1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 1995 Søren Schmidt 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 /* XXX we use functions that might not exist. */ 33 #include "opt_compat.h" 34 #include "opt_inet6.h" 35 36 #include <sys/param.h> 37 #include <sys/proc.h> 38 #include <sys/systm.h> 39 #include <sys/sysproto.h> 40 #include <sys/capsicum.h> 41 #include <sys/fcntl.h> 42 #include <sys/file.h> 43 #include <sys/filedesc.h> 44 #include <sys/limits.h> 45 #include <sys/lock.h> 46 #include <sys/malloc.h> 47 #include <sys/mutex.h> 48 #include <sys/mbuf.h> 49 #include <sys/socket.h> 50 #include <sys/socketvar.h> 51 #include <sys/syscallsubr.h> 52 #include <sys/uio.h> 53 #include <sys/stat.h> 54 #include <sys/syslog.h> 55 #include <sys/un.h> 56 #include <sys/unistd.h> 57 58 #include <security/audit/audit.h> 59 60 #include <net/if.h> 61 #include <net/vnet.h> 62 #include <netinet/in.h> 63 #include <netinet/in_systm.h> 64 #include <netinet/ip.h> 65 #include <netinet/tcp.h> 66 #ifdef INET6 67 #include <netinet/ip6.h> 68 #include <netinet6/ip6_var.h> 69 #endif 70 71 #ifdef COMPAT_LINUX32 72 #include <machine/../linux32/linux.h> 73 #include <machine/../linux32/linux32_proto.h> 74 #else 75 #include <machine/../linux/linux.h> 76 #include <machine/../linux/linux_proto.h> 77 #endif 78 #include <compat/linux/linux_common.h> 79 #include <compat/linux/linux_file.h> 80 #include <compat/linux/linux_mib.h> 81 #include <compat/linux/linux_socket.h> 82 #include <compat/linux/linux_timer.h> 83 #include <compat/linux/linux_util.h> 84 85 #define SECURITY_CONTEXT_STRING "unconfined" 86 87 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *, 88 l_uint); 89 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *, 90 l_uint, struct msghdr *); 91 static int linux_set_socket_flags(int, int *); 92 93 static int 94 linux_to_bsd_sockopt_level(int level) 95 { 96 97 if (level == LINUX_SOL_SOCKET) 98 return (SOL_SOCKET); 99 /* Remaining values are RFC-defined protocol numbers. */ 100 return (level); 101 } 102 103 static int 104 bsd_to_linux_sockopt_level(int level) 105 { 106 107 if (level == SOL_SOCKET) 108 return (LINUX_SOL_SOCKET); 109 return (level); 110 } 111 112 static int 113 linux_to_bsd_ip_sockopt(int opt) 114 { 115 116 switch (opt) { 117 /* known and translated sockopts */ 118 case LINUX_IP_TOS: 119 return (IP_TOS); 120 case LINUX_IP_TTL: 121 return (IP_TTL); 122 case LINUX_IP_HDRINCL: 123 return (IP_HDRINCL); 124 case LINUX_IP_OPTIONS: 125 return (IP_OPTIONS); 126 case LINUX_IP_RECVOPTS: 127 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVOPTS"); 128 return (IP_RECVOPTS); 129 case LINUX_IP_RETOPTS: 130 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_REETOPTS"); 131 return (IP_RETOPTS); 132 case LINUX_IP_RECVTTL: 133 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTTL"); 134 return (IP_RECVTTL); 135 case LINUX_IP_RECVTOS: 136 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTOS"); 137 return (IP_RECVTOS); 138 case LINUX_IP_FREEBIND: 139 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_FREEBIND"); 140 return (IP_BINDANY); 141 case LINUX_IP_IPSEC_POLICY: 142 /* we have this option, but not documented in ip(4) manpage */ 143 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_IPSEC_POLICY"); 144 return (IP_IPSEC_POLICY); 145 case LINUX_IP_MINTTL: 146 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MINTTL"); 147 return (IP_MINTTL); 148 case LINUX_IP_MULTICAST_IF: 149 return (IP_MULTICAST_IF); 150 case LINUX_IP_MULTICAST_TTL: 151 return (IP_MULTICAST_TTL); 152 case LINUX_IP_MULTICAST_LOOP: 153 return (IP_MULTICAST_LOOP); 154 case LINUX_IP_ADD_MEMBERSHIP: 155 return (IP_ADD_MEMBERSHIP); 156 case LINUX_IP_DROP_MEMBERSHIP: 157 return (IP_DROP_MEMBERSHIP); 158 case LINUX_IP_UNBLOCK_SOURCE: 159 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_UNBLOCK_SOURCE"); 160 return (IP_UNBLOCK_SOURCE); 161 case LINUX_IP_BLOCK_SOURCE: 162 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_BLOCK_SOURCE"); 163 return (IP_BLOCK_SOURCE); 164 case LINUX_IP_ADD_SOURCE_MEMBERSHIP: 165 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_ADD_SOURCE_MEMBERSHIP"); 166 return (IP_ADD_SOURCE_MEMBERSHIP); 167 case LINUX_IP_DROP_SOURCE_MEMBERSHIP: 168 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_DROP_SOURCE_MEMBERSHIP"); 169 return (IP_DROP_SOURCE_MEMBERSHIP); 170 case LINUX_MCAST_JOIN_GROUP: 171 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_JOIN_GROUP"); 172 return (MCAST_JOIN_GROUP); 173 case LINUX_MCAST_LEAVE_GROUP: 174 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_LEAVE_GROUP"); 175 return (MCAST_LEAVE_GROUP); 176 case LINUX_MCAST_JOIN_SOURCE_GROUP: 177 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_JOIN_SOURCE_GROUP"); 178 return (MCAST_JOIN_SOURCE_GROUP); 179 case LINUX_MCAST_LEAVE_SOURCE_GROUP: 180 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_LEAVE_SOURCE_GROUP"); 181 return (MCAST_LEAVE_SOURCE_GROUP); 182 183 /* known but not implemented sockopts */ 184 case LINUX_IP_ROUTER_ALERT: 185 LINUX_RATELIMIT_MSG_OPT1( 186 "unsupported IPv4 socket option IP_ROUTER_ALERT (%d), you can not do user-space routing from linux programs", 187 opt); 188 return (-2); 189 case LINUX_IP_PKTINFO: 190 LINUX_RATELIMIT_MSG_OPT1( 191 "unsupported IPv4 socket option IP_PKTINFO (%d), you can not get extended packet info for datagram sockets in linux programs", 192 opt); 193 return (-2); 194 case LINUX_IP_PKTOPTIONS: 195 LINUX_RATELIMIT_MSG_OPT1( 196 "unsupported IPv4 socket option IP_PKTOPTIONS (%d)", 197 opt); 198 return (-2); 199 case LINUX_IP_MTU_DISCOVER: 200 LINUX_RATELIMIT_MSG_OPT1( 201 "unsupported IPv4 socket option IP_MTU_DISCOVER (%d), your linux program can not control path-MTU discovery", 202 opt); 203 return (-2); 204 case LINUX_IP_RECVERR: 205 /* needed by steam */ 206 LINUX_RATELIMIT_MSG_OPT1( 207 "unsupported IPv4 socket option IP_RECVERR (%d), you can not get extended reliability info in linux programs", 208 opt); 209 return (-2); 210 case LINUX_IP_MTU: 211 LINUX_RATELIMIT_MSG_OPT1( 212 "unsupported IPv4 socket option IP_MTU (%d), your linux program can not control the MTU on this socket", 213 opt); 214 return (-2); 215 case LINUX_IP_XFRM_POLICY: 216 LINUX_RATELIMIT_MSG_OPT1( 217 "unsupported IPv4 socket option IP_XFRM_POLICY (%d)", 218 opt); 219 return (-2); 220 case LINUX_IP_PASSSEC: 221 /* needed by steam */ 222 LINUX_RATELIMIT_MSG_OPT1( 223 "unsupported IPv4 socket option IP_PASSSEC (%d), you can not get IPSEC related credential information associated with this socket in linux programs -- if you do not use IPSEC, you can ignore this", 224 opt); 225 return (-2); 226 case LINUX_IP_TRANSPARENT: 227 /* IP_BINDANY or more? */ 228 LINUX_RATELIMIT_MSG_OPT1( 229 "unsupported IPv4 socket option IP_TRANSPARENT (%d), you can not enable transparent proxying in linux programs -- note, IP_FREEBIND is supported, no idea if the FreeBSD IP_BINDANY is equivalent to the Linux IP_TRANSPARENT or not, any info is welcome", 230 opt); 231 return (-2); 232 case LINUX_IP_NODEFRAG: 233 LINUX_RATELIMIT_MSG_OPT1( 234 "unsupported IPv4 socket option IP_NODEFRAG (%d)", 235 opt); 236 return (-2); 237 case LINUX_IP_CHECKSUM: 238 LINUX_RATELIMIT_MSG_OPT1( 239 "unsupported IPv4 socket option IP_CHECKSUM (%d)", 240 opt); 241 return (-2); 242 case LINUX_IP_BIND_ADDRESS_NO_PORT: 243 LINUX_RATELIMIT_MSG_OPT1( 244 "unsupported IPv4 socket option IP_BIND_ADDRESS_NO_PORT (%d)", 245 opt); 246 return (-2); 247 case LINUX_IP_RECVFRAGSIZE: 248 LINUX_RATELIMIT_MSG_OPT1( 249 "unsupported IPv4 socket option IP_RECVFRAGSIZE (%d)", 250 opt); 251 return (-2); 252 case LINUX_MCAST_MSFILTER: 253 LINUX_RATELIMIT_MSG_OPT1( 254 "unsupported IPv4 socket option IP_MCAST_MSFILTER (%d)", 255 opt); 256 return (-2); 257 case LINUX_IP_MULTICAST_ALL: 258 LINUX_RATELIMIT_MSG_OPT1( 259 "unsupported IPv4 socket option IP_MULTICAST_ALL (%d), your linux program will not see all multicast groups joined by the entire system, only those the program joined itself on this socket", 260 opt); 261 return (-2); 262 case LINUX_IP_UNICAST_IF: 263 LINUX_RATELIMIT_MSG_OPT1( 264 "unsupported IPv4 socket option IP_UNICAST_IF (%d)", 265 opt); 266 return (-2); 267 268 /* unknown sockopts */ 269 default: 270 return (-1); 271 } 272 } 273 274 static int 275 linux_to_bsd_ip6_sockopt(int opt) 276 { 277 278 switch (opt) { 279 /* known and translated sockopts */ 280 case LINUX_IPV6_2292PKTINFO: 281 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292PKTINFO"); 282 return (IPV6_2292PKTINFO); 283 case LINUX_IPV6_2292HOPOPTS: 284 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292HOPOPTS"); 285 return (IPV6_2292HOPOPTS); 286 case LINUX_IPV6_2292DSTOPTS: 287 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292DSTOPTS"); 288 return (IPV6_2292DSTOPTS); 289 case LINUX_IPV6_2292RTHDR: 290 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292RTHDR"); 291 return (IPV6_2292RTHDR); 292 case LINUX_IPV6_2292PKTOPTIONS: 293 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292PKTOPTIONS"); 294 return (IPV6_2292PKTOPTIONS); 295 case LINUX_IPV6_CHECKSUM: 296 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_CHECKSUM"); 297 return (IPV6_CHECKSUM); 298 case LINUX_IPV6_2292HOPLIMIT: 299 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292HOPLIMIT"); 300 return (IPV6_2292HOPLIMIT); 301 case LINUX_IPV6_NEXTHOP: 302 return (IPV6_NEXTHOP); 303 case LINUX_IPV6_UNICAST_HOPS: 304 return (IPV6_UNICAST_HOPS); 305 case LINUX_IPV6_MULTICAST_IF: 306 return (IPV6_MULTICAST_IF); 307 case LINUX_IPV6_MULTICAST_HOPS: 308 return (IPV6_MULTICAST_HOPS); 309 case LINUX_IPV6_MULTICAST_LOOP: 310 return (IPV6_MULTICAST_LOOP); 311 case LINUX_IPV6_ADD_MEMBERSHIP: 312 return (IPV6_JOIN_GROUP); 313 case LINUX_IPV6_DROP_MEMBERSHIP: 314 return (IPV6_LEAVE_GROUP); 315 case LINUX_IPV6_V6ONLY: 316 return (IPV6_V6ONLY); 317 case LINUX_IPV6_IPSEC_POLICY: 318 /* we have this option, but not documented in ip6(4) manpage */ 319 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_IPSEC_POLICY"); 320 return (IPV6_IPSEC_POLICY); 321 case LINUX_MCAST_JOIN_GROUP: 322 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_JOIN_GROUP"); 323 return (IPV6_JOIN_GROUP); 324 case LINUX_MCAST_LEAVE_GROUP: 325 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_LEAVE_GROUP"); 326 return (IPV6_LEAVE_GROUP); 327 case LINUX_IPV6_RECVPKTINFO: 328 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVPKTINFO"); 329 return (IPV6_RECVPKTINFO); 330 case LINUX_IPV6_PKTINFO: 331 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_PKTINFO"); 332 return (IPV6_PKTINFO); 333 case LINUX_IPV6_RECVHOPLIMIT: 334 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVHOPLIMIT"); 335 return (IPV6_RECVHOPLIMIT); 336 case LINUX_IPV6_HOPLIMIT: 337 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_HOPLIMIT"); 338 return (IPV6_HOPLIMIT); 339 case LINUX_IPV6_RECVHOPOPTS: 340 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVHOPOPTS"); 341 return (IPV6_RECVHOPOPTS); 342 case LINUX_IPV6_HOPOPTS: 343 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_HOPOPTS"); 344 return (IPV6_HOPOPTS); 345 case LINUX_IPV6_RTHDRDSTOPTS: 346 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RTHDRDSTOPTS"); 347 return (IPV6_RTHDRDSTOPTS); 348 case LINUX_IPV6_RECVRTHDR: 349 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVRTHDR"); 350 return (IPV6_RECVRTHDR); 351 case LINUX_IPV6_RTHDR: 352 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RTHDR"); 353 return (IPV6_RTHDR); 354 case LINUX_IPV6_RECVDSTOPTS: 355 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVDSTOPTS"); 356 return (IPV6_RECVDSTOPTS); 357 case LINUX_IPV6_DSTOPTS: 358 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_DSTOPTS"); 359 return (IPV6_DSTOPTS); 360 case LINUX_IPV6_RECVPATHMTU: 361 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVPATHMTU"); 362 return (IPV6_RECVPATHMTU); 363 case LINUX_IPV6_PATHMTU: 364 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_PATHMTU"); 365 return (IPV6_PATHMTU); 366 case LINUX_IPV6_DONTFRAG: 367 return (IPV6_DONTFRAG); 368 case LINUX_IPV6_AUTOFLOWLABEL: 369 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_AUTOFLOWLABEL"); 370 return (IPV6_AUTOFLOWLABEL); 371 case LINUX_IPV6_ORIGDSTADDR: 372 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_ORIGDSTADDR"); 373 return (IPV6_ORIGDSTADDR); 374 case LINUX_IPV6_FREEBIND: 375 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_FREEBIND"); 376 return (IPV6_BINDANY); 377 378 /* known but not implemented sockopts */ 379 case LINUX_IPV6_ADDRFORM: 380 LINUX_RATELIMIT_MSG_OPT1( 381 "unsupported IPv6 socket option IPV6_ADDRFORM (%d), you linux program can not convert the socket to IPv4", 382 opt); 383 return (-2); 384 case LINUX_IPV6_AUTHHDR: 385 LINUX_RATELIMIT_MSG_OPT1( 386 "unsupported IPv6 socket option IPV6_AUTHHDR (%d), your linux program can not get the authentication header info of IPv6 packets", 387 opt); 388 return (-2); 389 case LINUX_IPV6_FLOWINFO: 390 LINUX_RATELIMIT_MSG_OPT1( 391 "unsupported IPv6 socket option IPV6_FLOWINFO (%d), your linux program can not get the flowid of IPv6 packets", 392 opt); 393 return (-2); 394 case LINUX_IPV6_ROUTER_ALERT: 395 LINUX_RATELIMIT_MSG_OPT1( 396 "unsupported IPv6 socket option IPV6_ROUTER_ALERT (%d), you can not do user-space routing from linux programs", 397 opt); 398 return (-2); 399 case LINUX_IPV6_MTU_DISCOVER: 400 LINUX_RATELIMIT_MSG_OPT1( 401 "unsupported IPv6 socket option IPV6_MTU_DISCOVER (%d), your linux program can not control path-MTU discovery", 402 opt); 403 return (-2); 404 case LINUX_IPV6_MTU: 405 LINUX_RATELIMIT_MSG_OPT1( 406 "unsupported IPv6 socket option IPV6_MTU (%d), your linux program can not control the MTU on this socket", 407 opt); 408 return (-2); 409 case LINUX_IPV6_JOIN_ANYCAST: 410 LINUX_RATELIMIT_MSG_OPT1( 411 "unsupported IPv6 socket option IPV6_JOIN_ANYCAST (%d)", 412 opt); 413 return (-2); 414 case LINUX_IPV6_LEAVE_ANYCAST: 415 LINUX_RATELIMIT_MSG_OPT1( 416 "unsupported IPv6 socket option IPV6_LEAVE_ANYCAST (%d)", 417 opt); 418 return (-2); 419 case LINUX_IPV6_MULTICAST_ALL: 420 LINUX_RATELIMIT_MSG_OPT1( 421 "unsupported IPv6 socket option IPV6_MULTICAST_ALL (%d)", 422 opt); 423 return (-2); 424 case LINUX_IPV6_ROUTER_ALERT_ISOLATE: 425 LINUX_RATELIMIT_MSG_OPT1( 426 "unsupported IPv6 socket option IPV6_ROUTER_ALERT_ISOLATE (%d)", 427 opt); 428 return (-2); 429 case LINUX_IPV6_FLOWLABEL_MGR: 430 LINUX_RATELIMIT_MSG_OPT1( 431 "unsupported IPv6 socket option IPV6_FLOWLABEL_MGR (%d)", 432 opt); 433 return (-2); 434 case LINUX_IPV6_FLOWINFO_SEND: 435 LINUX_RATELIMIT_MSG_OPT1( 436 "unsupported IPv6 socket option IPV6_FLOWINFO_SEND (%d)", 437 opt); 438 return (-2); 439 case LINUX_IPV6_XFRM_POLICY: 440 LINUX_RATELIMIT_MSG_OPT1( 441 "unsupported IPv6 socket option IPV6_XFRM_POLICY (%d)", 442 opt); 443 return (-2); 444 case LINUX_IPV6_HDRINCL: 445 LINUX_RATELIMIT_MSG_OPT1( 446 "unsupported IPv6 socket option IPV6_HDRINCL (%d)", 447 opt); 448 return (-2); 449 case LINUX_MCAST_BLOCK_SOURCE: 450 LINUX_RATELIMIT_MSG_OPT1( 451 "unsupported IPv6 socket option MCAST_BLOCK_SOURCE (%d), your linux program may see more multicast stuff than it wants", 452 opt); 453 return (-2); 454 case LINUX_MCAST_UNBLOCK_SOURCE: 455 LINUX_RATELIMIT_MSG_OPT1( 456 "unsupported IPv6 socket option MCAST_UNBLOCK_SOURCE (%d), your linux program may not see all the multicast stuff it wants", 457 opt); 458 return (-2); 459 case LINUX_MCAST_JOIN_SOURCE_GROUP: 460 LINUX_RATELIMIT_MSG_OPT1( 461 "unsupported IPv6 socket option MCAST_JOIN_SOURCE_GROUP (%d), your linux program is not able to join a multicast source group", 462 opt); 463 return (-2); 464 case LINUX_MCAST_LEAVE_SOURCE_GROUP: 465 LINUX_RATELIMIT_MSG_OPT1( 466 "unsupported IPv6 socket option MCAST_LEAVE_SOURCE_GROUP (%d), your linux program is not able to leave a multicast source group -- but it was also not able to join one, so no issue", 467 opt); 468 return (-2); 469 case LINUX_MCAST_MSFILTER: 470 LINUX_RATELIMIT_MSG_OPT1( 471 "unsupported IPv6 socket option MCAST_MSFILTER (%d), your linux program can not manipulate the multicast filter, it may see more multicast data than it wants to see", 472 opt); 473 return (-2); 474 case LINUX_IPV6_ADDR_PREFERENCES: 475 LINUX_RATELIMIT_MSG_OPT1( 476 "unsupported IPv6 socket option IPV6_ADDR_PREFERENCES (%d)", 477 opt); 478 return (-2); 479 case LINUX_IPV6_MINHOPCOUNT: 480 LINUX_RATELIMIT_MSG_OPT1( 481 "unsupported IPv6 socket option IPV6_MINHOPCOUNT (%d)", 482 opt); 483 return (-2); 484 case LINUX_IPV6_TRANSPARENT: 485 /* IP_BINDANY or more? */ 486 LINUX_RATELIMIT_MSG_OPT1( 487 "unsupported IPv6 socket option IPV6_TRANSPARENT (%d), you can not enable transparent proxying in linux programs -- note, IP_FREEBIND is supported, no idea if the FreeBSD IP_BINDANY is equivalent to the Linux IP_TRANSPARENT or not, any info is welcome", 488 opt); 489 return (-2); 490 case LINUX_IPV6_UNICAST_IF: 491 LINUX_RATELIMIT_MSG_OPT1( 492 "unsupported IPv6 socket option IPV6_UNICAST_IF (%d)", 493 opt); 494 return (-2); 495 case LINUX_IPV6_RECVFRAGSIZE: 496 LINUX_RATELIMIT_MSG_OPT1( 497 "unsupported IPv6 socket option IPV6_RECVFRAGSIZE (%d)", 498 opt); 499 return (-2); 500 501 /* unknown sockopts */ 502 default: 503 return (-1); 504 } 505 } 506 507 static int 508 linux_to_bsd_so_sockopt(int opt) 509 { 510 511 switch (opt) { 512 case LINUX_SO_DEBUG: 513 return (SO_DEBUG); 514 case LINUX_SO_REUSEADDR: 515 return (SO_REUSEADDR); 516 case LINUX_SO_TYPE: 517 return (SO_TYPE); 518 case LINUX_SO_ERROR: 519 return (SO_ERROR); 520 case LINUX_SO_DONTROUTE: 521 return (SO_DONTROUTE); 522 case LINUX_SO_BROADCAST: 523 return (SO_BROADCAST); 524 case LINUX_SO_SNDBUF: 525 case LINUX_SO_SNDBUFFORCE: 526 return (SO_SNDBUF); 527 case LINUX_SO_RCVBUF: 528 case LINUX_SO_RCVBUFFORCE: 529 return (SO_RCVBUF); 530 case LINUX_SO_KEEPALIVE: 531 return (SO_KEEPALIVE); 532 case LINUX_SO_OOBINLINE: 533 return (SO_OOBINLINE); 534 case LINUX_SO_LINGER: 535 return (SO_LINGER); 536 case LINUX_SO_REUSEPORT: 537 return (SO_REUSEPORT_LB); 538 case LINUX_SO_PASSCRED: 539 return (LOCAL_CREDS_PERSISTENT); 540 case LINUX_SO_PEERCRED: 541 return (LOCAL_PEERCRED); 542 case LINUX_SO_RCVLOWAT: 543 return (SO_RCVLOWAT); 544 case LINUX_SO_SNDLOWAT: 545 return (SO_SNDLOWAT); 546 case LINUX_SO_RCVTIMEO: 547 return (SO_RCVTIMEO); 548 case LINUX_SO_SNDTIMEO: 549 return (SO_SNDTIMEO); 550 case LINUX_SO_TIMESTAMP: 551 return (SO_TIMESTAMP); 552 case LINUX_SO_ACCEPTCONN: 553 return (SO_ACCEPTCONN); 554 case LINUX_SO_PROTOCOL: 555 return (SO_PROTOCOL); 556 } 557 return (-1); 558 } 559 560 static int 561 linux_to_bsd_tcp_sockopt(int opt) 562 { 563 564 switch (opt) { 565 case LINUX_TCP_NODELAY: 566 return (TCP_NODELAY); 567 case LINUX_TCP_MAXSEG: 568 return (TCP_MAXSEG); 569 case LINUX_TCP_CORK: 570 return (TCP_NOPUSH); 571 case LINUX_TCP_KEEPIDLE: 572 return (TCP_KEEPIDLE); 573 case LINUX_TCP_KEEPINTVL: 574 return (TCP_KEEPINTVL); 575 case LINUX_TCP_KEEPCNT: 576 return (TCP_KEEPCNT); 577 case LINUX_TCP_INFO: 578 LINUX_RATELIMIT_MSG_OPT1( 579 "unsupported TCP socket option TCP_INFO (%d)", opt); 580 return (-2); 581 case LINUX_TCP_MD5SIG: 582 return (TCP_MD5SIG); 583 } 584 return (-1); 585 } 586 587 static int 588 linux_to_bsd_msg_flags(int flags) 589 { 590 int ret_flags = 0; 591 592 if (flags & LINUX_MSG_OOB) 593 ret_flags |= MSG_OOB; 594 if (flags & LINUX_MSG_PEEK) 595 ret_flags |= MSG_PEEK; 596 if (flags & LINUX_MSG_DONTROUTE) 597 ret_flags |= MSG_DONTROUTE; 598 if (flags & LINUX_MSG_CTRUNC) 599 ret_flags |= MSG_CTRUNC; 600 if (flags & LINUX_MSG_TRUNC) 601 ret_flags |= MSG_TRUNC; 602 if (flags & LINUX_MSG_DONTWAIT) 603 ret_flags |= MSG_DONTWAIT; 604 if (flags & LINUX_MSG_EOR) 605 ret_flags |= MSG_EOR; 606 if (flags & LINUX_MSG_WAITALL) 607 ret_flags |= MSG_WAITALL; 608 if (flags & LINUX_MSG_NOSIGNAL) 609 ret_flags |= MSG_NOSIGNAL; 610 if (flags & LINUX_MSG_PROXY) 611 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_PROXY (%d) not handled", 612 LINUX_MSG_PROXY); 613 if (flags & LINUX_MSG_FIN) 614 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_FIN (%d) not handled", 615 LINUX_MSG_FIN); 616 if (flags & LINUX_MSG_SYN) 617 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_SYN (%d) not handled", 618 LINUX_MSG_SYN); 619 if (flags & LINUX_MSG_CONFIRM) 620 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_CONFIRM (%d) not handled", 621 LINUX_MSG_CONFIRM); 622 if (flags & LINUX_MSG_RST) 623 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_RST (%d) not handled", 624 LINUX_MSG_RST); 625 if (flags & LINUX_MSG_ERRQUEUE) 626 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_ERRQUEUE (%d) not handled", 627 LINUX_MSG_ERRQUEUE); 628 return (ret_flags); 629 } 630 631 static int 632 linux_to_bsd_cmsg_type(int cmsg_type) 633 { 634 635 switch (cmsg_type) { 636 case LINUX_SCM_RIGHTS: 637 return (SCM_RIGHTS); 638 case LINUX_SCM_CREDENTIALS: 639 return (SCM_CREDS); 640 } 641 return (-1); 642 } 643 644 static int 645 bsd_to_linux_cmsg_type(int cmsg_type) 646 { 647 648 switch (cmsg_type) { 649 case SCM_RIGHTS: 650 return (LINUX_SCM_RIGHTS); 651 case SCM_CREDS: 652 return (LINUX_SCM_CREDENTIALS); 653 case SCM_CREDS2: 654 return (LINUX_SCM_CREDENTIALS); 655 case SCM_TIMESTAMP: 656 return (LINUX_SCM_TIMESTAMP); 657 } 658 return (-1); 659 } 660 661 static int 662 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 663 { 664 if (lhdr->msg_controllen > INT_MAX) 665 return (ENOBUFS); 666 667 bhdr->msg_name = PTRIN(lhdr->msg_name); 668 bhdr->msg_namelen = lhdr->msg_namelen; 669 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 670 bhdr->msg_iovlen = lhdr->msg_iovlen; 671 bhdr->msg_control = PTRIN(lhdr->msg_control); 672 673 /* 674 * msg_controllen is skipped since BSD and LINUX control messages 675 * are potentially different sizes (e.g. the cred structure used 676 * by SCM_CREDS is different between the two operating system). 677 * 678 * The caller can set it (if necessary) after converting all the 679 * control messages. 680 */ 681 682 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 683 return (0); 684 } 685 686 static int 687 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 688 { 689 lhdr->msg_name = PTROUT(bhdr->msg_name); 690 lhdr->msg_namelen = bhdr->msg_namelen; 691 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 692 lhdr->msg_iovlen = bhdr->msg_iovlen; 693 lhdr->msg_control = PTROUT(bhdr->msg_control); 694 695 /* 696 * msg_controllen is skipped since BSD and LINUX control messages 697 * are potentially different sizes (e.g. the cred structure used 698 * by SCM_CREDS is different between the two operating system). 699 * 700 * The caller can set it (if necessary) after converting all the 701 * control messages. 702 */ 703 704 /* msg_flags skipped */ 705 return (0); 706 } 707 708 static int 709 linux_set_socket_flags(int lflags, int *flags) 710 { 711 712 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 713 return (EINVAL); 714 if (lflags & LINUX_SOCK_NONBLOCK) 715 *flags |= SOCK_NONBLOCK; 716 if (lflags & LINUX_SOCK_CLOEXEC) 717 *flags |= SOCK_CLOEXEC; 718 return (0); 719 } 720 721 static int 722 linux_copyout_sockaddr(const struct sockaddr *sa, void *uaddr, size_t len) 723 { 724 struct l_sockaddr *lsa; 725 int error; 726 727 error = bsd_to_linux_sockaddr(sa, &lsa, len); 728 if (error != 0) 729 return (error); 730 731 error = copyout(lsa, uaddr, len); 732 free(lsa, M_SONAME); 733 734 return (error); 735 } 736 737 static int 738 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 739 struct mbuf *control, enum uio_seg segflg) 740 { 741 struct sockaddr *to; 742 int error, len; 743 744 if (mp->msg_name != NULL) { 745 len = mp->msg_namelen; 746 error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len); 747 if (error != 0) 748 return (error); 749 mp->msg_name = to; 750 } else 751 to = NULL; 752 753 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 754 segflg); 755 756 if (to) 757 free(to, M_SONAME); 758 return (error); 759 } 760 761 /* Return 0 if IP_HDRINCL is set for the given socket. */ 762 static int 763 linux_check_hdrincl(struct thread *td, int s) 764 { 765 int error, optval; 766 socklen_t size_val; 767 768 size_val = sizeof(optval); 769 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 770 &optval, UIO_SYSSPACE, &size_val); 771 if (error != 0) 772 return (error); 773 774 return (optval == 0); 775 } 776 777 /* 778 * Updated sendto() when IP_HDRINCL is set: 779 * tweak endian-dependent fields in the IP packet. 780 */ 781 static int 782 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 783 { 784 /* 785 * linux_ip_copysize defines how many bytes we should copy 786 * from the beginning of the IP packet before we customize it for BSD. 787 * It should include all the fields we modify (ip_len and ip_off). 788 */ 789 #define linux_ip_copysize 8 790 791 struct ip *packet; 792 struct msghdr msg; 793 struct iovec aiov[1]; 794 int error; 795 796 /* Check that the packet isn't too big or too small. */ 797 if (linux_args->len < linux_ip_copysize || 798 linux_args->len > IP_MAXPACKET) 799 return (EINVAL); 800 801 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK); 802 803 /* Make kernel copy of the packet to be sent */ 804 if ((error = copyin(PTRIN(linux_args->msg), packet, 805 linux_args->len))) 806 goto goout; 807 808 /* Convert fields from Linux to BSD raw IP socket format */ 809 packet->ip_len = linux_args->len; 810 packet->ip_off = ntohs(packet->ip_off); 811 812 /* Prepare the msghdr and iovec structures describing the new packet */ 813 msg.msg_name = PTRIN(linux_args->to); 814 msg.msg_namelen = linux_args->tolen; 815 msg.msg_iov = aiov; 816 msg.msg_iovlen = 1; 817 msg.msg_control = NULL; 818 msg.msg_flags = 0; 819 aiov[0].iov_base = (char *)packet; 820 aiov[0].iov_len = linux_args->len; 821 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 822 NULL, UIO_SYSSPACE); 823 goout: 824 free(packet, M_LINUX); 825 return (error); 826 } 827 828 static const char *linux_netlink_names[] = { 829 [LINUX_NETLINK_ROUTE] = "ROUTE", 830 [LINUX_NETLINK_SOCK_DIAG] = "SOCK_DIAG", 831 [LINUX_NETLINK_NFLOG] = "NFLOG", 832 [LINUX_NETLINK_SELINUX] = "SELINUX", 833 [LINUX_NETLINK_AUDIT] = "AUDIT", 834 [LINUX_NETLINK_FIB_LOOKUP] = "FIB_LOOKUP", 835 [LINUX_NETLINK_NETFILTER] = "NETFILTER", 836 [LINUX_NETLINK_KOBJECT_UEVENT] = "KOBJECT_UEVENT", 837 }; 838 839 int 840 linux_socket(struct thread *td, struct linux_socket_args *args) 841 { 842 int domain, retval_socket, type; 843 844 type = args->type & LINUX_SOCK_TYPE_MASK; 845 if (type < 0 || type > LINUX_SOCK_MAX) 846 return (EINVAL); 847 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 848 &type); 849 if (retval_socket != 0) 850 return (retval_socket); 851 domain = linux_to_bsd_domain(args->domain); 852 if (domain == -1) { 853 /* Mask off SOCK_NONBLOCK / CLOEXEC for error messages. */ 854 type = args->type & LINUX_SOCK_TYPE_MASK; 855 if (args->domain == LINUX_AF_NETLINK && 856 args->protocol == LINUX_NETLINK_AUDIT) { 857 ; /* Do nothing, quietly. */ 858 } else if (args->domain == LINUX_AF_NETLINK) { 859 const char *nl_name; 860 861 if (args->protocol >= 0 && 862 args->protocol < nitems(linux_netlink_names)) 863 nl_name = linux_netlink_names[args->protocol]; 864 else 865 nl_name = NULL; 866 if (nl_name != NULL) 867 linux_msg(curthread, 868 "unsupported socket(AF_NETLINK, %d, " 869 "NETLINK_%s)", type, nl_name); 870 else 871 linux_msg(curthread, 872 "unsupported socket(AF_NETLINK, %d, %d)", 873 type, args->protocol); 874 } else { 875 linux_msg(curthread, "unsupported socket domain %d, " 876 "type %d, protocol %d", args->domain, type, 877 args->protocol); 878 } 879 return (EAFNOSUPPORT); 880 } 881 882 retval_socket = kern_socket(td, domain, type, args->protocol); 883 if (retval_socket) 884 return (retval_socket); 885 886 if (type == SOCK_RAW 887 && (args->protocol == IPPROTO_RAW || args->protocol == 0) 888 && domain == PF_INET) { 889 /* It's a raw IP socket: set the IP_HDRINCL option. */ 890 int hdrincl; 891 892 hdrincl = 1; 893 /* We ignore any error returned by kern_setsockopt() */ 894 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 895 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 896 } 897 #ifdef INET6 898 /* 899 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default 900 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps. 901 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only 902 * sysctl value. 903 */ 904 if (domain == PF_INET6) { 905 int v6only; 906 907 v6only = 0; 908 /* We ignore any error returned by setsockopt() */ 909 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 910 &v6only, UIO_SYSSPACE, sizeof(v6only)); 911 } 912 #endif 913 914 return (retval_socket); 915 } 916 917 int 918 linux_bind(struct thread *td, struct linux_bind_args *args) 919 { 920 struct sockaddr *sa; 921 int error; 922 923 error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa, 924 &args->namelen); 925 if (error != 0) 926 return (error); 927 928 error = kern_bindat(td, AT_FDCWD, args->s, sa); 929 free(sa, M_SONAME); 930 931 /* XXX */ 932 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 933 return (EINVAL); 934 return (error); 935 } 936 937 int 938 linux_connect(struct thread *td, struct linux_connect_args *args) 939 { 940 struct socket *so; 941 struct sockaddr *sa; 942 struct file *fp; 943 u_int fflag; 944 int error; 945 946 error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa, 947 &args->namelen); 948 if (error != 0) 949 return (error); 950 951 error = kern_connectat(td, AT_FDCWD, args->s, sa); 952 free(sa, M_SONAME); 953 if (error != EISCONN) 954 return (error); 955 956 /* 957 * Linux doesn't return EISCONN the first time it occurs, 958 * when on a non-blocking socket. Instead it returns the 959 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 960 */ 961 error = getsock_cap(td, args->s, &cap_connect_rights, 962 &fp, &fflag, NULL); 963 if (error != 0) 964 return (error); 965 966 error = EISCONN; 967 so = fp->f_data; 968 if (fflag & FNONBLOCK) { 969 SOCK_LOCK(so); 970 if (so->so_emuldata == 0) 971 error = so->so_error; 972 so->so_emuldata = (void *)1; 973 SOCK_UNLOCK(so); 974 } 975 fdrop(fp, td); 976 977 return (error); 978 } 979 980 int 981 linux_listen(struct thread *td, struct linux_listen_args *args) 982 { 983 984 return (kern_listen(td, args->s, args->backlog)); 985 } 986 987 static int 988 linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 989 l_uintptr_t namelen, int flags) 990 { 991 struct sockaddr *sa; 992 struct file *fp, *fp1; 993 int bflags, len; 994 struct socket *so; 995 int error, error1; 996 997 bflags = 0; 998 fp = NULL; 999 sa = NULL; 1000 1001 error = linux_set_socket_flags(flags, &bflags); 1002 if (error != 0) 1003 return (error); 1004 1005 if (PTRIN(addr) == NULL) { 1006 len = 0; 1007 error = kern_accept4(td, s, NULL, NULL, bflags, NULL); 1008 } else { 1009 error = copyin(PTRIN(namelen), &len, sizeof(len)); 1010 if (error != 0) 1011 return (error); 1012 if (len < 0) 1013 return (EINVAL); 1014 error = kern_accept4(td, s, &sa, &len, bflags, &fp); 1015 } 1016 1017 /* 1018 * Translate errno values into ones used by Linux. 1019 */ 1020 if (error != 0) { 1021 /* 1022 * XXX. This is wrong, different sockaddr structures 1023 * have different sizes. 1024 */ 1025 switch (error) { 1026 case EFAULT: 1027 if (namelen != sizeof(struct sockaddr_in)) 1028 error = EINVAL; 1029 break; 1030 case EINVAL: 1031 error1 = getsock_cap(td, s, &cap_accept_rights, &fp1, NULL, NULL); 1032 if (error1 != 0) { 1033 error = error1; 1034 break; 1035 } 1036 so = fp1->f_data; 1037 if (so->so_type == SOCK_DGRAM) 1038 error = EOPNOTSUPP; 1039 fdrop(fp1, td); 1040 break; 1041 } 1042 return (error); 1043 } 1044 1045 if (len != 0) { 1046 error = linux_copyout_sockaddr(sa, PTRIN(addr), len); 1047 1048 /* 1049 * XXX: We should also copyout the len, shouldn't we? 1050 */ 1051 1052 if (error != 0) { 1053 fdclose(td, fp, td->td_retval[0]); 1054 td->td_retval[0] = 0; 1055 } 1056 } 1057 if (fp != NULL) 1058 fdrop(fp, td); 1059 free(sa, M_SONAME); 1060 return (error); 1061 } 1062 1063 int 1064 linux_accept(struct thread *td, struct linux_accept_args *args) 1065 { 1066 1067 return (linux_accept_common(td, args->s, args->addr, 1068 args->namelen, 0)); 1069 } 1070 1071 int 1072 linux_accept4(struct thread *td, struct linux_accept4_args *args) 1073 { 1074 1075 return (linux_accept_common(td, args->s, args->addr, 1076 args->namelen, args->flags)); 1077 } 1078 1079 int 1080 linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 1081 { 1082 struct sockaddr *sa; 1083 int len, error; 1084 1085 error = copyin(PTRIN(args->namelen), &len, sizeof(len)); 1086 if (error != 0) 1087 return (error); 1088 1089 error = kern_getsockname(td, args->s, &sa, &len); 1090 if (error != 0) 1091 return (error); 1092 1093 if (len != 0) 1094 error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len); 1095 1096 free(sa, M_SONAME); 1097 if (error == 0) 1098 error = copyout(&len, PTRIN(args->namelen), sizeof(len)); 1099 return (error); 1100 } 1101 1102 int 1103 linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 1104 { 1105 struct sockaddr *sa; 1106 int len, error; 1107 1108 error = copyin(PTRIN(args->namelen), &len, sizeof(len)); 1109 if (error != 0) 1110 return (error); 1111 if (len < 0) 1112 return (EINVAL); 1113 1114 error = kern_getpeername(td, args->s, &sa, &len); 1115 if (error != 0) 1116 return (error); 1117 1118 if (len != 0) 1119 error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len); 1120 1121 free(sa, M_SONAME); 1122 if (error == 0) 1123 error = copyout(&len, PTRIN(args->namelen), sizeof(len)); 1124 return (error); 1125 } 1126 1127 int 1128 linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 1129 { 1130 int domain, error, sv[2], type; 1131 1132 domain = linux_to_bsd_domain(args->domain); 1133 if (domain != PF_LOCAL) 1134 return (EAFNOSUPPORT); 1135 type = args->type & LINUX_SOCK_TYPE_MASK; 1136 if (type < 0 || type > LINUX_SOCK_MAX) 1137 return (EINVAL); 1138 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 1139 &type); 1140 if (error != 0) 1141 return (error); 1142 if (args->protocol != 0 && args->protocol != PF_UNIX) { 1143 /* 1144 * Use of PF_UNIX as protocol argument is not right, 1145 * but Linux does it. 1146 * Do not map PF_UNIX as its Linux value is identical 1147 * to FreeBSD one. 1148 */ 1149 return (EPROTONOSUPPORT); 1150 } 1151 error = kern_socketpair(td, domain, type, 0, sv); 1152 if (error != 0) 1153 return (error); 1154 error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int)); 1155 if (error != 0) { 1156 (void)kern_close(td, sv[0]); 1157 (void)kern_close(td, sv[1]); 1158 } 1159 return (error); 1160 } 1161 1162 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1163 struct linux_send_args { 1164 register_t s; 1165 register_t msg; 1166 register_t len; 1167 register_t flags; 1168 }; 1169 1170 static int 1171 linux_send(struct thread *td, struct linux_send_args *args) 1172 { 1173 struct sendto_args /* { 1174 int s; 1175 caddr_t buf; 1176 int len; 1177 int flags; 1178 caddr_t to; 1179 int tolen; 1180 } */ bsd_args; 1181 struct file *fp; 1182 int error, fflag; 1183 1184 bsd_args.s = args->s; 1185 bsd_args.buf = (caddr_t)PTRIN(args->msg); 1186 bsd_args.len = args->len; 1187 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 1188 bsd_args.to = NULL; 1189 bsd_args.tolen = 0; 1190 error = sys_sendto(td, &bsd_args); 1191 if (error == ENOTCONN) { 1192 /* 1193 * Linux doesn't return ENOTCONN for non-blocking sockets. 1194 * Instead it returns the EAGAIN. 1195 */ 1196 error = getsock_cap(td, args->s, &cap_send_rights, &fp, 1197 &fflag, NULL); 1198 if (error == 0) { 1199 if (fflag & FNONBLOCK) 1200 error = EAGAIN; 1201 fdrop(fp, td); 1202 } 1203 } 1204 return (error); 1205 } 1206 1207 struct linux_recv_args { 1208 register_t s; 1209 register_t msg; 1210 register_t len; 1211 register_t flags; 1212 }; 1213 1214 static int 1215 linux_recv(struct thread *td, struct linux_recv_args *args) 1216 { 1217 struct recvfrom_args /* { 1218 int s; 1219 caddr_t buf; 1220 int len; 1221 int flags; 1222 struct sockaddr *from; 1223 socklen_t fromlenaddr; 1224 } */ bsd_args; 1225 1226 bsd_args.s = args->s; 1227 bsd_args.buf = (caddr_t)PTRIN(args->msg); 1228 bsd_args.len = args->len; 1229 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 1230 bsd_args.from = NULL; 1231 bsd_args.fromlenaddr = 0; 1232 return (sys_recvfrom(td, &bsd_args)); 1233 } 1234 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1235 1236 int 1237 linux_sendto(struct thread *td, struct linux_sendto_args *args) 1238 { 1239 struct msghdr msg; 1240 struct iovec aiov; 1241 1242 if (linux_check_hdrincl(td, args->s) == 0) 1243 /* IP_HDRINCL set, tweak the packet before sending */ 1244 return (linux_sendto_hdrincl(td, args)); 1245 1246 msg.msg_name = PTRIN(args->to); 1247 msg.msg_namelen = args->tolen; 1248 msg.msg_iov = &aiov; 1249 msg.msg_iovlen = 1; 1250 msg.msg_control = NULL; 1251 msg.msg_flags = 0; 1252 aiov.iov_base = PTRIN(args->msg); 1253 aiov.iov_len = args->len; 1254 return (linux_sendit(td, args->s, &msg, args->flags, NULL, 1255 UIO_USERSPACE)); 1256 } 1257 1258 int 1259 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 1260 { 1261 struct sockaddr *sa; 1262 struct msghdr msg; 1263 struct iovec aiov; 1264 int error, fromlen; 1265 1266 if (PTRIN(args->fromlen) != NULL) { 1267 error = copyin(PTRIN(args->fromlen), &fromlen, 1268 sizeof(fromlen)); 1269 if (error != 0) 1270 return (error); 1271 if (fromlen < 0) 1272 return (EINVAL); 1273 sa = malloc(fromlen, M_SONAME, M_WAITOK); 1274 } else { 1275 fromlen = 0; 1276 sa = NULL; 1277 } 1278 1279 msg.msg_name = sa; 1280 msg.msg_namelen = fromlen; 1281 msg.msg_iov = &aiov; 1282 msg.msg_iovlen = 1; 1283 aiov.iov_base = PTRIN(args->buf); 1284 aiov.iov_len = args->len; 1285 msg.msg_control = 0; 1286 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 1287 1288 error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL); 1289 if (error != 0) 1290 goto out; 1291 1292 if (PTRIN(args->from) != NULL) 1293 error = linux_copyout_sockaddr(sa, PTRIN(args->from), msg.msg_namelen); 1294 1295 if (error == 0 && PTRIN(args->fromlen) != NULL) 1296 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 1297 sizeof(msg.msg_namelen)); 1298 out: 1299 free(sa, M_SONAME); 1300 return (error); 1301 } 1302 1303 static int 1304 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1305 l_uint flags) 1306 { 1307 struct cmsghdr *cmsg; 1308 struct mbuf *control; 1309 struct msghdr msg; 1310 struct l_cmsghdr linux_cmsg; 1311 struct l_cmsghdr *ptr_cmsg; 1312 struct l_msghdr linux_msghdr; 1313 struct iovec *iov; 1314 socklen_t datalen; 1315 struct sockaddr *sa; 1316 struct socket *so; 1317 sa_family_t sa_family; 1318 struct file *fp; 1319 void *data; 1320 l_size_t len; 1321 l_size_t clen; 1322 int error, fflag; 1323 1324 error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr)); 1325 if (error != 0) 1326 return (error); 1327 1328 /* 1329 * Some Linux applications (ping) define a non-NULL control data 1330 * pointer, but a msg_controllen of 0, which is not allowed in the 1331 * FreeBSD system call interface. NULL the msg_control pointer in 1332 * order to handle this case. This should be checked, but allows the 1333 * Linux ping to work. 1334 */ 1335 if (PTRIN(linux_msghdr.msg_control) != NULL && 1336 linux_msghdr.msg_controllen == 0) 1337 linux_msghdr.msg_control = PTROUT(NULL); 1338 1339 error = linux_to_bsd_msghdr(&msg, &linux_msghdr); 1340 if (error != 0) 1341 return (error); 1342 1343 #ifdef COMPAT_LINUX32 1344 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1345 &iov, EMSGSIZE); 1346 #else 1347 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1348 #endif 1349 if (error != 0) 1350 return (error); 1351 1352 control = NULL; 1353 1354 error = kern_getsockname(td, s, &sa, &datalen); 1355 if (error != 0) 1356 goto bad; 1357 sa_family = sa->sa_family; 1358 free(sa, M_SONAME); 1359 1360 if (flags & LINUX_MSG_OOB) { 1361 error = EOPNOTSUPP; 1362 if (sa_family == AF_UNIX) 1363 goto bad; 1364 1365 error = getsock_cap(td, s, &cap_send_rights, &fp, 1366 &fflag, NULL); 1367 if (error != 0) 1368 goto bad; 1369 so = fp->f_data; 1370 if (so->so_type != SOCK_STREAM) 1371 error = EOPNOTSUPP; 1372 fdrop(fp, td); 1373 if (error != 0) 1374 goto bad; 1375 } 1376 1377 if (linux_msghdr.msg_controllen >= sizeof(struct l_cmsghdr)) { 1378 error = ENOBUFS; 1379 control = m_get(M_WAITOK, MT_CONTROL); 1380 MCLGET(control, M_WAITOK); 1381 data = mtod(control, void *); 1382 datalen = 0; 1383 1384 ptr_cmsg = PTRIN(linux_msghdr.msg_control); 1385 clen = linux_msghdr.msg_controllen; 1386 do { 1387 error = copyin(ptr_cmsg, &linux_cmsg, 1388 sizeof(struct l_cmsghdr)); 1389 if (error != 0) 1390 goto bad; 1391 1392 error = EINVAL; 1393 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) || 1394 linux_cmsg.cmsg_len > clen) 1395 goto bad; 1396 1397 if (datalen + CMSG_HDRSZ > MCLBYTES) 1398 goto bad; 1399 1400 /* 1401 * Now we support only SCM_RIGHTS and SCM_CRED, 1402 * so return EINVAL in any other cmsg_type 1403 */ 1404 cmsg = data; 1405 cmsg->cmsg_type = 1406 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1407 cmsg->cmsg_level = 1408 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1409 if (cmsg->cmsg_type == -1 1410 || cmsg->cmsg_level != SOL_SOCKET) { 1411 linux_msg(curthread, 1412 "unsupported sendmsg cmsg level %d type %d", 1413 linux_cmsg.cmsg_level, linux_cmsg.cmsg_type); 1414 goto bad; 1415 } 1416 1417 /* 1418 * Some applications (e.g. pulseaudio) attempt to 1419 * send ancillary data even if the underlying protocol 1420 * doesn't support it which is not allowed in the 1421 * FreeBSD system call interface. 1422 */ 1423 if (sa_family != AF_UNIX) 1424 goto next; 1425 1426 if (cmsg->cmsg_type == SCM_CREDS) { 1427 len = sizeof(struct cmsgcred); 1428 if (datalen + CMSG_SPACE(len) > MCLBYTES) 1429 goto bad; 1430 1431 /* 1432 * The lower levels will fill in the structure 1433 */ 1434 memset(CMSG_DATA(data), 0, len); 1435 } else { 1436 len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1437 if (datalen + CMSG_SPACE(len) < datalen || 1438 datalen + CMSG_SPACE(len) > MCLBYTES) 1439 goto bad; 1440 1441 error = copyin(LINUX_CMSG_DATA(ptr_cmsg), 1442 CMSG_DATA(data), len); 1443 if (error != 0) 1444 goto bad; 1445 } 1446 1447 cmsg->cmsg_len = CMSG_LEN(len); 1448 data = (char *)data + CMSG_SPACE(len); 1449 datalen += CMSG_SPACE(len); 1450 1451 next: 1452 if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len)) 1453 break; 1454 1455 clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len); 1456 ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg + 1457 LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len)); 1458 } while(clen >= sizeof(struct l_cmsghdr)); 1459 1460 control->m_len = datalen; 1461 if (datalen == 0) { 1462 m_freem(control); 1463 control = NULL; 1464 } 1465 } 1466 1467 msg.msg_iov = iov; 1468 msg.msg_flags = 0; 1469 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1470 control = NULL; 1471 1472 bad: 1473 m_freem(control); 1474 free(iov, M_IOV); 1475 return (error); 1476 } 1477 1478 int 1479 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1480 { 1481 1482 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1483 args->flags)); 1484 } 1485 1486 int 1487 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1488 { 1489 struct l_mmsghdr *msg; 1490 l_uint retval; 1491 int error, datagrams; 1492 1493 if (args->vlen > UIO_MAXIOV) 1494 args->vlen = UIO_MAXIOV; 1495 1496 msg = PTRIN(args->msg); 1497 datagrams = 0; 1498 while (datagrams < args->vlen) { 1499 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1500 args->flags); 1501 if (error != 0) 1502 break; 1503 1504 retval = td->td_retval[0]; 1505 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1506 if (error != 0) 1507 break; 1508 ++msg; 1509 ++datagrams; 1510 } 1511 if (error == 0) 1512 td->td_retval[0] = datagrams; 1513 return (error); 1514 } 1515 1516 static int 1517 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1518 l_uint flags, struct msghdr *msg) 1519 { 1520 struct cmsghdr *cm; 1521 struct cmsgcred *cmcred; 1522 struct sockcred2 *scred; 1523 struct l_cmsghdr *linux_cmsg = NULL; 1524 struct l_ucred linux_ucred; 1525 socklen_t datalen, maxlen, outlen; 1526 struct l_msghdr linux_msghdr; 1527 struct iovec *iov, *uiov; 1528 struct mbuf *control = NULL; 1529 struct mbuf **controlp; 1530 struct timeval *ftmvl; 1531 struct sockaddr *sa; 1532 l_timeval ltmvl; 1533 caddr_t outbuf; 1534 void *data; 1535 int error, i, fd, fds, *fdp; 1536 1537 error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr)); 1538 if (error != 0) 1539 return (error); 1540 1541 error = linux_to_bsd_msghdr(msg, &linux_msghdr); 1542 if (error != 0) 1543 return (error); 1544 1545 #ifdef COMPAT_LINUX32 1546 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1547 &iov, EMSGSIZE); 1548 #else 1549 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1550 #endif 1551 if (error != 0) 1552 return (error); 1553 1554 if (msg->msg_name != NULL && msg->msg_namelen > 0) { 1555 msg->msg_namelen = min(msg->msg_namelen, SOCK_MAXADDRLEN); 1556 sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK); 1557 msg->msg_name = sa; 1558 } else { 1559 sa = NULL; 1560 msg->msg_name = NULL; 1561 } 1562 1563 uiov = msg->msg_iov; 1564 msg->msg_iov = iov; 1565 controlp = (msg->msg_control != NULL) ? &control : NULL; 1566 error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp); 1567 msg->msg_iov = uiov; 1568 if (error != 0) 1569 goto bad; 1570 1571 /* 1572 * Note that kern_recvit() updates msg->msg_namelen. 1573 */ 1574 if (msg->msg_name != NULL && msg->msg_namelen > 0) { 1575 msg->msg_name = PTRIN(linux_msghdr.msg_name); 1576 error = linux_copyout_sockaddr(sa, 1577 PTRIN(msg->msg_name), msg->msg_namelen); 1578 if (error != 0) 1579 goto bad; 1580 } 1581 1582 error = bsd_to_linux_msghdr(msg, &linux_msghdr); 1583 if (error != 0) 1584 goto bad; 1585 1586 maxlen = linux_msghdr.msg_controllen; 1587 linux_msghdr.msg_controllen = 0; 1588 if (control) { 1589 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1590 1591 msg->msg_control = mtod(control, struct cmsghdr *); 1592 msg->msg_controllen = control->m_len; 1593 1594 cm = CMSG_FIRSTHDR(msg); 1595 outbuf = PTRIN(linux_msghdr.msg_control); 1596 outlen = 0; 1597 while (cm != NULL) { 1598 linux_cmsg->cmsg_type = 1599 bsd_to_linux_cmsg_type(cm->cmsg_type); 1600 linux_cmsg->cmsg_level = 1601 bsd_to_linux_sockopt_level(cm->cmsg_level); 1602 if (linux_cmsg->cmsg_type == -1 || 1603 cm->cmsg_level != SOL_SOCKET) { 1604 linux_msg(curthread, 1605 "unsupported recvmsg cmsg level %d type %d", 1606 cm->cmsg_level, cm->cmsg_type); 1607 error = EINVAL; 1608 goto bad; 1609 } 1610 1611 data = CMSG_DATA(cm); 1612 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1613 1614 switch (cm->cmsg_type) { 1615 case SCM_RIGHTS: 1616 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1617 fds = datalen / sizeof(int); 1618 fdp = data; 1619 for (i = 0; i < fds; i++) { 1620 fd = *fdp++; 1621 (void)kern_fcntl(td, fd, 1622 F_SETFD, FD_CLOEXEC); 1623 } 1624 } 1625 break; 1626 1627 case SCM_CREDS: 1628 /* 1629 * Currently LOCAL_CREDS is never in 1630 * effect for Linux so no need to worry 1631 * about sockcred 1632 */ 1633 if (datalen != sizeof(*cmcred)) { 1634 error = EMSGSIZE; 1635 goto bad; 1636 } 1637 cmcred = (struct cmsgcred *)data; 1638 bzero(&linux_ucred, sizeof(linux_ucred)); 1639 linux_ucred.pid = cmcred->cmcred_pid; 1640 linux_ucred.uid = cmcred->cmcred_uid; 1641 linux_ucred.gid = cmcred->cmcred_gid; 1642 data = &linux_ucred; 1643 datalen = sizeof(linux_ucred); 1644 break; 1645 1646 case SCM_CREDS2: 1647 scred = data; 1648 bzero(&linux_ucred, sizeof(linux_ucred)); 1649 linux_ucred.pid = scred->sc_pid; 1650 linux_ucred.uid = scred->sc_uid; 1651 linux_ucred.gid = scred->sc_gid; 1652 data = &linux_ucred; 1653 datalen = sizeof(linux_ucred); 1654 break; 1655 1656 case SCM_TIMESTAMP: 1657 if (datalen != sizeof(struct timeval)) { 1658 error = EMSGSIZE; 1659 goto bad; 1660 } 1661 ftmvl = (struct timeval *)data; 1662 ltmvl.tv_sec = ftmvl->tv_sec; 1663 ltmvl.tv_usec = ftmvl->tv_usec; 1664 data = <mvl; 1665 datalen = sizeof(ltmvl); 1666 break; 1667 } 1668 1669 if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) { 1670 if (outlen == 0) { 1671 error = EMSGSIZE; 1672 goto bad; 1673 } else { 1674 linux_msghdr.msg_flags |= LINUX_MSG_CTRUNC; 1675 m_dispose_extcontrolm(control); 1676 goto out; 1677 } 1678 } 1679 1680 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen); 1681 1682 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ); 1683 if (error != 0) 1684 goto bad; 1685 outbuf += L_CMSG_HDRSZ; 1686 1687 error = copyout(data, outbuf, datalen); 1688 if (error != 0) 1689 goto bad; 1690 1691 outbuf += LINUX_CMSG_ALIGN(datalen); 1692 outlen += LINUX_CMSG_LEN(datalen); 1693 1694 cm = CMSG_NXTHDR(msg, cm); 1695 } 1696 linux_msghdr.msg_controllen = outlen; 1697 } 1698 1699 out: 1700 error = copyout(&linux_msghdr, msghdr, sizeof(linux_msghdr)); 1701 1702 bad: 1703 if (control != NULL) { 1704 if (error != 0) 1705 m_dispose_extcontrolm(control); 1706 m_freem(control); 1707 } 1708 free(iov, M_IOV); 1709 free(linux_cmsg, M_LINUX); 1710 free(sa, M_SONAME); 1711 1712 return (error); 1713 } 1714 1715 int 1716 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1717 { 1718 struct msghdr bsd_msg; 1719 1720 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1721 args->flags, &bsd_msg)); 1722 } 1723 1724 int 1725 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 1726 { 1727 struct l_mmsghdr *msg; 1728 struct msghdr bsd_msg; 1729 struct l_timespec lts; 1730 struct timespec ts, tts; 1731 l_uint retval; 1732 int error, datagrams; 1733 1734 if (args->timeout) { 1735 error = copyin(args->timeout, <s, sizeof(struct l_timespec)); 1736 if (error != 0) 1737 return (error); 1738 error = linux_to_native_timespec(&ts, <s); 1739 if (error != 0) 1740 return (error); 1741 getnanotime(&tts); 1742 timespecadd(&tts, &ts, &tts); 1743 } 1744 1745 msg = PTRIN(args->msg); 1746 datagrams = 0; 1747 while (datagrams < args->vlen) { 1748 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr, 1749 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1750 if (error != 0) 1751 break; 1752 1753 retval = td->td_retval[0]; 1754 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1755 if (error != 0) 1756 break; 1757 ++msg; 1758 ++datagrams; 1759 1760 /* 1761 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1762 */ 1763 if (args->flags & LINUX_MSG_WAITFORONE) 1764 args->flags |= LINUX_MSG_DONTWAIT; 1765 1766 /* 1767 * See BUGS section of recvmmsg(2). 1768 */ 1769 if (args->timeout) { 1770 getnanotime(&ts); 1771 timespecsub(&ts, &tts, &ts); 1772 if (!timespecisset(&ts) || ts.tv_sec > 0) 1773 break; 1774 } 1775 /* Out of band data, return right away. */ 1776 if (bsd_msg.msg_flags & MSG_OOB) 1777 break; 1778 } 1779 if (error == 0) 1780 td->td_retval[0] = datagrams; 1781 return (error); 1782 } 1783 1784 int 1785 linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 1786 { 1787 1788 return (kern_shutdown(td, args->s, args->how)); 1789 } 1790 1791 int 1792 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 1793 { 1794 l_timeval linux_tv; 1795 struct sockaddr *sa; 1796 struct timeval tv; 1797 socklen_t len; 1798 int error, level, name; 1799 1800 level = linux_to_bsd_sockopt_level(args->level); 1801 switch (level) { 1802 case SOL_SOCKET: 1803 name = linux_to_bsd_so_sockopt(args->optname); 1804 switch (name) { 1805 case LOCAL_CREDS_PERSISTENT: 1806 level = SOL_LOCAL; 1807 break; 1808 case SO_RCVTIMEO: 1809 /* FALLTHROUGH */ 1810 case SO_SNDTIMEO: 1811 error = copyin(PTRIN(args->optval), &linux_tv, 1812 sizeof(linux_tv)); 1813 if (error != 0) 1814 return (error); 1815 tv.tv_sec = linux_tv.tv_sec; 1816 tv.tv_usec = linux_tv.tv_usec; 1817 return (kern_setsockopt(td, args->s, level, 1818 name, &tv, UIO_SYSSPACE, sizeof(tv))); 1819 /* NOTREACHED */ 1820 default: 1821 break; 1822 } 1823 break; 1824 case IPPROTO_IP: 1825 if (args->optname == LINUX_IP_RECVERR && 1826 linux_ignore_ip_recverr) { 1827 /* 1828 * XXX: This is a hack to unbreak DNS resolution 1829 * with glibc 2.30 and above. 1830 */ 1831 return (0); 1832 } 1833 name = linux_to_bsd_ip_sockopt(args->optname); 1834 break; 1835 case IPPROTO_IPV6: 1836 name = linux_to_bsd_ip6_sockopt(args->optname); 1837 break; 1838 case IPPROTO_TCP: 1839 name = linux_to_bsd_tcp_sockopt(args->optname); 1840 break; 1841 default: 1842 name = -1; 1843 break; 1844 } 1845 if (name < 0) { 1846 if (name == -1) 1847 linux_msg(curthread, 1848 "unsupported setsockopt level %d optname %d", 1849 args->level, args->optname); 1850 return (ENOPROTOOPT); 1851 } 1852 1853 if (name == IPV6_NEXTHOP) { 1854 len = args->optlen; 1855 error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len); 1856 if (error != 0) 1857 return (error); 1858 1859 error = kern_setsockopt(td, args->s, level, 1860 name, sa, UIO_SYSSPACE, len); 1861 free(sa, M_SONAME); 1862 } else { 1863 error = kern_setsockopt(td, args->s, level, 1864 name, PTRIN(args->optval), UIO_USERSPACE, args->optlen); 1865 } 1866 1867 return (error); 1868 } 1869 1870 static int 1871 linux_getsockopt_so_peergroups(struct thread *td, 1872 struct linux_getsockopt_args *args) 1873 { 1874 struct xucred xu; 1875 socklen_t xulen, len; 1876 int error, i; 1877 1878 xulen = sizeof(xu); 1879 error = kern_getsockopt(td, args->s, 0, 1880 LOCAL_PEERCRED, &xu, UIO_SYSSPACE, &xulen); 1881 if (error != 0) 1882 return (error); 1883 1884 len = xu.cr_ngroups * sizeof(l_gid_t); 1885 if (args->optlen < len) { 1886 error = copyout(&len, PTRIN(args->optlen), sizeof(len)); 1887 if (error == 0) 1888 error = ERANGE; 1889 return (error); 1890 } 1891 1892 /* 1893 * "- 1" to skip the primary group. 1894 */ 1895 for (i = 0; i < xu.cr_ngroups - 1; i++) { 1896 error = copyout(xu.cr_groups + i + 1, 1897 (void *)(args->optval + i * sizeof(l_gid_t)), 1898 sizeof(l_gid_t)); 1899 if (error != 0) 1900 return (error); 1901 } 1902 1903 error = copyout(&len, PTRIN(args->optlen), sizeof(len)); 1904 return (error); 1905 } 1906 1907 static int 1908 linux_getsockopt_so_peersec(struct thread *td, 1909 struct linux_getsockopt_args *args) 1910 { 1911 socklen_t len; 1912 int error; 1913 1914 len = sizeof(SECURITY_CONTEXT_STRING); 1915 if (args->optlen < len) { 1916 error = copyout(&len, PTRIN(args->optlen), sizeof(len)); 1917 if (error == 0) 1918 error = ERANGE; 1919 return (error); 1920 } 1921 1922 error = copyout(SECURITY_CONTEXT_STRING, 1923 PTRIN(args->optval), sizeof(SECURITY_CONTEXT_STRING)); 1924 if (error == 0) 1925 error = copyout(&len, PTRIN(args->optlen), sizeof(len)); 1926 return (error); 1927 } 1928 1929 int 1930 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 1931 { 1932 l_timeval linux_tv; 1933 struct timeval tv; 1934 socklen_t tv_len, xulen, len; 1935 struct sockaddr *sa; 1936 struct xucred xu; 1937 struct l_ucred lxu; 1938 int error, level, name, newval; 1939 1940 level = linux_to_bsd_sockopt_level(args->level); 1941 switch (level) { 1942 case SOL_SOCKET: 1943 switch (args->optname) { 1944 case LINUX_SO_PEERGROUPS: 1945 return (linux_getsockopt_so_peergroups(td, args)); 1946 case LINUX_SO_PEERSEC: 1947 return (linux_getsockopt_so_peersec(td, args)); 1948 default: 1949 break; 1950 } 1951 1952 name = linux_to_bsd_so_sockopt(args->optname); 1953 switch (name) { 1954 case LOCAL_CREDS_PERSISTENT: 1955 level = SOL_LOCAL; 1956 break; 1957 case SO_RCVTIMEO: 1958 /* FALLTHROUGH */ 1959 case SO_SNDTIMEO: 1960 tv_len = sizeof(tv); 1961 error = kern_getsockopt(td, args->s, level, 1962 name, &tv, UIO_SYSSPACE, &tv_len); 1963 if (error != 0) 1964 return (error); 1965 linux_tv.tv_sec = tv.tv_sec; 1966 linux_tv.tv_usec = tv.tv_usec; 1967 return (copyout(&linux_tv, PTRIN(args->optval), 1968 sizeof(linux_tv))); 1969 /* NOTREACHED */ 1970 case LOCAL_PEERCRED: 1971 if (args->optlen < sizeof(lxu)) 1972 return (EINVAL); 1973 /* 1974 * LOCAL_PEERCRED is not served at the SOL_SOCKET level, 1975 * but by the Unix socket's level 0. 1976 */ 1977 level = 0; 1978 xulen = sizeof(xu); 1979 error = kern_getsockopt(td, args->s, level, 1980 name, &xu, UIO_SYSSPACE, &xulen); 1981 if (error != 0) 1982 return (error); 1983 lxu.pid = xu.cr_pid; 1984 lxu.uid = xu.cr_uid; 1985 lxu.gid = xu.cr_gid; 1986 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu))); 1987 /* NOTREACHED */ 1988 case SO_ERROR: 1989 len = sizeof(newval); 1990 error = kern_getsockopt(td, args->s, level, 1991 name, &newval, UIO_SYSSPACE, &len); 1992 if (error != 0) 1993 return (error); 1994 newval = -bsd_to_linux_errno(newval); 1995 return (copyout(&newval, PTRIN(args->optval), len)); 1996 /* NOTREACHED */ 1997 default: 1998 break; 1999 } 2000 break; 2001 case IPPROTO_IP: 2002 name = linux_to_bsd_ip_sockopt(args->optname); 2003 break; 2004 case IPPROTO_IPV6: 2005 name = linux_to_bsd_ip6_sockopt(args->optname); 2006 break; 2007 case IPPROTO_TCP: 2008 name = linux_to_bsd_tcp_sockopt(args->optname); 2009 break; 2010 default: 2011 name = -1; 2012 break; 2013 } 2014 if (name < 0) { 2015 if (name == -1) 2016 linux_msg(curthread, 2017 "unsupported getsockopt level %d optname %d", 2018 args->level, args->optname); 2019 return (EINVAL); 2020 } 2021 2022 if (name == IPV6_NEXTHOP) { 2023 error = copyin(PTRIN(args->optlen), &len, sizeof(len)); 2024 if (error != 0) 2025 return (error); 2026 sa = malloc(len, M_SONAME, M_WAITOK); 2027 2028 error = kern_getsockopt(td, args->s, level, 2029 name, sa, UIO_SYSSPACE, &len); 2030 if (error != 0) 2031 goto out; 2032 2033 error = linux_copyout_sockaddr(sa, PTRIN(args->optval), len); 2034 if (error == 0) 2035 error = copyout(&len, PTRIN(args->optlen), 2036 sizeof(len)); 2037 out: 2038 free(sa, M_SONAME); 2039 } else { 2040 if (args->optval) { 2041 error = copyin(PTRIN(args->optlen), &len, sizeof(len)); 2042 if (error != 0) 2043 return (error); 2044 } 2045 error = kern_getsockopt(td, args->s, level, 2046 name, PTRIN(args->optval), UIO_USERSPACE, &len); 2047 if (error == 0) 2048 error = copyout(&len, PTRIN(args->optlen), 2049 sizeof(len)); 2050 } 2051 2052 return (error); 2053 } 2054 2055 static int 2056 linux_sendfile_common(struct thread *td, l_int out, l_int in, 2057 l_loff_t *offset, l_size_t count) 2058 { 2059 off_t bytes_read; 2060 int error; 2061 l_loff_t current_offset; 2062 struct file *fp; 2063 2064 AUDIT_ARG_FD(in); 2065 error = fget_read(td, in, &cap_pread_rights, &fp); 2066 if (error != 0) 2067 return (error); 2068 2069 if (offset != NULL) { 2070 current_offset = *offset; 2071 } else { 2072 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ? 2073 fo_seek(fp, 0, SEEK_CUR, td) : ESPIPE; 2074 if (error != 0) 2075 goto drop; 2076 current_offset = td->td_uretoff.tdu_off; 2077 } 2078 2079 bytes_read = 0; 2080 2081 /* Linux cannot have 0 count. */ 2082 if (count <= 0 || current_offset < 0) { 2083 error = EINVAL; 2084 goto drop; 2085 } 2086 2087 error = fo_sendfile(fp, out, NULL, NULL, current_offset, count, 2088 &bytes_read, 0, td); 2089 if (error != 0) 2090 goto drop; 2091 current_offset += bytes_read; 2092 2093 if (offset != NULL) { 2094 *offset = current_offset; 2095 } else { 2096 error = fo_seek(fp, current_offset, SEEK_SET, td); 2097 if (error != 0) 2098 goto drop; 2099 } 2100 2101 td->td_retval[0] = (ssize_t)bytes_read; 2102 drop: 2103 fdrop(fp, td); 2104 if (error == ENOTSOCK) 2105 error = EINVAL; 2106 return (error); 2107 } 2108 2109 int 2110 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg) 2111 { 2112 /* 2113 * Differences between FreeBSD and Linux sendfile: 2114 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to 2115 * mean send the whole file.) In linux_sendfile given fds are still 2116 * checked for validity when the count is 0. 2117 * - Linux can send to any fd whereas FreeBSD only supports sockets. 2118 * The same restriction follows for linux_sendfile. 2119 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr. 2120 * - Linux takes an offset pointer and updates it to the read location. 2121 * FreeBSD takes in an offset and a 'bytes read' parameter which is 2122 * only filled if it isn't NULL. We use this parameter to update the 2123 * offset pointer if it exists. 2124 * - Linux sendfile returns bytes read on success while FreeBSD 2125 * returns 0. We use the 'bytes read' parameter to get this value. 2126 */ 2127 2128 l_loff_t offset64; 2129 l_long offset; 2130 int ret; 2131 int error; 2132 2133 if (arg->offset != NULL) { 2134 error = copyin(arg->offset, &offset, sizeof(offset)); 2135 if (error != 0) 2136 return (error); 2137 offset64 = (l_loff_t)offset; 2138 } 2139 2140 ret = linux_sendfile_common(td, arg->out, arg->in, 2141 arg->offset != NULL ? &offset64 : NULL, arg->count); 2142 2143 if (arg->offset != NULL) { 2144 #if defined(__i386__) || defined(__arm__) || \ 2145 (defined(__amd64__) && defined(COMPAT_LINUX32)) 2146 if (offset64 > INT32_MAX) 2147 return (EOVERFLOW); 2148 #endif 2149 offset = (l_long)offset64; 2150 error = copyout(&offset, arg->offset, sizeof(offset)); 2151 if (error != 0) 2152 return (error); 2153 } 2154 2155 return (ret); 2156 } 2157 2158 #if defined(__i386__) || defined(__arm__) || \ 2159 (defined(__amd64__) && defined(COMPAT_LINUX32)) 2160 2161 int 2162 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg) 2163 { 2164 l_loff_t offset; 2165 int ret; 2166 int error; 2167 2168 if (arg->offset != NULL) { 2169 error = copyin(arg->offset, &offset, sizeof(offset)); 2170 if (error != 0) 2171 return (error); 2172 } 2173 2174 ret = linux_sendfile_common(td, arg->out, arg->in, 2175 arg->offset != NULL ? &offset : NULL, arg->count); 2176 2177 if (arg->offset != NULL) { 2178 error = copyout(&offset, arg->offset, sizeof(offset)); 2179 if (error != 0) 2180 return (error); 2181 } 2182 2183 return (ret); 2184 } 2185 2186 /* Argument list sizes for linux_socketcall */ 2187 static const unsigned char lxs_args_cnt[] = { 2188 0 /* unused*/, 3 /* socket */, 2189 3 /* bind */, 3 /* connect */, 2190 2 /* listen */, 3 /* accept */, 2191 3 /* getsockname */, 3 /* getpeername */, 2192 4 /* socketpair */, 4 /* send */, 2193 4 /* recv */, 6 /* sendto */, 2194 6 /* recvfrom */, 2 /* shutdown */, 2195 5 /* setsockopt */, 5 /* getsockopt */, 2196 3 /* sendmsg */, 3 /* recvmsg */, 2197 4 /* accept4 */, 5 /* recvmmsg */, 2198 4 /* sendmmsg */, 4 /* sendfile */ 2199 }; 2200 #define LINUX_ARGS_CNT (nitems(lxs_args_cnt) - 1) 2201 #define LINUX_ARG_SIZE(x) (lxs_args_cnt[x] * sizeof(l_ulong)) 2202 2203 int 2204 linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 2205 { 2206 l_ulong a[6]; 2207 #if defined(__amd64__) && defined(COMPAT_LINUX32) 2208 register_t l_args[6]; 2209 #endif 2210 void *arg; 2211 int error; 2212 2213 if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT) 2214 return (EINVAL); 2215 error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what)); 2216 if (error != 0) 2217 return (error); 2218 2219 #if defined(__amd64__) && defined(COMPAT_LINUX32) 2220 for (int i = 0; i < lxs_args_cnt[args->what]; ++i) 2221 l_args[i] = a[i]; 2222 arg = l_args; 2223 #else 2224 arg = a; 2225 #endif 2226 switch (args->what) { 2227 case LINUX_SOCKET: 2228 return (linux_socket(td, arg)); 2229 case LINUX_BIND: 2230 return (linux_bind(td, arg)); 2231 case LINUX_CONNECT: 2232 return (linux_connect(td, arg)); 2233 case LINUX_LISTEN: 2234 return (linux_listen(td, arg)); 2235 case LINUX_ACCEPT: 2236 return (linux_accept(td, arg)); 2237 case LINUX_GETSOCKNAME: 2238 return (linux_getsockname(td, arg)); 2239 case LINUX_GETPEERNAME: 2240 return (linux_getpeername(td, arg)); 2241 case LINUX_SOCKETPAIR: 2242 return (linux_socketpair(td, arg)); 2243 case LINUX_SEND: 2244 return (linux_send(td, arg)); 2245 case LINUX_RECV: 2246 return (linux_recv(td, arg)); 2247 case LINUX_SENDTO: 2248 return (linux_sendto(td, arg)); 2249 case LINUX_RECVFROM: 2250 return (linux_recvfrom(td, arg)); 2251 case LINUX_SHUTDOWN: 2252 return (linux_shutdown(td, arg)); 2253 case LINUX_SETSOCKOPT: 2254 return (linux_setsockopt(td, arg)); 2255 case LINUX_GETSOCKOPT: 2256 return (linux_getsockopt(td, arg)); 2257 case LINUX_SENDMSG: 2258 return (linux_sendmsg(td, arg)); 2259 case LINUX_RECVMSG: 2260 return (linux_recvmsg(td, arg)); 2261 case LINUX_ACCEPT4: 2262 return (linux_accept4(td, arg)); 2263 case LINUX_RECVMMSG: 2264 return (linux_recvmmsg(td, arg)); 2265 case LINUX_SENDMMSG: 2266 return (linux_sendmmsg(td, arg)); 2267 case LINUX_SENDFILE: 2268 return (linux_sendfile(td, arg)); 2269 } 2270 2271 linux_msg(td, "socket type %d not implemented", args->what); 2272 return (ENOSYS); 2273 } 2274 #endif /* __i386__ || __arm__ || (__amd64__ && COMPAT_LINUX32) */ 2275