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_PASSCRED: 226 return (LOCAL_CREDS_PERSISTENT); 227 case LINUX_SO_PEERCRED: 228 return (LOCAL_PEERCRED); 229 case LINUX_SO_RCVLOWAT: 230 return (SO_RCVLOWAT); 231 case LINUX_SO_SNDLOWAT: 232 return (SO_SNDLOWAT); 233 case LINUX_SO_RCVTIMEO: 234 return (SO_RCVTIMEO); 235 case LINUX_SO_SNDTIMEO: 236 return (SO_SNDTIMEO); 237 case LINUX_SO_TIMESTAMP: 238 return (SO_TIMESTAMP); 239 case LINUX_SO_ACCEPTCONN: 240 return (SO_ACCEPTCONN); 241 case LINUX_SO_PROTOCOL: 242 return (SO_PROTOCOL); 243 } 244 return (-1); 245 } 246 247 static int 248 linux_to_bsd_tcp_sockopt(int opt) 249 { 250 251 switch (opt) { 252 case LINUX_TCP_NODELAY: 253 return (TCP_NODELAY); 254 case LINUX_TCP_MAXSEG: 255 return (TCP_MAXSEG); 256 case LINUX_TCP_CORK: 257 return (TCP_NOPUSH); 258 case LINUX_TCP_KEEPIDLE: 259 return (TCP_KEEPIDLE); 260 case LINUX_TCP_KEEPINTVL: 261 return (TCP_KEEPINTVL); 262 case LINUX_TCP_KEEPCNT: 263 return (TCP_KEEPCNT); 264 case LINUX_TCP_MD5SIG: 265 return (TCP_MD5SIG); 266 } 267 return (-1); 268 } 269 270 static int 271 linux_to_bsd_msg_flags(int flags) 272 { 273 int ret_flags = 0; 274 275 if (flags & LINUX_MSG_OOB) 276 ret_flags |= MSG_OOB; 277 if (flags & LINUX_MSG_PEEK) 278 ret_flags |= MSG_PEEK; 279 if (flags & LINUX_MSG_DONTROUTE) 280 ret_flags |= MSG_DONTROUTE; 281 if (flags & LINUX_MSG_CTRUNC) 282 ret_flags |= MSG_CTRUNC; 283 if (flags & LINUX_MSG_TRUNC) 284 ret_flags |= MSG_TRUNC; 285 if (flags & LINUX_MSG_DONTWAIT) 286 ret_flags |= MSG_DONTWAIT; 287 if (flags & LINUX_MSG_EOR) 288 ret_flags |= MSG_EOR; 289 if (flags & LINUX_MSG_WAITALL) 290 ret_flags |= MSG_WAITALL; 291 if (flags & LINUX_MSG_NOSIGNAL) 292 ret_flags |= MSG_NOSIGNAL; 293 #if 0 /* not handled */ 294 if (flags & LINUX_MSG_PROXY) 295 ; 296 if (flags & LINUX_MSG_FIN) 297 ; 298 if (flags & LINUX_MSG_SYN) 299 ; 300 if (flags & LINUX_MSG_CONFIRM) 301 ; 302 if (flags & LINUX_MSG_RST) 303 ; 304 if (flags & LINUX_MSG_ERRQUEUE) 305 ; 306 #endif 307 return (ret_flags); 308 } 309 310 static int 311 linux_to_bsd_cmsg_type(int cmsg_type) 312 { 313 314 switch (cmsg_type) { 315 case LINUX_SCM_RIGHTS: 316 return (SCM_RIGHTS); 317 case LINUX_SCM_CREDENTIALS: 318 return (SCM_CREDS); 319 } 320 return (-1); 321 } 322 323 static int 324 bsd_to_linux_cmsg_type(int cmsg_type) 325 { 326 327 switch (cmsg_type) { 328 case SCM_RIGHTS: 329 return (LINUX_SCM_RIGHTS); 330 case SCM_CREDS: 331 return (LINUX_SCM_CREDENTIALS); 332 case SCM_TIMESTAMP: 333 return (LINUX_SCM_TIMESTAMP); 334 } 335 return (-1); 336 } 337 338 static int 339 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 340 { 341 if (lhdr->msg_controllen > INT_MAX) 342 return (ENOBUFS); 343 344 bhdr->msg_name = PTRIN(lhdr->msg_name); 345 bhdr->msg_namelen = lhdr->msg_namelen; 346 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 347 bhdr->msg_iovlen = lhdr->msg_iovlen; 348 bhdr->msg_control = PTRIN(lhdr->msg_control); 349 350 /* 351 * msg_controllen is skipped since BSD and LINUX control messages 352 * are potentially different sizes (e.g. the cred structure used 353 * by SCM_CREDS is different between the two operating system). 354 * 355 * The caller can set it (if necessary) after converting all the 356 * control messages. 357 */ 358 359 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 360 return (0); 361 } 362 363 static int 364 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 365 { 366 lhdr->msg_name = PTROUT(bhdr->msg_name); 367 lhdr->msg_namelen = bhdr->msg_namelen; 368 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 369 lhdr->msg_iovlen = bhdr->msg_iovlen; 370 lhdr->msg_control = PTROUT(bhdr->msg_control); 371 372 /* 373 * msg_controllen is skipped since BSD and LINUX control messages 374 * are potentially different sizes (e.g. the cred structure used 375 * by SCM_CREDS is different between the two operating system). 376 * 377 * The caller can set it (if necessary) after converting all the 378 * control messages. 379 */ 380 381 /* msg_flags skipped */ 382 return (0); 383 } 384 385 static int 386 linux_set_socket_flags(int lflags, int *flags) 387 { 388 389 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 390 return (EINVAL); 391 if (lflags & LINUX_SOCK_NONBLOCK) 392 *flags |= SOCK_NONBLOCK; 393 if (lflags & LINUX_SOCK_CLOEXEC) 394 *flags |= SOCK_CLOEXEC; 395 return (0); 396 } 397 398 static int 399 linux_copyout_sockaddr(const struct sockaddr *sa, void *uaddr, size_t len) 400 { 401 struct l_sockaddr *lsa; 402 int error; 403 404 error = bsd_to_linux_sockaddr(sa, &lsa, len); 405 if (error != 0) 406 return (error); 407 408 error = copyout(lsa, uaddr, len); 409 free(lsa, M_SONAME); 410 411 return (error); 412 } 413 414 static int 415 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 416 struct mbuf *control, enum uio_seg segflg) 417 { 418 struct sockaddr *to; 419 int error, len; 420 421 if (mp->msg_name != NULL) { 422 len = mp->msg_namelen; 423 error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len); 424 if (error != 0) 425 return (error); 426 mp->msg_name = to; 427 } else 428 to = NULL; 429 430 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 431 segflg); 432 433 if (to) 434 free(to, M_SONAME); 435 return (error); 436 } 437 438 /* Return 0 if IP_HDRINCL is set for the given socket. */ 439 static int 440 linux_check_hdrincl(struct thread *td, int s) 441 { 442 int error, optval; 443 socklen_t size_val; 444 445 size_val = sizeof(optval); 446 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 447 &optval, UIO_SYSSPACE, &size_val); 448 if (error != 0) 449 return (error); 450 451 return (optval == 0); 452 } 453 454 /* 455 * Updated sendto() when IP_HDRINCL is set: 456 * tweak endian-dependent fields in the IP packet. 457 */ 458 static int 459 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 460 { 461 /* 462 * linux_ip_copysize defines how many bytes we should copy 463 * from the beginning of the IP packet before we customize it for BSD. 464 * It should include all the fields we modify (ip_len and ip_off). 465 */ 466 #define linux_ip_copysize 8 467 468 struct ip *packet; 469 struct msghdr msg; 470 struct iovec aiov[1]; 471 int error; 472 473 /* Check that the packet isn't too big or too small. */ 474 if (linux_args->len < linux_ip_copysize || 475 linux_args->len > IP_MAXPACKET) 476 return (EINVAL); 477 478 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK); 479 480 /* Make kernel copy of the packet to be sent */ 481 if ((error = copyin(PTRIN(linux_args->msg), packet, 482 linux_args->len))) 483 goto goout; 484 485 /* Convert fields from Linux to BSD raw IP socket format */ 486 packet->ip_len = linux_args->len; 487 packet->ip_off = ntohs(packet->ip_off); 488 489 /* Prepare the msghdr and iovec structures describing the new packet */ 490 msg.msg_name = PTRIN(linux_args->to); 491 msg.msg_namelen = linux_args->tolen; 492 msg.msg_iov = aiov; 493 msg.msg_iovlen = 1; 494 msg.msg_control = NULL; 495 msg.msg_flags = 0; 496 aiov[0].iov_base = (char *)packet; 497 aiov[0].iov_len = linux_args->len; 498 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 499 NULL, UIO_SYSSPACE); 500 goout: 501 free(packet, M_LINUX); 502 return (error); 503 } 504 505 static const char *linux_netlink_names[] = { 506 [LINUX_NETLINK_ROUTE] = "ROUTE", 507 [LINUX_NETLINK_SOCK_DIAG] = "SOCK_DIAG", 508 [LINUX_NETLINK_NFLOG] = "NFLOG", 509 [LINUX_NETLINK_SELINUX] = "SELINUX", 510 [LINUX_NETLINK_AUDIT] = "AUDIT", 511 [LINUX_NETLINK_FIB_LOOKUP] = "FIB_LOOKUP", 512 [LINUX_NETLINK_NETFILTER] = "NETFILTER", 513 [LINUX_NETLINK_KOBJECT_UEVENT] = "KOBJECT_UEVENT", 514 }; 515 516 int 517 linux_socket(struct thread *td, struct linux_socket_args *args) 518 { 519 int domain, retval_socket, type; 520 521 type = args->type & LINUX_SOCK_TYPE_MASK; 522 if (type < 0 || type > LINUX_SOCK_MAX) 523 return (EINVAL); 524 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 525 &type); 526 if (retval_socket != 0) 527 return (retval_socket); 528 domain = linux_to_bsd_domain(args->domain); 529 if (domain == -1) { 530 /* Mask off SOCK_NONBLOCK / CLOEXEC for error messages. */ 531 type = args->type & LINUX_SOCK_TYPE_MASK; 532 if (args->domain == LINUX_AF_NETLINK) { 533 const char *nl_name; 534 535 if (args->protocol >= 0 && 536 args->protocol < nitems(linux_netlink_names)) 537 nl_name = linux_netlink_names[args->protocol]; 538 else 539 nl_name = NULL; 540 if (nl_name != NULL) 541 linux_msg(curthread, 542 "unsupported socket(AF_NETLINK, %d, " 543 "NETLINK_%s)", type, nl_name); 544 else 545 linux_msg(curthread, 546 "unsupported socket(AF_NETLINK, %d, %d)", 547 type, args->protocol); 548 } else { 549 linux_msg(curthread, "unsupported socket domain %d, " 550 "type %d, protocol %d", args->domain, type, 551 args->protocol); 552 } 553 return (EAFNOSUPPORT); 554 } 555 556 retval_socket = kern_socket(td, domain, type, args->protocol); 557 if (retval_socket) 558 return (retval_socket); 559 560 if (type == SOCK_RAW 561 && (args->protocol == IPPROTO_RAW || args->protocol == 0) 562 && domain == PF_INET) { 563 /* It's a raw IP socket: set the IP_HDRINCL option. */ 564 int hdrincl; 565 566 hdrincl = 1; 567 /* We ignore any error returned by kern_setsockopt() */ 568 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 569 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 570 } 571 #ifdef INET6 572 /* 573 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default 574 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps. 575 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only 576 * sysctl value. 577 */ 578 if (domain == PF_INET6) { 579 int v6only; 580 581 v6only = 0; 582 /* We ignore any error returned by setsockopt() */ 583 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 584 &v6only, UIO_SYSSPACE, sizeof(v6only)); 585 } 586 #endif 587 588 return (retval_socket); 589 } 590 591 int 592 linux_bind(struct thread *td, struct linux_bind_args *args) 593 { 594 struct sockaddr *sa; 595 int error; 596 597 error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa, 598 &args->namelen); 599 if (error != 0) 600 return (error); 601 602 error = kern_bindat(td, AT_FDCWD, args->s, sa); 603 free(sa, M_SONAME); 604 605 /* XXX */ 606 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 607 return (EINVAL); 608 return (error); 609 } 610 611 int 612 linux_connect(struct thread *td, struct linux_connect_args *args) 613 { 614 struct socket *so; 615 struct sockaddr *sa; 616 struct file *fp; 617 u_int fflag; 618 int error; 619 620 error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa, 621 &args->namelen); 622 if (error != 0) 623 return (error); 624 625 error = kern_connectat(td, AT_FDCWD, args->s, sa); 626 free(sa, M_SONAME); 627 if (error != EISCONN) 628 return (error); 629 630 /* 631 * Linux doesn't return EISCONN the first time it occurs, 632 * when on a non-blocking socket. Instead it returns the 633 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 634 */ 635 error = getsock_cap(td, args->s, &cap_connect_rights, 636 &fp, &fflag, NULL); 637 if (error != 0) 638 return (error); 639 640 error = EISCONN; 641 so = fp->f_data; 642 if (fflag & FNONBLOCK) { 643 SOCK_LOCK(so); 644 if (so->so_emuldata == 0) 645 error = so->so_error; 646 so->so_emuldata = (void *)1; 647 SOCK_UNLOCK(so); 648 } 649 fdrop(fp, td); 650 651 return (error); 652 } 653 654 int 655 linux_listen(struct thread *td, struct linux_listen_args *args) 656 { 657 658 return (kern_listen(td, args->s, args->backlog)); 659 } 660 661 static int 662 linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 663 l_uintptr_t namelen, int flags) 664 { 665 struct sockaddr *sa; 666 struct file *fp, *fp1; 667 int bflags, len; 668 struct socket *so; 669 int error, error1; 670 671 bflags = 0; 672 fp = NULL; 673 sa = NULL; 674 675 error = linux_set_socket_flags(flags, &bflags); 676 if (error != 0) 677 return (error); 678 679 if (PTRIN(addr) == NULL) { 680 len = 0; 681 error = kern_accept4(td, s, NULL, NULL, bflags, NULL); 682 } else { 683 error = copyin(PTRIN(namelen), &len, sizeof(len)); 684 if (error != 0) 685 return (error); 686 if (len < 0) 687 return (EINVAL); 688 error = kern_accept4(td, s, &sa, &len, bflags, &fp); 689 } 690 691 /* 692 * Translate errno values into ones used by Linux. 693 */ 694 if (error != 0) { 695 /* 696 * XXX. This is wrong, different sockaddr structures 697 * have different sizes. 698 */ 699 switch (error) { 700 case EFAULT: 701 if (namelen != sizeof(struct sockaddr_in)) 702 error = EINVAL; 703 break; 704 case EINVAL: 705 error1 = getsock_cap(td, s, &cap_accept_rights, &fp1, NULL, NULL); 706 if (error1 != 0) { 707 error = error1; 708 break; 709 } 710 so = fp1->f_data; 711 if (so->so_type == SOCK_DGRAM) 712 error = EOPNOTSUPP; 713 fdrop(fp1, td); 714 break; 715 } 716 return (error); 717 } 718 719 if (len != 0) { 720 error = linux_copyout_sockaddr(sa, PTRIN(addr), len); 721 722 /* 723 * XXX: We should also copyout the len, shouldn't we? 724 */ 725 726 if (error != 0) { 727 fdclose(td, fp, td->td_retval[0]); 728 td->td_retval[0] = 0; 729 } 730 } 731 if (fp != NULL) 732 fdrop(fp, td); 733 free(sa, M_SONAME); 734 return (error); 735 } 736 737 int 738 linux_accept(struct thread *td, struct linux_accept_args *args) 739 { 740 741 return (linux_accept_common(td, args->s, args->addr, 742 args->namelen, 0)); 743 } 744 745 int 746 linux_accept4(struct thread *td, struct linux_accept4_args *args) 747 { 748 749 return (linux_accept_common(td, args->s, args->addr, 750 args->namelen, args->flags)); 751 } 752 753 int 754 linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 755 { 756 struct sockaddr *sa; 757 int len, error; 758 759 error = copyin(PTRIN(args->namelen), &len, sizeof(len)); 760 if (error != 0) 761 return (error); 762 763 error = kern_getsockname(td, args->s, &sa, &len); 764 if (error != 0) 765 return (error); 766 767 if (len != 0) 768 error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len); 769 770 free(sa, M_SONAME); 771 if (error == 0) 772 error = copyout(&len, PTRIN(args->namelen), sizeof(len)); 773 return (error); 774 } 775 776 int 777 linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 778 { 779 struct sockaddr *sa; 780 int len, error; 781 782 error = copyin(PTRIN(args->namelen), &len, sizeof(len)); 783 if (error != 0) 784 return (error); 785 if (len < 0) 786 return (EINVAL); 787 788 error = kern_getpeername(td, args->s, &sa, &len); 789 if (error != 0) 790 return (error); 791 792 if (len != 0) 793 error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len); 794 795 free(sa, M_SONAME); 796 if (error == 0) 797 error = copyout(&len, PTRIN(args->namelen), sizeof(len)); 798 return (error); 799 } 800 801 int 802 linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 803 { 804 int domain, error, sv[2], type; 805 806 domain = linux_to_bsd_domain(args->domain); 807 if (domain != PF_LOCAL) 808 return (EAFNOSUPPORT); 809 type = args->type & LINUX_SOCK_TYPE_MASK; 810 if (type < 0 || type > LINUX_SOCK_MAX) 811 return (EINVAL); 812 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 813 &type); 814 if (error != 0) 815 return (error); 816 if (args->protocol != 0 && args->protocol != PF_UNIX) { 817 /* 818 * Use of PF_UNIX as protocol argument is not right, 819 * but Linux does it. 820 * Do not map PF_UNIX as its Linux value is identical 821 * to FreeBSD one. 822 */ 823 return (EPROTONOSUPPORT); 824 } 825 error = kern_socketpair(td, domain, type, 0, sv); 826 if (error != 0) 827 return (error); 828 error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int)); 829 if (error != 0) { 830 (void)kern_close(td, sv[0]); 831 (void)kern_close(td, sv[1]); 832 } 833 return (error); 834 } 835 836 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 837 struct linux_send_args { 838 register_t s; 839 register_t msg; 840 register_t len; 841 register_t flags; 842 }; 843 844 static int 845 linux_send(struct thread *td, struct linux_send_args *args) 846 { 847 struct sendto_args /* { 848 int s; 849 caddr_t buf; 850 int len; 851 int flags; 852 caddr_t to; 853 int tolen; 854 } */ bsd_args; 855 struct file *fp; 856 int error, fflag; 857 858 bsd_args.s = args->s; 859 bsd_args.buf = (caddr_t)PTRIN(args->msg); 860 bsd_args.len = args->len; 861 bsd_args.flags = args->flags; 862 bsd_args.to = NULL; 863 bsd_args.tolen = 0; 864 error = sys_sendto(td, &bsd_args); 865 if (error == ENOTCONN) { 866 /* 867 * Linux doesn't return ENOTCONN for non-blocking sockets. 868 * Instead it returns the EAGAIN. 869 */ 870 error = getsock_cap(td, args->s, &cap_send_rights, &fp, 871 &fflag, NULL); 872 if (error == 0) { 873 if (fflag & FNONBLOCK) 874 error = EAGAIN; 875 fdrop(fp, td); 876 } 877 } 878 return (error); 879 } 880 881 struct linux_recv_args { 882 register_t s; 883 register_t msg; 884 register_t len; 885 register_t flags; 886 }; 887 888 static int 889 linux_recv(struct thread *td, struct linux_recv_args *args) 890 { 891 struct recvfrom_args /* { 892 int s; 893 caddr_t buf; 894 int len; 895 int flags; 896 struct sockaddr *from; 897 socklen_t fromlenaddr; 898 } */ bsd_args; 899 900 bsd_args.s = args->s; 901 bsd_args.buf = (caddr_t)PTRIN(args->msg); 902 bsd_args.len = args->len; 903 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 904 bsd_args.from = NULL; 905 bsd_args.fromlenaddr = 0; 906 return (sys_recvfrom(td, &bsd_args)); 907 } 908 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 909 910 int 911 linux_sendto(struct thread *td, struct linux_sendto_args *args) 912 { 913 struct msghdr msg; 914 struct iovec aiov; 915 916 if (linux_check_hdrincl(td, args->s) == 0) 917 /* IP_HDRINCL set, tweak the packet before sending */ 918 return (linux_sendto_hdrincl(td, args)); 919 920 msg.msg_name = PTRIN(args->to); 921 msg.msg_namelen = args->tolen; 922 msg.msg_iov = &aiov; 923 msg.msg_iovlen = 1; 924 msg.msg_control = NULL; 925 msg.msg_flags = 0; 926 aiov.iov_base = PTRIN(args->msg); 927 aiov.iov_len = args->len; 928 return (linux_sendit(td, args->s, &msg, args->flags, NULL, 929 UIO_USERSPACE)); 930 } 931 932 int 933 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 934 { 935 struct sockaddr *sa; 936 struct msghdr msg; 937 struct iovec aiov; 938 int error, fromlen; 939 940 if (PTRIN(args->fromlen) != NULL) { 941 error = copyin(PTRIN(args->fromlen), &fromlen, 942 sizeof(fromlen)); 943 if (error != 0) 944 return (error); 945 if (fromlen < 0) 946 return (EINVAL); 947 sa = malloc(fromlen, M_SONAME, M_WAITOK); 948 } else { 949 fromlen = 0; 950 sa = NULL; 951 } 952 953 msg.msg_name = sa; 954 msg.msg_namelen = fromlen; 955 msg.msg_iov = &aiov; 956 msg.msg_iovlen = 1; 957 aiov.iov_base = PTRIN(args->buf); 958 aiov.iov_len = args->len; 959 msg.msg_control = 0; 960 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 961 962 error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL); 963 if (error != 0) 964 goto out; 965 966 if (PTRIN(args->from) != NULL) 967 error = linux_copyout_sockaddr(sa, PTRIN(args->from), msg.msg_namelen); 968 969 if (error == 0 && PTRIN(args->fromlen) != NULL) 970 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 971 sizeof(msg.msg_namelen)); 972 out: 973 free(sa, M_SONAME); 974 return (error); 975 } 976 977 static int 978 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 979 l_uint flags) 980 { 981 struct cmsghdr *cmsg; 982 struct mbuf *control; 983 struct msghdr msg; 984 struct l_cmsghdr linux_cmsg; 985 struct l_cmsghdr *ptr_cmsg; 986 struct l_msghdr linux_msghdr; 987 struct iovec *iov; 988 socklen_t datalen; 989 struct sockaddr *sa; 990 struct socket *so; 991 sa_family_t sa_family; 992 struct file *fp; 993 void *data; 994 l_size_t len; 995 l_size_t clen; 996 int error, fflag; 997 998 error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr)); 999 if (error != 0) 1000 return (error); 1001 1002 /* 1003 * Some Linux applications (ping) define a non-NULL control data 1004 * pointer, but a msg_controllen of 0, which is not allowed in the 1005 * FreeBSD system call interface. NULL the msg_control pointer in 1006 * order to handle this case. This should be checked, but allows the 1007 * Linux ping to work. 1008 */ 1009 if (PTRIN(linux_msghdr.msg_control) != NULL && 1010 linux_msghdr.msg_controllen == 0) 1011 linux_msghdr.msg_control = PTROUT(NULL); 1012 1013 error = linux_to_bsd_msghdr(&msg, &linux_msghdr); 1014 if (error != 0) 1015 return (error); 1016 1017 #ifdef COMPAT_LINUX32 1018 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1019 &iov, EMSGSIZE); 1020 #else 1021 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1022 #endif 1023 if (error != 0) 1024 return (error); 1025 1026 control = NULL; 1027 1028 error = kern_getsockname(td, s, &sa, &datalen); 1029 if (error != 0) 1030 goto bad; 1031 sa_family = sa->sa_family; 1032 free(sa, M_SONAME); 1033 1034 if (flags & LINUX_MSG_OOB) { 1035 error = EOPNOTSUPP; 1036 if (sa_family == AF_UNIX) 1037 goto bad; 1038 1039 error = getsock_cap(td, s, &cap_send_rights, &fp, 1040 &fflag, NULL); 1041 if (error != 0) 1042 goto bad; 1043 so = fp->f_data; 1044 if (so->so_type != SOCK_STREAM) 1045 error = EOPNOTSUPP; 1046 fdrop(fp, td); 1047 if (error != 0) 1048 goto bad; 1049 } 1050 1051 if (linux_msghdr.msg_controllen >= sizeof(struct l_cmsghdr)) { 1052 error = ENOBUFS; 1053 control = m_get(M_WAITOK, MT_CONTROL); 1054 MCLGET(control, M_WAITOK); 1055 data = mtod(control, void *); 1056 datalen = 0; 1057 1058 ptr_cmsg = PTRIN(linux_msghdr.msg_control); 1059 clen = linux_msghdr.msg_controllen; 1060 do { 1061 error = copyin(ptr_cmsg, &linux_cmsg, 1062 sizeof(struct l_cmsghdr)); 1063 if (error != 0) 1064 goto bad; 1065 1066 error = EINVAL; 1067 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) || 1068 linux_cmsg.cmsg_len > clen) 1069 goto bad; 1070 1071 if (datalen + CMSG_HDRSZ > MCLBYTES) 1072 goto bad; 1073 1074 /* 1075 * Now we support only SCM_RIGHTS and SCM_CRED, 1076 * so return EINVAL in any other cmsg_type 1077 */ 1078 cmsg = data; 1079 cmsg->cmsg_type = 1080 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1081 cmsg->cmsg_level = 1082 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1083 if (cmsg->cmsg_type == -1 1084 || cmsg->cmsg_level != SOL_SOCKET) { 1085 linux_msg(curthread, 1086 "unsupported sendmsg cmsg level %d type %d", 1087 linux_cmsg.cmsg_level, linux_cmsg.cmsg_type); 1088 goto bad; 1089 } 1090 1091 /* 1092 * Some applications (e.g. pulseaudio) attempt to 1093 * send ancillary data even if the underlying protocol 1094 * doesn't support it which is not allowed in the 1095 * FreeBSD system call interface. 1096 */ 1097 if (sa_family != AF_UNIX) 1098 goto next; 1099 1100 if (cmsg->cmsg_type == SCM_CREDS) { 1101 len = sizeof(struct cmsgcred); 1102 if (datalen + CMSG_SPACE(len) > MCLBYTES) 1103 goto bad; 1104 1105 /* 1106 * The lower levels will fill in the structure 1107 */ 1108 memset(CMSG_DATA(data), 0, len); 1109 } else { 1110 len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1111 if (datalen + CMSG_SPACE(len) < datalen || 1112 datalen + CMSG_SPACE(len) > MCLBYTES) 1113 goto bad; 1114 1115 error = copyin(LINUX_CMSG_DATA(ptr_cmsg), 1116 CMSG_DATA(data), len); 1117 if (error != 0) 1118 goto bad; 1119 } 1120 1121 cmsg->cmsg_len = CMSG_LEN(len); 1122 data = (char *)data + CMSG_SPACE(len); 1123 datalen += CMSG_SPACE(len); 1124 1125 next: 1126 if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len)) 1127 break; 1128 1129 clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len); 1130 ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg + 1131 LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len)); 1132 } while(clen >= sizeof(struct l_cmsghdr)); 1133 1134 control->m_len = datalen; 1135 if (datalen == 0) { 1136 m_freem(control); 1137 control = NULL; 1138 } 1139 } 1140 1141 msg.msg_iov = iov; 1142 msg.msg_flags = 0; 1143 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1144 control = NULL; 1145 1146 bad: 1147 m_freem(control); 1148 free(iov, M_IOV); 1149 return (error); 1150 } 1151 1152 int 1153 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1154 { 1155 1156 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1157 args->flags)); 1158 } 1159 1160 int 1161 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1162 { 1163 struct l_mmsghdr *msg; 1164 l_uint retval; 1165 int error, datagrams; 1166 1167 if (args->vlen > UIO_MAXIOV) 1168 args->vlen = UIO_MAXIOV; 1169 1170 msg = PTRIN(args->msg); 1171 datagrams = 0; 1172 while (datagrams < args->vlen) { 1173 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1174 args->flags); 1175 if (error != 0) 1176 break; 1177 1178 retval = td->td_retval[0]; 1179 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1180 if (error != 0) 1181 break; 1182 ++msg; 1183 ++datagrams; 1184 } 1185 if (error == 0) 1186 td->td_retval[0] = datagrams; 1187 return (error); 1188 } 1189 1190 static int 1191 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1192 l_uint flags, struct msghdr *msg) 1193 { 1194 struct cmsghdr *cm; 1195 struct cmsgcred *cmcred; 1196 struct l_cmsghdr *linux_cmsg = NULL; 1197 struct l_ucred linux_ucred; 1198 socklen_t datalen, maxlen, outlen; 1199 struct l_msghdr linux_msghdr; 1200 struct iovec *iov, *uiov; 1201 struct mbuf *control = NULL; 1202 struct mbuf **controlp; 1203 struct timeval *ftmvl; 1204 struct sockaddr *sa; 1205 l_timeval ltmvl; 1206 caddr_t outbuf; 1207 void *data; 1208 int error, i, fd, fds, *fdp; 1209 1210 error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr)); 1211 if (error != 0) 1212 return (error); 1213 1214 error = linux_to_bsd_msghdr(msg, &linux_msghdr); 1215 if (error != 0) 1216 return (error); 1217 1218 #ifdef COMPAT_LINUX32 1219 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1220 &iov, EMSGSIZE); 1221 #else 1222 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1223 #endif 1224 if (error != 0) 1225 return (error); 1226 1227 if (msg->msg_name != NULL && msg->msg_namelen > 0) { 1228 msg->msg_namelen = min(msg->msg_namelen, SOCK_MAXADDRLEN); 1229 sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK); 1230 msg->msg_name = sa; 1231 } else { 1232 sa = NULL; 1233 msg->msg_name = NULL; 1234 } 1235 1236 uiov = msg->msg_iov; 1237 msg->msg_iov = iov; 1238 controlp = (msg->msg_control != NULL) ? &control : NULL; 1239 error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp); 1240 msg->msg_iov = uiov; 1241 if (error != 0) 1242 goto bad; 1243 1244 /* 1245 * Note that kern_recvit() updates msg->msg_namelen. 1246 */ 1247 if (msg->msg_name != NULL && msg->msg_namelen > 0) { 1248 msg->msg_name = PTRIN(linux_msghdr.msg_name); 1249 error = linux_copyout_sockaddr(sa, 1250 PTRIN(msg->msg_name), msg->msg_namelen); 1251 if (error != 0) 1252 goto bad; 1253 } 1254 1255 error = bsd_to_linux_msghdr(msg, &linux_msghdr); 1256 if (error != 0) 1257 goto bad; 1258 1259 maxlen = linux_msghdr.msg_controllen; 1260 linux_msghdr.msg_controllen = 0; 1261 if (control) { 1262 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1263 1264 msg->msg_control = mtod(control, struct cmsghdr *); 1265 msg->msg_controllen = control->m_len; 1266 1267 cm = CMSG_FIRSTHDR(msg); 1268 outbuf = PTRIN(linux_msghdr.msg_control); 1269 outlen = 0; 1270 while (cm != NULL) { 1271 linux_cmsg->cmsg_type = 1272 bsd_to_linux_cmsg_type(cm->cmsg_type); 1273 linux_cmsg->cmsg_level = 1274 bsd_to_linux_sockopt_level(cm->cmsg_level); 1275 if (linux_cmsg->cmsg_type == -1 || 1276 cm->cmsg_level != SOL_SOCKET) { 1277 linux_msg(curthread, 1278 "unsupported recvmsg cmsg level %d type %d", 1279 cm->cmsg_level, cm->cmsg_type); 1280 error = EINVAL; 1281 goto bad; 1282 } 1283 1284 data = CMSG_DATA(cm); 1285 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1286 1287 switch (cm->cmsg_type) { 1288 case SCM_RIGHTS: 1289 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1290 fds = datalen / sizeof(int); 1291 fdp = data; 1292 for (i = 0; i < fds; i++) { 1293 fd = *fdp++; 1294 (void)kern_fcntl(td, fd, 1295 F_SETFD, FD_CLOEXEC); 1296 } 1297 } 1298 break; 1299 1300 case SCM_CREDS: 1301 /* 1302 * Currently LOCAL_CREDS is never in 1303 * effect for Linux so no need to worry 1304 * about sockcred 1305 */ 1306 if (datalen != sizeof(*cmcred)) { 1307 error = EMSGSIZE; 1308 goto bad; 1309 } 1310 cmcred = (struct cmsgcred *)data; 1311 bzero(&linux_ucred, sizeof(linux_ucred)); 1312 linux_ucred.pid = cmcred->cmcred_pid; 1313 linux_ucred.uid = cmcred->cmcred_uid; 1314 linux_ucred.gid = cmcred->cmcred_gid; 1315 data = &linux_ucred; 1316 datalen = sizeof(linux_ucred); 1317 break; 1318 1319 case SCM_TIMESTAMP: 1320 if (datalen != sizeof(struct timeval)) { 1321 error = EMSGSIZE; 1322 goto bad; 1323 } 1324 ftmvl = (struct timeval *)data; 1325 ltmvl.tv_sec = ftmvl->tv_sec; 1326 ltmvl.tv_usec = ftmvl->tv_usec; 1327 data = <mvl; 1328 datalen = sizeof(ltmvl); 1329 break; 1330 } 1331 1332 if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) { 1333 if (outlen == 0) { 1334 error = EMSGSIZE; 1335 goto bad; 1336 } else { 1337 linux_msghdr.msg_flags |= LINUX_MSG_CTRUNC; 1338 m_dispose_extcontrolm(control); 1339 goto out; 1340 } 1341 } 1342 1343 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1344 1345 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1346 if (error != 0) 1347 goto bad; 1348 outbuf += L_CMSG_HDRSZ; 1349 1350 error = copyout(data, outbuf, datalen); 1351 if (error != 0) 1352 goto bad; 1353 1354 outbuf += LINUX_CMSG_ALIGN(datalen); 1355 outlen += LINUX_CMSG_LEN(datalen); 1356 1357 cm = CMSG_NXTHDR(msg, cm); 1358 } 1359 linux_msghdr.msg_controllen = outlen; 1360 } 1361 1362 out: 1363 error = copyout(&linux_msghdr, msghdr, sizeof(linux_msghdr)); 1364 1365 bad: 1366 if (control != NULL) { 1367 if (error != 0) 1368 m_dispose_extcontrolm(control); 1369 m_freem(control); 1370 } 1371 free(iov, M_IOV); 1372 free(linux_cmsg, M_LINUX); 1373 free(sa, M_SONAME); 1374 1375 return (error); 1376 } 1377 1378 int 1379 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1380 { 1381 struct msghdr bsd_msg; 1382 1383 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1384 args->flags, &bsd_msg)); 1385 } 1386 1387 int 1388 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1389 { 1390 struct l_mmsghdr *msg; 1391 struct msghdr bsd_msg; 1392 struct l_timespec lts; 1393 struct timespec ts, tts; 1394 l_uint retval; 1395 int error, datagrams; 1396 1397 if (args->timeout) { 1398 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1399 if (error != 0) 1400 return (error); 1401 error = linux_to_native_timespec(&ts, <s); 1402 if (error != 0) 1403 return (error); 1404 getnanotime(&tts); 1405 timespecadd(&tts, &ts, &tts); 1406 } 1407 1408 msg = PTRIN(args->msg); 1409 datagrams = 0; 1410 while (datagrams < args->vlen) { 1411 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1412 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1413 if (error != 0) 1414 break; 1415 1416 retval = td->td_retval[0]; 1417 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1418 if (error != 0) 1419 break; 1420 ++msg; 1421 ++datagrams; 1422 1423 /* 1424 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1425 */ 1426 if (args->flags & LINUX_MSG_WAITFORONE) 1427 args->flags |= LINUX_MSG_DONTWAIT; 1428 1429 /* 1430 * See BUGS section of recvmmsg(2). 1431 */ 1432 if (args->timeout) { 1433 getnanotime(&ts); 1434 timespecsub(&ts, &tts, &ts); 1435 if (!timespecisset(&ts) || ts.tv_sec > 0) 1436 break; 1437 } 1438 /* Out of band data, return right away. */ 1439 if (bsd_msg.msg_flags & MSG_OOB) 1440 break; 1441 } 1442 if (error == 0) 1443 td->td_retval[0] = datagrams; 1444 return (error); 1445 } 1446 1447 int 1448 linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1449 { 1450 1451 return (kern_shutdown(td, args->s, args->how)); 1452 } 1453 1454 int 1455 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1456 { 1457 l_timeval linux_tv; 1458 struct sockaddr *sa; 1459 struct timeval tv; 1460 socklen_t len; 1461 int error, level, name; 1462 1463 level = linux_to_bsd_sockopt_level(args->level); 1464 switch (level) { 1465 case SOL_SOCKET: 1466 name = linux_to_bsd_so_sockopt(args->optname); 1467 switch (name) { 1468 case LOCAL_CREDS_PERSISTENT: 1469 level = SOL_LOCAL; 1470 break; 1471 case SO_RCVTIMEO: 1472 /* FALLTHROUGH */ 1473 case SO_SNDTIMEO: 1474 error = copyin(PTRIN(args->optval), &linux_tv, 1475 sizeof(linux_tv)); 1476 if (error != 0) 1477 return (error); 1478 tv.tv_sec = linux_tv.tv_sec; 1479 tv.tv_usec = linux_tv.tv_usec; 1480 return (kern_setsockopt(td, args->s, level, 1481 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1482 /* NOTREACHED */ 1483 default: 1484 break; 1485 } 1486 break; 1487 case IPPROTO_IP: 1488 if (args->optname == LINUX_IP_RECVERR && 1489 linux_ignore_ip_recverr) { 1490 /* 1491 * XXX: This is a hack to unbreak DNS resolution 1492 * with glibc 2.30 and above. 1493 */ 1494 return (0); 1495 } 1496 name = linux_to_bsd_ip_sockopt(args->optname); 1497 break; 1498 case IPPROTO_IPV6: 1499 name = linux_to_bsd_ip6_sockopt(args->optname); 1500 break; 1501 case IPPROTO_TCP: 1502 name = linux_to_bsd_tcp_sockopt(args->optname); 1503 break; 1504 default: 1505 name = -1; 1506 break; 1507 } 1508 if (name == -1) { 1509 linux_msg(curthread, 1510 "unsupported setsockopt level %d optname %d", 1511 args->level, args->optname); 1512 return (ENOPROTOOPT); 1513 } 1514 1515 if (name == IPV6_NEXTHOP) { 1516 len = args->optlen; 1517 error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len); 1518 if (error != 0) 1519 return (error); 1520 1521 error = kern_setsockopt(td, args->s, level, 1522 name, sa, UIO_SYSSPACE, len); 1523 free(sa, M_SONAME); 1524 } else { 1525 error = kern_setsockopt(td, args->s, level, 1526 name, PTRIN(args->optval), UIO_USERSPACE, args->optlen); 1527 } 1528 1529 return (error); 1530 } 1531 1532 int 1533 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1534 { 1535 l_timeval linux_tv; 1536 struct timeval tv; 1537 socklen_t tv_len, xulen, len; 1538 struct sockaddr *sa; 1539 struct xucred xu; 1540 struct l_ucred lxu; 1541 int error, level, name, newval; 1542 1543 level = linux_to_bsd_sockopt_level(args->level); 1544 switch (level) { 1545 case SOL_SOCKET: 1546 name = linux_to_bsd_so_sockopt(args->optname); 1547 switch (name) { 1548 case LOCAL_CREDS_PERSISTENT: 1549 level = SOL_LOCAL; 1550 break; 1551 case SO_RCVTIMEO: 1552 /* FALLTHROUGH */ 1553 case SO_SNDTIMEO: 1554 tv_len = sizeof(tv); 1555 error = kern_getsockopt(td, args->s, level, 1556 name, &tv, UIO_SYSSPACE, &tv_len); 1557 if (error != 0) 1558 return (error); 1559 linux_tv.tv_sec = tv.tv_sec; 1560 linux_tv.tv_usec = tv.tv_usec; 1561 return (copyout(&linux_tv, PTRIN(args->optval), 1562 sizeof(linux_tv))); 1563 /* NOTREACHED */ 1564 case LOCAL_PEERCRED: 1565 if (args->optlen < sizeof(lxu)) 1566 return (EINVAL); 1567 /* 1568 * LOCAL_PEERCRED is not served at the SOL_SOCKET level, 1569 * but by the Unix socket's level 0. 1570 */ 1571 level = 0; 1572 xulen = sizeof(xu); 1573 error = kern_getsockopt(td, args->s, level, 1574 name, &xu, UIO_SYSSPACE, &xulen); 1575 if (error != 0) 1576 return (error); 1577 lxu.pid = xu.cr_pid; 1578 lxu.uid = xu.cr_uid; 1579 lxu.gid = xu.cr_gid; 1580 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1581 /* NOTREACHED */ 1582 case SO_ERROR: 1583 len = sizeof(newval); 1584 error = kern_getsockopt(td, args->s, level, 1585 name, &newval, UIO_SYSSPACE, &len); 1586 if (error != 0) 1587 return (error); 1588 newval = -bsd_to_linux_errno(newval); 1589 return (copyout(&newval, PTRIN(args->optval), len)); 1590 /* NOTREACHED */ 1591 default: 1592 break; 1593 } 1594 break; 1595 case IPPROTO_IP: 1596 name = linux_to_bsd_ip_sockopt(args->optname); 1597 break; 1598 case IPPROTO_IPV6: 1599 name = linux_to_bsd_ip6_sockopt(args->optname); 1600 break; 1601 case IPPROTO_TCP: 1602 name = linux_to_bsd_tcp_sockopt(args->optname); 1603 break; 1604 default: 1605 name = -1; 1606 break; 1607 } 1608 if (name == -1) { 1609 linux_msg(curthread, 1610 "unsupported getsockopt level %d optname %d", 1611 args->level, args->optname); 1612 return (EINVAL); 1613 } 1614 1615 if (name == IPV6_NEXTHOP) { 1616 error = copyin(PTRIN(args->optlen), &len, sizeof(len)); 1617 if (error != 0) 1618 return (error); 1619 sa = malloc(len, M_SONAME, M_WAITOK); 1620 1621 error = kern_getsockopt(td, args->s, level, 1622 name, sa, UIO_SYSSPACE, &len); 1623 if (error != 0) 1624 goto out; 1625 1626 error = linux_copyout_sockaddr(sa, PTRIN(args->optval), len); 1627 if (error == 0) 1628 error = copyout(&len, PTRIN(args->optlen), 1629 sizeof(len)); 1630 out: 1631 free(sa, M_SONAME); 1632 } else { 1633 if (args->optval) { 1634 error = copyin(PTRIN(args->optlen), &len, sizeof(len)); 1635 if (error != 0) 1636 return (error); 1637 } 1638 error = kern_getsockopt(td, args->s, level, 1639 name, PTRIN(args->optval), UIO_USERSPACE, &len); 1640 if (error == 0) 1641 error = copyout(&len, PTRIN(args->optlen), 1642 sizeof(len)); 1643 } 1644 1645 return (error); 1646 } 1647 1648 static int 1649 linux_sendfile_common(struct thread *td, l_int out, l_int in, 1650 l_loff_t *offset, l_size_t count) 1651 { 1652 off_t bytes_read; 1653 int error; 1654 l_loff_t current_offset; 1655 struct file *fp; 1656 1657 AUDIT_ARG_FD(in); 1658 error = fget_read(td, in, &cap_pread_rights, &fp); 1659 if (error != 0) 1660 return (error); 1661 1662 if (offset != NULL) { 1663 current_offset = *offset; 1664 } else { 1665 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ? 1666 fo_seek(fp, 0, SEEK_CUR, td) : ESPIPE; 1667 if (error != 0) 1668 goto drop; 1669 current_offset = td->td_uretoff.tdu_off; 1670 } 1671 1672 bytes_read = 0; 1673 1674 /* Linux cannot have 0 count. */ 1675 if (count <= 0 || current_offset < 0) { 1676 error = EINVAL; 1677 goto drop; 1678 } 1679 1680 error = fo_sendfile(fp, out, NULL, NULL, current_offset, count, 1681 &bytes_read, 0, td); 1682 if (error != 0) 1683 goto drop; 1684 current_offset += bytes_read; 1685 1686 if (offset != NULL) { 1687 *offset = current_offset; 1688 } else { 1689 error = fo_seek(fp, current_offset, SEEK_SET, td); 1690 if (error != 0) 1691 goto drop; 1692 } 1693 1694 td->td_retval[0] = (ssize_t)bytes_read; 1695 drop: 1696 fdrop(fp, td); 1697 return (error); 1698 } 1699 1700 int 1701 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg) 1702 { 1703 /* 1704 * Differences between FreeBSD and Linux sendfile: 1705 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to 1706 * mean send the whole file.) In linux_sendfile given fds are still 1707 * checked for validity when the count is 0. 1708 * - Linux can send to any fd whereas FreeBSD only supports sockets. 1709 * The same restriction follows for linux_sendfile. 1710 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr. 1711 * - Linux takes an offset pointer and updates it to the read location. 1712 * FreeBSD takes in an offset and a 'bytes read' parameter which is 1713 * only filled if it isn't NULL. We use this parameter to update the 1714 * offset pointer if it exists. 1715 * - Linux sendfile returns bytes read on success while FreeBSD 1716 * returns 0. We use the 'bytes read' parameter to get this value. 1717 */ 1718 1719 l_loff_t offset64; 1720 l_long offset; 1721 int ret; 1722 int error; 1723 1724 if (arg->offset != NULL) { 1725 error = copyin(arg->offset, &offset, sizeof(offset)); 1726 if (error != 0) 1727 return (error); 1728 offset64 = (l_loff_t)offset; 1729 } 1730 1731 ret = linux_sendfile_common(td, arg->out, arg->in, 1732 arg->offset != NULL ? &offset64 : NULL, arg->count); 1733 1734 if (arg->offset != NULL) { 1735 #if defined(__i386__) || defined(__arm__) || \ 1736 (defined(__amd64__) && defined(COMPAT_LINUX32)) 1737 if (offset64 > INT32_MAX) 1738 return (EOVERFLOW); 1739 #endif 1740 offset = (l_long)offset64; 1741 error = copyout(&offset, arg->offset, sizeof(offset)); 1742 if (error != 0) 1743 return (error); 1744 } 1745 1746 return (ret); 1747 } 1748 1749 #if defined(__i386__) || defined(__arm__) || \ 1750 (defined(__amd64__) && defined(COMPAT_LINUX32)) 1751 1752 int 1753 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg) 1754 { 1755 l_loff_t offset; 1756 int ret; 1757 int error; 1758 1759 if (arg->offset != NULL) { 1760 error = copyin(arg->offset, &offset, sizeof(offset)); 1761 if (error != 0) 1762 return (error); 1763 } 1764 1765 ret = linux_sendfile_common(td, arg->out, arg->in, 1766 arg->offset != NULL ? &offset : NULL, arg->count); 1767 1768 if (arg->offset != NULL) { 1769 error = copyout(&offset, arg->offset, sizeof(offset)); 1770 if (error != 0) 1771 return (error); 1772 } 1773 1774 return (ret); 1775 } 1776 1777 /* Argument list sizes for linux_socketcall */ 1778 static const unsigned char lxs_args_cnt[] = { 1779 0 /* unused*/, 3 /* socket */, 1780 3 /* bind */, 3 /* connect */, 1781 2 /* listen */, 3 /* accept */, 1782 3 /* getsockname */, 3 /* getpeername */, 1783 4 /* socketpair */, 4 /* send */, 1784 4 /* recv */, 6 /* sendto */, 1785 6 /* recvfrom */, 2 /* shutdown */, 1786 5 /* setsockopt */, 5 /* getsockopt */, 1787 3 /* sendmsg */, 3 /* recvmsg */, 1788 4 /* accept4 */, 5 /* recvmmsg */, 1789 4 /* sendmmsg */, 4 /* sendfile */ 1790 }; 1791 #define LINUX_ARGS_CNT (nitems(lxs_args_cnt) - 1) 1792 #define LINUX_ARG_SIZE(x) (lxs_args_cnt[x] * sizeof(l_ulong)) 1793 1794 int 1795 linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1796 { 1797 l_ulong a[6]; 1798 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1799 register_t l_args[6]; 1800 #endif 1801 void *arg; 1802 int error; 1803 1804 if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT) 1805 return (EINVAL); 1806 error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what)); 1807 if (error != 0) 1808 return (error); 1809 1810 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1811 for (int i = 0; i < lxs_args_cnt[args->what]; ++i) 1812 l_args[i] = a[i]; 1813 arg = l_args; 1814 #else 1815 arg = a; 1816 #endif 1817 switch (args->what) { 1818 case LINUX_SOCKET: 1819 return (linux_socket(td, arg)); 1820 case LINUX_BIND: 1821 return (linux_bind(td, arg)); 1822 case LINUX_CONNECT: 1823 return (linux_connect(td, arg)); 1824 case LINUX_LISTEN: 1825 return (linux_listen(td, arg)); 1826 case LINUX_ACCEPT: 1827 return (linux_accept(td, arg)); 1828 case LINUX_GETSOCKNAME: 1829 return (linux_getsockname(td, arg)); 1830 case LINUX_GETPEERNAME: 1831 return (linux_getpeername(td, arg)); 1832 case LINUX_SOCKETPAIR: 1833 return (linux_socketpair(td, arg)); 1834 case LINUX_SEND: 1835 return (linux_send(td, arg)); 1836 case LINUX_RECV: 1837 return (linux_recv(td, arg)); 1838 case LINUX_SENDTO: 1839 return (linux_sendto(td, arg)); 1840 case LINUX_RECVFROM: 1841 return (linux_recvfrom(td, arg)); 1842 case LINUX_SHUTDOWN: 1843 return (linux_shutdown(td, arg)); 1844 case LINUX_SETSOCKOPT: 1845 return (linux_setsockopt(td, arg)); 1846 case LINUX_GETSOCKOPT: 1847 return (linux_getsockopt(td, arg)); 1848 case LINUX_SENDMSG: 1849 return (linux_sendmsg(td, arg)); 1850 case LINUX_RECVMSG: 1851 return (linux_recvmsg(td, arg)); 1852 case LINUX_ACCEPT4: 1853 return (linux_accept4(td, arg)); 1854 case LINUX_RECVMMSG: 1855 return (linux_recvmmsg(td, arg)); 1856 case LINUX_SENDMMSG: 1857 return (linux_sendmmsg(td, arg)); 1858 case LINUX_SENDFILE: 1859 return (linux_sendfile(td, arg)); 1860 } 1861 1862 linux_msg(td, "socket type %d not implemented", args->what); 1863 return (ENOSYS); 1864 } 1865 #endif /* __i386__ || __arm__ || (__amd64__ && COMPAT_LINUX32) */ 1866