blob: 4829c07ff6edbb1508946ad789031f51b2cb6035 [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"
Paolo Bonzinif582af52010-02-02 20:33:11 +010025
26/* Needed early for CONFIG_BSD etc. */
Paolo Bonzinif582af52010-02-02 20:33:11 +010027
Juan Quinteladfe5fff2009-07-27 16:12:40 +020028#ifdef CONFIG_SOLARIS
ths605686c2007-01-17 23:31:19 +000029#include <sys/statvfs.h>
Andreas Färbere78815a2010-09-25 11:26:05 +000030/* See MySQL bug #7156 (http://bugs.mysql.com/bug.php?id=7156) for
31 discussion about Solaris header problems */
Marc-André Lureaud7df0b42019-01-17 15:43:53 +040032extern int madvise(char *, size_t, int);
ths605686c2007-01-17 23:31:19 +000033#endif
bellardea888122004-02-16 22:12:40 +000034
blueswir1511d2b12009-03-07 15:32:56 +000035#include "qemu-common.h"
Veronia Bahaaf348b6d2016-03-20 19:16:19 +020036#include "qemu/cutils.h"
Paolo Bonzini1de7afc2012-12-17 18:20:00 +010037#include "qemu/sockets.h"
Markus Armbrusterd49b6832015-03-17 18:29:20 +010038#include "qemu/error-report.h"
Paolo Bonzini83c90892012-12-17 18:19:49 +010039#include "monitor/monitor.h"
aliguori03ff3ca2008-09-15 15:51:35 +000040
Paul Moore0f669982012-08-03 14:39:21 -040041static bool fips_enabled = false;
42
Eduardo Habkostd4943522016-04-09 16:42:44 -030043static const char *hw_version = QEMU_HW_VERSION;
Crístian Viana93bfef42012-05-30 00:35:51 -030044
Paolo Bonzini128aa582011-09-21 12:36:48 +020045int socket_set_cork(int fd, int v)
46{
47#if defined(SOL_TCP) && defined(TCP_CORK)
Lei Li4bd1afb2013-03-06 22:29:16 +080048 return qemu_setsockopt(fd, SOL_TCP, TCP_CORK, &v, sizeof(v));
Paolo Bonzini128aa582011-09-21 12:36:48 +020049#else
50 return 0;
51#endif
52}
53
MORITA Kazutakabf1c8522013-02-22 12:39:50 +090054int socket_set_nodelay(int fd)
55{
56 int v = 1;
Lei Li4bd1afb2013-03-06 22:29:16 +080057 return qemu_setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &v, sizeof(v));
MORITA Kazutakabf1c8522013-02-22 12:39:50 +090058}
59
Andreas Färbere78815a2010-09-25 11:26:05 +000060int qemu_madvise(void *addr, size_t len, int advice)
61{
62 if (advice == QEMU_MADV_INVALID) {
63 errno = EINVAL;
64 return -1;
65 }
66#if defined(CONFIG_MADVISE)
67 return madvise(addr, len, advice);
68#elif defined(CONFIG_POSIX_MADVISE)
69 return posix_madvise(addr, len, advice);
70#else
71 errno = EINVAL;
72 return -1;
73#endif
74}
75
Emilio G. Cota5fa64b32017-07-15 02:28:47 -040076static int qemu_mprotect__osdep(void *addr, size_t size, int prot)
77{
78 g_assert(!((uintptr_t)addr & ~qemu_real_host_page_mask));
79 g_assert(!(size & ~qemu_real_host_page_mask));
80
81#ifdef _WIN32
82 DWORD old_protect;
83
84 if (!VirtualProtect(addr, size, prot, &old_protect)) {
Philippe Mathieu-Daudécf0c76c2020-02-28 11:07:24 +010085 g_autofree gchar *emsg = g_win32_error_message(GetLastError());
86 error_report("%s: VirtualProtect failed: %s", __func__, emsg);
Emilio G. Cota5fa64b32017-07-15 02:28:47 -040087 return -1;
88 }
89 return 0;
90#else
91 if (mprotect(addr, size, prot)) {
92 error_report("%s: mprotect failed: %s", __func__, strerror(errno));
93 return -1;
94 }
95 return 0;
96#endif
97}
98
99int 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 */
125static int qemu_dup_flags(int fd, int flags)
126{
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();
Fam Zhengf86428a2017-12-26 14:53:00 +0800247 do {
248 ret = fcntl(fd, fcntl_op_setlk, &fl);
249 } while (ret == -1 && errno == EINTR);
Fam Zheng13461fd2017-05-03 00:35:54 +0800250 return ret == -1 ? -errno : 0;
Fam Zheng13461fd2017-05-03 00:35:54 +0800251}
252
253int qemu_lock_fd(int fd, int64_t start, int64_t len, bool exclusive)
254{
255 return qemu_lock_fcntl(fd, start, len, exclusive ? F_WRLCK : F_RDLCK);
256}
257
258int qemu_unlock_fd(int fd, int64_t start, int64_t len)
259{
260 return qemu_lock_fcntl(fd, start, len, F_UNLCK);
261}
262
263int qemu_lock_fd_test(int fd, int64_t start, int64_t len, bool exclusive)
264{
Fam Zheng13461fd2017-05-03 00:35:54 +0800265 int ret;
266 struct flock fl = {
267 .l_whence = SEEK_SET,
268 .l_start = start,
269 .l_len = len,
270 .l_type = exclusive ? F_WRLCK : F_RDLCK,
271 };
Fam Zhengca749952017-08-11 19:44:46 +0800272 qemu_probe_lock_ops();
273 ret = fcntl(fd, fcntl_op_getlk, &fl);
Fam Zheng13461fd2017-05-03 00:35:54 +0800274 if (ret == -1) {
275 return -errno;
276 } else {
277 return fl.l_type == F_UNLCK ? 0 : -EAGAIN;
278 }
Fam Zheng13461fd2017-05-03 00:35:54 +0800279}
Corey Bryantadb696f2012-08-14 16:43:47 -0400280#endif
aliguori03ff3ca2008-09-15 15:51:35 +0000281
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100282/*
283 * Opens a file with FD_CLOEXEC set
284 */
285int qemu_open(const char *name, int flags, ...)
286{
287 int ret;
288 int mode = 0;
289
Corey Bryantadb696f2012-08-14 16:43:47 -0400290#ifndef _WIN32
291 const char *fdset_id_str;
292
293 /* Attempt dup of fd from fd set */
294 if (strstart(name, "/dev/fdset/", &fdset_id_str)) {
295 int64_t fdset_id;
296 int fd, dupfd;
297
298 fdset_id = qemu_parse_fdset(fdset_id_str);
299 if (fdset_id == -1) {
300 errno = EINVAL;
301 return -1;
302 }
303
304 fd = monitor_fdset_get_fd(fdset_id, flags);
Peter Xu47451462018-06-08 11:55:11 +0800305 if (fd < 0) {
306 errno = -fd;
Corey Bryantadb696f2012-08-14 16:43:47 -0400307 return -1;
308 }
309
310 dupfd = qemu_dup_flags(fd, flags);
311 if (dupfd == -1) {
312 return -1;
313 }
314
315 ret = monitor_fdset_dup_fd_add(fdset_id, dupfd);
316 if (ret == -1) {
317 close(dupfd);
318 errno = EINVAL;
319 return -1;
320 }
321
322 return dupfd;
323 }
324#endif
325
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100326 if (flags & O_CREAT) {
327 va_list ap;
328
329 va_start(ap, flags);
330 mode = va_arg(ap, int);
331 va_end(ap);
332 }
333
334#ifdef O_CLOEXEC
335 ret = open(name, flags | O_CLOEXEC, mode);
336#else
337 ret = open(name, flags, mode);
338 if (ret >= 0) {
339 qemu_set_cloexec(ret);
340 }
341#endif
342
Stefan Hajnoczia5813072013-08-22 11:29:03 +0200343#ifdef O_DIRECT
344 if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
345 error_report("file system may not support O_DIRECT");
346 errno = EINVAL; /* in case it was clobbered */
347 }
348#endif /* O_DIRECT */
349
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100350 return ret;
351}
352
Corey Bryant2e1e79d2012-08-14 16:43:46 -0400353int qemu_close(int fd)
354{
Corey Bryantadb696f2012-08-14 16:43:47 -0400355 int64_t fdset_id;
356
357 /* Close fd that was dup'd from an fdset */
358 fdset_id = monitor_fdset_dup_fd_find(fd);
359 if (fdset_id != -1) {
360 int ret;
361
362 ret = close(fd);
363 if (ret == 0) {
364 monitor_fdset_dup_fd_remove(fd);
365 }
366
367 return ret;
368 }
369
Corey Bryant2e1e79d2012-08-14 16:43:46 -0400370 return close(fd);
371}
372
Kirill A. Shutemov7b5f6992010-01-20 00:56:08 +0100373/*
Marc-André Lureauee132402019-11-08 18:09:56 +0400374 * Delete a file from the filesystem, unless the filename is /dev/fdset/...
375 *
376 * Returns: On success, zero is returned. On error, -1 is returned,
377 * and errno is set appropriately.
378 */
379int qemu_unlink(const char *name)
380{
381 if (g_str_has_prefix(name, "/dev/fdset/")) {
382 return 0;
383 }
384
385 return unlink(name);
386}
387
388/*
Kirill A. Shutemov7b5f6992010-01-20 00:56:08 +0100389 * A variant of write(2) which handles partial write.
390 *
391 * Return the number of bytes transferred.
392 * Set errno if fewer than `count' bytes are written.
Juan Quintela1298cb62010-03-04 10:00:39 +0100393 *
394 * This function don't work with non-blocking fd's.
395 * Any of the possibilities with non-bloking fd's is bad:
396 * - return a short write (then name is wrong)
397 * - busy wait adding (errno == EAGAIN) to the loop
Kirill A. Shutemov7b5f6992010-01-20 00:56:08 +0100398 */
399ssize_t qemu_write_full(int fd, const void *buf, size_t count)
400{
401 ssize_t ret = 0;
402 ssize_t total = 0;
403
404 while (count) {
405 ret = write(fd, buf, count);
406 if (ret < 0) {
407 if (errno == EINTR)
408 continue;
409 break;
410 }
411
412 count -= ret;
413 buf += ret;
414 total += ret;
415 }
416
417 return total;
418}
419
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100420/*
421 * Opens a socket with FD_CLOEXEC set
422 */
423int qemu_socket(int domain, int type, int protocol)
424{
425 int ret;
426
427#ifdef SOCK_CLOEXEC
428 ret = socket(domain, type | SOCK_CLOEXEC, protocol);
Andre Przywara3a03bfa2009-12-18 10:45:07 +0100429 if (ret != -1 || errno != EINVAL) {
430 return ret;
431 }
432#endif
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100433 ret = socket(domain, type, protocol);
434 if (ret >= 0) {
435 qemu_set_cloexec(ret);
436 }
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100437
438 return ret;
439}
440
441/*
442 * Accept a connection and set FD_CLOEXEC
443 */
444int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
445{
446 int ret;
447
448#ifdef CONFIG_ACCEPT4
449 ret = accept4(s, addr, addrlen, SOCK_CLOEXEC);
Kevin Wolf347ed552010-01-13 16:20:56 +0100450 if (ret != -1 || errno != ENOSYS) {
Andre Przywara3a03bfa2009-12-18 10:45:07 +0100451 return ret;
452 }
453#endif
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100454 ret = accept(s, addr, addrlen);
455 if (ret >= 0) {
456 qemu_set_cloexec(ret);
457 }
Kevin Wolf40ff6d72009-12-02 12:24:42 +0100458
459 return ret;
460}
Paolo Bonzini993295f2011-09-17 16:27:59 +0200461
Eduardo Habkost35c2c8d2015-10-30 17:36:08 -0200462void qemu_set_hw_version(const char *version)
Crístian Viana93bfef42012-05-30 00:35:51 -0300463{
Eduardo Habkost35c2c8d2015-10-30 17:36:08 -0200464 hw_version = version;
Crístian Viana93bfef42012-05-30 00:35:51 -0300465}
466
Eduardo Habkost35c2c8d2015-10-30 17:36:08 -0200467const char *qemu_hw_version(void)
Crístian Viana93bfef42012-05-30 00:35:51 -0300468{
Eduardo Habkost35c2c8d2015-10-30 17:36:08 -0200469 return hw_version;
Crístian Viana93bfef42012-05-30 00:35:51 -0300470}
Paul Moore0f669982012-08-03 14:39:21 -0400471
472void fips_set_state(bool requested)
473{
474#ifdef __linux__
475 if (requested) {
476 FILE *fds = fopen("/proc/sys/crypto/fips_enabled", "r");
477 if (fds != NULL) {
478 fips_enabled = (fgetc(fds) == '1');
479 fclose(fds);
480 }
481 }
482#else
483 fips_enabled = false;
484#endif /* __linux__ */
485
486#ifdef _FIPS_DEBUG
487 fprintf(stderr, "FIPS mode %s (requested %s)\n",
Paolo Bonzini7d374352018-12-13 23:37:37 +0100488 (fips_enabled ? "enabled" : "disabled"),
489 (requested ? "enabled" : "disabled"));
Paul Moore0f669982012-08-03 14:39:21 -0400490#endif
491}
492
493bool fips_get_state(void)
494{
495 return fips_enabled;
496}
Paolo Bonzini0100fbb2012-10-29 15:19:18 +0100497
Marc-André Lureaud3bf8252013-02-25 23:31:11 +0100498#ifdef _WIN32
499static void socket_cleanup(void)
500{
501 WSACleanup();
502}
503#endif
504
505int socket_init(void)
506{
507#ifdef _WIN32
508 WSADATA Data;
509 int ret, err;
510
511 ret = WSAStartup(MAKEWORD(2, 2), &Data);
512 if (ret != 0) {
513 err = WSAGetLastError();
514 fprintf(stderr, "WSAStartup: %d\n", err);
515 return -1;
516 }
517 atexit(socket_cleanup);
518#endif
519 return 0;
520}
Paolo Bonzini9adea5f2013-04-21 12:01:06 +0200521
Stefan Hajnocziae2990c2013-10-08 11:58:31 +0200522
Paolo Bonzini9adea5f2013-04-21 12:01:06 +0200523#ifndef CONFIG_IOVEC
524/* helper function for iov_send_recv() */
525static ssize_t
526readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write)
527{
528 unsigned i = 0;
529 ssize_t ret = 0;
530 while (i < iov_cnt) {
531 ssize_t r = do_write
532 ? write(fd, iov[i].iov_base, iov[i].iov_len)
533 : read(fd, iov[i].iov_base, iov[i].iov_len);
534 if (r > 0) {
535 ret += r;
536 } else if (!r) {
537 break;
538 } else if (errno == EINTR) {
539 continue;
540 } else {
541 /* else it is some "other" error,
542 * only return if there was no data processed. */
543 if (ret == 0) {
544 ret = -1;
545 }
546 break;
547 }
548 i++;
549 }
550 return ret;
551}
552
553ssize_t
554readv(int fd, const struct iovec *iov, int iov_cnt)
555{
556 return readv_writev(fd, iov, iov_cnt, false);
557}
558
559ssize_t
560writev(int fd, const struct iovec *iov, int iov_cnt)
561{
562 return readv_writev(fd, iov, iov_cnt, true);
563}
564#endif