2 * QEMU low level functions
4 * Copyright (c) 2003 Fabrice Bellard
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24 #include "qemu/osdep.h"
25 #include "qapi/error.h"
26 #include "qemu/cutils.h"
27 #include "qemu/sockets.h"
28 #include "qemu/error-report.h"
29 #include "qemu/madvise.h"
30 #include "qemu/mprotect.h"
31 #include "qemu/hw-version.h"
32 #include "monitor/monitor.h"
34 static const char *hw_version = QEMU_HW_VERSION;
36 int socket_set_cork(int fd, int v)
38 #if defined(SOL_TCP) && defined(TCP_CORK)
39 return setsockopt(fd, SOL_TCP, TCP_CORK, &v, sizeof(v));
45 int socket_set_nodelay(int fd)
48 return setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &v, sizeof(v));
51 int qemu_madvise(void *addr, size_t len, int advice)
53 if (advice == QEMU_MADV_INVALID) {
57 #if defined(CONFIG_MADVISE)
58 return madvise(addr, len, advice);
59 #elif defined(CONFIG_POSIX_MADVISE)
60 return posix_madvise(addr, len, advice);
67 static int qemu_mprotect__osdep(void *addr, size_t size, int prot)
69 g_assert(!((uintptr_t)addr & ~qemu_real_host_page_mask()));
70 g_assert(!(size & ~qemu_real_host_page_mask()));
75 if (!VirtualProtect(addr, size, prot, &old_protect)) {
76 g_autofree gchar *emsg = g_win32_error_message(GetLastError());
77 error_report("%s: VirtualProtect failed: %s", __func__, emsg);
82 if (mprotect(addr, size, prot)) {
83 error_report("%s: mprotect failed: %s", __func__, strerror(errno));
90 int qemu_mprotect_rw(void *addr, size_t size)
93 return qemu_mprotect__osdep(addr, size, PAGE_READWRITE);
95 return qemu_mprotect__osdep(addr, size, PROT_READ | PROT_WRITE);
99 int qemu_mprotect_rwx(void *addr, size_t size)
102 return qemu_mprotect__osdep(addr, size, PAGE_EXECUTE_READWRITE);
104 return qemu_mprotect__osdep(addr, size, PROT_READ | PROT_WRITE | PROT_EXEC);
108 int qemu_mprotect_none(void *addr, size_t size)
111 return qemu_mprotect__osdep(addr, size, PAGE_NOACCESS);
113 return qemu_mprotect__osdep(addr, size, PROT_NONE);
119 static int fcntl_op_setlk = -1;
120 static int fcntl_op_getlk = -1;
123 * Dups an fd and sets the flags
125 int qemu_dup_flags(int fd, int flags)
136 dup_flags = fcntl(ret, F_GETFL);
137 if (dup_flags == -1) {
141 if ((flags & O_SYNC) != (dup_flags & O_SYNC)) {
146 /* Set/unset flags that we can with fcntl */
147 if (fcntl(ret, F_SETFL, flags) == -1) {
151 /* Truncate the file in the cases that open() would truncate it */
152 if (flags & O_TRUNC ||
153 ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))) {
154 if (ftruncate(ret, 0) == -1) {
173 #ifdef F_DUPFD_CLOEXEC
174 ret = fcntl(fd, F_DUPFD_CLOEXEC, 0);
178 qemu_set_cloexec(ret);
184 static int qemu_parse_fdset(const char *param)
186 return qemu_parse_fd(param);
189 static void qemu_probe_lock_ops(void)
191 if (fcntl_op_setlk == -1) {
196 .l_whence = SEEK_SET,
202 fd = open("/dev/null", O_RDWR);
205 "Failed to open /dev/null for OFD lock probing: %s\n",
207 fcntl_op_setlk = F_SETLK;
208 fcntl_op_getlk = F_GETLK;
211 ret = fcntl(fd, F_OFD_GETLK, &fl);
214 fcntl_op_setlk = F_OFD_SETLK;
215 fcntl_op_getlk = F_OFD_GETLK;
217 fcntl_op_setlk = F_SETLK;
218 fcntl_op_getlk = F_GETLK;
221 fcntl_op_setlk = F_SETLK;
222 fcntl_op_getlk = F_GETLK;
227 bool qemu_has_ofd_lock(void)
229 qemu_probe_lock_ops();
231 return fcntl_op_setlk == F_OFD_SETLK;
237 static int qemu_lock_fcntl(int fd, int64_t start, int64_t len, int fl_type)
241 .l_whence = SEEK_SET,
246 qemu_probe_lock_ops();
247 ret = RETRY_ON_EINTR(fcntl(fd, fcntl_op_setlk, &fl));
248 return ret == -1 ? -errno : 0;
251 int qemu_lock_fd(int fd, int64_t start, int64_t len, bool exclusive)
253 return qemu_lock_fcntl(fd, start, len, exclusive ? F_WRLCK : F_RDLCK);
256 int qemu_unlock_fd(int fd, int64_t start, int64_t len)
258 return qemu_lock_fcntl(fd, start, len, F_UNLCK);
261 int qemu_lock_fd_test(int fd, int64_t start, int64_t len, bool exclusive)
265 .l_whence = SEEK_SET,
268 .l_type = exclusive ? F_WRLCK : F_RDLCK,
270 qemu_probe_lock_ops();
271 ret = fcntl(fd, fcntl_op_getlk, &fl);
275 return fl.l_type == F_UNLCK ? 0 : -EAGAIN;
280 static int qemu_open_cloexec(const char *name, int flags, mode_t mode)
284 ret = open(name, flags | O_CLOEXEC, mode);
286 ret = open(name, flags, mode);
288 qemu_set_cloexec(ret);
295 * Opens a file with FD_CLOEXEC set
298 qemu_open_internal(const char *name, int flags, mode_t mode, Error **errp)
303 const char *fdset_id_str;
305 /* Attempt dup of fd from fd set */
306 if (strstart(name, "/dev/fdset/", &fdset_id_str)) {
310 fdset_id = qemu_parse_fdset(fdset_id_str);
311 if (fdset_id == -1) {
312 error_setg(errp, "Could not parse fdset %s", name);
317 dupfd = monitor_fdset_dup_fd_add(fdset_id, flags);
319 error_setg_errno(errp, errno, "Could not dup FD for %s flags %x",
328 ret = qemu_open_cloexec(name, flags, mode);
331 const char *action = flags & O_CREAT ? "create" : "open";
333 /* Give more helpful error message for O_DIRECT */
334 if (errno == EINVAL && (flags & O_DIRECT)) {
335 ret = open(name, flags & ~O_DIRECT, mode);
338 error_setg(errp, "Could not %s '%s': "
339 "filesystem does not support O_DIRECT",
341 errno = EINVAL; /* restore first open()'s errno */
345 #endif /* O_DIRECT */
346 error_setg_errno(errp, errno, "Could not %s '%s'",
354 int qemu_open(const char *name, int flags, Error **errp)
356 assert(!(flags & O_CREAT));
358 return qemu_open_internal(name, flags, 0, errp);
362 int qemu_create(const char *name, int flags, mode_t mode, Error **errp)
364 assert(!(flags & O_CREAT));
366 return qemu_open_internal(name, flags | O_CREAT, mode, errp);
370 int qemu_open_old(const char *name, int flags, ...)
377 if (flags & O_CREAT) {
378 mode = va_arg(ap, int);
382 ret = qemu_open_internal(name, flags, mode, NULL);
385 if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
386 error_report("file system may not support O_DIRECT");
387 errno = EINVAL; /* in case it was clobbered */
389 #endif /* O_DIRECT */
394 int qemu_close(int fd)
398 /* Close fd that was dup'd from an fdset */
399 fdset_id = monitor_fdset_dup_fd_find(fd);
400 if (fdset_id != -1) {
405 monitor_fdset_dup_fd_remove(fd);
415 * Delete a file from the filesystem, unless the filename is /dev/fdset/...
417 * Returns: On success, zero is returned. On error, -1 is returned,
418 * and errno is set appropriately.
420 int qemu_unlink(const char *name)
422 if (g_str_has_prefix(name, "/dev/fdset/")) {
430 * A variant of write(2) which handles partial write.
432 * Return the number of bytes transferred.
433 * Set errno if fewer than `count' bytes are written.
435 * This function don't work with non-blocking fd's.
436 * Any of the possibilities with non-blocking fd's is bad:
437 * - return a short write (then name is wrong)
438 * - busy wait adding (errno == EAGAIN) to the loop
440 ssize_t qemu_write_full(int fd, const void *buf, size_t count)
446 ret = write(fd, buf, count);
462 * Opens a socket with FD_CLOEXEC set
464 int qemu_socket(int domain, int type, int protocol)
469 ret = socket(domain, type | SOCK_CLOEXEC, protocol);
470 if (ret != -1 || errno != EINVAL) {
474 ret = socket(domain, type, protocol);
476 qemu_set_cloexec(ret);
483 * Accept a connection and set FD_CLOEXEC
485 int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
489 #ifdef CONFIG_ACCEPT4
490 ret = accept4(s, addr, addrlen, SOCK_CLOEXEC);
491 if (ret != -1 || errno != ENOSYS) {
495 ret = accept(s, addr, addrlen);
497 qemu_set_cloexec(ret);
503 ssize_t qemu_send_full(int s, const void *buf, size_t count)
509 ret = send(s, buf, count, 0);
511 if (errno == EINTR) {
525 void qemu_set_hw_version(const char *version)
527 hw_version = version;
530 const char *qemu_hw_version(void)
536 static void socket_cleanup(void)
542 int socket_init(void)
548 ret = WSAStartup(MAKEWORD(2, 2), &Data);
550 err = WSAGetLastError();
551 fprintf(stderr, "WSAStartup: %d\n", err);
554 atexit(socket_cleanup);
562 readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write)
567 while (i < iov_cnt) {
569 ? write(fd, iov[i].iov_base + off, iov[i].iov_len - off)
570 : read(fd, iov[i].iov_base + off, iov[i].iov_len - off);
574 if (off < iov[i].iov_len) {
579 } else if (errno == EINTR) {
582 /* else it is some "other" error,
583 * only return if there was no data processed. */
596 readv(int fd, const struct iovec *iov, int iov_cnt)
598 return readv_writev(fd, iov, iov_cnt, false);
602 writev(int fd, const struct iovec *iov, int iov_cnt)
604 return readv_writev(fd, iov, iov_cnt, true);
609 * Make sure data goes on disk, but if possible do not bother to
610 * write out the inode just for timestamp updates.
612 * Unfortunately even in 2009 many operating systems do not support
613 * fdatasync and have to fall back to fsync.
615 int qemu_fdatasync(int fd)
617 #ifdef CONFIG_FDATASYNC
618 return fdatasync(fd);