]>
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 | ||
052616eb ÇO |
24 | #include "config.h" |
25 | ||
7935833c | 26 | #define __STDC_FORMAT_MACROS /* Required for PRIu64 to work. */ |
643c1984 | 27 | #include <ctype.h> |
a1e5280d | 28 | #include <dirent.h> |
e3642c43 | 29 | #include <errno.h> |
a1e5280d | 30 | #include <fcntl.h> |
dbaf55a3 | 31 | #include <grp.h> |
7935833c | 32 | #include <inttypes.h> |
a1e5280d | 33 | #include <libgen.h> |
d983b93c | 34 | #include <stddef.h> |
a1e5280d CB |
35 | #include <stdio.h> |
36 | #include <stdlib.h> | |
61a1d519 | 37 | #include <string.h> |
981f6029 | 38 | #include <unistd.h> |
e3642c43 | 39 | #include <sys/mman.h> |
6e4bb2e0 | 40 | #include <sys/mount.h> |
a1e5280d CB |
41 | #include <sys/param.h> |
42 | #include <sys/prctl.h> | |
43 | #include <sys/stat.h> | |
9be53773 SH |
44 | #include <sys/types.h> |
45 | #include <sys/wait.h> | |
e3642c43 DL |
46 | |
47 | #include "log.h" | |
025ed0f3 | 48 | #include "lxclock.h" |
51d0854c | 49 | #include "namespace.h" |
981f6029 | 50 | #include "utils.h" |
e3642c43 | 51 | |
4928c718 SG |
52 | #ifndef O_PATH |
53 | #define O_PATH 010000000 | |
54 | #endif | |
55 | ||
56 | #ifndef O_NOFOLLOW | |
57 | #define O_NOFOLLOW 00400000 | |
58 | #endif | |
59 | ||
e3642c43 DL |
60 | lxc_log_define(lxc_utils, lxc); |
61 | ||
4295c5de SH |
62 | /* |
63 | * if path is btrfs, tries to remove it and any subvolumes beneath it | |
64 | */ | |
65 | extern bool btrfs_try_remove_subvol(const char *path); | |
66 | ||
0cc417b2 SH |
67 | static int _recursive_rmdir(char *dirname, dev_t pdev, |
68 | const char *exclude, int level, bool onedev) | |
60bf62d4 | 69 | { |
74f96976 | 70 | struct dirent *direntp; |
60bf62d4 SH |
71 | DIR *dir; |
72 | int ret, failed=0; | |
73 | char pathname[MAXPATHLEN]; | |
18aa217b | 74 | bool hadexclude = false; |
60bf62d4 SH |
75 | |
76 | dir = opendir(dirname); | |
77 | if (!dir) { | |
b103ceac | 78 | ERROR("failed to open %s", dirname); |
4355ab5f | 79 | return -1; |
60bf62d4 SH |
80 | } |
81 | ||
74f96976 | 82 | while ((direntp = readdir(dir))) { |
60bf62d4 SH |
83 | struct stat mystat; |
84 | int rc; | |
85 | ||
86 | if (!direntp) | |
87 | break; | |
88 | ||
89 | if (!strcmp(direntp->d_name, ".") || | |
90 | !strcmp(direntp->d_name, "..")) | |
91 | continue; | |
92 | ||
93 | rc = snprintf(pathname, MAXPATHLEN, "%s/%s", dirname, direntp->d_name); | |
94 | if (rc < 0 || rc >= MAXPATHLEN) { | |
95 | ERROR("pathname too long"); | |
96 | failed=1; | |
97 | continue; | |
98 | } | |
18aa217b SH |
99 | |
100 | if (!level && exclude && !strcmp(direntp->d_name, exclude)) { | |
101 | ret = rmdir(pathname); | |
102 | if (ret < 0) { | |
103 | switch(errno) { | |
104 | case ENOTEMPTY: | |
0cc417b2 | 105 | INFO("Not deleting snapshot %s", pathname); |
18aa217b SH |
106 | hadexclude = true; |
107 | break; | |
108 | case ENOTDIR: | |
109 | ret = unlink(pathname); | |
110 | if (ret) | |
b103ceac | 111 | INFO("Failed to remove %s", pathname); |
18aa217b SH |
112 | break; |
113 | default: | |
b103ceac | 114 | SYSERROR("Failed to rmdir %s", pathname); |
18aa217b SH |
115 | failed = 1; |
116 | break; | |
117 | } | |
118 | } | |
119 | continue; | |
120 | } | |
121 | ||
60bf62d4 SH |
122 | ret = lstat(pathname, &mystat); |
123 | if (ret) { | |
b103ceac | 124 | ERROR("Failed to stat %s", pathname); |
4295c5de | 125 | failed = 1; |
60bf62d4 SH |
126 | continue; |
127 | } | |
4295c5de SH |
128 | if (onedev && mystat.st_dev != pdev) { |
129 | /* TODO should we be checking /proc/self/mountinfo for | |
130 | * pathname and not doing this if found? */ | |
131 | if (btrfs_try_remove_subvol(pathname)) | |
132 | INFO("Removed btrfs subvolume at %s\n", pathname); | |
60bf62d4 | 133 | continue; |
4295c5de | 134 | } |
60bf62d4 | 135 | if (S_ISDIR(mystat.st_mode)) { |
0cc417b2 | 136 | if (_recursive_rmdir(pathname, pdev, exclude, level+1, onedev) < 0) |
60bf62d4 SH |
137 | failed=1; |
138 | } else { | |
139 | if (unlink(pathname) < 0) { | |
b103ceac | 140 | SYSERROR("Failed to delete %s", pathname); |
60bf62d4 SH |
141 | failed=1; |
142 | } | |
143 | } | |
144 | } | |
145 | ||
4295c5de | 146 | if (rmdir(dirname) < 0 && !btrfs_try_remove_subvol(dirname) && !hadexclude) { |
b103ceac | 147 | ERROR("Failed to delete %s", dirname); |
4295c5de | 148 | failed=1; |
60bf62d4 SH |
149 | } |
150 | ||
025ed0f3 | 151 | ret = closedir(dir); |
025ed0f3 | 152 | if (ret) { |
b103ceac | 153 | ERROR("Failed to close directory %s", dirname); |
60bf62d4 SH |
154 | failed=1; |
155 | } | |
156 | ||
4355ab5f | 157 | return failed ? -1 : 0; |
60bf62d4 SH |
158 | } |
159 | ||
29a11a7f CB |
160 | /* We have two different magic values for overlayfs, yay. */ |
161 | #ifndef OVERLAYFS_SUPER_MAGIC | |
0cc417b2 | 162 | #define OVERLAYFS_SUPER_MAGIC 0x794c764f |
29a11a7f CB |
163 | #endif |
164 | ||
165 | #ifndef OVERLAY_SUPER_MAGIC | |
0cc417b2 | 166 | #define OVERLAY_SUPER_MAGIC 0x794c7630 |
29a11a7f CB |
167 | #endif |
168 | ||
169 | /* In overlayfs, st_dev is unreliable. So on overlayfs we don't do the | |
170 | * lxc_rmdir_onedev() | |
0cc417b2 SH |
171 | */ |
172 | static bool is_native_overlayfs(const char *path) | |
173 | { | |
29a11a7f CB |
174 | if (has_fs_type(path, OVERLAY_SUPER_MAGIC) || |
175 | has_fs_type(path, OVERLAYFS_SUPER_MAGIC)) | |
0cc417b2 | 176 | return true; |
29a11a7f | 177 | |
0cc417b2 SH |
178 | return false; |
179 | } | |
180 | ||
4355ab5f | 181 | /* returns 0 on success, -1 if there were any failures */ |
18aa217b | 182 | extern int lxc_rmdir_onedev(char *path, const char *exclude) |
60bf62d4 SH |
183 | { |
184 | struct stat mystat; | |
0cc417b2 SH |
185 | bool onedev = true; |
186 | ||
187 | if (is_native_overlayfs(path)) { | |
188 | onedev = false; | |
189 | } | |
60bf62d4 SH |
190 | |
191 | if (lstat(path, &mystat) < 0) { | |
067650d0 SH |
192 | if (errno == ENOENT) |
193 | return 0; | |
b103ceac | 194 | ERROR("Failed to stat %s", path); |
4355ab5f | 195 | return -1; |
60bf62d4 SH |
196 | } |
197 | ||
0cc417b2 | 198 | return _recursive_rmdir(path, mystat.st_dev, exclude, 0, onedev); |
60bf62d4 SH |
199 | } |
200 | ||
9ddaf3bf | 201 | /* borrowed from iproute2 */ |
7c11d57a | 202 | extern int get_u16(unsigned short *val, const char *arg, int base) |
9ddaf3bf JHS |
203 | { |
204 | unsigned long res; | |
205 | char *ptr; | |
206 | ||
207 | if (!arg || !*arg) | |
208 | return -1; | |
209 | ||
09bbd745 | 210 | errno = 0; |
9ddaf3bf | 211 | res = strtoul(arg, &ptr, base); |
09bbd745 | 212 | if (!ptr || ptr == arg || *ptr || res > 0xFFFF || errno != 0) |
9ddaf3bf JHS |
213 | return -1; |
214 | ||
215 | *val = res; | |
216 | ||
217 | return 0; | |
218 | } | |
219 | ||
3ce74686 | 220 | extern int mkdir_p(const char *dir, mode_t mode) |
1b09f2c0 | 221 | { |
3ce74686 SH |
222 | const char *tmp = dir; |
223 | const char *orig = dir; | |
860fc865 RW |
224 | char *makeme; |
225 | ||
226 | do { | |
227 | dir = tmp + strspn(tmp, "/"); | |
228 | tmp = dir + strcspn(dir, "/"); | |
d74325c4 | 229 | makeme = strndup(orig, dir - orig); |
860fc865 RW |
230 | if (*makeme) { |
231 | if (mkdir(makeme, mode) && errno != EEXIST) { | |
959aee9c | 232 | SYSERROR("failed to create directory '%s'", makeme); |
d74325c4 | 233 | free(makeme); |
860fc865 RW |
234 | return -1; |
235 | } | |
236 | } | |
d74325c4 | 237 | free(makeme); |
860fc865 | 238 | } while(tmp != dir); |
1b09f2c0 | 239 | |
98663823 | 240 | return 0; |
1b09f2c0 | 241 | } |
2a59a681 | 242 | |
44b9ae4b | 243 | char *get_rundir() |
9e60f51d | 244 | { |
97a696c6 SG |
245 | char *rundir; |
246 | const char *homedir; | |
9e60f51d | 247 | |
d6470e71 | 248 | if (geteuid() == 0) { |
c580b8d2 | 249 | rundir = strdup(RUNTIME_PATH); |
d6470e71 SG |
250 | return rundir; |
251 | } | |
97a696c6 SG |
252 | |
253 | rundir = getenv("XDG_RUNTIME_DIR"); | |
44b9ae4b SG |
254 | if (rundir) { |
255 | rundir = strdup(rundir); | |
256 | return rundir; | |
257 | } | |
97a696c6 | 258 | |
44b9ae4b SG |
259 | INFO("XDG_RUNTIME_DIR isn't set in the environment."); |
260 | homedir = getenv("HOME"); | |
261 | if (!homedir) { | |
262 | ERROR("HOME isn't set in the environment."); | |
263 | return NULL; | |
97a696c6 SG |
264 | } |
265 | ||
44b9ae4b SG |
266 | rundir = malloc(sizeof(char) * (17 + strlen(homedir))); |
267 | sprintf(rundir, "%s/.cache/lxc/run/", homedir); | |
268 | ||
9e60f51d DE |
269 | return rundir; |
270 | } | |
271 | ||
9be53773 SH |
272 | int wait_for_pid(pid_t pid) |
273 | { | |
274 | int status, ret; | |
275 | ||
276 | again: | |
277 | ret = waitpid(pid, &status, 0); | |
278 | if (ret == -1) { | |
71b9b8ed | 279 | if (errno == EINTR) |
9be53773 SH |
280 | goto again; |
281 | return -1; | |
282 | } | |
283 | if (ret != pid) | |
284 | goto again; | |
285 | if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) | |
286 | return -1; | |
287 | return 0; | |
288 | } | |
c797a220 CS |
289 | |
290 | int lxc_wait_for_pid_status(pid_t pid) | |
291 | { | |
292 | int status, ret; | |
293 | ||
294 | again: | |
295 | ret = waitpid(pid, &status, 0); | |
296 | if (ret == -1) { | |
297 | if (errno == EINTR) | |
298 | goto again; | |
299 | return -1; | |
300 | } | |
301 | if (ret != pid) | |
302 | goto again; | |
303 | return status; | |
304 | } | |
92f023dc | 305 | |
650468bb | 306 | ssize_t lxc_write_nointr(int fd, const void* buf, size_t count) |
92f023dc | 307 | { |
650468bb | 308 | ssize_t ret; |
92f023dc CS |
309 | again: |
310 | ret = write(fd, buf, count); | |
311 | if (ret < 0 && errno == EINTR) | |
312 | goto again; | |
313 | return ret; | |
314 | } | |
315 | ||
650468bb | 316 | ssize_t lxc_read_nointr(int fd, void* buf, size_t count) |
92f023dc | 317 | { |
650468bb | 318 | ssize_t ret; |
92f023dc CS |
319 | again: |
320 | ret = read(fd, buf, count); | |
321 | if (ret < 0 && errno == EINTR) | |
322 | goto again; | |
323 | return ret; | |
324 | } | |
325 | ||
650468bb | 326 | ssize_t lxc_read_nointr_expect(int fd, void* buf, size_t count, const void* expected_buf) |
92f023dc | 327 | { |
650468bb | 328 | ssize_t ret; |
92f023dc CS |
329 | ret = lxc_read_nointr(fd, buf, count); |
330 | if (ret <= 0) | |
331 | return ret; | |
650468bb | 332 | if ((size_t)ret != count) |
92f023dc CS |
333 | return -1; |
334 | if (expected_buf && memcmp(buf, expected_buf, count) != 0) { | |
335 | errno = EINVAL; | |
336 | return -1; | |
337 | } | |
338 | return ret; | |
339 | } | |
3ce74686 SH |
340 | |
341 | #if HAVE_LIBGNUTLS | |
342 | #include <gnutls/gnutls.h> | |
343 | #include <gnutls/crypto.h> | |
41246cee DE |
344 | |
345 | __attribute__((constructor)) | |
346 | static void gnutls_lxc_init(void) | |
347 | { | |
348 | gnutls_global_init(); | |
349 | } | |
350 | ||
3ce74686 SH |
351 | int sha1sum_file(char *fnam, unsigned char *digest) |
352 | { | |
353 | char *buf; | |
354 | int ret; | |
355 | FILE *f; | |
356 | long flen; | |
357 | ||
358 | if (!fnam) | |
359 | return -1; | |
025ed0f3 | 360 | f = fopen_cloexec(fnam, "r"); |
7be677a8 | 361 | if (!f) { |
3ce74686 SH |
362 | SYSERROR("Error opening template"); |
363 | return -1; | |
364 | } | |
365 | if (fseek(f, 0, SEEK_END) < 0) { | |
366 | SYSERROR("Error seeking to end of template"); | |
dd1d77f9 | 367 | fclose(f); |
3ce74686 SH |
368 | return -1; |
369 | } | |
370 | if ((flen = ftell(f)) < 0) { | |
371 | SYSERROR("Error telling size of template"); | |
dd1d77f9 | 372 | fclose(f); |
3ce74686 SH |
373 | return -1; |
374 | } | |
375 | if (fseek(f, 0, SEEK_SET) < 0) { | |
376 | SYSERROR("Error seeking to start of template"); | |
dd1d77f9 | 377 | fclose(f); |
3ce74686 SH |
378 | return -1; |
379 | } | |
380 | if ((buf = malloc(flen+1)) == NULL) { | |
381 | SYSERROR("Out of memory"); | |
dd1d77f9 | 382 | fclose(f); |
3ce74686 SH |
383 | return -1; |
384 | } | |
385 | if (fread(buf, 1, flen, f) != flen) { | |
386 | SYSERROR("Failure reading template"); | |
387 | free(buf); | |
dd1d77f9 | 388 | fclose(f); |
3ce74686 SH |
389 | return -1; |
390 | } | |
dd1d77f9 | 391 | if (fclose(f) < 0) { |
3ce74686 SH |
392 | SYSERROR("Failre closing template"); |
393 | free(buf); | |
394 | return -1; | |
395 | } | |
396 | buf[flen] = '\0'; | |
397 | ret = gnutls_hash_fast(GNUTLS_DIG_SHA1, buf, flen, (void *)digest); | |
398 | free(buf); | |
399 | return ret; | |
400 | } | |
401 | #endif | |
61a1d519 CS |
402 | |
403 | char** lxc_va_arg_list_to_argv(va_list ap, size_t skip, int do_strdup) | |
404 | { | |
405 | va_list ap2; | |
406 | size_t count = 1 + skip; | |
407 | char **result; | |
408 | ||
409 | /* first determine size of argument list, we don't want to reallocate | |
410 | * constantly... | |
411 | */ | |
412 | va_copy(ap2, ap); | |
413 | while (1) { | |
414 | char* arg = va_arg(ap2, char*); | |
415 | if (!arg) | |
416 | break; | |
417 | count++; | |
418 | } | |
419 | va_end(ap2); | |
420 | ||
421 | result = calloc(count, sizeof(char*)); | |
422 | if (!result) | |
423 | return NULL; | |
424 | count = skip; | |
425 | while (1) { | |
426 | char* arg = va_arg(ap, char*); | |
427 | if (!arg) | |
428 | break; | |
429 | arg = do_strdup ? strdup(arg) : arg; | |
430 | if (!arg) | |
431 | goto oom; | |
432 | result[count++] = arg; | |
433 | } | |
434 | ||
435 | /* calloc has already set last element to NULL*/ | |
436 | return result; | |
437 | ||
438 | oom: | |
439 | free(result); | |
440 | return NULL; | |
441 | } | |
442 | ||
443 | const char** lxc_va_arg_list_to_argv_const(va_list ap, size_t skip) | |
444 | { | |
445 | return (const char**)lxc_va_arg_list_to_argv(ap, skip, 0); | |
446 | } | |
db27c8d7 | 447 | |
8bd8018e | 448 | struct lxc_popen_FILE *lxc_popen(const char *command) |
ebec9176 | 449 | { |
3f323207 | 450 | int ret; |
ebec9176 AM |
451 | int pipe_fds[2]; |
452 | pid_t child_pid; | |
8bd8018e | 453 | struct lxc_popen_FILE *fp = NULL; |
ebec9176 | 454 | |
8bd8018e CB |
455 | ret = pipe2(pipe_fds, O_CLOEXEC); |
456 | if (ret < 0) | |
ebec9176 | 457 | return NULL; |
ebec9176 AM |
458 | |
459 | child_pid = fork(); | |
8bd8018e CB |
460 | if (child_pid < 0) |
461 | goto on_error; | |
462 | ||
463 | if (!child_pid) { | |
464 | sigset_t mask; | |
465 | ||
466 | close(pipe_fds[0]); | |
467 | ||
468 | /* duplicate stdout */ | |
469 | if (pipe_fds[1] != STDOUT_FILENO) | |
470 | ret = dup2(pipe_fds[1], STDOUT_FILENO); | |
471 | else | |
472 | ret = fcntl(pipe_fds[1], F_SETFD, 0); | |
473 | if (ret < 0) { | |
474 | close(pipe_fds[1]); | |
475 | exit(EXIT_FAILURE); | |
3f323207 CB |
476 | } |
477 | ||
8bd8018e CB |
478 | /* duplicate stderr */ |
479 | if (pipe_fds[1] != STDERR_FILENO) | |
480 | ret = dup2(pipe_fds[1], STDERR_FILENO); | |
481 | else | |
482 | ret = fcntl(pipe_fds[1], F_SETFD, 0); | |
483 | close(pipe_fds[1]); | |
484 | if (ret < 0) | |
485 | exit(EXIT_FAILURE); | |
486 | ||
487 | /* unblock all signals */ | |
488 | ret = sigfillset(&mask); | |
489 | if (ret < 0) | |
490 | exit(EXIT_FAILURE); | |
491 | ||
492 | ret = sigprocmask(SIG_UNBLOCK, &mask, NULL); | |
493 | if (ret < 0) | |
494 | exit(EXIT_FAILURE); | |
495 | ||
496 | execl("/bin/sh", "sh", "-c", command, (char *)NULL); | |
ebec9176 AM |
497 | exit(127); |
498 | } | |
499 | ||
8bd8018e CB |
500 | close(pipe_fds[1]); |
501 | pipe_fds[1] = -1; | |
ebec9176 | 502 | |
8bd8018e CB |
503 | fp = malloc(sizeof(*fp)); |
504 | if (!fp) | |
505 | goto on_error; | |
ebec9176 AM |
506 | |
507 | fp->child_pid = child_pid; | |
8bd8018e | 508 | fp->pipe = pipe_fds[0]; |
ebec9176 | 509 | |
8bd8018e CB |
510 | fp->f = fdopen(pipe_fds[0], "r"); |
511 | if (!fp->f) | |
512 | goto on_error; | |
ebec9176 | 513 | |
8bd8018e | 514 | return fp; |
ebec9176 | 515 | |
8bd8018e CB |
516 | on_error: |
517 | if (fp) | |
ebec9176 | 518 | free(fp); |
ebec9176 | 519 | |
8bd8018e CB |
520 | if (pipe_fds[0] >= 0) |
521 | close(pipe_fds[0]); | |
522 | ||
523 | if (pipe_fds[1] >= 0) | |
524 | close(pipe_fds[1]); | |
ebec9176 AM |
525 | |
526 | return NULL; | |
527 | } | |
528 | ||
8bd8018e | 529 | int lxc_pclose(struct lxc_popen_FILE *fp) |
ebec9176 | 530 | { |
ebec9176 | 531 | pid_t wait_pid; |
8bd8018e | 532 | int wstatus = 0; |
ebec9176 | 533 | |
8bd8018e | 534 | if (!fp) |
ebec9176 | 535 | return -1; |
ebec9176 AM |
536 | |
537 | do { | |
8bd8018e CB |
538 | wait_pid = waitpid(fp->child_pid, &wstatus, 0); |
539 | } while (wait_pid < 0 && errno == EINTR); | |
ebec9176 | 540 | |
8bd8018e CB |
541 | close(fp->pipe); |
542 | fclose(fp->f); | |
543 | free(fp); | |
544 | ||
545 | if (wait_pid < 0) | |
ebec9176 | 546 | return -1; |
ebec9176 AM |
547 | |
548 | return wstatus; | |
549 | } | |
550 | ||
502657d5 CS |
551 | char *lxc_string_replace(const char *needle, const char *replacement, const char *haystack) |
552 | { | |
553 | ssize_t len = -1, saved_len = -1; | |
554 | char *result = NULL; | |
555 | size_t replacement_len = strlen(replacement); | |
556 | size_t needle_len = strlen(needle); | |
557 | ||
558 | /* should be executed exactly twice */ | |
559 | while (len == -1 || result == NULL) { | |
560 | char *p; | |
561 | char *last_p; | |
562 | ssize_t part_len; | |
563 | ||
564 | if (len != -1) { | |
565 | result = calloc(1, len + 1); | |
566 | if (!result) | |
567 | return NULL; | |
568 | saved_len = len; | |
569 | } | |
570 | ||
571 | len = 0; | |
572 | ||
573 | for (last_p = (char *)haystack, p = strstr(last_p, needle); p; last_p = p, p = strstr(last_p, needle)) { | |
574 | part_len = (ssize_t)(p - last_p); | |
575 | if (result && part_len > 0) | |
576 | memcpy(&result[len], last_p, part_len); | |
577 | len += part_len; | |
578 | if (result && replacement_len > 0) | |
579 | memcpy(&result[len], replacement, replacement_len); | |
580 | len += replacement_len; | |
581 | p += needle_len; | |
582 | } | |
583 | part_len = strlen(last_p); | |
584 | if (result && part_len > 0) | |
585 | memcpy(&result[len], last_p, part_len); | |
586 | len += part_len; | |
587 | } | |
588 | ||
589 | /* make sure we did the same thing twice, | |
590 | * once for calculating length, the other | |
591 | * time for copying data */ | |
97bc2422 CB |
592 | if (saved_len != len) { |
593 | free(result); | |
594 | return NULL; | |
595 | } | |
502657d5 CS |
596 | /* make sure we didn't overwrite any buffer, |
597 | * due to calloc the string should be 0-terminated */ | |
97bc2422 CB |
598 | if (result[len] != '\0') { |
599 | free(result); | |
600 | return NULL; | |
601 | } | |
502657d5 CS |
602 | |
603 | return result; | |
604 | } | |
605 | ||
606 | bool lxc_string_in_array(const char *needle, const char **haystack) | |
607 | { | |
608 | for (; haystack && *haystack; haystack++) | |
609 | if (!strcmp(needle, *haystack)) | |
610 | return true; | |
611 | return false; | |
612 | } | |
613 | ||
614 | char *lxc_string_join(const char *sep, const char **parts, bool use_as_prefix) | |
615 | { | |
616 | char *result; | |
617 | char **p; | |
618 | size_t sep_len = strlen(sep); | |
619 | size_t result_len = use_as_prefix * sep_len; | |
620 | ||
621 | /* calculate new string length */ | |
622 | for (p = (char **)parts; *p; p++) | |
623 | result_len += (p > (char **)parts) * sep_len + strlen(*p); | |
624 | ||
625 | result = calloc(result_len + 1, 1); | |
626 | if (!result) | |
627 | return NULL; | |
628 | ||
629 | if (use_as_prefix) | |
630 | strcpy(result, sep); | |
631 | for (p = (char **)parts; *p; p++) { | |
632 | if (p > (char **)parts) | |
633 | strcat(result, sep); | |
634 | strcat(result, *p); | |
635 | } | |
636 | ||
637 | return result; | |
638 | } | |
639 | ||
640 | char **lxc_normalize_path(const char *path) | |
641 | { | |
642 | char **components; | |
643 | char **p; | |
644 | size_t components_len = 0; | |
645 | size_t pos = 0; | |
646 | ||
647 | components = lxc_string_split(path, '/'); | |
648 | if (!components) | |
649 | return NULL; | |
650 | for (p = components; *p; p++) | |
651 | components_len++; | |
652 | ||
653 | /* resolve '.' and '..' */ | |
654 | for (pos = 0; pos < components_len; ) { | |
655 | if (!strcmp(components[pos], ".") || (!strcmp(components[pos], "..") && pos == 0)) { | |
656 | /* eat this element */ | |
657 | free(components[pos]); | |
658 | memmove(&components[pos], &components[pos+1], sizeof(char *) * (components_len - pos)); | |
659 | components_len--; | |
660 | } else if (!strcmp(components[pos], "..")) { | |
661 | /* eat this and the previous element */ | |
662 | free(components[pos - 1]); | |
663 | free(components[pos]); | |
664 | memmove(&components[pos-1], &components[pos+1], sizeof(char *) * (components_len - pos)); | |
665 | components_len -= 2; | |
666 | pos--; | |
667 | } else { | |
668 | pos++; | |
669 | } | |
670 | } | |
671 | ||
672 | return components; | |
673 | } | |
674 | ||
eda0afd4 | 675 | char *lxc_deslashify(const char *path) |
aeb3682f | 676 | { |
eda0afd4 | 677 | char *dup, *p; |
c56a9652 CB |
678 | char **parts = NULL; |
679 | size_t n, len; | |
aeb3682f | 680 | |
eda0afd4 CB |
681 | dup = strdup(path); |
682 | if (!dup) | |
683 | return NULL; | |
684 | ||
685 | parts = lxc_normalize_path(dup); | |
686 | if (!parts) { | |
687 | free(dup); | |
688 | return NULL; | |
689 | } | |
aeb3682f | 690 | |
c56a9652 CB |
691 | /* We'll end up here if path == "///" or path == "". */ |
692 | if (!*parts) { | |
eda0afd4 | 693 | len = strlen(dup); |
f85b16a1 | 694 | if (!len) { |
eda0afd4 CB |
695 | lxc_free_array((void **)parts, free); |
696 | return dup; | |
f85b16a1 | 697 | } |
eda0afd4 | 698 | n = strcspn(dup, "/"); |
c56a9652 | 699 | if (n == len) { |
eda0afd4 CB |
700 | free(dup); |
701 | lxc_free_array((void **)parts, free); | |
702 | ||
c56a9652 CB |
703 | p = strdup("/"); |
704 | if (!p) | |
eda0afd4 CB |
705 | return NULL; |
706 | ||
707 | return p; | |
c56a9652 CB |
708 | } |
709 | } | |
710 | ||
eda0afd4 CB |
711 | p = lxc_string_join("/", (const char **)parts, *dup == '/'); |
712 | free(dup); | |
f85b16a1 | 713 | lxc_free_array((void **)parts, free); |
eda0afd4 | 714 | return p; |
aeb3682f TA |
715 | } |
716 | ||
24b51482 CS |
717 | char *lxc_append_paths(const char *first, const char *second) |
718 | { | |
719 | size_t len = strlen(first) + strlen(second) + 1; | |
720 | const char *pattern = "%s%s"; | |
721 | char *result = NULL; | |
722 | ||
723 | if (second[0] != '/') { | |
724 | len += 1; | |
725 | pattern = "%s/%s"; | |
726 | } | |
727 | ||
728 | result = calloc(1, len); | |
729 | if (!result) | |
730 | return NULL; | |
731 | ||
732 | snprintf(result, len, pattern, first, second); | |
733 | return result; | |
734 | } | |
735 | ||
502657d5 CS |
736 | bool lxc_string_in_list(const char *needle, const char *haystack, char _sep) |
737 | { | |
738 | char *token, *str, *saveptr = NULL; | |
739 | char sep[2] = { _sep, '\0' }; | |
740 | ||
741 | if (!haystack || !needle) | |
742 | return 0; | |
743 | ||
744 | str = alloca(strlen(haystack)+1); | |
745 | strcpy(str, haystack); | |
746 | for (; (token = strtok_r(str, sep, &saveptr)); str = NULL) { | |
747 | if (strcmp(needle, token) == 0) | |
748 | return 1; | |
749 | } | |
750 | ||
751 | return 0; | |
752 | } | |
753 | ||
754 | char **lxc_string_split(const char *string, char _sep) | |
755 | { | |
756 | char *token, *str, *saveptr = NULL; | |
605ea1f7 CB |
757 | char sep[2] = {_sep, '\0'}; |
758 | char **tmp = NULL, **result = NULL; | |
502657d5 CS |
759 | size_t result_capacity = 0; |
760 | size_t result_count = 0; | |
761 | int r, saved_errno; | |
762 | ||
763 | if (!string) | |
764 | return calloc(1, sizeof(char *)); | |
765 | ||
605ea1f7 | 766 | str = alloca(strlen(string) + 1); |
502657d5 CS |
767 | strcpy(str, string); |
768 | for (; (token = strtok_r(str, sep, &saveptr)); str = NULL) { | |
769 | r = lxc_grow_array((void ***)&result, &result_capacity, result_count + 1, 16); | |
770 | if (r < 0) | |
771 | goto error_out; | |
772 | result[result_count] = strdup(token); | |
773 | if (!result[result_count]) | |
774 | goto error_out; | |
775 | result_count++; | |
776 | } | |
777 | ||
778 | /* if we allocated too much, reduce it */ | |
605ea1f7 CB |
779 | tmp = realloc(result, (result_count + 1) * sizeof(char *)); |
780 | if (!tmp) | |
781 | goto error_out; | |
782 | result = tmp; | |
783 | /* Make sure we don't return uninitialized memory. */ | |
784 | if (result_count == 0) | |
785 | *result = NULL; | |
786 | return result; | |
502657d5 CS |
787 | error_out: |
788 | saved_errno = errno; | |
789 | lxc_free_array((void **)result, free); | |
790 | errno = saved_errno; | |
791 | return NULL; | |
792 | } | |
793 | ||
3dca1af0 SH |
794 | static bool complete_word(char ***result, char *start, char *end, size_t *cap, size_t *cnt) |
795 | { | |
796 | int r; | |
797 | ||
798 | r = lxc_grow_array((void ***)result, cap, 2 + *cnt, 16); | |
799 | if (r < 0) | |
800 | return false; | |
801 | (*result)[*cnt] = strndup(start, end - start); | |
802 | if (!(*result)[*cnt]) | |
803 | return false; | |
804 | (*cnt)++; | |
805 | ||
806 | return true; | |
807 | } | |
808 | ||
809 | /* | |
810 | * Given a a string 'one two "three four"', split into three words, | |
811 | * one, two, and "three four" | |
812 | */ | |
813 | char **lxc_string_split_quoted(char *string) | |
814 | { | |
815 | char *nextword = string, *p, state; | |
816 | char **result = NULL; | |
817 | size_t result_capacity = 0; | |
818 | size_t result_count = 0; | |
819 | ||
820 | if (!string || !*string) | |
821 | return calloc(1, sizeof(char *)); | |
822 | ||
823 | // TODO I'm *not* handling escaped quote | |
824 | state = ' '; | |
825 | for (p = string; *p; p++) { | |
826 | switch(state) { | |
827 | case ' ': | |
828 | if (isspace(*p)) | |
829 | continue; | |
830 | else if (*p == '"' || *p == '\'') { | |
831 | nextword = p; | |
832 | state = *p; | |
833 | continue; | |
834 | } | |
835 | nextword = p; | |
836 | state = 'a'; | |
837 | continue; | |
838 | case 'a': | |
839 | if (isspace(*p)) { | |
840 | complete_word(&result, nextword, p, &result_capacity, &result_count); | |
841 | state = ' '; | |
842 | continue; | |
843 | } | |
844 | continue; | |
845 | case '"': | |
846 | case '\'': | |
847 | if (*p == state) { | |
848 | complete_word(&result, nextword+1, p, &result_capacity, &result_count); | |
849 | state = ' '; | |
850 | continue; | |
851 | } | |
852 | continue; | |
853 | } | |
854 | } | |
855 | ||
856 | if (state == 'a') | |
857 | complete_word(&result, nextword, p, &result_capacity, &result_count); | |
858 | ||
859 | return realloc(result, (result_count + 1) * sizeof(char *)); | |
860 | } | |
861 | ||
502657d5 CS |
862 | char **lxc_string_split_and_trim(const char *string, char _sep) |
863 | { | |
864 | char *token, *str, *saveptr = NULL; | |
865 | char sep[2] = { _sep, '\0' }; | |
866 | char **result = NULL; | |
867 | size_t result_capacity = 0; | |
868 | size_t result_count = 0; | |
869 | int r, saved_errno; | |
870 | size_t i = 0; | |
871 | ||
872 | if (!string) | |
873 | return calloc(1, sizeof(char *)); | |
874 | ||
875 | str = alloca(strlen(string)+1); | |
876 | strcpy(str, string); | |
877 | for (; (token = strtok_r(str, sep, &saveptr)); str = NULL) { | |
878 | while (token[0] == ' ' || token[0] == '\t') | |
879 | token++; | |
880 | i = strlen(token); | |
881 | while (i > 0 && (token[i - 1] == ' ' || token[i - 1] == '\t')) { | |
882 | token[i - 1] = '\0'; | |
883 | i--; | |
884 | } | |
885 | r = lxc_grow_array((void ***)&result, &result_capacity, result_count + 1, 16); | |
886 | if (r < 0) | |
887 | goto error_out; | |
888 | result[result_count] = strdup(token); | |
889 | if (!result[result_count]) | |
890 | goto error_out; | |
891 | result_count++; | |
892 | } | |
893 | ||
894 | /* if we allocated too much, reduce it */ | |
895 | return realloc(result, (result_count + 1) * sizeof(char *)); | |
896 | error_out: | |
897 | saved_errno = errno; | |
898 | lxc_free_array((void **)result, free); | |
899 | errno = saved_errno; | |
900 | return NULL; | |
901 | } | |
902 | ||
903 | void lxc_free_array(void **array, lxc_free_fn element_free_fn) | |
904 | { | |
905 | void **p; | |
906 | for (p = array; p && *p; p++) | |
907 | element_free_fn(*p); | |
908 | free((void*)array); | |
909 | } | |
910 | ||
911 | int lxc_grow_array(void ***array, size_t* capacity, size_t new_size, size_t capacity_increment) | |
912 | { | |
913 | size_t new_capacity; | |
914 | void **new_array; | |
915 | ||
916 | /* first time around, catch some trivial mistakes of the user | |
917 | * only initializing one of these */ | |
918 | if (!*array || !*capacity) { | |
919 | *array = NULL; | |
920 | *capacity = 0; | |
921 | } | |
922 | ||
923 | new_capacity = *capacity; | |
924 | while (new_size + 1 > new_capacity) | |
925 | new_capacity += capacity_increment; | |
926 | if (new_capacity != *capacity) { | |
927 | /* we have to reallocate */ | |
928 | new_array = realloc(*array, new_capacity * sizeof(void *)); | |
929 | if (!new_array) | |
930 | return -1; | |
931 | memset(&new_array[*capacity], 0, (new_capacity - (*capacity)) * sizeof(void *)); | |
932 | *array = new_array; | |
933 | *capacity = new_capacity; | |
934 | } | |
935 | ||
936 | /* array has sufficient elements */ | |
937 | return 0; | |
938 | } | |
939 | ||
940 | size_t lxc_array_len(void **array) | |
941 | { | |
942 | void **p; | |
943 | size_t result = 0; | |
944 | ||
945 | for (p = array; p && *p; p++) | |
946 | result++; | |
947 | ||
948 | return result; | |
949 | } | |
950 | ||
0e95426b CS |
951 | int lxc_write_to_file(const char *filename, const void* buf, size_t count, bool add_newline) |
952 | { | |
953 | int fd, saved_errno; | |
954 | ssize_t ret; | |
955 | ||
956 | fd = open(filename, O_WRONLY | O_TRUNC | O_CREAT | O_CLOEXEC, 0666); | |
957 | if (fd < 0) | |
958 | return -1; | |
959 | ret = lxc_write_nointr(fd, buf, count); | |
960 | if (ret < 0) | |
799f29ab | 961 | goto out_error; |
0e95426b CS |
962 | if ((size_t)ret != count) |
963 | goto out_error; | |
964 | if (add_newline) { | |
965 | ret = lxc_write_nointr(fd, "\n", 1); | |
966 | if (ret != 1) | |
967 | goto out_error; | |
968 | } | |
969 | close(fd); | |
970 | return 0; | |
971 | ||
972 | out_error: | |
973 | saved_errno = errno; | |
974 | close(fd); | |
975 | errno = saved_errno; | |
976 | return -1; | |
977 | } | |
978 | ||
979 | int lxc_read_from_file(const char *filename, void* buf, size_t count) | |
980 | { | |
981 | int fd = -1, saved_errno; | |
982 | ssize_t ret; | |
983 | ||
984 | fd = open(filename, O_RDONLY | O_CLOEXEC); | |
985 | if (fd < 0) | |
986 | return -1; | |
987 | ||
988 | if (!buf || !count) { | |
989 | char buf2[100]; | |
990 | size_t count2 = 0; | |
991 | while ((ret = read(fd, buf2, 100)) > 0) | |
992 | count2 += ret; | |
993 | if (ret >= 0) | |
994 | ret = count2; | |
995 | } else { | |
996 | memset(buf, 0, count); | |
997 | ret = read(fd, buf, count); | |
998 | } | |
999 | ||
1000 | if (ret < 0) | |
1001 | ERROR("read %s: %s", filename, strerror(errno)); | |
1002 | ||
1003 | saved_errno = errno; | |
1004 | close(fd); | |
1005 | errno = saved_errno; | |
1006 | return ret; | |
1007 | } | |
799f29ab ÇO |
1008 | |
1009 | void **lxc_append_null_to_array(void **array, size_t count) | |
1010 | { | |
1011 | void **temp; | |
1012 | ||
1013 | /* Append NULL to the array */ | |
1014 | if (count) { | |
1015 | temp = realloc(array, (count + 1) * sizeof(*array)); | |
1016 | if (!temp) { | |
84760c11 | 1017 | size_t i; |
799f29ab ÇO |
1018 | for (i = 0; i < count; i++) |
1019 | free(array[i]); | |
1020 | free(array); | |
1021 | return NULL; | |
1022 | } | |
1023 | array = temp; | |
1024 | array[count] = NULL; | |
1025 | } | |
1026 | return array; | |
1027 | } | |
508c263e SH |
1028 | |
1029 | int randseed(bool srand_it) | |
1030 | { | |
1031 | /* | |
1032 | srand pre-seed function based on /dev/urandom | |
1033 | */ | |
091045f8 | 1034 | unsigned int seed = time(NULL) + getpid(); |
508c263e SH |
1035 | |
1036 | FILE *f; | |
1037 | f = fopen("/dev/urandom", "r"); | |
1038 | if (f) { | |
1039 | int ret = fread(&seed, sizeof(seed), 1, f); | |
1040 | if (ret != 1) | |
1041 | DEBUG("unable to fread /dev/urandom, %s, fallback to time+pid rand seed", strerror(errno)); | |
1042 | fclose(f); | |
1043 | } | |
1044 | ||
1045 | if (srand_it) | |
1046 | srand(seed); | |
1047 | ||
1048 | return seed; | |
1049 | } | |
5d897655 SH |
1050 | |
1051 | uid_t get_ns_uid(uid_t orig) | |
1052 | { | |
1053 | char *line = NULL; | |
1054 | size_t sz = 0; | |
1055 | uid_t nsid, hostid, range; | |
1056 | FILE *f = fopen("/proc/self/uid_map", "r"); | |
1057 | if (!f) | |
1058 | return 0; | |
1059 | ||
1060 | while (getline(&line, &sz, f) != -1) { | |
1061 | if (sscanf(line, "%u %u %u", &nsid, &hostid, &range) != 3) | |
1062 | continue; | |
1063 | if (hostid <= orig && hostid + range > orig) { | |
1064 | nsid += orig - hostid; | |
1065 | goto found; | |
1066 | } | |
1067 | } | |
1068 | ||
1069 | nsid = 0; | |
1070 | found: | |
1071 | fclose(f); | |
1072 | free(line); | |
1073 | return nsid; | |
1074 | } | |
c476bdce SH |
1075 | |
1076 | bool dir_exists(const char *path) | |
1077 | { | |
1078 | struct stat sb; | |
1079 | int ret; | |
1080 | ||
1081 | ret = stat(path, &sb); | |
1082 | if (ret < 0) | |
1a0e70ac | 1083 | /* Could be something other than eexist, just say "no". */ |
c476bdce SH |
1084 | return false; |
1085 | return S_ISDIR(sb.st_mode); | |
1086 | } | |
93c379f0 ÇO |
1087 | |
1088 | /* Note we don't use SHA-1 here as we don't want to depend on HAVE_GNUTLS. | |
1089 | * FNV has good anti collision properties and we're not worried | |
1090 | * about pre-image resistance or one-way-ness, we're just trying to make | |
1091 | * the name unique in the 108 bytes of space we have. | |
1092 | */ | |
1093 | uint64_t fnv_64a_buf(void *buf, size_t len, uint64_t hval) | |
1094 | { | |
1095 | unsigned char *bp; | |
1096 | ||
1097 | for(bp = buf; bp < (unsigned char *)buf + len; bp++) | |
1098 | { | |
1099 | /* xor the bottom with the current octet */ | |
1100 | hval ^= (uint64_t)*bp; | |
1101 | ||
1102 | /* gcc optimised: | |
1103 | * multiply by the 64 bit FNV magic prime mod 2^64 | |
1104 | */ | |
1105 | hval += (hval << 1) + (hval << 4) + (hval << 5) + | |
1106 | (hval << 7) + (hval << 8) + (hval << 40); | |
1107 | } | |
1108 | ||
1109 | return hval; | |
1110 | } | |
2c6f3fc9 SH |
1111 | |
1112 | /* | |
1113 | * Detect whether / is mounted MS_SHARED. The only way I know of to | |
1114 | * check that is through /proc/self/mountinfo. | |
1115 | * I'm only checking for /. If the container rootfs or mount location | |
1116 | * is MS_SHARED, but not '/', then you're out of luck - figuring that | |
1117 | * out would be too much work to be worth it. | |
1118 | */ | |
2c6f3fc9 SH |
1119 | int detect_shared_rootfs(void) |
1120 | { | |
eab15c1e | 1121 | char buf[LXC_LINELEN], *p; |
2c6f3fc9 SH |
1122 | FILE *f; |
1123 | int i; | |
1124 | char *p2; | |
1125 | ||
1126 | f = fopen("/proc/self/mountinfo", "r"); | |
1127 | if (!f) | |
1128 | return 0; | |
eab15c1e CB |
1129 | while (fgets(buf, LXC_LINELEN, f)) { |
1130 | for (p = buf, i = 0; p && i < 4; i++) | |
1131 | p = strchr(p + 1, ' '); | |
2c6f3fc9 SH |
1132 | if (!p) |
1133 | continue; | |
eab15c1e | 1134 | p2 = strchr(p + 1, ' '); |
2c6f3fc9 SH |
1135 | if (!p2) |
1136 | continue; | |
1137 | *p2 = '\0'; | |
eab15c1e | 1138 | if (strcmp(p + 1, "/") == 0) { |
1a0e70ac | 1139 | /* This is '/'. Is it shared? */ |
eab15c1e | 1140 | p = strchr(p2 + 1, ' '); |
2c6f3fc9 SH |
1141 | if (p && strstr(p, "shared:")) { |
1142 | fclose(f); | |
1143 | return 1; | |
1144 | } | |
1145 | } | |
1146 | } | |
1147 | fclose(f); | |
1148 | return 0; | |
1149 | } | |
0e6e3a41 | 1150 | |
51d0854c DY |
1151 | bool switch_to_ns(pid_t pid, const char *ns) { |
1152 | int fd, ret; | |
1153 | char nspath[MAXPATHLEN]; | |
1154 | ||
1155 | /* Switch to new ns */ | |
1156 | ret = snprintf(nspath, MAXPATHLEN, "/proc/%d/ns/%s", pid, ns); | |
1157 | if (ret < 0 || ret >= MAXPATHLEN) | |
1158 | return false; | |
1159 | ||
1160 | fd = open(nspath, O_RDONLY); | |
1161 | if (fd < 0) { | |
1162 | SYSERROR("failed to open %s", nspath); | |
1163 | return false; | |
1164 | } | |
1165 | ||
1166 | ret = setns(fd, 0); | |
1167 | if (ret) { | |
1168 | SYSERROR("failed to set process %d to %s of %d.", pid, ns, fd); | |
1169 | close(fd); | |
1170 | return false; | |
1171 | } | |
1172 | close(fd); | |
1173 | return true; | |
1174 | } | |
1175 | ||
b7f954bb SH |
1176 | /* |
1177 | * looking at fs/proc_namespace.c, it appears we can | |
1178 | * actually expect the rootfs entry to very specifically contain | |
1179 | * " - rootfs rootfs " | |
1180 | * IIUC, so long as we've chrooted so that rootfs is not our root, | |
1181 | * the rootfs entry should always be skipped in mountinfo contents. | |
1182 | */ | |
fa454c8e | 1183 | bool detect_ramfs_rootfs(void) |
b7f954bb | 1184 | { |
b7f954bb | 1185 | FILE *f; |
fa454c8e CB |
1186 | char *p, *p2; |
1187 | char *line = NULL; | |
1188 | size_t len = 0; | |
b7f954bb | 1189 | int i; |
b7f954bb SH |
1190 | |
1191 | f = fopen("/proc/self/mountinfo", "r"); | |
1192 | if (!f) | |
fa454c8e CB |
1193 | return false; |
1194 | ||
1195 | while (getline(&line, &len, f) != -1) { | |
1196 | for (p = line, i = 0; p && i < 4; i++) | |
1197 | p = strchr(p + 1, ' '); | |
b7f954bb SH |
1198 | if (!p) |
1199 | continue; | |
fa454c8e | 1200 | p2 = strchr(p + 1, ' '); |
b7f954bb SH |
1201 | if (!p2) |
1202 | continue; | |
1203 | *p2 = '\0'; | |
fa454c8e | 1204 | if (strcmp(p + 1, "/") == 0) { |
1a0e70ac | 1205 | /* This is '/'. Is it the ramfs? */ |
fa454c8e | 1206 | p = strchr(p2 + 1, '-'); |
b7f954bb | 1207 | if (p && strncmp(p, "- rootfs rootfs ", 16) == 0) { |
fa454c8e | 1208 | free(line); |
b7f954bb | 1209 | fclose(f); |
fa454c8e | 1210 | return true; |
b7f954bb SH |
1211 | } |
1212 | } | |
1213 | } | |
fa454c8e | 1214 | free(line); |
b7f954bb | 1215 | fclose(f); |
fa454c8e | 1216 | return false; |
b7f954bb SH |
1217 | } |
1218 | ||
df6a2945 | 1219 | char *on_path(const char *cmd, const char *rootfs) { |
0e6e3a41 SG |
1220 | char *path = NULL; |
1221 | char *entry = NULL; | |
1222 | char *saveptr = NULL; | |
1223 | char cmdpath[MAXPATHLEN]; | |
1224 | int ret; | |
1225 | ||
1226 | path = getenv("PATH"); | |
1227 | if (!path) | |
8afb3e61 | 1228 | return NULL; |
0e6e3a41 SG |
1229 | |
1230 | path = strdup(path); | |
1231 | if (!path) | |
8afb3e61 | 1232 | return NULL; |
0e6e3a41 SG |
1233 | |
1234 | entry = strtok_r(path, ":", &saveptr); | |
1235 | while (entry) { | |
9d9c111c SH |
1236 | if (rootfs) |
1237 | ret = snprintf(cmdpath, MAXPATHLEN, "%s/%s/%s", rootfs, entry, cmd); | |
1238 | else | |
1239 | ret = snprintf(cmdpath, MAXPATHLEN, "%s/%s", entry, cmd); | |
0e6e3a41 SG |
1240 | |
1241 | if (ret < 0 || ret >= MAXPATHLEN) | |
1242 | goto next_loop; | |
1243 | ||
1244 | if (access(cmdpath, X_OK) == 0) { | |
1245 | free(path); | |
8afb3e61 | 1246 | return strdup(cmdpath); |
0e6e3a41 SG |
1247 | } |
1248 | ||
1249 | next_loop: | |
b707e368 | 1250 | entry = strtok_r(NULL, ":", &saveptr); |
0e6e3a41 SG |
1251 | } |
1252 | ||
1253 | free(path); | |
8afb3e61 | 1254 | return NULL; |
0e6e3a41 | 1255 | } |
76a26f55 SH |
1256 | |
1257 | bool file_exists(const char *f) | |
1258 | { | |
1259 | struct stat statbuf; | |
1260 | ||
1261 | return stat(f, &statbuf) == 0; | |
1262 | } | |
9d9c111c | 1263 | |
12983ba4 SH |
1264 | bool cgns_supported(void) |
1265 | { | |
1266 | return file_exists("/proc/self/ns/cgroup"); | |
1267 | } | |
1268 | ||
9d9c111c SH |
1269 | /* historically lxc-init has been under /usr/lib/lxc and under |
1270 | * /usr/lib/$ARCH/lxc. It now lives as $prefix/sbin/init.lxc. | |
1271 | */ | |
1272 | char *choose_init(const char *rootfs) | |
1273 | { | |
1274 | char *retv = NULL; | |
370ec268 SF |
1275 | const char *empty = "", |
1276 | *tmp; | |
9d9c111c | 1277 | int ret, env_set = 0; |
9d9c111c SH |
1278 | |
1279 | if (!getenv("PATH")) { | |
1280 | if (setenv("PATH", "/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", 0)) | |
1281 | SYSERROR("Failed to setenv"); | |
1282 | env_set = 1; | |
1283 | } | |
1284 | ||
1285 | retv = on_path("init.lxc", rootfs); | |
1286 | ||
1287 | if (env_set) { | |
1288 | if (unsetenv("PATH")) | |
1289 | SYSERROR("Failed to unsetenv"); | |
1290 | } | |
1291 | ||
1292 | if (retv) | |
1293 | return retv; | |
1294 | ||
1295 | retv = malloc(PATH_MAX); | |
1296 | if (!retv) | |
1297 | return NULL; | |
1298 | ||
1299 | if (rootfs) | |
370ec268 | 1300 | tmp = rootfs; |
9d9c111c | 1301 | else |
370ec268 SF |
1302 | tmp = empty; |
1303 | ||
1304 | ret = snprintf(retv, PATH_MAX, "%s/%s/%s", tmp, SBINDIR, "/init.lxc"); | |
9d9c111c SH |
1305 | if (ret < 0 || ret >= PATH_MAX) { |
1306 | ERROR("pathname too long"); | |
1307 | goto out1; | |
1308 | } | |
e57cd7e9 | 1309 | if (access(retv, X_OK) == 0) |
9d9c111c SH |
1310 | return retv; |
1311 | ||
370ec268 | 1312 | ret = snprintf(retv, PATH_MAX, "%s/%s/%s", tmp, LXCINITDIR, "/lxc/lxc-init"); |
9d9c111c SH |
1313 | if (ret < 0 || ret >= PATH_MAX) { |
1314 | ERROR("pathname too long"); | |
1315 | goto out1; | |
1316 | } | |
e57cd7e9 | 1317 | if (access(retv, X_OK) == 0) |
9d9c111c SH |
1318 | return retv; |
1319 | ||
370ec268 | 1320 | ret = snprintf(retv, PATH_MAX, "%s/usr/lib/lxc/lxc-init", tmp); |
9d9c111c SH |
1321 | if (ret < 0 || ret >= PATH_MAX) { |
1322 | ERROR("pathname too long"); | |
1323 | goto out1; | |
1324 | } | |
e57cd7e9 | 1325 | if (access(retv, X_OK) == 0) |
9d9c111c SH |
1326 | return retv; |
1327 | ||
370ec268 | 1328 | ret = snprintf(retv, PATH_MAX, "%s/sbin/lxc-init", tmp); |
9d9c111c SH |
1329 | if (ret < 0 || ret >= PATH_MAX) { |
1330 | ERROR("pathname too long"); | |
1331 | goto out1; | |
1332 | } | |
e57cd7e9 | 1333 | if (access(retv, X_OK) == 0) |
9d9c111c SH |
1334 | return retv; |
1335 | ||
1336 | /* | |
1337 | * Last resort, look for the statically compiled init.lxc which we | |
1338 | * hopefully bind-mounted in. | |
1339 | * If we are called during container setup, and we get to this point, | |
1340 | * then the init.lxc.static from the host will need to be bind-mounted | |
1341 | * in. So we return NULL here to indicate that. | |
1342 | */ | |
1343 | if (rootfs) | |
1344 | goto out1; | |
1345 | ||
1346 | ret = snprintf(retv, PATH_MAX, "/init.lxc.static"); | |
1347 | if (ret < 0 || ret >= PATH_MAX) { | |
1348 | WARN("Nonsense - name /lxc.init.static too long"); | |
1349 | goto out1; | |
1350 | } | |
e57cd7e9 | 1351 | if (access(retv, X_OK) == 0) |
9d9c111c SH |
1352 | return retv; |
1353 | ||
1354 | out1: | |
1355 | free(retv); | |
1356 | return NULL; | |
1357 | } | |
735f2c6e TA |
1358 | |
1359 | int print_to_file(const char *file, const char *content) | |
1360 | { | |
1361 | FILE *f; | |
1362 | int ret = 0; | |
1363 | ||
1364 | f = fopen(file, "w"); | |
1365 | if (!f) | |
1366 | return -1; | |
1367 | if (fprintf(f, "%s", content) != strlen(content)) | |
1368 | ret = -1; | |
1369 | fclose(f); | |
1370 | return ret; | |
1371 | } | |
e1daebd9 SH |
1372 | |
1373 | int is_dir(const char *path) | |
1374 | { | |
1375 | struct stat statbuf; | |
1376 | int ret = stat(path, &statbuf); | |
1377 | if (ret == 0 && S_ISDIR(statbuf.st_mode)) | |
1378 | return 1; | |
1379 | return 0; | |
1380 | } | |
6010a416 SG |
1381 | |
1382 | /* | |
1383 | * Given the '-t' template option to lxc-create, figure out what to | |
1384 | * do. If the template is a full executable path, use that. If it | |
1385 | * is something like 'sshd', then return $templatepath/lxc-sshd. | |
1386 | * On success return the template, on error return NULL. | |
1387 | */ | |
1388 | char *get_template_path(const char *t) | |
1389 | { | |
1390 | int ret, len; | |
1391 | char *tpath; | |
1392 | ||
1393 | if (t[0] == '/' && access(t, X_OK) == 0) { | |
1394 | tpath = strdup(t); | |
1395 | return tpath; | |
1396 | } | |
1397 | ||
1398 | len = strlen(LXCTEMPLATEDIR) + strlen(t) + strlen("/lxc-") + 1; | |
1399 | tpath = malloc(len); | |
1400 | if (!tpath) | |
1401 | return NULL; | |
1402 | ret = snprintf(tpath, len, "%s/lxc-%s", LXCTEMPLATEDIR, t); | |
1403 | if (ret < 0 || ret >= len) { | |
1404 | free(tpath); | |
1405 | return NULL; | |
1406 | } | |
1407 | if (access(tpath, X_OK) < 0) { | |
1408 | SYSERROR("bad template: %s", t); | |
1409 | free(tpath); | |
1410 | return NULL; | |
1411 | } | |
1412 | ||
1413 | return tpath; | |
1414 | } | |
0a4be28d | 1415 | |
592fd47a SH |
1416 | /* |
1417 | * @path: a pathname where / replaced with '\0'. | |
1418 | * @offsetp: pointer to int showing which path segment was last seen. | |
1419 | * Updated on return to reflect the next segment. | |
1420 | * @fulllen: full original path length. | |
1421 | * Returns a pointer to the next path segment, or NULL if done. | |
1422 | */ | |
1423 | static char *get_nextpath(char *path, int *offsetp, int fulllen) | |
1424 | { | |
1425 | int offset = *offsetp; | |
1426 | ||
1427 | if (offset >= fulllen) | |
1428 | return NULL; | |
1429 | ||
1430 | while (path[offset] != '\0' && offset < fulllen) | |
1431 | offset++; | |
1432 | while (path[offset] == '\0' && offset < fulllen) | |
1433 | offset++; | |
1434 | ||
1435 | *offsetp = offset; | |
1436 | return (offset < fulllen) ? &path[offset] : NULL; | |
1437 | } | |
1438 | ||
1439 | /* | |
1440 | * Check that @subdir is a subdir of @dir. @len is the length of | |
1441 | * @dir (to avoid having to recalculate it). | |
1442 | */ | |
1443 | static bool is_subdir(const char *subdir, const char *dir, size_t len) | |
1444 | { | |
1445 | size_t subdirlen = strlen(subdir); | |
1446 | ||
1447 | if (subdirlen < len) | |
1448 | return false; | |
1449 | if (strncmp(subdir, dir, len) != 0) | |
1450 | return false; | |
1451 | if (dir[len-1] == '/') | |
1452 | return true; | |
1453 | if (subdir[len] == '/' || subdirlen == len) | |
1454 | return true; | |
1455 | return false; | |
1456 | } | |
1457 | ||
1458 | /* | |
1459 | * Check if the open fd is a symlink. Return -ELOOP if it is. Return | |
1460 | * -ENOENT if we couldn't fstat. Return 0 if the fd is ok. | |
1461 | */ | |
1462 | static int check_symlink(int fd) | |
1463 | { | |
1464 | struct stat sb; | |
1465 | int ret = fstat(fd, &sb); | |
1466 | if (ret < 0) | |
1467 | return -ENOENT; | |
1468 | if (S_ISLNK(sb.st_mode)) | |
1469 | return -ELOOP; | |
1470 | return 0; | |
1471 | } | |
1472 | ||
1473 | /* | |
1474 | * Open a file or directory, provided that it contains no symlinks. | |
1475 | * | |
1476 | * CAVEAT: This function must not be used for other purposes than container | |
1477 | * setup before executing the container's init | |
1478 | */ | |
1479 | static int open_if_safe(int dirfd, const char *nextpath) | |
1480 | { | |
1481 | int newfd = openat(dirfd, nextpath, O_RDONLY | O_NOFOLLOW); | |
1a0e70ac | 1482 | if (newfd >= 0) /* Was not a symlink, all good. */ |
592fd47a SH |
1483 | return newfd; |
1484 | ||
1485 | if (errno == ELOOP) | |
1486 | return newfd; | |
1487 | ||
1488 | if (errno == EPERM || errno == EACCES) { | |
1a0e70ac CB |
1489 | /* We're not root (cause we got EPERM) so try opening with |
1490 | * O_PATH. | |
1491 | */ | |
592fd47a SH |
1492 | newfd = openat(dirfd, nextpath, O_PATH | O_NOFOLLOW); |
1493 | if (newfd >= 0) { | |
1a0e70ac CB |
1494 | /* O_PATH will return an fd for symlinks. We know |
1495 | * nextpath wasn't a symlink at last openat, so if fd is | |
1496 | * now a link, then something * fishy is going on. | |
592fd47a SH |
1497 | */ |
1498 | int ret = check_symlink(newfd); | |
1499 | if (ret < 0) { | |
1500 | close(newfd); | |
1501 | newfd = ret; | |
1502 | } | |
1503 | } | |
1504 | } | |
1505 | ||
1506 | return newfd; | |
1507 | } | |
1508 | ||
1509 | /* | |
1510 | * Open a path intending for mounting, ensuring that the final path | |
1511 | * is inside the container's rootfs. | |
1512 | * | |
1513 | * CAVEAT: This function must not be used for other purposes than container | |
1514 | * setup before executing the container's init | |
1515 | * | |
1516 | * @target: path to be opened | |
1517 | * @prefix_skip: a part of @target in which to ignore symbolic links. This | |
1518 | * would be the container's rootfs. | |
1519 | * | |
1520 | * Return an open fd for the path, or <0 on error. | |
1521 | */ | |
1522 | static int open_without_symlink(const char *target, const char *prefix_skip) | |
1523 | { | |
1524 | int curlen = 0, dirfd, fulllen, i; | |
1525 | char *dup = NULL; | |
1526 | ||
1527 | fulllen = strlen(target); | |
1528 | ||
1529 | /* make sure prefix-skip makes sense */ | |
01074e5b | 1530 | if (prefix_skip && strlen(prefix_skip) > 0) { |
592fd47a SH |
1531 | curlen = strlen(prefix_skip); |
1532 | if (!is_subdir(target, prefix_skip, curlen)) { | |
1533 | ERROR("WHOA there - target '%s' didn't start with prefix '%s'", | |
1534 | target, prefix_skip); | |
1535 | return -EINVAL; | |
1536 | } | |
1537 | /* | |
1538 | * get_nextpath() expects the curlen argument to be | |
1539 | * on a (turned into \0) / or before it, so decrement | |
1540 | * curlen to make sure that happens | |
1541 | */ | |
1542 | if (curlen) | |
1543 | curlen--; | |
1544 | } else { | |
1545 | prefix_skip = "/"; | |
1546 | curlen = 0; | |
1547 | } | |
1548 | ||
1549 | /* Make a copy of target which we can hack up, and tokenize it */ | |
1550 | if ((dup = strdup(target)) == NULL) { | |
1551 | SYSERROR("Out of memory checking for symbolic link"); | |
1552 | return -ENOMEM; | |
1553 | } | |
1554 | for (i = 0; i < fulllen; i++) { | |
1555 | if (dup[i] == '/') | |
1556 | dup[i] = '\0'; | |
1557 | } | |
1558 | ||
1559 | dirfd = open(prefix_skip, O_RDONLY); | |
1560 | if (dirfd < 0) | |
1561 | goto out; | |
1562 | while (1) { | |
1563 | int newfd, saved_errno; | |
1564 | char *nextpath; | |
1565 | ||
1566 | if ((nextpath = get_nextpath(dup, &curlen, fulllen)) == NULL) | |
1567 | goto out; | |
1568 | newfd = open_if_safe(dirfd, nextpath); | |
1569 | saved_errno = errno; | |
1570 | close(dirfd); | |
1571 | dirfd = newfd; | |
1572 | if (newfd < 0) { | |
1573 | errno = saved_errno; | |
1574 | if (errno == ELOOP) | |
1575 | SYSERROR("%s in %s was a symbolic link!", nextpath, target); | |
592fd47a SH |
1576 | goto out; |
1577 | } | |
1578 | } | |
1579 | ||
1580 | out: | |
1581 | free(dup); | |
1582 | return dirfd; | |
1583 | } | |
1584 | ||
1585 | /* | |
1586 | * Safely mount a path into a container, ensuring that the mount target | |
1587 | * is under the container's @rootfs. (If @rootfs is NULL, then the container | |
1588 | * uses the host's /) | |
1589 | * | |
1590 | * CAVEAT: This function must not be used for other purposes than container | |
1591 | * setup before executing the container's init | |
1592 | */ | |
1593 | int safe_mount(const char *src, const char *dest, const char *fstype, | |
1594 | unsigned long flags, const void *data, const char *rootfs) | |
1595 | { | |
1a0e70ac CB |
1596 | int destfd, ret, saved_errno; |
1597 | /* Only needs enough for /proc/self/fd/<fd>. */ | |
1598 | char srcbuf[50], destbuf[50]; | |
1599 | int srcfd = -1; | |
592fd47a SH |
1600 | const char *mntsrc = src; |
1601 | ||
1602 | if (!rootfs) | |
1603 | rootfs = ""; | |
1604 | ||
1605 | /* todo - allow symlinks for relative paths if 'allowsymlinks' option is passed */ | |
1606 | if (flags & MS_BIND && src && src[0] != '/') { | |
1607 | INFO("this is a relative bind mount"); | |
1608 | srcfd = open_without_symlink(src, NULL); | |
1609 | if (srcfd < 0) | |
1610 | return srcfd; | |
1611 | ret = snprintf(srcbuf, 50, "/proc/self/fd/%d", srcfd); | |
1612 | if (ret < 0 || ret > 50) { | |
1613 | close(srcfd); | |
1614 | ERROR("Out of memory"); | |
1615 | return -EINVAL; | |
1616 | } | |
1617 | mntsrc = srcbuf; | |
1618 | } | |
1619 | ||
1620 | destfd = open_without_symlink(dest, rootfs); | |
1621 | if (destfd < 0) { | |
88e078ba CB |
1622 | if (srcfd != -1) { |
1623 | saved_errno = errno; | |
592fd47a | 1624 | close(srcfd); |
88e078ba CB |
1625 | errno = saved_errno; |
1626 | } | |
592fd47a SH |
1627 | return destfd; |
1628 | } | |
1629 | ||
1630 | ret = snprintf(destbuf, 50, "/proc/self/fd/%d", destfd); | |
1631 | if (ret < 0 || ret > 50) { | |
1632 | if (srcfd != -1) | |
1633 | close(srcfd); | |
1634 | close(destfd); | |
1635 | ERROR("Out of memory"); | |
1636 | return -EINVAL; | |
1637 | } | |
1638 | ||
1639 | ret = mount(mntsrc, destbuf, fstype, flags, data); | |
1640 | saved_errno = errno; | |
1641 | if (srcfd != -1) | |
1642 | close(srcfd); | |
1643 | close(destfd); | |
1644 | if (ret < 0) { | |
1645 | errno = saved_errno; | |
1646 | SYSERROR("Failed to mount %s onto %s", src, dest); | |
1647 | return ret; | |
1648 | } | |
1649 | ||
1650 | return 0; | |
1651 | } | |
1652 | ||
ced03a01 SH |
1653 | /* |
1654 | * Mount a proc under @rootfs if proc self points to a pid other than | |
1655 | * my own. This is needed to have a known-good proc mount for setting | |
1656 | * up LSMs both at container startup and attach. | |
1657 | * | |
1658 | * @rootfs : the rootfs where proc should be mounted | |
1659 | * | |
1660 | * Returns < 0 on failure, 0 if the correct proc was already mounted | |
1661 | * and 1 if a new proc was mounted. | |
f267d666 BP |
1662 | * |
1663 | * NOTE: not to be called from inside the container namespace! | |
ced03a01 | 1664 | */ |
943144d9 | 1665 | int lxc_mount_proc_if_needed(const char *rootfs) |
ced03a01 SH |
1666 | { |
1667 | char path[MAXPATHLEN]; | |
6b1ba5d6 CB |
1668 | int link_to_pid, linklen, mypid, ret; |
1669 | char link[LXC_NUMSTRLEN64] = {0}; | |
ced03a01 SH |
1670 | |
1671 | ret = snprintf(path, MAXPATHLEN, "%s/proc/self", rootfs); | |
1672 | if (ret < 0 || ret >= MAXPATHLEN) { | |
1673 | SYSERROR("proc path name too long"); | |
1674 | return -1; | |
1675 | } | |
fc2ad9dc | 1676 | |
6b1ba5d6 | 1677 | linklen = readlink(path, link, LXC_NUMSTRLEN64); |
fc2ad9dc | 1678 | |
ced03a01 | 1679 | ret = snprintf(path, MAXPATHLEN, "%s/proc", rootfs); |
d539a2b2 CB |
1680 | if (ret < 0 || ret >= MAXPATHLEN) { |
1681 | SYSERROR("proc path name too long"); | |
1682 | return -1; | |
1683 | } | |
fc2ad9dc CB |
1684 | |
1685 | /* /proc not mounted */ | |
1686 | if (linklen < 0) { | |
1687 | if (mkdir(path, 0755) && errno != EEXIST) | |
1688 | return -1; | |
ced03a01 | 1689 | goto domount; |
6b1ba5d6 CB |
1690 | } else if (linklen >= LXC_NUMSTRLEN64) { |
1691 | link[linklen - 1] = '\0'; | |
1692 | ERROR("readlink returned truncated content: \"%s\"", link); | |
1693 | return -1; | |
fc2ad9dc CB |
1694 | } |
1695 | ||
6b1ba5d6 CB |
1696 | mypid = getpid(); |
1697 | INFO("I am %d, /proc/self points to \"%s\"", mypid, link); | |
1698 | ||
2d036cca CB |
1699 | if (lxc_safe_int(link, &link_to_pid) < 0) |
1700 | return -1; | |
fc2ad9dc | 1701 | |
6b1ba5d6 CB |
1702 | /* correct procfs is already mounted */ |
1703 | if (link_to_pid == mypid) | |
1704 | return 0; | |
fc2ad9dc | 1705 | |
6b1ba5d6 CB |
1706 | ret = umount2(path, MNT_DETACH); |
1707 | if (ret < 0) | |
1708 | WARN("failed to umount \"%s\" with MNT_DETACH", path); | |
ced03a01 SH |
1709 | |
1710 | domount: | |
fc2ad9dc | 1711 | /* rootfs is NULL */ |
6b1ba5d6 | 1712 | if (!strcmp(rootfs, "")) |
f267d666 BP |
1713 | ret = mount("proc", path, "proc", 0, NULL); |
1714 | else | |
1715 | ret = safe_mount("proc", path, "proc", 0, NULL, rootfs); | |
f267d666 | 1716 | if (ret < 0) |
ced03a01 | 1717 | return -1; |
f267d666 | 1718 | |
fc2ad9dc | 1719 | INFO("mounted /proc in container for security transition"); |
ced03a01 SH |
1720 | return 1; |
1721 | } | |
69aeabac | 1722 | |
f8dd0275 | 1723 | int open_devnull(void) |
69aeabac | 1724 | { |
f8dd0275 AM |
1725 | int fd = open("/dev/null", O_RDWR); |
1726 | ||
1727 | if (fd < 0) | |
1728 | SYSERROR("Can't open /dev/null"); | |
1729 | ||
1730 | return fd; | |
1731 | } | |
69aeabac | 1732 | |
f8dd0275 AM |
1733 | int set_stdfds(int fd) |
1734 | { | |
bbbf65ee CB |
1735 | int ret; |
1736 | ||
69aeabac TA |
1737 | if (fd < 0) |
1738 | return -1; | |
1739 | ||
bbbf65ee CB |
1740 | ret = dup2(fd, STDIN_FILENO); |
1741 | if (ret < 0) | |
f8dd0275 | 1742 | return -1; |
bbbf65ee CB |
1743 | |
1744 | ret = dup2(fd, STDOUT_FILENO); | |
1745 | if (ret < 0) | |
f8dd0275 | 1746 | return -1; |
bbbf65ee CB |
1747 | |
1748 | ret = dup2(fd, STDERR_FILENO); | |
1749 | if (ret < 0) | |
f8dd0275 AM |
1750 | return -1; |
1751 | ||
1752 | return 0; | |
1753 | } | |
1754 | ||
1755 | int null_stdfds(void) | |
1756 | { | |
1757 | int ret = -1; | |
1758 | int fd = open_devnull(); | |
1759 | ||
1760 | if (fd >= 0) { | |
1761 | ret = set_stdfds(fd); | |
1762 | close(fd); | |
1763 | } | |
69aeabac | 1764 | |
69aeabac TA |
1765 | return ret; |
1766 | } | |
ccb4cabe SH |
1767 | |
1768 | /* | |
1769 | * Return the number of lines in file @fn, or -1 on error | |
1770 | */ | |
1771 | int lxc_count_file_lines(const char *fn) | |
1772 | { | |
1773 | FILE *f; | |
1774 | char *line = NULL; | |
1775 | size_t sz = 0; | |
1776 | int n = 0; | |
1777 | ||
1778 | f = fopen_cloexec(fn, "r"); | |
1779 | if (!f) | |
1780 | return -1; | |
1781 | ||
1782 | while (getline(&line, &sz, f) != -1) { | |
1783 | n++; | |
1784 | } | |
1785 | free(line); | |
1786 | fclose(f); | |
1787 | return n; | |
1788 | } | |
1adbd020 | 1789 | |
25086a5f CB |
1790 | void *lxc_strmmap(void *addr, size_t length, int prot, int flags, int fd, |
1791 | off_t offset) | |
1adbd020 CB |
1792 | { |
1793 | void *tmp = NULL, *overlap = NULL; | |
1794 | ||
1795 | /* We establish an anonymous mapping that is one byte larger than the | |
1796 | * underlying file. The pages handed to us are zero filled. */ | |
1797 | tmp = mmap(addr, length + 1, PROT_READ, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); | |
1798 | if (tmp == MAP_FAILED) | |
a1e5280d | 1799 | return tmp; |
1adbd020 CB |
1800 | |
1801 | /* Now we establish a fixed-address mapping starting at the address we | |
1802 | * received from our anonymous mapping and replace all bytes excluding | |
1803 | * the additional \0-byte with the file. This allows us to use normal | |
a1e5280d | 1804 | * string-handling functions. */ |
1adbd020 CB |
1805 | overlap = mmap(tmp, length, prot, MAP_FIXED | flags, fd, offset); |
1806 | if (overlap == MAP_FAILED) | |
a1e5280d | 1807 | munmap(tmp, length + 1); |
1adbd020 | 1808 | |
1adbd020 CB |
1809 | return overlap; |
1810 | } | |
1811 | ||
25086a5f | 1812 | int lxc_strmunmap(void *addr, size_t length) |
1adbd020 CB |
1813 | { |
1814 | return munmap(addr, length + 1); | |
1815 | } | |
330ae3d3 CB |
1816 | |
1817 | /* Check whether a signal is blocked by a process. */ | |
de3c491b | 1818 | /* /proc/pid-to-str/status\0 = (5 + 21 + 7 + 1) */ |
eab15c1e | 1819 | #define __PROC_STATUS_LEN (5 + (LXC_NUMSTRLEN64) + 7 + 1) |
330ae3d3 CB |
1820 | bool task_blocking_signal(pid_t pid, int signal) |
1821 | { | |
1822 | bool bret = false; | |
1823 | char *line = NULL; | |
1824 | long unsigned int sigblk = 0; | |
1825 | size_t n = 0; | |
1826 | int ret; | |
1827 | FILE *f; | |
1828 | ||
de3c491b | 1829 | char status[__PROC_STATUS_LEN]; |
330ae3d3 | 1830 | |
de3c491b CB |
1831 | ret = snprintf(status, __PROC_STATUS_LEN, "/proc/%d/status", pid); |
1832 | if (ret < 0 || ret >= __PROC_STATUS_LEN) | |
330ae3d3 CB |
1833 | return bret; |
1834 | ||
1835 | f = fopen(status, "r"); | |
1836 | if (!f) | |
1837 | return bret; | |
1838 | ||
1839 | while (getline(&line, &n, f) != -1) { | |
6fbcbe3b CB |
1840 | if (strncmp(line, "SigBlk:\t", 8)) |
1841 | continue; | |
1842 | ||
1843 | if (sscanf(line + 8, "%lx", &sigblk) != 1) | |
1844 | goto out; | |
330ae3d3 CB |
1845 | } |
1846 | ||
6fbcbe3b | 1847 | if (sigblk & (1LU << (signal - 1))) |
330ae3d3 CB |
1848 | bret = true; |
1849 | ||
1850 | out: | |
1851 | free(line); | |
1852 | fclose(f); | |
1853 | return bret; | |
1854 | } | |
000dfda7 CB |
1855 | |
1856 | static int lxc_append_null_to_list(void ***list) | |
1857 | { | |
1858 | int newentry = 0; | |
1859 | void **tmp; | |
1860 | ||
1861 | if (*list) | |
1862 | for (; (*list)[newentry]; newentry++) { | |
1863 | ; | |
1864 | } | |
1865 | ||
1866 | tmp = realloc(*list, (newentry + 2) * sizeof(void **)); | |
1867 | if (!tmp) | |
1868 | return -1; | |
1869 | ||
1870 | *list = tmp; | |
1871 | (*list)[newentry + 1] = NULL; | |
1872 | ||
1873 | return newentry; | |
1874 | } | |
1875 | ||
1876 | int lxc_append_string(char ***list, char *entry) | |
1877 | { | |
000dfda7 | 1878 | char *copy; |
a54694f8 CB |
1879 | int newentry; |
1880 | ||
1881 | newentry = lxc_append_null_to_list((void ***)list); | |
1882 | if (newentry < 0) | |
1883 | return -1; | |
000dfda7 CB |
1884 | |
1885 | copy = strdup(entry); | |
1886 | if (!copy) | |
1887 | return -1; | |
1888 | ||
1889 | (*list)[newentry] = copy; | |
1890 | ||
1891 | return 0; | |
1892 | } | |
a687256f CB |
1893 | |
1894 | int lxc_preserve_ns(const int pid, const char *ns) | |
1895 | { | |
1896 | int ret; | |
a052913d CB |
1897 | /* 5 /proc + 21 /int_as_str + 3 /ns + 20 /NS_NAME + 1 \0 */ |
1898 | #define __NS_PATH_LEN 50 | |
1899 | char path[__NS_PATH_LEN]; | |
a687256f | 1900 | |
4d8ac866 CB |
1901 | /* This way we can use this function to also check whether namespaces |
1902 | * are supported by the kernel by passing in the NULL or the empty | |
1903 | * string. | |
1904 | */ | |
a052913d | 1905 | ret = snprintf(path, __NS_PATH_LEN, "/proc/%d/ns%s%s", pid, |
4d8ac866 CB |
1906 | !ns || strcmp(ns, "") == 0 ? "" : "/", |
1907 | !ns || strcmp(ns, "") == 0 ? "" : ns); | |
a052913d | 1908 | if (ret < 0 || (size_t)ret >= __NS_PATH_LEN) |
a687256f CB |
1909 | return -1; |
1910 | ||
1911 | return open(path, O_RDONLY | O_CLOEXEC); | |
1912 | } | |
6bc2eafe CB |
1913 | |
1914 | int lxc_safe_uint(const char *numstr, unsigned int *converted) | |
1915 | { | |
1916 | char *err = NULL; | |
1917 | unsigned long int uli; | |
1918 | ||
643c1984 CB |
1919 | while (isspace(*numstr)) |
1920 | numstr++; | |
1921 | ||
1922 | if (*numstr == '-') | |
1923 | return -EINVAL; | |
1924 | ||
6bc2eafe CB |
1925 | errno = 0; |
1926 | uli = strtoul(numstr, &err, 0); | |
643c1984 | 1927 | if (errno == ERANGE && uli == ULONG_MAX) |
ff0e49c7 | 1928 | return -ERANGE; |
6bc2eafe | 1929 | |
643c1984 | 1930 | if (err == numstr || *err != '\0') |
6bc2eafe CB |
1931 | return -EINVAL; |
1932 | ||
1933 | if (uli > UINT_MAX) | |
1934 | return -ERANGE; | |
1935 | ||
8c57d930 | 1936 | *converted = (unsigned int)uli; |
6bc2eafe CB |
1937 | return 0; |
1938 | } | |
b5f845e7 | 1939 | |
681188c1 CB |
1940 | int lxc_safe_ulong(const char *numstr, unsigned long *converted) |
1941 | { | |
1942 | char *err = NULL; | |
1943 | unsigned long int uli; | |
1944 | ||
1945 | while (isspace(*numstr)) | |
1946 | numstr++; | |
1947 | ||
1948 | if (*numstr == '-') | |
1949 | return -EINVAL; | |
1950 | ||
1951 | errno = 0; | |
1952 | uli = strtoul(numstr, &err, 0); | |
1953 | if (errno == ERANGE && uli == ULONG_MAX) | |
1954 | return -ERANGE; | |
1955 | ||
1956 | if (err == numstr || *err != '\0') | |
1957 | return -EINVAL; | |
1958 | ||
1959 | *converted = uli; | |
1960 | return 0; | |
1961 | } | |
1962 | ||
b5f845e7 CB |
1963 | int lxc_safe_int(const char *numstr, int *converted) |
1964 | { | |
1965 | char *err = NULL; | |
1966 | signed long int sli; | |
1967 | ||
1968 | errno = 0; | |
1969 | sli = strtol(numstr, &err, 0); | |
643c1984 | 1970 | if (errno == ERANGE && (sli == LONG_MAX || sli == LONG_MIN)) |
ff0e49c7 | 1971 | return -ERANGE; |
643c1984 CB |
1972 | |
1973 | if (errno != 0 && sli == 0) | |
ff0e49c7 | 1974 | return -EINVAL; |
b5f845e7 | 1975 | |
643c1984 | 1976 | if (err == numstr || *err != '\0') |
b5f845e7 CB |
1977 | return -EINVAL; |
1978 | ||
643c1984 | 1979 | if (sli > INT_MAX || sli < INT_MIN) |
b5f845e7 CB |
1980 | return -ERANGE; |
1981 | ||
1982 | *converted = (int)sli; | |
1983 | return 0; | |
1984 | } | |
8c57d930 CB |
1985 | |
1986 | int lxc_safe_long(const char *numstr, long int *converted) | |
1987 | { | |
1988 | char *err = NULL; | |
1989 | signed long int sli; | |
1990 | ||
1991 | errno = 0; | |
1992 | sli = strtol(numstr, &err, 0); | |
643c1984 | 1993 | if (errno == ERANGE && (sli == LONG_MAX || sli == LONG_MIN)) |
ff0e49c7 | 1994 | return -ERANGE; |
8c57d930 | 1995 | |
643c1984 | 1996 | if (errno != 0 && sli == 0) |
ff0e49c7 | 1997 | return -EINVAL; |
8c57d930 | 1998 | |
643c1984 CB |
1999 | if (err == numstr || *err != '\0') |
2000 | return -EINVAL; | |
8c57d930 CB |
2001 | |
2002 | *converted = sli; | |
2003 | return 0; | |
2004 | } | |
dbaf55a3 CB |
2005 | |
2006 | int lxc_switch_uid_gid(uid_t uid, gid_t gid) | |
2007 | { | |
2008 | if (setgid(gid) < 0) { | |
2009 | SYSERROR("Failed to switch to gid %d.", gid); | |
2010 | return -errno; | |
2011 | } | |
2012 | NOTICE("Switched to gid %d.", gid); | |
2013 | ||
2014 | if (setuid(uid) < 0) { | |
2015 | SYSERROR("Failed to switch to uid %d.", uid); | |
2016 | return -errno; | |
2017 | } | |
2018 | NOTICE("Switched to uid %d.", uid); | |
2019 | ||
2020 | return 0; | |
2021 | } | |
2022 | ||
2023 | /* Simple covenience function which enables uniform logging. */ | |
2024 | int lxc_setgroups(int size, gid_t list[]) | |
2025 | { | |
2026 | if (setgroups(size, list) < 0) { | |
2027 | SYSERROR("Failed to setgroups()."); | |
2028 | return -errno; | |
2029 | } | |
2030 | NOTICE("Dropped additional groups."); | |
2031 | ||
2032 | return 0; | |
2033 | } | |
c6868a1f CB |
2034 | |
2035 | static int lxc_get_unused_loop_dev_legacy(char *loop_name) | |
2036 | { | |
2037 | struct dirent *dp; | |
2038 | struct loop_info64 lo64; | |
2039 | DIR *dir; | |
2040 | int dfd = -1, fd = -1, ret = -1; | |
2041 | ||
2042 | dir = opendir("/dev"); | |
2043 | if (!dir) | |
2044 | return -1; | |
2045 | ||
2046 | while ((dp = readdir(dir))) { | |
2047 | if (!dp) | |
2048 | break; | |
2049 | ||
2050 | if (strncmp(dp->d_name, "loop", 4) != 0) | |
2051 | continue; | |
2052 | ||
2053 | dfd = dirfd(dir); | |
2054 | if (dfd < 0) | |
2055 | continue; | |
2056 | ||
2057 | fd = openat(dfd, dp->d_name, O_RDWR); | |
2058 | if (fd < 0) | |
2059 | continue; | |
2060 | ||
2061 | ret = ioctl(fd, LOOP_GET_STATUS64, &lo64); | |
2062 | if (ret < 0) { | |
2063 | if (ioctl(fd, LOOP_GET_STATUS64, &lo64) == 0 || | |
2064 | errno != ENXIO) { | |
2065 | close(fd); | |
2066 | fd = -1; | |
2067 | continue; | |
2068 | } | |
2069 | } | |
2070 | ||
2071 | ret = snprintf(loop_name, LO_NAME_SIZE, "/dev/%s", dp->d_name); | |
2072 | if (ret < 0 || ret >= LO_NAME_SIZE) { | |
2073 | close(fd); | |
2074 | fd = -1; | |
2075 | continue; | |
2076 | } | |
2077 | ||
2078 | break; | |
2079 | } | |
2080 | ||
2081 | closedir(dir); | |
2082 | ||
2083 | if (fd < 0) | |
2084 | return -1; | |
2085 | ||
2086 | return fd; | |
2087 | } | |
2088 | ||
2089 | static int lxc_get_unused_loop_dev(char *name_loop) | |
2090 | { | |
2091 | int loop_nr, ret; | |
2092 | int fd_ctl = -1, fd_tmp = -1; | |
2093 | ||
2094 | fd_ctl = open("/dev/loop-control", O_RDWR | O_CLOEXEC); | |
2095 | if (fd_ctl < 0) | |
2096 | return -ENODEV; | |
2097 | ||
2098 | loop_nr = ioctl(fd_ctl, LOOP_CTL_GET_FREE); | |
2099 | if (loop_nr < 0) | |
2100 | goto on_error; | |
2101 | ||
2102 | ret = snprintf(name_loop, LO_NAME_SIZE, "/dev/loop%d", loop_nr); | |
2103 | if (ret < 0 || ret >= LO_NAME_SIZE) | |
2104 | goto on_error; | |
2105 | ||
2106 | fd_tmp = open(name_loop, O_RDWR | O_CLOEXEC); | |
2107 | if (fd_tmp < 0) | |
2108 | goto on_error; | |
2109 | ||
2110 | on_error: | |
2111 | close(fd_ctl); | |
2112 | return fd_tmp; | |
2113 | } | |
2114 | ||
2115 | int lxc_prepare_loop_dev(const char *source, char *loop_dev, int flags) | |
2116 | { | |
2117 | int ret; | |
2118 | struct loop_info64 lo64; | |
2119 | int fd_img = -1, fret = -1, fd_loop = -1; | |
2120 | ||
2121 | fd_loop = lxc_get_unused_loop_dev(loop_dev); | |
2122 | if (fd_loop < 0) { | |
2123 | if (fd_loop == -ENODEV) | |
2124 | fd_loop = lxc_get_unused_loop_dev_legacy(loop_dev); | |
2125 | else | |
2126 | goto on_error; | |
2127 | } | |
2128 | ||
2129 | fd_img = open(source, O_RDWR | O_CLOEXEC); | |
2130 | if (fd_img < 0) | |
2131 | goto on_error; | |
2132 | ||
2133 | ret = ioctl(fd_loop, LOOP_SET_FD, fd_img); | |
2134 | if (ret < 0) | |
2135 | goto on_error; | |
2136 | ||
2137 | memset(&lo64, 0, sizeof(lo64)); | |
2138 | lo64.lo_flags = flags; | |
2139 | ||
2140 | ret = ioctl(fd_loop, LOOP_SET_STATUS64, &lo64); | |
2141 | if (ret < 0) | |
2142 | goto on_error; | |
2143 | ||
2144 | fret = 0; | |
2145 | ||
2146 | on_error: | |
2147 | if (fd_img >= 0) | |
2148 | close(fd_img); | |
2149 | ||
2150 | if (fret < 0 && fd_loop >= 0) { | |
2151 | close(fd_loop); | |
2152 | fd_loop = -1; | |
2153 | } | |
2154 | ||
2155 | return fd_loop; | |
2156 | } | |
74251e49 CB |
2157 | |
2158 | int lxc_unstack_mountpoint(const char *path, bool lazy) | |
2159 | { | |
2160 | int ret; | |
2161 | int umounts = 0; | |
2162 | ||
2163 | pop_stack: | |
2164 | ret = umount2(path, lazy ? MNT_DETACH : 0); | |
2165 | if (ret < 0) { | |
2166 | /* We consider anything else than EINVAL deadly to prevent going | |
2167 | * into an infinite loop. (The other alternative is constantly | |
2168 | * parsing /proc/self/mountinfo which is yucky and probably | |
2169 | * racy.) | |
2170 | */ | |
2171 | if (errno != EINVAL) | |
2172 | return -errno; | |
2173 | } else { | |
b4a40f7b CB |
2174 | /* Just stop counting when this happens. That'd just be so |
2175 | * stupid that we won't even bother trying to report back the | |
2176 | * correct value anymore. | |
2177 | */ | |
2178 | if (umounts != INT_MAX) | |
2179 | umounts++; | |
74251e49 CB |
2180 | /* We succeeded in umounting. Make sure that there's no other |
2181 | * mountpoint stacked underneath. | |
2182 | */ | |
74251e49 CB |
2183 | goto pop_stack; |
2184 | } | |
2185 | ||
2186 | return umounts; | |
2187 | } | |
ea3a694f CB |
2188 | |
2189 | int run_command(char *buf, size_t buf_size, int (*child_fn)(void *), void *args) | |
2190 | { | |
2191 | pid_t child; | |
2192 | int ret, fret, pipefd[2]; | |
2193 | ssize_t bytes; | |
2194 | ||
2195 | /* Make sure our callers do not receive unitialized memory. */ | |
2196 | if (buf_size > 0 && buf) | |
2197 | buf[0] = '\0'; | |
2198 | ||
2199 | if (pipe(pipefd) < 0) { | |
2200 | SYSERROR("failed to create pipe"); | |
2201 | return -1; | |
2202 | } | |
2203 | ||
2204 | child = fork(); | |
2205 | if (child < 0) { | |
2206 | close(pipefd[0]); | |
2207 | close(pipefd[1]); | |
2208 | SYSERROR("failed to create new process"); | |
2209 | return -1; | |
2210 | } | |
2211 | ||
2212 | if (child == 0) { | |
2213 | /* Close the read-end of the pipe. */ | |
2214 | close(pipefd[0]); | |
2215 | ||
2216 | /* Redirect std{err,out} to write-end of the | |
2217 | * pipe. | |
2218 | */ | |
2219 | ret = dup2(pipefd[1], STDOUT_FILENO); | |
2220 | if (ret >= 0) | |
2221 | ret = dup2(pipefd[1], STDERR_FILENO); | |
2222 | ||
2223 | /* Close the write-end of the pipe. */ | |
2224 | close(pipefd[1]); | |
2225 | ||
2226 | if (ret < 0) { | |
2227 | SYSERROR("failed to duplicate std{err,out} file descriptor"); | |
2228 | exit(EXIT_FAILURE); | |
2229 | } | |
2230 | ||
2231 | /* Does not return. */ | |
2232 | child_fn(args); | |
2233 | ERROR("failed to exec command"); | |
2234 | exit(EXIT_FAILURE); | |
2235 | } | |
2236 | ||
2237 | /* close the write-end of the pipe */ | |
2238 | close(pipefd[1]); | |
2239 | ||
7a643c7c CB |
2240 | if (buf && buf_size > 0) { |
2241 | bytes = read(pipefd[0], buf, buf_size - 1); | |
2242 | if (bytes > 0) | |
2243 | buf[bytes - 1] = '\0'; | |
2244 | } | |
ea3a694f CB |
2245 | |
2246 | fret = wait_for_pid(child); | |
2247 | /* close the read-end of the pipe */ | |
2248 | close(pipefd[0]); | |
2249 | ||
2250 | return fret; | |
2251 | } | |
04ad7ffe CB |
2252 | |
2253 | char *must_make_path(const char *first, ...) | |
2254 | { | |
2255 | va_list args; | |
2256 | char *cur, *dest; | |
2257 | size_t full_len = strlen(first); | |
2258 | ||
2259 | dest = must_copy_string(first); | |
2260 | ||
2261 | va_start(args, first); | |
2262 | while ((cur = va_arg(args, char *)) != NULL) { | |
2263 | full_len += strlen(cur); | |
2264 | if (cur[0] != '/') | |
2265 | full_len++; | |
2266 | dest = must_realloc(dest, full_len + 1); | |
2267 | if (cur[0] != '/') | |
2268 | strcat(dest, "/"); | |
2269 | strcat(dest, cur); | |
2270 | } | |
2271 | va_end(args); | |
2272 | ||
2273 | return dest; | |
2274 | } | |
2275 | ||
2276 | char *must_copy_string(const char *entry) | |
2277 | { | |
2278 | char *ret; | |
2279 | ||
2280 | if (!entry) | |
2281 | return NULL; | |
2282 | do { | |
2283 | ret = strdup(entry); | |
2284 | } while (!ret); | |
2285 | ||
2286 | return ret; | |
2287 | } | |
2288 | ||
2289 | void *must_realloc(void *orig, size_t sz) | |
2290 | { | |
2291 | void *ret; | |
2292 | ||
2293 | do { | |
2294 | ret = realloc(orig, sz); | |
2295 | } while (!ret); | |
2296 | ||
2297 | return ret; | |
2298 | } | |
a035c53a CB |
2299 | |
2300 | bool is_fs_type(const struct statfs *fs, fs_type_magic magic_val) | |
2301 | { | |
2302 | return (fs->f_type == (fs_type_magic)magic_val); | |
2303 | } | |
2304 | ||
2305 | bool has_fs_type(const char *path, fs_type_magic magic_val) | |
2306 | { | |
2307 | bool has_type; | |
2308 | int ret; | |
2309 | struct statfs sb; | |
2310 | ||
2311 | ret = statfs(path, &sb); | |
2312 | if (ret < 0) | |
2313 | return false; | |
2314 | ||
2315 | has_type = is_fs_type(&sb, magic_val); | |
2316 | if (!has_type && magic_val == RAMFS_MAGIC) | |
2317 | WARN("When the ramfs it a tmpfs statfs() might report tmpfs"); | |
2318 | ||
2319 | return has_type; | |
2320 | } | |
d75c14e2 CB |
2321 | |
2322 | bool lxc_nic_exists(char *nic) | |
2323 | { | |
2324 | #define __LXC_SYS_CLASS_NET_LEN 15 + IFNAMSIZ + 1 | |
2325 | char path[__LXC_SYS_CLASS_NET_LEN]; | |
2326 | int ret; | |
2327 | struct stat sb; | |
2328 | ||
2329 | if (!strcmp(nic, "none")) | |
2330 | return true; | |
2331 | ||
2332 | ret = snprintf(path, __LXC_SYS_CLASS_NET_LEN, "/sys/class/net/%s", nic); | |
2333 | if (ret < 0 || (size_t)ret >= __LXC_SYS_CLASS_NET_LEN) | |
2334 | return false; | |
2335 | ||
2336 | ret = stat(path, &sb); | |
2337 | if (ret < 0) | |
2338 | return false; | |
2339 | ||
2340 | return true; | |
2341 | } | |
127c6e70 CB |
2342 | |
2343 | int lxc_make_tmpfile(char *template, bool rm) | |
2344 | { | |
2345 | int fd, ret; | |
2346 | ||
2347 | fd = mkstemp(template); | |
2348 | if (fd < 0) | |
2349 | return -1; | |
2350 | ||
2351 | if (!rm) | |
2352 | return fd; | |
2353 | ||
2354 | ret = unlink(template); | |
2355 | if (ret < 0) { | |
2356 | close(fd); | |
2357 | return -1; | |
2358 | } | |
2359 | ||
2360 | return fd; | |
2361 | } |