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