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/filedesc.h> 44 #include <sys/limits.h> 45 #include <sys/lock.h> 46 #include <sys/malloc.h> 47 #include <sys/mutex.h> 48 #include <sys/mbuf.h> 49 #include <sys/socket.h> 50 #include <sys/socketvar.h> 51 #include <sys/syscallsubr.h> 52 #include <sys/uio.h> 53 #include <sys/stat.h> 54 #include <sys/syslog.h> 55 #include <sys/un.h> 56 #include <sys/unistd.h> 57 58 #include <security/audit/audit.h> 59 60 #include <net/if.h> 61 #include <net/vnet.h> 62 #include <netinet/in.h> 63 #include <netinet/in_systm.h> 64 #include <netinet/ip.h> 65 #include <netinet/tcp.h> 66 #ifdef INET6 67 #include <netinet/ip6.h> 68 #include <netinet6/ip6_var.h> 69 #endif 70 71 #ifdef COMPAT_LINUX32 72 #include <machine/../linux32/linux.h> 73 #include <machine/../linux32/linux32_proto.h> 74 #else 75 #include <machine/../linux/linux.h> 76 #include <machine/../linux/linux_proto.h> 77 #endif 78 #include <compat/linux/linux_common.h> 79 #include <compat/linux/linux_file.h> 80 #include <compat/linux/linux_mib.h> 81 #include <compat/linux/linux_socket.h> 82 #include <compat/linux/linux_timer.h> 83 #include <compat/linux/linux_util.h> 84 85 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *, 86 l_uint); 87 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *, 88 l_uint, struct msghdr *); 89 static int linux_set_socket_flags(int, int *); 90 91 static int 92 linux_to_bsd_sockopt_level(int level) 93 { 94 95 if (level == LINUX_SOL_SOCKET) 96 return (SOL_SOCKET); 97 /* Remaining values are RFC-defined protocol numbers. */ 98 return (level); 99 } 100 101 static int 102 bsd_to_linux_sockopt_level(int level) 103 { 104 105 if (level == SOL_SOCKET) 106 return (LINUX_SOL_SOCKET); 107 return (level); 108 } 109 110 static int 111 linux_to_bsd_ip_sockopt(int opt) 112 { 113 114 switch (opt) { 115 case LINUX_IP_TOS: 116 return (IP_TOS); 117 case LINUX_IP_TTL: 118 return (IP_TTL); 119 case LINUX_IP_OPTIONS: 120 return (IP_OPTIONS); 121 case LINUX_IP_MULTICAST_IF: 122 return (IP_MULTICAST_IF); 123 case LINUX_IP_MULTICAST_TTL: 124 return (IP_MULTICAST_TTL); 125 case LINUX_IP_MULTICAST_LOOP: 126 return (IP_MULTICAST_LOOP); 127 case LINUX_IP_ADD_MEMBERSHIP: 128 return (IP_ADD_MEMBERSHIP); 129 case LINUX_IP_DROP_MEMBERSHIP: 130 return (IP_DROP_MEMBERSHIP); 131 case LINUX_IP_HDRINCL: 132 return (IP_HDRINCL); 133 } 134 return (-1); 135 } 136 137 static int 138 linux_to_bsd_ip6_sockopt(int opt) 139 { 140 141 switch (opt) { 142 case LINUX_IPV6_NEXTHOP: 143 return (IPV6_NEXTHOP); 144 case LINUX_IPV6_UNICAST_HOPS: 145 return (IPV6_UNICAST_HOPS); 146 case LINUX_IPV6_MULTICAST_IF: 147 return (IPV6_MULTICAST_IF); 148 case LINUX_IPV6_MULTICAST_HOPS: 149 return (IPV6_MULTICAST_HOPS); 150 case LINUX_IPV6_MULTICAST_LOOP: 151 return (IPV6_MULTICAST_LOOP); 152 case LINUX_IPV6_ADD_MEMBERSHIP: 153 return (IPV6_JOIN_GROUP); 154 case LINUX_IPV6_DROP_MEMBERSHIP: 155 return (IPV6_LEAVE_GROUP); 156 case LINUX_IPV6_V6ONLY: 157 return (IPV6_V6ONLY); 158 case LINUX_IPV6_DONTFRAG: 159 return (IPV6_DONTFRAG); 160 #if 0 161 case LINUX_IPV6_CHECKSUM: 162 return (IPV6_CHECKSUM); 163 case LINUX_IPV6_RECVPKTINFO: 164 return (IPV6_RECVPKTINFO); 165 case LINUX_IPV6_PKTINFO: 166 return (IPV6_PKTINFO); 167 case LINUX_IPV6_RECVHOPLIMIT: 168 return (IPV6_RECVHOPLIMIT); 169 case LINUX_IPV6_HOPLIMIT: 170 return (IPV6_HOPLIMIT); 171 case LINUX_IPV6_RECVHOPOPTS: 172 return (IPV6_RECVHOPOPTS); 173 case LINUX_IPV6_HOPOPTS: 174 return (IPV6_HOPOPTS); 175 case LINUX_IPV6_RTHDRDSTOPTS: 176 return (IPV6_RTHDRDSTOPTS); 177 case LINUX_IPV6_RECVRTHDR: 178 return (IPV6_RECVRTHDR); 179 case LINUX_IPV6_RTHDR: 180 return (IPV6_RTHDR); 181 case LINUX_IPV6_RECVDSTOPTS: 182 return (IPV6_RECVDSTOPTS); 183 case LINUX_IPV6_DSTOPTS: 184 return (IPV6_DSTOPTS); 185 case LINUX_IPV6_RECVPATHMTU: 186 return (IPV6_RECVPATHMTU); 187 case LINUX_IPV6_PATHMTU: 188 return (IPV6_PATHMTU); 189 #endif 190 } 191 return (-1); 192 } 193 194 static int 195 linux_to_bsd_so_sockopt(int opt) 196 { 197 198 switch (opt) { 199 case LINUX_SO_DEBUG: 200 return (SO_DEBUG); 201 case LINUX_SO_REUSEADDR: 202 return (SO_REUSEADDR); 203 case LINUX_SO_TYPE: 204 return (SO_TYPE); 205 case LINUX_SO_ERROR: 206 return (SO_ERROR); 207 case LINUX_SO_DONTROUTE: 208 return (SO_DONTROUTE); 209 case LINUX_SO_BROADCAST: 210 return (SO_BROADCAST); 211 case LINUX_SO_SNDBUF: 212 case LINUX_SO_SNDBUFFORCE: 213 return (SO_SNDBUF); 214 case LINUX_SO_RCVBUF: 215 case LINUX_SO_RCVBUFFORCE: 216 return (SO_RCVBUF); 217 case LINUX_SO_KEEPALIVE: 218 return (SO_KEEPALIVE); 219 case LINUX_SO_OOBINLINE: 220 return (SO_OOBINLINE); 221 case LINUX_SO_LINGER: 222 return (SO_LINGER); 223 case LINUX_SO_REUSEPORT: 224 return (SO_REUSEPORT_LB); 225 case LINUX_SO_PEERCRED: 226 return (LOCAL_PEERCRED); 227 case LINUX_SO_RCVLOWAT: 228 return (SO_RCVLOWAT); 229 case LINUX_SO_SNDLOWAT: 230 return (SO_SNDLOWAT); 231 case LINUX_SO_RCVTIMEO: 232 return (SO_RCVTIMEO); 233 case LINUX_SO_SNDTIMEO: 234 return (SO_SNDTIMEO); 235 case LINUX_SO_TIMESTAMP: 236 return (SO_TIMESTAMP); 237 case LINUX_SO_ACCEPTCONN: 238 return (SO_ACCEPTCONN); 239 case LINUX_SO_PROTOCOL: 240 return (SO_PROTOCOL); 241 } 242 return (-1); 243 } 244 245 static int 246 linux_to_bsd_tcp_sockopt(int opt) 247 { 248 249 switch (opt) { 250 case LINUX_TCP_NODELAY: 251 return (TCP_NODELAY); 252 case LINUX_TCP_MAXSEG: 253 return (TCP_MAXSEG); 254 case LINUX_TCP_CORK: 255 return (TCP_NOPUSH); 256 case LINUX_TCP_KEEPIDLE: 257 return (TCP_KEEPIDLE); 258 case LINUX_TCP_KEEPINTVL: 259 return (TCP_KEEPINTVL); 260 case LINUX_TCP_KEEPCNT: 261 return (TCP_KEEPCNT); 262 case LINUX_TCP_MD5SIG: 263 return (TCP_MD5SIG); 264 } 265 return (-1); 266 } 267 268 static int 269 linux_to_bsd_msg_flags(int flags) 270 { 271 int ret_flags = 0; 272 273 if (flags & LINUX_MSG_OOB) 274 ret_flags |= MSG_OOB; 275 if (flags & LINUX_MSG_PEEK) 276 ret_flags |= MSG_PEEK; 277 if (flags & LINUX_MSG_DONTROUTE) 278 ret_flags |= MSG_DONTROUTE; 279 if (flags & LINUX_MSG_CTRUNC) 280 ret_flags |= MSG_CTRUNC; 281 if (flags & LINUX_MSG_TRUNC) 282 ret_flags |= MSG_TRUNC; 283 if (flags & LINUX_MSG_DONTWAIT) 284 ret_flags |= MSG_DONTWAIT; 285 if (flags & LINUX_MSG_EOR) 286 ret_flags |= MSG_EOR; 287 if (flags & LINUX_MSG_WAITALL) 288 ret_flags |= MSG_WAITALL; 289 if (flags & LINUX_MSG_NOSIGNAL) 290 ret_flags |= MSG_NOSIGNAL; 291 #if 0 /* not handled */ 292 if (flags & LINUX_MSG_PROXY) 293 ; 294 if (flags & LINUX_MSG_FIN) 295 ; 296 if (flags & LINUX_MSG_SYN) 297 ; 298 if (flags & LINUX_MSG_CONFIRM) 299 ; 300 if (flags & LINUX_MSG_RST) 301 ; 302 if (flags & LINUX_MSG_ERRQUEUE) 303 ; 304 #endif 305 return (ret_flags); 306 } 307 308 static int 309 linux_to_bsd_cmsg_type(int cmsg_type) 310 { 311 312 switch (cmsg_type) { 313 case LINUX_SCM_RIGHTS: 314 return (SCM_RIGHTS); 315 case LINUX_SCM_CREDENTIALS: 316 return (SCM_CREDS); 317 } 318 return (-1); 319 } 320 321 static int 322 bsd_to_linux_cmsg_type(int cmsg_type) 323 { 324 325 switch (cmsg_type) { 326 case SCM_RIGHTS: 327 return (LINUX_SCM_RIGHTS); 328 case SCM_CREDS: 329 return (LINUX_SCM_CREDENTIALS); 330 case SCM_TIMESTAMP: 331 return (LINUX_SCM_TIMESTAMP); 332 } 333 return (-1); 334 } 335 336 static int 337 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 338 { 339 if (lhdr->msg_controllen > INT_MAX) 340 return (ENOBUFS); 341 342 bhdr->msg_name = PTRIN(lhdr->msg_name); 343 bhdr->msg_namelen = lhdr->msg_namelen; 344 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 345 bhdr->msg_iovlen = lhdr->msg_iovlen; 346 bhdr->msg_control = PTRIN(lhdr->msg_control); 347 348 /* 349 * msg_controllen is skipped since BSD and LINUX control messages 350 * are potentially different sizes (e.g. the cred structure used 351 * by SCM_CREDS is different between the two operating system). 352 * 353 * The caller can set it (if necessary) after converting all the 354 * control messages. 355 */ 356 357 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 358 return (0); 359 } 360 361 static int 362 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 363 { 364 lhdr->msg_name = PTROUT(bhdr->msg_name); 365 lhdr->msg_namelen = bhdr->msg_namelen; 366 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 367 lhdr->msg_iovlen = bhdr->msg_iovlen; 368 lhdr->msg_control = PTROUT(bhdr->msg_control); 369 370 /* 371 * msg_controllen is skipped since BSD and LINUX control messages 372 * are potentially different sizes (e.g. the cred structure used 373 * by SCM_CREDS is different between the two operating system). 374 * 375 * The caller can set it (if necessary) after converting all the 376 * control messages. 377 */ 378 379 /* msg_flags skipped */ 380 return (0); 381 } 382 383 static int 384 linux_set_socket_flags(int lflags, int *flags) 385 { 386 387 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 388 return (EINVAL); 389 if (lflags & LINUX_SOCK_NONBLOCK) 390 *flags |= SOCK_NONBLOCK; 391 if (lflags & LINUX_SOCK_CLOEXEC) 392 *flags |= SOCK_CLOEXEC; 393 return (0); 394 } 395 396 static int 397 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 398 struct mbuf *control, enum uio_seg segflg) 399 { 400 struct sockaddr *to; 401 int error, len; 402 403 if (mp->msg_name != NULL) { 404 len = mp->msg_namelen; 405 error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len); 406 if (error != 0) 407 return (error); 408 mp->msg_name = to; 409 } else 410 to = NULL; 411 412 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 413 segflg); 414 415 if (to) 416 free(to, M_SONAME); 417 return (error); 418 } 419 420 /* Return 0 if IP_HDRINCL is set for the given socket. */ 421 static int 422 linux_check_hdrincl(struct thread *td, int s) 423 { 424 int error, optval; 425 socklen_t size_val; 426 427 size_val = sizeof(optval); 428 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 429 &optval, UIO_SYSSPACE, &size_val); 430 if (error != 0) 431 return (error); 432 433 return (optval == 0); 434 } 435 436 /* 437 * Updated sendto() when IP_HDRINCL is set: 438 * tweak endian-dependent fields in the IP packet. 439 */ 440 static int 441 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 442 { 443 /* 444 * linux_ip_copysize defines how many bytes we should copy 445 * from the beginning of the IP packet before we customize it for BSD. 446 * It should include all the fields we modify (ip_len and ip_off). 447 */ 448 #define linux_ip_copysize 8 449 450 struct ip *packet; 451 struct msghdr msg; 452 struct iovec aiov[1]; 453 int error; 454 455 /* Check that the packet isn't too big or too small. */ 456 if (linux_args->len < linux_ip_copysize || 457 linux_args->len > IP_MAXPACKET) 458 return (EINVAL); 459 460 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK); 461 462 /* Make kernel copy of the packet to be sent */ 463 if ((error = copyin(PTRIN(linux_args->msg), packet, 464 linux_args->len))) 465 goto goout; 466 467 /* Convert fields from Linux to BSD raw IP socket format */ 468 packet->ip_len = linux_args->len; 469 packet->ip_off = ntohs(packet->ip_off); 470 471 /* Prepare the msghdr and iovec structures describing the new packet */ 472 msg.msg_name = PTRIN(linux_args->to); 473 msg.msg_namelen = linux_args->tolen; 474 msg.msg_iov = aiov; 475 msg.msg_iovlen = 1; 476 msg.msg_control = NULL; 477 msg.msg_flags = 0; 478 aiov[0].iov_base = (char *)packet; 479 aiov[0].iov_len = linux_args->len; 480 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 481 NULL, UIO_SYSSPACE); 482 goout: 483 free(packet, M_LINUX); 484 return (error); 485 } 486 487 int 488 linux_socket(struct thread *td, struct linux_socket_args *args) 489 { 490 int domain, retval_socket, type; 491 492 type = args->type & LINUX_SOCK_TYPE_MASK; 493 if (type < 0 || type > LINUX_SOCK_MAX) 494 return (EINVAL); 495 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 496 &type); 497 if (retval_socket != 0) 498 return (retval_socket); 499 domain = linux_to_bsd_domain(args->domain); 500 if (domain == -1) 501 return (EAFNOSUPPORT); 502 503 retval_socket = kern_socket(td, domain, type, args->protocol); 504 if (retval_socket) 505 return (retval_socket); 506 507 if (type == SOCK_RAW 508 && (args->protocol == IPPROTO_RAW || args->protocol == 0) 509 && domain == PF_INET) { 510 /* It's a raw IP socket: set the IP_HDRINCL option. */ 511 int hdrincl; 512 513 hdrincl = 1; 514 /* We ignore any error returned by kern_setsockopt() */ 515 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 516 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 517 } 518 #ifdef INET6 519 /* 520 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default 521 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps. 522 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only 523 * sysctl value. 524 */ 525 if (domain == PF_INET6) { 526 int v6only; 527 528 v6only = 0; 529 /* We ignore any error returned by setsockopt() */ 530 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 531 &v6only, UIO_SYSSPACE, sizeof(v6only)); 532 } 533 #endif 534 535 return (retval_socket); 536 } 537 538 int 539 linux_bind(struct thread *td, struct linux_bind_args *args) 540 { 541 struct sockaddr *sa; 542 int error; 543 544 error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa, 545 &args->namelen); 546 if (error != 0) 547 return (error); 548 549 error = kern_bindat(td, AT_FDCWD, args->s, sa); 550 free(sa, M_SONAME); 551 552 /* XXX */ 553 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 554 return (EINVAL); 555 return (error); 556 } 557 558 int 559 linux_connect(struct thread *td, struct linux_connect_args *args) 560 { 561 struct socket *so; 562 struct sockaddr *sa; 563 struct file *fp; 564 u_int fflag; 565 int error; 566 567 error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa, 568 &args->namelen); 569 if (error != 0) 570 return (error); 571 572 error = kern_connectat(td, AT_FDCWD, args->s, sa); 573 free(sa, M_SONAME); 574 if (error != EISCONN) 575 return (error); 576 577 /* 578 * Linux doesn't return EISCONN the first time it occurs, 579 * when on a non-blocking socket. Instead it returns the 580 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 581 */ 582 error = getsock_cap(td, args->s, &cap_connect_rights, 583 &fp, &fflag, NULL); 584 if (error != 0) 585 return (error); 586 587 error = EISCONN; 588 so = fp->f_data; 589 if (fflag & FNONBLOCK) { 590 SOCK_LOCK(so); 591 if (so->so_emuldata == 0) 592 error = so->so_error; 593 so->so_emuldata = (void *)1; 594 SOCK_UNLOCK(so); 595 } 596 fdrop(fp, td); 597 598 return (error); 599 } 600 601 int 602 linux_listen(struct thread *td, struct linux_listen_args *args) 603 { 604 605 return (kern_listen(td, args->s, args->backlog)); 606 } 607 608 static int 609 linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 610 l_uintptr_t namelen, int flags) 611 { 612 struct l_sockaddr *lsa; 613 struct sockaddr *sa; 614 struct file *fp, *fp1; 615 int bflags, len; 616 struct socket *so; 617 int error, error1; 618 619 bflags = 0; 620 fp = NULL; 621 sa = NULL; 622 623 error = linux_set_socket_flags(flags, &bflags); 624 if (error != 0) 625 return (error); 626 627 if (PTRIN(addr) == NULL) { 628 len = 0; 629 error = kern_accept4(td, s, NULL, NULL, bflags, NULL); 630 } else { 631 error = copyin(PTRIN(namelen), &len, sizeof(len)); 632 if (error != 0) 633 return (error); 634 if (len < 0) 635 return (EINVAL); 636 error = kern_accept4(td, s, &sa, &len, bflags, &fp); 637 } 638 639 /* 640 * Translate errno values into ones used by Linux. 641 */ 642 if (error != 0) { 643 /* 644 * XXX. This is wrong, different sockaddr structures 645 * have different sizes. 646 */ 647 switch (error) { 648 case EFAULT: 649 if (namelen != sizeof(struct sockaddr_in)) 650 error = EINVAL; 651 break; 652 case EINVAL: 653 error1 = getsock_cap(td, s, &cap_accept_rights, &fp1, NULL, NULL); 654 if (error1 != 0) { 655 error = error1; 656 break; 657 } 658 so = fp1->f_data; 659 if (so->so_type == SOCK_DGRAM) 660 error = EOPNOTSUPP; 661 fdrop(fp1, td); 662 break; 663 } 664 return (error); 665 } 666 667 if (len != 0) { 668 error = bsd_to_linux_sockaddr(sa, &lsa, len); 669 if (error == 0) 670 error = copyout(lsa, PTRIN(addr), len); 671 free(lsa, M_SONAME); 672 673 /* 674 * XXX: We should also copyout the len, shouldn't we? 675 */ 676 677 if (error != 0) { 678 fdclose(td, fp, td->td_retval[0]); 679 td->td_retval[0] = 0; 680 } 681 } 682 if (fp != NULL) 683 fdrop(fp, td); 684 free(sa, M_SONAME); 685 return (error); 686 } 687 688 int 689 linux_accept(struct thread *td, struct linux_accept_args *args) 690 { 691 692 return (linux_accept_common(td, args->s, args->addr, 693 args->namelen, 0)); 694 } 695 696 int 697 linux_accept4(struct thread *td, struct linux_accept4_args *args) 698 { 699 700 return (linux_accept_common(td, args->s, args->addr, 701 args->namelen, args->flags)); 702 } 703 704 int 705 linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 706 { 707 struct l_sockaddr *lsa; 708 struct sockaddr *sa; 709 int len, error; 710 711 error = copyin(PTRIN(args->namelen), &len, sizeof(len)); 712 if (error != 0) 713 return (error); 714 715 error = kern_getsockname(td, args->s, &sa, &len); 716 if (error != 0) 717 return (error); 718 719 if (len != 0) { 720 error = bsd_to_linux_sockaddr(sa, &lsa, len); 721 if (error == 0) 722 error = copyout(lsa, PTRIN(args->addr), 723 len); 724 free(lsa, M_SONAME); 725 } 726 727 free(sa, M_SONAME); 728 if (error == 0) 729 error = copyout(&len, PTRIN(args->namelen), sizeof(len)); 730 return (error); 731 } 732 733 int 734 linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 735 { 736 struct l_sockaddr *lsa; 737 struct sockaddr *sa; 738 int len, error; 739 740 error = copyin(PTRIN(args->namelen), &len, sizeof(len)); 741 if (error != 0) 742 return (error); 743 if (len < 0) 744 return (EINVAL); 745 746 error = kern_getpeername(td, args->s, &sa, &len); 747 if (error != 0) 748 return (error); 749 750 if (len != 0) { 751 error = bsd_to_linux_sockaddr(sa, &lsa, len); 752 if (error == 0) 753 error = copyout(lsa, PTRIN(args->addr), 754 len); 755 free(lsa, M_SONAME); 756 } 757 758 free(sa, M_SONAME); 759 if (error == 0) 760 error = copyout(&len, PTRIN(args->namelen), sizeof(len)); 761 return (error); 762 } 763 764 int 765 linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 766 { 767 int domain, error, sv[2], type; 768 769 domain = linux_to_bsd_domain(args->domain); 770 if (domain != PF_LOCAL) 771 return (EAFNOSUPPORT); 772 type = args->type & LINUX_SOCK_TYPE_MASK; 773 if (type < 0 || type > LINUX_SOCK_MAX) 774 return (EINVAL); 775 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 776 &type); 777 if (error != 0) 778 return (error); 779 if (args->protocol != 0 && args->protocol != PF_UNIX) { 780 /* 781 * Use of PF_UNIX as protocol argument is not right, 782 * but Linux does it. 783 * Do not map PF_UNIX as its Linux value is identical 784 * to FreeBSD one. 785 */ 786 return (EPROTONOSUPPORT); 787 } 788 error = kern_socketpair(td, domain, type, 0, sv); 789 if (error != 0) 790 return (error); 791 error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int)); 792 if (error != 0) { 793 (void)kern_close(td, sv[0]); 794 (void)kern_close(td, sv[1]); 795 } 796 return (error); 797 } 798 799 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 800 struct linux_send_args { 801 register_t s; 802 register_t msg; 803 register_t len; 804 register_t flags; 805 }; 806 807 static int 808 linux_send(struct thread *td, struct linux_send_args *args) 809 { 810 struct sendto_args /* { 811 int s; 812 caddr_t buf; 813 int len; 814 int flags; 815 caddr_t to; 816 int tolen; 817 } */ bsd_args; 818 struct file *fp; 819 int error, fflag; 820 821 bsd_args.s = args->s; 822 bsd_args.buf = (caddr_t)PTRIN(args->msg); 823 bsd_args.len = args->len; 824 bsd_args.flags = args->flags; 825 bsd_args.to = NULL; 826 bsd_args.tolen = 0; 827 error = sys_sendto(td, &bsd_args); 828 if (error == ENOTCONN) { 829 /* 830 * Linux doesn't return ENOTCONN for non-blocking sockets. 831 * Instead it returns the EAGAIN. 832 */ 833 error = getsock_cap(td, args->s, &cap_send_rights, &fp, 834 &fflag, NULL); 835 if (error == 0) { 836 if (fflag & FNONBLOCK) 837 error = EAGAIN; 838 fdrop(fp, td); 839 } 840 } 841 return (error); 842 } 843 844 struct linux_recv_args { 845 register_t s; 846 register_t msg; 847 register_t len; 848 register_t flags; 849 }; 850 851 static int 852 linux_recv(struct thread *td, struct linux_recv_args *args) 853 { 854 struct recvfrom_args /* { 855 int s; 856 caddr_t buf; 857 int len; 858 int flags; 859 struct sockaddr *from; 860 socklen_t fromlenaddr; 861 } */ bsd_args; 862 863 bsd_args.s = args->s; 864 bsd_args.buf = (caddr_t)PTRIN(args->msg); 865 bsd_args.len = args->len; 866 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 867 bsd_args.from = NULL; 868 bsd_args.fromlenaddr = 0; 869 return (sys_recvfrom(td, &bsd_args)); 870 } 871 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 872 873 int 874 linux_sendto(struct thread *td, struct linux_sendto_args *args) 875 { 876 struct msghdr msg; 877 struct iovec aiov; 878 879 if (linux_check_hdrincl(td, args->s) == 0) 880 /* IP_HDRINCL set, tweak the packet before sending */ 881 return (linux_sendto_hdrincl(td, args)); 882 883 msg.msg_name = PTRIN(args->to); 884 msg.msg_namelen = args->tolen; 885 msg.msg_iov = &aiov; 886 msg.msg_iovlen = 1; 887 msg.msg_control = NULL; 888 msg.msg_flags = 0; 889 aiov.iov_base = PTRIN(args->msg); 890 aiov.iov_len = args->len; 891 return (linux_sendit(td, args->s, &msg, args->flags, NULL, 892 UIO_USERSPACE)); 893 } 894 895 int 896 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 897 { 898 struct l_sockaddr *lsa; 899 struct sockaddr *sa; 900 struct msghdr msg; 901 struct iovec aiov; 902 int error, fromlen; 903 904 if (PTRIN(args->fromlen) != NULL) { 905 error = copyin(PTRIN(args->fromlen), &fromlen, 906 sizeof(fromlen)); 907 if (error != 0) 908 return (error); 909 if (fromlen < 0) 910 return (EINVAL); 911 sa = malloc(fromlen, M_SONAME, M_WAITOK); 912 } else { 913 fromlen = 0; 914 sa = NULL; 915 } 916 917 msg.msg_name = sa; 918 msg.msg_namelen = fromlen; 919 msg.msg_iov = &aiov; 920 msg.msg_iovlen = 1; 921 aiov.iov_base = PTRIN(args->buf); 922 aiov.iov_len = args->len; 923 msg.msg_control = 0; 924 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 925 926 error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL); 927 if (error != 0) 928 goto out; 929 930 if (PTRIN(args->from) != NULL) { 931 error = bsd_to_linux_sockaddr(sa, &lsa, msg.msg_namelen); 932 if (error == 0) 933 error = copyout(lsa, PTRIN(args->from), 934 msg.msg_namelen); 935 free(lsa, M_SONAME); 936 } 937 938 if (error == 0 && PTRIN(args->fromlen) != NULL) 939 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 940 sizeof(msg.msg_namelen)); 941 out: 942 free(sa, M_SONAME); 943 return (error); 944 } 945 946 static int 947 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 948 l_uint flags) 949 { 950 struct cmsghdr *cmsg; 951 struct mbuf *control; 952 struct msghdr msg; 953 struct l_cmsghdr linux_cmsg; 954 struct l_cmsghdr *ptr_cmsg; 955 struct l_msghdr linux_msghdr; 956 struct iovec *iov; 957 socklen_t datalen; 958 struct sockaddr *sa; 959 struct socket *so; 960 sa_family_t sa_family; 961 struct file *fp; 962 void *data; 963 l_size_t len; 964 l_size_t clen; 965 int error, fflag; 966 967 error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr)); 968 if (error != 0) 969 return (error); 970 971 /* 972 * Some Linux applications (ping) define a non-NULL control data 973 * pointer, but a msg_controllen of 0, which is not allowed in the 974 * FreeBSD system call interface. NULL the msg_control pointer in 975 * order to handle this case. This should be checked, but allows the 976 * Linux ping to work. 977 */ 978 if (PTRIN(linux_msghdr.msg_control) != NULL && 979 linux_msghdr.msg_controllen == 0) 980 linux_msghdr.msg_control = PTROUT(NULL); 981 982 error = linux_to_bsd_msghdr(&msg, &linux_msghdr); 983 if (error != 0) 984 return (error); 985 986 #ifdef COMPAT_LINUX32 987 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 988 &iov, EMSGSIZE); 989 #else 990 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 991 #endif 992 if (error != 0) 993 return (error); 994 995 control = NULL; 996 997 error = kern_getsockname(td, s, &sa, &datalen); 998 if (error != 0) 999 goto bad; 1000 sa_family = sa->sa_family; 1001 free(sa, M_SONAME); 1002 1003 if (flags & LINUX_MSG_OOB) { 1004 error = EOPNOTSUPP; 1005 if (sa_family == AF_UNIX) 1006 goto bad; 1007 1008 error = getsock_cap(td, s, &cap_send_rights, &fp, 1009 &fflag, NULL); 1010 if (error != 0) 1011 goto bad; 1012 so = fp->f_data; 1013 if (so->so_type != SOCK_STREAM) 1014 error = EOPNOTSUPP; 1015 fdrop(fp, td); 1016 if (error != 0) 1017 goto bad; 1018 } 1019 1020 if (linux_msghdr.msg_controllen >= sizeof(struct l_cmsghdr)) { 1021 error = ENOBUFS; 1022 control = m_get(M_WAITOK, MT_CONTROL); 1023 MCLGET(control, M_WAITOK); 1024 data = mtod(control, void *); 1025 datalen = 0; 1026 1027 ptr_cmsg = PTRIN(linux_msghdr.msg_control); 1028 clen = linux_msghdr.msg_controllen; 1029 do { 1030 error = copyin(ptr_cmsg, &linux_cmsg, 1031 sizeof(struct l_cmsghdr)); 1032 if (error != 0) 1033 goto bad; 1034 1035 error = EINVAL; 1036 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) || 1037 linux_cmsg.cmsg_len > clen) 1038 goto bad; 1039 1040 if (datalen + CMSG_HDRSZ > MCLBYTES) 1041 goto bad; 1042 1043 /* 1044 * Now we support only SCM_RIGHTS and SCM_CRED, 1045 * so return EINVAL in any other cmsg_type 1046 */ 1047 cmsg = data; 1048 cmsg->cmsg_type = 1049 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1050 cmsg->cmsg_level = 1051 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1052 if (cmsg->cmsg_type == -1 1053 || cmsg->cmsg_level != SOL_SOCKET) { 1054 linux_msg(curthread, 1055 "unsupported sendmsg cmsg level %d type %d", 1056 linux_cmsg.cmsg_level, linux_cmsg.cmsg_type); 1057 goto bad; 1058 } 1059 1060 /* 1061 * Some applications (e.g. pulseaudio) attempt to 1062 * send ancillary data even if the underlying protocol 1063 * doesn't support it which is not allowed in the 1064 * FreeBSD system call interface. 1065 */ 1066 if (sa_family != AF_UNIX) 1067 goto next; 1068 1069 if (cmsg->cmsg_type == SCM_CREDS) { 1070 len = sizeof(struct cmsgcred); 1071 if (datalen + CMSG_SPACE(len) > MCLBYTES) 1072 goto bad; 1073 1074 /* 1075 * The lower levels will fill in the structure 1076 */ 1077 memset(CMSG_DATA(data), 0, len); 1078 } else { 1079 len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1080 if (datalen + CMSG_SPACE(len) < datalen || 1081 datalen + CMSG_SPACE(len) > MCLBYTES) 1082 goto bad; 1083 1084 error = copyin(LINUX_CMSG_DATA(ptr_cmsg), 1085 CMSG_DATA(data), len); 1086 if (error != 0) 1087 goto bad; 1088 } 1089 1090 cmsg->cmsg_len = CMSG_LEN(len); 1091 data = (char *)data + CMSG_SPACE(len); 1092 datalen += CMSG_SPACE(len); 1093 1094 next: 1095 if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len)) 1096 break; 1097 1098 clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len); 1099 ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg + 1100 LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len)); 1101 } while(clen >= sizeof(struct l_cmsghdr)); 1102 1103 control->m_len = datalen; 1104 if (datalen == 0) { 1105 m_freem(control); 1106 control = NULL; 1107 } 1108 } 1109 1110 msg.msg_iov = iov; 1111 msg.msg_flags = 0; 1112 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1113 control = NULL; 1114 1115 bad: 1116 m_freem(control); 1117 free(iov, M_IOV); 1118 return (error); 1119 } 1120 1121 int 1122 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1123 { 1124 1125 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1126 args->flags)); 1127 } 1128 1129 int 1130 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1131 { 1132 struct l_mmsghdr *msg; 1133 l_uint retval; 1134 int error, datagrams; 1135 1136 if (args->vlen > UIO_MAXIOV) 1137 args->vlen = UIO_MAXIOV; 1138 1139 msg = PTRIN(args->msg); 1140 datagrams = 0; 1141 while (datagrams < args->vlen) { 1142 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1143 args->flags); 1144 if (error != 0) 1145 break; 1146 1147 retval = td->td_retval[0]; 1148 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1149 if (error != 0) 1150 break; 1151 ++msg; 1152 ++datagrams; 1153 } 1154 if (error == 0) 1155 td->td_retval[0] = datagrams; 1156 return (error); 1157 } 1158 1159 static int 1160 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1161 l_uint flags, struct msghdr *msg) 1162 { 1163 struct cmsghdr *cm; 1164 struct cmsgcred *cmcred; 1165 struct l_cmsghdr *linux_cmsg = NULL; 1166 struct l_ucred linux_ucred; 1167 socklen_t datalen, maxlen, outlen; 1168 struct l_msghdr linux_msghdr; 1169 struct iovec *iov, *uiov; 1170 struct mbuf *control = NULL; 1171 struct mbuf **controlp; 1172 struct timeval *ftmvl; 1173 struct l_sockaddr *lsa; 1174 struct sockaddr *sa; 1175 l_timeval ltmvl; 1176 caddr_t outbuf; 1177 void *data; 1178 int error, i, fd, fds, *fdp; 1179 1180 error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr)); 1181 if (error != 0) 1182 return (error); 1183 1184 error = linux_to_bsd_msghdr(msg, &linux_msghdr); 1185 if (error != 0) 1186 return (error); 1187 1188 #ifdef COMPAT_LINUX32 1189 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1190 &iov, EMSGSIZE); 1191 #else 1192 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1193 #endif 1194 if (error != 0) 1195 return (error); 1196 1197 if (msg->msg_name != NULL && msg->msg_namelen > 0) { 1198 msg->msg_namelen = min(msg->msg_namelen, SOCK_MAXADDRLEN); 1199 sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK); 1200 msg->msg_name = sa; 1201 } else { 1202 sa = NULL; 1203 msg->msg_name = NULL; 1204 } 1205 1206 uiov = msg->msg_iov; 1207 msg->msg_iov = iov; 1208 controlp = (msg->msg_control != NULL) ? &control : NULL; 1209 error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp); 1210 msg->msg_iov = uiov; 1211 if (error != 0) 1212 goto bad; 1213 1214 /* 1215 * Note that kern_recvit() updates msg->msg_namelen. 1216 */ 1217 if (msg->msg_name != NULL && msg->msg_namelen > 0) { 1218 msg->msg_name = PTRIN(linux_msghdr.msg_name); 1219 error = bsd_to_linux_sockaddr(sa, &lsa, msg->msg_namelen); 1220 if (error == 0) 1221 error = copyout(lsa, PTRIN(msg->msg_name), 1222 msg->msg_namelen); 1223 free(lsa, M_SONAME); 1224 if (error != 0) 1225 goto bad; 1226 } 1227 1228 error = bsd_to_linux_msghdr(msg, &linux_msghdr); 1229 if (error != 0) 1230 goto bad; 1231 1232 maxlen = linux_msghdr.msg_controllen; 1233 linux_msghdr.msg_controllen = 0; 1234 if (control) { 1235 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1236 1237 msg->msg_control = mtod(control, struct cmsghdr *); 1238 msg->msg_controllen = control->m_len; 1239 1240 cm = CMSG_FIRSTHDR(msg); 1241 outbuf = PTRIN(linux_msghdr.msg_control); 1242 outlen = 0; 1243 while (cm != NULL) { 1244 linux_cmsg->cmsg_type = 1245 bsd_to_linux_cmsg_type(cm->cmsg_type); 1246 linux_cmsg->cmsg_level = 1247 bsd_to_linux_sockopt_level(cm->cmsg_level); 1248 if (linux_cmsg->cmsg_type == -1 || 1249 cm->cmsg_level != SOL_SOCKET) { 1250 linux_msg(curthread, 1251 "unsupported recvmsg cmsg level %d type %d", 1252 cm->cmsg_level, cm->cmsg_type); 1253 error = EINVAL; 1254 goto bad; 1255 } 1256 1257 data = CMSG_DATA(cm); 1258 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1259 1260 switch (cm->cmsg_type) { 1261 case SCM_RIGHTS: 1262 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1263 fds = datalen / sizeof(int); 1264 fdp = data; 1265 for (i = 0; i < fds; i++) { 1266 fd = *fdp++; 1267 (void)kern_fcntl(td, fd, 1268 F_SETFD, FD_CLOEXEC); 1269 } 1270 } 1271 break; 1272 1273 case SCM_CREDS: 1274 /* 1275 * Currently LOCAL_CREDS is never in 1276 * effect for Linux so no need to worry 1277 * about sockcred 1278 */ 1279 if (datalen != sizeof(*cmcred)) { 1280 error = EMSGSIZE; 1281 goto bad; 1282 } 1283 cmcred = (struct cmsgcred *)data; 1284 bzero(&linux_ucred, sizeof(linux_ucred)); 1285 linux_ucred.pid = cmcred->cmcred_pid; 1286 linux_ucred.uid = cmcred->cmcred_uid; 1287 linux_ucred.gid = cmcred->cmcred_gid; 1288 data = &linux_ucred; 1289 datalen = sizeof(linux_ucred); 1290 break; 1291 1292 case SCM_TIMESTAMP: 1293 if (datalen != sizeof(struct timeval)) { 1294 error = EMSGSIZE; 1295 goto bad; 1296 } 1297 ftmvl = (struct timeval *)data; 1298 ltmvl.tv_sec = ftmvl->tv_sec; 1299 ltmvl.tv_usec = ftmvl->tv_usec; 1300 data = <mvl; 1301 datalen = sizeof(ltmvl); 1302 break; 1303 } 1304 1305 if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) { 1306 if (outlen == 0) { 1307 error = EMSGSIZE; 1308 goto bad; 1309 } else { 1310 linux_msghdr.msg_flags |= LINUX_MSG_CTRUNC; 1311 m_dispose_extcontrolm(control); 1312 goto out; 1313 } 1314 } 1315 1316 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1317 1318 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1319 if (error != 0) 1320 goto bad; 1321 outbuf += L_CMSG_HDRSZ; 1322 1323 error = copyout(data, outbuf, datalen); 1324 if (error != 0) 1325 goto bad; 1326 1327 outbuf += LINUX_CMSG_ALIGN(datalen); 1328 outlen += LINUX_CMSG_LEN(datalen); 1329 1330 cm = CMSG_NXTHDR(msg, cm); 1331 } 1332 linux_msghdr.msg_controllen = outlen; 1333 } 1334 1335 out: 1336 error = copyout(&linux_msghdr, msghdr, sizeof(linux_msghdr)); 1337 1338 bad: 1339 if (control != NULL) { 1340 if (error != 0) 1341 m_dispose_extcontrolm(control); 1342 m_freem(control); 1343 } 1344 free(iov, M_IOV); 1345 free(linux_cmsg, M_LINUX); 1346 free(sa, M_SONAME); 1347 1348 return (error); 1349 } 1350 1351 int 1352 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1353 { 1354 struct msghdr bsd_msg; 1355 1356 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1357 args->flags, &bsd_msg)); 1358 } 1359 1360 int 1361 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1362 { 1363 struct l_mmsghdr *msg; 1364 struct msghdr bsd_msg; 1365 struct l_timespec lts; 1366 struct timespec ts, tts; 1367 l_uint retval; 1368 int error, datagrams; 1369 1370 if (args->timeout) { 1371 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1372 if (error != 0) 1373 return (error); 1374 error = linux_to_native_timespec(&ts, <s); 1375 if (error != 0) 1376 return (error); 1377 getnanotime(&tts); 1378 timespecadd(&tts, &ts, &tts); 1379 } 1380 1381 msg = PTRIN(args->msg); 1382 datagrams = 0; 1383 while (datagrams < args->vlen) { 1384 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1385 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1386 if (error != 0) 1387 break; 1388 1389 retval = td->td_retval[0]; 1390 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1391 if (error != 0) 1392 break; 1393 ++msg; 1394 ++datagrams; 1395 1396 /* 1397 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1398 */ 1399 if (args->flags & LINUX_MSG_WAITFORONE) 1400 args->flags |= LINUX_MSG_DONTWAIT; 1401 1402 /* 1403 * See BUGS section of recvmmsg(2). 1404 */ 1405 if (args->timeout) { 1406 getnanotime(&ts); 1407 timespecsub(&ts, &tts, &ts); 1408 if (!timespecisset(&ts) || ts.tv_sec > 0) 1409 break; 1410 } 1411 /* Out of band data, return right away. */ 1412 if (bsd_msg.msg_flags & MSG_OOB) 1413 break; 1414 } 1415 if (error == 0) 1416 td->td_retval[0] = datagrams; 1417 return (error); 1418 } 1419 1420 int 1421 linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1422 { 1423 1424 return (kern_shutdown(td, args->s, args->how)); 1425 } 1426 1427 int 1428 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1429 { 1430 l_timeval linux_tv; 1431 struct sockaddr *sa; 1432 struct timeval tv; 1433 socklen_t len; 1434 int error, level, name; 1435 1436 level = linux_to_bsd_sockopt_level(args->level); 1437 switch (level) { 1438 case SOL_SOCKET: 1439 name = linux_to_bsd_so_sockopt(args->optname); 1440 switch (name) { 1441 case SO_RCVTIMEO: 1442 /* FALLTHROUGH */ 1443 case SO_SNDTIMEO: 1444 error = copyin(PTRIN(args->optval), &linux_tv, 1445 sizeof(linux_tv)); 1446 if (error != 0) 1447 return (error); 1448 tv.tv_sec = linux_tv.tv_sec; 1449 tv.tv_usec = linux_tv.tv_usec; 1450 return (kern_setsockopt(td, args->s, level, 1451 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1452 /* NOTREACHED */ 1453 default: 1454 break; 1455 } 1456 break; 1457 case IPPROTO_IP: 1458 if (args->optname == LINUX_IP_RECVERR && 1459 linux_ignore_ip_recverr) { 1460 /* 1461 * XXX: This is a hack to unbreak DNS resolution 1462 * with glibc 2.30 and above. 1463 */ 1464 return (0); 1465 } 1466 name = linux_to_bsd_ip_sockopt(args->optname); 1467 break; 1468 case IPPROTO_IPV6: 1469 name = linux_to_bsd_ip6_sockopt(args->optname); 1470 break; 1471 case IPPROTO_TCP: 1472 name = linux_to_bsd_tcp_sockopt(args->optname); 1473 break; 1474 default: 1475 name = -1; 1476 break; 1477 } 1478 if (name == -1) { 1479 linux_msg(curthread, 1480 "unsupported setsockopt level %d optname %d", 1481 args->level, args->optname); 1482 return (ENOPROTOOPT); 1483 } 1484 1485 if (name == IPV6_NEXTHOP) { 1486 len = args->optlen; 1487 error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len); 1488 if (error != 0) 1489 return (error); 1490 1491 error = kern_setsockopt(td, args->s, level, 1492 name, sa, UIO_SYSSPACE, len); 1493 free(sa, M_SONAME); 1494 } else { 1495 error = kern_setsockopt(td, args->s, level, 1496 name, PTRIN(args->optval), UIO_USERSPACE, args->optlen); 1497 } 1498 1499 return (error); 1500 } 1501 1502 int 1503 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1504 { 1505 l_timeval linux_tv; 1506 struct timeval tv; 1507 socklen_t tv_len, xulen, len; 1508 struct l_sockaddr *lsa; 1509 struct sockaddr *sa; 1510 struct xucred xu; 1511 struct l_ucred lxu; 1512 int error, level, name, newval; 1513 1514 level = linux_to_bsd_sockopt_level(args->level); 1515 switch (level) { 1516 case SOL_SOCKET: 1517 name = linux_to_bsd_so_sockopt(args->optname); 1518 switch (name) { 1519 case SO_RCVTIMEO: 1520 /* FALLTHROUGH */ 1521 case SO_SNDTIMEO: 1522 tv_len = sizeof(tv); 1523 error = kern_getsockopt(td, args->s, level, 1524 name, &tv, UIO_SYSSPACE, &tv_len); 1525 if (error != 0) 1526 return (error); 1527 linux_tv.tv_sec = tv.tv_sec; 1528 linux_tv.tv_usec = tv.tv_usec; 1529 return (copyout(&linux_tv, PTRIN(args->optval), 1530 sizeof(linux_tv))); 1531 /* NOTREACHED */ 1532 case LOCAL_PEERCRED: 1533 if (args->optlen < sizeof(lxu)) 1534 return (EINVAL); 1535 /* 1536 * LOCAL_PEERCRED is not served at the SOL_SOCKET level, 1537 * but by the Unix socket's level 0. 1538 */ 1539 level = 0; 1540 xulen = sizeof(xu); 1541 error = kern_getsockopt(td, args->s, level, 1542 name, &xu, UIO_SYSSPACE, &xulen); 1543 if (error != 0) 1544 return (error); 1545 lxu.pid = xu.cr_pid; 1546 lxu.uid = xu.cr_uid; 1547 lxu.gid = xu.cr_gid; 1548 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1549 /* NOTREACHED */ 1550 case SO_ERROR: 1551 len = sizeof(newval); 1552 error = kern_getsockopt(td, args->s, level, 1553 name, &newval, UIO_SYSSPACE, &len); 1554 if (error != 0) 1555 return (error); 1556 newval = -SV_ABI_ERRNO(td->td_proc, newval); 1557 return (copyout(&newval, PTRIN(args->optval), len)); 1558 /* NOTREACHED */ 1559 default: 1560 break; 1561 } 1562 break; 1563 case IPPROTO_IP: 1564 name = linux_to_bsd_ip_sockopt(args->optname); 1565 break; 1566 case IPPROTO_IPV6: 1567 name = linux_to_bsd_ip6_sockopt(args->optname); 1568 break; 1569 case IPPROTO_TCP: 1570 name = linux_to_bsd_tcp_sockopt(args->optname); 1571 break; 1572 default: 1573 name = -1; 1574 break; 1575 } 1576 if (name == -1) { 1577 linux_msg(curthread, 1578 "unsupported getsockopt level %d optname %d", 1579 args->level, args->optname); 1580 return (EINVAL); 1581 } 1582 1583 if (name == IPV6_NEXTHOP) { 1584 error = copyin(PTRIN(args->optlen), &len, sizeof(len)); 1585 if (error != 0) 1586 return (error); 1587 sa = malloc(len, M_SONAME, M_WAITOK); 1588 1589 error = kern_getsockopt(td, args->s, level, 1590 name, sa, UIO_SYSSPACE, &len); 1591 if (error != 0) 1592 goto out; 1593 1594 error = bsd_to_linux_sockaddr(sa, &lsa, len); 1595 if (error == 0) 1596 error = copyout(lsa, PTRIN(args->optval), len); 1597 free(lsa, M_SONAME); 1598 if (error == 0) 1599 error = copyout(&len, PTRIN(args->optlen), 1600 sizeof(len)); 1601 out: 1602 free(sa, M_SONAME); 1603 } else { 1604 if (args->optval) { 1605 error = copyin(PTRIN(args->optlen), &len, sizeof(len)); 1606 if (error != 0) 1607 return (error); 1608 } 1609 error = kern_getsockopt(td, args->s, level, 1610 name, PTRIN(args->optval), UIO_USERSPACE, &len); 1611 if (error == 0) 1612 error = copyout(&len, PTRIN(args->optlen), 1613 sizeof(len)); 1614 } 1615 1616 return (error); 1617 } 1618 1619 static int 1620 linux_sendfile_common(struct thread *td, l_int out, l_int in, 1621 l_loff_t *offset, l_size_t count) 1622 { 1623 off_t bytes_read; 1624 int error; 1625 l_loff_t current_offset; 1626 struct file *fp; 1627 1628 AUDIT_ARG_FD(in); 1629 error = fget_read(td, in, &cap_pread_rights, &fp); 1630 if (error != 0) 1631 return (error); 1632 1633 if (offset != NULL) { 1634 current_offset = *offset; 1635 } else { 1636 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ? 1637 fo_seek(fp, 0, SEEK_CUR, td) : ESPIPE; 1638 if (error != 0) 1639 goto drop; 1640 current_offset = td->td_uretoff.tdu_off; 1641 } 1642 1643 bytes_read = 0; 1644 1645 /* Linux cannot have 0 count. */ 1646 if (count <= 0 || current_offset < 0) { 1647 error = EINVAL; 1648 goto drop; 1649 } 1650 1651 error = fo_sendfile(fp, out, NULL, NULL, current_offset, count, 1652 &bytes_read, 0, td); 1653 if (error != 0) 1654 goto drop; 1655 current_offset += bytes_read; 1656 1657 if (offset != NULL) { 1658 *offset = current_offset; 1659 } else { 1660 error = fo_seek(fp, current_offset, SEEK_SET, td); 1661 if (error != 0) 1662 goto drop; 1663 } 1664 1665 td->td_retval[0] = (ssize_t)bytes_read; 1666 drop: 1667 fdrop(fp, td); 1668 return (error); 1669 } 1670 1671 int 1672 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg) 1673 { 1674 /* 1675 * Differences between FreeBSD and Linux sendfile: 1676 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to 1677 * mean send the whole file.) In linux_sendfile given fds are still 1678 * checked for validity when the count is 0. 1679 * - Linux can send to any fd whereas FreeBSD only supports sockets. 1680 * The same restriction follows for linux_sendfile. 1681 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr. 1682 * - Linux takes an offset pointer and updates it to the read location. 1683 * FreeBSD takes in an offset and a 'bytes read' parameter which is 1684 * only filled if it isn't NULL. We use this parameter to update the 1685 * offset pointer if it exists. 1686 * - Linux sendfile returns bytes read on success while FreeBSD 1687 * returns 0. We use the 'bytes read' parameter to get this value. 1688 */ 1689 1690 l_loff_t offset64; 1691 l_long offset; 1692 int ret; 1693 int error; 1694 1695 if (arg->offset != NULL) { 1696 error = copyin(arg->offset, &offset, sizeof(offset)); 1697 if (error != 0) 1698 return (error); 1699 offset64 = (l_loff_t)offset; 1700 } 1701 1702 ret = linux_sendfile_common(td, arg->out, arg->in, 1703 arg->offset != NULL ? &offset64 : NULL, arg->count); 1704 1705 if (arg->offset != NULL) { 1706 #if defined(__i386__) || defined(__arm__) || \ 1707 (defined(__amd64__) && defined(COMPAT_LINUX32)) 1708 if (offset64 > INT32_MAX) 1709 return (EOVERFLOW); 1710 #endif 1711 offset = (l_long)offset64; 1712 error = copyout(&offset, arg->offset, sizeof(offset)); 1713 if (error != 0) 1714 return (error); 1715 } 1716 1717 return (ret); 1718 } 1719 1720 #if defined(__i386__) || defined(__arm__) || \ 1721 (defined(__amd64__) && defined(COMPAT_LINUX32)) 1722 1723 int 1724 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg) 1725 { 1726 l_loff_t offset; 1727 int ret; 1728 int error; 1729 1730 if (arg->offset != NULL) { 1731 error = copyin(arg->offset, &offset, sizeof(offset)); 1732 if (error != 0) 1733 return (error); 1734 } 1735 1736 ret = linux_sendfile_common(td, arg->out, arg->in, 1737 arg->offset != NULL ? &offset : NULL, arg->count); 1738 1739 if (arg->offset != NULL) { 1740 error = copyout(&offset, arg->offset, sizeof(offset)); 1741 if (error != 0) 1742 return (error); 1743 } 1744 1745 return (ret); 1746 } 1747 1748 /* Argument list sizes for linux_socketcall */ 1749 static const unsigned char lxs_args_cnt[] = { 1750 0 /* unused*/, 3 /* socket */, 1751 3 /* bind */, 3 /* connect */, 1752 2 /* listen */, 3 /* accept */, 1753 3 /* getsockname */, 3 /* getpeername */, 1754 4 /* socketpair */, 4 /* send */, 1755 4 /* recv */, 6 /* sendto */, 1756 6 /* recvfrom */, 2 /* shutdown */, 1757 5 /* setsockopt */, 5 /* getsockopt */, 1758 3 /* sendmsg */, 3 /* recvmsg */, 1759 4 /* accept4 */, 5 /* recvmmsg */, 1760 4 /* sendmmsg */, 4 /* sendfile */ 1761 }; 1762 #define LINUX_ARGS_CNT (nitems(lxs_args_cnt) - 1) 1763 #define LINUX_ARG_SIZE(x) (lxs_args_cnt[x] * sizeof(l_ulong)) 1764 1765 int 1766 linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1767 { 1768 l_ulong a[6]; 1769 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1770 register_t l_args[6]; 1771 #endif 1772 void *arg; 1773 int error; 1774 1775 if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT) 1776 return (EINVAL); 1777 error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what)); 1778 if (error != 0) 1779 return (error); 1780 1781 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1782 for (int i = 0; i < lxs_args_cnt[args->what]; ++i) 1783 l_args[i] = a[i]; 1784 arg = l_args; 1785 #else 1786 arg = a; 1787 #endif 1788 switch (args->what) { 1789 case LINUX_SOCKET: 1790 return (linux_socket(td, arg)); 1791 case LINUX_BIND: 1792 return (linux_bind(td, arg)); 1793 case LINUX_CONNECT: 1794 return (linux_connect(td, arg)); 1795 case LINUX_LISTEN: 1796 return (linux_listen(td, arg)); 1797 case LINUX_ACCEPT: 1798 return (linux_accept(td, arg)); 1799 case LINUX_GETSOCKNAME: 1800 return (linux_getsockname(td, arg)); 1801 case LINUX_GETPEERNAME: 1802 return (linux_getpeername(td, arg)); 1803 case LINUX_SOCKETPAIR: 1804 return (linux_socketpair(td, arg)); 1805 case LINUX_SEND: 1806 return (linux_send(td, arg)); 1807 case LINUX_RECV: 1808 return (linux_recv(td, arg)); 1809 case LINUX_SENDTO: 1810 return (linux_sendto(td, arg)); 1811 case LINUX_RECVFROM: 1812 return (linux_recvfrom(td, arg)); 1813 case LINUX_SHUTDOWN: 1814 return (linux_shutdown(td, arg)); 1815 case LINUX_SETSOCKOPT: 1816 return (linux_setsockopt(td, arg)); 1817 case LINUX_GETSOCKOPT: 1818 return (linux_getsockopt(td, arg)); 1819 case LINUX_SENDMSG: 1820 return (linux_sendmsg(td, arg)); 1821 case LINUX_RECVMSG: 1822 return (linux_recvmsg(td, arg)); 1823 case LINUX_ACCEPT4: 1824 return (linux_accept4(td, arg)); 1825 case LINUX_RECVMMSG: 1826 return (linux_recvmmsg(td, arg)); 1827 case LINUX_SENDMMSG: 1828 return (linux_sendmmsg(td, arg)); 1829 case LINUX_SENDFILE: 1830 return (linux_sendfile(td, arg)); 1831 } 1832 1833 linux_msg(td, "socket type %d not implemented", args->what); 1834 return (ENOSYS); 1835 } 1836 #endif /* __i386__ || __arm__ || (__amd64__ && COMPAT_LINUX32) */ 1837