xref: /openbmc/qemu/util/osdep.c (revision 0a9516c2)
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 "qemu/error-report.h"
51 #include "monitor/monitor.h"
52 
53 static bool fips_enabled = false;
54 
55 static const char *hw_version = QEMU_VERSION;
56 
57 int socket_set_cork(int fd, int v)
58 {
59 #if defined(SOL_TCP) && defined(TCP_CORK)
60     return qemu_setsockopt(fd, SOL_TCP, TCP_CORK, &v, sizeof(v));
61 #else
62     return 0;
63 #endif
64 }
65 
66 int socket_set_nodelay(int fd)
67 {
68     int v = 1;
69     return qemu_setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &v, sizeof(v));
70 }
71 
72 int qemu_madvise(void *addr, size_t len, int advice)
73 {
74     if (advice == QEMU_MADV_INVALID) {
75         errno = EINVAL;
76         return -1;
77     }
78 #if defined(CONFIG_MADVISE)
79     return madvise(addr, len, advice);
80 #elif defined(CONFIG_POSIX_MADVISE)
81     return posix_madvise(addr, len, advice);
82 #else
83     errno = EINVAL;
84     return -1;
85 #endif
86 }
87 
88 #ifndef _WIN32
89 /*
90  * Dups an fd and sets the flags
91  */
92 static int qemu_dup_flags(int fd, int flags)
93 {
94     int ret;
95     int serrno;
96     int dup_flags;
97 
98 #ifdef F_DUPFD_CLOEXEC
99     ret = fcntl(fd, F_DUPFD_CLOEXEC, 0);
100 #else
101     ret = dup(fd);
102     if (ret != -1) {
103         qemu_set_cloexec(ret);
104     }
105 #endif
106     if (ret == -1) {
107         goto fail;
108     }
109 
110     dup_flags = fcntl(ret, F_GETFL);
111     if (dup_flags == -1) {
112         goto fail;
113     }
114 
115     if ((flags & O_SYNC) != (dup_flags & O_SYNC)) {
116         errno = EINVAL;
117         goto fail;
118     }
119 
120     /* Set/unset flags that we can with fcntl */
121     if (fcntl(ret, F_SETFL, flags) == -1) {
122         goto fail;
123     }
124 
125     /* Truncate the file in the cases that open() would truncate it */
126     if (flags & O_TRUNC ||
127             ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))) {
128         if (ftruncate(ret, 0) == -1) {
129             goto fail;
130         }
131     }
132 
133     return ret;
134 
135 fail:
136     serrno = errno;
137     if (ret != -1) {
138         close(ret);
139     }
140     errno = serrno;
141     return -1;
142 }
143 
144 static int qemu_parse_fdset(const char *param)
145 {
146     return qemu_parse_fd(param);
147 }
148 #endif
149 
150 /*
151  * Opens a file with FD_CLOEXEC set
152  */
153 int qemu_open(const char *name, int flags, ...)
154 {
155     int ret;
156     int mode = 0;
157 
158 #ifndef _WIN32
159     const char *fdset_id_str;
160 
161     /* Attempt dup of fd from fd set */
162     if (strstart(name, "/dev/fdset/", &fdset_id_str)) {
163         int64_t fdset_id;
164         int fd, dupfd;
165 
166         fdset_id = qemu_parse_fdset(fdset_id_str);
167         if (fdset_id == -1) {
168             errno = EINVAL;
169             return -1;
170         }
171 
172         fd = monitor_fdset_get_fd(fdset_id, flags);
173         if (fd == -1) {
174             return -1;
175         }
176 
177         dupfd = qemu_dup_flags(fd, flags);
178         if (dupfd == -1) {
179             return -1;
180         }
181 
182         ret = monitor_fdset_dup_fd_add(fdset_id, dupfd);
183         if (ret == -1) {
184             close(dupfd);
185             errno = EINVAL;
186             return -1;
187         }
188 
189         return dupfd;
190     }
191 #endif
192 
193     if (flags & O_CREAT) {
194         va_list ap;
195 
196         va_start(ap, flags);
197         mode = va_arg(ap, int);
198         va_end(ap);
199     }
200 
201 #ifdef O_CLOEXEC
202     ret = open(name, flags | O_CLOEXEC, mode);
203 #else
204     ret = open(name, flags, mode);
205     if (ret >= 0) {
206         qemu_set_cloexec(ret);
207     }
208 #endif
209 
210 #ifdef O_DIRECT
211     if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
212         error_report("file system may not support O_DIRECT");
213         errno = EINVAL; /* in case it was clobbered */
214     }
215 #endif /* O_DIRECT */
216 
217     return ret;
218 }
219 
220 int qemu_close(int fd)
221 {
222     int64_t fdset_id;
223 
224     /* Close fd that was dup'd from an fdset */
225     fdset_id = monitor_fdset_dup_fd_find(fd);
226     if (fdset_id != -1) {
227         int ret;
228 
229         ret = close(fd);
230         if (ret == 0) {
231             monitor_fdset_dup_fd_remove(fd);
232         }
233 
234         return ret;
235     }
236 
237     return close(fd);
238 }
239 
240 /*
241  * A variant of write(2) which handles partial write.
242  *
243  * Return the number of bytes transferred.
244  * Set errno if fewer than `count' bytes are written.
245  *
246  * This function don't work with non-blocking fd's.
247  * Any of the possibilities with non-bloking fd's is bad:
248  *   - return a short write (then name is wrong)
249  *   - busy wait adding (errno == EAGAIN) to the loop
250  */
251 ssize_t qemu_write_full(int fd, const void *buf, size_t count)
252 {
253     ssize_t ret = 0;
254     ssize_t total = 0;
255 
256     while (count) {
257         ret = write(fd, buf, count);
258         if (ret < 0) {
259             if (errno == EINTR)
260                 continue;
261             break;
262         }
263 
264         count -= ret;
265         buf += ret;
266         total += ret;
267     }
268 
269     return total;
270 }
271 
272 /*
273  * Opens a socket with FD_CLOEXEC set
274  */
275 int qemu_socket(int domain, int type, int protocol)
276 {
277     int ret;
278 
279 #ifdef SOCK_CLOEXEC
280     ret = socket(domain, type | SOCK_CLOEXEC, protocol);
281     if (ret != -1 || errno != EINVAL) {
282         return ret;
283     }
284 #endif
285     ret = socket(domain, type, protocol);
286     if (ret >= 0) {
287         qemu_set_cloexec(ret);
288     }
289 
290     return ret;
291 }
292 
293 /*
294  * Accept a connection and set FD_CLOEXEC
295  */
296 int qemu_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
297 {
298     int ret;
299 
300 #ifdef CONFIG_ACCEPT4
301     ret = accept4(s, addr, addrlen, SOCK_CLOEXEC);
302     if (ret != -1 || errno != ENOSYS) {
303         return ret;
304     }
305 #endif
306     ret = accept(s, addr, addrlen);
307     if (ret >= 0) {
308         qemu_set_cloexec(ret);
309     }
310 
311     return ret;
312 }
313 
314 void qemu_set_hw_version(const char *version)
315 {
316     hw_version = version;
317 }
318 
319 const char *qemu_hw_version(void)
320 {
321     return hw_version;
322 }
323 
324 void fips_set_state(bool requested)
325 {
326 #ifdef __linux__
327     if (requested) {
328         FILE *fds = fopen("/proc/sys/crypto/fips_enabled", "r");
329         if (fds != NULL) {
330             fips_enabled = (fgetc(fds) == '1');
331             fclose(fds);
332         }
333     }
334 #else
335     fips_enabled = false;
336 #endif /* __linux__ */
337 
338 #ifdef _FIPS_DEBUG
339     fprintf(stderr, "FIPS mode %s (requested %s)\n",
340 	    (fips_enabled ? "enabled" : "disabled"),
341 	    (requested ? "enabled" : "disabled"));
342 #endif
343 }
344 
345 bool fips_get_state(void)
346 {
347     return fips_enabled;
348 }
349 
350 #ifdef _WIN32
351 static void socket_cleanup(void)
352 {
353     WSACleanup();
354 }
355 #endif
356 
357 int socket_init(void)
358 {
359 #ifdef _WIN32
360     WSADATA Data;
361     int ret, err;
362 
363     ret = WSAStartup(MAKEWORD(2, 2), &Data);
364     if (ret != 0) {
365         err = WSAGetLastError();
366         fprintf(stderr, "WSAStartup: %d\n", err);
367         return -1;
368     }
369     atexit(socket_cleanup);
370 #endif
371     return 0;
372 }
373 
374 #if !GLIB_CHECK_VERSION(2, 31, 0)
375 /* Ensure that glib is running in multi-threaded mode
376  * Old versions of glib require explicit initialization.  Failure to do
377  * this results in the single-threaded code paths being taken inside
378  * glib.  For example, the g_slice allocator will not be thread-safe
379  * and cause crashes.
380  */
381 static void __attribute__((constructor)) thread_init(void)
382 {
383     if (!g_thread_supported()) {
384        g_thread_init(NULL);
385     }
386 }
387 #endif
388 
389 #ifndef CONFIG_IOVEC
390 /* helper function for iov_send_recv() */
391 static ssize_t
392 readv_writev(int fd, const struct iovec *iov, int iov_cnt, bool do_write)
393 {
394     unsigned i = 0;
395     ssize_t ret = 0;
396     while (i < iov_cnt) {
397         ssize_t r = do_write
398             ? write(fd, iov[i].iov_base, iov[i].iov_len)
399             : read(fd, iov[i].iov_base, iov[i].iov_len);
400         if (r > 0) {
401             ret += r;
402         } else if (!r) {
403             break;
404         } else if (errno == EINTR) {
405             continue;
406         } else {
407             /* else it is some "other" error,
408              * only return if there was no data processed. */
409             if (ret == 0) {
410                 ret = -1;
411             }
412             break;
413         }
414         i++;
415     }
416     return ret;
417 }
418 
419 ssize_t
420 readv(int fd, const struct iovec *iov, int iov_cnt)
421 {
422     return readv_writev(fd, iov, iov_cnt, false);
423 }
424 
425 ssize_t
426 writev(int fd, const struct iovec *iov, int iov_cnt)
427 {
428     return readv_writev(fd, iov, iov_cnt, true);
429 }
430 #endif
431