linux_socket.c revision 315313
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/11/sys/compat/linux/linux_socket.c 315313 2017-03-15 17:06:04Z 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 <net/vnet.h> 57#include <netinet/in.h> 58#include <netinet/in_systm.h> 59#include <netinet/ip.h> 60#include <netinet/tcp.h> 61#ifdef INET6 62#include <netinet/ip6.h> 63#include <netinet6/ip6_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 != 0) 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 != 0) 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 != 0) 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 != 0) 763 return (error); 764 765 error = kern_bindat(td, AT_FDCWD, 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 struct file *fp; 779 u_int fflag; 780 int error; 781 782 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 783 args->namelen); 784 if (error != 0) 785 return (error); 786 787 error = kern_connectat(td, AT_FDCWD, args->s, sa); 788 free(sa, M_SONAME); 789 if (error != EISCONN) 790 return (error); 791 792 /* 793 * Linux doesn't return EISCONN the first time it occurs, 794 * when on a non-blocking socket. Instead it returns the 795 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 796 */ 797 error = getsock_cap(td, args->s, cap_rights_init(&rights, CAP_CONNECT), 798 &fp, &fflag, NULL); 799 if (error != 0) 800 return (error); 801 802 error = EISCONN; 803 so = fp->f_data; 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 fdrop(fp, td); 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 cap_rights_t rights; 840 struct socket *so; 841 struct file *fp; 842 int error, error1; 843 844 bsd_args.s = s; 845 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr); 846 bsd_args.anamelen = PTRIN(namelen); 847 bsd_args.flags = 0; 848 error = linux_set_socket_flags(flags, &bsd_args.flags); 849 if (error != 0) 850 return (error); 851 error = sys_accept4(td, &bsd_args); 852 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 853 if (error != 0) { 854 if (error == EFAULT && namelen != sizeof(struct sockaddr_in)) 855 return (EINVAL); 856 if (error == EINVAL) { 857 error1 = getsock_cap(td, s, 858 cap_rights_init(&rights, CAP_ACCEPT), &fp, NULL, NULL); 859 if (error1 != 0) 860 return (error1); 861 so = fp->f_data; 862 if (so->so_type == SOCK_DGRAM) { 863 fdrop(fp, td); 864 return (EOPNOTSUPP); 865 } 866 fdrop(fp, td); 867 } 868 return (error); 869 } 870 if (addr) 871 error = linux_sa_put(PTRIN(addr)); 872 if (error != 0) { 873 (void)kern_close(td, td->td_retval[0]); 874 td->td_retval[0] = 0; 875 } 876 return (error); 877} 878 879int 880linux_accept(struct thread *td, struct linux_accept_args *args) 881{ 882 883 return (linux_accept_common(td, args->s, args->addr, 884 args->namelen, 0)); 885} 886 887int 888linux_accept4(struct thread *td, struct linux_accept4_args *args) 889{ 890 891 return (linux_accept_common(td, args->s, args->addr, 892 args->namelen, args->flags)); 893} 894 895int 896linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 897{ 898 struct getsockname_args /* { 899 int fdes; 900 struct sockaddr * __restrict asa; 901 socklen_t * __restrict alen; 902 } */ bsd_args; 903 int error; 904 905 bsd_args.fdes = args->s; 906 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 907 bsd_args.alen = PTRIN(args->namelen); 908 error = sys_getsockname(td, &bsd_args); 909 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 910 if (error != 0) 911 return (error); 912 return (linux_sa_put(PTRIN(args->addr))); 913} 914 915int 916linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 917{ 918 struct getpeername_args /* { 919 int fdes; 920 caddr_t asa; 921 int *alen; 922 } */ bsd_args; 923 int error; 924 925 bsd_args.fdes = args->s; 926 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 927 bsd_args.alen = (socklen_t *)PTRIN(args->namelen); 928 error = sys_getpeername(td, &bsd_args); 929 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 930 if (error != 0) 931 return (error); 932 return (linux_sa_put(PTRIN(args->addr))); 933} 934 935int 936linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 937{ 938 struct socketpair_args /* { 939 int domain; 940 int type; 941 int protocol; 942 int *rsv; 943 } */ bsd_args; 944 int error; 945 946 bsd_args.domain = linux_to_bsd_domain(args->domain); 947 if (bsd_args.domain != PF_LOCAL) 948 return (EAFNOSUPPORT); 949 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 950 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 951 return (EINVAL); 952 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 953 &bsd_args.type); 954 if (error != 0) 955 return (error); 956 if (args->protocol != 0 && args->protocol != PF_UNIX) 957 958 /* 959 * Use of PF_UNIX as protocol argument is not right, 960 * but Linux does it. 961 * Do not map PF_UNIX as its Linux value is identical 962 * to FreeBSD one. 963 */ 964 return (EPROTONOSUPPORT); 965 else 966 bsd_args.protocol = 0; 967 bsd_args.rsv = (int *)PTRIN(args->rsv); 968 return (sys_socketpair(td, &bsd_args)); 969} 970 971#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 972struct linux_send_args { 973 register_t s; 974 register_t msg; 975 register_t len; 976 register_t flags; 977}; 978 979static int 980linux_send(struct thread *td, struct linux_send_args *args) 981{ 982 struct sendto_args /* { 983 int s; 984 caddr_t buf; 985 int len; 986 int flags; 987 caddr_t to; 988 int tolen; 989 } */ bsd_args; 990 991 bsd_args.s = args->s; 992 bsd_args.buf = (caddr_t)PTRIN(args->msg); 993 bsd_args.len = args->len; 994 bsd_args.flags = args->flags; 995 bsd_args.to = NULL; 996 bsd_args.tolen = 0; 997 return (sys_sendto(td, &bsd_args)); 998} 999 1000struct linux_recv_args { 1001 register_t s; 1002 register_t msg; 1003 register_t len; 1004 register_t flags; 1005}; 1006 1007static int 1008linux_recv(struct thread *td, struct linux_recv_args *args) 1009{ 1010 struct recvfrom_args /* { 1011 int s; 1012 caddr_t buf; 1013 int len; 1014 int flags; 1015 struct sockaddr *from; 1016 socklen_t fromlenaddr; 1017 } */ bsd_args; 1018 1019 bsd_args.s = args->s; 1020 bsd_args.buf = (caddr_t)PTRIN(args->msg); 1021 bsd_args.len = args->len; 1022 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 1023 bsd_args.from = NULL; 1024 bsd_args.fromlenaddr = 0; 1025 return (sys_recvfrom(td, &bsd_args)); 1026} 1027#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1028 1029int 1030linux_sendto(struct thread *td, struct linux_sendto_args *args) 1031{ 1032 struct msghdr msg; 1033 struct iovec aiov; 1034 1035 if (linux_check_hdrincl(td, args->s) == 0) 1036 /* IP_HDRINCL set, tweak the packet before sending */ 1037 return (linux_sendto_hdrincl(td, args)); 1038 1039 msg.msg_name = PTRIN(args->to); 1040 msg.msg_namelen = args->tolen; 1041 msg.msg_iov = &aiov; 1042 msg.msg_iovlen = 1; 1043 msg.msg_control = NULL; 1044 msg.msg_flags = 0; 1045 aiov.iov_base = PTRIN(args->msg); 1046 aiov.iov_len = args->len; 1047 return (linux_sendit(td, args->s, &msg, args->flags, NULL, 1048 UIO_USERSPACE)); 1049} 1050 1051int 1052linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 1053{ 1054 struct msghdr msg; 1055 struct iovec aiov; 1056 int error, fromlen; 1057 1058 if (PTRIN(args->fromlen) != NULL) { 1059 error = copyin(PTRIN(args->fromlen), &fromlen, 1060 sizeof(fromlen)); 1061 if (error != 0) 1062 return (error); 1063 if (fromlen < 0) 1064 return (EINVAL); 1065 msg.msg_namelen = fromlen; 1066 } else 1067 msg.msg_namelen = 0; 1068 1069 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from); 1070 msg.msg_iov = &aiov; 1071 msg.msg_iovlen = 1; 1072 aiov.iov_base = PTRIN(args->buf); 1073 aiov.iov_len = args->len; 1074 msg.msg_control = 0; 1075 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 1076 1077 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL); 1078 if (error != 0) 1079 return (error); 1080 1081 if (PTRIN(args->from) != NULL) { 1082 error = bsd_to_linux_sockaddr((struct sockaddr *) 1083 PTRIN(args->from)); 1084 if (error != 0) 1085 return (error); 1086 1087 error = linux_sa_put((struct osockaddr *) 1088 PTRIN(args->from)); 1089 } 1090 1091 if (PTRIN(args->fromlen) != NULL) 1092 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 1093 sizeof(msg.msg_namelen)); 1094 1095 return (error); 1096} 1097 1098static int 1099linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1100 l_uint flags) 1101{ 1102 struct cmsghdr *cmsg; 1103 struct cmsgcred cmcred; 1104 struct mbuf *control; 1105 struct msghdr msg; 1106 struct l_cmsghdr linux_cmsg; 1107 struct l_cmsghdr *ptr_cmsg; 1108 struct l_msghdr linux_msg; 1109 struct iovec *iov; 1110 socklen_t datalen; 1111 struct sockaddr *sa; 1112 sa_family_t sa_family; 1113 void *data; 1114 int error; 1115 1116 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1117 if (error != 0) 1118 return (error); 1119 1120 /* 1121 * Some Linux applications (ping) define a non-NULL control data 1122 * pointer, but a msg_controllen of 0, which is not allowed in the 1123 * FreeBSD system call interface. NULL the msg_control pointer in 1124 * order to handle this case. This should be checked, but allows the 1125 * Linux ping to work. 1126 */ 1127 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0) 1128 linux_msg.msg_control = PTROUT(NULL); 1129 1130 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1131 if (error != 0) 1132 return (error); 1133 1134#ifdef COMPAT_LINUX32 1135 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1136 &iov, EMSGSIZE); 1137#else 1138 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1139#endif 1140 if (error != 0) 1141 return (error); 1142 1143 control = NULL; 1144 cmsg = NULL; 1145 1146 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) { 1147 error = kern_getsockname(td, s, &sa, &datalen); 1148 if (error != 0) 1149 goto bad; 1150 sa_family = sa->sa_family; 1151 free(sa, M_SONAME); 1152 1153 error = ENOBUFS; 1154 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO); 1155 control = m_get(M_WAITOK, MT_CONTROL); 1156 1157 do { 1158 error = copyin(ptr_cmsg, &linux_cmsg, 1159 sizeof(struct l_cmsghdr)); 1160 if (error != 0) 1161 goto bad; 1162 1163 error = EINVAL; 1164 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1165 goto bad; 1166 1167 /* 1168 * Now we support only SCM_RIGHTS and SCM_CRED, 1169 * so return EINVAL in any other cmsg_type 1170 */ 1171 cmsg->cmsg_type = 1172 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1173 cmsg->cmsg_level = 1174 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1175 if (cmsg->cmsg_type == -1 1176 || cmsg->cmsg_level != SOL_SOCKET) 1177 goto bad; 1178 1179 /* 1180 * Some applications (e.g. pulseaudio) attempt to 1181 * send ancillary data even if the underlying protocol 1182 * doesn't support it which is not allowed in the 1183 * FreeBSD system call interface. 1184 */ 1185 if (sa_family != AF_UNIX) 1186 continue; 1187 1188 data = LINUX_CMSG_DATA(ptr_cmsg); 1189 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1190 1191 switch (cmsg->cmsg_type) 1192 { 1193 case SCM_RIGHTS: 1194 break; 1195 1196 case SCM_CREDS: 1197 data = &cmcred; 1198 datalen = sizeof(cmcred); 1199 1200 /* 1201 * The lower levels will fill in the structure 1202 */ 1203 bzero(data, datalen); 1204 break; 1205 } 1206 1207 cmsg->cmsg_len = CMSG_LEN(datalen); 1208 1209 error = ENOBUFS; 1210 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg)) 1211 goto bad; 1212 if (!m_append(control, datalen, (c_caddr_t)data)) 1213 goto bad; 1214 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg))); 1215 1216 if (m_length(control, NULL) == 0) { 1217 m_freem(control); 1218 control = NULL; 1219 } 1220 } 1221 1222 msg.msg_iov = iov; 1223 msg.msg_flags = 0; 1224 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1225 control = NULL; 1226 1227bad: 1228 m_freem(control); 1229 free(iov, M_IOV); 1230 if (cmsg) 1231 free(cmsg, M_LINUX); 1232 return (error); 1233} 1234 1235int 1236linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1237{ 1238 1239 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1240 args->flags)); 1241} 1242 1243int 1244linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1245{ 1246 struct l_mmsghdr *msg; 1247 l_uint retval; 1248 int error, datagrams; 1249 1250 if (args->vlen > UIO_MAXIOV) 1251 args->vlen = UIO_MAXIOV; 1252 1253 msg = PTRIN(args->msg); 1254 datagrams = 0; 1255 while (datagrams < args->vlen) { 1256 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1257 args->flags); 1258 if (error != 0) 1259 break; 1260 1261 retval = td->td_retval[0]; 1262 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1263 if (error != 0) 1264 break; 1265 ++msg; 1266 ++datagrams; 1267 } 1268 if (error == 0) 1269 td->td_retval[0] = datagrams; 1270 return (error); 1271} 1272 1273static int 1274linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1275 l_uint flags, struct msghdr *msg) 1276{ 1277 struct cmsghdr *cm; 1278 struct cmsgcred *cmcred; 1279 struct l_cmsghdr *linux_cmsg = NULL; 1280 struct l_ucred linux_ucred; 1281 socklen_t datalen, outlen; 1282 struct l_msghdr linux_msg; 1283 struct iovec *iov, *uiov; 1284 struct mbuf *control = NULL; 1285 struct mbuf **controlp; 1286 struct timeval *ftmvl; 1287 l_timeval ltmvl; 1288 caddr_t outbuf; 1289 void *data; 1290 int error, i, fd, fds, *fdp; 1291 1292 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1293 if (error != 0) 1294 return (error); 1295 1296 error = linux_to_bsd_msghdr(msg, &linux_msg); 1297 if (error != 0) 1298 return (error); 1299 1300#ifdef COMPAT_LINUX32 1301 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1302 &iov, EMSGSIZE); 1303#else 1304 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1305#endif 1306 if (error != 0) 1307 return (error); 1308 1309 if (msg->msg_name) { 1310 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name, 1311 msg->msg_namelen); 1312 if (error != 0) 1313 goto bad; 1314 } 1315 1316 uiov = msg->msg_iov; 1317 msg->msg_iov = iov; 1318 controlp = (msg->msg_control != NULL) ? &control : NULL; 1319 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp); 1320 msg->msg_iov = uiov; 1321 if (error != 0) 1322 goto bad; 1323 1324 error = bsd_to_linux_msghdr(msg, &linux_msg); 1325 if (error != 0) 1326 goto bad; 1327 1328 if (linux_msg.msg_name) { 1329 error = bsd_to_linux_sockaddr((struct sockaddr *) 1330 PTRIN(linux_msg.msg_name)); 1331 if (error != 0) 1332 goto bad; 1333 } 1334 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1335 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1336 if (error != 0) 1337 goto bad; 1338 } 1339 1340 outbuf = PTRIN(linux_msg.msg_control); 1341 outlen = 0; 1342 1343 if (control) { 1344 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1345 1346 msg->msg_control = mtod(control, struct cmsghdr *); 1347 msg->msg_controllen = control->m_len; 1348 1349 cm = CMSG_FIRSTHDR(msg); 1350 1351 while (cm != NULL) { 1352 linux_cmsg->cmsg_type = 1353 bsd_to_linux_cmsg_type(cm->cmsg_type); 1354 linux_cmsg->cmsg_level = 1355 bsd_to_linux_sockopt_level(cm->cmsg_level); 1356 if (linux_cmsg->cmsg_type == -1 1357 || cm->cmsg_level != SOL_SOCKET) 1358 { 1359 error = EINVAL; 1360 goto bad; 1361 } 1362 1363 data = CMSG_DATA(cm); 1364 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1365 1366 switch (cm->cmsg_type) 1367 { 1368 case SCM_RIGHTS: 1369 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1370 fds = datalen / sizeof(int); 1371 fdp = data; 1372 for (i = 0; i < fds; i++) { 1373 fd = *fdp++; 1374 (void)kern_fcntl(td, fd, 1375 F_SETFD, FD_CLOEXEC); 1376 } 1377 } 1378 break; 1379 1380 case SCM_CREDS: 1381 /* 1382 * Currently LOCAL_CREDS is never in 1383 * effect for Linux so no need to worry 1384 * about sockcred 1385 */ 1386 if (datalen != sizeof(*cmcred)) { 1387 error = EMSGSIZE; 1388 goto bad; 1389 } 1390 cmcred = (struct cmsgcred *)data; 1391 bzero(&linux_ucred, sizeof(linux_ucred)); 1392 linux_ucred.pid = cmcred->cmcred_pid; 1393 linux_ucred.uid = cmcred->cmcred_uid; 1394 linux_ucred.gid = cmcred->cmcred_gid; 1395 data = &linux_ucred; 1396 datalen = sizeof(linux_ucred); 1397 break; 1398 1399 case SCM_TIMESTAMP: 1400 if (datalen != sizeof(struct timeval)) { 1401 error = EMSGSIZE; 1402 goto bad; 1403 } 1404 ftmvl = (struct timeval *)data; 1405 ltmvl.tv_sec = ftmvl->tv_sec; 1406 ltmvl.tv_usec = ftmvl->tv_usec; 1407 data = <mvl; 1408 datalen = sizeof(ltmvl); 1409 break; 1410 } 1411 1412 if (outlen + LINUX_CMSG_LEN(datalen) > 1413 linux_msg.msg_controllen) { 1414 if (outlen == 0) { 1415 error = EMSGSIZE; 1416 goto bad; 1417 } else { 1418 linux_msg.msg_flags |= 1419 LINUX_MSG_CTRUNC; 1420 goto out; 1421 } 1422 } 1423 1424 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1425 1426 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1427 if (error != 0) 1428 goto bad; 1429 outbuf += L_CMSG_HDRSZ; 1430 1431 error = copyout(data, outbuf, datalen); 1432 if (error != 0) 1433 goto bad; 1434 1435 outbuf += LINUX_CMSG_ALIGN(datalen); 1436 outlen += LINUX_CMSG_LEN(datalen); 1437 1438 cm = CMSG_NXTHDR(msg, cm); 1439 } 1440 } 1441 1442out: 1443 linux_msg.msg_controllen = outlen; 1444 error = copyout(&linux_msg, msghdr, sizeof(linux_msg)); 1445 1446bad: 1447 free(iov, M_IOV); 1448 m_freem(control); 1449 free(linux_cmsg, M_LINUX); 1450 1451 return (error); 1452} 1453 1454int 1455linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1456{ 1457 struct msghdr bsd_msg; 1458 1459 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1460 args->flags, &bsd_msg)); 1461} 1462 1463int 1464linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1465{ 1466 struct l_mmsghdr *msg; 1467 struct msghdr bsd_msg; 1468 struct l_timespec lts; 1469 struct timespec ts, tts; 1470 l_uint retval; 1471 int error, datagrams; 1472 1473 if (args->timeout) { 1474 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1475 if (error != 0) 1476 return (error); 1477 error = linux_to_native_timespec(&ts, <s); 1478 if (error != 0) 1479 return (error); 1480 getnanotime(&tts); 1481 timespecadd(&tts, &ts); 1482 } 1483 1484 msg = PTRIN(args->msg); 1485 datagrams = 0; 1486 while (datagrams < args->vlen) { 1487 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1488 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1489 if (error != 0) 1490 break; 1491 1492 retval = td->td_retval[0]; 1493 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1494 if (error != 0) 1495 break; 1496 ++msg; 1497 ++datagrams; 1498 1499 /* 1500 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1501 */ 1502 if (args->flags & LINUX_MSG_WAITFORONE) 1503 args->flags |= LINUX_MSG_DONTWAIT; 1504 1505 /* 1506 * See BUGS section of recvmmsg(2). 1507 */ 1508 if (args->timeout) { 1509 getnanotime(&ts); 1510 timespecsub(&ts, &tts); 1511 if (!timespecisset(&ts) || ts.tv_sec > 0) 1512 break; 1513 } 1514 /* Out of band data, return right away. */ 1515 if (bsd_msg.msg_flags & MSG_OOB) 1516 break; 1517 } 1518 if (error == 0) 1519 td->td_retval[0] = datagrams; 1520 return (error); 1521} 1522 1523int 1524linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1525{ 1526 struct shutdown_args /* { 1527 int s; 1528 int how; 1529 } */ bsd_args; 1530 1531 bsd_args.s = args->s; 1532 bsd_args.how = args->how; 1533 return (sys_shutdown(td, &bsd_args)); 1534} 1535 1536int 1537linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1538{ 1539 struct setsockopt_args /* { 1540 int s; 1541 int level; 1542 int name; 1543 caddr_t val; 1544 int valsize; 1545 } */ bsd_args; 1546 l_timeval linux_tv; 1547 struct timeval tv; 1548 int error, name; 1549 1550 bsd_args.s = args->s; 1551 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1552 switch (bsd_args.level) { 1553 case SOL_SOCKET: 1554 name = linux_to_bsd_so_sockopt(args->optname); 1555 switch (name) { 1556 case SO_RCVTIMEO: 1557 /* FALLTHROUGH */ 1558 case SO_SNDTIMEO: 1559 error = copyin(PTRIN(args->optval), &linux_tv, 1560 sizeof(linux_tv)); 1561 if (error != 0) 1562 return (error); 1563 tv.tv_sec = linux_tv.tv_sec; 1564 tv.tv_usec = linux_tv.tv_usec; 1565 return (kern_setsockopt(td, args->s, bsd_args.level, 1566 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1567 /* NOTREACHED */ 1568 break; 1569 default: 1570 break; 1571 } 1572 break; 1573 case IPPROTO_IP: 1574 name = linux_to_bsd_ip_sockopt(args->optname); 1575 break; 1576 case IPPROTO_IPV6: 1577 name = linux_to_bsd_ip6_sockopt(args->optname); 1578 break; 1579 case IPPROTO_TCP: 1580 name = linux_to_bsd_tcp_sockopt(args->optname); 1581 break; 1582 default: 1583 name = -1; 1584 break; 1585 } 1586 if (name == -1) 1587 return (ENOPROTOOPT); 1588 1589 bsd_args.name = name; 1590 bsd_args.val = PTRIN(args->optval); 1591 bsd_args.valsize = args->optlen; 1592 1593 if (name == IPV6_NEXTHOP) { 1594 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1595 bsd_args.valsize); 1596 error = sys_setsockopt(td, &bsd_args); 1597 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1598 } else 1599 error = sys_setsockopt(td, &bsd_args); 1600 1601 return (error); 1602} 1603 1604int 1605linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1606{ 1607 struct getsockopt_args /* { 1608 int s; 1609 int level; 1610 int name; 1611 caddr_t val; 1612 int *avalsize; 1613 } */ bsd_args; 1614 l_timeval linux_tv; 1615 struct timeval tv; 1616 socklen_t tv_len, xulen, len; 1617 struct xucred xu; 1618 struct l_ucred lxu; 1619 int error, name, newval; 1620 1621 bsd_args.s = args->s; 1622 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1623 switch (bsd_args.level) { 1624 case SOL_SOCKET: 1625 name = linux_to_bsd_so_sockopt(args->optname); 1626 switch (name) { 1627 case SO_RCVTIMEO: 1628 /* FALLTHROUGH */ 1629 case SO_SNDTIMEO: 1630 tv_len = sizeof(tv); 1631 error = kern_getsockopt(td, args->s, bsd_args.level, 1632 name, &tv, UIO_SYSSPACE, &tv_len); 1633 if (error != 0) 1634 return (error); 1635 linux_tv.tv_sec = tv.tv_sec; 1636 linux_tv.tv_usec = tv.tv_usec; 1637 return (copyout(&linux_tv, PTRIN(args->optval), 1638 sizeof(linux_tv))); 1639 /* NOTREACHED */ 1640 break; 1641 case LOCAL_PEERCRED: 1642 if (args->optlen != sizeof(lxu)) 1643 return (EINVAL); 1644 xulen = sizeof(xu); 1645 error = kern_getsockopt(td, args->s, bsd_args.level, 1646 name, &xu, UIO_SYSSPACE, &xulen); 1647 if (error != 0) 1648 return (error); 1649 /* 1650 * XXX Use 0 for pid as the FreeBSD does not cache peer pid. 1651 */ 1652 lxu.pid = 0; 1653 lxu.uid = xu.cr_uid; 1654 lxu.gid = xu.cr_gid; 1655 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1656 /* NOTREACHED */ 1657 break; 1658 case SO_ERROR: 1659 len = sizeof(newval); 1660 error = kern_getsockopt(td, args->s, bsd_args.level, 1661 name, &newval, UIO_SYSSPACE, &len); 1662 if (error != 0) 1663 return (error); 1664 newval = -SV_ABI_ERRNO(td->td_proc, newval); 1665 return (copyout(&newval, PTRIN(args->optval), len)); 1666 /* NOTREACHED */ 1667 default: 1668 break; 1669 } 1670 break; 1671 case IPPROTO_IP: 1672 name = linux_to_bsd_ip_sockopt(args->optname); 1673 break; 1674 case IPPROTO_IPV6: 1675 name = linux_to_bsd_ip6_sockopt(args->optname); 1676 break; 1677 case IPPROTO_TCP: 1678 name = linux_to_bsd_tcp_sockopt(args->optname); 1679 break; 1680 default: 1681 name = -1; 1682 break; 1683 } 1684 if (name == -1) 1685 return (EINVAL); 1686 1687 bsd_args.name = name; 1688 bsd_args.val = PTRIN(args->optval); 1689 bsd_args.avalsize = PTRIN(args->optlen); 1690 1691 if (name == IPV6_NEXTHOP) { 1692 error = sys_getsockopt(td, &bsd_args); 1693 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1694 } else 1695 error = sys_getsockopt(td, &bsd_args); 1696 1697 return (error); 1698} 1699 1700#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1701 1702/* Argument list sizes for linux_socketcall */ 1703static const unsigned char lxs_args_cnt[] = { 1704 0 /* unused*/, 3 /* socket */, 1705 3 /* bind */, 3 /* connect */, 1706 2 /* listen */, 3 /* accept */, 1707 3 /* getsockname */, 3 /* getpeername */, 1708 4 /* socketpair */, 4 /* send */, 1709 4 /* recv */, 6 /* sendto */, 1710 6 /* recvfrom */, 2 /* shutdown */, 1711 5 /* setsockopt */, 5 /* getsockopt */, 1712 3 /* sendmsg */, 3 /* recvmsg */, 1713 4 /* accept4 */, 5 /* recvmmsg */, 1714 4 /* sendmmsg */ 1715}; 1716#define LINUX_ARGS_CNT (nitems(lxs_args_cnt) - 1) 1717#define LINUX_ARG_SIZE(x) (lxs_args_cnt[x] * sizeof(l_ulong)) 1718 1719int 1720linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1721{ 1722 l_ulong a[6]; 1723#if defined(__amd64__) && defined(COMPAT_LINUX32) 1724 register_t l_args[6]; 1725#endif 1726 void *arg; 1727 int error; 1728 1729 if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT) 1730 return (EINVAL); 1731 error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what)); 1732 if (error != 0) 1733 return (error); 1734 1735#if defined(__amd64__) && defined(COMPAT_LINUX32) 1736 for (int i = 0; i < lxs_args_cnt[args->what]; ++i) 1737 l_args[i] = a[i]; 1738 arg = l_args; 1739#else 1740 arg = a; 1741#endif 1742 switch (args->what) { 1743 case LINUX_SOCKET: 1744 return (linux_socket(td, arg)); 1745 case LINUX_BIND: 1746 return (linux_bind(td, arg)); 1747 case LINUX_CONNECT: 1748 return (linux_connect(td, arg)); 1749 case LINUX_LISTEN: 1750 return (linux_listen(td, arg)); 1751 case LINUX_ACCEPT: 1752 return (linux_accept(td, arg)); 1753 case LINUX_GETSOCKNAME: 1754 return (linux_getsockname(td, arg)); 1755 case LINUX_GETPEERNAME: 1756 return (linux_getpeername(td, arg)); 1757 case LINUX_SOCKETPAIR: 1758 return (linux_socketpair(td, arg)); 1759 case LINUX_SEND: 1760 return (linux_send(td, arg)); 1761 case LINUX_RECV: 1762 return (linux_recv(td, arg)); 1763 case LINUX_SENDTO: 1764 return (linux_sendto(td, arg)); 1765 case LINUX_RECVFROM: 1766 return (linux_recvfrom(td, arg)); 1767 case LINUX_SHUTDOWN: 1768 return (linux_shutdown(td, arg)); 1769 case LINUX_SETSOCKOPT: 1770 return (linux_setsockopt(td, arg)); 1771 case LINUX_GETSOCKOPT: 1772 return (linux_getsockopt(td, arg)); 1773 case LINUX_SENDMSG: 1774 return (linux_sendmsg(td, arg)); 1775 case LINUX_RECVMSG: 1776 return (linux_recvmsg(td, arg)); 1777 case LINUX_ACCEPT4: 1778 return (linux_accept4(td, arg)); 1779 case LINUX_RECVMMSG: 1780 return (linux_recvmmsg(td, arg)); 1781 case LINUX_SENDMMSG: 1782 return (linux_sendmmsg(td, arg)); 1783 } 1784 1785 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1786 return (ENOSYS); 1787} 1788#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1789