3 * Copyright © 2012 Serge Hallyn <serge.hallyn@ubuntu.com>.
4 * Copyright © 2012 Canonical Ltd.
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
26 #include <sys/types.h>
28 #include <sys/mount.h>
34 #include <arpa/inet.h>
38 #include <sys/syscall.h>
40 #include <lxc/lxccontainer.h>
41 #include <lxc/version.h>
56 #include "namespace.h"
62 #include <../include/ifaddrs.h>
65 #define MAX_BUFFER 4096
67 #define NOT_SUPPORTED_ERROR "the requested function %s is not currently supported with unprivileged containers"
69 /* Define faccessat() if missing from the C library */
70 #ifndef HAVE_FACCESSAT
71 static int faccessat(int __fd
, const char *__file
, int __type
, int __flag
)
74 return syscall(__NR_faccessat
, __fd
, __file
, __type
, __flag
);
83 lxc_log_define(lxc_container
, lxc
);
85 static bool config_file_exists(const char *lxcpath
, const char *cname
)
87 /* $lxcpath + '/' + $cname + '/config' + \0 */
88 int ret
, len
= strlen(lxcpath
) + strlen(cname
) + 9;
89 char *fname
= alloca(len
);
91 ret
= snprintf(fname
, len
, "%s/%s/config", lxcpath
, cname
);
92 if (ret
< 0 || ret
>= len
)
95 return file_exists(fname
);
99 * A few functions to help detect when a container creation failed.
100 * If a container creation was killed partway through, then trying
101 * to actually start that container could harm the host. We detect
102 * this by creating a 'partial' file under the container directory,
103 * and keeping an advisory lock. When container creation completes,
104 * we remove that file. When we load or try to start a container, if
105 * we find that file, without a flock, we remove the container.
107 static int ongoing_create(struct lxc_container
*c
)
109 int len
= strlen(c
->config_path
) + strlen(c
->name
) + 10;
110 char *path
= alloca(len
);
114 ret
= snprintf(path
, len
, "%s/%s/partial", c
->config_path
, c
->name
);
115 if (ret
< 0 || ret
>= len
) {
116 ERROR("Error writing partial pathname");
120 if (!file_exists(path
))
122 fd
= open(path
, O_RDWR
);
124 // give benefit of the doubt
125 SYSERROR("Error opening partial file");
129 lk
.l_whence
= SEEK_SET
;
133 if (fcntl(fd
, F_GETLK
, &lk
) == 0 && lk
.l_pid
!= -1) {
134 // create is still ongoing
138 // create completed but partial is still there.
143 static int create_partial(struct lxc_container
*c
)
145 // $lxcpath + '/' + $name + '/partial' + \0
146 int len
= strlen(c
->config_path
) + strlen(c
->name
) + 10;
147 char *path
= alloca(len
);
151 ret
= snprintf(path
, len
, "%s/%s/partial", c
->config_path
, c
->name
);
152 if (ret
< 0 || ret
>= len
) {
153 ERROR("Error writing partial pathname");
156 if ((fd
=open(path
, O_RDWR
| O_CREAT
| O_EXCL
, 0755)) < 0) {
157 SYSERROR("Erorr creating partial file");
161 lk
.l_whence
= SEEK_SET
;
164 if (fcntl(fd
, F_SETLKW
, &lk
) < 0) {
165 SYSERROR("Error locking partial file %s", path
);
173 static void remove_partial(struct lxc_container
*c
, int fd
)
175 // $lxcpath + '/' + $name + '/partial' + \0
176 int len
= strlen(c
->config_path
) + strlen(c
->name
) + 10;
177 char *path
= alloca(len
);
181 ret
= snprintf(path
, len
, "%s/%s/partial", c
->config_path
, c
->name
);
182 if (ret
< 0 || ret
>= len
) {
183 ERROR("Error writing partial pathname");
186 if (unlink(path
) < 0)
187 SYSERROR("Error unlink partial file %s", path
);
191 * 1. container_mem_lock(c) protects the struct lxc_container from multiple threads.
192 * 2. container_disk_lock(c) protects the on-disk container data - in particular the
193 * container configuration file.
194 * The container_disk_lock also takes the container_mem_lock.
195 * 3. thread_mutex protects process data (ex: fd table) from multiple threads.
196 * NOTHING mutexes two independent programs with their own struct
197 * lxc_container for the same c->name, between API calls. For instance,
198 * c->config_read(); c->start(); Between those calls, data on disk
199 * could change (which shouldn't bother the caller unless for instance
200 * the rootfs get moved). c->config_read(); update; c->config_write();
201 * Two such updaters could race. The callers should therefore check their
202 * results. Trying to prevent that would necessarily expose us to deadlocks
203 * due to hung callers. So I prefer to keep the locks only within our own
204 * functions, not across functions.
206 * If you're going to clone while holding a lxccontainer, increment
207 * c->numthreads (under privlock) before forking. When deleting,
208 * decrement numthreads under privlock, then if it hits 0 you can delete.
209 * Do not ever use a lxccontainer whose numthreads you did not bump.
212 static void lxc_container_free(struct lxc_container
*c
)
219 c
->configfile
= NULL
;
221 if (c
->error_string
) {
222 free(c
->error_string
);
223 c
->error_string
= NULL
;
226 lxc_putlock(c
->slock
);
230 lxc_putlock(c
->privlock
);
238 lxc_conf_free(c
->lxc_conf
);
241 if (c
->config_path
) {
242 free(c
->config_path
);
243 c
->config_path
= NULL
;
250 * Consider the following case:
251 freer | racing get()er
252 ==================================================================
253 lxc_container_put() | lxc_container_get()
254 \ lxclock(c->privlock) | c->numthreads < 1? (no)
255 \ c->numthreads = 0 | \ lxclock(c->privlock) -> waits
257 \ lxc_container_free() | \ lxclock() returns
258 | \ c->numthreads < 1 -> return 0
260 \ \ sem_destroy(privlock) |
262 * When the get()er checks numthreads the first time, one of the following
264 * 1. freer has set numthreads = 0. get() returns 0
265 * 2. freer is between lxclock and setting numthreads to 0. get()er will
266 * sem_wait on privlock, get lxclock after freer() drops it, then see
267 * numthreads is 0 and exit without touching lxclock again..
268 * 3. freer has not yet locked privlock. If get()er runs first, then put()er
269 * will see --numthreads = 1 and not call lxc_container_free().
272 int lxc_container_get(struct lxc_container
*c
)
277 // if someone else has already started freeing the container, don't
278 // try to take the lock, which may be invalid
279 if (c
->numthreads
< 1)
282 if (container_mem_lock(c
))
284 if (c
->numthreads
< 1) {
285 // bail without trying to unlock, bc the privlock is now probably
290 container_mem_unlock(c
);
294 int lxc_container_put(struct lxc_container
*c
)
298 if (container_mem_lock(c
))
300 if (--c
->numthreads
< 1) {
301 container_mem_unlock(c
);
302 lxc_container_free(c
);
305 container_mem_unlock(c
);
309 static bool lxcapi_is_defined(struct lxc_container
*c
)
318 if (container_mem_lock(c
))
322 statret
= stat(c
->configfile
, &statbuf
);
328 container_mem_unlock(c
);
332 static const char *lxcapi_state(struct lxc_container
*c
)
338 s
= lxc_getstate(c
->name
, c
->config_path
);
339 return lxc_state2str(s
);
342 static bool is_stopped(struct lxc_container
*c
)
345 s
= lxc_getstate(c
->name
, c
->config_path
);
346 return (s
== STOPPED
);
349 static bool lxcapi_is_running(struct lxc_container
*c
)
356 if (!s
|| strcmp(s
, "STOPPED") == 0)
361 static bool lxcapi_freeze(struct lxc_container
*c
)
367 ret
= lxc_freeze(c
->name
, c
->config_path
);
373 static bool lxcapi_unfreeze(struct lxc_container
*c
)
379 ret
= lxc_unfreeze(c
->name
, c
->config_path
);
385 static int lxcapi_console_getfd(struct lxc_container
*c
, int *ttynum
, int *masterfd
)
391 ttyfd
= lxc_console_getfd(c
, ttynum
, masterfd
);
395 static int lxcapi_console(struct lxc_container
*c
, int ttynum
, int stdinfd
,
396 int stdoutfd
, int stderrfd
, int escape
)
398 return lxc_console(c
, ttynum
, stdinfd
, stdoutfd
, stderrfd
, escape
);
401 static pid_t
lxcapi_init_pid(struct lxc_container
*c
)
406 return lxc_cmd_get_init_pid(c
->name
, c
->config_path
);
409 static bool load_config_locked(struct lxc_container
*c
, const char *fname
)
412 c
->lxc_conf
= lxc_conf_init();
415 if (lxc_config_read(fname
, c
->lxc_conf
, false) != 0)
417 if (!clone_update_unexp_network(c
->lxc_conf
))
422 static bool lxcapi_load_config(struct lxc_container
*c
, const char *alt_file
)
424 bool ret
= false, need_disklock
= false;
430 fname
= c
->configfile
;
436 * If we're reading something other than the container's config,
437 * we only need to lock the in-memory container. If loading the
438 * container's config file, take the disk lock.
440 if (strcmp(fname
, c
->configfile
) == 0)
441 need_disklock
= true;
444 lret
= container_disk_lock(c
);
446 lret
= container_mem_lock(c
);
450 ret
= load_config_locked(c
, fname
);
453 container_disk_unlock(c
);
455 container_mem_unlock(c
);
459 static bool lxcapi_want_daemonize(struct lxc_container
*c
, bool state
)
461 if (!c
|| !c
->lxc_conf
)
463 if (container_mem_lock(c
)) {
464 ERROR("Error getting mem lock");
467 c
->daemonize
= state
;
468 /* daemonize implies close_all_fds so set it */
470 c
->lxc_conf
->close_all_fds
= 1;
471 container_mem_unlock(c
);
475 static bool lxcapi_want_close_all_fds(struct lxc_container
*c
, bool state
)
477 if (!c
|| !c
->lxc_conf
)
479 if (container_mem_lock(c
)) {
480 ERROR("Error getting mem lock");
483 c
->lxc_conf
->close_all_fds
= state
;
484 container_mem_unlock(c
);
488 static bool lxcapi_wait(struct lxc_container
*c
, const char *state
, int timeout
)
495 ret
= lxc_wait(c
->name
, state
, timeout
, c
->config_path
);
500 static bool wait_on_daemonized_start(struct lxc_container
*c
, int pid
)
502 /* we'll probably want to make this timeout configurable? */
503 int timeout
= 5, ret
, status
;
506 * our child is going to fork again, then exit. reap the
509 ret
= waitpid(pid
, &status
, 0);
510 if (ret
== -1 || !WIFEXITED(status
) || WEXITSTATUS(status
) != 0)
511 DEBUG("failed waiting for first dual-fork child");
512 return lxcapi_wait(c
, "RUNNING", timeout
);
515 static bool am_single_threaded(void)
517 struct dirent dirent
, *direntp
;
521 dir
= opendir("/proc/self/task");
523 INFO("failed to open /proc/self/task");
527 while (!readdir_r(dir
, &dirent
, &direntp
)) {
531 if (!strcmp(direntp
->d_name
, "."))
534 if (!strcmp(direntp
->d_name
, ".."))
544 * I can't decide if it'd be more convenient for callers if we accept '...',
545 * or a null-terminated array (i.e. execl vs execv)
547 static bool lxcapi_start(struct lxc_container
*c
, int useinit
, char * const argv
[])
550 struct lxc_conf
*conf
;
551 bool daemonize
= false;
553 char *default_args
[] = {
558 /* container exists */
561 /* container has been setup */
565 if ((ret
= ongoing_create(c
)) < 0) {
566 ERROR("Error checking for incomplete creation");
570 ERROR("Error: %s creation was not completed", c
->name
);
573 } else if (ret
== 1) {
574 ERROR("Error: creation of %s is ongoing", c
->name
);
578 /* is this app meant to be run through lxcinit, as in lxc-execute? */
579 if (useinit
&& !argv
)
582 if (container_mem_lock(c
))
585 daemonize
= c
->daemonize
;
586 container_mem_unlock(c
);
589 ret
= lxc_execute(c
->name
, argv
, 1, conf
, c
->config_path
);
590 return ret
== 0 ? true : false;
597 * say, I'm not sure - what locks do we want here? Any?
598 * Is liblxc's locking enough here to protect the on disk
599 * container? We don't want to exclude things like lxc_info
600 * while container is running...
603 lxc_monitord_spawn(c
->config_path
);
610 /* Set to NULL because we don't want father unlink
611 * the PID file, child will do the free and unlink.
614 return wait_on_daemonized_start(c
, pid
);
617 /* second fork to be reparented by init */
620 SYSERROR("Error doing dual-fork");
625 /* like daemon(), chdir to / and redirect 0,1,2 to /dev/null */
627 SYSERROR("Error chdir()ing to /.");
630 lxc_check_inherited(conf
, -1);
634 open("/dev/zero", O_RDONLY
);
635 open("/dev/null", O_RDWR
);
636 open("/dev/null", O_RDWR
);
639 if (!am_single_threaded()) {
640 ERROR("Cannot start non-daemonized container when threaded");
645 /* We need to write PID file after daeminize, so we always
646 * write the right PID.
649 pid_fp
= fopen(c
->pidfile
, "w");
650 if (pid_fp
== NULL
) {
651 SYSERROR("Failed to create pidfile '%s' for '%s'",
652 c
->pidfile
, c
->name
);
656 if (fprintf(pid_fp
, "%d\n", getpid()) < 0) {
657 SYSERROR("Failed to write '%s'", c
->pidfile
);
669 ret
= lxc_start(c
->name
, argv
, conf
, c
->config_path
);
673 INFO("container requested reboot");
685 exit (ret
== 0 ? true : false);
687 return (ret
== 0 ? true : false);
691 * note there MUST be an ending NULL
693 static bool lxcapi_startl(struct lxc_container
*c
, int useinit
, ...)
696 char **inargs
= NULL
;
699 /* container exists */
703 va_start(ap
, useinit
);
704 inargs
= lxc_va_arg_list_to_argv(ap
, 0, 1);
708 ERROR("Memory allocation error.");
712 /* pass NULL if no arguments were supplied */
713 bret
= lxcapi_start(c
, useinit
, *inargs
? inargs
: NULL
);
718 for (arg
= inargs
; *arg
; arg
++)
726 static bool lxcapi_stop(struct lxc_container
*c
)
733 ret
= lxc_cmd_stop(c
->name
, c
->config_path
);
738 static int do_create_container_dir(const char *path
, struct lxc_conf
*conf
)
740 int ret
= -1, lasterr
;
741 char *p
= alloca(strlen(path
)+1);
742 mode_t mask
= umask(0002);
743 ret
= mkdir(path
, 0770);
751 SYSERROR("failed to create container path %s", path
);
756 if (!lxc_list_empty(&conf
->id_map
) && chown_mapped_root(p
, conf
) != 0) {
757 ERROR("Failed to chown container dir");
764 * create the standard expected container dir
766 static bool create_container_dir(struct lxc_container
*c
)
771 len
= strlen(c
->config_path
) + strlen(c
->name
) + 2;
775 ret
= snprintf(s
, len
, "%s/%s", c
->config_path
, c
->name
);
776 if (ret
< 0 || ret
>= len
) {
780 ret
= do_create_container_dir(s
, c
->lxc_conf
);
785 static const char *lxcapi_get_config_path(struct lxc_container
*c
);
786 static bool lxcapi_set_config_item(struct lxc_container
*c
, const char *key
, const char *v
);
789 * do_bdev_create: thin wrapper around bdev_create(). Like bdev_create(),
790 * it returns a mounted bdev on success, NULL on error.
792 static struct bdev
*do_bdev_create(struct lxc_container
*c
, const char *type
,
793 struct bdev_specs
*specs
)
800 /* rootfs.path or lxcpath/lxcname/rootfs */
801 if (c
->lxc_conf
->rootfs
.path
&& access(c
->lxc_conf
->rootfs
.path
, F_OK
) == 0) {
802 const char *rpath
= c
->lxc_conf
->rootfs
.path
;
803 len
= strlen(rpath
) + 1;
805 ret
= snprintf(dest
, len
, "%s", rpath
);
807 const char *lxcpath
= lxcapi_get_config_path(c
);
808 len
= strlen(c
->name
) + strlen(lxcpath
) + 9;
810 ret
= snprintf(dest
, len
, "%s/%s/rootfs", lxcpath
, c
->name
);
812 if (ret
< 0 || ret
>= len
)
815 bdev
= bdev_create(dest
, type
, c
->name
, specs
);
817 ERROR("Failed to create backing store type %s", type
);
821 lxcapi_set_config_item(c
, "lxc.rootfs", bdev
->src
);
823 /* if we are not root, chown the rootfs dir to root in the
826 if (geteuid() != 0 || (c
->lxc_conf
&& !lxc_list_empty(&c
->lxc_conf
->id_map
))) {
827 if (chown_mapped_root(bdev
->dest
, c
->lxc_conf
) < 0) {
828 ERROR("Error chowning %s to container root", bdev
->dest
);
829 suggest_default_idmap();
839 * Given the '-t' template option to lxc-create, figure out what to
840 * do. If the template is a full executable path, use that. If it
841 * is something like 'sshd', then return $templatepath/lxc-sshd.
842 * On success return the template, on error return NULL.
844 static char *get_template_path(const char *t
)
849 if (t
[0] == '/' && access(t
, X_OK
) == 0) {
854 len
= strlen(LXCTEMPLATEDIR
) + strlen(t
) + strlen("/lxc-") + 1;
858 ret
= snprintf(tpath
, len
, "%s/lxc-%s", LXCTEMPLATEDIR
, t
);
859 if (ret
< 0 || ret
>= len
) {
863 if (access(tpath
, X_OK
) < 0) {
864 SYSERROR("bad template: %s", t
);
872 static char *lxcbasename(char *path
)
874 char *p
= path
+ strlen(path
) - 1;
875 while (*p
!= '/' && p
> path
)
880 static bool create_run_template(struct lxc_container
*c
, char *tpath
, bool quiet
,
890 SYSERROR("failed to fork task for container creation template");
894 if (pid
== 0) { // child
895 char *patharg
, *namearg
, *rootfsarg
, *src
;
896 struct bdev
*bdev
= NULL
;
898 int ret
, len
, nargs
= 0;
900 struct lxc_conf
*conf
= c
->lxc_conf
;
906 open("/dev/zero", O_RDONLY
);
907 open("/dev/null", O_RDWR
);
908 open("/dev/null", O_RDWR
);
911 src
= c
->lxc_conf
->rootfs
.path
;
913 * for an overlay create, what the user wants is the template to fill
914 * in what will become the readonly lower layer. So don't mount for
917 if (strncmp(src
, "overlayfs:", 10) == 0)
918 src
= overlay_getlower(src
+10);
919 if (strncmp(src
, "aufs:", 5) == 0)
920 src
= overlay_getlower(src
+5);
922 bdev
= bdev_init(c
->lxc_conf
, src
, c
->lxc_conf
->rootfs
.mount
, NULL
);
924 ERROR("Error opening rootfs");
928 if (geteuid() == 0) {
929 if (unshare(CLONE_NEWNS
) < 0) {
930 ERROR("error unsharing mounts");
933 if (detect_shared_rootfs()) {
934 if (mount(NULL
, "/", NULL
, MS_SLAVE
|MS_REC
, NULL
)) {
935 SYSERROR("Failed to make / rslave to run template");
936 ERROR("Continuing...");
940 if (strcmp(bdev
->type
, "dir") && strcmp(bdev
->type
, "btrfs")) {
941 if (geteuid() != 0) {
942 ERROR("non-root users can only create btrfs and directory-backed containers");
945 if (bdev
->ops
->mount(bdev
) < 0) {
946 ERROR("Error mounting rootfs");
949 } else { // TODO come up with a better way here!
952 bdev
->dest
= strdup(bdev
->src
);
956 * create our new array, pre-pend the template name and
960 for (nargs
= 0; argv
[nargs
]; nargs
++) ;
961 nargs
+= 4; // template, path, rootfs and name args
963 newargv
= malloc(nargs
* sizeof(*newargv
));
966 newargv
[0] = lxcbasename(tpath
);
968 len
= strlen(c
->config_path
) + strlen(c
->name
) + strlen("--path=") + 2;
969 patharg
= malloc(len
);
972 ret
= snprintf(patharg
, len
, "--path=%s/%s", c
->config_path
, c
->name
);
973 if (ret
< 0 || ret
>= len
)
975 newargv
[1] = patharg
;
976 len
= strlen("--name=") + strlen(c
->name
) + 1;
977 namearg
= malloc(len
);
980 ret
= snprintf(namearg
, len
, "--name=%s", c
->name
);
981 if (ret
< 0 || ret
>= len
)
983 newargv
[2] = namearg
;
985 len
= strlen("--rootfs=") + 1 + strlen(bdev
->dest
);
986 rootfsarg
= malloc(len
);
989 ret
= snprintf(rootfsarg
, len
, "--rootfs=%s", bdev
->dest
);
990 if (ret
< 0 || ret
>= len
)
992 newargv
[3] = rootfsarg
;
994 /* add passed-in args */
996 for (i
= 4; i
< nargs
; i
++)
997 newargv
[i
] = argv
[i
-4];
999 /* add trailing NULL */
1001 newargv
= realloc(newargv
, nargs
* sizeof(*newargv
));
1004 newargv
[nargs
- 1] = NULL
;
1007 * If we're running the template in a mapped userns, then
1008 * we prepend the template command with:
1009 * lxc-usernsexec <-m map1> ... <-m mapn> --
1010 * and we append "--mapped-uid x", where x is the mapped uid
1013 if (!lxc_list_empty(&conf
->id_map
)) {
1017 char **n2
= malloc(n2args
* sizeof(*n2
));
1018 struct lxc_list
*it
;
1022 SYSERROR("out of memory");
1026 tpath
= "lxc-usernsexec";
1027 n2
[0] = "lxc-usernsexec";
1028 lxc_list_for_each(it
, &conf
->id_map
) {
1031 n2
= realloc(n2
, n2args
* sizeof(char *));
1034 n2
[n2args
-2] = "-m";
1035 n2
[n2args
-1] = malloc(200);
1038 ret
= snprintf(n2
[n2args
-1], 200, "%c:%lu:%lu:%lu",
1039 map
->idtype
== ID_TYPE_UID
? 'u' : 'g',
1040 map
->nsid
, map
->hostid
, map
->range
);
1041 if (ret
< 0 || ret
>= 200)
1044 int hostid_mapped
= mapped_hostid(geteuid(), conf
, ID_TYPE_UID
);
1045 int extraargs
= hostid_mapped
>= 0 ? 1 : 3;
1046 n2
= realloc(n2
, (nargs
+ n2args
+ extraargs
) * sizeof(char *));
1049 if (hostid_mapped
< 0) {
1050 hostid_mapped
= find_unmapped_nsuid(conf
, ID_TYPE_UID
);
1051 n2
[n2args
++] = "-m";
1052 if (hostid_mapped
< 0) {
1053 ERROR("Could not find free uid to map");
1056 n2
[n2args
++] = malloc(200);
1057 if (!n2
[n2args
-1]) {
1058 SYSERROR("out of memory");
1061 ret
= snprintf(n2
[n2args
-1], 200, "u:%d:%d:1",
1062 hostid_mapped
, geteuid());
1063 if (ret
< 0 || ret
>= 200) {
1064 ERROR("string too long");
1068 int hostgid_mapped
= mapped_hostid(getegid(), conf
, ID_TYPE_GID
);
1069 extraargs
= hostgid_mapped
>= 0 ? 1 : 3;
1070 n2
= realloc(n2
, (nargs
+ n2args
+ extraargs
) * sizeof(char *));
1073 if (hostgid_mapped
< 0) {
1074 hostgid_mapped
= find_unmapped_nsuid(conf
, ID_TYPE_GID
);
1075 n2
[n2args
++] = "-m";
1076 if (hostgid_mapped
< 0) {
1077 ERROR("Could not find free uid to map");
1080 n2
[n2args
++] = malloc(200);
1081 if (!n2
[n2args
-1]) {
1082 SYSERROR("out of memory");
1085 ret
= snprintf(n2
[n2args
-1], 200, "g:%d:%d:1",
1086 hostgid_mapped
, getegid());
1087 if (ret
< 0 || ret
>= 200) {
1088 ERROR("string too long");
1092 n2
[n2args
++] = "--";
1093 for (i
= 0; i
< nargs
; i
++)
1094 n2
[i
+ n2args
] = newargv
[i
];
1096 // Finally add "--mapped-uid $uid" to tell template what to chown
1099 n2
= realloc(n2
, n2args
* sizeof(char *));
1101 SYSERROR("out of memory");
1104 // note n2[n2args-1] is NULL
1105 n2
[n2args
-5] = "--mapped-uid";
1106 snprintf(txtuid
, 20, "%d", hostid_mapped
);
1107 n2
[n2args
-4] = txtuid
;
1108 n2
[n2args
-3] = "--mapped-gid";
1109 snprintf(txtgid
, 20, "%d", hostgid_mapped
);
1110 n2
[n2args
-2] = txtgid
;
1111 n2
[n2args
-1] = NULL
;
1116 execvp(tpath
, newargv
);
1117 SYSERROR("failed to execute template %s", tpath
);
1121 if (wait_for_pid(pid
) != 0) {
1122 ERROR("container creation template for %s failed", c
->name
);
1129 static bool prepend_lxc_header(char *path
, const char *t
, char *const argv
[])
1137 unsigned char md_value
[SHA_DIGEST_LENGTH
];
1141 f
= fopen(path
, "r");
1145 if (fseek(f
, 0, SEEK_END
) < 0)
1147 if ((flen
= ftell(f
)) < 0)
1149 if (fseek(f
, 0, SEEK_SET
) < 0)
1151 if ((contents
= malloc(flen
+ 1)) == NULL
)
1153 if (fread(contents
, 1, flen
, f
) != flen
)
1154 goto out_free_contents
;
1156 contents
[flen
] = '\0';
1160 goto out_free_contents
;
1163 tpath
= get_template_path(t
);
1165 ERROR("bad template: %s", t
);
1166 goto out_free_contents
;
1169 ret
= sha1sum_file(tpath
, md_value
);
1171 ERROR("Error getting sha1sum of %s", tpath
);
1173 goto out_free_contents
;
1178 f
= fopen(path
, "w");
1180 SYSERROR("reopening config for writing");
1184 fprintf(f
, "# Template used to create this container: %s\n", t
);
1186 fprintf(f
, "# Parameters passed to the template:");
1188 fprintf(f
, " %s", *argv
);
1194 fprintf(f
, "# Template script checksum (SHA-1): ");
1195 for (i
=0; i
<SHA_DIGEST_LENGTH
; i
++)
1196 fprintf(f
, "%02x", md_value
[i
]);
1199 fprintf(f
, "# For additional config options, please look at lxc.container.conf(5)\n");
1200 if (fwrite(contents
, 1, flen
, f
) != flen
) {
1201 SYSERROR("Writing original contents");
1217 SYSERROR("Error prepending header");
1223 static void lxcapi_clear_config(struct lxc_container
*c
)
1227 lxc_conf_free(c
->lxc_conf
);
1233 static bool lxcapi_destroy(struct lxc_container
*c
);
1234 static bool container_destroy(struct lxc_container
*c
);
1235 static bool get_snappath_dir(struct lxc_container
*c
, char *snappath
);
1238 * create a container with the given parameters.
1239 * @c: container to be created. It has the lxcpath, name, and a starting
1240 * configuration already set
1241 * @t: the template to execute to instantiate the root filesystem and
1242 * adjust the configuration.
1243 * @bdevtype: backing store type to use. If NULL, dir will be used.
1244 * @specs: additional parameters for the backing store, i.e. LVM vg to
1247 * @argv: the arguments to pass to the template, terminated by NULL. If no
1248 * arguments, you can just pass NULL.
1250 static bool lxcapi_create(struct lxc_container
*c
, const char *t
,
1251 const char *bdevtype
, struct bdev_specs
*specs
, int flags
,
1263 tpath
= get_template_path(t
);
1265 ERROR("bad template: %s", t
);
1271 * If a template is passed in, and the rootfs already is defined in
1272 * the container config and exists, then * caller is trying to create
1273 * an existing container. Return an error, but do NOT delete the
1276 if (lxcapi_is_defined(c
) && c
->lxc_conf
&& c
->lxc_conf
->rootfs
.path
&&
1277 access(c
->lxc_conf
->rootfs
.path
, F_OK
) == 0 && tpath
) {
1278 ERROR("Container %s:%s already exists", c
->config_path
, c
->name
);
1283 if (!c
->load_config(c
, lxc_global_config_value("lxc.default_config"))) {
1284 ERROR("Error loading default configuration file %s", lxc_global_config_value("lxc.default_config"));
1289 if (!create_container_dir(c
))
1293 * either template or rootfs.path should be set.
1294 * if both template and rootfs.path are set, template is setup as rootfs.path.
1295 * container is already created if we have a config and rootfs.path is accessible
1297 if (!c
->lxc_conf
->rootfs
.path
&& !tpath
)
1298 /* no template passed in and rootfs does not exist: error */
1300 if (c
->lxc_conf
->rootfs
.path
&& access(c
->lxc_conf
->rootfs
.path
, F_OK
) != 0)
1301 /* rootfs passed into configuration, but does not exist: error */
1303 if (lxcapi_is_defined(c
) && c
->lxc_conf
->rootfs
.path
&& !tpath
) {
1304 /* Rootfs already existed, user just wanted to save the
1305 * loaded configuration */
1310 /* Mark that this container is being created */
1311 if ((partial_fd
= create_partial(c
)) < 0)
1314 /* no need to get disk lock bc we have the partial locked */
1317 * Create the backing store
1318 * Note we can't do this in the same task as we use to execute the
1319 * template because of the way zfs works.
1320 * After you 'zfs create', zfs mounts the fs only in the initial
1325 SYSERROR("failed to fork task for container creation template");
1329 if (pid
== 0) { // child
1330 struct bdev
*bdev
= NULL
;
1332 if (!(bdev
= do_bdev_create(c
, bdevtype
, specs
))) {
1333 ERROR("Error creating backing store type %s for %s",
1334 bdevtype
? bdevtype
: "(none)", c
->name
);
1338 /* save config file again to store the new rootfs location */
1339 if (!c
->save_config(c
, NULL
)) {
1340 ERROR("failed to save starting configuration for %s", c
->name
);
1341 // parent task won't see bdev in config so we delete it
1342 bdev
->ops
->umount(bdev
);
1343 bdev
->ops
->destroy(bdev
);
1348 if (wait_for_pid(pid
) != 0)
1351 /* reload config to get the rootfs */
1352 lxc_conf_free(c
->lxc_conf
);
1354 if (!load_config_locked(c
, c
->configfile
))
1357 if (!create_run_template(c
, tpath
, !!(flags
& LXC_CREATE_QUIET
), argv
))
1360 // now clear out the lxc_conf we have, reload from the created
1362 lxcapi_clear_config(c
);
1365 if (!prepend_lxc_header(c
->configfile
, tpath
, argv
)) {
1366 ERROR("Error prepending header to configuration file");
1370 ret
= load_config_locked(c
, c
->configfile
);
1373 if (partial_fd
>= 0)
1374 remove_partial(c
, partial_fd
);
1377 container_destroy(c
);
1384 static bool lxcapi_reboot(struct lxc_container
*c
)
1390 if (!c
->is_running(c
))
1392 pid
= c
->init_pid(c
);
1395 if (kill(pid
, SIGINT
) < 0)
1401 static bool lxcapi_shutdown(struct lxc_container
*c
, int timeout
)
1405 int haltsignal
= SIGPWR
;
1410 if (!c
->is_running(c
))
1412 pid
= c
->init_pid(c
);
1415 if (c
->lxc_conf
&& c
->lxc_conf
->haltsignal
)
1416 haltsignal
= c
->lxc_conf
->haltsignal
;
1417 kill(pid
, haltsignal
);
1418 retv
= c
->wait(c
, "STOPPED", timeout
);
1422 static bool lxcapi_createl(struct lxc_container
*c
, const char *t
,
1423 const char *bdevtype
, struct bdev_specs
*specs
, int flags
, ...)
1433 * since we're going to wait for create to finish, I don't think we
1434 * need to get a copy of the arguments.
1436 va_start(ap
, flags
);
1437 args
= lxc_va_arg_list_to_argv(ap
, 0, 0);
1440 ERROR("Memory allocation error.");
1444 bret
= c
->create(c
, t
, bdevtype
, specs
, flags
, args
);
1451 static void do_clear_unexp_config_line(struct lxc_conf
*conf
, const char *key
)
1453 if (strcmp(key
, "lxc.cgroup") == 0)
1454 clear_unexp_config_line(conf
, key
, true);
1455 else if (strcmp(key
, "lxc.network") == 0)
1456 clear_unexp_config_line(conf
, key
, true);
1457 else if (strcmp(key
, "lxc.hook") == 0)
1458 clear_unexp_config_line(conf
, key
, true);
1460 clear_unexp_config_line(conf
, key
, false);
1461 if (!do_append_unexp_config_line(conf
, key
, ""))
1462 WARN("Error clearing configuration for %s", key
);
1465 static bool lxcapi_clear_config_item(struct lxc_container
*c
, const char *key
)
1469 if (!c
|| !c
->lxc_conf
)
1471 if (container_mem_lock(c
))
1473 ret
= lxc_clear_config_item(c
->lxc_conf
, key
);
1475 do_clear_unexp_config_line(c
->lxc_conf
, key
);
1476 container_mem_unlock(c
);
1480 static inline bool enter_to_ns(struct lxc_container
*c
) {
1481 int netns
, userns
, ret
= 0, init_pid
= 0;;
1482 char new_netns_path
[MAXPATHLEN
];
1483 char new_userns_path
[MAXPATHLEN
];
1485 if (!c
->is_running(c
))
1488 init_pid
= c
->init_pid(c
);
1490 /* Switch to new userns */
1491 if ((geteuid() != 0 || (c
->lxc_conf
&& !lxc_list_empty(&c
->lxc_conf
->id_map
))) && access("/proc/self/ns/user", F_OK
) == 0) {
1492 ret
= snprintf(new_userns_path
, MAXPATHLEN
, "/proc/%d/ns/user", init_pid
);
1493 if (ret
< 0 || ret
>= MAXPATHLEN
)
1496 userns
= open(new_userns_path
, O_RDONLY
);
1498 SYSERROR("failed to open %s", new_userns_path
);
1502 if (setns(userns
, CLONE_NEWUSER
)) {
1503 SYSERROR("failed to setns for CLONE_NEWUSER");
1510 /* Switch to new netns */
1511 ret
= snprintf(new_netns_path
, MAXPATHLEN
, "/proc/%d/ns/net", init_pid
);
1512 if (ret
< 0 || ret
>= MAXPATHLEN
)
1515 netns
= open(new_netns_path
, O_RDONLY
);
1517 SYSERROR("failed to open %s", new_netns_path
);
1521 if (setns(netns
, CLONE_NEWNET
)) {
1522 SYSERROR("failed to setns for CLONE_NEWNET");
1532 // used by qsort and bsearch functions for comparing names
1533 static inline int string_cmp(char **first
, char **second
)
1535 return strcmp(*first
, *second
);
1538 // used by qsort and bsearch functions for comparing container names
1539 static inline int container_cmp(struct lxc_container
**first
, struct lxc_container
**second
)
1541 return strcmp((*first
)->name
, (*second
)->name
);
1544 static bool add_to_array(char ***names
, char *cname
, int pos
)
1546 char **newnames
= realloc(*names
, (pos
+1) * sizeof(char *));
1548 ERROR("Out of memory");
1553 newnames
[pos
] = strdup(cname
);
1557 // sort the arrray as we will use binary search on it
1558 qsort(newnames
, pos
+ 1, sizeof(char *), (int (*)(const void *,const void *))string_cmp
);
1563 static bool add_to_clist(struct lxc_container
***list
, struct lxc_container
*c
, int pos
, bool sort
)
1565 struct lxc_container
**newlist
= realloc(*list
, (pos
+1) * sizeof(struct lxc_container
*));
1567 ERROR("Out of memory");
1574 // sort the arrray as we will use binary search on it
1576 qsort(newlist
, pos
+ 1, sizeof(struct lxc_container
*), (int (*)(const void *,const void *))container_cmp
);
1581 static char** get_from_array(char ***names
, char *cname
, int size
)
1583 return (char **)bsearch(&cname
, *names
, size
, sizeof(char *), (int (*)(const void *, const void *))string_cmp
);
1587 static bool array_contains(char ***names
, char *cname
, int size
) {
1588 if(get_from_array(names
, cname
, size
) != NULL
)
1593 static bool remove_from_array(char ***names
, char *cname
, int size
)
1595 char **result
= get_from_array(names
, cname
, size
);
1596 if (result
!= NULL
) {
1603 static char** lxcapi_get_interfaces(struct lxc_container
*c
)
1606 int i
, count
= 0, pipefd
[2];
1607 char **interfaces
= NULL
;
1608 char interface
[IFNAMSIZ
];
1610 if(pipe(pipefd
) < 0) {
1611 SYSERROR("pipe failed");
1617 SYSERROR("failed to fork task to get interfaces information");
1623 if (pid
== 0) { // child
1624 int ret
= 1, nbytes
;
1625 struct ifaddrs
*interfaceArray
= NULL
, *tempIfAddr
= NULL
;
1627 /* close the read-end of the pipe */
1630 if (!enter_to_ns(c
)) {
1631 SYSERROR("failed to enter namespace");
1635 /* Grab the list of interfaces */
1636 if (getifaddrs(&interfaceArray
)) {
1637 SYSERROR("failed to get interfaces list");
1641 /* Iterate through the interfaces */
1642 for (tempIfAddr
= interfaceArray
; tempIfAddr
!= NULL
; tempIfAddr
= tempIfAddr
->ifa_next
) {
1643 nbytes
= write(pipefd
[1], tempIfAddr
->ifa_name
, IFNAMSIZ
);
1645 ERROR("write failed");
1654 freeifaddrs(interfaceArray
);
1656 /* close the write-end of the pipe, thus sending EOF to the reader */
1661 /* close the write-end of the pipe */
1664 while (read(pipefd
[0], &interface
, IFNAMSIZ
) == IFNAMSIZ
) {
1665 if (array_contains(&interfaces
, interface
, count
))
1668 if(!add_to_array(&interfaces
, interface
, count
))
1669 ERROR("PARENT: add_to_array failed");
1673 if (wait_for_pid(pid
) != 0) {
1674 for(i
=0;i
<count
;i
++)
1675 free(interfaces
[i
]);
1680 /* close the read-end of the pipe */
1683 /* Append NULL to the array */
1685 interfaces
= (char **)lxc_append_null_to_array((void **)interfaces
, count
);
1690 static char** lxcapi_get_ips(struct lxc_container
*c
, const char* interface
, const char* family
, int scope
)
1693 int i
, count
= 0, pipefd
[2];
1694 char **addresses
= NULL
;
1695 char address
[INET6_ADDRSTRLEN
];
1697 if(pipe(pipefd
) < 0) {
1698 SYSERROR("pipe failed");
1704 SYSERROR("failed to fork task to get container ips");
1710 if (pid
== 0) { // child
1711 int ret
= 1, nbytes
;
1712 struct ifaddrs
*interfaceArray
= NULL
, *tempIfAddr
= NULL
;
1713 char addressOutputBuffer
[INET6_ADDRSTRLEN
];
1714 void *tempAddrPtr
= NULL
;
1715 char *address
= NULL
;
1717 /* close the read-end of the pipe */
1720 if (!enter_to_ns(c
)) {
1721 SYSERROR("failed to enter namespace");
1725 /* Grab the list of interfaces */
1726 if (getifaddrs(&interfaceArray
)) {
1727 SYSERROR("failed to get interfaces list");
1731 /* Iterate through the interfaces */
1732 for (tempIfAddr
= interfaceArray
; tempIfAddr
!= NULL
; tempIfAddr
= tempIfAddr
->ifa_next
) {
1733 if (tempIfAddr
->ifa_addr
== NULL
)
1736 if(tempIfAddr
->ifa_addr
->sa_family
== AF_INET
) {
1737 if (family
&& strcmp(family
, "inet"))
1739 tempAddrPtr
= &((struct sockaddr_in
*)tempIfAddr
->ifa_addr
)->sin_addr
;
1742 if (family
&& strcmp(family
, "inet6"))
1745 if (((struct sockaddr_in6
*)tempIfAddr
->ifa_addr
)->sin6_scope_id
!= scope
)
1748 tempAddrPtr
= &((struct sockaddr_in6
*)tempIfAddr
->ifa_addr
)->sin6_addr
;
1751 if (interface
&& strcmp(interface
, tempIfAddr
->ifa_name
))
1753 else if (!interface
&& strcmp("lo", tempIfAddr
->ifa_name
) == 0)
1756 address
= (char *)inet_ntop(tempIfAddr
->ifa_addr
->sa_family
,
1758 addressOutputBuffer
,
1759 sizeof(addressOutputBuffer
));
1763 nbytes
= write(pipefd
[1], address
, INET6_ADDRSTRLEN
);
1765 ERROR("write failed");
1774 freeifaddrs(interfaceArray
);
1776 /* close the write-end of the pipe, thus sending EOF to the reader */
1781 /* close the write-end of the pipe */
1784 while (read(pipefd
[0], &address
, INET6_ADDRSTRLEN
) == INET6_ADDRSTRLEN
) {
1785 if(!add_to_array(&addresses
, address
, count
))
1786 ERROR("PARENT: add_to_array failed");
1790 if (wait_for_pid(pid
) != 0) {
1791 for(i
=0;i
<count
;i
++)
1797 /* close the read-end of the pipe */
1800 /* Append NULL to the array */
1802 addresses
= (char **)lxc_append_null_to_array((void **)addresses
, count
);
1807 static int lxcapi_get_config_item(struct lxc_container
*c
, const char *key
, char *retv
, int inlen
)
1811 if (!c
|| !c
->lxc_conf
)
1813 if (container_mem_lock(c
))
1815 ret
= lxc_get_config_item(c
->lxc_conf
, key
, retv
, inlen
);
1816 container_mem_unlock(c
);
1820 static char* lxcapi_get_running_config_item(struct lxc_container
*c
, const char *key
)
1824 if (!c
|| !c
->lxc_conf
)
1826 if (container_mem_lock(c
))
1828 ret
= lxc_cmd_get_config_item(c
->name
, key
, c
->get_config_path(c
));
1829 container_mem_unlock(c
);
1833 static int lxcapi_get_keys(struct lxc_container
*c
, const char *key
, char *retv
, int inlen
)
1836 return lxc_listconfigs(retv
, inlen
);
1838 * Support 'lxc.network.<idx>', i.e. 'lxc.network.0'
1839 * This is an intelligent result to show which keys are valid given
1840 * the type of nic it is
1842 if (!c
|| !c
->lxc_conf
)
1844 if (container_mem_lock(c
))
1847 if (strncmp(key
, "lxc.network.", 12) == 0)
1848 ret
= lxc_list_nicconfigs(c
->lxc_conf
, key
, retv
, inlen
);
1849 container_mem_unlock(c
);
1853 static bool lxcapi_save_config(struct lxc_container
*c
, const char *alt_file
)
1856 bool ret
= false, need_disklock
= false;
1860 alt_file
= c
->configfile
;
1862 return false; // should we write to stdout if no file is specified?
1864 // If we haven't yet loaded a config, load the stock config
1866 if (!c
->load_config(c
, lxc_global_config_value("lxc.default_config"))) {
1867 ERROR("Error loading default configuration file %s while saving %s", lxc_global_config_value("lxc.default_config"), c
->name
);
1872 if (!create_container_dir(c
))
1876 * If we're writing to the container's config file, take the
1877 * disk lock. Otherwise just take the memlock to protect the
1878 * struct lxc_container while we're traversing it.
1880 if (strcmp(c
->configfile
, alt_file
) == 0)
1881 need_disklock
= true;
1884 lret
= container_disk_lock(c
);
1886 lret
= container_mem_lock(c
);
1891 fout
= fopen(alt_file
, "w");
1894 write_config(fout
, c
->lxc_conf
);
1900 container_disk_unlock(c
);
1902 container_mem_unlock(c
);
1906 static bool mod_rdep(struct lxc_container
*c
, bool inc
)
1908 char path
[MAXPATHLEN
];
1913 if (container_disk_lock(c
))
1915 ret
= snprintf(path
, MAXPATHLEN
, "%s/%s/lxc_snapshots", c
->config_path
,
1917 if (ret
< 0 || ret
> MAXPATHLEN
)
1919 f
= fopen(path
, "r");
1921 ret
= fscanf(f
, "%d", &v
);
1924 ERROR("Corrupted file %s", path
);
1929 f
= fopen(path
, "w");
1932 if (fprintf(f
, "%d\n", v
) < 0) {
1933 ERROR("Error writing new snapshots value");
1939 SYSERROR("Error writing to or closing snapshots file");
1946 container_disk_unlock(c
);
1950 static void strip_newline(char *p
)
1952 size_t len
= strlen(p
);
1955 if (p
[len
-1] == '\n')
1959 static void mod_all_rdeps(struct lxc_container
*c
, bool inc
)
1961 struct lxc_container
*p
;
1962 char *lxcpath
= NULL
, *lxcname
= NULL
, path
[MAXPATHLEN
];
1963 size_t pathlen
= 0, namelen
= 0;
1967 ret
= snprintf(path
, MAXPATHLEN
, "%s/%s/lxc_rdepends",
1968 c
->config_path
, c
->name
);
1969 if (ret
< 0 || ret
>= MAXPATHLEN
) {
1970 ERROR("Path name too long");
1973 f
= fopen(path
, "r");
1976 while (getline(&lxcpath
, &pathlen
, f
) != -1) {
1977 if (getline(&lxcname
, &namelen
, f
) == -1) {
1978 ERROR("badly formatted file %s", path
);
1981 strip_newline(lxcpath
);
1982 strip_newline(lxcname
);
1983 if ((p
= lxc_container_new(lxcname
, lxcpath
)) == NULL
) {
1984 ERROR("Unable to find dependent container %s:%s",
1988 if (!mod_rdep(p
, inc
))
1989 ERROR("Failed to increase numsnapshots for %s:%s",
1991 lxc_container_put(p
);
1994 if (lxcpath
) free(lxcpath
);
1995 if (lxcname
) free(lxcname
);
1999 static bool has_fs_snapshots(struct lxc_container
*c
)
2001 char path
[MAXPATHLEN
];
2006 ret
= snprintf(path
, MAXPATHLEN
, "%s/%s/lxc_snapshots", c
->config_path
,
2008 if (ret
< 0 || ret
> MAXPATHLEN
)
2010 f
= fopen(path
, "r");
2013 ret
= fscanf(f
, "%d", &v
);
2023 static bool has_snapshots(struct lxc_container
*c
)
2025 char path
[MAXPATHLEN
];
2026 struct dirent dirent
, *direntp
;
2030 if (!get_snappath_dir(c
, path
))
2032 dir
= opendir(path
);
2035 while (!readdir_r(dir
, &dirent
, &direntp
)) {
2039 if (!strcmp(direntp
->d_name
, "."))
2042 if (!strcmp(direntp
->d_name
, ".."))
2051 static int lxc_rmdir_onedev_wrapper(void *data
)
2053 char *arg
= (char *) data
;
2054 return lxc_rmdir_onedev(arg
, "snaps");
2057 static int do_bdev_destroy(struct lxc_conf
*conf
)
2062 r
= bdev_init(conf
, conf
->rootfs
.path
, conf
->rootfs
.mount
, NULL
);
2066 if (r
->ops
->destroy(r
) < 0)
2072 static int bdev_destroy_wrapper(void *data
)
2074 struct lxc_conf
*conf
= data
;
2076 if (setgid(0) < 0) {
2077 ERROR("Failed to setgid to 0");
2080 if (setgroups(0, NULL
) < 0)
2081 WARN("Failed to clear groups");
2082 if (setuid(0) < 0) {
2083 ERROR("Failed to setuid to 0");
2086 return do_bdev_destroy(conf
);
2089 static bool container_destroy(struct lxc_container
*c
)
2094 if (!c
|| !lxcapi_is_defined(c
))
2097 if (container_disk_lock(c
))
2100 if (!is_stopped(c
)) {
2101 // we should queue some sort of error - in c->error_string?
2102 ERROR("container %s is not stopped", c
->name
);
2106 if (c
->lxc_conf
&& c
->lxc_conf
->rootfs
.path
&& c
->lxc_conf
->rootfs
.mount
) {
2108 ret
= userns_exec_1(c
->lxc_conf
, bdev_destroy_wrapper
, c
->lxc_conf
);
2110 ret
= do_bdev_destroy(c
->lxc_conf
);
2112 ERROR("Error destroying rootfs for %s", c
->name
);
2117 mod_all_rdeps(c
, false);
2119 const char *p1
= lxcapi_get_config_path(c
);
2120 char *path
= alloca(strlen(p1
) + strlen(c
->name
) + 2);
2121 sprintf(path
, "%s/%s", p1
, c
->name
);
2123 ret
= userns_exec_1(c
->lxc_conf
, lxc_rmdir_onedev_wrapper
, path
);
2125 ret
= lxc_rmdir_onedev(path
, "snaps");
2127 ERROR("Error destroying container directory for %s", c
->name
);
2133 container_disk_unlock(c
);
2137 static bool lxcapi_destroy(struct lxc_container
*c
)
2139 if (!c
|| !lxcapi_is_defined(c
))
2141 if (has_snapshots(c
)) {
2142 ERROR("Container %s has snapshots; not removing", c
->name
);
2146 if (has_fs_snapshots(c
)) {
2147 ERROR("container %s has snapshots on its rootfs", c
->name
);
2151 return container_destroy(c
);
2154 static bool lxcapi_snapshot_destroy_all(struct lxc_container
*c
);
2156 static bool lxcapi_destroy_with_snapshots(struct lxc_container
*c
)
2158 if (!c
|| !lxcapi_is_defined(c
))
2160 if (!lxcapi_snapshot_destroy_all(c
)) {
2161 ERROR("Error deleting all snapshots");
2164 return lxcapi_destroy(c
);
2167 static bool set_config_item_locked(struct lxc_container
*c
, const char *key
, const char *v
)
2169 struct lxc_config_t
*config
;
2172 c
->lxc_conf
= lxc_conf_init();
2175 config
= lxc_getconfig(key
);
2178 if (config
->cb(key
, v
, c
->lxc_conf
) != 0)
2180 return do_append_unexp_config_line(c
->lxc_conf
, key
, v
);
2183 static bool lxcapi_set_config_item(struct lxc_container
*c
, const char *key
, const char *v
)
2190 if (container_mem_lock(c
))
2193 b
= set_config_item_locked(c
, key
, v
);
2195 container_mem_unlock(c
);
2199 static char *lxcapi_config_file_name(struct lxc_container
*c
)
2201 if (!c
|| !c
->configfile
)
2203 return strdup(c
->configfile
);
2206 static const char *lxcapi_get_config_path(struct lxc_container
*c
)
2208 if (!c
|| !c
->config_path
)
2210 return (const char *)(c
->config_path
);
2215 * Just recalculate the c->configfile based on the
2216 * c->config_path, which must be set.
2217 * The lxc_container must be locked or not yet public.
2219 static bool set_config_filename(struct lxc_container
*c
)
2224 if (!c
->config_path
)
2227 /* $lxc_path + "/" + c->name + "/" + "config" + '\0' */
2228 len
= strlen(c
->config_path
) + strlen(c
->name
) + strlen("config") + 3;
2229 newpath
= malloc(len
);
2233 ret
= snprintf(newpath
, len
, "%s/%s/config", c
->config_path
, c
->name
);
2234 if (ret
< 0 || ret
>= len
) {
2235 fprintf(stderr
, "Error printing out config file name\n");
2241 free(c
->configfile
);
2242 c
->configfile
= newpath
;
2247 static bool lxcapi_set_config_path(struct lxc_container
*c
, const char *path
)
2251 char *oldpath
= NULL
;
2256 if (container_mem_lock(c
))
2261 ERROR("Out of memory setting new lxc path");
2267 oldpath
= c
->config_path
;
2270 /* Since we've changed the config path, we have to change the
2271 * config file name too */
2272 if (!set_config_filename(c
)) {
2273 ERROR("Out of memory setting new config filename");
2275 free(c
->config_path
);
2276 c
->config_path
= oldpath
;
2282 container_mem_unlock(c
);
2287 static bool lxcapi_set_cgroup_item(struct lxc_container
*c
, const char *subsys
, const char *value
)
2297 if (container_disk_lock(c
))
2300 ret
= lxc_cgroup_set(subsys
, value
, c
->name
, c
->config_path
);
2302 container_disk_unlock(c
);
2306 static int lxcapi_get_cgroup_item(struct lxc_container
*c
, const char *subsys
, char *retv
, int inlen
)
2316 if (container_disk_lock(c
))
2319 ret
= lxc_cgroup_get(subsys
, retv
, inlen
, c
->name
, c
->config_path
);
2321 container_disk_unlock(c
);
2325 const char *lxc_get_global_config_item(const char *key
)
2327 return lxc_global_config_value(key
);
2330 const char *lxc_get_version(void)
2335 static int copy_file(const char *old
, const char *new)
2342 if (file_exists(new)) {
2343 ERROR("copy destination %s exists", new);
2346 ret
= stat(old
, &sbuf
);
2348 INFO("Error stat'ing %s", old
);
2352 in
= open(old
, O_RDONLY
);
2354 SYSERROR("Error opening original file %s", old
);
2357 out
= open(new, O_CREAT
| O_EXCL
| O_WRONLY
, 0644);
2359 SYSERROR("Error opening new file %s", new);
2365 len
= read(in
, buf
, 8096);
2367 SYSERROR("Error reading old file %s", old
);
2372 ret
= write(out
, buf
, len
);
2373 if (ret
< len
) { // should we retry?
2374 SYSERROR("Error: write to new file %s was interrupted", new);
2381 // we set mode, but not owner/group
2382 ret
= chmod(new, sbuf
.st_mode
);
2384 SYSERROR("Error setting mode on %s", new);
2396 static int copyhooks(struct lxc_container
*oldc
, struct lxc_container
*c
)
2399 struct lxc_list
*it
;
2402 len
= strlen(oldc
->config_path
) + strlen(oldc
->name
) + 3;
2403 cpath
= alloca(len
);
2404 ret
= snprintf(cpath
, len
, "%s/%s/", oldc
->config_path
, oldc
->name
);
2405 if (ret
< 0 || ret
>= len
)
2408 for (i
=0; i
<NUM_LXC_HOOKS
; i
++) {
2409 lxc_list_for_each(it
, &c
->lxc_conf
->hooks
[i
]) {
2410 char *hookname
= it
->elem
;
2411 char *fname
= strrchr(hookname
, '/');
2412 char tmppath
[MAXPATHLEN
];
2413 if (!fname
) // relative path - we don't support, but maybe we should
2415 if (strncmp(hookname
, cpath
, len
- 1) != 0) {
2416 // this hook is public - ignore
2419 // copy the script, and change the entry in confile
2420 ret
= snprintf(tmppath
, MAXPATHLEN
, "%s/%s/%s",
2421 c
->config_path
, c
->name
, fname
+1);
2422 if (ret
< 0 || ret
>= MAXPATHLEN
)
2424 ret
= copy_file(it
->elem
, tmppath
);
2428 it
->elem
= strdup(tmppath
);
2430 ERROR("out of memory copying hook path");
2436 if (!clone_update_unexp_hooks(c
->lxc_conf
)) {
2437 ERROR("Error saving new hooks in clone");
2440 c
->save_config(c
, NULL
);
2444 static void new_hwaddr(char *hwaddr
)
2447 f
= fopen("/dev/urandom", "r");
2450 int ret
= fread(&seed
, sizeof(seed
), 1, f
);
2457 snprintf(hwaddr
, 18, "00:16:3e:%02x:%02x:%02x",
2458 rand() % 255, rand() % 255, rand() % 255);
2461 static void network_new_hwaddrs(struct lxc_container
*c
)
2463 struct lxc_list
*it
;
2465 lxc_list_for_each(it
, &c
->lxc_conf
->network
) {
2466 struct lxc_netdev
*n
= it
->elem
;
2468 new_hwaddr(n
->hwaddr
);
2472 static int copy_fstab(struct lxc_container
*oldc
, struct lxc_container
*c
)
2474 char newpath
[MAXPATHLEN
];
2475 char *oldpath
= oldc
->lxc_conf
->fstab
;
2481 clear_unexp_config_line(c
->lxc_conf
, "lxc.mount", false);
2483 char *p
= strrchr(oldpath
, '/');
2486 ret
= snprintf(newpath
, MAXPATHLEN
, "%s/%s%s",
2487 c
->config_path
, c
->name
, p
);
2488 if (ret
< 0 || ret
>= MAXPATHLEN
) {
2489 ERROR("error printing new path for %s", oldpath
);
2492 if (file_exists(newpath
)) {
2493 ERROR("error: fstab file %s exists", newpath
);
2497 if (copy_file(oldpath
, newpath
) < 0) {
2498 ERROR("error: copying %s to %s", oldpath
, newpath
);
2501 free(c
->lxc_conf
->fstab
);
2502 c
->lxc_conf
->fstab
= strdup(newpath
);
2503 if (!c
->lxc_conf
->fstab
) {
2504 ERROR("error: allocating pathname");
2507 if (!do_append_unexp_config_line(c
->lxc_conf
, "lxc.mount", newpath
)) {
2508 ERROR("error saving new lxctab");
2515 static void copy_rdepends(struct lxc_container
*c
, struct lxc_container
*c0
)
2517 char path0
[MAXPATHLEN
], path1
[MAXPATHLEN
];
2520 ret
= snprintf(path0
, MAXPATHLEN
, "%s/%s/lxc_rdepends", c0
->config_path
,
2522 if (ret
< 0 || ret
>= MAXPATHLEN
) {
2523 WARN("Error copying reverse dependencies");
2526 ret
= snprintf(path1
, MAXPATHLEN
, "%s/%s/lxc_rdepends", c
->config_path
,
2528 if (ret
< 0 || ret
>= MAXPATHLEN
) {
2529 WARN("Error copying reverse dependencies");
2532 if (copy_file(path0
, path1
) < 0) {
2533 INFO("Error copying reverse dependencies");
2538 static bool add_rdepends(struct lxc_container
*c
, struct lxc_container
*c0
)
2541 char path
[MAXPATHLEN
];
2545 ret
= snprintf(path
, MAXPATHLEN
, "%s/%s/lxc_rdepends", c
->config_path
,
2547 if (ret
< 0 || ret
>= MAXPATHLEN
)
2549 f
= fopen(path
, "a");
2553 // if anything goes wrong, just return an error
2554 if (fprintf(f
, "%s\n%s\n", c0
->config_path
, c0
->name
) < 0)
2561 static int copy_storage(struct lxc_container
*c0
, struct lxc_container
*c
,
2562 const char *newtype
, int flags
, const char *bdevdata
, uint64_t newsize
)
2567 bdev
= bdev_copy(c0
, c
->name
, c
->config_path
, newtype
, flags
,
2568 bdevdata
, newsize
, &need_rdep
);
2570 ERROR("Error copying storage");
2573 free(c
->lxc_conf
->rootfs
.path
);
2574 c
->lxc_conf
->rootfs
.path
= strdup(bdev
->src
);
2576 if (!c
->lxc_conf
->rootfs
.path
) {
2577 ERROR("Out of memory while setting storage path");
2580 // We will simply append a new lxc.rootfs entry to the unexpanded config
2581 clear_unexp_config_line(c
->lxc_conf
, "lxc.rootfs", false);
2582 if (!do_append_unexp_config_line(c
->lxc_conf
, "lxc.rootfs", c
->lxc_conf
->rootfs
.path
)) {
2583 ERROR("Error saving new rootfs to cloend config");
2586 if (flags
& LXC_CLONE_SNAPSHOT
)
2587 copy_rdepends(c
, c0
);
2589 if (!add_rdepends(c
, c0
))
2590 WARN("Error adding reverse dependency from %s to %s",
2594 mod_all_rdeps(c
, true);
2599 struct clone_update_data
{
2600 struct lxc_container
*c0
;
2601 struct lxc_container
*c1
;
2606 static int clone_update_rootfs(struct clone_update_data
*data
)
2608 struct lxc_container
*c0
= data
->c0
;
2609 struct lxc_container
*c
= data
->c1
;
2610 int flags
= data
->flags
;
2611 char **hookargs
= data
->hookargs
;
2613 char path
[MAXPATHLEN
];
2616 struct lxc_conf
*conf
= c
->lxc_conf
;
2618 /* update hostname in rootfs */
2619 /* we're going to mount, so run in a clean namespace to simplify cleanup */
2621 if (setgid(0) < 0) {
2622 ERROR("Failed to setgid to 0");
2625 if (setuid(0) < 0) {
2626 ERROR("Failed to setuid to 0");
2629 if (setgroups(0, NULL
) < 0)
2630 WARN("Failed to clear groups");
2632 if (unshare(CLONE_NEWNS
) < 0)
2634 bdev
= bdev_init(c
->lxc_conf
, c
->lxc_conf
->rootfs
.path
, c
->lxc_conf
->rootfs
.mount
, NULL
);
2637 if (strcmp(bdev
->type
, "dir") != 0) {
2638 if (unshare(CLONE_NEWNS
) < 0) {
2639 ERROR("error unsharing mounts");
2643 if (detect_shared_rootfs()) {
2644 if (mount(NULL
, "/", NULL
, MS_SLAVE
|MS_REC
, NULL
)) {
2645 SYSERROR("Failed to make / rslave");
2646 ERROR("Continuing...");
2649 if (bdev
->ops
->mount(bdev
) < 0) {
2653 } else { // TODO come up with a better way
2656 bdev
->dest
= strdup(bdev
->src
);
2659 if (!lxc_list_empty(&conf
->hooks
[LXCHOOK_CLONE
])) {
2660 /* Start of environment variable setup for hooks */
2661 if (setenv("LXC_SRC_NAME", c0
->name
, 1)) {
2662 SYSERROR("failed to set environment variable for source container name");
2664 if (setenv("LXC_NAME", c
->name
, 1)) {
2665 SYSERROR("failed to set environment variable for container name");
2667 if (setenv("LXC_CONFIG_FILE", conf
->rcfile
, 1)) {
2668 SYSERROR("failed to set environment variable for config path");
2670 if (setenv("LXC_ROOTFS_MOUNT", bdev
->dest
, 1)) {
2671 SYSERROR("failed to set environment variable for rootfs mount");
2673 if (setenv("LXC_ROOTFS_PATH", conf
->rootfs
.path
, 1)) {
2674 SYSERROR("failed to set environment variable for rootfs mount");
2677 if (run_lxc_hooks(c
->name
, "clone", conf
, c
->get_config_path(c
), hookargs
)) {
2678 ERROR("Error executing clone hook for %s", c
->name
);
2684 if (!(flags
& LXC_CLONE_KEEPNAME
)) {
2685 ret
= snprintf(path
, MAXPATHLEN
, "%s/etc/hostname", bdev
->dest
);
2688 if (ret
< 0 || ret
>= MAXPATHLEN
)
2690 if (!file_exists(path
))
2692 if (!(fout
= fopen(path
, "w"))) {
2693 SYSERROR("unable to open %s: ignoring", path
);
2696 if (fprintf(fout
, "%s", c
->name
) < 0) {
2700 if (fclose(fout
) < 0)
2709 static int clone_update_rootfs_wrapper(void *data
)
2711 struct clone_update_data
*arg
= (struct clone_update_data
*) data
;
2712 return clone_update_rootfs(arg
);
2716 * We want to support:
2717 sudo lxc-clone -o o1 -n n1 -s -L|-fssize fssize -v|--vgname vgname \
2718 -p|--lvprefix lvprefix -t|--fstype fstype -B backingstore
2720 -s [ implies overlayfs]
2724 only rootfs gets converted (copied/snapshotted) on clone.
2727 static int create_file_dirname(char *path
, struct lxc_conf
*conf
)
2729 char *p
= strrchr(path
, '/');
2735 ret
= do_create_container_dir(path
, conf
);
2740 static struct lxc_container
*lxcapi_clone(struct lxc_container
*c
, const char *newname
,
2741 const char *lxcpath
, int flags
,
2742 const char *bdevtype
, const char *bdevdata
, uint64_t newsize
,
2745 struct lxc_container
*c2
= NULL
;
2746 char newpath
[MAXPATHLEN
];
2747 int ret
, storage_copied
= 0;
2748 char *origroot
= NULL
;
2749 struct clone_update_data data
;
2753 if (!c
|| !c
->is_defined(c
))
2756 if (container_mem_lock(c
))
2759 if (!is_stopped(c
)) {
2760 ERROR("error: Original container (%s) is running", c
->name
);
2764 // Make sure the container doesn't yet exist.
2768 lxcpath
= c
->get_config_path(c
);
2769 ret
= snprintf(newpath
, MAXPATHLEN
, "%s/%s/config", lxcpath
, newname
);
2770 if (ret
< 0 || ret
>= MAXPATHLEN
) {
2771 SYSERROR("clone: failed making config pathname");
2774 if (file_exists(newpath
)) {
2775 ERROR("error: clone: %s exists", newpath
);
2779 ret
= create_file_dirname(newpath
, c
->lxc_conf
);
2780 if (ret
< 0 && errno
!= EEXIST
) {
2781 ERROR("Error creating container dir for %s", newpath
);
2785 // copy the configuration, tweak it as needed,
2786 if (c
->lxc_conf
->rootfs
.path
) {
2787 origroot
= c
->lxc_conf
->rootfs
.path
;
2788 c
->lxc_conf
->rootfs
.path
= NULL
;
2790 fout
= fopen(newpath
, "w");
2792 SYSERROR("open %s", newpath
);
2795 write_config(fout
, c
->lxc_conf
);
2797 c
->lxc_conf
->rootfs
.path
= origroot
;
2799 sprintf(newpath
, "%s/%s/rootfs", lxcpath
, newname
);
2800 if (mkdir(newpath
, 0755) < 0) {
2801 SYSERROR("error creating %s", newpath
);
2806 if (chown_mapped_root(newpath
, c
->lxc_conf
) < 0) {
2807 ERROR("Error chowning %s to container root", newpath
);
2812 c2
= lxc_container_new(newname
, lxcpath
);
2814 ERROR("clone: failed to create new container (%s %s)", newname
,
2819 // copy/snapshot rootfs's
2820 ret
= copy_storage(c
, c2
, bdevtype
, flags
, bdevdata
, newsize
);
2824 clear_unexp_config_line(c2
->lxc_conf
, "lxc.utsname", false);
2827 if (!set_config_item_locked(c2
, "lxc.utsname", newname
)) {
2828 ERROR("Error setting new hostname");
2833 ret
= copyhooks(c
, c2
);
2835 ERROR("error copying hooks");
2839 if (copy_fstab(c
, c2
) < 0) {
2840 ERROR("error copying fstab");
2845 if (!(flags
& LXC_CLONE_KEEPMACADDR
)) {
2846 network_new_hwaddrs(c2
);
2847 if (!clone_update_unexp_network(c2
->lxc_conf
)) {
2848 ERROR("Error updating network for clone");
2853 // We've now successfully created c2's storage, so clear it out if we
2857 if (!c2
->save_config(c2
, NULL
))
2860 if ((pid
= fork()) < 0) {
2865 ret
= wait_for_pid(pid
);
2868 container_mem_unlock(c
);
2874 data
.hookargs
= hookargs
;
2876 ret
= userns_exec_1(c
->lxc_conf
, clone_update_rootfs_wrapper
,
2879 ret
= clone_update_rootfs(&data
);
2883 container_mem_unlock(c
);
2887 container_mem_unlock(c
);
2889 if (!storage_copied
)
2890 c2
->lxc_conf
->rootfs
.path
= NULL
;
2892 lxc_container_put(c2
);
2898 static bool lxcapi_rename(struct lxc_container
*c
, const char *newname
)
2901 struct lxc_container
*newc
;
2903 if (!c
|| !c
->name
|| !c
->config_path
|| !c
->lxc_conf
)
2906 if (has_fs_snapshots(c
) || has_snapshots(c
)) {
2907 ERROR("Renaming a container with snapshots is not supported");
2910 bdev
= bdev_init(c
->lxc_conf
, c
->lxc_conf
->rootfs
.path
, c
->lxc_conf
->rootfs
.mount
, NULL
);
2912 ERROR("Failed to find original backing store type");
2916 newc
= lxcapi_clone(c
, newname
, c
->config_path
, LXC_CLONE_KEEPMACADDR
, NULL
, bdev
->type
, 0, NULL
);
2919 lxc_container_put(newc
);
2923 if (newc
&& lxcapi_is_defined(newc
))
2924 lxc_container_put(newc
);
2926 if (!container_destroy(c
)) {
2927 ERROR("Could not destroy existing container %s", c
->name
);
2933 static int lxcapi_attach(struct lxc_container
*c
, lxc_attach_exec_t exec_function
, void *exec_payload
, lxc_attach_options_t
*options
, pid_t
*attached_process
)
2938 return lxc_attach(c
->name
, c
->config_path
, exec_function
, exec_payload
, options
, attached_process
);
2941 static int lxcapi_attach_run_wait(struct lxc_container
*c
, lxc_attach_options_t
*options
, const char *program
, const char * const argv
[])
2943 lxc_attach_command_t command
;
2950 command
.program
= (char*)program
;
2951 command
.argv
= (char**)argv
;
2952 r
= lxc_attach(c
->name
, c
->config_path
, lxc_attach_run_command
, &command
, options
, &pid
);
2957 return lxc_wait_for_pid_status(pid
);
2960 static int get_next_index(const char *lxcpath
, char *cname
)
2966 fname
= alloca(strlen(lxcpath
) + 20);
2968 sprintf(fname
, "%s/snap%d", lxcpath
, i
);
2969 ret
= stat(fname
, &sb
);
2976 static bool get_snappath_dir(struct lxc_container
*c
, char *snappath
)
2980 * If the old style snapshot path exists, use it
2981 * /var/lib/lxc -> /var/lib/lxcsnaps
2983 ret
= snprintf(snappath
, MAXPATHLEN
, "%ssnaps", c
->config_path
);
2984 if (ret
< 0 || ret
>= MAXPATHLEN
)
2986 if (dir_exists(snappath
)) {
2987 ret
= snprintf(snappath
, MAXPATHLEN
, "%ssnaps/%s", c
->config_path
, c
->name
);
2988 if (ret
< 0 || ret
>= MAXPATHLEN
)
2994 * Use the new style path
2995 * /var/lib/lxc -> /var/lib/lxc + c->name + /snaps + \0
2997 ret
= snprintf(snappath
, MAXPATHLEN
, "%s/%s/snaps", c
->config_path
, c
->name
);
2998 if (ret
< 0 || ret
>= MAXPATHLEN
)
3003 static int lxcapi_snapshot(struct lxc_container
*c
, const char *commentfile
)
3006 struct lxc_container
*c2
;
3007 char snappath
[MAXPATHLEN
], newname
[20];
3009 if (!c
|| !lxcapi_is_defined(c
))
3012 if (!bdev_can_backup(c
->lxc_conf
)) {
3013 ERROR("%s's backing store cannot be backed up.", c
->name
);
3014 ERROR("Your container must use another backing store type.");
3018 if (!get_snappath_dir(c
, snappath
))
3021 i
= get_next_index(snappath
, c
->name
);
3023 if (mkdir_p(snappath
, 0755) < 0) {
3024 ERROR("Failed to create snapshot directory %s", snappath
);
3028 ret
= snprintf(newname
, 20, "snap%d", i
);
3029 if (ret
< 0 || ret
>= 20)
3033 * We pass LXC_CLONE_SNAPSHOT to make sure that a rdepends file entry is
3034 * created in the original container
3036 flags
= LXC_CLONE_SNAPSHOT
| LXC_CLONE_KEEPMACADDR
| LXC_CLONE_KEEPNAME
|
3037 LXC_CLONE_KEEPBDEVTYPE
| LXC_CLONE_MAYBE_SNAPSHOT
;
3038 if (bdev_is_dir(c
->lxc_conf
, c
->lxc_conf
->rootfs
.path
)) {
3039 ERROR("Snapshot of directory-backed container requested.");
3040 ERROR("Making a copy-clone. If you do want snapshots, then");
3041 ERROR("please create an aufs or overlayfs clone first, snapshot that");
3042 ERROR("and keep the original container pristine.");
3043 flags
&= ~LXC_CLONE_SNAPSHOT
| LXC_CLONE_MAYBE_SNAPSHOT
;
3045 c2
= c
->clone(c
, newname
, snappath
, flags
, NULL
, NULL
, 0, NULL
);
3047 ERROR("clone of %s:%s failed", c
->config_path
, c
->name
);
3051 lxc_container_put(c2
);
3053 // Now write down the creation time
3060 tm_info
= localtime(&timer
);
3062 strftime(buffer
, 25, "%Y:%m:%d %H:%M:%S", tm_info
);
3064 char *dfnam
= alloca(strlen(snappath
) + strlen(newname
) + 5);
3065 sprintf(dfnam
, "%s/%s/ts", snappath
, newname
);
3066 f
= fopen(dfnam
, "w");
3068 ERROR("Failed to open %s", dfnam
);
3071 if (fprintf(f
, "%s", buffer
) < 0) {
3072 SYSERROR("Writing timestamp");
3078 SYSERROR("Writing timestamp");
3083 // $p / $name / comment \0
3084 int len
= strlen(snappath
) + strlen(newname
) + 10;
3085 char *path
= alloca(len
);
3086 sprintf(path
, "%s/%s/comment", snappath
, newname
);
3087 return copy_file(commentfile
, path
) < 0 ? -1 : i
;
3093 static void lxcsnap_free(struct lxc_snapshot
*s
)
3097 if (s
->comment_pathname
)
3098 free(s
->comment_pathname
);
3105 static char *get_snapcomment_path(char* snappath
, char *name
)
3107 // $snappath/$name/comment
3108 int ret
, len
= strlen(snappath
) + strlen(name
) + 10;
3109 char *s
= malloc(len
);
3112 ret
= snprintf(s
, len
, "%s/%s/comment", snappath
, name
);
3113 if (ret
< 0 || ret
>= len
) {
3121 static char *get_timestamp(char* snappath
, char *name
)
3123 char path
[MAXPATHLEN
], *s
= NULL
;
3127 ret
= snprintf(path
, MAXPATHLEN
, "%s/%s/ts", snappath
, name
);
3128 if (ret
< 0 || ret
>= MAXPATHLEN
)
3130 fin
= fopen(path
, "r");
3133 (void) fseek(fin
, 0, SEEK_END
);
3135 (void) fseek(fin
, 0, SEEK_SET
);
3140 if (fread(s
, 1, len
, fin
) != len
) {
3141 SYSERROR("reading timestamp");
3151 static int lxcapi_snapshot_list(struct lxc_container
*c
, struct lxc_snapshot
**ret_snaps
)
3153 char snappath
[MAXPATHLEN
], path2
[MAXPATHLEN
];
3155 struct dirent dirent
, *direntp
;
3156 struct lxc_snapshot
*snaps
=NULL
, *nsnaps
;
3159 if (!c
|| !lxcapi_is_defined(c
))
3162 if (!get_snappath_dir(c
, snappath
)) {
3163 ERROR("path name too long");
3166 dir
= opendir(snappath
);
3168 INFO("failed to open %s - assuming no snapshots", snappath
);
3172 while (!readdir_r(dir
, &dirent
, &direntp
)) {
3176 if (!strcmp(direntp
->d_name
, "."))
3179 if (!strcmp(direntp
->d_name
, ".."))
3182 ret
= snprintf(path2
, MAXPATHLEN
, "%s/%s/config", snappath
, direntp
->d_name
);
3183 if (ret
< 0 || ret
>= MAXPATHLEN
) {
3184 ERROR("pathname too long");
3187 if (!file_exists(path2
))
3189 nsnaps
= realloc(snaps
, (count
+ 1)*sizeof(*snaps
));
3191 SYSERROR("Out of memory");
3195 snaps
[count
].free
= lxcsnap_free
;
3196 snaps
[count
].name
= strdup(direntp
->d_name
);
3197 if (!snaps
[count
].name
)
3199 snaps
[count
].lxcpath
= strdup(snappath
);
3200 if (!snaps
[count
].lxcpath
) {
3201 free(snaps
[count
].name
);
3204 snaps
[count
].comment_pathname
= get_snapcomment_path(snappath
, direntp
->d_name
);
3205 snaps
[count
].timestamp
= get_timestamp(snappath
, direntp
->d_name
);
3210 WARN("failed to close directory");
3218 for (i
=0; i
<count
; i
++)
3219 lxcsnap_free(&snaps
[i
]);
3223 WARN("failed to close directory");
3227 static bool lxcapi_snapshot_restore(struct lxc_container
*c
, const char *snapname
, const char *newname
)
3229 char clonelxcpath
[MAXPATHLEN
];
3231 struct lxc_container
*snap
, *rest
;
3235 if (!c
|| !c
->name
|| !c
->config_path
)
3238 if (has_fs_snapshots(c
)) {
3239 ERROR("container rootfs has dependent snapshots");
3243 bdev
= bdev_init(c
->lxc_conf
, c
->lxc_conf
->rootfs
.path
, c
->lxc_conf
->rootfs
.mount
, NULL
);
3245 ERROR("Failed to find original backing store type");
3252 if (!get_snappath_dir(c
, clonelxcpath
)) {
3256 // how should we lock this?
3258 snap
= lxc_container_new(snapname
, clonelxcpath
);
3259 if (!snap
|| !lxcapi_is_defined(snap
)) {
3260 ERROR("Could not open snapshot %s", snapname
);
3261 if (snap
) lxc_container_put(snap
);
3266 if (strcmp(c
->name
, newname
) == 0) {
3267 if (!container_destroy(c
)) {
3268 ERROR("Could not destroy existing container %s", newname
);
3269 lxc_container_put(snap
);
3275 if (strcmp(bdev
->type
, "dir") != 0 && strcmp(bdev
->type
, "loop") != 0)
3276 flags
= LXC_CLONE_SNAPSHOT
| LXC_CLONE_MAYBE_SNAPSHOT
;
3277 rest
= lxcapi_clone(snap
, newname
, c
->config_path
, flags
,
3278 bdev
->type
, NULL
, 0, NULL
);
3280 if (rest
&& lxcapi_is_defined(rest
))
3283 lxc_container_put(rest
);
3284 lxc_container_put(snap
);
3288 static bool do_snapshot_destroy(const char *snapname
, const char *clonelxcpath
)
3290 struct lxc_container
*snap
= NULL
;
3293 snap
= lxc_container_new(snapname
, clonelxcpath
);
3295 ERROR("Could not find snapshot %s", snapname
);
3299 if (!lxcapi_destroy(snap
)) {
3300 ERROR("Could not destroy snapshot %s", snapname
);
3307 lxc_container_put(snap
);
3311 static bool remove_all_snapshots(const char *path
)
3314 struct dirent dirent
, *direntp
;
3317 dir
= opendir(path
);
3319 SYSERROR("opendir on snapshot path %s", path
);
3322 while (!readdir_r(dir
, &dirent
, &direntp
)) {
3325 if (!strcmp(direntp
->d_name
, "."))
3327 if (!strcmp(direntp
->d_name
, ".."))
3329 if (!do_snapshot_destroy(direntp
->d_name
, path
)) {
3338 SYSERROR("Error removing directory %s", path
);
3343 static bool lxcapi_snapshot_destroy(struct lxc_container
*c
, const char *snapname
)
3345 char clonelxcpath
[MAXPATHLEN
];
3347 if (!c
|| !c
->name
|| !c
->config_path
|| !snapname
)
3350 if (!get_snappath_dir(c
, clonelxcpath
))
3353 return do_snapshot_destroy(snapname
, clonelxcpath
);
3356 static bool lxcapi_snapshot_destroy_all(struct lxc_container
*c
)
3358 char clonelxcpath
[MAXPATHLEN
];
3360 if (!c
|| !c
->name
|| !c
->config_path
)
3363 if (!get_snappath_dir(c
, clonelxcpath
))
3366 return remove_all_snapshots(clonelxcpath
);
3369 static bool lxcapi_may_control(struct lxc_container
*c
)
3371 return lxc_try_cmd(c
->name
, c
->config_path
) == 0;
3374 static bool do_add_remove_node(pid_t init_pid
, const char *path
, bool add
,
3377 char chrootpath
[MAXPATHLEN
];
3378 char *directory_path
= NULL
;
3382 if ((pid
= fork()) < 0) {
3383 SYSERROR("failed to fork a child helper");
3387 if (wait_for_pid(pid
) != 0) {
3388 ERROR("Failed to create note in guest");
3394 /* prepare the path */
3395 ret
= snprintf(chrootpath
, MAXPATHLEN
, "/proc/%d/root", init_pid
);
3396 if (ret
< 0 || ret
>= MAXPATHLEN
)
3399 if (chroot(chrootpath
) < 0)
3403 /* remove path if it exists */
3404 if(faccessat(AT_FDCWD
, path
, F_OK
, AT_SYMLINK_NOFOLLOW
) == 0) {
3405 if (unlink(path
) < 0) {
3406 ERROR("unlink failed");
3413 /* create any missing directories */
3414 directory_path
= dirname(strdup(path
));
3415 if (mkdir_p(directory_path
, 0755) < 0 && errno
!= EEXIST
) {
3416 ERROR("failed to create directory");
3420 /* create the device node */
3421 if (mknod(path
, st
->st_mode
, st
->st_rdev
) < 0) {
3422 ERROR("mknod failed");
3429 static bool add_remove_device_node(struct lxc_container
*c
, const char *src_path
, const char *dest_path
, bool add
)
3433 char value
[MAX_BUFFER
];
3436 /* make sure container is running */
3437 if (!c
->is_running(c
)) {
3438 ERROR("container is not running");
3442 /* use src_path if dest_path is NULL otherwise use dest_path */
3443 p
= dest_path
? dest_path
: src_path
;
3445 /* make sure we can access p */
3446 if(access(p
, F_OK
) < 0 || stat(p
, &st
) < 0)
3449 /* continue if path is character device or block device */
3450 if (S_ISCHR(st
.st_mode
))
3451 ret
= snprintf(value
, MAX_BUFFER
, "c %d:%d rwm", major(st
.st_rdev
), minor(st
.st_rdev
));
3452 else if (S_ISBLK(st
.st_mode
))
3453 ret
= snprintf(value
, MAX_BUFFER
, "b %d:%d rwm", major(st
.st_rdev
), minor(st
.st_rdev
));
3457 /* check snprintf return code */
3458 if (ret
< 0 || ret
>= MAX_BUFFER
)
3461 if (!do_add_remove_node(c
->init_pid(c
), p
, add
, &st
))
3464 /* add or remove device to/from cgroup access list */
3466 if (!c
->set_cgroup_item(c
, "devices.allow", value
)) {
3467 ERROR("set_cgroup_item failed while adding the device node");
3471 if (!c
->set_cgroup_item(c
, "devices.deny", value
)) {
3472 ERROR("set_cgroup_item failed while removing the device node");
3480 static bool lxcapi_add_device_node(struct lxc_container
*c
, const char *src_path
, const char *dest_path
)
3483 ERROR(NOT_SUPPORTED_ERROR
, __FUNCTION__
);
3486 return add_remove_device_node(c
, src_path
, dest_path
, true);
3489 static bool lxcapi_remove_device_node(struct lxc_container
*c
, const char *src_path
, const char *dest_path
)
3492 ERROR(NOT_SUPPORTED_ERROR
, __FUNCTION__
);
3495 return add_remove_device_node(c
, src_path
, dest_path
, false);
3498 static int lxcapi_attach_run_waitl(struct lxc_container
*c
, lxc_attach_options_t
*options
, const char *program
, const char *arg
, ...)
3508 argv
= lxc_va_arg_list_to_argv_const(ap
, 1);
3512 ERROR("Memory allocation error.");
3517 ret
= lxcapi_attach_run_wait(c
, options
, program
, (const char * const *)argv
);
3522 struct lxc_container
*lxc_container_new(const char *name
, const char *configpath
)
3524 struct lxc_container
*c
;
3529 c
= malloc(sizeof(*c
));
3531 fprintf(stderr
, "failed to malloc lxc_container\n");
3534 memset(c
, 0, sizeof(*c
));
3537 c
->config_path
= strdup(configpath
);
3539 c
->config_path
= strdup(lxc_global_config_value("lxc.lxcpath"));
3541 if (!c
->config_path
) {
3542 fprintf(stderr
, "Out of memory\n");
3546 remove_trailing_slashes(c
->config_path
);
3547 c
->name
= malloc(strlen(name
)+1);
3549 fprintf(stderr
, "Error allocating lxc_container name\n");
3552 strcpy(c
->name
, name
);
3555 if (!(c
->slock
= lxc_newlock(c
->config_path
, name
))) {
3556 fprintf(stderr
, "failed to create lock\n");
3560 if (!(c
->privlock
= lxc_newlock(NULL
, NULL
))) {
3561 fprintf(stderr
, "failed to alloc privlock\n");
3565 if (!set_config_filename(c
)) {
3566 fprintf(stderr
, "Error allocating config file pathname\n");
3570 if (file_exists(c
->configfile
) && !lxcapi_load_config(c
, NULL
))
3573 if (ongoing_create(c
) == 2) {
3574 ERROR("Error: %s creation was not completed", c
->name
);
3575 container_destroy(c
);
3576 lxcapi_clear_config(c
);
3578 c
->daemonize
= true;
3581 // assign the member functions
3582 c
->is_defined
= lxcapi_is_defined
;
3583 c
->state
= lxcapi_state
;
3584 c
->is_running
= lxcapi_is_running
;
3585 c
->freeze
= lxcapi_freeze
;
3586 c
->unfreeze
= lxcapi_unfreeze
;
3587 c
->console
= lxcapi_console
;
3588 c
->console_getfd
= lxcapi_console_getfd
;
3589 c
->init_pid
= lxcapi_init_pid
;
3590 c
->load_config
= lxcapi_load_config
;
3591 c
->want_daemonize
= lxcapi_want_daemonize
;
3592 c
->want_close_all_fds
= lxcapi_want_close_all_fds
;
3593 c
->start
= lxcapi_start
;
3594 c
->startl
= lxcapi_startl
;
3595 c
->stop
= lxcapi_stop
;
3596 c
->config_file_name
= lxcapi_config_file_name
;
3597 c
->wait
= lxcapi_wait
;
3598 c
->set_config_item
= lxcapi_set_config_item
;
3599 c
->destroy
= lxcapi_destroy
;
3600 c
->destroy_with_snapshots
= lxcapi_destroy_with_snapshots
;
3601 c
->rename
= lxcapi_rename
;
3602 c
->save_config
= lxcapi_save_config
;
3603 c
->get_keys
= lxcapi_get_keys
;
3604 c
->create
= lxcapi_create
;
3605 c
->createl
= lxcapi_createl
;
3606 c
->shutdown
= lxcapi_shutdown
;
3607 c
->reboot
= lxcapi_reboot
;
3608 c
->clear_config
= lxcapi_clear_config
;
3609 c
->clear_config_item
= lxcapi_clear_config_item
;
3610 c
->get_config_item
= lxcapi_get_config_item
;
3611 c
->get_running_config_item
= lxcapi_get_running_config_item
;
3612 c
->get_cgroup_item
= lxcapi_get_cgroup_item
;
3613 c
->set_cgroup_item
= lxcapi_set_cgroup_item
;
3614 c
->get_config_path
= lxcapi_get_config_path
;
3615 c
->set_config_path
= lxcapi_set_config_path
;
3616 c
->clone
= lxcapi_clone
;
3617 c
->get_interfaces
= lxcapi_get_interfaces
;
3618 c
->get_ips
= lxcapi_get_ips
;
3619 c
->attach
= lxcapi_attach
;
3620 c
->attach_run_wait
= lxcapi_attach_run_wait
;
3621 c
->attach_run_waitl
= lxcapi_attach_run_waitl
;
3622 c
->snapshot
= lxcapi_snapshot
;
3623 c
->snapshot_list
= lxcapi_snapshot_list
;
3624 c
->snapshot_restore
= lxcapi_snapshot_restore
;
3625 c
->snapshot_destroy
= lxcapi_snapshot_destroy
;
3626 c
->snapshot_destroy_all
= lxcapi_snapshot_destroy_all
;
3627 c
->may_control
= lxcapi_may_control
;
3628 c
->add_device_node
= lxcapi_add_device_node
;
3629 c
->remove_device_node
= lxcapi_remove_device_node
;
3631 /* we'll allow the caller to update these later */
3632 if (lxc_log_init(NULL
, "none", NULL
, "lxc_container", 0, c
->config_path
)) {
3633 fprintf(stderr
, "failed to open log\n");
3640 lxc_container_free(c
);
3644 int lxc_get_wait_states(const char **states
)
3649 for (i
=0; i
<MAX_STATE
; i
++)
3650 states
[i
] = lxc_state2str(i
);
3655 * These next two could probably be done smarter with reusing a common function
3656 * with different iterators and tests...
3658 int list_defined_containers(const char *lxcpath
, char ***names
, struct lxc_container
***cret
)
3661 int i
, cfound
= 0, nfound
= 0;
3662 struct dirent dirent
, *direntp
;
3663 struct lxc_container
*c
;
3666 lxcpath
= lxc_global_config_value("lxc.lxcpath");
3668 dir
= opendir(lxcpath
);
3670 SYSERROR("opendir on lxcpath");
3679 while (!readdir_r(dir
, &dirent
, &direntp
)) {
3682 if (!strcmp(direntp
->d_name
, "."))
3684 if (!strcmp(direntp
->d_name
, ".."))
3687 if (!config_file_exists(lxcpath
, direntp
->d_name
))
3691 if (!add_to_array(names
, direntp
->d_name
, cfound
))
3701 c
= lxc_container_new(direntp
->d_name
, lxcpath
);
3703 INFO("Container %s:%s has a config but could not be loaded",
3704 lxcpath
, direntp
->d_name
);
3706 if(!remove_from_array(names
, direntp
->d_name
, cfound
--))
3710 if (!lxcapi_is_defined(c
)) {
3711 INFO("Container %s:%s has a config but is not defined",
3712 lxcpath
, direntp
->d_name
);
3714 if(!remove_from_array(names
, direntp
->d_name
, cfound
--))
3716 lxc_container_put(c
);
3720 if (!add_to_clist(cret
, c
, nfound
, true)) {
3721 lxc_container_put(c
);
3731 if (names
&& *names
) {
3732 for (i
=0; i
<cfound
; i
++)
3736 if (cret
&& *cret
) {
3737 for (i
=0; i
<nfound
; i
++)
3738 lxc_container_put((*cret
)[i
]);
3745 int list_active_containers(const char *lxcpath
, char ***nret
,
3746 struct lxc_container
***cret
)
3748 int i
, ret
= -1, cret_cnt
= 0, ct_name_cnt
= 0;
3751 char **ct_name
= NULL
;
3753 struct lxc_container
*c
;
3756 lxcpath
= lxc_global_config_value("lxc.lxcpath");
3757 lxcpath_len
= strlen(lxcpath
);
3764 FILE *f
= fopen("/proc/net/unix", "r");
3768 while (getline(&line
, &len
, f
) != -1) {
3769 char *p
= strrchr(line
, ' '), *p2
;
3776 if (strncmp(p
, lxcpath
, lxcpath_len
) != 0)
3782 // Now p is the start of lxc_name
3784 if (!p2
|| strncmp(p2
, "/command", 8) != 0)
3788 if (array_contains(&ct_name
, p
, ct_name_cnt
))
3791 if (!add_to_array(&ct_name
, p
, ct_name_cnt
))
3792 goto free_cret_list
;
3799 c
= lxc_container_new(p
, lxcpath
);
3801 INFO("Container %s:%s is running but could not be loaded",
3803 remove_from_array(&ct_name
, p
, ct_name_cnt
--);
3808 * If this is an anonymous container, then is_defined *can*
3809 * return false. So we don't do that check. Count on the
3810 * fact that the command socket exists.
3813 if (!add_to_clist(cret
, c
, cret_cnt
, true)) {
3814 lxc_container_put(c
);
3815 goto free_cret_list
;
3820 assert(!nret
|| !cret
|| cret_cnt
== ct_name_cnt
);
3829 if (cret
&& *cret
) {
3830 for (i
= 0; i
< cret_cnt
; i
++)
3831 lxc_container_put((*cret
)[i
]);
3837 for (i
= 0; i
< ct_name_cnt
; i
++)
3850 int list_all_containers(const char *lxcpath
, char ***nret
,
3851 struct lxc_container
***cret
)
3853 int i
, ret
, active_cnt
, ct_cnt
, ct_list_cnt
;
3856 struct lxc_container
**ct_list
= NULL
;
3858 ct_cnt
= list_defined_containers(lxcpath
, &ct_name
, NULL
);
3862 active_cnt
= list_active_containers(lxcpath
, &active_name
, NULL
);
3863 if (active_cnt
< 0) {
3868 for (i
= 0; i
< active_cnt
; i
++) {
3869 if (!array_contains(&ct_name
, active_name
[i
], ct_cnt
)) {
3870 if (!add_to_array(&ct_name
, active_name
[i
], ct_cnt
)) {
3872 goto free_active_name
;
3876 free(active_name
[i
]);
3877 active_name
[i
] = NULL
;
3883 for (i
= 0, ct_list_cnt
= 0; i
< ct_cnt
&& cret
; i
++) {
3884 struct lxc_container
*c
;
3886 c
= lxc_container_new(ct_name
[i
], lxcpath
);
3888 WARN("Container %s:%s could not be loaded", lxcpath
, ct_name
[i
]);
3889 remove_from_array(&ct_name
, ct_name
[i
], ct_cnt
--);
3893 if (!add_to_clist(&ct_list
, c
, ct_list_cnt
, false)) {
3894 lxc_container_put(c
);
3913 for (i
= 0; i
< ct_list_cnt
; i
++) {
3914 lxc_container_put(ct_list
[i
]);
3920 for (i
= 0; i
< active_cnt
; i
++) {
3922 free(active_name
[i
]);
3928 for (i
= 0; i
< ct_cnt
; i
++) {