linux_socket.c revision 301431
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 301431 2016-06-05 07:43:20Z 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, bdom; 466 467 if ((error = copyin(arg, &sa, sa_len))) 468 return (error); 469 470 bdom = bsd_to_linux_domain(sa.sa_family); 471 if (bdom == -1) 472 return (EAFNOSUPPORT); 473 474 *(u_short *)&sa = bdom; 475 return (copyout(&sa, arg, sa_len)); 476} 477 478static int 479linux_to_bsd_sockaddr(struct sockaddr *arg, int len) 480{ 481 struct sockaddr sa; 482 size_t sa_len = sizeof(struct sockaddr); 483 int error, bdom; 484 485 if ((error = copyin(arg, &sa, sa_len))) 486 return (error); 487 488 bdom = linux_to_bsd_domain(*(sa_family_t *)&sa); 489 if (bdom == -1) 490 return (EAFNOSUPPORT); 491 492 sa.sa_family = bdom; 493 sa.sa_len = len; 494 return (copyout(&sa, arg, sa_len)); 495} 496 497static int 498linux_sa_put(struct osockaddr *osa) 499{ 500 struct osockaddr sa; 501 int error, bdom; 502 503 /* 504 * Only read/write the osockaddr family part, the rest is 505 * not changed. 506 */ 507 error = copyin(osa, &sa, sizeof(sa.sa_family)); 508 if (error) 509 return (error); 510 511 bdom = bsd_to_linux_domain(sa.sa_family); 512 if (bdom == -1) 513 return (EINVAL); 514 515 sa.sa_family = bdom; 516 return (copyout(&sa, osa, sizeof(sa.sa_family))); 517} 518 519static int 520linux_to_bsd_cmsg_type(int cmsg_type) 521{ 522 523 switch (cmsg_type) { 524 case LINUX_SCM_RIGHTS: 525 return (SCM_RIGHTS); 526 case LINUX_SCM_CREDENTIALS: 527 return (SCM_CREDS); 528 } 529 return (-1); 530} 531 532static int 533bsd_to_linux_cmsg_type(int cmsg_type) 534{ 535 536 switch (cmsg_type) { 537 case SCM_RIGHTS: 538 return (LINUX_SCM_RIGHTS); 539 case SCM_CREDS: 540 return (LINUX_SCM_CREDENTIALS); 541 case SCM_TIMESTAMP: 542 return (LINUX_SCM_TIMESTAMP); 543 } 544 return (-1); 545} 546 547static int 548linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 549{ 550 if (lhdr->msg_controllen > INT_MAX) 551 return (ENOBUFS); 552 553 bhdr->msg_name = PTRIN(lhdr->msg_name); 554 bhdr->msg_namelen = lhdr->msg_namelen; 555 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 556 bhdr->msg_iovlen = lhdr->msg_iovlen; 557 bhdr->msg_control = PTRIN(lhdr->msg_control); 558 559 /* 560 * msg_controllen is skipped since BSD and LINUX control messages 561 * are potentially different sizes (e.g. the cred structure used 562 * by SCM_CREDS is different between the two operating system). 563 * 564 * The caller can set it (if necessary) after converting all the 565 * control messages. 566 */ 567 568 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 569 return (0); 570} 571 572static int 573bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 574{ 575 lhdr->msg_name = PTROUT(bhdr->msg_name); 576 lhdr->msg_namelen = bhdr->msg_namelen; 577 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 578 lhdr->msg_iovlen = bhdr->msg_iovlen; 579 lhdr->msg_control = PTROUT(bhdr->msg_control); 580 581 /* 582 * msg_controllen is skipped since BSD and LINUX control messages 583 * are potentially different sizes (e.g. the cred structure used 584 * by SCM_CREDS is different between the two operating system). 585 * 586 * The caller can set it (if necessary) after converting all the 587 * control messages. 588 */ 589 590 /* msg_flags skipped */ 591 return (0); 592} 593 594static int 595linux_set_socket_flags(int lflags, int *flags) 596{ 597 598 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 599 return (EINVAL); 600 if (lflags & LINUX_SOCK_NONBLOCK) 601 *flags |= SOCK_NONBLOCK; 602 if (lflags & LINUX_SOCK_CLOEXEC) 603 *flags |= SOCK_CLOEXEC; 604 return (0); 605} 606 607static int 608linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 609 struct mbuf *control, enum uio_seg segflg) 610{ 611 struct sockaddr *to; 612 int error; 613 614 if (mp->msg_name != NULL) { 615 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen); 616 if (error) 617 return (error); 618 mp->msg_name = to; 619 } else 620 to = NULL; 621 622 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 623 segflg); 624 625 if (to) 626 free(to, M_SONAME); 627 return (error); 628} 629 630/* Return 0 if IP_HDRINCL is set for the given socket. */ 631static int 632linux_check_hdrincl(struct thread *td, int s) 633{ 634 int error, optval; 635 socklen_t size_val; 636 637 size_val = sizeof(optval); 638 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 639 &optval, UIO_SYSSPACE, &size_val); 640 if (error) 641 return (error); 642 643 return (optval == 0); 644} 645 646/* 647 * Updated sendto() when IP_HDRINCL is set: 648 * tweak endian-dependent fields in the IP packet. 649 */ 650static int 651linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 652{ 653/* 654 * linux_ip_copysize defines how many bytes we should copy 655 * from the beginning of the IP packet before we customize it for BSD. 656 * It should include all the fields we modify (ip_len and ip_off). 657 */ 658#define linux_ip_copysize 8 659 660 struct ip *packet; 661 struct msghdr msg; 662 struct iovec aiov[1]; 663 int error; 664 665 /* Check that the packet isn't too big or too small. */ 666 if (linux_args->len < linux_ip_copysize || 667 linux_args->len > IP_MAXPACKET) 668 return (EINVAL); 669 670 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK); 671 672 /* Make kernel copy of the packet to be sent */ 673 if ((error = copyin(PTRIN(linux_args->msg), packet, 674 linux_args->len))) 675 goto goout; 676 677 /* Convert fields from Linux to BSD raw IP socket format */ 678 packet->ip_len = linux_args->len; 679 packet->ip_off = ntohs(packet->ip_off); 680 681 /* Prepare the msghdr and iovec structures describing the new packet */ 682 msg.msg_name = PTRIN(linux_args->to); 683 msg.msg_namelen = linux_args->tolen; 684 msg.msg_iov = aiov; 685 msg.msg_iovlen = 1; 686 msg.msg_control = NULL; 687 msg.msg_flags = 0; 688 aiov[0].iov_base = (char *)packet; 689 aiov[0].iov_len = linux_args->len; 690 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 691 NULL, UIO_SYSSPACE); 692goout: 693 free(packet, M_LINUX); 694 return (error); 695} 696 697int 698linux_socket(struct thread *td, struct linux_socket_args *args) 699{ 700 struct socket_args /* { 701 int domain; 702 int type; 703 int protocol; 704 } */ bsd_args; 705 int retval_socket; 706 707 bsd_args.protocol = args->protocol; 708 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 709 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 710 return (EINVAL); 711 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 712 &bsd_args.type); 713 if (retval_socket != 0) 714 return (retval_socket); 715 bsd_args.domain = linux_to_bsd_domain(args->domain); 716 if (bsd_args.domain == -1) 717 return (EAFNOSUPPORT); 718 719 retval_socket = sys_socket(td, &bsd_args); 720 if (retval_socket) 721 return (retval_socket); 722 723 if (bsd_args.type == SOCK_RAW 724 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0) 725 && bsd_args.domain == PF_INET) { 726 /* It's a raw IP socket: set the IP_HDRINCL option. */ 727 int hdrincl; 728 729 hdrincl = 1; 730 /* We ignore any error returned by kern_setsockopt() */ 731 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 732 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 733 } 734#ifdef INET6 735 /* 736 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default 737 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps. 738 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only 739 * sysctl value. 740 */ 741 if (bsd_args.domain == PF_INET6) { 742 int v6only; 743 744 v6only = 0; 745 /* We ignore any error returned by setsockopt() */ 746 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 747 &v6only, UIO_SYSSPACE, sizeof(v6only)); 748 } 749#endif 750 751 return (retval_socket); 752} 753 754int 755linux_bind(struct thread *td, struct linux_bind_args *args) 756{ 757 struct sockaddr *sa; 758 int error; 759 760 error = linux_getsockaddr(&sa, PTRIN(args->name), 761 args->namelen); 762 if (error) 763 return (error); 764 765 error = kern_bind(td, args->s, sa); 766 free(sa, M_SONAME); 767 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 768 return (EINVAL); 769 return (error); 770} 771 772int 773linux_connect(struct thread *td, struct linux_connect_args *args) 774{ 775 cap_rights_t rights; 776 struct socket *so; 777 struct sockaddr *sa; 778 u_int fflag; 779 int error; 780 781 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 782 args->namelen); 783 if (error) 784 return (error); 785 786 error = kern_connect(td, args->s, sa); 787 free(sa, M_SONAME); 788 if (error != EISCONN) 789 return (error); 790 791 /* 792 * Linux doesn't return EISCONN the first time it occurs, 793 * when on a non-blocking socket. Instead it returns the 794 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 795 * 796 * XXXRW: Instead of using fgetsock(), check that it is a 797 * socket and use the file descriptor reference instead of 798 * creating a new one. 799 */ 800 error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT), 801 &so, &fflag); 802 if (error == 0) { 803 error = EISCONN; 804 if (fflag & FNONBLOCK) { 805 SOCK_LOCK(so); 806 if (so->so_emuldata == 0) 807 error = so->so_error; 808 so->so_emuldata = (void *)1; 809 SOCK_UNLOCK(so); 810 } 811 fputsock(so); 812 } 813 return (error); 814} 815 816int 817linux_listen(struct thread *td, struct linux_listen_args *args) 818{ 819 struct listen_args /* { 820 int s; 821 int backlog; 822 } */ bsd_args; 823 824 bsd_args.s = args->s; 825 bsd_args.backlog = args->backlog; 826 return (sys_listen(td, &bsd_args)); 827} 828 829static int 830linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 831 l_uintptr_t namelen, int flags) 832{ 833 struct accept4_args /* { 834 int s; 835 struct sockaddr * __restrict name; 836 socklen_t * __restrict anamelen; 837 int flags; 838 } */ bsd_args; 839 int error; 840 841 bsd_args.s = s; 842 /* XXX: */ 843 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr); 844 bsd_args.anamelen = PTRIN(namelen);/* XXX */ 845 bsd_args.flags = 0; 846 error = linux_set_socket_flags(flags, &bsd_args.flags); 847 if (error != 0) 848 return (error); 849 error = sys_accept4(td, &bsd_args); 850 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 851 if (error) { 852 if (error == EFAULT && namelen != sizeof(struct sockaddr_in)) 853 return (EINVAL); 854 return (error); 855 } 856 if (addr) 857 error = linux_sa_put(PTRIN(addr)); 858 if (error) { 859 (void)kern_close(td, td->td_retval[0]); 860 td->td_retval[0] = 0; 861 } 862 return (error); 863} 864 865int 866linux_accept(struct thread *td, struct linux_accept_args *args) 867{ 868 869 return (linux_accept_common(td, args->s, args->addr, 870 args->namelen, 0)); 871} 872 873int 874linux_accept4(struct thread *td, struct linux_accept4_args *args) 875{ 876 877 return (linux_accept_common(td, args->s, args->addr, 878 args->namelen, args->flags)); 879} 880 881int 882linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 883{ 884 struct getsockname_args /* { 885 int fdes; 886 struct sockaddr * __restrict asa; 887 socklen_t * __restrict alen; 888 } */ bsd_args; 889 int error; 890 891 bsd_args.fdes = args->s; 892 /* XXX: */ 893 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 894 bsd_args.alen = PTRIN(args->namelen); /* XXX */ 895 error = sys_getsockname(td, &bsd_args); 896 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 897 if (error) 898 return (error); 899 return (linux_sa_put(PTRIN(args->addr))); 900} 901 902int 903linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 904{ 905 struct getpeername_args /* { 906 int fdes; 907 caddr_t asa; 908 int *alen; 909 } */ bsd_args; 910 int error; 911 912 bsd_args.fdes = args->s; 913 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 914 bsd_args.alen = (socklen_t *)PTRIN(args->namelen); 915 error = sys_getpeername(td, &bsd_args); 916 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 917 if (error) 918 return (error); 919 return (linux_sa_put(PTRIN(args->addr))); 920} 921 922int 923linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 924{ 925 struct socketpair_args /* { 926 int domain; 927 int type; 928 int protocol; 929 int *rsv; 930 } */ bsd_args; 931 int error; 932 933 bsd_args.domain = linux_to_bsd_domain(args->domain); 934 if (bsd_args.domain != PF_LOCAL) 935 return (EAFNOSUPPORT); 936 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 937 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 938 return (EINVAL); 939 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 940 &bsd_args.type); 941 if (error != 0) 942 return (error); 943 if (args->protocol != 0 && args->protocol != PF_UNIX) 944 945 /* 946 * Use of PF_UNIX as protocol argument is not right, 947 * but Linux does it. 948 * Do not map PF_UNIX as its Linux value is identical 949 * to FreeBSD one. 950 */ 951 return (EPROTONOSUPPORT); 952 else 953 bsd_args.protocol = 0; 954 bsd_args.rsv = (int *)PTRIN(args->rsv); 955 return (sys_socketpair(td, &bsd_args)); 956} 957 958#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 959struct linux_send_args { 960 int s; 961 l_uintptr_t msg; 962 int len; 963 int flags; 964}; 965 966static int 967linux_send(struct thread *td, struct linux_send_args *args) 968{ 969 struct sendto_args /* { 970 int s; 971 caddr_t buf; 972 int len; 973 int flags; 974 caddr_t to; 975 int tolen; 976 } */ bsd_args; 977 978 bsd_args.s = args->s; 979 bsd_args.buf = (caddr_t)PTRIN(args->msg); 980 bsd_args.len = args->len; 981 bsd_args.flags = args->flags; 982 bsd_args.to = NULL; 983 bsd_args.tolen = 0; 984 return (sys_sendto(td, &bsd_args)); 985} 986 987struct linux_recv_args { 988 int s; 989 l_uintptr_t msg; 990 int len; 991 int flags; 992}; 993 994static int 995linux_recv(struct thread *td, struct linux_recv_args *args) 996{ 997 struct recvfrom_args /* { 998 int s; 999 caddr_t buf; 1000 int len; 1001 int flags; 1002 struct sockaddr *from; 1003 socklen_t fromlenaddr; 1004 } */ bsd_args; 1005 1006 bsd_args.s = args->s; 1007 bsd_args.buf = (caddr_t)PTRIN(args->msg); 1008 bsd_args.len = args->len; 1009 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 1010 bsd_args.from = NULL; 1011 bsd_args.fromlenaddr = 0; 1012 return (sys_recvfrom(td, &bsd_args)); 1013} 1014#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1015 1016int 1017linux_sendto(struct thread *td, struct linux_sendto_args *args) 1018{ 1019 struct msghdr msg; 1020 struct iovec aiov; 1021 1022 if (linux_check_hdrincl(td, args->s) == 0) 1023 /* IP_HDRINCL set, tweak the packet before sending */ 1024 return (linux_sendto_hdrincl(td, args)); 1025 1026 msg.msg_name = PTRIN(args->to); 1027 msg.msg_namelen = args->tolen; 1028 msg.msg_iov = &aiov; 1029 msg.msg_iovlen = 1; 1030 msg.msg_control = NULL; 1031 msg.msg_flags = 0; 1032 aiov.iov_base = PTRIN(args->msg); 1033 aiov.iov_len = args->len; 1034 return (linux_sendit(td, args->s, &msg, args->flags, NULL, 1035 UIO_USERSPACE)); 1036} 1037 1038int 1039linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 1040{ 1041 struct msghdr msg; 1042 struct iovec aiov; 1043 int error; 1044 1045 if (PTRIN(args->fromlen) != NULL) { 1046 error = copyin(PTRIN(args->fromlen), &msg.msg_namelen, 1047 sizeof(msg.msg_namelen)); 1048 if (error != 0) 1049 return (error); 1050 1051 error = linux_to_bsd_sockaddr((struct sockaddr *)PTRIN(args->from), 1052 msg.msg_namelen); 1053 if (error != 0) 1054 return (error); 1055 } else 1056 msg.msg_namelen = 0; 1057 1058 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from); 1059 msg.msg_iov = &aiov; 1060 msg.msg_iovlen = 1; 1061 aiov.iov_base = PTRIN(args->buf); 1062 aiov.iov_len = args->len; 1063 msg.msg_control = 0; 1064 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 1065 1066 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL); 1067 if (error != 0) 1068 return (error); 1069 1070 if (PTRIN(args->from) != NULL) { 1071 error = bsd_to_linux_sockaddr((struct sockaddr *) 1072 PTRIN(args->from)); 1073 if (error != 0) 1074 return (error); 1075 1076 error = linux_sa_put((struct osockaddr *) 1077 PTRIN(args->from)); 1078 } 1079 1080 if (PTRIN(args->fromlen) != NULL) 1081 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 1082 sizeof(msg.msg_namelen)); 1083 1084 return (error); 1085} 1086 1087static int 1088linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1089 l_uint flags) 1090{ 1091 struct cmsghdr *cmsg; 1092 struct cmsgcred cmcred; 1093 struct mbuf *control; 1094 struct msghdr msg; 1095 struct l_cmsghdr linux_cmsg; 1096 struct l_cmsghdr *ptr_cmsg; 1097 struct l_msghdr linux_msg; 1098 struct iovec *iov; 1099 socklen_t datalen; 1100 struct sockaddr *sa; 1101 sa_family_t sa_family; 1102 void *data; 1103 int error; 1104 1105 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1106 if (error != 0) 1107 return (error); 1108 1109 /* 1110 * Some Linux applications (ping) define a non-NULL control data 1111 * pointer, but a msg_controllen of 0, which is not allowed in the 1112 * FreeBSD system call interface. NULL the msg_control pointer in 1113 * order to handle this case. This should be checked, but allows the 1114 * Linux ping to work. 1115 */ 1116 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0) 1117 linux_msg.msg_control = PTROUT(NULL); 1118 1119 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1120 if (error != 0) 1121 return (error); 1122 1123#ifdef COMPAT_LINUX32 1124 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1125 &iov, EMSGSIZE); 1126#else 1127 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1128#endif 1129 if (error != 0) 1130 return (error); 1131 1132 control = NULL; 1133 cmsg = NULL; 1134 1135 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) { 1136 error = kern_getsockname(td, s, &sa, &datalen); 1137 if (error != 0) 1138 goto bad; 1139 sa_family = sa->sa_family; 1140 free(sa, M_SONAME); 1141 1142 error = ENOBUFS; 1143 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO); 1144 control = m_get(M_WAITOK, MT_CONTROL); 1145 1146 do { 1147 error = copyin(ptr_cmsg, &linux_cmsg, 1148 sizeof(struct l_cmsghdr)); 1149 if (error != 0) 1150 goto bad; 1151 1152 error = EINVAL; 1153 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1154 goto bad; 1155 1156 /* 1157 * Now we support only SCM_RIGHTS and SCM_CRED, 1158 * so return EINVAL in any other cmsg_type 1159 */ 1160 cmsg->cmsg_type = 1161 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1162 cmsg->cmsg_level = 1163 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1164 if (cmsg->cmsg_type == -1 1165 || cmsg->cmsg_level != SOL_SOCKET) 1166 goto bad; 1167 1168 /* 1169 * Some applications (e.g. pulseaudio) attempt to 1170 * send ancillary data even if the underlying protocol 1171 * doesn't support it which is not allowed in the 1172 * FreeBSD system call interface. 1173 */ 1174 if (sa_family != AF_UNIX) 1175 continue; 1176 1177 data = LINUX_CMSG_DATA(ptr_cmsg); 1178 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1179 1180 switch (cmsg->cmsg_type) 1181 { 1182 case SCM_RIGHTS: 1183 break; 1184 1185 case SCM_CREDS: 1186 data = &cmcred; 1187 datalen = sizeof(cmcred); 1188 1189 /* 1190 * The lower levels will fill in the structure 1191 */ 1192 bzero(data, datalen); 1193 break; 1194 } 1195 1196 cmsg->cmsg_len = CMSG_LEN(datalen); 1197 1198 error = ENOBUFS; 1199 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg)) 1200 goto bad; 1201 if (!m_append(control, datalen, (c_caddr_t)data)) 1202 goto bad; 1203 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg))); 1204 1205 if (m_length(control, NULL) == 0) { 1206 m_freem(control); 1207 control = NULL; 1208 } 1209 } 1210 1211 msg.msg_iov = iov; 1212 msg.msg_flags = 0; 1213 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1214 control = NULL; 1215 1216bad: 1217 m_freem(control); 1218 free(iov, M_IOV); 1219 if (cmsg) 1220 free(cmsg, M_LINUX); 1221 return (error); 1222} 1223 1224int 1225linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1226{ 1227 1228 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1229 args->flags)); 1230} 1231 1232int 1233linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1234{ 1235 struct l_mmsghdr *msg; 1236 l_uint retval; 1237 int error, datagrams; 1238 1239 if (args->vlen > UIO_MAXIOV) 1240 args->vlen = UIO_MAXIOV; 1241 1242 msg = PTRIN(args->msg); 1243 datagrams = 0; 1244 while (datagrams < args->vlen) { 1245 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1246 args->flags); 1247 if (error != 0) 1248 break; 1249 1250 retval = td->td_retval[0]; 1251 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1252 if (error != 0) 1253 break; 1254 ++msg; 1255 ++datagrams; 1256 } 1257 if (error == 0) 1258 td->td_retval[0] = datagrams; 1259 return (error); 1260} 1261 1262static int 1263linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1264 l_uint flags, struct msghdr *msg) 1265{ 1266 struct cmsghdr *cm; 1267 struct cmsgcred *cmcred; 1268 struct l_cmsghdr *linux_cmsg = NULL; 1269 struct l_ucred linux_ucred; 1270 socklen_t datalen, outlen; 1271 struct l_msghdr linux_msg; 1272 struct iovec *iov, *uiov; 1273 struct mbuf *control = NULL; 1274 struct mbuf **controlp; 1275 struct timeval *ftmvl; 1276 l_timeval ltmvl; 1277 caddr_t outbuf; 1278 void *data; 1279 int error, i, fd, fds, *fdp; 1280 1281 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1282 if (error != 0) 1283 return (error); 1284 1285 error = linux_to_bsd_msghdr(msg, &linux_msg); 1286 if (error != 0) 1287 return (error); 1288 1289#ifdef COMPAT_LINUX32 1290 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1291 &iov, EMSGSIZE); 1292#else 1293 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1294#endif 1295 if (error != 0) 1296 return (error); 1297 1298 if (msg->msg_name) { 1299 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name, 1300 msg->msg_namelen); 1301 if (error != 0) 1302 goto bad; 1303 } 1304 1305 uiov = msg->msg_iov; 1306 msg->msg_iov = iov; 1307 controlp = (msg->msg_control != NULL) ? &control : NULL; 1308 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp); 1309 msg->msg_iov = uiov; 1310 if (error != 0) 1311 goto bad; 1312 1313 error = bsd_to_linux_msghdr(msg, &linux_msg); 1314 if (error != 0) 1315 goto bad; 1316 1317 if (linux_msg.msg_name) { 1318 error = bsd_to_linux_sockaddr((struct sockaddr *) 1319 PTRIN(linux_msg.msg_name)); 1320 if (error != 0) 1321 goto bad; 1322 } 1323 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1324 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1325 if (error != 0) 1326 goto bad; 1327 } 1328 1329 outbuf = PTRIN(linux_msg.msg_control); 1330 outlen = 0; 1331 1332 if (control) { 1333 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1334 1335 msg->msg_control = mtod(control, struct cmsghdr *); 1336 msg->msg_controllen = control->m_len; 1337 1338 cm = CMSG_FIRSTHDR(msg); 1339 1340 while (cm != NULL) { 1341 linux_cmsg->cmsg_type = 1342 bsd_to_linux_cmsg_type(cm->cmsg_type); 1343 linux_cmsg->cmsg_level = 1344 bsd_to_linux_sockopt_level(cm->cmsg_level); 1345 if (linux_cmsg->cmsg_type == -1 1346 || cm->cmsg_level != SOL_SOCKET) 1347 { 1348 error = EINVAL; 1349 goto bad; 1350 } 1351 1352 data = CMSG_DATA(cm); 1353 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1354 1355 switch (cm->cmsg_type) 1356 { 1357 case SCM_RIGHTS: 1358 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1359 fds = datalen / sizeof(int); 1360 fdp = data; 1361 for (i = 0; i < fds; i++) { 1362 fd = *fdp++; 1363 (void)kern_fcntl(td, fd, 1364 F_SETFD, FD_CLOEXEC); 1365 } 1366 } 1367 break; 1368 1369 case SCM_CREDS: 1370 /* 1371 * Currently LOCAL_CREDS is never in 1372 * effect for Linux so no need to worry 1373 * about sockcred 1374 */ 1375 if (datalen != sizeof(*cmcred)) { 1376 error = EMSGSIZE; 1377 goto bad; 1378 } 1379 cmcred = (struct cmsgcred *)data; 1380 bzero(&linux_ucred, sizeof(linux_ucred)); 1381 linux_ucred.pid = cmcred->cmcred_pid; 1382 linux_ucred.uid = cmcred->cmcred_uid; 1383 linux_ucred.gid = cmcred->cmcred_gid; 1384 data = &linux_ucred; 1385 datalen = sizeof(linux_ucred); 1386 break; 1387 1388 case SCM_TIMESTAMP: 1389 if (datalen != sizeof(struct timeval)) { 1390 error = EMSGSIZE; 1391 goto bad; 1392 } 1393 ftmvl = (struct timeval *)data; 1394 ltmvl.tv_sec = ftmvl->tv_sec; 1395 ltmvl.tv_usec = ftmvl->tv_usec; 1396 data = <mvl; 1397 datalen = sizeof(ltmvl); 1398 break; 1399 } 1400 1401 if (outlen + LINUX_CMSG_LEN(datalen) > 1402 linux_msg.msg_controllen) { 1403 if (outlen == 0) { 1404 error = EMSGSIZE; 1405 goto bad; 1406 } else { 1407 linux_msg.msg_flags |= 1408 LINUX_MSG_CTRUNC; 1409 goto out; 1410 } 1411 } 1412 1413 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1414 1415 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1416 if (error) 1417 goto bad; 1418 outbuf += L_CMSG_HDRSZ; 1419 1420 error = copyout(data, outbuf, datalen); 1421 if (error) 1422 goto bad; 1423 1424 outbuf += LINUX_CMSG_ALIGN(datalen); 1425 outlen += LINUX_CMSG_LEN(datalen); 1426 1427 cm = CMSG_NXTHDR(msg, cm); 1428 } 1429 } 1430 1431out: 1432 linux_msg.msg_controllen = outlen; 1433 error = copyout(&linux_msg, msghdr, sizeof(linux_msg)); 1434 1435bad: 1436 free(iov, M_IOV); 1437 m_freem(control); 1438 free(linux_cmsg, M_LINUX); 1439 1440 return (error); 1441} 1442 1443int 1444linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1445{ 1446 struct msghdr bsd_msg; 1447 1448 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1449 args->flags, &bsd_msg)); 1450} 1451 1452int 1453linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1454{ 1455 struct l_mmsghdr *msg; 1456 struct msghdr bsd_msg; 1457 struct l_timespec lts; 1458 struct timespec ts, tts; 1459 l_uint retval; 1460 int error, datagrams; 1461 1462 if (args->timeout) { 1463 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1464 if (error != 0) 1465 return (error); 1466 error = linux_to_native_timespec(&ts, <s); 1467 if (error != 0) 1468 return (error); 1469 getnanotime(&tts); 1470 timespecadd(&tts, &ts); 1471 } 1472 1473 msg = PTRIN(args->msg); 1474 datagrams = 0; 1475 while (datagrams < args->vlen) { 1476 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1477 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1478 if (error != 0) 1479 break; 1480 1481 retval = td->td_retval[0]; 1482 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1483 if (error != 0) 1484 break; 1485 ++msg; 1486 ++datagrams; 1487 1488 /* 1489 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1490 */ 1491 if (args->flags & LINUX_MSG_WAITFORONE) 1492 args->flags |= LINUX_MSG_DONTWAIT; 1493 1494 /* 1495 * See BUGS section of recvmmsg(2). 1496 */ 1497 if (args->timeout) { 1498 getnanotime(&ts); 1499 timespecsub(&ts, &tts); 1500 if (!timespecisset(&ts) || ts.tv_sec > 0) 1501 break; 1502 } 1503 /* Out of band data, return right away. */ 1504 if (bsd_msg.msg_flags & MSG_OOB) 1505 break; 1506 } 1507 if (error == 0) 1508 td->td_retval[0] = datagrams; 1509 return (error); 1510} 1511 1512int 1513linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1514{ 1515 struct shutdown_args /* { 1516 int s; 1517 int how; 1518 } */ bsd_args; 1519 1520 bsd_args.s = args->s; 1521 bsd_args.how = args->how; 1522 return (sys_shutdown(td, &bsd_args)); 1523} 1524 1525int 1526linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1527{ 1528 struct setsockopt_args /* { 1529 int s; 1530 int level; 1531 int name; 1532 caddr_t val; 1533 int valsize; 1534 } */ bsd_args; 1535 l_timeval linux_tv; 1536 struct timeval tv; 1537 int error, name; 1538 1539 bsd_args.s = args->s; 1540 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1541 switch (bsd_args.level) { 1542 case SOL_SOCKET: 1543 name = linux_to_bsd_so_sockopt(args->optname); 1544 switch (name) { 1545 case SO_RCVTIMEO: 1546 /* FALLTHROUGH */ 1547 case SO_SNDTIMEO: 1548 error = copyin(PTRIN(args->optval), &linux_tv, 1549 sizeof(linux_tv)); 1550 if (error) 1551 return (error); 1552 tv.tv_sec = linux_tv.tv_sec; 1553 tv.tv_usec = linux_tv.tv_usec; 1554 return (kern_setsockopt(td, args->s, bsd_args.level, 1555 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1556 /* NOTREACHED */ 1557 break; 1558 default: 1559 break; 1560 } 1561 break; 1562 case IPPROTO_IP: 1563 name = linux_to_bsd_ip_sockopt(args->optname); 1564 break; 1565 case IPPROTO_IPV6: 1566 name = linux_to_bsd_ip6_sockopt(args->optname); 1567 break; 1568 case IPPROTO_TCP: 1569 name = linux_to_bsd_tcp_sockopt(args->optname); 1570 break; 1571 default: 1572 name = -1; 1573 break; 1574 } 1575 if (name == -1) 1576 return (ENOPROTOOPT); 1577 1578 bsd_args.name = name; 1579 bsd_args.val = PTRIN(args->optval); 1580 bsd_args.valsize = args->optlen; 1581 1582 if (name == IPV6_NEXTHOP) { 1583 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1584 bsd_args.valsize); 1585 error = sys_setsockopt(td, &bsd_args); 1586 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1587 } else 1588 error = sys_setsockopt(td, &bsd_args); 1589 1590 return (error); 1591} 1592 1593int 1594linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1595{ 1596 struct getsockopt_args /* { 1597 int s; 1598 int level; 1599 int name; 1600 caddr_t val; 1601 int *avalsize; 1602 } */ bsd_args; 1603 l_timeval linux_tv; 1604 struct timeval tv; 1605 socklen_t tv_len, xulen, len; 1606 struct xucred xu; 1607 struct l_ucred lxu; 1608 int error, name, newval; 1609 1610 bsd_args.s = args->s; 1611 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1612 switch (bsd_args.level) { 1613 case SOL_SOCKET: 1614 name = linux_to_bsd_so_sockopt(args->optname); 1615 switch (name) { 1616 case SO_RCVTIMEO: 1617 /* FALLTHROUGH */ 1618 case SO_SNDTIMEO: 1619 tv_len = sizeof(tv); 1620 error = kern_getsockopt(td, args->s, bsd_args.level, 1621 name, &tv, UIO_SYSSPACE, &tv_len); 1622 if (error) 1623 return (error); 1624 linux_tv.tv_sec = tv.tv_sec; 1625 linux_tv.tv_usec = tv.tv_usec; 1626 return (copyout(&linux_tv, PTRIN(args->optval), 1627 sizeof(linux_tv))); 1628 /* NOTREACHED */ 1629 break; 1630 case LOCAL_PEERCRED: 1631 if (args->optlen != sizeof(lxu)) 1632 return (EINVAL); 1633 xulen = sizeof(xu); 1634 error = kern_getsockopt(td, args->s, bsd_args.level, 1635 name, &xu, UIO_SYSSPACE, &xulen); 1636 if (error) 1637 return (error); 1638 /* 1639 * XXX Use 0 for pid as the FreeBSD does not cache peer pid. 1640 */ 1641 lxu.pid = 0; 1642 lxu.uid = xu.cr_uid; 1643 lxu.gid = xu.cr_gid; 1644 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1645 /* NOTREACHED */ 1646 break; 1647 case SO_ERROR: 1648 len = sizeof(newval); 1649 error = kern_getsockopt(td, args->s, bsd_args.level, 1650 name, &newval, UIO_SYSSPACE, &len); 1651 if (error) 1652 return (error); 1653 newval = -SV_ABI_ERRNO(td->td_proc, newval); 1654 return (copyout(&newval, PTRIN(args->optval), len)); 1655 /* NOTREACHED */ 1656 default: 1657 break; 1658 } 1659 break; 1660 case IPPROTO_IP: 1661 name = linux_to_bsd_ip_sockopt(args->optname); 1662 break; 1663 case IPPROTO_IPV6: 1664 name = linux_to_bsd_ip6_sockopt(args->optname); 1665 break; 1666 case IPPROTO_TCP: 1667 name = linux_to_bsd_tcp_sockopt(args->optname); 1668 break; 1669 default: 1670 name = -1; 1671 break; 1672 } 1673 if (name == -1) 1674 return (EINVAL); 1675 1676 bsd_args.name = name; 1677 bsd_args.val = PTRIN(args->optval); 1678 bsd_args.avalsize = PTRIN(args->optlen); 1679 1680 if (name == IPV6_NEXTHOP) { 1681 error = sys_getsockopt(td, &bsd_args); 1682 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1683 } else 1684 error = sys_getsockopt(td, &bsd_args); 1685 1686 return (error); 1687} 1688 1689#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1690 1691/* Argument list sizes for linux_socketcall */ 1692 1693#define LINUX_AL(x) ((x) * sizeof(l_ulong)) 1694 1695static const unsigned char lxs_args[] = { 1696 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */, 1697 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */, 1698 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */, 1699 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */, 1700 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */, 1701 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */, 1702 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */, 1703 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */, 1704 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */, 1705 LINUX_AL(4) /* accept4 */, LINUX_AL(5) /* recvmmsg */, 1706 LINUX_AL(4) /* sendmmsg */ 1707}; 1708 1709#define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1 1710 1711int 1712linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1713{ 1714 l_ulong a[6]; 1715 void *arg; 1716 int error; 1717 1718 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE) 1719 return (EINVAL); 1720 error = copyin(PTRIN(args->args), a, lxs_args[args->what]); 1721 if (error) 1722 return (error); 1723 1724 arg = a; 1725 switch (args->what) { 1726 case LINUX_SOCKET: 1727 return (linux_socket(td, arg)); 1728 case LINUX_BIND: 1729 return (linux_bind(td, arg)); 1730 case LINUX_CONNECT: 1731 return (linux_connect(td, arg)); 1732 case LINUX_LISTEN: 1733 return (linux_listen(td, arg)); 1734 case LINUX_ACCEPT: 1735 return (linux_accept(td, arg)); 1736 case LINUX_GETSOCKNAME: 1737 return (linux_getsockname(td, arg)); 1738 case LINUX_GETPEERNAME: 1739 return (linux_getpeername(td, arg)); 1740 case LINUX_SOCKETPAIR: 1741 return (linux_socketpair(td, arg)); 1742 case LINUX_SEND: 1743 return (linux_send(td, arg)); 1744 case LINUX_RECV: 1745 return (linux_recv(td, arg)); 1746 case LINUX_SENDTO: 1747 return (linux_sendto(td, arg)); 1748 case LINUX_RECVFROM: 1749 return (linux_recvfrom(td, arg)); 1750 case LINUX_SHUTDOWN: 1751 return (linux_shutdown(td, arg)); 1752 case LINUX_SETSOCKOPT: 1753 return (linux_setsockopt(td, arg)); 1754 case LINUX_GETSOCKOPT: 1755 return (linux_getsockopt(td, arg)); 1756 case LINUX_SENDMSG: 1757 return (linux_sendmsg(td, arg)); 1758 case LINUX_RECVMSG: 1759 return (linux_recvmsg(td, arg)); 1760 case LINUX_ACCEPT4: 1761 return (linux_accept4(td, arg)); 1762 case LINUX_RECVMMSG: 1763 return (linux_recvmmsg(td, arg)); 1764 case LINUX_SENDMMSG: 1765 return (linux_sendmmsg(td, arg)); 1766 } 1767 1768 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1769 return (ENOSYS); 1770} 1771#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1772