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, 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) 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) 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) 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) 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 cap_rights_t rights; 770 struct socket *so; 771 struct sockaddr *sa; 772 struct file *fp; 773 u_int fflag; 774 int error; 775 776 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name), 777 args->namelen); 778 if (error) 779 return (error); 780 781 error = kern_connectat(td, AT_FDCWD, args->s, sa); 782 free(sa, M_SONAME); 783 if (error != EISCONN) 784 return (error); 785 786 /* 787 * Linux doesn't return EISCONN the first time it occurs, 788 * when on a non-blocking socket. Instead it returns the 789 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 790 */ 791 error = getsock_cap(td, args->s, cap_rights_init(&rights, CAP_CONNECT), 792 &fp, &fflag, NULL); 793 if (error != 0) 794 return (error); 795 796 error = EISCONN; 797 so = fp->f_data; 798 if (fflag & FNONBLOCK) { 799 SOCK_LOCK(so); 800 if (so->so_emuldata == 0) 801 error = so->so_error; 802 so->so_emuldata = (void *)1; 803 SOCK_UNLOCK(so); 804 } 805 fdrop(fp, td); 806 807 return (error); 808 } 809 810 int 811 linux_listen(struct thread *td, struct linux_listen_args *args) 812 { 813 814 return (kern_listen(td, args->s, args->backlog)); 815 } 816 817 static int 818 linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 819 l_uintptr_t namelen, int flags) 820 { 821 struct accept4_args /* { 822 int s; 823 struct sockaddr * __restrict name; 824 socklen_t * __restrict anamelen; 825 int flags; 826 } */ bsd_args; 827 cap_rights_t rights; 828 struct socket *so; 829 struct file *fp; 830 int error, error1; 831 832 bsd_args.s = s; 833 /* XXX: */ 834 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr); 835 bsd_args.anamelen = PTRIN(namelen);/* XXX */ 836 bsd_args.flags = 0; 837 error = linux_set_socket_flags(flags, &bsd_args.flags); 838 if (error != 0) 839 return (error); 840 error = sys_accept4(td, &bsd_args); 841 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name); 842 if (error) { 843 if (error == EFAULT && namelen != sizeof(struct sockaddr_in)) 844 return (EINVAL); 845 if (error == EINVAL) { 846 error1 = getsock_cap(td, s, 847 cap_rights_init(&rights, CAP_ACCEPT), &fp, NULL, NULL); 848 if (error1 != 0) 849 return (error1); 850 so = fp->f_data; 851 if (so->so_type == SOCK_DGRAM) { 852 fdrop(fp, td); 853 return (EOPNOTSUPP); 854 } 855 fdrop(fp, td); 856 } 857 return (error); 858 } 859 if (addr) 860 error = linux_sa_put(PTRIN(addr)); 861 if (error) { 862 (void)kern_close(td, td->td_retval[0]); 863 td->td_retval[0] = 0; 864 } 865 return (error); 866 } 867 868 int 869 linux_accept(struct thread *td, struct linux_accept_args *args) 870 { 871 872 return (linux_accept_common(td, args->s, args->addr, 873 args->namelen, 0)); 874 } 875 876 int 877 linux_accept4(struct thread *td, struct linux_accept4_args *args) 878 { 879 880 return (linux_accept_common(td, args->s, args->addr, 881 args->namelen, args->flags)); 882 } 883 884 int 885 linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 886 { 887 struct getsockname_args /* { 888 int fdes; 889 struct sockaddr * __restrict asa; 890 socklen_t * __restrict alen; 891 } */ bsd_args; 892 int error; 893 894 bsd_args.fdes = args->s; 895 /* XXX: */ 896 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr); 897 bsd_args.alen = PTRIN(args->namelen); /* XXX */ 898 error = sys_getsockname(td, &bsd_args); 899 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 900 if (error) 901 return (error); 902 return (linux_sa_put(PTRIN(args->addr))); 903 } 904 905 int 906 linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 907 { 908 struct getpeername_args /* { 909 int fdes; 910 caddr_t asa; 911 int *alen; 912 } */ bsd_args; 913 int error; 914 915 bsd_args.fdes = args->s; 916 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr); 917 bsd_args.alen = (socklen_t *)PTRIN(args->namelen); 918 error = sys_getpeername(td, &bsd_args); 919 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa); 920 if (error) 921 return (error); 922 return (linux_sa_put(PTRIN(args->addr))); 923 } 924 925 int 926 linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 927 { 928 struct socketpair_args /* { 929 int domain; 930 int type; 931 int protocol; 932 int *rsv; 933 } */ bsd_args; 934 int error; 935 936 bsd_args.domain = linux_to_bsd_domain(args->domain); 937 if (bsd_args.domain != PF_LOCAL) 938 return (EAFNOSUPPORT); 939 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK; 940 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX) 941 return (EINVAL); 942 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 943 &bsd_args.type); 944 if (error != 0) 945 return (error); 946 if (args->protocol != 0 && args->protocol != PF_UNIX) 947 948 /* 949 * Use of PF_UNIX as protocol argument is not right, 950 * but Linux does it. 951 * Do not map PF_UNIX as its Linux value is identical 952 * to FreeBSD one. 953 */ 954 return (EPROTONOSUPPORT); 955 else 956 bsd_args.protocol = 0; 957 bsd_args.rsv = (int *)PTRIN(args->rsv); 958 return (sys_socketpair(td, &bsd_args)); 959 } 960 961 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 962 struct linux_send_args { 963 register_t s; 964 register_t msg; 965 register_t len; 966 register_t flags; 967 }; 968 969 static int 970 linux_send(struct thread *td, struct linux_send_args *args) 971 { 972 struct sendto_args /* { 973 int s; 974 caddr_t buf; 975 int len; 976 int flags; 977 caddr_t to; 978 int tolen; 979 } */ bsd_args; 980 981 bsd_args.s = args->s; 982 bsd_args.buf = (caddr_t)PTRIN(args->msg); 983 bsd_args.len = args->len; 984 bsd_args.flags = args->flags; 985 bsd_args.to = NULL; 986 bsd_args.tolen = 0; 987 return (sys_sendto(td, &bsd_args)); 988 } 989 990 struct linux_recv_args { 991 register_t s; 992 register_t msg; 993 register_t len; 994 register_t flags; 995 }; 996 997 static int 998 linux_recv(struct thread *td, struct linux_recv_args *args) 999 { 1000 struct recvfrom_args /* { 1001 int s; 1002 caddr_t buf; 1003 int len; 1004 int flags; 1005 struct sockaddr *from; 1006 socklen_t fromlenaddr; 1007 } */ bsd_args; 1008 1009 bsd_args.s = args->s; 1010 bsd_args.buf = (caddr_t)PTRIN(args->msg); 1011 bsd_args.len = args->len; 1012 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 1013 bsd_args.from = NULL; 1014 bsd_args.fromlenaddr = 0; 1015 return (sys_recvfrom(td, &bsd_args)); 1016 } 1017 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1018 1019 int 1020 linux_sendto(struct thread *td, struct linux_sendto_args *args) 1021 { 1022 struct msghdr msg; 1023 struct iovec aiov; 1024 1025 if (linux_check_hdrincl(td, args->s) == 0) 1026 /* IP_HDRINCL set, tweak the packet before sending */ 1027 return (linux_sendto_hdrincl(td, args)); 1028 1029 msg.msg_name = PTRIN(args->to); 1030 msg.msg_namelen = args->tolen; 1031 msg.msg_iov = &aiov; 1032 msg.msg_iovlen = 1; 1033 msg.msg_control = NULL; 1034 msg.msg_flags = 0; 1035 aiov.iov_base = PTRIN(args->msg); 1036 aiov.iov_len = args->len; 1037 return (linux_sendit(td, args->s, &msg, args->flags, NULL, 1038 UIO_USERSPACE)); 1039 } 1040 1041 int 1042 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 1043 { 1044 struct msghdr msg; 1045 struct iovec aiov; 1046 int error, fromlen; 1047 1048 if (PTRIN(args->fromlen) != NULL) { 1049 error = copyin(PTRIN(args->fromlen), &fromlen, 1050 sizeof(fromlen)); 1051 if (error != 0) 1052 return (error); 1053 if (fromlen < 0) 1054 return (EINVAL); 1055 msg.msg_namelen = fromlen; 1056 } else 1057 msg.msg_namelen = 0; 1058 1059 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from); 1060 msg.msg_iov = &aiov; 1061 msg.msg_iovlen = 1; 1062 aiov.iov_base = PTRIN(args->buf); 1063 aiov.iov_len = args->len; 1064 msg.msg_control = 0; 1065 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 1066 1067 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL); 1068 if (error != 0) 1069 return (error); 1070 1071 if (PTRIN(args->from) != NULL) { 1072 error = bsd_to_linux_sockaddr((struct sockaddr *) 1073 PTRIN(args->from)); 1074 if (error != 0) 1075 return (error); 1076 1077 error = linux_sa_put((struct osockaddr *) 1078 PTRIN(args->from)); 1079 } 1080 1081 if (PTRIN(args->fromlen) != NULL) 1082 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 1083 sizeof(msg.msg_namelen)); 1084 1085 return (error); 1086 } 1087 1088 static int 1089 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1090 l_uint flags) 1091 { 1092 struct cmsghdr *cmsg; 1093 struct cmsgcred cmcred; 1094 struct mbuf *control; 1095 struct msghdr msg; 1096 struct l_cmsghdr linux_cmsg; 1097 struct l_cmsghdr *ptr_cmsg; 1098 struct l_msghdr linux_msg; 1099 struct iovec *iov; 1100 socklen_t datalen; 1101 struct sockaddr *sa; 1102 sa_family_t sa_family; 1103 void *data; 1104 int error; 1105 1106 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1107 if (error != 0) 1108 return (error); 1109 1110 /* 1111 * Some Linux applications (ping) define a non-NULL control data 1112 * pointer, but a msg_controllen of 0, which is not allowed in the 1113 * FreeBSD system call interface. NULL the msg_control pointer in 1114 * order to handle this case. This should be checked, but allows the 1115 * Linux ping to work. 1116 */ 1117 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0) 1118 linux_msg.msg_control = PTROUT(NULL); 1119 1120 error = linux_to_bsd_msghdr(&msg, &linux_msg); 1121 if (error != 0) 1122 return (error); 1123 1124 #ifdef COMPAT_LINUX32 1125 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1126 &iov, EMSGSIZE); 1127 #else 1128 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1129 #endif 1130 if (error != 0) 1131 return (error); 1132 1133 control = NULL; 1134 cmsg = NULL; 1135 1136 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) { 1137 error = kern_getsockname(td, s, &sa, &datalen); 1138 if (error != 0) 1139 goto bad; 1140 sa_family = sa->sa_family; 1141 free(sa, M_SONAME); 1142 1143 error = ENOBUFS; 1144 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO); 1145 control = m_get(M_WAITOK, MT_CONTROL); 1146 1147 do { 1148 error = copyin(ptr_cmsg, &linux_cmsg, 1149 sizeof(struct l_cmsghdr)); 1150 if (error != 0) 1151 goto bad; 1152 1153 error = EINVAL; 1154 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr)) 1155 goto bad; 1156 1157 /* 1158 * Now we support only SCM_RIGHTS and SCM_CRED, 1159 * so return EINVAL in any other cmsg_type 1160 */ 1161 cmsg->cmsg_type = 1162 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1163 cmsg->cmsg_level = 1164 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1165 if (cmsg->cmsg_type == -1 1166 || cmsg->cmsg_level != SOL_SOCKET) 1167 goto bad; 1168 1169 /* 1170 * Some applications (e.g. pulseaudio) attempt to 1171 * send ancillary data even if the underlying protocol 1172 * doesn't support it which is not allowed in the 1173 * FreeBSD system call interface. 1174 */ 1175 if (sa_family != AF_UNIX) 1176 continue; 1177 1178 data = LINUX_CMSG_DATA(ptr_cmsg); 1179 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1180 1181 switch (cmsg->cmsg_type) 1182 { 1183 case SCM_RIGHTS: 1184 break; 1185 1186 case SCM_CREDS: 1187 data = &cmcred; 1188 datalen = sizeof(cmcred); 1189 1190 /* 1191 * The lower levels will fill in the structure 1192 */ 1193 bzero(data, datalen); 1194 break; 1195 } 1196 1197 cmsg->cmsg_len = CMSG_LEN(datalen); 1198 1199 error = ENOBUFS; 1200 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg)) 1201 goto bad; 1202 if (!m_append(control, datalen, (c_caddr_t)data)) 1203 goto bad; 1204 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg))); 1205 1206 if (m_length(control, NULL) == 0) { 1207 m_freem(control); 1208 control = NULL; 1209 } 1210 } 1211 1212 msg.msg_iov = iov; 1213 msg.msg_flags = 0; 1214 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1215 control = NULL; 1216 1217 bad: 1218 m_freem(control); 1219 free(iov, M_IOV); 1220 if (cmsg) 1221 free(cmsg, M_LINUX); 1222 return (error); 1223 } 1224 1225 int 1226 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1227 { 1228 1229 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1230 args->flags)); 1231 } 1232 1233 int 1234 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1235 { 1236 struct l_mmsghdr *msg; 1237 l_uint retval; 1238 int error, datagrams; 1239 1240 if (args->vlen > UIO_MAXIOV) 1241 args->vlen = UIO_MAXIOV; 1242 1243 msg = PTRIN(args->msg); 1244 datagrams = 0; 1245 while (datagrams < args->vlen) { 1246 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1247 args->flags); 1248 if (error != 0) 1249 break; 1250 1251 retval = td->td_retval[0]; 1252 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1253 if (error != 0) 1254 break; 1255 ++msg; 1256 ++datagrams; 1257 } 1258 if (error == 0) 1259 td->td_retval[0] = datagrams; 1260 return (error); 1261 } 1262 1263 static int 1264 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1265 l_uint flags, struct msghdr *msg) 1266 { 1267 struct cmsghdr *cm; 1268 struct cmsgcred *cmcred; 1269 struct l_cmsghdr *linux_cmsg = NULL; 1270 struct l_ucred linux_ucred; 1271 socklen_t datalen, outlen; 1272 struct l_msghdr linux_msg; 1273 struct iovec *iov, *uiov; 1274 struct mbuf *control = NULL; 1275 struct mbuf **controlp; 1276 struct timeval *ftmvl; 1277 l_timeval ltmvl; 1278 caddr_t outbuf; 1279 void *data; 1280 int error, i, fd, fds, *fdp; 1281 1282 error = copyin(msghdr, &linux_msg, sizeof(linux_msg)); 1283 if (error != 0) 1284 return (error); 1285 1286 error = linux_to_bsd_msghdr(msg, &linux_msg); 1287 if (error != 0) 1288 return (error); 1289 1290 #ifdef COMPAT_LINUX32 1291 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1292 &iov, EMSGSIZE); 1293 #else 1294 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1295 #endif 1296 if (error != 0) 1297 return (error); 1298 1299 if (msg->msg_name) { 1300 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name, 1301 msg->msg_namelen); 1302 if (error != 0) 1303 goto bad; 1304 } 1305 1306 uiov = msg->msg_iov; 1307 msg->msg_iov = iov; 1308 controlp = (msg->msg_control != NULL) ? &control : NULL; 1309 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp); 1310 msg->msg_iov = uiov; 1311 if (error != 0) 1312 goto bad; 1313 1314 error = bsd_to_linux_msghdr(msg, &linux_msg); 1315 if (error != 0) 1316 goto bad; 1317 1318 if (linux_msg.msg_name) { 1319 error = bsd_to_linux_sockaddr((struct sockaddr *) 1320 PTRIN(linux_msg.msg_name)); 1321 if (error != 0) 1322 goto bad; 1323 } 1324 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) { 1325 error = linux_sa_put(PTRIN(linux_msg.msg_name)); 1326 if (error != 0) 1327 goto bad; 1328 } 1329 1330 outbuf = PTRIN(linux_msg.msg_control); 1331 outlen = 0; 1332 1333 if (control) { 1334 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1335 1336 msg->msg_control = mtod(control, struct cmsghdr *); 1337 msg->msg_controllen = control->m_len; 1338 1339 cm = CMSG_FIRSTHDR(msg); 1340 1341 while (cm != NULL) { 1342 linux_cmsg->cmsg_type = 1343 bsd_to_linux_cmsg_type(cm->cmsg_type); 1344 linux_cmsg->cmsg_level = 1345 bsd_to_linux_sockopt_level(cm->cmsg_level); 1346 if (linux_cmsg->cmsg_type == -1 1347 || cm->cmsg_level != SOL_SOCKET) 1348 { 1349 error = EINVAL; 1350 goto bad; 1351 } 1352 1353 data = CMSG_DATA(cm); 1354 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1355 1356 switch (cm->cmsg_type) 1357 { 1358 case SCM_RIGHTS: 1359 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1360 fds = datalen / sizeof(int); 1361 fdp = data; 1362 for (i = 0; i < fds; i++) { 1363 fd = *fdp++; 1364 (void)kern_fcntl(td, fd, 1365 F_SETFD, FD_CLOEXEC); 1366 } 1367 } 1368 break; 1369 1370 case SCM_CREDS: 1371 /* 1372 * Currently LOCAL_CREDS is never in 1373 * effect for Linux so no need to worry 1374 * about sockcred 1375 */ 1376 if (datalen != sizeof(*cmcred)) { 1377 error = EMSGSIZE; 1378 goto bad; 1379 } 1380 cmcred = (struct cmsgcred *)data; 1381 bzero(&linux_ucred, sizeof(linux_ucred)); 1382 linux_ucred.pid = cmcred->cmcred_pid; 1383 linux_ucred.uid = cmcred->cmcred_uid; 1384 linux_ucred.gid = cmcred->cmcred_gid; 1385 data = &linux_ucred; 1386 datalen = sizeof(linux_ucred); 1387 break; 1388 1389 case SCM_TIMESTAMP: 1390 if (datalen != sizeof(struct timeval)) { 1391 error = EMSGSIZE; 1392 goto bad; 1393 } 1394 ftmvl = (struct timeval *)data; 1395 ltmvl.tv_sec = ftmvl->tv_sec; 1396 ltmvl.tv_usec = ftmvl->tv_usec; 1397 data = <mvl; 1398 datalen = sizeof(ltmvl); 1399 break; 1400 } 1401 1402 if (outlen + LINUX_CMSG_LEN(datalen) > 1403 linux_msg.msg_controllen) { 1404 if (outlen == 0) { 1405 error = EMSGSIZE; 1406 goto bad; 1407 } else { 1408 linux_msg.msg_flags |= 1409 LINUX_MSG_CTRUNC; 1410 goto out; 1411 } 1412 } 1413 1414 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1415 1416 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1417 if (error) 1418 goto bad; 1419 outbuf += L_CMSG_HDRSZ; 1420 1421 error = copyout(data, outbuf, datalen); 1422 if (error) 1423 goto bad; 1424 1425 outbuf += LINUX_CMSG_ALIGN(datalen); 1426 outlen += LINUX_CMSG_LEN(datalen); 1427 1428 cm = CMSG_NXTHDR(msg, cm); 1429 } 1430 } 1431 1432 out: 1433 linux_msg.msg_controllen = outlen; 1434 error = copyout(&linux_msg, msghdr, sizeof(linux_msg)); 1435 1436 bad: 1437 free(iov, M_IOV); 1438 m_freem(control); 1439 free(linux_cmsg, M_LINUX); 1440 1441 return (error); 1442 } 1443 1444 int 1445 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1446 { 1447 struct msghdr bsd_msg; 1448 1449 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1450 args->flags, &bsd_msg)); 1451 } 1452 1453 int 1454 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1455 { 1456 struct l_mmsghdr *msg; 1457 struct msghdr bsd_msg; 1458 struct l_timespec lts; 1459 struct timespec ts, tts; 1460 l_uint retval; 1461 int error, datagrams; 1462 1463 if (args->timeout) { 1464 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1465 if (error != 0) 1466 return (error); 1467 error = linux_to_native_timespec(&ts, <s); 1468 if (error != 0) 1469 return (error); 1470 getnanotime(&tts); 1471 timespecadd(&tts, &ts); 1472 } 1473 1474 msg = PTRIN(args->msg); 1475 datagrams = 0; 1476 while (datagrams < args->vlen) { 1477 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1478 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1479 if (error != 0) 1480 break; 1481 1482 retval = td->td_retval[0]; 1483 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1484 if (error != 0) 1485 break; 1486 ++msg; 1487 ++datagrams; 1488 1489 /* 1490 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1491 */ 1492 if (args->flags & LINUX_MSG_WAITFORONE) 1493 args->flags |= LINUX_MSG_DONTWAIT; 1494 1495 /* 1496 * See BUGS section of recvmmsg(2). 1497 */ 1498 if (args->timeout) { 1499 getnanotime(&ts); 1500 timespecsub(&ts, &tts); 1501 if (!timespecisset(&ts) || ts.tv_sec > 0) 1502 break; 1503 } 1504 /* Out of band data, return right away. */ 1505 if (bsd_msg.msg_flags & MSG_OOB) 1506 break; 1507 } 1508 if (error == 0) 1509 td->td_retval[0] = datagrams; 1510 return (error); 1511 } 1512 1513 int 1514 linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1515 { 1516 1517 return (kern_shutdown(td, args->s, args->how)); 1518 } 1519 1520 int 1521 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1522 { 1523 struct setsockopt_args /* { 1524 int s; 1525 int level; 1526 int name; 1527 caddr_t val; 1528 int valsize; 1529 } */ bsd_args; 1530 l_timeval linux_tv; 1531 struct timeval tv; 1532 int error, name; 1533 1534 bsd_args.s = args->s; 1535 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1536 switch (bsd_args.level) { 1537 case SOL_SOCKET: 1538 name = linux_to_bsd_so_sockopt(args->optname); 1539 switch (name) { 1540 case SO_RCVTIMEO: 1541 /* FALLTHROUGH */ 1542 case SO_SNDTIMEO: 1543 error = copyin(PTRIN(args->optval), &linux_tv, 1544 sizeof(linux_tv)); 1545 if (error) 1546 return (error); 1547 tv.tv_sec = linux_tv.tv_sec; 1548 tv.tv_usec = linux_tv.tv_usec; 1549 return (kern_setsockopt(td, args->s, bsd_args.level, 1550 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1551 /* NOTREACHED */ 1552 break; 1553 default: 1554 break; 1555 } 1556 break; 1557 case IPPROTO_IP: 1558 name = linux_to_bsd_ip_sockopt(args->optname); 1559 break; 1560 case IPPROTO_IPV6: 1561 name = linux_to_bsd_ip6_sockopt(args->optname); 1562 break; 1563 case IPPROTO_TCP: 1564 name = linux_to_bsd_tcp_sockopt(args->optname); 1565 break; 1566 default: 1567 name = -1; 1568 break; 1569 } 1570 if (name == -1) 1571 return (ENOPROTOOPT); 1572 1573 bsd_args.name = name; 1574 bsd_args.val = PTRIN(args->optval); 1575 bsd_args.valsize = args->optlen; 1576 1577 if (name == IPV6_NEXTHOP) { 1578 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val, 1579 bsd_args.valsize); 1580 error = sys_setsockopt(td, &bsd_args); 1581 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1582 } else 1583 error = sys_setsockopt(td, &bsd_args); 1584 1585 return (error); 1586 } 1587 1588 int 1589 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1590 { 1591 struct getsockopt_args /* { 1592 int s; 1593 int level; 1594 int name; 1595 caddr_t val; 1596 int *avalsize; 1597 } */ bsd_args; 1598 l_timeval linux_tv; 1599 struct timeval tv; 1600 socklen_t tv_len, xulen, len; 1601 struct xucred xu; 1602 struct l_ucred lxu; 1603 int error, name, newval; 1604 1605 bsd_args.s = args->s; 1606 bsd_args.level = linux_to_bsd_sockopt_level(args->level); 1607 switch (bsd_args.level) { 1608 case SOL_SOCKET: 1609 name = linux_to_bsd_so_sockopt(args->optname); 1610 switch (name) { 1611 case SO_RCVTIMEO: 1612 /* FALLTHROUGH */ 1613 case SO_SNDTIMEO: 1614 tv_len = sizeof(tv); 1615 error = kern_getsockopt(td, args->s, bsd_args.level, 1616 name, &tv, UIO_SYSSPACE, &tv_len); 1617 if (error) 1618 return (error); 1619 linux_tv.tv_sec = tv.tv_sec; 1620 linux_tv.tv_usec = tv.tv_usec; 1621 return (copyout(&linux_tv, PTRIN(args->optval), 1622 sizeof(linux_tv))); 1623 /* NOTREACHED */ 1624 break; 1625 case LOCAL_PEERCRED: 1626 if (args->optlen != sizeof(lxu)) 1627 return (EINVAL); 1628 xulen = sizeof(xu); 1629 error = kern_getsockopt(td, args->s, bsd_args.level, 1630 name, &xu, UIO_SYSSPACE, &xulen); 1631 if (error) 1632 return (error); 1633 /* 1634 * XXX Use 0 for pid as the FreeBSD does not cache peer pid. 1635 */ 1636 lxu.pid = 0; 1637 lxu.uid = xu.cr_uid; 1638 lxu.gid = xu.cr_gid; 1639 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1640 /* NOTREACHED */ 1641 break; 1642 case SO_ERROR: 1643 len = sizeof(newval); 1644 error = kern_getsockopt(td, args->s, bsd_args.level, 1645 name, &newval, UIO_SYSSPACE, &len); 1646 if (error) 1647 return (error); 1648 newval = -SV_ABI_ERRNO(td->td_proc, newval); 1649 return (copyout(&newval, PTRIN(args->optval), len)); 1650 /* NOTREACHED */ 1651 default: 1652 break; 1653 } 1654 break; 1655 case IPPROTO_IP: 1656 name = linux_to_bsd_ip_sockopt(args->optname); 1657 break; 1658 case IPPROTO_IPV6: 1659 name = linux_to_bsd_ip6_sockopt(args->optname); 1660 break; 1661 case IPPROTO_TCP: 1662 name = linux_to_bsd_tcp_sockopt(args->optname); 1663 break; 1664 default: 1665 name = -1; 1666 break; 1667 } 1668 if (name == -1) 1669 return (EINVAL); 1670 1671 bsd_args.name = name; 1672 bsd_args.val = PTRIN(args->optval); 1673 bsd_args.avalsize = PTRIN(args->optlen); 1674 1675 if (name == IPV6_NEXTHOP) { 1676 error = sys_getsockopt(td, &bsd_args); 1677 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val); 1678 } else 1679 error = sys_getsockopt(td, &bsd_args); 1680 1681 return (error); 1682 } 1683 1684 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1685 1686 /* Argument list sizes for linux_socketcall */ 1687 static const unsigned char lxs_args_cnt[] = { 1688 0 /* unused*/, 3 /* socket */, 1689 3 /* bind */, 3 /* connect */, 1690 2 /* listen */, 3 /* accept */, 1691 3 /* getsockname */, 3 /* getpeername */, 1692 4 /* socketpair */, 4 /* send */, 1693 4 /* recv */, 6 /* sendto */, 1694 6 /* recvfrom */, 2 /* shutdown */, 1695 5 /* setsockopt */, 5 /* getsockopt */, 1696 3 /* sendmsg */, 3 /* recvmsg */, 1697 4 /* accept4 */, 5 /* recvmmsg */, 1698 4 /* sendmmsg */ 1699 }; 1700 #define LINUX_ARGS_CNT (nitems(lxs_args_cnt) - 1) 1701 #define LINUX_ARG_SIZE(x) (lxs_args_cnt[x] * sizeof(l_ulong)) 1702 1703 int 1704 linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 1705 { 1706 l_ulong a[6]; 1707 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1708 register_t l_args[6]; 1709 #endif 1710 void *arg; 1711 int error; 1712 1713 if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT) 1714 return (EINVAL); 1715 error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what)); 1716 if (error != 0) 1717 return (error); 1718 1719 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1720 for (int i = 0; i < lxs_args_cnt[args->what]; ++i) 1721 l_args[i] = a[i]; 1722 arg = l_args; 1723 #else 1724 arg = a; 1725 #endif 1726 switch (args->what) { 1727 case LINUX_SOCKET: 1728 return (linux_socket(td, arg)); 1729 case LINUX_BIND: 1730 return (linux_bind(td, arg)); 1731 case LINUX_CONNECT: 1732 return (linux_connect(td, arg)); 1733 case LINUX_LISTEN: 1734 return (linux_listen(td, arg)); 1735 case LINUX_ACCEPT: 1736 return (linux_accept(td, arg)); 1737 case LINUX_GETSOCKNAME: 1738 return (linux_getsockname(td, arg)); 1739 case LINUX_GETPEERNAME: 1740 return (linux_getpeername(td, arg)); 1741 case LINUX_SOCKETPAIR: 1742 return (linux_socketpair(td, arg)); 1743 case LINUX_SEND: 1744 return (linux_send(td, arg)); 1745 case LINUX_RECV: 1746 return (linux_recv(td, arg)); 1747 case LINUX_SENDTO: 1748 return (linux_sendto(td, arg)); 1749 case LINUX_RECVFROM: 1750 return (linux_recvfrom(td, arg)); 1751 case LINUX_SHUTDOWN: 1752 return (linux_shutdown(td, arg)); 1753 case LINUX_SETSOCKOPT: 1754 return (linux_setsockopt(td, arg)); 1755 case LINUX_GETSOCKOPT: 1756 return (linux_getsockopt(td, arg)); 1757 case LINUX_SENDMSG: 1758 return (linux_sendmsg(td, arg)); 1759 case LINUX_RECVMSG: 1760 return (linux_recvmsg(td, arg)); 1761 case LINUX_ACCEPT4: 1762 return (linux_accept4(td, arg)); 1763 case LINUX_RECVMMSG: 1764 return (linux_recvmmsg(td, arg)); 1765 case LINUX_SENDMMSG: 1766 return (linux_sendmmsg(td, arg)); 1767 } 1768 1769 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what); 1770 return (ENOSYS); 1771 } 1772 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1773