#include <sys/signalfd.h>
#include <grp.h>
#include <linux/fs.h>
+#include <sys/un.h>
+#include <sys/socket.h>
#include <systemd/sd-daemon.h>
#include "missing.h"
#include "cgroup-util.h"
#include "strv.h"
+#include "path-util.h"
#include "loopback-setup.h"
static char *arg_directory = NULL;
static char *arg_user = NULL;
static char **arg_controllers = NULL;
+static char *arg_uuid = NULL;
static bool arg_private_network = false;
+static bool arg_read_only = false;
+static bool arg_boot = false;
+static uint64_t arg_retain =
+ (1ULL << CAP_CHOWN) |
+ (1ULL << CAP_DAC_OVERRIDE) |
+ (1ULL << CAP_DAC_READ_SEARCH) |
+ (1ULL << CAP_FOWNER) |
+ (1ULL << CAP_FSETID) |
+ (1ULL << CAP_IPC_OWNER) |
+ (1ULL << CAP_KILL) |
+ (1ULL << CAP_LEASE) |
+ (1ULL << CAP_LINUX_IMMUTABLE) |
+ (1ULL << CAP_NET_BIND_SERVICE) |
+ (1ULL << CAP_NET_BROADCAST) |
+ (1ULL << CAP_NET_RAW) |
+ (1ULL << CAP_SETGID) |
+ (1ULL << CAP_SETFCAP) |
+ (1ULL << CAP_SETPCAP) |
+ (1ULL << CAP_SETUID) |
+ (1ULL << CAP_SYS_ADMIN) |
+ (1ULL << CAP_SYS_CHROOT) |
+ (1ULL << CAP_SYS_NICE) |
+ (1ULL << CAP_SYS_PTRACE) |
+ (1ULL << CAP_SYS_TTY_CONFIG) |
+ (1ULL << CAP_SYS_RESOURCE);
static int help(void) {
"Spawn a minimal namespace container for debugging, testing and building.\n\n"
" -h --help Show this help\n"
" -D --directory=NAME Root directory for the container\n"
+ " -b --boot Boot up full system (i.e. invoke init)\n"
" -u --user=USER Run the command under specified user or uid\n"
" -C --controllers=LIST Put the container in specified comma-separated cgroup hierarchies\n"
- " --private-network Disable network in container\n",
+ " --uuid=UUID Set a specific machine UUID for the container\n"
+ " --private-network Disable network in container\n"
+ " --read-only Mount the root directory read-only\n"
+ " --capability=CAP In addition to the default, retain specified capability\n",
program_invocation_short_name);
return 0;
static int parse_argv(int argc, char *argv[]) {
enum {
- ARG_PRIVATE_NETWORK = 0x100
+ ARG_PRIVATE_NETWORK = 0x100,
+ ARG_UUID,
+ ARG_READ_ONLY,
+ ARG_CAPABILITY
};
static const struct option options[] = {
{ "user", required_argument, NULL, 'u' },
{ "controllers", required_argument, NULL, 'C' },
{ "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK },
+ { "boot", no_argument, NULL, 'b' },
+ { "uuid", required_argument, NULL, ARG_UUID },
+ { "read-only", no_argument, NULL, ARG_READ_ONLY },
+ { "capability", required_argument, NULL, ARG_CAPABILITY },
{ NULL, 0, NULL, 0 }
};
assert(argc >= 0);
assert(argv);
- while ((c = getopt_long(argc, argv, "+hD:u:C:", options, NULL)) >= 0) {
+ while ((c = getopt_long(argc, argv, "+hD:u:C:b", options, NULL)) >= 0) {
switch (c) {
case 'D':
free(arg_directory);
- if (!(arg_directory = strdup(optarg))) {
- log_error("Failed to duplicate root directory.");
+ arg_directory = canonicalize_file_name(optarg);
+ if (!arg_directory) {
+ log_error("Failed to canonicalize root directory.");
return -ENOMEM;
}
arg_private_network = true;
break;
+ case 'b':
+ arg_boot = true;
+ break;
+
+ case ARG_UUID:
+ arg_uuid = optarg;
+ break;
+
+ case ARG_READ_ONLY:
+ arg_read_only = true;
+ break;
+
+ case ARG_CAPABILITY: {
+ char *state, *word;
+ size_t length;
+
+ FOREACH_WORD_SEPARATOR(word, length, optarg, ",", state) {
+ cap_value_t cap;
+ char *t;
+
+ t = strndup(word, length);
+ if (!t) {
+ log_error("Out of memory.");
+ return -ENOMEM;
+ }
+
+ if (cap_from_name(t, &cap) < 0) {
+ log_error("Failed to parse capability %s.", t);
+ free(t);
+ return -EINVAL;
+ }
+
+ free(t);
+ arg_retain |= 1ULL << (uint64_t) cap;
+ }
+
+ break;
+ }
+
case '?':
return -EINVAL;
continue;
}
- mkdir_p(where, 0755);
+ mkdir_p_label(where, 0755);
if (mount(mount_table[k].what,
where,
free(where);
}
- /* Fix the timezone, if possible */
- if (asprintf(&where, "%s/etc/localtime", dest) >= 0) {
+ return r;
+}
+
+static int setup_timezone(const char *dest) {
+ char *where;
- if (mount("/etc/localtime", where, "bind", MS_BIND, NULL) >= 0)
- mount("/etc/localtime", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
+ assert(dest);
- free(where);
+ /* Fix the timezone, if possible */
+ if (asprintf(&where, "%s/etc/localtime", dest) < 0) {
+ log_error("Out of memory");
+ return -ENOMEM;
}
- if (asprintf(&where, "%s/etc/timezone", dest) >= 0) {
+ if (mount("/etc/localtime", where, "bind", MS_BIND, NULL) >= 0)
+ mount("/etc/localtime", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
- if (mount("/etc/timezone", where, "bind", MS_BIND, NULL) >= 0)
- mount("/etc/timezone", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
+ free(where);
- free(where);
+ if (asprintf(&where, "%s/etc/timezone", dest) < 0) {
+ log_error("Out of memory");
+ return -ENOMEM;
}
- if (asprintf(&where, "%s/proc/kmsg", dest) >= 0) {
- mount("/dev/null", where, "bind", MS_BIND, NULL);
- free(where);
+ if (mount("/etc/timezone", where, "bind", MS_BIND, NULL) >= 0)
+ mount("/etc/timezone", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
+
+ free(where);
+
+ return 0;
+}
+
+static int setup_resolv_conf(const char *dest) {
+ char *where;
+
+ assert(dest);
+
+ if (arg_private_network)
+ return 0;
+
+ /* Fix resolv.conf, if possible */
+ if (asprintf(&where, "%s/etc/resolv.conf", dest) < 0) {
+ log_error("Out of memory");
+ return -ENOMEM;
}
- return r;
+ if (mount("/etc/resolv.conf", where, "bind", MS_BIND, NULL) >= 0)
+ mount("/etc/resolv.conf", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
+
+ free(where);
+
+ return 0;
}
-static int copy_devnodes(const char *dest, const char *console) {
+static int copy_devnodes(const char *dest) {
static const char devnodes[] =
"null\0"
"urandom\0"
"tty\0"
"ptmx\0"
- "kmsg\0"
"rtc0\0";
const char *d;
- int r = 0, k;
+ int r = 0;
mode_t u;
- struct stat st;
- char *from = NULL, *to = NULL;
assert(dest);
- assert(console);
u = umask(0000);
NULSTR_FOREACH(d, devnodes) {
- from = to = NULL;
+ struct stat st;
+ char *from = NULL, *to = NULL;
asprintf(&from, "/dev/%s", d);
asprintf(&to, "%s/dev/%s", dest, d);
free(to);
}
- if (stat(console, &st) < 0) {
+ umask(u);
- log_error("Failed to stat %s: %m", console);
- if (r == 0)
- r = -errno;
+ return r;
+}
+
+static int setup_dev_console(const char *dest, const char *console) {
+ struct stat st;
+ char *to = NULL;
+ int r;
+ mode_t u;
+ assert(dest);
+ assert(console);
+
+ u = umask(0000);
+
+ if (stat(console, &st) < 0) {
+ log_error("Failed to stat %s: %m", console);
+ r = -errno;
goto finish;
} else if (!S_ISCHR(st.st_mode)) {
-
log_error("/dev/console is not a char device.");
- if (r == 0)
- r = -EIO;
+ r = -EIO;
+ goto finish;
+ }
+ r = chmod_and_chown(console, 0600, 0, 0);
+ if (r < 0) {
+ log_error("Failed to correct access mode for TTY: %s", strerror(-r));
goto finish;
}
if (asprintf(&to, "%s/dev/console", dest) < 0) {
-
log_error("Out of memory");
- if (r == 0)
- r = -ENOMEM;
-
- goto finish;
+ r = -ENOMEM;
+ goto finish;
}
/* We need to bind mount the right tty to /dev/console since
* doesn't actually matter here, since we mount it over
* anyway). */
- if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0)
- log_error("mknod for /dev/console failed: %m");
+ if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) {
+ log_error("mknod() for /dev/console failed: %m");
+ r = -errno;
+ goto finish;
+ }
if (mount(console, to, "bind", MS_BIND, NULL) < 0) {
- log_error("bind mount for /dev/console failed: %m");
-
- if (r == 0)
- r = -errno;
+ log_error("Bind mount for /dev/console failed: %m");
+ r = -errno;
+ goto finish;
}
+finish:
free(to);
+ umask(u);
+
+ return r;
+}
+
+static int setup_kmsg(const char *dest, int kmsg_socket) {
+ char *from = NULL, *to = NULL;
+ int r, fd, k;
+ mode_t u;
+ union {
+ struct cmsghdr cmsghdr;
+ uint8_t buf[CMSG_SPACE(sizeof(int))];
+ } control;
+ struct msghdr mh;
+ struct cmsghdr *cmsg;
+
+ assert(dest);
+ assert(kmsg_socket >= 0);
+
+ u = umask(0000);
+
+ /* We create the kmsg FIFO as /dev/kmsg, but immediately
+ * delete it after bind mounting it to /proc/kmsg. While FIFOs
+ * on the reading side behave very similar to /proc/kmsg,
+ * their writing side behaves differently from /dev/kmsg in
+ * that writing blocks when nothing is reading. In order to
+ * avoid any problems with containers deadlocking due to this
+ * we simply make /dev/kmsg unavailable to the container. */
+ if (asprintf(&from, "%s/dev/kmsg", dest) < 0) {
+ log_error("Out of memory");
+ r = -ENOMEM;
+ goto finish;
+ }
+
+ if (asprintf(&to, "%s/proc/kmsg", dest) < 0) {
+ log_error("Out of memory");
+ r = -ENOMEM;
+ goto finish;
+ }
+
+ if (mkfifo(from, 0600) < 0) {
+ log_error("mkfifo() for /dev/kmsg failed: %m");
+ r = -errno;
+ goto finish;
+ }
- if ((k = chmod_and_chown(console, 0600, 0, 0)) < 0) {
- log_error("Failed to correct access mode for TTY: %s", strerror(-k));
+ r = chmod_and_chown(from, 0600, 0, 0);
+ if (r < 0) {
+ log_error("Failed to correct access mode for /dev/kmsg: %s", strerror(-r));
+ goto finish;
+ }
- if (r == 0)
- r = k;
+ if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
+ log_error("Bind mount for /proc/kmsg failed: %m");
+ r = -errno;
+ goto finish;
}
+ fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC);
+ if (fd < 0) {
+ log_error("Failed to open fifo: %m");
+ r = -errno;
+ goto finish;
+ }
+
+ zero(mh);
+ zero(control);
+
+ mh.msg_control = &control;
+ mh.msg_controllen = sizeof(control);
+
+ cmsg = CMSG_FIRSTHDR(&mh);
+ cmsg->cmsg_level = SOL_SOCKET;
+ cmsg->cmsg_type = SCM_RIGHTS;
+ cmsg->cmsg_len = CMSG_LEN(sizeof(int));
+ memcpy(CMSG_DATA(cmsg), &fd, sizeof(int));
+
+ mh.msg_controllen = cmsg->cmsg_len;
+
+ /* Store away the fd in the socket, so that it stays open as
+ * long as we run the child */
+ k = sendmsg(kmsg_socket, &mh, MSG_DONTWAIT|MSG_NOSIGNAL);
+ close_nointr_nofail(fd);
+
+ if (k < 0) {
+ log_error("Failed to send FIFO fd: %m");
+ r = -errno;
+ goto finish;
+ }
+
+ /* And now make the FIFO unavailable as /dev/kmsg... */
+ unlink(from);
+
finish:
+ free(from);
+ free(to);
umask(u);
return r;
}
-static int drop_capabilities(void) {
- static const unsigned long retain[] = {
- CAP_CHOWN,
- CAP_DAC_OVERRIDE,
- CAP_DAC_READ_SEARCH,
- CAP_FOWNER,
- CAP_FSETID,
- CAP_IPC_OWNER,
- CAP_KILL,
- CAP_LEASE,
- CAP_LINUX_IMMUTABLE,
- CAP_NET_BIND_SERVICE,
- CAP_NET_BROADCAST,
- CAP_NET_RAW,
- CAP_SETGID,
- CAP_SETFCAP,
- CAP_SETPCAP,
- CAP_SETUID,
- CAP_SYS_ADMIN,
- CAP_SYS_CHROOT,
- CAP_SYS_NICE,
- CAP_SYS_PTRACE,
- CAP_SYS_TTY_CONFIG
- };
-
- unsigned long l;
+static int setup_hostname(void) {
+ char *hn;
+ int r = 0;
- for (l = 0; l <= cap_last_cap(); l++) {
- unsigned i;
+ hn = path_get_file_name(arg_directory);
+ if (hn) {
+ hn = strdup(hn);
+ if (!hn)
+ return -ENOMEM;
- for (i = 0; i < ELEMENTSOF(retain); i++)
- if (retain[i] == l)
- break;
+ hostname_cleanup(hn);
- if (i < ELEMENTSOF(retain))
- continue;
+ if (!isempty(hn))
+ if (sethostname(hn, strlen(hn)) < 0)
+ r = -errno;
- if (prctl(PR_CAPBSET_DROP, l) < 0) {
- log_error("PR_CAPBSET_DROP failed: %m");
- return -errno;
- }
+ free(hn);
}
- return 0;
+ return r;
+}
+
+static int drop_capabilities(void) {
+ return capability_bounding_set_drop(~arg_retain, false);
}
static int is_os_tree(const char *path) {
sigset_t mask;
bool saved_attr_valid = false;
struct winsize ws;
+ int kmsg_socket_pair[2] = { -1, -1 };
log_parse_environment();
log_open();
goto finish;
}
+ if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) {
+ log_error("Failed to create kmsg socket pair");
+ goto finish;
+ }
+
assert_se(sigemptyset(&mask) == 0);
sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1);
assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0);
if (pid == 0) {
/* child */
- const char *hn;
const char *home = NULL;
uid_t uid = (uid_t) -1;
gid_t gid = (gid_t) -1;
NULL, /* HOME */
NULL, /* USER */
NULL, /* LOGNAME */
+ NULL, /* container_uuid */
NULL
};
close_nointr(STDOUT_FILENO);
close_nointr(STDERR_FILENO);
- close_all_fds(NULL, 0);
+ close_all_fds(&kmsg_socket_pair[1], 1);
reset_all_signal_handlers();
if (mount(NULL, "/", NULL, MS_PRIVATE|MS_REC, NULL) < 0)
goto child_fail;
+ /* Turn directory into bind mount */
+ if (mount(arg_directory, arg_directory, "bind", MS_BIND, NULL) < 0) {
+ log_error("Failed to make bind mount.");
+ goto child_fail;
+ }
+
+ if (arg_read_only)
+ if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL) < 0) {
+ log_error("Failed to make read-only.");
+ goto child_fail;
+ }
+
if (mount_all(arg_directory) < 0)
goto child_fail;
- if (copy_devnodes(arg_directory, console) < 0)
+ if (copy_devnodes(arg_directory) < 0)
+ goto child_fail;
+
+ if (setup_dev_console(arg_directory, console) < 0)
+ goto child_fail;
+
+ if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0)
+ goto child_fail;
+
+ close_nointr_nofail(kmsg_socket_pair[1]);
+
+ if (setup_timezone(arg_directory) < 0)
+ goto child_fail;
+
+ if (setup_resolv_conf(arg_directory) < 0)
goto child_fail;
if (chdir(arg_directory) < 0) {
dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO)
goto child_fail;
- if (mount(arg_directory, "/", "bind", MS_BIND|MS_MOVE, NULL) < 0) {
- log_error("mount(MS_MOVE) failed: %m");
+ if (mount(arg_directory, "/", "bind", MS_MOVE, NULL) < 0) {
+ log_error("mount(MS_BIND) failed: %m");
goto child_fail;
}
loopback_setup();
- if (drop_capabilities() < 0)
+ if (drop_capabilities() < 0) {
+ log_error("drop_capabilities() failed: %m");
goto child_fail;
+ }
if (arg_user) {
goto child_fail;
}
- if (mkdir_parents(home, 0775) < 0) {
- log_error("mkdir_parents() failed: %m");
+ if (mkdir_parents_label(home, 0775) < 0) {
+ log_error("mkdir_parents_label() failed: %m");
goto child_fail;
}
- if (safe_mkdir(home, 0775, uid, gid) < 0) {
- log_error("safe_mkdir() failed: %m");
+ if (mkdir_safe_label(home, 0775, uid, gid) < 0) {
+ log_error("mkdir_safe_label() failed: %m");
goto child_fail;
}
}
}
- if ((asprintf((char**)(envp + 3), "HOME=%s", home? home: "/root") < 0) ||
- (asprintf((char**)(envp + 4), "USER=%s", arg_user? arg_user : "root") < 0) ||
- (asprintf((char**)(envp + 5), "LOGNAME=%s", arg_user? arg_user : "root") < 0)) {
+ if ((asprintf((char**)(envp + 3), "HOME=%s", home ? home: "/root") < 0) ||
+ (asprintf((char**)(envp + 4), "USER=%s", arg_user ? arg_user : "root") < 0) ||
+ (asprintf((char**)(envp + 5), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) {
log_error("Out of memory");
goto child_fail;
}
- if ((hn = file_name_from_path(arg_directory)))
- sethostname(hn, strlen(hn));
+ if (arg_uuid) {
+ if (asprintf((char**)(envp + 6), "container_uuid=%s", arg_uuid) < 0) {
+ log_error("Out of memory");
+ goto child_fail;
+ }
+ }
+
+ setup_hostname();
+
+ if (arg_boot) {
+ char **a;
+ size_t l;
- if (argc > optind)
+ /* Automatically search for the init system */
+
+ l = 1 + argc - optind;
+ a = newa(char*, l + 1);
+ memcpy(a + 1, argv + optind, l * sizeof(char*));
+
+ a[0] = (char*) "/usr/lib/systemd/systemd";
+ execve(a[0], a, (char**) envp);
+
+ a[0] = (char*) "/lib/systemd/systemd";
+ execve(a[0], a, (char**) envp);
+
+ a[0] = (char*) "/sbin/init";
+ execve(a[0], a, (char**) envp);
+ } else if (argc > optind)
execvpe(argv[optind], argv + optind, (char**) envp);
else {
chdir(home ? home : "/root");
if (master >= 0)
close_nointr_nofail(master);
+ close_pipe(kmsg_socket_pair);
+
if (oldcg)
cg_attach(SYSTEMD_CGROUP_CONTROLLER, oldcg, 0);