Several follow-ups for recent SELinux changes (#38161)

Fixes #38157.

.github/workflows/coverage.yml

@@ -25,7 +25,7 @@ jobs:
 
     steps:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
-      - uses: systemd/mkosi@0d1143150835b21c1bfe64428df5f45b558280b1
+      - uses: systemd/mkosi@184472f0f1f831ca29953546ec01fd941ff763a6
 
       # Freeing up disk space with rm -rf can take multiple minutes. Since we don't need the extra free space
       # immediately, we remove the files in the background. However, we first move them to a different location
@@ -83,22 +83,22 @@ jobs:
         run: mkosi summary
 
       - name: Build tools tree
-        run: sudo mkosi -f sandbox -- true
+        run: sudo mkosi -f box -- true
 
       - name: Configure meson
         run: |
-          sudo mkosi sandbox -- \
+          sudo mkosi box -- \
             meson setup \
             --buildtype=debugoptimized \
             build
 
       - name: Build image
-        run: sudo mkosi sandbox -- meson compile -C build mkosi
+        run: sudo mkosi box -- meson compile -C build mkosi
 
       - name: Initial coverage report
         run: |
           sudo mkdir -p build/test/coverage
-          sudo mkosi sandbox -- \
+          sudo mkosi box -- \
             lcov \
             --directory build/mkosi.builddir/arch~rolling~x86-64 \
             --capture \
@@ -115,7 +115,7 @@ jobs:
           # --preserve-env makes sure all the github actions environment variables are propagated which are
           # used in integration-test-wrapper.py to construct the `gh` command line to download the journals
           # of failed tests.
-          sudo --preserve-env mkosi sandbox -- \
+          sudo --preserve-env mkosi box -- \
             env \
             TEST_RUNNER=ubuntu-24.04 \
             meson test \
@@ -147,10 +147,10 @@ jobs:
               lcov_args+=(--add-tracefile "${file}")
           done < <(find build/test/coverage -name "TEST-*.coverage-info")
 
-          sudo mkosi sandbox -- lcov --ignore-errors inconsistent,inconsistent "${lcov_args[@]}" --output-file build/test/coverage/everything.coverage-info
+          sudo mkosi box -- lcov --ignore-errors inconsistent,inconsistent "${lcov_args[@]}" --output-file build/test/coverage/everything.coverage-info
 
       - name: List coverage report
-        run: sudo mkosi sandbox -- lcov --ignore-errors inconsistent,inconsistent --list build/test/coverage/everything.coverage-info
+        run: sudo mkosi box -- lcov --ignore-errors inconsistent,inconsistent --list build/test/coverage/everything.coverage-info
 
       - name: Coveralls
         uses: coverallsapp/github-action@648a8eb78e6d50909eff900e4ec85cab4524a45b

.github/workflows/linter.yml

@@ -38,10 +38,10 @@ jobs:
           LINTER_RULES_PATH: .github/linters
           GITHUB_ACTIONS_CONFIG_FILE: actionlint.yml
 
-      - uses: systemd/mkosi@0d1143150835b21c1bfe64428df5f45b558280b1
+      - uses: systemd/mkosi@184472f0f1f831ca29953546ec01fd941ff763a6
 
       - name: Check that tabs are not used in Python code
-        run: sh -c '! git grep -P "\\t" -- src/basic/generate-bpf-delegate-configs.py src/boot/generate-hwids-section.py src/ukify/ukify.py test/integration-tests/integration-test-wrapper.py'
+        run: sh -c '! git grep -P "\\t" -- src/core/generate-bpf-delegate-configs.py src/boot/generate-hwids-section.py src/ukify/ukify.py test/integration-tests/integration-test-wrapper.py'
 
       - name: Build tools tree
         run: |
@@ -51,29 +51,29 @@ jobs:
           ToolsTreeRelease=rawhide
           EOF
 
-          mkosi -f sandbox -- true
+          mkosi -f box -- true
 
       - name: Run mypy
         run: |
-          mkosi sandbox -- mypy --version
+          mkosi box -- mypy --version
-          mkosi sandbox -- mypy src/basic/generate-bpf-delegate-configs.py src/boot/generate-hwids-section.py src/test/generate-sym-test.py src/ukify/ukify.py test/integration-tests/integration-test-wrapper.py
+          mkosi box -- mypy src/core/generate-bpf-delegate-configs.py src/boot/generate-hwids-section.py src/test/generate-sym-test.py src/ukify/ukify.py test/integration-tests/integration-test-wrapper.py
 
       - name: Run ruff check
         run: |
-          mkosi sandbox -- ruff --version
+          mkosi box -- ruff --version
-          mkosi sandbox -- ruff check src/basic/generate-bpf-delegate-configs.py src/boot/generate-hwids-section.py src/test/generate-sym-test.py src/ukify/ukify.py test/integration-tests/integration-test-wrapper.py
+          mkosi box -- ruff check src/core/generate-bpf-delegate-configs.py src/boot/generate-hwids-section.py src/test/generate-sym-test.py src/ukify/ukify.py test/integration-tests/integration-test-wrapper.py
 
       - name: Run ruff format
         run: |
-          mkosi sandbox -- ruff --version
+          mkosi box -- ruff --version
-          if ! mkosi sandbox -- ruff format --check src/basic/generate-bpf-delegate-configs.py src/boot/generate-hwids-section.py src/test/generate-sym-test.py src/ukify/ukify.py test/integration-tests/integration-test-wrapper.py
+          if ! mkosi box -- ruff format --check src/core/generate-bpf-delegate-configs.py src/boot/generate-hwids-section.py src/test/generate-sym-test.py src/ukify/ukify.py test/integration-tests/integration-test-wrapper.py
           then
               echo "Please run 'ruff format' on the above files or apply the diffs below manually"
-              mkosi sandbox -- ruff format --check --quiet --diff src/basic/generate-bpf-delegate-configs.py src/boot/generate-hwids-section.py src/test/generate-sym-test.py src/ukify/ukify.py test/integration-tests/integration-test-wrapper.py
+              mkosi box -- ruff format --check --quiet --diff src/core/generate-bpf-delegate-configs.py src/boot/generate-hwids-section.py src/test/generate-sym-test.py src/ukify/ukify.py test/integration-tests/integration-test-wrapper.py
           fi
 
       - name: Configure meson
-        run: mkosi sandbox -- env CC=clang CXX=clang++ meson setup -Dlocalegen-path=/usr/bin/locale-gen -Dcompat-mutable-uid-boundaries=true build
+        run: mkosi box -- env CC=clang CXX=clang++ meson setup -Dlocalegen-path=/usr/bin/locale-gen -Dcompat-mutable-uid-boundaries=true build
 
       - name: Run clang-tidy
-        run: mkosi sandbox -- meson test -C build --suite=clang-tidy --print-errorlogs --no-stdsplit
+        run: mkosi box -- meson test -C build --suite=clang-tidy --print-errorlogs --no-stdsplit

.github/workflows/mkosi.yml

@@ -147,7 +147,7 @@ jobs:
 
     steps:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
-      - uses: systemd/mkosi@0d1143150835b21c1bfe64428df5f45b558280b1
+      - uses: systemd/mkosi@184472f0f1f831ca29953546ec01fd941ff763a6
 
       # Freeing up disk space with rm -rf can take multiple minutes. Since we don't need the extra free space
       # immediately, we remove the files in the background. However, we first move them to a different location
@@ -214,21 +214,21 @@ jobs:
         run: mkosi summary
 
       - name: Build tools tree
-        run: sudo mkosi -f sandbox -- true
+        run: sudo mkosi -f box -- true
 
       - name: Configure meson
         run: |
           # /usr/sbin/bpftool is completely broken inside containers on Ubuntu which makes meson blow up so
           # disable the bpf-framework stuff to avoid the issue.
           # TODO: Drop when we move off Ubuntu Noble as this will be fixed in the next Ubuntu LTS release.
-          sudo mkosi sandbox -- \
+          sudo mkosi box -- \
             meson setup \
             --buildtype=debugoptimized \
             -Dbpf-framework=disabled \
             build
 
       - name: Build image
-        run: sudo mkosi sandbox -- meson compile -C build mkosi
+        run: sudo mkosi box -- meson compile -C build mkosi
 
       - name: Make sure sources weren't polluted by package build scripts
         run: |
@@ -243,7 +243,7 @@ jobs:
 
       - name: Run integration tests
         run: |
-          if [[ "$(sudo mkosi sandbox -- meson test --help)" == *"--max-lines"* ]]; then
+          if [[ "$(sudo mkosi box -- meson test --help)" == *"--max-lines"* ]]; then
               MAX_LINES=(--max-lines 300)
           else
               MAX_LINES=()
@@ -256,7 +256,7 @@ jobs:
           # --preserve-env makes sure all the github actions environment variables are propagated which are
           # used in integration-test-wrapper.py to construct the `gh` command line to download the journals
           # of failed tests.
-          sudo --preserve-env mkosi sandbox -- \
+          sudo --preserve-env mkosi box -- \
             env \
             TEST_PREFER_QEMU=${{ matrix.vm }} \
             TEST_NO_QEMU=${{ matrix.no_qemu }} \

NEWS

@@ -135,6 +135,9 @@ CHANGES WITH 258 in spe:
           incompatible with PrivateDevices= and resulted in automatic extension
           of the DeviceAllow= list. The latter behaviour has been removed.
 
+        * The command 'journalctl --follow' now exits with success on
+          SIGTERM/SIGINT, or its pipe STDOUT is disconnected.
+
         Announcements of Future Feature Removals:
 
         * Support for System V service scripts is deprecated and will be

docs/ARCHITECTURE.md

@@ -20,6 +20,9 @@ There are many, and more are constantly added, so we will not enumerate them all
 
 The code that is shared between components is split into a few directories, each with a different purpose:
 
+- 'src/include/uapi/' contains copy of kernel headers we use.
+  'src/include/override/' contains wrappers for libc and kernel headers, to provide several missing symbols.
+
 - `src/basic/` and `src/fundamental/` — those directories contain code primitives that are used by all other code.
   `src/fundamental/` is stricter, because it used for EFI and user-space code, while `src/basic/` is only used for user-space code.
   The code in `src/fundamental/` cannot depend on any other code in the tree, and `src/basic/` can depend only on itself and `src/fundamental/`.
@@ -43,6 +46,12 @@ Any code that is used only for EFI goes under `src/boot/efi/`, and `src/fundamen
 
 To summarize:
 
+`src/include/uapi/`
+- copy of kernel headers
+
+`src/include/override/`
+- wrappers for libc and kernel headers
+
 `src/fundamental/`
 - may be used by all code in the tree
 - may not use any code outside of `src/fundamental/`

docs/HACKING.md

@@ -40,16 +40,16 @@ Then, you can build, run and test systemd executables as follows:
 
 ```sh
 $ mkosi -f genkey                                  # Generate signing keys once.
-$ mkosi -f sandbox -- meson setup -Dbpf-framework=disabled build # bpftool detection inside mkosi sandbox is broken on Ubuntu Noble and older
+$ mkosi -f box -- meson setup -Dbpf-framework=disabled build # bpftool detection inside mkosi box is broken on Ubuntu Noble and older
-$ mkosi -f sandbox -- meson compile -C build
+$ mkosi -f box -- meson compile -C build
-$ mkosi -f sandbox -- build/systemctl --version
+$ mkosi -f box -- build/systemctl --version
-$ mkosi -f sandbox -- meson test -C build          # Run the unit tests
+$ mkosi -f box -- meson test -C build          # Run the unit tests
 ```
 
 To build and boot an OS image with the latest systemd installed:
 
 ```sh
-$ mkosi -f sandbox -- meson compile -C build mkosi # (re-)build the OS image
+$ mkosi -f box -- meson compile -C build mkosi # (re-)build the OS image
 $ mkosi boot                                       # Boot the image with systemd-nspawn.
 $ mkosi vm                                         # Boot the image with qemu.
 ```
@@ -65,8 +65,8 @@ $ cd systemd
 $ git checkout -b <BRANCH>                         # where BRANCH is the name of the branch
 $ $EDITOR src/core/main.c                          # or wherever you'd like to make your changes
 $ mkosi -f genkey                                  # Generate signing keys once.
-$ mkosi -f sandbox -- meson setup build            # Set up meson
+$ mkosi -f box -- meson setup build            # Set up meson
-$ mkosi -f sandbox -- meson compile -C build mkosi # (re-)build the test image
+$ mkosi -f box -- meson compile -C build mkosi # (re-)build the test image
 $ mkosi vm                                         # Boot the image in qemu
 $ git add -p                                       # interactively put together your patch
 $ git commit                                       # commit it
@@ -85,7 +85,7 @@ not required to write basic patches.
 ## Building the OS image without a tools tree
 
 By default, `mkosi` will first build a tools tree and use it build the image and
-provide the environment for `mkosi sandbox`. To disable the tools tree and use
+provide the environment for `mkosi box`. To disable the tools tree and use
 binaries from your host instead, write the following to `mkosi/mkosi.local.conf`:
 
 ```conf
@@ -311,7 +311,7 @@ To debug systemd-boot in an IDE such as VSCode we can use a launch configuration
 right in your editor of choice (with the right plugin installed). When using mkosi, we can run clangd in the
 mkosi tools tree to avoid needing to install clangd on the host machine.
 
-All that is required is to run `mkosi -f sandbox true` once to make sure the tools tree is available and to modify
+All that is required is to run `mkosi -f box true` once to make sure the tools tree is available and to modify
 the path of the clangd binary used by your editor to the `mkosi.clangd` script included in the systemd repository.
 For example, for VScode, you'd have to add the following to the VSCode workspace settings of the systemd repository:
 
@@ -329,7 +329,7 @@ When using clangd, it's recommended to setup the build directory containing the
 compilation database used by clangd to use clang as the compiler as well:
 
 ```sh
-$ mkosi sandbox -- env CC=clang CXX=clang++ meson setup build
+$ mkosi box -- env CC=clang CXX=clang++ meson setup build
 ```
 
 Additionally, the `gensources` target can be used to make sure all generated
@@ -337,5 +337,5 @@ sources are generated to avoid clangd complaining that these source files don't
 exist.
 
 ```sh
-$ mkosi sandbox -- ninja -C build gensources
+$ mkosi box -- ninja -C build gensources
 ```

man/meson.build

@@ -20,9 +20,10 @@ xsltproc_flags = [
         '--stringparam', 'man.copyright.section.enabled', '0',
         '--stringparam', 'systemd.version', '@0@'.format(meson.project_version()),
         '--path',
-        '@0@:@1@:@2@'.format(meson.current_build_dir(),
+        '@0@:@1@:@2@:@3@'.format(meson.current_build_dir(),
                              meson.current_source_dir(),
-                             libshared_build_dir)]
+                             libshared_build_dir,
+                             libcore_build_dir)]
 
 custom_man_xsl = files('custom-man.xsl')
 custom_html_xsl = files('custom-html.xsl')
@@ -35,17 +36,6 @@ custom_entities_ent = custom_target(
 
 man_page_depends += custom_entities_ent
 
-generate_bpf_delegate_configs = find_program('../src/basic/generate-bpf-delegate-configs.py')
-
-bpf_delegate_xml = custom_target(
-        input : files('../src/basic/include/linux/bpf.h'),
-        output : 'bpf-delegate.xml',
-        command : [generate_bpf_delegate_configs,
-                   'doc',
-                   '@INPUT@'],
-        capture : true)
-man_page_depends += bpf_delegate_xml
-
 man_pages = []
 html_pages = []
 source_xml_files = []

man/systemd.exec.xml

@@ -2567,7 +2567,7 @@ RestrictNamespaces=~cgroup net</programlisting>
         available to the unit's processes. When mounting the BPF filesystem with the fsopen() API, four mount
         options can be specified to set a list of BPF commands, maps, programs and attachment types that are
         allowed to be used. Processes needs to get a file descriptor for the bpffs mountpoint and use that to
-        get a token which will enable for that user namespace the BPF functionalities choosen upon bpffs mount.
+        get a token which will enable for that user namespace the BPF functionalities chosen upon bpffs mount.
         A more detailed explanation of the feature can be found in this
         <ulink url="https://lwn.net/Articles/947173/">LWN post</ulink>.</para>
 

meson.build

@@ -580,7 +580,6 @@ foreach ident : [
         ['fsconfig',          '''#include <sys/mount.h>'''],    # since glibc-2.36
         ['fsmount',           '''#include <sys/mount.h>'''],    # since glibc-2.36
         ['fsopen',            '''#include <sys/mount.h>'''],    # since glibc-2.36
-        ['fspick',            '''#include <sys/mount.h>'''],    # since glibc-2.36
         ['mount_setattr',     '''#include <sys/mount.h>'''],    # since glibc-2.36
         ['move_mount',        '''#include <sys/mount.h>'''],    # since glibc-2.36
         ['open_tree',         '''#include <sys/mount.h>'''],    # since glibc-2.36
@@ -591,15 +590,17 @@ foreach ident : [
         ['ioprio_get',        '''#include <sched.h>'''],        # no known header declares ioprio_get
         ['ioprio_set',        '''#include <sched.h>'''],        # no known header declares ioprio_set
         ['rt_tgsigqueueinfo', '''#include <signal.h>'''],       # no known header declares rt_tgsigqueueinfo
+        ['open_tree_attr',    '''#include <sys/mount.h>'''],    # no known header declares open_tree_attr
         ['quotactl_fd',       '''#include <sys/quota.h>'''],    # no known header declares quotactl_fd
         ['fchmodat2',         '''#include <sys/stat.h>'''],     # no known header declares fchmodat2
         ['bpf',               '''#include <sys/syscall.h>'''],  # no known header declares bpf
         ['kcmp',              '''#include <sys/syscall.h>'''],  # no known header declares kcmp
         ['keyctl',            '''#include <sys/syscall.h>'''],  # no known header declares keyctl
-        ['pivot_root',        '''#include <sys/syscall.h>'''],  # no known header declares pivot_root
+        ['add_key',           '''#include <sys/syscall.h>'''],  # no known header declares add_key
+        ['request_key',       '''#include <sys/syscall.h>'''],  # no known header declares request_key
         ['setxattrat',        '''#include <sys/xattr.h>'''],    # no known header declares setxattrat
         ['removexattrat',     '''#include <sys/xattr.h>'''],    # no known header declares removexattrat
-        ['open_tree_attr',    '''#include <sys/mount.h>'''],    # no known header declares open_tree_attr
+        ['pivot_root',        '''#include <unistd.h>'''],       # no known header declares pivot_root
 ]
 
         have = cc.has_function(ident[0], prefix : ident[1], args : '-D_GNU_SOURCE')
@@ -2008,16 +2009,33 @@ dbus_programs = []
 # A list of boot stubs. Required for testing of ukify.
 boot_stubs = []
 
+# This is similar to system_includes below, but for passing custom_target().
+system_include_args = [
+        '-isystem', meson.project_build_root()  / 'src/include/override',
+        '-isystem', meson.project_source_root() / 'src/include/override',
+        '-isystem', meson.project_build_root()  / 'src/include/uapi',
+        '-isystem', meson.project_source_root() / 'src/include/uapi',
+]
+
+system_includes = [
+        include_directories(
+                # gcc(1) says
+                # "Directories specified with -isystem options are scanned in left-to-right order",
+                # and meson puts the directories in the reversed order. Hence, a directory with a lower
+                # priority must be listed earlier.
+                'src/include/uapi',
+                'src/include/override',
+                is_system : true,
+        ),
+]
+
 basic_includes = [
         include_directories(
                 'src/basic',
                 'src/fundamental',
                 'src/systemd',
         ),
-        include_directories(
+        system_includes,
-                'src/basic/include',
-                is_system : true,
-        ),
         version_include,
 ]
 
@@ -2039,6 +2057,8 @@ includes = [libsystemd_includes, include_directories('src/shared')]
 
 subdir('po')
 subdir('catalog')
+subdir('src/include')
+subdir('src/libc')
 subdir('src/fundamental')
 subdir('src/basic')
 subdir('src/libsystemd')
@@ -2054,7 +2074,8 @@ libsystemd = shared_library(
                      # Make sure our library is never deleted from memory, so that our open logging fds don't leak on dlopen/dlclose cycles.
                      '-z', 'nodelete',
                      '-Wl,--version-script=' + libsystemd_sym_path],
-        link_with : [libbasic_static],
+        link_with : [libc_wrapper_static,
+                     libbasic_static],
         link_whole : [libsystemd_static],
         dependencies : [librt,
                         threads,
@@ -2067,6 +2088,7 @@ libsystemd = shared_library(
 if static_libsystemd != 'false'
         install_libsystemd_static = static_library(
                 'systemd',
+                libc_wrapper_sources,
                 libsystemd_sources,
                 basic_sources,
                 fundamental_sources,
@@ -2112,6 +2134,7 @@ libudev = shared_library(
 if static_libudev != 'false'
         install_libudev_static = static_library(
                 'udev',
+                libc_wrapper_sources,
                 basic_sources,
                 fundamental_sources,
                 shared_sources,
@@ -2216,6 +2239,7 @@ nss_template = {
         # Note that we link NSS modules with '-z nodelete' so that mempools never get orphaned
         'link_args' : ['-z', 'nodelete'],
         'link_with' : [
+                libc_wrapper_static,
                 libsystemd_static,
                 libshared_static,
                 libbasic_static,

mkosi/mkosi.clangd

@@ -9,7 +9,7 @@ else
 fi
 
 exec "${SPAWN[@]}" \
-    mkosi sandbox -- \
+    mkosi box -- \
         clangd \
             --compile-commands-dir=build \
             --path-mappings="\

mkosi/mkosi.conf

@@ -1,7 +1,7 @@
 # SPDX-License-Identifier: LGPL-2.1-or-later
 
 [Config]
-MinimumVersion=commit:0d1143150835b21c1bfe64428df5f45b558280b1
+MinimumVersion=commit:184472f0f1f831ca29953546ec01fd941ff763a6
 Dependencies=
         exitrd
         initrd

mkosi/mkosi.conf.d/arch/mkosi.conf.d/pkgenv.conf

@@ -1,8 +1,5 @@
 # SPDX-License-Identifier: LGPL-2.1-or-later
 
-[Match]
-Distribution=arch
-
 [Build]
 Environment=
         GIT_URL=https://gitlab.archlinux.org/archlinux/packaging/packages/systemd.git

mkosi/mkosi.conf.d/centos-fedora/mkosi.conf

@@ -27,6 +27,8 @@ Packages=
         cryptsetup
         device-mapper-event
         device-mapper-multipath
+        dfuzzer
+        erofs-utils
         git-core
         glibc-langpack-de
         glibc-langpack-en
@@ -38,6 +40,7 @@ Packages=
         iputils
         iscsi-initiator-utils
         kernel-core
+        knot
         libcap-ng-utils
         man-db
         nmap-ncat
@@ -52,6 +55,7 @@ Packages=
         python3-pexpect
         # needed to upgrade and downgrade systemd-ukify in tests
         python3-zstd
+        qrencode
         quota
         rpm
         softhsm

mkosi/mkosi.conf.d/centos-fedora/mkosi.conf.d/efi.conf

@@ -0,0 +1,11 @@
+# SPDX-License-Identifier: LGPL-2.1-or-later
+
+[Match]
+Architecture=uefi
+
+[Content]
+Packages=
+        sbsigntools
+
+VolatilePackages=
+        systemd-boot

mkosi/mkosi.conf.d/centos-fedora/mkosi.conf.d/pkgenv.conf

@@ -1,9 +1,5 @@
 # SPDX-License-Identifier: LGPL-2.1-or-later
 
-[Match]
-Distribution=|centos
-Distribution=|fedora
-
 [Build]
 Environment=
         GIT_URL=https://src.fedoraproject.org/rpms/systemd.git

mkosi/mkosi.conf.d/centos/mkosi.conf

@@ -5,6 +5,7 @@ Distribution=centos
 
 [Distribution]
 Release=10
+Repositories=epel,epel-next
 
 [Build]
 Environment=

mkosi/mkosi.conf.d/centos/mkosi.conf.d/10-epel.conf

@@ -1,9 +0,0 @@
-# SPDX-License-Identifier: LGPL-2.1-or-later
-
-[Match]
-Release=9
-
-[Distribution]
-Repositories=
-        epel
-        epel-next

mkosi/mkosi.conf.d/centos/mkosi.conf.d/20-efi.conf

@@ -1,18 +0,0 @@
-# SPDX-License-Identifier: LGPL-2.1-or-later
-
-[Match]
-Repositories=epel
-Architecture=|x86
-Architecture=|x86-64
-Architecture=|arm
-Architecture=|arm64
-Architecture=|riscv32
-Architecture=|riscv64
-Architecture=|loongarch64
-
-[Content]
-Packages=
-        sbsigntools
-
-VolatilePackages=
-        systemd-boot

mkosi/mkosi.conf.d/centos/mkosi.conf.d/20-epel-packages.conf

@@ -1,12 +0,0 @@
-# SPDX-License-Identifier: LGPL-2.1-or-later
-
-[Match]
-Repositories=epel
-
-[Content]
-Packages=
-        dfuzzer
-        dhcp-server
-        erofs-utils
-        knot
-        qrencode

mkosi/mkosi.conf.d/debian-ubuntu/mkosi.conf.d/efi-debug.conf

@@ -4,13 +4,7 @@
 Environment=WITH_DEBUG=1
 
 [Match]
-Architecture=|x86
+Architecture=uefi
-Architecture=|x86-64
-Architecture=|arm
-Architecture=|arm64
-Architecture=|riscv32
-Architecture=|riscv64
-Architecture=|loongarch64
 
 [Content]
 VolatilePackages=

mkosi/mkosi.conf.d/debian-ubuntu/mkosi.conf.d/efi.conf

@@ -2,13 +2,7 @@
 # sbsigntool exists only on UEFI architectures
 
 [Match]
-Architecture=|x86
+Architecture=uefi
-Architecture=|x86-64
-Architecture=|arm
-Architecture=|arm64
-Architecture=|riscv32
-Architecture=|riscv64
-Architecture=|loongarch64
 
 [Content]
 Packages=

mkosi/mkosi.conf.d/debian-ubuntu/mkosi.conf.d/pkgenv.conf

@@ -1,9 +1,5 @@
 # SPDX-License-Identifier: LGPL-2.1-or-later
 
-[Match]
-Distribution=|debian
-Distribution=|ubuntu
-
 [Build]
 Environment=
         GIT_URL=https://salsa.debian.org/systemd-team/systemd.git

mkosi/mkosi.conf.d/fedora/mkosi.conf

@@ -10,15 +10,11 @@ Release=rawhide
 Packages=
         btrfs-progs
         compsize
-        dfuzzer
         dhcp-server
         dnf5
-        erofs-utils
         f2fs-tools
         # Required for systemd-networkd-tests.py (netdevsim and sch_xxx modules)
         kernel-modules-extra
         kernel-modules-internal
-        knot
-        qrencode
         rpmautospec
         scsi-target-utils

mkosi/mkosi.conf.d/fedora/mkosi.conf.d/efi.conf

@@ -1,18 +0,0 @@
-# SPDX-License-Identifier: LGPL-2.1-or-later
-
-[Match]
-Distribution=fedora
-Architecture=|x86
-Architecture=|x86-64
-Architecture=|arm
-Architecture=|arm64
-Architecture=|riscv32
-Architecture=|riscv64
-Architecture=|loongarch64
-
-[Content]
-Packages=
-        sbsigntools
-
-VolatilePackages=
-        systemd-boot

mkosi/mkosi.conf.d/opensuse/mkosi.conf.d/efi-debug.conf

@@ -4,13 +4,7 @@
 Environment=WITH_DEBUG=1
 
 [Match]
-Architecture=|x86
+Architecture=uefi
-Architecture=|x86-64
-Architecture=|arm
-Architecture=|arm64
-Architecture=|riscv32
-Architecture=|riscv64
-Architecture=|loongarch64
 
 [Content]
 VolatilePackages=

mkosi/mkosi.conf.d/opensuse/mkosi.conf.d/efi.conf

@@ -1,14 +1,7 @@
 # SPDX-License-Identifier: LGPL-2.1-or-later
 
 [Match]
-Distribution=opensuse
+Architecture=uefi
-Architecture=|x86
-Architecture=|x86-64
-Architecture=|arm
-Architecture=|arm64
-Architecture=|riscv32
-Architecture=|riscv64
-Architecture=|loongarch64
 
 [Content]
 VolatilePackages=

mkosi/mkosi.conf.d/opensuse/mkosi.conf.d/pkgenv.conf

@@ -1,8 +1,5 @@
 # SPDX-License-Identifier: LGPL-2.1-or-later
 
-[Match]
-Distribution=opensuse
-
 [Build]
 Environment=
         GIT_URL=https://github.com/bmwiedemann/openSUSE

src/basic/cgroup-util.c

@@ -1,5 +1,7 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
+#include <linux/fs.h>
+#include <linux/magic.h>
 #include <signal.h>
 #include <stdlib.h>
 #include <sys/xattr.h>
@@ -18,8 +20,6 @@
 #include "fs-util.h"
 #include "log.h"
 #include "login-util.h"
-#include "missing_fs.h"
-#include "missing_magic.h"
 #include "parse-util.h"
 #include "path-util.h"
 #include "pidref.h"

src/basic/chase.c

@@ -126,10 +126,6 @@ int chaseat(int dir_fd, const char *path, ChaseFlags flags, char **ret_path, int
         assert(!FLAGS_SET(flags, CHASE_STEP|CHASE_EXTRACT_FILENAME));
         assert(dir_fd >= 0 || dir_fd == AT_FDCWD);
 
-        /* Either the file may be missing, or we return an fd to the final object, but both make no sense */
-        if (FLAGS_SET(flags, CHASE_NONEXISTENT))
-                assert(!ret_fd);
-
         if (FLAGS_SET(flags, CHASE_STEP))
                 assert(!ret_fd);
 
@@ -552,11 +548,13 @@ int chaseat(int dir_fd, const char *path, ChaseFlags flags, char **ret_path, int
         }
 
         if (ret_fd) {
-                /* Return the O_PATH fd we currently are looking to the caller. It can translate it to a
+                if (exists) {
-                 * proper fd by opening /proc/self/fd/xyz. */
+                        /* Return the O_PATH fd we currently are looking to the caller. It can translate it
-
+                         * to a proper fd by opening /proc/self/fd/xyz. */
                         assert(fd >= 0);
                         *ret_fd = TAKE_FD(fd);
+                } else
+                        *ret_fd = -EBADF;
         }
 
         if (FLAGS_SET(flags, CHASE_STEP))

src/basic/check-filesystems.sh

@@ -3,19 +3,14 @@
 set -eu
 set -o pipefail
 
-cpp="$1"
+cpp="${1:?}"
-filesystems_gperf="$2"
+filesystems_gperf="${2:?}"
 shift 2
 
-includes=""
-for i in "$@"; do
-    includes="$includes -include $i"
-done
-
 error=false
 
 # shellcheck disable=SC2086
-for fs in $($cpp -dM $includes - </dev/null | \
+for fs in $($cpp -dM -include linux/magic.h "$@" - </dev/null | \
             grep -E '_MAGIC' | \
             grep -vE 'LINUX_MAGIC' | \
             awk '/^#define[ \t]+[A-Z0-9_]+MAGIC[ \t]+/ { print $2; }'); do

src/basic/conf-files.c

@@ -20,316 +20,24 @@
 #include "string-util.h"
 #include "strv.h"
 
-static int files_add(
+ConfFile* conf_file_free(ConfFile *c) {
-                DIR *dir,
+        if (!c)
-                const char *dirpath,
+                return NULL;
-                int rfd,
-                const char *root, /* for logging, can be NULL */
-                Hashmap **files,
-                Set **masked,
-                const char *suffix,
-                ConfFilesFlags flags) {
 
-        int r;
+        free(c->name);
+        free(c->result);
+        free(c->original_path);
+        free(c->resolved_path);
+        safe_close(c->fd);
 
-        assert(dir);
+        return mfree(c);
-        assert(dirpath);
-        assert(rfd >= 0 || rfd == AT_FDCWD);
-        assert(files);
-        assert(masked);
-
-        root = strempty(root);
-
-        FOREACH_DIRENT(de, dir, return log_debug_errno(errno, "Failed to read directory '%s/%s': %m",
-                                                       root, skip_leading_slash(dirpath))) {
-
-                /* Does this match the suffix? */
-                if (suffix && !endswith(de->d_name, suffix))
-                        continue;
-
-                /* Has this file already been found in an earlier directory? */
-                if (hashmap_contains(*files, de->d_name)) {
-                        log_debug("Skipping overridden file '%s/%s/%s'.",
-                                  root, skip_leading_slash(dirpath), de->d_name);
-                        continue;
 }
 
-                /* Has this been masked in an earlier directory? */
+void conf_file_free_many(ConfFile **array, size_t n) {
-                if ((flags & CONF_FILES_FILTER_MASKED) != 0 && set_contains(*masked, de->d_name)) {
+        FOREACH_ARRAY(i, array, n)
-                        log_debug("File '%s/%s/%s' is masked by previous entry.",
+                conf_file_free(*i);
-                                  root, skip_leading_slash(dirpath), de->d_name);
-                        continue;
-                }
 
-                _cleanup_free_ char *p = path_join(dirpath, de->d_name);
+        free(array);
-                if (!p)
-                        return log_oom_debug();
-
-                _cleanup_free_ char *resolved_path = NULL;
-                bool need_stat = (flags & (CONF_FILES_FILTER_MASKED | CONF_FILES_REGULAR | CONF_FILES_DIRECTORY | CONF_FILES_EXECUTABLE)) != 0;
-                struct stat st;
-
-                if (!need_stat || FLAGS_SET(flags, CONF_FILES_FILTER_MASKED_BY_SYMLINK)) {
-
-                        /* Even if no verification is requested, let's unconditionally call chaseat(),
-                         * to drop unsafe symlinks. */
-
-                        r = chaseat(rfd, p, CHASE_AT_RESOLVE_IN_ROOT | CHASE_NONEXISTENT, &resolved_path, /* ret_fd = */ NULL);
-                        if (r < 0) {
-                                log_debug_errno(r, "Failed to chase '%s/%s', ignoring: %m",
-                                                root, skip_leading_slash(p));
-                                continue;
-                        }
-                        if (r == 0 && FLAGS_SET(flags, CONF_FILES_FILTER_MASKED_BY_SYMLINK)) {
-
-                                /* If the path points to /dev/null in a image or so, then the device node may not exist. */
-                                if (path_equal(skip_leading_slash(resolved_path), "dev/null")) {
-                                        /* Mark this one as masked */
-                                        r = set_put_strdup(masked, de->d_name);
-                                        if (r < 0)
-                                                return log_oom_debug();
-
-                                        log_debug("File '%s/%s' is a mask (symlink to /dev/null).",
-                                                  root, skip_leading_slash(p));
-                                        continue;
-                                }
-
-                                /* If the flag is set, we need to have stat, hence, skip the entry. */
-                                log_debug_errno(SYNTHETIC_ERRNO(ENOENT), "Failed to chase '%s/%s', ignoring: %m",
-                                                root, skip_leading_slash(p));
-                                continue;
-                        }
-
-                        if (need_stat && fstatat(rfd, resolved_path, &st, AT_SYMLINK_NOFOLLOW) < 0) {
-                                log_debug_errno(errno, "Failed to stat '%s/%s', ignoring: %m",
-                                                root, skip_leading_slash(p));
-                                continue;
-                        }
-
-                } else {
-                        r = chase_and_statat(rfd, p, CHASE_AT_RESOLVE_IN_ROOT, &resolved_path, &st);
-                        if (r < 0) {
-                                log_debug_errno(r, "Failed to chase and stat '%s/%s', ignoring: %m",
-                                                root, skip_leading_slash(p));
-                                continue;
-                        }
-                }
-
-                /* Is this a masking entry? */
-                if (FLAGS_SET(flags, CONF_FILES_FILTER_MASKED_BY_SYMLINK) && stat_may_be_dev_null(&st)) {
-                        /* Mark this one as masked */
-                        r = set_put_strdup(masked, de->d_name);
-                        if (r < 0)
-                                return log_oom_debug();
-
-                        log_debug("File '%s/%s' is a mask (symlink to /dev/null).", root, skip_leading_slash(p));
-                        continue;
-                }
-
-                if (FLAGS_SET(flags, CONF_FILES_FILTER_MASKED_BY_EMPTY) && stat_is_empty(&st)) {
-                        /* Mark this one as masked */
-                        r = set_put_strdup(masked, de->d_name);
-                        if (r < 0)
-                                return log_oom_debug();
-
-                        log_debug("File '%s/%s' is a mask (an empty file).", root, skip_leading_slash(p));
-                        continue;
-                }
-
-                if (FLAGS_SET(flags, CONF_FILES_REGULAR|CONF_FILES_DIRECTORY)) {
-                        if (!S_ISREG(st.st_mode) && !S_ISDIR(st.st_mode)) {
-                                log_debug("Ignoring '%s/%s', as it is neither a regular file or directory.", root, skip_leading_slash(p));
-                                continue;
-                        }
-                } else {
-                        /* Is this node a regular file? */
-                        if (FLAGS_SET(flags, CONF_FILES_REGULAR) && !S_ISREG(st.st_mode)) {
-                                log_debug("Ignoring '%s/%s', as it is not a regular file.", root, skip_leading_slash(p));
-                                continue;
-                        }
-
-                        /* Is this node a directory? */
-                        if (FLAGS_SET(flags, CONF_FILES_DIRECTORY) && !S_ISDIR(st.st_mode)) {
-                                log_debug("Ignoring '%s/%s', as it is not a directory.", root, skip_leading_slash(p));
-                                continue;
-                        }
-                }
-
-                /* Does this node have the executable bit set?
-                 * As requested: check if the file is marked executable. Note that we don't check access(X_OK)
-                 * here, as we care about whether the file is marked executable at all, and not whether it is
-                 * executable for us, because if so, such errors are stuff we should log about. */
-                if (FLAGS_SET(flags, CONF_FILES_EXECUTABLE) && (st.st_mode & 0111) == 0) {
-                        log_debug("Ignoring '%s/%s', as it is not marked executable.", root, skip_leading_slash(p));
-                        continue;
-                }
-
-                _cleanup_free_ char *n = strdup(de->d_name);
-                if (!n)
-                        return log_oom_debug();
-
-                r = hashmap_ensure_put(files, &string_hash_ops_free_free, n, p);
-                if (r < 0) {
-                        assert(r == -ENOMEM);
-                        return log_oom_debug();
-                }
-                assert(r > 0);
-
-                TAKE_PTR(n);
-                TAKE_PTR(p);
-        }
-
-        return 0;
-}
-
-static int copy_and_sort_files_from_hashmap(Hashmap *fh, const char *root, ConfFilesFlags flags, char ***ret) {
-        _cleanup_free_ char **sv = NULL;
-        _cleanup_strv_free_ char **files = NULL;
-        size_t n = 0;
-        int r;
-
-        assert(ret);
-
-        r = hashmap_dump_sorted(fh, (void***) &sv, /* ret_n = */ NULL);
-        if (r < 0)
-                return log_oom_debug();
-
-        /* The entries in the array given by hashmap_dump_sorted() are still owned by the hashmap. */
-        STRV_FOREACH(s, sv) {
-                _cleanup_free_ char *p = NULL;
-
-                if (FLAGS_SET(flags, CONF_FILES_BASENAME)) {
-                        r = path_extract_filename(*s, &p);
-                        if (r < 0)
-                                return log_debug_errno(r, "Failed to extract filename from '%s': %m", *s);
-                } else if (root) {
-                        r = chaseat_prefix_root(*s, root, &p);
-                        if (r < 0)
-                                return log_debug_errno(r, "Failed to prefix '%s' with root '%s': %m", *s, root);
-                }
-
-                if (p)
-                        r = strv_consume_with_size(&files, &n, TAKE_PTR(p));
-                else
-                        r = strv_extend_with_size(&files, &n, *s);
-                if (r < 0)
-                        return log_oom_debug();
-        }
-
-        *ret = TAKE_PTR(files);
-        return 0;
-}
-
-static int insert_replacement(Hashmap **fh, char *filename_replacement, char *resolved_replacement, char **ret) {
-        _cleanup_free_ char *fname = ASSERT_PTR(filename_replacement), *path = ASSERT_PTR(resolved_replacement);
-        int r;
-
-        assert(fh);
-
-        /* This consumes the input filename and path. */
-
-        const char *existing = hashmap_get(*fh, fname);
-        if (existing) {
-                log_debug("An entry with higher priority already exists ('%s' -> '%s'), ignoring the replacement: %s",
-                          fname, existing, path);
-                if (ret)
-                        *ret = NULL;
-                return 0;
-        }
-
-        _cleanup_free_ char *copy = NULL;
-        if (ret) {
-                copy = strdup(path);
-                if (!copy)
-                        return log_oom_debug();
-        }
-
-        r = hashmap_ensure_put(fh, &string_hash_ops_free_free, fname, path);
-        if (r < 0) {
-                assert(r == -ENOMEM);
-                return log_oom_debug();
-        }
-        assert(r > 0);
-
-        log_debug("Inserted replacement: '%s' -> '%s'", fname, path);
-
-        TAKE_PTR(fname);
-        TAKE_PTR(path);
-
-        if (ret)
-                *ret = TAKE_PTR(copy);
-        return 0;
-}
-
-static int conf_files_list_impl(
-                const char *suffix,
-                int rfd,
-                const char *root, /* for logging, can be NULL */
-                ConfFilesFlags flags,
-                const char * const *dirs,
-                const char *replacement,
-                Hashmap **ret,
-                char **ret_inserted) {
-
-        _cleanup_hashmap_free_ Hashmap *fh = NULL;
-        _cleanup_set_free_ Set *masked = NULL;
-        int r;
-
-        assert(rfd >= 0 || rfd == AT_FDCWD);
-        assert(ret);
-
-        _cleanup_free_ char *filename_replacement = NULL, *resolved_dirpath_replacement = NULL, *resolved_replacement = NULL, *inserted = NULL;
-        if (replacement) {
-                r = path_extract_filename(replacement, &filename_replacement);
-                if (r < 0)
-                        return r;
-
-                _cleanup_free_ char *d = NULL;
-                r = path_extract_directory(replacement, &d);
-                if (r < 0)
-                        return r;
-
-                r = chaseat(rfd, d, CHASE_AT_RESOLVE_IN_ROOT | CHASE_NONEXISTENT, &resolved_dirpath_replacement, /* ret_fd = */ NULL);
-                if (r < 0)
-                        return r;
-
-                resolved_replacement = path_join(resolved_dirpath_replacement, filename_replacement);
-                if (!resolved_replacement)
-                        return log_oom_debug();
-        }
-
-        STRV_FOREACH(p, dirs) {
-                _cleanup_closedir_ DIR *dir = NULL;
-                _cleanup_free_ char *path = NULL;
-
-                r = chase_and_opendirat(rfd, *p, CHASE_AT_RESOLVE_IN_ROOT, &path, &dir);
-                if (r < 0) {
-                        if (r != -ENOENT)
-                                log_debug_errno(r, "Failed to chase and open directory '%s%s', ignoring: %m", strempty(root), *p);
-                        continue;
-                }
-
-                if (resolved_replacement && path_equal(resolved_dirpath_replacement, path)) {
-                        r = insert_replacement(&fh, TAKE_PTR(filename_replacement), TAKE_PTR(resolved_replacement), ret_inserted ? &inserted : NULL);
-                        if (r < 0)
-                                return r;
-                }
-
-                r = files_add(dir, path, rfd, root, &fh, &masked, suffix, flags);
-                if (r == -ENOMEM)
-                        return r;
-        }
-
-        if (resolved_replacement) {
-                r = insert_replacement(&fh, TAKE_PTR(filename_replacement), TAKE_PTR(resolved_replacement), ret_inserted ? &inserted : NULL);
-                if (r < 0)
-                        return r;
-        }
-
-        *ret = TAKE_PTR(fh);
-        if (ret_inserted)
-                *ret_inserted = TAKE_PTR(inserted);
-        return 0;
 }
 
 static int prepare_dirs(const char *root, char * const *dirs, int *ret_rfd, char **ret_root, char ***ret_dirs) {
@@ -339,25 +47,27 @@ static int prepare_dirs(const char *root, char * const *dirs, int *ret_rfd, char
 
         assert(ret_rfd);
         assert(ret_root);
-        assert(ret_dirs);
+        assert(ret_dirs || strv_isempty(dirs));
 
         r = empty_or_root_harder_to_null(&root);
         if (r < 0)
                 return log_debug_errno(r, "Failed to determine if '%s' points to the root directory: %m", strempty(root));
 
+        if (ret_dirs) {
                 dirs_abs = strv_copy(dirs);
                 if (!dirs_abs)
                         return log_oom();
+        }
 
         if (root) {
-                /* WHen a non-trivial root is specified, we will prefix the result later. Hence, it is not
+                /* When a non-trivial root is specified, we will prefix the result later. Hence, it is not
                  * necessary to modify each config directories here. but needs to normalize the root directory. */
                 r = path_make_absolute_cwd(root, &root_abs);
                 if (r < 0)
                         return log_debug_errno(r, "Failed to make '%s' absolute: %m", root);
 
                 path_simplify(root_abs);
-        } else {
+        } else if (ret_dirs) {
                 /* When an empty root or "/" is specified, we will open "/" below, hence we need to make
                  * each config directory absolute if relative. */
                 r = path_strv_make_absolute_cwd(dirs_abs);
@@ -371,10 +81,465 @@ static int prepare_dirs(const char *root, char * const *dirs, int *ret_rfd, char
 
         *ret_rfd = TAKE_FD(rfd);
         *ret_root = TAKE_PTR(root_abs);
+        if (ret_dirs)
                 *ret_dirs = TAKE_PTR(dirs_abs);
         return 0;
 }
 
+static int conf_file_prefix_root(ConfFile *c, const char *root) {
+        char *p;
+        int r;
+
+        assert(c);
+
+        r = chaseat_prefix_root(c->result, root, &p);
+        if (r < 0)
+                return log_debug_errno(r, "Failed to prefix '%s' with root '%s': %m", c->result, root);
+        free_and_replace(c->result, p);
+
+        r = chaseat_prefix_root(c->resolved_path, root, &p);
+        if (r < 0)
+                return log_debug_errno(r, "Failed to prefix '%s' with root '%s': %m", c->resolved_path, root);
+        free_and_replace(c->resolved_path, p);
+
+        /* Do not use chaseat_prefix_root(), as it is for the result of chaseat(), but the path is not chased. */
+        p = path_join(empty_to_root(root), skip_leading_slash(c->original_path));
+        if (!p)
+                return log_oom_debug();
+        path_simplify(p);
+        free_and_replace(c->original_path, p);
+
+        return 0;
+}
+
+int conf_file_new_at(const char *path, int rfd, ChaseFlags chase_flags, ConfFile **ret) {
+        int r;
+
+        assert(path);
+        assert(rfd >= 0 || rfd == AT_FDCWD);
+        assert(ret);
+
+        _cleanup_free_ char *root = NULL;
+        if (rfd >= 0 && DEBUG_LOGGING)
+                (void) fd_get_path(rfd, &root);
+
+        _cleanup_(conf_file_freep) ConfFile *c = new(ConfFile, 1);
+        if (!c)
+                return log_oom_debug();
+
+        *c = (ConfFile) {
+                .original_path = strdup(path),
+                .fd = -EBADF,
+        };
+
+        if (!c->original_path)
+                return log_oom_debug();
+
+        r = path_extract_filename(path, &c->name);
+        if (r < 0)
+                return log_debug_errno(r, "Failed to extract filename from '%s': %m", path);
+
+        _cleanup_free_ char *dirpath = NULL, *resolved_dirpath = NULL;
+        r = path_extract_directory(path, &dirpath);
+        if (r < 0 && r != -EDESTADDRREQ)
+                return log_debug_errno(r, "Failed to extract directory from '%s': %m", path);
+        if (r >= 0) {
+                r = chaseat(rfd, dirpath,
+                            CHASE_AT_RESOLVE_IN_ROOT | (FLAGS_SET(chase_flags, CHASE_NONEXISTENT) ? CHASE_NONEXISTENT : CHASE_MUST_BE_DIRECTORY),
+                            &resolved_dirpath, /* ret_fd = */ NULL);
+                if (r < 0)
+                        return log_debug_errno(r, "Failed to chase '%s%s': %m", empty_to_root(root), skip_leading_slash(dirpath));
+        }
+
+        c->result = path_join(resolved_dirpath, c->name);
+        if (!c->result)
+                return log_oom_debug();
+
+        r = chaseat(rfd, c->result, CHASE_AT_RESOLVE_IN_ROOT | chase_flags, &c->resolved_path, &c->fd);
+        if (r < 0)
+                return log_debug_errno(r, "Failed to chase '%s%s': %m", empty_to_root(root), skip_leading_slash(c->original_path));
+
+        if (c->fd >= 0 && fstat(c->fd, &c->st) < 0)
+                return log_debug_errno(r, "Failed to stat '%s%s': %m", empty_to_root(root), skip_leading_slash(c->resolved_path));
+
+        *ret = TAKE_PTR(c);
+        return 0;
+}
+
+int conf_file_new(const char *path, const char *root, ChaseFlags chase_flags, ConfFile **ret) {
+        int r;
+
+        assert(path);
+        assert((chase_flags & (CHASE_PREFIX_ROOT | CHASE_STEP)) == 0);
+        assert(ret);
+
+        _cleanup_free_ char *root_abs = NULL;
+        _cleanup_close_ int rfd = -EBADF;
+        r = prepare_dirs(root, /* dirs = */ NULL, &rfd, &root_abs, /* ret_dirs = */ NULL);
+        if (r < 0)
+                return r;
+
+        _cleanup_free_ char *path_abs = NULL;
+        if (!root_abs) {
+                r = path_make_absolute_cwd(path, &path_abs);
+                if (r < 0)
+                        return log_debug_errno(r, "Failed to make '%s' absolute: %m", path);
+
+                path = path_abs;
+        }
+
+        _cleanup_(conf_file_freep) ConfFile *c = NULL;
+        r = conf_file_new_at(path, rfd, chase_flags, &c);
+        if (r < 0)
+                return r;
+
+        r = conf_file_prefix_root(c, root_abs);
+        if (r < 0)
+                return r;
+
+        *ret = TAKE_PTR(c);
+        return 0;
+}
+
+DEFINE_PRIVATE_HASH_OPS_WITH_VALUE_DESTRUCTOR(
+                conf_file_hash_ops,
+                char, string_hash_func, string_compare_func,
+                ConfFile, conf_file_free);
+
+static int files_add(
+                DIR *dir,
+                const char *original_dirpath,
+                const char *resolved_dirpath,
+                int rfd,
+                const char *root, /* for logging, can be NULL */
+                Hashmap **files,
+                Set **masked,
+                const char *suffix,
+                ConfFilesFlags flags) {
+
+        int r;
+
+        assert(dir);
+        assert(original_dirpath);
+        assert(resolved_dirpath);
+        assert(rfd >= 0 || rfd == AT_FDCWD);
+        assert(files);
+        assert(masked);
+
+        root = strempty(root);
+
+        FOREACH_DIRENT(de, dir, return log_debug_errno(errno, "Failed to read directory '%s/%s': %m",
+                                                       root, skip_leading_slash(original_dirpath))) {
+
+                _cleanup_free_ char *original_path = path_join(original_dirpath, de->d_name);
+                if (!original_path)
+                        return log_oom_debug();
+
+                /* Does this match the suffix? */
+                if (suffix && !endswith(de->d_name, suffix)) {
+                        log_debug("Skipping file '%s/%s' with unexpected suffix.", root, skip_leading_slash(original_path));
+                        continue;
+                }
+
+                /* Has this file already been found in an earlier directory? */
+                if (hashmap_contains(*files, de->d_name)) {
+                        log_debug("Skipping overridden file '%s/%s'.", root, skip_leading_slash(original_path));
+                        continue;
+                }
+
+                /* Has this been masked in an earlier directory? */
+                if ((flags & CONF_FILES_FILTER_MASKED) != 0 && set_contains(*masked, de->d_name)) {
+                        log_debug("File '%s/%s' is masked by previous entry.", root, skip_leading_slash(original_path));
+                        continue;
+                }
+
+                _cleanup_free_ char *p = path_join(resolved_dirpath, de->d_name);
+                if (!p)
+                        return log_oom_debug();
+
+                _cleanup_free_ char *resolved_path = NULL;
+                _cleanup_close_ int fd = -EBADF;
+                bool need_stat = (flags & (CONF_FILES_FILTER_MASKED | CONF_FILES_REGULAR | CONF_FILES_DIRECTORY | CONF_FILES_EXECUTABLE)) != 0;
+                ChaseFlags chase_flags = CHASE_AT_RESOLVE_IN_ROOT;
+
+                if (!need_stat || FLAGS_SET(flags, CONF_FILES_FILTER_MASKED_BY_SYMLINK))
+                        /* Even if no verification is requested, let's unconditionally call chaseat(),
+                         * to drop unsafe symlinks. */
+                        chase_flags |= CHASE_NONEXISTENT;
+
+                r = chaseat(rfd, p, chase_flags, &resolved_path, &fd);
+                if (r < 0) {
+                        log_debug_errno(r, "Failed to chase '%s/%s', ignoring: %m",
+                                        root, skip_leading_slash(original_path));
+                        continue;
+                }
+                if (r == 0) {
+                        if (FLAGS_SET(flags, CONF_FILES_FILTER_MASKED_BY_SYMLINK)) {
+
+                                /* If the path points to /dev/null in a image or so, then the device node may not exist. */
+                                if (path_equal(skip_leading_slash(resolved_path), "dev/null")) {
+                                        /* Mark this one as masked */
+                                        r = set_put_strdup(masked, de->d_name);
+                                        if (r < 0)
+                                                return log_oom_debug();
+
+                                        log_debug("File '%s/%s' is a mask (symlink to /dev/null).",
+                                                  root, skip_leading_slash(original_path));
+                                        continue;
+                                }
+                        }
+
+                        if (need_stat) {
+                                /* If we need to have stat, skip the entry. */
+                                log_debug_errno(SYNTHETIC_ERRNO(ENOENT), "Failed to chase '%s/%s', ignoring: %m",
+                                                root, skip_leading_slash(original_path));
+                                continue;
+                        }
+                }
+
+                /* Even if we do not need stat, let's take stat now. The caller may use the info later. */
+                struct stat st = {};
+                if (fd >= 0 && fstat(fd, &st) < 0) {
+                        log_debug_errno(errno, "Failed to stat '%s/%s', ignoring: %m",
+                                        root, skip_leading_slash(original_path));
+                        continue;
+                }
+
+                /* Is this a masking entry? */
+                if (FLAGS_SET(flags, CONF_FILES_FILTER_MASKED_BY_SYMLINK) && stat_may_be_dev_null(&st)) {
+                        /* Mark this one as masked */
+                        r = set_put_strdup(masked, de->d_name);
+                        if (r < 0)
+                                return log_oom_debug();
+
+                        log_debug("File '%s/%s' is a mask (symlink to /dev/null).", root, skip_leading_slash(original_path));
+                        continue;
+                }
+
+                if (FLAGS_SET(flags, CONF_FILES_FILTER_MASKED_BY_EMPTY) && stat_is_empty(&st)) {
+                        /* Mark this one as masked */
+                        r = set_put_strdup(masked, de->d_name);
+                        if (r < 0)
+                                return log_oom_debug();
+
+                        log_debug("File '%s/%s' is a mask (an empty file).", root, skip_leading_slash(original_path));
+                        continue;
+                }
+
+                if (FLAGS_SET(flags, CONF_FILES_REGULAR|CONF_FILES_DIRECTORY)) {
+                        if (!S_ISREG(st.st_mode) && !S_ISDIR(st.st_mode)) {
+                                log_debug("Ignoring '%s/%s', as it is neither a regular file or directory.", root, skip_leading_slash(original_path));
+                                continue;
+                        }
+                } else {
+                        /* Is this node a regular file? */
+                        if (FLAGS_SET(flags, CONF_FILES_REGULAR) && !S_ISREG(st.st_mode)) {
+                                log_debug("Ignoring '%s/%s', as it is not a regular file.", root, skip_leading_slash(original_path));
+                                continue;
+                        }
+
+                        /* Is this node a directory? */
+                        if (FLAGS_SET(flags, CONF_FILES_DIRECTORY) && !S_ISDIR(st.st_mode)) {
+                                log_debug("Ignoring '%s/%s', as it is not a directory.", root, skip_leading_slash(original_path));
+                                continue;
+                        }
+                }
+
+                /* Does this node have the executable bit set?
+                 * As requested: check if the file is marked executable. Note that we don't check access(X_OK)
+                 * here, as we care about whether the file is marked executable at all, and not whether it is
+                 * executable for us, because if so, such errors are stuff we should log about. */
+                if (FLAGS_SET(flags, CONF_FILES_EXECUTABLE) && (st.st_mode & 0111) == 0) {
+                        log_debug("Ignoring '%s/%s', as it is not marked executable.", root, skip_leading_slash(original_path));
+                        continue;
+                }
+
+                _cleanup_(conf_file_freep) ConfFile *c = new(ConfFile, 1);
+                if (!c)
+                        return log_oom_debug();
+
+                *c = (ConfFile) {
+                        .name = strdup(de->d_name),
+                        .result = TAKE_PTR(p),
+                        .original_path = TAKE_PTR(original_path),
+                        .resolved_path = TAKE_PTR(resolved_path),
+                        .fd = TAKE_FD(fd),
+                        .st = st,
+                };
+
+                if (!c->name)
+                        return log_oom_debug();
+
+                r = hashmap_ensure_put(files, &conf_file_hash_ops, c->name, c);
+                if (r < 0) {
+                        assert(r == -ENOMEM);
+                        return log_oom_debug();
+                }
+                assert(r > 0);
+
+                TAKE_PTR(c);
+        }
+
+        return 0;
+}
+
+static int dump_files(Hashmap *fh, const char *root, ConfFile ***ret_files, size_t *ret_n_files) {
+        ConfFile **files = NULL;
+        size_t n_files = 0;
+        int r;
+
+        CLEANUP_ARRAY(files, n_files, conf_file_free_many);
+
+        assert(ret_files);
+        assert(ret_n_files);
+
+        /* The entries in the array given by hashmap_dump_sorted() are still owned by the hashmap. */
+        r = hashmap_dump_sorted(fh, (void***) &files, &n_files);
+        if (r < 0)
+                return log_oom_debug();
+
+        /* Hence, we need to remove them from the hashmap. */
+        FOREACH_ARRAY(i, files, n_files)
+                assert_se(hashmap_remove(fh, (*i)->name) == *i);
+
+        if (root)
+                FOREACH_ARRAY(i, files, n_files) {
+                        r = conf_file_prefix_root(*i, root);
+                        if (r < 0)
+                                return r;
+                }
+
+        *ret_files = TAKE_PTR(files);
+        *ret_n_files = n_files;
+        return 0;
+}
+
+static int copy_and_sort_files_from_hashmap(Hashmap *fh, const char *root, ConfFilesFlags flags, char ***ret) {
+        _cleanup_strv_free_ char **results = NULL;
+        _cleanup_free_ ConfFile **files = NULL;
+        size_t n_files = 0, n_results = 0;
+        int r;
+
+        assert(ret);
+
+        /* The entries in the array given by hashmap_dump_sorted() are still owned by the hashmap.
+         * Hence, do not use conf_file_free_many() for 'entries' */
+        r = hashmap_dump_sorted(fh, (void***) &files, &n_files);
+        if (r < 0)
+                return log_oom_debug();
+
+        FOREACH_ARRAY(i, files, n_files) {
+                ConfFile *c = *i;
+
+                if (FLAGS_SET(flags, CONF_FILES_BASENAME))
+                        r = strv_extend_with_size(&results, &n_results, c->name);
+                else if (root) {
+                        char *p;
+
+                        r = chaseat_prefix_root(c->result, root, &p);
+                        if (r < 0)
+                                return log_debug_errno(r, "Failed to prefix '%s' with root '%s': %m", c->result, root);
+
+                        r = strv_consume_with_size(&results, &n_results, TAKE_PTR(p));
+                } else
+                        r = strv_extend_with_size(&results, &n_results, c->result);
+                if (r < 0)
+                        return log_oom_debug();
+        }
+
+        *ret = TAKE_PTR(results);
+        return 0;
+}
+
+static int insert_replacement(Hashmap **fh, ConfFile *replacement, const ConfFile **ret) {
+        _cleanup_(conf_file_freep) ConfFile *c = ASSERT_PTR(replacement);
+        int r;
+
+        assert(fh);
+        assert(ret);
+
+        /* This consumes the input ConfFile. */
+
+        ConfFile *existing = hashmap_get(*fh, c->name);
+        if (existing) {
+                log_debug("An entry with higher priority '%s' -> '%s' already exists, ignoring the replacement: %s",
+                          existing->name, existing->result, c->original_path);
+                *ret = NULL;
+                return 0;
+        }
+
+        r = hashmap_ensure_put(fh, &conf_file_hash_ops, c->name, c);
+        if (r < 0) {
+                assert(r == -ENOMEM);
+                return log_oom_debug();
+        }
+        assert(r > 0);
+
+        log_debug("Inserted replacement: '%s' -> '%s'", c->name, c->result);
+
+        *ret = TAKE_PTR(c);
+        return 0;
+}
+
+static int conf_files_list_impl(
+                const char *suffix,
+                int rfd,
+                const char *root, /* for logging, can be NULL */
+                ConfFilesFlags flags,
+                const char * const *dirs,
+                const char *replacement,
+                Hashmap **ret,
+                const ConfFile **ret_inserted) {
+
+        _cleanup_hashmap_free_ Hashmap *fh = NULL;
+        _cleanup_set_free_ Set *masked = NULL;
+        _cleanup_(conf_file_freep) ConfFile *c = NULL;
+        const ConfFile *inserted = NULL;
+        int r;
+
+        assert(rfd >= 0 || rfd == AT_FDCWD);
+        assert(ret);
+
+        if (replacement) {
+                r = conf_file_new_at(replacement, rfd, CHASE_NONEXISTENT, &c);
+                if (r < 0)
+                        return r;
+        }
+
+        STRV_FOREACH(p, dirs) {
+                _cleanup_closedir_ DIR *dir = NULL;
+                _cleanup_free_ char *path = NULL;
+
+                r = chase_and_opendirat(rfd, *p, CHASE_AT_RESOLVE_IN_ROOT, &path, &dir);
+                if (r < 0) {
+                        if (r != -ENOENT)
+                                log_debug_errno(r, "Failed to chase and open directory '%s/%s', ignoring: %m", strempty(root), skip_leading_slash(*p));
+                        continue;
+                }
+
+                if (c && streq_ptr(path_startswith(c->result, path), c->name)) {
+                        r = insert_replacement(&fh, TAKE_PTR(c), &inserted);
+                        if (r < 0)
+                                return r;
+                }
+
+                r = files_add(dir, *p, path, rfd, root, &fh, &masked, suffix, flags);
+                if (r == -ENOMEM)
+                        return r;
+        }
+
+        if (c) {
+                r = insert_replacement(&fh, TAKE_PTR(c), &inserted);
+                if (r < 0)
+                        return r;
+        }
+
+        *ret = TAKE_PTR(fh);
+        if (ret_inserted)
+                *ret_inserted = inserted;
+        return 0;
+}
+
 int conf_files_list_strv(
                 char ***ret,
                 const char *suffix,
@@ -402,6 +567,35 @@ int conf_files_list_strv(
         return copy_and_sort_files_from_hashmap(fh, empty_to_root(root_abs), flags, ret);
 }
 
+int conf_files_list_strv_full(
+                const char *suffix,
+                const char *root,
+                ConfFilesFlags flags,
+                const char * const *dirs,
+                ConfFile ***ret_files,
+                size_t *ret_n_files) {
+
+        _cleanup_hashmap_free_ Hashmap *fh = NULL;
+        _cleanup_close_ int rfd = -EBADF;
+        _cleanup_free_ char *root_abs = NULL;
+        _cleanup_strv_free_ char **dirs_abs = NULL;
+        int r;
+
+        assert(ret_files);
+        assert(ret_n_files);
+
+        r = prepare_dirs(root, (char**) dirs, &rfd, &root_abs, &dirs_abs);
+        if (r < 0)
+                return r;
+
+        r = conf_files_list_impl(suffix, rfd, root_abs, flags, (const char * const *) dirs_abs,
+                                 /* replacement = */ NULL, &fh, /* ret_inserted = */ NULL);
+        if (r < 0)
+                return r;
+
+        return dump_files(fh, empty_to_root(root_abs), ret_files, ret_n_files);
+}
+
 int conf_files_list_strv_at(
                 char ***ret,
                 const char *suffix,
@@ -426,14 +620,48 @@ int conf_files_list_strv_at(
         return copy_and_sort_files_from_hashmap(fh, /* root = */ NULL, flags, ret);
 }
 
+int conf_files_list_strv_at_full(
+                const char *suffix,
+                int rfd,
+                ConfFilesFlags flags,
+                const char * const *dirs,
+                ConfFile ***ret_files,
+                size_t *ret_n_files) {
+
+        _cleanup_hashmap_free_ Hashmap *fh = NULL;
+        _cleanup_free_ char *root = NULL;
+        int r;
+
+        assert(rfd >= 0 || rfd == AT_FDCWD);
+        assert(ret_files);
+        assert(ret_n_files);
+
+        if (rfd >= 0 && DEBUG_LOGGING)
+                (void) fd_get_path(rfd, &root); /* for logging */
+
+        r = conf_files_list_impl(suffix, rfd, root, flags, dirs, /* replacement = */ NULL, &fh, /* ret_inserted = */ NULL);
+        if (r < 0)
+                return r;
+
+        return dump_files(fh, /* root = */ NULL, ret_files, ret_n_files);
+}
+
 int conf_files_list(char ***ret, const char *suffix, const char *root, ConfFilesFlags flags, const char *dir) {
         return conf_files_list_strv(ret, suffix, root, flags, STRV_MAKE_CONST(dir));
 }
 
+int conf_files_list_full(const char *suffix, const char *root, ConfFilesFlags flags, const char *dir, ConfFile ***ret_files, size_t *ret_n_files) {
+        return conf_files_list_strv_full(suffix, root, flags, STRV_MAKE_CONST(dir), ret_files, ret_n_files);
+}
+
 int conf_files_list_at(char ***ret, const char *suffix, int rfd, ConfFilesFlags flags, const char *dir) {
         return conf_files_list_strv_at(ret, suffix, rfd, flags, STRV_MAKE_CONST(dir));
 }
 
+int conf_files_list_at_full(const char *suffix, int rfd, ConfFilesFlags flags, const char *dir, ConfFile ***ret_files, size_t *ret_n_files) {
+        return conf_files_list_strv_at_full(suffix, rfd, flags, STRV_MAKE_CONST(dir), ret_files, ret_n_files);
+}
+
 int conf_files_list_nulstr(char ***ret, const char *suffix, const char *root, ConfFilesFlags flags, const char *dirs) {
         _cleanup_strv_free_ char **d = NULL;
 
@@ -446,6 +674,19 @@ int conf_files_list_nulstr(char ***ret, const char *suffix, const char *root, Co
         return conf_files_list_strv(ret, suffix, root, flags, (const char**) d);
 }
 
+int conf_files_list_nulstr_full(const char *suffix, const char *root, ConfFilesFlags flags, const char *dirs, ConfFile ***ret_files, size_t *ret_n_files) {
+        _cleanup_strv_free_ char **d = NULL;
+
+        assert(ret_files);
+        assert(ret_n_files);
+
+        d = strv_split_nulstr(dirs);
+        if (!d)
+                return -ENOMEM;
+
+        return conf_files_list_strv_full(suffix, root, flags, (const char**) d, ret_files, ret_n_files);
+}
+
 int conf_files_list_nulstr_at(char ***ret, const char *suffix, int rfd, ConfFilesFlags flags, const char *dirs) {
         _cleanup_strv_free_ char **d = NULL;
 
@@ -458,6 +699,19 @@ int conf_files_list_nulstr_at(char ***ret, const char *suffix, int rfd, ConfFile
         return conf_files_list_strv_at(ret, suffix, rfd, flags, (const char**) d);
 }
 
+int conf_files_list_nulstr_at_full(const char *suffix, int rfd, ConfFilesFlags flags, const char *dirs, ConfFile ***ret_files, size_t *ret_n_files) {
+        _cleanup_strv_free_ char **d = NULL;
+
+        assert(ret_files);
+        assert(ret_n_files);
+
+        d = strv_split_nulstr(dirs);
+        if (!d)
+                return -ENOMEM;
+
+        return conf_files_list_strv_at_full(suffix, rfd, flags, (const char**) d, ret_files, ret_n_files);
+}
+
 int conf_files_list_with_replacement(
                 const char *root,
                 char **config_dirs,
@@ -471,6 +725,7 @@ int conf_files_list_with_replacement(
         _cleanup_close_ int rfd = -EBADF;
         _cleanup_free_ char *root_abs = NULL;
         _cleanup_strv_free_ char **dirs_abs = NULL;
+        const ConfFile *c = NULL;
         int r;
 
         assert(ret_files);
@@ -480,18 +735,14 @@ int conf_files_list_with_replacement(
                 return r;
 
         r = conf_files_list_impl(".conf", rfd, root_abs, flags, (const char * const *) dirs_abs,
-                                 replacement, &fh, ret_inserted ? &inserted : NULL);
+                                 replacement, &fh, ret_inserted ? &c : NULL);
         if (r < 0)
                 return r;
 
-        if (inserted) {
+        if (c) {
-                char *p;
+                r = chaseat_prefix_root(c->result, root_abs, &inserted);
-
-                r = chaseat_prefix_root(inserted, root_abs, &p);
                 if (r < 0)
-                        return log_debug_errno(r, "Failed to prefix '%s' with root '%s': %m", inserted, empty_to_root(root_abs));
+                        return log_debug_errno(r, "Failed to prefix '%s' with root '%s': %m", c->result, empty_to_root(root_abs));
-
-                free_and_replace(inserted, p);
         }
 
         r = copy_and_sort_files_from_hashmap(fh, empty_to_root(root_abs), flags, ret_files);

src/basic/conf-files.h

@@ -1,6 +1,8 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 #pragma once
 
+#include <sys/stat.h>
+
 #include "forward.h"
 
 typedef enum ConfFilesFlags {
@@ -13,12 +15,36 @@ typedef enum ConfFilesFlags {
         CONF_FILES_FILTER_MASKED            = CONF_FILES_FILTER_MASKED_BY_SYMLINK | CONF_FILES_FILTER_MASKED_BY_EMPTY,
 } ConfFilesFlags;
 
+typedef struct ConfFile {
+        char *name;          /* name of a file found in config directories */
+        char *result;        /* resolved config directory with the original file name found in the directory */
+        char *original_path; /* original config directory with the original file name found in the directory */
+        char *resolved_path; /* fully resolved path, where the filename part of the path may be different from the original name */
+        int fd;              /* O_PATH fd to resolved_path, -EBADF if the resolved_path does not exist */
+        struct stat st;      /* stat of the file. */
+} ConfFile;
+
+ConfFile* conf_file_free(ConfFile *c);
+DEFINE_TRIVIAL_CLEANUP_FUNC(ConfFile*, conf_file_free);
+void conf_file_free_many(ConfFile **array, size_t n);
+
+int conf_file_new_at(const char *path, int rfd, ChaseFlags chase_flags, ConfFile **ret);
+int conf_file_new(const char *path, const char *root, ChaseFlags chase_flags, ConfFile **ret);
+
 int conf_files_list(char ***ret, const char *suffix, const char *root, ConfFilesFlags flags, const char *dir);
 int conf_files_list_at(char ***ret, const char *suffix, int rfd, ConfFilesFlags flags, const char *dir);
 int conf_files_list_strv(char ***ret, const char *suffix, const char *root, ConfFilesFlags flags, const char* const* dirs);
 int conf_files_list_strv_at(char ***ret, const char *suffix, int rfd, ConfFilesFlags flags, const char * const *dirs);
 int conf_files_list_nulstr(char ***ret, const char *suffix, const char *root, ConfFilesFlags flags, const char *dirs);
 int conf_files_list_nulstr_at(char ***ret, const char *suffix, int rfd, ConfFilesFlags flags, const char *dirs);
+
+int conf_files_list_full(const char *suffix, const char *root, ConfFilesFlags flags, const char *dir, ConfFile ***ret_files, size_t *ret_n_files);
+int conf_files_list_at_full(const char *suffix, int rfd, ConfFilesFlags flags, const char *dir, ConfFile ***ret_files, size_t *ret_n_files);
+int conf_files_list_strv_full(const char *suffix, const char *root, ConfFilesFlags flags, const char * const *dirs, ConfFile ***ret_files, size_t *ret_n_files);
+int conf_files_list_strv_at_full(const char *suffix, int rfd, ConfFilesFlags flags, const char * const *dirs, ConfFile ***ret_files, size_t *ret_n_files);
+int conf_files_list_nulstr_full(const char *suffix, const char *root, ConfFilesFlags flags, const char *dirs, ConfFile ***ret_files, size_t *ret_n_files);
+int conf_files_list_nulstr_at_full(const char *suffix, int rfd, ConfFilesFlags flags, const char *dirs, ConfFile ***ret_files, size_t *ret_n_files);
+
 int conf_files_list_with_replacement(
                 const char *root,
                 char **config_dirs,

src/basic/fd-util.c

@@ -2,8 +2,8 @@
 
 #include <fcntl.h>
 #include <linux/fs.h>
-#include <linux/kcmp.h>
 #include <sys/ioctl.h>
+#include <sys/kcmp.h>
 #include <sys/resource.h>
 #include <sys/stat.h>
 #include <unistd.h>
@@ -16,7 +16,6 @@
 #include "format-util.h"
 #include "fs-util.h"
 #include "log.h"
-#include "missing_syscall.h"
 #include "mountpoint-util.h"
 #include "parse-util.h"
 #include "path-util.h"

src/basic/filesystems-gperf.gperf

@@ -6,7 +6,6 @@ _Pragma("GCC diagnostic ignored \"-Wzero-as-null-pointer-constant\"")
 #include <linux/magic.h>
 
 #include "filesystems.h"
-#include "missing_magic.h"
 #include "stat-util.h"
 
 struct FilesystemMagic {

src/basic/forward.h

@@ -104,6 +104,7 @@ typedef struct Set Set;
 
 typedef struct dual_timestamp dual_timestamp;
 typedef struct triple_timestamp triple_timestamp;
+typedef struct ConfFile ConfFile;
 typedef struct LockFile LockFile;
 typedef struct PidRef PidRef;
 typedef struct Prioq Prioq;

src/basic/fs-util.c

@@ -16,7 +16,6 @@
 #include "label.h"
 #include "lock-util.h"
 #include "log.h"
-#include "missing_syscall.h"
 #include "mkdir.h"
 #include "path-util.h"
 #include "process-util.h"

src/basic/generate-af-list.sh

@@ -3,6 +3,9 @@
 set -eu
 set -o pipefail
 
-${1:?} -E -dM -include sys/socket.h -include "${2:?}" - </dev/null | \
+CC=${1:?}
+shift
+
+$CC -E -dM -include sys/socket.h "$@" - </dev/null | \
        grep -Ev 'AF_UNSPEC|AF_MAX' | \
        awk '/^#define[ \t]+AF_[^ \t]+[ \t]+[AP]F_[^ \t]/ { print $2; }'

src/basic/generate-arphrd-list.sh

@@ -3,6 +3,9 @@
 set -eu
 set -o pipefail
 
-${1:?} -dM -include "${2:?}" - </dev/null | \
+CC=${1:?}
+shift
+
+$CC -dM -include linux/if_arp.h "$@" - </dev/null | \
        awk '/^#define[ \t]+ARPHRD_[^ \t]+[ \t]+[^ \t]/ { print $2; }' | \
        sed -e 's/ARPHRD_//'

src/basic/generate-cap-list.sh

@@ -3,6 +3,9 @@
 set -eu
 set -o pipefail
 
-${1:?} -dM -include "${2:?}" - </dev/null | \
+CC=${1:?}
+shift
+
+$CC -dM -include linux/capability.h "$@" - </dev/null | \
        awk '/^#define[ \t]+CAP_[A-Z_]+[ \t]+/ { print $2; }' | \
        grep -v CAP_LAST_CAP

src/basic/generate-errno-list.sh

@@ -6,6 +6,9 @@ set -o pipefail
 # In kernel's arch/parisc/include/uapi/asm/errno.h, ECANCELLED and EREFUSED are defined as aliases of
 # ECANCELED and ECONNREFUSED, respectively. Let's drop them.
 
-${1:?} -dM -include errno.h - </dev/null | \
+CC=${1:?}
+shift
+
+$CC -dM -include errno.h "$@" - </dev/null | \
        grep -Ev '^#define[[:space:]]+(ECANCELLED|EREFUSED)' | \
        awk '/^#define[ \t]+E[^ _]+[ \t]+/ { print $2; }'

src/basic/include/sys/xattr.h

@@ -1,8 +0,0 @@
-/* SPDX-License-Identifier: LGPL-2.1-or-later */
-#pragma once
-
-/* To make struct xattr_args defined, which is used by setxattrat(). Note, the kernel header must be
- * included before the glibc header, otherwise the struct will not be defined. */
-#include <linux/xattr.h>
-
-#include_next <sys/xattr.h>

src/basic/keyring-util.c

@@ -3,7 +3,6 @@
 #include "alloc-util.h"
 #include "keyring-util.h"
 #include "log.h"
-#include "missing_syscall.h"
 
 int keyring_read(key_serial_t serial, void **ret, size_t *ret_size) {
         size_t bufsize = 100;

src/basic/keyring-util.h

@@ -1,8 +1,9 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 #pragma once
 
+#include <sys/keyctl.h>         /* IWYU pragma: export */
+
 #include "forward.h"
-#include "missing_keyctl.h"
 
 /* Like TAKE_PTR() but for key_serial_t, resetting them to -1 */
 #define TAKE_KEY_SERIAL(key_serial) TAKE_GENERIC(key_serial, key_serial_t, -1)

src/basic/meson.build

@@ -84,6 +84,7 @@ basic_sources = files(
         'psi-util.c',
         'random-util.c',
         'ratelimit.c',
+        'raw-clone.c',
         'recurse-dir.c',
         'replace-var.c',
         'rlimit-util.c',
@@ -119,44 +120,26 @@ basic_sources = files(
 
 sources += basic_sources
 
-missing_audit_h = files('missing_audit.h')
-
-missing_syscall_def_h = files('missing_syscall_def.h')
-basic_sources += missing_syscall_def_h
-
-generate_af_list = find_program('generate-af-list.sh')
-af_list_txt = custom_target(
-        output : 'af-list.txt',
-        command : [generate_af_list, cpp, files('include/sys/socket.h')],
-        capture : true)
-
-generate_arphrd_list = find_program('generate-arphrd-list.sh')
-arphrd_list_txt = custom_target(
-        output : 'arphrd-list.txt',
-        command : [generate_arphrd_list, cpp, files('include/linux/if_arp.h')],
-        capture : true)
-
-generate_cap_list = find_program('generate-cap-list.sh')
-cap_list_txt = custom_target(
-        output : 'cap-list.txt',
-        command : [generate_cap_list, cpp, files('include/linux/capability.h')],
-        capture : true)
-
-generate_errno_list = find_program('generate-errno-list.sh')
-errno_list_txt = custom_target(
-        output : 'errno-list.txt',
-        command : [generate_errno_list, cpp],
-        capture : true)
-
 generated_gperf_headers = []
-foreach item : [['af',     af_list_txt,     'af',         '',        ['<sys/socket.h>']],
+foreach item : [
-                ['arphrd', arphrd_list_txt, 'arphrd',     'ARPHRD_', []],
+        # name,    source,         struct name,  prefix,    headers
-                ['cap',    cap_list_txt,    'capability', '',        []],
+        ['af',     af_sources,     'af',         '',        ['<sys/socket.h>'],       ],
-                ['errno',  errno_list_txt,  'errno',      '',        []]]
+        ['arphrd', arphrd_sources, 'arphrd',     'ARPHRD_', ['<linux/if_arp.h>'],     ],
+        ['cap',    cap_sources,    'capability', '',        ['<linux/capability.h>'], ],
+        ['errno',  [],             'errno',      '',        ['<errno.h>'],            ],
+]
+
+        fname = '@0@-list.txt'.format(item[0])
+        generate_list = files('generate-@0@-list.sh'.format(item[0]))
+        list_txt = custom_target(
+                input : [generate_list, item[1]],
+                output : fname,
+                command : [env, 'bash', generate_list, cpp, system_include_args],
+                capture : true)
 
         fname = '@0@-from-name.gperf'.format(item[0])
         gperf_file = custom_target(
-                input : item[1],
+                input : list_txt,
                 output : fname,
                 command : [generate_gperfs, item[2], item[3], '@INPUT@'] + item[4],
                 capture : true)
@@ -176,7 +159,7 @@ foreach item : [['af',     af_list_txt,     'af',         '',        ['<sys/sock
         fname = '@0@-to-name.inc'.format(item[0])
         awkscript = '@0@-to-name.awk'.format(item[0])
         target2 = custom_target(
-                input : [awkscript, item[1]],
+                input : [awkscript, list_txt],
                 output : fname,
                 command : [awk, '-f', '@INPUT0@', '@INPUT1@'],
                 capture : true)
@@ -189,61 +172,13 @@ basic_sources += generated_gperf_headers
 
 ############################################################
 
-arch_list = [
-        'alpha',
-        'arc',
-        'arm',
-        'arm64',
-        'i386',
-        'ia64',
-        'loongarch64',
-        'm68k',
-        'mips64',
-        'mips64n32',
-        'mipso32',
-        'parisc',
-        'powerpc',
-        'powerpc64',
-        'riscv32',
-        'riscv64',
-        's390',
-        's390x',
-        'sparc',
-        'x86_64'
-]
-
-run_target(
-        'update-syscall-tables',
-        command : [update_syscall_tables_sh, meson.current_source_dir()] + arch_list)
-
-syscall_list_txt = files('syscall-list.txt')
-
-syscall_lists = []
-foreach arch: arch_list
-        syscall_lists += files('syscalls-@0@.txt'.format(arch))
-endforeach
-
-missing_syscalls_py = find_program('missing_syscalls.py')
-
-run_target(
-        'update-syscall-header',
-        command : [missing_syscalls_py,
-                   missing_syscall_def_h,
-                   syscall_lists])
-
-############################################################
-
-filesystem_includes = files(
-        'include/linux/magic.h',
-        'missing_magic.h',
-)
-
 check_filesystems = find_program('check-filesystems.sh')
 r = run_command(
         [
                 'env', '--chdir', meson.project_build_root(),
-                check_filesystems, cpp, files('filesystems-gperf.gperf')
+                check_filesystems, cpp, files('filesystems-gperf.gperf'),
-        ] + filesystem_includes,
+                system_include_args,
+        ],
         check: false,
 )
 if r.returncode() != 0
@@ -273,17 +208,8 @@ filesystem_switch_case_inc = custom_target(
                    '@INPUT@'],
         capture : true)
 
-generate_bpf_delegate_configs = find_program('generate-bpf-delegate-configs.py')
+generated_sources += [filesystem_list_inc, filesystem_switch_case_inc, filesystems_gperf_h]
-bpf_delegate_configs_inc = custom_target(
+basic_sources += [filesystem_list_inc, filesystem_switch_case_inc, filesystems_gperf_h]
-        input : files('include/linux/bpf.h'),
-        output : 'bpf-delegate-configs.inc',
-        command : [generate_bpf_delegate_configs,
-                   'code',
-                   '@INPUT@'],
-        capture : true)
-
-generated_sources += [filesystem_list_inc, filesystem_switch_case_inc, filesystems_gperf_h, bpf_delegate_configs_inc]
-basic_sources += [filesystem_list_inc, filesystem_switch_case_inc, filesystems_gperf_h, bpf_delegate_configs_inc]
 
 libbasic_static = static_library(
         'basic',

src/basic/missing_syscall.h

@@ -1,249 +0,0 @@
-/* SPDX-License-Identifier: LGPL-2.1-or-later */
-#pragma once
-
-/* Missing glibc definitions to access certain kernel APIs */
-
-#include <linux/mempolicy.h>
-#include <sched.h>
-#include <signal.h>
-#include <sys/syscall.h>
-#include <sys/xattr.h>
-#include <unistd.h>
-
-#ifdef ARCH_MIPS
-#include <asm/sgidefs.h>
-#endif
-
-#include "forward.h"
-#include "missing_keyctl.h"
-#include "missing_syscall_def.h"
-
-/* ======================================================================= */
-
-#if !HAVE_FCHMODAT2
-/* since kernel v6.6 (78252deb023cf0879256fcfbafe37022c390762b) */
-static inline int missing_fchmodat2(int dirfd, const char *path, mode_t mode, int flags) {
-        return syscall(__NR_fchmodat2, dirfd, path, mode, flags);
-}
-
-#  define fchmodat2 missing_fchmodat2
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_PIVOT_ROOT
-static inline int missing_pivot_root(const char *new_root, const char *put_old) {
-        return syscall(__NR_pivot_root, new_root, put_old);
-}
-
-#  define pivot_root missing_pivot_root
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_IOPRIO_GET
-static inline int missing_ioprio_get(int which, int who) {
-        return syscall(__NR_ioprio_get, which, who);
-}
-
-#  define ioprio_get missing_ioprio_get
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_IOPRIO_SET
-static inline int missing_ioprio_set(int which, int who, int ioprio) {
-        return syscall(__NR_ioprio_set, which, who, ioprio);
-}
-
-#  define ioprio_set missing_ioprio_set
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_KCMP
-static inline int missing_kcmp(pid_t pid1, pid_t pid2, int type, unsigned long idx1, unsigned long idx2) {
-        return syscall(__NR_kcmp, pid1, pid2, type, idx1, idx2);
-}
-
-#  define kcmp missing_kcmp
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_KEYCTL
-static inline long missing_keyctl(int cmd, unsigned long arg2, unsigned long arg3, unsigned long arg4, unsigned long arg5) {
-        return syscall(__NR_keyctl, cmd, arg2, arg3, arg4, arg5);
-
-#  define keyctl missing_keyctl
-}
-
-/* ======================================================================= */
-
-static inline key_serial_t missing_add_key(const char *type, const char *description, const void *payload, size_t plen, key_serial_t ringid) {
-        return syscall(__NR_add_key, type, description, payload, plen, ringid);
-
-#  define add_key missing_add_key
-}
-
-/* ======================================================================= */
-
-static inline key_serial_t missing_request_key(const char *type, const char *description, const char * callout_info, key_serial_t destringid) {
-        return syscall(__NR_request_key, type, description, callout_info, destringid);
-
-#  define request_key missing_request_key
-}
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_BPF
-union bpf_attr;
-
-static inline int missing_bpf(int cmd, union bpf_attr *attr, size_t size) {
-        return (int) syscall(__NR_bpf, cmd, attr, size);
-}
-
-#  define bpf missing_bpf
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_SET_MEMPOLICY
-static inline long missing_set_mempolicy(int mode, const unsigned long *nodemask,
-                           unsigned long maxnode) {
-        return syscall(__NR_set_mempolicy, mode, nodemask, maxnode);
-}
-
-#  define set_mempolicy missing_set_mempolicy
-#endif
-
-#if !HAVE_GET_MEMPOLICY
-static inline long missing_get_mempolicy(int *mode, unsigned long *nodemask,
-                           unsigned long maxnode, void *addr,
-                           unsigned long flags) {
-        return syscall(__NR_get_mempolicy, mode, nodemask, maxnode, addr, flags);
-}
-
-#  define get_mempolicy missing_get_mempolicy
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_PIDFD_SEND_SIGNAL
-/* since kernel v5.1 (3eb39f47934f9d5a3027fe00d906a45fe3a15fad) */
-static inline int missing_pidfd_send_signal(int fd, int sig, siginfo_t *info, unsigned flags) {
-        return syscall(__NR_pidfd_send_signal, fd, sig, info, flags);
-}
-
-#  define pidfd_send_signal missing_pidfd_send_signal
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_PIDFD_OPEN
-/* since kernel v5.3 (7615d9e1780e26e0178c93c55b73309a5dc093d7) */
-static inline int missing_pidfd_open(pid_t pid, unsigned flags) {
-        return syscall(__NR_pidfd_open, pid, flags);
-}
-
-#  define pidfd_open missing_pidfd_open
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_RT_TGSIGQUEUEINFO
-static inline int missing_rt_tgsigqueueinfo(pid_t tgid, pid_t tid, int sig, siginfo_t *info) {
-        return syscall(__NR_rt_tgsigqueueinfo, tgid, tid, sig, info);
-}
-
-#  define rt_tgsigqueueinfo missing_rt_tgsigqueueinfo
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_EXECVEAT
-/* since kernel v3.19 (51f39a1f0cea1cacf8c787f652f26dfee9611874) */
-static inline int missing_execveat(int dirfd, const char *pathname,
-                                   char *const argv[], char *const envp[],
-                                   int flags) {
-        return syscall(__NR_execveat, dirfd, pathname, argv, envp, flags);
-}
-
-#  define execveat missing_execveat
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_CLOSE_RANGE
-/* since kernel v5.9 (9b4feb630e8e9801603f3cab3a36369e3c1cf88d) */
-static inline int missing_close_range(unsigned first_fd, unsigned end_fd, unsigned flags) {
-        /* Kernel-side the syscall expects fds as unsigned integers (just like close() actually), while
-         * userspace exclusively uses signed integers for fds. glibc chose to expose it 1:1 however, hence we
-         * do so here too, even if we end up passing signed fds to it most of the time. */
-        return syscall(__NR_close_range,
-                       first_fd,
-                       end_fd,
-                       flags);
-}
-
-#  define close_range missing_close_range
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_SCHED_SETATTR
-/* since kernel 3.14 (e6cfc0295c7d51b008999a8b13a44fb43f8685ea) */
-static inline ssize_t missing_sched_setattr(pid_t pid, struct sched_attr *attr, unsigned int flags) {
-        return syscall(__NR_sched_setattr, pid, attr, flags);
-}
-
-#  define sched_setattr missing_sched_setattr
-#endif
-
-/* ======================================================================= */
-
-/* glibc does not provide clone() on ia64, only clone2(). Not only that, but it also doesn't provide a
- * prototype, only the symbol in the shared library (it provides a prototype for clone(), but not the
- * symbol in the shared library). */
-#if defined(__ia64__)
-int __clone2(int (*fn)(void *), void *stack_base, size_t stack_size, int flags, void *arg);
-#define HAVE_CLONE 0
-#else
-/* We know that everywhere else clone() is available, so we don't bother with a meson check (that takes time
- * at build time) and just define it. Once the kernel drops ia64 support, we can drop this too. */
-#define HAVE_CLONE 1
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_QUOTACTL_FD
-/* since kernel v5.14 (64c2c2c62f92339b176ea24403d8db16db36f9e6) */
-static inline int missing_quotactl_fd(int fd, int cmd, int id, void *addr) {
-        return syscall(__NR_quotactl_fd, fd, cmd, id, addr);
-}
-
-#  define quotactl_fd missing_quotactl_fd
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_SETXATTRAT
-/* since kernel v6.13 (6140be90ec70c39fa844741ca3cc807dd0866394) */
-static inline int missing_setxattrat(int fd, const char *path, int at_flags, const char *name, const struct xattr_args *args, size_t size) {
-        return syscall(__NR_setxattrat, fd, path, at_flags, name, args, size);
-}
-
-#  define setxattrat missing_setxattrat
-#endif
-
-/* ======================================================================= */
-
-#if !HAVE_REMOVEXATTRAT
-/* since kernel v6.13 (6140be90ec70c39fa844741ca3cc807dd0866394) */
-static inline int missing_removexattrat(int fd, const char *path, int at_flags, const char *name) {
-        return syscall(__NR_removexattrat, fd, path, at_flags, name);
-}
-
-#  define removexattrat missing_removexattrat
-#endif

src/basic/namespace-util.c

@@ -1,6 +1,8 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
 #include <fcntl.h>
+#include <linux/magic.h>
+#include <linux/nsfs.h>
 #include <sched.h>
 #include <sys/ioctl.h>
 #include <sys/mount.h>
@@ -10,8 +12,6 @@
 #include "fd-util.h"
 #include "fileio.h"
 #include "log.h"
-#include "missing_magic.h"
-#include "missing_namespace.h"
 #include "mountpoint-util.h"
 #include "namespace-util.h"
 #include "parse-util.h"

src/basic/pidfd-util.c

@@ -1,5 +1,7 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
+#include <linux/fs.h>
+#include <linux/magic.h>
 #include <sys/ioctl.h>
 #include <threads.h>
 #include <unistd.h>
@@ -7,8 +9,6 @@
 #include "errno-util.h"
 #include "fd-util.h"
 #include "fileio.h"
-#include "missing_fs.h"
-#include "missing_magic.h"
 #include "mountpoint-util.h"
 #include "parse-util.h"
 #include "pidfd-util.h"

src/basic/pidfd-util.h

@@ -4,7 +4,6 @@
 #include <sys/pidfd.h>          /* IWYU pragma: export */
 
 #include "forward.h"
-#include "missing_syscall.h"    /* IWYU pragma: export */
 
 int pidfd_get_namespace(int fd, unsigned long ns_type_cmd);
 

src/basic/process-util.c

@@ -34,7 +34,6 @@
 #include "locale-util.h"
 #include "log.h"
 #include "memory-util.h"
-#include "missing_syscall.h"
 #include "mountpoint-util.h"
 #include "namespace-util.h"
 #include "nulstr-util.h"

src/basic/raw-clone.c

@@ -0,0 +1,84 @@
+/* SPDX-License-Identifier: LGPL-2.1-or-later */
+
+/***
+  Copyright © 2016 Michael Karcher
+***/
+
+#include <errno.h>
+#include <sched.h>
+#include <sys/syscall.h>
+#include <unistd.h>
+
+#include "process-util.h"
+#include "raw-clone.h"
+
+/**
+ * raw_clone() - uses clone to create a new process with clone flags
+ * @flags: Flags to pass to the clone system call
+ *
+ * Uses the clone system call to create a new process with the cloning flags and termination signal passed in the flags
+ * parameter. Opposed to glibc's clone function, using this function does not set up a separate stack for the child, but
+ * relies on copy-on-write semantics on the one stack at a common virtual address, just as fork does.
+ *
+ * To obtain copy-on-write semantics, flags must not contain CLONE_VM, and thus CLONE_THREAD and CLONE_SIGHAND
+ * (which require CLONE_VM) are not usable.
+ *
+ * Additionally, as this function does not pass the ptid (pidfd in the case of CLONE_PIDFD), newtls and ctid
+ * parameters to the kernel, flags must not contain CLONE_PARENT_SETTID, CLONE_CHILD_SETTID, CLONE_CHILD_CLEARTID,
+ * CLONE_SETTLS, or CLONE_PIDFD.
+ *
+ * WARNING: 💣 this call (just like glibc's own clone() wrapper) will not synchronize on glibc's malloc
+ *          locks, which means they will be in an undefined state in the child if the parent is
+ *          threaded. This means: the parent must either never use threads, or the child cannot use memory
+ *          allocation itself. This is a major pitfall, hence be careful! 💣
+ *
+ * Returns: 0 in the child process and the child process id in the parent.
+ */
+pid_t raw_clone(unsigned long flags) {
+        pid_t ret;
+
+        assert((flags & (CLONE_VM|CLONE_PARENT_SETTID|CLONE_CHILD_SETTID|CLONE_CHILD_CLEARTID|CLONE_SETTLS|CLONE_PIDFD)) == 0);
+#if defined(__s390x__) || defined(__s390__) || defined(__CRIS__)
+        /* On s390/s390x and cris the order of the first and second arguments
+         * of the raw clone() system call is reversed. */
+        ret = (pid_t) syscall(__NR_clone, NULL, flags);
+#elif defined(__sparc__)
+        {
+                /**
+                 * sparc always returns the other process id in %o0, and
+                 * a boolean flag whether this is the child or the parent in
+                 * %o1. Inline assembly is needed to get the flag returned
+                 * in %o1.
+                 */
+                int in_child, child_pid, error;
+
+                asm volatile("mov %3, %%g1\n\t"
+                             "mov %4, %%o0\n\t"
+                             "mov 0 , %%o1\n\t"
+#if defined(__arch64__)
+                             "t 0x6d\n\t"
+#else
+                             "t 0x10\n\t"
+#endif
+                             "addx %%g0, 0, %2\n\t"
+                             "mov %%o1, %0\n\t"
+                             "mov %%o0, %1" :
+                             "=r"(in_child), "=r"(child_pid), "=r"(error) :
+                             "i"(__NR_clone), "r"(flags) :
+                             "%o1", "%o0", "%g1", "cc" );
+
+                if (error) {
+                        errno = child_pid;
+                        ret = -1;
+                } else
+                        ret = in_child ? 0 : child_pid;
+        }
+#else
+        ret = (pid_t) syscall(__NR_clone, flags, NULL);
+#endif
+
+        if (ret == 0)
+                reset_cached_pid();
+
+        return ret;
+}

src/basic/raw-clone.h

@@ -1,85 +1,8 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 #pragma once
 
-/***
+#include <sched.h>              /* IWYU pragma: export */
-  Copyright © 2016 Michael Karcher
-***/
 
-#include <errno.h>
+#include "forward.h"
-#include <sched.h>
-#include <sys/syscall.h>
 
-#include "log.h"
+pid_t raw_clone(unsigned long flags);
-#include "macro.h"
-#include "process-util.h"
-
-/**
- * raw_clone() - uses clone to create a new process with clone flags
- * @flags: Flags to pass to the clone system call
- *
- * Uses the clone system call to create a new process with the cloning flags and termination signal passed in the flags
- * parameter. Opposed to glibc's clone function, using this function does not set up a separate stack for the child, but
- * relies on copy-on-write semantics on the one stack at a common virtual address, just as fork does.
- *
- * To obtain copy-on-write semantics, flags must not contain CLONE_VM, and thus CLONE_THREAD and CLONE_SIGHAND
- * (which require CLONE_VM) are not usable.
- *
- * Additionally, as this function does not pass the ptid (pidfd in the case of CLONE_PIDFD), newtls and ctid
- * parameters to the kernel, flags must not contain CLONE_PARENT_SETTID, CLONE_CHILD_SETTID, CLONE_CHILD_CLEARTID,
- * CLONE_SETTLS, or CLONE_PIDFD.
- *
- * WARNING: 💣 this call (just like glibc's own clone() wrapper) will not synchronize on glibc's malloc
- *          locks, which means they will be in an undefined state in the child if the parent is
- *          threaded. This means: the parent must either never use threads, or the child cannot use memory
- *          allocation itself. This is a major pitfall, hence be careful! 💣
- *
- * Returns: 0 in the child process and the child process id in the parent.
- */
-static inline pid_t raw_clone(unsigned long flags) {
-        pid_t ret;
-
-        assert((flags & (CLONE_VM|CLONE_PARENT_SETTID|CLONE_CHILD_SETTID|CLONE_CHILD_CLEARTID|CLONE_SETTLS|CLONE_PIDFD)) == 0);
-#if defined(__s390x__) || defined(__s390__) || defined(__CRIS__)
-        /* On s390/s390x and cris the order of the first and second arguments
-         * of the raw clone() system call is reversed. */
-        ret = (pid_t) syscall(__NR_clone, NULL, flags);
-#elif defined(__sparc__)
-        {
-                /**
-                 * sparc always returns the other process id in %o0, and
-                 * a boolean flag whether this is the child or the parent in
-                 * %o1. Inline assembly is needed to get the flag returned
-                 * in %o1.
-                 */
-                int in_child, child_pid, error;
-
-                asm volatile("mov %3, %%g1\n\t"
-                             "mov %4, %%o0\n\t"
-                             "mov 0 , %%o1\n\t"
-#if defined(__arch64__)
-                             "t 0x6d\n\t"
-#else
-                             "t 0x10\n\t"
-#endif
-                             "addx %%g0, 0, %2\n\t"
-                             "mov %%o1, %0\n\t"
-                             "mov %%o0, %1" :
-                             "=r"(in_child), "=r"(child_pid), "=r"(error) :
-                             "i"(__NR_clone), "r"(flags) :
-                             "%o1", "%o0", "%g1", "cc" );
-
-                if (error) {
-                        errno = child_pid;
-                        ret = -1;
-                } else
-                        ret = in_child ? 0 : child_pid;
-        }
-#else
-        ret = (pid_t) syscall(__NR_clone, flags, NULL);
-#endif
-
-        if (ret == 0)
-                reset_cached_pid();
-
-        return ret;
-}

src/basic/raw-reboot.h

@@ -1,15 +0,0 @@
-/* SPDX-License-Identifier: LGPL-2.1-or-later */
-#pragma once
-
-#include <linux/reboot.h>
-#include <sys/reboot.h>
-#include <sys/syscall.h>
-#include <unistd.h>
-
-/* glibc defines the reboot() API call, which is a wrapper around the system call of the same name, but without the
- * extra "arg" parameter. Since we need that parameter for some calls, let's add a "raw" wrapper that is defined the
- * same way, except it takes the additional argument. */
-
-static inline int raw_reboot(int cmd, const void *arg) {
-        return (int) syscall(SYS_reboot, LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, cmd, arg);
-}

src/basic/signal-util.c

@@ -4,7 +4,6 @@
 #include <unistd.h>
 
 #include "errno-util.h"
-#include "missing_syscall.h"
 #include "parse-util.h"
 #include "signal-util.h"
 #include "stdio-util.h"

src/basic/stat-util.c

@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
 #include <fcntl.h>
+#include <linux/magic.h>
 #include <sys/statvfs.h>
 #include <unistd.h>
 
@@ -13,7 +14,6 @@
 #include "fs-util.h"
 #include "hash-funcs.h"
 #include "log.h"
-#include "missing_magic.h"
 #include "mountpoint-util.h"
 #include "path-util.h"
 #include "siphash24.h"

src/basic/terminal-util.c

@@ -2,6 +2,7 @@
 
 #include <fcntl.h>
 #include <linux/kd.h>
+#include <linux/magic.h>
 #include <linux/tiocl.h>
 #include <linux/vt.h>
 #include <poll.h>
@@ -27,7 +28,6 @@
 #include "inotify-util.h"
 #include "io-util.h"
 #include "log.h"
-#include "missing_magic.h"
 #include "namespace-util.h"
 #include "parse-util.h"
 #include "path-util.h"

src/basic/xattr-util.c

@@ -8,7 +8,6 @@
 #include "errno-util.h"
 #include "fd-util.h"
 #include "fs-util.h"
-#include "missing_syscall.h"
 #include "nulstr-util.h"
 #include "parse-util.h"
 #include "sparse-endian.h"

src/core/bpf-devices.c

@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
 #include <fnmatch.h>
+#include <linux/bpf.h>
 #include <linux/bpf_insn.h>
 #include <sys/stat.h>
 
@@ -12,7 +13,6 @@
 #include "fd-util.h"
 #include "fileio.h"
 #include "log.h"
-#include "missing_bpf.h"
 #include "nulstr-util.h"
 #include "parse-util.h"
 #include "path-util.h"

src/core/bpf-foreign.c

@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
 #include <linux/bpf.h>
+#include <linux/magic.h>
 
 #include "alloc-util.h"
 #include "bpf-foreign.h"
@@ -8,7 +9,6 @@
 #include "cgroup.h"
 #include "hash-funcs.h"
 #include "hashmap.h"
-#include "missing_magic.h"
 #include "siphash24.h"
 #include "stat-util.h"
 #include "unit.h"

src/core/emergency-action.c

@@ -1,12 +1,10 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
-#include <sys/reboot.h>
 #include <unistd.h>
 
 #include "ansi-color.h"
 #include "emergency-action.h"
 #include "manager.h"
-#include "raw-reboot.h"
 #include "reboot-util.h"
 #include "special.h"
 #include "string-table.h"

src/core/exec-invoke.c

@@ -1,13 +1,14 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
 #include <grp.h>
-#include <linux/ioprio.h>
 #include <linux/prctl.h>
 #include <linux/sched.h>
 #include <linux/securebits.h>
 #include <poll.h>
 #include <sys/eventfd.h>
 #include <sys/ioctl.h>
+#include <sys/ioprio.h>
+#include <sys/keyctl.h>
 #include <sys/mount.h>
 #include <sys/prctl.h>
 #include <sys/statvfs.h>
@@ -54,7 +55,6 @@
 #include "journal-send.h"
 #include "manager.h"
 #include "memfd-util.h"
-#include "missing_syscall.h"
 #include "mkdir-label.h"
 #include "mount-util.h"
 #include "namespace-util.h"

src/core/meson.build

@@ -100,15 +100,34 @@ load_fragment_gperf_nulstr_c = custom_target(
         command : [awk, '-f', '@INPUT0@', '@INPUT1@'],
         capture : true)
 
-generated_sources += [load_fragment_gperf_c, load_fragment_gperf_nulstr_c]
+generate_bpf_delegate_configs = files('generate-bpf-delegate-configs.py')
+bpf_delegate_configs_inc = custom_target(
+        input : [generate_bpf_delegate_configs, bpf_delegate_sources],
+        output : 'bpf-delegate-configs.inc',
+        command : [python,
+                   generate_bpf_delegate_configs,
+                   'code',
+                   bpf_delegate_sources],
+        capture : true)
 
+bpf_delegate_xml = custom_target(
+        input : [generate_bpf_delegate_configs, bpf_delegate_sources],
+        output : 'bpf-delegate.xml',
+        command : [python,
+                   generate_bpf_delegate_configs,
+                   'doc',
+                   bpf_delegate_sources],
+        capture : true)
+man_page_depends += bpf_delegate_xml
+
+generated_sources += [load_fragment_gperf_c, load_fragment_gperf_nulstr_c, bpf_delegate_configs_inc]
+libcore_sources += [load_fragment_gperf_c, load_fragment_gperf_nulstr_c, bpf_delegate_configs_inc]
+libcore_build_dir = meson.current_build_dir()
 libcore_name = 'systemd-core-@0@'.format(shared_lib_tag)
 
 libcore_static = static_library(
         libcore_name,
         libcore_sources,
-        load_fragment_gperf_c,
-        load_fragment_gperf_nulstr_c,
         include_directories : [includes, include_directories('.')],
         implicit_include_directories : false,
         c_args : ['-fvisibility=default'],
@@ -162,6 +181,7 @@ executor_libs = get_option('link-executor-shared') ? \
                 libcore,
                 libshared,
         ] : [
+                libc_wrapper_static,
                 libcore_static,
                 libshared_static,
                 libbasic_static,

src/core/namespace.c

@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
 #include <linux/loop.h>
+#include <linux/magic.h>
 #include <sched.h>
 #include <stdio.h>
 #include <stdlib.h>
@@ -29,7 +30,6 @@
 #include "log.h"
 #include "loop-util.h"
 #include "loopback-setup.h"
-#include "missing_magic.h"
 #include "mkdir-label.h"
 #include "mount-util.h"
 #include "mountpoint-util.h"

src/core/selinux-access.c

@@ -9,6 +9,7 @@
 #include <unistd.h>
 
 #include "sd-bus.h"
+#include "sd-varlink.h"
 
 #include "alloc-util.h"
 #include "audit-fd.h"
@@ -26,6 +27,7 @@ static bool initialized = false;
 
 struct audit_info {
         sd_bus_creds *creds;
+        sd_varlink *link;
         const char *path;
         const char *cmdline;
         const char *function;
@@ -48,12 +50,23 @@ static int audit_callback(
         char uid_buf[DECIMAL_STR_MAX(uid_t) + 1] = "n/a";
         char gid_buf[DECIMAL_STR_MAX(gid_t) + 1] = "n/a";
 
+        if (audit->creds) {
+                /* DBus case */
                 if (sd_bus_creds_get_audit_login_uid(audit->creds, &login_uid) >= 0)
                         xsprintf(login_uid_buf, UID_FMT, login_uid);
                 if (sd_bus_creds_get_euid(audit->creds, &uid) >= 0)
                         xsprintf(uid_buf, UID_FMT, uid);
                 if (sd_bus_creds_get_egid(audit->creds, &gid) >= 0)
                         xsprintf(gid_buf, GID_FMT, gid);
+        }
+
+        if (audit->link) {
+                /* varlink */
+                if (sd_varlink_get_peer_uid(audit->link, &uid) >= 0)
+                        xsprintf(uid_buf, UID_FMT, uid);
+                if (sd_varlink_get_peer_gid(audit->link, &gid) >= 0)
+                        xsprintf(gid_buf, GID_FMT, gid);
+        }
 
         (void) snprintf(msgbuf, msgbufsize,
                         "auid=%s uid=%s gid=%s%s%s%s%s%s%s%s%s%s",
@@ -164,13 +177,85 @@ static int access_init(sd_bus_error *error) {
         return 1;
 }
 
+static int get_our_contexts(const Unit *unit, const char **ret_acon, const char **ret_tclass, char **ret_fcon) {
+        _cleanup_freecon_ char *fcon = NULL;
+
+        assert(ret_acon);
+        assert(ret_tclass);
+        assert(ret_fcon);
+
+        if (unit && unit->access_selinux_context) {
+                /* Nice! The unit comes with a SELinux context read from the unit file */
+                *ret_acon = unit->access_selinux_context;
+                *ret_tclass = "service";
+                *ret_fcon = NULL;
+                return 0;
+        }
+
+        /* If no unit context is known, use our own */
+
+        /* Ideally, we should call mac_selinux_get_our_label() here because it
+         * does exactly the same - call getcon_raw(). However, it involves
+         * selinux_init() which opens label DB. It was not part of the
+         * original code. I don't want to change it for now. */
+        if (getcon_raw(&fcon) < 0)
+                return log_debug_errno(errno, "SELinux getcon_raw() failed: %m");
+
+        if (!fcon)
+                return log_debug_errno(SYNTHETIC_ERRNO(EOPNOTSUPP), "SELinux returned no context of the current process");
+
+        *ret_acon = fcon;
+        *ret_tclass = "system";
+        *ret_fcon = TAKE_PTR(fcon);
+        return 0;
+}
+
+static int check_access(
+                const char *scon,
+                const char *tcon,
+                const char *tclass,
+                const char *permission,
+                struct audit_info *audit_info,
+                sd_bus_error *error) {
+
+        bool enforce = mac_selinux_enforcing();
+        int r;
+
+        assert(scon);
+        assert(tcon);
+        assert(tclass);
+        assert(permission);
+        assert(audit_info);
+        assert(audit_info->function);
+
+        r = selinux_check_access(scon, tcon, tclass, permission, &audit_info);
+        if (r < 0) {
+                errno = -(r = errno_or_else(EPERM));
+
+                if (enforce)
+                        sd_bus_error_setf(error, SD_BUS_ERROR_ACCESS_DENIED, "SELinux policy denies access: %m");
+        }
+
+        return log_selinux_enforcing_errno(
+                              r,
+                              "SELinux access check scon=%s tcon=%s tclass=%s perm=%s state=%s function=%s path=%s cmdline=%s: %m",
+                              scon,
+                              tcon,
+                              tclass,
+                              permission,
+                              enforce ? "enforcing" : "permissive",
+                              audit_info->function,
+                              empty_to_na(audit_info->path),
+                              empty_to_na(audit_info->cmdline));
+}
+
 /*
    This function communicates with the kernel to check whether or not it should
    allow the access.
    If the machine is in permissive mode it will return ok.  Audit messages will
    still be generated if the access would be denied in enforcing mode.
 */
-int mac_selinux_access_check_internal(
+int mac_selinux_access_check_bus_internal(
                 sd_bus_message *message,
                 const Unit *unit,
                 const char *permission,
@@ -216,31 +301,22 @@ int mac_selinux_access_check_internal(
         if (r < 0)
                 return r;
 
-        if (unit && unit->access_selinux_context) {
+        r = get_our_contexts(unit, &acon, &tclass, &fcon);
-                /* Nice! The unit comes with a SELinux context read from the unit file */
+        if (r < 0) {
-                acon = unit->access_selinux_context;
+                log_selinux_enforcing_errno(
-                tclass = "service";
+                                r,
-        } else {
+                                "Failed to retrieves SELinux context of current process (perm=%s)%s: %m",
-                /* If no unit context is known, use our own */
+                                permission,
-                if (getcon_raw(&fcon) < 0) {
+                                enforce ? "" : ", ignoring");
-                        log_warning_errno(errno, "SELinux getcon_raw() failed%s (perm=%s): %m",
+
-                                          enforce ? "" : ", ignoring",
-                                          permission);
                 if (!enforce)
                         return 0;
 
+                if (r == -EOPNOTSUPP)
+                        return sd_bus_error_setf(error, SD_BUS_ERROR_ACCESS_DENIED, "We appear not to have any SELinux context: %m");
+
                 return sd_bus_error_setf(error, SD_BUS_ERROR_ACCESS_DENIED, "Failed to get current context: %m");
         }
-                if (!fcon) {
-                        if (!enforce)
-                                return 0;
-
-                        return sd_bus_error_setf(error, SD_BUS_ERROR_ACCESS_DENIED, "We appear not to have any SELinux context: %m");
-                }
-
-                acon = fcon;
-                tclass = "system";
-        }
 
         (void) sd_bus_creds_get_cmdline(creds, &cmdline);
         cl = strv_join(cmdline, " ");
@@ -252,23 +328,67 @@ int mac_selinux_access_check_internal(
                 .function = function,
         };
 
-        r = selinux_check_access(scon, acon, tclass, permission, &audit_info);
+        return check_access(scon, acon, tclass, permission, &audit_info, error);
-        if (r < 0) {
-                errno = -(r = errno_or_else(EPERM));
-
-                if (enforce)
-                        sd_bus_error_setf(error, SD_BUS_ERROR_ACCESS_DENIED, "SELinux policy denies access: %m");
 }
 
-        log_full_errno_zerook(LOG_DEBUG, r,
+int mac_selinux_access_check_varlink_internal(
-                              "SELinux access check scon=%s tcon=%s tclass=%s perm=%s state=%s function=%s path=%s cmdline=%s: %m",
+                sd_varlink *link,
-                              scon, acon, tclass, permission, enforce ? "enforcing" : "permissive", function, strna(unit ? unit->fragment_path : NULL), empty_to_na(cl));
+                const Unit *unit,
-        return enforce ? r : 0;
+                const char *permission,
+                const char *function) {
+
+        _cleanup_freecon_ char *fcon = NULL, *scon = NULL;
+        const char *tclass, *acon;
+        int r;
+
+        assert(link);
+        assert(permission);
+        assert(function);
+
+        r = access_init(/* error= */ NULL);
+        if (r <= 0)
+                return log_debug_errno(r, "Failed to init SELinux: %m");
+
+        /* delay call until we checked in `access_init()` if SELinux is actually enabled */
+        bool enforce = mac_selinux_enforcing();
+
+        int fd = sd_varlink_get_fd(link);
+        if (fd < 0)
+                return log_debug_errno(fd, "Failed to get varlink peer fd: %m");
+
+        /* We should call mac_selinux_get_peer_label() here similarly to get_our_contexts().
+         * See the explanation there why not. */
+        if (getpeercon_raw(fd, &scon) < 0)
+                return log_selinux_enforcing_errno(
+                                errno,
+                                "Failed to get peer SELinux context%s: %m",
+                                enforce ? "" : ", ignoring");
+
+        if (!scon)
+                return log_selinux_enforcing_errno(
+                                SYNTHETIC_ERRNO(EOPNOTSUPP),
+                                "Peer does not have SELinux context");
+
+        r = get_our_contexts(unit, &acon, &tclass, &fcon);
+        if (r < 0)
+                return log_selinux_enforcing_errno(
+                                r,
+                                "Failed to retrieves SELinux context of current process (perm=%s)%s: %m",
+                                permission,
+                                enforce ? "" : ", ignoring");
+
+        struct audit_info audit_info = {
+                .link = link,
+                .path = unit ? unit->fragment_path : NULL,
+                .function = function,
+        };
+
+        return check_access(scon, acon, tclass, permission, &audit_info, /* error= */ NULL);
 }
 
 #else /* HAVE_SELINUX */
 
-int mac_selinux_access_check_internal(
+int mac_selinux_access_check_bus_internal(
                 sd_bus_message *message,
                 const Unit *unit,
                 const char *permission,
@@ -278,4 +398,12 @@ int mac_selinux_access_check_internal(
         return 0;
 }
 
+int mac_selinux_access_check_varlink_internal(
+                sd_varlink *link,
+                const Unit *unit,
+                const char *permission,
+                const char *function) {
+        return 0;
+}
+
 #endif /* HAVE_SELINUX */

src/core/selinux-access.h

@@ -3,10 +3,17 @@
 
 #include "core-forward.h"
 
-int mac_selinux_access_check_internal(sd_bus_message *message, const Unit *unit, const char *permission, const char *function, sd_bus_error *error);
+int mac_selinux_access_check_bus_internal(sd_bus_message *message, const Unit *unit, const char *permission, const char *function, sd_bus_error *error);
+int mac_selinux_access_check_varlink_internal(sd_varlink *link, const Unit *unit, const char *permission, const char *function);
 
 #define mac_selinux_access_check(message, permission, error) \
-        mac_selinux_access_check_internal((message), NULL, (permission), __func__, (error))
+        mac_selinux_access_check_bus_internal((message), NULL, (permission), __func__, (error))
 
 #define mac_selinux_unit_access_check(unit, message, permission, error) \
-        mac_selinux_access_check_internal((message), (unit), (permission), __func__, (error))
+        mac_selinux_access_check_bus_internal((message), (unit), (permission), __func__, (error))
+
+#define mac_selinux_access_check_varlink(link, permission) \
+        mac_selinux_access_check_varlink_internal((link), NULL, (permission), __func__)
+
+#define mac_selinux_unit_access_check_varlink(unit, link, permission) \
+        mac_selinux_access_check_bus_internal((link), (unit), (permission), __func__)

src/core/service.c

@@ -1,5 +1,6 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
+#include <linux/audit.h>
 #include <math.h>
 #include <sys/stat.h>
 #include <unistd.h>
@@ -31,7 +32,6 @@
 #include "image-policy.h"
 #include "log.h"
 #include "manager.h"
-#include "missing_audit.h"
 #include "mount-util.h"
 #include "namespace.h"
 #include "open-file.h"

src/core/varlink-manager.c

@@ -39,48 +39,7 @@ static int manager_environment_build_json(sd_json_variant **ret, const char *nam
         return sd_json_variant_new_array_strv(ret, l);
 }
 
-static int manager_context_build_json(sd_json_variant **ret, const char *name, void *userdata) {
+static int log_level_build_json(sd_json_variant **ret, const char *name, void *userdata) {
-        Manager *m = ASSERT_PTR(userdata);
-        _cleanup_strv_free_ char **taints = NULL;
-
-        taints = taint_strv();
-        if (!taints)
-                return -ENOMEM;
-
-        return sd_json_buildo(
-                        ASSERT_PTR(ret),
-                        SD_JSON_BUILD_PAIR_STRING("Version", GIT_VERSION),
-                        SD_JSON_BUILD_PAIR_STRING("Architecture", architecture_to_string(uname_architecture())),
-                        SD_JSON_BUILD_PAIR_STRING("Features", systemd_features),
-                        JSON_BUILD_PAIR_STRV_NON_EMPTY("Taints", taints),
-                        SD_JSON_BUILD_PAIR_STRV("UnitPath", m->lookup_paths.search_path),
-                        JSON_BUILD_PAIR_STRV_NON_EMPTY("Environment", m->transient_environment),
-                        SD_JSON_BUILD_PAIR_STRING("DefaultStandardOutput", exec_output_to_string(m->defaults.std_output)),
-                        SD_JSON_BUILD_PAIR_STRING("DefaultStandardError", exec_output_to_string(m->defaults.std_error)),
-                        SD_JSON_BUILD_PAIR_BOOLEAN("ServiceWatchdogs", m->service_watchdogs),
-                        JSON_BUILD_PAIR_FINITE_USEC("DefaultTimerAccuracyUSec", m->defaults.timer_accuracy_usec),
-                        JSON_BUILD_PAIR_FINITE_USEC("DefaultTimeoutStartUSec", m->defaults.timeout_start_usec),
-                        JSON_BUILD_PAIR_FINITE_USEC("DefaultTimeoutStopUSec", m->defaults.timeout_stop_usec),
-                        JSON_BUILD_PAIR_FINITE_USEC("DefaultTimeoutAbortUSec", manager_default_timeout_abort_usec(m)),
-                        JSON_BUILD_PAIR_FINITE_USEC("DefaultDeviceTimeoutUSec", m->defaults.device_timeout_usec),
-                        JSON_BUILD_PAIR_FINITE_USEC("DefaultRestartUSec", m->defaults.restart_usec),
-                        JSON_BUILD_PAIR_RATELIMIT("DefaultStartLimit", &m->defaults.start_limit),
-                        SD_JSON_BUILD_PAIR_BOOLEAN("DefaultIOAccounting", m->defaults.io_accounting),
-                        SD_JSON_BUILD_PAIR_BOOLEAN("DefaultIPAccounting", m->defaults.ip_accounting),
-                        SD_JSON_BUILD_PAIR_BOOLEAN("DefaultMemoryAccounting", m->defaults.memory_accounting),
-                        SD_JSON_BUILD_PAIR_BOOLEAN("DefaultTasksAccounting", m->defaults.tasks_accounting),
-                        SD_JSON_BUILD_PAIR_CALLBACK("DefaultLimits", rlimit_table_build_json, m->defaults.rlimit),
-                        SD_JSON_BUILD_PAIR_UNSIGNED("DefaultTasksMax", cgroup_tasks_max_resolve(&m->defaults.tasks_max)),
-                        JSON_BUILD_PAIR_FINITE_USEC("DefaultMemoryPressureThresholdUSec", m->defaults.memory_pressure_threshold_usec),
-                        SD_JSON_BUILD_PAIR_STRING("DefaultMemoryPressureWatch", cgroup_pressure_watch_to_string(m->defaults.memory_pressure_watch)),
-                        JSON_BUILD_PAIR_FINITE_USEC("TimerSlackNSec", (uint64_t) prctl(PR_GET_TIMERSLACK)),
-                        SD_JSON_BUILD_PAIR_STRING("DefaultOOMPolicy", oom_policy_to_string(m->defaults.oom_policy)),
-                        SD_JSON_BUILD_PAIR_INTEGER("DefaultOOMScoreAdjust", m->defaults.oom_score_adjust),
-                        SD_JSON_BUILD_PAIR_BOOLEAN("DefaultRestrictSUIDSGID", m->defaults.restrict_suid_sgid),
-                        SD_JSON_BUILD_PAIR_STRING("CtrlAltDelBurstAction", emergency_action_to_string(m->cad_burst_action)));
-}
-
-static int log_level_build_runtime_json(sd_json_variant **ret, const char *name, void *userdata) {
         _cleanup_(sd_json_variant_unrefp) sd_json_variant *v = NULL;
         int log_max_level = log_get_max_level();
         int r;
@@ -107,15 +66,70 @@ static int log_level_build_runtime_json(sd_json_variant **ret, const char *name,
         return 0;
 }
 
-static int manager_runtime_build_json(sd_json_variant **ret, const char *name, void *userdata) {
+static int manager_context_build_json(sd_json_variant **ret, const char *name, void *userdata) {
         Manager *m = ASSERT_PTR(userdata);
-        dual_timestamp watchdog_last_ping;
+
+        /* The main principle behind context/runtime split is the following:
+         * If it make sense to place a property into a config/unit file it belongs to Context.
+         * Otherwise it's a 'Runtime'. */
 
         return sd_json_buildo(
                         ASSERT_PTR(ret),
+                        SD_JSON_BUILD_PAIR_BOOLEAN("ShowStatus", manager_get_show_status_on(m)),
+                        JSON_BUILD_PAIR_CALLBACK_NON_NULL("LogLevel", log_level_build_json, m),
+                        SD_JSON_BUILD_PAIR_STRING("LogTarget", log_target_to_string(log_get_target())),
+                        JSON_BUILD_PAIR_CALLBACK_NON_NULL("Environment", manager_environment_build_json, m),
+                        SD_JSON_BUILD_PAIR_STRING("DefaultStandardOutput", exec_output_to_string(m->defaults.std_output)),
+                        SD_JSON_BUILD_PAIR_STRING("DefaultStandardError", exec_output_to_string(m->defaults.std_error)),
+                        SD_JSON_BUILD_PAIR_BOOLEAN("ServiceWatchdogs", m->service_watchdogs),
+                        JSON_BUILD_PAIR_FINITE_USEC("DefaultTimerAccuracyUSec", m->defaults.timer_accuracy_usec),
+                        JSON_BUILD_PAIR_FINITE_USEC("DefaultTimeoutStartUSec", m->defaults.timeout_start_usec),
+                        JSON_BUILD_PAIR_FINITE_USEC("DefaultTimeoutStopUSec", m->defaults.timeout_stop_usec),
+                        JSON_BUILD_PAIR_FINITE_USEC("DefaultTimeoutAbortUSec", manager_default_timeout_abort_usec(m)),
+                        JSON_BUILD_PAIR_FINITE_USEC("DefaultDeviceTimeoutUSec", m->defaults.device_timeout_usec),
+                        JSON_BUILD_PAIR_FINITE_USEC("DefaultRestartUSec", m->defaults.restart_usec),
+                        JSON_BUILD_PAIR_RATELIMIT("DefaultStartLimit", &m->defaults.start_limit),
+                        SD_JSON_BUILD_PAIR_BOOLEAN("DefaultIOAccounting", m->defaults.io_accounting),
+                        SD_JSON_BUILD_PAIR_BOOLEAN("DefaultIPAccounting", m->defaults.ip_accounting),
+                        SD_JSON_BUILD_PAIR_BOOLEAN("DefaultMemoryAccounting", m->defaults.memory_accounting),
+                        SD_JSON_BUILD_PAIR_BOOLEAN("DefaultTasksAccounting", m->defaults.tasks_accounting),
+                        SD_JSON_BUILD_PAIR_CALLBACK("DefaultLimits", rlimit_table_build_json, m->defaults.rlimit),
+                        SD_JSON_BUILD_PAIR_UNSIGNED("DefaultTasksMax", cgroup_tasks_max_resolve(&m->defaults.tasks_max)),
+                        JSON_BUILD_PAIR_FINITE_USEC("DefaultMemoryPressureThresholdUSec", m->defaults.memory_pressure_threshold_usec),
+                        SD_JSON_BUILD_PAIR_STRING("DefaultMemoryPressureWatch", cgroup_pressure_watch_to_string(m->defaults.memory_pressure_watch)),
+                        JSON_BUILD_PAIR_FINITE_USEC("RuntimeWatchdogUSec", manager_get_watchdog(m, WATCHDOG_RUNTIME)),
+                        JSON_BUILD_PAIR_FINITE_USEC("RebootWatchdogUSec", manager_get_watchdog(m, WATCHDOG_REBOOT)),
+                        JSON_BUILD_PAIR_FINITE_USEC("KExecWatchdogUSec", manager_get_watchdog(m, WATCHDOG_KEXEC)),
+                        JSON_BUILD_PAIR_FINITE_USEC("RuntimeWatchdogPreUSec", manager_get_watchdog(m, WATCHDOG_PRETIMEOUT)),
+                        JSON_BUILD_PAIR_STRING_NON_EMPTY("RuntimeWatchdogPreGovernor", m->watchdog_pretimeout_governor),
+                        JSON_BUILD_PAIR_STRING_NON_EMPTY("WatchdogDevice", watchdog_get_device()),
+                        JSON_BUILD_PAIR_FINITE_USEC("TimerSlackNSec", (uint64_t) prctl(PR_GET_TIMERSLACK)),
+                        SD_JSON_BUILD_PAIR_STRING("DefaultOOMPolicy", oom_policy_to_string(m->defaults.oom_policy)),
+                        SD_JSON_BUILD_PAIR_INTEGER("DefaultOOMScoreAdjust", m->defaults.oom_score_adjust),
+                        SD_JSON_BUILD_PAIR_BOOLEAN("DefaultRestrictSUIDSGID", m->defaults.restrict_suid_sgid),
+                        SD_JSON_BUILD_PAIR_STRING("CtrlAltDelBurstAction", emergency_action_to_string(m->cad_burst_action)),
+                        JSON_BUILD_PAIR_STRING_NON_EMPTY("ConfirmSpawn", manager_get_confirm_spawn(m)),
+                        JSON_BUILD_PAIR_STRING_NON_EMPTY("ControlGroup", m->cgroup_root));
+}
+
+static int manager_runtime_build_json(sd_json_variant **ret, const char *name, void *userdata) {
+        Manager *m = ASSERT_PTR(userdata);
+        dual_timestamp watchdog_last_ping;
+        _cleanup_strv_free_ char **taints = NULL;
+
+        taints = taint_strv();
+        if (!taints)
+                return -ENOMEM;
+
+        return sd_json_buildo(
+                ASSERT_PTR(ret),
+                SD_JSON_BUILD_PAIR_STRING("Version", GIT_VERSION),
+                SD_JSON_BUILD_PAIR_STRING("Architecture", architecture_to_string(uname_architecture())),
+                SD_JSON_BUILD_PAIR_STRING("Features", systemd_features),
+                JSON_BUILD_PAIR_STRV_NON_EMPTY("Taints", taints),
+                SD_JSON_BUILD_PAIR_STRV("UnitPath", m->lookup_paths.search_path),
                 SD_JSON_BUILD_PAIR_STRING("Virtualization", virtualization_to_string(detect_virtualization())),
                 SD_JSON_BUILD_PAIR_STRING("ConfidentialVirtualization", confidential_virtualization_to_string(detect_confidential_virtualization())),
-                JSON_BUILD_PAIR_STRING_NON_EMPTY("ConfirmSpawn", manager_get_confirm_spawn(m)),
                 JSON_BUILD_PAIR_DUAL_TIMESTAMP_NON_NULL("FirmwareTimestamp", &m->timestamps[MANAGER_TIMESTAMP_FIRMWARE]),
                 JSON_BUILD_PAIR_DUAL_TIMESTAMP_NON_NULL("LoaderTimestamp", &m->timestamps[MANAGER_TIMESTAMP_LOADER]),
                 JSON_BUILD_PAIR_DUAL_TIMESTAMP_NON_NULL("KernelTimestamp", &m->timestamps[MANAGER_TIMESTAMP_KERNEL]),
@@ -135,26 +149,15 @@ static int manager_runtime_build_json(sd_json_variant **ret, const char *name, v
                 JSON_BUILD_PAIR_DUAL_TIMESTAMP_NON_NULL("InitRDGeneratorsFinishTimestamp", &m->timestamps[MANAGER_TIMESTAMP_INITRD_GENERATORS_FINISH]),
                 JSON_BUILD_PAIR_DUAL_TIMESTAMP_NON_NULL("InitRDUnitsLoadStartTimestamp", &m->timestamps[MANAGER_TIMESTAMP_INITRD_UNITS_LOAD_START]),
                 JSON_BUILD_PAIR_DUAL_TIMESTAMP_NON_NULL("InitRDUnitsLoadFinishTimestamp", &m->timestamps[MANAGER_TIMESTAMP_INITRD_UNITS_LOAD_FINISH]),
-                SD_JSON_BUILD_PAIR_BOOLEAN("ShowStatus", manager_get_show_status_on(m)),
-                JSON_BUILD_PAIR_CALLBACK_NON_NULL("LogLevel", log_level_build_runtime_json, m),
-                SD_JSON_BUILD_PAIR_STRING("LogTarget", log_target_to_string(log_get_target())),
                 SD_JSON_BUILD_PAIR_UNSIGNED("NNames", hashmap_size(m->units)),
                 SD_JSON_BUILD_PAIR_UNSIGNED("NFailedUnits", set_size(m->failed_units)),
                 SD_JSON_BUILD_PAIR_UNSIGNED("NJobs", hashmap_size(m->jobs)),
                 SD_JSON_BUILD_PAIR_UNSIGNED("NInstalledJobs", m->n_installed_jobs),
                 SD_JSON_BUILD_PAIR_UNSIGNED("NFailedJobs", m->n_failed_jobs),
                 SD_JSON_BUILD_PAIR_REAL("Progress", manager_get_progress(m)),
-                JSON_BUILD_PAIR_CALLBACK_NON_NULL("Environment", manager_environment_build_json, m),
-                JSON_BUILD_PAIR_STRING_NON_EMPTY("WatchdogDevice", watchdog_get_device()),
                 JSON_BUILD_PAIR_DUAL_TIMESTAMP_NON_NULL("WatchdogLastPingTimestamp", watchdog_get_last_ping_as_dual_timestamp(&watchdog_last_ping)),
-                JSON_BUILD_PAIR_STRING_NON_EMPTY("ControlGroup", m->cgroup_root),
                 SD_JSON_BUILD_PAIR_STRING("SystemState", manager_state_to_string(manager_state(m))),
                 SD_JSON_BUILD_PAIR_UNSIGNED("ExitCode", m->return_value),
-                JSON_BUILD_PAIR_FINITE_USEC("RuntimeWatchdogUSec", manager_get_watchdog(m, WATCHDOG_RUNTIME)),
-                JSON_BUILD_PAIR_FINITE_USEC("RuntimeWatchdogPreUSec", manager_get_watchdog(m, WATCHDOG_PRETIMEOUT)),
-                JSON_BUILD_PAIR_STRING_NON_EMPTY("RuntimeWatchdogPreGovernor", m->watchdog_pretimeout_governor),
-                JSON_BUILD_PAIR_FINITE_USEC("RebootWatchdogUSec", manager_get_watchdog(m, WATCHDOG_REBOOT)),
-                JSON_BUILD_PAIR_FINITE_USEC("KExecWatchdogUSec", manager_get_watchdog(m, WATCHDOG_KEXEC)),
                 SD_JSON_BUILD_PAIR_UNSIGNED("SoftRebootsCount", m->soft_reboots_count));
 }
 

src/creds/creds.c

@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
 #include <getopt.h>
+#include <linux/magic.h>
 #include <unistd.h>
 
 #include "sd-json.h"
@@ -23,7 +24,6 @@
 #include "log.h"
 #include "main-func.h"
 #include "memory-util.h"
-#include "missing_magic.h"
 #include "pager.h"
 #include "parse-argument.h"
 #include "parse-util.h"

src/fuzz/fuzz-catalog.c

@@ -19,7 +19,7 @@ int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
         assert_se(fd >= 0);
         assert_se(write(fd, data, size) == (ssize_t) size);
 
-        (void) catalog_import_file(&h, name);
+        (void) catalog_import_file(&h, fd, name);
 
         return 0;
 }

src/home/homework-fscrypt.c

@@ -22,8 +22,6 @@
 #include "keyring-util.h"
 #include "log.h"
 #include "memory-util.h"
-#include "missing_keyctl.h"
-#include "missing_syscall.h"
 #include "mkdir.h"
 #include "mount-util.h"
 #include "nulstr-util.h"

src/home/homework-luks.c

@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
 #include <linux/loop.h>
+#include <linux/magic.h>
 #include <poll.h>
 #include <sys/file.h>
 #include <sys/ioctl.h>
@@ -45,8 +46,6 @@
 #include "keyring-util.h"
 #include "loop-util.h"
 #include "memory-util.h"
-#include "missing_magic.h"
-#include "missing_syscall.h"
 #include "mkdir.h"
 #include "mkfs-util.h"
 #include "openssl-util.h"

src/home/homework-password-cache.c

@@ -4,7 +4,6 @@
 #include "homework-password-cache.h"
 #include "keyring-util.h"
 #include "log.h"
-#include "missing_syscall.h"
 #include "string-util.h"
 #include "user-record.h"
 

src/home/homework-quota.c

@@ -1,5 +1,7 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
+#include <linux/magic.h>
+
 #include "btrfs-util.h"
 #include "errno-util.h"
 #include "fd-util.h"
@@ -7,7 +9,6 @@
 #include "homework-quota.h"
 #include "log.h"
 #include "memory-util.h"
-#include "missing_magic.h"
 #include "quota-util.h"
 #include "stat-util.h"
 #include "user-record.h"

src/home/homework.c

@@ -1,5 +1,6 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
+#include <linux/magic.h>
 #include <stdlib.h>
 #include <sys/mount.h>
 #include <unistd.h>
@@ -35,8 +36,6 @@
 #include "loop-util.h"
 #include "main-func.h"
 #include "memory-util.h"
-#include "missing_magic.h"
-#include "missing_syscall.h"
 #include "mount-util.h"
 #include "path-util.h"
 #include "recovery-key.h"

src/home/homework.h

@@ -2,11 +2,11 @@
 #pragma once
 
 #include <linux/fscrypt.h>
+#include <sys/keyctl.h>
 
 #include "sd-id128.h"
 
 #include "homework-forward.h"
-#include "missing_keyctl.h"
 #include "user-record-util.h"
 
 typedef struct HomeSetup {

src/include/meson.build

@@ -0,0 +1,45 @@
+# SPDX-License-Identifier: LGPL-2.1-or-later
+
+subdir('override/sys')
+
+# Source files that provides AF_XYZ
+af_sources = files(
+        'override/sys/socket.h',
+)
+
+# Source files that provides ARPHRD_XYZ
+arphrd_sources = files(
+        'uapi/linux/if_arp.h',
+)
+
+# Source files that provides CAP_XYZ
+cap_sources = files(
+        'uapi/linux/capability.h',
+)
+
+# Source files that provides BPF delegate options
+bpf_delegate_sources = files(
+        'uapi/linux/bpf.h',
+)
+
+# Source files that provides ETHTOOL_LINK_MODE_XYZ
+ethtool_link_mode_sources = files(
+        'uapi/linux/ethtool.h',
+)
+
+# Source files that provides IPPROTO_XYZ
+ipproto_sources = files(
+        'override/netinet/in.h',
+        'uapi/linux/in.h',
+)
+
+# Source files that provides AUDIT_XYZ
+audit_sources = files(
+        'override/linux/audit.h',
+)
+
+# Source files that provides KEY_XYZ
+keyboard_sources = files(
+        'uapi/linux/input.h',
+        'uapi/linux/input-event-codes.h',
+)

src/include/override/linux/audit.h

@@ -1,13 +1,15 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 #pragma once
 
-#include <linux/audit.h>        /* IWYU pragma: export */
+#include_next <linux/audit.h>   /* IWYU pragma: export */
 
 #if HAVE_AUDIT
 #  include <libaudit.h>         /* IWYU pragma: export */
 #endif
 
-/* We use _Static_assert() directly here instead of assert_cc()
+#include <assert.h>
+
+/* We use static_assert() directly here instead of assert_cc()
  * because if we include macro.h in this header, the invocation
  * of generate-audit_type-list.sh becomes more complex.
  */
@@ -15,17 +17,17 @@
 #ifndef AUDIT_SERVICE_START
 #  define AUDIT_SERVICE_START 1130 /* Service (daemon) start */
 #else
-_Static_assert(AUDIT_SERVICE_START == 1130, "");
+static_assert(AUDIT_SERVICE_START == 1130, "");
 #endif
 
 #ifndef AUDIT_SERVICE_STOP
 #  define AUDIT_SERVICE_STOP 1131 /* Service (daemon) stop */
 #else
-_Static_assert(AUDIT_SERVICE_STOP == 1131, "");
+static_assert(AUDIT_SERVICE_STOP == 1131, "");
 #endif
 
 #ifndef MAX_AUDIT_MESSAGE_LENGTH
 #  define MAX_AUDIT_MESSAGE_LENGTH 8970
 #else
-_Static_assert(MAX_AUDIT_MESSAGE_LENGTH == 8970, "");
+static_assert(MAX_AUDIT_MESSAGE_LENGTH == 8970, "");
 #endif

src/include/override/linux/bpf.h

@@ -1,7 +1,7 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 #pragma once
 
-#include <linux/bpf.h> /* IWYU pragma: export */
+#include_next <linux/bpf.h>     /* IWYU pragma: export */
 
 /* defined in linux/filter.h */
 /* Unconditional jumps, goto pc + off16 */

src/include/override/linux/fs.h

@@ -1,7 +1,7 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 #pragma once
 
-#include <linux/fs.h> /* IWYU pragma: export */
+#include_next <linux/fs.h>      /* IWYU pragma: export */
 
 /* Not exposed yet. Defined at fs/ext4/ext4.h */
 #ifndef EXT4_IOC_RESIZE_FS

src/include/override/linux/keyctl.h

@@ -1,9 +1,9 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 #pragma once
 
-#include <linux/keyctl.h> /* IWYU pragma: export */
+#include_next <linux/keyctl.h>  /* IWYU pragma: export */
 
-#include "forward.h"
+#include <stdint.h>
 
 /* From linux/key.h */
 #ifndef KEY_POS_VIEW
@@ -42,5 +42,5 @@ typedef int32_t key_serial_t;
 #  define KEY_OTH_SETATTR 0x00000020
 #  define KEY_OTH_ALL     0x0000003f
 #else
-assert_cc(KEY_OTH_ALL == 0x0000003f);
+static_assert(KEY_OTH_ALL == 0x0000003f, "");
 #endif

src/include/override/linux/magic.h

@@ -1,102 +1,104 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 #pragma once
 
-#include <linux/magic.h> /* IWYU pragma: export */
+#include_next <linux/magic.h>   /* IWYU pragma: export */
+
+#include <assert.h>
 
 /* Not exposed yet (4.20). Defined at ipc/mqueue.c */
 #ifndef MQUEUE_MAGIC
 #  define MQUEUE_MAGIC 0x19800202
 #else
-assert_cc(MQUEUE_MAGIC == 0x19800202);
+static_assert(MQUEUE_MAGIC == 0x19800202, "");
 #endif
 
 /* b1123ea6d3b3da25af5c8a9d843bd07ab63213f4 (4.8), dropped by 68f2736a858324c3ec852f6c2cddd9d1c777357d (v6.0) */
 #ifndef BALLOON_KVM_MAGIC
 #  define BALLOON_KVM_MAGIC 0x13661366
 #else
-assert_cc(BALLOON_KVM_MAGIC == 0x13661366);
+static_assert(BALLOON_KVM_MAGIC == 0x13661366, "");
 #endif
 
 /* 48b4800a1c6af2cdda344ea4e2c843dcc1f6afc9 (4.8), dropped by 68f2736a858324c3ec852f6c2cddd9d1c777357d (v6.0) */
 #ifndef ZSMALLOC_MAGIC
 #  define ZSMALLOC_MAGIC 0x58295829
 #else
-assert_cc(ZSMALLOC_MAGIC == 0x58295829);
+static_assert(ZSMALLOC_MAGIC == 0x58295829, "");
 #endif
 
 /* ea8157ab2ae5e914dd427e5cfab533b6da3819cd (5.3), dropped by 68f2736a858324c3ec852f6c2cddd9d1c777357d (v6.0) */
 #ifndef Z3FOLD_MAGIC
 #  define Z3FOLD_MAGIC 0x33
 #else
-assert_cc(Z3FOLD_MAGIC == 0x33);
+static_assert(Z3FOLD_MAGIC == 0x33, "");
 #endif
 
 /* fe030c9b85e6783bc52fe86449c0a4b8aa16c753 (5.5), dropped by 68f2736a858324c3ec852f6c2cddd9d1c777357d (v6.0) */
 #ifndef PPC_CMM_MAGIC
 #  define PPC_CMM_MAGIC 0xc7571590
 #else
-assert_cc(PPC_CMM_MAGIC == 0xc7571590);
+static_assert(PPC_CMM_MAGIC == 0xc7571590, "");
 #endif
 
 /* Not in mainline but included in Ubuntu */
 #ifndef SHIFTFS_MAGIC
 #  define SHIFTFS_MAGIC 0x6a656a62
 #else
-assert_cc(SHIFTFS_MAGIC == 0x6a656a62);
+static_assert(SHIFTFS_MAGIC == 0x6a656a62, "");
 #endif
 
 /* Not exposed yet. Defined at fs/fuse/control.c */
 #ifndef FUSE_CTL_SUPER_MAGIC
 #  define FUSE_CTL_SUPER_MAGIC 0x65735543
 #else
-assert_cc(FUSE_CTL_SUPER_MAGIC == 0x65735543);
+static_assert(FUSE_CTL_SUPER_MAGIC == 0x65735543, "");
 #endif
 
 /* Not exposed yet. Defined at fs/orangefs/orangefs-kernel.h */
 #ifndef ORANGEFS_DEVREQ_MAGIC
 #  define ORANGEFS_DEVREQ_MAGIC 0x20030529
 #else
-assert_cc(ORANGEFS_DEVREQ_MAGIC == 0x20030529);
+static_assert(ORANGEFS_DEVREQ_MAGIC == 0x20030529, "");
 #endif
 
 /* linux/gfs2_ondisk.h */
 #ifndef GFS2_MAGIC
 #  define GFS2_MAGIC 0x01161970
 #else
-assert_cc(GFS2_MAGIC == 0x01161970);
+static_assert(GFS2_MAGIC == 0x01161970, "");
 #endif
 
 /* Not exposed yet. Defined at fs/configfs/mount.c */
 #ifndef CONFIGFS_MAGIC
 #  define CONFIGFS_MAGIC 0x62656570
 #else
-assert_cc(CONFIGFS_MAGIC == 0x62656570);
+static_assert(CONFIGFS_MAGIC == 0x62656570, "");
 #endif
 
 /* Not exposed yet. Defined at fs/vboxsf/super.c */
 #ifndef VBOXSF_SUPER_MAGIC
 #  define VBOXSF_SUPER_MAGIC 0x786f4256
 #else
-assert_cc(VBOXSF_SUPER_MAGIC == 0x786f4256);
+static_assert(VBOXSF_SUPER_MAGIC == 0x786f4256, "");
 #endif
 
 /* Not exposed yet, internally actually called RPCAUTH_GSSMAGIC. Defined in net/sunrpc/rpc_pipe.c */
 #ifndef RPC_PIPEFS_SUPER_MAGIC
 #  define RPC_PIPEFS_SUPER_MAGIC 0x67596969
 #else
-assert_cc(RPC_PIPEFS_SUPER_MAGIC == 0x67596969);
+static_assert(RPC_PIPEFS_SUPER_MAGIC == 0x67596969, "");
 #endif
 
 /* Not exposed yet, defined at fs/ntfs/ntfs.h */
 #ifndef NTFS_SB_MAGIC
 #  define NTFS_SB_MAGIC 0x5346544e
 #else
-assert_cc(NTFS_SB_MAGIC == 0x5346544e);
+static_assert(NTFS_SB_MAGIC == 0x5346544e, "");
 #endif
 
 /* Not exposed yet, encoded literally in fs/ntfs3/super.c. */
 #ifndef NTFS3_SUPER_MAGIC
 #  define NTFS3_SUPER_MAGIC 0x7366746e
 #else
-assert_cc(NTFS3_SUPER_MAGIC == 0x7366746e);
+static_assert(NTFS3_SUPER_MAGIC == 0x7366746e, "");
 #endif

src/include/override/linux/nsfs.h

@@ -1,8 +1,7 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 #pragma once
 
-#include <linux/nsfs.h> /* IWYU pragma: export */
+#include_next <linux/nsfs.h>    /* IWYU pragma: export */
-#include <linux/types.h>
 
 /* Root namespace inode numbers, as per include/linux/proc_ns.h in the kernel source tree, since v3.8:
  * https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=98f842e675f96ffac96e6c50315790912b2812be */

src/include/override/linux/xfs.h

@@ -1,7 +1,8 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 #pragma once
 
-#include "forward.h"
+#include <linux/ioctl.h>
+#include <stdint.h>
 
 /* This is currently not exported in the public kernel headers, but the libxfs library code part of xfsprogs
  * defines it as public header */
@@ -22,7 +23,7 @@ typedef struct xfs_fsop_geom {
         uint64_t rtblocks;
         uint64_t rtextents;
         uint64_t logstart;
-        unsigned char uuid[16];
+        uint8_t  uuid[16];
         uint32_t sunit;
         uint32_t swidth;
         int32_t  version;

src/include/override/sched.h

@@ -35,3 +35,22 @@ static_assert(PF_KTHREAD == 0x00200000, "");
 #else
 static_assert(TASK_COMM_LEN == 16, "");
 #endif
+
+/* glibc does not provide clone() on ia64, only clone2(). Not only that, but it also doesn't provide a
+ * prototype, only the symbol in the shared library (it provides a prototype for clone(), but not the
+ * symbol in the shared library). */
+#if defined(__ia64__)
+int __clone2(int (*fn)(void *), void *stack_base, size_t stack_size, int flags, void *arg);
+#define HAVE_CLONE 0
+#else
+/* We know that everywhere else clone() is available, so we don't bother with a meson check (that takes time
+ * at build time) and just define it. Once the kernel drops ia64 support, we can drop this too. */
+#define HAVE_CLONE 1
+#endif
+
+/* Defined since glibc-2.41.
+ * Supported since kernel 3.14 (e6cfc0295c7d51b008999a8b13a44fb43f8685ea). */
+#if !HAVE_SCHED_SETATTR
+int missing_sched_setattr(pid_t pid, struct sched_attr *attr, unsigned flags);
+#  define sched_setattr missing_sched_setattr
+#endif

src/include/override/signal.h

@@ -0,0 +1,9 @@
+/* SPDX-License-Identifier: LGPL-2.1-or-later */
+#pragma once
+
+#include_next <signal.h>
+
+#if !HAVE_RT_TGSIGQUEUEINFO
+int missing_rt_tgsigqueueinfo(pid_t tgid, pid_t tid, int sig, siginfo_t *info);
+#  define rt_tgsigqueueinfo missing_rt_tgsigqueueinfo
+#endif

src/include/override/sys/bpf.h

@@ -0,0 +1,11 @@
+/* SPDX-License-Identifier: LGPL-2.1-or-later */
+#pragma once
+
+#include <linux/bpf.h>          /* IWYU pragma: export */
+#include <stddef.h>
+
+/* Supported since kernel v3.18 (749730ce42a2121e1c88350d69478bff3994b10a). */
+#if !HAVE_BPF
+int missing_bpf(int cmd, union bpf_attr *attr, size_t size);
+#  define bpf missing_bpf
+#endif

src/include/override/sys/generate-syscall.py

@@ -33,6 +33,20 @@ def parse_syscall_tables(filenames):
     return {filename.split('-')[-1][:-4]: parse_syscall_table(filename)
             for filename in filenames}
 
+HEADER = '''\
+/* SPDX-License-Identifier: LGPL-2.1-or-later
+ * This file is generated by src/include/override/sys/generate-syscall.py. Do not edit!
+ *
+ * Use 'ninja -C build update-syscall-tables' to download new syscall tables,
+ * and 'ninja -C build update-syscall-header' to regenerate this file.
+ */
+#pragma once
+
+#include_next <sys/syscall.h>
+
+#include <assert.h>
+'''
+
 DEF_TEMPLATE_A = '''\
 
 #ifndef __IGNORE_{syscall}
@@ -97,7 +111,7 @@ DEF_TEMPLATE_C = '''\
 /* may be an (invalid) negative number due to libseccomp, see PR 13319 */
 #  if defined __NR_{syscall} && __NR_{syscall} >= 0
 #    if defined systemd_NR_{syscall}
-_Static_assert(__NR_{syscall} == systemd_NR_{syscall}, "");
+static_assert(__NR_{syscall} == systemd_NR_{syscall}, "");
 #    endif
 #  else
 #    if defined __NR_{syscall}
@@ -128,16 +142,7 @@ def print_syscall_def(syscall, tables, out):
           file=out)
 
 def print_syscall_defs(syscalls, tables, out):
-    print('''\
+    print(HEADER, file=out)
-/* SPDX-License-Identifier: LGPL-2.1-or-later
- * This file is generated by src/basic/missing_syscalls.py. Do not edit!
- *
- * Use 'ninja -C build update-syscall-tables' to download new syscall tables,
- * and 'ninja -C build update-syscall-header' to regenerate this file.
- */
-#pragma once
-''',
-          file=out)
     print(ARCH_CHECK, file=out)
     for syscall in syscalls:
         print_syscall_def(syscall, tables, out)

src/include/override/sys/ioprio.h

@@ -0,0 +1,14 @@
+/* SPDX-License-Identifier: LGPL-2.1-or-later */
+#pragma once
+
+#include <linux/ioprio.h>       /* IWYU pragma: export */
+
+#if !HAVE_IOPRIO_GET
+int missing_ioprio_get(int which, int who);
+#  define ioprio_get missing_ioprio_get
+#endif
+
+#if !HAVE_IOPRIO_SET
+int missing_ioprio_set(int which, int who, int ioprio);
+#  define ioprio_set missing_ioprio_set
+#endif

src/include/override/sys/kcmp.h

@@ -0,0 +1,11 @@
+/* SPDX-License-Identifier: LGPL-2.1-or-later */
+#pragma once
+
+#include <linux/kcmp.h>         /* IWYU pragma: export */
+#include <sys/types.h>
+
+/* Supported since kernel v3.5 (d97b46a64674a267bc41c9e16132ee2a98c3347d). */
+#if !HAVE_KCMP
+int missing_kcmp(pid_t pid1, pid_t pid2, int type, unsigned long idx1, unsigned long idx2);
+#  define kcmp missing_kcmp
+#endif

src/include/override/sys/keyctl.h

@@ -0,0 +1,20 @@
+/* SPDX-License-Identifier: LGPL-2.1-or-later */
+#pragma once
+
+#include <linux/keyctl.h>       /* IWYU pragma: export */
+#include <stddef.h>
+
+#if !HAVE_KEYCTL
+long missing_keyctl(int cmd, unsigned long arg2, unsigned long arg3, unsigned long arg4, unsigned long arg5);
+#  define keyctl missing_keyctl
+#endif
+
+#if !HAVE_ADD_KEY
+key_serial_t missing_add_key(const char *type, const char *description, const void *payload, size_t plen, key_serial_t ringid);
+#  define add_key missing_add_key
+#endif
+
+#if !HAVE_REQUEST_KEY
+key_serial_t missing_request_key(const char *type, const char *description, const char *callout_info, key_serial_t destringid);
+#  define request_key missing_request_key
+#endif

src/include/override/sys/mempolicy.h

@@ -0,0 +1,14 @@
+/* SPDX-License-Identifier: LGPL-2.1-or-later */
+#pragma once
+
+#include <linux/mempolicy.h>    /* IWYU pragma: export */
+
+#if !HAVE_SET_MEMPOLICY
+int missing_set_mempolicy(int mode, const unsigned long *nodemask, unsigned long maxnode);
+#  define set_mempolicy missing_set_mempolicy
+#endif
+
+#if !HAVE_GET_MEMPOLICY
+int missing_get_mempolicy(int *mode, unsigned long *nodemask, unsigned long maxnode, void *addr, unsigned long flags);
+#  define get_mempolicy missing_get_mempolicy
+#endif

src/include/override/sys/meson.build

@@ -0,0 +1,41 @@
+# SPDX-License-Identifier: LGPL-2.1-or-later
+
+arch_list = [
+        'alpha',
+        'arc',
+        'arm',
+        'arm64',
+        'i386',
+        'ia64',
+        'loongarch64',
+        'm68k',
+        'mips64',
+        'mips64n32',
+        'mipso32',
+        'parisc',
+        'powerpc',
+        'powerpc64',
+        'riscv32',
+        'riscv64',
+        's390',
+        's390x',
+        'sparc',
+        'x86_64'
+]
+
+run_target(
+        'update-syscall-tables',
+        command : [update_syscall_tables_sh, meson.current_source_dir()] + arch_list)
+
+syscall_list_txt = files('syscall-list.txt')
+
+syscall_lists = []
+foreach arch: arch_list
+        syscall_lists += files('syscalls-@0@.txt'.format(arch))
+endforeach
+
+generate_syscall_py = find_program('generate-syscall.py')
+
+run_target(
+        'update-syscall-header',
+        command : [generate_syscall_py, files('syscall.h')] + syscall_lists)

src/include/override/sys/mount.h

@@ -8,10 +8,6 @@
 #include <stddef.h>
 #include <stdint.h>
 #include <sys/ioctl.h>
-#include <sys/syscall.h>
-#include <unistd.h>
-
-#include "missing_syscall_def.h"
 
 /* Since glibc-2.37 (774058d72942249f71d74e7f2b639f77184160a6), sys/mount.h includes linux/mount.h, and
  * we can safely include both headers in the same source file. However, we cannot do that with older glibc.
@@ -43,67 +39,56 @@ extern int umount2(const char *__special_file, int __flags) __THROW;
 
 /* Open the filesystem referenced by FS_NAME so it can be configured for
    mouting.  */
+/* Defined since glibc-2.36.
+ * Supported since kernel v5.2 (24dcb3d90a1f67fe08c68a004af37df059d74005). */
 #if HAVE_FSOPEN
 extern int fsopen(const char *__fs_name, unsigned int __flags) __THROW;
 #else
-static inline int missing_fsopen(const char *fsname, unsigned flags) {
+int missing_fsopen(const char *fsname, unsigned flags);
-        return syscall(__NR_fsopen, fsname, flags);
-}
 #  define fsopen missing_fsopen
 #endif
 
 /* Create a mount representation for the FD created by fsopen using
    FLAGS with ATTR_FLAGS describing how the mount is to be performed.  */
+/* Defined since glibc-2.36.
+ * Supported since kernel v5.2 (93766fbd2696c2c4453dd8e1070977e9cd4e6b6d). */
 #if HAVE_FSMOUNT
 extern int fsmount(int __fd, unsigned int __flags, unsigned int __ms_flags) __THROW;
 #else
-static inline int missing_fsmount(int fd, unsigned flags, unsigned ms_flags) {
+int missing_fsmount(int fd, unsigned flags, unsigned ms_flags);
-        return syscall(__NR_fsmount, fd, flags, ms_flags);
-}
 #  define fsmount missing_fsmount
 #endif
 
 /* Add the mounted FROM_DFD referenced by FROM_PATHNAME filesystem returned
    by fsmount in the hierarchy in the place TO_DFD reference by TO_PATHNAME
    using FLAGS.  */
+/* Defined since glibc-2.36.
+ * Supported since kernel v5.2 (2db154b3ea8e14b04fee23e3fdfd5e9d17fbc6ae). */
 #if HAVE_MOVE_MOUNT
 extern int move_mount(int __from_dfd, const char *__from_pathname, int __to_dfd, const char *__to_pathname, unsigned int flags) __THROW;
 #else
-static inline int missing_move_mount(
+int missing_move_mount(int from_dfd, const char *from_pathname, int to_dfd, const char *to_pathname, unsigned flags);
-                int from_dfd,
-                const char *from_pathname,
-                int to_dfd,
-                const char *to_pathname,
-                unsigned flags) {
-
-        return syscall(__NR_move_mount, from_dfd, from_pathname, to_dfd, to_pathname, flags);
-}
 #  define move_mount missing_move_mount
 #endif
 
 /* Set parameters and trigger CMD action on the FD context.  KEY, VALUE,
    and AUX are used depending ng of the CMD.  */
+/* Defined since glibc-2.36.
+ * Supported since kernel v5.2 (ecdab150fddb42fe6a739335257949220033b782). */
 #if HAVE_FSCONFIG
 extern int fsconfig(int __fd, unsigned int __cmd, const char *__key, const void *__value, int __aux) __THROW;
 #else
-static inline int missing_fsconfig(int fd, unsigned cmd, const char *key, const void *value, int aux) {
+int missing_fsconfig(int fd, unsigned cmd, const char *key, const void *value, int aux);
-        return syscall(__NR_fsconfig, fd, cmd, key, value, aux);
-}
 #  define fsconfig missing_fsconfig
 #endif
 
-/* Equivalent of fopen for an existing mount point.  */
-#if HAVE_FSPICK
-extern int fspick(int __dfd, const char *__path, unsigned int __flags) __THROW;
-#endif
-
 /* Open the mount point FILENAME in directory DFD using FLAGS.  */
+/* Defined since glibc-2.36.
+ * Supported since kernel v5.2 (a07b20004793d8926f78d63eb5980559f7813404). */
 #if HAVE_OPEN_TREE
 extern int open_tree(int __dfd, const char *__filename, unsigned int __flags) __THROW;
 #else
-static inline int missing_open_tree(int dfd, const char *filename, unsigned flags) {
+int missing_open_tree(int dfd, const char *filename, unsigned flags);
-        return syscall(__NR_open_tree, dfd, filename, flags);
-}
 #  define open_tree missing_open_tree
 #endif
 
@@ -112,21 +97,20 @@ static inline int missing_open_tree(int dfd, const char *filename, unsigned flag
    directory referred to by the file descriptor dirfd.  Otherwise if DFD is
    the special value AT_FDCWD then PATH is interpreted relative to the current
    working directory of the calling process.  */
+/* Defined since glibc-2.36.
+ * Supported since kernel v5.12 (2a1867219c7b27f928e2545782b86daaf9ad50bd). */
 #if HAVE_MOUNT_SETATTR
 extern int mount_setattr(int __dfd, const char *__path, unsigned int __flags, struct mount_attr *__uattr, size_t __usize) __THROW;
 #else
-static inline int missing_mount_setattr(int dfd, const char *path, unsigned flags, struct mount_attr *attr, size_t size) {
+int missing_mount_setattr(int dfd, const char *path, unsigned flags, struct mount_attr *attr, size_t size);
-        return syscall(__NR_mount_setattr, dfd, path, flags, attr, size);
-}
 #  define mount_setattr missing_mount_setattr
 #endif
 
+/* Not defined in glibc yet as of glibc-2.41.
+ * Supported since kernel v6.15 (c4a16820d90199409c9bf01c4f794e1e9e8d8fd8). */
 #if HAVE_OPEN_TREE_ATTR
 extern int open_tree_attr(int __dfd, const char *__filename, unsigned int __flags, struct mount_attr *__uattr, size_t __usize) __THROW;
 #else
-static inline int missing_open_tree_attr(int dfd, const char *filename, unsigned int flags, struct mount_attr *attr, size_t size) {
+int missing_open_tree_attr(int dfd, const char *filename, unsigned int flags, struct mount_attr *attr, size_t size);
-        return syscall(__NR_open_tree_attr, dfd, filename, flags, attr, size);
-}
-
 #  define open_tree_attr missing_open_tree_attr
 #endif