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 /* 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 */
32 #define daemon qemu_fake_daemon_function
35 extern int daemon(int, int);
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)
47 # define QEMU_VMALLOC_ALIGN getpagesize()
50 #include "qemu/osdep.h"
54 #include <glib/gprintf.h>
56 #include "sysemu/sysemu.h"
58 #include "qemu/sockets.h"
62 #include <sys/signal.h>
65 #include <sys/syscall.h>
69 #include <sys/sysctl.h>
72 #include <qemu/mmap-alloc.h>
74 int qemu_get_thread_id(void)
76 #if defined(__linux__)
77 return syscall(SYS_gettid);
83 int qemu_daemon(int nochdir, int noclose)
85 return daemon(nochdir, noclose);
88 void *qemu_oom_check(void *ptr)
91 fprintf(stderr, "Failed to allocate memory: %s\n", strerror(errno));
97 void *qemu_try_memalign(size_t alignment, size_t size)
101 if (alignment < sizeof(void*)) {
102 alignment = sizeof(void*);
105 #if defined(_POSIX_C_SOURCE) && !defined(__sun__)
107 ret = posix_memalign(&ptr, alignment, size);
112 #elif defined(CONFIG_BSD)
115 ptr = memalign(alignment, size);
117 trace_qemu_memalign(alignment, size, ptr);
121 void *qemu_memalign(size_t alignment, size_t size)
123 return qemu_oom_check(qemu_try_memalign(alignment, size));
126 /* alloc shared memory pages */
127 void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment)
129 size_t align = QEMU_VMALLOC_ALIGN;
130 void *ptr = qemu_ram_mmap(-1, size, align, false);
132 if (ptr == MAP_FAILED) {
140 trace_qemu_anon_ram_alloc(size, ptr);
144 void qemu_vfree(void *ptr)
146 trace_qemu_vfree(ptr);
150 void qemu_anon_ram_free(void *ptr, size_t size)
152 trace_qemu_anon_ram_free(ptr, size);
153 qemu_ram_munmap(ptr, size);
156 void qemu_set_block(int fd)
159 f = fcntl(fd, F_GETFL);
160 fcntl(fd, F_SETFL, f & ~O_NONBLOCK);
163 void qemu_set_nonblock(int fd)
166 f = fcntl(fd, F_GETFL);
167 fcntl(fd, F_SETFL, f | O_NONBLOCK);
170 int socket_set_fast_reuse(int fd)
174 ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
175 (const char *)&val, sizeof(val));
182 void qemu_set_cloexec(int fd)
185 f = fcntl(fd, F_GETFD);
186 fcntl(fd, F_SETFD, f | FD_CLOEXEC);
190 * Creates a pipe with FD_CLOEXEC set on both file descriptors
192 int qemu_pipe(int pipefd[2])
197 ret = pipe2(pipefd, O_CLOEXEC);
198 if (ret != -1 || errno != ENOSYS) {
204 qemu_set_cloexec(pipefd[0]);
205 qemu_set_cloexec(pipefd[1]);
211 int qemu_utimens(const char *path, const struct timespec *times)
213 struct timeval tv[2], tv_now;
216 #ifdef CONFIG_UTIMENSAT
219 ret = utimensat(AT_FDCWD, path, times, AT_SYMLINK_NOFOLLOW);
220 if (ret != -1 || errno != ENOSYS) {
224 /* Fallback: use utimes() instead of utimensat() */
226 /* happy if special cases */
227 if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT) {
230 if (times[0].tv_nsec == UTIME_NOW && times[1].tv_nsec == UTIME_NOW) {
231 return utimes(path, NULL);
234 /* prepare for hard cases */
235 if (times[0].tv_nsec == UTIME_NOW || times[1].tv_nsec == UTIME_NOW) {
236 gettimeofday(&tv_now, NULL);
238 if (times[0].tv_nsec == UTIME_OMIT || times[1].tv_nsec == UTIME_OMIT) {
242 for (i = 0; i < 2; i++) {
243 if (times[i].tv_nsec == UTIME_NOW) {
244 tv[i].tv_sec = tv_now.tv_sec;
245 tv[i].tv_usec = tv_now.tv_usec;
246 } else if (times[i].tv_nsec == UTIME_OMIT) {
247 tv[i].tv_sec = (i == 0) ? st.st_atime : st.st_mtime;
250 tv[i].tv_sec = times[i].tv_sec;
251 tv[i].tv_usec = times[i].tv_nsec / 1000;
255 return utimes(path, &tv[0]);
259 qemu_get_local_state_pathname(const char *relative_pathname)
261 return g_strdup_printf("%s/%s", CONFIG_QEMU_LOCALSTATEDIR,
265 void qemu_set_tty_echo(int fd, bool echo)
272 tty.c_lflag |= ECHO | ECHONL | ICANON | IEXTEN;
274 tty.c_lflag &= ~(ECHO | ECHONL | ICANON | IEXTEN);
277 tcsetattr(fd, TCSANOW, &tty);
280 static char exec_dir[PATH_MAX];
282 void qemu_init_exec_dir(const char *argv0)
288 assert(!exec_dir[0]);
290 #if defined(__linux__)
293 len = readlink("/proc/self/exe", buf, sizeof(buf) - 1);
299 #elif defined(__FreeBSD__)
301 static int mib[4] = {CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1};
302 size_t len = sizeof(buf) - 1;
305 if (!sysctl(mib, ARRAY_SIZE(mib), buf, &len, NULL, 0) &&
307 buf[sizeof(buf) - 1] = '\0';
312 /* If we don't have any way of figuring out the actual executable
313 location then try argv[0]. */
318 p = realpath(argv0, buf);
325 pstrcpy(exec_dir, sizeof(exec_dir), dir);
328 char *qemu_get_exec_dir(void)
330 return g_strdup(exec_dir);
333 static sigjmp_buf sigjump;
335 static void sigbus_handler(int signal)
337 siglongjmp(sigjump, 1);
340 void os_mem_prealloc(int fd, char *area, size_t memory)
343 struct sigaction act, oldact;
344 sigset_t set, oldset;
346 memset(&act, 0, sizeof(act));
347 act.sa_handler = &sigbus_handler;
350 ret = sigaction(SIGBUS, &act, &oldact);
352 perror("os_mem_prealloc: failed to install signal handler");
358 sigaddset(&set, SIGBUS);
359 pthread_sigmask(SIG_UNBLOCK, &set, &oldset);
361 if (sigsetjmp(sigjump, 1)) {
362 fprintf(stderr, "os_mem_prealloc: Insufficient free host memory "
363 "pages available to allocate guest RAM\n");
367 size_t hpagesize = qemu_fd_getpagesize(fd);
368 size_t numpages = DIV_ROUND_UP(memory, hpagesize);
370 /* MAP_POPULATE silently ignores failures */
371 for (i = 0; i < numpages; i++) {
372 memset(area + (hpagesize * i), 0, 1);
375 ret = sigaction(SIGBUS, &oldact, NULL);
377 perror("os_mem_prealloc: failed to reinstall signal handler");
381 pthread_sigmask(SIG_SETMASK, &oldset, NULL);
386 static struct termios oldtty;
388 static void term_exit(void)
390 tcsetattr(0, TCSANOW, &oldtty);
393 static void term_init(void)
400 tty.c_iflag &= ~(IGNBRK|BRKINT|PARMRK|ISTRIP
401 |INLCR|IGNCR|ICRNL|IXON);
402 tty.c_oflag |= OPOST;
403 tty.c_lflag &= ~(ECHO|ECHONL|ICANON|IEXTEN);
404 tty.c_cflag &= ~(CSIZE|PARENB);
409 tcsetattr(0, TCSANOW, &tty);
414 int qemu_read_password(char *buf, int buf_size)
419 printf("password: ");
424 ret = read(0, &ch, 1);
426 if (errno == EAGAIN || errno == EINTR) {
431 } else if (ret == 0) {
440 if (i < (buf_size - 1)) {
452 pid_t qemu_fork(Error **errp)
454 sigset_t oldmask, newmask;
455 struct sigaction sig_action;
460 * Need to block signals now, so that child process can safely
461 * kill off caller's signal handlers without a race.
463 sigfillset(&newmask);
464 if (pthread_sigmask(SIG_SETMASK, &newmask, &oldmask) != 0) {
465 error_setg_errno(errp, errno,
466 "cannot block signals");
474 /* attempt to restore signal mask, but ignore failure, to
475 * avoid obscuring the fork failure */
476 (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
477 error_setg_errno(errp, saved_errno,
478 "cannot fork child process");
484 /* Restore our original signal mask now that the child is
485 * safely running. Only documented failures are EFAULT (not
486 * possible, since we are using just-grabbed mask) or EINVAL
487 * (not possible, since we are using correct arguments). */
488 (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
493 /* Clear out all signal handlers from parent so nothing
494 * unexpected can happen in our child once we unblock
496 sig_action.sa_handler = SIG_DFL;
497 sig_action.sa_flags = 0;
498 sigemptyset(&sig_action.sa_mask);
500 for (i = 1; i < NSIG; i++) {
501 /* Only possible errors are EFAULT or EINVAL The former
502 * won't happen, the latter we expect, so no need to check
504 (void)sigaction(i, &sig_action, NULL);
507 /* Unmask all signals in child, since we've no idea what the
508 * caller's done with their signal mask and don't want to
509 * propagate that to children */
510 sigemptyset(&newmask);
511 if (pthread_sigmask(SIG_SETMASK, &newmask, NULL) != 0) {
512 Error *local_err = NULL;
513 error_setg_errno(&local_err, errno,
514 "cannot unblock signals");
515 error_report_err(local_err);