linux_socket.c revision 296503
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 296503 2016-03-08 15:15:34Z 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_so_sockopt(int opt) 293{ 294 295 switch (opt) { 296 case LINUX_SO_DEBUG: 297 return (SO_DEBUG); 298 case LINUX_SO_REUSEADDR: 299 return (SO_REUSEADDR); 300 case LINUX_SO_TYPE: 301 return (SO_TYPE); 302 case LINUX_SO_ERROR: 303 return (SO_ERROR); 304 case LINUX_SO_DONTROUTE: 305 return (SO_DONTROUTE); 306 case LINUX_SO_BROADCAST: 307 return (SO_BROADCAST); 308 case LINUX_SO_SNDBUF: 309 return (SO_SNDBUF); 310 case LINUX_SO_RCVBUF: 311 return (SO_RCVBUF); 312 case LINUX_SO_KEEPALIVE: 313 return (SO_KEEPALIVE); 314 case LINUX_SO_OOBINLINE: 315 return (SO_OOBINLINE); 316 case LINUX_SO_LINGER: 317 return (SO_LINGER); 318 case LINUX_SO_PEERCRED: 319 return (LOCAL_PEERCRED); 320 case LINUX_SO_RCVLOWAT: 321 return (SO_RCVLOWAT); 322 case LINUX_SO_SNDLOWAT: 323 return (SO_SNDLOWAT); 324 case LINUX_SO_RCVTIMEO: 325 return (SO_RCVTIMEO); 326 case LINUX_SO_SNDTIMEO: 327 return (SO_SNDTIMEO); 328 case LINUX_SO_TIMESTAMP: 329 return (SO_TIMESTAMP); 330 case LINUX_SO_ACCEPTCONN: 331 return (SO_ACCEPTCONN); 332 } 333 return (-1); 334} 335 336static int 337linux_to_bsd_tcp_sockopt(int opt) 338{ 339 340 switch (opt) { 341 case LINUX_TCP_NODELAY: 342 return (TCP_NODELAY); 343 case LINUX_TCP_MAXSEG: 344 return (TCP_MAXSEG); 345 case LINUX_TCP_KEEPIDLE: 346 return (TCP_KEEPIDLE); 347 case LINUX_TCP_KEEPINTVL: 348 return (TCP_KEEPINTVL); 349 case LINUX_TCP_KEEPCNT: 350 return (TCP_KEEPCNT); 351 case LINUX_TCP_MD5SIG: 352 return (TCP_MD5SIG); 353 } 354 return (-1); 355} 356 357static int 358linux_to_bsd_msg_flags(int flags) 359{ 360 int ret_flags = 0; 361 362 if (flags & LINUX_MSG_OOB) 363 ret_flags |= MSG_OOB; 364 if (flags & LINUX_MSG_PEEK) 365 ret_flags |= MSG_PEEK; 366 if (flags & LINUX_MSG_DONTROUTE) 367 ret_flags |= MSG_DONTROUTE; 368 if (flags & LINUX_MSG_CTRUNC) 369 ret_flags |= MSG_CTRUNC; 370 if (flags & LINUX_MSG_TRUNC) 371 ret_flags |= MSG_TRUNC; 372 if (flags & LINUX_MSG_DONTWAIT) 373 ret_flags |= MSG_DONTWAIT; 374 if (flags & LINUX_MSG_EOR) 375 ret_flags |= MSG_EOR; 376 if (flags & LINUX_MSG_WAITALL) 377 ret_flags |= MSG_WAITALL; 378 if (flags & LINUX_MSG_NOSIGNAL) 379 ret_flags |= MSG_NOSIGNAL; 380#if 0 /* not handled */ 381 if (flags & LINUX_MSG_PROXY) 382 ; 383 if (flags & LINUX_MSG_FIN) 384 ; 385 if (flags & LINUX_MSG_SYN) 386 ; 387 if (flags & LINUX_MSG_CONFIRM) 388 ; 389 if (flags & LINUX_MSG_RST) 390 ; 391 if (flags & LINUX_MSG_ERRQUEUE) 392 ; 393#endif 394 return ret_flags; 395} 396 397/* 398* If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the 399* native syscall will fault. Thus, we don't really need to check the 400* return values for these functions. 401*/ 402 403static int 404bsd_to_linux_sockaddr(struct sockaddr *arg) 405{ 406 struct sockaddr sa; 407 size_t sa_len = sizeof(struct sockaddr); 408 int error; 409 410 if ((error = copyin(arg, &sa, sa_len))) 411 return (error); 412 413 *(u_short *)&sa = sa.sa_family; 414 415 error = copyout(&sa, arg, sa_len); 416 417 return (error); 418} 419 420static int 421linux_to_bsd_sockaddr(struct sockaddr *arg, int len) 422{ 423 struct sockaddr sa; 424 size_t sa_len = sizeof(struct sockaddr); 425 int error; 426 427 if ((error = copyin(arg, &sa, sa_len))) 428 return (error); 429 430 sa.sa_family = *(sa_family_t *)&sa; 431 sa.sa_len = len; 432 433 error = copyout(&sa, arg, sa_len); 434 435 return (error); 436} 437 438static int 439linux_sa_put(struct osockaddr *osa) 440{ 441 struct osockaddr sa; 442 int error, bdom; 443 444 /* 445 * Only read/write the osockaddr family part, the rest is 446 * not changed. 447 */ 448 error = copyin(osa, &sa, sizeof(sa.sa_family)); 449 if (error) 450 return (error); 451 452 bdom = bsd_to_linux_domain(sa.sa_family); 453 if (bdom == -1) 454 return (EINVAL); 455 456 sa.sa_family = bdom; 457 error = copyout(&sa, osa, sizeof(sa.sa_family)); 458 if (error) 459 return (error); 460 461 return (0); 462} 463 464static int 465linux_to_bsd_cmsg_type(int cmsg_type) 466{ 467 468 switch (cmsg_type) { 469 case LINUX_SCM_RIGHTS: 470 return (SCM_RIGHTS); 471 case LINUX_SCM_CREDENTIALS: 472 return (SCM_CREDS); 473 } 474 return (-1); 475} 476 477static int 478bsd_to_linux_cmsg_type(int cmsg_type) 479{ 480 481 switch (cmsg_type) { 482 case SCM_RIGHTS: 483 return (LINUX_SCM_RIGHTS); 484 case SCM_CREDS: 485 return (LINUX_SCM_CREDENTIALS); 486 case SCM_TIMESTAMP: 487 return (LINUX_SCM_TIMESTAMP); 488 } 489 return (-1); 490} 491 492static int 493linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 494{ 495 if (lhdr->msg_controllen > INT_MAX) 496 return (ENOBUFS); 497 498 bhdr->msg_name = PTRIN(lhdr->msg_name); 499 bhdr->msg_namelen = lhdr->msg_namelen; 500 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 501 bhdr->msg_iovlen = lhdr->msg_iovlen; 502 bhdr->msg_control = PTRIN(lhdr->msg_control); 503 504 /* 505 * msg_controllen is skipped since BSD and LINUX control messages 506 * are potentially different sizes (e.g. the cred structure used 507 * by SCM_CREDS is different between the two operating system). 508 * 509 * The caller can set it (if necessary) after converting all the 510 * control messages. 511 */ 512 513 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 514 return (0); 515} 516 517static int 518bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 519{ 520 lhdr->msg_name = PTROUT(bhdr->msg_name); 521 lhdr->msg_namelen = bhdr->msg_namelen; 522 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 523 lhdr->msg_iovlen = bhdr->msg_iovlen; 524 lhdr->msg_control = PTROUT(bhdr->msg_control); 525 526 /* 527 * msg_controllen is skipped since BSD and LINUX control messages 528 * are potentially different sizes (e.g. the cred structure used 529 * by SCM_CREDS is different between the two operating system). 530 * 531 * The caller can set it (if necessary) after converting all the 532 * control messages. 533 */ 534 535 /* msg_flags skipped */ 536 return (0); 537} 538 539static int 540linux_set_socket_flags(int lflags, int *flags) 541{ 542 543 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 544 return (EINVAL); 545 if (lflags & LINUX_SOCK_NONBLOCK) 546 *flags |= SOCK_NONBLOCK; 547 if (lflags & LINUX_SOCK_CLOEXEC) 548 *flags |= SOCK_CLOEXEC; 549 return (0); 550} 551 552static int 553linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 554 struct mbuf *control, enum uio_seg segflg) 555{ 556 struct sockaddr *to; 557 int error; 558 559 if (mp->msg_name != NULL) { 560 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen); 561 if (error) 562 return (error); 563 mp->msg_name = to; 564 } else 565 to = NULL; 566 567 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 568 segflg); 569 570 if (to) 571 free(to, M_SONAME); 572 return (error); 573} 574 575/* Return 0 if IP_HDRINCL is set for the given socket. */ 576static int 577linux_check_hdrincl(struct thread *td, int s) 578{ 579 int error, optval; 580 socklen_t size_val; 581 582 size_val = sizeof(optval); 583 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 584 &optval, UIO_SYSSPACE, &size_val); 585 if (error) 586 return (error); 587 588 return (optval == 0); 589} 590 591/* 592 * Updated sendto() when IP_HDRINCL is set: 593 * tweak endian-dependent fields in the IP packet. 594 */ 595static int 596linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 597{ 598/* 599 * linux_ip_copysize defines how many bytes we should copy 600 * from the beginning of the IP packet before we customize it for BSD. 601 * It should include all the fields we modify (ip_len and ip_off). 602 */ 603#define linux_ip_copysize 8 604 605 struct ip *packet; 606 struct msghdr msg; 607 struct iovec aiov[1]; 608 int error; 609 610 /* Check that the packet isn't too big or too small. */ 611 if (linux_args->len < linux_ip_copysize || 612 linux_args->len > IP_MAXPACKET) 613 return (EINVAL); 614 615 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK); 616 617 /* Make kernel copy of the packet to be sent */ 618 if ((error = copyin(PTRIN(linux_args->msg), packet, 619 linux_args->len))) 620 goto goout; 621 622 /* Convert fields from Linux to BSD raw IP socket format */ 623 packet->ip_len = linux_args->len; 624 packet->ip_off = ntohs(packet->ip_off); 625 626 /* Prepare the msghdr and iovec structures describing the new packet */ 627 msg.msg_name = PTRIN(linux_args->to); 628 msg.msg_namelen = linux_args->tolen; 629 msg.msg_iov = aiov; 630 msg.msg_iovlen = 1; 631 msg.msg_control = NULL; 632 msg.msg_flags = 0; 633 aiov[0].iov_base = (char *)packet; 634 aiov[0].iov_len = linux_args->len; 635 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 636 NULL, UIO_SYSSPACE); 637goout: 638 free(packet, M_LINUX); 639 return (error); 640} 641 642int 643linux_socket(struct thread *td, struct linux_socket_args *args) 644{ 645 struct socket_args /* { 646 int domain; 647 int type; 648 int protocol; 649 } */ bsd_args; 650 int retval_socket; 651 652 bsd_args.protocol = args->protocol; 653 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 654 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 655 return (EINVAL); 656 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 657 &bsd_args.type); 658 if (retval_socket != 0) 659 return (retval_socket); 660 bsd_args.domain = linux_to_bsd_domain(args->domain); 661 if (bsd_args.domain == -1) 662 return (EAFNOSUPPORT); 663 664 retval_socket = sys_socket(td, &bsd_args); 665 if (retval_socket) 666 return (retval_socket); 667 668 if (bsd_args.type == SOCK_RAW 669 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0) 670 && bsd_args.domain == PF_INET) { 671 /* It's a raw IP socket: set the IP_HDRINCL option. */ 672 int hdrincl; 673 674 hdrincl = 1; 675 /* We ignore any error returned by kern_setsockopt() */ 676 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 677 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 678 } 679#ifdef INET6 680 /* 681 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default 682 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps. 683 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only 684 * sysctl value. 685 */ 686 if (bsd_args.domain == PF_INET6) { 687 int v6only; 688 689 v6only = 0; 690 /* We ignore any error returned by setsockopt() */ 691 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 692 &v6only, UIO_SYSSPACE, sizeof(v6only)); 693 } 694#endif 695 696 return (retval_socket); 697} 698 699int 700linux_bind(struct thread *td, struct linux_bind_args *args) 701{ 702 struct sockaddr *sa; 703 int error; 704 705 error = linux_getsockaddr(&sa, PTRIN(args->name), 706 args->namelen); 707 if (error) 708 return (error); 709 710 error = kern_bindat(td, AT_FDCWD, args->s, sa); 711 free(sa, M_SONAME); 712 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 713 return (EINVAL); 714 return (error); 715} 716 717int 718linux_connect(struct thread *td, struct linux_connect_args *args) 719{ 720 cap_rights_t rights; 721 struct socket *so; 722 struct sockaddr *sa; 723 u_int fflag; 724 int error; 725 726 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 727 args->namelen); 728 if (error) 729 return (error); 730 731 error = kern_connectat(td, AT_FDCWD, args->s, sa); 732 free(sa, M_SONAME); 733 if (error != EISCONN) 734 return (error); 735 736 /* 737 * Linux doesn't return EISCONN the first time it occurs, 738 * when on a non-blocking socket. Instead it returns the 739 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 740 * 741 * XXXRW: Instead of using fgetsock(), check that it is a 742 * socket and use the file descriptor reference instead of 743 * creating a new one. 744 */ 745 error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT), 746 &so, &fflag); 747 if (error == 0) { 748 error = EISCONN; 749 if (fflag & FNONBLOCK) { 750 SOCK_LOCK(so); 751 if (so->so_emuldata == 0) 752 error = so->so_error; 753 so->so_emuldata = (void *)1; 754 SOCK_UNLOCK(so); 755 } 756 fputsock(so); 757 } 758 return (error); 759} 760 761int 762linux_listen(struct thread *td, struct linux_listen_args *args) 763{ 764 struct listen_args /* { 765 int s; 766 int backlog; 767 } */ bsd_args; 768 769 bsd_args.s = args->s; 770 bsd_args.backlog = args->backlog; 771 return (sys_listen(td, &bsd_args)); 772} 773 774static int 775linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 776 l_uintptr_t namelen, int flags) 777{ 778 struct accept4_args /* { 779 int s; 780 struct sockaddr * __restrict name; 781 socklen_t * __restrict anamelen; 782 int flags; 783 } */ bsd_args; 784 cap_rights_t rights; 785 struct socket *so; 786 struct file *fp; 787 int error, error1; 788 789 bsd_args.s = s; 790 /* XXX: */ 791 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr); 792 bsd_args.anamelen = PTRIN(namelen);/* XXX */ 793 bsd_args.flags = 0; 794 error = linux_set_socket_flags(flags, &bsd_args.flags); 795 if (error != 0) 796 return (error); 797 error = sys_accept4(td, &bsd_args); 798 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 799 if (error) { 800 if (error == EFAULT && namelen != sizeof(struct sockaddr_in)) 801 return (EINVAL); 802 if (error == EINVAL) { 803 error1 = getsock_cap(td, s, &rights, &fp, NULL); 804 if (error1 != 0) 805 return (error1); 806 so = (struct socket *)fp->f_data; 807 if (so->so_type == SOCK_DGRAM) 808 return (EOPNOTSUPP); 809 } 810 return (error); 811 } 812 if (addr) 813 error = linux_sa_put(PTRIN(addr)); 814 if (error) { 815 (void)kern_close(td, td->td_retval[0]); 816 td->td_retval[0] = 0; 817 } 818 return (error); 819} 820 821int 822linux_accept(struct thread *td, struct linux_accept_args *args) 823{ 824 825 return (linux_accept_common(td, args->s, args->addr, 826 args->namelen, 0)); 827} 828 829int 830linux_accept4(struct thread *td, struct linux_accept4_args *args) 831{ 832 833 return (linux_accept_common(td, args->s, args->addr, 834 args->namelen, args->flags)); 835} 836 837int 838linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 839{ 840 struct getsockname_args /* { 841 int fdes; 842 struct sockaddr * __restrict asa; 843 socklen_t * __restrict alen; 844 } */ bsd_args; 845 int error; 846 847 bsd_args.fdes = args->s; 848 /* XXX: */ 849 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 850 bsd_args.alen = PTRIN(args->namelen); /* XXX */ 851 error = sys_getsockname(td, &bsd_args); 852 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 853 if (error) 854 return (error); 855 error = linux_sa_put(PTRIN(args->addr)); 856 if (error) 857 return (error); 858 return (0); 859} 860 861int 862linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 863{ 864 struct getpeername_args /* { 865 int fdes; 866 caddr_t asa; 867 int *alen; 868 } */ bsd_args; 869 int error; 870 871 bsd_args.fdes = args->s; 872 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 873 bsd_args.alen = (socklen_t *)PTRIN(args->namelen); 874 error = sys_getpeername(td, &bsd_args); 875 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 876 if (error) 877 return (error); 878 error = linux_sa_put(PTRIN(args->addr)); 879 if (error) 880 return (error); 881 return (0); 882} 883 884int 885linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 886{ 887 struct socketpair_args /* { 888 int domain; 889 int type; 890 int protocol; 891 int *rsv; 892 } */ bsd_args; 893 int error; 894 895 bsd_args.domain = linux_to_bsd_domain(args->domain); 896 if (bsd_args.domain != PF_LOCAL) 897 return (EAFNOSUPPORT); 898 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 899 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 900 return (EINVAL); 901 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 902 &bsd_args.type); 903 if (error != 0) 904 return (error); 905 if (args->protocol != 0 && args->protocol != PF_UNIX) 906 907 /* 908 * Use of PF_UNIX as protocol argument is not right, 909 * but Linux does it. 910 * Do not map PF_UNIX as its Linux value is identical 911 * to FreeBSD one. 912 */ 913 return (EPROTONOSUPPORT); 914 else 915 bsd_args.protocol = 0; 916 bsd_args.rsv = (int *)PTRIN(args->rsv); 917 return (sys_socketpair(td, &bsd_args)); 918} 919 920#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 921struct linux_send_args { 922 int s; 923 l_uintptr_t msg; 924 int len; 925 int flags; 926}; 927 928static int 929linux_send(struct thread *td, struct linux_send_args *args) 930{ 931 struct sendto_args /* { 932 int s; 933 caddr_t buf; 934 int len; 935 int flags; 936 caddr_t to; 937 int tolen; 938 } */ bsd_args; 939 940 bsd_args.s = args->s; 941 bsd_args.buf = (caddr_t)PTRIN(args->msg); 942 bsd_args.len = args->len; 943 bsd_args.flags = args->flags; 944 bsd_args.to = NULL; 945 bsd_args.tolen = 0; 946 return sys_sendto(td, &bsd_args); 947} 948 949struct linux_recv_args { 950 int s; 951 l_uintptr_t msg; 952 int len; 953 int flags; 954}; 955 956static int 957linux_recv(struct thread *td, struct linux_recv_args *args) 958{ 959 struct recvfrom_args /* { 960 int s; 961 caddr_t buf; 962 int len; 963 int flags; 964 struct sockaddr *from; 965 socklen_t fromlenaddr; 966 } */ bsd_args; 967 968 bsd_args.s = args->s; 969 bsd_args.buf = (caddr_t)PTRIN(args->msg); 970 bsd_args.len = args->len; 971 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 972 bsd_args.from = NULL; 973 bsd_args.fromlenaddr = 0; 974 return (sys_recvfrom(td, &bsd_args)); 975} 976#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 977 978int 979linux_sendto(struct thread *td, struct linux_sendto_args *args) 980{ 981 struct msghdr msg; 982 struct iovec aiov; 983 int error; 984 985 if (linux_check_hdrincl(td, args->s) == 0) 986 /* IP_HDRINCL set, tweak the packet before sending */ 987 return (linux_sendto_hdrincl(td, args)); 988 989 msg.msg_name = PTRIN(args->to); 990 msg.msg_namelen = args->tolen; 991 msg.msg_iov = &aiov; 992 msg.msg_iovlen = 1; 993 msg.msg_control = NULL; 994 msg.msg_flags = 0; 995 aiov.iov_base = PTRIN(args->msg); 996 aiov.iov_len = args->len; 997 error = linux_sendit(td, args->s, &msg, args->flags, NULL, 998 UIO_USERSPACE); 999 return (error); 1000} 1001 1002int 1003linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 1004{ 1005 struct msghdr msg; 1006 struct iovec aiov; 1007 int error; 1008 1009 if (PTRIN(args->fromlen) != NULL) { 1010 error = copyin(PTRIN(args->fromlen), &msg.msg_namelen, 1011 sizeof(msg.msg_namelen)); 1012 if (error != 0) 1013 return (error); 1014 1015 error = linux_to_bsd_sockaddr((struct sockaddr *)PTRIN(args->from), 1016 msg.msg_namelen); 1017 if (error != 0) 1018 return (error); 1019 } else 1020 msg.msg_namelen = 0; 1021 1022 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from); 1023 msg.msg_iov = &aiov; 1024 msg.msg_iovlen = 1; 1025 aiov.iov_base = PTRIN(args->buf); 1026 aiov.iov_len = args->len; 1027 msg.msg_control = 0; 1028 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 1029 1030 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL); 1031 if (error != 0) 1032 return (error); 1033 1034 if (PTRIN(args->from) != NULL) { 1035 error = bsd_to_linux_sockaddr((struct sockaddr *) 1036 PTRIN(args->from)); 1037 if (error != 0) 1038 return (error); 1039 1040 error = linux_sa_put((struct osockaddr *) 1041 PTRIN(args->from)); 1042 } 1043 1044 if (PTRIN(args->fromlen) != NULL) 1045 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 1046 sizeof(msg.msg_namelen)); 1047 1048 return (error); 1049} 1050 1051static int 1052linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1053 l_uint flags) 1054{ 1055 struct cmsghdr *cmsg; 1056 struct cmsgcred cmcred; 1057 struct mbuf *control; 1058 struct msghdr msg; 1059 struct l_cmsghdr linux_cmsg; 1060 struct l_cmsghdr *ptr_cmsg; 1061 struct l_msghdr linux_msg; 1062 struct iovec *iov; 1063 socklen_t datalen; 1064 struct sockaddr *sa; 1065 sa_family_t sa_family; 1066 void *data; 1067 int error; 1068 1069 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1070 if (error != 0) 1071 return (error); 1072 1073 /* 1074 * Some Linux applications (ping) define a non-NULL control data 1075 * pointer, but a msg_controllen of 0, which is not allowed in the 1076 * FreeBSD system call interface. NULL the msg_control pointer in 1077 * order to handle this case. This should be checked, but allows the 1078 * Linux ping to work. 1079 */ 1080 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0) 1081 linux_msg.msg_control = PTROUT(NULL); 1082 1083 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1084 if (error != 0) 1085 return (error); 1086 1087#ifdef COMPAT_LINUX32 1088 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1089 &iov, EMSGSIZE); 1090#else 1091 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1092#endif 1093 if (error != 0) 1094 return (error); 1095 1096 control = NULL; 1097 cmsg = NULL; 1098 1099 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) { 1100 error = kern_getsockname(td, s, &sa, &datalen); 1101 if (error != 0) 1102 goto bad; 1103 sa_family = sa->sa_family; 1104 free(sa, M_SONAME); 1105 1106 error = ENOBUFS; 1107 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO); 1108 control = m_get(M_WAITOK, MT_CONTROL); 1109 1110 do { 1111 error = copyin(ptr_cmsg, &linux_cmsg, 1112 sizeof(struct l_cmsghdr)); 1113 if (error != 0) 1114 goto bad; 1115 1116 error = EINVAL; 1117 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1118 goto bad; 1119 1120 /* 1121 * Now we support only SCM_RIGHTS and SCM_CRED, 1122 * so return EINVAL in any other cmsg_type 1123 */ 1124 cmsg->cmsg_type = 1125 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1126 cmsg->cmsg_level = 1127 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1128 if (cmsg->cmsg_type == -1 1129 || cmsg->cmsg_level != SOL_SOCKET) 1130 goto bad; 1131 1132 /* 1133 * Some applications (e.g. pulseaudio) attempt to 1134 * send ancillary data even if the underlying protocol 1135 * doesn't support it which is not allowed in the 1136 * FreeBSD system call interface. 1137 */ 1138 if (sa_family != AF_UNIX) 1139 continue; 1140 1141 data = LINUX_CMSG_DATA(ptr_cmsg); 1142 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1143 1144 switch (cmsg->cmsg_type) 1145 { 1146 case SCM_RIGHTS: 1147 break; 1148 1149 case SCM_CREDS: 1150 data = &cmcred; 1151 datalen = sizeof(cmcred); 1152 1153 /* 1154 * The lower levels will fill in the structure 1155 */ 1156 bzero(data, datalen); 1157 break; 1158 } 1159 1160 cmsg->cmsg_len = CMSG_LEN(datalen); 1161 1162 error = ENOBUFS; 1163 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg)) 1164 goto bad; 1165 if (!m_append(control, datalen, (c_caddr_t)data)) 1166 goto bad; 1167 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg))); 1168 1169 if (m_length(control, NULL) == 0) { 1170 m_freem(control); 1171 control = NULL; 1172 } 1173 } 1174 1175 msg.msg_iov = iov; 1176 msg.msg_flags = 0; 1177 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1178 control = NULL; 1179 1180bad: 1181 m_freem(control); 1182 free(iov, M_IOV); 1183 if (cmsg) 1184 free(cmsg, M_LINUX); 1185 return (error); 1186} 1187 1188int 1189linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1190{ 1191 1192 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1193 args->flags)); 1194} 1195 1196int 1197linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1198{ 1199 struct l_mmsghdr *msg; 1200 l_uint retval; 1201 int error, datagrams; 1202 1203 if (args->vlen > UIO_MAXIOV) 1204 args->vlen = UIO_MAXIOV; 1205 1206 msg = PTRIN(args->msg); 1207 datagrams = 0; 1208 while (datagrams < args->vlen) { 1209 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1210 args->flags); 1211 if (error != 0) 1212 break; 1213 1214 retval = td->td_retval[0]; 1215 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1216 if (error != 0) 1217 break; 1218 ++msg; 1219 ++datagrams; 1220 } 1221 if (error == 0) 1222 td->td_retval[0] = datagrams; 1223 return (error); 1224} 1225 1226static int 1227linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1228 l_uint flags, struct msghdr *msg) 1229{ 1230 struct cmsghdr *cm; 1231 struct cmsgcred *cmcred; 1232 struct l_cmsghdr *linux_cmsg = NULL; 1233 struct l_ucred linux_ucred; 1234 socklen_t datalen, outlen; 1235 struct l_msghdr linux_msg; 1236 struct iovec *iov, *uiov; 1237 struct mbuf *control = NULL; 1238 struct mbuf **controlp; 1239 struct timeval *ftmvl; 1240 l_timeval ltmvl; 1241 caddr_t outbuf; 1242 void *data; 1243 int error, i, fd, fds, *fdp; 1244 1245 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1246 if (error != 0) 1247 return (error); 1248 1249 error = linux_to_bsd_msghdr(msg, &linux_msg); 1250 if (error != 0) 1251 return (error); 1252 1253#ifdef COMPAT_LINUX32 1254 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1255 &iov, EMSGSIZE); 1256#else 1257 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1258#endif 1259 if (error != 0) 1260 return (error); 1261 1262 if (msg->msg_name) { 1263 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name, 1264 msg->msg_namelen); 1265 if (error != 0) 1266 goto bad; 1267 } 1268 1269 uiov = msg->msg_iov; 1270 msg->msg_iov = iov; 1271 controlp = (msg->msg_control != NULL) ? &control : NULL; 1272 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp); 1273 msg->msg_iov = uiov; 1274 if (error != 0) 1275 goto bad; 1276 1277 error = bsd_to_linux_msghdr(msg, &linux_msg); 1278 if (error != 0) 1279 goto bad; 1280 1281 if (linux_msg.msg_name) { 1282 error = bsd_to_linux_sockaddr((struct sockaddr *) 1283 PTRIN(linux_msg.msg_name)); 1284 if (error != 0) 1285 goto bad; 1286 } 1287 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1288 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1289 if (error != 0) 1290 goto bad; 1291 } 1292 1293 outbuf = PTRIN(linux_msg.msg_control); 1294 outlen = 0; 1295 1296 if (control) { 1297 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1298 1299 msg->msg_control = mtod(control, struct cmsghdr *); 1300 msg->msg_controllen = control->m_len; 1301 1302 cm = CMSG_FIRSTHDR(msg); 1303 1304 while (cm != NULL) { 1305 linux_cmsg->cmsg_type = 1306 bsd_to_linux_cmsg_type(cm->cmsg_type); 1307 linux_cmsg->cmsg_level = 1308 bsd_to_linux_sockopt_level(cm->cmsg_level); 1309 if (linux_cmsg->cmsg_type == -1 1310 || cm->cmsg_level != SOL_SOCKET) 1311 { 1312 error = EINVAL; 1313 goto bad; 1314 } 1315 1316 data = CMSG_DATA(cm); 1317 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1318 1319 switch (cm->cmsg_type) 1320 { 1321 case SCM_RIGHTS: 1322 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1323 fds = datalen / sizeof(int); 1324 fdp = data; 1325 for (i = 0; i < fds; i++) { 1326 fd = *fdp++; 1327 (void)kern_fcntl(td, fd, 1328 F_SETFD, FD_CLOEXEC); 1329 } 1330 } 1331 break; 1332 1333 case SCM_CREDS: 1334 /* 1335 * Currently LOCAL_CREDS is never in 1336 * effect for Linux so no need to worry 1337 * about sockcred 1338 */ 1339 if (datalen != sizeof(*cmcred)) { 1340 error = EMSGSIZE; 1341 goto bad; 1342 } 1343 cmcred = (struct cmsgcred *)data; 1344 bzero(&linux_ucred, sizeof(linux_ucred)); 1345 linux_ucred.pid = cmcred->cmcred_pid; 1346 linux_ucred.uid = cmcred->cmcred_uid; 1347 linux_ucred.gid = cmcred->cmcred_gid; 1348 data = &linux_ucred; 1349 datalen = sizeof(linux_ucred); 1350 break; 1351 1352 case SCM_TIMESTAMP: 1353 if (datalen != sizeof(struct timeval)) { 1354 error = EMSGSIZE; 1355 goto bad; 1356 } 1357 ftmvl = (struct timeval *)data; 1358 ltmvl.tv_sec = ftmvl->tv_sec; 1359 ltmvl.tv_usec = ftmvl->tv_usec; 1360 data = <mvl; 1361 datalen = sizeof(ltmvl); 1362 break; 1363 } 1364 1365 if (outlen + LINUX_CMSG_LEN(datalen) > 1366 linux_msg.msg_controllen) { 1367 if (outlen == 0) { 1368 error = EMSGSIZE; 1369 goto bad; 1370 } else { 1371 linux_msg.msg_flags |= 1372 LINUX_MSG_CTRUNC; 1373 goto out; 1374 } 1375 } 1376 1377 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1378 1379 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1380 if (error) 1381 goto bad; 1382 outbuf += L_CMSG_HDRSZ; 1383 1384 error = copyout(data, outbuf, datalen); 1385 if (error) 1386 goto bad; 1387 1388 outbuf += LINUX_CMSG_ALIGN(datalen); 1389 outlen += LINUX_CMSG_LEN(datalen); 1390 1391 cm = CMSG_NXTHDR(msg, cm); 1392 } 1393 } 1394 1395out: 1396 linux_msg.msg_controllen = outlen; 1397 error = copyout(&linux_msg, msghdr, sizeof(linux_msg)); 1398 1399bad: 1400 free(iov, M_IOV); 1401 m_freem(control); 1402 free(linux_cmsg, M_LINUX); 1403 1404 return (error); 1405} 1406 1407int 1408linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1409{ 1410 struct msghdr bsd_msg; 1411 1412 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1413 args->flags, &bsd_msg)); 1414} 1415 1416int 1417linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1418{ 1419 struct l_mmsghdr *msg; 1420 struct msghdr bsd_msg; 1421 struct l_timespec lts; 1422 struct timespec ts, tts; 1423 l_uint retval; 1424 int error, datagrams; 1425 1426 if (args->timeout) { 1427 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1428 if (error != 0) 1429 return (error); 1430 error = linux_to_native_timespec(&ts, <s); 1431 if (error != 0) 1432 return (error); 1433 getnanotime(&tts); 1434 timespecadd(&tts, &ts); 1435 } 1436 1437 msg = PTRIN(args->msg); 1438 datagrams = 0; 1439 while (datagrams < args->vlen) { 1440 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1441 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1442 if (error != 0) 1443 break; 1444 1445 retval = td->td_retval[0]; 1446 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1447 if (error != 0) 1448 break; 1449 ++msg; 1450 ++datagrams; 1451 1452 /* 1453 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1454 */ 1455 if (args->flags & LINUX_MSG_WAITFORONE) 1456 args->flags |= LINUX_MSG_DONTWAIT; 1457 1458 /* 1459 * See BUGS section of recvmmsg(2). 1460 */ 1461 if (args->timeout) { 1462 getnanotime(&ts); 1463 timespecsub(&ts, &tts); 1464 if (!timespecisset(&ts) || ts.tv_sec > 0) 1465 break; 1466 } 1467 /* Out of band data, return right away. */ 1468 if (bsd_msg.msg_flags & MSG_OOB) 1469 break; 1470 } 1471 if (error == 0) 1472 td->td_retval[0] = datagrams; 1473 return (error); 1474} 1475 1476int 1477linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1478{ 1479 struct shutdown_args /* { 1480 int s; 1481 int how; 1482 } */ bsd_args; 1483 1484 bsd_args.s = args->s; 1485 bsd_args.how = args->how; 1486 return (sys_shutdown(td, &bsd_args)); 1487} 1488 1489int 1490linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1491{ 1492 struct setsockopt_args /* { 1493 int s; 1494 int level; 1495 int name; 1496 caddr_t val; 1497 int valsize; 1498 } */ bsd_args; 1499 l_timeval linux_tv; 1500 struct timeval tv; 1501 int error, name; 1502 1503 bsd_args.s = args->s; 1504 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1505 switch (bsd_args.level) { 1506 case SOL_SOCKET: 1507 name = linux_to_bsd_so_sockopt(args->optname); 1508 switch (name) { 1509 case SO_RCVTIMEO: 1510 /* FALLTHROUGH */ 1511 case SO_SNDTIMEO: 1512 error = copyin(PTRIN(args->optval), &linux_tv, 1513 sizeof(linux_tv)); 1514 if (error) 1515 return (error); 1516 tv.tv_sec = linux_tv.tv_sec; 1517 tv.tv_usec = linux_tv.tv_usec; 1518 return (kern_setsockopt(td, args->s, bsd_args.level, 1519 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1520 /* NOTREACHED */ 1521 break; 1522 default: 1523 break; 1524 } 1525 break; 1526 case IPPROTO_IP: 1527 name = linux_to_bsd_ip_sockopt(args->optname); 1528 break; 1529 case IPPROTO_TCP: 1530 name = linux_to_bsd_tcp_sockopt(args->optname); 1531 break; 1532 default: 1533 name = -1; 1534 break; 1535 } 1536 if (name == -1) 1537 return (ENOPROTOOPT); 1538 1539 bsd_args.name = name; 1540 bsd_args.val = PTRIN(args->optval); 1541 bsd_args.valsize = args->optlen; 1542 1543 if (name == IPV6_NEXTHOP) { 1544 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1545 bsd_args.valsize); 1546 error = sys_setsockopt(td, &bsd_args); 1547 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1548 } else 1549 error = sys_setsockopt(td, &bsd_args); 1550 1551 return (error); 1552} 1553 1554int 1555linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1556{ 1557 struct getsockopt_args /* { 1558 int s; 1559 int level; 1560 int name; 1561 caddr_t val; 1562 int *avalsize; 1563 } */ bsd_args; 1564 l_timeval linux_tv; 1565 struct timeval tv; 1566 socklen_t tv_len, xulen; 1567 struct xucred xu; 1568 struct l_ucred lxu; 1569 int error, name; 1570 1571 bsd_args.s = args->s; 1572 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1573 switch (bsd_args.level) { 1574 case SOL_SOCKET: 1575 name = linux_to_bsd_so_sockopt(args->optname); 1576 switch (name) { 1577 case SO_RCVTIMEO: 1578 /* FALLTHROUGH */ 1579 case SO_SNDTIMEO: 1580 tv_len = sizeof(tv); 1581 error = kern_getsockopt(td, args->s, bsd_args.level, 1582 name, &tv, UIO_SYSSPACE, &tv_len); 1583 if (error) 1584 return (error); 1585 linux_tv.tv_sec = tv.tv_sec; 1586 linux_tv.tv_usec = tv.tv_usec; 1587 return (copyout(&linux_tv, PTRIN(args->optval), 1588 sizeof(linux_tv))); 1589 /* NOTREACHED */ 1590 break; 1591 case LOCAL_PEERCRED: 1592 if (args->optlen != sizeof(lxu)) 1593 return (EINVAL); 1594 xulen = sizeof(xu); 1595 error = kern_getsockopt(td, args->s, bsd_args.level, 1596 name, &xu, UIO_SYSSPACE, &xulen); 1597 if (error) 1598 return (error); 1599 /* 1600 * XXX Use 0 for pid as the FreeBSD does not cache peer pid. 1601 */ 1602 lxu.pid = 0; 1603 lxu.uid = xu.cr_uid; 1604 lxu.gid = xu.cr_gid; 1605 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1606 /* NOTREACHED */ 1607 break; 1608 default: 1609 break; 1610 } 1611 break; 1612 case IPPROTO_IP: 1613 name = linux_to_bsd_ip_sockopt(args->optname); 1614 break; 1615 case IPPROTO_TCP: 1616 name = linux_to_bsd_tcp_sockopt(args->optname); 1617 break; 1618 default: 1619 name = -1; 1620 break; 1621 } 1622 if (name == -1) 1623 return (EINVAL); 1624 1625 bsd_args.name = name; 1626 bsd_args.val = PTRIN(args->optval); 1627 bsd_args.avalsize = PTRIN(args->optlen); 1628 1629 if (name == IPV6_NEXTHOP) { 1630 error = sys_getsockopt(td, &bsd_args); 1631 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1632 } else 1633 error = sys_getsockopt(td, &bsd_args); 1634 1635 return (error); 1636} 1637 1638#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1639 1640/* Argument list sizes for linux_socketcall */ 1641 1642#define LINUX_AL(x) ((x) * sizeof(l_ulong)) 1643 1644static const unsigned char lxs_args[] = { 1645 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */, 1646 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */, 1647 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */, 1648 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */, 1649 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */, 1650 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */, 1651 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */, 1652 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */, 1653 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */, 1654 LINUX_AL(4) /* accept4 */, LINUX_AL(5) /* recvmmsg */, 1655 LINUX_AL(4) /* sendmmsg */ 1656}; 1657 1658#define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1 1659 1660int 1661linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1662{ 1663 l_ulong a[6]; 1664 void *arg; 1665 int error; 1666 1667 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE) 1668 return (EINVAL); 1669 error = copyin(PTRIN(args->args), a, lxs_args[args->what]); 1670 if (error) 1671 return (error); 1672 1673 arg = a; 1674 switch (args->what) { 1675 case LINUX_SOCKET: 1676 return (linux_socket(td, arg)); 1677 case LINUX_BIND: 1678 return (linux_bind(td, arg)); 1679 case LINUX_CONNECT: 1680 return (linux_connect(td, arg)); 1681 case LINUX_LISTEN: 1682 return (linux_listen(td, arg)); 1683 case LINUX_ACCEPT: 1684 return (linux_accept(td, arg)); 1685 case LINUX_GETSOCKNAME: 1686 return (linux_getsockname(td, arg)); 1687 case LINUX_GETPEERNAME: 1688 return (linux_getpeername(td, arg)); 1689 case LINUX_SOCKETPAIR: 1690 return (linux_socketpair(td, arg)); 1691 case LINUX_SEND: 1692 return (linux_send(td, arg)); 1693 case LINUX_RECV: 1694 return (linux_recv(td, arg)); 1695 case LINUX_SENDTO: 1696 return (linux_sendto(td, arg)); 1697 case LINUX_RECVFROM: 1698 return (linux_recvfrom(td, arg)); 1699 case LINUX_SHUTDOWN: 1700 return (linux_shutdown(td, arg)); 1701 case LINUX_SETSOCKOPT: 1702 return (linux_setsockopt(td, arg)); 1703 case LINUX_GETSOCKOPT: 1704 return (linux_getsockopt(td, arg)); 1705 case LINUX_SENDMSG: 1706 return (linux_sendmsg(td, arg)); 1707 case LINUX_RECVMSG: 1708 return (linux_recvmsg(td, arg)); 1709 case LINUX_ACCEPT4: 1710 return (linux_accept4(td, arg)); 1711 case LINUX_RECVMMSG: 1712 return (linux_recvmmsg(td, arg)); 1713 case LINUX_SENDMMSG: 1714 return (linux_sendmmsg(td, arg)); 1715 } 1716 1717 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1718 return (ENOSYS); 1719} 1720#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1721