1 /* $NetBSD: uipc_socket.c,v 1.288 2020/02/22 08:39:33 maxv 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 1.288 2020/02/22 08:39:33 maxv 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 static kauth_listener_t socket_listener; 170 171 #define SOCK_LOAN_CHUNK 65536 172 173 static void sopendfree_thread(void *); 174 static kcondvar_t pendfree_thread_cv; 175 static lwp_t *sopendfree_lwp; 176 177 static void sysctl_kern_socket_setup(void); 178 static struct sysctllog *socket_sysctllog; 179 180 static vsize_t 181 sokvareserve(struct socket *so, vsize_t len) 182 { 183 int error; 184 185 mutex_enter(&so_pendfree_lock); 186 while (socurkva + len > somaxkva) { 187 SOSEND_COUNTER_INCR(&sosend_kvalimit); 188 error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock); 189 if (error) { 190 len = 0; 191 break; 192 } 193 } 194 socurkva += len; 195 mutex_exit(&so_pendfree_lock); 196 return len; 197 } 198 199 static void 200 sokvaunreserve(vsize_t len) 201 { 202 203 mutex_enter(&so_pendfree_lock); 204 socurkva -= len; 205 cv_broadcast(&socurkva_cv); 206 mutex_exit(&so_pendfree_lock); 207 } 208 209 /* 210 * sokvaalloc: allocate kva for loan. 211 */ 212 vaddr_t 213 sokvaalloc(vaddr_t sva, vsize_t len, struct socket *so) 214 { 215 vaddr_t lva; 216 217 if (sokvareserve(so, len) == 0) 218 return 0; 219 220 lva = uvm_km_alloc(kernel_map, len, atop(sva) & uvmexp.colormask, 221 UVM_KMF_COLORMATCH | UVM_KMF_VAONLY | UVM_KMF_WAITVA); 222 if (lva == 0) { 223 sokvaunreserve(len); 224 return 0; 225 } 226 227 return lva; 228 } 229 230 /* 231 * sokvafree: free kva for loan. 232 */ 233 void 234 sokvafree(vaddr_t sva, vsize_t len) 235 { 236 237 uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY); 238 sokvaunreserve(len); 239 } 240 241 static void 242 sodoloanfree(struct vm_page **pgs, void *buf, size_t size) 243 { 244 vaddr_t sva, eva; 245 vsize_t len; 246 int npgs; 247 248 KASSERT(pgs != NULL); 249 250 eva = round_page((vaddr_t) buf + size); 251 sva = trunc_page((vaddr_t) buf); 252 len = eva - sva; 253 npgs = len >> PAGE_SHIFT; 254 255 pmap_kremove(sva, len); 256 pmap_update(pmap_kernel()); 257 uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE); 258 sokvafree(sva, len); 259 } 260 261 /* 262 * sopendfree_thread: free mbufs on "pendfree" list. Unlock and relock 263 * so_pendfree_lock when freeing mbufs. 264 */ 265 static void 266 sopendfree_thread(void *v) 267 { 268 struct mbuf *m, *next; 269 size_t rv; 270 271 mutex_enter(&so_pendfree_lock); 272 273 for (;;) { 274 rv = 0; 275 while (so_pendfree != NULL) { 276 m = so_pendfree; 277 so_pendfree = NULL; 278 mutex_exit(&so_pendfree_lock); 279 280 for (; m != NULL; m = next) { 281 next = m->m_next; 282 KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 283 0); 284 KASSERT(m->m_ext.ext_refcnt == 0); 285 286 rv += m->m_ext.ext_size; 287 sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf, 288 m->m_ext.ext_size); 289 pool_cache_put(mb_cache, m); 290 } 291 292 mutex_enter(&so_pendfree_lock); 293 } 294 if (rv) 295 cv_broadcast(&socurkva_cv); 296 cv_wait(&pendfree_thread_cv, &so_pendfree_lock); 297 } 298 panic("sopendfree_thread"); 299 /* NOTREACHED */ 300 } 301 302 void 303 soloanfree(struct mbuf *m, void *buf, size_t size, void *arg) 304 { 305 306 KASSERT(m != NULL); 307 308 /* 309 * postpone freeing mbuf. 310 * 311 * we can't do it in interrupt context 312 * because we need to put kva back to kernel_map. 313 */ 314 315 mutex_enter(&so_pendfree_lock); 316 m->m_next = so_pendfree; 317 so_pendfree = m; 318 cv_signal(&pendfree_thread_cv); 319 mutex_exit(&so_pendfree_lock); 320 } 321 322 static long 323 sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space) 324 { 325 struct iovec *iov = uio->uio_iov; 326 vaddr_t sva, eva; 327 vsize_t len; 328 vaddr_t lva; 329 int npgs, error; 330 vaddr_t va; 331 int i; 332 333 if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace)) 334 return 0; 335 336 if (iov->iov_len < (size_t) space) 337 space = iov->iov_len; 338 if (space > SOCK_LOAN_CHUNK) 339 space = SOCK_LOAN_CHUNK; 340 341 eva = round_page((vaddr_t) iov->iov_base + space); 342 sva = trunc_page((vaddr_t) iov->iov_base); 343 len = eva - sva; 344 npgs = len >> PAGE_SHIFT; 345 346 KASSERT(npgs <= M_EXT_MAXPAGES); 347 348 lva = sokvaalloc(sva, len, so); 349 if (lva == 0) 350 return 0; 351 352 error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len, 353 m->m_ext.ext_pgs, UVM_LOAN_TOPAGE); 354 if (error) { 355 sokvafree(lva, len); 356 return 0; 357 } 358 359 for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE) 360 pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]), 361 VM_PROT_READ, 0); 362 pmap_update(pmap_kernel()); 363 364 lva += (vaddr_t) iov->iov_base & PAGE_MASK; 365 366 MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so); 367 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 368 369 uio->uio_resid -= space; 370 /* uio_offset not updated, not set/used for write(2) */ 371 uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space; 372 uio->uio_iov->iov_len -= space; 373 if (uio->uio_iov->iov_len == 0) { 374 uio->uio_iov++; 375 uio->uio_iovcnt--; 376 } 377 378 return space; 379 } 380 381 static int 382 socket_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 383 void *arg0, void *arg1, void *arg2, void *arg3) 384 { 385 int result; 386 enum kauth_network_req req; 387 388 result = KAUTH_RESULT_DEFER; 389 req = (enum kauth_network_req)(uintptr_t)arg0; 390 391 if ((action != KAUTH_NETWORK_SOCKET) && 392 (action != KAUTH_NETWORK_BIND)) 393 return result; 394 395 switch (req) { 396 case KAUTH_REQ_NETWORK_BIND_PORT: 397 result = KAUTH_RESULT_ALLOW; 398 break; 399 400 case KAUTH_REQ_NETWORK_SOCKET_DROP: { 401 /* Normal users can only drop their own connections. */ 402 struct socket *so = (struct socket *)arg1; 403 404 if (so->so_cred && proc_uidmatch(cred, so->so_cred) == 0) 405 result = KAUTH_RESULT_ALLOW; 406 407 break; 408 } 409 410 case KAUTH_REQ_NETWORK_SOCKET_OPEN: 411 /* We allow "raw" routing/bluetooth sockets to anyone. */ 412 switch ((u_long)arg1) { 413 case PF_ROUTE: 414 case PF_OROUTE: 415 case PF_BLUETOOTH: 416 case PF_CAN: 417 result = KAUTH_RESULT_ALLOW; 418 break; 419 default: 420 /* Privileged, let secmodel handle this. */ 421 if ((u_long)arg2 == SOCK_RAW) 422 break; 423 result = KAUTH_RESULT_ALLOW; 424 break; 425 } 426 break; 427 428 case KAUTH_REQ_NETWORK_SOCKET_CANSEE: 429 result = KAUTH_RESULT_ALLOW; 430 431 break; 432 433 default: 434 break; 435 } 436 437 return result; 438 } 439 440 void 441 soinit(void) 442 { 443 444 sysctl_kern_socket_setup(); 445 446 #ifdef SCTP 447 /* Update the SCTP function hooks if necessary*/ 448 449 vec_sctp_add_ip_address = sctp_add_ip_address; 450 vec_sctp_delete_ip_address = sctp_delete_ip_address; 451 #endif 452 453 mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM); 454 softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 455 cv_init(&socurkva_cv, "sokva"); 456 cv_init(&pendfree_thread_cv, "sopendfr"); 457 soinit2(); 458 459 /* Set the initial adjusted socket buffer size. */ 460 if (sb_max_set(sb_max)) 461 panic("bad initial sb_max value: %lu", sb_max); 462 463 socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 464 socket_listener_cb, NULL); 465 } 466 467 void 468 soinit1(void) 469 { 470 int error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL, 471 sopendfree_thread, NULL, &sopendfree_lwp, "sopendfree"); 472 if (error) 473 panic("soinit1 %d", error); 474 } 475 476 /* 477 * socreate: create a new socket of the specified type and the protocol. 478 * 479 * => Caller may specify another socket for lock sharing (must not be held). 480 * => Returns the new socket without lock held. 481 */ 482 int 483 socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l, 484 struct socket *lockso) 485 { 486 const struct protosw *prp; 487 struct socket *so; 488 uid_t uid; 489 int error; 490 kmutex_t *lock; 491 492 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 493 KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type), 494 KAUTH_ARG(proto)); 495 if (error != 0) 496 return error; 497 498 if (proto) 499 prp = pffindproto(dom, proto, type); 500 else 501 prp = pffindtype(dom, type); 502 if (prp == NULL) { 503 /* no support for domain */ 504 if (pffinddomain(dom) == 0) 505 return EAFNOSUPPORT; 506 /* no support for socket type */ 507 if (proto == 0 && type != 0) 508 return EPROTOTYPE; 509 return EPROTONOSUPPORT; 510 } 511 if (prp->pr_usrreqs == NULL) 512 return EPROTONOSUPPORT; 513 if (prp->pr_type != type) 514 return EPROTOTYPE; 515 516 so = soget(true); 517 so->so_type = type; 518 so->so_proto = prp; 519 so->so_send = sosend; 520 so->so_receive = soreceive; 521 so->so_options = sooptions; 522 #ifdef MBUFTRACE 523 so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner; 524 so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner; 525 so->so_mowner = &prp->pr_domain->dom_mowner; 526 #endif 527 uid = kauth_cred_geteuid(l->l_cred); 528 so->so_uidinfo = uid_find(uid); 529 so->so_cpid = l->l_proc->p_pid; 530 531 /* 532 * Lock assigned and taken during PCB attach, unless we share 533 * the lock with another socket, e.g. socketpair(2) case. 534 */ 535 if (lockso) { 536 lock = lockso->so_lock; 537 so->so_lock = lock; 538 mutex_obj_hold(lock); 539 mutex_enter(lock); 540 } 541 542 /* Attach the PCB (returns with the socket lock held). */ 543 error = (*prp->pr_usrreqs->pr_attach)(so, proto); 544 KASSERT(solocked(so)); 545 546 if (error) { 547 KASSERT(so->so_pcb == NULL); 548 so->so_state |= SS_NOFDREF; 549 sofree(so); 550 return error; 551 } 552 so->so_cred = kauth_cred_dup(l->l_cred); 553 sounlock(so); 554 555 *aso = so; 556 return 0; 557 } 558 559 /* 560 * fsocreate: create a socket and a file descriptor associated with it. 561 * 562 * => On success, write file descriptor to fdout and return zero. 563 * => On failure, return non-zero; *fdout will be undefined. 564 */ 565 int 566 fsocreate(int domain, struct socket **sop, int type, int proto, int *fdout) 567 { 568 lwp_t *l = curlwp; 569 int error, fd, flags; 570 struct socket *so; 571 struct file *fp; 572 573 if ((error = fd_allocfile(&fp, &fd)) != 0) { 574 return error; 575 } 576 flags = type & SOCK_FLAGS_MASK; 577 fd_set_exclose(l, fd, (flags & SOCK_CLOEXEC) != 0); 578 fp->f_flag = FREAD|FWRITE|((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)| 579 ((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0); 580 fp->f_type = DTYPE_SOCKET; 581 fp->f_ops = &socketops; 582 583 type &= ~SOCK_FLAGS_MASK; 584 error = socreate(domain, &so, type, proto, l, NULL); 585 if (error) { 586 fd_abort(curproc, fp, fd); 587 return error; 588 } 589 if (flags & SOCK_NONBLOCK) { 590 so->so_state |= SS_NBIO; 591 } 592 fp->f_socket = so; 593 fd_affix(curproc, fp, fd); 594 595 if (sop != NULL) { 596 *sop = so; 597 } 598 *fdout = fd; 599 return error; 600 } 601 602 int 603 sofamily(const struct socket *so) 604 { 605 const struct protosw *pr; 606 const struct domain *dom; 607 608 if ((pr = so->so_proto) == NULL) 609 return AF_UNSPEC; 610 if ((dom = pr->pr_domain) == NULL) 611 return AF_UNSPEC; 612 return dom->dom_family; 613 } 614 615 int 616 sobind(struct socket *so, struct sockaddr *nam, struct lwp *l) 617 { 618 int error; 619 620 solock(so); 621 if (nam->sa_family != so->so_proto->pr_domain->dom_family) { 622 sounlock(so); 623 return EAFNOSUPPORT; 624 } 625 error = (*so->so_proto->pr_usrreqs->pr_bind)(so, nam, l); 626 sounlock(so); 627 return error; 628 } 629 630 int 631 solisten(struct socket *so, int backlog, struct lwp *l) 632 { 633 int error; 634 short oldopt, oldqlimit; 635 636 solock(so); 637 if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | 638 SS_ISDISCONNECTING)) != 0) { 639 sounlock(so); 640 return EINVAL; 641 } 642 oldopt = so->so_options; 643 oldqlimit = so->so_qlimit; 644 if (TAILQ_EMPTY(&so->so_q)) 645 so->so_options |= SO_ACCEPTCONN; 646 if (backlog < 0) 647 backlog = 0; 648 so->so_qlimit = uimin(backlog, somaxconn); 649 650 error = (*so->so_proto->pr_usrreqs->pr_listen)(so, l); 651 if (error != 0) { 652 so->so_options = oldopt; 653 so->so_qlimit = oldqlimit; 654 sounlock(so); 655 return error; 656 } 657 sounlock(so); 658 return 0; 659 } 660 661 void 662 sofree(struct socket *so) 663 { 664 u_int refs; 665 666 KASSERT(solocked(so)); 667 668 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) { 669 sounlock(so); 670 return; 671 } 672 if (so->so_head) { 673 /* 674 * We must not decommission a socket that's on the accept(2) 675 * queue. If we do, then accept(2) may hang after select(2) 676 * indicated that the listening socket was ready. 677 */ 678 if (!soqremque(so, 0)) { 679 sounlock(so); 680 return; 681 } 682 } 683 if (so->so_rcv.sb_hiwat) 684 (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, 685 RLIM_INFINITY); 686 if (so->so_snd.sb_hiwat) 687 (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, 688 RLIM_INFINITY); 689 sbrelease(&so->so_snd, so); 690 KASSERT(!cv_has_waiters(&so->so_cv)); 691 KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv)); 692 KASSERT(!cv_has_waiters(&so->so_snd.sb_cv)); 693 sorflush(so); 694 refs = so->so_aborting; /* XXX */ 695 /* Remove acccept filter if one is present. */ 696 if (so->so_accf != NULL) 697 (void)accept_filt_clear(so); 698 sounlock(so); 699 if (refs == 0) /* XXX */ 700 soput(so); 701 } 702 703 /* 704 * soclose: close a socket on last file table reference removal. 705 * Initiate disconnect if connected. Free socket when disconnect complete. 706 */ 707 int 708 soclose(struct socket *so) 709 { 710 struct socket *so2; 711 int error = 0; 712 713 solock(so); 714 if (so->so_options & SO_ACCEPTCONN) { 715 for (;;) { 716 if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) { 717 KASSERT(solocked2(so, so2)); 718 (void) soqremque(so2, 0); 719 /* soabort drops the lock. */ 720 (void) soabort(so2); 721 solock(so); 722 continue; 723 } 724 if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) { 725 KASSERT(solocked2(so, so2)); 726 (void) soqremque(so2, 1); 727 /* soabort drops the lock. */ 728 (void) soabort(so2); 729 solock(so); 730 continue; 731 } 732 break; 733 } 734 } 735 if (so->so_pcb == NULL) 736 goto discard; 737 if (so->so_state & SS_ISCONNECTED) { 738 if ((so->so_state & SS_ISDISCONNECTING) == 0) { 739 error = sodisconnect(so); 740 if (error) 741 goto drop; 742 } 743 if (so->so_options & SO_LINGER) { 744 if ((so->so_state & (SS_ISDISCONNECTING|SS_NBIO)) == 745 (SS_ISDISCONNECTING|SS_NBIO)) 746 goto drop; 747 while (so->so_state & SS_ISCONNECTED) { 748 error = sowait(so, true, so->so_linger * hz); 749 if (error) 750 break; 751 } 752 } 753 } 754 drop: 755 if (so->so_pcb) { 756 KASSERT(solocked(so)); 757 (*so->so_proto->pr_usrreqs->pr_detach)(so); 758 } 759 discard: 760 KASSERT((so->so_state & SS_NOFDREF) == 0); 761 kauth_cred_free(so->so_cred); 762 so->so_cred = NULL; 763 so->so_state |= SS_NOFDREF; 764 sofree(so); 765 return error; 766 } 767 768 /* 769 * Must be called with the socket locked.. Will return with it unlocked. 770 */ 771 int 772 soabort(struct socket *so) 773 { 774 u_int refs; 775 int error; 776 777 KASSERT(solocked(so)); 778 KASSERT(so->so_head == NULL); 779 780 so->so_aborting++; /* XXX */ 781 error = (*so->so_proto->pr_usrreqs->pr_abort)(so); 782 refs = --so->so_aborting; /* XXX */ 783 if (error || (refs == 0)) { 784 sofree(so); 785 } else { 786 sounlock(so); 787 } 788 return error; 789 } 790 791 int 792 soaccept(struct socket *so, struct sockaddr *nam) 793 { 794 int error; 795 796 KASSERT(solocked(so)); 797 KASSERT((so->so_state & SS_NOFDREF) != 0); 798 799 so->so_state &= ~SS_NOFDREF; 800 if ((so->so_state & SS_ISDISCONNECTED) == 0 || 801 (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0) 802 error = (*so->so_proto->pr_usrreqs->pr_accept)(so, nam); 803 else 804 error = ECONNABORTED; 805 806 return error; 807 } 808 809 int 810 soconnect(struct socket *so, struct sockaddr *nam, struct lwp *l) 811 { 812 int error; 813 814 KASSERT(solocked(so)); 815 816 if (so->so_options & SO_ACCEPTCONN) 817 return EOPNOTSUPP; 818 /* 819 * If protocol is connection-based, can only connect once. 820 * Otherwise, if connected, try to disconnect first. 821 * This allows user to disconnect by connecting to, e.g., 822 * a null address. 823 */ 824 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && 825 ((so->so_proto->pr_flags & PR_CONNREQUIRED) || 826 (error = sodisconnect(so)))) { 827 error = EISCONN; 828 } else { 829 if (nam->sa_family != so->so_proto->pr_domain->dom_family) { 830 return EAFNOSUPPORT; 831 } 832 error = (*so->so_proto->pr_usrreqs->pr_connect)(so, nam, l); 833 } 834 835 return error; 836 } 837 838 int 839 soconnect2(struct socket *so1, struct socket *so2) 840 { 841 KASSERT(solocked2(so1, so2)); 842 843 return (*so1->so_proto->pr_usrreqs->pr_connect2)(so1, so2); 844 } 845 846 int 847 sodisconnect(struct socket *so) 848 { 849 int error; 850 851 KASSERT(solocked(so)); 852 853 if ((so->so_state & SS_ISCONNECTED) == 0) { 854 error = ENOTCONN; 855 } else if (so->so_state & SS_ISDISCONNECTING) { 856 error = EALREADY; 857 } else { 858 error = (*so->so_proto->pr_usrreqs->pr_disconnect)(so); 859 } 860 return error; 861 } 862 863 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) 864 /* 865 * Send on a socket. 866 * If send must go all at once and message is larger than 867 * send buffering, then hard error. 868 * Lock against other senders. 869 * If must go all at once and not enough room now, then 870 * inform user that this would block and do nothing. 871 * Otherwise, if nonblocking, send as much as possible. 872 * The data to be sent is described by "uio" if nonzero, 873 * otherwise by the mbuf chain "top" (which must be null 874 * if uio is not). Data provided in mbuf chain must be small 875 * enough to send all at once. 876 * 877 * Returns nonzero on error, timeout or signal; callers 878 * must check for short counts if EINTR/ERESTART are returned. 879 * Data and control buffers are freed on return. 880 */ 881 int 882 sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, 883 struct mbuf *top, struct mbuf *control, int flags, struct lwp *l) 884 { 885 struct mbuf **mp, *m; 886 long space, len, resid, clen, mlen; 887 int error, s, dontroute, atomic; 888 short wakeup_state = 0; 889 890 clen = 0; 891 892 /* 893 * solock() provides atomicity of access. splsoftnet() prevents 894 * protocol processing soft interrupts from interrupting us and 895 * blocking (expensive). 896 */ 897 s = splsoftnet(); 898 solock(so); 899 atomic = sosendallatonce(so) || top; 900 if (uio) 901 resid = uio->uio_resid; 902 else 903 resid = top->m_pkthdr.len; 904 /* 905 * In theory resid should be unsigned. 906 * However, space must be signed, as it might be less than 0 907 * if we over-committed, and we must use a signed comparison 908 * of space and resid. On the other hand, a negative resid 909 * causes us to loop sending 0-length segments to the protocol. 910 */ 911 if (resid < 0) { 912 error = EINVAL; 913 goto out; 914 } 915 dontroute = 916 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && 917 (so->so_proto->pr_flags & PR_ATOMIC); 918 l->l_ru.ru_msgsnd++; 919 if (control) 920 clen = control->m_len; 921 restart: 922 if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) 923 goto out; 924 do { 925 if (so->so_state & SS_CANTSENDMORE) { 926 error = EPIPE; 927 goto release; 928 } 929 if (so->so_error) { 930 error = so->so_error; 931 if ((flags & MSG_PEEK) == 0) 932 so->so_error = 0; 933 goto release; 934 } 935 if ((so->so_state & SS_ISCONNECTED) == 0) { 936 if (so->so_proto->pr_flags & PR_CONNREQUIRED) { 937 if (resid || clen == 0) { 938 error = ENOTCONN; 939 goto release; 940 } 941 } else if (addr == NULL) { 942 error = EDESTADDRREQ; 943 goto release; 944 } 945 } 946 space = sbspace(&so->so_snd); 947 if (flags & MSG_OOB) 948 space += 1024; 949 if ((atomic && resid > so->so_snd.sb_hiwat) || 950 clen > so->so_snd.sb_hiwat) { 951 error = EMSGSIZE; 952 goto release; 953 } 954 if (space < resid + clen && 955 (atomic || space < so->so_snd.sb_lowat || space < clen)) { 956 if ((so->so_state & SS_NBIO) || (flags & MSG_NBIO)) { 957 error = EWOULDBLOCK; 958 goto release; 959 } 960 sbunlock(&so->so_snd); 961 if (wakeup_state & SS_RESTARTSYS) { 962 error = ERESTART; 963 goto out; 964 } 965 error = sbwait(&so->so_snd); 966 if (error) 967 goto out; 968 wakeup_state = so->so_state; 969 goto restart; 970 } 971 wakeup_state = 0; 972 mp = ⊤ 973 space -= clen; 974 do { 975 if (uio == NULL) { 976 /* 977 * Data is prepackaged in "top". 978 */ 979 resid = 0; 980 if (flags & MSG_EOR) 981 top->m_flags |= M_EOR; 982 } else do { 983 sounlock(so); 984 splx(s); 985 if (top == NULL) { 986 m = m_gethdr(M_WAIT, MT_DATA); 987 mlen = MHLEN; 988 m->m_pkthdr.len = 0; 989 m_reset_rcvif(m); 990 } else { 991 m = m_get(M_WAIT, MT_DATA); 992 mlen = MLEN; 993 } 994 MCLAIM(m, so->so_snd.sb_mowner); 995 if (sock_loan_thresh >= 0 && 996 uio->uio_iov->iov_len >= sock_loan_thresh && 997 space >= sock_loan_thresh && 998 (len = sosend_loan(so, uio, m, 999 space)) != 0) { 1000 SOSEND_COUNTER_INCR(&sosend_loan_big); 1001 space -= len; 1002 goto have_data; 1003 } 1004 if (resid >= MINCLSIZE && space >= MCLBYTES) { 1005 SOSEND_COUNTER_INCR(&sosend_copy_big); 1006 m_clget(m, M_DONTWAIT); 1007 if ((m->m_flags & M_EXT) == 0) 1008 goto nopages; 1009 mlen = MCLBYTES; 1010 if (atomic && top == 0) { 1011 len = lmin(MCLBYTES - max_hdr, 1012 resid); 1013 m->m_data += max_hdr; 1014 } else 1015 len = lmin(MCLBYTES, resid); 1016 space -= len; 1017 } else { 1018 nopages: 1019 SOSEND_COUNTER_INCR(&sosend_copy_small); 1020 len = lmin(lmin(mlen, resid), space); 1021 space -= len; 1022 /* 1023 * For datagram protocols, leave room 1024 * for protocol headers in first mbuf. 1025 */ 1026 if (atomic && top == 0 && len < mlen) 1027 m_align(m, len); 1028 } 1029 error = uiomove(mtod(m, void *), (int)len, uio); 1030 have_data: 1031 resid = uio->uio_resid; 1032 m->m_len = len; 1033 *mp = m; 1034 top->m_pkthdr.len += len; 1035 s = splsoftnet(); 1036 solock(so); 1037 if (error != 0) 1038 goto release; 1039 mp = &m->m_next; 1040 if (resid <= 0) { 1041 if (flags & MSG_EOR) 1042 top->m_flags |= M_EOR; 1043 break; 1044 } 1045 } while (space > 0 && atomic); 1046 1047 if (so->so_state & SS_CANTSENDMORE) { 1048 error = EPIPE; 1049 goto release; 1050 } 1051 if (dontroute) 1052 so->so_options |= SO_DONTROUTE; 1053 if (resid > 0) 1054 so->so_state |= SS_MORETOCOME; 1055 if (flags & MSG_OOB) { 1056 error = (*so->so_proto->pr_usrreqs->pr_sendoob)( 1057 so, top, control); 1058 } else { 1059 error = (*so->so_proto->pr_usrreqs->pr_send)(so, 1060 top, addr, control, l); 1061 } 1062 if (dontroute) 1063 so->so_options &= ~SO_DONTROUTE; 1064 if (resid > 0) 1065 so->so_state &= ~SS_MORETOCOME; 1066 clen = 0; 1067 control = NULL; 1068 top = NULL; 1069 mp = ⊤ 1070 if (error != 0) 1071 goto release; 1072 } while (resid && space > 0); 1073 } while (resid); 1074 1075 release: 1076 sbunlock(&so->so_snd); 1077 out: 1078 sounlock(so); 1079 splx(s); 1080 if (top) 1081 m_freem(top); 1082 if (control) 1083 m_freem(control); 1084 return error; 1085 } 1086 1087 /* 1088 * Following replacement or removal of the first mbuf on the first 1089 * mbuf chain of a socket buffer, push necessary state changes back 1090 * into the socket buffer so that other consumers see the values 1091 * consistently. 'nextrecord' is the caller's locally stored value of 1092 * the original value of sb->sb_mb->m_nextpkt which must be restored 1093 * when the lead mbuf changes. NOTE: 'nextrecord' may be NULL. 1094 */ 1095 static void 1096 sbsync(struct sockbuf *sb, struct mbuf *nextrecord) 1097 { 1098 1099 KASSERT(solocked(sb->sb_so)); 1100 1101 /* 1102 * First, update for the new value of nextrecord. If necessary, 1103 * make it the first record. 1104 */ 1105 if (sb->sb_mb != NULL) 1106 sb->sb_mb->m_nextpkt = nextrecord; 1107 else 1108 sb->sb_mb = nextrecord; 1109 1110 /* 1111 * Now update any dependent socket buffer fields to reflect 1112 * the new state. This is an inline of SB_EMPTY_FIXUP, with 1113 * the addition of a second clause that takes care of the 1114 * case where sb_mb has been updated, but remains the last 1115 * record. 1116 */ 1117 if (sb->sb_mb == NULL) { 1118 sb->sb_mbtail = NULL; 1119 sb->sb_lastrecord = NULL; 1120 } else if (sb->sb_mb->m_nextpkt == NULL) 1121 sb->sb_lastrecord = sb->sb_mb; 1122 } 1123 1124 /* 1125 * Implement receive operations on a socket. 1126 * 1127 * We depend on the way that records are added to the sockbuf by sbappend*. In 1128 * particular, each record (mbufs linked through m_next) must begin with an 1129 * address if the protocol so specifies, followed by an optional mbuf or mbufs 1130 * containing ancillary data, and then zero or more mbufs of data. 1131 * 1132 * In order to avoid blocking network interrupts for the entire time here, we 1133 * splx() while doing the actual copy to user space. Although the sockbuf is 1134 * locked, new data may still be appended, and thus we must maintain 1135 * consistency of the sockbuf during that time. 1136 * 1137 * The caller may receive the data as a single mbuf chain by supplying an mbuf 1138 * **mp0 for use in returning the chain. The uio is then used only for the 1139 * count in uio_resid. 1140 */ 1141 int 1142 soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, 1143 struct mbuf **mp0, struct mbuf **controlp, int *flagsp) 1144 { 1145 struct lwp *l = curlwp; 1146 struct mbuf *m, **mp, *mt; 1147 size_t len, offset, moff, orig_resid; 1148 int atomic, flags, error, s, type; 1149 const struct protosw *pr; 1150 struct mbuf *nextrecord; 1151 int mbuf_removed = 0; 1152 const struct domain *dom; 1153 short wakeup_state = 0; 1154 1155 pr = so->so_proto; 1156 atomic = pr->pr_flags & PR_ATOMIC; 1157 dom = pr->pr_domain; 1158 mp = mp0; 1159 type = 0; 1160 orig_resid = uio->uio_resid; 1161 1162 if (paddr != NULL) 1163 *paddr = NULL; 1164 if (controlp != NULL) 1165 *controlp = NULL; 1166 if (flagsp != NULL) 1167 flags = *flagsp &~ MSG_EOR; 1168 else 1169 flags = 0; 1170 1171 if (flags & MSG_OOB) { 1172 m = m_get(M_WAIT, MT_DATA); 1173 solock(so); 1174 error = (*pr->pr_usrreqs->pr_recvoob)(so, m, flags & MSG_PEEK); 1175 sounlock(so); 1176 if (error) 1177 goto bad; 1178 do { 1179 error = uiomove(mtod(m, void *), 1180 MIN(uio->uio_resid, m->m_len), uio); 1181 m = m_free(m); 1182 } while (uio->uio_resid > 0 && error == 0 && m); 1183 /* We consumed the oob data, no more oobmark. */ 1184 so->so_oobmark = 0; 1185 so->so_state &= ~SS_RCVATMARK; 1186 bad: 1187 if (m != NULL) 1188 m_freem(m); 1189 return error; 1190 } 1191 if (mp != NULL) 1192 *mp = NULL; 1193 1194 /* 1195 * solock() provides atomicity of access. splsoftnet() prevents 1196 * protocol processing soft interrupts from interrupting us and 1197 * blocking (expensive). 1198 */ 1199 s = splsoftnet(); 1200 solock(so); 1201 restart: 1202 if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) { 1203 sounlock(so); 1204 splx(s); 1205 return error; 1206 } 1207 m = so->so_rcv.sb_mb; 1208 1209 /* 1210 * If we have less data than requested, block awaiting more 1211 * (subject to any timeout) if: 1212 * 1. the current count is less than the low water mark, 1213 * 2. MSG_WAITALL is set, and it is possible to do the entire 1214 * receive operation at once if we block (resid <= hiwat), or 1215 * 3. MSG_DONTWAIT is not set. 1216 * If MSG_WAITALL is set but resid is larger than the receive buffer, 1217 * we have to do the receive in sections, and thus risk returning 1218 * a short count if a timeout or signal occurs after we start. 1219 */ 1220 if (m == NULL || 1221 ((flags & MSG_DONTWAIT) == 0 && 1222 so->so_rcv.sb_cc < uio->uio_resid && 1223 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || 1224 ((flags & MSG_WAITALL) && 1225 uio->uio_resid <= so->so_rcv.sb_hiwat)) && 1226 m->m_nextpkt == NULL && !atomic)) { 1227 #ifdef DIAGNOSTIC 1228 if (m == NULL && so->so_rcv.sb_cc) 1229 panic("receive 1"); 1230 #endif 1231 if (so->so_error || so->so_rerror) { 1232 u_short *e; 1233 if (m != NULL) 1234 goto dontblock; 1235 e = so->so_error ? &so->so_error : &so->so_rerror; 1236 error = *e; 1237 if ((flags & MSG_PEEK) == 0) 1238 *e = 0; 1239 goto release; 1240 } 1241 if (so->so_state & SS_CANTRCVMORE) { 1242 if (m != NULL) 1243 goto dontblock; 1244 else 1245 goto release; 1246 } 1247 for (; m != NULL; m = m->m_next) 1248 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { 1249 m = so->so_rcv.sb_mb; 1250 goto dontblock; 1251 } 1252 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && 1253 (so->so_proto->pr_flags & PR_CONNREQUIRED)) { 1254 error = ENOTCONN; 1255 goto release; 1256 } 1257 if (uio->uio_resid == 0) 1258 goto release; 1259 if ((so->so_state & SS_NBIO) || 1260 (flags & (MSG_DONTWAIT|MSG_NBIO))) { 1261 error = EWOULDBLOCK; 1262 goto release; 1263 } 1264 SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1"); 1265 SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1"); 1266 sbunlock(&so->so_rcv); 1267 if (wakeup_state & SS_RESTARTSYS) 1268 error = ERESTART; 1269 else 1270 error = sbwait(&so->so_rcv); 1271 if (error != 0) { 1272 sounlock(so); 1273 splx(s); 1274 return error; 1275 } 1276 wakeup_state = so->so_state; 1277 goto restart; 1278 } 1279 1280 dontblock: 1281 /* 1282 * On entry here, m points to the first record of the socket buffer. 1283 * From this point onward, we maintain 'nextrecord' as a cache of the 1284 * pointer to the next record in the socket buffer. We must keep the 1285 * various socket buffer pointers and local stack versions of the 1286 * pointers in sync, pushing out modifications before dropping the 1287 * socket lock, and re-reading them when picking it up. 1288 * 1289 * Otherwise, we will race with the network stack appending new data 1290 * or records onto the socket buffer by using inconsistent/stale 1291 * versions of the field, possibly resulting in socket buffer 1292 * corruption. 1293 * 1294 * By holding the high-level sblock(), we prevent simultaneous 1295 * readers from pulling off the front of the socket buffer. 1296 */ 1297 if (l != NULL) 1298 l->l_ru.ru_msgrcv++; 1299 KASSERT(m == so->so_rcv.sb_mb); 1300 SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); 1301 SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); 1302 nextrecord = m->m_nextpkt; 1303 1304 if (pr->pr_flags & PR_ADDR) { 1305 KASSERT(m->m_type == MT_SONAME); 1306 orig_resid = 0; 1307 if (flags & MSG_PEEK) { 1308 if (paddr) 1309 *paddr = m_copym(m, 0, m->m_len, M_DONTWAIT); 1310 m = m->m_next; 1311 } else { 1312 sbfree(&so->so_rcv, m); 1313 mbuf_removed = 1; 1314 if (paddr != NULL) { 1315 *paddr = m; 1316 so->so_rcv.sb_mb = m->m_next; 1317 m->m_next = NULL; 1318 m = so->so_rcv.sb_mb; 1319 } else { 1320 m = so->so_rcv.sb_mb = m_free(m); 1321 } 1322 sbsync(&so->so_rcv, nextrecord); 1323 } 1324 } 1325 1326 if (pr->pr_flags & PR_ADDR_OPT) { 1327 /* 1328 * For SCTP we may be getting a whole message OR a partial 1329 * delivery. 1330 */ 1331 if (m->m_type == MT_SONAME) { 1332 orig_resid = 0; 1333 if (flags & MSG_PEEK) { 1334 if (paddr) 1335 *paddr = m_copym(m, 0, m->m_len, M_DONTWAIT); 1336 m = m->m_next; 1337 } else { 1338 sbfree(&so->so_rcv, m); 1339 /* XXX XXX XXX: should set mbuf_removed? */ 1340 if (paddr) { 1341 *paddr = m; 1342 so->so_rcv.sb_mb = m->m_next; 1343 m->m_next = 0; 1344 m = so->so_rcv.sb_mb; 1345 } else { 1346 m = so->so_rcv.sb_mb = m_free(m); 1347 } 1348 /* XXX XXX XXX: isn't there an sbsync() 1349 * missing here? */ 1350 } 1351 } 1352 } 1353 1354 /* 1355 * Process one or more MT_CONTROL mbufs present before any data mbufs 1356 * in the first mbuf chain on the socket buffer. If MSG_PEEK, we 1357 * just copy the data; if !MSG_PEEK, we call into the protocol to 1358 * perform externalization (or freeing if controlp == NULL). 1359 */ 1360 if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) { 1361 struct mbuf *cm = NULL, *cmn; 1362 struct mbuf **cme = &cm; 1363 1364 do { 1365 if (flags & MSG_PEEK) { 1366 if (controlp != NULL) { 1367 *controlp = m_copym(m, 0, m->m_len, M_DONTWAIT); 1368 controlp = &(*controlp)->m_next; 1369 } 1370 m = m->m_next; 1371 } else { 1372 sbfree(&so->so_rcv, m); 1373 so->so_rcv.sb_mb = m->m_next; 1374 m->m_next = NULL; 1375 *cme = m; 1376 cme = &(*cme)->m_next; 1377 m = so->so_rcv.sb_mb; 1378 } 1379 } while (m != NULL && m->m_type == MT_CONTROL); 1380 if ((flags & MSG_PEEK) == 0) 1381 sbsync(&so->so_rcv, nextrecord); 1382 1383 for (; cm != NULL; cm = cmn) { 1384 cmn = cm->m_next; 1385 cm->m_next = NULL; 1386 type = mtod(cm, struct cmsghdr *)->cmsg_type; 1387 if (controlp != NULL) { 1388 if (dom->dom_externalize != NULL && 1389 type == SCM_RIGHTS) { 1390 sounlock(so); 1391 splx(s); 1392 error = (*dom->dom_externalize)(cm, l, 1393 (flags & MSG_CMSG_CLOEXEC) ? 1394 O_CLOEXEC : 0); 1395 s = splsoftnet(); 1396 solock(so); 1397 } 1398 *controlp = cm; 1399 while (*controlp != NULL) 1400 controlp = &(*controlp)->m_next; 1401 } else { 1402 /* 1403 * Dispose of any SCM_RIGHTS message that went 1404 * through the read path rather than recv. 1405 */ 1406 if (dom->dom_dispose != NULL && 1407 type == SCM_RIGHTS) { 1408 sounlock(so); 1409 (*dom->dom_dispose)(cm); 1410 solock(so); 1411 } 1412 m_freem(cm); 1413 } 1414 } 1415 if (m != NULL) 1416 nextrecord = so->so_rcv.sb_mb->m_nextpkt; 1417 else 1418 nextrecord = so->so_rcv.sb_mb; 1419 orig_resid = 0; 1420 } 1421 1422 /* If m is non-NULL, we have some data to read. */ 1423 if (__predict_true(m != NULL)) { 1424 type = m->m_type; 1425 if (type == MT_OOBDATA) 1426 flags |= MSG_OOB; 1427 } 1428 SBLASTRECORDCHK(&so->so_rcv, "soreceive 2"); 1429 SBLASTMBUFCHK(&so->so_rcv, "soreceive 2"); 1430 1431 moff = 0; 1432 offset = 0; 1433 while (m != NULL && uio->uio_resid > 0 && error == 0) { 1434 /* 1435 * If the type of mbuf has changed, end the receive 1436 * operation and do a short read. 1437 */ 1438 if (m->m_type == MT_OOBDATA) { 1439 if (type != MT_OOBDATA) 1440 break; 1441 } else if (type == MT_OOBDATA) { 1442 break; 1443 } else if (m->m_type == MT_CONTROL) { 1444 break; 1445 } 1446 #ifdef DIAGNOSTIC 1447 else if (m->m_type != MT_DATA && m->m_type != MT_HEADER) { 1448 panic("%s: m_type=%d", __func__, m->m_type); 1449 } 1450 #endif 1451 1452 so->so_state &= ~SS_RCVATMARK; 1453 wakeup_state = 0; 1454 len = uio->uio_resid; 1455 if (so->so_oobmark && len > so->so_oobmark - offset) 1456 len = so->so_oobmark - offset; 1457 if (len > m->m_len - moff) 1458 len = m->m_len - moff; 1459 1460 /* 1461 * If mp is set, just pass back the mbufs. 1462 * Otherwise copy them out via the uio, then free. 1463 * Sockbuf must be consistent here (points to current mbuf, 1464 * it points to next record) when we drop priority; 1465 * we must note any additions to the sockbuf when we 1466 * block interrupts again. 1467 */ 1468 if (mp == NULL) { 1469 SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); 1470 SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); 1471 sounlock(so); 1472 splx(s); 1473 error = uiomove(mtod(m, char *) + moff, len, uio); 1474 s = splsoftnet(); 1475 solock(so); 1476 if (error != 0) { 1477 /* 1478 * If any part of the record has been removed 1479 * (such as the MT_SONAME mbuf, which will 1480 * happen when PR_ADDR, and thus also 1481 * PR_ATOMIC, is set), then drop the entire 1482 * record to maintain the atomicity of the 1483 * receive operation. 1484 * 1485 * This avoids a later panic("receive 1a") 1486 * when compiled with DIAGNOSTIC. 1487 */ 1488 if (m && mbuf_removed && atomic) 1489 (void) sbdroprecord(&so->so_rcv); 1490 1491 goto release; 1492 } 1493 } else { 1494 uio->uio_resid -= len; 1495 } 1496 1497 if (len == m->m_len - moff) { 1498 if (m->m_flags & M_EOR) 1499 flags |= MSG_EOR; 1500 #ifdef SCTP 1501 if (m->m_flags & M_NOTIFICATION) 1502 flags |= MSG_NOTIFICATION; 1503 #endif 1504 if (flags & MSG_PEEK) { 1505 m = m->m_next; 1506 moff = 0; 1507 } else { 1508 nextrecord = m->m_nextpkt; 1509 sbfree(&so->so_rcv, m); 1510 if (mp) { 1511 *mp = m; 1512 mp = &m->m_next; 1513 so->so_rcv.sb_mb = m = m->m_next; 1514 *mp = NULL; 1515 } else { 1516 m = so->so_rcv.sb_mb = m_free(m); 1517 } 1518 /* 1519 * If m != NULL, we also know that 1520 * so->so_rcv.sb_mb != NULL. 1521 */ 1522 KASSERT(so->so_rcv.sb_mb == m); 1523 if (m) { 1524 m->m_nextpkt = nextrecord; 1525 if (nextrecord == NULL) 1526 so->so_rcv.sb_lastrecord = m; 1527 } else { 1528 so->so_rcv.sb_mb = nextrecord; 1529 SB_EMPTY_FIXUP(&so->so_rcv); 1530 } 1531 SBLASTRECORDCHK(&so->so_rcv, "soreceive 3"); 1532 SBLASTMBUFCHK(&so->so_rcv, "soreceive 3"); 1533 } 1534 } else if (flags & MSG_PEEK) { 1535 moff += len; 1536 } else { 1537 if (mp != NULL) { 1538 mt = m_copym(m, 0, len, M_NOWAIT); 1539 if (__predict_false(mt == NULL)) { 1540 sounlock(so); 1541 mt = m_copym(m, 0, len, M_WAIT); 1542 solock(so); 1543 } 1544 *mp = mt; 1545 } 1546 m->m_data += len; 1547 m->m_len -= len; 1548 so->so_rcv.sb_cc -= len; 1549 } 1550 1551 if (so->so_oobmark) { 1552 if ((flags & MSG_PEEK) == 0) { 1553 so->so_oobmark -= len; 1554 if (so->so_oobmark == 0) { 1555 so->so_state |= SS_RCVATMARK; 1556 break; 1557 } 1558 } else { 1559 offset += len; 1560 if (offset == so->so_oobmark) 1561 break; 1562 } 1563 } 1564 if (flags & MSG_EOR) 1565 break; 1566 1567 /* 1568 * If the MSG_WAITALL flag is set (for non-atomic socket), 1569 * we must not quit until "uio->uio_resid == 0" or an error 1570 * termination. If a signal/timeout occurs, return 1571 * with a short count but without error. 1572 * Keep sockbuf locked against other readers. 1573 */ 1574 while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 && 1575 !sosendallatonce(so) && !nextrecord) { 1576 if (so->so_error || so->so_rerror || 1577 so->so_state & SS_CANTRCVMORE) 1578 break; 1579 /* 1580 * If we are peeking and the socket receive buffer is 1581 * full, stop since we can't get more data to peek at. 1582 */ 1583 if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0) 1584 break; 1585 /* 1586 * If we've drained the socket buffer, tell the 1587 * protocol in case it needs to do something to 1588 * get it filled again. 1589 */ 1590 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) 1591 (*pr->pr_usrreqs->pr_rcvd)(so, flags, l); 1592 SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2"); 1593 SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2"); 1594 if (wakeup_state & SS_RESTARTSYS) 1595 error = ERESTART; 1596 else 1597 error = sbwait(&so->so_rcv); 1598 if (error != 0) { 1599 sbunlock(&so->so_rcv); 1600 sounlock(so); 1601 splx(s); 1602 return 0; 1603 } 1604 if ((m = so->so_rcv.sb_mb) != NULL) 1605 nextrecord = m->m_nextpkt; 1606 wakeup_state = so->so_state; 1607 } 1608 } 1609 1610 if (m && atomic) { 1611 flags |= MSG_TRUNC; 1612 if ((flags & MSG_PEEK) == 0) 1613 (void) sbdroprecord(&so->so_rcv); 1614 } 1615 if ((flags & MSG_PEEK) == 0) { 1616 if (m == NULL) { 1617 /* 1618 * First part is an inline SB_EMPTY_FIXUP(). Second 1619 * part makes sure sb_lastrecord is up-to-date if 1620 * there is still data in the socket buffer. 1621 */ 1622 so->so_rcv.sb_mb = nextrecord; 1623 if (so->so_rcv.sb_mb == NULL) { 1624 so->so_rcv.sb_mbtail = NULL; 1625 so->so_rcv.sb_lastrecord = NULL; 1626 } else if (nextrecord->m_nextpkt == NULL) 1627 so->so_rcv.sb_lastrecord = nextrecord; 1628 } 1629 SBLASTRECORDCHK(&so->so_rcv, "soreceive 4"); 1630 SBLASTMBUFCHK(&so->so_rcv, "soreceive 4"); 1631 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) 1632 (*pr->pr_usrreqs->pr_rcvd)(so, flags, l); 1633 } 1634 if (orig_resid == uio->uio_resid && orig_resid && 1635 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { 1636 sbunlock(&so->so_rcv); 1637 goto restart; 1638 } 1639 1640 if (flagsp != NULL) 1641 *flagsp |= flags; 1642 release: 1643 sbunlock(&so->so_rcv); 1644 sounlock(so); 1645 splx(s); 1646 return error; 1647 } 1648 1649 int 1650 soshutdown(struct socket *so, int how) 1651 { 1652 const struct protosw *pr; 1653 int error; 1654 1655 KASSERT(solocked(so)); 1656 1657 pr = so->so_proto; 1658 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) 1659 return EINVAL; 1660 1661 if (how == SHUT_RD || how == SHUT_RDWR) { 1662 sorflush(so); 1663 error = 0; 1664 } 1665 if (how == SHUT_WR || how == SHUT_RDWR) 1666 error = (*pr->pr_usrreqs->pr_shutdown)(so); 1667 1668 return error; 1669 } 1670 1671 void 1672 sorestart(struct socket *so) 1673 { 1674 /* 1675 * An application has called close() on an fd on which another 1676 * of its threads has called a socket system call. 1677 * Mark this and wake everyone up, and code that would block again 1678 * instead returns ERESTART. 1679 * On system call re-entry the fd is validated and EBADF returned. 1680 * Any other fd will block again on the 2nd syscall. 1681 */ 1682 solock(so); 1683 so->so_state |= SS_RESTARTSYS; 1684 cv_broadcast(&so->so_cv); 1685 cv_broadcast(&so->so_snd.sb_cv); 1686 cv_broadcast(&so->so_rcv.sb_cv); 1687 sounlock(so); 1688 } 1689 1690 void 1691 sorflush(struct socket *so) 1692 { 1693 struct sockbuf *sb, asb; 1694 const struct protosw *pr; 1695 1696 KASSERT(solocked(so)); 1697 1698 sb = &so->so_rcv; 1699 pr = so->so_proto; 1700 socantrcvmore(so); 1701 sb->sb_flags |= SB_NOINTR; 1702 (void )sblock(sb, M_WAITOK); 1703 sbunlock(sb); 1704 asb = *sb; 1705 /* 1706 * Clear most of the sockbuf structure, but leave some of the 1707 * fields valid. 1708 */ 1709 memset(&sb->sb_startzero, 0, 1710 sizeof(*sb) - offsetof(struct sockbuf, sb_startzero)); 1711 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) { 1712 sounlock(so); 1713 (*pr->pr_domain->dom_dispose)(asb.sb_mb); 1714 solock(so); 1715 } 1716 sbrelease(&asb, so); 1717 } 1718 1719 /* 1720 * internal set SOL_SOCKET options 1721 */ 1722 static int 1723 sosetopt1(struct socket *so, const struct sockopt *sopt) 1724 { 1725 int error, opt; 1726 int optval = 0; /* XXX: gcc */ 1727 struct linger l; 1728 struct timeval tv; 1729 1730 opt = sopt->sopt_name; 1731 1732 switch (opt) { 1733 1734 case SO_ACCEPTFILTER: 1735 error = accept_filt_setopt(so, sopt); 1736 KASSERT(solocked(so)); 1737 break; 1738 1739 case SO_LINGER: 1740 error = sockopt_get(sopt, &l, sizeof(l)); 1741 solock(so); 1742 if (error) 1743 break; 1744 if (l.l_linger < 0 || l.l_linger > USHRT_MAX || 1745 l.l_linger > (INT_MAX / hz)) { 1746 error = EDOM; 1747 break; 1748 } 1749 so->so_linger = l.l_linger; 1750 if (l.l_onoff) 1751 so->so_options |= SO_LINGER; 1752 else 1753 so->so_options &= ~SO_LINGER; 1754 break; 1755 1756 case SO_DEBUG: 1757 case SO_KEEPALIVE: 1758 case SO_DONTROUTE: 1759 case SO_USELOOPBACK: 1760 case SO_BROADCAST: 1761 case SO_REUSEADDR: 1762 case SO_REUSEPORT: 1763 case SO_OOBINLINE: 1764 case SO_TIMESTAMP: 1765 case SO_NOSIGPIPE: 1766 case SO_RERROR: 1767 error = sockopt_getint(sopt, &optval); 1768 solock(so); 1769 if (error) 1770 break; 1771 if (optval) 1772 so->so_options |= opt; 1773 else 1774 so->so_options &= ~opt; 1775 break; 1776 1777 case SO_SNDBUF: 1778 case SO_RCVBUF: 1779 case SO_SNDLOWAT: 1780 case SO_RCVLOWAT: 1781 error = sockopt_getint(sopt, &optval); 1782 solock(so); 1783 if (error) 1784 break; 1785 1786 /* 1787 * Values < 1 make no sense for any of these 1788 * options, so disallow them. 1789 */ 1790 if (optval < 1) { 1791 error = EINVAL; 1792 break; 1793 } 1794 1795 switch (opt) { 1796 case SO_SNDBUF: 1797 if (sbreserve(&so->so_snd, (u_long)optval, so) == 0) { 1798 error = ENOBUFS; 1799 break; 1800 } 1801 so->so_snd.sb_flags &= ~SB_AUTOSIZE; 1802 break; 1803 1804 case SO_RCVBUF: 1805 if (sbreserve(&so->so_rcv, (u_long)optval, so) == 0) { 1806 error = ENOBUFS; 1807 break; 1808 } 1809 so->so_rcv.sb_flags &= ~SB_AUTOSIZE; 1810 break; 1811 1812 /* 1813 * Make sure the low-water is never greater than 1814 * the high-water. 1815 */ 1816 case SO_SNDLOWAT: 1817 if (optval > so->so_snd.sb_hiwat) 1818 optval = so->so_snd.sb_hiwat; 1819 1820 so->so_snd.sb_lowat = optval; 1821 break; 1822 1823 case SO_RCVLOWAT: 1824 if (optval > so->so_rcv.sb_hiwat) 1825 optval = so->so_rcv.sb_hiwat; 1826 1827 so->so_rcv.sb_lowat = optval; 1828 break; 1829 } 1830 break; 1831 1832 case SO_SNDTIMEO: 1833 case SO_RCVTIMEO: 1834 solock(so); 1835 error = sockopt_get(sopt, &tv, sizeof(tv)); 1836 if (error) 1837 break; 1838 1839 if (tv.tv_sec < 0 || tv.tv_usec < 0 || tv.tv_usec >= 1000000) { 1840 error = EDOM; 1841 break; 1842 } 1843 if (tv.tv_sec > (INT_MAX - tv.tv_usec / tick) / hz) { 1844 error = EDOM; 1845 break; 1846 } 1847 1848 optval = tv.tv_sec * hz + tv.tv_usec / tick; 1849 if (optval == 0 && tv.tv_usec != 0) 1850 optval = 1; 1851 1852 switch (opt) { 1853 case SO_SNDTIMEO: 1854 so->so_snd.sb_timeo = optval; 1855 break; 1856 case SO_RCVTIMEO: 1857 so->so_rcv.sb_timeo = optval; 1858 break; 1859 } 1860 break; 1861 1862 default: 1863 MODULE_HOOK_CALL(uipc_socket_50_setopt1_hook, 1864 (opt, so, sopt), enosys(), error); 1865 if (error == ENOSYS || error == EPASSTHROUGH) { 1866 solock(so); 1867 error = ENOPROTOOPT; 1868 } 1869 break; 1870 } 1871 KASSERT(solocked(so)); 1872 return error; 1873 } 1874 1875 int 1876 sosetopt(struct socket *so, struct sockopt *sopt) 1877 { 1878 int error, prerr; 1879 1880 if (sopt->sopt_level == SOL_SOCKET) { 1881 error = sosetopt1(so, sopt); 1882 KASSERT(solocked(so)); 1883 } else { 1884 error = ENOPROTOOPT; 1885 solock(so); 1886 } 1887 1888 if ((error == 0 || error == ENOPROTOOPT) && 1889 so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) { 1890 /* give the protocol stack a shot */ 1891 prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, sopt); 1892 if (prerr == 0) 1893 error = 0; 1894 else if (prerr != ENOPROTOOPT) 1895 error = prerr; 1896 } 1897 sounlock(so); 1898 return error; 1899 } 1900 1901 /* 1902 * so_setsockopt() is a wrapper providing a sockopt structure for sosetopt() 1903 */ 1904 int 1905 so_setsockopt(struct lwp *l, struct socket *so, int level, int name, 1906 const void *val, size_t valsize) 1907 { 1908 struct sockopt sopt; 1909 int error; 1910 1911 KASSERT(valsize == 0 || val != NULL); 1912 1913 sockopt_init(&sopt, level, name, valsize); 1914 sockopt_set(&sopt, val, valsize); 1915 1916 error = sosetopt(so, &sopt); 1917 1918 sockopt_destroy(&sopt); 1919 1920 return error; 1921 } 1922 1923 /* 1924 * internal get SOL_SOCKET options 1925 */ 1926 static int 1927 sogetopt1(struct socket *so, struct sockopt *sopt) 1928 { 1929 int error, optval, opt; 1930 struct linger l; 1931 struct timeval tv; 1932 1933 switch ((opt = sopt->sopt_name)) { 1934 1935 case SO_ACCEPTFILTER: 1936 error = accept_filt_getopt(so, sopt); 1937 break; 1938 1939 case SO_LINGER: 1940 l.l_onoff = (so->so_options & SO_LINGER) ? 1 : 0; 1941 l.l_linger = so->so_linger; 1942 1943 error = sockopt_set(sopt, &l, sizeof(l)); 1944 break; 1945 1946 case SO_USELOOPBACK: 1947 case SO_DONTROUTE: 1948 case SO_DEBUG: 1949 case SO_KEEPALIVE: 1950 case SO_REUSEADDR: 1951 case SO_REUSEPORT: 1952 case SO_BROADCAST: 1953 case SO_OOBINLINE: 1954 case SO_TIMESTAMP: 1955 case SO_NOSIGPIPE: 1956 case SO_RERROR: 1957 case SO_ACCEPTCONN: 1958 error = sockopt_setint(sopt, (so->so_options & opt) ? 1 : 0); 1959 break; 1960 1961 case SO_TYPE: 1962 error = sockopt_setint(sopt, so->so_type); 1963 break; 1964 1965 case SO_ERROR: 1966 if (so->so_error == 0) { 1967 so->so_error = so->so_rerror; 1968 so->so_rerror = 0; 1969 } 1970 error = sockopt_setint(sopt, so->so_error); 1971 so->so_error = 0; 1972 break; 1973 1974 case SO_SNDBUF: 1975 error = sockopt_setint(sopt, so->so_snd.sb_hiwat); 1976 break; 1977 1978 case SO_RCVBUF: 1979 error = sockopt_setint(sopt, so->so_rcv.sb_hiwat); 1980 break; 1981 1982 case SO_SNDLOWAT: 1983 error = sockopt_setint(sopt, so->so_snd.sb_lowat); 1984 break; 1985 1986 case SO_RCVLOWAT: 1987 error = sockopt_setint(sopt, so->so_rcv.sb_lowat); 1988 break; 1989 1990 case SO_SNDTIMEO: 1991 case SO_RCVTIMEO: 1992 optval = (opt == SO_SNDTIMEO ? 1993 so->so_snd.sb_timeo : so->so_rcv.sb_timeo); 1994 1995 memset(&tv, 0, sizeof(tv)); 1996 tv.tv_sec = optval / hz; 1997 tv.tv_usec = (optval % hz) * tick; 1998 1999 error = sockopt_set(sopt, &tv, sizeof(tv)); 2000 break; 2001 2002 case SO_OVERFLOWED: 2003 error = sockopt_setint(sopt, so->so_rcv.sb_overflowed); 2004 break; 2005 2006 default: 2007 MODULE_HOOK_CALL(uipc_socket_50_getopt1_hook, 2008 (opt, so, sopt), enosys(), error); 2009 if (error) 2010 error = ENOPROTOOPT; 2011 break; 2012 } 2013 2014 return error; 2015 } 2016 2017 int 2018 sogetopt(struct socket *so, struct sockopt *sopt) 2019 { 2020 int error; 2021 2022 solock(so); 2023 if (sopt->sopt_level != SOL_SOCKET) { 2024 if (so->so_proto && so->so_proto->pr_ctloutput) { 2025 error = ((*so->so_proto->pr_ctloutput) 2026 (PRCO_GETOPT, so, sopt)); 2027 } else 2028 error = (ENOPROTOOPT); 2029 } else { 2030 error = sogetopt1(so, sopt); 2031 } 2032 sounlock(so); 2033 return error; 2034 } 2035 2036 /* 2037 * alloc sockopt data buffer buffer 2038 * - will be released at destroy 2039 */ 2040 static int 2041 sockopt_alloc(struct sockopt *sopt, size_t len, km_flag_t kmflag) 2042 { 2043 2044 KASSERT(sopt->sopt_size == 0); 2045 2046 if (len > sizeof(sopt->sopt_buf)) { 2047 sopt->sopt_data = kmem_zalloc(len, kmflag); 2048 if (sopt->sopt_data == NULL) 2049 return ENOMEM; 2050 } else 2051 sopt->sopt_data = sopt->sopt_buf; 2052 2053 sopt->sopt_size = len; 2054 return 0; 2055 } 2056 2057 /* 2058 * initialise sockopt storage 2059 * - MAY sleep during allocation 2060 */ 2061 void 2062 sockopt_init(struct sockopt *sopt, int level, int name, size_t size) 2063 { 2064 2065 memset(sopt, 0, sizeof(*sopt)); 2066 2067 sopt->sopt_level = level; 2068 sopt->sopt_name = name; 2069 (void)sockopt_alloc(sopt, size, KM_SLEEP); 2070 } 2071 2072 /* 2073 * destroy sockopt storage 2074 * - will release any held memory references 2075 */ 2076 void 2077 sockopt_destroy(struct sockopt *sopt) 2078 { 2079 2080 if (sopt->sopt_data != sopt->sopt_buf) 2081 kmem_free(sopt->sopt_data, sopt->sopt_size); 2082 2083 memset(sopt, 0, sizeof(*sopt)); 2084 } 2085 2086 /* 2087 * set sockopt value 2088 * - value is copied into sockopt 2089 * - memory is allocated when necessary, will not sleep 2090 */ 2091 int 2092 sockopt_set(struct sockopt *sopt, const void *buf, size_t len) 2093 { 2094 int error; 2095 2096 if (sopt->sopt_size == 0) { 2097 error = sockopt_alloc(sopt, len, KM_NOSLEEP); 2098 if (error) 2099 return error; 2100 } 2101 2102 sopt->sopt_retsize = MIN(sopt->sopt_size, len); 2103 if (sopt->sopt_retsize > 0) { 2104 memcpy(sopt->sopt_data, buf, sopt->sopt_retsize); 2105 } 2106 2107 return 0; 2108 } 2109 2110 /* 2111 * common case of set sockopt integer value 2112 */ 2113 int 2114 sockopt_setint(struct sockopt *sopt, int val) 2115 { 2116 2117 return sockopt_set(sopt, &val, sizeof(int)); 2118 } 2119 2120 /* 2121 * get sockopt value 2122 * - correct size must be given 2123 */ 2124 int 2125 sockopt_get(const struct sockopt *sopt, void *buf, size_t len) 2126 { 2127 2128 if (sopt->sopt_size != len) 2129 return EINVAL; 2130 2131 memcpy(buf, sopt->sopt_data, len); 2132 return 0; 2133 } 2134 2135 /* 2136 * common case of get sockopt integer value 2137 */ 2138 int 2139 sockopt_getint(const struct sockopt *sopt, int *valp) 2140 { 2141 2142 return sockopt_get(sopt, valp, sizeof(int)); 2143 } 2144 2145 /* 2146 * set sockopt value from mbuf 2147 * - ONLY for legacy code 2148 * - mbuf is released by sockopt 2149 * - will not sleep 2150 */ 2151 int 2152 sockopt_setmbuf(struct sockopt *sopt, struct mbuf *m) 2153 { 2154 size_t len; 2155 int error; 2156 2157 len = m_length(m); 2158 2159 if (sopt->sopt_size == 0) { 2160 error = sockopt_alloc(sopt, len, KM_NOSLEEP); 2161 if (error) 2162 return error; 2163 } 2164 2165 sopt->sopt_retsize = MIN(sopt->sopt_size, len); 2166 m_copydata(m, 0, sopt->sopt_retsize, sopt->sopt_data); 2167 m_freem(m); 2168 2169 return 0; 2170 } 2171 2172 /* 2173 * get sockopt value into mbuf 2174 * - ONLY for legacy code 2175 * - mbuf to be released by the caller 2176 * - will not sleep 2177 */ 2178 struct mbuf * 2179 sockopt_getmbuf(const struct sockopt *sopt) 2180 { 2181 struct mbuf *m; 2182 2183 if (sopt->sopt_size > MCLBYTES) 2184 return NULL; 2185 2186 m = m_get(M_DONTWAIT, MT_SOOPTS); 2187 if (m == NULL) 2188 return NULL; 2189 2190 if (sopt->sopt_size > MLEN) { 2191 MCLGET(m, M_DONTWAIT); 2192 if ((m->m_flags & M_EXT) == 0) { 2193 m_free(m); 2194 return NULL; 2195 } 2196 } 2197 2198 memcpy(mtod(m, void *), sopt->sopt_data, sopt->sopt_size); 2199 m->m_len = sopt->sopt_size; 2200 2201 return m; 2202 } 2203 2204 void 2205 sohasoutofband(struct socket *so) 2206 { 2207 2208 fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so); 2209 selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, NOTE_SUBMIT); 2210 } 2211 2212 static void 2213 filt_sordetach(struct knote *kn) 2214 { 2215 struct socket *so; 2216 2217 so = ((file_t *)kn->kn_obj)->f_socket; 2218 solock(so); 2219 SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext); 2220 if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist)) 2221 so->so_rcv.sb_flags &= ~SB_KNOTE; 2222 sounlock(so); 2223 } 2224 2225 /*ARGSUSED*/ 2226 static int 2227 filt_soread(struct knote *kn, long hint) 2228 { 2229 struct socket *so; 2230 int rv; 2231 2232 so = ((file_t *)kn->kn_obj)->f_socket; 2233 if (hint != NOTE_SUBMIT) 2234 solock(so); 2235 kn->kn_data = so->so_rcv.sb_cc; 2236 if (so->so_state & SS_CANTRCVMORE) { 2237 kn->kn_flags |= EV_EOF; 2238 kn->kn_fflags = so->so_error; 2239 rv = 1; 2240 } else if (so->so_error || so->so_rerror) 2241 rv = 1; 2242 else if (kn->kn_sfflags & NOTE_LOWAT) 2243 rv = (kn->kn_data >= kn->kn_sdata); 2244 else 2245 rv = (kn->kn_data >= so->so_rcv.sb_lowat); 2246 if (hint != NOTE_SUBMIT) 2247 sounlock(so); 2248 return rv; 2249 } 2250 2251 static void 2252 filt_sowdetach(struct knote *kn) 2253 { 2254 struct socket *so; 2255 2256 so = ((file_t *)kn->kn_obj)->f_socket; 2257 solock(so); 2258 SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext); 2259 if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist)) 2260 so->so_snd.sb_flags &= ~SB_KNOTE; 2261 sounlock(so); 2262 } 2263 2264 /*ARGSUSED*/ 2265 static int 2266 filt_sowrite(struct knote *kn, long hint) 2267 { 2268 struct socket *so; 2269 int rv; 2270 2271 so = ((file_t *)kn->kn_obj)->f_socket; 2272 if (hint != NOTE_SUBMIT) 2273 solock(so); 2274 kn->kn_data = sbspace(&so->so_snd); 2275 if (so->so_state & SS_CANTSENDMORE) { 2276 kn->kn_flags |= EV_EOF; 2277 kn->kn_fflags = so->so_error; 2278 rv = 1; 2279 } else if (so->so_error) 2280 rv = 1; 2281 else if (((so->so_state & SS_ISCONNECTED) == 0) && 2282 (so->so_proto->pr_flags & PR_CONNREQUIRED)) 2283 rv = 0; 2284 else if (kn->kn_sfflags & NOTE_LOWAT) 2285 rv = (kn->kn_data >= kn->kn_sdata); 2286 else 2287 rv = (kn->kn_data >= so->so_snd.sb_lowat); 2288 if (hint != NOTE_SUBMIT) 2289 sounlock(so); 2290 return rv; 2291 } 2292 2293 /*ARGSUSED*/ 2294 static int 2295 filt_solisten(struct knote *kn, long hint) 2296 { 2297 struct socket *so; 2298 int rv; 2299 2300 so = ((file_t *)kn->kn_obj)->f_socket; 2301 2302 /* 2303 * Set kn_data to number of incoming connections, not 2304 * counting partial (incomplete) connections. 2305 */ 2306 if (hint != NOTE_SUBMIT) 2307 solock(so); 2308 kn->kn_data = so->so_qlen; 2309 rv = (kn->kn_data > 0); 2310 if (hint != NOTE_SUBMIT) 2311 sounlock(so); 2312 return rv; 2313 } 2314 2315 static const struct filterops solisten_filtops = { 2316 .f_isfd = 1, 2317 .f_attach = NULL, 2318 .f_detach = filt_sordetach, 2319 .f_event = filt_solisten, 2320 }; 2321 2322 static const struct filterops soread_filtops = { 2323 .f_isfd = 1, 2324 .f_attach = NULL, 2325 .f_detach = filt_sordetach, 2326 .f_event = filt_soread, 2327 }; 2328 2329 static const struct filterops sowrite_filtops = { 2330 .f_isfd = 1, 2331 .f_attach = NULL, 2332 .f_detach = filt_sowdetach, 2333 .f_event = filt_sowrite, 2334 }; 2335 2336 int 2337 soo_kqfilter(struct file *fp, struct knote *kn) 2338 { 2339 struct socket *so; 2340 struct sockbuf *sb; 2341 2342 so = ((file_t *)kn->kn_obj)->f_socket; 2343 solock(so); 2344 switch (kn->kn_filter) { 2345 case EVFILT_READ: 2346 if (so->so_options & SO_ACCEPTCONN) 2347 kn->kn_fop = &solisten_filtops; 2348 else 2349 kn->kn_fop = &soread_filtops; 2350 sb = &so->so_rcv; 2351 break; 2352 case EVFILT_WRITE: 2353 kn->kn_fop = &sowrite_filtops; 2354 sb = &so->so_snd; 2355 break; 2356 default: 2357 sounlock(so); 2358 return EINVAL; 2359 } 2360 SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext); 2361 sb->sb_flags |= SB_KNOTE; 2362 sounlock(so); 2363 return 0; 2364 } 2365 2366 static int 2367 sodopoll(struct socket *so, int events) 2368 { 2369 int revents; 2370 2371 revents = 0; 2372 2373 if (events & (POLLIN | POLLRDNORM)) 2374 if (soreadable(so)) 2375 revents |= events & (POLLIN | POLLRDNORM); 2376 2377 if (events & (POLLOUT | POLLWRNORM)) 2378 if (sowritable(so)) 2379 revents |= events & (POLLOUT | POLLWRNORM); 2380 2381 if (events & (POLLPRI | POLLRDBAND)) 2382 if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) 2383 revents |= events & (POLLPRI | POLLRDBAND); 2384 2385 return revents; 2386 } 2387 2388 int 2389 sopoll(struct socket *so, int events) 2390 { 2391 int revents = 0; 2392 2393 #ifndef DIAGNOSTIC 2394 /* 2395 * Do a quick, unlocked check in expectation that the socket 2396 * will be ready for I/O. Don't do this check if DIAGNOSTIC, 2397 * as the solocked() assertions will fail. 2398 */ 2399 if ((revents = sodopoll(so, events)) != 0) 2400 return revents; 2401 #endif 2402 2403 solock(so); 2404 if ((revents = sodopoll(so, events)) == 0) { 2405 if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) { 2406 selrecord(curlwp, &so->so_rcv.sb_sel); 2407 so->so_rcv.sb_flags |= SB_NOTIFY; 2408 } 2409 2410 if (events & (POLLOUT | POLLWRNORM)) { 2411 selrecord(curlwp, &so->so_snd.sb_sel); 2412 so->so_snd.sb_flags |= SB_NOTIFY; 2413 } 2414 } 2415 sounlock(so); 2416 2417 return revents; 2418 } 2419 2420 struct mbuf ** 2421 sbsavetimestamp(int opt, struct mbuf **mp) 2422 { 2423 struct timeval tv; 2424 int error; 2425 2426 microtime(&tv); 2427 2428 MODULE_HOOK_CALL(uipc_socket_50_sbts_hook, (opt, &mp), enosys(), error); 2429 if (error == 0) 2430 return mp; 2431 2432 if (opt & SO_TIMESTAMP) { 2433 *mp = sbcreatecontrol(&tv, sizeof(tv), 2434 SCM_TIMESTAMP, SOL_SOCKET); 2435 if (*mp) 2436 mp = &(*mp)->m_next; 2437 } 2438 return mp; 2439 } 2440 2441 2442 #include <sys/sysctl.h> 2443 2444 static int sysctl_kern_somaxkva(SYSCTLFN_PROTO); 2445 static int sysctl_kern_sbmax(SYSCTLFN_PROTO); 2446 2447 /* 2448 * sysctl helper routine for kern.somaxkva. ensures that the given 2449 * value is not too small. 2450 * (XXX should we maybe make sure it's not too large as well?) 2451 */ 2452 static int 2453 sysctl_kern_somaxkva(SYSCTLFN_ARGS) 2454 { 2455 int error, new_somaxkva; 2456 struct sysctlnode node; 2457 2458 new_somaxkva = somaxkva; 2459 node = *rnode; 2460 node.sysctl_data = &new_somaxkva; 2461 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2462 if (error || newp == NULL) 2463 return error; 2464 2465 if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */ 2466 return EINVAL; 2467 2468 mutex_enter(&so_pendfree_lock); 2469 somaxkva = new_somaxkva; 2470 cv_broadcast(&socurkva_cv); 2471 mutex_exit(&so_pendfree_lock); 2472 2473 return error; 2474 } 2475 2476 /* 2477 * sysctl helper routine for kern.sbmax. Basically just ensures that 2478 * any new value is not too small. 2479 */ 2480 static int 2481 sysctl_kern_sbmax(SYSCTLFN_ARGS) 2482 { 2483 int error, new_sbmax; 2484 struct sysctlnode node; 2485 2486 new_sbmax = sb_max; 2487 node = *rnode; 2488 node.sysctl_data = &new_sbmax; 2489 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2490 if (error || newp == NULL) 2491 return error; 2492 2493 KERNEL_LOCK(1, NULL); 2494 error = sb_max_set(new_sbmax); 2495 KERNEL_UNLOCK_ONE(NULL); 2496 2497 return error; 2498 } 2499 2500 /* 2501 * sysctl helper routine for kern.sooptions. Ensures that only allowed 2502 * options can be set. 2503 */ 2504 static int 2505 sysctl_kern_sooptions(SYSCTLFN_ARGS) 2506 { 2507 int error, new_options; 2508 struct sysctlnode node; 2509 2510 new_options = sooptions; 2511 node = *rnode; 2512 node.sysctl_data = &new_options; 2513 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2514 if (error || newp == NULL) 2515 return error; 2516 2517 if (new_options & ~SO_DEFOPTS) 2518 return EINVAL; 2519 2520 sooptions = new_options; 2521 2522 return 0; 2523 } 2524 2525 static void 2526 sysctl_kern_socket_setup(void) 2527 { 2528 2529 KASSERT(socket_sysctllog == NULL); 2530 2531 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2532 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2533 CTLTYPE_INT, "somaxkva", 2534 SYSCTL_DESCR("Maximum amount of kernel memory to be " 2535 "used for socket buffers"), 2536 sysctl_kern_somaxkva, 0, NULL, 0, 2537 CTL_KERN, KERN_SOMAXKVA, CTL_EOL); 2538 2539 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2540 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2541 CTLTYPE_INT, "sbmax", 2542 SYSCTL_DESCR("Maximum socket buffer size"), 2543 sysctl_kern_sbmax, 0, NULL, 0, 2544 CTL_KERN, KERN_SBMAX, CTL_EOL); 2545 2546 sysctl_createv(&socket_sysctllog, 0, NULL, NULL, 2547 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2548 CTLTYPE_INT, "sooptions", 2549 SYSCTL_DESCR("Default socket options"), 2550 sysctl_kern_sooptions, 0, NULL, 0, 2551 CTL_KERN, CTL_CREATE, CTL_EOL); 2552 } 2553