#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 "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,
(1ULL << CAP_SYS_PTRACE) |
(1ULL << CAP_SYS_TTY_CONFIG) |
(1ULL << CAP_SYS_RESOURCE) |
- (1ULL << CAP_SYS_BOOT);
+ (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"
- " -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"
- " --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"
- " --link-journal=MODE Link up guest journal, one of no, auto, guest, host\n"
- " -j Equivalent to --link-journal=host\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_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' },
{ "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 }
};
help();
return 0;
+ case ARG_VERSION:
+ puts(PACKAGE_STRING);
+ puts(SYSTEMD_FEATURES);
+ return 0;
+
case 'D':
free(arg_directory);
arg_directory = canonicalize_file_name(optarg);
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;
{ 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", NULL, 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
char _cleanup_free_ *where = NULL;
int t;
- if (asprintf(&where, "%s/%s", dest, mount_table[k].where) < 0) {
- log_oom();
-
- if (r == 0)
- r = -ENOMEM;
-
- break;
- }
+ where = strjoin(dest, "/", mount_table[k].where, NULL);
+ if (!where)
+ return log_oom();
t = path_is_mount_point(where, true);
if (t < 0) {
if (mount_table[k].what && t > 0)
continue;
- mkdir_p_label(where, 0755);
+ mkdir_p(where, 0755);
if (mount(mount_table[k].what,
where,
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);
+
+ 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;
SD_ID128_FORMAT_VAL(rnd));
char_array_0(as_uuid);
- r = write_one_line_file(from, 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
- mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
+ } 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;
"full\0"
"random\0"
"urandom\0"
- "tty\0"
- "ptmx\0";
+ "tty\0";
const char *d;
int r = 0;
return r;
}
+static int setup_ptmx(const char *dest) {
+ _cleanup_free_ char *p = NULL;
+
+ p = strappend(dest, "/dev/ptmx");
+ if (!p)
+ return log_oom();
+
+ if (symlink("pts/ptmx", p) < 0) {
+ log_error("Failed to create /dev/ptmx symlink: %m");
+ return -errno;
+ }
+
+ return 0;
+}
+
static int setup_dev_console(const char *dest, const char *console) {
struct stat st;
char _cleanup_free_ *to = NULL;
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);
signal_ev.events = EPOLLIN;
signal_ev.data.fd = signal_fd;
- if (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 register fds in epoll: %m");
r = -errno;
/* 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;
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();
goto finish;
}
+ 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));
for (;;) {
siginfo_t status;
+ int pipefd[2];
- if (saved_attr_valid) {
- if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) {
- log_error("Failed to set terminal attributes: %m");
- goto finish;
- }
+ if (pipe2(pipefd, O_NONBLOCK|O_CLOEXEC) < 0) {
+ log_error("pipe2(): %m");
+ goto finish;
}
pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL);
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, /* USER */
NULL, /* LOGNAME */
NULL, /* container_uuid */
+ NULL, /* LISTEN_FDS */
+ NULL, /* LISTEN_PID */
NULL
};
- envp[2] = strv_find_prefix(environ, "TERM=");
+ 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(STDIN_FILENO);
close_nointr(STDOUT_FILENO);
close_nointr(STDERR_FILENO);
- close_all_fds(&kmsg_socket_pair[1], 1);
+ close_nointr_nofail(kmsg_socket_pair[0]);
+ kmsg_socket_pair[0] = -1;
reset_all_signal_handlers();
assert_se(sigemptyset(&mask) == 0);
assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
- if (open_terminal(console, O_RDWR) != STDIN_FILENO ||
- dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
- dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO)
+ k = open_terminal(console, O_RDWR);
+ if (k != STDIN_FILENO) {
+ if (k >= 0) {
+ close_nointr_nofail(k);
+ k = -EINVAL;
+ }
+
+ log_error("Failed to open console: %s", strerror(-k));
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;
+ }
if (setsid() < 0) {
log_error("setsid() failed: %m");
if (copy_devnodes(arg_directory) < 0)
goto child_fail;
+ if (setup_ptmx(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]);
+ kmsg_socket_pair[1] = -1;
if (setup_boot_id(arg_directory) < 0)
goto child_fail;
if (setup_journal(arg_directory) < 0)
goto child_fail;
+ if (mount_binds(arg_directory, arg_bind, 0) < 0)
+ goto child_fail;
+
+ if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0)
+ goto child_fail;
+
if (chdir(arg_directory) < 0) {
log_error("chdir(%s) failed: %m", arg_directory);
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 + 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 (arg_uuid) {
- if (asprintf((char**)(envp + 6), "container_uuid=%s", arg_uuid) < 0) {
+ 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;
}
_exit(EXIT_FAILURE);
}
- if (process_pty(master, &mask) < 0)
- goto finish;
+ log_info("Init process in the container running as PID %lu.", (unsigned long) pid);
+ close_nointr_nofail(pipefd[0]);
+ close_nointr_nofail(pipefd[1]);
+ fdset_free(fds);
+ fds = NULL;
+
+ if (process_pty(master, pid, &mask) < 0)
+ goto finish;
if (saved_attr_valid)
tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr);
free(oldcg);
free(newcg);
+ fdset_free(fds);
+
return r;
}
~ianmdlvl / elogind.git / blobdiff