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()
49 #define HUGETLBFS_MAGIC 0x958458f6
55 #include <glib/gprintf.h>
57 #include "config-host.h"
58 #include "sysemu/sysemu.h"
60 #include "qemu/sockets.h"
64 #include <sys/signal.h>
67 #include <sys/syscall.h>
72 #include <sys/sysctl.h>
75 int qemu_get_thread_id(void)
77 #if defined(__linux__)
78 return syscall(SYS_gettid);
84 int qemu_daemon(int nochdir, int noclose)
86 return daemon(nochdir, noclose);
89 void *qemu_oom_check(void *ptr)
92 fprintf(stderr, "Failed to allocate memory: %s\n", strerror(errno));
98 void *qemu_try_memalign(size_t alignment, size_t size)
102 if (alignment < sizeof(void*)) {
103 alignment = sizeof(void*);
106 #if defined(_POSIX_C_SOURCE) && !defined(__sun__)
108 ret = posix_memalign(&ptr, alignment, size);
113 #elif defined(CONFIG_BSD)
116 ptr = memalign(alignment, size);
118 trace_qemu_memalign(alignment, size, ptr);
122 void *qemu_memalign(size_t alignment, size_t size)
124 return qemu_oom_check(qemu_try_memalign(alignment, size));
127 /* alloc shared memory pages */
128 void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment)
130 size_t align = QEMU_VMALLOC_ALIGN;
131 size_t total = size + align;
132 void *ptr = mmap(0, total, PROT_NONE, MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
133 size_t offset = QEMU_ALIGN_UP((uintptr_t)ptr, align) - (uintptr_t)ptr;
136 if (ptr == MAP_FAILED) {
144 ptr1 = mmap(ptr + offset, size, PROT_READ | PROT_WRITE,
145 MAP_FIXED | MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
146 if (ptr1 == MAP_FAILED) {
155 munmap(ptr - offset, offset);
157 if (total > size + getpagesize()) {
158 munmap(ptr + size + getpagesize(), total - size - getpagesize());
161 trace_qemu_anon_ram_alloc(size, ptr);
165 void qemu_vfree(void *ptr)
167 trace_qemu_vfree(ptr);
171 void qemu_anon_ram_free(void *ptr, size_t size)
173 trace_qemu_anon_ram_free(ptr, size);
175 munmap(ptr, size + getpagesize());
179 void qemu_set_block(int fd)
182 f = fcntl(fd, F_GETFL);
183 fcntl(fd, F_SETFL, f & ~O_NONBLOCK);
186 void qemu_set_nonblock(int fd)
189 f = fcntl(fd, F_GETFL);
190 fcntl(fd, F_SETFL, f | O_NONBLOCK);
193 int socket_set_fast_reuse(int fd)
197 ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
198 (const char *)&val, sizeof(val));
205 void qemu_set_cloexec(int fd)
208 f = fcntl(fd, F_GETFD);
209 fcntl(fd, F_SETFD, f | FD_CLOEXEC);
213 * Creates a pipe with FD_CLOEXEC set on both file descriptors
215 int qemu_pipe(int pipefd[2])
220 ret = pipe2(pipefd, O_CLOEXEC);
221 if (ret != -1 || errno != ENOSYS) {
227 qemu_set_cloexec(pipefd[0]);
228 qemu_set_cloexec(pipefd[1]);
234 int qemu_utimens(const char *path, const struct timespec *times)
236 struct timeval tv[2], tv_now;
239 #ifdef CONFIG_UTIMENSAT
242 ret = utimensat(AT_FDCWD, path, times, AT_SYMLINK_NOFOLLOW);
243 if (ret != -1 || errno != ENOSYS) {
247 /* Fallback: use utimes() instead of utimensat() */
249 /* happy if special cases */
250 if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT) {
253 if (times[0].tv_nsec == UTIME_NOW && times[1].tv_nsec == UTIME_NOW) {
254 return utimes(path, NULL);
257 /* prepare for hard cases */
258 if (times[0].tv_nsec == UTIME_NOW || times[1].tv_nsec == UTIME_NOW) {
259 gettimeofday(&tv_now, NULL);
261 if (times[0].tv_nsec == UTIME_OMIT || times[1].tv_nsec == UTIME_OMIT) {
265 for (i = 0; i < 2; i++) {
266 if (times[i].tv_nsec == UTIME_NOW) {
267 tv[i].tv_sec = tv_now.tv_sec;
268 tv[i].tv_usec = tv_now.tv_usec;
269 } else if (times[i].tv_nsec == UTIME_OMIT) {
270 tv[i].tv_sec = (i == 0) ? st.st_atime : st.st_mtime;
273 tv[i].tv_sec = times[i].tv_sec;
274 tv[i].tv_usec = times[i].tv_nsec / 1000;
278 return utimes(path, &tv[0]);
282 qemu_get_local_state_pathname(const char *relative_pathname)
284 return g_strdup_printf("%s/%s", CONFIG_QEMU_LOCALSTATEDIR,
288 void qemu_set_tty_echo(int fd, bool echo)
295 tty.c_lflag |= ECHO | ECHONL | ICANON | IEXTEN;
297 tty.c_lflag &= ~(ECHO | ECHONL | ICANON | IEXTEN);
300 tcsetattr(fd, TCSANOW, &tty);
303 static char exec_dir[PATH_MAX];
305 void qemu_init_exec_dir(const char *argv0)
311 assert(!exec_dir[0]);
313 #if defined(__linux__)
316 len = readlink("/proc/self/exe", buf, sizeof(buf) - 1);
322 #elif defined(__FreeBSD__)
324 static int mib[4] = {CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1};
325 size_t len = sizeof(buf) - 1;
328 if (!sysctl(mib, ARRAY_SIZE(mib), buf, &len, NULL, 0) &&
330 buf[sizeof(buf) - 1] = '\0';
335 /* If we don't have any way of figuring out the actual executable
336 location then try argv[0]. */
341 p = realpath(argv0, buf);
348 pstrcpy(exec_dir, sizeof(exec_dir), dir);
351 char *qemu_get_exec_dir(void)
353 return g_strdup(exec_dir);
356 static sigjmp_buf sigjump;
358 static void sigbus_handler(int signal)
360 siglongjmp(sigjump, 1);
363 static size_t fd_getpagesize(int fd)
371 ret = fstatfs(fd, &fs);
372 } while (ret != 0 && errno == EINTR);
374 if (ret == 0 && fs.f_type == HUGETLBFS_MAGIC) {
380 return getpagesize();
383 void os_mem_prealloc(int fd, char *area, size_t memory)
386 struct sigaction act, oldact;
387 sigset_t set, oldset;
389 memset(&act, 0, sizeof(act));
390 act.sa_handler = &sigbus_handler;
393 ret = sigaction(SIGBUS, &act, &oldact);
395 perror("os_mem_prealloc: failed to install signal handler");
401 sigaddset(&set, SIGBUS);
402 pthread_sigmask(SIG_UNBLOCK, &set, &oldset);
404 if (sigsetjmp(sigjump, 1)) {
405 fprintf(stderr, "os_mem_prealloc: Insufficient free host memory "
406 "pages available to allocate guest RAM\n");
410 size_t hpagesize = fd_getpagesize(fd);
411 size_t numpages = DIV_ROUND_UP(memory, hpagesize);
413 /* MAP_POPULATE silently ignores failures */
414 for (i = 0; i < numpages; i++) {
415 memset(area + (hpagesize * i), 0, 1);
418 ret = sigaction(SIGBUS, &oldact, NULL);
420 perror("os_mem_prealloc: failed to reinstall signal handler");
424 pthread_sigmask(SIG_SETMASK, &oldset, NULL);
429 static struct termios oldtty;
431 static void term_exit(void)
433 tcsetattr(0, TCSANOW, &oldtty);
436 static void term_init(void)
443 tty.c_iflag &= ~(IGNBRK|BRKINT|PARMRK|ISTRIP
444 |INLCR|IGNCR|ICRNL|IXON);
445 tty.c_oflag |= OPOST;
446 tty.c_lflag &= ~(ECHO|ECHONL|ICANON|IEXTEN);
447 tty.c_cflag &= ~(CSIZE|PARENB);
452 tcsetattr(0, TCSANOW, &tty);
457 int qemu_read_password(char *buf, int buf_size)
462 printf("password: ");
467 ret = read(0, &ch, 1);
469 if (errno == EAGAIN || errno == EINTR) {
474 } else if (ret == 0) {
483 if (i < (buf_size - 1)) {
495 pid_t qemu_fork(Error **errp)
497 sigset_t oldmask, newmask;
498 struct sigaction sig_action;
503 * Need to block signals now, so that child process can safely
504 * kill off caller's signal handlers without a race.
506 sigfillset(&newmask);
507 if (pthread_sigmask(SIG_SETMASK, &newmask, &oldmask) != 0) {
508 error_setg_errno(errp, errno,
509 "cannot block signals");
517 /* attempt to restore signal mask, but ignore failure, to
518 * avoid obscuring the fork failure */
519 (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
520 error_setg_errno(errp, saved_errno,
521 "cannot fork child process");
527 /* Restore our original signal mask now that the child is
528 * safely running. Only documented failures are EFAULT (not
529 * possible, since we are using just-grabbed mask) or EINVAL
530 * (not possible, since we are using correct arguments). */
531 (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
536 /* Clear out all signal handlers from parent so nothing
537 * unexpected can happen in our child once we unblock
539 sig_action.sa_handler = SIG_DFL;
540 sig_action.sa_flags = 0;
541 sigemptyset(&sig_action.sa_mask);
543 for (i = 1; i < NSIG; i++) {
544 /* Only possible errors are EFAULT or EINVAL The former
545 * won't happen, the latter we expect, so no need to check
547 (void)sigaction(i, &sig_action, NULL);
550 /* Unmask all signals in child, since we've no idea what the
551 * caller's done with their signal mask and don't want to
552 * propagate that to children */
553 sigemptyset(&newmask);
554 if (pthread_sigmask(SIG_SETMASK, &newmask, NULL) != 0) {
555 Error *local_err = NULL;
556 error_setg_errno(&local_err, errno,
557 "cannot unblock signals");
558 error_report_err(local_err);