1 /* $NetBSD: linux_socket.c,v 1.60 2006/03/27 21:27:02 dyoung Exp $ */ 2 3 /*- 4 * Copyright (c) 1995, 1998 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Frank van der Linden and Eric Haszlakiewicz. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the NetBSD 21 * Foundation, Inc. and its contributors. 22 * 4. Neither the name of The NetBSD Foundation nor the names of its 23 * contributors may be used to endorse or promote products derived 24 * from this software without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 36 * POSSIBILITY OF SUCH DAMAGE. 37 */ 38 39 /* 40 * Functions in multiarch: 41 * linux_sys_socketcall : linux_socketcall.c 42 */ 43 44 #include <sys/cdefs.h> 45 __KERNEL_RCSID(0, "$NetBSD: linux_socket.c,v 1.60 2006/03/27 21:27:02 dyoung Exp $"); 46 47 #if defined(_KERNEL_OPT) 48 #include "opt_inet.h" 49 #endif 50 51 #include <sys/param.h> 52 #include <sys/kernel.h> 53 #include <sys/systm.h> 54 #include <sys/buf.h> 55 #include <sys/malloc.h> 56 #include <sys/ioctl.h> 57 #include <sys/tty.h> 58 #include <sys/file.h> 59 #include <sys/filedesc.h> 60 #include <sys/select.h> 61 #include <sys/socket.h> 62 #include <sys/socketvar.h> 63 #include <sys/domain.h> 64 #include <net/if.h> 65 #include <net/if_dl.h> 66 #include <net/if_types.h> 67 #include <netinet/in.h> 68 #include <netinet/tcp.h> 69 #include <sys/mount.h> 70 #include <sys/proc.h> 71 #include <sys/vnode.h> 72 #include <sys/device.h> 73 #include <sys/protosw.h> 74 #include <sys/mbuf.h> 75 #include <sys/syslog.h> 76 #include <sys/exec.h> 77 78 #include <sys/sa.h> 79 #include <sys/syscallargs.h> 80 81 #include <lib/libkern/libkern.h> 82 83 #ifdef INET6 84 #include <netinet/ip6.h> 85 #include <netinet6/ip6_var.h> 86 #endif 87 88 #include <compat/sys/socket.h> 89 90 #include <compat/linux/common/linux_types.h> 91 #include <compat/linux/common/linux_util.h> 92 #include <compat/linux/common/linux_signal.h> 93 #include <compat/linux/common/linux_ioctl.h> 94 #include <compat/linux/common/linux_socket.h> 95 #if !defined(__alpha__) && !defined(__amd64__) 96 #include <compat/linux/common/linux_socketcall.h> 97 #endif 98 #include <compat/linux/common/linux_sockio.h> 99 100 #include <compat/linux/linux_syscallargs.h> 101 102 #ifdef DEBUG_LINUX 103 #define DPRINTF(a) uprintf a 104 #else 105 #define DPRINTF(a) 106 #endif 107 108 /* 109 * The calls in this file are entered either via the linux_socketcall() 110 * interface or, on the Alpha, as individual syscalls. The 111 * linux_socketcall function does any massaging of arguments so that all 112 * the calls in here need not think that they are anything other 113 * than a normal syscall. 114 */ 115 116 static int linux_to_bsd_domain __P((int)); 117 static int bsd_to_linux_domain __P((int)); 118 int linux_to_bsd_sopt_level __P((int)); 119 int linux_to_bsd_so_sockopt __P((int)); 120 int linux_to_bsd_ip_sockopt __P((int)); 121 int linux_to_bsd_tcp_sockopt __P((int)); 122 int linux_to_bsd_udp_sockopt __P((int)); 123 int linux_getifhwaddr __P((struct lwp *, register_t *, u_int, void *)); 124 static int linux_sa_get __P((struct lwp *, int, caddr_t *, struct sockaddr **, 125 const struct osockaddr *, int *)); 126 static int linux_sa_put __P((struct osockaddr *osa)); 127 static int linux_to_bsd_msg_flags __P((int)); 128 static int bsd_to_linux_msg_flags __P((int)); 129 130 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = { 131 AF_UNSPEC, 132 AF_UNIX, 133 AF_INET, 134 AF_CCITT, /* LINUX_AF_AX25 */ 135 AF_IPX, 136 AF_APPLETALK, 137 -1, /* LINUX_AF_NETROM */ 138 -1, /* LINUX_AF_BRIDGE */ 139 -1, /* LINUX_AF_ATMPVC */ 140 AF_CCITT, /* LINUX_AF_X25 */ 141 AF_INET6, 142 -1, /* LINUX_AF_ROSE */ 143 AF_DECnet, 144 -1, /* LINUX_AF_NETBEUI */ 145 -1, /* LINUX_AF_SECURITY */ 146 pseudo_AF_KEY, 147 AF_ROUTE, /* LINUX_AF_NETLINK */ 148 -1, /* LINUX_AF_PACKET */ 149 -1, /* LINUX_AF_ASH */ 150 -1, /* LINUX_AF_ECONET */ 151 -1, /* LINUX_AF_ATMSVC */ 152 AF_SNA, 153 /* rest up to LINUX_AF_MAX-1 is not allocated */ 154 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 155 }; 156 157 static const int bsd_to_linux_domain_[AF_MAX] = { 158 LINUX_AF_UNSPEC, 159 LINUX_AF_UNIX, 160 LINUX_AF_INET, 161 -1, /* AF_IMPLINK */ 162 -1, /* AF_PUP */ 163 -1, /* AF_CHAOS */ 164 -1, /* AF_NS */ 165 -1, /* AF_ISO */ 166 -1, /* AF_ECMA */ 167 -1, /* AF_DATAKIT */ 168 LINUX_AF_AX25, /* AF_CCITT */ 169 LINUX_AF_SNA, 170 LINUX_AF_DECnet, 171 -1, /* AF_DLI */ 172 -1, /* AF_LAT */ 173 -1, /* AF_HYLINK */ 174 LINUX_AF_APPLETALK, 175 LINUX_AF_NETLINK, 176 -1, /* AF_LINK */ 177 -1, /* AF_XTP */ 178 -1, /* AF_COIP */ 179 -1, /* AF_CNT */ 180 -1, /* pseudo_AF_RTIP */ 181 LINUX_AF_IPX, 182 LINUX_AF_INET6, 183 -1, /* pseudo_AF_PIP */ 184 -1, /* AF_ISDN */ 185 -1, /* AF_NATM */ 186 -1, /* AF_ARP */ 187 LINUX_pseudo_AF_KEY, 188 -1, /* pseudo_AF_HDRCMPLT */ 189 }; 190 191 static const struct { 192 int bfl; 193 int lfl; 194 } bsd_to_linux_msg_flags_[] = { 195 {MSG_OOB, LINUX_MSG_OOB}, 196 {MSG_PEEK, LINUX_MSG_PEEK}, 197 {MSG_DONTROUTE, LINUX_MSG_DONTROUTE}, 198 {MSG_EOR, LINUX_MSG_EOR}, 199 {MSG_TRUNC, LINUX_MSG_TRUNC}, 200 {MSG_CTRUNC, LINUX_MSG_CTRUNC}, 201 {MSG_WAITALL, LINUX_MSG_WAITALL}, 202 {MSG_DONTWAIT, LINUX_MSG_DONTWAIT}, 203 {MSG_BCAST, 0}, /* not supported, clear */ 204 {MSG_MCAST, 0}, /* not supported, clear */ 205 {-1, /* not supp */ LINUX_MSG_PROBE}, 206 {-1, /* not supp */ LINUX_MSG_FIN}, 207 {-1, /* not supp */ LINUX_MSG_SYN}, 208 {-1, /* not supp */ LINUX_MSG_CONFIRM}, 209 {-1, /* not supp */ LINUX_MSG_RST}, 210 {-1, /* not supp */ LINUX_MSG_ERRQUEUE}, 211 {-1, /* not supp */ LINUX_MSG_NOSIGNAL}, 212 {-1, /* not supp */ LINUX_MSG_MORE}, 213 }; 214 215 /* 216 * Convert between Linux and BSD socket domain values 217 */ 218 static int 219 linux_to_bsd_domain(ldom) 220 int ldom; 221 { 222 if (ldom < 0 || ldom >= LINUX_AF_MAX) 223 return (-1); 224 225 return linux_to_bsd_domain_[ldom]; 226 } 227 228 /* 229 * Convert between BSD and Linux socket domain values 230 */ 231 static int 232 bsd_to_linux_domain(bdom) 233 int bdom; 234 { 235 if (bdom < 0 || bdom >= AF_MAX) 236 return (-1); 237 238 return bsd_to_linux_domain_[bdom]; 239 } 240 241 static int 242 linux_to_bsd_msg_flags(lflag) 243 int lflag; 244 { 245 int i, lfl, bfl; 246 int bflag = 0; 247 248 if (lflag == 0) 249 return (0); 250 251 for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) { 252 bfl = bsd_to_linux_msg_flags_[i].bfl; 253 lfl = bsd_to_linux_msg_flags_[i].lfl; 254 255 if (lfl == 0) 256 continue; 257 258 if (lflag & lfl) { 259 if (bfl < 0) 260 return (-1); 261 262 bflag |= bfl; 263 } 264 } 265 266 return (bflag); 267 } 268 269 static int 270 bsd_to_linux_msg_flags(bflag) 271 int bflag; 272 { 273 int i, lfl, bfl; 274 int lflag = 0; 275 276 if (bflag == 0) 277 return (0); 278 279 for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) { 280 bfl = bsd_to_linux_msg_flags_[i].bfl; 281 lfl = bsd_to_linux_msg_flags_[i].lfl; 282 283 if (bfl <= 0) 284 continue; 285 286 if (bflag & bfl) { 287 if (lfl < 0) 288 return (-1); 289 290 lflag |= lfl; 291 } 292 } 293 294 return (lflag); 295 } 296 297 int 298 linux_sys_socket(l, v, retval) 299 struct lwp *l; 300 void *v; 301 register_t *retval; 302 { 303 struct linux_sys_socket_args /* { 304 syscallarg(int) domain; 305 syscallarg(int) type; 306 syscallarg(int) protocol; 307 } */ *uap = v; 308 struct sys_socket_args bsa; 309 int error; 310 311 SCARG(&bsa, protocol) = SCARG(uap, protocol); 312 SCARG(&bsa, type) = SCARG(uap, type); 313 SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain)); 314 if (SCARG(&bsa, domain) == -1) 315 return EINVAL; 316 error = sys_socket(l, &bsa, retval); 317 318 #ifdef INET6 319 /* 320 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by 321 * default and some apps depend on this. So, set V6ONLY to 0 322 * for Linux apps if the sysctl value is set to 1. 323 */ 324 if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) { 325 struct proc *p = l->l_proc; 326 struct file *fp; 327 328 if (getsock(p->p_fd, *retval, &fp) == 0) { 329 struct mbuf *m; 330 331 m = m_get(M_WAIT, MT_SOOPTS); 332 m->m_len = sizeof(int); 333 *mtod(m, int *) = 0; 334 335 /* ignore error */ 336 (void) sosetopt((struct socket *)fp->f_data, 337 IPPROTO_IPV6, IPV6_V6ONLY, m); 338 339 FILE_UNUSE(fp, l); 340 } 341 } 342 #endif 343 344 return (error); 345 } 346 347 int 348 linux_sys_socketpair(l, v, retval) 349 struct lwp *l; 350 void *v; 351 register_t *retval; 352 { 353 struct linux_sys_socketpair_args /* { 354 syscallarg(int) domain; 355 syscallarg(int) type; 356 syscallarg(int) protocol; 357 syscallarg(int *) rsv; 358 } */ *uap = v; 359 struct sys_socketpair_args bsa; 360 361 SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain)); 362 if (SCARG(&bsa, domain) == -1) 363 return EINVAL; 364 SCARG(&bsa, type) = SCARG(uap, type); 365 SCARG(&bsa, protocol) = SCARG(uap, protocol); 366 SCARG(&bsa, rsv) = SCARG(uap, rsv); 367 368 return sys_socketpair(l, &bsa, retval); 369 } 370 371 int 372 linux_sys_sendto(l, v, retval) 373 struct lwp *l; 374 void *v; 375 register_t *retval; 376 { 377 struct linux_sys_sendto_args /* { 378 syscallarg(int) s; 379 syscallarg(void *) msg; 380 syscallarg(int) len; 381 syscallarg(int) flags; 382 syscallarg(struct osockaddr *) to; 383 syscallarg(int) tolen; 384 } */ *uap = v; 385 struct proc *p = l->l_proc; 386 struct sys_sendto_args bsa; 387 int tolen; 388 389 SCARG(&bsa, s) = SCARG(uap, s); 390 SCARG(&bsa, buf) = SCARG(uap, msg); 391 SCARG(&bsa, len) = (size_t) SCARG(uap, len); 392 SCARG(&bsa, flags) = SCARG(uap, flags); 393 tolen = SCARG(uap, tolen); 394 if (SCARG(uap, to)) { 395 struct sockaddr *sa; 396 int error; 397 caddr_t sg = stackgap_init(p, 0); 398 399 error = linux_sa_get(l, SCARG(uap, s), &sg, &sa, 400 SCARG(uap, to), &tolen); 401 if (error) 402 return (error); 403 404 SCARG(&bsa, to) = sa; 405 } else 406 SCARG(&bsa, to) = NULL; 407 SCARG(&bsa, tolen) = tolen; 408 409 return (sys_sendto(l, &bsa, retval)); 410 } 411 412 int 413 linux_sys_sendmsg(l, v, retval) 414 struct lwp *l; 415 void *v; 416 register_t *retval; 417 { 418 struct linux_sys_sendmsg_args /* { 419 syscallarg(int) s; 420 syscallarg(struct msghdr *) msg; 421 syscallarg(u_int) flags; 422 } */ *uap = v; 423 struct proc *p = l->l_proc; 424 struct msghdr msg; 425 int error; 426 struct iovec aiov[UIO_SMALLIOV], *iov; 427 caddr_t sg = 0; 428 int bflags; 429 u_int8_t *control=NULL; 430 431 error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg)); 432 if (error) 433 return (error); 434 if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) { 435 if ((unsigned int)msg.msg_iovlen > IOV_MAX) 436 return (EMSGSIZE); 437 iov = malloc(sizeof(struct iovec) * msg.msg_iovlen, 438 M_IOV, M_WAITOK); 439 } else 440 iov = aiov; 441 if ((unsigned int)msg.msg_iovlen > 0) { 442 error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov, 443 (size_t)(msg.msg_iovlen * sizeof(struct iovec))); 444 if (error) 445 goto done; 446 } 447 msg.msg_iov = iov; 448 msg.msg_flags = 0; 449 450 /* Convert the sockaddr if necessary */ 451 if (msg.msg_name) { 452 struct sockaddr *sa; 453 sg = stackgap_init(p, 0); 454 455 error = linux_sa_get(l, SCARG(uap, s), &sg, &sa, 456 (struct osockaddr *) msg.msg_name, &msg.msg_namelen); 457 if (error) 458 goto done; 459 msg.msg_name = sa; 460 } 461 462 /* 463 * Translate message flags. 464 */ 465 bflags = linux_to_bsd_msg_flags(SCARG(uap, flags)); 466 if (bflags < 0) { 467 /* Some supported flag */ 468 error = EINVAL; 469 goto done; 470 } 471 472 /* 473 * Handle cmsg if there is any. 474 */ 475 if (CMSG_FIRSTHDR(&msg)) { 476 struct cmsghdr cmsg, *cc; 477 int changed = 0; 478 ssize_t resid = msg.msg_controllen; 479 size_t clen, cidx = 0, cspace; 480 481 /* 482 * Limit the size even more than what sockargs() would do, 483 * We need to fit into stackgap space. 484 */ 485 if (msg.msg_controllen > (STACKGAPLEN / 2)) { 486 /* Sorry guys! */ 487 error = EMSGSIZE; 488 goto done; 489 } 490 491 control = malloc((clen = msg.msg_controllen), M_TEMP, M_WAITOK); 492 if (!control) { 493 error = ENOMEM; 494 goto done; 495 } 496 497 cc = CMSG_FIRSTHDR(&msg); 498 do { 499 error = copyin(cc, &cmsg, sizeof(cmsg)); 500 if (error) 501 goto done; 502 503 /* 504 * Sanity check the control message length. 505 */ 506 if (cmsg.cmsg_len > resid 507 || cmsg.cmsg_len < sizeof(struct cmsghdr)) { 508 error = EINVAL; 509 goto done; 510 } 511 512 /* 513 * Refuse unsupported control messages, and 514 * translate fields as appropriate. 515 */ 516 switch (cmsg.cmsg_level) { 517 case LINUX_SOL_SOCKET: 518 /* It only differs on some archs */ 519 if (LINUX_SOL_SOCKET != SOL_SOCKET) { 520 cmsg.cmsg_level = SOL_SOCKET; 521 changed = 1; 522 } 523 524 switch(cmsg.cmsg_type) { 525 case LINUX_SCM_RIGHTS: 526 /* Linux SCM_RIGHTS is same as NetBSD */ 527 break; 528 529 default: 530 /* other types not supported */ 531 error = EINVAL; 532 goto done; 533 } 534 break; 535 default: 536 /* pray and leave intact */ 537 break; 538 } 539 540 cspace = CMSG_SPACE(cmsg.cmsg_len - sizeof(cmsg)); 541 542 /* Check the buffer is big enough */ 543 if (__predict_false(cidx + cspace > clen)) { 544 u_int8_t *nc; 545 546 clen = cidx + cspace; 547 nc = realloc(control, clen, M_TEMP, M_WAITOK); 548 if (!nc) { 549 error = ENOMEM; 550 goto done; 551 } 552 control = nc; 553 } 554 555 /* Copy header */ 556 memcpy(&control[cidx], &cmsg, sizeof(cmsg)); 557 558 /* Zero are between header and data */ 559 memset(&control[cidx+sizeof(cmsg)], 0, 560 CMSG_ALIGN(sizeof(cmsg)) - sizeof(cmsg)); 561 562 /* Copyin the data */ 563 error = copyin(LINUX_CMSG_DATA(cc), 564 CMSG_DATA(control), 565 cmsg.cmsg_len - sizeof(cmsg)); 566 if (error) 567 goto done; 568 569 /* 570 * If there is alignment difference, we changed 571 * layout of cmsg. 572 */ 573 if (LINUX_CMSG_ALIGNDIFF) 574 changed = 1; 575 576 resid -= cspace; 577 cidx += cspace; 578 } while ((cc = LINUX_CMSG_NXTHDR(&msg, cc)) && resid > 0); 579 580 /* 581 * If any of the passed control message needed 582 * a change, put the changed data into stackgap 583 * and adjust msg appropriately. 584 */ 585 if (changed) { 586 char *newc; 587 588 /* 589 * Check again the total len is maximum half of 590 * stackgap. The length might change if the 591 * alignment is different. 592 */ 593 if (clen > STACKGAPLEN/2) { 594 error = EMSGSIZE; 595 goto done; 596 } 597 598 /* 599 * Allocate space on stack within stackgap, and 600 * copy changed data there. 601 */ 602 if (!sg) 603 sg = stackgap_init(p, STACKGAPLEN/3); 604 newc = stackgap_alloc(p, &sg, clen); 605 if (!newc) { 606 error = ENOMEM; 607 goto done; 608 } 609 610 error = copyout(control, newc, clen); 611 if (error) 612 goto done; 613 614 msg.msg_control = newc; 615 msg.msg_controllen = clen; 616 } 617 618 free(control, M_TEMP); 619 control = NULL; 620 } 621 622 error = sendit(l, SCARG(uap, s), &msg, bflags, retval); 623 624 done: 625 if (control) 626 free(control, M_TEMP); 627 if (iov != aiov) 628 free(iov, M_IOV); 629 return (error); 630 } 631 632 int 633 linux_sys_recvfrom(l, v, retval) 634 struct lwp *l; 635 void *v; 636 register_t *retval; 637 { 638 struct linux_sys_recvfrom_args /* { 639 syscallarg(int) s; 640 syscallarg(void *) buf; 641 syscallarg(int) len; 642 syscallarg(int) flags; 643 syscallarg(struct osockaddr *) from; 644 syscallarg(int *) fromlenaddr; 645 } */ *uap = v; 646 int error; 647 struct sys_recvfrom_args bra; 648 649 SCARG(&bra, s) = SCARG(uap, s); 650 SCARG(&bra, buf) = SCARG(uap, buf); 651 SCARG(&bra, len) = SCARG(uap, len); 652 SCARG(&bra, flags) = SCARG(uap, flags); 653 SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from); 654 SCARG(&bra, fromlenaddr) = SCARG(uap, fromlenaddr); 655 656 if ((error = sys_recvfrom(l, &bra, retval))) 657 return (error); 658 659 if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from)))) 660 return (error); 661 662 return (0); 663 } 664 665 int 666 linux_sys_recvmsg(l, v, retval) 667 struct lwp *l; 668 void *v; 669 register_t *retval; 670 { 671 struct linux_sys_recvmsg_args /* { 672 syscallarg(int) s; 673 syscallarg(struct msghdr *) msg; 674 syscallarg(u_int) flags; 675 } */ *uap = v; 676 struct msghdr msg; 677 int error; 678 struct sys_recvmsg_args bsa; 679 int lflags; 680 u_int8_t *ocontrol = NULL; /* XXX: gcc */ 681 socklen_t ocontrollen = 0; 682 683 /* 684 * Data alignment is different on some architectures. If control 685 * message is expected, we must arrange for the control message 686 * to be initially put elsewhere, and copy to target place 687 * with Linux alignment. 688 */ 689 if (LINUX_CMSG_ALIGNDIFF) { 690 error = copyin(SCARG(uap, msg), &msg, sizeof(msg)); 691 if (error) 692 return (error); 693 694 if (CMSG_FIRSTHDR(&msg)) { 695 caddr_t sg; 696 697 /* Need to fit within stackgap */ 698 if (msg.msg_controllen > STACKGAPLEN/2) { 699 /* Sorry guys! */ 700 return (EINVAL); 701 } 702 703 sg = stackgap_init(l->l_proc, STACKGAPLEN/3); 704 705 ocontrol = msg.msg_control; 706 ocontrollen = msg.msg_controllen; 707 708 /* space for at least one message's worth align */ 709 msg.msg_controllen += CMSG_ALIGN(1); 710 711 msg.msg_control = stackgap_alloc(l->l_proc, &sg, 712 msg.msg_controllen); 713 if (!msg.msg_control) 714 return (ENOMEM); 715 716 /* 717 * Okay to overwrite the original structure, it's 718 * supposed to be writable. 719 */ 720 error = copyout(&msg, SCARG(uap, msg), sizeof(msg)); 721 if (error) 722 return (error); 723 } 724 } 725 726 SCARG(&bsa, s) = SCARG(uap, s); 727 SCARG(&bsa, msg) = SCARG(uap, msg); 728 SCARG(&bsa, flags) = linux_to_bsd_msg_flags(SCARG(uap, flags)); 729 730 if (SCARG(&bsa, flags) < 0) { 731 /* Some unsupported flag */ 732 return (EINVAL); 733 } 734 735 if ((error = sys_recvmsg(l, &bsa, retval))) 736 goto done; 737 738 /* Fixup sockaddr */ 739 error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg, 740 sizeof(msg)); 741 if (error) 742 goto done; 743 744 if (msg.msg_name && msg.msg_namelen > 2) { 745 if ((error = linux_sa_put(msg.msg_name))) 746 goto done; 747 } 748 749 /* Fixup msg flags */ 750 lflags = bsd_to_linux_msg_flags(msg.msg_flags); 751 if (lflags < 0) { 752 /* Some flag unsupported by Linux */ 753 error = EINVAL; 754 goto done; 755 } 756 error = copyout(&lflags, (u_int8_t *) SCARG(uap, msg) + 757 offsetof(struct msghdr, msg_flags), sizeof(lflags)); 758 if (error) 759 goto done; 760 761 /* 762 * Fixup cmsg. We handle two things: 763 * 1. different values for level/type on some archs 764 * 2. different alignment of CMSG_DATA on some archs 765 */ 766 if (CMSG_FIRSTHDR(&msg)) { 767 struct cmsghdr cmsg, *cc; 768 int changed = 0; 769 size_t resid = ocontrollen; 770 771 cc = CMSG_FIRSTHDR(&msg); 772 do { 773 error = copyin(cc, &cmsg, sizeof(cmsg)); 774 if (error) 775 goto done; 776 777 switch (cmsg.cmsg_level) { 778 case SOL_SOCKET: 779 if (SOL_SOCKET != LINUX_SOL_SOCKET) { 780 cmsg.cmsg_level = LINUX_SOL_SOCKET; 781 changed = 1; 782 } 783 784 switch (cmsg.cmsg_type) { 785 case SCM_RIGHTS: 786 /* Linux SCM_RIGHTS is same as NetBSD */ 787 break; 788 789 default: 790 /* other types not supported */ 791 error = EINVAL; 792 goto done; 793 } 794 default: 795 /* pray and leave intact */ 796 break; 797 } 798 799 if (LINUX_CMSG_ALIGNDIFF) { 800 int i; 801 u_int8_t d, *sd, *td; 802 803 /* 804 * Sanity check. 805 */ 806 if (cmsg.cmsg_len > resid 807 || cmsg.cmsg_len < sizeof(cmsg)) { 808 error = EINVAL; 809 goto done; 810 } 811 812 /* 813 * Need to copy the cmsg from scratch area 814 * to the original place, converting data 815 * alignment from NetBSD to Linux one. 816 */ 817 error = copyout(&cmsg, ocontrol, sizeof(cmsg)); 818 if (error) 819 goto done; 820 /* zero pad */ 821 for(i=0; i < LINUX_CMSG_ALIGN(sizeof(cmsg)) - sizeof(cmsg); i++) { 822 copyout("",&ocontrol[sizeof(cmsg)+i],1); 823 } 824 825 sd = CMSG_DATA(cc); 826 td = LINUX_CMSG_DATA(ocontrol); 827 828 /* This is not particularily effective, but ..*/ 829 d = '\0'; 830 for(i=0; i < cmsg.cmsg_len - sizeof(cmsg); i++){ 831 copyin(sd++, &d, 1); 832 copyout(&d, td++, 1); 833 } 834 835 resid -= (td - ocontrol); 836 ocontrol = td; 837 } else if (changed) { 838 /* Update cmsghdr in-place */ 839 error = copyout(&cmsg, cc, sizeof(cmsg)); 840 if (error) 841 goto done; 842 changed = 0; 843 } 844 } while((cc = CMSG_NXTHDR(&msg, cc))); 845 } 846 847 done: 848 return (error); 849 } 850 851 /* 852 * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET 853 * is different, the rest matches IPPROTO_* on both systems. 854 */ 855 int 856 linux_to_bsd_sopt_level(llevel) 857 int llevel; 858 { 859 860 switch (llevel) { 861 case LINUX_SOL_SOCKET: 862 return SOL_SOCKET; 863 case LINUX_SOL_IP: 864 return IPPROTO_IP; 865 case LINUX_SOL_TCP: 866 return IPPROTO_TCP; 867 case LINUX_SOL_UDP: 868 return IPPROTO_UDP; 869 default: 870 return -1; 871 } 872 } 873 874 /* 875 * Convert Linux socket level socket option numbers to NetBSD values. 876 */ 877 int 878 linux_to_bsd_so_sockopt(lopt) 879 int lopt; 880 { 881 882 switch (lopt) { 883 case LINUX_SO_DEBUG: 884 return SO_DEBUG; 885 case LINUX_SO_REUSEADDR: 886 /* 887 * Linux does not implement SO_REUSEPORT, but allows reuse of a 888 * host:port pair through SO_REUSEADDR even if the address is not a 889 * multicast-address. Effectively, this means that we should use 890 * SO_REUSEPORT to allow Linux applications to not exit with 891 * EADDRINUSE 892 */ 893 return SO_REUSEPORT; 894 case LINUX_SO_TYPE: 895 return SO_TYPE; 896 case LINUX_SO_ERROR: 897 return SO_ERROR; 898 case LINUX_SO_DONTROUTE: 899 return SO_DONTROUTE; 900 case LINUX_SO_BROADCAST: 901 return SO_BROADCAST; 902 case LINUX_SO_SNDBUF: 903 return SO_SNDBUF; 904 case LINUX_SO_RCVBUF: 905 return SO_RCVBUF; 906 case LINUX_SO_KEEPALIVE: 907 return SO_KEEPALIVE; 908 case LINUX_SO_OOBINLINE: 909 return SO_OOBINLINE; 910 case LINUX_SO_LINGER: 911 return SO_LINGER; 912 case LINUX_SO_PRIORITY: 913 case LINUX_SO_NO_CHECK: 914 default: 915 return -1; 916 } 917 } 918 919 /* 920 * Convert Linux IP level socket option number to NetBSD values. 921 */ 922 int 923 linux_to_bsd_ip_sockopt(lopt) 924 int lopt; 925 { 926 927 switch (lopt) { 928 case LINUX_IP_TOS: 929 return IP_TOS; 930 case LINUX_IP_TTL: 931 return IP_TTL; 932 case LINUX_IP_MULTICAST_TTL: 933 return IP_MULTICAST_TTL; 934 case LINUX_IP_MULTICAST_LOOP: 935 return IP_MULTICAST_LOOP; 936 case LINUX_IP_MULTICAST_IF: 937 return IP_MULTICAST_IF; 938 case LINUX_IP_ADD_MEMBERSHIP: 939 return IP_ADD_MEMBERSHIP; 940 case LINUX_IP_DROP_MEMBERSHIP: 941 return IP_DROP_MEMBERSHIP; 942 default: 943 return -1; 944 } 945 } 946 947 /* 948 * Convert Linux TCP level socket option number to NetBSD values. 949 */ 950 int 951 linux_to_bsd_tcp_sockopt(lopt) 952 int lopt; 953 { 954 955 switch (lopt) { 956 case LINUX_TCP_NODELAY: 957 return TCP_NODELAY; 958 case LINUX_TCP_MAXSEG: 959 return TCP_MAXSEG; 960 default: 961 return -1; 962 } 963 } 964 965 /* 966 * Convert Linux UDP level socket option number to NetBSD values. 967 */ 968 int 969 linux_to_bsd_udp_sockopt(lopt) 970 int lopt; 971 { 972 973 switch (lopt) { 974 default: 975 return -1; 976 } 977 } 978 979 /* 980 * Another reasonably straightforward function: setsockopt(2). 981 * The level and option numbers are converted; the values passed 982 * are not (yet) converted, the ones currently implemented don't 983 * need conversion, as they are the same on both systems. 984 */ 985 int 986 linux_sys_setsockopt(l, v, retval) 987 struct lwp *l; 988 void *v; 989 register_t *retval; 990 { 991 struct linux_sys_setsockopt_args /* { 992 syscallarg(int) s; 993 syscallarg(int) level; 994 syscallarg(int) optname; 995 syscallarg(void *) optval; 996 syscallarg(int) optlen; 997 } */ *uap = v; 998 struct sys_setsockopt_args bsa; 999 int name; 1000 1001 SCARG(&bsa, s) = SCARG(uap, s); 1002 SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level)); 1003 SCARG(&bsa, val) = SCARG(uap, optval); 1004 SCARG(&bsa, valsize) = SCARG(uap, optlen); 1005 1006 switch (SCARG(&bsa, level)) { 1007 case SOL_SOCKET: 1008 name = linux_to_bsd_so_sockopt(SCARG(uap, optname)); 1009 break; 1010 case IPPROTO_IP: 1011 name = linux_to_bsd_ip_sockopt(SCARG(uap, optname)); 1012 break; 1013 case IPPROTO_TCP: 1014 name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname)); 1015 break; 1016 case IPPROTO_UDP: 1017 name = linux_to_bsd_udp_sockopt(SCARG(uap, optname)); 1018 break; 1019 default: 1020 return EINVAL; 1021 } 1022 1023 if (name == -1) 1024 return EINVAL; 1025 SCARG(&bsa, name) = name; 1026 1027 return sys_setsockopt(l, &bsa, retval); 1028 } 1029 1030 /* 1031 * getsockopt(2) is very much the same as setsockopt(2) (see above) 1032 */ 1033 int 1034 linux_sys_getsockopt(l, v, retval) 1035 struct lwp *l; 1036 void *v; 1037 register_t *retval; 1038 { 1039 struct linux_sys_getsockopt_args /* { 1040 syscallarg(int) s; 1041 syscallarg(int) level; 1042 syscallarg(int) optname; 1043 syscallarg(void *) optval; 1044 syscallarg(int *) optlen; 1045 } */ *uap = v; 1046 struct sys_getsockopt_args bga; 1047 int name; 1048 1049 SCARG(&bga, s) = SCARG(uap, s); 1050 SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level)); 1051 SCARG(&bga, val) = SCARG(uap, optval); 1052 SCARG(&bga, avalsize) = SCARG(uap, optlen); 1053 1054 switch (SCARG(&bga, level)) { 1055 case SOL_SOCKET: 1056 name = linux_to_bsd_so_sockopt(SCARG(uap, optname)); 1057 break; 1058 case IPPROTO_IP: 1059 name = linux_to_bsd_ip_sockopt(SCARG(uap, optname)); 1060 break; 1061 case IPPROTO_TCP: 1062 name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname)); 1063 break; 1064 case IPPROTO_UDP: 1065 name = linux_to_bsd_udp_sockopt(SCARG(uap, optname)); 1066 break; 1067 default: 1068 return EINVAL; 1069 } 1070 1071 if (name == -1) 1072 return EINVAL; 1073 SCARG(&bga, name) = name; 1074 1075 return sys_getsockopt(l, &bga, retval); 1076 } 1077 1078 #define IF_NAME_LEN 16 1079 1080 int 1081 linux_getifhwaddr(l, retval, fd, data) 1082 struct lwp *l; 1083 register_t *retval; 1084 u_int fd; 1085 void *data; 1086 { 1087 /* Not the full structure, just enough to map what we do here */ 1088 struct linux_ifreq { 1089 char if_name[IF_NAME_LEN]; 1090 struct osockaddr hwaddr; 1091 } lreq; 1092 struct proc *p = l->l_proc; 1093 struct filedesc *fdp; 1094 struct file *fp; 1095 struct ifaddr *ifa; 1096 struct ifnet *ifp; 1097 struct sockaddr_dl *sadl; 1098 int error, found; 1099 int index, ifnum; 1100 1101 /* 1102 * We can't emulate this ioctl by calling sys_ioctl() to run 1103 * SIOCGIFCONF, because the user buffer is not of the right 1104 * type to take those results. We can't use kernel buffers to 1105 * receive the results, as the implementation of sys_ioctl() 1106 * and ifconf() [which implements SIOCGIFCONF] use 1107 * copyin()/copyout() which will fail on kernel addresses. 1108 * 1109 * So, we must duplicate code from sys_ioctl() and ifconf(). Ugh. 1110 */ 1111 1112 fdp = p->p_fd; 1113 if ((fp = fd_getfile(fdp, fd)) == NULL) 1114 return (EBADF); 1115 1116 FILE_USE(fp); 1117 if ((fp->f_flag & (FREAD | FWRITE)) == 0) { 1118 error = EBADF; 1119 goto out; 1120 } 1121 1122 error = copyin(data, (caddr_t)&lreq, sizeof(lreq)); 1123 if (error) 1124 goto out; 1125 lreq.if_name[IF_NAME_LEN-1] = '\0'; /* just in case */ 1126 1127 /* 1128 * Try real interface name first, then fake "ethX" 1129 */ 1130 for (ifp = ifnet.tqh_first, found = 0; 1131 ifp != 0 && !found; 1132 ifp = ifp->if_list.tqe_next) { 1133 if (strcmp(lreq.if_name, ifp->if_xname)) 1134 /* not this interface */ 1135 continue; 1136 found=1; 1137 if ((ifa = ifp->if_addrlist.tqh_first) != 0) { 1138 for (; ifa != 0; ifa = ifa->ifa_list.tqe_next) { 1139 sadl = (struct sockaddr_dl *)ifa->ifa_addr; 1140 /* only return ethernet addresses */ 1141 /* XXX what about FDDI, etc. ? */ 1142 if (sadl->sdl_family != AF_LINK || 1143 sadl->sdl_type != IFT_ETHER) 1144 continue; 1145 memcpy((caddr_t)&lreq.hwaddr.sa_data, 1146 LLADDR(sadl), 1147 MIN(sadl->sdl_alen, 1148 sizeof(lreq.hwaddr.sa_data))); 1149 lreq.hwaddr.sa_family = 1150 sadl->sdl_family; 1151 error = copyout((caddr_t)&lreq, data, 1152 sizeof(lreq)); 1153 goto out; 1154 } 1155 } else { 1156 error = ENODEV; 1157 goto out; 1158 } 1159 } 1160 1161 if (strncmp(lreq.if_name, "eth", 3) == 0) { 1162 for (ifnum = 0, index = 3; 1163 lreq.if_name[index] != '\0' && index < IF_NAME_LEN; 1164 index++) { 1165 ifnum *= 10; 1166 ifnum += lreq.if_name[index] - '0'; 1167 } 1168 1169 error = EINVAL; /* in case we don't find one */ 1170 for (ifp = ifnet.tqh_first, found = 0; 1171 ifp != 0 && !found; 1172 ifp = ifp->if_list.tqe_next) { 1173 memcpy(lreq.if_name, ifp->if_xname, 1174 MIN(IF_NAME_LEN, IFNAMSIZ)); 1175 if ((ifa = ifp->if_addrlist.tqh_first) == 0) 1176 /* no addresses on this interface */ 1177 continue; 1178 else 1179 for (; ifa != 0; ifa = ifa->ifa_list.tqe_next) { 1180 sadl = (struct sockaddr_dl *)ifa->ifa_addr; 1181 /* only return ethernet addresses */ 1182 /* XXX what about FDDI, etc. ? */ 1183 if (sadl->sdl_family != AF_LINK || 1184 sadl->sdl_type != IFT_ETHER) 1185 continue; 1186 if (ifnum--) 1187 /* not the reqested iface */ 1188 continue; 1189 memcpy((caddr_t)&lreq.hwaddr.sa_data, 1190 LLADDR(sadl), 1191 MIN(sadl->sdl_alen, 1192 sizeof(lreq.hwaddr.sa_data))); 1193 lreq.hwaddr.sa_family = 1194 sadl->sdl_family; 1195 error = copyout((caddr_t)&lreq, data, 1196 sizeof(lreq)); 1197 found = 1; 1198 break; 1199 } 1200 } 1201 } else { 1202 /* unknown interface, not even an "eth*" name */ 1203 error = ENODEV; 1204 } 1205 1206 out: 1207 FILE_UNUSE(fp, l); 1208 return error; 1209 } 1210 #undef IF_NAME_LEN 1211 1212 int 1213 linux_ioctl_socket(l, uap, retval) 1214 struct lwp *l; 1215 struct linux_sys_ioctl_args /* { 1216 syscallarg(int) fd; 1217 syscallarg(u_long) com; 1218 syscallarg(caddr_t) data; 1219 } */ *uap; 1220 register_t *retval; 1221 { 1222 struct proc *p = l->l_proc; 1223 u_long com; 1224 int error = 0, isdev = 0, dosys = 1; 1225 struct sys_ioctl_args ia; 1226 struct file *fp; 1227 struct filedesc *fdp; 1228 struct vnode *vp; 1229 int (*ioctlf)(struct file *, u_long, void *, struct lwp *); 1230 struct ioctl_pt pt; 1231 1232 fdp = p->p_fd; 1233 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1234 return (EBADF); 1235 1236 FILE_USE(fp); 1237 1238 if (fp->f_type == DTYPE_VNODE) { 1239 vp = (struct vnode *)fp->f_data; 1240 isdev = vp->v_type == VCHR; 1241 } 1242 1243 /* 1244 * Don't try to interpret socket ioctl calls that are done 1245 * on a device filedescriptor, just pass them through, to 1246 * emulate Linux behaviour. Use PTIOCLINUX so that the 1247 * device will only handle these if it's prepared to do 1248 * so, to avoid unexpected things from happening. 1249 */ 1250 if (isdev) { 1251 dosys = 0; 1252 ioctlf = fp->f_ops->fo_ioctl; 1253 pt.com = SCARG(uap, com); 1254 pt.data = SCARG(uap, data); 1255 error = ioctlf(fp, PTIOCLINUX, (caddr_t)&pt, l); 1256 /* 1257 * XXX hack: if the function returns EJUSTRETURN, 1258 * it has stuffed a sysctl return value in pt.data. 1259 */ 1260 if (error == EJUSTRETURN) { 1261 retval[0] = (register_t)pt.data; 1262 error = 0; 1263 } 1264 goto out; 1265 } 1266 1267 com = SCARG(uap, com); 1268 retval[0] = 0; 1269 1270 switch (com) { 1271 case LINUX_SIOCGIFCONF: 1272 SCARG(&ia, com) = OSIOCGIFCONF; 1273 break; 1274 case LINUX_SIOCGIFFLAGS: 1275 SCARG(&ia, com) = SIOCGIFFLAGS; 1276 break; 1277 case LINUX_SIOCSIFFLAGS: 1278 SCARG(&ia, com) = SIOCSIFFLAGS; 1279 break; 1280 case LINUX_SIOCGIFADDR: 1281 SCARG(&ia, com) = OSIOCGIFADDR; 1282 break; 1283 case LINUX_SIOCGIFDSTADDR: 1284 SCARG(&ia, com) = OSIOCGIFDSTADDR; 1285 break; 1286 case LINUX_SIOCGIFBRDADDR: 1287 SCARG(&ia, com) = OSIOCGIFBRDADDR; 1288 break; 1289 case LINUX_SIOCGIFNETMASK: 1290 SCARG(&ia, com) = OSIOCGIFNETMASK; 1291 break; 1292 case LINUX_SIOCADDMULTI: 1293 SCARG(&ia, com) = SIOCADDMULTI; 1294 break; 1295 case LINUX_SIOCDELMULTI: 1296 SCARG(&ia, com) = SIOCDELMULTI; 1297 break; 1298 case LINUX_SIOCGIFHWADDR: 1299 error = linux_getifhwaddr(l, retval, SCARG(uap, fd), 1300 SCARG(uap, data)); 1301 dosys = 0; 1302 break; 1303 default: 1304 error = EINVAL; 1305 } 1306 1307 out: 1308 FILE_UNUSE(fp, l); 1309 1310 if (error ==0 && dosys) { 1311 SCARG(&ia, fd) = SCARG(uap, fd); 1312 SCARG(&ia, data) = SCARG(uap, data); 1313 /* XXX NJWLWP */ 1314 error = sys_ioctl(curlwp, &ia, retval); 1315 } 1316 1317 return error; 1318 } 1319 1320 int 1321 linux_sys_connect(l, v, retval) 1322 struct lwp *l; 1323 void *v; 1324 register_t *retval; 1325 { 1326 struct linux_sys_connect_args /* { 1327 syscallarg(int) s; 1328 syscallarg(const struct sockaddr *) name; 1329 syscallarg(int) namelen; 1330 } */ *uap = v; 1331 struct proc *p = l->l_proc; 1332 int error; 1333 struct sockaddr *sa; 1334 struct sys_connect_args bca; 1335 caddr_t sg = stackgap_init(p, 0); 1336 int namlen; 1337 1338 namlen = SCARG(uap, namelen); 1339 error = linux_sa_get(l, SCARG(uap, s), &sg, &sa, 1340 SCARG(uap, name), &namlen); 1341 if (error) 1342 return (error); 1343 1344 SCARG(&bca, s) = SCARG(uap, s); 1345 SCARG(&bca, name) = sa; 1346 SCARG(&bca, namelen) = (unsigned int) namlen; 1347 1348 error = sys_connect(l, &bca, retval); 1349 1350 if (error == EISCONN) { 1351 struct file *fp; 1352 struct socket *so; 1353 int s, state, prflags; 1354 1355 /* getsock() will use the descriptor for us */ 1356 if (getsock(p->p_fd, SCARG(uap, s), &fp) != 0) 1357 return EISCONN; 1358 1359 s = splsoftnet(); 1360 so = (struct socket *)fp->f_data; 1361 state = so->so_state; 1362 prflags = so->so_proto->pr_flags; 1363 splx(s); 1364 FILE_UNUSE(fp, l); 1365 /* 1366 * We should only let this call succeed once per 1367 * non-blocking connect; however we don't have 1368 * a convenient place to keep that state.. 1369 */ 1370 if ((state & SS_NBIO) && (state & SS_ISCONNECTED) && 1371 (prflags & PR_CONNREQUIRED)) 1372 return 0; 1373 } 1374 1375 return (error); 1376 } 1377 1378 int 1379 linux_sys_bind(l, v, retval) 1380 struct lwp *l; 1381 void *v; 1382 register_t *retval; 1383 { 1384 struct linux_sys_bind_args /* { 1385 syscallarg(int) s; 1386 syscallarg(const struct osockaddr *) name; 1387 syscallarg(int) namelen; 1388 } */ *uap = v; 1389 struct proc *p = l->l_proc; 1390 int error, namlen; 1391 struct sys_bind_args bsa; 1392 1393 namlen = SCARG(uap, namelen); 1394 SCARG(&bsa, s) = SCARG(uap, s); 1395 if (SCARG(uap, name)) { 1396 struct sockaddr *sa; 1397 caddr_t sg = stackgap_init(p, 0); 1398 1399 error = linux_sa_get(l, SCARG(uap, s), &sg, &sa, 1400 SCARG(uap, name), &namlen); 1401 if (error) 1402 return (error); 1403 1404 SCARG(&bsa, name) = sa; 1405 } else 1406 SCARG(&bsa, name) = NULL; 1407 SCARG(&bsa, namelen) = namlen; 1408 1409 return (sys_bind(l, &bsa, retval)); 1410 } 1411 1412 int 1413 linux_sys_getsockname(l, v, retval) 1414 struct lwp *l; 1415 void *v; 1416 register_t *retval; 1417 { 1418 struct linux_sys_getsockname_args /* { 1419 syscallarg(int) fdes; 1420 syscallarg(caddr_t) asa; 1421 syscallarg(int *) alen; 1422 } */ *uap = v; 1423 int error; 1424 1425 if ((error = sys_getsockname(l, uap, retval)) != 0) 1426 return (error); 1427 1428 if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa)))) 1429 return (error); 1430 1431 return (0); 1432 } 1433 1434 int 1435 linux_sys_getpeername(l, v, retval) 1436 struct lwp *l; 1437 void *v; 1438 register_t *retval; 1439 { 1440 struct sys_getpeername_args /* { 1441 syscallarg(int) fdes; 1442 syscallarg(caddr_t) asa; 1443 syscallarg(int *) alen; 1444 } */ *uap = v; 1445 int error; 1446 1447 if ((error = sys_getpeername(l, uap, retval)) != 0) 1448 return (error); 1449 1450 if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa)))) 1451 return (error); 1452 1453 return (0); 1454 } 1455 1456 /* 1457 * Copy the osockaddr structure pointed to by osa to kernel, adjust 1458 * family and convert to sockaddr, allocate stackgap and put the 1459 * the converted structure there, address on stackgap returned in sap. 1460 */ 1461 static int 1462 linux_sa_get(l, s, sgp, sap, osa, osalen) 1463 struct lwp *l; 1464 int s; 1465 caddr_t *sgp; 1466 struct sockaddr **sap; 1467 const struct osockaddr *osa; 1468 int *osalen; 1469 { 1470 int error=0, bdom; 1471 struct sockaddr *sa, *usa; 1472 struct osockaddr *kosa = (struct osockaddr *) &sa; 1473 struct proc *p = l->l_proc; 1474 int alloclen; 1475 #ifdef INET6 1476 int oldv6size; 1477 struct sockaddr_in6 *sin6; 1478 #endif 1479 1480 if (*osalen < 2 || *osalen > UCHAR_MAX || !osa) { 1481 DPRINTF(("bad osa=%p osalen=%d\n", osa, *osalen)); 1482 return (EINVAL); 1483 } 1484 1485 alloclen = *osalen; 1486 #ifdef INET6 1487 oldv6size = 0; 1488 /* 1489 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it 1490 * if it's a v4-mapped address, so reserve the proper space 1491 * for it. 1492 */ 1493 if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) { 1494 alloclen = sizeof (struct sockaddr_in6); 1495 oldv6size = 1; 1496 } 1497 #endif 1498 1499 kosa = (struct osockaddr *) malloc(alloclen, M_TEMP, M_WAITOK); 1500 1501 if ((error = copyin(osa, (caddr_t) kosa, *osalen))) { 1502 DPRINTF(("error copying osa %d\n", error)); 1503 goto out; 1504 } 1505 1506 bdom = linux_to_bsd_domain(kosa->sa_family); 1507 if (bdom == -1) { 1508 DPRINTF(("bad linux family=%d\n", kosa->sa_family)); 1509 error = EINVAL; 1510 goto out; 1511 } 1512 1513 #ifdef INET6 1514 /* 1515 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6, 1516 * which lacks the scope id compared with RFC2553 one. If we detect 1517 * the situation, reject the address and write a message to system log. 1518 * 1519 * Still accept addresses for which the scope id is not used. 1520 */ 1521 if (oldv6size && bdom == AF_INET6) { 1522 sin6 = (struct sockaddr_in6 *)kosa; 1523 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) || 1524 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) && 1525 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) && 1526 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) && 1527 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) && 1528 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) { 1529 sin6->sin6_scope_id = 0; 1530 } else { 1531 int uid = p->p_cred && p->p_ucred ? 1532 p->p_ucred->cr_uid : -1; 1533 1534 log(LOG_DEBUG, 1535 "pid %d (%s), uid %d: obsolete pre-RFC2553 " 1536 "sockaddr_in6 rejected", 1537 p->p_pid, p->p_comm, uid); 1538 error = EINVAL; 1539 goto out; 1540 } 1541 } 1542 #endif 1543 1544 /* 1545 * If the family is unspecified, use address family of the 1546 * socket. This avoid triggering COMPAT_43 struct socket family check 1547 * in sockargs() on little-endian machines, and strict family checks 1548 * in netinet/in_pcb.c et.al. 1549 */ 1550 if (bdom == AF_UNSPEC) { 1551 struct file *fp; 1552 struct socket *so; 1553 1554 /* getsock() will use the descriptor for us */ 1555 if ((error = getsock(p->p_fd, s, &fp)) != 0) 1556 goto out; 1557 1558 so = (struct socket *)fp->f_data; 1559 bdom = so->so_proto->pr_domain->dom_family; 1560 1561 FILE_UNUSE(fp, l); 1562 1563 DPRINTF(("AF_UNSPEC family adjusted to %d\n", bdom)); 1564 } 1565 1566 if (bdom == AF_INET) { 1567 alloclen = sizeof(struct sockaddr_in); 1568 } 1569 1570 sa = (struct sockaddr *) kosa; 1571 sa->sa_family = bdom; 1572 sa->sa_len = alloclen; 1573 #ifdef DEBUG_LINUX 1574 DPRINTF(("family %d, len = %d [ ", sa->sa_family, sa->sa_len)); 1575 for (bdom = 0; bdom < sizeof(sa->sa_data); bdom++) 1576 DPRINTF(("%02x ", (unsigned char) sa->sa_data[bdom])); 1577 DPRINTF(("\n")); 1578 #endif 1579 1580 usa = (struct sockaddr *) stackgap_alloc(p, sgp, alloclen); 1581 if (!usa) { 1582 error = ENOMEM; 1583 goto out; 1584 } 1585 1586 if ((error = copyout(sa, usa, alloclen))) { 1587 DPRINTF(("error copying out socket %d\n", error)); 1588 goto out; 1589 } 1590 1591 *sap = usa; 1592 1593 out: 1594 *osalen = alloclen; 1595 free(kosa, M_TEMP); 1596 return (error); 1597 } 1598 1599 static int 1600 linux_sa_put(osa) 1601 struct osockaddr *osa; 1602 { 1603 struct sockaddr sa; 1604 struct osockaddr *kosa; 1605 int error, bdom, len; 1606 1607 /* 1608 * Only read/write the sockaddr family and length part, the rest is 1609 * not changed. 1610 */ 1611 len = sizeof(sa.sa_len) + sizeof(sa.sa_family); 1612 1613 error = copyin((caddr_t) osa, (caddr_t) &sa, len); 1614 if (error) 1615 return (error); 1616 1617 bdom = bsd_to_linux_domain(sa.sa_family); 1618 if (bdom == -1) 1619 return (EINVAL); 1620 1621 /* Note: we convert from sockaddr to osockaddr here, too */ 1622 kosa = (struct osockaddr *) &sa; 1623 kosa->sa_family = bdom; 1624 error = copyout(kosa, osa, len); 1625 if (error) 1626 return (error); 1627 1628 return (0); 1629 } 1630 1631 #ifndef __amd64__ 1632 int 1633 linux_sys_recv(l, v, retval) 1634 struct lwp *l; 1635 void *v; 1636 register_t *retval; 1637 { 1638 struct linux_sys_recv_args /* { 1639 syscallarg(int) s; 1640 syscallarg(void *) buf; 1641 syscallarg(int) len; 1642 syscallarg(int) flags; 1643 } */ *uap = v; 1644 struct sys_recvfrom_args bra; 1645 1646 1647 SCARG(&bra, s) = SCARG(uap, s); 1648 SCARG(&bra, buf) = SCARG(uap, buf); 1649 SCARG(&bra, len) = (size_t) SCARG(uap, len); 1650 SCARG(&bra, flags) = SCARG(uap, flags); 1651 SCARG(&bra, from) = NULL; 1652 SCARG(&bra, fromlenaddr) = NULL; 1653 1654 return (sys_recvfrom(l, &bra, retval)); 1655 } 1656 1657 int 1658 linux_sys_send(l, v, retval) 1659 struct lwp *l; 1660 void *v; 1661 register_t *retval; 1662 { 1663 struct linux_sys_send_args /* { 1664 syscallarg(int) s; 1665 syscallarg(caddr_t) buf; 1666 syscallarg(int) len; 1667 syscallarg(int) flags; 1668 } */ *uap = v; 1669 struct sys_sendto_args bsa; 1670 1671 SCARG(&bsa, s) = SCARG(uap, s); 1672 SCARG(&bsa, buf) = SCARG(uap, buf); 1673 SCARG(&bsa, len) = SCARG(uap, len); 1674 SCARG(&bsa, flags) = SCARG(uap, flags); 1675 SCARG(&bsa, to) = NULL; 1676 SCARG(&bsa, tolen) = 0; 1677 1678 return (sys_sendto(l, &bsa, retval)); 1679 } 1680 #endif 1681 1682 int 1683 linux_sys_accept(l, v, retval) 1684 struct lwp *l; 1685 void *v; 1686 register_t *retval; 1687 { 1688 struct linux_sys_accept_args /* { 1689 syscallarg(int) s; 1690 syscallarg(struct osockaddr *) name; 1691 syscallarg(int *) anamelen; 1692 } */ *uap = v; 1693 int error; 1694 struct sys_accept_args baa; 1695 1696 SCARG(&baa, s) = SCARG(uap, s); 1697 SCARG(&baa, name) = (struct sockaddr *) SCARG(uap, name); 1698 SCARG(&baa, anamelen) = (unsigned int *) SCARG(uap, anamelen); 1699 1700 if ((error = sys_accept(l, &baa, retval))) 1701 return (error); 1702 1703 if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name)))) 1704 return (error); 1705 1706 return (0); 1707 } 1708