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>
39 #include <sys/syscall.h>
41 #include <lxc/lxccontainer.h>
42 #include <lxc/version.h>
43 #include <lxc/network.h>
59 #include "namespace.h"
67 #include <../include/ifaddrs.h>
71 #include <../include/lxcmntent.h>
76 #define MAX_BUFFER 4096
78 #define NOT_SUPPORTED_ERROR "the requested function %s is not currently supported with unprivileged containers"
80 /* Define faccessat() if missing from the C library */
81 #ifndef HAVE_FACCESSAT
82 static int faccessat(int __fd
, const char *__file
, int __type
, int __flag
)
85 return syscall(__NR_faccessat
, __fd
, __file
, __type
, __flag
);
93 lxc_log_define(lxc_container
, lxc
);
95 static bool config_file_exists(const char *lxcpath
, const char *cname
)
97 /* $lxcpath + '/' + $cname + '/config' + \0 */
98 int ret
, len
= strlen(lxcpath
) + strlen(cname
) + 9;
99 char *fname
= alloca(len
);
101 ret
= snprintf(fname
, len
, "%s/%s/config", lxcpath
, cname
);
102 if (ret
< 0 || ret
>= len
)
105 return file_exists(fname
);
109 * A few functions to help detect when a container creation failed.
110 * If a container creation was killed partway through, then trying
111 * to actually start that container could harm the host. We detect
112 * this by creating a 'partial' file under the container directory,
113 * and keeping an advisory lock. When container creation completes,
114 * we remove that file. When we load or try to start a container, if
115 * we find that file, without a flock, we remove the container.
117 static int ongoing_create(struct lxc_container
*c
)
119 int len
= strlen(c
->config_path
) + strlen(c
->name
) + 10;
120 char *path
= alloca(len
);
124 ret
= snprintf(path
, len
, "%s/%s/partial", c
->config_path
, c
->name
);
125 if (ret
< 0 || ret
>= len
) {
126 ERROR("Error writing partial pathname");
130 if (!file_exists(path
))
132 fd
= open(path
, O_RDWR
);
134 // give benefit of the doubt
135 SYSERROR("Error opening partial file");
139 lk
.l_whence
= SEEK_SET
;
143 if (fcntl(fd
, F_GETLK
, &lk
) == 0 && lk
.l_pid
!= -1) {
144 // create is still ongoing
148 // create completed but partial is still there.
153 static int create_partial(struct lxc_container
*c
)
155 // $lxcpath + '/' + $name + '/partial' + \0
156 int len
= strlen(c
->config_path
) + strlen(c
->name
) + 10;
157 char *path
= alloca(len
);
161 ret
= snprintf(path
, len
, "%s/%s/partial", c
->config_path
, c
->name
);
162 if (ret
< 0 || ret
>= len
) {
163 ERROR("Error writing partial pathname");
166 if ((fd
=open(path
, O_RDWR
| O_CREAT
| O_EXCL
, 0755)) < 0) {
167 SYSERROR("Erorr creating partial file");
171 lk
.l_whence
= SEEK_SET
;
174 if (fcntl(fd
, F_SETLKW
, &lk
) < 0) {
175 SYSERROR("Error locking partial file %s", path
);
183 static void remove_partial(struct lxc_container
*c
, int fd
)
185 // $lxcpath + '/' + $name + '/partial' + \0
186 int len
= strlen(c
->config_path
) + strlen(c
->name
) + 10;
187 char *path
= alloca(len
);
191 ret
= snprintf(path
, len
, "%s/%s/partial", c
->config_path
, c
->name
);
192 if (ret
< 0 || ret
>= len
) {
193 ERROR("Error writing partial pathname");
196 if (unlink(path
) < 0)
197 SYSERROR("Error unlink partial file %s", path
);
201 * 1. container_mem_lock(c) protects the struct lxc_container from multiple threads.
202 * 2. container_disk_lock(c) protects the on-disk container data - in particular the
203 * container configuration file.
204 * The container_disk_lock also takes the container_mem_lock.
205 * 3. thread_mutex protects process data (ex: fd table) from multiple threads.
206 * NOTHING mutexes two independent programs with their own struct
207 * lxc_container for the same c->name, between API calls. For instance,
208 * c->config_read(); c->start(); Between those calls, data on disk
209 * could change (which shouldn't bother the caller unless for instance
210 * the rootfs get moved). c->config_read(); update; c->config_write();
211 * Two such updaters could race. The callers should therefore check their
212 * results. Trying to prevent that would necessarily expose us to deadlocks
213 * due to hung callers. So I prefer to keep the locks only within our own
214 * functions, not across functions.
216 * If you're going to clone while holding a lxccontainer, increment
217 * c->numthreads (under privlock) before forking. When deleting,
218 * decrement numthreads under privlock, then if it hits 0 you can delete.
219 * Do not ever use a lxccontainer whose numthreads you did not bump.
222 static void lxc_container_free(struct lxc_container
*c
)
228 c
->configfile
= NULL
;
229 free(c
->error_string
);
230 c
->error_string
= NULL
;
232 lxc_putlock(c
->slock
);
236 lxc_putlock(c
->privlock
);
242 lxc_conf_free(c
->lxc_conf
);
245 free(c
->config_path
);
246 c
->config_path
= NULL
;
252 * Consider the following case:
253 freer | racing get()er
254 ==================================================================
255 lxc_container_put() | lxc_container_get()
256 \ lxclock(c->privlock) | c->numthreads < 1? (no)
257 \ c->numthreads = 0 | \ lxclock(c->privlock) -> waits
259 \ lxc_container_free() | \ lxclock() returns
260 | \ c->numthreads < 1 -> return 0
262 \ \ sem_destroy(privlock) |
264 * When the get()er checks numthreads the first time, one of the following
266 * 1. freer has set numthreads = 0. get() returns 0
267 * 2. freer is between lxclock and setting numthreads to 0. get()er will
268 * sem_wait on privlock, get lxclock after freer() drops it, then see
269 * numthreads is 0 and exit without touching lxclock again..
270 * 3. freer has not yet locked privlock. If get()er runs first, then put()er
271 * will see --numthreads = 1 and not call lxc_container_free().
274 int lxc_container_get(struct lxc_container
*c
)
279 // if someone else has already started freeing the container, don't
280 // try to take the lock, which may be invalid
281 if (c
->numthreads
< 1)
284 if (container_mem_lock(c
))
286 if (c
->numthreads
< 1) {
287 // bail without trying to unlock, bc the privlock is now probably
292 container_mem_unlock(c
);
296 int lxc_container_put(struct lxc_container
*c
)
300 if (container_mem_lock(c
))
302 if (--c
->numthreads
< 1) {
303 container_mem_unlock(c
);
304 lxc_container_free(c
);
307 container_mem_unlock(c
);
311 static bool lxcapi_is_defined(struct lxc_container
*c
)
320 if (container_mem_lock(c
))
324 statret
= stat(c
->configfile
, &statbuf
);
330 container_mem_unlock(c
);
334 static const char *lxcapi_state(struct lxc_container
*c
)
340 s
= lxc_getstate(c
->name
, c
->config_path
);
341 return lxc_state2str(s
);
344 static bool is_stopped(struct lxc_container
*c
)
347 s
= lxc_getstate(c
->name
, c
->config_path
);
348 return (s
== STOPPED
);
351 static bool lxcapi_is_running(struct lxc_container
*c
)
358 if (!s
|| strcmp(s
, "STOPPED") == 0)
363 static bool lxcapi_freeze(struct lxc_container
*c
)
369 ret
= lxc_freeze(c
->name
, c
->config_path
);
375 static bool lxcapi_unfreeze(struct lxc_container
*c
)
381 ret
= lxc_unfreeze(c
->name
, c
->config_path
);
387 static int lxcapi_console_getfd(struct lxc_container
*c
, int *ttynum
, int *masterfd
)
393 ttyfd
= lxc_console_getfd(c
, ttynum
, masterfd
);
397 static int lxcapi_console(struct lxc_container
*c
, int ttynum
, int stdinfd
,
398 int stdoutfd
, int stderrfd
, int escape
)
400 return lxc_console(c
, ttynum
, stdinfd
, stdoutfd
, stderrfd
, escape
);
403 static pid_t
lxcapi_init_pid(struct lxc_container
*c
)
408 return lxc_cmd_get_init_pid(c
->name
, c
->config_path
);
411 static bool load_config_locked(struct lxc_container
*c
, const char *fname
)
414 c
->lxc_conf
= lxc_conf_init();
417 if (lxc_config_read(fname
, c
->lxc_conf
, false) != 0)
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 container_mem_unlock(c
);
472 static bool lxcapi_want_close_all_fds(struct lxc_container
*c
, bool state
)
474 if (!c
|| !c
->lxc_conf
)
476 if (container_mem_lock(c
)) {
477 ERROR("Error getting mem lock");
480 c
->lxc_conf
->close_all_fds
= state
;
481 container_mem_unlock(c
);
485 static bool lxcapi_wait(struct lxc_container
*c
, const char *state
, int timeout
)
492 ret
= lxc_wait(c
->name
, state
, timeout
, c
->config_path
);
497 static bool wait_on_daemonized_start(struct lxc_container
*c
, int pid
)
499 /* we'll probably want to make this timeout configurable? */
500 int timeout
= 5, ret
, status
;
503 * our child is going to fork again, then exit. reap the
506 ret
= waitpid(pid
, &status
, 0);
507 if (ret
== -1 || !WIFEXITED(status
) || WEXITSTATUS(status
) != 0)
508 DEBUG("failed waiting for first dual-fork child");
509 return lxcapi_wait(c
, "RUNNING", timeout
);
512 static bool am_single_threaded(void)
514 struct dirent dirent
, *direntp
;
518 dir
= opendir("/proc/self/task");
520 INFO("failed to open /proc/self/task");
524 while (!readdir_r(dir
, &dirent
, &direntp
)) {
528 if (!strcmp(direntp
->d_name
, "."))
531 if (!strcmp(direntp
->d_name
, ".."))
541 * I can't decide if it'd be more convenient for callers if we accept '...',
542 * or a null-terminated array (i.e. execl vs execv)
544 static bool lxcapi_start(struct lxc_container
*c
, int useinit
, char * const argv
[])
547 struct lxc_conf
*conf
;
548 bool daemonize
= false;
550 char *default_args
[] = {
556 /* container exists */
559 /* container has been setup */
563 if ((ret
= ongoing_create(c
)) < 0) {
564 ERROR("Error checking for incomplete creation");
568 ERROR("Error: %s creation was not completed", c
->name
);
571 } else if (ret
== 1) {
572 ERROR("Error: creation of %s is ongoing", c
->name
);
576 /* is this app meant to be run through lxcinit, as in lxc-execute? */
577 if (useinit
&& !argv
)
580 if (container_mem_lock(c
))
583 daemonize
= c
->daemonize
;
584 container_mem_unlock(c
);
587 ret
= lxc_execute(c
->name
, argv
, 1, conf
, c
->config_path
, daemonize
);
588 return ret
== 0 ? true : false;
592 if (conf
->init_cmd
) {
593 init_cmd
[0] = conf
->init_cmd
;
602 * say, I'm not sure - what locks do we want here? Any?
603 * Is liblxc's locking enough here to protect the on disk
604 * container? We don't want to exclude things like lxc_info
605 * while container is running...
609 lxc_monitord_spawn(c
->config_path
);
616 /* Set to NULL because we don't want father unlink
617 * the PID file, child will do the free and unlink.
620 return wait_on_daemonized_start(c
, pid
);
623 /* We don't really care if this doesn't print all the
624 * characters; all that it means is that the proctitle will be
625 * ugly. Similarly, we also don't care if setproctitle()
627 snprintf(title
, sizeof(title
), "[lxc monitor] %s %s", c
->config_path
, c
->name
);
628 INFO("Attempting to set proc title to %s", title
);
631 /* second fork to be reparented by init */
634 SYSERROR("Error doing dual-fork");
639 /* like daemon(), chdir to / and redirect 0,1,2 to /dev/null */
641 SYSERROR("Error chdir()ing to /.");
644 lxc_check_inherited(conf
, true, -1);
647 if (!am_single_threaded()) {
648 ERROR("Cannot start non-daemonized container when threaded");
653 /* We need to write PID file after daeminize, so we always
654 * write the right PID.
657 pid_fp
= fopen(c
->pidfile
, "w");
658 if (pid_fp
== NULL
) {
659 SYSERROR("Failed to create pidfile '%s' for '%s'",
660 c
->pidfile
, c
->name
);
664 if (fprintf(pid_fp
, "%d\n", getpid()) < 0) {
665 SYSERROR("Failed to write '%s'", c
->pidfile
);
678 if (lxc_check_inherited(conf
, daemonize
, -1)) {
679 ERROR("Inherited fds found");
684 ret
= lxc_start(c
->name
, argv
, conf
, c
->config_path
, daemonize
);
688 INFO("container requested reboot");
701 exit (ret
== 0 ? true : false);
703 return (ret
== 0 ? true : false);
707 * note there MUST be an ending NULL
709 static bool lxcapi_startl(struct lxc_container
*c
, int useinit
, ...)
712 char **inargs
= NULL
;
715 /* container exists */
719 va_start(ap
, useinit
);
720 inargs
= lxc_va_arg_list_to_argv(ap
, 0, 1);
724 ERROR("Memory allocation error.");
728 /* pass NULL if no arguments were supplied */
729 bret
= lxcapi_start(c
, useinit
, *inargs
? inargs
: NULL
);
734 for (arg
= inargs
; *arg
; arg
++)
742 static bool lxcapi_stop(struct lxc_container
*c
)
749 ret
= lxc_cmd_stop(c
->name
, c
->config_path
);
754 static int do_create_container_dir(const char *path
, struct lxc_conf
*conf
)
756 int ret
= -1, lasterr
;
757 char *p
= alloca(strlen(path
)+1);
758 mode_t mask
= umask(0002);
759 ret
= mkdir(path
, 0770);
767 SYSERROR("failed to create container path %s", path
);
772 if (!lxc_list_empty(&conf
->id_map
) && chown_mapped_root(p
, conf
) != 0) {
773 ERROR("Failed to chown container dir");
780 * create the standard expected container dir
782 static bool create_container_dir(struct lxc_container
*c
)
787 len
= strlen(c
->config_path
) + strlen(c
->name
) + 2;
791 ret
= snprintf(s
, len
, "%s/%s", c
->config_path
, c
->name
);
792 if (ret
< 0 || ret
>= len
) {
796 ret
= do_create_container_dir(s
, c
->lxc_conf
);
801 static const char *lxcapi_get_config_path(struct lxc_container
*c
);
802 static bool lxcapi_set_config_item(struct lxc_container
*c
, const char *key
, const char *v
);
805 * do_bdev_create: thin wrapper around bdev_create(). Like bdev_create(),
806 * it returns a mounted bdev on success, NULL on error.
808 static struct bdev
*do_bdev_create(struct lxc_container
*c
, const char *type
,
809 struct bdev_specs
*specs
)
816 /* rootfs.path or lxcpath/lxcname/rootfs */
817 if (c
->lxc_conf
->rootfs
.path
&& access(c
->lxc_conf
->rootfs
.path
, F_OK
) == 0) {
818 const char *rpath
= c
->lxc_conf
->rootfs
.path
;
819 len
= strlen(rpath
) + 1;
821 ret
= snprintf(dest
, len
, "%s", rpath
);
823 const char *lxcpath
= lxcapi_get_config_path(c
);
824 len
= strlen(c
->name
) + strlen(lxcpath
) + 9;
826 ret
= snprintf(dest
, len
, "%s/%s/rootfs", lxcpath
, c
->name
);
828 if (ret
< 0 || ret
>= len
)
831 bdev
= bdev_create(dest
, type
, c
->name
, specs
);
833 ERROR("Failed to create backing store type %s", type
);
837 lxcapi_set_config_item(c
, "lxc.rootfs", bdev
->src
);
839 /* if we are not root, chown the rootfs dir to root in the
842 if (geteuid() != 0 || (c
->lxc_conf
&& !lxc_list_empty(&c
->lxc_conf
->id_map
))) {
843 if (chown_mapped_root(bdev
->dest
, c
->lxc_conf
) < 0) {
844 ERROR("Error chowning %s to container root", bdev
->dest
);
845 suggest_default_idmap();
854 static char *lxcbasename(char *path
)
856 char *p
= path
+ strlen(path
) - 1;
857 while (*p
!= '/' && p
> path
)
862 static bool create_run_template(struct lxc_container
*c
, char *tpath
, bool quiet
,
872 SYSERROR("failed to fork task for container creation template");
876 if (pid
== 0) { // child
877 char *patharg
, *namearg
, *rootfsarg
, *src
;
878 struct bdev
*bdev
= NULL
;
880 int ret
, len
, nargs
= 0;
882 struct lxc_conf
*conf
= c
->lxc_conf
;
888 open("/dev/zero", O_RDONLY
);
889 open("/dev/null", O_RDWR
);
890 open("/dev/null", O_RDWR
);
893 src
= c
->lxc_conf
->rootfs
.path
;
895 * for an overlay create, what the user wants is the template to fill
896 * in what will become the readonly lower layer. So don't mount for
899 if (strncmp(src
, "overlayfs:", 10) == 0)
900 src
= overlay_getlower(src
+10);
901 if (strncmp(src
, "aufs:", 5) == 0)
902 src
= overlay_getlower(src
+5);
904 bdev
= bdev_init(c
->lxc_conf
, src
, c
->lxc_conf
->rootfs
.mount
, NULL
);
906 ERROR("Error opening rootfs");
910 if (geteuid() == 0) {
911 if (unshare(CLONE_NEWNS
) < 0) {
912 ERROR("error unsharing mounts");
915 if (detect_shared_rootfs()) {
916 if (mount(NULL
, "/", NULL
, MS_SLAVE
|MS_REC
, NULL
)) {
917 SYSERROR("Failed to make / rslave to run template");
918 ERROR("Continuing...");
922 if (strcmp(bdev
->type
, "dir") && strcmp(bdev
->type
, "btrfs")) {
923 if (geteuid() != 0) {
924 ERROR("non-root users can only create btrfs and directory-backed containers");
927 if (bdev
->ops
->mount(bdev
) < 0) {
928 ERROR("Error mounting rootfs");
931 } else { // TODO come up with a better way here!
933 bdev
->dest
= strdup(bdev
->src
);
937 * create our new array, pre-pend the template name and
941 for (nargs
= 0; argv
[nargs
]; nargs
++) ;
942 nargs
+= 4; // template, path, rootfs and name args
944 newargv
= malloc(nargs
* sizeof(*newargv
));
947 newargv
[0] = lxcbasename(tpath
);
949 len
= strlen(c
->config_path
) + strlen(c
->name
) + strlen("--path=") + 2;
950 patharg
= malloc(len
);
953 ret
= snprintf(patharg
, len
, "--path=%s/%s", c
->config_path
, c
->name
);
954 if (ret
< 0 || ret
>= len
)
956 newargv
[1] = patharg
;
957 len
= strlen("--name=") + strlen(c
->name
) + 1;
958 namearg
= malloc(len
);
961 ret
= snprintf(namearg
, len
, "--name=%s", c
->name
);
962 if (ret
< 0 || ret
>= len
)
964 newargv
[2] = namearg
;
966 len
= strlen("--rootfs=") + 1 + strlen(bdev
->dest
);
967 rootfsarg
= malloc(len
);
970 ret
= snprintf(rootfsarg
, len
, "--rootfs=%s", bdev
->dest
);
971 if (ret
< 0 || ret
>= len
)
973 newargv
[3] = rootfsarg
;
975 /* add passed-in args */
977 for (i
= 4; i
< nargs
; i
++)
978 newargv
[i
] = argv
[i
-4];
980 /* add trailing NULL */
982 newargv
= realloc(newargv
, nargs
* sizeof(*newargv
));
985 newargv
[nargs
- 1] = NULL
;
988 * If we're running the template in a mapped userns, then
989 * we prepend the template command with:
990 * lxc-usernsexec <-m map1> ... <-m mapn> --
991 * and we append "--mapped-uid x", where x is the mapped uid
994 if (!lxc_list_empty(&conf
->id_map
)) {
998 char **n2
= malloc(n2args
* sizeof(*n2
));
1003 SYSERROR("out of memory");
1007 tpath
= "lxc-usernsexec";
1008 n2
[0] = "lxc-usernsexec";
1009 lxc_list_for_each(it
, &conf
->id_map
) {
1012 n2
= realloc(n2
, n2args
* sizeof(char *));
1015 n2
[n2args
-2] = "-m";
1016 n2
[n2args
-1] = malloc(200);
1019 ret
= snprintf(n2
[n2args
-1], 200, "%c:%lu:%lu:%lu",
1020 map
->idtype
== ID_TYPE_UID
? 'u' : 'g',
1021 map
->nsid
, map
->hostid
, map
->range
);
1022 if (ret
< 0 || ret
>= 200)
1025 int hostid_mapped
= mapped_hostid(geteuid(), conf
, ID_TYPE_UID
);
1026 int extraargs
= hostid_mapped
>= 0 ? 1 : 3;
1027 n2
= realloc(n2
, (nargs
+ n2args
+ extraargs
) * sizeof(char *));
1030 if (hostid_mapped
< 0) {
1031 hostid_mapped
= find_unmapped_nsuid(conf
, ID_TYPE_UID
);
1032 n2
[n2args
++] = "-m";
1033 if (hostid_mapped
< 0) {
1034 ERROR("Could not find free uid to map");
1037 n2
[n2args
++] = malloc(200);
1038 if (!n2
[n2args
-1]) {
1039 SYSERROR("out of memory");
1042 ret
= snprintf(n2
[n2args
-1], 200, "u:%d:%d:1",
1043 hostid_mapped
, geteuid());
1044 if (ret
< 0 || ret
>= 200) {
1045 ERROR("string too long");
1049 int hostgid_mapped
= mapped_hostid(getegid(), conf
, ID_TYPE_GID
);
1050 extraargs
= hostgid_mapped
>= 0 ? 1 : 3;
1051 n2
= realloc(n2
, (nargs
+ n2args
+ extraargs
) * sizeof(char *));
1054 if (hostgid_mapped
< 0) {
1055 hostgid_mapped
= find_unmapped_nsuid(conf
, ID_TYPE_GID
);
1056 n2
[n2args
++] = "-m";
1057 if (hostgid_mapped
< 0) {
1058 ERROR("Could not find free uid to map");
1061 n2
[n2args
++] = malloc(200);
1062 if (!n2
[n2args
-1]) {
1063 SYSERROR("out of memory");
1066 ret
= snprintf(n2
[n2args
-1], 200, "g:%d:%d:1",
1067 hostgid_mapped
, getegid());
1068 if (ret
< 0 || ret
>= 200) {
1069 ERROR("string too long");
1073 n2
[n2args
++] = "--";
1074 for (i
= 0; i
< nargs
; i
++)
1075 n2
[i
+ n2args
] = newargv
[i
];
1077 // Finally add "--mapped-uid $uid" to tell template what to chown
1080 n2
= realloc(n2
, n2args
* sizeof(char *));
1082 SYSERROR("out of memory");
1085 // note n2[n2args-1] is NULL
1086 n2
[n2args
-5] = "--mapped-uid";
1087 snprintf(txtuid
, 20, "%d", hostid_mapped
);
1088 n2
[n2args
-4] = txtuid
;
1089 n2
[n2args
-3] = "--mapped-gid";
1090 snprintf(txtgid
, 20, "%d", hostgid_mapped
);
1091 n2
[n2args
-2] = txtgid
;
1092 n2
[n2args
-1] = NULL
;
1097 execvp(tpath
, newargv
);
1098 SYSERROR("failed to execute template %s", tpath
);
1102 if (wait_for_pid(pid
) != 0) {
1103 ERROR("container creation template for %s failed", c
->name
);
1110 static bool prepend_lxc_header(char *path
, const char *t
, char *const argv
[])
1118 unsigned char md_value
[SHA_DIGEST_LENGTH
];
1122 f
= fopen(path
, "r");
1126 if (fseek(f
, 0, SEEK_END
) < 0)
1128 if ((flen
= ftell(f
)) < 0)
1130 if (fseek(f
, 0, SEEK_SET
) < 0)
1132 if ((contents
= malloc(flen
+ 1)) == NULL
)
1134 if (fread(contents
, 1, flen
, f
) != flen
)
1135 goto out_free_contents
;
1137 contents
[flen
] = '\0';
1141 goto out_free_contents
;
1144 tpath
= get_template_path(t
);
1146 ERROR("bad template: %s", t
);
1147 goto out_free_contents
;
1150 ret
= sha1sum_file(tpath
, md_value
);
1152 ERROR("Error getting sha1sum of %s", tpath
);
1154 goto out_free_contents
;
1159 f
= fopen(path
, "w");
1161 SYSERROR("reopening config for writing");
1165 fprintf(f
, "# Template used to create this container: %s\n", t
);
1167 fprintf(f
, "# Parameters passed to the template:");
1169 fprintf(f
, " %s", *argv
);
1175 fprintf(f
, "# Template script checksum (SHA-1): ");
1176 for (i
=0; i
<SHA_DIGEST_LENGTH
; i
++)
1177 fprintf(f
, "%02x", md_value
[i
]);
1180 fprintf(f
, "# For additional config options, please look at lxc.container.conf(5)\n");
1181 if (fwrite(contents
, 1, flen
, f
) != flen
) {
1182 SYSERROR("Writing original contents");
1198 SYSERROR("Error prepending header");
1204 static void lxcapi_clear_config(struct lxc_container
*c
)
1208 lxc_conf_free(c
->lxc_conf
);
1214 static bool lxcapi_destroy(struct lxc_container
*c
);
1215 static bool container_destroy(struct lxc_container
*c
);
1216 static bool get_snappath_dir(struct lxc_container
*c
, char *snappath
);
1219 * create a container with the given parameters.
1220 * @c: container to be created. It has the lxcpath, name, and a starting
1221 * configuration already set
1222 * @t: the template to execute to instantiate the root filesystem and
1223 * adjust the configuration.
1224 * @bdevtype: backing store type to use. If NULL, dir will be used.
1225 * @specs: additional parameters for the backing store, i.e. LVM vg to
1228 * @argv: the arguments to pass to the template, terminated by NULL. If no
1229 * arguments, you can just pass NULL.
1231 static bool lxcapi_create(struct lxc_container
*c
, const char *t
,
1232 const char *bdevtype
, struct bdev_specs
*specs
, int flags
,
1244 tpath
= get_template_path(t
);
1246 ERROR("bad template: %s", t
);
1252 * If a template is passed in, and the rootfs already is defined in
1253 * the container config and exists, then * caller is trying to create
1254 * an existing container. Return an error, but do NOT delete the
1257 if (lxcapi_is_defined(c
) && c
->lxc_conf
&& c
->lxc_conf
->rootfs
.path
&&
1258 access(c
->lxc_conf
->rootfs
.path
, F_OK
) == 0 && tpath
) {
1259 ERROR("Container %s:%s already exists", c
->config_path
, c
->name
);
1264 if (!c
->load_config(c
, lxc_global_config_value("lxc.default_config"))) {
1265 ERROR("Error loading default configuration file %s", lxc_global_config_value("lxc.default_config"));
1270 if (!create_container_dir(c
))
1274 * either template or rootfs.path should be set.
1275 * if both template and rootfs.path are set, template is setup as rootfs.path.
1276 * container is already created if we have a config and rootfs.path is accessible
1278 if (!c
->lxc_conf
->rootfs
.path
&& !tpath
)
1279 /* no template passed in and rootfs does not exist: error */
1281 if (c
->lxc_conf
->rootfs
.path
&& access(c
->lxc_conf
->rootfs
.path
, F_OK
) != 0)
1282 /* rootfs passed into configuration, but does not exist: error */
1284 if (lxcapi_is_defined(c
) && c
->lxc_conf
->rootfs
.path
&& !tpath
) {
1285 /* Rootfs already existed, user just wanted to save the
1286 * loaded configuration */
1291 /* Mark that this container is being created */
1292 if ((partial_fd
= create_partial(c
)) < 0)
1295 /* no need to get disk lock bc we have the partial locked */
1298 * Create the backing store
1299 * Note we can't do this in the same task as we use to execute the
1300 * template because of the way zfs works.
1301 * After you 'zfs create', zfs mounts the fs only in the initial
1306 SYSERROR("failed to fork task for container creation template");
1310 if (pid
== 0) { // child
1311 struct bdev
*bdev
= NULL
;
1313 if (!(bdev
= do_bdev_create(c
, bdevtype
, specs
))) {
1314 ERROR("Error creating backing store type %s for %s",
1315 bdevtype
? bdevtype
: "(none)", c
->name
);
1319 /* save config file again to store the new rootfs location */
1320 if (!c
->save_config(c
, NULL
)) {
1321 ERROR("failed to save starting configuration for %s", c
->name
);
1322 // parent task won't see bdev in config so we delete it
1323 bdev
->ops
->umount(bdev
);
1324 bdev
->ops
->destroy(bdev
);
1329 if (wait_for_pid(pid
) != 0)
1332 /* reload config to get the rootfs */
1333 lxc_conf_free(c
->lxc_conf
);
1335 if (!load_config_locked(c
, c
->configfile
))
1338 if (!create_run_template(c
, tpath
, !!(flags
& LXC_CREATE_QUIET
), argv
))
1341 // now clear out the lxc_conf we have, reload from the created
1343 lxcapi_clear_config(c
);
1346 if (!prepend_lxc_header(c
->configfile
, tpath
, argv
)) {
1347 ERROR("Error prepending header to configuration file");
1351 ret
= load_config_locked(c
, c
->configfile
);
1354 if (partial_fd
>= 0)
1355 remove_partial(c
, partial_fd
);
1358 container_destroy(c
);
1364 static bool lxcapi_reboot(struct lxc_container
*c
)
1367 int rebootsignal
= SIGINT
;
1371 if (!c
->is_running(c
))
1373 pid
= c
->init_pid(c
);
1376 if (c
->lxc_conf
&& c
->lxc_conf
->rebootsignal
)
1377 rebootsignal
= c
->lxc_conf
->rebootsignal
;
1378 if (kill(pid
, rebootsignal
) < 0)
1384 static bool lxcapi_shutdown(struct lxc_container
*c
, int timeout
)
1388 int haltsignal
= SIGPWR
;
1393 if (!c
->is_running(c
))
1395 pid
= c
->init_pid(c
);
1398 if (c
->lxc_conf
&& c
->lxc_conf
->haltsignal
)
1399 haltsignal
= c
->lxc_conf
->haltsignal
;
1400 kill(pid
, haltsignal
);
1401 retv
= c
->wait(c
, "STOPPED", timeout
);
1405 static bool lxcapi_createl(struct lxc_container
*c
, const char *t
,
1406 const char *bdevtype
, struct bdev_specs
*specs
, int flags
, ...)
1416 * since we're going to wait for create to finish, I don't think we
1417 * need to get a copy of the arguments.
1419 va_start(ap
, flags
);
1420 args
= lxc_va_arg_list_to_argv(ap
, 0, 0);
1423 ERROR("Memory allocation error.");
1427 bret
= c
->create(c
, t
, bdevtype
, specs
, flags
, args
);
1434 static void do_clear_unexp_config_line(struct lxc_conf
*conf
, const char *key
)
1436 if (strcmp(key
, "lxc.cgroup") == 0)
1437 clear_unexp_config_line(conf
, key
, true);
1438 else if (strcmp(key
, "lxc.network") == 0)
1439 clear_unexp_config_line(conf
, key
, true);
1440 else if (strcmp(key
, "lxc.hook") == 0)
1441 clear_unexp_config_line(conf
, key
, true);
1443 clear_unexp_config_line(conf
, key
, false);
1444 if (!do_append_unexp_config_line(conf
, key
, ""))
1445 WARN("Error clearing configuration for %s", key
);
1448 static bool lxcapi_clear_config_item(struct lxc_container
*c
, const char *key
)
1452 if (!c
|| !c
->lxc_conf
)
1454 if (container_mem_lock(c
))
1456 ret
= lxc_clear_config_item(c
->lxc_conf
, key
);
1458 do_clear_unexp_config_line(c
->lxc_conf
, key
);
1459 container_mem_unlock(c
);
1463 static inline bool enter_net_ns(struct lxc_container
*c
)
1465 pid_t pid
= c
->init_pid(c
);
1467 if ((geteuid() != 0 || (c
->lxc_conf
&& !lxc_list_empty(&c
->lxc_conf
->id_map
))) && access("/proc/self/ns/user", F_OK
) == 0) {
1468 if (!switch_to_ns(pid
, "user"))
1471 return switch_to_ns(pid
, "net");
1474 // used by qsort and bsearch functions for comparing names
1475 static inline int string_cmp(char **first
, char **second
)
1477 return strcmp(*first
, *second
);
1480 // used by qsort and bsearch functions for comparing container names
1481 static inline int container_cmp(struct lxc_container
**first
, struct lxc_container
**second
)
1483 return strcmp((*first
)->name
, (*second
)->name
);
1486 static bool add_to_array(char ***names
, char *cname
, int pos
)
1488 char **newnames
= realloc(*names
, (pos
+1) * sizeof(char *));
1490 ERROR("Out of memory");
1495 newnames
[pos
] = strdup(cname
);
1499 // sort the arrray as we will use binary search on it
1500 qsort(newnames
, pos
+ 1, sizeof(char *), (int (*)(const void *,const void *))string_cmp
);
1505 static bool add_to_clist(struct lxc_container
***list
, struct lxc_container
*c
, int pos
, bool sort
)
1507 struct lxc_container
**newlist
= realloc(*list
, (pos
+1) * sizeof(struct lxc_container
*));
1509 ERROR("Out of memory");
1516 // sort the arrray as we will use binary search on it
1518 qsort(newlist
, pos
+ 1, sizeof(struct lxc_container
*), (int (*)(const void *,const void *))container_cmp
);
1523 static char** get_from_array(char ***names
, char *cname
, int size
)
1525 return (char **)bsearch(&cname
, *names
, size
, sizeof(char *), (int (*)(const void *, const void *))string_cmp
);
1529 static bool array_contains(char ***names
, char *cname
, int size
) {
1530 if(get_from_array(names
, cname
, size
) != NULL
)
1535 static bool remove_from_array(char ***names
, char *cname
, int size
)
1537 char **result
= get_from_array(names
, cname
, size
);
1538 if (result
!= NULL
) {
1545 static char** lxcapi_get_interfaces(struct lxc_container
*c
)
1548 int i
, count
= 0, pipefd
[2];
1549 char **interfaces
= NULL
;
1550 char interface
[IFNAMSIZ
];
1552 if(pipe(pipefd
) < 0) {
1553 SYSERROR("pipe failed");
1559 SYSERROR("failed to fork task to get interfaces information");
1565 if (pid
== 0) { // child
1566 int ret
= 1, nbytes
;
1567 struct ifaddrs
*interfaceArray
= NULL
, *tempIfAddr
= NULL
;
1569 /* close the read-end of the pipe */
1572 if (!enter_net_ns(c
)) {
1573 SYSERROR("failed to enter namespace");
1577 /* Grab the list of interfaces */
1578 if (getifaddrs(&interfaceArray
)) {
1579 SYSERROR("failed to get interfaces list");
1583 /* Iterate through the interfaces */
1584 for (tempIfAddr
= interfaceArray
; tempIfAddr
!= NULL
; tempIfAddr
= tempIfAddr
->ifa_next
) {
1585 nbytes
= write(pipefd
[1], tempIfAddr
->ifa_name
, IFNAMSIZ
);
1587 ERROR("write failed");
1596 freeifaddrs(interfaceArray
);
1598 /* close the write-end of the pipe, thus sending EOF to the reader */
1603 /* close the write-end of the pipe */
1606 while (read(pipefd
[0], &interface
, IFNAMSIZ
) == IFNAMSIZ
) {
1607 if (array_contains(&interfaces
, interface
, count
))
1610 if(!add_to_array(&interfaces
, interface
, count
))
1611 ERROR("PARENT: add_to_array failed");
1615 if (wait_for_pid(pid
) != 0) {
1616 for(i
=0;i
<count
;i
++)
1617 free(interfaces
[i
]);
1622 /* close the read-end of the pipe */
1625 /* Append NULL to the array */
1627 interfaces
= (char **)lxc_append_null_to_array((void **)interfaces
, count
);
1632 static char** lxcapi_get_ips(struct lxc_container
*c
, const char* interface
, const char* family
, int scope
)
1635 int i
, count
= 0, pipefd
[2];
1636 char **addresses
= NULL
;
1637 char address
[INET6_ADDRSTRLEN
];
1639 if(pipe(pipefd
) < 0) {
1640 SYSERROR("pipe failed");
1646 SYSERROR("failed to fork task to get container ips");
1652 if (pid
== 0) { // child
1653 int ret
= 1, nbytes
;
1654 struct ifaddrs
*interfaceArray
= NULL
, *tempIfAddr
= NULL
;
1655 char addressOutputBuffer
[INET6_ADDRSTRLEN
];
1656 void *tempAddrPtr
= NULL
;
1657 char *address
= NULL
;
1659 /* close the read-end of the pipe */
1662 if (!enter_net_ns(c
)) {
1663 SYSERROR("failed to enter namespace");
1667 /* Grab the list of interfaces */
1668 if (getifaddrs(&interfaceArray
)) {
1669 SYSERROR("failed to get interfaces list");
1673 /* Iterate through the interfaces */
1674 for (tempIfAddr
= interfaceArray
; tempIfAddr
!= NULL
; tempIfAddr
= tempIfAddr
->ifa_next
) {
1675 if (tempIfAddr
->ifa_addr
== NULL
)
1678 if(tempIfAddr
->ifa_addr
->sa_family
== AF_INET
) {
1679 if (family
&& strcmp(family
, "inet"))
1681 tempAddrPtr
= &((struct sockaddr_in
*)tempIfAddr
->ifa_addr
)->sin_addr
;
1684 if (family
&& strcmp(family
, "inet6"))
1687 if (((struct sockaddr_in6
*)tempIfAddr
->ifa_addr
)->sin6_scope_id
!= scope
)
1690 tempAddrPtr
= &((struct sockaddr_in6
*)tempIfAddr
->ifa_addr
)->sin6_addr
;
1693 if (interface
&& strcmp(interface
, tempIfAddr
->ifa_name
))
1695 else if (!interface
&& strcmp("lo", tempIfAddr
->ifa_name
) == 0)
1698 address
= (char *)inet_ntop(tempIfAddr
->ifa_addr
->sa_family
,
1700 addressOutputBuffer
,
1701 sizeof(addressOutputBuffer
));
1705 nbytes
= write(pipefd
[1], address
, INET6_ADDRSTRLEN
);
1707 ERROR("write failed");
1716 freeifaddrs(interfaceArray
);
1718 /* close the write-end of the pipe, thus sending EOF to the reader */
1723 /* close the write-end of the pipe */
1726 while (read(pipefd
[0], &address
, INET6_ADDRSTRLEN
) == INET6_ADDRSTRLEN
) {
1727 if(!add_to_array(&addresses
, address
, count
))
1728 ERROR("PARENT: add_to_array failed");
1732 if (wait_for_pid(pid
) != 0) {
1733 for(i
=0;i
<count
;i
++)
1739 /* close the read-end of the pipe */
1742 /* Append NULL to the array */
1744 addresses
= (char **)lxc_append_null_to_array((void **)addresses
, count
);
1749 static int lxcapi_get_config_item(struct lxc_container
*c
, const char *key
, char *retv
, int inlen
)
1753 if (!c
|| !c
->lxc_conf
)
1755 if (container_mem_lock(c
))
1757 ret
= lxc_get_config_item(c
->lxc_conf
, key
, retv
, inlen
);
1758 container_mem_unlock(c
);
1762 static char* lxcapi_get_running_config_item(struct lxc_container
*c
, const char *key
)
1766 if (!c
|| !c
->lxc_conf
)
1768 if (container_mem_lock(c
))
1770 ret
= lxc_cmd_get_config_item(c
->name
, key
, c
->get_config_path(c
));
1771 container_mem_unlock(c
);
1775 static int lxcapi_get_keys(struct lxc_container
*c
, const char *key
, char *retv
, int inlen
)
1778 return lxc_listconfigs(retv
, inlen
);
1780 * Support 'lxc.network.<idx>', i.e. 'lxc.network.0'
1781 * This is an intelligent result to show which keys are valid given
1782 * the type of nic it is
1784 if (!c
|| !c
->lxc_conf
)
1786 if (container_mem_lock(c
))
1789 if (strncmp(key
, "lxc.network.", 12) == 0)
1790 ret
= lxc_list_nicconfigs(c
->lxc_conf
, key
, retv
, inlen
);
1791 container_mem_unlock(c
);
1795 static bool lxcapi_save_config(struct lxc_container
*c
, const char *alt_file
)
1798 bool ret
= false, need_disklock
= false;
1802 alt_file
= c
->configfile
;
1804 return false; // should we write to stdout if no file is specified?
1806 // If we haven't yet loaded a config, load the stock config
1808 if (!c
->load_config(c
, lxc_global_config_value("lxc.default_config"))) {
1809 ERROR("Error loading default configuration file %s while saving %s", lxc_global_config_value("lxc.default_config"), c
->name
);
1814 if (!create_container_dir(c
))
1818 * If we're writing to the container's config file, take the
1819 * disk lock. Otherwise just take the memlock to protect the
1820 * struct lxc_container while we're traversing it.
1822 if (strcmp(c
->configfile
, alt_file
) == 0)
1823 need_disklock
= true;
1826 lret
= container_disk_lock(c
);
1828 lret
= container_mem_lock(c
);
1833 fout
= fopen(alt_file
, "w");
1836 write_config(fout
, c
->lxc_conf
);
1842 container_disk_unlock(c
);
1844 container_mem_unlock(c
);
1848 static bool mod_rdep(struct lxc_container
*c
, bool inc
)
1850 char path
[MAXPATHLEN
];
1855 if (container_disk_lock(c
))
1857 ret
= snprintf(path
, MAXPATHLEN
, "%s/%s/lxc_snapshots", c
->config_path
,
1859 if (ret
< 0 || ret
> MAXPATHLEN
)
1861 f
= fopen(path
, "r");
1863 ret
= fscanf(f
, "%d", &v
);
1866 ERROR("Corrupted file %s", path
);
1871 f
= fopen(path
, "w");
1874 if (fprintf(f
, "%d\n", v
) < 0) {
1875 ERROR("Error writing new snapshots value");
1881 SYSERROR("Error writing to or closing snapshots file");
1888 container_disk_unlock(c
);
1892 static void strip_newline(char *p
)
1894 size_t len
= strlen(p
);
1897 if (p
[len
-1] == '\n')
1901 static void mod_all_rdeps(struct lxc_container
*c
, bool inc
)
1903 struct lxc_container
*p
;
1904 char *lxcpath
= NULL
, *lxcname
= NULL
, path
[MAXPATHLEN
];
1905 size_t pathlen
= 0, namelen
= 0;
1909 ret
= snprintf(path
, MAXPATHLEN
, "%s/%s/lxc_rdepends",
1910 c
->config_path
, c
->name
);
1911 if (ret
< 0 || ret
>= MAXPATHLEN
) {
1912 ERROR("Path name too long");
1915 f
= fopen(path
, "r");
1918 while (getline(&lxcpath
, &pathlen
, f
) != -1) {
1919 if (getline(&lxcname
, &namelen
, f
) == -1) {
1920 ERROR("badly formatted file %s", path
);
1923 strip_newline(lxcpath
);
1924 strip_newline(lxcname
);
1925 if ((p
= lxc_container_new(lxcname
, lxcpath
)) == NULL
) {
1926 ERROR("Unable to find dependent container %s:%s",
1930 if (!mod_rdep(p
, inc
))
1931 ERROR("Failed to increase numsnapshots for %s:%s",
1933 lxc_container_put(p
);
1941 static bool has_fs_snapshots(struct lxc_container
*c
)
1943 char path
[MAXPATHLEN
];
1948 ret
= snprintf(path
, MAXPATHLEN
, "%s/%s/lxc_snapshots", c
->config_path
,
1950 if (ret
< 0 || ret
> MAXPATHLEN
)
1952 f
= fopen(path
, "r");
1955 ret
= fscanf(f
, "%d", &v
);
1965 static bool has_snapshots(struct lxc_container
*c
)
1967 char path
[MAXPATHLEN
];
1968 struct dirent dirent
, *direntp
;
1972 if (!get_snappath_dir(c
, path
))
1974 dir
= opendir(path
);
1977 while (!readdir_r(dir
, &dirent
, &direntp
)) {
1981 if (!strcmp(direntp
->d_name
, "."))
1984 if (!strcmp(direntp
->d_name
, ".."))
1993 static int lxc_rmdir_onedev_wrapper(void *data
)
1995 char *arg
= (char *) data
;
1996 return lxc_rmdir_onedev(arg
, "snaps");
1999 static int do_bdev_destroy(struct lxc_conf
*conf
)
2004 r
= bdev_init(conf
, conf
->rootfs
.path
, conf
->rootfs
.mount
, NULL
);
2008 if (r
->ops
->destroy(r
) < 0)
2014 static int bdev_destroy_wrapper(void *data
)
2016 struct lxc_conf
*conf
= data
;
2018 if (setgid(0) < 0) {
2019 ERROR("Failed to setgid to 0");
2022 if (setgroups(0, NULL
) < 0)
2023 WARN("Failed to clear groups");
2024 if (setuid(0) < 0) {
2025 ERROR("Failed to setuid to 0");
2028 return do_bdev_destroy(conf
);
2031 static bool container_destroy(struct lxc_container
*c
)
2036 if (!c
|| !lxcapi_is_defined(c
))
2039 if (container_disk_lock(c
))
2042 if (!is_stopped(c
)) {
2043 // we should queue some sort of error - in c->error_string?
2044 ERROR("container %s is not stopped", c
->name
);
2048 if (c
->lxc_conf
&& c
->lxc_conf
->rootfs
.path
&& c
->lxc_conf
->rootfs
.mount
) {
2050 ret
= userns_exec_1(c
->lxc_conf
, bdev_destroy_wrapper
, c
->lxc_conf
);
2052 ret
= do_bdev_destroy(c
->lxc_conf
);
2054 ERROR("Error destroying rootfs for %s", c
->name
);
2059 mod_all_rdeps(c
, false);
2061 const char *p1
= lxcapi_get_config_path(c
);
2062 char *path
= alloca(strlen(p1
) + strlen(c
->name
) + 2);
2063 sprintf(path
, "%s/%s", p1
, c
->name
);
2065 ret
= userns_exec_1(c
->lxc_conf
, lxc_rmdir_onedev_wrapper
, path
);
2067 ret
= lxc_rmdir_onedev(path
, "snaps");
2069 ERROR("Error destroying container directory for %s", c
->name
);
2075 container_disk_unlock(c
);
2079 static bool lxcapi_destroy(struct lxc_container
*c
)
2081 if (!c
|| !lxcapi_is_defined(c
))
2083 if (has_snapshots(c
)) {
2084 ERROR("Container %s has snapshots; not removing", c
->name
);
2088 if (has_fs_snapshots(c
)) {
2089 ERROR("container %s has snapshots on its rootfs", c
->name
);
2093 return container_destroy(c
);
2096 static bool lxcapi_snapshot_destroy_all(struct lxc_container
*c
);
2098 static bool lxcapi_destroy_with_snapshots(struct lxc_container
*c
)
2100 if (!c
|| !lxcapi_is_defined(c
))
2102 if (!lxcapi_snapshot_destroy_all(c
)) {
2103 ERROR("Error deleting all snapshots");
2106 return lxcapi_destroy(c
);
2109 static bool set_config_item_locked(struct lxc_container
*c
, const char *key
, const char *v
)
2111 struct lxc_config_t
*config
;
2114 c
->lxc_conf
= lxc_conf_init();
2117 config
= lxc_getconfig(key
);
2120 if (config
->cb(key
, v
, c
->lxc_conf
) != 0)
2122 return do_append_unexp_config_line(c
->lxc_conf
, key
, v
);
2125 static bool lxcapi_set_config_item(struct lxc_container
*c
, const char *key
, const char *v
)
2132 if (container_mem_lock(c
))
2135 b
= set_config_item_locked(c
, key
, v
);
2137 container_mem_unlock(c
);
2141 static char *lxcapi_config_file_name(struct lxc_container
*c
)
2143 if (!c
|| !c
->configfile
)
2145 return strdup(c
->configfile
);
2148 static const char *lxcapi_get_config_path(struct lxc_container
*c
)
2150 if (!c
|| !c
->config_path
)
2152 return (const char *)(c
->config_path
);
2157 * Just recalculate the c->configfile based on the
2158 * c->config_path, which must be set.
2159 * The lxc_container must be locked or not yet public.
2161 static bool set_config_filename(struct lxc_container
*c
)
2166 if (!c
->config_path
)
2169 /* $lxc_path + "/" + c->name + "/" + "config" + '\0' */
2170 len
= strlen(c
->config_path
) + strlen(c
->name
) + strlen("config") + 3;
2171 newpath
= malloc(len
);
2175 ret
= snprintf(newpath
, len
, "%s/%s/config", c
->config_path
, c
->name
);
2176 if (ret
< 0 || ret
>= len
) {
2177 fprintf(stderr
, "Error printing out config file name\n");
2182 free(c
->configfile
);
2183 c
->configfile
= newpath
;
2188 static bool lxcapi_set_config_path(struct lxc_container
*c
, const char *path
)
2192 char *oldpath
= NULL
;
2197 if (container_mem_lock(c
))
2202 ERROR("Out of memory setting new lxc path");
2208 oldpath
= c
->config_path
;
2211 /* Since we've changed the config path, we have to change the
2212 * config file name too */
2213 if (!set_config_filename(c
)) {
2214 ERROR("Out of memory setting new config filename");
2216 free(c
->config_path
);
2217 c
->config_path
= oldpath
;
2222 container_mem_unlock(c
);
2227 static bool lxcapi_set_cgroup_item(struct lxc_container
*c
, const char *subsys
, const char *value
)
2237 if (container_disk_lock(c
))
2240 ret
= lxc_cgroup_set(subsys
, value
, c
->name
, c
->config_path
);
2242 container_disk_unlock(c
);
2246 static int lxcapi_get_cgroup_item(struct lxc_container
*c
, const char *subsys
, char *retv
, int inlen
)
2256 if (container_disk_lock(c
))
2259 ret
= lxc_cgroup_get(subsys
, retv
, inlen
, c
->name
, c
->config_path
);
2261 container_disk_unlock(c
);
2265 const char *lxc_get_global_config_item(const char *key
)
2267 return lxc_global_config_value(key
);
2270 const char *lxc_get_version(void)
2275 static int copy_file(const char *old
, const char *new)
2282 if (file_exists(new)) {
2283 ERROR("copy destination %s exists", new);
2286 ret
= stat(old
, &sbuf
);
2288 INFO("Error stat'ing %s", old
);
2292 in
= open(old
, O_RDONLY
);
2294 SYSERROR("Error opening original file %s", old
);
2297 out
= open(new, O_CREAT
| O_EXCL
| O_WRONLY
, 0644);
2299 SYSERROR("Error opening new file %s", new);
2305 len
= read(in
, buf
, 8096);
2307 SYSERROR("Error reading old file %s", old
);
2312 ret
= write(out
, buf
, len
);
2313 if (ret
< len
) { // should we retry?
2314 SYSERROR("Error: write to new file %s was interrupted", new);
2321 // we set mode, but not owner/group
2322 ret
= chmod(new, sbuf
.st_mode
);
2324 SYSERROR("Error setting mode on %s", new);
2336 static int copyhooks(struct lxc_container
*oldc
, struct lxc_container
*c
)
2339 struct lxc_list
*it
;
2342 len
= strlen(oldc
->config_path
) + strlen(oldc
->name
) + 3;
2343 cpath
= alloca(len
);
2344 ret
= snprintf(cpath
, len
, "%s/%s/", oldc
->config_path
, oldc
->name
);
2345 if (ret
< 0 || ret
>= len
)
2348 for (i
=0; i
<NUM_LXC_HOOKS
; i
++) {
2349 lxc_list_for_each(it
, &c
->lxc_conf
->hooks
[i
]) {
2350 char *hookname
= it
->elem
;
2351 char *fname
= strrchr(hookname
, '/');
2352 char tmppath
[MAXPATHLEN
];
2353 if (!fname
) // relative path - we don't support, but maybe we should
2355 if (strncmp(hookname
, cpath
, len
- 1) != 0) {
2356 // this hook is public - ignore
2359 // copy the script, and change the entry in confile
2360 ret
= snprintf(tmppath
, MAXPATHLEN
, "%s/%s/%s",
2361 c
->config_path
, c
->name
, fname
+1);
2362 if (ret
< 0 || ret
>= MAXPATHLEN
)
2364 ret
= copy_file(it
->elem
, tmppath
);
2368 it
->elem
= strdup(tmppath
);
2370 ERROR("out of memory copying hook path");
2376 if (!clone_update_unexp_hooks(c
->lxc_conf
, oldc
->config_path
,
2377 c
->config_path
, oldc
->name
, c
->name
)) {
2378 ERROR("Error saving new hooks in clone");
2381 c
->save_config(c
, NULL
);
2386 static int copy_fstab(struct lxc_container
*oldc
, struct lxc_container
*c
)
2388 char newpath
[MAXPATHLEN
];
2389 char *oldpath
= oldc
->lxc_conf
->fstab
;
2395 clear_unexp_config_line(c
->lxc_conf
, "lxc.mount", false);
2397 char *p
= strrchr(oldpath
, '/');
2400 ret
= snprintf(newpath
, MAXPATHLEN
, "%s/%s%s",
2401 c
->config_path
, c
->name
, p
);
2402 if (ret
< 0 || ret
>= MAXPATHLEN
) {
2403 ERROR("error printing new path for %s", oldpath
);
2406 if (file_exists(newpath
)) {
2407 ERROR("error: fstab file %s exists", newpath
);
2411 if (copy_file(oldpath
, newpath
) < 0) {
2412 ERROR("error: copying %s to %s", oldpath
, newpath
);
2415 free(c
->lxc_conf
->fstab
);
2416 c
->lxc_conf
->fstab
= strdup(newpath
);
2417 if (!c
->lxc_conf
->fstab
) {
2418 ERROR("error: allocating pathname");
2421 if (!do_append_unexp_config_line(c
->lxc_conf
, "lxc.mount", newpath
)) {
2422 ERROR("error saving new lxctab");
2429 static void copy_rdepends(struct lxc_container
*c
, struct lxc_container
*c0
)
2431 char path0
[MAXPATHLEN
], path1
[MAXPATHLEN
];
2434 ret
= snprintf(path0
, MAXPATHLEN
, "%s/%s/lxc_rdepends", c0
->config_path
,
2436 if (ret
< 0 || ret
>= MAXPATHLEN
) {
2437 WARN("Error copying reverse dependencies");
2440 ret
= snprintf(path1
, MAXPATHLEN
, "%s/%s/lxc_rdepends", c
->config_path
,
2442 if (ret
< 0 || ret
>= MAXPATHLEN
) {
2443 WARN("Error copying reverse dependencies");
2446 if (copy_file(path0
, path1
) < 0) {
2447 INFO("Error copying reverse dependencies");
2452 static bool add_rdepends(struct lxc_container
*c
, struct lxc_container
*c0
)
2455 char path
[MAXPATHLEN
];
2459 ret
= snprintf(path
, MAXPATHLEN
, "%s/%s/lxc_rdepends", c
->config_path
,
2461 if (ret
< 0 || ret
>= MAXPATHLEN
)
2463 f
= fopen(path
, "a");
2467 // if anything goes wrong, just return an error
2468 if (fprintf(f
, "%s\n%s\n", c0
->config_path
, c0
->name
) < 0)
2475 static int copy_storage(struct lxc_container
*c0
, struct lxc_container
*c
,
2476 const char *newtype
, int flags
, const char *bdevdata
, uint64_t newsize
)
2481 bdev
= bdev_copy(c0
, c
->name
, c
->config_path
, newtype
, flags
,
2482 bdevdata
, newsize
, &need_rdep
);
2484 ERROR("Error copying storage");
2487 free(c
->lxc_conf
->rootfs
.path
);
2488 c
->lxc_conf
->rootfs
.path
= strdup(bdev
->src
);
2490 if (!c
->lxc_conf
->rootfs
.path
) {
2491 ERROR("Out of memory while setting storage path");
2494 // We will simply append a new lxc.rootfs entry to the unexpanded config
2495 clear_unexp_config_line(c
->lxc_conf
, "lxc.rootfs", false);
2496 if (!do_append_unexp_config_line(c
->lxc_conf
, "lxc.rootfs", c
->lxc_conf
->rootfs
.path
)) {
2497 ERROR("Error saving new rootfs to cloend config");
2500 if (flags
& LXC_CLONE_SNAPSHOT
)
2501 copy_rdepends(c
, c0
);
2503 if (!add_rdepends(c
, c0
))
2504 WARN("Error adding reverse dependency from %s to %s",
2508 mod_all_rdeps(c
, true);
2513 struct clone_update_data
{
2514 struct lxc_container
*c0
;
2515 struct lxc_container
*c1
;
2520 static int clone_update_rootfs(struct clone_update_data
*data
)
2522 struct lxc_container
*c0
= data
->c0
;
2523 struct lxc_container
*c
= data
->c1
;
2524 int flags
= data
->flags
;
2525 char **hookargs
= data
->hookargs
;
2527 char path
[MAXPATHLEN
];
2530 struct lxc_conf
*conf
= c
->lxc_conf
;
2532 /* update hostname in rootfs */
2533 /* we're going to mount, so run in a clean namespace to simplify cleanup */
2535 if (setgid(0) < 0) {
2536 ERROR("Failed to setgid to 0");
2539 if (setuid(0) < 0) {
2540 ERROR("Failed to setuid to 0");
2543 if (setgroups(0, NULL
) < 0)
2544 WARN("Failed to clear groups");
2546 if (unshare(CLONE_NEWNS
) < 0)
2548 bdev
= bdev_init(c
->lxc_conf
, c
->lxc_conf
->rootfs
.path
, c
->lxc_conf
->rootfs
.mount
, NULL
);
2551 if (strcmp(bdev
->type
, "dir") != 0) {
2552 if (unshare(CLONE_NEWNS
) < 0) {
2553 ERROR("error unsharing mounts");
2557 if (detect_shared_rootfs()) {
2558 if (mount(NULL
, "/", NULL
, MS_SLAVE
|MS_REC
, NULL
)) {
2559 SYSERROR("Failed to make / rslave");
2560 ERROR("Continuing...");
2563 if (bdev
->ops
->mount(bdev
) < 0) {
2567 } else { // TODO come up with a better way
2569 bdev
->dest
= strdup(bdev
->src
);
2572 if (!lxc_list_empty(&conf
->hooks
[LXCHOOK_CLONE
])) {
2573 /* Start of environment variable setup for hooks */
2574 if (setenv("LXC_SRC_NAME", c0
->name
, 1)) {
2575 SYSERROR("failed to set environment variable for source container name");
2577 if (setenv("LXC_NAME", c
->name
, 1)) {
2578 SYSERROR("failed to set environment variable for container name");
2580 if (setenv("LXC_CONFIG_FILE", conf
->rcfile
, 1)) {
2581 SYSERROR("failed to set environment variable for config path");
2583 if (setenv("LXC_ROOTFS_MOUNT", bdev
->dest
, 1)) {
2584 SYSERROR("failed to set environment variable for rootfs mount");
2586 if (setenv("LXC_ROOTFS_PATH", conf
->rootfs
.path
, 1)) {
2587 SYSERROR("failed to set environment variable for rootfs mount");
2590 if (run_lxc_hooks(c
->name
, "clone", conf
, c
->get_config_path(c
), hookargs
)) {
2591 ERROR("Error executing clone hook for %s", c
->name
);
2597 if (!(flags
& LXC_CLONE_KEEPNAME
)) {
2598 ret
= snprintf(path
, MAXPATHLEN
, "%s/etc/hostname", bdev
->dest
);
2601 if (ret
< 0 || ret
>= MAXPATHLEN
)
2603 if (!file_exists(path
))
2605 if (!(fout
= fopen(path
, "w"))) {
2606 SYSERROR("unable to open %s: ignoring", path
);
2609 if (fprintf(fout
, "%s", c
->name
) < 0) {
2613 if (fclose(fout
) < 0)
2622 static int clone_update_rootfs_wrapper(void *data
)
2624 struct clone_update_data
*arg
= (struct clone_update_data
*) data
;
2625 return clone_update_rootfs(arg
);
2629 * We want to support:
2630 sudo lxc-clone -o o1 -n n1 -s -L|-fssize fssize -v|--vgname vgname \
2631 -p|--lvprefix lvprefix -t|--fstype fstype -B backingstore
2633 -s [ implies overlayfs]
2637 only rootfs gets converted (copied/snapshotted) on clone.
2640 static int create_file_dirname(char *path
, struct lxc_conf
*conf
)
2642 char *p
= strrchr(path
, '/');
2648 ret
= do_create_container_dir(path
, conf
);
2653 static struct lxc_container
*lxcapi_clone(struct lxc_container
*c
, const char *newname
,
2654 const char *lxcpath
, int flags
,
2655 const char *bdevtype
, const char *bdevdata
, uint64_t newsize
,
2658 struct lxc_container
*c2
= NULL
;
2659 char newpath
[MAXPATHLEN
];
2660 int ret
, storage_copied
= 0;
2661 char *origroot
= NULL
;
2662 struct clone_update_data data
;
2666 if (!c
|| !c
->is_defined(c
))
2669 if (container_mem_lock(c
))
2672 if (!is_stopped(c
)) {
2673 ERROR("error: Original container (%s) is running", c
->name
);
2677 // Make sure the container doesn't yet exist.
2681 lxcpath
= c
->get_config_path(c
);
2682 ret
= snprintf(newpath
, MAXPATHLEN
, "%s/%s/config", lxcpath
, newname
);
2683 if (ret
< 0 || ret
>= MAXPATHLEN
) {
2684 SYSERROR("clone: failed making config pathname");
2687 if (file_exists(newpath
)) {
2688 ERROR("error: clone: %s exists", newpath
);
2692 ret
= create_file_dirname(newpath
, c
->lxc_conf
);
2693 if (ret
< 0 && errno
!= EEXIST
) {
2694 ERROR("Error creating container dir for %s", newpath
);
2698 // copy the configuration, tweak it as needed,
2699 if (c
->lxc_conf
->rootfs
.path
) {
2700 origroot
= c
->lxc_conf
->rootfs
.path
;
2701 c
->lxc_conf
->rootfs
.path
= NULL
;
2703 fout
= fopen(newpath
, "w");
2705 SYSERROR("open %s", newpath
);
2708 write_config(fout
, c
->lxc_conf
);
2710 c
->lxc_conf
->rootfs
.path
= origroot
;
2712 sprintf(newpath
, "%s/%s/rootfs", lxcpath
, newname
);
2713 if (mkdir(newpath
, 0755) < 0) {
2714 SYSERROR("error creating %s", newpath
);
2719 if (chown_mapped_root(newpath
, c
->lxc_conf
) < 0) {
2720 ERROR("Error chowning %s to container root", newpath
);
2725 c2
= lxc_container_new(newname
, lxcpath
);
2727 ERROR("clone: failed to create new container (%s %s)", newname
,
2732 // copy/snapshot rootfs's
2733 ret
= copy_storage(c
, c2
, bdevtype
, flags
, bdevdata
, newsize
);
2737 clear_unexp_config_line(c2
->lxc_conf
, "lxc.utsname", false);
2740 if (!set_config_item_locked(c2
, "lxc.utsname", newname
)) {
2741 ERROR("Error setting new hostname");
2746 ret
= copyhooks(c
, c2
);
2748 ERROR("error copying hooks");
2752 if (copy_fstab(c
, c2
) < 0) {
2753 ERROR("error copying fstab");
2758 if (!(flags
& LXC_CLONE_KEEPMACADDR
)) {
2759 if (!network_new_hwaddrs(c2
->lxc_conf
)) {
2760 ERROR("Error updating mac addresses");
2765 // We've now successfully created c2's storage, so clear it out if we
2769 if (!c2
->save_config(c2
, NULL
))
2772 if ((pid
= fork()) < 0) {
2777 ret
= wait_for_pid(pid
);
2780 container_mem_unlock(c
);
2786 data
.hookargs
= hookargs
;
2788 ret
= userns_exec_1(c
->lxc_conf
, clone_update_rootfs_wrapper
,
2791 ret
= clone_update_rootfs(&data
);
2795 container_mem_unlock(c
);
2799 container_mem_unlock(c
);
2801 if (!storage_copied
)
2802 c2
->lxc_conf
->rootfs
.path
= NULL
;
2804 lxc_container_put(c2
);
2810 static bool lxcapi_rename(struct lxc_container
*c
, const char *newname
)
2813 struct lxc_container
*newc
;
2815 if (!c
|| !c
->name
|| !c
->config_path
|| !c
->lxc_conf
)
2818 if (has_fs_snapshots(c
) || has_snapshots(c
)) {
2819 ERROR("Renaming a container with snapshots is not supported");
2822 bdev
= bdev_init(c
->lxc_conf
, c
->lxc_conf
->rootfs
.path
, c
->lxc_conf
->rootfs
.mount
, NULL
);
2824 ERROR("Failed to find original backing store type");
2828 newc
= lxcapi_clone(c
, newname
, c
->config_path
, LXC_CLONE_KEEPMACADDR
, NULL
, bdev
->type
, 0, NULL
);
2831 lxc_container_put(newc
);
2835 if (newc
&& lxcapi_is_defined(newc
))
2836 lxc_container_put(newc
);
2838 if (!container_destroy(c
)) {
2839 ERROR("Could not destroy existing container %s", c
->name
);
2845 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
)
2850 return lxc_attach(c
->name
, c
->config_path
, exec_function
, exec_payload
, options
, attached_process
);
2853 static int lxcapi_attach_run_wait(struct lxc_container
*c
, lxc_attach_options_t
*options
, const char *program
, const char * const argv
[])
2855 lxc_attach_command_t command
;
2862 command
.program
= (char*)program
;
2863 command
.argv
= (char**)argv
;
2864 r
= lxc_attach(c
->name
, c
->config_path
, lxc_attach_run_command
, &command
, options
, &pid
);
2869 return lxc_wait_for_pid_status(pid
);
2872 static int get_next_index(const char *lxcpath
, char *cname
)
2878 fname
= alloca(strlen(lxcpath
) + 20);
2880 sprintf(fname
, "%s/snap%d", lxcpath
, i
);
2881 ret
= stat(fname
, &sb
);
2888 static bool get_snappath_dir(struct lxc_container
*c
, char *snappath
)
2892 * If the old style snapshot path exists, use it
2893 * /var/lib/lxc -> /var/lib/lxcsnaps
2895 ret
= snprintf(snappath
, MAXPATHLEN
, "%ssnaps", c
->config_path
);
2896 if (ret
< 0 || ret
>= MAXPATHLEN
)
2898 if (dir_exists(snappath
)) {
2899 ret
= snprintf(snappath
, MAXPATHLEN
, "%ssnaps/%s", c
->config_path
, c
->name
);
2900 if (ret
< 0 || ret
>= MAXPATHLEN
)
2906 * Use the new style path
2907 * /var/lib/lxc -> /var/lib/lxc + c->name + /snaps + \0
2909 ret
= snprintf(snappath
, MAXPATHLEN
, "%s/%s/snaps", c
->config_path
, c
->name
);
2910 if (ret
< 0 || ret
>= MAXPATHLEN
)
2915 static int lxcapi_snapshot(struct lxc_container
*c
, const char *commentfile
)
2918 struct lxc_container
*c2
;
2919 char snappath
[MAXPATHLEN
], newname
[20];
2921 if (!c
|| !lxcapi_is_defined(c
))
2924 if (!bdev_can_backup(c
->lxc_conf
)) {
2925 ERROR("%s's backing store cannot be backed up.", c
->name
);
2926 ERROR("Your container must use another backing store type.");
2930 if (!get_snappath_dir(c
, snappath
))
2933 i
= get_next_index(snappath
, c
->name
);
2935 if (mkdir_p(snappath
, 0755) < 0) {
2936 ERROR("Failed to create snapshot directory %s", snappath
);
2940 ret
= snprintf(newname
, 20, "snap%d", i
);
2941 if (ret
< 0 || ret
>= 20)
2945 * We pass LXC_CLONE_SNAPSHOT to make sure that a rdepends file entry is
2946 * created in the original container
2948 flags
= LXC_CLONE_SNAPSHOT
| LXC_CLONE_KEEPMACADDR
| LXC_CLONE_KEEPNAME
|
2949 LXC_CLONE_KEEPBDEVTYPE
| LXC_CLONE_MAYBE_SNAPSHOT
;
2950 if (bdev_is_dir(c
->lxc_conf
, c
->lxc_conf
->rootfs
.path
)) {
2951 ERROR("Snapshot of directory-backed container requested.");
2952 ERROR("Making a copy-clone. If you do want snapshots, then");
2953 ERROR("please create an aufs or overlayfs clone first, snapshot that");
2954 ERROR("and keep the original container pristine.");
2955 flags
&= ~LXC_CLONE_SNAPSHOT
| LXC_CLONE_MAYBE_SNAPSHOT
;
2957 c2
= c
->clone(c
, newname
, snappath
, flags
, NULL
, NULL
, 0, NULL
);
2959 ERROR("clone of %s:%s failed", c
->config_path
, c
->name
);
2963 lxc_container_put(c2
);
2965 // Now write down the creation time
2972 tm_info
= localtime(&timer
);
2974 strftime(buffer
, 25, "%Y:%m:%d %H:%M:%S", tm_info
);
2976 char *dfnam
= alloca(strlen(snappath
) + strlen(newname
) + 5);
2977 sprintf(dfnam
, "%s/%s/ts", snappath
, newname
);
2978 f
= fopen(dfnam
, "w");
2980 ERROR("Failed to open %s", dfnam
);
2983 if (fprintf(f
, "%s", buffer
) < 0) {
2984 SYSERROR("Writing timestamp");
2990 SYSERROR("Writing timestamp");
2995 // $p / $name / comment \0
2996 int len
= strlen(snappath
) + strlen(newname
) + 10;
2997 char *path
= alloca(len
);
2998 sprintf(path
, "%s/%s/comment", snappath
, newname
);
2999 return copy_file(commentfile
, path
) < 0 ? -1 : i
;
3005 static void lxcsnap_free(struct lxc_snapshot
*s
)
3008 free(s
->comment_pathname
);
3013 static char *get_snapcomment_path(char* snappath
, char *name
)
3015 // $snappath/$name/comment
3016 int ret
, len
= strlen(snappath
) + strlen(name
) + 10;
3017 char *s
= malloc(len
);
3020 ret
= snprintf(s
, len
, "%s/%s/comment", snappath
, name
);
3021 if (ret
< 0 || ret
>= len
) {
3029 static char *get_timestamp(char* snappath
, char *name
)
3031 char path
[MAXPATHLEN
], *s
= NULL
;
3035 ret
= snprintf(path
, MAXPATHLEN
, "%s/%s/ts", snappath
, name
);
3036 if (ret
< 0 || ret
>= MAXPATHLEN
)
3038 fin
= fopen(path
, "r");
3041 (void) fseek(fin
, 0, SEEK_END
);
3043 (void) fseek(fin
, 0, SEEK_SET
);
3048 if (fread(s
, 1, len
, fin
) != len
) {
3049 SYSERROR("reading timestamp");
3059 static int lxcapi_snapshot_list(struct lxc_container
*c
, struct lxc_snapshot
**ret_snaps
)
3061 char snappath
[MAXPATHLEN
], path2
[MAXPATHLEN
];
3063 struct dirent dirent
, *direntp
;
3064 struct lxc_snapshot
*snaps
=NULL
, *nsnaps
;
3067 if (!c
|| !lxcapi_is_defined(c
))
3070 if (!get_snappath_dir(c
, snappath
)) {
3071 ERROR("path name too long");
3074 dir
= opendir(snappath
);
3076 INFO("failed to open %s - assuming no snapshots", snappath
);
3080 while (!readdir_r(dir
, &dirent
, &direntp
)) {
3084 if (!strcmp(direntp
->d_name
, "."))
3087 if (!strcmp(direntp
->d_name
, ".."))
3090 ret
= snprintf(path2
, MAXPATHLEN
, "%s/%s/config", snappath
, direntp
->d_name
);
3091 if (ret
< 0 || ret
>= MAXPATHLEN
) {
3092 ERROR("pathname too long");
3095 if (!file_exists(path2
))
3097 nsnaps
= realloc(snaps
, (count
+ 1)*sizeof(*snaps
));
3099 SYSERROR("Out of memory");
3103 snaps
[count
].free
= lxcsnap_free
;
3104 snaps
[count
].name
= strdup(direntp
->d_name
);
3105 if (!snaps
[count
].name
)
3107 snaps
[count
].lxcpath
= strdup(snappath
);
3108 if (!snaps
[count
].lxcpath
) {
3109 free(snaps
[count
].name
);
3112 snaps
[count
].comment_pathname
= get_snapcomment_path(snappath
, direntp
->d_name
);
3113 snaps
[count
].timestamp
= get_timestamp(snappath
, direntp
->d_name
);
3118 WARN("failed to close directory");
3126 for (i
=0; i
<count
; i
++)
3127 lxcsnap_free(&snaps
[i
]);
3131 WARN("failed to close directory");
3135 static bool lxcapi_snapshot_restore(struct lxc_container
*c
, const char *snapname
, const char *newname
)
3137 char clonelxcpath
[MAXPATHLEN
];
3139 struct lxc_container
*snap
, *rest
;
3143 if (!c
|| !c
->name
|| !c
->config_path
)
3146 if (has_fs_snapshots(c
)) {
3147 ERROR("container rootfs has dependent snapshots");
3151 bdev
= bdev_init(c
->lxc_conf
, c
->lxc_conf
->rootfs
.path
, c
->lxc_conf
->rootfs
.mount
, NULL
);
3153 ERROR("Failed to find original backing store type");
3160 if (!get_snappath_dir(c
, clonelxcpath
)) {
3164 // how should we lock this?
3166 snap
= lxc_container_new(snapname
, clonelxcpath
);
3167 if (!snap
|| !lxcapi_is_defined(snap
)) {
3168 ERROR("Could not open snapshot %s", snapname
);
3169 if (snap
) lxc_container_put(snap
);
3174 if (strcmp(c
->name
, newname
) == 0) {
3175 if (!container_destroy(c
)) {
3176 ERROR("Could not destroy existing container %s", newname
);
3177 lxc_container_put(snap
);
3183 if (strcmp(bdev
->type
, "dir") != 0 && strcmp(bdev
->type
, "loop") != 0)
3184 flags
= LXC_CLONE_SNAPSHOT
| LXC_CLONE_MAYBE_SNAPSHOT
;
3185 rest
= lxcapi_clone(snap
, newname
, c
->config_path
, flags
,
3186 bdev
->type
, NULL
, 0, NULL
);
3188 if (rest
&& lxcapi_is_defined(rest
))
3191 lxc_container_put(rest
);
3192 lxc_container_put(snap
);
3196 static bool do_snapshot_destroy(const char *snapname
, const char *clonelxcpath
)
3198 struct lxc_container
*snap
= NULL
;
3201 snap
= lxc_container_new(snapname
, clonelxcpath
);
3203 ERROR("Could not find snapshot %s", snapname
);
3207 if (!lxcapi_destroy(snap
)) {
3208 ERROR("Could not destroy snapshot %s", snapname
);
3215 lxc_container_put(snap
);
3219 static bool remove_all_snapshots(const char *path
)
3222 struct dirent dirent
, *direntp
;
3225 dir
= opendir(path
);
3227 SYSERROR("opendir on snapshot path %s", path
);
3230 while (!readdir_r(dir
, &dirent
, &direntp
)) {
3233 if (!strcmp(direntp
->d_name
, "."))
3235 if (!strcmp(direntp
->d_name
, ".."))
3237 if (!do_snapshot_destroy(direntp
->d_name
, path
)) {
3246 SYSERROR("Error removing directory %s", path
);
3251 static bool lxcapi_snapshot_destroy(struct lxc_container
*c
, const char *snapname
)
3253 char clonelxcpath
[MAXPATHLEN
];
3255 if (!c
|| !c
->name
|| !c
->config_path
|| !snapname
)
3258 if (!get_snappath_dir(c
, clonelxcpath
))
3261 return do_snapshot_destroy(snapname
, clonelxcpath
);
3264 static bool lxcapi_snapshot_destroy_all(struct lxc_container
*c
)
3266 char clonelxcpath
[MAXPATHLEN
];
3268 if (!c
|| !c
->name
|| !c
->config_path
)
3271 if (!get_snappath_dir(c
, clonelxcpath
))
3274 return remove_all_snapshots(clonelxcpath
);
3277 static bool lxcapi_may_control(struct lxc_container
*c
)
3279 return lxc_try_cmd(c
->name
, c
->config_path
) == 0;
3282 static bool do_add_remove_node(pid_t init_pid
, const char *path
, bool add
,
3285 char chrootpath
[MAXPATHLEN
];
3286 char *directory_path
= NULL
;
3290 if ((pid
= fork()) < 0) {
3291 SYSERROR("failed to fork a child helper");
3295 if (wait_for_pid(pid
) != 0) {
3296 ERROR("Failed to create note in guest");
3302 /* prepare the path */
3303 ret
= snprintf(chrootpath
, MAXPATHLEN
, "/proc/%d/root", init_pid
);
3304 if (ret
< 0 || ret
>= MAXPATHLEN
)
3307 if (chroot(chrootpath
) < 0)
3311 /* remove path if it exists */
3312 if(faccessat(AT_FDCWD
, path
, F_OK
, AT_SYMLINK_NOFOLLOW
) == 0) {
3313 if (unlink(path
) < 0) {
3314 ERROR("unlink failed");
3321 /* create any missing directories */
3322 directory_path
= dirname(strdup(path
));
3323 if (mkdir_p(directory_path
, 0755) < 0 && errno
!= EEXIST
) {
3324 ERROR("failed to create directory");
3328 /* create the device node */
3329 if (mknod(path
, st
->st_mode
, st
->st_rdev
) < 0) {
3330 ERROR("mknod failed");
3337 static bool add_remove_device_node(struct lxc_container
*c
, const char *src_path
, const char *dest_path
, bool add
)
3341 char value
[MAX_BUFFER
];
3344 /* make sure container is running */
3345 if (!c
->is_running(c
)) {
3346 ERROR("container is not running");
3350 /* use src_path if dest_path is NULL otherwise use dest_path */
3351 p
= dest_path
? dest_path
: src_path
;
3353 /* make sure we can access p */
3354 if(access(p
, F_OK
) < 0 || stat(p
, &st
) < 0)
3357 /* continue if path is character device or block device */
3358 if (S_ISCHR(st
.st_mode
))
3359 ret
= snprintf(value
, MAX_BUFFER
, "c %d:%d rwm", major(st
.st_rdev
), minor(st
.st_rdev
));
3360 else if (S_ISBLK(st
.st_mode
))
3361 ret
= snprintf(value
, MAX_BUFFER
, "b %d:%d rwm", major(st
.st_rdev
), minor(st
.st_rdev
));
3365 /* check snprintf return code */
3366 if (ret
< 0 || ret
>= MAX_BUFFER
)
3369 if (!do_add_remove_node(c
->init_pid(c
), p
, add
, &st
))
3372 /* add or remove device to/from cgroup access list */
3374 if (!c
->set_cgroup_item(c
, "devices.allow", value
)) {
3375 ERROR("set_cgroup_item failed while adding the device node");
3379 if (!c
->set_cgroup_item(c
, "devices.deny", value
)) {
3380 ERROR("set_cgroup_item failed while removing the device node");
3388 static bool lxcapi_add_device_node(struct lxc_container
*c
, const char *src_path
, const char *dest_path
)
3391 ERROR(NOT_SUPPORTED_ERROR
, __FUNCTION__
);
3394 return add_remove_device_node(c
, src_path
, dest_path
, true);
3397 static bool lxcapi_remove_device_node(struct lxc_container
*c
, const char *src_path
, const char *dest_path
)
3400 ERROR(NOT_SUPPORTED_ERROR
, __FUNCTION__
);
3403 return add_remove_device_node(c
, src_path
, dest_path
, false);
3406 static bool lxcapi_attach_interface(struct lxc_container
*c
, const char *ifname
,
3407 const char *dst_ifname
)
3411 ERROR(NOT_SUPPORTED_ERROR
, __FUNCTION__
);
3416 ERROR("No source interface name given");
3420 ret
= lxc_netdev_isup(ifname
);
3423 /* netdev of ifname is up. */
3424 ret
= lxc_netdev_down(ifname
);
3429 ret
= lxc_netdev_move_by_name(ifname
, c
->init_pid(c
), dst_ifname
);
3439 static bool lxcapi_detach_interface(struct lxc_container
*c
, const char *ifname
,
3440 const char *dst_ifname
)
3442 pid_t pid
, pid_outside
;
3445 ERROR(NOT_SUPPORTED_ERROR
, __FUNCTION__
);
3450 ERROR("No source interface name given");
3454 pid_outside
= getpid();
3457 ERROR("failed to fork task to get interfaces information");
3461 if (pid
== 0) { // child
3463 if (!enter_net_ns(c
)) {
3464 ERROR("failed to enter namespace");
3468 ret
= lxc_netdev_isup(ifname
);
3472 /* netdev of ifname is up. */
3474 ret
= lxc_netdev_down(ifname
);
3479 ret
= lxc_netdev_move_by_name(ifname
, pid_outside
, dst_ifname
);
3481 /* -EINVAL means there is no netdev named as ifanme. */
3482 if (ret
== -EINVAL
) {
3483 ERROR("No network device named as %s.", ifname
);
3488 if (wait_for_pid(pid
) != 0)
3494 static bool lxcapi_checkpoint(struct lxc_container
*c
, char *directory
, bool stop
, bool verbose
)
3502 if (mkdir(directory
, 0700) < 0 && errno
!= EEXIST
)
3505 if (!dump_net_info(c
, directory
))
3513 struct criu_opts os
;
3516 os
.directory
= directory
;
3519 os
.verbose
= verbose
;
3521 /* exec_criu() returning is an error */
3525 pid_t w
= waitpid(pid
, &status
, 0);
3527 SYSERROR("waitpid");
3531 if (WIFEXITED(status
)) {
3532 return !WEXITSTATUS(status
);
3539 static bool lxcapi_restore(struct lxc_container
*c
, char *directory
, bool verbose
)
3549 ERROR("Must be root to restore\n");
3554 ERROR("failed to create pipe");
3567 // this never returns
3568 do_restore(c
, pipefd
[1], directory
, verbose
);
3573 nread
= read(pipefd
[0], &status
, sizeof(status
));
3575 if (sizeof(status
) != nread
) {
3576 ERROR("reading status from pipe failed");
3580 // If the criu process was killed or exited nonzero, wait() for the
3581 // handler, since the restore process died. Otherwise, we don't need to
3582 // wait, since the child becomes the monitor process.
3583 if (!WIFEXITED(status
) || WEXITSTATUS(status
))
3588 if (wait_for_pid(pid
))
3589 ERROR("restore process died");
3593 static int lxcapi_attach_run_waitl(struct lxc_container
*c
, lxc_attach_options_t
*options
, const char *program
, const char *arg
, ...)
3603 argv
= lxc_va_arg_list_to_argv_const(ap
, 1);
3607 ERROR("Memory allocation error.");
3612 ret
= lxcapi_attach_run_wait(c
, options
, program
, (const char * const *)argv
);
3617 struct lxc_container
*lxc_container_new(const char *name
, const char *configpath
)
3619 struct lxc_container
*c
;
3624 c
= malloc(sizeof(*c
));
3626 fprintf(stderr
, "failed to malloc lxc_container\n");
3629 memset(c
, 0, sizeof(*c
));
3632 c
->config_path
= strdup(configpath
);
3634 c
->config_path
= strdup(lxc_global_config_value("lxc.lxcpath"));
3636 if (!c
->config_path
) {
3637 fprintf(stderr
, "Out of memory\n");
3641 remove_trailing_slashes(c
->config_path
);
3642 c
->name
= malloc(strlen(name
)+1);
3644 fprintf(stderr
, "Error allocating lxc_container name\n");
3647 strcpy(c
->name
, name
);
3650 if (!(c
->slock
= lxc_newlock(c
->config_path
, name
))) {
3651 fprintf(stderr
, "failed to create lock\n");
3655 if (!(c
->privlock
= lxc_newlock(NULL
, NULL
))) {
3656 fprintf(stderr
, "failed to alloc privlock\n");
3660 if (!set_config_filename(c
)) {
3661 fprintf(stderr
, "Error allocating config file pathname\n");
3665 if (file_exists(c
->configfile
) && !lxcapi_load_config(c
, NULL
))
3668 if (ongoing_create(c
) == 2) {
3669 ERROR("Error: %s creation was not completed", c
->name
);
3670 container_destroy(c
);
3671 lxcapi_clear_config(c
);
3673 c
->daemonize
= true;
3676 // assign the member functions
3677 c
->is_defined
= lxcapi_is_defined
;
3678 c
->state
= lxcapi_state
;
3679 c
->is_running
= lxcapi_is_running
;
3680 c
->freeze
= lxcapi_freeze
;
3681 c
->unfreeze
= lxcapi_unfreeze
;
3682 c
->console
= lxcapi_console
;
3683 c
->console_getfd
= lxcapi_console_getfd
;
3684 c
->init_pid
= lxcapi_init_pid
;
3685 c
->load_config
= lxcapi_load_config
;
3686 c
->want_daemonize
= lxcapi_want_daemonize
;
3687 c
->want_close_all_fds
= lxcapi_want_close_all_fds
;
3688 c
->start
= lxcapi_start
;
3689 c
->startl
= lxcapi_startl
;
3690 c
->stop
= lxcapi_stop
;
3691 c
->config_file_name
= lxcapi_config_file_name
;
3692 c
->wait
= lxcapi_wait
;
3693 c
->set_config_item
= lxcapi_set_config_item
;
3694 c
->destroy
= lxcapi_destroy
;
3695 c
->destroy_with_snapshots
= lxcapi_destroy_with_snapshots
;
3696 c
->rename
= lxcapi_rename
;
3697 c
->save_config
= lxcapi_save_config
;
3698 c
->get_keys
= lxcapi_get_keys
;
3699 c
->create
= lxcapi_create
;
3700 c
->createl
= lxcapi_createl
;
3701 c
->shutdown
= lxcapi_shutdown
;
3702 c
->reboot
= lxcapi_reboot
;
3703 c
->clear_config
= lxcapi_clear_config
;
3704 c
->clear_config_item
= lxcapi_clear_config_item
;
3705 c
->get_config_item
= lxcapi_get_config_item
;
3706 c
->get_running_config_item
= lxcapi_get_running_config_item
;
3707 c
->get_cgroup_item
= lxcapi_get_cgroup_item
;
3708 c
->set_cgroup_item
= lxcapi_set_cgroup_item
;
3709 c
->get_config_path
= lxcapi_get_config_path
;
3710 c
->set_config_path
= lxcapi_set_config_path
;
3711 c
->clone
= lxcapi_clone
;
3712 c
->get_interfaces
= lxcapi_get_interfaces
;
3713 c
->get_ips
= lxcapi_get_ips
;
3714 c
->attach
= lxcapi_attach
;
3715 c
->attach_run_wait
= lxcapi_attach_run_wait
;
3716 c
->attach_run_waitl
= lxcapi_attach_run_waitl
;
3717 c
->snapshot
= lxcapi_snapshot
;
3718 c
->snapshot_list
= lxcapi_snapshot_list
;
3719 c
->snapshot_restore
= lxcapi_snapshot_restore
;
3720 c
->snapshot_destroy
= lxcapi_snapshot_destroy
;
3721 c
->snapshot_destroy_all
= lxcapi_snapshot_destroy_all
;
3722 c
->may_control
= lxcapi_may_control
;
3723 c
->add_device_node
= lxcapi_add_device_node
;
3724 c
->remove_device_node
= lxcapi_remove_device_node
;
3725 c
->attach_interface
= lxcapi_attach_interface
;
3726 c
->detach_interface
= lxcapi_detach_interface
;
3727 c
->checkpoint
= lxcapi_checkpoint
;
3728 c
->restore
= lxcapi_restore
;
3730 /* we'll allow the caller to update these later */
3731 if (lxc_log_init(NULL
, "none", NULL
, "lxc_container", 0, c
->config_path
)) {
3732 fprintf(stderr
, "failed to open log\n");
3739 lxc_container_free(c
);
3743 int lxc_get_wait_states(const char **states
)
3748 for (i
=0; i
<MAX_STATE
; i
++)
3749 states
[i
] = lxc_state2str(i
);
3754 * These next two could probably be done smarter with reusing a common function
3755 * with different iterators and tests...
3757 int list_defined_containers(const char *lxcpath
, char ***names
, struct lxc_container
***cret
)
3760 int i
, cfound
= 0, nfound
= 0;
3761 struct dirent dirent
, *direntp
;
3762 struct lxc_container
*c
;
3765 lxcpath
= lxc_global_config_value("lxc.lxcpath");
3767 dir
= opendir(lxcpath
);
3769 SYSERROR("opendir on lxcpath");
3778 while (!readdir_r(dir
, &dirent
, &direntp
)) {
3781 if (!strcmp(direntp
->d_name
, "."))
3783 if (!strcmp(direntp
->d_name
, ".."))
3786 if (!config_file_exists(lxcpath
, direntp
->d_name
))
3790 if (!add_to_array(names
, direntp
->d_name
, cfound
))
3800 c
= lxc_container_new(direntp
->d_name
, lxcpath
);
3802 INFO("Container %s:%s has a config but could not be loaded",
3803 lxcpath
, direntp
->d_name
);
3805 if(!remove_from_array(names
, direntp
->d_name
, cfound
--))
3809 if (!lxcapi_is_defined(c
)) {
3810 INFO("Container %s:%s has a config but is not defined",
3811 lxcpath
, direntp
->d_name
);
3813 if(!remove_from_array(names
, direntp
->d_name
, cfound
--))
3815 lxc_container_put(c
);
3819 if (!add_to_clist(cret
, c
, nfound
, true)) {
3820 lxc_container_put(c
);
3830 if (names
&& *names
) {
3831 for (i
=0; i
<cfound
; i
++)
3835 if (cret
&& *cret
) {
3836 for (i
=0; i
<nfound
; i
++)
3837 lxc_container_put((*cret
)[i
]);
3844 int list_active_containers(const char *lxcpath
, char ***nret
,
3845 struct lxc_container
***cret
)
3847 int i
, ret
= -1, cret_cnt
= 0, ct_name_cnt
= 0;
3850 char **ct_name
= NULL
;
3852 struct lxc_container
*c
;
3856 lxcpath
= lxc_global_config_value("lxc.lxcpath");
3857 lxcpath_len
= strlen(lxcpath
);
3864 FILE *f
= fopen("/proc/net/unix", "r");
3868 while (getline(&line
, &len
, f
) != -1) {
3870 char *p
= strrchr(line
, ' '), *p2
;
3879 if (strncmp(p
, lxcpath
, lxcpath_len
) == 0) {
3881 } else if (strncmp(p
, "lxc/", 4) == 0) {
3891 // Now p is the start of lxc_name
3892 p2
= strchr(p
, '/');
3893 if (!p2
|| strncmp(p2
, "/command", 8) != 0)
3898 if (strncmp(lxcpath
, lxc_cmd_get_lxcpath(p
), lxcpath_len
) != 0)
3900 p
= lxc_cmd_get_name(p
);
3903 if (array_contains(&ct_name
, p
, ct_name_cnt
))
3906 if (!add_to_array(&ct_name
, p
, ct_name_cnt
))
3907 goto free_cret_list
;
3914 c
= lxc_container_new(p
, lxcpath
);
3916 INFO("Container %s:%s is running but could not be loaded",
3918 remove_from_array(&ct_name
, p
, ct_name_cnt
--);
3923 * If this is an anonymous container, then is_defined *can*
3924 * return false. So we don't do that check. Count on the
3925 * fact that the command socket exists.
3928 if (!add_to_clist(cret
, c
, cret_cnt
, true)) {
3929 lxc_container_put(c
);
3930 goto free_cret_list
;
3935 assert(!nret
|| !cret
|| cret_cnt
== ct_name_cnt
);
3944 if (cret
&& *cret
) {
3945 for (i
= 0; i
< cret_cnt
; i
++)
3946 lxc_container_put((*cret
)[i
]);
3952 for (i
= 0; i
< ct_name_cnt
; i
++)
3964 int list_all_containers(const char *lxcpath
, char ***nret
,
3965 struct lxc_container
***cret
)
3967 int i
, ret
, active_cnt
, ct_cnt
, ct_list_cnt
;
3970 struct lxc_container
**ct_list
= NULL
;
3972 ct_cnt
= list_defined_containers(lxcpath
, &ct_name
, NULL
);
3976 active_cnt
= list_active_containers(lxcpath
, &active_name
, NULL
);
3977 if (active_cnt
< 0) {
3982 for (i
= 0; i
< active_cnt
; i
++) {
3983 if (!array_contains(&ct_name
, active_name
[i
], ct_cnt
)) {
3984 if (!add_to_array(&ct_name
, active_name
[i
], ct_cnt
)) {
3986 goto free_active_name
;
3990 free(active_name
[i
]);
3991 active_name
[i
] = NULL
;
3997 for (i
= 0, ct_list_cnt
= 0; i
< ct_cnt
&& cret
; i
++) {
3998 struct lxc_container
*c
;
4000 c
= lxc_container_new(ct_name
[i
], lxcpath
);
4002 WARN("Container %s:%s could not be loaded", lxcpath
, ct_name
[i
]);
4003 remove_from_array(&ct_name
, ct_name
[i
], ct_cnt
--);
4007 if (!add_to_clist(&ct_list
, c
, ct_list_cnt
, false)) {
4008 lxc_container_put(c
);
4027 for (i
= 0; i
< ct_list_cnt
; i
++) {
4028 lxc_container_put(ct_list
[i
]);
4033 for (i
= 0; i
< active_cnt
; i
++) {
4034 free(active_name
[i
]);
4039 for (i
= 0; i
< ct_cnt
; i
++) {