1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 1995 Søren Schmidt 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 /* XXX we use functions that might not exist. */ 33 #include "opt_compat.h" 34 #include "opt_inet6.h" 35 36 #include <sys/param.h> 37 #include <sys/proc.h> 38 #include <sys/systm.h> 39 #include <sys/sysproto.h> 40 #include <sys/capsicum.h> 41 #include <sys/fcntl.h> 42 #include <sys/file.h> 43 #include <sys/limits.h> 44 #include <sys/lock.h> 45 #include <sys/malloc.h> 46 #include <sys/mutex.h> 47 #include <sys/mbuf.h> 48 #include <sys/socket.h> 49 #include <sys/socketvar.h> 50 #include <sys/syscallsubr.h> 51 #include <sys/uio.h> 52 #include <sys/syslog.h> 53 #include <sys/un.h> 54 55 #include <net/if.h> 56 #include <net/vnet.h> 57 #include <netinet/in.h> 58 #include <netinet/in_systm.h> 59 #include <netinet/ip.h> 60 #include <netinet/tcp.h> 61 #ifdef INET6 62 #include <netinet/ip6.h> 63 #include <netinet6/ip6_var.h> 64 #endif 65 66 #ifdef COMPAT_LINUX32 67 #include <machine/../linux32/linux.h> 68 #include <machine/../linux32/linux32_proto.h> 69 #else 70 #include <machine/../linux/linux.h> 71 #include <machine/../linux/linux_proto.h> 72 #endif 73 #include <compat/linux/linux_file.h> 74 #include <compat/linux/linux_socket.h> 75 #include <compat/linux/linux_timer.h> 76 #include <compat/linux/linux_util.h> 77 78 static int linux_to_bsd_domain(int); 79 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *, 80 l_uint); 81 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *, 82 l_uint, struct msghdr *); 83 static int linux_set_socket_flags(int, int *); 84 85 /* 86 * Reads a Linux sockaddr and does any necessary translation. 87 * Linux sockaddrs don't have a length field, only a family. 88 * Copy the osockaddr structure pointed to by osa to kernel, adjust 89 * family and convert to sockaddr. 90 */ 91 static int 92 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen) 93 { 94 struct sockaddr *sa; 95 struct osockaddr *kosa; 96 #ifdef INET6 97 struct sockaddr_in6 *sin6; 98 int oldv6size; 99 #endif 100 char *name; 101 int bdom, error, hdrlen, namelen; 102 103 if (salen < 2 || salen > UCHAR_MAX || !osa) 104 return (EINVAL); 105 106 #ifdef INET6 107 oldv6size = 0; 108 /* 109 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it 110 * if it's a v4-mapped address, so reserve the proper space 111 * for it. 112 */ 113 if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) { 114 salen += sizeof(uint32_t); 115 oldv6size = 1; 116 } 117 #endif 118 119 kosa = malloc(salen, M_SONAME, M_WAITOK); 120 121 if ((error = copyin(osa, kosa, salen))) 122 goto out; 123 124 bdom = linux_to_bsd_domain(kosa->sa_family); 125 if (bdom == -1) { 126 error = EAFNOSUPPORT; 127 goto out; 128 } 129 130 #ifdef INET6 131 /* 132 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6, 133 * which lacks the scope id compared with RFC2553 one. If we detect 134 * the situation, reject the address and write a message to system log. 135 * 136 * Still accept addresses for which the scope id is not used. 137 */ 138 if (oldv6size) { 139 if (bdom == AF_INET6) { 140 sin6 = (struct sockaddr_in6 *)kosa; 141 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) || 142 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) && 143 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) && 144 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) && 145 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) && 146 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) { 147 sin6->sin6_scope_id = 0; 148 } else { 149 log(LOG_DEBUG, 150 "obsolete pre-RFC2553 sockaddr_in6 rejected\n"); 151 error = EINVAL; 152 goto out; 153 } 154 } else 155 salen -= sizeof(uint32_t); 156 } 157 #endif 158 if (bdom == AF_INET) { 159 if (salen < sizeof(struct sockaddr_in)) { 160 error = EINVAL; 161 goto out; 162 } 163 salen = sizeof(struct sockaddr_in); 164 } 165 166 if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) { 167 hdrlen = offsetof(struct sockaddr_un, sun_path); 168 name = ((struct sockaddr_un *)kosa)->sun_path; 169 if (*name == '\0') { 170 /* 171 * Linux abstract namespace starts with a NULL byte. 172 * XXX We do not support abstract namespace yet. 173 */ 174 namelen = strnlen(name + 1, salen - hdrlen - 1) + 1; 175 } else 176 namelen = strnlen(name, salen - hdrlen); 177 salen = hdrlen + namelen; 178 if (salen > sizeof(struct sockaddr_un)) { 179 error = ENAMETOOLONG; 180 goto out; 181 } 182 } 183 184 sa = (struct sockaddr *)kosa; 185 sa->sa_family = bdom; 186 sa->sa_len = salen; 187 188 *sap = sa; 189 return (0); 190 191 out: 192 free(kosa, M_SONAME); 193 return (error); 194 } 195 196 static int 197 linux_to_bsd_domain(int domain) 198 { 199 200 switch (domain) { 201 case LINUX_AF_UNSPEC: 202 return (AF_UNSPEC); 203 case LINUX_AF_UNIX: 204 return (AF_LOCAL); 205 case LINUX_AF_INET: 206 return (AF_INET); 207 case LINUX_AF_INET6: 208 return (AF_INET6); 209 case LINUX_AF_AX25: 210 return (AF_CCITT); 211 case LINUX_AF_IPX: 212 return (AF_IPX); 213 case LINUX_AF_APPLETALK: 214 return (AF_APPLETALK); 215 } 216 return (-1); 217 } 218 219 static int 220 bsd_to_linux_domain(int domain) 221 { 222 223 switch (domain) { 224 case AF_UNSPEC: 225 return (LINUX_AF_UNSPEC); 226 case AF_LOCAL: 227 return (LINUX_AF_UNIX); 228 case AF_INET: 229 return (LINUX_AF_INET); 230 case AF_INET6: 231 return (LINUX_AF_INET6); 232 case AF_CCITT: 233 return (LINUX_AF_AX25); 234 case AF_IPX: 235 return (LINUX_AF_IPX); 236 case AF_APPLETALK: 237 return (LINUX_AF_APPLETALK); 238 } 239 return (-1); 240 } 241 242 static int 243 linux_to_bsd_sockopt_level(int level) 244 { 245 246 switch (level) { 247 case LINUX_SOL_SOCKET: 248 return (SOL_SOCKET); 249 } 250 return (level); 251 } 252 253 static int 254 bsd_to_linux_sockopt_level(int level) 255 { 256 257 switch (level) { 258 case SOL_SOCKET: 259 return (LINUX_SOL_SOCKET); 260 } 261 return (level); 262 } 263 264 static int 265 linux_to_bsd_ip_sockopt(int opt) 266 { 267 268 switch (opt) { 269 case LINUX_IP_TOS: 270 return (IP_TOS); 271 case LINUX_IP_TTL: 272 return (IP_TTL); 273 case LINUX_IP_OPTIONS: 274 return (IP_OPTIONS); 275 case LINUX_IP_MULTICAST_IF: 276 return (IP_MULTICAST_IF); 277 case LINUX_IP_MULTICAST_TTL: 278 return (IP_MULTICAST_TTL); 279 case LINUX_IP_MULTICAST_LOOP: 280 return (IP_MULTICAST_LOOP); 281 case LINUX_IP_ADD_MEMBERSHIP: 282 return (IP_ADD_MEMBERSHIP); 283 case LINUX_IP_DROP_MEMBERSHIP: 284 return (IP_DROP_MEMBERSHIP); 285 case LINUX_IP_HDRINCL: 286 return (IP_HDRINCL); 287 } 288 return (-1); 289 } 290 291 static int 292 linux_to_bsd_ip6_sockopt(int opt) 293 { 294 295 switch (opt) { 296 case LINUX_IPV6_NEXTHOP: 297 return (IPV6_NEXTHOP); 298 case LINUX_IPV6_UNICAST_HOPS: 299 return (IPV6_UNICAST_HOPS); 300 case LINUX_IPV6_MULTICAST_IF: 301 return (IPV6_MULTICAST_IF); 302 case LINUX_IPV6_MULTICAST_HOPS: 303 return (IPV6_MULTICAST_HOPS); 304 case LINUX_IPV6_MULTICAST_LOOP: 305 return (IPV6_MULTICAST_LOOP); 306 case LINUX_IPV6_ADD_MEMBERSHIP: 307 return (IPV6_JOIN_GROUP); 308 case LINUX_IPV6_DROP_MEMBERSHIP: 309 return (IPV6_LEAVE_GROUP); 310 case LINUX_IPV6_V6ONLY: 311 return (IPV6_V6ONLY); 312 case LINUX_IPV6_DONTFRAG: 313 return (IPV6_DONTFRAG); 314 #if 0 315 case LINUX_IPV6_CHECKSUM: 316 return (IPV6_CHECKSUM); 317 case LINUX_IPV6_RECVPKTINFO: 318 return (IPV6_RECVPKTINFO); 319 case LINUX_IPV6_PKTINFO: 320 return (IPV6_PKTINFO); 321 case LINUX_IPV6_RECVHOPLIMIT: 322 return (IPV6_RECVHOPLIMIT); 323 case LINUX_IPV6_HOPLIMIT: 324 return (IPV6_HOPLIMIT); 325 case LINUX_IPV6_RECVHOPOPTS: 326 return (IPV6_RECVHOPOPTS); 327 case LINUX_IPV6_HOPOPTS: 328 return (IPV6_HOPOPTS); 329 case LINUX_IPV6_RTHDRDSTOPTS: 330 return (IPV6_RTHDRDSTOPTS); 331 case LINUX_IPV6_RECVRTHDR: 332 return (IPV6_RECVRTHDR); 333 case LINUX_IPV6_RTHDR: 334 return (IPV6_RTHDR); 335 case LINUX_IPV6_RECVDSTOPTS: 336 return (IPV6_RECVDSTOPTS); 337 case LINUX_IPV6_DSTOPTS: 338 return (IPV6_DSTOPTS); 339 case LINUX_IPV6_RECVPATHMTU: 340 return (IPV6_RECVPATHMTU); 341 case LINUX_IPV6_PATHMTU: 342 return (IPV6_PATHMTU); 343 #endif 344 } 345 return (-1); 346 } 347 348 static int 349 linux_to_bsd_so_sockopt(int opt) 350 { 351 352 switch (opt) { 353 case LINUX_SO_DEBUG: 354 return (SO_DEBUG); 355 case LINUX_SO_REUSEADDR: 356 return (SO_REUSEADDR); 357 case LINUX_SO_TYPE: 358 return (SO_TYPE); 359 case LINUX_SO_ERROR: 360 return (SO_ERROR); 361 case LINUX_SO_DONTROUTE: 362 return (SO_DONTROUTE); 363 case LINUX_SO_BROADCAST: 364 return (SO_BROADCAST); 365 case LINUX_SO_SNDBUF: 366 return (SO_SNDBUF); 367 case LINUX_SO_RCVBUF: 368 return (SO_RCVBUF); 369 case LINUX_SO_KEEPALIVE: 370 return (SO_KEEPALIVE); 371 case LINUX_SO_OOBINLINE: 372 return (SO_OOBINLINE); 373 case LINUX_SO_LINGER: 374 return (SO_LINGER); 375 case LINUX_SO_PEERCRED: 376 return (LOCAL_PEERCRED); 377 case LINUX_SO_RCVLOWAT: 378 return (SO_RCVLOWAT); 379 case LINUX_SO_SNDLOWAT: 380 return (SO_SNDLOWAT); 381 case LINUX_SO_RCVTIMEO: 382 return (SO_RCVTIMEO); 383 case LINUX_SO_SNDTIMEO: 384 return (SO_SNDTIMEO); 385 case LINUX_SO_TIMESTAMP: 386 return (SO_TIMESTAMP); 387 case LINUX_SO_ACCEPTCONN: 388 return (SO_ACCEPTCONN); 389 } 390 return (-1); 391 } 392 393 static int 394 linux_to_bsd_tcp_sockopt(int opt) 395 { 396 397 switch (opt) { 398 case LINUX_TCP_NODELAY: 399 return (TCP_NODELAY); 400 case LINUX_TCP_MAXSEG: 401 return (TCP_MAXSEG); 402 case LINUX_TCP_KEEPIDLE: 403 return (TCP_KEEPIDLE); 404 case LINUX_TCP_KEEPINTVL: 405 return (TCP_KEEPINTVL); 406 case LINUX_TCP_KEEPCNT: 407 return (TCP_KEEPCNT); 408 case LINUX_TCP_MD5SIG: 409 return (TCP_MD5SIG); 410 } 411 return (-1); 412 } 413 414 static int 415 linux_to_bsd_msg_flags(int flags) 416 { 417 int ret_flags = 0; 418 419 if (flags & LINUX_MSG_OOB) 420 ret_flags |= MSG_OOB; 421 if (flags & LINUX_MSG_PEEK) 422 ret_flags |= MSG_PEEK; 423 if (flags & LINUX_MSG_DONTROUTE) 424 ret_flags |= MSG_DONTROUTE; 425 if (flags & LINUX_MSG_CTRUNC) 426 ret_flags |= MSG_CTRUNC; 427 if (flags & LINUX_MSG_TRUNC) 428 ret_flags |= MSG_TRUNC; 429 if (flags & LINUX_MSG_DONTWAIT) 430 ret_flags |= MSG_DONTWAIT; 431 if (flags & LINUX_MSG_EOR) 432 ret_flags |= MSG_EOR; 433 if (flags & LINUX_MSG_WAITALL) 434 ret_flags |= MSG_WAITALL; 435 if (flags & LINUX_MSG_NOSIGNAL) 436 ret_flags |= MSG_NOSIGNAL; 437 #if 0 /* not handled */ 438 if (flags & LINUX_MSG_PROXY) 439 ; 440 if (flags & LINUX_MSG_FIN) 441 ; 442 if (flags & LINUX_MSG_SYN) 443 ; 444 if (flags & LINUX_MSG_CONFIRM) 445 ; 446 if (flags & LINUX_MSG_RST) 447 ; 448 if (flags & LINUX_MSG_ERRQUEUE) 449 ; 450 #endif 451 return (ret_flags); 452 } 453 454 /* 455 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the 456 * native syscall will fault. Thus, we don't really need to check the 457 * return values for these functions. 458 */ 459 460 static int 461 bsd_to_linux_sockaddr(struct sockaddr *arg) 462 { 463 struct sockaddr sa; 464 size_t sa_len = sizeof(struct sockaddr); 465 int error, bdom; 466 467 if ((error = copyin(arg, &sa, sa_len))) 468 return (error); 469 470 bdom = bsd_to_linux_domain(sa.sa_family); 471 if (bdom == -1) 472 return (EAFNOSUPPORT); 473 474 *(u_short *)&sa = bdom; 475 return (copyout(&sa, arg, sa_len)); 476 } 477 478 static int 479 linux_to_bsd_sockaddr(struct sockaddr *arg, int len) 480 { 481 struct sockaddr sa; 482 size_t sa_len = sizeof(struct sockaddr); 483 int error, bdom; 484 485 if ((error = copyin(arg, &sa, sa_len))) 486 return (error); 487 488 bdom = linux_to_bsd_domain(*(sa_family_t *)&sa); 489 if (bdom == -1) 490 return (EAFNOSUPPORT); 491 492 sa.sa_family = bdom; 493 sa.sa_len = len; 494 return (copyout(&sa, arg, sa_len)); 495 } 496 497 static int 498 linux_sa_put(struct osockaddr *osa) 499 { 500 struct osockaddr sa; 501 int error, bdom; 502 503 /* 504 * Only read/write the osockaddr family part, the rest is 505 * not changed. 506 */ 507 error = copyin(osa, &sa, sizeof(sa.sa_family)); 508 if (error != 0) 509 return (error); 510 511 bdom = bsd_to_linux_domain(sa.sa_family); 512 if (bdom == -1) 513 return (EINVAL); 514 515 sa.sa_family = bdom; 516 return (copyout(&sa, osa, sizeof(sa.sa_family))); 517 } 518 519 static int 520 linux_to_bsd_cmsg_type(int cmsg_type) 521 { 522 523 switch (cmsg_type) { 524 case LINUX_SCM_RIGHTS: 525 return (SCM_RIGHTS); 526 case LINUX_SCM_CREDENTIALS: 527 return (SCM_CREDS); 528 } 529 return (-1); 530 } 531 532 static int 533 bsd_to_linux_cmsg_type(int cmsg_type) 534 { 535 536 switch (cmsg_type) { 537 case SCM_RIGHTS: 538 return (LINUX_SCM_RIGHTS); 539 case SCM_CREDS: 540 return (LINUX_SCM_CREDENTIALS); 541 case SCM_TIMESTAMP: 542 return (LINUX_SCM_TIMESTAMP); 543 } 544 return (-1); 545 } 546 547 static int 548 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 549 { 550 if (lhdr->msg_controllen > INT_MAX) 551 return (ENOBUFS); 552 553 bhdr->msg_name = PTRIN(lhdr->msg_name); 554 bhdr->msg_namelen = lhdr->msg_namelen; 555 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 556 bhdr->msg_iovlen = lhdr->msg_iovlen; 557 bhdr->msg_control = PTRIN(lhdr->msg_control); 558 559 /* 560 * msg_controllen is skipped since BSD and LINUX control messages 561 * are potentially different sizes (e.g. the cred structure used 562 * by SCM_CREDS is different between the two operating system). 563 * 564 * The caller can set it (if necessary) after converting all the 565 * control messages. 566 */ 567 568 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 569 return (0); 570 } 571 572 static int 573 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 574 { 575 lhdr->msg_name = PTROUT(bhdr->msg_name); 576 lhdr->msg_namelen = bhdr->msg_namelen; 577 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 578 lhdr->msg_iovlen = bhdr->msg_iovlen; 579 lhdr->msg_control = PTROUT(bhdr->msg_control); 580 581 /* 582 * msg_controllen is skipped since BSD and LINUX control messages 583 * are potentially different sizes (e.g. the cred structure used 584 * by SCM_CREDS is different between the two operating system). 585 * 586 * The caller can set it (if necessary) after converting all the 587 * control messages. 588 */ 589 590 /* msg_flags skipped */ 591 return (0); 592 } 593 594 static int 595 linux_set_socket_flags(int lflags, int *flags) 596 { 597 598 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 599 return (EINVAL); 600 if (lflags & LINUX_SOCK_NONBLOCK) 601 *flags |= SOCK_NONBLOCK; 602 if (lflags & LINUX_SOCK_CLOEXEC) 603 *flags |= SOCK_CLOEXEC; 604 return (0); 605 } 606 607 static int 608 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 609 struct mbuf *control, enum uio_seg segflg) 610 { 611 struct sockaddr *to; 612 int error; 613 614 if (mp->msg_name != NULL) { 615 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen); 616 if (error != 0) 617 return (error); 618 mp->msg_name = to; 619 } else 620 to = NULL; 621 622 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 623 segflg); 624 625 if (to) 626 free(to, M_SONAME); 627 return (error); 628 } 629 630 /* Return 0 if IP_HDRINCL is set for the given socket. */ 631 static int 632 linux_check_hdrincl(struct thread *td, int s) 633 { 634 int error, optval; 635 socklen_t size_val; 636 637 size_val = sizeof(optval); 638 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 639 &optval, UIO_SYSSPACE, &size_val); 640 if (error != 0) 641 return (error); 642 643 return (optval == 0); 644 } 645 646 /* 647 * Updated sendto() when IP_HDRINCL is set: 648 * tweak endian-dependent fields in the IP packet. 649 */ 650 static int 651 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 652 { 653 /* 654 * linux_ip_copysize defines how many bytes we should copy 655 * from the beginning of the IP packet before we customize it for BSD. 656 * It should include all the fields we modify (ip_len and ip_off). 657 */ 658 #define linux_ip_copysize 8 659 660 struct ip *packet; 661 struct msghdr msg; 662 struct iovec aiov[1]; 663 int error; 664 665 /* Check that the packet isn't too big or too small. */ 666 if (linux_args->len < linux_ip_copysize || 667 linux_args->len > IP_MAXPACKET) 668 return (EINVAL); 669 670 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK); 671 672 /* Make kernel copy of the packet to be sent */ 673 if ((error = copyin(PTRIN(linux_args->msg), packet, 674 linux_args->len))) 675 goto goout; 676 677 /* Convert fields from Linux to BSD raw IP socket format */ 678 packet->ip_len = linux_args->len; 679 packet->ip_off = ntohs(packet->ip_off); 680 681 /* Prepare the msghdr and iovec structures describing the new packet */ 682 msg.msg_name = PTRIN(linux_args->to); 683 msg.msg_namelen = linux_args->tolen; 684 msg.msg_iov = aiov; 685 msg.msg_iovlen = 1; 686 msg.msg_control = NULL; 687 msg.msg_flags = 0; 688 aiov[0].iov_base = (char *)packet; 689 aiov[0].iov_len = linux_args->len; 690 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 691 NULL, UIO_SYSSPACE); 692 goout: 693 free(packet, M_LINUX); 694 return (error); 695 } 696 697 int 698 linux_socket(struct thread *td, struct linux_socket_args *args) 699 { 700 int domain, retval_socket, type; 701 702 type = args->type & LINUX_SOCK_TYPE_MASK; 703 if (type < 0 || type > LINUX_SOCK_MAX) 704 return (EINVAL); 705 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 706 &type); 707 if (retval_socket != 0) 708 return (retval_socket); 709 domain = linux_to_bsd_domain(args->domain); 710 if (domain == -1) 711 return (EAFNOSUPPORT); 712 713 retval_socket = kern_socket(td, domain, type, args->protocol); 714 if (retval_socket) 715 return (retval_socket); 716 717 if (type == SOCK_RAW 718 && (args->protocol == IPPROTO_RAW || args->protocol == 0) 719 && domain == PF_INET) { 720 /* It's a raw IP socket: set the IP_HDRINCL option. */ 721 int hdrincl; 722 723 hdrincl = 1; 724 /* We ignore any error returned by kern_setsockopt() */ 725 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 726 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 727 } 728 #ifdef INET6 729 /* 730 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default 731 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps. 732 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only 733 * sysctl value. 734 */ 735 if (domain == PF_INET6) { 736 int v6only; 737 738 v6only = 0; 739 /* We ignore any error returned by setsockopt() */ 740 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 741 &v6only, UIO_SYSSPACE, sizeof(v6only)); 742 } 743 #endif 744 745 return (retval_socket); 746 } 747 748 int 749 linux_bind(struct thread *td, struct linux_bind_args *args) 750 { 751 struct sockaddr *sa; 752 int error; 753 754 error = linux_getsockaddr(&sa, PTRIN(args->name), 755 args->namelen); 756 if (error != 0) 757 return (error); 758 759 error = kern_bindat(td, AT_FDCWD, args->s, sa); 760 free(sa, M_SONAME); 761 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 762 return (EINVAL); 763 return (error); 764 } 765 766 int 767 linux_connect(struct thread *td, struct linux_connect_args *args) 768 { 769 struct socket *so; 770 struct sockaddr *sa; 771 struct file *fp; 772 u_int fflag; 773 int error; 774 775 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 776 args->namelen); 777 if (error != 0) 778 return (error); 779 780 error = kern_connectat(td, AT_FDCWD, args->s, sa); 781 free(sa, M_SONAME); 782 if (error != EISCONN) 783 return (error); 784 785 /* 786 * Linux doesn't return EISCONN the first time it occurs, 787 * when on a non-blocking socket. Instead it returns the 788 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 789 */ 790 error = getsock_cap(td, args->s, &cap_connect_rights, 791 &fp, &fflag, NULL); 792 if (error != 0) 793 return (error); 794 795 error = EISCONN; 796 so = fp->f_data; 797 if (fflag & FNONBLOCK) { 798 SOCK_LOCK(so); 799 if (so->so_emuldata == 0) 800 error = so->so_error; 801 so->so_emuldata = (void *)1; 802 SOCK_UNLOCK(so); 803 } 804 fdrop(fp, td); 805 806 return (error); 807 } 808 809 int 810 linux_listen(struct thread *td, struct linux_listen_args *args) 811 { 812 813 return (kern_listen(td, args->s, args->backlog)); 814 } 815 816 static int 817 linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 818 l_uintptr_t namelen, int flags) 819 { 820 struct accept4_args /* { 821 int s; 822 struct sockaddr * __restrict name; 823 socklen_t * __restrict anamelen; 824 int flags; 825 } */ bsd_args; 826 struct socket *so; 827 struct file *fp; 828 int error, error1; 829 830 bsd_args.s = s; 831 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr); 832 bsd_args.anamelen = PTRIN(namelen); 833 bsd_args.flags = 0; 834 error = linux_set_socket_flags(flags, &bsd_args.flags); 835 if (error != 0) 836 return (error); 837 error = sys_accept4(td, &bsd_args); 838 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 839 if (error != 0) { 840 if (error == EFAULT && namelen != sizeof(struct sockaddr_in)) 841 return (EINVAL); 842 if (error == EINVAL) { 843 error1 = getsock_cap(td, s, &cap_accept_rights, &fp, NULL, NULL); 844 if (error1 != 0) 845 return (error1); 846 so = fp->f_data; 847 if (so->so_type == SOCK_DGRAM) { 848 fdrop(fp, td); 849 return (EOPNOTSUPP); 850 } 851 fdrop(fp, td); 852 } 853 return (error); 854 } 855 if (addr) 856 error = linux_sa_put(PTRIN(addr)); 857 if (error != 0) { 858 (void)kern_close(td, td->td_retval[0]); 859 td->td_retval[0] = 0; 860 } 861 return (error); 862 } 863 864 int 865 linux_accept(struct thread *td, struct linux_accept_args *args) 866 { 867 868 return (linux_accept_common(td, args->s, args->addr, 869 args->namelen, 0)); 870 } 871 872 int 873 linux_accept4(struct thread *td, struct linux_accept4_args *args) 874 { 875 876 return (linux_accept_common(td, args->s, args->addr, 877 args->namelen, args->flags)); 878 } 879 880 int 881 linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 882 { 883 struct getsockname_args /* { 884 int fdes; 885 struct sockaddr * __restrict asa; 886 socklen_t * __restrict alen; 887 } */ bsd_args; 888 int error; 889 890 bsd_args.fdes = args->s; 891 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 892 bsd_args.alen = PTRIN(args->namelen); 893 error = sys_getsockname(td, &bsd_args); 894 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 895 if (error != 0) 896 return (error); 897 return (linux_sa_put(PTRIN(args->addr))); 898 } 899 900 int 901 linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 902 { 903 struct getpeername_args /* { 904 int fdes; 905 caddr_t asa; 906 int *alen; 907 } */ bsd_args; 908 int error; 909 910 bsd_args.fdes = args->s; 911 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 912 bsd_args.alen = (socklen_t *)PTRIN(args->namelen); 913 error = sys_getpeername(td, &bsd_args); 914 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 915 if (error != 0) 916 return (error); 917 return (linux_sa_put(PTRIN(args->addr))); 918 } 919 920 int 921 linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 922 { 923 struct socketpair_args /* { 924 int domain; 925 int type; 926 int protocol; 927 int *rsv; 928 } */ bsd_args; 929 int error; 930 931 bsd_args.domain = linux_to_bsd_domain(args->domain); 932 if (bsd_args.domain != PF_LOCAL) 933 return (EAFNOSUPPORT); 934 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 935 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 936 return (EINVAL); 937 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 938 &bsd_args.type); 939 if (error != 0) 940 return (error); 941 if (args->protocol != 0 && args->protocol != PF_UNIX) 942 943 /* 944 * Use of PF_UNIX as protocol argument is not right, 945 * but Linux does it. 946 * Do not map PF_UNIX as its Linux value is identical 947 * to FreeBSD one. 948 */ 949 return (EPROTONOSUPPORT); 950 else 951 bsd_args.protocol = 0; 952 bsd_args.rsv = (int *)PTRIN(args->rsv); 953 return (sys_socketpair(td, &bsd_args)); 954 } 955 956 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 957 struct linux_send_args { 958 register_t s; 959 register_t msg; 960 register_t len; 961 register_t flags; 962 }; 963 964 static int 965 linux_send(struct thread *td, struct linux_send_args *args) 966 { 967 struct sendto_args /* { 968 int s; 969 caddr_t buf; 970 int len; 971 int flags; 972 caddr_t to; 973 int tolen; 974 } */ bsd_args; 975 976 bsd_args.s = args->s; 977 bsd_args.buf = (caddr_t)PTRIN(args->msg); 978 bsd_args.len = args->len; 979 bsd_args.flags = args->flags; 980 bsd_args.to = NULL; 981 bsd_args.tolen = 0; 982 return (sys_sendto(td, &bsd_args)); 983 } 984 985 struct linux_recv_args { 986 register_t s; 987 register_t msg; 988 register_t len; 989 register_t flags; 990 }; 991 992 static int 993 linux_recv(struct thread *td, struct linux_recv_args *args) 994 { 995 struct recvfrom_args /* { 996 int s; 997 caddr_t buf; 998 int len; 999 int flags; 1000 struct sockaddr *from; 1001 socklen_t fromlenaddr; 1002 } */ bsd_args; 1003 1004 bsd_args.s = args->s; 1005 bsd_args.buf = (caddr_t)PTRIN(args->msg); 1006 bsd_args.len = args->len; 1007 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 1008 bsd_args.from = NULL; 1009 bsd_args.fromlenaddr = 0; 1010 return (sys_recvfrom(td, &bsd_args)); 1011 } 1012 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1013 1014 int 1015 linux_sendto(struct thread *td, struct linux_sendto_args *args) 1016 { 1017 struct msghdr msg; 1018 struct iovec aiov; 1019 1020 if (linux_check_hdrincl(td, args->s) == 0) 1021 /* IP_HDRINCL set, tweak the packet before sending */ 1022 return (linux_sendto_hdrincl(td, args)); 1023 1024 msg.msg_name = PTRIN(args->to); 1025 msg.msg_namelen = args->tolen; 1026 msg.msg_iov = &aiov; 1027 msg.msg_iovlen = 1; 1028 msg.msg_control = NULL; 1029 msg.msg_flags = 0; 1030 aiov.iov_base = PTRIN(args->msg); 1031 aiov.iov_len = args->len; 1032 return (linux_sendit(td, args->s, &msg, args->flags, NULL, 1033 UIO_USERSPACE)); 1034 } 1035 1036 int 1037 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 1038 { 1039 struct msghdr msg; 1040 struct iovec aiov; 1041 int error, fromlen; 1042 1043 if (PTRIN(args->fromlen) != NULL) { 1044 error = copyin(PTRIN(args->fromlen), &fromlen, 1045 sizeof(fromlen)); 1046 if (error != 0) 1047 return (error); 1048 if (fromlen < 0) 1049 return (EINVAL); 1050 msg.msg_namelen = fromlen; 1051 } else 1052 msg.msg_namelen = 0; 1053 1054 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from); 1055 msg.msg_iov = &aiov; 1056 msg.msg_iovlen = 1; 1057 aiov.iov_base = PTRIN(args->buf); 1058 aiov.iov_len = args->len; 1059 msg.msg_control = 0; 1060 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 1061 1062 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL); 1063 if (error != 0) 1064 return (error); 1065 1066 if (PTRIN(args->from) != NULL) { 1067 error = bsd_to_linux_sockaddr((struct sockaddr *) 1068 PTRIN(args->from)); 1069 if (error != 0) 1070 return (error); 1071 1072 error = linux_sa_put((struct osockaddr *) 1073 PTRIN(args->from)); 1074 } 1075 1076 if (PTRIN(args->fromlen) != NULL) 1077 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 1078 sizeof(msg.msg_namelen)); 1079 1080 return (error); 1081 } 1082 1083 static int 1084 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1085 l_uint flags) 1086 { 1087 struct cmsghdr *cmsg; 1088 struct mbuf *control; 1089 struct msghdr msg; 1090 struct l_cmsghdr linux_cmsg; 1091 struct l_cmsghdr *ptr_cmsg; 1092 struct l_msghdr linux_msg; 1093 struct iovec *iov; 1094 socklen_t datalen; 1095 struct sockaddr *sa; 1096 sa_family_t sa_family; 1097 void *data; 1098 l_size_t len; 1099 l_size_t clen; 1100 int error; 1101 1102 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1103 if (error != 0) 1104 return (error); 1105 1106 /* 1107 * Some Linux applications (ping) define a non-NULL control data 1108 * pointer, but a msg_controllen of 0, which is not allowed in the 1109 * FreeBSD system call interface. NULL the msg_control pointer in 1110 * order to handle this case. This should be checked, but allows the 1111 * Linux ping to work. 1112 */ 1113 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0) 1114 linux_msg.msg_control = PTROUT(NULL); 1115 1116 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1117 if (error != 0) 1118 return (error); 1119 1120 #ifdef COMPAT_LINUX32 1121 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1122 &iov, EMSGSIZE); 1123 #else 1124 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1125 #endif 1126 if (error != 0) 1127 return (error); 1128 1129 control = NULL; 1130 1131 if (linux_msg.msg_controllen >= sizeof(struct l_cmsghdr)) { 1132 error = kern_getsockname(td, s, &sa, &datalen); 1133 if (error != 0) 1134 goto bad; 1135 sa_family = sa->sa_family; 1136 free(sa, M_SONAME); 1137 1138 error = ENOBUFS; 1139 control = m_get(M_WAITOK, MT_CONTROL); 1140 MCLGET(control, M_WAITOK); 1141 data = mtod(control, void *); 1142 datalen = 0; 1143 1144 ptr_cmsg = PTRIN(linux_msg.msg_control); 1145 clen = linux_msg.msg_controllen; 1146 do { 1147 error = copyin(ptr_cmsg, &linux_cmsg, 1148 sizeof(struct l_cmsghdr)); 1149 if (error != 0) 1150 goto bad; 1151 1152 error = EINVAL; 1153 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) || 1154 linux_cmsg.cmsg_len > clen) 1155 goto bad; 1156 1157 if (datalen + CMSG_HDRSZ > MCLBYTES) 1158 goto bad; 1159 1160 /* 1161 * Now we support only SCM_RIGHTS and SCM_CRED, 1162 * so return EINVAL in any other cmsg_type 1163 */ 1164 cmsg = data; 1165 cmsg->cmsg_type = 1166 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1167 cmsg->cmsg_level = 1168 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1169 if (cmsg->cmsg_type == -1 1170 || cmsg->cmsg_level != SOL_SOCKET) 1171 goto bad; 1172 1173 /* 1174 * Some applications (e.g. pulseaudio) attempt to 1175 * send ancillary data even if the underlying protocol 1176 * doesn't support it which is not allowed in the 1177 * FreeBSD system call interface. 1178 */ 1179 if (sa_family != AF_UNIX) 1180 continue; 1181 1182 if (cmsg->cmsg_type == SCM_CREDS) { 1183 len = sizeof(struct cmsgcred); 1184 if (datalen + CMSG_SPACE(len) > MCLBYTES) 1185 goto bad; 1186 1187 /* 1188 * The lower levels will fill in the structure 1189 */ 1190 memset(CMSG_DATA(data), 0, len); 1191 } else { 1192 len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1193 if (datalen + CMSG_SPACE(len) < datalen || 1194 datalen + CMSG_SPACE(len) > MCLBYTES) 1195 goto bad; 1196 1197 error = copyin(LINUX_CMSG_DATA(ptr_cmsg), 1198 CMSG_DATA(data), len); 1199 if (error != 0) 1200 goto bad; 1201 } 1202 1203 cmsg->cmsg_len = CMSG_LEN(len); 1204 data = (char *)data + CMSG_SPACE(len); 1205 datalen += CMSG_SPACE(len); 1206 1207 if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len)) 1208 break; 1209 1210 clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len); 1211 ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg + 1212 LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len)); 1213 } while(clen >= sizeof(struct l_cmsghdr)); 1214 1215 control->m_len = datalen; 1216 if (datalen == 0) { 1217 m_freem(control); 1218 control = NULL; 1219 } 1220 } 1221 1222 msg.msg_iov = iov; 1223 msg.msg_flags = 0; 1224 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1225 control = NULL; 1226 1227 bad: 1228 m_freem(control); 1229 free(iov, M_IOV); 1230 return (error); 1231 } 1232 1233 int 1234 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1235 { 1236 1237 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1238 args->flags)); 1239 } 1240 1241 int 1242 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1243 { 1244 struct l_mmsghdr *msg; 1245 l_uint retval; 1246 int error, datagrams; 1247 1248 if (args->vlen > UIO_MAXIOV) 1249 args->vlen = UIO_MAXIOV; 1250 1251 msg = PTRIN(args->msg); 1252 datagrams = 0; 1253 while (datagrams < args->vlen) { 1254 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1255 args->flags); 1256 if (error != 0) 1257 break; 1258 1259 retval = td->td_retval[0]; 1260 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1261 if (error != 0) 1262 break; 1263 ++msg; 1264 ++datagrams; 1265 } 1266 if (error == 0) 1267 td->td_retval[0] = datagrams; 1268 return (error); 1269 } 1270 1271 static int 1272 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1273 l_uint flags, struct msghdr *msg) 1274 { 1275 struct cmsghdr *cm; 1276 struct cmsgcred *cmcred; 1277 struct l_cmsghdr *linux_cmsg = NULL; 1278 struct l_ucred linux_ucred; 1279 socklen_t datalen, maxlen, outlen; 1280 struct l_msghdr linux_msg; 1281 struct iovec *iov, *uiov; 1282 struct mbuf *control = NULL; 1283 struct mbuf **controlp; 1284 struct timeval *ftmvl; 1285 l_timeval ltmvl; 1286 caddr_t outbuf; 1287 void *data; 1288 int error, i, fd, fds, *fdp; 1289 1290 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1291 if (error != 0) 1292 return (error); 1293 1294 error = linux_to_bsd_msghdr(msg, &linux_msg); 1295 if (error != 0) 1296 return (error); 1297 1298 #ifdef COMPAT_LINUX32 1299 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1300 &iov, EMSGSIZE); 1301 #else 1302 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1303 #endif 1304 if (error != 0) 1305 return (error); 1306 1307 if (msg->msg_name) { 1308 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name, 1309 msg->msg_namelen); 1310 if (error != 0) 1311 goto bad; 1312 } 1313 1314 uiov = msg->msg_iov; 1315 msg->msg_iov = iov; 1316 controlp = (msg->msg_control != NULL) ? &control : NULL; 1317 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp); 1318 msg->msg_iov = uiov; 1319 if (error != 0) 1320 goto bad; 1321 1322 error = bsd_to_linux_msghdr(msg, &linux_msg); 1323 if (error != 0) 1324 goto bad; 1325 1326 if (linux_msg.msg_name) { 1327 error = bsd_to_linux_sockaddr((struct sockaddr *) 1328 PTRIN(linux_msg.msg_name)); 1329 if (error != 0) 1330 goto bad; 1331 } 1332 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1333 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1334 if (error != 0) 1335 goto bad; 1336 } 1337 1338 maxlen = linux_msg.msg_controllen; 1339 linux_msg.msg_controllen = 0; 1340 if (control) { 1341 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1342 1343 msg->msg_control = mtod(control, struct cmsghdr *); 1344 msg->msg_controllen = control->m_len; 1345 1346 cm = CMSG_FIRSTHDR(msg); 1347 outbuf = PTRIN(linux_msg.msg_control); 1348 outlen = 0; 1349 while (cm != NULL) { 1350 linux_cmsg->cmsg_type = 1351 bsd_to_linux_cmsg_type(cm->cmsg_type); 1352 linux_cmsg->cmsg_level = 1353 bsd_to_linux_sockopt_level(cm->cmsg_level); 1354 if (linux_cmsg->cmsg_type == -1 || 1355 cm->cmsg_level != SOL_SOCKET) { 1356 error = EINVAL; 1357 goto bad; 1358 } 1359 1360 data = CMSG_DATA(cm); 1361 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1362 1363 switch (cm->cmsg_type) { 1364 case SCM_RIGHTS: 1365 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1366 fds = datalen / sizeof(int); 1367 fdp = data; 1368 for (i = 0; i < fds; i++) { 1369 fd = *fdp++; 1370 (void)kern_fcntl(td, fd, 1371 F_SETFD, FD_CLOEXEC); 1372 } 1373 } 1374 break; 1375 1376 case SCM_CREDS: 1377 /* 1378 * Currently LOCAL_CREDS is never in 1379 * effect for Linux so no need to worry 1380 * about sockcred 1381 */ 1382 if (datalen != sizeof(*cmcred)) { 1383 error = EMSGSIZE; 1384 goto bad; 1385 } 1386 cmcred = (struct cmsgcred *)data; 1387 bzero(&linux_ucred, sizeof(linux_ucred)); 1388 linux_ucred.pid = cmcred->cmcred_pid; 1389 linux_ucred.uid = cmcred->cmcred_uid; 1390 linux_ucred.gid = cmcred->cmcred_gid; 1391 data = &linux_ucred; 1392 datalen = sizeof(linux_ucred); 1393 break; 1394 1395 case SCM_TIMESTAMP: 1396 if (datalen != sizeof(struct timeval)) { 1397 error = EMSGSIZE; 1398 goto bad; 1399 } 1400 ftmvl = (struct timeval *)data; 1401 ltmvl.tv_sec = ftmvl->tv_sec; 1402 ltmvl.tv_usec = ftmvl->tv_usec; 1403 data = <mvl; 1404 datalen = sizeof(ltmvl); 1405 break; 1406 } 1407 1408 if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) { 1409 if (outlen == 0) { 1410 error = EMSGSIZE; 1411 goto bad; 1412 } else { 1413 linux_msg.msg_flags |= LINUX_MSG_CTRUNC; 1414 m_dispose_extcontrolm(control); 1415 goto out; 1416 } 1417 } 1418 1419 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1420 1421 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1422 if (error != 0) 1423 goto bad; 1424 outbuf += L_CMSG_HDRSZ; 1425 1426 error = copyout(data, outbuf, datalen); 1427 if (error != 0) 1428 goto bad; 1429 1430 outbuf += LINUX_CMSG_ALIGN(datalen); 1431 outlen += LINUX_CMSG_LEN(datalen); 1432 1433 cm = CMSG_NXTHDR(msg, cm); 1434 } 1435 linux_msg.msg_controllen = outlen; 1436 } 1437 1438 out: 1439 error = copyout(&linux_msg, msghdr, sizeof(linux_msg)); 1440 1441 bad: 1442 if (control != NULL) { 1443 if (error != 0) 1444 m_dispose_extcontrolm(control); 1445 m_freem(control); 1446 } 1447 free(iov, M_IOV); 1448 free(linux_cmsg, M_LINUX); 1449 1450 return (error); 1451 } 1452 1453 int 1454 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1455 { 1456 struct msghdr bsd_msg; 1457 1458 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1459 args->flags, &bsd_msg)); 1460 } 1461 1462 int 1463 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1464 { 1465 struct l_mmsghdr *msg; 1466 struct msghdr bsd_msg; 1467 struct l_timespec lts; 1468 struct timespec ts, tts; 1469 l_uint retval; 1470 int error, datagrams; 1471 1472 if (args->timeout) { 1473 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1474 if (error != 0) 1475 return (error); 1476 error = linux_to_native_timespec(&ts, <s); 1477 if (error != 0) 1478 return (error); 1479 getnanotime(&tts); 1480 timespecadd(&tts, &ts, &tts); 1481 } 1482 1483 msg = PTRIN(args->msg); 1484 datagrams = 0; 1485 while (datagrams < args->vlen) { 1486 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1487 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1488 if (error != 0) 1489 break; 1490 1491 retval = td->td_retval[0]; 1492 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1493 if (error != 0) 1494 break; 1495 ++msg; 1496 ++datagrams; 1497 1498 /* 1499 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1500 */ 1501 if (args->flags & LINUX_MSG_WAITFORONE) 1502 args->flags |= LINUX_MSG_DONTWAIT; 1503 1504 /* 1505 * See BUGS section of recvmmsg(2). 1506 */ 1507 if (args->timeout) { 1508 getnanotime(&ts); 1509 timespecsub(&ts, &tts, &ts); 1510 if (!timespecisset(&ts) || ts.tv_sec > 0) 1511 break; 1512 } 1513 /* Out of band data, return right away. */ 1514 if (bsd_msg.msg_flags & MSG_OOB) 1515 break; 1516 } 1517 if (error == 0) 1518 td->td_retval[0] = datagrams; 1519 return (error); 1520 } 1521 1522 int 1523 linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1524 { 1525 1526 return (kern_shutdown(td, args->s, args->how)); 1527 } 1528 1529 int 1530 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1531 { 1532 struct setsockopt_args /* { 1533 int s; 1534 int level; 1535 int name; 1536 const void *val; 1537 int valsize; 1538 } */ bsd_args; 1539 l_timeval linux_tv; 1540 struct timeval tv; 1541 int error, name; 1542 1543 bsd_args.s = args->s; 1544 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1545 switch (bsd_args.level) { 1546 case SOL_SOCKET: 1547 name = linux_to_bsd_so_sockopt(args->optname); 1548 switch (name) { 1549 case SO_RCVTIMEO: 1550 /* FALLTHROUGH */ 1551 case SO_SNDTIMEO: 1552 error = copyin(PTRIN(args->optval), &linux_tv, 1553 sizeof(linux_tv)); 1554 if (error != 0) 1555 return (error); 1556 tv.tv_sec = linux_tv.tv_sec; 1557 tv.tv_usec = linux_tv.tv_usec; 1558 return (kern_setsockopt(td, args->s, bsd_args.level, 1559 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1560 /* NOTREACHED */ 1561 default: 1562 break; 1563 } 1564 break; 1565 case IPPROTO_IP: 1566 name = linux_to_bsd_ip_sockopt(args->optname); 1567 break; 1568 case IPPROTO_IPV6: 1569 name = linux_to_bsd_ip6_sockopt(args->optname); 1570 break; 1571 case IPPROTO_TCP: 1572 name = linux_to_bsd_tcp_sockopt(args->optname); 1573 break; 1574 default: 1575 name = -1; 1576 break; 1577 } 1578 if (name == -1) 1579 return (ENOPROTOOPT); 1580 1581 bsd_args.name = name; 1582 bsd_args.val = PTRIN(args->optval); 1583 bsd_args.valsize = args->optlen; 1584 1585 if (name == IPV6_NEXTHOP) { 1586 linux_to_bsd_sockaddr(__DECONST(struct sockaddr *, 1587 bsd_args.val), bsd_args.valsize); 1588 error = sys_setsockopt(td, &bsd_args); 1589 bsd_to_linux_sockaddr(__DECONST(struct sockaddr *, 1590 bsd_args.val)); 1591 } else 1592 error = sys_setsockopt(td, &bsd_args); 1593 1594 return (error); 1595 } 1596 1597 int 1598 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1599 { 1600 struct getsockopt_args /* { 1601 int s; 1602 int level; 1603 int name; 1604 caddr_t val; 1605 int *avalsize; 1606 } */ bsd_args; 1607 l_timeval linux_tv; 1608 struct timeval tv; 1609 socklen_t tv_len, xulen, len; 1610 struct xucred xu; 1611 struct l_ucred lxu; 1612 int error, name, newval; 1613 1614 bsd_args.s = args->s; 1615 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1616 switch (bsd_args.level) { 1617 case SOL_SOCKET: 1618 name = linux_to_bsd_so_sockopt(args->optname); 1619 switch (name) { 1620 case SO_RCVTIMEO: 1621 /* FALLTHROUGH */ 1622 case SO_SNDTIMEO: 1623 tv_len = sizeof(tv); 1624 error = kern_getsockopt(td, args->s, bsd_args.level, 1625 name, &tv, UIO_SYSSPACE, &tv_len); 1626 if (error != 0) 1627 return (error); 1628 linux_tv.tv_sec = tv.tv_sec; 1629 linux_tv.tv_usec = tv.tv_usec; 1630 return (copyout(&linux_tv, PTRIN(args->optval), 1631 sizeof(linux_tv))); 1632 /* NOTREACHED */ 1633 case LOCAL_PEERCRED: 1634 if (args->optlen < sizeof(lxu)) 1635 return (EINVAL); 1636 /* 1637 * LOCAL_PEERCRED is not served at the SOL_SOCKET level, 1638 * but by the Unix socket's level 0. 1639 */ 1640 bsd_args.level = 0; 1641 xulen = sizeof(xu); 1642 error = kern_getsockopt(td, args->s, bsd_args.level, 1643 name, &xu, UIO_SYSSPACE, &xulen); 1644 if (error != 0) 1645 return (error); 1646 /* 1647 * XXX Use 0 for pid as the FreeBSD does not cache peer pid. 1648 */ 1649 lxu.pid = 0; 1650 lxu.uid = xu.cr_uid; 1651 lxu.gid = xu.cr_gid; 1652 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1653 /* NOTREACHED */ 1654 case SO_ERROR: 1655 len = sizeof(newval); 1656 error = kern_getsockopt(td, args->s, bsd_args.level, 1657 name, &newval, UIO_SYSSPACE, &len); 1658 if (error != 0) 1659 return (error); 1660 newval = -SV_ABI_ERRNO(td->td_proc, newval); 1661 return (copyout(&newval, PTRIN(args->optval), len)); 1662 /* NOTREACHED */ 1663 default: 1664 break; 1665 } 1666 break; 1667 case IPPROTO_IP: 1668 name = linux_to_bsd_ip_sockopt(args->optname); 1669 break; 1670 case IPPROTO_IPV6: 1671 name = linux_to_bsd_ip6_sockopt(args->optname); 1672 break; 1673 case IPPROTO_TCP: 1674 name = linux_to_bsd_tcp_sockopt(args->optname); 1675 break; 1676 default: 1677 name = -1; 1678 break; 1679 } 1680 if (name == -1) 1681 return (EINVAL); 1682 1683 bsd_args.name = name; 1684 bsd_args.val = PTRIN(args->optval); 1685 bsd_args.avalsize = PTRIN(args->optlen); 1686 1687 if (name == IPV6_NEXTHOP) { 1688 error = sys_getsockopt(td, &bsd_args); 1689 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1690 } else 1691 error = sys_getsockopt(td, &bsd_args); 1692 1693 return (error); 1694 } 1695 1696 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1697 1698 /* Argument list sizes for linux_socketcall */ 1699 static const unsigned char lxs_args_cnt[] = { 1700 0 /* unused*/, 3 /* socket */, 1701 3 /* bind */, 3 /* connect */, 1702 2 /* listen */, 3 /* accept */, 1703 3 /* getsockname */, 3 /* getpeername */, 1704 4 /* socketpair */, 4 /* send */, 1705 4 /* recv */, 6 /* sendto */, 1706 6 /* recvfrom */, 2 /* shutdown */, 1707 5 /* setsockopt */, 5 /* getsockopt */, 1708 3 /* sendmsg */, 3 /* recvmsg */, 1709 4 /* accept4 */, 5 /* recvmmsg */, 1710 4 /* sendmmsg */ 1711 }; 1712 #define LINUX_ARGS_CNT (nitems(lxs_args_cnt) - 1) 1713 #define LINUX_ARG_SIZE(x) (lxs_args_cnt[x] * sizeof(l_ulong)) 1714 1715 int 1716 linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1717 { 1718 l_ulong a[6]; 1719 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1720 register_t l_args[6]; 1721 #endif 1722 void *arg; 1723 int error; 1724 1725 if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT) 1726 return (EINVAL); 1727 error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what)); 1728 if (error != 0) 1729 return (error); 1730 1731 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1732 for (int i = 0; i < lxs_args_cnt[args->what]; ++i) 1733 l_args[i] = a[i]; 1734 arg = l_args; 1735 #else 1736 arg = a; 1737 #endif 1738 switch (args->what) { 1739 case LINUX_SOCKET: 1740 return (linux_socket(td, arg)); 1741 case LINUX_BIND: 1742 return (linux_bind(td, arg)); 1743 case LINUX_CONNECT: 1744 return (linux_connect(td, arg)); 1745 case LINUX_LISTEN: 1746 return (linux_listen(td, arg)); 1747 case LINUX_ACCEPT: 1748 return (linux_accept(td, arg)); 1749 case LINUX_GETSOCKNAME: 1750 return (linux_getsockname(td, arg)); 1751 case LINUX_GETPEERNAME: 1752 return (linux_getpeername(td, arg)); 1753 case LINUX_SOCKETPAIR: 1754 return (linux_socketpair(td, arg)); 1755 case LINUX_SEND: 1756 return (linux_send(td, arg)); 1757 case LINUX_RECV: 1758 return (linux_recv(td, arg)); 1759 case LINUX_SENDTO: 1760 return (linux_sendto(td, arg)); 1761 case LINUX_RECVFROM: 1762 return (linux_recvfrom(td, arg)); 1763 case LINUX_SHUTDOWN: 1764 return (linux_shutdown(td, arg)); 1765 case LINUX_SETSOCKOPT: 1766 return (linux_setsockopt(td, arg)); 1767 case LINUX_GETSOCKOPT: 1768 return (linux_getsockopt(td, arg)); 1769 case LINUX_SENDMSG: 1770 return (linux_sendmsg(td, arg)); 1771 case LINUX_RECVMSG: 1772 return (linux_recvmsg(td, arg)); 1773 case LINUX_ACCEPT4: 1774 return (linux_accept4(td, arg)); 1775 case LINUX_RECVMMSG: 1776 return (linux_recvmmsg(td, arg)); 1777 case LINUX_SENDMMSG: 1778 return (linux_sendmmsg(td, arg)); 1779 } 1780 1781 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1782 return (ENOSYS); 1783 } 1784 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1785