xref: /openbmc/qemu/util/osdep.c (revision 489653b5)
1 /*
2  * QEMU low level functions
3  *
4  * Copyright (c) 2003 Fabrice Bellard
5  *
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:
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  */
24 #include <stdlib.h>
25 #include <stdio.h>
26 #include <stdarg.h>
27 #include <stdbool.h>
28 #include <string.h>
29 #include <errno.h>
30 #include <unistd.h>
31 #include <fcntl.h>
32 
33 /* Needed early for CONFIG_BSD etc. */
34 #include "config-host.h"
35 
36 #if defined(CONFIG_MADVISE) || defined(CONFIG_POSIX_MADVISE)
37 #include <sys/mman.h>
38 #endif
39 
40 #ifdef CONFIG_SOLARIS
41 #include <sys/types.h>
42 #include <sys/statvfs.h>
43 /* See MySQL bug #7156 (http://bugs.mysql.com/bug.php?id=7156) for
44    discussion about Solaris header problems */
45 extern int madvise(caddr_t, size_t, int);
46 #endif
47 
48 #include "qemu-common.h"
49 #include "qemu/sockets.h"
50 #include "monitor/monitor.h"
51 
52 static bool fips_enabled = false;
53 
54 static const char *qemu_version = QEMU_VERSION;
55 
56 int socket_set_cork(int fd, int v)
57 {
58 #if defined(SOL_TCP) && defined(TCP_CORK)
59     return qemu_setsockopt(fd, SOL_TCP, TCP_CORK, &v, sizeof(v));
60 #else
61     return 0;
62 #endif
63 }
64 
65 int socket_set_nodelay(int fd)
66 {
67     int v = 1;
68     return qemu_setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &v, sizeof(v));
69 }
70 
71 int qemu_madvise(void *addr, size_t len, int advice)
72 {
73     if (advice == QEMU_MADV_INVALID) {
74         errno = EINVAL;
75         return -1;
76     }
77 #if defined(CONFIG_MADVISE)
78     return madvise(addr, len, advice);
79 #elif defined(CONFIG_POSIX_MADVISE)
80     return posix_madvise(addr, len, advice);
81 #else
82     errno = EINVAL;
83     return -1;
84 #endif
85 }
86 
87 #ifndef _WIN32
88 /*
89  * Dups an fd and sets the flags
90  */
91 static int qemu_dup_flags(int fd, int flags)
92 {
93     int ret;
94     int serrno;
95     int dup_flags;
96 
97 #ifdef F_DUPFD_CLOEXEC
98     ret = fcntl(fd, F_DUPFD_CLOEXEC, 0);
99 #else
100     ret = dup(fd);
101     if (ret != -1) {
102         qemu_set_cloexec(ret);
103     }
104 #endif
105     if (ret == -1) {
106         goto fail;
107     }
108 
109     dup_flags = fcntl(ret, F_GETFL);
110     if (dup_flags == -1) {
111         goto fail;
112     }
113 
114     if ((flags & O_SYNC) != (dup_flags & O_SYNC)) {
115         errno = EINVAL;
116         goto fail;
117     }
118 
119     /* Set/unset flags that we can with fcntl */
120     if (fcntl(ret, F_SETFL, flags) == -1) {
121         goto fail;
122     }
123 
124     /* Truncate the file in the cases that open() would truncate it */
125     if (flags & O_TRUNC ||
126             ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))) {
127         if (ftruncate(ret, 0) == -1) {
128             goto fail;
129         }
130     }
131 
132     return ret;
133 
134 fail:
135     serrno = errno;
136     if (ret != -1) {
137         close(ret);
138     }
139     errno = serrno;
140     return -1;
141 }
142 
143 static int qemu_parse_fdset(const char *param)
144 {
145     return qemu_parse_fd(param);
146 }
147 #endif
148 
149 /*
150  * Opens a file with FD_CLOEXEC set
151  */
152 int qemu_open(const char *name, int flags, ...)
153 {
154     int ret;
155     int mode = 0;
156 
157 #ifndef _WIN32
158     const char *fdset_id_str;
159 
160     /* Attempt dup of fd from fd set */
161     if (strstart(name, "/dev/fdset/", &fdset_id_str)) {
162         int64_t fdset_id;
163         int fd, dupfd;
164 
165         fdset_id = qemu_parse_fdset(fdset_id_str);
166         if (fdset_id == -1) {
167             errno = EINVAL;
168             return -1;
169         }
170 
171         fd = monitor_fdset_get_fd(fdset_id, flags);
172         if (fd == -1) {
173             return -1;
174         }
175 
176         dupfd = qemu_dup_flags(fd, flags);
177         if (dupfd == -1) {
178             return -1;
179         }
180 
181         ret = monitor_fdset_dup_fd_add(fdset_id, dupfd);
182         if (ret == -1) {
183             close(dupfd);
184             errno = EINVAL;
185             return -1;
186         }
187 
188         return dupfd;
189     }
190 #endif
191 
192     if (flags & O_CREAT) {
193         va_list ap;
194 
195         va_start(ap, flags);
196         mode = va_arg(ap, int);
197         va_end(ap);
198     }
199 
200 #ifdef O_CLOEXEC
201     ret = open(name, flags | O_CLOEXEC, mode);
202 #else
203     ret = open(name, flags, mode);
204     if (ret >= 0) {
205         qemu_set_cloexec(ret);
206     }
207 #endif
208 
209 #ifdef O_DIRECT
210     if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
211         error_report("file system may not support O_DIRECT");
212         errno = EINVAL; /* in case it was clobbered */
213     }
214 #endif /* O_DIRECT */
215 
216     return ret;
217 }
218 
219 int qemu_close(int fd)
220 {
221     int64_t fdset_id;
222 
223     /* Close fd that was dup'd from an fdset */
224     fdset_id = monitor_fdset_dup_fd_find(fd);
225     if (fdset_id != -1) {
226         int ret;
227 
228         ret = close(fd);
229         if (ret == 0) {
230             monitor_fdset_dup_fd_remove(fd);
231         }
232 
233         return ret;
234     }
235 
236     return close(fd);
237 }
238 
239 /*
240  * A variant of write(2) which handles partial write.
241  *
242  * Return the number of bytes transferred.
243  * Set errno if fewer than `count' bytes are written.
244  *
245  * This function don't work with non-blocking fd's.
246  * Any of the possibilities with non-bloking fd's is bad:
247  *   - return a short write (then name is wrong)
248  *   - busy wait adding (errno == EAGAIN) to the loop
249  */
250 ssize_t qemu_write_full(int fd, const void *buf, size_t count)
251 {
252     ssize_t ret = 0;
253     ssize_t total = 0;
254 
255     while (count) {
256         ret = write(fd, buf, count);
257         if (ret < 0) {
258             if (errno == EINTR)
259                 continue;
260             break;
261         }
262 
263         count -= ret;
264         buf += ret;
265         total += ret;
266     }
267 
268     return total;
269 }
270 
271 /*
272  * Opens a socket with FD_CLOEXEC set
273  */
274 int qemu_socket(int domain, int type, int protocol)
275 {
276     int ret;
277 
278 #ifdef SOCK_CLOEXEC
279     ret = socket(domain, type | SOCK_CLOEXEC, protocol);
280     if (ret != -1 || errno != EINVAL) {
281         return ret;
282     }
283 #endif
284     ret = socket(domain, type, protocol);
285     if (ret >= 0) {
286         qemu_set_cloexec(ret);
287     }
288 
289     return ret;
290 }
291 
292 /*
293  * Accept a connection and set FD_CLOEXEC
294  */
295 int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
296 {
297     int ret;
298 
299 #ifdef CONFIG_ACCEPT4
300     ret = accept4(s, addr, addrlen, SOCK_CLOEXEC);
301     if (ret != -1 || errno != ENOSYS) {
302         return ret;
303     }
304 #endif
305     ret = accept(s, addr, addrlen);
306     if (ret >= 0) {
307         qemu_set_cloexec(ret);
308     }
309 
310     return ret;
311 }
312 
313 void qemu_set_version(const char *version)
314 {
315     qemu_version = version;
316 }
317 
318 const char *qemu_get_version(void)
319 {
320     return qemu_version;
321 }
322 
323 void fips_set_state(bool requested)
324 {
325 #ifdef __linux__
326     if (requested) {
327         FILE *fds = fopen("/proc/sys/crypto/fips_enabled", "r");
328         if (fds != NULL) {
329             fips_enabled = (fgetc(fds) == '1');
330             fclose(fds);
331         }
332     }
333 #else
334     fips_enabled = false;
335 #endif /* __linux__ */
336 
337 #ifdef _FIPS_DEBUG
338     fprintf(stderr, "FIPS mode %s (requested %s)\n",
339 	    (fips_enabled ? "enabled" : "disabled"),
340 	    (requested ? "enabled" : "disabled"));
341 #endif
342 }
343 
344 bool fips_get_state(void)
345 {
346     return fips_enabled;
347 }
348 
349 #ifdef _WIN32
350 static void socket_cleanup(void)
351 {
352     WSACleanup();
353 }
354 #endif
355 
356 int socket_init(void)
357 {
358 #ifdef _WIN32
359     WSADATA Data;
360     int ret, err;
361 
362     ret = WSAStartup(MAKEWORD(2, 2), &Data);
363     if (ret != 0) {
364         err = WSAGetLastError();
365         fprintf(stderr, "WSAStartup: %d\n", err);
366         return -1;
367     }
368     atexit(socket_cleanup);
369 #endif
370     return 0;
371 }
372 
373 #if !GLIB_CHECK_VERSION(2, 31, 0)
374 /* Ensure that glib is running in multi-threaded mode
375  * Old versions of glib require explicit initialization.  Failure to do
376  * this results in the single-threaded code paths being taken inside
377  * glib.  For example, the g_slice allocator will not be thread-safe
378  * and cause crashes.
379  */
380 static void __attribute__((constructor)) thread_init(void)
381 {
382     if (!g_thread_supported()) {
383        g_thread_init(NULL);
384     }
385 }
386 #endif
387 
388 #ifndef CONFIG_IOVEC
389 /* helper function for iov_send_recv() */
390 static ssize_t
391 readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write)
392 {
393     unsigned i = 0;
394     ssize_t ret = 0;
395     while (i < iov_cnt) {
396         ssize_t r = do_write
397             ? write(fd, iov[i].iov_base, iov[i].iov_len)
398             : read(fd, iov[i].iov_base, iov[i].iov_len);
399         if (r > 0) {
400             ret += r;
401         } else if (!r) {
402             break;
403         } else if (errno == EINTR) {
404             continue;
405         } else {
406             /* else it is some "other" error,
407              * only return if there was no data processed. */
408             if (ret == 0) {
409                 ret = -1;
410             }
411             break;
412         }
413         i++;
414     }
415     return ret;
416 }
417 
418 ssize_t
419 readv(int fd, const struct iovec *iov, int iov_cnt)
420 {
421     return readv_writev(fd, iov, iov_cnt, false);
422 }
423 
424 ssize_t
425 writev(int fd, const struct iovec *iov, int iov_cnt)
426 {
427     return readv_writev(fd, iov, iov_cnt, true);
428 }
429 #endif
430