]>
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 | ||
9c17d615 | 34 | #include "sysemu/sysemu.h" |
c1b0b93b | 35 | #include "trace.h" |
da34e65c | 36 | #include "qapi/error.h" |
1de7afc9 | 37 | #include "qemu/sockets.h" |
10f5bff6 | 38 | #include <libgen.h> |
38183310 | 39 | #include <sys/signal.h> |
f348b6d1 | 40 | #include "qemu/cutils.h" |
c1b0b93b | 41 | |
cbcfa041 PB |
42 | #ifdef CONFIG_LINUX |
43 | #include <sys/syscall.h> | |
44 | #endif | |
cbcfa041 | 45 | |
41975b26 AF |
46 | #ifdef __FreeBSD__ |
47 | #include <sys/sysctl.h> | |
a7764f15 | 48 | #include <sys/user.h> |
7dc9ae43 | 49 | #include <libutil.h> |
41975b26 AF |
50 | #endif |
51 | ||
094611b4 KR |
52 | #ifdef __NetBSD__ |
53 | #include <sys/sysctl.h> | |
54 | #endif | |
55 | ||
a9c94277 | 56 | #include "qemu/mmap-alloc.h" |
794e8f30 | 57 | |
7d992e4d PL |
58 | #ifdef CONFIG_DEBUG_STACK_USAGE |
59 | #include "qemu/error-report.h" | |
60 | #endif | |
61 | ||
dfd0dcc7 | 62 | #define MAX_MEM_PREALLOC_THREAD_COUNT 16 |
1e356fc1 JK |
63 | |
64 | struct MemsetThread { | |
65 | char *addr; | |
e947d47d SW |
66 | size_t numpages; |
67 | size_t hpagesize; | |
1e356fc1 JK |
68 | QemuThread pgthread; |
69 | sigjmp_buf env; | |
70 | }; | |
71 | typedef struct MemsetThread MemsetThread; | |
72 | ||
73 | static MemsetThread *memset_thread; | |
74 | static int memset_num_threads; | |
75 | static bool memset_thread_failed; | |
76 | ||
cbcfa041 PB |
77 | int qemu_get_thread_id(void) |
78 | { | |
79 | #if defined(__linux__) | |
80 | return syscall(SYS_gettid); | |
81 | #else | |
82 | return getpid(); | |
83 | #endif | |
84 | } | |
f97742d0 AR |
85 | |
86 | int qemu_daemon(int nochdir, int noclose) | |
87 | { | |
88 | return daemon(nochdir, noclose); | |
89 | } | |
90 | ||
9e6bdef2 MAL |
91 | bool qemu_write_pidfile(const char *path, Error **errp) |
92 | { | |
93 | int fd; | |
94 | char pidstr[32]; | |
95 | ||
96 | while (1) { | |
97 | struct stat a, b; | |
35f7f3fb MAL |
98 | struct flock lock = { |
99 | .l_type = F_WRLCK, | |
100 | .l_whence = SEEK_SET, | |
101 | .l_len = 0, | |
102 | }; | |
9e6bdef2 MAL |
103 | |
104 | fd = qemu_open(path, O_CREAT | O_WRONLY, S_IRUSR | S_IWUSR); | |
105 | if (fd == -1) { | |
106 | error_setg_errno(errp, errno, "Cannot open pid file"); | |
107 | return false; | |
108 | } | |
109 | ||
110 | if (fstat(fd, &b) < 0) { | |
111 | error_setg_errno(errp, errno, "Cannot stat file"); | |
112 | goto fail_close; | |
113 | } | |
114 | ||
35f7f3fb | 115 | if (fcntl(fd, F_SETLK, &lock)) { |
9e6bdef2 MAL |
116 | error_setg_errno(errp, errno, "Cannot lock pid file"); |
117 | goto fail_close; | |
118 | } | |
119 | ||
120 | /* | |
121 | * Now make sure the path we locked is the same one that now | |
122 | * exists on the filesystem. | |
123 | */ | |
124 | if (stat(path, &a) < 0) { | |
125 | /* | |
126 | * PID file disappeared, someone else must be racing with | |
127 | * us, so try again. | |
128 | */ | |
129 | close(fd); | |
130 | continue; | |
131 | } | |
132 | ||
133 | if (a.st_ino == b.st_ino) { | |
134 | break; | |
135 | } | |
136 | ||
137 | /* | |
138 | * PID file was recreated, someone else must be racing with | |
139 | * us, so try again. | |
140 | */ | |
141 | close(fd); | |
142 | } | |
143 | ||
144 | if (ftruncate(fd, 0) < 0) { | |
145 | error_setg_errno(errp, errno, "Failed to truncate pid file"); | |
146 | goto fail_unlink; | |
147 | } | |
148 | ||
149 | snprintf(pidstr, sizeof(pidstr), FMT_pid "\n", getpid()); | |
150 | if (write(fd, pidstr, strlen(pidstr)) != strlen(pidstr)) { | |
151 | error_setg(errp, "Failed to write pid file"); | |
152 | goto fail_unlink; | |
153 | } | |
154 | ||
155 | return true; | |
156 | ||
157 | fail_unlink: | |
158 | unlink(path); | |
159 | fail_close: | |
160 | close(fd); | |
161 | return false; | |
162 | } | |
163 | ||
b152aa84 | 164 | void *qemu_oom_check(void *ptr) |
c1b0b93b JS |
165 | { |
166 | if (ptr == NULL) { | |
167 | fprintf(stderr, "Failed to allocate memory: %s\n", strerror(errno)); | |
168 | abort(); | |
169 | } | |
170 | return ptr; | |
171 | } | |
c1b0b93b | 172 | |
7d2a35cc | 173 | void *qemu_try_memalign(size_t alignment, size_t size) |
c1b0b93b JS |
174 | { |
175 | void *ptr; | |
e5354657 KW |
176 | |
177 | if (alignment < sizeof(void*)) { | |
178 | alignment = sizeof(void*); | |
179 | } | |
180 | ||
9bc5a719 | 181 | #if defined(CONFIG_POSIX_MEMALIGN) |
c1b0b93b JS |
182 | int ret; |
183 | ret = posix_memalign(&ptr, alignment, size); | |
184 | if (ret != 0) { | |
7d2a35cc KW |
185 | errno = ret; |
186 | ptr = NULL; | |
c1b0b93b JS |
187 | } |
188 | #elif defined(CONFIG_BSD) | |
7d2a35cc | 189 | ptr = valloc(size); |
c1b0b93b | 190 | #else |
7d2a35cc | 191 | ptr = memalign(alignment, size); |
c1b0b93b JS |
192 | #endif |
193 | trace_qemu_memalign(alignment, size, ptr); | |
194 | return ptr; | |
195 | } | |
196 | ||
7d2a35cc KW |
197 | void *qemu_memalign(size_t alignment, size_t size) |
198 | { | |
199 | return qemu_oom_check(qemu_try_memalign(alignment, size)); | |
200 | } | |
201 | ||
c1b0b93b | 202 | /* alloc shared memory pages */ |
06329cce | 203 | void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment, bool shared) |
c1b0b93b | 204 | { |
36b58628 | 205 | size_t align = QEMU_VMALLOC_ALIGN; |
06329cce | 206 | void *ptr = qemu_ram_mmap(-1, size, align, shared); |
36b58628 | 207 | |
7dda5dc8 | 208 | if (ptr == MAP_FAILED) { |
39228250 | 209 | return NULL; |
c2a8238a | 210 | } |
c2a8238a | 211 | |
a2b257d6 IM |
212 | if (alignment) { |
213 | *alignment = align; | |
214 | } | |
c2dfc5ba | 215 | |
6eebf958 | 216 | trace_qemu_anon_ram_alloc(size, ptr); |
c7f4111a | 217 | return ptr; |
c1b0b93b JS |
218 | } |
219 | ||
220 | void qemu_vfree(void *ptr) | |
221 | { | |
222 | trace_qemu_vfree(ptr); | |
223 | free(ptr); | |
224 | } | |
9549e764 | 225 | |
e7a09b92 PB |
226 | void qemu_anon_ram_free(void *ptr, size_t size) |
227 | { | |
228 | trace_qemu_anon_ram_free(ptr, size); | |
53adb9d4 | 229 | qemu_ram_munmap(-1, ptr, size); |
e7a09b92 PB |
230 | } |
231 | ||
f9e8cacc | 232 | void qemu_set_block(int fd) |
154b9a0c PB |
233 | { |
234 | int f; | |
235 | f = fcntl(fd, F_GETFL); | |
da93b820 LQ |
236 | assert(f != -1); |
237 | f = fcntl(fd, F_SETFL, f & ~O_NONBLOCK); | |
238 | assert(f != -1); | |
154b9a0c PB |
239 | } |
240 | ||
f9e8cacc | 241 | void qemu_set_nonblock(int fd) |
9549e764 JS |
242 | { |
243 | int f; | |
244 | f = fcntl(fd, F_GETFL); | |
da93b820 LQ |
245 | assert(f != -1); |
246 | f = fcntl(fd, F_SETFL, f | O_NONBLOCK); | |
02cdcc96 PMD |
247 | #ifdef __OpenBSD__ |
248 | if (f == -1) { | |
249 | /* | |
250 | * Previous to OpenBSD 6.3, fcntl(F_SETFL) is not permitted on | |
251 | * memory devices and sets errno to ENODEV. | |
252 | * It's OK if we fail to set O_NONBLOCK on devices like /dev/null, | |
253 | * because they will never block anyway. | |
254 | */ | |
255 | assert(errno == ENODEV); | |
256 | } | |
257 | #else | |
da93b820 | 258 | assert(f != -1); |
02cdcc96 | 259 | #endif |
9549e764 JS |
260 | } |
261 | ||
606600a1 SO |
262 | int socket_set_fast_reuse(int fd) |
263 | { | |
264 | int val = 1, ret; | |
265 | ||
266 | ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, | |
267 | (const char *)&val, sizeof(val)); | |
268 | ||
269 | assert(ret == 0); | |
270 | ||
271 | return ret; | |
272 | } | |
273 | ||
9549e764 JS |
274 | void qemu_set_cloexec(int fd) |
275 | { | |
276 | int f; | |
277 | f = fcntl(fd, F_GETFD); | |
7e6478e7 SS |
278 | assert(f != -1); |
279 | f = fcntl(fd, F_SETFD, f | FD_CLOEXEC); | |
280 | assert(f != -1); | |
9549e764 | 281 | } |
70e72ce4 JS |
282 | |
283 | /* | |
284 | * Creates a pipe with FD_CLOEXEC set on both file descriptors | |
285 | */ | |
286 | int qemu_pipe(int pipefd[2]) | |
287 | { | |
288 | int ret; | |
289 | ||
290 | #ifdef CONFIG_PIPE2 | |
291 | ret = pipe2(pipefd, O_CLOEXEC); | |
292 | if (ret != -1 || errno != ENOSYS) { | |
293 | return ret; | |
294 | } | |
295 | #endif | |
296 | ret = pipe(pipefd); | |
297 | if (ret == 0) { | |
298 | qemu_set_cloexec(pipefd[0]); | |
299 | qemu_set_cloexec(pipefd[1]); | |
300 | } | |
301 | ||
302 | return ret; | |
303 | } | |
38671423 | 304 | |
e2ea3515 LE |
305 | char * |
306 | qemu_get_local_state_pathname(const char *relative_pathname) | |
307 | { | |
308 | return g_strdup_printf("%s/%s", CONFIG_QEMU_LOCALSTATEDIR, | |
309 | relative_pathname); | |
310 | } | |
13401ba0 SH |
311 | |
312 | void qemu_set_tty_echo(int fd, bool echo) | |
313 | { | |
314 | struct termios tty; | |
315 | ||
316 | tcgetattr(fd, &tty); | |
317 | ||
318 | if (echo) { | |
319 | tty.c_lflag |= ECHO | ECHONL | ICANON | IEXTEN; | |
320 | } else { | |
321 | tty.c_lflag &= ~(ECHO | ECHONL | ICANON | IEXTEN); | |
322 | } | |
323 | ||
324 | tcsetattr(fd, TCSANOW, &tty); | |
325 | } | |
10f5bff6 FZ |
326 | |
327 | static char exec_dir[PATH_MAX]; | |
328 | ||
329 | void qemu_init_exec_dir(const char *argv0) | |
330 | { | |
331 | char *dir; | |
332 | char *p = NULL; | |
333 | char buf[PATH_MAX]; | |
334 | ||
335 | assert(!exec_dir[0]); | |
336 | ||
337 | #if defined(__linux__) | |
338 | { | |
339 | int len; | |
340 | len = readlink("/proc/self/exe", buf, sizeof(buf) - 1); | |
341 | if (len > 0) { | |
342 | buf[len] = 0; | |
343 | p = buf; | |
344 | } | |
345 | } | |
094611b4 KR |
346 | #elif defined(__FreeBSD__) \ |
347 | || (defined(__NetBSD__) && defined(KERN_PROC_PATHNAME)) | |
10f5bff6 | 348 | { |
094611b4 | 349 | #if defined(__FreeBSD__) |
10f5bff6 | 350 | static int mib[4] = {CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1}; |
094611b4 KR |
351 | #else |
352 | static int mib[4] = {CTL_KERN, KERN_PROC_ARGS, -1, KERN_PROC_PATHNAME}; | |
353 | #endif | |
10f5bff6 FZ |
354 | size_t len = sizeof(buf) - 1; |
355 | ||
356 | *buf = '\0'; | |
357 | if (!sysctl(mib, ARRAY_SIZE(mib), buf, &len, NULL, 0) && | |
358 | *buf) { | |
359 | buf[sizeof(buf) - 1] = '\0'; | |
360 | p = buf; | |
361 | } | |
362 | } | |
363 | #endif | |
364 | /* If we don't have any way of figuring out the actual executable | |
365 | location then try argv[0]. */ | |
366 | if (!p) { | |
367 | if (!argv0) { | |
368 | return; | |
369 | } | |
370 | p = realpath(argv0, buf); | |
371 | if (!p) { | |
372 | return; | |
373 | } | |
374 | } | |
55ad781c | 375 | dir = g_path_get_dirname(p); |
10f5bff6 FZ |
376 | |
377 | pstrcpy(exec_dir, sizeof(exec_dir), dir); | |
55ad781c WJ |
378 | |
379 | g_free(dir); | |
10f5bff6 FZ |
380 | } |
381 | ||
382 | char *qemu_get_exec_dir(void) | |
383 | { | |
384 | return g_strdup(exec_dir); | |
385 | } | |
38183310 | 386 | |
38183310 PB |
387 | static void sigbus_handler(int signal) |
388 | { | |
1e356fc1 JK |
389 | int i; |
390 | if (memset_thread) { | |
391 | for (i = 0; i < memset_num_threads; i++) { | |
392 | if (qemu_thread_is_self(&memset_thread[i].pgthread)) { | |
393 | siglongjmp(memset_thread[i].env, 1); | |
394 | } | |
395 | } | |
396 | } | |
38183310 PB |
397 | } |
398 | ||
1e356fc1 JK |
399 | static void *do_touch_pages(void *arg) |
400 | { | |
401 | MemsetThread *memset_args = (MemsetThread *)arg; | |
1e356fc1 | 402 | sigset_t set, oldset; |
1e356fc1 JK |
403 | |
404 | /* unblock SIGBUS */ | |
405 | sigemptyset(&set); | |
406 | sigaddset(&set, SIGBUS); | |
407 | pthread_sigmask(SIG_UNBLOCK, &set, &oldset); | |
408 | ||
409 | if (sigsetjmp(memset_args->env, 1)) { | |
410 | memset_thread_failed = true; | |
411 | } else { | |
e947d47d SW |
412 | char *addr = memset_args->addr; |
413 | size_t numpages = memset_args->numpages; | |
414 | size_t hpagesize = memset_args->hpagesize; | |
415 | size_t i; | |
1e356fc1 | 416 | for (i = 0; i < numpages; i++) { |
9dc44aa5 DB |
417 | /* |
418 | * Read & write back the same value, so we don't | |
419 | * corrupt existing user/app data that might be | |
420 | * stored. | |
421 | * | |
422 | * 'volatile' to stop compiler optimizing this away | |
423 | * to a no-op | |
424 | * | |
425 | * TODO: get a better solution from kernel so we | |
426 | * don't need to write at all so we don't cause | |
427 | * wear on the storage backing the region... | |
428 | */ | |
429 | *(volatile char *)addr = *addr; | |
1e356fc1 JK |
430 | addr += hpagesize; |
431 | } | |
432 | } | |
433 | pthread_sigmask(SIG_SETMASK, &oldset, NULL); | |
434 | return NULL; | |
435 | } | |
436 | ||
dfd0dcc7 JK |
437 | static inline int get_memset_num_threads(int smp_cpus) |
438 | { | |
439 | long host_procs = sysconf(_SC_NPROCESSORS_ONLN); | |
440 | int ret = 1; | |
441 | ||
442 | if (host_procs > 0) { | |
443 | ret = MIN(MIN(host_procs, MAX_MEM_PREALLOC_THREAD_COUNT), smp_cpus); | |
444 | } | |
445 | /* In case sysconf() fails, we fall back to single threaded */ | |
446 | return ret; | |
447 | } | |
448 | ||
1e356fc1 JK |
449 | static bool touch_all_pages(char *area, size_t hpagesize, size_t numpages, |
450 | int smp_cpus) | |
451 | { | |
e947d47d SW |
452 | size_t numpages_per_thread; |
453 | size_t size_per_thread; | |
1e356fc1 JK |
454 | char *addr = area; |
455 | int i = 0; | |
456 | ||
457 | memset_thread_failed = false; | |
dfd0dcc7 | 458 | memset_num_threads = get_memset_num_threads(smp_cpus); |
1e356fc1 JK |
459 | memset_thread = g_new0(MemsetThread, memset_num_threads); |
460 | numpages_per_thread = (numpages / memset_num_threads); | |
461 | size_per_thread = (hpagesize * numpages_per_thread); | |
462 | for (i = 0; i < memset_num_threads; i++) { | |
463 | memset_thread[i].addr = addr; | |
464 | memset_thread[i].numpages = (i == (memset_num_threads - 1)) ? | |
465 | numpages : numpages_per_thread; | |
466 | memset_thread[i].hpagesize = hpagesize; | |
467 | qemu_thread_create(&memset_thread[i].pgthread, "touch_pages", | |
468 | do_touch_pages, &memset_thread[i], | |
469 | QEMU_THREAD_JOINABLE); | |
470 | addr += size_per_thread; | |
471 | numpages -= numpages_per_thread; | |
472 | } | |
473 | for (i = 0; i < memset_num_threads; i++) { | |
474 | qemu_thread_join(&memset_thread[i].pgthread); | |
475 | } | |
476 | g_free(memset_thread); | |
477 | memset_thread = NULL; | |
478 | ||
479 | return memset_thread_failed; | |
480 | } | |
481 | ||
482 | void os_mem_prealloc(int fd, char *area, size_t memory, int smp_cpus, | |
483 | Error **errp) | |
38183310 | 484 | { |
b7bf8f56 | 485 | int ret; |
38183310 | 486 | struct sigaction act, oldact; |
1e356fc1 JK |
487 | size_t hpagesize = qemu_fd_getpagesize(fd); |
488 | size_t numpages = DIV_ROUND_UP(memory, hpagesize); | |
38183310 PB |
489 | |
490 | memset(&act, 0, sizeof(act)); | |
491 | act.sa_handler = &sigbus_handler; | |
492 | act.sa_flags = 0; | |
493 | ||
494 | ret = sigaction(SIGBUS, &act, &oldact); | |
495 | if (ret) { | |
056b68af IM |
496 | error_setg_errno(errp, errno, |
497 | "os_mem_prealloc: failed to install signal handler"); | |
498 | return; | |
38183310 PB |
499 | } |
500 | ||
1e356fc1 JK |
501 | /* touch pages simultaneously */ |
502 | if (touch_all_pages(area, hpagesize, numpages, smp_cpus)) { | |
056b68af | 503 | error_setg(errp, "os_mem_prealloc: Insufficient free host memory " |
462e5d50 | 504 | "pages available to allocate guest RAM"); |
056b68af | 505 | } |
38183310 | 506 | |
056b68af IM |
507 | ret = sigaction(SIGBUS, &oldact, NULL); |
508 | if (ret) { | |
509 | /* Terminate QEMU since it can't recover from error */ | |
510 | perror("os_mem_prealloc: failed to reinstall signal handler"); | |
511 | exit(1); | |
b7bf8f56 | 512 | } |
38183310 | 513 | } |
d57e4e48 | 514 | |
314aec4a SH |
515 | uint64_t qemu_get_pmem_size(const char *filename, Error **errp) |
516 | { | |
517 | struct stat st; | |
518 | ||
519 | if (stat(filename, &st) < 0) { | |
520 | error_setg(errp, "unable to stat pmem file \"%s\"", filename); | |
521 | return 0; | |
522 | } | |
523 | ||
524 | #if defined(__linux__) | |
525 | /* Special handling for devdax character devices */ | |
526 | if (S_ISCHR(st.st_mode)) { | |
527 | char *subsystem_path = NULL; | |
528 | char *subsystem = NULL; | |
529 | char *size_path = NULL; | |
530 | char *size_str = NULL; | |
531 | uint64_t ret = 0; | |
532 | ||
533 | subsystem_path = g_strdup_printf("/sys/dev/char/%d:%d/subsystem", | |
534 | major(st.st_rdev), minor(st.st_rdev)); | |
535 | subsystem = g_file_read_link(subsystem_path, NULL); | |
536 | if (!subsystem) { | |
537 | error_setg(errp, "unable to read subsystem for pmem file \"%s\"", | |
538 | filename); | |
539 | goto devdax_err; | |
540 | } | |
541 | ||
542 | if (!g_str_has_suffix(subsystem, "/dax")) { | |
543 | error_setg(errp, "pmem file \"%s\" is not a dax device", filename); | |
544 | goto devdax_err; | |
545 | } | |
546 | ||
547 | size_path = g_strdup_printf("/sys/dev/char/%d:%d/size", | |
548 | major(st.st_rdev), minor(st.st_rdev)); | |
549 | if (!g_file_get_contents(size_path, &size_str, NULL, NULL)) { | |
550 | error_setg(errp, "unable to read size for pmem file \"%s\"", | |
551 | size_path); | |
552 | goto devdax_err; | |
553 | } | |
554 | ||
555 | ret = g_ascii_strtoull(size_str, NULL, 0); | |
556 | ||
557 | devdax_err: | |
558 | g_free(size_str); | |
559 | g_free(size_path); | |
560 | g_free(subsystem); | |
561 | g_free(subsystem_path); | |
562 | return ret; | |
563 | } | |
564 | #endif /* defined(__linux__) */ | |
565 | ||
566 | return st.st_size; | |
567 | } | |
d57e4e48 | 568 | |
7dc9ae43 MP |
569 | char *qemu_get_pid_name(pid_t pid) |
570 | { | |
571 | char *name = NULL; | |
572 | ||
573 | #if defined(__FreeBSD__) | |
574 | /* BSDs don't have /proc, but they provide a nice substitute */ | |
575 | struct kinfo_proc *proc = kinfo_getproc(pid); | |
576 | ||
577 | if (proc) { | |
578 | name = g_strdup(proc->ki_comm); | |
579 | free(proc); | |
580 | } | |
581 | #else | |
582 | /* Assume a system with reasonable procfs */ | |
583 | char *pid_path; | |
584 | size_t len; | |
585 | ||
586 | pid_path = g_strdup_printf("/proc/%d/cmdline", pid); | |
587 | g_file_get_contents(pid_path, &name, &len, NULL); | |
588 | g_free(pid_path); | |
589 | #endif | |
590 | ||
591 | return name; | |
592 | } | |
593 | ||
594 | ||
57cb38b3 DB |
595 | pid_t qemu_fork(Error **errp) |
596 | { | |
597 | sigset_t oldmask, newmask; | |
598 | struct sigaction sig_action; | |
599 | int saved_errno; | |
600 | pid_t pid; | |
601 | ||
602 | /* | |
603 | * Need to block signals now, so that child process can safely | |
604 | * kill off caller's signal handlers without a race. | |
605 | */ | |
606 | sigfillset(&newmask); | |
607 | if (pthread_sigmask(SIG_SETMASK, &newmask, &oldmask) != 0) { | |
608 | error_setg_errno(errp, errno, | |
609 | "cannot block signals"); | |
610 | return -1; | |
611 | } | |
612 | ||
613 | pid = fork(); | |
614 | saved_errno = errno; | |
615 | ||
616 | if (pid < 0) { | |
617 | /* attempt to restore signal mask, but ignore failure, to | |
618 | * avoid obscuring the fork failure */ | |
619 | (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL); | |
620 | error_setg_errno(errp, saved_errno, | |
621 | "cannot fork child process"); | |
622 | errno = saved_errno; | |
623 | return -1; | |
624 | } else if (pid) { | |
625 | /* parent process */ | |
626 | ||
627 | /* Restore our original signal mask now that the child is | |
628 | * safely running. Only documented failures are EFAULT (not | |
629 | * possible, since we are using just-grabbed mask) or EINVAL | |
630 | * (not possible, since we are using correct arguments). */ | |
631 | (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL); | |
632 | } else { | |
633 | /* child process */ | |
634 | size_t i; | |
635 | ||
636 | /* Clear out all signal handlers from parent so nothing | |
637 | * unexpected can happen in our child once we unblock | |
638 | * signals */ | |
639 | sig_action.sa_handler = SIG_DFL; | |
640 | sig_action.sa_flags = 0; | |
641 | sigemptyset(&sig_action.sa_mask); | |
642 | ||
643 | for (i = 1; i < NSIG; i++) { | |
644 | /* Only possible errors are EFAULT or EINVAL The former | |
645 | * won't happen, the latter we expect, so no need to check | |
646 | * return value */ | |
647 | (void)sigaction(i, &sig_action, NULL); | |
648 | } | |
649 | ||
650 | /* Unmask all signals in child, since we've no idea what the | |
651 | * caller's done with their signal mask and don't want to | |
652 | * propagate that to children */ | |
653 | sigemptyset(&newmask); | |
654 | if (pthread_sigmask(SIG_SETMASK, &newmask, NULL) != 0) { | |
655 | Error *local_err = NULL; | |
656 | error_setg_errno(&local_err, errno, | |
657 | "cannot unblock signals"); | |
658 | error_report_err(local_err); | |
659 | _exit(1); | |
660 | } | |
661 | } | |
662 | return pid; | |
663 | } | |
8737d9e0 PL |
664 | |
665 | void *qemu_alloc_stack(size_t *sz) | |
666 | { | |
667 | void *ptr, *guardpage; | |
fc3d1bad | 668 | int flags; |
7d992e4d PL |
669 | #ifdef CONFIG_DEBUG_STACK_USAGE |
670 | void *ptr2; | |
671 | #endif | |
8737d9e0 PL |
672 | size_t pagesz = getpagesize(); |
673 | #ifdef _SC_THREAD_STACK_MIN | |
674 | /* avoid stacks smaller than _SC_THREAD_STACK_MIN */ | |
675 | long min_stack_sz = sysconf(_SC_THREAD_STACK_MIN); | |
676 | *sz = MAX(MAX(min_stack_sz, 0), *sz); | |
677 | #endif | |
678 | /* adjust stack size to a multiple of the page size */ | |
679 | *sz = ROUND_UP(*sz, pagesz); | |
680 | /* allocate one extra page for the guard page */ | |
681 | *sz += pagesz; | |
682 | ||
fc3d1bad BS |
683 | flags = MAP_PRIVATE | MAP_ANONYMOUS; |
684 | #if defined(MAP_STACK) && defined(__OpenBSD__) | |
685 | /* Only enable MAP_STACK on OpenBSD. Other OS's such as | |
686 | * Linux/FreeBSD/NetBSD have a flag with the same name | |
687 | * but have differing functionality. OpenBSD will SEGV | |
688 | * if it spots execution with a stack pointer pointing | |
689 | * at memory that was not allocated with MAP_STACK. | |
690 | */ | |
691 | flags |= MAP_STACK; | |
692 | #endif | |
693 | ||
694 | ptr = mmap(NULL, *sz, PROT_READ | PROT_WRITE, flags, -1, 0); | |
8737d9e0 | 695 | if (ptr == MAP_FAILED) { |
e916a6e8 | 696 | perror("failed to allocate memory for stack"); |
8737d9e0 PL |
697 | abort(); |
698 | } | |
699 | ||
700 | #if defined(HOST_IA64) | |
701 | /* separate register stack */ | |
702 | guardpage = ptr + (((*sz - pagesz) / 2) & ~pagesz); | |
703 | #elif defined(HOST_HPPA) | |
704 | /* stack grows up */ | |
705 | guardpage = ptr + *sz - pagesz; | |
706 | #else | |
707 | /* stack grows down */ | |
708 | guardpage = ptr; | |
709 | #endif | |
710 | if (mprotect(guardpage, pagesz, PROT_NONE) != 0) { | |
e916a6e8 | 711 | perror("failed to set up stack guard page"); |
8737d9e0 PL |
712 | abort(); |
713 | } | |
714 | ||
7d992e4d PL |
715 | #ifdef CONFIG_DEBUG_STACK_USAGE |
716 | for (ptr2 = ptr + pagesz; ptr2 < ptr + *sz; ptr2 += sizeof(uint32_t)) { | |
717 | *(uint32_t *)ptr2 = 0xdeadbeaf; | |
718 | } | |
719 | #endif | |
720 | ||
8737d9e0 PL |
721 | return ptr; |
722 | } | |
723 | ||
7d992e4d PL |
724 | #ifdef CONFIG_DEBUG_STACK_USAGE |
725 | static __thread unsigned int max_stack_usage; | |
726 | #endif | |
727 | ||
8737d9e0 PL |
728 | void qemu_free_stack(void *stack, size_t sz) |
729 | { | |
7d992e4d PL |
730 | #ifdef CONFIG_DEBUG_STACK_USAGE |
731 | unsigned int usage; | |
732 | void *ptr; | |
733 | ||
734 | for (ptr = stack + getpagesize(); ptr < stack + sz; | |
735 | ptr += sizeof(uint32_t)) { | |
736 | if (*(uint32_t *)ptr != 0xdeadbeaf) { | |
737 | break; | |
738 | } | |
739 | } | |
740 | usage = sz - (uintptr_t) (ptr - stack); | |
741 | if (usage > max_stack_usage) { | |
742 | error_report("thread %d max stack usage increased from %u to %u", | |
743 | qemu_get_thread_id(), max_stack_usage, usage); | |
744 | max_stack_usage = usage; | |
745 | } | |
746 | #endif | |
747 | ||
8737d9e0 PL |
748 | munmap(stack, sz); |
749 | } | |
d98d4072 PB |
750 | |
751 | void sigaction_invoke(struct sigaction *action, | |
752 | struct qemu_signalfd_siginfo *info) | |
753 | { | |
02ffa034 | 754 | siginfo_t si = {}; |
d98d4072 PB |
755 | si.si_signo = info->ssi_signo; |
756 | si.si_errno = info->ssi_errno; | |
757 | si.si_code = info->ssi_code; | |
758 | ||
759 | /* Convert the minimal set of fields defined by POSIX. | |
760 | * Positive si_code values are reserved for kernel-generated | |
761 | * signals, where the valid siginfo fields are determined by | |
762 | * the signal number. But according to POSIX, it is unspecified | |
763 | * whether SI_USER and SI_QUEUE have values less than or equal to | |
764 | * zero. | |
765 | */ | |
766 | if (info->ssi_code == SI_USER || info->ssi_code == SI_QUEUE || | |
767 | info->ssi_code <= 0) { | |
768 | /* SIGTERM, etc. */ | |
769 | si.si_pid = info->ssi_pid; | |
770 | si.si_uid = info->ssi_uid; | |
771 | } else if (info->ssi_signo == SIGILL || info->ssi_signo == SIGFPE || | |
772 | info->ssi_signo == SIGSEGV || info->ssi_signo == SIGBUS) { | |
773 | si.si_addr = (void *)(uintptr_t)info->ssi_addr; | |
774 | } else if (info->ssi_signo == SIGCHLD) { | |
775 | si.si_pid = info->ssi_pid; | |
776 | si.si_status = info->ssi_status; | |
777 | si.si_uid = info->ssi_uid; | |
d98d4072 PB |
778 | } |
779 | action->sa_sigaction(info->ssi_signo, &si, NULL); | |
780 | } |