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 * When an attempt at a new connection is noted on a socket 222 * which accepts connections, sonewconn is called. If the 223 * connection is possible (subject to space constraints, etc.) 224 * then we allocate a new structure, propoerly linked into the 225 * data structure of the original socket, and return this. 226 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED. 227 */ 228 struct socket * 229 sonewconn(struct socket *head, int connstatus) 230 { 231 struct socket *so; 232 struct socket *sp; 233 struct pru_attach_info ai; 234 235 if (head->so_qlen > 3 * head->so_qlimit / 2) 236 return (NULL); 237 so = soalloc(1); 238 if (so == NULL) 239 return (NULL); 240 if ((head->so_options & SO_ACCEPTFILTER) != 0) 241 connstatus = 0; 242 so->so_head = head; 243 so->so_type = head->so_type; 244 so->so_options = head->so_options &~ SO_ACCEPTCONN; 245 so->so_linger = head->so_linger; 246 so->so_state = head->so_state | SS_NOFDREF; 247 so->so_proto = head->so_proto; 248 so->so_cred = crhold(head->so_cred); 249 ai.sb_rlimit = NULL; 250 ai.p_ucred = NULL; 251 ai.fd_rdir = NULL; /* jail code cruft XXX JH */ 252 if (soreserve(so, head->so_snd.ssb_hiwat, head->so_rcv.ssb_hiwat, NULL) || 253 /* Directly call function since we're already at protocol level. */ 254 (*so->so_proto->pr_usrreqs->pru_attach)(so, 0, &ai)) { 255 sodealloc(so); 256 return (NULL); 257 } 258 so->so_rcv.ssb_lowat = head->so_rcv.ssb_lowat; 259 so->so_snd.ssb_lowat = head->so_snd.ssb_lowat; 260 so->so_rcv.ssb_timeo = head->so_rcv.ssb_timeo; 261 so->so_snd.ssb_timeo = head->so_snd.ssb_timeo; 262 so->so_rcv.ssb_flags |= head->so_rcv.ssb_flags & 263 (SSB_AUTOSIZE | SSB_AUTOLOWAT); 264 so->so_snd.ssb_flags |= head->so_snd.ssb_flags & 265 (SSB_AUTOSIZE | SSB_AUTOLOWAT); 266 if (connstatus) { 267 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list); 268 so->so_state |= SS_COMP; 269 head->so_qlen++; 270 } else { 271 if (head->so_incqlen > head->so_qlimit) { 272 sp = TAILQ_FIRST(&head->so_incomp); 273 TAILQ_REMOVE(&head->so_incomp, sp, so_list); 274 head->so_incqlen--; 275 sp->so_state &= ~SS_INCOMP; 276 sp->so_head = NULL; 277 soaborta(sp); 278 } 279 TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list); 280 so->so_state |= SS_INCOMP; 281 head->so_incqlen++; 282 } 283 if (connstatus) { 284 sorwakeup(head); 285 wakeup((caddr_t)&head->so_timeo); 286 so->so_state |= connstatus; 287 } 288 return (so); 289 } 290 291 /* 292 * Socantsendmore indicates that no more data will be sent on the 293 * socket; it would normally be applied to a socket when the user 294 * informs the system that no more data is to be sent, by the protocol 295 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data 296 * will be received, and will normally be applied to the socket by a 297 * protocol when it detects that the peer will send no more data. 298 * Data queued for reading in the socket may yet be read. 299 */ 300 void 301 socantsendmore(struct socket *so) 302 { 303 so->so_state |= SS_CANTSENDMORE; 304 sowwakeup(so); 305 } 306 307 void 308 socantrcvmore(struct socket *so) 309 { 310 so->so_state |= SS_CANTRCVMORE; 311 sorwakeup(so); 312 } 313 314 /* 315 * Wakeup processes waiting on a socket buffer. Do asynchronous notification 316 * via SIGIO if the socket has the SS_ASYNC flag set. 317 * 318 * For users waiting on send/recv try to avoid unnecessary context switch 319 * thrashing. Particularly for senders of large buffers (needs to be 320 * extended to sel and aio? XXX) 321 */ 322 void 323 sowakeup(struct socket *so, struct signalsockbuf *ssb) 324 { 325 struct selinfo *selinfo = &ssb->ssb_sel; 326 327 selwakeup(selinfo); 328 ssb->ssb_flags &= ~SSB_SEL; 329 if (ssb->ssb_flags & SSB_WAIT) { 330 if ((ssb == &so->so_snd && ssb_space(ssb) >= ssb->ssb_lowat) || 331 (ssb == &so->so_rcv && ssb->ssb_cc >= ssb->ssb_lowat) || 332 (ssb == &so->so_snd && (so->so_state & SS_CANTSENDMORE)) || 333 (ssb == &so->so_rcv && (so->so_state & SS_CANTRCVMORE)) 334 ) { 335 ssb->ssb_flags &= ~SSB_WAIT; 336 wakeup((caddr_t)&ssb->ssb_cc); 337 } 338 } 339 if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL) 340 pgsigio(so->so_sigio, SIGIO, 0); 341 if (ssb->ssb_flags & SSB_UPCALL) 342 (*so->so_upcall)(so, so->so_upcallarg, MB_DONTWAIT); 343 if (ssb->ssb_flags & SSB_AIO) 344 aio_swake(so, ssb); 345 KNOTE(&selinfo->si_note, 0); 346 if (ssb->ssb_flags & SSB_MEVENT) { 347 struct netmsg_so_notify *msg, *nmsg; 348 349 TAILQ_FOREACH_MUTABLE(msg, &selinfo->si_mlist, nm_list, nmsg) { 350 if (msg->nm_predicate(&msg->nm_netmsg)) { 351 TAILQ_REMOVE(&selinfo->si_mlist, msg, nm_list); 352 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 353 msg->nm_netmsg.nm_lmsg.ms_error); 354 } 355 } 356 if (TAILQ_EMPTY(&ssb->ssb_sel.si_mlist)) 357 ssb->ssb_flags &= ~SSB_MEVENT; 358 } 359 } 360 361 /* 362 * Socket buffer (struct signalsockbuf) utility routines. 363 * 364 * Each socket contains two socket buffers: one for sending data and 365 * one for receiving data. Each buffer contains a queue of mbufs, 366 * information about the number of mbufs and amount of data in the 367 * queue, and other fields allowing select() statements and notification 368 * on data availability to be implemented. 369 * 370 * Data stored in a socket buffer is maintained as a list of records. 371 * Each record is a list of mbufs chained together with the m_next 372 * field. Records are chained together with the m_nextpkt field. The upper 373 * level routine soreceive() expects the following conventions to be 374 * observed when placing information in the receive buffer: 375 * 376 * 1. If the protocol requires each message be preceded by the sender's 377 * name, then a record containing that name must be present before 378 * any associated data (mbuf's must be of type MT_SONAME). 379 * 2. If the protocol supports the exchange of ``access rights'' (really 380 * just additional data associated with the message), and there are 381 * ``rights'' to be received, then a record containing this data 382 * should be present (mbuf's must be of type MT_RIGHTS). 383 * 3. If a name or rights record exists, then it must be followed by 384 * a data record, perhaps of zero length. 385 * 386 * Before using a new socket structure it is first necessary to reserve 387 * buffer space to the socket, by calling sbreserve(). This should commit 388 * some of the available buffer space in the system buffer pool for the 389 * socket (currently, it does nothing but enforce limits). The space 390 * should be released by calling ssb_release() when the socket is destroyed. 391 */ 392 int 393 soreserve(struct socket *so, u_long sndcc, u_long rcvcc, struct rlimit *rl) 394 { 395 if (so->so_snd.ssb_lowat == 0) 396 so->so_snd.ssb_flags |= SSB_AUTOLOWAT; 397 if (ssb_reserve(&so->so_snd, sndcc, so, rl) == 0) 398 goto bad; 399 if (ssb_reserve(&so->so_rcv, rcvcc, so, rl) == 0) 400 goto bad2; 401 if (so->so_rcv.ssb_lowat == 0) 402 so->so_rcv.ssb_lowat = 1; 403 if (so->so_snd.ssb_lowat == 0) 404 so->so_snd.ssb_lowat = MCLBYTES; 405 if (so->so_snd.ssb_lowat > so->so_snd.ssb_hiwat) 406 so->so_snd.ssb_lowat = so->so_snd.ssb_hiwat; 407 return (0); 408 bad2: 409 ssb_release(&so->so_snd, so); 410 bad: 411 return (ENOBUFS); 412 } 413 414 static int 415 sysctl_handle_sb_max(SYSCTL_HANDLER_ARGS) 416 { 417 int error = 0; 418 u_long old_sb_max = sb_max; 419 420 error = SYSCTL_OUT(req, arg1, sizeof(int)); 421 if (error || !req->newptr) 422 return (error); 423 error = SYSCTL_IN(req, arg1, sizeof(int)); 424 if (error) 425 return (error); 426 if (sb_max < MSIZE + MCLBYTES) { 427 sb_max = old_sb_max; 428 return (EINVAL); 429 } 430 sb_max_adj = (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES); 431 return (0); 432 } 433 434 /* 435 * Allot mbufs to a signalsockbuf. 436 * 437 * Attempt to scale mbmax so that mbcnt doesn't become limiting 438 * if buffering efficiency is near the normal case. 439 * 440 * sb_max only applies to user-sockets (where rl != NULL). It does 441 * not apply to kernel sockets or kernel-controlled sockets. Note 442 * that NFS overrides the sockbuf limits created when nfsd creates 443 * a socket. 444 */ 445 int 446 ssb_reserve(struct signalsockbuf *ssb, u_long cc, struct socket *so, 447 struct rlimit *rl) 448 { 449 /* 450 * rl will only be NULL when we're in an interrupt (eg, in tcp_input) 451 * or when called from netgraph (ie, ngd_attach) 452 */ 453 if (rl && cc > sb_max_adj) 454 cc = sb_max_adj; 455 if (!chgsbsize(so->so_cred->cr_uidinfo, &ssb->ssb_hiwat, cc, 456 rl ? rl->rlim_cur : RLIM_INFINITY)) { 457 return (0); 458 } 459 if (rl) 460 ssb->ssb_mbmax = min(cc * sb_efficiency, sb_max); 461 else 462 ssb->ssb_mbmax = cc * sb_efficiency; 463 464 /* 465 * AUTOLOWAT is set on send buffers and prevents large writes 466 * from generating a huge number of context switches. 467 */ 468 if (ssb->ssb_flags & SSB_AUTOLOWAT) { 469 ssb->ssb_lowat = ssb->ssb_hiwat / 2; 470 if (ssb->ssb_lowat < MCLBYTES) 471 ssb->ssb_lowat = MCLBYTES; 472 } 473 if (ssb->ssb_lowat > ssb->ssb_hiwat) 474 ssb->ssb_lowat = ssb->ssb_hiwat; 475 return (1); 476 } 477 478 /* 479 * Free mbufs held by a socket, and reserved mbuf space. 480 */ 481 void 482 ssb_release(struct signalsockbuf *ssb, struct socket *so) 483 { 484 sbflush(&ssb->sb); 485 (void)chgsbsize(so->so_cred->cr_uidinfo, &ssb->ssb_hiwat, 0, 486 RLIM_INFINITY); 487 ssb->ssb_mbmax = 0; 488 } 489 490 /* 491 * Some routines that return EOPNOTSUPP for entry points that are not 492 * supported by a protocol. Fill in as needed. 493 */ 494 int 495 pru_accept_notsupp(struct socket *so, struct sockaddr **nam) 496 { 497 return EOPNOTSUPP; 498 } 499 500 int 501 pru_bind_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td) 502 { 503 return EOPNOTSUPP; 504 } 505 506 int 507 pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td) 508 { 509 return EOPNOTSUPP; 510 } 511 512 int 513 pru_connect2_notsupp(struct socket *so1, struct socket *so2) 514 { 515 return EOPNOTSUPP; 516 } 517 518 int 519 pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data, 520 struct ifnet *ifp, struct thread *td) 521 { 522 return EOPNOTSUPP; 523 } 524 525 int 526 pru_disconnect_notsupp(struct socket *so) 527 { 528 return EOPNOTSUPP; 529 } 530 531 int 532 pru_listen_notsupp(struct socket *so, struct thread *td) 533 { 534 return EOPNOTSUPP; 535 } 536 537 int 538 pru_peeraddr_notsupp(struct socket *so, struct sockaddr **nam) 539 { 540 return EOPNOTSUPP; 541 } 542 543 int 544 pru_rcvd_notsupp(struct socket *so, int flags) 545 { 546 return EOPNOTSUPP; 547 } 548 549 int 550 pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags) 551 { 552 return EOPNOTSUPP; 553 } 554 555 int 556 pru_shutdown_notsupp(struct socket *so) 557 { 558 return EOPNOTSUPP; 559 } 560 561 int 562 pru_sockaddr_notsupp(struct socket *so, struct sockaddr **nam) 563 { 564 return EOPNOTSUPP; 565 } 566 567 int 568 pru_sosend_notsupp(struct socket *so, struct sockaddr *addr, struct uio *uio, 569 struct mbuf *top, struct mbuf *control, int flags, 570 struct thread *td) 571 { 572 if (top) 573 m_freem(top); 574 if (control) 575 m_freem(control); 576 return (EOPNOTSUPP); 577 } 578 579 int 580 pru_soreceive_notsupp(struct socket *so, struct sockaddr **paddr, 581 struct uio *uio, struct sockbuf *sio, 582 struct mbuf **controlp, int *flagsp) 583 { 584 return (EOPNOTSUPP); 585 } 586 587 int 588 pru_sopoll_notsupp(struct socket *so, int events, 589 struct ucred *cred, struct thread *td) 590 { 591 return (EOPNOTSUPP); 592 } 593 594 int 595 pru_ctloutput_notsupp(struct socket *so, struct sockopt *sopt) 596 { 597 return (EOPNOTSUPP); 598 } 599 600 /* 601 * This isn't really a ``null'' operation, but it's the default one 602 * and doesn't do anything destructive. 603 */ 604 int 605 pru_sense_null(struct socket *so, struct stat *sb) 606 { 607 sb->st_blksize = so->so_snd.ssb_hiwat; 608 return 0; 609 } 610 611 /* 612 * Make a copy of a sockaddr in a malloced buffer of type M_SONAME. Callers 613 * of this routine assume that it always succeeds, so we have to use a 614 * blockable allocation even though we might be called from a critical thread. 615 */ 616 struct sockaddr * 617 dup_sockaddr(const struct sockaddr *sa) 618 { 619 struct sockaddr *sa2; 620 621 sa2 = kmalloc(sa->sa_len, M_SONAME, M_INTWAIT); 622 bcopy(sa, sa2, sa->sa_len); 623 return (sa2); 624 } 625 626 /* 627 * Create an external-format (``xsocket'') structure using the information 628 * in the kernel-format socket structure pointed to by so. This is done 629 * to reduce the spew of irrelevant information over this interface, 630 * to isolate user code from changes in the kernel structure, and 631 * potentially to provide information-hiding if we decide that 632 * some of this information should be hidden from users. 633 */ 634 void 635 sotoxsocket(struct socket *so, struct xsocket *xso) 636 { 637 xso->xso_len = sizeof *xso; 638 xso->xso_so = so; 639 xso->so_type = so->so_type; 640 xso->so_options = so->so_options; 641 xso->so_linger = so->so_linger; 642 xso->so_state = so->so_state; 643 xso->so_pcb = so->so_pcb; 644 xso->xso_protocol = so->so_proto->pr_protocol; 645 xso->xso_family = so->so_proto->pr_domain->dom_family; 646 xso->so_qlen = so->so_qlen; 647 xso->so_incqlen = so->so_incqlen; 648 xso->so_qlimit = so->so_qlimit; 649 xso->so_timeo = so->so_timeo; 650 xso->so_error = so->so_error; 651 xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0; 652 xso->so_oobmark = so->so_oobmark; 653 ssbtoxsockbuf(&so->so_snd, &xso->so_snd); 654 ssbtoxsockbuf(&so->so_rcv, &xso->so_rcv); 655 xso->so_uid = so->so_cred->cr_uid; 656 } 657 658 /* 659 * Here is the definition of some of the basic objects in the kern.ipc 660 * branch of the MIB. 661 */ 662 SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC"); 663 664 /* 665 * This takes the place of kern.maxsockbuf, which moved to kern.ipc. 666 * 667 * NOTE! sb_max only applies to user-created socket buffers. 668 */ 669 static int dummy; 670 SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, ""); 671 SYSCTL_OID(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLTYPE_INT|CTLFLAG_RW, 672 &sb_max, 0, sysctl_handle_sb_max, "I", "Maximum socket buffer size"); 673 SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD, 674 &maxsockets, 0, "Maximum number of sockets available"); 675 SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW, 676 &sb_efficiency, 0, ""); 677 678 /* 679 * Initialize maxsockets 680 */ 681 static void 682 init_maxsockets(void *ignored) 683 { 684 TUNABLE_INT_FETCH("kern.ipc.maxsockets", &maxsockets); 685 maxsockets = imax(maxsockets, imax(maxfiles, nmbclusters)); 686 } 687 SYSINIT(param, SI_BOOT1_TUNABLES, SI_ORDER_ANY, 688 init_maxsockets, NULL); 689 690