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, uint64_t *alignment)
 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     if (alignment) {
 140         *alignment = align;
 141     }
 142     ptr += offset;
 143     total -= offset;
 144
 145     if (offset > 0) {
 146         munmap(ptr - offset, offset);
 147     }
 148     if (total > size) {
 149         munmap(ptr + size, total - size);
 150     }
 151
 152     trace_qemu_anon_ram_alloc(size, ptr);
 153     return ptr;
 154 }
 155
 156 void qemu_vfree(void *ptr)
 157 {
 158     trace_qemu_vfree(ptr);
 159     free(ptr);
 160 }
 161
 162 void qemu_anon_ram_free(void *ptr, size_t size)
 163 {
 164     trace_qemu_anon_ram_free(ptr, size);
 165     if (ptr) {
 166         munmap(ptr, size);
 167     }
 168 }
 169
 170 void qemu_set_block(int fd)
 171 {
 172     int f;
 173     f = fcntl(fd, F_GETFL);
 174     fcntl(fd, F_SETFL, f & ~O_NONBLOCK);
 175 }
 176
 177 void qemu_set_nonblock(int fd)
 178 {
 179     int f;
 180     f = fcntl(fd, F_GETFL);
 181     fcntl(fd, F_SETFL, f | O_NONBLOCK);
 182 }
 183
 184 int socket_set_fast_reuse(int fd)
 185 {
 186     int val = 1, ret;
 187
 188     ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
 189                      (const char *)&val, sizeof(val));
 190
 191     assert(ret == 0);
 192
 193     return ret;
 194 }
 195
 196 void qemu_set_cloexec(int fd)
 197 {
 198     int f;
 199     f = fcntl(fd, F_GETFD);
 200     fcntl(fd, F_SETFD, f | FD_CLOEXEC);
 201 }
 202
 203 /*
 204  * Creates a pipe with FD_CLOEXEC set on both file descriptors
 205  */
 206 int qemu_pipe(int pipefd[2])
 207 {
 208     int ret;
 209
 210 #ifdef CONFIG_PIPE2
 211     ret = pipe2(pipefd, O_CLOEXEC);
 212     if (ret != -1 || errno != ENOSYS) {
 213         return ret;
 214     }
 215 #endif
 216     ret = pipe(pipefd);
 217     if (ret == 0) {
 218         qemu_set_cloexec(pipefd[0]);
 219         qemu_set_cloexec(pipefd[1]);
 220     }
 221
 222     return ret;
 223 }
 224
 225 int qemu_utimens(const char *path, const struct timespec *times)
 226 {
 227     struct timeval tv[2], tv_now;
 228     struct stat st;
 229     int i;
 230 #ifdef CONFIG_UTIMENSAT
 231     int ret;
 232
 233     ret = utimensat(AT_FDCWD, path, times, AT_SYMLINK_NOFOLLOW);
 234     if (ret != -1 || errno != ENOSYS) {
 235         return ret;
 236     }
 237 #endif
 238     /* Fallback: use utimes() instead of utimensat() */
 239
 240     /* happy if special cases */
 241     if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT) {
 242         return 0;
 243     }
 244     if (times[0].tv_nsec == UTIME_NOW && times[1].tv_nsec == UTIME_NOW) {
 245         return utimes(path, NULL);
 246     }
 247
 248     /* prepare for hard cases */
 249     if (times[0].tv_nsec == UTIME_NOW || times[1].tv_nsec == UTIME_NOW) {
 250         gettimeofday(&tv_now, NULL);
 251     }
 252     if (times[0].tv_nsec == UTIME_OMIT || times[1].tv_nsec == UTIME_OMIT) {
 253         stat(path, &st);
 254     }
 255
 256     for (i = 0; i < 2; i++) {
 257         if (times[i].tv_nsec == UTIME_NOW) {
 258             tv[i].tv_sec = tv_now.tv_sec;
 259             tv[i].tv_usec = tv_now.tv_usec;
 260         } else if (times[i].tv_nsec == UTIME_OMIT) {
 261             tv[i].tv_sec = (i == 0) ? st.st_atime : st.st_mtime;
 262             tv[i].tv_usec = 0;
 263         } else {
 264             tv[i].tv_sec = times[i].tv_sec;
 265             tv[i].tv_usec = times[i].tv_nsec / 1000;
 266         }
 267     }
 268
 269     return utimes(path, &tv[0]);
 270 }
 271
 272 char *
 273 qemu_get_local_state_pathname(const char *relative_pathname)
 274 {
 275     return g_strdup_printf("%s/%s", CONFIG_QEMU_LOCALSTATEDIR,
 276                            relative_pathname);
 277 }
 278
 279 void qemu_set_tty_echo(int fd, bool echo)
 280 {
 281     struct termios tty;
 282
 283     tcgetattr(fd, &tty);
 284
 285     if (echo) {
 286         tty.c_lflag |= ECHO | ECHONL | ICANON | IEXTEN;
 287     } else {
 288         tty.c_lflag &= ~(ECHO | ECHONL | ICANON | IEXTEN);
 289     }
 290
 291     tcsetattr(fd, TCSANOW, &tty);
 292 }
 293
 294 static char exec_dir[PATH_MAX];
 295
 296 void qemu_init_exec_dir(const char *argv0)
 297 {
 298     char *dir;
 299     char *p = NULL;
 300     char buf[PATH_MAX];
 301
 302     assert(!exec_dir[0]);
 303
 304 #if defined(__linux__)
 305     {
 306         int len;
 307         len = readlink("/proc/self/exe", buf, sizeof(buf) - 1);
 308         if (len > 0) {
 309             buf[len] = 0;
 310             p = buf;
 311         }
 312     }
 313 #elif defined(__FreeBSD__)
 314     {
 315         static int mib[4] = {CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1};
 316         size_t len = sizeof(buf) - 1;
 317
 318         *buf = '\0';
 319         if (!sysctl(mib, ARRAY_SIZE(mib), buf, &len, NULL, 0) &&
 320             *buf) {
 321             buf[sizeof(buf) - 1] = '\0';
 322             p = buf;
 323         }
 324     }
 325 #endif
 326     /* If we don't have any way of figuring out the actual executable
 327        location then try argv[0].  */
 328     if (!p) {
 329         if (!argv0) {
 330             return;
 331         }
 332         p = realpath(argv0, buf);
 333         if (!p) {
 334             return;
 335         }
 336     }
 337     dir = dirname(p);
 338
 339     pstrcpy(exec_dir, sizeof(exec_dir), dir);
 340 }
 341
 342 char *qemu_get_exec_dir(void)
 343 {
 344     return g_strdup(exec_dir);
 345 }
 346
 347 static sigjmp_buf sigjump;
 348
 349 static void sigbus_handler(int signal)
 350 {
 351     siglongjmp(sigjump, 1);
 352 }
 353
 354 static size_t fd_getpagesize(int fd)
 355 {
 356 #ifdef CONFIG_LINUX
 357     struct statfs fs;
 358     int ret;
 359
 360     if (fd != -1) {
 361         do {
 362             ret = fstatfs(fd, &fs);
 363         } while (ret != 0 && errno == EINTR);
 364
 365         if (ret == 0 && fs.f_type == HUGETLBFS_MAGIC) {
 366             return fs.f_bsize;
 367         }
 368     }
 369 #endif
 370
 371     return getpagesize();
 372 }
 373
 374 void os_mem_prealloc(int fd, char *area, size_t memory)
 375 {
 376     int ret;
 377     struct sigaction act, oldact;
 378     sigset_t set, oldset;
 379
 380     memset(&act, 0, sizeof(act));
 381     act.sa_handler = &sigbus_handler;
 382     act.sa_flags = 0;
 383
 384     ret = sigaction(SIGBUS, &act, &oldact);
 385     if (ret) {
 386         perror("os_mem_prealloc: failed to install signal handler");
 387         exit(1);
 388     }
 389
 390     /* unblock SIGBUS */
 391     sigemptyset(&set);
 392     sigaddset(&set, SIGBUS);
 393     pthread_sigmask(SIG_UNBLOCK, &set, &oldset);
 394
 395     if (sigsetjmp(sigjump, 1)) {
 396         fprintf(stderr, "os_mem_prealloc: Insufficient free host memory "
 397                         "pages available to allocate guest RAM\n");
 398         exit(1);
 399     } else {
 400         int i;
 401         size_t hpagesize = fd_getpagesize(fd);
 402         size_t numpages = DIV_ROUND_UP(memory, hpagesize);
 403
 404         /* MAP_POPULATE silently ignores failures */
 405         for (i = 0; i < numpages; i++) {
 406             memset(area + (hpagesize * i), 0, 1);
 407         }
 408
 409         ret = sigaction(SIGBUS, &oldact, NULL);
 410         if (ret) {
 411             perror("os_mem_prealloc: failed to reinstall signal handler");
 412             exit(1);
 413         }
 414
 415         pthread_sigmask(SIG_SETMASK, &oldset, NULL);
 416     }
 417 }