1 /*
2 * OS includes and handling of OS dependencies
3 *
4 * This header exists to pull in some common system headers that
5 * most code in QEMU will want, and to fix up some possible issues with
6 * it (missing defines, Windows weirdness, and so on).
7 *
8 * To avoid getting into possible circular include dependencies, this
9 * file should not include any other QEMU headers, with the exceptions
10 * of config-host.h, config-target.h, qemu/compiler.h,
11 * sysemu/os-posix.h, sysemu/os-win32.h, glib-compat.h and
12 * qemu/typedefs.h, all of which are doing a similar job to this file
13 * and are under similar constraints.
14 *
15 * This header also contains prototypes for functions defined in
16 * os-*.c and util/oslib-*.c; those would probably be better split
17 * out into separate header files.
18 *
19 * In an ideal world this header would contain only:
20 * (1) things which everybody needs
21 * (2) things without which code would work on most platforms but
22 * fail to compile or misbehave on a minority of host OSes
23 *
24 * This work is licensed under the terms of the GNU GPL, version 2 or later.
25 * See the COPYING file in the top-level directory.
26 */
27 #ifndef QEMU_OSDEP_H
28 #define QEMU_OSDEP_H
29
30 #if !defined _FORTIFY_SOURCE && defined __OPTIMIZE__ && __OPTIMIZE__ && defined __linux__
31 # define _FORTIFY_SOURCE 2
32 #endif
33
34 #include "config-host.h"
35 #ifdef COMPILING_PER_TARGET
36 #include CONFIG_TARGET
37 #else
38 #include "exec/poison.h"
39 #endif
40
41 /*
42 * HOST_WORDS_BIGENDIAN was replaced with HOST_BIG_ENDIAN. Prevent it from
43 * creeping back in.
44 */
45 #pragma GCC poison HOST_WORDS_BIGENDIAN
46
47 /*
48 * TARGET_WORDS_BIGENDIAN was replaced with TARGET_BIG_ENDIAN. Prevent it from
49 * creeping back in.
50 */
51 #pragma GCC poison TARGET_WORDS_BIGENDIAN
52
53 #include "qemu/compiler.h"
54
55 /* Older versions of C++ don't get definitions of various macros from
56 * stdlib.h unless we define these macros before first inclusion of
57 * that system header.
58 */
59 #ifndef __STDC_CONSTANT_MACROS
60 #define __STDC_CONSTANT_MACROS
61 #endif
62 #ifndef __STDC_LIMIT_MACROS
63 #define __STDC_LIMIT_MACROS
64 #endif
65 #ifndef __STDC_FORMAT_MACROS
66 #define __STDC_FORMAT_MACROS
67 #endif
68
69 /* The following block of code temporarily renames the daemon() function so the
70 * compiler does not see the warning associated with it in stdlib.h on OSX
71 */
72 #ifdef __APPLE__
73 #define daemon qemu_fake_daemon_function
74 #include <stdlib.h>
75 #undef daemon
76 QEMU_EXTERN_C int daemon(int, int);
77 #endif
78
79 #ifdef _WIN32
80 /* as defined in sdkddkver.h */
81 #ifndef _WIN32_WINNT
82 #define _WIN32_WINNT 0x0602 /* Windows 8 API (should be >= the one from glib) */
83 #endif
84 /* reduces the number of implicitly included headers */
85 #ifndef WIN32_LEAN_AND_MEAN
86 #define WIN32_LEAN_AND_MEAN
87 #endif
88 #endif
89
90 /* enable C99/POSIX format strings (needs mingw32-runtime 3.15 or later) */
91 #ifdef __MINGW32__
92 #define __USE_MINGW_ANSI_STDIO 1
93 #endif
94
95 /*
96 * We need the FreeBSD "legacy" definitions. Rust needs the FreeBSD 11 system
97 * calls since it doesn't use libc at all, so we have to emulate that despite
98 * FreeBSD 11 being EOL'd.
99 */
100 #ifdef __FreeBSD__
101 #define _WANT_FREEBSD11_STAT
102 #define _WANT_FREEBSD11_STATFS
103 #define _WANT_FREEBSD11_DIRENT
104 #define _WANT_KERNEL_ERRNO
105 #define _WANT_SEMUN
106 #endif
107
108 #include <stdarg.h>
109 #include <stddef.h>
110 #include <stdbool.h>
111 #include <stdint.h>
112 #include <sys/types.h>
113 #include <stdlib.h>
114 #include <stdio.h>
115
116 #include <string.h>
117 #include <strings.h>
118 #include <inttypes.h>
119 #include <limits.h>
120 /* Put unistd.h before time.h as that triggers localtime_r/gmtime_r
121 * function availability on recentish Mingw-w64 platforms. */
122 #include <unistd.h>
123 #include <time.h>
124 #include <ctype.h>
125 #include <errno.h>
126 #include <fcntl.h>
127 #include <getopt.h>
128 #include <sys/stat.h>
129 #include <sys/time.h>
130 #include <assert.h>
131 /* setjmp must be declared before sysemu/os-win32.h
132 * because it is redefined there. */
133 #include <setjmp.h>
134 #include <signal.h>
135
136 #ifdef CONFIG_IOVEC
137 #include <sys/uio.h>
138 #endif
139
140 #if defined(__linux__) && defined(__sparc__)
141 /* The SPARC definition of QEMU_VMALLOC_ALIGN needs SHMLBA */
142 #include <sys/shm.h>
143 #endif
144
145 #ifndef _WIN32
146 #include <sys/wait.h>
147 #else
148 #define WIFEXITED(x) 1
149 #define WEXITSTATUS(x) (x)
150 #endif
151
152 #ifdef __APPLE__
153 #include <AvailabilityMacros.h>
154 #endif
155
156 /*
157 * This is somewhat like a system header; it must be outside any extern "C"
158 * block because it includes system headers itself, including glib.h,
159 * which will not compile if inside an extern "C" block.
160 */
161 #include "glib-compat.h"
162
163 #ifdef _WIN32
164 #include "sysemu/os-win32.h"
165 #endif
166
167 #ifdef CONFIG_POSIX
168 #include "sysemu/os-posix.h"
169 #endif
170
171 #ifdef __cplusplus
172 extern "C" {
173 #endif
174
175 #include "qemu/typedefs.h"
176
177 /**
178 * Mark a function that executes in coroutine context
179 *
180 * Functions that execute in coroutine context cannot be called directly from
181 * normal functions. In the future it would be nice to enable compiler or
182 * static checker support for catching such errors. This annotation might make
183 * it possible and in the meantime it serves as documentation.
184 *
185 * For example:
186 *
187 * static void coroutine_fn foo(void) {
188 * ....
189 * }
190 */
191 #ifdef __clang__
192 #define coroutine_fn QEMU_ANNOTATE("coroutine_fn")
193 #else
194 #define coroutine_fn
195 #endif
196
197 /**
198 * Mark a function that can suspend when executed in coroutine context,
199 * but can handle running in non-coroutine context too.
200 */
201 #ifdef __clang__
202 #define coroutine_mixed_fn QEMU_ANNOTATE("coroutine_mixed_fn")
203 #else
204 #define coroutine_mixed_fn
205 #endif
206
207 /**
208 * Mark a function that should not be called from a coroutine context.
209 * Usually there will be an analogous, coroutine_fn function that should
210 * be used instead.
211 *
212 * When the function is also marked as coroutine_mixed_fn, the function should
213 * only be called if the caller does not know whether it is in coroutine
214 * context.
215 *
216 * Functions that are only no_coroutine_fn, on the other hand, should not
217 * be called from within coroutines at all. This for example includes
218 * functions that block.
219 *
220 * In the future it would be nice to enable compiler or static checker
221 * support for catching such errors. This annotation is the first step
222 * towards this, and in the meantime it serves as documentation.
223 *
224 * For example:
225 *
226 * static void no_coroutine_fn foo(void) {
227 * ....
228 * }
229 */
230 #ifdef __clang__
231 #define no_coroutine_fn QEMU_ANNOTATE("no_coroutine_fn")
232 #else
233 #define no_coroutine_fn
234 #endif
235
236
237 /*
238 * For mingw, as of v6.0.0, the function implementing the assert macro is
239 * not marked as noreturn, so the compiler cannot delete code following an
240 * assert(false) as unused. We rely on this within the code base to delete
241 * code that is unreachable when features are disabled.
242 * All supported versions of Glib's g_assert() satisfy this requirement.
243 */
244 #ifdef __MINGW32__
245 #undef assert
246 #define assert(x) g_assert(x)
247 #endif
248
249 /**
250 * qemu_build_not_reached()
251 *
252 * The compiler, during optimization, is expected to prove that a call
253 * to this function cannot be reached and remove it. If the compiler
254 * supports QEMU_ERROR, this will be reported at compile time; otherwise
255 * this will be reported at link time due to the missing symbol.
256 */
257 G_NORETURN
258 void QEMU_ERROR("code path is reachable")
259 qemu_build_not_reached_always(void);
260 #if defined(__OPTIMIZE__) && !defined(__NO_INLINE__)
261 #define qemu_build_not_reached() qemu_build_not_reached_always()
262 #else
263 #define qemu_build_not_reached() g_assert_not_reached()
264 #endif
265
266 /**
267 * qemu_build_assert()
268 *
269 * The compiler, during optimization, is expected to prove that the
270 * assertion is true.
271 */
272 #define qemu_build_assert(test) while (!(test)) qemu_build_not_reached()
273
274 /*
275 * According to waitpid man page:
276 * WCOREDUMP
277 * This macro is not specified in POSIX.1-2001 and is not
278 * available on some UNIX implementations (e.g., AIX, SunOS).
279 * Therefore, enclose its use inside #ifdef WCOREDUMP ... #endif.
280 */
281 #ifndef WCOREDUMP
282 #define WCOREDUMP(status) 0
283 #endif
284 /*
285 * We have a lot of unaudited code that may fail in strange ways, or
286 * even be a security risk during migration, if you disable assertions
287 * at compile-time. You may comment out these safety checks if you
288 * absolutely want to disable assertion overhead, but it is not
289 * supported upstream so the risk is all yours. Meanwhile, please
290 * submit patches to remove any side-effects inside an assertion, or
291 * fixing error handling that should use Error instead of assert.
292 */
293 #ifdef NDEBUG
294 #error building with NDEBUG is not supported
295 #endif
296 #ifdef G_DISABLE_ASSERT
297 #error building with G_DISABLE_ASSERT is not supported
298 #endif
299
300 #ifndef O_LARGEFILE
301 #define O_LARGEFILE 0
302 #endif
303 #ifndef O_BINARY
304 #define O_BINARY 0
305 #endif
306 #ifndef MAP_ANONYMOUS
307 #define MAP_ANONYMOUS MAP_ANON
308 #endif
309 #ifndef MAP_NORESERVE
310 #define MAP_NORESERVE 0
311 #endif
312 #ifndef ENOMEDIUM
313 #define ENOMEDIUM ENODEV
314 #endif
315 #if !defined(ENOTSUP)
316 #define ENOTSUP 4096
317 #endif
318 #if !defined(ECANCELED)
319 #define ECANCELED 4097
320 #endif
321 #if !defined(EMEDIUMTYPE)
322 #define EMEDIUMTYPE 4098
323 #endif
324 #if !defined(ESHUTDOWN)
325 #define ESHUTDOWN 4099
326 #endif
327
328 #define RETRY_ON_EINTR(expr) \
329 (__extension__ \
330 ({ typeof(expr) __result; \
331 do { \
332 __result = (expr); \
333 } while (__result == -1 && errno == EINTR); \
334 __result; }))
335
336 /* time_t may be either 32 or 64 bits depending on the host OS, and
337 * can be either signed or unsigned, so we can't just hardcode a
338 * specific maximum value. This is not a C preprocessor constant,
339 * so you can't use TIME_MAX in an #ifdef, but for our purposes
340 * this isn't a problem.
341 */
342
343 /* The macros TYPE_SIGNED, TYPE_WIDTH, and TYPE_MAXIMUM are from
344 * Gnulib, and are under the LGPL v2.1 or (at your option) any
345 * later version.
346 */
347
348 /* True if the real type T is signed. */
349 #define TYPE_SIGNED(t) (!((t)0 < (t)-1))
350
351 /* The width in bits of the integer type or expression T.
352 * Padding bits are not supported.
353 */
354 #define TYPE_WIDTH(t) (sizeof(t) * CHAR_BIT)
355
356 /* The maximum and minimum values for the integer type T. */
357 #define TYPE_MAXIMUM(t) \
358 ((t) (!TYPE_SIGNED(t) \
359 ? (t)-1 \
360 : ((((t)1 << (TYPE_WIDTH(t) - 2)) - 1) * 2 + 1)))
361
362 #ifndef TIME_MAX
363 #define TIME_MAX TYPE_MAXIMUM(time_t)
364 #endif
365
366 /* Mac OSX has a <stdint.h> bug that incorrectly defines SIZE_MAX with
367 * the wrong type. Our replacement isn't usable in preprocessor
368 * expressions, but it is sufficient for our needs. */
369 #ifdef HAVE_BROKEN_SIZE_MAX
370 #undef SIZE_MAX
371 #define SIZE_MAX ((size_t)-1)
372 #endif
373
374 /*
375 * Two variations of MIN/MAX macros. The first is for runtime use, and
376 * evaluates arguments only once (so it is safe even with side
377 * effects), but will not work in constant contexts (such as array
378 * size declarations) because of the '{}'. The second is for constant
379 * expression use, where evaluating arguments twice is safe because
380 * the result is going to be constant anyway, but will not work in a
381 * runtime context because of a void expression where a value is
382 * expected. Thus, both gcc and clang will fail to compile if you use
383 * the wrong macro (even if the error may seem a bit cryptic).
384 *
385 * Note that neither form is usable as an #if condition; if you truly
386 * need to write conditional code that depends on a minimum or maximum
387 * determined by the pre-processor instead of the compiler, you'll
388 * have to open-code it. Sadly, Coverity is severely confused by the
389 * constant variants, so we have to dumb things down there.
390 *
391 * Preprocessor sorcery ahead: use different identifiers for the local
392 * variables in each expansion, so we can nest macro calls without
393 * shadowing variables.
394 */
395 #define MIN_INTERNAL(a, b, _a, _b) \
396 ({ \
397 typeof(1 ? (a) : (b)) _a = (a), _b = (b); \
398 _a < _b ? _a : _b; \
399 })
400 #undef MIN
401 #define MIN(a, b) \
402 MIN_INTERNAL((a), (b), MAKE_IDENTFIER(_a), MAKE_IDENTFIER(_b))
403
404 #define MAX_INTERNAL(a, b, _a, _b) \
405 ({ \
406 typeof(1 ? (a) : (b)) _a = (a), _b = (b); \
407 _a > _b ? _a : _b; \
408 })
409 #undef MAX
410 #define MAX(a, b) \
411 MAX_INTERNAL((a), (b), MAKE_IDENTFIER(_a), MAKE_IDENTFIER(_b))
412
413 #ifdef __COVERITY__
414 # define MIN_CONST(a, b) ((a) < (b) ? (a) : (b))
415 # define MAX_CONST(a, b) ((a) > (b) ? (a) : (b))
416 #else
417 # define MIN_CONST(a, b) \
418 __builtin_choose_expr( \
419 __builtin_constant_p(a) && __builtin_constant_p(b), \
420 (a) < (b) ? (a) : (b), \
421 ((void)0))
422 # define MAX_CONST(a, b) \
423 __builtin_choose_expr( \
424 __builtin_constant_p(a) && __builtin_constant_p(b), \
425 (a) > (b) ? (a) : (b), \
426 ((void)0))
427 #endif
428
429 /*
430 * Minimum function that returns zero only if both values are zero.
431 * Intended for use with unsigned values only.
432 *
433 * Preprocessor sorcery ahead: use different identifiers for the local
434 * variables in each expansion, so we can nest macro calls without
435 * shadowing variables.
436 */
437 #define MIN_NON_ZERO_INTERNAL(a, b, _a, _b) \
438 ({ \
439 typeof(1 ? (a) : (b)) _a = (a), _b = (b); \
440 _a == 0 ? _b : (_b == 0 || _b > _a) ? _a : _b; \
441 })
442 #define MIN_NON_ZERO(a, b) \
443 MIN_NON_ZERO_INTERNAL((a), (b), MAKE_IDENTFIER(_a), MAKE_IDENTFIER(_b))
444
445 /*
446 * Round number down to multiple. Safe when m is not a power of 2 (see
447 * ROUND_DOWN for a faster version when a power of 2 is guaranteed).
448 */
449 #define QEMU_ALIGN_DOWN(n, m) ((n) / (m) * (m))
450
451 /*
452 * Round number up to multiple. Safe when m is not a power of 2 (see
453 * ROUND_UP for a faster version when a power of 2 is guaranteed).
454 */
455 #define QEMU_ALIGN_UP(n, m) QEMU_ALIGN_DOWN((n) + (m) - 1, (m))
456
457 /* Check if n is a multiple of m */
458 #define QEMU_IS_ALIGNED(n, m) (((n) % (m)) == 0)
459
460 /* n-byte align pointer down */
461 #define QEMU_ALIGN_PTR_DOWN(p, n) \
462 ((typeof(p))QEMU_ALIGN_DOWN((uintptr_t)(p), (n)))
463
464 /* n-byte align pointer up */
465 #define QEMU_ALIGN_PTR_UP(p, n) \
466 ((typeof(p))QEMU_ALIGN_UP((uintptr_t)(p), (n)))
467
468 /* Check if pointer p is n-bytes aligned */
469 #define QEMU_PTR_IS_ALIGNED(p, n) QEMU_IS_ALIGNED((uintptr_t)(p), (n))
470
471 /*
472 * Round number down to multiple. Requires that d be a power of 2 (see
473 * QEMU_ALIGN_UP for a safer but slower version on arbitrary
474 * numbers); works even if d is a smaller type than n.
475 */
476 #ifndef ROUND_DOWN
477 #define ROUND_DOWN(n, d) ((n) & -(0 ? (n) : (d)))
478 #endif
479
480 /*
481 * Round number up to multiple. Requires that d be a power of 2 (see
482 * QEMU_ALIGN_UP for a safer but slower version on arbitrary
483 * numbers); works even if d is a smaller type than n.
484 */
485 #ifndef ROUND_UP
486 #define ROUND_UP(n, d) ROUND_DOWN((n) + (d) - 1, (d))
487 #endif
488
489 #ifndef DIV_ROUND_UP
490 #define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
491 #endif
492
493 /*
494 * &(x)[0] is always a pointer - if it's same type as x then the argument is a
495 * pointer, not an array.
496 */
497 #define QEMU_IS_ARRAY(x) (!__builtin_types_compatible_p(typeof(x), \
498 typeof(&(x)[0])))
499 #ifndef ARRAY_SIZE
500 #define ARRAY_SIZE(x) ((sizeof(x) / sizeof((x)[0])) + \
501 QEMU_BUILD_BUG_ON_ZERO(!QEMU_IS_ARRAY(x)))
502 #endif
503
504 int qemu_daemon(int nochdir, int noclose);
505 void *qemu_anon_ram_alloc(size_t size, uint64_t *align, bool shared,
506 bool noreserve);
507 void qemu_anon_ram_free(void *ptr, size_t size);
508
509 #ifdef _WIN32
510 #define HAVE_CHARDEV_SERIAL 1
511 #define HAVE_CHARDEV_PARALLEL 1
512 #else
513 #if defined(__linux__) || defined(__sun__) || defined(__FreeBSD__) \
514 || defined(__NetBSD__) || defined(__OpenBSD__) || defined(__DragonFly__) \
515 || defined(__GLIBC__) || defined(__APPLE__)
516 #define HAVE_CHARDEV_SERIAL 1
517 #endif
518 #if defined(__linux__) || defined(__FreeBSD__) \
519 || defined(__FreeBSD_kernel__) || defined(__DragonFly__)
520 #define HAVE_CHARDEV_PARALLEL 1
521 #endif
522 #endif
523
524 #if defined(__HAIKU__)
525 #define SIGIO SIGPOLL
526 #endif
527
528 #ifdef HAVE_MADVISE_WITHOUT_PROTOTYPE
529 /*
530 * See MySQL bug #7156 (http://bugs.mysql.com/bug.php?id=7156) for discussion
531 * about Solaris missing the madvise() prototype.
532 */
533 int madvise(char *, size_t, int);
534 #endif
535
536 #if defined(CONFIG_LINUX)
537 #ifndef BUS_MCEERR_AR
538 #define BUS_MCEERR_AR 4
539 #endif
540 #ifndef BUS_MCEERR_AO
541 #define BUS_MCEERR_AO 5
542 #endif
543 #endif
544
545 #if defined(__linux__) && \
546 (defined(__x86_64__) || defined(__arm__) || defined(__aarch64__) \
547 || defined(__powerpc64__))
548 /* Use 2 MiB alignment so transparent hugepages can be used by KVM.
549 Valgrind does not support alignments larger than 1 MiB,
550 therefore we need special code which handles running on Valgrind. */
551 # define QEMU_VMALLOC_ALIGN (512 * 4096)
552 #elif defined(__linux__) && defined(__s390x__)
553 /* Use 1 MiB (segment size) alignment so gmap can be used by KVM. */
554 # define QEMU_VMALLOC_ALIGN (256 * 4096)
555 #elif defined(__linux__) && defined(__sparc__)
556 # define QEMU_VMALLOC_ALIGN MAX(qemu_real_host_page_size(), SHMLBA)
557 #elif defined(__linux__) && defined(__loongarch__)
558 /*
559 * For transparent hugepage optimization, it has better be huge page
560 * aligned. LoongArch host system supports two kinds of pagesize: 4K
561 * and 16K, here calculate huge page size from host page size
562 */
563 # define QEMU_VMALLOC_ALIGN (qemu_real_host_page_size() * \
564 qemu_real_host_page_size() / sizeof(long))
565 #else
566 # define QEMU_VMALLOC_ALIGN qemu_real_host_page_size()
567 #endif
568
569 #ifdef CONFIG_POSIX
570 struct qemu_signalfd_siginfo {
571 uint32_t ssi_signo; /* Signal number */
572 int32_t ssi_errno; /* Error number (unused) */
573 int32_t ssi_code; /* Signal code */
574 uint32_t ssi_pid; /* PID of sender */
575 uint32_t ssi_uid; /* Real UID of sender */
576 int32_t ssi_fd; /* File descriptor (SIGIO) */
577 uint32_t ssi_tid; /* Kernel timer ID (POSIX timers) */
578 uint32_t ssi_band; /* Band event (SIGIO) */
579 uint32_t ssi_overrun; /* POSIX timer overrun count */
580 uint32_t ssi_trapno; /* Trap number that caused signal */
581 int32_t ssi_status; /* Exit status or signal (SIGCHLD) */
582 int32_t ssi_int; /* Integer sent by sigqueue(2) */
583 uint64_t ssi_ptr; /* Pointer sent by sigqueue(2) */
584 uint64_t ssi_utime; /* User CPU time consumed (SIGCHLD) */
585 uint64_t ssi_stime; /* System CPU time consumed (SIGCHLD) */
586 uint64_t ssi_addr; /* Address that generated signal
587 (for hardware-generated signals) */
588 uint8_t pad[48]; /* Pad size to 128 bytes (allow for
589 additional fields in the future) */
590 };
591
592 int qemu_signalfd(const sigset_t *mask);
593 void sigaction_invoke(struct sigaction *action,
594 struct qemu_signalfd_siginfo *info);
595 #endif
596
597 /*
598 * Don't introduce new usage of this function, prefer the following
599 * qemu_open/qemu_create that take an "Error **errp"
600 */
601 int qemu_open_old(const char *name, int flags, ...);
602 int qemu_open(const char *name, int flags, Error **errp);
603 int qemu_create(const char *name, int flags, mode_t mode, Error **errp);
604 int qemu_close(int fd);
605 int qemu_unlink(const char *name);
606 #ifndef _WIN32
607 int qemu_dup_flags(int fd, int flags);
608 int qemu_dup(int fd);
609 int qemu_lock_fd(int fd, int64_t start, int64_t len, bool exclusive);
610 int qemu_unlock_fd(int fd, int64_t start, int64_t len);
611 int qemu_lock_fd_test(int fd, int64_t start, int64_t len, bool exclusive);
612 bool qemu_has_ofd_lock(void);
613 #endif
614
615 #if defined(__HAIKU__) && defined(__i386__)
616 #define FMT_pid "%ld"
617 #elif defined(WIN64)
618 #define FMT_pid "%" PRId64
619 #else
620 #define FMT_pid "%d"
621 #endif
622
623 bool qemu_write_pidfile(const char *pidfile, Error **errp);
624
625 int qemu_get_thread_id(void);
626
627 #ifndef CONFIG_IOVEC
628 struct iovec {
629 void *iov_base;
630 size_t iov_len;
631 };
632 /*
633 * Use the same value as Linux for now.
634 */
635 #define IOV_MAX 1024
636
637 ssize_t readv(int fd, const struct iovec *iov, int iov_cnt);
638 ssize_t writev(int fd, const struct iovec *iov, int iov_cnt);
639 #endif
640
641 #ifdef _WIN32
qemu_timersub(const struct timeval * val1,const struct timeval * val2,struct timeval * res)642 static inline void qemu_timersub(const struct timeval *val1,
643 const struct timeval *val2,
644 struct timeval *res)
645 {
646 res->tv_sec = val1->tv_sec - val2->tv_sec;
647 if (val1->tv_usec < val2->tv_usec) {
648 res->tv_sec--;
649 res->tv_usec = val1->tv_usec - val2->tv_usec + 1000 * 1000;
650 } else {
651 res->tv_usec = val1->tv_usec - val2->tv_usec;
652 }
653 }
654 #else
655 #define qemu_timersub timersub
656 #endif
657
658 ssize_t qemu_write_full(int fd, const void *buf, size_t count)
659 G_GNUC_WARN_UNUSED_RESULT;
660
661 void qemu_set_cloexec(int fd);
662
663 /* Return a dynamically allocated directory path that is appropriate for storing
664 * local state.
665 *
666 * The caller is responsible for releasing the value returned with g_free()
667 * after use.
668 */
669 char *qemu_get_local_state_dir(void);
670
671 /**
672 * qemu_getauxval:
673 * @type: the auxiliary vector key to lookup
674 *
675 * Search the auxiliary vector for @type, returning the value
676 * or 0 if @type is not present.
677 */
678 unsigned long qemu_getauxval(unsigned long type);
679
680 void qemu_set_tty_echo(int fd, bool echo);
681
682 typedef struct ThreadContext ThreadContext;
683
684 /**
685 * qemu_prealloc_mem:
686 * @fd: the fd mapped into the area, -1 for anonymous memory
687 * @area: start address of the are to preallocate
688 * @sz: the size of the area to preallocate
689 * @max_threads: maximum number of threads to use
690 * @tc: prealloc context threads pointer, NULL if not in use
691 * @async: request asynchronous preallocation, requires @tc
692 * @errp: returns an error if this function fails
693 *
694 * Preallocate memory (populate/prefault page tables writable) for the virtual
695 * memory area starting at @area with the size of @sz. After a successful call,
696 * each page in the area was faulted in writable at least once, for example,
697 * after allocating file blocks for mapped files.
698 *
699 * When setting @async, allocation might be performed asynchronously.
700 * qemu_finish_async_prealloc_mem() must be called to finish any asynchronous
701 * preallocation.
702 *
703 * Return: true on success, else false setting @errp with error.
704 */
705 bool qemu_prealloc_mem(int fd, char *area, size_t sz, int max_threads,
706 ThreadContext *tc, bool async, Error **errp);
707
708 /**
709 * qemu_finish_async_prealloc_mem:
710 * @errp: returns an error if this function fails
711 *
712 * Finish all outstanding asynchronous memory preallocation.
713 *
714 * Return: true on success, else false setting @errp with error.
715 */
716 bool qemu_finish_async_prealloc_mem(Error **errp);
717
718 /**
719 * qemu_get_pid_name:
720 * @pid: pid of a process
721 *
722 * For given @pid fetch its name. Caller is responsible for
723 * freeing the string when no longer needed.
724 * Returns allocated string on success, NULL on failure.
725 */
726 char *qemu_get_pid_name(pid_t pid);
727
728 /* Using intptr_t ensures that qemu_*_page_mask is sign-extended even
729 * when intptr_t is 32-bit and we are aligning a long long.
730 */
qemu_real_host_page_size(void)731 static inline uintptr_t qemu_real_host_page_size(void)
732 {
733 return getpagesize();
734 }
735
qemu_real_host_page_mask(void)736 static inline intptr_t qemu_real_host_page_mask(void)
737 {
738 return -(intptr_t)qemu_real_host_page_size();
739 }
740
741 /*
742 * After using getopt or getopt_long, if you need to parse another set
743 * of options, then you must reset optind. Unfortunately the way to
744 * do this varies between implementations of getopt.
745 */
qemu_reset_optind(void)746 static inline void qemu_reset_optind(void)
747 {
748 #ifdef HAVE_OPTRESET
749 optind = 1;
750 optreset = 1;
751 #else
752 optind = 0;
753 #endif
754 }
755
756 int qemu_fdatasync(int fd);
757
758 /**
759 * Sync changes made to the memory mapped file back to the backing
760 * storage. For POSIX compliant systems this will fallback
761 * to regular msync call. Otherwise it will trigger whole file sync
762 * (including the metadata case there is no support to skip that otherwise)
763 *
764 * @addr - start of the memory area to be synced
765 * @length - length of the are to be synced
766 * @fd - file descriptor for the file to be synced
767 * (mandatory only for POSIX non-compliant systems)
768 */
769 int qemu_msync(void *addr, size_t length, int fd);
770
771 /**
772 * qemu_get_host_physmem:
773 *
774 * Operating system agnostic way of querying host memory.
775 *
776 * Returns amount of physical memory on the system. This is purely
777 * advisery and may return 0 if we can't work it out. At the other
778 * end we saturate to SIZE_MAX if you are lucky enough to have that
779 * much memory.
780 */
781 size_t qemu_get_host_physmem(void);
782
783 /*
784 * Toggle write/execute on the pages marked MAP_JIT
785 * for the current thread.
786 */
787 #if defined(MAC_OS_VERSION_11_0) && \
788 MAC_OS_X_VERSION_MIN_REQUIRED >= MAC_OS_VERSION_11_0
qemu_thread_jit_execute(void)789 static inline void qemu_thread_jit_execute(void)
790 {
791 pthread_jit_write_protect_np(true);
792 }
793
qemu_thread_jit_write(void)794 static inline void qemu_thread_jit_write(void)
795 {
796 pthread_jit_write_protect_np(false);
797 }
798 #else
qemu_thread_jit_write(void)799 static inline void qemu_thread_jit_write(void) {}
qemu_thread_jit_execute(void)800 static inline void qemu_thread_jit_execute(void) {}
801 #endif
802
803 /**
804 * Platforms which do not support system() return ENOSYS
805 */
806 #ifndef HAVE_SYSTEM_FUNCTION
807 #define system platform_does_not_support_system
platform_does_not_support_system(const char * command)808 static inline int platform_does_not_support_system(const char *command)
809 {
810 errno = ENOSYS;
811 return -1;
812 }
813 #endif /* !HAVE_SYSTEM_FUNCTION */
814
815 #ifdef __cplusplus
816 }
817 #endif
818
819 #endif
820