1 /* $NetBSD: uipc_socket2.c,v 1.37 2001/02/27 05:19:15 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/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 netio[] = "netio"; 56 const char netcon[] = "netcon"; 57 const char netcls[] = "netcls"; 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 89 void 90 soisconnecting(struct socket *so) 91 { 92 93 so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING); 94 so->so_state |= SS_ISCONNECTING; 95 } 96 97 void 98 soisconnected(struct socket *so) 99 { 100 struct socket *head; 101 102 head = so->so_head; 103 so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING); 104 so->so_state |= SS_ISCONNECTED; 105 if (head && soqremque(so, 0)) { 106 soqinsque(head, so, 1); 107 sorwakeup(head); 108 wakeup((caddr_t)&head->so_timeo); 109 } else { 110 wakeup((caddr_t)&so->so_timeo); 111 sorwakeup(so); 112 sowwakeup(so); 113 } 114 } 115 116 void 117 soisdisconnecting(struct socket *so) 118 { 119 120 so->so_state &= ~SS_ISCONNECTING; 121 so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE); 122 wakeup((caddr_t)&so->so_timeo); 123 sowwakeup(so); 124 sorwakeup(so); 125 } 126 127 void 128 soisdisconnected(struct socket *so) 129 { 130 131 so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING); 132 so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED); 133 wakeup((caddr_t)&so->so_timeo); 134 sowwakeup(so); 135 sorwakeup(so); 136 } 137 138 /* 139 * When an attempt at a new connection is noted on a socket 140 * which accepts connections, sonewconn is called. If the 141 * connection is possible (subject to space constraints, etc.) 142 * then we allocate a new structure, propoerly linked into the 143 * data structure of the original socket, and return this. 144 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED. 145 * 146 * Currently, sonewconn() is defined as sonewconn1() in socketvar.h 147 * to catch calls that are missing the (new) second parameter. 148 */ 149 struct socket * 150 sonewconn1(struct socket *head, int connstatus) 151 { 152 struct socket *so; 153 int soqueue; 154 155 soqueue = connstatus ? 1 : 0; 156 if (head->so_qlen + head->so_q0len > 3 * head->so_qlimit / 2) 157 return ((struct socket *)0); 158 so = pool_get(&socket_pool, PR_NOWAIT); 159 if (so == NULL) 160 return (NULL); 161 memset((caddr_t)so, 0, sizeof(*so)); 162 so->so_type = head->so_type; 163 so->so_options = head->so_options &~ SO_ACCEPTCONN; 164 so->so_linger = head->so_linger; 165 so->so_state = head->so_state | SS_NOFDREF; 166 so->so_proto = head->so_proto; 167 so->so_timeo = head->so_timeo; 168 so->so_pgid = head->so_pgid; 169 so->so_send = head->so_send; 170 so->so_receive = head->so_receive; 171 so->so_uid = head->so_uid; 172 (void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat); 173 soqinsque(head, so, soqueue); 174 if ((*so->so_proto->pr_usrreq)(so, PRU_ATTACH, 175 (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0, 176 (struct proc *)0)) { 177 (void) soqremque(so, soqueue); 178 pool_put(&socket_pool, so); 179 return (NULL); 180 } 181 if (connstatus) { 182 sorwakeup(head); 183 wakeup((caddr_t)&head->so_timeo); 184 so->so_state |= connstatus; 185 } 186 return (so); 187 } 188 189 void 190 soqinsque(struct socket *head, struct socket *so, int q) 191 { 192 193 #ifdef DIAGNOSTIC 194 if (so->so_onq != NULL) 195 panic("soqinsque"); 196 #endif 197 198 so->so_head = head; 199 if (q == 0) { 200 head->so_q0len++; 201 so->so_onq = &head->so_q0; 202 } else { 203 head->so_qlen++; 204 so->so_onq = &head->so_q; 205 } 206 TAILQ_INSERT_TAIL(so->so_onq, so, so_qe); 207 } 208 209 int 210 soqremque(struct socket *so, int q) 211 { 212 struct socket *head; 213 214 head = so->so_head; 215 if (q == 0) { 216 if (so->so_onq != &head->so_q0) 217 return (0); 218 head->so_q0len--; 219 } else { 220 if (so->so_onq != &head->so_q) 221 return (0); 222 head->so_qlen--; 223 } 224 TAILQ_REMOVE(so->so_onq, so, so_qe); 225 so->so_onq = NULL; 226 so->so_head = NULL; 227 return (1); 228 } 229 230 /* 231 * Socantsendmore indicates that no more data will be sent on the 232 * socket; it would normally be applied to a socket when the user 233 * informs the system that no more data is to be sent, by the protocol 234 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data 235 * will be received, and will normally be applied to the socket by a 236 * protocol when it detects that the peer will send no more data. 237 * Data queued for reading in the socket may yet be read. 238 */ 239 240 void 241 socantsendmore(struct socket *so) 242 { 243 244 so->so_state |= SS_CANTSENDMORE; 245 sowwakeup(so); 246 } 247 248 void 249 socantrcvmore(struct socket *so) 250 { 251 252 so->so_state |= SS_CANTRCVMORE; 253 sorwakeup(so); 254 } 255 256 /* 257 * Wait for data to arrive at/drain from a socket buffer. 258 */ 259 int 260 sbwait(struct sockbuf *sb) 261 { 262 263 sb->sb_flags |= SB_WAIT; 264 return (tsleep((caddr_t)&sb->sb_cc, 265 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, netio, 266 sb->sb_timeo)); 267 } 268 269 /* 270 * Lock a sockbuf already known to be locked; 271 * return any error returned from sleep (EINTR). 272 */ 273 int 274 sb_lock(struct sockbuf *sb) 275 { 276 int error; 277 278 while (sb->sb_flags & SB_LOCK) { 279 sb->sb_flags |= SB_WANT; 280 error = tsleep((caddr_t)&sb->sb_flags, 281 (sb->sb_flags & SB_NOINTR) ? 282 PSOCK : PSOCK|PCATCH, netio, 0); 283 if (error) 284 return (error); 285 } 286 sb->sb_flags |= SB_LOCK; 287 return (0); 288 } 289 290 /* 291 * Wakeup processes waiting on a socket buffer. 292 * Do asynchronous notification via SIGIO 293 * if the socket has the SS_ASYNC flag set. 294 */ 295 void 296 sowakeup(struct socket *so, struct sockbuf *sb) 297 { 298 struct proc *p; 299 300 selwakeup(&sb->sb_sel); 301 sb->sb_flags &= ~SB_SEL; 302 if (sb->sb_flags & SB_WAIT) { 303 sb->sb_flags &= ~SB_WAIT; 304 wakeup((caddr_t)&sb->sb_cc); 305 } 306 if (so->so_state & SS_ASYNC) { 307 if (so->so_pgid < 0) 308 gsignal(-so->so_pgid, SIGIO); 309 else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0) 310 psignal(p, SIGIO); 311 } 312 if (sb->sb_flags & SB_UPCALL) 313 (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT); 314 } 315 316 /* 317 * Socket buffer (struct sockbuf) utility routines. 318 * 319 * Each socket contains two socket buffers: one for sending data and 320 * one for receiving data. Each buffer contains a queue of mbufs, 321 * information about the number of mbufs and amount of data in the 322 * queue, and other fields allowing poll() statements and notification 323 * on data availability to be implemented. 324 * 325 * Data stored in a socket buffer is maintained as a list of records. 326 * Each record is a list of mbufs chained together with the m_next 327 * field. Records are chained together with the m_nextpkt field. The upper 328 * level routine soreceive() expects the following conventions to be 329 * observed when placing information in the receive buffer: 330 * 331 * 1. If the protocol requires each message be preceded by the sender's 332 * name, then a record containing that name must be present before 333 * any associated data (mbuf's must be of type MT_SONAME). 334 * 2. If the protocol supports the exchange of ``access rights'' (really 335 * just additional data associated with the message), and there are 336 * ``rights'' to be received, then a record containing this data 337 * should be present (mbuf's must be of type MT_CONTROL). 338 * 3. If a name or rights record exists, then it must be followed by 339 * a data record, perhaps of zero length. 340 * 341 * Before using a new socket structure it is first necessary to reserve 342 * buffer space to the socket, by calling sbreserve(). This should commit 343 * some of the available buffer space in the system buffer pool for the 344 * socket (currently, it does nothing but enforce limits). The space 345 * should be released by calling sbrelease() when the socket is destroyed. 346 */ 347 348 int 349 soreserve(struct socket *so, u_long sndcc, u_long rcvcc) 350 { 351 352 if (sbreserve(&so->so_snd, sndcc) == 0) 353 goto bad; 354 if (sbreserve(&so->so_rcv, rcvcc) == 0) 355 goto bad2; 356 if (so->so_rcv.sb_lowat == 0) 357 so->so_rcv.sb_lowat = 1; 358 if (so->so_snd.sb_lowat == 0) 359 so->so_snd.sb_lowat = MCLBYTES; 360 if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat) 361 so->so_snd.sb_lowat = so->so_snd.sb_hiwat; 362 return (0); 363 bad2: 364 sbrelease(&so->so_snd); 365 bad: 366 return (ENOBUFS); 367 } 368 369 /* 370 * Allot mbufs to a sockbuf. 371 * Attempt to scale mbmax so that mbcnt doesn't become limiting 372 * if buffering efficiency is near the normal case. 373 */ 374 int 375 sbreserve(struct sockbuf *sb, u_long cc) 376 { 377 378 if (cc == 0 || cc > sb_max * MCLBYTES / (MSIZE + MCLBYTES)) 379 return (0); 380 sb->sb_hiwat = cc; 381 sb->sb_mbmax = min(cc * 2, sb_max); 382 if (sb->sb_lowat > sb->sb_hiwat) 383 sb->sb_lowat = sb->sb_hiwat; 384 return (1); 385 } 386 387 /* 388 * Free mbufs held by a socket, and reserved mbuf space. 389 */ 390 void 391 sbrelease(struct sockbuf *sb) 392 { 393 394 sbflush(sb); 395 sb->sb_hiwat = sb->sb_mbmax = 0; 396 } 397 398 /* 399 * Routines to add and remove 400 * data from an mbuf queue. 401 * 402 * The routines sbappend() or sbappendrecord() are normally called to 403 * append new mbufs to a socket buffer, after checking that adequate 404 * space is available, comparing the function sbspace() with the amount 405 * of data to be added. sbappendrecord() differs from sbappend() in 406 * that data supplied is treated as the beginning of a new record. 407 * To place a sender's address, optional access rights, and data in a 408 * socket receive buffer, sbappendaddr() should be used. To place 409 * access rights and data in a socket receive buffer, sbappendrights() 410 * should be used. In either case, the new data begins a new record. 411 * Note that unlike sbappend() and sbappendrecord(), these routines check 412 * for the caller that there will be enough space to store the data. 413 * Each fails if there is not enough space, or if it cannot find mbufs 414 * to store additional information in. 415 * 416 * Reliable protocols may use the socket send buffer to hold data 417 * awaiting acknowledgement. Data is normally copied from a socket 418 * send buffer in a protocol with m_copy for output to a peer, 419 * and then removing the data from the socket buffer with sbdrop() 420 * or sbdroprecord() when the data is acknowledged by the peer. 421 */ 422 423 /* 424 * Append mbuf chain m to the last record in the 425 * socket buffer sb. The additional space associated 426 * the mbuf chain is recorded in sb. Empty mbufs are 427 * discarded and mbufs are compacted where possible. 428 */ 429 void 430 sbappend(struct sockbuf *sb, struct mbuf *m) 431 { 432 struct mbuf *n; 433 434 if (m == 0) 435 return; 436 if ((n = sb->sb_mb) != NULL) { 437 while (n->m_nextpkt) 438 n = n->m_nextpkt; 439 do { 440 if (n->m_flags & M_EOR) { 441 sbappendrecord(sb, m); /* XXXXXX!!!! */ 442 return; 443 } 444 } while (n->m_next && (n = n->m_next)); 445 } 446 sbcompress(sb, m, n); 447 } 448 449 #ifdef SOCKBUF_DEBUG 450 void 451 sbcheck(struct sockbuf *sb) 452 { 453 struct mbuf *m; 454 int len, mbcnt; 455 456 len = 0; 457 mbcnt = 0; 458 for (m = sb->sb_mb; m; m = m->m_next) { 459 len += m->m_len; 460 mbcnt += MSIZE; 461 if (m->m_flags & M_EXT) 462 mbcnt += m->m_ext.ext_size; 463 if (m->m_nextpkt) 464 panic("sbcheck nextpkt"); 465 } 466 if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) { 467 printf("cc %d != %d || mbcnt %d != %d\n", len, sb->sb_cc, 468 mbcnt, sb->sb_mbcnt); 469 panic("sbcheck"); 470 } 471 } 472 #endif 473 474 /* 475 * As above, except the mbuf chain 476 * begins a new record. 477 */ 478 void 479 sbappendrecord(struct sockbuf *sb, struct mbuf *m0) 480 { 481 struct mbuf *m; 482 483 if (m0 == 0) 484 return; 485 if ((m = sb->sb_mb) != NULL) 486 while (m->m_nextpkt) 487 m = m->m_nextpkt; 488 /* 489 * Put the first mbuf on the queue. 490 * Note this permits zero length records. 491 */ 492 sballoc(sb, m0); 493 if (m) 494 m->m_nextpkt = m0; 495 else 496 sb->sb_mb = m0; 497 m = m0->m_next; 498 m0->m_next = 0; 499 if (m && (m0->m_flags & M_EOR)) { 500 m0->m_flags &= ~M_EOR; 501 m->m_flags |= M_EOR; 502 } 503 sbcompress(sb, m, m0); 504 } 505 506 /* 507 * As above except that OOB data 508 * is inserted at the beginning of the sockbuf, 509 * but after any other OOB data. 510 */ 511 void 512 sbinsertoob(struct sockbuf *sb, struct mbuf *m0) 513 { 514 struct mbuf *m, **mp; 515 516 if (m0 == 0) 517 return; 518 for (mp = &sb->sb_mb; (m = *mp) != NULL; mp = &((*mp)->m_nextpkt)) { 519 again: 520 switch (m->m_type) { 521 522 case MT_OOBDATA: 523 continue; /* WANT next train */ 524 525 case MT_CONTROL: 526 if ((m = m->m_next) != NULL) 527 goto again; /* inspect THIS train further */ 528 } 529 break; 530 } 531 /* 532 * Put the first mbuf on the queue. 533 * Note this permits zero length records. 534 */ 535 sballoc(sb, m0); 536 m0->m_nextpkt = *mp; 537 *mp = m0; 538 m = m0->m_next; 539 m0->m_next = 0; 540 if (m && (m0->m_flags & M_EOR)) { 541 m0->m_flags &= ~M_EOR; 542 m->m_flags |= M_EOR; 543 } 544 sbcompress(sb, m, m0); 545 } 546 547 /* 548 * Append address and data, and optionally, control (ancillary) data 549 * to the receive queue of a socket. If present, 550 * m0 must include a packet header with total length. 551 * Returns 0 if no space in sockbuf or insufficient mbufs. 552 */ 553 int 554 sbappendaddr(struct sockbuf *sb, struct sockaddr *asa, struct mbuf *m0, 555 struct mbuf *control) 556 { 557 struct mbuf *m, *n; 558 int space; 559 560 space = asa->sa_len; 561 562 if (m0 && (m0->m_flags & M_PKTHDR) == 0) 563 panic("sbappendaddr"); 564 if (m0) 565 space += m0->m_pkthdr.len; 566 for (n = control; n; n = n->m_next) { 567 space += n->m_len; 568 if (n->m_next == 0) /* keep pointer to last control buf */ 569 break; 570 } 571 if (space > sbspace(sb)) 572 return (0); 573 MGET(m, M_DONTWAIT, MT_SONAME); 574 if (m == 0) 575 return (0); 576 if (asa->sa_len > MLEN) { 577 MEXTMALLOC(m, asa->sa_len, M_NOWAIT); 578 if ((m->m_flags & M_EXT) == 0) { 579 m_free(m); 580 return (0); 581 } 582 } 583 m->m_len = asa->sa_len; 584 memcpy(mtod(m, caddr_t), (caddr_t)asa, asa->sa_len); 585 if (n) 586 n->m_next = m0; /* concatenate data to control */ 587 else 588 control = m0; 589 m->m_next = control; 590 for (n = m; n; n = n->m_next) 591 sballoc(sb, n); 592 if ((n = sb->sb_mb) != NULL) { 593 while (n->m_nextpkt) 594 n = n->m_nextpkt; 595 n->m_nextpkt = m; 596 } else 597 sb->sb_mb = m; 598 return (1); 599 } 600 601 int 602 sbappendcontrol(struct sockbuf *sb, struct mbuf *m0, struct mbuf *control) 603 { 604 struct mbuf *m, *n; 605 int space; 606 607 space = 0; 608 if (control == 0) 609 panic("sbappendcontrol"); 610 for (m = control; ; m = m->m_next) { 611 space += m->m_len; 612 if (m->m_next == 0) 613 break; 614 } 615 n = m; /* save pointer to last control buffer */ 616 for (m = m0; m; m = m->m_next) 617 space += m->m_len; 618 if (space > sbspace(sb)) 619 return (0); 620 n->m_next = m0; /* concatenate data to control */ 621 for (m = control; m; m = m->m_next) 622 sballoc(sb, m); 623 if ((n = sb->sb_mb) != NULL) { 624 while (n->m_nextpkt) 625 n = n->m_nextpkt; 626 n->m_nextpkt = control; 627 } else 628 sb->sb_mb = control; 629 return (1); 630 } 631 632 /* 633 * Compress mbuf chain m into the socket 634 * buffer sb following mbuf n. If n 635 * is null, the buffer is presumed empty. 636 */ 637 void 638 sbcompress(struct sockbuf *sb, struct mbuf *m, struct mbuf *n) 639 { 640 int eor; 641 struct mbuf *o; 642 643 eor = 0; 644 while (m) { 645 eor |= m->m_flags & M_EOR; 646 if (m->m_len == 0 && 647 (eor == 0 || 648 (((o = m->m_next) || (o = n)) && 649 o->m_type == m->m_type))) { 650 m = m_free(m); 651 continue; 652 } 653 if (n && (n->m_flags & M_EOR) == 0 && n->m_type == m->m_type && 654 (((n->m_flags & M_EXT) == 0 && 655 n->m_data + n->m_len + m->m_len <= &n->m_dat[MLEN]) || 656 ((~n->m_flags & (M_EXT|M_CLUSTER)) == 0 && 657 !MCLISREFERENCED(n) && 658 n->m_data + n->m_len + m->m_len <= 659 &n->m_ext.ext_buf[MCLBYTES]))) { 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