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