1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 1995 Søren Schmidt 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 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/stat.h> 53 #include <sys/syslog.h> 54 #include <sys/un.h> 55 #include <sys/unistd.h> 56 57 #include <security/audit/audit.h> 58 59 #include <net/if.h> 60 #include <net/vnet.h> 61 #include <netinet/in.h> 62 #include <netinet/in_systm.h> 63 #include <netinet/ip.h> 64 #include <netinet/tcp.h> 65 #ifdef INET6 66 #include <netinet/ip6.h> 67 #include <netinet6/ip6_var.h> 68 #endif 69 70 #ifdef COMPAT_LINUX32 71 #include <machine/../linux32/linux.h> 72 #include <machine/../linux32/linux32_proto.h> 73 #else 74 #include <machine/../linux/linux.h> 75 #include <machine/../linux/linux_proto.h> 76 #endif 77 #include <compat/linux/linux_common.h> 78 #include <compat/linux/linux_file.h> 79 #include <compat/linux/linux_mib.h> 80 #include <compat/linux/linux_socket.h> 81 #include <compat/linux/linux_timer.h> 82 #include <compat/linux/linux_util.h> 83 84 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *, 85 l_uint); 86 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *, 87 l_uint, struct msghdr *); 88 static int linux_set_socket_flags(int, int *); 89 90 91 static int 92 linux_to_bsd_sockopt_level(int level) 93 { 94 95 switch (level) { 96 case LINUX_SOL_SOCKET: 97 return (SOL_SOCKET); 98 } 99 return (level); 100 } 101 102 static int 103 bsd_to_linux_sockopt_level(int level) 104 { 105 106 switch (level) { 107 case SOL_SOCKET: 108 return (LINUX_SOL_SOCKET); 109 } 110 return (level); 111 } 112 113 static int 114 linux_to_bsd_ip_sockopt(int opt) 115 { 116 117 switch (opt) { 118 case LINUX_IP_TOS: 119 return (IP_TOS); 120 case LINUX_IP_TTL: 121 return (IP_TTL); 122 case LINUX_IP_OPTIONS: 123 return (IP_OPTIONS); 124 case LINUX_IP_MULTICAST_IF: 125 return (IP_MULTICAST_IF); 126 case LINUX_IP_MULTICAST_TTL: 127 return (IP_MULTICAST_TTL); 128 case LINUX_IP_MULTICAST_LOOP: 129 return (IP_MULTICAST_LOOP); 130 case LINUX_IP_ADD_MEMBERSHIP: 131 return (IP_ADD_MEMBERSHIP); 132 case LINUX_IP_DROP_MEMBERSHIP: 133 return (IP_DROP_MEMBERSHIP); 134 case LINUX_IP_HDRINCL: 135 return (IP_HDRINCL); 136 } 137 return (-1); 138 } 139 140 static int 141 linux_to_bsd_ip6_sockopt(int opt) 142 { 143 144 switch (opt) { 145 case LINUX_IPV6_NEXTHOP: 146 return (IPV6_NEXTHOP); 147 case LINUX_IPV6_UNICAST_HOPS: 148 return (IPV6_UNICAST_HOPS); 149 case LINUX_IPV6_MULTICAST_IF: 150 return (IPV6_MULTICAST_IF); 151 case LINUX_IPV6_MULTICAST_HOPS: 152 return (IPV6_MULTICAST_HOPS); 153 case LINUX_IPV6_MULTICAST_LOOP: 154 return (IPV6_MULTICAST_LOOP); 155 case LINUX_IPV6_ADD_MEMBERSHIP: 156 return (IPV6_JOIN_GROUP); 157 case LINUX_IPV6_DROP_MEMBERSHIP: 158 return (IPV6_LEAVE_GROUP); 159 case LINUX_IPV6_V6ONLY: 160 return (IPV6_V6ONLY); 161 case LINUX_IPV6_DONTFRAG: 162 return (IPV6_DONTFRAG); 163 #if 0 164 case LINUX_IPV6_CHECKSUM: 165 return (IPV6_CHECKSUM); 166 case LINUX_IPV6_RECVPKTINFO: 167 return (IPV6_RECVPKTINFO); 168 case LINUX_IPV6_PKTINFO: 169 return (IPV6_PKTINFO); 170 case LINUX_IPV6_RECVHOPLIMIT: 171 return (IPV6_RECVHOPLIMIT); 172 case LINUX_IPV6_HOPLIMIT: 173 return (IPV6_HOPLIMIT); 174 case LINUX_IPV6_RECVHOPOPTS: 175 return (IPV6_RECVHOPOPTS); 176 case LINUX_IPV6_HOPOPTS: 177 return (IPV6_HOPOPTS); 178 case LINUX_IPV6_RTHDRDSTOPTS: 179 return (IPV6_RTHDRDSTOPTS); 180 case LINUX_IPV6_RECVRTHDR: 181 return (IPV6_RECVRTHDR); 182 case LINUX_IPV6_RTHDR: 183 return (IPV6_RTHDR); 184 case LINUX_IPV6_RECVDSTOPTS: 185 return (IPV6_RECVDSTOPTS); 186 case LINUX_IPV6_DSTOPTS: 187 return (IPV6_DSTOPTS); 188 case LINUX_IPV6_RECVPATHMTU: 189 return (IPV6_RECVPATHMTU); 190 case LINUX_IPV6_PATHMTU: 191 return (IPV6_PATHMTU); 192 #endif 193 } 194 return (-1); 195 } 196 197 static int 198 linux_to_bsd_so_sockopt(int opt) 199 { 200 201 switch (opt) { 202 case LINUX_SO_DEBUG: 203 return (SO_DEBUG); 204 case LINUX_SO_REUSEADDR: 205 return (SO_REUSEADDR); 206 case LINUX_SO_TYPE: 207 return (SO_TYPE); 208 case LINUX_SO_ERROR: 209 return (SO_ERROR); 210 case LINUX_SO_DONTROUTE: 211 return (SO_DONTROUTE); 212 case LINUX_SO_BROADCAST: 213 return (SO_BROADCAST); 214 case LINUX_SO_SNDBUF: 215 case LINUX_SO_SNDBUFFORCE: 216 return (SO_SNDBUF); 217 case LINUX_SO_RCVBUF: 218 case LINUX_SO_RCVBUFFORCE: 219 return (SO_RCVBUF); 220 case LINUX_SO_KEEPALIVE: 221 return (SO_KEEPALIVE); 222 case LINUX_SO_OOBINLINE: 223 return (SO_OOBINLINE); 224 case LINUX_SO_LINGER: 225 return (SO_LINGER); 226 case LINUX_SO_REUSEPORT: 227 return (SO_REUSEPORT_LB); 228 case LINUX_SO_PEERCRED: 229 return (LOCAL_PEERCRED); 230 case LINUX_SO_RCVLOWAT: 231 return (SO_RCVLOWAT); 232 case LINUX_SO_SNDLOWAT: 233 return (SO_SNDLOWAT); 234 case LINUX_SO_RCVTIMEO: 235 return (SO_RCVTIMEO); 236 case LINUX_SO_SNDTIMEO: 237 return (SO_SNDTIMEO); 238 case LINUX_SO_TIMESTAMP: 239 return (SO_TIMESTAMP); 240 case LINUX_SO_ACCEPTCONN: 241 return (SO_ACCEPTCONN); 242 } 243 return (-1); 244 } 245 246 static int 247 linux_to_bsd_tcp_sockopt(int opt) 248 { 249 250 switch (opt) { 251 case LINUX_TCP_NODELAY: 252 return (TCP_NODELAY); 253 case LINUX_TCP_MAXSEG: 254 return (TCP_MAXSEG); 255 case LINUX_TCP_CORK: 256 return (TCP_NOPUSH); 257 case LINUX_TCP_KEEPIDLE: 258 return (TCP_KEEPIDLE); 259 case LINUX_TCP_KEEPINTVL: 260 return (TCP_KEEPINTVL); 261 case LINUX_TCP_KEEPCNT: 262 return (TCP_KEEPCNT); 263 case LINUX_TCP_MD5SIG: 264 return (TCP_MD5SIG); 265 } 266 return (-1); 267 } 268 269 static int 270 linux_to_bsd_msg_flags(int flags) 271 { 272 int ret_flags = 0; 273 274 if (flags & LINUX_MSG_OOB) 275 ret_flags |= MSG_OOB; 276 if (flags & LINUX_MSG_PEEK) 277 ret_flags |= MSG_PEEK; 278 if (flags & LINUX_MSG_DONTROUTE) 279 ret_flags |= MSG_DONTROUTE; 280 if (flags & LINUX_MSG_CTRUNC) 281 ret_flags |= MSG_CTRUNC; 282 if (flags & LINUX_MSG_TRUNC) 283 ret_flags |= MSG_TRUNC; 284 if (flags & LINUX_MSG_DONTWAIT) 285 ret_flags |= MSG_DONTWAIT; 286 if (flags & LINUX_MSG_EOR) 287 ret_flags |= MSG_EOR; 288 if (flags & LINUX_MSG_WAITALL) 289 ret_flags |= MSG_WAITALL; 290 if (flags & LINUX_MSG_NOSIGNAL) 291 ret_flags |= MSG_NOSIGNAL; 292 #if 0 /* not handled */ 293 if (flags & LINUX_MSG_PROXY) 294 ; 295 if (flags & LINUX_MSG_FIN) 296 ; 297 if (flags & LINUX_MSG_SYN) 298 ; 299 if (flags & LINUX_MSG_CONFIRM) 300 ; 301 if (flags & LINUX_MSG_RST) 302 ; 303 if (flags & LINUX_MSG_ERRQUEUE) 304 ; 305 #endif 306 return (ret_flags); 307 } 308 309 static int 310 linux_to_bsd_cmsg_type(int cmsg_type) 311 { 312 313 switch (cmsg_type) { 314 case LINUX_SCM_RIGHTS: 315 return (SCM_RIGHTS); 316 case LINUX_SCM_CREDENTIALS: 317 return (SCM_CREDS); 318 } 319 return (-1); 320 } 321 322 static int 323 bsd_to_linux_cmsg_type(int cmsg_type) 324 { 325 326 switch (cmsg_type) { 327 case SCM_RIGHTS: 328 return (LINUX_SCM_RIGHTS); 329 case SCM_CREDS: 330 return (LINUX_SCM_CREDENTIALS); 331 case SCM_TIMESTAMP: 332 return (LINUX_SCM_TIMESTAMP); 333 } 334 return (-1); 335 } 336 337 static int 338 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 339 { 340 if (lhdr->msg_controllen > INT_MAX) 341 return (ENOBUFS); 342 343 bhdr->msg_name = PTRIN(lhdr->msg_name); 344 bhdr->msg_namelen = lhdr->msg_namelen; 345 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 346 bhdr->msg_iovlen = lhdr->msg_iovlen; 347 bhdr->msg_control = PTRIN(lhdr->msg_control); 348 349 /* 350 * msg_controllen is skipped since BSD and LINUX control messages 351 * are potentially different sizes (e.g. the cred structure used 352 * by SCM_CREDS is different between the two operating system). 353 * 354 * The caller can set it (if necessary) after converting all the 355 * control messages. 356 */ 357 358 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 359 return (0); 360 } 361 362 static int 363 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 364 { 365 lhdr->msg_name = PTROUT(bhdr->msg_name); 366 lhdr->msg_namelen = bhdr->msg_namelen; 367 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 368 lhdr->msg_iovlen = bhdr->msg_iovlen; 369 lhdr->msg_control = PTROUT(bhdr->msg_control); 370 371 /* 372 * msg_controllen is skipped since BSD and LINUX control messages 373 * are potentially different sizes (e.g. the cred structure used 374 * by SCM_CREDS is different between the two operating system). 375 * 376 * The caller can set it (if necessary) after converting all the 377 * control messages. 378 */ 379 380 /* msg_flags skipped */ 381 return (0); 382 } 383 384 static int 385 linux_set_socket_flags(int lflags, int *flags) 386 { 387 388 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 389 return (EINVAL); 390 if (lflags & LINUX_SOCK_NONBLOCK) 391 *flags |= SOCK_NONBLOCK; 392 if (lflags & LINUX_SOCK_CLOEXEC) 393 *flags |= SOCK_CLOEXEC; 394 return (0); 395 } 396 397 static int 398 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 399 struct mbuf *control, enum uio_seg segflg) 400 { 401 struct sockaddr *to; 402 int error, len; 403 404 if (mp->msg_name != NULL) { 405 len = mp->msg_namelen; 406 error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len); 407 if (error != 0) 408 return (error); 409 mp->msg_name = to; 410 } else 411 to = NULL; 412 413 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 414 segflg); 415 416 if (to) 417 free(to, M_SONAME); 418 return (error); 419 } 420 421 /* Return 0 if IP_HDRINCL is set for the given socket. */ 422 static int 423 linux_check_hdrincl(struct thread *td, int s) 424 { 425 int error, optval; 426 socklen_t size_val; 427 428 size_val = sizeof(optval); 429 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 430 &optval, UIO_SYSSPACE, &size_val); 431 if (error != 0) 432 return (error); 433 434 return (optval == 0); 435 } 436 437 /* 438 * Updated sendto() when IP_HDRINCL is set: 439 * tweak endian-dependent fields in the IP packet. 440 */ 441 static int 442 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 443 { 444 /* 445 * linux_ip_copysize defines how many bytes we should copy 446 * from the beginning of the IP packet before we customize it for BSD. 447 * It should include all the fields we modify (ip_len and ip_off). 448 */ 449 #define linux_ip_copysize 8 450 451 struct ip *packet; 452 struct msghdr msg; 453 struct iovec aiov[1]; 454 int error; 455 456 /* Check that the packet isn't too big or too small. */ 457 if (linux_args->len < linux_ip_copysize || 458 linux_args->len > IP_MAXPACKET) 459 return (EINVAL); 460 461 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK); 462 463 /* Make kernel copy of the packet to be sent */ 464 if ((error = copyin(PTRIN(linux_args->msg), packet, 465 linux_args->len))) 466 goto goout; 467 468 /* Convert fields from Linux to BSD raw IP socket format */ 469 packet->ip_len = linux_args->len; 470 packet->ip_off = ntohs(packet->ip_off); 471 472 /* Prepare the msghdr and iovec structures describing the new packet */ 473 msg.msg_name = PTRIN(linux_args->to); 474 msg.msg_namelen = linux_args->tolen; 475 msg.msg_iov = aiov; 476 msg.msg_iovlen = 1; 477 msg.msg_control = NULL; 478 msg.msg_flags = 0; 479 aiov[0].iov_base = (char *)packet; 480 aiov[0].iov_len = linux_args->len; 481 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 482 NULL, UIO_SYSSPACE); 483 goout: 484 free(packet, M_LINUX); 485 return (error); 486 } 487 488 int 489 linux_socket(struct thread *td, struct linux_socket_args *args) 490 { 491 int domain, retval_socket, type; 492 493 type = args->type & LINUX_SOCK_TYPE_MASK; 494 if (type < 0 || type > LINUX_SOCK_MAX) 495 return (EINVAL); 496 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 497 &type); 498 if (retval_socket != 0) 499 return (retval_socket); 500 domain = linux_to_bsd_domain(args->domain); 501 if (domain == -1) 502 return (EAFNOSUPPORT); 503 504 retval_socket = kern_socket(td, domain, type, args->protocol); 505 if (retval_socket) 506 return (retval_socket); 507 508 if (type == SOCK_RAW 509 && (args->protocol == IPPROTO_RAW || args->protocol == 0) 510 && domain == PF_INET) { 511 /* It's a raw IP socket: set the IP_HDRINCL option. */ 512 int hdrincl; 513 514 hdrincl = 1; 515 /* We ignore any error returned by kern_setsockopt() */ 516 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 517 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 518 } 519 #ifdef INET6 520 /* 521 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default 522 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps. 523 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only 524 * sysctl value. 525 */ 526 if (domain == PF_INET6) { 527 int v6only; 528 529 v6only = 0; 530 /* We ignore any error returned by setsockopt() */ 531 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 532 &v6only, UIO_SYSSPACE, sizeof(v6only)); 533 } 534 #endif 535 536 return (retval_socket); 537 } 538 539 int 540 linux_bind(struct thread *td, struct linux_bind_args *args) 541 { 542 struct sockaddr *sa; 543 int error; 544 545 error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa, 546 &args->namelen); 547 if (error != 0) 548 return (error); 549 550 error = kern_bindat(td, AT_FDCWD, args->s, sa); 551 free(sa, M_SONAME); 552 553 /* XXX */ 554 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 555 return (EINVAL); 556 return (error); 557 } 558 559 int 560 linux_connect(struct thread *td, struct linux_connect_args *args) 561 { 562 struct socket *so; 563 struct sockaddr *sa; 564 struct file *fp; 565 u_int fflag; 566 int error; 567 568 error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa, 569 &args->namelen); 570 if (error != 0) 571 return (error); 572 573 error = kern_connectat(td, AT_FDCWD, args->s, sa); 574 free(sa, M_SONAME); 575 if (error != EISCONN) 576 return (error); 577 578 /* 579 * Linux doesn't return EISCONN the first time it occurs, 580 * when on a non-blocking socket. Instead it returns the 581 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 582 */ 583 error = getsock_cap(td, args->s, &cap_connect_rights, 584 &fp, &fflag, NULL); 585 if (error != 0) 586 return (error); 587 588 error = EISCONN; 589 so = fp->f_data; 590 if (fflag & FNONBLOCK) { 591 SOCK_LOCK(so); 592 if (so->so_emuldata == 0) 593 error = so->so_error; 594 so->so_emuldata = (void *)1; 595 SOCK_UNLOCK(so); 596 } 597 fdrop(fp, td); 598 599 return (error); 600 } 601 602 int 603 linux_listen(struct thread *td, struct linux_listen_args *args) 604 { 605 606 return (kern_listen(td, args->s, args->backlog)); 607 } 608 609 static int 610 linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 611 l_uintptr_t namelen, int flags) 612 { 613 struct l_sockaddr *lsa; 614 struct sockaddr *sa; 615 struct file *fp; 616 int bflags, len; 617 struct socket *so; 618 int error, error1; 619 620 bflags = 0; 621 error = linux_set_socket_flags(flags, &bflags); 622 if (error != 0) 623 return (error); 624 625 sa = NULL; 626 if (PTRIN(addr) == NULL) { 627 len = 0; 628 error = kern_accept4(td, s, NULL, NULL, bflags, NULL); 629 } else { 630 error = copyin(PTRIN(namelen), &len, sizeof(len)); 631 if (error != 0) 632 return (error); 633 if (len < 0) 634 return (EINVAL); 635 error = kern_accept4(td, s, &sa, &len, bflags, &fp); 636 if (error == 0) 637 fdrop(fp, td); 638 } 639 640 if (error != 0) { 641 /* 642 * XXX. This is wrong, different sockaddr structures 643 * have different sizes. 644 */ 645 if (error == EFAULT && namelen != sizeof(struct sockaddr_in)) 646 { 647 error = EINVAL; 648 goto out; 649 } 650 if (error == EINVAL) { 651 error1 = getsock_cap(td, s, &cap_accept_rights, &fp, NULL, NULL); 652 if (error1 != 0) { 653 error = error1; 654 goto out; 655 } 656 so = fp->f_data; 657 if (so->so_type == SOCK_DGRAM) 658 error = EOPNOTSUPP; 659 fdrop(fp, td); 660 } 661 goto out; 662 } 663 664 if (len != 0 && error == 0) { 665 error = bsd_to_linux_sockaddr(sa, &lsa, len); 666 if (error == 0) 667 error = copyout(lsa, PTRIN(addr), len); 668 free(lsa, M_SONAME); 669 } 670 671 free(sa, M_SONAME); 672 673 out: 674 if (error != 0) { 675 (void)kern_close(td, td->td_retval[0]); 676 td->td_retval[0] = 0; 677 } 678 return (error); 679 } 680 681 int 682 linux_accept(struct thread *td, struct linux_accept_args *args) 683 { 684 685 return (linux_accept_common(td, args->s, args->addr, 686 args->namelen, 0)); 687 } 688 689 int 690 linux_accept4(struct thread *td, struct linux_accept4_args *args) 691 { 692 693 return (linux_accept_common(td, args->s, args->addr, 694 args->namelen, args->flags)); 695 } 696 697 int 698 linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 699 { 700 struct l_sockaddr *lsa; 701 struct sockaddr *sa; 702 int len, error; 703 704 error = copyin(PTRIN(args->namelen), &len, sizeof(len)); 705 if (error != 0) 706 return (error); 707 708 error = kern_getsockname(td, args->s, &sa, &len); 709 if (error != 0) 710 return (error); 711 712 if (len != 0) { 713 error = bsd_to_linux_sockaddr(sa, &lsa, len); 714 if (error == 0) 715 error = copyout(lsa, PTRIN(args->addr), 716 len); 717 free(lsa, M_SONAME); 718 } 719 720 free(sa, M_SONAME); 721 if (error == 0) 722 error = copyout(&len, PTRIN(args->namelen), sizeof(len)); 723 return (error); 724 } 725 726 int 727 linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 728 { 729 struct l_sockaddr *lsa; 730 struct sockaddr *sa; 731 int len, error; 732 733 error = copyin(PTRIN(args->namelen), &len, sizeof(len)); 734 if (error != 0) 735 return (error); 736 if (len < 0) 737 return (EINVAL); 738 739 error = kern_getpeername(td, args->s, &sa, &len); 740 if (error != 0) 741 return (error); 742 743 if (len != 0) { 744 error = bsd_to_linux_sockaddr(sa, &lsa, len); 745 if (error == 0) 746 error = copyout(lsa, PTRIN(args->addr), 747 len); 748 free(lsa, M_SONAME); 749 } 750 751 free(sa, M_SONAME); 752 if (error == 0) 753 error = copyout(&len, PTRIN(args->namelen), sizeof(len)); 754 return (error); 755 } 756 757 int 758 linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 759 { 760 int domain, error, sv[2], type; 761 762 domain = linux_to_bsd_domain(args->domain); 763 if (domain != PF_LOCAL) 764 return (EAFNOSUPPORT); 765 type = args->type & LINUX_SOCK_TYPE_MASK; 766 if (type < 0 || type > LINUX_SOCK_MAX) 767 return (EINVAL); 768 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 769 &type); 770 if (error != 0) 771 return (error); 772 if (args->protocol != 0 && args->protocol != PF_UNIX) { 773 774 /* 775 * Use of PF_UNIX as protocol argument is not right, 776 * but Linux does it. 777 * Do not map PF_UNIX as its Linux value is identical 778 * to FreeBSD one. 779 */ 780 return (EPROTONOSUPPORT); 781 } 782 error = kern_socketpair(td, domain, type, 0, sv); 783 if (error != 0) 784 return (error); 785 error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int)); 786 if (error != 0) { 787 (void)kern_close(td, sv[0]); 788 (void)kern_close(td, sv[1]); 789 } 790 return (error); 791 } 792 793 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 794 struct linux_send_args { 795 register_t s; 796 register_t msg; 797 register_t len; 798 register_t flags; 799 }; 800 801 static int 802 linux_send(struct thread *td, struct linux_send_args *args) 803 { 804 struct sendto_args /* { 805 int s; 806 caddr_t buf; 807 int len; 808 int flags; 809 caddr_t to; 810 int tolen; 811 } */ bsd_args; 812 struct file *fp; 813 int error, fflag; 814 815 bsd_args.s = args->s; 816 bsd_args.buf = (caddr_t)PTRIN(args->msg); 817 bsd_args.len = args->len; 818 bsd_args.flags = args->flags; 819 bsd_args.to = NULL; 820 bsd_args.tolen = 0; 821 error = sys_sendto(td, &bsd_args); 822 if (error == ENOTCONN) { 823 /* 824 * Linux doesn't return ENOTCONN for non-blocking sockets. 825 * Instead it returns the EAGAIN. 826 */ 827 error = getsock_cap(td, args->s, &cap_send_rights, &fp, 828 &fflag, NULL); 829 if (error == 0) { 830 if (fflag & FNONBLOCK) 831 error = EAGAIN; 832 fdrop(fp, td); 833 } 834 } 835 return (error); 836 } 837 838 struct linux_recv_args { 839 register_t s; 840 register_t msg; 841 register_t len; 842 register_t flags; 843 }; 844 845 static int 846 linux_recv(struct thread *td, struct linux_recv_args *args) 847 { 848 struct recvfrom_args /* { 849 int s; 850 caddr_t buf; 851 int len; 852 int flags; 853 struct sockaddr *from; 854 socklen_t fromlenaddr; 855 } */ bsd_args; 856 857 bsd_args.s = args->s; 858 bsd_args.buf = (caddr_t)PTRIN(args->msg); 859 bsd_args.len = args->len; 860 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 861 bsd_args.from = NULL; 862 bsd_args.fromlenaddr = 0; 863 return (sys_recvfrom(td, &bsd_args)); 864 } 865 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 866 867 int 868 linux_sendto(struct thread *td, struct linux_sendto_args *args) 869 { 870 struct msghdr msg; 871 struct iovec aiov; 872 873 if (linux_check_hdrincl(td, args->s) == 0) 874 /* IP_HDRINCL set, tweak the packet before sending */ 875 return (linux_sendto_hdrincl(td, args)); 876 877 msg.msg_name = PTRIN(args->to); 878 msg.msg_namelen = args->tolen; 879 msg.msg_iov = &aiov; 880 msg.msg_iovlen = 1; 881 msg.msg_control = NULL; 882 msg.msg_flags = 0; 883 aiov.iov_base = PTRIN(args->msg); 884 aiov.iov_len = args->len; 885 return (linux_sendit(td, args->s, &msg, args->flags, NULL, 886 UIO_USERSPACE)); 887 } 888 889 int 890 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 891 { 892 struct l_sockaddr *lsa; 893 struct sockaddr *sa; 894 struct msghdr msg; 895 struct iovec aiov; 896 int error, fromlen; 897 898 if (PTRIN(args->fromlen) != NULL) { 899 error = copyin(PTRIN(args->fromlen), &fromlen, 900 sizeof(fromlen)); 901 if (error != 0) 902 return (error); 903 if (fromlen < 0) 904 return (EINVAL); 905 sa = malloc(fromlen, M_SONAME, M_WAITOK); 906 } else { 907 fromlen = 0; 908 sa = NULL; 909 } 910 911 msg.msg_name = sa; 912 msg.msg_namelen = fromlen; 913 msg.msg_iov = &aiov; 914 msg.msg_iovlen = 1; 915 aiov.iov_base = PTRIN(args->buf); 916 aiov.iov_len = args->len; 917 msg.msg_control = 0; 918 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 919 920 error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL); 921 if (error != 0) 922 goto out; 923 924 if (PTRIN(args->from) != NULL) { 925 error = bsd_to_linux_sockaddr(sa, &lsa, msg.msg_namelen); 926 if (error == 0) 927 error = copyout(lsa, PTRIN(args->from), 928 msg.msg_namelen); 929 free(lsa, M_SONAME); 930 } 931 932 if (error == 0 && PTRIN(args->fromlen) != NULL) 933 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 934 sizeof(msg.msg_namelen)); 935 out: 936 free(sa, M_SONAME); 937 return (error); 938 } 939 940 static int 941 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 942 l_uint flags) 943 { 944 struct cmsghdr *cmsg; 945 struct mbuf *control; 946 struct msghdr msg; 947 struct l_cmsghdr linux_cmsg; 948 struct l_cmsghdr *ptr_cmsg; 949 struct l_msghdr linux_msg; 950 struct iovec *iov; 951 socklen_t datalen; 952 struct sockaddr *sa; 953 struct socket *so; 954 sa_family_t sa_family; 955 struct file *fp; 956 void *data; 957 l_size_t len; 958 l_size_t clen; 959 int error, fflag; 960 961 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 962 if (error != 0) 963 return (error); 964 965 /* 966 * Some Linux applications (ping) define a non-NULL control data 967 * pointer, but a msg_controllen of 0, which is not allowed in the 968 * FreeBSD system call interface. NULL the msg_control pointer in 969 * order to handle this case. This should be checked, but allows the 970 * Linux ping to work. 971 */ 972 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0) 973 linux_msg.msg_control = PTROUT(NULL); 974 975 error = linux_to_bsd_msghdr(&msg, &linux_msg); 976 if (error != 0) 977 return (error); 978 979 #ifdef COMPAT_LINUX32 980 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 981 &iov, EMSGSIZE); 982 #else 983 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 984 #endif 985 if (error != 0) 986 return (error); 987 988 control = NULL; 989 990 error = kern_getsockname(td, s, &sa, &datalen); 991 if (error != 0) 992 goto bad; 993 sa_family = sa->sa_family; 994 free(sa, M_SONAME); 995 996 if (flags & LINUX_MSG_OOB) { 997 error = EOPNOTSUPP; 998 if (sa_family == AF_UNIX) 999 goto bad; 1000 1001 error = getsock_cap(td, s, &cap_send_rights, &fp, 1002 &fflag, NULL); 1003 if (error != 0) 1004 goto bad; 1005 so = fp->f_data; 1006 if (so->so_type != SOCK_STREAM) 1007 error = EOPNOTSUPP; 1008 fdrop(fp, td); 1009 if (error != 0) 1010 goto bad; 1011 } 1012 1013 if (linux_msg.msg_controllen >= sizeof(struct l_cmsghdr)) { 1014 1015 error = ENOBUFS; 1016 control = m_get(M_WAITOK, MT_CONTROL); 1017 MCLGET(control, M_WAITOK); 1018 data = mtod(control, void *); 1019 datalen = 0; 1020 1021 ptr_cmsg = PTRIN(linux_msg.msg_control); 1022 clen = linux_msg.msg_controllen; 1023 do { 1024 error = copyin(ptr_cmsg, &linux_cmsg, 1025 sizeof(struct l_cmsghdr)); 1026 if (error != 0) 1027 goto bad; 1028 1029 error = EINVAL; 1030 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) || 1031 linux_cmsg.cmsg_len > clen) 1032 goto bad; 1033 1034 if (datalen + CMSG_HDRSZ > MCLBYTES) 1035 goto bad; 1036 1037 /* 1038 * Now we support only SCM_RIGHTS and SCM_CRED, 1039 * so return EINVAL in any other cmsg_type 1040 */ 1041 cmsg = data; 1042 cmsg->cmsg_type = 1043 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1044 cmsg->cmsg_level = 1045 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1046 if (cmsg->cmsg_type == -1 1047 || cmsg->cmsg_level != SOL_SOCKET) 1048 goto bad; 1049 1050 /* 1051 * Some applications (e.g. pulseaudio) attempt to 1052 * send ancillary data even if the underlying protocol 1053 * doesn't support it which is not allowed in the 1054 * FreeBSD system call interface. 1055 */ 1056 if (sa_family != AF_UNIX) 1057 continue; 1058 1059 if (cmsg->cmsg_type == SCM_CREDS) { 1060 len = sizeof(struct cmsgcred); 1061 if (datalen + CMSG_SPACE(len) > MCLBYTES) 1062 goto bad; 1063 1064 /* 1065 * The lower levels will fill in the structure 1066 */ 1067 memset(CMSG_DATA(data), 0, len); 1068 } else { 1069 len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1070 if (datalen + CMSG_SPACE(len) < datalen || 1071 datalen + CMSG_SPACE(len) > MCLBYTES) 1072 goto bad; 1073 1074 error = copyin(LINUX_CMSG_DATA(ptr_cmsg), 1075 CMSG_DATA(data), len); 1076 if (error != 0) 1077 goto bad; 1078 } 1079 1080 cmsg->cmsg_len = CMSG_LEN(len); 1081 data = (char *)data + CMSG_SPACE(len); 1082 datalen += CMSG_SPACE(len); 1083 1084 if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len)) 1085 break; 1086 1087 clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len); 1088 ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg + 1089 LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len)); 1090 } while(clen >= sizeof(struct l_cmsghdr)); 1091 1092 control->m_len = datalen; 1093 if (datalen == 0) { 1094 m_freem(control); 1095 control = NULL; 1096 } 1097 } 1098 1099 msg.msg_iov = iov; 1100 msg.msg_flags = 0; 1101 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1102 control = NULL; 1103 1104 bad: 1105 m_freem(control); 1106 free(iov, M_IOV); 1107 return (error); 1108 } 1109 1110 int 1111 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1112 { 1113 1114 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1115 args->flags)); 1116 } 1117 1118 int 1119 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1120 { 1121 struct l_mmsghdr *msg; 1122 l_uint retval; 1123 int error, datagrams; 1124 1125 if (args->vlen > UIO_MAXIOV) 1126 args->vlen = UIO_MAXIOV; 1127 1128 msg = PTRIN(args->msg); 1129 datagrams = 0; 1130 while (datagrams < args->vlen) { 1131 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1132 args->flags); 1133 if (error != 0) 1134 break; 1135 1136 retval = td->td_retval[0]; 1137 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1138 if (error != 0) 1139 break; 1140 ++msg; 1141 ++datagrams; 1142 } 1143 if (error == 0) 1144 td->td_retval[0] = datagrams; 1145 return (error); 1146 } 1147 1148 static int 1149 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1150 l_uint flags, struct msghdr *msg) 1151 { 1152 struct cmsghdr *cm; 1153 struct cmsgcred *cmcred; 1154 struct l_cmsghdr *linux_cmsg = NULL; 1155 struct l_ucred linux_ucred; 1156 socklen_t datalen, maxlen, outlen; 1157 struct l_msghdr linux_msg; 1158 struct iovec *iov, *uiov; 1159 struct mbuf *control = NULL; 1160 struct mbuf **controlp; 1161 struct timeval *ftmvl; 1162 struct l_sockaddr *lsa; 1163 struct sockaddr *sa; 1164 l_timeval ltmvl; 1165 caddr_t outbuf; 1166 void *data; 1167 int error, i, fd, fds, *fdp; 1168 1169 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1170 if (error != 0) 1171 return (error); 1172 1173 error = linux_to_bsd_msghdr(msg, &linux_msg); 1174 if (error != 0) 1175 return (error); 1176 1177 #ifdef COMPAT_LINUX32 1178 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1179 &iov, EMSGSIZE); 1180 #else 1181 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1182 #endif 1183 if (error != 0) 1184 return (error); 1185 1186 if (msg->msg_name) { 1187 sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK); 1188 msg->msg_name = sa; 1189 } else 1190 sa = NULL; 1191 1192 uiov = msg->msg_iov; 1193 msg->msg_iov = iov; 1194 controlp = (msg->msg_control != NULL) ? &control : NULL; 1195 error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp); 1196 msg->msg_iov = uiov; 1197 if (error != 0) 1198 goto bad; 1199 1200 if (msg->msg_name) { 1201 msg->msg_name = PTRIN(linux_msg.msg_name); 1202 error = bsd_to_linux_sockaddr(sa, &lsa, msg->msg_namelen); 1203 if (error == 0) 1204 error = copyout(lsa, PTRIN(msg->msg_name), 1205 msg->msg_namelen); 1206 free(lsa, M_SONAME); 1207 if (error != 0) 1208 goto bad; 1209 } 1210 1211 error = bsd_to_linux_msghdr(msg, &linux_msg); 1212 if (error != 0) 1213 goto bad; 1214 1215 maxlen = linux_msg.msg_controllen; 1216 linux_msg.msg_controllen = 0; 1217 if (control) { 1218 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1219 1220 msg->msg_control = mtod(control, struct cmsghdr *); 1221 msg->msg_controllen = control->m_len; 1222 1223 cm = CMSG_FIRSTHDR(msg); 1224 outbuf = PTRIN(linux_msg.msg_control); 1225 outlen = 0; 1226 while (cm != NULL) { 1227 linux_cmsg->cmsg_type = 1228 bsd_to_linux_cmsg_type(cm->cmsg_type); 1229 linux_cmsg->cmsg_level = 1230 bsd_to_linux_sockopt_level(cm->cmsg_level); 1231 if (linux_cmsg->cmsg_type == -1 || 1232 cm->cmsg_level != SOL_SOCKET) { 1233 error = EINVAL; 1234 goto bad; 1235 } 1236 1237 data = CMSG_DATA(cm); 1238 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1239 1240 switch (cm->cmsg_type) { 1241 case SCM_RIGHTS: 1242 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1243 fds = datalen / sizeof(int); 1244 fdp = data; 1245 for (i = 0; i < fds; i++) { 1246 fd = *fdp++; 1247 (void)kern_fcntl(td, fd, 1248 F_SETFD, FD_CLOEXEC); 1249 } 1250 } 1251 break; 1252 1253 case SCM_CREDS: 1254 /* 1255 * Currently LOCAL_CREDS is never in 1256 * effect for Linux so no need to worry 1257 * about sockcred 1258 */ 1259 if (datalen != sizeof(*cmcred)) { 1260 error = EMSGSIZE; 1261 goto bad; 1262 } 1263 cmcred = (struct cmsgcred *)data; 1264 bzero(&linux_ucred, sizeof(linux_ucred)); 1265 linux_ucred.pid = cmcred->cmcred_pid; 1266 linux_ucred.uid = cmcred->cmcred_uid; 1267 linux_ucred.gid = cmcred->cmcred_gid; 1268 data = &linux_ucred; 1269 datalen = sizeof(linux_ucred); 1270 break; 1271 1272 case SCM_TIMESTAMP: 1273 if (datalen != sizeof(struct timeval)) { 1274 error = EMSGSIZE; 1275 goto bad; 1276 } 1277 ftmvl = (struct timeval *)data; 1278 ltmvl.tv_sec = ftmvl->tv_sec; 1279 ltmvl.tv_usec = ftmvl->tv_usec; 1280 data = <mvl; 1281 datalen = sizeof(ltmvl); 1282 break; 1283 } 1284 1285 if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) { 1286 if (outlen == 0) { 1287 error = EMSGSIZE; 1288 goto bad; 1289 } else { 1290 linux_msg.msg_flags |= LINUX_MSG_CTRUNC; 1291 m_dispose_extcontrolm(control); 1292 goto out; 1293 } 1294 } 1295 1296 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1297 1298 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1299 if (error != 0) 1300 goto bad; 1301 outbuf += L_CMSG_HDRSZ; 1302 1303 error = copyout(data, outbuf, datalen); 1304 if (error != 0) 1305 goto bad; 1306 1307 outbuf += LINUX_CMSG_ALIGN(datalen); 1308 outlen += LINUX_CMSG_LEN(datalen); 1309 1310 cm = CMSG_NXTHDR(msg, cm); 1311 } 1312 linux_msg.msg_controllen = outlen; 1313 } 1314 1315 out: 1316 error = copyout(&linux_msg, msghdr, sizeof(linux_msg)); 1317 1318 bad: 1319 if (control != NULL) { 1320 if (error != 0) 1321 m_dispose_extcontrolm(control); 1322 m_freem(control); 1323 } 1324 free(iov, M_IOV); 1325 free(linux_cmsg, M_LINUX); 1326 free(sa, M_SONAME); 1327 1328 return (error); 1329 } 1330 1331 int 1332 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1333 { 1334 struct msghdr bsd_msg; 1335 1336 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1337 args->flags, &bsd_msg)); 1338 } 1339 1340 int 1341 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1342 { 1343 struct l_mmsghdr *msg; 1344 struct msghdr bsd_msg; 1345 struct l_timespec lts; 1346 struct timespec ts, tts; 1347 l_uint retval; 1348 int error, datagrams; 1349 1350 if (args->timeout) { 1351 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1352 if (error != 0) 1353 return (error); 1354 error = linux_to_native_timespec(&ts, <s); 1355 if (error != 0) 1356 return (error); 1357 getnanotime(&tts); 1358 timespecadd(&tts, &ts, &tts); 1359 } 1360 1361 msg = PTRIN(args->msg); 1362 datagrams = 0; 1363 while (datagrams < args->vlen) { 1364 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1365 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1366 if (error != 0) 1367 break; 1368 1369 retval = td->td_retval[0]; 1370 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1371 if (error != 0) 1372 break; 1373 ++msg; 1374 ++datagrams; 1375 1376 /* 1377 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1378 */ 1379 if (args->flags & LINUX_MSG_WAITFORONE) 1380 args->flags |= LINUX_MSG_DONTWAIT; 1381 1382 /* 1383 * See BUGS section of recvmmsg(2). 1384 */ 1385 if (args->timeout) { 1386 getnanotime(&ts); 1387 timespecsub(&ts, &tts, &ts); 1388 if (!timespecisset(&ts) || ts.tv_sec > 0) 1389 break; 1390 } 1391 /* Out of band data, return right away. */ 1392 if (bsd_msg.msg_flags & MSG_OOB) 1393 break; 1394 } 1395 if (error == 0) 1396 td->td_retval[0] = datagrams; 1397 return (error); 1398 } 1399 1400 int 1401 linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1402 { 1403 1404 return (kern_shutdown(td, args->s, args->how)); 1405 } 1406 1407 int 1408 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1409 { 1410 l_timeval linux_tv; 1411 struct sockaddr *sa; 1412 struct timeval tv; 1413 socklen_t len; 1414 int error, level, name; 1415 1416 level = linux_to_bsd_sockopt_level(args->level); 1417 switch (level) { 1418 case SOL_SOCKET: 1419 name = linux_to_bsd_so_sockopt(args->optname); 1420 switch (name) { 1421 case SO_RCVTIMEO: 1422 /* FALLTHROUGH */ 1423 case SO_SNDTIMEO: 1424 error = copyin(PTRIN(args->optval), &linux_tv, 1425 sizeof(linux_tv)); 1426 if (error != 0) 1427 return (error); 1428 tv.tv_sec = linux_tv.tv_sec; 1429 tv.tv_usec = linux_tv.tv_usec; 1430 return (kern_setsockopt(td, args->s, level, 1431 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1432 /* NOTREACHED */ 1433 default: 1434 break; 1435 } 1436 break; 1437 case IPPROTO_IP: 1438 if (args->optname == LINUX_IP_RECVERR && 1439 linux_ignore_ip_recverr) { 1440 /* 1441 * XXX: This is a hack to unbreak DNS resolution 1442 * with glibc 2.30 and above. 1443 */ 1444 return (0); 1445 } 1446 name = linux_to_bsd_ip_sockopt(args->optname); 1447 break; 1448 case IPPROTO_IPV6: 1449 name = linux_to_bsd_ip6_sockopt(args->optname); 1450 break; 1451 case IPPROTO_TCP: 1452 name = linux_to_bsd_tcp_sockopt(args->optname); 1453 break; 1454 default: 1455 name = -1; 1456 break; 1457 } 1458 if (name == -1) { 1459 linux_msg(curthread, 1460 "unsupported setsockopt level %d optname %d", 1461 args->level, args->optname); 1462 return (ENOPROTOOPT); 1463 } 1464 1465 if (name == IPV6_NEXTHOP) { 1466 len = args->optlen; 1467 error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len); 1468 if (error != 0) 1469 return (error); 1470 1471 error = kern_setsockopt(td, args->s, level, 1472 name, sa, UIO_SYSSPACE, len); 1473 free(sa, M_SONAME); 1474 } else { 1475 error = kern_setsockopt(td, args->s, level, 1476 name, PTRIN(args->optval), UIO_USERSPACE, args->optlen); 1477 } 1478 1479 return (error); 1480 } 1481 1482 int 1483 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1484 { 1485 l_timeval linux_tv; 1486 struct timeval tv; 1487 socklen_t tv_len, xulen, len; 1488 struct l_sockaddr *lsa; 1489 struct sockaddr *sa; 1490 struct xucred xu; 1491 struct l_ucred lxu; 1492 int error, level, name, newval; 1493 1494 level = linux_to_bsd_sockopt_level(args->level); 1495 switch (level) { 1496 case SOL_SOCKET: 1497 name = linux_to_bsd_so_sockopt(args->optname); 1498 switch (name) { 1499 case SO_RCVTIMEO: 1500 /* FALLTHROUGH */ 1501 case SO_SNDTIMEO: 1502 tv_len = sizeof(tv); 1503 error = kern_getsockopt(td, args->s, level, 1504 name, &tv, UIO_SYSSPACE, &tv_len); 1505 if (error != 0) 1506 return (error); 1507 linux_tv.tv_sec = tv.tv_sec; 1508 linux_tv.tv_usec = tv.tv_usec; 1509 return (copyout(&linux_tv, PTRIN(args->optval), 1510 sizeof(linux_tv))); 1511 /* NOTREACHED */ 1512 case LOCAL_PEERCRED: 1513 if (args->optlen < sizeof(lxu)) 1514 return (EINVAL); 1515 /* 1516 * LOCAL_PEERCRED is not served at the SOL_SOCKET level, 1517 * but by the Unix socket's level 0. 1518 */ 1519 level = 0; 1520 xulen = sizeof(xu); 1521 error = kern_getsockopt(td, args->s, level, 1522 name, &xu, UIO_SYSSPACE, &xulen); 1523 if (error != 0) 1524 return (error); 1525 lxu.pid = xu.cr_pid; 1526 lxu.uid = xu.cr_uid; 1527 lxu.gid = xu.cr_gid; 1528 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1529 /* NOTREACHED */ 1530 case SO_ERROR: 1531 len = sizeof(newval); 1532 error = kern_getsockopt(td, args->s, level, 1533 name, &newval, UIO_SYSSPACE, &len); 1534 if (error != 0) 1535 return (error); 1536 newval = -SV_ABI_ERRNO(td->td_proc, newval); 1537 return (copyout(&newval, PTRIN(args->optval), len)); 1538 /* NOTREACHED */ 1539 default: 1540 break; 1541 } 1542 break; 1543 case IPPROTO_IP: 1544 name = linux_to_bsd_ip_sockopt(args->optname); 1545 break; 1546 case IPPROTO_IPV6: 1547 name = linux_to_bsd_ip6_sockopt(args->optname); 1548 break; 1549 case IPPROTO_TCP: 1550 name = linux_to_bsd_tcp_sockopt(args->optname); 1551 break; 1552 default: 1553 name = -1; 1554 break; 1555 } 1556 if (name == -1) { 1557 linux_msg(curthread, 1558 "unsupported getsockopt level %d optname %d", 1559 args->level, args->optname); 1560 return (EINVAL); 1561 } 1562 1563 if (name == IPV6_NEXTHOP) { 1564 error = copyin(PTRIN(args->optlen), &len, sizeof(len)); 1565 if (error != 0) 1566 return (error); 1567 sa = malloc(len, M_SONAME, M_WAITOK); 1568 1569 error = kern_getsockopt(td, args->s, level, 1570 name, sa, UIO_SYSSPACE, &len); 1571 if (error != 0) 1572 goto out; 1573 1574 error = bsd_to_linux_sockaddr(sa, &lsa, len); 1575 if (error == 0) 1576 error = copyout(lsa, PTRIN(args->optval), len); 1577 free(lsa, M_SONAME); 1578 if (error == 0) 1579 error = copyout(&len, PTRIN(args->optlen), 1580 sizeof(len)); 1581 out: 1582 free(sa, M_SONAME); 1583 } else { 1584 if (args->optval) { 1585 error = copyin(PTRIN(args->optlen), &len, sizeof(len)); 1586 if (error != 0) 1587 return (error); 1588 } 1589 error = kern_getsockopt(td, args->s, level, 1590 name, PTRIN(args->optval), UIO_USERSPACE, &len); 1591 if (error == 0) 1592 error = copyout(&len, PTRIN(args->optlen), 1593 sizeof(len)); 1594 } 1595 1596 return (error); 1597 } 1598 1599 static int 1600 linux_sendfile_common(struct thread *td, l_int out, l_int in, 1601 l_loff_t *offset, l_size_t count) 1602 { 1603 off_t bytes_read; 1604 int error; 1605 l_loff_t current_offset; 1606 struct file *fp; 1607 1608 AUDIT_ARG_FD(in); 1609 error = fget_read(td, in, &cap_pread_rights, &fp); 1610 if (error != 0) 1611 return (error); 1612 1613 if (offset != NULL) { 1614 current_offset = *offset; 1615 } else { 1616 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ? 1617 fo_seek(fp, 0, SEEK_CUR, td) : ESPIPE; 1618 if (error != 0) 1619 goto drop; 1620 current_offset = td->td_uretoff.tdu_off; 1621 } 1622 1623 bytes_read = 0; 1624 1625 /* Linux cannot have 0 count. */ 1626 if (count <= 0 || current_offset < 0) { 1627 error = EINVAL; 1628 goto drop; 1629 } 1630 1631 error = fo_sendfile(fp, out, NULL, NULL, current_offset, count, 1632 &bytes_read, 0, td); 1633 if (error != 0) 1634 goto drop; 1635 current_offset += bytes_read; 1636 1637 if (offset != NULL) { 1638 *offset = current_offset; 1639 } else { 1640 error = fo_seek(fp, current_offset, SEEK_SET, td); 1641 if (error != 0) 1642 goto drop; 1643 } 1644 1645 td->td_retval[0] = (ssize_t)bytes_read; 1646 drop: 1647 fdrop(fp, td); 1648 return (error); 1649 } 1650 1651 int 1652 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg) 1653 { 1654 /* 1655 * Differences between FreeBSD and Linux sendfile: 1656 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to 1657 * mean send the whole file.) In linux_sendfile given fds are still 1658 * checked for validity when the count is 0. 1659 * - Linux can send to any fd whereas FreeBSD only supports sockets. 1660 * The same restriction follows for linux_sendfile. 1661 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr. 1662 * - Linux takes an offset pointer and updates it to the read location. 1663 * FreeBSD takes in an offset and a 'bytes read' parameter which is 1664 * only filled if it isn't NULL. We use this parameter to update the 1665 * offset pointer if it exists. 1666 * - Linux sendfile returns bytes read on success while FreeBSD 1667 * returns 0. We use the 'bytes read' parameter to get this value. 1668 */ 1669 1670 l_loff_t offset64; 1671 l_long offset; 1672 int ret; 1673 int error; 1674 1675 if (arg->offset != NULL) { 1676 error = copyin(arg->offset, &offset, sizeof(offset)); 1677 if (error != 0) 1678 return (error); 1679 offset64 = (l_loff_t)offset; 1680 } 1681 1682 ret = linux_sendfile_common(td, arg->out, arg->in, 1683 arg->offset != NULL ? &offset64 : NULL, arg->count); 1684 1685 if (arg->offset != NULL) { 1686 #if defined(__i386__) || defined(__arm__) || \ 1687 (defined(__amd64__) && defined(COMPAT_LINUX32)) 1688 if (offset64 > INT32_MAX) 1689 return (EOVERFLOW); 1690 #endif 1691 offset = (l_long)offset64; 1692 error = copyout(&offset, arg->offset, sizeof(offset)); 1693 if (error != 0) 1694 return (error); 1695 } 1696 1697 return (ret); 1698 } 1699 1700 #if defined(__i386__) || defined(__arm__) || \ 1701 (defined(__amd64__) && defined(COMPAT_LINUX32)) 1702 1703 int 1704 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg) 1705 { 1706 l_loff_t offset; 1707 int ret; 1708 int error; 1709 1710 if (arg->offset != NULL) { 1711 error = copyin(arg->offset, &offset, sizeof(offset)); 1712 if (error != 0) 1713 return (error); 1714 } 1715 1716 ret = linux_sendfile_common(td, arg->out, arg->in, 1717 arg->offset != NULL ? &offset : NULL, arg->count); 1718 1719 if (arg->offset != NULL) { 1720 error = copyout(&offset, arg->offset, sizeof(offset)); 1721 if (error != 0) 1722 return (error); 1723 } 1724 1725 return (ret); 1726 } 1727 1728 /* Argument list sizes for linux_socketcall */ 1729 static const unsigned char lxs_args_cnt[] = { 1730 0 /* unused*/, 3 /* socket */, 1731 3 /* bind */, 3 /* connect */, 1732 2 /* listen */, 3 /* accept */, 1733 3 /* getsockname */, 3 /* getpeername */, 1734 4 /* socketpair */, 4 /* send */, 1735 4 /* recv */, 6 /* sendto */, 1736 6 /* recvfrom */, 2 /* shutdown */, 1737 5 /* setsockopt */, 5 /* getsockopt */, 1738 3 /* sendmsg */, 3 /* recvmsg */, 1739 4 /* accept4 */, 5 /* recvmmsg */, 1740 4 /* sendmmsg */, 4 /* sendfile */ 1741 }; 1742 #define LINUX_ARGS_CNT (nitems(lxs_args_cnt) - 1) 1743 #define LINUX_ARG_SIZE(x) (lxs_args_cnt[x] * sizeof(l_ulong)) 1744 1745 int 1746 linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1747 { 1748 l_ulong a[6]; 1749 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1750 register_t l_args[6]; 1751 #endif 1752 void *arg; 1753 int error; 1754 1755 if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT) 1756 return (EINVAL); 1757 error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what)); 1758 if (error != 0) 1759 return (error); 1760 1761 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1762 for (int i = 0; i < lxs_args_cnt[args->what]; ++i) 1763 l_args[i] = a[i]; 1764 arg = l_args; 1765 #else 1766 arg = a; 1767 #endif 1768 switch (args->what) { 1769 case LINUX_SOCKET: 1770 return (linux_socket(td, arg)); 1771 case LINUX_BIND: 1772 return (linux_bind(td, arg)); 1773 case LINUX_CONNECT: 1774 return (linux_connect(td, arg)); 1775 case LINUX_LISTEN: 1776 return (linux_listen(td, arg)); 1777 case LINUX_ACCEPT: 1778 return (linux_accept(td, arg)); 1779 case LINUX_GETSOCKNAME: 1780 return (linux_getsockname(td, arg)); 1781 case LINUX_GETPEERNAME: 1782 return (linux_getpeername(td, arg)); 1783 case LINUX_SOCKETPAIR: 1784 return (linux_socketpair(td, arg)); 1785 case LINUX_SEND: 1786 return (linux_send(td, arg)); 1787 case LINUX_RECV: 1788 return (linux_recv(td, arg)); 1789 case LINUX_SENDTO: 1790 return (linux_sendto(td, arg)); 1791 case LINUX_RECVFROM: 1792 return (linux_recvfrom(td, arg)); 1793 case LINUX_SHUTDOWN: 1794 return (linux_shutdown(td, arg)); 1795 case LINUX_SETSOCKOPT: 1796 return (linux_setsockopt(td, arg)); 1797 case LINUX_GETSOCKOPT: 1798 return (linux_getsockopt(td, arg)); 1799 case LINUX_SENDMSG: 1800 return (linux_sendmsg(td, arg)); 1801 case LINUX_RECVMSG: 1802 return (linux_recvmsg(td, arg)); 1803 case LINUX_ACCEPT4: 1804 return (linux_accept4(td, arg)); 1805 case LINUX_RECVMMSG: 1806 return (linux_recvmmsg(td, arg)); 1807 case LINUX_SENDMMSG: 1808 return (linux_sendmmsg(td, arg)); 1809 case LINUX_SENDFILE: 1810 return (linux_sendfile(td, arg)); 1811 } 1812 1813 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1814 return (ENOSYS); 1815 } 1816 #endif /* __i386__ || __arm__ || (__amd64__ && COMPAT_LINUX32) */ 1817