#include <sys/syscall.h>
#include <limits.h>
#include <linux/fs.h>
+#include <sys/file.h>
#include "strv.h"
#include "util.h"
#include "path-util.h"
#include "namespace.h"
#include "missing.h"
+#include "execute.h"
+#include "loopback-setup.h"
+#include "mkdir.h"
+#include "dev-setup.h"
+#include "def.h"
-typedef enum PathMode {
+typedef enum MountMode {
/* This is ordered by priority! */
INACCESSIBLE,
READONLY,
- PRIVATE,
+ PRIVATE_TMP,
+ PRIVATE_VAR_TMP,
+ PRIVATE_DEV,
READWRITE
-} PathMode;
+} MountMode;
-typedef struct Path {
+typedef struct BindMount {
const char *path;
- PathMode mode;
-} Path;
+ MountMode mode;
+ bool done;
+ bool ignore;
+} BindMount;
-static int append_paths(Path **p, char **strv, PathMode mode) {
+static int append_mounts(BindMount **p, char **strv, MountMode mode) {
char **i;
+ assert(p);
+
STRV_FOREACH(i, strv) {
+ (*p)->ignore = false;
+
+ if ((mode == INACCESSIBLE || mode == READONLY) && (*i)[0] == '-') {
+ (*p)->ignore = true;
+ (*i)++;
+ }
+
if (!path_is_absolute(*i))
return -EINVAL;
return 0;
}
-static int path_compare(const void *a, const void *b) {
- const Path *p = a, *q = b;
+static int mount_path_compare(const void *a, const void *b) {
+ const BindMount *p = a, *q = b;
if (path_equal(p->path, q->path)) {
return 0;
}
-static void drop_duplicates(Path *p, unsigned *n, bool *need_inaccessible, bool *need_private) {
- Path *f, *t, *previous;
+static void drop_duplicates(BindMount *m, unsigned *n) {
+ BindMount *f, *t, *previous;
- assert(p);
+ assert(m);
assert(n);
- assert(need_inaccessible);
- assert(need_private);
- for (f = p, t = p, previous = NULL; f < p+*n; f++) {
+ for (f = m, t = m, previous = NULL; f < m+*n; f++) {
+ /* The first one wins */
if (previous && path_equal(f->path, previous->path))
continue;
t->path = f->path;
t->mode = f->mode;
- if (t->mode == PRIVATE)
- *need_private = true;
-
- if (t->mode == INACCESSIBLE)
- *need_inaccessible = true;
-
previous = t;
t++;
}
- *n = t - p;
+ *n = t - m;
}
-static int apply_mount(Path *p, const char *root_dir, const char *inaccessible_dir, const char *private_dir, unsigned long flags) {
+static int mount_dev(BindMount *m) {
+ static const char devnodes[] =
+ "/dev/null\0"
+ "/dev/zero\0"
+ "/dev/full\0"
+ "/dev/random\0"
+ "/dev/urandom\0"
+ "/dev/tty\0";
+
+ struct stat devnodes_stat[6] = {};
+ const char *d;
+ unsigned n = 0;
+ _cleanup_umask_ mode_t u;
+ int r;
+
+ assert(m);
+
+ u = umask(0000);
+
+ /* First: record device mode_t and dev_t */
+ NULSTR_FOREACH(d, devnodes) {
+ r = stat(d, &devnodes_stat[n]);
+ if (r < 0) {
+ if (errno != ENOENT)
+ return -errno;
+ } else {
+ if (!S_ISBLK(devnodes_stat[n].st_mode) &&
+ !S_ISCHR(devnodes_stat[n].st_mode))
+ return -EINVAL;
+ }
+
+ n++;
+ }
+
+ assert(n == ELEMENTSOF(devnodes_stat));
+
+ r = mount("tmpfs", "/dev", "tmpfs", MS_NOSUID|MS_STRICTATIME, "mode=755");
+ if (r < 0)
+ return m->ignore ? 0 : -errno;
+
+
+ mkdir_p("/dev/pts", 0755);
+
+ r = mount("devpts", "/dev/pts", "devpts", MS_NOSUID|MS_NOEXEC, "newinstance,ptmxmode=0666,mode=620,gid=" STRINGIFY(TTY_GID));
+ if (r < 0)
+ return m->ignore ? 0 : -errno;
+
+ mkdir_p("/dev/shm", 0755);
+
+ r = mount("tmpfs", "/dev/shm", "tmpfs", MS_NOSUID|MS_NODEV|MS_STRICTATIME, "mode=1777");
+ if (r < 0)
+ return m->ignore ? 0 : -errno;
+
+ /* Second: actually create it */
+ n = 0;
+ NULSTR_FOREACH(d, devnodes) {
+ if (devnodes_stat[n].st_rdev == 0)
+ continue;
+
+ r = mknod(d, devnodes_stat[n].st_mode, devnodes_stat[n].st_rdev);
+ if (r < 0)
+ return m->ignore ? 0 : -errno;
+
+ n++;
+ }
+
+ dev_setup(NULL);
+
+ return 0;
+}
+
+static int apply_mount(
+ BindMount *m,
+ const char *tmp_dir,
+ const char *var_tmp_dir) {
+
const char *what;
- char *where;
int r;
- assert(p);
- assert(root_dir);
- assert(inaccessible_dir);
- assert(private_dir);
+ assert(m);
- where = strappend(root_dir, p->path);
- if (!where)
- return -ENOMEM;
+ switch (m->mode) {
- switch (p->mode) {
+ case PRIVATE_DEV:
+ return mount_dev(m);
case INACCESSIBLE:
- what = inaccessible_dir;
- flags |= MS_RDONLY;
+ what = "/run/systemd/inaccessible";
break;
case READONLY:
- flags |= MS_RDONLY;
- /* Fall through */
-
case READWRITE:
- what = p->path;
+ what = m->path;
+ break;
+
+ case PRIVATE_TMP:
+ what = tmp_dir;
break;
- case PRIVATE:
- what = private_dir;
+ case PRIVATE_VAR_TMP:
+ what = var_tmp_dir;
break;
default:
assert_not_reached("Unknown mode");
}
- r = mount(what, where, NULL, MS_BIND|MS_REC, NULL);
- if (r >= 0) {
- log_debug("Successfully mounted %s to %s", what, where);
-
- /* The bind mount will always inherit the original
- * flags. If we want to set any flag we need
- * to do so in a second independent step. */
- if (flags)
- r = mount(NULL, where, NULL, MS_REMOUNT|MS_BIND|MS_REC|flags, NULL);
-
- /* Avoid exponential growth of trees */
- if (r >= 0 && path_equal(p->path, "/"))
- r = mount(NULL, where, NULL, MS_REMOUNT|MS_BIND|flags, NULL);
+ assert(what);
- if (r < 0) {
- r = -errno;
- umount2(where, MNT_DETACH);
- }
- }
+ r = mount(what, m->path, NULL, MS_BIND|MS_REC, NULL);
+ if (r >= 0)
+ log_debug("Successfully mounted %s to %s", what, m->path);
+ else if (m->ignore && errno == ENOENT)
+ r = 0;
- free(where);
return r;
}
-int setup_namespace(
- char **writable,
- char **readable,
- char **inaccessible,
- bool private_tmp,
- unsigned long flags) {
-
- char
- tmp_dir[] = "/tmp/systemd-namespace-XXXXXX",
- root_dir[] = "/tmp/systemd-namespace-XXXXXX/root",
- old_root_dir[] = "/tmp/systemd-namespace-XXXXXX/root/tmp/old-root-XXXXXX",
- inaccessible_dir[] = "/tmp/systemd-namespace-XXXXXX/inaccessible",
- private_dir[] = "/tmp/systemd-namespace-XXXXXX/private";
-
- Path *paths, *p;
- unsigned n;
- bool need_private = false, need_inaccessible = false;
- bool remove_tmp = false, remove_root = false, remove_old_root = false, remove_inaccessible = false, remove_private = false;
+static int make_read_only(BindMount *m) {
int r;
- const char *t;
- n =
- strv_length(writable) +
- strv_length(readable) +
- strv_length(inaccessible) +
- (private_tmp ? 3 : 1);
+ assert(m);
- paths = new(Path, n);
- if (!paths)
- return -ENOMEM;
-
- p = paths;
- if ((r = append_paths(&p, writable, READWRITE)) < 0 ||
- (r = append_paths(&p, readable, READONLY)) < 0 ||
- (r = append_paths(&p, inaccessible, INACCESSIBLE)) < 0)
- goto fail;
+ if (m->mode != INACCESSIBLE && m->mode != READONLY)
+ return 0;
- if (private_tmp) {
- p->path = "/tmp";
- p->mode = PRIVATE;
- p++;
+ r = mount(NULL, m->path, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_REC, NULL);
+ if (r < 0 && !(m->ignore && errno == ENOENT))
+ return -errno;
- p->path = "/var/tmp";
- p->mode = PRIVATE;
- p++;
- }
+ return 0;
+}
- p->path = "/";
- p->mode = READWRITE;
- p++;
+int setup_namespace(
+ char** read_write_dirs,
+ char** read_only_dirs,
+ char** inaccessible_dirs,
+ char* tmp_dir,
+ char* var_tmp_dir,
+ bool private_dev,
+ unsigned mount_flags) {
+
+ BindMount *m, *mounts = NULL;
+ unsigned n;
+ int r = 0;
- assert(paths + n == p);
+ if (mount_flags == 0)
+ mount_flags = MS_SHARED;
- qsort(paths, n, sizeof(Path), path_compare);
- drop_duplicates(paths, &n, &need_inaccessible, &need_private);
+ if (unshare(CLONE_NEWNS) < 0)
+ return -errno;
- if (!mkdtemp(tmp_dir)) {
- r = -errno;
- goto fail;
- }
- remove_tmp = true;
+ n = !!tmp_dir + !!var_tmp_dir +
+ strv_length(read_write_dirs) +
+ strv_length(read_only_dirs) +
+ strv_length(inaccessible_dirs) +
+ private_dev;
- memcpy(root_dir, tmp_dir, sizeof(tmp_dir)-1);
- if (mkdir(root_dir, 0777) < 0) {
- r = -errno;
- goto fail;
- }
- remove_root = true;
+ if (n > 0) {
+ m = mounts = (BindMount *) alloca(n * sizeof(BindMount));
+ r = append_mounts(&m, read_write_dirs, READWRITE);
+ if (r < 0)
+ return r;
- if (need_inaccessible) {
- memcpy(inaccessible_dir, tmp_dir, sizeof(tmp_dir)-1);
- if (mkdir(inaccessible_dir, 0) < 0) {
- r = -errno;
- goto fail;
- }
- remove_inaccessible = true;
- }
+ r = append_mounts(&m, read_only_dirs, READONLY);
+ if (r < 0)
+ return r;
- if (need_private) {
- mode_t u;
+ r = append_mounts(&m, inaccessible_dirs, INACCESSIBLE);
+ if (r < 0)
+ return r;
- memcpy(private_dir, tmp_dir, sizeof(tmp_dir)-1);
+ if (tmp_dir) {
+ m->path = "/tmp";
+ m->mode = PRIVATE_TMP;
+ m++;
+ }
- u = umask(0000);
- if (mkdir(private_dir, 0777 + S_ISVTX) < 0) {
- umask(u);
+ if (var_tmp_dir) {
+ m->path = "/var/tmp";
+ m->mode = PRIVATE_VAR_TMP;
+ m++;
+ }
- r = -errno;
- goto fail;
+ if (private_dev) {
+ m->path = "/dev";
+ m->mode = PRIVATE_DEV;
+ m++;
}
- umask(u);
- remove_private = true;
- }
+ assert(mounts + n == m);
- if (unshare(CLONE_NEWNS) < 0) {
- r = -errno;
- goto fail;
+ qsort(mounts, n, sizeof(BindMount), mount_path_compare);
+ drop_duplicates(mounts, &n);
}
- /* Remount / as SLAVE so that nothing mounted in the namespace
+ /* Remount / as SLAVE so that nothing now mounted in the namespace
shows up in the parent */
- if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) {
- r = -errno;
- goto fail;
+ if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0)
+ return -errno;
+
+ for (m = mounts; m < mounts + n; ++m) {
+ r = apply_mount(m, tmp_dir, var_tmp_dir);
+ if (r < 0)
+ goto fail;
}
- for (p = paths; p < paths + n; p++) {
- r = apply_mount(p, root_dir, inaccessible_dir, private_dir, flags);
+ for (m = mounts; m < mounts + n; ++m) {
+ r = make_read_only(m);
if (r < 0)
- goto undo_mounts;
+ goto fail;
}
- memcpy(old_root_dir, tmp_dir, sizeof(tmp_dir)-1);
- if (!mkdtemp(old_root_dir)) {
+ /* Remount / as the desired mode */
+ if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) {
r = -errno;
- goto undo_mounts;
+ goto fail;
}
- remove_old_root = true;
- if (chdir(root_dir) < 0) {
- r = -errno;
- goto undo_mounts;
+ return 0;
+
+fail:
+ for (m = mounts; m < mounts + n; ++m)
+ if (m->done)
+ umount2(m->path, MNT_DETACH);
+
+ return r;
+}
+
+static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) {
+ _cleanup_free_ char *x = NULL;
+ char bid[SD_ID128_STRING_MAX];
+ sd_id128_t boot_id;
+ int r;
+
+ assert(id);
+ assert(prefix);
+ assert(path);
+
+ /* We include the boot id in the directory so that after a
+ * reboot we can easily identify obsolete directories. */
+
+ r = sd_id128_get_boot(&boot_id);
+ if (r < 0)
+ return r;
+
+ x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX", NULL);
+ if (!x)
+ return -ENOMEM;
+
+ RUN_WITH_UMASK(0077)
+ if (!mkdtemp(x))
+ return -errno;
+
+ RUN_WITH_UMASK(0000) {
+ char *y;
+
+ y = strappenda(x, "/tmp");
+
+ if (mkdir(y, 0777 | S_ISVTX) < 0)
+ return -errno;
}
- if (pivot_root(root_dir, old_root_dir) < 0) {
- r = -errno;
- goto undo_mounts;
+ *path = x;
+ x = NULL;
+
+ return 0;
+}
+
+int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) {
+ char *a, *b;
+ int r;
+
+ assert(id);
+ assert(tmp_dir);
+ assert(var_tmp_dir);
+
+ r = setup_one_tmp_dir(id, "/tmp", &a);
+ if (r < 0)
+ return r;
+
+ r = setup_one_tmp_dir(id, "/var/tmp", &b);
+ if (r < 0) {
+ char *t;
+
+ t = strappenda(a, "/tmp");
+ rmdir(t);
+ rmdir(a);
+
+ free(a);
+ return r;
}
- t = old_root_dir + sizeof(root_dir) - 1;
- if (umount2(t, MNT_DETACH) < 0)
- /* At this point it's too late to turn anything back,
- * since we are already in the new root. */
- return -errno;
+ *tmp_dir = a;
+ *var_tmp_dir = b;
+
+ return 0;
+}
+
+int setup_netns(int netns_storage_socket[2]) {
+ _cleanup_close_ int netns = -1;
+ union {
+ struct cmsghdr cmsghdr;
+ uint8_t buf[CMSG_SPACE(sizeof(int))];
+ } control = {};
+ struct msghdr mh = {
+ .msg_control = &control,
+ .msg_controllen = sizeof(control),
+ };
+ struct cmsghdr *cmsg;
+ int r;
+
+ assert(netns_storage_socket);
+ assert(netns_storage_socket[0] >= 0);
+ assert(netns_storage_socket[1] >= 0);
+
+ /* We use the passed socketpair as a storage buffer for our
+ * namespace reference fd. Whatever process runs this first
+ * shall create a new namespace, all others should just join
+ * it. To serialize that we use a file lock on the socket
+ * pair.
+ *
+ * It's a bit crazy, but hey, works great! */
- if (rmdir(t) < 0)
+ if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0)
return -errno;
- return 0;
+ if (recvmsg(netns_storage_socket[0], &mh, MSG_DONTWAIT|MSG_CMSG_CLOEXEC) < 0) {
+ if (errno != EAGAIN) {
+ r = -errno;
+ goto fail;
+ }
-undo_mounts:
+ /* Nothing stored yet, so let's create a new namespace */
- for (p--; p >= paths; p--) {
- char full_path[PATH_MAX];
+ if (unshare(CLONE_NEWNET) < 0) {
+ r = -errno;
+ goto fail;
+ }
- snprintf(full_path, sizeof(full_path), "%s%s", root_dir, p->path);
- char_array_0(full_path);
+ loopback_setup();
- umount2(full_path, MNT_DETACH);
- }
+ netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY);
+ if (netns < 0) {
+ r = -errno;
+ goto fail;
+ }
-fail:
- if (remove_old_root)
- rmdir(old_root_dir);
+ r = 1;
+ } else {
+ /* Yay, found something, so let's join the namespace */
+
+ for (cmsg = CMSG_FIRSTHDR(&mh); cmsg; cmsg = CMSG_NXTHDR(&mh, cmsg)) {
+ if (cmsg->cmsg_level == SOL_SOCKET && cmsg->cmsg_type == SCM_RIGHTS) {
+ assert(cmsg->cmsg_len == CMSG_LEN(sizeof(int)));
+ netns = *(int*) CMSG_DATA(cmsg);
+ }
+ }
- if (remove_inaccessible)
- rmdir(inaccessible_dir);
+ if (setns(netns, CLONE_NEWNET) < 0) {
+ r = -errno;
+ goto fail;
+ }
- if (remove_private)
- rmdir(private_dir);
+ r = 0;
+ }
- if (remove_root)
- rmdir(root_dir);
+ 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), &netns, sizeof(int));
+ mh.msg_controllen = cmsg->cmsg_len;
- if (remove_tmp)
- rmdir(tmp_dir);
+ if (sendmsg(netns_storage_socket[1], &mh, MSG_DONTWAIT|MSG_NOSIGNAL) < 0) {
+ r = -errno;
+ goto fail;
+ }
- free(paths);
+fail:
+ lockf(netns_storage_socket[0], F_ULOCK, 0);
return r;
}