util/osdep.c

   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
  27 /* Needed early for CONFIG_BSD etc. */
  28
  29 #ifdef CONFIG_SOLARIS
  30 #include <sys/statvfs.h>
  31 /* See MySQL bug #7156 (http://bugs.mysql.com/bug.php?id=7156) for
  32    discussion about Solaris header problems */
  33 extern int madvise(char *, size_t, int);
  34 #endif
  35
  36 #include "qemu-common.h"
  37 #include "qemu/cutils.h"
  38 #include "qemu/sockets.h"
  39 #include "qemu/error-report.h"
  40 #include "monitor/monitor.h"
  41
  42 static bool fips_enabled = false;
  43
  44 static const char *hw_version = QEMU_HW_VERSION;
  45
  46 int socket_set_cork(int fd, int v)
  47 {
  48 #if defined(SOL_TCP) && defined(TCP_CORK)
  49     return qemu_setsockopt(fd, SOL_TCP, TCP_CORK, &v, sizeof(v));
  50 #else
  51     return 0;
  52 #endif
  53 }
  54
  55 int socket_set_nodelay(int fd)
  56 {
  57     int v = 1;
  58     return qemu_setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &v, sizeof(v));
  59 }
  60
  61 int qemu_madvise(void *addr, size_t len, int advice)
  62 {
  63     if (advice == QEMU_MADV_INVALID) {
  64         errno = EINVAL;
  65         return -1;
  66     }
  67 #if defined(CONFIG_MADVISE)
  68     return madvise(addr, len, advice);
  69 #elif defined(CONFIG_POSIX_MADVISE)
  70     return posix_madvise(addr, len, advice);
  71 #else
  72     errno = EINVAL;
  73     return -1;
  74 #endif
  75 }
  76
  77 static int qemu_mprotect__osdep(void *addr, size_t size, int prot)
  78 {
  79     g_assert(!((uintptr_t)addr & ~qemu_real_host_page_mask));
  80     g_assert(!(size & ~qemu_real_host_page_mask));
  81
  82 #ifdef _WIN32
  83     DWORD old_protect;
  84
  85     if (!VirtualProtect(addr, size, prot, &old_protect)) {
  86         g_autofree gchar *emsg = g_win32_error_message(GetLastError());
  87         error_report("%s: VirtualProtect failed: %s", __func__, emsg);
  88         return -1;
  89     }
  90     return 0;
  91 #else
  92     if (mprotect(addr, size, prot)) {
  93         error_report("%s: mprotect failed: %s", __func__, strerror(errno));
  94         return -1;
  95     }
  96     return 0;
  97 #endif
  98 }
  99
 100 int qemu_mprotect_rwx(void *addr, size_t size)
 101 {
 102 #ifdef _WIN32
 103     return qemu_mprotect__osdep(addr, size, PAGE_EXECUTE_READWRITE);
 104 #else
 105     return qemu_mprotect__osdep(addr, size, PROT_READ | PROT_WRITE | PROT_EXEC);
 106 #endif
 107 }
 108
 109 int qemu_mprotect_none(void *addr, size_t size)
 110 {
 111 #ifdef _WIN32
 112     return qemu_mprotect__osdep(addr, size, PAGE_NOACCESS);
 113 #else
 114     return qemu_mprotect__osdep(addr, size, PROT_NONE);
 115 #endif
 116 }
 117
 118 #ifndef _WIN32
 119
 120 static int fcntl_op_setlk = -1;
 121 static int fcntl_op_getlk = -1;
 122
 123 /*
 124  * Dups an fd and sets the flags
 125  */
 126 int qemu_dup_flags(int fd, int flags)
 127 {
 128     int ret;
 129     int serrno;
 130     int dup_flags;
 131
 132     ret = qemu_dup(fd);
 133     if (ret == -1) {
 134         goto fail;
 135     }
 136
 137     dup_flags = fcntl(ret, F_GETFL);
 138     if (dup_flags == -1) {
 139         goto fail;
 140     }
 141
 142     if ((flags & O_SYNC) != (dup_flags & O_SYNC)) {
 143         errno = EINVAL;
 144         goto fail;
 145     }
 146
 147     /* Set/unset flags that we can with fcntl */
 148     if (fcntl(ret, F_SETFL, flags) == -1) {
 149         goto fail;
 150     }
 151
 152     /* Truncate the file in the cases that open() would truncate it */
 153     if (flags & O_TRUNC ||
 154             ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))) {
 155         if (ftruncate(ret, 0) == -1) {
 156             goto fail;
 157         }
 158     }
 159
 160     return ret;
 161
 162 fail:
 163     serrno = errno;
 164     if (ret != -1) {
 165         close(ret);
 166     }
 167     errno = serrno;
 168     return -1;
 169 }
 170
 171 int qemu_dup(int fd)
 172 {
 173     int ret;
 174 #ifdef F_DUPFD_CLOEXEC
 175     ret = fcntl(fd, F_DUPFD_CLOEXEC, 0);
 176 #else
 177     ret = dup(fd);
 178     if (ret != -1) {
 179         qemu_set_cloexec(ret);
 180     }
 181 #endif
 182     return ret;
 183 }
 184
 185 static int qemu_parse_fdset(const char *param)
 186 {
 187     return qemu_parse_fd(param);
 188 }
 189
 190 static void qemu_probe_lock_ops(void)
 191 {
 192     if (fcntl_op_setlk == -1) {
 193 #ifdef F_OFD_SETLK
 194         int fd;
 195         int ret;
 196         struct flock fl = {
 197             .l_whence = SEEK_SET,
 198             .l_start  = 0,
 199             .l_len    = 0,
 200             .l_type   = F_WRLCK,
 201         };
 202
 203         fd = open("/dev/null", O_RDWR);
 204         if (fd < 0) {
 205             fprintf(stderr,
 206                     "Failed to open /dev/null for OFD lock probing: %s\n",
 207                     strerror(errno));
 208             fcntl_op_setlk = F_SETLK;
 209             fcntl_op_getlk = F_GETLK;
 210             return;
 211         }
 212         ret = fcntl(fd, F_OFD_GETLK, &fl);
 213         close(fd);
 214         if (!ret) {
 215             fcntl_op_setlk = F_OFD_SETLK;
 216             fcntl_op_getlk = F_OFD_GETLK;
 217         } else {
 218             fcntl_op_setlk = F_SETLK;
 219             fcntl_op_getlk = F_GETLK;
 220         }
 221 #else
 222         fcntl_op_setlk = F_SETLK;
 223         fcntl_op_getlk = F_GETLK;
 224 #endif
 225     }
 226 }
 227
 228 bool qemu_has_ofd_lock(void)
 229 {
 230     qemu_probe_lock_ops();
 231 #ifdef F_OFD_SETLK
 232     return fcntl_op_setlk == F_OFD_SETLK;
 233 #else
 234     return false;
 235 #endif
 236 }
 237
 238 static int qemu_lock_fcntl(int fd, int64_t start, int64_t len, int fl_type)
 239 {
 240     int ret;
 241     struct flock fl = {
 242         .l_whence = SEEK_SET,
 243         .l_start  = start,
 244         .l_len    = len,
 245         .l_type   = fl_type,
 246     };
 247     qemu_probe_lock_ops();
 248     do {
 249         ret = fcntl(fd, fcntl_op_setlk, &fl);
 250     } while (ret == -1 && errno == EINTR);
 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 #endif
 282
 283 static int qemu_open_cloexec(const char *name, int flags, mode_t mode)
 284 {
 285     int ret;
 286 #ifdef O_CLOEXEC
 287     ret = open(name, flags | O_CLOEXEC, mode);
 288 #else
 289     ret = open(name, flags, mode);
 290     if (ret >= 0) {
 291         qemu_set_cloexec(ret);
 292     }
 293 #endif
 294     return ret;
 295 }
 296
 297 /*
 298  * Opens a file with FD_CLOEXEC set
 299  */
 300 static int
 301 qemu_open_internal(const char *name, int flags, mode_t mode, Error **errp)
 302 {
 303     int ret;
 304
 305 #ifndef _WIN32
 306     const char *fdset_id_str;
 307
 308     /* Attempt dup of fd from fd set */
 309     if (strstart(name, "/dev/fdset/", &fdset_id_str)) {
 310         int64_t fdset_id;
 311         int dupfd;
 312
 313         fdset_id = qemu_parse_fdset(fdset_id_str);
 314         if (fdset_id == -1) {
 315             error_setg(errp, "Could not parse fdset %s", name);
 316             errno = EINVAL;
 317             return -1;
 318         }
 319
 320         dupfd = monitor_fdset_dup_fd_add(fdset_id, flags);
 321         if (dupfd == -1) {
 322             error_setg_errno(errp, errno, "Could not dup FD for %s flags %x",
 323                              name, flags);
 324             return -1;
 325         }
 326
 327         return dupfd;
 328     }
 329 #endif
 330
 331     ret = qemu_open_cloexec(name, flags, mode);
 332
 333     if (ret == -1) {
 334         const char *action = flags & O_CREAT ? "create" : "open";
 335 #ifdef O_DIRECT
 336         /* Give more helpful error message for O_DIRECT */
 337         if (errno == EINVAL && (flags & O_DIRECT)) {
 338             ret = open(name, flags & ~O_DIRECT, mode);
 339             if (ret != -1) {
 340                 close(ret);
 341                 error_setg(errp, "Could not %s '%s': "
 342                            "filesystem does not support O_DIRECT",
 343                            action, name);
 344                 errno = EINVAL; /* restore first open()'s errno */
 345                 return -1;
 346             }
 347         }
 348 #endif /* O_DIRECT */
 349         error_setg_errno(errp, errno, "Could not %s '%s'",
 350                          action, name);
 351     }
 352
 353     return ret;
 354 }
 355
 356
 357 int qemu_open(const char *name, int flags, Error **errp)
 358 {
 359     assert(!(flags & O_CREAT));
 360
 361     return qemu_open_internal(name, flags, 0, errp);
 362 }
 363
 364
 365 int qemu_create(const char *name, int flags, mode_t mode, Error **errp)
 366 {
 367     assert(!(flags & O_CREAT));
 368
 369     return qemu_open_internal(name, flags | O_CREAT, mode, errp);
 370 }
 371
 372
 373 int qemu_open_old(const char *name, int flags, ...)
 374 {
 375     va_list ap;
 376     mode_t mode = 0;
 377     int ret;
 378
 379     va_start(ap, flags);
 380     if (flags & O_CREAT) {
 381         mode = va_arg(ap, int);
 382     }
 383     va_end(ap);
 384
 385     ret = qemu_open_internal(name, flags, mode, NULL);
 386
 387 #ifdef O_DIRECT
 388     if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
 389         error_report("file system may not support O_DIRECT");
 390         errno = EINVAL; /* in case it was clobbered */
 391     }
 392 #endif /* O_DIRECT */
 393
 394     return ret;
 395 }
 396
 397 int qemu_close(int fd)
 398 {
 399     int64_t fdset_id;
 400
 401     /* Close fd that was dup'd from an fdset */
 402     fdset_id = monitor_fdset_dup_fd_find(fd);
 403     if (fdset_id != -1) {
 404         int ret;
 405
 406         ret = close(fd);
 407         if (ret == 0) {
 408             monitor_fdset_dup_fd_remove(fd);
 409         }
 410
 411         return ret;
 412     }
 413
 414     return close(fd);
 415 }
 416
 417 /*
 418  * Delete a file from the filesystem, unless the filename is /dev/fdset/...
 419  *
 420  * Returns: On success, zero is returned.  On error, -1 is returned,
 421  * and errno is set appropriately.
 422  */
 423 int qemu_unlink(const char *name)
 424 {
 425     if (g_str_has_prefix(name, "/dev/fdset/")) {
 426         return 0;
 427     }
 428
 429     return unlink(name);
 430 }
 431
 432 /*
 433  * A variant of write(2) which handles partial write.
 434  *
 435  * Return the number of bytes transferred.
 436  * Set errno if fewer than `count' bytes are written.
 437  *
 438  * This function don't work with non-blocking fd's.
 439  * Any of the possibilities with non-blocking fd's is bad:
 440  *   - return a short write (then name is wrong)
 441  *   - busy wait adding (errno == EAGAIN) to the loop
 442  */
 443 ssize_t qemu_write_full(int fd, const void *buf, size_t count)
 444 {
 445     ssize_t ret = 0;
 446     ssize_t total = 0;
 447
 448     while (count) {
 449         ret = write(fd, buf, count);
 450         if (ret < 0) {
 451             if (errno == EINTR)
 452                 continue;
 453             break;
 454         }
 455
 456         count -= ret;
 457         buf += ret;
 458         total += ret;
 459     }
 460
 461     return total;
 462 }
 463
 464 /*
 465  * Opens a socket with FD_CLOEXEC set
 466  */
 467 int qemu_socket(int domain, int type, int protocol)
 468 {
 469     int ret;
 470
 471 #ifdef SOCK_CLOEXEC
 472     ret = socket(domain, type | SOCK_CLOEXEC, protocol);
 473     if (ret != -1 || errno != EINVAL) {
 474         return ret;
 475     }
 476 #endif
 477     ret = socket(domain, type, protocol);
 478     if (ret >= 0) {
 479         qemu_set_cloexec(ret);
 480     }
 481
 482     return ret;
 483 }
 484
 485 /*
 486  * Accept a connection and set FD_CLOEXEC
 487  */
 488 int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
 489 {
 490     int ret;
 491
 492 #ifdef CONFIG_ACCEPT4
 493     ret = accept4(s, addr, addrlen, SOCK_CLOEXEC);
 494     if (ret != -1 || errno != ENOSYS) {
 495         return ret;
 496     }
 497 #endif
 498     ret = accept(s, addr, addrlen);
 499     if (ret >= 0) {
 500         qemu_set_cloexec(ret);
 501     }
 502
 503     return ret;
 504 }
 505
 506 void qemu_set_hw_version(const char *version)
 507 {
 508     hw_version = version;
 509 }
 510
 511 const char *qemu_hw_version(void)
 512 {
 513     return hw_version;
 514 }
 515
 516 void fips_set_state(bool requested)
 517 {
 518 #ifdef __linux__
 519     if (requested) {
 520         FILE *fds = fopen("/proc/sys/crypto/fips_enabled", "r");
 521         if (fds != NULL) {
 522             fips_enabled = (fgetc(fds) == '1');
 523             fclose(fds);
 524         }
 525     }
 526 #else
 527     fips_enabled = false;
 528 #endif /* __linux__ */
 529
 530 #ifdef _FIPS_DEBUG
 531     fprintf(stderr, "FIPS mode %s (requested %s)\n",
 532             (fips_enabled ? "enabled" : "disabled"),
 533             (requested ? "enabled" : "disabled"));
 534 #endif
 535 }
 536
 537 bool fips_get_state(void)
 538 {
 539     return fips_enabled;
 540 }
 541
 542 #ifdef _WIN32
 543 static void socket_cleanup(void)
 544 {
 545     WSACleanup();
 546 }
 547 #endif
 548
 549 int socket_init(void)
 550 {
 551 #ifdef _WIN32
 552     WSADATA Data;
 553     int ret, err;
 554
 555     ret = WSAStartup(MAKEWORD(2, 2), &Data);
 556     if (ret != 0) {
 557         err = WSAGetLastError();
 558         fprintf(stderr, "WSAStartup: %d\n", err);
 559         return -1;
 560     }
 561     atexit(socket_cleanup);
 562 #endif
 563     return 0;
 564 }
 565
 566
 567 #ifndef CONFIG_IOVEC
 568 /* helper function for iov_send_recv() */
 569 static ssize_t
 570 readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write)
 571 {
 572     unsigned i = 0;
 573     ssize_t ret = 0;
 574     while (i < iov_cnt) {
 575         ssize_t r = do_write
 576             ? write(fd, iov[i].iov_base, iov[i].iov_len)
 577             : read(fd, iov[i].iov_base, iov[i].iov_len);
 578         if (r > 0) {
 579             ret += r;
 580         } else if (!r) {
 581             break;
 582         } else if (errno == EINTR) {
 583             continue;
 584         } else {
 585             /* else it is some "other" error,
 586              * only return if there was no data processed. */
 587             if (ret == 0) {
 588                 ret = -1;
 589             }
 590             break;
 591         }
 592         i++;
 593     }
 594     return ret;
 595 }
 596
 597 ssize_t
 598 readv(int fd, const struct iovec *iov, int iov_cnt)
 599 {
 600     return readv_writev(fd, iov, iov_cnt, false);
 601 }
 602
 603 ssize_t
 604 writev(int fd, const struct iovec *iov, int iov_cnt)
 605 {
 606     return readv_writev(fd, iov, iov_cnt, true);
 607 }
 608 #endif