summary refs log tree commit diff
path: root/doc
diff options
context:
space:
mode:
authorLaurent Bercot <ska-skaware@skarnet.org>2017-11-20 14:51:36 +0000
committerLaurent Bercot <ska-skaware@skarnet.org>2017-11-20 14:51:36 +0000
commit4ba5ae5776c2e9ba4f297115c19923a928cf3e87 (patch)
tree5a67f39a64dc5df84fc20307aa95d7536890b8b5 /doc
downloadutmps-4ba5ae5776c2e9ba4f297115c19923a928cf3e87.tar.gz
utmps-4ba5ae5776c2e9ba4f297115c19923a928cf3e87.tar.xz
utmps-4ba5ae5776c2e9ba4f297115c19923a928cf3e87.zip
Initial release / rename to utmps
Diffstat (limited to 'doc')
-rw-r--r--doc/index.html132
-rw-r--r--doc/libutmps/index.html171
-rw-r--r--doc/overview.html109
-rw-r--r--doc/upgrade.html28
-rw-r--r--doc/utmps-utmpd.html77
-rw-r--r--doc/utmps-wtmpd.html78
6 files changed, 595 insertions, 0 deletions
diff --git a/doc/index.html b/doc/index.html
new file mode 100644
index 0000000..8d080da
--- /dev/null
+++ b/doc/index.html
@@ -0,0 +1,132 @@
+<html>
+  <head>
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
+    <meta http-equiv="Content-Language" content="en" />
+    <title>utmps - a secure utmp implementation</title>
+    <meta name="Description" content="utmps - a secure utmp implementation" />
+    <meta name="Keywords" content="utmps utmp utmpx unix login accounting wtmp laurent bercot skarnet" />
+    <!-- <link rel="stylesheet" type="text/css" href="//skarnet.org/default.css" /> -->
+  </head>
+<body>
+
+<p>
+<a href="//skarnet.org/software/">Software</a><br />
+<a href="//skarnet.org/">skarnet.org</a>
+</p>
+
+<h1> utmps </h1>
+
+<h2> What is it&nbsp;? </h2>
+
+<p>
+ utmps is an implementation of the <tt>utmp.h</tt> and <tt>utmpx.h</tt>
+family of functions performing user accounting on Unix systems.
+</p>
+
+<p>
+ Traditionally, <tt>utmp</tt> functionality is provided by the system's
+libc. However, not all libcs implement utmp: for instance the
+<a href="https://musl-libc.org/">musl</a> libc, on Linux, does not. The
+main reason for it is that <tt>utmp</tt> functionality is difficult to
+implement in a secure way; in particular, it is impossible to implement
+without either running a daemon or allowing arbitrary programs to tamper
+with user accounting.
+</p>
+
+<p>
+ <tt>utmps</tt> is a secure implementation of user accounting, using
+a daemon as the only authority to manage the utmp and wtmp data; programs
+running utmp functions are just clients to this daemon.
+</p>
+
+<hr />
+
+<ul>
+ <li> <a href="overview.html">An overview of utmps</a> </li>
+</ul>
+
+<hr />
+
+<h2> Installation </h2>
+
+<h3> Requirements </h3>
+
+<ul>
+ <li> A POSIX-compliant system with a standard C development environment </li>
+ <li> GNU make, version 3.81 or later </li>
+ <li> <a href="//skarnet.org/software/skalibs/">skalibs</a> version
+2.6.1.0 or later. It's a build-time requirement. It's also a run-time
+requirement if you link against the shared version of the skalibs
+library. </li>
+</ul>
+
+<h3> Licensing </h3>
+
+<p>
+ s6 is free software. It is available under the
+<a href="http://opensource.org/licenses/ISC">ISC license</a>.
+</p>
+
+<h3> Download </h3>
+
+<ul>
+ <li> The current released version of utmps is <a href="utmps-0.0.1.0.tar.gz">0.0.1.0</a>. </li>
+ <li> Alternatively, you can checkout a copy of the
+<a href="//git.skarnet.org/cgi-bin/cgit.cgi/utmps/">utmps
+git repository</a>:
+<pre> git clone git://git.skarnet.org/utmps </pre> </li>
+ <li> There's also a
+<a href="https://github.com/skarnet/utmps">GitHub mirror</a>
+of the utmps git repository. </li>
+</ul>
+
+<h3> Compilation </h3>
+
+<ul>
+ <li> See the enclosed INSTALL file for installation details. </li>
+</ul>
+
+<h3> Upgrade notes </h3>
+
+<ul>
+ <li> <a href="upgrade.html">This page</a> lists the differences to be aware of between
+the previous versions of utmps and the current one. </li>
+</ul>
+
+<hr />
+
+<h2> Reference </h2>
+
+<h3> Commands </h3>
+
+<ul>
+<li><a href="utmps-utmpd.html">The <tt>utmps-utmpd</tt> program</a></li>
+<li><a href="utmps-wtmpd.html">The <tt>utmps-wtmpd</tt> program</a></li>
+</ul>
+
+<h3> Libraries </h3>
+
+<ul>
+<li> <a href="libutmps/">The <tt>utmps</tt> library interface</a> </li>
+<li> <a href="http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/utmpx.h.html">The
+<tt>utmpx.h</tt> library interface</a> is implemented on top of the utmps library. </li>
+<li> <a href="http://man7.org/linux/man-pages/man3/logwtmp.3.html">The <tt>logwtmp()</tt>
+and <tt>updwtmpx()</tt> functions</a> are also implemented. </li>
+</ul>
+
+<hr />
+
+<a name="related">
+<h2> Related resources </h2>
+</a>
+
+<h3> utmps discussion </h3>
+
+<ul>
+ <li> <tt>utmps</tt> is discussed on the
+<a href="//skarnet.org/lists.html#skaware">skaware</a> mailing-list. </li>
+</ul>
+
+</body>
+</html>
diff --git a/doc/libutmps/index.html b/doc/libutmps/index.html
new file mode 100644
index 0000000..0372726
--- /dev/null
+++ b/doc/libutmps/index.html
@@ -0,0 +1,171 @@
+<html>
+  <head>
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
+    <meta http-equiv="Content-Language" content="en" />
+    <title>utmps: the utmps library interface</title>
+    <meta name="Description" content="utmps: the utmps library interface" />
+    <meta name="Keywords" content="utmps utmp wtmp library libutmps" />
+    <!-- <link rel="stylesheet" type="text/css" href="//skarnet.org/default.css" /> -->
+  </head>
+<body>
+
+<p>
+<a href="../">utmps</a><br />
+<a href="//skarnet.org/software/">Software</a><br />
+<a href="//skarnet.org/">skarnet.org</a>
+</p>
+
+<h1> The <tt>utmps</tt> library interface </h1>
+
+<h2> General information </h2>
+
+<p>
+ <tt>libutmps</tt> is a client library meant to be used by client
+programs needing utmp functionality. It interacts with the
+<a href="../utmps-utmpd.html">utmps-utmpd</a> and
+<a href="../utmps-wtmpd.html">utmps-wtmpd</a> daemons.
+</p>
+
+<p>
+ Application programs can use it directly, but most existing programs
+simply use the standard
+<a href="http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/utmpx.h.html">utmpx.h</a>
+interface, which in utmps is implemented as a series of thin wrappers
+around the utmps library.
+</p>
+
+<h2> Compiling </h2>
+
+<ul>
+ <li> Make sure the utmps headers, as well as the skalibs headers,
+are visible in your header search path. </li>
+ <li> Use <tt>#include &lt;utmps/utmps.h&gt;</tt> </li>
+ <li> To use the standard <tt>utmpx.h</tt> interface, you can
+just <tt>#include &lt;utmpx.h&gt;</tt>, which will work as long
+as the <tt>utmps/utmpx.h</tt> header is accessible in your header
+search path. </li>
+</ul>
+
+<h2> Linking </h2>
+
+<ul>
+ <li> Make sure the utmps library, as well as the skalibs library,
+are visible in your library search path. </li>
+ <li> Link against <tt>-lutmps</tt>, <tt>-lskarnet</tt>, </li>
+<tt>`cat $SYSDEPS/socket.lib`</tt> and
+<tt>`cat $SYSDEPS/tainnow.lib`</tt>, $SYSDEPS being your skalibs
+sysdeps directory. </li>
+</ul>
+
+<h2> Programming </h2>
+
+<p>
+ Check the <tt>utmps/utmps.h</tt> header for the exact function list,
+and the <tt>utmps/utmpx.h</tt> header for the definition of the standard
+<tt>struct utmpx</tt> data type.
+</p>
+
+<h3> Synchronous functions with a specified maximum execution time </h3>
+
+<p>
+ The standard utmpx.h functions are fully synchronous. They were not
+initially meant to perform inter-processus communication; however, in
+utmps, they do. Their synchronous nature is obviously not changed here,
+but the underlying utmps functions use a safety mechanism to bound their
+execution time in case daemons fail to respond. This mechanism is described,
+for instance,
+<a href="//skarnet.org/libs6/ftrigr.html#synctimed">here</a>.
+</p>
+
+<h3> Starting and ending a session </h3>
+
+<p>
+<code>int utmps_start (utmps *a, char const *path, tain_t const *deadline, tain_t *stamp)</code> <br />
+Connects to a <tt>utmps-utmpd</tt> service listening on a Unix domain socket at <em>path</em>.
+<em>a</em> must point to a previously allocated <em>utmps</em> object, which is flat and can
+be allocated in the stack. This object must have been initialized to UTMPS_ZERO before the call.
+<em>a</em> will be a handle describing the session, and must be given to all utmps functions
+called in that session.
+<em>deadline</em> and <em>stamp</em> are used to bound the execution time as described in the
+above link. The function returns 1 if it succeeds; it returns 0, and sets errno, if it fails.
+</p>
+
+<p>
+<code>void utmps_end (utmps *a)</code> <br />
+Ends the session described by <em>a</em>, and releases all used resources.
+</p>
+
+<h3> Reading from the utmp database </h3>
+
+<p>
+ Any user authorized to connect to the utmpd service can call these functions. In other
+words, if <tt>utmps_start()</tt> succeeded, then these functions should not fail due to
+insufficient permissions.
+</p>
+
+<p>
+<code>int utmps_rewind (utmps *a, tain_t const *deadline, tain_t *stamp)</code> <br />
+Performs the <tt>setutxent()</tt> functionality on the utmp database addressed via <em>a</em>,
+i.e. sets the internal pointer at the start of the database.
+On success, stores the result into <em>*b</em> and returns 1. On failure, returns 0 and sets errno.
+</p>
+
+<p>
+<code>int utmps_getent (utmps *a, struct utmpx *b, tain_t const *deadline, tain_t *stamp)</code> <br />
+Performs the <tt>getutxent()</tt> functionality on the utmp database addressed via <em>a</em>.
+On success, stores the result into <em>*b</em> and returns 1. On failure, returns 0 and sets errno.
+</p>
+
+<p>
+<code>int utmps_getid (utmps *a, unsigned short type, char const *id, struct utmpx *b, tain_t const *deadline, tain_t *stamp)</code> <br />
+Performs the <tt>getutxid()</tt> functionality on the utmp database addressed via <em>a</em>,
+using ut_type <em>type</em> and ut_id <em>id</em>. <em>id</em> must be a null-terminated
+string; only its first UTMPS_UT_IDSIZE-1 characters will be taken into account.
+On success, the function stores the result into <em>*b</em> and returns 1. On failure,
+it returns 0 and sets errno.
+</p>
+
+<p>
+<code>int utmps_getline (utmps *a, char const *line, struct utmpx *b, tain_t const *deadline, tain_t *stamp)</code> <br />
+Performs the <tt>getutxline()</tt> functionality on the utmp database addressed via <em>a</em>,
+using ut_line <em>line</em>. <em>line</em> must be a null-terminated
+string; only its first UTMPS_UT_LINESIZE-1 characters will be taken into account.
+On success, the function stores the result into <em>*b</em> and returns 1. On failure,
+it returns 0 and sets errno.
+</p>
+
+<h3> Writing to the utmp database </h3>
+
+<p>
+ Currently, only the super-user is allowed to use this function.
+</p>
+
+<p>
+<code>int utmps_putline (utmps *a, struct utmpx const *b, tain_t const *deadline, tain_t *stamp)</code> <br />
+Performs the <tt>pututxline()</tt> functionality on the utmp database addressed via <em>a</em>,
+i.e. writes the <em>*b</em> structure into the utmp database looking for an appropriate
+record to replace, and appending to the database if no such record can be found.
+On success, the function returns 1. On failure, it returns 0 and sets errno.
+</p>
+
+<h3> Writing to the wtmp database </h3>
+
+<p>
+<code>int utmps_updwtmpx (char const *path, struct utmpx const *b, tain_t const *deadline, tain_t *stamp)</code> <br />
+Unlike the previous functions, <tt>utmps_updwtmpx()</tt> does not use a utmps handle, because
+it does not connect to an utmpd service. Instead, it connects to a wtmpd service listening
+on Unix domain socket <em>path</em>, once for every call. It appends the <em>*b</em> structure
+to the wtmp database, returning 1 on success and 0 (and setting errno) on failure.
+</p>
+
+<p>
+ <tt>utmps_updwtmpx()</tt> will only succeed if the caller is root, or if
+<em>b&rarr;ut_user</em> resolves (according to <tt>getpwnam()</tt>) to the
+effective uid of the caller. In other words: users can append phony records
+for themselves, but not for others, and only root can spoof the whole
+wtmp database.
+</p>
+
+</body>
+</html>
diff --git a/doc/overview.html b/doc/overview.html
new file mode 100644
index 0000000..22e9210
--- /dev/null
+++ b/doc/overview.html
@@ -0,0 +1,109 @@
+<html>
+  <head>
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
+    <meta http-equiv="Content-Language" content="en" />
+    <title>utmps: an overview</title>
+    <meta name="Description" content="s6: an overview" />
+    <meta name="Keywords" content="utmps overview utmp wtmp utmpx login user accounting unix" />
+    <!-- <link rel="stylesheet" type="text/css" href="//skarnet.org/default.css" /> -->
+  </head>
+<body>
+
+<p>
+<a href="index.html">utmps</a><br />
+<a href="//skarnet.org/software/">Software</a><br />
+<a href="//skarnet.org/">skarnet.org</a>
+</p>
+
+<h1> An overview of utmps </h1>
+
+<p>
+ utmps is a secure implementation of the <em>utmp</em> functionality, i.e.
+user accounting on Unix systems. It includes full POSIX
+<a href="http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/utmpx.h.html">utmpx.h</a>
+functionality, a few extensions created by GNU, and an underlying
+<a href="libutmps/">C client library</a> with better error reporting
+than the POSIX interface specifies.
+</p>
+
+<h2> The issues with traditional utmp </h2>
+
+<p>
+ Traditional <em>utmp</em> implementations, as performed by most Unix
+libcs, are woefully insecure. The fundamental issue with <em>utmp</em>
+is that it requires user programs to write to files (the utmp or wtmp
+databases) owned by either root or a specific system user. That means
+having the suid bit set on programs using it.
+</p>
+
+<h2> The utmps solution </h2>
+
+<p>
+ utmps uses the age old Unix client-server model, following the
+adage "one resource &rarr; one daemon". It provides two daemons,
+<a href="utmps-utmpd.html">utmps-utmpd</a> and
+<a href="utmps-wtmpd.html">utmps-wtmpd</a>, which should be the only
+programs allowed to access the utmp and wtmp databases respectively.
+It provides the
+<a href="libutmps/">utmps client library</a> to communicate with
+those daemons; and it implements the
+<a href="http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/utmpx.h.html">utmpx.h</a>
+interfaces, and the extensions, as wrappers for this client library.
+</p>
+
+<h2> Authentication, local services and superservers </h2>
+
+<p>
+<a href="utmps-utmpd.html">utmps-utmpd</a> and
+<a href="utmps-wtmpd.html">utmps-wtmpd</a> do not listen to the
+network themselves. They are designed to serve only one client
+connection, following the
+<a href="http://cr.yp.to/proto/ucspi.txt">UCSPI model</a> -
+also known as the inetd model. To implement the utmpd and wtmpd
+<a href="//skarnet.org/software/s6/localservice.html">local services</a>,
+a Unix domain superserver such as
+<a href="//skarnet.org/software/s6/s6-ipcserver.html">s6-ipcserver</a>
+is required. s6-ipcserver listens to a socket, and spawns a
+<a href="utmps-utmpd.html">utmps-utmpd</a> process when a client
+calls <a href="http://pubs.opengroup.org/onlinepubs/9699919799/functions/setutxent.html">setutxent()</a>
+for instance.
+</p>
+
+<p>
+ The utmpd and wtmpd services must be started at boot time in
+order for utmp calls to succeed. The <tt>examples/</tt> subdirectory
+of the utmps package has examples on how to start those services
+when using the
+<a href="//skarnet.org/software/s6/">s6</a> supervision suite,
+the <a href="//skarnet.org/software/s6-rc/">s6-rc</a> service manager, or
+the <a href="https://wiki.gentoo.org/wiki/OpenRC">OpenRC</a> service manager.
+</p>
+
+<p>
+ This model has advantages and drawbacks. The main drawback is that it requires
+a daemon to be running in order for the system to provide full POSIX
+functionality. The main advantage, on the other hand, is that no program
+needs to be suid or sgid, and permissions can actually be quite fine-grained.
+</p>
+
+<ul>
+ <li> <a href="utmps-utmpd.html">utmps-utmpd</a> will allow any user to
+read from the utmp database, but will only allow root to write to it. </li>
+ <li> <a href="utmps-wtmpd.html">utmps-wtmpd</a> will only allow a user
+to add an entry to the wtmp database if the user is root, or if the
+<tt>ut_user</tt> field of the added entry resolves to the user's effective
+uid. </li>
+  <li> The <a href="//skarnet.org/software/s6/s6-ipcserver.html">s6-ipcserver</a>
+superserver, which is recommended to implement the utmpd and wtmpd services,
+allows fine-tuning the permissions: it is possible to deny users from
+connecting to the service, or to only allow certain groups, etc. </li>
+</ul>
+
+<p>
+ All in all, I believe the flexibility it offers overweighs the inconvenience
+of having to run services before providing utmp/wtmp.
+</p>
+
+</body>
+</html>
diff --git a/doc/upgrade.html b/doc/upgrade.html
new file mode 100644
index 0000000..b5c9db9
--- /dev/null
+++ b/doc/upgrade.html
@@ -0,0 +1,28 @@
+<html>
+  <head>
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
+    <meta http-equiv="Content-Language" content="en" />
+    <title>How to upgrade utmps</title>
+    <meta name="Description" content="How to upgrade utmps" />
+    <meta name="Keywords" content="utmps installation upgrade" />
+    <!-- <link rel="stylesheet" type="text/css" href="//skarnet.org/default.css" /> -->
+  </head>
+<body>
+
+<p>
+<a href="index.html">utmps</a><br />
+<a href="//skarnet.org/software/">Software</a><br />
+<a href="//skarnet.org/">skarnet.org</a>
+</p>
+
+<h1> What has changed in utmps </h1>
+
+<h2> in 0.0.1.0 </h2>
+
+<p>
+ Initial release.
+</p>
+
+</body>
+</html>
diff --git a/doc/utmps-utmpd.html b/doc/utmps-utmpd.html
new file mode 100644
index 0000000..57fa6ac
--- /dev/null
+++ b/doc/utmps-utmpd.html
@@ -0,0 +1,77 @@
+<html>
+  <head>
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
+    <meta http-equiv="Content-Language" content="en" />
+    <title>utmps: the utmps-utmpd program</title>
+    <meta name="Description" content="utmps: the utmps-utmpd program" />
+    <meta name="Keywords" content="utmps utmp utmpd daemon service utmps-utmpd" />
+    <!-- <link rel="stylesheet" type="text/css" href="//skarnet.org/default.css" /> -->
+  </head>
+<body>
+
+<p>
+<a href="index.html">utmps</a><br />
+<a href="//skarnet.org/software/">Software</a><br />
+<a href="//skarnet.org/">skarnet.org</a>
+</p>
+
+<h1> The utmps-utmpd program </h1>
+
+<p>
+utmps-utmpd manages the utmp database. It expects to be able to create, read and
+modify a file named <tt>utmp</tt> in the directory it is launched in.
+</p>
+
+<p>
+ It is not meant to be called directly; instead, it is expected to be run from
+a script as a part of a "utmpd"
+<a href="//skarnet.org/software/s6/localservice.html">local service</a>.
+</p>
+
+<p>
+ The <tt>examples/</tt> subdirectory of the utmps package provides examples
+on how to run such a service.
+ The simplest way to do so, for testing purposes, is a command line such as:
+</p>
+<pre>s6-ipcserver -l0 utmpd-socket utmps-utmpd</pre>
+<p>
+ while being in the <tt>/run/utmps</tt> directory.
+</p>
+
+<p>
+<tt>/run/utmps/utmpd-socket</tt> is the default place where utmps's
+implementation of the <tt>utmpx.h</tt> functions expects the utmpd
+service to be. It can be changed at build time by giving the
+<tt>--with-utmp-socket=PATH</tt> option to configure.
+</p>
+
+<p>
+ utmps-utmpd does not listen to the socket itself: it reads from its
+standard input and writes to its standard output. It relies
+on a superserver such as
+<a href="//skarnet.org/software/s6/s6-ipcserver.html">s6-ipcserver</a>
+to manage connections to the socket. An instance of utmps-utmpd is run
+for every client connection;
+every instance reads the effective uid of the client in an environment
+variable set by the superserver, which allows it to filter operations -
+for instance, it allows any user to read from the database but it only
+allows root to write to it.
+</p>
+
+<p>
+ If more fine-grained authorizations are required (only allowing
+certain users and groups to connect to the service), the superserver
+can be configured to enforce them.
+</p>
+
+<p>
+ utmps-utmpd does not need to run as root, provided it can write its file;
+it is recommended to create a <em>utmp</em> user and group, dedicated to
+utmps-utmpd and
+<a href="utmps-wtmpd.html">utmps-wtmpd</a> operation, and run the
+superserver as this user and group.
+</p>
+
+</body>
+</html>
diff --git a/doc/utmps-wtmpd.html b/doc/utmps-wtmpd.html
new file mode 100644
index 0000000..a1e090b
--- /dev/null
+++ b/doc/utmps-wtmpd.html
@@ -0,0 +1,78 @@
+<html>
+  <head>
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
+    <meta http-equiv="Content-Language" content="en" />
+    <title>utmps: the utmps-wtmpd program</title>
+    <meta name="Description" content="utmps: the utmps-wtmpd program" />
+    <meta name="Keywords" content="utmps wtmp wtmpd daemon service utmps-wtmpd" />
+    <!-- <link rel="stylesheet" type="text/css" href="//skarnet.org/default.css" /> -->
+  </head>
+<body>
+
+<p>
+<a href="index.html">utmps</a><br />
+<a href="//skarnet.org/software/">Software</a><br />
+<a href="//skarnet.org/">skarnet.org</a>
+</p>
+
+<h1> The utmps-wtmpd program </h1>
+
+<p>
+utmps-wtmpd manages the wtmp database. It expects to be able to create, read and
+modify a file named <tt>wtmp</tt> in the directory it is launched in.
+</p>
+
+<p>
+ It is not meant to be called directly; instead, it is expected to be run from
+a script as a part of a "wtmpd"
+<a href="//skarnet.org/software/s6/localservice.html">local service</a>.
+</p>
+
+<p>
+ The <tt>examples/</tt> subdirectory of the utmps package provides examples
+on how to run such a service.
+ The simplest way to do so, for testing purposes, is a command line such as:
+</p>
+<pre>s6-ipcserver -l0 utmpd-socket utmps-wtmpd</pre>
+<p>
+ while being in the <tt>/run/utmps</tt> directory.
+</p>
+
+<p>
+<tt>/run/utmps/wtmpd-socket</tt> is the default place where utmps's
+implementation of the <tt>updwtmpx()</tt> function expects the wtmpd
+service to be. It can be changed at build time by giving the
+<tt>--with-wtmp-socket=PATH</tt> option to configure.
+</p>
+
+<p>
+ utmps-wtmpd does not listen to the socket itself: it reads from its
+standard input and writes to its standard output. It relies
+on a superserver such as
+<a href="//skarnet.org/software/s6/s6-ipcserver.html">s6-ipcserver</a>
+to manage connections to the socket. An instance of utmps-wtmpd is run
+for every client connection;
+every instance reads the effective uid of the client in an environment
+variable set by the superserver, which allows it to filter operations:
+only lines whose <tt>ut_user</tt> field resolves to the effective uid of
+the client will be appended to the database.
+</p>
+
+<p>
+ If more fine-grained authorizations are required (only allowing
+certain users and groups to connect to the service), the superserver
+can be configured to enforce them. For instance, it is possible to
+tell s6-ipcserver to only accept connections from root.
+</p>
+
+<p>
+ utmps-wtmpd does not need to run as root, provided it can write its file;
+it is recommended to create a <em>utmp</em> user and group, dedicated to
+utmps-wtmpd and
+<a href="utmps-utmpd.html">utmps-utmpd</a> operation, and run the
+superserver as this user and group.
+</p>
+
+</body>
+</html>