#include "path-util.h"
#include "loopback-setup.h"
#include "sd-id128.h"
+#include "dev-setup.h"
typedef enum LinkJournal {
LINK_NO,
char *t;
t = strndup(word, length);
- if (!t) {
- log_error("Out of memory.");
- return -ENOMEM;
- }
+ if (!t)
+ return log_oom();
if (cap_from_name(t, &cap) < 0) {
log_error("Failed to parse capability %s.", t);
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 },
+ { "/dev/pts", "/dev/pts", NULL, NULL, MS_BIND, 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
};
int t;
if (asprintf(&where, "%s/%s", dest, mount_table[k].where) < 0) {
- log_error("Out of memory");
+ 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);
continue;
}
+ /* Skip this entry if it is not a remount. */
+ if (mount_table[k].what && t > 0)
+ continue;
+
mkdir_p_label(where, 0755);
if (mount(mount_table[k].what,
assert(dest);
/* Fix the timezone, if possible */
- if (asprintf(&where, "%s/etc/localtime", dest) < 0) {
- log_error("Out of memory");
- return -ENOMEM;
- }
+ where = strappend(dest, "/etc/localtime");
+ if (!where)
+ return log_oom();
if (mount("/etc/localtime", where, "bind", MS_BIND, NULL) >= 0)
mount("/etc/localtime", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
free(where);
- if (asprintf(&where, "%s/etc/timezone", dest) < 0) {
- log_error("Out of memory");
- return -ENOMEM;
- }
+ where = strappend(dest, "/etc/timezone");
+ if (!where)
+ return log_oom();
if (mount("/etc/timezone", where, "bind", MS_BIND, NULL) >= 0)
mount("/etc/timezone", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
return 0;
/* Fix resolv.conf, if possible */
- if (asprintf(&where, "%s/etc/resolv.conf", dest) < 0) {
- log_error("Out of memory");
- return -ENOMEM;
- }
+ where = strappend(dest, "/etc/resolv.conf");
+ if (!where)
+ return log_oom();
if (mount("/etc/resolv.conf", where, "bind", MS_BIND, NULL) >= 0)
mount("/etc/resolv.conf", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
return 0;
}
+static int setup_boot_id(const char *dest) {
+ char *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");
+ if (!from) {
+ r = log_oom();
+ goto finish;
+ }
+
+ to = strappend(dest, "/proc/sys/kernel/random/boot_id");
+ if (!to) {
+ r = log_oom();
+ goto finish;
+ }
+
+ r = sd_id128_randomize(&rnd);
+ if (r < 0) {
+ log_error("Failed to generate random boot id: %s", strerror(-r));
+ goto finish;
+ }
+
+ 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_one_line_file(from, as_uuid);
+ if (r < 0) {
+ log_error("Failed to write boot id: %s", strerror(-r));
+ goto finish;
+ }
+
+ if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
+ log_error("Failed to bind mount boot id: %m");
+ r = -errno;
+ } else
+ mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
+
+ unlink(from);
+
+finish:
+ free(from);
+ free(to);
+
+ return r;
+}
+
static int copy_devnodes(const char *dest) {
static const char devnodes[] =
}
if (asprintf(&to, "%s/dev/console", dest) < 0) {
- log_error("Out of memory");
- r = -ENOMEM;
+ r = log_oom();
goto finish;
}
* 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;
+ r = log_oom();
goto finish;
}
if (asprintf(&to, "%s/proc/kmsg", dest) < 0) {
- log_error("Out of memory");
- r = -ENOMEM;
+ r = log_oom();
goto finish;
}
p = strappend(directory, "/etc/machine-id");
if (!p) {
- log_error("Out of memory");
- r = -ENOMEM;
+ r = log_oom();
goto finish;
}
p = strappend("/var/log/journal/", l);
q = strjoin(directory, "/var/log/journal/", l, NULL);
if (!p || !q) {
- log_error("Out of memory");
- r = -ENOMEM;
+ r = log_oom();
goto finish;
}
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;
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");
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;
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) {
+ 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;
}
assert_se(sigemptyset(&mask) == 0);
assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
- if (setsid() < 0)
+ if (open_terminal(console, O_RDWR) != STDIN_FILENO ||
+ dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
+ dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO)
goto child_fail;
- if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0)
+ if (setsid() < 0) {
+ log_error("setsid() failed: %m");
goto child_fail;
+ }
- /* Mark / as private, in case somebody marked it shared */
- if (mount(NULL, "/", NULL, MS_PRIVATE|MS_REC, NULL) < 0)
+ if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) {
+ log_error("PR_SET_PDEATHSIG failed: %m");
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;
+ }
/* Turn directory into bind mount */
- if (mount(arg_directory, arg_directory, "bind", MS_BIND, NULL) < 0) {
+ if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, 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) {
+ 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 (copy_devnodes(arg_directory) < 0)
goto child_fail;
+ dev_setup(arg_directory);
+
if (setup_dev_console(arg_directory, console) < 0)
goto child_fail;
close_nointr_nofail(kmsg_socket_pair[1]);
+ if (setup_boot_id(arg_directory) < 0)
+ goto child_fail;
+
if (setup_timezone(arg_directory) < 0)
goto child_fail;
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 (mount(arg_directory, "/", "bind", MS_MOVE, NULL) < 0) {
- log_error("mount(MS_BIND) failed: %m");
+ if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) {
+ log_error("mount(MS_MOVE) 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)) {
- log_error("Out of memory");
+ log_oom();
goto child_fail;
}
if (arg_uuid) {
if (asprintf((char**)(envp + 6), "container_uuid=%s", arg_uuid) < 0) {
- log_error("Out of memory");
+ log_oom();
goto child_fail;
}
}