]>
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> | |
e5adb2b5 | 27 | #include <termios.h> |
c476bdce | 28 | #include <grp.h> |
6f4f1937 CB |
29 | #include <pwd.h> |
30 | #include <signal.h> | |
31 | #include <stdio.h> | |
32 | #include <stdlib.h> | |
33 | #include <string.h> | |
34 | #include <unistd.h> | |
35 | #include <linux/unistd.h> | |
36 | #include <sys/mount.h> | |
e0732705 CS |
37 | #include <sys/param.h> |
38 | #include <sys/prctl.h> | |
5ec27989 | 39 | #include <sys/socket.h> |
1ba0013f | 40 | #include <sys/syscall.h> |
905022f7 | 41 | #include <sys/wait.h> |
6f4f1937 CB |
42 | |
43 | #include <lxc/lxccontainer.h> | |
e0732705 | 44 | |
955e2a02 | 45 | #ifndef HAVE_DECL_PR_CAPBSET_DROP |
e0732705 CS |
46 | #define PR_CAPBSET_DROP 24 |
47 | #endif | |
48 | ||
955e2a02 CB |
49 | #ifndef HAVE_DECL_PR_SET_NO_NEW_PRIVS |
50 | #define PR_SET_NO_NEW_PRIVS 38 | |
51 | #endif | |
52 | ||
53 | #ifndef HAVE_DECL_PR_GET_NO_NEW_PRIVS | |
54 | #define PR_GET_NO_NEW_PRIVS 39 | |
55 | #endif | |
56 | ||
81f466d0 | 57 | #include "af_unix.h" |
e0732705 CS |
58 | #include "attach.h" |
59 | #include "caps.h" | |
9c4693b8 | 60 | #include "cgroup.h" |
6f4f1937 | 61 | #include "commands.h" |
2c4ea790 | 62 | #include "conf.h" |
6f4f1937 | 63 | #include "config.h" |
9b8e3c96 | 64 | #include "confile.h" |
6f4f1937 CB |
65 | #include "log.h" |
66 | #include "lsm/lsm.h" | |
67 | #include "lxclock.h" | |
68 | #include "lxcseccomp.h" | |
ba2be1a8 | 69 | #include "mainloop.h" |
6f4f1937 | 70 | #include "namespace.h" |
0ed9b1bc | 71 | #include "terminal.h" |
6f4f1937 | 72 | #include "utils.h" |
9c4693b8 CS |
73 | |
74 | #if HAVE_SYS_PERSONALITY_H | |
75 | #include <sys/personality.h> | |
76 | #endif | |
e0732705 | 77 | |
a3da2f3b | 78 | #ifndef SOCK_CLOEXEC |
6f4f1937 | 79 | #define SOCK_CLOEXEC 02000000 |
a3da2f3b SG |
80 | #endif |
81 | ||
d6a3c917 SG |
82 | #ifndef MS_REC |
83 | #define MS_REC 16384 | |
84 | #endif | |
85 | ||
86 | #ifndef MS_SLAVE | |
6f4f1937 | 87 | #define MS_SLAVE (1 << 19) |
d6a3c917 SG |
88 | #endif |
89 | ||
ac2cecc4 | 90 | lxc_log_define(attach, lxc); |
e0732705 | 91 | |
8ce83369 | 92 | /* /proc/pid-to-str/status\0 = (5 + 21 + 7 + 1) */ |
eab15c1e | 93 | #define __PROC_STATUS_LEN (5 + (LXC_NUMSTRLEN64) + 7 + 1) |
74a3920a | 94 | static struct lxc_proc_context_info *lxc_proc_get_context_info(pid_t pid) |
e0732705 | 95 | { |
6f4f1937 CB |
96 | int ret; |
97 | bool found; | |
e0732705 | 98 | FILE *proc_file; |
8ce83369 | 99 | char proc_fn[__PROC_STATUS_LEN]; |
e0732705 | 100 | size_t line_bufsz = 0; |
6f4f1937 | 101 | char *line = NULL; |
8ce83369 | 102 | struct lxc_proc_context_info *info = NULL; |
e0732705 | 103 | |
8ce83369 CB |
104 | /* Read capabilities. */ |
105 | ret = snprintf(proc_fn, __PROC_STATUS_LEN, "/proc/%d/status", pid); | |
106 | if (ret < 0 || ret >= __PROC_STATUS_LEN) | |
107 | goto on_error; | |
e0732705 CS |
108 | |
109 | proc_file = fopen(proc_fn, "r"); | |
110 | if (!proc_file) { | |
ea918412 | 111 | SYSERROR("Could not open %s", proc_fn); |
8ce83369 | 112 | goto on_error; |
e0732705 CS |
113 | } |
114 | ||
8ce83369 CB |
115 | info = calloc(1, sizeof(*info)); |
116 | if (!info) { | |
ea918412 | 117 | SYSERROR("Could not allocate memory"); |
17ac5301 | 118 | fclose(proc_file); |
8ce83369 CB |
119 | return NULL; |
120 | } | |
121 | ||
122 | found = false; | |
ea918412 | 123 | |
e0732705 CS |
124 | while (getline(&line, &line_bufsz, proc_file) != -1) { |
125 | ret = sscanf(line, "CapBnd: %llx", &info->capability_mask); | |
8ce83369 CB |
126 | if (ret != EOF && ret == 1) { |
127 | found = true; | |
e0732705 CS |
128 | break; |
129 | } | |
130 | } | |
131 | ||
f10fad2f | 132 | free(line); |
e0732705 CS |
133 | fclose(proc_file); |
134 | ||
135 | if (!found) { | |
ea918412 | 136 | ERROR("Could not read capability bounding set from %s", |
137 | proc_fn); | |
8ce83369 | 138 | goto on_error; |
e0732705 CS |
139 | } |
140 | ||
fe4de9a6 | 141 | info->lsm_label = lsm_process_label_get(pid); |
877f3a04 CB |
142 | info->ns_inherited = 0; |
143 | memset(info->ns_fd, -1, sizeof(int) * LXC_NS_MAX); | |
e0732705 | 144 | |
e0732705 CS |
145 | return info; |
146 | ||
8ce83369 | 147 | on_error: |
460a1cf0 | 148 | free(info); |
e0732705 CS |
149 | return NULL; |
150 | } | |
151 | ||
877f3a04 CB |
152 | static inline void lxc_proc_close_ns_fd(struct lxc_proc_context_info *ctx) |
153 | { | |
154 | int i; | |
155 | ||
156 | for (i = 0; i < LXC_NS_MAX; i++) { | |
157 | if (ctx->ns_fd[i] < 0) | |
158 | continue; | |
ea918412 | 159 | |
877f3a04 CB |
160 | close(ctx->ns_fd[i]); |
161 | ctx->ns_fd[i] = -EBADF; | |
162 | } | |
163 | } | |
164 | ||
fe4de9a6 DE |
165 | static void lxc_proc_put_context_info(struct lxc_proc_context_info *ctx) |
166 | { | |
f10fad2f | 167 | free(ctx->lsm_label); |
08ea9270 CB |
168 | ctx->lsm_label = NULL; |
169 | ||
170 | if (ctx->container) { | |
2c4ea790 | 171 | lxc_container_put(ctx->container); |
08ea9270 CB |
172 | ctx->container = NULL; |
173 | } | |
174 | ||
877f3a04 | 175 | lxc_proc_close_ns_fd(ctx); |
fe4de9a6 DE |
176 | free(ctx); |
177 | } | |
178 | ||
299d1198 CB |
179 | /** |
180 | * in_same_namespace - Check whether two processes are in the same namespace. | |
181 | * @pid1 - PID of the first process. | |
182 | * @pid2 - PID of the second process. | |
183 | * @ns - Name of the namespace to check. Must correspond to one of the names | |
184 | * for the namespaces as shown in /proc/<pid/ns/ | |
185 | * | |
186 | * If the two processes are not in the same namespace returns an fd to the | |
187 | * namespace of the second process identified by @pid2. If the two processes are | |
188 | * in the same namespace returns -EINVAL, -1 if an error occurred. | |
189 | */ | |
190 | static int in_same_namespace(pid_t pid1, pid_t pid2, const char *ns) | |
191 | { | |
192 | int ns_fd1 = -1, ns_fd2 = -1, ret = -1; | |
21d0acc2 | 193 | int saved_errno; |
299d1198 CB |
194 | struct stat ns_st1, ns_st2; |
195 | ||
196 | ns_fd1 = lxc_preserve_ns(pid1, ns); | |
134284c3 CB |
197 | if (ns_fd1 < 0) { |
198 | /* The kernel does not support this namespace. This is not an | |
199 | * error. | |
200 | */ | |
201 | if (errno == ENOENT) | |
202 | return -EINVAL; | |
203 | ||
299d1198 | 204 | goto out; |
134284c3 | 205 | } |
299d1198 CB |
206 | |
207 | ns_fd2 = lxc_preserve_ns(pid2, ns); | |
21d0acc2 | 208 | if (ns_fd2 < 0) |
299d1198 CB |
209 | goto out; |
210 | ||
211 | ret = fstat(ns_fd1, &ns_st1); | |
21d0acc2 | 212 | if (ret < 0) |
299d1198 CB |
213 | goto out; |
214 | ||
215 | ret = fstat(ns_fd2, &ns_st2); | |
21d0acc2 | 216 | if (ret < 0) |
299d1198 CB |
217 | goto out; |
218 | ||
219 | /* processes are in the same namespace */ | |
ea918412 | 220 | if ((ns_st1.st_dev == ns_st2.st_dev ) && (ns_st1.st_ino == ns_st2.st_ino)) { |
221 | ret = -EINVAL; | |
299d1198 | 222 | goto out; |
ea918412 | 223 | } |
299d1198 CB |
224 | |
225 | /* processes are in different namespaces */ | |
226 | ret = ns_fd2; | |
227 | ns_fd2 = -1; | |
228 | ||
229 | out: | |
21d0acc2 | 230 | saved_errno = errno; |
231 | ||
299d1198 CB |
232 | if (ns_fd1 >= 0) |
233 | close(ns_fd1); | |
ea918412 | 234 | |
299d1198 CB |
235 | if (ns_fd2 >= 0) |
236 | close(ns_fd2); | |
237 | ||
ea918412 | 238 | errno = saved_errno; |
299d1198 CB |
239 | return ret; |
240 | } | |
241 | ||
877f3a04 | 242 | static int lxc_attach_to_ns(pid_t pid, struct lxc_proc_context_info *ctx) |
99d50954 | 243 | { |
21d0acc2 | 244 | int i, ret; |
99d50954 | 245 | |
26818618 | 246 | for (i = 0; i < LXC_NS_MAX; i++) { |
877f3a04 | 247 | if (ctx->ns_fd[i] < 0) |
26818618 CB |
248 | continue; |
249 | ||
21d0acc2 | 250 | ret = setns(ctx->ns_fd[i], ns_info[i].clone_flag); |
251 | if (ret < 0) { | |
299d1198 | 252 | SYSERROR("Failed to attach to %s namespace of %d", |
ea918412 | 253 | ns_info[i].proc_name, pid); |
99d50954 CS |
254 | return -1; |
255 | } | |
256 | ||
299d1198 | 257 | DEBUG("Attached to %s namespace of %d", ns_info[i].proc_name, pid); |
99d50954 CS |
258 | } |
259 | ||
260 | return 0; | |
261 | } | |
262 | ||
74a3920a | 263 | static int lxc_attach_remount_sys_proc(void) |
7a0b0b56 CS |
264 | { |
265 | int ret; | |
266 | ||
267 | ret = unshare(CLONE_NEWNS); | |
268 | if (ret < 0) { | |
ea918412 | 269 | SYSERROR("Failed to unshare mount namespace"); |
7a0b0b56 CS |
270 | return -1; |
271 | } | |
272 | ||
2c6f3fc9 | 273 | if (detect_shared_rootfs()) { |
6f4f1937 | 274 | if (mount(NULL, "/", NULL, MS_SLAVE | MS_REC, NULL)) { |
ea918412 | 275 | SYSERROR("Failed to make / rslave"); |
2c6f3fc9 SH |
276 | ERROR("Continuing..."); |
277 | } | |
278 | } | |
279 | ||
8ce83369 | 280 | /* Assume /proc is always mounted, so remount it. */ |
7a0b0b56 CS |
281 | ret = umount2("/proc", MNT_DETACH); |
282 | if (ret < 0) { | |
ea918412 | 283 | SYSERROR("Failed to unmount /proc"); |
7a0b0b56 CS |
284 | return -1; |
285 | } | |
286 | ||
287 | ret = mount("none", "/proc", "proc", 0, NULL); | |
288 | if (ret < 0) { | |
ea918412 | 289 | SYSERROR("Failed to remount /proc"); |
7a0b0b56 CS |
290 | return -1; |
291 | } | |
292 | ||
8ce83369 CB |
293 | /* Try to umount /sys. If it's not a mount point, we'll get EINVAL, then |
294 | * we ignore it because it may not have been mounted in the first place. | |
7a0b0b56 CS |
295 | */ |
296 | ret = umount2("/sys", MNT_DETACH); | |
297 | if (ret < 0 && errno != EINVAL) { | |
ea918412 | 298 | SYSERROR("Failed to unmount /sys"); |
7a0b0b56 CS |
299 | return -1; |
300 | } else if (ret == 0) { | |
8ce83369 | 301 | /* Remount it. */ |
7a0b0b56 CS |
302 | ret = mount("none", "/sys", "sysfs", 0, NULL); |
303 | if (ret < 0) { | |
ea918412 | 304 | SYSERROR("Failed to remount /sys"); |
7a0b0b56 CS |
305 | return -1; |
306 | } | |
307 | } | |
308 | ||
309 | return 0; | |
310 | } | |
311 | ||
74a3920a | 312 | static int lxc_attach_drop_privs(struct lxc_proc_context_info *ctx) |
e0732705 | 313 | { |
6f4f1937 | 314 | int cap, last_cap; |
e0732705 | 315 | |
6f4f1937 | 316 | last_cap = lxc_caps_last_cap(); |
e0732705 CS |
317 | for (cap = 0; cap <= last_cap; cap++) { |
318 | if (ctx->capability_mask & (1LL << cap)) | |
319 | continue; | |
320 | ||
321 | if (prctl(PR_CAPBSET_DROP, cap, 0, 0, 0)) { | |
94ac256f | 322 | SYSERROR("Failed to drop capability %d", cap); |
e0732705 CS |
323 | return -1; |
324 | } | |
ea918412 | 325 | |
94ac256f | 326 | TRACE("Dropped capability %d", cap); |
e0732705 CS |
327 | } |
328 | ||
329 | return 0; | |
330 | } | |
905022f7 | 331 | |
7385273f | 332 | static int lxc_attach_set_environment(struct lxc_proc_context_info *init_ctx, |
333 | enum lxc_attach_env_policy_t policy, | |
6f4f1937 | 334 | char **extra_env, char **extra_keep) |
b3a39ba6 | 335 | { |
3d55242a | 336 | int ret; |
7385273f | 337 | struct lxc_list *iterator; |
338 | ||
799f96fd | 339 | if (policy == LXC_ATTACH_CLEAR_ENV) { |
3d5e9f48 | 340 | int path_kept = 0; |
6f4f1937 | 341 | char **extra_keep_store = NULL; |
3d5e9f48 CS |
342 | |
343 | if (extra_keep) { | |
344 | size_t count, i; | |
345 | ||
3d55242a CB |
346 | for (count = 0; extra_keep[count]; count++) |
347 | ; | |
3d5e9f48 CS |
348 | |
349 | extra_keep_store = calloc(count, sizeof(char *)); | |
3d55242a | 350 | if (!extra_keep_store) |
3d5e9f48 | 351 | return -1; |
3d55242a | 352 | |
3d5e9f48 CS |
353 | for (i = 0; i < count; i++) { |
354 | char *v = getenv(extra_keep[i]); | |
355 | if (v) { | |
356 | extra_keep_store[i] = strdup(v); | |
357 | if (!extra_keep_store[i]) { | |
3d5e9f48 CS |
358 | while (i > 0) |
359 | free(extra_keep_store[--i]); | |
ea918412 | 360 | |
3d5e9f48 CS |
361 | free(extra_keep_store); |
362 | return -1; | |
363 | } | |
3d55242a | 364 | |
3d5e9f48 CS |
365 | if (strcmp(extra_keep[i], "PATH") == 0) |
366 | path_kept = 1; | |
367 | } | |
3d5e9f48 CS |
368 | } |
369 | } | |
370 | ||
799f96fd | 371 | if (clearenv()) { |
a9cab7e3 | 372 | if (extra_keep_store) { |
3d55242a CB |
373 | char **p; |
374 | ||
a9cab7e3 CS |
375 | for (p = extra_keep_store; *p; p++) |
376 | free(*p); | |
3d55242a | 377 | |
a9cab7e3 CS |
378 | free(extra_keep_store); |
379 | } | |
3d55242a | 380 | |
ea918412 | 381 | ERROR("Failed to clear environment"); |
3d5e9f48 CS |
382 | return -1; |
383 | } | |
384 | ||
385 | if (extra_keep_store) { | |
386 | size_t i; | |
6f4f1937 | 387 | |
3d5e9f48 | 388 | for (i = 0; extra_keep[i]; i++) { |
acd4922e | 389 | if (extra_keep_store[i]) { |
3d55242a CB |
390 | ret = setenv(extra_keep[i], extra_keep_store[i], 1); |
391 | if (ret < 0) | |
a24c5678 | 392 | SYSWARN("Failed to set environment variable"); |
acd4922e | 393 | } |
ea918412 | 394 | |
3d5e9f48 CS |
395 | free(extra_keep_store[i]); |
396 | } | |
ea918412 | 397 | |
3d5e9f48 CS |
398 | free(extra_keep_store); |
399 | } | |
400 | ||
8ce83369 CB |
401 | /* Always set a default path; shells and execlp tend to be fine |
402 | * without it, but there is a disturbing number of C programs | |
403 | * out there that just assume that getenv("PATH") is never NULL | |
404 | * and then die a painful segfault death. | |
405 | */ | |
3d55242a CB |
406 | if (!path_kept) { |
407 | ret = setenv("PATH", "/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", 1); | |
408 | if (ret < 0) | |
a24c5678 | 409 | SYSWARN("Failed to set environment variable"); |
3d55242a | 410 | } |
b3a39ba6 DW |
411 | } |
412 | ||
3d55242a CB |
413 | ret = putenv("container=lxc"); |
414 | if (ret < 0) { | |
a24c5678 | 415 | SYSWARN("Failed to set environment variable"); |
b3a39ba6 DW |
416 | return -1; |
417 | } | |
418 | ||
7385273f | 419 | /* Set container environment variables.*/ |
420 | if (init_ctx && init_ctx->container && init_ctx->container->lxc_conf) { | |
421 | lxc_list_for_each(iterator, &init_ctx->container->lxc_conf->environment) { | |
3d55242a CB |
422 | char *env_tmp; |
423 | ||
424 | env_tmp = strdup((char *)iterator->elem); | |
425 | if (!env_tmp) | |
7385273f | 426 | return -1; |
7385273f | 427 | |
3d55242a CB |
428 | ret = putenv(env_tmp); |
429 | if (ret < 0) { | |
430 | SYSERROR("Failed to set environment variable: %s", (char *)iterator->elem); | |
7385273f | 431 | return -1; |
432 | } | |
433 | } | |
434 | } | |
435 | ||
8ce83369 | 436 | /* Set extra environment variables. */ |
3d5e9f48 CS |
437 | if (extra_env) { |
438 | for (; *extra_env; extra_env++) { | |
3d55242a | 439 | char *p; |
ea918412 | 440 | |
8ce83369 CB |
441 | /* We just assume the user knows what they are doing, so |
442 | * we don't do any checks. | |
443 | */ | |
3d55242a CB |
444 | p = strdup(*extra_env); |
445 | if (!p) | |
3d5e9f48 | 446 | return -1; |
3d55242a CB |
447 | |
448 | ret = putenv(p); | |
449 | if (ret < 0) | |
a24c5678 | 450 | SYSWARN("Failed to set environment variable"); |
3d5e9f48 CS |
451 | } |
452 | } | |
453 | ||
b3a39ba6 DW |
454 | return 0; |
455 | } | |
456 | ||
74a3920a | 457 | static char *lxc_attach_getpwshell(uid_t uid) |
905022f7 | 458 | { |
6f4f1937 | 459 | int fd, ret; |
905022f7 CS |
460 | pid_t pid; |
461 | int pipes[2]; | |
3fa23ac3 CB |
462 | FILE *pipe_f; |
463 | bool found = false; | |
464 | size_t line_bufsz = 0; | |
465 | char *line = NULL, *result = NULL; | |
905022f7 | 466 | |
8ce83369 CB |
467 | /* We need to fork off a process that runs the getent program, and we |
468 | * need to capture its output, so we use a pipe for that purpose. | |
905022f7 | 469 | */ |
3fa23ac3 | 470 | ret = pipe2(pipes, O_CLOEXEC); |
905022f7 CS |
471 | if (ret < 0) |
472 | return NULL; | |
473 | ||
474 | pid = fork(); | |
475 | if (pid < 0) { | |
476 | close(pipes[0]); | |
477 | close(pipes[1]); | |
478 | return NULL; | |
479 | } | |
480 | ||
3fa23ac3 | 481 | if (!pid) { |
905022f7 CS |
482 | char uid_buf[32]; |
483 | char *arguments[] = { | |
484 | "getent", | |
485 | "passwd", | |
486 | uid_buf, | |
487 | NULL | |
488 | }; | |
489 | ||
490 | close(pipes[0]); | |
491 | ||
8ce83369 | 492 | /* We want to capture stdout. */ |
3fa23ac3 | 493 | ret = dup2(pipes[1], STDOUT_FILENO); |
905022f7 | 494 | close(pipes[1]); |
3fa23ac3 | 495 | if (ret < 0) |
ea918412 | 496 | _exit(EXIT_FAILURE); |
905022f7 | 497 | |
8ce83369 CB |
498 | /* Get rid of stdin/stderr, so we try to associate it with |
499 | * /dev/null. | |
905022f7 | 500 | */ |
3fa23ac3 | 501 | fd = open_devnull(); |
905022f7 | 502 | if (fd < 0) { |
3fa23ac3 CB |
503 | close(STDIN_FILENO); |
504 | close(STDERR_FILENO); | |
905022f7 | 505 | } else { |
3fa23ac3 CB |
506 | (void)dup3(fd, STDIN_FILENO, O_CLOEXEC); |
507 | (void)dup3(fd, STDOUT_FILENO, O_CLOEXEC); | |
905022f7 CS |
508 | close(fd); |
509 | } | |
510 | ||
8ce83369 | 511 | /* Finish argument list. */ |
3fa23ac3 CB |
512 | ret = snprintf(uid_buf, sizeof(uid_buf), "%ld", (long)uid); |
513 | if (ret <= 0 || ret >= sizeof(uid_buf)) | |
ea918412 | 514 | _exit(EXIT_FAILURE); |
905022f7 | 515 | |
8ce83369 | 516 | /* Try to run getent program. */ |
3fa23ac3 | 517 | (void)execvp("getent", arguments); |
ea918412 | 518 | _exit(EXIT_FAILURE); |
905022f7 | 519 | } |
3fa23ac3 CB |
520 | |
521 | close(pipes[1]); | |
522 | ||
523 | pipe_f = fdopen(pipes[0], "r"); | |
524 | while (getline(&line, &line_bufsz, pipe_f) != -1) { | |
525 | int i; | |
526 | long value; | |
527 | char *token; | |
528 | char *endptr = NULL, *saveptr = NULL; | |
529 | ||
530 | /* If we already found something, just continue to read | |
531 | * until the pipe doesn't deliver any more data, but | |
532 | * don't modify the existing data structure. | |
533 | */ | |
534 | if (found) | |
535 | continue; | |
536 | ||
18d4ffde | 537 | if (!line) |
538 | continue; | |
539 | ||
3fa23ac3 CB |
540 | /* Trim line on the right hand side. */ |
541 | for (i = strlen(line); i > 0 && (line[i - 1] == '\n' || line[i - 1] == '\r'); --i) | |
542 | line[i - 1] = '\0'; | |
543 | ||
544 | /* Split into tokens: first: user name. */ | |
545 | token = strtok_r(line, ":", &saveptr); | |
546 | if (!token) | |
547 | continue; | |
548 | ||
549 | /* next: dummy password field */ | |
550 | token = strtok_r(NULL, ":", &saveptr); | |
551 | if (!token) | |
552 | continue; | |
553 | ||
554 | /* next: user id */ | |
555 | token = strtok_r(NULL, ":", &saveptr); | |
556 | value = token ? strtol(token, &endptr, 10) : 0; | |
557 | if (!token || !endptr || *endptr || value == LONG_MIN || | |
ea918412 | 558 | value == LONG_MAX) |
3fa23ac3 CB |
559 | continue; |
560 | ||
561 | /* dummy sanity check: user id matches */ | |
562 | if ((uid_t)value != uid) | |
563 | continue; | |
564 | ||
565 | /* skip fields: gid, gecos, dir, go to next field 'shell' */ | |
566 | for (i = 0; i < 4; i++) { | |
567 | token = strtok_r(NULL, ":", &saveptr); | |
568 | if (!token) | |
569 | continue; | |
570 | } | |
ea918412 | 571 | |
3fa23ac3 CB |
572 | if (!token) |
573 | continue; | |
ea918412 | 574 | |
3fa23ac3 CB |
575 | free(result); |
576 | result = strdup(token); | |
577 | ||
578 | /* Sanity check that there are no fields after that. */ | |
579 | token = strtok_r(NULL, ":", &saveptr); | |
580 | if (token) | |
581 | continue; | |
582 | ||
583 | found = true; | |
584 | } | |
ea918412 | 585 | |
3fa23ac3 CB |
586 | free(line); |
587 | fclose(pipe_f); | |
588 | ||
589 | ret = wait_for_pid(pid); | |
590 | if (ret < 0) { | |
591 | free(result); | |
592 | return NULL; | |
593 | } | |
594 | ||
595 | if (!found) { | |
596 | free(result); | |
597 | return NULL; | |
598 | } | |
599 | ||
600 | return result; | |
905022f7 | 601 | } |
cb3e61fa | 602 | |
6f4f1937 | 603 | static void lxc_attach_get_init_uidgid(uid_t *init_uid, gid_t *init_gid) |
cb3e61fa CS |
604 | { |
605 | FILE *proc_file; | |
8ce83369 CB |
606 | char proc_fn[__PROC_STATUS_LEN]; |
607 | int ret; | |
cb3e61fa CS |
608 | char *line = NULL; |
609 | size_t line_bufsz = 0; | |
cb3e61fa CS |
610 | long value = -1; |
611 | uid_t uid = (uid_t)-1; | |
612 | gid_t gid = (gid_t)-1; | |
613 | ||
7fb45c93 CB |
614 | ret = snprintf(proc_fn, __PROC_STATUS_LEN, "/proc/%d/status", 1); |
615 | if (ret < 0 || ret >= __PROC_STATUS_LEN) | |
616 | return; | |
cb3e61fa CS |
617 | |
618 | proc_file = fopen(proc_fn, "r"); | |
619 | if (!proc_file) | |
620 | return; | |
621 | ||
622 | while (getline(&line, &line_bufsz, proc_file) != -1) { | |
8ce83369 CB |
623 | /* Format is: real, effective, saved set user, fs we only care |
624 | * about real uid. | |
cb3e61fa CS |
625 | */ |
626 | ret = sscanf(line, "Uid: %ld", &value); | |
8ce83369 | 627 | if (ret != EOF && ret == 1) { |
6f4f1937 | 628 | uid = (uid_t)value; |
cb3e61fa CS |
629 | } else { |
630 | ret = sscanf(line, "Gid: %ld", &value); | |
8ce83369 | 631 | if (ret != EOF && ret == 1) |
6f4f1937 | 632 | gid = (gid_t)value; |
cb3e61fa | 633 | } |
ea918412 | 634 | |
cb3e61fa CS |
635 | if (uid != (uid_t)-1 && gid != (gid_t)-1) |
636 | break; | |
637 | } | |
638 | ||
639 | fclose(proc_file); | |
640 | free(line); | |
641 | ||
8ce83369 | 642 | /* Only override arguments if we found something. */ |
cb3e61fa CS |
643 | if (uid != (uid_t)-1) |
644 | *init_uid = uid; | |
ea918412 | 645 | |
cb3e61fa CS |
646 | if (gid != (gid_t)-1) |
647 | *init_gid = gid; | |
648 | ||
649 | /* TODO: we should also parse supplementary groups and use | |
8ce83369 CB |
650 | * setgroups() to set them. |
651 | */ | |
cb3e61fa | 652 | } |
9c4693b8 | 653 | |
8ce83369 | 654 | /* Help the optimizer along if it doesn't know that exit always exits. */ |
6f4f1937 CB |
655 | #define rexit(c) \ |
656 | do { \ | |
657 | int __c = (c); \ | |
658 | _exit(__c); \ | |
659 | return __c; \ | |
660 | } while (0) | |
9c4693b8 | 661 | |
8ce83369 | 662 | /* Define default options if no options are supplied by the user. */ |
9c4693b8 CS |
663 | static lxc_attach_options_t attach_static_default_options = LXC_ATTACH_OPTIONS_DEFAULT; |
664 | ||
d4db3d14 | 665 | static bool fetch_seccomp(struct lxc_container *c, lxc_attach_options_t *options) |
2c4ea790 | 666 | { |
d4db3d14 CB |
667 | int ret; |
668 | bool bret; | |
bd7b4e28 | 669 | char *path; |
2eef2bda | 670 | |
6f4f1937 CB |
671 | if (!(options->namespaces & CLONE_NEWNS) || |
672 | !(options->attach_flags & LXC_ATTACH_LSM)) { | |
bd4307f0 CB |
673 | free(c->lxc_conf->seccomp); |
674 | c->lxc_conf->seccomp = NULL; | |
2c4ea790 | 675 | return true; |
bd4307f0 | 676 | } |
bd7b4e28 | 677 | |
2e812c16 | 678 | /* Remove current setting. */ |
d4db3d14 | 679 | if (!c->set_config_item(c, "lxc.seccomp.profile", "") && |
ea918412 | 680 | !c->set_config_item(c, "lxc.seccomp", "")) |
2c4ea790 | 681 | return false; |
bd7b4e28 | 682 | |
8ce83369 | 683 | /* Fetch the current profile path over the cmd interface. */ |
0b427da0 | 684 | path = c->get_running_config_item(c, "lxc.seccomp.profile"); |
bd7b4e28 | 685 | if (!path) { |
d4db3d14 | 686 | INFO("Failed to retrieve lxc.seccomp.profile"); |
ea918412 | 687 | |
0b427da0 | 688 | path = c->get_running_config_item(c, "lxc.seccomp"); |
d4db3d14 CB |
689 | if (!path) { |
690 | INFO("Failed to retrieve lxc.seccomp"); | |
691 | return true; | |
692 | } | |
bd7b4e28 SG |
693 | } |
694 | ||
8ce83369 | 695 | /* Copy the value into the new lxc_conf. */ |
d4db3d14 | 696 | bret = c->set_config_item(c, "lxc.seccomp.profile", path); |
bd7b4e28 | 697 | free(path); |
d4db3d14 CB |
698 | if (!bret) |
699 | return false; | |
bd7b4e28 | 700 | |
8ce83369 | 701 | /* Attempt to parse the resulting config. */ |
d4db3d14 CB |
702 | ret = lxc_read_seccomp_config(c->lxc_conf); |
703 | if (ret < 0) { | |
704 | ERROR("Failed to retrieve seccomp policy"); | |
2c4ea790 SH |
705 | return false; |
706 | } | |
707 | ||
d4db3d14 | 708 | INFO("Retrieved seccomp policy"); |
2e812c16 CB |
709 | return true; |
710 | } | |
711 | ||
6f4f1937 | 712 | static bool no_new_privs(struct lxc_container *c, lxc_attach_options_t *options) |
2e812c16 | 713 | { |
bcbef733 | 714 | bool bret; |
2e812c16 CB |
715 | char *val; |
716 | ||
2e812c16 | 717 | /* Remove current setting. */ |
bcbef733 CB |
718 | if (!c->set_config_item(c, "lxc.no_new_privs", "")) { |
719 | INFO("Failed to unset lxc.no_new_privs"); | |
2e812c16 | 720 | return false; |
bcbef733 | 721 | } |
2e812c16 CB |
722 | |
723 | /* Retrieve currently active setting. */ | |
724 | val = c->get_running_config_item(c, "lxc.no_new_privs"); | |
725 | if (!val) { | |
bcbef733 | 726 | INFO("Failed to retrieve lxc.no_new_privs"); |
2e812c16 CB |
727 | return false; |
728 | } | |
729 | ||
730 | /* Set currently active setting. */ | |
bcbef733 | 731 | bret = c->set_config_item(c, "lxc.no_new_privs", val); |
2e812c16 | 732 | free(val); |
bcbef733 | 733 | return bret; |
2c4ea790 SH |
734 | } |
735 | ||
9b8e3c96 SH |
736 | static signed long get_personality(const char *name, const char *lxcpath) |
737 | { | |
6f4f1937 | 738 | char *p; |
9b8e3c96 SH |
739 | signed long ret; |
740 | ||
6f4f1937 | 741 | p = lxc_cmd_get_config_item(name, "lxc.arch", lxcpath); |
9b8e3c96 SH |
742 | if (!p) |
743 | return -1; | |
6f4f1937 | 744 | |
9b8e3c96 SH |
745 | ret = lxc_config_parse_arch(p); |
746 | free(p); | |
6f4f1937 | 747 | |
9b8e3c96 SH |
748 | return ret; |
749 | } | |
750 | ||
a998454a CB |
751 | struct attach_clone_payload { |
752 | int ipc_socket; | |
9e84479f | 753 | int terminal_slave_fd; |
a998454a CB |
754 | lxc_attach_options_t *options; |
755 | struct lxc_proc_context_info *init_ctx; | |
756 | lxc_attach_exec_t exec_function; | |
757 | void *exec_payload; | |
758 | }; | |
759 | ||
ba2be1a8 CB |
760 | static void lxc_put_attach_clone_payload(struct attach_clone_payload *p) |
761 | { | |
762 | if (p->ipc_socket >= 0) { | |
763 | shutdown(p->ipc_socket, SHUT_RDWR); | |
764 | close(p->ipc_socket); | |
765 | p->ipc_socket = -EBADF; | |
766 | } | |
767 | ||
9e84479f CB |
768 | if (p->terminal_slave_fd >= 0) { |
769 | close(p->terminal_slave_fd); | |
770 | p->terminal_slave_fd = -EBADF; | |
ba2be1a8 CB |
771 | } |
772 | ||
b21da190 | 773 | if (p->init_ctx) { |
ba2be1a8 | 774 | lxc_proc_put_context_info(p->init_ctx); |
b21da190 CB |
775 | p->init_ctx = NULL; |
776 | } | |
ba2be1a8 CB |
777 | } |
778 | ||
a998454a CB |
779 | static int attach_child_main(struct attach_clone_payload *payload) |
780 | { | |
57de839f | 781 | int fd, lsm_fd, ret; |
a998454a CB |
782 | uid_t new_uid; |
783 | gid_t new_gid; | |
a998454a CB |
784 | lxc_attach_options_t* options = payload->options; |
785 | struct lxc_proc_context_info* init_ctx = payload->init_ctx; | |
57de839f CB |
786 | bool needs_lsm = (options->namespaces & CLONE_NEWNS) && |
787 | (options->attach_flags & LXC_ATTACH_LSM) && | |
788 | init_ctx->lsm_label; | |
a998454a CB |
789 | |
790 | /* A description of the purpose of this functionality is provided in the | |
791 | * lxc-attach(1) manual page. We have to remount here and not in the | |
792 | * parent process, otherwise /proc may not properly reflect the new pid | |
793 | * namespace. | |
794 | */ | |
795 | if (!(options->namespaces & CLONE_NEWNS) && | |
796 | (options->attach_flags & LXC_ATTACH_REMOUNT_PROC_SYS)) { | |
797 | ret = lxc_attach_remount_sys_proc(); | |
b75c344c CB |
798 | if (ret < 0) |
799 | goto on_error; | |
ea918412 | 800 | |
b75c344c | 801 | TRACE("Remounted \"/proc\" and \"/sys\""); |
a998454a CB |
802 | } |
803 | ||
b75c344c | 804 | /* Now perform additional attachments. */ |
a998454a | 805 | #if HAVE_SYS_PERSONALITY_H |
a998454a | 806 | if (options->attach_flags & LXC_ATTACH_SET_PERSONALITY) { |
b75c344c CB |
807 | long new_personality; |
808 | ||
809 | if (options->personality < 0) | |
810 | new_personality = init_ctx->personality; | |
811 | else | |
812 | new_personality = options->personality; | |
ea918412 | 813 | |
a998454a | 814 | ret = personality(new_personality); |
b75c344c CB |
815 | if (ret < 0) |
816 | goto on_error; | |
ea918412 | 817 | |
b75c344c | 818 | TRACE("Set new personality"); |
a998454a CB |
819 | } |
820 | #endif | |
821 | ||
822 | if (options->attach_flags & LXC_ATTACH_DROP_CAPABILITIES) { | |
823 | ret = lxc_attach_drop_privs(init_ctx); | |
b75c344c CB |
824 | if (ret < 0) |
825 | goto on_error; | |
ea918412 | 826 | |
b75c344c | 827 | TRACE("Dropped capabilities"); |
a998454a CB |
828 | } |
829 | ||
830 | /* Always set the environment (specify (LXC_ATTACH_KEEP_ENV, NULL, NULL) | |
831 | * if you want this to be a no-op). | |
832 | */ | |
7385273f | 833 | ret = lxc_attach_set_environment(init_ctx, |
834 | options->env_policy, | |
a998454a CB |
835 | options->extra_env_vars, |
836 | options->extra_keep_env); | |
b75c344c CB |
837 | if (ret < 0) |
838 | goto on_error; | |
ea918412 | 839 | |
b75c344c | 840 | TRACE("Set up environment"); |
a998454a | 841 | |
57de839f CB |
842 | /* This remark only affects fully unprivileged containers: |
843 | * Receive fd for LSM security module before we set{g,u}id(). The reason | |
844 | * is that on set{g,u}id() the kernel will a) make us undumpable and b) | |
845 | * we will change our effective uid. This means our effective uid will | |
846 | * be different from the effective uid of the process that created us | |
847 | * which means that this processs no longer has capabilities in our | |
848 | * namespace including CAP_SYS_PTRACE. This means we will not be able to | |
849 | * read and /proc/<pid> files for the process anymore when /proc is | |
850 | * mounted with hidepid={1,2}. So let's get the lsm label fd before the | |
851 | * set{g,u}id(). | |
852 | */ | |
853 | if (needs_lsm) { | |
b75c344c | 854 | ret = lxc_abstract_unix_recv_fds(payload->ipc_socket, &lsm_fd, 1, NULL, 0); |
9044b79e | 855 | if (ret <= 0) { |
856 | if (ret < 0) | |
857 | SYSERROR("Failed to receive lsm label fd"); | |
858 | ||
b75c344c | 859 | goto on_error; |
9044b79e | 860 | } |
861 | ||
57de839f CB |
862 | TRACE("Received LSM label file descriptor %d from parent", lsm_fd); |
863 | } | |
864 | ||
08ea9270 | 865 | if (options->stdin_fd > 0 && isatty(options->stdin_fd)) { |
cd0a2b2f | 866 | ret = lxc_make_controlling_terminal(options->stdin_fd); |
08ea9270 CB |
867 | if (ret < 0) |
868 | goto on_error; | |
869 | } | |
870 | ||
a998454a CB |
871 | /* Set {u,g}id. */ |
872 | new_uid = 0; | |
873 | new_gid = 0; | |
ea918412 | 874 | |
a998454a CB |
875 | /* Ignore errors, we will fall back to root in that case (/proc was not |
876 | * mounted etc.). | |
877 | */ | |
878 | if (options->namespaces & CLONE_NEWUSER) | |
879 | lxc_attach_get_init_uidgid(&new_uid, &new_gid); | |
880 | ||
881 | if (options->uid != (uid_t)-1) | |
882 | new_uid = options->uid; | |
ea918412 | 883 | |
a998454a CB |
884 | if (options->gid != (gid_t)-1) |
885 | new_gid = options->gid; | |
886 | ||
a998454a | 887 | /* Try to set the {u,g}id combination. */ |
b75c344c CB |
888 | if (new_uid != 0 || new_gid != 0 || options->namespaces & CLONE_NEWUSER) { |
889 | ret = lxc_switch_uid_gid(new_uid, new_gid); | |
890 | if (ret < 0) | |
891 | goto on_error; | |
a998454a CB |
892 | } |
893 | ||
24927339 | 894 | ret = lxc_setgroups(0, NULL); |
96ec54ac | 895 | if (ret < 0 && errno != EPERM) |
24927339 CB |
896 | goto on_error; |
897 | ||
a998454a CB |
898 | if ((init_ctx->container && init_ctx->container->lxc_conf && |
899 | init_ctx->container->lxc_conf->no_new_privs) || | |
900 | (options->attach_flags & LXC_ATTACH_NO_NEW_PRIVS)) { | |
b75c344c CB |
901 | ret = prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0); |
902 | if (ret < 0) | |
903 | goto on_error; | |
ea918412 | 904 | |
b75c344c | 905 | TRACE("Set PR_SET_NO_NEW_PRIVS"); |
a998454a CB |
906 | } |
907 | ||
57de839f | 908 | if (needs_lsm) { |
d3ba7c98 | 909 | bool on_exec; |
a998454a CB |
910 | |
911 | /* Change into our new LSM profile. */ | |
d3ba7c98 | 912 | on_exec = options->attach_flags & LXC_ATTACH_LSM_EXEC ? true : false; |
ea918412 | 913 | |
d3ba7c98 | 914 | ret = lsm_process_label_set_at(lsm_fd, init_ctx->lsm_label, on_exec); |
57de839f | 915 | close(lsm_fd); |
b75c344c CB |
916 | if (ret < 0) |
917 | goto on_error; | |
ea918412 | 918 | |
d3ba7c98 | 919 | TRACE("Set %s LSM label to \"%s\"", lsm_name(), init_ctx->lsm_label); |
a998454a CB |
920 | } |
921 | ||
922 | if (init_ctx->container && init_ctx->container->lxc_conf && | |
b75c344c CB |
923 | init_ctx->container->lxc_conf->seccomp) { |
924 | ret = lxc_seccomp_load(init_ctx->container->lxc_conf); | |
925 | if (ret < 0) | |
926 | goto on_error; | |
ea918412 | 927 | |
b75c344c | 928 | TRACE("Loaded seccomp profile"); |
a998454a | 929 | } |
ea918412 | 930 | |
b75c344c CB |
931 | shutdown(payload->ipc_socket, SHUT_RDWR); |
932 | close(payload->ipc_socket); | |
ba2be1a8 CB |
933 | payload->ipc_socket = -EBADF; |
934 | lxc_proc_put_context_info(init_ctx); | |
d35b372a | 935 | payload->init_ctx = NULL; |
a998454a CB |
936 | |
937 | /* The following is done after the communication socket is shut down. | |
938 | * That way, all errors that might (though unlikely) occur up until this | |
939 | * point will have their messages printed to the original stderr (if | |
940 | * logging is so configured) and not the fd the user supplied, if any. | |
941 | */ | |
942 | ||
943 | /* Fd handling for stdin, stdout and stderr; ignore errors here, user | |
944 | * may want to make sure the fds are closed, for example. | |
945 | */ | |
08ea9270 | 946 | if (options->stdin_fd >= 0 && options->stdin_fd != STDIN_FILENO) |
00c72a93 | 947 | (void)dup2(options->stdin_fd, STDIN_FILENO); |
08ea9270 CB |
948 | |
949 | if (options->stdout_fd >= 0 && options->stdout_fd != STDOUT_FILENO) | |
00c72a93 | 950 | (void)dup2(options->stdout_fd, STDOUT_FILENO); |
08ea9270 CB |
951 | |
952 | if (options->stderr_fd >= 0 && options->stderr_fd != STDERR_FILENO) | |
00c72a93 | 953 | (void)dup2(options->stderr_fd, STDERR_FILENO); |
a998454a CB |
954 | |
955 | /* close the old fds */ | |
08ea9270 | 956 | if (options->stdin_fd > STDERR_FILENO) |
a998454a | 957 | close(options->stdin_fd); |
08ea9270 CB |
958 | |
959 | if (options->stdout_fd > STDERR_FILENO) | |
a998454a | 960 | close(options->stdout_fd); |
08ea9270 CB |
961 | |
962 | if (options->stderr_fd > STDERR_FILENO) | |
a998454a CB |
963 | close(options->stderr_fd); |
964 | ||
965 | /* Try to remove FD_CLOEXEC flag from stdin/stdout/stderr, but also | |
966 | * here, ignore errors. | |
967 | */ | |
b75c344c | 968 | for (fd = STDIN_FILENO; fd <= STDERR_FILENO; fd++) { |
3f62938a | 969 | ret = fd_cloexec(fd, false); |
b75c344c CB |
970 | if (ret < 0) { |
971 | SYSERROR("Failed to clear FD_CLOEXEC from file descriptor %d", fd); | |
972 | goto on_error; | |
973 | } | |
a998454a CB |
974 | } |
975 | ||
9e84479f CB |
976 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
977 | ret = lxc_terminal_prepare_login(payload->terminal_slave_fd); | |
ba2be1a8 | 978 | if (ret < 0) { |
9e84479f | 979 | SYSERROR("Failed to prepare terminal file descriptor %d", payload->terminal_slave_fd); |
ba2be1a8 CB |
980 | goto on_error; |
981 | } | |
ea918412 | 982 | |
9e84479f | 983 | TRACE("Prepared terminal file descriptor %d", payload->terminal_slave_fd); |
ba2be1a8 CB |
984 | } |
985 | ||
a998454a CB |
986 | /* We're done, so we can now do whatever the user intended us to do. */ |
987 | rexit(payload->exec_function(payload->exec_payload)); | |
b75c344c CB |
988 | |
989 | on_error: | |
ba2be1a8 | 990 | lxc_put_attach_clone_payload(payload); |
b75c344c | 991 | rexit(EXIT_FAILURE); |
a998454a CB |
992 | } |
993 | ||
9e84479f CB |
994 | static int lxc_attach_terminal(struct lxc_conf *conf, |
995 | struct lxc_terminal *terminal) | |
ba2be1a8 CB |
996 | { |
997 | int ret; | |
998 | ||
9e84479f | 999 | lxc_terminal_init(terminal); |
ba2be1a8 | 1000 | |
9e84479f | 1001 | ret = lxc_terminal_create(terminal); |
ba2be1a8 | 1002 | if (ret < 0) { |
ea918412 | 1003 | ERROR("Failed to create terminal"); |
ba2be1a8 CB |
1004 | return -1; |
1005 | } | |
1006 | ||
1007 | /* Shift ttys to container. */ | |
9e84479f | 1008 | ret = lxc_terminal_map_ids(conf, terminal); |
ba2be1a8 | 1009 | if (ret < 0) { |
9e84479f | 1010 | ERROR("Failed to chown terminal"); |
ba2be1a8 CB |
1011 | goto on_error; |
1012 | } | |
1013 | ||
1014 | return 0; | |
1015 | ||
1016 | on_error: | |
9e84479f CB |
1017 | lxc_terminal_delete(terminal); |
1018 | lxc_terminal_conf_free(terminal); | |
ba2be1a8 CB |
1019 | return -1; |
1020 | } | |
1021 | ||
9e84479f CB |
1022 | static int lxc_attach_terminal_mainloop_init(struct lxc_terminal *terminal, |
1023 | struct lxc_epoll_descr *descr) | |
ba2be1a8 CB |
1024 | { |
1025 | int ret; | |
1026 | ||
1027 | ret = lxc_mainloop_open(descr); | |
1028 | if (ret < 0) { | |
1029 | ERROR("Failed to create mainloop"); | |
1030 | return -1; | |
1031 | } | |
1032 | ||
9e84479f | 1033 | ret = lxc_terminal_mainloop_add(descr, terminal); |
ba2be1a8 CB |
1034 | if (ret < 0) { |
1035 | ERROR("Failed to add handlers to mainloop"); | |
1036 | lxc_mainloop_close(descr); | |
1037 | return -1; | |
1038 | } | |
1039 | ||
1040 | return 0; | |
1041 | } | |
1042 | ||
9e84479f | 1043 | static inline void lxc_attach_terminal_close_master(struct lxc_terminal *terminal) |
ba2be1a8 | 1044 | { |
9e84479f | 1045 | if (terminal->master < 0) |
ba2be1a8 CB |
1046 | return; |
1047 | ||
9e84479f CB |
1048 | close(terminal->master); |
1049 | terminal->master = -EBADF; | |
ba2be1a8 CB |
1050 | } |
1051 | ||
9e84479f | 1052 | static inline void lxc_attach_terminal_close_slave(struct lxc_terminal *terminal) |
ba2be1a8 | 1053 | { |
9e84479f | 1054 | if (terminal->slave < 0) |
ba2be1a8 CB |
1055 | return; |
1056 | ||
9e84479f CB |
1057 | close(terminal->slave); |
1058 | terminal->slave = -EBADF; | |
ba2be1a8 CB |
1059 | } |
1060 | ||
9e84479f | 1061 | static inline void lxc_attach_terminal_close_peer(struct lxc_terminal *terminal) |
ba2be1a8 | 1062 | { |
9e84479f | 1063 | if (terminal->peer < 0) |
ba2be1a8 CB |
1064 | return; |
1065 | ||
9e84479f CB |
1066 | close(terminal->peer); |
1067 | terminal->peer = -EBADF; | |
ba2be1a8 CB |
1068 | } |
1069 | ||
9e84479f | 1070 | static inline void lxc_attach_terminal_close_log(struct lxc_terminal *terminal) |
ba2be1a8 | 1071 | { |
9e84479f | 1072 | if (terminal->log_fd < 0) |
ba2be1a8 CB |
1073 | return; |
1074 | ||
9e84479f CB |
1075 | close(terminal->log_fd); |
1076 | terminal->log_fd = -EBADF; | |
ba2be1a8 CB |
1077 | } |
1078 | ||
6f4f1937 CB |
1079 | int lxc_attach(const char *name, const char *lxcpath, |
1080 | lxc_attach_exec_t exec_function, void *exec_payload, | |
1081 | lxc_attach_options_t *options, pid_t *attached_process) | |
9c4693b8 | 1082 | { |
877f3a04 | 1083 | int i, ret, status; |
9c4693b8 | 1084 | int ipc_sockets[2]; |
6f4f1937 | 1085 | char *cwd, *new_cwd; |
9b8e3c96 | 1086 | signed long personality; |
ba2be1a8 | 1087 | pid_t attached_pid, init_pid, pid; |
6f4f1937 | 1088 | struct lxc_proc_context_info *init_ctx; |
9e84479f | 1089 | struct lxc_terminal terminal; |
1cce35e6 | 1090 | struct lxc_conf *conf; |
a998454a | 1091 | struct attach_clone_payload payload = {0}; |
9c4693b8 | 1092 | |
877f3a04 CB |
1093 | ret = access("/proc/self/ns", X_OK); |
1094 | if (ret) { | |
ea918412 | 1095 | SYSERROR("Does this kernel version support namespaces?"); |
877f3a04 CB |
1096 | return -1; |
1097 | } | |
1098 | ||
9c4693b8 CS |
1099 | if (!options) |
1100 | options = &attach_static_default_options; | |
1101 | ||
1102 | init_pid = lxc_cmd_get_init_pid(name, lxcpath); | |
1103 | if (init_pid < 0) { | |
ae026f55 | 1104 | ERROR("Failed to get init pid"); |
9c4693b8 CS |
1105 | return -1; |
1106 | } | |
1107 | ||
1108 | init_ctx = lxc_proc_get_context_info(init_pid); | |
1109 | if (!init_ctx) { | |
6f4f1937 | 1110 | ERROR("Failed to get context of init process: %ld", (long)init_pid); |
9c4693b8 CS |
1111 | return -1; |
1112 | } | |
1113 | ||
9b8e3c96 SH |
1114 | personality = get_personality(name, lxcpath); |
1115 | if (init_ctx->personality < 0) { | |
6f4f1937 | 1116 | ERROR("Failed to get personality of the container"); |
9b8e3c96 SH |
1117 | lxc_proc_put_context_info(init_ctx); |
1118 | return -1; | |
1119 | } | |
1120 | init_ctx->personality = personality; | |
1121 | ||
ff07d7bb | 1122 | init_ctx->container = lxc_container_new(name, lxcpath); |
62de1db6 CB |
1123 | if (!init_ctx->container) { |
1124 | lxc_proc_put_context_info(init_ctx); | |
ff07d7bb | 1125 | return -1; |
62de1db6 | 1126 | } |
ff07d7bb | 1127 | |
ba773996 CB |
1128 | if (!init_ctx->container->lxc_conf) { |
1129 | init_ctx->container->lxc_conf = lxc_conf_init(); | |
62de1db6 CB |
1130 | if (!init_ctx->container->lxc_conf) { |
1131 | lxc_proc_put_context_info(init_ctx); | |
ea918412 | 1132 | return -1; |
62de1db6 | 1133 | } |
ba773996 | 1134 | } |
1cce35e6 | 1135 | conf = init_ctx->container->lxc_conf; |
ba773996 | 1136 | |
bd4307f0 | 1137 | if (!fetch_seccomp(init_ctx->container, options)) |
ae026f55 | 1138 | WARN("Failed to get seccomp policy"); |
2c4ea790 | 1139 | |
bd4307f0 | 1140 | if (!no_new_privs(init_ctx->container, options)) |
ae026f55 | 1141 | WARN("Could not determine whether PR_SET_NO_NEW_PRIVS is set"); |
2e812c16 | 1142 | |
9c4693b8 CS |
1143 | cwd = getcwd(NULL, 0); |
1144 | ||
8ce83369 CB |
1145 | /* Determine which namespaces the container was created with |
1146 | * by asking lxc-start, if necessary. | |
9c4693b8 CS |
1147 | */ |
1148 | if (options->namespaces == -1) { | |
1149 | options->namespaces = lxc_cmd_get_clone_flags(name, lxcpath); | |
1150 | /* call failed */ | |
1151 | if (options->namespaces == -1) { | |
8ce83369 | 1152 | ERROR("Failed to automatically determine the " |
877f3a04 | 1153 | "namespaces which the container uses"); |
9c4693b8 | 1154 | free(cwd); |
fe4de9a6 | 1155 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1156 | return -1; |
1157 | } | |
877f3a04 CB |
1158 | |
1159 | for (i = 0; i < LXC_NS_MAX; i++) { | |
1160 | if (ns_info[i].clone_flag & CLONE_NEWCGROUP) | |
1161 | if (!(options->attach_flags & LXC_ATTACH_MOVE_TO_CGROUP) || | |
1162 | !cgns_supported()) | |
1163 | continue; | |
1164 | ||
1165 | if (ns_info[i].clone_flag & options->namespaces) | |
1166 | continue; | |
1167 | ||
1168 | init_ctx->ns_inherited |= ns_info[i].clone_flag; | |
1169 | } | |
1170 | } | |
1171 | ||
0059379f | 1172 | pid = lxc_raw_getpid(); |
ea918412 | 1173 | |
877f3a04 | 1174 | for (i = 0; i < LXC_NS_MAX; i++) { |
ea918412 | 1175 | int j; |
877f3a04 CB |
1176 | |
1177 | if (options->namespaces & ns_info[i].clone_flag) | |
1178 | init_ctx->ns_fd[i] = lxc_preserve_ns(init_pid, ns_info[i].proc_name); | |
1179 | else if (init_ctx->ns_inherited & ns_info[i].clone_flag) | |
1180 | init_ctx->ns_fd[i] = in_same_namespace(pid, init_pid, ns_info[i].proc_name); | |
1181 | else | |
1182 | continue; | |
ea918412 | 1183 | |
877f3a04 CB |
1184 | if (init_ctx->ns_fd[i] >= 0) |
1185 | continue; | |
1186 | ||
1187 | if (init_ctx->ns_fd[i] == -EINVAL) { | |
1188 | DEBUG("Inheriting %s namespace from %d", | |
1189 | ns_info[i].proc_name, pid); | |
1190 | init_ctx->ns_inherited &= ~ns_info[i].clone_flag; | |
1191 | continue; | |
1192 | } | |
1193 | ||
1194 | /* We failed to preserve the namespace. */ | |
ea918412 | 1195 | SYSERROR("Failed to attach to %s namespace of %d", |
1196 | ns_info[i].proc_name, pid); | |
1197 | ||
877f3a04 CB |
1198 | /* Close all already opened file descriptors before we return an |
1199 | * error, so we don't leak them. | |
1200 | */ | |
1201 | for (j = 0; j < i; j++) | |
1202 | close(init_ctx->ns_fd[j]); | |
1203 | ||
877f3a04 CB |
1204 | free(cwd); |
1205 | lxc_proc_put_context_info(init_ctx); | |
1206 | return -1; | |
9c4693b8 CS |
1207 | } |
1208 | ||
9e84479f CB |
1209 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
1210 | ret = lxc_attach_terminal(conf, &terminal); | |
ba2be1a8 | 1211 | if (ret < 0) { |
9e84479f | 1212 | ERROR("Failed to setup new terminal"); |
ba2be1a8 CB |
1213 | free(cwd); |
1214 | lxc_proc_put_context_info(init_ctx); | |
1215 | return -1; | |
1216 | } | |
1217 | ||
9e84479f | 1218 | terminal.log_fd = options->log_fd; |
c948657b | 1219 | } else { |
9e84479f | 1220 | lxc_terminal_init(&terminal); |
ba2be1a8 CB |
1221 | } |
1222 | ||
8ce83369 CB |
1223 | /* Create a socket pair for IPC communication; set SOCK_CLOEXEC in order |
1224 | * to make sure we don't irritate other threads that want to fork+exec | |
1225 | * away | |
9c4693b8 CS |
1226 | * |
1227 | * IMPORTANT: if the initial process is multithreaded and another call | |
1228 | * just fork()s away without exec'ing directly after, the socket fd will | |
1229 | * exist in the forked process from the other thread and any close() in | |
8ce83369 CB |
1230 | * our own child process will not really cause the socket to close |
1231 | * properly, potentiall causing the parent to hang. | |
9c4693b8 CS |
1232 | * |
1233 | * For this reason, while IPC is still active, we have to use shutdown() | |
8ce83369 CB |
1234 | * if the child exits prematurely in order to signal that the socket is |
1235 | * closed and cannot assume that the child exiting will automatically do | |
1236 | * that. | |
9c4693b8 CS |
1237 | * |
1238 | * IPC mechanism: (X is receiver) | |
1239 | * initial process intermediate attached | |
1240 | * X <--- send pid of | |
1241 | * attached proc, | |
1242 | * then exit | |
1243 | * send 0 ------------------------------------> X | |
1244 | * [do initialization] | |
1245 | * X <------------------------------------ send 1 | |
1246 | * [add to cgroup, ...] | |
1247 | * send 2 ------------------------------------> X | |
81f466d0 CB |
1248 | * [set LXC_ATTACH_NO_NEW_PRIVS] |
1249 | * X <------------------------------------ send 3 | |
1250 | * [open LSM label fd] | |
1251 | * send 4 ------------------------------------> X | |
1252 | * [set LSM label] | |
9c4693b8 CS |
1253 | * close socket close socket |
1254 | * run program | |
1255 | */ | |
1256 | ret = socketpair(PF_LOCAL, SOCK_STREAM | SOCK_CLOEXEC, 0, ipc_sockets); | |
1257 | if (ret < 0) { | |
ae026f55 | 1258 | SYSERROR("Could not set up required IPC mechanism for attaching"); |
9c4693b8 | 1259 | free(cwd); |
fe4de9a6 | 1260 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1261 | return -1; |
1262 | } | |
1263 | ||
e3f0e436 CB |
1264 | /* Create intermediate subprocess, two reasons: |
1265 | * 1. We can't setns() in the child itself, since we want to make | |
8ce83369 | 1266 | * sure we are properly attached to the pidns. |
e3f0e436 | 1267 | * 2. Also, the initial thread has to put the attached process |
8ce83369 CB |
1268 | * into the cgroup, which we can only do if we didn't already |
1269 | * setns() (otherwise, user namespaces will hate us). | |
9c4693b8 CS |
1270 | */ |
1271 | pid = fork(); | |
9c4693b8 | 1272 | if (pid < 0) { |
ae026f55 | 1273 | SYSERROR("Failed to create first subprocess"); |
9c4693b8 | 1274 | free(cwd); |
fe4de9a6 | 1275 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1276 | return -1; |
1277 | } | |
1278 | ||
1279 | if (pid) { | |
ba2be1a8 | 1280 | int ret_parent = -1; |
9c4693b8 | 1281 | pid_t to_cleanup_pid = pid; |
ba2be1a8 | 1282 | struct lxc_epoll_descr descr = {0}; |
9c4693b8 | 1283 | |
ba2be1a8 | 1284 | /* close unneeded file descriptors */ |
9c4693b8 CS |
1285 | close(ipc_sockets[1]); |
1286 | free(cwd); | |
ba2be1a8 | 1287 | lxc_proc_close_ns_fd(init_ctx); |
9e84479f CB |
1288 | if (options->attach_flags & LXC_ATTACH_TERMINAL) |
1289 | lxc_attach_terminal_close_slave(&terminal); | |
9c4693b8 | 1290 | |
8ce83369 | 1291 | /* Attach to cgroup, if requested. */ |
f4364484 | 1292 | if (options->attach_flags & LXC_ATTACH_MOVE_TO_CGROUP) { |
2202afc9 CB |
1293 | struct cgroup_ops *cgroup_ops; |
1294 | ||
1295 | cgroup_ops = cgroup_init(NULL); | |
1296 | if (!cgroup_ops) | |
1297 | goto on_error; | |
1298 | ||
1299 | if (!cgroup_ops->attach(cgroup_ops, name, lxcpath, pid)) | |
8ce83369 | 1300 | goto on_error; |
2202afc9 CB |
1301 | |
1302 | cgroup_exit(cgroup_ops); | |
1303 | TRACE("Moved intermediate process %d into container's cgroups", pid); | |
f4364484 SG |
1304 | } |
1305 | ||
bb2ada6f | 1306 | /* Setup /proc limits */ |
1cce35e6 CB |
1307 | if (!lxc_list_empty(&conf->procs)) { |
1308 | ret = setup_proc_filesystem(&conf->procs, pid); | |
bb2ada6f CB |
1309 | if (ret < 0) |
1310 | goto on_error; | |
1311 | } | |
1312 | ||
c6d09e15 | 1313 | /* Setup resource limits */ |
1cce35e6 CB |
1314 | if (!lxc_list_empty(&conf->limits)) { |
1315 | ret = setup_resource_limits(&conf->limits, pid); | |
ba2be1a8 CB |
1316 | if (ret < 0) |
1317 | goto on_error; | |
1318 | } | |
1319 | ||
9e84479f CB |
1320 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
1321 | ret = lxc_attach_terminal_mainloop_init(&terminal, &descr); | |
ba2be1a8 | 1322 | if (ret < 0) |
6f4f1937 | 1323 | goto on_error; |
ea918412 | 1324 | |
9e84479f | 1325 | TRACE("Initialized terminal mainloop"); |
ba2be1a8 | 1326 | } |
c6d09e15 | 1327 | |
8ce83369 | 1328 | /* Let the child process know to go ahead. */ |
f4364484 SG |
1329 | status = 0; |
1330 | ret = lxc_write_nointr(ipc_sockets[0], &status, sizeof(status)); | |
94ac256f | 1331 | if (ret != sizeof(status)) |
ba2be1a8 | 1332 | goto close_mainloop; |
ea918412 | 1333 | |
94ac256f | 1334 | TRACE("Told intermediate process to start initializing"); |
f4364484 | 1335 | |
8ce83369 | 1336 | /* Get pid of attached process from intermediate process. */ |
94ac256f CB |
1337 | ret = lxc_read_nointr(ipc_sockets[0], &attached_pid, sizeof(attached_pid)); |
1338 | if (ret != sizeof(attached_pid)) | |
ba2be1a8 | 1339 | goto close_mainloop; |
ea918412 | 1340 | |
94ac256f | 1341 | TRACE("Received pid %d of attached process in parent pid namespace", attached_pid); |
9c4693b8 | 1342 | |
8ce83369 | 1343 | /* Ignore SIGKILL (CTRL-C) and SIGQUIT (CTRL-\) - issue #313. */ |
62183f1a SH |
1344 | if (options->stdin_fd == 0) { |
1345 | signal(SIGINT, SIG_IGN); | |
1346 | signal(SIGQUIT, SIG_IGN); | |
1347 | } | |
2eef2bda | 1348 | |
8ce83369 | 1349 | /* Reap intermediate process. */ |
9c4693b8 CS |
1350 | ret = wait_for_pid(pid); |
1351 | if (ret < 0) | |
ba2be1a8 | 1352 | goto close_mainloop; |
ea918412 | 1353 | |
94ac256f | 1354 | TRACE("Intermediate process %d exited", pid); |
9c4693b8 | 1355 | |
8ce83369 | 1356 | /* We will always have to reap the attached process now. */ |
9c4693b8 CS |
1357 | to_cleanup_pid = attached_pid; |
1358 | ||
81f466d0 | 1359 | /* Open LSM fd and send it to child. */ |
6f4f1937 CB |
1360 | if ((options->namespaces & CLONE_NEWNS) && |
1361 | (options->attach_flags & LXC_ATTACH_LSM) && | |
1362 | init_ctx->lsm_label) { | |
94ac256f | 1363 | int ret = -1; |
47ce2cb7 CB |
1364 | int labelfd; |
1365 | bool on_exec; | |
6f4f1937 | 1366 | |
47ce2cb7 CB |
1367 | on_exec = options->attach_flags & LXC_ATTACH_LSM_EXEC ? true : false; |
1368 | labelfd = lsm_process_label_fd_get(attached_pid, on_exec); | |
81f466d0 | 1369 | if (labelfd < 0) |
ba2be1a8 | 1370 | goto close_mainloop; |
ea918412 | 1371 | |
94ac256f | 1372 | TRACE("Opened LSM label file descriptor %d", labelfd); |
81f466d0 CB |
1373 | |
1374 | /* Send child fd of the LSM security module to write to. */ | |
ae467c54 | 1375 | ret = lxc_abstract_unix_send_fds(ipc_sockets[0], &labelfd, 1, NULL, 0); |
81f466d0 | 1376 | if (ret <= 0) { |
9044b79e | 1377 | if (ret < 0) |
1378 | SYSERROR("Failed to send lsm label fd"); | |
1379 | ||
1380 | close(labelfd); | |
ba2be1a8 | 1381 | goto close_mainloop; |
81f466d0 | 1382 | } |
9044b79e | 1383 | |
1384 | close(labelfd); | |
94ac256f | 1385 | TRACE("Sent LSM label file descriptor %d to child", labelfd); |
81f466d0 CB |
1386 | } |
1387 | ||
8ce83369 CB |
1388 | /* We're done, the child process should now execute whatever it |
1389 | * is that the user requested. The parent can now track it with | |
1390 | * waitpid() or similar. | |
9c4693b8 CS |
1391 | */ |
1392 | ||
1393 | *attached_process = attached_pid; | |
9c4693b8 | 1394 | |
ba2be1a8 | 1395 | /* Now shut down communication with child, we're done. */ |
9c4693b8 CS |
1396 | shutdown(ipc_sockets[0], SHUT_RDWR); |
1397 | close(ipc_sockets[0]); | |
ba2be1a8 CB |
1398 | ipc_sockets[0] = -1; |
1399 | ||
1400 | ret_parent = 0; | |
1401 | to_cleanup_pid = -1; | |
ea918412 | 1402 | |
9e84479f | 1403 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
ba2be1a8 CB |
1404 | ret = lxc_mainloop(&descr, -1); |
1405 | if (ret < 0) { | |
1406 | ret_parent = -1; | |
1407 | to_cleanup_pid = attached_pid; | |
1408 | } | |
1409 | } | |
1410 | ||
1411 | close_mainloop: | |
9e84479f | 1412 | if (options->attach_flags & LXC_ATTACH_TERMINAL) |
ba2be1a8 CB |
1413 | lxc_mainloop_close(&descr); |
1414 | ||
1415 | on_error: | |
1416 | if (ipc_sockets[0] >= 0) { | |
1417 | shutdown(ipc_sockets[0], SHUT_RDWR); | |
1418 | close(ipc_sockets[0]); | |
1419 | } | |
1420 | ||
1421 | if (to_cleanup_pid > 0) | |
6f4f1937 | 1422 | (void)wait_for_pid(to_cleanup_pid); |
ba2be1a8 | 1423 | |
9e84479f CB |
1424 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
1425 | lxc_terminal_delete(&terminal); | |
1426 | lxc_terminal_conf_free(&terminal); | |
ba2be1a8 | 1427 | } |
ea918412 | 1428 | |
fe4de9a6 | 1429 | lxc_proc_put_context_info(init_ctx); |
ba2be1a8 | 1430 | return ret_parent; |
9c4693b8 CS |
1431 | } |
1432 | ||
ba2be1a8 | 1433 | /* close unneeded file descriptors */ |
9c4693b8 | 1434 | close(ipc_sockets[0]); |
ba2be1a8 | 1435 | ipc_sockets[0] = -EBADF; |
ea918412 | 1436 | |
9e84479f CB |
1437 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
1438 | lxc_attach_terminal_close_master(&terminal); | |
1439 | lxc_attach_terminal_close_peer(&terminal); | |
1440 | lxc_attach_terminal_close_log(&terminal); | |
ba2be1a8 | 1441 | } |
9c4693b8 | 1442 | |
8ce83369 | 1443 | /* Wait for the parent to have setup cgroups. */ |
94ac256f | 1444 | ret = lxc_read_nointr(ipc_sockets[1], &status, sizeof(status)); |
ba2be1a8 | 1445 | if (ret != sizeof(status)) { |
f4364484 | 1446 | shutdown(ipc_sockets[1], SHUT_RDWR); |
62de1db6 | 1447 | lxc_proc_put_context_info(init_ctx); |
f4364484 SG |
1448 | rexit(-1); |
1449 | } | |
ea918412 | 1450 | |
94ac256f | 1451 | TRACE("Intermediate process starting to initialize"); |
f4364484 | 1452 | |
8ce83369 CB |
1453 | /* Attach now, create another subprocess later, since pid namespaces |
1454 | * only really affect the children of the current process. | |
9c4693b8 | 1455 | */ |
877f3a04 | 1456 | ret = lxc_attach_to_ns(init_pid, init_ctx); |
9c4693b8 | 1457 | if (ret < 0) { |
94ac256f | 1458 | ERROR("Failed to enter namespaces"); |
9c4693b8 | 1459 | shutdown(ipc_sockets[1], SHUT_RDWR); |
62de1db6 | 1460 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1461 | rexit(-1); |
1462 | } | |
ea918412 | 1463 | |
877f3a04 CB |
1464 | /* close namespace file descriptors */ |
1465 | lxc_proc_close_ns_fd(init_ctx); | |
9c4693b8 | 1466 | |
8ce83369 | 1467 | /* Attach succeeded, try to cwd. */ |
9c4693b8 CS |
1468 | if (options->initial_cwd) |
1469 | new_cwd = options->initial_cwd; | |
1470 | else | |
1471 | new_cwd = cwd; | |
d6d979bc CB |
1472 | if (new_cwd) { |
1473 | ret = chdir(new_cwd); | |
1474 | if (ret < 0) | |
1475 | WARN("Could not change directory to \"%s\"", new_cwd); | |
1476 | } | |
9c4693b8 CS |
1477 | free(cwd); |
1478 | ||
a998454a CB |
1479 | /* Create attached process. */ |
1480 | payload.ipc_socket = ipc_sockets[1]; | |
1481 | payload.options = options; | |
1482 | payload.init_ctx = init_ctx; | |
9e84479f | 1483 | payload.terminal_slave_fd = terminal.slave; |
a998454a CB |
1484 | payload.exec_function = exec_function; |
1485 | payload.exec_payload = exec_payload; | |
9c4693b8 | 1486 | |
a998454a CB |
1487 | pid = lxc_raw_clone(CLONE_PARENT); |
1488 | if (pid < 0) { | |
94ac256f | 1489 | SYSERROR("Failed to clone attached process"); |
9c4693b8 | 1490 | shutdown(ipc_sockets[1], SHUT_RDWR); |
62de1db6 | 1491 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1492 | rexit(-1); |
1493 | } | |
a998454a CB |
1494 | |
1495 | if (pid == 0) { | |
1496 | ret = attach_child_main(&payload); | |
1497 | if (ret < 0) | |
1498 | ERROR("Failed to exec"); | |
ea918412 | 1499 | |
a998454a CB |
1500 | _exit(EXIT_FAILURE); |
1501 | } | |
ea918412 | 1502 | |
9e84479f CB |
1503 | if (options->attach_flags & LXC_ATTACH_TERMINAL) |
1504 | lxc_attach_terminal_close_slave(&terminal); | |
9c4693b8 | 1505 | |
8ce83369 | 1506 | /* Tell grandparent the pid of the pid of the newly created child. */ |
9c4693b8 CS |
1507 | ret = lxc_write_nointr(ipc_sockets[1], &pid, sizeof(pid)); |
1508 | if (ret != sizeof(pid)) { | |
8ce83369 CB |
1509 | /* If this really happens here, this is very unfortunate, since |
1510 | * the parent will not know the pid of the attached process and | |
1511 | * will not be able to wait for it (and we won't either due to | |
1512 | * CLONE_PARENT) so the parent won't be able to reap it and the | |
1513 | * attached process will remain a zombie. | |
9c4693b8 | 1514 | */ |
9c4693b8 | 1515 | shutdown(ipc_sockets[1], SHUT_RDWR); |
62de1db6 | 1516 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1517 | rexit(-1); |
1518 | } | |
ea918412 | 1519 | |
94ac256f | 1520 | TRACE("Sending pid %d of attached process", pid); |
9c4693b8 | 1521 | |
8ce83369 | 1522 | /* The rest is in the hands of the initial and the attached process. */ |
62de1db6 | 1523 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1524 | rexit(0); |
1525 | } | |
1526 | ||
9c4693b8 CS |
1527 | int lxc_attach_run_command(void* payload) |
1528 | { | |
1529 | lxc_attach_command_t* cmd = (lxc_attach_command_t*)payload; | |
1530 | ||
1531 | execvp(cmd->program, cmd->argv); | |
ea918412 | 1532 | |
1533 | SYSERROR("Failed to exec \"%s\"", cmd->program); | |
9c4693b8 CS |
1534 | return -1; |
1535 | } | |
1536 | ||
1537 | int lxc_attach_run_shell(void* payload) | |
1538 | { | |
1539 | uid_t uid; | |
cb7aa5e8 DJ |
1540 | struct passwd pwent; |
1541 | struct passwd *pwentp = NULL; | |
9c4693b8 | 1542 | char *user_shell; |
cb7aa5e8 DJ |
1543 | char *buf; |
1544 | size_t bufsize; | |
1545 | int ret; | |
9c4693b8 | 1546 | |
8ce83369 | 1547 | /* Ignore payload parameter. */ |
9c4693b8 CS |
1548 | (void)payload; |
1549 | ||
1550 | uid = getuid(); | |
cb7aa5e8 DJ |
1551 | |
1552 | bufsize = sysconf(_SC_GETPW_R_SIZE_MAX); | |
1553 | if (bufsize == -1) | |
1554 | bufsize = 1024; | |
1555 | ||
1556 | buf = malloc(bufsize); | |
1557 | if (buf) { | |
1558 | ret = getpwuid_r(uid, &pwent, buf, bufsize, &pwentp); | |
1559 | if (!pwentp) { | |
1560 | if (ret == 0) | |
ea918412 | 1561 | WARN("Could not find matched password record"); |
cb7aa5e8 DJ |
1562 | |
1563 | WARN("Failed to get password record - %u", uid); | |
1564 | } | |
1565 | } | |
9c4693b8 | 1566 | |
8ce83369 CB |
1567 | /* This probably happens because of incompatible nss implementations in |
1568 | * host and container (remember, this code is still using the host's | |
1569 | * glibc but our mount namespace is in the container) we may try to get | |
1570 | * the information by spawning a [getent passwd uid] process and parsing | |
1571 | * the result. | |
9c4693b8 | 1572 | */ |
cb7aa5e8 | 1573 | if (!pwentp) |
9c4693b8 CS |
1574 | user_shell = lxc_attach_getpwshell(uid); |
1575 | else | |
cb7aa5e8 | 1576 | user_shell = pwent.pw_shell; |
ea918412 | 1577 | |
9c4693b8 | 1578 | if (user_shell) |
acf47e1b | 1579 | execlp(user_shell, user_shell, (char *)NULL); |
9c4693b8 | 1580 | |
8ce83369 CB |
1581 | /* Executed if either no passwd entry or execvp fails, we will fall back |
1582 | * on /bin/sh as a default shell. | |
9c4693b8 | 1583 | */ |
acf47e1b | 1584 | execlp("/bin/sh", "/bin/sh", (char *)NULL); |
ea918412 | 1585 | |
edeb1836 | 1586 | SYSERROR("Failed to execute shell"); |
cb7aa5e8 | 1587 | if (!pwentp) |
edeb1836 | 1588 | free(user_shell); |
ea918412 | 1589 | |
cb7aa5e8 | 1590 | free(buf); |
9c4693b8 CS |
1591 | return -1; |
1592 | } |