]> git.proxmox.com Git - mirror_qemu.git/blob - util/oslib-posix.c
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
[mirror_qemu.git] / 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 #include <termios.h>
54
55 #include <glib/gprintf.h>
56
57 #include "config-host.h"
58 #include "sysemu/sysemu.h"
59 #include "trace.h"
60 #include "qemu/sockets.h"
61 #include <sys/mman.h>
62 #include <libgen.h>
63 #include <setjmp.h>
64 #include <sys/signal.h>
65
66 #ifdef CONFIG_LINUX
67 #include <sys/syscall.h>
68 #include <sys/vfs.h>
69 #endif
70
71 #ifdef __FreeBSD__
72 #include <sys/sysctl.h>
73 #endif
74
75 int qemu_get_thread_id(void)
76 {
77 #if defined(__linux__)
78 return syscall(SYS_gettid);
79 #else
80 return getpid();
81 #endif
82 }
83
84 int qemu_daemon(int nochdir, int noclose)
85 {
86 return daemon(nochdir, noclose);
87 }
88
89 void *qemu_oom_check(void *ptr)
90 {
91 if (ptr == NULL) {
92 fprintf(stderr, "Failed to allocate memory: %s\n", strerror(errno));
93 abort();
94 }
95 return ptr;
96 }
97
98 void *qemu_try_memalign(size_t alignment, size_t size)
99 {
100 void *ptr;
101
102 if (alignment < sizeof(void*)) {
103 alignment = sizeof(void*);
104 }
105
106 #if defined(_POSIX_C_SOURCE) && !defined(__sun__)
107 int ret;
108 ret = posix_memalign(&ptr, alignment, size);
109 if (ret != 0) {
110 errno = ret;
111 ptr = NULL;
112 }
113 #elif defined(CONFIG_BSD)
114 ptr = valloc(size);
115 #else
116 ptr = memalign(alignment, size);
117 #endif
118 trace_qemu_memalign(alignment, size, ptr);
119 return ptr;
120 }
121
122 void *qemu_memalign(size_t alignment, size_t size)
123 {
124 return qemu_oom_check(qemu_try_memalign(alignment, size));
125 }
126
127 /* alloc shared memory pages */
128 void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment)
129 {
130 size_t align = QEMU_VMALLOC_ALIGN;
131 size_t total = size + align - getpagesize();
132 void *ptr = mmap(0, total, PROT_READ | PROT_WRITE,
133 MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
134 size_t offset = QEMU_ALIGN_UP((uintptr_t)ptr, align) - (uintptr_t)ptr;
135
136 if (ptr == MAP_FAILED) {
137 return NULL;
138 }
139
140 if (alignment) {
141 *alignment = align;
142 }
143 ptr += offset;
144 total -= offset;
145
146 if (offset > 0) {
147 munmap(ptr - offset, offset);
148 }
149 if (total > size) {
150 munmap(ptr + size, total - size);
151 }
152
153 trace_qemu_anon_ram_alloc(size, ptr);
154 return ptr;
155 }
156
157 void qemu_vfree(void *ptr)
158 {
159 trace_qemu_vfree(ptr);
160 free(ptr);
161 }
162
163 void qemu_anon_ram_free(void *ptr, size_t size)
164 {
165 trace_qemu_anon_ram_free(ptr, size);
166 if (ptr) {
167 munmap(ptr, size);
168 }
169 }
170
171 void qemu_set_block(int fd)
172 {
173 int f;
174 f = fcntl(fd, F_GETFL);
175 fcntl(fd, F_SETFL, f & ~O_NONBLOCK);
176 }
177
178 void qemu_set_nonblock(int fd)
179 {
180 int f;
181 f = fcntl(fd, F_GETFL);
182 fcntl(fd, F_SETFL, f | O_NONBLOCK);
183 }
184
185 int socket_set_fast_reuse(int fd)
186 {
187 int val = 1, ret;
188
189 ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
190 (const char *)&val, sizeof(val));
191
192 assert(ret == 0);
193
194 return ret;
195 }
196
197 void qemu_set_cloexec(int fd)
198 {
199 int f;
200 f = fcntl(fd, F_GETFD);
201 fcntl(fd, F_SETFD, f | FD_CLOEXEC);
202 }
203
204 /*
205 * Creates a pipe with FD_CLOEXEC set on both file descriptors
206 */
207 int qemu_pipe(int pipefd[2])
208 {
209 int ret;
210
211 #ifdef CONFIG_PIPE2
212 ret = pipe2(pipefd, O_CLOEXEC);
213 if (ret != -1 || errno != ENOSYS) {
214 return ret;
215 }
216 #endif
217 ret = pipe(pipefd);
218 if (ret == 0) {
219 qemu_set_cloexec(pipefd[0]);
220 qemu_set_cloexec(pipefd[1]);
221 }
222
223 return ret;
224 }
225
226 int qemu_utimens(const char *path, const struct timespec *times)
227 {
228 struct timeval tv[2], tv_now;
229 struct stat st;
230 int i;
231 #ifdef CONFIG_UTIMENSAT
232 int ret;
233
234 ret = utimensat(AT_FDCWD, path, times, AT_SYMLINK_NOFOLLOW);
235 if (ret != -1 || errno != ENOSYS) {
236 return ret;
237 }
238 #endif
239 /* Fallback: use utimes() instead of utimensat() */
240
241 /* happy if special cases */
242 if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT) {
243 return 0;
244 }
245 if (times[0].tv_nsec == UTIME_NOW && times[1].tv_nsec == UTIME_NOW) {
246 return utimes(path, NULL);
247 }
248
249 /* prepare for hard cases */
250 if (times[0].tv_nsec == UTIME_NOW || times[1].tv_nsec == UTIME_NOW) {
251 gettimeofday(&tv_now, NULL);
252 }
253 if (times[0].tv_nsec == UTIME_OMIT || times[1].tv_nsec == UTIME_OMIT) {
254 stat(path, &st);
255 }
256
257 for (i = 0; i < 2; i++) {
258 if (times[i].tv_nsec == UTIME_NOW) {
259 tv[i].tv_sec = tv_now.tv_sec;
260 tv[i].tv_usec = tv_now.tv_usec;
261 } else if (times[i].tv_nsec == UTIME_OMIT) {
262 tv[i].tv_sec = (i == 0) ? st.st_atime : st.st_mtime;
263 tv[i].tv_usec = 0;
264 } else {
265 tv[i].tv_sec = times[i].tv_sec;
266 tv[i].tv_usec = times[i].tv_nsec / 1000;
267 }
268 }
269
270 return utimes(path, &tv[0]);
271 }
272
273 char *
274 qemu_get_local_state_pathname(const char *relative_pathname)
275 {
276 return g_strdup_printf("%s/%s", CONFIG_QEMU_LOCALSTATEDIR,
277 relative_pathname);
278 }
279
280 void qemu_set_tty_echo(int fd, bool echo)
281 {
282 struct termios tty;
283
284 tcgetattr(fd, &tty);
285
286 if (echo) {
287 tty.c_lflag |= ECHO | ECHONL | ICANON | IEXTEN;
288 } else {
289 tty.c_lflag &= ~(ECHO | ECHONL | ICANON | IEXTEN);
290 }
291
292 tcsetattr(fd, TCSANOW, &tty);
293 }
294
295 static char exec_dir[PATH_MAX];
296
297 void qemu_init_exec_dir(const char *argv0)
298 {
299 char *dir;
300 char *p = NULL;
301 char buf[PATH_MAX];
302
303 assert(!exec_dir[0]);
304
305 #if defined(__linux__)
306 {
307 int len;
308 len = readlink("/proc/self/exe", buf, sizeof(buf) - 1);
309 if (len > 0) {
310 buf[len] = 0;
311 p = buf;
312 }
313 }
314 #elif defined(__FreeBSD__)
315 {
316 static int mib[4] = {CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1};
317 size_t len = sizeof(buf) - 1;
318
319 *buf = '\0';
320 if (!sysctl(mib, ARRAY_SIZE(mib), buf, &len, NULL, 0) &&
321 *buf) {
322 buf[sizeof(buf) - 1] = '\0';
323 p = buf;
324 }
325 }
326 #endif
327 /* If we don't have any way of figuring out the actual executable
328 location then try argv[0]. */
329 if (!p) {
330 if (!argv0) {
331 return;
332 }
333 p = realpath(argv0, buf);
334 if (!p) {
335 return;
336 }
337 }
338 dir = dirname(p);
339
340 pstrcpy(exec_dir, sizeof(exec_dir), dir);
341 }
342
343 char *qemu_get_exec_dir(void)
344 {
345 return g_strdup(exec_dir);
346 }
347
348 static sigjmp_buf sigjump;
349
350 static void sigbus_handler(int signal)
351 {
352 siglongjmp(sigjump, 1);
353 }
354
355 static size_t fd_getpagesize(int fd)
356 {
357 #ifdef CONFIG_LINUX
358 struct statfs fs;
359 int ret;
360
361 if (fd != -1) {
362 do {
363 ret = fstatfs(fd, &fs);
364 } while (ret != 0 && errno == EINTR);
365
366 if (ret == 0 && fs.f_type == HUGETLBFS_MAGIC) {
367 return fs.f_bsize;
368 }
369 }
370 #endif
371
372 return getpagesize();
373 }
374
375 void os_mem_prealloc(int fd, char *area, size_t memory)
376 {
377 int ret;
378 struct sigaction act, oldact;
379 sigset_t set, oldset;
380
381 memset(&act, 0, sizeof(act));
382 act.sa_handler = &sigbus_handler;
383 act.sa_flags = 0;
384
385 ret = sigaction(SIGBUS, &act, &oldact);
386 if (ret) {
387 perror("os_mem_prealloc: failed to install signal handler");
388 exit(1);
389 }
390
391 /* unblock SIGBUS */
392 sigemptyset(&set);
393 sigaddset(&set, SIGBUS);
394 pthread_sigmask(SIG_UNBLOCK, &set, &oldset);
395
396 if (sigsetjmp(sigjump, 1)) {
397 fprintf(stderr, "os_mem_prealloc: Insufficient free host memory "
398 "pages available to allocate guest RAM\n");
399 exit(1);
400 } else {
401 int i;
402 size_t hpagesize = fd_getpagesize(fd);
403 size_t numpages = DIV_ROUND_UP(memory, hpagesize);
404
405 /* MAP_POPULATE silently ignores failures */
406 for (i = 0; i < numpages; i++) {
407 memset(area + (hpagesize * i), 0, 1);
408 }
409
410 ret = sigaction(SIGBUS, &oldact, NULL);
411 if (ret) {
412 perror("os_mem_prealloc: failed to reinstall signal handler");
413 exit(1);
414 }
415
416 pthread_sigmask(SIG_SETMASK, &oldset, NULL);
417 }
418 }
419
420
421 static struct termios oldtty;
422
423 static void term_exit(void)
424 {
425 tcsetattr(0, TCSANOW, &oldtty);
426 }
427
428 static void term_init(void)
429 {
430 struct termios tty;
431
432 tcgetattr(0, &tty);
433 oldtty = tty;
434
435 tty.c_iflag &= ~(IGNBRK|BRKINT|PARMRK|ISTRIP
436 |INLCR|IGNCR|ICRNL|IXON);
437 tty.c_oflag |= OPOST;
438 tty.c_lflag &= ~(ECHO|ECHONL|ICANON|IEXTEN);
439 tty.c_cflag &= ~(CSIZE|PARENB);
440 tty.c_cflag |= CS8;
441 tty.c_cc[VMIN] = 1;
442 tty.c_cc[VTIME] = 0;
443
444 tcsetattr(0, TCSANOW, &tty);
445
446 atexit(term_exit);
447 }
448
449 int qemu_read_password(char *buf, int buf_size)
450 {
451 uint8_t ch;
452 int i, ret;
453
454 printf("password: ");
455 fflush(stdout);
456 term_init();
457 i = 0;
458 for (;;) {
459 ret = read(0, &ch, 1);
460 if (ret == -1) {
461 if (errno == EAGAIN || errno == EINTR) {
462 continue;
463 } else {
464 break;
465 }
466 } else if (ret == 0) {
467 ret = -1;
468 break;
469 } else {
470 if (ch == '\r' ||
471 ch == '\n') {
472 ret = 0;
473 break;
474 }
475 if (i < (buf_size - 1)) {
476 buf[i++] = ch;
477 }
478 }
479 }
480 term_exit();
481 buf[i] = '\0';
482 printf("\n");
483 return ret;
484 }