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_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 fprintf(stderr, "Failed to allocate %zu B: %s\n", 110 size, strerror(ret)); 111 abort(); 112 } 113 #elif defined(CONFIG_BSD) 114 ptr = qemu_oom_check(valloc(size)); 115 #else 116 ptr = qemu_oom_check(memalign(alignment, size)); 117 #endif 118 trace_qemu_memalign(alignment, size, ptr); 119 return ptr; 120 } 121 122 /* alloc shared memory pages */ 123 void *qemu_anon_ram_alloc(size_t size) 124 { 125 size_t align = QEMU_VMALLOC_ALIGN; 126 size_t total = size + align - getpagesize(); 127 void *ptr = mmap(0, total, PROT_READ | PROT_WRITE, 128 MAP_ANONYMOUS | MAP_PRIVATE, -1, 0); 129 size_t offset = QEMU_ALIGN_UP((uintptr_t)ptr, align) - (uintptr_t)ptr; 130 131 if (ptr == MAP_FAILED) { 132 return NULL; 133 } 134 135 ptr += offset; 136 total -= offset; 137 138 if (offset > 0) { 139 munmap(ptr - offset, offset); 140 } 141 if (total > size) { 142 munmap(ptr + size, total - size); 143 } 144 145 trace_qemu_anon_ram_alloc(size, ptr); 146 return ptr; 147 } 148 149 void qemu_vfree(void *ptr) 150 { 151 trace_qemu_vfree(ptr); 152 free(ptr); 153 } 154 155 void qemu_anon_ram_free(void *ptr, size_t size) 156 { 157 trace_qemu_anon_ram_free(ptr, size); 158 if (ptr) { 159 munmap(ptr, size); 160 } 161 } 162 163 void qemu_set_block(int fd) 164 { 165 int f; 166 f = fcntl(fd, F_GETFL); 167 fcntl(fd, F_SETFL, f & ~O_NONBLOCK); 168 } 169 170 void qemu_set_nonblock(int fd) 171 { 172 int f; 173 f = fcntl(fd, F_GETFL); 174 fcntl(fd, F_SETFL, f | O_NONBLOCK); 175 } 176 177 int socket_set_fast_reuse(int fd) 178 { 179 int val = 1, ret; 180 181 ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, 182 (const char *)&val, sizeof(val)); 183 184 assert(ret == 0); 185 186 return ret; 187 } 188 189 void qemu_set_cloexec(int fd) 190 { 191 int f; 192 f = fcntl(fd, F_GETFD); 193 fcntl(fd, F_SETFD, f | FD_CLOEXEC); 194 } 195 196 /* 197 * Creates a pipe with FD_CLOEXEC set on both file descriptors 198 */ 199 int qemu_pipe(int pipefd[2]) 200 { 201 int ret; 202 203 #ifdef CONFIG_PIPE2 204 ret = pipe2(pipefd, O_CLOEXEC); 205 if (ret != -1 || errno != ENOSYS) { 206 return ret; 207 } 208 #endif 209 ret = pipe(pipefd); 210 if (ret == 0) { 211 qemu_set_cloexec(pipefd[0]); 212 qemu_set_cloexec(pipefd[1]); 213 } 214 215 return ret; 216 } 217 218 int qemu_utimens(const char *path, const struct timespec *times) 219 { 220 struct timeval tv[2], tv_now; 221 struct stat st; 222 int i; 223 #ifdef CONFIG_UTIMENSAT 224 int ret; 225 226 ret = utimensat(AT_FDCWD, path, times, AT_SYMLINK_NOFOLLOW); 227 if (ret != -1 || errno != ENOSYS) { 228 return ret; 229 } 230 #endif 231 /* Fallback: use utimes() instead of utimensat() */ 232 233 /* happy if special cases */ 234 if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT) { 235 return 0; 236 } 237 if (times[0].tv_nsec == UTIME_NOW && times[1].tv_nsec == UTIME_NOW) { 238 return utimes(path, NULL); 239 } 240 241 /* prepare for hard cases */ 242 if (times[0].tv_nsec == UTIME_NOW || times[1].tv_nsec == UTIME_NOW) { 243 gettimeofday(&tv_now, NULL); 244 } 245 if (times[0].tv_nsec == UTIME_OMIT || times[1].tv_nsec == UTIME_OMIT) { 246 stat(path, &st); 247 } 248 249 for (i = 0; i < 2; i++) { 250 if (times[i].tv_nsec == UTIME_NOW) { 251 tv[i].tv_sec = tv_now.tv_sec; 252 tv[i].tv_usec = tv_now.tv_usec; 253 } else if (times[i].tv_nsec == UTIME_OMIT) { 254 tv[i].tv_sec = (i == 0) ? st.st_atime : st.st_mtime; 255 tv[i].tv_usec = 0; 256 } else { 257 tv[i].tv_sec = times[i].tv_sec; 258 tv[i].tv_usec = times[i].tv_nsec / 1000; 259 } 260 } 261 262 return utimes(path, &tv[0]); 263 } 264 265 char * 266 qemu_get_local_state_pathname(const char *relative_pathname) 267 { 268 return g_strdup_printf("%s/%s", CONFIG_QEMU_LOCALSTATEDIR, 269 relative_pathname); 270 } 271 272 void qemu_set_tty_echo(int fd, bool echo) 273 { 274 struct termios tty; 275 276 tcgetattr(fd, &tty); 277 278 if (echo) { 279 tty.c_lflag |= ECHO | ECHONL | ICANON | IEXTEN; 280 } else { 281 tty.c_lflag &= ~(ECHO | ECHONL | ICANON | IEXTEN); 282 } 283 284 tcsetattr(fd, TCSANOW, &tty); 285 } 286 287 static char exec_dir[PATH_MAX]; 288 289 void qemu_init_exec_dir(const char *argv0) 290 { 291 char *dir; 292 char *p = NULL; 293 char buf[PATH_MAX]; 294 295 assert(!exec_dir[0]); 296 297 #if defined(__linux__) 298 { 299 int len; 300 len = readlink("/proc/self/exe", buf, sizeof(buf) - 1); 301 if (len > 0) { 302 buf[len] = 0; 303 p = buf; 304 } 305 } 306 #elif defined(__FreeBSD__) 307 { 308 static int mib[4] = {CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1}; 309 size_t len = sizeof(buf) - 1; 310 311 *buf = '\0'; 312 if (!sysctl(mib, ARRAY_SIZE(mib), buf, &len, NULL, 0) && 313 *buf) { 314 buf[sizeof(buf) - 1] = '\0'; 315 p = buf; 316 } 317 } 318 #endif 319 /* If we don't have any way of figuring out the actual executable 320 location then try argv[0]. */ 321 if (!p) { 322 if (!argv0) { 323 return; 324 } 325 p = realpath(argv0, buf); 326 if (!p) { 327 return; 328 } 329 } 330 dir = dirname(p); 331 332 pstrcpy(exec_dir, sizeof(exec_dir), dir); 333 } 334 335 char *qemu_get_exec_dir(void) 336 { 337 return g_strdup(exec_dir); 338 } 339 340 static sigjmp_buf sigjump; 341 342 static void sigbus_handler(int signal) 343 { 344 siglongjmp(sigjump, 1); 345 } 346 347 static size_t fd_getpagesize(int fd) 348 { 349 #ifdef CONFIG_LINUX 350 struct statfs fs; 351 int ret; 352 353 if (fd != -1) { 354 do { 355 ret = fstatfs(fd, &fs); 356 } while (ret != 0 && errno == EINTR); 357 358 if (ret == 0 && fs.f_type == HUGETLBFS_MAGIC) { 359 return fs.f_bsize; 360 } 361 } 362 #endif 363 364 return getpagesize(); 365 } 366 367 void os_mem_prealloc(int fd, char *area, size_t memory) 368 { 369 int ret, i; 370 struct sigaction act, oldact; 371 sigset_t set, oldset; 372 size_t hpagesize = fd_getpagesize(fd); 373 374 memset(&act, 0, sizeof(act)); 375 act.sa_handler = &sigbus_handler; 376 act.sa_flags = 0; 377 378 ret = sigaction(SIGBUS, &act, &oldact); 379 if (ret) { 380 perror("os_mem_prealloc: failed to install signal handler"); 381 exit(1); 382 } 383 384 /* unblock SIGBUS */ 385 sigemptyset(&set); 386 sigaddset(&set, SIGBUS); 387 pthread_sigmask(SIG_UNBLOCK, &set, &oldset); 388 389 if (sigsetjmp(sigjump, 1)) { 390 fprintf(stderr, "os_mem_prealloc: failed to preallocate pages\n"); 391 exit(1); 392 } 393 394 /* MAP_POPULATE silently ignores failures */ 395 memory = (memory + hpagesize - 1) & -hpagesize; 396 for (i = 0; i < (memory/hpagesize); i++) { 397 memset(area + (hpagesize*i), 0, 1); 398 } 399 400 ret = sigaction(SIGBUS, &oldact, NULL); 401 if (ret) { 402 perror("os_mem_prealloc: failed to reinstall signal handler"); 403 exit(1); 404 } 405 406 pthread_sigmask(SIG_SETMASK, &oldset, NULL); 407 } 408