1 /*- 2 * Copyright (c) 1995 Søren Schmidt 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer 10 * in this position and unchanged. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. The name of the author may not be used to endorse or promote products 15 * derived from this software without specific prior written permission 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 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; 466 467 if ((error = copyin(arg, &sa, sa_len))) 468 return (error); 469 470 *(u_short *)&sa = sa.sa_family; 471 return (copyout(&sa, arg, sa_len)); 472 } 473 474 static int 475 linux_to_bsd_sockaddr(struct sockaddr *arg, int len) 476 { 477 struct sockaddr sa; 478 size_t sa_len = sizeof(struct sockaddr); 479 int error; 480 481 if ((error = copyin(arg, &sa, sa_len))) 482 return (error); 483 484 sa.sa_family = *(sa_family_t *)&sa; 485 sa.sa_len = len; 486 return (copyout(&sa, arg, sa_len)); 487 } 488 489 static int 490 linux_sa_put(struct osockaddr *osa) 491 { 492 struct osockaddr sa; 493 int error, bdom; 494 495 /* 496 * Only read/write the osockaddr family part, the rest is 497 * not changed. 498 */ 499 error = copyin(osa, &sa, sizeof(sa.sa_family)); 500 if (error) 501 return (error); 502 503 bdom = bsd_to_linux_domain(sa.sa_family); 504 if (bdom == -1) 505 return (EINVAL); 506 507 sa.sa_family = bdom; 508 return (copyout(&sa, osa, sizeof(sa.sa_family))); 509 } 510 511 static int 512 linux_to_bsd_cmsg_type(int cmsg_type) 513 { 514 515 switch (cmsg_type) { 516 case LINUX_SCM_RIGHTS: 517 return (SCM_RIGHTS); 518 case LINUX_SCM_CREDENTIALS: 519 return (SCM_CREDS); 520 } 521 return (-1); 522 } 523 524 static int 525 bsd_to_linux_cmsg_type(int cmsg_type) 526 { 527 528 switch (cmsg_type) { 529 case SCM_RIGHTS: 530 return (LINUX_SCM_RIGHTS); 531 case SCM_CREDS: 532 return (LINUX_SCM_CREDENTIALS); 533 case SCM_TIMESTAMP: 534 return (LINUX_SCM_TIMESTAMP); 535 } 536 return (-1); 537 } 538 539 static int 540 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 541 { 542 if (lhdr->msg_controllen > INT_MAX) 543 return (ENOBUFS); 544 545 bhdr->msg_name = PTRIN(lhdr->msg_name); 546 bhdr->msg_namelen = lhdr->msg_namelen; 547 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 548 bhdr->msg_iovlen = lhdr->msg_iovlen; 549 bhdr->msg_control = PTRIN(lhdr->msg_control); 550 551 /* 552 * msg_controllen is skipped since BSD and LINUX control messages 553 * are potentially different sizes (e.g. the cred structure used 554 * by SCM_CREDS is different between the two operating system). 555 * 556 * The caller can set it (if necessary) after converting all the 557 * control messages. 558 */ 559 560 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 561 return (0); 562 } 563 564 static int 565 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 566 { 567 lhdr->msg_name = PTROUT(bhdr->msg_name); 568 lhdr->msg_namelen = bhdr->msg_namelen; 569 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 570 lhdr->msg_iovlen = bhdr->msg_iovlen; 571 lhdr->msg_control = PTROUT(bhdr->msg_control); 572 573 /* 574 * msg_controllen is skipped since BSD and LINUX control messages 575 * are potentially different sizes (e.g. the cred structure used 576 * by SCM_CREDS is different between the two operating system). 577 * 578 * The caller can set it (if necessary) after converting all the 579 * control messages. 580 */ 581 582 /* msg_flags skipped */ 583 return (0); 584 } 585 586 static int 587 linux_set_socket_flags(int lflags, int *flags) 588 { 589 590 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 591 return (EINVAL); 592 if (lflags & LINUX_SOCK_NONBLOCK) 593 *flags |= SOCK_NONBLOCK; 594 if (lflags & LINUX_SOCK_CLOEXEC) 595 *flags |= SOCK_CLOEXEC; 596 return (0); 597 } 598 599 static int 600 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 601 struct mbuf *control, enum uio_seg segflg) 602 { 603 struct sockaddr *to; 604 int error; 605 606 if (mp->msg_name != NULL) { 607 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen); 608 if (error) 609 return (error); 610 mp->msg_name = to; 611 } else 612 to = NULL; 613 614 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 615 segflg); 616 617 if (to) 618 free(to, M_SONAME); 619 return (error); 620 } 621 622 /* Return 0 if IP_HDRINCL is set for the given socket. */ 623 static int 624 linux_check_hdrincl(struct thread *td, int s) 625 { 626 int error, optval; 627 socklen_t size_val; 628 629 size_val = sizeof(optval); 630 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 631 &optval, UIO_SYSSPACE, &size_val); 632 if (error) 633 return (error); 634 635 return (optval == 0); 636 } 637 638 /* 639 * Updated sendto() when IP_HDRINCL is set: 640 * tweak endian-dependent fields in the IP packet. 641 */ 642 static int 643 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 644 { 645 /* 646 * linux_ip_copysize defines how many bytes we should copy 647 * from the beginning of the IP packet before we customize it for BSD. 648 * It should include all the fields we modify (ip_len and ip_off). 649 */ 650 #define linux_ip_copysize 8 651 652 struct ip *packet; 653 struct msghdr msg; 654 struct iovec aiov[1]; 655 int error; 656 657 /* Check that the packet isn't too big or too small. */ 658 if (linux_args->len < linux_ip_copysize || 659 linux_args->len > IP_MAXPACKET) 660 return (EINVAL); 661 662 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK); 663 664 /* Make kernel copy of the packet to be sent */ 665 if ((error = copyin(PTRIN(linux_args->msg), packet, 666 linux_args->len))) 667 goto goout; 668 669 /* Convert fields from Linux to BSD raw IP socket format */ 670 packet->ip_len = linux_args->len; 671 packet->ip_off = ntohs(packet->ip_off); 672 673 /* Prepare the msghdr and iovec structures describing the new packet */ 674 msg.msg_name = PTRIN(linux_args->to); 675 msg.msg_namelen = linux_args->tolen; 676 msg.msg_iov = aiov; 677 msg.msg_iovlen = 1; 678 msg.msg_control = NULL; 679 msg.msg_flags = 0; 680 aiov[0].iov_base = (char *)packet; 681 aiov[0].iov_len = linux_args->len; 682 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 683 NULL, UIO_SYSSPACE); 684 goout: 685 free(packet, M_LINUX); 686 return (error); 687 } 688 689 int 690 linux_socket(struct thread *td, struct linux_socket_args *args) 691 { 692 struct socket_args /* { 693 int domain; 694 int type; 695 int protocol; 696 } */ bsd_args; 697 int retval_socket; 698 699 bsd_args.protocol = args->protocol; 700 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 701 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 702 return (EINVAL); 703 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 704 &bsd_args.type); 705 if (retval_socket != 0) 706 return (retval_socket); 707 bsd_args.domain = linux_to_bsd_domain(args->domain); 708 if (bsd_args.domain == -1) 709 return (EAFNOSUPPORT); 710 711 retval_socket = sys_socket(td, &bsd_args); 712 if (retval_socket) 713 return (retval_socket); 714 715 if (bsd_args.type == SOCK_RAW 716 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0) 717 && bsd_args.domain == PF_INET) { 718 /* It's a raw IP socket: set the IP_HDRINCL option. */ 719 int hdrincl; 720 721 hdrincl = 1; 722 /* We ignore any error returned by kern_setsockopt() */ 723 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 724 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 725 } 726 #ifdef INET6 727 /* 728 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default 729 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps. 730 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only 731 * sysctl value. 732 */ 733 if (bsd_args.domain == PF_INET6) { 734 int v6only; 735 736 v6only = 0; 737 /* We ignore any error returned by setsockopt() */ 738 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 739 &v6only, UIO_SYSSPACE, sizeof(v6only)); 740 } 741 #endif 742 743 return (retval_socket); 744 } 745 746 int 747 linux_bind(struct thread *td, struct linux_bind_args *args) 748 { 749 struct sockaddr *sa; 750 int error; 751 752 error = linux_getsockaddr(&sa, PTRIN(args->name), 753 args->namelen); 754 if (error) 755 return (error); 756 757 error = kern_bindat(td, AT_FDCWD, args->s, sa); 758 free(sa, M_SONAME); 759 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 760 return (EINVAL); 761 return (error); 762 } 763 764 int 765 linux_connect(struct thread *td, struct linux_connect_args *args) 766 { 767 cap_rights_t rights; 768 struct socket *so; 769 struct sockaddr *sa; 770 u_int fflag; 771 int error; 772 773 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 774 args->namelen); 775 if (error) 776 return (error); 777 778 error = kern_connectat(td, AT_FDCWD, args->s, sa); 779 free(sa, M_SONAME); 780 if (error != EISCONN) 781 return (error); 782 783 /* 784 * Linux doesn't return EISCONN the first time it occurs, 785 * when on a non-blocking socket. Instead it returns the 786 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 787 * 788 * XXXRW: Instead of using fgetsock(), check that it is a 789 * socket and use the file descriptor reference instead of 790 * creating a new one. 791 */ 792 error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT), 793 &so, &fflag); 794 if (error == 0) { 795 error = EISCONN; 796 if (fflag & FNONBLOCK) { 797 SOCK_LOCK(so); 798 if (so->so_emuldata == 0) 799 error = so->so_error; 800 so->so_emuldata = (void *)1; 801 SOCK_UNLOCK(so); 802 } 803 fputsock(so); 804 } 805 return (error); 806 } 807 808 int 809 linux_listen(struct thread *td, struct linux_listen_args *args) 810 { 811 struct listen_args /* { 812 int s; 813 int backlog; 814 } */ bsd_args; 815 816 bsd_args.s = args->s; 817 bsd_args.backlog = args->backlog; 818 return (sys_listen(td, &bsd_args)); 819 } 820 821 static int 822 linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 823 l_uintptr_t namelen, int flags) 824 { 825 struct accept4_args /* { 826 int s; 827 struct sockaddr * __restrict name; 828 socklen_t * __restrict anamelen; 829 int flags; 830 } */ bsd_args; 831 cap_rights_t rights; 832 struct socket *so; 833 struct file *fp; 834 int error, error1; 835 836 bsd_args.s = s; 837 /* XXX: */ 838 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr); 839 bsd_args.anamelen = PTRIN(namelen);/* XXX */ 840 bsd_args.flags = 0; 841 error = linux_set_socket_flags(flags, &bsd_args.flags); 842 if (error != 0) 843 return (error); 844 error = sys_accept4(td, &bsd_args); 845 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 846 if (error) { 847 if (error == EFAULT && namelen != sizeof(struct sockaddr_in)) 848 return (EINVAL); 849 if (error == EINVAL) { 850 error1 = getsock_cap(td, s, &rights, &fp, NULL); 851 if (error1 != 0) 852 return (error1); 853 so = fp->f_data; 854 if (so->so_type == SOCK_DGRAM) { 855 fdrop(fp, td); 856 return (EOPNOTSUPP); 857 } 858 fdrop(fp, td); 859 } 860 return (error); 861 } 862 if (addr) 863 error = linux_sa_put(PTRIN(addr)); 864 if (error) { 865 (void)kern_close(td, td->td_retval[0]); 866 td->td_retval[0] = 0; 867 } 868 return (error); 869 } 870 871 int 872 linux_accept(struct thread *td, struct linux_accept_args *args) 873 { 874 875 return (linux_accept_common(td, args->s, args->addr, 876 args->namelen, 0)); 877 } 878 879 int 880 linux_accept4(struct thread *td, struct linux_accept4_args *args) 881 { 882 883 return (linux_accept_common(td, args->s, args->addr, 884 args->namelen, args->flags)); 885 } 886 887 int 888 linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 889 { 890 struct getsockname_args /* { 891 int fdes; 892 struct sockaddr * __restrict asa; 893 socklen_t * __restrict alen; 894 } */ bsd_args; 895 int error; 896 897 bsd_args.fdes = args->s; 898 /* XXX: */ 899 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 900 bsd_args.alen = PTRIN(args->namelen); /* XXX */ 901 error = sys_getsockname(td, &bsd_args); 902 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 903 if (error) 904 return (error); 905 return (linux_sa_put(PTRIN(args->addr))); 906 } 907 908 int 909 linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 910 { 911 struct getpeername_args /* { 912 int fdes; 913 caddr_t asa; 914 int *alen; 915 } */ bsd_args; 916 int error; 917 918 bsd_args.fdes = args->s; 919 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 920 bsd_args.alen = (socklen_t *)PTRIN(args->namelen); 921 error = sys_getpeername(td, &bsd_args); 922 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 923 if (error) 924 return (error); 925 return (linux_sa_put(PTRIN(args->addr))); 926 } 927 928 int 929 linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 930 { 931 struct socketpair_args /* { 932 int domain; 933 int type; 934 int protocol; 935 int *rsv; 936 } */ bsd_args; 937 int error; 938 939 bsd_args.domain = linux_to_bsd_domain(args->domain); 940 if (bsd_args.domain != PF_LOCAL) 941 return (EAFNOSUPPORT); 942 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 943 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 944 return (EINVAL); 945 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 946 &bsd_args.type); 947 if (error != 0) 948 return (error); 949 if (args->protocol != 0 && args->protocol != PF_UNIX) 950 951 /* 952 * Use of PF_UNIX as protocol argument is not right, 953 * but Linux does it. 954 * Do not map PF_UNIX as its Linux value is identical 955 * to FreeBSD one. 956 */ 957 return (EPROTONOSUPPORT); 958 else 959 bsd_args.protocol = 0; 960 bsd_args.rsv = (int *)PTRIN(args->rsv); 961 return (sys_socketpair(td, &bsd_args)); 962 } 963 964 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 965 struct linux_send_args { 966 int s; 967 l_uintptr_t msg; 968 int len; 969 int flags; 970 }; 971 972 static int 973 linux_send(struct thread *td, struct linux_send_args *args) 974 { 975 struct sendto_args /* { 976 int s; 977 caddr_t buf; 978 int len; 979 int flags; 980 caddr_t to; 981 int tolen; 982 } */ bsd_args; 983 984 bsd_args.s = args->s; 985 bsd_args.buf = (caddr_t)PTRIN(args->msg); 986 bsd_args.len = args->len; 987 bsd_args.flags = args->flags; 988 bsd_args.to = NULL; 989 bsd_args.tolen = 0; 990 return (sys_sendto(td, &bsd_args)); 991 } 992 993 struct linux_recv_args { 994 int s; 995 l_uintptr_t msg; 996 int len; 997 int flags; 998 }; 999 1000 static int 1001 linux_recv(struct thread *td, struct linux_recv_args *args) 1002 { 1003 struct recvfrom_args /* { 1004 int s; 1005 caddr_t buf; 1006 int len; 1007 int flags; 1008 struct sockaddr *from; 1009 socklen_t fromlenaddr; 1010 } */ bsd_args; 1011 1012 bsd_args.s = args->s; 1013 bsd_args.buf = (caddr_t)PTRIN(args->msg); 1014 bsd_args.len = args->len; 1015 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 1016 bsd_args.from = NULL; 1017 bsd_args.fromlenaddr = 0; 1018 return (sys_recvfrom(td, &bsd_args)); 1019 } 1020 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1021 1022 int 1023 linux_sendto(struct thread *td, struct linux_sendto_args *args) 1024 { 1025 struct msghdr msg; 1026 struct iovec aiov; 1027 1028 if (linux_check_hdrincl(td, args->s) == 0) 1029 /* IP_HDRINCL set, tweak the packet before sending */ 1030 return (linux_sendto_hdrincl(td, args)); 1031 1032 msg.msg_name = PTRIN(args->to); 1033 msg.msg_namelen = args->tolen; 1034 msg.msg_iov = &aiov; 1035 msg.msg_iovlen = 1; 1036 msg.msg_control = NULL; 1037 msg.msg_flags = 0; 1038 aiov.iov_base = PTRIN(args->msg); 1039 aiov.iov_len = args->len; 1040 return (linux_sendit(td, args->s, &msg, args->flags, NULL, 1041 UIO_USERSPACE)); 1042 } 1043 1044 int 1045 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 1046 { 1047 struct msghdr msg; 1048 struct iovec aiov; 1049 int error; 1050 1051 if (PTRIN(args->fromlen) != NULL) { 1052 error = copyin(PTRIN(args->fromlen), &msg.msg_namelen, 1053 sizeof(msg.msg_namelen)); 1054 if (error != 0) 1055 return (error); 1056 1057 error = linux_to_bsd_sockaddr((struct sockaddr *)PTRIN(args->from), 1058 msg.msg_namelen); 1059 if (error != 0) 1060 return (error); 1061 } else 1062 msg.msg_namelen = 0; 1063 1064 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from); 1065 msg.msg_iov = &aiov; 1066 msg.msg_iovlen = 1; 1067 aiov.iov_base = PTRIN(args->buf); 1068 aiov.iov_len = args->len; 1069 msg.msg_control = 0; 1070 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 1071 1072 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL); 1073 if (error != 0) 1074 return (error); 1075 1076 if (PTRIN(args->from) != NULL) { 1077 error = bsd_to_linux_sockaddr((struct sockaddr *) 1078 PTRIN(args->from)); 1079 if (error != 0) 1080 return (error); 1081 1082 error = linux_sa_put((struct osockaddr *) 1083 PTRIN(args->from)); 1084 } 1085 1086 if (PTRIN(args->fromlen) != NULL) 1087 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 1088 sizeof(msg.msg_namelen)); 1089 1090 return (error); 1091 } 1092 1093 static int 1094 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1095 l_uint flags) 1096 { 1097 struct cmsghdr *cmsg; 1098 struct cmsgcred cmcred; 1099 struct mbuf *control; 1100 struct msghdr msg; 1101 struct l_cmsghdr linux_cmsg; 1102 struct l_cmsghdr *ptr_cmsg; 1103 struct l_msghdr linux_msg; 1104 struct iovec *iov; 1105 socklen_t datalen; 1106 struct sockaddr *sa; 1107 sa_family_t sa_family; 1108 void *data; 1109 int error; 1110 1111 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1112 if (error != 0) 1113 return (error); 1114 1115 /* 1116 * Some Linux applications (ping) define a non-NULL control data 1117 * pointer, but a msg_controllen of 0, which is not allowed in the 1118 * FreeBSD system call interface. NULL the msg_control pointer in 1119 * order to handle this case. This should be checked, but allows the 1120 * Linux ping to work. 1121 */ 1122 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0) 1123 linux_msg.msg_control = PTROUT(NULL); 1124 1125 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1126 if (error != 0) 1127 return (error); 1128 1129 #ifdef COMPAT_LINUX32 1130 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1131 &iov, EMSGSIZE); 1132 #else 1133 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1134 #endif 1135 if (error != 0) 1136 return (error); 1137 1138 control = NULL; 1139 cmsg = NULL; 1140 1141 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) { 1142 error = kern_getsockname(td, s, &sa, &datalen); 1143 if (error != 0) 1144 goto bad; 1145 sa_family = sa->sa_family; 1146 free(sa, M_SONAME); 1147 1148 error = ENOBUFS; 1149 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO); 1150 control = m_get(M_WAITOK, MT_CONTROL); 1151 1152 do { 1153 error = copyin(ptr_cmsg, &linux_cmsg, 1154 sizeof(struct l_cmsghdr)); 1155 if (error != 0) 1156 goto bad; 1157 1158 error = EINVAL; 1159 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1160 goto bad; 1161 1162 /* 1163 * Now we support only SCM_RIGHTS and SCM_CRED, 1164 * so return EINVAL in any other cmsg_type 1165 */ 1166 cmsg->cmsg_type = 1167 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1168 cmsg->cmsg_level = 1169 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1170 if (cmsg->cmsg_type == -1 1171 || cmsg->cmsg_level != SOL_SOCKET) 1172 goto bad; 1173 1174 /* 1175 * Some applications (e.g. pulseaudio) attempt to 1176 * send ancillary data even if the underlying protocol 1177 * doesn't support it which is not allowed in the 1178 * FreeBSD system call interface. 1179 */ 1180 if (sa_family != AF_UNIX) 1181 continue; 1182 1183 data = LINUX_CMSG_DATA(ptr_cmsg); 1184 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1185 1186 switch (cmsg->cmsg_type) 1187 { 1188 case SCM_RIGHTS: 1189 break; 1190 1191 case SCM_CREDS: 1192 data = &cmcred; 1193 datalen = sizeof(cmcred); 1194 1195 /* 1196 * The lower levels will fill in the structure 1197 */ 1198 bzero(data, datalen); 1199 break; 1200 } 1201 1202 cmsg->cmsg_len = CMSG_LEN(datalen); 1203 1204 error = ENOBUFS; 1205 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg)) 1206 goto bad; 1207 if (!m_append(control, datalen, (c_caddr_t)data)) 1208 goto bad; 1209 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg))); 1210 1211 if (m_length(control, NULL) == 0) { 1212 m_freem(control); 1213 control = NULL; 1214 } 1215 } 1216 1217 msg.msg_iov = iov; 1218 msg.msg_flags = 0; 1219 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1220 control = NULL; 1221 1222 bad: 1223 m_freem(control); 1224 free(iov, M_IOV); 1225 if (cmsg) 1226 free(cmsg, M_LINUX); 1227 return (error); 1228 } 1229 1230 int 1231 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1232 { 1233 1234 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1235 args->flags)); 1236 } 1237 1238 int 1239 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1240 { 1241 struct l_mmsghdr *msg; 1242 l_uint retval; 1243 int error, datagrams; 1244 1245 if (args->vlen > UIO_MAXIOV) 1246 args->vlen = UIO_MAXIOV; 1247 1248 msg = PTRIN(args->msg); 1249 datagrams = 0; 1250 while (datagrams < args->vlen) { 1251 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1252 args->flags); 1253 if (error != 0) 1254 break; 1255 1256 retval = td->td_retval[0]; 1257 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1258 if (error != 0) 1259 break; 1260 ++msg; 1261 ++datagrams; 1262 } 1263 if (error == 0) 1264 td->td_retval[0] = datagrams; 1265 return (error); 1266 } 1267 1268 static int 1269 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1270 l_uint flags, struct msghdr *msg) 1271 { 1272 struct cmsghdr *cm; 1273 struct cmsgcred *cmcred; 1274 struct l_cmsghdr *linux_cmsg = NULL; 1275 struct l_ucred linux_ucred; 1276 socklen_t datalen, outlen; 1277 struct l_msghdr linux_msg; 1278 struct iovec *iov, *uiov; 1279 struct mbuf *control = NULL; 1280 struct mbuf **controlp; 1281 struct timeval *ftmvl; 1282 l_timeval ltmvl; 1283 caddr_t outbuf; 1284 void *data; 1285 int error, i, fd, fds, *fdp; 1286 1287 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1288 if (error != 0) 1289 return (error); 1290 1291 error = linux_to_bsd_msghdr(msg, &linux_msg); 1292 if (error != 0) 1293 return (error); 1294 1295 #ifdef COMPAT_LINUX32 1296 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1297 &iov, EMSGSIZE); 1298 #else 1299 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1300 #endif 1301 if (error != 0) 1302 return (error); 1303 1304 if (msg->msg_name) { 1305 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name, 1306 msg->msg_namelen); 1307 if (error != 0) 1308 goto bad; 1309 } 1310 1311 uiov = msg->msg_iov; 1312 msg->msg_iov = iov; 1313 controlp = (msg->msg_control != NULL) ? &control : NULL; 1314 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp); 1315 msg->msg_iov = uiov; 1316 if (error != 0) 1317 goto bad; 1318 1319 error = bsd_to_linux_msghdr(msg, &linux_msg); 1320 if (error != 0) 1321 goto bad; 1322 1323 if (linux_msg.msg_name) { 1324 error = bsd_to_linux_sockaddr((struct sockaddr *) 1325 PTRIN(linux_msg.msg_name)); 1326 if (error != 0) 1327 goto bad; 1328 } 1329 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1330 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1331 if (error != 0) 1332 goto bad; 1333 } 1334 1335 outbuf = PTRIN(linux_msg.msg_control); 1336 outlen = 0; 1337 1338 if (control) { 1339 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1340 1341 msg->msg_control = mtod(control, struct cmsghdr *); 1342 msg->msg_controllen = control->m_len; 1343 1344 cm = CMSG_FIRSTHDR(msg); 1345 1346 while (cm != NULL) { 1347 linux_cmsg->cmsg_type = 1348 bsd_to_linux_cmsg_type(cm->cmsg_type); 1349 linux_cmsg->cmsg_level = 1350 bsd_to_linux_sockopt_level(cm->cmsg_level); 1351 if (linux_cmsg->cmsg_type == -1 1352 || cm->cmsg_level != SOL_SOCKET) 1353 { 1354 error = EINVAL; 1355 goto bad; 1356 } 1357 1358 data = CMSG_DATA(cm); 1359 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1360 1361 switch (cm->cmsg_type) 1362 { 1363 case SCM_RIGHTS: 1364 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1365 fds = datalen / sizeof(int); 1366 fdp = data; 1367 for (i = 0; i < fds; i++) { 1368 fd = *fdp++; 1369 (void)kern_fcntl(td, fd, 1370 F_SETFD, FD_CLOEXEC); 1371 } 1372 } 1373 break; 1374 1375 case SCM_CREDS: 1376 /* 1377 * Currently LOCAL_CREDS is never in 1378 * effect for Linux so no need to worry 1379 * about sockcred 1380 */ 1381 if (datalen != sizeof(*cmcred)) { 1382 error = EMSGSIZE; 1383 goto bad; 1384 } 1385 cmcred = (struct cmsgcred *)data; 1386 bzero(&linux_ucred, sizeof(linux_ucred)); 1387 linux_ucred.pid = cmcred->cmcred_pid; 1388 linux_ucred.uid = cmcred->cmcred_uid; 1389 linux_ucred.gid = cmcred->cmcred_gid; 1390 data = &linux_ucred; 1391 datalen = sizeof(linux_ucred); 1392 break; 1393 1394 case SCM_TIMESTAMP: 1395 if (datalen != sizeof(struct timeval)) { 1396 error = EMSGSIZE; 1397 goto bad; 1398 } 1399 ftmvl = (struct timeval *)data; 1400 ltmvl.tv_sec = ftmvl->tv_sec; 1401 ltmvl.tv_usec = ftmvl->tv_usec; 1402 data = <mvl; 1403 datalen = sizeof(ltmvl); 1404 break; 1405 } 1406 1407 if (outlen + LINUX_CMSG_LEN(datalen) > 1408 linux_msg.msg_controllen) { 1409 if (outlen == 0) { 1410 error = EMSGSIZE; 1411 goto bad; 1412 } else { 1413 linux_msg.msg_flags |= 1414 LINUX_MSG_CTRUNC; 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) 1423 goto bad; 1424 outbuf += L_CMSG_HDRSZ; 1425 1426 error = copyout(data, outbuf, datalen); 1427 if (error) 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 } 1436 1437 out: 1438 linux_msg.msg_controllen = outlen; 1439 error = copyout(&linux_msg, msghdr, sizeof(linux_msg)); 1440 1441 bad: 1442 free(iov, M_IOV); 1443 m_freem(control); 1444 free(linux_cmsg, M_LINUX); 1445 1446 return (error); 1447 } 1448 1449 int 1450 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1451 { 1452 struct msghdr bsd_msg; 1453 1454 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1455 args->flags, &bsd_msg)); 1456 } 1457 1458 int 1459 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1460 { 1461 struct l_mmsghdr *msg; 1462 struct msghdr bsd_msg; 1463 struct l_timespec lts; 1464 struct timespec ts, tts; 1465 l_uint retval; 1466 int error, datagrams; 1467 1468 if (args->timeout) { 1469 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1470 if (error != 0) 1471 return (error); 1472 error = linux_to_native_timespec(&ts, <s); 1473 if (error != 0) 1474 return (error); 1475 getnanotime(&tts); 1476 timespecadd(&tts, &ts); 1477 } 1478 1479 msg = PTRIN(args->msg); 1480 datagrams = 0; 1481 while (datagrams < args->vlen) { 1482 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1483 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1484 if (error != 0) 1485 break; 1486 1487 retval = td->td_retval[0]; 1488 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1489 if (error != 0) 1490 break; 1491 ++msg; 1492 ++datagrams; 1493 1494 /* 1495 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1496 */ 1497 if (args->flags & LINUX_MSG_WAITFORONE) 1498 args->flags |= LINUX_MSG_DONTWAIT; 1499 1500 /* 1501 * See BUGS section of recvmmsg(2). 1502 */ 1503 if (args->timeout) { 1504 getnanotime(&ts); 1505 timespecsub(&ts, &tts); 1506 if (!timespecisset(&ts) || ts.tv_sec > 0) 1507 break; 1508 } 1509 /* Out of band data, return right away. */ 1510 if (bsd_msg.msg_flags & MSG_OOB) 1511 break; 1512 } 1513 if (error == 0) 1514 td->td_retval[0] = datagrams; 1515 return (error); 1516 } 1517 1518 int 1519 linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1520 { 1521 struct shutdown_args /* { 1522 int s; 1523 int how; 1524 } */ bsd_args; 1525 1526 bsd_args.s = args->s; 1527 bsd_args.how = args->how; 1528 return (sys_shutdown(td, &bsd_args)); 1529 } 1530 1531 int 1532 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1533 { 1534 struct setsockopt_args /* { 1535 int s; 1536 int level; 1537 int name; 1538 caddr_t val; 1539 int valsize; 1540 } */ bsd_args; 1541 l_timeval linux_tv; 1542 struct timeval tv; 1543 int error, name; 1544 1545 bsd_args.s = args->s; 1546 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1547 switch (bsd_args.level) { 1548 case SOL_SOCKET: 1549 name = linux_to_bsd_so_sockopt(args->optname); 1550 switch (name) { 1551 case SO_RCVTIMEO: 1552 /* FALLTHROUGH */ 1553 case SO_SNDTIMEO: 1554 error = copyin(PTRIN(args->optval), &linux_tv, 1555 sizeof(linux_tv)); 1556 if (error) 1557 return (error); 1558 tv.tv_sec = linux_tv.tv_sec; 1559 tv.tv_usec = linux_tv.tv_usec; 1560 return (kern_setsockopt(td, args->s, bsd_args.level, 1561 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1562 /* NOTREACHED */ 1563 break; 1564 default: 1565 break; 1566 } 1567 break; 1568 case IPPROTO_IP: 1569 name = linux_to_bsd_ip_sockopt(args->optname); 1570 break; 1571 case IPPROTO_IPV6: 1572 name = linux_to_bsd_ip6_sockopt(args->optname); 1573 break; 1574 case IPPROTO_TCP: 1575 name = linux_to_bsd_tcp_sockopt(args->optname); 1576 break; 1577 default: 1578 name = -1; 1579 break; 1580 } 1581 if (name == -1) 1582 return (ENOPROTOOPT); 1583 1584 bsd_args.name = name; 1585 bsd_args.val = PTRIN(args->optval); 1586 bsd_args.valsize = args->optlen; 1587 1588 if (name == IPV6_NEXTHOP) { 1589 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1590 bsd_args.valsize); 1591 error = sys_setsockopt(td, &bsd_args); 1592 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1593 } else 1594 error = sys_setsockopt(td, &bsd_args); 1595 1596 return (error); 1597 } 1598 1599 int 1600 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1601 { 1602 struct getsockopt_args /* { 1603 int s; 1604 int level; 1605 int name; 1606 caddr_t val; 1607 int *avalsize; 1608 } */ bsd_args; 1609 l_timeval linux_tv; 1610 struct timeval tv; 1611 socklen_t tv_len, xulen; 1612 struct xucred xu; 1613 struct l_ucred lxu; 1614 int error, name; 1615 1616 bsd_args.s = args->s; 1617 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1618 switch (bsd_args.level) { 1619 case SOL_SOCKET: 1620 name = linux_to_bsd_so_sockopt(args->optname); 1621 switch (name) { 1622 case SO_RCVTIMEO: 1623 /* FALLTHROUGH */ 1624 case SO_SNDTIMEO: 1625 tv_len = sizeof(tv); 1626 error = kern_getsockopt(td, args->s, bsd_args.level, 1627 name, &tv, UIO_SYSSPACE, &tv_len); 1628 if (error) 1629 return (error); 1630 linux_tv.tv_sec = tv.tv_sec; 1631 linux_tv.tv_usec = tv.tv_usec; 1632 return (copyout(&linux_tv, PTRIN(args->optval), 1633 sizeof(linux_tv))); 1634 /* NOTREACHED */ 1635 break; 1636 case LOCAL_PEERCRED: 1637 if (args->optlen != sizeof(lxu)) 1638 return (EINVAL); 1639 xulen = sizeof(xu); 1640 error = kern_getsockopt(td, args->s, bsd_args.level, 1641 name, &xu, UIO_SYSSPACE, &xulen); 1642 if (error) 1643 return (error); 1644 /* 1645 * XXX Use 0 for pid as the FreeBSD does not cache peer pid. 1646 */ 1647 lxu.pid = 0; 1648 lxu.uid = xu.cr_uid; 1649 lxu.gid = xu.cr_gid; 1650 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1651 /* NOTREACHED */ 1652 break; 1653 default: 1654 break; 1655 } 1656 break; 1657 case IPPROTO_IP: 1658 name = linux_to_bsd_ip_sockopt(args->optname); 1659 break; 1660 case IPPROTO_IPV6: 1661 name = linux_to_bsd_ip6_sockopt(args->optname); 1662 break; 1663 case IPPROTO_TCP: 1664 name = linux_to_bsd_tcp_sockopt(args->optname); 1665 break; 1666 default: 1667 name = -1; 1668 break; 1669 } 1670 if (name == -1) 1671 return (EINVAL); 1672 1673 bsd_args.name = name; 1674 bsd_args.val = PTRIN(args->optval); 1675 bsd_args.avalsize = PTRIN(args->optlen); 1676 1677 if (name == IPV6_NEXTHOP) { 1678 error = sys_getsockopt(td, &bsd_args); 1679 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1680 } else 1681 error = sys_getsockopt(td, &bsd_args); 1682 1683 return (error); 1684 } 1685 1686 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1687 1688 /* Argument list sizes for linux_socketcall */ 1689 1690 #define LINUX_AL(x) ((x) * sizeof(l_ulong)) 1691 1692 static const unsigned char lxs_args[] = { 1693 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */, 1694 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */, 1695 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */, 1696 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */, 1697 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */, 1698 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */, 1699 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */, 1700 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */, 1701 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */, 1702 LINUX_AL(4) /* accept4 */, LINUX_AL(5) /* recvmmsg */, 1703 LINUX_AL(4) /* sendmmsg */ 1704 }; 1705 1706 #define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1 1707 1708 int 1709 linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1710 { 1711 l_ulong a[6]; 1712 void *arg; 1713 int error; 1714 1715 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE) 1716 return (EINVAL); 1717 error = copyin(PTRIN(args->args), a, lxs_args[args->what]); 1718 if (error) 1719 return (error); 1720 1721 arg = a; 1722 switch (args->what) { 1723 case LINUX_SOCKET: 1724 return (linux_socket(td, arg)); 1725 case LINUX_BIND: 1726 return (linux_bind(td, arg)); 1727 case LINUX_CONNECT: 1728 return (linux_connect(td, arg)); 1729 case LINUX_LISTEN: 1730 return (linux_listen(td, arg)); 1731 case LINUX_ACCEPT: 1732 return (linux_accept(td, arg)); 1733 case LINUX_GETSOCKNAME: 1734 return (linux_getsockname(td, arg)); 1735 case LINUX_GETPEERNAME: 1736 return (linux_getpeername(td, arg)); 1737 case LINUX_SOCKETPAIR: 1738 return (linux_socketpair(td, arg)); 1739 case LINUX_SEND: 1740 return (linux_send(td, arg)); 1741 case LINUX_RECV: 1742 return (linux_recv(td, arg)); 1743 case LINUX_SENDTO: 1744 return (linux_sendto(td, arg)); 1745 case LINUX_RECVFROM: 1746 return (linux_recvfrom(td, arg)); 1747 case LINUX_SHUTDOWN: 1748 return (linux_shutdown(td, arg)); 1749 case LINUX_SETSOCKOPT: 1750 return (linux_setsockopt(td, arg)); 1751 case LINUX_GETSOCKOPT: 1752 return (linux_getsockopt(td, arg)); 1753 case LINUX_SENDMSG: 1754 return (linux_sendmsg(td, arg)); 1755 case LINUX_RECVMSG: 1756 return (linux_recvmsg(td, arg)); 1757 case LINUX_ACCEPT4: 1758 return (linux_accept4(td, arg)); 1759 case LINUX_RECVMMSG: 1760 return (linux_recvmmsg(td, arg)); 1761 case LINUX_SENDMMSG: 1762 return (linux_sendmmsg(td, arg)); 1763 } 1764 1765 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1766 return (ENOSYS); 1767 } 1768 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1769