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 #if defined(CONFIG_MADVISE) || defined(CONFIG_POSIX_MADVISE)
  29 #include <sys/mman.h>
  30 #endif
  31
  32 #ifdef CONFIG_SOLARIS
  33 #include <sys/statvfs.h>
  34 /* See MySQL bug #7156 (http://bugs.mysql.com/bug.php?id=7156) for
  35    discussion about Solaris header problems */
  36 extern int madvise(caddr_t, size_t, int);
  37 #endif
  38
  39 #include "qemu-common.h"
  40 #include "qemu/cutils.h"
  41 #include "qemu/sockets.h"
  42 #include "qemu/error-report.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 #ifndef _WIN32
  81 /*
  82  * Dups an fd and sets the flags
  83  */
  84 static int qemu_dup_flags(int fd, int flags)
  85 {
  86     int ret;
  87     int serrno;
  88     int dup_flags;
  89
  90 #ifdef F_DUPFD_CLOEXEC
  91     ret = fcntl(fd, F_DUPFD_CLOEXEC, 0);
  92 #else
  93     ret = dup(fd);
  94     if (ret != -1) {
  95         qemu_set_cloexec(ret);
  96     }
  97 #endif
  98     if (ret == -1) {
  99         goto fail;
 100     }
 101
 102     dup_flags = fcntl(ret, F_GETFL);
 103     if (dup_flags == -1) {
 104         goto fail;
 105     }
 106
 107     if ((flags & O_SYNC) != (dup_flags & O_SYNC)) {
 108         errno = EINVAL;
 109         goto fail;
 110     }
 111
 112     /* Set/unset flags that we can with fcntl */
 113     if (fcntl(ret, F_SETFL, flags) == -1) {
 114         goto fail;
 115     }
 116
 117     /* Truncate the file in the cases that open() would truncate it */
 118     if (flags & O_TRUNC ||
 119             ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))) {
 120         if (ftruncate(ret, 0) == -1) {
 121             goto fail;
 122         }
 123     }
 124
 125     return ret;
 126
 127 fail:
 128     serrno = errno;
 129     if (ret != -1) {
 130         close(ret);
 131     }
 132     errno = serrno;
 133     return -1;
 134 }
 135
 136 static int qemu_parse_fdset(const char *param)
 137 {
 138     return qemu_parse_fd(param);
 139 }
 140 #endif
 141
 142 /*
 143  * Opens a file with FD_CLOEXEC set
 144  */
 145 int qemu_open(const char *name, int flags, ...)
 146 {
 147     int ret;
 148     int mode = 0;
 149
 150 #ifndef _WIN32
 151     const char *fdset_id_str;
 152
 153     /* Attempt dup of fd from fd set */
 154     if (strstart(name, "/dev/fdset/", &fdset_id_str)) {
 155         int64_t fdset_id;
 156         int fd, dupfd;
 157
 158         fdset_id = qemu_parse_fdset(fdset_id_str);
 159         if (fdset_id == -1) {
 160             errno = EINVAL;
 161             return -1;
 162         }
 163
 164         fd = monitor_fdset_get_fd(fdset_id, flags);
 165         if (fd == -1) {
 166             return -1;
 167         }
 168
 169         dupfd = qemu_dup_flags(fd, flags);
 170         if (dupfd == -1) {
 171             return -1;
 172         }
 173
 174         ret = monitor_fdset_dup_fd_add(fdset_id, dupfd);
 175         if (ret == -1) {
 176             close(dupfd);
 177             errno = EINVAL;
 178             return -1;
 179         }
 180
 181         return dupfd;
 182     }
 183 #endif
 184
 185     if (flags & O_CREAT) {
 186         va_list ap;
 187
 188         va_start(ap, flags);
 189         mode = va_arg(ap, int);
 190         va_end(ap);
 191     }
 192
 193 #ifdef O_CLOEXEC
 194     ret = open(name, flags | O_CLOEXEC, mode);
 195 #else
 196     ret = open(name, flags, mode);
 197     if (ret >= 0) {
 198         qemu_set_cloexec(ret);
 199     }
 200 #endif
 201
 202 #ifdef O_DIRECT
 203     if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
 204         error_report("file system may not support O_DIRECT");
 205         errno = EINVAL; /* in case it was clobbered */
 206     }
 207 #endif /* O_DIRECT */
 208
 209     return ret;
 210 }
 211
 212 int qemu_close(int fd)
 213 {
 214     int64_t fdset_id;
 215
 216     /* Close fd that was dup'd from an fdset */
 217     fdset_id = monitor_fdset_dup_fd_find(fd);
 218     if (fdset_id != -1) {
 219         int ret;
 220
 221         ret = close(fd);
 222         if (ret == 0) {
 223             monitor_fdset_dup_fd_remove(fd);
 224         }
 225
 226         return ret;
 227     }
 228
 229     return close(fd);
 230 }
 231
 232 /*
 233  * A variant of write(2) which handles partial write.
 234  *
 235  * Return the number of bytes transferred.
 236  * Set errno if fewer than `count' bytes are written.
 237  *
 238  * This function don't work with non-blocking fd's.
 239  * Any of the possibilities with non-bloking fd's is bad:
 240  *   - return a short write (then name is wrong)
 241  *   - busy wait adding (errno == EAGAIN) to the loop
 242  */
 243 ssize_t qemu_write_full(int fd, const void *buf, size_t count)
 244 {
 245     ssize_t ret = 0;
 246     ssize_t total = 0;
 247
 248     while (count) {
 249         ret = write(fd, buf, count);
 250         if (ret < 0) {
 251             if (errno == EINTR)
 252                 continue;
 253             break;
 254         }
 255
 256         count -= ret;
 257         buf += ret;
 258         total += ret;
 259     }
 260
 261     return total;
 262 }
 263
 264 /*
 265  * Opens a socket with FD_CLOEXEC set
 266  */
 267 int qemu_socket(int domain, int type, int protocol)
 268 {
 269     int ret;
 270
 271 #ifdef SOCK_CLOEXEC
 272     ret = socket(domain, type | SOCK_CLOEXEC, protocol);
 273     if (ret != -1 || errno != EINVAL) {
 274         return ret;
 275     }
 276 #endif
 277     ret = socket(domain, type, protocol);
 278     if (ret >= 0) {
 279         qemu_set_cloexec(ret);
 280     }
 281
 282     return ret;
 283 }
 284
 285 /*
 286  * Accept a connection and set FD_CLOEXEC
 287  */
 288 int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
 289 {
 290     int ret;
 291
 292 #ifdef CONFIG_ACCEPT4
 293     ret = accept4(s, addr, addrlen, SOCK_CLOEXEC);
 294     if (ret != -1 || errno != ENOSYS) {
 295         return ret;
 296     }
 297 #endif
 298     ret = accept(s, addr, addrlen);
 299     if (ret >= 0) {
 300         qemu_set_cloexec(ret);
 301     }
 302
 303     return ret;
 304 }
 305
 306 void qemu_set_hw_version(const char *version)
 307 {
 308     hw_version = version;
 309 }
 310
 311 const char *qemu_hw_version(void)
 312 {
 313     return hw_version;
 314 }
 315
 316 void fips_set_state(bool requested)
 317 {
 318 #ifdef __linux__
 319     if (requested) {
 320         FILE *fds = fopen("/proc/sys/crypto/fips_enabled", "r");
 321         if (fds != NULL) {
 322             fips_enabled = (fgetc(fds) == '1');
 323             fclose(fds);
 324         }
 325     }
 326 #else
 327     fips_enabled = false;
 328 #endif /* __linux__ */
 329
 330 #ifdef _FIPS_DEBUG
 331     fprintf(stderr, "FIPS mode %s (requested %s)\n",
 332             (fips_enabled ? "enabled" : "disabled"),
 333             (requested ? "enabled" : "disabled"));
 334 #endif
 335 }
 336
 337 bool fips_get_state(void)
 338 {
 339     return fips_enabled;
 340 }
 341
 342 #ifdef _WIN32
 343 static void socket_cleanup(void)
 344 {
 345     WSACleanup();
 346 }
 347 #endif
 348
 349 int socket_init(void)
 350 {
 351 #ifdef _WIN32
 352     WSADATA Data;
 353     int ret, err;
 354
 355     ret = WSAStartup(MAKEWORD(2, 2), &Data);
 356     if (ret != 0) {
 357         err = WSAGetLastError();
 358         fprintf(stderr, "WSAStartup: %d\n", err);
 359         return -1;
 360     }
 361     atexit(socket_cleanup);
 362 #endif
 363     return 0;
 364 }
 365
 366 #if !GLIB_CHECK_VERSION(2, 31, 0)
 367 /* Ensure that glib is running in multi-threaded mode
 368  * Old versions of glib require explicit initialization.  Failure to do
 369  * this results in the single-threaded code paths being taken inside
 370  * glib.  For example, the g_slice allocator will not be thread-safe
 371  * and cause crashes.
 372  */
 373 static void __attribute__((constructor)) thread_init(void)
 374 {
 375     if (!g_thread_supported()) {
 376        g_thread_init(NULL);
 377     }
 378 }
 379 #endif
 380
 381 #ifndef CONFIG_IOVEC
 382 /* helper function for iov_send_recv() */
 383 static ssize_t
 384 readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write)
 385 {
 386     unsigned i = 0;
 387     ssize_t ret = 0;
 388     while (i < iov_cnt) {
 389         ssize_t r = do_write
 390             ? write(fd, iov[i].iov_base, iov[i].iov_len)
 391             : read(fd, iov[i].iov_base, iov[i].iov_len);
 392         if (r > 0) {
 393             ret += r;
 394         } else if (!r) {
 395             break;
 396         } else if (errno == EINTR) {
 397             continue;
 398         } else {
 399             /* else it is some "other" error,
 400              * only return if there was no data processed. */
 401             if (ret == 0) {
 402                 ret = -1;
 403             }
 404             break;
 405         }
 406         i++;
 407     }
 408     return ret;
 409 }
 410
 411 ssize_t
 412 readv(int fd, const struct iovec *iov, int iov_cnt)
 413 {
 414     return readv_writev(fd, iov, iov_cnt, false);
 415 }
 416
 417 ssize_t
 418 writev(int fd, const struct iovec *iov, int iov_cnt)
 419 {
 420     return readv_writev(fd, iov, iov_cnt, true);
 421 }
 422 #endif