uipc_sockbuf.c revision 40913
1/* 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)uipc_socket2.c 8.1 (Berkeley) 6/10/93 34 * $Id: uipc_socket2.c,v 1.39 1998/09/05 13:24:39 bde Exp $ 35 */ 36 37#include <sys/param.h> 38#include <sys/systm.h> 39#include <sys/domain.h> 40#include <sys/kernel.h> 41#include <sys/proc.h> 42#include <sys/malloc.h> 43#include <sys/mbuf.h> 44#include <sys/protosw.h> 45#include <sys/stat.h> 46#include <sys/socket.h> 47#include <sys/socketvar.h> 48#include <sys/signalvar.h> 49#include <sys/sysctl.h> 50 51/* 52 * Primitive routines for operating on sockets and socket buffers 53 */ 54 55u_long sb_max = SB_MAX; /* XXX should be static */ 56 57static u_long sb_efficiency = 8; /* parameter for sbreserve() */ 58 59/* 60 * Procedures to manipulate state flags of socket 61 * and do appropriate wakeups. Normal sequence from the 62 * active (originating) side is that soisconnecting() is 63 * called during processing of connect() call, 64 * resulting in an eventual call to soisconnected() if/when the 65 * connection is established. When the connection is torn down 66 * soisdisconnecting() is called during processing of disconnect() call, 67 * and soisdisconnected() is called when the connection to the peer 68 * is totally severed. The semantics of these routines are such that 69 * connectionless protocols can call soisconnected() and soisdisconnected() 70 * only, bypassing the in-progress calls when setting up a ``connection'' 71 * takes no time. 72 * 73 * From the passive side, a socket is created with 74 * two queues of sockets: so_q0 for connections in progress 75 * and so_q for connections already made and awaiting user acceptance. 76 * As a protocol is preparing incoming connections, it creates a socket 77 * structure queued on so_q0 by calling sonewconn(). When the connection 78 * is established, soisconnected() is called, and transfers the 79 * socket structure to so_q, making it available to accept(). 80 * 81 * If a socket is closed with sockets on either 82 * so_q0 or so_q, these sockets are dropped. 83 * 84 * If higher level protocols are implemented in 85 * the kernel, the wakeups done here will sometimes 86 * cause software-interrupt process scheduling. 87 */ 88 89void 90soisconnecting(so) 91 register struct socket *so; 92{ 93 94 so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING); 95 so->so_state |= SS_ISCONNECTING; 96} 97 98void 99soisconnected(so) 100 register struct socket *so; 101{ 102 register struct socket *head = so->so_head; 103 104 so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING); 105 so->so_state |= SS_ISCONNECTED; 106 if (head && (so->so_state & SS_INCOMP)) { 107 TAILQ_REMOVE(&head->so_incomp, so, so_list); 108 head->so_incqlen--; 109 so->so_state &= ~SS_INCOMP; 110 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list); 111 so->so_state |= SS_COMP; 112 sorwakeup(head); 113 wakeup_one(&head->so_timeo); 114 } else { 115 wakeup(&so->so_timeo); 116 sorwakeup(so); 117 sowwakeup(so); 118 } 119} 120 121void 122soisdisconnecting(so) 123 register struct socket *so; 124{ 125 126 so->so_state &= ~SS_ISCONNECTING; 127 so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE); 128 wakeup((caddr_t)&so->so_timeo); 129 sowwakeup(so); 130 sorwakeup(so); 131} 132 133void 134soisdisconnected(so) 135 register struct socket *so; 136{ 137 138 so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING); 139 so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE); 140 wakeup((caddr_t)&so->so_timeo); 141 sowwakeup(so); 142 sorwakeup(so); 143} 144 145/* 146 * Return a random connection that hasn't been serviced yet and 147 * is eligible for discard. There is a one in qlen chance that 148 * we will return a null, saying that there are no dropable 149 * requests. In this case, the protocol specific code should drop 150 * the new request. This insures fairness. 151 * 152 * This may be used in conjunction with protocol specific queue 153 * congestion routines. 154 */ 155struct socket * 156sodropablereq(head) 157 register struct socket *head; 158{ 159 register struct socket *so; 160 unsigned int i, j, qlen; 161 static int rnd; 162 static struct timeval old_runtime; 163 static unsigned int cur_cnt, old_cnt; 164 struct timeval tv; 165 166 getmicrouptime(&tv); 167 if ((i = (tv.tv_sec - old_runtime.tv_sec)) != 0) { 168 old_runtime = tv; 169 old_cnt = cur_cnt / i; 170 cur_cnt = 0; 171 } 172 173 so = TAILQ_FIRST(&head->so_incomp); 174 if (!so) 175 return (so); 176 177 qlen = head->so_incqlen; 178 if (++cur_cnt > qlen || old_cnt > qlen) { 179 rnd = (314159 * rnd + 66329) & 0xffff; 180 j = ((qlen + 1) * rnd) >> 16; 181 182 while (j-- && so) 183 so = TAILQ_NEXT(so, so_list); 184 } 185 186 return (so); 187} 188 189/* 190 * When an attempt at a new connection is noted on a socket 191 * which accepts connections, sonewconn is called. If the 192 * connection is possible (subject to space constraints, etc.) 193 * then we allocate a new structure, propoerly linked into the 194 * data structure of the original socket, and return this. 195 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED. 196 */ 197struct socket * 198sonewconn(head, connstatus) 199 register struct socket *head; 200 int connstatus; 201{ 202 register struct socket *so; 203 204 if (head->so_qlen > 3 * head->so_qlimit / 2) 205 return ((struct socket *)0); 206 so = soalloc(0); 207 if (so == NULL) 208 return ((struct socket *)0); 209 so->so_head = head; 210 so->so_type = head->so_type; 211 so->so_options = head->so_options &~ SO_ACCEPTCONN; 212 so->so_linger = head->so_linger; 213 so->so_state = head->so_state | SS_NOFDREF; 214 so->so_proto = head->so_proto; 215 so->so_timeo = head->so_timeo; 216 so->so_pgid = head->so_pgid; 217 so->so_uid = head->so_uid; 218 (void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat); 219 220 if ((*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) { 221 sodealloc(so); 222 return ((struct socket *)0); 223 } 224 225 if (connstatus) { 226 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list); 227 so->so_state |= SS_COMP; 228 } else { 229 TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list); 230 so->so_state |= SS_INCOMP; 231 head->so_incqlen++; 232 } 233 head->so_qlen++; 234 if (connstatus) { 235 sorwakeup(head); 236 wakeup((caddr_t)&head->so_timeo); 237 so->so_state |= connstatus; 238 } 239 return (so); 240} 241 242/* 243 * Socantsendmore indicates that no more data will be sent on the 244 * socket; it would normally be applied to a socket when the user 245 * informs the system that no more data is to be sent, by the protocol 246 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data 247 * will be received, and will normally be applied to the socket by a 248 * protocol when it detects that the peer will send no more data. 249 * Data queued for reading in the socket may yet be read. 250 */ 251 252void 253socantsendmore(so) 254 struct socket *so; 255{ 256 257 so->so_state |= SS_CANTSENDMORE; 258 sowwakeup(so); 259} 260 261void 262socantrcvmore(so) 263 struct socket *so; 264{ 265 266 so->so_state |= SS_CANTRCVMORE; 267 sorwakeup(so); 268} 269 270/* 271 * Wait for data to arrive at/drain from a socket buffer. 272 */ 273int 274sbwait(sb) 275 struct sockbuf *sb; 276{ 277 278 sb->sb_flags |= SB_WAIT; 279 return (tsleep((caddr_t)&sb->sb_cc, 280 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait", 281 sb->sb_timeo)); 282} 283 284/* 285 * Lock a sockbuf already known to be locked; 286 * return any error returned from sleep (EINTR). 287 */ 288int 289sb_lock(sb) 290 register struct sockbuf *sb; 291{ 292 int error; 293 294 while (sb->sb_flags & SB_LOCK) { 295 sb->sb_flags |= SB_WANT; 296 error = tsleep((caddr_t)&sb->sb_flags, 297 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH, 298 "sblock", 0); 299 if (error) 300 return (error); 301 } 302 sb->sb_flags |= SB_LOCK; 303 return (0); 304} 305 306/* 307 * Wakeup processes waiting on a socket buffer. 308 * Do asynchronous notification via SIGIO 309 * if the socket has the SS_ASYNC flag set. 310 */ 311void 312sowakeup(so, sb) 313 register struct socket *so; 314 register struct sockbuf *sb; 315{ 316 struct proc *p; 317 318 selwakeup(&sb->sb_sel); 319 sb->sb_flags &= ~SB_SEL; 320 if (sb->sb_flags & SB_WAIT) { 321 sb->sb_flags &= ~SB_WAIT; 322 wakeup((caddr_t)&sb->sb_cc); 323 } 324 if (so->so_state & SS_ASYNC) { 325 if (so->so_pgid < 0) 326 gsignal(-so->so_pgid, SIGIO); 327 else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0) 328 psignal(p, SIGIO); 329 } 330 if (sb->sb_flags & SB_UPCALL) 331 (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT); 332} 333 334/* 335 * Socket buffer (struct sockbuf) utility routines. 336 * 337 * Each socket contains two socket buffers: one for sending data and 338 * one for receiving data. Each buffer contains a queue of mbufs, 339 * information about the number of mbufs and amount of data in the 340 * queue, and other fields allowing select() statements and notification 341 * on data availability to be implemented. 342 * 343 * Data stored in a socket buffer is maintained as a list of records. 344 * Each record is a list of mbufs chained together with the m_next 345 * field. Records are chained together with the m_nextpkt field. The upper 346 * level routine soreceive() expects the following conventions to be 347 * observed when placing information in the receive buffer: 348 * 349 * 1. If the protocol requires each message be preceded by the sender's 350 * name, then a record containing that name must be present before 351 * any associated data (mbuf's must be of type MT_SONAME). 352 * 2. If the protocol supports the exchange of ``access rights'' (really 353 * just additional data associated with the message), and there are 354 * ``rights'' to be received, then a record containing this data 355 * should be present (mbuf's must be of type MT_RIGHTS). 356 * 3. If a name or rights record exists, then it must be followed by 357 * a data record, perhaps of zero length. 358 * 359 * Before using a new socket structure it is first necessary to reserve 360 * buffer space to the socket, by calling sbreserve(). This should commit 361 * some of the available buffer space in the system buffer pool for the 362 * socket (currently, it does nothing but enforce limits). The space 363 * should be released by calling sbrelease() when the socket is destroyed. 364 */ 365 366int 367soreserve(so, sndcc, rcvcc) 368 register struct socket *so; 369 u_long sndcc, rcvcc; 370{ 371 372 if (sbreserve(&so->so_snd, sndcc) == 0) 373 goto bad; 374 if (sbreserve(&so->so_rcv, rcvcc) == 0) 375 goto bad2; 376 if (so->so_rcv.sb_lowat == 0) 377 so->so_rcv.sb_lowat = 1; 378 if (so->so_snd.sb_lowat == 0) 379 so->so_snd.sb_lowat = MCLBYTES; 380 if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat) 381 so->so_snd.sb_lowat = so->so_snd.sb_hiwat; 382 return (0); 383bad2: 384 sbrelease(&so->so_snd); 385bad: 386 return (ENOBUFS); 387} 388 389/* 390 * Allot mbufs to a sockbuf. 391 * Attempt to scale mbmax so that mbcnt doesn't become limiting 392 * if buffering efficiency is near the normal case. 393 */ 394int 395sbreserve(sb, cc) 396 struct sockbuf *sb; 397 u_long cc; 398{ 399 if ((u_quad_t)cc > (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES)) 400 return (0); 401 sb->sb_hiwat = cc; 402 sb->sb_mbmax = min(cc * sb_efficiency, sb_max); 403 if (sb->sb_lowat > sb->sb_hiwat) 404 sb->sb_lowat = sb->sb_hiwat; 405 return (1); 406} 407 408/* 409 * Free mbufs held by a socket, and reserved mbuf space. 410 */ 411void 412sbrelease(sb) 413 struct sockbuf *sb; 414{ 415 416 sbflush(sb); 417 sb->sb_hiwat = sb->sb_mbmax = 0; 418} 419 420/* 421 * Routines to add and remove 422 * data from an mbuf queue. 423 * 424 * The routines sbappend() or sbappendrecord() are normally called to 425 * append new mbufs to a socket buffer, after checking that adequate 426 * space is available, comparing the function sbspace() with the amount 427 * of data to be added. sbappendrecord() differs from sbappend() in 428 * that data supplied is treated as the beginning of a new record. 429 * To place a sender's address, optional access rights, and data in a 430 * socket receive buffer, sbappendaddr() should be used. To place 431 * access rights and data in a socket receive buffer, sbappendrights() 432 * should be used. In either case, the new data begins a new record. 433 * Note that unlike sbappend() and sbappendrecord(), these routines check 434 * for the caller that there will be enough space to store the data. 435 * Each fails if there is not enough space, or if it cannot find mbufs 436 * to store additional information in. 437 * 438 * Reliable protocols may use the socket send buffer to hold data 439 * awaiting acknowledgement. Data is normally copied from a socket 440 * send buffer in a protocol with m_copy for output to a peer, 441 * and then removing the data from the socket buffer with sbdrop() 442 * or sbdroprecord() when the data is acknowledged by the peer. 443 */ 444 445/* 446 * Append mbuf chain m to the last record in the 447 * socket buffer sb. The additional space associated 448 * the mbuf chain is recorded in sb. Empty mbufs are 449 * discarded and mbufs are compacted where possible. 450 */ 451void 452sbappend(sb, m) 453 struct sockbuf *sb; 454 struct mbuf *m; 455{ 456 register struct mbuf *n; 457 458 if (m == 0) 459 return; 460 n = sb->sb_mb; 461 if (n) { 462 while (n->m_nextpkt) 463 n = n->m_nextpkt; 464 do { 465 if (n->m_flags & M_EOR) { 466 sbappendrecord(sb, m); /* XXXXXX!!!! */ 467 return; 468 } 469 } while (n->m_next && (n = n->m_next)); 470 } 471 sbcompress(sb, m, n); 472} 473 474#ifdef SOCKBUF_DEBUG 475void 476sbcheck(sb) 477 register struct sockbuf *sb; 478{ 479 register struct mbuf *m; 480 register struct mbuf *n = 0; 481 register u_long len = 0, mbcnt = 0; 482 483 for (m = sb->sb_mb; m; m = n) { 484 n = m->m_nextpkt; 485 for (; m; m = m->m_next) { 486 len += m->m_len; 487 mbcnt += MSIZE; 488 if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */ 489 mbcnt += m->m_ext.ext_size; 490 } 491 } 492 if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) { 493 printf("cc %ld != %ld || mbcnt %ld != %ld\n", len, sb->sb_cc, 494 mbcnt, sb->sb_mbcnt); 495 panic("sbcheck"); 496 } 497} 498#endif 499 500/* 501 * As above, except the mbuf chain 502 * begins a new record. 503 */ 504void 505sbappendrecord(sb, m0) 506 register struct sockbuf *sb; 507 register struct mbuf *m0; 508{ 509 register struct mbuf *m; 510 511 if (m0 == 0) 512 return; 513 m = sb->sb_mb; 514 if (m) 515 while (m->m_nextpkt) 516 m = m->m_nextpkt; 517 /* 518 * Put the first mbuf on the queue. 519 * Note this permits zero length records. 520 */ 521 sballoc(sb, m0); 522 if (m) 523 m->m_nextpkt = m0; 524 else 525 sb->sb_mb = m0; 526 m = m0->m_next; 527 m0->m_next = 0; 528 if (m && (m0->m_flags & M_EOR)) { 529 m0->m_flags &= ~M_EOR; 530 m->m_flags |= M_EOR; 531 } 532 sbcompress(sb, m, m0); 533} 534 535/* 536 * As above except that OOB data 537 * is inserted at the beginning of the sockbuf, 538 * but after any other OOB data. 539 */ 540void 541sbinsertoob(sb, m0) 542 register struct sockbuf *sb; 543 register struct mbuf *m0; 544{ 545 register struct mbuf *m; 546 register struct mbuf **mp; 547 548 if (m0 == 0) 549 return; 550 for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) { 551 m = *mp; 552 again: 553 switch (m->m_type) { 554 555 case MT_OOBDATA: 556 continue; /* WANT next train */ 557 558 case MT_CONTROL: 559 m = m->m_next; 560 if (m) 561 goto again; /* inspect THIS train further */ 562 } 563 break; 564 } 565 /* 566 * Put the first mbuf on the queue. 567 * Note this permits zero length records. 568 */ 569 sballoc(sb, m0); 570 m0->m_nextpkt = *mp; 571 *mp = m0; 572 m = m0->m_next; 573 m0->m_next = 0; 574 if (m && (m0->m_flags & M_EOR)) { 575 m0->m_flags &= ~M_EOR; 576 m->m_flags |= M_EOR; 577 } 578 sbcompress(sb, m, m0); 579} 580 581/* 582 * Append address and data, and optionally, control (ancillary) data 583 * to the receive queue of a socket. If present, 584 * m0 must include a packet header with total length. 585 * Returns 0 if no space in sockbuf or insufficient mbufs. 586 */ 587int 588sbappendaddr(sb, asa, m0, control) 589 register struct sockbuf *sb; 590 struct sockaddr *asa; 591 struct mbuf *m0, *control; 592{ 593 register struct mbuf *m, *n; 594 int space = asa->sa_len; 595 596if (m0 && (m0->m_flags & M_PKTHDR) == 0) 597panic("sbappendaddr"); 598 if (m0) 599 space += m0->m_pkthdr.len; 600 for (n = control; n; n = n->m_next) { 601 space += n->m_len; 602 if (n->m_next == 0) /* keep pointer to last control buf */ 603 break; 604 } 605 if (space > sbspace(sb)) 606 return (0); 607 if (asa->sa_len > MLEN) 608 return (0); 609 MGET(m, M_DONTWAIT, MT_SONAME); 610 if (m == 0) 611 return (0); 612 m->m_len = asa->sa_len; 613 bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len); 614 if (n) 615 n->m_next = m0; /* concatenate data to control */ 616 else 617 control = m0; 618 m->m_next = control; 619 for (n = m; n; n = n->m_next) 620 sballoc(sb, n); 621 n = sb->sb_mb; 622 if (n) { 623 while (n->m_nextpkt) 624 n = n->m_nextpkt; 625 n->m_nextpkt = m; 626 } else 627 sb->sb_mb = m; 628 return (1); 629} 630 631int 632sbappendcontrol(sb, m0, control) 633 struct sockbuf *sb; 634 struct mbuf *control, *m0; 635{ 636 register struct mbuf *m, *n; 637 int space = 0; 638 639 if (control == 0) 640 panic("sbappendcontrol"); 641 for (m = control; ; m = m->m_next) { 642 space += m->m_len; 643 if (m->m_next == 0) 644 break; 645 } 646 n = m; /* save pointer to last control buffer */ 647 for (m = m0; m; m = m->m_next) 648 space += m->m_len; 649 if (space > sbspace(sb)) 650 return (0); 651 n->m_next = m0; /* concatenate data to control */ 652 for (m = control; m; m = m->m_next) 653 sballoc(sb, m); 654 n = sb->sb_mb; 655 if (n) { 656 while (n->m_nextpkt) 657 n = n->m_nextpkt; 658 n->m_nextpkt = control; 659 } else 660 sb->sb_mb = control; 661 return (1); 662} 663 664/* 665 * Compress mbuf chain m into the socket 666 * buffer sb following mbuf n. If n 667 * is null, the buffer is presumed empty. 668 */ 669void 670sbcompress(sb, m, n) 671 register struct sockbuf *sb; 672 register struct mbuf *m, *n; 673{ 674 register int eor = 0; 675 register struct mbuf *o; 676 677 while (m) { 678 eor |= m->m_flags & M_EOR; 679 if (m->m_len == 0 && 680 (eor == 0 || 681 (((o = m->m_next) || (o = n)) && 682 o->m_type == m->m_type))) { 683 m = m_free(m); 684 continue; 685 } 686 if (n && (n->m_flags & (M_EXT | M_EOR)) == 0 && 687 (n->m_data + n->m_len + m->m_len) < &n->m_dat[MLEN] && 688 n->m_type == m->m_type) { 689 bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len, 690 (unsigned)m->m_len); 691 n->m_len += m->m_len; 692 sb->sb_cc += m->m_len; 693 m = m_free(m); 694 continue; 695 } 696 if (n) 697 n->m_next = m; 698 else 699 sb->sb_mb = m; 700 sballoc(sb, m); 701 n = m; 702 m->m_flags &= ~M_EOR; 703 m = m->m_next; 704 n->m_next = 0; 705 } 706 if (eor) { 707 if (n) 708 n->m_flags |= eor; 709 else 710 printf("semi-panic: sbcompress\n"); 711 } 712} 713 714/* 715 * Free all mbufs in a sockbuf. 716 * Check that all resources are reclaimed. 717 */ 718void 719sbflush(sb) 720 register struct sockbuf *sb; 721{ 722 723 if (sb->sb_flags & SB_LOCK) 724 panic("sbflush: locked"); 725 while (sb->sb_mbcnt && sb->sb_cc) 726 sbdrop(sb, (int)sb->sb_cc); 727 if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt) 728 panic("sbflush: cc %ld || mb %p || mbcnt %ld", sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt); 729} 730 731/* 732 * Drop data from (the front of) a sockbuf. 733 */ 734void 735sbdrop(sb, len) 736 register struct sockbuf *sb; 737 register int len; 738{ 739 register struct mbuf *m, *mn; 740 struct mbuf *next; 741 742 next = (m = sb->sb_mb) ? m->m_nextpkt : 0; 743 while (len > 0) { 744 if (m == 0) { 745 if (next == 0) 746 panic("sbdrop"); 747 m = next; 748 next = m->m_nextpkt; 749 continue; 750 } 751 if (m->m_len > len) { 752 m->m_len -= len; 753 m->m_data += len; 754 sb->sb_cc -= len; 755 break; 756 } 757 len -= m->m_len; 758 sbfree(sb, m); 759 MFREE(m, mn); 760 m = mn; 761 } 762 while (m && m->m_len == 0) { 763 sbfree(sb, m); 764 MFREE(m, mn); 765 m = mn; 766 } 767 if (m) { 768 sb->sb_mb = m; 769 m->m_nextpkt = next; 770 } else 771 sb->sb_mb = next; 772} 773 774/* 775 * Drop a record off the front of a sockbuf 776 * and move the next record to the front. 777 */ 778void 779sbdroprecord(sb) 780 register struct sockbuf *sb; 781{ 782 register struct mbuf *m, *mn; 783 784 m = sb->sb_mb; 785 if (m) { 786 sb->sb_mb = m->m_nextpkt; 787 do { 788 sbfree(sb, m); 789 MFREE(m, mn); 790 m = mn; 791 } while (m); 792 } 793} 794 795/* 796 * Create a "control" mbuf containing the specified data 797 * with the specified type for presentation on a socket buffer. 798 */ 799struct mbuf * 800sbcreatecontrol(p, size, type, level) 801 caddr_t p; 802 register int size; 803 int type, level; 804{ 805 register struct cmsghdr *cp; 806 struct mbuf *m; 807 808 if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL) 809 return ((struct mbuf *) NULL); 810 cp = mtod(m, struct cmsghdr *); 811 /* XXX check size? */ 812 (void)memcpy(CMSG_DATA(cp), p, size); 813 size += sizeof(*cp); 814 m->m_len = size; 815 cp->cmsg_len = size; 816 cp->cmsg_level = level; 817 cp->cmsg_type = type; 818 return (m); 819} 820 821/* 822 * Some routines that return EOPNOTSUPP for entry points that are not 823 * supported by a protocol. Fill in as needed. 824 */ 825int 826pru_accept_notsupp(struct socket *so, struct sockaddr **nam) 827{ 828 return EOPNOTSUPP; 829} 830 831int 832pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct proc *p) 833{ 834 return EOPNOTSUPP; 835} 836 837int 838pru_connect2_notsupp(struct socket *so1, struct socket *so2) 839{ 840 return EOPNOTSUPP; 841} 842 843int 844pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data, 845 struct ifnet *ifp, struct proc *p) 846{ 847 return EOPNOTSUPP; 848} 849 850int 851pru_listen_notsupp(struct socket *so, struct proc *p) 852{ 853 return EOPNOTSUPP; 854} 855 856int 857pru_rcvd_notsupp(struct socket *so, int flags) 858{ 859 return EOPNOTSUPP; 860} 861 862int 863pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags) 864{ 865 return EOPNOTSUPP; 866} 867 868/* 869 * This isn't really a ``null'' operation, but it's the default one 870 * and doesn't do anything destructive. 871 */ 872int 873pru_sense_null(struct socket *so, struct stat *sb) 874{ 875 sb->st_blksize = so->so_snd.sb_hiwat; 876 return 0; 877} 878 879/* 880 * Make a copy of a sockaddr in a malloced buffer of type M_SONAME. 881 */ 882struct sockaddr * 883dup_sockaddr(sa, canwait) 884 struct sockaddr *sa; 885 int canwait; 886{ 887 struct sockaddr *sa2; 888 889 MALLOC(sa2, struct sockaddr *, sa->sa_len, M_SONAME, 890 canwait ? M_WAITOK : M_NOWAIT); 891 if (sa2) 892 bcopy(sa, sa2, sa->sa_len); 893 return sa2; 894} 895 896/* 897 * Create an external-format (``xsocket'') structure using the information 898 * in the kernel-format socket structure pointed to by so. This is done 899 * to reduce the spew of irrelevant information over this interface, 900 * to isolate user code from changes in the kernel structure, and 901 * potentially to provide information-hiding if we decide that 902 * some of this information should be hidden from users. 903 */ 904void 905sotoxsocket(struct socket *so, struct xsocket *xso) 906{ 907 xso->xso_len = sizeof *xso; 908 xso->xso_so = so; 909 xso->so_type = so->so_type; 910 xso->so_options = so->so_options; 911 xso->so_linger = so->so_linger; 912 xso->so_state = so->so_state; 913 xso->so_pcb = so->so_pcb; 914 xso->xso_protocol = so->so_proto->pr_protocol; 915 xso->xso_family = so->so_proto->pr_domain->dom_family; 916 xso->so_qlen = so->so_qlen; 917 xso->so_incqlen = so->so_incqlen; 918 xso->so_qlimit = so->so_qlimit; 919 xso->so_timeo = so->so_timeo; 920 xso->so_error = so->so_error; 921 xso->so_pgid = so->so_pgid; 922 xso->so_oobmark = so->so_oobmark; 923 sbtoxsockbuf(&so->so_snd, &xso->so_snd); 924 sbtoxsockbuf(&so->so_rcv, &xso->so_rcv); 925 xso->so_uid = so->so_uid; 926} 927 928/* 929 * This does the same for sockbufs. Note that the xsockbuf structure, 930 * since it is always embedded in a socket, does not include a self 931 * pointer nor a length. We make this entry point public in case 932 * some other mechanism needs it. 933 */ 934void 935sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb) 936{ 937 xsb->sb_cc = sb->sb_cc; 938 xsb->sb_hiwat = sb->sb_hiwat; 939 xsb->sb_mbcnt = sb->sb_mbcnt; 940 xsb->sb_mbmax = sb->sb_mbmax; 941 xsb->sb_lowat = sb->sb_lowat; 942 xsb->sb_flags = sb->sb_flags; 943 xsb->sb_timeo = sb->sb_timeo; 944} 945 946/* 947 * Here is the definition of some of the basic objects in the kern.ipc 948 * branch of the MIB. 949 */ 950SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC"); 951 952/* This takes the place of kern.maxsockbuf, which moved to kern.ipc. */ 953static int dummy; 954SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, ""); 955 956SYSCTL_INT(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLFLAG_RW, &sb_max, 0, ""); 957SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD, &maxsockets, 0, ""); 958SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW, 959 &sb_efficiency, 0, ""); 960SYSCTL_INT(_kern_ipc, KIPC_NMBCLUSTERS, nmbclusters, CTLFLAG_RD, &nmbclusters, 0, ""); 961 962