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