* properly handle .mount unit state tracking when two mount points are stacked one on top of another on the exact same mount point.
-* When we detect invalid UTF-8, we cannot use it in an error message:
- log...("Path is not UTF-8 clean, ignoring assignment: %s", rvalue);
-
* shorten the message to sane length:
Cannot add dependency job for unit display-manager.service, ignoring: Unit display-manager.service failed to load: No such file or directory. See system logs and 'systemctl status display-manager.service' for details.
-External:
+* systemctl --root=container/ set-default ... is totally borked.
-* ps should gain colums for slice
+External:
* Fedora: when installing fedora with yum --installroot /var/run is a directory, not a symlink
https://bugzilla.redhat.com/show_bug.cgi?id=975864
Features:
-* libsystemd-journal returns the object created as first param in sd_journal_new(), sd_bus_new() and suchlike as last...
+* man: move .link, .network and .netdev documentation into their own
+ man pages maybe called "systemd.link(5)", "systemd.network(5)" and
+ "systemd.netdev(5)" or so.
+
+* extend device cgroup controller support to allow enabling groups of
+ device nodes, so that pts can be allowed in whole. For that,
+ introduce the syntax "block-XYZ" and "char-XYZ" in addition to the
+ existing "/dev/foobar", where XYZ then is looked up in /proc/devices
+ and mapped to one or more majors.
+
+* Add all pts device nodes to those allowed by default when the device
+ controller is used.
+
+* "busctl status" works only as root on dbus1, since we cannot read
+ /proc/$PID/exe
+
+* systemctl (and possibly related tools): support a new switch that
+ allows enumerating units in local containers recursively. "systemctl
+ list-units -R" or so should not only lists on the host, but also the
+ services in all containers in a pretty way, to give an overview of
+ the entire system. Also, maybe add "systemctl list-machines" which
+ works like "machinectl list" but includes information about the
+ health status of each registered machine. For that we should
+ probably implement something that encodes the system health status
+ in a single enum state, i.e. something like a system-wide state
+ starting → running → failed → stopping, that is based on the current
+ job queue and a check for failed services. Maybe then change
+ "systemctl status" without args to output this state along with a
+ selection of other data, such as the uptime or so.
+
+* Add a seccomp-based filter for socket() calls to limit services to
+ specific address families (for example: AF_UNIX), inspired by
+ Android's sandboxing
+
+* implement Distribute= in socket units to allow running multiple
+ service instances processing the listening socket, and open this up
+ for ReusePort=
+
+* add a timelimit to generator invocation
+
+* socket units: support creating sockets in different namespace,
+ opening it up for JoinsNamespaceOf=. This would require to fork off
+ a tiny process that joins the namespace and creates/binds the socket
+ and passes this back to PID1 via SCM_RIGHTS. This also could be used
+ to allow Chown/chgrp on sockets without requiring NSS in PID 1.
+
+* New service property: maximum CPU and wallclock runtime for a service
+
+* introduce bus call FreezeUnit(s, b), as well as "systemctl freeze
+ $UNIT" and "systemctl thaw $UNIT" as wrappers around this. The calls
+ should SIGSTOP all unit processes in a loop until all processes of
+ it are fully stopped. This can later be used for app management by
+ desktop UIs such as gnome-shell to freeze apps that are not visible
+ on screen, not unlike how job control works on the shell
+
+* completions:
+ - busctl zsh completion is outdated
+ - systemd-nspawn -Z/-L/-q is missing for zsh
+ - systemd-nspawn completion is missing for bash
+ - manager property enumeration was broken when systemd moved to /usr/lib/
* cgroups:
- implement system-wide DefaultCPUAccounting=1 switch (and similar for blockio, memory?)
- add field to transient units that indicate whether systemd or somebody else saves/restores its settings, for integration with libvirt
- ensure scope units may be started only a single time
-* code cleanup
- - get rid of readdir_r/dirent_storage stuff, it's unnecessary on Linux
- - we probably should replace the left-over uses of strv_append() and replace them by strv_push() or strv_extend()
-
-* switch to SipHash for hashmaps/sets?
+* Automatically configure swap partition to use for hibernation by looking for largest swap partition on the root disk?
* when we detect low battery and no AC on boot, show pretty splash and refuse boot
-* move libasyncns into systemd as libsystemd-asyncns
-
* machined, localed: when we try to kill an empty cgroup, generate an ESRCH error over the bus
* libsystemd-journal, libsystemd-login, libudev: add calls to easily attach these objects to sd-event event loops
* Automatically configure swap partition to use for hibernation by looking for largest swap partition on the root disk?
-* remove NSS usage from PID 1 (notably the specifiers)
-
-* socket-proxyd:Use a nonblocking alternative to getaddrinfo
+* socket-proxyd: Use sd-resolve to resolve the server address
* 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
-* Add a new Distribute=$NUMBER key to socket units that makes use of SO_REUSEPORT to distribute network traffic on $NUMBER instances
-
* move config_parse_path_strv() out of conf-parser.c
* After coming back from hibernation reset hibernation swap partition using the /dev/snapshot ioctl APIs
error. Currently, we just ignore it and read the unit from the search
path anyway.
-* When a Type=forking service fails and needed another service, that
- service is not cleaned up again when it has StopWhenUnneeded=yes
- http://lists.freedesktop.org/archives/systemd-devel/2013-July/012141.html
-
* refuse boot if /etc/os-release is missing or /etc/machine-id cannot be set up
-* given that logind/machined now let PID 1 do all nasty work, we can
- probably reduce the capability set they retain substantially.
+* given that logind now lets PID 1 do all nasty work, we can
+ probably reduce the capability set it retains substantially.
+ (we need CAP_SYS_ADMIN for drmSetMaster(), so maybe not worth it)
* btrfs raid assembly: some .device jobs stay stuck in the queue
Maybe take a BSD lock at the disk device node and teach udev to
check for that and suppress event handling.
-* document logic of auto/noauto and fail/nofail in fstab in systemd.mount or systemd-fstab-generator man page
-
* something pulls in pcre as shared object dependency into our daemons such as hostnamed.
* allow implementation of InaccessibleDirectories=/ plus
ReadOnlyDirectories=... for whitelisting files for a service.
-* libsystemd-bus:
+* sd-bus:
- when kdbus doesn't take our message without memfds, try again with memfds
- - implement translator service
- implement monitor logic
- - properly map matches with well-known names against messages with unique names
- - when triggering property change events, allow a NULL strv indicate that all properties listed as such are send out as changed
- see if we can drop more message validation on the sending side
- - support "const" properties as flag
- add API to clone sd_bus_message objects
- - SD_BUS_COMMENT() macro for inclusion in vtables, syntax inspired by gdbus
- - sort out error codes for sd_bus_release_name(), distuingish: successful removal from foreign name, from non-existing name
+ - systemd-bus-proxyd needs to enforce good old XML policy
+ - upload minimal kdbus policy into the kernel at boot
- kdbus: matches against source or destination pids for an "strace -p"-like feel. Problem: The PID info needs to be available in userspace too...
- - kdbus: we need a way to distuingish messages we got due to monitoring from normal messages, since we want to bind methods only to the latter
- - kdbus: when we do "systemctl daemon-reexec" the call doesn't get properly cancelled
- - figure out what to do when fields in the kdbus header and in the payload header do not match
- - longer term:
- * priority queues
- * priority inheritance
+ - longer term: priority inheritance
+ - check sender of response messages
+ - dbus spec updates:
+ - kdbus mapping
+ - NameLost/NameAcquired obsolete
+ - GVariant
+ - "const" properties (posted)
+ - port exit-on-idle logic to byebye ioctl
+ - update systemd.special(7) to mention that dbus.socket is only about the compatibility socket now
+ - allow updating attach flags during runtime
+ - pid1: peek into activating message when activating a service
+ - test bloom filter generation indexes
+ - introduce sd_bus_emit_object_added()/sd_bus_emit_object_removed() that automatically includes the build-in interfaces in the list
+ - port to sd-resolve for connecting to TCP dbus servers
+ - constructors for bus messages should probably not be OK with a NULL bus pointer
* sd-event
- - allow multiple signal handlers per signal
- - when a handler returns an error, just turn off its event source,
- but do not return anything up to the event loop caller. Instead
- add parameter to sd_event_request_quit() to take retval. This way
- errors rippling upwards are the option, not the default
- - when dispatching an event source then _unref() on it should remove it from the epoll
+ - allow multiple signal handlers per signal?
* in the final killing spree, detect processes from the root directory, and
complain loudly if they have argv[0][0] == '@' set.
https://bugzilla.redhat.com/show_bug.cgi?id=961044
-* Introduce a way how we can kill the main process of a service with KillSignal, but all processes with SIGKILL later on
- https://bugzilla.redhat.com/show_bug.cgi?id=952634
-
-* maybe add a warning to the unit file parses where the access mode of unit files is nonsensical.
-
* investigate endianness issues of UUID vs. GUID
* dbus: when a unit failed to load (i.e. is in UNIT_ERROR state), we
* timedatctl, localectl: possibly make some commands work without the daemon, for chroot situations...
+* timedatectl: print a nicer message when enabling ntp fails because ntpd/chrony are not installed
+
* cgtop: make cgtop useful in a container
* test/:
* when isolating, try to figure out a way how we implicitly can order
all units we stop before the isolating unit...
-* add ConditionArchitecture= or so
-
* teach ConditionKernelCommandLine= globs or regexes (in order to match foobar={no,0,off})
-* Support SO_REUSEPORT with socket activation:
- - Let systemd maintain a pool of servers.
- - Use for seamless upgrades, by running the new server before stopping the
- old.
-
* after all byte-wise realloc() is slow, even on glibc, so i guess we
need manual exponential loops after all
- logind: wakelock/opportunistic suspend support
- Add pretty name for seats in logind
- logind: allow showing logout dialog from system?
- - logind: non-local X11 server handling
- logind: add equivalent to sd_pid_get_owner_uid() to the D-Bus API
- - pam: when leaving a session explicitly exclude the ReleaseSession() caller process from the killing spree
- we should probably handle SIGTERM/SIGINT to not leave dot files around, just in case
- when logging out, remove user-owned sysv and posix IPC objects
- session scopes/user unit: add RequiresMountsFor for the home directory of the user
- 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?
- - journalctl: support -M to read journal of containers and determine journal directory from root directory of container
- tmpfiles: when applying ownership to /run/log/journal, also do this for the journal fails contained in it
- 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
hence doing this via signals is not going to work.
* document:
- - document that deps in [Unit] sections ignore Alias= fileds in
+ - 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 word splitting syntax for ExecStart= and friends
- - document that units from /etc override those from /usr and /run
- document the exit codes when services fail before they are exec()ed
- document that service reload may be implemented as service reexec
- document in wiki how to map ical recurrence events to systemd timer unit calendar specifications
- 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
- - man: remove .include documentation, and instead push people to use .d/*.conf
- 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: extend runlevel(8) to mention that runlevels suck, and are dead. Maybe add runlevel(7) with a note about that too
- rework wait filter to not require match callback
- better error message if you run systemctl without systemd running
- systemctl status output should should include list of triggering units and their status
- - in systemctl list-timers show time trggering units ran last
+ - in systemctl list-timers show time triggering units ran last
* unit install:
- "systemctl mask" should find all names by which a unit is accessible
* currently x-systemd.timeout is lost in the initrd, since crypttab is copied into dracut, but fstab isn't
* nspawn:
- - nspawn: consider changing users for -u with su, so that NSS resolving works correctly
+ - nspawn: consider changing users for -u with /usr/bin/getent, so that NSS resolving works correctly
- nspawn: implement personality changes a la linux32(8)
- nspawn: --read-only is not applied recursively to submounts
- nspawn: make use of device cgroup controller by default
- bind mount read-only the cgroup tree higher than nspawn
- - nspawn: investigate whether we can support the same as LXC's
- lxc.network.type=phys mode, and pass through entire network
- interfaces to the container
- - nspawn: maybe add a way to drop additional caps, in addition to add additional caps
- - nspawn: maybe explicitly reset loginuid?
- nspawn: make it work for dwalsh and shared /usr containers -- tmpfs mounts as command line parameters, selinux exec context
- refuses to boot containers without /etc/machine-id (OK?), and with empty /etc/machine-id (not OK).
- - add an option to nspawn that uses seccomp to make socket(AF_NETLINK,
- SOCK_RAW, NETLINK_AUDIT) fail the the appropriate error code that
- makes the audit userspace to think auditing is not available in the
- kernel.
+ - support taking a btrfs snapshot at startup and dropping it afterwards
+ - maybe: hookup nspawn and PrivateNetwork=yes with "ip netns"
+ - allow booting disk images with a GPT signature using the bootloaderspec partition uuids
* cryptsetup:
- cryptsetup-generator: allow specification of passwords in crypttab itself
* explore multiple service instances per listening socket idea
-* shutdown: don't read-only mount anything when running in container
-
* MountFlags=shared acts as MountFlags=slave right now.
* ReadOnlyDirectories= is not applied recursively to submounts
* rename "userspace" to "core-os"
-* syscall filter:
- - syscall filter: add knowledge about compat syscalls
- - syscall filter: don't enforce no new privs?
- - syscall filter: option to return EPERM rather than SIGSYS?
- - syscall filter: port to libseccomp
- - system-wide seccomp filter
+* syscall filter: optionally don't enforce no new privs?
* load-fragment: when loading a unit file via a chain of symlinks
verify that it isn't masked via any of the names traversed.
- move to LGPL
- kill scsi_id
- add trigger --subsystem-match=usb/usb_device device
-
-* cleanup syslog 'priority' vs. 'level' wording
+ - reimport udev db after MOVE events for devices without dev_t
* when a service has the same env var set twice we actually store it twice and return that in systemctl show -p... We should only show the last setting
-* support container_ttys=
-
* introduce mix of BindTo and Requisite
* add DeleteSocketsOnStop=yes|no option to socket units
* dot output for --test showing the 'initial transaction'
-* port over to LISTEN_FDS/LISTEN_PID:
- - postfix, saslauthd
- - apache/samba
- - libvirtd (/var/run/libvirt/libvirt-sock-ro)
- - bluetoothd (/var/run/sdp! @/org/bluez/audio!)
- - distccd
-
* fingerprint.target, wireless.target, gps.target, netdevice.target
* io priority during initialization
- Make sure ID_PATH is always exported and complete for
network devices where possible, so we can safely rely
on Path= matching
- - Check if Driver= is broken, or just my driver (bcma)
* sd-rtnl:
- - replace sd_rtnl_message_append() by type-safe versions
- - improve container support, and add support for entering containers when reading
+ - add support for exiting containers without reading them fully first
- add support for more attribute types
* networkd:
- - unify and clarify logging, also use proper structured log messages
- - add [Route] sections
- - add more keys to [Address] sections
- - add DHCPv4 support (and, longer term, other kinds of dynamic config)
+ - add more keys to [Route] and [Address] sections
+ - add support for more DHCPv4 options (and, longer term, other kinds of dynamic config)
- add proper initrd support (in particular generate .network/.link files based on /proc/cmdline)
+ - add reduced [Link] support to .network files
External: