1 /* $NetBSD: linux_socket.c,v 1.163 2026/09/20 13:43:51 riastradh 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 1.163 2026/09/20 13:43:51 riastradh 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 struct sys_recvfrom_args bra; 718 719 memset(&bra, 0, sizeof(bra)); 720 SCARG(&bra, s) = SCARG(uap, s); 721 SCARG(&bra, buf) = SCARG(uap, buf); 722 SCARG(&bra, len) = SCARG(uap, len); 723 SCARG(&bra, flags) = SCARG(uap, flags); 724 SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from); 725 SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr); 726 727 if ((error = sys_recvfrom(l, &bra, retval))) 728 return (error); 729 730 if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from)))) 731 return (error); 732 733 return (0); 734 } 735 736 static int 737 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control) 738 { 739 socklen_t dlen; 740 int error = 0; 741 struct cmsghdr *cmsg; 742 struct linux_cmsghdr linux_cmsg; 743 struct mbuf *m; 744 char *q, *q_end; 745 746 if (mp->msg_controllen <= 0 || control == 0) { 747 mp->msg_controllen = 0; 748 free_control_mbuf(l, control, control); 749 return 0; 750 } 751 752 ktrkuser("msgcontrol", mtod(control, void *), mp->msg_controllen); 753 754 q = (char *)mp->msg_control; 755 q_end = q + mp->msg_controllen; 756 757 for (m = control; m != NULL; ) { 758 cmsg = mtod(m, struct cmsghdr *); 759 760 /* 761 * Fixup cmsg. We handle two things: 762 * 0. different sizeof cmsg_len. 763 * 1. different values for level/type on some archs 764 * 2. different alignment of CMSG_DATA on some archs 765 */ 766 memset(&linux_cmsg, 0, sizeof(linux_cmsg)); 767 CTASSERT(LINUX_CMSG_ALIGN_DELTA >= 0); 768 CTASSERT(CMSG_LEN(0) == 769 sizeof(linux_cmsg) + LINUX_CMSG_ALIGN_DELTA); 770 KASSERTMSG(cmsg->cmsg_len >= CMSG_LEN(0), 771 "cmsg->cmsg_len=%u", cmsg->cmsg_len); 772 linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA; 773 KASSERTMSG(linux_cmsg.cmsg_len >= sizeof(linux_cmsg), 774 "linux_cmsg.cmsg_len=%zu", linux_cmsg.cmsg_len); 775 linux_cmsg.cmsg_level = cmsg->cmsg_level; 776 linux_cmsg.cmsg_type = cmsg->cmsg_type; 777 778 dlen = q_end - q; 779 if (linux_cmsg.cmsg_len > dlen) { 780 /* Not enough room for the parameter */ 781 if (dlen < sizeof linux_cmsg) 782 /* Discard if header wont fit */ 783 break; 784 dlen -= sizeof linux_cmsg; 785 mp->msg_flags |= MSG_CTRUNC; 786 if (linux_cmsg.cmsg_level == SOL_SOCKET 787 && linux_cmsg.cmsg_type == SCM_RIGHTS) 788 /* Do not truncate me ... */ 789 break; 790 } else 791 dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg; 792 793 /* 794 * NetBSD will not create a cmsg record whose header 795 * length plus aligned content length would overflow 796 * socklen_t. This justifies computing 797 * LINUX_CMSG_SPACE(dlen) (and LINUX_CMSG_LEN) below 798 * without overflow. 799 */ 800 CTASSERT(LINUX_CMSG_SPACE(1) <= CMSG_SPACE(1)); 801 CTASSERT(CMSG_SPACE(0) <= 802 __type_max(__typeof(cmsg->cmsg_len))); 803 KASSERTMSG((dlen <= __type_max(__typeof(dlen)) - 804 CMSG_SPACE(0)), 805 "dlen=%u", dlen); 806 CTASSERT(LINUX_CMSG_LEN(1) <= LINUX_CMSG_SPACE(1)); 807 808 switch (linux_cmsg.cmsg_level) { 809 case SOL_SOCKET: 810 linux_cmsg.cmsg_level = LINUX_SOL_SOCKET; 811 switch (linux_cmsg.cmsg_type) { 812 case SCM_RIGHTS: 813 /* Linux SCM_RIGHTS is same as NetBSD */ 814 break; 815 816 default: 817 /* other types not supported */ 818 error = EINVAL; 819 goto done; 820 } 821 /* machine dependent ! */ 822 break; 823 default: 824 /* pray and leave intact */ 825 break; 826 } 827 828 /* There can be padding between the header and data... */ 829 error = copyout(&linux_cmsg, q, sizeof linux_cmsg); 830 if (error != 0) { 831 break; 832 } 833 error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg, dlen); 834 if (error != 0) { 835 break; 836 } 837 m = m->m_next; 838 if (m == NULL || LINUX_CMSG_SPACE(dlen) > q_end - q) { 839 q += LINUX_CMSG_LEN(dlen); 840 break; 841 } 842 q += LINUX_CMSG_SPACE(dlen); 843 } 844 845 done: 846 if (error) { 847 /* We must free all the SCM_RIGHTS */ 848 m = control; 849 } 850 free_control_mbuf(l, control, m); 851 852 mp->msg_controllen = q - (char *)mp->msg_control; 853 return error; 854 } 855 856 int 857 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval) 858 { 859 /* { 860 syscallarg(int) s; 861 syscallarg(struct linux_msghdr *) msg; 862 syscallarg(u_int) flags; 863 } */ 864 struct msghdr msg; 865 struct linux_msghdr lmsg; 866 int error; 867 struct mbuf *from, *control; 868 869 error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg)); 870 if (error) 871 return error; 872 error = linux_to_bsd_msghdr(&lmsg, &msg); 873 if (error) 874 return error; 875 876 msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags)); 877 if (msg.msg_flags < 0) { 878 /* Some unsupported flag */ 879 return (EINVAL); 880 } 881 msg.msg_flags |= MSG_IOVUSRSPACE; 882 883 error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from, 884 msg.msg_control != NULL ? &control : NULL, retval); 885 if (error != 0) 886 return error; 887 888 if (msg.msg_control != NULL) 889 error = linux_copyout_msg_control(l, &msg, control); 890 891 if (error == 0 && from != 0) { 892 mtod(from, struct osockaddr *)->sa_family = 893 bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family); 894 error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0, 895 from); 896 } else 897 msg.msg_namelen = 0; 898 899 if (from != NULL) 900 m_free(from); 901 902 if (error == 0) { 903 msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags); 904 if (msg.msg_flags < 0) 905 /* Some flag unsupported by Linux */ 906 error = EINVAL; 907 else { 908 ktrkuser("msghdr", &msg, sizeof(msg)); 909 bsd_to_linux_msghdr(&msg, &lmsg); 910 error = copyout(&lmsg, SCARG(uap, msg), sizeof(lmsg)); 911 } 912 } 913 914 return (error); 915 } 916 917 /* 918 * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET 919 * is different, the rest matches IPPROTO_* on both systems. 920 */ 921 int 922 linux_to_bsd_sopt_level(int llevel) 923 { 924 925 switch (llevel) { 926 case LINUX_SOL_SOCKET: 927 return SOL_SOCKET; 928 case LINUX_SOL_IP: 929 return IPPROTO_IP; 930 #ifdef INET6 931 case LINUX_SOL_IPV6: 932 return IPPROTO_IPV6; 933 #endif 934 case LINUX_SOL_TCP: 935 return IPPROTO_TCP; 936 case LINUX_SOL_UDP: 937 return IPPROTO_UDP; 938 default: 939 return -1; 940 } 941 } 942 943 /* 944 * Convert Linux socket level socket option numbers to NetBSD values. 945 */ 946 int 947 linux_to_bsd_so_sockopt(int lopt) 948 { 949 950 switch (lopt) { 951 case LINUX_SO_DEBUG: 952 return SO_DEBUG; 953 case LINUX_SO_REUSEADDR: 954 /* 955 * Linux does not implement SO_REUSEPORT, but allows reuse of 956 * a host:port pair through SO_REUSEADDR even if the address 957 * is not a multicast-address. Effectively, this means that we 958 * should use SO_REUSEPORT to allow Linux applications to not 959 * exit with EADDRINUSE 960 */ 961 return SO_REUSEPORT; 962 case LINUX_SO_TYPE: 963 return SO_TYPE; 964 case LINUX_SO_ERROR: 965 return SO_ERROR; 966 case LINUX_SO_DONTROUTE: 967 return SO_DONTROUTE; 968 case LINUX_SO_BROADCAST: 969 return SO_BROADCAST; 970 case LINUX_SO_SNDBUF: 971 return SO_SNDBUF; 972 case LINUX_SO_RCVBUF: 973 return SO_RCVBUF; 974 case LINUX_SO_KEEPALIVE: 975 return SO_KEEPALIVE; 976 case LINUX_SO_OOBINLINE: 977 return SO_OOBINLINE; 978 case LINUX_SO_NO_CHECK: 979 case LINUX_SO_PRIORITY: 980 return -1; 981 case LINUX_SO_LINGER: 982 return SO_LINGER; 983 case LINUX_SO_BSDCOMPAT: 984 case LINUX_SO_PASSCRED: 985 case LINUX_SO_PEERCRED: 986 return -1; 987 case LINUX_SO_RCVLOWAT: 988 return SO_RCVLOWAT; 989 case LINUX_SO_SNDLOWAT: 990 return SO_SNDLOWAT; 991 case LINUX_SO_RCVTIMEO: 992 return SO_RCVTIMEO; 993 case LINUX_SO_SNDTIMEO: 994 return SO_SNDTIMEO; 995 case LINUX_SO_SECURITY_AUTHENTICATION: 996 case LINUX_SO_SECURITY_ENCRYPTION_TRANSPORT: 997 case LINUX_SO_SECURITY_ENCRYPTION_NETWORK: 998 case LINUX_SO_BINDTODEVICE: 999 case LINUX_SO_ATTACH_FILTER: 1000 case LINUX_SO_DETACH_FILTER: 1001 case LINUX_SO_PEERNAME: 1002 return -1; 1003 case LINUX_SO_TIMESTAMP: 1004 return SO_TIMESTAMP; 1005 case LINUX_SO_ACCEPTCONN: 1006 case LINUX_SO_PEERSEC: 1007 case LINUX_SO_SNDBUFFORCE: 1008 case LINUX_SO_RCVBUFFORCE: 1009 case LINUX_SO_PASSSEC: 1010 case LINUX_SO_TIMESTAMPNS: 1011 case LINUX_SO_MARK: 1012 case LINUX_SO_TIMESTAMPING: 1013 case LINUX_SO_PROTOCOL: 1014 case LINUX_SO_DOMAIN: 1015 case LINUX_SO_RXQ_OVFL: 1016 case LINUX_SO_WIFI_STATUS: 1017 case LINUX_SO_PEEK_OFF: 1018 case LINUX_SO_NOFCS: 1019 default: 1020 return -1; 1021 } 1022 } 1023 1024 /* 1025 * Convert Linux IP level socket option number to NetBSD values. 1026 */ 1027 int 1028 linux_to_bsd_ip_sockopt(int lopt) 1029 { 1030 1031 switch (lopt) { 1032 case LINUX_IP_TOS: 1033 return IP_TOS; 1034 case LINUX_IP_TTL: 1035 return IP_TTL; 1036 case LINUX_IP_RETOPTS: 1037 return IP_RETOPTS; 1038 case LINUX_IP_PKTINFO: 1039 return IP_PKTINFO; 1040 case LINUX_IP_RECVOPTS: 1041 return IP_RECVOPTS; 1042 case LINUX_IP_HDRINCL: 1043 return IP_HDRINCL; 1044 case LINUX_IP_MULTICAST_TTL: 1045 return IP_MULTICAST_TTL; 1046 case LINUX_IP_MULTICAST_LOOP: 1047 return IP_MULTICAST_LOOP; 1048 case LINUX_IP_MULTICAST_IF: 1049 return IP_MULTICAST_IF; 1050 case LINUX_IP_ADD_MEMBERSHIP: 1051 return IP_ADD_MEMBERSHIP; 1052 case LINUX_IP_DROP_MEMBERSHIP: 1053 return IP_DROP_MEMBERSHIP; 1054 case LINUX_IP_RECVERR: 1055 case LINUX_IP_FREEBIND: 1056 return -2; /* ignored */ 1057 case LINUX_IP_MULTICAST_ALL: 1058 return -3; /* noprotoopt */ 1059 default: 1060 return -1; 1061 } 1062 } 1063 1064 /* 1065 * Convert Linux IPV6 level socket option number to NetBSD values. 1066 */ 1067 #ifdef INET6 1068 int 1069 linux_to_bsd_ipv6_sockopt(int lopt) 1070 { 1071 1072 switch (lopt) { 1073 case LINUX_IPV6_V6ONLY: 1074 return IPV6_V6ONLY; 1075 case LINUX_IPV6_MULTICAST_HOPS: 1076 return IPV6_MULTICAST_HOPS; 1077 case LINUX_IPV6_MULTICAST_ALL: 1078 return -3; /* noprotoopt */ 1079 default: 1080 return -1; 1081 } 1082 } 1083 #endif 1084 1085 /* 1086 * Convert Linux TCP level socket option number to NetBSD values. 1087 */ 1088 int 1089 linux_to_bsd_tcp_sockopt(int lopt) 1090 { 1091 1092 switch (lopt) { 1093 case LINUX_TCP_NODELAY: 1094 return TCP_NODELAY; 1095 case LINUX_TCP_MAXSEG: 1096 return TCP_MAXSEG; 1097 default: 1098 return -1; 1099 } 1100 } 1101 1102 /* 1103 * Convert Linux UDP level socket option number to NetBSD values. 1104 */ 1105 int 1106 linux_to_bsd_udp_sockopt(int lopt) 1107 { 1108 1109 switch (lopt) { 1110 default: 1111 return -1; 1112 } 1113 } 1114 1115 /* 1116 * Another reasonably straightforward function: setsockopt(2). 1117 * The level and option numbers are converted; the values passed 1118 * are not (yet) converted, the ones currently implemented don't 1119 * need conversion, as they are the same on both systems. 1120 */ 1121 int 1122 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval) 1123 { 1124 /* { 1125 syscallarg(int) s; 1126 syscallarg(int) level; 1127 syscallarg(int) optname; 1128 syscallarg(void *) optval; 1129 syscallarg(int) optlen; 1130 } */ 1131 struct sys_setsockopt_args bsa; 1132 int name; 1133 1134 memset(&bsa, 0, sizeof(bsa)); 1135 SCARG(&bsa, s) = SCARG(uap, s); 1136 SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level)); 1137 SCARG(&bsa, val) = SCARG(uap, optval); 1138 SCARG(&bsa, valsize) = SCARG(uap, optlen); 1139 1140 /* 1141 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets 1142 * and returns EOPNOTSUPP for other levels 1143 */ 1144 if (SCARG(&bsa, level) != SOL_SOCKET) { 1145 struct socket *so; 1146 int error, family; 1147 1148 /* fd_getsock() will use the descriptor for us */ 1149 if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0) 1150 return error; 1151 family = so->so_proto->pr_domain->dom_family; 1152 fd_putfile(SCARG(&bsa, s)); 1153 1154 if (family == AF_LOCAL) 1155 return EOPNOTSUPP; 1156 } 1157 1158 switch (SCARG(&bsa, level)) { 1159 case SOL_SOCKET: 1160 name = linux_to_bsd_so_sockopt(SCARG(uap, optname)); 1161 break; 1162 case IPPROTO_IP: 1163 name = linux_to_bsd_ip_sockopt(SCARG(uap, optname)); 1164 break; 1165 #ifdef INET6 1166 case IPPROTO_IPV6: 1167 name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname)); 1168 break; 1169 #endif 1170 case IPPROTO_TCP: 1171 name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname)); 1172 break; 1173 case IPPROTO_UDP: 1174 name = linux_to_bsd_udp_sockopt(SCARG(uap, optname)); 1175 break; 1176 default: 1177 return EINVAL; 1178 } 1179 1180 switch (name) { 1181 case -1: 1182 return EINVAL; 1183 case -2: 1184 return 0; 1185 case -3: 1186 return ENOPROTOOPT; 1187 } 1188 SCARG(&bsa, name) = name; 1189 1190 return sys_setsockopt(l, &bsa, retval); 1191 } 1192 1193 /* 1194 * getsockopt(2) is very much the same as setsockopt(2) (see above) 1195 */ 1196 int 1197 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval) 1198 { 1199 /* { 1200 syscallarg(int) s; 1201 syscallarg(int) level; 1202 syscallarg(int) optname; 1203 syscallarg(void *) optval; 1204 syscallarg(int *) optlen; 1205 } */ 1206 struct sys_getsockopt_args bga; 1207 int name; 1208 1209 memset(&bga, 0, sizeof(bga)); 1210 SCARG(&bga, s) = SCARG(uap, s); 1211 SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level)); 1212 SCARG(&bga, val) = SCARG(uap, optval); 1213 SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen); 1214 1215 switch (SCARG(&bga, level)) { 1216 case SOL_SOCKET: 1217 name = linux_to_bsd_so_sockopt(SCARG(uap, optname)); 1218 break; 1219 case IPPROTO_IP: 1220 name = linux_to_bsd_ip_sockopt(SCARG(uap, optname)); 1221 break; 1222 #ifdef INET6 1223 case IPPROTO_IPV6: 1224 name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname)); 1225 break; 1226 #endif 1227 case IPPROTO_TCP: 1228 name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname)); 1229 break; 1230 case IPPROTO_UDP: 1231 name = linux_to_bsd_udp_sockopt(SCARG(uap, optname)); 1232 break; 1233 default: 1234 return EINVAL; 1235 } 1236 1237 switch (name) { 1238 case -1: 1239 case -2: /* we can't ignore, since we don't know what to return */ 1240 return EINVAL; 1241 case -3: 1242 return ENOPROTOOPT; 1243 } 1244 SCARG(&bga, name) = name; 1245 1246 return sys_getsockopt(l, &bga, retval); 1247 } 1248 1249 int 1250 linux_getifname(struct lwp *l, register_t *retval, void *data) 1251 { 1252 struct ifnet *ifp; 1253 struct linux_ifreq ifr; 1254 int error; 1255 int s; 1256 1257 error = copyin(data, &ifr, sizeof(ifr)); 1258 if (error) 1259 return error; 1260 1261 s = pserialize_read_enter(); 1262 ifp = if_byindex(ifr.ifr_ifru.ifru_ifindex); 1263 if (ifp == NULL) { 1264 pserialize_read_exit(s); 1265 return ENODEV; 1266 } 1267 1268 strncpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)); 1269 pserialize_read_exit(s); 1270 1271 return copyout(&ifr, data, sizeof(ifr)); 1272 } 1273 1274 int 1275 linux_getifconf(struct lwp *l, register_t *retval, void *data) 1276 { 1277 struct linux_ifreq ifr, *ifrp = NULL; 1278 struct linux_ifconf ifc; 1279 struct ifnet *ifp; 1280 struct sockaddr *sa; 1281 struct osockaddr *osa; 1282 int space = 0, error; 1283 const int sz = (int)sizeof(ifr); 1284 bool docopy; 1285 int s; 1286 int bound; 1287 struct psref psref; 1288 1289 error = copyin(data, &ifc, sizeof(ifc)); 1290 if (error) 1291 return error; 1292 1293 docopy = ifc.ifc_req != NULL; 1294 if (docopy) { 1295 if (ifc.ifc_len < 0) 1296 return EINVAL; 1297 1298 space = ifc.ifc_len; 1299 ifrp = ifc.ifc_req; 1300 } 1301 memset(&ifr, 0, sizeof(ifr)); 1302 1303 bound = curlwp_bind(); 1304 s = pserialize_read_enter(); 1305 IFNET_READER_FOREACH(ifp) { 1306 struct ifaddr *ifa; 1307 if_acquire(ifp, &psref); 1308 pserialize_read_exit(s); 1309 1310 (void)strncpy(ifr.ifr_name, ifp->if_xname, 1311 sizeof(ifr.ifr_name)); 1312 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') { 1313 error = ENAMETOOLONG; 1314 goto release_exit; 1315 } 1316 1317 s = pserialize_read_enter(); 1318 IFADDR_READER_FOREACH(ifa, ifp) { 1319 struct psref psref_ifa; 1320 ifa_acquire(ifa, &psref_ifa); 1321 pserialize_read_exit(s); 1322 1323 sa = ifa->ifa_addr; 1324 if (sa->sa_family != AF_INET || 1325 sa->sa_len > sizeof(*osa)) 1326 goto next; 1327 memcpy(&ifr.ifr_addr, sa, sa->sa_len); 1328 osa = (struct osockaddr *)&ifr.ifr_addr; 1329 osa->sa_family = sa->sa_family; 1330 if (space >= sz) { 1331 error = copyout(&ifr, ifrp, sz); 1332 if (error != 0) { 1333 ifa_release(ifa, &psref_ifa); 1334 goto release_exit; 1335 } 1336 ifrp++; 1337 } 1338 space -= sz; 1339 next: 1340 s = pserialize_read_enter(); 1341 ifa_release(ifa, &psref_ifa); 1342 } 1343 1344 KASSERT(pserialize_in_read_section()); 1345 if_release(ifp, &psref); 1346 } 1347 pserialize_read_exit(s); 1348 curlwp_bindx(bound); 1349 1350 if (docopy) 1351 ifc.ifc_len -= space; 1352 else 1353 ifc.ifc_len = -space; 1354 1355 return copyout(&ifc, data, sizeof(ifc)); 1356 1357 release_exit: 1358 if_release(ifp, &psref); 1359 curlwp_bindx(bound); 1360 return error; 1361 } 1362 1363 int 1364 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd, 1365 void *data) 1366 { 1367 /* Not the full structure, just enough to map what we do here */ 1368 struct linux_ifreq lreq; 1369 file_t *fp; 1370 struct ifaddr *ifa; 1371 struct ifnet *ifp; 1372 struct sockaddr_dl *sadl; 1373 int error, found; 1374 int index, ifnum; 1375 int s; 1376 1377 /* 1378 * We can't emulate this ioctl by calling sys_ioctl() to run 1379 * SIOCGIFCONF, because the user buffer is not of the right 1380 * type to take those results. We can't use kernel buffers to 1381 * receive the results, as the implementation of sys_ioctl() 1382 * and ifconf() [which implements SIOCGIFCONF] use 1383 * copyin()/copyout() which will fail on kernel addresses. 1384 * 1385 * So, we must duplicate code from sys_ioctl() and ifconf(). Ugh. 1386 */ 1387 1388 if ((fp = fd_getfile(fd)) == NULL) 1389 return (EBADF); 1390 1391 KERNEL_LOCK(1, NULL); 1392 1393 if ((fp->f_flag & (FREAD | FWRITE)) == 0) { 1394 error = EBADF; 1395 goto out; 1396 } 1397 1398 error = copyin(data, &lreq, sizeof(lreq)); 1399 if (error) 1400 goto out; 1401 lreq.ifr_name[LINUX_IFNAMSIZ-1] = '\0'; /* just in case */ 1402 1403 /* 1404 * Try real interface name first, then fake "ethX" 1405 */ 1406 found = 0; 1407 s = pserialize_read_enter(); 1408 IFNET_READER_FOREACH(ifp) { 1409 if (found) 1410 break; 1411 if (strcmp(lreq.ifr_name, ifp->if_xname)) 1412 /* not this interface */ 1413 continue; 1414 1415 found=1; 1416 if (IFADDR_READER_EMPTY(ifp)) { 1417 pserialize_read_exit(s); 1418 error = ENODEV; 1419 goto out; 1420 } 1421 IFADDR_READER_FOREACH(ifa, ifp) { 1422 sadl = satosdl(ifa->ifa_addr); 1423 /* only return ethernet addresses */ 1424 /* XXX what about FDDI, etc. ? */ 1425 if (sadl->sdl_family != AF_LINK || 1426 sadl->sdl_type != IFT_ETHER) 1427 continue; 1428 memcpy(&lreq.ifr_hwaddr.sa_data, CLLADDR(sadl), 1429 MIN(sadl->sdl_alen, 1430 sizeof(lreq.ifr_hwaddr.sa_data))); 1431 lreq.ifr_hwaddr.sa_family = 1432 sadl->sdl_family; 1433 pserialize_read_exit(s); 1434 1435 error = copyout(&lreq, data, sizeof(lreq)); 1436 goto out; 1437 } 1438 } 1439 pserialize_read_exit(s); 1440 1441 if (strncmp(lreq.ifr_name, "eth", 3) != 0) { 1442 /* unknown interface, not even an "eth*" name */ 1443 error = ENODEV; 1444 goto out; 1445 } 1446 1447 for (ifnum = 0, index = 3; 1448 index < LINUX_IFNAMSIZ && lreq.ifr_name[index] != '\0'; 1449 index++) { 1450 ifnum *= 10; 1451 ifnum += lreq.ifr_name[index] - '0'; 1452 } 1453 1454 error = EINVAL; /* in case we don't find one */ 1455 s = pserialize_read_enter(); 1456 IFNET_READER_FOREACH(ifp) { 1457 memcpy(lreq.ifr_name, ifp->if_xname, 1458 MIN(LINUX_IFNAMSIZ, IFNAMSIZ)); 1459 IFADDR_READER_FOREACH(ifa, ifp) { 1460 sadl = satosdl(ifa->ifa_addr); 1461 /* only return ethernet addresses */ 1462 /* XXX what about FDDI, etc. ? */ 1463 if (sadl->sdl_family != AF_LINK || 1464 sadl->sdl_type != IFT_ETHER) 1465 continue; 1466 if (ifnum--) 1467 /* not the requested iface */ 1468 continue; 1469 memcpy(&lreq.ifr_hwaddr.sa_data, 1470 CLLADDR(sadl), 1471 MIN(sadl->sdl_alen, 1472 sizeof(lreq.ifr_hwaddr.sa_data))); 1473 lreq.ifr_hwaddr.sa_family = 1474 sadl->sdl_family; 1475 pserialize_read_exit(s); 1476 1477 error = copyout(&lreq, data, sizeof(lreq)); 1478 goto out; 1479 } 1480 } 1481 pserialize_read_exit(s); 1482 1483 out: 1484 KERNEL_UNLOCK_ONE(NULL); 1485 fd_putfile(fd); 1486 return error; 1487 } 1488 1489 int 1490 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval) 1491 { 1492 /* { 1493 syscallarg(int) fd; 1494 syscallarg(u_long) com; 1495 syscallarg(void *) data; 1496 } */ 1497 u_long com; 1498 int error = 0, isdev = 0, dosys = 1; 1499 struct sys_ioctl_args ia; 1500 file_t *fp; 1501 struct vnode *vp; 1502 int (*ioctlf)(file_t *, u_long, void *); 1503 struct ioctl_pt pt; 1504 1505 if ((fp = fd_getfile(SCARG(uap, fd))) == NULL) 1506 return (EBADF); 1507 1508 if (fp->f_type == DTYPE_VNODE) { 1509 vp = (struct vnode *)fp->f_data; 1510 isdev = vp->v_type == VCHR; 1511 } 1512 1513 /* 1514 * Don't try to interpret socket ioctl calls that are done 1515 * on a device filedescriptor, just pass them through, to 1516 * emulate Linux behaviour. Use PTIOCLINUX so that the 1517 * device will only handle these if it's prepared to do 1518 * so, to avoid unexpected things from happening. 1519 */ 1520 if (isdev) { 1521 dosys = 0; 1522 ioctlf = fp->f_ops->fo_ioctl; 1523 pt.com = SCARG(uap, com); 1524 pt.data = SCARG(uap, data); 1525 error = ioctlf(fp, PTIOCLINUX, &pt); 1526 /* 1527 * XXX hack: if the function returns EJUSTRETURN, 1528 * it has stuffed a sysctl return value in pt.data. 1529 */ 1530 if (error == EJUSTRETURN) { 1531 retval[0] = (register_t)pt.data; 1532 error = 0; 1533 } 1534 goto out; 1535 } 1536 1537 com = SCARG(uap, com); 1538 retval[0] = 0; 1539 1540 memset(&ia, 0, sizeof(ia)); 1541 1542 switch (com) { 1543 case LINUX_SIOCGIFNAME: 1544 error = linux_getifname(l, retval, SCARG(uap, data)); 1545 dosys = 0; 1546 break; 1547 case LINUX_SIOCGIFCONF: 1548 error = linux_getifconf(l, retval, SCARG(uap, data)); 1549 dosys = 0; 1550 break; 1551 case LINUX_SIOCGIFFLAGS: 1552 SCARG(&ia, com) = OSIOCGIFFLAGS; 1553 break; 1554 case LINUX_SIOCSIFFLAGS: 1555 SCARG(&ia, com) = OSIOCSIFFLAGS; 1556 break; 1557 case LINUX_SIOCGIFADDR: 1558 SCARG(&ia, com) = OOSIOCGIFADDR; 1559 break; 1560 case LINUX_SIOCGIFDSTADDR: 1561 SCARG(&ia, com) = OOSIOCGIFDSTADDR; 1562 break; 1563 case LINUX_SIOCGIFBRDADDR: 1564 SCARG(&ia, com) = OOSIOCGIFBRDADDR; 1565 break; 1566 case LINUX_SIOCGIFNETMASK: 1567 SCARG(&ia, com) = OOSIOCGIFNETMASK; 1568 break; 1569 case LINUX_SIOCGIFMTU: 1570 SCARG(&ia, com) = OSIOCGIFMTU; 1571 break; 1572 case LINUX_SIOCADDMULTI: 1573 SCARG(&ia, com) = OSIOCADDMULTI; 1574 break; 1575 case LINUX_SIOCDELMULTI: 1576 SCARG(&ia, com) = OSIOCDELMULTI; 1577 break; 1578 case LINUX_SIOCGIFHWADDR: 1579 error = linux_getifhwaddr(l, retval, SCARG(uap, fd), 1580 SCARG(uap, data)); 1581 dosys = 0; 1582 break; 1583 default: 1584 error = EINVAL; 1585 } 1586 1587 out: 1588 fd_putfile(SCARG(uap, fd)); 1589 1590 if (error ==0 && dosys) { 1591 SCARG(&ia, fd) = SCARG(uap, fd); 1592 SCARG(&ia, data) = SCARG(uap, data); 1593 error = sys_ioctl(curlwp, &ia, retval); 1594 } 1595 1596 return error; 1597 } 1598 1599 int 1600 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval) 1601 { 1602 /* { 1603 syscallarg(int) s; 1604 syscallarg(const struct sockaddr *) name; 1605 syscallarg(int) namelen; 1606 } */ 1607 int error; 1608 struct sockaddr_big sb; 1609 1610 error = linux_get_sa(l, SCARG(uap, s), &sb, SCARG(uap, name), 1611 SCARG(uap, namelen)); 1612 if (error) 1613 return (error); 1614 1615 error = do_sys_connect(l, SCARG(uap, s), (struct sockaddr *)&sb); 1616 1617 if (error == EISCONN) { 1618 struct socket *so; 1619 int state, prflags; 1620 1621 /* fd_getsock() will use the descriptor for us */ 1622 if (fd_getsock(SCARG(uap, s), &so) != 0) 1623 return EISCONN; 1624 1625 solock(so); 1626 state = so->so_state; 1627 prflags = so->so_proto->pr_flags; 1628 sounlock(so); 1629 fd_putfile(SCARG(uap, s)); 1630 /* 1631 * We should only let this call succeed once per 1632 * non-blocking connect; however we don't have 1633 * a convenient place to keep that state.. 1634 */ 1635 if ((state & (SS_ISCONNECTED|SS_NBIO)) == 1636 (SS_ISCONNECTED|SS_NBIO) && 1637 (prflags & PR_CONNREQUIRED)) 1638 return 0; 1639 } 1640 1641 return (error); 1642 } 1643 1644 int 1645 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval) 1646 { 1647 /* { 1648 syscallarg(int) s; 1649 syscallarg(const struct osockaddr *) name; 1650 syscallarg(int) namelen; 1651 } */ 1652 int error; 1653 struct sockaddr_big sb; 1654 1655 error = linux_get_sa(l, SCARG(uap, s), &sb, SCARG(uap, name), 1656 SCARG(uap, namelen)); 1657 if (error) 1658 return (error); 1659 1660 return do_sys_bind(l, SCARG(uap, s), (struct sockaddr *)&sb); 1661 } 1662 1663 int 1664 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval) 1665 { 1666 /* { 1667 syscallarg(int) fdes; 1668 syscallarg(void *) asa; 1669 syscallarg(int *) alen; 1670 } */ 1671 int error; 1672 1673 if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0) 1674 return (error); 1675 1676 if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa)))) 1677 return (error); 1678 1679 return (0); 1680 } 1681 1682 int 1683 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval) 1684 { 1685 /* { 1686 syscallarg(int) fdes; 1687 syscallarg(void *) asa; 1688 syscallarg(int *) alen; 1689 } */ 1690 int error; 1691 1692 if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0) 1693 return (error); 1694 1695 if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa)))) 1696 return (error); 1697 1698 return (0); 1699 } 1700 1701 /* 1702 * Copy the osockaddr structure pointed to by name to sb, adjust 1703 * family and convert to sockaddr. 1704 */ 1705 static int 1706 linux_get_sa(struct lwp *l, int s, struct sockaddr_big *sb, 1707 const struct osockaddr *name, socklen_t namelen) 1708 { 1709 int error, bdom; 1710 1711 if (namelen > UCHAR_MAX || 1712 namelen <= offsetof(struct sockaddr_big, sb_data)) 1713 return EINVAL; 1714 1715 error = copyin(name, sb, namelen); 1716 if (error) 1717 return error; 1718 1719 bdom = linux_to_bsd_domain(sb->sb_family); 1720 if (bdom == -1) 1721 return EINVAL; 1722 1723 /* 1724 * If the family is unspecified, use address family of the socket. 1725 * This avoid triggering strict family checks in netinet/in_pcb.c et.al. 1726 */ 1727 if (bdom == AF_UNSPEC) { 1728 struct socket *so; 1729 1730 /* fd_getsock() will use the descriptor for us */ 1731 if ((error = fd_getsock(s, &so)) != 0) 1732 return error; 1733 1734 bdom = so->so_proto->pr_domain->dom_family; 1735 fd_putfile(s); 1736 } 1737 1738 /* 1739 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6, 1740 * which lacks the scope id compared with RFC2553 one. If we detect 1741 * the situation, reject the address and write a message to system log. 1742 * 1743 * Still accept addresses for which the scope id is not used. 1744 */ 1745 if (bdom == AF_INET6 && 1746 namelen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) { 1747 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sb; 1748 if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) && 1749 (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) || 1750 IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) || 1751 IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) || 1752 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) || 1753 IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) { 1754 struct proc *p = l->l_proc; 1755 int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1; 1756 1757 log(LOG_DEBUG, 1758 "pid %d (%s), uid %d: obsolete pre-RFC2553 " 1759 "sockaddr_in6 rejected", 1760 p->p_pid, p->p_comm, uid); 1761 return EINVAL; 1762 } 1763 namelen = sizeof(struct sockaddr_in6); 1764 sin6->sin6_scope_id = 0; 1765 } 1766 1767 /* 1768 * Linux is less strict than NetBSD and permits namelen to be larger 1769 * than valid struct sockaddr_in*. If this is the case, truncate 1770 * the value to the correct size, so that NetBSD networking does not 1771 * return an error. 1772 */ 1773 switch (bdom) { 1774 case AF_INET: 1775 namelen = MIN(namelen, sizeof(struct sockaddr_in)); 1776 break; 1777 case AF_INET6: 1778 namelen = MIN(namelen, sizeof(struct sockaddr_in6)); 1779 break; 1780 } 1781 1782 sb->sb_family = bdom; 1783 sb->sb_len = namelen; 1784 ktrkuser("mbsoname", sb, namelen); 1785 return 0; 1786 } 1787 1788 static int 1789 linux_sa_put(struct osockaddr *osa) 1790 { 1791 struct sockaddr sa; 1792 struct osockaddr *kosa; 1793 int error, bdom, len; 1794 1795 /* 1796 * Only read/write the sockaddr family and length part, the rest is 1797 * not changed. 1798 */ 1799 len = sizeof(sa.sa_len) + sizeof(sa.sa_family); 1800 1801 error = copyin(osa, &sa, len); 1802 if (error) 1803 return (error); 1804 1805 bdom = bsd_to_linux_domain(sa.sa_family); 1806 if (bdom == -1) 1807 return (EINVAL); 1808 1809 /* Note: we convert from sockaddr to osockaddr here, too */ 1810 kosa = (struct osockaddr *) &sa; 1811 kosa->sa_family = bdom; 1812 error = copyout(kosa, osa, len); 1813 if (error) 1814 return (error); 1815 1816 return (0); 1817 } 1818 1819 #if !defined(__aarch64__) && !defined(__amd64__) 1820 int 1821 linux_sys_recv(struct lwp *l, const struct linux_sys_recv_args *uap, register_t *retval) 1822 { 1823 /* { 1824 syscallarg(int) s; 1825 syscallarg(void *) buf; 1826 syscallarg(int) len; 1827 syscallarg(int) flags; 1828 } */ 1829 struct sys_recvfrom_args bra; 1830 1831 memset(&bra, 0, sizeof(bra)); 1832 SCARG(&bra, s) = SCARG(uap, s); 1833 SCARG(&bra, buf) = SCARG(uap, buf); 1834 SCARG(&bra, len) = (size_t) SCARG(uap, len); 1835 SCARG(&bra, flags) = SCARG(uap, flags); 1836 SCARG(&bra, from) = NULL; 1837 SCARG(&bra, fromlenaddr) = NULL; 1838 1839 return (sys_recvfrom(l, &bra, retval)); 1840 } 1841 1842 int 1843 linux_sys_send(struct lwp *l, const struct linux_sys_send_args *uap, register_t *retval) 1844 { 1845 /* { 1846 syscallarg(int) s; 1847 syscallarg(void *) buf; 1848 syscallarg(int) len; 1849 syscallarg(int) flags; 1850 } */ 1851 struct sys_sendto_args bsa; 1852 1853 memset(&bsa, 0, sizeof(bsa)); 1854 SCARG(&bsa, s) = SCARG(uap, s); 1855 SCARG(&bsa, buf) = SCARG(uap, buf); 1856 SCARG(&bsa, len) = SCARG(uap, len); 1857 SCARG(&bsa, flags) = SCARG(uap, flags); 1858 SCARG(&bsa, to) = NULL; 1859 SCARG(&bsa, tolen) = 0; 1860 1861 return (sys_sendto(l, &bsa, retval)); 1862 } 1863 #endif 1864 1865 int 1866 linux_sys_accept(struct lwp *l, const struct linux_sys_accept_args *uap, register_t *retval) 1867 { 1868 /* { 1869 syscallarg(int) s; 1870 syscallarg(struct osockaddr *) name; 1871 syscallarg(int *) anamelen; 1872 } */ 1873 int error; 1874 struct sys_accept_args baa; 1875 1876 memset(&baa, 0, sizeof(baa)); 1877 SCARG(&baa, s) = SCARG(uap, s); 1878 SCARG(&baa, name) = (struct sockaddr *) SCARG(uap, name); 1879 SCARG(&baa, anamelen) = (unsigned int *) SCARG(uap, anamelen); 1880 1881 if ((error = sys_accept(l, &baa, retval))) 1882 return (error); 1883 1884 if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name)))) 1885 return (error); 1886 1887 return (0); 1888 } 1889 1890 int 1891 linux_sys_accept4(struct lwp *l, const struct linux_sys_accept4_args *uap, register_t *retval) 1892 { 1893 /* { 1894 syscallarg(int) s; 1895 syscallarg(struct osockaddr *) name; 1896 syscallarg(int *) anamelen; 1897 syscallarg(int) flags; 1898 } */ 1899 int error, flags; 1900 struct sockaddr_big name; 1901 1902 if ((flags = linux_to_bsd_type(SCARG(uap, flags))) == -1) 1903 return EINVAL; 1904 1905 name.sb_len = UCHAR_MAX; 1906 error = do_sys_accept(l, SCARG(uap, s), (struct sockaddr *)&name, 1907 retval, NULL, flags, 0); 1908 if (error != 0) 1909 return error; 1910 1911 error = copyout_sockname_sb((struct sockaddr *)SCARG(uap, name), 1912 SCARG(uap, anamelen), MSG_LENUSRSPACE, &name); 1913 if (error != 0) { 1914 int fd = (int)*retval; 1915 if (fd_getfile(fd) != NULL) 1916 (void)fd_close(fd); 1917 return error; 1918 } 1919 if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name)))) 1920 return error; 1921 1922 return 0; 1923 } 1924 1925 int 1926 linux_sys_sendmmsg(struct lwp *l, const struct linux_sys_sendmmsg_args *uap, 1927 register_t *retval) 1928 { 1929 /* { 1930 syscallarg(int) s; 1931 syscallarg(struct linux_mmsghdr *) msgvec; 1932 syscallarg(unsigned int) vlen; 1933 syscallarg(unsigned int) flags; 1934 } */ 1935 struct linux_mmsghdr lmsg; 1936 struct mmsghdr bmsg; 1937 struct socket *so; 1938 file_t *fp; 1939 struct msghdr *msg = &bmsg.msg_hdr; 1940 int error, s; 1941 unsigned int vlen, flags, dg; 1942 1943 if ((flags = linux_to_bsd_msg_flags(SCARG(uap, flags))) == -1) 1944 return EINVAL; 1945 1946 flags = (flags & MSG_USERFLAGS) | MSG_IOVUSRSPACE; 1947 1948 s = SCARG(uap, s); 1949 if ((error = fd_getsock1(s, &so, &fp)) != 0) 1950 return error; 1951 1952 vlen = SCARG(uap, vlen); 1953 if (vlen > 1024) 1954 vlen = 1024; 1955 1956 for (dg = 0; dg < vlen;) { 1957 error = copyin(SCARG(uap, msgvec) + dg, &lmsg, sizeof(lmsg)); 1958 if (error) 1959 break; 1960 error = linux_to_bsd_msghdr(&lmsg.msg_hdr, &bmsg.msg_hdr); 1961 if (error) 1962 break; 1963 1964 msg->msg_flags = flags; 1965 1966 error = do_sys_sendmsg_so(l, s, so, fp, msg, flags, retval); 1967 if (error) 1968 break; 1969 1970 ktrkuser("msghdr", msg, sizeof *msg); 1971 lmsg.msg_len = *retval; 1972 error = copyout(&lmsg, SCARG(uap, msgvec) + dg, sizeof(lmsg)); 1973 if (error) 1974 break; 1975 dg++; 1976 1977 } 1978 1979 *retval = dg; 1980 1981 fd_putfile(s); 1982 1983 /* 1984 * If we succeeded at least once, return 0. 1985 */ 1986 if (dg) 1987 return 0; 1988 return error; 1989 } 1990 1991 int 1992 linux_sys_recvmmsg(struct lwp *l, const struct linux_sys_recvmmsg_args *uap, 1993 register_t *retval) 1994 { 1995 /* { 1996 syscallarg(int) s; 1997 syscallarg(struct linux_mmsghdr *) msgvec; 1998 syscallarg(unsigned int) vlen; 1999 syscallarg(unsigned int) flags; 2000 syscallarg(struct linux_timespec *) timeout; 2001 } */ 2002 struct linux_mmsghdr lmsg; 2003 struct mmsghdr bmsg; 2004 struct socket *so; 2005 struct msghdr *msg = &bmsg.msg_hdr; 2006 int error, s; 2007 struct mbuf *from, *control; 2008 struct timespec ts = {0}, now; 2009 struct linux_timespec lts; 2010 unsigned int vlen, flags, dg; 2011 2012 if (SCARG(uap, timeout)) { 2013 error = copyin(SCARG(uap, timeout), <s, sizeof(lts)); 2014 if (error) 2015 return error; 2016 ts.tv_sec = lts.tv_sec; 2017 ts.tv_nsec = lts.tv_nsec; 2018 getnanotime(&now); 2019 timespecadd(&now, &ts, &ts); 2020 } 2021 2022 s = SCARG(uap, s); 2023 if ((error = fd_getsock(s, &so)) != 0) 2024 return error; 2025 2026 /* 2027 * If so->so_rerror holds a deferred error return it now. 2028 */ 2029 if (so->so_rerror) { 2030 error = so->so_rerror; 2031 so->so_rerror = 0; 2032 fd_putfile(s); 2033 return error; 2034 } 2035 2036 vlen = SCARG(uap, vlen); 2037 if (vlen > 1024) 2038 vlen = 1024; 2039 2040 from = NULL; 2041 flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE; 2042 2043 for (dg = 0; dg < vlen;) { 2044 error = copyin(SCARG(uap, msgvec) + dg, &lmsg, sizeof(lmsg)); 2045 if (error) 2046 break; 2047 error = linux_to_bsd_msghdr(&lmsg.msg_hdr, &bmsg.msg_hdr); 2048 if (error) 2049 break; 2050 msg->msg_flags = flags & ~MSG_WAITFORONE; 2051 2052 if (from != NULL) { 2053 m_free(from); 2054 from = NULL; 2055 } 2056 2057 error = do_sys_recvmsg_so(l, s, so, msg, &from, 2058 msg->msg_control != NULL ? &control : NULL, retval); 2059 if (error) { 2060 if (error == EAGAIN && dg > 0) 2061 error = 0; 2062 break; 2063 } 2064 2065 if (msg->msg_control != NULL) 2066 error = linux_copyout_msg_control(l, msg, control); 2067 if (error) 2068 break; 2069 2070 if (from != NULL) { 2071 mtod(from, struct osockaddr *)->sa_family = 2072 bsd_to_linux_domain(mtod(from, 2073 struct sockaddr *)->sa_family); 2074 error = copyout_sockname(msg->msg_name, 2075 &msg->msg_namelen, 0, from); 2076 if (error) 2077 break; 2078 } 2079 2080 2081 lmsg.msg_len = *retval; 2082 ktrkuser("msghdr", msg, sizeof(*msg)); 2083 bsd_to_linux_msghdr(msg, &lmsg.msg_hdr); 2084 error = copyout(&lmsg, SCARG(uap, msgvec) + dg, sizeof(lmsg)); 2085 if (error) 2086 break; 2087 2088 dg++; 2089 if (msg->msg_flags & MSG_OOB) 2090 break; 2091 2092 if (SCARG(uap, timeout)) { 2093 getnanotime(&now); 2094 timespecsub(&now, &ts, &now); 2095 if (now.tv_sec > 0) 2096 break; 2097 } 2098 2099 if (flags & MSG_WAITFORONE) 2100 flags |= MSG_DONTWAIT; 2101 2102 } 2103 2104 if (from != NULL) 2105 m_free(from); 2106 2107 *retval = dg; 2108 2109 /* 2110 * If we succeeded at least once, return 0, hopefully so->so_rerror 2111 * will catch it next time. 2112 */ 2113 if (error && dg > 0) { 2114 so->so_rerror = error; 2115 error = 0; 2116 } 2117 2118 fd_putfile(s); 2119 2120 return error; 2121 } 2122
Indexes created Fri Sep 25 00:27:33 UTC 2026