]>
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> |
604ca1c0 | 28 | #include <linux/unistd.h> |
6f4f1937 CB |
29 | #include <pwd.h> |
30 | #include <signal.h> | |
31 | #include <stdio.h> | |
32 | #include <stdlib.h> | |
33 | #include <string.h> | |
6f4f1937 | 34 | #include <sys/mount.h> |
e0732705 CS |
35 | #include <sys/param.h> |
36 | #include <sys/prctl.h> | |
5ec27989 | 37 | #include <sys/socket.h> |
1ba0013f | 38 | #include <sys/syscall.h> |
905022f7 | 39 | #include <sys/wait.h> |
604ca1c0 CB |
40 | #include <termios.h> |
41 | #include <unistd.h> | |
6f4f1937 CB |
42 | |
43 | #include <lxc/lxccontainer.h> | |
e0732705 | 44 | |
81f466d0 | 45 | #include "af_unix.h" |
e0732705 CS |
46 | #include "attach.h" |
47 | #include "caps.h" | |
9c4693b8 | 48 | #include "cgroup.h" |
6f4f1937 | 49 | #include "commands.h" |
2c4ea790 | 50 | #include "conf.h" |
6f4f1937 | 51 | #include "config.h" |
9b8e3c96 | 52 | #include "confile.h" |
6f4f1937 CB |
53 | #include "log.h" |
54 | #include "lsm/lsm.h" | |
55 | #include "lxclock.h" | |
56 | #include "lxcseccomp.h" | |
604ca1c0 | 57 | #include "macro.h" |
ba2be1a8 | 58 | #include "mainloop.h" |
6f4f1937 | 59 | #include "namespace.h" |
0ed9b1bc | 60 | #include "terminal.h" |
6f4f1937 | 61 | #include "utils.h" |
9c4693b8 CS |
62 | |
63 | #if HAVE_SYS_PERSONALITY_H | |
64 | #include <sys/personality.h> | |
65 | #endif | |
e0732705 | 66 | |
ac2cecc4 | 67 | lxc_log_define(attach, lxc); |
e0732705 | 68 | |
ef05d368 CB |
69 | /* Define default options if no options are supplied by the user. */ |
70 | static lxc_attach_options_t attach_static_default_options = LXC_ATTACH_OPTIONS_DEFAULT; | |
71 | ||
74a3920a | 72 | static struct lxc_proc_context_info *lxc_proc_get_context_info(pid_t pid) |
e0732705 | 73 | { |
6f4f1937 CB |
74 | int ret; |
75 | bool found; | |
e0732705 | 76 | FILE *proc_file; |
604ca1c0 | 77 | char proc_fn[LXC_PROC_STATUS_LEN]; |
e0732705 | 78 | size_t line_bufsz = 0; |
6f4f1937 | 79 | char *line = NULL; |
8ce83369 | 80 | struct lxc_proc_context_info *info = NULL; |
e0732705 | 81 | |
8ce83369 | 82 | /* Read capabilities. */ |
604ca1c0 CB |
83 | ret = snprintf(proc_fn, LXC_PROC_STATUS_LEN, "/proc/%d/status", pid); |
84 | if (ret < 0 || ret >= LXC_PROC_STATUS_LEN) | |
8ce83369 | 85 | goto on_error; |
e0732705 CS |
86 | |
87 | proc_file = fopen(proc_fn, "r"); | |
88 | if (!proc_file) { | |
ea918412 | 89 | SYSERROR("Could not open %s", proc_fn); |
8ce83369 | 90 | goto on_error; |
e0732705 CS |
91 | } |
92 | ||
8ce83369 CB |
93 | info = calloc(1, sizeof(*info)); |
94 | if (!info) { | |
ea918412 | 95 | SYSERROR("Could not allocate memory"); |
17ac5301 | 96 | fclose(proc_file); |
8ce83369 CB |
97 | return NULL; |
98 | } | |
99 | ||
100 | found = false; | |
ea918412 | 101 | |
e0732705 CS |
102 | while (getline(&line, &line_bufsz, proc_file) != -1) { |
103 | ret = sscanf(line, "CapBnd: %llx", &info->capability_mask); | |
8ce83369 CB |
104 | if (ret != EOF && ret == 1) { |
105 | found = true; | |
e0732705 CS |
106 | break; |
107 | } | |
108 | } | |
109 | ||
f10fad2f | 110 | free(line); |
e0732705 CS |
111 | fclose(proc_file); |
112 | ||
113 | if (!found) { | |
ea918412 | 114 | ERROR("Could not read capability bounding set from %s", |
115 | proc_fn); | |
8ce83369 | 116 | goto on_error; |
e0732705 CS |
117 | } |
118 | ||
fe4de9a6 | 119 | info->lsm_label = lsm_process_label_get(pid); |
877f3a04 CB |
120 | info->ns_inherited = 0; |
121 | memset(info->ns_fd, -1, sizeof(int) * LXC_NS_MAX); | |
e0732705 | 122 | |
e0732705 CS |
123 | return info; |
124 | ||
8ce83369 | 125 | on_error: |
460a1cf0 | 126 | free(info); |
e0732705 CS |
127 | return NULL; |
128 | } | |
129 | ||
877f3a04 CB |
130 | static inline void lxc_proc_close_ns_fd(struct lxc_proc_context_info *ctx) |
131 | { | |
132 | int i; | |
133 | ||
134 | for (i = 0; i < LXC_NS_MAX; i++) { | |
135 | if (ctx->ns_fd[i] < 0) | |
136 | continue; | |
ea918412 | 137 | |
877f3a04 CB |
138 | close(ctx->ns_fd[i]); |
139 | ctx->ns_fd[i] = -EBADF; | |
140 | } | |
141 | } | |
142 | ||
fe4de9a6 DE |
143 | static void lxc_proc_put_context_info(struct lxc_proc_context_info *ctx) |
144 | { | |
f10fad2f | 145 | free(ctx->lsm_label); |
08ea9270 CB |
146 | ctx->lsm_label = NULL; |
147 | ||
148 | if (ctx->container) { | |
2c4ea790 | 149 | lxc_container_put(ctx->container); |
08ea9270 CB |
150 | ctx->container = NULL; |
151 | } | |
152 | ||
877f3a04 | 153 | lxc_proc_close_ns_fd(ctx); |
fe4de9a6 DE |
154 | free(ctx); |
155 | } | |
156 | ||
299d1198 CB |
157 | /** |
158 | * in_same_namespace - Check whether two processes are in the same namespace. | |
159 | * @pid1 - PID of the first process. | |
160 | * @pid2 - PID of the second process. | |
161 | * @ns - Name of the namespace to check. Must correspond to one of the names | |
162 | * for the namespaces as shown in /proc/<pid/ns/ | |
163 | * | |
164 | * If the two processes are not in the same namespace returns an fd to the | |
165 | * namespace of the second process identified by @pid2. If the two processes are | |
166 | * in the same namespace returns -EINVAL, -1 if an error occurred. | |
167 | */ | |
168 | static int in_same_namespace(pid_t pid1, pid_t pid2, const char *ns) | |
169 | { | |
170 | int ns_fd1 = -1, ns_fd2 = -1, ret = -1; | |
21d0acc2 | 171 | int saved_errno; |
299d1198 CB |
172 | struct stat ns_st1, ns_st2; |
173 | ||
174 | ns_fd1 = lxc_preserve_ns(pid1, ns); | |
134284c3 CB |
175 | if (ns_fd1 < 0) { |
176 | /* The kernel does not support this namespace. This is not an | |
177 | * error. | |
178 | */ | |
179 | if (errno == ENOENT) | |
180 | return -EINVAL; | |
181 | ||
299d1198 | 182 | goto out; |
134284c3 | 183 | } |
299d1198 CB |
184 | |
185 | ns_fd2 = lxc_preserve_ns(pid2, ns); | |
21d0acc2 | 186 | if (ns_fd2 < 0) |
299d1198 CB |
187 | goto out; |
188 | ||
189 | ret = fstat(ns_fd1, &ns_st1); | |
21d0acc2 | 190 | if (ret < 0) |
299d1198 CB |
191 | goto out; |
192 | ||
193 | ret = fstat(ns_fd2, &ns_st2); | |
21d0acc2 | 194 | if (ret < 0) |
299d1198 CB |
195 | goto out; |
196 | ||
197 | /* processes are in the same namespace */ | |
ea918412 | 198 | if ((ns_st1.st_dev == ns_st2.st_dev ) && (ns_st1.st_ino == ns_st2.st_ino)) { |
199 | ret = -EINVAL; | |
299d1198 | 200 | goto out; |
ea918412 | 201 | } |
299d1198 CB |
202 | |
203 | /* processes are in different namespaces */ | |
204 | ret = ns_fd2; | |
205 | ns_fd2 = -1; | |
206 | ||
207 | out: | |
21d0acc2 | 208 | saved_errno = errno; |
209 | ||
299d1198 CB |
210 | if (ns_fd1 >= 0) |
211 | close(ns_fd1); | |
ea918412 | 212 | |
299d1198 CB |
213 | if (ns_fd2 >= 0) |
214 | close(ns_fd2); | |
215 | ||
ea918412 | 216 | errno = saved_errno; |
299d1198 CB |
217 | return ret; |
218 | } | |
219 | ||
877f3a04 | 220 | static int lxc_attach_to_ns(pid_t pid, struct lxc_proc_context_info *ctx) |
99d50954 | 221 | { |
21d0acc2 | 222 | int i, ret; |
99d50954 | 223 | |
26818618 | 224 | for (i = 0; i < LXC_NS_MAX; i++) { |
877f3a04 | 225 | if (ctx->ns_fd[i] < 0) |
26818618 CB |
226 | continue; |
227 | ||
21d0acc2 | 228 | ret = setns(ctx->ns_fd[i], ns_info[i].clone_flag); |
229 | if (ret < 0) { | |
299d1198 | 230 | SYSERROR("Failed to attach to %s namespace of %d", |
ea918412 | 231 | ns_info[i].proc_name, pid); |
99d50954 CS |
232 | return -1; |
233 | } | |
234 | ||
299d1198 | 235 | DEBUG("Attached to %s namespace of %d", ns_info[i].proc_name, pid); |
99d50954 CS |
236 | } |
237 | ||
238 | return 0; | |
239 | } | |
240 | ||
74a3920a | 241 | static int lxc_attach_remount_sys_proc(void) |
7a0b0b56 CS |
242 | { |
243 | int ret; | |
244 | ||
245 | ret = unshare(CLONE_NEWNS); | |
246 | if (ret < 0) { | |
ea918412 | 247 | SYSERROR("Failed to unshare mount namespace"); |
7a0b0b56 CS |
248 | return -1; |
249 | } | |
250 | ||
2c6f3fc9 | 251 | if (detect_shared_rootfs()) { |
6f4f1937 | 252 | if (mount(NULL, "/", NULL, MS_SLAVE | MS_REC, NULL)) { |
ea918412 | 253 | SYSERROR("Failed to make / rslave"); |
2c6f3fc9 SH |
254 | ERROR("Continuing..."); |
255 | } | |
256 | } | |
257 | ||
8ce83369 | 258 | /* Assume /proc is always mounted, so remount it. */ |
7a0b0b56 CS |
259 | ret = umount2("/proc", MNT_DETACH); |
260 | if (ret < 0) { | |
ea918412 | 261 | SYSERROR("Failed to unmount /proc"); |
7a0b0b56 CS |
262 | return -1; |
263 | } | |
264 | ||
265 | ret = mount("none", "/proc", "proc", 0, NULL); | |
266 | if (ret < 0) { | |
ea918412 | 267 | SYSERROR("Failed to remount /proc"); |
7a0b0b56 CS |
268 | return -1; |
269 | } | |
270 | ||
8ce83369 CB |
271 | /* Try to umount /sys. If it's not a mount point, we'll get EINVAL, then |
272 | * we ignore it because it may not have been mounted in the first place. | |
7a0b0b56 CS |
273 | */ |
274 | ret = umount2("/sys", MNT_DETACH); | |
275 | if (ret < 0 && errno != EINVAL) { | |
ea918412 | 276 | SYSERROR("Failed to unmount /sys"); |
7a0b0b56 CS |
277 | return -1; |
278 | } else if (ret == 0) { | |
8ce83369 | 279 | /* Remount it. */ |
7a0b0b56 CS |
280 | ret = mount("none", "/sys", "sysfs", 0, NULL); |
281 | if (ret < 0) { | |
ea918412 | 282 | SYSERROR("Failed to remount /sys"); |
7a0b0b56 CS |
283 | return -1; |
284 | } | |
285 | } | |
286 | ||
287 | return 0; | |
288 | } | |
289 | ||
74a3920a | 290 | static int lxc_attach_drop_privs(struct lxc_proc_context_info *ctx) |
e0732705 | 291 | { |
6f4f1937 | 292 | int cap, last_cap; |
e0732705 | 293 | |
6f4f1937 | 294 | last_cap = lxc_caps_last_cap(); |
e0732705 CS |
295 | for (cap = 0; cap <= last_cap; cap++) { |
296 | if (ctx->capability_mask & (1LL << cap)) | |
297 | continue; | |
298 | ||
b81689a1 CB |
299 | if (prctl(PR_CAPBSET_DROP, prctl_arg(cap), prctl_arg(0), |
300 | prctl_arg(0), prctl_arg(0))) { | |
94ac256f | 301 | SYSERROR("Failed to drop capability %d", cap); |
e0732705 CS |
302 | return -1; |
303 | } | |
ea918412 | 304 | |
94ac256f | 305 | TRACE("Dropped capability %d", cap); |
e0732705 CS |
306 | } |
307 | ||
308 | return 0; | |
309 | } | |
905022f7 | 310 | |
7385273f | 311 | static int lxc_attach_set_environment(struct lxc_proc_context_info *init_ctx, |
312 | enum lxc_attach_env_policy_t policy, | |
6f4f1937 | 313 | char **extra_env, char **extra_keep) |
b3a39ba6 | 314 | { |
3d55242a | 315 | int ret; |
7385273f | 316 | struct lxc_list *iterator; |
317 | ||
799f96fd | 318 | if (policy == LXC_ATTACH_CLEAR_ENV) { |
3d5e9f48 | 319 | int path_kept = 0; |
6f4f1937 | 320 | char **extra_keep_store = NULL; |
3d5e9f48 CS |
321 | |
322 | if (extra_keep) { | |
323 | size_t count, i; | |
324 | ||
3d55242a CB |
325 | for (count = 0; extra_keep[count]; count++) |
326 | ; | |
3d5e9f48 CS |
327 | |
328 | extra_keep_store = calloc(count, sizeof(char *)); | |
3d55242a | 329 | if (!extra_keep_store) |
3d5e9f48 | 330 | return -1; |
3d55242a | 331 | |
3d5e9f48 CS |
332 | for (i = 0; i < count; i++) { |
333 | char *v = getenv(extra_keep[i]); | |
334 | if (v) { | |
335 | extra_keep_store[i] = strdup(v); | |
336 | if (!extra_keep_store[i]) { | |
3d5e9f48 CS |
337 | while (i > 0) |
338 | free(extra_keep_store[--i]); | |
ea918412 | 339 | |
3d5e9f48 CS |
340 | free(extra_keep_store); |
341 | return -1; | |
342 | } | |
3d55242a | 343 | |
3d5e9f48 CS |
344 | if (strcmp(extra_keep[i], "PATH") == 0) |
345 | path_kept = 1; | |
346 | } | |
3d5e9f48 CS |
347 | } |
348 | } | |
349 | ||
799f96fd | 350 | if (clearenv()) { |
a9cab7e3 | 351 | if (extra_keep_store) { |
3d55242a CB |
352 | char **p; |
353 | ||
a9cab7e3 CS |
354 | for (p = extra_keep_store; *p; p++) |
355 | free(*p); | |
3d55242a | 356 | |
a9cab7e3 CS |
357 | free(extra_keep_store); |
358 | } | |
3d55242a | 359 | |
ea918412 | 360 | ERROR("Failed to clear environment"); |
3d5e9f48 CS |
361 | return -1; |
362 | } | |
363 | ||
364 | if (extra_keep_store) { | |
365 | size_t i; | |
6f4f1937 | 366 | |
3d5e9f48 | 367 | for (i = 0; extra_keep[i]; i++) { |
acd4922e | 368 | if (extra_keep_store[i]) { |
3d55242a CB |
369 | ret = setenv(extra_keep[i], extra_keep_store[i], 1); |
370 | if (ret < 0) | |
a24c5678 | 371 | SYSWARN("Failed to set environment variable"); |
acd4922e | 372 | } |
ea918412 | 373 | |
3d5e9f48 CS |
374 | free(extra_keep_store[i]); |
375 | } | |
ea918412 | 376 | |
3d5e9f48 CS |
377 | free(extra_keep_store); |
378 | } | |
379 | ||
8ce83369 CB |
380 | /* Always set a default path; shells and execlp tend to be fine |
381 | * without it, but there is a disturbing number of C programs | |
382 | * out there that just assume that getenv("PATH") is never NULL | |
383 | * and then die a painful segfault death. | |
384 | */ | |
3d55242a CB |
385 | if (!path_kept) { |
386 | ret = setenv("PATH", "/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", 1); | |
387 | if (ret < 0) | |
a24c5678 | 388 | SYSWARN("Failed to set environment variable"); |
3d55242a | 389 | } |
b3a39ba6 DW |
390 | } |
391 | ||
3d55242a CB |
392 | ret = putenv("container=lxc"); |
393 | if (ret < 0) { | |
a24c5678 | 394 | SYSWARN("Failed to set environment variable"); |
b3a39ba6 DW |
395 | return -1; |
396 | } | |
397 | ||
7385273f | 398 | /* Set container environment variables.*/ |
399 | if (init_ctx && init_ctx->container && init_ctx->container->lxc_conf) { | |
400 | lxc_list_for_each(iterator, &init_ctx->container->lxc_conf->environment) { | |
3d55242a CB |
401 | char *env_tmp; |
402 | ||
403 | env_tmp = strdup((char *)iterator->elem); | |
404 | if (!env_tmp) | |
7385273f | 405 | return -1; |
7385273f | 406 | |
3d55242a CB |
407 | ret = putenv(env_tmp); |
408 | if (ret < 0) { | |
409 | SYSERROR("Failed to set environment variable: %s", (char *)iterator->elem); | |
7385273f | 410 | return -1; |
411 | } | |
412 | } | |
413 | } | |
414 | ||
8ce83369 | 415 | /* Set extra environment variables. */ |
3d5e9f48 CS |
416 | if (extra_env) { |
417 | for (; *extra_env; extra_env++) { | |
3d55242a | 418 | char *p; |
ea918412 | 419 | |
8ce83369 CB |
420 | /* We just assume the user knows what they are doing, so |
421 | * we don't do any checks. | |
422 | */ | |
3d55242a CB |
423 | p = strdup(*extra_env); |
424 | if (!p) | |
3d5e9f48 | 425 | return -1; |
3d55242a CB |
426 | |
427 | ret = putenv(p); | |
428 | if (ret < 0) | |
a24c5678 | 429 | SYSWARN("Failed to set environment variable"); |
3d5e9f48 CS |
430 | } |
431 | } | |
432 | ||
b3a39ba6 DW |
433 | return 0; |
434 | } | |
435 | ||
74a3920a | 436 | static char *lxc_attach_getpwshell(uid_t uid) |
905022f7 | 437 | { |
6f4f1937 | 438 | int fd, ret; |
905022f7 CS |
439 | pid_t pid; |
440 | int pipes[2]; | |
3fa23ac3 CB |
441 | FILE *pipe_f; |
442 | bool found = false; | |
443 | size_t line_bufsz = 0; | |
444 | char *line = NULL, *result = NULL; | |
905022f7 | 445 | |
8ce83369 CB |
446 | /* We need to fork off a process that runs the getent program, and we |
447 | * need to capture its output, so we use a pipe for that purpose. | |
905022f7 | 448 | */ |
3fa23ac3 | 449 | ret = pipe2(pipes, O_CLOEXEC); |
905022f7 CS |
450 | if (ret < 0) |
451 | return NULL; | |
452 | ||
453 | pid = fork(); | |
454 | if (pid < 0) { | |
455 | close(pipes[0]); | |
456 | close(pipes[1]); | |
457 | return NULL; | |
458 | } | |
459 | ||
3fa23ac3 | 460 | if (!pid) { |
905022f7 CS |
461 | char uid_buf[32]; |
462 | char *arguments[] = { | |
463 | "getent", | |
464 | "passwd", | |
465 | uid_buf, | |
466 | NULL | |
467 | }; | |
468 | ||
469 | close(pipes[0]); | |
470 | ||
8ce83369 | 471 | /* We want to capture stdout. */ |
3fa23ac3 | 472 | ret = dup2(pipes[1], STDOUT_FILENO); |
905022f7 | 473 | close(pipes[1]); |
3fa23ac3 | 474 | if (ret < 0) |
ea918412 | 475 | _exit(EXIT_FAILURE); |
905022f7 | 476 | |
8ce83369 CB |
477 | /* Get rid of stdin/stderr, so we try to associate it with |
478 | * /dev/null. | |
905022f7 | 479 | */ |
3fa23ac3 | 480 | fd = open_devnull(); |
905022f7 | 481 | if (fd < 0) { |
3fa23ac3 CB |
482 | close(STDIN_FILENO); |
483 | close(STDERR_FILENO); | |
905022f7 | 484 | } else { |
3fa23ac3 CB |
485 | (void)dup3(fd, STDIN_FILENO, O_CLOEXEC); |
486 | (void)dup3(fd, STDOUT_FILENO, O_CLOEXEC); | |
905022f7 CS |
487 | close(fd); |
488 | } | |
489 | ||
8ce83369 | 490 | /* Finish argument list. */ |
3fa23ac3 CB |
491 | ret = snprintf(uid_buf, sizeof(uid_buf), "%ld", (long)uid); |
492 | if (ret <= 0 || ret >= sizeof(uid_buf)) | |
ea918412 | 493 | _exit(EXIT_FAILURE); |
905022f7 | 494 | |
8ce83369 | 495 | /* Try to run getent program. */ |
3fa23ac3 | 496 | (void)execvp("getent", arguments); |
ea918412 | 497 | _exit(EXIT_FAILURE); |
905022f7 | 498 | } |
3fa23ac3 CB |
499 | |
500 | close(pipes[1]); | |
501 | ||
502 | pipe_f = fdopen(pipes[0], "r"); | |
503 | while (getline(&line, &line_bufsz, pipe_f) != -1) { | |
504 | int i; | |
505 | long value; | |
506 | char *token; | |
507 | char *endptr = NULL, *saveptr = NULL; | |
508 | ||
509 | /* If we already found something, just continue to read | |
510 | * until the pipe doesn't deliver any more data, but | |
511 | * don't modify the existing data structure. | |
512 | */ | |
513 | if (found) | |
514 | continue; | |
515 | ||
18d4ffde | 516 | if (!line) |
517 | continue; | |
518 | ||
3fa23ac3 CB |
519 | /* Trim line on the right hand side. */ |
520 | for (i = strlen(line); i > 0 && (line[i - 1] == '\n' || line[i - 1] == '\r'); --i) | |
521 | line[i - 1] = '\0'; | |
522 | ||
523 | /* Split into tokens: first: user name. */ | |
524 | token = strtok_r(line, ":", &saveptr); | |
525 | if (!token) | |
526 | continue; | |
527 | ||
528 | /* next: dummy password field */ | |
529 | token = strtok_r(NULL, ":", &saveptr); | |
530 | if (!token) | |
531 | continue; | |
532 | ||
533 | /* next: user id */ | |
534 | token = strtok_r(NULL, ":", &saveptr); | |
535 | value = token ? strtol(token, &endptr, 10) : 0; | |
536 | if (!token || !endptr || *endptr || value == LONG_MIN || | |
ea918412 | 537 | value == LONG_MAX) |
3fa23ac3 CB |
538 | continue; |
539 | ||
540 | /* dummy sanity check: user id matches */ | |
541 | if ((uid_t)value != uid) | |
542 | continue; | |
543 | ||
544 | /* skip fields: gid, gecos, dir, go to next field 'shell' */ | |
545 | for (i = 0; i < 4; i++) { | |
546 | token = strtok_r(NULL, ":", &saveptr); | |
547 | if (!token) | |
548 | continue; | |
549 | } | |
ea918412 | 550 | |
3fa23ac3 CB |
551 | if (!token) |
552 | continue; | |
ea918412 | 553 | |
3fa23ac3 CB |
554 | free(result); |
555 | result = strdup(token); | |
556 | ||
557 | /* Sanity check that there are no fields after that. */ | |
558 | token = strtok_r(NULL, ":", &saveptr); | |
559 | if (token) | |
560 | continue; | |
561 | ||
562 | found = true; | |
563 | } | |
ea918412 | 564 | |
3fa23ac3 CB |
565 | free(line); |
566 | fclose(pipe_f); | |
567 | ||
568 | ret = wait_for_pid(pid); | |
569 | if (ret < 0) { | |
570 | free(result); | |
571 | return NULL; | |
572 | } | |
573 | ||
574 | if (!found) { | |
575 | free(result); | |
576 | return NULL; | |
577 | } | |
578 | ||
579 | return result; | |
905022f7 | 580 | } |
cb3e61fa | 581 | |
6f4f1937 | 582 | static void lxc_attach_get_init_uidgid(uid_t *init_uid, gid_t *init_gid) |
cb3e61fa CS |
583 | { |
584 | FILE *proc_file; | |
604ca1c0 | 585 | char proc_fn[LXC_PROC_STATUS_LEN]; |
8ce83369 | 586 | int ret; |
cb3e61fa CS |
587 | char *line = NULL; |
588 | size_t line_bufsz = 0; | |
cb3e61fa CS |
589 | long value = -1; |
590 | uid_t uid = (uid_t)-1; | |
591 | gid_t gid = (gid_t)-1; | |
592 | ||
604ca1c0 CB |
593 | ret = snprintf(proc_fn, LXC_PROC_STATUS_LEN, "/proc/%d/status", 1); |
594 | if (ret < 0 || ret >= LXC_PROC_STATUS_LEN) | |
7fb45c93 | 595 | return; |
cb3e61fa CS |
596 | |
597 | proc_file = fopen(proc_fn, "r"); | |
598 | if (!proc_file) | |
599 | return; | |
600 | ||
601 | while (getline(&line, &line_bufsz, proc_file) != -1) { | |
8ce83369 CB |
602 | /* Format is: real, effective, saved set user, fs we only care |
603 | * about real uid. | |
cb3e61fa CS |
604 | */ |
605 | ret = sscanf(line, "Uid: %ld", &value); | |
8ce83369 | 606 | if (ret != EOF && ret == 1) { |
6f4f1937 | 607 | uid = (uid_t)value; |
cb3e61fa CS |
608 | } else { |
609 | ret = sscanf(line, "Gid: %ld", &value); | |
8ce83369 | 610 | if (ret != EOF && ret == 1) |
6f4f1937 | 611 | gid = (gid_t)value; |
cb3e61fa | 612 | } |
ea918412 | 613 | |
cb3e61fa CS |
614 | if (uid != (uid_t)-1 && gid != (gid_t)-1) |
615 | break; | |
616 | } | |
617 | ||
618 | fclose(proc_file); | |
619 | free(line); | |
620 | ||
8ce83369 | 621 | /* Only override arguments if we found something. */ |
cb3e61fa CS |
622 | if (uid != (uid_t)-1) |
623 | *init_uid = uid; | |
ea918412 | 624 | |
cb3e61fa CS |
625 | if (gid != (gid_t)-1) |
626 | *init_gid = gid; | |
627 | ||
628 | /* TODO: we should also parse supplementary groups and use | |
8ce83369 CB |
629 | * setgroups() to set them. |
630 | */ | |
cb3e61fa | 631 | } |
9c4693b8 | 632 | |
d4db3d14 | 633 | static bool fetch_seccomp(struct lxc_container *c, lxc_attach_options_t *options) |
2c4ea790 | 634 | { |
d4db3d14 CB |
635 | int ret; |
636 | bool bret; | |
bd7b4e28 | 637 | char *path; |
2eef2bda | 638 | |
6f4f1937 CB |
639 | if (!(options->namespaces & CLONE_NEWNS) || |
640 | !(options->attach_flags & LXC_ATTACH_LSM)) { | |
bd4307f0 CB |
641 | free(c->lxc_conf->seccomp); |
642 | c->lxc_conf->seccomp = NULL; | |
2c4ea790 | 643 | return true; |
bd4307f0 | 644 | } |
bd7b4e28 | 645 | |
2e812c16 | 646 | /* Remove current setting. */ |
d4db3d14 | 647 | if (!c->set_config_item(c, "lxc.seccomp.profile", "") && |
ea918412 | 648 | !c->set_config_item(c, "lxc.seccomp", "")) |
2c4ea790 | 649 | return false; |
bd7b4e28 | 650 | |
8ce83369 | 651 | /* Fetch the current profile path over the cmd interface. */ |
0b427da0 | 652 | path = c->get_running_config_item(c, "lxc.seccomp.profile"); |
bd7b4e28 | 653 | if (!path) { |
d4db3d14 | 654 | INFO("Failed to retrieve lxc.seccomp.profile"); |
ea918412 | 655 | |
0b427da0 | 656 | path = c->get_running_config_item(c, "lxc.seccomp"); |
d4db3d14 CB |
657 | if (!path) { |
658 | INFO("Failed to retrieve lxc.seccomp"); | |
659 | return true; | |
660 | } | |
bd7b4e28 SG |
661 | } |
662 | ||
8ce83369 | 663 | /* Copy the value into the new lxc_conf. */ |
d4db3d14 | 664 | bret = c->set_config_item(c, "lxc.seccomp.profile", path); |
bd7b4e28 | 665 | free(path); |
d4db3d14 CB |
666 | if (!bret) |
667 | return false; | |
bd7b4e28 | 668 | |
8ce83369 | 669 | /* Attempt to parse the resulting config. */ |
d4db3d14 CB |
670 | ret = lxc_read_seccomp_config(c->lxc_conf); |
671 | if (ret < 0) { | |
672 | ERROR("Failed to retrieve seccomp policy"); | |
2c4ea790 SH |
673 | return false; |
674 | } | |
675 | ||
d4db3d14 | 676 | INFO("Retrieved seccomp policy"); |
2e812c16 CB |
677 | return true; |
678 | } | |
679 | ||
6f4f1937 | 680 | static bool no_new_privs(struct lxc_container *c, lxc_attach_options_t *options) |
2e812c16 | 681 | { |
bcbef733 | 682 | bool bret; |
2e812c16 CB |
683 | char *val; |
684 | ||
2e812c16 | 685 | /* Remove current setting. */ |
bcbef733 CB |
686 | if (!c->set_config_item(c, "lxc.no_new_privs", "")) { |
687 | INFO("Failed to unset lxc.no_new_privs"); | |
2e812c16 | 688 | return false; |
bcbef733 | 689 | } |
2e812c16 CB |
690 | |
691 | /* Retrieve currently active setting. */ | |
692 | val = c->get_running_config_item(c, "lxc.no_new_privs"); | |
693 | if (!val) { | |
bcbef733 | 694 | INFO("Failed to retrieve lxc.no_new_privs"); |
2e812c16 CB |
695 | return false; |
696 | } | |
697 | ||
698 | /* Set currently active setting. */ | |
bcbef733 | 699 | bret = c->set_config_item(c, "lxc.no_new_privs", val); |
2e812c16 | 700 | free(val); |
bcbef733 | 701 | return bret; |
2c4ea790 SH |
702 | } |
703 | ||
9b8e3c96 SH |
704 | static signed long get_personality(const char *name, const char *lxcpath) |
705 | { | |
6f4f1937 | 706 | char *p; |
9b8e3c96 SH |
707 | signed long ret; |
708 | ||
6f4f1937 | 709 | p = lxc_cmd_get_config_item(name, "lxc.arch", lxcpath); |
9b8e3c96 SH |
710 | if (!p) |
711 | return -1; | |
6f4f1937 | 712 | |
9b8e3c96 SH |
713 | ret = lxc_config_parse_arch(p); |
714 | free(p); | |
6f4f1937 | 715 | |
9b8e3c96 SH |
716 | return ret; |
717 | } | |
718 | ||
a998454a CB |
719 | struct attach_clone_payload { |
720 | int ipc_socket; | |
9e84479f | 721 | int terminal_slave_fd; |
a998454a CB |
722 | lxc_attach_options_t *options; |
723 | struct lxc_proc_context_info *init_ctx; | |
724 | lxc_attach_exec_t exec_function; | |
725 | void *exec_payload; | |
726 | }; | |
727 | ||
ba2be1a8 CB |
728 | static void lxc_put_attach_clone_payload(struct attach_clone_payload *p) |
729 | { | |
730 | if (p->ipc_socket >= 0) { | |
ba2be1a8 CB |
731 | close(p->ipc_socket); |
732 | p->ipc_socket = -EBADF; | |
733 | } | |
734 | ||
9e84479f CB |
735 | if (p->terminal_slave_fd >= 0) { |
736 | close(p->terminal_slave_fd); | |
737 | p->terminal_slave_fd = -EBADF; | |
ba2be1a8 CB |
738 | } |
739 | ||
b21da190 | 740 | if (p->init_ctx) { |
ba2be1a8 | 741 | lxc_proc_put_context_info(p->init_ctx); |
b21da190 CB |
742 | p->init_ctx = NULL; |
743 | } | |
ba2be1a8 CB |
744 | } |
745 | ||
a998454a CB |
746 | static int attach_child_main(struct attach_clone_payload *payload) |
747 | { | |
57de839f | 748 | int fd, lsm_fd, ret; |
a998454a CB |
749 | uid_t new_uid; |
750 | gid_t new_gid; | |
936efc72 CB |
751 | uid_t ns_root_uid = 0; |
752 | gid_t ns_root_gid = 0; | |
a998454a CB |
753 | lxc_attach_options_t* options = payload->options; |
754 | struct lxc_proc_context_info* init_ctx = payload->init_ctx; | |
57de839f CB |
755 | bool needs_lsm = (options->namespaces & CLONE_NEWNS) && |
756 | (options->attach_flags & LXC_ATTACH_LSM) && | |
757 | init_ctx->lsm_label; | |
a998454a CB |
758 | |
759 | /* A description of the purpose of this functionality is provided in the | |
760 | * lxc-attach(1) manual page. We have to remount here and not in the | |
761 | * parent process, otherwise /proc may not properly reflect the new pid | |
762 | * namespace. | |
763 | */ | |
764 | if (!(options->namespaces & CLONE_NEWNS) && | |
765 | (options->attach_flags & LXC_ATTACH_REMOUNT_PROC_SYS)) { | |
766 | ret = lxc_attach_remount_sys_proc(); | |
b75c344c CB |
767 | if (ret < 0) |
768 | goto on_error; | |
ea918412 | 769 | |
b75c344c | 770 | TRACE("Remounted \"/proc\" and \"/sys\""); |
a998454a CB |
771 | } |
772 | ||
b75c344c | 773 | /* Now perform additional attachments. */ |
a998454a | 774 | #if HAVE_SYS_PERSONALITY_H |
a998454a | 775 | if (options->attach_flags & LXC_ATTACH_SET_PERSONALITY) { |
b75c344c CB |
776 | long new_personality; |
777 | ||
778 | if (options->personality < 0) | |
779 | new_personality = init_ctx->personality; | |
780 | else | |
781 | new_personality = options->personality; | |
ea918412 | 782 | |
a998454a | 783 | ret = personality(new_personality); |
b75c344c CB |
784 | if (ret < 0) |
785 | goto on_error; | |
ea918412 | 786 | |
b75c344c | 787 | TRACE("Set new personality"); |
a998454a CB |
788 | } |
789 | #endif | |
790 | ||
791 | if (options->attach_flags & LXC_ATTACH_DROP_CAPABILITIES) { | |
792 | ret = lxc_attach_drop_privs(init_ctx); | |
b75c344c CB |
793 | if (ret < 0) |
794 | goto on_error; | |
ea918412 | 795 | |
b75c344c | 796 | TRACE("Dropped capabilities"); |
a998454a CB |
797 | } |
798 | ||
799 | /* Always set the environment (specify (LXC_ATTACH_KEEP_ENV, NULL, NULL) | |
800 | * if you want this to be a no-op). | |
801 | */ | |
7385273f | 802 | ret = lxc_attach_set_environment(init_ctx, |
803 | options->env_policy, | |
a998454a CB |
804 | options->extra_env_vars, |
805 | options->extra_keep_env); | |
b75c344c CB |
806 | if (ret < 0) |
807 | goto on_error; | |
ea918412 | 808 | |
b75c344c | 809 | TRACE("Set up environment"); |
a998454a | 810 | |
57de839f CB |
811 | /* This remark only affects fully unprivileged containers: |
812 | * Receive fd for LSM security module before we set{g,u}id(). The reason | |
813 | * is that on set{g,u}id() the kernel will a) make us undumpable and b) | |
814 | * we will change our effective uid. This means our effective uid will | |
815 | * be different from the effective uid of the process that created us | |
816 | * which means that this processs no longer has capabilities in our | |
817 | * namespace including CAP_SYS_PTRACE. This means we will not be able to | |
818 | * read and /proc/<pid> files for the process anymore when /proc is | |
819 | * mounted with hidepid={1,2}. So let's get the lsm label fd before the | |
820 | * set{g,u}id(). | |
821 | */ | |
822 | if (needs_lsm) { | |
b75c344c | 823 | ret = lxc_abstract_unix_recv_fds(payload->ipc_socket, &lsm_fd, 1, NULL, 0); |
9044b79e | 824 | if (ret <= 0) { |
825 | if (ret < 0) | |
826 | SYSERROR("Failed to receive lsm label fd"); | |
827 | ||
b75c344c | 828 | goto on_error; |
9044b79e | 829 | } |
830 | ||
57de839f CB |
831 | TRACE("Received LSM label file descriptor %d from parent", lsm_fd); |
832 | } | |
833 | ||
08ea9270 | 834 | if (options->stdin_fd > 0 && isatty(options->stdin_fd)) { |
cd0a2b2f | 835 | ret = lxc_make_controlling_terminal(options->stdin_fd); |
08ea9270 CB |
836 | if (ret < 0) |
837 | goto on_error; | |
838 | } | |
839 | ||
936efc72 CB |
840 | if (options->namespaces & CLONE_NEWUSER) { |
841 | /* Check whether nsuid 0 has a mapping. */ | |
842 | ns_root_uid = get_ns_uid(0); | |
ea918412 | 843 | |
936efc72 CB |
844 | /* Check whether nsgid 0 has a mapping. */ |
845 | ns_root_gid = get_ns_gid(0); | |
a998454a | 846 | |
936efc72 CB |
847 | /* If there's no mapping for nsuid 0 try to retrieve the nsuid |
848 | * init was started with. | |
849 | */ | |
850 | if (ns_root_uid == LXC_INVALID_UID) | |
851 | lxc_attach_get_init_uidgid(&ns_root_uid, &ns_root_gid); | |
ea918412 | 852 | |
936efc72 CB |
853 | if (ns_root_uid == LXC_INVALID_UID) |
854 | goto on_error; | |
a998454a | 855 | |
464c4611 | 856 | if (!lxc_switch_uid_gid(ns_root_uid, ns_root_gid)) |
b75c344c | 857 | goto on_error; |
a998454a CB |
858 | } |
859 | ||
8af07f82 | 860 | if (!lxc_setgroups(0, NULL) && errno != EPERM) |
24927339 CB |
861 | goto on_error; |
862 | ||
936efc72 CB |
863 | /* Set {u,g}id. */ |
864 | if (options->uid != LXC_INVALID_UID) | |
865 | new_uid = options->uid; | |
866 | else | |
867 | new_uid = ns_root_uid; | |
868 | ||
869 | if (options->gid != LXC_INVALID_GID) | |
870 | new_gid = options->gid; | |
871 | else | |
872 | new_gid = ns_root_gid; | |
873 | ||
a998454a CB |
874 | if ((init_ctx->container && init_ctx->container->lxc_conf && |
875 | init_ctx->container->lxc_conf->no_new_privs) || | |
876 | (options->attach_flags & LXC_ATTACH_NO_NEW_PRIVS)) { | |
b81689a1 CB |
877 | ret = prctl(PR_SET_NO_NEW_PRIVS, prctl_arg(1), prctl_arg(0), |
878 | prctl_arg(0), prctl_arg(0)); | |
b75c344c CB |
879 | if (ret < 0) |
880 | goto on_error; | |
ea918412 | 881 | |
b75c344c | 882 | TRACE("Set PR_SET_NO_NEW_PRIVS"); |
a998454a CB |
883 | } |
884 | ||
57de839f | 885 | if (needs_lsm) { |
d3ba7c98 | 886 | bool on_exec; |
a998454a CB |
887 | |
888 | /* Change into our new LSM profile. */ | |
d3ba7c98 | 889 | on_exec = options->attach_flags & LXC_ATTACH_LSM_EXEC ? true : false; |
ea918412 | 890 | |
d3ba7c98 | 891 | ret = lsm_process_label_set_at(lsm_fd, init_ctx->lsm_label, on_exec); |
57de839f | 892 | close(lsm_fd); |
b75c344c CB |
893 | if (ret < 0) |
894 | goto on_error; | |
ea918412 | 895 | |
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; | |
ea918412 | 904 | |
b75c344c | 905 | TRACE("Loaded seccomp profile"); |
a998454a | 906 | } |
ea918412 | 907 | |
b75c344c | 908 | close(payload->ipc_socket); |
ba2be1a8 CB |
909 | payload->ipc_socket = -EBADF; |
910 | lxc_proc_put_context_info(init_ctx); | |
d35b372a | 911 | payload->init_ctx = NULL; |
a998454a CB |
912 | |
913 | /* The following is done after the communication socket is shut down. | |
914 | * That way, all errors that might (though unlikely) occur up until this | |
915 | * point will have their messages printed to the original stderr (if | |
916 | * logging is so configured) and not the fd the user supplied, if any. | |
917 | */ | |
918 | ||
919 | /* Fd handling for stdin, stdout and stderr; ignore errors here, user | |
920 | * may want to make sure the fds are closed, for example. | |
921 | */ | |
08ea9270 | 922 | if (options->stdin_fd >= 0 && options->stdin_fd != STDIN_FILENO) |
00c72a93 | 923 | (void)dup2(options->stdin_fd, STDIN_FILENO); |
08ea9270 CB |
924 | |
925 | if (options->stdout_fd >= 0 && options->stdout_fd != STDOUT_FILENO) | |
00c72a93 | 926 | (void)dup2(options->stdout_fd, STDOUT_FILENO); |
08ea9270 CB |
927 | |
928 | if (options->stderr_fd >= 0 && options->stderr_fd != STDERR_FILENO) | |
00c72a93 | 929 | (void)dup2(options->stderr_fd, STDERR_FILENO); |
a998454a CB |
930 | |
931 | /* close the old fds */ | |
08ea9270 | 932 | if (options->stdin_fd > STDERR_FILENO) |
a998454a | 933 | close(options->stdin_fd); |
08ea9270 CB |
934 | |
935 | if (options->stdout_fd > STDERR_FILENO) | |
a998454a | 936 | close(options->stdout_fd); |
08ea9270 CB |
937 | |
938 | if (options->stderr_fd > STDERR_FILENO) | |
a998454a CB |
939 | close(options->stderr_fd); |
940 | ||
941 | /* Try to remove FD_CLOEXEC flag from stdin/stdout/stderr, but also | |
942 | * here, ignore errors. | |
943 | */ | |
b75c344c | 944 | for (fd = STDIN_FILENO; fd <= STDERR_FILENO; fd++) { |
3f62938a | 945 | ret = fd_cloexec(fd, false); |
b75c344c CB |
946 | if (ret < 0) { |
947 | SYSERROR("Failed to clear FD_CLOEXEC from file descriptor %d", fd); | |
948 | goto on_error; | |
949 | } | |
a998454a CB |
950 | } |
951 | ||
9e84479f CB |
952 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
953 | ret = lxc_terminal_prepare_login(payload->terminal_slave_fd); | |
ba2be1a8 | 954 | if (ret < 0) { |
9e84479f | 955 | SYSERROR("Failed to prepare terminal file descriptor %d", payload->terminal_slave_fd); |
ba2be1a8 CB |
956 | goto on_error; |
957 | } | |
ea918412 | 958 | |
9e84479f | 959 | TRACE("Prepared terminal file descriptor %d", payload->terminal_slave_fd); |
ba2be1a8 CB |
960 | } |
961 | ||
936efc72 CB |
962 | /* Avoid unnecessary syscalls. */ |
963 | if (new_uid == ns_root_uid) | |
964 | new_uid = LXC_INVALID_UID; | |
965 | ||
966 | if (new_gid == ns_root_gid) | |
967 | new_gid = LXC_INVALID_GID; | |
968 | ||
464c4611 | 969 | if (!lxc_switch_uid_gid(new_uid, new_gid)) |
936efc72 CB |
970 | goto on_error; |
971 | ||
a998454a | 972 | /* We're done, so we can now do whatever the user intended us to do. */ |
c7ac2e1c | 973 | _exit(payload->exec_function(payload->exec_payload)); |
b75c344c CB |
974 | |
975 | on_error: | |
ba2be1a8 | 976 | lxc_put_attach_clone_payload(payload); |
c7ac2e1c | 977 | _exit(EXIT_FAILURE); |
a998454a CB |
978 | } |
979 | ||
9e84479f CB |
980 | static int lxc_attach_terminal(struct lxc_conf *conf, |
981 | struct lxc_terminal *terminal) | |
ba2be1a8 CB |
982 | { |
983 | int ret; | |
984 | ||
9e84479f | 985 | lxc_terminal_init(terminal); |
ba2be1a8 | 986 | |
9e84479f | 987 | ret = lxc_terminal_create(terminal); |
ba2be1a8 | 988 | if (ret < 0) { |
ea918412 | 989 | ERROR("Failed to create terminal"); |
ba2be1a8 CB |
990 | return -1; |
991 | } | |
992 | ||
993 | /* Shift ttys to container. */ | |
9e84479f | 994 | ret = lxc_terminal_map_ids(conf, terminal); |
ba2be1a8 | 995 | if (ret < 0) { |
9e84479f | 996 | ERROR("Failed to chown terminal"); |
ba2be1a8 CB |
997 | goto on_error; |
998 | } | |
999 | ||
1000 | return 0; | |
1001 | ||
1002 | on_error: | |
9e84479f CB |
1003 | lxc_terminal_delete(terminal); |
1004 | lxc_terminal_conf_free(terminal); | |
ba2be1a8 CB |
1005 | return -1; |
1006 | } | |
1007 | ||
9e84479f CB |
1008 | static int lxc_attach_terminal_mainloop_init(struct lxc_terminal *terminal, |
1009 | struct lxc_epoll_descr *descr) | |
ba2be1a8 CB |
1010 | { |
1011 | int ret; | |
1012 | ||
1013 | ret = lxc_mainloop_open(descr); | |
1014 | if (ret < 0) { | |
1015 | ERROR("Failed to create mainloop"); | |
1016 | return -1; | |
1017 | } | |
1018 | ||
9e84479f | 1019 | ret = lxc_terminal_mainloop_add(descr, terminal); |
ba2be1a8 CB |
1020 | if (ret < 0) { |
1021 | ERROR("Failed to add handlers to mainloop"); | |
1022 | lxc_mainloop_close(descr); | |
1023 | return -1; | |
1024 | } | |
1025 | ||
1026 | return 0; | |
1027 | } | |
1028 | ||
9e84479f | 1029 | static inline void lxc_attach_terminal_close_master(struct lxc_terminal *terminal) |
ba2be1a8 | 1030 | { |
9e84479f | 1031 | if (terminal->master < 0) |
ba2be1a8 CB |
1032 | return; |
1033 | ||
9e84479f CB |
1034 | close(terminal->master); |
1035 | terminal->master = -EBADF; | |
ba2be1a8 CB |
1036 | } |
1037 | ||
9e84479f | 1038 | static inline void lxc_attach_terminal_close_slave(struct lxc_terminal *terminal) |
ba2be1a8 | 1039 | { |
9e84479f | 1040 | if (terminal->slave < 0) |
ba2be1a8 CB |
1041 | return; |
1042 | ||
9e84479f CB |
1043 | close(terminal->slave); |
1044 | terminal->slave = -EBADF; | |
ba2be1a8 CB |
1045 | } |
1046 | ||
9e84479f | 1047 | static inline void lxc_attach_terminal_close_peer(struct lxc_terminal *terminal) |
ba2be1a8 | 1048 | { |
9e84479f | 1049 | if (terminal->peer < 0) |
ba2be1a8 CB |
1050 | return; |
1051 | ||
9e84479f CB |
1052 | close(terminal->peer); |
1053 | terminal->peer = -EBADF; | |
ba2be1a8 CB |
1054 | } |
1055 | ||
9e84479f | 1056 | static inline void lxc_attach_terminal_close_log(struct lxc_terminal *terminal) |
ba2be1a8 | 1057 | { |
9e84479f | 1058 | if (terminal->log_fd < 0) |
ba2be1a8 CB |
1059 | return; |
1060 | ||
9e84479f CB |
1061 | close(terminal->log_fd); |
1062 | terminal->log_fd = -EBADF; | |
ba2be1a8 CB |
1063 | } |
1064 | ||
6f4f1937 CB |
1065 | int lxc_attach(const char *name, const char *lxcpath, |
1066 | lxc_attach_exec_t exec_function, void *exec_payload, | |
1067 | lxc_attach_options_t *options, pid_t *attached_process) | |
9c4693b8 | 1068 | { |
877f3a04 | 1069 | int i, ret, status; |
9c4693b8 | 1070 | int ipc_sockets[2]; |
6f4f1937 | 1071 | char *cwd, *new_cwd; |
9b8e3c96 | 1072 | signed long personality; |
ba2be1a8 | 1073 | pid_t attached_pid, init_pid, pid; |
6f4f1937 | 1074 | struct lxc_proc_context_info *init_ctx; |
9e84479f | 1075 | struct lxc_terminal terminal; |
1cce35e6 | 1076 | struct lxc_conf *conf; |
a998454a | 1077 | struct attach_clone_payload payload = {0}; |
9c4693b8 | 1078 | |
877f3a04 CB |
1079 | ret = access("/proc/self/ns", X_OK); |
1080 | if (ret) { | |
ea918412 | 1081 | SYSERROR("Does this kernel version support namespaces?"); |
877f3a04 CB |
1082 | return -1; |
1083 | } | |
1084 | ||
9c4693b8 CS |
1085 | if (!options) |
1086 | options = &attach_static_default_options; | |
1087 | ||
1088 | init_pid = lxc_cmd_get_init_pid(name, lxcpath); | |
1089 | if (init_pid < 0) { | |
ae026f55 | 1090 | ERROR("Failed to get init pid"); |
9c4693b8 CS |
1091 | return -1; |
1092 | } | |
1093 | ||
1094 | init_ctx = lxc_proc_get_context_info(init_pid); | |
1095 | if (!init_ctx) { | |
6f4f1937 | 1096 | ERROR("Failed to get context of init process: %ld", (long)init_pid); |
9c4693b8 CS |
1097 | return -1; |
1098 | } | |
1099 | ||
9b8e3c96 SH |
1100 | personality = get_personality(name, lxcpath); |
1101 | if (init_ctx->personality < 0) { | |
6f4f1937 | 1102 | ERROR("Failed to get personality of the container"); |
9b8e3c96 SH |
1103 | lxc_proc_put_context_info(init_ctx); |
1104 | return -1; | |
1105 | } | |
1106 | init_ctx->personality = personality; | |
1107 | ||
ff07d7bb | 1108 | init_ctx->container = lxc_container_new(name, lxcpath); |
62de1db6 CB |
1109 | if (!init_ctx->container) { |
1110 | lxc_proc_put_context_info(init_ctx); | |
ff07d7bb | 1111 | return -1; |
62de1db6 | 1112 | } |
ff07d7bb | 1113 | |
ba773996 CB |
1114 | if (!init_ctx->container->lxc_conf) { |
1115 | init_ctx->container->lxc_conf = lxc_conf_init(); | |
62de1db6 CB |
1116 | if (!init_ctx->container->lxc_conf) { |
1117 | lxc_proc_put_context_info(init_ctx); | |
ea918412 | 1118 | return -1; |
62de1db6 | 1119 | } |
ba773996 | 1120 | } |
1cce35e6 | 1121 | conf = init_ctx->container->lxc_conf; |
ba773996 | 1122 | |
bd4307f0 | 1123 | if (!fetch_seccomp(init_ctx->container, options)) |
ae026f55 | 1124 | WARN("Failed to get seccomp policy"); |
2c4ea790 | 1125 | |
bd4307f0 | 1126 | if (!no_new_privs(init_ctx->container, options)) |
ae026f55 | 1127 | WARN("Could not determine whether PR_SET_NO_NEW_PRIVS is set"); |
2e812c16 | 1128 | |
9c4693b8 CS |
1129 | cwd = getcwd(NULL, 0); |
1130 | ||
8ce83369 CB |
1131 | /* Determine which namespaces the container was created with |
1132 | * by asking lxc-start, if necessary. | |
9c4693b8 CS |
1133 | */ |
1134 | if (options->namespaces == -1) { | |
1135 | options->namespaces = lxc_cmd_get_clone_flags(name, lxcpath); | |
1136 | /* call failed */ | |
1137 | if (options->namespaces == -1) { | |
8ce83369 | 1138 | ERROR("Failed to automatically determine the " |
877f3a04 | 1139 | "namespaces which the container uses"); |
9c4693b8 | 1140 | free(cwd); |
fe4de9a6 | 1141 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1142 | return -1; |
1143 | } | |
877f3a04 CB |
1144 | |
1145 | for (i = 0; i < LXC_NS_MAX; i++) { | |
1146 | if (ns_info[i].clone_flag & CLONE_NEWCGROUP) | |
1147 | if (!(options->attach_flags & LXC_ATTACH_MOVE_TO_CGROUP) || | |
1148 | !cgns_supported()) | |
1149 | continue; | |
1150 | ||
1151 | if (ns_info[i].clone_flag & options->namespaces) | |
1152 | continue; | |
1153 | ||
1154 | init_ctx->ns_inherited |= ns_info[i].clone_flag; | |
1155 | } | |
1156 | } | |
1157 | ||
0059379f | 1158 | pid = lxc_raw_getpid(); |
ea918412 | 1159 | |
877f3a04 | 1160 | for (i = 0; i < LXC_NS_MAX; i++) { |
ea918412 | 1161 | int j; |
877f3a04 CB |
1162 | |
1163 | if (options->namespaces & ns_info[i].clone_flag) | |
1164 | init_ctx->ns_fd[i] = lxc_preserve_ns(init_pid, ns_info[i].proc_name); | |
1165 | else if (init_ctx->ns_inherited & ns_info[i].clone_flag) | |
1166 | init_ctx->ns_fd[i] = in_same_namespace(pid, init_pid, ns_info[i].proc_name); | |
1167 | else | |
1168 | continue; | |
ea918412 | 1169 | |
877f3a04 CB |
1170 | if (init_ctx->ns_fd[i] >= 0) |
1171 | continue; | |
1172 | ||
1173 | if (init_ctx->ns_fd[i] == -EINVAL) { | |
1174 | DEBUG("Inheriting %s namespace from %d", | |
1175 | ns_info[i].proc_name, pid); | |
1176 | init_ctx->ns_inherited &= ~ns_info[i].clone_flag; | |
1177 | continue; | |
1178 | } | |
1179 | ||
1180 | /* We failed to preserve the namespace. */ | |
ea918412 | 1181 | SYSERROR("Failed to attach to %s namespace of %d", |
1182 | ns_info[i].proc_name, pid); | |
1183 | ||
877f3a04 CB |
1184 | /* Close all already opened file descriptors before we return an |
1185 | * error, so we don't leak them. | |
1186 | */ | |
1187 | for (j = 0; j < i; j++) | |
1188 | close(init_ctx->ns_fd[j]); | |
1189 | ||
877f3a04 CB |
1190 | free(cwd); |
1191 | lxc_proc_put_context_info(init_ctx); | |
1192 | return -1; | |
9c4693b8 CS |
1193 | } |
1194 | ||
9e84479f CB |
1195 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
1196 | ret = lxc_attach_terminal(conf, &terminal); | |
ba2be1a8 | 1197 | if (ret < 0) { |
9e84479f | 1198 | ERROR("Failed to setup new terminal"); |
ba2be1a8 CB |
1199 | free(cwd); |
1200 | lxc_proc_put_context_info(init_ctx); | |
1201 | return -1; | |
1202 | } | |
1203 | ||
9e84479f | 1204 | terminal.log_fd = options->log_fd; |
c948657b | 1205 | } else { |
9e84479f | 1206 | lxc_terminal_init(&terminal); |
ba2be1a8 CB |
1207 | } |
1208 | ||
8ce83369 CB |
1209 | /* Create a socket pair for IPC communication; set SOCK_CLOEXEC in order |
1210 | * to make sure we don't irritate other threads that want to fork+exec | |
1211 | * away | |
9c4693b8 CS |
1212 | * |
1213 | * IMPORTANT: if the initial process is multithreaded and another call | |
1214 | * just fork()s away without exec'ing directly after, the socket fd will | |
1215 | * exist in the forked process from the other thread and any close() in | |
8ce83369 CB |
1216 | * our own child process will not really cause the socket to close |
1217 | * properly, potentiall causing the parent to hang. | |
9c4693b8 CS |
1218 | * |
1219 | * For this reason, while IPC is still active, we have to use shutdown() | |
8ce83369 CB |
1220 | * if the child exits prematurely in order to signal that the socket is |
1221 | * closed and cannot assume that the child exiting will automatically do | |
1222 | * that. | |
9c4693b8 CS |
1223 | * |
1224 | * IPC mechanism: (X is receiver) | |
1225 | * initial process intermediate attached | |
1226 | * X <--- send pid of | |
1227 | * attached proc, | |
1228 | * then exit | |
1229 | * send 0 ------------------------------------> X | |
1230 | * [do initialization] | |
1231 | * X <------------------------------------ send 1 | |
1232 | * [add to cgroup, ...] | |
1233 | * send 2 ------------------------------------> X | |
81f466d0 CB |
1234 | * [set LXC_ATTACH_NO_NEW_PRIVS] |
1235 | * X <------------------------------------ send 3 | |
1236 | * [open LSM label fd] | |
1237 | * send 4 ------------------------------------> X | |
1238 | * [set LSM label] | |
9c4693b8 CS |
1239 | * close socket close socket |
1240 | * run program | |
1241 | */ | |
1242 | ret = socketpair(PF_LOCAL, SOCK_STREAM | SOCK_CLOEXEC, 0, ipc_sockets); | |
1243 | if (ret < 0) { | |
ae026f55 | 1244 | SYSERROR("Could not set up required IPC mechanism for attaching"); |
9c4693b8 | 1245 | free(cwd); |
fe4de9a6 | 1246 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1247 | return -1; |
1248 | } | |
1249 | ||
e3f0e436 CB |
1250 | /* Create intermediate subprocess, two reasons: |
1251 | * 1. We can't setns() in the child itself, since we want to make | |
8ce83369 | 1252 | * sure we are properly attached to the pidns. |
e3f0e436 | 1253 | * 2. Also, the initial thread has to put the attached process |
8ce83369 CB |
1254 | * into the cgroup, which we can only do if we didn't already |
1255 | * setns() (otherwise, user namespaces will hate us). | |
9c4693b8 CS |
1256 | */ |
1257 | pid = fork(); | |
9c4693b8 | 1258 | if (pid < 0) { |
ae026f55 | 1259 | SYSERROR("Failed to create first subprocess"); |
9c4693b8 | 1260 | free(cwd); |
fe4de9a6 | 1261 | lxc_proc_put_context_info(init_ctx); |
9c4693b8 CS |
1262 | return -1; |
1263 | } | |
1264 | ||
1265 | if (pid) { | |
ba2be1a8 | 1266 | int ret_parent = -1; |
9c4693b8 | 1267 | pid_t to_cleanup_pid = pid; |
ba2be1a8 | 1268 | struct lxc_epoll_descr descr = {0}; |
9c4693b8 | 1269 | |
ba2be1a8 | 1270 | /* close unneeded file descriptors */ |
9c4693b8 CS |
1271 | close(ipc_sockets[1]); |
1272 | free(cwd); | |
ba2be1a8 | 1273 | lxc_proc_close_ns_fd(init_ctx); |
9e84479f CB |
1274 | if (options->attach_flags & LXC_ATTACH_TERMINAL) |
1275 | lxc_attach_terminal_close_slave(&terminal); | |
9c4693b8 | 1276 | |
8ce83369 | 1277 | /* Attach to cgroup, if requested. */ |
f4364484 | 1278 | if (options->attach_flags & LXC_ATTACH_MOVE_TO_CGROUP) { |
2202afc9 CB |
1279 | struct cgroup_ops *cgroup_ops; |
1280 | ||
5a087e05 | 1281 | cgroup_ops = cgroup_init(conf); |
2202afc9 CB |
1282 | if (!cgroup_ops) |
1283 | goto on_error; | |
1284 | ||
1285 | if (!cgroup_ops->attach(cgroup_ops, name, lxcpath, pid)) | |
8ce83369 | 1286 | goto on_error; |
2202afc9 CB |
1287 | |
1288 | cgroup_exit(cgroup_ops); | |
1289 | TRACE("Moved intermediate process %d into container's cgroups", pid); | |
f4364484 SG |
1290 | } |
1291 | ||
bb2ada6f | 1292 | /* Setup /proc limits */ |
1cce35e6 CB |
1293 | if (!lxc_list_empty(&conf->procs)) { |
1294 | ret = setup_proc_filesystem(&conf->procs, pid); | |
bb2ada6f CB |
1295 | if (ret < 0) |
1296 | goto on_error; | |
1297 | } | |
1298 | ||
c6d09e15 | 1299 | /* Setup resource limits */ |
1cce35e6 CB |
1300 | if (!lxc_list_empty(&conf->limits)) { |
1301 | ret = setup_resource_limits(&conf->limits, pid); | |
ba2be1a8 CB |
1302 | if (ret < 0) |
1303 | goto on_error; | |
1304 | } | |
1305 | ||
9e84479f CB |
1306 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
1307 | ret = lxc_attach_terminal_mainloop_init(&terminal, &descr); | |
ba2be1a8 | 1308 | if (ret < 0) |
6f4f1937 | 1309 | goto on_error; |
ea918412 | 1310 | |
9e84479f | 1311 | TRACE("Initialized terminal mainloop"); |
ba2be1a8 | 1312 | } |
c6d09e15 | 1313 | |
8ce83369 | 1314 | /* Let the child process know to go ahead. */ |
f4364484 SG |
1315 | status = 0; |
1316 | ret = lxc_write_nointr(ipc_sockets[0], &status, sizeof(status)); | |
94ac256f | 1317 | if (ret != sizeof(status)) |
ba2be1a8 | 1318 | goto close_mainloop; |
ea918412 | 1319 | |
94ac256f | 1320 | TRACE("Told intermediate process to start initializing"); |
f4364484 | 1321 | |
8ce83369 | 1322 | /* Get pid of attached process from intermediate process. */ |
94ac256f CB |
1323 | ret = lxc_read_nointr(ipc_sockets[0], &attached_pid, sizeof(attached_pid)); |
1324 | if (ret != sizeof(attached_pid)) | |
ba2be1a8 | 1325 | goto close_mainloop; |
ea918412 | 1326 | |
94ac256f | 1327 | TRACE("Received pid %d of attached process in parent pid namespace", attached_pid); |
9c4693b8 | 1328 | |
8ce83369 | 1329 | /* Ignore SIGKILL (CTRL-C) and SIGQUIT (CTRL-\) - issue #313. */ |
62183f1a SH |
1330 | if (options->stdin_fd == 0) { |
1331 | signal(SIGINT, SIG_IGN); | |
1332 | signal(SIGQUIT, SIG_IGN); | |
1333 | } | |
2eef2bda | 1334 | |
8ce83369 | 1335 | /* Reap intermediate process. */ |
9c4693b8 CS |
1336 | ret = wait_for_pid(pid); |
1337 | if (ret < 0) | |
ba2be1a8 | 1338 | goto close_mainloop; |
ea918412 | 1339 | |
94ac256f | 1340 | TRACE("Intermediate process %d exited", pid); |
9c4693b8 | 1341 | |
8ce83369 | 1342 | /* We will always have to reap the attached process now. */ |
9c4693b8 CS |
1343 | to_cleanup_pid = attached_pid; |
1344 | ||
81f466d0 | 1345 | /* Open LSM fd and send it to child. */ |
6f4f1937 CB |
1346 | if ((options->namespaces & CLONE_NEWNS) && |
1347 | (options->attach_flags & LXC_ATTACH_LSM) && | |
1348 | init_ctx->lsm_label) { | |
94ac256f | 1349 | int ret = -1; |
47ce2cb7 CB |
1350 | int labelfd; |
1351 | bool on_exec; | |
6f4f1937 | 1352 | |
47ce2cb7 CB |
1353 | on_exec = options->attach_flags & LXC_ATTACH_LSM_EXEC ? true : false; |
1354 | labelfd = lsm_process_label_fd_get(attached_pid, on_exec); | |
81f466d0 | 1355 | if (labelfd < 0) |
ba2be1a8 | 1356 | goto close_mainloop; |
ea918412 | 1357 | |
94ac256f | 1358 | TRACE("Opened LSM label file descriptor %d", labelfd); |
81f466d0 CB |
1359 | |
1360 | /* Send child fd of the LSM security module to write to. */ | |
ae467c54 | 1361 | ret = lxc_abstract_unix_send_fds(ipc_sockets[0], &labelfd, 1, NULL, 0); |
81f466d0 | 1362 | if (ret <= 0) { |
9044b79e | 1363 | if (ret < 0) |
1364 | SYSERROR("Failed to send lsm label fd"); | |
1365 | ||
1366 | close(labelfd); | |
ba2be1a8 | 1367 | goto close_mainloop; |
81f466d0 | 1368 | } |
9044b79e | 1369 | |
1370 | close(labelfd); | |
94ac256f | 1371 | TRACE("Sent LSM label file descriptor %d to child", labelfd); |
81f466d0 CB |
1372 | } |
1373 | ||
8ce83369 CB |
1374 | /* We're done, the child process should now execute whatever it |
1375 | * is that the user requested. The parent can now track it with | |
1376 | * waitpid() or similar. | |
9c4693b8 CS |
1377 | */ |
1378 | ||
1379 | *attached_process = attached_pid; | |
9c4693b8 | 1380 | |
ba2be1a8 | 1381 | /* Now shut down communication with child, we're done. */ |
9c4693b8 CS |
1382 | shutdown(ipc_sockets[0], SHUT_RDWR); |
1383 | close(ipc_sockets[0]); | |
ba2be1a8 CB |
1384 | ipc_sockets[0] = -1; |
1385 | ||
1386 | ret_parent = 0; | |
1387 | to_cleanup_pid = -1; | |
ea918412 | 1388 | |
9e84479f | 1389 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
ba2be1a8 CB |
1390 | ret = lxc_mainloop(&descr, -1); |
1391 | if (ret < 0) { | |
1392 | ret_parent = -1; | |
1393 | to_cleanup_pid = attached_pid; | |
1394 | } | |
1395 | } | |
1396 | ||
1397 | close_mainloop: | |
9e84479f | 1398 | if (options->attach_flags & LXC_ATTACH_TERMINAL) |
ba2be1a8 CB |
1399 | lxc_mainloop_close(&descr); |
1400 | ||
1401 | on_error: | |
1402 | if (ipc_sockets[0] >= 0) { | |
1403 | shutdown(ipc_sockets[0], SHUT_RDWR); | |
1404 | close(ipc_sockets[0]); | |
1405 | } | |
1406 | ||
1407 | if (to_cleanup_pid > 0) | |
6f4f1937 | 1408 | (void)wait_for_pid(to_cleanup_pid); |
ba2be1a8 | 1409 | |
9e84479f CB |
1410 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
1411 | lxc_terminal_delete(&terminal); | |
1412 | lxc_terminal_conf_free(&terminal); | |
ba2be1a8 | 1413 | } |
ea918412 | 1414 | |
fe4de9a6 | 1415 | lxc_proc_put_context_info(init_ctx); |
ba2be1a8 | 1416 | return ret_parent; |
9c4693b8 CS |
1417 | } |
1418 | ||
ba2be1a8 | 1419 | /* close unneeded file descriptors */ |
9c4693b8 | 1420 | close(ipc_sockets[0]); |
ba2be1a8 | 1421 | ipc_sockets[0] = -EBADF; |
ea918412 | 1422 | |
9e84479f CB |
1423 | if (options->attach_flags & LXC_ATTACH_TERMINAL) { |
1424 | lxc_attach_terminal_close_master(&terminal); | |
1425 | lxc_attach_terminal_close_peer(&terminal); | |
1426 | lxc_attach_terminal_close_log(&terminal); | |
ba2be1a8 | 1427 | } |
9c4693b8 | 1428 | |
8ce83369 | 1429 | /* Wait for the parent to have setup cgroups. */ |
94ac256f | 1430 | ret = lxc_read_nointr(ipc_sockets[1], &status, sizeof(status)); |
ba2be1a8 | 1431 | if (ret != sizeof(status)) { |
f4364484 | 1432 | shutdown(ipc_sockets[1], SHUT_RDWR); |
62de1db6 | 1433 | lxc_proc_put_context_info(init_ctx); |
c7ac2e1c | 1434 | _exit(EXIT_FAILURE); |
f4364484 | 1435 | } |
ea918412 | 1436 | |
94ac256f | 1437 | TRACE("Intermediate process starting to initialize"); |
f4364484 | 1438 | |
8ce83369 CB |
1439 | /* Attach now, create another subprocess later, since pid namespaces |
1440 | * only really affect the children of the current process. | |
9c4693b8 | 1441 | */ |
877f3a04 | 1442 | ret = lxc_attach_to_ns(init_pid, init_ctx); |
9c4693b8 | 1443 | if (ret < 0) { |
94ac256f | 1444 | ERROR("Failed to enter namespaces"); |
9c4693b8 | 1445 | shutdown(ipc_sockets[1], SHUT_RDWR); |
62de1db6 | 1446 | lxc_proc_put_context_info(init_ctx); |
c7ac2e1c | 1447 | _exit(EXIT_FAILURE); |
9c4693b8 | 1448 | } |
ea918412 | 1449 | |
877f3a04 CB |
1450 | /* close namespace file descriptors */ |
1451 | lxc_proc_close_ns_fd(init_ctx); | |
9c4693b8 | 1452 | |
8ce83369 | 1453 | /* Attach succeeded, try to cwd. */ |
9c4693b8 CS |
1454 | if (options->initial_cwd) |
1455 | new_cwd = options->initial_cwd; | |
1456 | else | |
1457 | new_cwd = cwd; | |
d6d979bc CB |
1458 | if (new_cwd) { |
1459 | ret = chdir(new_cwd); | |
1460 | if (ret < 0) | |
1461 | WARN("Could not change directory to \"%s\"", new_cwd); | |
1462 | } | |
9c4693b8 CS |
1463 | free(cwd); |
1464 | ||
a998454a CB |
1465 | /* Create attached process. */ |
1466 | payload.ipc_socket = ipc_sockets[1]; | |
1467 | payload.options = options; | |
1468 | payload.init_ctx = init_ctx; | |
9e84479f | 1469 | payload.terminal_slave_fd = terminal.slave; |
a998454a CB |
1470 | payload.exec_function = exec_function; |
1471 | payload.exec_payload = exec_payload; | |
9c4693b8 | 1472 | |
a998454a CB |
1473 | pid = lxc_raw_clone(CLONE_PARENT); |
1474 | if (pid < 0) { | |
94ac256f | 1475 | SYSERROR("Failed to clone attached process"); |
9c4693b8 | 1476 | shutdown(ipc_sockets[1], SHUT_RDWR); |
62de1db6 | 1477 | lxc_proc_put_context_info(init_ctx); |
c7ac2e1c | 1478 | _exit(EXIT_FAILURE); |
9c4693b8 | 1479 | } |
a998454a CB |
1480 | |
1481 | if (pid == 0) { | |
1482 | ret = attach_child_main(&payload); | |
1483 | if (ret < 0) | |
1484 | ERROR("Failed to exec"); | |
ea918412 | 1485 | |
a998454a CB |
1486 | _exit(EXIT_FAILURE); |
1487 | } | |
ea918412 | 1488 | |
9e84479f CB |
1489 | if (options->attach_flags & LXC_ATTACH_TERMINAL) |
1490 | lxc_attach_terminal_close_slave(&terminal); | |
9c4693b8 | 1491 | |
8ce83369 | 1492 | /* Tell grandparent the pid of the pid of the newly created child. */ |
9c4693b8 CS |
1493 | ret = lxc_write_nointr(ipc_sockets[1], &pid, sizeof(pid)); |
1494 | if (ret != sizeof(pid)) { | |
8ce83369 CB |
1495 | /* If this really happens here, this is very unfortunate, since |
1496 | * the parent will not know the pid of the attached process and | |
1497 | * will not be able to wait for it (and we won't either due to | |
1498 | * CLONE_PARENT) so the parent won't be able to reap it and the | |
1499 | * attached process will remain a zombie. | |
9c4693b8 | 1500 | */ |
9c4693b8 | 1501 | shutdown(ipc_sockets[1], SHUT_RDWR); |
62de1db6 | 1502 | lxc_proc_put_context_info(init_ctx); |
c7ac2e1c | 1503 | _exit(EXIT_FAILURE); |
9c4693b8 | 1504 | } |
ea918412 | 1505 | |
94ac256f | 1506 | TRACE("Sending pid %d of attached process", pid); |
9c4693b8 | 1507 | |
8ce83369 | 1508 | /* The rest is in the hands of the initial and the attached process. */ |
62de1db6 | 1509 | lxc_proc_put_context_info(init_ctx); |
c7ac2e1c | 1510 | _exit(0); |
9c4693b8 CS |
1511 | } |
1512 | ||
9c4693b8 CS |
1513 | int lxc_attach_run_command(void* payload) |
1514 | { | |
1515 | lxc_attach_command_t* cmd = (lxc_attach_command_t*)payload; | |
1516 | ||
1517 | execvp(cmd->program, cmd->argv); | |
ea918412 | 1518 | |
1519 | SYSERROR("Failed to exec \"%s\"", cmd->program); | |
9c4693b8 CS |
1520 | return -1; |
1521 | } | |
1522 | ||
1523 | int lxc_attach_run_shell(void* payload) | |
1524 | { | |
1525 | uid_t uid; | |
cb7aa5e8 DJ |
1526 | struct passwd pwent; |
1527 | struct passwd *pwentp = NULL; | |
9c4693b8 | 1528 | char *user_shell; |
cb7aa5e8 DJ |
1529 | char *buf; |
1530 | size_t bufsize; | |
1531 | int ret; | |
9c4693b8 | 1532 | |
8ce83369 | 1533 | /* Ignore payload parameter. */ |
9c4693b8 CS |
1534 | (void)payload; |
1535 | ||
1536 | uid = getuid(); | |
cb7aa5e8 DJ |
1537 | |
1538 | bufsize = sysconf(_SC_GETPW_R_SIZE_MAX); | |
1539 | if (bufsize == -1) | |
1540 | bufsize = 1024; | |
1541 | ||
1542 | buf = malloc(bufsize); | |
1543 | if (buf) { | |
1544 | ret = getpwuid_r(uid, &pwent, buf, bufsize, &pwentp); | |
1545 | if (!pwentp) { | |
1546 | if (ret == 0) | |
ea918412 | 1547 | WARN("Could not find matched password record"); |
cb7aa5e8 DJ |
1548 | |
1549 | WARN("Failed to get password record - %u", uid); | |
1550 | } | |
1551 | } | |
9c4693b8 | 1552 | |
8ce83369 CB |
1553 | /* This probably happens because of incompatible nss implementations in |
1554 | * host and container (remember, this code is still using the host's | |
1555 | * glibc but our mount namespace is in the container) we may try to get | |
1556 | * the information by spawning a [getent passwd uid] process and parsing | |
1557 | * the result. | |
9c4693b8 | 1558 | */ |
cb7aa5e8 | 1559 | if (!pwentp) |
9c4693b8 CS |
1560 | user_shell = lxc_attach_getpwshell(uid); |
1561 | else | |
cb7aa5e8 | 1562 | user_shell = pwent.pw_shell; |
ea918412 | 1563 | |
9c4693b8 | 1564 | if (user_shell) |
acf47e1b | 1565 | execlp(user_shell, user_shell, (char *)NULL); |
9c4693b8 | 1566 | |
8ce83369 CB |
1567 | /* Executed if either no passwd entry or execvp fails, we will fall back |
1568 | * on /bin/sh as a default shell. | |
9c4693b8 | 1569 | */ |
acf47e1b | 1570 | execlp("/bin/sh", "/bin/sh", (char *)NULL); |
ea918412 | 1571 | |
edeb1836 | 1572 | SYSERROR("Failed to execute shell"); |
cb7aa5e8 | 1573 | if (!pwentp) |
edeb1836 | 1574 | free(user_shell); |
ea918412 | 1575 | |
cb7aa5e8 | 1576 | free(buf); |
9c4693b8 CS |
1577 | return -1; |
1578 | } |