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 #include "qapi/error.h"
  26 #include "qemu/cutils.h"
  27 #include "qemu/sockets.h"
  28 #include "qemu/error-report.h"
  29 #include "qemu/madvise.h"
  30 #include "qemu/mprotect.h"
  31 #include "qemu/hw-version.h"
  32 #include "monitor/monitor.h"
  33
  34 static const char *hw_version = QEMU_HW_VERSION;
  35
  36 int socket_set_cork(int fd, int v)
  37 {
  38 #if defined(SOL_TCP) && defined(TCP_CORK)
  39     return setsockopt(fd, SOL_TCP, TCP_CORK, &v, sizeof(v));
  40 #else
  41     return 0;
  42 #endif
  43 }
  44
  45 int socket_set_nodelay(int fd)
  46 {
  47     int v = 1;
  48     return setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &v, sizeof(v));
  49 }
  50
  51 int qemu_madvise(void *addr, size_t len, int advice)
  52 {
  53     if (advice == QEMU_MADV_INVALID) {
  54         errno = EINVAL;
  55         return -1;
  56     }
  57 #if defined(CONFIG_MADVISE)
  58     return madvise(addr, len, advice);
  59 #elif defined(CONFIG_POSIX_MADVISE)
  60     return posix_madvise(addr, len, advice);
  61 #else
  62     errno = EINVAL;
  63     return -1;
  64 #endif
  65 }
  66
  67 static int qemu_mprotect__osdep(void *addr, size_t size, int prot)
  68 {
  69     g_assert(!((uintptr_t)addr & ~qemu_real_host_page_mask()));
  70     g_assert(!(size & ~qemu_real_host_page_mask()));
  71
  72 #ifdef _WIN32
  73     DWORD old_protect;
  74
  75     if (!VirtualProtect(addr, size, prot, &old_protect)) {
  76         g_autofree gchar *emsg = g_win32_error_message(GetLastError());
  77         error_report("%s: VirtualProtect failed: %s", __func__, emsg);
  78         return -1;
  79     }
  80     return 0;
  81 #else
  82     if (mprotect(addr, size, prot)) {
  83         error_report("%s: mprotect failed: %s", __func__, strerror(errno));
  84         return -1;
  85     }
  86     return 0;
  87 #endif
  88 }
  89
  90 int qemu_mprotect_rw(void *addr, size_t size)
  91 {
  92 #ifdef _WIN32
  93     return qemu_mprotect__osdep(addr, size, PAGE_READWRITE);
  94 #else
  95     return qemu_mprotect__osdep(addr, size, PROT_READ | PROT_WRITE);
  96 #endif
  97 }
  98
  99 int qemu_mprotect_rwx(void *addr, size_t size)
 100 {
 101 #ifdef _WIN32
 102     return qemu_mprotect__osdep(addr, size, PAGE_EXECUTE_READWRITE);
 103 #else
 104     return qemu_mprotect__osdep(addr, size, PROT_READ | PROT_WRITE | PROT_EXEC);
 105 #endif
 106 }
 107
 108 int qemu_mprotect_none(void *addr, size_t size)
 109 {
 110 #ifdef _WIN32
 111     return qemu_mprotect__osdep(addr, size, PAGE_NOACCESS);
 112 #else
 113     return qemu_mprotect__osdep(addr, size, PROT_NONE);
 114 #endif
 115 }
 116
 117 #ifndef _WIN32
 118
 119 static int fcntl_op_setlk = -1;
 120 static int fcntl_op_getlk = -1;
 121
 122 /*
 123  * Dups an fd and sets the flags
 124  */
 125 int qemu_dup_flags(int fd, int flags)
 126 {
 127     int ret;
 128     int serrno;
 129     int dup_flags;
 130
 131     ret = qemu_dup(fd);
 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 int qemu_dup(int fd)
 171 {
 172     int ret;
 173 #ifdef F_DUPFD_CLOEXEC
 174     ret = fcntl(fd, F_DUPFD_CLOEXEC, 0);
 175 #else
 176     ret = dup(fd);
 177     if (ret != -1) {
 178         qemu_set_cloexec(ret);
 179     }
 180 #endif
 181     return ret;
 182 }
 183
 184 static int qemu_parse_fdset(const char *param)
 185 {
 186     return qemu_parse_fd(param);
 187 }
 188
 189 static void qemu_probe_lock_ops(void)
 190 {
 191     if (fcntl_op_setlk == -1) {
 192 #ifdef F_OFD_SETLK
 193         int fd;
 194         int ret;
 195         struct flock fl = {
 196             .l_whence = SEEK_SET,
 197             .l_start  = 0,
 198             .l_len    = 0,
 199             .l_type   = F_WRLCK,
 200         };
 201
 202         fd = open("/dev/null", O_RDWR);
 203         if (fd < 0) {
 204             fprintf(stderr,
 205                     "Failed to open /dev/null for OFD lock probing: %s\n",
 206                     strerror(errno));
 207             fcntl_op_setlk = F_SETLK;
 208             fcntl_op_getlk = F_GETLK;
 209             return;
 210         }
 211         ret = fcntl(fd, F_OFD_GETLK, &fl);
 212         close(fd);
 213         if (!ret) {
 214             fcntl_op_setlk = F_OFD_SETLK;
 215             fcntl_op_getlk = F_OFD_GETLK;
 216         } else {
 217             fcntl_op_setlk = F_SETLK;
 218             fcntl_op_getlk = F_GETLK;
 219         }
 220 #else
 221         fcntl_op_setlk = F_SETLK;
 222         fcntl_op_getlk = F_GETLK;
 223 #endif
 224     }
 225 }
 226
 227 bool qemu_has_ofd_lock(void)
 228 {
 229     qemu_probe_lock_ops();
 230 #ifdef F_OFD_SETLK
 231     return fcntl_op_setlk == F_OFD_SETLK;
 232 #else
 233     return false;
 234 #endif
 235 }
 236
 237 static int qemu_lock_fcntl(int fd, int64_t start, int64_t len, int fl_type)
 238 {
 239     int ret;
 240     struct flock fl = {
 241         .l_whence = SEEK_SET,
 242         .l_start  = start,
 243         .l_len    = len,
 244         .l_type   = fl_type,
 245     };
 246     qemu_probe_lock_ops();
 247     do {
 248         ret = fcntl(fd, fcntl_op_setlk, &fl);
 249     } while (ret == -1 && errno == EINTR);
 250     return ret == -1 ? -errno : 0;
 251 }
 252
 253 int qemu_lock_fd(int fd, int64_t start, int64_t len, bool exclusive)
 254 {
 255     return qemu_lock_fcntl(fd, start, len, exclusive ? F_WRLCK : F_RDLCK);
 256 }
 257
 258 int qemu_unlock_fd(int fd, int64_t start, int64_t len)
 259 {
 260     return qemu_lock_fcntl(fd, start, len, F_UNLCK);
 261 }
 262
 263 int qemu_lock_fd_test(int fd, int64_t start, int64_t len, bool exclusive)
 264 {
 265     int ret;
 266     struct flock fl = {
 267         .l_whence = SEEK_SET,
 268         .l_start  = start,
 269         .l_len    = len,
 270         .l_type   = exclusive ? F_WRLCK : F_RDLCK,
 271     };
 272     qemu_probe_lock_ops();
 273     ret = fcntl(fd, fcntl_op_getlk, &fl);
 274     if (ret == -1) {
 275         return -errno;
 276     } else {
 277         return fl.l_type == F_UNLCK ? 0 : -EAGAIN;
 278     }
 279 }
 280 #endif
 281
 282 static int qemu_open_cloexec(const char *name, int flags, mode_t mode)
 283 {
 284     int ret;
 285 #ifdef O_CLOEXEC
 286     ret = open(name, flags | O_CLOEXEC, mode);
 287 #else
 288     ret = open(name, flags, mode);
 289     if (ret >= 0) {
 290         qemu_set_cloexec(ret);
 291     }
 292 #endif
 293     return ret;
 294 }
 295
 296 /*
 297  * Opens a file with FD_CLOEXEC set
 298  */
 299 static int
 300 qemu_open_internal(const char *name, int flags, mode_t mode, Error **errp)
 301 {
 302     int ret;
 303
 304 #ifndef _WIN32
 305     const char *fdset_id_str;
 306
 307     /* Attempt dup of fd from fd set */
 308     if (strstart(name, "/dev/fdset/", &fdset_id_str)) {
 309         int64_t fdset_id;
 310         int dupfd;
 311
 312         fdset_id = qemu_parse_fdset(fdset_id_str);
 313         if (fdset_id == -1) {
 314             error_setg(errp, "Could not parse fdset %s", name);
 315             errno = EINVAL;
 316             return -1;
 317         }
 318
 319         dupfd = monitor_fdset_dup_fd_add(fdset_id, flags);
 320         if (dupfd == -1) {
 321             error_setg_errno(errp, errno, "Could not dup FD for %s flags %x",
 322                              name, flags);
 323             return -1;
 324         }
 325
 326         return dupfd;
 327     }
 328 #endif
 329
 330     ret = qemu_open_cloexec(name, flags, mode);
 331
 332     if (ret == -1) {
 333         const char *action = flags & O_CREAT ? "create" : "open";
 334 #ifdef O_DIRECT
 335         /* Give more helpful error message for O_DIRECT */
 336         if (errno == EINVAL && (flags & O_DIRECT)) {
 337             ret = open(name, flags & ~O_DIRECT, mode);
 338             if (ret != -1) {
 339                 close(ret);
 340                 error_setg(errp, "Could not %s '%s': "
 341                            "filesystem does not support O_DIRECT",
 342                            action, name);
 343                 errno = EINVAL; /* restore first open()'s errno */
 344                 return -1;
 345             }
 346         }
 347 #endif /* O_DIRECT */
 348         error_setg_errno(errp, errno, "Could not %s '%s'",
 349                          action, name);
 350     }
 351
 352     return ret;
 353 }
 354
 355
 356 int qemu_open(const char *name, int flags, Error **errp)
 357 {
 358     assert(!(flags & O_CREAT));
 359
 360     return qemu_open_internal(name, flags, 0, errp);
 361 }
 362
 363
 364 int qemu_create(const char *name, int flags, mode_t mode, Error **errp)
 365 {
 366     assert(!(flags & O_CREAT));
 367
 368     return qemu_open_internal(name, flags | O_CREAT, mode, errp);
 369 }
 370
 371
 372 int qemu_open_old(const char *name, int flags, ...)
 373 {
 374     va_list ap;
 375     mode_t mode = 0;
 376     int ret;
 377
 378     va_start(ap, flags);
 379     if (flags & O_CREAT) {
 380         mode = va_arg(ap, int);
 381     }
 382     va_end(ap);
 383
 384     ret = qemu_open_internal(name, flags, mode, NULL);
 385
 386 #ifdef O_DIRECT
 387     if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
 388         error_report("file system may not support O_DIRECT");
 389         errno = EINVAL; /* in case it was clobbered */
 390     }
 391 #endif /* O_DIRECT */
 392
 393     return ret;
 394 }
 395
 396 int qemu_close(int fd)
 397 {
 398     int64_t fdset_id;
 399
 400     /* Close fd that was dup'd from an fdset */
 401     fdset_id = monitor_fdset_dup_fd_find(fd);
 402     if (fdset_id != -1) {
 403         int ret;
 404
 405         ret = close(fd);
 406         if (ret == 0) {
 407             monitor_fdset_dup_fd_remove(fd);
 408         }
 409
 410         return ret;
 411     }
 412
 413     return close(fd);
 414 }
 415
 416 /*
 417  * Delete a file from the filesystem, unless the filename is /dev/fdset/...
 418  *
 419  * Returns: On success, zero is returned.  On error, -1 is returned,
 420  * and errno is set appropriately.
 421  */
 422 int qemu_unlink(const char *name)
 423 {
 424     if (g_str_has_prefix(name, "/dev/fdset/")) {
 425         return 0;
 426     }
 427
 428     return unlink(name);
 429 }
 430
 431 /*
 432  * A variant of write(2) which handles partial write.
 433  *
 434  * Return the number of bytes transferred.
 435  * Set errno if fewer than `count' bytes are written.
 436  *
 437  * This function don't work with non-blocking fd's.
 438  * Any of the possibilities with non-blocking fd's is bad:
 439  *   - return a short write (then name is wrong)
 440  *   - busy wait adding (errno == EAGAIN) to the loop
 441  */
 442 ssize_t qemu_write_full(int fd, const void *buf, size_t count)
 443 {
 444     ssize_t ret = 0;
 445     ssize_t total = 0;
 446
 447     while (count) {
 448         ret = write(fd, buf, count);
 449         if (ret < 0) {
 450             if (errno == EINTR)
 451                 continue;
 452             break;
 453         }
 454
 455         count -= ret;
 456         buf += ret;
 457         total += ret;
 458     }
 459
 460     return total;
 461 }
 462
 463 /*
 464  * Opens a socket with FD_CLOEXEC set
 465  */
 466 int qemu_socket(int domain, int type, int protocol)
 467 {
 468     int ret;
 469
 470 #ifdef SOCK_CLOEXEC
 471     ret = socket(domain, type | SOCK_CLOEXEC, protocol);
 472     if (ret != -1 || errno != EINVAL) {
 473         return ret;
 474     }
 475 #endif
 476     ret = socket(domain, type, protocol);
 477     if (ret >= 0) {
 478         qemu_set_cloexec(ret);
 479     }
 480
 481     return ret;
 482 }
 483
 484 /*
 485  * Accept a connection and set FD_CLOEXEC
 486  */
 487 int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
 488 {
 489     int ret;
 490
 491 #ifdef CONFIG_ACCEPT4
 492     ret = accept4(s, addr, addrlen, SOCK_CLOEXEC);
 493     if (ret != -1 || errno != ENOSYS) {
 494         return ret;
 495     }
 496 #endif
 497     ret = accept(s, addr, addrlen);
 498     if (ret >= 0) {
 499         qemu_set_cloexec(ret);
 500     }
 501
 502     return ret;
 503 }
 504
 505 ssize_t qemu_send_full(int s, const void *buf, size_t count)
 506 {
 507     ssize_t ret = 0;
 508     ssize_t total = 0;
 509
 510     while (count) {
 511         ret = send(s, buf, count, 0);
 512         if (ret < 0) {
 513             if (errno == EINTR) {
 514                 continue;
 515             }
 516             break;
 517         }
 518
 519         count -= ret;
 520         buf += ret;
 521         total += ret;
 522     }
 523
 524     return total;
 525 }
 526
 527 void qemu_set_hw_version(const char *version)
 528 {
 529     hw_version = version;
 530 }
 531
 532 const char *qemu_hw_version(void)
 533 {
 534     return hw_version;
 535 }
 536
 537 #ifdef _WIN32
 538 static void socket_cleanup(void)
 539 {
 540     WSACleanup();
 541 }
 542 #endif
 543
 544 int socket_init(void)
 545 {
 546 #ifdef _WIN32
 547     WSADATA Data;
 548     int ret, err;
 549
 550     ret = WSAStartup(MAKEWORD(2, 2), &Data);
 551     if (ret != 0) {
 552         err = WSAGetLastError();
 553         fprintf(stderr, "WSAStartup: %d\n", err);
 554         return -1;
 555     }
 556     atexit(socket_cleanup);
 557 #endif
 558     return 0;
 559 }
 560
 561
 562 #ifndef CONFIG_IOVEC
 563 static ssize_t
 564 readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write)
 565 {
 566     unsigned i = 0;
 567     ssize_t ret = 0;
 568     ssize_t off = 0;
 569     while (i < iov_cnt) {
 570         ssize_t r = do_write
 571             ? write(fd, iov[i].iov_base + off, iov[i].iov_len - off)
 572             : read(fd, iov[i].iov_base + off, iov[i].iov_len - off);
 573         if (r > 0) {
 574             ret += r;
 575             off += r;
 576             if (off < iov[i].iov_len) {
 577                 continue;
 578             }
 579         } else if (!r) {
 580             break;
 581         } else if (errno == EINTR) {
 582             continue;
 583         } else {
 584             /* else it is some "other" error,
 585              * only return if there was no data processed. */
 586             if (ret == 0) {
 587                 ret = -1;
 588             }
 589             break;
 590         }
 591         off = 0;
 592         i++;
 593     }
 594     return ret;
 595 }
 596
 597 ssize_t
 598 readv(int fd, const struct iovec *iov, int iov_cnt)
 599 {
 600     return readv_writev(fd, iov, iov_cnt, false);
 601 }
 602
 603 ssize_t
 604 writev(int fd, const struct iovec *iov, int iov_cnt)
 605 {
 606     return readv_writev(fd, iov, iov_cnt, true);
 607 }
 608 #endif
 609
 610 /*
 611  * Make sure data goes on disk, but if possible do not bother to
 612  * write out the inode just for timestamp updates.
 613  *
 614  * Unfortunately even in 2009 many operating systems do not support
 615  * fdatasync and have to fall back to fsync.
 616  */
 617 int qemu_fdatasync(int fd)
 618 {
 619 #ifdef CONFIG_FDATASYNC
 620     return fdatasync(fd);
 621 #else
 622     return fsync(fd);
 623 #endif
 624 }