+/* SPDX-License-Identifier: LGPL-2.1+ */
/***
- This file is part of systemd.
-
Copyright 2010 Lennart Poettering
-
- systemd is free software; you can redistribute it and/or modify it
- under the terms of the GNU Lesser General Public License as published by
- the Free Software Foundation; either version 2.1 of the License, or
- (at your option) any later version.
-
- systemd is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- Lesser General Public License for more details.
-
- You should have received a copy of the GNU Lesser General Public License
- along with systemd; If not, see <http://www.gnu.org/licenses/>.
***/
#include <errno.h>
#include <ftw.h>
#include <stdlib.h>
#include <sys/mount.h>
+//#include <sys/statvfs.h>
#include <unistd.h>
#include "alloc-util.h"
#include "cgroup-util.h"
//#include "dev-setup.h"
//#include "efivars.h"
+//#include "fileio.h"
+#include "fs-util.h"
#include "label.h"
//#include "log.h"
#include "macro.h"
#include "path-util.h"
//#include "set.h"
//#include "smack-util.h"
-#include "string-util.h"
//#include "strv.h"
#include "user-util.h"
//#include "util.h"
#include "virt.h"
+/// Additional includes needed by elogind
+#include "string-util.h"
+
typedef enum MountMode {
- MNT_NONE = 0,
- MNT_FATAL = 1 << 0,
- MNT_IN_CONTAINER = 1 << 1,
+ MNT_NONE = 0,
+ MNT_FATAL = 1 << 0,
+ MNT_IN_CONTAINER = 1 << 1,
+ MNT_CHECK_WRITABLE = 1 << 2,
} MountMode;
typedef struct MountPoint {
* fourth (securityfs) is needed by IMA to load a custom policy. The
* other ones we can delay until SELinux and IMA are loaded. When
* SMACK is enabled we need smackfs, too, so it's a fifth one. */
-#ifdef HAVE_SMACK
+#if ENABLE_SMACK
#define N_EARLY_MOUNT 5
#else
#define N_EARLY_MOUNT 4
NULL, MNT_FATAL|MNT_IN_CONTAINER },
{ "securityfs", "/sys/kernel/security", "securityfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
NULL, MNT_NONE },
-#ifdef HAVE_SMACK
+#if ENABLE_SMACK
{ "smackfs", "/sys/fs/smackfs", "smackfs", "smackfsdef=*", MS_NOSUID|MS_NOEXEC|MS_NODEV,
mac_smack_use, MNT_FATAL },
{ "tmpfs", "/dev/shm", "tmpfs", "mode=1777,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
NULL, MNT_FATAL|MNT_IN_CONTAINER },
{ "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC,
NULL, MNT_IN_CONTAINER },
-#ifdef HAVE_SMACK
+#if ENABLE_SMACK
{ "tmpfs", "/run", "tmpfs", "mode=755,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
mac_smack_use, MNT_FATAL },
#endif
{ "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
NULL, MNT_FATAL|MNT_IN_CONTAINER },
- { "cgroup", "/sys/fs/cgroup", "cgroup", "__DEVEL__sane_behavior", MS_NOSUID|MS_NOEXEC|MS_NODEV,
- cg_is_unified_wanted, MNT_FATAL|MNT_IN_CONTAINER },
#endif // 0
+ { "cgroup2", "/sys/fs/cgroup", "cgroup2", "nsdelegate", MS_NOSUID|MS_NOEXEC|MS_NODEV,
+ cg_is_unified_wanted, MNT_IN_CONTAINER|MNT_CHECK_WRITABLE },
+ { "cgroup2", "/sys/fs/cgroup", "cgroup2", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
+ cg_is_unified_wanted, MNT_IN_CONTAINER|MNT_CHECK_WRITABLE },
{ "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
+ { "cgroup2", "/sys/fs/cgroup/unified", "cgroup2", "nsdelegate", MS_NOSUID|MS_NOEXEC|MS_NODEV,
+ cg_is_hybrid_wanted, MNT_IN_CONTAINER|MNT_CHECK_WRITABLE },
+ { "cgroup2", "/sys/fs/cgroup/unified", "cgroup2", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
+ cg_is_hybrid_wanted, MNT_IN_CONTAINER|MNT_CHECK_WRITABLE },
#if 0 /// UNNEEDED by elogind
{ "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
- cg_is_legacy_wanted, MNT_IN_CONTAINER },
+ cg_is_legacy_wanted, MNT_IN_CONTAINER },
{ "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV,
cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
-#else
- { "cgroup", "/sys/fs/cgroup/elogind", "cgroup", "none,name=elogind,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
- cg_is_legacy_wanted, MNT_IN_CONTAINER },
- { "cgroup", "/sys/fs/cgroup/elogind", "cgroup", "none,name=elogind", MS_NOSUID|MS_NOEXEC|MS_NODEV,
- cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
-#endif // 0
-#if 0 /// UNNEEDED by elogind
{ "pstore", "/sys/fs/pstore", "pstore", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
NULL, MNT_NONE },
-#ifdef ENABLE_EFI
+#if ENABLE_EFI
{ "efivarfs", "/sys/firmware/efi/efivars", "efivarfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
is_efi_boot, MNT_NONE },
#endif
- { "kdbusfs", "/sys/fs/kdbus", "kdbusfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
- is_kdbus_wanted, MNT_IN_CONTAINER },
+ { "bpf", "/sys/fs/bpf", "bpf", "mode=700", MS_NOSUID|MS_NOEXEC|MS_NODEV,
+ NULL, MNT_NONE, },
+#else
+ { "cgroup", "/sys/fs/cgroup/elogind", "cgroup", "none,name=elogind,release_agent="SYSTEMD_CGROUP_AGENT_PATH",xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
+ cg_is_legacy_wanted, MNT_IN_CONTAINER },
+ { "cgroup", "/sys/fs/cgroup/elogind", "cgroup", "none,name=elogind,release_agent="SYSTEMD_CGROUP_AGENT_PATH, MS_NOSUID|MS_NOEXEC|MS_NODEV,
+ cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
#endif // 0
};
#endif // 0
static int mount_one(const MountPoint *p, bool relabel) {
- int r;
+ int r, priority;
assert(p);
+ priority = (p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG;
+
if (p->condition_fn && !p->condition_fn())
return 0;
/* Relabel first, just in case */
if (relabel)
- (void) label_fix(p->where, true, true);
+ (void) label_fix(p->where, LABEL_IGNORE_ENOENT|LABEL_IGNORE_EROFS);
- r = path_is_mount_point(p->where, AT_SYMLINK_FOLLOW);
+ r = path_is_mount_point(p->where, NULL, AT_SYMLINK_FOLLOW);
if (r < 0 && r != -ENOENT) {
- log_full_errno((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, r, "Failed to determine whether %s is a mount point: %m", p->where);
+ log_full_errno(priority, r, "Failed to determine whether %s is a mount point: %m", p->where);
return (p->mode & MNT_FATAL) ? r : 0;
}
if (r > 0)
p->type,
p->flags,
p->options) < 0) {
- log_full_errno((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, errno, "Failed to mount %s at %s: %m", p->type, p->where);
+ log_full_errno(priority, errno, "Failed to mount %s at %s: %m", p->type, p->where);
return (p->mode & MNT_FATAL) ? -errno : 0;
}
/* Relabel again, since we now mounted something fresh here */
if (relabel)
- (void) label_fix(p->where, false, false);
+ (void) label_fix(p->where, 0);
+
+ if (p->mode & MNT_CHECK_WRITABLE) {
+ if (access(p->where, W_OK) < 0) {
+ r = -errno;
+
+ (void) umount(p->where);
+ (void) rmdir(p->where);
+
+ log_full_errno(priority, r, "Mount point %s not writable after mounting: %m", p->where);
+ return (p->mode & MNT_FATAL) ? r : 0;
+ }
+ }
return 1;
}
int mount_cgroup_controllers(char ***join_controllers) {
_cleanup_set_free_free_ Set *controllers = NULL;
+ bool has_argument = !!join_controllers;
int r;
if (!cg_is_legacy_wanted())
/* Mount all available cgroup controllers that are built into the kernel. */
- controllers = set_new(&string_hash_ops);
- if (!controllers)
- return log_oom();
+ if (!has_argument)
+ /* The defaults:
+ * mount "cpu" + "cpuacct" together, and "net_cls" + "net_prio".
+ *
+ * We'd like to add "cpuset" to the mix, but "cpuset" doesn't really
+ * work for groups with no initialized attributes.
+ */
+ join_controllers = (char**[]) {
+ STRV_MAKE("cpu", "cpuacct"),
+ STRV_MAKE("net_cls", "net_prio"),
+ NULL,
+ };
- r = cg_kernel_controllers(controllers);
+ r = cg_kernel_controllers(&controllers);
if (r < 0)
return log_error_errno(r, "Failed to enumerate cgroup controllers: %m");
if (!controller)
break;
- if (join_controllers)
- for (k = join_controllers; *k; k++)
- if (strv_find(*k, controller))
- break;
+ for (k = join_controllers; *k; k++)
+ if (strv_find(*k, controller))
+ break;
- if (k && *k) {
+ if (*k) {
char **i, **j;
for (i = *k, j = *k; *i; i++) {
t = set_remove(controllers, *i);
if (!t) {
- free(*i);
+ if (has_argument)
+ free(*i);
continue;
}
}
options = strv_join(*k, ",");
if (!options)
return log_oom();
- } else {
- options = controller;
- controller = NULL;
- }
+ } else
+ options = TAKE_PTR(controller);
where = strappend("/sys/fs/cgroup/", options);
if (!where)
if (r < 0)
return r;
- if (r > 0 && k && *k) {
+ if (r > 0 && *k) {
char **i;
for (i = *k; *i; i++) {
return 0;
}
-#if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
+#if HAVE_SELINUX || ENABLE_SMACK
static int nftw_cb(
const char *fpath,
const struct stat *sb,
if (_unlikely_(ftwbuf->level == 0))
return FTW_CONTINUE;
- label_fix(fpath, false, false);
+ (void) label_fix(fpath, 0);
/* /run/initramfs is static data and big, no need to
* dynamically relabel its contents at boot... */
return FTW_CONTINUE;
};
+
+static int relabel_cgroup_filesystems(void) {
+ int r;
+ struct statfs st;
+
+ r = cg_all_unified();
+ if (r == 0) {
+ /* Temporarily remount the root cgroup filesystem to give it a proper label. Do this
+ only when the filesystem has been already populated by a previous instance of systemd
+ running from initrd. Otherwise don't remount anything and leave the filesystem read-write
+ for the cgroup filesystems to be mounted inside. */
+ if (statfs("/sys/fs/cgroup", &st) < 0)
+ return log_error_errno(errno, "Failed to determine mount flags for /sys/fs/cgroup: %m");
+
+ if (st.f_flags & ST_RDONLY)
+ (void) mount(NULL, "/sys/fs/cgroup", NULL, MS_REMOUNT, NULL);
+
+ (void) label_fix("/sys/fs/cgroup", 0);
+ (void) nftw("/sys/fs/cgroup", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
+
+ if (st.f_flags & ST_RDONLY)
+ (void) mount(NULL, "/sys/fs/cgroup", NULL, MS_REMOUNT|MS_RDONLY, NULL);
+
+ } else if (r < 0)
+ return log_error_errno(r, "Failed to determine whether we are in all unified mode: %m");
+
+ return 0;
+}
#endif
#endif // 0
int r = 0;
r = mount_points_setup(ELEMENTSOF(mount_table), loaded_policy);
-
if (r < 0)
return r;
#if 0 /// elogind does not control /, /dev, /run and /run/systemd/* are setup elsewhere.
-#if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
+#if HAVE_SELINUX || ENABLE_SMACK
/* Nodes in devtmpfs and /run need to be manually updated for
* the appropriate labels, after mounting. The other virtual
* API file systems like /sys and /proc do not need that, they
before_relabel = now(CLOCK_MONOTONIC);
- nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
- nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
+ (void) nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
+ (void) nftw("/dev/shm", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
+ (void) nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
+
+ r = relabel_cgroup_filesystems();
+ if (r < 0)
+ return r;
after_relabel = now(CLOCK_MONOTONIC);
- log_info("Relabelled /dev and /run in %s.",
+ log_info("Relabelled /dev, /run and /sys/fs/cgroup in %s.",
format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel, 0));
}
#endif
* udevd. */
dev_setup(NULL, UID_INVALID, GID_INVALID);
- /* Mark the root directory as shared in regards to mount
- * propagation. The kernel defaults to "private", but we think
- * it makes more sense to have a default of "shared" so that
- * nspawn and the container tools work out of the box. If
- * specific setups need other settings they can reset the
- * propagation mode to private if needed. */
+ /* Mark the root directory as shared in regards to mount propagation. The kernel defaults to "private", but we
+ * think it makes more sense to have a default of "shared" so that nspawn and the container tools work out of
+ * the box. If specific setups need other settings they can reset the propagation mode to private if
+ * needed. Note that we set this only when we are invoked directly by the kernel. If we are invoked by a
+ * container manager we assume the container manager knows what it is doing (for example, because it set up
+ * some directories with different propagation modes). */
if (detect_container() <= 0)
if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
log_warning_errno(errno, "Failed to set up the root directory for shared mount propagation: %m");
- /* Create a few directories we always want around, Note that
- * sd_booted() checks for /run/systemd/system, so this mkdir
- * really needs to stay for good, otherwise software that
- * copied sd-daemon.c into their sources will misdetect
- * systemd. */
- mkdir_label("/run/systemd", 0755);
- mkdir_label("/run/systemd/system", 0755);
- mkdir_label("/run/systemd/inaccessible", 0000);
#endif // 0
+ /* Create a few directories we always want around, Note that sd_booted() checks for /run/systemd/system, so
+ * this mkdir really needs to stay for good, otherwise software that copied sd-daemon.c into their sources will
+ * misdetect systemd. */
+ (void) mkdir_label("/run/systemd", 0755);
+#if 0 /// Yeah, but elogind is not used with systemd, so this directory would be toxic.
+ (void) mkdir_label("/run/systemd/system", 0755);
+#endif // 0
+
+ /* Set up inaccessible (and empty) file nodes of all types */
+ (void) mkdir_label("/run/systemd/inaccessible", 0000);
+ (void) mknod("/run/systemd/inaccessible/reg", S_IFREG | 0000, 0);
+ (void) mkdir_label("/run/systemd/inaccessible/dir", 0000);
+ (void) mkfifo("/run/systemd/inaccessible/fifo", 0000);
+ (void) mknod("/run/systemd/inaccessible/sock", S_IFSOCK | 0000, 0);
+
+ /* The following two are likely to fail if we lack the privs for it (for example in an userns environment, if
+ * CAP_SYS_MKNOD is missing, or if a device node policy prohibit major/minor of 0 device nodes to be
+ * created). But that's entirely fine. Consumers of these files should carry fallback to use a different node
+ * then, for example /run/systemd/inaccessible/sock, which is close enough in behaviour and semantics for most
+ * uses. */
+ (void) mknod("/run/systemd/inaccessible/chr", S_IFCHR | 0000, makedev(0, 0));
+ (void) mknod("/run/systemd/inaccessible/blk", S_IFBLK | 0000, makedev(0, 0));
return 0;
}