - will release any held memory references 2083 */ 2084 void 2085 sockopt_destroy(struct sockopt *sopt) 2086 { 2087 2088 if (sopt->sopt_data != sopt->sopt_buf) 2089 kmem_free(sopt->sopt_data, sopt->sopt_size); 2090 2091 memset(sopt, 0, sizeof(*sopt)); 2092 } 2093 2094 /* 2095 * set sockopt value 2096 * - value is copied into sockopt 2097 * - memory is allocated when necessary, will not sleep 2098 */ 2099 int 2100 sockopt_set(struct sockopt *sopt, const void *buf, size_t len) 2101 { 2102 int error; 2103 2104 if (sopt->sopt_size == 0) { 2105 error = sockopt_alloc(sopt, len, KM_NOSLEEP); 2106 if (error) 2107 return error; 2108 } 2109 2110 sopt->sopt_retsize = MIN(sopt->sopt_size, len); 2111 if (sopt->sopt_retsize > 0) { 2112 memcpy(sopt->sopt_data, buf, sopt->sopt_retsize); 2113 } 2114 2115 return 0; 2116 } 2117 2118 /* 2119 * common case of set sockopt integer value 2120 */ 2121 int 2122 sockopt_setint(struct sockopt *sopt, int val) 2123 { 2124 2125 return sockopt_set(sopt, &val, sizeof(int)); 2126 } 2127 2128 /* 2129 * get sockopt value 2130 * - correct size must be given 2131 */ 2132 int 2133 sockopt_get(const struct sockopt *sopt, void *buf, size_t len) 2134 { 2135 2136 if (sopt->sopt_size != len) 2137 return EINVAL; 2138 2139 memcpy(buf, sopt->sopt_data, len); 2140 return 0; 2141 } 2142 2143 /* 2144 * common case of get sockopt integer value 2145 */ 2146 int 2147 sockopt_getint(const struct sockopt *sopt, int *valp) 2148 { 2149 2150 return sockopt_get(sopt, valp, sizeof(int)); 2151 } 2152 2153 /* 2154 * set sockopt value from mbuf 2155 * - ONLY for legacy code 2156 * - mbuf is released by sockopt 2157 * - will not sleep 2158 */ 2159 int 2160 sockopt_setmbuf(struct sockopt *sopt, struct mbuf *m) 2161 { 2162 size_t len; 2163 int error; 2164 2165 len = m_length(m); 2166 2167 if (sopt->sopt_size == 0) { 2168 error = sockopt_alloc(sopt, len, KM_NOSLEEP); 2169 if (error) 2170 return error; 2171 } 2172 2173 sopt->sopt_retsize = MIN(sopt->sopt_size, len); 2174 m_copydata(m, 0, sopt->sopt_retsize, sopt->sopt_data); 2175 m_freem(m); 2176 2177 return 0; 2178 } 2179 2180 /* 2181 * get sockopt value into mbuf 2182 * - ONLY for legacy code 2183 * - mbuf to be released by the caller 2184 * - will not sleep 2185 */ 2186 struct mbuf * 2187 sockopt_getmbuf(const struct sockopt *sopt) 2188 { 2189 struct mbuf *m; 2190 2191 if (sopt->sopt_size > MCLBYTES) 2192 return NULL; 2193 2194 m = m_get(M_DONTWAIT, MT_SOOPTS); 2195 if (m == NULL) 2196 return NULL; 2197 2198 if (sopt->sopt_size > MLEN) { 2199 MCLGET(m, M_DONTWAIT); 2200 if ((m->m_flags & M_EXT) == 0) { 2201 m_free(m); 2202 return NULL; 2203 } 2204 } 2205 2206 memcpy(mtod(m, void *), sopt->sopt_data, sopt->sopt_size); 2207 m->m_len = sopt->sopt_size; 2208 2209 return m; 2210 } 2211 2212 void 2213 sohasoutofband(struct socket *so) 2214 { 2215 2216 so->so_state |= SS_POLLRDBAND; 2217 fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so); 2218 selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, NOTE_SUBMIT); 2219 } 2220 2221 static void 2222 filt_sordetach(struct knote *kn) 2223 { 2224 struct socket *so; 2225 2226 so = ((file_t *)kn->kn_obj)->f_socket; 2227 solock(so); 2228 if (selremove_knote(&so->so_rcv.sb_sel, kn)) 2229 so->so_rcv.sb_flags &= ~SB_KNOTE; 2230 sounlock(so); 2231 } 2232 2233 /*ARGSUSED*/ 2234 static int 2235 filt_soread(struct knote *kn, long hint) 2236 { 2237 struct socket *so; 2238 int rv; 2239 2240 so = ((file_t *)kn->kn_obj)->f_socket; 2241 if (hint != NOTE_SUBMIT) 2242 solock(so); 2243 kn->kn_data = so->so_rcv.sb_cc; 2244 if (so->so_state & SS_CANTRCVMORE) { 2245 kn->kn_flags |= EV_EOF; 2246 kn->kn_fflags = so->so_error; 2247 rv = 1; 2248 } else if (so->so_error || so->so_rerror) 2249 rv = 1; 2250 else if (kn->kn_sfflags & NOTE_LOWAT) 2251 rv = (kn->kn_data >= kn->kn_sdata); 2252 else 2253 rv = (kn->kn_data >= so->so_rcv.sb_lowat); 2254 if (hint != NOTE_SUBMIT) 2255 sounlock(so); 2256 return rv; 2257 } 2258 2259 static void 2260 filt_sowdetach(struct knote *kn) 2261 { 2262 struct socket *so; 2263 2264 so = ((file_t *)kn->kn_obj)->f_socket; 2265 solock(so); 2266 if (selremove_knote(&so->so_snd.sb_sel, kn)) 2267 so->so_snd.sb_flags &= ~SB_KNOTE; 2268 sounlock(so); 2269 } 2270 2271 /*ARGSUSED*/ 2272 static int 2273 filt_sowrite(struct knote *kn, long hint) 2274 { 2275 struct socket *so; 2276 int rv; 2277 2278 so = ((file_t *)kn->kn_obj)->f_socket; 2279 if (hint != NOTE_SUBMIT) 2280 solock(so); 2281 kn->kn_data = sbspace(&so->so_snd); 2282 if (so->so_state & SS_CANTSENDMORE) { 2283 kn->kn_flags |= EV_EOF; 2284 kn->kn_fflags = so->so_error; 2285 rv = 1; 2286 } else if (so->so_error) 2287 rv = 1; 2288 else if (((so->so_state & SS_ISCONNECTED) == 0) && 2289 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 2290 rv = 0; 2291 else if (kn->kn_sfflags & NOTE_LOWAT) 2292 rv = (kn->kn_data >= kn->kn_sdata); 2293 else 2294 rv = (kn->kn_data >= so->so_snd.sb_lowat); 2295 if (hint != NOTE_SUBMIT) 2296 sounlock(so); 2297 return rv; 2298 } 2299 2300 /*ARGSUSED*/ 2301 static int 2302 filt_solisten(struct knote *kn, long hint) 2303 { 2304 struct socket *so; 2305 int rv; 2306 2307 so = ((file_t *)kn->kn_obj)->f_socket; 2308 2309 /* 2310 * Set kn_data to number of incoming connections, not 2311 * counting partial (incomplete) connections. 2312 */ 2313 if (hint != NOTE_SUBMIT) 2314 solock(so); 2315 kn->kn_data = so->so_qlen; 2316 rv = (kn->kn_data > 0); 2317 if (hint != NOTE_SUBMIT) 2318 sounlock(so); 2319 return rv; 2320 } 2321 2322 static const struct filterops solisten_filtops = { 2323 .f_flags = FILTEROP_ISFD, 2324 .f_attach = NULL, 2325 .f_detach = filt_sordetach, 2326 .f_event = filt_solisten, 2327 }; 2328 2329 static const struct filterops soread_filtops = { 2330 .f_flags = FILTEROP_ISFD, 2331 .f_attach = NULL, 2332 .f_detach = filt_sordetach, 2333 .f_event = filt_soread, 2334 }; 2335 2336 static const struct filterops sowrite_filtops = { 2337 .f_flags = FILTEROP_ISFD, 2338 .f_attach = NULL, 2339 .f_detach = filt_sowdetach, 2340 .f_event = filt_sowrite, 2341 }; 2342 2343 int 2344 soo_kqfilter(struct file *fp, struct knote *kn) 2345 { 2346 struct socket *so; 2347 struct sockbuf *sb; 2348 2349 so = ((file_t *)kn->kn_obj)->f_socket; 2350 solock(so); 2351 switch (kn->kn_filter) { 2352 case EVFILT_READ: 2353 if (so->so_options & SO_ACCEPTCONN) 2354 kn->kn_fop = &solisten_filtops; 2355 else 2356 kn->kn_fop = &soread_filtops; 2357 sb = &so->so_rcv; 2358 break; 2359 case EVFILT_WRITE: 2360 kn->kn_fop = &sowrite_filtops; 2361 sb = &so->so_snd; 2362 break; 2363 default: 2364 sounlock(so); 2365 return EINVAL; 2366 } 2367 selrecord_knote(&sb->sb_sel, kn); 2368 sb->sb_flags |= SB_KNOTE; 2369 sounlock(so); 2370 return 0; 2371 } 2372 2373 static int 2374 sodopoll(struct socket *so, int events) 2375 { 2376 int revents; 2377 2378 revents = 0; 2379 2380 if (events & (POLLIN | POLLRDNORM)) 2381 if (soreadable(so)) 2382 revents |= events & (POLLIN | POLLRDNORM); 2383 2384 if (events & (POLLOUT | POLLWRNORM)) 2385 if (sowritable(so)) 2386 revents |= events & (POLLOUT | POLLWRNORM); 2387 2388 if (events & (POLLPRI | POLLRDBAND)) 2389 if (so->so_state & SS_POLLRDBAND) 2390 revents |= events & (POLLPRI | POLLRDBAND); 2391 2392 return revents; 2393 } 2394 2395 int 2396 sopoll(struct socket *so, int events) 2397 { 2398 int revents = 0; 2399 2400 #ifndef DIAGNOSTIC 2401 /* 2402 * Do a quick, unlocked check in expectation that the socket 2403 * will be ready for I/O. Don't do this check if DIAGNOSTIC, 2404 * as the solocked() assertions will fail. 2405 */ 2406 if ((revents = sodopoll(so, events)) != 0) 2407 return revents; 2408 #endif 2409 2410 solock(so); 2411 if ((revents = sodopoll(so, events)) == 0) { 2412 if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) { 2413 selrecord(curlwp, &so->so_rcv.sb_sel); 2414 so->so_rcv.sb_flags |= SB_NOTIFY; 2415 } 2416 2417 if (events & (POLLOUT | POLLWRNORM)) { 2418 selrecord(curlwp, &so->so_snd.sb_sel); 2419 so->so_snd.sb_flags |= SB_NOTIFY; 2420 } 2421 } 2422 sounlock(so); 2423 2424 return revents; 2425 } 2426 2427 struct mbuf ** 2428 sbsavetimestamp(int opt, struct mbuf **mp) 2429 { 2430 struct timeval tv; 2431 int error; 2432 2433 memset(&tv, 0, sizeof(tv)); 2434 microtime(&tv); 2435 2436 MODULE_HOOK_CALL(uipc_socket_50_sbts_hook, (opt, &mp), enosys(), error); 2437 if (error == 0) 2438 return mp; 2439 2440 if (opt & SO_TIMESTAMP) { 2441 *mp = sbcreatecontrol(&tv, sizeof(tv), 2442 SCM_TIMESTAMP, SOL_SOCKET); 2443 if (*mp) 2444 mp = &(*mp)->m_next; 2445 } 2446 return mp; 2447 } 2448 2449 2450 #include <sys/sysctl.h> 2451 2452 static int sysctl_kern_somaxkva(SYSCTLFN_PROTO); 2453 static int sysctl_kern_sbmax(SYSCTLFN_PROTO); 2454 2455 /* 2456 * sysctl helper routine for kern.somaxkva. ensures that the given 2457 * value is not too small. 2458 * (XXX should we maybe make sure it's not too large as well?) 2459 */ 2460 static int 2461 sysctl_kern_somaxkva(SYSCTLFN_ARGS) 2462 { 2463 int error, new_somaxkva; 2464 struct sysctlnode node; 2465 2466 new_somaxkva = somaxkva; 2467 node = *rnode; 2468 node.sysctl_data = &new_somaxkva; 2469 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2470 if (error || newp == NULL) 2471 return error; 2472 2473 if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */ 2474 return EINVAL; 2475 2476 mutex_enter(&so_pendfree_lock); 2477 somaxkva = new_somaxkva; 2478 cv_broadcast(&socurkva_cv); 2479 mutex_exit(&so_pendfree_lock); 2480 2481 return error; 2482 } 2483 2484 /* 2485 * sysctl helper routine for kern.sbmax. Basically just ensures that 2486 * any new value is not too small. 2487 */ 2488 static int 2489 sysctl_kern_sbmax(SYSCTLFN_ARGS) 2490 { 2491 int error, new_sbmax; 2492 struct sysctlnode node; 2493 2494 new_sbmax = sb_max; 2495 node = *rnode; 2496 node.sysctl_data = &new_sbmax; 2497 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2498 if (error || newp == NULL) 2499 return error; 2500 2501 KERNEL_LOCK(1, NULL); 2502 error = sb_max_set(new_sbmax); 2503 KERNEL_UNLOCK_ONE(NULL); 2504 2505 return error; 2506 } 2507 2508 /* 2509 * sysctl helper routine for kern.sooptions. Ensures that only allowed 2510 * options can be set. 2511 */ 2512 static int 2513 sysctl_kern_sooptions(SYSCTLFN_ARGS) 2514 { 2515 int error, new_options; 2516 struct sysctlnode node; 2517 2518 new_options = sooptions; 2519 node = *rnode; 2520 node.sysctl_data = &new_options; 2521 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2522 if (error || newp == NULL) 2523 return error; 2524 2525 if (new_options & ~SO_DEFOPTS) 2526 return EINVAL; 2527 2528 sooptions = new_options; 2529 2530 return 0; 2531 } 2532 2533 static void 2534 sysctl_kern_socket_setup(void) 2535 { 2536 2537 KASSERT(socket_sysctllog == NULL); 2538 2539 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2540 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2541 CTLTYPE_INT, "somaxkva", 2542 SYSCTL_DESCR("Maximum amount of kernel memory to be " 2543 "used for socket buffers"), 2544 sysctl_kern_somaxkva, 0, NULL, 0, 2545 CTL_KERN, KERN_SOMAXKVA, CTL_EOL); 2546 2547 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2548 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2549 CTLTYPE_BOOL, "sofixedbuf", 2550 SYSCTL_DESCR("Prevent scaling of fixed socket buffers"), 2551 NULL, 0, &sofixedbuf, 0, 2552 CTL_KERN, KERN_SOFIXEDBUF, CTL_EOL); 2553 2554 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2555 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2556 CTLTYPE_INT, "sbmax", 2557 SYSCTL_DESCR("Maximum socket buffer size"), 2558 sysctl_kern_sbmax, 0, NULL, 0, 2559 CTL_KERN, KERN_SBMAX, CTL_EOL); 2560 2561 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2562 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2563 CTLTYPE_INT, "sooptions", 2564 SYSCTL_DESCR("Default socket options"), 2565 sysctl_kern_sooptions, 0, NULL, 0, 2566 CTL_KERN, CTL_CREATE, CTL_EOL); 2567 } 2568
Indexes created Sat Sep 20 22:09:52 GMT 2025