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 #ifndef _WIN32
  77 /*
  78  * Dups an fd and sets the flags
  79  */
  80 static int qemu_dup_flags(int fd, int flags)
  81 {
  82     int ret;
  83     int serrno;
  84     int dup_flags;
  85
  86     ret = qemu_dup(fd);
  87     if (ret == -1) {
  88         goto fail;
  89     }
  90
  91     dup_flags = fcntl(ret, F_GETFL);
  92     if (dup_flags == -1) {
  93         goto fail;
  94     }
  95
  96     if ((flags & O_SYNC) != (dup_flags & O_SYNC)) {
  97         errno = EINVAL;
  98         goto fail;
  99     }
 100
 101     /* Set/unset flags that we can with fcntl */
 102     if (fcntl(ret, F_SETFL, flags) == -1) {
 103         goto fail;
 104     }
 105
 106     /* Truncate the file in the cases that open() would truncate it */
 107     if (flags & O_TRUNC ||
 108             ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))) {
 109         if (ftruncate(ret, 0) == -1) {
 110             goto fail;
 111         }
 112     }
 113
 114     return ret;
 115
 116 fail:
 117     serrno = errno;
 118     if (ret != -1) {
 119         close(ret);
 120     }
 121     errno = serrno;
 122     return -1;
 123 }
 124
 125 int qemu_dup(int fd)
 126 {
 127     int ret;
 128 #ifdef F_DUPFD_CLOEXEC
 129     ret = fcntl(fd, F_DUPFD_CLOEXEC, 0);
 130 #else
 131     ret = dup(fd);
 132     if (ret != -1) {
 133         qemu_set_cloexec(ret);
 134     }
 135 #endif
 136     return ret;
 137 }
 138
 139 static int qemu_parse_fdset(const char *param)
 140 {
 141     return qemu_parse_fd(param);
 142 }
 143 #endif
 144
 145 /*
 146  * Opens a file with FD_CLOEXEC set
 147  */
 148 int qemu_open(const char *name, int flags, ...)
 149 {
 150     int ret;
 151     int mode = 0;
 152
 153 #ifndef _WIN32
 154     const char *fdset_id_str;
 155
 156     /* Attempt dup of fd from fd set */
 157     if (strstart(name, "/dev/fdset/", &fdset_id_str)) {
 158         int64_t fdset_id;
 159         int fd, dupfd;
 160
 161         fdset_id = qemu_parse_fdset(fdset_id_str);
 162         if (fdset_id == -1) {
 163             errno = EINVAL;
 164             return -1;
 165         }
 166
 167         fd = monitor_fdset_get_fd(fdset_id, flags);
 168         if (fd == -1) {
 169             return -1;
 170         }
 171
 172         dupfd = qemu_dup_flags(fd, flags);
 173         if (dupfd == -1) {
 174             return -1;
 175         }
 176
 177         ret = monitor_fdset_dup_fd_add(fdset_id, dupfd);
 178         if (ret == -1) {
 179             close(dupfd);
 180             errno = EINVAL;
 181             return -1;
 182         }
 183
 184         return dupfd;
 185     }
 186 #endif
 187
 188     if (flags & O_CREAT) {
 189         va_list ap;
 190
 191         va_start(ap, flags);
 192         mode = va_arg(ap, int);
 193         va_end(ap);
 194     }
 195
 196 #ifdef O_CLOEXEC
 197     ret = open(name, flags | O_CLOEXEC, mode);
 198 #else
 199     ret = open(name, flags, mode);
 200     if (ret >= 0) {
 201         qemu_set_cloexec(ret);
 202     }
 203 #endif
 204
 205 #ifdef O_DIRECT
 206     if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
 207         error_report("file system may not support O_DIRECT");
 208         errno = EINVAL; /* in case it was clobbered */
 209     }
 210 #endif /* O_DIRECT */
 211
 212     return ret;
 213 }
 214
 215 int qemu_close(int fd)
 216 {
 217     int64_t fdset_id;
 218
 219     /* Close fd that was dup'd from an fdset */
 220     fdset_id = monitor_fdset_dup_fd_find(fd);
 221     if (fdset_id != -1) {
 222         int ret;
 223
 224         ret = close(fd);
 225         if (ret == 0) {
 226             monitor_fdset_dup_fd_remove(fd);
 227         }
 228
 229         return ret;
 230     }
 231
 232     return close(fd);
 233 }
 234
 235 /*
 236  * A variant of write(2) which handles partial write.
 237  *
 238  * Return the number of bytes transferred.
 239  * Set errno if fewer than `count' bytes are written.
 240  *
 241  * This function don't work with non-blocking fd's.
 242  * Any of the possibilities with non-bloking fd's is bad:
 243  *   - return a short write (then name is wrong)
 244  *   - busy wait adding (errno == EAGAIN) to the loop
 245  */
 246 ssize_t qemu_write_full(int fd, const void *buf, size_t count)
 247 {
 248     ssize_t ret = 0;
 249     ssize_t total = 0;
 250
 251     while (count) {
 252         ret = write(fd, buf, count);
 253         if (ret < 0) {
 254             if (errno == EINTR)
 255                 continue;
 256             break;
 257         }
 258
 259         count -= ret;
 260         buf += ret;
 261         total += ret;
 262     }
 263
 264     return total;
 265 }
 266
 267 /*
 268  * Opens a socket with FD_CLOEXEC set
 269  */
 270 int qemu_socket(int domain, int type, int protocol)
 271 {
 272     int ret;
 273
 274 #ifdef SOCK_CLOEXEC
 275     ret = socket(domain, type | SOCK_CLOEXEC, protocol);
 276     if (ret != -1 || errno != EINVAL) {
 277         return ret;
 278     }
 279 #endif
 280     ret = socket(domain, type, protocol);
 281     if (ret >= 0) {
 282         qemu_set_cloexec(ret);
 283     }
 284
 285     return ret;
 286 }
 287
 288 /*
 289  * Accept a connection and set FD_CLOEXEC
 290  */
 291 int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
 292 {
 293     int ret;
 294
 295 #ifdef CONFIG_ACCEPT4
 296     ret = accept4(s, addr, addrlen, SOCK_CLOEXEC);
 297     if (ret != -1 || errno != ENOSYS) {
 298         return ret;
 299     }
 300 #endif
 301     ret = accept(s, addr, addrlen);
 302     if (ret >= 0) {
 303         qemu_set_cloexec(ret);
 304     }
 305
 306     return ret;
 307 }
 308
 309 void qemu_set_hw_version(const char *version)
 310 {
 311     hw_version = version;
 312 }
 313
 314 const char *qemu_hw_version(void)
 315 {
 316     return hw_version;
 317 }
 318
 319 void fips_set_state(bool requested)
 320 {
 321 #ifdef __linux__
 322     if (requested) {
 323         FILE *fds = fopen("/proc/sys/crypto/fips_enabled", "r");
 324         if (fds != NULL) {
 325             fips_enabled = (fgetc(fds) == '1');
 326             fclose(fds);
 327         }
 328     }
 329 #else
 330     fips_enabled = false;
 331 #endif /* __linux__ */
 332
 333 #ifdef _FIPS_DEBUG
 334     fprintf(stderr, "FIPS mode %s (requested %s)\n",
 335             (fips_enabled ? "enabled" : "disabled"),
 336             (requested ? "enabled" : "disabled"));
 337 #endif
 338 }
 339
 340 bool fips_get_state(void)
 341 {
 342     return fips_enabled;
 343 }
 344
 345 #ifdef _WIN32
 346 static void socket_cleanup(void)
 347 {
 348     WSACleanup();
 349 }
 350 #endif
 351
 352 int socket_init(void)
 353 {
 354 #ifdef _WIN32
 355     WSADATA Data;
 356     int ret, err;
 357
 358     ret = WSAStartup(MAKEWORD(2, 2), &Data);
 359     if (ret != 0) {
 360         err = WSAGetLastError();
 361         fprintf(stderr, "WSAStartup: %d\n", err);
 362         return -1;
 363     }
 364     atexit(socket_cleanup);
 365 #endif
 366     return 0;
 367 }
 368
 369 #if !GLIB_CHECK_VERSION(2, 31, 0)
 370 /* Ensure that glib is running in multi-threaded mode
 371  * Old versions of glib require explicit initialization.  Failure to do
 372  * this results in the single-threaded code paths being taken inside
 373  * glib.  For example, the g_slice allocator will not be thread-safe
 374  * and cause crashes.
 375  */
 376 static void __attribute__((constructor)) thread_init(void)
 377 {
 378     if (!g_thread_supported()) {
 379        g_thread_init(NULL);
 380     }
 381 }
 382 #endif
 383
 384 #ifndef CONFIG_IOVEC
 385 /* helper function for iov_send_recv() */
 386 static ssize_t
 387 readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write)
 388 {
 389     unsigned i = 0;
 390     ssize_t ret = 0;
 391     while (i < iov_cnt) {
 392         ssize_t r = do_write
 393             ? write(fd, iov[i].iov_base, iov[i].iov_len)
 394             : read(fd, iov[i].iov_base, iov[i].iov_len);
 395         if (r > 0) {
 396             ret += r;
 397         } else if (!r) {
 398             break;
 399         } else if (errno == EINTR) {
 400             continue;
 401         } else {
 402             /* else it is some "other" error,
 403              * only return if there was no data processed. */
 404             if (ret == 0) {
 405                 ret = -1;
 406             }
 407             break;
 408         }
 409         i++;
 410     }
 411     return ret;
 412 }
 413
 414 ssize_t
 415 readv(int fd, const struct iovec *iov, int iov_cnt)
 416 {
 417     return readv_writev(fd, iov, iov_cnt, false);
 418 }
 419
 420 ssize_t
 421 writev(int fd, const struct iovec *iov, int iov_cnt)
 422 {
 423     return readv_writev(fd, iov, iov_cnt, true);
 424 }
 425 #endif