4 * Copyright (c) 2003-2008 Fabrice Bellard
5 * Copyright (c) 2010 Red Hat, Inc.
7 * QEMU library functions on POSIX which are shared between QEMU and
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:
17 * The above copyright notice and this permission notice shall be included in
18 * all copies or substantial portions of the Software.
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
29 #if defined(__linux__) && (defined(__x86_64__) || defined(__arm__))
30 /* Use 2 MiB alignment so transparent hugepages can be used by KVM.
31 Valgrind does not support alignments larger than 1 MiB,
32 therefore we need special code which handles running on Valgrind. */
33 # define QEMU_VMALLOC_ALIGN (512 * 4096)
34 #elif defined(__linux__) && defined(__s390x__)
35 /* Use 1 MiB (segment size) alignment so gmap can be used by KVM. */
36 # define QEMU_VMALLOC_ALIGN (256 * 4096)
38 # define QEMU_VMALLOC_ALIGN getpagesize()
41 #include "qemu/osdep.h"
45 #include <glib/gprintf.h>
47 #include "sysemu/sysemu.h"
49 #include "qemu/sockets.h"
52 #include <sys/signal.h>
55 #include <sys/syscall.h>
59 #include <sys/sysctl.h>
62 #include <qemu/mmap-alloc.h>
64 int qemu_get_thread_id(void)
66 #if defined(__linux__)
67 return syscall(SYS_gettid);
73 int qemu_daemon(int nochdir, int noclose)
75 return daemon(nochdir, noclose);
78 void *qemu_oom_check(void *ptr)
81 fprintf(stderr, "Failed to allocate memory: %s\n", strerror(errno));
87 void *qemu_try_memalign(size_t alignment, size_t size)
91 if (alignment < sizeof(void*)) {
92 alignment = sizeof(void*);
95 #if defined(_POSIX_C_SOURCE) && !defined(__sun__)
97 ret = posix_memalign(&ptr, alignment, size);
102 #elif defined(CONFIG_BSD)
105 ptr = memalign(alignment, size);
107 trace_qemu_memalign(alignment, size, ptr);
111 void *qemu_memalign(size_t alignment, size_t size)
113 return qemu_oom_check(qemu_try_memalign(alignment, size));
116 /* alloc shared memory pages */
117 void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment)
119 size_t align = QEMU_VMALLOC_ALIGN;
120 void *ptr = qemu_ram_mmap(-1, size, align, false);
122 if (ptr == MAP_FAILED) {
130 trace_qemu_anon_ram_alloc(size, ptr);
134 void qemu_vfree(void *ptr)
136 trace_qemu_vfree(ptr);
140 void qemu_anon_ram_free(void *ptr, size_t size)
142 trace_qemu_anon_ram_free(ptr, size);
143 qemu_ram_munmap(ptr, size);
146 void qemu_set_block(int fd)
149 f = fcntl(fd, F_GETFL);
150 fcntl(fd, F_SETFL, f & ~O_NONBLOCK);
153 void qemu_set_nonblock(int fd)
156 f = fcntl(fd, F_GETFL);
157 fcntl(fd, F_SETFL, f | O_NONBLOCK);
160 int socket_set_fast_reuse(int fd)
164 ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
165 (const char *)&val, sizeof(val));
172 void qemu_set_cloexec(int fd)
175 f = fcntl(fd, F_GETFD);
176 fcntl(fd, F_SETFD, f | FD_CLOEXEC);
180 * Creates a pipe with FD_CLOEXEC set on both file descriptors
182 int qemu_pipe(int pipefd[2])
187 ret = pipe2(pipefd, O_CLOEXEC);
188 if (ret != -1 || errno != ENOSYS) {
194 qemu_set_cloexec(pipefd[0]);
195 qemu_set_cloexec(pipefd[1]);
201 int qemu_utimens(const char *path, const struct timespec *times)
203 struct timeval tv[2], tv_now;
206 #ifdef CONFIG_UTIMENSAT
209 ret = utimensat(AT_FDCWD, path, times, AT_SYMLINK_NOFOLLOW);
210 if (ret != -1 || errno != ENOSYS) {
214 /* Fallback: use utimes() instead of utimensat() */
216 /* happy if special cases */
217 if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT) {
220 if (times[0].tv_nsec == UTIME_NOW && times[1].tv_nsec == UTIME_NOW) {
221 return utimes(path, NULL);
224 /* prepare for hard cases */
225 if (times[0].tv_nsec == UTIME_NOW || times[1].tv_nsec == UTIME_NOW) {
226 gettimeofday(&tv_now, NULL);
228 if (times[0].tv_nsec == UTIME_OMIT || times[1].tv_nsec == UTIME_OMIT) {
232 for (i = 0; i < 2; i++) {
233 if (times[i].tv_nsec == UTIME_NOW) {
234 tv[i].tv_sec = tv_now.tv_sec;
235 tv[i].tv_usec = tv_now.tv_usec;
236 } else if (times[i].tv_nsec == UTIME_OMIT) {
237 tv[i].tv_sec = (i == 0) ? st.st_atime : st.st_mtime;
240 tv[i].tv_sec = times[i].tv_sec;
241 tv[i].tv_usec = times[i].tv_nsec / 1000;
245 return utimes(path, &tv[0]);
249 qemu_get_local_state_pathname(const char *relative_pathname)
251 return g_strdup_printf("%s/%s", CONFIG_QEMU_LOCALSTATEDIR,
255 void qemu_set_tty_echo(int fd, bool echo)
262 tty.c_lflag |= ECHO | ECHONL | ICANON | IEXTEN;
264 tty.c_lflag &= ~(ECHO | ECHONL | ICANON | IEXTEN);
267 tcsetattr(fd, TCSANOW, &tty);
270 static char exec_dir[PATH_MAX];
272 void qemu_init_exec_dir(const char *argv0)
278 assert(!exec_dir[0]);
280 #if defined(__linux__)
283 len = readlink("/proc/self/exe", buf, sizeof(buf) - 1);
289 #elif defined(__FreeBSD__)
291 static int mib[4] = {CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1};
292 size_t len = sizeof(buf) - 1;
295 if (!sysctl(mib, ARRAY_SIZE(mib), buf, &len, NULL, 0) &&
297 buf[sizeof(buf) - 1] = '\0';
302 /* If we don't have any way of figuring out the actual executable
303 location then try argv[0]. */
308 p = realpath(argv0, buf);
315 pstrcpy(exec_dir, sizeof(exec_dir), dir);
318 char *qemu_get_exec_dir(void)
320 return g_strdup(exec_dir);
323 static sigjmp_buf sigjump;
325 static void sigbus_handler(int signal)
327 siglongjmp(sigjump, 1);
330 void os_mem_prealloc(int fd, char *area, size_t memory)
333 struct sigaction act, oldact;
334 sigset_t set, oldset;
336 memset(&act, 0, sizeof(act));
337 act.sa_handler = &sigbus_handler;
340 ret = sigaction(SIGBUS, &act, &oldact);
342 perror("os_mem_prealloc: failed to install signal handler");
348 sigaddset(&set, SIGBUS);
349 pthread_sigmask(SIG_UNBLOCK, &set, &oldset);
351 if (sigsetjmp(sigjump, 1)) {
352 fprintf(stderr, "os_mem_prealloc: Insufficient free host memory "
353 "pages available to allocate guest RAM\n");
357 size_t hpagesize = qemu_fd_getpagesize(fd);
358 size_t numpages = DIV_ROUND_UP(memory, hpagesize);
360 /* MAP_POPULATE silently ignores failures */
361 for (i = 0; i < numpages; i++) {
362 memset(area + (hpagesize * i), 0, 1);
365 ret = sigaction(SIGBUS, &oldact, NULL);
367 perror("os_mem_prealloc: failed to reinstall signal handler");
371 pthread_sigmask(SIG_SETMASK, &oldset, NULL);
376 static struct termios oldtty;
378 static void term_exit(void)
380 tcsetattr(0, TCSANOW, &oldtty);
383 static void term_init(void)
390 tty.c_iflag &= ~(IGNBRK|BRKINT|PARMRK|ISTRIP
391 |INLCR|IGNCR|ICRNL|IXON);
392 tty.c_oflag |= OPOST;
393 tty.c_lflag &= ~(ECHO|ECHONL|ICANON|IEXTEN);
394 tty.c_cflag &= ~(CSIZE|PARENB);
399 tcsetattr(0, TCSANOW, &tty);
404 int qemu_read_password(char *buf, int buf_size)
409 printf("password: ");
414 ret = read(0, &ch, 1);
416 if (errno == EAGAIN || errno == EINTR) {
421 } else if (ret == 0) {
430 if (i < (buf_size - 1)) {
442 pid_t qemu_fork(Error **errp)
444 sigset_t oldmask, newmask;
445 struct sigaction sig_action;
450 * Need to block signals now, so that child process can safely
451 * kill off caller's signal handlers without a race.
453 sigfillset(&newmask);
454 if (pthread_sigmask(SIG_SETMASK, &newmask, &oldmask) != 0) {
455 error_setg_errno(errp, errno,
456 "cannot block signals");
464 /* attempt to restore signal mask, but ignore failure, to
465 * avoid obscuring the fork failure */
466 (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
467 error_setg_errno(errp, saved_errno,
468 "cannot fork child process");
474 /* Restore our original signal mask now that the child is
475 * safely running. Only documented failures are EFAULT (not
476 * possible, since we are using just-grabbed mask) or EINVAL
477 * (not possible, since we are using correct arguments). */
478 (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
483 /* Clear out all signal handlers from parent so nothing
484 * unexpected can happen in our child once we unblock
486 sig_action.sa_handler = SIG_DFL;
487 sig_action.sa_flags = 0;
488 sigemptyset(&sig_action.sa_mask);
490 for (i = 1; i < NSIG; i++) {
491 /* Only possible errors are EFAULT or EINVAL The former
492 * won't happen, the latter we expect, so no need to check
494 (void)sigaction(i, &sig_action, NULL);
497 /* Unmask all signals in child, since we've no idea what the
498 * caller's done with their signal mask and don't want to
499 * propagate that to children */
500 sigemptyset(&newmask);
501 if (pthread_sigmask(SIG_SETMASK, &newmask, NULL) != 0) {
502 Error *local_err = NULL;
503 error_setg_errno(&local_err, errno,
504 "cannot unblock signals");
505 error_report_err(local_err);