1 /* $NetBSD: uipc_socket.c,v 1.313 2024/12/06 18:44:00 riastradh Exp $ */ 2 3 /* 4 * Copyright (c) 2002, 2007, 2008, 2009, 2023 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 1.313 2024/12/06 18:44:00 riastradh Exp $"); 75 76 #ifdef _KERNEL_OPT 77 #include "opt_compat_netbsd.h" 78 #include "opt_mbuftrace.h" 79 #include "opt_multiprocessor.h" /* XXX */ 80 #include "opt_pipe.h" 81 #include "opt_sctp.h" 82 #include "opt_sock_counters.h" 83 #include "opt_somaxkva.h" 84 #include "opt_sosend_loan.h" 85 #endif 86 87 #include <sys/param.h> 88 #include <sys/types.h> 89 90 #include <sys/compat_stub.h> 91 #include <sys/condvar.h> 92 #include <sys/domain.h> 93 #include <sys/event.h> 94 #include <sys/file.h> 95 #include <sys/filedesc.h> 96 #include <sys/kauth.h> 97 #include <sys/kernel.h> 98 #include <sys/kmem.h> 99 #include <sys/kthread.h> 100 #include <sys/mbuf.h> 101 #include <sys/mutex.h> 102 #include <sys/poll.h> 103 #include <sys/proc.h> 104 #include <sys/protosw.h> 105 #include <sys/resourcevar.h> 106 #include <sys/sdt.h> 107 #include <sys/signalvar.h> 108 #include <sys/socket.h> 109 #include <sys/socketvar.h> 110 #include <sys/systm.h> 111 #include <sys/uidinfo.h> 112 113 #include <compat/sys/socket.h> 114 #include <compat/sys/time.h> 115 116 #include <uvm/uvm_extern.h> 117 #include <uvm/uvm_loan.h> 118 #include <uvm/uvm_page.h> 119 120 #ifdef SCTP 121 #include <netinet/sctp_route.h> 122 #endif 123 124 MALLOC_DEFINE(M_SONAME, "soname", "socket name"); 125 126 extern const struct fileops socketops; 127 128 static int sooptions; 129 extern int somaxconn; /* patchable (XXX sysctl) */ 130 int somaxconn = SOMAXCONN; 131 kmutex_t *softnet_lock; 132 133 #ifdef SOSEND_COUNTERS 134 #include <sys/device.h> 135 136 static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 137 NULL, "sosend", "loan big"); 138 static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 139 NULL, "sosend", "copy big"); 140 static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 141 NULL, "sosend", "copy small"); 142 static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 143 NULL, "sosend", "kva limit"); 144 145 #define SOSEND_COUNTER_INCR(ev) (ev)->ev_count++ 146 147 EVCNT_ATTACH_STATIC(sosend_loan_big); 148 EVCNT_ATTACH_STATIC(sosend_copy_big); 149 EVCNT_ATTACH_STATIC(sosend_copy_small); 150 EVCNT_ATTACH_STATIC(sosend_kvalimit); 151 #else 152 153 #define SOSEND_COUNTER_INCR(ev) /* nothing */ 154 155 #endif /* SOSEND_COUNTERS */ 156 157 #if defined(SOSEND_NO_LOAN) || defined(MULTIPROCESSOR) 158 int sock_loan_thresh = -1; 159 #else 160 int sock_loan_thresh = 4096; 161 #endif 162 163 static kmutex_t so_pendfree_lock; 164 static struct mbuf *so_pendfree = NULL; 165 166 #ifndef SOMAXKVA 167 #define SOMAXKVA (16 * 1024 * 1024) 168 #endif 169 int somaxkva = SOMAXKVA; 170 static int socurkva; 171 static kcondvar_t socurkva_cv; 172 173 #ifndef SOFIXEDBUF 174 #define SOFIXEDBUF true 175 #endif 176 bool sofixedbuf = SOFIXEDBUF; 177 178 static kauth_listener_t socket_listener; 179 180 #define SOCK_LOAN_CHUNK 65536 181 182 static void sopendfree_thread(void *); 183 static kcondvar_t pendfree_thread_cv; 184 static lwp_t *sopendfree_lwp; 185 186 static void sysctl_kern_socket_setup(void); 187 static struct sysctllog *socket_sysctllog; 188 189 static vsize_t 190 sokvareserve(struct socket *so, vsize_t len) 191 { 192 int error; 193 194 mutex_enter(&so_pendfree_lock); 195 while (socurkva + len > somaxkva) { 196 SOSEND_COUNTER_INCR(&sosend_kvalimit); 197 error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock); 198 if (error) { 199 len = 0; 200 break; 201 } 202 } 203 socurkva += len; 204 mutex_exit(&so_pendfree_lock); 205 return len; 206 } 207 208 static void 209 sokvaunreserve(vsize_t len) 210 { 211 212 mutex_enter(&so_pendfree_lock); 213 socurkva -= len; 214 cv_broadcast(&socurkva_cv); 215 mutex_exit(&so_pendfree_lock); 216 } 217 218 /* 219 * sokvaalloc: allocate kva for loan. 220 */ 221 vaddr_t 222 sokvaalloc(vaddr_t sva, vsize_t len, struct socket *so) 223 { 224 vaddr_t lva; 225 226 if (sokvareserve(so, len) == 0) 227 return 0; 228 229 lva = uvm_km_alloc(kernel_map, len, atop(sva) & uvmexp.colormask, 230 UVM_KMF_COLORMATCH | UVM_KMF_VAONLY | UVM_KMF_WAITVA); 231 if (lva == 0) { 232 sokvaunreserve(len); 233 return 0; 234 } 235 236 return lva; 237 } 238 239 /* 240 * sokvafree: free kva for loan. 241 */ 242 void 243 sokvafree(vaddr_t sva, vsize_t len) 244 { 245 246 uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY); 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 /* 271 * sopendfree_thread: free mbufs on "pendfree" list. Unlock and relock 272 * so_pendfree_lock when freeing mbufs. 273 */ 274 static void 275 sopendfree_thread(void *v) 276 { 277 struct mbuf *m, *next; 278 size_t rv; 279 280 mutex_enter(&so_pendfree_lock); 281 282 for (;;) { 283 rv = 0; 284 while (so_pendfree != NULL) { 285 m = so_pendfree; 286 so_pendfree = NULL; 287 mutex_exit(&so_pendfree_lock); 288 289 for (; m != NULL; m = next) { 290 next = m->m_next; 291 KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 292 0); 293 KASSERT(m->m_ext.ext_refcnt == 0); 294 295 rv += m->m_ext.ext_size; 296 sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf, 297 m->m_ext.ext_size); 298 pool_cache_put(mb_cache, m); 299 } 300 301 mutex_enter(&so_pendfree_lock); 302 } 303 if (rv) 304 cv_broadcast(&socurkva_cv); 305 cv_wait(&pendfree_thread_cv, &so_pendfree_lock); 306 } 307 panic("sopendfree_thread"); 308 /* NOTREACHED */ 309 } 310 311 void 312 soloanfree(struct mbuf *m, void *buf, size_t size, void *arg) 313 { 314 315 KASSERT(m != NULL); 316 317 /* 318 * postpone freeing mbuf. 319 * 320 * we can't do it in interrupt context 321 * because we need to put kva back to kernel_map. 322 */ 323 324 mutex_enter(&so_pendfree_lock); 325 m->m_next = so_pendfree; 326 so_pendfree = m; 327 cv_signal(&pendfree_thread_cv); 328 mutex_exit(&so_pendfree_lock); 329 } 330 331 static long 332 sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space) 333 { 334 struct iovec *iov = uio->uio_iov; 335 vaddr_t sva, eva; 336 vsize_t len; 337 vaddr_t lva; 338 int npgs, error; 339 vaddr_t va; 340 int i; 341 342 if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace)) 343 return 0; 344 345 if (iov->iov_len < (size_t) space) 346 space = iov->iov_len; 347 if (space > SOCK_LOAN_CHUNK) 348 space = SOCK_LOAN_CHUNK; 349 350 eva = round_page((vaddr_t) iov->iov_base + space); 351 sva = trunc_page((vaddr_t) iov->iov_base); 352 len = eva - sva; 353 npgs = len >> PAGE_SHIFT; 354 355 KASSERT(npgs <= M_EXT_MAXPAGES); 356 357 lva = sokvaalloc(sva, len, so); 358 if (lva == 0) 359 return 0; 360 361 error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len, 362 m->m_ext.ext_pgs, UVM_LOAN_TOPAGE); 363 if (error) { 364 sokvafree(lva, len); 365 return 0; 366 } 367 368 for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE) 369 pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]), 370 VM_PROT_READ, 0); 371 pmap_update(pmap_kernel()); 372 373 lva += (vaddr_t) iov->iov_base & PAGE_MASK; 374 375 MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so); 376 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 377 378 uio->uio_resid -= space; 379 /* uio_offset not updated, not set/used for write(2) */ 380 uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space; 381 uio->uio_iov->iov_len -= space; 382 if (uio->uio_iov->iov_len == 0) { 383 uio->uio_iov++; 384 uio->uio_iovcnt--; 385 } 386 387 return space; 388 } 389 390 static int 391 socket_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 392 void *arg0, void *arg1, void *arg2, void *arg3) 393 { 394 int result; 395 enum kauth_network_req req; 396 397 result = KAUTH_RESULT_DEFER; 398 req = (enum kauth_network_req)(uintptr_t)arg0; 399 400 if ((action != KAUTH_NETWORK_SOCKET) && 401 (action != KAUTH_NETWORK_BIND)) 402 return result; 403 404 switch (req) { 405 case KAUTH_REQ_NETWORK_BIND_PORT: 406 result = KAUTH_RESULT_ALLOW; 407 break; 408 409 case KAUTH_REQ_NETWORK_SOCKET_DROP: { 410 /* Normal users can only drop their own connections. */ 411 struct socket *so = (struct socket *)arg1; 412 413 if (so->so_cred && proc_uidmatch(cred, so->so_cred) == 0) 414 result = KAUTH_RESULT_ALLOW; 415 416 break; 417 } 418 419 case KAUTH_REQ_NETWORK_SOCKET_OPEN: 420 /* We allow "raw" routing/bluetooth sockets to anyone. */ 421 switch ((u_long)arg1) { 422 case PF_ROUTE: 423 case PF_OROUTE: 424 case PF_BLUETOOTH: 425 case PF_CAN: 426 result = KAUTH_RESULT_ALLOW; 427 break; 428 default: 429 /* Privileged, let secmodel handle this. */ 430 if ((u_long)arg2 == SOCK_RAW) 431 break; 432 result = KAUTH_RESULT_ALLOW; 433 break; 434 } 435 break; 436 437 case KAUTH_REQ_NETWORK_SOCKET_CANSEE: 438 result = KAUTH_RESULT_ALLOW; 439 440 break; 441 442 default: 443 break; 444 } 445 446 return result; 447 } 448 449 void 450 soinit(void) 451 { 452 453 sysctl_kern_socket_setup(); 454 455 #ifdef SCTP 456 /* Update the SCTP function hooks if necessary*/ 457 458 vec_sctp_add_ip_address = sctp_add_ip_address; 459 vec_sctp_delete_ip_address = sctp_delete_ip_address; 460 #endif 461 462 mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM); 463 softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 464 cv_init(&socurkva_cv, "sokva"); 465 cv_init(&pendfree_thread_cv, "sopendfr"); 466 soinit2(); 467 468 /* Set the initial adjusted socket buffer size. */ 469 if (sb_max_set(sb_max)) 470 panic("bad initial sb_max value: %lu", sb_max); 471 472 socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 473 socket_listener_cb, NULL); 474 } 475 476 void 477 soinit1(void) 478 { 479 int error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL, 480 sopendfree_thread, NULL, &sopendfree_lwp, "sopendfree"); 481 if (error) 482 panic("soinit1 %d", error); 483 } 484 485 /* 486 * socreate: create a new socket of the specified type and the protocol. 487 * 488 * => Caller may specify another socket for lock sharing (must not be held). 489 * => Returns the new socket without lock held. 490 */ 491 int 492 socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l, 493 struct socket *lockso) 494 { 495 const struct protosw *prp; 496 struct socket *so; 497 uid_t uid; 498 int error; 499 kmutex_t *lock; 500 501 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 502 KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type), 503 KAUTH_ARG(proto)); 504 if (error != 0) 505 return error; 506 507 if (proto) 508 prp = pffindproto(dom, proto, type); 509 else 510 prp = pffindtype(dom, type); 511 if (prp == NULL) { 512 /* no support for domain */ 513 if (pffinddomain(dom) == 0) 514 return SET_ERROR(EAFNOSUPPORT); 515 /* no support for socket type */ 516 if (proto == 0 && type != 0) 517 return SET_ERROR(EPROTOTYPE); 518 return SET_ERROR(EPROTONOSUPPORT); 519 } 520 if (prp->pr_usrreqs == NULL) 521 return SET_ERROR(EPROTONOSUPPORT); 522 if (prp->pr_type != type) 523 return SET_ERROR(EPROTOTYPE); 524 525 so = soget(true); 526 so->so_type = type; 527 so->so_proto = prp; 528 so->so_send = sosend; 529 so->so_receive = soreceive; 530 so->so_options = sooptions; 531 #ifdef MBUFTRACE 532 so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner; 533 so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner; 534 so->so_mowner = &prp->pr_domain->dom_mowner; 535 #endif 536 uid = kauth_cred_geteuid(l->l_cred); 537 so->so_uidinfo = uid_find(uid); 538 so->so_egid = kauth_cred_getegid(l->l_cred); 539 so->so_cpid = l->l_proc->p_pid; 540 541 /* 542 * Lock assigned and taken during PCB attach, unless we share 543 * the lock with another socket, e.g. socketpair(2) case. 544 */ 545 if (lockso) { 546 /* 547 * lockso->so_lock should be stable at this point, so 548 * no need for atomic_load_*. 549 */ 550 lock = lockso->so_lock; 551 so->so_lock = lock; 552 mutex_obj_hold(lock); 553 mutex_enter(lock); 554 } 555 556 /* Attach the PCB (returns with the socket lock held). */ 557 error = (*prp->pr_usrreqs->pr_attach)(so, proto); 558 KASSERT(solocked(so)); 559 560 if (error) { 561 KASSERT(so->so_pcb == NULL); 562 so->so_state |= SS_NOFDREF; 563 sofree(so); 564 return error; 565 } 566 so->so_cred = kauth_cred_hold(l->l_cred); 567 sounlock(so); 568 569 *aso = so; 570 return 0; 571 } 572 573 /* 574 * fsocreate: create a socket and a file descriptor associated with it. 575 * Returns the allocated file structure in *fpp, but the descriptor 576 * is not visible yet for the process. 577 * Caller is responsible for calling fd_affix() for the returned *fpp once 578 * it's socket initialization is finished successfully, or fd_abort() if it's 579 * initialization fails. 580 * 581 * 582 * => On success, write file descriptor to *fdout and *fpp and return zero. 583 * => On failure, return non-zero; *fdout and *fpp will be undefined. 584 */ 585 int 586 fsocreate(int domain, struct socket **sop, int type, int proto, int *fdout, 587 file_t **fpp, struct socket *lockso) 588 { 589 lwp_t *l = curlwp; 590 int error, fd, flags; 591 struct socket *so; 592 file_t *fp; 593 594 flags = type & SOCK_FLAGS_MASK; 595 type &= ~SOCK_FLAGS_MASK; 596 error = socreate(domain, &so, type, proto, l, lockso); 597 if (error) { 598 return error; 599 } 600 601 if ((error = fd_allocfile(&fp, &fd)) != 0) { 602 soclose(so); 603 return error; 604 } 605 fd_set_exclose(l, fd, (flags & SOCK_CLOEXEC) != 0); 606 fp->f_flag = FREAD|FWRITE|((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)| 607 ((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0); 608 fp->f_type = DTYPE_SOCKET; 609 fp->f_ops = &socketops; 610 if (flags & SOCK_NONBLOCK) { 611 so->so_state |= SS_NBIO; 612 } 613 fp->f_socket = so; 614 615 if (sop != NULL) { 616 *sop = so; 617 } 618 *fdout = fd; 619 *fpp = fp; 620 return error; 621 } 622 623 int 624 sofamily(const struct socket *so) 625 { 626 const struct protosw *pr; 627 const struct domain *dom; 628 629 if ((pr = so->so_proto) == NULL) 630 return AF_UNSPEC; 631 if ((dom = pr->pr_domain) == NULL) 632 return AF_UNSPEC; 633 return dom->dom_family; 634 } 635 636 int 637 sobind(struct socket *so, struct sockaddr *nam, struct lwp *l) 638 { 639 int error; 640 641 solock(so); 642 if (nam->sa_family != so->so_proto->pr_domain->dom_family) { 643 sounlock(so); 644 return SET_ERROR(EAFNOSUPPORT); 645 } 646 error = (*so->so_proto->pr_usrreqs->pr_bind)(so, nam, l); 647 sounlock(so); 648 return error; 649 } 650 651 int 652 solisten(struct socket *so, int backlog, struct lwp *l) 653 { 654 int error; 655 short oldopt, oldqlimit; 656 657 solock(so); 658 if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | 659 SS_ISDISCONNECTING)) != 0) { 660 sounlock(so); 661 return SET_ERROR(EINVAL); 662 } 663 oldopt = so->so_options; 664 oldqlimit = so->so_qlimit; 665 if (TAILQ_EMPTY(&so->so_q)) 666 so->so_options |= SO_ACCEPTCONN; 667 if (backlog < 0) 668 backlog = 0; 669 so->so_qlimit = uimin(backlog, somaxconn); 670 671 error = (*so->so_proto->pr_usrreqs->pr_listen)(so, l); 672 if (error != 0) { 673 so->so_options = oldopt; 674 so->so_qlimit = oldqlimit; 675 sounlock(so); 676 return error; 677 } 678 sounlock(so); 679 return 0; 680 } 681 682 void 683 sofree(struct socket *so) 684 { 685 u_int refs; 686 687 KASSERT(solocked(so)); 688 689 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) { 690 sounlock(so); 691 return; 692 } 693 if (so->so_head) { 694 /* 695 * We must not decommission a socket that's on the accept(2) 696 * queue. If we do, then accept(2) may hang after select(2) 697 * indicated that the listening socket was ready. 698 */ 699 if (!soqremque(so, 0)) { 700 sounlock(so); 701 return; 702 } 703 } 704 if (so->so_rcv.sb_hiwat) 705 (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, 706 RLIM_INFINITY); 707 if (so->so_snd.sb_hiwat) 708 (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, 709 RLIM_INFINITY); 710 sbrelease(&so->so_snd, so); 711 KASSERT(!cv_has_waiters(&so->so_cv)); 712 KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv)); 713 KASSERT(!cv_has_waiters(&so->so_snd.sb_cv)); 714 sorflush(so); 715 refs = so->so_aborting; /* XXX */ 716 /* Remove accept filter if one is present. */ 717 if (so->so_accf != NULL) 718 (void)accept_filt_clear(so); 719 sounlock(so); 720 if (refs == 0) /* XXX */ 721 soput(so); 722 } 723 724 /* 725 * soclose: close a socket on last file table reference removal. 726 * Initiate disconnect if connected. Free socket when disconnect complete. 727 */ 728 int 729 soclose(struct socket *so) 730 { 731 struct socket *so2; 732 int error = 0; 733 734 solock(so); 735 if (so->so_options & SO_ACCEPTCONN) { 736 for (;;) { 737 if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) { 738 KASSERT(solocked2(so, so2)); 739 (void) soqremque(so2, 0); 740 /* soabort drops the lock. */ 741 (void) soabort(so2); 742 solock(so); 743 continue; 744 } 745 if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) { 746 KASSERT(solocked2(so, so2)); 747 (void) soqremque(so2, 1); 748 /* soabort drops the lock. */ 749 (void) soabort(so2); 750 solock(so); 751 continue; 752 } 753 break; 754 } 755 } 756 if (so->so_pcb == NULL) 757 goto discard; 758 if (so->so_state & SS_ISCONNECTED) { 759 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 760 error = sodisconnect(so); 761 if (error) 762 goto drop; 763 } 764 if (so->so_options & SO_LINGER) { 765 if ((so->so_state & (SS_ISDISCONNECTING|SS_NBIO)) == 766 (SS_ISDISCONNECTING|SS_NBIO)) 767 goto drop; 768 while (so->so_state & SS_ISCONNECTED) { 769 error = sowait(so, true, so->so_linger * hz); 770 if (error) 771 break; 772 } 773 } 774 } 775 drop: 776 if (so->so_pcb) { 777 KASSERT(solocked(so)); 778 (*so->so_proto->pr_usrreqs->pr_detach)(so); 779 } 780 discard: 781 KASSERT((so->so_state & SS_NOFDREF) == 0); 782 kauth_cred_free(so->so_cred); 783 so->so_cred = NULL; 784 so->so_state |= SS_NOFDREF; 785 sofree(so); 786 return error; 787 } 788 789 /* 790 * Must be called with the socket locked.. Will return with it unlocked. 791 */ 792 int 793 soabort(struct socket *so) 794 { 795 u_int refs; 796 int error; 797 798 KASSERT(solocked(so)); 799 KASSERT(so->so_head == NULL); 800 801 so->so_aborting++; /* XXX */ 802 error = (*so->so_proto->pr_usrreqs->pr_abort)(so); 803 refs = --so->so_aborting; /* XXX */ 804 if (error || (refs == 0)) { 805 sofree(so); 806 } else { 807 sounlock(so); 808 } 809 return error; 810 } 811 812 int 813 soaccept(struct socket *so, struct sockaddr *nam) 814 { 815 int error; 816 817 KASSERT(solocked(so)); 818 KASSERT((so->so_state & SS_NOFDREF) != 0); 819 820 so->so_state &= ~SS_NOFDREF; 821 if ((so->so_state & SS_ISDISCONNECTED) == 0 || 822 (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0) 823 error = (*so->so_proto->pr_usrreqs->pr_accept)(so, nam); 824 else 825 error = SET_ERROR(ECONNABORTED); 826 827 return error; 828 } 829 830 int 831 soconnect(struct socket *so, struct sockaddr *nam, struct lwp *l) 832 { 833 int error; 834 835 KASSERT(solocked(so)); 836 837 if (so->so_options & SO_ACCEPTCONN) 838 return SET_ERROR(EOPNOTSUPP); 839 /* 840 * If protocol is connection-based, can only connect once. 841 * Otherwise, if connected, try to disconnect first. 842 * This allows user to disconnect by connecting to, e.g., 843 * a null address. 844 */ 845 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 846 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 847 (error = sodisconnect(so)))) { 848 error = SET_ERROR(EISCONN); 849 } else { 850 if (nam->sa_family != so->so_proto->pr_domain->dom_family) { 851 return SET_ERROR(EAFNOSUPPORT); 852 } 853 error = (*so->so_proto->pr_usrreqs->pr_connect)(so, nam, l); 854 } 855 856 return error; 857 } 858 859 int 860 soconnect2(struct socket *so1, struct socket *so2) 861 { 862 KASSERT(solocked2(so1, so2)); 863 864 return (*so1->so_proto->pr_usrreqs->pr_connect2)(so1, so2); 865 } 866 867 int 868 sodisconnect(struct socket *so) 869 { 870 int error; 871 872 KASSERT(solocked(so)); 873 874 if ((so->so_state & SS_ISCONNECTED) == 0) { 875 error = SET_ERROR(ENOTCONN); 876 } else if (so->so_state & SS_ISDISCONNECTING) { 877 error = SET_ERROR(EALREADY); 878 } else { 879 error = (*so->so_proto->pr_usrreqs->pr_disconnect)(so); 880 } 881 return error; 882 } 883 884 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 885 /* 886 * Send on a socket. 887 * If send must go all at once and message is larger than 888 * send buffering, then hard error. 889 * Lock against other senders. 890 * If must go all at once and not enough room now, then 891 * inform user that this would block and do nothing. 892 * Otherwise, if nonblocking, send as much as possible. 893 * The data to be sent is described by "uio" if nonzero, 894 * otherwise by the mbuf chain "top" (which must be null 895 * if uio is not). Data provided in mbuf chain must be small 896 * enough to send all at once. 897 * 898 * Returns nonzero on error, timeout or signal; callers 899 * must check for short counts if EINTR/ERESTART are returned. 900 * Data and control buffers are freed on return. 901 */ 902 int 903 sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, 904 struct mbuf *top, struct mbuf *control, int flags, struct lwp *l) 905 { 906 struct mbuf **mp, *m; 907 long space, len, resid, clen, mlen; 908 int error, s, dontroute, atomic; 909 short wakeup_state = 0; 910 911 clen = 0; 912 913 /* 914 * solock() provides atomicity of access. splsoftnet() prevents 915 * protocol processing soft interrupts from interrupting us and 916 * blocking (expensive). 917 */ 918 s = splsoftnet(); 919 solock(so); 920 atomic = sosendallatonce(so) || top; 921 if (uio) 922 resid = uio->uio_resid; 923 else 924 resid = top->m_pkthdr.len; 925 /* 926 * In theory resid should be unsigned. 927 * However, space must be signed, as it might be less than 0 928 * if we over-committed, and we must use a signed comparison 929 * of space and resid. On the other hand, a negative resid 930 * causes us to loop sending 0-length segments to the protocol. 931 */ 932 if (resid < 0) { 933 error = SET_ERROR(EINVAL); 934 goto out; 935 } 936 dontroute = 937 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 938 (so->so_proto->pr_flags & PR_ATOMIC); 939 l->l_ru.ru_msgsnd++; 940 if (control) 941 clen = control->m_len; 942 restart: 943 if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) 944 goto out; 945 do { 946 if (so->so_state & SS_CANTSENDMORE) { 947 error = SET_ERROR(EPIPE); 948 goto release; 949 } 950 if (so->so_error) { 951 error = SET_ERROR(so->so_error); 952 if ((flags & MSG_PEEK) == 0) 953 so->so_error = 0; 954 goto release; 955 } 956 if ((so->so_state & SS_ISCONNECTED) == 0) { 957 if (so->so_proto->pr_flags & PR_CONNREQUIRED) { 958 if (resid || clen == 0) { 959 error = SET_ERROR(ENOTCONN); 960 goto release; 961 } 962 } else if (addr == NULL) { 963 error = SET_ERROR(EDESTADDRREQ); 964 goto release; 965 } 966 } 967 space = sbspace(&so->so_snd); 968 if (flags & MSG_OOB) 969 space += 1024; 970 if ((atomic && resid > so->so_snd.sb_hiwat) || 971 clen > so->so_snd.sb_hiwat) { 972 error = SET_ERROR(EMSGSIZE); 973 goto release; 974 } 975 if (space < resid + clen && 976 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 977 if ((so->so_state & SS_NBIO) || (flags & MSG_NBIO)) { 978 error = SET_ERROR(EWOULDBLOCK); 979 goto release; 980 } 981 sbunlock(&so->so_snd); 982 if (wakeup_state & SS_RESTARTSYS) { 983 error = SET_ERROR(ERESTART); 984 goto out; 985 } 986 error = sbwait(&so->so_snd); 987 if (error) 988 goto out; 989 wakeup_state = so->so_state; 990 goto restart; 991 } 992 wakeup_state = 0; 993 mp = ⊤ 994 space -= clen; 995 do { 996 if (uio == NULL) { 997 /* 998 * Data is prepackaged in "top". 999 */ 1000 resid = 0; 1001 if (flags & MSG_EOR) 1002 top->m_flags |= M_EOR; 1003 } else do { 1004 sounlock(so); 1005 splx(s); 1006 if (top == NULL) { 1007 m = m_gethdr(M_WAIT, MT_DATA); 1008 mlen = MHLEN; 1009 m->m_pkthdr.len = 0; 1010 m_reset_rcvif(m); 1011 } else { 1012 m = m_get(M_WAIT, MT_DATA); 1013 mlen = MLEN; 1014 } 1015 MCLAIM(m, so->so_snd.sb_mowner); 1016 if (sock_loan_thresh >= 0 && 1017 uio->uio_iov->iov_len >= sock_loan_thresh && 1018 space >= sock_loan_thresh && 1019 (len = sosend_loan(so, uio, m, 1020 space)) != 0) { 1021 SOSEND_COUNTER_INCR(&sosend_loan_big); 1022 space -= len; 1023 goto have_data; 1024 } 1025 if (resid >= MINCLSIZE && space >= MCLBYTES) { 1026 SOSEND_COUNTER_INCR(&sosend_copy_big); 1027 m_clget(m, M_DONTWAIT); 1028 if ((m->m_flags & M_EXT) == 0) 1029 goto nopages; 1030 mlen = MCLBYTES; 1031 if (atomic && top == 0) { 1032 len = lmin(MCLBYTES - max_hdr, 1033 resid); 1034 m->m_data += max_hdr; 1035 } else 1036 len = lmin(MCLBYTES, resid); 1037 space -= len; 1038 } else { 1039 nopages: 1040 SOSEND_COUNTER_INCR(&sosend_copy_small); 1041 len = lmin(lmin(mlen, resid), space); 1042 space -= len; 1043 /* 1044 * For datagram protocols, leave room 1045 * for protocol headers in first mbuf. 1046 */ 1047 if (atomic && top == 0 && len < mlen) 1048 m_align(m, len); 1049 } 1050 error = uiomove(mtod(m, void *), (int)len, uio); 1051 have_data: 1052 resid = uio->uio_resid; 1053 m->m_len = len; 1054 *mp = m; 1055 top->m_pkthdr.len += len; 1056 s = splsoftnet(); 1057 solock(so); 1058 if (error != 0) 1059 goto release; 1060 mp = &m->m_next; 1061 if (resid <= 0) { 1062 if (flags & MSG_EOR) 1063 top->m_flags |= M_EOR; 1064 break; 1065 } 1066 } while (space > 0 && atomic); 1067 1068 if (so->so_state & SS_CANTSENDMORE) { 1069 error = SET_ERROR(EPIPE); 1070 goto release; 1071 } 1072 if (dontroute) 1073 so->so_options |= SO_DONTROUTE; 1074 if (resid > 0) 1075 so->so_state |= SS_MORETOCOME; 1076 if (flags & MSG_OOB) { 1077 error = (*so->so_proto->pr_usrreqs->pr_sendoob)( 1078 so, top, control); 1079 } else { 1080 error = (*so->so_proto->pr_usrreqs->pr_send)(so, 1081 top, addr, control, l); 1082 } 1083 if (dontroute) 1084 so->so_options &= ~SO_DONTROUTE; 1085 if (resid > 0) 1086 so->so_state &= ~SS_MORETOCOME; 1087 clen = 0; 1088 control = NULL; 1089 top = NULL; 1090 mp = ⊤ 1091 if (error != 0) 1092 goto release; 1093 } while (resid && space > 0); 1094 } while (resid); 1095 1096 release: 1097 sbunlock(&so->so_snd); 1098 out: 1099 sounlock(so); 1100 splx(s); 1101 m_freem(top); 1102 m_freem(control); 1103 return error; 1104 } 1105 1106 /* 1107 * Following replacement or removal of the first mbuf on the first 1108 * mbuf chain of a socket buffer, push necessary state changes back 1109 * into the socket buffer so that other consumers see the values 1110 * consistently. 'nextrecord' is the caller's locally stored value of 1111 * the original value of sb->sb_mb->m_nextpkt which must be restored 1112 * when the lead mbuf changes. NOTE: 'nextrecord' may be NULL. 1113 */ 1114 static void 1115 sbsync(struct sockbuf *sb, struct mbuf *nextrecord) 1116 { 1117 1118 KASSERT(solocked(sb->sb_so)); 1119 1120 /* 1121 * First, update for the new value of nextrecord. If necessary, 1122 * make it the first record. 1123 */ 1124 if (sb->sb_mb != NULL) 1125 sb->sb_mb->m_nextpkt = nextrecord; 1126 else 1127 sb->sb_mb = nextrecord; 1128 1129 /* 1130 * Now update any dependent socket buffer fields to reflect 1131 * the new state. This is an inline of SB_EMPTY_FIXUP, with 1132 * the addition of a second clause that takes care of the 1133 * case where sb_mb has been updated, but remains the last 1134 * record. 1135 */ 1136 if (sb->sb_mb == NULL) { 1137 sb->sb_mbtail = NULL; 1138 sb->sb_lastrecord = NULL; 1139 } else if (sb->sb_mb->m_nextpkt == NULL) 1140 sb->sb_lastrecord = sb->sb_mb; 1141 } 1142 1143 /* 1144 * Implement receive operations on a socket. 1145 * 1146 * We depend on the way that records are added to the sockbuf by sbappend*. In 1147 * particular, each record (mbufs linked through m_next) must begin with an 1148 * address if the protocol so specifies, followed by an optional mbuf or mbufs 1149 * containing ancillary data, and then zero or more mbufs of data. 1150 * 1151 * In order to avoid blocking network interrupts for the entire time here, we 1152 * splx() while doing the actual copy to user space. Although the sockbuf is 1153 * locked, new data may still be appended, and thus we must maintain 1154 * consistency of the sockbuf during that time. 1155 * 1156 * The caller may receive the data as a single mbuf chain by supplying an mbuf 1157 * **mp0 for use in returning the chain. The uio is then used only for the 1158 * count in uio_resid. 1159 */ 1160 int 1161 soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, 1162 struct mbuf **mp0, struct mbuf **controlp, int *flagsp) 1163 { 1164 struct lwp *l = curlwp; 1165 struct mbuf *m, **mp, *mt; 1166 size_t len, offset, moff, orig_resid; 1167 int atomic, flags, error, s, type; 1168 const struct protosw *pr; 1169 struct mbuf *nextrecord; 1170 int mbuf_removed = 0; 1171 const struct domain *dom; 1172 short wakeup_state = 0; 1173 1174 pr = so->so_proto; 1175 atomic = pr->pr_flags & PR_ATOMIC; 1176 dom = pr->pr_domain; 1177 mp = mp0; 1178 type = 0; 1179 orig_resid = uio->uio_resid; 1180 1181 if (paddr != NULL) 1182 *paddr = NULL; 1183 if (controlp != NULL) 1184 *controlp = NULL; 1185 if (flagsp != NULL) 1186 flags = *flagsp &~ MSG_EOR; 1187 else 1188 flags = 0; 1189 1190 if (flags & MSG_OOB) { 1191 m = m_get(M_WAIT, MT_DATA); 1192 solock(so); 1193 error = (*pr->pr_usrreqs->pr_recvoob)(so, m, flags & MSG_PEEK); 1194 sounlock(so); 1195 if (error) 1196 goto bad; 1197 do { 1198 error = uiomove(mtod(m, void *), 1199 MIN(uio->uio_resid, m->m_len), uio); 1200 m = m_free(m); 1201 } while (uio->uio_resid > 0 && error == 0 && m); 1202 bad: 1203 m_freem(m); 1204 return error; 1205 } 1206 if (mp != NULL) 1207 *mp = NULL; 1208 1209 /* 1210 * solock() provides atomicity of access. splsoftnet() prevents 1211 * protocol processing soft interrupts from interrupting us and 1212 * blocking (expensive). 1213 */ 1214 s = splsoftnet(); 1215 solock(so); 1216 restart: 1217 if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) { 1218 sounlock(so); 1219 splx(s); 1220 return error; 1221 } 1222 m = so->so_rcv.sb_mb; 1223 1224 /* 1225 * If we have less data than requested, block awaiting more 1226 * (subject to any timeout) if: 1227 * 1. the current count is less than the low water mark, 1228 * 2. MSG_WAITALL is set, and it is possible to do the entire 1229 * receive operation at once if we block (resid <= hiwat), or 1230 * 3. MSG_DONTWAIT is not set. 1231 * If MSG_WAITALL is set but resid is larger than the receive buffer, 1232 * we have to do the receive in sections, and thus risk returning 1233 * a short count if a timeout or signal occurs after we start. 1234 */ 1235 if (m == NULL || 1236 ((flags & MSG_DONTWAIT) == 0 && 1237 so->so_rcv.sb_cc < uio->uio_resid && 1238 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 1239 ((flags & MSG_WAITALL) && 1240 uio->uio_resid <= so->so_rcv.sb_hiwat)) && 1241 m->m_nextpkt == NULL && !atomic)) { 1242 #ifdef DIAGNOSTIC 1243 if (m == NULL && so->so_rcv.sb_cc) 1244 panic("receive 1"); 1245 #endif 1246 if (so->so_error || so->so_rerror) { 1247 u_short *e; 1248 if (m != NULL) 1249 goto dontblock; 1250 e = so->so_error ? &so->so_error : &so->so_rerror; 1251 error = SET_ERROR(*e); 1252 if ((flags & MSG_PEEK) == 0) 1253 *e = 0; 1254 goto release; 1255 } 1256 if (so->so_state & SS_CANTRCVMORE) { 1257 if (m != NULL) 1258 goto dontblock; 1259 else 1260 goto release; 1261 } 1262 for (; m != NULL; m = m->m_next) 1263 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 1264 m = so->so_rcv.sb_mb; 1265 goto dontblock; 1266 } 1267 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 1268 (so->so_proto->pr_flags & PR_CONNREQUIRED)) { 1269 error = SET_ERROR(ENOTCONN); 1270 goto release; 1271 } 1272 if (uio->uio_resid == 0) 1273 goto release; 1274 if ((so->so_state & SS_NBIO) || 1275 (flags & (MSG_DONTWAIT|MSG_NBIO))) { 1276 error = SET_ERROR(EWOULDBLOCK); 1277 goto release; 1278 } 1279 SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1"); 1280 SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1"); 1281 sbunlock(&so->so_rcv); 1282 if (wakeup_state & SS_RESTARTSYS) 1283 error = SET_ERROR(ERESTART); 1284 else 1285 error = sbwait(&so->so_rcv); 1286 if (error != 0) { 1287 sounlock(so); 1288 splx(s); 1289 return error; 1290 } 1291 wakeup_state = so->so_state; 1292 goto restart; 1293 } 1294 1295 dontblock: 1296 /* 1297 * On entry here, m points to the first record of the socket buffer. 1298 * From this point onward, we maintain 'nextrecord' as a cache of the 1299 * pointer to the next record in the socket buffer. We must keep the 1300 * various socket buffer pointers and local stack versions of the 1301 * pointers in sync, pushing out modifications before dropping the 1302 * socket lock, and re-reading them when picking it up. 1303 * 1304 * Otherwise, we will race with the network stack appending new data 1305 * or records onto the socket buffer by using inconsistent/stale 1306 * versions of the field, possibly resulting in socket buffer 1307 * corruption. 1308 * 1309 * By holding the high-level sblock(), we prevent simultaneous 1310 * readers from pulling off the front of the socket buffer. 1311 */ 1312 if (l != NULL) 1313 l->l_ru.ru_msgrcv++; 1314 KASSERT(m == so->so_rcv.sb_mb); 1315 SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); 1316 SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); 1317 nextrecord = m->m_nextpkt; 1318 1319 if (pr->pr_flags & PR_ADDR) { 1320 KASSERT(m->m_type == MT_SONAME); 1321 orig_resid = 0; 1322 if (flags & MSG_PEEK) { 1323 if (paddr) 1324 *paddr = m_copym(m, 0, m->m_len, M_DONTWAIT); 1325 m = m->m_next; 1326 } else { 1327 sbfree(&so->so_rcv, m); 1328 mbuf_removed = 1; 1329 if (paddr != NULL) { 1330 *paddr = m; 1331 so->so_rcv.sb_mb = m->m_next; 1332 m->m_next = NULL; 1333 m = so->so_rcv.sb_mb; 1334 } else { 1335 m = so->so_rcv.sb_mb = m_free(m); 1336 } 1337 sbsync(&so->so_rcv, nextrecord); 1338 } 1339 } 1340 1341 if (pr->pr_flags & PR_ADDR_OPT) { 1342 /* 1343 * For SCTP we may be getting a whole message OR a partial 1344 * delivery. 1345 */ 1346 if (m->m_type == MT_SONAME) { 1347 orig_resid = 0; 1348 if (flags & MSG_PEEK) { 1349 if (paddr) 1350 *paddr = m_copym(m, 0, m->m_len, M_DONTWAIT); 1351 m = m->m_next; 1352 } else { 1353 sbfree(&so->so_rcv, m); 1354 mbuf_removed = 1; 1355 if (paddr) { 1356 *paddr = m; 1357 so->so_rcv.sb_mb = m->m_next; 1358 m->m_next = 0; 1359 m = so->so_rcv.sb_mb; 1360 } else { 1361 m = so->so_rcv.sb_mb = m_free(m); 1362 } 1363 sbsync(&so->so_rcv, nextrecord); 1364 } 1365 } 1366 } 1367 1368 /* 1369 * Process one or more MT_CONTROL mbufs present before any data mbufs 1370 * in the first mbuf chain on the socket buffer. If MSG_PEEK, we 1371 * just copy the data; if !MSG_PEEK, we call into the protocol to 1372 * perform externalization (or freeing if controlp == NULL). 1373 */ 1374 if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) { 1375 struct mbuf *cm = NULL, *cmn; 1376 struct mbuf **cme = &cm; 1377 1378 do { 1379 if (flags & MSG_PEEK) { 1380 if (controlp != NULL) { 1381 *controlp = m_copym(m, 0, m->m_len, M_DONTWAIT); 1382 controlp = (*controlp == NULL ? NULL : 1383 &(*controlp)->m_next); 1384 } 1385 m = m->m_next; 1386 } else { 1387 sbfree(&so->so_rcv, m); 1388 so->so_rcv.sb_mb = m->m_next; 1389 m->m_next = NULL; 1390 *cme = m; 1391 cme = &(*cme)->m_next; 1392 m = so->so_rcv.sb_mb; 1393 } 1394 } while (m != NULL && m->m_type == MT_CONTROL); 1395 if ((flags & MSG_PEEK) == 0) 1396 sbsync(&so->so_rcv, nextrecord); 1397 1398 for (; cm != NULL; cm = cmn) { 1399 cmn = cm->m_next; 1400 cm->m_next = NULL; 1401 type = mtod(cm, struct cmsghdr *)->cmsg_type; 1402 if (controlp != NULL) { 1403 if (dom->dom_externalize != NULL && 1404 type == SCM_RIGHTS) { 1405 sounlock(so); 1406 splx(s); 1407 error = (*dom->dom_externalize)(cm, l, 1408 (flags & MSG_CMSG_CLOEXEC) ? 1409 O_CLOEXEC : 0); 1410 s = splsoftnet(); 1411 solock(so); 1412 } 1413 *controlp = cm; 1414 while (*controlp != NULL) 1415 controlp = &(*controlp)->m_next; 1416 } else { 1417 /* 1418 * Dispose of any SCM_RIGHTS message that went 1419 * through the read path rather than recv. 1420 */ 1421 if (dom->dom_dispose != NULL && 1422 type == SCM_RIGHTS) { 1423 sounlock(so); 1424 (*dom->dom_dispose)(cm); 1425 solock(so); 1426 } 1427 m_freem(cm); 1428 } 1429 } 1430 if (m != NULL) 1431 nextrecord = so->so_rcv.sb_mb->m_nextpkt; 1432 else 1433 nextrecord = so->so_rcv.sb_mb; 1434 orig_resid = 0; 1435 } 1436 1437 /* If m is non-NULL, we have some data to read. */ 1438 if (__predict_true(m != NULL)) { 1439 type = m->m_type; 1440 if (type == MT_OOBDATA) 1441 flags |= MSG_OOB; 1442 } 1443 SBLASTRECORDCHK(&so->so_rcv, "soreceive 2"); 1444 SBLASTMBUFCHK(&so->so_rcv, "soreceive 2"); 1445 1446 moff = 0; 1447 offset = 0; 1448 while (m != NULL && uio->uio_resid > 0 && error == 0) { 1449 /* 1450 * If the type of mbuf has changed, end the receive 1451 * operation and do a short read. 1452 */ 1453 if (m->m_type == MT_OOBDATA) { 1454 if (type != MT_OOBDATA) 1455 break; 1456 } else if (type == MT_OOBDATA) { 1457 break; 1458 } else if (m->m_type == MT_CONTROL) { 1459 break; 1460 } 1461 #ifdef DIAGNOSTIC 1462 else if (m->m_type != MT_DATA && m->m_type != MT_HEADER) { 1463 panic("%s: m_type=%d", __func__, m->m_type); 1464 } 1465 #endif 1466 1467 so->so_state &= ~SS_RCVATMARK; 1468 wakeup_state = 0; 1469 len = uio->uio_resid; 1470 if (so->so_oobmark && len > so->so_oobmark - offset) 1471 len = so->so_oobmark - offset; 1472 if (len > m->m_len - moff) 1473 len = m->m_len - moff; 1474 1475 /* 1476 * If mp is set, just pass back the mbufs. 1477 * Otherwise copy them out via the uio, then free. 1478 * Sockbuf must be consistent here (points to current mbuf, 1479 * it points to next record) when we drop priority; 1480 * we must note any additions to the sockbuf when we 1481 * block interrupts again. 1482 */ 1483 if (mp == NULL) { 1484 SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); 1485 SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); 1486 sounlock(so); 1487 splx(s); 1488 error = uiomove(mtod(m, char *) + moff, len, uio); 1489 s = splsoftnet(); 1490 solock(so); 1491 if (error != 0) { 1492 /* 1493 * If any part of the record has been removed 1494 * (such as the MT_SONAME mbuf, which will 1495 * happen when PR_ADDR, and thus also 1496 * PR_ATOMIC, is set), then drop the entire 1497 * record to maintain the atomicity of the 1498 * receive operation. 1499 * 1500 * This avoids a later panic("receive 1a") 1501 * when compiled with DIAGNOSTIC. 1502 */ 1503 if (m && mbuf_removed && atomic) 1504 (void) sbdroprecord(&so->so_rcv); 1505 1506 goto release; 1507 } 1508 } else { 1509 uio->uio_resid -= len; 1510 } 1511 1512 if (len == m->m_len - moff) { 1513 if (m->m_flags & M_EOR) 1514 flags |= MSG_EOR; 1515 #ifdef SCTP 1516 if (m->m_flags & M_NOTIFICATION) 1517 flags |= MSG_NOTIFICATION; 1518 #endif 1519 if (flags & MSG_PEEK) { 1520 m = m->m_next; 1521 moff = 0; 1522 } else { 1523 nextrecord = m->m_nextpkt; 1524 sbfree(&so->so_rcv, m); 1525 if (mp) { 1526 *mp = m; 1527 mp = &m->m_next; 1528 so->so_rcv.sb_mb = m = m->m_next; 1529 *mp = NULL; 1530 } else { 1531 m = so->so_rcv.sb_mb = m_free(m); 1532 } 1533 /* 1534 * If m != NULL, we also know that 1535 * so->so_rcv.sb_mb != NULL. 1536 */ 1537 KASSERT(so->so_rcv.sb_mb == m); 1538 if (m) { 1539 m->m_nextpkt = nextrecord; 1540 if (nextrecord == NULL) 1541 so->so_rcv.sb_lastrecord = m; 1542 } else { 1543 so->so_rcv.sb_mb = nextrecord; 1544 SB_EMPTY_FIXUP(&so->so_rcv); 1545 } 1546 SBLASTRECORDCHK(&so->so_rcv, "soreceive 3"); 1547 SBLASTMBUFCHK(&so->so_rcv, "soreceive 3"); 1548 } 1549 } else if (flags & MSG_PEEK) { 1550 moff += len; 1551 } else { 1552 if (mp != NULL) { 1553 mt = m_copym(m, 0, len, M_NOWAIT); 1554 if (__predict_false(mt == NULL)) { 1555 sounlock(so); 1556 mt = m_copym(m, 0, len, M_WAIT); 1557 solock(so); 1558 } 1559 *mp = mt; 1560 } 1561 m->m_data += len; 1562 m->m_len -= len; 1563 so->so_rcv.sb_cc -= len; 1564 } 1565 1566 if (so->so_oobmark) { 1567 if ((flags & MSG_PEEK) == 0) { 1568 so->so_oobmark -= len; 1569 if (so->so_oobmark == 0) { 1570 so->so_state |= SS_RCVATMARK; 1571 break; 1572 } 1573 } else { 1574 offset += len; 1575 if (offset == so->so_oobmark) 1576 break; 1577 } 1578 } else { 1579 so->so_state &= ~SS_POLLRDBAND; 1580 } 1581 if (flags & MSG_EOR) 1582 break; 1583 1584 /* 1585 * If the MSG_WAITALL flag is set (for non-atomic socket), 1586 * we must not quit until "uio->uio_resid == 0" or an error 1587 * termination. If a signal/timeout occurs, return 1588 * with a short count but without error. 1589 * Keep sockbuf locked against other readers. 1590 */ 1591 while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 && 1592 !sosendallatonce(so) && !nextrecord) { 1593 if (so->so_error || so->so_rerror || 1594 so->so_state & SS_CANTRCVMORE) 1595 break; 1596 /* 1597 * If we are peeking and the socket receive buffer is 1598 * full, stop since we can't get more data to peek at. 1599 */ 1600 if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0) 1601 break; 1602 /* 1603 * If we've drained the socket buffer, tell the 1604 * protocol in case it needs to do something to 1605 * get it filled again. 1606 */ 1607 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) 1608 (*pr->pr_usrreqs->pr_rcvd)(so, flags, l); 1609 SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2"); 1610 SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2"); 1611 if (wakeup_state & SS_RESTARTSYS) 1612 error = SET_ERROR(ERESTART); 1613 else 1614 error = sbwait(&so->so_rcv); 1615 if (error != 0) { 1616 sbunlock(&so->so_rcv); 1617 sounlock(so); 1618 splx(s); 1619 return 0; 1620 } 1621 if ((m = so->so_rcv.sb_mb) != NULL) 1622 nextrecord = m->m_nextpkt; 1623 wakeup_state = so->so_state; 1624 } 1625 } 1626 1627 if (m && atomic) { 1628 flags |= MSG_TRUNC; 1629 if ((flags & MSG_PEEK) == 0) 1630 (void) sbdroprecord(&so->so_rcv); 1631 } 1632 if ((flags & MSG_PEEK) == 0) { 1633 if (m == NULL) { 1634 /* 1635 * First part is an inline SB_EMPTY_FIXUP(). Second 1636 * part makes sure sb_lastrecord is up-to-date if 1637 * there is still data in the socket buffer. 1638 */ 1639 so->so_rcv.sb_mb = nextrecord; 1640 if (so->so_rcv.sb_mb == NULL) { 1641 so->so_rcv.sb_mbtail = NULL; 1642 so->so_rcv.sb_lastrecord = NULL; 1643 } else if (nextrecord->m_nextpkt == NULL) 1644 so->so_rcv.sb_lastrecord = nextrecord; 1645 } 1646 SBLASTRECORDCHK(&so->so_rcv, "soreceive 4"); 1647 SBLASTMBUFCHK(&so->so_rcv, "soreceive 4"); 1648 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1649 (*pr->pr_usrreqs->pr_rcvd)(so, flags, l); 1650 } 1651 if (orig_resid == uio->uio_resid && orig_resid && 1652 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 1653 sbunlock(&so->so_rcv); 1654 goto restart; 1655 } 1656 1657 if (flagsp != NULL) 1658 *flagsp |= flags; 1659 release: 1660 sbunlock(&so->so_rcv); 1661 sounlock(so); 1662 splx(s); 1663 return error; 1664 } 1665 1666 int 1667 soshutdown(struct socket *so, int how) 1668 { 1669 const struct protosw *pr; 1670 int error; 1671 1672 KASSERT(solocked(so)); 1673 1674 pr = so->so_proto; 1675 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 1676 return SET_ERROR(EINVAL); 1677 1678 if (how == SHUT_RD || how == SHUT_RDWR) { 1679 sorflush(so); 1680 error = 0; 1681 } 1682 if (how == SHUT_WR || how == SHUT_RDWR) 1683 error = (*pr->pr_usrreqs->pr_shutdown)(so); 1684 1685 return error; 1686 } 1687 1688 void 1689 sorestart(struct socket *so) 1690 { 1691 /* 1692 * An application has called close() on an fd on which another 1693 * of its threads has called a socket system call. 1694 * Mark this and wake everyone up, and code that would block again 1695 * instead returns ERESTART. 1696 * On system call re-entry the fd is validated and EBADF returned. 1697 * Any other fd will block again on the 2nd syscall. 1698 */ 1699 solock(so); 1700 so->so_state |= SS_RESTARTSYS; 1701 cv_broadcast(&so->so_cv); 1702 cv_broadcast(&so->so_snd.sb_cv); 1703 cv_broadcast(&so->so_rcv.sb_cv); 1704 sounlock(so); 1705 } 1706 1707 void 1708 sorflush(struct socket *so) 1709 { 1710 struct sockbuf *sb, asb; 1711 const struct protosw *pr; 1712 1713 KASSERT(solocked(so)); 1714 1715 sb = &so->so_rcv; 1716 pr = so->so_proto; 1717 socantrcvmore(so); 1718 sb->sb_flags |= SB_NOINTR; 1719 (void )sblock(sb, M_WAITOK); 1720 sbunlock(sb); 1721 asb = *sb; 1722 /* 1723 * Clear most of the sockbuf structure, but leave some of the 1724 * fields valid. 1725 */ 1726 memset(&sb->sb_startzero, 0, 1727 sizeof(*sb) - offsetof(struct sockbuf, sb_startzero)); 1728 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) { 1729 sounlock(so); 1730 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 1731 solock(so); 1732 } 1733 sbrelease(&asb, so); 1734 } 1735 1736 /* 1737 * internal set SOL_SOCKET options 1738 */ 1739 static int 1740 sosetopt1(struct socket *so, const struct sockopt *sopt) 1741 { 1742 int error, opt; 1743 int optval = 0; /* XXX: gcc */ 1744 struct linger l; 1745 struct timeval tv; 1746 1747 opt = sopt->sopt_name; 1748 1749 switch (opt) { 1750 1751 case SO_ACCEPTFILTER: 1752 error = accept_filt_setopt(so, sopt); 1753 KASSERT(solocked(so)); 1754 break; 1755 1756 case SO_LINGER: 1757 error = sockopt_get(sopt, &l, sizeof(l)); 1758 solock(so); 1759 if (error) 1760 break; 1761 if (l.l_linger < 0 || l.l_linger > USHRT_MAX || 1762 l.l_linger > (INT_MAX / hz)) { 1763 error = SET_ERROR(EDOM); 1764 break; 1765 } 1766 so->so_linger = l.l_linger; 1767 if (l.l_onoff) 1768 so->so_options |= SO_LINGER; 1769 else 1770 so->so_options &= ~SO_LINGER; 1771 break; 1772 1773 case SO_DEBUG: 1774 case SO_KEEPALIVE: 1775 case SO_DONTROUTE: 1776 case SO_USELOOPBACK: 1777 case SO_BROADCAST: 1778 case SO_REUSEADDR: 1779 case SO_REUSEPORT: 1780 case SO_OOBINLINE: 1781 case SO_TIMESTAMP: 1782 case SO_NOSIGPIPE: 1783 case SO_RERROR: 1784 error = sockopt_getint(sopt, &optval); 1785 solock(so); 1786 if (error) 1787 break; 1788 if (optval) 1789 so->so_options |= opt; 1790 else 1791 so->so_options &= ~opt; 1792 break; 1793 1794 case SO_SNDBUF: 1795 case SO_RCVBUF: 1796 case SO_SNDLOWAT: 1797 case SO_RCVLOWAT: 1798 error = sockopt_getint(sopt, &optval); 1799 solock(so); 1800 if (error) 1801 break; 1802 1803 /* 1804 * Values < 1 make no sense for any of these 1805 * options, so disallow them. 1806 */ 1807 if (optval < 1) { 1808 error = SET_ERROR(EINVAL); 1809 break; 1810 } 1811 1812 switch (opt) { 1813 case SO_SNDBUF: 1814 if (sbreserve(&so->so_snd, (u_long)optval, so) == 0) { 1815 error = SET_ERROR(ENOBUFS); 1816 break; 1817 } 1818 if (sofixedbuf) 1819 so->so_snd.sb_flags &= ~SB_AUTOSIZE; 1820 break; 1821 1822 case SO_RCVBUF: 1823 if (sbreserve(&so->so_rcv, (u_long)optval, so) == 0) { 1824 error = SET_ERROR(ENOBUFS); 1825 break; 1826 } 1827 if (sofixedbuf) 1828 so->so_rcv.sb_flags &= ~SB_AUTOSIZE; 1829 break; 1830 1831 /* 1832 * Make sure the low-water is never greater than 1833 * the high-water. 1834 */ 1835 case SO_SNDLOWAT: 1836 if (optval > so->so_snd.sb_hiwat) 1837 optval = so->so_snd.sb_hiwat; 1838 1839 so->so_snd.sb_lowat = optval; 1840 break; 1841 1842 case SO_RCVLOWAT: 1843 if (optval > so->so_rcv.sb_hiwat) 1844 optval = so->so_rcv.sb_hiwat; 1845 1846 so->so_rcv.sb_lowat = optval; 1847 break; 1848 } 1849 break; 1850 1851 case SO_SNDTIMEO: 1852 case SO_RCVTIMEO: 1853 solock(so); 1854 error = sockopt_get(sopt, &tv, sizeof(tv)); 1855 if (error) 1856 break; 1857 1858 if (tv.tv_sec < 0 || tv.tv_usec < 0 || tv.tv_usec >= 1000000) { 1859 error = SET_ERROR(EDOM); 1860 break; 1861 } 1862 if (tv.tv_sec > (INT_MAX - tv.tv_usec / tick) / hz) { 1863 error = SET_ERROR(EDOM); 1864 break; 1865 } 1866 1867 optval = tv.tv_sec * hz + tv.tv_usec / tick; 1868 if (optval == 0 && tv.tv_usec != 0) 1869 optval = 1; 1870 1871 switch (opt) { 1872 case SO_SNDTIMEO: 1873 so->so_snd.sb_timeo = optval; 1874 break; 1875 case SO_RCVTIMEO: 1876 so->so_rcv.sb_timeo = optval; 1877 break; 1878 } 1879 break; 1880 1881 default: 1882 MODULE_HOOK_CALL(uipc_socket_50_setopt1_hook, 1883 (opt, so, sopt), enosys(), error); 1884 if (error == ENOSYS || error == EPASSTHROUGH) { 1885 solock(so); 1886 error = SET_ERROR(ENOPROTOOPT); 1887 } 1888 break; 1889 } 1890 KASSERT(solocked(so)); 1891 return error; 1892 } 1893 1894 int 1895 sosetopt(struct socket *so, struct sockopt *sopt) 1896 { 1897 int error, prerr; 1898 1899 if (sopt->sopt_level == SOL_SOCKET) { 1900 error = sosetopt1(so, sopt); 1901 KASSERT(solocked(so)); 1902 } else { 1903 error = SET_ERROR(ENOPROTOOPT); 1904 solock(so); 1905 } 1906 1907 if ((error == 0 || error == ENOPROTOOPT) && 1908 so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) { 1909 /* give the protocol stack a shot */ 1910 prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, sopt); 1911 if (prerr == 0) 1912 error = 0; 1913 else if (prerr != ENOPROTOOPT) 1914 error = prerr; 1915 } 1916 sounlock(so); 1917 return error; 1918 } 1919 1920 /* 1921 * so_setsockopt() is a wrapper providing a sockopt structure for sosetopt() 1922 */ 1923 int 1924 so_setsockopt(struct lwp *l, struct socket *so, int level, int name, 1925 const void *val, size_t valsize) 1926 { 1927 struct sockopt sopt; 1928 int error; 1929 1930 KASSERT(valsize == 0 || val != NULL); 1931 1932 sockopt_init(&sopt, level, name, valsize); 1933 sockopt_set(&sopt, val, valsize); 1934 1935 error = sosetopt(so, &sopt); 1936 1937 sockopt_destroy(&sopt); 1938 1939 return error; 1940 } 1941 1942 /* 1943 * internal get SOL_SOCKET options 1944 */ 1945 static int 1946 sogetopt1(struct socket *so, struct sockopt *sopt) 1947 { 1948 int error, optval, opt; 1949 struct linger l; 1950 struct timeval tv; 1951 1952 switch ((opt = sopt->sopt_name)) { 1953 1954 case SO_ACCEPTFILTER: 1955 error = accept_filt_getopt(so, sopt); 1956 break; 1957 1958 case SO_LINGER: 1959 l.l_onoff = (so->so_options & SO_LINGER) ? 1 : 0; 1960 l.l_linger = so->so_linger; 1961 1962 error = sockopt_set(sopt, &l, sizeof(l)); 1963 break; 1964 1965 case SO_USELOOPBACK: 1966 case SO_DONTROUTE: 1967 case SO_DEBUG: 1968 case SO_KEEPALIVE: 1969 case SO_REUSEADDR: 1970 case SO_REUSEPORT: 1971 case SO_BROADCAST: 1972 case SO_OOBINLINE: 1973 case SO_TIMESTAMP: 1974 case SO_NOSIGPIPE: 1975 case SO_RERROR: 1976 case SO_ACCEPTCONN: 1977 error = sockopt_setint(sopt, (so->so_options & opt) ? 1 : 0); 1978 break; 1979 1980 case SO_TYPE: 1981 error = sockopt_setint(sopt, so->so_type); 1982 break; 1983 1984 case SO_ERROR: 1985 if (so->so_error == 0) { 1986 so->so_error = so->so_rerror; 1987 so->so_rerror = 0; 1988 } 1989 error = sockopt_setint(sopt, so->so_error); 1990 so->so_error = 0; 1991 break; 1992 1993 case SO_SNDBUF: 1994 error = sockopt_setint(sopt, so->so_snd.sb_hiwat); 1995 break; 1996 1997 case SO_RCVBUF: 1998 error = sockopt_setint(sopt, so->so_rcv.sb_hiwat); 1999 break; 2000 2001 case SO_SNDLOWAT: 2002 error = sockopt_setint(sopt, so->so_snd.sb_lowat); 2003 break; 2004 2005 case SO_RCVLOWAT: 2006 error = sockopt_setint(sopt, so->so_rcv.sb_lowat); 2007 break; 2008 2009 case SO_SNDTIMEO: 2010 case SO_RCVTIMEO: 2011 optval = (opt == SO_SNDTIMEO ? 2012 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 2013 2014 memset(&tv, 0, sizeof(tv)); 2015 tv.tv_sec = optval / hz; 2016 tv.tv_usec = (optval % hz) * tick; 2017 2018 error = sockopt_set(sopt, &tv, sizeof(tv)); 2019 break; 2020 2021 case SO_OVERFLOWED: 2022 error = sockopt_setint(sopt, so->so_rcv.sb_overflowed); 2023 break; 2024 2025 default: 2026 MODULE_HOOK_CALL(uipc_socket_50_getopt1_hook, 2027 (opt, so, sopt), enosys(), error); 2028 if (error) 2029 error = SET_ERROR(ENOPROTOOPT); 2030 break; 2031 } 2032 2033 return error; 2034 } 2035 2036 int 2037 sogetopt(struct socket *so, struct sockopt *sopt) 2038 { 2039 int error; 2040 2041 solock(so); 2042 if (sopt->sopt_level != SOL_SOCKET) { 2043 if (so->so_proto && so->so_proto->pr_ctloutput) { 2044 error = ((*so->so_proto->pr_ctloutput) 2045 (PRCO_GETOPT, so, sopt)); 2046 } else 2047 error = SET_ERROR(ENOPROTOOPT); 2048 } else { 2049 error = sogetopt1(so, sopt); 2050 } 2051 sounlock(so); 2052 return error; 2053 } 2054 2055 /* 2056 * alloc sockopt data buffer buffer 2057 * - will be released at destroy 2058 */ 2059 static int 2060 sockopt_alloc(struct sockopt *sopt, size_t len, km_flag_t kmflag) 2061 { 2062 void *data; 2063 2064 KASSERT(sopt->sopt_size == 0); 2065 2066 if (len > sizeof(sopt->sopt_buf)) { 2067 data = kmem_zalloc(len, kmflag); 2068 if (data == NULL) 2069 return SET_ERROR(ENOMEM); 2070 sopt->sopt_data = data; 2071 } else 2072 sopt->sopt_data = sopt->sopt_buf; 2073 2074 sopt->sopt_size = len; 2075 return 0; 2076 } 2077 2078 /* 2079 * initialise sockopt storage 2080 * - MAY sleep during allocation 2081 */ 2082 void 2083 sockopt_init(struct sockopt *sopt, int level, int name, size_t size) 2084 { 2085 2086 memset(sopt, 0, sizeof(*sopt)); 2087 2088 sopt->sopt_level = level; 2089 sopt->sopt_name = name; 2090 (void)sockopt_alloc(sopt, size, KM_SLEEP); 2091 } 2092 2093 /* 2094 * destroy sockopt storage 2095 * - will release any held memory references 2096 */ 2097 void 2098 sockopt_destroy(struct sockopt *sopt) 2099 { 2100 2101 if (sopt->sopt_data != sopt->sopt_buf) 2102 kmem_free(sopt->sopt_data, sopt->sopt_size); 2103 2104 memset(sopt, 0, sizeof(*sopt)); 2105 } 2106 2107 /* 2108 * set sockopt value 2109 * - value is copied into sockopt 2110 * - memory is allocated when necessary, will not sleep 2111 */ 2112 int 2113 sockopt_set(struct sockopt *sopt, const void *buf, size_t len) 2114 { 2115 int error; 2116 2117 if (sopt->sopt_size == 0) { 2118 error = sockopt_alloc(sopt, len, KM_NOSLEEP); 2119 if (error) 2120 return error; 2121 } 2122 2123 sopt->sopt_retsize = MIN(sopt->sopt_size, len); 2124 if (sopt->sopt_retsize > 0) { 2125 memcpy(sopt->sopt_data, buf, sopt->sopt_retsize); 2126 } 2127 2128 return 0; 2129 } 2130 2131 /* 2132 * common case of set sockopt integer value 2133 */ 2134 int 2135 sockopt_setint(struct sockopt *sopt, int val) 2136 { 2137 2138 return sockopt_set(sopt, &val, sizeof(int)); 2139 } 2140 2141 /* 2142 * get sockopt value 2143 * - correct size must be given 2144 */ 2145 int 2146 sockopt_get(const struct sockopt *sopt, void *buf, size_t len) 2147 { 2148 2149 if (sopt->sopt_size != len) 2150 return SET_ERROR(EINVAL); 2151 2152 memcpy(buf, sopt->sopt_data, len); 2153 return 0; 2154 } 2155 2156 /* 2157 * common case of get sockopt integer value 2158 */ 2159 int 2160 sockopt_getint(const struct sockopt *sopt, int *valp) 2161 { 2162 2163 return sockopt_get(sopt, valp, sizeof(int)); 2164 } 2165 2166 /* 2167 * set sockopt value from mbuf 2168 * - ONLY for legacy code 2169 * - mbuf is released by sockopt 2170 * - will not sleep 2171 */ 2172 int 2173 sockopt_setmbuf(struct sockopt *sopt, struct mbuf *m) 2174 { 2175 size_t len; 2176 int error; 2177 2178 len = m_length(m); 2179 2180 if (sopt->sopt_size == 0) { 2181 error = sockopt_alloc(sopt, len, KM_NOSLEEP); 2182 if (error) 2183 return error; 2184 } 2185 2186 sopt->sopt_retsize = MIN(sopt->sopt_size, len); 2187 m_copydata(m, 0, sopt->sopt_retsize, sopt->sopt_data); 2188 m_freem(m); 2189 2190 return 0; 2191 } 2192 2193 /* 2194 * get sockopt value into mbuf 2195 * - ONLY for legacy code 2196 * - mbuf to be released by the caller 2197 * - will not sleep 2198 */ 2199 struct mbuf * 2200 sockopt_getmbuf(const struct sockopt *sopt) 2201 { 2202 struct mbuf *m; 2203 2204 if (sopt->sopt_size > MCLBYTES) 2205 return NULL; 2206 2207 m = m_get(M_DONTWAIT, MT_SOOPTS); 2208 if (m == NULL) 2209 return NULL; 2210 2211 if (sopt->sopt_size > MLEN) { 2212 MCLGET(m, M_DONTWAIT); 2213 if ((m->m_flags & M_EXT) == 0) { 2214 m_free(m); 2215 return NULL; 2216 } 2217 } 2218 2219 memcpy(mtod(m, void *), sopt->sopt_data, sopt->sopt_size); 2220 m->m_len = sopt->sopt_size; 2221 2222 return m; 2223 } 2224 2225 void 2226 sohasoutofband(struct socket *so) 2227 { 2228 2229 so->so_state |= SS_POLLRDBAND; 2230 fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so); 2231 selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, NOTE_SUBMIT); 2232 } 2233 2234 static void 2235 filt_sordetach(struct knote *kn) 2236 { 2237 struct socket *so; 2238 2239 so = ((file_t *)kn->kn_obj)->f_socket; 2240 solock(so); 2241 if (selremove_knote(&so->so_rcv.sb_sel, kn)) 2242 so->so_rcv.sb_flags &= ~SB_KNOTE; 2243 sounlock(so); 2244 } 2245 2246 /*ARGSUSED*/ 2247 static int 2248 filt_soread(struct knote *kn, long hint) 2249 { 2250 struct socket *so; 2251 int rv; 2252 2253 so = ((file_t *)kn->kn_obj)->f_socket; 2254 if (hint != NOTE_SUBMIT) 2255 solock(so); 2256 kn->kn_data = so->so_rcv.sb_cc; 2257 if (so->so_state & SS_CANTRCVMORE) { 2258 knote_set_eof(kn, 0); 2259 kn->kn_fflags = so->so_error; 2260 rv = 1; 2261 } else if (so->so_error || so->so_rerror) 2262 rv = 1; 2263 else if (kn->kn_sfflags & NOTE_LOWAT) 2264 rv = (kn->kn_data >= kn->kn_sdata); 2265 else 2266 rv = (kn->kn_data >= so->so_rcv.sb_lowat); 2267 if (hint != NOTE_SUBMIT) 2268 sounlock(so); 2269 return rv; 2270 } 2271 2272 static void 2273 filt_sowdetach(struct knote *kn) 2274 { 2275 struct socket *so; 2276 2277 so = ((file_t *)kn->kn_obj)->f_socket; 2278 solock(so); 2279 if (selremove_knote(&so->so_snd.sb_sel, kn)) 2280 so->so_snd.sb_flags &= ~SB_KNOTE; 2281 sounlock(so); 2282 } 2283 2284 /*ARGSUSED*/ 2285 static int 2286 filt_sowrite(struct knote *kn, long hint) 2287 { 2288 struct socket *so; 2289 int rv; 2290 2291 so = ((file_t *)kn->kn_obj)->f_socket; 2292 if (hint != NOTE_SUBMIT) 2293 solock(so); 2294 kn->kn_data = sbspace(&so->so_snd); 2295 if (so->so_state & SS_CANTSENDMORE) { 2296 knote_set_eof(kn, 0); 2297 kn->kn_fflags = so->so_error; 2298 rv = 1; 2299 } else if (so->so_error) 2300 rv = 1; 2301 else if (((so->so_state & SS_ISCONNECTED) == 0) && 2302 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 2303 rv = 0; 2304 else if (kn->kn_sfflags & NOTE_LOWAT) 2305 rv = (kn->kn_data >= kn->kn_sdata); 2306 else 2307 rv = (kn->kn_data >= so->so_snd.sb_lowat); 2308 if (hint != NOTE_SUBMIT) 2309 sounlock(so); 2310 return rv; 2311 } 2312 2313 static int 2314 filt_soempty(struct knote *kn, long hint) 2315 { 2316 struct socket *so; 2317 int rv; 2318 2319 so = ((file_t *)kn->kn_obj)->f_socket; 2320 if (hint != NOTE_SUBMIT) 2321 solock(so); 2322 rv = (kn->kn_data = sbused(&so->so_snd)) == 0 || 2323 (so->so_options & SO_ACCEPTCONN) != 0; 2324 if (hint != NOTE_SUBMIT) 2325 sounlock(so); 2326 return rv; 2327 } 2328 2329 /*ARGSUSED*/ 2330 static int 2331 filt_solisten(struct knote *kn, long hint) 2332 { 2333 struct socket *so; 2334 int rv; 2335 2336 so = ((file_t *)kn->kn_obj)->f_socket; 2337 2338 /* 2339 * Set kn_data to number of incoming connections, not 2340 * counting partial (incomplete) connections. 2341 */ 2342 if (hint != NOTE_SUBMIT) 2343 solock(so); 2344 kn->kn_data = so->so_qlen; 2345 rv = (kn->kn_data > 0); 2346 if (hint != NOTE_SUBMIT) 2347 sounlock(so); 2348 return rv; 2349 } 2350 2351 static const struct filterops solisten_filtops = { 2352 .f_flags = FILTEROP_ISFD | FILTEROP_MPSAFE, 2353 .f_attach = NULL, 2354 .f_detach = filt_sordetach, 2355 .f_event = filt_solisten, 2356 }; 2357 2358 static const struct filterops soread_filtops = { 2359 .f_flags = FILTEROP_ISFD | FILTEROP_MPSAFE, 2360 .f_attach = NULL, 2361 .f_detach = filt_sordetach, 2362 .f_event = filt_soread, 2363 }; 2364 2365 static const struct filterops sowrite_filtops = { 2366 .f_flags = FILTEROP_ISFD | FILTEROP_MPSAFE, 2367 .f_attach = NULL, 2368 .f_detach = filt_sowdetach, 2369 .f_event = filt_sowrite, 2370 }; 2371 2372 static const struct filterops soempty_filtops = { 2373 .f_flags = FILTEROP_ISFD | FILTEROP_MPSAFE, 2374 .f_attach = NULL, 2375 .f_detach = filt_sowdetach, 2376 .f_event = filt_soempty, 2377 }; 2378 2379 int 2380 soo_kqfilter(struct file *fp, struct knote *kn) 2381 { 2382 struct socket *so; 2383 struct sockbuf *sb; 2384 2385 so = ((file_t *)kn->kn_obj)->f_socket; 2386 solock(so); 2387 switch (kn->kn_filter) { 2388 case EVFILT_READ: 2389 if (so->so_options & SO_ACCEPTCONN) 2390 kn->kn_fop = &solisten_filtops; 2391 else 2392 kn->kn_fop = &soread_filtops; 2393 sb = &so->so_rcv; 2394 break; 2395 case EVFILT_WRITE: 2396 kn->kn_fop = &sowrite_filtops; 2397 sb = &so->so_snd; 2398 2399 #ifdef PIPE_SOCKETPAIR 2400 if (so->so_state & SS_ISAPIPE) { 2401 /* Other end of pipe has been closed. */ 2402 if (so->so_state & SS_ISDISCONNECTED) { 2403 sounlock(so); 2404 return SET_ERROR(EBADF); 2405 } 2406 } 2407 #endif 2408 break; 2409 case EVFILT_EMPTY: 2410 kn->kn_fop = &soempty_filtops; 2411 sb = &so->so_snd; 2412 break; 2413 default: 2414 sounlock(so); 2415 return SET_ERROR(EINVAL); 2416 } 2417 selrecord_knote(&sb->sb_sel, kn); 2418 sb->sb_flags |= SB_KNOTE; 2419 sounlock(so); 2420 return 0; 2421 } 2422 2423 static int 2424 sodopoll(struct socket *so, int events) 2425 { 2426 int revents; 2427 2428 revents = 0; 2429 2430 if (events & (POLLIN | POLLRDNORM)) 2431 if (soreadable(so)) 2432 revents |= events & (POLLIN | POLLRDNORM); 2433 2434 if (events & (POLLOUT | POLLWRNORM)) 2435 if (sowritable(so)) 2436 revents |= events & (POLLOUT | POLLWRNORM); 2437 2438 if (events & (POLLPRI | POLLRDBAND)) 2439 if (so->so_state & SS_POLLRDBAND) 2440 revents |= events & (POLLPRI | POLLRDBAND); 2441 2442 return revents; 2443 } 2444 2445 int 2446 sopoll(struct socket *so, int events) 2447 { 2448 int revents = 0; 2449 2450 #ifndef DIAGNOSTIC 2451 /* 2452 * Do a quick, unlocked check in expectation that the socket 2453 * will be ready for I/O. Don't do this check if DIAGNOSTIC, 2454 * as the solocked() assertions will fail. 2455 */ 2456 if ((revents = sodopoll(so, events)) != 0) 2457 return revents; 2458 #endif 2459 2460 solock(so); 2461 if ((revents = sodopoll(so, events)) == 0) { 2462 if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) { 2463 selrecord(curlwp, &so->so_rcv.sb_sel); 2464 so->so_rcv.sb_flags |= SB_NOTIFY; 2465 } 2466 2467 if (events & (POLLOUT | POLLWRNORM)) { 2468 selrecord(curlwp, &so->so_snd.sb_sel); 2469 so->so_snd.sb_flags |= SB_NOTIFY; 2470 } 2471 } 2472 sounlock(so); 2473 2474 return revents; 2475 } 2476 2477 struct mbuf ** 2478 sbsavetimestamp(int opt, struct mbuf **mp) 2479 { 2480 struct timeval tv; 2481 int error; 2482 2483 memset(&tv, 0, sizeof(tv)); 2484 microtime(&tv); 2485 2486 MODULE_HOOK_CALL(uipc_socket_50_sbts_hook, (opt, &mp), enosys(), error); 2487 if (error == 0) 2488 return mp; 2489 2490 if (opt & SO_TIMESTAMP) { 2491 *mp = sbcreatecontrol(&tv, sizeof(tv), 2492 SCM_TIMESTAMP, SOL_SOCKET); 2493 if (*mp) 2494 mp = &(*mp)->m_next; 2495 } 2496 return mp; 2497 } 2498 2499 2500 #include <sys/sysctl.h> 2501 2502 static int sysctl_kern_somaxkva(SYSCTLFN_PROTO); 2503 static int sysctl_kern_sbmax(SYSCTLFN_PROTO); 2504 2505 /* 2506 * sysctl helper routine for kern.somaxkva. ensures that the given 2507 * value is not too small. 2508 * (XXX should we maybe make sure it's not too large as well?) 2509 */ 2510 static int 2511 sysctl_kern_somaxkva(SYSCTLFN_ARGS) 2512 { 2513 int error, new_somaxkva; 2514 struct sysctlnode node; 2515 2516 new_somaxkva = somaxkva; 2517 node = *rnode; 2518 node.sysctl_data = &new_somaxkva; 2519 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2520 if (error || newp == NULL) 2521 return error; 2522 2523 if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */ 2524 return SET_ERROR(EINVAL); 2525 2526 mutex_enter(&so_pendfree_lock); 2527 somaxkva = new_somaxkva; 2528 cv_broadcast(&socurkva_cv); 2529 mutex_exit(&so_pendfree_lock); 2530 2531 return error; 2532 } 2533 2534 /* 2535 * sysctl helper routine for kern.sbmax. Basically just ensures that 2536 * any new value is not too small. 2537 */ 2538 static int 2539 sysctl_kern_sbmax(SYSCTLFN_ARGS) 2540 { 2541 int error, new_sbmax; 2542 struct sysctlnode node; 2543 2544 new_sbmax = sb_max; 2545 node = *rnode; 2546 node.sysctl_data = &new_sbmax; 2547 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2548 if (error || newp == NULL) 2549 return error; 2550 2551 KERNEL_LOCK(1, NULL); 2552 error = sb_max_set(new_sbmax); 2553 KERNEL_UNLOCK_ONE(NULL); 2554 2555 return error; 2556 } 2557 2558 /* 2559 * sysctl helper routine for kern.sooptions. Ensures that only allowed 2560 * options can be set. 2561 */ 2562 static int 2563 sysctl_kern_sooptions(SYSCTLFN_ARGS) 2564 { 2565 int error, new_options; 2566 struct sysctlnode node; 2567 2568 new_options = sooptions; 2569 node = *rnode; 2570 node.sysctl_data = &new_options; 2571 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2572 if (error || newp == NULL) 2573 return error; 2574 2575 if (new_options & ~SO_DEFOPTS) 2576 return SET_ERROR(EINVAL); 2577 2578 sooptions = new_options; 2579 2580 return 0; 2581 } 2582 2583 static void 2584 sysctl_kern_socket_setup(void) 2585 { 2586 2587 KASSERT(socket_sysctllog == NULL); 2588 2589 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2590 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2591 CTLTYPE_INT, "somaxkva", 2592 SYSCTL_DESCR("Maximum amount of kernel memory to be " 2593 "used for socket buffers"), 2594 sysctl_kern_somaxkva, 0, NULL, 0, 2595 CTL_KERN, KERN_SOMAXKVA, CTL_EOL); 2596 2597 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2598 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2599 CTLTYPE_BOOL, "sofixedbuf", 2600 SYSCTL_DESCR("Prevent scaling of fixed socket buffers"), 2601 NULL, 0, &sofixedbuf, 0, 2602 CTL_KERN, KERN_SOFIXEDBUF, CTL_EOL); 2603 2604 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2605 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2606 CTLTYPE_INT, "sbmax", 2607 SYSCTL_DESCR("Maximum socket buffer size"), 2608 sysctl_kern_sbmax, 0, NULL, 0, 2609 CTL_KERN, KERN_SBMAX, CTL_EOL); 2610 2611 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2612 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2613 CTLTYPE_INT, "sooptions", 2614 SYSCTL_DESCR("Default socket options"), 2615 sysctl_kern_sooptions, 0, NULL, 0, 2616 CTL_KERN, CTL_CREATE, CTL_EOL); 2617 } 2618