linux_socket.c revision 192204
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 192204 2009-05-16 18:44:56Z dchagin $"); 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 (EAFNOSUPPORT); 603 604 retval_socket = socket(td, &bsd_args); 605 if (retval_socket) 606 return (retval_socket); 607 608 if (bsd_args.type == SOCK_RAW 609 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0) 610 && bsd_args.domain == PF_INET) { 611 /* It's a raw IP socket: set the IP_HDRINCL option. */ 612 int hdrincl; 613 614 hdrincl = 1; 615 /* We ignore any error returned by kern_setsockopt() */ 616 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 617 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 618 } 619#ifdef INET6 620 /* 621 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by 622 * default and some apps depend on this. So, set V6ONLY to 0 623 * for Linux apps if the sysctl value is set to 1. 624 */ 625 if (bsd_args.domain == PF_INET6 626#ifndef KLD_MODULE 627 /* 628 * XXX: Avoid undefined symbol error with an IPv4 only 629 * kernel. 630 */ 631 && V_ip6_v6only 632#endif 633 ) { 634 int v6only; 635 636 v6only = 0; 637 /* We ignore any error returned by setsockopt() */ 638 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 639 &v6only, UIO_SYSSPACE, sizeof(v6only)); 640 } 641#endif 642 643 return (retval_socket); 644} 645 646struct linux_bind_args { 647 int s; 648 l_uintptr_t name; 649 int namelen; 650}; 651 652static int 653linux_bind(struct thread *td, struct linux_bind_args *args) 654{ 655 struct sockaddr *sa; 656 int error; 657 658 error = linux_getsockaddr(&sa, PTRIN(args->name), 659 args->namelen); 660 if (error) 661 return (error); 662 663 error = kern_bind(td, args->s, sa); 664 free(sa, M_SONAME); 665 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 666 return (EINVAL); 667 return (error); 668} 669 670struct linux_connect_args { 671 int s; 672 l_uintptr_t name; 673 int namelen; 674}; 675int linux_connect(struct thread *, struct linux_connect_args *); 676 677int 678linux_connect(struct thread *td, struct linux_connect_args *args) 679{ 680 struct socket *so; 681 struct sockaddr *sa; 682 u_int fflag; 683 int error; 684 685 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 686 args->namelen); 687 if (error) 688 return (error); 689 690 error = kern_connect(td, args->s, sa); 691 free(sa, M_SONAME); 692 if (error != EISCONN) 693 return (error); 694 695 /* 696 * Linux doesn't return EISCONN the first time it occurs, 697 * when on a non-blocking socket. Instead it returns the 698 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 699 * 700 * XXXRW: Instead of using fgetsock(), check that it is a 701 * socket and use the file descriptor reference instead of 702 * creating a new one. 703 */ 704 error = fgetsock(td, args->s, &so, &fflag); 705 if (error == 0) { 706 error = EISCONN; 707 if (fflag & FNONBLOCK) { 708 SOCK_LOCK(so); 709 if (so->so_emuldata == 0) 710 error = so->so_error; 711 so->so_emuldata = (void *)1; 712 SOCK_UNLOCK(so); 713 } 714 fputsock(so); 715 } 716 return (error); 717} 718 719struct linux_listen_args { 720 int s; 721 int backlog; 722}; 723 724static int 725linux_listen(struct thread *td, struct linux_listen_args *args) 726{ 727 struct listen_args /* { 728 int s; 729 int backlog; 730 } */ bsd_args; 731 732 bsd_args.s = args->s; 733 bsd_args.backlog = args->backlog; 734 return (listen(td, &bsd_args)); 735} 736 737struct linux_accept_args { 738 int s; 739 l_uintptr_t addr; 740 l_uintptr_t namelen; 741}; 742 743static int 744linux_accept(struct thread *td, struct linux_accept_args *args) 745{ 746 struct accept_args /* { 747 int s; 748 struct sockaddr * __restrict name; 749 socklen_t * __restrict anamelen; 750 } */ bsd_args; 751 int error, fd; 752 753 bsd_args.s = args->s; 754 /* XXX: */ 755 bsd_args.name = (struct sockaddr * __restrict)PTRIN(args->addr); 756 bsd_args.anamelen = PTRIN(args->namelen);/* XXX */ 757 error = accept(td, &bsd_args); 758 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 759 if (error) { 760 if (error == EFAULT && args->namelen != sizeof(struct sockaddr_in)) 761 return (EINVAL); 762 return (error); 763 } 764 if (args->addr) { 765 error = linux_sa_put(PTRIN(args->addr)); 766 if (error) { 767 (void)kern_close(td, td->td_retval[0]); 768 return (error); 769 } 770 } 771 772 /* 773 * linux appears not to copy flags from the parent socket to the 774 * accepted one, so we must clear the flags in the new descriptor. 775 * Ignore any errors, because we already have an open fd. 776 */ 777 fd = td->td_retval[0]; 778 (void)kern_fcntl(td, fd, F_SETFL, 0); 779 td->td_retval[0] = fd; 780 return (0); 781} 782 783struct linux_getsockname_args { 784 int s; 785 l_uintptr_t addr; 786 l_uintptr_t namelen; 787}; 788 789static int 790linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 791{ 792 struct getsockname_args /* { 793 int fdes; 794 struct sockaddr * __restrict asa; 795 socklen_t * __restrict alen; 796 } */ bsd_args; 797 int error; 798 799 bsd_args.fdes = args->s; 800 /* XXX: */ 801 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 802 bsd_args.alen = PTRIN(args->namelen); /* XXX */ 803 error = getsockname(td, &bsd_args); 804 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 805 if (error) 806 return (error); 807 error = linux_sa_put(PTRIN(args->addr)); 808 if (error) 809 return (error); 810 return (0); 811} 812 813struct linux_getpeername_args { 814 int s; 815 l_uintptr_t addr; 816 l_uintptr_t namelen; 817}; 818 819static int 820linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 821{ 822 struct getpeername_args /* { 823 int fdes; 824 caddr_t asa; 825 int *alen; 826 } */ bsd_args; 827 int error; 828 829 bsd_args.fdes = args->s; 830 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 831 bsd_args.alen = (int *)PTRIN(args->namelen); 832 error = getpeername(td, &bsd_args); 833 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 834 if (error) 835 return (error); 836 error = linux_sa_put(PTRIN(args->addr)); 837 if (error) 838 return (error); 839 return (0); 840} 841 842struct linux_socketpair_args { 843 int domain; 844 int type; 845 int protocol; 846 l_uintptr_t rsv; 847}; 848 849static int 850linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 851{ 852 struct socketpair_args /* { 853 int domain; 854 int type; 855 int protocol; 856 int *rsv; 857 } */ bsd_args; 858 859 bsd_args.domain = linux_to_bsd_domain(args->domain); 860 if (bsd_args.domain != PF_LOCAL) 861 return (EAFNOSUPPORT); 862 863 bsd_args.type = args->type; 864 if (args->protocol != 0 && args->protocol != PF_UNIX) 865 866 /* 867 * Use of PF_UNIX as protocol argument is not right, 868 * but Linux does it. 869 * Do not map PF_UNIX as its Linux value is identical 870 * to FreeBSD one. 871 */ 872 return (EPROTONOSUPPORT); 873 else 874 bsd_args.protocol = 0; 875 bsd_args.rsv = (int *)PTRIN(args->rsv); 876 return (socketpair(td, &bsd_args)); 877} 878 879struct linux_send_args { 880 int s; 881 l_uintptr_t msg; 882 int len; 883 int flags; 884}; 885 886static int 887linux_send(struct thread *td, struct linux_send_args *args) 888{ 889 struct sendto_args /* { 890 int s; 891 caddr_t buf; 892 int len; 893 int flags; 894 caddr_t to; 895 int tolen; 896 } */ bsd_args; 897 898 bsd_args.s = args->s; 899 bsd_args.buf = (caddr_t)PTRIN(args->msg); 900 bsd_args.len = args->len; 901 bsd_args.flags = args->flags; 902 bsd_args.to = NULL; 903 bsd_args.tolen = 0; 904 return sendto(td, &bsd_args); 905} 906 907struct linux_recv_args { 908 int s; 909 l_uintptr_t msg; 910 int len; 911 int flags; 912}; 913 914static int 915linux_recv(struct thread *td, struct linux_recv_args *args) 916{ 917 struct recvfrom_args /* { 918 int s; 919 caddr_t buf; 920 int len; 921 int flags; 922 struct sockaddr *from; 923 socklen_t fromlenaddr; 924 } */ bsd_args; 925 926 bsd_args.s = args->s; 927 bsd_args.buf = (caddr_t)PTRIN(args->msg); 928 bsd_args.len = args->len; 929 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 930 bsd_args.from = NULL; 931 bsd_args.fromlenaddr = 0; 932 return (recvfrom(td, &bsd_args)); 933} 934 935static int 936linux_sendto(struct thread *td, struct linux_sendto_args *args) 937{ 938 struct msghdr msg; 939 struct iovec aiov; 940 int error; 941 942 if (linux_check_hdrincl(td, args->s) == 0) 943 /* IP_HDRINCL set, tweak the packet before sending */ 944 return (linux_sendto_hdrincl(td, args)); 945 946 msg.msg_name = PTRIN(args->to); 947 msg.msg_namelen = args->tolen; 948 msg.msg_iov = &aiov; 949 msg.msg_iovlen = 1; 950 msg.msg_control = NULL; 951 msg.msg_flags = 0; 952 aiov.iov_base = PTRIN(args->msg); 953 aiov.iov_len = args->len; 954 error = linux_sendit(td, args->s, &msg, args->flags, NULL, 955 UIO_USERSPACE); 956 return (error); 957} 958 959struct linux_recvfrom_args { 960 int s; 961 l_uintptr_t buf; 962 int len; 963 int flags; 964 l_uintptr_t from; 965 l_uintptr_t fromlen; 966}; 967 968static int 969linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 970{ 971 struct recvfrom_args /* { 972 int s; 973 caddr_t buf; 974 size_t len; 975 int flags; 976 struct sockaddr * __restrict from; 977 socklen_t * __restrict fromlenaddr; 978 } */ bsd_args; 979 size_t len; 980 int error; 981 982 if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t)))) 983 return (error); 984 985 bsd_args.s = args->s; 986 bsd_args.buf = PTRIN(args->buf); 987 bsd_args.len = args->len; 988 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 989 /* XXX: */ 990 bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from); 991 bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */ 992 993 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len); 994 error = recvfrom(td, &bsd_args); 995 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from); 996 997 if (error) 998 return (error); 999 if (args->from) { 1000 error = linux_sa_put((struct osockaddr *) 1001 PTRIN(args->from)); 1002 if (error) 1003 return (error); 1004 } 1005 return (0); 1006} 1007 1008struct linux_sendmsg_args { 1009 int s; 1010 l_uintptr_t msg; 1011 int flags; 1012}; 1013 1014static int 1015linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1016{ 1017 struct cmsghdr *cmsg; 1018 struct mbuf *control; 1019 struct msghdr msg; 1020 struct l_cmsghdr linux_cmsg; 1021 struct l_cmsghdr *ptr_cmsg; 1022 struct l_msghdr linux_msg; 1023 struct iovec *iov; 1024 socklen_t datalen; 1025 void *data; 1026 int error; 1027 1028 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg)); 1029 if (error) 1030 return (error); 1031 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1032 if (error) 1033 return (error); 1034 1035 /* 1036 * Some Linux applications (ping) define a non-NULL control data 1037 * pointer, but a msg_controllen of 0, which is not allowed in the 1038 * FreeBSD system call interface. NULL the msg_control pointer in 1039 * order to handle this case. This should be checked, but allows the 1040 * Linux ping to work. 1041 */ 1042 if (msg.msg_control != NULL && msg.msg_controllen == 0) 1043 msg.msg_control = NULL; 1044 1045#ifdef COMPAT_LINUX32 1046 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1047 &iov, EMSGSIZE); 1048#else 1049 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1050#endif 1051 if (error) 1052 return (error); 1053 1054 if (msg.msg_control != NULL) { 1055 error = ENOBUFS; 1056 cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO); 1057 control = m_get(M_WAIT, MT_CONTROL); 1058 if (control == NULL) 1059 goto bad; 1060 ptr_cmsg = LINUX_CMSG_FIRSTHDR(&msg); 1061 1062 do { 1063 error = copyin(ptr_cmsg, &linux_cmsg, 1064 sizeof(struct l_cmsghdr)); 1065 if (error) 1066 goto bad; 1067 1068 error = EINVAL; 1069 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1070 goto bad; 1071 1072 /* 1073 * Now we support only SCM_RIGHTS, so return EINVAL 1074 * in any other cmsg_type 1075 */ 1076 if ((cmsg->cmsg_type = 1077 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type)) == -1) 1078 goto bad; 1079 cmsg->cmsg_level = 1080 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1081 1082 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1083 cmsg->cmsg_len = CMSG_LEN(datalen); 1084 data = LINUX_CMSG_DATA(ptr_cmsg); 1085 1086 error = ENOBUFS; 1087 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t) cmsg)) 1088 goto bad; 1089 if (!m_append(control, datalen, (c_caddr_t) data)) 1090 goto bad; 1091 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&msg, ptr_cmsg))); 1092 } else { 1093 control = NULL; 1094 cmsg = NULL; 1095 } 1096 1097 msg.msg_iov = iov; 1098 msg.msg_flags = 0; 1099 error = linux_sendit(td, args->s, &msg, args->flags, control, 1100 UIO_USERSPACE); 1101 1102bad: 1103 free(iov, M_IOV); 1104 if (cmsg) 1105 free(cmsg, M_TEMP); 1106 return (error); 1107} 1108 1109struct linux_recvmsg_args { 1110 int s; 1111 l_uintptr_t msg; 1112 int flags; 1113}; 1114 1115static int 1116linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1117{ 1118 struct cmsghdr *cm; 1119 struct msghdr msg; 1120 struct l_cmsghdr *linux_cmsg = NULL; 1121 socklen_t datalen, outlen, clen; 1122 struct l_msghdr linux_msg; 1123 struct iovec *iov, *uiov; 1124 struct mbuf *control = NULL; 1125 struct mbuf **controlp; 1126 caddr_t outbuf; 1127 void *data; 1128 int error; 1129 1130 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg)); 1131 if (error) 1132 return (error); 1133 1134 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1135 if (error) 1136 return (error); 1137 1138#ifdef COMPAT_LINUX32 1139 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1140 &iov, EMSGSIZE); 1141#else 1142 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1143#endif 1144 if (error) 1145 return (error); 1146 1147 if (msg.msg_name) { 1148 error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name, 1149 msg.msg_namelen); 1150 if (error) 1151 goto bad; 1152 } 1153 1154 uiov = msg.msg_iov; 1155 msg.msg_iov = iov; 1156 controlp = (msg.msg_control != NULL) ? &control : NULL; 1157 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp); 1158 msg.msg_iov = uiov; 1159 if (error) 1160 goto bad; 1161 1162 error = bsd_to_linux_msghdr(&msg, &linux_msg); 1163 if (error) 1164 goto bad; 1165 1166 if (linux_msg.msg_name) { 1167 error = bsd_to_linux_sockaddr((struct sockaddr *) 1168 PTRIN(linux_msg.msg_name)); 1169 if (error) 1170 goto bad; 1171 } 1172 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1173 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1174 if (error) 1175 goto bad; 1176 } 1177 1178 if (control) { 1179 1180 linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO); 1181 outbuf = PTRIN(linux_msg.msg_control); 1182 cm = mtod(control, struct cmsghdr *); 1183 outlen = 0; 1184 clen = control->m_len; 1185 1186 while (cm != NULL) { 1187 1188 if ((linux_cmsg->cmsg_type = 1189 bsd_to_linux_cmsg_type(cm->cmsg_type)) == -1) 1190 { 1191 error = EINVAL; 1192 goto bad; 1193 } 1194 data = CMSG_DATA(cm); 1195 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1196 1197 if (outlen + LINUX_CMSG_LEN(datalen) > 1198 linux_msg.msg_controllen) { 1199 if (outlen == 0) { 1200 error = EMSGSIZE; 1201 goto bad; 1202 } else { 1203 linux_msg.msg_flags |= LINUX_MSG_CTRUNC; 1204 goto out; 1205 } 1206 } 1207 1208 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1209 linux_cmsg->cmsg_level = 1210 bsd_to_linux_sockopt_level(cm->cmsg_level); 1211 1212 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1213 if (error) 1214 goto bad; 1215 outbuf += L_CMSG_HDRSZ; 1216 1217 error = copyout(data, outbuf, datalen); 1218 if (error) 1219 goto bad; 1220 1221 outbuf += LINUX_CMSG_ALIGN(datalen); 1222 outlen += LINUX_CMSG_LEN(datalen); 1223 linux_msg.msg_controllen = outlen; 1224 1225 if (CMSG_SPACE(datalen) < clen) { 1226 clen -= CMSG_SPACE(datalen); 1227 cm = (struct cmsghdr *) 1228 ((caddr_t)cm + CMSG_SPACE(datalen)); 1229 } else 1230 cm = NULL; 1231 } 1232 } 1233 1234out: 1235 error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg)); 1236 1237bad: 1238 free(iov, M_IOV); 1239 if (control != NULL) 1240 m_freem(control); 1241 if (linux_cmsg != NULL) 1242 free(linux_cmsg, M_TEMP); 1243 1244 return (error); 1245} 1246 1247struct linux_shutdown_args { 1248 int s; 1249 int how; 1250}; 1251 1252static int 1253linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1254{ 1255 struct shutdown_args /* { 1256 int s; 1257 int how; 1258 } */ bsd_args; 1259 1260 bsd_args.s = args->s; 1261 bsd_args.how = args->how; 1262 return (shutdown(td, &bsd_args)); 1263} 1264 1265struct linux_setsockopt_args { 1266 int s; 1267 int level; 1268 int optname; 1269 l_uintptr_t optval; 1270 int optlen; 1271}; 1272 1273static int 1274linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1275{ 1276 struct setsockopt_args /* { 1277 int s; 1278 int level; 1279 int name; 1280 caddr_t val; 1281 int valsize; 1282 } */ bsd_args; 1283 l_timeval linux_tv; 1284 struct timeval tv; 1285 int error, name; 1286 1287 bsd_args.s = args->s; 1288 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1289 switch (bsd_args.level) { 1290 case SOL_SOCKET: 1291 name = linux_to_bsd_so_sockopt(args->optname); 1292 switch (name) { 1293 case SO_RCVTIMEO: 1294 /* FALLTHROUGH */ 1295 case SO_SNDTIMEO: 1296 error = copyin(PTRIN(args->optval), &linux_tv, 1297 sizeof(linux_tv)); 1298 if (error) 1299 return (error); 1300 tv.tv_sec = linux_tv.tv_sec; 1301 tv.tv_usec = linux_tv.tv_usec; 1302 return (kern_setsockopt(td, args->s, bsd_args.level, 1303 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1304 /* NOTREACHED */ 1305 break; 1306 default: 1307 break; 1308 } 1309 break; 1310 case IPPROTO_IP: 1311 name = linux_to_bsd_ip_sockopt(args->optname); 1312 break; 1313 case IPPROTO_TCP: 1314 /* Linux TCP option values match BSD's */ 1315 name = args->optname; 1316 break; 1317 default: 1318 name = -1; 1319 break; 1320 } 1321 if (name == -1) 1322 return (ENOPROTOOPT); 1323 1324 bsd_args.name = name; 1325 bsd_args.val = PTRIN(args->optval); 1326 bsd_args.valsize = args->optlen; 1327 1328 if (name == IPV6_NEXTHOP) { 1329 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1330 bsd_args.valsize); 1331 error = setsockopt(td, &bsd_args); 1332 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1333 } else 1334 error = setsockopt(td, &bsd_args); 1335 1336 return (error); 1337} 1338 1339struct linux_getsockopt_args { 1340 int s; 1341 int level; 1342 int optname; 1343 l_uintptr_t optval; 1344 l_uintptr_t optlen; 1345}; 1346 1347static int 1348linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1349{ 1350 struct getsockopt_args /* { 1351 int s; 1352 int level; 1353 int name; 1354 caddr_t val; 1355 int *avalsize; 1356 } */ bsd_args; 1357 l_timeval linux_tv; 1358 struct timeval tv; 1359 socklen_t tv_len, xulen; 1360 struct xucred xu; 1361 struct l_ucred lxu; 1362 int error, name; 1363 1364 bsd_args.s = args->s; 1365 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1366 switch (bsd_args.level) { 1367 case SOL_SOCKET: 1368 name = linux_to_bsd_so_sockopt(args->optname); 1369 switch (name) { 1370 case SO_RCVTIMEO: 1371 /* FALLTHROUGH */ 1372 case SO_SNDTIMEO: 1373 tv_len = sizeof(tv); 1374 error = kern_getsockopt(td, args->s, bsd_args.level, 1375 name, &tv, UIO_SYSSPACE, &tv_len); 1376 if (error) 1377 return (error); 1378 linux_tv.tv_sec = tv.tv_sec; 1379 linux_tv.tv_usec = tv.tv_usec; 1380 return (copyout(&linux_tv, PTRIN(args->optval), 1381 sizeof(linux_tv))); 1382 /* NOTREACHED */ 1383 break; 1384 case LOCAL_PEERCRED: 1385 if (args->optlen != sizeof(lxu)) 1386 return (EINVAL); 1387 xulen = sizeof(xu); 1388 error = kern_getsockopt(td, args->s, bsd_args.level, 1389 name, &xu, UIO_SYSSPACE, &xulen); 1390 if (error) 1391 return (error); 1392 /* 1393 * XXX Use 0 for pid as the FreeBSD does not cache peer pid. 1394 */ 1395 lxu.pid = 0; 1396 lxu.uid = xu.cr_uid; 1397 lxu.gid = xu.cr_gid; 1398 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1399 /* NOTREACHED */ 1400 break; 1401 default: 1402 break; 1403 } 1404 break; 1405 case IPPROTO_IP: 1406 name = linux_to_bsd_ip_sockopt(args->optname); 1407 break; 1408 case IPPROTO_TCP: 1409 /* Linux TCP option values match BSD's */ 1410 name = args->optname; 1411 break; 1412 default: 1413 name = -1; 1414 break; 1415 } 1416 if (name == -1) 1417 return (EINVAL); 1418 1419 bsd_args.name = name; 1420 bsd_args.val = PTRIN(args->optval); 1421 bsd_args.avalsize = PTRIN(args->optlen); 1422 1423 if (name == IPV6_NEXTHOP) { 1424 error = getsockopt(td, &bsd_args); 1425 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1426 } else 1427 error = getsockopt(td, &bsd_args); 1428 1429 return (error); 1430} 1431 1432int 1433linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1434{ 1435 void *arg = (void *)(intptr_t)args->args; 1436 1437 switch (args->what) { 1438 case LINUX_SOCKET: 1439 return (linux_socket(td, arg)); 1440 case LINUX_BIND: 1441 return (linux_bind(td, arg)); 1442 case LINUX_CONNECT: 1443 return (linux_connect(td, arg)); 1444 case LINUX_LISTEN: 1445 return (linux_listen(td, arg)); 1446 case LINUX_ACCEPT: 1447 return (linux_accept(td, arg)); 1448 case LINUX_GETSOCKNAME: 1449 return (linux_getsockname(td, arg)); 1450 case LINUX_GETPEERNAME: 1451 return (linux_getpeername(td, arg)); 1452 case LINUX_SOCKETPAIR: 1453 return (linux_socketpair(td, arg)); 1454 case LINUX_SEND: 1455 return (linux_send(td, arg)); 1456 case LINUX_RECV: 1457 return (linux_recv(td, arg)); 1458 case LINUX_SENDTO: 1459 return (linux_sendto(td, arg)); 1460 case LINUX_RECVFROM: 1461 return (linux_recvfrom(td, arg)); 1462 case LINUX_SHUTDOWN: 1463 return (linux_shutdown(td, arg)); 1464 case LINUX_SETSOCKOPT: 1465 return (linux_setsockopt(td, arg)); 1466 case LINUX_GETSOCKOPT: 1467 return (linux_getsockopt(td, arg)); 1468 case LINUX_SENDMSG: 1469 return (linux_sendmsg(td, arg)); 1470 case LINUX_RECVMSG: 1471 return (linux_recvmsg(td, arg)); 1472 } 1473 1474 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1475 return (ENOSYS); 1476} 1477