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 "monitor/monitor.h"
  51
  52 static bool fips_enabled = false;
  53
  54 static const char *qemu_version = QEMU_VERSION;
  55
  56 int socket_set_cork(int fd, int v)
  57 {
  58 #if defined(SOL_TCP) && defined(TCP_CORK)
  59     return qemu_setsockopt(fd, SOL_TCP, TCP_CORK, &v, sizeof(v));
  60 #else
  61     return 0;
  62 #endif
  63 }
  64
  65 int socket_set_nodelay(int fd)
  66 {
  67     int v = 1;
  68     return qemu_setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &v, sizeof(v));
  69 }
  70
  71 int qemu_madvise(void *addr, size_t len, int advice)
  72 {
  73     if (advice == QEMU_MADV_INVALID) {
  74         errno = EINVAL;
  75         return -1;
  76     }
  77 #if defined(CONFIG_MADVISE)
  78     return madvise(addr, len, advice);
  79 #elif defined(CONFIG_POSIX_MADVISE)
  80     return posix_madvise(addr, len, advice);
  81 #else
  82     errno = EINVAL;
  83     return -1;
  84 #endif
  85 }
  86
  87 #ifndef _WIN32
  88 /*
  89  * Dups an fd and sets the flags
  90  */
  91 static int qemu_dup_flags(int fd, int flags)
  92 {
  93     int ret;
  94     int serrno;
  95     int dup_flags;
  96
  97 #ifdef F_DUPFD_CLOEXEC
  98     ret = fcntl(fd, F_DUPFD_CLOEXEC, 0);
  99 #else
 100     ret = dup(fd);
 101     if (ret != -1) {
 102         qemu_set_cloexec(ret);
 103     }
 104 #endif
 105     if (ret == -1) {
 106         goto fail;
 107     }
 108
 109     dup_flags = fcntl(ret, F_GETFL);
 110     if (dup_flags == -1) {
 111         goto fail;
 112     }
 113
 114     if ((flags & O_SYNC) != (dup_flags & O_SYNC)) {
 115         errno = EINVAL;
 116         goto fail;
 117     }
 118
 119     /* Set/unset flags that we can with fcntl */
 120     if (fcntl(ret, F_SETFL, flags) == -1) {
 121         goto fail;
 122     }
 123
 124     /* Truncate the file in the cases that open() would truncate it */
 125     if (flags & O_TRUNC ||
 126             ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))) {
 127         if (ftruncate(ret, 0) == -1) {
 128             goto fail;
 129         }
 130     }
 131
 132     return ret;
 133
 134 fail:
 135     serrno = errno;
 136     if (ret != -1) {
 137         close(ret);
 138     }
 139     errno = serrno;
 140     return -1;
 141 }
 142
 143 static int qemu_parse_fdset(const char *param)
 144 {
 145     return qemu_parse_fd(param);
 146 }
 147 #endif
 148
 149 /*
 150  * Opens a file with FD_CLOEXEC set
 151  */
 152 int qemu_open(const char *name, int flags, ...)
 153 {
 154     int ret;
 155     int mode = 0;
 156
 157 #ifndef _WIN32
 158     const char *fdset_id_str;
 159
 160     /* Attempt dup of fd from fd set */
 161     if (strstart(name, "/dev/fdset/", &fdset_id_str)) {
 162         int64_t fdset_id;
 163         int fd, dupfd;
 164
 165         fdset_id = qemu_parse_fdset(fdset_id_str);
 166         if (fdset_id == -1) {
 167             errno = EINVAL;
 168             return -1;
 169         }
 170
 171         fd = monitor_fdset_get_fd(fdset_id, flags);
 172         if (fd == -1) {
 173             return -1;
 174         }
 175
 176         dupfd = qemu_dup_flags(fd, flags);
 177         if (dupfd == -1) {
 178             return -1;
 179         }
 180
 181         ret = monitor_fdset_dup_fd_add(fdset_id, dupfd);
 182         if (ret == -1) {
 183             close(dupfd);
 184             errno = EINVAL;
 185             return -1;
 186         }
 187
 188         return dupfd;
 189     }
 190 #endif
 191
 192     if (flags & O_CREAT) {
 193         va_list ap;
 194
 195         va_start(ap, flags);
 196         mode = va_arg(ap, int);
 197         va_end(ap);
 198     }
 199
 200 #ifdef O_CLOEXEC
 201     ret = open(name, flags | O_CLOEXEC, mode);
 202 #else
 203     ret = open(name, flags, mode);
 204     if (ret >= 0) {
 205         qemu_set_cloexec(ret);
 206     }
 207 #endif
 208
 209 #ifdef O_DIRECT
 210     if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
 211         error_report("file system may not support O_DIRECT");
 212         errno = EINVAL; /* in case it was clobbered */
 213     }
 214 #endif /* O_DIRECT */
 215
 216     return ret;
 217 }
 218
 219 int qemu_close(int fd)
 220 {
 221     int64_t fdset_id;
 222
 223     /* Close fd that was dup'd from an fdset */
 224     fdset_id = monitor_fdset_dup_fd_find(fd);
 225     if (fdset_id != -1) {
 226         int ret;
 227
 228         ret = close(fd);
 229         if (ret == 0) {
 230             monitor_fdset_dup_fd_remove(fd);
 231         }
 232
 233         return ret;
 234     }
 235
 236     return close(fd);
 237 }
 238
 239 /*
 240  * A variant of write(2) which handles partial write.
 241  *
 242  * Return the number of bytes transferred.
 243  * Set errno if fewer than `count' bytes are written.
 244  *
 245  * This function don't work with non-blocking fd's.
 246  * Any of the possibilities with non-bloking fd's is bad:
 247  *   - return a short write (then name is wrong)
 248  *   - busy wait adding (errno == EAGAIN) to the loop
 249  */
 250 ssize_t qemu_write_full(int fd, const void *buf, size_t count)
 251 {
 252     ssize_t ret = 0;
 253     ssize_t total = 0;
 254
 255     while (count) {
 256         ret = write(fd, buf, count);
 257         if (ret < 0) {
 258             if (errno == EINTR)
 259                 continue;
 260             break;
 261         }
 262
 263         count -= ret;
 264         buf += ret;
 265         total += ret;
 266     }
 267
 268     return total;
 269 }
 270
 271 /*
 272  * Opens a socket with FD_CLOEXEC set
 273  */
 274 int qemu_socket(int domain, int type, int protocol)
 275 {
 276     int ret;
 277
 278 #ifdef SOCK_CLOEXEC
 279     ret = socket(domain, type | SOCK_CLOEXEC, protocol);
 280     if (ret != -1 || errno != EINVAL) {
 281         return ret;
 282     }
 283 #endif
 284     ret = socket(domain, type, protocol);
 285     if (ret >= 0) {
 286         qemu_set_cloexec(ret);
 287     }
 288
 289     return ret;
 290 }
 291
 292 /*
 293  * Accept a connection and set FD_CLOEXEC
 294  */
 295 int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
 296 {
 297     int ret;
 298
 299 #ifdef CONFIG_ACCEPT4
 300     ret = accept4(s, addr, addrlen, SOCK_CLOEXEC);
 301     if (ret != -1 || errno != ENOSYS) {
 302         return ret;
 303     }
 304 #endif
 305     ret = accept(s, addr, addrlen);
 306     if (ret >= 0) {
 307         qemu_set_cloexec(ret);
 308     }
 309
 310     return ret;
 311 }
 312
 313 /*
 314  * A variant of send(2) which handles partial write.
 315  *
 316  * Return the number of bytes transferred, which is only
 317  * smaller than `count' if there is an error.
 318  *
 319  * This function won't work with non-blocking fd's.
 320  * Any of the possibilities with non-bloking fd's is bad:
 321  *   - return a short write (then name is wrong)
 322  *   - busy wait adding (errno == EAGAIN) to the loop
 323  */
 324 ssize_t qemu_send_full(int fd, const void *buf, size_t count, int flags)
 325 {
 326     ssize_t ret = 0;
 327     ssize_t total = 0;
 328
 329     while (count) {
 330         ret = send(fd, buf, count, flags);
 331         if (ret < 0) {
 332             if (errno == EINTR) {
 333                 continue;
 334             }
 335             break;
 336         }
 337
 338         count -= ret;
 339         buf += ret;
 340         total += ret;
 341     }
 342
 343     return total;
 344 }
 345
 346 /*
 347  * A variant of recv(2) which handles partial write.
 348  *
 349  * Return the number of bytes transferred, which is only
 350  * smaller than `count' if there is an error.
 351  *
 352  * This function won't work with non-blocking fd's.
 353  * Any of the possibilities with non-bloking fd's is bad:
 354  *   - return a short write (then name is wrong)
 355  *   - busy wait adding (errno == EAGAIN) to the loop
 356  */
 357 ssize_t qemu_recv_full(int fd, void *buf, size_t count, int flags)
 358 {
 359     ssize_t ret = 0;
 360     ssize_t total = 0;
 361
 362     while (count) {
 363         ret = qemu_recv(fd, buf, count, flags);
 364         if (ret <= 0) {
 365             if (ret < 0 && errno == EINTR) {
 366                 continue;
 367             }
 368             break;
 369         }
 370
 371         count -= ret;
 372         buf += ret;
 373         total += ret;
 374     }
 375
 376     return total;
 377 }
 378
 379 void qemu_set_version(const char *version)
 380 {
 381     qemu_version = version;
 382 }
 383
 384 const char *qemu_get_version(void)
 385 {
 386     return qemu_version;
 387 }
 388
 389 void fips_set_state(bool requested)
 390 {
 391 #ifdef __linux__
 392     if (requested) {
 393         FILE *fds = fopen("/proc/sys/crypto/fips_enabled", "r");
 394         if (fds != NULL) {
 395             fips_enabled = (fgetc(fds) == '1');
 396             fclose(fds);
 397         }
 398     }
 399 #else
 400     fips_enabled = false;
 401 #endif /* __linux__ */
 402
 403 #ifdef _FIPS_DEBUG
 404     fprintf(stderr, "FIPS mode %s (requested %s)\n",
 405             (fips_enabled ? "enabled" : "disabled"),
 406             (requested ? "enabled" : "disabled"));
 407 #endif
 408 }
 409
 410 bool fips_get_state(void)
 411 {
 412     return fips_enabled;
 413 }
 414
 415 #ifdef _WIN32
 416 static void socket_cleanup(void)
 417 {
 418     WSACleanup();
 419 }
 420 #endif
 421
 422 int socket_init(void)
 423 {
 424 #ifdef _WIN32
 425     WSADATA Data;
 426     int ret, err;
 427
 428     ret = WSAStartup(MAKEWORD(2, 2), &Data);
 429     if (ret != 0) {
 430         err = WSAGetLastError();
 431         fprintf(stderr, "WSAStartup: %d\n", err);
 432         return -1;
 433     }
 434     atexit(socket_cleanup);
 435 #endif
 436     return 0;
 437 }
 438
 439 #ifndef CONFIG_IOVEC
 440 /* helper function for iov_send_recv() */
 441 static ssize_t
 442 readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write)
 443 {
 444     unsigned i = 0;
 445     ssize_t ret = 0;
 446     while (i < iov_cnt) {
 447         ssize_t r = do_write
 448             ? write(fd, iov[i].iov_base, iov[i].iov_len)
 449             : read(fd, iov[i].iov_base, iov[i].iov_len);
 450         if (r > 0) {
 451             ret += r;
 452         } else if (!r) {
 453             break;
 454         } else if (errno == EINTR) {
 455             continue;
 456         } else {
 457             /* else it is some "other" error,
 458              * only return if there was no data processed. */
 459             if (ret == 0) {
 460                 ret = -1;
 461             }
 462             break;
 463         }
 464         i++;
 465     }
 466     return ret;
 467 }
 468
 469 ssize_t
 470 readv(int fd, const struct iovec *iov, int iov_cnt)
 471 {
 472     return readv_writev(fd, iov, iov_cnt, false);
 473 }
 474
 475 ssize_t
 476 writev(int fd, const struct iovec *iov, int iov_cnt)
 477 {
 478     return readv_writev(fd, iov, iov_cnt, true);
 479 }
 480 #endif