linux_socket.c revision 192205
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 192205 2009-05-16 18:46:51Z 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 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 601 return (EINVAL); 602 bsd_args.domain = linux_to_bsd_domain(args->domain); 603 if (bsd_args.domain == -1) 604 return (EAFNOSUPPORT); 605 606 retval_socket = socket(td, &bsd_args); 607 if (retval_socket) 608 return (retval_socket); 609 610 if (bsd_args.type == SOCK_RAW 611 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0) 612 && bsd_args.domain == PF_INET) { 613 /* It's a raw IP socket: set the IP_HDRINCL option. */ 614 int hdrincl; 615 616 hdrincl = 1; 617 /* We ignore any error returned by kern_setsockopt() */ 618 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 619 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 620 } 621#ifdef INET6 622 /* 623 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by 624 * default and some apps depend on this. So, set V6ONLY to 0 625 * for Linux apps if the sysctl value is set to 1. 626 */ 627 if (bsd_args.domain == PF_INET6 628#ifndef KLD_MODULE 629 /* 630 * XXX: Avoid undefined symbol error with an IPv4 only 631 * kernel. 632 */ 633 && V_ip6_v6only 634#endif 635 ) { 636 int v6only; 637 638 v6only = 0; 639 /* We ignore any error returned by setsockopt() */ 640 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 641 &v6only, UIO_SYSSPACE, sizeof(v6only)); 642 } 643#endif 644 645 return (retval_socket); 646} 647 648struct linux_bind_args { 649 int s; 650 l_uintptr_t name; 651 int namelen; 652}; 653 654static int 655linux_bind(struct thread *td, struct linux_bind_args *args) 656{ 657 struct sockaddr *sa; 658 int error; 659 660 error = linux_getsockaddr(&sa, PTRIN(args->name), 661 args->namelen); 662 if (error) 663 return (error); 664 665 error = kern_bind(td, args->s, sa); 666 free(sa, M_SONAME); 667 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 668 return (EINVAL); 669 return (error); 670} 671 672struct linux_connect_args { 673 int s; 674 l_uintptr_t name; 675 int namelen; 676}; 677int linux_connect(struct thread *, struct linux_connect_args *); 678 679int 680linux_connect(struct thread *td, struct linux_connect_args *args) 681{ 682 struct socket *so; 683 struct sockaddr *sa; 684 u_int fflag; 685 int error; 686 687 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 688 args->namelen); 689 if (error) 690 return (error); 691 692 error = kern_connect(td, args->s, sa); 693 free(sa, M_SONAME); 694 if (error != EISCONN) 695 return (error); 696 697 /* 698 * Linux doesn't return EISCONN the first time it occurs, 699 * when on a non-blocking socket. Instead it returns the 700 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 701 * 702 * XXXRW: Instead of using fgetsock(), check that it is a 703 * socket and use the file descriptor reference instead of 704 * creating a new one. 705 */ 706 error = fgetsock(td, args->s, &so, &fflag); 707 if (error == 0) { 708 error = EISCONN; 709 if (fflag & FNONBLOCK) { 710 SOCK_LOCK(so); 711 if (so->so_emuldata == 0) 712 error = so->so_error; 713 so->so_emuldata = (void *)1; 714 SOCK_UNLOCK(so); 715 } 716 fputsock(so); 717 } 718 return (error); 719} 720 721struct linux_listen_args { 722 int s; 723 int backlog; 724}; 725 726static int 727linux_listen(struct thread *td, struct linux_listen_args *args) 728{ 729 struct listen_args /* { 730 int s; 731 int backlog; 732 } */ bsd_args; 733 734 bsd_args.s = args->s; 735 bsd_args.backlog = args->backlog; 736 return (listen(td, &bsd_args)); 737} 738 739struct linux_accept_args { 740 int s; 741 l_uintptr_t addr; 742 l_uintptr_t namelen; 743}; 744 745static int 746linux_accept(struct thread *td, struct linux_accept_args *args) 747{ 748 struct accept_args /* { 749 int s; 750 struct sockaddr * __restrict name; 751 socklen_t * __restrict anamelen; 752 } */ bsd_args; 753 int error, fd; 754 755 bsd_args.s = args->s; 756 /* XXX: */ 757 bsd_args.name = (struct sockaddr * __restrict)PTRIN(args->addr); 758 bsd_args.anamelen = PTRIN(args->namelen);/* XXX */ 759 error = accept(td, &bsd_args); 760 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 761 if (error) { 762 if (error == EFAULT && args->namelen != sizeof(struct sockaddr_in)) 763 return (EINVAL); 764 return (error); 765 } 766 if (args->addr) { 767 error = linux_sa_put(PTRIN(args->addr)); 768 if (error) { 769 (void)kern_close(td, td->td_retval[0]); 770 return (error); 771 } 772 } 773 774 /* 775 * linux appears not to copy flags from the parent socket to the 776 * accepted one, so we must clear the flags in the new descriptor. 777 * Ignore any errors, because we already have an open fd. 778 */ 779 fd = td->td_retval[0]; 780 (void)kern_fcntl(td, fd, F_SETFL, 0); 781 td->td_retval[0] = fd; 782 return (0); 783} 784 785struct linux_getsockname_args { 786 int s; 787 l_uintptr_t addr; 788 l_uintptr_t namelen; 789}; 790 791static int 792linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 793{ 794 struct getsockname_args /* { 795 int fdes; 796 struct sockaddr * __restrict asa; 797 socklen_t * __restrict alen; 798 } */ bsd_args; 799 int error; 800 801 bsd_args.fdes = args->s; 802 /* XXX: */ 803 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 804 bsd_args.alen = PTRIN(args->namelen); /* XXX */ 805 error = getsockname(td, &bsd_args); 806 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 807 if (error) 808 return (error); 809 error = linux_sa_put(PTRIN(args->addr)); 810 if (error) 811 return (error); 812 return (0); 813} 814 815struct linux_getpeername_args { 816 int s; 817 l_uintptr_t addr; 818 l_uintptr_t namelen; 819}; 820 821static int 822linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 823{ 824 struct getpeername_args /* { 825 int fdes; 826 caddr_t asa; 827 int *alen; 828 } */ bsd_args; 829 int error; 830 831 bsd_args.fdes = args->s; 832 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 833 bsd_args.alen = (int *)PTRIN(args->namelen); 834 error = getpeername(td, &bsd_args); 835 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 836 if (error) 837 return (error); 838 error = linux_sa_put(PTRIN(args->addr)); 839 if (error) 840 return (error); 841 return (0); 842} 843 844struct linux_socketpair_args { 845 int domain; 846 int type; 847 int protocol; 848 l_uintptr_t rsv; 849}; 850 851static int 852linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 853{ 854 struct socketpair_args /* { 855 int domain; 856 int type; 857 int protocol; 858 int *rsv; 859 } */ bsd_args; 860 861 bsd_args.domain = linux_to_bsd_domain(args->domain); 862 if (bsd_args.domain != PF_LOCAL) 863 return (EAFNOSUPPORT); 864 865 bsd_args.type = args->type; 866 if (args->protocol != 0 && args->protocol != PF_UNIX) 867 868 /* 869 * Use of PF_UNIX as protocol argument is not right, 870 * but Linux does it. 871 * Do not map PF_UNIX as its Linux value is identical 872 * to FreeBSD one. 873 */ 874 return (EPROTONOSUPPORT); 875 else 876 bsd_args.protocol = 0; 877 bsd_args.rsv = (int *)PTRIN(args->rsv); 878 return (socketpair(td, &bsd_args)); 879} 880 881struct linux_send_args { 882 int s; 883 l_uintptr_t msg; 884 int len; 885 int flags; 886}; 887 888static int 889linux_send(struct thread *td, struct linux_send_args *args) 890{ 891 struct sendto_args /* { 892 int s; 893 caddr_t buf; 894 int len; 895 int flags; 896 caddr_t to; 897 int tolen; 898 } */ bsd_args; 899 900 bsd_args.s = args->s; 901 bsd_args.buf = (caddr_t)PTRIN(args->msg); 902 bsd_args.len = args->len; 903 bsd_args.flags = args->flags; 904 bsd_args.to = NULL; 905 bsd_args.tolen = 0; 906 return sendto(td, &bsd_args); 907} 908 909struct linux_recv_args { 910 int s; 911 l_uintptr_t msg; 912 int len; 913 int flags; 914}; 915 916static int 917linux_recv(struct thread *td, struct linux_recv_args *args) 918{ 919 struct recvfrom_args /* { 920 int s; 921 caddr_t buf; 922 int len; 923 int flags; 924 struct sockaddr *from; 925 socklen_t fromlenaddr; 926 } */ bsd_args; 927 928 bsd_args.s = args->s; 929 bsd_args.buf = (caddr_t)PTRIN(args->msg); 930 bsd_args.len = args->len; 931 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 932 bsd_args.from = NULL; 933 bsd_args.fromlenaddr = 0; 934 return (recvfrom(td, &bsd_args)); 935} 936 937static int 938linux_sendto(struct thread *td, struct linux_sendto_args *args) 939{ 940 struct msghdr msg; 941 struct iovec aiov; 942 int error; 943 944 if (linux_check_hdrincl(td, args->s) == 0) 945 /* IP_HDRINCL set, tweak the packet before sending */ 946 return (linux_sendto_hdrincl(td, args)); 947 948 msg.msg_name = PTRIN(args->to); 949 msg.msg_namelen = args->tolen; 950 msg.msg_iov = &aiov; 951 msg.msg_iovlen = 1; 952 msg.msg_control = NULL; 953 msg.msg_flags = 0; 954 aiov.iov_base = PTRIN(args->msg); 955 aiov.iov_len = args->len; 956 error = linux_sendit(td, args->s, &msg, args->flags, NULL, 957 UIO_USERSPACE); 958 return (error); 959} 960 961struct linux_recvfrom_args { 962 int s; 963 l_uintptr_t buf; 964 int len; 965 int flags; 966 l_uintptr_t from; 967 l_uintptr_t fromlen; 968}; 969 970static int 971linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 972{ 973 struct recvfrom_args /* { 974 int s; 975 caddr_t buf; 976 size_t len; 977 int flags; 978 struct sockaddr * __restrict from; 979 socklen_t * __restrict fromlenaddr; 980 } */ bsd_args; 981 size_t len; 982 int error; 983 984 if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t)))) 985 return (error); 986 987 bsd_args.s = args->s; 988 bsd_args.buf = PTRIN(args->buf); 989 bsd_args.len = args->len; 990 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 991 /* XXX: */ 992 bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from); 993 bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */ 994 995 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len); 996 error = recvfrom(td, &bsd_args); 997 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from); 998 999 if (error) 1000 return (error); 1001 if (args->from) { 1002 error = linux_sa_put((struct osockaddr *) 1003 PTRIN(args->from)); 1004 if (error) 1005 return (error); 1006 } 1007 return (0); 1008} 1009 1010struct linux_sendmsg_args { 1011 int s; 1012 l_uintptr_t msg; 1013 int flags; 1014}; 1015 1016static int 1017linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1018{ 1019 struct cmsghdr *cmsg; 1020 struct mbuf *control; 1021 struct msghdr msg; 1022 struct l_cmsghdr linux_cmsg; 1023 struct l_cmsghdr *ptr_cmsg; 1024 struct l_msghdr linux_msg; 1025 struct iovec *iov; 1026 socklen_t datalen; 1027 void *data; 1028 int error; 1029 1030 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg)); 1031 if (error) 1032 return (error); 1033 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1034 if (error) 1035 return (error); 1036 1037 /* 1038 * Some Linux applications (ping) define a non-NULL control data 1039 * pointer, but a msg_controllen of 0, which is not allowed in the 1040 * FreeBSD system call interface. NULL the msg_control pointer in 1041 * order to handle this case. This should be checked, but allows the 1042 * Linux ping to work. 1043 */ 1044 if (msg.msg_control != NULL && msg.msg_controllen == 0) 1045 msg.msg_control = NULL; 1046 1047#ifdef COMPAT_LINUX32 1048 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1049 &iov, EMSGSIZE); 1050#else 1051 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1052#endif 1053 if (error) 1054 return (error); 1055 1056 if (msg.msg_control != NULL) { 1057 error = ENOBUFS; 1058 cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO); 1059 control = m_get(M_WAIT, MT_CONTROL); 1060 if (control == NULL) 1061 goto bad; 1062 ptr_cmsg = LINUX_CMSG_FIRSTHDR(&msg); 1063 1064 do { 1065 error = copyin(ptr_cmsg, &linux_cmsg, 1066 sizeof(struct l_cmsghdr)); 1067 if (error) 1068 goto bad; 1069 1070 error = EINVAL; 1071 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1072 goto bad; 1073 1074 /* 1075 * Now we support only SCM_RIGHTS, so return EINVAL 1076 * in any other cmsg_type 1077 */ 1078 if ((cmsg->cmsg_type = 1079 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type)) == -1) 1080 goto bad; 1081 cmsg->cmsg_level = 1082 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1083 1084 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1085 cmsg->cmsg_len = CMSG_LEN(datalen); 1086 data = LINUX_CMSG_DATA(ptr_cmsg); 1087 1088 error = ENOBUFS; 1089 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t) cmsg)) 1090 goto bad; 1091 if (!m_append(control, datalen, (c_caddr_t) data)) 1092 goto bad; 1093 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&msg, ptr_cmsg))); 1094 } else { 1095 control = NULL; 1096 cmsg = NULL; 1097 } 1098 1099 msg.msg_iov = iov; 1100 msg.msg_flags = 0; 1101 error = linux_sendit(td, args->s, &msg, args->flags, control, 1102 UIO_USERSPACE); 1103 1104bad: 1105 free(iov, M_IOV); 1106 if (cmsg) 1107 free(cmsg, M_TEMP); 1108 return (error); 1109} 1110 1111struct linux_recvmsg_args { 1112 int s; 1113 l_uintptr_t msg; 1114 int flags; 1115}; 1116 1117static int 1118linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1119{ 1120 struct cmsghdr *cm; 1121 struct msghdr msg; 1122 struct l_cmsghdr *linux_cmsg = NULL; 1123 socklen_t datalen, outlen, clen; 1124 struct l_msghdr linux_msg; 1125 struct iovec *iov, *uiov; 1126 struct mbuf *control = NULL; 1127 struct mbuf **controlp; 1128 caddr_t outbuf; 1129 void *data; 1130 int error; 1131 1132 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg)); 1133 if (error) 1134 return (error); 1135 1136 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1137 if (error) 1138 return (error); 1139 1140#ifdef COMPAT_LINUX32 1141 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1142 &iov, EMSGSIZE); 1143#else 1144 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1145#endif 1146 if (error) 1147 return (error); 1148 1149 if (msg.msg_name) { 1150 error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name, 1151 msg.msg_namelen); 1152 if (error) 1153 goto bad; 1154 } 1155 1156 uiov = msg.msg_iov; 1157 msg.msg_iov = iov; 1158 controlp = (msg.msg_control != NULL) ? &control : NULL; 1159 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp); 1160 msg.msg_iov = uiov; 1161 if (error) 1162 goto bad; 1163 1164 error = bsd_to_linux_msghdr(&msg, &linux_msg); 1165 if (error) 1166 goto bad; 1167 1168 if (linux_msg.msg_name) { 1169 error = bsd_to_linux_sockaddr((struct sockaddr *) 1170 PTRIN(linux_msg.msg_name)); 1171 if (error) 1172 goto bad; 1173 } 1174 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1175 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1176 if (error) 1177 goto bad; 1178 } 1179 1180 if (control) { 1181 1182 linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO); 1183 outbuf = PTRIN(linux_msg.msg_control); 1184 cm = mtod(control, struct cmsghdr *); 1185 outlen = 0; 1186 clen = control->m_len; 1187 1188 while (cm != NULL) { 1189 1190 if ((linux_cmsg->cmsg_type = 1191 bsd_to_linux_cmsg_type(cm->cmsg_type)) == -1) 1192 { 1193 error = EINVAL; 1194 goto bad; 1195 } 1196 data = CMSG_DATA(cm); 1197 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1198 1199 if (outlen + LINUX_CMSG_LEN(datalen) > 1200 linux_msg.msg_controllen) { 1201 if (outlen == 0) { 1202 error = EMSGSIZE; 1203 goto bad; 1204 } else { 1205 linux_msg.msg_flags |= LINUX_MSG_CTRUNC; 1206 goto out; 1207 } 1208 } 1209 1210 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1211 linux_cmsg->cmsg_level = 1212 bsd_to_linux_sockopt_level(cm->cmsg_level); 1213 1214 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1215 if (error) 1216 goto bad; 1217 outbuf += L_CMSG_HDRSZ; 1218 1219 error = copyout(data, outbuf, datalen); 1220 if (error) 1221 goto bad; 1222 1223 outbuf += LINUX_CMSG_ALIGN(datalen); 1224 outlen += LINUX_CMSG_LEN(datalen); 1225 linux_msg.msg_controllen = outlen; 1226 1227 if (CMSG_SPACE(datalen) < clen) { 1228 clen -= CMSG_SPACE(datalen); 1229 cm = (struct cmsghdr *) 1230 ((caddr_t)cm + CMSG_SPACE(datalen)); 1231 } else 1232 cm = NULL; 1233 } 1234 } 1235 1236out: 1237 error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg)); 1238 1239bad: 1240 free(iov, M_IOV); 1241 if (control != NULL) 1242 m_freem(control); 1243 if (linux_cmsg != NULL) 1244 free(linux_cmsg, M_TEMP); 1245 1246 return (error); 1247} 1248 1249struct linux_shutdown_args { 1250 int s; 1251 int how; 1252}; 1253 1254static int 1255linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1256{ 1257 struct shutdown_args /* { 1258 int s; 1259 int how; 1260 } */ bsd_args; 1261 1262 bsd_args.s = args->s; 1263 bsd_args.how = args->how; 1264 return (shutdown(td, &bsd_args)); 1265} 1266 1267struct linux_setsockopt_args { 1268 int s; 1269 int level; 1270 int optname; 1271 l_uintptr_t optval; 1272 int optlen; 1273}; 1274 1275static int 1276linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1277{ 1278 struct setsockopt_args /* { 1279 int s; 1280 int level; 1281 int name; 1282 caddr_t val; 1283 int valsize; 1284 } */ bsd_args; 1285 l_timeval linux_tv; 1286 struct timeval tv; 1287 int error, name; 1288 1289 bsd_args.s = args->s; 1290 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1291 switch (bsd_args.level) { 1292 case SOL_SOCKET: 1293 name = linux_to_bsd_so_sockopt(args->optname); 1294 switch (name) { 1295 case SO_RCVTIMEO: 1296 /* FALLTHROUGH */ 1297 case SO_SNDTIMEO: 1298 error = copyin(PTRIN(args->optval), &linux_tv, 1299 sizeof(linux_tv)); 1300 if (error) 1301 return (error); 1302 tv.tv_sec = linux_tv.tv_sec; 1303 tv.tv_usec = linux_tv.tv_usec; 1304 return (kern_setsockopt(td, args->s, bsd_args.level, 1305 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1306 /* NOTREACHED */ 1307 break; 1308 default: 1309 break; 1310 } 1311 break; 1312 case IPPROTO_IP: 1313 name = linux_to_bsd_ip_sockopt(args->optname); 1314 break; 1315 case IPPROTO_TCP: 1316 /* Linux TCP option values match BSD's */ 1317 name = args->optname; 1318 break; 1319 default: 1320 name = -1; 1321 break; 1322 } 1323 if (name == -1) 1324 return (ENOPROTOOPT); 1325 1326 bsd_args.name = name; 1327 bsd_args.val = PTRIN(args->optval); 1328 bsd_args.valsize = args->optlen; 1329 1330 if (name == IPV6_NEXTHOP) { 1331 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1332 bsd_args.valsize); 1333 error = setsockopt(td, &bsd_args); 1334 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1335 } else 1336 error = setsockopt(td, &bsd_args); 1337 1338 return (error); 1339} 1340 1341struct linux_getsockopt_args { 1342 int s; 1343 int level; 1344 int optname; 1345 l_uintptr_t optval; 1346 l_uintptr_t optlen; 1347}; 1348 1349static int 1350linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1351{ 1352 struct getsockopt_args /* { 1353 int s; 1354 int level; 1355 int name; 1356 caddr_t val; 1357 int *avalsize; 1358 } */ bsd_args; 1359 l_timeval linux_tv; 1360 struct timeval tv; 1361 socklen_t tv_len, xulen; 1362 struct xucred xu; 1363 struct l_ucred lxu; 1364 int error, name; 1365 1366 bsd_args.s = args->s; 1367 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1368 switch (bsd_args.level) { 1369 case SOL_SOCKET: 1370 name = linux_to_bsd_so_sockopt(args->optname); 1371 switch (name) { 1372 case SO_RCVTIMEO: 1373 /* FALLTHROUGH */ 1374 case SO_SNDTIMEO: 1375 tv_len = sizeof(tv); 1376 error = kern_getsockopt(td, args->s, bsd_args.level, 1377 name, &tv, UIO_SYSSPACE, &tv_len); 1378 if (error) 1379 return (error); 1380 linux_tv.tv_sec = tv.tv_sec; 1381 linux_tv.tv_usec = tv.tv_usec; 1382 return (copyout(&linux_tv, PTRIN(args->optval), 1383 sizeof(linux_tv))); 1384 /* NOTREACHED */ 1385 break; 1386 case LOCAL_PEERCRED: 1387 if (args->optlen != sizeof(lxu)) 1388 return (EINVAL); 1389 xulen = sizeof(xu); 1390 error = kern_getsockopt(td, args->s, bsd_args.level, 1391 name, &xu, UIO_SYSSPACE, &xulen); 1392 if (error) 1393 return (error); 1394 /* 1395 * XXX Use 0 for pid as the FreeBSD does not cache peer pid. 1396 */ 1397 lxu.pid = 0; 1398 lxu.uid = xu.cr_uid; 1399 lxu.gid = xu.cr_gid; 1400 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1401 /* NOTREACHED */ 1402 break; 1403 default: 1404 break; 1405 } 1406 break; 1407 case IPPROTO_IP: 1408 name = linux_to_bsd_ip_sockopt(args->optname); 1409 break; 1410 case IPPROTO_TCP: 1411 /* Linux TCP option values match BSD's */ 1412 name = args->optname; 1413 break; 1414 default: 1415 name = -1; 1416 break; 1417 } 1418 if (name == -1) 1419 return (EINVAL); 1420 1421 bsd_args.name = name; 1422 bsd_args.val = PTRIN(args->optval); 1423 bsd_args.avalsize = PTRIN(args->optlen); 1424 1425 if (name == IPV6_NEXTHOP) { 1426 error = getsockopt(td, &bsd_args); 1427 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1428 } else 1429 error = getsockopt(td, &bsd_args); 1430 1431 return (error); 1432} 1433 1434int 1435linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1436{ 1437 void *arg = (void *)(intptr_t)args->args; 1438 1439 switch (args->what) { 1440 case LINUX_SOCKET: 1441 return (linux_socket(td, arg)); 1442 case LINUX_BIND: 1443 return (linux_bind(td, arg)); 1444 case LINUX_CONNECT: 1445 return (linux_connect(td, arg)); 1446 case LINUX_LISTEN: 1447 return (linux_listen(td, arg)); 1448 case LINUX_ACCEPT: 1449 return (linux_accept(td, arg)); 1450 case LINUX_GETSOCKNAME: 1451 return (linux_getsockname(td, arg)); 1452 case LINUX_GETPEERNAME: 1453 return (linux_getpeername(td, arg)); 1454 case LINUX_SOCKETPAIR: 1455 return (linux_socketpair(td, arg)); 1456 case LINUX_SEND: 1457 return (linux_send(td, arg)); 1458 case LINUX_RECV: 1459 return (linux_recv(td, arg)); 1460 case LINUX_SENDTO: 1461 return (linux_sendto(td, arg)); 1462 case LINUX_RECVFROM: 1463 return (linux_recvfrom(td, arg)); 1464 case LINUX_SHUTDOWN: 1465 return (linux_shutdown(td, arg)); 1466 case LINUX_SETSOCKOPT: 1467 return (linux_setsockopt(td, arg)); 1468 case LINUX_GETSOCKOPT: 1469 return (linux_getsockopt(td, arg)); 1470 case LINUX_SENDMSG: 1471 return (linux_sendmsg(td, arg)); 1472 case LINUX_RECVMSG: 1473 return (linux_recvmsg(td, arg)); 1474 } 1475 1476 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1477 return (ENOSYS); 1478} 1479