linux_socket.c revision 297306
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: stable/10/sys/compat/linux/linux_socket.c 297306 2016-03-27 07:34:42Z 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/capsicum.h> 41#include <sys/fcntl.h> 42#include <sys/file.h> 43#include <sys/limits.h> 44#include <sys/lock.h> 45#include <sys/malloc.h> 46#include <sys/mutex.h> 47#include <sys/mbuf.h> 48#include <sys/socket.h> 49#include <sys/socketvar.h> 50#include <sys/syscallsubr.h> 51#include <sys/uio.h> 52#include <sys/syslog.h> 53#include <sys/un.h> 54 55#include <net/if.h> 56#include <netinet/in.h> 57#include <netinet/in_systm.h> 58#include <netinet/ip.h> 59#include <netinet/tcp.h> 60#ifdef INET6 61#include <netinet/ip6.h> 62#include <netinet6/ip6_var.h> 63#include <netinet6/in6_var.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_file.h> 74#include <compat/linux/linux_socket.h> 75#include <compat/linux/linux_timer.h> 76#include <compat/linux/linux_util.h> 77 78static int linux_to_bsd_domain(int); 79static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *, 80 l_uint); 81static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *, 82 l_uint, struct msghdr *); 83static int linux_set_socket_flags(int, int *); 84 85/* 86 * Reads a linux sockaddr and does any necessary translation. 87 * Linux sockaddrs don't have a length field, only a family. 88 * Copy the osockaddr structure pointed to by osa to kernel, adjust 89 * family and convert to sockaddr. 90 */ 91static int 92linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen) 93{ 94 struct sockaddr *sa; 95 struct osockaddr *kosa; 96#ifdef INET6 97 struct sockaddr_in6 *sin6; 98 int oldv6size; 99#endif 100 char *name; 101 int bdom, error, hdrlen, namelen; 102 103 if (salen < 2 || salen > UCHAR_MAX || !osa) 104 return (EINVAL); 105 106#ifdef INET6 107 oldv6size = 0; 108 /* 109 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it 110 * if it's a v4-mapped address, so reserve the proper space 111 * for it. 112 */ 113 if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) { 114 salen += sizeof(uint32_t); 115 oldv6size = 1; 116 } 117#endif 118 119 kosa = malloc(salen, M_SONAME, M_WAITOK); 120 121 if ((error = copyin(osa, kosa, salen))) 122 goto out; 123 124 bdom = linux_to_bsd_domain(kosa->sa_family); 125 if (bdom == -1) { 126 error = EAFNOSUPPORT; 127 goto out; 128 } 129 130#ifdef INET6 131 /* 132 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6, 133 * which lacks the scope id compared with RFC2553 one. If we detect 134 * the situation, reject the address and write a message to system log. 135 * 136 * Still accept addresses for which the scope id is not used. 137 */ 138 if (oldv6size) { 139 if (bdom == AF_INET6) { 140 sin6 = (struct sockaddr_in6 *)kosa; 141 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) || 142 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) && 143 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) && 144 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) && 145 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) && 146 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) { 147 sin6->sin6_scope_id = 0; 148 } else { 149 log(LOG_DEBUG, 150 "obsolete pre-RFC2553 sockaddr_in6 rejected\n"); 151 error = EINVAL; 152 goto out; 153 } 154 } else 155 salen -= sizeof(uint32_t); 156 } 157#endif 158 if (bdom == AF_INET) { 159 if (salen < sizeof(struct sockaddr_in)) { 160 error = EINVAL; 161 goto out; 162 } 163 salen = sizeof(struct sockaddr_in); 164 } 165 166 if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) { 167 hdrlen = offsetof(struct sockaddr_un, sun_path); 168 name = ((struct sockaddr_un *)kosa)->sun_path; 169 if (*name == '\0') { 170 /* 171 * Linux abstract namespace starts with a NULL byte. 172 * XXX We do not support abstract namespace yet. 173 */ 174 namelen = strnlen(name + 1, salen - hdrlen - 1) + 1; 175 } else 176 namelen = strnlen(name, salen - hdrlen); 177 salen = hdrlen + namelen; 178 if (salen > sizeof(struct sockaddr_un)) { 179 error = ENAMETOOLONG; 180 goto out; 181 } 182 } 183 184 sa = (struct sockaddr *)kosa; 185 sa->sa_family = bdom; 186 sa->sa_len = salen; 187 188 *sap = sa; 189 return (0); 190 191out: 192 free(kosa, M_SONAME); 193 return (error); 194} 195 196static int 197linux_to_bsd_domain(int domain) 198{ 199 200 switch (domain) { 201 case LINUX_AF_UNSPEC: 202 return (AF_UNSPEC); 203 case LINUX_AF_UNIX: 204 return (AF_LOCAL); 205 case LINUX_AF_INET: 206 return (AF_INET); 207 case LINUX_AF_INET6: 208 return (AF_INET6); 209 case LINUX_AF_AX25: 210 return (AF_CCITT); 211 case LINUX_AF_IPX: 212 return (AF_IPX); 213 case LINUX_AF_APPLETALK: 214 return (AF_APPLETALK); 215 } 216 return (-1); 217} 218 219static int 220bsd_to_linux_domain(int domain) 221{ 222 223 switch (domain) { 224 case AF_UNSPEC: 225 return (LINUX_AF_UNSPEC); 226 case AF_LOCAL: 227 return (LINUX_AF_UNIX); 228 case AF_INET: 229 return (LINUX_AF_INET); 230 case AF_INET6: 231 return (LINUX_AF_INET6); 232 case AF_CCITT: 233 return (LINUX_AF_AX25); 234 case AF_IPX: 235 return (LINUX_AF_IPX); 236 case AF_APPLETALK: 237 return (LINUX_AF_APPLETALK); 238 } 239 return (-1); 240} 241 242static int 243linux_to_bsd_sockopt_level(int level) 244{ 245 246 switch (level) { 247 case LINUX_SOL_SOCKET: 248 return (SOL_SOCKET); 249 } 250 return (level); 251} 252 253static int 254bsd_to_linux_sockopt_level(int level) 255{ 256 257 switch (level) { 258 case SOL_SOCKET: 259 return (LINUX_SOL_SOCKET); 260 } 261 return (level); 262} 263 264static int 265linux_to_bsd_ip_sockopt(int opt) 266{ 267 268 switch (opt) { 269 case LINUX_IP_TOS: 270 return (IP_TOS); 271 case LINUX_IP_TTL: 272 return (IP_TTL); 273 case LINUX_IP_OPTIONS: 274 return (IP_OPTIONS); 275 case LINUX_IP_MULTICAST_IF: 276 return (IP_MULTICAST_IF); 277 case LINUX_IP_MULTICAST_TTL: 278 return (IP_MULTICAST_TTL); 279 case LINUX_IP_MULTICAST_LOOP: 280 return (IP_MULTICAST_LOOP); 281 case LINUX_IP_ADD_MEMBERSHIP: 282 return (IP_ADD_MEMBERSHIP); 283 case LINUX_IP_DROP_MEMBERSHIP: 284 return (IP_DROP_MEMBERSHIP); 285 case LINUX_IP_HDRINCL: 286 return (IP_HDRINCL); 287 } 288 return (-1); 289} 290 291static int 292linux_to_bsd_ip6_sockopt(int opt) 293{ 294 295 switch (opt) { 296 case LINUX_IPV6_NEXTHOP: 297 return (IPV6_NEXTHOP); 298 case LINUX_IPV6_UNICAST_HOPS: 299 return (IPV6_UNICAST_HOPS); 300 case LINUX_IPV6_MULTICAST_IF: 301 return (IPV6_MULTICAST_IF); 302 case LINUX_IPV6_MULTICAST_HOPS: 303 return (IPV6_MULTICAST_HOPS); 304 case LINUX_IPV6_MULTICAST_LOOP: 305 return (IPV6_MULTICAST_LOOP); 306 case LINUX_IPV6_ADD_MEMBERSHIP: 307 return (IPV6_JOIN_GROUP); 308 case LINUX_IPV6_DROP_MEMBERSHIP: 309 return (IPV6_LEAVE_GROUP); 310 case LINUX_IPV6_V6ONLY: 311 return (IPV6_V6ONLY); 312 case LINUX_IPV6_DONTFRAG: 313 return (IPV6_DONTFRAG); 314#if 0 315 case LINUX_IPV6_CHECKSUM: 316 return (IPV6_CHECKSUM); 317 case LINUX_IPV6_RECVPKTINFO: 318 return (IPV6_RECVPKTINFO); 319 case LINUX_IPV6_PKTINFO: 320 return (IPV6_PKTINFO); 321 case LINUX_IPV6_RECVHOPLIMIT: 322 return (IPV6_RECVHOPLIMIT); 323 case LINUX_IPV6_HOPLIMIT: 324 return (IPV6_HOPLIMIT); 325 case LINUX_IPV6_RECVHOPOPTS: 326 return (IPV6_RECVHOPOPTS); 327 case LINUX_IPV6_HOPOPTS: 328 return (IPV6_HOPOPTS); 329 case LINUX_IPV6_RTHDRDSTOPTS: 330 return (IPV6_RTHDRDSTOPTS); 331 case LINUX_IPV6_RECVRTHDR: 332 return (IPV6_RECVRTHDR); 333 case LINUX_IPV6_RTHDR: 334 return (IPV6_RTHDR); 335 case LINUX_IPV6_RECVDSTOPTS: 336 return (IPV6_RECVDSTOPTS); 337 case LINUX_IPV6_DSTOPTS: 338 return (IPV6_DSTOPTS); 339 case LINUX_IPV6_RECVPATHMTU: 340 return (IPV6_RECVPATHMTU); 341 case LINUX_IPV6_PATHMTU: 342 return (IPV6_PATHMTU); 343#endif 344 } 345 return (-1); 346} 347 348static int 349linux_to_bsd_so_sockopt(int opt) 350{ 351 352 switch (opt) { 353 case LINUX_SO_DEBUG: 354 return (SO_DEBUG); 355 case LINUX_SO_REUSEADDR: 356 return (SO_REUSEADDR); 357 case LINUX_SO_TYPE: 358 return (SO_TYPE); 359 case LINUX_SO_ERROR: 360 return (SO_ERROR); 361 case LINUX_SO_DONTROUTE: 362 return (SO_DONTROUTE); 363 case LINUX_SO_BROADCAST: 364 return (SO_BROADCAST); 365 case LINUX_SO_SNDBUF: 366 return (SO_SNDBUF); 367 case LINUX_SO_RCVBUF: 368 return (SO_RCVBUF); 369 case LINUX_SO_KEEPALIVE: 370 return (SO_KEEPALIVE); 371 case LINUX_SO_OOBINLINE: 372 return (SO_OOBINLINE); 373 case LINUX_SO_LINGER: 374 return (SO_LINGER); 375 case LINUX_SO_PEERCRED: 376 return (LOCAL_PEERCRED); 377 case LINUX_SO_RCVLOWAT: 378 return (SO_RCVLOWAT); 379 case LINUX_SO_SNDLOWAT: 380 return (SO_SNDLOWAT); 381 case LINUX_SO_RCVTIMEO: 382 return (SO_RCVTIMEO); 383 case LINUX_SO_SNDTIMEO: 384 return (SO_SNDTIMEO); 385 case LINUX_SO_TIMESTAMP: 386 return (SO_TIMESTAMP); 387 case LINUX_SO_ACCEPTCONN: 388 return (SO_ACCEPTCONN); 389 } 390 return (-1); 391} 392 393static int 394linux_to_bsd_tcp_sockopt(int opt) 395{ 396 397 switch (opt) { 398 case LINUX_TCP_NODELAY: 399 return (TCP_NODELAY); 400 case LINUX_TCP_MAXSEG: 401 return (TCP_MAXSEG); 402 case LINUX_TCP_KEEPIDLE: 403 return (TCP_KEEPIDLE); 404 case LINUX_TCP_KEEPINTVL: 405 return (TCP_KEEPINTVL); 406 case LINUX_TCP_KEEPCNT: 407 return (TCP_KEEPCNT); 408 case LINUX_TCP_MD5SIG: 409 return (TCP_MD5SIG); 410 } 411 return (-1); 412} 413 414static int 415linux_to_bsd_msg_flags(int flags) 416{ 417 int ret_flags = 0; 418 419 if (flags & LINUX_MSG_OOB) 420 ret_flags |= MSG_OOB; 421 if (flags & LINUX_MSG_PEEK) 422 ret_flags |= MSG_PEEK; 423 if (flags & LINUX_MSG_DONTROUTE) 424 ret_flags |= MSG_DONTROUTE; 425 if (flags & LINUX_MSG_CTRUNC) 426 ret_flags |= MSG_CTRUNC; 427 if (flags & LINUX_MSG_TRUNC) 428 ret_flags |= MSG_TRUNC; 429 if (flags & LINUX_MSG_DONTWAIT) 430 ret_flags |= MSG_DONTWAIT; 431 if (flags & LINUX_MSG_EOR) 432 ret_flags |= MSG_EOR; 433 if (flags & LINUX_MSG_WAITALL) 434 ret_flags |= MSG_WAITALL; 435 if (flags & LINUX_MSG_NOSIGNAL) 436 ret_flags |= MSG_NOSIGNAL; 437#if 0 /* not handled */ 438 if (flags & LINUX_MSG_PROXY) 439 ; 440 if (flags & LINUX_MSG_FIN) 441 ; 442 if (flags & LINUX_MSG_SYN) 443 ; 444 if (flags & LINUX_MSG_CONFIRM) 445 ; 446 if (flags & LINUX_MSG_RST) 447 ; 448 if (flags & LINUX_MSG_ERRQUEUE) 449 ; 450#endif 451 return ret_flags; 452} 453 454/* 455* If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the 456* native syscall will fault. Thus, we don't really need to check the 457* return values for these functions. 458*/ 459 460static int 461bsd_to_linux_sockaddr(struct sockaddr *arg) 462{ 463 struct sockaddr sa; 464 size_t sa_len = sizeof(struct sockaddr); 465 int error; 466 467 if ((error = copyin(arg, &sa, sa_len))) 468 return (error); 469 470 *(u_short *)&sa = sa.sa_family; 471 472 error = copyout(&sa, arg, sa_len); 473 474 return (error); 475} 476 477static int 478linux_to_bsd_sockaddr(struct sockaddr *arg, int len) 479{ 480 struct sockaddr sa; 481 size_t sa_len = sizeof(struct sockaddr); 482 int error; 483 484 if ((error = copyin(arg, &sa, sa_len))) 485 return (error); 486 487 sa.sa_family = *(sa_family_t *)&sa; 488 sa.sa_len = len; 489 490 error = copyout(&sa, arg, sa_len); 491 492 return (error); 493} 494 495static int 496linux_sa_put(struct osockaddr *osa) 497{ 498 struct osockaddr sa; 499 int error, bdom; 500 501 /* 502 * Only read/write the osockaddr family part, the rest is 503 * not changed. 504 */ 505 error = copyin(osa, &sa, sizeof(sa.sa_family)); 506 if (error) 507 return (error); 508 509 bdom = bsd_to_linux_domain(sa.sa_family); 510 if (bdom == -1) 511 return (EINVAL); 512 513 sa.sa_family = bdom; 514 error = copyout(&sa, osa, sizeof(sa.sa_family)); 515 if (error) 516 return (error); 517 518 return (0); 519} 520 521static int 522linux_to_bsd_cmsg_type(int cmsg_type) 523{ 524 525 switch (cmsg_type) { 526 case LINUX_SCM_RIGHTS: 527 return (SCM_RIGHTS); 528 case LINUX_SCM_CREDENTIALS: 529 return (SCM_CREDS); 530 } 531 return (-1); 532} 533 534static int 535bsd_to_linux_cmsg_type(int cmsg_type) 536{ 537 538 switch (cmsg_type) { 539 case SCM_RIGHTS: 540 return (LINUX_SCM_RIGHTS); 541 case SCM_CREDS: 542 return (LINUX_SCM_CREDENTIALS); 543 case SCM_TIMESTAMP: 544 return (LINUX_SCM_TIMESTAMP); 545 } 546 return (-1); 547} 548 549static int 550linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 551{ 552 if (lhdr->msg_controllen > INT_MAX) 553 return (ENOBUFS); 554 555 bhdr->msg_name = PTRIN(lhdr->msg_name); 556 bhdr->msg_namelen = lhdr->msg_namelen; 557 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 558 bhdr->msg_iovlen = lhdr->msg_iovlen; 559 bhdr->msg_control = PTRIN(lhdr->msg_control); 560 561 /* 562 * msg_controllen is skipped since BSD and LINUX control messages 563 * are potentially different sizes (e.g. the cred structure used 564 * by SCM_CREDS is different between the two operating system). 565 * 566 * The caller can set it (if necessary) after converting all the 567 * control messages. 568 */ 569 570 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 571 return (0); 572} 573 574static int 575bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 576{ 577 lhdr->msg_name = PTROUT(bhdr->msg_name); 578 lhdr->msg_namelen = bhdr->msg_namelen; 579 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 580 lhdr->msg_iovlen = bhdr->msg_iovlen; 581 lhdr->msg_control = PTROUT(bhdr->msg_control); 582 583 /* 584 * msg_controllen is skipped since BSD and LINUX control messages 585 * are potentially different sizes (e.g. the cred structure used 586 * by SCM_CREDS is different between the two operating system). 587 * 588 * The caller can set it (if necessary) after converting all the 589 * control messages. 590 */ 591 592 /* msg_flags skipped */ 593 return (0); 594} 595 596static int 597linux_set_socket_flags(int lflags, int *flags) 598{ 599 600 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 601 return (EINVAL); 602 if (lflags & LINUX_SOCK_NONBLOCK) 603 *flags |= SOCK_NONBLOCK; 604 if (lflags & LINUX_SOCK_CLOEXEC) 605 *flags |= SOCK_CLOEXEC; 606 return (0); 607} 608 609static int 610linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 611 struct mbuf *control, enum uio_seg segflg) 612{ 613 struct sockaddr *to; 614 int error; 615 616 if (mp->msg_name != NULL) { 617 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen); 618 if (error) 619 return (error); 620 mp->msg_name = to; 621 } else 622 to = NULL; 623 624 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 625 segflg); 626 627 if (to) 628 free(to, M_SONAME); 629 return (error); 630} 631 632/* Return 0 if IP_HDRINCL is set for the given socket. */ 633static int 634linux_check_hdrincl(struct thread *td, int s) 635{ 636 int error, optval; 637 socklen_t size_val; 638 639 size_val = sizeof(optval); 640 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 641 &optval, UIO_SYSSPACE, &size_val); 642 if (error) 643 return (error); 644 645 return (optval == 0); 646} 647 648/* 649 * Updated sendto() when IP_HDRINCL is set: 650 * tweak endian-dependent fields in the IP packet. 651 */ 652static int 653linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 654{ 655/* 656 * linux_ip_copysize defines how many bytes we should copy 657 * from the beginning of the IP packet before we customize it for BSD. 658 * It should include all the fields we modify (ip_len and ip_off). 659 */ 660#define linux_ip_copysize 8 661 662 struct ip *packet; 663 struct msghdr msg; 664 struct iovec aiov[1]; 665 int error; 666 667 /* Check that the packet isn't too big or too small. */ 668 if (linux_args->len < linux_ip_copysize || 669 linux_args->len > IP_MAXPACKET) 670 return (EINVAL); 671 672 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK); 673 674 /* Make kernel copy of the packet to be sent */ 675 if ((error = copyin(PTRIN(linux_args->msg), packet, 676 linux_args->len))) 677 goto goout; 678 679 /* Convert fields from Linux to BSD raw IP socket format */ 680 packet->ip_len = linux_args->len; 681 packet->ip_off = ntohs(packet->ip_off); 682 683 /* Prepare the msghdr and iovec structures describing the new packet */ 684 msg.msg_name = PTRIN(linux_args->to); 685 msg.msg_namelen = linux_args->tolen; 686 msg.msg_iov = aiov; 687 msg.msg_iovlen = 1; 688 msg.msg_control = NULL; 689 msg.msg_flags = 0; 690 aiov[0].iov_base = (char *)packet; 691 aiov[0].iov_len = linux_args->len; 692 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 693 NULL, UIO_SYSSPACE); 694goout: 695 free(packet, M_LINUX); 696 return (error); 697} 698 699int 700linux_socket(struct thread *td, struct linux_socket_args *args) 701{ 702 struct socket_args /* { 703 int domain; 704 int type; 705 int protocol; 706 } */ bsd_args; 707 int retval_socket; 708 709 bsd_args.protocol = args->protocol; 710 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 711 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 712 return (EINVAL); 713 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 714 &bsd_args.type); 715 if (retval_socket != 0) 716 return (retval_socket); 717 bsd_args.domain = linux_to_bsd_domain(args->domain); 718 if (bsd_args.domain == -1) 719 return (EAFNOSUPPORT); 720 721 retval_socket = sys_socket(td, &bsd_args); 722 if (retval_socket) 723 return (retval_socket); 724 725 if (bsd_args.type == SOCK_RAW 726 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0) 727 && bsd_args.domain == PF_INET) { 728 /* It's a raw IP socket: set the IP_HDRINCL option. */ 729 int hdrincl; 730 731 hdrincl = 1; 732 /* We ignore any error returned by kern_setsockopt() */ 733 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 734 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 735 } 736#ifdef INET6 737 /* 738 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default 739 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps. 740 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only 741 * sysctl value. 742 */ 743 if (bsd_args.domain == PF_INET6) { 744 int v6only; 745 746 v6only = 0; 747 /* We ignore any error returned by setsockopt() */ 748 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 749 &v6only, UIO_SYSSPACE, sizeof(v6only)); 750 } 751#endif 752 753 return (retval_socket); 754} 755 756int 757linux_bind(struct thread *td, struct linux_bind_args *args) 758{ 759 struct sockaddr *sa; 760 int error; 761 762 error = linux_getsockaddr(&sa, PTRIN(args->name), 763 args->namelen); 764 if (error) 765 return (error); 766 767 error = kern_bind(td, args->s, sa); 768 free(sa, M_SONAME); 769 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 770 return (EINVAL); 771 return (error); 772} 773 774int 775linux_connect(struct thread *td, struct linux_connect_args *args) 776{ 777 cap_rights_t rights; 778 struct socket *so; 779 struct sockaddr *sa; 780 u_int fflag; 781 int error; 782 783 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 784 args->namelen); 785 if (error) 786 return (error); 787 788 error = kern_connect(td, args->s, sa); 789 free(sa, M_SONAME); 790 if (error != EISCONN) 791 return (error); 792 793 /* 794 * Linux doesn't return EISCONN the first time it occurs, 795 * when on a non-blocking socket. Instead it returns the 796 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 797 * 798 * XXXRW: Instead of using fgetsock(), check that it is a 799 * socket and use the file descriptor reference instead of 800 * creating a new one. 801 */ 802 error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT), 803 &so, &fflag); 804 if (error == 0) { 805 error = EISCONN; 806 if (fflag & FNONBLOCK) { 807 SOCK_LOCK(so); 808 if (so->so_emuldata == 0) 809 error = so->so_error; 810 so->so_emuldata = (void *)1; 811 SOCK_UNLOCK(so); 812 } 813 fputsock(so); 814 } 815 return (error); 816} 817 818int 819linux_listen(struct thread *td, struct linux_listen_args *args) 820{ 821 struct listen_args /* { 822 int s; 823 int backlog; 824 } */ bsd_args; 825 826 bsd_args.s = args->s; 827 bsd_args.backlog = args->backlog; 828 return (sys_listen(td, &bsd_args)); 829} 830 831static int 832linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 833 l_uintptr_t namelen, int flags) 834{ 835 struct accept4_args /* { 836 int s; 837 struct sockaddr * __restrict name; 838 socklen_t * __restrict anamelen; 839 int flags; 840 } */ bsd_args; 841 int error; 842 843 bsd_args.s = s; 844 /* XXX: */ 845 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr); 846 bsd_args.anamelen = PTRIN(namelen);/* XXX */ 847 bsd_args.flags = 0; 848 error = linux_set_socket_flags(flags, &bsd_args.flags); 849 if (error != 0) 850 return (error); 851 error = sys_accept4(td, &bsd_args); 852 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 853 if (error) { 854 if (error == EFAULT && namelen != sizeof(struct sockaddr_in)) 855 return (EINVAL); 856 return (error); 857 } 858 if (addr) 859 error = linux_sa_put(PTRIN(addr)); 860 if (error) { 861 (void)kern_close(td, td->td_retval[0]); 862 td->td_retval[0] = 0; 863 } 864 return (error); 865} 866 867int 868linux_accept(struct thread *td, struct linux_accept_args *args) 869{ 870 871 return (linux_accept_common(td, args->s, args->addr, 872 args->namelen, 0)); 873} 874 875int 876linux_accept4(struct thread *td, struct linux_accept4_args *args) 877{ 878 879 return (linux_accept_common(td, args->s, args->addr, 880 args->namelen, args->flags)); 881} 882 883int 884linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 885{ 886 struct getsockname_args /* { 887 int fdes; 888 struct sockaddr * __restrict asa; 889 socklen_t * __restrict alen; 890 } */ bsd_args; 891 int error; 892 893 bsd_args.fdes = args->s; 894 /* XXX: */ 895 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 896 bsd_args.alen = PTRIN(args->namelen); /* XXX */ 897 error = sys_getsockname(td, &bsd_args); 898 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 899 if (error) 900 return (error); 901 error = linux_sa_put(PTRIN(args->addr)); 902 if (error) 903 return (error); 904 return (0); 905} 906 907int 908linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 909{ 910 struct getpeername_args /* { 911 int fdes; 912 caddr_t asa; 913 int *alen; 914 } */ bsd_args; 915 int error; 916 917 bsd_args.fdes = args->s; 918 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 919 bsd_args.alen = (socklen_t *)PTRIN(args->namelen); 920 error = sys_getpeername(td, &bsd_args); 921 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 922 if (error) 923 return (error); 924 error = linux_sa_put(PTRIN(args->addr)); 925 if (error) 926 return (error); 927 return (0); 928} 929 930int 931linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 932{ 933 struct socketpair_args /* { 934 int domain; 935 int type; 936 int protocol; 937 int *rsv; 938 } */ bsd_args; 939 int error; 940 941 bsd_args.domain = linux_to_bsd_domain(args->domain); 942 if (bsd_args.domain != PF_LOCAL) 943 return (EAFNOSUPPORT); 944 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 945 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 946 return (EINVAL); 947 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 948 &bsd_args.type); 949 if (error != 0) 950 return (error); 951 if (args->protocol != 0 && args->protocol != PF_UNIX) 952 953 /* 954 * Use of PF_UNIX as protocol argument is not right, 955 * but Linux does it. 956 * Do not map PF_UNIX as its Linux value is identical 957 * to FreeBSD one. 958 */ 959 return (EPROTONOSUPPORT); 960 else 961 bsd_args.protocol = 0; 962 bsd_args.rsv = (int *)PTRIN(args->rsv); 963 return (sys_socketpair(td, &bsd_args)); 964} 965 966#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 967struct linux_send_args { 968 int s; 969 l_uintptr_t msg; 970 int len; 971 int flags; 972}; 973 974static int 975linux_send(struct thread *td, struct linux_send_args *args) 976{ 977 struct sendto_args /* { 978 int s; 979 caddr_t buf; 980 int len; 981 int flags; 982 caddr_t to; 983 int tolen; 984 } */ bsd_args; 985 986 bsd_args.s = args->s; 987 bsd_args.buf = (caddr_t)PTRIN(args->msg); 988 bsd_args.len = args->len; 989 bsd_args.flags = args->flags; 990 bsd_args.to = NULL; 991 bsd_args.tolen = 0; 992 return sys_sendto(td, &bsd_args); 993} 994 995struct linux_recv_args { 996 int s; 997 l_uintptr_t msg; 998 int len; 999 int flags; 1000}; 1001 1002static int 1003linux_recv(struct thread *td, struct linux_recv_args *args) 1004{ 1005 struct recvfrom_args /* { 1006 int s; 1007 caddr_t buf; 1008 int len; 1009 int flags; 1010 struct sockaddr *from; 1011 socklen_t fromlenaddr; 1012 } */ bsd_args; 1013 1014 bsd_args.s = args->s; 1015 bsd_args.buf = (caddr_t)PTRIN(args->msg); 1016 bsd_args.len = args->len; 1017 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 1018 bsd_args.from = NULL; 1019 bsd_args.fromlenaddr = 0; 1020 return (sys_recvfrom(td, &bsd_args)); 1021} 1022#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1023 1024int 1025linux_sendto(struct thread *td, struct linux_sendto_args *args) 1026{ 1027 struct msghdr msg; 1028 struct iovec aiov; 1029 int error; 1030 1031 if (linux_check_hdrincl(td, args->s) == 0) 1032 /* IP_HDRINCL set, tweak the packet before sending */ 1033 return (linux_sendto_hdrincl(td, args)); 1034 1035 msg.msg_name = PTRIN(args->to); 1036 msg.msg_namelen = args->tolen; 1037 msg.msg_iov = &aiov; 1038 msg.msg_iovlen = 1; 1039 msg.msg_control = NULL; 1040 msg.msg_flags = 0; 1041 aiov.iov_base = PTRIN(args->msg); 1042 aiov.iov_len = args->len; 1043 error = linux_sendit(td, args->s, &msg, args->flags, NULL, 1044 UIO_USERSPACE); 1045 return (error); 1046} 1047 1048int 1049linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 1050{ 1051 struct msghdr msg; 1052 struct iovec aiov; 1053 int error; 1054 1055 if (PTRIN(args->fromlen) != NULL) { 1056 error = copyin(PTRIN(args->fromlen), &msg.msg_namelen, 1057 sizeof(msg.msg_namelen)); 1058 if (error != 0) 1059 return (error); 1060 1061 error = linux_to_bsd_sockaddr((struct sockaddr *)PTRIN(args->from), 1062 msg.msg_namelen); 1063 if (error != 0) 1064 return (error); 1065 } else 1066 msg.msg_namelen = 0; 1067 1068 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from); 1069 msg.msg_iov = &aiov; 1070 msg.msg_iovlen = 1; 1071 aiov.iov_base = PTRIN(args->buf); 1072 aiov.iov_len = args->len; 1073 msg.msg_control = 0; 1074 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 1075 1076 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL); 1077 if (error != 0) 1078 return (error); 1079 1080 if (PTRIN(args->from) != NULL) { 1081 error = bsd_to_linux_sockaddr((struct sockaddr *) 1082 PTRIN(args->from)); 1083 if (error != 0) 1084 return (error); 1085 1086 error = linux_sa_put((struct osockaddr *) 1087 PTRIN(args->from)); 1088 } 1089 1090 if (PTRIN(args->fromlen) != NULL) 1091 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 1092 sizeof(msg.msg_namelen)); 1093 1094 return (error); 1095} 1096 1097static int 1098linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1099 l_uint flags) 1100{ 1101 struct cmsghdr *cmsg; 1102 struct cmsgcred cmcred; 1103 struct mbuf *control; 1104 struct msghdr msg; 1105 struct l_cmsghdr linux_cmsg; 1106 struct l_cmsghdr *ptr_cmsg; 1107 struct l_msghdr linux_msg; 1108 struct iovec *iov; 1109 socklen_t datalen; 1110 struct sockaddr *sa; 1111 sa_family_t sa_family; 1112 void *data; 1113 int error; 1114 1115 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1116 if (error != 0) 1117 return (error); 1118 1119 /* 1120 * Some Linux applications (ping) define a non-NULL control data 1121 * pointer, but a msg_controllen of 0, which is not allowed in the 1122 * FreeBSD system call interface. NULL the msg_control pointer in 1123 * order to handle this case. This should be checked, but allows the 1124 * Linux ping to work. 1125 */ 1126 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0) 1127 linux_msg.msg_control = PTROUT(NULL); 1128 1129 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1130 if (error != 0) 1131 return (error); 1132 1133#ifdef COMPAT_LINUX32 1134 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1135 &iov, EMSGSIZE); 1136#else 1137 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1138#endif 1139 if (error != 0) 1140 return (error); 1141 1142 control = NULL; 1143 cmsg = NULL; 1144 1145 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) { 1146 error = kern_getsockname(td, s, &sa, &datalen); 1147 if (error != 0) 1148 goto bad; 1149 sa_family = sa->sa_family; 1150 free(sa, M_SONAME); 1151 1152 error = ENOBUFS; 1153 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO); 1154 control = m_get(M_WAITOK, MT_CONTROL); 1155 1156 do { 1157 error = copyin(ptr_cmsg, &linux_cmsg, 1158 sizeof(struct l_cmsghdr)); 1159 if (error != 0) 1160 goto bad; 1161 1162 error = EINVAL; 1163 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1164 goto bad; 1165 1166 /* 1167 * Now we support only SCM_RIGHTS and SCM_CRED, 1168 * so return EINVAL in any other cmsg_type 1169 */ 1170 cmsg->cmsg_type = 1171 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1172 cmsg->cmsg_level = 1173 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1174 if (cmsg->cmsg_type == -1 1175 || cmsg->cmsg_level != SOL_SOCKET) 1176 goto bad; 1177 1178 /* 1179 * Some applications (e.g. pulseaudio) attempt to 1180 * send ancillary data even if the underlying protocol 1181 * doesn't support it which is not allowed in the 1182 * FreeBSD system call interface. 1183 */ 1184 if (sa_family != AF_UNIX) 1185 continue; 1186 1187 data = LINUX_CMSG_DATA(ptr_cmsg); 1188 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1189 1190 switch (cmsg->cmsg_type) 1191 { 1192 case SCM_RIGHTS: 1193 break; 1194 1195 case SCM_CREDS: 1196 data = &cmcred; 1197 datalen = sizeof(cmcred); 1198 1199 /* 1200 * The lower levels will fill in the structure 1201 */ 1202 bzero(data, datalen); 1203 break; 1204 } 1205 1206 cmsg->cmsg_len = CMSG_LEN(datalen); 1207 1208 error = ENOBUFS; 1209 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg)) 1210 goto bad; 1211 if (!m_append(control, datalen, (c_caddr_t)data)) 1212 goto bad; 1213 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg))); 1214 1215 if (m_length(control, NULL) == 0) { 1216 m_freem(control); 1217 control = NULL; 1218 } 1219 } 1220 1221 msg.msg_iov = iov; 1222 msg.msg_flags = 0; 1223 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1224 control = NULL; 1225 1226bad: 1227 m_freem(control); 1228 free(iov, M_IOV); 1229 if (cmsg) 1230 free(cmsg, M_LINUX); 1231 return (error); 1232} 1233 1234int 1235linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1236{ 1237 1238 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1239 args->flags)); 1240} 1241 1242int 1243linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1244{ 1245 struct l_mmsghdr *msg; 1246 l_uint retval; 1247 int error, datagrams; 1248 1249 if (args->vlen > UIO_MAXIOV) 1250 args->vlen = UIO_MAXIOV; 1251 1252 msg = PTRIN(args->msg); 1253 datagrams = 0; 1254 while (datagrams < args->vlen) { 1255 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1256 args->flags); 1257 if (error != 0) 1258 break; 1259 1260 retval = td->td_retval[0]; 1261 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1262 if (error != 0) 1263 break; 1264 ++msg; 1265 ++datagrams; 1266 } 1267 if (error == 0) 1268 td->td_retval[0] = datagrams; 1269 return (error); 1270} 1271 1272static int 1273linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1274 l_uint flags, struct msghdr *msg) 1275{ 1276 struct cmsghdr *cm; 1277 struct cmsgcred *cmcred; 1278 struct l_cmsghdr *linux_cmsg = NULL; 1279 struct l_ucred linux_ucred; 1280 socklen_t datalen, outlen; 1281 struct l_msghdr linux_msg; 1282 struct iovec *iov, *uiov; 1283 struct mbuf *control = NULL; 1284 struct mbuf **controlp; 1285 struct timeval *ftmvl; 1286 l_timeval ltmvl; 1287 caddr_t outbuf; 1288 void *data; 1289 int error, i, fd, fds, *fdp; 1290 1291 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1292 if (error != 0) 1293 return (error); 1294 1295 error = linux_to_bsd_msghdr(msg, &linux_msg); 1296 if (error != 0) 1297 return (error); 1298 1299#ifdef COMPAT_LINUX32 1300 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1301 &iov, EMSGSIZE); 1302#else 1303 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1304#endif 1305 if (error != 0) 1306 return (error); 1307 1308 if (msg->msg_name) { 1309 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name, 1310 msg->msg_namelen); 1311 if (error != 0) 1312 goto bad; 1313 } 1314 1315 uiov = msg->msg_iov; 1316 msg->msg_iov = iov; 1317 controlp = (msg->msg_control != NULL) ? &control : NULL; 1318 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp); 1319 msg->msg_iov = uiov; 1320 if (error != 0) 1321 goto bad; 1322 1323 error = bsd_to_linux_msghdr(msg, &linux_msg); 1324 if (error != 0) 1325 goto bad; 1326 1327 if (linux_msg.msg_name) { 1328 error = bsd_to_linux_sockaddr((struct sockaddr *) 1329 PTRIN(linux_msg.msg_name)); 1330 if (error != 0) 1331 goto bad; 1332 } 1333 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1334 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1335 if (error != 0) 1336 goto bad; 1337 } 1338 1339 outbuf = PTRIN(linux_msg.msg_control); 1340 outlen = 0; 1341 1342 if (control) { 1343 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1344 1345 msg->msg_control = mtod(control, struct cmsghdr *); 1346 msg->msg_controllen = control->m_len; 1347 1348 cm = CMSG_FIRSTHDR(msg); 1349 1350 while (cm != NULL) { 1351 linux_cmsg->cmsg_type = 1352 bsd_to_linux_cmsg_type(cm->cmsg_type); 1353 linux_cmsg->cmsg_level = 1354 bsd_to_linux_sockopt_level(cm->cmsg_level); 1355 if (linux_cmsg->cmsg_type == -1 1356 || cm->cmsg_level != SOL_SOCKET) 1357 { 1358 error = EINVAL; 1359 goto bad; 1360 } 1361 1362 data = CMSG_DATA(cm); 1363 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1364 1365 switch (cm->cmsg_type) 1366 { 1367 case SCM_RIGHTS: 1368 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1369 fds = datalen / sizeof(int); 1370 fdp = data; 1371 for (i = 0; i < fds; i++) { 1372 fd = *fdp++; 1373 (void)kern_fcntl(td, fd, 1374 F_SETFD, FD_CLOEXEC); 1375 } 1376 } 1377 break; 1378 1379 case SCM_CREDS: 1380 /* 1381 * Currently LOCAL_CREDS is never in 1382 * effect for Linux so no need to worry 1383 * about sockcred 1384 */ 1385 if (datalen != sizeof(*cmcred)) { 1386 error = EMSGSIZE; 1387 goto bad; 1388 } 1389 cmcred = (struct cmsgcred *)data; 1390 bzero(&linux_ucred, sizeof(linux_ucred)); 1391 linux_ucred.pid = cmcred->cmcred_pid; 1392 linux_ucred.uid = cmcred->cmcred_uid; 1393 linux_ucred.gid = cmcred->cmcred_gid; 1394 data = &linux_ucred; 1395 datalen = sizeof(linux_ucred); 1396 break; 1397 1398 case SCM_TIMESTAMP: 1399 if (datalen != sizeof(struct timeval)) { 1400 error = EMSGSIZE; 1401 goto bad; 1402 } 1403 ftmvl = (struct timeval *)data; 1404 ltmvl.tv_sec = ftmvl->tv_sec; 1405 ltmvl.tv_usec = ftmvl->tv_usec; 1406 data = <mvl; 1407 datalen = sizeof(ltmvl); 1408 break; 1409 } 1410 1411 if (outlen + LINUX_CMSG_LEN(datalen) > 1412 linux_msg.msg_controllen) { 1413 if (outlen == 0) { 1414 error = EMSGSIZE; 1415 goto bad; 1416 } else { 1417 linux_msg.msg_flags |= 1418 LINUX_MSG_CTRUNC; 1419 goto out; 1420 } 1421 } 1422 1423 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1424 1425 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1426 if (error) 1427 goto bad; 1428 outbuf += L_CMSG_HDRSZ; 1429 1430 error = copyout(data, outbuf, datalen); 1431 if (error) 1432 goto bad; 1433 1434 outbuf += LINUX_CMSG_ALIGN(datalen); 1435 outlen += LINUX_CMSG_LEN(datalen); 1436 1437 cm = CMSG_NXTHDR(msg, cm); 1438 } 1439 } 1440 1441out: 1442 linux_msg.msg_controllen = outlen; 1443 error = copyout(&linux_msg, msghdr, sizeof(linux_msg)); 1444 1445bad: 1446 free(iov, M_IOV); 1447 m_freem(control); 1448 free(linux_cmsg, M_LINUX); 1449 1450 return (error); 1451} 1452 1453int 1454linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1455{ 1456 struct msghdr bsd_msg; 1457 1458 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1459 args->flags, &bsd_msg)); 1460} 1461 1462int 1463linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1464{ 1465 struct l_mmsghdr *msg; 1466 struct msghdr bsd_msg; 1467 struct l_timespec lts; 1468 struct timespec ts, tts; 1469 l_uint retval; 1470 int error, datagrams; 1471 1472 if (args->timeout) { 1473 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1474 if (error != 0) 1475 return (error); 1476 error = linux_to_native_timespec(&ts, <s); 1477 if (error != 0) 1478 return (error); 1479 getnanotime(&tts); 1480 timespecadd(&tts, &ts); 1481 } 1482 1483 msg = PTRIN(args->msg); 1484 datagrams = 0; 1485 while (datagrams < args->vlen) { 1486 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1487 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1488 if (error != 0) 1489 break; 1490 1491 retval = td->td_retval[0]; 1492 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1493 if (error != 0) 1494 break; 1495 ++msg; 1496 ++datagrams; 1497 1498 /* 1499 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1500 */ 1501 if (args->flags & LINUX_MSG_WAITFORONE) 1502 args->flags |= LINUX_MSG_DONTWAIT; 1503 1504 /* 1505 * See BUGS section of recvmmsg(2). 1506 */ 1507 if (args->timeout) { 1508 getnanotime(&ts); 1509 timespecsub(&ts, &tts); 1510 if (!timespecisset(&ts) || ts.tv_sec > 0) 1511 break; 1512 } 1513 /* Out of band data, return right away. */ 1514 if (bsd_msg.msg_flags & MSG_OOB) 1515 break; 1516 } 1517 if (error == 0) 1518 td->td_retval[0] = datagrams; 1519 return (error); 1520} 1521 1522int 1523linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1524{ 1525 struct shutdown_args /* { 1526 int s; 1527 int how; 1528 } */ bsd_args; 1529 1530 bsd_args.s = args->s; 1531 bsd_args.how = args->how; 1532 return (sys_shutdown(td, &bsd_args)); 1533} 1534 1535int 1536linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1537{ 1538 struct setsockopt_args /* { 1539 int s; 1540 int level; 1541 int name; 1542 caddr_t val; 1543 int valsize; 1544 } */ bsd_args; 1545 l_timeval linux_tv; 1546 struct timeval tv; 1547 int error, name; 1548 1549 bsd_args.s = args->s; 1550 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1551 switch (bsd_args.level) { 1552 case SOL_SOCKET: 1553 name = linux_to_bsd_so_sockopt(args->optname); 1554 switch (name) { 1555 case SO_RCVTIMEO: 1556 /* FALLTHROUGH */ 1557 case SO_SNDTIMEO: 1558 error = copyin(PTRIN(args->optval), &linux_tv, 1559 sizeof(linux_tv)); 1560 if (error) 1561 return (error); 1562 tv.tv_sec = linux_tv.tv_sec; 1563 tv.tv_usec = linux_tv.tv_usec; 1564 return (kern_setsockopt(td, args->s, bsd_args.level, 1565 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1566 /* NOTREACHED */ 1567 break; 1568 default: 1569 break; 1570 } 1571 break; 1572 case IPPROTO_IP: 1573 name = linux_to_bsd_ip_sockopt(args->optname); 1574 break; 1575 case IPPROTO_IPV6: 1576 name = linux_to_bsd_ip6_sockopt(args->optname); 1577 break; 1578 case IPPROTO_TCP: 1579 name = linux_to_bsd_tcp_sockopt(args->optname); 1580 break; 1581 default: 1582 name = -1; 1583 break; 1584 } 1585 if (name == -1) 1586 return (ENOPROTOOPT); 1587 1588 bsd_args.name = name; 1589 bsd_args.val = PTRIN(args->optval); 1590 bsd_args.valsize = args->optlen; 1591 1592 if (name == IPV6_NEXTHOP) { 1593 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1594 bsd_args.valsize); 1595 error = sys_setsockopt(td, &bsd_args); 1596 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1597 } else 1598 error = sys_setsockopt(td, &bsd_args); 1599 1600 return (error); 1601} 1602 1603int 1604linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1605{ 1606 struct getsockopt_args /* { 1607 int s; 1608 int level; 1609 int name; 1610 caddr_t val; 1611 int *avalsize; 1612 } */ bsd_args; 1613 l_timeval linux_tv; 1614 struct timeval tv; 1615 socklen_t tv_len, xulen; 1616 struct xucred xu; 1617 struct l_ucred lxu; 1618 int error, name; 1619 1620 bsd_args.s = args->s; 1621 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1622 switch (bsd_args.level) { 1623 case SOL_SOCKET: 1624 name = linux_to_bsd_so_sockopt(args->optname); 1625 switch (name) { 1626 case SO_RCVTIMEO: 1627 /* FALLTHROUGH */ 1628 case SO_SNDTIMEO: 1629 tv_len = sizeof(tv); 1630 error = kern_getsockopt(td, args->s, bsd_args.level, 1631 name, &tv, UIO_SYSSPACE, &tv_len); 1632 if (error) 1633 return (error); 1634 linux_tv.tv_sec = tv.tv_sec; 1635 linux_tv.tv_usec = tv.tv_usec; 1636 return (copyout(&linux_tv, PTRIN(args->optval), 1637 sizeof(linux_tv))); 1638 /* NOTREACHED */ 1639 break; 1640 case LOCAL_PEERCRED: 1641 if (args->optlen != sizeof(lxu)) 1642 return (EINVAL); 1643 xulen = sizeof(xu); 1644 error = kern_getsockopt(td, args->s, bsd_args.level, 1645 name, &xu, UIO_SYSSPACE, &xulen); 1646 if (error) 1647 return (error); 1648 /* 1649 * XXX Use 0 for pid as the FreeBSD does not cache peer pid. 1650 */ 1651 lxu.pid = 0; 1652 lxu.uid = xu.cr_uid; 1653 lxu.gid = xu.cr_gid; 1654 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1655 /* NOTREACHED */ 1656 break; 1657 default: 1658 break; 1659 } 1660 break; 1661 case IPPROTO_IP: 1662 name = linux_to_bsd_ip_sockopt(args->optname); 1663 break; 1664 case IPPROTO_IPV6: 1665 name = linux_to_bsd_ip6_sockopt(args->optname); 1666 break; 1667 case IPPROTO_TCP: 1668 name = linux_to_bsd_tcp_sockopt(args->optname); 1669 break; 1670 default: 1671 name = -1; 1672 break; 1673 } 1674 if (name == -1) 1675 return (EINVAL); 1676 1677 bsd_args.name = name; 1678 bsd_args.val = PTRIN(args->optval); 1679 bsd_args.avalsize = PTRIN(args->optlen); 1680 1681 if (name == IPV6_NEXTHOP) { 1682 error = sys_getsockopt(td, &bsd_args); 1683 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1684 } else 1685 error = sys_getsockopt(td, &bsd_args); 1686 1687 return (error); 1688} 1689 1690#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1691 1692/* Argument list sizes for linux_socketcall */ 1693 1694#define LINUX_AL(x) ((x) * sizeof(l_ulong)) 1695 1696static const unsigned char lxs_args[] = { 1697 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */, 1698 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */, 1699 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */, 1700 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */, 1701 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */, 1702 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */, 1703 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */, 1704 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */, 1705 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */, 1706 LINUX_AL(4) /* accept4 */, LINUX_AL(5) /* recvmmsg */, 1707 LINUX_AL(4) /* sendmmsg */ 1708}; 1709 1710#define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1 1711 1712int 1713linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1714{ 1715 l_ulong a[6]; 1716 void *arg; 1717 int error; 1718 1719 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE) 1720 return (EINVAL); 1721 error = copyin(PTRIN(args->args), a, lxs_args[args->what]); 1722 if (error) 1723 return (error); 1724 1725 arg = a; 1726 switch (args->what) { 1727 case LINUX_SOCKET: 1728 return (linux_socket(td, arg)); 1729 case LINUX_BIND: 1730 return (linux_bind(td, arg)); 1731 case LINUX_CONNECT: 1732 return (linux_connect(td, arg)); 1733 case LINUX_LISTEN: 1734 return (linux_listen(td, arg)); 1735 case LINUX_ACCEPT: 1736 return (linux_accept(td, arg)); 1737 case LINUX_GETSOCKNAME: 1738 return (linux_getsockname(td, arg)); 1739 case LINUX_GETPEERNAME: 1740 return (linux_getpeername(td, arg)); 1741 case LINUX_SOCKETPAIR: 1742 return (linux_socketpair(td, arg)); 1743 case LINUX_SEND: 1744 return (linux_send(td, arg)); 1745 case LINUX_RECV: 1746 return (linux_recv(td, arg)); 1747 case LINUX_SENDTO: 1748 return (linux_sendto(td, arg)); 1749 case LINUX_RECVFROM: 1750 return (linux_recvfrom(td, arg)); 1751 case LINUX_SHUTDOWN: 1752 return (linux_shutdown(td, arg)); 1753 case LINUX_SETSOCKOPT: 1754 return (linux_setsockopt(td, arg)); 1755 case LINUX_GETSOCKOPT: 1756 return (linux_getsockopt(td, arg)); 1757 case LINUX_SENDMSG: 1758 return (linux_sendmsg(td, arg)); 1759 case LINUX_RECVMSG: 1760 return (linux_recvmsg(td, arg)); 1761 case LINUX_ACCEPT4: 1762 return (linux_accept4(td, arg)); 1763 case LINUX_RECVMMSG: 1764 return (linux_recvmmsg(td, arg)); 1765 case LINUX_SENDMMSG: 1766 return (linux_sendmmsg(td, arg)); 1767 } 1768 1769 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1770 return (ENOSYS); 1771} 1772#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1773