| 1 | /* |
| 2 | * QEMU low level functions |
| 3 | * |
| 4 | * Copyright (c) 2003 Fabrice Bellard |
| 5 | * |
| 6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
| 7 | * of this software and associated documentation files (the "Software"), to deal |
| 8 | * in the Software without restriction, including without limitation the rights |
| 9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell |
| 10 | * copies of the Software, and to permit persons to whom the Software is |
| 11 | * furnished to do so, subject to the following conditions: |
| 12 | * |
| 13 | * The above copyright notice and this permission notice shall be included in |
| 14 | * all copies or substantial portions of the Software. |
| 15 | * |
| 16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
| 19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN |
| 22 | * THE SOFTWARE. |
| 23 | */ |
| 24 | #include "qemu/osdep.h" |
| 25 | #include "qapi/error.h" |
| 26 | #include "qemu/cutils.h" |
| 27 | #include "qemu/sockets.h" |
| 28 | #include "qemu/error-report.h" |
| 29 | #include "qemu/madvise.h" |
| 30 | #include "qemu/mprotect.h" |
| 31 | #include "qemu/hw-version.h" |
| 32 | #include "monitor/monitor.h" |
| 33 | |
| 34 | int socket_set_cork(int fd, int v) |
| 35 | { |
| 36 | #if defined(SOL_TCP) && defined(TCP_CORK) |
| 37 | return setsockopt(fd, SOL_TCP, TCP_CORK, &v, sizeof(v)); |
| 38 | #else |
| 39 | return 0; |
| 40 | #endif |
| 41 | } |
| 42 | |
| 43 | int socket_set_nodelay(int fd) |
| 44 | { |
| 45 | int v = 1; |
| 46 | return setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &v, sizeof(v)); |
| 47 | } |
| 48 | |
| 49 | int qemu_madvise(void *addr, size_t len, int advice) |
| 50 | { |
| 51 | if (advice == QEMU_MADV_INVALID) { |
| 52 | errno = EINVAL; |
| 53 | return -1; |
| 54 | } |
| 55 | #if defined(CONFIG_MADVISE) |
| 56 | return madvise(addr, len, advice); |
| 57 | #elif defined(CONFIG_POSIX_MADVISE) |
| 58 | int rc = posix_madvise(addr, len, advice); |
| 59 | if (rc) { |
| 60 | errno = rc; |
| 61 | return -1; |
| 62 | } |
| 63 | return 0; |
| 64 | #else |
| 65 | errno = ENOSYS; |
| 66 | return -1; |
| 67 | #endif |
| 68 | } |
| 69 | |
| 70 | static int qemu_mprotect__osdep(void *addr, size_t size, int prot) |
| 71 | { |
| 72 | g_assert(!((uintptr_t)addr & ~qemu_real_host_page_mask())); |
| 73 | g_assert(!(size & ~qemu_real_host_page_mask())); |
| 74 | |
| 75 | #ifdef _WIN32 |
| 76 | DWORD old_protect; |
| 77 | |
| 78 | if (!VirtualProtect(addr, size, prot, &old_protect)) { |
| 79 | g_autofree gchar *emsg = g_win32_error_message(GetLastError()); |
| 80 | error_report("%s: VirtualProtect failed: %s", __func__, emsg); |
| 81 | return -1; |
| 82 | } |
| 83 | return 0; |
| 84 | #else |
| 85 | if (mprotect(addr, size, prot)) { |
| 86 | error_report("%s: mprotect failed: %s", __func__, strerror(errno)); |
| 87 | return -1; |
| 88 | } |
| 89 | return 0; |
| 90 | #endif |
| 91 | } |
| 92 | |
| 93 | int qemu_mprotect_rw(void *addr, size_t size) |
| 94 | { |
| 95 | #ifdef _WIN32 |
| 96 | return qemu_mprotect__osdep(addr, size, PAGE_READWRITE); |
| 97 | #else |
| 98 | return qemu_mprotect__osdep(addr, size, PROT_READ | PROT_WRITE); |
| 99 | #endif |
| 100 | } |
| 101 | |
| 102 | int qemu_mprotect_rwx(void *addr, size_t size) |
| 103 | { |
| 104 | #ifdef _WIN32 |
| 105 | return qemu_mprotect__osdep(addr, size, PAGE_EXECUTE_READWRITE); |
| 106 | #else |
| 107 | return qemu_mprotect__osdep(addr, size, PROT_READ | PROT_WRITE | PROT_EXEC); |
| 108 | #endif |
| 109 | } |
| 110 | |
| 111 | int qemu_mprotect_none(void *addr, size_t size) |
| 112 | { |
| 113 | #ifdef _WIN32 |
| 114 | return qemu_mprotect__osdep(addr, size, PAGE_NOACCESS); |
| 115 | #else |
| 116 | return qemu_mprotect__osdep(addr, size, PROT_NONE); |
| 117 | #endif |
| 118 | } |
| 119 | |
| 120 | #ifndef _WIN32 |
| 121 | |
| 122 | static int fcntl_op_setlk = -1; |
| 123 | static int fcntl_op_getlk = -1; |
| 124 | |
| 125 | /* |
| 126 | * Dups an fd and sets the flags |
| 127 | */ |
| 128 | int qemu_dup_flags(int fd, int flags) |
| 129 | { |
| 130 | int ret; |
| 131 | int serrno; |
| 132 | int dup_flags; |
| 133 | |
| 134 | ret = qemu_dup(fd); |
| 135 | if (ret == -1) { |
| 136 | goto fail; |
| 137 | } |
| 138 | |
| 139 | dup_flags = fcntl(ret, F_GETFL); |
| 140 | if (dup_flags == -1) { |
| 141 | goto fail; |
| 142 | } |
| 143 | |
| 144 | if ((flags & O_SYNC) != (dup_flags & O_SYNC)) { |
| 145 | errno = EINVAL; |
| 146 | goto fail; |
| 147 | } |
| 148 | |
| 149 | /* Set/unset flags that we can with fcntl */ |
| 150 | if (fcntl(ret, F_SETFL, flags) == -1) { |
| 151 | goto fail; |
| 152 | } |
| 153 | |
| 154 | /* Truncate the file in the cases that open() would truncate it */ |
| 155 | if (flags & O_TRUNC || |
| 156 | ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))) { |
| 157 | if (ftruncate(ret, 0) == -1) { |
| 158 | goto fail; |
| 159 | } |
| 160 | } |
| 161 | |
| 162 | return ret; |
| 163 | |
| 164 | fail: |
| 165 | serrno = errno; |
| 166 | if (ret != -1) { |
| 167 | close(ret); |
| 168 | } |
| 169 | errno = serrno; |
| 170 | return -1; |
| 171 | } |
| 172 | |
| 173 | int qemu_dup(int fd) |
| 174 | { |
| 175 | int ret; |
| 176 | #ifdef F_DUPFD_CLOEXEC |
| 177 | ret = fcntl(fd, F_DUPFD_CLOEXEC, 0); |
| 178 | #else |
| 179 | ret = dup(fd); |
| 180 | if (ret != -1) { |
| 181 | qemu_set_cloexec(ret); |
| 182 | } |
| 183 | #endif |
| 184 | return ret; |
| 185 | } |
| 186 | |
| 187 | static int qemu_parse_fdset(const char *param) |
| 188 | { |
| 189 | return qemu_parse_fd(param); |
| 190 | } |
| 191 | |
| 192 | static void qemu_probe_lock_ops(void) |
| 193 | { |
| 194 | if (fcntl_op_setlk == -1) { |
| 195 | #ifdef F_OFD_SETLK |
| 196 | int fd; |
| 197 | int ret; |
| 198 | struct flock fl = { |
| 199 | .l_whence = SEEK_SET, |
| 200 | .l_start = 0, |
| 201 | .l_len = 0, |
| 202 | .l_type = F_WRLCK, |
| 203 | }; |
| 204 | |
| 205 | fd = open("/dev/null", O_RDWR); |
| 206 | if (fd < 0) { |
| 207 | fprintf(stderr, |
| 208 | "Failed to open /dev/null for OFD lock probing: %s\n", |
| 209 | strerror(errno)); |
| 210 | fcntl_op_setlk = F_SETLK; |
| 211 | fcntl_op_getlk = F_GETLK; |
| 212 | return; |
| 213 | } |
| 214 | ret = fcntl(fd, F_OFD_GETLK, &fl); |
| 215 | close(fd); |
| 216 | if (!ret) { |
| 217 | fcntl_op_setlk = F_OFD_SETLK; |
| 218 | fcntl_op_getlk = F_OFD_GETLK; |
| 219 | } else { |
| 220 | fcntl_op_setlk = F_SETLK; |
| 221 | fcntl_op_getlk = F_GETLK; |
| 222 | } |
| 223 | #else |
| 224 | fcntl_op_setlk = F_SETLK; |
| 225 | fcntl_op_getlk = F_GETLK; |
| 226 | #endif |
| 227 | } |
| 228 | } |
| 229 | |
| 230 | bool qemu_has_ofd_lock(void) |
| 231 | { |
| 232 | qemu_probe_lock_ops(); |
| 233 | #ifdef F_OFD_SETLK |
| 234 | return fcntl_op_setlk == F_OFD_SETLK; |
| 235 | #else |
| 236 | return false; |
| 237 | #endif |
| 238 | } |
| 239 | |
| 240 | static int qemu_lock_fcntl(int fd, int64_t start, int64_t len, int fl_type) |
| 241 | { |
| 242 | int ret; |
| 243 | struct flock fl = { |
| 244 | .l_whence = SEEK_SET, |
| 245 | .l_start = start, |
| 246 | .l_len = len, |
| 247 | .l_type = fl_type, |
| 248 | }; |
| 249 | qemu_probe_lock_ops(); |
| 250 | ret = RETRY_ON_EINTR(fcntl(fd, fcntl_op_setlk, &fl)); |
| 251 | return ret == -1 ? -errno : 0; |
| 252 | } |
| 253 | |
| 254 | int qemu_lock_fd(int fd, int64_t start, int64_t len, bool exclusive) |
| 255 | { |
| 256 | return qemu_lock_fcntl(fd, start, len, exclusive ? F_WRLCK : F_RDLCK); |
| 257 | } |
| 258 | |
| 259 | int qemu_unlock_fd(int fd, int64_t start, int64_t len) |
| 260 | { |
| 261 | return qemu_lock_fcntl(fd, start, len, F_UNLCK); |
| 262 | } |
| 263 | |
| 264 | int qemu_lock_fd_test(int fd, int64_t start, int64_t len, bool exclusive) |
| 265 | { |
| 266 | int ret; |
| 267 | struct flock fl = { |
| 268 | .l_whence = SEEK_SET, |
| 269 | .l_start = start, |
| 270 | .l_len = len, |
| 271 | .l_type = exclusive ? F_WRLCK : F_RDLCK, |
| 272 | }; |
| 273 | qemu_probe_lock_ops(); |
| 274 | ret = fcntl(fd, fcntl_op_getlk, &fl); |
| 275 | if (ret == -1) { |
| 276 | return -errno; |
| 277 | } else { |
| 278 | return fl.l_type == F_UNLCK ? 0 : -EAGAIN; |
| 279 | } |
| 280 | } |
| 281 | |
| 282 | /** |
| 283 | * Set the given flag(s) (fcntl GETFL/SETFL) on the given FD, while retaining |
| 284 | * other flags. |
| 285 | */ |
| 286 | int qemu_fcntl_addfl(int fd, int flag) |
| 287 | { |
| 288 | int flags; |
| 289 | |
| 290 | flags = fcntl(fd, F_GETFL); |
| 291 | if (flags == -1) { |
| 292 | return -errno; |
| 293 | } |
| 294 | if (fcntl(fd, F_SETFL, flags | flag) == -1) { |
| 295 | return -errno; |
| 296 | } |
| 297 | return 0; |
| 298 | } |
| 299 | #endif |
| 300 | |
| 301 | bool qemu_has_direct_io(void) |
| 302 | { |
| 303 | #ifdef O_DIRECT |
| 304 | return true; |
| 305 | #else |
| 306 | return false; |
| 307 | #endif |
| 308 | } |
| 309 | |
| 310 | static int qemu_open_cloexec(const char *name, int flags, mode_t mode) |
| 311 | { |
| 312 | int ret; |
| 313 | #ifdef O_CLOEXEC |
| 314 | ret = open(name, flags | O_CLOEXEC, mode); |
| 315 | #else |
| 316 | ret = open(name, flags, mode); |
| 317 | if (ret >= 0) { |
| 318 | qemu_set_cloexec(ret); |
| 319 | } |
| 320 | #endif |
| 321 | return ret; |
| 322 | } |
| 323 | |
| 324 | /* |
| 325 | * Opens a file with FD_CLOEXEC set |
| 326 | */ |
| 327 | static int |
| 328 | qemu_open_internal(const char *name, int flags, mode_t mode, Error **errp) |
| 329 | { |
| 330 | int ret; |
| 331 | |
| 332 | #ifndef _WIN32 |
| 333 | const char *fdset_id_str; |
| 334 | |
| 335 | /* Attempt dup of fd from fd set */ |
| 336 | if (strstart(name, "/dev/fdset/", &fdset_id_str)) { |
| 337 | int64_t fdset_id; |
| 338 | |
| 339 | fdset_id = qemu_parse_fdset(fdset_id_str); |
| 340 | if (fdset_id == -1) { |
| 341 | error_setg(errp, "Could not parse fdset %s", name); |
| 342 | errno = EINVAL; |
| 343 | return -1; |
| 344 | } |
| 345 | |
| 346 | return monitor_fdset_dup_fd_add(fdset_id, flags, errp); |
| 347 | } |
| 348 | #endif |
| 349 | |
| 350 | ret = qemu_open_cloexec(name, flags, mode); |
| 351 | |
| 352 | if (ret == -1) { |
| 353 | const char *action = flags & O_CREAT ? "create" : "open"; |
| 354 | #ifdef O_DIRECT |
| 355 | /* Give more helpful error message for O_DIRECT */ |
| 356 | if (errno == EINVAL && (flags & O_DIRECT)) { |
| 357 | ret = open(name, flags & ~O_DIRECT, mode); |
| 358 | if (ret != -1) { |
| 359 | close(ret); |
| 360 | error_setg(errp, "Could not %s '%s': " |
| 361 | "filesystem does not support O_DIRECT", |
| 362 | action, name); |
| 363 | errno = EINVAL; /* restore first open()'s errno */ |
| 364 | return -1; |
| 365 | } |
| 366 | } |
| 367 | #endif /* O_DIRECT */ |
| 368 | error_setg_errno(errp, errno, "Could not %s '%s'", |
| 369 | action, name); |
| 370 | } |
| 371 | |
| 372 | return ret; |
| 373 | } |
| 374 | |
| 375 | |
| 376 | int qemu_open(const char *name, int flags, Error **errp) |
| 377 | { |
| 378 | assert(!(flags & O_CREAT)); |
| 379 | |
| 380 | return qemu_open_internal(name, flags, 0, errp); |
| 381 | } |
| 382 | |
| 383 | |
| 384 | int qemu_create(const char *name, int flags, mode_t mode, Error **errp) |
| 385 | { |
| 386 | assert(!(flags & O_CREAT)); |
| 387 | |
| 388 | return qemu_open_internal(name, flags | O_CREAT, mode, errp); |
| 389 | } |
| 390 | |
| 391 | |
| 392 | int qemu_open_old(const char *name, int flags, ...) |
| 393 | { |
| 394 | va_list ap; |
| 395 | mode_t mode = 0; |
| 396 | int ret; |
| 397 | |
| 398 | va_start(ap, flags); |
| 399 | if (flags & O_CREAT) { |
| 400 | mode = va_arg(ap, int); |
| 401 | } |
| 402 | va_end(ap); |
| 403 | |
| 404 | ret = qemu_open_internal(name, flags, mode, NULL); |
| 405 | |
| 406 | #ifdef O_DIRECT |
| 407 | if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) { |
| 408 | error_report("file system may not support O_DIRECT"); |
| 409 | errno = EINVAL; /* in case it was clobbered */ |
| 410 | } |
| 411 | #endif /* O_DIRECT */ |
| 412 | |
| 413 | return ret; |
| 414 | } |
| 415 | |
| 416 | int qemu_close(int fd) |
| 417 | { |
| 418 | /* Close fd that was dup'd from an fdset */ |
| 419 | monitor_fdset_dup_fd_remove(fd); |
| 420 | return close(fd); |
| 421 | } |
| 422 | |
| 423 | /* |
| 424 | * Delete a file from the filesystem, unless the filename is /dev/fdset/... |
| 425 | * |
| 426 | * Returns: On success, zero is returned. On error, -1 is returned, |
| 427 | * and errno is set appropriately. |
| 428 | */ |
| 429 | int qemu_unlink(const char *name) |
| 430 | { |
| 431 | if (g_str_has_prefix(name, "/dev/fdset/")) { |
| 432 | return 0; |
| 433 | } |
| 434 | |
| 435 | return unlink(name); |
| 436 | } |
| 437 | |
| 438 | /* |
| 439 | * A variant of write(2) which handles partial write. |
| 440 | * |
| 441 | * Return the number of bytes transferred. |
| 442 | * Set errno if fewer than `count' bytes are written. |
| 443 | * |
| 444 | * This function don't work with non-blocking fd's. |
| 445 | * Any of the possibilities with non-blocking fd's is bad: |
| 446 | * - return a short write (then name is wrong) |
| 447 | * - busy wait adding (errno == EAGAIN) to the loop |
| 448 | */ |
| 449 | ssize_t qemu_write_full(int fd, const void *buf, size_t count) |
| 450 | { |
| 451 | ssize_t ret = 0; |
| 452 | ssize_t total = 0; |
| 453 | |
| 454 | while (count) { |
| 455 | ret = write(fd, buf, count); |
| 456 | if (ret < 0) { |
| 457 | if (errno == EINTR) |
| 458 | continue; |
| 459 | break; |
| 460 | } |
| 461 | |
| 462 | count -= ret; |
| 463 | buf += ret; |
| 464 | total += ret; |
| 465 | } |
| 466 | |
| 467 | return total; |
| 468 | } |
| 469 | |
| 470 | /* |
| 471 | * Opens a socket with FD_CLOEXEC set |
| 472 | */ |
| 473 | int qemu_socket(int domain, int type, int protocol) |
| 474 | { |
| 475 | int ret; |
| 476 | |
| 477 | #ifdef SOCK_CLOEXEC |
| 478 | ret = socket(domain, type | SOCK_CLOEXEC, protocol); |
| 479 | if (ret != -1 || errno != EINVAL) { |
| 480 | return ret; |
| 481 | } |
| 482 | #endif |
| 483 | ret = socket(domain, type, protocol); |
| 484 | if (ret >= 0) { |
| 485 | qemu_set_cloexec(ret); |
| 486 | } |
| 487 | |
| 488 | return ret; |
| 489 | } |
| 490 | |
| 491 | /* |
| 492 | * Accept a connection and set FD_CLOEXEC |
| 493 | */ |
| 494 | int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen) |
| 495 | { |
| 496 | int ret; |
| 497 | |
| 498 | #ifdef CONFIG_ACCEPT4 |
| 499 | ret = accept4(s, addr, addrlen, SOCK_CLOEXEC); |
| 500 | if (ret != -1 || errno != ENOSYS) { |
| 501 | return ret; |
| 502 | } |
| 503 | #endif |
| 504 | ret = accept(s, addr, addrlen); |
| 505 | if (ret >= 0) { |
| 506 | qemu_set_cloexec(ret); |
| 507 | } |
| 508 | |
| 509 | return ret; |
| 510 | } |
| 511 | |
| 512 | ssize_t qemu_send_full(int s, const void *buf, size_t count) |
| 513 | { |
| 514 | ssize_t ret = 0; |
| 515 | ssize_t total = 0; |
| 516 | |
| 517 | while (count) { |
| 518 | ret = send(s, buf, count, 0); |
| 519 | if (ret < 0) { |
| 520 | if (errno == EINTR) { |
| 521 | continue; |
| 522 | } |
| 523 | break; |
| 524 | } |
| 525 | |
| 526 | count -= ret; |
| 527 | buf += ret; |
| 528 | total += ret; |
| 529 | } |
| 530 | |
| 531 | return total; |
| 532 | } |
| 533 | |
| 534 | #ifdef _WIN32 |
| 535 | static void socket_cleanup(void) |
| 536 | { |
| 537 | WSACleanup(); |
| 538 | } |
| 539 | #endif |
| 540 | |
| 541 | int socket_init(void) |
| 542 | { |
| 543 | #ifdef _WIN32 |
| 544 | WSADATA Data; |
| 545 | int ret, err; |
| 546 | |
| 547 | ret = WSAStartup(MAKEWORD(2, 2), &Data); |
| 548 | if (ret != 0) { |
| 549 | err = WSAGetLastError(); |
| 550 | fprintf(stderr, "WSAStartup: %d\n", err); |
| 551 | return -1; |
| 552 | } |
| 553 | atexit(socket_cleanup); |
| 554 | #endif |
| 555 | return 0; |
| 556 | } |
| 557 | |
| 558 | |
| 559 | #ifndef CONFIG_IOVEC |
| 560 | static ssize_t |
| 561 | readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write) |
| 562 | { |
| 563 | unsigned i = 0; |
| 564 | ssize_t ret = 0; |
| 565 | ssize_t off = 0; |
| 566 | while (i < iov_cnt) { |
| 567 | ssize_t r = do_write |
| 568 | ? write(fd, iov[i].iov_base + off, iov[i].iov_len - off) |
| 569 | : read(fd, iov[i].iov_base + off, iov[i].iov_len - off); |
| 570 | if (r > 0) { |
| 571 | ret += r; |
| 572 | off += r; |
| 573 | if (off < iov[i].iov_len) { |
| 574 | continue; |
| 575 | } |
| 576 | } else if (!r) { |
| 577 | break; |
| 578 | } else if (errno == EINTR) { |
| 579 | continue; |
| 580 | } else { |
| 581 | /* else it is some "other" error, |
| 582 | * only return if there was no data processed. */ |
| 583 | if (ret == 0) { |
| 584 | ret = -1; |
| 585 | } |
| 586 | break; |
| 587 | } |
| 588 | off = 0; |
| 589 | i++; |
| 590 | } |
| 591 | return ret; |
| 592 | } |
| 593 | |
| 594 | ssize_t |
| 595 | readv(int fd, const struct iovec *iov, int iov_cnt) |
| 596 | { |
| 597 | return readv_writev(fd, iov, iov_cnt, false); |
| 598 | } |
| 599 | |
| 600 | ssize_t |
| 601 | writev(int fd, const struct iovec *iov, int iov_cnt) |
| 602 | { |
| 603 | return readv_writev(fd, iov, iov_cnt, true); |
| 604 | } |
| 605 | #endif |
| 606 | |
| 607 | /* |
| 608 | * Make sure data goes on disk, but if possible do not bother to |
| 609 | * write out the inode just for timestamp updates. |
| 610 | * |
| 611 | * Unfortunately even in 2009 many operating systems do not support |
| 612 | * fdatasync and have to fall back to fsync. |
| 613 | */ |
| 614 | int qemu_fdatasync(int fd) |
| 615 | { |
| 616 | #ifdef CONFIG_FDATASYNC |
| 617 | return fdatasync(fd); |
| 618 | #else |
| 619 | return fsync(fd); |
| 620 | #endif |
| 621 | } |