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 #include <libutil.h>
  50 #endif
  51
  52 #include "qemu/mmap-alloc.h"
  53
  54 int qemu_get_thread_id(void)
  55 {
  56 #if defined(__linux__)
  57     return syscall(SYS_gettid);
  58 #else
  59     return getpid();
  60 #endif
  61 }
  62
  63 int qemu_daemon(int nochdir, int noclose)
  64 {
  65     return daemon(nochdir, noclose);
  66 }
  67
  68 void *qemu_oom_check(void *ptr)
  69 {
  70     if (ptr == NULL) {
  71         fprintf(stderr, "Failed to allocate memory: %s\n", strerror(errno));
  72         abort();
  73     }
  74     return ptr;
  75 }
  76
  77 void *qemu_try_memalign(size_t alignment, size_t size)
  78 {
  79     void *ptr;
  80
  81     if (alignment < sizeof(void*)) {
  82         alignment = sizeof(void*);
  83     }
  84
  85 #if defined(_POSIX_C_SOURCE) && !defined(__sun__)
  86     int ret;
  87     ret = posix_memalign(&ptr, alignment, size);
  88     if (ret != 0) {
  89         errno = ret;
  90         ptr = NULL;
  91     }
  92 #elif defined(CONFIG_BSD)
  93     ptr = valloc(size);
  94 #else
  95     ptr = memalign(alignment, size);
  96 #endif
  97     trace_qemu_memalign(alignment, size, ptr);
  98     return ptr;
  99 }
 100
 101 void *qemu_memalign(size_t alignment, size_t size)
 102 {
 103     return qemu_oom_check(qemu_try_memalign(alignment, size));
 104 }
 105
 106 /* alloc shared memory pages */
 107 void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment)
 108 {
 109     size_t align = QEMU_VMALLOC_ALIGN;
 110     void *ptr = qemu_ram_mmap(-1, size, align, false);
 111
 112     if (ptr == MAP_FAILED) {
 113         return NULL;
 114     }
 115
 116     if (alignment) {
 117         *alignment = align;
 118     }
 119
 120     trace_qemu_anon_ram_alloc(size, ptr);
 121     return ptr;
 122 }
 123
 124 void qemu_vfree(void *ptr)
 125 {
 126     trace_qemu_vfree(ptr);
 127     free(ptr);
 128 }
 129
 130 void qemu_anon_ram_free(void *ptr, size_t size)
 131 {
 132     trace_qemu_anon_ram_free(ptr, size);
 133     qemu_ram_munmap(ptr, size);
 134 }
 135
 136 void qemu_set_block(int fd)
 137 {
 138     int f;
 139     f = fcntl(fd, F_GETFL);
 140     fcntl(fd, F_SETFL, f & ~O_NONBLOCK);
 141 }
 142
 143 void qemu_set_nonblock(int fd)
 144 {
 145     int f;
 146     f = fcntl(fd, F_GETFL);
 147     fcntl(fd, F_SETFL, f | O_NONBLOCK);
 148 }
 149
 150 int socket_set_fast_reuse(int fd)
 151 {
 152     int val = 1, ret;
 153
 154     ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
 155                      (const char *)&val, sizeof(val));
 156
 157     assert(ret == 0);
 158
 159     return ret;
 160 }
 161
 162 void qemu_set_cloexec(int fd)
 163 {
 164     int f;
 165     f = fcntl(fd, F_GETFD);
 166     fcntl(fd, F_SETFD, f | FD_CLOEXEC);
 167 }
 168
 169 /*
 170  * Creates a pipe with FD_CLOEXEC set on both file descriptors
 171  */
 172 int qemu_pipe(int pipefd[2])
 173 {
 174     int ret;
 175
 176 #ifdef CONFIG_PIPE2
 177     ret = pipe2(pipefd, O_CLOEXEC);
 178     if (ret != -1 || errno != ENOSYS) {
 179         return ret;
 180     }
 181 #endif
 182     ret = pipe(pipefd);
 183     if (ret == 0) {
 184         qemu_set_cloexec(pipefd[0]);
 185         qemu_set_cloexec(pipefd[1]);
 186     }
 187
 188     return ret;
 189 }
 190
 191 int qemu_utimens(const char *path, const struct timespec *times)
 192 {
 193     struct timeval tv[2], tv_now;
 194     struct stat st;
 195     int i;
 196 #ifdef CONFIG_UTIMENSAT
 197     int ret;
 198
 199     ret = utimensat(AT_FDCWD, path, times, AT_SYMLINK_NOFOLLOW);
 200     if (ret != -1 || errno != ENOSYS) {
 201         return ret;
 202     }
 203 #endif
 204     /* Fallback: use utimes() instead of utimensat() */
 205
 206     /* happy if special cases */
 207     if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT) {
 208         return 0;
 209     }
 210     if (times[0].tv_nsec == UTIME_NOW && times[1].tv_nsec == UTIME_NOW) {
 211         return utimes(path, NULL);
 212     }
 213
 214     /* prepare for hard cases */
 215     if (times[0].tv_nsec == UTIME_NOW || times[1].tv_nsec == UTIME_NOW) {
 216         gettimeofday(&tv_now, NULL);
 217     }
 218     if (times[0].tv_nsec == UTIME_OMIT || times[1].tv_nsec == UTIME_OMIT) {
 219         stat(path, &st);
 220     }
 221
 222     for (i = 0; i < 2; i++) {
 223         if (times[i].tv_nsec == UTIME_NOW) {
 224             tv[i].tv_sec = tv_now.tv_sec;
 225             tv[i].tv_usec = tv_now.tv_usec;
 226         } else if (times[i].tv_nsec == UTIME_OMIT) {
 227             tv[i].tv_sec = (i == 0) ? st.st_atime : st.st_mtime;
 228             tv[i].tv_usec = 0;
 229         } else {
 230             tv[i].tv_sec = times[i].tv_sec;
 231             tv[i].tv_usec = times[i].tv_nsec / 1000;
 232         }
 233     }
 234
 235     return utimes(path, &tv[0]);
 236 }
 237
 238 char *
 239 qemu_get_local_state_pathname(const char *relative_pathname)
 240 {
 241     return g_strdup_printf("%s/%s", CONFIG_QEMU_LOCALSTATEDIR,
 242                            relative_pathname);
 243 }
 244
 245 void qemu_set_tty_echo(int fd, bool echo)
 246 {
 247     struct termios tty;
 248
 249     tcgetattr(fd, &tty);
 250
 251     if (echo) {
 252         tty.c_lflag |= ECHO | ECHONL | ICANON | IEXTEN;
 253     } else {
 254         tty.c_lflag &= ~(ECHO | ECHONL | ICANON | IEXTEN);
 255     }
 256
 257     tcsetattr(fd, TCSANOW, &tty);
 258 }
 259
 260 static char exec_dir[PATH_MAX];
 261
 262 void qemu_init_exec_dir(const char *argv0)
 263 {
 264     char *dir;
 265     char *p = NULL;
 266     char buf[PATH_MAX];
 267
 268     assert(!exec_dir[0]);
 269
 270 #if defined(__linux__)
 271     {
 272         int len;
 273         len = readlink("/proc/self/exe", buf, sizeof(buf) - 1);
 274         if (len > 0) {
 275             buf[len] = 0;
 276             p = buf;
 277         }
 278     }
 279 #elif defined(__FreeBSD__)
 280     {
 281         static int mib[4] = {CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1};
 282         size_t len = sizeof(buf) - 1;
 283
 284         *buf = '\0';
 285         if (!sysctl(mib, ARRAY_SIZE(mib), buf, &len, NULL, 0) &&
 286             *buf) {
 287             buf[sizeof(buf) - 1] = '\0';
 288             p = buf;
 289         }
 290     }
 291 #endif
 292     /* If we don't have any way of figuring out the actual executable
 293        location then try argv[0].  */
 294     if (!p) {
 295         if (!argv0) {
 296             return;
 297         }
 298         p = realpath(argv0, buf);
 299         if (!p) {
 300             return;
 301         }
 302     }
 303     dir = g_path_get_dirname(p);
 304
 305     pstrcpy(exec_dir, sizeof(exec_dir), dir);
 306
 307     g_free(dir);
 308 }
 309
 310 char *qemu_get_exec_dir(void)
 311 {
 312     return g_strdup(exec_dir);
 313 }
 314
 315 static sigjmp_buf sigjump;
 316
 317 static void sigbus_handler(int signal)
 318 {
 319     siglongjmp(sigjump, 1);
 320 }
 321
 322 void os_mem_prealloc(int fd, char *area, size_t memory, Error **errp)
 323 {
 324     int ret;
 325     struct sigaction act, oldact;
 326     sigset_t set, oldset;
 327
 328     memset(&act, 0, sizeof(act));
 329     act.sa_handler = &sigbus_handler;
 330     act.sa_flags = 0;
 331
 332     ret = sigaction(SIGBUS, &act, &oldact);
 333     if (ret) {
 334         error_setg_errno(errp, errno,
 335             "os_mem_prealloc: failed to install signal handler");
 336         return;
 337     }
 338
 339     /* unblock SIGBUS */
 340     sigemptyset(&set);
 341     sigaddset(&set, SIGBUS);
 342     pthread_sigmask(SIG_UNBLOCK, &set, &oldset);
 343
 344     if (sigsetjmp(sigjump, 1)) {
 345         error_setg(errp, "os_mem_prealloc: Insufficient free host memory "
 346             "pages available to allocate guest RAM\n");
 347     } else {
 348         int i;
 349         size_t hpagesize = qemu_fd_getpagesize(fd);
 350         size_t numpages = DIV_ROUND_UP(memory, hpagesize);
 351
 352         /* MAP_POPULATE silently ignores failures */
 353         for (i = 0; i < numpages; i++) {
 354             memset(area + (hpagesize * i), 0, 1);
 355         }
 356     }
 357
 358     ret = sigaction(SIGBUS, &oldact, NULL);
 359     if (ret) {
 360         /* Terminate QEMU since it can't recover from error */
 361         perror("os_mem_prealloc: failed to reinstall signal handler");
 362         exit(1);
 363     }
 364     pthread_sigmask(SIG_SETMASK, &oldset, NULL);
 365 }
 366
 367
 368 static struct termios oldtty;
 369
 370 static void term_exit(void)
 371 {
 372     tcsetattr(0, TCSANOW, &oldtty);
 373 }
 374
 375 static void term_init(void)
 376 {
 377     struct termios tty;
 378
 379     tcgetattr(0, &tty);
 380     oldtty = tty;
 381
 382     tty.c_iflag &= ~(IGNBRK|BRKINT|PARMRK|ISTRIP
 383                           |INLCR|IGNCR|ICRNL|IXON);
 384     tty.c_oflag |= OPOST;
 385     tty.c_lflag &= ~(ECHO|ECHONL|ICANON|IEXTEN);
 386     tty.c_cflag &= ~(CSIZE|PARENB);
 387     tty.c_cflag |= CS8;
 388     tty.c_cc[VMIN] = 1;
 389     tty.c_cc[VTIME] = 0;
 390
 391     tcsetattr(0, TCSANOW, &tty);
 392
 393     atexit(term_exit);
 394 }
 395
 396 int qemu_read_password(char *buf, int buf_size)
 397 {
 398     uint8_t ch;
 399     int i, ret;
 400
 401     printf("password: ");
 402     fflush(stdout);
 403     term_init();
 404     i = 0;
 405     for (;;) {
 406         ret = read(0, &ch, 1);
 407         if (ret == -1) {
 408             if (errno == EAGAIN || errno == EINTR) {
 409                 continue;
 410             } else {
 411                 break;
 412             }
 413         } else if (ret == 0) {
 414             ret = -1;
 415             break;
 416         } else {
 417             if (ch == '\r' ||
 418                 ch == '\n') {
 419                 ret = 0;
 420                 break;
 421             }
 422             if (i < (buf_size - 1)) {
 423                 buf[i++] = ch;
 424             }
 425         }
 426     }
 427     term_exit();
 428     buf[i] = '\0';
 429     printf("\n");
 430     return ret;
 431 }
 432
 433
 434 char *qemu_get_pid_name(pid_t pid)
 435 {
 436     char *name = NULL;
 437
 438 #if defined(__FreeBSD__)
 439     /* BSDs don't have /proc, but they provide a nice substitute */
 440     struct kinfo_proc *proc = kinfo_getproc(pid);
 441
 442     if (proc) {
 443         name = g_strdup(proc->ki_comm);
 444         free(proc);
 445     }
 446 #else
 447     /* Assume a system with reasonable procfs */
 448     char *pid_path;
 449     size_t len;
 450
 451     pid_path = g_strdup_printf("/proc/%d/cmdline", pid);
 452     g_file_get_contents(pid_path, &name, &len, NULL);
 453     g_free(pid_path);
 454 #endif
 455
 456     return name;
 457 }
 458
 459
 460 pid_t qemu_fork(Error **errp)
 461 {
 462     sigset_t oldmask, newmask;
 463     struct sigaction sig_action;
 464     int saved_errno;
 465     pid_t pid;
 466
 467     /*
 468      * Need to block signals now, so that child process can safely
 469      * kill off caller's signal handlers without a race.
 470      */
 471     sigfillset(&newmask);
 472     if (pthread_sigmask(SIG_SETMASK, &newmask, &oldmask) != 0) {
 473         error_setg_errno(errp, errno,
 474                          "cannot block signals");
 475         return -1;
 476     }
 477
 478     pid = fork();
 479     saved_errno = errno;
 480
 481     if (pid < 0) {
 482         /* attempt to restore signal mask, but ignore failure, to
 483          * avoid obscuring the fork failure */
 484         (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
 485         error_setg_errno(errp, saved_errno,
 486                          "cannot fork child process");
 487         errno = saved_errno;
 488         return -1;
 489     } else if (pid) {
 490         /* parent process */
 491
 492         /* Restore our original signal mask now that the child is
 493          * safely running. Only documented failures are EFAULT (not
 494          * possible, since we are using just-grabbed mask) or EINVAL
 495          * (not possible, since we are using correct arguments).  */
 496         (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
 497     } else {
 498         /* child process */
 499         size_t i;
 500
 501         /* Clear out all signal handlers from parent so nothing
 502          * unexpected can happen in our child once we unblock
 503          * signals */
 504         sig_action.sa_handler = SIG_DFL;
 505         sig_action.sa_flags = 0;
 506         sigemptyset(&sig_action.sa_mask);
 507
 508         for (i = 1; i < NSIG; i++) {
 509             /* Only possible errors are EFAULT or EINVAL The former
 510              * won't happen, the latter we expect, so no need to check
 511              * return value */
 512             (void)sigaction(i, &sig_action, NULL);
 513         }
 514
 515         /* Unmask all signals in child, since we've no idea what the
 516          * caller's done with their signal mask and don't want to
 517          * propagate that to children */
 518         sigemptyset(&newmask);
 519         if (pthread_sigmask(SIG_SETMASK, &newmask, NULL) != 0) {
 520             Error *local_err = NULL;
 521             error_setg_errno(&local_err, errno,
 522                              "cannot unblock signals");
 523             error_report_err(local_err);
 524             _exit(1);
 525         }
 526     }
 527     return pid;
 528 }