]>
Commit | Line | Data |
---|---|---|
c1b0b93b JS |
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 | ||
aafd7584 | 29 | #include "qemu/osdep.h" |
13401ba0 | 30 | #include <termios.h> |
13401ba0 | 31 | |
e2ea3515 LE |
32 | #include <glib/gprintf.h> |
33 | ||
a8d25326 | 34 | #include "qemu-common.h" |
9c17d615 | 35 | #include "sysemu/sysemu.h" |
c1b0b93b | 36 | #include "trace.h" |
da34e65c | 37 | #include "qapi/error.h" |
1de7afc9 | 38 | #include "qemu/sockets.h" |
db725815 | 39 | #include "qemu/thread.h" |
10f5bff6 | 40 | #include <libgen.h> |
38183310 | 41 | #include <sys/signal.h> |
f348b6d1 | 42 | #include "qemu/cutils.h" |
c1b0b93b | 43 | |
cbcfa041 PB |
44 | #ifdef CONFIG_LINUX |
45 | #include <sys/syscall.h> | |
46 | #endif | |
cbcfa041 | 47 | |
41975b26 AF |
48 | #ifdef __FreeBSD__ |
49 | #include <sys/sysctl.h> | |
a7764f15 | 50 | #include <sys/user.h> |
7dc9ae43 | 51 | #include <libutil.h> |
41975b26 AF |
52 | #endif |
53 | ||
094611b4 KR |
54 | #ifdef __NetBSD__ |
55 | #include <sys/sysctl.h> | |
56 | #endif | |
57 | ||
a9c94277 | 58 | #include "qemu/mmap-alloc.h" |
794e8f30 | 59 | |
7d992e4d PL |
60 | #ifdef CONFIG_DEBUG_STACK_USAGE |
61 | #include "qemu/error-report.h" | |
62 | #endif | |
63 | ||
dfd0dcc7 | 64 | #define MAX_MEM_PREALLOC_THREAD_COUNT 16 |
1e356fc1 JK |
65 | |
66 | struct MemsetThread { | |
67 | char *addr; | |
e947d47d SW |
68 | size_t numpages; |
69 | size_t hpagesize; | |
1e356fc1 JK |
70 | QemuThread pgthread; |
71 | sigjmp_buf env; | |
72 | }; | |
73 | typedef struct MemsetThread MemsetThread; | |
74 | ||
75 | static MemsetThread *memset_thread; | |
76 | static int memset_num_threads; | |
77 | static bool memset_thread_failed; | |
78 | ||
037fb5eb | 79 | static QemuMutex page_mutex; |
80 | static QemuCond page_cond; | |
81 | static bool threads_created_flag; | |
82 | ||
cbcfa041 PB |
83 | int qemu_get_thread_id(void) |
84 | { | |
85 | #if defined(__linux__) | |
86 | return syscall(SYS_gettid); | |
87 | #else | |
88 | return getpid(); | |
89 | #endif | |
90 | } | |
f97742d0 AR |
91 | |
92 | int qemu_daemon(int nochdir, int noclose) | |
93 | { | |
94 | return daemon(nochdir, noclose); | |
95 | } | |
96 | ||
9e6bdef2 MAL |
97 | bool qemu_write_pidfile(const char *path, Error **errp) |
98 | { | |
99 | int fd; | |
100 | char pidstr[32]; | |
101 | ||
102 | while (1) { | |
103 | struct stat a, b; | |
35f7f3fb MAL |
104 | struct flock lock = { |
105 | .l_type = F_WRLCK, | |
106 | .l_whence = SEEK_SET, | |
107 | .l_len = 0, | |
108 | }; | |
9e6bdef2 MAL |
109 | |
110 | fd = qemu_open(path, O_CREAT | O_WRONLY, S_IRUSR | S_IWUSR); | |
111 | if (fd == -1) { | |
112 | error_setg_errno(errp, errno, "Cannot open pid file"); | |
113 | return false; | |
114 | } | |
115 | ||
116 | if (fstat(fd, &b) < 0) { | |
117 | error_setg_errno(errp, errno, "Cannot stat file"); | |
118 | goto fail_close; | |
119 | } | |
120 | ||
35f7f3fb | 121 | if (fcntl(fd, F_SETLK, &lock)) { |
9e6bdef2 MAL |
122 | error_setg_errno(errp, errno, "Cannot lock pid file"); |
123 | goto fail_close; | |
124 | } | |
125 | ||
126 | /* | |
127 | * Now make sure the path we locked is the same one that now | |
128 | * exists on the filesystem. | |
129 | */ | |
130 | if (stat(path, &a) < 0) { | |
131 | /* | |
132 | * PID file disappeared, someone else must be racing with | |
133 | * us, so try again. | |
134 | */ | |
135 | close(fd); | |
136 | continue; | |
137 | } | |
138 | ||
139 | if (a.st_ino == b.st_ino) { | |
140 | break; | |
141 | } | |
142 | ||
143 | /* | |
144 | * PID file was recreated, someone else must be racing with | |
145 | * us, so try again. | |
146 | */ | |
147 | close(fd); | |
148 | } | |
149 | ||
150 | if (ftruncate(fd, 0) < 0) { | |
151 | error_setg_errno(errp, errno, "Failed to truncate pid file"); | |
152 | goto fail_unlink; | |
153 | } | |
154 | ||
155 | snprintf(pidstr, sizeof(pidstr), FMT_pid "\n", getpid()); | |
156 | if (write(fd, pidstr, strlen(pidstr)) != strlen(pidstr)) { | |
157 | error_setg(errp, "Failed to write pid file"); | |
158 | goto fail_unlink; | |
159 | } | |
160 | ||
161 | return true; | |
162 | ||
163 | fail_unlink: | |
164 | unlink(path); | |
165 | fail_close: | |
166 | close(fd); | |
167 | return false; | |
168 | } | |
169 | ||
b152aa84 | 170 | void *qemu_oom_check(void *ptr) |
c1b0b93b JS |
171 | { |
172 | if (ptr == NULL) { | |
173 | fprintf(stderr, "Failed to allocate memory: %s\n", strerror(errno)); | |
174 | abort(); | |
175 | } | |
176 | return ptr; | |
177 | } | |
c1b0b93b | 178 | |
7d2a35cc | 179 | void *qemu_try_memalign(size_t alignment, size_t size) |
c1b0b93b JS |
180 | { |
181 | void *ptr; | |
e5354657 KW |
182 | |
183 | if (alignment < sizeof(void*)) { | |
184 | alignment = sizeof(void*); | |
185 | } | |
186 | ||
9bc5a719 | 187 | #if defined(CONFIG_POSIX_MEMALIGN) |
c1b0b93b JS |
188 | int ret; |
189 | ret = posix_memalign(&ptr, alignment, size); | |
190 | if (ret != 0) { | |
7d2a35cc KW |
191 | errno = ret; |
192 | ptr = NULL; | |
c1b0b93b JS |
193 | } |
194 | #elif defined(CONFIG_BSD) | |
7d2a35cc | 195 | ptr = valloc(size); |
c1b0b93b | 196 | #else |
7d2a35cc | 197 | ptr = memalign(alignment, size); |
c1b0b93b JS |
198 | #endif |
199 | trace_qemu_memalign(alignment, size, ptr); | |
200 | return ptr; | |
201 | } | |
202 | ||
7d2a35cc KW |
203 | void *qemu_memalign(size_t alignment, size_t size) |
204 | { | |
205 | return qemu_oom_check(qemu_try_memalign(alignment, size)); | |
206 | } | |
207 | ||
c1b0b93b | 208 | /* alloc shared memory pages */ |
06329cce | 209 | void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment, bool shared) |
c1b0b93b | 210 | { |
36b58628 | 211 | size_t align = QEMU_VMALLOC_ALIGN; |
2ac0f162 | 212 | void *ptr = qemu_ram_mmap(-1, size, align, shared, false); |
36b58628 | 213 | |
7dda5dc8 | 214 | if (ptr == MAP_FAILED) { |
39228250 | 215 | return NULL; |
c2a8238a | 216 | } |
c2a8238a | 217 | |
a2b257d6 IM |
218 | if (alignment) { |
219 | *alignment = align; | |
220 | } | |
c2dfc5ba | 221 | |
6eebf958 | 222 | trace_qemu_anon_ram_alloc(size, ptr); |
c7f4111a | 223 | return ptr; |
c1b0b93b JS |
224 | } |
225 | ||
226 | void qemu_vfree(void *ptr) | |
227 | { | |
228 | trace_qemu_vfree(ptr); | |
229 | free(ptr); | |
230 | } | |
9549e764 | 231 | |
e7a09b92 PB |
232 | void qemu_anon_ram_free(void *ptr, size_t size) |
233 | { | |
234 | trace_qemu_anon_ram_free(ptr, size); | |
53adb9d4 | 235 | qemu_ram_munmap(-1, ptr, size); |
e7a09b92 PB |
236 | } |
237 | ||
f9e8cacc | 238 | void qemu_set_block(int fd) |
154b9a0c PB |
239 | { |
240 | int f; | |
241 | f = fcntl(fd, F_GETFL); | |
da93b820 LQ |
242 | assert(f != -1); |
243 | f = fcntl(fd, F_SETFL, f & ~O_NONBLOCK); | |
244 | assert(f != -1); | |
154b9a0c PB |
245 | } |
246 | ||
f9e8cacc | 247 | void qemu_set_nonblock(int fd) |
9549e764 JS |
248 | { |
249 | int f; | |
250 | f = fcntl(fd, F_GETFL); | |
da93b820 LQ |
251 | assert(f != -1); |
252 | f = fcntl(fd, F_SETFL, f | O_NONBLOCK); | |
02cdcc96 PMD |
253 | #ifdef __OpenBSD__ |
254 | if (f == -1) { | |
255 | /* | |
256 | * Previous to OpenBSD 6.3, fcntl(F_SETFL) is not permitted on | |
257 | * memory devices and sets errno to ENODEV. | |
258 | * It's OK if we fail to set O_NONBLOCK on devices like /dev/null, | |
259 | * because they will never block anyway. | |
260 | */ | |
261 | assert(errno == ENODEV); | |
262 | } | |
263 | #else | |
da93b820 | 264 | assert(f != -1); |
02cdcc96 | 265 | #endif |
9549e764 JS |
266 | } |
267 | ||
606600a1 SO |
268 | int socket_set_fast_reuse(int fd) |
269 | { | |
270 | int val = 1, ret; | |
271 | ||
272 | ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, | |
273 | (const char *)&val, sizeof(val)); | |
274 | ||
275 | assert(ret == 0); | |
276 | ||
277 | return ret; | |
278 | } | |
279 | ||
9549e764 JS |
280 | void qemu_set_cloexec(int fd) |
281 | { | |
282 | int f; | |
283 | f = fcntl(fd, F_GETFD); | |
7e6478e7 SS |
284 | assert(f != -1); |
285 | f = fcntl(fd, F_SETFD, f | FD_CLOEXEC); | |
286 | assert(f != -1); | |
9549e764 | 287 | } |
70e72ce4 JS |
288 | |
289 | /* | |
290 | * Creates a pipe with FD_CLOEXEC set on both file descriptors | |
291 | */ | |
292 | int qemu_pipe(int pipefd[2]) | |
293 | { | |
294 | int ret; | |
295 | ||
296 | #ifdef CONFIG_PIPE2 | |
297 | ret = pipe2(pipefd, O_CLOEXEC); | |
298 | if (ret != -1 || errno != ENOSYS) { | |
299 | return ret; | |
300 | } | |
301 | #endif | |
302 | ret = pipe(pipefd); | |
303 | if (ret == 0) { | |
304 | qemu_set_cloexec(pipefd[0]); | |
305 | qemu_set_cloexec(pipefd[1]); | |
306 | } | |
307 | ||
308 | return ret; | |
309 | } | |
38671423 | 310 | |
e2ea3515 LE |
311 | char * |
312 | qemu_get_local_state_pathname(const char *relative_pathname) | |
313 | { | |
314 | return g_strdup_printf("%s/%s", CONFIG_QEMU_LOCALSTATEDIR, | |
315 | relative_pathname); | |
316 | } | |
13401ba0 SH |
317 | |
318 | void qemu_set_tty_echo(int fd, bool echo) | |
319 | { | |
320 | struct termios tty; | |
321 | ||
322 | tcgetattr(fd, &tty); | |
323 | ||
324 | if (echo) { | |
325 | tty.c_lflag |= ECHO | ECHONL | ICANON | IEXTEN; | |
326 | } else { | |
327 | tty.c_lflag &= ~(ECHO | ECHONL | ICANON | IEXTEN); | |
328 | } | |
329 | ||
330 | tcsetattr(fd, TCSANOW, &tty); | |
331 | } | |
10f5bff6 FZ |
332 | |
333 | static char exec_dir[PATH_MAX]; | |
334 | ||
335 | void qemu_init_exec_dir(const char *argv0) | |
336 | { | |
337 | char *dir; | |
338 | char *p = NULL; | |
339 | char buf[PATH_MAX]; | |
340 | ||
341 | assert(!exec_dir[0]); | |
342 | ||
343 | #if defined(__linux__) | |
344 | { | |
345 | int len; | |
346 | len = readlink("/proc/self/exe", buf, sizeof(buf) - 1); | |
347 | if (len > 0) { | |
348 | buf[len] = 0; | |
349 | p = buf; | |
350 | } | |
351 | } | |
094611b4 KR |
352 | #elif defined(__FreeBSD__) \ |
353 | || (defined(__NetBSD__) && defined(KERN_PROC_PATHNAME)) | |
10f5bff6 | 354 | { |
094611b4 | 355 | #if defined(__FreeBSD__) |
10f5bff6 | 356 | static int mib[4] = {CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1}; |
094611b4 KR |
357 | #else |
358 | static int mib[4] = {CTL_KERN, KERN_PROC_ARGS, -1, KERN_PROC_PATHNAME}; | |
359 | #endif | |
10f5bff6 FZ |
360 | size_t len = sizeof(buf) - 1; |
361 | ||
362 | *buf = '\0'; | |
363 | if (!sysctl(mib, ARRAY_SIZE(mib), buf, &len, NULL, 0) && | |
364 | *buf) { | |
365 | buf[sizeof(buf) - 1] = '\0'; | |
366 | p = buf; | |
367 | } | |
368 | } | |
369 | #endif | |
370 | /* If we don't have any way of figuring out the actual executable | |
371 | location then try argv[0]. */ | |
372 | if (!p) { | |
373 | if (!argv0) { | |
374 | return; | |
375 | } | |
376 | p = realpath(argv0, buf); | |
377 | if (!p) { | |
378 | return; | |
379 | } | |
380 | } | |
55ad781c | 381 | dir = g_path_get_dirname(p); |
10f5bff6 FZ |
382 | |
383 | pstrcpy(exec_dir, sizeof(exec_dir), dir); | |
55ad781c WJ |
384 | |
385 | g_free(dir); | |
10f5bff6 FZ |
386 | } |
387 | ||
388 | char *qemu_get_exec_dir(void) | |
389 | { | |
390 | return g_strdup(exec_dir); | |
391 | } | |
38183310 | 392 | |
38183310 PB |
393 | static void sigbus_handler(int signal) |
394 | { | |
1e356fc1 JK |
395 | int i; |
396 | if (memset_thread) { | |
397 | for (i = 0; i < memset_num_threads; i++) { | |
398 | if (qemu_thread_is_self(&memset_thread[i].pgthread)) { | |
399 | siglongjmp(memset_thread[i].env, 1); | |
400 | } | |
401 | } | |
402 | } | |
38183310 PB |
403 | } |
404 | ||
1e356fc1 JK |
405 | static void *do_touch_pages(void *arg) |
406 | { | |
407 | MemsetThread *memset_args = (MemsetThread *)arg; | |
1e356fc1 | 408 | sigset_t set, oldset; |
1e356fc1 | 409 | |
037fb5eb | 410 | /* |
411 | * On Linux, the page faults from the loop below can cause mmap_sem | |
412 | * contention with allocation of the thread stacks. Do not start | |
413 | * clearing until all threads have been created. | |
414 | */ | |
415 | qemu_mutex_lock(&page_mutex); | |
416 | while(!threads_created_flag){ | |
417 | qemu_cond_wait(&page_cond, &page_mutex); | |
418 | } | |
419 | qemu_mutex_unlock(&page_mutex); | |
420 | ||
1e356fc1 JK |
421 | /* unblock SIGBUS */ |
422 | sigemptyset(&set); | |
423 | sigaddset(&set, SIGBUS); | |
424 | pthread_sigmask(SIG_UNBLOCK, &set, &oldset); | |
425 | ||
426 | if (sigsetjmp(memset_args->env, 1)) { | |
427 | memset_thread_failed = true; | |
428 | } else { | |
e947d47d SW |
429 | char *addr = memset_args->addr; |
430 | size_t numpages = memset_args->numpages; | |
431 | size_t hpagesize = memset_args->hpagesize; | |
432 | size_t i; | |
1e356fc1 | 433 | for (i = 0; i < numpages; i++) { |
9dc44aa5 DB |
434 | /* |
435 | * Read & write back the same value, so we don't | |
436 | * corrupt existing user/app data that might be | |
437 | * stored. | |
438 | * | |
439 | * 'volatile' to stop compiler optimizing this away | |
440 | * to a no-op | |
441 | * | |
442 | * TODO: get a better solution from kernel so we | |
443 | * don't need to write at all so we don't cause | |
444 | * wear on the storage backing the region... | |
445 | */ | |
446 | *(volatile char *)addr = *addr; | |
1e356fc1 JK |
447 | addr += hpagesize; |
448 | } | |
449 | } | |
450 | pthread_sigmask(SIG_SETMASK, &oldset, NULL); | |
451 | return NULL; | |
452 | } | |
453 | ||
dfd0dcc7 JK |
454 | static inline int get_memset_num_threads(int smp_cpus) |
455 | { | |
456 | long host_procs = sysconf(_SC_NPROCESSORS_ONLN); | |
457 | int ret = 1; | |
458 | ||
459 | if (host_procs > 0) { | |
460 | ret = MIN(MIN(host_procs, MAX_MEM_PREALLOC_THREAD_COUNT), smp_cpus); | |
461 | } | |
462 | /* In case sysconf() fails, we fall back to single threaded */ | |
463 | return ret; | |
464 | } | |
465 | ||
1e356fc1 JK |
466 | static bool touch_all_pages(char *area, size_t hpagesize, size_t numpages, |
467 | int smp_cpus) | |
468 | { | |
78b3f67a | 469 | static gsize initialized = 0; |
037fb5eb | 470 | size_t numpages_per_thread, leftover; |
1e356fc1 JK |
471 | char *addr = area; |
472 | int i = 0; | |
473 | ||
78b3f67a PB |
474 | if (g_once_init_enter(&initialized)) { |
475 | qemu_mutex_init(&page_mutex); | |
476 | qemu_cond_init(&page_cond); | |
477 | g_once_init_leave(&initialized, 1); | |
478 | } | |
479 | ||
1e356fc1 | 480 | memset_thread_failed = false; |
037fb5eb | 481 | threads_created_flag = false; |
dfd0dcc7 | 482 | memset_num_threads = get_memset_num_threads(smp_cpus); |
1e356fc1 | 483 | memset_thread = g_new0(MemsetThread, memset_num_threads); |
037fb5eb | 484 | numpages_per_thread = numpages / memset_num_threads; |
485 | leftover = numpages % memset_num_threads; | |
1e356fc1 JK |
486 | for (i = 0; i < memset_num_threads; i++) { |
487 | memset_thread[i].addr = addr; | |
037fb5eb | 488 | memset_thread[i].numpages = numpages_per_thread + (i < leftover); |
1e356fc1 JK |
489 | memset_thread[i].hpagesize = hpagesize; |
490 | qemu_thread_create(&memset_thread[i].pgthread, "touch_pages", | |
491 | do_touch_pages, &memset_thread[i], | |
492 | QEMU_THREAD_JOINABLE); | |
037fb5eb | 493 | addr += memset_thread[i].numpages * hpagesize; |
1e356fc1 | 494 | } |
278fb162 B |
495 | |
496 | qemu_mutex_lock(&page_mutex); | |
037fb5eb | 497 | threads_created_flag = true; |
498 | qemu_cond_broadcast(&page_cond); | |
278fb162 | 499 | qemu_mutex_unlock(&page_mutex); |
037fb5eb | 500 | |
1e356fc1 JK |
501 | for (i = 0; i < memset_num_threads; i++) { |
502 | qemu_thread_join(&memset_thread[i].pgthread); | |
503 | } | |
504 | g_free(memset_thread); | |
505 | memset_thread = NULL; | |
506 | ||
507 | return memset_thread_failed; | |
508 | } | |
509 | ||
510 | void os_mem_prealloc(int fd, char *area, size_t memory, int smp_cpus, | |
511 | Error **errp) | |
38183310 | 512 | { |
b7bf8f56 | 513 | int ret; |
38183310 | 514 | struct sigaction act, oldact; |
1e356fc1 JK |
515 | size_t hpagesize = qemu_fd_getpagesize(fd); |
516 | size_t numpages = DIV_ROUND_UP(memory, hpagesize); | |
38183310 PB |
517 | |
518 | memset(&act, 0, sizeof(act)); | |
519 | act.sa_handler = &sigbus_handler; | |
520 | act.sa_flags = 0; | |
521 | ||
522 | ret = sigaction(SIGBUS, &act, &oldact); | |
523 | if (ret) { | |
056b68af IM |
524 | error_setg_errno(errp, errno, |
525 | "os_mem_prealloc: failed to install signal handler"); | |
526 | return; | |
38183310 PB |
527 | } |
528 | ||
1e356fc1 JK |
529 | /* touch pages simultaneously */ |
530 | if (touch_all_pages(area, hpagesize, numpages, smp_cpus)) { | |
056b68af | 531 | error_setg(errp, "os_mem_prealloc: Insufficient free host memory " |
462e5d50 | 532 | "pages available to allocate guest RAM"); |
056b68af | 533 | } |
38183310 | 534 | |
056b68af IM |
535 | ret = sigaction(SIGBUS, &oldact, NULL); |
536 | if (ret) { | |
537 | /* Terminate QEMU since it can't recover from error */ | |
538 | perror("os_mem_prealloc: failed to reinstall signal handler"); | |
539 | exit(1); | |
b7bf8f56 | 540 | } |
38183310 | 541 | } |
d57e4e48 | 542 | |
7dc9ae43 MP |
543 | char *qemu_get_pid_name(pid_t pid) |
544 | { | |
545 | char *name = NULL; | |
546 | ||
547 | #if defined(__FreeBSD__) | |
548 | /* BSDs don't have /proc, but they provide a nice substitute */ | |
549 | struct kinfo_proc *proc = kinfo_getproc(pid); | |
550 | ||
551 | if (proc) { | |
552 | name = g_strdup(proc->ki_comm); | |
553 | free(proc); | |
554 | } | |
555 | #else | |
556 | /* Assume a system with reasonable procfs */ | |
557 | char *pid_path; | |
558 | size_t len; | |
559 | ||
560 | pid_path = g_strdup_printf("/proc/%d/cmdline", pid); | |
561 | g_file_get_contents(pid_path, &name, &len, NULL); | |
562 | g_free(pid_path); | |
563 | #endif | |
564 | ||
565 | return name; | |
566 | } | |
567 | ||
568 | ||
57cb38b3 DB |
569 | pid_t qemu_fork(Error **errp) |
570 | { | |
571 | sigset_t oldmask, newmask; | |
572 | struct sigaction sig_action; | |
573 | int saved_errno; | |
574 | pid_t pid; | |
575 | ||
576 | /* | |
577 | * Need to block signals now, so that child process can safely | |
578 | * kill off caller's signal handlers without a race. | |
579 | */ | |
580 | sigfillset(&newmask); | |
581 | if (pthread_sigmask(SIG_SETMASK, &newmask, &oldmask) != 0) { | |
582 | error_setg_errno(errp, errno, | |
583 | "cannot block signals"); | |
584 | return -1; | |
585 | } | |
586 | ||
587 | pid = fork(); | |
588 | saved_errno = errno; | |
589 | ||
590 | if (pid < 0) { | |
591 | /* attempt to restore signal mask, but ignore failure, to | |
592 | * avoid obscuring the fork failure */ | |
593 | (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL); | |
594 | error_setg_errno(errp, saved_errno, | |
595 | "cannot fork child process"); | |
596 | errno = saved_errno; | |
597 | return -1; | |
598 | } else if (pid) { | |
599 | /* parent process */ | |
600 | ||
601 | /* Restore our original signal mask now that the child is | |
602 | * safely running. Only documented failures are EFAULT (not | |
603 | * possible, since we are using just-grabbed mask) or EINVAL | |
604 | * (not possible, since we are using correct arguments). */ | |
605 | (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL); | |
606 | } else { | |
607 | /* child process */ | |
608 | size_t i; | |
609 | ||
610 | /* Clear out all signal handlers from parent so nothing | |
611 | * unexpected can happen in our child once we unblock | |
612 | * signals */ | |
613 | sig_action.sa_handler = SIG_DFL; | |
614 | sig_action.sa_flags = 0; | |
615 | sigemptyset(&sig_action.sa_mask); | |
616 | ||
617 | for (i = 1; i < NSIG; i++) { | |
618 | /* Only possible errors are EFAULT or EINVAL The former | |
619 | * won't happen, the latter we expect, so no need to check | |
620 | * return value */ | |
621 | (void)sigaction(i, &sig_action, NULL); | |
622 | } | |
623 | ||
624 | /* Unmask all signals in child, since we've no idea what the | |
625 | * caller's done with their signal mask and don't want to | |
626 | * propagate that to children */ | |
627 | sigemptyset(&newmask); | |
628 | if (pthread_sigmask(SIG_SETMASK, &newmask, NULL) != 0) { | |
629 | Error *local_err = NULL; | |
630 | error_setg_errno(&local_err, errno, | |
631 | "cannot unblock signals"); | |
632 | error_report_err(local_err); | |
633 | _exit(1); | |
634 | } | |
635 | } | |
636 | return pid; | |
637 | } | |
8737d9e0 PL |
638 | |
639 | void *qemu_alloc_stack(size_t *sz) | |
640 | { | |
641 | void *ptr, *guardpage; | |
fc3d1bad | 642 | int flags; |
7d992e4d PL |
643 | #ifdef CONFIG_DEBUG_STACK_USAGE |
644 | void *ptr2; | |
645 | #endif | |
038adc2f | 646 | size_t pagesz = qemu_real_host_page_size; |
8737d9e0 PL |
647 | #ifdef _SC_THREAD_STACK_MIN |
648 | /* avoid stacks smaller than _SC_THREAD_STACK_MIN */ | |
649 | long min_stack_sz = sysconf(_SC_THREAD_STACK_MIN); | |
650 | *sz = MAX(MAX(min_stack_sz, 0), *sz); | |
651 | #endif | |
652 | /* adjust stack size to a multiple of the page size */ | |
653 | *sz = ROUND_UP(*sz, pagesz); | |
654 | /* allocate one extra page for the guard page */ | |
655 | *sz += pagesz; | |
656 | ||
fc3d1bad BS |
657 | flags = MAP_PRIVATE | MAP_ANONYMOUS; |
658 | #if defined(MAP_STACK) && defined(__OpenBSD__) | |
659 | /* Only enable MAP_STACK on OpenBSD. Other OS's such as | |
660 | * Linux/FreeBSD/NetBSD have a flag with the same name | |
661 | * but have differing functionality. OpenBSD will SEGV | |
662 | * if it spots execution with a stack pointer pointing | |
663 | * at memory that was not allocated with MAP_STACK. | |
664 | */ | |
665 | flags |= MAP_STACK; | |
666 | #endif | |
667 | ||
668 | ptr = mmap(NULL, *sz, PROT_READ | PROT_WRITE, flags, -1, 0); | |
8737d9e0 | 669 | if (ptr == MAP_FAILED) { |
e916a6e8 | 670 | perror("failed to allocate memory for stack"); |
8737d9e0 PL |
671 | abort(); |
672 | } | |
673 | ||
674 | #if defined(HOST_IA64) | |
675 | /* separate register stack */ | |
676 | guardpage = ptr + (((*sz - pagesz) / 2) & ~pagesz); | |
677 | #elif defined(HOST_HPPA) | |
678 | /* stack grows up */ | |
679 | guardpage = ptr + *sz - pagesz; | |
680 | #else | |
681 | /* stack grows down */ | |
682 | guardpage = ptr; | |
683 | #endif | |
684 | if (mprotect(guardpage, pagesz, PROT_NONE) != 0) { | |
e916a6e8 | 685 | perror("failed to set up stack guard page"); |
8737d9e0 PL |
686 | abort(); |
687 | } | |
688 | ||
7d992e4d PL |
689 | #ifdef CONFIG_DEBUG_STACK_USAGE |
690 | for (ptr2 = ptr + pagesz; ptr2 < ptr + *sz; ptr2 += sizeof(uint32_t)) { | |
691 | *(uint32_t *)ptr2 = 0xdeadbeaf; | |
692 | } | |
693 | #endif | |
694 | ||
8737d9e0 PL |
695 | return ptr; |
696 | } | |
697 | ||
7d992e4d PL |
698 | #ifdef CONFIG_DEBUG_STACK_USAGE |
699 | static __thread unsigned int max_stack_usage; | |
700 | #endif | |
701 | ||
8737d9e0 PL |
702 | void qemu_free_stack(void *stack, size_t sz) |
703 | { | |
7d992e4d PL |
704 | #ifdef CONFIG_DEBUG_STACK_USAGE |
705 | unsigned int usage; | |
706 | void *ptr; | |
707 | ||
038adc2f | 708 | for (ptr = stack + qemu_real_host_page_size; ptr < stack + sz; |
7d992e4d PL |
709 | ptr += sizeof(uint32_t)) { |
710 | if (*(uint32_t *)ptr != 0xdeadbeaf) { | |
711 | break; | |
712 | } | |
713 | } | |
714 | usage = sz - (uintptr_t) (ptr - stack); | |
715 | if (usage > max_stack_usage) { | |
716 | error_report("thread %d max stack usage increased from %u to %u", | |
717 | qemu_get_thread_id(), max_stack_usage, usage); | |
718 | max_stack_usage = usage; | |
719 | } | |
720 | #endif | |
721 | ||
8737d9e0 PL |
722 | munmap(stack, sz); |
723 | } | |
d98d4072 PB |
724 | |
725 | void sigaction_invoke(struct sigaction *action, | |
726 | struct qemu_signalfd_siginfo *info) | |
727 | { | |
02ffa034 | 728 | siginfo_t si = {}; |
d98d4072 PB |
729 | si.si_signo = info->ssi_signo; |
730 | si.si_errno = info->ssi_errno; | |
731 | si.si_code = info->ssi_code; | |
732 | ||
733 | /* Convert the minimal set of fields defined by POSIX. | |
734 | * Positive si_code values are reserved for kernel-generated | |
735 | * signals, where the valid siginfo fields are determined by | |
736 | * the signal number. But according to POSIX, it is unspecified | |
737 | * whether SI_USER and SI_QUEUE have values less than or equal to | |
738 | * zero. | |
739 | */ | |
740 | if (info->ssi_code == SI_USER || info->ssi_code == SI_QUEUE || | |
741 | info->ssi_code <= 0) { | |
742 | /* SIGTERM, etc. */ | |
743 | si.si_pid = info->ssi_pid; | |
744 | si.si_uid = info->ssi_uid; | |
745 | } else if (info->ssi_signo == SIGILL || info->ssi_signo == SIGFPE || | |
746 | info->ssi_signo == SIGSEGV || info->ssi_signo == SIGBUS) { | |
747 | si.si_addr = (void *)(uintptr_t)info->ssi_addr; | |
748 | } else if (info->ssi_signo == SIGCHLD) { | |
749 | si.si_pid = info->ssi_pid; | |
750 | si.si_status = info->ssi_status; | |
751 | si.si_uid = info->ssi_uid; | |
d98d4072 PB |
752 | } |
753 | action->sa_sigaction(info->ssi_signo, &si, NULL); | |
754 | } |