1 /* $NetBSD: uipc_socket2.c,v 1.41 2001/08/05 08:25:39 enami 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/param.h> 39 #include <sys/systm.h> 40 #include <sys/proc.h> 41 #include <sys/file.h> 42 #include <sys/buf.h> 43 #include <sys/malloc.h> 44 #include <sys/mbuf.h> 45 #include <sys/protosw.h> 46 #include <sys/socket.h> 47 #include <sys/socketvar.h> 48 #include <sys/signalvar.h> 49 50 /* 51 * Primitive routines for operating on sockets and socket buffers 52 */ 53 54 /* strings for sleep message: */ 55 const char netcon[] = "netcon"; 56 const char netcls[] = "netcls"; 57 const char netio[] = "netio"; 58 const char netlck[] = "netlck"; 59 60 /* 61 * Procedures to manipulate state flags of socket 62 * and do appropriate wakeups. Normal sequence from the 63 * active (originating) side is that soisconnecting() is 64 * called during processing of connect() call, 65 * resulting in an eventual call to soisconnected() if/when the 66 * connection is established. When the connection is torn down 67 * soisdisconnecting() is called during processing of disconnect() call, 68 * and soisdisconnected() is called when the connection to the peer 69 * is totally severed. The semantics of these routines are such that 70 * connectionless protocols can call soisconnected() and soisdisconnected() 71 * only, bypassing the in-progress calls when setting up a ``connection'' 72 * takes no time. 73 * 74 * From the passive side, a socket is created with 75 * two queues of sockets: so_q0 for connections in progress 76 * and so_q for connections already made and awaiting user acceptance. 77 * As a protocol is preparing incoming connections, it creates a socket 78 * structure queued on so_q0 by calling sonewconn(). When the connection 79 * is established, soisconnected() is called, and transfers the 80 * socket structure to so_q, making it available to accept(). 81 * 82 * If a socket is closed with sockets on either 83 * so_q0 or so_q, these sockets are dropped. 84 * 85 * If higher level protocols are implemented in 86 * the kernel, the wakeups done here will sometimes 87 * cause software-interrupt process scheduling. 88 */ 89 90 void 91 soisconnecting(struct socket *so) 92 { 93 94 so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING); 95 so->so_state |= SS_ISCONNECTING; 96 } 97 98 void 99 soisconnected(struct socket *so) 100 { 101 struct socket *head; 102 103 head = so->so_head; 104 so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING); 105 so->so_state |= SS_ISCONNECTED; 106 if (head && soqremque(so, 0)) { 107 soqinsque(head, so, 1); 108 sorwakeup(head); 109 wakeup((caddr_t)&head->so_timeo); 110 } else { 111 wakeup((caddr_t)&so->so_timeo); 112 sorwakeup(so); 113 sowwakeup(so); 114 } 115 } 116 117 void 118 soisdisconnecting(struct socket *so) 119 { 120 121 so->so_state &= ~SS_ISCONNECTING; 122 so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE); 123 wakeup((caddr_t)&so->so_timeo); 124 sowwakeup(so); 125 sorwakeup(so); 126 } 127 128 void 129 soisdisconnected(struct socket *so) 130 { 131 132 so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING); 133 so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED); 134 wakeup((caddr_t)&so->so_timeo); 135 sowwakeup(so); 136 sorwakeup(so); 137 } 138 139 /* 140 * When an attempt at a new connection is noted on a socket 141 * which accepts connections, sonewconn is called. If the 142 * connection is possible (subject to space constraints, etc.) 143 * then we allocate a new structure, propoerly linked into the 144 * data structure of the original socket, and return this. 145 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED. 146 * 147 * Currently, sonewconn() is defined as sonewconn1() in socketvar.h 148 * to catch calls that are missing the (new) second parameter. 149 */ 150 struct socket * 151 sonewconn1(struct socket *head, int connstatus) 152 { 153 struct socket *so; 154 int soqueue; 155 156 soqueue = connstatus ? 1 : 0; 157 if (head->so_qlen + head->so_q0len > 3 * head->so_qlimit / 2) 158 return ((struct socket *)0); 159 so = pool_get(&socket_pool, PR_NOWAIT); 160 if (so == NULL) 161 return (NULL); 162 memset((caddr_t)so, 0, sizeof(*so)); 163 so->so_type = head->so_type; 164 so->so_options = head->so_options &~ SO_ACCEPTCONN; 165 so->so_linger = head->so_linger; 166 so->so_state = head->so_state | SS_NOFDREF; 167 so->so_proto = head->so_proto; 168 so->so_timeo = head->so_timeo; 169 so->so_pgid = head->so_pgid; 170 so->so_send = head->so_send; 171 so->so_receive = head->so_receive; 172 so->so_uid = head->so_uid; 173 (void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat); 174 soqinsque(head, so, soqueue); 175 if ((*so->so_proto->pr_usrreq)(so, PRU_ATTACH, 176 (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0, 177 (struct proc *)0)) { 178 (void) soqremque(so, soqueue); 179 pool_put(&socket_pool, so); 180 return (NULL); 181 } 182 if (connstatus) { 183 sorwakeup(head); 184 wakeup((caddr_t)&head->so_timeo); 185 so->so_state |= connstatus; 186 } 187 return (so); 188 } 189 190 void 191 soqinsque(struct socket *head, struct socket *so, int q) 192 { 193 194 #ifdef DIAGNOSTIC 195 if (so->so_onq != NULL) 196 panic("soqinsque"); 197 #endif 198 199 so->so_head = head; 200 if (q == 0) { 201 head->so_q0len++; 202 so->so_onq = &head->so_q0; 203 } else { 204 head->so_qlen++; 205 so->so_onq = &head->so_q; 206 } 207 TAILQ_INSERT_TAIL(so->so_onq, so, so_qe); 208 } 209 210 int 211 soqremque(struct socket *so, int q) 212 { 213 struct socket *head; 214 215 head = so->so_head; 216 if (q == 0) { 217 if (so->so_onq != &head->so_q0) 218 return (0); 219 head->so_q0len--; 220 } else { 221 if (so->so_onq != &head->so_q) 222 return (0); 223 head->so_qlen--; 224 } 225 TAILQ_REMOVE(so->so_onq, so, so_qe); 226 so->so_onq = NULL; 227 so->so_head = NULL; 228 return (1); 229 } 230 231 /* 232 * Socantsendmore indicates that no more data will be sent on the 233 * socket; it would normally be applied to a socket when the user 234 * informs the system that no more data is to be sent, by the protocol 235 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data 236 * will be received, and will normally be applied to the socket by a 237 * protocol when it detects that the peer will send no more data. 238 * Data queued for reading in the socket may yet be read. 239 */ 240 241 void 242 socantsendmore(struct socket *so) 243 { 244 245 so->so_state |= SS_CANTSENDMORE; 246 sowwakeup(so); 247 } 248 249 void 250 socantrcvmore(struct socket *so) 251 { 252 253 so->so_state |= SS_CANTRCVMORE; 254 sorwakeup(so); 255 } 256 257 /* 258 * Wait for data to arrive at/drain from a socket buffer. 259 */ 260 int 261 sbwait(struct sockbuf *sb) 262 { 263 264 sb->sb_flags |= SB_WAIT; 265 return (tsleep((caddr_t)&sb->sb_cc, 266 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, netio, 267 sb->sb_timeo)); 268 } 269 270 /* 271 * Lock a sockbuf already known to be locked; 272 * return any error returned from sleep (EINTR). 273 */ 274 int 275 sb_lock(struct sockbuf *sb) 276 { 277 int error; 278 279 while (sb->sb_flags & SB_LOCK) { 280 sb->sb_flags |= SB_WANT; 281 error = tsleep((caddr_t)&sb->sb_flags, 282 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH, 283 netlck, 0); 284 if (error) 285 return (error); 286 } 287 sb->sb_flags |= SB_LOCK; 288 return (0); 289 } 290 291 /* 292 * Wakeup processes waiting on a socket buffer. 293 * Do asynchronous notification via SIGIO 294 * if the socket buffer has the SB_ASYNC flag set. 295 */ 296 void 297 sowakeup(struct socket *so, struct sockbuf *sb) 298 { 299 struct proc *p; 300 301 selwakeup(&sb->sb_sel); 302 sb->sb_flags &= ~SB_SEL; 303 if (sb->sb_flags & SB_WAIT) { 304 sb->sb_flags &= ~SB_WAIT; 305 wakeup((caddr_t)&sb->sb_cc); 306 } 307 if (sb->sb_flags & SB_ASYNC) { 308 if (so->so_pgid < 0) 309 gsignal(-so->so_pgid, SIGIO); 310 else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0) 311 psignal(p, SIGIO); 312 } 313 if (sb->sb_flags & SB_UPCALL) 314 (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT); 315 } 316 317 /* 318 * Socket buffer (struct sockbuf) utility routines. 319 * 320 * Each socket contains two socket buffers: one for sending data and 321 * one for receiving data. Each buffer contains a queue of mbufs, 322 * information about the number of mbufs and amount of data in the 323 * queue, and other fields allowing poll() statements and notification 324 * on data availability to be implemented. 325 * 326 * Data stored in a socket buffer is maintained as a list of records. 327 * Each record is a list of mbufs chained together with the m_next 328 * field. Records are chained together with the m_nextpkt field. The upper 329 * level routine soreceive() expects the following conventions to be 330 * observed when placing information in the receive buffer: 331 * 332 * 1. If the protocol requires each message be preceded by the sender's 333 * name, then a record containing that name must be present before 334 * any associated data (mbuf's must be of type MT_SONAME). 335 * 2. If the protocol supports the exchange of ``access rights'' (really 336 * just additional data associated with the message), and there are 337 * ``rights'' to be received, then a record containing this data 338 * should be present (mbuf's must be of type MT_CONTROL). 339 * 3. If a name or rights record exists, then it must be followed by 340 * a data record, perhaps of zero length. 341 * 342 * Before using a new socket structure it is first necessary to reserve 343 * buffer space to the socket, by calling sbreserve(). This should commit 344 * some of the available buffer space in the system buffer pool for the 345 * socket (currently, it does nothing but enforce limits). The space 346 * should be released by calling sbrelease() when the socket is destroyed. 347 */ 348 349 int 350 soreserve(struct socket *so, u_long sndcc, u_long rcvcc) 351 { 352 353 if (sbreserve(&so->so_snd, sndcc) == 0) 354 goto bad; 355 if (sbreserve(&so->so_rcv, rcvcc) == 0) 356 goto bad2; 357 if (so->so_rcv.sb_lowat == 0) 358 so->so_rcv.sb_lowat = 1; 359 if (so->so_snd.sb_lowat == 0) 360 so->so_snd.sb_lowat = MCLBYTES; 361 if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat) 362 so->so_snd.sb_lowat = so->so_snd.sb_hiwat; 363 return (0); 364 bad2: 365 sbrelease(&so->so_snd); 366 bad: 367 return (ENOBUFS); 368 } 369 370 /* 371 * Allot mbufs to a sockbuf. 372 * Attempt to scale mbmax so that mbcnt doesn't become limiting 373 * if buffering efficiency is near the normal case. 374 */ 375 int 376 sbreserve(struct sockbuf *sb, u_long cc) 377 { 378 379 if (cc == 0 || 380 (u_quad_t) cc > (u_quad_t) sb_max * MCLBYTES / (MSIZE + MCLBYTES)) 381 return (0); 382 sb->sb_hiwat = cc; 383 sb->sb_mbmax = min(cc * 2, sb_max); 384 if (sb->sb_lowat > sb->sb_hiwat) 385 sb->sb_lowat = sb->sb_hiwat; 386 return (1); 387 } 388 389 /* 390 * Free mbufs held by a socket, and reserved mbuf space. 391 */ 392 void 393 sbrelease(struct sockbuf *sb) 394 { 395 396 sbflush(sb); 397 sb->sb_hiwat = sb->sb_mbmax = 0; 398 } 399 400 /* 401 * Routines to add and remove 402 * data from an mbuf queue. 403 * 404 * The routines sbappend() or sbappendrecord() are normally called to 405 * append new mbufs to a socket buffer, after checking that adequate 406 * space is available, comparing the function sbspace() with the amount 407 * of data to be added. sbappendrecord() differs from sbappend() in 408 * that data supplied is treated as the beginning of a new record. 409 * To place a sender's address, optional access rights, and data in a 410 * socket receive buffer, sbappendaddr() should be used. To place 411 * access rights and data in a socket receive buffer, sbappendrights() 412 * should be used. In either case, the new data begins a new record. 413 * Note that unlike sbappend() and sbappendrecord(), these routines check 414 * for the caller that there will be enough space to store the data. 415 * Each fails if there is not enough space, or if it cannot find mbufs 416 * to store additional information in. 417 * 418 * Reliable protocols may use the socket send buffer to hold data 419 * awaiting acknowledgement. Data is normally copied from a socket 420 * send buffer in a protocol with m_copy for output to a peer, 421 * and then removing the data from the socket buffer with sbdrop() 422 * or sbdroprecord() when the data is acknowledged by the peer. 423 */ 424 425 /* 426 * Append mbuf chain m to the last record in the 427 * socket buffer sb. The additional space associated 428 * the mbuf chain is recorded in sb. Empty mbufs are 429 * discarded and mbufs are compacted where possible. 430 */ 431 void 432 sbappend(struct sockbuf *sb, struct mbuf *m) 433 { 434 struct mbuf *n; 435 436 if (m == 0) 437 return; 438 if ((n = sb->sb_mb) != NULL) { 439 while (n->m_nextpkt) 440 n = n->m_nextpkt; 441 do { 442 if (n->m_flags & M_EOR) { 443 sbappendrecord(sb, m); /* XXXXXX!!!! */ 444 return; 445 } 446 } while (n->m_next && (n = n->m_next)); 447 } 448 sbcompress(sb, m, n); 449 } 450 451 #ifdef SOCKBUF_DEBUG 452 void 453 sbcheck(struct sockbuf *sb) 454 { 455 struct mbuf *m; 456 int len, mbcnt; 457 458 len = 0; 459 mbcnt = 0; 460 for (m = sb->sb_mb; m; m = m->m_next) { 461 len += m->m_len; 462 mbcnt += MSIZE; 463 if (m->m_flags & M_EXT) 464 mbcnt += m->m_ext.ext_size; 465 if (m->m_nextpkt) 466 panic("sbcheck nextpkt"); 467 } 468 if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) { 469 printf("cc %d != %d || mbcnt %d != %d\n", len, sb->sb_cc, 470 mbcnt, sb->sb_mbcnt); 471 panic("sbcheck"); 472 } 473 } 474 #endif 475 476 /* 477 * As above, except the mbuf chain 478 * begins a new record. 479 */ 480 void 481 sbappendrecord(struct sockbuf *sb, struct mbuf *m0) 482 { 483 struct mbuf *m; 484 485 if (m0 == 0) 486 return; 487 if ((m = sb->sb_mb) != NULL) 488 while (m->m_nextpkt) 489 m = m->m_nextpkt; 490 /* 491 * Put the first mbuf on the queue. 492 * Note this permits zero length records. 493 */ 494 sballoc(sb, m0); 495 if (m) 496 m->m_nextpkt = m0; 497 else 498 sb->sb_mb = m0; 499 m = m0->m_next; 500 m0->m_next = 0; 501 if (m && (m0->m_flags & M_EOR)) { 502 m0->m_flags &= ~M_EOR; 503 m->m_flags |= M_EOR; 504 } 505 sbcompress(sb, m, m0); 506 } 507 508 /* 509 * As above except that OOB data 510 * is inserted at the beginning of the sockbuf, 511 * but after any other OOB data. 512 */ 513 void 514 sbinsertoob(struct sockbuf *sb, struct mbuf *m0) 515 { 516 struct mbuf *m, **mp; 517 518 if (m0 == 0) 519 return; 520 for (mp = &sb->sb_mb; (m = *mp) != NULL; mp = &((*mp)->m_nextpkt)) { 521 again: 522 switch (m->m_type) { 523 524 case MT_OOBDATA: 525 continue; /* WANT next train */ 526 527 case MT_CONTROL: 528 if ((m = m->m_next) != NULL) 529 goto again; /* inspect THIS train further */ 530 } 531 break; 532 } 533 /* 534 * Put the first mbuf on the queue. 535 * Note this permits zero length records. 536 */ 537 sballoc(sb, m0); 538 m0->m_nextpkt = *mp; 539 *mp = m0; 540 m = m0->m_next; 541 m0->m_next = 0; 542 if (m && (m0->m_flags & M_EOR)) { 543 m0->m_flags &= ~M_EOR; 544 m->m_flags |= M_EOR; 545 } 546 sbcompress(sb, m, m0); 547 } 548 549 /* 550 * Append address and data, and optionally, control (ancillary) data 551 * to the receive queue of a socket. If present, 552 * m0 must include a packet header with total length. 553 * Returns 0 if no space in sockbuf or insufficient mbufs. 554 */ 555 int 556 sbappendaddr(struct sockbuf *sb, struct sockaddr *asa, struct mbuf *m0, 557 struct mbuf *control) 558 { 559 struct mbuf *m, *n; 560 int space; 561 562 space = asa->sa_len; 563 564 if (m0 && (m0->m_flags & M_PKTHDR) == 0) 565 panic("sbappendaddr"); 566 if (m0) 567 space += m0->m_pkthdr.len; 568 for (n = control; n; n = n->m_next) { 569 space += n->m_len; 570 if (n->m_next == 0) /* keep pointer to last control buf */ 571 break; 572 } 573 if (space > sbspace(sb)) 574 return (0); 575 MGET(m, M_DONTWAIT, MT_SONAME); 576 if (m == 0) 577 return (0); 578 if (asa->sa_len > MLEN) { 579 MEXTMALLOC(m, asa->sa_len, M_NOWAIT); 580 if ((m->m_flags & M_EXT) == 0) { 581 m_free(m); 582 return (0); 583 } 584 } 585 m->m_len = asa->sa_len; 586 memcpy(mtod(m, caddr_t), (caddr_t)asa, asa->sa_len); 587 if (n) 588 n->m_next = m0; /* concatenate data to control */ 589 else 590 control = m0; 591 m->m_next = control; 592 for (n = m; n; n = n->m_next) 593 sballoc(sb, n); 594 if ((n = sb->sb_mb) != NULL) { 595 while (n->m_nextpkt) 596 n = n->m_nextpkt; 597 n->m_nextpkt = m; 598 } else 599 sb->sb_mb = m; 600 return (1); 601 } 602 603 int 604 sbappendcontrol(struct sockbuf *sb, struct mbuf *m0, struct mbuf *control) 605 { 606 struct mbuf *m, *n; 607 int space; 608 609 space = 0; 610 if (control == 0) 611 panic("sbappendcontrol"); 612 for (m = control; ; m = m->m_next) { 613 space += m->m_len; 614 if (m->m_next == 0) 615 break; 616 } 617 n = m; /* save pointer to last control buffer */ 618 for (m = m0; m; m = m->m_next) 619 space += m->m_len; 620 if (space > sbspace(sb)) 621 return (0); 622 n->m_next = m0; /* concatenate data to control */ 623 for (m = control; m; m = m->m_next) 624 sballoc(sb, m); 625 if ((n = sb->sb_mb) != NULL) { 626 while (n->m_nextpkt) 627 n = n->m_nextpkt; 628 n->m_nextpkt = control; 629 } else 630 sb->sb_mb = control; 631 return (1); 632 } 633 634 /* 635 * Compress mbuf chain m into the socket 636 * buffer sb following mbuf n. If n 637 * is null, the buffer is presumed empty. 638 */ 639 void 640 sbcompress(struct sockbuf *sb, struct mbuf *m, struct mbuf *n) 641 { 642 int eor; 643 struct mbuf *o; 644 645 eor = 0; 646 while (m) { 647 eor |= m->m_flags & M_EOR; 648 if (m->m_len == 0 && 649 (eor == 0 || 650 (((o = m->m_next) || (o = n)) && 651 o->m_type == m->m_type))) { 652 m = m_free(m); 653 continue; 654 } 655 if (n && (n->m_flags & M_EOR) == 0 && 656 /* M_TRAILINGSPACE() checks buffer writeability */ 657 m->m_len <= MCLBYTES / 4 && /* XXX Don't copy too much */ 658 m->m_len <= M_TRAILINGSPACE(n) && 659 n->m_type == m->m_type) { 660 memcpy(mtod(n, caddr_t) + n->m_len, mtod(m, caddr_t), 661 (unsigned)m->m_len); 662 n->m_len += m->m_len; 663 sb->sb_cc += m->m_len; 664 m = m_free(m); 665 continue; 666 } 667 if (n) 668 n->m_next = m; 669 else 670 sb->sb_mb = m; 671 sballoc(sb, m); 672 n = m; 673 m->m_flags &= ~M_EOR; 674 m = m->m_next; 675 n->m_next = 0; 676 } 677 if (eor) { 678 if (n) 679 n->m_flags |= eor; 680 else 681 printf("semi-panic: sbcompress\n"); 682 } 683 } 684 685 /* 686 * Free all mbufs in a sockbuf. 687 * Check that all resources are reclaimed. 688 */ 689 void 690 sbflush(struct sockbuf *sb) 691 { 692 693 if (sb->sb_flags & SB_LOCK) 694 panic("sbflush"); 695 while (sb->sb_mbcnt) 696 sbdrop(sb, (int)sb->sb_cc); 697 if (sb->sb_cc || sb->sb_mb) 698 panic("sbflush 2"); 699 } 700 701 /* 702 * Drop data from (the front of) a sockbuf. 703 */ 704 void 705 sbdrop(struct sockbuf *sb, int len) 706 { 707 struct mbuf *m, *mn, *next; 708 709 next = (m = sb->sb_mb) ? m->m_nextpkt : 0; 710 while (len > 0) { 711 if (m == 0) { 712 if (next == 0) 713 panic("sbdrop"); 714 m = next; 715 next = m->m_nextpkt; 716 continue; 717 } 718 if (m->m_len > len) { 719 m->m_len -= len; 720 m->m_data += len; 721 sb->sb_cc -= len; 722 break; 723 } 724 len -= m->m_len; 725 sbfree(sb, m); 726 MFREE(m, mn); 727 m = mn; 728 } 729 while (m && m->m_len == 0) { 730 sbfree(sb, m); 731 MFREE(m, mn); 732 m = mn; 733 } 734 if (m) { 735 sb->sb_mb = m; 736 m->m_nextpkt = next; 737 } else 738 sb->sb_mb = next; 739 } 740 741 /* 742 * Drop a record off the front of a sockbuf 743 * and move the next record to the front. 744 */ 745 void 746 sbdroprecord(struct sockbuf *sb) 747 { 748 struct mbuf *m, *mn; 749 750 m = sb->sb_mb; 751 if (m) { 752 sb->sb_mb = m->m_nextpkt; 753 do { 754 sbfree(sb, m); 755 MFREE(m, mn); 756 } while ((m = mn) != NULL); 757 } 758 } 759 760 /* 761 * Create a "control" mbuf containing the specified data 762 * with the specified type for presentation on a socket buffer. 763 */ 764 struct mbuf * 765 sbcreatecontrol(caddr_t p, int size, int type, int level) 766 { 767 struct cmsghdr *cp; 768 struct mbuf *m; 769 770 if (CMSG_SPACE(size) > MCLBYTES) { 771 printf("sbcreatecontrol: message too large %d\n", size); 772 return NULL; 773 } 774 775 if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL) 776 return ((struct mbuf *) NULL); 777 if (CMSG_SPACE(size) > MLEN) { 778 MCLGET(m, M_DONTWAIT); 779 if ((m->m_flags & M_EXT) == 0) { 780 m_free(m); 781 return NULL; 782 } 783 } 784 cp = mtod(m, struct cmsghdr *); 785 memcpy(CMSG_DATA(cp), p, size); 786 m->m_len = CMSG_SPACE(size); 787 cp->cmsg_len = CMSG_LEN(size); 788 cp->cmsg_level = level; 789 cp->cmsg_type = type; 790 return (m); 791 } 792