util/oslib-posix.c

   1 /*
   2  * os-posix-lib.c
   3  *
   4  * Copyright (c) 2003-2008 Fabrice Bellard
   5  * Copyright (c) 2010 Red Hat, Inc.
   6  *
   7  * QEMU library functions on POSIX which are shared between QEMU and
   8  * the QEMU tools.
   9  *
  10  * Permission is hereby granted, free of charge, to any person obtaining a copy
  11  * of this software and associated documentation files (the "Software"), to deal
  12  * in the Software without restriction, including without limitation the rights
  13  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  14  * copies of the Software, and to permit persons to whom the Software is
  15  * furnished to do so, subject to the following conditions:
  16  *
  17  * The above copyright notice and this permission notice shall be included in
  18  * all copies or substantial portions of the Software.
  19  *
  20  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  21  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  23  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  24  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  25  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  26  * THE SOFTWARE.
  27  */
  28
  29 #include "qemu/osdep.h"
  30 #include <termios.h>
  31 #include <termios.h>
  32
  33 #include <glib/gprintf.h>
  34
  35 #include "sysemu/sysemu.h"
  36 #include "trace.h"
  37 #include "qapi/error.h"
  38 #include "qemu/sockets.h"
  39 #include <libgen.h>
  40 #include <sys/signal.h>
  41 #include "qemu/cutils.h"
  42
  43 #ifdef CONFIG_LINUX
  44 #include <sys/syscall.h>
  45 #endif
  46
  47 #ifdef __FreeBSD__
  48 #include <sys/sysctl.h>
  49 #endif
  50
  51 #include "qemu/mmap-alloc.h"
  52
  53 int qemu_get_thread_id(void)
  54 {
  55 #if defined(__linux__)
  56     return syscall(SYS_gettid);
  57 #else
  58     return getpid();
  59 #endif
  60 }
  61
  62 int qemu_daemon(int nochdir, int noclose)
  63 {
  64     return daemon(nochdir, noclose);
  65 }
  66
  67 void *qemu_oom_check(void *ptr)
  68 {
  69     if (ptr == NULL) {
  70         fprintf(stderr, "Failed to allocate memory: %s\n", strerror(errno));
  71         abort();
  72     }
  73     return ptr;
  74 }
  75
  76 void *qemu_try_memalign(size_t alignment, size_t size)
  77 {
  78     void *ptr;
  79
  80     if (alignment < sizeof(void*)) {
  81         alignment = sizeof(void*);
  82     }
  83
  84 #if defined(_POSIX_C_SOURCE) && !defined(__sun__)
  85     int ret;
  86     ret = posix_memalign(&ptr, alignment, size);
  87     if (ret != 0) {
  88         errno = ret;
  89         ptr = NULL;
  90     }
  91 #elif defined(CONFIG_BSD)
  92     ptr = valloc(size);
  93 #else
  94     ptr = memalign(alignment, size);
  95 #endif
  96     trace_qemu_memalign(alignment, size, ptr);
  97     return ptr;
  98 }
  99
 100 void *qemu_memalign(size_t alignment, size_t size)
 101 {
 102     return qemu_oom_check(qemu_try_memalign(alignment, size));
 103 }
 104
 105 /* alloc shared memory pages */
 106 void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment)
 107 {
 108     size_t align = QEMU_VMALLOC_ALIGN;
 109     void *ptr = qemu_ram_mmap(-1, size, align, false);
 110
 111     if (ptr == MAP_FAILED) {
 112         return NULL;
 113     }
 114
 115     if (alignment) {
 116         *alignment = align;
 117     }
 118
 119     trace_qemu_anon_ram_alloc(size, ptr);
 120     return ptr;
 121 }
 122
 123 void qemu_vfree(void *ptr)
 124 {
 125     trace_qemu_vfree(ptr);
 126     free(ptr);
 127 }
 128
 129 void qemu_anon_ram_free(void *ptr, size_t size)
 130 {
 131     trace_qemu_anon_ram_free(ptr, size);
 132     qemu_ram_munmap(ptr, size);
 133 }
 134
 135 void qemu_set_block(int fd)
 136 {
 137     int f;
 138     f = fcntl(fd, F_GETFL);
 139     fcntl(fd, F_SETFL, f & ~O_NONBLOCK);
 140 }
 141
 142 void qemu_set_nonblock(int fd)
 143 {
 144     int f;
 145     f = fcntl(fd, F_GETFL);
 146     fcntl(fd, F_SETFL, f | O_NONBLOCK);
 147 }
 148
 149 int socket_set_fast_reuse(int fd)
 150 {
 151     int val = 1, ret;
 152
 153     ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
 154                      (const char *)&val, sizeof(val));
 155
 156     assert(ret == 0);
 157
 158     return ret;
 159 }
 160
 161 void qemu_set_cloexec(int fd)
 162 {
 163     int f;
 164     f = fcntl(fd, F_GETFD);
 165     fcntl(fd, F_SETFD, f | FD_CLOEXEC);
 166 }
 167
 168 /*
 169  * Creates a pipe with FD_CLOEXEC set on both file descriptors
 170  */
 171 int qemu_pipe(int pipefd[2])
 172 {
 173     int ret;
 174
 175 #ifdef CONFIG_PIPE2
 176     ret = pipe2(pipefd, O_CLOEXEC);
 177     if (ret != -1 || errno != ENOSYS) {
 178         return ret;
 179     }
 180 #endif
 181     ret = pipe(pipefd);
 182     if (ret == 0) {
 183         qemu_set_cloexec(pipefd[0]);
 184         qemu_set_cloexec(pipefd[1]);
 185     }
 186
 187     return ret;
 188 }
 189
 190 int qemu_utimens(const char *path, const struct timespec *times)
 191 {
 192     struct timeval tv[2], tv_now;
 193     struct stat st;
 194     int i;
 195 #ifdef CONFIG_UTIMENSAT
 196     int ret;
 197
 198     ret = utimensat(AT_FDCWD, path, times, AT_SYMLINK_NOFOLLOW);
 199     if (ret != -1 || errno != ENOSYS) {
 200         return ret;
 201     }
 202 #endif
 203     /* Fallback: use utimes() instead of utimensat() */
 204
 205     /* happy if special cases */
 206     if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT) {
 207         return 0;
 208     }
 209     if (times[0].tv_nsec == UTIME_NOW && times[1].tv_nsec == UTIME_NOW) {
 210         return utimes(path, NULL);
 211     }
 212
 213     /* prepare for hard cases */
 214     if (times[0].tv_nsec == UTIME_NOW || times[1].tv_nsec == UTIME_NOW) {
 215         gettimeofday(&tv_now, NULL);
 216     }
 217     if (times[0].tv_nsec == UTIME_OMIT || times[1].tv_nsec == UTIME_OMIT) {
 218         stat(path, &st);
 219     }
 220
 221     for (i = 0; i < 2; i++) {
 222         if (times[i].tv_nsec == UTIME_NOW) {
 223             tv[i].tv_sec = tv_now.tv_sec;
 224             tv[i].tv_usec = tv_now.tv_usec;
 225         } else if (times[i].tv_nsec == UTIME_OMIT) {
 226             tv[i].tv_sec = (i == 0) ? st.st_atime : st.st_mtime;
 227             tv[i].tv_usec = 0;
 228         } else {
 229             tv[i].tv_sec = times[i].tv_sec;
 230             tv[i].tv_usec = times[i].tv_nsec / 1000;
 231         }
 232     }
 233
 234     return utimes(path, &tv[0]);
 235 }
 236
 237 char *
 238 qemu_get_local_state_pathname(const char *relative_pathname)
 239 {
 240     return g_strdup_printf("%s/%s", CONFIG_QEMU_LOCALSTATEDIR,
 241                            relative_pathname);
 242 }
 243
 244 void qemu_set_tty_echo(int fd, bool echo)
 245 {
 246     struct termios tty;
 247
 248     tcgetattr(fd, &tty);
 249
 250     if (echo) {
 251         tty.c_lflag |= ECHO | ECHONL | ICANON | IEXTEN;
 252     } else {
 253         tty.c_lflag &= ~(ECHO | ECHONL | ICANON | IEXTEN);
 254     }
 255
 256     tcsetattr(fd, TCSANOW, &tty);
 257 }
 258
 259 static char exec_dir[PATH_MAX];
 260
 261 void qemu_init_exec_dir(const char *argv0)
 262 {
 263     char *dir;
 264     char *p = NULL;
 265     char buf[PATH_MAX];
 266
 267     assert(!exec_dir[0]);
 268
 269 #if defined(__linux__)
 270     {
 271         int len;
 272         len = readlink("/proc/self/exe", buf, sizeof(buf) - 1);
 273         if (len > 0) {
 274             buf[len] = 0;
 275             p = buf;
 276         }
 277     }
 278 #elif defined(__FreeBSD__)
 279     {
 280         static int mib[4] = {CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1};
 281         size_t len = sizeof(buf) - 1;
 282
 283         *buf = '\0';
 284         if (!sysctl(mib, ARRAY_SIZE(mib), buf, &len, NULL, 0) &&
 285             *buf) {
 286             buf[sizeof(buf) - 1] = '\0';
 287             p = buf;
 288         }
 289     }
 290 #endif
 291     /* If we don't have any way of figuring out the actual executable
 292        location then try argv[0].  */
 293     if (!p) {
 294         if (!argv0) {
 295             return;
 296         }
 297         p = realpath(argv0, buf);
 298         if (!p) {
 299             return;
 300         }
 301     }
 302     dir = g_path_get_dirname(p);
 303
 304     pstrcpy(exec_dir, sizeof(exec_dir), dir);
 305
 306     g_free(dir);
 307 }
 308
 309 char *qemu_get_exec_dir(void)
 310 {
 311     return g_strdup(exec_dir);
 312 }
 313
 314 static sigjmp_buf sigjump;
 315
 316 static void sigbus_handler(int signal)
 317 {
 318     siglongjmp(sigjump, 1);
 319 }
 320
 321 void os_mem_prealloc(int fd, char *area, size_t memory)
 322 {
 323     int ret;
 324     struct sigaction act, oldact;
 325     sigset_t set, oldset;
 326
 327     memset(&act, 0, sizeof(act));
 328     act.sa_handler = &sigbus_handler;
 329     act.sa_flags = 0;
 330
 331     ret = sigaction(SIGBUS, &act, &oldact);
 332     if (ret) {
 333         perror("os_mem_prealloc: failed to install signal handler");
 334         exit(1);
 335     }
 336
 337     /* unblock SIGBUS */
 338     sigemptyset(&set);
 339     sigaddset(&set, SIGBUS);
 340     pthread_sigmask(SIG_UNBLOCK, &set, &oldset);
 341
 342     if (sigsetjmp(sigjump, 1)) {
 343         fprintf(stderr, "os_mem_prealloc: Insufficient free host memory "
 344                         "pages available to allocate guest RAM\n");
 345         exit(1);
 346     } else {
 347         int i;
 348         size_t hpagesize = qemu_fd_getpagesize(fd);
 349         size_t numpages = DIV_ROUND_UP(memory, hpagesize);
 350
 351         /* MAP_POPULATE silently ignores failures */
 352         for (i = 0; i < numpages; i++) {
 353             memset(area + (hpagesize * i), 0, 1);
 354         }
 355
 356         ret = sigaction(SIGBUS, &oldact, NULL);
 357         if (ret) {
 358             perror("os_mem_prealloc: failed to reinstall signal handler");
 359             exit(1);
 360         }
 361
 362         pthread_sigmask(SIG_SETMASK, &oldset, NULL);
 363     }
 364 }
 365
 366
 367 static struct termios oldtty;
 368
 369 static void term_exit(void)
 370 {
 371     tcsetattr(0, TCSANOW, &oldtty);
 372 }
 373
 374 static void term_init(void)
 375 {
 376     struct termios tty;
 377
 378     tcgetattr(0, &tty);
 379     oldtty = tty;
 380
 381     tty.c_iflag &= ~(IGNBRK|BRKINT|PARMRK|ISTRIP
 382                           |INLCR|IGNCR|ICRNL|IXON);
 383     tty.c_oflag |= OPOST;
 384     tty.c_lflag &= ~(ECHO|ECHONL|ICANON|IEXTEN);
 385     tty.c_cflag &= ~(CSIZE|PARENB);
 386     tty.c_cflag |= CS8;
 387     tty.c_cc[VMIN] = 1;
 388     tty.c_cc[VTIME] = 0;
 389
 390     tcsetattr(0, TCSANOW, &tty);
 391
 392     atexit(term_exit);
 393 }
 394
 395 int qemu_read_password(char *buf, int buf_size)
 396 {
 397     uint8_t ch;
 398     int i, ret;
 399
 400     printf("password: ");
 401     fflush(stdout);
 402     term_init();
 403     i = 0;
 404     for (;;) {
 405         ret = read(0, &ch, 1);
 406         if (ret == -1) {
 407             if (errno == EAGAIN || errno == EINTR) {
 408                 continue;
 409             } else {
 410                 break;
 411             }
 412         } else if (ret == 0) {
 413             ret = -1;
 414             break;
 415         } else {
 416             if (ch == '\r' ||
 417                 ch == '\n') {
 418                 ret = 0;
 419                 break;
 420             }
 421             if (i < (buf_size - 1)) {
 422                 buf[i++] = ch;
 423             }
 424         }
 425     }
 426     term_exit();
 427     buf[i] = '\0';
 428     printf("\n");
 429     return ret;
 430 }
 431
 432
 433 pid_t qemu_fork(Error **errp)
 434 {
 435     sigset_t oldmask, newmask;
 436     struct sigaction sig_action;
 437     int saved_errno;
 438     pid_t pid;
 439
 440     /*
 441      * Need to block signals now, so that child process can safely
 442      * kill off caller's signal handlers without a race.
 443      */
 444     sigfillset(&newmask);
 445     if (pthread_sigmask(SIG_SETMASK, &newmask, &oldmask) != 0) {
 446         error_setg_errno(errp, errno,
 447                          "cannot block signals");
 448         return -1;
 449     }
 450
 451     pid = fork();
 452     saved_errno = errno;
 453
 454     if (pid < 0) {
 455         /* attempt to restore signal mask, but ignore failure, to
 456          * avoid obscuring the fork failure */
 457         (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
 458         error_setg_errno(errp, saved_errno,
 459                          "cannot fork child process");
 460         errno = saved_errno;
 461         return -1;
 462     } else if (pid) {
 463         /* parent process */
 464
 465         /* Restore our original signal mask now that the child is
 466          * safely running. Only documented failures are EFAULT (not
 467          * possible, since we are using just-grabbed mask) or EINVAL
 468          * (not possible, since we are using correct arguments).  */
 469         (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
 470     } else {
 471         /* child process */
 472         size_t i;
 473
 474         /* Clear out all signal handlers from parent so nothing
 475          * unexpected can happen in our child once we unblock
 476          * signals */
 477         sig_action.sa_handler = SIG_DFL;
 478         sig_action.sa_flags = 0;
 479         sigemptyset(&sig_action.sa_mask);
 480
 481         for (i = 1; i < NSIG; i++) {
 482             /* Only possible errors are EFAULT or EINVAL The former
 483              * won't happen, the latter we expect, so no need to check
 484              * return value */
 485             (void)sigaction(i, &sig_action, NULL);
 486         }
 487
 488         /* Unmask all signals in child, since we've no idea what the
 489          * caller's done with their signal mask and don't want to
 490          * propagate that to children */
 491         sigemptyset(&newmask);
 492         if (pthread_sigmask(SIG_SETMASK, &newmask, NULL) != 0) {
 493             Error *local_err = NULL;
 494             error_setg_errno(&local_err, errno,
 495                              "cannot unblock signals");
 496             error_report_err(local_err);
 497             _exit(1);
 498         }
 499     }
 500     return pid;
 501 }