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 /* The following block of code temporarily renames the daemon() function so the
  30    compiler does not see the warning associated with it in stdlib.h on OSX */
  31 #ifdef __APPLE__
  32 #define daemon qemu_fake_daemon_function
  33 #include <stdlib.h>
  34 #undef daemon
  35 extern int daemon(int, int);
  36 #endif
  37
  38 #if defined(__linux__) && (defined(__x86_64__) || defined(__arm__))
  39    /* Use 2 MiB alignment so transparent hugepages can be used by KVM.
  40       Valgrind does not support alignments larger than 1 MiB,
  41       therefore we need special code which handles running on Valgrind. */
  42 #  define QEMU_VMALLOC_ALIGN (512 * 4096)
  43 #elif defined(__linux__) && defined(__s390x__)
  44    /* Use 1 MiB (segment size) alignment so gmap can be used by KVM. */
  45 #  define QEMU_VMALLOC_ALIGN (256 * 4096)
  46 #else
  47 #  define QEMU_VMALLOC_ALIGN getpagesize()
  48 #endif
  49 #define HUGETLBFS_MAGIC       0x958458f6
  50
  51 #include <termios.h>
  52 #include <unistd.h>
  53
  54 #include <glib/gprintf.h>
  55
  56 #include "config-host.h"
  57 #include "sysemu/sysemu.h"
  58 #include "trace.h"
  59 #include "qemu/sockets.h"
  60 #include <sys/mman.h>
  61 #include <libgen.h>
  62 #include <setjmp.h>
  63 #include <sys/signal.h>
  64
  65 #ifdef CONFIG_LINUX
  66 #include <sys/syscall.h>
  67 #include <sys/vfs.h>
  68 #endif
  69
  70 #ifdef __FreeBSD__
  71 #include <sys/sysctl.h>
  72 #endif
  73
  74 int qemu_get_thread_id(void)
  75 {
  76 #if defined(__linux__)
  77     return syscall(SYS_gettid);
  78 #else
  79     return getpid();
  80 #endif
  81 }
  82
  83 int qemu_daemon(int nochdir, int noclose)
  84 {
  85     return daemon(nochdir, noclose);
  86 }
  87
  88 void *qemu_oom_check(void *ptr)
  89 {
  90     if (ptr == NULL) {
  91         fprintf(stderr, "Failed to allocate memory: %s\n", strerror(errno));
  92         abort();
  93     }
  94     return ptr;
  95 }
  96
  97 void *qemu_try_memalign(size_t alignment, size_t size)
  98 {
  99     void *ptr;
 100
 101     if (alignment < sizeof(void*)) {
 102         alignment = sizeof(void*);
 103     }
 104
 105 #if defined(_POSIX_C_SOURCE) && !defined(__sun__)
 106     int ret;
 107     ret = posix_memalign(&ptr, alignment, size);
 108     if (ret != 0) {
 109         errno = ret;
 110         ptr = NULL;
 111     }
 112 #elif defined(CONFIG_BSD)
 113     ptr = valloc(size);
 114 #else
 115     ptr = memalign(alignment, size);
 116 #endif
 117     trace_qemu_memalign(alignment, size, ptr);
 118     return ptr;
 119 }
 120
 121 void *qemu_memalign(size_t alignment, size_t size)
 122 {
 123     return qemu_oom_check(qemu_try_memalign(alignment, size));
 124 }
 125
 126 /* alloc shared memory pages */
 127 void *qemu_anon_ram_alloc(size_t size)
 128 {
 129     size_t align = QEMU_VMALLOC_ALIGN;
 130     size_t total = size + align - getpagesize();
 131     void *ptr = mmap(0, total, PROT_READ | PROT_WRITE,
 132                      MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
 133     size_t offset = QEMU_ALIGN_UP((uintptr_t)ptr, align) - (uintptr_t)ptr;
 134
 135     if (ptr == MAP_FAILED) {
 136         return NULL;
 137     }
 138
 139     ptr += offset;
 140     total -= offset;
 141
 142     if (offset > 0) {
 143         munmap(ptr - offset, offset);
 144     }
 145     if (total > size) {
 146         munmap(ptr + size, total - size);
 147     }
 148
 149     trace_qemu_anon_ram_alloc(size, ptr);
 150     return ptr;
 151 }
 152
 153 void qemu_vfree(void *ptr)
 154 {
 155     trace_qemu_vfree(ptr);
 156     free(ptr);
 157 }
 158
 159 void qemu_anon_ram_free(void *ptr, size_t size)
 160 {
 161     trace_qemu_anon_ram_free(ptr, size);
 162     if (ptr) {
 163         munmap(ptr, size);
 164     }
 165 }
 166
 167 void qemu_set_block(int fd)
 168 {
 169     int f;
 170     f = fcntl(fd, F_GETFL);
 171     fcntl(fd, F_SETFL, f & ~O_NONBLOCK);
 172 }
 173
 174 void qemu_set_nonblock(int fd)
 175 {
 176     int f;
 177     f = fcntl(fd, F_GETFL);
 178     fcntl(fd, F_SETFL, f | O_NONBLOCK);
 179 }
 180
 181 int socket_set_fast_reuse(int fd)
 182 {
 183     int val = 1, ret;
 184
 185     ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
 186                      (const char *)&val, sizeof(val));
 187
 188     assert(ret == 0);
 189
 190     return ret;
 191 }
 192
 193 void qemu_set_cloexec(int fd)
 194 {
 195     int f;
 196     f = fcntl(fd, F_GETFD);
 197     fcntl(fd, F_SETFD, f | FD_CLOEXEC);
 198 }
 199
 200 /*
 201  * Creates a pipe with FD_CLOEXEC set on both file descriptors
 202  */
 203 int qemu_pipe(int pipefd[2])
 204 {
 205     int ret;
 206
 207 #ifdef CONFIG_PIPE2
 208     ret = pipe2(pipefd, O_CLOEXEC);
 209     if (ret != -1 || errno != ENOSYS) {
 210         return ret;
 211     }
 212 #endif
 213     ret = pipe(pipefd);
 214     if (ret == 0) {
 215         qemu_set_cloexec(pipefd[0]);
 216         qemu_set_cloexec(pipefd[1]);
 217     }
 218
 219     return ret;
 220 }
 221
 222 int qemu_utimens(const char *path, const struct timespec *times)
 223 {
 224     struct timeval tv[2], tv_now;
 225     struct stat st;
 226     int i;
 227 #ifdef CONFIG_UTIMENSAT
 228     int ret;
 229
 230     ret = utimensat(AT_FDCWD, path, times, AT_SYMLINK_NOFOLLOW);
 231     if (ret != -1 || errno != ENOSYS) {
 232         return ret;
 233     }
 234 #endif
 235     /* Fallback: use utimes() instead of utimensat() */
 236
 237     /* happy if special cases */
 238     if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT) {
 239         return 0;
 240     }
 241     if (times[0].tv_nsec == UTIME_NOW && times[1].tv_nsec == UTIME_NOW) {
 242         return utimes(path, NULL);
 243     }
 244
 245     /* prepare for hard cases */
 246     if (times[0].tv_nsec == UTIME_NOW || times[1].tv_nsec == UTIME_NOW) {
 247         gettimeofday(&tv_now, NULL);
 248     }
 249     if (times[0].tv_nsec == UTIME_OMIT || times[1].tv_nsec == UTIME_OMIT) {
 250         stat(path, &st);
 251     }
 252
 253     for (i = 0; i < 2; i++) {
 254         if (times[i].tv_nsec == UTIME_NOW) {
 255             tv[i].tv_sec = tv_now.tv_sec;
 256             tv[i].tv_usec = tv_now.tv_usec;
 257         } else if (times[i].tv_nsec == UTIME_OMIT) {
 258             tv[i].tv_sec = (i == 0) ? st.st_atime : st.st_mtime;
 259             tv[i].tv_usec = 0;
 260         } else {
 261             tv[i].tv_sec = times[i].tv_sec;
 262             tv[i].tv_usec = times[i].tv_nsec / 1000;
 263         }
 264     }
 265
 266     return utimes(path, &tv[0]);
 267 }
 268
 269 char *
 270 qemu_get_local_state_pathname(const char *relative_pathname)
 271 {
 272     return g_strdup_printf("%s/%s", CONFIG_QEMU_LOCALSTATEDIR,
 273                            relative_pathname);
 274 }
 275
 276 void qemu_set_tty_echo(int fd, bool echo)
 277 {
 278     struct termios tty;
 279
 280     tcgetattr(fd, &tty);
 281
 282     if (echo) {
 283         tty.c_lflag |= ECHO | ECHONL | ICANON | IEXTEN;
 284     } else {
 285         tty.c_lflag &= ~(ECHO | ECHONL | ICANON | IEXTEN);
 286     }
 287
 288     tcsetattr(fd, TCSANOW, &tty);
 289 }
 290
 291 static char exec_dir[PATH_MAX];
 292
 293 void qemu_init_exec_dir(const char *argv0)
 294 {
 295     char *dir;
 296     char *p = NULL;
 297     char buf[PATH_MAX];
 298
 299     assert(!exec_dir[0]);
 300
 301 #if defined(__linux__)
 302     {
 303         int len;
 304         len = readlink("/proc/self/exe", buf, sizeof(buf) - 1);
 305         if (len > 0) {
 306             buf[len] = 0;
 307             p = buf;
 308         }
 309     }
 310 #elif defined(__FreeBSD__)
 311     {
 312         static int mib[4] = {CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1};
 313         size_t len = sizeof(buf) - 1;
 314
 315         *buf = '\0';
 316         if (!sysctl(mib, ARRAY_SIZE(mib), buf, &len, NULL, 0) &&
 317             *buf) {
 318             buf[sizeof(buf) - 1] = '\0';
 319             p = buf;
 320         }
 321     }
 322 #endif
 323     /* If we don't have any way of figuring out the actual executable
 324        location then try argv[0].  */
 325     if (!p) {
 326         if (!argv0) {
 327             return;
 328         }
 329         p = realpath(argv0, buf);
 330         if (!p) {
 331             return;
 332         }
 333     }
 334     dir = dirname(p);
 335
 336     pstrcpy(exec_dir, sizeof(exec_dir), dir);
 337 }
 338
 339 char *qemu_get_exec_dir(void)
 340 {
 341     return g_strdup(exec_dir);
 342 }
 343
 344 static sigjmp_buf sigjump;
 345
 346 static void sigbus_handler(int signal)
 347 {
 348     siglongjmp(sigjump, 1);
 349 }
 350
 351 static size_t fd_getpagesize(int fd)
 352 {
 353 #ifdef CONFIG_LINUX
 354     struct statfs fs;
 355     int ret;
 356
 357     if (fd != -1) {
 358         do {
 359             ret = fstatfs(fd, &fs);
 360         } while (ret != 0 && errno == EINTR);
 361
 362         if (ret == 0 && fs.f_type == HUGETLBFS_MAGIC) {
 363             return fs.f_bsize;
 364         }
 365     }
 366 #endif
 367
 368     return getpagesize();
 369 }
 370
 371 void os_mem_prealloc(int fd, char *area, size_t memory)
 372 {
 373     int ret;
 374     struct sigaction act, oldact;
 375     sigset_t set, oldset;
 376
 377     memset(&act, 0, sizeof(act));
 378     act.sa_handler = &sigbus_handler;
 379     act.sa_flags = 0;
 380
 381     ret = sigaction(SIGBUS, &act, &oldact);
 382     if (ret) {
 383         perror("os_mem_prealloc: failed to install signal handler");
 384         exit(1);
 385     }
 386
 387     /* unblock SIGBUS */
 388     sigemptyset(&set);
 389     sigaddset(&set, SIGBUS);
 390     pthread_sigmask(SIG_UNBLOCK, &set, &oldset);
 391
 392     if (sigsetjmp(sigjump, 1)) {
 393         fprintf(stderr, "os_mem_prealloc: failed to preallocate pages\n");
 394         exit(1);
 395     } else {
 396         int i;
 397         size_t hpagesize = fd_getpagesize(fd);
 398
 399         /* MAP_POPULATE silently ignores failures */
 400         memory = (memory + hpagesize - 1) & -hpagesize;
 401         for (i = 0; i < (memory / hpagesize); i++) {
 402             memset(area + (hpagesize * i), 0, 1);
 403         }
 404
 405         ret = sigaction(SIGBUS, &oldact, NULL);
 406         if (ret) {
 407             perror("os_mem_prealloc: failed to reinstall signal handler");
 408             exit(1);
 409         }
 410
 411         pthread_sigmask(SIG_SETMASK, &oldset, NULL);
 412     }
 413 }