core: port unit_fork_helper_process() and unit_fork_and_watch_rm_rf() to PidRef

[thirdparty/systemd.git] / TODO

Bugfixes:

* Many manager configuration settings that are only applicable to user
  manager or system manager can be always set. It would be better to reject
  them when parsing config.

* Jun 01 09:43:02 krowka systemd[1]: Unit user@1000.service has alias user@.service.
  Jun 01 09:43:02 krowka systemd[1]: Unit user@6.service has alias user@.service.
  Jun 01 09:43:02 krowka systemd[1]: Unit user-runtime-dir@6.service has alias user-runtime-dir@.service.

External:

* Fedora: add an rpmlint check that verifies that all unit files in the RPM are listed in %systemd_post macros.

* dbus:
   - natively watch for dbus-*.service symlinks (PENDING)
   - teach dbus to activate all services it finds in /etc/systemd/services/org-*.service

* kernel: add device_type = "fb", "fbcon" to class "graphics"

* /usr/bin/service should actually show the new command line

* fedora: suggest auto-restart on failure, but not on success and not on coredump. also, ask people to think about changing the start limit logic. Also point people to RestartPreventExitStatus=, SuccessExitStatus=

* neither pkexec nor sudo initialize environ[] from the PAM environment?

* fedora: update policy to declare access mode and ownership of unit files to root:root 0644, and add an rpmlint check for it

* register catalog database signature as file magic

* zsh shell completion:
  - <command> <verb> -<TAB> should complete options, but currently does not
  - systemctl add-wants,add-requires
  - systemctl reboot --boot-loader-entry=

* systemctl status should know about 'systemd-analyze calendar ... --iterations='
* If timer has just OnInactiveSec=..., it should fire after a specified time
  after being started.

* write blog stories about:
  - hwdb: what belongs into it, lsusb
  - enabling dbus services
  - how to make changes to sysctl and sysfs attributes
  - remote access
  - how to pass throw-away units to systemd, or dynamically change properties of existing units
  - testing with Harald's awesome test kit
  - auto-restart
  - how to develop against journal browsing APIs
  - the journal HTTP iface
  - non-cgroup resource management
  - dynamic resource management with cgroups
  - refreshed, longer missions statement
  - calendar time events
  - init=/bin/sh vs. "emergency" mode, vs. "rescue" mode, vs. "multi-user" mode, vs. "graphical" mode, and the debug shell
  - how to create your own target
  - instantiated apache, dovecot and so on
  - hooking a script into various stages of shutdown/rearly booot

Regularly:

* look for close() vs. close_nointr() vs. close_nointr_nofail()

* check for strerror(r) instead of strerror(-r)

* pahole

* set_put(), hashmap_put() return values check. i.e. == 0 does not free()!

* use secure_getenv() instead of getenv() where appropriate

* link up selected blog stories from man pages and unit files Documentation= fields

Janitorial Clean-ups:

* rework mount.c and swap.c to follow proper state enumeration/deserialization
  semantics, like we do for device.c now

* get rid of prefix_roota() and similar, only use chase() and related
  calls instead.

* get rid of basename() and replace by path_extract_filename()

* Replace our fstype_is_network() with a call to libmount's mnt_fstype_is_netfs()?
  Having two lists is not nice, but maybe it's now worth making a dependency on
  libmount for something so trivial.

* drop set_free_free() and switch things over from string_hash_ops to
  string_hash_ops_free everywhere, so that destruction is implicit rather than
  explicit. Similar, for other special hashmap/set/ordered_hashmap destructors.

* generators sometimes apply C escaping and somethines specifier escaping to
  paths and similar strings they write out. Sometimes both. We should clean
  this up, and should probably always apply both, i.e. introduce
  unit_file_escape() or so, which applies both.

Deprecations and removals:

* Remove any support for booting without /usr pre-mounted in the initrd entirely.
  Update INITRD_INTERFACE.md accordingly.

* remove cgrouspv1 support EOY 2023. As per
  https://lists.freedesktop.org/archives/systemd-devel/2022-July/048120.html
  and then rework cgroupsv2 support around fds, i.e. keep one fd per active
  unit around, and always operate on that, instead of cgroup fs paths.

* drop support for kernels that lack ambient capabilities support (i.e. make
  4.3 new baseline). Then drop support for "!!" modifier for ExecStart= which
  is only supported for such old kernels.

* drop support for kernels lacking memfd_create() (i.e. make 3.17 new
  baseline), then drop all pipe() based fallbacks.

* drop support for getrandom()-less kernels. (GRND_INSECURE means once kernel
  5.6 becomes our baseline). See
  https://github.com/systemd/systemd/pull/24101#issuecomment-1193966468 for
  details. Maybe before that: at taint-flags/warn about kernels that lack
  getrandom()/environments where it is blocked.

* drop support for LOOP_CONFIGURE-less loopback block devices, once kernel
  baseline is 5.8.

* drop fd_is_mount_point() fallback mess once we can rely on
  STATX_ATTR_MOUNT_ROOT to exist i.e. kernel baseline 5.8

* Remove /dev/mem ACPI FPDT parsing when /sys/firmware/acpi/fpdt is ubiquitous.
  That requires distros to enable CONFIG_ACPI_FPDT, and have kernels v5.12 for
  x86 and v6.2 for arm.

* Once baseline is 4.13, remove support for INTERFACE_OLD= checks in "udevadm
  trigger"'s waiting logic, since we can then rely on uuid-tagged uevents

* remove remaining tpm1.2 support from sd-stub

Features:

* make systemd-pcrphase accessible via varlink so that clients can measure
  things reasonably freely, and get a proper entry in the TPM event log we
  maintain, correctly synchronized

* add some service that makes an atomic snapshot of PCR state and event log up
  to that point available, possibly even with quote by the TPM.

* encode type1 entries in some UKI section to add additional entries to the
  menu.

* extend the various CLI tools we have that output JSON already to also read
  their command to execute from JSON in varlink format, so that you can fork
  them off and talk to them fully via varlink.

* add a "varlinkctl" tool that allows interacting with varlink services from
  the shell. In particular add a "--exec=" switch, which allows specifying a
  binary to invoke to which to talk via stdin/stdout

* make tools that speak varlink over stdin/stdout trivially sockect
  activatable. i.e. once bootctl, kernel-install, systemd-measure and similar
  speak varlink make them available via a .socket unit with Accept=yes, so that
  they can be talked to via IPC out-of-process

* beef up .service units that are socket activated with Accept=yes with options
  AllowPeerUser= + AllowPeerGroup= to allow trivially simple access control
  when invoked via socket as IPC services

* when systemd-sysext learns mutable /usr/ (and systemd-confext mutable /etc/)
  then allow them to store the result in a .v/ versioned subdir, for some basic
  snapshot logic

* add a new PE binary section ".mokkeys" or so which sd-stub will insert into
  Mok keyring, by overriding/extending whatever shim sets in the EFI
  var. Benefit: we can extend the kernel module keyring at ukify time,
  i.e. without recompiling the kernel, taking an upstrem OS' kernel and adding
  a local key to it.

* PidRef conversion work:
  - pid_is_unwaited() → pidref_is_unwaited()
  - pid_is_alive() → pidref_is_alive()
  - unit_watch_pid() → unit_watch_pidref()
  - actually wait for POLLIN on pidref's pidfd in service logic
  - exec_spawn()
  - serialization of control/main pid in service, socket, mount, swap units
  - cg_pid_get_unit()
  - openpt_allocate_in_namespace()
  - scope dbus PIDs property needs to gain PIDFDs companion

* ddi must be listed as block device fstype

* measure some string via pcrphase whenever we end up booting into emergency
  mode.

* homed: add a basic form of of secrets management to homed, that stores
  secrets in $HOME somewhere, is protected by the accounts own authentication
  mechanisms. Should implement something PKCS#11-like that can be used to
  implement emulated FIDO2 in unpriv userspace on top (which should happen
  outside of homed), emulated PKCS11, and libsecrets support. Operate with a
  2nd key derived from volume key of the user, with which to wrap all
  keys. maintain keys in kernel keyring if possible.

* add ConditionSecurity=stub-measured or so that checks if we are booted with
  systemd-stub and its measurements

* sd-boot should probably measure its configuration file to PCR 5 at boot, as
  per TCG PC Client Platform Firmware Profile Spec.

* use sd-event ratelimit feature optionally for .socket units to "pause" overly
  busy sockets temporarily. (as a less drastic version of the trigger
  ratelimit)

* similar, add the same for journal stream clients that log too much

* systemd-mount sould only consider modern file systems when mounting, similar
  to systemd-dissect

* new "systemd-pcrlock" component for dealing with PCR4. Design idea:
  1. define /{etc,usr,var/lib}/pcrlock.d/<component>/<version>.pcrlock
  2. these files contain list of hashes that will be measured when component is
     run, per PCR
  3. each component involved in the boot that is deterministically measured can
     place one or more of these files in those dirs (shim, sd-boot,
     sd-stub/UKI, cryptsetup, pcrphase, pcrfs, …)
  4. since each component has its own dir, with multiple files in them, package
     such as kernels (of which there can be multiple installed at the same
     time) can be grouped together: only one of them is measured at a time.
  5. whenever a new component is added or an old one removed, or the PCR lock
     shall be relaxed or tightened the systemd-pcrlock tool is invoked.
  6. tool iterates through all these files, orders them alphabetically by
     component, then matches them up with current measurements (as per uefi
     event log), identifying by hash, accepting that the "beginning" of the
     measurements might not be recognizable.
  7. Then calculates expected PCR values starting with the "unrecognized
     head" from the event log, then continuing with all of components
     defined via the .pcrlock files (but dropping out the "recognized tail"
     from the uefi event log). (This might mean combinatorial explosion, if
     there are multiple shims, multiple sd-boot, and so on.)
  8. Generates a public/private key pair on the TPM
  9. Generates a counter object in the TPM, with a policy that allows only
     one-by-one increase with signature policy by the public/private key pair.
  10. now signs policies of all expected PCR values with the generated keypair,
      using all combinations of components defined in the .pcrlock files
      restricting it to the counter + 1.
  11. locks down the keypair with a signed policy with its own public key
  12. generates JSON file of all these policies with their signatures, drops
      them as singleton in ESP
  13. increases the counter by one.
  14. after boot sd-stub picks JSON up from ESP, passes it to userspace via
      .extra
  15. JSON contained policies can now be used to unlock disk as well as the
      public/key itself for signing further policies, as well as increment for
      the counter
  16. whenever any of the components above is added/removed new JSON file with
      signatures for counter + 1 is generated, dropped in ESP, then counter
      increased. (i.e. this means the "recognized tail" of the event log is
      deterministically swapped out)
  17. when firmware update is expected, relaxed signed policy is generated for
      next boot only valid if counter is increased (this means the
      "unrecognized head" for the event log can change without losing access)
  18. on every boot checks if releaxed policy is in effect, if so, new strict
      policy is generated and counter increased.
  Net result: Removes downgrade attack surface + Locks OS to firmware + Allows
              downgrades within bounds

* add another PE section ".fname" or so that encodes the intended filename for
  PE file, and validate that when loading add-ons and similar before using
  it. This is particularly relevant when we load multiple add-ons and want to
  sort them to apply them in a define order. The order should not be under
  control of the attacker.

* also include packaging metadata (á la
  https://systemd.io/ELF_PACKAGE_METADATA/) in our UEFI PE binaries, using the
  same JSON format.

* make "bootctl install" + "bootctl update" useful for installing shim too. For
  that introduce new dir /usr/lib/systemd/efi/extra/ which we copy mostly 1:1
  into the ESP at install time. Then make the logic smart enough so that we
  don't overwrite bootx64.efi with our own if the extra tree already contains
  one. Also, follow symlinks when copying, so that shim rpm can symlink their
  stuff into our dir (which is safe since the target ESP is generally VFAT and
  thus does not have symlinks anyway). Later, teach the update logic to look at
  the ELF package metadata (which we also should include in all PE files, see
  above) for version info in all *.EFI files, and use it to only update if
  newer.

* in sd-stub: optionally add support for a new PE section .keyring or so that
  contains additional certificates to include in the Mok keyring, extending
  what shim might have placed there. why? let's say I use "ukify" to build +
  sign my own fedora-based UKIs, and only enroll my personal lennart key via
  shim.  Then, I want to include the fedora keyring in it, so that kmods work.
  But I might not want to enroll the fedora key in shim, because this would
  also mean that the key would be in effect whenever I boot an archlinux UKI
  built the same way, signed with the same lennart key.

* resolved: take possession of some IPv6 ULA address (let's say
  fd00:5353:5353:5353:5353:5353:5353:5353), and listen on port 53 on it for the
  local stubs, so that we can make the stub available via ipv6 too.

* introduce a .microcode PE section for sd-stub which we'll pass as first initrd
  to the kernel which will then upload it to the CPU. This should be distinct
  from .initrd to guarantee right ordering. also, and maybe more importantly
  support .microcode in PE add-ons, so that a microcode update can be shipped
  independently of any kernel.

* Maybe add SwitchRootEx() as new bus call that takes env vars to set for new
  PID 1 as argument. When adding SwitchRootEx() we should maybe also add a
  flags param that allows disabling and enabling whether serialization is
  requested during switch root.

* introduce a .acpitable section for early ACPI table override

* add proper .osrel matching for PE addons. i.e. refuse applying an addon
  intended for a different OS. Take inspiration from how confext/sysext are
  matched against OS.

* use different sbat for sd-boot and sd-stub (so that people can revoke one
  without the other)

* in ukify merge sbat info from kernel (if it has any, upstream kernels so far
  dont), of sd-stub and data supplied by user. Then measure sbat too in
  sd-stub, explicitly.

* figure out what to do about credentials sealed to PCRs in kexec + soft-reboot
  scenarios. Maybe insist sealing is done additionally against some keypair in
  the TPM to which access is updated on each boot, for the next, or so?

* logind: when logging in, always take an fd to the home dir, to keep the dir
  busy, so that autofs release can never happen. (this is generally a good
  idea, and specifically works around the fact the autofs ignores busy by mount
  namespaces)

* mount most file systems with a restrictive uidmap. e.g. mount /usr/ with a
  uidmap that blocks out anything outside 0…1000 (i.e. system users) and similar.

* mount the root fs with MS_NOSUID by default, and then mount /usr/ without
  both so that suid executables can only be placed there. Do this already in
  the initrd. If /usr/ is not split out create a bind mount automatically.

* fix our various hwdb lookup keys to end with ":" again. The original idea was
  that hwdb patterns can match arbitrary fields with expressions like
  "*:foobar:*", to wildcard match both the start and the end of the string.
  This only works safely for later extensions of the string if the strings
  always end in a colon. This requires updating our udev rules, as well as
  checking if the various hwdb files are fine with that.

* mount /tmp/ and /var/tmp with a uidmap applied that blocks out "nobody" user
  among other things such as dynamic uid ranges for containers and so on. That
  way no one can create files there with these uids and we enforce they are only
  used transiently, never persistently.

* rework loopback support in fstab: when "loop" option is used, then
  instantiate a new systemd-loop@.service for the source path, set the
  lo_file_name field for it to something recognizable derived from the fstab
  line, and then generate a mount unit for it using a udev generated symlink
  based on lo_file_name.

* remove tomoyo support, it's obsolete and unmaintained apparently

* In .socket units, add ConnectStream=, ConnectDatagram=,
  ConnectSequentialPacket= that create a socket, and then *connect to* rather than
  listen on some socket. Then, add a new setting WriteData= that takes some
  base64 data that systemd will write into the socket early on. This can then
  be used to create connections to arbitrary services and issue requests into
  them, as long as the data is static. This can then be combined with the
  aforementioned journald subscription varlink service, to enable
  activation-by-message id and similar.

* .service with invalid Sockets= starts successfully.

* landlock: lock down RuntimeDirectory= via landlock, so that services lose
  ability to write anywehere else below /run/. Similar for
  StateDirectory=. Benefit would be clear delegation via unit files: services
  get the directories they get, and nothing else even if they wanted to.

* landlock: for unprivileged systemd (i.e. systemd --user), use landlock to
  implement ProtectSystem=, ProtectHome= and so on. Landlock does not require
  privs, and we can implement pretty similar behaviour. Also, maybe add a mode
  where ProtectSystem= combined with an explicit PrivateMounts=no could request
  similar behaviour for system services, too.

* Add systemd-mount@.service which is instantiated for a block device and
  invokes systemd-mount and exits. This is then useful to use in
  ENV{SYSTEMD_WANTS} in udev rules, and a bit prettier than using RUN+=

* udevd: extend memory pressure logic: also kill any idle worker processes

* SIGRTMIN+18 and memory pressure handling should still be added to: hostnamed,
  localed, oomd, timedated.

* in order to make binding to PCR 4 realistic:
  - generate one keypair "U" and store it in a tpm2 nvindex.
  - Generate another keypair "P" and store it in a second tpm2 nvindex.
  - allocate a persistent counter object "C" in the tpm2
  - Enroll all user objects (i.e. luks volumes, creds, …) to a tpm2 policy
    signed by U.
  - Lock both U and P down with a tpm2 policy signed by P (yes, P can only be
    used if a signature by P itself can be provided)
  - For regular reboots generate a signature for a restrictive PCR4 + counter C
    based policy with key P. Place signature in EFI var, so it can be found on
    next boot
  - For reboots where a firmware update is expected generate a signature with a
    more open policy against just counter C. Place signature in same EFI var.
  - Increase C whenever switching between these two signature types.
  - During early boot, use the signature from the EFI var to unlock U and P.
    Use it to generate a signature for unlocking user objects given the current
    PCR 4 value, store that away into /run somewhere, for user during the whole
    later boot.
  - When booting up automatically update the mentioned efi var so that it
    contains the restrictive signature. But also generate a signature ahead of
    time that could be used in case during the current boot we later detect we might
    need to reboot for a firmware update. Store that in /run somewhere, so that
    it can be placed in the EFI var, if needed.

* repart/gpt-auto/DDIs: maybe introduce a concept of "extension" partitions,
  that have a new type uuid and can "extend" earlier partitions, to work around
  the fact that systemd-repart can only grow the last partition defined. During
  activation we'd simply set up a dm-linear mapping to merge them again. A
  partition that is to be extended would just set a bit in the partition flags
  field to indicate that there's another extension partition to look for. The
  identifying UUID of the extension partition would be hashed in counter mode
  from the uuid of the original partition it extends. Inspiration for this is
  the "dynamic partitions" concept of new Android. This would be a minimalistic
  concept of a volume manager, with the extents it manages being exposes as GPT
  partitions. I a partition is extended multiple times they should probably
  grow exponentially in size to ensure O(log(n)) time for finding them on
  access.

* split out execute.c into new "systemd-executor" binary. Then make PID 1 fork
  that off via vfork(), and then let that executor do the hard work. Ultimately
  the executor then gets replaced by the real binary sooner or later. Reason:
  currently the intermediary "stub" process is a CoW trap that doubles memory
  usage of PID 1 on each service start. Also, strictly speaking we are not
  allowed to do NSS from the stub process yet we do anyway. Next steps would
  then be maybe use CLONE_INTO_CGROUP for the executor, given that we don't
  need glibc anymore in the stub process then. Then, switch nspawn to just be a
  frontend for this too, so that we have to ways into the executor: via unit
  files/dbus/varlin through PID1 and via cmdline/OCI through nspawn.

* sd-stub: detect if we are running with uefi console output on serial, and if so
  automatically add console= to kernel cmdline matching the same port.

* add a utility that can be used with the kernel's
  CONFIG_STATIC_USERMODEHELPER_PATH and then handles them within pid1 so that
  security, resource management and cgroup settings can be enforced properly
  for all umh processes.

* systemd-shutdown: keep sending sd_notify() status updates immediately before
  going down, in particular include the "reboot param" string.

* homed: when resizing an fs don't sync identity beforehand there might simply
  not be enough disk space for that. try to be defensive and sync only after
  resize.

* homed: if for some reason the partition ended up being much smaller than
  whole disk, recover from that, and grow it again.

* timesyncd: when saving/restoring clock try to take boot time into account.
  Specifically, along with the saved clock, store the current boot ID. When
  starting, check if the boot id matches. If so, don't do anything (we are on
  the same boot and clock just kept running anyway). If not, then read
  CLOCK_BOOTTIME (which started at boot), and add it to the saved clock
  timestamp, to compensate for the time we spent booting. If EFI timestamps are
  available, also include that in the calculation. With this we'll then only
  miss the time spent during shutdown after timesync stopped and before the
  system actually reset.

* systemd-stub: maybe store a "boot counter" in the ESP, and pass it down to
  userspace to allow ordering boots (for example in journalctl). The counter
  would be monotonically increased on every boot.

* pam_systemd_home: add module parameter to control whether to only accept
  only password or only pcks11/fido2 auth, and then use this to hook nicely
  into two of the three PAM stacks gdm provides.
  See discussion at https://github.com/authselect/authselect/pull/311

* sd-boot: make boot loader spec type #1 accept http urls in "linux"
  lines. Then, do the uefi http dance to download kernels and boot them. This
  is then useful for network boot, by embdedding a cpio with type #1 snippets
  in sd-boot, which reference remote kernels.

* maybe prohibit setuid() to the nobody user, to lock things down, via seccomp.
  the nobody is not a user any code should run under, ever, as that user would
  possibly get a lot of access to resources it really shouldn't be getting
  access to due to the userns + nfs semantics of the user. Alternatively: use
  the seccomp log action, and allow it.

* sd-boot: add a new PE section .bls or so that carries a cpio with additional
  boot loader entries (both type1 and type2). Then when initializing, find this
  section, iterate through it and populate menu with it. cpio is simple enough
  to make a parser for this reasonably robust. use same path structures as in
  the ESP. Similar add one for signature key drop-ins.

* sd-boot: also allow passing in the cpio as in the previous item via SMBIOS

* add a new EFI tool "sd-fetch" or so. It looks in a PE section ".url" for an
  URL, then downloads the file from it using UEFI HTTP APIs, and executes it.
  Usecase: provide a minimal ESP with sd-boot and a couple of these sd-fetch
  binaries in place of UKIs, and download them on-the-fly.

* maybe: systemd-loop-generator that sets up loopback devices if requested via kernel
  cmdline. usecase: include encrypted/verity root fs in UKI.

* systemd-gpt-auto-generator: add kernel cmdline option to override block
  device to dissect. also support dissecting a regular file. useccase: include
  encrypted/verity root fs in UKI.

* sd-stub: add ".bootcfg" section for kernel bootconfig data (as per
  https://docs.kernel.org/admin-guide/bootconfig.html)

* tpm2: add (optional) support for generating a local signing key from PCR 15
  state. use private key part to sign PCR 7+14 policies. stash signatures for
  expected PCR7+14 policies in EFI var. use public key part in disk encryption.
  generate new sigs whenever db/dbx/mok/mokx gets updated. that way we can
  securely bind against SecureBoot/shim state, without having to renroll
  everything on each update (but we still have to generate one sig on each
  update, but that should be robust/idempotent). needs rollback protection, as
  usual.

* Lennart: big blog story about DDIs

* Lennart: big blog story about building initrds

* Lennart: big blog story about "why systemd-boot"

* bpf: see if we can use BPF to solve the syslog message cgroup source problem:
  one idea would be to patch source sockaddr of all AF_UNIX/SOCK_DGRAM to
  implicitly contain the source cgroup id. Another idea would be to patch
  sendto()/connect()/sendmsg() sockaddr on-the-fly to use a different target
  sockaddr.

* bpf: see if we can address opportunistic inode sharing of immutable fs images
  with BPF. i.e. if bpf gives us power to hook into openat() and return a
  different inode than is requested for which we however it has same contents
  then we can use that to implement opportunistic inode sharing among DDIs:
  make all DDIs ship xattr on all reg files with a SHA256 hash. Then, also
  dictate that DDIs should come with a top-level subdir where all reg files are
  linked into by their SHA256 sum. Then, whenever an inode is opened with the
  xattr set, check bpf table to find dirs with hashes for other prior DDIs and
  try to use inode from there.

* extend the verity signature partition to permit multiple signatures for the
  same root hash, so that people can sign a single image with multiple keys.

* consider adding a new partition type, just for /opt/ for usage in system
  extensions

* gpt-auto-discovery: also use the pkcs7 signature stuff, and pass signature to
  kernel. So far we only did this for the various --image= switches, but not
  for the root fs or /usr/.

* dissection policy should enforce that unlocking can only take place by
  certain means, i.e. only via pw, only via tpm2, or only via fido, or a
  combination thereof.

* make the systemd-repart "seed" value provisionable via credentials, so that
  confidential computing environments can set it and deterministically
  enforce the uuids for partitions created, so that they can calculate PCR 15
  ahead of time.

* systemd-repart: also derive the volume key from the seed value, for the
  aforementioned purpose.

* in the initrd: derive the default machine ID to pass to the host PID 1 via
  $machine_id from the same seed credential.

* Add systemd-sysupdate-initrd.service or so that runs systemd-sysupdate in the
  initrd to bootstrap the initrd to populate the initial partitions. Some things
  to figure out:
  - Should it run on firstboot or on every boot?
  - If run on every boot, should it use the sysupdate config from the host on
    subsequent boots?

* provide an API (probably IPC) to apps to encrypt/decrypt
  credentials. usecase: allow bluez bluetooth daemon to pass pairings to initrd
  that way, without shelling out to our tools.

* revisit default PCR bindings in cryptenroll and systemd-creds. Currently they
  use PCR 7 which should contain secureboot state db/dbx. Which sounded like a
  safe bet, given that it should change only on policy changes, and not
  software updates. But that's wrong. Recent fwupd (rightfully) contains code
  for updating the dbx denylist. This means even without any active policy
  change PCR 7 might change. Hence, better idea might be in systemd-creds to
  default to PCR 15 at least if sd-stub is used (i.e. bind to system identity),
  and in cryptsetup simply the empty list? Also, PCR 14 almost certainly should
  be included as much as PCR 7 (as it contains shim's policy, which is
  certainly as relevant as PCR 7 on many systems)

* To mimic the new tpm2-measure-pcr= crypttab option add the same to veritytab
  (measuring the root hash) and integritytab (measuring the HMAC key if one is
  used)

* We should start measuring all services, containers, and system extensions we
  activate. probably into PCR 13. i.e. add --tpm2-measure-pcr= or so to
  systemd-nspawn, and MeasurePCR= to unit files. Should contain a measurement
  of the activated configuration and the image that is being activated (in case
  verity is used, hash of the root hash).

* whenever we measure something into a TPM PCR from userspace, write a record in
  TCG's "Canonical Event Log" format to some file, so that we can reason about
  how PCR values we manage came to
  be. https://trustedcomputinggroup.org/resource/canonical-event-log-format/

* bootspec: permit graceful "update" from type #2 to type #1. If both a type #1
  and a type #2 entry exist under otherwise the exact same name, then use the
  type #1 entry, and ignore the type #2 entry. This way, people can "upgrade"
  from the UKI with all parameters baked in to a Type #1 .conf file with manual
  parametrization, if needed. This matches our usual rule that admin config
  should win over vendor defaults.

* write a "search path" spec, that documents the prefixes to search in
  (i.e. the usual /etc/, /run/, /usr/lib/ dance, potentially /usr/etc/), how to
  sort found entries, how masking works and overriding.

* automatic boot assessment: add one more default success check that just waits
  for a bit after boot, and blesses the boot if the system stayed up that long.

* implement concept of "versioned" resources inside a dir, and write a spec for
  it. Make all tools in systemd, in particular
  RootImage=/RootDirectory=/--image=/--directory= implement this. Idea:
  directories ending in ".v/" indicate a directory with versioned resources in
  them. Versioned resources inside a .v dir are always named in the pattern
  <prefix>_<version>[+<tries-left>[-<tries-done>]].<suffix>

* add support for using this .v/ logic on the root fs itself: in the initrd,
  after mounting the rootfs, look for root-<arch>.v/ in the root fs, and then
  apply the logic, moving the switch root logic there.

* systemd-repart: add support for generating ISO9660 images

* systemd-repart: in addition to the existing "factory reset" mode (which
  simply empties existing partitions marked for that). add a mode where
  partitions marked for it are entirely removed. Usecase: remove secondary OS
  copy, and redundant partitions entirely, and recreate them anew.

* systemd-boot: maybe add support for collapsing menu entries of the same OS
  into one item that can be opened (like in a "tree view" UI element) or
  collapsed. If only a single OS is installed, disable this mode, but if
  multiple OSes are installed might make sense to default to it, so that user
  is not immediately bombarded with a multitude of Linux kernel versions but
  only one for each OS.

* systemd-repart: if the GPT *disk* UUID (i.e. the one global for the entire
  disk) is set to all FFFFF then use this as trigger for factory reset, in
  addition to the existing mechanisms via EFI variables and kernel command
  line. Benefit: works also on non-EFI systems, and can be requested on one
  boot, for the next.

* figure out a sane way when building UKIs how to extract SBAT data from inner
  kernel, extend it with component info, and add to outer kernel.

* systemd-sysupdate: make transport pluggable, so people can plug casync or
  similar behind it, instead of http.

* systemd-tmpfiles: add concept for conditionalizing lines on factory reset
  boot, or on first boot.

* in UKIs: add way to define allowlist of additional words that can be added to
  the kernel cmdline even in SecureBoot mode

* we probably needs .pcrpkeyrd or so as additional PE section in UKIs,
  which contains a separate public key for PCR values that only apply in the
  initrd, i.e. in the boot phase "enter-initrd". Then, consumers in userspace
  can easily bind resources to just the initrd. Similar, maybe one more for
  "enter-initrd:leave-initrd" for resources that shall be accessible only
  before unprivileged user code is allowed. (we only need this for .pcrpkey,
  not for .pcrsig, since the latter is a list of signatures anyway). With that,
  when you enroll a LUKS volume or similar, pick either the .pcrkey (for
  coverage through all phases of the boot, but excluding shutdown), the
  .pcrpkeyrd (for coverage in the initrd only) and .pcrpkeybt (for coverage
  until users are allowed to log in).

* Once the root fs LUKS volume key is measured into PCR 15, default to binding
  credentials to PCR 15 in "systemd-creds"

* add support for asymmetric LUKS2 TPM based encryption. i.e. allow preparing
  an encrypted image on some host given a public key belonging to a specific
  other host, so that only hosts possessing the private key in the TPM2 chip
  can decrypt the volume key and activate the volume. Usecase: systemd-confext
  for a central orchestrator to generate confext images securely that can only
  be activated on one specific host (which can be used for installing a bunch
  of creds in /etc/credstore/ for example). Extending on this: allow binding
  LUKS2 TPM based encryption also to the TPM2 internal clock. Net result:
  prepare a confext image that can only be activated on a specific host that
  runs a specific software in a specific time window. confext would be
  automatically invalidated outside of it.

* maybe add a "systemd-report" tool, that generates a TPM2-backed "report" of
  current system state, i.e. a combination of PCR information, local system
  time and TPM clock, running services, recent high-priority log
  messages/coredumps, system load/PSI, signed by the local TPM chip, to form an
  enhanced remote attestation quote. Usecase: a simple orchestrator could use
  this: have the report tool upload these reports every 3min somewhere. Then
  have the orchestrator collect these reports centrally over a 3min time
  window, and use them to determine what which node should now start/stop what,
  and generate a small confext for each node, that uses Uphold= to pin services
  on each node.  The confext would be encrypted using the asymmetric encryption
  proposed above, so that it can only be activated on the specific host, if the
  software is in a good state, and within a specific time frame. Then run a
  loop on each node that sends report to orchestrator and then sysupdate to
  update confext.  Orchestrator would be stateless, i.e. operate on desired
  config and collected reports in the last 3min time window only, and thus can
  be trivially scaled up since all instances of the orchestrator should come to
  the same conclusions given the same inputs of reports/desired workload info.
  Could also be used to deliver Wireguard secrets and thus to clients, thus
  permitting zero-trust networking: secrets are rolled over via confext updates,
  and via the time window TPM logic invalidated if node doesn't keep itself
  updated, or becomes corrupted in some way.

* in the initrd, once the rootfs encryption key has been measured to PCR 15,
  derive default machine ID to use from it, and pass it to host PID 1.

* tree-wide: convert as much as possible over to use sd_event_set_signal_exit(), instead
  of manually hooking into SIGINT/SIGTERM

* tree-wide: convert as much as possible over to SD_EVENT_SIGNAL_PROCMASK
  instead of manual blocking.

* sd-boot: for each installed OS, grey out older entries (i.e. all but the
  newest), to indicate they are obsolete

* automatically propagate LUKS password credential into cryptsetup from host
  (i.e. SMBIOS type #11, …), so that one can unlock LUKS via VM hypervisor
  supplied password.

* add ability to path_is_valid() to classify paths that refer to a dir from
  those which may refer to anything, and use that in various places to filter
  early. i.e. stuff ending in "/", "/." and "/.." definitely refers to a
  directory, and paths ending that way can be refused early in many contexts.

* systemd-measure: allow operating with PEM certificates in addition to PEM
  public keys when signing PCR values. SecureBoot and our Verity signatures
  operate with certificates already, hence I guess we should also just deal for
  convencience with certificates for the PCR stuff too.

* systemd-measure: add --pcrpkey-auto as an alternative to --pcrpkey=, where it
  would just use the same public key specified with --public-key= (or the one
  automatically derived from --private-key=).

* push people to use ".sysext.raw" as suffix for sysext DDIs (DDI =
  discoverable disk images, i.e. the new name for gpt disk images following the
  discoverable disk spec). [Also: just ".sysext/" for directory-based sysext]

* Add "purpose" flag to partition flags in discoverable partition spec that
  indicate if partition is intended for sysext, for portable service, for
  booting and so on. Then, when dissecting DDI allow specifying a purpose to
  use as additional search condition. Usecase: images that combined a sysext
  partition with a portable service partition in one.

* On boot, auto-generate an asymmetric key pair from the TPM,
  and use it for validating DDIs and credentials. Maybe upload it to the kernel
  keyring, so that the kernel does this validation for us for verity and kernel
  modules

* for systemd-confext: add a tool that can generate suitable DDIs with verity +
  sig using squashfs-tools-ng's library. Maybe just systemd-repart called under
  a new name with a built-in config?

* lock down acceptable encrypted credentials at boot, via simple allowlist,
  maybe on kernel command line:
  systemd.import_encrypted_creds=foobar.waldo,tmpfiles.extra to protect locked
  down kernels from credentials generated on the host with a weak kernel

* Add support for extra verity configuration options to systemd-repart (FEC,
  hash type, etc)

* chase(): take inspiration from path_extract_filename() and return
  O_DIRECTORY if input path contains trailing slash.

* chase(): refuse resolution if trailing slash is specified on input,
  but final node is not a directory

* document in boot loader spec that symlinks in XBOOTLDR/ESP are not OK even if
  non-VFAT fs is used.

* measure credentials picked up from SMBIOS to some suitable PCR

* measure GPT and LUKS headers somewhere when we use them (i.e. in
  systemd-gpt-auto-generator/systemd-repart and in systemd-cryptsetup?)

* pick up creds from EFI vars

* Add and pickup tpm2 metadata for creds structure.

* sd-boot: we probably should include all BootXY EFI variable defined boot
  entries in our menu, and then suppress ourselves. Benefit: instant
  compatibility with all other OSes which register things there, in particular
  on other disks. Always boot into them via NextBoot EFI variable, to not
  affect PCR values.

* systemd-measure tool:
  - pre-calculate PCR 12 (command line) + PCR 13 (sysext) the same way we can precalculate PCR 11

* in sd-boot: load EFI drivers from a new PE section. That way, one can have a
  "supercharged" sd-boot binary, that could carry ext4 drivers built-in.

* sd-bus: document that sd_bus_process() only returns messages that non of the
  filters/handlers installed on the connection took possession of.

* sd-device: add an API for acquiring list of child devices, given a device
  objects (i.e. all child dirents that dirs or symlinks to dirs)

* sd-device: maybe pin the sysfs dir with an fd, during the entire runtime of
  an sd_device, then always work based on that.

* add small wrapper around qemu that implements sd_notify/AF_VSOCK + machined and
  maybe some other stuff and boots it. Should implement command line roughly
  equivalent to nspawn's. Maybe be called "systemd-vmspawn". Should imply good
  settings, i.e. RNG + HyperV enlightenments. Should also result in swtpm
  instance, plus virtiofsd instances. Translate credentials into smbios type
  11 strings. Correctly translate SIGTERM into ACPI shutdown events.
  Listen to logind suspend events and turn these into suspend key pressed +
  ACPI resume events.

* maybe add new flags to gpt partition tables for rootfs and usrfs indicating
  purpose, i.e. whether something is supposed to be bootable in a VM, on
  baremetal, on an nspawn-style container, if it is a portable service image,
  or a sysext for initrd, for host os, or for portable container. Then hook
  portabled/… up to udev to watch block devices coming up with the flags set, and
  use it.

* sd-boot should look for information what to boot in SMBIOS, too, so that VM
  managers can tell sd-boot what to boot into and suchlike

* add "systemd-sysext identify" verb, that you can point on any file in /usr/
  and that determines from which overlayfs layer it originates, which image, and with
  what it was signed.

* systemd-creds: extend encryption logic to support asymmetric
  encryption/authentication. Idea: add new verb "systemd-creds public-key"
  which generates a priv/pub key pair on the TPM2 and stores the priv key
  locally in /var. It then outputs a certificate for the pub part to stdout.
  This can then be copied/taken elsewhere, and can be used for encrypting creds
  that only the host on its specific hw can decrypt. Then, support a drop-in
  dir with certificates that can be used to authenticate credentials. Flow of
  operations is then this: build image with owner certificate, then after
  boot up issue "systemd-creds public-key" to acquire pubkey of the machine.
  Then, when passing data to the machine, sign with privkey belonging to one of
  the dropped in certs and encrypted with machine pubkey, and pass to machine.
  Machine is then able to authenticate you, and confidentiality is guaranteed.

* building on top of the above, the pub/priv key pair generated on the TPM2
  should probably also one you can use to get a remote attestation quote.

* Process credentials in:
  • networkd/udevd: add a way to define additional .link, .network, .netdev files
    via the credentials logic.
  • crypttab-generator: allow defining additional crypttab-like volumes via
    credentials (similar: verity-generator, integrity-generator). Use
    fstab-generator logic as inspiration.
  • run-generator: allow defining additional commands to run via a credential
  • resolved: allow defining additional /etc/hosts entries via a credential (it
    might make sense to then synthesize a new combined /etc/hosts file in /run
    and bind mount it on /etc/hosts for other clients that want to read it.
  • repart: allow defining additional partitions via credential
  • timesyncd: pick NTP server info from credential
  • portabled: read a credential "portable.extra" or so, that takes a list of
    file system paths to enable on start.
  • make systemd-fstab-generator look for a system credential encoding root= or
    usr=
  • systemd-homed: when initializing, look for a credential
    systemd.homed.register or so with JSON user records to automatically
    register if not registered yet.  Usecase: deploy a system, and add an
    account one can directly log into.
  • in gpt-auto-generator: check partition uuids against such uuids supplied via
    sd-stub credentials. That way, we can support parallel OS installations with
    pre-built kernels.

* define a JSON format for units, separating out unit definitions from unit
  runtime state. Then, expose it:

  1. Add Describe() method to Unit D-Bus object that returns a JSON object
     about the unit.
  2. Expose this natively via Varlink, in similar style
  3. Use it when invoking binaries (i.e. make PID 1 fork off systemd-executor
     binary which reads the JSON definition and runs it), to address the cow
     trap issue and the fact that NSS is actually forbidden in
     forked-but-not-exec'ed children
  4. Add varlink API to run transient units based on provided JSON definitions

* Add SUPPORT_END_URL= field to os-release with more *actionable* information
  what to do if support ended

* pam_systemd: on interactive logins, maybe show SUPPORT_END information at
  login time, à la motd

* sd-boot: instead of unconditionally deriving the ESP to search boot loader
  spec entries in from the paths of sd-boot binary, let's optionally allow it
  to be configured on sd-boot cmdline + efi var. Usecase: embed sd-boot in the
  UEFI firmware (for example, ovmf supports that via qemu cmdline option), and
  use it to load stuff from the ESP.

* mount /var/ from initrd, so that we can apply sysext and stuff before the
  initrd transition. Specifically:
  1. There should be a var= kernel cmdline option, matching root= and usr=
  2. systemd-gpt-auto-generator should auto-mount /var if it finds it on disk
  3. mount.x-initrd mount option in fstab should be implied for /var

* implement varlink introspection

* make persistent restarts easier by adding a new setting OpenPersistentFile=
  or so, which allows opening one or more files that is "persistent" across
  service restarts, hot reboot, cold reboots (depending on configuration): the
  files are created empty on first invocation, and on subsequent invocations
  the files are reboot. The files would be backed by tmpfs, pmem or /var
  depending on desired level of persistency.

* sd-event: add ability to "chain" event sources. Specifically, add a call
  sd_event_source_chain(x, y), which will automatically enable event source y
  in oneshot mode once x is triggered. Use case: in src/core/mount.c implement
  the /proc/self/mountinfo rescan on SIGCHLD with this: whenever a SIGCHLD is
  seen, trigger the rescan defer event source automatically, and allow it to be
  dispatched *before* the SIGCHLD is handled (based on priorities). Benefit:
  dispatch order is strictly controlled by priorities again. (next step: chain
  event sources to the ratelimit being over)

* if we fork of a service with StandardOutput=journal, and it forks off a
  subprocess that quickly dies, we might not be able to identify the cgroup it
  comes from, but we can still derive that from the stdin socket its output
  came from. We apparently don't do that right now.

* add ability to set hostname with suffix derived from machine id at boot

* add PR_SET_DUMPABLE service setting

* homed/userdb: maybe define a "companion" dir for home directories where apps
  can safely put privileged stuff in. Would not be writable by the user, but
  still conceptually belong to the user. Would be included in user's quota if
  possible, even if files are not owned by UID of user. Usecase: container
  images that owned by arbitrary UIDs, and are owned/managed by the users, but
  are not directly belonging to the user's UID. Goal: we shouldn't place more
  privileged dirs inside of unprivileged dirs, and thus containers really
  should not be placed inside of traditional UNIX home dirs (which are owned by
  users themselves) but somewhere else, that is separate, but still close
  by. Inform user code about path to this companion dir via env var, so that
  container managers find it. the ~/.identity file is also a candidate for a
  file to move there, since it is managed by privileged code (i.e. homed) and
  not unprivileged code.

* given that /etc/ssh/ssh_config.d/ is a thing now, ship a drop-in for that
  that hooks up userdbctl ssh-key stuff.

* maybe add support for binding and connecting AF_UNIX sockets in the file
  system outside of the 108ch limit. When connecting, open O_PATH fd to socket
  inode first, then connect to /proc/self/fd/XYZ. When binding, create symlink
  to target dir in /tmp, and bind through it.

* add a proper concept of a "developer" mode, i.e. where cryptographic
  protections of the root OS are weakened after interactive confirmation, to
  allow hackers to allow their own stuff. idea: allow entering developer mode
  only via explicit choice in boot menu: i.e. add explicit boot menu item for
  it. When developer mode is entered, generate a key pair in the TPM2, and add
  the public part of it automatically to keychain of valid code signature keys
  on subsequent boots. Then provide a tool to sign code with the key in the
  TPM2. Ensure that boot menu item is the only way to enter developer mode, by
  binding it to locality/PCRs so that keys cannot be generated otherwise.

* services: add support for cryptographically unlocking per-service directories
  via TPM2. Specifically, for StateDirectory= (and related dirs) use fscrypt to
  set up the directory so that it can only be accessed if host and app are in
  order.

* TPM2: extend unlock policy to protect against version downgrades in signed
  policies: policy probably must take some nvram based generation counter into
  account that can only monotonically increase and can be used to invalidate
  old PCR signatures. Otherwise people could downgrade to old signed PCR sets
  whenever they want.

* update HACKING.md to suggest developing systemd with the ideas from:
  https://0pointer.net/blog/testing-my-system-code-in-usr-without-modifying-usr.html
  https://0pointer.net/blog/running-an-container-off-the-host-usr.html

* sd-event: compat wd reuse in inotify code: keep a set of removed watch
  descriptors, and clear this set piecemeal when we see the IN_IGNORED event
  for it, or when read() returns EAGAIN or on IN_Q_OVERFLOW. Then, whenever we
  see an inotify wd event check against this set, and if it is contained ignore
  the event. (to be fully correct this would have to count the occurrences, in
  case the same wd is reused multiple times before we start processing
  IN_IGNORED again)

* for vendor-built signed initrds:
  - kernel-install should be able to install encrypted creds automatically for
    machine id, root pw, rootfs uuid, resume partition uuid, and place next to
    EFI kernel, for sd-stub to pick them up. These creds should be locked to
    the TPM, and bind to the right PCR the kernel is measured to.
  - kernel-install should be able to pick up initrd sysexts automatically and
    place them next to EFI kernel, for sd-stub to pick them up.
  - systemd-fstab-generator should look for rootfs device to mount in creds
  - systemd-resume-generator should look for resume partition uuid in creds
  - sd-stub: automatically pick up microcode from ESP (/loader/microcode/*)
    and synthesize initrd from it, and measure it. Signing is not necessary, as
    microcode does that on its own. Pass as first initrd to kernel.

* Maybe extend the service protocol to support handling of some specific SIGRT
  signal for setting service log level, that carries the level via the
  sigqueue() data parameter. Enable this via unit file setting.

* sd_notify/vsock: maybe support binding to AF_VSOCK in Type=notify services,
  then passing $NOTIFY_SOCKET and $NOTIFY_GUESTCID with PID1's cid (typically
  fixed to "2", i.e. the official host cid) and the expected guest cid, for the
  two sides of the channel. The latter env var could then be used in an
  appropriate qemu cmdline. That way qemu payloads could talk sd_notify()
  directly to host service manager.

* sd-device has an API to create an sd_device object from a device id, but has
  no api to query the device id

* sd-device should return the devnum type (i.e. 'b' or 'c') via some API for an
  sd_device object, so that data passed into sd_device_new_from_devnum() can
  also be queried.

* sd-event: optionally, if per-event source rate limit is hit, downgrade
  priority, but leave enabled, and once ratelimit window is over, upgrade
  priority again. That way we can combat event source starvation without
  stopping processing events from one source entirely.

* sd-event: similar to existing inotify support add fanotify support (given
  that apparently new features in this area are only going to be added to the
  latter).

* sd-event: add 1st class event source for clock changes

* sd-event: add 1st class event source for timezone changes

* support uefi/http boots with sd-boot: instead of looking for dropin files in
  /loader/entries/ dir, look for a file /loader/entries/SHA256SUMS and use that
  as directory manifest. The file would be a standard directory listing as
  generated by GNU sha256sums.

* sd-boot: maybe add support for embedding the various auxiliary resources we
  look for right in the sd-boot binary. i.e. take inspiration from sd-stub
  logic: allow combining sd-boot via ukify with kernels to enumerate, .conf
  files, drivers, keys to enroll and so on. Then, add whatever we find that way
  to the menu. Usecase: allow building a single PE image you can boot into via
  UEFI HTTP boot.

* maybe add a new UEFI stub binary "sd-http". It works similar to sd-stub, but
  all it does is download a file from a http server, and execute it, after
  optionally checking its hash sum. idea would be: combine this "sd-http" stub
  binary with some minimal info about a URL + hash sum, plus .osrel data, and
  drop it into the unified kernel dir in the ESP. And bam you have something
  that is tiny, feels a lot like a unified kernel, but all it does is chainload
  the real kernel. benefit: downloading these stubs would be tiny and quick,
  hence cheap for enumeration.

* sysext: measure all activated sysext into a TPM PCR

* systemd-dissect: show available versions inside of a disk image, i.e. if
  multiple versions are around of the same resource, show which ones. (in other
  words: show partition labels).

* maybe add a generator that reads /proc/cmdline, looks for
  systemd.pull-raw-portable=, systemd-pull-raw-sysext= and similar switches
  that take a URL as parameter. It then generates service units for
  systemd-pull calls that download these URLs if not installed yet. usecase:
  invoke a VM or nspawn container in a way it automatically deploys/runs these
  images as OS payloads. i.e. have a generic OS image you can point to any
  payload you like, which is then downloaded, securely verified and run.

* improve scope units to support creation by pidfd instead of by PID

* deprecate cgroupsv1 further (print log message at boot)

* systemd-dissect: add --cat switch for dumping files such as /etc/os-release

* per-service sandboxing option: ProtectIds=. If used, will overmount
  /etc/machine-id and /proc/sys/kernel/random/boot_id with synthetic files, to
  make it harder for the service to identify the host. Depending on the user
  setting it should be fully randomized at invocation time, or a hash of the
  real thing, keyed by the unit name or so. Of course, there are other ways to
  get these IDs (e.g. journal) or similar ids (e.g. MAC addresses, DMI ids, CPU
  ids), so this knob would only be useful in combination with other lockdown
  options. Particularly useful for portable services, and anything else that
  uses RootDirectory= or RootImage=. (Might also over-mount
  /sys/class/dmi/id/*{uuid,serial} with /dev/null).

* kernel-install:
  - add --all switch for rerunning kernel-install for all installed kernels

* doc: prep a document explaining resolved's internal objects, i.e. Query
  vs. Question vs. Transaction vs. Stream and so on.

* doc: prep a document explaining PID 1's internal logic, i.e. transactions,
  jobs, units

* bootspec: bring UEFI and userspace enumeration of bootspec entries back into
  sync, i.e. parse out architecture field in sd-boot (currently only done in
  userspace)

* automatically ignore threaded cgroups in cg_xyz().

* add linker script that implicitly adds symbol for build ID and new coredump
  json package metadata, and use that when logging

* Enable RestrictFileSystems= for all our long-running services (similar:
  RestrictNetworkInterfaces=)

* Add systemd-analyze security checks for RestrictFileSystems= and
  RestrictNetworkInterfaces=

* cryptsetup/homed: implement TOTP authentication backed by TPM2 and its
  internal clock.

* man: rework os-release(5), and clearly separate our extension-release.d/ and
  initrd-release parts, i.e. list explicitly which fields are about what.

* sysext: before applying a sysext, do a superficial validation run so that
  things are not rearranged to wildy. I.e. protect against accidental fuckups,
  such as masking out /usr/lib/ or so. We should probably refuse if existing
  inodes are replaced by other types of inodes or so.

* userdb: when synthesizing NSS records, pick "best" password from defined
  passwords, not just the first. i.e. if there are multiple defined, prefer
  unlocked over locked and prefer non-empty over empty.

* maybe add a tool inspired by the GPT auto discovery spec that runs in the
  initrd and rearranges the rootfs hierarchy via bind mounts, if
  enabled. Specifically in some top-level dir /@auto/ it will look for
  dirs/symlinks/subvolumes that are named after their purpose, and optionally
  encode a version as well as assessment counters, and then mount them into the
  file system tree to boot into, similar to how we do that for the gpt auto
  logic. Maybe then bind mount the original root into /.superior or something
  like that (so that update tools can look there). Further discussion in this
  thread:
  https://lists.freedesktop.org/archives/systemd-devel/2021-November/047059.html
  The GPT dissection logic should automatically enable this tool whenever we
  detect a specially marked root fs (i.e introduce a new generic root gpt type
  for this, that is arch independent). The also implement this in the image
  dissection logic, so that nspawn/RootImage= and so on grok it. Maybe make
  generic enough so that it can also work for ostrees arrangements.

* if a path ending in ".auto.d/" is set for RootDirectory=/RootImage= then do a
  strverscmp() of everything inside that dir and use that. i.e. implement very
  simple version control. Also use this in systemd-nspawn --image= and so on.

* homed: while a home dir is not activated generate slightly different NSS
  records for it, that reports the home dir as "/" and the shell as some binary
  provided by us. Then, when an SSH login happens and SSH permits it our binary
  is invoked. This binary can then talk to homed and activate the homedir if
  it's not around yet, prompting the user for a password. Once that succeeded
  we'll switch to the real user record, i.e. home dir and shell, and our tool
  exec()s the latter. Net effect: ssh'ing into a homed account will just work:
  we'll neatly prompt for the homedir's password if its needed. –– Building on
  this we could take this even further: since this tool will potentially have
  access to the client's ssh-agent (if ssh-agent forwarding is enabled) we
  could implement SSH unlocking of a homedir with that: when enrolling a new
  ssh pubkey in a user record we'd ask the ssh-agent to sign some random value
  with the privkey, then use that as luks key to unlock the home dir. Will not
  work for ECDSA keys since their signatures contain a random component, but
  will work for RSA and Ed25519 keys.

* add tiny service that decrypts encrypted user records passed via initrd
  credential logic and drops them into /run where nss-systemd can pick them up,
  similar to /run/host/userdb/. Usecase: drop a root user JSON record there,
  and use it in the initrd to log in as root with locally selected password,
  for debugging purposes. Other usecase: boot into qemu with regular user
  mounted from host. maybe put this in systemd-user-sessions.service?

* drop dependency on libcap, replace by direct syscalls based on
  CapabilityQuintet we already have. (This likely allows us to drop libcap
  dep in the base OS image)

* add concept for "exitrd" as inverse of "initrd", that we can transition to at
  shutdown, and has similar security semantics. This should then take the place
  of dracut's shutdown logic. Should probably support sysexts too. Care needs
  to be taken that the resulting logic ends up in RAM, i.e. is copied out of
  on-disk storage.

* userdbd: implement an additional varlink service socket that provides the
  host user db in restricted form, then allow this to be bind mounted into
  sandboxed environments that want the host database in minimal form. All
  records would be stripped of all meta info, except the basic UID/name
  info. Then use this in portabled environments that do not use PrivateUsers=1.

* portabled: when extracting unit files and copying to system.attached, if a
  .p7s is available in the image, use it to protect the system.attached copy
  with fs-verity, so that it cannot be tampered with

* logind introduce two types of sessions: "heavy" and "light". The former would
  be our current sessions. But the latter would be a new type of session that
  is mostly the same but does not pull in user@.service or wait for it. Then,
  allow configuration which type of session is desired via pam_systemd
  parameters, and then make user@.service's session one of these "light" ones.
  People could then choose to make FTP sessions and suchlike "light" if they
  don't want the service manager to be started for that.

* /etc/veritytab: allow that the roothash column can be specified as fs path
  including a path to an AF_UNIX path, similar to how we do things with the
  keys of /etc/crypttab. That way people can store/provide the roothash
  externally and provide to us on demand only.

* we probably should extend the root verity hash of the root fs into some PCR
  on boot. (i.e. maybe add a veritytab option tpm2-measure=12 or so to measure
  it into PCR 12); Similar: we probably should extend the LUKS volume key of
  the root fs into some PCR on boot. (i.e. maybe add a crypttab option
  tpm2-measure=15 or so to measure it into PCR 15); once both are in place
  update gpt-auto-discovery to generate these by default for the partitions it
  discovers. Static vendor stuff should probably end up in PCR 12 (i.e. the
  verity hash), with local keys in PCR 15 (i.e. the encryption volume
  key). That way, we nicely distinguish resources supplied by the OS vendor
  (i.e. sysext, root verity) from those inherently local (i.e. encryption key),
  which is useful if they shall be signed separately.

* in uefi stub: query firmware regarding which PCR banks are being used, store
  that in EFI var. then use this when enrolling TPM2 in cryptsetup to verify
  that the selected PCRs actually are used by firmware.

* rework recursive read-only remount to use new mount API

* PAM: pick up authentication token from credentials

* when mounting disk images: if IMAGE_ID/IMAGE_VERSION is set in os-release
  data in the image, make sure the image filename actually matches this, so
  that images cannot be misused.

* New udev block device symlink names:
  /dev/disk/by-parttypelabel/<pttype>-<ptlabel>. Use case: if pt label is used
  as partition image version string, this is a safe way to reference a specific
  version of a specific partition type, in particular where related partitions
  are processed (e.g. verity + rootfs both named "LennartOS_0.7").

* sysupdate:
  - add fuzzing to the pattern parser
  - support casync as download mechanism
  - "systemd-sysupdate update --all" support, that iterates through all components
    defined on the host, plus all images installed into /var/lib/machines/,
    /var/lib/portable/ and so on.
  - figure out what to do about system extensions (i.e. they need to imply an
    update component, since otherwise system extenion' sysupdate.d/ files would
    override the host's update files.)
  - Allow invocation with a single transfer definition, i.e. with
    --definitions= pointing to a file rather than a dir.
  - add ability to disable implicit decompression of downloaded artifacts,
    i.e. a Compress=no option in the transfer definitions

* in sd-id128: also parse UUIDs in RFC4122 URN syntax (i.e. chop off urn:uuid: prefix)

* DynamicUser= + StateDirectory= → use uid mapping mounts, too, in order to
  make dirs appear under right UID.

* systemd-sysext: optionally, run it in initrd already, before transitioning
  into host, to open up possibility for services shipped like that.

* introduce /dev/disk/root/* symlinks that allow referencing partitions on the
  disk the rootfs is on in a reasonably secure way. (or maybe: add
  /dev/gpt-auto-{home,srv,boot,…} similar in style to /dev/gpt-auto-root as we
  already have it.

* whenever we receive fds via SCM_RIGHTS make sure none got dropped due to the
  reception limit the kernel silently enforces.

* Add service unit setting ConnectStream= which takes IP addresses and connects to them.

* Similar, Load= which takes literal data in text or base64 format, and puts it
  into a memfd, and passes that. This enables some fun stuff, such as embedding
  bash scripts in unit files, by combining Load= with ExecStart=/bin/bash
  /proc/self/fd/3

* add a ConnectSocket= setting to service unit files, that may reference a
  socket unit, and which will connect to the socket defined therein, and pass
  the resulting fd to the service program via socket activation proto.

* Add a concept of ListenStream=anonymous to socket units: listen on a socket
  that is deleted in the fs. Usecase would be with ConnectSocket= above.

* importd: support image signature verification with PKCS#7 + OpenBSD signify
  logic, as alternative to crummy gpg

* add "systemd-analyze debug" + AttachDebugger= in unit files: The former
  specifies a command to execute; the latter specifies that an already running
  "systemd-analyze debug" instance shall be contacted and execution paused
  until it gives an OK. That way, tools like gdb or strace can be safely be
  invoked on processes forked off PID 1.

* expose MS_NOSYMFOLLOW in various places

* credentials system:
  - acquire from EFI variable?
  - acquire via ask-password?
  - acquire creds via keyring?
  - pass creds via keyring?
  - pass creds via memfd?
  - acquire + decrypt creds from pkcs11?
  - make systemd-cryptsetup acquire pw via creds logic
  - make PAMName= acquire pw via creds logic
  - make macsec/wireguard code in networkd read key via creds logic
  - make gatwayd/remote read key via creds logic
  - add sd_notify() command for flushing out creds not needed anymore
  - make user manager instances create and use a user-specific key (the one in
    /var/lib is root-only) and add --user switch to systemd-creds to use it

* add tpm.target or so which is delayed until TPM2 device showed up in case
  firmware indicates there is one.

* TPM2: auto-reenroll in cryptsetup, as fallback for hosed firmware upgrades
  and such

* introduce a new group to own TPM devices

* cryptsetup: add option for automatically removing empty password slot on boot

* cryptsetup: optionally, when run during boot-up and password is never
  entered, and we are on battery power (or so), power off machine again

* cryptsetup: when waiting for FIDO2/PKCS#11 token, tell plymouth that, and
  allow plymouth to abort the waiting and enter pw instead

* make cryptsetup lower --iter-time

* cryptsetup: allow encoding key directly in /etc/crypttab, maybe with a
  "base64:" prefix. Useful in particular for pkcs11 mode.

* cryptsetup: reimplement the mkswap/mke2fs in cryptsetup-generator to use
  systemd-makefs.service instead.

* cryptsetup:
  - cryptsetup-generator: allow specification of passwords in crypttab itself
  - support rd.luks.allow-discards= kernel cmdline params in cryptsetup generator

* systemd-analyze netif that explains predictable interface (or networkctl)

* Add service setting to run a service within the specified VRF. i.e. do the
  equivalent of "ip vrf exec".

* special case some calls of chase() to use openat2() internally, so
  that the kernel does what we otherwise do.

* add a new flag to chase() that stops chasing once the first missing
  component is found and then allows the caller to create the rest.

* make use of new glibc 2.32 APIs sigabbrev_np() and strerrorname_np().

* if /usr/bin/swapoff fails due to OOM, log a friendly explanatory message about it

* pid1: support new clone3() fork-into-cgroup feature

* pid1: also remove PID files of a service when the service starts, not just
  when it exits

* make us use dynamically fewer deps for containers in general purpose distros:
  o turn into dlopen() deps:
    - kmod-libs (only when called from PID 1)
    - libblkid (only in RootImage= handling in PID 1, but not elsewhere)
    - libpam (only when called from PID 1)
    - bzip2, xz, lz4 (always — gzip and zstd should probably stay static deps the way they are,
      since they are so basic and our defaults)
  o move into separate libsystemd-shared-iptables.so .so
    - iptables-libs (only used by nspawn + networkd)

* seccomp: maybe use seccomp_merge() to merge our filters per-arch if we can.
  Apparently kernel performance is much better with fewer larger seccomp
  filters than with more smaller seccomp filters.

* systemd-path: add ESP and XBOOTLDR path. Add "private" runtime/state/cache dir enum,
  mapping to $RUNTIME_DIRECTORY, $STATE_DIRECTORY and such

* seccomp: by default mask x32 ABI system wide on x86-64. it's on its way out

* seccomp: don't install filters for ABIs that are masked anyway for the
  specific service

* busctl: maybe expose a verb "ping" for pinging a dbus service to see if it
  exists and responds.

* socket units: allow creating a udev monitor socket with ListenDevices= or so,
  with matches, then activate app through that passing socket over

* unify on openssl:
  - kill gnutls support in resolved
  - figure out what to do about libmicrohttpd, which has a hard dependency on
    gnutls
  - port fsprg over to a dlopen lib, then switch it to openssl

* add growvol and makevol options for /etc/crypttab, similar to
  x-systemd.growfs and x-systemd-makefs.

* userdb: allow username prefix searches in varlink API, allow realname and
  realname substr searches in varlink API

* userdb: allow uid/gid range checks

* userdb: allow existence checks

* pid1: activation by journal search expression

* when switching root from initrd to host, set the machine_id env var so that
  if the host has no machine ID set yet we continue to use the random one the
  initrd had set.

* sd-event: add native support for P_ALL waitid() watching, then move PID 1 to
  it for reaping assigned but unknown children. This needs to some special care
  to operate somewhat sensibly in light of priorities: P_ALL will return
  arbitrary processes, regardless of the priority we want to watch them with,
  hence on each event loop iteration check all processes which we shall watch
  with higher prio explicitly, and then watch the entire rest with P_ALL.

* tweak sd-event's child watching: keep a prioq of children to watch and use
  waitid() only on the children with the highest priority until one is waitable
  and ignore all lower-prio ones from that point on

* maybe introduce xattrs that can be set on the root dir of the root fs
  partition that declare the volatility mode to use the image in. Previously I
  thought marking this via GPT partition flags but that's not ideal since
  that's outside of the LUKS encryption/verity verification, and we probably
  shouldn't operate in a volatile mode unless we got told so from a trusted
  source.

* coredump: maybe when coredumping read a new xattr from /proc/$PID/exe that
  may be used to mark a whole binary as non-coredumpable. Would fix:
  https://bugs.freedesktop.org/show_bug.cgi?id=69447

* teach parse_timestamp() timezones like the calendar spec already knows it

* beef up s2h to implement a battery watch loop: instead of entering
  hibernation unconditionally after coming back from resume make a decision
  based on the battery load level: if battery level is above a specific
  threshold, go to suspend again, only hibernate if below it. This means we'd
  stick to suspend usually, but fall back to hibernation only when battery runs
  empty (well, subject to our sampling interval). Related to this, check if we
  can make ACPI _BTP (i.e. /sys/class/power_supply/*/alarm) work for us too,
  i.e. see if it can wake up machines from suspend, so that we could resume
  automatically when the system is low on power and move automatically to
  hibernation mode. (see
  https://uefi.org/sites/default/files/resources/ACPI%206_2_A_Sept29.pdf
  section 10.2.2.8 and
  https://docs.microsoft.com/en-us/windows-hardware/design/device-experiences/modern-standby-wake-sources
  at the end).

* We should probably replace /etc/rc.d/README with a symlink to doc
  content. After all it is constant vendor data.

* maybe add kernel cmdline params: to force random seed crediting

* introduce a new per-process uuid, similar to the boot id, the machine id, the
  invocation id, that is derived from process creds, specifically a hashed
  combination of AT_RANDOM + getpid() + the starttime from
  /proc/self/status. Then add these ids implicitly when logging. Deriving this
  uuid from these three things has the benefit that it can be derived easily
  from /proc/$PID/ in a stable, and unique way that changes on both fork() and
  exec().

* let's not GC a unit while its ratelimits are still pending

* when killing due to service watchdog timeout maybe detect whether target
  process is under ptracing and then log loudly and continue instead.

* make rfkill uaccess controllable by default, i.e. steal rule from
  gnome-bluetooth and friends

* make MAINPID= message reception checks even stricter: if service uses User=,
  then check sending UID and ignore message if it doesn't match the user or
  root.

* maybe trigger a uevent "change" on a device if "systemctl reload xyz.device"
  is issued.

* when importing an fs tree with machined, optionally apply userns-rec-chown

* when importing an fs tree with machined, complain if image is not an OS

* Maybe introduce a helper safe_exec() or so, which is to execve() which
  safe_fork() is to fork(). And then make revert the RLIMIT_NOFILE soft limit
  to 1K implicitly, unless explicitly opted-out.

* rework seccomp/nnp logic that even if User= is used in combination with
  a seccomp option we don't have to set NNP. For that, change uid first whil
  keeping CAP_SYS_ADMIN, then apply seccomp, the drop cap.

* when no locale is configured, default to UEFI's PlatformLang variable

* add a new syscall group "@esoteric" for more esoteric stuff such as bpf() and
  usefaultd() and make systemd-analyze check for it.

* paranoia: whenever we process passwords, call mlock() on the memory
  first. i.e. look for all places we use free_and_erasep() and
  augment them with mlock(). Also use MADV_DONTDUMP.
  Alternatively (preferably?) use memfd_secret().

* Move RestrictAddressFamily= to the new cgroup create socket

* optionally: turn on cgroup delegation for per-session scope units

* sd-boot: optionally, show boot menu when previous default boot item has
  non-zero "tries done" count

* augment CODE_FILE=, CODE_LINE= with something like CODE_BASE= or so which
  contains some identifier for the project, which allows us to include
  clickable links to source files generating these log messages. The identifier
  could be some abberviated URL prefix or so (taking inspiration from Go
  imports). For example, for systemd we could use
  CODE_BASE=github.com/systemd/systemd/blob/98b0b1123cc or so which is
  sufficient to build a link by prefixing "http://" and suffixing the
  CODE_FILE.

* Augment MESSAGE_ID with MESSAGE_BASE, in a similar fashion so that we can
  make clickable links from log messages carrying a MESSAGE_ID, that lead to
  some explanatory text online.

* maybe extend .path units to expose fanotify() per-mount change events

* When reloading configuration PID 1 should reset all its properties to the
  original defaults before calling parse_config()

* hibernate/s2h: if swap is on weird storage and refuse if so

* cgroups: use inotify to get notified when somebody else modifies cgroups
  owned by us, then log a friendly warning.

* beef up log.c with support for stripping ANSI sequences from strings, so that
  it is OK to include them in log strings. This would be particularly useful so
  that our log messages could contain clickable links for example for unit
  files and suchlike we operate on.

* importd: add ability download images for portabled + sysext

* add support for "portablectl attach http://foobar.com/waaa.raw (i.e. importd integration)

* sync dynamic uids/gids between host+portable srvice (i.e. if DynamicUser=1 is set for a service, make sure that the
  selected user is resolvable in the service even if it ships its own /etc/passwd)

* Fix DECIMAL_STR_MAX or DECIMAL_STR_WIDTH. One includes a trailing NUL, the
  other doesn't. What a disaster. Probably to exclude it.

* Check that users of inotify's IN_DELETE_SELF flag are using it properly, as
  usually IN_ATTRIB is the right way to watch deleted files, as the former only
  fires when a file is actually removed from disk, i.e. the link count drops to
  zero and is not open anymore, while the latter happens when a file is
  unlinked from any dir.

* port systemctl, busctl, … over to format-table.[ch]'s table formatters

* pid1: lock image configured with RootDirectory=/RootImage= using the usual nspawn semantics while the unit is up

* add --vacuum-xyz options to coredumpctl, matching those journalctl already has.

* add CopyFile= or so as unit file setting that may be used to copy files or
  directory trees from the host to the services RootImage= and RootDirectory=
  environment. Which we can use for /etc/machine-id and in particular
  /etc/resolv.conf. Should be smart and do something useful on read-only
  images, for example fall back to read-only bind mounting the file instead.

* show invocation ID in systemd-run output

* bypass SIGTERM state in unit files if KillSignal is SIGKILL

* add proper dbus APIs for the various sd_notify() commands, such as MAINPID=1
  and so on, which would mean we could report errors and such.

* introduce DefaultSlice= or so in system.conf that allows changing where we
  place our units by default, i.e. change system.slice to something
  else. Similar, ManagerSlice= should exist so that PID1's own scope unit could
  be moved somewhere else too. Finally machined and logind should get similar
  options so that it is possible to move user session scopes and machines to a
  different slice too by default. Usecase: people who want to put resources on
  the entire system, with the exception of one specific service. See:
  https://lists.freedesktop.org/archives/systemd-devel/2018-February/040369.html

* maybe rework get_user_creds() to query the user database if $SHELL is used
  for root, but only then.

* calenderspec: add support for week numbers and day numbers within a
  year. This would allow us to define "bi-weekly" triggers safely.

* sd-bus: add vtable flag, that may be used to request client creds implicitly
  and asynchronously before dispatching the operation

* sd-bus: parse addresses given in sd_bus_set_addresses immediately and not
  only when used. Add unit tests.

* make use of ethtool veth peer info in machined, for automatically finding out
  host-side interface pointing to the container.

* add some special mode to LogsDirectory=/StateDirectory=… that allows
  declaring these directories without necessarily pulling in deps for them, or
  creating them when starting up. That way, we could declare that
  systemd-journald writes to /var/log/journal, which could be useful when we
  doing disk usage calculations and so on.

* deprecate RootDirectoryStartOnly= in favour of a new ExecStart= prefix char

* support projid-based quota in machinectl for containers

* add a way to lock down cgroup migration: a boolean, which when set for a unit
  makes sure the processes in it can never migrate out of it

* blog about fd store and restartable services

* document Environment=SYSTEMD_LOG_LEVEL=debug drop-in in debugging document

* rework ExecOutput and ExecInput enums so that EXEC_OUTPUT_NULL loses its
  magic meaning and is no longer upgraded to something else if set explicitly.

* in the long run: permit a system with /etc/machine-id linked to /dev/null, to
  make it lose its identity, i.e. be anonymous. For this we'd have to patch
  through the whole tree to make all code deal with the case where no machine
  ID is available.

* optionally, collect cgroup resource data, and store it in per-unit RRD files,
  suitable for processing with rrdtool. Add bus API to access this data, and
  possibly implement a CPULoad property based on it.

* beef up pam_systemd to take unit file settings such as cgroups properties as
  parameters

* maybe hook up xfs/ext4 quotactl() with services? i.e. automatically manage
  the quota of the user indicated in User= via unit file settings, like the
  other resource management concepts. Would mix nicely with DynamicUser=1. Or
  alternatively, do this with projids, so that we can also cover services
  running as root. Quota should probably cover all the special dirs such as
  StateDirectory=, LogsDirectory=, CacheDirectory=, as well as RootDirectory= if it
  is set, plus the whole disk space any image configured with RootImage=.

* In DynamicUser= mode: before selecting a UID, use disk quota APIs on relevant
  disks to see if the UID is already in use.

* expose IO accounting data on the bus, show it in systemd-run --wait and log
  about it in the resource log message

* Add AddUser= setting to unit files, similar to DynamicUser=1 which however
  creates a static, persistent user rather than a dynamic, transient user. We
  can leverage code from sysusers.d for this.

* add some optional flag to ReadWritePaths= and friends, that has the effect
  that we create the dir in question when the service is started. Example:

  ReadWritePaths=:/var/lib/foobar

* Add ExecMonitor= setting. May be used multiple times. Forks off a process in
  the service cgroup, which is supposed to monitor the service, and when it
  exits the service is considered failed by its monitor.

* track the per-service PAM process properly (i.e. as an additional control
  process), so that it may be queried on the bus and everything.

* add a new "debug" job mode, that is propagated to unit_start() and for
  services results in two things: we raise SIGSTOP right before invoking
  execve() and turn off watchdog support. Then, use that to implement
  "systemd-gdb" for attaching to the start-up of any system service in its
  natural habitat.

* gpt-auto logic: support encrypted swap, add kernel cmdline option to force
  it, and honour a gpt bit about it, plus maybe a configuration file

* add a percentage syntax for TimeoutStopSec=, e.g. TimeoutStopSec=150%, and
  then use that for the setting used in user@.service. It should be understood
  relative to the configured default value.

* enable LockMLOCK to take a percentage value relative to physical memory

* Permit masking specific netlink APIs with RestrictAddressFamily=

* define gpt header bits to select volatility mode

* ProtectClock= (drops CAP_SYS_TIMES, adds seecomp filters for settimeofday, adjtimex), sets DeviceAllow o /dev/rtc

* ProtectTracing= (drops CAP_SYS_PTRACE, blocks ptrace syscall, makes /sys/kernel/tracing go away)

* ProtectMount= (drop mount/umount/pivot_root from seccomp, disallow fuse via DeviceAllow, imply Mountflags=slave)

* ProtectKeyRing= to take keyring calls away

* RemoveKeyRing= to remove all keyring entries of the specified user

* ProtectReboot= that masks reboot() and kexec_load() syscalls, prohibits kill
  on PID 1 with the relevant signals, and makes relevant files in /sys and
  /proc (such as the sysrq stuff) unavailable

* Support ReadWritePaths/ReadOnlyPaths/InaccessiblePaths in systemd --user instances
  via the new unprivileged Landlock LSM (https://landlock.io)

* make sure the ratelimit object can deal with USEC_INFINITY as way to turn off things

* in nss-systemd, if we run inside of RootDirectory= with PrivateUsers= set,
  find a way to map the User=/Group= of the service to the right name. This way
  a user/group for a service only has to exist on the host for the right
  mapping to work.

* add bus API for creating unit files in /etc, reusing the code for transient units

* add bus API to remove unit files from /etc

* add bus API to retrieve current unit file contents (i.e. implement "systemctl cat" on the bus only)

* rework fopen_temporary() to make use of open_tmpfile_linkable() (problem: the
  kernel doesn't support linkat() that replaces existing files, currently)

* transient units: don't bother with actually setting unit properties, we
  reload the unit file anyway

* optionally, also require WATCHDOG=1 notifications during service start-up and shutdown

* cache sd_event_now() result from before the first iteration...

* PID1: find a way how we can reload unit file configuration for
  specific units only, without reloading the whole of systemd

* add an explicit parser for LimitRTPRIO= that verifies
  the specified range and generates sane error messages for incorrect
  specifications.

* when we detect that there are waiting jobs but no running jobs, do something

* PID 1 should send out sd_notify("WATCHDOG=1") messages (for usage in the --user mode, and when run via nspawn)

* there's probably something wrong with having user mounts below /sys,
  as we have for debugfs. for example, src/core/mount.c handles mounts
  prefixed with /sys generally special.
  https://lists.freedesktop.org/archives/systemd-devel/2015-June/032962.html

* fstab-generator: default to tmpfs-as-root if only usr= is specified on the kernel cmdline

* docs: bring https://www.freedesktop.org/wiki/Software/systemd/MyServiceCantGetRealtime up to date

* add a job mode that will fail if a transaction would mean stopping
  running units. Use this in timedated to manage the NTP service
  state.
  https://lists.freedesktop.org/archives/systemd-devel/2015-April/030229.html

* The udev blkid built-in should expose a property that reflects
  whether media was sensed in USB CF/SD card readers. This should then
  be used to control SYSTEMD_READY=1/0 so that USB card readers aren't
  picked up by systemd unless they contain a medium. This would mirror
  the behaviour we already have for CD drives.

* hostnamectl: show root image uuid

* Find a solution for SMACK capabilities stuff:
  https://lists.freedesktop.org/archives/systemd-devel/2014-December/026188.html

* synchronize console access with BSD locks:
  https://lists.freedesktop.org/archives/systemd-devel/2014-October/024582.html

* as soon as we have sender timestamps, revisit coalescing multiple parallel daemon reloads:
  https://lists.freedesktop.org/archives/systemd-devel/2014-December/025862.html

* figure out when we can use the coarse timers

* maybe allow timer units with an empty Units= setting, so that they
  can be used for resuming the system but nothing else.

* what to do about udev db binary stability for apps? (raw access is not an option)

* exponential backoff in timesyncd when we cannot reach a server

* timesyncd: add ugly bus calls to set NTP servers per-interface, for usage by NM

* add systemd.abort_on_kill or some other such flag to send SIGABRT instead of SIGKILL
  (throughout the codebase, not only PID1)

* drop nss-myhostname in favour of nss-resolve?

* resolved:
  - mDNS/DNS-SD
        - service registration
        - service/domain/types browsing
        - avahi compat
  - DNS-SD service registration from socket units
  - resolved should optionally register additional per-interface LLMNR
    names, so that for the container case we can establish the same name
    (maybe "host") for referencing the server, everywhere.
  - allow clients to request DNSSEC for a single lookup even if DNSSEC is off (?)
  - hook up resolved with machined-based address resolution

* refcounting in sd-resolve is borked

* add new gpt type for btrfs volumes

* generator that automatically discovers btrfs subvolumes, identifies their purpose based on some xattr on them.

* a way for container managers to turn off getty starting via $container_headless= or so...

* figure out a nice way how we can let the admin know what child/sibling unit causes cgroup membership for a specific unit

* For timer units: add some mechanisms so that timer units that trigger immediately on boot do not have the services
  they run added to the initial transaction and thus confuse Type=idle.

* add bus api to query unit file's X fields.

* gpt-auto-generator:
  - Define new partition type for encrypted swap? Support probed LUKS for encrypted swap?
  - Make /home automount rather than mount?

* add generator that pulls in systemd-network from containers when
  CAP_NET_ADMIN is set, more than the loopback device is defined, even
  when it is otherwise off

* MessageQueueMessageSize= (and suchlike) should use parse_iec_size().

* implement Distribute= in socket units to allow running multiple
  service instances processing the listening socket, and open this up
  for ReusePort=

* cgroups:
  - implement per-slice CPUFairScheduling=1 switch
  - introduce high-level settings for RT budget, swappiness
  - how to reset dynamically changed unit cgroup attributes sanely?
  - when reloading configuration, apply new cgroup configuration
  - when recursively showing the cgroup hierarchy, optionally also show
    the hierarchies of child processes
- add settings for cgroup.max.descendants and cgroup.max.depth,
  maybe use them for user@.service

* transient units:
  - add field to transient units that indicate whether systemd or somebody else saves/restores its settings, for integration with libvirt

* when we detect low battery and no AC on boot, show pretty splash and refuse boot

* libsystemd-journal, libsystemd-login, libudev: add calls to easily attach these objects to sd-event event loops

* be more careful what we export on the bus as (usec_t) 0 and (usec_t) -1

* rfkill,backlight: we probably should run the load tools inside of the udev rules so that the state is properly initialized by the time other software sees it

* After coming back from hibernation reset hibernation swap partition using the /dev/snapshot ioctl APIs

* If we try to find a unit via a dangling symlink, generate a clean
  error. Currently, we just ignore it and read the unit from the search
  path anyway.

* refuse boot if /usr/lib/os-release is missing or /etc/machine-id cannot be set up

* man: the documentation of Restart= currently is very misleading and suggests the tools from ExecStartPre= might get restarted.

* load .d/*.conf dropins for device units

* There's currently no way to cancel fsck (used to be possible via C-c or c on the console)

* add option to sockets to avoid activation. Instead just drop packets/connections, see http://cyberelk.net/tim/2012/02/15/portreserve-systemd-solution/

* make sure systemd-ask-password-wall does not shutdown systemd-ask-password-console too early

* verify that the AF_UNIX sockets of a service in the fs still exist
  when we start a service in order to avoid confusion when a user
  assumes starting a service is enough to make it accessible

* Make it possible to set the keymap independently from the font on
  the kernel cmdline. Right now setting one resets also the other.

* and a dbus call to generate target from current state

* investigate whether the gnome pty helper should be moved into systemd, to provide cgroup support.

* dot output for --test showing the 'initial transaction'

* be able to specify a forced restart of service A where service B depends on, in case B
  needs to be auto-respawned?

* pid1:
  - When logging about multiple units (stopping BoundTo units, conflicts, etc.),
    log both units as UNIT=, so that journalctl -u triggers on both.
  - generate better errors when people try to set transient properties
    that are not supported...
    https://lists.freedesktop.org/archives/systemd-devel/2015-February/028076.html
  - maybe introduce WantsMountsFor=? Usecase:
    https://lists.freedesktop.org/archives/systemd-devel/2015-January/027729.html
  - recreate systemd's D-Bus private socket file on SIGUSR2
  - move PAM code into its own binary
  - when we automatically restart a service, ensure we restart its rdeps, too.
  - hide PAM options in fragment parser when compile time disabled
  - Support --test based on current system state
  - If we show an error about a unit (such as not showing up) and it has no Description string, then show a description string generated form the reverse of unit_name_mangle().
  - after deserializing sockets in socket.c we should reapply sockopts and things
  - drop PID 1 reloading, only do reexecing (difficult: Reload()
    currently is properly synchronous, Reexec() is weird, because we
    cannot delay the response properly until we are back, so instead of
    being properly synchronous we just keep open the fd and close it
    when done. That means clients do not get a successful method reply,
    but much rather a disconnect on success.
  - when breaking cycles drop sysv services first, then services from /run, then from /etc, then from /usr
  - when a bus name of a service disappears from the bus make sure to queue further activation requests
  - maybe introduce CoreScheduling=yes/no to optionally set a PR_SCHED_CORE cookie, so that all
    processes in a service's cgroup share the same cookie and are guaranteed not to share SMT cores
    with other units https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/admin-guide/hw-vuln/core-scheduling.rst

* unit files:
  - allow port=0 in .socket units
  - maybe introduce ExecRestartPre=
  - implement Register= switch in .socket units to enable registration
    in Avahi, RPC and other socket registration services.
  - allow Type=simple with PIDFile=
    https://bugzilla.redhat.com/show_bug.cgi?id=723942
  - allow writing multiple conditions in unit files on one line
  - introduce Type=pid-file
  - add a concept of RemainAfterExit= to scope units
  - Allow multiple ExecStart= for all Type= settings, so that we can cover rescue.service nicely
  - add verification of [Install] section to systemd-analyze verify

* timer units:
  - timer units should get the ability to trigger when DST changes
  - Modulate timer frequency based on battery state

* add libsystemd-password or so to query passwords during boot using the password agent logic

* clean up date formatting and parsing so that all absolute/relative timestamps we format can also be parsed

* on shutdown: move utmp, wall, audit logic all into PID 1 (or logind?), get rid of systemd-update-utmp-runlevel

* make repeated alt-ctrl-del presses printing a dump

* currently x-systemd.timeout is lost in the initrd, since crypttab is copied into dracut, but fstab is not

* add a pam module that passes the hdd passphrase into the PAM stack and then expires it, for usage by gdm auto-login.

* add a pam module that on password changes updates any LUKS slot where the password matches

* test/:
  - add unit tests for config_parse_device_allow()

* seems that when we follow symlinks to units we prefer the symlink
  destination path over /etc and /usr. We should not do that. Instead
  /etc should always override /run+/usr and also any symlink
  destination.

* when isolating, try to figure out a way how we implicitly can order
  all units we stop before the isolating unit...

* teach ConditionKernelCommandLine= globs or regexes (in order to match foobar={no,0,off})

* Add ConditionDirectoryNotEmpty= handle non-absoute paths as a search path or add
  ConditionConfigSearchPathNotEmpty= or different syntax? See the discussion starting at
  https://github.com/systemd/systemd/pull/15109#issuecomment-607740136.

* BootLoaderSpec: Define a way how an installer can figure out whether a BLS
  compliant boot loader is installed.

* think about requeuing jobs when daemon-reload is issued? usecase:
  the initrd issues a reload after fstab from the host is accessible
  and we might want to requeue the mounts local-fs acquired through
  that automatically.

* systemd-inhibit: make taking delay locks useful: support sending SIGINT or SIGTERM on PrepareForSleep()

* remove any syslog support from log.c — we probably cannot do this before split-off udev is gone for good

* shutdown logging: store to EFI var, and store to USB stick?

* merge unit_kill_common() and unit_kill_context()

* add a dependency on standard-conf.xml and other included files to man pages

* MountFlags=shared acts as MountFlags=slave right now.

* properly handle loop back mounts via fstab, especially regards to fsck/passno

* initialize the hostname from the fs label of /, if /etc/hostname does not exist?

* sd-bus:
  - EBADSLT handling
  - GetAllProperties() on a non-existing object does not result in a failure currently
  - port to sd-resolve for connecting to TCP dbus servers
  - see if we can introduce a new sd_bus_get_owner_machine_id() call to retrieve the machine ID of the machine of the bus itself
  - see if we can drop more message validation on the sending side
  - add API to clone sd_bus_message objects
  - longer term: priority inheritance
  - dbus spec updates:
       - NameLost/NameAcquired obsolete
       - path escaping
  - update systemd.special(7) to mention that dbus.socket is only about the compatibility socket now

* sd-event
  - allow multiple signal handlers per signal?
  - document chaining of signal handler for SIGCHLD and child handlers
  - define more intervals where we will shift wakeup intervals around in, 1h, 6h, 24h, ...
  - maybe support iouring as backend, so that we allow hooking read and write
    operations instead of IO ready events into event loops. See considerations
    here:
    http://blog.vmsplice.net/2020/07/rethinking-event-loop-integration-for.html

* dbus: when a unit failed to load (i.e. is in UNIT_ERROR state), we
  should be able to safely try another attempt when the bus call LoadUnit() is invoked.

* document org.freedesktop.MemoryAllocation1

* maybe do not install getty@tty1.service symlink in /etc but in /usr?

* print a nicer explanation if people use variable/specifier expansion in ExecStart= for the first word

* mount: turn dependency information from /proc/self/mountinfo into dependency information between systemd units.

* EFI:
  - honor language efi variables for default language selection (if there are any?)
  - honor timezone efi variables for default timezone selection (if there are any?)
  - change bootctl to be backed by systemd-bootd to control temporary and persistent default boot goal plus efi variables
* bootctl
  - recognize the case when not booted on EFI

* bootctl,sd-boot: actually honour the "architecture" key

* bootctl:
  - show whether UEFI audit mode is available
  - teach it to prepare an ESP wholesale, i.e. with mkfs.vfat invocation
  - teach it to copy in unified kernel images and maybe type #1 boot loader spec entries from host

* logind:
  - logind: optionally, ignore idle-hint logic for autosuspend, block suspend as long as a session is around
  - logind: wakelock/opportunistic suspend support
  - Add pretty name for seats in logind
  - logind: allow showing logout dialog from system?
  - add Suspend() bus calls which take timestamps to fix double suspend issues when somebody hits suspend and closes laptop quickly.
  - if pam_systemd is invoked by su from a process that is outside of a
    any session we should probably just become a NOP, since that's
    usually not a real user session but just some system code that just
    needs setuid().
  - logind: make the Suspend()/Hibernate() bus calls wait for the for
    the job to be completed. before returning, so that clients can wait
    for "systemctl suspend" to finish to know when the suspending is
    complete.
  - logind: when the power button is pressed short, just popup a
    logout dialog. If it is pressed for 1s, do the usual
    shutdown. Inspiration are Macs here.
  - expose "Locked" property on logind session objects
  - maybe allow configuration of the StopTimeout for session scopes
  - rename session scope so that it includes the UID. THat way
    the session scope can be arranged freely in slices and we don't have
    make assumptions about their slice anymore.
  - follow PropertiesChanged state more closely, to deal with quick logouts and
    relogins
  - (optionally?) spawn seat-manager@$SEAT.service whenever a seat shows up that as CanGraphical set
  - expose details of boot entries on the bus. In particular, it should be possible
    to query the list of boot entry titles that bootctl / sd-boot would show.
    Currently we only expose their identifiers.

* move multiseat vid/pid matches from logind udev rule to hwdb

* logind: rework pam_logind to also do a bus call in case of invocation from
  user@.service, which returns the XDG_RUNTIME_DIR value, and make this
  behaviour selectable via pam module option.

* delay activation of logind until somebody logs in, or when /dev/tty0 pulls it
  in or lingering is on (so that containers don't bother with it until PAM is used). also exit-on-idle

* journal:
  - consider introducing implicit _TTY= + _PPID= + _EUID= + _EGID= + _FSUID= + _FSGID= fields
  - journald: also get thread ID from client, plus thread name
  - journal: when waiting for journal additions in the client always sleep at least 1s or so, in order to minimize wakeups
  - add API to close/reopen/get fd for journal client fd in libsystemd-journal.
  - fall back to /dev/log based logging in libsystemd-journal, if we cannot log natively?
  - declare the local journal protocol stable in the wiki interface chart
  - sd-journal: speed up sd_journal_get_data() with transparent hash table in bg
  - journald: when dropping msgs due to ratelimit make sure to write
    "dropped %u messages" not only when we are about to print the next
    message that works, but already after a short timeout
  - check if we can make journalctl by default use --follow mode inside of less if called without args?
  - maybe add API to send pairs of iovecs via sd_journal_send
  - journal: add a setgid "systemd-journal" utility to invoke from libsystemd-journal, which passes fds via STDOUT and does PK access
  - journalctl: support negative filtering, i.e. FOOBAR!="waldo",
    and !FOOBAR for events without FOOBAR.
  - journal: store timestamp of journal_file_set_offline() in the header,
    so it is possible to display when the file was last synced.
  - journal-send.c, log.c: when the log socket is clogged, and we drop, count this and write a message about this when it gets unclogged again.
  - journal: find a way to allow dropping history early, based on priority, other rules
  - journal: When used on NFS, check payload hashes
  - journald: add kernel cmdline option to disable ratelimiting for debug purposes
  - refuse taking lower-case variable names in sd_journal_send() and friends.
  - journald: we currently rotate only after MaxUse+MaxFilesize has been reached.
  - journal: deal nicely with byte-by-byte copied files, especially regards header
  - journal: sanely deal with entries which are larger than the individual file size, but where the components would fit
  - Replace utmp, wtmp, btmp, and lastlog completely with journal
  - journalctl: instead --after-cursor= maybe have a --cursor=XYZ+1 syntax?
  - when a kernel driver logs in a tight loop, we should ratelimit that too.
  - journald: optionally, log debug messages to /run but everything else to /var
  - journald: when we drop syslog messages because the syslog socket is
    full, make sure to write how many messages are lost as first thing
    to syslog when it works again.
  - journald: allow per-priority and per-service retention times when rotating/vacuuming
  - journald: make use of uid-range.h to managed uid ranges to split
    journals in.
  - journalctl: add the ability to look for the most recent process of a binary. journalctl /usr/bin/X11 --pid=-1 or so...
  - improve journalctl performance by loading journal files
    lazily. Encode just enough information in the file name, so that we
    do not have to open it to know that it is not interesting for us, for
    the most common operations.
  - man: document that corrupted journal files is nothing to act on
  - rework journald sigbus stuff to use mutex
  - Set RLIMIT_NPROC for systemd-journal-xyz, and all other of our
    services that run under their own user ids, and use User= (but only
    in a world where userns is ubiquitous since otherwise we cannot
    invoke those daemons on the host AND in a container anymore). Also,
    if LimitNPROC= is used without User= we should warn and refuse
    operation.
  - journalctl --verify: don't show files that are currently being
    written to as FAIL, but instead show that their are being written to.
  - add journalctl -H that talks via ssh to a remote peer and passes through
    binary logs data
  - add a version of --merge which also merges /var/log/journal/remote
  - journalctl: -m should access container journals directly by enumerating
    them via machined, and also watch containers coming and going.
    Benefit: nspawn --ephemeral would start working nicely with the journal.
  - assign MESSAGE_ID to log messages about failed services
  - check if loop in decompress_blob_xz() is necessary

* journald: support RFC3164 fully for the incoming syslog transport, see
  https://github.com/systemd/systemd/issues/19251#issuecomment-816601955

* Hook up journald's FSS logic with TPM2: seal the verification disk by
  time-based policy, so that the verification key can remain on host and ve
  validated via TPM.

* rework journalctl -M to be based on a machined method that generates a mount
  fd of the relevant journal dirs in the container with uidmapping applied to
  allow the host to read it, while making everything read-only.

* journald: add varlink service that allows subscribing to certain log events,
  for example matching by message ID, or log level returns a list of journal
  cursors as they happen.

* journald: also collect CLOCK_BOOTTIME timestamps per log entry. Then, derive
  "corrected" CLOCK_REALTIME information on display from that and the timestamp
  info of the newest entry of the specific boot (as identified by the boot
  ID). This way, if a system comes up without a valid clock but acquires a
  better clock later, we can "fix" older entry timestamps on display, by
  calculating backwards. We cannot use CLOCK_MONOTONIC for this, since it does
  not account for suspend phases. This would then also enable us to correct the
  kmsg timestamping we consume (where we erroneously assume the clock was in
  CLOCK_MONOTONIC, but it actually is CLOCK_BOOTTIME as per kernel).

* in journald, write out a recognizable log record whenever the system clock is
  changed ("stepped"), and in timesyncd whenever we acquire an NTP fix
  ("slewing"). Then, in journalctl for each boot time we come across, find
  these records, and use the structured info they include to display
  "corrected" wallclock time, as calculated from the monotonic timestamp in the
  log record, adjusted by the delta declared in the structured log record.

* in journald: whenever we start a new journal file because the boot ID
  changed, let's generate a recognizable log record containing info about old
  and new ID. Then, when displaying log stream in journalctl look for these
  records, to be able to order them.

* journald: generate recognizable log events whenever we shutdown journald
  cleanly, and when we migrate run → var. This way tools can verify that a
  previous boot terminated cleanly, because either of these two messages must
  be safely written to disk, then.

* hook up journald with TPMs? measure new journal records to the TPM in regular
  intervals, validate the journal against current TPM state with that. (taking
  inspiration from IMA log)

* sd-journal puts a limit on parallel journal files to view at once. journald
  should probably honour that same limit (JOURNAL_FILES_MAX) when vacuuming to
  ensure we never generate more files than we can actually view.

* maybe add a tool that displays most recent journal logs as QR code to scan
  off screen and run it automatically on boot failures, emergency logs and
  such. Use DRM APIs directly, see
  https://github.com/dvdhrm/docs/blob/master/drm-howto/modeset.c for an example
  for doing that.

* maybe implicitly attach monotonic+realtime timestamps to outgoing messages in
  log.c and sd-journal-send

* journalctl/timesyncd: whenever timesyncd acquires a synchronization from NTP,
  create a structured log entry that contains boot ID, monotonic clock and
  realtime clock (I mean, this requires no special work, as these three fields
  are implicit). Then in journalctl when attempting to display the realtime
  timestamp of a log entry, first search for the closest later log entry
  of this kinda that has a matching boot id, and convert the monotonic clock
  timestamp of the entry to the realtime clock using this info. This way we can
  retroactively correct the wallclock timestamps, in particular for systems
  without RTC, i.e. where initially wallclock timestamps carry rubbish, until
  an NTP sync is acquired.

* introduce per-unit (i.e. per-slice, per-service) journal log size limits.

* journald: do journal file writing out-of-process, with one writer process per
  client UID, so that synthetic hash table collisions can slow down a specific
  user's journal stream down but not the others.

* tweak journald context caching. In addition to caching per-process attributes
  keyed by PID, cache per-cgroup attributes (i.e. the various xattrs we read)
  keyed by cgroup path, and guarded by ctime changes. This should provide us
  with a nice speed-up on services that have many processes running in the same
  cgroup.

* maybe add call sd_journal_set_block_timeout() or so to set SO_SNDTIMEO for
  the sd-journal logging socket, and, if the timeout is set to 0, sets
  O_NONBLOCK on it. That way people can control if and when to block for
  logging.

* journalctl: make sure -f ends when the container indicated by -M terminates

* journald: sigbus API via a signal-handler safe function that people may call
  from the SIGBUS handler

* add a test if all entries in the catalog are properly formatted.
    (Adding dashes in a catalog entry currently results in the catalog entry
     being silently skipped. journalctl --update-catalog must warn about this,
     and we should also have a unit test to check that all our message are OK.)

* build short web pages out of each catalog entry, build them along with man
  pages, and include hyperlinks to them in the journal output

* homed:
  - when user tries to log into record signed by unrecognized key, automatically add key to our chain after polkit auth
  - rollback when resize fails mid-operation
  - GNOME's side for forget key on suspend (requires rework so that lock screen runs outside of uid)
  - update LUKS password on login if we find there's a password that unlocks the JSON record but not the LUKS device.
  - create on activate?
  - properties: icon url?, preferred session type?, administrator bool (which translates to 'wheel' membership)?, address?, telephone?, vcard?, samba stuff?, parental controls?
  - communicate clearly when usb stick is safe to remove. probably involves
    beefing up logind to make pam session close hook synchronous and wait until
    systemd --user is shut down.
  - logind: maybe keep a "busy fd" as long as there's a non-released session around or the user@.service
  - maybe make automatic, read-only, time-based reflink-copies of LUKS disk
    images (and btrfs snapshots of subvolumes) (think: time machine)
  - distinguish destroy / remove (i.e. currently we can unregister a user, unregister+remove their home directory, but not just remove their home directory)
  - in systemd's PAMName= logic: query passwords with ssh-askpassword, so that we can make "loginctl set-linger" mode work
  - fingerprint authentication, pattern authentication, …
  - make sure "classic" user records can also be managed by homed
  - make size of $XDG_RUNTIME_DIR configurable in user record
  - query password from kernel keyring first
  - update even if record is "absent"
  - move acct mgmt stuff from pam_systemd_home to pam_systemd?
  - when "homectl --pkcs11-token-uri=" is used, synthesize ssh-authorized-keys records for all keys we have private keys on the stick for
  - make slice for users configurable (requires logind rework)
  - logind: populate auto-login list bus property from PKCS#11 token
  - when determining state of a LUKS home directory, check DM suspended sysfs file
  - when homed is in use, maybe start the user session manager in a mount namespace with MS_SLAVE,
    so that mounts propagate down but not up - eg, user A setting up a backup volume
    doesn't mean user B sees it
  - use credentials logic/TPM2 logic to store homed signing key
  - permit multiple user record signing keys to be used locally, and pick
    the right one for signing records automatically depending on a pre-existing
    signature
  - add a way to "adopt" a home directory, i.e. strip foreign signatures
    and insert a local signature instead.
  - as an extension to the directory+subvolume backend: if located on
    especially marked fs, then sync down password into LUKS header of that fs,
    and always verify passwords against it too. Bootstrapping is a problem
    though: if no one is logged in (or no other user even exists yet), how do you
    unlock the volume in order to create the first user and add the first pw.
  - support new FS_IOC_ADD_ENCRYPTION_KEY ioctl for setting up fscrypt
  - maybe pre-create ~/.cache as subvol so that it can have separate quota
    easily?
  - add a switch to homectl (maybe called --first-boot) where it will check if
    any non-system users exist, and if not prompts interactively for basic user
    info, mimicking systemd-firstboot. Then, place this in a service that runs
    after systemd-homed, but before gdm and friends, as a simple, barebones
    fallback logic to get a regular user created on uninitialized systems.
  - store PKCS#11 + FIDO2 token info in LUKS2 header, compatible with
    systemd-cryptsetup, so that it can unlock homed volumes
  - maybe make all *.home files owned by `systemd-home` user or so, so that we
    can easily set overall quota for all users
  - on login, if we can't fallocate initially, but rebalance is on, then allow
    login in discard mode, then immediately rebalance, then turn off discard
  - extend user records with optional "bulk" data. Specifically, a user
    avatar/photo or so. This data should be stored along with the user record,
    but probably shouldn't be part of the record itself, since it might be
    large.

* add a new switch --auto-definitions=yes/no or so to systemd-repart. If
  specified, synthesize a definition automatically if we can: enlarge last
  partition on disk, but only if it is marked for growing and not read-only.

* systemd-repart: read LUKS encryption key from $CREDENTIALS_DIRECTORY

* systemd-repart: add a switch to factory reset the partition table without
  immediately applying the new configuration again. i.e. --factory-reset=leave
  or so. (this is useful to factory reset an image, then putting it into
  another machine, ensuring that luks key is generated on new machine, not old)

* systemd-repart: support setting up dm-integrity with HMAC

* systemd-repart: maybe remove half-initialized image on failure. It fails
  if the output file exists, so a repeated invocation will usually fail if
  something goes wrong on the way.

* systemd-repart: drop pager mode on normal operation?

* systemd-repart: by default generate minimized partition tables (i.e. tables
  that only cover the space actually used, excluding any free space at the
  end), in order to maximize dd'ability. Requires libfdisk work, see
  https://github.com/karelzak/util-linux/issues/907

* systemd-repart: MBR partition table support. Care needs to be taken regarding
  Type=, so that partition definitions can sanely apply to both the GPT and the
  MBR case. Idea: accept syntax "Type=gpt:home mbr:0x83" for setting the types
  for the two partition types explicitly. And provide an internal mapping so
  that "Type=linux-generic" maps to the right types for both partition tables
  automatically.

* systemd-repart: allow sizing partitions as factor of available RAM, so that
  we can reasonably size swap partitions for hibernation.

* systemd-repart: allow boolean option that ensures that if existing partition
  doesn't exist within the configured size bounds the whole command fails. This
  is useful to implement ESP vs. XBOOTLDR schemes in installers: have one set
  of repart files for the case where ESP is large enough and one where it isn't
  and XBOOTLDR is added in instead.  Then apply the former first, and if it
  fails to apply use the latter.

* systemd-repart: add per-partition option to never reuse existing partition
  and always create anew even if matching partition already exists.

* systemd-repart: add per-partition option to fail if partition already exist,
  i.e. is not added new. Similar, add option to fail if partition does not exist yet.

* systemd-repart: allow disabling growing of specific partitions, or making
  them (think ESP: we don't ever want to grow it, since we cannot resize vfat)
  Also add option to disable operation via kernel command line.

* systemd-repart: make it a static checker during early boot for existence and
  absence of other partitions for trusted boot environments

* systemd-repart: add support for SD_GPT_FLAG_GROWFS also on real systems, i.e.
  generate some unit to actually enlarge the fs after growing the partition
  during boot.

* systemd-repart: do not print "Successfully resized …" when no change was done.

* document:
  - document that deps in [Unit] sections ignore Alias= fields in
    [Install] units of other units, unless those units are disabled
  - man: clarify that time-sync.target is not only sysv compat but also useful otherwise. Same for similar targets
  - document that service reload may be implemented as service reexec
  - add a man page containing packaging guidelines and recommending usage of things like Documentation=, PrivateTmp=, PrivateNetwork= and ReadOnlyDirectories=/etc /usr.
  - document systemd-journal-flush.service properly
  - documentation: recommend to connect the timer units of a service to the service via Also= in [Install]
  - man: document the very specific env the shutdown drop-in tools live in
  - man: add more examples to man pages,
  -      in particular an example how to do the equivalent of switching runlevels
  - man: maybe sort directives in man pages, and take sections from --help and apply them to man too
  - document root=gpt-auto properly

* systemctl:
  - add systemctl switch to dump transaction without executing it
  - Add a verbose mode to "systemctl start" and friends that explains what is being done or not done
  - "systemctl disable" on a static unit prints no message and does
    nothing. "systemctl enable" does nothing, and gives a bad message
    about it. Should fix both to print nice actionable messages.
  - print nice message from systemctl --failed if there are no entries shown, and hook that into ExecStartPre of rescue.service/emergency.service
  - add new command to systemctl: "systemctl system-reexec" which reexecs as many daemons as virtually possible
  - systemctl enable: fail if target to alias into does not exist? maybe show how many units are enabled afterwards?
  - systemctl: "Journal has been rotated since unit was started." message is misleading
  - systemctl status output should include list of triggering units and their status

* introduce an option (or replacement) for "systemctl show" that outputs all
  properties as JSON, similar to busctl's new JSON output. In contrast to that
  it should skip the variant type string though.

* Add a "systemctl list-units --by-slice" mode or so, which rearranges the
  output of "systemctl list-units" slightly by showing the tree structure of
  the slices, and the units attached to them.

* add "systemctl wait" or so, which does what "systemd-run --wait" does, but
  for all units. It should be both a way to pin units into memory as well as a
  wait to retrieve their exit data.

* show whether a service has out-of-date configuration in "systemctl status" by
  using mtime data of ConfigurationDirectory=.

* "systemctl preset-all" should probably order the unit files it
  operates on lexicographically before starting to work, in order to
  ensure deterministic behaviour if two unit files conflict (like DMs
  do, for example)

* add "systemctl start -v foobar.service" that shows logs of a service
  while the start command runs. This is non-trivial to do without
  races though, since we should flush out all journal messages before
  returning from the "systemctl stop".

* systemctl: if some operation fails, show log output?

* Add a new verb "systemctl top"

* unit install:
  - "systemctl mask" should find all names by which a unit is accessible
    (i.e. by scanning for symlinks to it) and link them all to /dev/null

* nspawn:
  - emulate /dev/kmsg using CUSE and turn off the syslog syscall
    with seccomp. That should provide us with a useful log buffer that
    systemd can log to during early boot, and disconnect container logs
    from the kernel's logs.
  - as soon as networkd has a bus interface, hook up --network-interface=,
    --network-bridge= with networkd, to trigger netdev creation should an
    interface be missing
  - a nice way to boot up without machine id set, so that it is set at boot
    automatically for supporting --ephemeral. Maybe hash the host machine id
    together with the machine name to generate the machine id for the container
  - fix logic always print a final newline on output.
    https://github.com/systemd/systemd/pull/272#issuecomment-113153176
  - should optionally support receiving WATCHDOG=1 messages from its payload
    PID 1...
  - optionally automatically add FORWARD rules to iptables whenever nspawn is
    running, remove them when shut down.
  - add support for sysext extensions, too. i.e. a new --extension= switch that
    takes one or more arguments, and applies the extensions already during
    startup.
  - when main nspawn supervisor process gets suspended due to SIGSTOP/SIGTTOU
    or so, freeze the payload too.
  - support time namespaces
  - on cgroupsv1 issue cgroup empty handler process based on host events, so
    that we make cgroup agent logic safe
  - add API to invoke binary in container, then use that as fallback in
    "machinectl shell"
  - make nspawn suitable for shell pipelines: instead of triggering a hangup
    when input is finished, send ^D, which synthesizes an EOF. Then wait for
    hangup or ^D before passing on the EOF.
  - greater control over selinux label?
  - support that /proc, /sys/, /dev are pre-mounted
  - maybe allow TPM passthrough, backed by swtpm, and measure --image= hash
    into its PCR 11, so that nspawn instances can be TPM enabled, and partake
    in measurements/remote attestation and such. swtpm would run outside of
    control of container, and ideally would itself bind its encryption keys to
    host TPM.
  - make boot assessment do something sensible in a container. i.e send an
    sd_notify() from payload to container manager once boot-up is completed
    successfully, and use that in nspawn for dealing with boot counting,
    implemented in the partition table labels and directory names.
  - optionally set up nftables/iptables routes that forward UDP/TCP traffic on
    port 53 to resolved stub 127.0.0.54
  - maybe optionally insert .nspawn file as GPT partition into images, so that
    such container images are entirely stand-alone and can be updated as one.
  - The subreaper logic we currently have seems overly complex. We should
    investigate whether creating the inner child with CLONE_PARENT isn't better.
  - Reduce the number of sockets that are currently in use and just rely on one
    or two sockets.
  - Support running nspawn as an unprivileged user.

* machined: add API to acquire UID range. add API to mount/dissect loopback
  file. Both protected by PK. Then make nspawn use these APIs to run
  unprivileged containers. i.e. push the truly privileged bits into machined,
  so that the client side can remain entirely unprivileged, with SUID or
  anything like that.

* machined:
  - add an API so that libvirt-lxc can inform us about network interfaces being
    removed or added to an existing machine
  - "machinectl migrate" or similar to copy a container from or to a
    difference host, via ssh
  - introduce systemd-nspawn-ephemeral@.service, and hook it into
    "machinectl start" with a new --ephemeral switch
  - "machinectl status" should also show internal logs of the container in
    question
  - "machinectl history"
  - "machinectl diff"
  - "machinectl commit" that takes a writable snapshot of a tree, invokes a
    shell in it, and marks it read-only after use

* udev:
  - move to LGPL
  - kill scsi_id
  - add trigger --subsystem-match=usb/usb_device device
  - reimport udev db after MOVE events for devices without dev_t
  - re-enable ProtectClock= once only cgroupsv2 is supported.
    See f562abe2963bad241d34e0b308e48cf114672c84.

* coredump:
  - save coredump in Windows/Mozilla minidump format
  - when truncating coredumps, also log the full size that the process had, and make a metadata field so we can report truncated coredumps
  - add examples for other distros in ELF_PACKAGE_METADATA

* support crash reporting operation modes (https://live.gnome.org/GnomeOS/Design/Whiteboards/ProblemReporting)

* tmpfiles:
  - apply "x" on "D" too (see patch from William Douglas)
  - allow time-based cleanup in r and R too
  - instead of ignoring unknown fields, reject them.
  - creating new directories/subvolumes/fifos/device nodes
    should not follow symlinks. None of the other adjustment or creation
    calls follow symlinks.
  - add --test mode
  - teach tmpfiles.d q/Q logic something sensible in the context of XFS/ext4
    project quota
  - teach tmpfiles.d m/M to move / atomic move + symlink old -> new
  - add new line type for setting btrfs subvolume attributes (i.e. rw/ro)
  - tmpfiles: add new line type for setting fcaps

* udev-link-config:
   - Make sure ID_PATH is always exported and complete for
     network devices where possible, so we can safely rely
     on Path= matching

* sd-rtnl:
   - add support for more attribute types
   - inbuilt piping support (essentially degenerate async)? see loopback-setup.c and other places

* networkd:
   - add more keys to [Route] and [Address] sections
   - add support for more DHCPv4 options (and, longer term, other kinds of dynamic config)
   - add reduced [Link] support to .network files
   - properly handle routerless dhcp leases
   - work with non-Ethernet devices
   - dhcp: do we allow configuring dhcp routes on interfaces that are not the one we got the dhcp info from?
   - the DHCP lease data (such as NTP/DNS) is still made available when
     a carrier is lost on a link. It should be removed instantly.
   - expose in the API the following bits:
         - option 15, domain name
         - option 12, hostname and/or option 81, fqdn
         - option 123, 144, geolocation
         - option 252, configure http proxy (PAC/wpad)
   - provide a way to define a per-network interface default metric value
     for all routes to it. possibly a second default for DHCP routes.
   - allow Name= to be specified repeatedly in the [Match] section. Maybe also
     support Name=foo*|bar*|baz ?
   - whenever uplink info changes, make DHCP server send out FORCERENEW

* in networkd, when matching device types, fix up DEVTYPE rubbish the kernel passes to us

* Figure out how to do unittests of networkd's state serialization

* dhcp:
   - figure out how much we can increase Maximum Message Size

* dhcp6:
   - add functions to set previously stored IPv6 addresses on startup and get
     them at shutdown; store them in client->ia_na
   - write more test cases
   - implement reconfigure support, see 5.3., 15.11. and 22.20.
   - implement support for temporary adressess (IA_TA)
   - implement dhcpv6 authentication
   - investigate the usefulness of Confirm messages; i.e. are there any
     situations where the link changes without any loss in carrier detection
     or interface down
   - some servers don't do rapid commit without a filled in IA_NA, verify
     this behavior
   - RouteTable= ?

* shared/wall: Once more programs are taught to prefer sd-login over utmp,
  switch the default wall implementation to wall_logind
  (https://github.com/systemd/systemd/pull/29051#issuecomment-1704917074)