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