1 /* $NetBSD: linux_socket.c,v 1.164 2026/10/17 14:59:45 alnsn Exp $ */ 2 3 /*- 4 * Copyright (c) 1995, 1998, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Frank van der Linden and Eric Haszlakiewicz. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Functions in multiarch: 34 * linux_sys_socketcall : linux_socketcall.c 35 */ 36 37 #include <sys/cdefs.h> 38 __KERNEL_RCSID(0, "$NetBSD: linux_socket.c,v 1.164 2026/10/17 14:59:45 alnsn Exp $"); 39 40 #if defined(_KERNEL_OPT) 41 #include "opt_inet.h" 42 #endif /* defined(_KERNEL_OPT) */ 43 44 #include <sys/param.h> 45 #include <sys/kernel.h> 46 #include <sys/systm.h> 47 #include <sys/buf.h> 48 #include <sys/ioctl.h> 49 #include <sys/tty.h> 50 #include <sys/file.h> 51 #include <sys/filedesc.h> 52 #include <sys/select.h> 53 #include <sys/socket.h> 54 #include <sys/socketvar.h> 55 #include <sys/domain.h> 56 #include <net/if.h> 57 #include <net/if_dl.h> 58 #include <net/if_types.h> 59 #include <netinet/in.h> 60 #include <netinet/tcp.h> 61 #include <sys/mount.h> 62 #include <sys/proc.h> 63 #include <sys/vnode.h> 64 #include <sys/device.h> 65 #include <sys/protosw.h> 66 #include <sys/mbuf.h> 67 #include <sys/syslog.h> 68 #include <sys/exec.h> 69 #include <sys/kauth.h> 70 #include <sys/syscallargs.h> 71 #include <sys/ktrace.h> 72 73 #include <lib/libkern/libkern.h> 74 75 #include <netinet/ip6.h> 76 #include <netinet6/ip6_var.h> 77 78 #include <compat/sys/socket.h> 79 #include <compat/sys/sockio.h> 80 81 #include <compat/linux/common/linux_types.h> 82 #include <compat/linux/common/linux_util.h> 83 #include <compat/linux/common/linux_signal.h> 84 #include <compat/linux/common/linux_ioctl.h> 85 #include <compat/linux/common/linux_sched.h> 86 #include <compat/linux/common/linux_socket.h> 87 #include <compat/linux/common/linux_fcntl.h> 88 #if !defined(__aarch64__) && !defined(__alpha__) && !defined(__amd64__) 89 #include <compat/linux/common/linux_socketcall.h> 90 #endif 91 #include <compat/linux/common/linux_sockio.h> 92 #include <compat/linux/common/linux_ipc.h> 93 #include <compat/linux/common/linux_sem.h> 94 95 #include <compat/linux/linux_syscallargs.h> 96 97 #ifdef DEBUG_LINUX 98 #define DPRINTF(a) uprintf a 99 #else 100 #define DPRINTF(a) 101 #endif 102 103 /* 104 * The calls in this file are entered either via the linux_socketcall() 105 * interface or, on the Alpha, as individual syscalls. The 106 * linux_socketcall function does any massaging of arguments so that all 107 * the calls in here need not think that they are anything other 108 * than a normal syscall. 109 */ 110 111 static int linux_to_bsd_domain(int); 112 static int bsd_to_linux_domain(int); 113 static int linux_to_bsd_type(int); 114 int linux_to_bsd_sopt_level(int); 115 int linux_to_bsd_so_sockopt(int); 116 int linux_to_bsd_ip_sockopt(int); 117 int linux_to_bsd_ipv6_sockopt(int); 118 int linux_to_bsd_tcp_sockopt(int); 119 int linux_to_bsd_udp_sockopt(int); 120 int linux_getifname(struct lwp *, register_t *, void *); 121 int linux_getifconf(struct lwp *, register_t *, void *); 122 int linux_getifhwaddr(struct lwp *, register_t *, u_int, void *); 123 static int linux_get_sa(struct lwp *, int, struct sockaddr_big *, 124 const struct osockaddr *, socklen_t); 125 static int linux_sa_put(struct osockaddr *osa); 126 static int linux_to_bsd_msg_flags(int); 127 static int bsd_to_linux_msg_flags(int); 128 static int linux_to_bsd_msghdr(const struct linux_msghdr *, struct msghdr *); 129 static void bsd_to_linux_msghdr(const struct msghdr *, struct linux_msghdr *); 130 131 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = { 132 AF_UNSPEC, 133 AF_UNIX, 134 AF_INET, 135 AF_CCITT, /* LINUX_AF_AX25 */ 136 AF_IPX, 137 AF_APPLETALK, 138 -1, /* LINUX_AF_NETROM */ 139 -1, /* LINUX_AF_BRIDGE */ 140 -1, /* LINUX_AF_ATMPVC */ 141 AF_CCITT, /* LINUX_AF_X25 */ 142 AF_INET6, 143 -1, /* LINUX_AF_ROSE */ 144 AF_DECnet, 145 -1, /* LINUX_AF_NETBEUI */ 146 -1, /* LINUX_AF_SECURITY */ 147 pseudo_AF_KEY, 148 AF_ROUTE, /* LINUX_AF_NETLINK */ 149 -1, /* LINUX_AF_PACKET */ 150 -1, /* LINUX_AF_ASH */ 151 -1, /* LINUX_AF_ECONET */ 152 -1, /* LINUX_AF_ATMSVC */ 153 AF_SNA, 154 /* rest up to LINUX_AF_MAX-1 is not allocated */ 155 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 156 }; 157 158 static const int bsd_to_linux_domain_[AF_MAX] = { 159 LINUX_AF_UNSPEC, 160 LINUX_AF_UNIX, 161 LINUX_AF_INET, 162 -1, /* AF_IMPLINK */ 163 -1, /* AF_PUP */ 164 -1, /* AF_CHAOS */ 165 -1, /* AF_NS */ 166 -1, /* AF_ISO */ 167 -1, /* AF_ECMA */ 168 -1, /* AF_DATAKIT */ 169 LINUX_AF_AX25, /* AF_CCITT */ 170 LINUX_AF_SNA, 171 LINUX_AF_DECnet, 172 -1, /* AF_DLI */ 173 -1, /* AF_LAT */ 174 -1, /* AF_HYLINK */ 175 LINUX_AF_APPLETALK, 176 LINUX_AF_NETLINK, 177 -1, /* AF_LINK */ 178 -1, /* AF_XTP */ 179 -1, /* AF_COIP */ 180 -1, /* AF_CNT */ 181 -1, /* pseudo_AF_RTIP */ 182 LINUX_AF_IPX, 183 LINUX_AF_INET6, 184 -1, /* pseudo_AF_PIP */ 185 -1, /* AF_ISDN */ 186 -1, /* AF_NATM */ 187 -1, /* AF_ARP */ 188 LINUX_pseudo_AF_KEY, 189 -1, /* pseudo_AF_HDRCMPLT */ 190 }; 191 192 static const struct { 193 int bfl; 194 int lfl; 195 } bsd_to_linux_msg_flags_[] = { 196 {MSG_OOB, LINUX_MSG_OOB}, 197 {MSG_PEEK, LINUX_MSG_PEEK}, 198 {MSG_DONTROUTE, LINUX_MSG_DONTROUTE}, 199 {MSG_EOR, LINUX_MSG_EOR}, 200 {MSG_TRUNC, LINUX_MSG_TRUNC}, 201 {MSG_CTRUNC, LINUX_MSG_CTRUNC}, 202 {MSG_WAITALL, LINUX_MSG_WAITALL}, 203 {MSG_DONTWAIT, LINUX_MSG_DONTWAIT}, 204 {MSG_BCAST, 0}, /* not supported, clear */ 205 {MSG_MCAST, 0}, /* not supported, clear */ 206 {MSG_NOSIGNAL, LINUX_MSG_NOSIGNAL}, 207 {-1, /* not supp */ LINUX_MSG_PROBE}, 208 {-1, /* not supp */ LINUX_MSG_FIN}, 209 {-1, /* not supp */ LINUX_MSG_SYN}, 210 {-1, /* not supp */ LINUX_MSG_CONFIRM}, 211 {-1, /* not supp */ LINUX_MSG_RST}, 212 {-1, /* not supp */ LINUX_MSG_ERRQUEUE}, 213 {-1, /* not supp */ LINUX_MSG_MORE}, 214 }; 215 216 /* 217 * Convert between Linux and BSD socket domain values 218 */ 219 static int 220 linux_to_bsd_domain(int ldom) 221 { 222 if (ldom < 0 || ldom >= LINUX_AF_MAX) 223 return (-1); 224 225 return linux_to_bsd_domain_[ldom]; 226 } 227 228 /* 229 * Convert between BSD and Linux socket domain values 230 */ 231 static int 232 bsd_to_linux_domain(int bdom) 233 { 234 if (bdom < 0 || bdom >= AF_MAX) 235 return (-1); 236 237 return bsd_to_linux_domain_[bdom]; 238 } 239 240 static int 241 linux_to_bsd_type(int ltype) 242 { 243 int type, flags; 244 245 /* Real types are identical between Linux and NetBSD */ 246 type = ltype & LINUX_SOCK_TYPE_MASK; 247 248 /* But flags are not .. */ 249 flags = ltype & ~LINUX_SOCK_TYPE_MASK; 250 if (flags & ~(LINUX_SOCK_CLOEXEC|LINUX_SOCK_NONBLOCK)) 251 return -1; 252 253 if (flags & LINUX_SOCK_CLOEXEC) 254 type |= SOCK_CLOEXEC; 255 if (flags & LINUX_SOCK_NONBLOCK) 256 type |= SOCK_NONBLOCK; 257 258 return type; 259 } 260 261 static int 262 linux_to_bsd_msg_flags(int lflag) 263 { 264 int i, lfl, bfl; 265 int bflag = 0; 266 267 if (lflag == 0) 268 return (0); 269 270 for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) { 271 bfl = bsd_to_linux_msg_flags_[i].bfl; 272 lfl = bsd_to_linux_msg_flags_[i].lfl; 273 274 if (lfl == 0) 275 continue; 276 277 if (lflag & lfl) { 278 if (bfl < 0) 279 return (-1); 280 281 bflag |= bfl; 282 } 283 } 284 285 return (bflag); 286 } 287 288 static int 289 bsd_to_linux_msg_flags(int bflag) 290 { 291 int i, lfl, bfl; 292 int lflag = 0; 293 294 if (bflag == 0) 295 return (0); 296 297 for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) { 298 bfl = bsd_to_linux_msg_flags_[i].bfl; 299 lfl = bsd_to_linux_msg_flags_[i].lfl; 300 301 if (bfl <= 0) 302 continue; 303 304 if (bflag & bfl) { 305 if (lfl < 0) 306 return (-1); 307 308 lflag |= lfl; 309 } 310 } 311 312 return (lflag); 313 } 314 315 int 316 linux_sys_socket(struct lwp *l, const struct linux_sys_socket_args *uap, register_t *retval) 317 { 318 /* { 319 syscallarg(int) domain; 320 syscallarg(int) type; 321 syscallarg(int) protocol; 322 } */ 323 struct sys___socket30_args bsa; 324 int error; 325 326 memset(&bsa, 0, sizeof(bsa)); 327 SCARG(&bsa, protocol) = SCARG(uap, protocol); 328 SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain)); 329 if (SCARG(&bsa, domain) == -1) 330 return EINVAL; 331 SCARG(&bsa, type) = linux_to_bsd_type(SCARG(uap, type)); 332 if (SCARG(&bsa, type) == -1) 333 return EINVAL; 334 /* 335 * Apparently linux uses this to talk to ISDN sockets. If we fail 336 * now programs seems to handle it, but if we don't we are going 337 * to fail when we bind and programs don't handle this well. 338 */ 339 if (SCARG(&bsa, domain) == AF_ROUTE && SCARG(&bsa, type) == SOCK_RAW) 340 return ENOTSUP; 341 error = sys___socket30(l, &bsa, retval); 342 343 #ifdef INET6 344 /* 345 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by 346 * default and some apps depend on this. So, set V6ONLY to 0 347 * for Linux apps if the sysctl value is set to 1. 348 */ 349 if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) { 350 struct socket *so; 351 352 if (fd_getsock(*retval, &so) == 0) { 353 int val = 0; 354 355 /* ignore error */ 356 (void)so_setsockopt(l, so, IPPROTO_IPV6, IPV6_V6ONLY, 357 &val, sizeof(val)); 358 359 fd_putfile(*retval); 360 } 361 } 362 #endif 363 364 return (error); 365 } 366 367 int 368 linux_sys_socketpair(struct lwp *l, const struct linux_sys_socketpair_args *uap, register_t *retval) 369 { 370 /* { 371 syscallarg(int) domain; 372 syscallarg(int) type; 373 syscallarg(int) protocol; 374 syscallarg(int *) rsv; 375 } */ 376 struct sys_socketpair_args bsa; 377 378 memset(&bsa, 0, sizeof(bsa)); 379 SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain)); 380 if (SCARG(&bsa, domain) == -1) 381 return EINVAL; 382 SCARG(&bsa, type) = linux_to_bsd_type(SCARG(uap, type)); 383 if (SCARG(&bsa, type) == -1) 384 return EINVAL; 385 SCARG(&bsa, protocol) = SCARG(uap, protocol); 386 SCARG(&bsa, rsv) = SCARG(uap, rsv); 387 388 return sys_socketpair(l, &bsa, retval); 389 } 390 391 int 392 linux_sys_sendto(struct lwp *l, const struct linux_sys_sendto_args *uap, register_t *retval) 393 { 394 /* { 395 syscallarg(int) s; 396 syscallarg(void *) msg; 397 syscallarg(int) len; 398 syscallarg(int) flags; 399 syscallarg(struct osockaddr *) to; 400 syscallarg(int) tolen; 401 } */ 402 struct msghdr msg; 403 struct iovec aiov; 404 struct sockaddr_big nam; 405 struct mbuf *m; 406 int bflags; 407 int error; 408 409 /* Translate message flags. */ 410 bflags = linux_to_bsd_msg_flags(SCARG(uap, flags)); 411 if (bflags < 0) 412 /* Some supported flag */ 413 return EINVAL; 414 415 msg.msg_flags = 0; 416 msg.msg_name = NULL; 417 msg.msg_control = NULL; 418 419 if (SCARG(uap, tolen)) { 420 /* Read in and convert the sockaddr */ 421 error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to), 422 SCARG(uap, tolen)); 423 if (error) 424 return error; 425 error = sockargs(&m, &nam, nam.sb_len, UIO_SYSSPACE, MT_SONAME); 426 if (error) 427 return error; 428 msg.msg_flags |= MSG_NAMEMBUF; 429 msg.msg_name = m; 430 msg.msg_namelen = nam.sb_len; 431 } 432 433 msg.msg_iov = &aiov; 434 msg.msg_iovlen = 1; 435 aiov.iov_base = __UNCONST(SCARG(uap, msg)); 436 aiov.iov_len = SCARG(uap, len); 437 438 return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval); 439 } 440 441 static int 442 linux_to_bsd_msghdr(const struct linux_msghdr *lmsg, struct msghdr *bmsg) 443 { 444 445 /* 446 * Verify no integer truncation of the inputs. For overly 447 * large inputs, fail with the same error codes Linux uses. 448 */ 449 if (lmsg->msg_namelen < 0) 450 return EINVAL; 451 CTASSERT(!__type_is_signed(__typeof(lmsg->msg_iovlen))); 452 if (lmsg->msg_iovlen > UIO_MAXIOV) 453 return EMSGSIZE; 454 CTASSERT(!__type_is_signed(__typeof(lmsg->msg_controllen))); 455 if (lmsg->msg_controllen > __type_max(__typeof(bmsg->msg_controllen))) 456 return ENOBUFS; 457 458 memset(bmsg, 0, sizeof(*bmsg)); 459 bmsg->msg_name = lmsg->msg_name; 460 bmsg->msg_namelen = MIN(lmsg->msg_namelen, 461 sizeof(struct sockaddr_storage)); 462 bmsg->msg_iov = lmsg->msg_iov; 463 bmsg->msg_iovlen = lmsg->msg_iovlen; 464 bmsg->msg_control = lmsg->msg_control; 465 bmsg->msg_controllen = lmsg->msg_controllen; 466 bmsg->msg_flags = lmsg->msg_flags; 467 468 return 0; 469 } 470 471 static void 472 bsd_to_linux_msghdr(const struct msghdr *bmsg, struct linux_msghdr *lmsg) 473 { 474 475 /* 476 * Assert no truncation on the way out. 477 * 478 * - msg_namelen should have been preserved or shortened by 479 * copyout_sockname. 480 * - msg_iovlen should have been preserved and previously 481 * confirmed by linux_to_bsd_msghdr to lie in [0,UIO_MAXIOV]. 482 * - msg_controllen can always be faithfully converted from BSD 483 * struct msghdr to Linux struct msghdr without truncation. 484 */ 485 CTASSERT(__type_min(__typeof(bmsg->msg_namelen)) == 0); 486 KASSERT(bmsg->msg_namelen <= __type_max(__typeof(lmsg->msg_namelen))); 487 CTASSERT(__type_min(__typeof(lmsg->msg_iovlen)) <= 0); 488 KASSERT(lmsg->msg_iovlen >= 0); 489 KASSERT(lmsg->msg_iovlen <= UIO_MAXIOV); 490 CTASSERT(__type_min(__typeof(lmsg->msg_controllen)) <= 491 __type_min(__typeof(bmsg->msg_controllen))); 492 CTASSERT(__type_max(__typeof(lmsg->msg_controllen)) >= 493 __type_max(__typeof(bmsg->msg_controllen))); 494 495 memset(lmsg, 0, sizeof(*lmsg)); 496 lmsg->msg_name = bmsg->msg_name; 497 lmsg->msg_namelen = bmsg->msg_namelen; 498 lmsg->msg_iov = bmsg->msg_iov; 499 lmsg->msg_iovlen = bmsg->msg_iovlen; 500 lmsg->msg_control = bmsg->msg_control; 501 lmsg->msg_controllen = bmsg->msg_controllen; 502 lmsg->msg_flags = bmsg->msg_flags; 503 } 504 505 int 506 linux_sys_sendmsg(struct lwp *l, const struct linux_sys_sendmsg_args *uap, register_t *retval) 507 { 508 /* { 509 syscallarg(int) s; 510 syscallarg(struct linux_msghdr *) msg; 511 syscallarg(u_int) flags; 512 } */ 513 struct msghdr msg; 514 struct linux_msghdr lmsg; 515 int error; 516 int bflags; 517 struct sockaddr_big nam; 518 u_int8_t *control; 519 struct mbuf *ctl_mbuf = NULL; 520 521 error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg)); 522 if (error) 523 return error; 524 error = linux_to_bsd_msghdr(&lmsg, &msg); 525 if (error) 526 return error; 527 528 msg.msg_flags = MSG_IOVUSRSPACE; 529 530 /* 531 * Translate message flags. 532 */ 533 bflags = linux_to_bsd_msg_flags(SCARG(uap, flags)); 534 if (bflags < 0) 535 /* Some supported flag */ 536 return EINVAL; 537 538 if (lmsg.msg_name) { 539 /* Read in and convert the sockaddr */ 540 error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name, 541 msg.msg_namelen); 542 if (error) 543 return (error); 544 msg.msg_name = &nam; 545 } 546 547 /* 548 * Handle cmsg if there is any. 549 */ 550 if (LINUX_CMSG_FIRSTHDR(&lmsg)) { 551 struct linux_cmsghdr l_cmsg, *l_cc; 552 struct cmsghdr *cmsg; 553 socklen_t resid = msg.msg_controllen; 554 size_t clen, cidx = 0, cspace; 555 556 ctl_mbuf = m_get(M_WAIT, MT_CONTROL); 557 clen = MLEN; 558 control = mtod(ctl_mbuf, void *); 559 560 l_cc = LINUX_CMSG_FIRSTHDR(&lmsg); 561 do { 562 if (resid < sizeof(l_cmsg)) { 563 error = EINVAL; 564 goto done; 565 } 566 error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg)); 567 if (error) 568 goto done; 569 570 /* 571 * Sanity check the control message length. 572 * If either: 573 * 574 * 1. aligning it would overflow socklen_t, or 575 * 2. aligning it exceeds msg_controllen, or 576 * 3. adding the delta would overflow socklen_t, or 577 * 4. it's not enough for a struct cmsghdr, 578 * 579 * then fail with EINVAL. 580 */ 581 CTASSERT(LINUX_CMSG_ALIGN(1) <= 582 __type_max(__typeof(resid))); 583 if ((l_cmsg.cmsg_len >= __type_max(__typeof(resid)) - 584 LINUX_CMSG_ALIGN(1)) || 585 (LINUX_CMSG_ALIGN(l_cmsg.cmsg_len) > resid) || 586 (l_cmsg.cmsg_len > 587 __type_max(__typeof(cmsg->cmsg_len)) - 588 LINUX_CMSG_ALIGN_DELTA) || 589 (l_cmsg.cmsg_len < sizeof(l_cmsg))) { 590 error = EINVAL; 591 goto done; 592 } 593 594 /* 595 * Refuse unsupported control messages, and 596 * translate fields as appropriate. 597 */ 598 switch (l_cmsg.cmsg_level) { 599 case LINUX_SOL_SOCKET: 600 /* It only differs on some archs */ 601 if (LINUX_SOL_SOCKET != SOL_SOCKET) 602 l_cmsg.cmsg_level = SOL_SOCKET; 603 604 switch(l_cmsg.cmsg_type) { 605 case LINUX_SCM_RIGHTS: 606 /* Linux SCM_RIGHTS is same as NetBSD */ 607 break; 608 609 case LINUX_SCM_CREDENTIALS: 610 /* no native equivalent, just drop it */ 611 if (control != mtod(ctl_mbuf, void *)) 612 free(control, M_MBUF); 613 m_free(ctl_mbuf); 614 ctl_mbuf = NULL; 615 msg.msg_control = NULL; 616 msg.msg_controllen = 0; 617 goto skipcmsg; 618 619 default: 620 /* other types not supported */ 621 error = EINVAL; 622 goto done; 623 } 624 break; 625 default: 626 /* pray and leave intact */ 627 break; 628 } 629 630 cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg)); 631 632 /* Check the buffer is big enough */ 633 if (__predict_false(cidx + cspace > clen)) { 634 u_int8_t *nc; 635 size_t nclen; 636 637 nclen = cidx + cspace; 638 if (nclen >= PAGE_SIZE) { 639 error = EINVAL; 640 goto done; 641 } 642 nc = realloc(clen <= MLEN ? NULL : control, 643 nclen, M_TEMP, M_WAITOK); 644 if (!nc) { 645 error = ENOMEM; 646 goto done; 647 } 648 if (cidx <= MLEN) 649 /* Old buffer was in mbuf... */ 650 memcpy(nc, control, cidx); 651 control = nc; 652 clen = nclen; 653 } 654 655 /* Copy header */ 656 cmsg = (void *)&control[cidx]; 657 cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA; 658 cmsg->cmsg_level = l_cmsg.cmsg_level; 659 cmsg->cmsg_type = l_cmsg.cmsg_type; 660 661 /* Zero area between header and data */ 662 memset(cmsg + 1, 0, 663 CMSG_ALIGN(sizeof(*cmsg)) - sizeof(*cmsg)); 664 665 /* Copyin the data */ 666 error = copyin(LINUX_CMSG_DATA(l_cc), 667 CMSG_DATA(cmsg), 668 l_cmsg.cmsg_len - sizeof(l_cmsg)); 669 if (error) 670 goto done; 671 672 resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len); 673 cidx += cspace; 674 } while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc, &l_cmsg)) && resid > 0); 675 676 /* If we allocated a buffer, attach to mbuf */ 677 if (cidx > MLEN) { 678 MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL); 679 ctl_mbuf->m_flags |= M_EXT_RW; 680 } 681 control = NULL; 682 ctl_mbuf->m_len = cidx; 683 684 msg.msg_control = ctl_mbuf; 685 msg.msg_flags |= MSG_CONTROLMBUF; 686 687 ktrkuser("mbcontrol", mtod(ctl_mbuf, void *), 688 msg.msg_controllen); 689 } 690 691 skipcmsg: 692 error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval); 693 /* Freed internally */ 694 ctl_mbuf = NULL; 695 696 done: 697 if (ctl_mbuf != NULL) { 698 if (control != NULL && control != mtod(ctl_mbuf, void *)) 699 free(control, M_MBUF); 700 m_free(ctl_mbuf); 701 } 702 return (error); 703 } 704 705 int 706 linux_sys_recvfrom(struct lwp *l, const struct linux_sys_recvfrom_args *uap, register_t *retval) 707 { 708 /* { 709 syscallarg(int) s; 710 syscallarg(void *) buf; 711 syscallarg(int) len; 712 syscallarg(int) flags; 713 syscallarg(struct osockaddr *) from; 714 syscallarg(int *) fromlenaddr; 715 } */ 716 int error; 717 int bflags; 718 struct sys_recvfrom_args bra; 719 720 /* 721 * Translate message flags. 722 */ 723 bflags = linux_to_bsd_msg_flags(SCARG(uap, flags)); 724 if (bflags < 0) 725 return EINVAL; 726 727 memset(&bra, 0, sizeof(bra)); 728 SCARG(&bra, s) = SCARG(uap, s); 729 SCARG(&bra, buf) = SCARG(uap, buf); 730 SCARG(&bra, len) = SCARG(uap, len); 731 SCARG(&bra, flags) = bflags; 732 SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from); 733 SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr); 734 735 if ((error = sys_recvfrom(l, &bra, retval))) 736 return (error); 737 738 if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from)))) 739 return (error); 740 741 return (0); 742 } 743 744 static int 745 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control) 746 { 747 socklen_t dlen; 748 int error = 0; 749 struct cmsghdr *cmsg; 750 struct linux_cmsghdr linux_cmsg; 751 struct mbuf *m; 752 char *q, *q_end; 753 754 if (mp->msg_controllen <= 0 || control == 0) { 755 mp->msg_controllen = 0; 756 free_control_mbuf(l, control, control); 757 return 0; 758 } 759 760 ktrkuser("msgcontrol", mtod(control, void *), mp->msg_controllen); 761 762 q = (char *)mp->msg_control; 763 q_end = q + mp->msg_controllen; 764 765 for (m = control; m != NULL; ) { 766 cmsg = mtod(m, struct cmsghdr *); 767 768 /* 769 * Fixup cmsg. We handle two things: 770 * 0. different sizeof cmsg_len. 771 * 1. different values for level/type on some archs 772 * 2. different alignment of CMSG_DATA on some archs 773 */ 774 memset(&linux_cmsg, 0, sizeof(linux_cmsg)); 775 CTASSERT(LINUX_CMSG_ALIGN_DELTA >= 0); 776 CTASSERT(CMSG_LEN(0) == 777 sizeof(linux_cmsg) + LINUX_CMSG_ALIGN_DELTA); 778 KASSERTMSG(cmsg->cmsg_len >= CMSG_LEN(0), 779 "cmsg->cmsg_len=%u", cmsg->cmsg_len); 780 linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA; 781 KASSERTMSG(linux_cmsg.cmsg_len >= sizeof(linux_cmsg), 782 "linux_cmsg.cmsg_len=%zu", linux_cmsg.cmsg_len); 783 linux_cmsg.cmsg_level = cmsg->cmsg_level; 784 linux_cmsg.cmsg_type = cmsg->cmsg_type; 785 786 dlen = q_end - q; 787 if (linux_cmsg.cmsg_len > dlen) { 788 /* Not enough room for the parameter */ 789 if (dlen < sizeof linux_cmsg) 790 /* Discard if header wont fit */ 791 break; 792 dlen -= sizeof linux_cmsg; 793 mp->msg_flags |= MSG_CTRUNC; 794 if (linux_cmsg.cmsg_level == SOL_SOCKET 795 && linux_cmsg.cmsg_type == SCM_RIGHTS) 796 /* Do not truncate me ... */ 797 break; 798 } else 799 dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg; 800 801 /* 802 * NetBSD will not create a cmsg record whose header 803 * length plus aligned content length would overflow 804 * socklen_t. This justifies computing 805 * LINUX_CMSG_SPACE(dlen) (and LINUX_CMSG_LEN) below 806 * without overflow. 807 */ 808 CTASSERT(LINUX_CMSG_SPACE(1) <= CMSG_SPACE(1)); 809 CTASSERT(CMSG_SPACE(0) <= 810 __type_max(__typeof(cmsg->cmsg_len))); 811 KASSERTMSG((dlen <= __type_max(__typeof(dlen)) - 812 CMSG_SPACE(0)), 813 "dlen=%u", dlen); 814 CTASSERT(LINUX_CMSG_LEN(1) <= LINUX_CMSG_SPACE(1)); 815 816 switch (linux_cmsg.cmsg_level) { 817 case SOL_SOCKET: 818 linux_cmsg.cmsg_level = LINUX_SOL_SOCKET; 819 switch (linux_cmsg.cmsg_type) { 820 case SCM_RIGHTS: 821 /* Linux SCM_RIGHTS is same as NetBSD */ 822 break; 823 824 default: 825 /* other types not supported */ 826 error = EINVAL; 827 goto done; 828 } 829 /* machine dependent ! */ 830 break; 831 default: 832 /* pray and leave intact */ 833 break; 834 } 835 836 /* There can be padding between the header and data... */ 837 error = copyout(&linux_cmsg, q, sizeof linux_cmsg); 838 if (error != 0) { 839 break; 840 } 841 error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg, dlen); 842 if (error != 0) { 843 break; 844 } 845 m = m->m_next; 846 if (m == NULL || LINUX_CMSG_SPACE(dlen) > q_end - q) { 847 q += LINUX_CMSG_LEN(dlen); 848 break; 849 } 850 q += LINUX_CMSG_SPACE(dlen); 851 } 852 853 done: 854 if (error) { 855 /* We must free all the SCM_RIGHTS */ 856 m = control; 857 } 858 free_control_mbuf(l, control, m); 859 860 mp->msg_controllen = q - (char *)mp->msg_control; 861 return error; 862 } 863 864 int 865 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval) 866 { 867 /* { 868 syscallarg(int) s; 869 syscallarg(struct linux_msghdr *) msg; 870 syscallarg(u_int) flags; 871 } */ 872 struct msghdr msg; 873 struct linux_msghdr lmsg; 874 int error; 875 struct mbuf *from, *control; 876 877 error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg)); 878 if (error) 879 return error; 880 error = linux_to_bsd_msghdr(&lmsg, &msg); 881 if (error) 882 return error; 883 884 msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags)); 885 if (msg.msg_flags < 0) { 886 /* Some unsupported flag */ 887 return (EINVAL); 888 } 889 msg.msg_flags |= MSG_IOVUSRSPACE; 890 891 error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from, 892 msg.msg_control != NULL ? &control : NULL, retval); 893 if (error != 0) 894 return error; 895 896 if (msg.msg_control != NULL) 897 error = linux_copyout_msg_control(l, &msg, control); 898 899 if (error == 0 && from != 0) { 900 mtod(from, struct osockaddr *)->sa_family = 901 bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family); 902 error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0, 903 from); 904 } else 905 msg.msg_namelen = 0; 906 907 if (from != NULL) 908 m_free(from); 909 910 if (error == 0) { 911 msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags); 912 if (msg.msg_flags < 0) 913 /* Some flag unsupported by Linux */ 914 error = EINVAL; 915 else { 916 ktrkuser("msghdr", &msg, sizeof(msg)); 917 bsd_to_linux_msghdr(&msg, &lmsg); 918 error = copyout(&lmsg, SCARG(uap, msg), sizeof(lmsg)); 919 } 920 } 921 922 return (error); 923 } 924 925 /* 926 * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET 927 * is different, the rest matches IPPROTO_* on both systems. 928 */ 929 int 930 linux_to_bsd_sopt_level(int llevel) 931 { 932 933 switch (llevel) { 934 case LINUX_SOL_SOCKET: 935 return SOL_SOCKET; 936 case LINUX_SOL_IP: 937 return IPPROTO_IP; 938 #ifdef INET6 939 case LINUX_SOL_IPV6: 940 return IPPROTO_IPV6; 941 #endif 942 case LINUX_SOL_TCP: 943 return IPPROTO_TCP; 944 case LINUX_SOL_UDP: 945 return IPPROTO_UDP; 946 default: 947 return -1; 948 } 949 } 950 951 /* 952 * Convert Linux socket level socket option numbers to NetBSD values. 953 */ 954 int 955 linux_to_bsd_so_sockopt(int lopt) 956 { 957 958 switch (lopt) { 959 case LINUX_SO_DEBUG: 960 return SO_DEBUG; 961 case LINUX_SO_REUSEADDR: 962 /* 963 * Linux does not implement SO_REUSEPORT, but allows reuse of 964 * a host:port pair through SO_REUSEADDR even if the address 965 * is not a multicast-address. Effectively, this means that we 966 * should use SO_REUSEPORT to allow Linux applications to not 967 * exit with EADDRINUSE 968 */ 969 return SO_REUSEPORT; 970 case LINUX_SO_TYPE: 971 return SO_TYPE; 972 case LINUX_SO_ERROR: 973 return SO_ERROR; 974 case LINUX_SO_DONTROUTE: 975 return SO_DONTROUTE; 976 case LINUX_SO_BROADCAST: 977 return SO_BROADCAST; 978 case LINUX_SO_SNDBUF: 979 return SO_SNDBUF; 980 case LINUX_SO_RCVBUF: 981 return SO_RCVBUF; 982 case LINUX_SO_KEEPALIVE: 983 return SO_KEEPALIVE; 984 case LINUX_SO_OOBINLINE: 985 return SO_OOBINLINE; 986 case LINUX_SO_NO_CHECK: 987 case LINUX_SO_PRIORITY: 988 return -1; 989 case LINUX_SO_LINGER: 990 return SO_LINGER; 991 case LINUX_SO_BSDCOMPAT: 992 case LINUX_SO_PASSCRED: 993 case LINUX_SO_PEERCRED: 994 return -1; 995 case LINUX_SO_RCVLOWAT: 996 return SO_RCVLOWAT; 997 case LINUX_SO_SNDLOWAT: 998 return SO_SNDLOWAT; 999 case LINUX_SO_RCVTIMEO: 1000 return SO_RCVTIMEO; 1001 case LINUX_SO_SNDTIMEO: 1002 return SO_SNDTIMEO; 1003 case LINUX_SO_SECURITY_AUTHENTICATION: 1004 case LINUX_SO_SECURITY_ENCRYPTION_TRANSPORT: 1005 case LINUX_SO_SECURITY_ENCRYPTION_NETWORK: 1006 case LINUX_SO_BINDTODEVICE: 1007 case LINUX_SO_ATTACH_FILTER: 1008 case LINUX_SO_DETACH_FILTER: 1009 case LINUX_SO_PEERNAME: 1010 return -1; 1011 case LINUX_SO_TIMESTAMP: 1012 return SO_TIMESTAMP; 1013 case LINUX_SO_ACCEPTCONN: 1014 case LINUX_SO_PEERSEC: 1015 case LINUX_SO_SNDBUFFORCE: 1016 case LINUX_SO_RCVBUFFORCE: 1017 case LINUX_SO_PASSSEC: 1018 case LINUX_SO_TIMESTAMPNS: 1019 case LINUX_SO_MARK: 1020 case LINUX_SO_TIMESTAMPING: 1021 case LINUX_SO_PROTOCOL: 1022 case LINUX_SO_DOMAIN: 1023 case LINUX_SO_RXQ_OVFL: 1024 case LINUX_SO_WIFI_STATUS: 1025 case LINUX_SO_PEEK_OFF: 1026 case LINUX_SO_NOFCS: 1027 default: 1028 return -1; 1029 } 1030 } 1031 1032 /* 1033 * Convert Linux IP level socket option number to NetBSD values. 1034 */ 1035 int 1036 linux_to_bsd_ip_sockopt(int lopt) 1037 { 1038 1039 switch (lopt) { 1040 case LINUX_IP_TOS: 1041 return IP_TOS; 1042 case LINUX_IP_TTL: 1043 return IP_TTL; 1044 case LINUX_IP_RETOPTS: 1045 return IP_RETOPTS; 1046 case LINUX_IP_PKTINFO: 1047 return IP_PKTINFO; 1048 case LINUX_IP_RECVOPTS: 1049 return IP_RECVOPTS; 1050 case LINUX_IP_HDRINCL: 1051 return IP_HDRINCL; 1052 case LINUX_IP_MULTICAST_TTL: 1053 return IP_MULTICAST_TTL; 1054 case LINUX_IP_MULTICAST_LOOP: 1055 return IP_MULTICAST_LOOP; 1056 case LINUX_IP_MULTICAST_IF: 1057 return IP_MULTICAST_IF; 1058 case LINUX_IP_ADD_MEMBERSHIP: 1059 return IP_ADD_MEMBERSHIP; 1060 case LINUX_IP_DROP_MEMBERSHIP: 1061 return IP_DROP_MEMBERSHIP; 1062 case LINUX_IP_RECVERR: 1063 case LINUX_IP_FREEBIND: 1064 return -2; /* ignored */ 1065 case LINUX_IP_MULTICAST_ALL: 1066 return -3; /* noprotoopt */ 1067 default: 1068 return -1; 1069 } 1070 } 1071 1072 /* 1073 * Convert Linux IPV6 level socket option number to NetBSD values. 1074 */ 1075 #ifdef INET6 1076 int 1077 linux_to_bsd_ipv6_sockopt(int lopt) 1078 { 1079 1080 switch (lopt) { 1081 case LINUX_IPV6_V6ONLY: 1082 return IPV6_V6ONLY; 1083 case LINUX_IPV6_MULTICAST_HOPS: 1084 return IPV6_MULTICAST_HOPS; 1085 case LINUX_IPV6_MULTICAST_ALL: 1086 return -3; /* noprotoopt */ 1087 default: 1088 return -1; 1089 } 1090 } 1091 #endif 1092 1093 /* 1094 * Convert Linux TCP level socket option number to NetBSD values. 1095 */ 1096 int 1097 linux_to_bsd_tcp_sockopt(int lopt) 1098 { 1099 1100 switch (lopt) { 1101 case LINUX_TCP_NODELAY: 1102 return TCP_NODELAY; 1103 case LINUX_TCP_MAXSEG: 1104 return TCP_MAXSEG; 1105 default: 1106 return -1; 1107 } 1108 } 1109 1110 /* 1111 * Convert Linux UDP level socket option number to NetBSD values. 1112 */ 1113 int 1114 linux_to_bsd_udp_sockopt(int lopt) 1115 { 1116 1117 switch (lopt) { 1118 default: 1119 return -1; 1120 } 1121 } 1122 1123 /* 1124 * Another reasonably straightforward function: setsockopt(2). 1125 * The level and option numbers are converted; the values passed 1126 * are not (yet) converted, the ones currently implemented don't 1127 * need conversion, as they are the same on both systems. 1128 */ 1129 int 1130 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval) 1131 { 1132 /* { 1133 syscallarg(int) s; 1134 syscallarg(int) level; 1135 syscallarg(int) optname; 1136 syscallarg(void *) optval; 1137 syscallarg(int) optlen; 1138 } */ 1139 struct sys_setsockopt_args bsa; 1140 int name; 1141 1142 memset(&bsa, 0, sizeof(bsa)); 1143 SCARG(&bsa, s) = SCARG(uap, s); 1144 SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level)); 1145 SCARG(&bsa, val) = SCARG(uap, optval); 1146 SCARG(&bsa, valsize) = SCARG(uap, optlen); 1147 1148 /* 1149 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets 1150 * and returns EOPNOTSUPP for other levels 1151 */ 1152 if (SCARG(&bsa, level) != SOL_SOCKET) { 1153 struct socket *so; 1154 int error, family; 1155 1156 /* fd_getsock() will use the descriptor for us */ 1157 if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0) 1158 return error; 1159 family = so->so_proto->pr_domain->dom_family; 1160 fd_putfile(SCARG(&bsa, s)); 1161 1162 if (family == AF_LOCAL) 1163 return EOPNOTSUPP; 1164 } 1165 1166 switch (SCARG(&bsa, level)) { 1167 case SOL_SOCKET: 1168 name = linux_to_bsd_so_sockopt(SCARG(uap, optname)); 1169 break; 1170 case IPPROTO_IP: 1171 name = linux_to_bsd_ip_sockopt(SCARG(uap, optname)); 1172 break; 1173 #ifdef INET6 1174 case IPPROTO_IPV6: 1175 name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname)); 1176 break; 1177 #endif 1178 case IPPROTO_TCP: 1179 name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname)); 1180 break; 1181 case IPPROTO_UDP: 1182 name = linux_to_bsd_udp_sockopt(SCARG(uap, optname)); 1183 break; 1184 default: 1185 return EINVAL; 1186 } 1187 1188 switch (name) { 1189 case -1: 1190 return EINVAL; 1191 case -2: 1192 return 0; 1193 case -3: 1194 return ENOPROTOOPT; 1195 } 1196 SCARG(&bsa, name) = name; 1197 1198 return sys_setsockopt(l, &bsa, retval); 1199 } 1200 1201 /* 1202 * getsockopt(2) is very much the same as setsockopt(2) (see above) 1203 */ 1204 int 1205 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval) 1206 { 1207 /* { 1208 syscallarg(int) s; 1209 syscallarg(int) level; 1210 syscallarg(int) optname; 1211 syscallarg(void *) optval; 1212 syscallarg(int *) optlen; 1213 } */ 1214 struct sys_getsockopt_args bga; 1215 int name; 1216 1217 memset(&bga, 0, sizeof(bga)); 1218 SCARG(&bga, s) = SCARG(uap, s); 1219 SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level)); 1220 SCARG(&bga, val) = SCARG(uap, optval); 1221 SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen); 1222 1223 switch (SCARG(&bga, level)) { 1224 case SOL_SOCKET: 1225 name = linux_to_bsd_so_sockopt(SCARG(uap, optname)); 1226 break; 1227 case IPPROTO_IP: 1228 name = linux_to_bsd_ip_sockopt(SCARG(uap, optname)); 1229 break; 1230 #ifdef INET6 1231 case IPPROTO_IPV6: 1232 name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname)); 1233 break; 1234 #endif 1235 case IPPROTO_TCP: 1236 name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname)); 1237 break; 1238 case IPPROTO_UDP: 1239 name = linux_to_bsd_udp_sockopt(SCARG(uap, optname)); 1240 break; 1241 default: 1242 return EINVAL; 1243 } 1244 1245 switch (name) { 1246 case -1: 1247 case -2: /* we can't ignore, since we don't know what to return */ 1248 return EINVAL; 1249 case -3: 1250 return ENOPROTOOPT; 1251 } 1252 SCARG(&bga, name) = name; 1253 1254 return sys_getsockopt(l, &bga, retval); 1255 } 1256 1257 int 1258 linux_getifname(struct lwp *l, register_t *retval, void *data) 1259 { 1260 struct ifnet *ifp; 1261 struct linux_ifreq ifr; 1262 int error; 1263 int s; 1264 1265 error = copyin(data, &ifr, sizeof(ifr)); 1266 if (error) 1267 return error; 1268 1269 s = pserialize_read_enter(); 1270 ifp = if_byindex(ifr.ifr_ifru.ifru_ifindex); 1271 if (ifp == NULL) { 1272 pserialize_read_exit(s); 1273 return ENODEV; 1274 } 1275 1276 strncpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)); 1277 pserialize_read_exit(s); 1278 1279 return copyout(&ifr, data, sizeof(ifr)); 1280 } 1281 1282 int 1283 linux_getifconf(struct lwp *l, register_t *retval, void *data) 1284 { 1285 struct linux_ifreq ifr, *ifrp = NULL; 1286 struct linux_ifconf ifc; 1287 struct ifnet *ifp; 1288 struct sockaddr *sa; 1289 struct osockaddr *osa; 1290 int space = 0, error; 1291 const int sz = (int)sizeof(ifr); 1292 bool docopy; 1293 int s; 1294 int bound; 1295 struct psref psref; 1296 1297 error = copyin(data, &ifc, sizeof(ifc)); 1298 if (error) 1299 return error; 1300 1301 docopy = ifc.ifc_req != NULL; 1302 if (docopy) { 1303 if (ifc.ifc_len < 0) 1304 return EINVAL; 1305 1306 space = ifc.ifc_len; 1307 ifrp = ifc.ifc_req; 1308 } 1309 memset(&ifr, 0, sizeof(ifr)); 1310 1311 bound = curlwp_bind(); 1312 s = pserialize_read_enter(); 1313 IFNET_READER_FOREACH(ifp) { 1314 struct ifaddr *ifa; 1315 if_acquire(ifp, &psref); 1316 pserialize_read_exit(s); 1317 1318 (void)strncpy(ifr.ifr_name, ifp->if_xname, 1319 sizeof(ifr.ifr_name)); 1320 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') { 1321 error = ENAMETOOLONG; 1322 goto release_exit; 1323 } 1324 1325 s = pserialize_read_enter(); 1326 IFADDR_READER_FOREACH(ifa, ifp) { 1327 struct psref psref_ifa; 1328 ifa_acquire(ifa, &psref_ifa); 1329 pserialize_read_exit(s); 1330 1331 sa = ifa->ifa_addr; 1332 if (sa->sa_family != AF_INET || 1333 sa->sa_len > sizeof(*osa)) 1334 goto next; 1335 memcpy(&ifr.ifr_addr, sa, sa->sa_len); 1336 osa = (struct osockaddr *)&ifr.ifr_addr; 1337 osa->sa_family = sa->sa_family; 1338 if (space >= sz) { 1339 error = copyout(&ifr, ifrp, sz); 1340 if (error != 0) { 1341 ifa_release(ifa, &psref_ifa); 1342 goto release_exit; 1343 } 1344 ifrp++; 1345 } 1346 space -= sz; 1347 next: 1348 s = pserialize_read_enter(); 1349 ifa_release(ifa, &psref_ifa); 1350 } 1351 1352 KASSERT(pserialize_in_read_section()); 1353 if_release(ifp, &psref); 1354 } 1355 pserialize_read_exit(s); 1356 curlwp_bindx(bound); 1357 1358 if (docopy) 1359 ifc.ifc_len -= space; 1360 else 1361 ifc.ifc_len = -space; 1362 1363 return copyout(&ifc, data, sizeof(ifc)); 1364 1365 release_exit: 1366 if_release(ifp, &psref); 1367 curlwp_bindx(bound); 1368 return error; 1369 } 1370 1371 int 1372 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd, 1373 void *data) 1374 { 1375 /* Not the full structure, just enough to map what we do here */ 1376 struct linux_ifreq lreq; 1377 file_t *fp; 1378 struct ifaddr *ifa; 1379 struct ifnet *ifp; 1380 struct sockaddr_dl *sadl; 1381 int error, found; 1382 int index, ifnum; 1383 int s; 1384 1385 /* 1386 * We can't emulate this ioctl by calling sys_ioctl() to run 1387 * SIOCGIFCONF, because the user buffer is not of the right 1388 * type to take those results. We can't use kernel buffers to 1389 * receive the results, as the implementation of sys_ioctl() 1390 * and ifconf() [which implements SIOCGIFCONF] use 1391 * copyin()/copyout() which will fail on kernel addresses. 1392 * 1393 * So, we must duplicate code from sys_ioctl() and ifconf(). Ugh. 1394 */ 1395 1396 if ((fp = fd_getfile(fd)) == NULL) 1397 return (EBADF); 1398 1399 KERNEL_LOCK(1, NULL); 1400 1401 if ((fp->f_flag & (FREAD | FWRITE)) == 0) { 1402 error = EBADF; 1403 goto out; 1404 } 1405 1406 error = copyin(data, &lreq, sizeof(lreq)); 1407 if (error) 1408 goto out; 1409 lreq.ifr_name[LINUX_IFNAMSIZ-1] = '\0'; /* just in case */ 1410 1411 /* 1412 * Try real interface name first, then fake "ethX" 1413 */ 1414 found = 0; 1415 s = pserialize_read_enter(); 1416 IFNET_READER_FOREACH(ifp) { 1417 if (found) 1418 break; 1419 if (strcmp(lreq.ifr_name, ifp->if_xname)) 1420 /* not this interface */ 1421 continue; 1422 1423 found=1; 1424 if (IFADDR_READER_EMPTY(ifp)) { 1425 pserialize_read_exit(s); 1426 error = ENODEV; 1427 goto out; 1428 } 1429 IFADDR_READER_FOREACH(ifa, ifp) { 1430 sadl = satosdl(ifa->ifa_addr); 1431 /* only return ethernet addresses */ 1432 /* XXX what about FDDI, etc. ? */ 1433 if (sadl->sdl_family != AF_LINK || 1434 sadl->sdl_type != IFT_ETHER) 1435 continue; 1436 memcpy(&lreq.ifr_hwaddr.sa_data, CLLADDR(sadl), 1437 MIN(sadl->sdl_alen, 1438 sizeof(lreq.ifr_hwaddr.sa_data))); 1439 lreq.ifr_hwaddr.sa_family = 1440 sadl->sdl_family; 1441 pserialize_read_exit(s); 1442 1443 error = copyout(&lreq, data, sizeof(lreq)); 1444 goto out; 1445 } 1446 } 1447 pserialize_read_exit(s); 1448 1449 if (strncmp(lreq.ifr_name, "eth", 3) != 0) { 1450 /* unknown interface, not even an "eth*" name */ 1451 error = ENODEV; 1452 goto out; 1453 } 1454 1455 for (ifnum = 0, index = 3; 1456 index < LINUX_IFNAMSIZ && lreq.ifr_name[index] != '\0'; 1457 index++) { 1458 ifnum *= 10; 1459 ifnum += lreq.ifr_name[index] - '0'; 1460 } 1461 1462 error = EINVAL; /* in case we don't find one */ 1463 s = pserialize_read_enter(); 1464 IFNET_READER_FOREACH(ifp) { 1465 memcpy(lreq.ifr_name, ifp->if_xname, 1466 MIN(LINUX_IFNAMSIZ, IFNAMSIZ)); 1467 IFADDR_READER_FOREACH(ifa, ifp) { 1468 sadl = satosdl(ifa->ifa_addr); 1469 /* only return ethernet addresses */ 1470 /* XXX what about FDDI, etc. ? */ 1471 if (sadl->sdl_family != AF_LINK || 1472 sadl->sdl_type != IFT_ETHER) 1473 continue; 1474 if (ifnum--) 1475 /* not the requested iface */ 1476 continue; 1477 memcpy(&lreq.ifr_hwaddr.sa_data, 1478 CLLADDR(sadl), 1479 MIN(sadl->sdl_alen, 1480 sizeof(lreq.ifr_hwaddr.sa_data))); 1481 lreq.ifr_hwaddr.sa_family = 1482 sadl->sdl_family; 1483 pserialize_read_exit(s); 1484 1485 error = copyout(&lreq, data, sizeof(lreq)); 1486 goto out; 1487 } 1488 } 1489 pserialize_read_exit(s); 1490 1491 out: 1492 KERNEL_UNLOCK_ONE(NULL); 1493 fd_putfile(fd); 1494 return error; 1495 } 1496 1497 int 1498 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval) 1499 { 1500 /* { 1501 syscallarg(int) fd; 1502 syscallarg(u_long) com; 1503 syscallarg(void *) data; 1504 } */ 1505 u_long com; 1506 int error = 0, isdev = 0, dosys = 1; 1507 struct sys_ioctl_args ia; 1508 file_t *fp; 1509 struct vnode *vp; 1510 int (*ioctlf)(file_t *, u_long, void *); 1511 struct ioctl_pt pt; 1512 1513 if ((fp = fd_getfile(SCARG(uap, fd))) == NULL) 1514 return (EBADF); 1515 1516 if (fp->f_type == DTYPE_VNODE) { 1517 vp = (struct vnode *)fp->f_data; 1518 isdev = vp->v_type == VCHR; 1519 } 1520 1521 /* 1522 * Don't try to interpret socket ioctl calls that are done 1523 * on a device filedescriptor, just pass them through, to 1524 * emulate Linux behaviour. Use PTIOCLINUX so that the 1525 * device will only handle these if it's prepared to do 1526 * so, to avoid unexpected things from happening. 1527 */ 1528 if (isdev) { 1529 dosys = 0; 1530 ioctlf = fp->f_ops->fo_ioctl; 1531 pt.com = SCARG(uap, com); 1532 pt.data = SCARG(uap, data); 1533 error = ioctlf(fp, PTIOCLINUX, &pt); 1534 /* 1535 * XXX hack: if the function returns EJUSTRETURN, 1536 * it has stuffed a sysctl return value in pt.data. 1537 */ 1538 if (error == EJUSTRETURN) { 1539 retval[0] = (register_t)pt.data; 1540 error = 0; 1541 } 1542 goto out; 1543 } 1544 1545 com = SCARG(uap, com); 1546 retval[0] = 0; 1547 1548 memset(&ia, 0, sizeof(ia)); 1549 1550 switch (com) { 1551 case LINUX_SIOCGIFNAME: 1552 error = linux_getifname(l, retval, SCARG(uap, data)); 1553 dosys = 0; 1554 break; 1555 case LINUX_SIOCGIFCONF: 1556 error = linux_getifconf(l, retval, SCARG(uap, data)); 1557 dosys = 0; 1558 break; 1559 case LINUX_SIOCGIFFLAGS: 1560 SCARG(&ia, com) = OSIOCGIFFLAGS; 1561 break; 1562 case LINUX_SIOCSIFFLAGS: 1563 SCARG(&ia, com) = OSIOCSIFFLAGS; 1564 break; 1565 case LINUX_SIOCGIFADDR: 1566 SCARG(&ia, com) = OOSIOCGIFADDR; 1567 break; 1568 case LINUX_SIOCGIFDSTADDR: 1569 SCARG(&ia, com) = OOSIOCGIFDSTADDR; 1570 break; 1571 case LINUX_SIOCGIFBRDADDR: 1572 SCARG(&ia, com) = OOSIOCGIFBRDADDR; 1573 break; 1574 case LINUX_SIOCGIFNETMASK: 1575 SCARG(&ia, com) = OOSIOCGIFNETMASK; 1576 break; 1577 case LINUX_SIOCGIFMTU: 1578 SCARG(&ia, com) = OSIOCGIFMTU; 1579 break; 1580 case LINUX_SIOCADDMULTI: 1581 SCARG(&ia, com) = OSIOCADDMULTI; 1582 break; 1583 case LINUX_SIOCDELMULTI: 1584 SCARG(&ia, com) = OSIOCDELMULTI; 1585 break; 1586 case LINUX_SIOCGIFHWADDR: 1587 error = linux_getifhwaddr(l, retval, SCARG(uap, fd), 1588 SCARG(uap, data)); 1589 dosys = 0; 1590 break; 1591 default: 1592 error = EINVAL; 1593 } 1594 1595 out: 1596 fd_putfile(SCARG(uap, fd)); 1597 1598 if (error ==0 && dosys) { 1599 SCARG(&ia, fd) = SCARG(uap, fd); 1600 SCARG(&ia, data) = SCARG(uap, data); 1601 error = sys_ioctl(curlwp, &ia, retval); 1602 } 1603 1604 return error; 1605 } 1606 1607 int 1608 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval) 1609 { 1610 /* { 1611 syscallarg(int) s; 1612 syscallarg(const struct sockaddr *) name; 1613 syscallarg(int) namelen; 1614 } */ 1615 int error; 1616 struct sockaddr_big sb; 1617 1618 error = linux_get_sa(l, SCARG(uap, s), &sb, SCARG(uap, name), 1619 SCARG(uap, namelen)); 1620 if (error) 1621 return (error); 1622 1623 error = do_sys_connect(l, SCARG(uap, s), (struct sockaddr *)&sb); 1624 1625 if (error == EISCONN) { 1626 struct socket *so; 1627 int state, prflags; 1628 1629 /* fd_getsock() will use the descriptor for us */ 1630 if (fd_getsock(SCARG(uap, s), &so) != 0) 1631 return EISCONN; 1632 1633 solock(so); 1634 state = so->so_state; 1635 prflags = so->so_proto->pr_flags; 1636 sounlock(so); 1637 fd_putfile(SCARG(uap, s)); 1638 /* 1639 * We should only let this call succeed once per 1640 * non-blocking connect; however we don't have 1641 * a convenient place to keep that state.. 1642 */ 1643 if ((state & (SS_ISCONNECTED|SS_NBIO)) == 1644 (SS_ISCONNECTED|SS_NBIO) && 1645 (prflags & PR_CONNREQUIRED)) 1646 return 0; 1647 } 1648 1649 return (error); 1650 } 1651 1652 int 1653 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval) 1654 { 1655 /* { 1656 syscallarg(int) s; 1657 syscallarg(const struct osockaddr *) name; 1658 syscallarg(int) namelen; 1659 } */ 1660 int error; 1661 struct sockaddr_big sb; 1662 1663 error = linux_get_sa(l, SCARG(uap, s), &sb, SCARG(uap, name), 1664 SCARG(uap, namelen)); 1665 if (error) 1666 return (error); 1667 1668 return do_sys_bind(l, SCARG(uap, s), (struct sockaddr *)&sb); 1669 } 1670 1671 int 1672 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval) 1673 { 1674 /* { 1675 syscallarg(int) fdes; 1676 syscallarg(void *) asa; 1677 syscallarg(int *) alen; 1678 } */ 1679 int error; 1680 1681 if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0) 1682 return (error); 1683 1684 if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa)))) 1685 return (error); 1686 1687 return (0); 1688 } 1689 1690 int 1691 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval) 1692 { 1693 /* { 1694 syscallarg(int) fdes; 1695 syscallarg(void *) asa; 1696 syscallarg(int *) alen; 1697 } */ 1698 int error; 1699 1700 if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0) 1701 return (error); 1702 1703 if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa)))) 1704 return (error); 1705 1706 return (0); 1707 } 1708 1709 /* 1710 * Copy the osockaddr structure pointed to by name to sb, adjust 1711 * family and convert to sockaddr. 1712 */ 1713 static int 1714 linux_get_sa(struct lwp *l, int s, struct sockaddr_big *sb, 1715 const struct osockaddr *name, socklen_t namelen) 1716 { 1717 int error, bdom; 1718 1719 if (namelen > UCHAR_MAX || 1720 namelen <= offsetof(struct sockaddr_big, sb_data)) 1721 return EINVAL; 1722 1723 error = copyin(name, sb, namelen); 1724 if (error) 1725 return error; 1726 1727 bdom = linux_to_bsd_domain(sb->sb_family); 1728 if (bdom == -1) 1729 return EINVAL; 1730 1731 /* 1732 * If the family is unspecified, use address family of the socket. 1733 * This avoid triggering strict family checks in netinet/in_pcb.c et.al. 1734 */ 1735 if (bdom == AF_UNSPEC) { 1736 struct socket *so; 1737 1738 /* fd_getsock() will use the descriptor for us */ 1739 if ((error = fd_getsock(s, &so)) != 0) 1740 return error; 1741 1742 bdom = so->so_proto->pr_domain->dom_family; 1743 fd_putfile(s); 1744 } 1745 1746 /* 1747 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6, 1748 * which lacks the scope id compared with RFC2553 one. If we detect 1749 * the situation, reject the address and write a message to system log. 1750 * 1751 * Still accept addresses for which the scope id is not used. 1752 */ 1753 if (bdom == AF_INET6 && 1754 namelen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) { 1755 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sb; 1756 if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) && 1757 (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) || 1758 IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) || 1759 IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) || 1760 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) || 1761 IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) { 1762 struct proc *p = l->l_proc; 1763 int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1; 1764 1765 log(LOG_DEBUG, 1766 "pid %d (%s), uid %d: obsolete pre-RFC2553 " 1767 "sockaddr_in6 rejected", 1768 p->p_pid, p->p_comm, uid); 1769 return EINVAL; 1770 } 1771 namelen = sizeof(struct sockaddr_in6); 1772 sin6->sin6_scope_id = 0; 1773 } 1774 1775 /* 1776 * Linux is less strict than NetBSD and permits namelen to be larger 1777 * than valid struct sockaddr_in*. If this is the case, truncate 1778 * the value to the correct size, so that NetBSD networking does not 1779 * return an error. 1780 */ 1781 switch (bdom) { 1782 case AF_INET: 1783 namelen = MIN(namelen, sizeof(struct sockaddr_in)); 1784 break; 1785 case AF_INET6: 1786 namelen = MIN(namelen, sizeof(struct sockaddr_in6)); 1787 break; 1788 } 1789 1790 sb->sb_family = bdom; 1791 sb->sb_len = namelen; 1792 ktrkuser("mbsoname", sb, namelen); 1793 return 0; 1794 } 1795 1796 static int 1797 linux_sa_put(struct osockaddr *osa) 1798 { 1799 struct sockaddr sa; 1800 struct osockaddr *kosa; 1801 int error, bdom, len; 1802 1803 /* 1804 * Only read/write the sockaddr family and length part, the rest is 1805 * not changed. 1806 */ 1807 len = sizeof(sa.sa_len) + sizeof(sa.sa_family); 1808 1809 error = copyin(osa, &sa, len); 1810 if (error) 1811 return (error); 1812 1813 bdom = bsd_to_linux_domain(sa.sa_family); 1814 if (bdom == -1) 1815 return (EINVAL); 1816 1817 /* Note: we convert from sockaddr to osockaddr here, too */ 1818 kosa = (struct osockaddr *) &sa; 1819 kosa->sa_family = bdom; 1820 error = copyout(kosa, osa, len); 1821 if (error) 1822 return (error); 1823 1824 return (0); 1825 } 1826 1827 #if !defined(__aarch64__) && !defined(__amd64__) 1828 int 1829 linux_sys_recv(struct lwp *l, const struct linux_sys_recv_args *uap, register_t *retval) 1830 { 1831 /* { 1832 syscallarg(int) s; 1833 syscallarg(void *) buf; 1834 syscallarg(int) len; 1835 syscallarg(int) flags; 1836 } */ 1837 struct sys_recvfrom_args bra; 1838 1839 memset(&bra, 0, sizeof(bra)); 1840 SCARG(&bra, s) = SCARG(uap, s); 1841 SCARG(&bra, buf) = SCARG(uap, buf); 1842 SCARG(&bra, len) = (size_t) SCARG(uap, len); 1843 SCARG(&bra, flags) = SCARG(uap, flags); 1844 SCARG(&bra, from) = NULL; 1845 SCARG(&bra, fromlenaddr) = NULL; 1846 1847 return (sys_recvfrom(l, &bra, retval)); 1848 } 1849 1850 int 1851 linux_sys_send(struct lwp *l, const struct linux_sys_send_args *uap, register_t *retval) 1852 { 1853 /* { 1854 syscallarg(int) s; 1855 syscallarg(void *) buf; 1856 syscallarg(int) len; 1857 syscallarg(int) flags; 1858 } */ 1859 struct sys_sendto_args bsa; 1860 1861 memset(&bsa, 0, sizeof(bsa)); 1862 SCARG(&bsa, s) = SCARG(uap, s); 1863 SCARG(&bsa, buf) = SCARG(uap, buf); 1864 SCARG(&bsa, len) = SCARG(uap, len); 1865 SCARG(&bsa, flags) = SCARG(uap, flags); 1866 SCARG(&bsa, to) = NULL; 1867 SCARG(&bsa, tolen) = 0; 1868 1869 return (sys_sendto(l, &bsa, retval)); 1870 } 1871 #endif 1872 1873 int 1874 linux_sys_accept(struct lwp *l, const struct linux_sys_accept_args *uap, register_t *retval) 1875 { 1876 /* { 1877 syscallarg(int) s; 1878 syscallarg(struct osockaddr *) name; 1879 syscallarg(int *) anamelen; 1880 } */ 1881 int error; 1882 struct sys_accept_args baa; 1883 1884 memset(&baa, 0, sizeof(baa)); 1885 SCARG(&baa, s) = SCARG(uap, s); 1886 SCARG(&baa, name) = (struct sockaddr *) SCARG(uap, name); 1887 SCARG(&baa, anamelen) = (unsigned int *) SCARG(uap, anamelen); 1888 1889 if ((error = sys_accept(l, &baa, retval))) 1890 return (error); 1891 1892 if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name)))) 1893 return (error); 1894 1895 return (0); 1896 } 1897 1898 int 1899 linux_sys_accept4(struct lwp *l, const struct linux_sys_accept4_args *uap, register_t *retval) 1900 { 1901 /* { 1902 syscallarg(int) s; 1903 syscallarg(struct osockaddr *) name; 1904 syscallarg(int *) anamelen; 1905 syscallarg(int) flags; 1906 } */ 1907 int error, flags; 1908 struct sockaddr_big name; 1909 1910 if ((flags = linux_to_bsd_type(SCARG(uap, flags))) == -1) 1911 return EINVAL; 1912 1913 name.sb_len = UCHAR_MAX; 1914 error = do_sys_accept(l, SCARG(uap, s), (struct sockaddr *)&name, 1915 retval, NULL, flags, 0); 1916 if (error != 0) 1917 return error; 1918 1919 error = copyout_sockname_sb((struct sockaddr *)SCARG(uap, name), 1920 SCARG(uap, anamelen), MSG_LENUSRSPACE, &name); 1921 if (error != 0) { 1922 int fd = (int)*retval; 1923 if (fd_getfile(fd) != NULL) 1924 (void)fd_close(fd); 1925 return error; 1926 } 1927 if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name)))) 1928 return error; 1929 1930 return 0; 1931 } 1932 1933 int 1934 linux_sys_sendmmsg(struct lwp *l, const struct linux_sys_sendmmsg_args *uap, 1935 register_t *retval) 1936 { 1937 /* { 1938 syscallarg(int) s; 1939 syscallarg(struct linux_mmsghdr *) msgvec; 1940 syscallarg(unsigned int) vlen; 1941 syscallarg(unsigned int) flags; 1942 } */ 1943 struct linux_mmsghdr lmsg; 1944 struct mmsghdr bmsg; 1945 struct socket *so; 1946 file_t *fp; 1947 struct msghdr *msg = &bmsg.msg_hdr; 1948 int error, s; 1949 unsigned int vlen, flags, dg; 1950 1951 if ((flags = linux_to_bsd_msg_flags(SCARG(uap, flags))) == -1) 1952 return EINVAL; 1953 1954 flags = (flags & MSG_USERFLAGS) | MSG_IOVUSRSPACE; 1955 1956 s = SCARG(uap, s); 1957 if ((error = fd_getsock1(s, &so, &fp)) != 0) 1958 return error; 1959 1960 vlen = SCARG(uap, vlen); 1961 if (vlen > 1024) 1962 vlen = 1024; 1963 1964 for (dg = 0; dg < vlen;) { 1965 error = copyin(SCARG(uap, msgvec) + dg, &lmsg, sizeof(lmsg)); 1966 if (error) 1967 break; 1968 error = linux_to_bsd_msghdr(&lmsg.msg_hdr, &bmsg.msg_hdr); 1969 if (error) 1970 break; 1971 1972 msg->msg_flags = flags; 1973 1974 error = do_sys_sendmsg_so(l, s, so, fp, msg, flags, retval); 1975 if (error) 1976 break; 1977 1978 ktrkuser("msghdr", msg, sizeof *msg); 1979 lmsg.msg_len = *retval; 1980 error = copyout(&lmsg, SCARG(uap, msgvec) + dg, sizeof(lmsg)); 1981 if (error) 1982 break; 1983 dg++; 1984 1985 } 1986 1987 *retval = dg; 1988 1989 fd_putfile(s); 1990 1991 /* 1992 * If we succeeded at least once, return 0. 1993 */ 1994 if (dg) 1995 return 0; 1996 return error; 1997 } 1998 1999 int 2000 linux_sys_recvmmsg(struct lwp *l, const struct linux_sys_recvmmsg_args *uap, 2001 register_t *retval) 2002 { 2003 /* { 2004 syscallarg(int) s; 2005 syscallarg(struct linux_mmsghdr *) msgvec; 2006 syscallarg(unsigned int) vlen; 2007 syscallarg(unsigned int) flags; 2008 syscallarg(struct linux_timespec *) timeout; 2009 } */ 2010 struct linux_mmsghdr lmsg; 2011 struct mmsghdr bmsg; 2012 struct socket *so; 2013 struct msghdr *msg = &bmsg.msg_hdr; 2014 int error, s; 2015 struct mbuf *from, *control; 2016 struct timespec ts = {0}, now; 2017 struct linux_timespec lts; 2018 unsigned int vlen, flags, dg; 2019 2020 if (SCARG(uap, timeout)) { 2021 error = copyin(SCARG(uap, timeout), <s, sizeof(lts)); 2022 if (error) 2023 return error; 2024 ts.tv_sec = lts.tv_sec; 2025 ts.tv_nsec = lts.tv_nsec; 2026 getnanotime(&now); 2027 timespecadd(&now, &ts, &ts); 2028 } 2029 2030 s = SCARG(uap, s); 2031 if ((error = fd_getsock(s, &so)) != 0) 2032 return error; 2033 2034 /* 2035 * If so->so_rerror holds a deferred error return it now. 2036 */ 2037 if (so->so_rerror) { 2038 error = so->so_rerror; 2039 so->so_rerror = 0; 2040 fd_putfile(s); 2041 return error; 2042 } 2043 2044 vlen = SCARG(uap, vlen); 2045 if (vlen > 1024) 2046 vlen = 1024; 2047 2048 from = NULL; 2049 flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE; 2050 2051 for (dg = 0; dg < vlen;) { 2052 error = copyin(SCARG(uap, msgvec) + dg, &lmsg, sizeof(lmsg)); 2053 if (error) 2054 break; 2055 error = linux_to_bsd_msghdr(&lmsg.msg_hdr, &bmsg.msg_hdr); 2056 if (error) 2057 break; 2058 msg->msg_flags = flags & ~MSG_WAITFORONE; 2059 2060 if (from != NULL) { 2061 m_free(from); 2062 from = NULL; 2063 } 2064 2065 error = do_sys_recvmsg_so(l, s, so, msg, &from, 2066 msg->msg_control != NULL ? &control : NULL, retval); 2067 if (error) { 2068 if (error == EAGAIN && dg > 0) 2069 error = 0; 2070 break; 2071 } 2072 2073 if (msg->msg_control != NULL) 2074 error = linux_copyout_msg_control(l, msg, control); 2075 if (error) 2076 break; 2077 2078 if (from != NULL) { 2079 mtod(from, struct osockaddr *)->sa_family = 2080 bsd_to_linux_domain(mtod(from, 2081 struct sockaddr *)->sa_family); 2082 error = copyout_sockname(msg->msg_name, 2083 &msg->msg_namelen, 0, from); 2084 if (error) 2085 break; 2086 } 2087 2088 2089 lmsg.msg_len = *retval; 2090 ktrkuser("msghdr", msg, sizeof(*msg)); 2091 bsd_to_linux_msghdr(msg, &lmsg.msg_hdr); 2092 error = copyout(&lmsg, SCARG(uap, msgvec) + dg, sizeof(lmsg)); 2093 if (error) 2094 break; 2095 2096 dg++; 2097 if (msg->msg_flags & MSG_OOB) 2098 break; 2099 2100 if (SCARG(uap, timeout)) { 2101 getnanotime(&now); 2102 timespecsub(&now, &ts, &now); 2103 if (now.tv_sec > 0) 2104 break; 2105 } 2106 2107 if (flags & MSG_WAITFORONE) 2108 flags |= MSG_DONTWAIT; 2109 2110 } 2111 2112 if (from != NULL) 2113 m_free(from); 2114 2115 *retval = dg; 2116 2117 /* 2118 * If we succeeded at least once, return 0, hopefully so->so_rerror 2119 * will catch it next time. 2120 */ 2121 if (error && dg > 0) { 2122 so->so_rerror = error; 2123 error = 0; 2124 } 2125 2126 fd_putfile(s); 2127 2128 return error; 2129 } 2130
Indexes created Sun Oct 25 00:26:45 UTC 2026