1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 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 "opt_inet6.h" 30 31 #include <sys/param.h> 32 #include <sys/capsicum.h> 33 #include <sys/filedesc.h> 34 #include <sys/limits.h> 35 #include <sys/malloc.h> 36 #include <sys/mbuf.h> 37 #include <sys/proc.h> 38 #include <sys/protosw.h> 39 #include <sys/socket.h> 40 #include <sys/socketvar.h> 41 #include <sys/syscallsubr.h> 42 #include <sys/sysproto.h> 43 #include <sys/vnode.h> 44 #include <sys/un.h> 45 #include <sys/unistd.h> 46 47 #include <security/audit/audit.h> 48 49 #include <net/if.h> 50 #include <net/vnet.h> 51 #include <netinet/in.h> 52 #include <netinet/ip.h> 53 #include <netinet/tcp.h> 54 #ifdef INET6 55 #include <netinet/ip6.h> 56 #include <netinet6/ip6_var.h> 57 #endif 58 59 #ifdef COMPAT_LINUX32 60 #include <compat/freebsd32/freebsd32_util.h> 61 #include <machine/../linux32/linux.h> 62 #include <machine/../linux32/linux32_proto.h> 63 #else 64 #include <machine/../linux/linux.h> 65 #include <machine/../linux/linux_proto.h> 66 #endif 67 #include <compat/linux/linux_common.h> 68 #include <compat/linux/linux_emul.h> 69 #include <compat/linux/linux_file.h> 70 #include <compat/linux/linux_mib.h> 71 #include <compat/linux/linux_socket.h> 72 #include <compat/linux/linux_time.h> 73 #include <compat/linux/linux_util.h> 74 75 _Static_assert(offsetof(struct l_ifreq, ifr_ifru) == 76 offsetof(struct ifreq, ifr_ifru), 77 "Linux ifreq members names should be equal to FreeeBSD"); 78 _Static_assert(offsetof(struct l_ifreq, ifr_index) == 79 offsetof(struct ifreq, ifr_index), 80 "Linux ifreq members names should be equal to FreeeBSD"); 81 _Static_assert(offsetof(struct l_ifreq, ifr_name) == 82 offsetof(struct ifreq, ifr_name), 83 "Linux ifreq members names should be equal to FreeeBSD"); 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 #define SOL_NETLINK 270 94 95 static int 96 linux_to_bsd_sockopt_level(int level) 97 { 98 99 if (level == LINUX_SOL_SOCKET) 100 return (SOL_SOCKET); 101 /* Remaining values are RFC-defined protocol numbers. */ 102 return (level); 103 } 104 105 static int 106 bsd_to_linux_sockopt_level(int level) 107 { 108 109 if (level == SOL_SOCKET) 110 return (LINUX_SOL_SOCKET); 111 return (level); 112 } 113 114 static int 115 linux_to_bsd_ip_sockopt(int opt) 116 { 117 118 switch (opt) { 119 /* known and translated sockopts */ 120 case LINUX_IP_TOS: 121 return (IP_TOS); 122 case LINUX_IP_TTL: 123 return (IP_TTL); 124 case LINUX_IP_HDRINCL: 125 return (IP_HDRINCL); 126 case LINUX_IP_OPTIONS: 127 return (IP_OPTIONS); 128 case LINUX_IP_RECVOPTS: 129 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVOPTS"); 130 return (IP_RECVOPTS); 131 case LINUX_IP_RETOPTS: 132 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_REETOPTS"); 133 return (IP_RETOPTS); 134 case LINUX_IP_RECVTTL: 135 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTTL"); 136 return (IP_RECVTTL); 137 case LINUX_IP_RECVTOS: 138 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTOS"); 139 return (IP_RECVTOS); 140 case LINUX_IP_FREEBIND: 141 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_FREEBIND"); 142 return (IP_BINDANY); 143 case LINUX_IP_IPSEC_POLICY: 144 /* we have this option, but not documented in ip(4) manpage */ 145 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_IPSEC_POLICY"); 146 return (IP_IPSEC_POLICY); 147 case LINUX_IP_MINTTL: 148 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MINTTL"); 149 return (IP_MINTTL); 150 case LINUX_IP_MULTICAST_IF: 151 return (IP_MULTICAST_IF); 152 case LINUX_IP_MULTICAST_TTL: 153 return (IP_MULTICAST_TTL); 154 case LINUX_IP_MULTICAST_LOOP: 155 return (IP_MULTICAST_LOOP); 156 case LINUX_IP_ADD_MEMBERSHIP: 157 return (IP_ADD_MEMBERSHIP); 158 case LINUX_IP_DROP_MEMBERSHIP: 159 return (IP_DROP_MEMBERSHIP); 160 case LINUX_IP_UNBLOCK_SOURCE: 161 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_UNBLOCK_SOURCE"); 162 return (IP_UNBLOCK_SOURCE); 163 case LINUX_IP_BLOCK_SOURCE: 164 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_BLOCK_SOURCE"); 165 return (IP_BLOCK_SOURCE); 166 case LINUX_IP_ADD_SOURCE_MEMBERSHIP: 167 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_ADD_SOURCE_MEMBERSHIP"); 168 return (IP_ADD_SOURCE_MEMBERSHIP); 169 case LINUX_IP_DROP_SOURCE_MEMBERSHIP: 170 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_DROP_SOURCE_MEMBERSHIP"); 171 return (IP_DROP_SOURCE_MEMBERSHIP); 172 case LINUX_MCAST_JOIN_GROUP: 173 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_JOIN_GROUP"); 174 return (MCAST_JOIN_GROUP); 175 case LINUX_MCAST_LEAVE_GROUP: 176 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_LEAVE_GROUP"); 177 return (MCAST_LEAVE_GROUP); 178 case LINUX_MCAST_JOIN_SOURCE_GROUP: 179 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_JOIN_SOURCE_GROUP"); 180 return (MCAST_JOIN_SOURCE_GROUP); 181 case LINUX_MCAST_LEAVE_SOURCE_GROUP: 182 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_LEAVE_SOURCE_GROUP"); 183 return (MCAST_LEAVE_SOURCE_GROUP); 184 case LINUX_IP_RECVORIGDSTADDR: 185 return (IP_RECVORIGDSTADDR); 186 187 /* known but not implemented sockopts */ 188 case LINUX_IP_ROUTER_ALERT: 189 LINUX_RATELIMIT_MSG_OPT1( 190 "unsupported IPv4 socket option IP_ROUTER_ALERT (%d), you can not do user-space routing from linux programs", 191 opt); 192 return (-2); 193 case LINUX_IP_PKTINFO: 194 LINUX_RATELIMIT_MSG_OPT1( 195 "unsupported IPv4 socket option IP_PKTINFO (%d), you can not get extended packet info for datagram sockets in linux programs", 196 opt); 197 return (-2); 198 case LINUX_IP_PKTOPTIONS: 199 LINUX_RATELIMIT_MSG_OPT1( 200 "unsupported IPv4 socket option IP_PKTOPTIONS (%d)", 201 opt); 202 return (-2); 203 case LINUX_IP_MTU_DISCOVER: 204 LINUX_RATELIMIT_MSG_OPT1( 205 "unsupported IPv4 socket option IP_MTU_DISCOVER (%d), your linux program can not control path-MTU discovery", 206 opt); 207 return (-2); 208 case LINUX_IP_RECVERR: 209 /* needed by steam */ 210 LINUX_RATELIMIT_MSG_OPT1( 211 "unsupported IPv4 socket option IP_RECVERR (%d), you can not get extended reliability info in linux programs", 212 opt); 213 return (-2); 214 case LINUX_IP_MTU: 215 LINUX_RATELIMIT_MSG_OPT1( 216 "unsupported IPv4 socket option IP_MTU (%d), your linux program can not control the MTU on this socket", 217 opt); 218 return (-2); 219 case LINUX_IP_XFRM_POLICY: 220 LINUX_RATELIMIT_MSG_OPT1( 221 "unsupported IPv4 socket option IP_XFRM_POLICY (%d)", 222 opt); 223 return (-2); 224 case LINUX_IP_PASSSEC: 225 /* needed by steam */ 226 LINUX_RATELIMIT_MSG_OPT1( 227 "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", 228 opt); 229 return (-2); 230 case LINUX_IP_TRANSPARENT: 231 /* IP_BINDANY or more? */ 232 LINUX_RATELIMIT_MSG_OPT1( 233 "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", 234 opt); 235 return (-2); 236 case LINUX_IP_NODEFRAG: 237 LINUX_RATELIMIT_MSG_OPT1( 238 "unsupported IPv4 socket option IP_NODEFRAG (%d)", 239 opt); 240 return (-2); 241 case LINUX_IP_CHECKSUM: 242 LINUX_RATELIMIT_MSG_OPT1( 243 "unsupported IPv4 socket option IP_CHECKSUM (%d)", 244 opt); 245 return (-2); 246 case LINUX_IP_BIND_ADDRESS_NO_PORT: 247 LINUX_RATELIMIT_MSG_OPT1( 248 "unsupported IPv4 socket option IP_BIND_ADDRESS_NO_PORT (%d)", 249 opt); 250 return (-2); 251 case LINUX_IP_RECVFRAGSIZE: 252 LINUX_RATELIMIT_MSG_OPT1( 253 "unsupported IPv4 socket option IP_RECVFRAGSIZE (%d)", 254 opt); 255 return (-2); 256 case LINUX_MCAST_MSFILTER: 257 LINUX_RATELIMIT_MSG_OPT1( 258 "unsupported IPv4 socket option IP_MCAST_MSFILTER (%d)", 259 opt); 260 return (-2); 261 case LINUX_IP_MULTICAST_ALL: 262 LINUX_RATELIMIT_MSG_OPT1( 263 "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", 264 opt); 265 return (-2); 266 case LINUX_IP_UNICAST_IF: 267 LINUX_RATELIMIT_MSG_OPT1( 268 "unsupported IPv4 socket option IP_UNICAST_IF (%d)", 269 opt); 270 return (-2); 271 272 /* unknown sockopts */ 273 default: 274 return (-1); 275 } 276 } 277 278 static int 279 linux_to_bsd_ip6_sockopt(int opt) 280 { 281 282 switch (opt) { 283 /* known and translated sockopts */ 284 case LINUX_IPV6_2292PKTINFO: 285 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292PKTINFO"); 286 return (IPV6_2292PKTINFO); 287 case LINUX_IPV6_2292HOPOPTS: 288 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292HOPOPTS"); 289 return (IPV6_2292HOPOPTS); 290 case LINUX_IPV6_2292DSTOPTS: 291 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292DSTOPTS"); 292 return (IPV6_2292DSTOPTS); 293 case LINUX_IPV6_2292RTHDR: 294 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292RTHDR"); 295 return (IPV6_2292RTHDR); 296 case LINUX_IPV6_2292PKTOPTIONS: 297 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292PKTOPTIONS"); 298 return (IPV6_2292PKTOPTIONS); 299 case LINUX_IPV6_CHECKSUM: 300 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_CHECKSUM"); 301 return (IPV6_CHECKSUM); 302 case LINUX_IPV6_2292HOPLIMIT: 303 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292HOPLIMIT"); 304 return (IPV6_2292HOPLIMIT); 305 case LINUX_IPV6_NEXTHOP: 306 return (IPV6_NEXTHOP); 307 case LINUX_IPV6_UNICAST_HOPS: 308 return (IPV6_UNICAST_HOPS); 309 case LINUX_IPV6_MULTICAST_IF: 310 return (IPV6_MULTICAST_IF); 311 case LINUX_IPV6_MULTICAST_HOPS: 312 return (IPV6_MULTICAST_HOPS); 313 case LINUX_IPV6_MULTICAST_LOOP: 314 return (IPV6_MULTICAST_LOOP); 315 case LINUX_IPV6_ADD_MEMBERSHIP: 316 return (IPV6_JOIN_GROUP); 317 case LINUX_IPV6_DROP_MEMBERSHIP: 318 return (IPV6_LEAVE_GROUP); 319 case LINUX_IPV6_V6ONLY: 320 return (IPV6_V6ONLY); 321 case LINUX_IPV6_IPSEC_POLICY: 322 /* we have this option, but not documented in ip6(4) manpage */ 323 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_IPSEC_POLICY"); 324 return (IPV6_IPSEC_POLICY); 325 case LINUX_MCAST_JOIN_GROUP: 326 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_JOIN_GROUP"); 327 return (IPV6_JOIN_GROUP); 328 case LINUX_MCAST_LEAVE_GROUP: 329 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_LEAVE_GROUP"); 330 return (IPV6_LEAVE_GROUP); 331 case LINUX_IPV6_RECVPKTINFO: 332 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVPKTINFO"); 333 return (IPV6_RECVPKTINFO); 334 case LINUX_IPV6_PKTINFO: 335 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_PKTINFO"); 336 return (IPV6_PKTINFO); 337 case LINUX_IPV6_RECVHOPLIMIT: 338 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVHOPLIMIT"); 339 return (IPV6_RECVHOPLIMIT); 340 case LINUX_IPV6_HOPLIMIT: 341 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_HOPLIMIT"); 342 return (IPV6_HOPLIMIT); 343 case LINUX_IPV6_RECVHOPOPTS: 344 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVHOPOPTS"); 345 return (IPV6_RECVHOPOPTS); 346 case LINUX_IPV6_HOPOPTS: 347 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_HOPOPTS"); 348 return (IPV6_HOPOPTS); 349 case LINUX_IPV6_RTHDRDSTOPTS: 350 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RTHDRDSTOPTS"); 351 return (IPV6_RTHDRDSTOPTS); 352 case LINUX_IPV6_RECVRTHDR: 353 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVRTHDR"); 354 return (IPV6_RECVRTHDR); 355 case LINUX_IPV6_RTHDR: 356 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RTHDR"); 357 return (IPV6_RTHDR); 358 case LINUX_IPV6_RECVDSTOPTS: 359 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVDSTOPTS"); 360 return (IPV6_RECVDSTOPTS); 361 case LINUX_IPV6_DSTOPTS: 362 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_DSTOPTS"); 363 return (IPV6_DSTOPTS); 364 case LINUX_IPV6_RECVPATHMTU: 365 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVPATHMTU"); 366 return (IPV6_RECVPATHMTU); 367 case LINUX_IPV6_PATHMTU: 368 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_PATHMTU"); 369 return (IPV6_PATHMTU); 370 case LINUX_IPV6_DONTFRAG: 371 return (IPV6_DONTFRAG); 372 case LINUX_IPV6_AUTOFLOWLABEL: 373 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_AUTOFLOWLABEL"); 374 return (IPV6_AUTOFLOWLABEL); 375 case LINUX_IPV6_ORIGDSTADDR: 376 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_ORIGDSTADDR"); 377 return (IPV6_ORIGDSTADDR); 378 case LINUX_IPV6_FREEBIND: 379 LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_FREEBIND"); 380 return (IPV6_BINDANY); 381 382 /* known but not implemented sockopts */ 383 case LINUX_IPV6_ADDRFORM: 384 LINUX_RATELIMIT_MSG_OPT1( 385 "unsupported IPv6 socket option IPV6_ADDRFORM (%d), you linux program can not convert the socket to IPv4", 386 opt); 387 return (-2); 388 case LINUX_IPV6_AUTHHDR: 389 LINUX_RATELIMIT_MSG_OPT1( 390 "unsupported IPv6 socket option IPV6_AUTHHDR (%d), your linux program can not get the authentication header info of IPv6 packets", 391 opt); 392 return (-2); 393 case LINUX_IPV6_FLOWINFO: 394 LINUX_RATELIMIT_MSG_OPT1( 395 "unsupported IPv6 socket option IPV6_FLOWINFO (%d), your linux program can not get the flowid of IPv6 packets", 396 opt); 397 return (-2); 398 case LINUX_IPV6_ROUTER_ALERT: 399 LINUX_RATELIMIT_MSG_OPT1( 400 "unsupported IPv6 socket option IPV6_ROUTER_ALERT (%d), you can not do user-space routing from linux programs", 401 opt); 402 return (-2); 403 case LINUX_IPV6_MTU_DISCOVER: 404 LINUX_RATELIMIT_MSG_OPT1( 405 "unsupported IPv6 socket option IPV6_MTU_DISCOVER (%d), your linux program can not control path-MTU discovery", 406 opt); 407 return (-2); 408 case LINUX_IPV6_MTU: 409 LINUX_RATELIMIT_MSG_OPT1( 410 "unsupported IPv6 socket option IPV6_MTU (%d), your linux program can not control the MTU on this socket", 411 opt); 412 return (-2); 413 case LINUX_IPV6_JOIN_ANYCAST: 414 LINUX_RATELIMIT_MSG_OPT1( 415 "unsupported IPv6 socket option IPV6_JOIN_ANYCAST (%d)", 416 opt); 417 return (-2); 418 case LINUX_IPV6_LEAVE_ANYCAST: 419 LINUX_RATELIMIT_MSG_OPT1( 420 "unsupported IPv6 socket option IPV6_LEAVE_ANYCAST (%d)", 421 opt); 422 return (-2); 423 case LINUX_IPV6_MULTICAST_ALL: 424 LINUX_RATELIMIT_MSG_OPT1( 425 "unsupported IPv6 socket option IPV6_MULTICAST_ALL (%d)", 426 opt); 427 return (-2); 428 case LINUX_IPV6_ROUTER_ALERT_ISOLATE: 429 LINUX_RATELIMIT_MSG_OPT1( 430 "unsupported IPv6 socket option IPV6_ROUTER_ALERT_ISOLATE (%d)", 431 opt); 432 return (-2); 433 case LINUX_IPV6_FLOWLABEL_MGR: 434 LINUX_RATELIMIT_MSG_OPT1( 435 "unsupported IPv6 socket option IPV6_FLOWLABEL_MGR (%d)", 436 opt); 437 return (-2); 438 case LINUX_IPV6_FLOWINFO_SEND: 439 LINUX_RATELIMIT_MSG_OPT1( 440 "unsupported IPv6 socket option IPV6_FLOWINFO_SEND (%d)", 441 opt); 442 return (-2); 443 case LINUX_IPV6_XFRM_POLICY: 444 LINUX_RATELIMIT_MSG_OPT1( 445 "unsupported IPv6 socket option IPV6_XFRM_POLICY (%d)", 446 opt); 447 return (-2); 448 case LINUX_IPV6_HDRINCL: 449 LINUX_RATELIMIT_MSG_OPT1( 450 "unsupported IPv6 socket option IPV6_HDRINCL (%d)", 451 opt); 452 return (-2); 453 case LINUX_MCAST_BLOCK_SOURCE: 454 LINUX_RATELIMIT_MSG_OPT1( 455 "unsupported IPv6 socket option MCAST_BLOCK_SOURCE (%d), your linux program may see more multicast stuff than it wants", 456 opt); 457 return (-2); 458 case LINUX_MCAST_UNBLOCK_SOURCE: 459 LINUX_RATELIMIT_MSG_OPT1( 460 "unsupported IPv6 socket option MCAST_UNBLOCK_SOURCE (%d), your linux program may not see all the multicast stuff it wants", 461 opt); 462 return (-2); 463 case LINUX_MCAST_JOIN_SOURCE_GROUP: 464 LINUX_RATELIMIT_MSG_OPT1( 465 "unsupported IPv6 socket option MCAST_JOIN_SOURCE_GROUP (%d), your linux program is not able to join a multicast source group", 466 opt); 467 return (-2); 468 case LINUX_MCAST_LEAVE_SOURCE_GROUP: 469 LINUX_RATELIMIT_MSG_OPT1( 470 "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", 471 opt); 472 return (-2); 473 case LINUX_MCAST_MSFILTER: 474 LINUX_RATELIMIT_MSG_OPT1( 475 "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", 476 opt); 477 return (-2); 478 case LINUX_IPV6_ADDR_PREFERENCES: 479 LINUX_RATELIMIT_MSG_OPT1( 480 "unsupported IPv6 socket option IPV6_ADDR_PREFERENCES (%d)", 481 opt); 482 return (-2); 483 case LINUX_IPV6_MINHOPCOUNT: 484 LINUX_RATELIMIT_MSG_OPT1( 485 "unsupported IPv6 socket option IPV6_MINHOPCOUNT (%d)", 486 opt); 487 return (-2); 488 case LINUX_IPV6_TRANSPARENT: 489 /* IP_BINDANY or more? */ 490 LINUX_RATELIMIT_MSG_OPT1( 491 "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", 492 opt); 493 return (-2); 494 case LINUX_IPV6_UNICAST_IF: 495 LINUX_RATELIMIT_MSG_OPT1( 496 "unsupported IPv6 socket option IPV6_UNICAST_IF (%d)", 497 opt); 498 return (-2); 499 case LINUX_IPV6_RECVFRAGSIZE: 500 LINUX_RATELIMIT_MSG_OPT1( 501 "unsupported IPv6 socket option IPV6_RECVFRAGSIZE (%d)", 502 opt); 503 return (-2); 504 case LINUX_IPV6_RECVERR: 505 LINUX_RATELIMIT_MSG_OPT1( 506 "unsupported IPv6 socket option IPV6_RECVERR (%d), you can not get extended reliability info in linux programs", 507 opt); 508 return (-2); 509 510 /* unknown sockopts */ 511 default: 512 return (-1); 513 } 514 } 515 516 static int 517 linux_to_bsd_so_sockopt(int opt) 518 { 519 520 switch (opt) { 521 case LINUX_SO_DEBUG: 522 return (SO_DEBUG); 523 case LINUX_SO_REUSEADDR: 524 return (SO_REUSEADDR); 525 case LINUX_SO_TYPE: 526 return (SO_TYPE); 527 case LINUX_SO_ERROR: 528 return (SO_ERROR); 529 case LINUX_SO_DONTROUTE: 530 return (SO_DONTROUTE); 531 case LINUX_SO_BROADCAST: 532 return (SO_BROADCAST); 533 case LINUX_SO_SNDBUF: 534 case LINUX_SO_SNDBUFFORCE: 535 return (SO_SNDBUF); 536 case LINUX_SO_RCVBUF: 537 case LINUX_SO_RCVBUFFORCE: 538 return (SO_RCVBUF); 539 case LINUX_SO_KEEPALIVE: 540 return (SO_KEEPALIVE); 541 case LINUX_SO_OOBINLINE: 542 return (SO_OOBINLINE); 543 case LINUX_SO_LINGER: 544 return (SO_LINGER); 545 case LINUX_SO_REUSEPORT: 546 return (SO_REUSEPORT_LB); 547 case LINUX_SO_PASSCRED: 548 return (LOCAL_CREDS_PERSISTENT); 549 case LINUX_SO_PEERCRED: 550 return (LOCAL_PEERCRED); 551 case LINUX_SO_RCVLOWAT: 552 return (SO_RCVLOWAT); 553 case LINUX_SO_SNDLOWAT: 554 return (SO_SNDLOWAT); 555 case LINUX_SO_RCVTIMEO: 556 return (SO_RCVTIMEO); 557 case LINUX_SO_SNDTIMEO: 558 return (SO_SNDTIMEO); 559 case LINUX_SO_TIMESTAMPO: 560 case LINUX_SO_TIMESTAMPN: 561 return (SO_TIMESTAMP); 562 case LINUX_SO_TIMESTAMPNSO: 563 case LINUX_SO_TIMESTAMPNSN: 564 return (SO_BINTIME); 565 case LINUX_SO_ACCEPTCONN: 566 return (SO_ACCEPTCONN); 567 case LINUX_SO_PROTOCOL: 568 return (SO_PROTOCOL); 569 case LINUX_SO_DOMAIN: 570 return (SO_DOMAIN); 571 } 572 return (-1); 573 } 574 575 static int 576 linux_to_bsd_tcp_sockopt(int opt) 577 { 578 579 switch (opt) { 580 case LINUX_TCP_NODELAY: 581 return (TCP_NODELAY); 582 case LINUX_TCP_MAXSEG: 583 return (TCP_MAXSEG); 584 case LINUX_TCP_CORK: 585 return (TCP_NOPUSH); 586 case LINUX_TCP_KEEPIDLE: 587 return (TCP_KEEPIDLE); 588 case LINUX_TCP_KEEPINTVL: 589 return (TCP_KEEPINTVL); 590 case LINUX_TCP_KEEPCNT: 591 return (TCP_KEEPCNT); 592 case LINUX_TCP_INFO: 593 LINUX_RATELIMIT_MSG_OPT1( 594 "unsupported TCP socket option TCP_INFO (%d)", opt); 595 return (-2); 596 case LINUX_TCP_MD5SIG: 597 return (TCP_MD5SIG); 598 } 599 return (-1); 600 } 601 602 static int 603 linux_to_bsd_msg_flags(int flags) 604 { 605 int ret_flags = 0; 606 607 if (flags & LINUX_MSG_OOB) 608 ret_flags |= MSG_OOB; 609 if (flags & LINUX_MSG_PEEK) 610 ret_flags |= MSG_PEEK; 611 if (flags & LINUX_MSG_DONTROUTE) 612 ret_flags |= MSG_DONTROUTE; 613 if (flags & LINUX_MSG_CTRUNC) 614 ret_flags |= MSG_CTRUNC; 615 if (flags & LINUX_MSG_TRUNC) 616 ret_flags |= MSG_TRUNC; 617 if (flags & LINUX_MSG_DONTWAIT) 618 ret_flags |= MSG_DONTWAIT; 619 if (flags & LINUX_MSG_EOR) 620 ret_flags |= MSG_EOR; 621 if (flags & LINUX_MSG_WAITALL) 622 ret_flags |= MSG_WAITALL; 623 if (flags & LINUX_MSG_NOSIGNAL) 624 ret_flags |= MSG_NOSIGNAL; 625 if (flags & LINUX_MSG_PROXY) 626 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_PROXY (%d) not handled", 627 LINUX_MSG_PROXY); 628 if (flags & LINUX_MSG_FIN) 629 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_FIN (%d) not handled", 630 LINUX_MSG_FIN); 631 if (flags & LINUX_MSG_SYN) 632 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_SYN (%d) not handled", 633 LINUX_MSG_SYN); 634 if (flags & LINUX_MSG_CONFIRM) 635 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_CONFIRM (%d) not handled", 636 LINUX_MSG_CONFIRM); 637 if (flags & LINUX_MSG_RST) 638 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_RST (%d) not handled", 639 LINUX_MSG_RST); 640 if (flags & LINUX_MSG_ERRQUEUE) 641 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_ERRQUEUE (%d) not handled", 642 LINUX_MSG_ERRQUEUE); 643 return (ret_flags); 644 } 645 646 static int 647 linux_to_bsd_cmsg_type(int cmsg_type) 648 { 649 650 switch (cmsg_type) { 651 case LINUX_SCM_RIGHTS: 652 return (SCM_RIGHTS); 653 case LINUX_SCM_CREDENTIALS: 654 return (SCM_CREDS); 655 } 656 return (-1); 657 } 658 659 static int 660 bsd_to_linux_ip_cmsg_type(int cmsg_type) 661 { 662 663 switch (cmsg_type) { 664 case IP_RECVORIGDSTADDR: 665 return (LINUX_IP_RECVORIGDSTADDR); 666 } 667 return (-1); 668 } 669 670 static int 671 bsd_to_linux_cmsg_type(struct proc *p, int cmsg_type, int cmsg_level) 672 { 673 struct linux_pemuldata *pem; 674 675 if (cmsg_level == IPPROTO_IP) 676 return (bsd_to_linux_ip_cmsg_type(cmsg_type)); 677 if (cmsg_level != SOL_SOCKET) 678 return (-1); 679 680 pem = pem_find(p); 681 682 switch (cmsg_type) { 683 case SCM_RIGHTS: 684 return (LINUX_SCM_RIGHTS); 685 case SCM_CREDS: 686 return (LINUX_SCM_CREDENTIALS); 687 case SCM_CREDS2: 688 return (LINUX_SCM_CREDENTIALS); 689 case SCM_TIMESTAMP: 690 return (pem->so_timestamp); 691 case SCM_BINTIME: 692 return (pem->so_timestampns); 693 } 694 return (-1); 695 } 696 697 static int 698 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr) 699 { 700 if (lhdr->msg_controllen > INT_MAX) 701 return (ENOBUFS); 702 703 bhdr->msg_name = PTRIN(lhdr->msg_name); 704 bhdr->msg_namelen = lhdr->msg_namelen; 705 bhdr->msg_iov = PTRIN(lhdr->msg_iov); 706 bhdr->msg_iovlen = lhdr->msg_iovlen; 707 bhdr->msg_control = PTRIN(lhdr->msg_control); 708 709 /* 710 * msg_controllen is skipped since BSD and LINUX control messages 711 * are potentially different sizes (e.g. the cred structure used 712 * by SCM_CREDS is different between the two operating system). 713 * 714 * The caller can set it (if necessary) after converting all the 715 * control messages. 716 */ 717 718 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags); 719 return (0); 720 } 721 722 static int 723 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr) 724 { 725 lhdr->msg_name = PTROUT(bhdr->msg_name); 726 lhdr->msg_namelen = bhdr->msg_namelen; 727 lhdr->msg_iov = PTROUT(bhdr->msg_iov); 728 lhdr->msg_iovlen = bhdr->msg_iovlen; 729 lhdr->msg_control = PTROUT(bhdr->msg_control); 730 731 /* 732 * msg_controllen is skipped since BSD and LINUX control messages 733 * are potentially different sizes (e.g. the cred structure used 734 * by SCM_CREDS is different between the two operating system). 735 * 736 * The caller can set it (if necessary) after converting all the 737 * control messages. 738 */ 739 740 /* msg_flags skipped */ 741 return (0); 742 } 743 744 static int 745 linux_set_socket_flags(int lflags, int *flags) 746 { 747 748 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK)) 749 return (EINVAL); 750 if (lflags & LINUX_SOCK_NONBLOCK) 751 *flags |= SOCK_NONBLOCK; 752 if (lflags & LINUX_SOCK_CLOEXEC) 753 *flags |= SOCK_CLOEXEC; 754 return (0); 755 } 756 757 static int 758 linux_copyout_sockaddr(const struct sockaddr *sa, void *uaddr, size_t len) 759 { 760 struct l_sockaddr *lsa; 761 int error; 762 763 error = bsd_to_linux_sockaddr(sa, &lsa, len); 764 if (error != 0) 765 return (error); 766 767 error = copyout(lsa, uaddr, len); 768 free(lsa, M_LINUX); 769 770 return (error); 771 } 772 773 static int 774 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags, 775 struct mbuf *control, enum uio_seg segflg) 776 { 777 struct sockaddr *to; 778 int error, len; 779 780 if (mp->msg_name != NULL) { 781 len = mp->msg_namelen; 782 error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len); 783 if (error != 0) 784 return (error); 785 mp->msg_name = to; 786 } else 787 to = NULL; 788 789 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control, 790 segflg); 791 792 if (to) 793 free(to, M_SONAME); 794 return (error); 795 } 796 797 /* Return 0 if IP_HDRINCL is set for the given socket. */ 798 static int 799 linux_check_hdrincl(struct thread *td, int s) 800 { 801 int error, optval; 802 socklen_t size_val; 803 804 size_val = sizeof(optval); 805 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL, 806 &optval, UIO_SYSSPACE, &size_val); 807 if (error != 0) 808 return (error); 809 810 return (optval == 0); 811 } 812 813 /* 814 * Updated sendto() when IP_HDRINCL is set: 815 * tweak endian-dependent fields in the IP packet. 816 */ 817 static int 818 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args) 819 { 820 /* 821 * linux_ip_copysize defines how many bytes we should copy 822 * from the beginning of the IP packet before we customize it for BSD. 823 * It should include all the fields we modify (ip_len and ip_off). 824 */ 825 #define linux_ip_copysize 8 826 827 struct ip *packet; 828 struct msghdr msg; 829 struct iovec aiov[1]; 830 int error; 831 832 /* Check that the packet isn't too big or too small. */ 833 if (linux_args->len < linux_ip_copysize || 834 linux_args->len > IP_MAXPACKET) 835 return (EINVAL); 836 837 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK); 838 839 /* Make kernel copy of the packet to be sent */ 840 if ((error = copyin(PTRIN(linux_args->msg), packet, 841 linux_args->len))) 842 goto goout; 843 844 /* Convert fields from Linux to BSD raw IP socket format */ 845 packet->ip_len = linux_args->len; 846 packet->ip_off = ntohs(packet->ip_off); 847 848 /* Prepare the msghdr and iovec structures describing the new packet */ 849 msg.msg_name = PTRIN(linux_args->to); 850 msg.msg_namelen = linux_args->tolen; 851 msg.msg_iov = aiov; 852 msg.msg_iovlen = 1; 853 msg.msg_control = NULL; 854 msg.msg_flags = 0; 855 aiov[0].iov_base = (char *)packet; 856 aiov[0].iov_len = linux_args->len; 857 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags, 858 NULL, UIO_SYSSPACE); 859 goout: 860 free(packet, M_LINUX); 861 return (error); 862 } 863 864 static const char *linux_netlink_names[] = { 865 [LINUX_NETLINK_ROUTE] = "ROUTE", 866 [LINUX_NETLINK_SOCK_DIAG] = "SOCK_DIAG", 867 [LINUX_NETLINK_NFLOG] = "NFLOG", 868 [LINUX_NETLINK_SELINUX] = "SELINUX", 869 [LINUX_NETLINK_AUDIT] = "AUDIT", 870 [LINUX_NETLINK_FIB_LOOKUP] = "FIB_LOOKUP", 871 [LINUX_NETLINK_NETFILTER] = "NETFILTER", 872 [LINUX_NETLINK_KOBJECT_UEVENT] = "KOBJECT_UEVENT", 873 }; 874 875 int 876 linux_socket(struct thread *td, struct linux_socket_args *args) 877 { 878 int domain, retval_socket, type; 879 880 type = args->type & LINUX_SOCK_TYPE_MASK; 881 if (type < 0 || type > LINUX_SOCK_MAX) 882 return (EINVAL); 883 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 884 &type); 885 if (retval_socket != 0) 886 return (retval_socket); 887 domain = linux_to_bsd_domain(args->domain); 888 if (domain == -1) { 889 /* Mask off SOCK_NONBLOCK / CLOEXEC for error messages. */ 890 type = args->type & LINUX_SOCK_TYPE_MASK; 891 if (args->domain == LINUX_AF_NETLINK && 892 args->protocol == LINUX_NETLINK_AUDIT) { 893 ; /* Do nothing, quietly. */ 894 } else if (args->domain == LINUX_AF_NETLINK) { 895 const char *nl_name; 896 897 if (args->protocol >= 0 && 898 args->protocol < nitems(linux_netlink_names)) 899 nl_name = linux_netlink_names[args->protocol]; 900 else 901 nl_name = NULL; 902 if (nl_name != NULL) 903 linux_msg(curthread, 904 "unsupported socket(AF_NETLINK, %d, " 905 "NETLINK_%s)", type, nl_name); 906 else 907 linux_msg(curthread, 908 "unsupported socket(AF_NETLINK, %d, %d)", 909 type, args->protocol); 910 } else { 911 linux_msg(curthread, "unsupported socket domain %d, " 912 "type %d, protocol %d", args->domain, type, 913 args->protocol); 914 } 915 return (EAFNOSUPPORT); 916 } 917 918 retval_socket = kern_socket(td, domain, type, args->protocol); 919 if (retval_socket) 920 return (retval_socket); 921 922 if (type == SOCK_RAW 923 && (args->protocol == IPPROTO_RAW || args->protocol == 0) 924 && domain == PF_INET) { 925 /* It's a raw IP socket: set the IP_HDRINCL option. */ 926 int hdrincl; 927 928 hdrincl = 1; 929 /* We ignore any error returned by kern_setsockopt() */ 930 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL, 931 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl)); 932 } 933 #ifdef INET6 934 /* 935 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default 936 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps. 937 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only 938 * sysctl value. 939 */ 940 if (domain == PF_INET6) { 941 int v6only; 942 943 v6only = 0; 944 /* We ignore any error returned by setsockopt() */ 945 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY, 946 &v6only, UIO_SYSSPACE, sizeof(v6only)); 947 } 948 #endif 949 950 return (retval_socket); 951 } 952 953 int 954 linux_bind(struct thread *td, struct linux_bind_args *args) 955 { 956 struct sockaddr *sa; 957 int error; 958 959 error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa, 960 &args->namelen); 961 if (error != 0) 962 return (error); 963 964 error = kern_bindat(td, AT_FDCWD, args->s, sa); 965 free(sa, M_SONAME); 966 967 /* XXX */ 968 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in)) 969 return (EINVAL); 970 return (error); 971 } 972 973 int 974 linux_connect(struct thread *td, struct linux_connect_args *args) 975 { 976 struct socket *so; 977 struct sockaddr *sa; 978 struct file *fp; 979 int error; 980 981 error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa, 982 &args->namelen); 983 if (error != 0) 984 return (error); 985 986 error = kern_connectat(td, AT_FDCWD, args->s, sa); 987 free(sa, M_SONAME); 988 if (error != EISCONN) 989 return (error); 990 991 /* 992 * Linux doesn't return EISCONN the first time it occurs, 993 * when on a non-blocking socket. Instead it returns the 994 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD. 995 */ 996 error = getsock(td, args->s, &cap_connect_rights, &fp); 997 if (error != 0) 998 return (error); 999 1000 error = EISCONN; 1001 so = fp->f_data; 1002 if (atomic_load_int(&fp->f_flag) & FNONBLOCK) { 1003 SOCK_LOCK(so); 1004 if (so->so_emuldata == 0) 1005 error = so->so_error; 1006 so->so_emuldata = (void *)1; 1007 SOCK_UNLOCK(so); 1008 } 1009 fdrop(fp, td); 1010 1011 return (error); 1012 } 1013 1014 int 1015 linux_listen(struct thread *td, struct linux_listen_args *args) 1016 { 1017 1018 return (kern_listen(td, args->s, args->backlog)); 1019 } 1020 1021 static int 1022 linux_accept_common(struct thread *td, int s, l_uintptr_t addr, 1023 l_uintptr_t namelen, int flags) 1024 { 1025 struct sockaddr *sa; 1026 struct file *fp, *fp1; 1027 int bflags, len; 1028 struct socket *so; 1029 int error, error1; 1030 1031 bflags = 0; 1032 fp = NULL; 1033 sa = NULL; 1034 1035 error = linux_set_socket_flags(flags, &bflags); 1036 if (error != 0) 1037 return (error); 1038 1039 if (PTRIN(addr) == NULL) { 1040 len = 0; 1041 error = kern_accept4(td, s, NULL, NULL, bflags, NULL); 1042 } else { 1043 error = copyin(PTRIN(namelen), &len, sizeof(len)); 1044 if (error != 0) 1045 return (error); 1046 if (len < 0) 1047 return (EINVAL); 1048 error = kern_accept4(td, s, &sa, &len, bflags, &fp); 1049 } 1050 1051 /* 1052 * Translate errno values into ones used by Linux. 1053 */ 1054 if (error != 0) { 1055 /* 1056 * XXX. This is wrong, different sockaddr structures 1057 * have different sizes. 1058 */ 1059 switch (error) { 1060 case EFAULT: 1061 if (namelen != sizeof(struct sockaddr_in)) 1062 error = EINVAL; 1063 break; 1064 case EINVAL: 1065 error1 = getsock(td, s, &cap_accept_rights, &fp1); 1066 if (error1 != 0) { 1067 error = error1; 1068 break; 1069 } 1070 so = fp1->f_data; 1071 if (so->so_type == SOCK_DGRAM) 1072 error = EOPNOTSUPP; 1073 fdrop(fp1, td); 1074 break; 1075 } 1076 return (error); 1077 } 1078 1079 if (len != 0) { 1080 error = linux_copyout_sockaddr(sa, PTRIN(addr), len); 1081 if (error == 0) 1082 error = copyout(&len, PTRIN(namelen), 1083 sizeof(len)); 1084 if (error != 0) { 1085 fdclose(td, fp, td->td_retval[0]); 1086 td->td_retval[0] = 0; 1087 } 1088 } 1089 if (fp != NULL) 1090 fdrop(fp, td); 1091 free(sa, M_SONAME); 1092 return (error); 1093 } 1094 1095 int 1096 linux_accept(struct thread *td, struct linux_accept_args *args) 1097 { 1098 1099 return (linux_accept_common(td, args->s, args->addr, 1100 args->namelen, 0)); 1101 } 1102 1103 int 1104 linux_accept4(struct thread *td, struct linux_accept4_args *args) 1105 { 1106 1107 return (linux_accept_common(td, args->s, args->addr, 1108 args->namelen, args->flags)); 1109 } 1110 1111 int 1112 linux_getsockname(struct thread *td, struct linux_getsockname_args *args) 1113 { 1114 struct sockaddr *sa; 1115 int len, error; 1116 1117 error = copyin(PTRIN(args->namelen), &len, sizeof(len)); 1118 if (error != 0) 1119 return (error); 1120 1121 error = kern_getsockname(td, args->s, &sa, &len); 1122 if (error != 0) 1123 return (error); 1124 1125 if (len != 0) 1126 error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len); 1127 1128 free(sa, M_SONAME); 1129 if (error == 0) 1130 error = copyout(&len, PTRIN(args->namelen), sizeof(len)); 1131 return (error); 1132 } 1133 1134 int 1135 linux_getpeername(struct thread *td, struct linux_getpeername_args *args) 1136 { 1137 struct sockaddr *sa; 1138 int len, error; 1139 1140 error = copyin(PTRIN(args->namelen), &len, sizeof(len)); 1141 if (error != 0) 1142 return (error); 1143 if (len < 0) 1144 return (EINVAL); 1145 1146 error = kern_getpeername(td, args->s, &sa, &len); 1147 if (error != 0) 1148 return (error); 1149 1150 if (len != 0) 1151 error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len); 1152 1153 free(sa, M_SONAME); 1154 if (error == 0) 1155 error = copyout(&len, PTRIN(args->namelen), sizeof(len)); 1156 return (error); 1157 } 1158 1159 int 1160 linux_socketpair(struct thread *td, struct linux_socketpair_args *args) 1161 { 1162 int domain, error, sv[2], type; 1163 1164 domain = linux_to_bsd_domain(args->domain); 1165 if (domain != PF_LOCAL) 1166 return (EAFNOSUPPORT); 1167 type = args->type & LINUX_SOCK_TYPE_MASK; 1168 if (type < 0 || type > LINUX_SOCK_MAX) 1169 return (EINVAL); 1170 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK, 1171 &type); 1172 if (error != 0) 1173 return (error); 1174 if (args->protocol != 0 && args->protocol != PF_UNIX) { 1175 /* 1176 * Use of PF_UNIX as protocol argument is not right, 1177 * but Linux does it. 1178 * Do not map PF_UNIX as its Linux value is identical 1179 * to FreeBSD one. 1180 */ 1181 return (EPROTONOSUPPORT); 1182 } 1183 error = kern_socketpair(td, domain, type, 0, sv); 1184 if (error != 0) 1185 return (error); 1186 error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int)); 1187 if (error != 0) { 1188 (void)kern_close(td, sv[0]); 1189 (void)kern_close(td, sv[1]); 1190 } 1191 return (error); 1192 } 1193 1194 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1195 struct linux_send_args { 1196 register_t s; 1197 register_t msg; 1198 register_t len; 1199 register_t flags; 1200 }; 1201 1202 static int 1203 linux_send(struct thread *td, struct linux_send_args *args) 1204 { 1205 struct sendto_args /* { 1206 int s; 1207 caddr_t buf; 1208 int len; 1209 int flags; 1210 caddr_t to; 1211 int tolen; 1212 } */ bsd_args; 1213 struct file *fp; 1214 int error; 1215 1216 bsd_args.s = args->s; 1217 bsd_args.buf = (caddr_t)PTRIN(args->msg); 1218 bsd_args.len = args->len; 1219 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 1220 bsd_args.to = NULL; 1221 bsd_args.tolen = 0; 1222 error = sys_sendto(td, &bsd_args); 1223 if (error == ENOTCONN) { 1224 /* 1225 * Linux doesn't return ENOTCONN for non-blocking sockets. 1226 * Instead it returns the EAGAIN. 1227 */ 1228 error = getsock(td, args->s, &cap_send_rights, &fp); 1229 if (error == 0) { 1230 if (atomic_load_int(&fp->f_flag) & FNONBLOCK) 1231 error = EAGAIN; 1232 fdrop(fp, td); 1233 } 1234 } 1235 return (error); 1236 } 1237 1238 struct linux_recv_args { 1239 register_t s; 1240 register_t msg; 1241 register_t len; 1242 register_t flags; 1243 }; 1244 1245 static int 1246 linux_recv(struct thread *td, struct linux_recv_args *args) 1247 { 1248 struct recvfrom_args /* { 1249 int s; 1250 caddr_t buf; 1251 int len; 1252 int flags; 1253 struct sockaddr *from; 1254 socklen_t fromlenaddr; 1255 } */ bsd_args; 1256 1257 bsd_args.s = args->s; 1258 bsd_args.buf = (caddr_t)PTRIN(args->msg); 1259 bsd_args.len = args->len; 1260 bsd_args.flags = linux_to_bsd_msg_flags(args->flags); 1261 bsd_args.from = NULL; 1262 bsd_args.fromlenaddr = 0; 1263 return (sys_recvfrom(td, &bsd_args)); 1264 } 1265 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1266 1267 int 1268 linux_sendto(struct thread *td, struct linux_sendto_args *args) 1269 { 1270 struct msghdr msg; 1271 struct iovec aiov; 1272 struct socket *so; 1273 struct file *fp; 1274 int error; 1275 1276 if (linux_check_hdrincl(td, args->s) == 0) 1277 /* IP_HDRINCL set, tweak the packet before sending */ 1278 return (linux_sendto_hdrincl(td, args)); 1279 1280 bzero(&msg, sizeof(msg)); 1281 error = getsock(td, args->s, &cap_send_connect_rights, &fp); 1282 if (error != 0) 1283 return (error); 1284 so = fp->f_data; 1285 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0) { 1286 msg.msg_name = PTRIN(args->to); 1287 msg.msg_namelen = args->tolen; 1288 } 1289 msg.msg_iov = &aiov; 1290 msg.msg_iovlen = 1; 1291 aiov.iov_base = PTRIN(args->msg); 1292 aiov.iov_len = args->len; 1293 fdrop(fp, td); 1294 return (linux_sendit(td, args->s, &msg, args->flags, NULL, 1295 UIO_USERSPACE)); 1296 } 1297 1298 int 1299 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args) 1300 { 1301 struct sockaddr *sa; 1302 struct msghdr msg; 1303 struct iovec aiov; 1304 int error, fromlen; 1305 1306 if (PTRIN(args->fromlen) != NULL) { 1307 error = copyin(PTRIN(args->fromlen), &fromlen, 1308 sizeof(fromlen)); 1309 if (error != 0) 1310 return (error); 1311 if (fromlen < 0) 1312 return (EINVAL); 1313 fromlen = min(fromlen, SOCK_MAXADDRLEN); 1314 sa = malloc(fromlen, M_SONAME, M_WAITOK); 1315 } else { 1316 fromlen = 0; 1317 sa = NULL; 1318 } 1319 1320 msg.msg_name = sa; 1321 msg.msg_namelen = fromlen; 1322 msg.msg_iov = &aiov; 1323 msg.msg_iovlen = 1; 1324 aiov.iov_base = PTRIN(args->buf); 1325 aiov.iov_len = args->len; 1326 msg.msg_control = 0; 1327 msg.msg_flags = linux_to_bsd_msg_flags(args->flags); 1328 1329 error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL); 1330 if (error != 0) 1331 goto out; 1332 1333 /* 1334 * XXX. Seems that FreeBSD is different from Linux here. Linux 1335 * fill source address if underlying protocol provides it, while 1336 * FreeBSD fill it if underlying protocol is not connection-oriented. 1337 * So, kern_recvit() set msg.msg_namelen to 0 if protocol pr_flags 1338 * does not contains PR_ADDR flag. 1339 */ 1340 if (PTRIN(args->from) != NULL && msg.msg_namelen != 0) 1341 error = linux_copyout_sockaddr(sa, PTRIN(args->from), 1342 msg.msg_namelen); 1343 1344 if (error == 0 && PTRIN(args->fromlen) != NULL) 1345 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen), 1346 sizeof(msg.msg_namelen)); 1347 out: 1348 free(sa, M_SONAME); 1349 return (error); 1350 } 1351 1352 static int 1353 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1354 l_uint flags) 1355 { 1356 struct cmsghdr *cmsg; 1357 struct mbuf *control; 1358 struct msghdr msg; 1359 struct l_cmsghdr linux_cmsg; 1360 struct l_cmsghdr *ptr_cmsg; 1361 struct l_msghdr linux_msghdr; 1362 struct iovec *iov; 1363 socklen_t datalen; 1364 struct sockaddr *sa; 1365 struct socket *so; 1366 sa_family_t sa_family; 1367 struct file *fp; 1368 void *data; 1369 l_size_t len; 1370 l_size_t clen; 1371 int error; 1372 1373 error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr)); 1374 if (error != 0) 1375 return (error); 1376 1377 /* 1378 * Some Linux applications (ping) define a non-NULL control data 1379 * pointer, but a msg_controllen of 0, which is not allowed in the 1380 * FreeBSD system call interface. NULL the msg_control pointer in 1381 * order to handle this case. This should be checked, but allows the 1382 * Linux ping to work. 1383 */ 1384 if (PTRIN(linux_msghdr.msg_control) != NULL && 1385 linux_msghdr.msg_controllen == 0) 1386 linux_msghdr.msg_control = PTROUT(NULL); 1387 1388 error = linux_to_bsd_msghdr(&msg, &linux_msghdr); 1389 if (error != 0) 1390 return (error); 1391 1392 #ifdef COMPAT_LINUX32 1393 error = freebsd32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen, 1394 &iov, EMSGSIZE); 1395 #else 1396 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE); 1397 #endif 1398 if (error != 0) 1399 return (error); 1400 1401 control = NULL; 1402 1403 error = kern_getsockname(td, s, &sa, &datalen); 1404 if (error != 0) 1405 goto bad; 1406 sa_family = sa->sa_family; 1407 free(sa, M_SONAME); 1408 1409 if (flags & LINUX_MSG_OOB) { 1410 error = EOPNOTSUPP; 1411 if (sa_family == AF_UNIX) 1412 goto bad; 1413 1414 error = getsock(td, s, &cap_send_rights, &fp); 1415 if (error != 0) 1416 goto bad; 1417 so = fp->f_data; 1418 if (so->so_type != SOCK_STREAM) 1419 error = EOPNOTSUPP; 1420 fdrop(fp, td); 1421 if (error != 0) 1422 goto bad; 1423 } 1424 1425 if (linux_msghdr.msg_controllen >= sizeof(struct l_cmsghdr)) { 1426 error = ENOBUFS; 1427 control = m_get(M_WAITOK, MT_CONTROL); 1428 MCLGET(control, M_WAITOK); 1429 data = mtod(control, void *); 1430 datalen = 0; 1431 1432 ptr_cmsg = PTRIN(linux_msghdr.msg_control); 1433 clen = linux_msghdr.msg_controllen; 1434 do { 1435 error = copyin(ptr_cmsg, &linux_cmsg, 1436 sizeof(struct l_cmsghdr)); 1437 if (error != 0) 1438 goto bad; 1439 1440 error = EINVAL; 1441 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) || 1442 linux_cmsg.cmsg_len > clen) 1443 goto bad; 1444 1445 if (datalen + CMSG_HDRSZ > MCLBYTES) 1446 goto bad; 1447 1448 /* 1449 * Now we support only SCM_RIGHTS and SCM_CRED, 1450 * so return EINVAL in any other cmsg_type 1451 */ 1452 cmsg = data; 1453 cmsg->cmsg_type = 1454 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type); 1455 cmsg->cmsg_level = 1456 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level); 1457 if (cmsg->cmsg_type == -1 1458 || cmsg->cmsg_level != SOL_SOCKET) { 1459 linux_msg(curthread, 1460 "unsupported sendmsg cmsg level %d type %d", 1461 linux_cmsg.cmsg_level, linux_cmsg.cmsg_type); 1462 goto bad; 1463 } 1464 1465 /* 1466 * Some applications (e.g. pulseaudio) attempt to 1467 * send ancillary data even if the underlying protocol 1468 * doesn't support it which is not allowed in the 1469 * FreeBSD system call interface. 1470 */ 1471 if (sa_family != AF_UNIX) 1472 goto next; 1473 1474 if (cmsg->cmsg_type == SCM_CREDS) { 1475 len = sizeof(struct cmsgcred); 1476 if (datalen + CMSG_SPACE(len) > MCLBYTES) 1477 goto bad; 1478 1479 /* 1480 * The lower levels will fill in the structure 1481 */ 1482 memset(CMSG_DATA(data), 0, len); 1483 } else { 1484 len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ; 1485 if (datalen + CMSG_SPACE(len) < datalen || 1486 datalen + CMSG_SPACE(len) > MCLBYTES) 1487 goto bad; 1488 1489 error = copyin(LINUX_CMSG_DATA(ptr_cmsg), 1490 CMSG_DATA(data), len); 1491 if (error != 0) 1492 goto bad; 1493 } 1494 1495 cmsg->cmsg_len = CMSG_LEN(len); 1496 data = (char *)data + CMSG_SPACE(len); 1497 datalen += CMSG_SPACE(len); 1498 1499 next: 1500 if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len)) 1501 break; 1502 1503 clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len); 1504 ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg + 1505 LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len)); 1506 } while(clen >= sizeof(struct l_cmsghdr)); 1507 1508 control->m_len = datalen; 1509 if (datalen == 0) { 1510 m_freem(control); 1511 control = NULL; 1512 } 1513 } 1514 1515 msg.msg_iov = iov; 1516 msg.msg_flags = 0; 1517 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE); 1518 control = NULL; 1519 1520 bad: 1521 m_freem(control); 1522 free(iov, M_IOV); 1523 return (error); 1524 } 1525 1526 int 1527 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args) 1528 { 1529 1530 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg), 1531 args->flags)); 1532 } 1533 1534 int 1535 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args) 1536 { 1537 struct l_mmsghdr *msg; 1538 l_uint retval; 1539 int error, datagrams; 1540 1541 if (args->vlen > UIO_MAXIOV) 1542 args->vlen = UIO_MAXIOV; 1543 1544 msg = PTRIN(args->msg); 1545 datagrams = 0; 1546 while (datagrams < args->vlen) { 1547 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr, 1548 args->flags); 1549 if (error != 0) 1550 break; 1551 1552 retval = td->td_retval[0]; 1553 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1554 if (error != 0) 1555 break; 1556 ++msg; 1557 ++datagrams; 1558 } 1559 if (error == 0) 1560 td->td_retval[0] = datagrams; 1561 return (error); 1562 } 1563 1564 static int 1565 recvmsg_scm_rights(struct thread *td, l_uint flags, socklen_t *datalen, 1566 void **data, void **udata) 1567 { 1568 int i, fd, fds, *fdp; 1569 1570 if (flags & LINUX_MSG_CMSG_CLOEXEC) { 1571 fds = *datalen / sizeof(int); 1572 fdp = *data; 1573 for (i = 0; i < fds; i++) { 1574 fd = *fdp++; 1575 (void)kern_fcntl(td, fd, F_SETFD, FD_CLOEXEC); 1576 } 1577 } 1578 return (0); 1579 } 1580 1581 1582 static int 1583 recvmsg_scm_creds(socklen_t *datalen, void **data, void **udata) 1584 { 1585 struct cmsgcred *cmcred; 1586 struct l_ucred lu; 1587 1588 cmcred = *data; 1589 lu.pid = cmcred->cmcred_pid; 1590 lu.uid = cmcred->cmcred_uid; 1591 lu.gid = cmcred->cmcred_gid; 1592 memmove(*data, &lu, sizeof(lu)); 1593 *datalen = sizeof(lu); 1594 return (0); 1595 } 1596 _Static_assert(sizeof(struct cmsgcred) >= sizeof(struct l_ucred), 1597 "scm_creds sizeof l_ucred"); 1598 1599 static int 1600 recvmsg_scm_creds2(socklen_t *datalen, void **data, void **udata) 1601 { 1602 struct sockcred2 *scred; 1603 struct l_ucred lu; 1604 1605 scred = *data; 1606 lu.pid = scred->sc_pid; 1607 lu.uid = scred->sc_uid; 1608 lu.gid = scred->sc_gid; 1609 memmove(*data, &lu, sizeof(lu)); 1610 *datalen = sizeof(lu); 1611 return (0); 1612 } 1613 _Static_assert(sizeof(struct sockcred2) >= sizeof(struct l_ucred), 1614 "scm_creds2 sizeof l_ucred"); 1615 1616 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1617 static int 1618 recvmsg_scm_timestamp(l_int msg_type, socklen_t *datalen, void **data, 1619 void **udata) 1620 { 1621 l_sock_timeval ltv64; 1622 l_timeval ltv; 1623 struct timeval *tv; 1624 socklen_t len; 1625 void *buf; 1626 1627 if (*datalen != sizeof(struct timeval)) 1628 return (EMSGSIZE); 1629 1630 tv = *data; 1631 #if defined(COMPAT_LINUX32) 1632 if (msg_type == LINUX_SCM_TIMESTAMPO && 1633 (tv->tv_sec > INT_MAX || tv->tv_sec < INT_MIN)) 1634 return (EOVERFLOW); 1635 #endif 1636 if (msg_type == LINUX_SCM_TIMESTAMPN) 1637 len = sizeof(ltv64); 1638 else 1639 len = sizeof(ltv); 1640 1641 buf = malloc(len, M_LINUX, M_WAITOK); 1642 if (msg_type == LINUX_SCM_TIMESTAMPN) { 1643 ltv64.tv_sec = tv->tv_sec; 1644 ltv64.tv_usec = tv->tv_usec; 1645 memmove(buf, <v64, len); 1646 } else { 1647 ltv.tv_sec = tv->tv_sec; 1648 ltv.tv_usec = tv->tv_usec; 1649 memmove(buf, <v, len); 1650 } 1651 *data = *udata = buf; 1652 *datalen = len; 1653 return (0); 1654 } 1655 #else 1656 _Static_assert(sizeof(struct timeval) == sizeof(l_timeval), 1657 "scm_timestamp sizeof l_timeval"); 1658 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1659 1660 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1661 static int 1662 recvmsg_scm_timestampns(l_int msg_type, socklen_t *datalen, void **data, 1663 void **udata) 1664 { 1665 struct l_timespec64 ts64; 1666 struct l_timespec ts32; 1667 struct timespec ts; 1668 socklen_t len; 1669 void *buf; 1670 1671 if (msg_type == LINUX_SCM_TIMESTAMPNSO) 1672 len = sizeof(ts32); 1673 else 1674 len = sizeof(ts64); 1675 1676 buf = malloc(len, M_LINUX, M_WAITOK); 1677 bintime2timespec(*data, &ts); 1678 if (msg_type == LINUX_SCM_TIMESTAMPNSO) { 1679 ts32.tv_sec = ts.tv_sec; 1680 ts32.tv_nsec = ts.tv_nsec; 1681 memmove(buf, &ts32, len); 1682 } else { 1683 ts64.tv_sec = ts.tv_sec; 1684 ts64.tv_nsec = ts.tv_nsec; 1685 memmove(buf, &ts64, len); 1686 } 1687 *data = *udata = buf; 1688 *datalen = len; 1689 return (0); 1690 } 1691 #else 1692 static int 1693 recvmsg_scm_timestampns(l_int msg_type, socklen_t *datalen, void **data, 1694 void **udata) 1695 { 1696 struct timespec ts; 1697 1698 bintime2timespec(*data, &ts); 1699 memmove(*data, &ts, sizeof(struct timespec)); 1700 *datalen = sizeof(struct timespec); 1701 return (0); 1702 } 1703 _Static_assert(sizeof(struct bintime) >= sizeof(struct timespec), 1704 "scm_timestampns sizeof timespec"); 1705 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1706 1707 static int 1708 recvmsg_scm_sol_socket(struct thread *td, l_int msg_type, l_int lmsg_type, 1709 l_uint flags, socklen_t *datalen, void **data, void **udata) 1710 { 1711 int error; 1712 1713 error = 0; 1714 switch (msg_type) { 1715 case SCM_RIGHTS: 1716 error = recvmsg_scm_rights(td, flags, datalen, 1717 data, udata); 1718 break; 1719 case SCM_CREDS: 1720 error = recvmsg_scm_creds(datalen, data, udata); 1721 break; 1722 case SCM_CREDS2: 1723 error = recvmsg_scm_creds2(datalen, data, udata); 1724 break; 1725 case SCM_TIMESTAMP: 1726 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1727 error = recvmsg_scm_timestamp(lmsg_type, datalen, 1728 data, udata); 1729 #endif 1730 break; 1731 case SCM_BINTIME: 1732 error = recvmsg_scm_timestampns(lmsg_type, datalen, 1733 data, udata); 1734 break; 1735 } 1736 1737 return (error); 1738 } 1739 1740 static int 1741 recvmsg_scm_ip_origdstaddr(socklen_t *datalen, void **data, void **udata) 1742 { 1743 struct l_sockaddr *lsa; 1744 int error; 1745 1746 error = bsd_to_linux_sockaddr(*data, &lsa, *datalen); 1747 if (error == 0) { 1748 *data = *udata = lsa; 1749 *datalen = sizeof(*lsa); 1750 } 1751 return (error); 1752 } 1753 1754 static int 1755 recvmsg_scm_ipproto_ip(l_int msg_type, l_int lmsg_type, socklen_t *datalen, 1756 void **data, void **udata) 1757 { 1758 int error; 1759 1760 error = 0; 1761 switch (msg_type) { 1762 case IP_ORIGDSTADDR: 1763 error = recvmsg_scm_ip_origdstaddr(datalen, data, 1764 udata); 1765 break; 1766 } 1767 1768 return (error); 1769 } 1770 1771 static int 1772 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr, 1773 l_uint flags, struct msghdr *msg) 1774 { 1775 struct proc *p = td->td_proc; 1776 struct cmsghdr *cm; 1777 struct l_cmsghdr *lcm = NULL; 1778 socklen_t datalen, maxlen, outlen; 1779 struct l_msghdr l_msghdr; 1780 struct iovec *iov, *uiov; 1781 struct mbuf *m, *control = NULL; 1782 struct mbuf **controlp; 1783 struct sockaddr *sa; 1784 caddr_t outbuf; 1785 void *data, *udata; 1786 int error, skiped; 1787 1788 error = copyin(msghdr, &l_msghdr, sizeof(l_msghdr)); 1789 if (error != 0) 1790 return (error); 1791 1792 /* 1793 * Pass user-supplied recvmsg() flags in msg_flags field, 1794 * following sys_recvmsg() convention. 1795 */ 1796 l_msghdr.msg_flags = flags; 1797 1798 error = linux_to_bsd_msghdr(msg, &l_msghdr); 1799 if (error != 0) 1800 return (error); 1801 1802 #ifdef COMPAT_LINUX32 1803 error = freebsd32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen, 1804 &iov, EMSGSIZE); 1805 #else 1806 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE); 1807 #endif 1808 if (error != 0) 1809 return (error); 1810 1811 if (msg->msg_name != NULL && msg->msg_namelen > 0) { 1812 msg->msg_namelen = min(msg->msg_namelen, SOCK_MAXADDRLEN); 1813 sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK); 1814 msg->msg_name = sa; 1815 } else { 1816 sa = NULL; 1817 msg->msg_name = NULL; 1818 } 1819 1820 uiov = msg->msg_iov; 1821 msg->msg_iov = iov; 1822 controlp = (msg->msg_control != NULL) ? &control : NULL; 1823 error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp); 1824 msg->msg_iov = uiov; 1825 if (error != 0) 1826 goto bad; 1827 1828 /* 1829 * Note that kern_recvit() updates msg->msg_namelen. 1830 */ 1831 if (msg->msg_name != NULL && msg->msg_namelen > 0) { 1832 msg->msg_name = PTRIN(l_msghdr.msg_name); 1833 error = linux_copyout_sockaddr(sa, msg->msg_name, 1834 msg->msg_namelen); 1835 if (error != 0) 1836 goto bad; 1837 } 1838 1839 error = bsd_to_linux_msghdr(msg, &l_msghdr); 1840 if (error != 0) 1841 goto bad; 1842 1843 skiped = outlen = 0; 1844 maxlen = l_msghdr.msg_controllen; 1845 if (control == NULL) 1846 goto out; 1847 1848 lcm = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO); 1849 msg->msg_control = mtod(control, struct cmsghdr *); 1850 msg->msg_controllen = control->m_len; 1851 outbuf = PTRIN(l_msghdr.msg_control); 1852 for (m = control; m != NULL; m = m->m_next) { 1853 cm = mtod(m, struct cmsghdr *); 1854 lcm->cmsg_type = bsd_to_linux_cmsg_type(p, cm->cmsg_type, 1855 cm->cmsg_level); 1856 lcm->cmsg_level = bsd_to_linux_sockopt_level(cm->cmsg_level); 1857 1858 if (lcm->cmsg_type == -1 || 1859 cm->cmsg_level == -1) { 1860 LINUX_RATELIMIT_MSG_OPT2( 1861 "unsupported recvmsg cmsg level %d type %d", 1862 cm->cmsg_level, cm->cmsg_type); 1863 /* Skip unsupported messages */ 1864 skiped++; 1865 continue; 1866 } 1867 data = CMSG_DATA(cm); 1868 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; 1869 udata = NULL; 1870 error = 0; 1871 1872 switch (cm->cmsg_level) { 1873 case IPPROTO_IP: 1874 error = recvmsg_scm_ipproto_ip(cm->cmsg_type, 1875 lcm->cmsg_type, &datalen, &data, &udata); 1876 break; 1877 case SOL_SOCKET: 1878 error = recvmsg_scm_sol_socket(td, cm->cmsg_type, 1879 lcm->cmsg_type, flags, &datalen, &data, &udata); 1880 break; 1881 } 1882 1883 /* The recvmsg_scm_ is responsible to free udata on error. */ 1884 if (error != 0) 1885 goto bad; 1886 1887 if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) { 1888 if (outlen == 0) { 1889 error = EMSGSIZE; 1890 goto err; 1891 } else { 1892 l_msghdr.msg_flags |= LINUX_MSG_CTRUNC; 1893 m_dispose_extcontrolm(control); 1894 free(udata, M_LINUX); 1895 goto out; 1896 } 1897 } 1898 1899 lcm->cmsg_len = LINUX_CMSG_LEN(datalen); 1900 error = copyout(lcm, outbuf, L_CMSG_HDRSZ); 1901 if (error == 0) { 1902 error = copyout(data, LINUX_CMSG_DATA(outbuf), datalen); 1903 if (error == 0) { 1904 outbuf += LINUX_CMSG_SPACE(datalen); 1905 outlen += LINUX_CMSG_SPACE(datalen); 1906 } 1907 } 1908 err: 1909 free(udata, M_LINUX); 1910 if (error != 0) 1911 goto bad; 1912 } 1913 if (outlen == 0 && skiped > 0) { 1914 error = EINVAL; 1915 goto bad; 1916 } 1917 1918 out: 1919 l_msghdr.msg_controllen = outlen; 1920 error = copyout(&l_msghdr, msghdr, sizeof(l_msghdr)); 1921 1922 bad: 1923 if (control != NULL) { 1924 if (error != 0) 1925 m_dispose_extcontrolm(control); 1926 m_freem(control); 1927 } 1928 free(iov, M_IOV); 1929 free(lcm, M_LINUX); 1930 free(sa, M_SONAME); 1931 1932 return (error); 1933 } 1934 1935 int 1936 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args) 1937 { 1938 struct msghdr bsd_msg; 1939 struct file *fp; 1940 int error; 1941 1942 error = getsock(td, args->s, &cap_recv_rights, &fp); 1943 if (error != 0) 1944 return (error); 1945 fdrop(fp, td); 1946 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg), 1947 args->flags, &bsd_msg)); 1948 } 1949 1950 static int 1951 linux_recvmmsg_common(struct thread *td, l_int s, struct l_mmsghdr *msg, 1952 l_uint vlen, l_uint flags, struct timespec *tts) 1953 { 1954 struct msghdr bsd_msg; 1955 struct timespec ts; 1956 struct file *fp; 1957 l_uint retval; 1958 int error, datagrams; 1959 1960 error = getsock(td, s, &cap_recv_rights, &fp); 1961 if (error != 0) 1962 return (error); 1963 datagrams = 0; 1964 while (datagrams < vlen) { 1965 error = linux_recvmsg_common(td, s, &msg->msg_hdr, 1966 flags & ~LINUX_MSG_WAITFORONE, &bsd_msg); 1967 if (error != 0) 1968 break; 1969 1970 retval = td->td_retval[0]; 1971 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len)); 1972 if (error != 0) 1973 break; 1974 ++msg; 1975 ++datagrams; 1976 1977 /* 1978 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet. 1979 */ 1980 if (flags & LINUX_MSG_WAITFORONE) 1981 flags |= LINUX_MSG_DONTWAIT; 1982 1983 /* 1984 * See BUGS section of recvmmsg(2). 1985 */ 1986 if (tts) { 1987 getnanotime(&ts); 1988 timespecsub(&ts, tts, &ts); 1989 if (!timespecisset(&ts) || ts.tv_sec > 0) 1990 break; 1991 } 1992 /* Out of band data, return right away. */ 1993 if (bsd_msg.msg_flags & MSG_OOB) 1994 break; 1995 } 1996 if (error == 0) 1997 td->td_retval[0] = datagrams; 1998 fdrop(fp, td); 1999 return (error); 2000 } 2001 2002 int 2003 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args) 2004 { 2005 struct timespec ts, tts, *ptts; 2006 int error; 2007 2008 if (args->timeout) { 2009 error = linux_get_timespec(&ts, args->timeout); 2010 if (error != 0) 2011 return (error); 2012 getnanotime(&tts); 2013 timespecadd(&tts, &ts, &tts); 2014 ptts = &tts; 2015 } 2016 else ptts = NULL; 2017 2018 return (linux_recvmmsg_common(td, args->s, PTRIN(args->msg), 2019 args->vlen, args->flags, ptts)); 2020 } 2021 2022 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 2023 int 2024 linux_recvmmsg_time64(struct thread *td, struct linux_recvmmsg_time64_args *args) 2025 { 2026 struct timespec ts, tts, *ptts; 2027 int error; 2028 2029 if (args->timeout) { 2030 error = linux_get_timespec64(&ts, args->timeout); 2031 if (error != 0) 2032 return (error); 2033 getnanotime(&tts); 2034 timespecadd(&tts, &ts, &tts); 2035 ptts = &tts; 2036 } 2037 else ptts = NULL; 2038 2039 return (linux_recvmmsg_common(td, args->s, PTRIN(args->msg), 2040 args->vlen, args->flags, ptts)); 2041 } 2042 #endif 2043 2044 int 2045 linux_shutdown(struct thread *td, struct linux_shutdown_args *args) 2046 { 2047 2048 return (kern_shutdown(td, args->s, args->how)); 2049 } 2050 2051 int 2052 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args) 2053 { 2054 struct proc *p = td->td_proc; 2055 struct linux_pemuldata *pem; 2056 l_timeval linux_tv; 2057 struct sockaddr *sa; 2058 struct timeval tv; 2059 socklen_t len; 2060 int error, level, name, val; 2061 2062 level = linux_to_bsd_sockopt_level(args->level); 2063 switch (level) { 2064 case SOL_SOCKET: 2065 name = linux_to_bsd_so_sockopt(args->optname); 2066 switch (name) { 2067 case LOCAL_CREDS_PERSISTENT: 2068 level = SOL_LOCAL; 2069 break; 2070 case SO_RCVTIMEO: 2071 /* FALLTHROUGH */ 2072 case SO_SNDTIMEO: 2073 error = copyin(PTRIN(args->optval), &linux_tv, 2074 sizeof(linux_tv)); 2075 if (error != 0) 2076 return (error); 2077 tv.tv_sec = linux_tv.tv_sec; 2078 tv.tv_usec = linux_tv.tv_usec; 2079 return (kern_setsockopt(td, args->s, level, 2080 name, &tv, UIO_SYSSPACE, sizeof(tv))); 2081 /* NOTREACHED */ 2082 case SO_TIMESTAMP: 2083 /* overwrite SO_BINTIME */ 2084 val = 0; 2085 error = kern_setsockopt(td, args->s, level, 2086 SO_BINTIME, &val, UIO_SYSSPACE, sizeof(val)); 2087 if (error != 0) 2088 return (error); 2089 pem = pem_find(p); 2090 pem->so_timestamp = args->optname; 2091 break; 2092 case SO_BINTIME: 2093 /* overwrite SO_TIMESTAMP */ 2094 val = 0; 2095 error = kern_setsockopt(td, args->s, level, 2096 SO_TIMESTAMP, &val, UIO_SYSSPACE, sizeof(val)); 2097 if (error != 0) 2098 return (error); 2099 pem = pem_find(p); 2100 pem->so_timestampns = args->optname; 2101 break; 2102 default: 2103 break; 2104 } 2105 break; 2106 case IPPROTO_IP: 2107 if (args->optname == LINUX_IP_RECVERR && 2108 linux_ignore_ip_recverr) { 2109 /* 2110 * XXX: This is a hack to unbreak DNS resolution 2111 * with glibc 2.30 and above. 2112 */ 2113 return (0); 2114 } 2115 name = linux_to_bsd_ip_sockopt(args->optname); 2116 break; 2117 case IPPROTO_IPV6: 2118 if (args->optname == LINUX_IPV6_RECVERR && 2119 linux_ignore_ip_recverr) { 2120 /* 2121 * XXX: This is a hack to unbreak DNS resolution 2122 * with glibc 2.30 and above. 2123 */ 2124 return (0); 2125 } 2126 name = linux_to_bsd_ip6_sockopt(args->optname); 2127 break; 2128 case IPPROTO_TCP: 2129 name = linux_to_bsd_tcp_sockopt(args->optname); 2130 break; 2131 case SOL_NETLINK: 2132 level = SOL_SOCKET; 2133 name = args->optname; 2134 break; 2135 default: 2136 name = -1; 2137 break; 2138 } 2139 if (name < 0) { 2140 if (name == -1) 2141 linux_msg(curthread, 2142 "unsupported setsockopt level %d optname %d", 2143 args->level, args->optname); 2144 return (ENOPROTOOPT); 2145 } 2146 2147 if (name == IPV6_NEXTHOP) { 2148 len = args->optlen; 2149 error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len); 2150 if (error != 0) 2151 return (error); 2152 2153 error = kern_setsockopt(td, args->s, level, 2154 name, sa, UIO_SYSSPACE, len); 2155 free(sa, M_SONAME); 2156 } else { 2157 error = kern_setsockopt(td, args->s, level, 2158 name, PTRIN(args->optval), UIO_USERSPACE, args->optlen); 2159 } 2160 2161 return (error); 2162 } 2163 2164 static int 2165 linux_sockopt_copyout(struct thread *td, void *val, socklen_t len, 2166 struct linux_getsockopt_args *args) 2167 { 2168 int error; 2169 2170 error = copyout(val, PTRIN(args->optval), len); 2171 if (error == 0) 2172 error = copyout(&len, PTRIN(args->optlen), sizeof(len)); 2173 return (error); 2174 } 2175 2176 static int 2177 linux_getsockopt_so_peergroups(struct thread *td, 2178 struct linux_getsockopt_args *args) 2179 { 2180 struct xucred xu; 2181 socklen_t xulen, len; 2182 int error, i; 2183 2184 xulen = sizeof(xu); 2185 error = kern_getsockopt(td, args->s, 0, 2186 LOCAL_PEERCRED, &xu, UIO_SYSSPACE, &xulen); 2187 if (error != 0) 2188 return (error); 2189 2190 len = xu.cr_ngroups * sizeof(l_gid_t); 2191 if (args->optlen < len) { 2192 error = copyout(&len, PTRIN(args->optlen), sizeof(len)); 2193 if (error == 0) 2194 error = ERANGE; 2195 return (error); 2196 } 2197 2198 /* 2199 * "- 1" to skip the primary group. 2200 */ 2201 for (i = 0; i < xu.cr_ngroups - 1; i++) { 2202 error = copyout(xu.cr_groups + i + 1, 2203 (void *)(args->optval + i * sizeof(l_gid_t)), 2204 sizeof(l_gid_t)); 2205 if (error != 0) 2206 return (error); 2207 } 2208 2209 error = copyout(&len, PTRIN(args->optlen), sizeof(len)); 2210 return (error); 2211 } 2212 2213 static int 2214 linux_getsockopt_so_peersec(struct thread *td, 2215 struct linux_getsockopt_args *args) 2216 { 2217 socklen_t len; 2218 int error; 2219 2220 len = sizeof(SECURITY_CONTEXT_STRING); 2221 if (args->optlen < len) { 2222 error = copyout(&len, PTRIN(args->optlen), sizeof(len)); 2223 if (error == 0) 2224 error = ERANGE; 2225 return (error); 2226 } 2227 2228 return (linux_sockopt_copyout(td, SECURITY_CONTEXT_STRING, 2229 len, args)); 2230 } 2231 2232 static int 2233 linux_getsockopt_so_linger(struct thread *td, 2234 struct linux_getsockopt_args *args) 2235 { 2236 struct linger ling; 2237 socklen_t len; 2238 int error; 2239 2240 len = sizeof(ling); 2241 error = kern_getsockopt(td, args->s, SOL_SOCKET, 2242 SO_LINGER, &ling, UIO_SYSSPACE, &len); 2243 if (error != 0) 2244 return (error); 2245 ling.l_onoff = ((ling.l_onoff & SO_LINGER) != 0); 2246 return (linux_sockopt_copyout(td, &ling, len, args)); 2247 } 2248 2249 int 2250 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args) 2251 { 2252 l_timeval linux_tv; 2253 struct timeval tv; 2254 socklen_t tv_len, xulen, len; 2255 struct sockaddr *sa; 2256 struct xucred xu; 2257 struct l_ucred lxu; 2258 int error, level, name, newval; 2259 2260 level = linux_to_bsd_sockopt_level(args->level); 2261 switch (level) { 2262 case SOL_SOCKET: 2263 switch (args->optname) { 2264 case LINUX_SO_PEERGROUPS: 2265 return (linux_getsockopt_so_peergroups(td, args)); 2266 case LINUX_SO_PEERSEC: 2267 return (linux_getsockopt_so_peersec(td, args)); 2268 default: 2269 break; 2270 } 2271 2272 name = linux_to_bsd_so_sockopt(args->optname); 2273 switch (name) { 2274 case LOCAL_CREDS_PERSISTENT: 2275 level = SOL_LOCAL; 2276 break; 2277 case SO_RCVTIMEO: 2278 /* FALLTHROUGH */ 2279 case SO_SNDTIMEO: 2280 tv_len = sizeof(tv); 2281 error = kern_getsockopt(td, args->s, level, 2282 name, &tv, UIO_SYSSPACE, &tv_len); 2283 if (error != 0) 2284 return (error); 2285 linux_tv.tv_sec = tv.tv_sec; 2286 linux_tv.tv_usec = tv.tv_usec; 2287 return (linux_sockopt_copyout(td, &linux_tv, 2288 sizeof(linux_tv), args)); 2289 /* NOTREACHED */ 2290 case LOCAL_PEERCRED: 2291 if (args->optlen < sizeof(lxu)) 2292 return (EINVAL); 2293 /* 2294 * LOCAL_PEERCRED is not served at the SOL_SOCKET level, 2295 * but by the Unix socket's level 0. 2296 */ 2297 level = 0; 2298 xulen = sizeof(xu); 2299 error = kern_getsockopt(td, args->s, level, 2300 name, &xu, UIO_SYSSPACE, &xulen); 2301 if (error != 0) 2302 return (error); 2303 lxu.pid = xu.cr_pid; 2304 lxu.uid = xu.cr_uid; 2305 lxu.gid = xu.cr_gid; 2306 return (linux_sockopt_copyout(td, &lxu, 2307 sizeof(lxu), args)); 2308 /* NOTREACHED */ 2309 case SO_ERROR: 2310 len = sizeof(newval); 2311 error = kern_getsockopt(td, args->s, level, 2312 name, &newval, UIO_SYSSPACE, &len); 2313 if (error != 0) 2314 return (error); 2315 newval = -bsd_to_linux_errno(newval); 2316 return (linux_sockopt_copyout(td, &newval, 2317 len, args)); 2318 /* NOTREACHED */ 2319 case SO_DOMAIN: 2320 len = sizeof(newval); 2321 error = kern_getsockopt(td, args->s, level, 2322 name, &newval, UIO_SYSSPACE, &len); 2323 if (error != 0) 2324 return (error); 2325 newval = bsd_to_linux_domain(newval); 2326 if (newval == -1) 2327 return (ENOPROTOOPT); 2328 return (linux_sockopt_copyout(td, &newval, 2329 len, args)); 2330 /* NOTREACHED */ 2331 case SO_LINGER: 2332 return (linux_getsockopt_so_linger(td, args)); 2333 /* NOTREACHED */ 2334 default: 2335 break; 2336 } 2337 break; 2338 case IPPROTO_IP: 2339 name = linux_to_bsd_ip_sockopt(args->optname); 2340 break; 2341 case IPPROTO_IPV6: 2342 name = linux_to_bsd_ip6_sockopt(args->optname); 2343 break; 2344 case IPPROTO_TCP: 2345 name = linux_to_bsd_tcp_sockopt(args->optname); 2346 break; 2347 default: 2348 name = -1; 2349 break; 2350 } 2351 if (name < 0) { 2352 if (name == -1) 2353 linux_msg(curthread, 2354 "unsupported getsockopt level %d optname %d", 2355 args->level, args->optname); 2356 return (EINVAL); 2357 } 2358 2359 if (name == IPV6_NEXTHOP) { 2360 error = copyin(PTRIN(args->optlen), &len, sizeof(len)); 2361 if (error != 0) 2362 return (error); 2363 sa = malloc(len, M_SONAME, M_WAITOK); 2364 2365 error = kern_getsockopt(td, args->s, level, 2366 name, sa, UIO_SYSSPACE, &len); 2367 if (error != 0) 2368 goto out; 2369 2370 error = linux_copyout_sockaddr(sa, PTRIN(args->optval), len); 2371 if (error == 0) 2372 error = copyout(&len, PTRIN(args->optlen), 2373 sizeof(len)); 2374 out: 2375 free(sa, M_SONAME); 2376 } else { 2377 if (args->optval) { 2378 error = copyin(PTRIN(args->optlen), &len, sizeof(len)); 2379 if (error != 0) 2380 return (error); 2381 } 2382 error = kern_getsockopt(td, args->s, level, 2383 name, PTRIN(args->optval), UIO_USERSPACE, &len); 2384 if (error == 0) 2385 error = copyout(&len, PTRIN(args->optlen), 2386 sizeof(len)); 2387 } 2388 2389 return (error); 2390 } 2391 2392 /* 2393 * Based on sendfile_getsock from kern_sendfile.c 2394 * Determines whether an fd is a stream socket that can be used 2395 * with FreeBSD sendfile. 2396 */ 2397 static bool 2398 is_sendfile(struct file *fp, struct file *ofp) 2399 { 2400 struct socket *so; 2401 2402 /* 2403 * FreeBSD sendfile() system call sends a regular file or 2404 * shared memory object out a stream socket. 2405 */ 2406 if ((fp->f_type != DTYPE_SHM && fp->f_type != DTYPE_VNODE) || 2407 (fp->f_type == DTYPE_VNODE && 2408 (fp->f_vnode == NULL || fp->f_vnode->v_type != VREG))) 2409 return (false); 2410 /* 2411 * The socket must be a stream socket and connected. 2412 */ 2413 if (ofp->f_type != DTYPE_SOCKET) 2414 return (false); 2415 so = ofp->f_data; 2416 if (so->so_type != SOCK_STREAM) 2417 return (false); 2418 /* 2419 * SCTP one-to-one style sockets currently don't work with 2420 * sendfile(). 2421 */ 2422 if (so->so_proto->pr_protocol == IPPROTO_SCTP) 2423 return (false); 2424 return (!SOLISTENING(so)); 2425 } 2426 2427 static bool 2428 is_regular_file(struct file *fp) 2429 { 2430 2431 return (fp->f_type == DTYPE_VNODE && fp->f_vnode != NULL && 2432 fp->f_vnode->v_type == VREG); 2433 } 2434 2435 static int 2436 sendfile_fallback(struct thread *td, struct file *fp, l_int out, 2437 off_t *offset, l_size_t count, off_t *sbytes) 2438 { 2439 off_t current_offset, out_offset, to_send; 2440 l_size_t bytes_sent, n_read; 2441 struct file *ofp; 2442 struct iovec aiov; 2443 struct uio auio; 2444 bool seekable; 2445 size_t bufsz; 2446 void *buf; 2447 int flags, error; 2448 2449 if (offset == NULL) { 2450 if ((error = fo_seek(fp, 0, SEEK_CUR, td)) != 0) 2451 return (error); 2452 current_offset = td->td_uretoff.tdu_off; 2453 } else { 2454 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) 2455 return (ESPIPE); 2456 current_offset = *offset; 2457 } 2458 error = fget_write(td, out, &cap_pwrite_rights, &ofp); 2459 if (error != 0) 2460 return (error); 2461 seekable = (ofp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0; 2462 if (seekable) { 2463 if ((error = fo_seek(ofp, 0, SEEK_CUR, td)) != 0) 2464 goto drop; 2465 out_offset = td->td_uretoff.tdu_off; 2466 } else 2467 out_offset = 0; 2468 2469 flags = FOF_OFFSET | FOF_NOUPDATE; 2470 bufsz = min(count, MAXPHYS); 2471 buf = malloc(bufsz, M_LINUX, M_WAITOK); 2472 bytes_sent = 0; 2473 while (bytes_sent < count) { 2474 to_send = min(count - bytes_sent, bufsz); 2475 aiov.iov_base = buf; 2476 aiov.iov_len = bufsz; 2477 auio.uio_iov = &aiov; 2478 auio.uio_iovcnt = 1; 2479 auio.uio_segflg = UIO_SYSSPACE; 2480 auio.uio_td = td; 2481 auio.uio_rw = UIO_READ; 2482 auio.uio_offset = current_offset; 2483 auio.uio_resid = to_send; 2484 error = fo_read(fp, &auio, fp->f_cred, flags, td); 2485 if (error != 0) 2486 break; 2487 n_read = to_send - auio.uio_resid; 2488 if (n_read == 0) 2489 break; 2490 aiov.iov_base = buf; 2491 aiov.iov_len = bufsz; 2492 auio.uio_iov = &aiov; 2493 auio.uio_iovcnt = 1; 2494 auio.uio_segflg = UIO_SYSSPACE; 2495 auio.uio_td = td; 2496 auio.uio_rw = UIO_WRITE; 2497 auio.uio_offset = (seekable) ? out_offset : 0; 2498 auio.uio_resid = n_read; 2499 error = fo_write(ofp, &auio, ofp->f_cred, flags, td); 2500 if (error != 0) 2501 break; 2502 bytes_sent += n_read; 2503 current_offset += n_read; 2504 out_offset += n_read; 2505 } 2506 free(buf, M_LINUX); 2507 2508 if (error == 0) { 2509 *sbytes = bytes_sent; 2510 if (offset != NULL) 2511 *offset = current_offset; 2512 else 2513 error = fo_seek(fp, current_offset, SEEK_SET, td); 2514 } 2515 if (error == 0 && seekable) 2516 error = fo_seek(ofp, out_offset, SEEK_SET, td); 2517 2518 drop: 2519 fdrop(ofp, td); 2520 return (error); 2521 } 2522 2523 static int 2524 sendfile_sendfile(struct thread *td, struct file *fp, l_int out, 2525 off_t *offset, l_size_t count, off_t *sbytes) 2526 { 2527 off_t current_offset; 2528 int error; 2529 2530 if (offset == NULL) { 2531 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) 2532 return (ESPIPE); 2533 if ((error = fo_seek(fp, 0, SEEK_CUR, td)) != 0) 2534 return (error); 2535 current_offset = td->td_uretoff.tdu_off; 2536 } else 2537 current_offset = *offset; 2538 error = fo_sendfile(fp, out, NULL, NULL, current_offset, count, 2539 sbytes, 0, td); 2540 if (error == 0) { 2541 current_offset += *sbytes; 2542 if (offset != NULL) 2543 *offset = current_offset; 2544 else 2545 error = fo_seek(fp, current_offset, SEEK_SET, td); 2546 } 2547 return (error); 2548 } 2549 2550 static int 2551 linux_sendfile_common(struct thread *td, l_int out, l_int in, 2552 off_t *offset, l_size_t count) 2553 { 2554 struct file *fp, *ofp; 2555 off_t sbytes; 2556 int error; 2557 2558 /* Linux cannot have 0 count. */ 2559 if (count <= 0 || (offset != NULL && *offset < 0)) 2560 return (EINVAL); 2561 2562 AUDIT_ARG_FD(in); 2563 error = fget_read(td, in, &cap_pread_rights, &fp); 2564 if (error != 0) 2565 return (error); 2566 if ((fp->f_type != DTYPE_SHM && fp->f_type != DTYPE_VNODE) || 2567 (fp->f_type == DTYPE_VNODE && 2568 (fp->f_vnode == NULL || fp->f_vnode->v_type != VREG))) { 2569 error = EINVAL; 2570 goto drop; 2571 } 2572 error = fget_unlocked(td, out, &cap_no_rights, &ofp); 2573 if (error != 0) 2574 goto drop; 2575 2576 if (is_regular_file(fp) && is_regular_file(ofp)) { 2577 error = kern_copy_file_range(td, in, offset, out, NULL, count, 2578 0); 2579 } else { 2580 sbytes = 0; 2581 if (is_sendfile(fp, ofp)) 2582 error = sendfile_sendfile(td, fp, out, offset, count, 2583 &sbytes); 2584 else 2585 error = sendfile_fallback(td, fp, out, offset, count, 2586 &sbytes); 2587 if (error == ENOBUFS && (ofp->f_flag & FNONBLOCK) != 0) 2588 error = EAGAIN; 2589 if (error == 0) 2590 td->td_retval[0] = sbytes; 2591 } 2592 fdrop(ofp, td); 2593 2594 drop: 2595 fdrop(fp, td); 2596 return (error); 2597 } 2598 2599 int 2600 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg) 2601 { 2602 /* 2603 * Differences between FreeBSD and Linux sendfile: 2604 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to 2605 * mean send the whole file). 2606 * - Linux can send to any fd whereas FreeBSD only supports sockets. 2607 * We therefore use FreeBSD sendfile where possible for performance, 2608 * but fall back on a manual copy (sendfile_fallback). 2609 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr. 2610 * - Linux takes an offset pointer and updates it to the read location. 2611 * FreeBSD takes in an offset and a 'bytes read' parameter which is 2612 * only filled if it isn't NULL. We use this parameter to update the 2613 * offset pointer if it exists. 2614 * - Linux sendfile returns bytes read on success while FreeBSD 2615 * returns 0. We use the 'bytes read' parameter to get this value. 2616 */ 2617 2618 off_t offset64; 2619 l_off_t offset; 2620 int error; 2621 2622 if (arg->offset != NULL) { 2623 error = copyin(arg->offset, &offset, sizeof(offset)); 2624 if (error != 0) 2625 return (error); 2626 offset64 = offset; 2627 } 2628 2629 error = linux_sendfile_common(td, arg->out, arg->in, 2630 arg->offset != NULL ? &offset64 : NULL, arg->count); 2631 2632 if (error == 0 && arg->offset != NULL) { 2633 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 2634 if (offset64 > INT32_MAX) 2635 return (EOVERFLOW); 2636 #endif 2637 offset = (l_off_t)offset64; 2638 error = copyout(&offset, arg->offset, sizeof(offset)); 2639 } 2640 2641 return (error); 2642 } 2643 2644 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 2645 int 2646 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg) 2647 { 2648 off_t offset; 2649 int error; 2650 2651 if (arg->offset != NULL) { 2652 error = copyin(arg->offset, &offset, sizeof(offset)); 2653 if (error != 0) 2654 return (error); 2655 } 2656 2657 error = linux_sendfile_common(td, arg->out, arg->in, 2658 arg->offset != NULL ? &offset : NULL, arg->count); 2659 2660 if (error == 0 && arg->offset != NULL) 2661 error = copyout(&offset, arg->offset, sizeof(offset)); 2662 2663 return (error); 2664 } 2665 2666 /* Argument list sizes for linux_socketcall */ 2667 static const unsigned char lxs_args_cnt[] = { 2668 0 /* unused*/, 3 /* socket */, 2669 3 /* bind */, 3 /* connect */, 2670 2 /* listen */, 3 /* accept */, 2671 3 /* getsockname */, 3 /* getpeername */, 2672 4 /* socketpair */, 4 /* send */, 2673 4 /* recv */, 6 /* sendto */, 2674 6 /* recvfrom */, 2 /* shutdown */, 2675 5 /* setsockopt */, 5 /* getsockopt */, 2676 3 /* sendmsg */, 3 /* recvmsg */, 2677 4 /* accept4 */, 5 /* recvmmsg */, 2678 4 /* sendmmsg */, 4 /* sendfile */ 2679 }; 2680 #define LINUX_ARGS_CNT (nitems(lxs_args_cnt) - 1) 2681 #define LINUX_ARG_SIZE(x) (lxs_args_cnt[x] * sizeof(l_ulong)) 2682 2683 int 2684 linux_socketcall(struct thread *td, struct linux_socketcall_args *args) 2685 { 2686 l_ulong a[6]; 2687 #if defined(__amd64__) && defined(COMPAT_LINUX32) 2688 register_t l_args[6]; 2689 #endif 2690 void *arg; 2691 int error; 2692 2693 if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT) 2694 return (EINVAL); 2695 error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what)); 2696 if (error != 0) 2697 return (error); 2698 2699 #if defined(__amd64__) && defined(COMPAT_LINUX32) 2700 for (int i = 0; i < lxs_args_cnt[args->what]; ++i) 2701 l_args[i] = a[i]; 2702 arg = l_args; 2703 #else 2704 arg = a; 2705 #endif 2706 switch (args->what) { 2707 case LINUX_SOCKET: 2708 return (linux_socket(td, arg)); 2709 case LINUX_BIND: 2710 return (linux_bind(td, arg)); 2711 case LINUX_CONNECT: 2712 return (linux_connect(td, arg)); 2713 case LINUX_LISTEN: 2714 return (linux_listen(td, arg)); 2715 case LINUX_ACCEPT: 2716 return (linux_accept(td, arg)); 2717 case LINUX_GETSOCKNAME: 2718 return (linux_getsockname(td, arg)); 2719 case LINUX_GETPEERNAME: 2720 return (linux_getpeername(td, arg)); 2721 case LINUX_SOCKETPAIR: 2722 return (linux_socketpair(td, arg)); 2723 case LINUX_SEND: 2724 return (linux_send(td, arg)); 2725 case LINUX_RECV: 2726 return (linux_recv(td, arg)); 2727 case LINUX_SENDTO: 2728 return (linux_sendto(td, arg)); 2729 case LINUX_RECVFROM: 2730 return (linux_recvfrom(td, arg)); 2731 case LINUX_SHUTDOWN: 2732 return (linux_shutdown(td, arg)); 2733 case LINUX_SETSOCKOPT: 2734 return (linux_setsockopt(td, arg)); 2735 case LINUX_GETSOCKOPT: 2736 return (linux_getsockopt(td, arg)); 2737 case LINUX_SENDMSG: 2738 return (linux_sendmsg(td, arg)); 2739 case LINUX_RECVMSG: 2740 return (linux_recvmsg(td, arg)); 2741 case LINUX_ACCEPT4: 2742 return (linux_accept4(td, arg)); 2743 case LINUX_RECVMMSG: 2744 return (linux_recvmmsg(td, arg)); 2745 case LINUX_SENDMMSG: 2746 return (linux_sendmmsg(td, arg)); 2747 case LINUX_SENDFILE: 2748 return (linux_sendfile(td, arg)); 2749 } 2750 2751 linux_msg(td, "socket type %d not implemented", args->what); 2752 return (ENOSYS); 2753 } 2754 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 2755