linux_socket.c revision 185571
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: head/sys/compat/linux/linux_socket.c 185571 2008-12-02 21:37:28Z bz $"); 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 76static int do_sa_get(struct sockaddr **, const struct osockaddr *, int *, 77 struct malloc_type *); 78static 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 */ 84static int 85linux_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 */ 96static int 97do_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 173out: 174 free(kosa, mtype); 175 return (error); 176} 177 178static int 179linux_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 201static int 202bsd_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 224static int 225linux_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 235static int 236bsd_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 246static int 247linux_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 273static int 274linux_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 318static int 319linux_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 364static int 365bsd_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 381static int 382linux_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 400static int 401linux_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 426static int 427linux_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 437static int 438bsd_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 450static int 451linux_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 466static int 467bsd_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 479static int 480linux_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. */ 503static int 504linux_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 517struct 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 */ 530static int 531linux_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); 572goout: 573 free(packet, M_TEMP); 574 return (error); 575} 576 577struct linux_socket_args { 578 int domain; 579 int type; 580 int protocol; 581}; 582 583static int 584linux_socket(struct thread *td, struct linux_socket_args *args) 585{ 586#ifdef INET6 587 INIT_VNET_INET6(curvnet); 588#endif 589 struct socket_args /* { 590 int domain; 591 int type; 592 int protocol; 593 } */ bsd_args; 594 int retval_socket; 595 596 bsd_args.protocol = args->protocol; 597 bsd_args.type = args->type; 598 bsd_args.domain = linux_to_bsd_domain(args->domain); 599 if (bsd_args.domain == -1) 600 return (EINVAL); 601 602 retval_socket = socket(td, &bsd_args); 603 if (bsd_args.type == SOCK_RAW 604 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0) 605 && bsd_args.domain == AF_INET 606 && retval_socket >= 0) { 607 /* It's a raw IP socket: set the IP_HDRINCL option. */ 608 int hdrincl; 609 610 hdrincl = 1; 611 /* We ignore any error returned by kern_setsockopt() */ 612 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 613 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 614 } 615#ifdef INET6 616 /* 617 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by 618 * default and some apps depend on this. So, set V6ONLY to 0 619 * for Linux apps if the sysctl value is set to 1. 620 */ 621 if (bsd_args.domain == PF_INET6 && retval_socket >= 0 622#ifndef KLD_MODULE 623 /* 624 * XXX: Avoid undefined symbol error with an IPv4 only 625 * kernel. 626 */ 627 && V_ip6_v6only 628#endif 629 ) { 630 int v6only; 631 632 v6only = 0; 633 /* We ignore any error returned by setsockopt() */ 634 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 635 &v6only, UIO_SYSSPACE, sizeof(v6only)); 636 } 637#endif 638 639 return (retval_socket); 640} 641 642struct linux_bind_args { 643 int s; 644 l_uintptr_t name; 645 int namelen; 646}; 647 648static int 649linux_bind(struct thread *td, struct linux_bind_args *args) 650{ 651 struct sockaddr *sa; 652 int error; 653 654 error = linux_getsockaddr(&sa, PTRIN(args->name), 655 args->namelen); 656 if (error) 657 return (error); 658 659 error = kern_bind(td, args->s, sa); 660 free(sa, M_SONAME); 661 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 662 return (EINVAL); 663 return (error); 664} 665 666struct linux_connect_args { 667 int s; 668 l_uintptr_t name; 669 int namelen; 670}; 671int linux_connect(struct thread *, struct linux_connect_args *); 672 673int 674linux_connect(struct thread *td, struct linux_connect_args *args) 675{ 676 struct socket *so; 677 struct sockaddr *sa; 678 u_int fflag; 679 int error; 680 681 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 682 args->namelen); 683 if (error) 684 return (error); 685 686 error = kern_connect(td, args->s, sa); 687 free(sa, M_SONAME); 688 if (error != EISCONN) 689 return (error); 690 691 /* 692 * Linux doesn't return EISCONN the first time it occurs, 693 * when on a non-blocking socket. Instead it returns the 694 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 695 * 696 * XXXRW: Instead of using fgetsock(), check that it is a 697 * socket and use the file descriptor reference instead of 698 * creating a new one. 699 */ 700 error = fgetsock(td, args->s, &so, &fflag); 701 if (error == 0) { 702 error = EISCONN; 703 if (fflag & FNONBLOCK) { 704 SOCK_LOCK(so); 705 if (so->so_emuldata == 0) 706 error = so->so_error; 707 so->so_emuldata = (void *)1; 708 SOCK_UNLOCK(so); 709 } 710 fputsock(so); 711 } 712 return (error); 713} 714 715struct linux_listen_args { 716 int s; 717 int backlog; 718}; 719 720static int 721linux_listen(struct thread *td, struct linux_listen_args *args) 722{ 723 struct listen_args /* { 724 int s; 725 int backlog; 726 } */ bsd_args; 727 728 bsd_args.s = args->s; 729 bsd_args.backlog = args->backlog; 730 return (listen(td, &bsd_args)); 731} 732 733struct linux_accept_args { 734 int s; 735 l_uintptr_t addr; 736 l_uintptr_t namelen; 737}; 738 739static int 740linux_accept(struct thread *td, struct linux_accept_args *args) 741{ 742 struct accept_args /* { 743 int s; 744 struct sockaddr * __restrict name; 745 socklen_t * __restrict anamelen; 746 } */ bsd_args; 747 int error, fd; 748 749 bsd_args.s = args->s; 750 /* XXX: */ 751 bsd_args.name = (struct sockaddr * __restrict)PTRIN(args->addr); 752 bsd_args.anamelen = PTRIN(args->namelen);/* XXX */ 753 error = accept(td, &bsd_args); 754 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 755 if (error) { 756 if (error == EFAULT && args->namelen != sizeof(struct sockaddr_in)) 757 return (EINVAL); 758 return (error); 759 } 760 if (args->addr) { 761 error = linux_sa_put(PTRIN(args->addr)); 762 if (error) { 763 (void)kern_close(td, td->td_retval[0]); 764 return (error); 765 } 766 } 767 768 /* 769 * linux appears not to copy flags from the parent socket to the 770 * accepted one, so we must clear the flags in the new descriptor. 771 * Ignore any errors, because we already have an open fd. 772 */ 773 fd = td->td_retval[0]; 774 (void)kern_fcntl(td, fd, F_SETFL, 0); 775 td->td_retval[0] = fd; 776 return (0); 777} 778 779struct linux_getsockname_args { 780 int s; 781 l_uintptr_t addr; 782 l_uintptr_t namelen; 783}; 784 785static int 786linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 787{ 788 struct getsockname_args /* { 789 int fdes; 790 struct sockaddr * __restrict asa; 791 socklen_t * __restrict alen; 792 } */ bsd_args; 793 int error; 794 795 bsd_args.fdes = args->s; 796 /* XXX: */ 797 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 798 bsd_args.alen = PTRIN(args->namelen); /* XXX */ 799 error = getsockname(td, &bsd_args); 800 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 801 if (error) 802 return (error); 803 error = linux_sa_put(PTRIN(args->addr)); 804 if (error) 805 return (error); 806 return (0); 807} 808 809struct linux_getpeername_args { 810 int s; 811 l_uintptr_t addr; 812 l_uintptr_t namelen; 813}; 814 815static int 816linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 817{ 818 struct getpeername_args /* { 819 int fdes; 820 caddr_t asa; 821 int *alen; 822 } */ bsd_args; 823 int error; 824 825 bsd_args.fdes = args->s; 826 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 827 bsd_args.alen = (int *)PTRIN(args->namelen); 828 error = getpeername(td, &bsd_args); 829 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 830 if (error) 831 return (error); 832 error = linux_sa_put(PTRIN(args->addr)); 833 if (error) 834 return (error); 835 return (0); 836} 837 838struct linux_socketpair_args { 839 int domain; 840 int type; 841 int protocol; 842 l_uintptr_t rsv; 843}; 844 845static int 846linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 847{ 848 struct socketpair_args /* { 849 int domain; 850 int type; 851 int protocol; 852 int *rsv; 853 } */ bsd_args; 854 855 bsd_args.domain = linux_to_bsd_domain(args->domain); 856 if (bsd_args.domain == -1) 857 return (EINVAL); 858 859 bsd_args.type = args->type; 860 bsd_args.protocol = args->protocol; 861 bsd_args.rsv = (int *)PTRIN(args->rsv); 862 return (socketpair(td, &bsd_args)); 863} 864 865struct linux_send_args { 866 int s; 867 l_uintptr_t msg; 868 int len; 869 int flags; 870}; 871 872static int 873linux_send(struct thread *td, struct linux_send_args *args) 874{ 875 struct sendto_args /* { 876 int s; 877 caddr_t buf; 878 int len; 879 int flags; 880 caddr_t to; 881 int tolen; 882 } */ bsd_args; 883 884 bsd_args.s = args->s; 885 bsd_args.buf = (caddr_t)PTRIN(args->msg); 886 bsd_args.len = args->len; 887 bsd_args.flags = args->flags; 888 bsd_args.to = NULL; 889 bsd_args.tolen = 0; 890 return sendto(td, &bsd_args); 891} 892 893struct linux_recv_args { 894 int s; 895 l_uintptr_t msg; 896 int len; 897 int flags; 898}; 899 900static int 901linux_recv(struct thread *td, struct linux_recv_args *args) 902{ 903 struct recvfrom_args /* { 904 int s; 905 caddr_t buf; 906 int len; 907 int flags; 908 struct sockaddr *from; 909 socklen_t fromlenaddr; 910 } */ bsd_args; 911 912 bsd_args.s = args->s; 913 bsd_args.buf = (caddr_t)PTRIN(args->msg); 914 bsd_args.len = args->len; 915 bsd_args.flags = args->flags; 916 bsd_args.from = NULL; 917 bsd_args.fromlenaddr = 0; 918 return (recvfrom(td, &bsd_args)); 919} 920 921static int 922linux_sendto(struct thread *td, struct linux_sendto_args *args) 923{ 924 struct msghdr msg; 925 struct iovec aiov; 926 int error; 927 928 if (linux_check_hdrincl(td, args->s) == 0) 929 /* IP_HDRINCL set, tweak the packet before sending */ 930 return (linux_sendto_hdrincl(td, args)); 931 932 msg.msg_name = PTRIN(args->to); 933 msg.msg_namelen = args->tolen; 934 msg.msg_iov = &aiov; 935 msg.msg_iovlen = 1; 936 msg.msg_control = NULL; 937 msg.msg_flags = 0; 938 aiov.iov_base = PTRIN(args->msg); 939 aiov.iov_len = args->len; 940 error = linux_sendit(td, args->s, &msg, args->flags, NULL, 941 UIO_USERSPACE); 942 return (error); 943} 944 945struct linux_recvfrom_args { 946 int s; 947 l_uintptr_t buf; 948 int len; 949 int flags; 950 l_uintptr_t from; 951 l_uintptr_t fromlen; 952}; 953 954static int 955linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 956{ 957 struct recvfrom_args /* { 958 int s; 959 caddr_t buf; 960 size_t len; 961 int flags; 962 struct sockaddr * __restrict from; 963 socklen_t * __restrict fromlenaddr; 964 } */ bsd_args; 965 size_t len; 966 int error; 967 968 if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t)))) 969 return (error); 970 971 bsd_args.s = args->s; 972 bsd_args.buf = PTRIN(args->buf); 973 bsd_args.len = args->len; 974 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 975 /* XXX: */ 976 bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from); 977 bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */ 978 979 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len); 980 error = recvfrom(td, &bsd_args); 981 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from); 982 983 if (error) 984 return (error); 985 if (args->from) { 986 error = linux_sa_put((struct osockaddr *) 987 PTRIN(args->from)); 988 if (error) 989 return (error); 990 } 991 return (0); 992} 993 994struct linux_sendmsg_args { 995 int s; 996 l_uintptr_t msg; 997 int flags; 998}; 999 1000static int 1001linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1002{ 1003 struct cmsghdr *cmsg; 1004 struct mbuf *control; 1005 struct msghdr msg; 1006 struct l_cmsghdr linux_cmsg; 1007 struct l_cmsghdr *ptr_cmsg; 1008 struct l_msghdr linux_msg; 1009 struct iovec *iov; 1010 socklen_t datalen; 1011 void *data; 1012 int error; 1013 1014 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg)); 1015 if (error) 1016 return (error); 1017 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1018 if (error) 1019 return (error); 1020 1021 /* 1022 * Some Linux applications (ping) define a non-NULL control data 1023 * pointer, but a msg_controllen of 0, which is not allowed in the 1024 * FreeBSD system call interface. NULL the msg_control pointer in 1025 * order to handle this case. This should be checked, but allows the 1026 * Linux ping to work. 1027 */ 1028 if (msg.msg_control != NULL && msg.msg_controllen == 0) 1029 msg.msg_control = NULL; 1030 1031#ifdef COMPAT_LINUX32 1032 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1033 &iov, EMSGSIZE); 1034#else 1035 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1036#endif 1037 if (error) 1038 return (error); 1039 1040 if (msg.msg_control != NULL) { 1041 error = ENOBUFS; 1042 cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO); 1043 control = m_get(M_WAIT, MT_CONTROL); 1044 if (control == NULL) 1045 goto bad; 1046 ptr_cmsg = LINUX_CMSG_FIRSTHDR(&msg); 1047 1048 do { 1049 error = copyin(ptr_cmsg, &linux_cmsg, 1050 sizeof(struct l_cmsghdr)); 1051 if (error) 1052 goto bad; 1053 1054 error = EINVAL; 1055 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1056 goto bad; 1057 1058 /* 1059 * Now we support only SCM_RIGHTS, so return EINVAL 1060 * in any other cmsg_type 1061 */ 1062 if ((cmsg->cmsg_type = 1063 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type)) == -1) 1064 goto bad; 1065 cmsg->cmsg_level = 1066 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1067 1068 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1069 cmsg->cmsg_len = CMSG_LEN(datalen); 1070 data = LINUX_CMSG_DATA(ptr_cmsg); 1071 1072 error = ENOBUFS; 1073 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t) cmsg)) 1074 goto bad; 1075 if (!m_append(control, datalen, (c_caddr_t) data)) 1076 goto bad; 1077 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&msg, ptr_cmsg))); 1078 } else { 1079 control = NULL; 1080 cmsg = NULL; 1081 } 1082 1083 msg.msg_iov = iov; 1084 msg.msg_flags = 0; 1085 error = linux_sendit(td, args->s, &msg, args->flags, control, 1086 UIO_USERSPACE); 1087 1088bad: 1089 free(iov, M_IOV); 1090 if (cmsg) 1091 free(cmsg, M_TEMP); 1092 return (error); 1093} 1094 1095struct linux_recvmsg_args { 1096 int s; 1097 l_uintptr_t msg; 1098 int flags; 1099}; 1100 1101static int 1102linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1103{ 1104 struct cmsghdr *cm; 1105 struct msghdr msg; 1106 struct l_cmsghdr *linux_cmsg = NULL; 1107 socklen_t datalen, outlen, clen; 1108 struct l_msghdr linux_msg; 1109 struct iovec *iov, *uiov; 1110 struct mbuf *control = NULL; 1111 struct mbuf **controlp; 1112 caddr_t outbuf; 1113 void *data; 1114 int error; 1115 1116 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg)); 1117 if (error) 1118 return (error); 1119 1120 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1121 if (error) 1122 return (error); 1123 1124#ifdef COMPAT_LINUX32 1125 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1126 &iov, EMSGSIZE); 1127#else 1128 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1129#endif 1130 if (error) 1131 return (error); 1132 1133 if (msg.msg_name) { 1134 error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name, 1135 msg.msg_namelen); 1136 if (error) 1137 goto bad; 1138 } 1139 1140 uiov = msg.msg_iov; 1141 msg.msg_iov = iov; 1142 controlp = (msg.msg_control != NULL) ? &control : NULL; 1143 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp); 1144 msg.msg_iov = uiov; 1145 if (error) 1146 goto bad; 1147 1148 error = bsd_to_linux_msghdr(&msg, &linux_msg); 1149 if (error) 1150 goto bad; 1151 1152 if (linux_msg.msg_name) { 1153 error = bsd_to_linux_sockaddr((struct sockaddr *) 1154 PTRIN(linux_msg.msg_name)); 1155 if (error) 1156 goto bad; 1157 } 1158 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1159 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1160 if (error) 1161 goto bad; 1162 } 1163 1164 if (control) { 1165 1166 linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO); 1167 outbuf = PTRIN(linux_msg.msg_control); 1168 cm = mtod(control, struct cmsghdr *); 1169 outlen = 0; 1170 clen = control->m_len; 1171 1172 while (cm != NULL) { 1173 1174 if ((linux_cmsg->cmsg_type = 1175 bsd_to_linux_cmsg_type(cm->cmsg_type)) == -1) 1176 { 1177 error = EINVAL; 1178 goto bad; 1179 } 1180 data = CMSG_DATA(cm); 1181 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1182 1183 if (outlen + LINUX_CMSG_LEN(datalen) > 1184 linux_msg.msg_controllen) { 1185 if (outlen == 0) { 1186 error = EMSGSIZE; 1187 goto bad; 1188 } else { 1189 linux_msg.msg_flags |= LINUX_MSG_CTRUNC; 1190 goto out; 1191 } 1192 } 1193 1194 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1195 linux_cmsg->cmsg_level = 1196 bsd_to_linux_sockopt_level(cm->cmsg_level); 1197 1198 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1199 if (error) 1200 goto bad; 1201 outbuf += L_CMSG_HDRSZ; 1202 1203 error = copyout(data, outbuf, datalen); 1204 if (error) 1205 goto bad; 1206 1207 outbuf += LINUX_CMSG_ALIGN(datalen); 1208 outlen += LINUX_CMSG_LEN(datalen); 1209 linux_msg.msg_controllen = outlen; 1210 1211 if (CMSG_SPACE(datalen) < clen) { 1212 clen -= CMSG_SPACE(datalen); 1213 cm = (struct cmsghdr *) 1214 ((caddr_t)cm + CMSG_SPACE(datalen)); 1215 } else 1216 cm = NULL; 1217 } 1218 } 1219 1220out: 1221 error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg)); 1222 1223bad: 1224 free(iov, M_IOV); 1225 if (control != NULL) 1226 m_freem(control); 1227 if (linux_cmsg != NULL) 1228 free(linux_cmsg, M_TEMP); 1229 1230 return (error); 1231} 1232 1233struct linux_shutdown_args { 1234 int s; 1235 int how; 1236}; 1237 1238static int 1239linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1240{ 1241 struct shutdown_args /* { 1242 int s; 1243 int how; 1244 } */ bsd_args; 1245 1246 bsd_args.s = args->s; 1247 bsd_args.how = args->how; 1248 return (shutdown(td, &bsd_args)); 1249} 1250 1251struct linux_setsockopt_args { 1252 int s; 1253 int level; 1254 int optname; 1255 l_uintptr_t optval; 1256 int optlen; 1257}; 1258 1259static int 1260linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1261{ 1262 struct setsockopt_args /* { 1263 int s; 1264 int level; 1265 int name; 1266 caddr_t val; 1267 int valsize; 1268 } */ bsd_args; 1269 int error, name; 1270 1271 bsd_args.s = args->s; 1272 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1273 switch (bsd_args.level) { 1274 case SOL_SOCKET: 1275 name = linux_to_bsd_so_sockopt(args->optname); 1276 break; 1277 case IPPROTO_IP: 1278 name = linux_to_bsd_ip_sockopt(args->optname); 1279 break; 1280 case IPPROTO_TCP: 1281 /* Linux TCP option values match BSD's */ 1282 name = args->optname; 1283 break; 1284 default: 1285 name = -1; 1286 break; 1287 } 1288 if (name == -1) 1289 return (ENOPROTOOPT); 1290 1291 bsd_args.name = name; 1292 bsd_args.val = PTRIN(args->optval); 1293 bsd_args.valsize = args->optlen; 1294 1295 if (name == IPV6_NEXTHOP) { 1296 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1297 bsd_args.valsize); 1298 error = setsockopt(td, &bsd_args); 1299 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1300 } else 1301 error = setsockopt(td, &bsd_args); 1302 1303 return (error); 1304} 1305 1306struct linux_getsockopt_args { 1307 int s; 1308 int level; 1309 int optname; 1310 l_uintptr_t optval; 1311 l_uintptr_t optlen; 1312}; 1313 1314static int 1315linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1316{ 1317 struct getsockopt_args /* { 1318 int s; 1319 int level; 1320 int name; 1321 caddr_t val; 1322 int *avalsize; 1323 } */ bsd_args; 1324 int error, name; 1325 1326 bsd_args.s = args->s; 1327 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1328 switch (bsd_args.level) { 1329 case SOL_SOCKET: 1330 name = linux_to_bsd_so_sockopt(args->optname); 1331 break; 1332 case IPPROTO_IP: 1333 name = linux_to_bsd_ip_sockopt(args->optname); 1334 break; 1335 case IPPROTO_TCP: 1336 /* Linux TCP option values match BSD's */ 1337 name = args->optname; 1338 break; 1339 default: 1340 name = -1; 1341 break; 1342 } 1343 if (name == -1) 1344 return (EINVAL); 1345 1346 bsd_args.name = name; 1347 bsd_args.val = PTRIN(args->optval); 1348 bsd_args.avalsize = PTRIN(args->optlen); 1349 1350 if (name == IPV6_NEXTHOP) { 1351 error = getsockopt(td, &bsd_args); 1352 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1353 } else 1354 error = getsockopt(td, &bsd_args); 1355 1356 return (error); 1357} 1358 1359int 1360linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1361{ 1362 void *arg = (void *)(intptr_t)args->args; 1363 1364 switch (args->what) { 1365 case LINUX_SOCKET: 1366 return (linux_socket(td, arg)); 1367 case LINUX_BIND: 1368 return (linux_bind(td, arg)); 1369 case LINUX_CONNECT: 1370 return (linux_connect(td, arg)); 1371 case LINUX_LISTEN: 1372 return (linux_listen(td, arg)); 1373 case LINUX_ACCEPT: 1374 return (linux_accept(td, arg)); 1375 case LINUX_GETSOCKNAME: 1376 return (linux_getsockname(td, arg)); 1377 case LINUX_GETPEERNAME: 1378 return (linux_getpeername(td, arg)); 1379 case LINUX_SOCKETPAIR: 1380 return (linux_socketpair(td, arg)); 1381 case LINUX_SEND: 1382 return (linux_send(td, arg)); 1383 case LINUX_RECV: 1384 return (linux_recv(td, arg)); 1385 case LINUX_SENDTO: 1386 return (linux_sendto(td, arg)); 1387 case LINUX_RECVFROM: 1388 return (linux_recvfrom(td, arg)); 1389 case LINUX_SHUTDOWN: 1390 return (linux_shutdown(td, arg)); 1391 case LINUX_SETSOCKOPT: 1392 return (linux_setsockopt(td, arg)); 1393 case LINUX_GETSOCKOPT: 1394 return (linux_getsockopt(td, arg)); 1395 case LINUX_SENDMSG: 1396 return (linux_sendmsg(td, arg)); 1397 case LINUX_RECVMSG: 1398 return (linux_recvmsg(td, arg)); 1399 } 1400 1401 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1402 return (ENOSYS); 1403} 1404