linux_socket.c revision 297518
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 297518 2016-04-03 06:22:36Z 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 return (copyout(&sa, arg, sa_len)); 472} 473 474static int 475linux_to_bsd_sockaddr(struct sockaddr *arg, int len) 476{ 477 struct sockaddr sa; 478 size_t sa_len = sizeof(struct sockaddr); 479 int error; 480 481 if ((error = copyin(arg, &sa, sa_len))) 482 return (error); 483 484 sa.sa_family = *(sa_family_t *)&sa; 485 sa.sa_len = len; 486 return (copyout(&sa, arg, sa_len)); 487} 488 489static int 490linux_sa_put(struct osockaddr *osa) 491{ 492 struct osockaddr sa; 493 int error, bdom; 494 495 /* 496 * Only read/write the osockaddr family part, the rest is 497 * not changed. 498 */ 499 error = copyin(osa, &sa, sizeof(sa.sa_family)); 500 if (error) 501 return (error); 502 503 bdom = bsd_to_linux_domain(sa.sa_family); 504 if (bdom == -1) 505 return (EINVAL); 506 507 sa.sa_family = bdom; 508 return (copyout(&sa, osa, sizeof(sa.sa_family))); 509} 510 511static int 512linux_to_bsd_cmsg_type(int cmsg_type) 513{ 514 515 switch (cmsg_type) { 516 case LINUX_SCM_RIGHTS: 517 return (SCM_RIGHTS); 518 case LINUX_SCM_CREDENTIALS: 519 return (SCM_CREDS); 520 } 521 return (-1); 522} 523 524static int 525bsd_to_linux_cmsg_type(int cmsg_type) 526{ 527 528 switch (cmsg_type) { 529 case SCM_RIGHTS: 530 return (LINUX_SCM_RIGHTS); 531 case SCM_CREDS: 532 return (LINUX_SCM_CREDENTIALS); 533 case SCM_TIMESTAMP: 534 return (LINUX_SCM_TIMESTAMP); 535 } 536 return (-1); 537} 538 539static int 540linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 541{ 542 if (lhdr->msg_controllen > INT_MAX) 543 return (ENOBUFS); 544 545 bhdr->msg_name = PTRIN(lhdr->msg_name); 546 bhdr->msg_namelen = lhdr->msg_namelen; 547 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 548 bhdr->msg_iovlen = lhdr->msg_iovlen; 549 bhdr->msg_control = PTRIN(lhdr->msg_control); 550 551 /* 552 * msg_controllen is skipped since BSD and LINUX control messages 553 * are potentially different sizes (e.g. the cred structure used 554 * by SCM_CREDS is different between the two operating system). 555 * 556 * The caller can set it (if necessary) after converting all the 557 * control messages. 558 */ 559 560 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 561 return (0); 562} 563 564static int 565bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 566{ 567 lhdr->msg_name = PTROUT(bhdr->msg_name); 568 lhdr->msg_namelen = bhdr->msg_namelen; 569 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 570 lhdr->msg_iovlen = bhdr->msg_iovlen; 571 lhdr->msg_control = PTROUT(bhdr->msg_control); 572 573 /* 574 * msg_controllen is skipped since BSD and LINUX control messages 575 * are potentially different sizes (e.g. the cred structure used 576 * by SCM_CREDS is different between the two operating system). 577 * 578 * The caller can set it (if necessary) after converting all the 579 * control messages. 580 */ 581 582 /* msg_flags skipped */ 583 return (0); 584} 585 586static int 587linux_set_socket_flags(int lflags, int *flags) 588{ 589 590 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 591 return (EINVAL); 592 if (lflags & LINUX_SOCK_NONBLOCK) 593 *flags |= SOCK_NONBLOCK; 594 if (lflags & LINUX_SOCK_CLOEXEC) 595 *flags |= SOCK_CLOEXEC; 596 return (0); 597} 598 599static int 600linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 601 struct mbuf *control, enum uio_seg segflg) 602{ 603 struct sockaddr *to; 604 int error; 605 606 if (mp->msg_name != NULL) { 607 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen); 608 if (error) 609 return (error); 610 mp->msg_name = to; 611 } else 612 to = NULL; 613 614 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 615 segflg); 616 617 if (to) 618 free(to, M_SONAME); 619 return (error); 620} 621 622/* Return 0 if IP_HDRINCL is set for the given socket. */ 623static int 624linux_check_hdrincl(struct thread *td, int s) 625{ 626 int error, optval; 627 socklen_t size_val; 628 629 size_val = sizeof(optval); 630 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 631 &optval, UIO_SYSSPACE, &size_val); 632 if (error) 633 return (error); 634 635 return (optval == 0); 636} 637 638/* 639 * Updated sendto() when IP_HDRINCL is set: 640 * tweak endian-dependent fields in the IP packet. 641 */ 642static int 643linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 644{ 645/* 646 * linux_ip_copysize defines how many bytes we should copy 647 * from the beginning of the IP packet before we customize it for BSD. 648 * It should include all the fields we modify (ip_len and ip_off). 649 */ 650#define linux_ip_copysize 8 651 652 struct ip *packet; 653 struct msghdr msg; 654 struct iovec aiov[1]; 655 int error; 656 657 /* Check that the packet isn't too big or too small. */ 658 if (linux_args->len < linux_ip_copysize || 659 linux_args->len > IP_MAXPACKET) 660 return (EINVAL); 661 662 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK); 663 664 /* Make kernel copy of the packet to be sent */ 665 if ((error = copyin(PTRIN(linux_args->msg), packet, 666 linux_args->len))) 667 goto goout; 668 669 /* Convert fields from Linux to BSD raw IP socket format */ 670 packet->ip_len = linux_args->len; 671 packet->ip_off = ntohs(packet->ip_off); 672 673 /* Prepare the msghdr and iovec structures describing the new packet */ 674 msg.msg_name = PTRIN(linux_args->to); 675 msg.msg_namelen = linux_args->tolen; 676 msg.msg_iov = aiov; 677 msg.msg_iovlen = 1; 678 msg.msg_control = NULL; 679 msg.msg_flags = 0; 680 aiov[0].iov_base = (char *)packet; 681 aiov[0].iov_len = linux_args->len; 682 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 683 NULL, UIO_SYSSPACE); 684goout: 685 free(packet, M_LINUX); 686 return (error); 687} 688 689int 690linux_socket(struct thread *td, struct linux_socket_args *args) 691{ 692 struct socket_args /* { 693 int domain; 694 int type; 695 int protocol; 696 } */ bsd_args; 697 int retval_socket; 698 699 bsd_args.protocol = args->protocol; 700 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 701 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 702 return (EINVAL); 703 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 704 &bsd_args.type); 705 if (retval_socket != 0) 706 return (retval_socket); 707 bsd_args.domain = linux_to_bsd_domain(args->domain); 708 if (bsd_args.domain == -1) 709 return (EAFNOSUPPORT); 710 711 retval_socket = sys_socket(td, &bsd_args); 712 if (retval_socket) 713 return (retval_socket); 714 715 if (bsd_args.type == SOCK_RAW 716 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0) 717 && bsd_args.domain == PF_INET) { 718 /* It's a raw IP socket: set the IP_HDRINCL option. */ 719 int hdrincl; 720 721 hdrincl = 1; 722 /* We ignore any error returned by kern_setsockopt() */ 723 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 724 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 725 } 726#ifdef INET6 727 /* 728 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default 729 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps. 730 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only 731 * sysctl value. 732 */ 733 if (bsd_args.domain == PF_INET6) { 734 int v6only; 735 736 v6only = 0; 737 /* We ignore any error returned by setsockopt() */ 738 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 739 &v6only, UIO_SYSSPACE, sizeof(v6only)); 740 } 741#endif 742 743 return (retval_socket); 744} 745 746int 747linux_bind(struct thread *td, struct linux_bind_args *args) 748{ 749 struct sockaddr *sa; 750 int error; 751 752 error = linux_getsockaddr(&sa, PTRIN(args->name), 753 args->namelen); 754 if (error) 755 return (error); 756 757 error = kern_bind(td, args->s, sa); 758 free(sa, M_SONAME); 759 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 760 return (EINVAL); 761 return (error); 762} 763 764int 765linux_connect(struct thread *td, struct linux_connect_args *args) 766{ 767 cap_rights_t rights; 768 struct socket *so; 769 struct sockaddr *sa; 770 u_int fflag; 771 int error; 772 773 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 774 args->namelen); 775 if (error) 776 return (error); 777 778 error = kern_connect(td, args->s, sa); 779 free(sa, M_SONAME); 780 if (error != EISCONN) 781 return (error); 782 783 /* 784 * Linux doesn't return EISCONN the first time it occurs, 785 * when on a non-blocking socket. Instead it returns the 786 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 787 * 788 * XXXRW: Instead of using fgetsock(), check that it is a 789 * socket and use the file descriptor reference instead of 790 * creating a new one. 791 */ 792 error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT), 793 &so, &fflag); 794 if (error == 0) { 795 error = EISCONN; 796 if (fflag & FNONBLOCK) { 797 SOCK_LOCK(so); 798 if (so->so_emuldata == 0) 799 error = so->so_error; 800 so->so_emuldata = (void *)1; 801 SOCK_UNLOCK(so); 802 } 803 fputsock(so); 804 } 805 return (error); 806} 807 808int 809linux_listen(struct thread *td, struct linux_listen_args *args) 810{ 811 struct listen_args /* { 812 int s; 813 int backlog; 814 } */ bsd_args; 815 816 bsd_args.s = args->s; 817 bsd_args.backlog = args->backlog; 818 return (sys_listen(td, &bsd_args)); 819} 820 821static int 822linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 823 l_uintptr_t namelen, int flags) 824{ 825 struct accept4_args /* { 826 int s; 827 struct sockaddr * __restrict name; 828 socklen_t * __restrict anamelen; 829 int flags; 830 } */ bsd_args; 831 int error; 832 833 bsd_args.s = s; 834 /* XXX: */ 835 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr); 836 bsd_args.anamelen = PTRIN(namelen);/* XXX */ 837 bsd_args.flags = 0; 838 error = linux_set_socket_flags(flags, &bsd_args.flags); 839 if (error != 0) 840 return (error); 841 error = sys_accept4(td, &bsd_args); 842 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 843 if (error) { 844 if (error == EFAULT && namelen != sizeof(struct sockaddr_in)) 845 return (EINVAL); 846 return (error); 847 } 848 if (addr) 849 error = linux_sa_put(PTRIN(addr)); 850 if (error) { 851 (void)kern_close(td, td->td_retval[0]); 852 td->td_retval[0] = 0; 853 } 854 return (error); 855} 856 857int 858linux_accept(struct thread *td, struct linux_accept_args *args) 859{ 860 861 return (linux_accept_common(td, args->s, args->addr, 862 args->namelen, 0)); 863} 864 865int 866linux_accept4(struct thread *td, struct linux_accept4_args *args) 867{ 868 869 return (linux_accept_common(td, args->s, args->addr, 870 args->namelen, args->flags)); 871} 872 873int 874linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 875{ 876 struct getsockname_args /* { 877 int fdes; 878 struct sockaddr * __restrict asa; 879 socklen_t * __restrict alen; 880 } */ bsd_args; 881 int error; 882 883 bsd_args.fdes = args->s; 884 /* XXX: */ 885 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 886 bsd_args.alen = PTRIN(args->namelen); /* XXX */ 887 error = sys_getsockname(td, &bsd_args); 888 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 889 if (error) 890 return (error); 891 return (linux_sa_put(PTRIN(args->addr))); 892} 893 894int 895linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 896{ 897 struct getpeername_args /* { 898 int fdes; 899 caddr_t asa; 900 int *alen; 901 } */ bsd_args; 902 int error; 903 904 bsd_args.fdes = args->s; 905 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 906 bsd_args.alen = (socklen_t *)PTRIN(args->namelen); 907 error = sys_getpeername(td, &bsd_args); 908 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 909 if (error) 910 return (error); 911 return (linux_sa_put(PTRIN(args->addr))); 912} 913 914int 915linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 916{ 917 struct socketpair_args /* { 918 int domain; 919 int type; 920 int protocol; 921 int *rsv; 922 } */ bsd_args; 923 int error; 924 925 bsd_args.domain = linux_to_bsd_domain(args->domain); 926 if (bsd_args.domain != PF_LOCAL) 927 return (EAFNOSUPPORT); 928 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 929 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 930 return (EINVAL); 931 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 932 &bsd_args.type); 933 if (error != 0) 934 return (error); 935 if (args->protocol != 0 && args->protocol != PF_UNIX) 936 937 /* 938 * Use of PF_UNIX as protocol argument is not right, 939 * but Linux does it. 940 * Do not map PF_UNIX as its Linux value is identical 941 * to FreeBSD one. 942 */ 943 return (EPROTONOSUPPORT); 944 else 945 bsd_args.protocol = 0; 946 bsd_args.rsv = (int *)PTRIN(args->rsv); 947 return (sys_socketpair(td, &bsd_args)); 948} 949 950#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 951struct linux_send_args { 952 int s; 953 l_uintptr_t msg; 954 int len; 955 int flags; 956}; 957 958static int 959linux_send(struct thread *td, struct linux_send_args *args) 960{ 961 struct sendto_args /* { 962 int s; 963 caddr_t buf; 964 int len; 965 int flags; 966 caddr_t to; 967 int tolen; 968 } */ bsd_args; 969 970 bsd_args.s = args->s; 971 bsd_args.buf = (caddr_t)PTRIN(args->msg); 972 bsd_args.len = args->len; 973 bsd_args.flags = args->flags; 974 bsd_args.to = NULL; 975 bsd_args.tolen = 0; 976 return (sys_sendto(td, &bsd_args)); 977} 978 979struct linux_recv_args { 980 int s; 981 l_uintptr_t msg; 982 int len; 983 int flags; 984}; 985 986static int 987linux_recv(struct thread *td, struct linux_recv_args *args) 988{ 989 struct recvfrom_args /* { 990 int s; 991 caddr_t buf; 992 int len; 993 int flags; 994 struct sockaddr *from; 995 socklen_t fromlenaddr; 996 } */ bsd_args; 997 998 bsd_args.s = args->s; 999 bsd_args.buf = (caddr_t)PTRIN(args->msg); 1000 bsd_args.len = args->len; 1001 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 1002 bsd_args.from = NULL; 1003 bsd_args.fromlenaddr = 0; 1004 return (sys_recvfrom(td, &bsd_args)); 1005} 1006#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1007 1008int 1009linux_sendto(struct thread *td, struct linux_sendto_args *args) 1010{ 1011 struct msghdr msg; 1012 struct iovec aiov; 1013 1014 if (linux_check_hdrincl(td, args->s) == 0) 1015 /* IP_HDRINCL set, tweak the packet before sending */ 1016 return (linux_sendto_hdrincl(td, args)); 1017 1018 msg.msg_name = PTRIN(args->to); 1019 msg.msg_namelen = args->tolen; 1020 msg.msg_iov = &aiov; 1021 msg.msg_iovlen = 1; 1022 msg.msg_control = NULL; 1023 msg.msg_flags = 0; 1024 aiov.iov_base = PTRIN(args->msg); 1025 aiov.iov_len = args->len; 1026 return (linux_sendit(td, args->s, &msg, args->flags, NULL, 1027 UIO_USERSPACE)); 1028} 1029 1030int 1031linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 1032{ 1033 struct msghdr msg; 1034 struct iovec aiov; 1035 int error; 1036 1037 if (PTRIN(args->fromlen) != NULL) { 1038 error = copyin(PTRIN(args->fromlen), &msg.msg_namelen, 1039 sizeof(msg.msg_namelen)); 1040 if (error != 0) 1041 return (error); 1042 1043 error = linux_to_bsd_sockaddr((struct sockaddr *)PTRIN(args->from), 1044 msg.msg_namelen); 1045 if (error != 0) 1046 return (error); 1047 } else 1048 msg.msg_namelen = 0; 1049 1050 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from); 1051 msg.msg_iov = &aiov; 1052 msg.msg_iovlen = 1; 1053 aiov.iov_base = PTRIN(args->buf); 1054 aiov.iov_len = args->len; 1055 msg.msg_control = 0; 1056 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 1057 1058 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL); 1059 if (error != 0) 1060 return (error); 1061 1062 if (PTRIN(args->from) != NULL) { 1063 error = bsd_to_linux_sockaddr((struct sockaddr *) 1064 PTRIN(args->from)); 1065 if (error != 0) 1066 return (error); 1067 1068 error = linux_sa_put((struct osockaddr *) 1069 PTRIN(args->from)); 1070 } 1071 1072 if (PTRIN(args->fromlen) != NULL) 1073 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 1074 sizeof(msg.msg_namelen)); 1075 1076 return (error); 1077} 1078 1079static int 1080linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1081 l_uint flags) 1082{ 1083 struct cmsghdr *cmsg; 1084 struct cmsgcred cmcred; 1085 struct mbuf *control; 1086 struct msghdr msg; 1087 struct l_cmsghdr linux_cmsg; 1088 struct l_cmsghdr *ptr_cmsg; 1089 struct l_msghdr linux_msg; 1090 struct iovec *iov; 1091 socklen_t datalen; 1092 struct sockaddr *sa; 1093 sa_family_t sa_family; 1094 void *data; 1095 int error; 1096 1097 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1098 if (error != 0) 1099 return (error); 1100 1101 /* 1102 * Some Linux applications (ping) define a non-NULL control data 1103 * pointer, but a msg_controllen of 0, which is not allowed in the 1104 * FreeBSD system call interface. NULL the msg_control pointer in 1105 * order to handle this case. This should be checked, but allows the 1106 * Linux ping to work. 1107 */ 1108 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0) 1109 linux_msg.msg_control = PTROUT(NULL); 1110 1111 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1112 if (error != 0) 1113 return (error); 1114 1115#ifdef COMPAT_LINUX32 1116 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1117 &iov, EMSGSIZE); 1118#else 1119 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1120#endif 1121 if (error != 0) 1122 return (error); 1123 1124 control = NULL; 1125 cmsg = NULL; 1126 1127 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) { 1128 error = kern_getsockname(td, s, &sa, &datalen); 1129 if (error != 0) 1130 goto bad; 1131 sa_family = sa->sa_family; 1132 free(sa, M_SONAME); 1133 1134 error = ENOBUFS; 1135 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO); 1136 control = m_get(M_WAITOK, MT_CONTROL); 1137 1138 do { 1139 error = copyin(ptr_cmsg, &linux_cmsg, 1140 sizeof(struct l_cmsghdr)); 1141 if (error != 0) 1142 goto bad; 1143 1144 error = EINVAL; 1145 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1146 goto bad; 1147 1148 /* 1149 * Now we support only SCM_RIGHTS and SCM_CRED, 1150 * so return EINVAL in any other cmsg_type 1151 */ 1152 cmsg->cmsg_type = 1153 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1154 cmsg->cmsg_level = 1155 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1156 if (cmsg->cmsg_type == -1 1157 || cmsg->cmsg_level != SOL_SOCKET) 1158 goto bad; 1159 1160 /* 1161 * Some applications (e.g. pulseaudio) attempt to 1162 * send ancillary data even if the underlying protocol 1163 * doesn't support it which is not allowed in the 1164 * FreeBSD system call interface. 1165 */ 1166 if (sa_family != AF_UNIX) 1167 continue; 1168 1169 data = LINUX_CMSG_DATA(ptr_cmsg); 1170 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1171 1172 switch (cmsg->cmsg_type) 1173 { 1174 case SCM_RIGHTS: 1175 break; 1176 1177 case SCM_CREDS: 1178 data = &cmcred; 1179 datalen = sizeof(cmcred); 1180 1181 /* 1182 * The lower levels will fill in the structure 1183 */ 1184 bzero(data, datalen); 1185 break; 1186 } 1187 1188 cmsg->cmsg_len = CMSG_LEN(datalen); 1189 1190 error = ENOBUFS; 1191 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg)) 1192 goto bad; 1193 if (!m_append(control, datalen, (c_caddr_t)data)) 1194 goto bad; 1195 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg))); 1196 1197 if (m_length(control, NULL) == 0) { 1198 m_freem(control); 1199 control = NULL; 1200 } 1201 } 1202 1203 msg.msg_iov = iov; 1204 msg.msg_flags = 0; 1205 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1206 control = NULL; 1207 1208bad: 1209 m_freem(control); 1210 free(iov, M_IOV); 1211 if (cmsg) 1212 free(cmsg, M_LINUX); 1213 return (error); 1214} 1215 1216int 1217linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1218{ 1219 1220 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1221 args->flags)); 1222} 1223 1224int 1225linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1226{ 1227 struct l_mmsghdr *msg; 1228 l_uint retval; 1229 int error, datagrams; 1230 1231 if (args->vlen > UIO_MAXIOV) 1232 args->vlen = UIO_MAXIOV; 1233 1234 msg = PTRIN(args->msg); 1235 datagrams = 0; 1236 while (datagrams < args->vlen) { 1237 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1238 args->flags); 1239 if (error != 0) 1240 break; 1241 1242 retval = td->td_retval[0]; 1243 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1244 if (error != 0) 1245 break; 1246 ++msg; 1247 ++datagrams; 1248 } 1249 if (error == 0) 1250 td->td_retval[0] = datagrams; 1251 return (error); 1252} 1253 1254static int 1255linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1256 l_uint flags, struct msghdr *msg) 1257{ 1258 struct cmsghdr *cm; 1259 struct cmsgcred *cmcred; 1260 struct l_cmsghdr *linux_cmsg = NULL; 1261 struct l_ucred linux_ucred; 1262 socklen_t datalen, outlen; 1263 struct l_msghdr linux_msg; 1264 struct iovec *iov, *uiov; 1265 struct mbuf *control = NULL; 1266 struct mbuf **controlp; 1267 struct timeval *ftmvl; 1268 l_timeval ltmvl; 1269 caddr_t outbuf; 1270 void *data; 1271 int error, i, fd, fds, *fdp; 1272 1273 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1274 if (error != 0) 1275 return (error); 1276 1277 error = linux_to_bsd_msghdr(msg, &linux_msg); 1278 if (error != 0) 1279 return (error); 1280 1281#ifdef COMPAT_LINUX32 1282 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1283 &iov, EMSGSIZE); 1284#else 1285 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1286#endif 1287 if (error != 0) 1288 return (error); 1289 1290 if (msg->msg_name) { 1291 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name, 1292 msg->msg_namelen); 1293 if (error != 0) 1294 goto bad; 1295 } 1296 1297 uiov = msg->msg_iov; 1298 msg->msg_iov = iov; 1299 controlp = (msg->msg_control != NULL) ? &control : NULL; 1300 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp); 1301 msg->msg_iov = uiov; 1302 if (error != 0) 1303 goto bad; 1304 1305 error = bsd_to_linux_msghdr(msg, &linux_msg); 1306 if (error != 0) 1307 goto bad; 1308 1309 if (linux_msg.msg_name) { 1310 error = bsd_to_linux_sockaddr((struct sockaddr *) 1311 PTRIN(linux_msg.msg_name)); 1312 if (error != 0) 1313 goto bad; 1314 } 1315 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1316 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1317 if (error != 0) 1318 goto bad; 1319 } 1320 1321 outbuf = PTRIN(linux_msg.msg_control); 1322 outlen = 0; 1323 1324 if (control) { 1325 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1326 1327 msg->msg_control = mtod(control, struct cmsghdr *); 1328 msg->msg_controllen = control->m_len; 1329 1330 cm = CMSG_FIRSTHDR(msg); 1331 1332 while (cm != NULL) { 1333 linux_cmsg->cmsg_type = 1334 bsd_to_linux_cmsg_type(cm->cmsg_type); 1335 linux_cmsg->cmsg_level = 1336 bsd_to_linux_sockopt_level(cm->cmsg_level); 1337 if (linux_cmsg->cmsg_type == -1 1338 || cm->cmsg_level != SOL_SOCKET) 1339 { 1340 error = EINVAL; 1341 goto bad; 1342 } 1343 1344 data = CMSG_DATA(cm); 1345 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1346 1347 switch (cm->cmsg_type) 1348 { 1349 case SCM_RIGHTS: 1350 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1351 fds = datalen / sizeof(int); 1352 fdp = data; 1353 for (i = 0; i < fds; i++) { 1354 fd = *fdp++; 1355 (void)kern_fcntl(td, fd, 1356 F_SETFD, FD_CLOEXEC); 1357 } 1358 } 1359 break; 1360 1361 case SCM_CREDS: 1362 /* 1363 * Currently LOCAL_CREDS is never in 1364 * effect for Linux so no need to worry 1365 * about sockcred 1366 */ 1367 if (datalen != sizeof(*cmcred)) { 1368 error = EMSGSIZE; 1369 goto bad; 1370 } 1371 cmcred = (struct cmsgcred *)data; 1372 bzero(&linux_ucred, sizeof(linux_ucred)); 1373 linux_ucred.pid = cmcred->cmcred_pid; 1374 linux_ucred.uid = cmcred->cmcred_uid; 1375 linux_ucred.gid = cmcred->cmcred_gid; 1376 data = &linux_ucred; 1377 datalen = sizeof(linux_ucred); 1378 break; 1379 1380 case SCM_TIMESTAMP: 1381 if (datalen != sizeof(struct timeval)) { 1382 error = EMSGSIZE; 1383 goto bad; 1384 } 1385 ftmvl = (struct timeval *)data; 1386 ltmvl.tv_sec = ftmvl->tv_sec; 1387 ltmvl.tv_usec = ftmvl->tv_usec; 1388 data = <mvl; 1389 datalen = sizeof(ltmvl); 1390 break; 1391 } 1392 1393 if (outlen + LINUX_CMSG_LEN(datalen) > 1394 linux_msg.msg_controllen) { 1395 if (outlen == 0) { 1396 error = EMSGSIZE; 1397 goto bad; 1398 } else { 1399 linux_msg.msg_flags |= 1400 LINUX_MSG_CTRUNC; 1401 goto out; 1402 } 1403 } 1404 1405 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1406 1407 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1408 if (error) 1409 goto bad; 1410 outbuf += L_CMSG_HDRSZ; 1411 1412 error = copyout(data, outbuf, datalen); 1413 if (error) 1414 goto bad; 1415 1416 outbuf += LINUX_CMSG_ALIGN(datalen); 1417 outlen += LINUX_CMSG_LEN(datalen); 1418 1419 cm = CMSG_NXTHDR(msg, cm); 1420 } 1421 } 1422 1423out: 1424 linux_msg.msg_controllen = outlen; 1425 error = copyout(&linux_msg, msghdr, sizeof(linux_msg)); 1426 1427bad: 1428 free(iov, M_IOV); 1429 m_freem(control); 1430 free(linux_cmsg, M_LINUX); 1431 1432 return (error); 1433} 1434 1435int 1436linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1437{ 1438 struct msghdr bsd_msg; 1439 1440 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1441 args->flags, &bsd_msg)); 1442} 1443 1444int 1445linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1446{ 1447 struct l_mmsghdr *msg; 1448 struct msghdr bsd_msg; 1449 struct l_timespec lts; 1450 struct timespec ts, tts; 1451 l_uint retval; 1452 int error, datagrams; 1453 1454 if (args->timeout) { 1455 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1456 if (error != 0) 1457 return (error); 1458 error = linux_to_native_timespec(&ts, <s); 1459 if (error != 0) 1460 return (error); 1461 getnanotime(&tts); 1462 timespecadd(&tts, &ts); 1463 } 1464 1465 msg = PTRIN(args->msg); 1466 datagrams = 0; 1467 while (datagrams < args->vlen) { 1468 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1469 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1470 if (error != 0) 1471 break; 1472 1473 retval = td->td_retval[0]; 1474 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1475 if (error != 0) 1476 break; 1477 ++msg; 1478 ++datagrams; 1479 1480 /* 1481 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1482 */ 1483 if (args->flags & LINUX_MSG_WAITFORONE) 1484 args->flags |= LINUX_MSG_DONTWAIT; 1485 1486 /* 1487 * See BUGS section of recvmmsg(2). 1488 */ 1489 if (args->timeout) { 1490 getnanotime(&ts); 1491 timespecsub(&ts, &tts); 1492 if (!timespecisset(&ts) || ts.tv_sec > 0) 1493 break; 1494 } 1495 /* Out of band data, return right away. */ 1496 if (bsd_msg.msg_flags & MSG_OOB) 1497 break; 1498 } 1499 if (error == 0) 1500 td->td_retval[0] = datagrams; 1501 return (error); 1502} 1503 1504int 1505linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1506{ 1507 struct shutdown_args /* { 1508 int s; 1509 int how; 1510 } */ bsd_args; 1511 1512 bsd_args.s = args->s; 1513 bsd_args.how = args->how; 1514 return (sys_shutdown(td, &bsd_args)); 1515} 1516 1517int 1518linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1519{ 1520 struct setsockopt_args /* { 1521 int s; 1522 int level; 1523 int name; 1524 caddr_t val; 1525 int valsize; 1526 } */ bsd_args; 1527 l_timeval linux_tv; 1528 struct timeval tv; 1529 int error, name; 1530 1531 bsd_args.s = args->s; 1532 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1533 switch (bsd_args.level) { 1534 case SOL_SOCKET: 1535 name = linux_to_bsd_so_sockopt(args->optname); 1536 switch (name) { 1537 case SO_RCVTIMEO: 1538 /* FALLTHROUGH */ 1539 case SO_SNDTIMEO: 1540 error = copyin(PTRIN(args->optval), &linux_tv, 1541 sizeof(linux_tv)); 1542 if (error) 1543 return (error); 1544 tv.tv_sec = linux_tv.tv_sec; 1545 tv.tv_usec = linux_tv.tv_usec; 1546 return (kern_setsockopt(td, args->s, bsd_args.level, 1547 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1548 /* NOTREACHED */ 1549 break; 1550 default: 1551 break; 1552 } 1553 break; 1554 case IPPROTO_IP: 1555 name = linux_to_bsd_ip_sockopt(args->optname); 1556 break; 1557 case IPPROTO_IPV6: 1558 name = linux_to_bsd_ip6_sockopt(args->optname); 1559 break; 1560 case IPPROTO_TCP: 1561 name = linux_to_bsd_tcp_sockopt(args->optname); 1562 break; 1563 default: 1564 name = -1; 1565 break; 1566 } 1567 if (name == -1) 1568 return (ENOPROTOOPT); 1569 1570 bsd_args.name = name; 1571 bsd_args.val = PTRIN(args->optval); 1572 bsd_args.valsize = args->optlen; 1573 1574 if (name == IPV6_NEXTHOP) { 1575 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1576 bsd_args.valsize); 1577 error = sys_setsockopt(td, &bsd_args); 1578 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1579 } else 1580 error = sys_setsockopt(td, &bsd_args); 1581 1582 return (error); 1583} 1584 1585int 1586linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1587{ 1588 struct getsockopt_args /* { 1589 int s; 1590 int level; 1591 int name; 1592 caddr_t val; 1593 int *avalsize; 1594 } */ bsd_args; 1595 l_timeval linux_tv; 1596 struct timeval tv; 1597 socklen_t tv_len, xulen; 1598 struct xucred xu; 1599 struct l_ucred lxu; 1600 int error, name; 1601 1602 bsd_args.s = args->s; 1603 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1604 switch (bsd_args.level) { 1605 case SOL_SOCKET: 1606 name = linux_to_bsd_so_sockopt(args->optname); 1607 switch (name) { 1608 case SO_RCVTIMEO: 1609 /* FALLTHROUGH */ 1610 case SO_SNDTIMEO: 1611 tv_len = sizeof(tv); 1612 error = kern_getsockopt(td, args->s, bsd_args.level, 1613 name, &tv, UIO_SYSSPACE, &tv_len); 1614 if (error) 1615 return (error); 1616 linux_tv.tv_sec = tv.tv_sec; 1617 linux_tv.tv_usec = tv.tv_usec; 1618 return (copyout(&linux_tv, PTRIN(args->optval), 1619 sizeof(linux_tv))); 1620 /* NOTREACHED */ 1621 break; 1622 case LOCAL_PEERCRED: 1623 if (args->optlen != sizeof(lxu)) 1624 return (EINVAL); 1625 xulen = sizeof(xu); 1626 error = kern_getsockopt(td, args->s, bsd_args.level, 1627 name, &xu, UIO_SYSSPACE, &xulen); 1628 if (error) 1629 return (error); 1630 /* 1631 * XXX Use 0 for pid as the FreeBSD does not cache peer pid. 1632 */ 1633 lxu.pid = 0; 1634 lxu.uid = xu.cr_uid; 1635 lxu.gid = xu.cr_gid; 1636 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1637 /* NOTREACHED */ 1638 break; 1639 default: 1640 break; 1641 } 1642 break; 1643 case IPPROTO_IP: 1644 name = linux_to_bsd_ip_sockopt(args->optname); 1645 break; 1646 case IPPROTO_IPV6: 1647 name = linux_to_bsd_ip6_sockopt(args->optname); 1648 break; 1649 case IPPROTO_TCP: 1650 name = linux_to_bsd_tcp_sockopt(args->optname); 1651 break; 1652 default: 1653 name = -1; 1654 break; 1655 } 1656 if (name == -1) 1657 return (EINVAL); 1658 1659 bsd_args.name = name; 1660 bsd_args.val = PTRIN(args->optval); 1661 bsd_args.avalsize = PTRIN(args->optlen); 1662 1663 if (name == IPV6_NEXTHOP) { 1664 error = sys_getsockopt(td, &bsd_args); 1665 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1666 } else 1667 error = sys_getsockopt(td, &bsd_args); 1668 1669 return (error); 1670} 1671 1672#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1673 1674/* Argument list sizes for linux_socketcall */ 1675 1676#define LINUX_AL(x) ((x) * sizeof(l_ulong)) 1677 1678static const unsigned char lxs_args[] = { 1679 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */, 1680 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */, 1681 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */, 1682 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */, 1683 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */, 1684 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */, 1685 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */, 1686 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */, 1687 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */, 1688 LINUX_AL(4) /* accept4 */, LINUX_AL(5) /* recvmmsg */, 1689 LINUX_AL(4) /* sendmmsg */ 1690}; 1691 1692#define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1 1693 1694int 1695linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1696{ 1697 l_ulong a[6]; 1698 void *arg; 1699 int error; 1700 1701 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE) 1702 return (EINVAL); 1703 error = copyin(PTRIN(args->args), a, lxs_args[args->what]); 1704 if (error) 1705 return (error); 1706 1707 arg = a; 1708 switch (args->what) { 1709 case LINUX_SOCKET: 1710 return (linux_socket(td, arg)); 1711 case LINUX_BIND: 1712 return (linux_bind(td, arg)); 1713 case LINUX_CONNECT: 1714 return (linux_connect(td, arg)); 1715 case LINUX_LISTEN: 1716 return (linux_listen(td, arg)); 1717 case LINUX_ACCEPT: 1718 return (linux_accept(td, arg)); 1719 case LINUX_GETSOCKNAME: 1720 return (linux_getsockname(td, arg)); 1721 case LINUX_GETPEERNAME: 1722 return (linux_getpeername(td, arg)); 1723 case LINUX_SOCKETPAIR: 1724 return (linux_socketpair(td, arg)); 1725 case LINUX_SEND: 1726 return (linux_send(td, arg)); 1727 case LINUX_RECV: 1728 return (linux_recv(td, arg)); 1729 case LINUX_SENDTO: 1730 return (linux_sendto(td, arg)); 1731 case LINUX_RECVFROM: 1732 return (linux_recvfrom(td, arg)); 1733 case LINUX_SHUTDOWN: 1734 return (linux_shutdown(td, arg)); 1735 case LINUX_SETSOCKOPT: 1736 return (linux_setsockopt(td, arg)); 1737 case LINUX_GETSOCKOPT: 1738 return (linux_getsockopt(td, arg)); 1739 case LINUX_SENDMSG: 1740 return (linux_sendmsg(td, arg)); 1741 case LINUX_RECVMSG: 1742 return (linux_recvmsg(td, arg)); 1743 case LINUX_ACCEPT4: 1744 return (linux_accept4(td, arg)); 1745 case LINUX_RECVMMSG: 1746 return (linux_recvmmsg(td, arg)); 1747 case LINUX_SENDMMSG: 1748 return (linux_sendmmsg(td, arg)); 1749 } 1750 1751 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1752 return (ENOSYS); 1753} 1754#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1755