X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=README;h=0cd868ade28b70e40163bbb3ddfca44f0e536b8c;hp=350a22b45f66e17cbb59d35b34ece0baaf610111;hb=cac2e345596b2743053c0285280b81794b3aaf10;hpb=fd752623faf6a3a4f2ac17e1a9d24f589119084d diff --git a/README b/README index 350a22b45..0cd868ade 100644 --- a/README +++ b/README @@ -1,96 +1,183 @@ -udev - Linux userspace device management - -Integrating udev in the system has complex dependencies and may differ from -distribution to distribution. A system may not be able to boot up or work -reliably without a properly installed udev version. The upstream udev project -does not recommend replacing a distro's udev installation with the upstream -version. - -The upstream udev project's set of default rules may require a most recent -kernel release to work properly. This is currently version 2.6.31. - -Tools and rules shipped by udev are not public API and may change at any time. -Never call any private tool in /lib/udev from any external application; it might -just go away in the next release. Access to udev information is only offered -by udevadm and libudev. Tools and rules in /lib/udev and the entire contents of -the /dev/.udev directory are private to udev and do change whenever needed. - -Requirements: - - Version 2.6.27 of the Linux kernel with sysfs, procfs, signalfd, inotify, - unix domain sockets, networking and hotplug enabled: - CONFIG_HOTPLUG=y - CONFIG_UEVENT_HELPER_PATH="" - CONFIG_NET=y - CONFIG_UNIX=y - CONFIG_SYSFS=y - CONFIG_SYSFS_DEPRECATED*=n - CONFIG_PROC_FS=y - CONFIG_TMPFS=y - CONFIG_INOTIFY_USER=y - CONFIG_SIGNALFD=y - CONFIG_TMPFS_POSIX_ACL=y (user ACLs for device nodes) - CONFIG_BLK_DEV_BSG=y (SCSI devices) - - - Udev does not work with the CONFIG_SYSFS_DEPRECATED* option. - - - Unix domain sockets (CONFIG_UNIX) as a loadable kernel module may work, - but it is not supported. - - - The deprecated hotplug helper /sbin/hotplug should be disabled in the - kernel configuration, it is not needed today, and may render the system - unusable because the kernel may create too many processes in parallel - so that the system runs out-of-memory. - - - The proc filesystem must be mounted on /proc, and the sysfs filesystem must - be mounted at /sys. No other locations are supported by a standard - udev installation. - - - The system must have the following group names resolvable at udev startup: - disk, cdrom, floppy, tape, audio, video, lp, tty, dialout, and kmem. - Especially in LDAP setups, it is required that getgrnam() be able to resolve - these group names with only the rootfs mounted and while no network is - available. - - - The 'udev extras' has the following dependencies: - libacl, libglib2, libusb, usbutils, pciutils, and gperf. - These dependencies can be disabled with the --disable-extras configure option. - -Setup: - - At bootup, the /dev directory should get the 'devtmpfs' filesystem - mounted. Udev manages the permissions and ownership of the kernel-created - device nodes, and udev possibly creates additional symlinks. If needed, udev also - works on an empty 'tmpfs' filesystem, but some static device nodes like - /dev/null, /dev/console, /dev/kmsg are needed to be able to start udev itself. - - - The udev daemon should be started to handle device events sent by the kernel. - During bootup, the kernel can be asked to send events for all already existing - devices so that they too can be configured by udev. This is usually done by: - /sbin/udevadm trigger --type=subsystems - /sbin/udevadm trigger --type=devices - - - Restarting the daemon never applies any rules to existing devices. - - - New/changed rule files are picked up automatically; there is no daemon - restart or signal needed. - -Operation: - - Based on events the kernel sends out on device creation/removal, udev - creates/removes device nodes in the /dev directory. - - - All kernel events are matched against a set of specified rules, which - possibly hook into the event processing and load required kernel - modules to setup devices. For all devices the kernel exports a major/minor - number; if needed, udev creates a device node with the default kernel - name. If specified, udev applies permissions/ownership to the device - node, creates additional symlinks pointing to the node, and executes - programs to handle the device. - - - The events udev handles, and the information udev merges into its device - database, can be accessed with libudev: - http://www.kernel.org/pub/linux/utils/kernel/hotplug/libudev/ - http://www.kernel.org/pub/linux/utils/kernel/hotplug/gudev/ - -For more details about udev and udev rules see the udev(7) man page. - -Please direct any comment/question to the linux-hotplug mailing list at: - linux-hotplug@vger.kernel.org +Elogind User, Seat and Session Manager + +Introduction +------------ + +Elogind is the systemd project's "logind", extracted out to be a +standalone daemon. It integrates with PAM to know the set of users +that are logged in to a system and whether they are logged in +graphically, on the console, or remotely. Elogind exposes this +information via the standard org.freedesktop.login1 D-Bus interface, +as well as through the file system using systemd's standard +/run/systemd layout. Elogind also provides "libelogind", which is a +subset of the facilities offered by "libsystemd". There is a +"libelogind.pc" pkg-config file as well. + +All of the credit for elogind should go to the systemd developers. +For more on systemd, see +http://www.freedesktop.org/wiki/Software/systemd. All of the blame +should go to Andy Wingo, who extracted elogind from systemd. + +Contributing +------------ + +Elogind was branched from systemd version 219, and preserves the git +history of the systemd project. The version of elogind is the +upstream systemd version, followed by the patchlevel of elogind. For +example version 219.12 is the twelfth elogind release, which aims to +provide a subset of the interfaces of systemd 219. + +To contribute to elogind, fork the current source code from github: + + https://github.com/elogind/elogind + +Send a pull request for the changes you like. + +To chat about elogind: + + #guix on irc.freenode.org + +Finally, bug reports: + + https://github.com/elogind/elogind/issues + +Why bother? +----------- + +Elogind has been developed for use in GuixSD, the OS distribution of +GNU Guix. See http://gnu.org/s/guix for more on Guix. GuixSD uses a +specific init manager (DMD), for reasons that are not relevant here, +but still aims to eventually be a full-featured distribution that can +run GNOME and other desktop environments. However, to run GNOME these +days means that you need to have support for the login1 D-Bus +interface, which is currently only provided by systemd. That is the +origin of this project: to take the excellent logind functionality +from systemd and provide it as a standalone package. + +We like systemd. We realize that there are people out there that hate +it. You're welcome to use elogind for whatever purpose you like -- +as-is, or as a jumping-off point for other things -- but please don't +use it as part of some anti-systemd vendetta. Systemd hackers are +smart folks that are trying to solve interesting problems on the free +desktop, and their large adoption is largely because they solve +problems that users and developers of user-focused applications care +about. We are appreciative of their logind effort and think that +everyone deserves to run it if they like, even if they use a different +PID 1. + +Differences relative to systemd +------------------------------- + +The pkg-config file is called libelogind, not libsystemd or +libsystemd-logind. + +The headers are in , so like instead +of . + +Libelogind just implements login-related functionality. It also +provides the sd-bus API. + +Unlike systemd, whose logind arranges to manage resources for user +sessions via RPC calls to systemd, in elogind there is no systemd so +there is no global cgroup-based resource management. This has a few +implications: + + * Elogind does not create "slices" for users. Elogind will not + record that users are associated with slices. + + * The /run/systemd/slices directory will always be empty. + + * Elogind does not have the concept of a "scope", internally, as + it's the same as a session. Any API that refers to scopes will + always return an error code. + +On the other hand, elogind does use a similar strategy to systemd in +that it places processes in a private cgroup for organizational +purposes, without installing any controllers (see +http://0pointer.de/blog/projects/cgroups-vs-cgroups.html). This +allows elogind to map arbitrary processes to sessions, even if the +process does the usual double-fork to be reparented to PID 1. + +Elogind does not manage virtual terminals. + +Elogind does monitor power button and the lid switch, like systemd, +but instead of doing RPC to systemd to suspend, poweroff, or restart +the machine, elogind just does this directly. For suspend, hibernate, +and hybrid-sleep, elogind uses the same code as systemd-sleep. +Instead of using a separate sleep.conf file to configure the sleep +behavior, this is included in the [Sleep] section of +/etc/elogind/login.conf. See the example login.conf for more. For +shutdown, reboot, and kexec, elogind shells out to "halt", "reboot", +and "kexec" binaries. + +The loginctl command has the poweroff, reboot, sleep, hibernate, and +hybrid-sleep commands from systemd, as well as the --ignore-inhibitors +flag. + +The PAM module is called pam_elogind.so, not pam_systemd.so. + +Elogind and the running cgroup controller +----------------------------------------- +While 'configure' runs, it will detect which controller is in place. +If no controller is in place, configure will determine, that elogind +should be its own controller, which will be a very limited one. + +This approach should generally work, but if you just have no cgroup +controller in place, yet, or if you are currently switching to +another one, this approach will fail. + +In this case you can do one of the two following things: + + 1) Boot your system with the target init system and cgroup + controller, before configuring and building elogind, or + 2) Use the --with-cgroup-controller=name option. + +Example: If you plan to use openrc, but openrc has not yet booted + the machine, you can use + --with-cgroup-controller=openrc + to let elogind know that openrc will be the controller + in charge. + +However, if you set the controller at configure time to something +different than what is in place, elogind will not start until that +controller is actively used as the primary controller. + +License +------- + +LGPLv2.1+ for all code + + - except src/shared/MurmurHash2.c which is Public Domain + - except src/shared/siphash24.c which is CC0 Public Domain + - except src/journal/lookup3.c which is Public Domain + +Dependencies +------------ + + glibc >= 2.14 + libcap + libmount >= 2.20 (from util-linux) + libseccomp >= 1.0.0 (optional) + libblkid >= 2.24 (from util-linux) (optional) + PAM >= 1.1.2 (optional) + libacl (optional) + libselinux (optional) + make, gcc, and similar tools + +During runtime, you need the following additional dependencies: + + dbus >= 1.4.0 (strictly speaking optional, but recommended) + PolicyKit (optional) + +When building from git, you need the following additional +dependencies: + + pkg-config + docbook-xsl + xsltproc + automake + autoconf + libtool + intltool + gperf + gtkdocize (optional)