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
  26 /* Needed early for CONFIG_BSD etc. */
  27
  28 #ifdef CONFIG_SOLARIS
  29 #include <sys/statvfs.h>
  30 /* See MySQL bug #7156 (http://bugs.mysql.com/bug.php?id=7156) for
  31    discussion about Solaris header problems */
  32 extern int madvise(caddr_t, size_t, int);
  33 #endif
  34
  35 #include "qemu-common.h"
  36 #include "qemu/cutils.h"
  37 #include "qemu/sockets.h"
  38 #include "qemu/error-report.h"
  39 #include "monitor/monitor.h"
  40
  41 static bool fips_enabled = false;
  42
  43 static const char *hw_version = QEMU_HW_VERSION;
  44
  45 int socket_set_cork(int fd, int v)
  46 {
  47 #if defined(SOL_TCP) && defined(TCP_CORK)
  48     return qemu_setsockopt(fd, SOL_TCP, TCP_CORK, &v, sizeof(v));
  49 #else
  50     return 0;
  51 #endif
  52 }
  53
  54 int socket_set_nodelay(int fd)
  55 {
  56     int v = 1;
  57     return qemu_setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &v, sizeof(v));
  58 }
  59
  60 int qemu_madvise(void *addr, size_t len, int advice)
  61 {
  62     if (advice == QEMU_MADV_INVALID) {
  63         errno = EINVAL;
  64         return -1;
  65     }
  66 #if defined(CONFIG_MADVISE)
  67     return madvise(addr, len, advice);
  68 #elif defined(CONFIG_POSIX_MADVISE)
  69     return posix_madvise(addr, len, advice);
  70 #else
  71     errno = EINVAL;
  72     return -1;
  73 #endif
  74 }
  75
  76 static int qemu_mprotect__osdep(void *addr, size_t size, int prot)
  77 {
  78     g_assert(!((uintptr_t)addr & ~qemu_real_host_page_mask));
  79     g_assert(!(size & ~qemu_real_host_page_mask));
  80
  81 #ifdef _WIN32
  82     DWORD old_protect;
  83
  84     if (!VirtualProtect(addr, size, prot, &old_protect)) {
  85         error_report("%s: VirtualProtect failed with error code %ld",
  86                      __func__, GetLastError());
  87         return -1;
  88     }
  89     return 0;
  90 #else
  91     if (mprotect(addr, size, prot)) {
  92         error_report("%s: mprotect failed: %s", __func__, strerror(errno));
  93         return -1;
  94     }
  95     return 0;
  96 #endif
  97 }
  98
  99 int qemu_mprotect_rwx(void *addr, size_t size)
 100 {
 101 #ifdef _WIN32
 102     return qemu_mprotect__osdep(addr, size, PAGE_EXECUTE_READWRITE);
 103 #else
 104     return qemu_mprotect__osdep(addr, size, PROT_READ | PROT_WRITE | PROT_EXEC);
 105 #endif
 106 }
 107
 108 int qemu_mprotect_none(void *addr, size_t size)
 109 {
 110 #ifdef _WIN32
 111     return qemu_mprotect__osdep(addr, size, PAGE_NOACCESS);
 112 #else
 113     return qemu_mprotect__osdep(addr, size, PROT_NONE);
 114 #endif
 115 }
 116
 117 #ifndef _WIN32
 118
 119 static int fcntl_op_setlk = -1;
 120 static int fcntl_op_getlk = -1;
 121
 122 /*
 123  * Dups an fd and sets the flags
 124  */
 125 static int qemu_dup_flags(int fd, int flags)
 126 {
 127     int ret;
 128     int serrno;
 129     int dup_flags;
 130
 131     ret = qemu_dup(fd);
 132     if (ret == -1) {
 133         goto fail;
 134     }
 135
 136     dup_flags = fcntl(ret, F_GETFL);
 137     if (dup_flags == -1) {
 138         goto fail;
 139     }
 140
 141     if ((flags & O_SYNC) != (dup_flags & O_SYNC)) {
 142         errno = EINVAL;
 143         goto fail;
 144     }
 145
 146     /* Set/unset flags that we can with fcntl */
 147     if (fcntl(ret, F_SETFL, flags) == -1) {
 148         goto fail;
 149     }
 150
 151     /* Truncate the file in the cases that open() would truncate it */
 152     if (flags & O_TRUNC ||
 153             ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))) {
 154         if (ftruncate(ret, 0) == -1) {
 155             goto fail;
 156         }
 157     }
 158
 159     return ret;
 160
 161 fail:
 162     serrno = errno;
 163     if (ret != -1) {
 164         close(ret);
 165     }
 166     errno = serrno;
 167     return -1;
 168 }
 169
 170 int qemu_dup(int fd)
 171 {
 172     int ret;
 173 #ifdef F_DUPFD_CLOEXEC
 174     ret = fcntl(fd, F_DUPFD_CLOEXEC, 0);
 175 #else
 176     ret = dup(fd);
 177     if (ret != -1) {
 178         qemu_set_cloexec(ret);
 179     }
 180 #endif
 181     return ret;
 182 }
 183
 184 static int qemu_parse_fdset(const char *param)
 185 {
 186     return qemu_parse_fd(param);
 187 }
 188
 189 static void qemu_probe_lock_ops(void)
 190 {
 191     if (fcntl_op_setlk == -1) {
 192 #ifdef F_OFD_SETLK
 193         int fd;
 194         int ret;
 195         struct flock fl = {
 196             .l_whence = SEEK_SET,
 197             .l_start  = 0,
 198             .l_len    = 0,
 199             .l_type   = F_WRLCK,
 200         };
 201
 202         fd = open("/dev/null", O_RDWR);
 203         if (fd < 0) {
 204             fprintf(stderr,
 205                     "Failed to open /dev/null for OFD lock probing: %s\n",
 206                     strerror(errno));
 207             fcntl_op_setlk = F_SETLK;
 208             fcntl_op_getlk = F_GETLK;
 209             return;
 210         }
 211         ret = fcntl(fd, F_OFD_GETLK, &fl);
 212         close(fd);
 213         if (!ret) {
 214             fcntl_op_setlk = F_OFD_SETLK;
 215             fcntl_op_getlk = F_OFD_GETLK;
 216         } else {
 217             fcntl_op_setlk = F_SETLK;
 218             fcntl_op_getlk = F_GETLK;
 219         }
 220 #else
 221         fcntl_op_setlk = F_SETLK;
 222         fcntl_op_getlk = F_GETLK;
 223 #endif
 224     }
 225 }
 226
 227 bool qemu_has_ofd_lock(void)
 228 {
 229     qemu_probe_lock_ops();
 230 #ifdef F_OFD_SETLK
 231     return fcntl_op_setlk == F_OFD_SETLK;
 232 #else
 233     return false;
 234 #endif
 235 }
 236
 237 static int qemu_lock_fcntl(int fd, int64_t start, int64_t len, int fl_type)
 238 {
 239     int ret;
 240     struct flock fl = {
 241         .l_whence = SEEK_SET,
 242         .l_start  = start,
 243         .l_len    = len,
 244         .l_type   = fl_type,
 245     };
 246     qemu_probe_lock_ops();
 247     ret = fcntl(fd, fcntl_op_setlk, &fl);
 248     return ret == -1 ? -errno : 0;
 249 }
 250
 251 int qemu_lock_fd(int fd, int64_t start, int64_t len, bool exclusive)
 252 {
 253     return qemu_lock_fcntl(fd, start, len, exclusive ? F_WRLCK : F_RDLCK);
 254 }
 255
 256 int qemu_unlock_fd(int fd, int64_t start, int64_t len)
 257 {
 258     return qemu_lock_fcntl(fd, start, len, F_UNLCK);
 259 }
 260
 261 int qemu_lock_fd_test(int fd, int64_t start, int64_t len, bool exclusive)
 262 {
 263     int ret;
 264     struct flock fl = {
 265         .l_whence = SEEK_SET,
 266         .l_start  = start,
 267         .l_len    = len,
 268         .l_type   = exclusive ? F_WRLCK : F_RDLCK,
 269     };
 270     qemu_probe_lock_ops();
 271     ret = fcntl(fd, fcntl_op_getlk, &fl);
 272     if (ret == -1) {
 273         return -errno;
 274     } else {
 275         return fl.l_type == F_UNLCK ? 0 : -EAGAIN;
 276     }
 277 }
 278 #endif
 279
 280 /*
 281  * Opens a file with FD_CLOEXEC set
 282  */
 283 int qemu_open(const char *name, int flags, ...)
 284 {
 285     int ret;
 286     int mode = 0;
 287
 288 #ifndef _WIN32
 289     const char *fdset_id_str;
 290
 291     /* Attempt dup of fd from fd set */
 292     if (strstart(name, "/dev/fdset/", &fdset_id_str)) {
 293         int64_t fdset_id;
 294         int fd, dupfd;
 295
 296         fdset_id = qemu_parse_fdset(fdset_id_str);
 297         if (fdset_id == -1) {
 298             errno = EINVAL;
 299             return -1;
 300         }
 301
 302         fd = monitor_fdset_get_fd(fdset_id, flags);
 303         if (fd == -1) {
 304             return -1;
 305         }
 306
 307         dupfd = qemu_dup_flags(fd, flags);
 308         if (dupfd == -1) {
 309             return -1;
 310         }
 311
 312         ret = monitor_fdset_dup_fd_add(fdset_id, dupfd);
 313         if (ret == -1) {
 314             close(dupfd);
 315             errno = EINVAL;
 316             return -1;
 317         }
 318
 319         return dupfd;
 320     }
 321 #endif
 322
 323     if (flags & O_CREAT) {
 324         va_list ap;
 325
 326         va_start(ap, flags);
 327         mode = va_arg(ap, int);
 328         va_end(ap);
 329     }
 330
 331 #ifdef O_CLOEXEC
 332     ret = open(name, flags | O_CLOEXEC, mode);
 333 #else
 334     ret = open(name, flags, mode);
 335     if (ret >= 0) {
 336         qemu_set_cloexec(ret);
 337     }
 338 #endif
 339
 340 #ifdef O_DIRECT
 341     if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
 342         error_report("file system may not support O_DIRECT");
 343         errno = EINVAL; /* in case it was clobbered */
 344     }
 345 #endif /* O_DIRECT */
 346
 347     return ret;
 348 }
 349
 350 int qemu_close(int fd)
 351 {
 352     int64_t fdset_id;
 353
 354     /* Close fd that was dup'd from an fdset */
 355     fdset_id = monitor_fdset_dup_fd_find(fd);
 356     if (fdset_id != -1) {
 357         int ret;
 358
 359         ret = close(fd);
 360         if (ret == 0) {
 361             monitor_fdset_dup_fd_remove(fd);
 362         }
 363
 364         return ret;
 365     }
 366
 367     return close(fd);
 368 }
 369
 370 /*
 371  * A variant of write(2) which handles partial write.
 372  *
 373  * Return the number of bytes transferred.
 374  * Set errno if fewer than `count' bytes are written.
 375  *
 376  * This function don't work with non-blocking fd's.
 377  * Any of the possibilities with non-bloking fd's is bad:
 378  *   - return a short write (then name is wrong)
 379  *   - busy wait adding (errno == EAGAIN) to the loop
 380  */
 381 ssize_t qemu_write_full(int fd, const void *buf, size_t count)
 382 {
 383     ssize_t ret = 0;
 384     ssize_t total = 0;
 385
 386     while (count) {
 387         ret = write(fd, buf, count);
 388         if (ret < 0) {
 389             if (errno == EINTR)
 390                 continue;
 391             break;
 392         }
 393
 394         count -= ret;
 395         buf += ret;
 396         total += ret;
 397     }
 398
 399     return total;
 400 }
 401
 402 /*
 403  * Opens a socket with FD_CLOEXEC set
 404  */
 405 int qemu_socket(int domain, int type, int protocol)
 406 {
 407     int ret;
 408
 409 #ifdef SOCK_CLOEXEC
 410     ret = socket(domain, type | SOCK_CLOEXEC, protocol);
 411     if (ret != -1 || errno != EINVAL) {
 412         return ret;
 413     }
 414 #endif
 415     ret = socket(domain, type, protocol);
 416     if (ret >= 0) {
 417         qemu_set_cloexec(ret);
 418     }
 419
 420     return ret;
 421 }
 422
 423 /*
 424  * Accept a connection and set FD_CLOEXEC
 425  */
 426 int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
 427 {
 428     int ret;
 429
 430 #ifdef CONFIG_ACCEPT4
 431     ret = accept4(s, addr, addrlen, SOCK_CLOEXEC);
 432     if (ret != -1 || errno != ENOSYS) {
 433         return ret;
 434     }
 435 #endif
 436     ret = accept(s, addr, addrlen);
 437     if (ret >= 0) {
 438         qemu_set_cloexec(ret);
 439     }
 440
 441     return ret;
 442 }
 443
 444 void qemu_set_hw_version(const char *version)
 445 {
 446     hw_version = version;
 447 }
 448
 449 const char *qemu_hw_version(void)
 450 {
 451     return hw_version;
 452 }
 453
 454 void fips_set_state(bool requested)
 455 {
 456 #ifdef __linux__
 457     if (requested) {
 458         FILE *fds = fopen("/proc/sys/crypto/fips_enabled", "r");
 459         if (fds != NULL) {
 460             fips_enabled = (fgetc(fds) == '1');
 461             fclose(fds);
 462         }
 463     }
 464 #else
 465     fips_enabled = false;
 466 #endif /* __linux__ */
 467
 468 #ifdef _FIPS_DEBUG
 469     fprintf(stderr, "FIPS mode %s (requested %s)\n",
 470             (fips_enabled ? "enabled" : "disabled"),
 471             (requested ? "enabled" : "disabled"));
 472 #endif
 473 }
 474
 475 bool fips_get_state(void)
 476 {
 477     return fips_enabled;
 478 }
 479
 480 #ifdef _WIN32
 481 static void socket_cleanup(void)
 482 {
 483     WSACleanup();
 484 }
 485 #endif
 486
 487 int socket_init(void)
 488 {
 489 #ifdef _WIN32
 490     WSADATA Data;
 491     int ret, err;
 492
 493     ret = WSAStartup(MAKEWORD(2, 2), &Data);
 494     if (ret != 0) {
 495         err = WSAGetLastError();
 496         fprintf(stderr, "WSAStartup: %d\n", err);
 497         return -1;
 498     }
 499     atexit(socket_cleanup);
 500 #endif
 501     return 0;
 502 }
 503
 504 #if !GLIB_CHECK_VERSION(2, 31, 0)
 505 /* Ensure that glib is running in multi-threaded mode
 506  * Old versions of glib require explicit initialization.  Failure to do
 507  * this results in the single-threaded code paths being taken inside
 508  * glib.  For example, the g_slice allocator will not be thread-safe
 509  * and cause crashes.
 510  */
 511 static void __attribute__((constructor)) thread_init(void)
 512 {
 513     if (!g_thread_supported()) {
 514        g_thread_init(NULL);
 515     }
 516 }
 517 #endif
 518
 519 #ifndef CONFIG_IOVEC
 520 /* helper function for iov_send_recv() */
 521 static ssize_t
 522 readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write)
 523 {
 524     unsigned i = 0;
 525     ssize_t ret = 0;
 526     while (i < iov_cnt) {
 527         ssize_t r = do_write
 528             ? write(fd, iov[i].iov_base, iov[i].iov_len)
 529             : read(fd, iov[i].iov_base, iov[i].iov_len);
 530         if (r > 0) {
 531             ret += r;
 532         } else if (!r) {
 533             break;
 534         } else if (errno == EINTR) {
 535             continue;
 536         } else {
 537             /* else it is some "other" error,
 538              * only return if there was no data processed. */
 539             if (ret == 0) {
 540                 ret = -1;
 541             }
 542             break;
 543         }
 544         i++;
 545     }
 546     return ret;
 547 }
 548
 549 ssize_t
 550 readv(int fd, const struct iovec *iov, int iov_cnt)
 551 {
 552     return readv_writev(fd, iov, iov_cnt, false);
 553 }
 554
 555 ssize_t
 556 writev(int fd, const struct iovec *iov, int iov_cnt)
 557 {
 558     return readv_writev(fd, iov, iov_cnt, true);
 559 }
 560 #endif