#include <sys/prctl.h>
#include <sys/capability.h>
#include <getopt.h>
+#include <sys/poll.h>
#include <sys/epoll.h>
#include <termios.h>
#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 "log.h"
#include "util.h"
#include "mkdir.h"
+#include "macro.h"
#include "audit.h"
#include "missing.h"
#include "cgroup-util.h"
#include "strv.h"
+#include "path-util.h"
#include "loopback-setup.h"
+#include "sd-id128.h"
+#include "dev-setup.h"
+#include "fdset.h"
+#include "build.h"
+#include "fileio.h"
+
+#ifndef TTY_GID
+#define TTY_GID 5
+#endif
+
+typedef enum LinkJournal {
+ LINK_NO,
+ LINK_AUTO,
+ LINK_HOST,
+ LINK_GUEST
+} LinkJournal;
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 LinkJournal arg_link_journal = LINK_AUTO;
+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) |
+ (1ULL << CAP_SYS_BOOT) |
+ (1ULL << CAP_AUDIT_WRITE) |
+ (1ULL << CAP_AUDIT_CONTROL);
+static char **arg_bind = NULL;
+static char **arg_bind_ro = NULL;
static int help(void) {
printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n"
"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"
- " -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",
+ " -h --help Show this help\n"
+ " --version Print version string\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\n"
+ " cgroup hierarchies\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\n"
+ " capability\n"
+ " --link-journal=MODE Link up guest journal, one of no, auto, guest, host\n"
+ " -j Equivalent to --link-journal=host\n"
+ " --bind=PATH[:PATH] Bind mount a file or directory from the host into\n"
+ " the container\n"
+ " --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n",
program_invocation_short_name);
return 0;
static int parse_argv(int argc, char *argv[]) {
enum {
- ARG_PRIVATE_NETWORK = 0x100
+ ARG_VERSION = 0x100,
+ ARG_PRIVATE_NETWORK,
+ ARG_UUID,
+ ARG_READ_ONLY,
+ ARG_CAPABILITY,
+ ARG_LINK_JOURNAL,
+ ARG_BIND,
+ ARG_BIND_RO
};
static const struct option options[] = {
{ "help", no_argument, NULL, 'h' },
+ { "version", no_argument, NULL, ARG_VERSION },
{ "directory", required_argument, NULL, 'D' },
{ "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 },
+ { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL },
+ { "bind", required_argument, NULL, ARG_BIND },
+ { "bind-ro", required_argument, NULL, ARG_BIND_RO },
{ 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:bj", options, NULL)) >= 0) {
switch (c) {
help();
return 0;
+ case ARG_VERSION:
+ puts(PACKAGE_STRING);
+ puts(SYSTEMD_FEATURES);
+ return 0;
+
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)
+ return log_oom();
+
+ 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 'j':
+ arg_link_journal = LINK_GUEST;
+ break;
+
+ case ARG_LINK_JOURNAL:
+ if (streq(optarg, "auto"))
+ arg_link_journal = LINK_AUTO;
+ else if (streq(optarg, "no"))
+ arg_link_journal = LINK_NO;
+ else if (streq(optarg, "guest"))
+ arg_link_journal = LINK_GUEST;
+ else if (streq(optarg, "host"))
+ arg_link_journal = LINK_HOST;
+ else {
+ log_error("Failed to parse link journal mode %s", optarg);
+ return -EINVAL;
+ }
+
+ break;
+
+ case ARG_BIND:
+ case ARG_BIND_RO: {
+ _cleanup_free_ char *a = NULL, *b = NULL;
+ char *e;
+ char ***x;
+ int r;
+
+ x = c == ARG_BIND ? &arg_bind : &arg_bind_ro;
+
+ e = strchr(optarg, ':');
+ if (e) {
+ a = strndup(optarg, e - optarg);
+ b = strdup(e + 1);
+ } else {
+ a = strdup(optarg);
+ b = strdup(optarg);
+ }
+
+ if (!a || !b)
+ return log_oom();
+
+ if (!path_is_absolute(a) || !path_is_absolute(b)) {
+ log_error("Invalid bind mount specification: %s", optarg);
+ return -EINVAL;
+ }
+
+ r = strv_extend(x, a);
+ if (r < 0)
+ return r;
+
+ r = strv_extend(x, b);
+ if (r < 0)
+ return r;
+
+ break;
+ }
+
case '?':
return -EINVAL;
static const MountPoint mount_table[] = {
{ "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
- { "/proc/sys", "/proc/sys", "bind", NULL, MS_BIND, true }, /* Bind mount first */
- { "/proc/sys", "/proc/sys", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */
- { "/sys", "/sys", "bind", NULL, MS_BIND, true }, /* Bind mount first */
- { "/sys", "/sys", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */
+ { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true }, /* Bind mount first */
+ { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */
+ { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
{ "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true },
- { "/dev/pts", "/dev/pts", "bind", NULL, MS_BIND, true },
+ { "devpts", "/dev/pts", "devpts","newinstance,ptmxmode=0666,mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, true },
+ { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
{ "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
#ifdef HAVE_SELINUX
- { "/sys/fs/selinux", "/sys/fs/selinux", "bind", NULL, MS_BIND, false }, /* Bind mount first */
- { "/sys/fs/selinux", "/sys/fs/selinux", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */
+ { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false }, /* Bind mount first */
+ { NULL, "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */
#endif
};
unsigned k;
int r = 0;
- char *where;
for (k = 0; k < ELEMENTSOF(mount_table); k++) {
+ char _cleanup_free_ *where = NULL;
int t;
- if (asprintf(&where, "%s/%s", dest, mount_table[k].where) < 0) {
- log_error("Out of memory");
+ where = strjoin(dest, "/", mount_table[k].where, NULL);
+ if (!where)
+ return log_oom();
- if (r == 0)
- r = -ENOMEM;
-
- break;
- }
-
- t = path_is_mount_point(where, false);
+ t = path_is_mount_point(where, true);
if (t < 0) {
log_error("Failed to detect whether %s is a mount point: %s", where, strerror(-t));
- free(where);
if (r == 0)
r = t;
continue;
}
+ /* Skip this entry if it is not a remount. */
+ if (mount_table[k].what && t > 0)
+ continue;
+
mkdir_p(where, 0755);
if (mount(mount_table[k].what,
if (r == 0)
r = -errno;
}
+ }
+
+ return r;
+}
+
+static int mount_binds(const char *dest, char **l, unsigned long flags) {
+ char **x, **y;
+
+ STRV_FOREACH_PAIR(x, y, l) {
+ _cleanup_free_ char *where = NULL;
+
+ where = strjoin(dest, "/", *y, NULL);
+ if (!where)
+ return log_oom();
+
+ mkdir_p_label(where, 0755);
- free(where);
+ if (mount(*x, where, "bind", MS_BIND, NULL) < 0) {
+ log_error("mount(%s) failed: %m", where);
+ return -errno;
+ }
+
+ if (flags && mount(NULL, where, NULL, MS_REMOUNT|MS_BIND|flags, NULL) < 0) {
+ log_error("mount(%s) failed: %m", where);
+ return -errno;
+ }
}
+ return 0;
+}
+
+static int setup_timezone(const char *dest) {
+ _cleanup_free_ char *where = NULL, *p = NULL, *q = NULL, *check = NULL, *what = NULL;
+ char *z, *y;
+ int r;
+
+ assert(dest);
+
/* Fix the timezone, if possible */
- if (asprintf(&where, "%s/etc/localtime", dest) >= 0) {
+ r = readlink_malloc("/etc/localtime", &p);
+ if (r < 0) {
+ log_warning("/etc/localtime is not a symlink, not updating container timezone.");
+ return 0;
+ }
+
+ z = path_startswith(p, "../usr/share/zoneinfo/");
+ if (!z)
+ z = path_startswith(p, "/usr/share/zoneinfo/");
+ if (!z) {
+ log_warning("/etc/localtime does not point into /usr/share/zoneinfo/, not updating container timezone.");
+ return 0;
+ }
+
+ where = strappend(dest, "/etc/localtime");
+ if (!where)
+ return log_oom();
+
+ r = readlink_malloc(where, &q);
+ if (r >= 0) {
+ y = path_startswith(q, "../usr/share/zoneinfo/");
+ if (!y)
+ y = path_startswith(q, "/usr/share/zoneinfo/");
+
- if (mount("/etc/localtime", where, "bind", MS_BIND, NULL) >= 0)
- mount("/etc/localtime", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
+ /* Already pointing to the right place? Then do nothing .. */
+ if (y && streq(y, z))
+ return 0;
+ }
+
+ check = strjoin(dest, "/usr/share/zoneinfo/", z, NULL);
+ if (!check)
+ return log_oom();
+
+ if (access(check, F_OK) < 0) {
+ log_warning("Timezone %s does not exist in container, not updating container timezone.", z);
+ return 0;
+ }
- free(where);
+ what = strappend("../usr/share/zoneinfo/", z);
+ if (!what)
+ return log_oom();
+
+ unlink(where);
+ if (symlink(what, where) < 0) {
+ log_error("Failed to correct timezone of container: %m");
+ return 0;
}
- if (asprintf(&where, "%s/etc/timezone", dest) >= 0) {
+ return 0;
+}
+
+static int setup_resolv_conf(const char *dest) {
+ char *where;
+
+ assert(dest);
+
+ if (arg_private_network)
+ return 0;
- if (mount("/etc/timezone", where, "bind", MS_BIND, NULL) >= 0)
- mount("/etc/timezone", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
+ /* Fix resolv.conf, if possible */
+ where = strappend(dest, "/etc/resolv.conf");
+ if (!where)
+ return log_oom();
- free(where);
+ /* We don't really care for the results of this really. If it
+ * fails, it fails, but meh... */
+ 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 setup_boot_id(const char *dest) {
+ char _cleanup_free_ *from = NULL, *to = NULL;
+ sd_id128_t rnd;
+ char as_uuid[37];
+ int r;
+
+ assert(dest);
+
+ /* Generate a new randomized boot ID, so that each boot-up of
+ * the container gets a new one */
+
+ from = strappend(dest, "/dev/proc-sys-kernel-random-boot-id");
+ to = strappend(dest, "/proc/sys/kernel/random/boot_id");
+ if (!from || !to)
+ return log_oom();
+
+ r = sd_id128_randomize(&rnd);
+ if (r < 0) {
+ log_error("Failed to generate random boot id: %s", strerror(-r));
+ return r;
}
- if (asprintf(&where, "%s/proc/kmsg", dest) >= 0) {
- mount("/dev/null", where, "bind", MS_BIND, NULL);
- free(where);
+ snprintf(as_uuid, sizeof(as_uuid),
+ "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x",
+ SD_ID128_FORMAT_VAL(rnd));
+ char_array_0(as_uuid);
+
+ r = write_string_file(from, as_uuid);
+ if (r < 0) {
+ log_error("Failed to write boot id: %s", strerror(-r));
+ return r;
}
+ if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
+ log_error("Failed to bind mount boot id: %m");
+ r = -errno;
+ } else if (mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL))
+ log_warning("Failed to make boot id read-only: %m");
+
+ unlink(from);
return r;
}
-static int copy_devnodes(const char *dest, const char *console) {
+static int copy_devnodes(const char *dest) {
static const char devnodes[] =
"null\0"
"full\0"
"random\0"
"urandom\0"
- "tty\0"
- "ptmx\0"
- "kmsg\0"
- "rtc0\0";
+ "tty\0";
const char *d;
- int r = 0, k;
- mode_t u;
- struct stat st;
- char *from = NULL, *to = NULL;
+ int r = 0;
+ mode_t _cleanup_umask_ u;
assert(dest);
- assert(console);
u = umask(0000);
NULSTR_FOREACH(d, devnodes) {
- from = to = NULL;
+ struct stat st;
+ char _cleanup_free_ *from = NULL, *to = NULL;
asprintf(&from, "/dev/%s", d);
asprintf(&to, "%s/dev/%s", dest, d);
if (!from || !to) {
- log_error("Failed to allocate devnode path");
-
- free(from);
- free(to);
-
- from = to = NULL;
+ log_oom();
if (r == 0)
r = -ENOMEM;
} else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) {
- log_error("%s is not a char or block device, cannot copy.", from);
+ log_error("%s is not a char or block device, cannot copy", from);
if (r == 0)
r = -EIO;
if (r == 0)
r = -errno;
}
-
- free(from);
- free(to);
}
- if (stat(console, &st) < 0) {
+ return r;
+}
- log_error("Failed to stat %s: %m", console);
- if (r == 0)
- r = -errno;
+static int setup_ptmx(const char *dest) {
+ _cleanup_free_ char *p = NULL;
- goto finish;
+ p = strappend(dest, "/dev/ptmx");
+ if (!p)
+ return log_oom();
- } else if (!S_ISCHR(st.st_mode)) {
+ if (symlink("pts/ptmx", p) < 0) {
+ log_error("Failed to create /dev/ptmx symlink: %m");
+ return -errno;
+ }
- log_error("/dev/console is not a char device.");
- if (r == 0)
- r = -EIO;
+ return 0;
+}
- goto finish;
- }
+static int setup_dev_console(const char *dest, const char *console) {
+ struct stat st;
+ char _cleanup_free_ *to = NULL;
+ int r;
+ mode_t _cleanup_umask_ u;
- if (asprintf(&to, "%s/dev/console", dest) < 0) {
+ assert(dest);
+ assert(console);
- log_error("Out of memory");
- if (r == 0)
- r = -ENOMEM;
+ u = umask(0000);
- goto finish;
+ if (stat(console, &st) < 0) {
+ log_error("Failed to stat %s: %m", console);
+ return -errno;
+
+ } else if (!S_ISCHR(st.st_mode)) {
+ log_error("/dev/console is not a char device");
+ return -EIO;
+ }
+
+ r = chmod_and_chown(console, 0600, 0, 0);
+ if (r < 0) {
+ log_error("Failed to correct access mode for TTY: %s", strerror(-r));
+ return r;
}
+ if (asprintf(&to, "%s/dev/console", dest) < 0)
+ return log_oom();
+
/* We need to bind mount the right tty to /dev/console since
* ptys can only exist on pts file systems. To have something
* to bind mount things on we create a device node first, that
* 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");
+ return -errno;
+ }
if (mount(console, to, "bind", MS_BIND, NULL) < 0) {
- log_error("bind mount for /dev/console failed: %m");
+ log_error("Bind mount for /dev/console failed: %m");
+ return -errno;
+ }
- if (r == 0)
- r = -errno;
+ return 0;
+}
+
+static int setup_kmsg(const char *dest, int kmsg_socket) {
+ char _cleanup_free_ *from = NULL, *to = NULL;
+ int r, fd, k;
+ mode_t _cleanup_umask_ 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 ||
+ asprintf(&to, "%s/proc/kmsg", dest) < 0)
+ return log_oom();
+
+ if (mkfifo(from, 0600) < 0) {
+ log_error("mkfifo() for /dev/kmsg failed: %m");
+ return -errno;
}
- free(to);
+ r = chmod_and_chown(from, 0600, 0, 0);
+ if (r < 0) {
+ log_error("Failed to correct access mode for /dev/kmsg: %s", strerror(-r));
+ return r;
+ }
- if ((k = chmod_and_chown(console, 0600, 0, 0)) < 0) {
- log_error("Failed to correct access mode for TTY: %s", strerror(-k));
+ if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
+ log_error("Bind mount for /proc/kmsg failed: %m");
+ return -errno;
+ }
- if (r == 0)
- r = k;
+ fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC);
+ if (fd < 0) {
+ log_error("Failed to open fifo: %m");
+ return -errno;
}
-finish:
- umask(u);
+ 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");
+ return -errno;
+ }
+
+ /* And now make the FIFO unavailable as /dev/kmsg... */
+ unlink(from);
+ return 0;
+}
+
+static int setup_hostname(void) {
+ char *hn;
+ int r = 0;
+
+ hn = path_get_file_name(arg_directory);
+ if (hn) {
+ hn = strdup(hn);
+ if (!hn)
+ return -ENOMEM;
+
+ hostname_cleanup(hn);
+
+ if (!isempty(hn))
+ if (sethostname(hn, strlen(hn)) < 0)
+ r = -errno;
+
+ free(hn);
+ }
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
- };
+static int setup_journal(const char *directory) {
+ sd_id128_t machine_id;
+ char _cleanup_free_ *p = NULL, *b = NULL, *q = NULL, *d = NULL;
+ char *id;
+ int r;
- unsigned long l;
+ if (arg_link_journal == LINK_NO)
+ return 0;
- for (l = 0; l <= cap_last_cap(); l++) {
- unsigned i;
+ p = strappend(directory, "/etc/machine-id");
+ if (!p)
+ return log_oom();
- for (i = 0; i < ELEMENTSOF(retain); i++)
- if (retain[i] == l)
- break;
+ r = read_one_line_file(p, &b);
+ if (r == -ENOENT && arg_link_journal == LINK_AUTO)
+ return 0;
+ else if (r < 0) {
+ log_error("Failed to read machine ID from %s: %s", p, strerror(-r));
+ return r;
+ }
- if (i < ELEMENTSOF(retain))
- continue;
+ id = strstrip(b);
+ if (isempty(id) && arg_link_journal == LINK_AUTO)
+ return 0;
+
+ /* Verify validity */
+ r = sd_id128_from_string(id, &machine_id);
+ if (r < 0) {
+ log_error("Failed to parse machine ID from %s: %s", p, strerror(-r));
+ return r;
+ }
+
+ free(p);
+ p = strappend("/var/log/journal/", id);
+ q = strjoin(directory, "/var/log/journal/", id, NULL);
+ if (!p || !q)
+ return log_oom();
+
+ if (path_is_mount_point(p, false) > 0) {
+ if (arg_link_journal != LINK_AUTO) {
+ log_error("%s: already a mount point, refusing to use for journal", p);
+ return -EEXIST;
+ }
+
+ return 0;
+ }
+
+ if (path_is_mount_point(q, false) > 0) {
+ if (arg_link_journal != LINK_AUTO) {
+ log_error("%s: already a mount point, refusing to use for journal", q);
+ return -EEXIST;
+ }
+
+ return 0;
+ }
+
+ r = readlink_and_make_absolute(p, &d);
+ if (r >= 0) {
+ if ((arg_link_journal == LINK_GUEST ||
+ arg_link_journal == LINK_AUTO) &&
+ path_equal(d, q)) {
+
+ r = mkdir_p(q, 0755);
+ if (r < 0)
+ log_warning("failed to create directory %s: %m", q);
+ return 0;
+ }
+
+ if (unlink(p) < 0) {
+ log_error("Failed to remove symlink %s: %m", p);
+ return -errno;
+ }
+ } else if (r == -EINVAL) {
+
+ if (arg_link_journal == LINK_GUEST &&
+ rmdir(p) < 0) {
+
+ if (errno == ENOTDIR) {
+ log_error("%s already exists and is neither a symlink nor a directory", p);
+ return r;
+ } else {
+ log_error("Failed to remove %s: %m", p);
+ return -errno;
+ }
+ }
+ } else if (r != -ENOENT) {
+ log_error("readlink(%s) failed: %m", p);
+ return r;
+ }
- if (prctl(PR_CAPBSET_DROP, l) < 0) {
- log_error("PR_CAPBSET_DROP failed: %m");
+ if (arg_link_journal == LINK_GUEST) {
+
+ if (symlink(q, p) < 0) {
+ log_error("Failed to symlink %s to %s: %m", q, p);
return -errno;
}
+
+ r = mkdir_p(q, 0755);
+ if (r < 0)
+ log_warning("failed to create directory %s: %m", q);
+ return 0;
+ }
+
+ if (arg_link_journal == LINK_HOST) {
+ r = mkdir_p(p, 0755);
+ if (r < 0) {
+ log_error("Failed to create %s: %m", p);
+ return r;
+ }
+
+ } else if (access(p, F_OK) < 0)
+ return 0;
+
+ if (dir_is_empty(q) == 0) {
+ log_error("%s not empty.", q);
+ return -ENOTEMPTY;
+ }
+
+ r = mkdir_p(q, 0755);
+ if (r < 0) {
+ log_error("Failed to create %s: %m", q);
+ return r;
+ }
+
+ if (mount(p, q, "bind", MS_BIND, NULL) < 0) {
+ log_error("Failed to bind mount journal from host into guest: %m");
+ return -errno;
}
return 0;
}
+static int drop_capabilities(void) {
+ return capability_bounding_set_drop(~arg_retain, false);
+}
+
static int is_os_tree(const char *path) {
int r;
char *p;
return r < 0 ? 0 : 1;
}
-static int process_pty(int master, sigset_t *mask) {
+static int process_pty(int master, pid_t pid, sigset_t *mask) {
char in_buffer[LINE_MAX], out_buffer[LINE_MAX];
size_t in_buffer_full = 0, out_buffer_full = 0;
struct epoll_event stdin_ev, stdout_ev, master_ev, signal_ev;
bool stdin_readable = false, stdout_writable = false, master_readable = false, master_writable = false;
int ep = -1, signal_fd = -1, r;
+ bool tried_orderly_shutdown = false;
+
+ assert(master >= 0);
+ assert(pid > 0);
+ assert(mask);
fd_nonblock(STDIN_FILENO, 1);
fd_nonblock(STDOUT_FILENO, 1);
fd_nonblock(master, 1);
- if ((signal_fd = signalfd(-1, mask, SFD_NONBLOCK|SFD_CLOEXEC)) < 0) {
+ signal_fd = signalfd(-1, mask, SFD_NONBLOCK|SFD_CLOEXEC);
+ if (signal_fd < 0) {
log_error("signalfd(): %m");
r = -errno;
goto finish;
}
- if ((ep = epoll_create1(EPOLL_CLOEXEC)) < 0) {
+ ep = epoll_create1(EPOLL_CLOEXEC);
+ if (ep < 0) {
log_error("Failed to create epoll: %m");
r = -errno;
goto finish;
}
- zero(stdin_ev);
- stdin_ev.events = EPOLLIN|EPOLLET;
- stdin_ev.data.fd = STDIN_FILENO;
+ /* We read from STDIN only if this is actually a TTY,
+ * otherwise we assume non-interactivity. */
+ if (isatty(STDIN_FILENO)) {
+ zero(stdin_ev);
+ stdin_ev.events = EPOLLIN|EPOLLET;
+ stdin_ev.data.fd = STDIN_FILENO;
+
+ if (epoll_ctl(ep, EPOLL_CTL_ADD, STDIN_FILENO, &stdin_ev) < 0) {
+ log_error("Failed to register STDIN in epoll: %m");
+ r = -errno;
+ goto finish;
+ }
+ }
zero(stdout_ev);
stdout_ev.events = EPOLLOUT|EPOLLET;
signal_ev.events = EPOLLIN;
signal_ev.data.fd = signal_fd;
- if (epoll_ctl(ep, EPOLL_CTL_ADD, STDIN_FILENO, &stdin_ev) < 0 ||
- epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0 ||
- epoll_ctl(ep, EPOLL_CTL_ADD, master, &master_ev) < 0 ||
+ if (epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0) {
+ if (errno != EPERM) {
+ log_error("Failed to register stdout in epoll: %m");
+ r = -errno;
+ goto finish;
+ }
+ /* stdout without epoll support. Likely redirected to regular file. */
+ stdout_writable = true;
+ }
+
+ if (epoll_ctl(ep, EPOLL_CTL_ADD, master, &master_ev) < 0 ||
epoll_ctl(ep, EPOLL_CTL_ADD, signal_fd, &signal_ev) < 0) {
- log_error("Failed to regiser fds in epoll: %m");
+ log_error("Failed to register fds in epoll: %m");
r = -errno;
goto finish;
}
ssize_t k;
int i, nfds;
- if ((nfds = epoll_wait(ep, ev, ELEMENTSOF(ev), -1)) < 0) {
+ nfds = epoll_wait(ep, ev, ELEMENTSOF(ev), -1);
+ if (nfds < 0) {
if (errno == EINTR || errno == EAGAIN)
continue;
struct signalfd_siginfo sfsi;
ssize_t n;
- if ((n = read(signal_fd, &sfsi, sizeof(sfsi))) != sizeof(sfsi)) {
+ n = read(signal_fd, &sfsi, sizeof(sfsi));
+ if (n != sizeof(sfsi)) {
if (n >= 0) {
log_error("Failed to read from signalfd: invalid block size");
/* The window size changed, let's forward that. */
if (ioctl(STDIN_FILENO, TIOCGWINSZ, &ws) >= 0)
ioctl(master, TIOCSWINSZ, &ws);
+ } else if (sfsi.ssi_signo == SIGTERM && arg_boot && !tried_orderly_shutdown) {
+
+ log_info("Trying to halt container. Send SIGTERM again to trigger immediate termination.");
+
+ /* This only works for systemd... */
+ tried_orderly_shutdown = true;
+ kill(pid, SIGRTMIN+3);
+
} else {
r = 0;
goto finish;
if (stdin_readable && in_buffer_full < LINE_MAX) {
- if ((k = read(STDIN_FILENO, in_buffer + in_buffer_full, LINE_MAX - in_buffer_full)) < 0) {
+ k = read(STDIN_FILENO, in_buffer + in_buffer_full, LINE_MAX - in_buffer_full);
+ if (k < 0) {
if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO)
stdin_readable = false;
if (master_writable && in_buffer_full > 0) {
- if ((k = write(master, in_buffer, in_buffer_full)) < 0) {
+ k = write(master, in_buffer, in_buffer_full);
+ if (k < 0) {
if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO)
master_writable = false;
if (master_readable && out_buffer_full < LINE_MAX) {
- if ((k = read(master, out_buffer + out_buffer_full, LINE_MAX - out_buffer_full)) < 0) {
+ k = read(master, out_buffer + out_buffer_full, LINE_MAX - out_buffer_full);
+ if (k < 0) {
if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO)
master_readable = false;
if (stdout_writable && out_buffer_full > 0) {
- if ((k = write(STDOUT_FILENO, out_buffer, out_buffer_full)) < 0) {
+ k = write(STDOUT_FILENO, out_buffer, out_buffer_full);
+ if (k < 0) {
if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO)
stdout_writable = false;
int r = EXIT_FAILURE, k;
char *oldcg = NULL, *newcg = NULL;
char **controller = NULL;
- int master = -1;
+ int master = -1, n_fd_passed;
const char *console = NULL;
struct termios saved_attr, raw_attr;
sigset_t mask;
bool saved_attr_valid = false;
struct winsize ws;
+ int kmsg_socket_pair[2] = { -1, -1 };
+ FDSet *fds = NULL;
log_parse_environment();
log_open();
- if ((r = parse_argv(argc, argv)) <= 0)
+ r = parse_argv(argc, argv);
+ if (r <= 0)
goto finish;
if (arg_directory) {
goto finish;
}
- if ((k = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 0, &oldcg)) < 0) {
+ log_close();
+ n_fd_passed = sd_listen_fds(false);
+ if (n_fd_passed > 0) {
+ k = fdset_new_listen_fds(&fds, false);
+ if (k < 0) {
+ log_error("Failed to collect file descriptors: %s", strerror(-k));
+ goto finish;
+ }
+ }
+ fdset_close_others(fds);
+ log_open();
+
+ k = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 0, &oldcg);
+ if (k < 0) {
log_error("Failed to determine current cgroup: %s", strerror(-k));
goto finish;
}
goto finish;
}
- STRV_FOREACH(controller,arg_controllers) {
+ STRV_FOREACH(controller, arg_controllers) {
k = cg_create_and_attach(*controller, newcg, 0);
if (k < 0)
log_warning("Failed to create cgroup in controller %s: %s", *controller, strerror(-k));
}
- if ((master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY)) < 0) {
+ master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY);
+ if (master < 0) {
log_error("Failed to acquire pseudo tty: %m");
goto finish;
}
- if (!(console = ptsname(master))) {
+ console = ptsname(master);
+ if (!console) {
log_error("Failed to determine tty name: %m");
goto finish;
}
goto finish;
}
- if (tcgetattr(STDIN_FILENO, &saved_attr) < 0) {
- log_error("Failed to get terminal attributes: %m");
- goto finish;
- }
-
- saved_attr_valid = true;
+ if (tcgetattr(STDIN_FILENO, &saved_attr) >= 0) {
+ saved_attr_valid = true;
- raw_attr = saved_attr;
- cfmakeraw(&raw_attr);
- raw_attr.c_lflag &= ~ECHO;
+ raw_attr = saved_attr;
+ cfmakeraw(&raw_attr);
+ raw_attr.c_lflag &= ~ECHO;
+ }
- if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) {
- log_error("Failed to set terminal attributes: %m");
+ 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;
}
sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1);
assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0);
- pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL);
- if (pid < 0) {
- if (errno == EINVAL)
- log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m");
- else
- log_error("clone() failed: %m");
+ for (;;) {
+ siginfo_t status;
+ int pipefd[2];
- goto finish;
- }
+ if (pipe2(pipefd, O_NONBLOCK|O_CLOEXEC) < 0) {
+ log_error("pipe2(): %m");
+ goto finish;
+ }
- if (pid == 0) {
- /* child */
+ pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL);
+ if (pid < 0) {
+ if (errno == EINVAL)
+ log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m");
+ else
+ log_error("clone() failed: %m");
- const char *hn;
- const char *home = NULL;
- uid_t uid = (uid_t) -1;
- gid_t gid = (gid_t) -1;
- const char *envp[] = {
- "PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin",
- "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */
- NULL, /* TERM */
- NULL, /* HOME */
- NULL, /* USER */
- NULL, /* LOGNAME */
- NULL
- };
+ goto finish;
+ }
- envp[2] = strv_find_prefix(environ, "TERM=");
+ if (pid == 0) {
+ /* child */
+ const char *home = NULL;
+ uid_t uid = (uid_t) -1;
+ gid_t gid = (gid_t) -1;
+ unsigned n_env = 2;
+ const char *envp[] = {
+ "PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin",
+ "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */
+ NULL, /* TERM */
+ NULL, /* HOME */
+ NULL, /* USER */
+ NULL, /* LOGNAME */
+ NULL, /* container_uuid */
+ NULL, /* LISTEN_FDS */
+ NULL, /* LISTEN_PID */
+ NULL
+ };
+
+ envp[n_env] = strv_find_prefix(environ, "TERM=");
+ if (envp[n_env])
+ n_env ++;
+
+ close_nointr_nofail(pipefd[1]);
+ fd_wait_for_event(pipefd[0], POLLHUP, -1);
+ close_nointr_nofail(pipefd[0]);
+
+ close_nointr_nofail(master);
+ master = -1;
+
+ if (saved_attr_valid) {
+ if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) {
+ log_error("Failed to set terminal attributes: %m");
+ goto child_fail;
+ }
+ }
- close_nointr_nofail(master);
+ close_nointr(STDIN_FILENO);
+ close_nointr(STDOUT_FILENO);
+ close_nointr(STDERR_FILENO);
- close_nointr(STDIN_FILENO);
- close_nointr(STDOUT_FILENO);
- close_nointr(STDERR_FILENO);
+ close_nointr_nofail(kmsg_socket_pair[0]);
+ kmsg_socket_pair[0] = -1;
- close_all_fds(NULL, 0);
+ reset_all_signal_handlers();
- reset_all_signal_handlers();
+ assert_se(sigemptyset(&mask) == 0);
+ assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
- assert_se(sigemptyset(&mask) == 0);
- assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
+ k = open_terminal(console, O_RDWR);
+ if (k != STDIN_FILENO) {
+ if (k >= 0) {
+ close_nointr_nofail(k);
+ k = -EINVAL;
+ }
- if (setsid() < 0)
- goto child_fail;
+ log_error("Failed to open console: %s", strerror(-k));
+ goto child_fail;
+ }
- if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0)
- goto child_fail;
+ if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
+ dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) {
+ log_error("Failed to duplicate console: %m");
+ goto child_fail;
+ }
- /* Mark / as private, in case somebody marked it shared */
- if (mount(NULL, "/", NULL, MS_PRIVATE|MS_REC, NULL) < 0)
- goto child_fail;
+ if (setsid() < 0) {
+ log_error("setsid() failed: %m");
+ goto child_fail;
+ }
- if (mount_all(arg_directory) < 0)
- goto child_fail;
+ if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) {
+ log_error("PR_SET_PDEATHSIG failed: %m");
+ goto child_fail;
+ }
- if (copy_devnodes(arg_directory, console) < 0)
- goto child_fail;
+ /* Mark everything as slave, so that we still
+ * receive mounts from the real root, but don't
+ * propagate mounts to the real root. */
+ if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) {
+ log_error("MS_SLAVE|MS_REC failed: %m");
+ goto child_fail;
+ }
- if (chdir(arg_directory) < 0) {
- log_error("chdir(%s) failed: %m", arg_directory);
- goto child_fail;
- }
+ /* Turn directory into bind mount */
+ if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, NULL) < 0) {
+ log_error("Failed to make bind mount.");
+ goto child_fail;
+ }
- if (open_terminal("dev/console", O_RDWR) != STDIN_FILENO ||
- dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
- dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO)
- goto child_fail;
+ if (arg_read_only)
+ if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY|MS_REC, NULL) < 0) {
+ log_error("Failed to make read-only.");
+ goto child_fail;
+ }
- if (mount(arg_directory, "/", "bind", MS_BIND|MS_MOVE, NULL) < 0) {
- log_error("mount(MS_MOVE) failed: %m");
- goto child_fail;
- }
+ if (mount_all(arg_directory) < 0)
+ goto child_fail;
- if (chroot(".") < 0) {
- log_error("chroot() failed: %m");
- goto child_fail;
- }
+ if (copy_devnodes(arg_directory) < 0)
+ goto child_fail;
- if (chdir("/") < 0) {
- log_error("chdir() failed: %m");
- goto child_fail;
- }
+ if (setup_ptmx(arg_directory) < 0)
+ goto child_fail;
- umask(0022);
+ dev_setup(arg_directory);
+
+ if (setup_dev_console(arg_directory, console) < 0)
+ goto child_fail;
- loopback_setup();
+ if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0)
+ goto child_fail;
+
+ close_nointr_nofail(kmsg_socket_pair[1]);
+ kmsg_socket_pair[1] = -1;
+
+ if (setup_boot_id(arg_directory) < 0)
+ goto child_fail;
+
+ if (setup_timezone(arg_directory) < 0)
+ goto child_fail;
+
+ if (setup_resolv_conf(arg_directory) < 0)
+ goto child_fail;
+
+ if (setup_journal(arg_directory) < 0)
+ goto child_fail;
- if (drop_capabilities() < 0)
- goto child_fail;
+ if (mount_binds(arg_directory, arg_bind, 0) < 0)
+ goto child_fail;
- if (arg_user) {
+ if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0)
+ goto child_fail;
- if (get_user_creds((const char**)&arg_user, &uid, &gid, &home) < 0) {
- log_error("get_user_creds() failed: %m");
+ if (chdir(arg_directory) < 0) {
+ log_error("chdir(%s) failed: %m", arg_directory);
goto child_fail;
}
- if (mkdir_parents(home, 0775) < 0) {
- log_error("mkdir_parents() failed: %m");
+ if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) {
+ log_error("mount(MS_MOVE) failed: %m");
goto child_fail;
}
- if (safe_mkdir(home, 0775, uid, gid) < 0) {
- log_error("safe_mkdir() failed: %m");
+ if (chroot(".") < 0) {
+ log_error("chroot() failed: %m");
goto child_fail;
}
- if (initgroups((const char*)arg_user, gid) < 0) {
- log_error("initgroups() failed: %m");
+ if (chdir("/") < 0) {
+ log_error("chdir() failed: %m");
goto child_fail;
}
- if (setresgid(gid, gid, gid) < 0) {
- log_error("setregid() failed: %m");
+ umask(0022);
+
+ loopback_setup();
+
+ if (drop_capabilities() < 0) {
+ log_error("drop_capabilities() failed: %m");
goto child_fail;
}
- if (setresuid(uid, uid, uid) < 0) {
- log_error("setreuid() failed: %m");
+ if (arg_user) {
+
+ /* Note that this resolves user names
+ * inside the container, and hence
+ * accesses the NSS modules from the
+ * container and not the host. This is
+ * a bit weird... */
+
+ if (get_user_creds((const char**)&arg_user, &uid, &gid, &home, NULL) < 0) {
+ log_error("get_user_creds() failed: %m");
+ goto child_fail;
+ }
+
+ if (mkdir_parents_label(home, 0775) < 0) {
+ log_error("mkdir_parents_label() failed: %m");
+ goto child_fail;
+ }
+
+ if (mkdir_safe_label(home, 0775, uid, gid) < 0) {
+ log_error("mkdir_safe_label() failed: %m");
+ goto child_fail;
+ }
+
+ if (initgroups((const char*)arg_user, gid) < 0) {
+ log_error("initgroups() failed: %m");
+ goto child_fail;
+ }
+
+ if (setresgid(gid, gid, gid) < 0) {
+ log_error("setregid() failed: %m");
+ goto child_fail;
+ }
+
+ if (setresuid(uid, uid, uid) < 0) {
+ log_error("setreuid() failed: %m");
+ goto child_fail;
+ }
+ } else {
+ /* Reset everything fully to 0, just in case */
+
+ if (setgroups(0, NULL) < 0) {
+ log_error("setgroups() failed: %m");
+ goto child_fail;
+ }
+
+ if (setresgid(0, 0, 0) < 0) {
+ log_error("setregid() failed: %m");
+ goto child_fail;
+ }
+
+ if (setresuid(0, 0, 0) < 0) {
+ log_error("setreuid() failed: %m");
+ goto child_fail;
+ }
+ }
+
+ if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) ||
+ (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) ||
+ (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) {
+ log_oom();
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)) {
- log_error("Out of memory");
- goto child_fail;
- }
+ if (arg_uuid) {
+ if (asprintf((char**)(envp + n_env++), "container_uuid=%s", arg_uuid) < 0) {
+ log_oom();
+ goto child_fail;
+ }
+ }
+
+ if (fdset_size(fds) > 0) {
+ k = fdset_cloexec(fds, false);
+ if (k < 0) {
+ log_error("Failed to unset O_CLOEXEC for file descriptors.");
+ goto child_fail;
+ }
+
+ if ((asprintf((char **)(envp + n_env++), "LISTEN_FDS=%u", n_fd_passed) < 0) ||
+ (asprintf((char **)(envp + n_env++), "LISTEN_PID=%lu", (unsigned long) getpid()) < 0)) {
+ log_oom();
+ goto child_fail;
+ }
+ }
+
+ setup_hostname();
+
+ if (arg_boot) {
+ char **a;
+ size_t l;
+
+ /* 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");
+ execle("/bin/bash", "-bash", NULL, (char**) envp);
+ }
- if ((hn = file_name_from_path(arg_directory)))
- sethostname(hn, strlen(hn));
+ log_error("execv() failed: %m");
- if (argc > optind)
- execvpe(argv[optind], argv + optind, (char**) envp);
- else {
- chdir(home ? home : "/root");
- execle("/bin/bash", "-bash", NULL, (char**) envp);
+ child_fail:
+ _exit(EXIT_FAILURE);
}
- log_error("execv() failed: %m");
+ log_info("Init process in the container running as PID %lu.", (unsigned long) pid);
+ close_nointr_nofail(pipefd[0]);
+ close_nointr_nofail(pipefd[1]);
- child_fail:
- _exit(EXIT_FAILURE);
- }
+ fdset_free(fds);
+ fds = NULL;
- if (process_pty(master, &mask) < 0)
- goto finish;
+ if (process_pty(master, pid, &mask) < 0)
+ goto finish;
- if (saved_attr_valid) {
- tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr);
- saved_attr_valid = false;
- }
+ if (saved_attr_valid)
+ tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr);
+
+ r = wait_for_terminate(pid, &status);
+ if (r < 0) {
+ r = EXIT_FAILURE;
+ break;
+ }
+
+ if (status.si_code == CLD_EXITED) {
+ if (status.si_status != 0) {
+ log_error("Container failed with error code %i.", status.si_status);
+ r = status.si_status;
+ break;
+ }
- r = wait_for_terminate_and_warn(argc > optind ? argv[optind] : "bash", pid);
+ log_debug("Container exited successfully.");
+ break;
+ } else if (status.si_code == CLD_KILLED &&
+ status.si_status == SIGINT) {
+ log_info("Container has been shut down.");
+ r = 0;
+ break;
+ } else if (status.si_code == CLD_KILLED &&
+ status.si_status == SIGHUP) {
+ log_info("Container is being rebooted.");
+ continue;
+ } else if (status.si_code == CLD_KILLED ||
+ status.si_code == CLD_DUMPED) {
- if (r < 0)
- r = EXIT_FAILURE;
+ log_error("Container terminated by signal %s.", signal_to_string(status.si_status));
+ r = EXIT_FAILURE;
+ break;
+ } else {
+ log_error("Container failed due to unknown reason.");
+ r = EXIT_FAILURE;
+ break;
+ }
+ }
finish:
if (saved_attr_valid)
if (master >= 0)
close_nointr_nofail(master);
+ close_pipe(kmsg_socket_pair);
+
if (oldcg)
cg_attach(SYSTEMD_CGROUP_CONTROLLER, oldcg, 0);
free(oldcg);
free(newcg);
+ fdset_free(fds);
+
return r;
}