]>
Commit | Line | Data |
---|---|---|
e0732705 CS |
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> |
e0732705 CS |
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 |
e0732705 CS |
22 | */ |
23 | ||
24 | #define _GNU_SOURCE | |
25 | #include <unistd.h> | |
26 | #include <stdio.h> | |
27 | #include <string.h> | |
28 | #include <stdlib.h> | |
2eef2bda | 29 | #include <signal.h> |
e0732705 CS |
30 | #include <errno.h> |
31 | #include <fcntl.h> | |
c476bdce | 32 | #include <grp.h> |
e0732705 CS |
33 | #include <sys/param.h> |
34 | #include <sys/prctl.h> | |
7a0b0b56 | 35 | #include <sys/mount.h> |
5ec27989 | 36 | #include <sys/socket.h> |
1ba0013f | 37 | #include <sys/syscall.h> |
905022f7 | 38 | #include <sys/wait.h> |
910bb4fa | 39 | #include <linux/unistd.h> |
905022f7 | 40 | #include <pwd.h> |
e0732705 CS |
41 | |
42 | #if !HAVE_DECL_PR_CAPBSET_DROP | |
43 | #define PR_CAPBSET_DROP 24 | |
44 | #endif | |
45 | ||
46 | #include "namespace.h" | |
47 | #include "log.h" | |
48 | #include "attach.h" | |
49 | #include "caps.h" | |
e0732705 | 50 | #include "config.h" |
6a44839f | 51 | #include "utils.h" |
9c4693b8 CS |
52 | #include "commands.h" |
53 | #include "cgroup.h" | |
025ed0f3 | 54 | #include "lxclock.h" |
2c4ea790 SH |
55 | #include "conf.h" |
56 | #include "lxcseccomp.h" | |
57 | #include <lxc/lxccontainer.h> | |
fe4de9a6 | 58 | #include "lsm/lsm.h" |
9b8e3c96 | 59 | #include "confile.h" |
9c4693b8 CS |
60 | |
61 | #if HAVE_SYS_PERSONALITY_H | |
62 | #include <sys/personality.h> | |
63 | #endif | |
e0732705 | 64 | |
a3da2f3b SG |
65 | #ifndef SOCK_CLOEXEC |
66 | # define SOCK_CLOEXEC 02000000 | |
67 | #endif | |
68 | ||
d6a3c917 SG |
69 | #ifndef MS_REC |
70 | #define MS_REC 16384 | |
71 | #endif | |
72 | ||
73 | #ifndef MS_SLAVE | |
74 | #define MS_SLAVE (1<<19) | |
75 | #endif | |
76 | ||
e0732705 CS |
77 | lxc_log_define(lxc_attach, lxc); |
78 | ||
5c3fcae7 SG |
79 | int lsm_set_label_at(int procfd, int on_exec, char* lsm_label) { |
80 | int labelfd = -1; | |
81 | int ret = 0; | |
82 | const char* name; | |
83 | char* command = NULL; | |
84 | ||
85 | name = lsm_name(); | |
86 | ||
87 | if (strcmp(name, "nop") == 0) | |
88 | goto out; | |
89 | ||
90 | if (strcmp(name, "none") == 0) | |
91 | goto out; | |
92 | ||
93 | /* We don't support on-exec with AppArmor */ | |
94 | if (strcmp(name, "AppArmor") == 0) | |
95 | on_exec = 0; | |
96 | ||
97 | if (on_exec) { | |
98 | labelfd = openat(procfd, "self/attr/exec", O_RDWR); | |
99 | } | |
100 | else { | |
101 | labelfd = openat(procfd, "self/attr/current", O_RDWR); | |
102 | } | |
103 | ||
104 | if (labelfd < 0) { | |
105 | SYSERROR("Unable to open LSM label"); | |
106 | ret = -1; | |
107 | goto out; | |
108 | } | |
109 | ||
110 | if (strcmp(name, "AppArmor") == 0) { | |
111 | int size; | |
112 | ||
113 | command = malloc(strlen(lsm_label) + strlen("changeprofile ") + 1); | |
114 | if (!command) { | |
115 | SYSERROR("Failed to write apparmor profile"); | |
116 | ret = -1; | |
117 | goto out; | |
118 | } | |
119 | ||
120 | size = sprintf(command, "changeprofile %s", lsm_label); | |
121 | if (size < 0) { | |
122 | SYSERROR("Failed to write apparmor profile"); | |
123 | ret = -1; | |
124 | goto out; | |
125 | } | |
126 | ||
127 | if (write(labelfd, command, size + 1) < 0) { | |
128 | SYSERROR("Unable to set LSM label"); | |
129 | ret = -1; | |
130 | goto out; | |
131 | } | |
132 | } | |
133 | else if (strcmp(name, "SELinux") == 0) { | |
134 | if (write(labelfd, lsm_label, strlen(lsm_label) + 1) < 0) { | |
135 | SYSERROR("Unable to set LSM label"); | |
136 | ret = -1; | |
137 | goto out; | |
138 | } | |
139 | } | |
140 | else { | |
141 | ERROR("Unable to restore label for unknown LSM: %s", name); | |
142 | ret = -1; | |
143 | goto out; | |
144 | } | |
145 | ||
146 | out: | |
147 | free(command); | |
148 | ||
149 | if (labelfd != -1) | |
150 | close(labelfd); | |
151 | ||
152 | return ret; | |
153 | } | |
154 | ||
74a3920a | 155 | static struct lxc_proc_context_info *lxc_proc_get_context_info(pid_t pid) |
e0732705 CS |
156 | { |
157 | struct lxc_proc_context_info *info = calloc(1, sizeof(*info)); | |
158 | FILE *proc_file; | |
159 | char proc_fn[MAXPATHLEN]; | |
460a1cf0 | 160 | char *line = NULL; |
e0732705 | 161 | size_t line_bufsz = 0; |
460a1cf0 | 162 | int ret, found; |
e0732705 CS |
163 | |
164 | if (!info) { | |
165 | SYSERROR("Could not allocate memory."); | |
166 | return NULL; | |
167 | } | |
168 | ||
169 | /* read capabilities */ | |
170 | snprintf(proc_fn, MAXPATHLEN, "/proc/%d/status", pid); | |
171 | ||
172 | proc_file = fopen(proc_fn, "r"); | |
173 | if (!proc_file) { | |
174 | SYSERROR("Could not open %s", proc_fn); | |
175 | goto out_error; | |
176 | } | |
177 | ||
178 | found = 0; | |
179 | while (getline(&line, &line_bufsz, proc_file) != -1) { | |
180 | ret = sscanf(line, "CapBnd: %llx", &info->capability_mask); | |
181 | if (ret != EOF && ret > 0) { | |
182 | found = 1; | |
183 | break; | |
184 | } | |
185 | } | |
186 | ||
f10fad2f | 187 | free(line); |
e0732705 CS |
188 | fclose(proc_file); |
189 | ||
190 | if (!found) { | |
191 | SYSERROR("Could not read capability bounding set from %s", proc_fn); | |
192 | errno = ENOENT; | |
193 | goto out_error; | |
194 | } | |
195 | ||
fe4de9a6 | 196 | info->lsm_label = lsm_process_label_get(pid); |
e0732705 | 197 | |
e0732705 CS |
198 | return info; |
199 | ||
200 | out_error: | |
460a1cf0 | 201 | free(info); |
e0732705 CS |
202 | return NULL; |
203 | } | |
204 | ||
fe4de9a6 DE |
205 | static void lxc_proc_put_context_info(struct lxc_proc_context_info *ctx) |
206 | { | |
f10fad2f | 207 | free(ctx->lsm_label); |
2c4ea790 SH |
208 | if (ctx->container) |
209 | lxc_container_put(ctx->container); | |
fe4de9a6 DE |
210 | free(ctx); |
211 | } | |
212 | ||
74a3920a | 213 | static int lxc_attach_to_ns(pid_t pid, int which) |
99d50954 CS |
214 | { |
215 | char path[MAXPATHLEN]; | |
fc763ab7 CS |
216 | /* according to <http://article.gmane.org/gmane.linux.kernel.containers.lxc.devel/1429>, |
217 | * the file for user namepsaces in /proc/$pid/ns will be called | |
218 | * 'user' once the kernel supports it | |
219 | */ | |
dac862c0 | 220 | static char *ns[] = { "user", "mnt", "pid", "uts", "ipc", "net", "cgroup" }; |
fc763ab7 | 221 | static int flags[] = { |
f4364484 | 222 | CLONE_NEWUSER, CLONE_NEWNS, CLONE_NEWPID, CLONE_NEWUTS, CLONE_NEWIPC, |
dac862c0 | 223 | CLONE_NEWNET, CLONE_NEWCGROUP |
fc763ab7 CS |
224 | }; |
225 | static const int size = sizeof(ns) / sizeof(char *); | |
99d50954 | 226 | int fd[size]; |
fc763ab7 CS |
227 | int i, j, saved_errno; |
228 | ||
99d50954 CS |
229 | |
230 | snprintf(path, MAXPATHLEN, "/proc/%d/ns", pid); | |
231 | if (access(path, X_OK)) { | |
232 | ERROR("Does this kernel version support 'attach' ?"); | |
233 | return -1; | |
234 | } | |
235 | ||
236 | for (i = 0; i < size; i++) { | |
fc763ab7 CS |
237 | /* ignore if we are not supposed to attach to that |
238 | * namespace | |
239 | */ | |
240 | if (which != -1 && !(which & flags[i])) { | |
241 | fd[i] = -1; | |
242 | continue; | |
243 | } | |
244 | ||
99d50954 | 245 | snprintf(path, MAXPATHLEN, "/proc/%d/ns/%s", pid, ns[i]); |
9c4693b8 | 246 | fd[i] = open(path, O_RDONLY | O_CLOEXEC); |
99d50954 | 247 | if (fd[i] < 0) { |
fc763ab7 CS |
248 | saved_errno = errno; |
249 | ||
250 | /* close all already opened file descriptors before | |
251 | * we return an error, so we don't leak them | |
252 | */ | |
253 | for (j = 0; j < i; j++) | |
254 | close(fd[j]); | |
255 | ||
256 | errno = saved_errno; | |
99d50954 CS |
257 | SYSERROR("failed to open '%s'", path); |
258 | return -1; | |
259 | } | |
260 | } | |
261 | ||
262 | for (i = 0; i < size; i++) { | |
fc763ab7 CS |
263 | if (fd[i] >= 0 && setns(fd[i], 0) != 0) { |
264 | saved_errno = errno; | |
265 | ||
266 | for (j = i; j < size; j++) | |
267 | close(fd[j]); | |
268 | ||
269 | errno = saved_errno; | |
99d50954 CS |
270 | SYSERROR("failed to set namespace '%s'", ns[i]); |
271 | return -1; | |
272 | } | |
273 | ||
274 | close(fd[i]); | |
275 | } | |
276 | ||
277 | return 0; | |
278 | } | |
279 | ||
74a3920a | 280 | static int lxc_attach_remount_sys_proc(void) |
7a0b0b56 CS |
281 | { |
282 | int ret; | |
283 | ||
284 | ret = unshare(CLONE_NEWNS); | |
285 | if (ret < 0) { | |
286 | SYSERROR("failed to unshare mount namespace"); | |
287 | return -1; | |
288 | } | |
289 | ||
2c6f3fc9 SH |
290 | if (detect_shared_rootfs()) { |
291 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL)) { | |
292 | SYSERROR("Failed to make / rslave"); | |
293 | ERROR("Continuing..."); | |
294 | } | |
295 | } | |
296 | ||
7a0b0b56 CS |
297 | /* assume /proc is always mounted, so remount it */ |
298 | ret = umount2("/proc", MNT_DETACH); | |
299 | if (ret < 0) { | |
300 | SYSERROR("failed to unmount /proc"); | |
301 | return -1; | |
302 | } | |
303 | ||
304 | ret = mount("none", "/proc", "proc", 0, NULL); | |
305 | if (ret < 0) { | |
306 | SYSERROR("failed to remount /proc"); | |
307 | return -1; | |
308 | } | |
309 | ||
310 | /* try to umount /sys - if it's not a mount point, | |
311 | * we'll get EINVAL, then we ignore it because it | |
312 | * may not have been mounted in the first place | |
313 | */ | |
314 | ret = umount2("/sys", MNT_DETACH); | |
315 | if (ret < 0 && errno != EINVAL) { | |
316 | SYSERROR("failed to unmount /sys"); | |
317 | return -1; | |
318 | } else if (ret == 0) { | |
319 | /* remount it */ | |
320 | ret = mount("none", "/sys", "sysfs", 0, NULL); | |
321 | if (ret < 0) { | |
322 | SYSERROR("failed to remount /sys"); | |
323 | return -1; | |
324 | } | |
325 | } | |
326 | ||
327 | return 0; | |
328 | } | |
329 | ||
74a3920a | 330 | static int lxc_attach_drop_privs(struct lxc_proc_context_info *ctx) |
e0732705 CS |
331 | { |
332 | int last_cap = lxc_caps_last_cap(); | |
333 | int cap; | |
334 | ||
335 | for (cap = 0; cap <= last_cap; cap++) { | |
336 | if (ctx->capability_mask & (1LL << cap)) | |
337 | continue; | |
338 | ||
339 | if (prctl(PR_CAPBSET_DROP, cap, 0, 0, 0)) { | |
340 | SYSERROR("failed to remove capability id %d", cap); | |
341 | return -1; | |
342 | } | |
343 | } | |
344 | ||
345 | return 0; | |
346 | } | |
905022f7 | 347 | |
74a3920a | 348 | static int lxc_attach_set_environment(enum lxc_attach_env_policy_t policy, char** extra_env, char** extra_keep) |
b3a39ba6 | 349 | { |
799f96fd | 350 | if (policy == LXC_ATTACH_CLEAR_ENV) { |
3d5e9f48 | 351 | char **extra_keep_store = NULL; |
3d5e9f48 CS |
352 | int path_kept = 0; |
353 | ||
354 | if (extra_keep) { | |
355 | size_t count, i; | |
356 | ||
357 | for (count = 0; extra_keep[count]; count++); | |
358 | ||
359 | extra_keep_store = calloc(count, sizeof(char *)); | |
360 | if (!extra_keep_store) { | |
361 | SYSERROR("failed to allocate memory for storing current " | |
362 | "environment variable values that will be kept"); | |
363 | return -1; | |
364 | } | |
365 | for (i = 0; i < count; i++) { | |
366 | char *v = getenv(extra_keep[i]); | |
367 | if (v) { | |
368 | extra_keep_store[i] = strdup(v); | |
369 | if (!extra_keep_store[i]) { | |
370 | SYSERROR("failed to allocate memory for storing current " | |
371 | "environment variable values that will be kept"); | |
372 | while (i > 0) | |
373 | free(extra_keep_store[--i]); | |
374 | free(extra_keep_store); | |
375 | return -1; | |
376 | } | |
377 | if (strcmp(extra_keep[i], "PATH") == 0) | |
378 | path_kept = 1; | |
379 | } | |
380 | /* calloc sets entire array to zero, so we don't | |
381 | * need an else */ | |
382 | } | |
383 | } | |
384 | ||
799f96fd | 385 | if (clearenv()) { |
a9cab7e3 | 386 | char **p; |
799f96fd | 387 | SYSERROR("failed to clear environment"); |
a9cab7e3 CS |
388 | if (extra_keep_store) { |
389 | for (p = extra_keep_store; *p; p++) | |
390 | free(*p); | |
391 | free(extra_keep_store); | |
392 | } | |
3d5e9f48 CS |
393 | return -1; |
394 | } | |
395 | ||
396 | if (extra_keep_store) { | |
397 | size_t i; | |
398 | for (i = 0; extra_keep[i]; i++) { | |
acd4922e SG |
399 | if (extra_keep_store[i]) { |
400 | if (setenv(extra_keep[i], extra_keep_store[i], 1) < 0) | |
401 | SYSERROR("Unable to set environment variable"); | |
402 | } | |
3d5e9f48 CS |
403 | free(extra_keep_store[i]); |
404 | } | |
405 | free(extra_keep_store); | |
406 | } | |
407 | ||
408 | /* always set a default path; shells and execlp tend | |
409 | * to be fine without it, but there is a disturbing | |
410 | * number of C programs out there that just assume | |
411 | * that getenv("PATH") is never NULL and then die a | |
412 | * painful segfault death. */ | |
cfa70b88 | 413 | if (!path_kept) |
511a6936 | 414 | setenv("PATH", "/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", 1); |
b3a39ba6 DW |
415 | } |
416 | ||
417 | if (putenv("container=lxc")) { | |
418 | SYSERROR("failed to set environment variable"); | |
419 | return -1; | |
420 | } | |
421 | ||
3d5e9f48 CS |
422 | /* set extra environment variables */ |
423 | if (extra_env) { | |
424 | for (; *extra_env; extra_env++) { | |
425 | /* duplicate the string, just to be on | |
426 | * the safe side, because putenv does not | |
427 | * do it for us */ | |
428 | char *p = strdup(*extra_env); | |
429 | /* we just assume the user knows what they | |
430 | * are doing, so we don't do any checks */ | |
431 | if (!p) { | |
432 | SYSERROR("failed to allocate memory for additional environment " | |
433 | "variables"); | |
434 | return -1; | |
435 | } | |
436 | putenv(p); | |
437 | } | |
438 | } | |
439 | ||
b3a39ba6 DW |
440 | return 0; |
441 | } | |
442 | ||
74a3920a | 443 | static char *lxc_attach_getpwshell(uid_t uid) |
905022f7 CS |
444 | { |
445 | /* local variables */ | |
446 | pid_t pid; | |
447 | int pipes[2]; | |
448 | int ret; | |
449 | int fd; | |
450 | char *result = NULL; | |
451 | ||
452 | /* we need to fork off a process that runs the | |
453 | * getent program, and we need to capture its | |
454 | * output, so we use a pipe for that purpose | |
455 | */ | |
456 | ret = pipe(pipes); | |
457 | if (ret < 0) | |
458 | return NULL; | |
459 | ||
460 | pid = fork(); | |
461 | if (pid < 0) { | |
462 | close(pipes[0]); | |
463 | close(pipes[1]); | |
464 | return NULL; | |
465 | } | |
466 | ||
467 | if (pid) { | |
468 | /* parent process */ | |
469 | FILE *pipe_f; | |
470 | char *line = NULL; | |
471 | size_t line_bufsz = 0; | |
472 | int found = 0; | |
473 | int status; | |
474 | ||
475 | close(pipes[1]); | |
476 | ||
477 | pipe_f = fdopen(pipes[0], "r"); | |
478 | while (getline(&line, &line_bufsz, pipe_f) != -1) { | |
479 | char *token; | |
480 | char *saveptr = NULL; | |
481 | long value; | |
482 | char *endptr = NULL; | |
483 | int i; | |
484 | ||
485 | /* if we already found something, just continue | |
486 | * to read until the pipe doesn't deliver any more | |
487 | * data, but don't modify the existing data | |
488 | * structure | |
489 | */ | |
490 | if (found) | |
491 | continue; | |
492 | ||
493 | /* trim line on the right hand side */ | |
bbb8a488 | 494 | for (i = strlen(line); i > 0 && (line[i - 1] == '\n' || line[i - 1] == '\r'); --i) |
905022f7 CS |
495 | line[i - 1] = '\0'; |
496 | ||
497 | /* split into tokens: first user name */ | |
498 | token = strtok_r(line, ":", &saveptr); | |
499 | if (!token) | |
500 | continue; | |
501 | /* next: dummy password field */ | |
502 | token = strtok_r(NULL, ":", &saveptr); | |
503 | if (!token) | |
504 | continue; | |
505 | /* next: user id */ | |
506 | token = strtok_r(NULL, ":", &saveptr); | |
507 | value = token ? strtol(token, &endptr, 10) : 0; | |
508 | if (!token || !endptr || *endptr || value == LONG_MIN || value == LONG_MAX) | |
509 | continue; | |
510 | /* dummy sanity check: user id matches */ | |
511 | if ((uid_t) value != uid) | |
512 | continue; | |
513 | /* skip fields: gid, gecos, dir, go to next field 'shell' */ | |
514 | for (i = 0; i < 4; i++) { | |
515 | token = strtok_r(NULL, ":", &saveptr); | |
516 | if (!token) | |
517 | break; | |
518 | } | |
519 | if (!token) | |
520 | continue; | |
f10fad2f | 521 | free(result); |
905022f7 CS |
522 | result = strdup(token); |
523 | ||
524 | /* sanity check that there are no fields after that */ | |
525 | token = strtok_r(NULL, ":", &saveptr); | |
526 | if (token) | |
527 | continue; | |
528 | ||
529 | found = 1; | |
530 | } | |
531 | ||
532 | free(line); | |
533 | fclose(pipe_f); | |
534 | again: | |
535 | if (waitpid(pid, &status, 0) < 0) { | |
536 | if (errno == EINTR) | |
537 | goto again; | |
538 | return NULL; | |
539 | } | |
540 | ||
541 | /* some sanity checks: if anything even hinted at going | |
542 | * wrong: we can't be sure we have a valid result, so | |
543 | * we assume we don't | |
544 | */ | |
545 | ||
546 | if (!WIFEXITED(status)) | |
547 | return NULL; | |
548 | ||
549 | if (WEXITSTATUS(status) != 0) | |
550 | return NULL; | |
551 | ||
552 | if (!found) | |
553 | return NULL; | |
554 | ||
555 | return result; | |
556 | } else { | |
557 | /* child process */ | |
558 | char uid_buf[32]; | |
559 | char *arguments[] = { | |
560 | "getent", | |
561 | "passwd", | |
562 | uid_buf, | |
563 | NULL | |
564 | }; | |
565 | ||
566 | close(pipes[0]); | |
567 | ||
568 | /* we want to capture stdout */ | |
569 | dup2(pipes[1], 1); | |
570 | close(pipes[1]); | |
571 | ||
572 | /* get rid of stdin/stderr, so we try to associate it | |
573 | * with /dev/null | |
574 | */ | |
575 | fd = open("/dev/null", O_RDWR); | |
576 | if (fd < 0) { | |
577 | close(0); | |
578 | close(2); | |
579 | } else { | |
580 | dup2(fd, 0); | |
581 | dup2(fd, 2); | |
582 | close(fd); | |
583 | } | |
584 | ||
585 | /* finish argument list */ | |
586 | ret = snprintf(uid_buf, sizeof(uid_buf), "%ld", (long) uid); | |
587 | if (ret <= 0) | |
588 | exit(-1); | |
589 | ||
590 | /* try to run getent program */ | |
591 | (void) execvp("getent", arguments); | |
592 | exit(-1); | |
593 | } | |
594 | } | |
cb3e61fa | 595 | |
74a3920a | 596 | static void lxc_attach_get_init_uidgid(uid_t* init_uid, gid_t* init_gid) |
cb3e61fa CS |
597 | { |
598 | FILE *proc_file; | |
599 | char proc_fn[MAXPATHLEN]; | |
600 | char *line = NULL; | |
601 | size_t line_bufsz = 0; | |
602 | int ret; | |
603 | long value = -1; | |
604 | uid_t uid = (uid_t)-1; | |
605 | gid_t gid = (gid_t)-1; | |
606 | ||
607 | /* read capabilities */ | |
608 | snprintf(proc_fn, MAXPATHLEN, "/proc/%d/status", 1); | |
609 | ||
610 | proc_file = fopen(proc_fn, "r"); | |
611 | if (!proc_file) | |
612 | return; | |
613 | ||
614 | while (getline(&line, &line_bufsz, proc_file) != -1) { | |
615 | /* format is: real, effective, saved set user, fs | |
616 | * we only care about real uid | |
617 | */ | |
618 | ret = sscanf(line, "Uid: %ld", &value); | |
619 | if (ret != EOF && ret > 0) { | |
620 | uid = (uid_t) value; | |
621 | } else { | |
622 | ret = sscanf(line, "Gid: %ld", &value); | |
623 | if (ret != EOF && ret > 0) | |
624 | gid = (gid_t) value; | |
625 | } | |
626 | if (uid != (uid_t)-1 && gid != (gid_t)-1) | |
627 | break; | |
628 | } | |
629 | ||
630 | fclose(proc_file); | |
631 | free(line); | |
632 | ||
633 | /* only override arguments if we found something */ | |
634 | if (uid != (uid_t)-1) | |
635 | *init_uid = uid; | |
636 | if (gid != (gid_t)-1) | |
637 | *init_gid = gid; | |
638 | ||
639 | /* TODO: we should also parse supplementary groups and use | |
640 | * setgroups() to set them */ | |
641 | } | |
9c4693b8 CS |
642 | |
643 | struct attach_clone_payload { | |
644 | int ipc_socket; | |
645 | lxc_attach_options_t* options; | |
646 | struct lxc_proc_context_info* init_ctx; | |
647 | lxc_attach_exec_t exec_function; | |
648 | void* exec_payload; | |
5c3fcae7 | 649 | int procfd; |
9c4693b8 CS |
650 | }; |
651 | ||
652 | static int attach_child_main(void* data); | |
653 | ||
654 | /* help the optimizer along if it doesn't know that exit always exits */ | |
5dcc1ca6 | 655 | #define rexit(c) do { int __c = (c); _exit(__c); return __c; } while(0) |
9c4693b8 CS |
656 | |
657 | /* define default options if no options are supplied by the user */ | |
658 | static lxc_attach_options_t attach_static_default_options = LXC_ATTACH_OPTIONS_DEFAULT; | |
659 | ||
ff07d7bb CB |
660 | static bool fetch_seccomp(struct lxc_proc_context_info *i, |
661 | lxc_attach_options_t *options) | |
2c4ea790 SH |
662 | { |
663 | struct lxc_container *c; | |
bd7b4e28 | 664 | char *path; |
2eef2bda | 665 | |
2c4ea790 SH |
666 | if (!(options->namespaces & CLONE_NEWNS) || !(options->attach_flags & LXC_ATTACH_LSM)) |
667 | return true; | |
668 | ||
ff07d7bb | 669 | c = i->container; |
bd7b4e28 SG |
670 | |
671 | /* Initialize an empty lxc_conf */ | |
672 | if (!c->set_config_item(c, "lxc.seccomp", "")) { | |
2c4ea790 | 673 | return false; |
bd7b4e28 SG |
674 | } |
675 | ||
676 | /* Fetch the current profile path over the cmd interface */ | |
677 | path = c->get_running_config_item(c, "lxc.seccomp"); | |
678 | if (!path) { | |
679 | return true; | |
680 | } | |
681 | ||
682 | /* Copy the value into the new lxc_conf */ | |
683 | if (!c->set_config_item(c, "lxc.seccomp", path)) { | |
684 | free(path); | |
685 | return false; | |
686 | } | |
687 | free(path); | |
688 | ||
689 | /* Attempt to parse the resulting config */ | |
2c4ea790 | 690 | if (lxc_read_seccomp_config(c->lxc_conf) < 0) { |
442f5c0f | 691 | ERROR("Error reading seccomp policy"); |
2c4ea790 SH |
692 | return false; |
693 | } | |
694 | ||
695 | return true; | |
696 | } | |
697 | ||
9b8e3c96 SH |
698 | static signed long get_personality(const char *name, const char *lxcpath) |
699 | { | |
0d7cf7e9 | 700 | char *p = lxc_cmd_get_config_item(name, "lxc.arch", lxcpath); |
9b8e3c96 SH |
701 | signed long ret; |
702 | ||
703 | if (!p) | |
704 | return -1; | |
705 | ret = lxc_config_parse_arch(p); | |
706 | free(p); | |
707 | return ret; | |
708 | } | |
709 | ||
9c4693b8 CS |
710 | int lxc_attach(const char* name, const char* lxcpath, lxc_attach_exec_t exec_function, void* exec_payload, lxc_attach_options_t* options, pid_t* attached_process) |
711 | { | |
712 | int ret, status; | |
f4364484 | 713 | pid_t init_pid, pid, attached_pid, expected; |
9c4693b8 CS |
714 | struct lxc_proc_context_info *init_ctx; |
715 | char* cwd; | |
716 | char* new_cwd; | |
717 | int ipc_sockets[2]; | |
5c3fcae7 | 718 | int procfd; |
9b8e3c96 | 719 | signed long personality; |
9c4693b8 CS |
720 | |
721 | if (!options) | |
722 | options = &attach_static_default_options; | |
723 | ||
724 | init_pid = lxc_cmd_get_init_pid(name, lxcpath); | |
725 | if (init_pid < 0) { | |
726 | ERROR("failed to get the init pid"); | |
727 | return -1; | |
728 | } | |
729 | ||
730 | init_ctx = lxc_proc_get_context_info(init_pid); | |
731 | if (!init_ctx) { | |
732 | ERROR("failed to get context of the init process, pid = %ld", (long)init_pid); | |
733 | return -1; | |
734 | } | |
735 | ||
9b8e3c96 SH |
736 | personality = get_personality(name, lxcpath); |
737 | if (init_ctx->personality < 0) { | |
738 | ERROR("Failed to get personality of the container"); | |
739 | lxc_proc_put_context_info(init_ctx); | |
740 | return -1; | |
741 | } | |
742 | init_ctx->personality = personality; | |
743 | ||
ff07d7bb CB |
744 | init_ctx->container = lxc_container_new(name, lxcpath); |
745 | if (!init_ctx->container) | |
746 | return -1; | |
747 | ||
748 | if (!fetch_seccomp(init_ctx, options)) | |
2c4ea790 SH |
749 | WARN("Failed to get seccomp policy"); |
750 | ||
9c4693b8 CS |
751 | cwd = getcwd(NULL, 0); |
752 | ||
753 | /* determine which namespaces the container was created with | |
754 | * by asking lxc-start, if necessary | |
755 | */ | |
756 | if (options->namespaces == -1) { | |
757 | options->namespaces = lxc_cmd_get_clone_flags(name, lxcpath); | |
758 | /* call failed */ | |
759 | if (options->namespaces == -1) { | |
760 | ERROR("failed to automatically determine the " | |
761 | "namespaces which the container unshared"); | |
762 | free(cwd); | |
fe4de9a6 | 763 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
764 | return -1; |
765 | } | |
766 | } | |
767 | ||
768 | /* create a socket pair for IPC communication; set SOCK_CLOEXEC in order | |
769 | * to make sure we don't irritate other threads that want to fork+exec away | |
770 | * | |
771 | * IMPORTANT: if the initial process is multithreaded and another call | |
772 | * just fork()s away without exec'ing directly after, the socket fd will | |
773 | * exist in the forked process from the other thread and any close() in | |
774 | * our own child process will not really cause the socket to close properly, | |
775 | * potentiall causing the parent to hang. | |
776 | * | |
777 | * For this reason, while IPC is still active, we have to use shutdown() | |
778 | * if the child exits prematurely in order to signal that the socket | |
779 | * is closed and cannot assume that the child exiting will automatically | |
780 | * do that. | |
781 | * | |
782 | * IPC mechanism: (X is receiver) | |
783 | * initial process intermediate attached | |
784 | * X <--- send pid of | |
785 | * attached proc, | |
786 | * then exit | |
787 | * send 0 ------------------------------------> X | |
788 | * [do initialization] | |
789 | * X <------------------------------------ send 1 | |
790 | * [add to cgroup, ...] | |
791 | * send 2 ------------------------------------> X | |
792 | * close socket close socket | |
793 | * run program | |
794 | */ | |
795 | ret = socketpair(PF_LOCAL, SOCK_STREAM | SOCK_CLOEXEC, 0, ipc_sockets); | |
796 | if (ret < 0) { | |
797 | SYSERROR("could not set up required IPC mechanism for attaching"); | |
798 | free(cwd); | |
fe4de9a6 | 799 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
800 | return -1; |
801 | } | |
802 | ||
803 | /* create intermediate subprocess, three reasons: | |
804 | * 1. runs all pthread_atfork handlers and the | |
805 | * child will no longer be threaded | |
806 | * (we can't properly setns() in a threaded process) | |
807 | * 2. we can't setns() in the child itself, since | |
808 | * we want to make sure we are properly attached to | |
809 | * the pidns | |
810 | * 3. also, the initial thread has to put the attached | |
811 | * process into the cgroup, which we can only do if | |
812 | * we didn't already setns() (otherwise, user | |
813 | * namespaces will hate us) | |
814 | */ | |
815 | pid = fork(); | |
816 | ||
817 | if (pid < 0) { | |
818 | SYSERROR("failed to create first subprocess"); | |
819 | free(cwd); | |
fe4de9a6 | 820 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
821 | return -1; |
822 | } | |
823 | ||
824 | if (pid) { | |
825 | pid_t to_cleanup_pid = pid; | |
9c4693b8 | 826 | |
ec64264d | 827 | /* initial thread, we close the socket that is for the |
9c4693b8 CS |
828 | * subprocesses |
829 | */ | |
830 | close(ipc_sockets[1]); | |
831 | free(cwd); | |
832 | ||
f4364484 SG |
833 | /* attach to cgroup, if requested */ |
834 | if (options->attach_flags & LXC_ATTACH_MOVE_TO_CGROUP) { | |
4fb3cba5 | 835 | if (!cgroup_attach(name, lxcpath, pid)) |
f4364484 | 836 | goto cleanup_error; |
f4364484 SG |
837 | } |
838 | ||
839 | /* Let the child process know to go ahead */ | |
840 | status = 0; | |
841 | ret = lxc_write_nointr(ipc_sockets[0], &status, sizeof(status)); | |
842 | if (ret <= 0) { | |
843 | ERROR("error using IPC to notify attached process for initialization (0)"); | |
844 | goto cleanup_error; | |
845 | } | |
846 | ||
9c4693b8 CS |
847 | /* get pid from intermediate process */ |
848 | ret = lxc_read_nointr_expect(ipc_sockets[0], &attached_pid, sizeof(attached_pid), NULL); | |
849 | if (ret <= 0) { | |
850 | if (ret != 0) | |
851 | ERROR("error using IPC to receive pid of attached process"); | |
852 | goto cleanup_error; | |
853 | } | |
854 | ||
2eef2bda | 855 | /* ignore SIGKILL (CTRL-C) and SIGQUIT (CTRL-\) - issue #313 */ |
62183f1a SH |
856 | if (options->stdin_fd == 0) { |
857 | signal(SIGINT, SIG_IGN); | |
858 | signal(SIGQUIT, SIG_IGN); | |
859 | } | |
2eef2bda | 860 | |
9c4693b8 CS |
861 | /* reap intermediate process */ |
862 | ret = wait_for_pid(pid); | |
863 | if (ret < 0) | |
864 | goto cleanup_error; | |
865 | ||
866 | /* we will always have to reap the grandchild now */ | |
867 | to_cleanup_pid = attached_pid; | |
868 | ||
869 | /* tell attached process it may start initializing */ | |
870 | status = 0; | |
871 | ret = lxc_write_nointr(ipc_sockets[0], &status, sizeof(status)); | |
872 | if (ret <= 0) { | |
873 | ERROR("error using IPC to notify attached process for initialization (0)"); | |
874 | goto cleanup_error; | |
875 | } | |
876 | ||
877 | /* wait for the attached process to finish initializing */ | |
878 | expected = 1; | |
879 | ret = lxc_read_nointr_expect(ipc_sockets[0], &status, sizeof(status), &expected); | |
880 | if (ret <= 0) { | |
881 | if (ret != 0) | |
882 | ERROR("error using IPC to receive notification from attached process (1)"); | |
883 | goto cleanup_error; | |
884 | } | |
885 | ||
9c4693b8 CS |
886 | /* tell attached process we're done */ |
887 | status = 2; | |
888 | ret = lxc_write_nointr(ipc_sockets[0], &status, sizeof(status)); | |
889 | if (ret <= 0) { | |
890 | ERROR("error using IPC to notify attached process for initialization (2)"); | |
891 | goto cleanup_error; | |
892 | } | |
893 | ||
894 | /* now shut down communication with child, we're done */ | |
895 | shutdown(ipc_sockets[0], SHUT_RDWR); | |
896 | close(ipc_sockets[0]); | |
fe4de9a6 | 897 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
898 | |
899 | /* we're done, the child process should now execute whatever | |
900 | * it is that the user requested. The parent can now track it | |
901 | * with waitpid() or similar. | |
902 | */ | |
903 | ||
904 | *attached_process = attached_pid; | |
905 | return 0; | |
906 | ||
907 | cleanup_error: | |
908 | /* first shut down the socket, then wait for the pid, | |
909 | * otherwise the pid we're waiting for may never exit | |
910 | */ | |
911 | shutdown(ipc_sockets[0], SHUT_RDWR); | |
912 | close(ipc_sockets[0]); | |
913 | if (to_cleanup_pid) | |
914 | (void) wait_for_pid(to_cleanup_pid); | |
fe4de9a6 | 915 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
916 | return -1; |
917 | } | |
918 | ||
919 | /* first subprocess begins here, we close the socket that is for the | |
920 | * initial thread | |
921 | */ | |
922 | close(ipc_sockets[0]); | |
923 | ||
f4364484 SG |
924 | /* Wait for the parent to have setup cgroups */ |
925 | expected = 0; | |
926 | status = -1; | |
927 | ret = lxc_read_nointr_expect(ipc_sockets[1], &status, sizeof(status), &expected); | |
928 | if (ret <= 0) { | |
929 | ERROR("error communicating with child process"); | |
930 | shutdown(ipc_sockets[1], SHUT_RDWR); | |
931 | rexit(-1); | |
932 | } | |
933 | ||
dac862c0 | 934 | if ((options->attach_flags & LXC_ATTACH_MOVE_TO_CGROUP) && cgns_supported()) |
935 | options->namespaces |= CLONE_NEWCGROUP; | |
fe3c80af | 936 | |
5c3fcae7 SG |
937 | procfd = open("/proc", O_DIRECTORY | O_RDONLY); |
938 | if (procfd < 0) { | |
939 | SYSERROR("Unable to open /proc"); | |
940 | shutdown(ipc_sockets[1], SHUT_RDWR); | |
941 | rexit(-1); | |
942 | } | |
943 | ||
9c4693b8 CS |
944 | /* attach now, create another subprocess later, since pid namespaces |
945 | * only really affect the children of the current process | |
946 | */ | |
947 | ret = lxc_attach_to_ns(init_pid, options->namespaces); | |
948 | if (ret < 0) { | |
949 | ERROR("failed to enter the namespace"); | |
950 | shutdown(ipc_sockets[1], SHUT_RDWR); | |
951 | rexit(-1); | |
952 | } | |
953 | ||
954 | /* attach succeeded, try to cwd */ | |
955 | if (options->initial_cwd) | |
956 | new_cwd = options->initial_cwd; | |
957 | else | |
958 | new_cwd = cwd; | |
959 | ret = chdir(new_cwd); | |
960 | if (ret < 0) | |
961 | WARN("could not change directory to '%s'", new_cwd); | |
962 | free(cwd); | |
963 | ||
964 | /* now create the real child process */ | |
965 | { | |
966 | struct attach_clone_payload payload = { | |
967 | .ipc_socket = ipc_sockets[1], | |
968 | .options = options, | |
969 | .init_ctx = init_ctx, | |
970 | .exec_function = exec_function, | |
5c3fcae7 SG |
971 | .exec_payload = exec_payload, |
972 | .procfd = procfd | |
9c4693b8 CS |
973 | }; |
974 | /* We use clone_parent here to make this subprocess a direct child of | |
975 | * the initial process. Then this intermediate process can exit and | |
976 | * the parent can directly track the attached process. | |
977 | */ | |
978 | pid = lxc_clone(attach_child_main, &payload, CLONE_PARENT); | |
979 | } | |
980 | ||
981 | /* shouldn't happen, clone() should always return positive pid */ | |
982 | if (pid <= 0) { | |
983 | SYSERROR("failed to create subprocess"); | |
984 | shutdown(ipc_sockets[1], SHUT_RDWR); | |
985 | rexit(-1); | |
986 | } | |
987 | ||
988 | /* tell grandparent the pid of the pid of the newly created child */ | |
989 | ret = lxc_write_nointr(ipc_sockets[1], &pid, sizeof(pid)); | |
990 | if (ret != sizeof(pid)) { | |
991 | /* if this really happens here, this is very unfortunate, since the | |
992 | * parent will not know the pid of the attached process and will | |
993 | * not be able to wait for it (and we won't either due to CLONE_PARENT) | |
994 | * so the parent won't be able to reap it and the attached process | |
995 | * will remain a zombie | |
996 | */ | |
997 | ERROR("error using IPC to notify main process of pid of the attached process"); | |
998 | shutdown(ipc_sockets[1], SHUT_RDWR); | |
999 | rexit(-1); | |
1000 | } | |
1001 | ||
1002 | /* the rest is in the hands of the initial and the attached process */ | |
1003 | rexit(0); | |
1004 | } | |
1005 | ||
74a3920a | 1006 | static int attach_child_main(void* data) |
9c4693b8 CS |
1007 | { |
1008 | struct attach_clone_payload* payload = (struct attach_clone_payload*)data; | |
1009 | int ipc_socket = payload->ipc_socket; | |
5c3fcae7 | 1010 | int procfd = payload->procfd; |
9c4693b8 CS |
1011 | lxc_attach_options_t* options = payload->options; |
1012 | struct lxc_proc_context_info* init_ctx = payload->init_ctx; | |
1a2e58cf | 1013 | #if HAVE_SYS_PERSONALITY_H |
9c4693b8 | 1014 | long new_personality; |
1a2e58cf | 1015 | #endif |
9c4693b8 CS |
1016 | int ret; |
1017 | int status; | |
1018 | int expected; | |
1019 | long flags; | |
1020 | int fd; | |
1021 | uid_t new_uid; | |
1022 | gid_t new_gid; | |
1023 | ||
1024 | /* wait for the initial thread to signal us that it's ready | |
1025 | * for us to start initializing | |
1026 | */ | |
1027 | expected = 0; | |
1028 | status = -1; | |
1029 | ret = lxc_read_nointr_expect(ipc_socket, &status, sizeof(status), &expected); | |
1030 | if (ret <= 0) { | |
1031 | ERROR("error using IPC to receive notification from initial process (0)"); | |
1032 | shutdown(ipc_socket, SHUT_RDWR); | |
1033 | rexit(-1); | |
1034 | } | |
1035 | ||
9c4693b8 CS |
1036 | /* A description of the purpose of this functionality is |
1037 | * provided in the lxc-attach(1) manual page. We have to | |
1038 | * remount here and not in the parent process, otherwise | |
1039 | * /proc may not properly reflect the new pid namespace. | |
1040 | */ | |
1041 | if (!(options->namespaces & CLONE_NEWNS) && (options->attach_flags & LXC_ATTACH_REMOUNT_PROC_SYS)) { | |
1042 | ret = lxc_attach_remount_sys_proc(); | |
1043 | if (ret < 0) { | |
1044 | shutdown(ipc_socket, SHUT_RDWR); | |
1045 | rexit(-1); | |
1046 | } | |
1047 | } | |
1048 | ||
1049 | /* now perform additional attachments*/ | |
1050 | #if HAVE_SYS_PERSONALITY_H | |
1051 | if (options->personality < 0) | |
1052 | new_personality = init_ctx->personality; | |
1053 | else | |
1054 | new_personality = options->personality; | |
1055 | ||
1056 | if (options->attach_flags & LXC_ATTACH_SET_PERSONALITY) { | |
1057 | ret = personality(new_personality); | |
1058 | if (ret < 0) { | |
1059 | SYSERROR("could not ensure correct architecture"); | |
1060 | shutdown(ipc_socket, SHUT_RDWR); | |
1061 | rexit(-1); | |
1062 | } | |
1063 | } | |
1064 | #endif | |
1065 | ||
1066 | if (options->attach_flags & LXC_ATTACH_DROP_CAPABILITIES) { | |
1067 | ret = lxc_attach_drop_privs(init_ctx); | |
1068 | if (ret < 0) { | |
1069 | ERROR("could not drop privileges"); | |
1070 | shutdown(ipc_socket, SHUT_RDWR); | |
1071 | rexit(-1); | |
1072 | } | |
1073 | } | |
1074 | ||
1075 | /* always set the environment (specify (LXC_ATTACH_KEEP_ENV, NULL, NULL) if you want this to be a no-op) */ | |
1076 | ret = lxc_attach_set_environment(options->env_policy, options->extra_env_vars, options->extra_keep_env); | |
1077 | if (ret < 0) { | |
1078 | ERROR("could not set initial environment for attached process"); | |
1079 | shutdown(ipc_socket, SHUT_RDWR); | |
1080 | rexit(-1); | |
1081 | } | |
1082 | ||
1083 | /* set user / group id */ | |
1084 | new_uid = 0; | |
1085 | new_gid = 0; | |
1086 | /* ignore errors, we will fall back to root in that case | |
1087 | * (/proc was not mounted etc.) | |
1088 | */ | |
1089 | if (options->namespaces & CLONE_NEWUSER) | |
1090 | lxc_attach_get_init_uidgid(&new_uid, &new_gid); | |
1091 | ||
1092 | if (options->uid != (uid_t)-1) | |
1093 | new_uid = options->uid; | |
1094 | if (options->gid != (gid_t)-1) | |
1095 | new_gid = options->gid; | |
1096 | ||
82e28fe0 | 1097 | /* setup the control tty */ |
d3b63011 | 1098 | if (options->stdin_fd && isatty(options->stdin_fd)) { |
82e28fe0 SG |
1099 | if (setsid() < 0) { |
1100 | SYSERROR("unable to setsid"); | |
1101 | shutdown(ipc_socket, SHUT_RDWR); | |
1102 | rexit(-1); | |
1103 | } | |
1104 | ||
1105 | if (ioctl(options->stdin_fd, TIOCSCTTY, (char *)NULL) < 0) { | |
1106 | SYSERROR("unable to TIOCSTTY"); | |
1107 | shutdown(ipc_socket, SHUT_RDWR); | |
1108 | rexit(-1); | |
1109 | } | |
1110 | } | |
1111 | ||
9c4693b8 | 1112 | /* try to set the uid/gid combination */ |
c476bdce SH |
1113 | if ((new_gid != 0 || options->namespaces & CLONE_NEWUSER)) { |
1114 | if (setgid(new_gid) || setgroups(0, NULL)) { | |
1115 | SYSERROR("switching to container gid"); | |
1116 | shutdown(ipc_socket, SHUT_RDWR); | |
1117 | rexit(-1); | |
1118 | } | |
9c4693b8 CS |
1119 | } |
1120 | if ((new_uid != 0 || options->namespaces & CLONE_NEWUSER) && setuid(new_uid)) { | |
1121 | SYSERROR("switching to container uid"); | |
1122 | shutdown(ipc_socket, SHUT_RDWR); | |
1123 | rexit(-1); | |
1124 | } | |
1125 | ||
1126 | /* tell initial process it may now put us into the cgroups */ | |
1127 | status = 1; | |
1128 | ret = lxc_write_nointr(ipc_socket, &status, sizeof(status)); | |
1129 | if (ret != sizeof(status)) { | |
1130 | ERROR("error using IPC to notify initial process for initialization (1)"); | |
1131 | shutdown(ipc_socket, SHUT_RDWR); | |
1132 | rexit(-1); | |
1133 | } | |
1134 | ||
1135 | /* wait for the initial thread to signal us that it has done | |
1136 | * everything for us when it comes to cgroups etc. | |
1137 | */ | |
1138 | expected = 2; | |
1139 | status = -1; | |
1140 | ret = lxc_read_nointr_expect(ipc_socket, &status, sizeof(status), &expected); | |
1141 | if (ret <= 0) { | |
1142 | ERROR("error using IPC to receive final notification from initial process (2)"); | |
1143 | shutdown(ipc_socket, SHUT_RDWR); | |
1144 | rexit(-1); | |
1145 | } | |
1146 | ||
1147 | shutdown(ipc_socket, SHUT_RDWR); | |
1148 | close(ipc_socket); | |
72863294 DE |
1149 | |
1150 | /* set new apparmor profile/selinux context */ | |
5c3fcae7 | 1151 | if ((options->namespaces & CLONE_NEWNS) && (options->attach_flags & LXC_ATTACH_LSM) && init_ctx->lsm_label) { |
72863294 DE |
1152 | int on_exec; |
1153 | ||
1154 | on_exec = options->attach_flags & LXC_ATTACH_LSM_EXEC ? 1 : 0; | |
5c3fcae7 | 1155 | if (lsm_set_label_at(procfd, on_exec, init_ctx->lsm_label) < 0) { |
72863294 DE |
1156 | rexit(-1); |
1157 | } | |
1158 | } | |
2c4ea790 SH |
1159 | |
1160 | if (init_ctx->container && init_ctx->container->lxc_conf && | |
1161 | lxc_seccomp_load(init_ctx->container->lxc_conf) != 0) { | |
1162 | ERROR("Loading seccomp policy"); | |
1163 | rexit(-1); | |
1164 | } | |
1165 | ||
fe4de9a6 | 1166 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1167 | |
1168 | /* The following is done after the communication socket is | |
1169 | * shut down. That way, all errors that might (though | |
1170 | * unlikely) occur up until this point will have their messages | |
1171 | * printed to the original stderr (if logging is so configured) | |
1172 | * and not the fd the user supplied, if any. | |
1173 | */ | |
1174 | ||
1175 | /* fd handling for stdin, stdout and stderr; | |
1176 | * ignore errors here, user may want to make sure | |
1177 | * the fds are closed, for example */ | |
1178 | if (options->stdin_fd >= 0 && options->stdin_fd != 0) | |
1179 | dup2(options->stdin_fd, 0); | |
1180 | if (options->stdout_fd >= 0 && options->stdout_fd != 1) | |
1181 | dup2(options->stdout_fd, 1); | |
1182 | if (options->stderr_fd >= 0 && options->stderr_fd != 2) | |
1183 | dup2(options->stderr_fd, 2); | |
1184 | ||
1185 | /* close the old fds */ | |
1186 | if (options->stdin_fd > 2) | |
1187 | close(options->stdin_fd); | |
1188 | if (options->stdout_fd > 2) | |
1189 | close(options->stdout_fd); | |
1190 | if (options->stderr_fd > 2) | |
1191 | close(options->stderr_fd); | |
1192 | ||
1193 | /* try to remove CLOEXEC flag from stdin/stdout/stderr, | |
1194 | * but also here, ignore errors */ | |
1195 | for (fd = 0; fd <= 2; fd++) { | |
1196 | flags = fcntl(fd, F_GETFL); | |
1197 | if (flags < 0) | |
1198 | continue; | |
71b2940d SG |
1199 | if (flags & FD_CLOEXEC) { |
1200 | if (fcntl(fd, F_SETFL, flags & ~FD_CLOEXEC) < 0) { | |
1201 | SYSERROR("Unable to clear CLOEXEC from fd"); | |
1202 | } | |
1203 | } | |
9c4693b8 CS |
1204 | } |
1205 | ||
5c3fcae7 SG |
1206 | /* we don't need proc anymore */ |
1207 | close(procfd); | |
1208 | ||
9c4693b8 CS |
1209 | /* we're done, so we can now do whatever the user intended us to do */ |
1210 | rexit(payload->exec_function(payload->exec_payload)); | |
1211 | } | |
1212 | ||
1213 | int lxc_attach_run_command(void* payload) | |
1214 | { | |
1215 | lxc_attach_command_t* cmd = (lxc_attach_command_t*)payload; | |
1216 | ||
1217 | execvp(cmd->program, cmd->argv); | |
1218 | SYSERROR("failed to exec '%s'", cmd->program); | |
1219 | return -1; | |
1220 | } | |
1221 | ||
1222 | int lxc_attach_run_shell(void* payload) | |
1223 | { | |
1224 | uid_t uid; | |
1225 | struct passwd *passwd; | |
1226 | char *user_shell; | |
1227 | ||
1228 | /* ignore payload parameter */ | |
1229 | (void)payload; | |
1230 | ||
1231 | uid = getuid(); | |
1232 | passwd = getpwuid(uid); | |
1233 | ||
1234 | /* this probably happens because of incompatible nss | |
1235 | * implementations in host and container (remember, this | |
1236 | * code is still using the host's glibc but our mount | |
1237 | * namespace is in the container) | |
1238 | * we may try to get the information by spawning a | |
1239 | * [getent passwd uid] process and parsing the result | |
1240 | */ | |
1241 | if (!passwd) | |
1242 | user_shell = lxc_attach_getpwshell(uid); | |
1243 | else | |
1244 | user_shell = passwd->pw_shell; | |
1245 | ||
1246 | if (user_shell) | |
acf47e1b | 1247 | execlp(user_shell, user_shell, (char *)NULL); |
9c4693b8 CS |
1248 | |
1249 | /* executed if either no passwd entry or execvp fails, | |
1250 | * we will fall back on /bin/sh as a default shell | |
1251 | */ | |
acf47e1b | 1252 | execlp("/bin/sh", "/bin/sh", (char *)NULL); |
9c4693b8 CS |
1253 | SYSERROR("failed to exec shell"); |
1254 | return -1; | |
1255 | } |