linux_socket.c revision 283488
1/*- 2 * Copyright (c) 1995 S��ren Schmidt 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer 10 * in this position and unchanged. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. The name of the author may not be used to endorse or promote products 15 * derived from this software without specific prior written permission 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29#include <sys/cdefs.h> 30__FBSDID("$FreeBSD: head/sys/compat/linux/linux_socket.c 283488 2015-05-24 18:04:04Z 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 *); 83 84/* 85 * Reads a linux sockaddr and does any necessary translation. 86 * Linux sockaddrs don't have a length field, only a family. 87 * Copy the osockaddr structure pointed to by osa to kernel, adjust 88 * family and convert to sockaddr. 89 */ 90static int 91linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen) 92{ 93 struct sockaddr *sa; 94 struct osockaddr *kosa; 95#ifdef INET6 96 struct sockaddr_in6 *sin6; 97 int oldv6size; 98#endif 99 char *name; 100 int bdom, error, hdrlen, namelen; 101 102 if (salen < 2 || salen > UCHAR_MAX || !osa) 103 return (EINVAL); 104 105#ifdef INET6 106 oldv6size = 0; 107 /* 108 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it 109 * if it's a v4-mapped address, so reserve the proper space 110 * for it. 111 */ 112 if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) { 113 salen += sizeof(uint32_t); 114 oldv6size = 1; 115 } 116#endif 117 118 kosa = malloc(salen, M_SONAME, M_WAITOK); 119 120 if ((error = copyin(osa, kosa, salen))) 121 goto out; 122 123 bdom = linux_to_bsd_domain(kosa->sa_family); 124 if (bdom == -1) { 125 error = EAFNOSUPPORT; 126 goto out; 127 } 128 129#ifdef INET6 130 /* 131 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6, 132 * which lacks the scope id compared with RFC2553 one. If we detect 133 * the situation, reject the address and write a message to system log. 134 * 135 * Still accept addresses for which the scope id is not used. 136 */ 137 if (oldv6size) { 138 if (bdom == AF_INET6) { 139 sin6 = (struct sockaddr_in6 *)kosa; 140 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) || 141 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) && 142 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) && 143 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) && 144 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) && 145 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) { 146 sin6->sin6_scope_id = 0; 147 } else { 148 log(LOG_DEBUG, 149 "obsolete pre-RFC2553 sockaddr_in6 rejected\n"); 150 error = EINVAL; 151 goto out; 152 } 153 } else 154 salen -= sizeof(uint32_t); 155 } 156#endif 157 if (bdom == AF_INET) { 158 if (salen < sizeof(struct sockaddr_in)) { 159 error = EINVAL; 160 goto out; 161 } 162 salen = sizeof(struct sockaddr_in); 163 } 164 165 if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) { 166 hdrlen = offsetof(struct sockaddr_un, sun_path); 167 name = ((struct sockaddr_un *)kosa)->sun_path; 168 if (*name == '\0') { 169 /* 170 * Linux abstract namespace starts with a NULL byte. 171 * XXX We do not support abstract namespace yet. 172 */ 173 namelen = strnlen(name + 1, salen - hdrlen - 1) + 1; 174 } else 175 namelen = strnlen(name, salen - hdrlen); 176 salen = hdrlen + namelen; 177 if (salen > sizeof(struct sockaddr_un)) { 178 error = ENAMETOOLONG; 179 goto out; 180 } 181 } 182 183 sa = (struct sockaddr *)kosa; 184 sa->sa_family = bdom; 185 sa->sa_len = salen; 186 187 *sap = sa; 188 return (0); 189 190out: 191 free(kosa, M_SONAME); 192 return (error); 193} 194 195static int 196linux_to_bsd_domain(int domain) 197{ 198 199 switch (domain) { 200 case LINUX_AF_UNSPEC: 201 return (AF_UNSPEC); 202 case LINUX_AF_UNIX: 203 return (AF_LOCAL); 204 case LINUX_AF_INET: 205 return (AF_INET); 206 case LINUX_AF_INET6: 207 return (AF_INET6); 208 case LINUX_AF_AX25: 209 return (AF_CCITT); 210 case LINUX_AF_IPX: 211 return (AF_IPX); 212 case LINUX_AF_APPLETALK: 213 return (AF_APPLETALK); 214 } 215 return (-1); 216} 217 218static int 219bsd_to_linux_domain(int domain) 220{ 221 222 switch (domain) { 223 case AF_UNSPEC: 224 return (LINUX_AF_UNSPEC); 225 case AF_LOCAL: 226 return (LINUX_AF_UNIX); 227 case AF_INET: 228 return (LINUX_AF_INET); 229 case AF_INET6: 230 return (LINUX_AF_INET6); 231 case AF_CCITT: 232 return (LINUX_AF_AX25); 233 case AF_IPX: 234 return (LINUX_AF_IPX); 235 case AF_APPLETALK: 236 return (LINUX_AF_APPLETALK); 237 } 238 return (-1); 239} 240 241static int 242linux_to_bsd_sockopt_level(int level) 243{ 244 245 switch (level) { 246 case LINUX_SOL_SOCKET: 247 return (SOL_SOCKET); 248 } 249 return (level); 250} 251 252static int 253bsd_to_linux_sockopt_level(int level) 254{ 255 256 switch (level) { 257 case SOL_SOCKET: 258 return (LINUX_SOL_SOCKET); 259 } 260 return (level); 261} 262 263static int 264linux_to_bsd_ip_sockopt(int opt) 265{ 266 267 switch (opt) { 268 case LINUX_IP_TOS: 269 return (IP_TOS); 270 case LINUX_IP_TTL: 271 return (IP_TTL); 272 case LINUX_IP_OPTIONS: 273 return (IP_OPTIONS); 274 case LINUX_IP_MULTICAST_IF: 275 return (IP_MULTICAST_IF); 276 case LINUX_IP_MULTICAST_TTL: 277 return (IP_MULTICAST_TTL); 278 case LINUX_IP_MULTICAST_LOOP: 279 return (IP_MULTICAST_LOOP); 280 case LINUX_IP_ADD_MEMBERSHIP: 281 return (IP_ADD_MEMBERSHIP); 282 case LINUX_IP_DROP_MEMBERSHIP: 283 return (IP_DROP_MEMBERSHIP); 284 case LINUX_IP_HDRINCL: 285 return (IP_HDRINCL); 286 } 287 return (-1); 288} 289 290static int 291linux_to_bsd_so_sockopt(int opt) 292{ 293 294 switch (opt) { 295 case LINUX_SO_DEBUG: 296 return (SO_DEBUG); 297 case LINUX_SO_REUSEADDR: 298 return (SO_REUSEADDR); 299 case LINUX_SO_TYPE: 300 return (SO_TYPE); 301 case LINUX_SO_ERROR: 302 return (SO_ERROR); 303 case LINUX_SO_DONTROUTE: 304 return (SO_DONTROUTE); 305 case LINUX_SO_BROADCAST: 306 return (SO_BROADCAST); 307 case LINUX_SO_SNDBUF: 308 return (SO_SNDBUF); 309 case LINUX_SO_RCVBUF: 310 return (SO_RCVBUF); 311 case LINUX_SO_KEEPALIVE: 312 return (SO_KEEPALIVE); 313 case LINUX_SO_OOBINLINE: 314 return (SO_OOBINLINE); 315 case LINUX_SO_LINGER: 316 return (SO_LINGER); 317 case LINUX_SO_PEERCRED: 318 return (LOCAL_PEERCRED); 319 case LINUX_SO_RCVLOWAT: 320 return (SO_RCVLOWAT); 321 case LINUX_SO_SNDLOWAT: 322 return (SO_SNDLOWAT); 323 case LINUX_SO_RCVTIMEO: 324 return (SO_RCVTIMEO); 325 case LINUX_SO_SNDTIMEO: 326 return (SO_SNDTIMEO); 327 case LINUX_SO_TIMESTAMP: 328 return (SO_TIMESTAMP); 329 case LINUX_SO_ACCEPTCONN: 330 return (SO_ACCEPTCONN); 331 } 332 return (-1); 333} 334 335static int 336linux_to_bsd_tcp_sockopt(int opt) 337{ 338 339 switch (opt) { 340 case LINUX_TCP_NODELAY: 341 return (TCP_NODELAY); 342 case LINUX_TCP_MAXSEG: 343 return (TCP_MAXSEG); 344 case LINUX_TCP_KEEPIDLE: 345 return (TCP_KEEPIDLE); 346 case LINUX_TCP_KEEPINTVL: 347 return (TCP_KEEPINTVL); 348 case LINUX_TCP_KEEPCNT: 349 return (TCP_KEEPCNT); 350 case LINUX_TCP_MD5SIG: 351 return (TCP_MD5SIG); 352 } 353 return (-1); 354} 355 356static int 357linux_to_bsd_msg_flags(int flags) 358{ 359 int ret_flags = 0; 360 361 if (flags & LINUX_MSG_OOB) 362 ret_flags |= MSG_OOB; 363 if (flags & LINUX_MSG_PEEK) 364 ret_flags |= MSG_PEEK; 365 if (flags & LINUX_MSG_DONTROUTE) 366 ret_flags |= MSG_DONTROUTE; 367 if (flags & LINUX_MSG_CTRUNC) 368 ret_flags |= MSG_CTRUNC; 369 if (flags & LINUX_MSG_TRUNC) 370 ret_flags |= MSG_TRUNC; 371 if (flags & LINUX_MSG_DONTWAIT) 372 ret_flags |= MSG_DONTWAIT; 373 if (flags & LINUX_MSG_EOR) 374 ret_flags |= MSG_EOR; 375 if (flags & LINUX_MSG_WAITALL) 376 ret_flags |= MSG_WAITALL; 377 if (flags & LINUX_MSG_NOSIGNAL) 378 ret_flags |= MSG_NOSIGNAL; 379#if 0 /* not handled */ 380 if (flags & LINUX_MSG_PROXY) 381 ; 382 if (flags & LINUX_MSG_FIN) 383 ; 384 if (flags & LINUX_MSG_SYN) 385 ; 386 if (flags & LINUX_MSG_CONFIRM) 387 ; 388 if (flags & LINUX_MSG_RST) 389 ; 390 if (flags & LINUX_MSG_ERRQUEUE) 391 ; 392#endif 393 return ret_flags; 394} 395 396/* 397* If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the 398* native syscall will fault. Thus, we don't really need to check the 399* return values for these functions. 400*/ 401 402static int 403bsd_to_linux_sockaddr(struct sockaddr *arg) 404{ 405 struct sockaddr sa; 406 size_t sa_len = sizeof(struct sockaddr); 407 int error; 408 409 if ((error = copyin(arg, &sa, sa_len))) 410 return (error); 411 412 *(u_short *)&sa = sa.sa_family; 413 414 error = copyout(&sa, arg, sa_len); 415 416 return (error); 417} 418 419static int 420linux_to_bsd_sockaddr(struct sockaddr *arg, int len) 421{ 422 struct sockaddr sa; 423 size_t sa_len = sizeof(struct sockaddr); 424 int error; 425 426 if ((error = copyin(arg, &sa, sa_len))) 427 return (error); 428 429 sa.sa_family = *(sa_family_t *)&sa; 430 sa.sa_len = len; 431 432 error = copyout(&sa, arg, sa_len); 433 434 return (error); 435} 436 437static int 438linux_sa_put(struct osockaddr *osa) 439{ 440 struct osockaddr sa; 441 int error, bdom; 442 443 /* 444 * Only read/write the osockaddr family part, the rest is 445 * not changed. 446 */ 447 error = copyin(osa, &sa, sizeof(sa.sa_family)); 448 if (error) 449 return (error); 450 451 bdom = bsd_to_linux_domain(sa.sa_family); 452 if (bdom == -1) 453 return (EINVAL); 454 455 sa.sa_family = bdom; 456 error = copyout(&sa, osa, sizeof(sa.sa_family)); 457 if (error) 458 return (error); 459 460 return (0); 461} 462 463static int 464linux_to_bsd_cmsg_type(int cmsg_type) 465{ 466 467 switch (cmsg_type) { 468 case LINUX_SCM_RIGHTS: 469 return (SCM_RIGHTS); 470 case LINUX_SCM_CREDENTIALS: 471 return (SCM_CREDS); 472 } 473 return (-1); 474} 475 476static int 477bsd_to_linux_cmsg_type(int cmsg_type) 478{ 479 480 switch (cmsg_type) { 481 case SCM_RIGHTS: 482 return (LINUX_SCM_RIGHTS); 483 case SCM_CREDS: 484 return (LINUX_SCM_CREDENTIALS); 485 } 486 return (-1); 487} 488 489static int 490linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 491{ 492 if (lhdr->msg_controllen > INT_MAX) 493 return (ENOBUFS); 494 495 bhdr->msg_name = PTRIN(lhdr->msg_name); 496 bhdr->msg_namelen = lhdr->msg_namelen; 497 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 498 bhdr->msg_iovlen = lhdr->msg_iovlen; 499 bhdr->msg_control = PTRIN(lhdr->msg_control); 500 501 /* 502 * msg_controllen is skipped since BSD and LINUX control messages 503 * are potentially different sizes (e.g. the cred structure used 504 * by SCM_CREDS is different between the two operating system). 505 * 506 * The caller can set it (if necessary) after converting all the 507 * control messages. 508 */ 509 510 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 511 return (0); 512} 513 514static int 515bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 516{ 517 lhdr->msg_name = PTROUT(bhdr->msg_name); 518 lhdr->msg_namelen = bhdr->msg_namelen; 519 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 520 lhdr->msg_iovlen = bhdr->msg_iovlen; 521 lhdr->msg_control = PTROUT(bhdr->msg_control); 522 523 /* 524 * msg_controllen is skipped since BSD and LINUX control messages 525 * are potentially different sizes (e.g. the cred structure used 526 * by SCM_CREDS is different between the two operating system). 527 * 528 * The caller can set it (if necessary) after converting all the 529 * control messages. 530 */ 531 532 /* msg_flags skipped */ 533 return (0); 534} 535 536static int 537linux_set_socket_flags(struct thread *td, int s, int flags) 538{ 539 int error; 540 541 if (flags & LINUX_SOCK_NONBLOCK) { 542 error = kern_fcntl(td, s, F_SETFL, O_NONBLOCK); 543 if (error) 544 return (error); 545 } 546 if (flags & LINUX_SOCK_CLOEXEC) { 547 error = kern_fcntl(td, s, F_SETFD, FD_CLOEXEC); 548 if (error) 549 return (error); 550 } 551 return (0); 552} 553 554static int 555linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 556 struct mbuf *control, enum uio_seg segflg) 557{ 558 struct sockaddr *to; 559 int error; 560 561 if (mp->msg_name != NULL) { 562 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen); 563 if (error) 564 return (error); 565 mp->msg_name = to; 566 } else 567 to = NULL; 568 569 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 570 segflg); 571 572 if (to) 573 free(to, M_SONAME); 574 return (error); 575} 576 577/* Return 0 if IP_HDRINCL is set for the given socket. */ 578static int 579linux_check_hdrincl(struct thread *td, int s) 580{ 581 int error, optval; 582 socklen_t size_val; 583 584 size_val = sizeof(optval); 585 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 586 &optval, UIO_SYSSPACE, &size_val); 587 if (error) 588 return (error); 589 590 return (optval == 0); 591} 592 593/* 594 * Updated sendto() when IP_HDRINCL is set: 595 * tweak endian-dependent fields in the IP packet. 596 */ 597static int 598linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 599{ 600/* 601 * linux_ip_copysize defines how many bytes we should copy 602 * from the beginning of the IP packet before we customize it for BSD. 603 * It should include all the fields we modify (ip_len and ip_off). 604 */ 605#define linux_ip_copysize 8 606 607 struct ip *packet; 608 struct msghdr msg; 609 struct iovec aiov[1]; 610 int error; 611 612 /* Check that the packet isn't too big or too small. */ 613 if (linux_args->len < linux_ip_copysize || 614 linux_args->len > IP_MAXPACKET) 615 return (EINVAL); 616 617 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK); 618 619 /* Make kernel copy of the packet to be sent */ 620 if ((error = copyin(PTRIN(linux_args->msg), packet, 621 linux_args->len))) 622 goto goout; 623 624 /* Convert fields from Linux to BSD raw IP socket format */ 625 packet->ip_len = linux_args->len; 626 packet->ip_off = ntohs(packet->ip_off); 627 628 /* Prepare the msghdr and iovec structures describing the new packet */ 629 msg.msg_name = PTRIN(linux_args->to); 630 msg.msg_namelen = linux_args->tolen; 631 msg.msg_iov = aiov; 632 msg.msg_iovlen = 1; 633 msg.msg_control = NULL; 634 msg.msg_flags = 0; 635 aiov[0].iov_base = (char *)packet; 636 aiov[0].iov_len = linux_args->len; 637 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 638 NULL, UIO_SYSSPACE); 639goout: 640 free(packet, M_LINUX); 641 return (error); 642} 643 644int 645linux_socket(struct thread *td, struct linux_socket_args *args) 646{ 647 struct socket_args /* { 648 int domain; 649 int type; 650 int protocol; 651 } */ bsd_args; 652 int retval_socket, socket_flags; 653 654 bsd_args.protocol = args->protocol; 655 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK; 656 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 657 return (EINVAL); 658 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 659 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 660 return (EINVAL); 661 bsd_args.domain = linux_to_bsd_domain(args->domain); 662 if (bsd_args.domain == -1) 663 return (EAFNOSUPPORT); 664 665 retval_socket = sys_socket(td, &bsd_args); 666 if (retval_socket) 667 return (retval_socket); 668 669 retval_socket = linux_set_socket_flags(td, td->td_retval[0], 670 socket_flags); 671 if (retval_socket) { 672 (void)kern_close(td, td->td_retval[0]); 673 goto out; 674 } 675 676 if (bsd_args.type == SOCK_RAW 677 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0) 678 && bsd_args.domain == PF_INET) { 679 /* It's a raw IP socket: set the IP_HDRINCL option. */ 680 int hdrincl; 681 682 hdrincl = 1; 683 /* We ignore any error returned by kern_setsockopt() */ 684 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 685 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 686 } 687#ifdef INET6 688 /* 689 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default 690 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps. 691 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only 692 * sysctl value. 693 */ 694 if (bsd_args.domain == PF_INET6) { 695 int v6only; 696 697 v6only = 0; 698 /* We ignore any error returned by setsockopt() */ 699 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 700 &v6only, UIO_SYSSPACE, sizeof(v6only)); 701 } 702#endif 703 704out: 705 return (retval_socket); 706} 707 708int 709linux_bind(struct thread *td, struct linux_bind_args *args) 710{ 711 struct sockaddr *sa; 712 int error; 713 714 error = linux_getsockaddr(&sa, PTRIN(args->name), 715 args->namelen); 716 if (error) 717 return (error); 718 719 error = kern_bindat(td, AT_FDCWD, args->s, sa); 720 free(sa, M_SONAME); 721 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 722 return (EINVAL); 723 return (error); 724} 725 726int 727linux_connect(struct thread *td, struct linux_connect_args *args) 728{ 729 cap_rights_t rights; 730 struct socket *so; 731 struct sockaddr *sa; 732 u_int fflag; 733 int error; 734 735 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 736 args->namelen); 737 if (error) 738 return (error); 739 740 error = kern_connectat(td, AT_FDCWD, args->s, sa); 741 free(sa, M_SONAME); 742 if (error != EISCONN) 743 return (error); 744 745 /* 746 * Linux doesn't return EISCONN the first time it occurs, 747 * when on a non-blocking socket. Instead it returns the 748 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 749 * 750 * XXXRW: Instead of using fgetsock(), check that it is a 751 * socket and use the file descriptor reference instead of 752 * creating a new one. 753 */ 754 error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT), 755 &so, &fflag); 756 if (error == 0) { 757 error = EISCONN; 758 if (fflag & FNONBLOCK) { 759 SOCK_LOCK(so); 760 if (so->so_emuldata == 0) 761 error = so->so_error; 762 so->so_emuldata = (void *)1; 763 SOCK_UNLOCK(so); 764 } 765 fputsock(so); 766 } 767 return (error); 768} 769 770int 771linux_listen(struct thread *td, struct linux_listen_args *args) 772{ 773 struct listen_args /* { 774 int s; 775 int backlog; 776 } */ bsd_args; 777 778 bsd_args.s = args->s; 779 bsd_args.backlog = args->backlog; 780 return (sys_listen(td, &bsd_args)); 781} 782 783static int 784linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 785 l_uintptr_t namelen, int flags) 786{ 787 struct accept_args /* { 788 int s; 789 struct sockaddr * __restrict name; 790 socklen_t * __restrict anamelen; 791 } */ bsd_args; 792 int error; 793 794 if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 795 return (EINVAL); 796 797 bsd_args.s = s; 798 /* XXX: */ 799 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr); 800 bsd_args.anamelen = PTRIN(namelen);/* XXX */ 801 error = sys_accept(td, &bsd_args); 802 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 803 if (error) { 804 if (error == EFAULT && namelen != sizeof(struct sockaddr_in)) 805 return (EINVAL); 806 return (error); 807 } 808 809 /* 810 * linux appears not to copy flags from the parent socket to the 811 * accepted one, so we must clear the flags in the new descriptor 812 * and apply the requested flags. 813 */ 814 error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0); 815 if (error) 816 goto out; 817 error = linux_set_socket_flags(td, td->td_retval[0], flags); 818 if (error) 819 goto out; 820 if (addr) 821 error = linux_sa_put(PTRIN(addr)); 822 823out: 824 if (error) { 825 (void)kern_close(td, td->td_retval[0]); 826 td->td_retval[0] = 0; 827 } 828 return (error); 829} 830 831int 832linux_accept(struct thread *td, struct linux_accept_args *args) 833{ 834 835 return (linux_accept_common(td, args->s, args->addr, 836 args->namelen, 0)); 837} 838 839int 840linux_accept4(struct thread *td, struct linux_accept4_args *args) 841{ 842 843 return (linux_accept_common(td, args->s, args->addr, 844 args->namelen, args->flags)); 845} 846 847int 848linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 849{ 850 struct getsockname_args /* { 851 int fdes; 852 struct sockaddr * __restrict asa; 853 socklen_t * __restrict alen; 854 } */ bsd_args; 855 int error; 856 857 bsd_args.fdes = args->s; 858 /* XXX: */ 859 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 860 bsd_args.alen = PTRIN(args->namelen); /* XXX */ 861 error = sys_getsockname(td, &bsd_args); 862 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 863 if (error) 864 return (error); 865 error = linux_sa_put(PTRIN(args->addr)); 866 if (error) 867 return (error); 868 return (0); 869} 870 871int 872linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 873{ 874 struct getpeername_args /* { 875 int fdes; 876 caddr_t asa; 877 int *alen; 878 } */ bsd_args; 879 int error; 880 881 bsd_args.fdes = args->s; 882 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 883 bsd_args.alen = (socklen_t *)PTRIN(args->namelen); 884 error = sys_getpeername(td, &bsd_args); 885 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 886 if (error) 887 return (error); 888 error = linux_sa_put(PTRIN(args->addr)); 889 if (error) 890 return (error); 891 return (0); 892} 893 894int 895linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 896{ 897 struct socketpair_args /* { 898 int domain; 899 int type; 900 int protocol; 901 int *rsv; 902 } */ bsd_args; 903 int error, socket_flags; 904 int sv[2]; 905 906 bsd_args.domain = linux_to_bsd_domain(args->domain); 907 if (bsd_args.domain != PF_LOCAL) 908 return (EAFNOSUPPORT); 909 910 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK; 911 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 912 return (EINVAL); 913 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 914 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 915 return (EINVAL); 916 917 if (args->protocol != 0 && args->protocol != PF_UNIX) 918 919 /* 920 * Use of PF_UNIX as protocol argument is not right, 921 * but Linux does it. 922 * Do not map PF_UNIX as its Linux value is identical 923 * to FreeBSD one. 924 */ 925 return (EPROTONOSUPPORT); 926 else 927 bsd_args.protocol = 0; 928 bsd_args.rsv = (int *)PTRIN(args->rsv); 929 error = kern_socketpair(td, bsd_args.domain, bsd_args.type, 930 bsd_args.protocol, sv); 931 if (error) 932 return (error); 933 error = linux_set_socket_flags(td, sv[0], socket_flags); 934 if (error) 935 goto out; 936 error = linux_set_socket_flags(td, sv[1], socket_flags); 937 if (error) 938 goto out; 939 940 error = copyout(sv, bsd_args.rsv, 2 * sizeof(int)); 941 942out: 943 if (error) { 944 (void)kern_close(td, sv[0]); 945 (void)kern_close(td, sv[1]); 946 } 947 return (error); 948} 949 950#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 951struct linux_send_args { 952 int s; 953 l_uintptr_t msg; 954 int len; 955 int flags; 956}; 957 958static int 959linux_send(struct thread *td, struct linux_send_args *args) 960{ 961 struct sendto_args /* { 962 int s; 963 caddr_t buf; 964 int len; 965 int flags; 966 caddr_t to; 967 int tolen; 968 } */ bsd_args; 969 970 bsd_args.s = args->s; 971 bsd_args.buf = (caddr_t)PTRIN(args->msg); 972 bsd_args.len = args->len; 973 bsd_args.flags = args->flags; 974 bsd_args.to = NULL; 975 bsd_args.tolen = 0; 976 return sys_sendto(td, &bsd_args); 977} 978 979struct linux_recv_args { 980 int s; 981 l_uintptr_t msg; 982 int len; 983 int flags; 984}; 985 986static int 987linux_recv(struct thread *td, struct linux_recv_args *args) 988{ 989 struct recvfrom_args /* { 990 int s; 991 caddr_t buf; 992 int len; 993 int flags; 994 struct sockaddr *from; 995 socklen_t fromlenaddr; 996 } */ bsd_args; 997 998 bsd_args.s = args->s; 999 bsd_args.buf = (caddr_t)PTRIN(args->msg); 1000 bsd_args.len = args->len; 1001 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 1002 bsd_args.from = NULL; 1003 bsd_args.fromlenaddr = 0; 1004 return (sys_recvfrom(td, &bsd_args)); 1005} 1006#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1007 1008int 1009linux_sendto(struct thread *td, struct linux_sendto_args *args) 1010{ 1011 struct msghdr msg; 1012 struct iovec aiov; 1013 int error; 1014 1015 if (linux_check_hdrincl(td, args->s) == 0) 1016 /* IP_HDRINCL set, tweak the packet before sending */ 1017 return (linux_sendto_hdrincl(td, args)); 1018 1019 msg.msg_name = PTRIN(args->to); 1020 msg.msg_namelen = args->tolen; 1021 msg.msg_iov = &aiov; 1022 msg.msg_iovlen = 1; 1023 msg.msg_control = NULL; 1024 msg.msg_flags = 0; 1025 aiov.iov_base = PTRIN(args->msg); 1026 aiov.iov_len = args->len; 1027 error = linux_sendit(td, args->s, &msg, args->flags, NULL, 1028 UIO_USERSPACE); 1029 return (error); 1030} 1031 1032int 1033linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 1034{ 1035 struct msghdr msg; 1036 struct iovec aiov; 1037 int error; 1038 1039 if (PTRIN(args->fromlen) != NULL) { 1040 error = copyin(PTRIN(args->fromlen), &msg.msg_namelen, 1041 sizeof(msg.msg_namelen)); 1042 if (error != 0) 1043 return (error); 1044 1045 error = linux_to_bsd_sockaddr((struct sockaddr *)PTRIN(args->from), 1046 msg.msg_namelen); 1047 if (error != 0) 1048 return (error); 1049 } else 1050 msg.msg_namelen = 0; 1051 1052 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from); 1053 msg.msg_iov = &aiov; 1054 msg.msg_iovlen = 1; 1055 aiov.iov_base = PTRIN(args->buf); 1056 aiov.iov_len = args->len; 1057 msg.msg_control = 0; 1058 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 1059 1060 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL); 1061 if (error != 0) 1062 return (error); 1063 1064 if (PTRIN(args->from) != NULL) { 1065 error = bsd_to_linux_sockaddr((struct sockaddr *) 1066 PTRIN(args->from)); 1067 if (error != 0) 1068 return (error); 1069 1070 error = linux_sa_put((struct osockaddr *) 1071 PTRIN(args->from)); 1072 } 1073 1074 if (PTRIN(args->fromlen) != NULL) 1075 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 1076 sizeof(msg.msg_namelen)); 1077 1078 return (error); 1079} 1080 1081static int 1082linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1083 l_uint flags) 1084{ 1085 struct cmsghdr *cmsg; 1086 struct cmsgcred cmcred; 1087 struct mbuf *control; 1088 struct msghdr msg; 1089 struct l_cmsghdr linux_cmsg; 1090 struct l_cmsghdr *ptr_cmsg; 1091 struct l_msghdr linux_msg; 1092 struct iovec *iov; 1093 socklen_t datalen; 1094 struct sockaddr *sa; 1095 sa_family_t sa_family; 1096 void *data; 1097 int error; 1098 1099 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1100 if (error != 0) 1101 return (error); 1102 1103 /* 1104 * Some Linux applications (ping) define a non-NULL control data 1105 * pointer, but a msg_controllen of 0, which is not allowed in the 1106 * FreeBSD system call interface. NULL the msg_control pointer in 1107 * order to handle this case. This should be checked, but allows the 1108 * Linux ping to work. 1109 */ 1110 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0) 1111 linux_msg.msg_control = PTROUT(NULL); 1112 1113 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1114 if (error != 0) 1115 return (error); 1116 1117#ifdef COMPAT_LINUX32 1118 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1119 &iov, EMSGSIZE); 1120#else 1121 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1122#endif 1123 if (error != 0) 1124 return (error); 1125 1126 control = NULL; 1127 cmsg = NULL; 1128 1129 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) { 1130 error = kern_getsockname(td, s, &sa, &datalen); 1131 if (error != 0) 1132 goto bad; 1133 sa_family = sa->sa_family; 1134 free(sa, M_SONAME); 1135 1136 error = ENOBUFS; 1137 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO); 1138 control = m_get(M_WAITOK, MT_CONTROL); 1139 if (control == NULL) 1140 goto bad; 1141 1142 do { 1143 error = copyin(ptr_cmsg, &linux_cmsg, 1144 sizeof(struct l_cmsghdr)); 1145 if (error != 0) 1146 goto bad; 1147 1148 error = EINVAL; 1149 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1150 goto bad; 1151 1152 /* 1153 * Now we support only SCM_RIGHTS and SCM_CRED, 1154 * so return EINVAL in any other cmsg_type 1155 */ 1156 cmsg->cmsg_type = 1157 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1158 cmsg->cmsg_level = 1159 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1160 if (cmsg->cmsg_type == -1 1161 || cmsg->cmsg_level != SOL_SOCKET) 1162 goto bad; 1163 1164 /* 1165 * Some applications (e.g. pulseaudio) attempt to 1166 * send ancillary data even if the underlying protocol 1167 * doesn't support it which is not allowed in the 1168 * FreeBSD system call interface. 1169 */ 1170 if (sa_family != AF_UNIX) 1171 continue; 1172 1173 data = LINUX_CMSG_DATA(ptr_cmsg); 1174 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1175 1176 switch (cmsg->cmsg_type) 1177 { 1178 case SCM_RIGHTS: 1179 break; 1180 1181 case SCM_CREDS: 1182 data = &cmcred; 1183 datalen = sizeof(cmcred); 1184 1185 /* 1186 * The lower levels will fill in the structure 1187 */ 1188 bzero(data, datalen); 1189 break; 1190 } 1191 1192 cmsg->cmsg_len = CMSG_LEN(datalen); 1193 1194 error = ENOBUFS; 1195 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg)) 1196 goto bad; 1197 if (!m_append(control, datalen, (c_caddr_t)data)) 1198 goto bad; 1199 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg))); 1200 1201 if (m_length(control, NULL) == 0) { 1202 m_freem(control); 1203 control = NULL; 1204 } 1205 } 1206 1207 msg.msg_iov = iov; 1208 msg.msg_flags = 0; 1209 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1210 1211bad: 1212 free(iov, M_IOV); 1213 if (cmsg) 1214 free(cmsg, M_LINUX); 1215 return (error); 1216} 1217 1218int 1219linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1220{ 1221 1222 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1223 args->flags)); 1224} 1225 1226int 1227linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1228{ 1229 struct l_mmsghdr *msg; 1230 l_uint retval; 1231 int error, datagrams; 1232 1233 if (args->vlen > UIO_MAXIOV) 1234 args->vlen = UIO_MAXIOV; 1235 1236 msg = PTRIN(args->msg); 1237 datagrams = 0; 1238 while (datagrams < args->vlen) { 1239 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1240 args->flags); 1241 if (error != 0) 1242 break; 1243 1244 retval = td->td_retval[0]; 1245 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1246 if (error != 0) 1247 break; 1248 ++msg; 1249 ++datagrams; 1250 } 1251 if (error == 0) 1252 td->td_retval[0] = datagrams; 1253 return (error); 1254} 1255 1256static int 1257linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1258 l_uint flags, struct msghdr *msg) 1259{ 1260 struct cmsghdr *cm; 1261 struct cmsgcred *cmcred; 1262 struct l_cmsghdr *linux_cmsg = NULL; 1263 struct l_ucred linux_ucred; 1264 socklen_t datalen, outlen; 1265 struct l_msghdr linux_msg; 1266 struct iovec *iov, *uiov; 1267 struct mbuf *control = NULL; 1268 struct mbuf **controlp; 1269 caddr_t outbuf; 1270 void *data; 1271 int error, i, fd, fds, *fdp; 1272 1273 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1274 if (error != 0) 1275 return (error); 1276 1277 error = linux_to_bsd_msghdr(msg, &linux_msg); 1278 if (error != 0) 1279 return (error); 1280 1281#ifdef COMPAT_LINUX32 1282 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1283 &iov, EMSGSIZE); 1284#else 1285 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1286#endif 1287 if (error != 0) 1288 return (error); 1289 1290 if (msg->msg_name) { 1291 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name, 1292 msg->msg_namelen); 1293 if (error != 0) 1294 goto bad; 1295 } 1296 1297 uiov = msg->msg_iov; 1298 msg->msg_iov = iov; 1299 controlp = (msg->msg_control != NULL) ? &control : NULL; 1300 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp); 1301 msg->msg_iov = uiov; 1302 if (error != 0) 1303 goto bad; 1304 1305 error = bsd_to_linux_msghdr(msg, &linux_msg); 1306 if (error != 0) 1307 goto bad; 1308 1309 if (linux_msg.msg_name) { 1310 error = bsd_to_linux_sockaddr((struct sockaddr *) 1311 PTRIN(linux_msg.msg_name)); 1312 if (error != 0) 1313 goto bad; 1314 } 1315 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1316 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1317 if (error != 0) 1318 goto bad; 1319 } 1320 1321 outbuf = PTRIN(linux_msg.msg_control); 1322 outlen = 0; 1323 1324 if (control) { 1325 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1326 1327 msg->msg_control = mtod(control, struct cmsghdr *); 1328 msg->msg_controllen = control->m_len; 1329 1330 cm = CMSG_FIRSTHDR(msg); 1331 1332 while (cm != NULL) { 1333 linux_cmsg->cmsg_type = 1334 bsd_to_linux_cmsg_type(cm->cmsg_type); 1335 linux_cmsg->cmsg_level = 1336 bsd_to_linux_sockopt_level(cm->cmsg_level); 1337 if (linux_cmsg->cmsg_type == -1 1338 || cm->cmsg_level != SOL_SOCKET) 1339 { 1340 error = EINVAL; 1341 goto bad; 1342 } 1343 1344 data = CMSG_DATA(cm); 1345 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1346 1347 switch (cm->cmsg_type) 1348 { 1349 case SCM_RIGHTS: 1350 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1351 fds = datalen / sizeof(int); 1352 fdp = data; 1353 for (i = 0; i < fds; i++) { 1354 fd = *fdp++; 1355 (void)kern_fcntl(td, fd, 1356 F_SETFD, FD_CLOEXEC); 1357 } 1358 } 1359 break; 1360 1361 case SCM_CREDS: 1362 /* 1363 * Currently LOCAL_CREDS is never in 1364 * effect for Linux so no need to worry 1365 * about sockcred 1366 */ 1367 if (datalen != sizeof(*cmcred)) { 1368 error = EMSGSIZE; 1369 goto bad; 1370 } 1371 cmcred = (struct cmsgcred *)data; 1372 bzero(&linux_ucred, sizeof(linux_ucred)); 1373 linux_ucred.pid = cmcred->cmcred_pid; 1374 linux_ucred.uid = cmcred->cmcred_uid; 1375 linux_ucred.gid = cmcred->cmcred_gid; 1376 data = &linux_ucred; 1377 datalen = sizeof(linux_ucred); 1378 break; 1379 } 1380 1381 if (outlen + LINUX_CMSG_LEN(datalen) > 1382 linux_msg.msg_controllen) { 1383 if (outlen == 0) { 1384 error = EMSGSIZE; 1385 goto bad; 1386 } else { 1387 linux_msg.msg_flags |= 1388 LINUX_MSG_CTRUNC; 1389 goto out; 1390 } 1391 } 1392 1393 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1394 1395 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1396 if (error) 1397 goto bad; 1398 outbuf += L_CMSG_HDRSZ; 1399 1400 error = copyout(data, outbuf, datalen); 1401 if (error) 1402 goto bad; 1403 1404 outbuf += LINUX_CMSG_ALIGN(datalen); 1405 outlen += LINUX_CMSG_LEN(datalen); 1406 1407 cm = CMSG_NXTHDR(msg, cm); 1408 } 1409 } 1410 1411out: 1412 linux_msg.msg_controllen = outlen; 1413 error = copyout(&linux_msg, msghdr, sizeof(linux_msg)); 1414 1415bad: 1416 free(iov, M_IOV); 1417 m_freem(control); 1418 free(linux_cmsg, M_LINUX); 1419 1420 return (error); 1421} 1422 1423int 1424linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1425{ 1426 struct msghdr bsd_msg; 1427 1428 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1429 args->flags, &bsd_msg)); 1430} 1431 1432int 1433linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1434{ 1435 struct l_mmsghdr *msg; 1436 struct msghdr bsd_msg; 1437 struct l_timespec lts; 1438 struct timespec ts, tts; 1439 l_uint retval; 1440 int error, datagrams; 1441 1442 if (args->timeout) { 1443 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1444 if (error != 0) 1445 return (error); 1446 error = linux_to_native_timespec(&ts, <s); 1447 if (error != 0) 1448 return (error); 1449 getnanotime(&tts); 1450 timespecadd(&tts, &ts); 1451 } 1452 1453 msg = PTRIN(args->msg); 1454 datagrams = 0; 1455 while (datagrams < args->vlen) { 1456 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1457 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1458 if (error != 0) 1459 break; 1460 1461 retval = td->td_retval[0]; 1462 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1463 if (error != 0) 1464 break; 1465 ++msg; 1466 ++datagrams; 1467 1468 /* 1469 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1470 */ 1471 if (args->flags & LINUX_MSG_WAITFORONE) 1472 args->flags |= LINUX_MSG_DONTWAIT; 1473 1474 /* 1475 * See BUGS section of recvmmsg(2). 1476 */ 1477 if (args->timeout) { 1478 getnanotime(&ts); 1479 timespecsub(&ts, &tts); 1480 if (!timespecisset(&ts) || ts.tv_sec > 0) 1481 break; 1482 } 1483 /* Out of band data, return right away. */ 1484 if (bsd_msg.msg_flags & MSG_OOB) 1485 break; 1486 } 1487 if (error == 0) 1488 td->td_retval[0] = datagrams; 1489 return (error); 1490} 1491 1492int 1493linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1494{ 1495 struct shutdown_args /* { 1496 int s; 1497 int how; 1498 } */ bsd_args; 1499 1500 bsd_args.s = args->s; 1501 bsd_args.how = args->how; 1502 return (sys_shutdown(td, &bsd_args)); 1503} 1504 1505int 1506linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1507{ 1508 struct setsockopt_args /* { 1509 int s; 1510 int level; 1511 int name; 1512 caddr_t val; 1513 int valsize; 1514 } */ bsd_args; 1515 l_timeval linux_tv; 1516 struct timeval tv; 1517 int error, name; 1518 1519 bsd_args.s = args->s; 1520 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1521 switch (bsd_args.level) { 1522 case SOL_SOCKET: 1523 name = linux_to_bsd_so_sockopt(args->optname); 1524 switch (name) { 1525 case SO_RCVTIMEO: 1526 /* FALLTHROUGH */ 1527 case SO_SNDTIMEO: 1528 error = copyin(PTRIN(args->optval), &linux_tv, 1529 sizeof(linux_tv)); 1530 if (error) 1531 return (error); 1532 tv.tv_sec = linux_tv.tv_sec; 1533 tv.tv_usec = linux_tv.tv_usec; 1534 return (kern_setsockopt(td, args->s, bsd_args.level, 1535 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1536 /* NOTREACHED */ 1537 break; 1538 default: 1539 break; 1540 } 1541 break; 1542 case IPPROTO_IP: 1543 name = linux_to_bsd_ip_sockopt(args->optname); 1544 break; 1545 case IPPROTO_TCP: 1546 name = linux_to_bsd_tcp_sockopt(args->optname); 1547 break; 1548 default: 1549 name = -1; 1550 break; 1551 } 1552 if (name == -1) 1553 return (ENOPROTOOPT); 1554 1555 bsd_args.name = name; 1556 bsd_args.val = PTRIN(args->optval); 1557 bsd_args.valsize = args->optlen; 1558 1559 if (name == IPV6_NEXTHOP) { 1560 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1561 bsd_args.valsize); 1562 error = sys_setsockopt(td, &bsd_args); 1563 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1564 } else 1565 error = sys_setsockopt(td, &bsd_args); 1566 1567 return (error); 1568} 1569 1570int 1571linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1572{ 1573 struct getsockopt_args /* { 1574 int s; 1575 int level; 1576 int name; 1577 caddr_t val; 1578 int *avalsize; 1579 } */ bsd_args; 1580 l_timeval linux_tv; 1581 struct timeval tv; 1582 socklen_t tv_len, xulen; 1583 struct xucred xu; 1584 struct l_ucred lxu; 1585 int error, name; 1586 1587 bsd_args.s = args->s; 1588 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1589 switch (bsd_args.level) { 1590 case SOL_SOCKET: 1591 name = linux_to_bsd_so_sockopt(args->optname); 1592 switch (name) { 1593 case SO_RCVTIMEO: 1594 /* FALLTHROUGH */ 1595 case SO_SNDTIMEO: 1596 tv_len = sizeof(tv); 1597 error = kern_getsockopt(td, args->s, bsd_args.level, 1598 name, &tv, UIO_SYSSPACE, &tv_len); 1599 if (error) 1600 return (error); 1601 linux_tv.tv_sec = tv.tv_sec; 1602 linux_tv.tv_usec = tv.tv_usec; 1603 return (copyout(&linux_tv, PTRIN(args->optval), 1604 sizeof(linux_tv))); 1605 /* NOTREACHED */ 1606 break; 1607 case LOCAL_PEERCRED: 1608 if (args->optlen != sizeof(lxu)) 1609 return (EINVAL); 1610 xulen = sizeof(xu); 1611 error = kern_getsockopt(td, args->s, bsd_args.level, 1612 name, &xu, UIO_SYSSPACE, &xulen); 1613 if (error) 1614 return (error); 1615 /* 1616 * XXX Use 0 for pid as the FreeBSD does not cache peer pid. 1617 */ 1618 lxu.pid = 0; 1619 lxu.uid = xu.cr_uid; 1620 lxu.gid = xu.cr_gid; 1621 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1622 /* NOTREACHED */ 1623 break; 1624 default: 1625 break; 1626 } 1627 break; 1628 case IPPROTO_IP: 1629 name = linux_to_bsd_ip_sockopt(args->optname); 1630 break; 1631 case IPPROTO_TCP: 1632 name = linux_to_bsd_tcp_sockopt(args->optname); 1633 break; 1634 default: 1635 name = -1; 1636 break; 1637 } 1638 if (name == -1) 1639 return (EINVAL); 1640 1641 bsd_args.name = name; 1642 bsd_args.val = PTRIN(args->optval); 1643 bsd_args.avalsize = PTRIN(args->optlen); 1644 1645 if (name == IPV6_NEXTHOP) { 1646 error = sys_getsockopt(td, &bsd_args); 1647 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1648 } else 1649 error = sys_getsockopt(td, &bsd_args); 1650 1651 return (error); 1652} 1653 1654#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1655 1656/* Argument list sizes for linux_socketcall */ 1657 1658#define LINUX_AL(x) ((x) * sizeof(l_ulong)) 1659 1660static const unsigned char lxs_args[] = { 1661 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */, 1662 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */, 1663 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */, 1664 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */, 1665 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */, 1666 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */, 1667 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */, 1668 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */, 1669 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */, 1670 LINUX_AL(4) /* accept4 */, LINUX_AL(5) /* recvmmsg */, 1671 LINUX_AL(4) /* sendmmsg */ 1672}; 1673 1674#define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1 1675 1676int 1677linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1678{ 1679 l_ulong a[6]; 1680 void *arg; 1681 int error; 1682 1683 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE) 1684 return (EINVAL); 1685 error = copyin(PTRIN(args->args), a, lxs_args[args->what]); 1686 if (error) 1687 return (error); 1688 1689 arg = a; 1690 switch (args->what) { 1691 case LINUX_SOCKET: 1692 return (linux_socket(td, arg)); 1693 case LINUX_BIND: 1694 return (linux_bind(td, arg)); 1695 case LINUX_CONNECT: 1696 return (linux_connect(td, arg)); 1697 case LINUX_LISTEN: 1698 return (linux_listen(td, arg)); 1699 case LINUX_ACCEPT: 1700 return (linux_accept(td, arg)); 1701 case LINUX_GETSOCKNAME: 1702 return (linux_getsockname(td, arg)); 1703 case LINUX_GETPEERNAME: 1704 return (linux_getpeername(td, arg)); 1705 case LINUX_SOCKETPAIR: 1706 return (linux_socketpair(td, arg)); 1707 case LINUX_SEND: 1708 return (linux_send(td, arg)); 1709 case LINUX_RECV: 1710 return (linux_recv(td, arg)); 1711 case LINUX_SENDTO: 1712 return (linux_sendto(td, arg)); 1713 case LINUX_RECVFROM: 1714 return (linux_recvfrom(td, arg)); 1715 case LINUX_SHUTDOWN: 1716 return (linux_shutdown(td, arg)); 1717 case LINUX_SETSOCKOPT: 1718 return (linux_setsockopt(td, arg)); 1719 case LINUX_GETSOCKOPT: 1720 return (linux_getsockopt(td, arg)); 1721 case LINUX_SENDMSG: 1722 return (linux_sendmsg(td, arg)); 1723 case LINUX_RECVMSG: 1724 return (linux_recvmsg(td, arg)); 1725 case LINUX_ACCEPT4: 1726 return (linux_accept4(td, arg)); 1727 case LINUX_RECVMMSG: 1728 return (linux_recvmmsg(td, arg)); 1729 case LINUX_SENDMMSG: 1730 return (linux_sendmmsg(td, arg)); 1731 } 1732 1733 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1734 return (ENOSYS); 1735} 1736#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1737