1 /* 2 * Copyright (c) 1982, 1986, 1989 Regents of the University of California. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms are permitted 6 * provided that the above copyright notice and this paragraph are 7 * duplicated in all such forms and that any documentation, 8 * advertising materials, and other materials related to such 9 * distribution and use acknowledge that the software was developed 10 * by the University of California, Berkeley. The name of the 11 * University may not be used to endorse or promote products derived 12 * from this software without specific prior written permission. 13 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR 14 * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED 15 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. 16 * 17 * @(#)uipc_usrreq.c 7.13 (Berkeley) 10/19/89 18 */ 19 20 #include "param.h" 21 #include "user.h" 22 #include "domain.h" 23 #include "protosw.h" 24 #include "socket.h" 25 #include "socketvar.h" 26 #include "unpcb.h" 27 #include "un.h" 28 #include "vnode.h" 29 #include "mount.h" 30 #include "file.h" 31 #include "stat.h" 32 #include "mbuf.h" /* XXX must appear after mount.h */ 33 34 /* 35 * Unix communications domain. 36 * 37 * TODO: 38 * SEQPACKET, RDM 39 * rethink name space problems 40 * need a proper out-of-band 41 */ 42 struct sockaddr sun_noname = { sizeof(sun_noname), AF_UNIX }; 43 ino_t unp_vno; /* prototype for fake vnode numbers */ 44 45 /*ARGSUSED*/ 46 uipc_usrreq(so, req, m, nam, rights) 47 struct socket *so; 48 int req; 49 struct mbuf *m, *nam, *rights; 50 { 51 struct unpcb *unp = sotounpcb(so); 52 register struct socket *so2; 53 int error = 0; 54 55 if (req == PRU_CONTROL) 56 return (EOPNOTSUPP); 57 if (req != PRU_SEND && rights && rights->m_len) { 58 error = EOPNOTSUPP; 59 goto release; 60 } 61 if (unp == 0 && req != PRU_ATTACH) { 62 error = EINVAL; 63 goto release; 64 } 65 switch (req) { 66 67 case PRU_ATTACH: 68 if (unp) { 69 error = EISCONN; 70 break; 71 } 72 error = unp_attach(so); 73 break; 74 75 case PRU_DETACH: 76 unp_detach(unp); 77 break; 78 79 case PRU_BIND: 80 error = unp_bind(unp, nam); 81 break; 82 83 case PRU_LISTEN: 84 if (unp->unp_vnode == 0) 85 error = EINVAL; 86 break; 87 88 case PRU_CONNECT: 89 error = unp_connect(so, nam); 90 break; 91 92 case PRU_CONNECT2: 93 error = unp_connect2(so, (struct socket *)nam); 94 break; 95 96 case PRU_DISCONNECT: 97 unp_disconnect(unp); 98 break; 99 100 case PRU_ACCEPT: 101 /* 102 * Pass back name of connected socket, 103 * if it was bound and we are still connected 104 * (our peer may have closed already!). 105 */ 106 if (unp->unp_conn && unp->unp_conn->unp_addr) { 107 nam->m_len = unp->unp_conn->unp_addr->m_len; 108 bcopy(mtod(unp->unp_conn->unp_addr, caddr_t), 109 mtod(nam, caddr_t), (unsigned)nam->m_len); 110 } else { 111 nam->m_len = sizeof(sun_noname); 112 *(mtod(nam, struct sockaddr *)) = sun_noname; 113 } 114 break; 115 116 case PRU_SHUTDOWN: 117 socantsendmore(so); 118 unp_usrclosed(unp); 119 break; 120 121 case PRU_RCVD: 122 switch (so->so_type) { 123 124 case SOCK_DGRAM: 125 panic("uipc 1"); 126 /*NOTREACHED*/ 127 128 case SOCK_STREAM: 129 #define rcv (&so->so_rcv) 130 #define snd (&so2->so_snd) 131 if (unp->unp_conn == 0) 132 break; 133 so2 = unp->unp_conn->unp_socket; 134 /* 135 * Adjust backpressure on sender 136 * and wakeup any waiting to write. 137 */ 138 snd->sb_mbmax += unp->unp_mbcnt - rcv->sb_mbcnt; 139 unp->unp_mbcnt = rcv->sb_mbcnt; 140 snd->sb_hiwat += unp->unp_cc - rcv->sb_cc; 141 unp->unp_cc = rcv->sb_cc; 142 sowwakeup(so2); 143 #undef snd 144 #undef rcv 145 break; 146 147 default: 148 panic("uipc 2"); 149 } 150 break; 151 152 case PRU_SEND: 153 if (rights) { 154 error = unp_internalize(rights); 155 if (error) 156 break; 157 } 158 switch (so->so_type) { 159 160 case SOCK_DGRAM: { 161 struct sockaddr *from; 162 163 if (nam) { 164 if (unp->unp_conn) { 165 error = EISCONN; 166 break; 167 } 168 error = unp_connect(so, nam); 169 if (error) 170 break; 171 } else { 172 if (unp->unp_conn == 0) { 173 error = ENOTCONN; 174 break; 175 } 176 } 177 so2 = unp->unp_conn->unp_socket; 178 if (unp->unp_addr) 179 from = mtod(unp->unp_addr, struct sockaddr *); 180 else 181 from = &sun_noname; 182 if (sbspace(&so2->so_rcv) > 0 && 183 sbappendaddr(&so2->so_rcv, from, m, rights)) { 184 sorwakeup(so2); 185 m = 0; 186 } else 187 error = ENOBUFS; 188 if (nam) 189 unp_disconnect(unp); 190 break; 191 } 192 193 case SOCK_STREAM: 194 #define rcv (&so2->so_rcv) 195 #define snd (&so->so_snd) 196 if (so->so_state & SS_CANTSENDMORE) { 197 error = EPIPE; 198 break; 199 } 200 if (unp->unp_conn == 0) 201 panic("uipc 3"); 202 so2 = unp->unp_conn->unp_socket; 203 /* 204 * Send to paired receive port, and then reduce 205 * send buffer hiwater marks to maintain backpressure. 206 * Wake up readers. 207 */ 208 if (rights) 209 (void)sbappendrights(rcv, m, rights); 210 else 211 sbappend(rcv, m); 212 snd->sb_mbmax -= 213 rcv->sb_mbcnt - unp->unp_conn->unp_mbcnt; 214 unp->unp_conn->unp_mbcnt = rcv->sb_mbcnt; 215 snd->sb_hiwat -= rcv->sb_cc - unp->unp_conn->unp_cc; 216 unp->unp_conn->unp_cc = rcv->sb_cc; 217 sorwakeup(so2); 218 m = 0; 219 #undef snd 220 #undef rcv 221 break; 222 223 default: 224 panic("uipc 4"); 225 } 226 break; 227 228 case PRU_ABORT: 229 unp_drop(unp, ECONNABORTED); 230 break; 231 232 case PRU_SENSE: 233 ((struct stat *) m)->st_blksize = so->so_snd.sb_hiwat; 234 if (so->so_type == SOCK_STREAM && unp->unp_conn != 0) { 235 so2 = unp->unp_conn->unp_socket; 236 ((struct stat *) m)->st_blksize += so2->so_rcv.sb_cc; 237 } 238 ((struct stat *) m)->st_dev = NODEV; 239 if (unp->unp_vno == 0) 240 unp->unp_vno = unp_vno++; 241 ((struct stat *) m)->st_ino = unp->unp_vno; 242 return (0); 243 244 case PRU_RCVOOB: 245 return (EOPNOTSUPP); 246 247 case PRU_SENDOOB: 248 error = EOPNOTSUPP; 249 break; 250 251 case PRU_SOCKADDR: 252 if (unp->unp_addr) { 253 nam->m_len = unp->unp_addr->m_len; 254 bcopy(mtod(unp->unp_addr, caddr_t), 255 mtod(nam, caddr_t), (unsigned)nam->m_len); 256 } else 257 nam->m_len = 0; 258 break; 259 260 case PRU_PEERADDR: 261 if (unp->unp_conn && unp->unp_conn->unp_addr) { 262 nam->m_len = unp->unp_conn->unp_addr->m_len; 263 bcopy(mtod(unp->unp_conn->unp_addr, caddr_t), 264 mtod(nam, caddr_t), (unsigned)nam->m_len); 265 } else 266 nam->m_len = 0; 267 break; 268 269 case PRU_SLOWTIMO: 270 break; 271 272 default: 273 panic("piusrreq"); 274 } 275 release: 276 if (m) 277 m_freem(m); 278 return (error); 279 } 280 281 /* 282 * Both send and receive buffers are allocated PIPSIZ bytes of buffering 283 * for stream sockets, although the total for sender and receiver is 284 * actually only PIPSIZ. 285 * Datagram sockets really use the sendspace as the maximum datagram size, 286 * and don't really want to reserve the sendspace. Their recvspace should 287 * be large enough for at least one max-size datagram plus address. 288 */ 289 #define PIPSIZ 4096 290 u_long unpst_sendspace = PIPSIZ; 291 u_long unpst_recvspace = PIPSIZ; 292 u_long unpdg_sendspace = 2*1024; /* really max datagram size */ 293 u_long unpdg_recvspace = 4*1024; 294 295 int unp_rights; /* file descriptors in flight */ 296 297 unp_attach(so) 298 struct socket *so; 299 { 300 register struct mbuf *m; 301 register struct unpcb *unp; 302 int error; 303 304 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 305 switch (so->so_type) { 306 307 case SOCK_STREAM: 308 error = soreserve(so, unpst_sendspace, unpst_recvspace); 309 break; 310 311 case SOCK_DGRAM: 312 error = soreserve(so, unpdg_sendspace, unpdg_recvspace); 313 break; 314 } 315 if (error) 316 return (error); 317 } 318 m = m_getclr(M_DONTWAIT, MT_PCB); 319 if (m == NULL) 320 return (ENOBUFS); 321 unp = mtod(m, struct unpcb *); 322 so->so_pcb = (caddr_t)unp; 323 unp->unp_socket = so; 324 return (0); 325 } 326 327 unp_detach(unp) 328 register struct unpcb *unp; 329 { 330 331 if (unp->unp_vnode) { 332 unp->unp_vnode->v_socket = 0; 333 vrele(unp->unp_vnode); 334 unp->unp_vnode = 0; 335 } 336 if (unp->unp_conn) 337 unp_disconnect(unp); 338 while (unp->unp_refs) 339 unp_drop(unp->unp_refs, ECONNRESET); 340 soisdisconnected(unp->unp_socket); 341 unp->unp_socket->so_pcb = 0; 342 m_freem(unp->unp_addr); 343 (void) m_free(dtom(unp)); 344 if (unp_rights) 345 unp_gc(); 346 } 347 348 unp_bind(unp, nam) 349 struct unpcb *unp; 350 struct mbuf *nam; 351 { 352 struct sockaddr_un *soun = mtod(nam, struct sockaddr_un *); 353 register struct vnode *vp; 354 register struct nameidata *ndp = &u.u_nd; 355 struct vattr vattr; 356 int error; 357 358 ndp->ni_dirp = soun->sun_path; 359 if (unp->unp_vnode != NULL) 360 return (EINVAL); 361 if (nam->m_len == MLEN) { 362 if (*(mtod(nam, caddr_t) + nam->m_len - 1) != 0) 363 return (EINVAL); 364 } else 365 *(mtod(nam, caddr_t) + nam->m_len) = 0; 366 /* SHOULD BE ABLE TO ADOPT EXISTING AND wakeup() ALA FIFO's */ 367 ndp->ni_nameiop = CREATE | FOLLOW | LOCKPARENT; 368 ndp->ni_segflg = UIO_SYSSPACE; 369 if (error = namei(ndp)) 370 return (error); 371 vp = ndp->ni_vp; 372 if (vp != NULL) { 373 VOP_ABORTOP(ndp); 374 return (EADDRINUSE); 375 } 376 vattr_null(&vattr); 377 vattr.va_type = VSOCK; 378 vattr.va_mode = 0777; 379 if (error = VOP_CREATE(ndp, &vattr)) 380 return (error); 381 vp = ndp->ni_vp; 382 vp->v_socket = unp->unp_socket; 383 unp->unp_vnode = vp; 384 unp->unp_addr = m_copy(nam, 0, (int)M_COPYALL); 385 VOP_UNLOCK(vp); 386 return (0); 387 } 388 389 unp_connect(so, nam) 390 struct socket *so; 391 struct mbuf *nam; 392 { 393 register struct sockaddr_un *soun = mtod(nam, struct sockaddr_un *); 394 register struct vnode *vp; 395 register struct socket *so2, *so3; 396 register struct nameidata *ndp = &u.u_nd; 397 struct unpcb *unp2, *unp3; 398 int error; 399 400 ndp->ni_dirp = soun->sun_path; 401 if (nam->m_data + nam->m_len == &nam->m_dat[MLEN]) { /* XXX */ 402 if (*(mtod(nam, caddr_t) + nam->m_len - 1) != 0) 403 return (EMSGSIZE); 404 } else 405 *(mtod(nam, caddr_t) + nam->m_len) = 0; 406 ndp->ni_nameiop = LOOKUP | FOLLOW | LOCKLEAF; 407 ndp->ni_segflg = UIO_SYSSPACE; 408 if (error = namei(ndp)) 409 return (error); 410 vp = ndp->ni_vp; 411 if (vp->v_type != VSOCK) { 412 error = ENOTSOCK; 413 goto bad; 414 } 415 if (error = VOP_ACCESS(vp, VWRITE, ndp->ni_cred)) 416 goto bad; 417 so2 = vp->v_socket; 418 if (so2 == 0) { 419 error = ECONNREFUSED; 420 goto bad; 421 } 422 if (so->so_type != so2->so_type) { 423 error = EPROTOTYPE; 424 goto bad; 425 } 426 if (so->so_proto->pr_flags & PR_CONNREQUIRED) { 427 if ((so2->so_options & SO_ACCEPTCONN) == 0 || 428 (so3 = sonewconn(so2)) == 0) { 429 error = ECONNREFUSED; 430 goto bad; 431 } 432 unp2 = sotounpcb(so2); 433 unp3 = sotounpcb(so3); 434 if (unp2->unp_addr) 435 unp3->unp_addr = 436 m_copy(unp2->unp_addr, 0, (int)M_COPYALL); 437 so2 = so3; 438 } 439 error = unp_connect2(so, so2); 440 bad: 441 vput(vp); 442 return (error); 443 } 444 445 unp_connect2(so, so2) 446 register struct socket *so; 447 register struct socket *so2; 448 { 449 register struct unpcb *unp = sotounpcb(so); 450 register struct unpcb *unp2; 451 452 if (so2->so_type != so->so_type) 453 return (EPROTOTYPE); 454 unp2 = sotounpcb(so2); 455 unp->unp_conn = unp2; 456 switch (so->so_type) { 457 458 case SOCK_DGRAM: 459 unp->unp_nextref = unp2->unp_refs; 460 unp2->unp_refs = unp; 461 soisconnected(so); 462 break; 463 464 case SOCK_STREAM: 465 unp2->unp_conn = unp; 466 soisconnected(so2); 467 soisconnected(so); 468 break; 469 470 default: 471 panic("unp_connect2"); 472 } 473 return (0); 474 } 475 476 unp_disconnect(unp) 477 struct unpcb *unp; 478 { 479 register struct unpcb *unp2 = unp->unp_conn; 480 481 if (unp2 == 0) 482 return; 483 unp->unp_conn = 0; 484 switch (unp->unp_socket->so_type) { 485 486 case SOCK_DGRAM: 487 if (unp2->unp_refs == unp) 488 unp2->unp_refs = unp->unp_nextref; 489 else { 490 unp2 = unp2->unp_refs; 491 for (;;) { 492 if (unp2 == 0) 493 panic("unp_disconnect"); 494 if (unp2->unp_nextref == unp) 495 break; 496 unp2 = unp2->unp_nextref; 497 } 498 unp2->unp_nextref = unp->unp_nextref; 499 } 500 unp->unp_nextref = 0; 501 unp->unp_socket->so_state &= ~SS_ISCONNECTED; 502 break; 503 504 case SOCK_STREAM: 505 soisdisconnected(unp->unp_socket); 506 unp2->unp_conn = 0; 507 soisdisconnected(unp2->unp_socket); 508 break; 509 } 510 } 511 512 #ifdef notdef 513 unp_abort(unp) 514 struct unpcb *unp; 515 { 516 517 unp_detach(unp); 518 } 519 #endif 520 521 /*ARGSUSED*/ 522 unp_usrclosed(unp) 523 struct unpcb *unp; 524 { 525 526 } 527 528 unp_drop(unp, errno) 529 struct unpcb *unp; 530 int errno; 531 { 532 struct socket *so = unp->unp_socket; 533 534 so->so_error = errno; 535 unp_disconnect(unp); 536 if (so->so_head) { 537 so->so_pcb = (caddr_t) 0; 538 m_freem(unp->unp_addr); 539 (void) m_free(dtom(unp)); 540 sofree(so); 541 } 542 } 543 544 #ifdef notdef 545 unp_drain() 546 { 547 548 } 549 #endif 550 551 unp_externalize(rights) 552 struct mbuf *rights; 553 { 554 int newfds = rights->m_len / sizeof (int); 555 register int i; 556 register struct file **rp = mtod(rights, struct file **); 557 register struct file *fp; 558 int f; 559 560 if (newfds > ufavail()) { 561 for (i = 0; i < newfds; i++) { 562 fp = *rp; 563 unp_discard(fp); 564 *rp++ = 0; 565 } 566 return (EMSGSIZE); 567 } 568 for (i = 0; i < newfds; i++) { 569 if (ufalloc(0, &f)) 570 panic("unp_externalize"); 571 fp = *rp; 572 u.u_ofile[f] = fp; 573 fp->f_msgcount--; 574 unp_rights--; 575 *(int *)rp++ = f; 576 } 577 return (0); 578 } 579 580 unp_internalize(rights) 581 struct mbuf *rights; 582 { 583 register struct file **rp; 584 int oldfds = rights->m_len / sizeof (int); 585 register int i, fd; 586 register struct file *fp; 587 588 rp = mtod(rights, struct file **); 589 for (i = 0; i < oldfds; i++) { 590 fd = *(int *)rp++; 591 if ((unsigned)fd >= NOFILE || u.u_ofile[fd] == NULL) 592 return (EBADF); 593 } 594 rp = mtod(rights, struct file **); 595 for (i = 0; i < oldfds; i++) { 596 fp = u.u_ofile[*(int *)rp]; 597 *rp++ = fp; 598 fp->f_count++; 599 fp->f_msgcount++; 600 unp_rights++; 601 } 602 return (0); 603 } 604 605 int unp_defer, unp_gcing; 606 int unp_mark(); 607 extern struct domain unixdomain; 608 609 unp_gc() 610 { 611 register struct file *fp; 612 register struct socket *so; 613 614 if (unp_gcing) 615 return; 616 unp_gcing = 1; 617 restart: 618 unp_defer = 0; 619 for (fp = file; fp < fileNFILE; fp++) 620 fp->f_flag &= ~(FMARK|FDEFER); 621 do { 622 for (fp = file; fp < fileNFILE; fp++) { 623 if (fp->f_count == 0) 624 continue; 625 if (fp->f_flag & FDEFER) { 626 fp->f_flag &= ~FDEFER; 627 unp_defer--; 628 } else { 629 if (fp->f_flag & FMARK) 630 continue; 631 if (fp->f_count == fp->f_msgcount) 632 continue; 633 fp->f_flag |= FMARK; 634 } 635 if (fp->f_type != DTYPE_SOCKET || 636 (so = (struct socket *)fp->f_data) == 0) 637 continue; 638 if (so->so_proto->pr_domain != &unixdomain || 639 (so->so_proto->pr_flags&PR_RIGHTS) == 0) 640 continue; 641 if (so->so_rcv.sb_flags & SB_LOCK) { 642 sbwait(&so->so_rcv); 643 goto restart; 644 } 645 unp_scan(so->so_rcv.sb_mb, unp_mark); 646 } 647 } while (unp_defer); 648 for (fp = file; fp < fileNFILE; fp++) { 649 if (fp->f_count == 0) 650 continue; 651 if (fp->f_count == fp->f_msgcount && (fp->f_flag & FMARK) == 0) 652 while (fp->f_msgcount) 653 unp_discard(fp); 654 } 655 unp_gcing = 0; 656 } 657 658 unp_dispose(m) 659 struct mbuf *m; 660 { 661 int unp_discard(); 662 663 if (m) 664 unp_scan(m, unp_discard); 665 } 666 667 unp_scan(m0, op) 668 register struct mbuf *m0; 669 int (*op)(); 670 { 671 register struct mbuf *m; 672 register struct file **rp; 673 register int i; 674 int qfds; 675 676 while (m0) { 677 for (m = m0; m; m = m->m_next) 678 if (m->m_type == MT_RIGHTS && m->m_len) { 679 qfds = m->m_len / sizeof (struct file *); 680 rp = mtod(m, struct file **); 681 for (i = 0; i < qfds; i++) 682 (*op)(*rp++); 683 break; /* XXX, but saves time */ 684 } 685 m0 = m0->m_act; 686 } 687 } 688 689 unp_mark(fp) 690 struct file *fp; 691 { 692 693 if (fp->f_flag & FMARK) 694 return; 695 unp_defer++; 696 fp->f_flag |= (FMARK|FDEFER); 697 } 698 699 unp_discard(fp) 700 struct file *fp; 701 { 702 703 fp->f_msgcount--; 704 unp_rights--; 705 (void) closef(fp); 706 } 707