#include "def.h"
#include "rtnl-util.h"
#include "udev-util.h"
+#include "eventfd-util.h"
#include "blkid-util.h"
#include "gpt.h"
#include "siphash24.h"
+#include "copy.h"
#ifdef HAVE_SECCOMP
#include "seccomp-util.h"
(1ULL << CAP_MKNOD);
static char **arg_bind = NULL;
static char **arg_bind_ro = NULL;
+static char **arg_tmpfs = NULL;
static char **arg_setenv = NULL;
static bool arg_quiet = false;
static bool arg_share_system = false;
" --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"
+ " --tmpfs=PATH:[OPTIONS] Mount an empty tmpfs to the specified directory\n"
" --setenv=NAME=VALUE Pass an environment variable to PID 1\n"
" --share-system Share system namespaces with host\n"
" --register=BOOLEAN Register container as machine\n"
ARG_LINK_JOURNAL,
ARG_BIND,
ARG_BIND_RO,
+ ARG_TMPFS,
ARG_SETENV,
ARG_SHARE_SYSTEM,
ARG_REGISTER,
{ "link-journal", required_argument, NULL, ARG_LINK_JOURNAL },
{ "bind", required_argument, NULL, ARG_BIND },
{ "bind-ro", required_argument, NULL, ARG_BIND_RO },
+ { "tmpfs", required_argument, NULL, ARG_TMPFS },
{ "machine", required_argument, NULL, 'M' },
{ "slice", required_argument, NULL, 'S' },
{ "setenv", required_argument, NULL, ARG_SETENV },
break;
}
+ case ARG_TMPFS: {
+ _cleanup_free_ char *a = NULL, *b = NULL;
+ char *e;
+
+ e = strchr(optarg, ':');
+ if (e) {
+ a = strndup(optarg, e - optarg);
+ b = strdup(e + 1);
+ } else {
+ a = strdup(optarg);
+ b = strdup("mode=0755");
+ }
+
+ if (!a || !b)
+ return log_oom();
+
+ if (!path_is_absolute(a)) {
+ log_error("Invalid tmpfs specification: %s", optarg);
+ return -EINVAL;
+ }
+
+ r = strv_push(&arg_tmpfs, a);
+ if (r < 0)
+ return log_oom();
+
+ a = NULL;
+
+ r = strv_push(&arg_tmpfs, b);
+ if (r < 0)
+ return log_oom();
+
+ b = NULL;
+
+ break;
+ }
+
case ARG_SETENV: {
char **n;
} MountPoint;
static const MountPoint mount_table[] = {
- { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
- { "/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 },
+ { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
+ { "/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 },
{ "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 },
+ { "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", 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 */
+ { "/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
};
return r;
}
-static int mount_binds(const char *dest, char **l, unsigned long flags) {
+static int mount_binds(const char *dest, char **l, bool ro) {
char **x, **y;
STRV_FOREACH_PAIR(x, y, l) {
- char *where;
+ _cleanup_free_ char *where = NULL;
struct stat source_st, dest_st;
int r;
return -errno;
}
- where = strappenda(dest, *y);
+ where = strappend(dest, *y);
+ if (!where)
+ return log_oom();
+
r = stat(where, &dest_st);
if (r == 0) {
if ((source_st.st_mode & S_IFMT) != (dest_st.st_mode & S_IFMT)) {
- log_error("The file types of %s and %s do not match. Refusing bind mount",
- *x, where);
+ log_error("The file types of %s and %s do not match. Refusing bind mount", *x, where);
return -EINVAL;
}
} else if (errno == ENOENT) {
log_error("Failed to bind mount %s: %m", *x);
return -errno;
}
+
/* Create the mount point, but be conservative -- refuse to create block
* and char devices. */
if (S_ISDIR(source_st.st_mode))
return -errno;
}
- if (flags && mount(NULL, where, NULL, MS_REMOUNT|MS_BIND|flags, NULL) < 0) {
- log_error("mount(%s) failed: %m", where);
+ if (ro) {
+ r = bind_remount_recursive(where, true);
+ if (r < 0) {
+ log_error("Read-Only bind mount failed: %s", strerror(-r));
+ return r;
+ }
+ }
+ }
+
+ return 0;
+}
+
+static int mount_tmpfs(const char *dest) {
+ char **i, **o;
+
+ STRV_FOREACH_PAIR(i, o, arg_tmpfs) {
+ _cleanup_free_ char *where = NULL;
+
+ where = strappend(dest, *i);
+ if (!where)
+ return log_oom();
+
+ mkdir_label(where, 0755);
+
+ if (mount("tmpfs", where, "tmpfs", MS_NODEV|MS_STRICTATIME, *o) < 0) {
+ log_error("tmpfs mount to %s failed: %m", where);
return -errno;
}
}
/* We don't really care for the results of this really. If it
* fails, it fails, but meh... */
- copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW);
+ copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW, 0644);
return 0;
}
return r;
}
+static void nop_handler(int sig) {}
+
int main(int argc, char *argv[]) {
_cleanup_free_ char *kdbus_domain = NULL, *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL;
const char *console = NULL;
char veth_name[IFNAMSIZ];
bool secondary = false;
+ sigset_t mask, mask_chld;
pid_t pid = 0;
- sigset_t mask;
log_parse_environment();
log_open();
if (arg_boot) {
if (path_is_os_tree(arg_directory) <= 0) {
- log_error("Directory %s doesn't look like an OS root directory (/etc/os-release is missing). Refusing.", arg_directory);
+ log_error("Directory %s doesn't look like an OS root directory (os-release file is missing). Refusing.", arg_directory);
goto finish;
}
} else {
}
if (!arg_quiet)
- log_info("Spawning container %s on %s. Press ^] three times within 1s to abort execution.", arg_machine, arg_image ? arg_image : arg_directory);
+ log_info("Spawning container %s on %s.\nPress ^] three times within 1s to kill container.",
+ arg_machine, arg_image ? arg_image : arg_directory);
if (unlockpt(master) < 0) {
log_error("Failed to unlock tty: %m");
sd_notify(0, "READY=1");
assert_se(sigemptyset(&mask) == 0);
+ assert_se(sigemptyset(&mask_chld) == 0);
+ sigaddset(&mask_chld, SIGCHLD);
sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1);
assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0);
for (;;) {
ContainerStatus container_status;
- int parent_ready_fd = -1, child_ready_fd = -1;
- eventfd_t x;
-
- parent_ready_fd = eventfd(0, EFD_CLOEXEC);
- if (parent_ready_fd < 0) {
- log_error("Failed to create event fd: %m");
+ int eventfds[2] = { -1, -1 };
+ struct sigaction sa = {
+ .sa_handler = nop_handler,
+ .sa_flags = SA_NOCLDSTOP,
+ };
+
+ /* Child can be killed before execv(), so handle SIGCHLD
+ * in order to interrupt parent's blocking calls and
+ * give it a chance to call wait() and terminate. */
+ r = sigprocmask(SIG_UNBLOCK, &mask_chld, NULL);
+ if (r < 0) {
+ log_error("Failed to change the signal mask: %m");
goto finish;
}
- child_ready_fd = eventfd(0, EFD_CLOEXEC);
- if (child_ready_fd < 0) {
- log_error("Failed to create event fd: %m");
+ r = sigaction(SIGCHLD, &sa, NULL);
+ if (r < 0) {
+ log_error("Failed to install SIGCHLD handler: %m");
goto finish;
}
- pid = syscall(__NR_clone,
- SIGCHLD|CLONE_NEWNS|
- (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)|
- (arg_private_network ? CLONE_NEWNET : 0), NULL);
+ pid = clone_with_eventfd(SIGCHLD|CLONE_NEWNS|
+ (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)|
+ (arg_private_network ? CLONE_NEWNET : 0), eventfds);
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");
+ r = pid;
goto finish;
}
/* 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.");
+ log_error("Failed to make bind mount: %m");
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.");
+ if (arg_read_only) {
+ k = bind_remount_recursive(arg_directory, true);
+ if (k < 0) {
+ log_error("Failed to make tree read-only: %s", strerror(-k));
goto child_fail;
}
+ }
if (mount_all(arg_directory) < 0)
goto child_fail;
if (setup_journal(arg_directory) < 0)
goto child_fail;
- if (mount_binds(arg_directory, arg_bind, 0) < 0)
+ if (mount_binds(arg_directory, arg_bind, false) < 0)
goto child_fail;
- if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0)
+ if (mount_binds(arg_directory, arg_bind_ro, true) < 0)
+ goto child_fail;
+
+ if (mount_tmpfs(arg_directory) < 0)
goto child_fail;
if (setup_kdbus(arg_directory, kdbus_domain) < 0)
/* Tell the parent that we are ready, and that
* it can cgroupify us to that we lack access
* to certain devices and resources. */
- eventfd_write(child_ready_fd, 1);
- child_ready_fd = safe_close(child_ready_fd);
+ r = eventfd_send_state(eventfds[1],
+ EVENTFD_CHILD_SUCCEEDED);
+ eventfds[1] = safe_close(eventfds[1]);
+ if (r < 0)
+ goto child_fail;
if (chdir(arg_directory) < 0) {
log_error("chdir(%s) failed: %m", arg_directory);
env_use = (char**) envp;
/* Wait until the parent is ready with the setup, too... */
- eventfd_read(parent_ready_fd, &x);
- parent_ready_fd = safe_close(parent_ready_fd);
+ r = eventfd_parent_succeeded(eventfds[0]);
+ eventfds[0] = safe_close(eventfds[0]);
+ if (r < 0)
+ goto child_fail;
if (arg_boot) {
char **a;
log_error("execv() failed: %m");
child_fail:
+ /* Tell the parent that the setup failed, so he
+ * can clean up resources and terminate. */
+ if (eventfds[1] != -1)
+ eventfd_send_state(eventfds[1],
+ EVENTFD_CHILD_FAILED);
_exit(EXIT_FAILURE);
}
fdset_free(fds);
fds = NULL;
- /* Wait until the child reported that it is ready with
- * all it needs to do with privileges. After we got
- * the notification we can make the process join its
- * cgroup which might limit what it can do */
- eventfd_read(child_ready_fd, &x);
+ /* Wait for the child event:
+ * If EVENTFD_CHILD_FAILED, the child will terminate soon.
+ * If EVENTFD_CHILD_SUCCEEDED, the child is reporting that
+ * it is ready with all it needs to do with priviliges.
+ * After we got the notification we can make the process
+ * join its cgroup which might limit what it can do */
+ r = eventfd_child_succeeded(eventfds[1]);
+ eventfds[1] = safe_close(eventfds[1]);
+ if (r < 0)
+ goto check_container_status;
r = register_machine(pid);
if (r < 0)
if (r < 0)
goto finish;
+ /* Block SIGCHLD here, before notifying child.
+ * process_pty() will handle it with the other signals. */
+ r = sigprocmask(SIG_BLOCK, &mask_chld, NULL);
+ if (r < 0)
+ goto finish;
+
+ /* Reset signal to default */
+ r = default_signals(SIGCHLD, -1);
+ if (r < 0)
+ goto finish;
+
/* Notify the child that the parent is ready with all
- * its setup, and thtat the child can now hand over
+ * its setup, and that the child can now hand over
* control to the code to run inside the container. */
- eventfd_write(parent_ready_fd, 1);
+ r = eventfd_send_state(eventfds[0],
+ EVENTFD_PARENT_SUCCEEDED);
+ eventfds[0] = safe_close(eventfds[0]);
+ if (r < 0)
+ goto finish;
k = process_pty(master, &mask, arg_boot ? pid : 0, SIGRTMIN+3);
if (k < 0) {
/* Kill if it is not dead yet anyway */
terminate_machine(pid);
+check_container_status:
/* Redundant, but better safe than sorry */
kill(pid, SIGKILL);
strv_free(arg_network_macvlan);
strv_free(arg_bind);
strv_free(arg_bind_ro);
+ strv_free(arg_tmpfs);
return r;
}