blob: e996c4744af1aee49a0ae0ff2f92430360ea3d36 [file] [log] [blame]
bellardea888122004-02-16 22:12:40 +00001/*
2 * QEMU low level functions
ths5fafdf22007-09-16 21:08:06 +00003 *
bellardea888122004-02-16 22:12:40 +00004 * Copyright (c) 2003 Fabrice Bellard
ths5fafdf22007-09-16 21:08:06 +00005 *
bellardea888122004-02-16 22:12:40 +00006 * 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:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
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
22 * THE SOFTWARE.
23 */
Peter Maydellaafd7582016-01-29 17:49:55 +000024#include "qemu/osdep.h"
Daniel P. Berrangéebb3d492020-08-21 17:45:02 +010025#include "qapi/error.h"
Veronia Bahaaf348b6d2016-03-20 19:16:19 +020026#include "qemu/cutils.h"
Paolo Bonzini1de7afc2012-12-17 18:20:00 +010027#include "qemu/sockets.h"
Markus Armbrusterd49b6832015-03-17 18:29:20 +010028#include "qemu/error-report.h"
Peter Maydellb85ea5f2022-02-08 20:08:52 +000029#include "qemu/madvise.h"
Peter Maydellf2241d12022-02-08 20:08:53 +000030#include "qemu/mprotect.h"
Peter Maydell15e09912022-02-08 20:08:56 +000031#include "qemu/hw-version.h"
Paolo Bonzini83c90892012-12-17 18:19:49 +010032#include "monitor/monitor.h"
aliguori03ff3ca2008-09-15 15:51:35 +000033
Eduardo Habkostd4943522016-04-09 16:42:44 -030034static const char *hw_version = QEMU_HW_VERSION;
Crístian Viana93bfef42012-05-30 00:35:51 -030035
Paolo Bonzini128aa582011-09-21 12:36:48 +020036int socket_set_cork(int fd, int v)
37{
38#if defined(SOL_TCP) && defined(TCP_CORK)
Marc-André Lureaue7b79422022-02-19 01:34:50 +040039 return setsockopt(fd, SOL_TCP, TCP_CORK, &v, sizeof(v));
Paolo Bonzini128aa582011-09-21 12:36:48 +020040#else
41 return 0;
42#endif
43}
44
MORITA Kazutakabf1c8522013-02-22 12:39:50 +090045int socket_set_nodelay(int fd)
46{
47 int v = 1;
Marc-André Lureaue7b79422022-02-19 01:34:50 +040048 return setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &v, sizeof(v));
MORITA Kazutakabf1c8522013-02-22 12:39:50 +090049}
50
Andreas Färbere78815a2010-09-25 11:26:05 +000051int qemu_madvise(void *addr, size_t len, int advice)
52{
53 if (advice == QEMU_MADV_INVALID) {
54 errno = EINVAL;
55 return -1;
56 }
57#if defined(CONFIG_MADVISE)
58 return madvise(addr, len, advice);
59#elif defined(CONFIG_POSIX_MADVISE)
60 return posix_madvise(addr, len, advice);
61#else
62 errno = EINVAL;
63 return -1;
64#endif
65}
66
Emilio G. Cota5fa64b32017-07-15 02:28:47 -040067static int qemu_mprotect__osdep(void *addr, size_t size, int prot)
68{
Marc-André Lureau8e3b0cb2022-03-23 19:57:22 +040069 g_assert(!((uintptr_t)addr & ~qemu_real_host_page_mask()));
70 g_assert(!(size & ~qemu_real_host_page_mask()));
Emilio G. Cota5fa64b32017-07-15 02:28:47 -040071
72#ifdef _WIN32
73 DWORD old_protect;
74
75 if (!VirtualProtect(addr, size, prot, &old_protect)) {
Philippe Mathieu-Daudécf0c76c2020-02-28 11:07:24 +010076 g_autofree gchar *emsg = g_win32_error_message(GetLastError());
77 error_report("%s: VirtualProtect failed: %s", __func__, emsg);
Emilio G. Cota5fa64b32017-07-15 02:28:47 -040078 return -1;
79 }
80 return 0;
81#else
82 if (mprotect(addr, size, prot)) {
83 error_report("%s: mprotect failed: %s", __func__, strerror(errno));
84 return -1;
85 }
86 return 0;
87#endif
88}
89
Richard Hendersond7107fc2021-03-10 17:47:33 -060090int qemu_mprotect_rw(void *addr, size_t size)
91{
92#ifdef _WIN32
93 return qemu_mprotect__osdep(addr, size, PAGE_READWRITE);
94#else
95 return qemu_mprotect__osdep(addr, size, PROT_READ | PROT_WRITE);
96#endif
97}
98
Emilio G. Cota5fa64b32017-07-15 02:28:47 -040099int qemu_mprotect_rwx(void *addr, size_t size)
100{
101#ifdef _WIN32
102 return qemu_mprotect__osdep(addr, size, PAGE_EXECUTE_READWRITE);
103#else
104 return qemu_mprotect__osdep(addr, size, PROT_READ | PROT_WRITE | PROT_EXEC);
105#endif
106}
107
108int qemu_mprotect_none(void *addr, size_t size)
109{
110#ifdef _WIN32
111 return qemu_mprotect__osdep(addr, size, PAGE_NOACCESS);
112#else
113 return qemu_mprotect__osdep(addr, size, PROT_NONE);
114#endif
115}
116
Corey Bryantadb696f2012-08-14 16:43:47 -0400117#ifndef _WIN32
Fam Zhengca749952017-08-11 19:44:46 +0800118
119static int fcntl_op_setlk = -1;
120static int fcntl_op_getlk = -1;
121
Corey Bryantadb696f2012-08-14 16:43:47 -0400122/*
123 * Dups an fd and sets the flags
124 */
Daniel P. Berrangé60efffa2020-08-27 13:27:00 +0100125int qemu_dup_flags(int fd, int flags)
Corey Bryantadb696f2012-08-14 16:43:47 -0400126{
127 int ret;
128 int serrno;
129 int dup_flags;
Corey Bryantadb696f2012-08-14 16:43:47 -0400130
Fam Zheng761d1dd2016-06-22 20:53:19 +0800131 ret = qemu_dup(fd);
Corey Bryantadb696f2012-08-14 16:43:47 -0400132 if (ret == -1) {
133 goto fail;
134 }
135
136 dup_flags = fcntl(ret, F_GETFL);
137 if (dup_flags == -1) {
138 goto fail;
139 }
140
141 if ((flags & O_SYNC) != (dup_flags & O_SYNC)) {
142 errno = EINVAL;
143 goto fail;
144 }
145
146 /* Set/unset flags that we can with fcntl */
Corey Bryant3b6eda22012-10-18 16:41:04 -0400147 if (fcntl(ret, F_SETFL, flags) == -1) {
Corey Bryantadb696f2012-08-14 16:43:47 -0400148 goto fail;
149 }
150
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) {
155 goto fail;
156 }
157 }
158
159 return ret;
160
161fail:
162 serrno = errno;
163 if (ret != -1) {
164 close(ret);
165 }
166 errno = serrno;
167 return -1;
168}
Paolo Bonzini0100fbb2012-10-29 15:19:18 +0100169
Fam Zheng761d1dd2016-06-22 20:53:19 +0800170int qemu_dup(int fd)
171{
172 int ret;
173#ifdef F_DUPFD_CLOEXEC
174 ret = fcntl(fd, F_DUPFD_CLOEXEC, 0);
175#else
176 ret = dup(fd);
177 if (ret != -1) {
178 qemu_set_cloexec(ret);
179 }
180#endif
181 return ret;
182}
183
Paolo Bonzini0100fbb2012-10-29 15:19:18 +0100184static int qemu_parse_fdset(const char *param)
185{
186 return qemu_parse_fd(param);
187}
Fam Zheng13461fd2017-05-03 00:35:54 +0800188
Fam Zhengca749952017-08-11 19:44:46 +0800189static void qemu_probe_lock_ops(void)
190{
191 if (fcntl_op_setlk == -1) {
192#ifdef F_OFD_SETLK
193 int fd;
194 int ret;
195 struct flock fl = {
196 .l_whence = SEEK_SET,
197 .l_start = 0,
198 .l_len = 0,
199 .l_type = F_WRLCK,
200 };
201
202 fd = open("/dev/null", O_RDWR);
203 if (fd < 0) {
204 fprintf(stderr,
205 "Failed to open /dev/null for OFD lock probing: %s\n",
206 strerror(errno));
207 fcntl_op_setlk = F_SETLK;
208 fcntl_op_getlk = F_GETLK;
209 return;
210 }
211 ret = fcntl(fd, F_OFD_GETLK, &fl);
212 close(fd);
213 if (!ret) {
214 fcntl_op_setlk = F_OFD_SETLK;
215 fcntl_op_getlk = F_OFD_GETLK;
216 } else {
217 fcntl_op_setlk = F_SETLK;
218 fcntl_op_getlk = F_GETLK;
219 }
220#else
221 fcntl_op_setlk = F_SETLK;
222 fcntl_op_getlk = F_GETLK;
223#endif
224 }
225}
226
227bool qemu_has_ofd_lock(void)
228{
229 qemu_probe_lock_ops();
230#ifdef F_OFD_SETLK
231 return fcntl_op_setlk == F_OFD_SETLK;
232#else
233 return false;
234#endif
235}
236
Fam Zheng13461fd2017-05-03 00:35:54 +0800237static int qemu_lock_fcntl(int fd, int64_t start, int64_t len, int fl_type)
238{
Fam Zheng13461fd2017-05-03 00:35:54 +0800239 int ret;
240 struct flock fl = {
241 .l_whence = SEEK_SET,
242 .l_start = start,
243 .l_len = len,
244 .l_type = fl_type,
245 };
Fam Zhengca749952017-08-11 19:44:46 +0800246 qemu_probe_lock_ops();
Nikita Ivanov37b0b242022-10-23 12:04:22 +0300247 ret = RETRY_ON_EINTR(fcntl(fd, fcntl_op_setlk, &fl));
Fam Zheng13461fd2017-05-03 00:35:54 +0800248 return ret == -1 ? -errno : 0;
Fam Zheng13461fd2017-05-03 00:35:54 +0800249}
250
251int qemu_lock_fd(int fd, int64_t start, int64_t len, bool exclusive)
252{
253 return qemu_lock_fcntl(fd, start, len, exclusive ? F_WRLCK : F_RDLCK);
254}
255
256int qemu_unlock_fd(int fd, int64_t start, int64_t len)
257{
258 return qemu_lock_fcntl(fd, start, len, F_UNLCK);
259}
260
261int qemu_lock_fd_test(int fd, int64_t start, int64_t len, bool exclusive)
262{
Fam Zheng13461fd2017-05-03 00:35:54 +0800263 int ret;
264 struct flock fl = {
265 .l_whence = SEEK_SET,
266 .l_start = start,
267 .l_len = len,
268 .l_type = exclusive ? F_WRLCK : F_RDLCK,
269 };
Fam Zhengca749952017-08-11 19:44:46 +0800270 qemu_probe_lock_ops();
271 ret = fcntl(fd, fcntl_op_getlk, &fl);
Fam Zheng13461fd2017-05-03 00:35:54 +0800272 if (ret == -1) {
273 return -errno;
274 } else {
275 return fl.l_type == F_UNLCK ? 0 : -EAGAIN;
276 }
Fam Zheng13461fd2017-05-03 00:35:54 +0800277}
Corey Bryantadb696f2012-08-14 16:43:47 -0400278#endif
aliguori03ff3ca2008-09-15 15:51:35 +0000279
Daniel P. Berrangéc2069ff2020-09-02 16:58:06 +0100280static int qemu_open_cloexec(const char *name, int flags, mode_t mode)
281{
282 int ret;
283#ifdef O_CLOEXEC
284 ret = open(name, flags | O_CLOEXEC, mode);
285#else
286 ret = open(name, flags, mode);
287 if (ret >= 0) {
288 qemu_set_cloexec(ret);
289 }
290#endif
291 return ret;
292}
293
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100294/*
295 * Opens a file with FD_CLOEXEC set
296 */
Daniel P. Berrangébf93d2a2020-07-01 16:30:35 +0100297static int
Daniel P. Berrangéebb3d492020-08-21 17:45:02 +0100298qemu_open_internal(const char *name, int flags, mode_t mode, Error **errp)
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100299{
300 int ret;
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100301
Corey Bryantadb696f2012-08-14 16:43:47 -0400302#ifndef _WIN32
303 const char *fdset_id_str;
304
305 /* Attempt dup of fd from fd set */
306 if (strstart(name, "/dev/fdset/", &fdset_id_str)) {
307 int64_t fdset_id;
Daniel P. Berrangé60efffa2020-08-27 13:27:00 +0100308 int dupfd;
Corey Bryantadb696f2012-08-14 16:43:47 -0400309
310 fdset_id = qemu_parse_fdset(fdset_id_str);
311 if (fdset_id == -1) {
Daniel P. Berrangéebb3d492020-08-21 17:45:02 +0100312 error_setg(errp, "Could not parse fdset %s", name);
Corey Bryantadb696f2012-08-14 16:43:47 -0400313 errno = EINVAL;
314 return -1;
315 }
316
Daniel P. Berrangé60efffa2020-08-27 13:27:00 +0100317 dupfd = monitor_fdset_dup_fd_add(fdset_id, flags);
Corey Bryantadb696f2012-08-14 16:43:47 -0400318 if (dupfd == -1) {
Daniel P. Berrangéebb3d492020-08-21 17:45:02 +0100319 error_setg_errno(errp, errno, "Could not dup FD for %s flags %x",
320 name, flags);
Corey Bryantadb696f2012-08-14 16:43:47 -0400321 return -1;
322 }
323
Corey Bryantadb696f2012-08-14 16:43:47 -0400324 return dupfd;
325 }
326#endif
327
Daniel P. Berrangéc2069ff2020-09-02 16:58:06 +0100328 ret = qemu_open_cloexec(name, flags, mode);
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100329
Daniel P. Berrangéebb3d492020-08-21 17:45:02 +0100330 if (ret == -1) {
331 const char *action = flags & O_CREAT ? "create" : "open";
Daniel P. Berrangé661b3e82020-07-21 16:17:35 +0100332#ifdef O_DIRECT
333 /* Give more helpful error message for O_DIRECT */
334 if (errno == EINVAL && (flags & O_DIRECT)) {
335 ret = open(name, flags & ~O_DIRECT, mode);
336 if (ret != -1) {
337 close(ret);
338 error_setg(errp, "Could not %s '%s': "
339 "filesystem does not support O_DIRECT",
340 action, name);
341 errno = EINVAL; /* restore first open()'s errno */
342 return -1;
343 }
344 }
345#endif /* O_DIRECT */
Daniel P. Berrangéebb3d492020-08-21 17:45:02 +0100346 error_setg_errno(errp, errno, "Could not %s '%s'",
347 action, name);
348 }
349
Daniel P. Berrangébf93d2a2020-07-01 16:30:35 +0100350 return ret;
351}
352
353
Daniel P. Berrangéc490af52020-07-01 16:30:35 +0100354int qemu_open(const char *name, int flags, Error **errp)
355{
356 assert(!(flags & O_CREAT));
357
358 return qemu_open_internal(name, flags, 0, errp);
359}
360
361
362int qemu_create(const char *name, int flags, mode_t mode, Error **errp)
363{
364 assert(!(flags & O_CREAT));
365
366 return qemu_open_internal(name, flags | O_CREAT, mode, errp);
367}
368
369
Daniel P. Berrangébf93d2a2020-07-01 16:30:35 +0100370int qemu_open_old(const char *name, int flags, ...)
371{
372 va_list ap;
373 mode_t mode = 0;
374 int ret;
375
376 va_start(ap, flags);
377 if (flags & O_CREAT) {
378 mode = va_arg(ap, int);
379 }
380 va_end(ap);
381
Daniel P. Berrangéebb3d492020-08-21 17:45:02 +0100382 ret = qemu_open_internal(name, flags, mode, NULL);
Daniel P. Berrangébf93d2a2020-07-01 16:30:35 +0100383
Stefan Hajnoczia5813072013-08-22 11:29:03 +0200384#ifdef O_DIRECT
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 */
388 }
389#endif /* O_DIRECT */
390
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100391 return ret;
392}
393
Corey Bryant2e1e79d2012-08-14 16:43:46 -0400394int qemu_close(int fd)
395{
Corey Bryantadb696f2012-08-14 16:43:47 -0400396 int64_t fdset_id;
397
398 /* Close fd that was dup'd from an fdset */
399 fdset_id = monitor_fdset_dup_fd_find(fd);
400 if (fdset_id != -1) {
401 int ret;
402
403 ret = close(fd);
404 if (ret == 0) {
405 monitor_fdset_dup_fd_remove(fd);
406 }
407
408 return ret;
409 }
410
Corey Bryant2e1e79d2012-08-14 16:43:46 -0400411 return close(fd);
412}
413
Kirill A. Shutemov7b5f6992010-01-20 00:56:08 +0100414/*
Marc-André Lureauee132402019-11-08 18:09:56 +0400415 * Delete a file from the filesystem, unless the filename is /dev/fdset/...
416 *
417 * Returns: On success, zero is returned. On error, -1 is returned,
418 * and errno is set appropriately.
419 */
420int qemu_unlink(const char *name)
421{
422 if (g_str_has_prefix(name, "/dev/fdset/")) {
423 return 0;
424 }
425
426 return unlink(name);
427}
428
429/*
Kirill A. Shutemov7b5f6992010-01-20 00:56:08 +0100430 * A variant of write(2) which handles partial write.
431 *
432 * Return the number of bytes transferred.
433 * Set errno if fewer than `count' bytes are written.
Juan Quintela1298cb62010-03-04 10:00:39 +0100434 *
435 * This function don't work with non-blocking fd's.
zhaolichang8cc360b2020-09-17 15:50:24 +0800436 * Any of the possibilities with non-blocking fd's is bad:
Juan Quintela1298cb62010-03-04 10:00:39 +0100437 * - return a short write (then name is wrong)
438 * - busy wait adding (errno == EAGAIN) to the loop
Kirill A. Shutemov7b5f6992010-01-20 00:56:08 +0100439 */
440ssize_t qemu_write_full(int fd, const void *buf, size_t count)
441{
442 ssize_t ret = 0;
443 ssize_t total = 0;
444
445 while (count) {
446 ret = write(fd, buf, count);
447 if (ret < 0) {
448 if (errno == EINTR)
449 continue;
450 break;
451 }
452
453 count -= ret;
454 buf += ret;
455 total += ret;
456 }
457
458 return total;
459}
460
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100461/*
462 * Opens a socket with FD_CLOEXEC set
463 */
464int qemu_socket(int domain, int type, int protocol)
465{
466 int ret;
467
468#ifdef SOCK_CLOEXEC
469 ret = socket(domain, type | SOCK_CLOEXEC, protocol);
Andre Przywara3a03bfa2009-12-18 10:45:07 +0100470 if (ret != -1 || errno != EINVAL) {
471 return ret;
472 }
473#endif
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100474 ret = socket(domain, type, protocol);
475 if (ret >= 0) {
476 qemu_set_cloexec(ret);
477 }
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100478
479 return ret;
480}
481
482/*
483 * Accept a connection and set FD_CLOEXEC
484 */
485int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
486{
487 int ret;
488
489#ifdef CONFIG_ACCEPT4
490 ret = accept4(s, addr, addrlen, SOCK_CLOEXEC);
Kevin Wolf347ed552010-01-13 16:20:56 +0100491 if (ret != -1 || errno != ENOSYS) {
Andre Przywara3a03bfa2009-12-18 10:45:07 +0100492 return ret;
493 }
494#endif
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100495 ret = accept(s, addr, addrlen);
496 if (ret >= 0) {
497 qemu_set_cloexec(ret);
498 }
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100499
500 return ret;
501}
Paolo Bonzini993295f2011-09-17 16:27:59 +0200502
Xuzhou Cheng84c662d2022-10-28 12:57:27 +0800503ssize_t qemu_send_full(int s, const void *buf, size_t count)
504{
505 ssize_t ret = 0;
506 ssize_t total = 0;
507
508 while (count) {
509 ret = send(s, buf, count, 0);
510 if (ret < 0) {
511 if (errno == EINTR) {
512 continue;
513 }
514 break;
515 }
516
517 count -= ret;
518 buf += ret;
519 total += ret;
520 }
521
522 return total;
523}
524
Eduardo Habkost35c2c8d2015-10-30 17:36:08 -0200525void qemu_set_hw_version(const char *version)
Crístian Viana93bfef42012-05-30 00:35:51 -0300526{
Eduardo Habkost35c2c8d2015-10-30 17:36:08 -0200527 hw_version = version;
Crístian Viana93bfef42012-05-30 00:35:51 -0300528}
529
Eduardo Habkost35c2c8d2015-10-30 17:36:08 -0200530const char *qemu_hw_version(void)
Crístian Viana93bfef42012-05-30 00:35:51 -0300531{
Eduardo Habkost35c2c8d2015-10-30 17:36:08 -0200532 return hw_version;
Crístian Viana93bfef42012-05-30 00:35:51 -0300533}
Paul Moore0f669982012-08-03 14:39:21 -0400534
Marc-André Lureaud3bf8252013-02-25 23:31:11 +0100535#ifdef _WIN32
536static void socket_cleanup(void)
537{
538 WSACleanup();
539}
540#endif
541
542int socket_init(void)
543{
544#ifdef _WIN32
545 WSADATA Data;
546 int ret, err;
547
548 ret = WSAStartup(MAKEWORD(2, 2), &Data);
549 if (ret != 0) {
550 err = WSAGetLastError();
551 fprintf(stderr, "WSAStartup: %d\n", err);
552 return -1;
553 }
554 atexit(socket_cleanup);
555#endif
556 return 0;
557}
Paolo Bonzini9adea5f2013-04-21 12:01:06 +0200558
Stefan Hajnocziae2990c2013-10-08 11:58:31 +0200559
Paolo Bonzini9adea5f2013-04-21 12:01:06 +0200560#ifndef CONFIG_IOVEC
Paolo Bonzini9adea5f2013-04-21 12:01:06 +0200561static ssize_t
562readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write)
563{
564 unsigned i = 0;
565 ssize_t ret = 0;
Marc-André Lureauc1f79802022-10-06 15:36:52 +0400566 ssize_t off = 0;
Paolo Bonzini9adea5f2013-04-21 12:01:06 +0200567 while (i < iov_cnt) {
568 ssize_t r = do_write
Marc-André Lureauc1f79802022-10-06 15:36:52 +0400569 ? 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);
Paolo Bonzini9adea5f2013-04-21 12:01:06 +0200571 if (r > 0) {
572 ret += r;
Marc-André Lureauc1f79802022-10-06 15:36:52 +0400573 off += r;
574 if (off < iov[i].iov_len) {
575 continue;
576 }
Paolo Bonzini9adea5f2013-04-21 12:01:06 +0200577 } else if (!r) {
578 break;
579 } else if (errno == EINTR) {
580 continue;
581 } else {
582 /* else it is some "other" error,
583 * only return if there was no data processed. */
584 if (ret == 0) {
585 ret = -1;
586 }
587 break;
588 }
Marc-André Lureauc1f79802022-10-06 15:36:52 +0400589 off = 0;
Paolo Bonzini9adea5f2013-04-21 12:01:06 +0200590 i++;
591 }
592 return ret;
593}
594
595ssize_t
596readv(int fd, const struct iovec *iov, int iov_cnt)
597{
598 return readv_writev(fd, iov, iov_cnt, false);
599}
600
601ssize_t
602writev(int fd, const struct iovec *iov, int iov_cnt)
603{
604 return readv_writev(fd, iov, iov_cnt, true);
605}
606#endif
Marc-André Lureau282468c2022-04-20 17:26:04 +0400607
608/*
609 * Make sure data goes on disk, but if possible do not bother to
610 * write out the inode just for timestamp updates.
611 *
612 * Unfortunately even in 2009 many operating systems do not support
613 * fdatasync and have to fall back to fsync.
614 */
615int qemu_fdatasync(int fd)
616{
617#ifdef CONFIG_FDATASYNC
618 return fdatasync(fd);
619#else
620 return fsync(fd);
621#endif
622}