1 /* $NetBSD: uipc_socket2.c,v 1.42 2001/11/12 15:25:33 lukem Exp $ */ 2 3 /* 4 * Copyright (c) 1982, 1986, 1988, 1990, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed by the University of 18 * California, Berkeley and its contributors. 19 * 4. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * @(#)uipc_socket2.c 8.2 (Berkeley) 2/14/95 36 */ 37 38 #include <sys/cdefs.h> 39 __KERNEL_RCSID(0, "$NetBSD: uipc_socket2.c,v 1.42 2001/11/12 15:25:33 lukem Exp $"); 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/proc.h> 44 #include <sys/file.h> 45 #include <sys/buf.h> 46 #include <sys/malloc.h> 47 #include <sys/mbuf.h> 48 #include <sys/protosw.h> 49 #include <sys/socket.h> 50 #include <sys/socketvar.h> 51 #include <sys/signalvar.h> 52 53 /* 54 * Primitive routines for operating on sockets and socket buffers 55 */ 56 57 /* strings for sleep message: */ 58 const char netcon[] = "netcon"; 59 const char netcls[] = "netcls"; 60 const char netio[] = "netio"; 61 const char netlck[] = "netlck"; 62 63 /* 64 * Procedures to manipulate state flags of socket 65 * and do appropriate wakeups. Normal sequence from the 66 * active (originating) side is that soisconnecting() is 67 * called during processing of connect() call, 68 * resulting in an eventual call to soisconnected() if/when the 69 * connection is established. When the connection is torn down 70 * soisdisconnecting() is called during processing of disconnect() call, 71 * and soisdisconnected() is called when the connection to the peer 72 * is totally severed. The semantics of these routines are such that 73 * connectionless protocols can call soisconnected() and soisdisconnected() 74 * only, bypassing the in-progress calls when setting up a ``connection'' 75 * takes no time. 76 * 77 * From the passive side, a socket is created with 78 * two queues of sockets: so_q0 for connections in progress 79 * and so_q for connections already made and awaiting user acceptance. 80 * As a protocol is preparing incoming connections, it creates a socket 81 * structure queued on so_q0 by calling sonewconn(). When the connection 82 * is established, soisconnected() is called, and transfers the 83 * socket structure to so_q, making it available to accept(). 84 * 85 * If a socket is closed with sockets on either 86 * so_q0 or so_q, these sockets are dropped. 87 * 88 * If higher level protocols are implemented in 89 * the kernel, the wakeups done here will sometimes 90 * cause software-interrupt process scheduling. 91 */ 92 93 void 94 soisconnecting(struct socket *so) 95 { 96 97 so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING); 98 so->so_state |= SS_ISCONNECTING; 99 } 100 101 void 102 soisconnected(struct socket *so) 103 { 104 struct socket *head; 105 106 head = so->so_head; 107 so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING); 108 so->so_state |= SS_ISCONNECTED; 109 if (head && soqremque(so, 0)) { 110 soqinsque(head, so, 1); 111 sorwakeup(head); 112 wakeup((caddr_t)&head->so_timeo); 113 } else { 114 wakeup((caddr_t)&so->so_timeo); 115 sorwakeup(so); 116 sowwakeup(so); 117 } 118 } 119 120 void 121 soisdisconnecting(struct socket *so) 122 { 123 124 so->so_state &= ~SS_ISCONNECTING; 125 so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE); 126 wakeup((caddr_t)&so->so_timeo); 127 sowwakeup(so); 128 sorwakeup(so); 129 } 130 131 void 132 soisdisconnected(struct socket *so) 133 { 134 135 so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING); 136 so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED); 137 wakeup((caddr_t)&so->so_timeo); 138 sowwakeup(so); 139 sorwakeup(so); 140 } 141 142 /* 143 * When an attempt at a new connection is noted on a socket 144 * which accepts connections, sonewconn is called. If the 145 * connection is possible (subject to space constraints, etc.) 146 * then we allocate a new structure, propoerly linked into the 147 * data structure of the original socket, and return this. 148 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED. 149 * 150 * Currently, sonewconn() is defined as sonewconn1() in socketvar.h 151 * to catch calls that are missing the (new) second parameter. 152 */ 153 struct socket * 154 sonewconn1(struct socket *head, int connstatus) 155 { 156 struct socket *so; 157 int soqueue; 158 159 soqueue = connstatus ? 1 : 0; 160 if (head->so_qlen + head->so_q0len > 3 * head->so_qlimit / 2) 161 return ((struct socket *)0); 162 so = pool_get(&socket_pool, PR_NOWAIT); 163 if (so == NULL) 164 return (NULL); 165 memset((caddr_t)so, 0, sizeof(*so)); 166 so->so_type = head->so_type; 167 so->so_options = head->so_options &~ SO_ACCEPTCONN; 168 so->so_linger = head->so_linger; 169 so->so_state = head->so_state | SS_NOFDREF; 170 so->so_proto = head->so_proto; 171 so->so_timeo = head->so_timeo; 172 so->so_pgid = head->so_pgid; 173 so->so_send = head->so_send; 174 so->so_receive = head->so_receive; 175 so->so_uid = head->so_uid; 176 (void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat); 177 soqinsque(head, so, soqueue); 178 if ((*so->so_proto->pr_usrreq)(so, PRU_ATTACH, 179 (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0, 180 (struct proc *)0)) { 181 (void) soqremque(so, soqueue); 182 pool_put(&socket_pool, so); 183 return (NULL); 184 } 185 if (connstatus) { 186 sorwakeup(head); 187 wakeup((caddr_t)&head->so_timeo); 188 so->so_state |= connstatus; 189 } 190 return (so); 191 } 192 193 void 194 soqinsque(struct socket *head, struct socket *so, int q) 195 { 196 197 #ifdef DIAGNOSTIC 198 if (so->so_onq != NULL) 199 panic("soqinsque"); 200 #endif 201 202 so->so_head = head; 203 if (q == 0) { 204 head->so_q0len++; 205 so->so_onq = &head->so_q0; 206 } else { 207 head->so_qlen++; 208 so->so_onq = &head->so_q; 209 } 210 TAILQ_INSERT_TAIL(so->so_onq, so, so_qe); 211 } 212 213 int 214 soqremque(struct socket *so, int q) 215 { 216 struct socket *head; 217 218 head = so->so_head; 219 if (q == 0) { 220 if (so->so_onq != &head->so_q0) 221 return (0); 222 head->so_q0len--; 223 } else { 224 if (so->so_onq != &head->so_q) 225 return (0); 226 head->so_qlen--; 227 } 228 TAILQ_REMOVE(so->so_onq, so, so_qe); 229 so->so_onq = NULL; 230 so->so_head = NULL; 231 return (1); 232 } 233 234 /* 235 * Socantsendmore indicates that no more data will be sent on the 236 * socket; it would normally be applied to a socket when the user 237 * informs the system that no more data is to be sent, by the protocol 238 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data 239 * will be received, and will normally be applied to the socket by a 240 * protocol when it detects that the peer will send no more data. 241 * Data queued for reading in the socket may yet be read. 242 */ 243 244 void 245 socantsendmore(struct socket *so) 246 { 247 248 so->so_state |= SS_CANTSENDMORE; 249 sowwakeup(so); 250 } 251 252 void 253 socantrcvmore(struct socket *so) 254 { 255 256 so->so_state |= SS_CANTRCVMORE; 257 sorwakeup(so); 258 } 259 260 /* 261 * Wait for data to arrive at/drain from a socket buffer. 262 */ 263 int 264 sbwait(struct sockbuf *sb) 265 { 266 267 sb->sb_flags |= SB_WAIT; 268 return (tsleep((caddr_t)&sb->sb_cc, 269 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, netio, 270 sb->sb_timeo)); 271 } 272 273 /* 274 * Lock a sockbuf already known to be locked; 275 * return any error returned from sleep (EINTR). 276 */ 277 int 278 sb_lock(struct sockbuf *sb) 279 { 280 int error; 281 282 while (sb->sb_flags & SB_LOCK) { 283 sb->sb_flags |= SB_WANT; 284 error = tsleep((caddr_t)&sb->sb_flags, 285 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH, 286 netlck, 0); 287 if (error) 288 return (error); 289 } 290 sb->sb_flags |= SB_LOCK; 291 return (0); 292 } 293 294 /* 295 * Wakeup processes waiting on a socket buffer. 296 * Do asynchronous notification via SIGIO 297 * if the socket buffer has the SB_ASYNC flag set. 298 */ 299 void 300 sowakeup(struct socket *so, struct sockbuf *sb) 301 { 302 struct proc *p; 303 304 selwakeup(&sb->sb_sel); 305 sb->sb_flags &= ~SB_SEL; 306 if (sb->sb_flags & SB_WAIT) { 307 sb->sb_flags &= ~SB_WAIT; 308 wakeup((caddr_t)&sb->sb_cc); 309 } 310 if (sb->sb_flags & SB_ASYNC) { 311 if (so->so_pgid < 0) 312 gsignal(-so->so_pgid, SIGIO); 313 else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0) 314 psignal(p, SIGIO); 315 } 316 if (sb->sb_flags & SB_UPCALL) 317 (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT); 318 } 319 320 /* 321 * Socket buffer (struct sockbuf) utility routines. 322 * 323 * Each socket contains two socket buffers: one for sending data and 324 * one for receiving data. Each buffer contains a queue of mbufs, 325 * information about the number of mbufs and amount of data in the 326 * queue, and other fields allowing poll() statements and notification 327 * on data availability to be implemented. 328 * 329 * Data stored in a socket buffer is maintained as a list of records. 330 * Each record is a list of mbufs chained together with the m_next 331 * field. Records are chained together with the m_nextpkt field. The upper 332 * level routine soreceive() expects the following conventions to be 333 * observed when placing information in the receive buffer: 334 * 335 * 1. If the protocol requires each message be preceded by the sender's 336 * name, then a record containing that name must be present before 337 * any associated data (mbuf's must be of type MT_SONAME). 338 * 2. If the protocol supports the exchange of ``access rights'' (really 339 * just additional data associated with the message), and there are 340 * ``rights'' to be received, then a record containing this data 341 * should be present (mbuf's must be of type MT_CONTROL). 342 * 3. If a name or rights record exists, then it must be followed by 343 * a data record, perhaps of zero length. 344 * 345 * Before using a new socket structure it is first necessary to reserve 346 * buffer space to the socket, by calling sbreserve(). This should commit 347 * some of the available buffer space in the system buffer pool for the 348 * socket (currently, it does nothing but enforce limits). The space 349 * should be released by calling sbrelease() when the socket is destroyed. 350 */ 351 352 int 353 soreserve(struct socket *so, u_long sndcc, u_long rcvcc) 354 { 355 356 if (sbreserve(&so->so_snd, sndcc) == 0) 357 goto bad; 358 if (sbreserve(&so->so_rcv, rcvcc) == 0) 359 goto bad2; 360 if (so->so_rcv.sb_lowat == 0) 361 so->so_rcv.sb_lowat = 1; 362 if (so->so_snd.sb_lowat == 0) 363 so->so_snd.sb_lowat = MCLBYTES; 364 if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat) 365 so->so_snd.sb_lowat = so->so_snd.sb_hiwat; 366 return (0); 367 bad2: 368 sbrelease(&so->so_snd); 369 bad: 370 return (ENOBUFS); 371 } 372 373 /* 374 * Allot mbufs to a sockbuf. 375 * Attempt to scale mbmax so that mbcnt doesn't become limiting 376 * if buffering efficiency is near the normal case. 377 */ 378 int 379 sbreserve(struct sockbuf *sb, u_long cc) 380 { 381 382 if (cc == 0 || 383 (u_quad_t) cc > (u_quad_t) sb_max * MCLBYTES / (MSIZE + MCLBYTES)) 384 return (0); 385 sb->sb_hiwat = cc; 386 sb->sb_mbmax = min(cc * 2, sb_max); 387 if (sb->sb_lowat > sb->sb_hiwat) 388 sb->sb_lowat = sb->sb_hiwat; 389 return (1); 390 } 391 392 /* 393 * Free mbufs held by a socket, and reserved mbuf space. 394 */ 395 void 396 sbrelease(struct sockbuf *sb) 397 { 398 399 sbflush(sb); 400 sb->sb_hiwat = sb->sb_mbmax = 0; 401 } 402 403 /* 404 * Routines to add and remove 405 * data from an mbuf queue. 406 * 407 * The routines sbappend() or sbappendrecord() are normally called to 408 * append new mbufs to a socket buffer, after checking that adequate 409 * space is available, comparing the function sbspace() with the amount 410 * of data to be added. sbappendrecord() differs from sbappend() in 411 * that data supplied is treated as the beginning of a new record. 412 * To place a sender's address, optional access rights, and data in a 413 * socket receive buffer, sbappendaddr() should be used. To place 414 * access rights and data in a socket receive buffer, sbappendrights() 415 * should be used. In either case, the new data begins a new record. 416 * Note that unlike sbappend() and sbappendrecord(), these routines check 417 * for the caller that there will be enough space to store the data. 418 * Each fails if there is not enough space, or if it cannot find mbufs 419 * to store additional information in. 420 * 421 * Reliable protocols may use the socket send buffer to hold data 422 * awaiting acknowledgement. Data is normally copied from a socket 423 * send buffer in a protocol with m_copy for output to a peer, 424 * and then removing the data from the socket buffer with sbdrop() 425 * or sbdroprecord() when the data is acknowledged by the peer. 426 */ 427 428 /* 429 * Append mbuf chain m to the last record in the 430 * socket buffer sb. The additional space associated 431 * the mbuf chain is recorded in sb. Empty mbufs are 432 * discarded and mbufs are compacted where possible. 433 */ 434 void 435 sbappend(struct sockbuf *sb, struct mbuf *m) 436 { 437 struct mbuf *n; 438 439 if (m == 0) 440 return; 441 if ((n = sb->sb_mb) != NULL) { 442 while (n->m_nextpkt) 443 n = n->m_nextpkt; 444 do { 445 if (n->m_flags & M_EOR) { 446 sbappendrecord(sb, m); /* XXXXXX!!!! */ 447 return; 448 } 449 } while (n->m_next && (n = n->m_next)); 450 } 451 sbcompress(sb, m, n); 452 } 453 454 #ifdef SOCKBUF_DEBUG 455 void 456 sbcheck(struct sockbuf *sb) 457 { 458 struct mbuf *m; 459 int len, mbcnt; 460 461 len = 0; 462 mbcnt = 0; 463 for (m = sb->sb_mb; m; m = m->m_next) { 464 len += m->m_len; 465 mbcnt += MSIZE; 466 if (m->m_flags & M_EXT) 467 mbcnt += m->m_ext.ext_size; 468 if (m->m_nextpkt) 469 panic("sbcheck nextpkt"); 470 } 471 if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) { 472 printf("cc %d != %d || mbcnt %d != %d\n", len, sb->sb_cc, 473 mbcnt, sb->sb_mbcnt); 474 panic("sbcheck"); 475 } 476 } 477 #endif 478 479 /* 480 * As above, except the mbuf chain 481 * begins a new record. 482 */ 483 void 484 sbappendrecord(struct sockbuf *sb, struct mbuf *m0) 485 { 486 struct mbuf *m; 487 488 if (m0 == 0) 489 return; 490 if ((m = sb->sb_mb) != NULL) 491 while (m->m_nextpkt) 492 m = m->m_nextpkt; 493 /* 494 * Put the first mbuf on the queue. 495 * Note this permits zero length records. 496 */ 497 sballoc(sb, m0); 498 if (m) 499 m->m_nextpkt = m0; 500 else 501 sb->sb_mb = m0; 502 m = m0->m_next; 503 m0->m_next = 0; 504 if (m && (m0->m_flags & M_EOR)) { 505 m0->m_flags &= ~M_EOR; 506 m->m_flags |= M_EOR; 507 } 508 sbcompress(sb, m, m0); 509 } 510 511 /* 512 * As above except that OOB data 513 * is inserted at the beginning of the sockbuf, 514 * but after any other OOB data. 515 */ 516 void 517 sbinsertoob(struct sockbuf *sb, struct mbuf *m0) 518 { 519 struct mbuf *m, **mp; 520 521 if (m0 == 0) 522 return; 523 for (mp = &sb->sb_mb; (m = *mp) != NULL; mp = &((*mp)->m_nextpkt)) { 524 again: 525 switch (m->m_type) { 526 527 case MT_OOBDATA: 528 continue; /* WANT next train */ 529 530 case MT_CONTROL: 531 if ((m = m->m_next) != NULL) 532 goto again; /* inspect THIS train further */ 533 } 534 break; 535 } 536 /* 537 * Put the first mbuf on the queue. 538 * Note this permits zero length records. 539 */ 540 sballoc(sb, m0); 541 m0->m_nextpkt = *mp; 542 *mp = m0; 543 m = m0->m_next; 544 m0->m_next = 0; 545 if (m && (m0->m_flags & M_EOR)) { 546 m0->m_flags &= ~M_EOR; 547 m->m_flags |= M_EOR; 548 } 549 sbcompress(sb, m, m0); 550 } 551 552 /* 553 * Append address and data, and optionally, control (ancillary) data 554 * to the receive queue of a socket. If present, 555 * m0 must include a packet header with total length. 556 * Returns 0 if no space in sockbuf or insufficient mbufs. 557 */ 558 int 559 sbappendaddr(struct sockbuf *sb, struct sockaddr *asa, struct mbuf *m0, 560 struct mbuf *control) 561 { 562 struct mbuf *m, *n; 563 int space; 564 565 space = asa->sa_len; 566 567 if (m0 && (m0->m_flags & M_PKTHDR) == 0) 568 panic("sbappendaddr"); 569 if (m0) 570 space += m0->m_pkthdr.len; 571 for (n = control; n; n = n->m_next) { 572 space += n->m_len; 573 if (n->m_next == 0) /* keep pointer to last control buf */ 574 break; 575 } 576 if (space > sbspace(sb)) 577 return (0); 578 MGET(m, M_DONTWAIT, MT_SONAME); 579 if (m == 0) 580 return (0); 581 if (asa->sa_len > MLEN) { 582 MEXTMALLOC(m, asa->sa_len, M_NOWAIT); 583 if ((m->m_flags & M_EXT) == 0) { 584 m_free(m); 585 return (0); 586 } 587 } 588 m->m_len = asa->sa_len; 589 memcpy(mtod(m, caddr_t), (caddr_t)asa, asa->sa_len); 590 if (n) 591 n->m_next = m0; /* concatenate data to control */ 592 else 593 control = m0; 594 m->m_next = control; 595 for (n = m; n; n = n->m_next) 596 sballoc(sb, n); 597 if ((n = sb->sb_mb) != NULL) { 598 while (n->m_nextpkt) 599 n = n->m_nextpkt; 600 n->m_nextpkt = m; 601 } else 602 sb->sb_mb = m; 603 return (1); 604 } 605 606 int 607 sbappendcontrol(struct sockbuf *sb, struct mbuf *m0, struct mbuf *control) 608 { 609 struct mbuf *m, *n; 610 int space; 611 612 space = 0; 613 if (control == 0) 614 panic("sbappendcontrol"); 615 for (m = control; ; m = m->m_next) { 616 space += m->m_len; 617 if (m->m_next == 0) 618 break; 619 } 620 n = m; /* save pointer to last control buffer */ 621 for (m = m0; m; m = m->m_next) 622 space += m->m_len; 623 if (space > sbspace(sb)) 624 return (0); 625 n->m_next = m0; /* concatenate data to control */ 626 for (m = control; m; m = m->m_next) 627 sballoc(sb, m); 628 if ((n = sb->sb_mb) != NULL) { 629 while (n->m_nextpkt) 630 n = n->m_nextpkt; 631 n->m_nextpkt = control; 632 } else 633 sb->sb_mb = control; 634 return (1); 635 } 636 637 /* 638 * Compress mbuf chain m into the socket 639 * buffer sb following mbuf n. If n 640 * is null, the buffer is presumed empty. 641 */ 642 void 643 sbcompress(struct sockbuf *sb, struct mbuf *m, struct mbuf *n) 644 { 645 int eor; 646 struct mbuf *o; 647 648 eor = 0; 649 while (m) { 650 eor |= m->m_flags & M_EOR; 651 if (m->m_len == 0 && 652 (eor == 0 || 653 (((o = m->m_next) || (o = n)) && 654 o->m_type == m->m_type))) { 655 m = m_free(m); 656 continue; 657 } 658 if (n && (n->m_flags & M_EOR) == 0 && 659 /* M_TRAILINGSPACE() checks buffer writeability */ 660 m->m_len <= MCLBYTES / 4 && /* XXX Don't copy too much */ 661 m->m_len <= M_TRAILINGSPACE(n) && 662 n->m_type == m->m_type) { 663 memcpy(mtod(n, caddr_t) + n->m_len, mtod(m, caddr_t), 664 (unsigned)m->m_len); 665 n->m_len += m->m_len; 666 sb->sb_cc += m->m_len; 667 m = m_free(m); 668 continue; 669 } 670 if (n) 671 n->m_next = m; 672 else 673 sb->sb_mb = m; 674 sballoc(sb, m); 675 n = m; 676 m->m_flags &= ~M_EOR; 677 m = m->m_next; 678 n->m_next = 0; 679 } 680 if (eor) { 681 if (n) 682 n->m_flags |= eor; 683 else 684 printf("semi-panic: sbcompress\n"); 685 } 686 } 687 688 /* 689 * Free all mbufs in a sockbuf. 690 * Check that all resources are reclaimed. 691 */ 692 void 693 sbflush(struct sockbuf *sb) 694 { 695 696 if (sb->sb_flags & SB_LOCK) 697 panic("sbflush"); 698 while (sb->sb_mbcnt) 699 sbdrop(sb, (int)sb->sb_cc); 700 if (sb->sb_cc || sb->sb_mb) 701 panic("sbflush 2"); 702 } 703 704 /* 705 * Drop data from (the front of) a sockbuf. 706 */ 707 void 708 sbdrop(struct sockbuf *sb, int len) 709 { 710 struct mbuf *m, *mn, *next; 711 712 next = (m = sb->sb_mb) ? m->m_nextpkt : 0; 713 while (len > 0) { 714 if (m == 0) { 715 if (next == 0) 716 panic("sbdrop"); 717 m = next; 718 next = m->m_nextpkt; 719 continue; 720 } 721 if (m->m_len > len) { 722 m->m_len -= len; 723 m->m_data += len; 724 sb->sb_cc -= len; 725 break; 726 } 727 len -= m->m_len; 728 sbfree(sb, m); 729 MFREE(m, mn); 730 m = mn; 731 } 732 while (m && m->m_len == 0) { 733 sbfree(sb, m); 734 MFREE(m, mn); 735 m = mn; 736 } 737 if (m) { 738 sb->sb_mb = m; 739 m->m_nextpkt = next; 740 } else 741 sb->sb_mb = next; 742 } 743 744 /* 745 * Drop a record off the front of a sockbuf 746 * and move the next record to the front. 747 */ 748 void 749 sbdroprecord(struct sockbuf *sb) 750 { 751 struct mbuf *m, *mn; 752 753 m = sb->sb_mb; 754 if (m) { 755 sb->sb_mb = m->m_nextpkt; 756 do { 757 sbfree(sb, m); 758 MFREE(m, mn); 759 } while ((m = mn) != NULL); 760 } 761 } 762 763 /* 764 * Create a "control" mbuf containing the specified data 765 * with the specified type for presentation on a socket buffer. 766 */ 767 struct mbuf * 768 sbcreatecontrol(caddr_t p, int size, int type, int level) 769 { 770 struct cmsghdr *cp; 771 struct mbuf *m; 772 773 if (CMSG_SPACE(size) > MCLBYTES) { 774 printf("sbcreatecontrol: message too large %d\n", size); 775 return NULL; 776 } 777 778 if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL) 779 return ((struct mbuf *) NULL); 780 if (CMSG_SPACE(size) > MLEN) { 781 MCLGET(m, M_DONTWAIT); 782 if ((m->m_flags & M_EXT) == 0) { 783 m_free(m); 784 return NULL; 785 } 786 } 787 cp = mtod(m, struct cmsghdr *); 788 memcpy(CMSG_DATA(cp), p, size); 789 m->m_len = CMSG_SPACE(size); 790 cp->cmsg_len = CMSG_LEN(size); 791 cp->cmsg_level = level; 792 cp->cmsg_type = type; 793 return (m); 794 } 795