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 <stdlib.h>
  25 #include <stdio.h>
  26 #include <stdarg.h>
  27 #include <stdbool.h>
  28 #include <string.h>
  29 #include <errno.h>
  30 #include <unistd.h>
  31 #include <fcntl.h>
  32
  33 /* Needed early for CONFIG_BSD etc. */
  34 #include "config-host.h"
  35
  36 #if defined(CONFIG_MADVISE) || defined(CONFIG_POSIX_MADVISE)
  37 #include <sys/mman.h>
  38 #endif
  39
  40 #ifdef CONFIG_SOLARIS
  41 #include <sys/types.h>
  42 #include <sys/statvfs.h>
  43 /* See MySQL bug #7156 (http://bugs.mysql.com/bug.php?id=7156) for
  44    discussion about Solaris header problems */
  45 extern int madvise(caddr_t, size_t, int);
  46 #endif
  47
  48 #include "qemu-common.h"
  49 #include "qemu/sockets.h"
  50 #include "qemu/error-report.h"
  51 #include "monitor/monitor.h"
  52
  53 static bool fips_enabled = false;
  54
  55 /* Starting on QEMU 2.5, qemu_hw_version() returns "2.5+" by default
  56  * instead of QEMU_VERSION, so setting hw_version on MachineClass
  57  * is no longer mandatory.
  58  *
  59  * Do NOT change this string, or it will break compatibility on all
  60  * machine classes that don't set hw_version.
  61  */
  62 static const char *hw_version = "2.5+";
  63
  64 int socket_set_cork(int fd, int v)
  65 {
  66 #if defined(SOL_TCP) && defined(TCP_CORK)
  67     return qemu_setsockopt(fd, SOL_TCP, TCP_CORK, &v, sizeof(v));
  68 #else
  69     return 0;
  70 #endif
  71 }
  72
  73 int socket_set_nodelay(int fd)
  74 {
  75     int v = 1;
  76     return qemu_setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &v, sizeof(v));
  77 }
  78
  79 int qemu_madvise(void *addr, size_t len, int advice)
  80 {
  81     if (advice == QEMU_MADV_INVALID) {
  82         errno = EINVAL;
  83         return -1;
  84     }
  85 #if defined(CONFIG_MADVISE)
  86     return madvise(addr, len, advice);
  87 #elif defined(CONFIG_POSIX_MADVISE)
  88     return posix_madvise(addr, len, advice);
  89 #else
  90     errno = EINVAL;
  91     return -1;
  92 #endif
  93 }
  94
  95 #ifndef _WIN32
  96 /*
  97  * Dups an fd and sets the flags
  98  */
  99 static int qemu_dup_flags(int fd, int flags)
 100 {
 101     int ret;
 102     int serrno;
 103     int dup_flags;
 104
 105 #ifdef F_DUPFD_CLOEXEC
 106     ret = fcntl(fd, F_DUPFD_CLOEXEC, 0);
 107 #else
 108     ret = dup(fd);
 109     if (ret != -1) {
 110         qemu_set_cloexec(ret);
 111     }
 112 #endif
 113     if (ret == -1) {
 114         goto fail;
 115     }
 116
 117     dup_flags = fcntl(ret, F_GETFL);
 118     if (dup_flags == -1) {
 119         goto fail;
 120     }
 121
 122     if ((flags & O_SYNC) != (dup_flags & O_SYNC)) {
 123         errno = EINVAL;
 124         goto fail;
 125     }
 126
 127     /* Set/unset flags that we can with fcntl */
 128     if (fcntl(ret, F_SETFL, flags) == -1) {
 129         goto fail;
 130     }
 131
 132     /* Truncate the file in the cases that open() would truncate it */
 133     if (flags & O_TRUNC ||
 134             ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))) {
 135         if (ftruncate(ret, 0) == -1) {
 136             goto fail;
 137         }
 138     }
 139
 140     return ret;
 141
 142 fail:
 143     serrno = errno;
 144     if (ret != -1) {
 145         close(ret);
 146     }
 147     errno = serrno;
 148     return -1;
 149 }
 150
 151 static int qemu_parse_fdset(const char *param)
 152 {
 153     return qemu_parse_fd(param);
 154 }
 155 #endif
 156
 157 /*
 158  * Opens a file with FD_CLOEXEC set
 159  */
 160 int qemu_open(const char *name, int flags, ...)
 161 {
 162     int ret;
 163     int mode = 0;
 164
 165 #ifndef _WIN32
 166     const char *fdset_id_str;
 167
 168     /* Attempt dup of fd from fd set */
 169     if (strstart(name, "/dev/fdset/", &fdset_id_str)) {
 170         int64_t fdset_id;
 171         int fd, dupfd;
 172
 173         fdset_id = qemu_parse_fdset(fdset_id_str);
 174         if (fdset_id == -1) {
 175             errno = EINVAL;
 176             return -1;
 177         }
 178
 179         fd = monitor_fdset_get_fd(fdset_id, flags);
 180         if (fd == -1) {
 181             return -1;
 182         }
 183
 184         dupfd = qemu_dup_flags(fd, flags);
 185         if (dupfd == -1) {
 186             return -1;
 187         }
 188
 189         ret = monitor_fdset_dup_fd_add(fdset_id, dupfd);
 190         if (ret == -1) {
 191             close(dupfd);
 192             errno = EINVAL;
 193             return -1;
 194         }
 195
 196         return dupfd;
 197     }
 198 #endif
 199
 200     if (flags & O_CREAT) {
 201         va_list ap;
 202
 203         va_start(ap, flags);
 204         mode = va_arg(ap, int);
 205         va_end(ap);
 206     }
 207
 208 #ifdef O_CLOEXEC
 209     ret = open(name, flags | O_CLOEXEC, mode);
 210 #else
 211     ret = open(name, flags, mode);
 212     if (ret >= 0) {
 213         qemu_set_cloexec(ret);
 214     }
 215 #endif
 216
 217 #ifdef O_DIRECT
 218     if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
 219         error_report("file system may not support O_DIRECT");
 220         errno = EINVAL; /* in case it was clobbered */
 221     }
 222 #endif /* O_DIRECT */
 223
 224     return ret;
 225 }
 226
 227 int qemu_close(int fd)
 228 {
 229     int64_t fdset_id;
 230
 231     /* Close fd that was dup'd from an fdset */
 232     fdset_id = monitor_fdset_dup_fd_find(fd);
 233     if (fdset_id != -1) {
 234         int ret;
 235
 236         ret = close(fd);
 237         if (ret == 0) {
 238             monitor_fdset_dup_fd_remove(fd);
 239         }
 240
 241         return ret;
 242     }
 243
 244     return close(fd);
 245 }
 246
 247 /*
 248  * A variant of write(2) which handles partial write.
 249  *
 250  * Return the number of bytes transferred.
 251  * Set errno if fewer than `count' bytes are written.
 252  *
 253  * This function don't work with non-blocking fd's.
 254  * Any of the possibilities with non-bloking fd's is bad:
 255  *   - return a short write (then name is wrong)
 256  *   - busy wait adding (errno == EAGAIN) to the loop
 257  */
 258 ssize_t qemu_write_full(int fd, const void *buf, size_t count)
 259 {
 260     ssize_t ret = 0;
 261     ssize_t total = 0;
 262
 263     while (count) {
 264         ret = write(fd, buf, count);
 265         if (ret < 0) {
 266             if (errno == EINTR)
 267                 continue;
 268             break;
 269         }
 270
 271         count -= ret;
 272         buf += ret;
 273         total += ret;
 274     }
 275
 276     return total;
 277 }
 278
 279 /*
 280  * Opens a socket with FD_CLOEXEC set
 281  */
 282 int qemu_socket(int domain, int type, int protocol)
 283 {
 284     int ret;
 285
 286 #ifdef SOCK_CLOEXEC
 287     ret = socket(domain, type | SOCK_CLOEXEC, protocol);
 288     if (ret != -1 || errno != EINVAL) {
 289         return ret;
 290     }
 291 #endif
 292     ret = socket(domain, type, protocol);
 293     if (ret >= 0) {
 294         qemu_set_cloexec(ret);
 295     }
 296
 297     return ret;
 298 }
 299
 300 /*
 301  * Accept a connection and set FD_CLOEXEC
 302  */
 303 int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
 304 {
 305     int ret;
 306
 307 #ifdef CONFIG_ACCEPT4
 308     ret = accept4(s, addr, addrlen, SOCK_CLOEXEC);
 309     if (ret != -1 || errno != ENOSYS) {
 310         return ret;
 311     }
 312 #endif
 313     ret = accept(s, addr, addrlen);
 314     if (ret >= 0) {
 315         qemu_set_cloexec(ret);
 316     }
 317
 318     return ret;
 319 }
 320
 321 void qemu_set_hw_version(const char *version)
 322 {
 323     hw_version = version;
 324 }
 325
 326 const char *qemu_hw_version(void)
 327 {
 328     return hw_version;
 329 }
 330
 331 void fips_set_state(bool requested)
 332 {
 333 #ifdef __linux__
 334     if (requested) {
 335         FILE *fds = fopen("/proc/sys/crypto/fips_enabled", "r");
 336         if (fds != NULL) {
 337             fips_enabled = (fgetc(fds) == '1');
 338             fclose(fds);
 339         }
 340     }
 341 #else
 342     fips_enabled = false;
 343 #endif /* __linux__ */
 344
 345 #ifdef _FIPS_DEBUG
 346     fprintf(stderr, "FIPS mode %s (requested %s)\n",
 347             (fips_enabled ? "enabled" : "disabled"),
 348             (requested ? "enabled" : "disabled"));
 349 #endif
 350 }
 351
 352 bool fips_get_state(void)
 353 {
 354     return fips_enabled;
 355 }
 356
 357 #ifdef _WIN32
 358 static void socket_cleanup(void)
 359 {
 360     WSACleanup();
 361 }
 362 #endif
 363
 364 int socket_init(void)
 365 {
 366 #ifdef _WIN32
 367     WSADATA Data;
 368     int ret, err;
 369
 370     ret = WSAStartup(MAKEWORD(2, 2), &Data);
 371     if (ret != 0) {
 372         err = WSAGetLastError();
 373         fprintf(stderr, "WSAStartup: %d\n", err);
 374         return -1;
 375     }
 376     atexit(socket_cleanup);
 377 #endif
 378     return 0;
 379 }
 380
 381 #if !GLIB_CHECK_VERSION(2, 31, 0)
 382 /* Ensure that glib is running in multi-threaded mode
 383  * Old versions of glib require explicit initialization.  Failure to do
 384  * this results in the single-threaded code paths being taken inside
 385  * glib.  For example, the g_slice allocator will not be thread-safe
 386  * and cause crashes.
 387  */
 388 static void __attribute__((constructor)) thread_init(void)
 389 {
 390     if (!g_thread_supported()) {
 391        g_thread_init(NULL);
 392     }
 393 }
 394 #endif
 395
 396 #ifndef CONFIG_IOVEC
 397 /* helper function for iov_send_recv() */
 398 static ssize_t
 399 readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write)
 400 {
 401     unsigned i = 0;
 402     ssize_t ret = 0;
 403     while (i < iov_cnt) {
 404         ssize_t r = do_write
 405             ? write(fd, iov[i].iov_base, iov[i].iov_len)
 406             : read(fd, iov[i].iov_base, iov[i].iov_len);
 407         if (r > 0) {
 408             ret += r;
 409         } else if (!r) {
 410             break;
 411         } else if (errno == EINTR) {
 412             continue;
 413         } else {
 414             /* else it is some "other" error,
 415              * only return if there was no data processed. */
 416             if (ret == 0) {
 417                 ret = -1;
 418             }
 419             break;
 420         }
 421         i++;
 422     }
 423     return ret;
 424 }
 425
 426 ssize_t
 427 readv(int fd, const struct iovec *iov, int iov_cnt)
 428 {
 429     return readv_writev(fd, iov, iov_cnt, false);
 430 }
 431
 432 ssize_t
 433 writev(int fd, const struct iovec *iov, int iov_cnt)
 434 {
 435     return readv_writev(fd, iov, iov_cnt, true);
 436 }
 437 #endif