linux_socket.c revision 191548
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 191548 2009-04-26 22:06:42Z zec $"); 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#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 (EINVAL); 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 644struct linux_bind_args { 645 int s; 646 l_uintptr_t name; 647 int namelen; 648}; 649 650static int 651linux_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 668struct linux_connect_args { 669 int s; 670 l_uintptr_t name; 671 int namelen; 672}; 673int linux_connect(struct thread *, struct linux_connect_args *); 674 675int 676linux_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 717struct linux_listen_args { 718 int s; 719 int backlog; 720}; 721 722static int 723linux_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 735struct linux_accept_args { 736 int s; 737 l_uintptr_t addr; 738 l_uintptr_t namelen; 739}; 740 741static int 742linux_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 781struct linux_getsockname_args { 782 int s; 783 l_uintptr_t addr; 784 l_uintptr_t namelen; 785}; 786 787static int 788linux_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 811struct linux_getpeername_args { 812 int s; 813 l_uintptr_t addr; 814 l_uintptr_t namelen; 815}; 816 817static int 818linux_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 840struct linux_socketpair_args { 841 int domain; 842 int type; 843 int protocol; 844 l_uintptr_t rsv; 845}; 846 847static int 848linux_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 == -1) 859 return (EINVAL); 860 861 bsd_args.type = args->type; 862 bsd_args.protocol = args->protocol; 863 bsd_args.rsv = (int *)PTRIN(args->rsv); 864 return (socketpair(td, &bsd_args)); 865} 866 867struct linux_send_args { 868 int s; 869 l_uintptr_t msg; 870 int len; 871 int flags; 872}; 873 874static int 875linux_send(struct thread *td, struct linux_send_args *args) 876{ 877 struct sendto_args /* { 878 int s; 879 caddr_t buf; 880 int len; 881 int flags; 882 caddr_t to; 883 int tolen; 884 } */ bsd_args; 885 886 bsd_args.s = args->s; 887 bsd_args.buf = (caddr_t)PTRIN(args->msg); 888 bsd_args.len = args->len; 889 bsd_args.flags = args->flags; 890 bsd_args.to = NULL; 891 bsd_args.tolen = 0; 892 return sendto(td, &bsd_args); 893} 894 895struct linux_recv_args { 896 int s; 897 l_uintptr_t msg; 898 int len; 899 int flags; 900}; 901 902static int 903linux_recv(struct thread *td, struct linux_recv_args *args) 904{ 905 struct recvfrom_args /* { 906 int s; 907 caddr_t buf; 908 int len; 909 int flags; 910 struct sockaddr *from; 911 socklen_t fromlenaddr; 912 } */ bsd_args; 913 914 bsd_args.s = args->s; 915 bsd_args.buf = (caddr_t)PTRIN(args->msg); 916 bsd_args.len = args->len; 917 bsd_args.flags = args->flags; 918 bsd_args.from = NULL; 919 bsd_args.fromlenaddr = 0; 920 return (recvfrom(td, &bsd_args)); 921} 922 923static int 924linux_sendto(struct thread *td, struct linux_sendto_args *args) 925{ 926 struct msghdr msg; 927 struct iovec aiov; 928 int error; 929 930 if (linux_check_hdrincl(td, args->s) == 0) 931 /* IP_HDRINCL set, tweak the packet before sending */ 932 return (linux_sendto_hdrincl(td, args)); 933 934 msg.msg_name = PTRIN(args->to); 935 msg.msg_namelen = args->tolen; 936 msg.msg_iov = &aiov; 937 msg.msg_iovlen = 1; 938 msg.msg_control = NULL; 939 msg.msg_flags = 0; 940 aiov.iov_base = PTRIN(args->msg); 941 aiov.iov_len = args->len; 942 error = linux_sendit(td, args->s, &msg, args->flags, NULL, 943 UIO_USERSPACE); 944 return (error); 945} 946 947struct linux_recvfrom_args { 948 int s; 949 l_uintptr_t buf; 950 int len; 951 int flags; 952 l_uintptr_t from; 953 l_uintptr_t fromlen; 954}; 955 956static int 957linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 958{ 959 struct recvfrom_args /* { 960 int s; 961 caddr_t buf; 962 size_t len; 963 int flags; 964 struct sockaddr * __restrict from; 965 socklen_t * __restrict fromlenaddr; 966 } */ bsd_args; 967 size_t len; 968 int error; 969 970 if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t)))) 971 return (error); 972 973 bsd_args.s = args->s; 974 bsd_args.buf = PTRIN(args->buf); 975 bsd_args.len = args->len; 976 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 977 /* XXX: */ 978 bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from); 979 bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */ 980 981 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len); 982 error = recvfrom(td, &bsd_args); 983 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from); 984 985 if (error) 986 return (error); 987 if (args->from) { 988 error = linux_sa_put((struct osockaddr *) 989 PTRIN(args->from)); 990 if (error) 991 return (error); 992 } 993 return (0); 994} 995 996struct linux_sendmsg_args { 997 int s; 998 l_uintptr_t msg; 999 int flags; 1000}; 1001 1002static int 1003linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1004{ 1005 struct cmsghdr *cmsg; 1006 struct mbuf *control; 1007 struct msghdr msg; 1008 struct l_cmsghdr linux_cmsg; 1009 struct l_cmsghdr *ptr_cmsg; 1010 struct l_msghdr linux_msg; 1011 struct iovec *iov; 1012 socklen_t datalen; 1013 void *data; 1014 int error; 1015 1016 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg)); 1017 if (error) 1018 return (error); 1019 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1020 if (error) 1021 return (error); 1022 1023 /* 1024 * Some Linux applications (ping) define a non-NULL control data 1025 * pointer, but a msg_controllen of 0, which is not allowed in the 1026 * FreeBSD system call interface. NULL the msg_control pointer in 1027 * order to handle this case. This should be checked, but allows the 1028 * Linux ping to work. 1029 */ 1030 if (msg.msg_control != NULL && msg.msg_controllen == 0) 1031 msg.msg_control = NULL; 1032 1033#ifdef COMPAT_LINUX32 1034 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1035 &iov, EMSGSIZE); 1036#else 1037 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1038#endif 1039 if (error) 1040 return (error); 1041 1042 if (msg.msg_control != NULL) { 1043 error = ENOBUFS; 1044 cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO); 1045 control = m_get(M_WAIT, MT_CONTROL); 1046 if (control == NULL) 1047 goto bad; 1048 ptr_cmsg = LINUX_CMSG_FIRSTHDR(&msg); 1049 1050 do { 1051 error = copyin(ptr_cmsg, &linux_cmsg, 1052 sizeof(struct l_cmsghdr)); 1053 if (error) 1054 goto bad; 1055 1056 error = EINVAL; 1057 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1058 goto bad; 1059 1060 /* 1061 * Now we support only SCM_RIGHTS, so return EINVAL 1062 * in any other cmsg_type 1063 */ 1064 if ((cmsg->cmsg_type = 1065 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type)) == -1) 1066 goto bad; 1067 cmsg->cmsg_level = 1068 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1069 1070 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1071 cmsg->cmsg_len = CMSG_LEN(datalen); 1072 data = LINUX_CMSG_DATA(ptr_cmsg); 1073 1074 error = ENOBUFS; 1075 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t) cmsg)) 1076 goto bad; 1077 if (!m_append(control, datalen, (c_caddr_t) data)) 1078 goto bad; 1079 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&msg, ptr_cmsg))); 1080 } else { 1081 control = NULL; 1082 cmsg = NULL; 1083 } 1084 1085 msg.msg_iov = iov; 1086 msg.msg_flags = 0; 1087 error = linux_sendit(td, args->s, &msg, args->flags, control, 1088 UIO_USERSPACE); 1089 1090bad: 1091 free(iov, M_IOV); 1092 if (cmsg) 1093 free(cmsg, M_TEMP); 1094 return (error); 1095} 1096 1097struct linux_recvmsg_args { 1098 int s; 1099 l_uintptr_t msg; 1100 int flags; 1101}; 1102 1103static int 1104linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1105{ 1106 struct cmsghdr *cm; 1107 struct msghdr msg; 1108 struct l_cmsghdr *linux_cmsg = NULL; 1109 socklen_t datalen, outlen, clen; 1110 struct l_msghdr linux_msg; 1111 struct iovec *iov, *uiov; 1112 struct mbuf *control = NULL; 1113 struct mbuf **controlp; 1114 caddr_t outbuf; 1115 void *data; 1116 int error; 1117 1118 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg)); 1119 if (error) 1120 return (error); 1121 1122 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1123 if (error) 1124 return (error); 1125 1126#ifdef COMPAT_LINUX32 1127 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1128 &iov, EMSGSIZE); 1129#else 1130 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1131#endif 1132 if (error) 1133 return (error); 1134 1135 if (msg.msg_name) { 1136 error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name, 1137 msg.msg_namelen); 1138 if (error) 1139 goto bad; 1140 } 1141 1142 uiov = msg.msg_iov; 1143 msg.msg_iov = iov; 1144 controlp = (msg.msg_control != NULL) ? &control : NULL; 1145 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp); 1146 msg.msg_iov = uiov; 1147 if (error) 1148 goto bad; 1149 1150 error = bsd_to_linux_msghdr(&msg, &linux_msg); 1151 if (error) 1152 goto bad; 1153 1154 if (linux_msg.msg_name) { 1155 error = bsd_to_linux_sockaddr((struct sockaddr *) 1156 PTRIN(linux_msg.msg_name)); 1157 if (error) 1158 goto bad; 1159 } 1160 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1161 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1162 if (error) 1163 goto bad; 1164 } 1165 1166 if (control) { 1167 1168 linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO); 1169 outbuf = PTRIN(linux_msg.msg_control); 1170 cm = mtod(control, struct cmsghdr *); 1171 outlen = 0; 1172 clen = control->m_len; 1173 1174 while (cm != NULL) { 1175 1176 if ((linux_cmsg->cmsg_type = 1177 bsd_to_linux_cmsg_type(cm->cmsg_type)) == -1) 1178 { 1179 error = EINVAL; 1180 goto bad; 1181 } 1182 data = CMSG_DATA(cm); 1183 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1184 1185 if (outlen + LINUX_CMSG_LEN(datalen) > 1186 linux_msg.msg_controllen) { 1187 if (outlen == 0) { 1188 error = EMSGSIZE; 1189 goto bad; 1190 } else { 1191 linux_msg.msg_flags |= LINUX_MSG_CTRUNC; 1192 goto out; 1193 } 1194 } 1195 1196 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1197 linux_cmsg->cmsg_level = 1198 bsd_to_linux_sockopt_level(cm->cmsg_level); 1199 1200 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1201 if (error) 1202 goto bad; 1203 outbuf += L_CMSG_HDRSZ; 1204 1205 error = copyout(data, outbuf, datalen); 1206 if (error) 1207 goto bad; 1208 1209 outbuf += LINUX_CMSG_ALIGN(datalen); 1210 outlen += LINUX_CMSG_LEN(datalen); 1211 linux_msg.msg_controllen = outlen; 1212 1213 if (CMSG_SPACE(datalen) < clen) { 1214 clen -= CMSG_SPACE(datalen); 1215 cm = (struct cmsghdr *) 1216 ((caddr_t)cm + CMSG_SPACE(datalen)); 1217 } else 1218 cm = NULL; 1219 } 1220 } 1221 1222out: 1223 error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg)); 1224 1225bad: 1226 free(iov, M_IOV); 1227 if (control != NULL) 1228 m_freem(control); 1229 if (linux_cmsg != NULL) 1230 free(linux_cmsg, M_TEMP); 1231 1232 return (error); 1233} 1234 1235struct linux_shutdown_args { 1236 int s; 1237 int how; 1238}; 1239 1240static int 1241linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1242{ 1243 struct shutdown_args /* { 1244 int s; 1245 int how; 1246 } */ bsd_args; 1247 1248 bsd_args.s = args->s; 1249 bsd_args.how = args->how; 1250 return (shutdown(td, &bsd_args)); 1251} 1252 1253struct linux_setsockopt_args { 1254 int s; 1255 int level; 1256 int optname; 1257 l_uintptr_t optval; 1258 int optlen; 1259}; 1260 1261static int 1262linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1263{ 1264 struct setsockopt_args /* { 1265 int s; 1266 int level; 1267 int name; 1268 caddr_t val; 1269 int valsize; 1270 } */ bsd_args; 1271 int error, name; 1272 1273 bsd_args.s = args->s; 1274 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1275 switch (bsd_args.level) { 1276 case SOL_SOCKET: 1277 name = linux_to_bsd_so_sockopt(args->optname); 1278 break; 1279 case IPPROTO_IP: 1280 name = linux_to_bsd_ip_sockopt(args->optname); 1281 break; 1282 case IPPROTO_TCP: 1283 /* Linux TCP option values match BSD's */ 1284 name = args->optname; 1285 break; 1286 default: 1287 name = -1; 1288 break; 1289 } 1290 if (name == -1) 1291 return (ENOPROTOOPT); 1292 1293 bsd_args.name = name; 1294 bsd_args.val = PTRIN(args->optval); 1295 bsd_args.valsize = args->optlen; 1296 1297 if (name == IPV6_NEXTHOP) { 1298 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1299 bsd_args.valsize); 1300 error = setsockopt(td, &bsd_args); 1301 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1302 } else 1303 error = setsockopt(td, &bsd_args); 1304 1305 return (error); 1306} 1307 1308struct linux_getsockopt_args { 1309 int s; 1310 int level; 1311 int optname; 1312 l_uintptr_t optval; 1313 l_uintptr_t optlen; 1314}; 1315 1316static int 1317linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1318{ 1319 struct getsockopt_args /* { 1320 int s; 1321 int level; 1322 int name; 1323 caddr_t val; 1324 int *avalsize; 1325 } */ bsd_args; 1326 int error, name; 1327 1328 bsd_args.s = args->s; 1329 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1330 switch (bsd_args.level) { 1331 case SOL_SOCKET: 1332 name = linux_to_bsd_so_sockopt(args->optname); 1333 break; 1334 case IPPROTO_IP: 1335 name = linux_to_bsd_ip_sockopt(args->optname); 1336 break; 1337 case IPPROTO_TCP: 1338 /* Linux TCP option values match BSD's */ 1339 name = args->optname; 1340 break; 1341 default: 1342 name = -1; 1343 break; 1344 } 1345 if (name == -1) 1346 return (EINVAL); 1347 1348 bsd_args.name = name; 1349 bsd_args.val = PTRIN(args->optval); 1350 bsd_args.avalsize = PTRIN(args->optlen); 1351 1352 if (name == IPV6_NEXTHOP) { 1353 error = getsockopt(td, &bsd_args); 1354 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1355 } else 1356 error = getsockopt(td, &bsd_args); 1357 1358 return (error); 1359} 1360 1361int 1362linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1363{ 1364 void *arg = (void *)(intptr_t)args->args; 1365 1366 switch (args->what) { 1367 case LINUX_SOCKET: 1368 return (linux_socket(td, arg)); 1369 case LINUX_BIND: 1370 return (linux_bind(td, arg)); 1371 case LINUX_CONNECT: 1372 return (linux_connect(td, arg)); 1373 case LINUX_LISTEN: 1374 return (linux_listen(td, arg)); 1375 case LINUX_ACCEPT: 1376 return (linux_accept(td, arg)); 1377 case LINUX_GETSOCKNAME: 1378 return (linux_getsockname(td, arg)); 1379 case LINUX_GETPEERNAME: 1380 return (linux_getpeername(td, arg)); 1381 case LINUX_SOCKETPAIR: 1382 return (linux_socketpair(td, arg)); 1383 case LINUX_SEND: 1384 return (linux_send(td, arg)); 1385 case LINUX_RECV: 1386 return (linux_recv(td, arg)); 1387 case LINUX_SENDTO: 1388 return (linux_sendto(td, arg)); 1389 case LINUX_RECVFROM: 1390 return (linux_recvfrom(td, arg)); 1391 case LINUX_SHUTDOWN: 1392 return (linux_shutdown(td, arg)); 1393 case LINUX_SETSOCKOPT: 1394 return (linux_setsockopt(td, arg)); 1395 case LINUX_GETSOCKOPT: 1396 return (linux_getsockopt(td, arg)); 1397 case LINUX_SENDMSG: 1398 return (linux_sendmsg(td, arg)); 1399 case LINUX_RECVMSG: 1400 return (linux_recvmsg(td, arg)); 1401 } 1402 1403 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1404 return (ENOSYS); 1405} 1406