]>
Commit | Line | Data |
---|---|---|
e3642c43 DL |
1 | /* |
2 | * lxc: linux Container library | |
3 | * | |
4 | * (C) Copyright IBM Corp. 2007, 2008 | |
5 | * | |
6 | * Authors: | |
9afe19d6 | 7 | * Daniel Lezcano <daniel.lezcano at free.fr> |
e3642c43 DL |
8 | * |
9 | * This library is free software; you can redistribute it and/or | |
10 | * modify it under the terms of the GNU Lesser General Public | |
11 | * License as published by the Free Software Foundation; either | |
12 | * version 2.1 of the License, or (at your option) any later version. | |
13 | * | |
14 | * This library is distributed in the hope that it will be useful, | |
15 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
17 | * Lesser General Public License for more details. | |
18 | * | |
19 | * You should have received a copy of the GNU Lesser General Public | |
20 | * License along with this library; if not, write to the Free Software | |
250b1eec | 21 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
e3642c43 DL |
22 | */ |
23 | ||
d38dd64a CB |
24 | #ifndef _GNU_SOURCE |
25 | #define _GNU_SOURCE 1 | |
26 | #endif | |
7935833c | 27 | #define __STDC_FORMAT_MACROS /* Required for PRIu64 to work. */ |
643c1984 | 28 | #include <ctype.h> |
a1e5280d | 29 | #include <dirent.h> |
e3642c43 | 30 | #include <errno.h> |
a1e5280d | 31 | #include <fcntl.h> |
dbaf55a3 | 32 | #include <grp.h> |
7935833c | 33 | #include <inttypes.h> |
a1e5280d | 34 | #include <libgen.h> |
b467714b | 35 | #include <pthread.h> |
d983b93c | 36 | #include <stddef.h> |
a1e5280d CB |
37 | #include <stdio.h> |
38 | #include <stdlib.h> | |
61a1d519 | 39 | #include <string.h> |
e3642c43 | 40 | #include <sys/mman.h> |
6e4bb2e0 | 41 | #include <sys/mount.h> |
a1e5280d CB |
42 | #include <sys/param.h> |
43 | #include <sys/prctl.h> | |
44 | #include <sys/stat.h> | |
9be53773 SH |
45 | #include <sys/types.h> |
46 | #include <sys/wait.h> | |
d38dd64a | 47 | #include <unistd.h> |
e3642c43 | 48 | |
d38dd64a | 49 | #include "config.h" |
e3642c43 | 50 | #include "log.h" |
025ed0f3 | 51 | #include "lxclock.h" |
51d0854c | 52 | #include "namespace.h" |
e3db0162 | 53 | #include "parse.h" |
981f6029 | 54 | #include "utils.h" |
e3642c43 | 55 | |
43f984ea DJ |
56 | #ifndef HAVE_STRLCPY |
57 | #include "include/strlcpy.h" | |
58 | #endif | |
59 | ||
bd583214 DJ |
60 | #ifndef HAVE_STRLCAT |
61 | #include "include/strlcat.h" | |
62 | #endif | |
63 | ||
4928c718 SG |
64 | #ifndef O_PATH |
65 | #define O_PATH 010000000 | |
66 | #endif | |
67 | ||
68 | #ifndef O_NOFOLLOW | |
69 | #define O_NOFOLLOW 00400000 | |
70 | #endif | |
71 | ||
ac2cecc4 | 72 | lxc_log_define(utils, lxc); |
e3642c43 | 73 | |
4295c5de SH |
74 | /* |
75 | * if path is btrfs, tries to remove it and any subvolumes beneath it | |
76 | */ | |
77 | extern bool btrfs_try_remove_subvol(const char *path); | |
78 | ||
41dc7155 | 79 | static int _recursive_rmdir(const char *dirname, dev_t pdev, |
0cc417b2 | 80 | const char *exclude, int level, bool onedev) |
60bf62d4 | 81 | { |
74f96976 | 82 | struct dirent *direntp; |
60bf62d4 SH |
83 | DIR *dir; |
84 | int ret, failed=0; | |
85 | char pathname[MAXPATHLEN]; | |
18aa217b | 86 | bool hadexclude = false; |
60bf62d4 SH |
87 | |
88 | dir = opendir(dirname); | |
89 | if (!dir) { | |
b103ceac | 90 | ERROR("failed to open %s", dirname); |
4355ab5f | 91 | return -1; |
60bf62d4 SH |
92 | } |
93 | ||
74f96976 | 94 | while ((direntp = readdir(dir))) { |
60bf62d4 SH |
95 | struct stat mystat; |
96 | int rc; | |
97 | ||
60bf62d4 SH |
98 | if (!strcmp(direntp->d_name, ".") || |
99 | !strcmp(direntp->d_name, "..")) | |
100 | continue; | |
101 | ||
102 | rc = snprintf(pathname, MAXPATHLEN, "%s/%s", dirname, direntp->d_name); | |
103 | if (rc < 0 || rc >= MAXPATHLEN) { | |
104 | ERROR("pathname too long"); | |
105 | failed=1; | |
106 | continue; | |
107 | } | |
18aa217b SH |
108 | |
109 | if (!level && exclude && !strcmp(direntp->d_name, exclude)) { | |
110 | ret = rmdir(pathname); | |
111 | if (ret < 0) { | |
112 | switch(errno) { | |
113 | case ENOTEMPTY: | |
0cc417b2 | 114 | INFO("Not deleting snapshot %s", pathname); |
18aa217b SH |
115 | hadexclude = true; |
116 | break; | |
117 | case ENOTDIR: | |
118 | ret = unlink(pathname); | |
119 | if (ret) | |
b103ceac | 120 | INFO("Failed to remove %s", pathname); |
18aa217b SH |
121 | break; |
122 | default: | |
b103ceac | 123 | SYSERROR("Failed to rmdir %s", pathname); |
18aa217b SH |
124 | failed = 1; |
125 | break; | |
126 | } | |
127 | } | |
128 | continue; | |
129 | } | |
130 | ||
60bf62d4 SH |
131 | ret = lstat(pathname, &mystat); |
132 | if (ret) { | |
b103ceac | 133 | ERROR("Failed to stat %s", pathname); |
4295c5de | 134 | failed = 1; |
60bf62d4 SH |
135 | continue; |
136 | } | |
b14fc100 | 137 | |
4295c5de SH |
138 | if (onedev && mystat.st_dev != pdev) { |
139 | /* TODO should we be checking /proc/self/mountinfo for | |
140 | * pathname and not doing this if found? */ | |
141 | if (btrfs_try_remove_subvol(pathname)) | |
142 | INFO("Removed btrfs subvolume at %s\n", pathname); | |
60bf62d4 | 143 | continue; |
4295c5de | 144 | } |
b14fc100 | 145 | |
60bf62d4 | 146 | if (S_ISDIR(mystat.st_mode)) { |
0cc417b2 | 147 | if (_recursive_rmdir(pathname, pdev, exclude, level+1, onedev) < 0) |
60bf62d4 SH |
148 | failed=1; |
149 | } else { | |
150 | if (unlink(pathname) < 0) { | |
b103ceac | 151 | SYSERROR("Failed to delete %s", pathname); |
60bf62d4 SH |
152 | failed=1; |
153 | } | |
154 | } | |
155 | } | |
156 | ||
4295c5de | 157 | if (rmdir(dirname) < 0 && !btrfs_try_remove_subvol(dirname) && !hadexclude) { |
b103ceac | 158 | ERROR("Failed to delete %s", dirname); |
4295c5de | 159 | failed=1; |
60bf62d4 SH |
160 | } |
161 | ||
025ed0f3 | 162 | ret = closedir(dir); |
025ed0f3 | 163 | if (ret) { |
b103ceac | 164 | ERROR("Failed to close directory %s", dirname); |
60bf62d4 SH |
165 | failed=1; |
166 | } | |
167 | ||
4355ab5f | 168 | return failed ? -1 : 0; |
60bf62d4 SH |
169 | } |
170 | ||
29a11a7f CB |
171 | /* In overlayfs, st_dev is unreliable. So on overlayfs we don't do the |
172 | * lxc_rmdir_onedev() | |
0cc417b2 SH |
173 | */ |
174 | static bool is_native_overlayfs(const char *path) | |
175 | { | |
29a11a7f CB |
176 | if (has_fs_type(path, OVERLAY_SUPER_MAGIC) || |
177 | has_fs_type(path, OVERLAYFS_SUPER_MAGIC)) | |
0cc417b2 | 178 | return true; |
29a11a7f | 179 | |
0cc417b2 SH |
180 | return false; |
181 | } | |
182 | ||
4355ab5f | 183 | /* returns 0 on success, -1 if there were any failures */ |
41dc7155 | 184 | extern int lxc_rmdir_onedev(const char *path, const char *exclude) |
60bf62d4 SH |
185 | { |
186 | struct stat mystat; | |
0cc417b2 SH |
187 | bool onedev = true; |
188 | ||
41dc7155 | 189 | if (is_native_overlayfs(path)) |
0cc417b2 | 190 | onedev = false; |
60bf62d4 SH |
191 | |
192 | if (lstat(path, &mystat) < 0) { | |
067650d0 SH |
193 | if (errno == ENOENT) |
194 | return 0; | |
41dc7155 | 195 | |
b103ceac | 196 | ERROR("Failed to stat %s", path); |
4355ab5f | 197 | return -1; |
60bf62d4 SH |
198 | } |
199 | ||
0cc417b2 | 200 | return _recursive_rmdir(path, mystat.st_dev, exclude, 0, onedev); |
60bf62d4 SH |
201 | } |
202 | ||
9ddaf3bf | 203 | /* borrowed from iproute2 */ |
7c11d57a | 204 | extern int get_u16(unsigned short *val, const char *arg, int base) |
9ddaf3bf JHS |
205 | { |
206 | unsigned long res; | |
207 | char *ptr; | |
208 | ||
209 | if (!arg || !*arg) | |
210 | return -1; | |
211 | ||
09bbd745 | 212 | errno = 0; |
9ddaf3bf | 213 | res = strtoul(arg, &ptr, base); |
09bbd745 | 214 | if (!ptr || ptr == arg || *ptr || res > 0xFFFF || errno != 0) |
9ddaf3bf JHS |
215 | return -1; |
216 | ||
217 | *val = res; | |
218 | ||
219 | return 0; | |
220 | } | |
221 | ||
3ce74686 | 222 | extern int mkdir_p(const char *dir, mode_t mode) |
1b09f2c0 | 223 | { |
3ce74686 SH |
224 | const char *tmp = dir; |
225 | const char *orig = dir; | |
860fc865 RW |
226 | char *makeme; |
227 | ||
228 | do { | |
229 | dir = tmp + strspn(tmp, "/"); | |
230 | tmp = dir + strcspn(dir, "/"); | |
b14fc100 | 231 | |
d74325c4 | 232 | makeme = strndup(orig, dir - orig); |
860fc865 RW |
233 | if (*makeme) { |
234 | if (mkdir(makeme, mode) && errno != EEXIST) { | |
959aee9c | 235 | SYSERROR("failed to create directory '%s'", makeme); |
d74325c4 | 236 | free(makeme); |
860fc865 RW |
237 | return -1; |
238 | } | |
239 | } | |
d74325c4 | 240 | free(makeme); |
860fc865 | 241 | } while(tmp != dir); |
1b09f2c0 | 242 | |
98663823 | 243 | return 0; |
1b09f2c0 | 244 | } |
2a59a681 | 245 | |
44b9ae4b | 246 | char *get_rundir() |
9e60f51d | 247 | { |
97a696c6 SG |
248 | char *rundir; |
249 | const char *homedir; | |
9650c735 | 250 | struct stat sb; |
9e60f51d | 251 | |
b14fc100 | 252 | if (stat(RUNTIME_PATH, &sb) < 0) |
9650c735 | 253 | return NULL; |
9650c735 TA |
254 | |
255 | if (geteuid() == sb.st_uid || getegid() == sb.st_gid) { | |
c580b8d2 | 256 | rundir = strdup(RUNTIME_PATH); |
d6470e71 SG |
257 | return rundir; |
258 | } | |
97a696c6 SG |
259 | |
260 | rundir = getenv("XDG_RUNTIME_DIR"); | |
44b9ae4b SG |
261 | if (rundir) { |
262 | rundir = strdup(rundir); | |
263 | return rundir; | |
264 | } | |
97a696c6 | 265 | |
44b9ae4b SG |
266 | INFO("XDG_RUNTIME_DIR isn't set in the environment."); |
267 | homedir = getenv("HOME"); | |
268 | if (!homedir) { | |
269 | ERROR("HOME isn't set in the environment."); | |
270 | return NULL; | |
97a696c6 SG |
271 | } |
272 | ||
44b9ae4b | 273 | rundir = malloc(sizeof(char) * (17 + strlen(homedir))); |
b14fc100 | 274 | if (!rundir) |
275 | return NULL; | |
276 | ||
44b9ae4b SG |
277 | sprintf(rundir, "%s/.cache/lxc/run/", homedir); |
278 | ||
9e60f51d DE |
279 | return rundir; |
280 | } | |
281 | ||
9be53773 SH |
282 | int wait_for_pid(pid_t pid) |
283 | { | |
284 | int status, ret; | |
285 | ||
286 | again: | |
287 | ret = waitpid(pid, &status, 0); | |
288 | if (ret == -1) { | |
71b9b8ed | 289 | if (errno == EINTR) |
9be53773 | 290 | goto again; |
b14fc100 | 291 | |
9be53773 SH |
292 | return -1; |
293 | } | |
b14fc100 | 294 | |
9be53773 SH |
295 | if (ret != pid) |
296 | goto again; | |
b14fc100 | 297 | |
9be53773 SH |
298 | if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) |
299 | return -1; | |
b14fc100 | 300 | |
9be53773 SH |
301 | return 0; |
302 | } | |
c797a220 CS |
303 | |
304 | int lxc_wait_for_pid_status(pid_t pid) | |
305 | { | |
306 | int status, ret; | |
307 | ||
308 | again: | |
309 | ret = waitpid(pid, &status, 0); | |
310 | if (ret == -1) { | |
311 | if (errno == EINTR) | |
312 | goto again; | |
b14fc100 | 313 | |
c797a220 CS |
314 | return -1; |
315 | } | |
b14fc100 | 316 | |
c797a220 CS |
317 | if (ret != pid) |
318 | goto again; | |
b14fc100 | 319 | |
c797a220 CS |
320 | return status; |
321 | } | |
92f023dc | 322 | |
3ce74686 SH |
323 | #if HAVE_LIBGNUTLS |
324 | #include <gnutls/gnutls.h> | |
325 | #include <gnutls/crypto.h> | |
41246cee DE |
326 | |
327 | __attribute__((constructor)) | |
328 | static void gnutls_lxc_init(void) | |
329 | { | |
330 | gnutls_global_init(); | |
331 | } | |
332 | ||
3ce74686 SH |
333 | int sha1sum_file(char *fnam, unsigned char *digest) |
334 | { | |
335 | char *buf; | |
336 | int ret; | |
337 | FILE *f; | |
338 | long flen; | |
339 | ||
340 | if (!fnam) | |
341 | return -1; | |
b14fc100 | 342 | |
025ed0f3 | 343 | f = fopen_cloexec(fnam, "r"); |
7be677a8 | 344 | if (!f) { |
3ce74686 SH |
345 | SYSERROR("Error opening template"); |
346 | return -1; | |
347 | } | |
b14fc100 | 348 | |
3ce74686 SH |
349 | if (fseek(f, 0, SEEK_END) < 0) { |
350 | SYSERROR("Error seeking to end of template"); | |
dd1d77f9 | 351 | fclose(f); |
3ce74686 SH |
352 | return -1; |
353 | } | |
b14fc100 | 354 | |
3ce74686 SH |
355 | if ((flen = ftell(f)) < 0) { |
356 | SYSERROR("Error telling size of template"); | |
dd1d77f9 | 357 | fclose(f); |
3ce74686 SH |
358 | return -1; |
359 | } | |
b14fc100 | 360 | |
3ce74686 SH |
361 | if (fseek(f, 0, SEEK_SET) < 0) { |
362 | SYSERROR("Error seeking to start of template"); | |
dd1d77f9 | 363 | fclose(f); |
3ce74686 SH |
364 | return -1; |
365 | } | |
b14fc100 | 366 | |
3ce74686 SH |
367 | if ((buf = malloc(flen+1)) == NULL) { |
368 | SYSERROR("Out of memory"); | |
dd1d77f9 | 369 | fclose(f); |
3ce74686 SH |
370 | return -1; |
371 | } | |
b14fc100 | 372 | |
3ce74686 SH |
373 | if (fread(buf, 1, flen, f) != flen) { |
374 | SYSERROR("Failure reading template"); | |
375 | free(buf); | |
dd1d77f9 | 376 | fclose(f); |
3ce74686 SH |
377 | return -1; |
378 | } | |
b14fc100 | 379 | |
dd1d77f9 | 380 | if (fclose(f) < 0) { |
3ce74686 SH |
381 | SYSERROR("Failre closing template"); |
382 | free(buf); | |
383 | return -1; | |
384 | } | |
b14fc100 | 385 | |
3ce74686 SH |
386 | buf[flen] = '\0'; |
387 | ret = gnutls_hash_fast(GNUTLS_DIG_SHA1, buf, flen, (void *)digest); | |
388 | free(buf); | |
389 | return ret; | |
390 | } | |
391 | #endif | |
61a1d519 | 392 | |
8bd8018e | 393 | struct lxc_popen_FILE *lxc_popen(const char *command) |
ebec9176 | 394 | { |
3f323207 | 395 | int ret; |
ebec9176 AM |
396 | int pipe_fds[2]; |
397 | pid_t child_pid; | |
8bd8018e | 398 | struct lxc_popen_FILE *fp = NULL; |
ebec9176 | 399 | |
8bd8018e CB |
400 | ret = pipe2(pipe_fds, O_CLOEXEC); |
401 | if (ret < 0) | |
ebec9176 | 402 | return NULL; |
ebec9176 AM |
403 | |
404 | child_pid = fork(); | |
8bd8018e CB |
405 | if (child_pid < 0) |
406 | goto on_error; | |
407 | ||
408 | if (!child_pid) { | |
409 | sigset_t mask; | |
410 | ||
411 | close(pipe_fds[0]); | |
412 | ||
413 | /* duplicate stdout */ | |
414 | if (pipe_fds[1] != STDOUT_FILENO) | |
415 | ret = dup2(pipe_fds[1], STDOUT_FILENO); | |
416 | else | |
417 | ret = fcntl(pipe_fds[1], F_SETFD, 0); | |
418 | if (ret < 0) { | |
419 | close(pipe_fds[1]); | |
03f618af | 420 | _exit(EXIT_FAILURE); |
3f323207 CB |
421 | } |
422 | ||
8bd8018e CB |
423 | /* duplicate stderr */ |
424 | if (pipe_fds[1] != STDERR_FILENO) | |
425 | ret = dup2(pipe_fds[1], STDERR_FILENO); | |
426 | else | |
427 | ret = fcntl(pipe_fds[1], F_SETFD, 0); | |
428 | close(pipe_fds[1]); | |
429 | if (ret < 0) | |
03f618af | 430 | _exit(EXIT_FAILURE); |
8bd8018e CB |
431 | |
432 | /* unblock all signals */ | |
433 | ret = sigfillset(&mask); | |
434 | if (ret < 0) | |
03f618af | 435 | _exit(EXIT_FAILURE); |
8bd8018e | 436 | |
b467714b | 437 | ret = pthread_sigmask(SIG_UNBLOCK, &mask, NULL); |
8bd8018e | 438 | if (ret < 0) |
03f618af | 439 | _exit(EXIT_FAILURE); |
8bd8018e CB |
440 | |
441 | execl("/bin/sh", "sh", "-c", command, (char *)NULL); | |
03f618af | 442 | _exit(127); |
ebec9176 AM |
443 | } |
444 | ||
8bd8018e CB |
445 | close(pipe_fds[1]); |
446 | pipe_fds[1] = -1; | |
ebec9176 | 447 | |
8bd8018e CB |
448 | fp = malloc(sizeof(*fp)); |
449 | if (!fp) | |
450 | goto on_error; | |
b14fc100 | 451 | |
7e50ec0b | 452 | memset(fp, 0, sizeof(*fp)); |
ebec9176 AM |
453 | |
454 | fp->child_pid = child_pid; | |
8bd8018e | 455 | fp->pipe = pipe_fds[0]; |
ebec9176 | 456 | |
7e50ec0b CB |
457 | /* From now on, closing fp->f will also close fp->pipe. So only ever |
458 | * call fclose(fp->f). | |
459 | */ | |
8bd8018e CB |
460 | fp->f = fdopen(pipe_fds[0], "r"); |
461 | if (!fp->f) | |
462 | goto on_error; | |
ebec9176 | 463 | |
8bd8018e | 464 | return fp; |
ebec9176 | 465 | |
8bd8018e | 466 | on_error: |
7e50ec0b CB |
467 | /* We can only close pipe_fds[0] if fdopen() didn't succeed or wasn't |
468 | * called yet. Otherwise the fd belongs to the file opened by fdopen() | |
469 | * since it isn't dup()ed. | |
470 | */ | |
471 | if (fp && !fp->f && pipe_fds[0] >= 0) | |
8bd8018e CB |
472 | close(pipe_fds[0]); |
473 | ||
474 | if (pipe_fds[1] >= 0) | |
475 | close(pipe_fds[1]); | |
ebec9176 | 476 | |
7e50ec0b CB |
477 | if (fp && fp->f) |
478 | fclose(fp->f); | |
479 | ||
480 | if (fp) | |
481 | free(fp); | |
482 | ||
ebec9176 AM |
483 | return NULL; |
484 | } | |
485 | ||
8bd8018e | 486 | int lxc_pclose(struct lxc_popen_FILE *fp) |
ebec9176 | 487 | { |
ebec9176 | 488 | pid_t wait_pid; |
8bd8018e | 489 | int wstatus = 0; |
ebec9176 | 490 | |
8bd8018e | 491 | if (!fp) |
ebec9176 | 492 | return -1; |
ebec9176 AM |
493 | |
494 | do { | |
8bd8018e CB |
495 | wait_pid = waitpid(fp->child_pid, &wstatus, 0); |
496 | } while (wait_pid < 0 && errno == EINTR); | |
ebec9176 | 497 | |
8bd8018e CB |
498 | fclose(fp->f); |
499 | free(fp); | |
500 | ||
501 | if (wait_pid < 0) | |
ebec9176 | 502 | return -1; |
ebec9176 AM |
503 | |
504 | return wstatus; | |
505 | } | |
506 | ||
508c263e SH |
507 | int randseed(bool srand_it) |
508 | { | |
509 | /* | |
510 | srand pre-seed function based on /dev/urandom | |
511 | */ | |
091045f8 | 512 | unsigned int seed = time(NULL) + getpid(); |
508c263e SH |
513 | |
514 | FILE *f; | |
515 | f = fopen("/dev/urandom", "r"); | |
516 | if (f) { | |
517 | int ret = fread(&seed, sizeof(seed), 1, f); | |
518 | if (ret != 1) | |
7874d81a | 519 | SYSDEBUG("unable to fread /dev/urandom, fallback to time+pid rand seed"); |
520 | ||
508c263e SH |
521 | fclose(f); |
522 | } | |
523 | ||
524 | if (srand_it) | |
525 | srand(seed); | |
526 | ||
527 | return seed; | |
528 | } | |
5d897655 SH |
529 | |
530 | uid_t get_ns_uid(uid_t orig) | |
531 | { | |
532 | char *line = NULL; | |
533 | size_t sz = 0; | |
534 | uid_t nsid, hostid, range; | |
535 | FILE *f = fopen("/proc/self/uid_map", "r"); | |
536 | if (!f) | |
537 | return 0; | |
538 | ||
539 | while (getline(&line, &sz, f) != -1) { | |
540 | if (sscanf(line, "%u %u %u", &nsid, &hostid, &range) != 3) | |
541 | continue; | |
b14fc100 | 542 | |
5d897655 SH |
543 | if (hostid <= orig && hostid + range > orig) { |
544 | nsid += orig - hostid; | |
545 | goto found; | |
546 | } | |
547 | } | |
548 | ||
b962868f CB |
549 | nsid = LXC_INVALID_UID; |
550 | ||
551 | found: | |
552 | fclose(f); | |
553 | free(line); | |
554 | return nsid; | |
555 | } | |
556 | ||
557 | gid_t get_ns_gid(gid_t orig) | |
558 | { | |
559 | char *line = NULL; | |
560 | size_t sz = 0; | |
561 | gid_t nsid, hostid, range; | |
562 | FILE *f = fopen("/proc/self/gid_map", "r"); | |
563 | if (!f) | |
564 | return 0; | |
565 | ||
566 | while (getline(&line, &sz, f) != -1) { | |
567 | if (sscanf(line, "%u %u %u", &nsid, &hostid, &range) != 3) | |
568 | continue; | |
569 | ||
570 | if (hostid <= orig && hostid + range > orig) { | |
571 | nsid += orig - hostid; | |
572 | goto found; | |
573 | } | |
574 | } | |
575 | ||
576 | nsid = LXC_INVALID_GID; | |
b14fc100 | 577 | |
5d897655 SH |
578 | found: |
579 | fclose(f); | |
580 | free(line); | |
581 | return nsid; | |
582 | } | |
c476bdce SH |
583 | |
584 | bool dir_exists(const char *path) | |
585 | { | |
586 | struct stat sb; | |
587 | int ret; | |
588 | ||
589 | ret = stat(path, &sb); | |
590 | if (ret < 0) | |
1a0e70ac | 591 | /* Could be something other than eexist, just say "no". */ |
c476bdce | 592 | return false; |
b14fc100 | 593 | |
c476bdce SH |
594 | return S_ISDIR(sb.st_mode); |
595 | } | |
93c379f0 ÇO |
596 | |
597 | /* Note we don't use SHA-1 here as we don't want to depend on HAVE_GNUTLS. | |
598 | * FNV has good anti collision properties and we're not worried | |
599 | * about pre-image resistance or one-way-ness, we're just trying to make | |
600 | * the name unique in the 108 bytes of space we have. | |
601 | */ | |
602 | uint64_t fnv_64a_buf(void *buf, size_t len, uint64_t hval) | |
603 | { | |
604 | unsigned char *bp; | |
605 | ||
606 | for(bp = buf; bp < (unsigned char *)buf + len; bp++) | |
607 | { | |
608 | /* xor the bottom with the current octet */ | |
609 | hval ^= (uint64_t)*bp; | |
610 | ||
611 | /* gcc optimised: | |
612 | * multiply by the 64 bit FNV magic prime mod 2^64 | |
613 | */ | |
614 | hval += (hval << 1) + (hval << 4) + (hval << 5) + | |
615 | (hval << 7) + (hval << 8) + (hval << 40); | |
616 | } | |
617 | ||
618 | return hval; | |
619 | } | |
2c6f3fc9 | 620 | |
f6310f18 | 621 | bool is_shared_mountpoint(const char *path) |
2c6f3fc9 | 622 | { |
f6310f18 | 623 | char buf[LXC_LINELEN]; |
2c6f3fc9 SH |
624 | FILE *f; |
625 | int i; | |
f6310f18 | 626 | char *p, *p2; |
2c6f3fc9 SH |
627 | |
628 | f = fopen("/proc/self/mountinfo", "r"); | |
629 | if (!f) | |
630 | return 0; | |
b14fc100 | 631 | |
eab15c1e CB |
632 | while (fgets(buf, LXC_LINELEN, f)) { |
633 | for (p = buf, i = 0; p && i < 4; i++) | |
634 | p = strchr(p + 1, ' '); | |
2c6f3fc9 SH |
635 | if (!p) |
636 | continue; | |
b14fc100 | 637 | |
eab15c1e | 638 | p2 = strchr(p + 1, ' '); |
2c6f3fc9 SH |
639 | if (!p2) |
640 | continue; | |
b14fc100 | 641 | |
2c6f3fc9 | 642 | *p2 = '\0'; |
f6310f18 LT |
643 | if (strcmp(p + 1, path) == 0) { |
644 | /* This is the path. Is it shared? */ | |
eab15c1e | 645 | p = strchr(p2 + 1, ' '); |
2c6f3fc9 SH |
646 | if (p && strstr(p, "shared:")) { |
647 | fclose(f); | |
f6310f18 | 648 | return true; |
2c6f3fc9 SH |
649 | } |
650 | } | |
651 | } | |
b14fc100 | 652 | |
2c6f3fc9 | 653 | fclose(f); |
f6310f18 LT |
654 | return false; |
655 | } | |
656 | ||
657 | /* | |
658 | * Detect whether / is mounted MS_SHARED. The only way I know of to | |
659 | * check that is through /proc/self/mountinfo. | |
660 | * I'm only checking for /. If the container rootfs or mount location | |
661 | * is MS_SHARED, but not '/', then you're out of luck - figuring that | |
662 | * out would be too much work to be worth it. | |
663 | */ | |
664 | int detect_shared_rootfs(void) | |
665 | { | |
666 | if (is_shared_mountpoint("/")) | |
667 | return 1; | |
2c6f3fc9 SH |
668 | return 0; |
669 | } | |
0e6e3a41 | 670 | |
37ef15bb CB |
671 | bool switch_to_ns(pid_t pid, const char *ns) |
672 | { | |
51d0854c DY |
673 | int fd, ret; |
674 | char nspath[MAXPATHLEN]; | |
675 | ||
676 | /* Switch to new ns */ | |
677 | ret = snprintf(nspath, MAXPATHLEN, "/proc/%d/ns/%s", pid, ns); | |
678 | if (ret < 0 || ret >= MAXPATHLEN) | |
679 | return false; | |
680 | ||
681 | fd = open(nspath, O_RDONLY); | |
682 | if (fd < 0) { | |
a9cb0fb8 | 683 | SYSERROR("Failed to open %s", nspath); |
51d0854c DY |
684 | return false; |
685 | } | |
686 | ||
687 | ret = setns(fd, 0); | |
688 | if (ret) { | |
a9cb0fb8 | 689 | SYSERROR("Failed to set process %d to %s of %d.", pid, ns, fd); |
51d0854c DY |
690 | close(fd); |
691 | return false; | |
692 | } | |
b14fc100 | 693 | |
51d0854c DY |
694 | close(fd); |
695 | return true; | |
696 | } | |
697 | ||
b7f954bb SH |
698 | /* |
699 | * looking at fs/proc_namespace.c, it appears we can | |
700 | * actually expect the rootfs entry to very specifically contain | |
701 | * " - rootfs rootfs " | |
702 | * IIUC, so long as we've chrooted so that rootfs is not our root, | |
703 | * the rootfs entry should always be skipped in mountinfo contents. | |
704 | */ | |
fa454c8e | 705 | bool detect_ramfs_rootfs(void) |
b7f954bb | 706 | { |
b7f954bb | 707 | FILE *f; |
fa454c8e CB |
708 | char *p, *p2; |
709 | char *line = NULL; | |
710 | size_t len = 0; | |
b7f954bb | 711 | int i; |
b7f954bb SH |
712 | |
713 | f = fopen("/proc/self/mountinfo", "r"); | |
714 | if (!f) | |
fa454c8e CB |
715 | return false; |
716 | ||
717 | while (getline(&line, &len, f) != -1) { | |
718 | for (p = line, i = 0; p && i < 4; i++) | |
719 | p = strchr(p + 1, ' '); | |
b7f954bb SH |
720 | if (!p) |
721 | continue; | |
b14fc100 | 722 | |
fa454c8e | 723 | p2 = strchr(p + 1, ' '); |
b7f954bb SH |
724 | if (!p2) |
725 | continue; | |
b14fc100 | 726 | |
b7f954bb | 727 | *p2 = '\0'; |
fa454c8e | 728 | if (strcmp(p + 1, "/") == 0) { |
1a0e70ac | 729 | /* This is '/'. Is it the ramfs? */ |
fa454c8e | 730 | p = strchr(p2 + 1, '-'); |
b7f954bb | 731 | if (p && strncmp(p, "- rootfs rootfs ", 16) == 0) { |
fa454c8e | 732 | free(line); |
b7f954bb | 733 | fclose(f); |
8ce1abc2 | 734 | INFO("Rootfs is located on ramfs"); |
fa454c8e | 735 | return true; |
b7f954bb SH |
736 | } |
737 | } | |
738 | } | |
b14fc100 | 739 | |
fa454c8e | 740 | free(line); |
b7f954bb | 741 | fclose(f); |
fa454c8e | 742 | return false; |
b7f954bb SH |
743 | } |
744 | ||
37ef15bb CB |
745 | char *on_path(const char *cmd, const char *rootfs) |
746 | { | |
84c5549b | 747 | char *entry = NULL, *path = NULL; |
0e6e3a41 SG |
748 | char cmdpath[MAXPATHLEN]; |
749 | int ret; | |
750 | ||
751 | path = getenv("PATH"); | |
752 | if (!path) | |
8afb3e61 | 753 | return NULL; |
0e6e3a41 SG |
754 | |
755 | path = strdup(path); | |
756 | if (!path) | |
8afb3e61 | 757 | return NULL; |
0e6e3a41 | 758 | |
37ef15bb | 759 | lxc_iterate_parts (entry, path, ":") { |
9d9c111c | 760 | if (rootfs) |
37ef15bb CB |
761 | ret = snprintf(cmdpath, MAXPATHLEN, "%s/%s/%s", rootfs, |
762 | entry, cmd); | |
9d9c111c SH |
763 | else |
764 | ret = snprintf(cmdpath, MAXPATHLEN, "%s/%s", entry, cmd); | |
0e6e3a41 | 765 | if (ret < 0 || ret >= MAXPATHLEN) |
84c5549b | 766 | continue; |
0e6e3a41 SG |
767 | |
768 | if (access(cmdpath, X_OK) == 0) { | |
769 | free(path); | |
8afb3e61 | 770 | return strdup(cmdpath); |
0e6e3a41 | 771 | } |
0e6e3a41 SG |
772 | } |
773 | ||
774 | free(path); | |
8afb3e61 | 775 | return NULL; |
0e6e3a41 | 776 | } |
76a26f55 | 777 | |
12983ba4 SH |
778 | bool cgns_supported(void) |
779 | { | |
780 | return file_exists("/proc/self/ns/cgroup"); | |
781 | } | |
782 | ||
9d9c111c SH |
783 | /* historically lxc-init has been under /usr/lib/lxc and under |
784 | * /usr/lib/$ARCH/lxc. It now lives as $prefix/sbin/init.lxc. | |
785 | */ | |
786 | char *choose_init(const char *rootfs) | |
787 | { | |
788 | char *retv = NULL; | |
370ec268 SF |
789 | const char *empty = "", |
790 | *tmp; | |
9d9c111c | 791 | int ret, env_set = 0; |
9d9c111c SH |
792 | |
793 | if (!getenv("PATH")) { | |
794 | if (setenv("PATH", "/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", 0)) | |
795 | SYSERROR("Failed to setenv"); | |
b14fc100 | 796 | |
9d9c111c SH |
797 | env_set = 1; |
798 | } | |
799 | ||
800 | retv = on_path("init.lxc", rootfs); | |
801 | ||
802 | if (env_set) { | |
803 | if (unsetenv("PATH")) | |
804 | SYSERROR("Failed to unsetenv"); | |
805 | } | |
806 | ||
807 | if (retv) | |
808 | return retv; | |
809 | ||
810 | retv = malloc(PATH_MAX); | |
811 | if (!retv) | |
812 | return NULL; | |
813 | ||
814 | if (rootfs) | |
370ec268 | 815 | tmp = rootfs; |
9d9c111c | 816 | else |
370ec268 SF |
817 | tmp = empty; |
818 | ||
819 | ret = snprintf(retv, PATH_MAX, "%s/%s/%s", tmp, SBINDIR, "/init.lxc"); | |
9d9c111c SH |
820 | if (ret < 0 || ret >= PATH_MAX) { |
821 | ERROR("pathname too long"); | |
822 | goto out1; | |
823 | } | |
b14fc100 | 824 | |
e57cd7e9 | 825 | if (access(retv, X_OK) == 0) |
9d9c111c SH |
826 | return retv; |
827 | ||
370ec268 | 828 | ret = snprintf(retv, PATH_MAX, "%s/%s/%s", tmp, LXCINITDIR, "/lxc/lxc-init"); |
9d9c111c SH |
829 | if (ret < 0 || ret >= PATH_MAX) { |
830 | ERROR("pathname too long"); | |
831 | goto out1; | |
832 | } | |
b14fc100 | 833 | |
e57cd7e9 | 834 | if (access(retv, X_OK) == 0) |
9d9c111c SH |
835 | return retv; |
836 | ||
370ec268 | 837 | ret = snprintf(retv, PATH_MAX, "%s/usr/lib/lxc/lxc-init", tmp); |
9d9c111c SH |
838 | if (ret < 0 || ret >= PATH_MAX) { |
839 | ERROR("pathname too long"); | |
840 | goto out1; | |
841 | } | |
b14fc100 | 842 | |
e57cd7e9 | 843 | if (access(retv, X_OK) == 0) |
9d9c111c SH |
844 | return retv; |
845 | ||
370ec268 | 846 | ret = snprintf(retv, PATH_MAX, "%s/sbin/lxc-init", tmp); |
9d9c111c SH |
847 | if (ret < 0 || ret >= PATH_MAX) { |
848 | ERROR("pathname too long"); | |
849 | goto out1; | |
850 | } | |
b14fc100 | 851 | |
e57cd7e9 | 852 | if (access(retv, X_OK) == 0) |
9d9c111c SH |
853 | return retv; |
854 | ||
855 | /* | |
856 | * Last resort, look for the statically compiled init.lxc which we | |
857 | * hopefully bind-mounted in. | |
858 | * If we are called during container setup, and we get to this point, | |
859 | * then the init.lxc.static from the host will need to be bind-mounted | |
860 | * in. So we return NULL here to indicate that. | |
861 | */ | |
862 | if (rootfs) | |
863 | goto out1; | |
864 | ||
865 | ret = snprintf(retv, PATH_MAX, "/init.lxc.static"); | |
866 | if (ret < 0 || ret >= PATH_MAX) { | |
867 | WARN("Nonsense - name /lxc.init.static too long"); | |
868 | goto out1; | |
869 | } | |
b14fc100 | 870 | |
e57cd7e9 | 871 | if (access(retv, X_OK) == 0) |
9d9c111c SH |
872 | return retv; |
873 | ||
874 | out1: | |
875 | free(retv); | |
876 | return NULL; | |
877 | } | |
735f2c6e | 878 | |
6010a416 SG |
879 | /* |
880 | * Given the '-t' template option to lxc-create, figure out what to | |
881 | * do. If the template is a full executable path, use that. If it | |
882 | * is something like 'sshd', then return $templatepath/lxc-sshd. | |
883 | * On success return the template, on error return NULL. | |
884 | */ | |
885 | char *get_template_path(const char *t) | |
886 | { | |
887 | int ret, len; | |
888 | char *tpath; | |
889 | ||
890 | if (t[0] == '/' && access(t, X_OK) == 0) { | |
891 | tpath = strdup(t); | |
892 | return tpath; | |
893 | } | |
894 | ||
895 | len = strlen(LXCTEMPLATEDIR) + strlen(t) + strlen("/lxc-") + 1; | |
b14fc100 | 896 | |
6010a416 SG |
897 | tpath = malloc(len); |
898 | if (!tpath) | |
899 | return NULL; | |
b14fc100 | 900 | |
6010a416 SG |
901 | ret = snprintf(tpath, len, "%s/lxc-%s", LXCTEMPLATEDIR, t); |
902 | if (ret < 0 || ret >= len) { | |
903 | free(tpath); | |
904 | return NULL; | |
905 | } | |
b14fc100 | 906 | |
6010a416 SG |
907 | if (access(tpath, X_OK) < 0) { |
908 | SYSERROR("bad template: %s", t); | |
909 | free(tpath); | |
910 | return NULL; | |
911 | } | |
912 | ||
913 | return tpath; | |
914 | } | |
0a4be28d | 915 | |
592fd47a SH |
916 | /* |
917 | * @path: a pathname where / replaced with '\0'. | |
918 | * @offsetp: pointer to int showing which path segment was last seen. | |
919 | * Updated on return to reflect the next segment. | |
920 | * @fulllen: full original path length. | |
921 | * Returns a pointer to the next path segment, or NULL if done. | |
922 | */ | |
923 | static char *get_nextpath(char *path, int *offsetp, int fulllen) | |
924 | { | |
925 | int offset = *offsetp; | |
926 | ||
927 | if (offset >= fulllen) | |
928 | return NULL; | |
929 | ||
91d9cab6 | 930 | while (offset < fulllen && path[offset] != '\0') |
592fd47a | 931 | offset++; |
b14fc100 | 932 | |
91d9cab6 | 933 | while (offset < fulllen && path[offset] == '\0') |
592fd47a SH |
934 | offset++; |
935 | ||
936 | *offsetp = offset; | |
937 | return (offset < fulllen) ? &path[offset] : NULL; | |
938 | } | |
939 | ||
940 | /* | |
941 | * Check that @subdir is a subdir of @dir. @len is the length of | |
942 | * @dir (to avoid having to recalculate it). | |
943 | */ | |
944 | static bool is_subdir(const char *subdir, const char *dir, size_t len) | |
945 | { | |
946 | size_t subdirlen = strlen(subdir); | |
947 | ||
948 | if (subdirlen < len) | |
949 | return false; | |
b14fc100 | 950 | |
592fd47a SH |
951 | if (strncmp(subdir, dir, len) != 0) |
952 | return false; | |
b14fc100 | 953 | |
592fd47a SH |
954 | if (dir[len-1] == '/') |
955 | return true; | |
b14fc100 | 956 | |
592fd47a SH |
957 | if (subdir[len] == '/' || subdirlen == len) |
958 | return true; | |
b14fc100 | 959 | |
592fd47a SH |
960 | return false; |
961 | } | |
962 | ||
963 | /* | |
964 | * Check if the open fd is a symlink. Return -ELOOP if it is. Return | |
965 | * -ENOENT if we couldn't fstat. Return 0 if the fd is ok. | |
966 | */ | |
967 | static int check_symlink(int fd) | |
968 | { | |
969 | struct stat sb; | |
b14fc100 | 970 | int ret; |
971 | ||
972 | ret = fstat(fd, &sb); | |
592fd47a SH |
973 | if (ret < 0) |
974 | return -ENOENT; | |
b14fc100 | 975 | |
592fd47a SH |
976 | if (S_ISLNK(sb.st_mode)) |
977 | return -ELOOP; | |
b14fc100 | 978 | |
592fd47a SH |
979 | return 0; |
980 | } | |
981 | ||
982 | /* | |
983 | * Open a file or directory, provided that it contains no symlinks. | |
984 | * | |
985 | * CAVEAT: This function must not be used for other purposes than container | |
986 | * setup before executing the container's init | |
987 | */ | |
988 | static int open_if_safe(int dirfd, const char *nextpath) | |
989 | { | |
990 | int newfd = openat(dirfd, nextpath, O_RDONLY | O_NOFOLLOW); | |
1a0e70ac | 991 | if (newfd >= 0) /* Was not a symlink, all good. */ |
592fd47a SH |
992 | return newfd; |
993 | ||
994 | if (errno == ELOOP) | |
995 | return newfd; | |
996 | ||
997 | if (errno == EPERM || errno == EACCES) { | |
1a0e70ac CB |
998 | /* We're not root (cause we got EPERM) so try opening with |
999 | * O_PATH. | |
1000 | */ | |
592fd47a SH |
1001 | newfd = openat(dirfd, nextpath, O_PATH | O_NOFOLLOW); |
1002 | if (newfd >= 0) { | |
1a0e70ac CB |
1003 | /* O_PATH will return an fd for symlinks. We know |
1004 | * nextpath wasn't a symlink at last openat, so if fd is | |
1005 | * now a link, then something * fishy is going on. | |
592fd47a SH |
1006 | */ |
1007 | int ret = check_symlink(newfd); | |
1008 | if (ret < 0) { | |
1009 | close(newfd); | |
1010 | newfd = ret; | |
1011 | } | |
1012 | } | |
1013 | } | |
1014 | ||
1015 | return newfd; | |
1016 | } | |
1017 | ||
1018 | /* | |
1019 | * Open a path intending for mounting, ensuring that the final path | |
1020 | * is inside the container's rootfs. | |
1021 | * | |
1022 | * CAVEAT: This function must not be used for other purposes than container | |
1023 | * setup before executing the container's init | |
1024 | * | |
1025 | * @target: path to be opened | |
1026 | * @prefix_skip: a part of @target in which to ignore symbolic links. This | |
1027 | * would be the container's rootfs. | |
1028 | * | |
1029 | * Return an open fd for the path, or <0 on error. | |
1030 | */ | |
1031 | static int open_without_symlink(const char *target, const char *prefix_skip) | |
1032 | { | |
1033 | int curlen = 0, dirfd, fulllen, i; | |
1034 | char *dup = NULL; | |
1035 | ||
1036 | fulllen = strlen(target); | |
1037 | ||
1038 | /* make sure prefix-skip makes sense */ | |
01074e5b | 1039 | if (prefix_skip && strlen(prefix_skip) > 0) { |
592fd47a SH |
1040 | curlen = strlen(prefix_skip); |
1041 | if (!is_subdir(target, prefix_skip, curlen)) { | |
1042 | ERROR("WHOA there - target '%s' didn't start with prefix '%s'", | |
1043 | target, prefix_skip); | |
1044 | return -EINVAL; | |
1045 | } | |
b14fc100 | 1046 | |
592fd47a SH |
1047 | /* |
1048 | * get_nextpath() expects the curlen argument to be | |
1049 | * on a (turned into \0) / or before it, so decrement | |
1050 | * curlen to make sure that happens | |
1051 | */ | |
1052 | if (curlen) | |
1053 | curlen--; | |
1054 | } else { | |
1055 | prefix_skip = "/"; | |
1056 | curlen = 0; | |
1057 | } | |
1058 | ||
1059 | /* Make a copy of target which we can hack up, and tokenize it */ | |
1060 | if ((dup = strdup(target)) == NULL) { | |
1061 | SYSERROR("Out of memory checking for symbolic link"); | |
1062 | return -ENOMEM; | |
1063 | } | |
b14fc100 | 1064 | |
592fd47a SH |
1065 | for (i = 0; i < fulllen; i++) { |
1066 | if (dup[i] == '/') | |
1067 | dup[i] = '\0'; | |
1068 | } | |
1069 | ||
1070 | dirfd = open(prefix_skip, O_RDONLY); | |
1071 | if (dirfd < 0) | |
1072 | goto out; | |
b14fc100 | 1073 | |
592fd47a SH |
1074 | while (1) { |
1075 | int newfd, saved_errno; | |
1076 | char *nextpath; | |
1077 | ||
1078 | if ((nextpath = get_nextpath(dup, &curlen, fulllen)) == NULL) | |
1079 | goto out; | |
b14fc100 | 1080 | |
592fd47a SH |
1081 | newfd = open_if_safe(dirfd, nextpath); |
1082 | saved_errno = errno; | |
1083 | close(dirfd); | |
b14fc100 | 1084 | |
592fd47a SH |
1085 | dirfd = newfd; |
1086 | if (newfd < 0) { | |
1087 | errno = saved_errno; | |
1088 | if (errno == ELOOP) | |
1089 | SYSERROR("%s in %s was a symbolic link!", nextpath, target); | |
b14fc100 | 1090 | |
592fd47a SH |
1091 | goto out; |
1092 | } | |
1093 | } | |
1094 | ||
1095 | out: | |
1096 | free(dup); | |
1097 | return dirfd; | |
1098 | } | |
1099 | ||
1100 | /* | |
1101 | * Safely mount a path into a container, ensuring that the mount target | |
1102 | * is under the container's @rootfs. (If @rootfs is NULL, then the container | |
1103 | * uses the host's /) | |
1104 | * | |
1105 | * CAVEAT: This function must not be used for other purposes than container | |
1106 | * setup before executing the container's init | |
1107 | */ | |
1108 | int safe_mount(const char *src, const char *dest, const char *fstype, | |
1109 | unsigned long flags, const void *data, const char *rootfs) | |
1110 | { | |
1a0e70ac CB |
1111 | int destfd, ret, saved_errno; |
1112 | /* Only needs enough for /proc/self/fd/<fd>. */ | |
1113 | char srcbuf[50], destbuf[50]; | |
1114 | int srcfd = -1; | |
592fd47a SH |
1115 | const char *mntsrc = src; |
1116 | ||
1117 | if (!rootfs) | |
1118 | rootfs = ""; | |
1119 | ||
1120 | /* todo - allow symlinks for relative paths if 'allowsymlinks' option is passed */ | |
1121 | if (flags & MS_BIND && src && src[0] != '/') { | |
1122 | INFO("this is a relative bind mount"); | |
b14fc100 | 1123 | |
592fd47a SH |
1124 | srcfd = open_without_symlink(src, NULL); |
1125 | if (srcfd < 0) | |
1126 | return srcfd; | |
b14fc100 | 1127 | |
592fd47a SH |
1128 | ret = snprintf(srcbuf, 50, "/proc/self/fd/%d", srcfd); |
1129 | if (ret < 0 || ret > 50) { | |
1130 | close(srcfd); | |
1131 | ERROR("Out of memory"); | |
1132 | return -EINVAL; | |
1133 | } | |
1134 | mntsrc = srcbuf; | |
1135 | } | |
1136 | ||
1137 | destfd = open_without_symlink(dest, rootfs); | |
1138 | if (destfd < 0) { | |
88e078ba CB |
1139 | if (srcfd != -1) { |
1140 | saved_errno = errno; | |
592fd47a | 1141 | close(srcfd); |
88e078ba CB |
1142 | errno = saved_errno; |
1143 | } | |
b14fc100 | 1144 | |
592fd47a SH |
1145 | return destfd; |
1146 | } | |
1147 | ||
1148 | ret = snprintf(destbuf, 50, "/proc/self/fd/%d", destfd); | |
1149 | if (ret < 0 || ret > 50) { | |
1150 | if (srcfd != -1) | |
1151 | close(srcfd); | |
b14fc100 | 1152 | |
592fd47a SH |
1153 | close(destfd); |
1154 | ERROR("Out of memory"); | |
1155 | return -EINVAL; | |
1156 | } | |
1157 | ||
1158 | ret = mount(mntsrc, destbuf, fstype, flags, data); | |
1159 | saved_errno = errno; | |
1160 | if (srcfd != -1) | |
1161 | close(srcfd); | |
b14fc100 | 1162 | |
592fd47a SH |
1163 | close(destfd); |
1164 | if (ret < 0) { | |
1165 | errno = saved_errno; | |
0103eb53 | 1166 | SYSERROR("Failed to mount %s onto %s", src ? src : "(null)", dest); |
592fd47a SH |
1167 | return ret; |
1168 | } | |
1169 | ||
1170 | return 0; | |
1171 | } | |
1172 | ||
ced03a01 SH |
1173 | /* |
1174 | * Mount a proc under @rootfs if proc self points to a pid other than | |
1175 | * my own. This is needed to have a known-good proc mount for setting | |
1176 | * up LSMs both at container startup and attach. | |
1177 | * | |
1178 | * @rootfs : the rootfs where proc should be mounted | |
1179 | * | |
1180 | * Returns < 0 on failure, 0 if the correct proc was already mounted | |
1181 | * and 1 if a new proc was mounted. | |
f267d666 BP |
1182 | * |
1183 | * NOTE: not to be called from inside the container namespace! | |
ced03a01 | 1184 | */ |
943144d9 | 1185 | int lxc_mount_proc_if_needed(const char *rootfs) |
ced03a01 SH |
1186 | { |
1187 | char path[MAXPATHLEN]; | |
6b1ba5d6 | 1188 | int link_to_pid, linklen, mypid, ret; |
40464e8a | 1189 | char link[INTTYPE_TO_STRLEN(pid_t)] = {0}; |
ced03a01 SH |
1190 | |
1191 | ret = snprintf(path, MAXPATHLEN, "%s/proc/self", rootfs); | |
1192 | if (ret < 0 || ret >= MAXPATHLEN) { | |
1193 | SYSERROR("proc path name too long"); | |
1194 | return -1; | |
1195 | } | |
fc2ad9dc | 1196 | |
979a0d93 | 1197 | linklen = readlink(path, link, sizeof(link)); |
fc2ad9dc | 1198 | |
ced03a01 | 1199 | ret = snprintf(path, MAXPATHLEN, "%s/proc", rootfs); |
d539a2b2 CB |
1200 | if (ret < 0 || ret >= MAXPATHLEN) { |
1201 | SYSERROR("proc path name too long"); | |
1202 | return -1; | |
1203 | } | |
fc2ad9dc CB |
1204 | |
1205 | /* /proc not mounted */ | |
1206 | if (linklen < 0) { | |
1207 | if (mkdir(path, 0755) && errno != EEXIST) | |
1208 | return -1; | |
b14fc100 | 1209 | |
ced03a01 | 1210 | goto domount; |
979a0d93 | 1211 | } else if (linklen >= sizeof(link)) { |
6b1ba5d6 CB |
1212 | link[linklen - 1] = '\0'; |
1213 | ERROR("readlink returned truncated content: \"%s\"", link); | |
1214 | return -1; | |
fc2ad9dc CB |
1215 | } |
1216 | ||
0059379f | 1217 | mypid = lxc_raw_getpid(); |
6b1ba5d6 CB |
1218 | INFO("I am %d, /proc/self points to \"%s\"", mypid, link); |
1219 | ||
2d036cca CB |
1220 | if (lxc_safe_int(link, &link_to_pid) < 0) |
1221 | return -1; | |
fc2ad9dc | 1222 | |
6b1ba5d6 CB |
1223 | /* correct procfs is already mounted */ |
1224 | if (link_to_pid == mypid) | |
1225 | return 0; | |
fc2ad9dc | 1226 | |
6b1ba5d6 CB |
1227 | ret = umount2(path, MNT_DETACH); |
1228 | if (ret < 0) | |
1229 | WARN("failed to umount \"%s\" with MNT_DETACH", path); | |
ced03a01 SH |
1230 | |
1231 | domount: | |
fc2ad9dc | 1232 | /* rootfs is NULL */ |
6b1ba5d6 | 1233 | if (!strcmp(rootfs, "")) |
f267d666 BP |
1234 | ret = mount("proc", path, "proc", 0, NULL); |
1235 | else | |
1236 | ret = safe_mount("proc", path, "proc", 0, NULL, rootfs); | |
f267d666 | 1237 | if (ret < 0) |
ced03a01 | 1238 | return -1; |
f267d666 | 1239 | |
fc2ad9dc | 1240 | INFO("mounted /proc in container for security transition"); |
ced03a01 SH |
1241 | return 1; |
1242 | } | |
69aeabac | 1243 | |
f8dd0275 | 1244 | int open_devnull(void) |
69aeabac | 1245 | { |
f8dd0275 AM |
1246 | int fd = open("/dev/null", O_RDWR); |
1247 | ||
1248 | if (fd < 0) | |
1249 | SYSERROR("Can't open /dev/null"); | |
1250 | ||
1251 | return fd; | |
1252 | } | |
69aeabac | 1253 | |
f8dd0275 AM |
1254 | int set_stdfds(int fd) |
1255 | { | |
bbbf65ee CB |
1256 | int ret; |
1257 | ||
69aeabac TA |
1258 | if (fd < 0) |
1259 | return -1; | |
1260 | ||
bbbf65ee CB |
1261 | ret = dup2(fd, STDIN_FILENO); |
1262 | if (ret < 0) | |
f8dd0275 | 1263 | return -1; |
bbbf65ee CB |
1264 | |
1265 | ret = dup2(fd, STDOUT_FILENO); | |
1266 | if (ret < 0) | |
f8dd0275 | 1267 | return -1; |
bbbf65ee CB |
1268 | |
1269 | ret = dup2(fd, STDERR_FILENO); | |
1270 | if (ret < 0) | |
f8dd0275 AM |
1271 | return -1; |
1272 | ||
1273 | return 0; | |
1274 | } | |
1275 | ||
1276 | int null_stdfds(void) | |
1277 | { | |
1278 | int ret = -1; | |
b14fc100 | 1279 | int fd; |
f8dd0275 | 1280 | |
b14fc100 | 1281 | fd = open_devnull(); |
f8dd0275 AM |
1282 | if (fd >= 0) { |
1283 | ret = set_stdfds(fd); | |
1284 | close(fd); | |
1285 | } | |
69aeabac | 1286 | |
69aeabac TA |
1287 | return ret; |
1288 | } | |
ccb4cabe | 1289 | |
330ae3d3 | 1290 | /* Check whether a signal is blocked by a process. */ |
de3c491b | 1291 | /* /proc/pid-to-str/status\0 = (5 + 21 + 7 + 1) */ |
40464e8a | 1292 | #define __PROC_STATUS_LEN (6 + INTTYPE_TO_STRLEN(pid_t) + 7 + 1) |
573ad77f | 1293 | bool task_blocks_signal(pid_t pid, int signal) |
330ae3d3 | 1294 | { |
330ae3d3 | 1295 | int ret; |
de3c491b | 1296 | char status[__PROC_STATUS_LEN]; |
eabf1ea9 | 1297 | FILE *f; |
573ad77f | 1298 | uint64_t sigblk = 0, one = 1; |
eabf1ea9 CB |
1299 | size_t n = 0; |
1300 | bool bret = false; | |
1301 | char *line = NULL; | |
330ae3d3 | 1302 | |
de3c491b CB |
1303 | ret = snprintf(status, __PROC_STATUS_LEN, "/proc/%d/status", pid); |
1304 | if (ret < 0 || ret >= __PROC_STATUS_LEN) | |
330ae3d3 CB |
1305 | return bret; |
1306 | ||
1307 | f = fopen(status, "r"); | |
1308 | if (!f) | |
1309 | return bret; | |
1310 | ||
1311 | while (getline(&line, &n, f) != -1) { | |
573ad77f CB |
1312 | char *numstr; |
1313 | ||
eabf1ea9 | 1314 | if (strncmp(line, "SigBlk:", 7)) |
6fbcbe3b CB |
1315 | continue; |
1316 | ||
573ad77f CB |
1317 | numstr = lxc_trim_whitespace_in_place(line + 7); |
1318 | ret = lxc_safe_uint64(numstr, &sigblk, 16); | |
1319 | if (ret < 0) | |
6fbcbe3b | 1320 | goto out; |
573ad77f CB |
1321 | |
1322 | break; | |
330ae3d3 CB |
1323 | } |
1324 | ||
573ad77f | 1325 | if (sigblk & (one << (signal - 1))) |
330ae3d3 CB |
1326 | bret = true; |
1327 | ||
1328 | out: | |
1329 | free(line); | |
1330 | fclose(f); | |
1331 | return bret; | |
1332 | } | |
000dfda7 | 1333 | |
a687256f CB |
1334 | int lxc_preserve_ns(const int pid, const char *ns) |
1335 | { | |
1336 | int ret; | |
a052913d CB |
1337 | /* 5 /proc + 21 /int_as_str + 3 /ns + 20 /NS_NAME + 1 \0 */ |
1338 | #define __NS_PATH_LEN 50 | |
1339 | char path[__NS_PATH_LEN]; | |
a687256f | 1340 | |
4d8ac866 CB |
1341 | /* This way we can use this function to also check whether namespaces |
1342 | * are supported by the kernel by passing in the NULL or the empty | |
1343 | * string. | |
1344 | */ | |
a052913d | 1345 | ret = snprintf(path, __NS_PATH_LEN, "/proc/%d/ns%s%s", pid, |
4d8ac866 CB |
1346 | !ns || strcmp(ns, "") == 0 ? "" : "/", |
1347 | !ns || strcmp(ns, "") == 0 ? "" : ns); | |
ea918412 | 1348 | if (ret < 0 || (size_t)ret >= __NS_PATH_LEN) { |
1349 | errno = EFBIG; | |
1350 | return -1; | |
1351 | } | |
a687256f CB |
1352 | |
1353 | return open(path, O_RDONLY | O_CLOEXEC); | |
1354 | } | |
6bc2eafe | 1355 | |
464c4611 | 1356 | bool lxc_switch_uid_gid(uid_t uid, gid_t gid) |
dbaf55a3 | 1357 | { |
db2d1af1 CB |
1358 | int ret = 0; |
1359 | ||
1360 | if (gid != LXC_INVALID_GID) { | |
1361 | ret = setgid(gid); | |
1362 | if (ret < 0) { | |
1363 | SYSERROR("Failed to switch to gid %d", gid); | |
464c4611 | 1364 | return false; |
db2d1af1 CB |
1365 | } |
1366 | NOTICE("Switched to gid %d", gid); | |
dbaf55a3 | 1367 | } |
dbaf55a3 | 1368 | |
db2d1af1 CB |
1369 | if (uid != LXC_INVALID_UID) { |
1370 | ret = setuid(uid); | |
1371 | if (ret < 0) { | |
1372 | SYSERROR("Failed to switch to uid %d", uid); | |
464c4611 | 1373 | return false; |
db2d1af1 CB |
1374 | } |
1375 | NOTICE("Switched to uid %d", uid); | |
dbaf55a3 | 1376 | } |
dbaf55a3 | 1377 | |
464c4611 | 1378 | return true; |
dbaf55a3 CB |
1379 | } |
1380 | ||
1381 | /* Simple covenience function which enables uniform logging. */ | |
8af07f82 | 1382 | bool lxc_setgroups(int size, gid_t list[]) |
dbaf55a3 CB |
1383 | { |
1384 | if (setgroups(size, list) < 0) { | |
8af07f82 CB |
1385 | SYSERROR("Failed to setgroups()"); |
1386 | return false; | |
dbaf55a3 | 1387 | } |
8af07f82 | 1388 | NOTICE("Dropped additional groups"); |
dbaf55a3 | 1389 | |
8af07f82 | 1390 | return true; |
dbaf55a3 | 1391 | } |
c6868a1f CB |
1392 | |
1393 | static int lxc_get_unused_loop_dev_legacy(char *loop_name) | |
1394 | { | |
1395 | struct dirent *dp; | |
1396 | struct loop_info64 lo64; | |
1397 | DIR *dir; | |
1398 | int dfd = -1, fd = -1, ret = -1; | |
1399 | ||
1400 | dir = opendir("/dev"); | |
1401 | if (!dir) | |
1402 | return -1; | |
1403 | ||
1404 | while ((dp = readdir(dir))) { | |
c6868a1f CB |
1405 | if (strncmp(dp->d_name, "loop", 4) != 0) |
1406 | continue; | |
1407 | ||
1408 | dfd = dirfd(dir); | |
1409 | if (dfd < 0) | |
1410 | continue; | |
1411 | ||
1412 | fd = openat(dfd, dp->d_name, O_RDWR); | |
1413 | if (fd < 0) | |
1414 | continue; | |
1415 | ||
1416 | ret = ioctl(fd, LOOP_GET_STATUS64, &lo64); | |
1417 | if (ret < 0) { | |
1418 | if (ioctl(fd, LOOP_GET_STATUS64, &lo64) == 0 || | |
1419 | errno != ENXIO) { | |
1420 | close(fd); | |
1421 | fd = -1; | |
1422 | continue; | |
1423 | } | |
1424 | } | |
1425 | ||
1426 | ret = snprintf(loop_name, LO_NAME_SIZE, "/dev/%s", dp->d_name); | |
1427 | if (ret < 0 || ret >= LO_NAME_SIZE) { | |
1428 | close(fd); | |
1429 | fd = -1; | |
1430 | continue; | |
1431 | } | |
1432 | ||
1433 | break; | |
1434 | } | |
1435 | ||
1436 | closedir(dir); | |
1437 | ||
1438 | if (fd < 0) | |
1439 | return -1; | |
1440 | ||
1441 | return fd; | |
1442 | } | |
1443 | ||
1444 | static int lxc_get_unused_loop_dev(char *name_loop) | |
1445 | { | |
1446 | int loop_nr, ret; | |
1447 | int fd_ctl = -1, fd_tmp = -1; | |
1448 | ||
1449 | fd_ctl = open("/dev/loop-control", O_RDWR | O_CLOEXEC); | |
1450 | if (fd_ctl < 0) | |
1451 | return -ENODEV; | |
1452 | ||
1453 | loop_nr = ioctl(fd_ctl, LOOP_CTL_GET_FREE); | |
1454 | if (loop_nr < 0) | |
1455 | goto on_error; | |
1456 | ||
1457 | ret = snprintf(name_loop, LO_NAME_SIZE, "/dev/loop%d", loop_nr); | |
1458 | if (ret < 0 || ret >= LO_NAME_SIZE) | |
1459 | goto on_error; | |
1460 | ||
1461 | fd_tmp = open(name_loop, O_RDWR | O_CLOEXEC); | |
1462 | if (fd_tmp < 0) | |
1463 | goto on_error; | |
1464 | ||
1465 | on_error: | |
1466 | close(fd_ctl); | |
1467 | return fd_tmp; | |
1468 | } | |
1469 | ||
1470 | int lxc_prepare_loop_dev(const char *source, char *loop_dev, int flags) | |
1471 | { | |
1472 | int ret; | |
1473 | struct loop_info64 lo64; | |
1474 | int fd_img = -1, fret = -1, fd_loop = -1; | |
1475 | ||
1476 | fd_loop = lxc_get_unused_loop_dev(loop_dev); | |
1477 | if (fd_loop < 0) { | |
1478 | if (fd_loop == -ENODEV) | |
1479 | fd_loop = lxc_get_unused_loop_dev_legacy(loop_dev); | |
1480 | else | |
1481 | goto on_error; | |
1482 | } | |
1483 | ||
1484 | fd_img = open(source, O_RDWR | O_CLOEXEC); | |
1485 | if (fd_img < 0) | |
1486 | goto on_error; | |
1487 | ||
1488 | ret = ioctl(fd_loop, LOOP_SET_FD, fd_img); | |
1489 | if (ret < 0) | |
1490 | goto on_error; | |
1491 | ||
1492 | memset(&lo64, 0, sizeof(lo64)); | |
1493 | lo64.lo_flags = flags; | |
1494 | ||
1495 | ret = ioctl(fd_loop, LOOP_SET_STATUS64, &lo64); | |
1496 | if (ret < 0) | |
1497 | goto on_error; | |
1498 | ||
1499 | fret = 0; | |
1500 | ||
1501 | on_error: | |
1502 | if (fd_img >= 0) | |
1503 | close(fd_img); | |
1504 | ||
1505 | if (fret < 0 && fd_loop >= 0) { | |
1506 | close(fd_loop); | |
1507 | fd_loop = -1; | |
1508 | } | |
1509 | ||
1510 | return fd_loop; | |
1511 | } | |
74251e49 CB |
1512 | |
1513 | int lxc_unstack_mountpoint(const char *path, bool lazy) | |
1514 | { | |
1515 | int ret; | |
1516 | int umounts = 0; | |
1517 | ||
1518 | pop_stack: | |
1519 | ret = umount2(path, lazy ? MNT_DETACH : 0); | |
1520 | if (ret < 0) { | |
1521 | /* We consider anything else than EINVAL deadly to prevent going | |
1522 | * into an infinite loop. (The other alternative is constantly | |
1523 | * parsing /proc/self/mountinfo which is yucky and probably | |
1524 | * racy.) | |
1525 | */ | |
1526 | if (errno != EINVAL) | |
1527 | return -errno; | |
1528 | } else { | |
b4a40f7b CB |
1529 | /* Just stop counting when this happens. That'd just be so |
1530 | * stupid that we won't even bother trying to report back the | |
1531 | * correct value anymore. | |
1532 | */ | |
1533 | if (umounts != INT_MAX) | |
1534 | umounts++; | |
b14fc100 | 1535 | |
74251e49 CB |
1536 | /* We succeeded in umounting. Make sure that there's no other |
1537 | * mountpoint stacked underneath. | |
1538 | */ | |
74251e49 CB |
1539 | goto pop_stack; |
1540 | } | |
1541 | ||
1542 | return umounts; | |
1543 | } | |
ea3a694f CB |
1544 | |
1545 | int run_command(char *buf, size_t buf_size, int (*child_fn)(void *), void *args) | |
1546 | { | |
1547 | pid_t child; | |
1548 | int ret, fret, pipefd[2]; | |
1549 | ssize_t bytes; | |
1550 | ||
46210729 | 1551 | /* Make sure our callers do not receive uninitialized memory. */ |
ea3a694f CB |
1552 | if (buf_size > 0 && buf) |
1553 | buf[0] = '\0'; | |
1554 | ||
1555 | if (pipe(pipefd) < 0) { | |
1556 | SYSERROR("failed to create pipe"); | |
1557 | return -1; | |
1558 | } | |
1559 | ||
2d728b2f | 1560 | child = lxc_raw_clone(0); |
ea3a694f CB |
1561 | if (child < 0) { |
1562 | close(pipefd[0]); | |
1563 | close(pipefd[1]); | |
1564 | SYSERROR("failed to create new process"); | |
1565 | return -1; | |
1566 | } | |
1567 | ||
1568 | if (child == 0) { | |
1569 | /* Close the read-end of the pipe. */ | |
1570 | close(pipefd[0]); | |
1571 | ||
1572 | /* Redirect std{err,out} to write-end of the | |
1573 | * pipe. | |
1574 | */ | |
1575 | ret = dup2(pipefd[1], STDOUT_FILENO); | |
1576 | if (ret >= 0) | |
1577 | ret = dup2(pipefd[1], STDERR_FILENO); | |
1578 | ||
1579 | /* Close the write-end of the pipe. */ | |
1580 | close(pipefd[1]); | |
1581 | ||
1582 | if (ret < 0) { | |
1583 | SYSERROR("failed to duplicate std{err,out} file descriptor"); | |
d8b3f9c3 | 1584 | _exit(EXIT_FAILURE); |
ea3a694f CB |
1585 | } |
1586 | ||
1587 | /* Does not return. */ | |
1588 | child_fn(args); | |
1589 | ERROR("failed to exec command"); | |
d8b3f9c3 | 1590 | _exit(EXIT_FAILURE); |
ea3a694f CB |
1591 | } |
1592 | ||
1593 | /* close the write-end of the pipe */ | |
1594 | close(pipefd[1]); | |
1595 | ||
7a643c7c | 1596 | if (buf && buf_size > 0) { |
a5bc6cb0 | 1597 | bytes = lxc_read_nointr(pipefd[0], buf, buf_size - 1); |
7a643c7c CB |
1598 | if (bytes > 0) |
1599 | buf[bytes - 1] = '\0'; | |
1600 | } | |
ea3a694f CB |
1601 | |
1602 | fret = wait_for_pid(child); | |
1603 | /* close the read-end of the pipe */ | |
1604 | close(pipefd[0]); | |
1605 | ||
1606 | return fret; | |
1607 | } | |
04ad7ffe | 1608 | |
d75c14e2 CB |
1609 | bool lxc_nic_exists(char *nic) |
1610 | { | |
1611 | #define __LXC_SYS_CLASS_NET_LEN 15 + IFNAMSIZ + 1 | |
1612 | char path[__LXC_SYS_CLASS_NET_LEN]; | |
1613 | int ret; | |
1614 | struct stat sb; | |
1615 | ||
1616 | if (!strcmp(nic, "none")) | |
1617 | return true; | |
1618 | ||
1619 | ret = snprintf(path, __LXC_SYS_CLASS_NET_LEN, "/sys/class/net/%s", nic); | |
1620 | if (ret < 0 || (size_t)ret >= __LXC_SYS_CLASS_NET_LEN) | |
1621 | return false; | |
1622 | ||
1623 | ret = stat(path, &sb); | |
1624 | if (ret < 0) | |
1625 | return false; | |
1626 | ||
1627 | return true; | |
1628 | } | |
127c6e70 | 1629 | |
6222c3f4 CB |
1630 | uint64_t lxc_find_next_power2(uint64_t n) |
1631 | { | |
1632 | /* 0 is not valid input. We return 0 to the caller since 0 is not a | |
1633 | * valid power of two. | |
1634 | */ | |
1635 | if (n == 0) | |
1636 | return 0; | |
1637 | ||
1638 | if (!(n & (n - 1))) | |
1639 | return n; | |
1640 | ||
1641 | while (n & (n - 1)) | |
1642 | n = n & (n - 1); | |
1643 | ||
1644 | n = n << 1; | |
1645 | return n; | |
1646 | } | |
1fd0f41e CB |
1647 | |
1648 | int lxc_set_death_signal(int signal) | |
1649 | { | |
1650 | int ret; | |
1651 | pid_t ppid; | |
1652 | ||
b81689a1 CB |
1653 | ret = prctl(PR_SET_PDEATHSIG, prctl_arg(signal), prctl_arg(0), |
1654 | prctl_arg(0), prctl_arg(0)); | |
1fd0f41e CB |
1655 | |
1656 | /* Check whether we have been orphaned. */ | |
1657 | ppid = (pid_t)syscall(SYS_getppid); | |
1658 | if (ppid == 1) { | |
1659 | pid_t self; | |
1660 | ||
1661 | self = lxc_raw_getpid(); | |
1662 | ret = kill(self, SIGKILL); | |
1663 | if (ret < 0) | |
1664 | return -1; | |
1665 | } | |
1666 | ||
1667 | if (ret < 0) { | |
1668 | SYSERROR("Failed to set PR_SET_PDEATHSIG to %d", signal); | |
1669 | return -1; | |
1670 | } | |
1671 | ||
1672 | return 0; | |
1673 | } | |
7ad37670 | 1674 | |
a9d4ebc1 CB |
1675 | int fd_cloexec(int fd, bool cloexec) |
1676 | { | |
1677 | int oflags, nflags; | |
1678 | ||
1679 | oflags = fcntl(fd, F_GETFD, 0); | |
1680 | if (oflags < 0) | |
1681 | return -errno; | |
1682 | ||
1683 | if (cloexec) | |
1684 | nflags = oflags | FD_CLOEXEC; | |
1685 | else | |
1686 | nflags = oflags & ~FD_CLOEXEC; | |
1687 | ||
1688 | if (nflags == oflags) | |
1689 | return 0; | |
1690 | ||
1691 | if (fcntl(fd, F_SETFD, nflags) < 0) | |
1692 | return -errno; | |
1693 | ||
1694 | return 0; | |
1695 | } | |
d7ab0375 | 1696 | |
1697 | int recursive_destroy(char *dirname) | |
1698 | { | |
1699 | int ret; | |
1700 | struct dirent *direntp; | |
1701 | DIR *dir; | |
1702 | int r = 0; | |
1703 | ||
1704 | dir = opendir(dirname); | |
1705 | if (!dir) | |
1706 | return -1; | |
1707 | ||
1708 | while ((direntp = readdir(dir))) { | |
1709 | char *pathname; | |
1710 | struct stat mystat; | |
1711 | ||
1712 | if (!strcmp(direntp->d_name, ".") || | |
1713 | !strcmp(direntp->d_name, "..")) | |
1714 | continue; | |
1715 | ||
1716 | pathname = must_make_path(dirname, direntp->d_name, NULL); | |
1717 | ||
1718 | ret = lstat(pathname, &mystat); | |
1719 | if (ret < 0) { | |
1720 | if (!r) | |
1721 | WARN("Failed to stat \"%s\"", pathname); | |
1722 | ||
1723 | r = -1; | |
1724 | goto next; | |
1725 | } | |
1726 | ||
1727 | if (!S_ISDIR(mystat.st_mode)) | |
1728 | goto next; | |
1729 | ||
1730 | ret = recursive_destroy(pathname); | |
1731 | if (ret < 0) | |
1732 | r = -1; | |
1733 | ||
1734 | next: | |
1735 | free(pathname); | |
1736 | } | |
1737 | ||
1738 | ret = rmdir(dirname); | |
1739 | if (ret < 0) { | |
1740 | if (!r) | |
1741 | SYSWARN("Failed to delete \"%s\"", dirname); | |
1742 | ||
1743 | r = -1; | |
1744 | } | |
1745 | ||
1746 | ret = closedir(dir); | |
1747 | if (ret < 0) { | |
1748 | if (!r) | |
1749 | SYSWARN("Failed to delete \"%s\"", dirname); | |
1750 | ||
1751 | r = -1; | |
1752 | } | |
1753 | ||
1754 | return r; | |
1755 | } |