]>
Commit | Line | Data |
---|---|---|
0ad19a3f | 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> |
0ad19a3f | 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 |
0ad19a3f | 22 | */ |
23 | ||
f3787121 | 24 | #define _GNU_SOURCE |
f549edcc GK |
25 | #include "config.h" |
26 | ||
f3787121 | 27 | #include <alloca.h> |
0ad19a3f | 28 | #include <dirent.h> |
29 | #include <errno.h> | |
b0a33c1e | 30 | #include <fcntl.h> |
c476bdce | 31 | #include <grp.h> |
37515ebd | 32 | #include <poll.h> |
f3787121 CB |
33 | #include <signal.h> |
34 | #include <stdio.h> | |
35 | #include <stdlib.h> | |
36 | #include <string.h> | |
37 | #include <unistd.h> | |
0ad19a3f | 38 | #include <sys/file.h> |
f4d507d5 | 39 | #include <sys/mount.h> |
f3787121 | 40 | #include <sys/param.h> |
0ad19a3f | 41 | #include <sys/prctl.h> |
f3787121 CB |
42 | #include <sys/socket.h> |
43 | #include <sys/stat.h> | |
44 | #include <sys/syscall.h> | |
ddceb1f9 | 45 | #include <sys/types.h> |
b0a33c1e | 46 | #include <sys/un.h> |
f3787121 | 47 | #include <sys/wait.h> |
ff218c25 | 48 | |
495d2046 SG |
49 | #if HAVE_SYS_CAPABILITY_H |
50 | #include <sys/capability.h> | |
51 | #endif | |
52 | ||
656994bb MH |
53 | #if !HAVE_DECL_PR_CAPBSET_DROP |
54 | #define PR_CAPBSET_DROP 24 | |
55 | #endif | |
56 | ||
f3787121 | 57 | #include "af_unix.h" |
d8e48992 | 58 | #include "bdev.h" |
f3787121 | 59 | #include "caps.h" |
563f2f2c | 60 | #include "cgroup.h" |
f3787121 CB |
61 | #include "commands.h" |
62 | #include "conf.h" | |
63 | #include "console.h" | |
e2bcd7db | 64 | #include "error.h" |
f3787121 CB |
65 | #include "log.h" |
66 | #include "lxclock.h" | |
67 | #include "lxcseccomp.h" | |
565c2d76 | 68 | #include "lxcutmp.h" |
f3787121 | 69 | #include "mainloop.h" |
63376d7d | 70 | #include "monitor.h" |
f549edcc | 71 | #include "namespace.h" |
f3787121 CB |
72 | #include "start.h" |
73 | #include "sync.h" | |
74 | #include "utils.h" | |
fe4de9a6 | 75 | #include "lsm/lsm.h" |
36eb9bde CLG |
76 | |
77 | lxc_log_define(lxc_start, lxc); | |
78 | ||
9f30a190 MM |
79 | const struct ns_info ns_info[LXC_NS_MAX] = { |
80 | [LXC_NS_MNT] = {"mnt", CLONE_NEWNS}, | |
81 | [LXC_NS_PID] = {"pid", CLONE_NEWPID}, | |
82 | [LXC_NS_UTS] = {"uts", CLONE_NEWUTS}, | |
83 | [LXC_NS_IPC] = {"ipc", CLONE_NEWIPC}, | |
84 | [LXC_NS_USER] = {"user", CLONE_NEWUSER}, | |
ceecc92c CB |
85 | [LXC_NS_NET] = {"net", CLONE_NEWNET}, |
86 | [LXC_NS_CGROUP] = {"cgroup", CLONE_NEWCGROUP} | |
9f30a190 MM |
87 | }; |
88 | ||
f01f7975 | 89 | extern void mod_all_rdeps(struct lxc_container *c, bool inc); |
28272964 CB |
90 | static bool do_destroy_container(struct lxc_conf *conf); |
91 | static int lxc_rmdir_onedev_wrapper(void *data); | |
92 | static void lxc_destroy_container_on_signal(struct lxc_handler *handler, | |
93 | const char *name); | |
94 | ||
c8154066 SH |
95 | static void print_top_failing_dir(const char *path) |
96 | { | |
97 | size_t len = strlen(path); | |
f3787121 | 98 | char *copy = alloca(len + 1), *p, *e, saved; |
c8154066 SH |
99 | strcpy(copy, path); |
100 | ||
101 | p = copy; | |
102 | e = copy + len; | |
103 | while (p < e) { | |
f3787121 CB |
104 | while (p < e && *p == '/') |
105 | p++; | |
106 | while (p < e && *p != '/') | |
107 | p++; | |
c8154066 SH |
108 | saved = *p; |
109 | *p = '\0'; | |
110 | if (access(copy, X_OK)) { | |
111 | SYSERROR("could not access %s. Please grant it 'x' " \ | |
112 | "access, or add an ACL for the container root.", | |
113 | copy); | |
114 | return; | |
115 | } | |
116 | *p = saved; | |
117 | } | |
118 | } | |
119 | ||
9f30a190 MM |
120 | static void close_ns(int ns_fd[LXC_NS_MAX]) { |
121 | int i; | |
122 | ||
9f30a190 MM |
123 | for (i = 0; i < LXC_NS_MAX; i++) { |
124 | if (ns_fd[i] > -1) { | |
125 | close(ns_fd[i]); | |
126 | ns_fd[i] = -1; | |
127 | } | |
128 | } | |
9f30a190 MM |
129 | } |
130 | ||
62d05d9b SH |
131 | /* |
132 | * preserve_ns: open /proc/@pid/ns/@ns for each namespace specified | |
133 | * in clone_flags. | |
134 | * Return true on success, false on failure. On failure, leave an error | |
135 | * message in *errmsg, which caller must free. | |
136 | */ | |
f3787121 CB |
137 | static bool preserve_ns(int ns_fd[LXC_NS_MAX], int clone_flags, pid_t pid, |
138 | char **errmsg) | |
139 | { | |
62d05d9b | 140 | int i, ret; |
9f30a190 MM |
141 | char path[MAXPATHLEN]; |
142 | ||
9f30a190 MM |
143 | for (i = 0; i < LXC_NS_MAX; i++) |
144 | ns_fd[i] = -1; | |
145 | ||
035a38fc WB |
146 | snprintf(path, MAXPATHLEN, "/proc/%d/ns", pid); |
147 | if (access(path, X_OK)) { | |
62d05d9b SH |
148 | if (asprintf(errmsg, "Kernel does not support setns.") == -1) |
149 | *errmsg = NULL; | |
150 | return false; | |
cd43d2d1 SH |
151 | } |
152 | ||
9f30a190 MM |
153 | for (i = 0; i < LXC_NS_MAX; i++) { |
154 | if ((clone_flags & ns_info[i].clone_flag) == 0) | |
155 | continue; | |
035a38fc WB |
156 | snprintf(path, MAXPATHLEN, "/proc/%d/ns/%s", pid, |
157 | ns_info[i].proc_name); | |
9f30a190 | 158 | ns_fd[i] = open(path, O_RDONLY | O_CLOEXEC); |
9f30a190 MM |
159 | if (ns_fd[i] < 0) |
160 | goto error; | |
161 | } | |
162 | ||
62d05d9b | 163 | return true; |
9f30a190 MM |
164 | |
165 | error: | |
62d05d9b SH |
166 | if (errno == ENOENT) { |
167 | ret = asprintf(errmsg, "Kernel does not support setns for %s", | |
168 | ns_info[i].proc_name); | |
169 | } else { | |
170 | ret = asprintf(errmsg, "Failed to open %s: %s", | |
171 | path, strerror(errno)); | |
172 | } | |
173 | if (ret == -1) | |
174 | *errmsg = NULL; | |
9f30a190 | 175 | close_ns(ns_fd); |
62d05d9b | 176 | return false; |
9f30a190 MM |
177 | } |
178 | ||
179 | static int attach_ns(const int ns_fd[LXC_NS_MAX]) { | |
180 | int i; | |
181 | ||
182 | for (i = 0; i < LXC_NS_MAX; i++) { | |
183 | if (ns_fd[i] < 0) | |
184 | continue; | |
185 | ||
186 | if (setns(ns_fd[i], 0) != 0) | |
187 | goto error; | |
188 | } | |
189 | return 0; | |
190 | ||
191 | error: | |
192 | SYSERROR("failed to set namespace '%s'", ns_info[i].proc_name); | |
193 | return -1; | |
194 | } | |
195 | ||
80090207 CLG |
196 | static int match_fd(int fd) |
197 | { | |
198 | return (fd == 0 || fd == 1 || fd == 2); | |
199 | } | |
200 | ||
d2cf4c37 SH |
201 | /* |
202 | * Check for any fds we need to close | |
203 | * * if fd_to_ignore != -1, then if we find that fd open we will ignore it. | |
204 | * * By default we warn about open fds we find. | |
205 | * * If closeall is true, we will close open fds. | |
206 | * * If lxc-start was passed "-C", then conf->close_all_fds will be true, | |
207 | * in which case we also close all open fds. | |
208 | * * A daemonized container will always pass closeall=true. | |
209 | */ | |
210 | int lxc_check_inherited(struct lxc_conf *conf, bool closeall, int fd_to_ignore) | |
80090207 CLG |
211 | { |
212 | struct dirent dirent, *direntp; | |
213 | int fd, fddir; | |
214 | DIR *dir; | |
80090207 | 215 | |
d2cf4c37 SH |
216 | if (conf && conf->close_all_fds) |
217 | closeall = true; | |
218 | ||
b119f362 | 219 | restart: |
80090207 CLG |
220 | dir = opendir("/proc/self/fd"); |
221 | if (!dir) { | |
222 | WARN("failed to open directory: %m"); | |
223 | return -1; | |
224 | } | |
225 | ||
226 | fddir = dirfd(dir); | |
227 | ||
228 | while (!readdir_r(dir, &dirent, &direntp)) { | |
80090207 CLG |
229 | if (!direntp) |
230 | break; | |
231 | ||
232 | if (!strcmp(direntp->d_name, ".")) | |
233 | continue; | |
234 | ||
235 | if (!strcmp(direntp->d_name, "..")) | |
236 | continue; | |
237 | ||
238 | fd = atoi(direntp->d_name); | |
239 | ||
f2faa8fa | 240 | if (fd == fddir || fd == lxc_log_fd || fd == fd_to_ignore) |
80090207 CLG |
241 | continue; |
242 | ||
858377e4 SH |
243 | if (current_config && fd == current_config->logfd) |
244 | continue; | |
245 | ||
80090207 CLG |
246 | if (match_fd(fd)) |
247 | continue; | |
80090207 | 248 | |
d2cf4c37 | 249 | if (closeall) { |
b119f362 SH |
250 | close(fd); |
251 | closedir(dir); | |
252 | INFO("closed inherited fd %d", fd); | |
253 | goto restart; | |
254 | } | |
92c7f629 | 255 | WARN("inherited fd %d", fd); |
80090207 CLG |
256 | } |
257 | ||
64c57ea1 BD |
258 | /* |
259 | * only enable syslog at this point to avoid the above logging function | |
260 | * to open a new fd and make the check_inherited function enter an | |
261 | * infinite loop. | |
262 | */ | |
263 | lxc_log_enable_syslog(); | |
264 | ||
92c7f629 GK |
265 | closedir(dir); /* cannot fail */ |
266 | return 0; | |
80090207 CLG |
267 | } |
268 | ||
83ee7875 | 269 | static int setup_signal_fd(sigset_t *oldmask) |
b0a33c1e | 270 | { |
271 | sigset_t mask; | |
272 | int fd; | |
273 | ||
f3304a29 FW |
274 | /* Block everything except serious error signals */ |
275 | if (sigfillset(&mask) || | |
276 | sigdelset(&mask, SIGILL) || | |
277 | sigdelset(&mask, SIGSEGV) || | |
278 | sigdelset(&mask, SIGBUS) || | |
b5159817 | 279 | sigdelset(&mask, SIGWINCH) || |
f3304a29 FW |
280 | sigprocmask(SIG_BLOCK, &mask, oldmask)) { |
281 | SYSERROR("failed to set signal mask"); | |
b0a33c1e | 282 | return -1; |
283 | } | |
284 | ||
285 | fd = signalfd(-1, &mask, 0); | |
286 | if (fd < 0) { | |
36eb9bde | 287 | SYSERROR("failed to create the signal fd"); |
b0a33c1e | 288 | return -1; |
289 | } | |
290 | ||
291 | if (fcntl(fd, F_SETFD, FD_CLOEXEC)) { | |
36eb9bde | 292 | SYSERROR("failed to set sigfd to close-on-exec"); |
b0a33c1e | 293 | close(fd); |
294 | return -1; | |
295 | } | |
296 | ||
1ac470c0 DL |
297 | DEBUG("sigchild handler set"); |
298 | ||
b0a33c1e | 299 | return fd; |
300 | } | |
301 | ||
84c92abd | 302 | static int signal_handler(int fd, uint32_t events, void *data, |
f3787121 | 303 | struct lxc_epoll_descr *descr) |
b0a33c1e | 304 | { |
15cd25fd | 305 | struct signalfd_siginfo siginfo; |
80507ee8 | 306 | siginfo_t info; |
15cd25fd | 307 | int ret; |
82d89dce | 308 | pid_t *pid = data; |
80507ee8 | 309 | bool init_died = false; |
15cd25fd DL |
310 | |
311 | ret = read(fd, &siginfo, sizeof(siginfo)); | |
312 | if (ret < 0) { | |
f3304a29 | 313 | ERROR("failed to read signal info"); |
15cd25fd DL |
314 | return -1; |
315 | } | |
316 | ||
317 | if (ret != sizeof(siginfo)) { | |
318 | ERROR("unexpected siginfo size"); | |
319 | return -1; | |
320 | } | |
321 | ||
80507ee8 SH |
322 | // check whether init is running |
323 | info.si_pid = 0; | |
324 | ret = waitid(P_PID, *pid, &info, WEXITED | WNOWAIT | WNOHANG); | |
325 | if (ret == 0 && info.si_pid == *pid) { | |
326 | init_died = true; | |
327 | } | |
328 | ||
f3304a29 FW |
329 | if (siginfo.ssi_signo != SIGCHLD) { |
330 | kill(*pid, siginfo.ssi_signo); | |
331 | INFO("forwarded signal %d to pid %d", siginfo.ssi_signo, *pid); | |
80507ee8 | 332 | return init_died ? 1 : 0; |
f3304a29 FW |
333 | } |
334 | ||
15cd25fd DL |
335 | if (siginfo.ssi_code == CLD_STOPPED || |
336 | siginfo.ssi_code == CLD_CONTINUED) { | |
337 | INFO("container init process was stopped/continued"); | |
80507ee8 | 338 | return init_died ? 1 : 0; |
15cd25fd | 339 | } |
1ac470c0 | 340 | |
82d89dce DL |
341 | /* more robustness, protect ourself from a SIGCHLD sent |
342 | * by a process different from the container init | |
343 | */ | |
344 | if (siginfo.ssi_pid != *pid) { | |
345 | WARN("invalid pid for SIGCHLD"); | |
80507ee8 | 346 | return init_died ? 1 : 0; |
82d89dce DL |
347 | } |
348 | ||
15cd25fd | 349 | DEBUG("container init process exited"); |
b0a33c1e | 350 | return 1; |
351 | } | |
352 | ||
735f2c6e | 353 | int lxc_set_state(const char *name, struct lxc_handler *handler, lxc_state_t state) |
66aeffc7 DL |
354 | { |
355 | handler->state = state; | |
9123e471 | 356 | lxc_monitor_send_state(name, state, handler->lxcpath); |
66aeffc7 DL |
357 | return 0; |
358 | } | |
359 | ||
735f2c6e | 360 | int lxc_poll(const char *name, struct lxc_handler *handler) |
b0a33c1e | 361 | { |
ca5f7926 DL |
362 | int sigfd = handler->sigfd; |
363 | int pid = handler->pid; | |
b0a33c1e | 364 | struct lxc_epoll_descr descr; |
365 | ||
a9e61274 | 366 | if (lxc_mainloop_open(&descr)) { |
36eb9bde | 367 | ERROR("failed to create mainloop"); |
50c8bf05 | 368 | goto out_sigfd; |
b0a33c1e | 369 | } |
370 | ||
83ee7875 | 371 | if (lxc_mainloop_add_handler(&descr, sigfd, signal_handler, &pid)) { |
36eb9bde | 372 | ERROR("failed to add handler for the signal"); |
b0a33c1e | 373 | goto out_mainloop_open; |
374 | } | |
375 | ||
da41561c | 376 | if (lxc_console_mainloop_add(&descr, handler->conf)) { |
63376d7d DL |
377 | ERROR("failed to add console handler to mainloop"); |
378 | goto out_mainloop_open; | |
379 | } | |
380 | ||
ef6e34ee | 381 | if (lxc_cmd_mainloop_add(name, &descr, handler)) { |
563f2f2c | 382 | ERROR("failed to add command handler to mainloop"); |
96fa1ff0 | 383 | goto out_mainloop_open; |
563f2f2c DL |
384 | } |
385 | ||
828695d9 | 386 | if (handler->conf->need_utmp_watch) { |
495d2046 | 387 | #if HAVE_SYS_CAPABILITY_H |
828695d9 SH |
388 | if (lxc_utmp_mainloop_add(&descr, handler)) { |
389 | ERROR("failed to add utmp handler to mainloop"); | |
390 | goto out_mainloop_open; | |
391 | } | |
495d2046 | 392 | #else |
959aee9c | 393 | DEBUG("not starting utmp handler as cap_sys_boot cannot be dropped without capabilities support"); |
495d2046 | 394 | #endif |
563f2f2c | 395 | } |
b0a33c1e | 396 | |
e51d4895 | 397 | return lxc_mainloop(&descr, -1); |
b0a33c1e | 398 | |
399 | out_mainloop_open: | |
400 | lxc_mainloop_close(&descr); | |
b0a33c1e | 401 | out_sigfd: |
402 | close(sigfd); | |
c3e13372 | 403 | return -1; |
b0a33c1e | 404 | } |
405 | ||
13f5be62 | 406 | struct lxc_handler *lxc_init(const char *name, struct lxc_conf *conf, const char *lxcpath) |
59eb99ba | 407 | { |
b6b2b194 | 408 | int i; |
3a0f472d DL |
409 | struct lxc_handler *handler; |
410 | ||
411 | handler = malloc(sizeof(*handler)); | |
412 | if (!handler) | |
413 | return NULL; | |
59eb99ba DL |
414 | |
415 | memset(handler, 0, sizeof(*handler)); | |
416 | ||
e8bd4e43 | 417 | handler->ttysock[0] = handler->ttysock[1] = -1; |
fae349da | 418 | handler->conf = conf; |
9123e471 | 419 | handler->lxcpath = lxcpath; |
5c068da9 | 420 | handler->pinfd = -1; |
fae349da | 421 | |
b6b2b194 WB |
422 | for (i = 0; i < LXC_NS_MAX; i++) |
423 | handler->nsfd[i] = -1; | |
424 | ||
fe4de9a6 DE |
425 | lsm_init(); |
426 | ||
3bdf52d7 DL |
427 | handler->name = strdup(name); |
428 | if (!handler->name) { | |
429 | ERROR("failed to allocate memory"); | |
430 | goto out_free; | |
431 | } | |
432 | ||
ef6e34ee | 433 | if (lxc_cmd_init(name, handler, lxcpath)) |
d2e30e99 DE |
434 | goto out_free_name; |
435 | ||
8f2c3a70 SH |
436 | if (lxc_read_seccomp_config(conf) != 0) { |
437 | ERROR("failed loading seccomp policy"); | |
d2e30e99 | 438 | goto out_close_maincmd_fd; |
8f2c3a70 SH |
439 | } |
440 | ||
051151de | 441 | /* Begin by setting the state to STARTING */ |
25c2aca5 | 442 | if (lxc_set_state(name, handler, STARTING)) { |
59eb99ba | 443 | ERROR("failed to set state '%s'", lxc_state2str(STARTING)); |
051151de | 444 | goto out_close_maincmd_fd; |
0ad19a3f | 445 | } |
446 | ||
f7bee6c6 | 447 | /* Start of environment variable setup for hooks */ |
ab7efcf5 | 448 | if (name && setenv("LXC_NAME", name, 1)) { |
f7bee6c6 MW |
449 | SYSERROR("failed to set environment variable for container name"); |
450 | } | |
ab7efcf5 | 451 | if (conf->rcfile && setenv("LXC_CONFIG_FILE", conf->rcfile, 1)) { |
f7bee6c6 MW |
452 | SYSERROR("failed to set environment variable for config path"); |
453 | } | |
ab7efcf5 | 454 | if (conf->rootfs.mount && setenv("LXC_ROOTFS_MOUNT", conf->rootfs.mount, 1)) { |
f7bee6c6 MW |
455 | SYSERROR("failed to set environment variable for rootfs mount"); |
456 | } | |
ab7efcf5 | 457 | if (conf->rootfs.path && setenv("LXC_ROOTFS_PATH", conf->rootfs.path, 1)) { |
f7bee6c6 MW |
458 | SYSERROR("failed to set environment variable for rootfs mount"); |
459 | } | |
460 | if (conf->console.path && setenv("LXC_CONSOLE", conf->console.path, 1)) { | |
461 | SYSERROR("failed to set environment variable for console path"); | |
462 | } | |
463 | if (conf->console.log_path && setenv("LXC_CONSOLE_LOGPATH", conf->console.log_path, 1)) { | |
464 | SYSERROR("failed to set environment variable for console log"); | |
465 | } | |
c4cafa08 SH |
466 | if (setenv("LXC_CGNS_AWARE", "1", 1)) { |
467 | SYSERROR("failed to set LXC_CGNS_AWARE environment variable"); | |
468 | } | |
f7bee6c6 MW |
469 | /* End of environment variable setup for hooks */ |
470 | ||
283678ed | 471 | if (run_lxc_hooks(name, "pre-start", conf, handler->lxcpath, NULL)) { |
773fb9ca SH |
472 | ERROR("failed to run pre-start hooks for container '%s'.", name); |
473 | goto out_aborting; | |
474 | } | |
26ddeedd | 475 | |
b0a33c1e | 476 | /* the signal fd has to be created before forking otherwise |
477 | * if the child process exits before we setup the signal fd, | |
478 | * the event will be lost and the command will be stuck */ | |
83ee7875 | 479 | handler->sigfd = setup_signal_fd(&handler->oldmask); |
59eb99ba | 480 | if (handler->sigfd < 0) { |
36eb9bde | 481 | ERROR("failed to set sigchild fd handler"); |
b5159817 DE |
482 | goto out_delete_tty; |
483 | } | |
484 | ||
485 | /* do this after setting up signals since it might unblock SIGWINCH */ | |
486 | if (lxc_console_create(conf)) { | |
487 | ERROR("failed to create console"); | |
488 | goto out_restore_sigmask; | |
b0a33c1e | 489 | } |
490 | ||
c4d10a05 SH |
491 | if (ttys_shift_ids(conf) < 0) { |
492 | ERROR("Failed to shift tty into container"); | |
493 | goto out_restore_sigmask; | |
494 | } | |
495 | ||
c3e13372 | 496 | INFO("'%s' is initialized", name); |
3a0f472d | 497 | return handler; |
59eb99ba | 498 | |
b5159817 DE |
499 | out_restore_sigmask: |
500 | sigprocmask(SIG_SETMASK, &handler->oldmask, NULL); | |
59eb99ba | 501 | out_delete_tty: |
fae349da | 502 | lxc_delete_tty(&conf->tty_info); |
59eb99ba | 503 | out_aborting: |
25c2aca5 | 504 | lxc_set_state(name, handler, ABORTING); |
d2e30e99 DE |
505 | out_close_maincmd_fd: |
506 | close(conf->maincmd_fd); | |
507 | conf->maincmd_fd = -1; | |
3bdf52d7 DL |
508 | out_free_name: |
509 | free(handler->name); | |
510 | handler->name = NULL; | |
3a0f472d DL |
511 | out_free: |
512 | free(handler); | |
c3e13372 | 513 | return NULL; |
59eb99ba DL |
514 | } |
515 | ||
3b72c4a0 | 516 | void lxc_fini(const char *name, struct lxc_handler *handler) |
59eb99ba | 517 | { |
b3286b62 WB |
518 | int i, rc; |
519 | pid_t self = getpid(); | |
520 | char *namespaces[LXC_NS_MAX+1]; | |
521 | size_t namespace_count = 0; | |
b6b2b194 | 522 | |
59eb99ba DL |
523 | /* The STOPPING state is there for future cleanup code |
524 | * which can take awhile | |
525 | */ | |
25c2aca5 | 526 | lxc_set_state(name, handler, STOPPING); |
b6b2b194 | 527 | |
b3286b62 WB |
528 | for (i = 0; i < LXC_NS_MAX; i++) { |
529 | if (handler->nsfd[i] != -1) { | |
530 | rc = asprintf(&namespaces[namespace_count], "%s:/proc/%d/fd/%d", | |
531 | ns_info[i].proc_name, self, handler->nsfd[i]); | |
532 | if (rc == -1) { | |
533 | SYSERROR("failed to allocate memory"); | |
534 | break; | |
535 | } | |
536 | ++namespace_count; | |
537 | } | |
538 | } | |
539 | namespaces[namespace_count] = NULL; | |
c154af98 SG |
540 | |
541 | if (handler->conf->reboot && setenv("LXC_TARGET", "reboot", 1)) { | |
542 | SYSERROR("failed to set environment variable for stop target"); | |
543 | } | |
544 | if (!handler->conf->reboot && setenv("LXC_TARGET", "stop", 1)) { | |
545 | SYSERROR("failed to set environment variable for stop target"); | |
546 | } | |
547 | ||
b3286b62 | 548 | if (run_lxc_hooks(name, "stop", handler->conf, handler->lxcpath, namespaces)) |
8438bfbd | 549 | ERROR("failed to run stop hooks for container '%s'.", name); |
c154af98 | 550 | |
b3286b62 WB |
551 | while (namespace_count--) |
552 | free(namespaces[namespace_count]); | |
b6b2b194 WB |
553 | for (i = 0; i < LXC_NS_MAX; i++) { |
554 | if (handler->nsfd[i] != -1) { | |
555 | close(handler->nsfd[i]); | |
556 | handler->nsfd[i] = -1; | |
557 | } | |
558 | } | |
25c2aca5 | 559 | lxc_set_state(name, handler, STOPPED); |
59eb99ba | 560 | |
f3787121 | 561 | if (run_lxc_hooks(name, "post-stop", handler->conf, handler->lxcpath, NULL)) { |
773fb9ca | 562 | ERROR("failed to run post-stop hooks for container '%s'.", name); |
f3787121 CB |
563 | if (handler->conf->reboot) { |
564 | WARN("Container will be stopped instead of rebooted."); | |
565 | handler->conf->reboot = 0; | |
566 | setenv("LXC_TARGET", "stop", 1); | |
567 | } | |
568 | } | |
26ddeedd | 569 | |
83ee7875 | 570 | /* reset mask set by setup_signal_fd */ |
8f64a3f6 MN |
571 | if (sigprocmask(SIG_SETMASK, &handler->oldmask, NULL)) |
572 | WARN("failed to restore sigprocmask"); | |
573 | ||
b5159817 | 574 | lxc_console_delete(&handler->conf->console); |
b2431939 | 575 | lxc_delete_tty(&handler->conf->tty_info); |
d2e30e99 DE |
576 | close(handler->conf->maincmd_fd); |
577 | handler->conf->maincmd_fd = -1; | |
3bdf52d7 | 578 | free(handler->name); |
e8bd4e43 SH |
579 | if (handler->ttysock[0] != -1) { |
580 | close(handler->ttysock[0]); | |
581 | close(handler->ttysock[1]); | |
582 | } | |
2c5f2ede CB |
583 | |
584 | if (handler->conf->ephemeral == 1 && handler->conf->reboot != 1) | |
28272964 | 585 | lxc_destroy_container_on_signal(handler, name); |
2c5f2ede | 586 | |
d4ef7c50 | 587 | cgroup_destroy(handler); |
b2431939 | 588 | free(handler); |
59eb99ba DL |
589 | } |
590 | ||
735f2c6e | 591 | void lxc_abort(const char *name, struct lxc_handler *handler) |
59eb99ba | 592 | { |
73e608b2 SH |
593 | int ret, status; |
594 | ||
25c2aca5 | 595 | lxc_set_state(name, handler, ABORTING); |
7d9fb3e9 DL |
596 | if (handler->pid > 0) |
597 | kill(handler->pid, SIGKILL); | |
73e608b2 | 598 | while ((ret = waitpid(-1, &status, 0)) > 0) ; |
59eb99ba DL |
599 | } |
600 | ||
828695d9 SH |
601 | #include <sys/reboot.h> |
602 | #include <linux/reboot.h> | |
603 | ||
e2fa1520 SH |
604 | /* |
605 | * reboot(LINUX_REBOOT_CMD_CAD_ON) will return -EINVAL | |
606 | * in a child pid namespace if container reboot support exists. | |
607 | * Otherwise, it will either succeed or return -EPERM. | |
608 | */ | |
609 | static int container_reboot_supported(void *arg) | |
828695d9 | 610 | { |
d028235d | 611 | int *cmd = arg; |
828695d9 | 612 | int ret; |
828695d9 | 613 | |
d028235d | 614 | ret = reboot(*cmd); |
e2fa1520 SH |
615 | if (ret == -1 && errno == EINVAL) |
616 | return 1; | |
617 | return 0; | |
618 | } | |
619 | ||
b60ed720 | 620 | static int must_drop_cap_sys_boot(struct lxc_conf *conf) |
e2fa1520 | 621 | { |
025ed0f3 | 622 | FILE *f; |
b60ed720 | 623 | int ret, cmd, v, flags; |
d028235d SG |
624 | long stack_size = 4096; |
625 | void *stack = alloca(stack_size); | |
626 | int status; | |
627 | pid_t pid; | |
e2fa1520 | 628 | |
025ed0f3 | 629 | f = fopen("/proc/sys/kernel/ctrl-alt-del", "r"); |
e2fa1520 SH |
630 | if (!f) { |
631 | DEBUG("failed to open /proc/sys/kernel/ctrl-alt-del"); | |
828695d9 | 632 | return 1; |
e2fa1520 | 633 | } |
828695d9 SH |
634 | |
635 | ret = fscanf(f, "%d", &v); | |
636 | fclose(f); | |
e2fa1520 SH |
637 | if (ret != 1) { |
638 | DEBUG("Failed to read /proc/sys/kernel/ctrl-alt-del"); | |
828695d9 | 639 | return 1; |
e2fa1520 SH |
640 | } |
641 | cmd = v ? LINUX_REBOOT_CMD_CAD_ON : LINUX_REBOOT_CMD_CAD_OFF; | |
642 | ||
b60ed720 SH |
643 | flags = CLONE_NEWPID | SIGCHLD; |
644 | if (!lxc_list_empty(&conf->id_map)) | |
645 | flags |= CLONE_NEWUSER; | |
646 | ||
7f145a6d | 647 | #ifdef __ia64__ |
959aee9c | 648 | pid = __clone2(container_reboot_supported, stack, stack_size, flags, &cmd); |
7f145a6d | 649 | #else |
959aee9c SG |
650 | stack += stack_size; |
651 | pid = clone(container_reboot_supported, stack, flags, &cmd); | |
7f145a6d | 652 | #endif |
959aee9c | 653 | if (pid < 0) { |
c08220e9 | 654 | if (flags & CLONE_NEWUSER) |
4a7e5f4f | 655 | ERROR("failed to clone (%#x): %s (includes CLONE_NEWUSER)", flags, strerror(errno)); |
c08220e9 | 656 | else |
4a7e5f4f | 657 | ERROR("failed to clone (%#x): %s", flags, strerror(errno)); |
959aee9c SG |
658 | return -1; |
659 | } | |
660 | if (wait(&status) < 0) { | |
661 | SYSERROR("unexpected wait error: %m"); | |
662 | return -1; | |
663 | } | |
e2fa1520 SH |
664 | |
665 | if (WEXITSTATUS(status) != 1) | |
828695d9 | 666 | return 1; |
e2fa1520 | 667 | |
828695d9 SH |
668 | return 0; |
669 | } | |
670 | ||
658979c5 SH |
671 | /* |
672 | * netpipe is used in the unprivileged case to transfer the ifindexes | |
673 | * from parent to child | |
674 | */ | |
675 | static int netpipe = -1; | |
676 | ||
677 | static inline int count_veths(struct lxc_list *network) | |
678 | { | |
679 | struct lxc_list *iterator; | |
680 | struct lxc_netdev *netdev; | |
681 | int count = 0; | |
682 | ||
683 | lxc_list_for_each(iterator, network) { | |
684 | netdev = iterator->elem; | |
685 | if (netdev->type != LXC_NET_VETH) | |
686 | continue; | |
687 | count++; | |
688 | } | |
689 | return count; | |
690 | } | |
691 | ||
692 | static int read_unpriv_netifindex(struct lxc_list *network) | |
693 | { | |
694 | struct lxc_list *iterator; | |
695 | struct lxc_netdev *netdev; | |
696 | ||
697 | if (netpipe == -1) | |
698 | return 0; | |
699 | lxc_list_for_each(iterator, network) { | |
700 | netdev = iterator->elem; | |
701 | if (netdev->type != LXC_NET_VETH) | |
702 | continue; | |
703 | if (!(netdev->name = malloc(IFNAMSIZ))) { | |
704 | ERROR("Out of memory"); | |
705 | close(netpipe); | |
706 | return -1; | |
707 | } | |
708 | if (read(netpipe, netdev->name, IFNAMSIZ) != IFNAMSIZ) { | |
709 | close(netpipe); | |
710 | return -1; | |
711 | } | |
712 | } | |
713 | close(netpipe); | |
714 | return 0; | |
715 | } | |
716 | ||
ffe1e01a | 717 | static int do_start(void *data) |
50e98013 | 718 | { |
7c661726 | 719 | struct lxc_list *iterator; |
23c53af9 | 720 | struct lxc_handler *handler = data; |
7a55c157 TA |
721 | int devnull_fd = -1, ret; |
722 | char path[PATH_MAX]; | |
50e98013 DL |
723 | |
724 | if (sigprocmask(SIG_SETMASK, &handler->oldmask, NULL)) { | |
725 | SYSERROR("failed to set sigprocmask"); | |
9d7f9e52 | 726 | return -1; |
50e98013 DL |
727 | } |
728 | ||
d028235d | 729 | /* This prctl must be before the synchro, so if the parent |
743ecd2e DL |
730 | * dies before we set the parent death signal, we will detect |
731 | * its death with the synchro right after, otherwise we have | |
732 | * a window where the parent can exit before we set the pdeath | |
733 | * signal leading to a unsupervized container. | |
734 | */ | |
735 | if (prctl(PR_SET_PDEATHSIG, SIGKILL, 0, 0, 0)) { | |
736 | SYSERROR("failed to set pdeath signal"); | |
737 | return -1; | |
738 | } | |
739 | ||
3c22086f | 740 | lxc_sync_fini_parent(handler); |
50e98013 | 741 | |
2b0e17e4 | 742 | /* don't leak the pinfd to the container */ |
025ed0f3 | 743 | if (handler->pinfd >= 0) { |
0d03360a | 744 | close(handler->pinfd); |
025ed0f3 | 745 | } |
2b0e17e4 | 746 | |
5b1e83cb SH |
747 | if (lxc_sync_wait_parent(handler, LXC_SYNC_STARTUP)) |
748 | return -1; | |
749 | ||
750 | /* Unshare CLONE_NEWNET after CLONE_NEWUSER - see | |
751 | https://github.com/lxc/lxd/issues/1978 */ | |
752 | if ((handler->clone_flags & (CLONE_NEWNET | CLONE_NEWUSER)) == | |
753 | (CLONE_NEWNET | CLONE_NEWUSER)) { | |
754 | ret = unshare(CLONE_NEWNET); | |
755 | if (ret < 0) { | |
756 | SYSERROR("Error unsharing network namespace"); | |
757 | goto out_warn_father; | |
758 | } | |
759 | } | |
760 | ||
3c22086f CLG |
761 | /* Tell the parent task it can begin to configure the |
762 | * container and wait for it to finish | |
763 | */ | |
764 | if (lxc_sync_barrier_parent(handler, LXC_SYNC_CONFIGURE)) | |
9d7f9e52 | 765 | return -1; |
50e98013 | 766 | |
658979c5 SH |
767 | if (read_unpriv_netifindex(&handler->conf->network) < 0) |
768 | goto out_warn_father; | |
769 | ||
f6d3e3e4 SH |
770 | /* |
771 | * if we are in a new user namespace, become root there to have | |
62d00539 PT |
772 | * privilege over our namespace. When using lxc-execute we default to root, |
773 | * but this can be overriden using the lxc.init_uid and lxc.init_gid | |
774 | * configuration options. | |
f6d3e3e4 SH |
775 | */ |
776 | if (!lxc_list_empty(&handler->conf->id_map)) { | |
c5cd20ce PT |
777 | gid_t new_gid = 0; |
778 | if (handler->conf->is_execute && handler->conf->init_gid) | |
779 | new_gid = handler->conf->init_gid; | |
780 | ||
781 | uid_t new_uid = 0; | |
782 | if (handler->conf->is_execute && handler->conf->init_uid) | |
783 | new_uid = handler->conf->init_uid; | |
784 | ||
56f8ff00 PT |
785 | NOTICE("switching to gid/uid %d/%d in new user namespace", new_gid, new_uid); |
786 | if (setgid(new_gid)) { | |
f6d3e3e4 SH |
787 | SYSERROR("setgid"); |
788 | goto out_warn_father; | |
789 | } | |
56f8ff00 | 790 | if (setuid(new_uid)) { |
f6d3e3e4 SH |
791 | SYSERROR("setuid"); |
792 | goto out_warn_father; | |
793 | } | |
c476bdce SH |
794 | if (setgroups(0, NULL)) { |
795 | SYSERROR("setgroups"); | |
796 | goto out_warn_father; | |
797 | } | |
f6d3e3e4 SH |
798 | } |
799 | ||
99b71824 | 800 | if (access(handler->lxcpath, X_OK)) { |
c8154066 SH |
801 | print_top_failing_dir(handler->lxcpath); |
802 | goto out_warn_father; | |
803 | } | |
804 | ||
495d2046 | 805 | #if HAVE_SYS_CAPABILITY_H |
69182a31 | 806 | if (handler->conf->need_utmp_watch) { |
828695d9 SH |
807 | if (prctl(PR_CAPBSET_DROP, CAP_SYS_BOOT, 0, 0, 0)) { |
808 | SYSERROR("failed to remove CAP_SYS_BOOT capability"); | |
c4ea60df | 809 | goto out_warn_father; |
828695d9 | 810 | } |
959aee9c | 811 | DEBUG("Dropped cap_sys_boot"); |
e2fa1520 | 812 | } |
495d2046 | 813 | #endif |
e2fa1520 | 814 | |
a91cde21 | 815 | ret = snprintf(path, sizeof(path), "%s/dev/null", handler->conf->rootfs.mount); |
7a55c157 TA |
816 | if (ret < 0 || ret >= sizeof(path)) { |
817 | SYSERROR("sprintf'd too many chars"); | |
818 | goto out_warn_father; | |
819 | } | |
820 | ||
821 | /* In order to checkpoint restore, we need to have everything in the | |
822 | * same mount namespace. However, some containers may not have a | |
823 | * reasonable /dev (in particular, they may not have /dev/null), so we | |
824 | * can't set init's std fds to /dev/null by opening it from inside the | |
825 | * container. | |
826 | * | |
827 | * If that's the case, fall back to using the host's /dev/null. This | |
828 | * means that migration won't work, but at least we won't spew output | |
829 | * where it isn't wanted. | |
830 | */ | |
831 | if (handler->backgrounded && !handler->conf->autodev && access(path, F_OK) < 0) { | |
c44de748 AM |
832 | devnull_fd = open_devnull(); |
833 | ||
834 | if (devnull_fd < 0) | |
835 | goto out_warn_father; | |
7a55c157 | 836 | WARN("using host's /dev/null for container init's std fds, migraiton won't work"); |
c44de748 AM |
837 | } |
838 | ||
e2fa1520 | 839 | /* Setup the container, ip, names, utsname, ... */ |
d4ef7c50 | 840 | if (lxc_setup(handler)) { |
e2fa1520 SH |
841 | ERROR("failed to setup the container"); |
842 | goto out_warn_father; | |
843 | } | |
50e98013 | 844 | |
544a48a0 SH |
845 | /* ask father to setup cgroups and wait for him to finish */ |
846 | if (lxc_sync_barrier_parent(handler, LXC_SYNC_CGROUP)) | |
c44de748 | 847 | goto out_error; |
544a48a0 | 848 | |
72863294 | 849 | /* Set the label to change to when we exec(2) the container's init */ |
7aff4f43 | 850 | if (lsm_process_label_set(NULL, handler->conf, 1, 1) < 0) |
e075f5d9 | 851 | goto out_warn_father; |
5112cd70 | 852 | |
0d9acb99 DE |
853 | /* Some init's such as busybox will set sane tty settings on stdin, |
854 | * stdout, stderr which it thinks is the console. We already set them | |
855 | * the way we wanted on the real terminal, and we want init to do its | |
856 | * setup on its console ie. the pty allocated in lxc_console_create() | |
857 | * so make sure that that pty is stdin,stdout,stderr. | |
858 | */ | |
39a78bbe | 859 | if (lxc_console_set_stdfds(handler->conf->console.slave) < 0) |
0d9acb99 DE |
860 | goto out_warn_father; |
861 | ||
5112cd70 SH |
862 | /* If we mounted a temporary proc, then unmount it now */ |
863 | tmp_proc_unmount(handler->conf); | |
e075f5d9 | 864 | |
8f2c3a70 SH |
865 | if (lxc_seccomp_load(handler->conf) != 0) |
866 | goto out_warn_father; | |
867 | ||
283678ed | 868 | if (run_lxc_hooks(handler->name, "start", handler->conf, handler->lxcpath, NULL)) { |
773fb9ca SH |
869 | ERROR("failed to run start hooks for container '%s'.", handler->name); |
870 | goto out_warn_father; | |
871 | } | |
fc25b815 | 872 | |
f7bee6c6 | 873 | /* The clearenv() and putenv() calls have been moved here |
ec64264d | 874 | * to allow us to use environment variables passed to the various |
f7bee6c6 MW |
875 | * hooks, such as the start hook above. Not all of the |
876 | * variables like CONFIG_PATH or ROOTFS are valid in this | |
877 | * context but others are. */ | |
878 | if (clearenv()) { | |
879 | SYSERROR("failed to clear environment"); | |
880 | /* don't error out though */ | |
881 | } | |
882 | ||
7c661726 MP |
883 | lxc_list_for_each(iterator, &handler->conf->environment) { |
884 | if (putenv((char *)iterator->elem)) { | |
885 | SYSERROR("failed to set environment variable '%s'", (char *)iterator->elem); | |
886 | goto out_warn_father; | |
887 | } | |
888 | } | |
889 | ||
f7bee6c6 | 890 | if (putenv("container=lxc")) { |
7c661726 | 891 | SYSERROR("failed to set environment variable 'container=lxc'"); |
c4ea60df | 892 | goto out_warn_father; |
f7bee6c6 MW |
893 | } |
894 | ||
393903d1 SH |
895 | if (handler->conf->pty_names) { |
896 | if (putenv(handler->conf->pty_names)) { | |
897 | SYSERROR("failed to set environment variable for container ptys"); | |
898 | goto out_warn_father; | |
899 | } | |
900 | } | |
901 | ||
773fb9ca | 902 | close(handler->sigfd); |
26ddeedd | 903 | |
7a55c157 TA |
904 | if (devnull_fd < 0) { |
905 | devnull_fd = open_devnull(); | |
906 | ||
907 | if (devnull_fd < 0) | |
908 | goto out_warn_father; | |
909 | } | |
910 | ||
c44de748 | 911 | if (handler->backgrounded && set_stdfds(devnull_fd)) |
69aeabac | 912 | goto out_warn_father; |
507cee36 | 913 | |
c44de748 AM |
914 | if (devnull_fd >= 0) { |
915 | close(devnull_fd); | |
916 | devnull_fd = -1; | |
917 | } | |
918 | ||
12983ba4 SH |
919 | if (cgns_supported() && unshare(CLONE_NEWCGROUP) != 0) { |
920 | SYSERROR("Failed to unshare cgroup namespace"); | |
921 | goto out_warn_father; | |
922 | } | |
923 | ||
8c9a7665 TA |
924 | setsid(); |
925 | ||
e6126dbe MN |
926 | /* after this call, we are in error because this |
927 | * ops should not return as it execs */ | |
c4ea60df | 928 | handler->ops->start(handler, handler->data); |
50e98013 DL |
929 | |
930 | out_warn_father: | |
d1ccb562 AM |
931 | /* we want the parent to know something went wrong, so we return a special |
932 | * error code. */ | |
933 | lxc_sync_wake_parent(handler, LXC_SYNC_ERROR); | |
c44de748 AM |
934 | |
935 | out_error: | |
936 | if (devnull_fd >= 0) | |
937 | close(devnull_fd); | |
938 | ||
50e98013 DL |
939 | return -1; |
940 | } | |
941 | ||
74a3920a | 942 | static int save_phys_nics(struct lxc_conf *conf) |
7b35f3d6 SH |
943 | { |
944 | struct lxc_list *iterator; | |
40f2f8a2 | 945 | int am_root = (getuid() == 0); |
7b35f3d6 | 946 | |
40f2f8a2 LQ |
947 | if (!am_root) |
948 | return 0; | |
2c5f2ede | 949 | |
7b35f3d6 SH |
950 | lxc_list_for_each(iterator, &conf->network) { |
951 | struct lxc_netdev *netdev = iterator->elem; | |
952 | ||
953 | if (netdev->type != LXC_NET_PHYS) | |
954 | continue; | |
955 | conf->saved_nics = realloc(conf->saved_nics, | |
956 | (conf->num_savednics+1)*sizeof(struct saved_nic)); | |
957 | if (!conf->saved_nics) { | |
958 | SYSERROR("failed to allocate memory"); | |
959 | return -1; | |
960 | } | |
961 | conf->saved_nics[conf->num_savednics].ifindex = netdev->ifindex; | |
962 | conf->saved_nics[conf->num_savednics].orig_name = strdup(netdev->link); | |
963 | if (!conf->saved_nics[conf->num_savednics].orig_name) { | |
964 | SYSERROR("failed to allocate memory"); | |
965 | return -1; | |
966 | } | |
959aee9c | 967 | INFO("stored saved_nic #%d idx %d name %s", conf->num_savednics, |
7b35f3d6 SH |
968 | conf->saved_nics[conf->num_savednics].ifindex, |
969 | conf->saved_nics[conf->num_savednics].orig_name); | |
970 | conf->num_savednics++; | |
971 | } | |
972 | ||
973 | return 0; | |
974 | } | |
975 | ||
e8bd4e43 SH |
976 | static int recv_fd(int sock, int *fd) |
977 | { | |
978 | if (lxc_abstract_unix_recv_fd(sock, fd, NULL, 0) < 0) { | |
979 | SYSERROR("Error receiving tty fd from child"); | |
980 | return -1; | |
981 | } | |
982 | if (*fd == -1) | |
983 | return -1; | |
984 | return 0; | |
985 | } | |
986 | ||
987 | static int recv_ttys_from_child(struct lxc_handler *handler) | |
988 | { | |
989 | struct lxc_conf *conf = handler->conf; | |
990 | int i, sock = handler->ttysock[1]; | |
991 | struct lxc_tty_info *tty_info = &conf->tty_info; | |
992 | ||
993 | if (!conf->tty) | |
994 | return 0; | |
995 | ||
996 | tty_info->pty_info = malloc(sizeof(*tty_info->pty_info)*conf->tty); | |
997 | if (!tty_info->pty_info) { | |
998 | SYSERROR("failed to allocate pty_info"); | |
999 | return -1; | |
1000 | } | |
1001 | ||
1002 | for (i = 0; i < conf->tty; i++) { | |
1003 | struct lxc_pty_info *pty_info = &tty_info->pty_info[i]; | |
1004 | pty_info->busy = 0; | |
1005 | if (recv_fd(sock, &pty_info->slave) < 0 || | |
1006 | recv_fd(sock, &pty_info->master) < 0) { | |
1007 | ERROR("Error receiving tty info from child"); | |
1008 | return -1; | |
1009 | } | |
1010 | } | |
1011 | tty_info->nbtty = conf->tty; | |
1012 | ||
1013 | return 0; | |
1014 | } | |
1015 | ||
f813849c TA |
1016 | void resolve_clone_flags(struct lxc_handler *handler) |
1017 | { | |
1018 | handler->clone_flags = CLONE_NEWPID | CLONE_NEWNS; | |
1019 | ||
1020 | if (!lxc_list_empty(&handler->conf->id_map)) { | |
1021 | INFO("Cloning a new user namespace"); | |
1022 | handler->clone_flags |= CLONE_NEWUSER; | |
1023 | } | |
1024 | ||
1025 | if (handler->conf->inherit_ns_fd[LXC_NS_NET] == -1) { | |
1026 | if (!lxc_requests_empty_network(handler)) | |
1027 | handler->clone_flags |= CLONE_NEWNET; | |
1028 | } else { | |
1029 | INFO("Inheriting a net namespace"); | |
1030 | } | |
1031 | ||
1032 | if (handler->conf->inherit_ns_fd[LXC_NS_IPC] == -1) { | |
1033 | handler->clone_flags |= CLONE_NEWIPC; | |
1034 | } else { | |
1035 | INFO("Inheriting an IPC namespace"); | |
1036 | } | |
1037 | ||
1038 | if (handler->conf->inherit_ns_fd[LXC_NS_UTS] == -1) { | |
1039 | handler->clone_flags |= CLONE_NEWUTS; | |
1040 | } else { | |
1041 | INFO("Inheriting a UTS namespace"); | |
1042 | } | |
1043 | } | |
1044 | ||
74a3920a | 1045 | static int lxc_spawn(struct lxc_handler *handler) |
59eb99ba | 1046 | { |
b98f7d6e | 1047 | int failed_before_rename = 0; |
ffe1e01a | 1048 | const char *name = handler->name; |
62d05d9b | 1049 | char *errmsg = NULL; |
7e4dfe0b | 1050 | bool cgroups_connected = false; |
9f30a190 | 1051 | int saved_ns_fd[LXC_NS_MAX]; |
5b1e83cb | 1052 | int preserve_mask = 0, i, flags; |
658979c5 | 1053 | int netpipepair[2], nveths; |
9f30a190 | 1054 | |
2f2623ec SH |
1055 | netpipe = -1; |
1056 | ||
9f30a190 | 1057 | for (i = 0; i < LXC_NS_MAX; i++) |
6c544cb3 | 1058 | if (handler->conf->inherit_ns_fd[i] != -1) |
9f30a190 | 1059 | preserve_mask |= ns_info[i].clone_flag; |
50e98013 | 1060 | |
3c22086f | 1061 | if (lxc_sync_init(handler)) |
9d7f9e52 | 1062 | return -1; |
0ad19a3f | 1063 | |
e8bd4e43 SH |
1064 | if (socketpair(AF_UNIX, SOCK_DGRAM, 0, handler->ttysock) < 0) { |
1065 | lxc_sync_fini(handler); | |
1066 | return -1; | |
1067 | } | |
1068 | ||
f813849c | 1069 | resolve_clone_flags(handler); |
9f30a190 | 1070 | |
f813849c | 1071 | if (handler->clone_flags & CLONE_NEWNET) { |
26b797f3 SH |
1072 | if (!lxc_list_empty(&handler->conf->network)) { |
1073 | ||
9f30a190 MM |
1074 | /* Find gateway addresses from the link device, which is |
1075 | * no longer accessible inside the container. Do this | |
1076 | * before creating network interfaces, since goto | |
1077 | * out_delete_net does not work before lxc_clone. */ | |
1078 | if (lxc_find_gateway_addresses(handler)) { | |
1079 | ERROR("failed to find gateway addresses"); | |
1080 | lxc_sync_fini(handler); | |
1081 | return -1; | |
1082 | } | |
1083 | ||
1084 | /* that should be done before the clone because we will | |
1085 | * fill the netdev index and use them in the child | |
1086 | */ | |
1087 | if (lxc_create_network(handler)) { | |
38521a3b | 1088 | ERROR("failed to create the network"); |
9f30a190 MM |
1089 | lxc_sync_fini(handler); |
1090 | return -1; | |
1091 | } | |
19a26f82 MK |
1092 | } |
1093 | ||
9f30a190 MM |
1094 | if (save_phys_nics(handler->conf)) { |
1095 | ERROR("failed to save physical nic info"); | |
1096 | goto out_abort; | |
82d5ae15 DL |
1097 | } |
1098 | } | |
1099 | ||
d4ef7c50 SH |
1100 | if (!cgroup_init(handler)) { |
1101 | ERROR("failed initializing cgroup support"); | |
33ad9f1a CS |
1102 | goto out_delete_net; |
1103 | } | |
1104 | ||
7e4dfe0b SH |
1105 | cgroups_connected = true; |
1106 | ||
d4ef7c50 SH |
1107 | if (!cgroup_create(handler)) { |
1108 | ERROR("failed creating cgroups"); | |
47d8fb3b CS |
1109 | goto out_delete_net; |
1110 | } | |
1111 | ||
0c547523 SH |
1112 | /* |
1113 | * if the rootfs is not a blockdev, prevent the container from | |
1114 | * marking it readonly. | |
5e32a990 ÇO |
1115 | * |
1116 | * if the container is unprivileged then skip rootfs pinning | |
0c547523 | 1117 | */ |
5e32a990 ÇO |
1118 | if (lxc_list_empty(&handler->conf->id_map)) { |
1119 | handler->pinfd = pin_rootfs(handler->conf->rootfs.path); | |
1120 | if (handler->pinfd == -1) | |
1121 | INFO("failed to pin the container's rootfs"); | |
1122 | } | |
0c547523 | 1123 | |
62d05d9b SH |
1124 | if (!preserve_ns(saved_ns_fd, preserve_mask, getpid(), &errmsg)) { |
1125 | SYSERROR("Failed to preserve requested namespaces: %s", | |
1126 | errmsg ? errmsg : "(Out of memory)"); | |
1127 | free(errmsg); | |
cd43d2d1 | 1128 | goto out_delete_net; |
62d05d9b | 1129 | } |
cd43d2d1 SH |
1130 | if (attach_ns(handler->conf->inherit_ns_fd) < 0) |
1131 | goto out_delete_net; | |
9f30a190 | 1132 | |
658979c5 SH |
1133 | if (am_unpriv() && (nveths = count_veths(&handler->conf->network))) { |
1134 | if (pipe(netpipepair) < 0) { | |
1135 | SYSERROR("Error creating pipe"); | |
1136 | goto out_delete_net; | |
1137 | } | |
1138 | /* store netpipe in the global var for do_start's use */ | |
1139 | netpipe = netpipepair[0]; | |
1140 | } | |
1141 | ||
0ad19a3f | 1142 | /* Create a process in a new set of namespaces */ |
5b1e83cb SH |
1143 | flags = handler->clone_flags; |
1144 | if (handler->clone_flags & CLONE_NEWUSER) | |
1145 | flags &= ~CLONE_NEWNET; | |
d5088cf2 | 1146 | handler->pid = lxc_clone(do_start, handler, handler->clone_flags); |
59eb99ba | 1147 | if (handler->pid < 0) { |
36eb9bde | 1148 | SYSERROR("failed to fork into a new namespace"); |
7fef7a06 | 1149 | goto out_delete_net; |
0ad19a3f | 1150 | } |
1151 | ||
330cbe36 | 1152 | if (!preserve_ns(handler->nsfd, handler->clone_flags | preserve_mask, handler->pid, &errmsg)) { |
62d05d9b SH |
1153 | INFO("Failed to store namespace references for stop hook: %s", |
1154 | errmsg ? errmsg : "(Out of memory)"); | |
1155 | free(errmsg); | |
b6b2b194 WB |
1156 | } |
1157 | ||
cd43d2d1 SH |
1158 | if (attach_ns(saved_ns_fd)) |
1159 | WARN("failed to restore saved namespaces"); | |
9f30a190 | 1160 | |
3c22086f CLG |
1161 | lxc_sync_fini_child(handler); |
1162 | ||
5b1e83cb SH |
1163 | /* map the container uids - the container became an invalid |
1164 | * userid the moment it was cloned with CLONE_NEWUSER - this | |
1165 | * call doesn't change anything immediately, but allows the | |
1166 | * container to setuid(0) (0 being mapped to something else on | |
1167 | * the host) later to become a valid uid again */ | |
1168 | if (lxc_map_ids(&handler->conf->id_map, handler->pid)) { | |
1169 | ERROR("failed to set up id mapping"); | |
1170 | goto out_delete_net; | |
1171 | } | |
1172 | ||
1173 | if (lxc_sync_wake_child(handler, LXC_SYNC_STARTUP)) { | |
99a6af52 | 1174 | failed_before_rename = 1; |
5b1e83cb SH |
1175 | goto out_delete_net; |
1176 | } | |
1177 | ||
1178 | if (lxc_sync_wait_child(handler, LXC_SYNC_CONFIGURE)) { | |
1179 | failed_before_rename = 1; | |
1180 | goto out_delete_net; | |
1181 | } | |
0ad19a3f | 1182 | |
d4ef7c50 SH |
1183 | if (!cgroup_create_legacy(handler)) { |
1184 | ERROR("failed to setup the legacy cgroups for %s", name); | |
ae5c8b8e | 1185 | goto out_delete_net; |
33ad9f1a | 1186 | } |
9daf6f5d SH |
1187 | if (!cgroup_setup_limits(handler, false)) { |
1188 | ERROR("failed to setup the cgroup limits for '%s'", name); | |
6031a6e5 DE |
1189 | goto out_delete_net; |
1190 | } | |
1191 | ||
d4ef7c50 | 1192 | if (!cgroup_enter(handler)) |
7fef7a06 | 1193 | goto out_delete_net; |
218d4250 | 1194 | |
0996e18a SH |
1195 | if (!cgroup_chown(handler)) |
1196 | goto out_delete_net; | |
1197 | ||
99a6af52 MN |
1198 | if (failed_before_rename) |
1199 | goto out_delete_net; | |
1200 | ||
0ad19a3f | 1201 | /* Create the network configuration */ |
d5088cf2 | 1202 | if (handler->clone_flags & CLONE_NEWNET) { |
c43cbc04 SH |
1203 | if (lxc_assign_network(handler->lxcpath, handler->name, |
1204 | &handler->conf->network, handler->pid)) { | |
82d5ae15 | 1205 | ERROR("failed to create the configured network"); |
7fef7a06 | 1206 | goto out_delete_net; |
82d5ae15 | 1207 | } |
0ad19a3f | 1208 | } |
1209 | ||
658979c5 SH |
1210 | if (netpipe != -1) { |
1211 | struct lxc_list *iterator; | |
1212 | struct lxc_netdev *netdev; | |
1213 | ||
1214 | close(netpipe); | |
1215 | lxc_list_for_each(iterator, &handler->conf->network) { | |
1216 | netdev = iterator->elem; | |
1217 | if (netdev->type != LXC_NET_VETH) | |
1218 | continue; | |
1219 | if (write(netpipepair[1], netdev->name, IFNAMSIZ) != IFNAMSIZ) { | |
1220 | ERROR("Error writing veth name to container"); | |
1221 | goto out_delete_net; | |
1222 | } | |
1223 | } | |
1224 | close(netpipepair[1]); | |
1225 | } | |
1226 | ||
544a48a0 SH |
1227 | /* Tell the child to continue its initialization. we'll get |
1228 | * LXC_SYNC_CGROUP when it is ready for us to setup cgroups | |
3c22086f CLG |
1229 | */ |
1230 | if (lxc_sync_barrier_child(handler, LXC_SYNC_POST_CONFIGURE)) | |
544a48a0 SH |
1231 | goto out_delete_net; |
1232 | ||
9daf6f5d | 1233 | if (!cgroup_setup_limits(handler, true)) { |
b98f7d6e SH |
1234 | ERROR("failed to setup the devices cgroup for '%s'", name); |
1235 | goto out_delete_net; | |
544a48a0 SH |
1236 | } |
1237 | ||
73d28d42 | 1238 | cgroup_disconnect(); |
7e4dfe0b | 1239 | cgroups_connected = false; |
73d28d42 | 1240 | |
e8bd4e43 SH |
1241 | /* read tty fds allocated by child */ |
1242 | if (recv_ttys_from_child(handler) < 0) { | |
1243 | ERROR("failed to receive tty info from child"); | |
1244 | goto out_delete_net; | |
1245 | } | |
1246 | ||
544a48a0 SH |
1247 | /* Tell the child to complete its initialization and wait for |
1248 | * it to exec or return an error. (the child will never | |
1249 | * return LXC_SYNC_POST_CGROUP+1. It will either close the | |
1250 | * sync pipe, causing lxc_sync_barrier_child to return | |
1251 | * success, or return a different value, causing us to error | |
1252 | * out). | |
1253 | */ | |
1254 | if (lxc_sync_barrier_child(handler, LXC_SYNC_POST_CGROUP)) | |
3c22086f | 1255 | return -1; |
0ad19a3f | 1256 | |
cc28d0b0 SH |
1257 | if (detect_shared_rootfs()) |
1258 | umount2(handler->conf->rootfs.mount, MNT_DETACH); | |
1259 | ||
23c53af9 | 1260 | if (handler->ops->post_start(handler, handler->data)) |
e6126dbe MN |
1261 | goto out_abort; |
1262 | ||
25c2aca5 | 1263 | if (lxc_set_state(name, handler, RUNNING)) { |
59eb99ba DL |
1264 | ERROR("failed to set state to %s", |
1265 | lxc_state2str(RUNNING)); | |
1266 | goto out_abort; | |
3f21c114 | 1267 | } |
22ebac19 | 1268 | |
3c22086f | 1269 | lxc_sync_fini(handler); |
0c547523 | 1270 | |
e6126dbe | 1271 | return 0; |
1ac470c0 | 1272 | |
7fef7a06 | 1273 | out_delete_net: |
7e4dfe0b SH |
1274 | if (cgroups_connected) |
1275 | cgroup_disconnect(); | |
d5088cf2 | 1276 | if (handler->clone_flags & CLONE_NEWNET) |
74a2b586 | 1277 | lxc_delete_network(handler); |
59eb99ba DL |
1278 | out_abort: |
1279 | lxc_abort(name, handler); | |
3c22086f | 1280 | lxc_sync_fini(handler); |
5c068da9 SH |
1281 | if (handler->pinfd >= 0) { |
1282 | close(handler->pinfd); | |
1283 | handler->pinfd = -1; | |
1284 | } | |
1285 | ||
b79fcd86 | 1286 | return -1; |
59eb99ba | 1287 | } |
0ad19a3f | 1288 | |
2af6bd1b SH |
1289 | int get_netns_fd(int pid) |
1290 | { | |
1291 | char path[MAXPATHLEN]; | |
1292 | int ret, fd; | |
1293 | ||
1294 | ret = snprintf(path, MAXPATHLEN, "/proc/%d/ns/net", pid); | |
1295 | if (ret < 0 || ret >= MAXPATHLEN) { | |
1296 | WARN("Failed to pin netns file for pid %d", pid); | |
1297 | return -1; | |
1298 | } | |
1299 | ||
1300 | fd = open(path, O_RDONLY); | |
1301 | if (fd < 0) { | |
1302 | WARN("Failed to pin netns file %s for pid %d: %s", | |
1303 | path, pid, strerror(errno)); | |
1304 | return -1; | |
1305 | } | |
1306 | return fd; | |
1307 | } | |
1308 | ||
ee70bf78 | 1309 | int __lxc_start(const char *name, struct lxc_conf *conf, |
507cee36 TA |
1310 | struct lxc_operations* ops, void *data, const char *lxcpath, |
1311 | bool backgrounded) | |
59eb99ba | 1312 | { |
3a0f472d | 1313 | struct lxc_handler *handler; |
e043236e | 1314 | int err = -1; |
59eb99ba | 1315 | int status; |
2af6bd1b | 1316 | int netnsfd = -1; |
80090207 | 1317 | |
13f5be62 | 1318 | handler = lxc_init(name, conf, lxcpath); |
3a0f472d | 1319 | if (!handler) { |
59eb99ba | 1320 | ERROR("failed to initialize the container"); |
66aeffc7 | 1321 | return -1; |
0ad19a3f | 1322 | } |
ee70bf78 CLG |
1323 | handler->ops = ops; |
1324 | handler->data = data; | |
507cee36 | 1325 | handler->backgrounded = backgrounded; |
e6126dbe | 1326 | |
b60ed720 | 1327 | if (must_drop_cap_sys_boot(handler->conf)) { |
495d2046 | 1328 | #if HAVE_SYS_CAPABILITY_H |
959aee9c | 1329 | DEBUG("Dropping cap_sys_boot"); |
495d2046 | 1330 | #else |
959aee9c | 1331 | DEBUG("Can't drop cap_sys_boot as capabilities aren't supported"); |
495d2046 | 1332 | #endif |
69182a31 | 1333 | } else { |
959aee9c | 1334 | DEBUG("Not dropping cap_sys_boot or watching utmp"); |
69182a31 SH |
1335 | handler->conf->need_utmp_watch = 0; |
1336 | } | |
1337 | ||
76a26f55 SH |
1338 | if (!attach_block_device(handler->conf)) { |
1339 | ERROR("Failure attaching block device"); | |
1340 | goto out_fini_nonet; | |
1341 | } | |
1342 | ||
35120d9c SH |
1343 | if (geteuid() == 0 && !lxc_list_empty(&conf->id_map)) { |
1344 | /* if the backing store is a device, mount it here and now */ | |
1345 | if (rootfs_is_blockdev(conf)) { | |
1346 | if (unshare(CLONE_NEWNS) < 0) { | |
1347 | ERROR("Error unsharing mounts"); | |
1348 | goto out_fini_nonet; | |
1349 | } | |
6a0c909a | 1350 | remount_all_slave(); |
35120d9c SH |
1351 | if (do_rootfs_setup(conf, name, lxcpath) < 0) { |
1352 | ERROR("Error setting up rootfs mount as root before spawn"); | |
1353 | goto out_fini_nonet; | |
1354 | } | |
1355 | INFO("Set up container rootfs as host root"); | |
1356 | } | |
1357 | } | |
1358 | ||
23c53af9 | 1359 | err = lxc_spawn(handler); |
59eb99ba | 1360 | if (err) { |
ee70bf78 | 1361 | ERROR("failed to spawn '%s'", name); |
76a26f55 | 1362 | goto out_detach_blockdev; |
0ad19a3f | 1363 | } |
1364 | ||
8bee8851 WB |
1365 | handler->conf->reboot = 0; |
1366 | ||
2af6bd1b SH |
1367 | netnsfd = get_netns_fd(handler->pid); |
1368 | ||
3a0f472d | 1369 | err = lxc_poll(name, handler); |
e043236e | 1370 | if (err) { |
59eb99ba | 1371 | ERROR("mainloop exited with an error"); |
4b2f98f5 SH |
1372 | if (netnsfd >= 0) |
1373 | close(netnsfd); | |
59eb99ba DL |
1374 | goto out_abort; |
1375 | } | |
0ad19a3f | 1376 | |
3a0f472d | 1377 | while (waitpid(handler->pid, &status, 0) < 0 && errno == EINTR) |
1bc5cc8c | 1378 | continue; |
e043236e | 1379 | |
8b004f07 SH |
1380 | /* |
1381 | * If the child process exited but was not signaled, | |
1382 | * it didn't call reboot. This should mean it was an | |
1383 | * lxc-execute which simply exited. In any case, treat | |
1384 | * it as a 'halt' | |
1385 | */ | |
d028235d | 1386 | if (WIFSIGNALED(status)) { |
8b004f07 SH |
1387 | switch(WTERMSIG(status)) { |
1388 | case SIGINT: /* halt */ | |
1389 | DEBUG("Container halting"); | |
1390 | break; | |
1391 | case SIGHUP: /* reboot */ | |
1392 | DEBUG("Container rebooting"); | |
1393 | handler->conf->reboot = 1; | |
1394 | break; | |
c2b9bd9e SH |
1395 | case SIGSYS: /* seccomp */ |
1396 | DEBUG("Container violated its seccomp policy"); | |
1397 | break; | |
8b004f07 | 1398 | default: |
959aee9c | 1399 | DEBUG("unknown exit status for init: %d", WTERMSIG(status)); |
8b004f07 SH |
1400 | break; |
1401 | } | |
d028235d | 1402 | } |
828695d9 | 1403 | |
ce5782df | 1404 | DEBUG("Pushing physical nics back to host namespace"); |
2af6bd1b | 1405 | lxc_rename_phys_nics_on_shutdown(netnsfd, handler->conf); |
ce5782df MH |
1406 | |
1407 | DEBUG("Tearing down virtual network devices used by container"); | |
1408 | lxc_delete_network(handler); | |
1409 | ||
4b2f98f5 SH |
1410 | if (netnsfd >= 0) |
1411 | close(netnsfd); | |
7b35f3d6 | 1412 | |
5c068da9 SH |
1413 | if (handler->pinfd >= 0) { |
1414 | close(handler->pinfd); | |
1415 | handler->pinfd = -1; | |
1416 | } | |
1417 | ||
1787abca | 1418 | lxc_monitor_send_exit_code(name, status, handler->lxcpath); |
3a0f472d | 1419 | err = lxc_error_set_and_log(handler->pid, status); |
9d7f9e52 | 1420 | out_fini: |
74a2b586 JK |
1421 | lxc_delete_network(handler); |
1422 | ||
76a26f55 SH |
1423 | out_detach_blockdev: |
1424 | detach_block_device(handler->conf); | |
1425 | ||
74a2b586 | 1426 | out_fini_nonet: |
3a0f472d | 1427 | lxc_fini(name, handler); |
0ad19a3f | 1428 | return err; |
1429 | ||
59eb99ba | 1430 | out_abort: |
3a0f472d | 1431 | lxc_abort(name, handler); |
9d7f9e52 | 1432 | goto out_fini; |
0ad19a3f | 1433 | } |
ee70bf78 CLG |
1434 | |
1435 | struct start_args { | |
1436 | char *const *argv; | |
1437 | }; | |
1438 | ||
1439 | static int start(struct lxc_handler *handler, void* data) | |
1440 | { | |
1441 | struct start_args *arg = data; | |
1442 | ||
1443 | NOTICE("exec'ing '%s'", arg->argv[0]); | |
1444 | ||
1445 | execvp(arg->argv[0], arg->argv); | |
1446 | SYSERROR("failed to exec %s", arg->argv[0]); | |
1447 | return 0; | |
1448 | } | |
1449 | ||
1450 | static int post_start(struct lxc_handler *handler, void* data) | |
1451 | { | |
1452 | struct start_args *arg = data; | |
1453 | ||
1454 | NOTICE("'%s' started with pid '%d'", arg->argv[0], handler->pid); | |
1455 | return 0; | |
1456 | } | |
1457 | ||
1458 | static struct lxc_operations start_ops = { | |
1459 | .start = start, | |
1460 | .post_start = post_start | |
1461 | }; | |
1462 | ||
13f5be62 | 1463 | int lxc_start(const char *name, char *const argv[], struct lxc_conf *conf, |
507cee36 | 1464 | const char *lxcpath, bool backgrounded) |
ee70bf78 CLG |
1465 | { |
1466 | struct start_args start_arg = { | |
1467 | .argv = argv, | |
1468 | }; | |
1469 | ||
828695d9 | 1470 | conf->need_utmp_watch = 1; |
507cee36 | 1471 | return __lxc_start(name, conf, &start_ops, &start_arg, lxcpath, backgrounded); |
ee70bf78 | 1472 | } |
28272964 CB |
1473 | |
1474 | static void lxc_destroy_container_on_signal(struct lxc_handler *handler, | |
1475 | const char *name) | |
1476 | { | |
1477 | char destroy[MAXPATHLEN]; | |
1478 | bool bret = true; | |
1479 | int ret = 0; | |
f01f7975 | 1480 | struct lxc_container *c; |
df31363a | 1481 | if (handler->conf->rootfs.path && handler->conf->rootfs.mount) { |
28272964 CB |
1482 | bret = do_destroy_container(handler->conf); |
1483 | if (!bret) { | |
1484 | ERROR("Error destroying rootfs for %s", name); | |
1485 | return; | |
1486 | } | |
1487 | } | |
1488 | INFO("Destroyed rootfs for %s", name); | |
1489 | ||
1490 | ret = snprintf(destroy, MAXPATHLEN, "%s/%s", handler->lxcpath, name); | |
1491 | if (ret < 0 || ret >= MAXPATHLEN) { | |
1492 | ERROR("Error printing path for %s", name); | |
1493 | ERROR("Error destroying directory for %s", name); | |
1494 | return; | |
1495 | } | |
1496 | ||
f01f7975 CB |
1497 | c = lxc_container_new(name, handler->lxcpath); |
1498 | if (c) { | |
1499 | if (container_disk_lock(c)) { | |
1500 | INFO("Could not update lxc_snapshots file"); | |
1501 | lxc_container_put(c); | |
1502 | } else { | |
1503 | mod_all_rdeps(c, false); | |
1504 | container_disk_unlock(c); | |
1505 | lxc_container_put(c); | |
1506 | } | |
1507 | } | |
1508 | ||
28272964 CB |
1509 | if (am_unpriv()) |
1510 | ret = userns_exec_1(handler->conf, lxc_rmdir_onedev_wrapper, destroy); | |
1511 | else | |
1512 | ret = lxc_rmdir_onedev(destroy, NULL); | |
1513 | ||
1514 | if (ret < 0) { | |
1515 | ERROR("Error destroying directory for %s", name); | |
1516 | return; | |
1517 | } | |
1518 | INFO("Destroyed directory for %s", name); | |
1519 | } | |
1520 | ||
1521 | static int lxc_rmdir_onedev_wrapper(void *data) | |
1522 | { | |
1523 | char *arg = (char *) data; | |
1524 | return lxc_rmdir_onedev(arg, NULL); | |
1525 | } | |
1526 | ||
1527 | static bool do_destroy_container(struct lxc_conf *conf) { | |
d028235d SG |
1528 | if (am_unpriv()) { |
1529 | if (userns_exec_1(conf, bdev_destroy_wrapper, conf) < 0) | |
1530 | return false; | |
1531 | return true; | |
1532 | } | |
1533 | return bdev_destroy(conf); | |
28272964 | 1534 | } |