]>
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 | |
e0732705 CS |
25 | #include <errno.h> |
26 | #include <fcntl.h> | |
c476bdce | 27 | #include <grp.h> |
6f4f1937 CB |
28 | #include <pwd.h> |
29 | #include <signal.h> | |
30 | #include <stdio.h> | |
31 | #include <stdlib.h> | |
32 | #include <string.h> | |
33 | #include <unistd.h> | |
34 | #include <linux/unistd.h> | |
35 | #include <sys/mount.h> | |
e0732705 CS |
36 | #include <sys/param.h> |
37 | #include <sys/prctl.h> | |
5ec27989 | 38 | #include <sys/socket.h> |
1ba0013f | 39 | #include <sys/syscall.h> |
905022f7 | 40 | #include <sys/wait.h> |
6f4f1937 CB |
41 | |
42 | #include <lxc/lxccontainer.h> | |
e0732705 | 43 | |
955e2a02 | 44 | #ifndef HAVE_DECL_PR_CAPBSET_DROP |
e0732705 CS |
45 | #define PR_CAPBSET_DROP 24 |
46 | #endif | |
47 | ||
955e2a02 CB |
48 | #ifndef HAVE_DECL_PR_SET_NO_NEW_PRIVS |
49 | #define PR_SET_NO_NEW_PRIVS 38 | |
50 | #endif | |
51 | ||
52 | #ifndef HAVE_DECL_PR_GET_NO_NEW_PRIVS | |
53 | #define PR_GET_NO_NEW_PRIVS 39 | |
54 | #endif | |
55 | ||
81f466d0 | 56 | #include "af_unix.h" |
e0732705 CS |
57 | #include "attach.h" |
58 | #include "caps.h" | |
9c4693b8 | 59 | #include "cgroup.h" |
6f4f1937 | 60 | #include "commands.h" |
2c4ea790 | 61 | #include "conf.h" |
6f4f1937 | 62 | #include "config.h" |
9b8e3c96 | 63 | #include "confile.h" |
6f4f1937 CB |
64 | #include "log.h" |
65 | #include "lsm/lsm.h" | |
66 | #include "lxclock.h" | |
67 | #include "lxcseccomp.h" | |
ba2be1a8 | 68 | #include "mainloop.h" |
6f4f1937 | 69 | #include "namespace.h" |
0ed9b1bc | 70 | #include "terminal.h" |
6f4f1937 | 71 | #include "utils.h" |
9c4693b8 CS |
72 | |
73 | #if HAVE_SYS_PERSONALITY_H | |
74 | #include <sys/personality.h> | |
75 | #endif | |
e0732705 | 76 | |
a3da2f3b | 77 | #ifndef SOCK_CLOEXEC |
6f4f1937 | 78 | #define SOCK_CLOEXEC 02000000 |
a3da2f3b SG |
79 | #endif |
80 | ||
d6a3c917 SG |
81 | #ifndef MS_REC |
82 | #define MS_REC 16384 | |
83 | #endif | |
84 | ||
85 | #ifndef MS_SLAVE | |
6f4f1937 | 86 | #define MS_SLAVE (1 << 19) |
d6a3c917 SG |
87 | #endif |
88 | ||
e0732705 CS |
89 | lxc_log_define(lxc_attach, lxc); |
90 | ||
8ce83369 | 91 | /* /proc/pid-to-str/status\0 = (5 + 21 + 7 + 1) */ |
eab15c1e | 92 | #define __PROC_STATUS_LEN (5 + (LXC_NUMSTRLEN64) + 7 + 1) |
74a3920a | 93 | static struct lxc_proc_context_info *lxc_proc_get_context_info(pid_t pid) |
e0732705 | 94 | { |
6f4f1937 CB |
95 | int ret; |
96 | bool found; | |
e0732705 | 97 | FILE *proc_file; |
8ce83369 | 98 | char proc_fn[__PROC_STATUS_LEN]; |
e0732705 | 99 | size_t line_bufsz = 0; |
6f4f1937 | 100 | char *line = NULL; |
8ce83369 | 101 | struct lxc_proc_context_info *info = NULL; |
e0732705 | 102 | |
8ce83369 CB |
103 | /* Read capabilities. */ |
104 | ret = snprintf(proc_fn, __PROC_STATUS_LEN, "/proc/%d/status", pid); | |
105 | if (ret < 0 || ret >= __PROC_STATUS_LEN) | |
106 | goto on_error; | |
e0732705 CS |
107 | |
108 | proc_file = fopen(proc_fn, "r"); | |
109 | if (!proc_file) { | |
8ce83369 CB |
110 | SYSERROR("Could not open %s.", proc_fn); |
111 | goto on_error; | |
e0732705 CS |
112 | } |
113 | ||
8ce83369 CB |
114 | info = calloc(1, sizeof(*info)); |
115 | if (!info) { | |
116 | SYSERROR("Could not allocate memory."); | |
17ac5301 | 117 | fclose(proc_file); |
8ce83369 CB |
118 | return NULL; |
119 | } | |
120 | ||
121 | found = false; | |
e0732705 CS |
122 | while (getline(&line, &line_bufsz, proc_file) != -1) { |
123 | ret = sscanf(line, "CapBnd: %llx", &info->capability_mask); | |
8ce83369 CB |
124 | if (ret != EOF && ret == 1) { |
125 | found = true; | |
e0732705 CS |
126 | break; |
127 | } | |
128 | } | |
129 | ||
f10fad2f | 130 | free(line); |
e0732705 CS |
131 | fclose(proc_file); |
132 | ||
133 | if (!found) { | |
6f4f1937 CB |
134 | SYSERROR("Could not read capability bounding set from %s.", |
135 | proc_fn); | |
e0732705 | 136 | errno = ENOENT; |
8ce83369 | 137 | goto on_error; |
e0732705 CS |
138 | } |
139 | ||
fe4de9a6 | 140 | info->lsm_label = lsm_process_label_get(pid); |
877f3a04 CB |
141 | info->ns_inherited = 0; |
142 | memset(info->ns_fd, -1, sizeof(int) * LXC_NS_MAX); | |
e0732705 | 143 | |
e0732705 CS |
144 | return info; |
145 | ||
8ce83369 | 146 | on_error: |
460a1cf0 | 147 | free(info); |
e0732705 CS |
148 | return NULL; |
149 | } | |
150 | ||
877f3a04 CB |
151 | static inline void lxc_proc_close_ns_fd(struct lxc_proc_context_info *ctx) |
152 | { | |
153 | int i; | |
154 | ||
155 | for (i = 0; i < LXC_NS_MAX; i++) { | |
156 | if (ctx->ns_fd[i] < 0) | |
157 | continue; | |
158 | close(ctx->ns_fd[i]); | |
159 | ctx->ns_fd[i] = -EBADF; | |
160 | } | |
161 | } | |
162 | ||
fe4de9a6 DE |
163 | static void lxc_proc_put_context_info(struct lxc_proc_context_info *ctx) |
164 | { | |
f10fad2f | 165 | free(ctx->lsm_label); |
08ea9270 CB |
166 | ctx->lsm_label = NULL; |
167 | ||
168 | if (ctx->container) { | |
2c4ea790 | 169 | lxc_container_put(ctx->container); |
08ea9270 CB |
170 | ctx->container = NULL; |
171 | } | |
172 | ||
877f3a04 | 173 | lxc_proc_close_ns_fd(ctx); |
fe4de9a6 DE |
174 | free(ctx); |
175 | } | |
176 | ||
299d1198 CB |
177 | /** |
178 | * in_same_namespace - Check whether two processes are in the same namespace. | |
179 | * @pid1 - PID of the first process. | |
180 | * @pid2 - PID of the second process. | |
181 | * @ns - Name of the namespace to check. Must correspond to one of the names | |
182 | * for the namespaces as shown in /proc/<pid/ns/ | |
183 | * | |
184 | * If the two processes are not in the same namespace returns an fd to the | |
185 | * namespace of the second process identified by @pid2. If the two processes are | |
186 | * in the same namespace returns -EINVAL, -1 if an error occurred. | |
187 | */ | |
188 | static int in_same_namespace(pid_t pid1, pid_t pid2, const char *ns) | |
189 | { | |
190 | int ns_fd1 = -1, ns_fd2 = -1, ret = -1; | |
191 | struct stat ns_st1, ns_st2; | |
192 | ||
193 | ns_fd1 = lxc_preserve_ns(pid1, ns); | |
134284c3 CB |
194 | if (ns_fd1 < 0) { |
195 | /* The kernel does not support this namespace. This is not an | |
196 | * error. | |
197 | */ | |
198 | if (errno == ENOENT) | |
199 | return -EINVAL; | |
200 | ||
299d1198 | 201 | goto out; |
134284c3 | 202 | } |
299d1198 CB |
203 | |
204 | ns_fd2 = lxc_preserve_ns(pid2, ns); | |
205 | if (ns_fd2 < 0) | |
206 | goto out; | |
207 | ||
208 | ret = fstat(ns_fd1, &ns_st1); | |
209 | if (ret < 0) | |
210 | goto out; | |
211 | ||
212 | ret = fstat(ns_fd2, &ns_st2); | |
213 | if (ret < 0) | |
214 | goto out; | |
215 | ||
216 | /* processes are in the same namespace */ | |
217 | ret = -EINVAL; | |
218 | if ((ns_st1.st_dev == ns_st2.st_dev ) && (ns_st1.st_ino == ns_st2.st_ino)) | |
219 | goto out; | |
220 | ||
221 | /* processes are in different namespaces */ | |
222 | ret = ns_fd2; | |
223 | ns_fd2 = -1; | |
224 | ||
225 | out: | |
226 | ||
227 | if (ns_fd1 >= 0) | |
228 | close(ns_fd1); | |
229 | if (ns_fd2 >= 0) | |
230 | close(ns_fd2); | |
231 | ||
232 | return ret; | |
233 | } | |
234 | ||
877f3a04 | 235 | static int lxc_attach_to_ns(pid_t pid, struct lxc_proc_context_info *ctx) |
99d50954 | 236 | { |
877f3a04 | 237 | int i, ret; |
99d50954 | 238 | |
26818618 | 239 | for (i = 0; i < LXC_NS_MAX; i++) { |
877f3a04 | 240 | if (ctx->ns_fd[i] < 0) |
26818618 CB |
241 | continue; |
242 | ||
877f3a04 CB |
243 | ret = setns(ctx->ns_fd[i], ns_info[i].clone_flag); |
244 | if (ret < 0) { | |
299d1198 CB |
245 | SYSERROR("Failed to attach to %s namespace of %d", |
246 | ns_info[i].proc_name, pid); | |
99d50954 CS |
247 | return -1; |
248 | } | |
249 | ||
299d1198 | 250 | DEBUG("Attached to %s namespace of %d", ns_info[i].proc_name, pid); |
99d50954 CS |
251 | } |
252 | ||
253 | return 0; | |
254 | } | |
255 | ||
74a3920a | 256 | static int lxc_attach_remount_sys_proc(void) |
7a0b0b56 CS |
257 | { |
258 | int ret; | |
259 | ||
260 | ret = unshare(CLONE_NEWNS); | |
261 | if (ret < 0) { | |
8ce83369 | 262 | SYSERROR("Failed to unshare mount namespace."); |
7a0b0b56 CS |
263 | return -1; |
264 | } | |
265 | ||
2c6f3fc9 | 266 | if (detect_shared_rootfs()) { |
6f4f1937 | 267 | if (mount(NULL, "/", NULL, MS_SLAVE | MS_REC, NULL)) { |
8ce83369 | 268 | SYSERROR("Failed to make / rslave."); |
2c6f3fc9 SH |
269 | ERROR("Continuing..."); |
270 | } | |
271 | } | |
272 | ||
8ce83369 | 273 | /* Assume /proc is always mounted, so remount it. */ |
7a0b0b56 CS |
274 | ret = umount2("/proc", MNT_DETACH); |
275 | if (ret < 0) { | |
8ce83369 | 276 | SYSERROR("Failed to unmount /proc."); |
7a0b0b56 CS |
277 | return -1; |
278 | } | |
279 | ||
280 | ret = mount("none", "/proc", "proc", 0, NULL); | |
281 | if (ret < 0) { | |
8ce83369 | 282 | SYSERROR("Failed to remount /proc."); |
7a0b0b56 CS |
283 | return -1; |
284 | } | |
285 | ||
8ce83369 CB |
286 | /* Try to umount /sys. If it's not a mount point, we'll get EINVAL, then |
287 | * we ignore it because it may not have been mounted in the first place. | |
7a0b0b56 CS |
288 | */ |
289 | ret = umount2("/sys", MNT_DETACH); | |
290 | if (ret < 0 && errno != EINVAL) { | |
8ce83369 | 291 | SYSERROR("Failed to unmount /sys."); |
7a0b0b56 CS |
292 | return -1; |
293 | } else if (ret == 0) { | |
8ce83369 | 294 | /* Remount it. */ |
7a0b0b56 CS |
295 | ret = mount("none", "/sys", "sysfs", 0, NULL); |
296 | if (ret < 0) { | |
8ce83369 | 297 | SYSERROR("Failed to remount /sys."); |
7a0b0b56 CS |
298 | return -1; |
299 | } | |
300 | } | |
301 | ||
302 | return 0; | |
303 | } | |
304 | ||
74a3920a | 305 | static int lxc_attach_drop_privs(struct lxc_proc_context_info *ctx) |
e0732705 | 306 | { |
6f4f1937 | 307 | int cap, last_cap; |
e0732705 | 308 | |
6f4f1937 | 309 | last_cap = lxc_caps_last_cap(); |
e0732705 CS |
310 | for (cap = 0; cap <= last_cap; cap++) { |
311 | if (ctx->capability_mask & (1LL << cap)) | |
312 | continue; | |
313 | ||
314 | if (prctl(PR_CAPBSET_DROP, cap, 0, 0, 0)) { | |
94ac256f | 315 | SYSERROR("Failed to drop capability %d", cap); |
e0732705 CS |
316 | return -1; |
317 | } | |
94ac256f | 318 | TRACE("Dropped capability %d", cap); |
e0732705 CS |
319 | } |
320 | ||
321 | return 0; | |
322 | } | |
905022f7 | 323 | |
7385273f | 324 | static int lxc_attach_set_environment(struct lxc_proc_context_info *init_ctx, |
325 | enum lxc_attach_env_policy_t policy, | |
6f4f1937 | 326 | char **extra_env, char **extra_keep) |
b3a39ba6 | 327 | { |
3d55242a | 328 | int ret; |
7385273f | 329 | struct lxc_list *iterator; |
330 | ||
799f96fd | 331 | if (policy == LXC_ATTACH_CLEAR_ENV) { |
3d5e9f48 | 332 | int path_kept = 0; |
6f4f1937 | 333 | char **extra_keep_store = NULL; |
3d5e9f48 CS |
334 | |
335 | if (extra_keep) { | |
336 | size_t count, i; | |
337 | ||
3d55242a CB |
338 | for (count = 0; extra_keep[count]; count++) |
339 | ; | |
3d5e9f48 CS |
340 | |
341 | extra_keep_store = calloc(count, sizeof(char *)); | |
3d55242a | 342 | if (!extra_keep_store) |
3d5e9f48 | 343 | return -1; |
3d55242a | 344 | |
3d5e9f48 CS |
345 | for (i = 0; i < count; i++) { |
346 | char *v = getenv(extra_keep[i]); | |
347 | if (v) { | |
348 | extra_keep_store[i] = strdup(v); | |
349 | if (!extra_keep_store[i]) { | |
3d5e9f48 CS |
350 | while (i > 0) |
351 | free(extra_keep_store[--i]); | |
352 | free(extra_keep_store); | |
353 | return -1; | |
354 | } | |
3d55242a | 355 | |
3d5e9f48 CS |
356 | if (strcmp(extra_keep[i], "PATH") == 0) |
357 | path_kept = 1; | |
358 | } | |
3d5e9f48 CS |
359 | } |
360 | } | |
361 | ||
799f96fd | 362 | if (clearenv()) { |
a9cab7e3 | 363 | if (extra_keep_store) { |
3d55242a CB |
364 | char **p; |
365 | ||
a9cab7e3 CS |
366 | for (p = extra_keep_store; *p; p++) |
367 | free(*p); | |
3d55242a | 368 | |
a9cab7e3 CS |
369 | free(extra_keep_store); |
370 | } | |
3d55242a CB |
371 | |
372 | SYSERROR("Failed to clear environment"); | |
3d5e9f48 CS |
373 | return -1; |
374 | } | |
375 | ||
376 | if (extra_keep_store) { | |
377 | size_t i; | |
6f4f1937 | 378 | |
3d5e9f48 | 379 | for (i = 0; extra_keep[i]; i++) { |
acd4922e | 380 | if (extra_keep_store[i]) { |
3d55242a CB |
381 | ret = setenv(extra_keep[i], extra_keep_store[i], 1); |
382 | if (ret < 0) | |
383 | WARN("%s - Failed to set environment variable", strerror(errno)); | |
acd4922e | 384 | } |
3d5e9f48 CS |
385 | free(extra_keep_store[i]); |
386 | } | |
387 | free(extra_keep_store); | |
388 | } | |
389 | ||
8ce83369 CB |
390 | /* Always set a default path; shells and execlp tend to be fine |
391 | * without it, but there is a disturbing number of C programs | |
392 | * out there that just assume that getenv("PATH") is never NULL | |
393 | * and then die a painful segfault death. | |
394 | */ | |
3d55242a CB |
395 | if (!path_kept) { |
396 | ret = setenv("PATH", "/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", 1); | |
397 | if (ret < 0) | |
398 | WARN("%s - Failed to set environment variable", strerror(errno)); | |
399 | } | |
b3a39ba6 DW |
400 | } |
401 | ||
3d55242a CB |
402 | ret = putenv("container=lxc"); |
403 | if (ret < 0) { | |
404 | WARN("%s - Failed to set environment variable", strerror(errno)); | |
b3a39ba6 DW |
405 | return -1; |
406 | } | |
407 | ||
7385273f | 408 | /* Set container environment variables.*/ |
409 | if (init_ctx && init_ctx->container && init_ctx->container->lxc_conf) { | |
410 | lxc_list_for_each(iterator, &init_ctx->container->lxc_conf->environment) { | |
3d55242a CB |
411 | char *env_tmp; |
412 | ||
413 | env_tmp = strdup((char *)iterator->elem); | |
414 | if (!env_tmp) | |
7385273f | 415 | return -1; |
7385273f | 416 | |
3d55242a CB |
417 | ret = putenv(env_tmp); |
418 | if (ret < 0) { | |
419 | SYSERROR("Failed to set environment variable: %s", (char *)iterator->elem); | |
7385273f | 420 | return -1; |
421 | } | |
422 | } | |
423 | } | |
424 | ||
8ce83369 | 425 | /* Set extra environment variables. */ |
3d5e9f48 CS |
426 | if (extra_env) { |
427 | for (; *extra_env; extra_env++) { | |
3d55242a | 428 | char *p; |
8ce83369 CB |
429 | /* We just assume the user knows what they are doing, so |
430 | * we don't do any checks. | |
431 | */ | |
3d55242a CB |
432 | p = strdup(*extra_env); |
433 | if (!p) | |
3d5e9f48 | 434 | return -1; |
3d55242a CB |
435 | |
436 | ret = putenv(p); | |
437 | if (ret < 0) | |
438 | WARN("%s - Failed to set environment variable", strerror(errno)); | |
3d5e9f48 CS |
439 | } |
440 | } | |
441 | ||
b3a39ba6 DW |
442 | return 0; |
443 | } | |
444 | ||
74a3920a | 445 | static char *lxc_attach_getpwshell(uid_t uid) |
905022f7 | 446 | { |
6f4f1937 | 447 | int fd, ret; |
905022f7 CS |
448 | pid_t pid; |
449 | int pipes[2]; | |
905022f7 CS |
450 | char *result = NULL; |
451 | ||
8ce83369 CB |
452 | /* We need to fork off a process that runs the getent program, and we |
453 | * need to capture its output, so we use a pipe for that purpose. | |
905022f7 CS |
454 | */ |
455 | ret = pipe(pipes); | |
456 | if (ret < 0) | |
457 | return NULL; | |
458 | ||
459 | pid = fork(); | |
460 | if (pid < 0) { | |
461 | close(pipes[0]); | |
462 | close(pipes[1]); | |
463 | return NULL; | |
464 | } | |
465 | ||
466 | if (pid) { | |
6f4f1937 | 467 | int status; |
905022f7 | 468 | FILE *pipe_f; |
905022f7 | 469 | int found = 0; |
6f4f1937 CB |
470 | size_t line_bufsz = 0; |
471 | char *line = NULL; | |
905022f7 CS |
472 | |
473 | close(pipes[1]); | |
474 | ||
475 | pipe_f = fdopen(pipes[0], "r"); | |
476 | while (getline(&line, &line_bufsz, pipe_f) != -1) { | |
905022f7 | 477 | int i; |
6f4f1937 CB |
478 | long value; |
479 | char *token; | |
480 | char *endptr = NULL, *saveptr = NULL; | |
905022f7 | 481 | |
8ce83369 CB |
482 | /* If we already found something, just continue to read |
483 | * until the pipe doesn't deliver any more data, but | |
484 | * don't modify the existing data structure. | |
905022f7 CS |
485 | */ |
486 | if (found) | |
487 | continue; | |
488 | ||
8ce83369 | 489 | /* Trim line on the right hand side. */ |
bbb8a488 | 490 | for (i = strlen(line); i > 0 && (line[i - 1] == '\n' || line[i - 1] == '\r'); --i) |
905022f7 CS |
491 | line[i - 1] = '\0'; |
492 | ||
8ce83369 | 493 | /* Split into tokens: first: user name. */ |
905022f7 CS |
494 | token = strtok_r(line, ":", &saveptr); |
495 | if (!token) | |
496 | continue; | |
497 | /* next: dummy password field */ | |
498 | token = strtok_r(NULL, ":", &saveptr); | |
499 | if (!token) | |
500 | continue; | |
501 | /* next: user id */ | |
502 | token = strtok_r(NULL, ":", &saveptr); | |
503 | value = token ? strtol(token, &endptr, 10) : 0; | |
504 | if (!token || !endptr || *endptr || value == LONG_MIN || value == LONG_MAX) | |
505 | continue; | |
506 | /* dummy sanity check: user id matches */ | |
507 | if ((uid_t) value != uid) | |
508 | continue; | |
509 | /* skip fields: gid, gecos, dir, go to next field 'shell' */ | |
510 | for (i = 0; i < 4; i++) { | |
511 | token = strtok_r(NULL, ":", &saveptr); | |
512 | if (!token) | |
513 | break; | |
514 | } | |
515 | if (!token) | |
516 | continue; | |
f10fad2f | 517 | free(result); |
905022f7 CS |
518 | result = strdup(token); |
519 | ||
8ce83369 | 520 | /* Sanity check that there are no fields after that. */ |
905022f7 CS |
521 | token = strtok_r(NULL, ":", &saveptr); |
522 | if (token) | |
523 | continue; | |
524 | ||
525 | found = 1; | |
526 | } | |
527 | ||
528 | free(line); | |
529 | fclose(pipe_f); | |
530 | again: | |
531 | if (waitpid(pid, &status, 0) < 0) { | |
532 | if (errno == EINTR) | |
533 | goto again; | |
f4e1fa60 | 534 | free(result); |
905022f7 CS |
535 | return NULL; |
536 | } | |
537 | ||
8ce83369 CB |
538 | /* Some sanity checks. If anything even hinted at going wrong, |
539 | * we can't be sure we have a valid result, so we assume we | |
540 | * don't. | |
905022f7 CS |
541 | */ |
542 | ||
f4e1fa60 CB |
543 | if (!WIFEXITED(status)) { |
544 | free(result); | |
905022f7 | 545 | return NULL; |
f4e1fa60 | 546 | } |
905022f7 | 547 | |
f4e1fa60 CB |
548 | if (WEXITSTATUS(status) != 0) { |
549 | free(result); | |
905022f7 | 550 | return NULL; |
f4e1fa60 | 551 | } |
905022f7 | 552 | |
f4e1fa60 CB |
553 | if (!found) { |
554 | free(result); | |
905022f7 | 555 | return NULL; |
f4e1fa60 | 556 | } |
905022f7 CS |
557 | |
558 | return result; | |
559 | } else { | |
905022f7 CS |
560 | char uid_buf[32]; |
561 | char *arguments[] = { | |
562 | "getent", | |
563 | "passwd", | |
564 | uid_buf, | |
565 | NULL | |
566 | }; | |
567 | ||
568 | close(pipes[0]); | |
569 | ||
8ce83369 | 570 | /* We want to capture stdout. */ |
905022f7 CS |
571 | dup2(pipes[1], 1); |
572 | close(pipes[1]); | |
573 | ||
8ce83369 CB |
574 | /* Get rid of stdin/stderr, so we try to associate it with |
575 | * /dev/null. | |
905022f7 CS |
576 | */ |
577 | fd = open("/dev/null", O_RDWR); | |
578 | if (fd < 0) { | |
579 | close(0); | |
580 | close(2); | |
581 | } else { | |
582 | dup2(fd, 0); | |
583 | dup2(fd, 2); | |
584 | close(fd); | |
585 | } | |
586 | ||
8ce83369 | 587 | /* Finish argument list. */ |
905022f7 CS |
588 | ret = snprintf(uid_buf, sizeof(uid_buf), "%ld", (long) uid); |
589 | if (ret <= 0) | |
590 | exit(-1); | |
591 | ||
8ce83369 | 592 | /* Try to run getent program. */ |
905022f7 CS |
593 | (void) execvp("getent", arguments); |
594 | exit(-1); | |
595 | } | |
596 | } | |
cb3e61fa | 597 | |
6f4f1937 | 598 | static void lxc_attach_get_init_uidgid(uid_t *init_uid, gid_t *init_gid) |
cb3e61fa CS |
599 | { |
600 | FILE *proc_file; | |
8ce83369 CB |
601 | char proc_fn[__PROC_STATUS_LEN]; |
602 | int ret; | |
cb3e61fa CS |
603 | char *line = NULL; |
604 | size_t line_bufsz = 0; | |
cb3e61fa CS |
605 | long value = -1; |
606 | uid_t uid = (uid_t)-1; | |
607 | gid_t gid = (gid_t)-1; | |
608 | ||
8ce83369 CB |
609 | /* Read capabilities. */ |
610 | snprintf(proc_fn, __PROC_STATUS_LEN, "/proc/%d/status", 1); | |
cb3e61fa CS |
611 | |
612 | proc_file = fopen(proc_fn, "r"); | |
613 | if (!proc_file) | |
614 | return; | |
615 | ||
616 | while (getline(&line, &line_bufsz, proc_file) != -1) { | |
8ce83369 CB |
617 | /* Format is: real, effective, saved set user, fs we only care |
618 | * about real uid. | |
cb3e61fa CS |
619 | */ |
620 | ret = sscanf(line, "Uid: %ld", &value); | |
8ce83369 | 621 | if (ret != EOF && ret == 1) { |
6f4f1937 | 622 | uid = (uid_t)value; |
cb3e61fa CS |
623 | } else { |
624 | ret = sscanf(line, "Gid: %ld", &value); | |
8ce83369 | 625 | if (ret != EOF && ret == 1) |
6f4f1937 | 626 | gid = (gid_t)value; |
cb3e61fa CS |
627 | } |
628 | if (uid != (uid_t)-1 && gid != (gid_t)-1) | |
629 | break; | |
630 | } | |
631 | ||
632 | fclose(proc_file); | |
633 | free(line); | |
634 | ||
8ce83369 | 635 | /* Only override arguments if we found something. */ |
cb3e61fa CS |
636 | if (uid != (uid_t)-1) |
637 | *init_uid = uid; | |
638 | if (gid != (gid_t)-1) | |
639 | *init_gid = gid; | |
640 | ||
641 | /* TODO: we should also parse supplementary groups and use | |
8ce83369 CB |
642 | * setgroups() to set them. |
643 | */ | |
cb3e61fa | 644 | } |
9c4693b8 | 645 | |
8ce83369 | 646 | /* Help the optimizer along if it doesn't know that exit always exits. */ |
6f4f1937 CB |
647 | #define rexit(c) \ |
648 | do { \ | |
649 | int __c = (c); \ | |
650 | _exit(__c); \ | |
651 | return __c; \ | |
652 | } while (0) | |
9c4693b8 | 653 | |
8ce83369 | 654 | /* Define default options if no options are supplied by the user. */ |
9c4693b8 CS |
655 | static lxc_attach_options_t attach_static_default_options = LXC_ATTACH_OPTIONS_DEFAULT; |
656 | ||
bd4307f0 | 657 | static bool fetch_seccomp(struct lxc_container *c, |
ff07d7bb | 658 | lxc_attach_options_t *options) |
2c4ea790 | 659 | { |
bd7b4e28 | 660 | char *path; |
2eef2bda | 661 | |
6f4f1937 CB |
662 | if (!(options->namespaces & CLONE_NEWNS) || |
663 | !(options->attach_flags & LXC_ATTACH_LSM)) { | |
bd4307f0 CB |
664 | free(c->lxc_conf->seccomp); |
665 | c->lxc_conf->seccomp = NULL; | |
2c4ea790 | 666 | return true; |
bd4307f0 | 667 | } |
bd7b4e28 | 668 | |
2e812c16 | 669 | /* Remove current setting. */ |
0b427da0 CB |
670 | if (!c->set_config_item(c, "lxc.seccomp", "") && |
671 | !c->set_config_item(c, "lxc.seccomp.profile", "")) { | |
2c4ea790 | 672 | return false; |
bd7b4e28 SG |
673 | } |
674 | ||
8ce83369 | 675 | /* Fetch the current profile path over the cmd interface. */ |
0b427da0 | 676 | path = c->get_running_config_item(c, "lxc.seccomp.profile"); |
bd7b4e28 | 677 | if (!path) { |
0b427da0 CB |
678 | INFO("Failed to get running config item for lxc.seccomp.profile"); |
679 | path = c->get_running_config_item(c, "lxc.seccomp"); | |
680 | } | |
681 | if (!path) { | |
682 | INFO("Failed to get running config item for lxc.seccomp"); | |
bd7b4e28 SG |
683 | return true; |
684 | } | |
685 | ||
8ce83369 | 686 | /* Copy the value into the new lxc_conf. */ |
0b427da0 | 687 | if (!c->set_config_item(c, "lxc.seccomp.profile", path)) { |
bd7b4e28 SG |
688 | free(path); |
689 | return false; | |
690 | } | |
691 | free(path); | |
692 | ||
8ce83369 | 693 | /* Attempt to parse the resulting config. */ |
2c4ea790 | 694 | if (lxc_read_seccomp_config(c->lxc_conf) < 0) { |
8ce83369 | 695 | ERROR("Error reading seccomp policy."); |
2c4ea790 SH |
696 | return false; |
697 | } | |
698 | ||
2e812c16 CB |
699 | INFO("Retrieved seccomp policy."); |
700 | return true; | |
701 | } | |
702 | ||
6f4f1937 | 703 | static bool no_new_privs(struct lxc_container *c, lxc_attach_options_t *options) |
2e812c16 | 704 | { |
2e812c16 CB |
705 | char *val; |
706 | ||
2e812c16 | 707 | /* Remove current setting. */ |
6f4f1937 | 708 | if (!c->set_config_item(c, "lxc.no_new_privs", "")) |
2e812c16 | 709 | return false; |
2e812c16 CB |
710 | |
711 | /* Retrieve currently active setting. */ | |
712 | val = c->get_running_config_item(c, "lxc.no_new_privs"); | |
713 | if (!val) { | |
714 | INFO("Failed to get running config item for lxc.no_new_privs."); | |
715 | return false; | |
716 | } | |
717 | ||
718 | /* Set currently active setting. */ | |
719 | if (!c->set_config_item(c, "lxc.no_new_privs", val)) { | |
720 | free(val); | |
721 | return false; | |
722 | } | |
723 | free(val); | |
724 | ||
2c4ea790 SH |
725 | return true; |
726 | } | |
727 | ||
9b8e3c96 SH |
728 | static signed long get_personality(const char *name, const char *lxcpath) |
729 | { | |
6f4f1937 | 730 | char *p; |
9b8e3c96 SH |
731 | signed long ret; |
732 | ||
6f4f1937 | 733 | p = lxc_cmd_get_config_item(name, "lxc.arch", lxcpath); |
9b8e3c96 SH |
734 | if (!p) |
735 | return -1; | |
6f4f1937 | 736 | |
9b8e3c96 SH |
737 | ret = lxc_config_parse_arch(p); |
738 | free(p); | |
6f4f1937 | 739 | |
9b8e3c96 SH |
740 | return ret; |
741 | } | |
742 | ||
a998454a CB |
743 | struct attach_clone_payload { |
744 | int ipc_socket; | |
9e84479f | 745 | int terminal_slave_fd; |
a998454a CB |
746 | lxc_attach_options_t *options; |
747 | struct lxc_proc_context_info *init_ctx; | |
748 | lxc_attach_exec_t exec_function; | |
749 | void *exec_payload; | |
750 | }; | |
751 | ||
ba2be1a8 CB |
752 | static void lxc_put_attach_clone_payload(struct attach_clone_payload *p) |
753 | { | |
754 | if (p->ipc_socket >= 0) { | |
755 | shutdown(p->ipc_socket, SHUT_RDWR); | |
756 | close(p->ipc_socket); | |
757 | p->ipc_socket = -EBADF; | |
758 | } | |
759 | ||
9e84479f CB |
760 | if (p->terminal_slave_fd >= 0) { |
761 | close(p->terminal_slave_fd); | |
762 | p->terminal_slave_fd = -EBADF; | |
ba2be1a8 CB |
763 | } |
764 | ||
b21da190 | 765 | if (p->init_ctx) { |
ba2be1a8 | 766 | lxc_proc_put_context_info(p->init_ctx); |
b21da190 CB |
767 | p->init_ctx = NULL; |
768 | } | |
ba2be1a8 CB |
769 | } |
770 | ||
a998454a CB |
771 | static int attach_child_main(struct attach_clone_payload *payload) |
772 | { | |
57de839f | 773 | int fd, lsm_fd, ret; |
a998454a CB |
774 | uid_t new_uid; |
775 | gid_t new_gid; | |
a998454a CB |
776 | lxc_attach_options_t* options = payload->options; |
777 | struct lxc_proc_context_info* init_ctx = payload->init_ctx; | |
57de839f CB |
778 | bool needs_lsm = (options->namespaces & CLONE_NEWNS) && |
779 | (options->attach_flags & LXC_ATTACH_LSM) && | |
780 | init_ctx->lsm_label; | |
a998454a CB |
781 | |
782 | /* A description of the purpose of this functionality is provided in the | |
783 | * lxc-attach(1) manual page. We have to remount here and not in the | |
784 | * parent process, otherwise /proc may not properly reflect the new pid | |
785 | * namespace. | |
786 | */ | |
787 | if (!(options->namespaces & CLONE_NEWNS) && | |
788 | (options->attach_flags & LXC_ATTACH_REMOUNT_PROC_SYS)) { | |
789 | ret = lxc_attach_remount_sys_proc(); | |
b75c344c CB |
790 | if (ret < 0) |
791 | goto on_error; | |
792 | TRACE("Remounted \"/proc\" and \"/sys\""); | |
a998454a CB |
793 | } |
794 | ||
b75c344c | 795 | /* Now perform additional attachments. */ |
a998454a | 796 | #if HAVE_SYS_PERSONALITY_H |
a998454a | 797 | if (options->attach_flags & LXC_ATTACH_SET_PERSONALITY) { |
b75c344c CB |
798 | long new_personality; |
799 | ||
800 | if (options->personality < 0) | |
801 | new_personality = init_ctx->personality; | |
802 | else | |
803 | new_personality = options->personality; | |
a998454a | 804 | ret = personality(new_personality); |
b75c344c CB |
805 | if (ret < 0) |
806 | goto on_error; | |
807 | TRACE("Set new personality"); | |
a998454a CB |
808 | } |
809 | #endif | |
810 | ||
811 | if (options->attach_flags & LXC_ATTACH_DROP_CAPABILITIES) { | |
812 | ret = lxc_attach_drop_privs(init_ctx); | |
b75c344c CB |
813 | if (ret < 0) |
814 | goto on_error; | |
815 | TRACE("Dropped capabilities"); | |
a998454a CB |
816 | } |
817 | ||
818 | /* Always set the environment (specify (LXC_ATTACH_KEEP_ENV, NULL, NULL) | |
819 | * if you want this to be a no-op). | |
820 | */ | |
7385273f | 821 | ret = lxc_attach_set_environment(init_ctx, |
822 | options->env_policy, | |
a998454a CB |
823 | options->extra_env_vars, |
824 | options->extra_keep_env); | |
b75c344c CB |
825 | if (ret < 0) |
826 | goto on_error; | |
827 | TRACE("Set up environment"); | |
a998454a | 828 | |
57de839f CB |
829 | /* This remark only affects fully unprivileged containers: |
830 | * Receive fd for LSM security module before we set{g,u}id(). The reason | |
831 | * is that on set{g,u}id() the kernel will a) make us undumpable and b) | |
832 | * we will change our effective uid. This means our effective uid will | |
833 | * be different from the effective uid of the process that created us | |
834 | * which means that this processs no longer has capabilities in our | |
835 | * namespace including CAP_SYS_PTRACE. This means we will not be able to | |
836 | * read and /proc/<pid> files for the process anymore when /proc is | |
837 | * mounted with hidepid={1,2}. So let's get the lsm label fd before the | |
838 | * set{g,u}id(). | |
839 | */ | |
840 | if (needs_lsm) { | |
b75c344c CB |
841 | ret = lxc_abstract_unix_recv_fds(payload->ipc_socket, &lsm_fd, 1, NULL, 0); |
842 | if (ret <= 0) | |
843 | goto on_error; | |
57de839f CB |
844 | TRACE("Received LSM label file descriptor %d from parent", lsm_fd); |
845 | } | |
846 | ||
08ea9270 | 847 | if (options->stdin_fd > 0 && isatty(options->stdin_fd)) { |
cd0a2b2f | 848 | ret = lxc_make_controlling_terminal(options->stdin_fd); |
08ea9270 CB |
849 | if (ret < 0) |
850 | goto on_error; | |
851 | } | |
852 | ||
a998454a CB |
853 | /* Set {u,g}id. */ |
854 | new_uid = 0; | |
855 | new_gid = 0; | |
856 | /* Ignore errors, we will fall back to root in that case (/proc was not | |
857 | * mounted etc.). | |
858 | */ | |
859 | if (options->namespaces & CLONE_NEWUSER) | |
860 | lxc_attach_get_init_uidgid(&new_uid, &new_gid); | |
861 | ||
862 | if (options->uid != (uid_t)-1) | |
863 | new_uid = options->uid; | |
864 | if (options->gid != (gid_t)-1) | |
865 | new_gid = options->gid; | |
866 | ||
a998454a | 867 | /* Try to set the {u,g}id combination. */ |
b75c344c CB |
868 | if (new_uid != 0 || new_gid != 0 || options->namespaces & CLONE_NEWUSER) { |
869 | ret = lxc_switch_uid_gid(new_uid, new_gid); | |
870 | if (ret < 0) | |
871 | goto on_error; | |
a998454a CB |
872 | } |
873 | ||
24927339 | 874 | ret = lxc_setgroups(0, NULL); |
96ec54ac | 875 | if (ret < 0 && errno != EPERM) |
24927339 CB |
876 | goto on_error; |
877 | ||
a998454a CB |
878 | if ((init_ctx->container && init_ctx->container->lxc_conf && |
879 | init_ctx->container->lxc_conf->no_new_privs) || | |
880 | (options->attach_flags & LXC_ATTACH_NO_NEW_PRIVS)) { | |
b75c344c CB |
881 | ret = prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0); |
882 | if (ret < 0) | |
883 | goto on_error; | |
884 | TRACE("Set PR_SET_NO_NEW_PRIVS"); | |
a998454a CB |
885 | } |
886 | ||
57de839f | 887 | if (needs_lsm) { |
d3ba7c98 | 888 | bool on_exec; |
a998454a CB |
889 | |
890 | /* Change into our new LSM profile. */ | |
d3ba7c98 CB |
891 | on_exec = options->attach_flags & LXC_ATTACH_LSM_EXEC ? true : false; |
892 | ret = lsm_process_label_set_at(lsm_fd, init_ctx->lsm_label, on_exec); | |
57de839f | 893 | close(lsm_fd); |
b75c344c CB |
894 | if (ret < 0) |
895 | goto on_error; | |
d3ba7c98 | 896 | TRACE("Set %s LSM label to \"%s\"", lsm_name(), init_ctx->lsm_label); |
a998454a CB |
897 | } |
898 | ||
899 | if (init_ctx->container && init_ctx->container->lxc_conf && | |
b75c344c CB |
900 | init_ctx->container->lxc_conf->seccomp) { |
901 | ret = lxc_seccomp_load(init_ctx->container->lxc_conf); | |
902 | if (ret < 0) | |
903 | goto on_error; | |
904 | TRACE("Loaded seccomp profile"); | |
a998454a | 905 | } |
b75c344c CB |
906 | shutdown(payload->ipc_socket, SHUT_RDWR); |
907 | close(payload->ipc_socket); | |
ba2be1a8 CB |
908 | payload->ipc_socket = -EBADF; |
909 | lxc_proc_put_context_info(init_ctx); | |
d35b372a | 910 | payload->init_ctx = NULL; |
a998454a CB |
911 | |
912 | /* The following is done after the communication socket is shut down. | |
913 | * That way, all errors that might (though unlikely) occur up until this | |
914 | * point will have their messages printed to the original stderr (if | |
915 | * logging is so configured) and not the fd the user supplied, if any. | |
916 | */ | |
917 | ||
918 | /* Fd handling for stdin, stdout and stderr; ignore errors here, user | |
919 | * may want to make sure the fds are closed, for example. | |
920 | */ | |
08ea9270 CB |
921 | if (options->stdin_fd >= 0 && options->stdin_fd != STDIN_FILENO) |
922 | dup2(options->stdin_fd, STDIN_FILENO); | |
923 | ||
924 | if (options->stdout_fd >= 0 && options->stdout_fd != STDOUT_FILENO) | |
925 | dup2(options->stdout_fd, STDOUT_FILENO); | |
926 | ||
927 | if (options->stderr_fd >= 0 && options->stderr_fd != STDERR_FILENO) | |
928 | dup2(options->stderr_fd, STDERR_FILENO); | |
a998454a CB |
929 | |
930 | /* close the old fds */ | |
08ea9270 | 931 | if (options->stdin_fd > STDERR_FILENO) |
a998454a | 932 | close(options->stdin_fd); |
08ea9270 CB |
933 | |
934 | if (options->stdout_fd > STDERR_FILENO) | |
a998454a | 935 | close(options->stdout_fd); |
08ea9270 CB |
936 | |
937 | if (options->stderr_fd > STDERR_FILENO) | |
a998454a CB |
938 | close(options->stderr_fd); |
939 | ||
940 | /* Try to remove FD_CLOEXEC flag from stdin/stdout/stderr, but also | |
941 | * here, ignore errors. | |
942 | */ | |
b75c344c CB |
943 | for (fd = STDIN_FILENO; fd <= STDERR_FILENO; fd++) { |
944 | int flags; | |
945 | ||
a998454a CB |
946 | flags = fcntl(fd, F_GETFL); |
947 | if (flags < 0) | |
948 | continue; | |
b75c344c CB |
949 | |
950 | if ((flags & FD_CLOEXEC) == 0) | |
951 | continue; | |
952 | ||
953 | ret = fcntl(fd, F_SETFL, flags & ~FD_CLOEXEC); | |
954 | if (ret < 0) { | |
955 | SYSERROR("Failed to clear FD_CLOEXEC from file descriptor %d", fd); | |
956 | goto on_error; | |
957 | } | |
a998454a CB |
958 | } |
959 | ||
9e84479f CB |
960 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
961 | ret = lxc_terminal_prepare_login(payload->terminal_slave_fd); | |
ba2be1a8 | 962 | if (ret < 0) { |
9e84479f | 963 | SYSERROR("Failed to prepare terminal file descriptor %d", payload->terminal_slave_fd); |
ba2be1a8 CB |
964 | goto on_error; |
965 | } | |
9e84479f | 966 | TRACE("Prepared terminal file descriptor %d", payload->terminal_slave_fd); |
ba2be1a8 CB |
967 | } |
968 | ||
a998454a CB |
969 | /* We're done, so we can now do whatever the user intended us to do. */ |
970 | rexit(payload->exec_function(payload->exec_payload)); | |
b75c344c CB |
971 | |
972 | on_error: | |
ba2be1a8 | 973 | lxc_put_attach_clone_payload(payload); |
b75c344c | 974 | rexit(EXIT_FAILURE); |
a998454a CB |
975 | } |
976 | ||
9e84479f CB |
977 | static int lxc_attach_terminal(struct lxc_conf *conf, |
978 | struct lxc_terminal *terminal) | |
ba2be1a8 CB |
979 | { |
980 | int ret; | |
981 | ||
9e84479f | 982 | lxc_terminal_init(terminal); |
ba2be1a8 | 983 | |
9e84479f | 984 | ret = lxc_terminal_create(terminal); |
ba2be1a8 | 985 | if (ret < 0) { |
9e84479f | 986 | SYSERROR("Failed to create terminal"); |
ba2be1a8 CB |
987 | return -1; |
988 | } | |
989 | ||
990 | /* Shift ttys to container. */ | |
9e84479f | 991 | ret = lxc_terminal_map_ids(conf, terminal); |
ba2be1a8 | 992 | if (ret < 0) { |
9e84479f | 993 | ERROR("Failed to chown terminal"); |
ba2be1a8 CB |
994 | goto on_error; |
995 | } | |
996 | ||
997 | return 0; | |
998 | ||
999 | on_error: | |
9e84479f CB |
1000 | lxc_terminal_delete(terminal); |
1001 | lxc_terminal_conf_free(terminal); | |
ba2be1a8 CB |
1002 | return -1; |
1003 | } | |
1004 | ||
9e84479f CB |
1005 | static int lxc_attach_terminal_mainloop_init(struct lxc_terminal *terminal, |
1006 | struct lxc_epoll_descr *descr) | |
ba2be1a8 CB |
1007 | { |
1008 | int ret; | |
1009 | ||
1010 | ret = lxc_mainloop_open(descr); | |
1011 | if (ret < 0) { | |
1012 | ERROR("Failed to create mainloop"); | |
1013 | return -1; | |
1014 | } | |
1015 | ||
9e84479f | 1016 | ret = lxc_terminal_mainloop_add(descr, terminal); |
ba2be1a8 CB |
1017 | if (ret < 0) { |
1018 | ERROR("Failed to add handlers to mainloop"); | |
1019 | lxc_mainloop_close(descr); | |
1020 | return -1; | |
1021 | } | |
1022 | ||
1023 | return 0; | |
1024 | } | |
1025 | ||
9e84479f | 1026 | static inline void lxc_attach_terminal_close_master(struct lxc_terminal *terminal) |
ba2be1a8 | 1027 | { |
9e84479f | 1028 | if (terminal->master < 0) |
ba2be1a8 CB |
1029 | return; |
1030 | ||
9e84479f CB |
1031 | close(terminal->master); |
1032 | terminal->master = -EBADF; | |
ba2be1a8 CB |
1033 | } |
1034 | ||
9e84479f | 1035 | static inline void lxc_attach_terminal_close_slave(struct lxc_terminal *terminal) |
ba2be1a8 | 1036 | { |
9e84479f | 1037 | if (terminal->slave < 0) |
ba2be1a8 CB |
1038 | return; |
1039 | ||
9e84479f CB |
1040 | close(terminal->slave); |
1041 | terminal->slave = -EBADF; | |
ba2be1a8 CB |
1042 | } |
1043 | ||
9e84479f | 1044 | static inline void lxc_attach_terminal_close_peer(struct lxc_terminal *terminal) |
ba2be1a8 | 1045 | { |
9e84479f | 1046 | if (terminal->peer < 0) |
ba2be1a8 CB |
1047 | return; |
1048 | ||
9e84479f CB |
1049 | close(terminal->peer); |
1050 | terminal->peer = -EBADF; | |
ba2be1a8 CB |
1051 | } |
1052 | ||
9e84479f | 1053 | static inline void lxc_attach_terminal_close_log(struct lxc_terminal *terminal) |
ba2be1a8 | 1054 | { |
9e84479f | 1055 | if (terminal->log_fd < 0) |
ba2be1a8 CB |
1056 | return; |
1057 | ||
9e84479f CB |
1058 | close(terminal->log_fd); |
1059 | terminal->log_fd = -EBADF; | |
ba2be1a8 CB |
1060 | } |
1061 | ||
6f4f1937 CB |
1062 | int lxc_attach(const char *name, const char *lxcpath, |
1063 | lxc_attach_exec_t exec_function, void *exec_payload, | |
1064 | lxc_attach_options_t *options, pid_t *attached_process) | |
9c4693b8 | 1065 | { |
877f3a04 | 1066 | int i, ret, status; |
9c4693b8 | 1067 | int ipc_sockets[2]; |
6f4f1937 | 1068 | char *cwd, *new_cwd; |
9b8e3c96 | 1069 | signed long personality; |
ba2be1a8 | 1070 | pid_t attached_pid, init_pid, pid; |
6f4f1937 | 1071 | struct lxc_proc_context_info *init_ctx; |
9e84479f | 1072 | struct lxc_terminal terminal; |
1cce35e6 | 1073 | struct lxc_conf *conf; |
a998454a | 1074 | struct attach_clone_payload payload = {0}; |
9c4693b8 | 1075 | |
877f3a04 CB |
1076 | ret = access("/proc/self/ns", X_OK); |
1077 | if (ret) { | |
1078 | ERROR("Does this kernel version support namespaces?"); | |
1079 | return -1; | |
1080 | } | |
1081 | ||
9c4693b8 CS |
1082 | if (!options) |
1083 | options = &attach_static_default_options; | |
1084 | ||
1085 | init_pid = lxc_cmd_get_init_pid(name, lxcpath); | |
1086 | if (init_pid < 0) { | |
8ce83369 | 1087 | ERROR("Failed to get init pid."); |
9c4693b8 CS |
1088 | return -1; |
1089 | } | |
1090 | ||
1091 | init_ctx = lxc_proc_get_context_info(init_pid); | |
1092 | if (!init_ctx) { | |
6f4f1937 | 1093 | ERROR("Failed to get context of init process: %ld", (long)init_pid); |
9c4693b8 CS |
1094 | return -1; |
1095 | } | |
1096 | ||
9b8e3c96 SH |
1097 | personality = get_personality(name, lxcpath); |
1098 | if (init_ctx->personality < 0) { | |
6f4f1937 | 1099 | ERROR("Failed to get personality of the container"); |
9b8e3c96 SH |
1100 | lxc_proc_put_context_info(init_ctx); |
1101 | return -1; | |
1102 | } | |
1103 | init_ctx->personality = personality; | |
1104 | ||
ff07d7bb | 1105 | init_ctx->container = lxc_container_new(name, lxcpath); |
62de1db6 CB |
1106 | if (!init_ctx->container) { |
1107 | lxc_proc_put_context_info(init_ctx); | |
ff07d7bb | 1108 | return -1; |
62de1db6 | 1109 | } |
ff07d7bb | 1110 | |
ba773996 CB |
1111 | if (!init_ctx->container->lxc_conf) { |
1112 | init_ctx->container->lxc_conf = lxc_conf_init(); | |
62de1db6 CB |
1113 | if (!init_ctx->container->lxc_conf) { |
1114 | lxc_proc_put_context_info(init_ctx); | |
ba773996 | 1115 | return -ENOMEM; |
62de1db6 | 1116 | } |
ba773996 | 1117 | } |
1cce35e6 | 1118 | conf = init_ctx->container->lxc_conf; |
ba773996 | 1119 | |
bd4307f0 | 1120 | if (!fetch_seccomp(init_ctx->container, options)) |
8ce83369 | 1121 | WARN("Failed to get seccomp policy."); |
2c4ea790 | 1122 | |
bd4307f0 | 1123 | if (!no_new_privs(init_ctx->container, options)) |
2e812c16 CB |
1124 | WARN("Could not determine whether PR_SET_NO_NEW_PRIVS is set."); |
1125 | ||
9c4693b8 CS |
1126 | cwd = getcwd(NULL, 0); |
1127 | ||
8ce83369 CB |
1128 | /* Determine which namespaces the container was created with |
1129 | * by asking lxc-start, if necessary. | |
9c4693b8 CS |
1130 | */ |
1131 | if (options->namespaces == -1) { | |
1132 | options->namespaces = lxc_cmd_get_clone_flags(name, lxcpath); | |
1133 | /* call failed */ | |
1134 | if (options->namespaces == -1) { | |
8ce83369 | 1135 | ERROR("Failed to automatically determine the " |
877f3a04 | 1136 | "namespaces which the container uses"); |
9c4693b8 | 1137 | free(cwd); |
fe4de9a6 | 1138 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1139 | return -1; |
1140 | } | |
877f3a04 CB |
1141 | |
1142 | for (i = 0; i < LXC_NS_MAX; i++) { | |
1143 | if (ns_info[i].clone_flag & CLONE_NEWCGROUP) | |
1144 | if (!(options->attach_flags & LXC_ATTACH_MOVE_TO_CGROUP) || | |
1145 | !cgns_supported()) | |
1146 | continue; | |
1147 | ||
1148 | if (ns_info[i].clone_flag & options->namespaces) | |
1149 | continue; | |
1150 | ||
1151 | init_ctx->ns_inherited |= ns_info[i].clone_flag; | |
1152 | } | |
1153 | } | |
1154 | ||
0059379f | 1155 | pid = lxc_raw_getpid(); |
877f3a04 CB |
1156 | for (i = 0; i < LXC_NS_MAX; i++) { |
1157 | int j, saved_errno; | |
1158 | ||
1159 | if (options->namespaces & ns_info[i].clone_flag) | |
1160 | init_ctx->ns_fd[i] = lxc_preserve_ns(init_pid, ns_info[i].proc_name); | |
1161 | else if (init_ctx->ns_inherited & ns_info[i].clone_flag) | |
1162 | init_ctx->ns_fd[i] = in_same_namespace(pid, init_pid, ns_info[i].proc_name); | |
1163 | else | |
1164 | continue; | |
1165 | if (init_ctx->ns_fd[i] >= 0) | |
1166 | continue; | |
1167 | ||
1168 | if (init_ctx->ns_fd[i] == -EINVAL) { | |
1169 | DEBUG("Inheriting %s namespace from %d", | |
1170 | ns_info[i].proc_name, pid); | |
1171 | init_ctx->ns_inherited &= ~ns_info[i].clone_flag; | |
1172 | continue; | |
1173 | } | |
1174 | ||
1175 | /* We failed to preserve the namespace. */ | |
1176 | saved_errno = errno; | |
1177 | /* Close all already opened file descriptors before we return an | |
1178 | * error, so we don't leak them. | |
1179 | */ | |
1180 | for (j = 0; j < i; j++) | |
1181 | close(init_ctx->ns_fd[j]); | |
1182 | ||
1183 | errno = saved_errno; | |
1184 | SYSERROR("Failed to attach to %s namespace of %d", | |
1185 | ns_info[i].proc_name, pid); | |
1186 | free(cwd); | |
1187 | lxc_proc_put_context_info(init_ctx); | |
1188 | return -1; | |
9c4693b8 CS |
1189 | } |
1190 | ||
9e84479f CB |
1191 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
1192 | ret = lxc_attach_terminal(conf, &terminal); | |
ba2be1a8 | 1193 | if (ret < 0) { |
9e84479f | 1194 | ERROR("Failed to setup new terminal"); |
ba2be1a8 CB |
1195 | free(cwd); |
1196 | lxc_proc_put_context_info(init_ctx); | |
1197 | return -1; | |
1198 | } | |
1199 | ||
9e84479f | 1200 | terminal.log_fd = options->log_fd; |
c948657b | 1201 | } else { |
9e84479f | 1202 | lxc_terminal_init(&terminal); |
ba2be1a8 CB |
1203 | } |
1204 | ||
8ce83369 CB |
1205 | /* Create a socket pair for IPC communication; set SOCK_CLOEXEC in order |
1206 | * to make sure we don't irritate other threads that want to fork+exec | |
1207 | * away | |
9c4693b8 CS |
1208 | * |
1209 | * IMPORTANT: if the initial process is multithreaded and another call | |
1210 | * just fork()s away without exec'ing directly after, the socket fd will | |
1211 | * exist in the forked process from the other thread and any close() in | |
8ce83369 CB |
1212 | * our own child process will not really cause the socket to close |
1213 | * properly, potentiall causing the parent to hang. | |
9c4693b8 CS |
1214 | * |
1215 | * For this reason, while IPC is still active, we have to use shutdown() | |
8ce83369 CB |
1216 | * if the child exits prematurely in order to signal that the socket is |
1217 | * closed and cannot assume that the child exiting will automatically do | |
1218 | * that. | |
9c4693b8 CS |
1219 | * |
1220 | * IPC mechanism: (X is receiver) | |
1221 | * initial process intermediate attached | |
1222 | * X <--- send pid of | |
1223 | * attached proc, | |
1224 | * then exit | |
1225 | * send 0 ------------------------------------> X | |
1226 | * [do initialization] | |
1227 | * X <------------------------------------ send 1 | |
1228 | * [add to cgroup, ...] | |
1229 | * send 2 ------------------------------------> X | |
81f466d0 CB |
1230 | * [set LXC_ATTACH_NO_NEW_PRIVS] |
1231 | * X <------------------------------------ send 3 | |
1232 | * [open LSM label fd] | |
1233 | * send 4 ------------------------------------> X | |
1234 | * [set LSM label] | |
9c4693b8 CS |
1235 | * close socket close socket |
1236 | * run program | |
1237 | */ | |
1238 | ret = socketpair(PF_LOCAL, SOCK_STREAM | SOCK_CLOEXEC, 0, ipc_sockets); | |
1239 | if (ret < 0) { | |
8ce83369 | 1240 | SYSERROR("Could not set up required IPC mechanism for attaching."); |
9c4693b8 | 1241 | free(cwd); |
fe4de9a6 | 1242 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1243 | return -1; |
1244 | } | |
1245 | ||
e3f0e436 CB |
1246 | /* Create intermediate subprocess, two reasons: |
1247 | * 1. We can't setns() in the child itself, since we want to make | |
8ce83369 | 1248 | * sure we are properly attached to the pidns. |
e3f0e436 | 1249 | * 2. Also, the initial thread has to put the attached process |
8ce83369 CB |
1250 | * into the cgroup, which we can only do if we didn't already |
1251 | * setns() (otherwise, user namespaces will hate us). | |
9c4693b8 CS |
1252 | */ |
1253 | pid = fork(); | |
9c4693b8 | 1254 | if (pid < 0) { |
8ce83369 | 1255 | SYSERROR("Failed to create first subprocess."); |
9c4693b8 | 1256 | free(cwd); |
fe4de9a6 | 1257 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1258 | return -1; |
1259 | } | |
1260 | ||
1261 | if (pid) { | |
ba2be1a8 | 1262 | int ret_parent = -1; |
9c4693b8 | 1263 | pid_t to_cleanup_pid = pid; |
ba2be1a8 | 1264 | struct lxc_epoll_descr descr = {0}; |
9c4693b8 | 1265 | |
ba2be1a8 | 1266 | /* close unneeded file descriptors */ |
9c4693b8 CS |
1267 | close(ipc_sockets[1]); |
1268 | free(cwd); | |
ba2be1a8 | 1269 | lxc_proc_close_ns_fd(init_ctx); |
9e84479f CB |
1270 | if (options->attach_flags & LXC_ATTACH_TERMINAL) |
1271 | lxc_attach_terminal_close_slave(&terminal); | |
9c4693b8 | 1272 | |
8ce83369 | 1273 | /* Attach to cgroup, if requested. */ |
f4364484 | 1274 | if (options->attach_flags & LXC_ATTACH_MOVE_TO_CGROUP) { |
2202afc9 CB |
1275 | struct cgroup_ops *cgroup_ops; |
1276 | ||
1277 | cgroup_ops = cgroup_init(NULL); | |
1278 | if (!cgroup_ops) | |
1279 | goto on_error; | |
1280 | ||
1281 | if (!cgroup_ops->attach(cgroup_ops, name, lxcpath, pid)) | |
8ce83369 | 1282 | goto on_error; |
2202afc9 CB |
1283 | |
1284 | cgroup_exit(cgroup_ops); | |
1285 | TRACE("Moved intermediate process %d into container's cgroups", pid); | |
f4364484 SG |
1286 | } |
1287 | ||
bb2ada6f | 1288 | /* Setup /proc limits */ |
1cce35e6 CB |
1289 | if (!lxc_list_empty(&conf->procs)) { |
1290 | ret = setup_proc_filesystem(&conf->procs, pid); | |
bb2ada6f CB |
1291 | if (ret < 0) |
1292 | goto on_error; | |
1293 | } | |
1294 | ||
c6d09e15 | 1295 | /* Setup resource limits */ |
1cce35e6 CB |
1296 | if (!lxc_list_empty(&conf->limits)) { |
1297 | ret = setup_resource_limits(&conf->limits, pid); | |
ba2be1a8 CB |
1298 | if (ret < 0) |
1299 | goto on_error; | |
1300 | } | |
1301 | ||
9e84479f CB |
1302 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
1303 | ret = lxc_attach_terminal_mainloop_init(&terminal, &descr); | |
ba2be1a8 | 1304 | if (ret < 0) |
6f4f1937 | 1305 | goto on_error; |
9e84479f | 1306 | TRACE("Initialized terminal mainloop"); |
ba2be1a8 | 1307 | } |
c6d09e15 | 1308 | |
8ce83369 | 1309 | /* Let the child process know to go ahead. */ |
f4364484 SG |
1310 | status = 0; |
1311 | ret = lxc_write_nointr(ipc_sockets[0], &status, sizeof(status)); | |
94ac256f | 1312 | if (ret != sizeof(status)) |
ba2be1a8 | 1313 | goto close_mainloop; |
94ac256f | 1314 | TRACE("Told intermediate process to start initializing"); |
f4364484 | 1315 | |
8ce83369 | 1316 | /* Get pid of attached process from intermediate process. */ |
94ac256f CB |
1317 | ret = lxc_read_nointr(ipc_sockets[0], &attached_pid, sizeof(attached_pid)); |
1318 | if (ret != sizeof(attached_pid)) | |
ba2be1a8 | 1319 | goto close_mainloop; |
94ac256f | 1320 | TRACE("Received pid %d of attached process in parent pid namespace", attached_pid); |
9c4693b8 | 1321 | |
8ce83369 | 1322 | /* Ignore SIGKILL (CTRL-C) and SIGQUIT (CTRL-\) - issue #313. */ |
62183f1a SH |
1323 | if (options->stdin_fd == 0) { |
1324 | signal(SIGINT, SIG_IGN); | |
1325 | signal(SIGQUIT, SIG_IGN); | |
1326 | } | |
2eef2bda | 1327 | |
8ce83369 | 1328 | /* Reap intermediate process. */ |
9c4693b8 CS |
1329 | ret = wait_for_pid(pid); |
1330 | if (ret < 0) | |
ba2be1a8 | 1331 | goto close_mainloop; |
94ac256f | 1332 | TRACE("Intermediate process %d exited", pid); |
9c4693b8 | 1333 | |
8ce83369 | 1334 | /* We will always have to reap the attached process now. */ |
9c4693b8 CS |
1335 | to_cleanup_pid = attached_pid; |
1336 | ||
81f466d0 | 1337 | /* Open LSM fd and send it to child. */ |
6f4f1937 CB |
1338 | if ((options->namespaces & CLONE_NEWNS) && |
1339 | (options->attach_flags & LXC_ATTACH_LSM) && | |
1340 | init_ctx->lsm_label) { | |
94ac256f | 1341 | int ret = -1; |
47ce2cb7 CB |
1342 | int labelfd; |
1343 | bool on_exec; | |
6f4f1937 | 1344 | |
47ce2cb7 CB |
1345 | on_exec = options->attach_flags & LXC_ATTACH_LSM_EXEC ? true : false; |
1346 | labelfd = lsm_process_label_fd_get(attached_pid, on_exec); | |
81f466d0 | 1347 | if (labelfd < 0) |
ba2be1a8 | 1348 | goto close_mainloop; |
94ac256f | 1349 | TRACE("Opened LSM label file descriptor %d", labelfd); |
81f466d0 CB |
1350 | |
1351 | /* Send child fd of the LSM security module to write to. */ | |
ae467c54 | 1352 | ret = lxc_abstract_unix_send_fds(ipc_sockets[0], &labelfd, 1, NULL, 0); |
1d8e5ca2 | 1353 | close(labelfd); |
81f466d0 | 1354 | if (ret <= 0) { |
94ac256f | 1355 | SYSERROR("%d", (int)ret); |
ba2be1a8 | 1356 | goto close_mainloop; |
81f466d0 | 1357 | } |
94ac256f | 1358 | TRACE("Sent LSM label file descriptor %d to child", labelfd); |
81f466d0 CB |
1359 | } |
1360 | ||
8ce83369 CB |
1361 | /* We're done, the child process should now execute whatever it |
1362 | * is that the user requested. The parent can now track it with | |
1363 | * waitpid() or similar. | |
9c4693b8 CS |
1364 | */ |
1365 | ||
1366 | *attached_process = attached_pid; | |
9c4693b8 | 1367 | |
ba2be1a8 | 1368 | /* Now shut down communication with child, we're done. */ |
9c4693b8 CS |
1369 | shutdown(ipc_sockets[0], SHUT_RDWR); |
1370 | close(ipc_sockets[0]); | |
ba2be1a8 CB |
1371 | ipc_sockets[0] = -1; |
1372 | ||
1373 | ret_parent = 0; | |
1374 | to_cleanup_pid = -1; | |
9e84479f | 1375 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
ba2be1a8 CB |
1376 | ret = lxc_mainloop(&descr, -1); |
1377 | if (ret < 0) { | |
1378 | ret_parent = -1; | |
1379 | to_cleanup_pid = attached_pid; | |
1380 | } | |
1381 | } | |
1382 | ||
1383 | close_mainloop: | |
9e84479f | 1384 | if (options->attach_flags & LXC_ATTACH_TERMINAL) |
ba2be1a8 CB |
1385 | lxc_mainloop_close(&descr); |
1386 | ||
1387 | on_error: | |
1388 | if (ipc_sockets[0] >= 0) { | |
1389 | shutdown(ipc_sockets[0], SHUT_RDWR); | |
1390 | close(ipc_sockets[0]); | |
1391 | } | |
1392 | ||
1393 | if (to_cleanup_pid > 0) | |
6f4f1937 | 1394 | (void)wait_for_pid(to_cleanup_pid); |
ba2be1a8 | 1395 | |
9e84479f CB |
1396 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
1397 | lxc_terminal_delete(&terminal); | |
1398 | lxc_terminal_conf_free(&terminal); | |
ba2be1a8 | 1399 | } |
fe4de9a6 | 1400 | lxc_proc_put_context_info(init_ctx); |
ba2be1a8 | 1401 | return ret_parent; |
9c4693b8 CS |
1402 | } |
1403 | ||
ba2be1a8 | 1404 | /* close unneeded file descriptors */ |
9c4693b8 | 1405 | close(ipc_sockets[0]); |
ba2be1a8 | 1406 | ipc_sockets[0] = -EBADF; |
9e84479f CB |
1407 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
1408 | lxc_attach_terminal_close_master(&terminal); | |
1409 | lxc_attach_terminal_close_peer(&terminal); | |
1410 | lxc_attach_terminal_close_log(&terminal); | |
ba2be1a8 | 1411 | } |
9c4693b8 | 1412 | |
8ce83369 | 1413 | /* Wait for the parent to have setup cgroups. */ |
94ac256f | 1414 | ret = lxc_read_nointr(ipc_sockets[1], &status, sizeof(status)); |
ba2be1a8 | 1415 | if (ret != sizeof(status)) { |
f4364484 | 1416 | shutdown(ipc_sockets[1], SHUT_RDWR); |
62de1db6 | 1417 | lxc_proc_put_context_info(init_ctx); |
f4364484 SG |
1418 | rexit(-1); |
1419 | } | |
94ac256f | 1420 | TRACE("Intermediate process starting to initialize"); |
f4364484 | 1421 | |
8ce83369 CB |
1422 | /* Attach now, create another subprocess later, since pid namespaces |
1423 | * only really affect the children of the current process. | |
9c4693b8 | 1424 | */ |
877f3a04 | 1425 | ret = lxc_attach_to_ns(init_pid, init_ctx); |
9c4693b8 | 1426 | if (ret < 0) { |
94ac256f | 1427 | ERROR("Failed to enter namespaces"); |
9c4693b8 | 1428 | shutdown(ipc_sockets[1], SHUT_RDWR); |
62de1db6 | 1429 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1430 | rexit(-1); |
1431 | } | |
877f3a04 CB |
1432 | /* close namespace file descriptors */ |
1433 | lxc_proc_close_ns_fd(init_ctx); | |
9c4693b8 | 1434 | |
8ce83369 | 1435 | /* Attach succeeded, try to cwd. */ |
9c4693b8 CS |
1436 | if (options->initial_cwd) |
1437 | new_cwd = options->initial_cwd; | |
1438 | else | |
1439 | new_cwd = cwd; | |
d6d979bc CB |
1440 | if (new_cwd) { |
1441 | ret = chdir(new_cwd); | |
1442 | if (ret < 0) | |
1443 | WARN("Could not change directory to \"%s\"", new_cwd); | |
1444 | } | |
9c4693b8 CS |
1445 | free(cwd); |
1446 | ||
a998454a CB |
1447 | /* Create attached process. */ |
1448 | payload.ipc_socket = ipc_sockets[1]; | |
1449 | payload.options = options; | |
1450 | payload.init_ctx = init_ctx; | |
9e84479f | 1451 | payload.terminal_slave_fd = terminal.slave; |
a998454a CB |
1452 | payload.exec_function = exec_function; |
1453 | payload.exec_payload = exec_payload; | |
9c4693b8 | 1454 | |
a998454a CB |
1455 | pid = lxc_raw_clone(CLONE_PARENT); |
1456 | if (pid < 0) { | |
94ac256f | 1457 | SYSERROR("Failed to clone attached process"); |
9c4693b8 | 1458 | shutdown(ipc_sockets[1], SHUT_RDWR); |
62de1db6 | 1459 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1460 | rexit(-1); |
1461 | } | |
a998454a CB |
1462 | |
1463 | if (pid == 0) { | |
1464 | ret = attach_child_main(&payload); | |
1465 | if (ret < 0) | |
1466 | ERROR("Failed to exec"); | |
1467 | _exit(EXIT_FAILURE); | |
1468 | } | |
9e84479f CB |
1469 | if (options->attach_flags & LXC_ATTACH_TERMINAL) |
1470 | lxc_attach_terminal_close_slave(&terminal); | |
9c4693b8 | 1471 | |
8ce83369 | 1472 | /* Tell grandparent the pid of the pid of the newly created child. */ |
9c4693b8 CS |
1473 | ret = lxc_write_nointr(ipc_sockets[1], &pid, sizeof(pid)); |
1474 | if (ret != sizeof(pid)) { | |
8ce83369 CB |
1475 | /* If this really happens here, this is very unfortunate, since |
1476 | * the parent will not know the pid of the attached process and | |
1477 | * will not be able to wait for it (and we won't either due to | |
1478 | * CLONE_PARENT) so the parent won't be able to reap it and the | |
1479 | * attached process will remain a zombie. | |
9c4693b8 | 1480 | */ |
9c4693b8 | 1481 | shutdown(ipc_sockets[1], SHUT_RDWR); |
62de1db6 | 1482 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1483 | rexit(-1); |
1484 | } | |
94ac256f | 1485 | TRACE("Sending pid %d of attached process", pid); |
9c4693b8 | 1486 | |
8ce83369 | 1487 | /* The rest is in the hands of the initial and the attached process. */ |
62de1db6 | 1488 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1489 | rexit(0); |
1490 | } | |
1491 | ||
9c4693b8 CS |
1492 | int lxc_attach_run_command(void* payload) |
1493 | { | |
1494 | lxc_attach_command_t* cmd = (lxc_attach_command_t*)payload; | |
1495 | ||
1496 | execvp(cmd->program, cmd->argv); | |
8ce83369 | 1497 | SYSERROR("Failed to exec \"%s\".", cmd->program); |
9c4693b8 CS |
1498 | return -1; |
1499 | } | |
1500 | ||
1501 | int lxc_attach_run_shell(void* payload) | |
1502 | { | |
1503 | uid_t uid; | |
1504 | struct passwd *passwd; | |
1505 | char *user_shell; | |
1506 | ||
8ce83369 | 1507 | /* Ignore payload parameter. */ |
9c4693b8 CS |
1508 | (void)payload; |
1509 | ||
1510 | uid = getuid(); | |
1511 | passwd = getpwuid(uid); | |
1512 | ||
8ce83369 CB |
1513 | /* This probably happens because of incompatible nss implementations in |
1514 | * host and container (remember, this code is still using the host's | |
1515 | * glibc but our mount namespace is in the container) we may try to get | |
1516 | * the information by spawning a [getent passwd uid] process and parsing | |
1517 | * the result. | |
9c4693b8 CS |
1518 | */ |
1519 | if (!passwd) | |
1520 | user_shell = lxc_attach_getpwshell(uid); | |
1521 | else | |
1522 | user_shell = passwd->pw_shell; | |
9c4693b8 | 1523 | if (user_shell) |
acf47e1b | 1524 | execlp(user_shell, user_shell, (char *)NULL); |
9c4693b8 | 1525 | |
8ce83369 CB |
1526 | /* Executed if either no passwd entry or execvp fails, we will fall back |
1527 | * on /bin/sh as a default shell. | |
9c4693b8 | 1528 | */ |
acf47e1b | 1529 | execlp("/bin/sh", "/bin/sh", (char *)NULL); |
edeb1836 CB |
1530 | SYSERROR("Failed to execute shell"); |
1531 | if (!passwd) | |
1532 | free(user_shell); | |
9c4693b8 CS |
1533 | return -1; |
1534 | } |