Merge bf05e30acc into bbec1c87d3

Continuation of 4ebbb5bfe8.
Closes #35307.

man/importctl.xml

@@ -128,7 +128,8 @@
 
         <para>If <option>-keep-download=yes</option> is specified the image will be downloaded and stored in
         a read-only subvolume/directory in the image directory that is named after the specified URL and its
-        HTTP etag. A writable snapshot is then taken from this subvolume, and named after the specified local
+        HTTP etag (see <ulink url="https://en.wikipedia.org/wiki/HTTP_ETag">HTTP ETag</ulink> for more
+        information). A writable snapshot is then taken from this subvolume, and named after the specified local
         name. This behavior ensures that creating multiple instances of the same URL is efficient, as
         multiple downloads are not necessary. In order to create only the read-only image, and avoid creating
         its writable snapshot, specify <literal>-</literal> as local name.</para>

man/pam_systemd_loadkey.xml

@@ -28,7 +28,9 @@
     <title>Description</title>
 
     <para><command>pam_systemd_loadkey</command> reads a NUL-separated password list from the kernel keyring,
-    and sets the last password in the list as the PAM authtok.</para>
+    and sets the last password in the list as the PAM authtok, which can be used by e.g.
+    <citerefentry project='man-pages'><refentrytitle>pam_get_authtok</refentrytitle><manvolnum>3</manvolnum></citerefentry>.
+    </para>
 
     <para>The password list is supposed to be stored in the "user" keyring of the root user,
     by an earlier call to

man/run0.xml

@@ -61,7 +61,10 @@
     <literal>systemd-run0</literal> PAM stack.</para>
 
     <para>Note that <command>run0</command> is implemented as an alternative multi-call invocation of
-    <citerefentry><refentrytitle>systemd-run</refentrytitle><manvolnum>1</manvolnum></citerefentry>.</para>
+    <citerefentry><refentrytitle>systemd-run</refentrytitle><manvolnum>1</manvolnum></citerefentry>. That is,
+    <command>run0</command> is a symbolic link to <command>systemd-run</command> executable file, and it
+    behaves as <command>run0</command> if it is invoked through the symbolic link, otherwise behaves as
+    <command>systemd-run</command>.</para>
   </refsect1>
 
   <refsect1>

man/systemd-rfkill.service.xml

@@ -41,8 +41,10 @@
   <refsect1>
     <title>Kernel Command Line</title>
 
-    <para><filename>systemd-rfkill</filename> understands the
+    <para>
-    following kernel command line parameter:</para>
+      <command>systemd-rfkill</command> understands the following kernel command line parameter. See also
+      <citerefentry><refentrytitle>kernel-command-line</refentrytitle><manvolnum>7</manvolnum></citerefentry>.
+    </para>
 
     <variablelist class='kernel-commandline-options'>
       <varlistentry>

man/systemd.exec.xml

@@ -2009,8 +2009,8 @@ BindReadOnlyPaths=/var/lib/systemd</programlisting>
       <varlistentry>
         <term><varname>PrivateUsers=</varname></term>
 
-        <listitem><para>Takes a boolean argument or one of <literal>self</literal> or
+        <listitem><para>Takes a boolean argument or one of <literal>self</literal>, <literal>identity</literal>,
-        <literal>identity</literal>. Defaults to false. If enabled, sets up a new user namespace for the
+        or <literal>full</literal>. Defaults to false. If enabled, sets up a new user namespace for the
         executed processes and configures a user and group mapping. If set to a true value or
         <literal>self</literal>, a minimal user and group mapping is configured that maps the
         <literal>root</literal> user and group as well as the unit's own user and group to themselves and
@@ -2026,6 +2026,10 @@ BindReadOnlyPaths=/var/lib/systemd</programlisting>
         since all UIDs/GIDs are chosen identically it does provide process capability isolation, and hence is
         often a good choice if proper user namespacing with distinct UID maps is not appropriate.</para>
 
+        <para>If the parameter is <literal>full</literal>, user namespacing is set up with an identity
+        mapping for all UIDs/GIDs. Similar to <literal>identity</literal>, this does not provide UID/GID
+        isolation, but it does provide process capability isolation.</para>
+
         <para>If this mode is enabled, all unit processes are run without privileges in the host user
         namespace (regardless if the unit's own user/group is <literal>root</literal> or not). Specifically
         this means that the process will have zero process capabilities on the host's user namespace, but

man/ukify.xml

@@ -394,9 +394,9 @@
           <listitem><para>SBAT metadata associated with the UKI or addon. SBAT policies are useful to revoke
           whole groups of UKIs or addons with a single, static policy update that does not take space in
           DBX/MOKX. If not specified manually, a default metadata entry consisting of
-          <literal>uki,1,UKI,uki,1,https://uapi-group.org/specifications/specs/unified_kernel_image/</literal>
+          <programlisting>uki,1,UKI,uki,1,https://uapi-group.org/specifications/specs/unified_kernel_image/</programlisting>
           for UKIs and
-          <literal>uki-addon,1,UKI Addon,addon,1,https://www.freedesktop.org/software/systemd/man/latest/systemd-stub.html</literal>
+          <programlisting>uki-addon,1,UKI Addon,addon,1,https://www.freedesktop.org/software/systemd/man/latest/systemd-stub.html</programlisting>
           for addons will be used, to ensure it is always possible to revoke them. For more information on
           SBAT see <ulink url="https://github.com/rhboot/shim/blob/main/SBAT.md">Shim documentation</ulink>.
           </para>

src/basic/fileio.c

@@ -289,7 +289,8 @@ int write_string_file_full(
                 const char *fn,
                 const char *line,
                 WriteStringFileFlags flags,
-                const struct timespec *ts) {
+                const struct timespec *ts,
+                const char *label_fn) {
 
         bool call_label_ops_post = false, made_file = false;
         _cleanup_fclose_ FILE *f = NULL;
@@ -321,7 +322,8 @@ int write_string_file_full(
         mode_t mode = write_string_file_flags_to_mode(flags);
 
         if (FLAGS_SET(flags, WRITE_STRING_FILE_LABEL|WRITE_STRING_FILE_CREATE)) {
-                r = label_ops_pre(dir_fd, fn, mode);
+                const char *lookup = label_fn ? label_fn : fn;
+                r = label_ops_pre(dir_fd, lookup, mode);
                 if (r < 0)
                         goto fail;
 

src/basic/fileio.h

@@ -51,12 +51,13 @@ int write_string_stream_full(FILE *f, const char *line, WriteStringFileFlags fla
 static inline int write_string_stream(FILE *f, const char *line, WriteStringFileFlags flags) {
         return write_string_stream_full(f, line, flags, /* ts= */ NULL);
 }
-int write_string_file_full(int dir_fd, const char *fn, const char *line, WriteStringFileFlags flags, const struct timespec *ts);
+
+int write_string_file_full(int dir_fd, const char *fn, const char *line, WriteStringFileFlags flags, const struct timespec *ts, const char *label_fn);
 static inline int write_string_file(const char *fn, const char *line, WriteStringFileFlags flags) {
-        return write_string_file_full(AT_FDCWD, fn, line, flags, /* ts= */ NULL);
+        return write_string_file_full(AT_FDCWD, fn, line, flags, /* ts= */ NULL, /*label_fn=*/ NULL);
 }
 static inline int write_string_file_at(int dir_fd, const char *fn, const char *line, WriteStringFileFlags flags) {
-        return write_string_file_full(dir_fd, fn, line, flags, /* ts= */ NULL);
+        return write_string_file_full(dir_fd, fn, line, flags, /* ts= */ NULL, /*label_fn=*/ NULL);
 }
 int write_string_filef(const char *fn, WriteStringFileFlags flags, const char *format, ...) _printf_(3, 4);
 

src/core/exec-invoke.c

@@ -2103,6 +2103,23 @@ static int setup_private_users(PrivateUsers private_users, uid_t ouid, gid_t ogi
                 uid_map = strdup("0 0 65536\n");
                 if (!uid_map)
                         return -ENOMEM;
+        } else if (private_users == PRIVATE_USERS_FULL) {
+                /* Map all UID/GID from original to new user namespace. We can't use `0 0 UINT32_MAX` because
+                 * this is the same UID/GID map as the init user namespace and there are various applications
+                 * (i.e. systemd's running_in_userns()) that check whether they are in a user namespace by
+                 * comparing uid_map/gid_map to `0 0 UINT32_MAX`. Thus, we still map all UIDs/GIDs but do it
+                 * using two extents to differentiate the new user namespace from the init namespace:
+                 *   0 0 1
+                 *   1 1 UINT32_MAX - 1
+                 *
+                 * Note the kernel defines the UID range between 0 and UINT32_MAX so we map all UIDs even though
+                 * the UID range beyond INT32_MAX (e.g. i.e. the range above the signed 32-bit range) is
+                 * icky. For example, setfsuid() returns the old UID as signed integer. But units can decide to
+                 * use these UIDs/GIDs so we need to map them. */
+                r = asprintf(&uid_map, "0 0 1\n"
+                                       "1 1 " UID_FMT "\n", UINT32_MAX - 1);
+                if (r < 0)
+                        return -ENOMEM;
         /* Can only set up multiple mappings with CAP_SETUID. */
         } else if (have_effective_cap(CAP_SETUID) > 0 && uid != ouid && uid_is_valid(uid)) {
                 r = asprintf(&uid_map,
@@ -2123,6 +2140,11 @@ static int setup_private_users(PrivateUsers private_users, uid_t ouid, gid_t ogi
                 gid_map = strdup("0 0 65536\n");
                 if (!gid_map)
                         return -ENOMEM;
+        } else if (private_users == PRIVATE_USERS_FULL) {
+                r = asprintf(&gid_map, "0 0 1\n"
+                                       "1 1 " UID_FMT "\n", UINT32_MAX - 1);
+                if (r < 0)
+                        return -ENOMEM;
         /* Can only set up multiple mappings with CAP_SETGID. */
         } else if (have_effective_cap(CAP_SETGID) > 0 && gid != ogid && gid_is_valid(gid)) {
                 r = asprintf(&gid_map,

src/core/namespace.c

@@ -3364,6 +3364,7 @@ static const char* const private_users_table[_PRIVATE_USERS_MAX] = {
         [PRIVATE_USERS_NO]       = "no",
         [PRIVATE_USERS_SELF]     = "self",
         [PRIVATE_USERS_IDENTITY] = "identity",
+        [PRIVATE_USERS_FULL]     = "full",
 };
 
 DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(private_users, PrivateUsers, PRIVATE_USERS_SELF);

src/core/namespace.h

@@ -65,6 +65,7 @@ typedef enum PrivateUsers {
         PRIVATE_USERS_NO,
         PRIVATE_USERS_SELF,
         PRIVATE_USERS_IDENTITY,
+        PRIVATE_USERS_FULL,
         _PRIVATE_USERS_MAX,
         _PRIVATE_USERS_INVALID = -EINVAL,
 } PrivateUsers;

src/libsystemd/sd-varlink/sd-varlink.c

@@ -1698,7 +1698,8 @@ _public_ int sd_varlink_get_events(sd_varlink *v) {
                 ret |= EPOLLIN;
 
         if (!v->write_disconnected &&
-            v->output_buffer_size > 0)
+            (v->output_queue ||
+             v->output_buffer_size > 0))
                 ret |= EPOLLOUT;
 
         return ret;

src/sysext/sysext.c

@@ -1,5 +1,6 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
+#include <errno.h>
 #include <fcntl.h>
 #include <getopt.h>
 #include <linux/loop.h>
@@ -45,6 +46,7 @@
 #include "process-util.h"
 #include "rm-rf.h"
 #include "sort-util.h"
+#include "selinux-util.h"
 #include "string-table.h"
 #include "string-util.h"
 #include "terminal-util.h"
@@ -899,6 +901,7 @@ static int resolve_mutable_directory(
         _cleanup_free_ char *path = NULL, *resolved_path = NULL, *dir_name = NULL;
         const char *root = arg_root, *base = MUTABLE_EXTENSIONS_BASE_DIR;
         int r;
+        _cleanup_close_ int atfd = -EBADF;
 
         assert(hierarchy);
         assert(ret_resolved_mutable_directory);
@@ -943,6 +946,14 @@ static int resolve_mutable_directory(
                 r = mkdir_p(path_in_root, 0700);
                 if (r < 0)
                         return log_error_errno(r, "Failed to create a directory '%s': %m", path_in_root);
+
+                atfd = open(path_in_root, O_DIRECTORY|O_CLOEXEC);
+                if (atfd < 0)
+                        return log_error_errno(errno, "Failed to open directory '%s': %m", path_in_root);
+
+                r = mac_selinux_fix_full(atfd, NULL, hierarchy, 0);
+                if (r < 0)
+                        return log_error_errno(r, "Failed to fix SELinux label for '%s': %m", path_in_root);
         }
 
         r = chase(path, root, CHASE_PREFIX_ROOT, &resolved_path, NULL);
@@ -1289,6 +1300,7 @@ static int mount_overlayfs_with_op(
 
         int r;
         const char *top_layer = NULL;
+        _cleanup_close_ int atfd = -EBADF;
 
         assert(op);
         assert(overlay_path);
@@ -1301,10 +1313,28 @@ static int mount_overlayfs_with_op(
         if (r < 0)
                 return log_error_errno(r, "Failed to make directory '%s': %m", meta_path);
 
+        atfd = open(meta_path, O_DIRECTORY|O_CLOEXEC);
+        if (atfd < 0)
+                return log_error_errno(errno, "Failed to open directory '%s': %m", meta_path);
+
+        r = mac_selinux_fix_full(atfd, NULL, op->hierarchy, 0);
+        if (r < 0)
+                return log_error_errno(r, "Failed to fix SELinux label for '%s': %m", meta_path);
+
         if (op->upper_dir && op->work_dir) {
                 r = mkdir_p(op->work_dir, 0700);
                 if (r < 0)
                         return log_error_errno(r, "Failed to make directory '%s': %m", op->work_dir);
+                _cleanup_close_ int dfd = -EBADF;
+
+                dfd = open(op->work_dir, O_DIRECTORY|O_CLOEXEC);
+                if (dfd < 0)
+                        return log_error_errno(errno, "Failed to open directory '%s': %m", op->work_dir);
+
+                r = mac_selinux_fix_full(dfd, NULL, op->hierarchy, 0);
+                if (r < 0)
+                        return log_error_errno(r, "Failed to fix SELinux label for '%s': %m", op->work_dir);
+
                 top_layer = op->upper_dir;
         } else {
                 assert(!strv_isempty(op->lower_dirs));
@@ -1325,7 +1355,7 @@ static int mount_overlayfs_with_op(
         return 0;
 }
 
-static int write_extensions_file(ImageClass image_class, char **extensions, const char *meta_path) {
+static int write_extensions_file(ImageClass image_class, char **extensions, const char *meta_path, const char *hierarchy) {
         _cleanup_free_ char *f = NULL, *buf = NULL;
         int r;
 
@@ -1343,14 +1373,15 @@ static int write_extensions_file(ImageClass image_class, char **extensions, cons
         if (!buf)
                 return log_oom();
 
-        r = write_string_file(f, buf, WRITE_STRING_FILE_CREATE|WRITE_STRING_FILE_MKDIR_0755);
+        const char *hierarchy_path = path_join(hierarchy, image_class_info[image_class].dot_directory_name, image_class_info[image_class].short_identifier_plural);
+        r = write_string_file_full(AT_FDCWD,f, buf, WRITE_STRING_FILE_CREATE|WRITE_STRING_FILE_MKDIR_0755|WRITE_STRING_FILE_LABEL, NULL, hierarchy_path);
         if (r < 0)
                 return log_error_errno(r, "Failed to write extension meta file '%s': %m", f);
 
         return 0;
 }
 
-static int write_dev_file(ImageClass image_class, const char *meta_path, const char *overlay_path) {
+static int write_dev_file(ImageClass image_class, const char *meta_path, const char *overlay_path, const char *hierarchy) {
         _cleanup_free_ char *f = NULL;
         struct stat st;
         int r;
@@ -1372,14 +1403,15 @@ static int write_dev_file(ImageClass image_class, const char *meta_path, const c
         /* Modifying the underlying layers while the overlayfs is mounted is technically undefined, but at
          * least it won't crash or deadlock, as per the kernel docs about overlayfs:
          * https://www.kernel.org/doc/html/latest/filesystems/overlayfs.html#changes-to-underlying-filesystems */
-        r = write_string_file(f, FORMAT_DEVNUM(st.st_dev), WRITE_STRING_FILE_CREATE);
+        const char *hierarchy_path = path_join(hierarchy, image_class_info[image_class].dot_directory_name, image_class_info[image_class].short_identifier_plural);
+        r = write_string_file_full(AT_FDCWD, f, FORMAT_DEVNUM(st.st_dev), WRITE_STRING_FILE_CREATE|WRITE_STRING_FILE_LABEL, NULL, hierarchy_path);
         if (r < 0)
                 return log_error_errno(r, "Failed to write '%s': %m", f);
 
         return 0;
 }
 
-static int write_work_dir_file(ImageClass image_class, const char *meta_path, const char *work_dir) {
+static int write_work_dir_file(ImageClass image_class, const char *meta_path, const char *work_dir, const char* hierarchy) {
         _cleanup_free_ char *escaped_work_dir_in_root = NULL, *f = NULL;
         char *work_dir_in_root = NULL;
         int r;
@@ -1406,7 +1438,8 @@ static int write_work_dir_file(ImageClass image_class, const char *meta_path, co
         escaped_work_dir_in_root = cescape(work_dir_in_root);
         if (!escaped_work_dir_in_root)
                 return log_oom();
-        r = write_string_file(f, escaped_work_dir_in_root, WRITE_STRING_FILE_CREATE);
+        const char *hierarchy_path = path_join(hierarchy, image_class_info[image_class].dot_directory_name, "work_dir");
+        r = write_string_file_full(AT_FDCWD, f, escaped_work_dir_in_root, WRITE_STRING_FILE_CREATE|WRITE_STRING_FILE_LABEL, NULL, hierarchy_path);
         if (r < 0)
                 return log_error_errno(r, "Failed to write '%s': %m", f);
 
@@ -1418,8 +1451,10 @@ static int store_info_in_meta(
                 char **extensions,
                 const char *meta_path,
                 const char *overlay_path,
-                const char *work_dir) {
+                const char *work_dir,
-
+                const char *hierarchy) {
+        _cleanup_free_ char *f = NULL;
+        _cleanup_close_ int atfd = -EBADF;
         int r;
 
         assert(extensions);
@@ -1427,15 +1462,32 @@ static int store_info_in_meta(
         assert(overlay_path);
         /* work_dir may be NULL */
 
-        r = write_extensions_file(image_class, extensions, meta_path);
+        f = path_join(meta_path, image_class_info[image_class].dot_directory_name);
+        if (!f)
+                return log_oom();
+
+        r = mkdir_p(f, 0755);
         if (r < 0)
                 return r;
 
-        r = write_dev_file(image_class, meta_path, overlay_path);
+        atfd = open(f, O_DIRECTORY|O_CLOEXEC);
+        if (atfd < 0)
+                return log_error_errno(errno, "Failed to open directory '%s': %m", f);
+
+        r = mac_selinux_fix_full(atfd, NULL, hierarchy, 0);
+
+        if (r < 0)
+                return log_error_errno(r, "Failed to fix SELinux label for '%s': %m", hierarchy);
+
+        r = write_extensions_file(image_class, extensions, meta_path, hierarchy);
         if (r < 0)
                 return r;
 
-        r = write_work_dir_file(image_class, meta_path, work_dir);
+        r = write_dev_file(image_class, meta_path, overlay_path, hierarchy);
+        if (r < 0)
+                return r;
+
+        r = write_work_dir_file(image_class, meta_path, work_dir, hierarchy);
         if (r < 0)
                 return r;
 
@@ -1501,6 +1553,8 @@ static int merge_hierarchy(
         assert(overlay_path);
         assert(workspace_path);
 
+        mac_selinux_init();
+
         r = determine_used_extensions(hierarchy, paths, &used_paths, &extensions_used);
         if (r < 0)
                 return r;
@@ -1528,7 +1582,7 @@ static int merge_hierarchy(
         if (r < 0)
                 return r;
 
-        r = store_info_in_meta(image_class, extensions, meta_path, overlay_path, op->work_dir);
+        r = store_info_in_meta(image_class, extensions, meta_path, overlay_path, op->work_dir, op->hierarchy);
         if (r < 0)
                 return r;
 

src/update-done/update-done.c

@@ -29,7 +29,7 @@ static int apply_timestamp(const char *path, struct timespec *ts) {
                      timespec_load_nsec(ts)) < 0)
                 return log_oom();
 
-        r = write_string_file_full(AT_FDCWD, path, message, WRITE_STRING_FILE_CREATE|WRITE_STRING_FILE_ATOMIC|WRITE_STRING_FILE_LABEL, ts);
+        r = write_string_file_full(AT_FDCWD, path, message, WRITE_STRING_FILE_CREATE|WRITE_STRING_FILE_ATOMIC|WRITE_STRING_FILE_LABEL, ts, NULL);
         if (r == -EROFS)
                 log_debug_errno(r, "Cannot create \"%s\", file system is read-only.", path);
         else if (r < 0)

test/units/TEST-07-PID1.private-users.sh

@@ -10,3 +10,5 @@ systemd-run -p PrivateUsersEx=self --wait bash -c 'test "$(cat /proc/self/uid_ma
 systemd-run -p PrivateUsersEx=self --wait bash -c 'test "$(cat /proc/self/gid_map)" == "         0          0          1"'
 systemd-run -p PrivateUsersEx=identity --wait bash -c 'test "$(cat /proc/self/uid_map)" == "         0          0      65536"'
 systemd-run -p PrivateUsersEx=identity --wait bash -c 'test "$(cat /proc/self/gid_map)" == "         0          0      65536"'
+systemd-run -p PrivateUsersEx=full --wait bash -c 'test "$(cat /proc/self/uid_map | tr -d "\n")" == "         0          0          1         1          1 4294967294"'
+systemd-run -p PrivateUsersEx=full --wait bash -c 'test "$(cat /proc/self/gid_map | tr -d "\n")" == "         0          0          1         1          1 4294967294"'

test/units/TEST-13-NSPAWN.machined.sh

@@ -22,6 +22,11 @@ trap at_exit EXIT
 
 systemctl service-log-level systemd-machined debug
 systemctl service-log-level systemd-importd debug
+# per request in https://github.com/systemd/systemd/pull/35117
+systemctl edit --runtime --stdin 'systemd-nspawn@.service' --drop-in=debug.conf <<EOF
+[Service]
+Environment=SYSTEMD_LOG_LEVEL=debug
+EOF
 
 # Mount temporary directory over /var/lib/machines to not pollute the image
 mkdir -p /var/lib/machines
@@ -278,13 +283,13 @@ varlinkctl call /run/systemd/machine/io.systemd.Machine io.systemd.Machine.List
 # sending TRAP signal
 rm -f /var/lib/machines/long-running/trap
 varlinkctl call /run/systemd/machine/io.systemd.Machine io.systemd.Machine.Kill '{"name":"long-running", "whom": "leader", "signal": 5}'
-timeout 30 bash -c "until test -e /var/lib/machines/long-running/trap; do sleep .5; done"
+timeout 120 bash -c "until test -e /var/lib/machines/long-running/trap; do sleep .5; done"
 
 # test io.systemd.Machine.Terminate
 long_running_machine_start
 rm -f /var/lib/machines/long-running/terminate
 varlinkctl call /run/systemd/machine/io.systemd.Machine io.systemd.Machine.Terminate '{"name":"long-running"}'
-timeout 10 bash -c "until test -e /var/lib/machines/long-running/terminate; do sleep .5; done"
+timeout 30 bash -c "until test -e /var/lib/machines/long-running/terminate; do sleep .5; done"
 timeout 30 bash -c "while varlinkctl call /run/systemd/machine/io.systemd.Machine io.systemd.Machine.List '{\"name\":\"long-running\"}'; do sleep 0.5; done"
 
 # test io.systemd.Machine.Register
@@ -356,7 +361,7 @@ journalctl --sync
 machinectl terminate container-without-os-release
 machinectl terminate long-running
 # wait for the container being stopped, otherwise acquiring image metadata by io.systemd.MachineImage.List may fail in the below.
-timeout 10 bash -c "while machinectl status long-running &>/dev/null; do sleep .5; done"
+timeout 30 bash -c "while machinectl status long-running &>/dev/null; do sleep .5; done"
 systemctl kill --signal=KILL systemd-nspawn@long-running.service || :
 
 (ip addr show lo | grep -q 192.168.1.100) || ip address add 192.168.1.100/24 dev lo