1
0
mirror of https://github.com/systemd/systemd.git synced 2024-12-22 17:35:35 +03:00
systemd/TODO
Lennart Poettering ecedd90fcd service: place control command in subcgroup control/
Previously, we were brutally and onconditionally killing all processes
in a service's cgroup before starting the service anew, in order to
ensure that StartPre lines cannot be misused to spawn long-running
processes.

On logind-less systems this has the effect that restarting sshd
necessarily calls all active ssh sessions, which is usually not
desirable.

With this patch control processes for a service are placed in a
sub-cgroup called "control/". When starting a service anew we simply
kill this cgroup, but not the main cgroup, in order to avoid killing any
long-running non-control processes from previous runs.

https://bugzilla.redhat.com/show_bug.cgi?id=805942
2012-04-13 23:29:59 +02:00

348 lines
12 KiB
Plaintext

Bugfixes:
* fix Fedora's cups.rpm to use tmpfiles instead of /lib/udev/devices/
* swap units that are activated by one name but shown in the kernel under another are semi-broken
* make anaconda write timeout=0 for encrypted devices
* make sure timeouts are applied to Type=oneshot services.
* Dangling symlinks of .automount unit files in .wants/ directories, set up
automount points even when the original .automount file did not exist
anymore. Only the .mount unit was still around.
* make polkit checks async
* properly handle .mount unit state tracking when two mount points are stacked one on top of another on the exact same mount point.
Features:
* cg_create_and_attach() should fail for non-available controllers
* make gtk-doc optional (like kmod?)
* udev: find a way to tell udev to not cancel firmware requests in initramfs
* udev: scsi_id -> sg3_utils?
* udev: add trigger --subsystem-match=usb/usb_device
* udev: libudev.so.1
- remove deprecated symbols
- use symbol versioning
- return object with *_unref()
- kill: udev_monitor_from_socket()
- kill: udev_queue_get_failed_list_entry()
* allow configuration of console width/height in vconsole.conf
* PrivateTmp should apply to both /tmp and /var/tmp
* fstab should take priority over units in /usr
* cleanup syslog 'priority' vs. 'level' wording
* journal: if mmap() fails for mapping window try to unmap a a few older maps
* dbus upstream still refers to dbus.target and shouldn'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
* add man page documenting all kernel cmdline options, including stuff like fsck.mode=
* show getty in container mode, not sulogin
* support container_ttys=
* journald: make configurable "store-on-var", "store-on-run", "dont-store", "auto"
(store-persistent, store-volatile?)
* Add ConditionReadWriteFileSystem= so that systemd-sysctl doesn't get executed when /proc/sys is read-only
* introduce mix of BindTo and Requisite
* journalctl: show multiline log messages sanely, expand tabs, and show all valid utf8 messages
* introduce NeedsMounts= or so to create .mount dependencies automatically for a specific path
* add DeleteSocketsOnStop=yes|no option to socket units
* add shutdown inhibit API for usage by libvirt and friends
* journal: store euid in journal if it differs from uid
* support chrony in addition to ntpd in timedated
* document crypttab(5)
* There's currently no way to cancel fsck (used to be possible via C-c or c on the console)
* when dumping cgroup contents, include main/control PID of a service, explicitly
* keep an eye on https://bugzilla.gnome.org/show_bug.cgi?id=670100
* D-Bus: always pass cred data along each message
* journal: allow turning off logging entirely
* journal: sanely deal with entries which are larger than the individual file size, but where the componets would fit
* add command to systemctl to plot dependency graph as tree (see rhbz 795365)
* make logind reserve tty10 or so for text logins, so that gdm never picks it up
* add option to sockets to avoid activation. Instead just drop packets/connections, see http://cyberelk.net/tim/2012/02/15/portreserve-systemd-solution/
* default unix qlen is too small (10). bump sysctl? add sockopt?
* support units generated by a generator and placed in /run/systemd/system/; the directory is
currently ignored because it is empty before the generatores are executed
* Possibly, detect whether SysV init scripts can do reloading by looking for "echo Usage:" lines
* figure out whether we should leave dbus around during shutdown
* add interface to allow immediate rotation of the journal, and even flushing.
* dbus: in fedora, make the machine a symlink to /etc/machine-id
* journald: reuse XZ context
* logind: add equivalent to sd_pid_get_owner_uid() to the D-Bus API
* write RPM spec macros for presets
* journal: write man pages for API
* journal: OR matches are borked
* journal: extend hash tables as we go
* journal: API for looking for retrieving "all values of this field"
* journal: deal nicely with byte-by-byte copied files, especially regards header
* journal: local deserializer of export mode, http server
* journal: message catalog
* journal: forward-secure signatures
* document the exit codes when services fail before they are exec()ed
* rework namespace support, don't use pivot_root, and mount things after creating the namespace, not before
* systemctl journal command
* journalctl: --cursor support, priority filtering
* systemctl status: show coredumps
* systemctl status: show whether journal was rotated since service started
* save coredump in Windows/Mozilla minidump format
* support crash reporting operation modes (https://live.gnome.org/GnomeOS/Design/Whiteboards/ProblemReporting)
* journal: allow per-entry control on /var vs. /run (think incognito browser mode)
* clean up session cgroups that remain after logout (think sshd), but eventually run empty
* support "systemctl stop foobar@.service" to stop all units matching a certain template
* move to LGPL2+
* logind: allow showing logout dialog from system
* document that %% can be used to write % in a string that is specifier extended
* when an instanced service exits, remove its parent cgroup too if possible.
* as Tom Gundersen pointed out there's a always a dep loop if people use crypto file systems with random keys
* automatically escape unit names passed on the service (i.e. think "systemctl start serial-getty.service@serial/by-path/jshdfjsdfhkjh" being automatically escaped as necessary.
* if we can not get user quota for tmpfs, mount a separate tmpfs instance
for every user in /run/user/$USER with a configured maximum size
* default to actual 32bit PIDs, via /proc/sys/kernel/pid_max
* add an option to make mounts private/shareable and so on, enable this for root by default
* be able to specify a forced restart of service A where service B depends on, in case B
needs to be auto-respawned?
* Something is wrong with symlink handling of "autovt@.service" in "systemctl list-unit-files"
* when a bus name of a service disappears from the bus make sure to queue further activation requests
* something like ConditionExec= or ExecStartPre= without failure state
* tmpfiles: apply "x" on "D" too (see patch from William Douglas)
* don't set $HOME in services unless requested
* hide PAM/TCPWrap options in fragment parser when compile time disabled
* when we automatically restart a service, ensure we restart its rdeps, too.
* allow Type=simple with PIDFile=
https://bugzilla.redhat.com/show_bug.cgi?id=723942
* move PAM code into its own binary
* warn if the user stops a service but not its associated socket
* logind: spawn user@..service on login
* logind: non-local X11 server handling
* implement Register= switch in .socket units to enable registration
in Avahi, RPC and other socket registration services.
* make sure systemd-ask-password-wall does not shutdown systemd-ask-password-console too early
* readahead: use BTRFS_IOC_DEFRAG_RANGE instead of BTRFS_IOC_DEFRAG ioctl, with START_IO
* readahead: check whether a btrfs volume includes ssd by checking mount flag "ssd"
* support sd_notify() style notification when reload begins (RELOADING=1), reload is finished (READY=1), and add ReloadSignal= then to use in combination
* support sd_notify() style notification when shutting down, to make auto-exit bus services work (STOPPING=1)
* 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.
* move nss-myhostname into systemd
* and a dbus call to generate target from current state
* drop /.readahead on bigger upgrades with yum
* add inode nr check to readahead to suppress preloading changed files
* add support for /bin/mount -s
* GC unreferenced jobs (such as .device jobs)
* when failing to start a service due to ratelimiting, try again later, if restart=always is set
* write blog stories about:
- enabling dbus services
- status update
- 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
- how to integrate cgconfig and suchlike with systemd
- resource control in systemd
* allow port=0 in .socket units
* move readahead files into /var, look for them with .path units
* teach dbus to activate all services it finds in /etc/systemd/services/org-*.service
* support systemd.mask= on the kernel command line.
* when key file cannot be found, read it from kbd in cryptsetup
* reuse mkdtemp namespace dirs in /tmp?
* recreate systemd's D-Bus private socket file on SIGUSR2
* Support --test based on current system state
* investigate whether the gnome pty helper should be moved into systemd, to provide cgroup support.
* maybe introduce ExecRestartPre=
* configurable jitter for timer events
* timer events with system resume
* timer events on calendar time
* dot output for --test showing the 'initial transaction'
* calendar time support in timer, iCalendar semantics for the timer stuff (RFC2445)
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=99ee5315dac6211e972fa3f23bcc9a0343ff58c4
* implicitly import "defaults" settings file into all types
* exec settings override
* writable cgroups dbus properties for live changes
* read config fragments for all units from /lib/systemd/system/foobar.service.d/ to override/extend specific settings
* port over to LISTEN_FDS/LISTEN_PID:
- rpcbind (/var/run/rpcbind.sock!) HAVEPATCH
- cups HAVEPATCH
- postfix, saslauthd
- apache/samba
- libvirtd (/var/run/libvirt/libvirt-sock-ro)
- bluetoothd (/var/run/sdp! @/org/bluez/audio!)
- distccd
* auditd service files
* fingerprint.target, wireless.target, gps.target, netdevice.target
* io priority during initialization
* systemctl list-jobs - show dependencies
* add systemctl switch to dump transaction without executing it
* suspend, resume support?
* drop cap bounding set in readahead and other services
External:
* dbus:
- dbus --user
- natively watch for dbus-*.service symlinks (PENDING)
- allow specification of socket mode/umask when allocating DBusServer
- allow disabling of fd passing when connecting a AF_UNIX connection
- allow disabling of UID passing for AUTH EXTERNAL
* systemd --user
PR_SET_CHILD_REAPER patch: https://lkml.org/lkml/2011/7/28/426
(patch in linux-next, on the way to the next kernel)
* fix alsa mixer restore to not print error when no config is stored
* gnome-shell python script/glxinfo/is-accelerated must die
* make cryptsetup lower --iter-time
* patch kernel for xattr support in /dev, /proc/, /sys and /sys/fs/cgroup?
* NTP: the kernel's 11-minutes-mode syncs the system time to the RTC, but only
in an ~30 minutes window. It does not adjust larger differences. Find a way
to tell the kernel, to always do a full time sync when the RTC is in UTC and
we are in 11-minutes-mode. When we trust the system time to NTP we also want
the RTC to sync up.
* patch kernel for cpu feature modalias for autoloading aes/kvm/...
(patches in linux-next, on the way to the next kernel)
* kernel: add device_type = "fb", "fbcon" to class "graphics"
Regularly:
* look for close() vs. close_nointr() vs. close_nointr_nofail()
* check for strerror(r) instead of strerror(-r)
* Use PR_SET_PROCTITLE_AREA if it becomes available in the kernel
* %m in printf() instead of strerror();
* pahole
* set_put(), hashmap_put() return values check. i.e. == 0 doesn't free()!