1 /* 2 * Copyright (c) 2005 Jeffrey M. Hsu. All rights reserved. 3 * Copyright (c) 1982, 1986, 1988, 1990, 1993 4 * The Regents of the University of California. All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. All advertising materials mentioning features or use of this software 15 * must display the following acknowledgement: 16 * This product includes software developed by the University of 17 * California, Berkeley and its contributors. 18 * 4. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)uipc_socket2.c 8.1 (Berkeley) 6/10/93 35 * $FreeBSD: src/sys/kern/uipc_socket2.c,v 1.55.2.17 2002/08/31 19:04:55 dwmalone Exp $ 36 * $DragonFly: src/sys/kern/uipc_socket2.c,v 1.33 2008/09/02 16:17:52 dillon Exp $ 37 */ 38 39 #include "opt_param.h" 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/domain.h> 43 #include <sys/file.h> /* for maxfiles */ 44 #include <sys/kernel.h> 45 #include <sys/proc.h> 46 #include <sys/malloc.h> 47 #include <sys/mbuf.h> 48 #include <sys/protosw.h> 49 #include <sys/resourcevar.h> 50 #include <sys/stat.h> 51 #include <sys/socket.h> 52 #include <sys/socketvar.h> 53 #include <sys/signalvar.h> 54 #include <sys/sysctl.h> 55 #include <sys/aio.h> /* for aio_swake proto */ 56 #include <sys/event.h> 57 58 #include <sys/thread2.h> 59 #include <sys/msgport2.h> 60 61 int maxsockets; 62 63 /* 64 * Primitive routines for operating on sockets and socket buffers 65 */ 66 67 u_long sb_max = SB_MAX; 68 u_long sb_max_adj = 69 SB_MAX * MCLBYTES / (MSIZE + MCLBYTES); /* adjusted sb_max */ 70 71 static u_long sb_efficiency = 8; /* parameter for sbreserve() */ 72 73 /************************************************************************ 74 * signalsockbuf procedures * 75 ************************************************************************/ 76 77 /* 78 * Wait for data to arrive at/drain from a socket buffer. 79 */ 80 int 81 ssb_wait(struct signalsockbuf *ssb) 82 { 83 84 ssb->ssb_flags |= SSB_WAIT; 85 return (tsleep((caddr_t)&ssb->ssb_cc, 86 ((ssb->ssb_flags & SSB_NOINTR) ? 0 : PCATCH), 87 "sbwait", 88 ssb->ssb_timeo)); 89 } 90 91 /* 92 * Lock a sockbuf already known to be locked; 93 * return any error returned from sleep (EINTR). 94 */ 95 int 96 _ssb_lock(struct signalsockbuf *ssb) 97 { 98 int error; 99 100 while (ssb->ssb_flags & SSB_LOCK) { 101 ssb->ssb_flags |= SSB_WANT; 102 error = tsleep((caddr_t)&ssb->ssb_flags, 103 ((ssb->ssb_flags & SSB_NOINTR) ? 0 : PCATCH), 104 "sblock", 0); 105 if (error) 106 return (error); 107 } 108 ssb->ssb_flags |= SSB_LOCK; 109 return (0); 110 } 111 112 /* 113 * This does the same for sockbufs. Note that the xsockbuf structure, 114 * since it is always embedded in a socket, does not include a self 115 * pointer nor a length. We make this entry point public in case 116 * some other mechanism needs it. 117 */ 118 void 119 ssbtoxsockbuf(struct signalsockbuf *ssb, struct xsockbuf *xsb) 120 { 121 xsb->sb_cc = ssb->ssb_cc; 122 xsb->sb_hiwat = ssb->ssb_hiwat; 123 xsb->sb_mbcnt = ssb->ssb_mbcnt; 124 xsb->sb_mbmax = ssb->ssb_mbmax; 125 xsb->sb_lowat = ssb->ssb_lowat; 126 xsb->sb_flags = ssb->ssb_flags; 127 xsb->sb_timeo = ssb->ssb_timeo; 128 } 129 130 131 /************************************************************************ 132 * Procedures which manipulate socket state flags, wakeups, etc. * 133 ************************************************************************ 134 * 135 * Normal sequence from the active (originating) side is that 136 * soisconnecting() is called during processing of connect() call, resulting 137 * in an eventual call to soisconnected() if/when the connection is 138 * established. When the connection is torn down soisdisconnecting() is 139 * called during processing of disconnect() call, and soisdisconnected() is 140 * called when the connection to the peer is totally severed. 141 * 142 * The semantics of these routines are such that connectionless protocols 143 * can call soisconnected() and soisdisconnected() only, bypassing the 144 * in-progress calls when setting up a ``connection'' takes no time. 145 * 146 * From the passive side, a socket is created with two queues of sockets: 147 * so_incomp for connections in progress and so_comp for connections 148 * already made and awaiting user acceptance. As a protocol is preparing 149 * incoming connections, it creates a socket structure queued on so_incomp 150 * by calling sonewconn(). When the connection is established, 151 * soisconnected() is called, and transfers the socket structure to so_comp, 152 * making it available to accept(). 153 * 154 * If a socket is closed with sockets on either so_incomp or so_comp, these 155 * sockets are dropped. 156 * 157 * If higher level protocols are implemented in the kernel, the wakeups 158 * done here will sometimes cause software-interrupt process scheduling. 159 */ 160 161 void 162 soisconnecting(struct socket *so) 163 { 164 so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING); 165 so->so_state |= SS_ISCONNECTING; 166 } 167 168 void 169 soisconnected(struct socket *so) 170 { 171 struct socket *head = so->so_head; 172 173 so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING); 174 so->so_state |= SS_ISCONNECTED; 175 if (head && (so->so_state & SS_INCOMP)) { 176 if ((so->so_options & SO_ACCEPTFILTER) != 0) { 177 so->so_upcall = head->so_accf->so_accept_filter->accf_callback; 178 so->so_upcallarg = head->so_accf->so_accept_filter_arg; 179 so->so_rcv.ssb_flags |= SSB_UPCALL; 180 so->so_options &= ~SO_ACCEPTFILTER; 181 so->so_upcall(so, so->so_upcallarg, 0); 182 return; 183 } 184 TAILQ_REMOVE(&head->so_incomp, so, so_list); 185 head->so_incqlen--; 186 so->so_state &= ~SS_INCOMP; 187 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list); 188 head->so_qlen++; 189 so->so_state |= SS_COMP; 190 sorwakeup(head); 191 wakeup_one(&head->so_timeo); 192 } else { 193 wakeup(&so->so_timeo); 194 sorwakeup(so); 195 sowwakeup(so); 196 } 197 } 198 199 void 200 soisdisconnecting(struct socket *so) 201 { 202 so->so_state &= ~SS_ISCONNECTING; 203 so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE); 204 wakeup((caddr_t)&so->so_timeo); 205 sowwakeup(so); 206 sorwakeup(so); 207 } 208 209 void 210 soisdisconnected(struct socket *so) 211 { 212 so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING); 213 so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED); 214 wakeup((caddr_t)&so->so_timeo); 215 sbdrop(&so->so_snd.sb, so->so_snd.ssb_cc); 216 sowwakeup(so); 217 sorwakeup(so); 218 } 219 220 /* 221 * Set or change the message port a socket receives commands on. 222 * 223 * XXX 224 */ 225 void 226 sosetport(struct socket *so, lwkt_port_t port) 227 { 228 so->so_port = port; 229 } 230 231 /* 232 * When an attempt at a new connection is noted on a socket 233 * which accepts connections, sonewconn is called. If the 234 * connection is possible (subject to space constraints, etc.) 235 * then we allocate a new structure, propoerly linked into the 236 * data structure of the original socket, and return this. 237 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED. 238 */ 239 struct socket * 240 sonewconn(struct socket *head, int connstatus) 241 { 242 struct socket *so; 243 struct socket *sp; 244 struct pru_attach_info ai; 245 246 if (head->so_qlen > 3 * head->so_qlimit / 2) 247 return (NULL); 248 so = soalloc(1); 249 if (so == NULL) 250 return (NULL); 251 if ((head->so_options & SO_ACCEPTFILTER) != 0) 252 connstatus = 0; 253 so->so_head = head; 254 so->so_type = head->so_type; 255 so->so_options = head->so_options &~ SO_ACCEPTCONN; 256 so->so_linger = head->so_linger; 257 so->so_state = head->so_state | SS_NOFDREF; 258 so->so_proto = head->so_proto; 259 so->so_cred = crhold(head->so_cred); 260 ai.sb_rlimit = NULL; 261 ai.p_ucred = NULL; 262 ai.fd_rdir = NULL; /* jail code cruft XXX JH */ 263 if (soreserve(so, head->so_snd.ssb_hiwat, head->so_rcv.ssb_hiwat, NULL) || 264 /* Directly call function since we're already at protocol level. */ 265 (*so->so_proto->pr_usrreqs->pru_attach)(so, 0, &ai)) { 266 sodealloc(so); 267 return (NULL); 268 } 269 KKASSERT(so->so_port != NULL); 270 so->so_rcv.ssb_lowat = head->so_rcv.ssb_lowat; 271 so->so_snd.ssb_lowat = head->so_snd.ssb_lowat; 272 so->so_rcv.ssb_timeo = head->so_rcv.ssb_timeo; 273 so->so_snd.ssb_timeo = head->so_snd.ssb_timeo; 274 so->so_rcv.ssb_flags |= head->so_rcv.ssb_flags & 275 (SSB_AUTOSIZE | SSB_AUTOLOWAT); 276 so->so_snd.ssb_flags |= head->so_snd.ssb_flags & 277 (SSB_AUTOSIZE | SSB_AUTOLOWAT); 278 if (connstatus) { 279 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list); 280 so->so_state |= SS_COMP; 281 head->so_qlen++; 282 } else { 283 if (head->so_incqlen > head->so_qlimit) { 284 sp = TAILQ_FIRST(&head->so_incomp); 285 TAILQ_REMOVE(&head->so_incomp, sp, so_list); 286 head->so_incqlen--; 287 sp->so_state &= ~SS_INCOMP; 288 sp->so_head = NULL; 289 soaborta(sp); 290 } 291 TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list); 292 so->so_state |= SS_INCOMP; 293 head->so_incqlen++; 294 } 295 if (connstatus) { 296 sorwakeup(head); 297 wakeup((caddr_t)&head->so_timeo); 298 so->so_state |= connstatus; 299 } 300 return (so); 301 } 302 303 /* 304 * Socantsendmore indicates that no more data will be sent on the 305 * socket; it would normally be applied to a socket when the user 306 * informs the system that no more data is to be sent, by the protocol 307 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data 308 * will be received, and will normally be applied to the socket by a 309 * protocol when it detects that the peer will send no more data. 310 * Data queued for reading in the socket may yet be read. 311 */ 312 void 313 socantsendmore(struct socket *so) 314 { 315 so->so_state |= SS_CANTSENDMORE; 316 sowwakeup(so); 317 } 318 319 void 320 socantrcvmore(struct socket *so) 321 { 322 so->so_state |= SS_CANTRCVMORE; 323 sorwakeup(so); 324 } 325 326 /* 327 * Wakeup processes waiting on a socket buffer. Do asynchronous notification 328 * via SIGIO if the socket has the SS_ASYNC flag set. 329 * 330 * For users waiting on send/recv try to avoid unnecessary context switch 331 * thrashing. Particularly for senders of large buffers (needs to be 332 * extended to sel and aio? XXX) 333 */ 334 void 335 sowakeup(struct socket *so, struct signalsockbuf *ssb) 336 { 337 struct selinfo *selinfo = &ssb->ssb_sel; 338 339 selwakeup(selinfo); 340 ssb->ssb_flags &= ~SSB_SEL; 341 if (ssb->ssb_flags & SSB_WAIT) { 342 if ((ssb == &so->so_snd && ssb_space(ssb) >= ssb->ssb_lowat) || 343 (ssb == &so->so_rcv && ssb->ssb_cc >= ssb->ssb_lowat) || 344 (ssb == &so->so_snd && (so->so_state & SS_CANTSENDMORE)) || 345 (ssb == &so->so_rcv && (so->so_state & SS_CANTRCVMORE)) 346 ) { 347 ssb->ssb_flags &= ~SSB_WAIT; 348 wakeup((caddr_t)&ssb->ssb_cc); 349 } 350 } 351 if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL) 352 pgsigio(so->so_sigio, SIGIO, 0); 353 if (ssb->ssb_flags & SSB_UPCALL) 354 (*so->so_upcall)(so, so->so_upcallarg, MB_DONTWAIT); 355 if (ssb->ssb_flags & SSB_AIO) 356 aio_swake(so, ssb); 357 KNOTE(&selinfo->si_note, 0); 358 if (ssb->ssb_flags & SSB_MEVENT) { 359 struct netmsg_so_notify *msg, *nmsg; 360 361 TAILQ_FOREACH_MUTABLE(msg, &selinfo->si_mlist, nm_list, nmsg) { 362 if (msg->nm_predicate(&msg->nm_netmsg)) { 363 TAILQ_REMOVE(&selinfo->si_mlist, msg, nm_list); 364 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 365 msg->nm_netmsg.nm_lmsg.ms_error); 366 } 367 } 368 if (TAILQ_EMPTY(&ssb->ssb_sel.si_mlist)) 369 ssb->ssb_flags &= ~SSB_MEVENT; 370 } 371 } 372 373 /* 374 * Socket buffer (struct signalsockbuf) utility routines. 375 * 376 * Each socket contains two socket buffers: one for sending data and 377 * one for receiving data. Each buffer contains a queue of mbufs, 378 * information about the number of mbufs and amount of data in the 379 * queue, and other fields allowing select() statements and notification 380 * on data availability to be implemented. 381 * 382 * Data stored in a socket buffer is maintained as a list of records. 383 * Each record is a list of mbufs chained together with the m_next 384 * field. Records are chained together with the m_nextpkt field. The upper 385 * level routine soreceive() expects the following conventions to be 386 * observed when placing information in the receive buffer: 387 * 388 * 1. If the protocol requires each message be preceded by the sender's 389 * name, then a record containing that name must be present before 390 * any associated data (mbuf's must be of type MT_SONAME). 391 * 2. If the protocol supports the exchange of ``access rights'' (really 392 * just additional data associated with the message), and there are 393 * ``rights'' to be received, then a record containing this data 394 * should be present (mbuf's must be of type MT_RIGHTS). 395 * 3. If a name or rights record exists, then it must be followed by 396 * a data record, perhaps of zero length. 397 * 398 * Before using a new socket structure it is first necessary to reserve 399 * buffer space to the socket, by calling sbreserve(). This should commit 400 * some of the available buffer space in the system buffer pool for the 401 * socket (currently, it does nothing but enforce limits). The space 402 * should be released by calling ssb_release() when the socket is destroyed. 403 */ 404 int 405 soreserve(struct socket *so, u_long sndcc, u_long rcvcc, struct rlimit *rl) 406 { 407 if (so->so_snd.ssb_lowat == 0) 408 so->so_snd.ssb_flags |= SSB_AUTOLOWAT; 409 if (ssb_reserve(&so->so_snd, sndcc, so, rl) == 0) 410 goto bad; 411 if (ssb_reserve(&so->so_rcv, rcvcc, so, rl) == 0) 412 goto bad2; 413 if (so->so_rcv.ssb_lowat == 0) 414 so->so_rcv.ssb_lowat = 1; 415 if (so->so_snd.ssb_lowat == 0) 416 so->so_snd.ssb_lowat = MCLBYTES; 417 if (so->so_snd.ssb_lowat > so->so_snd.ssb_hiwat) 418 so->so_snd.ssb_lowat = so->so_snd.ssb_hiwat; 419 return (0); 420 bad2: 421 ssb_release(&so->so_snd, so); 422 bad: 423 return (ENOBUFS); 424 } 425 426 static int 427 sysctl_handle_sb_max(SYSCTL_HANDLER_ARGS) 428 { 429 int error = 0; 430 u_long old_sb_max = sb_max; 431 432 error = SYSCTL_OUT(req, arg1, sizeof(int)); 433 if (error || !req->newptr) 434 return (error); 435 error = SYSCTL_IN(req, arg1, sizeof(int)); 436 if (error) 437 return (error); 438 if (sb_max < MSIZE + MCLBYTES) { 439 sb_max = old_sb_max; 440 return (EINVAL); 441 } 442 sb_max_adj = (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES); 443 return (0); 444 } 445 446 /* 447 * Allot mbufs to a signalsockbuf. 448 * 449 * Attempt to scale mbmax so that mbcnt doesn't become limiting 450 * if buffering efficiency is near the normal case. 451 * 452 * sb_max only applies to user-sockets (where rl != NULL). It does 453 * not apply to kernel sockets or kernel-controlled sockets. Note 454 * that NFS overrides the sockbuf limits created when nfsd creates 455 * a socket. 456 */ 457 int 458 ssb_reserve(struct signalsockbuf *ssb, u_long cc, struct socket *so, 459 struct rlimit *rl) 460 { 461 /* 462 * rl will only be NULL when we're in an interrupt (eg, in tcp_input) 463 * or when called from netgraph (ie, ngd_attach) 464 */ 465 if (rl && cc > sb_max_adj) 466 cc = sb_max_adj; 467 if (!chgsbsize(so->so_cred->cr_uidinfo, &ssb->ssb_hiwat, cc, 468 rl ? rl->rlim_cur : RLIM_INFINITY)) { 469 return (0); 470 } 471 if (rl) 472 ssb->ssb_mbmax = min(cc * sb_efficiency, sb_max); 473 else 474 ssb->ssb_mbmax = cc * sb_efficiency; 475 476 /* 477 * AUTOLOWAT is set on send buffers and prevents large writes 478 * from generating a huge number of context switches. 479 */ 480 if (ssb->ssb_flags & SSB_AUTOLOWAT) { 481 ssb->ssb_lowat = ssb->ssb_hiwat / 2; 482 if (ssb->ssb_lowat < MCLBYTES) 483 ssb->ssb_lowat = MCLBYTES; 484 } 485 if (ssb->ssb_lowat > ssb->ssb_hiwat) 486 ssb->ssb_lowat = ssb->ssb_hiwat; 487 return (1); 488 } 489 490 /* 491 * Free mbufs held by a socket, and reserved mbuf space. 492 */ 493 void 494 ssb_release(struct signalsockbuf *ssb, struct socket *so) 495 { 496 sbflush(&ssb->sb); 497 (void)chgsbsize(so->so_cred->cr_uidinfo, &ssb->ssb_hiwat, 0, 498 RLIM_INFINITY); 499 ssb->ssb_mbmax = 0; 500 } 501 502 /* 503 * Some routines that return EOPNOTSUPP for entry points that are not 504 * supported by a protocol. Fill in as needed. 505 */ 506 int 507 pru_accept_notsupp(struct socket *so, struct sockaddr **nam) 508 { 509 return EOPNOTSUPP; 510 } 511 512 int 513 pru_bind_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td) 514 { 515 return EOPNOTSUPP; 516 } 517 518 int 519 pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td) 520 { 521 return EOPNOTSUPP; 522 } 523 524 int 525 pru_connect2_notsupp(struct socket *so1, struct socket *so2) 526 { 527 return EOPNOTSUPP; 528 } 529 530 int 531 pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data, 532 struct ifnet *ifp, struct thread *td) 533 { 534 return EOPNOTSUPP; 535 } 536 537 int 538 pru_disconnect_notsupp(struct socket *so) 539 { 540 return EOPNOTSUPP; 541 } 542 543 int 544 pru_listen_notsupp(struct socket *so, struct thread *td) 545 { 546 return EOPNOTSUPP; 547 } 548 549 int 550 pru_peeraddr_notsupp(struct socket *so, struct sockaddr **nam) 551 { 552 return EOPNOTSUPP; 553 } 554 555 int 556 pru_rcvd_notsupp(struct socket *so, int flags) 557 { 558 return EOPNOTSUPP; 559 } 560 561 int 562 pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags) 563 { 564 return EOPNOTSUPP; 565 } 566 567 int 568 pru_shutdown_notsupp(struct socket *so) 569 { 570 return EOPNOTSUPP; 571 } 572 573 int 574 pru_sockaddr_notsupp(struct socket *so, struct sockaddr **nam) 575 { 576 return EOPNOTSUPP; 577 } 578 579 int 580 pru_sosend_notsupp(struct socket *so, struct sockaddr *addr, struct uio *uio, 581 struct mbuf *top, struct mbuf *control, int flags, 582 struct thread *td) 583 { 584 if (top) 585 m_freem(top); 586 if (control) 587 m_freem(control); 588 return (EOPNOTSUPP); 589 } 590 591 int 592 pru_soreceive_notsupp(struct socket *so, struct sockaddr **paddr, 593 struct uio *uio, struct sockbuf *sio, 594 struct mbuf **controlp, int *flagsp) 595 { 596 return (EOPNOTSUPP); 597 } 598 599 int 600 pru_sopoll_notsupp(struct socket *so, int events, 601 struct ucred *cred, struct thread *td) 602 { 603 return (EOPNOTSUPP); 604 } 605 606 int 607 pru_ctloutput_notsupp(struct socket *so, struct sockopt *sopt) 608 { 609 return (EOPNOTSUPP); 610 } 611 612 /* 613 * This isn't really a ``null'' operation, but it's the default one 614 * and doesn't do anything destructive. 615 */ 616 int 617 pru_sense_null(struct socket *so, struct stat *sb) 618 { 619 sb->st_blksize = so->so_snd.ssb_hiwat; 620 return 0; 621 } 622 623 /* 624 * Make a copy of a sockaddr in a malloced buffer of type M_SONAME. Callers 625 * of this routine assume that it always succeeds, so we have to use a 626 * blockable allocation even though we might be called from a critical thread. 627 */ 628 struct sockaddr * 629 dup_sockaddr(const struct sockaddr *sa) 630 { 631 struct sockaddr *sa2; 632 633 sa2 = kmalloc(sa->sa_len, M_SONAME, M_INTWAIT); 634 bcopy(sa, sa2, sa->sa_len); 635 return (sa2); 636 } 637 638 /* 639 * Create an external-format (``xsocket'') structure using the information 640 * in the kernel-format socket structure pointed to by so. This is done 641 * to reduce the spew of irrelevant information over this interface, 642 * to isolate user code from changes in the kernel structure, and 643 * potentially to provide information-hiding if we decide that 644 * some of this information should be hidden from users. 645 */ 646 void 647 sotoxsocket(struct socket *so, struct xsocket *xso) 648 { 649 xso->xso_len = sizeof *xso; 650 xso->xso_so = so; 651 xso->so_type = so->so_type; 652 xso->so_options = so->so_options; 653 xso->so_linger = so->so_linger; 654 xso->so_state = so->so_state; 655 xso->so_pcb = so->so_pcb; 656 xso->xso_protocol = so->so_proto->pr_protocol; 657 xso->xso_family = so->so_proto->pr_domain->dom_family; 658 xso->so_qlen = so->so_qlen; 659 xso->so_incqlen = so->so_incqlen; 660 xso->so_qlimit = so->so_qlimit; 661 xso->so_timeo = so->so_timeo; 662 xso->so_error = so->so_error; 663 xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0; 664 xso->so_oobmark = so->so_oobmark; 665 ssbtoxsockbuf(&so->so_snd, &xso->so_snd); 666 ssbtoxsockbuf(&so->so_rcv, &xso->so_rcv); 667 xso->so_uid = so->so_cred->cr_uid; 668 } 669 670 /* 671 * Here is the definition of some of the basic objects in the kern.ipc 672 * branch of the MIB. 673 */ 674 SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC"); 675 676 /* 677 * This takes the place of kern.maxsockbuf, which moved to kern.ipc. 678 * 679 * NOTE! sb_max only applies to user-created socket buffers. 680 */ 681 static int dummy; 682 SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, ""); 683 SYSCTL_OID(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLTYPE_INT|CTLFLAG_RW, 684 &sb_max, 0, sysctl_handle_sb_max, "I", "Maximum socket buffer size"); 685 SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD, 686 &maxsockets, 0, "Maximum number of sockets available"); 687 SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW, 688 &sb_efficiency, 0, ""); 689 690 /* 691 * Initialize maxsockets 692 */ 693 static void 694 init_maxsockets(void *ignored) 695 { 696 TUNABLE_INT_FETCH("kern.ipc.maxsockets", &maxsockets); 697 maxsockets = imax(maxsockets, imax(maxfiles, nmbclusters)); 698 } 699 SYSINIT(param, SI_BOOT1_TUNABLES, SI_ORDER_ANY, 700 init_maxsockets, NULL); 701 702