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