]>
Commit | Line | Data |
---|---|---|
1 | systemd System and Service Manager | |
2 | ||
3 | DETAILS: | |
4 | http://0pointer.de/blog/projects/systemd.html | |
5 | ||
6 | WEB SITE: | |
7 | https://www.freedesktop.org/wiki/Software/systemd | |
8 | ||
9 | GIT: | |
10 | git@github.com:systemd/systemd.git | |
11 | https://github.com/systemd/systemd | |
12 | ||
13 | MAILING LIST: | |
14 | https://lists.freedesktop.org/mailman/listinfo/systemd-devel | |
15 | ||
16 | IRC: | |
17 | #systemd on irc.freenode.org | |
18 | ||
19 | BUG REPORTS: | |
20 | https://github.com/systemd/systemd/issues | |
21 | ||
22 | AUTHOR: | |
23 | Lennart Poettering | |
24 | Kay Sievers | |
25 | ...and many others | |
26 | ||
27 | LICENSE: | |
28 | LGPLv2.1+ for all code | |
29 | - except src/basic/MurmurHash2.c which is Public Domain | |
30 | - except src/basic/siphash24.c which is CC0 Public Domain | |
31 | - except src/journal/lookup3.c which is Public Domain | |
32 | - except src/udev/* which is (currently still) GPLv2, GPLv2+ | |
33 | ||
34 | REQUIREMENTS: | |
35 | Linux kernel >= 3.13 | |
36 | Linux kernel >= 4.2 for unified cgroup hierarchy support | |
37 | ||
38 | Kernel Config Options: | |
39 | CONFIG_DEVTMPFS | |
40 | CONFIG_CGROUPS (it is OK to disable all controllers) | |
41 | CONFIG_INOTIFY_USER | |
42 | CONFIG_SIGNALFD | |
43 | CONFIG_TIMERFD | |
44 | CONFIG_EPOLL | |
45 | CONFIG_NET | |
46 | CONFIG_SYSFS | |
47 | CONFIG_PROC_FS | |
48 | CONFIG_FHANDLE (libudev, mount and bind mount handling) | |
49 | ||
50 | Kernel crypto/hash API | |
51 | CONFIG_CRYPTO_USER_API_HASH | |
52 | CONFIG_CRYPTO_HMAC | |
53 | CONFIG_CRYPTO_SHA256 | |
54 | ||
55 | udev will fail to work with the legacy sysfs layout: | |
56 | CONFIG_SYSFS_DEPRECATED=n | |
57 | ||
58 | Legacy hotplug slows down the system and confuses udev: | |
59 | CONFIG_UEVENT_HELPER_PATH="" | |
60 | ||
61 | Userspace firmware loading is not supported and should | |
62 | be disabled in the kernel: | |
63 | CONFIG_FW_LOADER_USER_HELPER=n | |
64 | ||
65 | Some udev rules and virtualization detection relies on it: | |
66 | CONFIG_DMIID | |
67 | ||
68 | Support for some SCSI devices serial number retrieval, to | |
69 | create additional symlinks in /dev/disk/ and /dev/tape: | |
70 | CONFIG_BLK_DEV_BSG | |
71 | ||
72 | Required for PrivateNetwork= in service units: | |
73 | CONFIG_NET_NS | |
74 | Note that systemd-localed.service and other systemd units use | |
75 | PrivateNetwork so this is effectively required. | |
76 | ||
77 | Required for PrivateUsers= in service units: | |
78 | CONFIG_USER_NS | |
79 | ||
80 | Optional but strongly recommended: | |
81 | CONFIG_IPV6 | |
82 | CONFIG_AUTOFS4_FS | |
83 | CONFIG_TMPFS_XATTR | |
84 | CONFIG_{TMPFS,EXT4_FS,XFS,BTRFS_FS,...}_POSIX_ACL | |
85 | CONFIG_SECCOMP | |
86 | CONFIG_SECCOMP_FILTER (required for seccomp support) | |
87 | CONFIG_CHECKPOINT_RESTORE (for the kcmp() syscall) | |
88 | ||
89 | Required for CPUShares= in resource control unit settings | |
90 | CONFIG_CGROUP_SCHED | |
91 | CONFIG_FAIR_GROUP_SCHED | |
92 | ||
93 | Required for CPUQuota= in resource control unit settings | |
94 | CONFIG_CFS_BANDWIDTH | |
95 | ||
96 | Required for IPAddressDeny= and IPAddressAllow= in resource control | |
97 | unit settings | |
98 | CONFIG_CGROUP_BPF | |
99 | ||
100 | For UEFI systems: | |
101 | CONFIG_EFIVAR_FS | |
102 | CONFIG_EFI_PARTITION | |
103 | ||
104 | We recommend to turn off Real-Time group scheduling in the | |
105 | kernel when using systemd. RT group scheduling effectively | |
106 | makes RT scheduling unavailable for most userspace, since it | |
107 | requires explicit assignment of RT budgets to each unit whose | |
108 | processes making use of RT. As there's no sensible way to | |
109 | assign these budgets automatically this cannot really be | |
110 | fixed, and it's best to disable group scheduling hence. | |
111 | CONFIG_RT_GROUP_SCHED=n | |
112 | ||
113 | It's a good idea to disable the implicit creation of networking bonding | |
114 | devices by the kernel networking bonding module, so that the | |
115 | automatically created "bond0" interface doesn't conflict with any such | |
116 | device created by systemd-networkd (or other tools). Ideally there | |
117 | would be a kernel compile-time option for this, but there currently | |
118 | isn't. The next best thing is to make this change through a modprobe.d | |
119 | drop-in. This is shipped by default, see modprobe.d/systemd.conf. | |
120 | ||
121 | Required for systemd-nspawn: | |
122 | CONFIG_DEVPTS_MULTIPLE_INSTANCES or Linux kernel >= 4.7 | |
123 | ||
124 | Note that kernel auditing is broken when used with systemd's | |
125 | container code. When using systemd in conjunction with | |
126 | containers, please make sure to either turn off auditing at | |
127 | runtime using the kernel command line option "audit=0", or | |
128 | turn it off at kernel compile time using: | |
129 | CONFIG_AUDIT=n | |
130 | If systemd is compiled with libseccomp support on | |
131 | architectures which do not use socketcall() and where seccomp | |
132 | is supported (this effectively means x86-64 and ARM, but | |
133 | excludes 32-bit x86!), then nspawn will now install a | |
134 | work-around seccomp filter that makes containers boot even | |
135 | with audit being enabled. This works correctly only on kernels | |
136 | 3.14 and newer though. TL;DR: turn audit off, still. | |
137 | ||
138 | glibc >= 2.16 | |
139 | libcap | |
140 | libmount >= 2.30 (from util-linux) | |
141 | (util-linux *must* be built without --enable-libmount-support-mtab) | |
142 | libseccomp >= 2.3.1 (optional) | |
143 | libblkid >= 2.24 (from util-linux) (optional) | |
144 | libkmod >= 15 (optional) | |
145 | PAM >= 1.1.2 (optional) | |
146 | libcryptsetup (optional) | |
147 | libaudit (optional) | |
148 | libacl (optional) | |
149 | libselinux (optional) | |
150 | liblzma (optional) | |
151 | liblz4 >= 119 (optional) | |
152 | libgcrypt (optional) | |
153 | libqrencode (optional) | |
154 | libmicrohttpd (optional) | |
155 | libpython (optional) | |
156 | libidn2 or libidn (optional) | |
157 | elfutils >= 158 (optional) | |
158 | polkit (optional) | |
159 | pkg-config | |
160 | gperf | |
161 | docbook-xsl (optional, required for documentation) | |
162 | xsltproc (optional, required for documentation) | |
163 | python-lxml (optional, required to build the indices) | |
164 | python, meson, ninja | |
165 | gcc, awk, sed, grep, m4, and similar tools | |
166 | ||
167 | During runtime, you need the following additional | |
168 | dependencies: | |
169 | ||
170 | util-linux >= v2.27.1 required | |
171 | dbus >= 1.4.0 (strictly speaking optional, but recommended) | |
172 | NOTE: If using dbus < 1.9.18, you should override the default | |
173 | policy directory (--with-dbuspolicydir=/etc/dbus-1/system.d). | |
174 | dracut (optional) | |
175 | PolicyKit (optional) | |
176 | ||
177 | To build in directory build/: | |
178 | meson build/ && ninja -C build | |
179 | ||
180 | Any configuration options can be specfied as -Darg=value... arguments | |
181 | to meson. After the build directory is initially configured, meson will | |
182 | refuse to run again, and options must be changed with: | |
183 | mesonconf -Darg=value... | |
184 | mesonconf without any arguments will print out available options and | |
185 | their current values. | |
186 | ||
187 | Useful commands: | |
188 | ninja -v some/target | |
189 | ninja test | |
190 | sudo ninja install | |
191 | DESTDIR=... ninja install | |
192 | ||
193 | A tarball can be created with: | |
194 | git archive --format=tar --prefix=systemd-222/ v222 | xz > systemd-222.tar.xz | |
195 | ||
196 | When systemd-hostnamed is used, it is strongly recommended to | |
197 | install nss-myhostname to ensure that, in a world of | |
198 | dynamically changing hostnames, the hostname stays resolvable | |
199 | under all circumstances. In fact, systemd-hostnamed will warn | |
200 | if nss-myhostname is not installed. | |
201 | ||
202 | nss-systemd must be enabled on systemd systems, as that's required for | |
203 | DynamicUser= to work. Note that we ship services out-of-the-box that | |
204 | make use of DynamicUser= now, hence enabling nss-systemd is not | |
205 | optional. | |
206 | ||
207 | Note that the build prefix for systemd must be /usr. -Dsplit-usr=false | |
208 | (which is the default and does not need to be specified) is the | |
209 | recommended setting, and -Dsplit-usr=true should be used on systems | |
210 | which have /usr on a separate partition. | |
211 | ||
212 | Additional packages are necessary to run some tests: | |
213 | - busybox (used by test/TEST-13-NSPAWN-SMOKE) | |
214 | - nc (used by test/TEST-12-ISSUE-3171) | |
215 | - python3-pyparsing | |
216 | - python3-evdev (used by hwdb parsing tests) | |
217 | - strace (used by test/test-functions) | |
218 | - capsh (optional, used by test-execute) | |
219 | ||
220 | USERS AND GROUPS: | |
221 | Default udev rules use the following standard system group | |
222 | names, which need to be resolvable by getgrnam() at any time, | |
223 | even in the very early boot stages, where no other databases | |
224 | and network are available: | |
225 | ||
226 | audio, cdrom, dialout, disk, input, kmem, kvm, lp, render, tape, tty, video | |
227 | ||
228 | During runtime, the journal daemon requires the | |
229 | "systemd-journal" system group to exist. New journal files will | |
230 | be readable by this group (but not writable), which may be used | |
231 | to grant specific users read access. In addition, system | |
232 | groups "wheel" and "adm" will be given read-only access to | |
233 | journal files using systemd-tmpfiles.service. | |
234 | ||
235 | The journal gateway daemon requires the | |
236 | "systemd-journal-gateway" system user and group to | |
237 | exist. During execution this network facing service will drop | |
238 | privileges and assume this uid/gid for security reasons. | |
239 | ||
240 | Similarly, the NTP daemon requires the "systemd-timesync" system | |
241 | user and group to exist. | |
242 | ||
243 | Similarly, the network management daemon requires the | |
244 | "systemd-network" system user and group to exist. | |
245 | ||
246 | Similarly, the name resolution daemon requires the | |
247 | "systemd-resolve" system user and group to exist. | |
248 | ||
249 | Similarly, the coredump support requires the | |
250 | "systemd-coredump" system user and group to exist. | |
251 | ||
252 | NSS: | |
253 | systemd ships with four glibc NSS modules: | |
254 | ||
255 | nss-myhostname resolves the local hostname to locally | |
256 | configured IP addresses, as well as "localhost" to | |
257 | 127.0.0.1/::1. | |
258 | ||
259 | nss-resolve enables DNS resolution via the systemd-resolved | |
260 | DNS/LLMNR caching stub resolver "systemd-resolved". | |
261 | ||
262 | nss-mymachines enables resolution of all local containers registered | |
263 | with machined to their respective IP addresses. It also maps UID/GIDs | |
264 | ranges used by containers to useful names. | |
265 | ||
266 | nss-systemd enables resolution of all dynamically allocated service | |
267 | users. (See the DynamicUser= setting in unit files.) | |
268 | ||
269 | To make use of these NSS modules, please add them to the "hosts:", | |
270 | "passwd:" and "group:" lines in /etc/nsswitch.conf. The "resolve" | |
271 | module should replace the glibc "dns" module in this file (and don't | |
272 | worry, it chain-loads the "dns" module if it can't talk to resolved). | |
273 | ||
274 | The four modules should be used in the following order: | |
275 | ||
276 | passwd: compat mymachines systemd | |
277 | group: compat mymachines systemd | |
278 | hosts: files mymachines resolve myhostname | |
279 | ||
280 | SYSV INIT.D SCRIPTS: | |
281 | When calling "systemctl enable/disable/is-enabled" on a unit which is a | |
282 | SysV init.d script, it calls /usr/lib/systemd/systemd-sysv-install; | |
283 | this needs to translate the action into the distribution specific | |
284 | mechanism such as chkconfig or update-rc.d. Packagers need to provide | |
285 | this script if you need this functionality (you don't if you disabled | |
286 | SysV init support). | |
287 | ||
288 | Please see src/systemctl/systemd-sysv-install.SKELETON for how this | |
289 | needs to look like, and provide an implementation at the marked places. | |
290 | ||
291 | WARNINGS: | |
292 | systemd will warn during early boot if /usr is not already mounted at | |
293 | this point (that means: either located on the same file system as / or | |
294 | already mounted in the initrd). While in systemd itself very little | |
295 | will break if /usr is on a separate, late-mounted partition, many of | |
296 | its dependencies very likely will break sooner or later in one form or | |
297 | another. For example, udev rules tend to refer to binaries in /usr, | |
298 | binaries that link to libraries in /usr or binaries that refer to data | |
299 | files in /usr. Since these breakages are not always directly visible, | |
300 | systemd will warn about this, since this kind of file system setup is | |
301 | not really supported anymore by the basic set of Linux OS components. | |
302 | ||
303 | systemd requires that the /run mount point exists. systemd also | |
304 | requires that /var/run is a symlink to /run. | |
305 | ||
306 | For more information on this issue consult | |
307 | https://www.freedesktop.org/wiki/Software/systemd/separate-usr-is-broken | |
308 | ||
309 | To run systemd under valgrind, compile with VALGRIND defined | |
310 | (e.g. CPPFLAGS='... -DVALGRIND=1' meson <options>) and have valgrind | |
311 | development headers installed (i.e. valgrind-devel or | |
312 | equivalent). Otherwise, false positives will be triggered by code which | |
313 | violates some rules but is actually safe. Note that valgrind generates | |
314 | nice output only on exit(), hence on shutdown we don't execve() | |
315 | systemd-shutdown. | |
316 | ||
317 | ENGINEERING AND CONSULTING SERVICES: | |
318 | Kinvolk (https://kinvolk.io) offers professional engineering | |
319 | and consulting services for systemd. Please contact Chris Kühl | |
320 | <chris@kinvolk.io> for more information. |