1 /*- 2 * Copyright (c) 1995 S�ren Schmidt 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer 10 * in this position and unchanged. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. The name of the author may not be used to endorse or promote products 15 * derived from this software without specific prior written permission 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 /* XXX we use functions that might not exist. */ 33 #include "opt_compat.h" 34 #include "opt_inet6.h" 35 36 #include <sys/param.h> 37 #include <sys/proc.h> 38 #include <sys/systm.h> 39 #include <sys/sysproto.h> 40 #include <sys/fcntl.h> 41 #include <sys/file.h> 42 #include <sys/limits.h> 43 #include <sys/lock.h> 44 #include <sys/malloc.h> 45 #include <sys/mutex.h> 46 #include <sys/mbuf.h> 47 #include <sys/socket.h> 48 #include <sys/socketvar.h> 49 #include <sys/syscallsubr.h> 50 #include <sys/uio.h> 51 #include <sys/syslog.h> 52 #include <sys/un.h> 53 #include <sys/vimage.h> 54 55 #include <net/if.h> 56 #include <netinet/in.h> 57 #include <netinet/in_systm.h> 58 #include <netinet/ip.h> 59 #ifdef INET6 60 #include <netinet/ip6.h> 61 #include <netinet6/ip6_var.h> 62 #include <netinet6/in6_var.h> 63 #include <netinet6/vinet6.h> 64 #endif 65 66 #ifdef COMPAT_LINUX32 67 #include <machine/../linux32/linux.h> 68 #include <machine/../linux32/linux32_proto.h> 69 #else 70 #include <machine/../linux/linux.h> 71 #include <machine/../linux/linux_proto.h> 72 #endif 73 #include <compat/linux/linux_socket.h> 74 #include <compat/linux/linux_util.h> 75 76 static int do_sa_get(struct sockaddr **, const struct osockaddr *, int *, 77 struct malloc_type *); 78 static int linux_to_bsd_domain(int); 79 80 /* 81 * Reads a linux sockaddr and does any necessary translation. 82 * Linux sockaddrs don't have a length field, only a family. 83 */ 84 static int 85 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len) 86 { 87 int osalen = len; 88 89 return (do_sa_get(sap, osa, &osalen, M_SONAME)); 90 } 91 92 /* 93 * Copy the osockaddr structure pointed to by osa to kernel, adjust 94 * family and convert to sockaddr. 95 */ 96 static int 97 do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen, 98 struct malloc_type *mtype) 99 { 100 int error=0, bdom; 101 struct sockaddr *sa; 102 struct osockaddr *kosa; 103 int alloclen; 104 #ifdef INET6 105 int oldv6size; 106 struct sockaddr_in6 *sin6; 107 #endif 108 109 if (*osalen < 2 || *osalen > UCHAR_MAX || !osa) 110 return (EINVAL); 111 112 alloclen = *osalen; 113 #ifdef INET6 114 oldv6size = 0; 115 /* 116 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it 117 * if it's a v4-mapped address, so reserve the proper space 118 * for it. 119 */ 120 if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) { 121 alloclen = sizeof (struct sockaddr_in6); 122 oldv6size = 1; 123 } 124 #endif 125 126 kosa = malloc(alloclen, mtype, M_WAITOK); 127 128 if ((error = copyin(osa, kosa, *osalen))) 129 goto out; 130 131 bdom = linux_to_bsd_domain(kosa->sa_family); 132 if (bdom == -1) { 133 error = EINVAL; 134 goto out; 135 } 136 137 #ifdef INET6 138 /* 139 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6, 140 * which lacks the scope id compared with RFC2553 one. If we detect 141 * the situation, reject the address and write a message to system log. 142 * 143 * Still accept addresses for which the scope id is not used. 144 */ 145 if (oldv6size && bdom == AF_INET6) { 146 sin6 = (struct sockaddr_in6 *)kosa; 147 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) || 148 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) && 149 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) && 150 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) && 151 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) && 152 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) { 153 sin6->sin6_scope_id = 0; 154 } else { 155 log(LOG_DEBUG, 156 "obsolete pre-RFC2553 sockaddr_in6 rejected\n"); 157 error = EINVAL; 158 goto out; 159 } 160 } else 161 #endif 162 if (bdom == AF_INET) 163 alloclen = sizeof(struct sockaddr_in); 164 165 sa = (struct sockaddr *) kosa; 166 sa->sa_family = bdom; 167 sa->sa_len = alloclen; 168 169 *sap = sa; 170 *osalen = alloclen; 171 return (0); 172 173 out: 174 free(kosa, mtype); 175 return (error); 176 } 177 178 static int 179 linux_to_bsd_domain(int domain) 180 { 181 182 switch (domain) { 183 case LINUX_AF_UNSPEC: 184 return (AF_UNSPEC); 185 case LINUX_AF_UNIX: 186 return (AF_LOCAL); 187 case LINUX_AF_INET: 188 return (AF_INET); 189 case LINUX_AF_INET6: 190 return (AF_INET6); 191 case LINUX_AF_AX25: 192 return (AF_CCITT); 193 case LINUX_AF_IPX: 194 return (AF_IPX); 195 case LINUX_AF_APPLETALK: 196 return (AF_APPLETALK); 197 } 198 return (-1); 199 } 200 201 static int 202 bsd_to_linux_domain(int domain) 203 { 204 205 switch (domain) { 206 case AF_UNSPEC: 207 return (LINUX_AF_UNSPEC); 208 case AF_LOCAL: 209 return (LINUX_AF_UNIX); 210 case AF_INET: 211 return (LINUX_AF_INET); 212 case AF_INET6: 213 return (LINUX_AF_INET6); 214 case AF_CCITT: 215 return (LINUX_AF_AX25); 216 case AF_IPX: 217 return (LINUX_AF_IPX); 218 case AF_APPLETALK: 219 return (LINUX_AF_APPLETALK); 220 } 221 return (-1); 222 } 223 224 static int 225 linux_to_bsd_sockopt_level(int level) 226 { 227 228 switch (level) { 229 case LINUX_SOL_SOCKET: 230 return (SOL_SOCKET); 231 } 232 return (level); 233 } 234 235 static int 236 bsd_to_linux_sockopt_level(int level) 237 { 238 239 switch (level) { 240 case SOL_SOCKET: 241 return (LINUX_SOL_SOCKET); 242 } 243 return (level); 244 } 245 246 static int 247 linux_to_bsd_ip_sockopt(int opt) 248 { 249 250 switch (opt) { 251 case LINUX_IP_TOS: 252 return (IP_TOS); 253 case LINUX_IP_TTL: 254 return (IP_TTL); 255 case LINUX_IP_OPTIONS: 256 return (IP_OPTIONS); 257 case LINUX_IP_MULTICAST_IF: 258 return (IP_MULTICAST_IF); 259 case LINUX_IP_MULTICAST_TTL: 260 return (IP_MULTICAST_TTL); 261 case LINUX_IP_MULTICAST_LOOP: 262 return (IP_MULTICAST_LOOP); 263 case LINUX_IP_ADD_MEMBERSHIP: 264 return (IP_ADD_MEMBERSHIP); 265 case LINUX_IP_DROP_MEMBERSHIP: 266 return (IP_DROP_MEMBERSHIP); 267 case LINUX_IP_HDRINCL: 268 return (IP_HDRINCL); 269 } 270 return (-1); 271 } 272 273 static int 274 linux_to_bsd_so_sockopt(int opt) 275 { 276 277 switch (opt) { 278 case LINUX_SO_DEBUG: 279 return (SO_DEBUG); 280 case LINUX_SO_REUSEADDR: 281 return (SO_REUSEADDR); 282 case LINUX_SO_TYPE: 283 return (SO_TYPE); 284 case LINUX_SO_ERROR: 285 return (SO_ERROR); 286 case LINUX_SO_DONTROUTE: 287 return (SO_DONTROUTE); 288 case LINUX_SO_BROADCAST: 289 return (SO_BROADCAST); 290 case LINUX_SO_SNDBUF: 291 return (SO_SNDBUF); 292 case LINUX_SO_RCVBUF: 293 return (SO_RCVBUF); 294 case LINUX_SO_KEEPALIVE: 295 return (SO_KEEPALIVE); 296 case LINUX_SO_OOBINLINE: 297 return (SO_OOBINLINE); 298 case LINUX_SO_LINGER: 299 return (SO_LINGER); 300 case LINUX_SO_PEERCRED: 301 return (LOCAL_PEERCRED); 302 case LINUX_SO_RCVLOWAT: 303 return (SO_RCVLOWAT); 304 case LINUX_SO_SNDLOWAT: 305 return (SO_SNDLOWAT); 306 case LINUX_SO_RCVTIMEO: 307 return (SO_RCVTIMEO); 308 case LINUX_SO_SNDTIMEO: 309 return (SO_SNDTIMEO); 310 case LINUX_SO_TIMESTAMP: 311 return (SO_TIMESTAMP); 312 case LINUX_SO_ACCEPTCONN: 313 return (SO_ACCEPTCONN); 314 } 315 return (-1); 316 } 317 318 static int 319 linux_to_bsd_msg_flags(int flags) 320 { 321 int ret_flags = 0; 322 323 if (flags & LINUX_MSG_OOB) 324 ret_flags |= MSG_OOB; 325 if (flags & LINUX_MSG_PEEK) 326 ret_flags |= MSG_PEEK; 327 if (flags & LINUX_MSG_DONTROUTE) 328 ret_flags |= MSG_DONTROUTE; 329 if (flags & LINUX_MSG_CTRUNC) 330 ret_flags |= MSG_CTRUNC; 331 if (flags & LINUX_MSG_TRUNC) 332 ret_flags |= MSG_TRUNC; 333 if (flags & LINUX_MSG_DONTWAIT) 334 ret_flags |= MSG_DONTWAIT; 335 if (flags & LINUX_MSG_EOR) 336 ret_flags |= MSG_EOR; 337 if (flags & LINUX_MSG_WAITALL) 338 ret_flags |= MSG_WAITALL; 339 if (flags & LINUX_MSG_NOSIGNAL) 340 ret_flags |= MSG_NOSIGNAL; 341 #if 0 /* not handled */ 342 if (flags & LINUX_MSG_PROXY) 343 ; 344 if (flags & LINUX_MSG_FIN) 345 ; 346 if (flags & LINUX_MSG_SYN) 347 ; 348 if (flags & LINUX_MSG_CONFIRM) 349 ; 350 if (flags & LINUX_MSG_RST) 351 ; 352 if (flags & LINUX_MSG_ERRQUEUE) 353 ; 354 #endif 355 return ret_flags; 356 } 357 358 /* 359 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the 360 * native syscall will fault. Thus, we don't really need to check the 361 * return values for these functions. 362 */ 363 364 static int 365 bsd_to_linux_sockaddr(struct sockaddr *arg) 366 { 367 struct sockaddr sa; 368 size_t sa_len = sizeof(struct sockaddr); 369 int error; 370 371 if ((error = copyin(arg, &sa, sa_len))) 372 return (error); 373 374 *(u_short *)&sa = sa.sa_family; 375 376 error = copyout(&sa, arg, sa_len); 377 378 return (error); 379 } 380 381 static int 382 linux_to_bsd_sockaddr(struct sockaddr *arg, int len) 383 { 384 struct sockaddr sa; 385 size_t sa_len = sizeof(struct sockaddr); 386 int error; 387 388 if ((error = copyin(arg, &sa, sa_len))) 389 return (error); 390 391 sa.sa_family = *(sa_family_t *)&sa; 392 sa.sa_len = len; 393 394 error = copyout(&sa, arg, sa_len); 395 396 return (error); 397 } 398 399 400 static int 401 linux_sa_put(struct osockaddr *osa) 402 { 403 struct osockaddr sa; 404 int error, bdom; 405 406 /* 407 * Only read/write the osockaddr family part, the rest is 408 * not changed. 409 */ 410 error = copyin(osa, &sa, sizeof(sa.sa_family)); 411 if (error) 412 return (error); 413 414 bdom = bsd_to_linux_domain(sa.sa_family); 415 if (bdom == -1) 416 return (EINVAL); 417 418 sa.sa_family = bdom; 419 error = copyout(&sa, osa, sizeof(sa.sa_family)); 420 if (error) 421 return (error); 422 423 return (0); 424 } 425 426 static int 427 linux_to_bsd_cmsg_type(int cmsg_type) 428 { 429 430 switch (cmsg_type) { 431 case LINUX_SCM_RIGHTS: 432 return (SCM_RIGHTS); 433 } 434 return (-1); 435 } 436 437 static int 438 bsd_to_linux_cmsg_type(int cmsg_type) 439 { 440 441 switch (cmsg_type) { 442 case SCM_RIGHTS: 443 return (LINUX_SCM_RIGHTS); 444 } 445 return (-1); 446 } 447 448 449 450 static int 451 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 452 { 453 if (lhdr->msg_controllen > INT_MAX) 454 return (ENOBUFS); 455 456 bhdr->msg_name = PTRIN(lhdr->msg_name); 457 bhdr->msg_namelen = lhdr->msg_namelen; 458 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 459 bhdr->msg_iovlen = lhdr->msg_iovlen; 460 bhdr->msg_control = PTRIN(lhdr->msg_control); 461 bhdr->msg_controllen = lhdr->msg_controllen; 462 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 463 return (0); 464 } 465 466 static int 467 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 468 { 469 lhdr->msg_name = PTROUT(bhdr->msg_name); 470 lhdr->msg_namelen = bhdr->msg_namelen; 471 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 472 lhdr->msg_iovlen = bhdr->msg_iovlen; 473 lhdr->msg_control = PTROUT(bhdr->msg_control); 474 lhdr->msg_controllen = bhdr->msg_controllen; 475 /* msg_flags skipped */ 476 return (0); 477 } 478 479 static int 480 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 481 struct mbuf *control, enum uio_seg segflg) 482 { 483 struct sockaddr *to; 484 int error; 485 486 if (mp->msg_name != NULL) { 487 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen); 488 if (error) 489 return (error); 490 mp->msg_name = to; 491 } else 492 to = NULL; 493 494 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 495 segflg); 496 497 if (to) 498 free(to, M_SONAME); 499 return (error); 500 } 501 502 /* Return 0 if IP_HDRINCL is set for the given socket. */ 503 static int 504 linux_check_hdrincl(struct thread *td, int s) 505 { 506 int error, optval, size_val; 507 508 size_val = sizeof(optval); 509 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 510 &optval, UIO_SYSSPACE, &size_val); 511 if (error) 512 return (error); 513 514 return (optval == 0); 515 } 516 517 struct linux_sendto_args { 518 int s; 519 l_uintptr_t msg; 520 int len; 521 int flags; 522 l_uintptr_t to; 523 int tolen; 524 }; 525 526 /* 527 * Updated sendto() when IP_HDRINCL is set: 528 * tweak endian-dependent fields in the IP packet. 529 */ 530 static int 531 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 532 { 533 /* 534 * linux_ip_copysize defines how many bytes we should copy 535 * from the beginning of the IP packet before we customize it for BSD. 536 * It should include all the fields we modify (ip_len and ip_off). 537 */ 538 #define linux_ip_copysize 8 539 540 struct ip *packet; 541 struct msghdr msg; 542 struct iovec aiov[1]; 543 int error; 544 545 /* Check that the packet isn't too big or too small. */ 546 if (linux_args->len < linux_ip_copysize || 547 linux_args->len > IP_MAXPACKET) 548 return (EINVAL); 549 550 packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK); 551 552 /* Make kernel copy of the packet to be sent */ 553 if ((error = copyin(PTRIN(linux_args->msg), packet, 554 linux_args->len))) 555 goto goout; 556 557 /* Convert fields from Linux to BSD raw IP socket format */ 558 packet->ip_len = linux_args->len; 559 packet->ip_off = ntohs(packet->ip_off); 560 561 /* Prepare the msghdr and iovec structures describing the new packet */ 562 msg.msg_name = PTRIN(linux_args->to); 563 msg.msg_namelen = linux_args->tolen; 564 msg.msg_iov = aiov; 565 msg.msg_iovlen = 1; 566 msg.msg_control = NULL; 567 msg.msg_flags = 0; 568 aiov[0].iov_base = (char *)packet; 569 aiov[0].iov_len = linux_args->len; 570 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 571 NULL, UIO_SYSSPACE); 572 goout: 573 free(packet, M_TEMP); 574 return (error); 575 } 576 577 struct linux_socket_args { 578 int domain; 579 int type; 580 int protocol; 581 }; 582 583 static int 584 linux_socket(struct thread *td, struct linux_socket_args *args) 585 { 586 #ifdef INET6 587 #ifndef KLD_MODULE 588 INIT_VNET_INET6(curvnet); 589 #endif 590 #endif 591 struct socket_args /* { 592 int domain; 593 int type; 594 int protocol; 595 } */ bsd_args; 596 int retval_socket; 597 598 bsd_args.protocol = args->protocol; 599 bsd_args.type = args->type; 600 bsd_args.domain = linux_to_bsd_domain(args->domain); 601 if (bsd_args.domain == -1) 602 return (EAFNOSUPPORT); 603 604 retval_socket = socket(td, &bsd_args); 605 if (bsd_args.type == SOCK_RAW 606 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0) 607 && bsd_args.domain == AF_INET 608 && retval_socket >= 0) { 609 /* It's a raw IP socket: set the IP_HDRINCL option. */ 610 int hdrincl; 611 612 hdrincl = 1; 613 /* We ignore any error returned by kern_setsockopt() */ 614 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 615 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 616 } 617 #ifdef INET6 618 /* 619 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by 620 * default and some apps depend on this. So, set V6ONLY to 0 621 * for Linux apps if the sysctl value is set to 1. 622 */ 623 if (bsd_args.domain == PF_INET6 && retval_socket >= 0 624 #ifndef KLD_MODULE 625 /* 626 * XXX: Avoid undefined symbol error with an IPv4 only 627 * kernel. 628 */ 629 && V_ip6_v6only 630 #endif 631 ) { 632 int v6only; 633 634 v6only = 0; 635 /* We ignore any error returned by setsockopt() */ 636 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 637 &v6only, UIO_SYSSPACE, sizeof(v6only)); 638 } 639 #endif 640 641 return (retval_socket); 642 } 643 644 struct linux_bind_args { 645 int s; 646 l_uintptr_t name; 647 int namelen; 648 }; 649 650 static int 651 linux_bind(struct thread *td, struct linux_bind_args *args) 652 { 653 struct sockaddr *sa; 654 int error; 655 656 error = linux_getsockaddr(&sa, PTRIN(args->name), 657 args->namelen); 658 if (error) 659 return (error); 660 661 error = kern_bind(td, args->s, sa); 662 free(sa, M_SONAME); 663 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 664 return (EINVAL); 665 return (error); 666 } 667 668 struct linux_connect_args { 669 int s; 670 l_uintptr_t name; 671 int namelen; 672 }; 673 int linux_connect(struct thread *, struct linux_connect_args *); 674 675 int 676 linux_connect(struct thread *td, struct linux_connect_args *args) 677 { 678 struct socket *so; 679 struct sockaddr *sa; 680 u_int fflag; 681 int error; 682 683 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 684 args->namelen); 685 if (error) 686 return (error); 687 688 error = kern_connect(td, args->s, sa); 689 free(sa, M_SONAME); 690 if (error != EISCONN) 691 return (error); 692 693 /* 694 * Linux doesn't return EISCONN the first time it occurs, 695 * when on a non-blocking socket. Instead it returns the 696 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 697 * 698 * XXXRW: Instead of using fgetsock(), check that it is a 699 * socket and use the file descriptor reference instead of 700 * creating a new one. 701 */ 702 error = fgetsock(td, args->s, &so, &fflag); 703 if (error == 0) { 704 error = EISCONN; 705 if (fflag & FNONBLOCK) { 706 SOCK_LOCK(so); 707 if (so->so_emuldata == 0) 708 error = so->so_error; 709 so->so_emuldata = (void *)1; 710 SOCK_UNLOCK(so); 711 } 712 fputsock(so); 713 } 714 return (error); 715 } 716 717 struct linux_listen_args { 718 int s; 719 int backlog; 720 }; 721 722 static int 723 linux_listen(struct thread *td, struct linux_listen_args *args) 724 { 725 struct listen_args /* { 726 int s; 727 int backlog; 728 } */ bsd_args; 729 730 bsd_args.s = args->s; 731 bsd_args.backlog = args->backlog; 732 return (listen(td, &bsd_args)); 733 } 734 735 struct linux_accept_args { 736 int s; 737 l_uintptr_t addr; 738 l_uintptr_t namelen; 739 }; 740 741 static int 742 linux_accept(struct thread *td, struct linux_accept_args *args) 743 { 744 struct accept_args /* { 745 int s; 746 struct sockaddr * __restrict name; 747 socklen_t * __restrict anamelen; 748 } */ bsd_args; 749 int error, fd; 750 751 bsd_args.s = args->s; 752 /* XXX: */ 753 bsd_args.name = (struct sockaddr * __restrict)PTRIN(args->addr); 754 bsd_args.anamelen = PTRIN(args->namelen);/* XXX */ 755 error = accept(td, &bsd_args); 756 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 757 if (error) { 758 if (error == EFAULT && args->namelen != sizeof(struct sockaddr_in)) 759 return (EINVAL); 760 return (error); 761 } 762 if (args->addr) { 763 error = linux_sa_put(PTRIN(args->addr)); 764 if (error) { 765 (void)kern_close(td, td->td_retval[0]); 766 return (error); 767 } 768 } 769 770 /* 771 * linux appears not to copy flags from the parent socket to the 772 * accepted one, so we must clear the flags in the new descriptor. 773 * Ignore any errors, because we already have an open fd. 774 */ 775 fd = td->td_retval[0]; 776 (void)kern_fcntl(td, fd, F_SETFL, 0); 777 td->td_retval[0] = fd; 778 return (0); 779 } 780 781 struct linux_getsockname_args { 782 int s; 783 l_uintptr_t addr; 784 l_uintptr_t namelen; 785 }; 786 787 static int 788 linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 789 { 790 struct getsockname_args /* { 791 int fdes; 792 struct sockaddr * __restrict asa; 793 socklen_t * __restrict alen; 794 } */ bsd_args; 795 int error; 796 797 bsd_args.fdes = args->s; 798 /* XXX: */ 799 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 800 bsd_args.alen = PTRIN(args->namelen); /* XXX */ 801 error = getsockname(td, &bsd_args); 802 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 803 if (error) 804 return (error); 805 error = linux_sa_put(PTRIN(args->addr)); 806 if (error) 807 return (error); 808 return (0); 809 } 810 811 struct linux_getpeername_args { 812 int s; 813 l_uintptr_t addr; 814 l_uintptr_t namelen; 815 }; 816 817 static int 818 linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 819 { 820 struct getpeername_args /* { 821 int fdes; 822 caddr_t asa; 823 int *alen; 824 } */ bsd_args; 825 int error; 826 827 bsd_args.fdes = args->s; 828 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 829 bsd_args.alen = (int *)PTRIN(args->namelen); 830 error = getpeername(td, &bsd_args); 831 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 832 if (error) 833 return (error); 834 error = linux_sa_put(PTRIN(args->addr)); 835 if (error) 836 return (error); 837 return (0); 838 } 839 840 struct linux_socketpair_args { 841 int domain; 842 int type; 843 int protocol; 844 l_uintptr_t rsv; 845 }; 846 847 static int 848 linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 849 { 850 struct socketpair_args /* { 851 int domain; 852 int type; 853 int protocol; 854 int *rsv; 855 } */ bsd_args; 856 857 bsd_args.domain = linux_to_bsd_domain(args->domain); 858 if (bsd_args.domain != PF_LOCAL) 859 return (EAFNOSUPPORT); 860 861 bsd_args.type = args->type; 862 if (args->protocol != 0 && args->protocol != PF_UNIX) 863 864 /* 865 * Use of PF_UNIX as protocol argument is not right, 866 * but Linux does it. 867 * Do not map PF_UNIX as its Linux value is identical 868 * to FreeBSD one. 869 */ 870 return (EPROTONOSUPPORT); 871 else 872 bsd_args.protocol = 0; 873 bsd_args.rsv = (int *)PTRIN(args->rsv); 874 return (socketpair(td, &bsd_args)); 875 } 876 877 struct linux_send_args { 878 int s; 879 l_uintptr_t msg; 880 int len; 881 int flags; 882 }; 883 884 static int 885 linux_send(struct thread *td, struct linux_send_args *args) 886 { 887 struct sendto_args /* { 888 int s; 889 caddr_t buf; 890 int len; 891 int flags; 892 caddr_t to; 893 int tolen; 894 } */ bsd_args; 895 896 bsd_args.s = args->s; 897 bsd_args.buf = (caddr_t)PTRIN(args->msg); 898 bsd_args.len = args->len; 899 bsd_args.flags = args->flags; 900 bsd_args.to = NULL; 901 bsd_args.tolen = 0; 902 return sendto(td, &bsd_args); 903 } 904 905 struct linux_recv_args { 906 int s; 907 l_uintptr_t msg; 908 int len; 909 int flags; 910 }; 911 912 static int 913 linux_recv(struct thread *td, struct linux_recv_args *args) 914 { 915 struct recvfrom_args /* { 916 int s; 917 caddr_t buf; 918 int len; 919 int flags; 920 struct sockaddr *from; 921 socklen_t fromlenaddr; 922 } */ bsd_args; 923 924 bsd_args.s = args->s; 925 bsd_args.buf = (caddr_t)PTRIN(args->msg); 926 bsd_args.len = args->len; 927 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 928 bsd_args.from = NULL; 929 bsd_args.fromlenaddr = 0; 930 return (recvfrom(td, &bsd_args)); 931 } 932 933 static int 934 linux_sendto(struct thread *td, struct linux_sendto_args *args) 935 { 936 struct msghdr msg; 937 struct iovec aiov; 938 int error; 939 940 if (linux_check_hdrincl(td, args->s) == 0) 941 /* IP_HDRINCL set, tweak the packet before sending */ 942 return (linux_sendto_hdrincl(td, args)); 943 944 msg.msg_name = PTRIN(args->to); 945 msg.msg_namelen = args->tolen; 946 msg.msg_iov = &aiov; 947 msg.msg_iovlen = 1; 948 msg.msg_control = NULL; 949 msg.msg_flags = 0; 950 aiov.iov_base = PTRIN(args->msg); 951 aiov.iov_len = args->len; 952 error = linux_sendit(td, args->s, &msg, args->flags, NULL, 953 UIO_USERSPACE); 954 return (error); 955 } 956 957 struct linux_recvfrom_args { 958 int s; 959 l_uintptr_t buf; 960 int len; 961 int flags; 962 l_uintptr_t from; 963 l_uintptr_t fromlen; 964 }; 965 966 static int 967 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 968 { 969 struct recvfrom_args /* { 970 int s; 971 caddr_t buf; 972 size_t len; 973 int flags; 974 struct sockaddr * __restrict from; 975 socklen_t * __restrict fromlenaddr; 976 } */ bsd_args; 977 size_t len; 978 int error; 979 980 if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t)))) 981 return (error); 982 983 bsd_args.s = args->s; 984 bsd_args.buf = PTRIN(args->buf); 985 bsd_args.len = args->len; 986 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 987 /* XXX: */ 988 bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from); 989 bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */ 990 991 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len); 992 error = recvfrom(td, &bsd_args); 993 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from); 994 995 if (error) 996 return (error); 997 if (args->from) { 998 error = linux_sa_put((struct osockaddr *) 999 PTRIN(args->from)); 1000 if (error) 1001 return (error); 1002 } 1003 return (0); 1004 } 1005 1006 struct linux_sendmsg_args { 1007 int s; 1008 l_uintptr_t msg; 1009 int flags; 1010 }; 1011 1012 static int 1013 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1014 { 1015 struct cmsghdr *cmsg; 1016 struct mbuf *control; 1017 struct msghdr msg; 1018 struct l_cmsghdr linux_cmsg; 1019 struct l_cmsghdr *ptr_cmsg; 1020 struct l_msghdr linux_msg; 1021 struct iovec *iov; 1022 socklen_t datalen; 1023 void *data; 1024 int error; 1025 1026 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg)); 1027 if (error) 1028 return (error); 1029 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1030 if (error) 1031 return (error); 1032 1033 /* 1034 * Some Linux applications (ping) define a non-NULL control data 1035 * pointer, but a msg_controllen of 0, which is not allowed in the 1036 * FreeBSD system call interface. NULL the msg_control pointer in 1037 * order to handle this case. This should be checked, but allows the 1038 * Linux ping to work. 1039 */ 1040 if (msg.msg_control != NULL && msg.msg_controllen == 0) 1041 msg.msg_control = NULL; 1042 1043 #ifdef COMPAT_LINUX32 1044 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1045 &iov, EMSGSIZE); 1046 #else 1047 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1048 #endif 1049 if (error) 1050 return (error); 1051 1052 if (msg.msg_control != NULL) { 1053 error = ENOBUFS; 1054 cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO); 1055 control = m_get(M_WAIT, MT_CONTROL); 1056 if (control == NULL) 1057 goto bad; 1058 ptr_cmsg = LINUX_CMSG_FIRSTHDR(&msg); 1059 1060 do { 1061 error = copyin(ptr_cmsg, &linux_cmsg, 1062 sizeof(struct l_cmsghdr)); 1063 if (error) 1064 goto bad; 1065 1066 error = EINVAL; 1067 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1068 goto bad; 1069 1070 /* 1071 * Now we support only SCM_RIGHTS, so return EINVAL 1072 * in any other cmsg_type 1073 */ 1074 if ((cmsg->cmsg_type = 1075 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type)) == -1) 1076 goto bad; 1077 cmsg->cmsg_level = 1078 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1079 1080 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1081 cmsg->cmsg_len = CMSG_LEN(datalen); 1082 data = LINUX_CMSG_DATA(ptr_cmsg); 1083 1084 error = ENOBUFS; 1085 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t) cmsg)) 1086 goto bad; 1087 if (!m_append(control, datalen, (c_caddr_t) data)) 1088 goto bad; 1089 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&msg, ptr_cmsg))); 1090 } else { 1091 control = NULL; 1092 cmsg = NULL; 1093 } 1094 1095 msg.msg_iov = iov; 1096 msg.msg_flags = 0; 1097 error = linux_sendit(td, args->s, &msg, args->flags, control, 1098 UIO_USERSPACE); 1099 1100 bad: 1101 free(iov, M_IOV); 1102 if (cmsg) 1103 free(cmsg, M_TEMP); 1104 return (error); 1105 } 1106 1107 struct linux_recvmsg_args { 1108 int s; 1109 l_uintptr_t msg; 1110 int flags; 1111 }; 1112 1113 static int 1114 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1115 { 1116 struct cmsghdr *cm; 1117 struct msghdr msg; 1118 struct l_cmsghdr *linux_cmsg = NULL; 1119 socklen_t datalen, outlen, clen; 1120 struct l_msghdr linux_msg; 1121 struct iovec *iov, *uiov; 1122 struct mbuf *control = NULL; 1123 struct mbuf **controlp; 1124 caddr_t outbuf; 1125 void *data; 1126 int error; 1127 1128 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg)); 1129 if (error) 1130 return (error); 1131 1132 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1133 if (error) 1134 return (error); 1135 1136 #ifdef COMPAT_LINUX32 1137 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1138 &iov, EMSGSIZE); 1139 #else 1140 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1141 #endif 1142 if (error) 1143 return (error); 1144 1145 if (msg.msg_name) { 1146 error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name, 1147 msg.msg_namelen); 1148 if (error) 1149 goto bad; 1150 } 1151 1152 uiov = msg.msg_iov; 1153 msg.msg_iov = iov; 1154 controlp = (msg.msg_control != NULL) ? &control : NULL; 1155 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp); 1156 msg.msg_iov = uiov; 1157 if (error) 1158 goto bad; 1159 1160 error = bsd_to_linux_msghdr(&msg, &linux_msg); 1161 if (error) 1162 goto bad; 1163 1164 if (linux_msg.msg_name) { 1165 error = bsd_to_linux_sockaddr((struct sockaddr *) 1166 PTRIN(linux_msg.msg_name)); 1167 if (error) 1168 goto bad; 1169 } 1170 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1171 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1172 if (error) 1173 goto bad; 1174 } 1175 1176 if (control) { 1177 1178 linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO); 1179 outbuf = PTRIN(linux_msg.msg_control); 1180 cm = mtod(control, struct cmsghdr *); 1181 outlen = 0; 1182 clen = control->m_len; 1183 1184 while (cm != NULL) { 1185 1186 if ((linux_cmsg->cmsg_type = 1187 bsd_to_linux_cmsg_type(cm->cmsg_type)) == -1) 1188 { 1189 error = EINVAL; 1190 goto bad; 1191 } 1192 data = CMSG_DATA(cm); 1193 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1194 1195 if (outlen + LINUX_CMSG_LEN(datalen) > 1196 linux_msg.msg_controllen) { 1197 if (outlen == 0) { 1198 error = EMSGSIZE; 1199 goto bad; 1200 } else { 1201 linux_msg.msg_flags |= LINUX_MSG_CTRUNC; 1202 goto out; 1203 } 1204 } 1205 1206 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1207 linux_cmsg->cmsg_level = 1208 bsd_to_linux_sockopt_level(cm->cmsg_level); 1209 1210 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1211 if (error) 1212 goto bad; 1213 outbuf += L_CMSG_HDRSZ; 1214 1215 error = copyout(data, outbuf, datalen); 1216 if (error) 1217 goto bad; 1218 1219 outbuf += LINUX_CMSG_ALIGN(datalen); 1220 outlen += LINUX_CMSG_LEN(datalen); 1221 linux_msg.msg_controllen = outlen; 1222 1223 if (CMSG_SPACE(datalen) < clen) { 1224 clen -= CMSG_SPACE(datalen); 1225 cm = (struct cmsghdr *) 1226 ((caddr_t)cm + CMSG_SPACE(datalen)); 1227 } else 1228 cm = NULL; 1229 } 1230 } 1231 1232 out: 1233 error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg)); 1234 1235 bad: 1236 free(iov, M_IOV); 1237 if (control != NULL) 1238 m_freem(control); 1239 if (linux_cmsg != NULL) 1240 free(linux_cmsg, M_TEMP); 1241 1242 return (error); 1243 } 1244 1245 struct linux_shutdown_args { 1246 int s; 1247 int how; 1248 }; 1249 1250 static int 1251 linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1252 { 1253 struct shutdown_args /* { 1254 int s; 1255 int how; 1256 } */ bsd_args; 1257 1258 bsd_args.s = args->s; 1259 bsd_args.how = args->how; 1260 return (shutdown(td, &bsd_args)); 1261 } 1262 1263 struct linux_setsockopt_args { 1264 int s; 1265 int level; 1266 int optname; 1267 l_uintptr_t optval; 1268 int optlen; 1269 }; 1270 1271 static int 1272 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1273 { 1274 struct setsockopt_args /* { 1275 int s; 1276 int level; 1277 int name; 1278 caddr_t val; 1279 int valsize; 1280 } */ bsd_args; 1281 int error, name; 1282 1283 bsd_args.s = args->s; 1284 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1285 switch (bsd_args.level) { 1286 case SOL_SOCKET: 1287 name = linux_to_bsd_so_sockopt(args->optname); 1288 break; 1289 case IPPROTO_IP: 1290 name = linux_to_bsd_ip_sockopt(args->optname); 1291 break; 1292 case IPPROTO_TCP: 1293 /* Linux TCP option values match BSD's */ 1294 name = args->optname; 1295 break; 1296 default: 1297 name = -1; 1298 break; 1299 } 1300 if (name == -1) 1301 return (ENOPROTOOPT); 1302 1303 bsd_args.name = name; 1304 bsd_args.val = PTRIN(args->optval); 1305 bsd_args.valsize = args->optlen; 1306 1307 if (name == IPV6_NEXTHOP) { 1308 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1309 bsd_args.valsize); 1310 error = setsockopt(td, &bsd_args); 1311 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1312 } else 1313 error = setsockopt(td, &bsd_args); 1314 1315 return (error); 1316 } 1317 1318 struct linux_getsockopt_args { 1319 int s; 1320 int level; 1321 int optname; 1322 l_uintptr_t optval; 1323 l_uintptr_t optlen; 1324 }; 1325 1326 static int 1327 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1328 { 1329 struct getsockopt_args /* { 1330 int s; 1331 int level; 1332 int name; 1333 caddr_t val; 1334 int *avalsize; 1335 } */ bsd_args; 1336 int error, name; 1337 1338 bsd_args.s = args->s; 1339 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1340 switch (bsd_args.level) { 1341 case SOL_SOCKET: 1342 name = linux_to_bsd_so_sockopt(args->optname); 1343 break; 1344 case IPPROTO_IP: 1345 name = linux_to_bsd_ip_sockopt(args->optname); 1346 break; 1347 case IPPROTO_TCP: 1348 /* Linux TCP option values match BSD's */ 1349 name = args->optname; 1350 break; 1351 default: 1352 name = -1; 1353 break; 1354 } 1355 if (name == -1) 1356 return (EINVAL); 1357 1358 bsd_args.name = name; 1359 bsd_args.val = PTRIN(args->optval); 1360 bsd_args.avalsize = PTRIN(args->optlen); 1361 1362 if (name == IPV6_NEXTHOP) { 1363 error = getsockopt(td, &bsd_args); 1364 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1365 } else 1366 error = getsockopt(td, &bsd_args); 1367 1368 return (error); 1369 } 1370 1371 int 1372 linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1373 { 1374 void *arg = (void *)(intptr_t)args->args; 1375 1376 switch (args->what) { 1377 case LINUX_SOCKET: 1378 return (linux_socket(td, arg)); 1379 case LINUX_BIND: 1380 return (linux_bind(td, arg)); 1381 case LINUX_CONNECT: 1382 return (linux_connect(td, arg)); 1383 case LINUX_LISTEN: 1384 return (linux_listen(td, arg)); 1385 case LINUX_ACCEPT: 1386 return (linux_accept(td, arg)); 1387 case LINUX_GETSOCKNAME: 1388 return (linux_getsockname(td, arg)); 1389 case LINUX_GETPEERNAME: 1390 return (linux_getpeername(td, arg)); 1391 case LINUX_SOCKETPAIR: 1392 return (linux_socketpair(td, arg)); 1393 case LINUX_SEND: 1394 return (linux_send(td, arg)); 1395 case LINUX_RECV: 1396 return (linux_recv(td, arg)); 1397 case LINUX_SENDTO: 1398 return (linux_sendto(td, arg)); 1399 case LINUX_RECVFROM: 1400 return (linux_recvfrom(td, arg)); 1401 case LINUX_SHUTDOWN: 1402 return (linux_shutdown(td, arg)); 1403 case LINUX_SETSOCKOPT: 1404 return (linux_setsockopt(td, arg)); 1405 case LINUX_GETSOCKOPT: 1406 return (linux_getsockopt(td, arg)); 1407 case LINUX_SENDMSG: 1408 return (linux_sendmsg(td, arg)); 1409 case LINUX_RECVMSG: 1410 return (linux_recvmsg(td, arg)); 1411 } 1412 1413 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1414 return (ENOSYS); 1415 } 1416