Implement SNI when using DNS-over-TLS

Some DNS providers need SNI to identify client.

This can be used by adding #name to a DNS.
Example:
[Resolve]
DNS=192.168.1.1#example.com

TODO

@@ -701,7 +701,6 @@ Features:
   - allow multiple signal handlers per signal?
   - document chaining of signal handler for SIGCHLD and child handlers
   - define more intervals where we will shift wakeup intervals around in, 1h, 6h, 24h, ...
-  - generate a failure of a default event loop is executed out-of-thread
 
 * investigate endianness issues of UUID vs. GUID
 

hwdb.d/60-keyboard.hwdb

@@ -207,6 +207,10 @@ evdev:atkbd:dmi:bvn*:bvr*:bd*:svnAlienware*:pn*:pvr*
  KEYBOARD_KEY_c1=!prog2                                 #graphics amplifier, undock-button event
  KEYBOARD_KEY_c2=!power                                 #graphics amplifier, surprise undock event
 
+# Alienware M17xR3 laptops
+evdev:atkbd:dmi:bvn*:bvr*:bd*:svnAlienware*:pnM17xR3:*
+ KEYBOARD_KEY_89=ejectcd
+
 ###########################################################
 # Asus
 ###########################################################

hwdb.d/70-mouse.hwdb

@@ -471,7 +471,7 @@ mouse:usb:v046dp4069:name:Logitech MX Master 2s:
 # Logitech MX Master 2S (via Bluetooth)
 # Horiz wheel has 14 stops, angle is rounded up
 mouse:bluetooth:v046dpb019:name:MX Master 2S Mouse:
- MOUSE_DPI=2000@2000
+ MOUSE_DPI=1000@2000
  MOUSE_WHEEL_CLICK_ANGLE=15
  MOUSE_WHEEL_CLICK_ANGLE_HORIZONTAL=26
  MOUSE_WHEEL_CLICK_COUNT=24

man/journald.conf.xml

@@ -139,6 +139,51 @@
         <literal>us</literal>. To turn off any kind of rate limiting,
         set either value to 0.</para>
 
+        <para>Note that the effective rate limit is multiplied with a
+        factor derived from the available free disk space for the journal.
+        Currently, this factor is calculated using the base 2 logarithm.</para>
+
+        <table>
+          <title>Example <varname>RateLimitBurst=</varname> rate
+            modifications by the available disk space</title>
+          <tgroup cols='2'>
+            <colspec colname='freespace' />
+            <colspec colname='multiplier' />
+            <thead>
+              <row>
+                <entry>Available Disk Space</entry>
+                <entry>Burst Multiplier</entry>
+              </row>
+            </thead>
+            <tbody>
+              <row>
+                <entry>&lt;= 1MB</entry>
+                <entry>1</entry>
+              </row>
+              <row>
+                <entry>&lt;= 16MB</entry>
+                <entry>2</entry>
+              </row>
+              <row>
+                <entry>&lt;= 256MB</entry>
+                <entry>3</entry>
+              </row>
+              <row>
+                <entry>&lt;= 4GB</entry>
+                <entry>4</entry>
+              </row>
+              <row>
+                <entry>&lt;= 64GB</entry>
+                <entry>5</entry>
+              </row>
+              <row>
+                <entry>&lt;= 1TB</entry>
+                <entry>6</entry>
+              </row>
+            </tbody>
+          </tgroup>
+        </table>
+
         <para>If a service provides rate limits for itself through
         <varname>LogRateLimitIntervalSec=</varname> and/or <varname>LogRateLimitBurst=</varname>
         in <citerefentry><refentrytitle>systemd.exec</refentrytitle><manvolnum>5</manvolnum></citerefentry>,

man/kernel-command-line.xml

@@ -393,6 +393,17 @@
         </listitem>
       </varlistentry>
 
+      <varlistentry>
+        <term><varname>systemd.cpu_affinity=</varname></term>
+
+        <listitem>
+          <para>Overrides the CPU affinity mask for the service manager and the default for all child
+          processes it forks. This takes precedence over <varname>CPUAffinity=</varname>, see
+          <citerefentry><refentrytitle>systemd-system.conf</refentrytitle><manvolnum>5</manvolnum></citerefentry>
+          for details.</para>
+        </listitem>
+      </varlistentry>
+
       <varlistentry>
         <term><varname>modules_load=</varname></term>
         <term><varname>rd.modules_load=</varname></term>

man/resolved.conf.xml

@@ -214,6 +214,9 @@
         resolver is not capable of authenticating the server, so it is
         vulnerable to "man-in-the-middle" attacks.</para>
 
+        <para>Server Name Indication (SNI) can be used when opening a TLS connection.
+        Entries in <varname>DNS=</varname> should be in format <literal>address#server_name</literal>.</para>
+
         <para>In addition to this global DNSOverTLS setting
         <citerefentry><refentrytitle>systemd-networkd.service</refentrytitle><manvolnum>8</manvolnum></citerefentry>
         also maintains per-link DNSOverTLS settings. For system DNS

man/rules/meson.build

@@ -371,7 +371,15 @@ manpages = [
  ['sd_bus_wait', '3', [], ''],
  ['sd_event_add_child',
   '3',
-  ['sd_event_child_handler_t', 'sd_event_source_get_child_pid'],
+  ['sd_event_add_child_pidfd',
+   'sd_event_child_handler_t',
+   'sd_event_source_get_child_pid',
+   'sd_event_source_get_child_pidfd',
+   'sd_event_source_get_child_pidfd_own',
+   'sd_event_source_get_child_process_own',
+   'sd_event_source_send_child_signal',
+   'sd_event_source_set_child_pidfd_own',
+   'sd_event_source_set_child_process_own'],
   ''],
  ['sd_event_add_defer',
   '3',

man/sd_event_add_child.xml

@@ -17,7 +17,14 @@
 
   <refnamediv>
     <refname>sd_event_add_child</refname>
+    <refname>sd_event_add_child_pidfd</refname>
     <refname>sd_event_source_get_child_pid</refname>
+    <refname>sd_event_source_get_child_pidfd</refname>
+    <refname>sd_event_source_get_child_pidfd_own</refname>
+    <refname>sd_event_source_set_child_pidfd_own</refname>
+    <refname>sd_event_source_get_child_process_own</refname>
+    <refname>sd_event_source_set_child_process_own</refname>
+    <refname>sd_event_source_send_child_signal</refname>
     <refname>sd_event_child_handler_t</refname>
 
     <refpurpose>Add a child process state change event source to an event loop</refpurpose>
@@ -46,40 +53,77 @@
         <paramdef>void *<parameter>userdata</parameter></paramdef>
       </funcprototype>
 
+      <funcprototype>
+        <funcdef>int <function>sd_event_add_child_pidfd</function></funcdef>
+        <paramdef>sd_event *<parameter>event</parameter></paramdef>
+        <paramdef>sd_event_source **<parameter>source</parameter></paramdef>
+        <paramdef>int <parameter>pidfd</parameter></paramdef>
+        <paramdef>int <parameter>options</parameter></paramdef>
+        <paramdef>sd_event_child_handler_t <parameter>handler</parameter></paramdef>
+        <paramdef>void *<parameter>userdata</parameter></paramdef>
+      </funcprototype>
+
       <funcprototype>
         <funcdef>int <function>sd_event_source_get_child_pid</function></funcdef>
         <paramdef>sd_event_source *<parameter>source</parameter></paramdef>
         <paramdef>pid_t *<parameter>pid</parameter></paramdef>
       </funcprototype>
 
+      <funcprototype>
+        <funcdef>int <function>sd_event_source_get_child_pidfd</function></funcdef>
+        <paramdef>sd_event_source *<parameter>source</parameter></paramdef>
+      </funcprototype>
+
+      <funcprototype>
+        <funcdef>int <function>sd_event_source_get_child_pidfd_own</function></funcdef>
+        <paramdef>sd_event_source *<parameter>source</parameter></paramdef>
+      </funcprototype>
+
+      <funcprototype>
+        <funcdef>int <function>sd_event_source_set_child_pidfd_own</function></funcdef>
+        <paramdef>sd_event_source *<parameter>source</parameter></paramdef>
+        <paramdef>int <parameter>own</parameter></paramdef>
+      </funcprototype>
+
+      <funcprototype>
+        <funcdef>int <function>sd_event_source_get_child_process_own</function></funcdef>
+        <paramdef>sd_event_source *<parameter>source</parameter></paramdef>
+      </funcprototype>
+
+      <funcprototype>
+        <funcdef>int <function>sd_event_source_set_child_process_own</function></funcdef>
+        <paramdef>sd_event_source *<parameter>source</parameter></paramdef>
+        <paramdef>int <parameter>own</parameter></paramdef>
+      </funcprototype>
+
+      <funcprototype>
+        <funcdef>int <function>sd_event_source_send_child_signal</function></funcdef>
+        <paramdef>sd_event_source *<parameter>source</parameter></paramdef>
+        <paramdef>int <parameter>sig</parameter></paramdef>
+        <paramdef>const siginfo_t *<parameter>info</parameter></paramdef>
+        <paramdef>unsigned <parameter>flags</parameter></paramdef>
+      </funcprototype>
+
     </funcsynopsis>
   </refsynopsisdiv>
 
   <refsect1>
     <title>Description</title>
 
-    <para><function>sd_event_add_child()</function> adds a new child
-    process state change event source to an event loop. The event loop
-    object is specified in the <parameter>event</parameter> parameter,
-    the event source object is returned in the
-    <parameter>source</parameter> parameter. The
-    <parameter>pid</parameter> parameter specifies the PID of the
-    process to watch. The <parameter>handler</parameter> must
-    reference a function to call when the process changes state. The
-    handler function will be passed the
-    <parameter>userdata</parameter> pointer, which may be chosen
-    freely by the caller. The handler also receives a pointer to a
-    <structname>siginfo_t</structname> structure containing
-    information about the child process event. The
-    <parameter>options</parameter> parameter determines which state
-    changes will be watched for. It must contain an OR-ed mask of
-    <constant>WEXITED</constant> (watch for the child process
-    terminating), <constant>WSTOPPED</constant> (watch for the child
-    process being stopped by a signal), and
-    <constant>WCONTINUED</constant> (watch for the child process being
-    resumed by a signal). See <citerefentry
-    project='man-pages'><refentrytitle>waitid</refentrytitle><manvolnum>2</manvolnum></citerefentry>
-    for further information.</para>
+    <para><function>sd_event_add_child()</function> adds a new child process state change event source to an
+    event loop. The event loop object is specified in the <parameter>event</parameter> parameter, the event
+    source object is returned in the <parameter>source</parameter> parameter. The <parameter>pid</parameter>
+    parameter specifies the PID of the process to watch, which must be a direct child process of the invoking
+    process. The <parameter>handler</parameter> must reference a function to call when the process changes
+    state. The handler function will be passed the <parameter>userdata</parameter> pointer, which may be
+    chosen freely by the caller. The handler also receives a pointer to a <structname>siginfo_t</structname>
+    structure containing information about the child process event. The <parameter>options</parameter>
+    parameter determines which state changes will be watched for. It must contain an OR-ed mask of
+    <constant>WEXITED</constant> (watch for the child process terminating), <constant>WSTOPPED</constant>
+    (watch for the child process being stopped by a signal), and <constant>WCONTINUED</constant> (watch for
+    the child process being resumed by a signal). See <citerefentry
+    project='man-pages'><refentrytitle>waitid</refentrytitle><manvolnum>2</manvolnum></citerefentry> for
+    further information.</para>
 
     <para>Only a single handler may be installed for a specific
     child process. The handler is enabled for a single event
@@ -100,6 +144,12 @@
     <constant>SD_EVENT_OFF</constant> with
     <citerefentry><refentrytitle>sd_event_source_set_enabled</refentrytitle><manvolnum>3</manvolnum></citerefentry>.</para>
 
+    <para>The <constant>SIGCHLD</constant> signal must be blocked in all threads before this function is
+    called (using <citerefentry
+    project='man-pages'><refentrytitle>sigprocmask</refentrytitle><manvolnum>2</manvolnum></citerefentry> or
+    <citerefentry
+    project='man-pages'><refentrytitle>pthread_sigmask</refentrytitle><manvolnum>3</manvolnum></citerefentry>).</para>
+
     <para>If the second parameter of
     <function>sd_event_add_child()</function> is passed as NULL no
     reference to the event source object is returned. In this case the
@@ -121,6 +171,17 @@
     processed first, it should leave the child processes for which
     child process state change event sources are installed unreaped.</para>
 
+    <para><function>sd_event_add_child_pidfd()</function> is similar to
+    <function>sd_event_add_child()</function> but takes a file descriptor referencing the process ("pidfd")
+    instead of the numeric PID. A suitable file descriptor may be acquired via <citerefentry
+    project='man-pages'><refentrytitle>pidfd_open</refentrytitle><manvolnum>2</manvolnum></citerefentry> and
+    related calls. The passed file descriptor is not closed when the event source is freed again, unless
+    <function>sd_event_source_set_child_pidfd_own()</function> is used to turn this behaviour on. Note that
+    regardless which of <function>sd_event_add_child()</function> and
+    <function>sd_event_add_child_pidfd()</function> is used for allocating an event source, the watched
+    process has to be a direct child process of the invoking process. Also in both cases
+    <constant>SIGCHLD</constant> has to be blocked in the invoking process.</para>
+
     <para><function>sd_event_source_get_child_pid()</function>
     retrieves the configured PID of a child process state change event
     source created previously with
@@ -129,6 +190,45 @@
     pointer to a <type>pid_t</type> variable to return the process ID
     in.
     </para>
+
+    <para><function>sd_event_source_get_child_pidfd()</function> retrieves the file descriptor referencing
+    the watched process ("pidfd") if this functionality is available. On kernels that support the concept the
+    event loop will make use of pidfds to watch child processes, regardless if the individual event sources
+    are allocated via <function>sd_event_add_child()</function> or
+    <function>sd_event_add_child_pidfd()</function>. If the latter call was used to allocate the event
+    source, this function returns the file descriptor used for allocation. On kernels that do not support the
+    pidfd concept this function will fail with <constant>EOPNOTSUPP</constant>. This call takes the event
+    source object as the <parameter>source</parameter> parameter and returns the numeric file descriptor.
+    </para>
+
+    <para><function>sd_event_source_get_child_pidfd_own()</function> may be used to query whether the pidfd
+    the event source encapsulates shall be closed when the event source is freed. This function returns zero
+    if the pidfd shall be left open, and positive if it shall be closed automatically. By default this
+    setting defaults to on if the event source was allocated via <function>sd_event_add_child()</function>
+    and off if it was allocated via <function>sd_event_add_child_pidfd()</function>. The
+    <function>sd_event_source_set_child_pidfd_own()</function> function may be used to change the setting and
+    takes a boolean parameter with the new setting.</para>
+
+    <para><function>sd_event_source_get_child_process_own()</function> may be used to query whether the
+    process the event source watches shall be killed (with <constant>SIGKILL</constant>) and reaped when the
+    event source is freed. This function returns zero if the process shell be left running, and positive if
+    it shall be killed and reaped automatically. By default this setting defaults to off. The
+    <function>sd_event_source_set_child_process_own()</function> function may be used to change the setting
+    and takes a boolean parameter with the new setting. Note that currently if the calling process is
+    terminated abnormally the watched process might survive even thought the event source ceases to
+    exist. This behaviour might change eventually.</para>
+
+    <para><function>sd_event_source_send_child_signal()</function> may be used to send a UNIX signal to the
+    watched process. If the pidfd concept is supported in the kernel, this is implemented via <citerefentry
+    project='man-pages'><refentrytitle>pidfd_send_signal</refentrytitle><manvolnum>2</manvolnum></citerefentry>
+    and otherwise via <citerefentry
+    project='man-pages'><refentrytitle>rt_sigqueueinfo</refentrytitle><manvolnum>2</manvolnum></citerefentry>
+    (or via <citerefentry
+    project='man-pages'><refentrytitle>kill</refentrytitle><manvolnum>2</manvolnum></citerefentry> in case
+    <parameter>info</parameter> is <constant>NULL</constant>). The specified parameters match those of these
+    underlying system calls, except that the <parameter>info</parameter> is never modified (and is thus
+    declared constant). Like for the underlying system calls, the <parameter>flags</parameter> parameter
+    currently must be zero.</para>
   </refsect1>
 
   <refsect1>
@@ -165,8 +265,8 @@
         <varlistentry>
           <term><constant>-EBUSY</constant></term>
 
-          <listitem><para>A handler is already installed for this
-          child process.</para></listitem>
+          <listitem><para>A handler is already installed for this child process, or
+          <constant>SIGCHLD</constant> is not blocked.</para></listitem>
 
         </varlistentry>
 
@@ -190,6 +290,12 @@
           <listitem><para>The passed event source is not a child process event source.</para></listitem>
         </varlistentry>
 
+        <varlistentry>
+          <term><constant>-EOPNOTSUPP</constant></term>
+
+          <listitem><para>A pidfd was requested but the kernel does not support this concept.</para></listitem>
+        </varlistentry>
+
       </variablelist>
     </refsect2>
   </refsect1>
@@ -214,7 +320,13 @@
       <citerefentry><refentrytitle>sd_event_source_set_userdata</refentrytitle><manvolnum>3</manvolnum></citerefentry>,
       <citerefentry><refentrytitle>sd_event_source_set_description</refentrytitle><manvolnum>3</manvolnum></citerefentry>,
       <citerefentry><refentrytitle>sd_event_source_set_floating</refentrytitle><manvolnum>3</manvolnum></citerefentry>,
-      <citerefentry project='man-pages'><refentrytitle>waitid</refentrytitle><manvolnum>2</manvolnum></citerefentry>
+      <citerefentry project='man-pages'><refentrytitle>waitid</refentrytitle><manvolnum>2</manvolnum></citerefentry>,
+      <citerefentry project='man-pages'><refentrytitle>sigprocmask</refentrytitle><manvolnum>2</manvolnum></citerefentry>,
+      <citerefentry project='man-pages'><refentrytitle>pthread_sigmask</refentrytitle><manvolnum>3</manvolnum></citerefentry>,
+      <citerefentry project='man-pages'><refentrytitle>pidfd_open</refentrytitle><manvolnum>2</manvolnum></citerefentry>,
+      <citerefentry project='man-pages'><refentrytitle>pidfd_send_signal</refentrytitle><manvolnum>2</manvolnum></citerefentry>,
+      <citerefentry project='man-pages'><refentrytitle>rt_sigqueueinfo</refentrytitle><manvolnum>2</manvolnum></citerefentry>,
+      <citerefentry project='man-pages'><refentrytitle>kill</refentrytitle><manvolnum>2</manvolnum></citerefentry>
     </para>
   </refsect1>
 

man/sd_event_add_signal.xml

@@ -75,14 +75,13 @@
     project='man-pages'><refentrytitle>signalfd</refentrytitle><manvolnum>2</manvolnum></citerefentry>
     for further information.</para>
 
-    <para>Only a single handler may be installed for a specific
-    signal. The signal will be unblocked by this call, and must be
-    blocked before this function is called in all threads (using
+    <para>Only a single handler may be installed for a specific signal. The signal must be blocked in all
+    threads before this function is called (using <citerefentry
+    project='man-pages'><refentrytitle>sigprocmask</refentrytitle><manvolnum>2</manvolnum></citerefentry> or
     <citerefentry
-    project='man-pages'><refentrytitle>sigprocmask</refentrytitle><manvolnum>2</manvolnum></citerefentry>). If
-    the handler is not specified (<parameter>handler</parameter> is
-    <constant>NULL</constant>), a default handler which causes the
-    program to exit cleanly will be used.</para>
+    project='man-pages'><refentrytitle>pthread_sigmask</refentrytitle><manvolnum>3</manvolnum></citerefentry>). If
+    the handler is not specified (<parameter>handler</parameter> is <constant>NULL</constant>), a default
+    handler which causes the program to exit cleanly will be used.</para>
 
     <para>By default, the event source is enabled permanently
     (<constant>SD_EVENT_ON</constant>), but this may be changed with
@@ -189,7 +188,9 @@
       <citerefentry><refentrytitle>sd_event_source_set_userdata</refentrytitle><manvolnum>3</manvolnum></citerefentry>,
       <citerefentry><refentrytitle>sd_event_source_set_floating</refentrytitle><manvolnum>3</manvolnum></citerefentry>,
       <citerefentry project='man-pages'><refentrytitle>signal</refentrytitle><manvolnum>7</manvolnum></citerefentry>,
-      <citerefentry project='man-pages'><refentrytitle>signalfd</refentrytitle><manvolnum>2</manvolnum></citerefentry>
+      <citerefentry project='man-pages'><refentrytitle>signalfd</refentrytitle><manvolnum>2</manvolnum></citerefentry>,
+      <citerefentry project='man-pages'><refentrytitle>sigprocmask</refentrytitle><manvolnum>2</manvolnum></citerefentry>,
+      <citerefentry project='man-pages'><refentrytitle>pthread_sigmask</refentrytitle><manvolnum>3</manvolnum></citerefentry>
     </para>
   </refsect1>
 

man/systemd.special.xml

@@ -38,6 +38,7 @@
     <filename>hibernate.target</filename>,
     <filename>hybrid-sleep.target</filename>,
     <filename>suspend-then-hibernate.target</filename>,
+    <filename>initrd.target</filename>,
     <filename>initrd-fs.target</filename>,
     <filename>initrd-root-device.target</filename>,
     <filename>initrd-root-fs.target</filename>,
@@ -202,14 +203,16 @@
         <varlistentry>
           <term><filename>default.target</filename></term>
           <listitem>
-            <para>The default unit systemd starts at bootup. Usually,
-            this should be aliased (symlinked) to
-            <filename>multi-user.target</filename> or
-            <filename>graphical.target</filename>.</para>
+            <para>The default unit systemd starts at bootup. Usually, this should be aliased (symlinked) to
+            <filename>multi-user.target</filename> or <filename>graphical.target</filename>. See
+            <citerefentry><refentrytitle>bootup</refentrytitle><manvolnum>7</manvolnum></citerefentry> for
+            more discussion.</para>
 
-            <para>The default unit systemd starts at bootup can be
-            overridden with the <varname>systemd.unit=</varname> kernel
-            command line option.</para>
+            <para>The default unit systemd starts at bootup can be overridden with the
+            <varname>systemd.unit=</varname> kernel command line option, or more conveniently, with the short
+            names like <varname>single</varname>, <varname>rescue</varname>, <varname>1</varname>,
+            <varname>3</varname>, <varname>5</varname>, …; see
+            <citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry>.</para>
           </listitem>
         </varlistentry>
         <varlistentry>
@@ -341,6 +344,15 @@
             is active as long as the system is running.</para>
           </listitem>
         </varlistentry>
+        <varlistentry>
+          <term><filename>initrd.target</filename></term>
+          <listitem>
+            <para>This is the default target in the initramfs, similar to <filename>default.target</filename>
+            in the main system. It is used to mount the real root and transition to it. See
+            <citerefentry><refentrytitle>bootup</refentrytitle><manvolnum>7</manvolnum></citerefentry> for
+            more discussion.</para>
+          </listitem>
+        </varlistentry>
         <varlistentry>
           <term><filename>initrd-fs.target</filename></term>
           <listitem>

meson.build

@@ -517,6 +517,18 @@ foreach ident : [
                                  #include <unistd.h>'''],
         ['get_mempolicy',     '''#include <stdlib.h>
                                  #include <unistd.h>'''],
+        ['pidfd_send_signal', '''#include <stdlib.h>
+                                 #include <unistd.h>
+                                 #include <signal.h>
+                                 #include <sys/wait.h>'''],
+        ['pidfd_open',        '''#include <stdlib.h>
+                                 #include <unistd.h>
+                                 #include <signal.h>
+                                 #include <sys/wait.h>'''],
+        ['rt_sigqueueinfo',   '''#include <stdlib.h>
+                                 #include <unistd.h>
+                                 #include <signal.h>
+                                 #include <sys/wait.h>'''],
 ]
 
         have = cc.has_function(ident[0], prefix : ident[1], args : '-D_GNU_SOURCE')

src/basic/capability-util.c

@@ -86,20 +86,17 @@ unsigned long cap_last_cap(void) {
 int capability_update_inherited_set(cap_t caps, uint64_t set) {
         unsigned long i;
 
-        /* Add capabilities in the set to the inherited caps. Do not apply
-         * them yet. */
+        /* Add capabilities in the set to the inherited caps, drops capabilities not in the set.
+         * Do not apply them yet. */
 
         for (i = 0; i <= cap_last_cap(); i++) {
+                cap_flag_value_t flag = set & (UINT64_C(1) << i) ? CAP_SET : CAP_CLEAR;
+                cap_value_t v;
 
-                if (set & (UINT64_C(1) << i)) {
-                        cap_value_t v;
+                v = (cap_value_t) i;
 
-                        v = (cap_value_t) i;
-
-                        /* Make the capability inheritable. */
-                        if (cap_set_flag(caps, CAP_INHERITABLE, 1, &v, CAP_SET) < 0)
-                                return -errno;
-                }
+                if (cap_set_flag(caps, CAP_INHERITABLE, 1, &v, flag) < 0)
+                        return -errno;
         }
 
         return 0;
@@ -132,6 +129,17 @@ int capability_ambient_set_apply(uint64_t set, bool also_inherit) {
                         /* Add the capability to the ambient set. */
                         if (prctl(PR_CAP_AMBIENT, PR_CAP_AMBIENT_RAISE, i, 0, 0) < 0)
                                 return -errno;
+                } else {
+
+                        /* Drop the capability so we don't inherit capabilities we didn't ask for. */
+                        r = prctl(PR_CAP_AMBIENT, PR_CAP_AMBIENT_IS_SET, i, 0, 0);
+                        if (r < 0)
+                                return -errno;
+
+                        if (r)
+                                if (prctl(PR_CAP_AMBIENT, PR_CAP_AMBIENT_LOWER, i, 0, 0) < 0)
+                                        return -errno;
+
                 }
         }
 

src/basic/copy.c

@@ -21,6 +21,7 @@
 #include "missing_syscall.h"
 #include "mountpoint-util.h"
 #include "nulstr-util.h"
+#include "selinux-util.h"
 #include "stat-util.h"
 #include "string-util.h"
 #include "strv.h"
@@ -372,7 +373,15 @@ static int fd_copy_symlink(
         if (r < 0)
                 return r;
 
-        if (symlinkat(target, dt, to) < 0)
+        if (copy_flags & COPY_MAC_CREATE) {
+                r = mac_selinux_create_file_prepare_at(dt, to, S_IFLNK);
+                if (r < 0)
+                        return r;
+        }
+        r = symlinkat(target, dt, to);
+        if (copy_flags & COPY_MAC_CREATE)
+                mac_selinux_create_file_clear();
+        if (r < 0)
                 return -errno;
 
         if (fchownat(dt, to,
@@ -408,7 +417,14 @@ static int fd_copy_regular(
         if (fdf < 0)
                 return -errno;
 
+        if (copy_flags & COPY_MAC_CREATE) {
+                r = mac_selinux_create_file_prepare_at(dt, to, S_IFREG);
+                if (r < 0)
+                        return r;
+        }
         fdt = openat(dt, to, O_WRONLY|O_CREAT|O_EXCL|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW, st->st_mode & 07777);
+        if (copy_flags & COPY_MAC_CREATE)
+                mac_selinux_create_file_clear();
         if (fdt < 0)
                 return -errno;
 
@@ -457,7 +473,14 @@ static int fd_copy_fifo(
         assert(st);
         assert(to);
 
+        if (copy_flags & COPY_MAC_CREATE) {
+                r = mac_selinux_create_file_prepare_at(dt, to, S_IFIFO);
+                if (r < 0)
+                        return r;
+        }
         r = mkfifoat(dt, to, st->st_mode & 07777);
+        if (copy_flags & COPY_MAC_CREATE)
+                mac_selinux_create_file_clear();
         if (r < 0)
                 return -errno;
 
@@ -488,7 +511,14 @@ static int fd_copy_node(
         assert(st);
         assert(to);
 
+        if (copy_flags & COPY_MAC_CREATE) {
+                r = mac_selinux_create_file_prepare_at(dt, to, st->st_mode & S_IFMT);
+                if (r < 0)
+                        return r;
+        }
         r = mknodat(dt, to, st->st_mode, st->st_rdev);
+        if (copy_flags & COPY_MAC_CREATE)
+                mac_selinux_create_file_clear();
         if (r < 0)
                 return -errno;
 
@@ -556,7 +586,10 @@ static int fd_copy_directory(
         if (exists)
                 created = false;
         else {
-                r = mkdirat(dt, to, st->st_mode & 07777);
+                if (copy_flags & COPY_MAC_CREATE)
+                        r = mkdirat_label(dt, to, st->st_mode & 07777);
+                else
+                        r = mkdirat(dt, to, st->st_mode & 07777);
                 if (r >= 0)
                         created = true;
                 else if (errno == EEXIST && (copy_flags & COPY_MERGE))
@@ -796,7 +829,14 @@ int copy_file_full(
         assert(to);
 
         RUN_WITH_UMASK(0000) {
+                if (copy_flags & COPY_MAC_CREATE) {
+                        r = mac_selinux_create_file_prepare(to, S_IFREG);
+                        if (r < 0)
+                                return r;
+                }
                 fdt = open(to, flags|O_WRONLY|O_CREAT|O_CLOEXEC|O_NOCTTY, mode);
+                if (copy_flags & COPY_MAC_CREATE)
+                        mac_selinux_create_file_clear();
                 if (fdt < 0)
                         return -errno;
         }
@@ -850,13 +890,29 @@ int copy_file_atomic_full(
                 if (r < 0)
                         return r;
 
+                if (copy_flags & COPY_MAC_CREATE) {
+                        r = mac_selinux_create_file_prepare(to, S_IFREG);
+                        if (r < 0) {
+                                t = mfree(t);
+                                return r;
+                        }
+                }
                 fdt = open(t, O_CREAT|O_EXCL|O_NOFOLLOW|O_NOCTTY|O_WRONLY|O_CLOEXEC, 0600);
+                if (copy_flags & COPY_MAC_CREATE)
+                        mac_selinux_create_file_clear();
                 if (fdt < 0) {
                         t = mfree(t);
                         return -errno;
                 }
         } else {
+                if (copy_flags & COPY_MAC_CREATE) {
+                        r = mac_selinux_create_file_prepare(to, S_IFREG);
+                        if (r < 0)
+                                return r;
+                }
                 fdt = open_tmpfile_linkable(to, O_WRONLY|O_CLOEXEC, &t);
+                if (copy_flags & COPY_MAC_CREATE)
+                        mac_selinux_create_file_clear();
                 if (fdt < 0)
                         return fdt;
         }

src/basic/copy.h

@@ -16,6 +16,7 @@ typedef enum CopyFlags {
         COPY_MERGE_EMPTY = 1 << 4, /* Merge an existing, empty directory with our new tree to copy */
         COPY_CRTIME      = 1 << 5, /* Generate a user.crtime_usec xattr off the source crtime if there is one, on copying */
         COPY_SIGINT      = 1 << 6, /* Check for SIGINT regularly and return EINTR if seen (caller needs to block SIGINT) */
+        COPY_MAC_CREATE  = 1 << 7, /* Create files with the correct MAC label (currently SELinux only) */
 } CopyFlags;
 
 typedef int (*copy_progress_bytes_t)(uint64_t n_bytes, void *userdata);

src/basic/errno-util.h

@@ -101,3 +101,11 @@ static inline bool ERRNO_IS_PRIVILEGE(int r) {
                       EACCES,
                       EPERM);
 }
+
+/* Three difference errors for "not enough disk space" */
+static inline bool ERRNO_IS_DISK_SPACE(int r) {
+        return IN_SET(abs(r),
+                      ENOSPC,
+                      EDQUOT,
+                      EFBIG);
+}

src/basic/memory-util.h

@@ -80,14 +80,21 @@ static inline void* explicit_bzero_safe(void *p, size_t l) {
 void *explicit_bzero_safe(void *p, size_t l);
 #endif
 
-static inline void erase_and_freep(void *p) {
-        void *ptr = *(void**) p;
+static inline void* erase_and_free(void *p) {
+        size_t l;
 
-        if (ptr) {
-                size_t l = malloc_usable_size(ptr);
-                explicit_bzero_safe(ptr, l);
-                free(ptr);
-        }
+        if (!p)
+                return NULL;
+
+        l = malloc_usable_size(p);
+        explicit_bzero_safe(p, l);
+        free(p);
+
+        return NULL;
+}
+
+static inline void erase_and_freep(void *p) {
+        erase_and_free(*(void**) p);
 }
 
 /* Use with _cleanup_ to erase a single 'char' when leaving scope */

src/basic/missing_syscall.h

@@ -5,8 +5,10 @@
 
 #include <errno.h>
 #include <fcntl.h>
+#include <signal.h>
 #include <sys/syscall.h>
 #include <sys/types.h>
+#include <sys/wait.h>
 #include <unistd.h>
 
 #ifdef ARCH_MIPS
@@ -524,3 +526,45 @@ static inline long missing_get_mempolicy(int *mode, unsigned long *nodemask,
 
 #define get_mempolicy missing_get_mempolicy
 #endif
+
+#if !HAVE_PIDFD_OPEN
+/* may be (invalid) negative number due to libseccomp, see PR 13319 */
+#  if ! (defined __NR_pidfd_open && __NR_pidfd_open > 0)
+#    if defined __NR_pidfd_open
+#      undef __NR_pidfd_open
+#    endif
+#    define __NR_pidfd_open 434
+#endif
+static inline int pidfd_open(pid_t pid, unsigned flags) {
+#ifdef __NR_pidfd_open
+        return syscall(__NR_pidfd_open, pid, flags);
+#else
+        errno = ENOSYS;
+        return -1;
+#endif
+}
+#endif
+
+#if !HAVE_PIDFD_SEND_SIGNAL
+/* may be (invalid) negative number due to libseccomp, see PR 13319 */
+#  if ! (defined __NR_pidfd_send_signal && __NR_pidfd_send_signal > 0)
+#    if defined __NR_pidfd_send_signal
+#      undef __NR_pidfd_send_signal
+#    endif
+#    define __NR_pidfd_send_signal 424
+#endif
+static inline int pidfd_send_signal(int fd, int sig, siginfo_t *info, unsigned flags) {
+#ifdef __NR_pidfd_open
+        return syscall(__NR_pidfd_send_signal, fd, sig, info, flags);
+#else
+        errno = ENOSYS;
+        return -1;
+#endif
+}
+#endif
+
+#if !HAVE_RT_SIGQUEUEINFO
+static inline int rt_sigqueueinfo(pid_t tgid, int sig, siginfo_t *info) {
+        return syscall(__NR_rt_sigqueueinfo, tgid, sig, info);
+}
+#endif

src/basic/ordered-set.h

@@ -50,6 +50,10 @@ static inline void* ordered_set_remove(OrderedSet *s, void *p) {
         return ordered_hashmap_remove((OrderedHashmap*) s, p);
 }
 
+static inline void* ordered_set_first(OrderedSet *s) {
+        return ordered_hashmap_first((OrderedHashmap*) s);
+}
+
 static inline void* ordered_set_steal_first(OrderedSet *s) {
         return ordered_hashmap_steal_first((OrderedHashmap*) s);
 }

src/basic/parse-util.c

@@ -365,7 +365,6 @@ int safe_atou_full(const char *s, unsigned base, unsigned *ret_u) {
         unsigned long l;
 
         assert(s);
-        assert(ret_u);
         assert(base <= 16);
 
         /* strtoul() is happy to parse negative values, and silently
@@ -389,7 +388,9 @@ int safe_atou_full(const char *s, unsigned base, unsigned *ret_u) {
         if ((unsigned long) (unsigned) l != l)
                 return -ERANGE;
 
-        *ret_u = (unsigned) l;
+        if (ret_u)
+                *ret_u = (unsigned) l;
+
         return 0;
 }
 
@@ -398,7 +399,6 @@ int safe_atoi(const char *s, int *ret_i) {
         long l;
 
         assert(s);
-        assert(ret_i);
 
         errno = 0;
         l = strtol(s, &x, 0);
@@ -409,7 +409,9 @@ int safe_atoi(const char *s, int *ret_i) {
         if ((long) (int) l != l)
                 return -ERANGE;
 
-        *ret_i = (int) l;
+        if (ret_i)
+                *ret_i = (int) l;
+
         return 0;
 }
 
@@ -418,7 +420,6 @@ int safe_atollu(const char *s, long long unsigned *ret_llu) {
         unsigned long long l;
 
         assert(s);
-        assert(ret_llu);
 
         s += strspn(s, WHITESPACE);
 
@@ -431,7 +432,9 @@ int safe_atollu(const char *s, long long unsigned *ret_llu) {
         if (*s == '-')
                 return -ERANGE;
 
-        *ret_llu = l;
+        if (ret_llu)
+                *ret_llu = l;
+
         return 0;
 }
 
@@ -440,7 +443,6 @@ int safe_atolli(const char *s, long long int *ret_lli) {
         long long l;
 
         assert(s);
-        assert(ret_lli);
 
         errno = 0;
         l = strtoll(s, &x, 0);
@@ -449,7 +451,9 @@ int safe_atolli(const char *s, long long int *ret_lli) {
         if (!x || x == s || *x != 0)
                 return -EINVAL;
 
-        *ret_lli = l;
+        if (ret_lli)
+                *ret_lli = l;
+
         return 0;
 }
 
@@ -458,7 +462,6 @@ int safe_atou8(const char *s, uint8_t *ret) {
         unsigned long l;
 
         assert(s);
-        assert(ret);
 
         s += strspn(s, WHITESPACE);
 
@@ -473,7 +476,8 @@ int safe_atou8(const char *s, uint8_t *ret) {
         if ((unsigned long) (uint8_t) l != l)
                 return -ERANGE;
 
-        *ret = (uint8_t) l;
+        if (ret)
+                *ret = (uint8_t) l;
         return 0;
 }
 
@@ -507,7 +511,6 @@ int safe_atoi16(const char *s, int16_t *ret) {
         long l;
 
         assert(s);
-        assert(ret);
 
         errno = 0;
         l = strtol(s, &x, 0);
@@ -518,7 +521,9 @@ int safe_atoi16(const char *s, int16_t *ret) {
         if ((long) (int16_t) l != l)
                 return -ERANGE;
 
-        *ret = (int16_t) l;
+        if (ret)
+                *ret = (int16_t) l;
+
         return 0;
 }
 
@@ -528,7 +533,6 @@ int safe_atod(const char *s, double *ret_d) {
         double d = 0;
 
         assert(s);
-        assert(ret_d);
 
         loc = newlocale(LC_NUMERIC_MASK, "C", (locale_t) 0);
         if (loc == (locale_t) 0)
@@ -541,7 +545,9 @@ int safe_atod(const char *s, double *ret_d) {
         if (!x || x == s || *x != 0)
                 return -EINVAL;
 
-        *ret_d = (double) d;
+        if (ret_d)
+                *ret_d = (double) d;
+
         return 0;
 }
 

src/basic/process-util.c

@@ -40,6 +40,7 @@
 #include "rlimit-util.h"
 #include "signal-util.h"
 #include "stat-util.h"
+#include "stdio-util.h"
 #include "string-table.h"
 #include "string-util.h"
 #include "terminal-util.h"
@@ -1337,6 +1338,13 @@ int safe_fork_full(
                         log_full_errno(prio, r, "Failed to connect stdin/stdout to /dev/null: %m");
                         _exit(EXIT_FAILURE);
                 }
+
+        } else if (flags & FORK_STDOUT_TO_STDERR) {
+
+                if (dup2(STDERR_FILENO, STDOUT_FILENO) < 0) {
+                        log_full_errno(prio, r, "Failed to connect stdout to stderr: %m");
+                        _exit(EXIT_FAILURE);
+                }
         }
 
         if (flags & FORK_RLIMIT_NOFILE_SAFE) {
@@ -1488,6 +1496,38 @@ int set_oom_score_adjust(int value) {
                                  WRITE_STRING_FILE_VERIFY_ON_FAILURE|WRITE_STRING_FILE_DISABLE_BUFFER);
 }
 
+int pidfd_get_pid(int fd, pid_t *ret) {
+        char path[STRLEN("/proc/self/fdinfo/") + DECIMAL_STR_MAX(int)];
+        _cleanup_free_ char *fdinfo = NULL;
+        char *p;
+        int r;
+
+        if (fd < 0)
+                return -EBADF;
+
+        xsprintf(path, "/proc/self/fdinfo/%i", fd);
+
+        r = read_full_file(path, &fdinfo, NULL);
+        if (r == -ENOENT) /* if fdinfo doesn't exist we assume the process does not exist */
+                return -ESRCH;
+        if (r < 0)
+                return r;
+
+        p = startswith(fdinfo, "Pid:");
+        if (!p) {
+                p = strstr(fdinfo, "\nPid:");
+                if (!p)
+                        return -ENOTTY; /* not a pidfd? */
+
+                p += 5;
+        }
+
+        p += strspn(p, WHITESPACE);
+        p[strcspn(p, WHITESPACE)] = 0;
+
+        return parse_pid(p, ret);
+}
+
 static const char *const ioprio_class_table[] = {
         [IOPRIO_CLASS_NONE] = "none",
         [IOPRIO_CLASS_RT] = "realtime",

src/basic/process-util.h

@@ -147,16 +147,17 @@ void reset_cached_pid(void);
 int must_be_root(void);
 
 typedef enum ForkFlags {
-        FORK_RESET_SIGNALS      = 1 << 0, /* Reset all signal handlers and signal mask */
-        FORK_CLOSE_ALL_FDS      = 1 << 1, /* Close all open file descriptors in the child, except for 0,1,2 */
-        FORK_DEATHSIG           = 1 << 2, /* Set PR_DEATHSIG in the child */
-        FORK_NULL_STDIO         = 1 << 3, /* Connect 0,1,2 to /dev/null */
-        FORK_REOPEN_LOG         = 1 << 4, /* Reopen log connection */
-        FORK_LOG                = 1 << 5, /* Log above LOG_DEBUG log level about failures */
-        FORK_WAIT               = 1 << 6, /* Wait until child exited */
-        FORK_NEW_MOUNTNS        = 1 << 7, /* Run child in its own mount namespace */
-        FORK_MOUNTNS_SLAVE      = 1 << 8, /* Make child's mount namespace MS_SLAVE */
-        FORK_RLIMIT_NOFILE_SAFE = 1 << 9, /* Set RLIMIT_NOFILE soft limit to 1K for select() compat */
+        FORK_RESET_SIGNALS      = 1 <<  0, /* Reset all signal handlers and signal mask */
+        FORK_CLOSE_ALL_FDS      = 1 <<  1, /* Close all open file descriptors in the child, except for 0,1,2 */
+        FORK_DEATHSIG           = 1 <<  2, /* Set PR_DEATHSIG in the child */
+        FORK_NULL_STDIO         = 1 <<  3, /* Connect 0,1,2 to /dev/null */
+        FORK_REOPEN_LOG         = 1 <<  4, /* Reopen log connection */
+        FORK_LOG                = 1 <<  5, /* Log above LOG_DEBUG log level about failures */
+        FORK_WAIT               = 1 <<  6, /* Wait until child exited */
+        FORK_NEW_MOUNTNS        = 1 <<  7, /* Run child in its own mount namespace */
+        FORK_MOUNTNS_SLAVE      = 1 <<  8, /* Make child's mount namespace MS_SLAVE */
+        FORK_RLIMIT_NOFILE_SAFE = 1 <<  9, /* Set RLIMIT_NOFILE soft limit to 1K for select() compat */
+        FORK_STDOUT_TO_STDERR   = 1 << 10, /* Make stdout a copy of stderr */
 } ForkFlags;
 
 int safe_fork_full(const char *name, const int except_fds[], size_t n_except_fds, ForkFlags flags, pid_t *ret_pid);
@@ -197,3 +198,5 @@ assert_cc(TASKS_MAX <= (unsigned long) PID_T_MAX);
                 (pid) = 0;                      \
                 _pid_;                          \
         })
+
+int pidfd_get_pid(int fd, pid_t *ret);

src/basic/selinux-util.c

@@ -107,6 +107,26 @@ void mac_selinux_finish(void) {
 #endif
 }
 
+void mac_selinux_reload(void) {
+
+#if HAVE_SELINUX
+        struct selabel_handle *backup_label_hnd;
+
+        if (!label_hnd)
+                return;
+
+        backup_label_hnd = TAKE_PTR(label_hnd);
+
+        /* try to initialize new handle
+         *    on success close backup
+         *    on failure restore backup */
+        if (mac_selinux_init() == 0)
+                selabel_close(backup_label_hnd);
+        else
+                label_hnd = backup_label_hnd;
+#endif
+}
+
 int mac_selinux_fix(const char *path, LabelFixFlags flags) {
 
 #if HAVE_SELINUX

src/basic/selinux-util.h

@@ -13,6 +13,7 @@ void mac_selinux_retest(void);
 
 int mac_selinux_init(void);
 void mac_selinux_finish(void);
+void mac_selinux_reload(void);
 
 int mac_selinux_fix(const char *path, LabelFixFlags flags);
 int mac_selinux_apply(const char *path, const char *label);

src/basic/signal-util.c

@@ -287,3 +287,18 @@ int signal_from_string(const char *s) {
 void nop_signal_handler(int sig) {
         /* nothing here */
 }
+
+int signal_is_blocked(int sig) {
+        sigset_t ss;
+        int r;
+
+        r = pthread_sigmask(SIG_SETMASK, NULL, &ss);
+        if (r != 0)
+                return -r;
+
+        r = sigismember(&ss, sig);
+        if (r < 0)
+                return -errno;
+
+        return r;
+}

src/basic/signal-util.h

@@ -41,3 +41,5 @@ static inline const char* signal_to_string_with_check(int n) {
 
         return signal_to_string(n);
 }
+
+int signal_is_blocked(int sig);

src/basic/special.h

@@ -2,6 +2,7 @@
 #pragma once
 
 #define SPECIAL_DEFAULT_TARGET "default.target"
+#define SPECIAL_INITRD_TARGET "initrd.target"
 
 /* Shutdown targets */
 #define SPECIAL_UMOUNT_TARGET "umount.target"

src/basic/string-util.c

@@ -1064,3 +1064,13 @@ bool string_is_safe(const char *p) {
 
         return true;
 }
+
+char* string_erase(char *x) {
+        if (!x)
+                return NULL;
+
+        /* A delicious drop of snake-oil! To be called on memory where we stored passphrases or so, after we
+         * used them. */
+        explicit_bzero_safe(x, strlen(x));
+        return x;
+}

src/basic/string-util.h

@@ -278,3 +278,5 @@ static inline char* str_realloc(char **p) {
 
         return (*p = t);
 }
+
+char* string_erase(char *x);

src/basic/tmpfile-util.c

@@ -19,50 +19,60 @@
 #include "tmpfile-util.h"
 #include "umask-util.h"
 
-int fopen_temporary(const char *path, FILE **_f, char **_temp_path) {
-        FILE *f;
-        char *t;
-        int r, fd;
+int fopen_temporary(const char *path, FILE **ret_f, char **ret_temp_path) {
+        _cleanup_fclose_ FILE *f = NULL;
+        _cleanup_free_ char *t = NULL;
+        _cleanup_close_ int fd = -1;
+        int r;
 
-        assert(path);
-        assert(_f);
-        assert(_temp_path);
+        if (path) {
+                r = tempfn_xxxxxx(path, NULL, &t);
+                if (r < 0)
+                        return r;
+        } else {
+                const char *d;
 
-        r = tempfn_xxxxxx(path, NULL, &t);
-        if (r < 0)
-                return r;
+                r = tmp_dir(&d);
+                if (r < 0)
+                        return r;
+
+                t = path_join(d, "XXXXXX");
+                if (!t)
+                        return -ENOMEM;
+        }
 
         fd = mkostemp_safe(t);
-        if (fd < 0) {
-                free(t);
+        if (fd < 0)
                 return -errno;
-        }
 
         /* This assumes that returned FILE object is short-lived and used within the same single-threaded
          * context and never shared externally, hence locking is not necessary. */
 
         r = fdopen_unlocked(fd, "w", &f);
         if (r < 0) {
-                unlink(t);
-                free(t);
-                safe_close(fd);
+                (void) unlink(t);
                 return r;
         }
 
-        *_f = f;
-        *_temp_path = t;
+        TAKE_FD(fd);
+
+        if (ret_f)
+                *ret_f = TAKE_PTR(f);
+
+        if (ret_temp_path)
+                *ret_temp_path = TAKE_PTR(t);
 
         return 0;
 }
 
 /* This is much like mkostemp() but is subject to umask(). */
 int mkostemp_safe(char *pattern) {
-        _unused_ _cleanup_umask_ mode_t u = umask(0077);
         int fd;
 
         assert(pattern);
 
-        fd = mkostemp(pattern, O_CLOEXEC);
+        RUN_WITH_UMASK(0077)
+                fd = mkostemp(pattern, O_CLOEXEC);
         if (fd < 0)
                 return -errno;
 

src/basic/user-util.c

@@ -84,7 +84,7 @@ char *getusername_malloc(void) {
         return uid_to_name(getuid());
 }
 
-static bool is_nologin_shell(const char *shell) {
+bool is_nologin_shell(const char *shell) {
 
         return PATH_IN_SET(shell,
                            /* 'nologin' is the friendliest way to disable logins for a user account. It prints a nice

src/basic/user-util.h

@@ -57,6 +57,14 @@ int take_etc_passwd_lock(const char *root);
 
 #define ETC_PASSWD_LOCK_PATH "/etc/.pwd.lock"
 
+static inline bool uid_is_system(uid_t uid) {
+        return uid <= SYSTEM_UID_MAX;
+}
+
+static inline bool gid_is_system(gid_t gid) {
+        return gid <= SYSTEM_GID_MAX;
+}
+
 static inline bool uid_is_dynamic(uid_t uid) {
         return DYNAMIC_UID_MIN <= uid && uid <= DYNAMIC_UID_MAX;
 }
@@ -65,12 +73,12 @@ static inline bool gid_is_dynamic(gid_t gid) {
         return uid_is_dynamic((uid_t) gid);
 }
 
-static inline bool uid_is_system(uid_t uid) {
-        return uid <= SYSTEM_UID_MAX;
+static inline bool uid_is_container(uid_t uid) {
+        return CONTAINER_UID_BASE_MIN <= uid && uid <= CONTAINER_UID_BASE_MAX;
 }
 
-static inline bool gid_is_system(gid_t gid) {
-        return gid <= SYSTEM_GID_MAX;
+static inline bool gid_is_container(gid_t gid) {
+        return uid_is_container((uid_t) gid);
 }
 
 /* The following macros add 1 when converting things, since UID 0 is a valid UID, while the pointer
@@ -127,3 +135,5 @@ int putsgent_sane(const struct sgrp *sg, FILE *stream);
 #endif
 
 int make_salt(char **ret);
+
+bool is_nologin_shell(const char *shell);

src/boot/efi/boot.c

@@ -32,7 +32,7 @@ enum loader_type {
 };
 
 typedef struct {
-        CHAR16 *id; /* The identifier for this entry (note that this id is not necessarily unique though!) */
+        CHAR16 *id; /* The unique identifier for this entry */
         CHAR16 *title_show;
         CHAR16 *title;
         CHAR16 *version;
@@ -1310,7 +1310,6 @@ static VOID config_entry_add_from_file(
         CHAR8 *line;
         UINTN pos = 0;
         CHAR8 *key, *value;
-        UINTN len;
         EFI_STATUS err;
         EFI_FILE_HANDLE handle;
         _cleanup_freepool_ CHAR16 *initrd = NULL;
@@ -1431,10 +1430,6 @@ static VOID config_entry_add_from_file(
 
         entry->device = device;
         entry->id = StrDuplicate(file);
-        len = StrLen(entry->id);
-        /* remove ".conf" */
-        if (len > 5)
-                entry->id[len - 5] = '\0';
         StrLwr(entry->id);
 
         config_add_entry(config, entry);
@@ -1775,7 +1770,8 @@ static ConfigEntry *config_entry_add_loader(
                 CHAR16 *id,
                 CHAR16 key,
                 CHAR16 *title,
-                CHAR16 *loader) {
+                CHAR16 *loader,
+                CHAR16 *version) {
 
         ConfigEntry *entry;
 
@@ -1783,6 +1779,7 @@ static ConfigEntry *config_entry_add_loader(
         *entry = (ConfigEntry) {
                 .type = type,
                 .title = StrDuplicate(title),
+                .version = StrDuplicate(version),
                 .device = device,
                 .loader = StrDuplicate(loader),
                 .id = StrDuplicate(id),
@@ -1840,7 +1837,7 @@ static BOOLEAN config_entry_add_loader_auto(
                 return FALSE;
         uefi_call_wrapper(handle->Close, 1, handle);
 
-        entry = config_entry_add_loader(config, device, LOADER_UNDEFINED, id, key, title, loader);
+        entry = config_entry_add_loader(config, device, LOADER_UNDEFINED, id, key, title, loader, NULL);
         if (!entry)
                 return FALSE;
 
@@ -1908,10 +1905,12 @@ static VOID config_entry_add_linux(
                 CHAR8 *line;
                 UINTN pos = 0;
                 CHAR8 *key, *value;
+                CHAR16 *os_name_pretty = NULL;
                 CHAR16 *os_name = NULL;
                 CHAR16 *os_id = NULL;
                 CHAR16 *os_version = NULL;
-                CHAR16 *os_build = NULL;
+                CHAR16 *os_version_id = NULL;
+                CHAR16 *os_build_id = NULL;
 
                 err = uefi_call_wrapper(linux_dir->Read, 3, linux_dir, &bufsize, buf);
                 if (bufsize == 0 || EFI_ERROR(err))
@@ -1927,6 +1926,8 @@ static VOID config_entry_add_linux(
                         continue;
                 if (StriCmp(f->FileName + len - 4, L".efi") != 0)
                         continue;
+                if (StrnCmp(f->FileName, L"auto-", 5) == 0)
+                        continue;
 
                 /* look for .osrel and .cmdline sections in the .efi binary */
                 err = pe_file_locate_sections(linux_dir, f->FileName, sections, addrs, offs, szs);
@@ -1940,6 +1941,12 @@ static VOID config_entry_add_linux(
                 /* read properties from the embedded os-release file */
                 while ((line = line_get_key_value(content, (CHAR8 *)"=", &pos, &key, &value))) {
                         if (strcmpa((CHAR8 *)"PRETTY_NAME", key) == 0) {
+                                FreePool(os_name_pretty);
+                                os_name_pretty = stra_to_str(value);
+                                continue;
+                        }
+
+                        if (strcmpa((CHAR8 *)"NAME", key) == 0) {
                                 FreePool(os_name);
                                 os_name = stra_to_str(value);
                                 continue;
@@ -1957,20 +1964,27 @@ static VOID config_entry_add_linux(
                                 continue;
                         }
 
+                        if (strcmpa((CHAR8 *)"VERSION_ID", key) == 0) {
+                                FreePool(os_version_id);
+                                os_version_id = stra_to_str(value);
+                                continue;
+                        }
+
                         if (strcmpa((CHAR8 *)"BUILD_ID", key) == 0) {
-                                FreePool(os_build);
-                                os_build = stra_to_str(value);
+                                FreePool(os_build_id);
+                                os_build_id = stra_to_str(value);
                                 continue;
                         }
                 }
 
-                if (os_name && os_id && (os_version || os_build)) {
-                        _cleanup_freepool_ CHAR16 *conf = NULL, *path = NULL;
+                if ((os_name_pretty || os_name) && os_id && (os_version || os_version_id || os_build_id)) {
+                        _cleanup_freepool_ CHAR16 *path = NULL;
 
-                        conf = PoolPrint(L"%s-%s", os_id, os_version ? : os_build);
                         path = PoolPrint(L"\\EFI\\Linux\\%s", f->FileName);
 
-                        entry = config_entry_add_loader(config, device, LOADER_LINUX, conf, 'l', os_name, path);
+                        entry = config_entry_add_loader(config, device, LOADER_LINUX, f->FileName, 'l',
+                                                        os_name_pretty ? : (os_name ? : os_id), path,
+                                                        os_version ? : (os_version_id ? : os_build_id));
 
                         FreePool(content);
                         content = NULL;
@@ -1989,10 +2003,12 @@ static VOID config_entry_add_linux(
                         config_entry_parse_tries(entry, L"\\EFI\\Linux", f->FileName, L".efi");
                 }
 
+                FreePool(os_name_pretty);
                 FreePool(os_name);
                 FreePool(os_id);
                 FreePool(os_version);
-                FreePool(os_build);
+                FreePool(os_version_id);
+                FreePool(os_build_id);
                 FreePool(content);
         }
 
@@ -2453,6 +2469,12 @@ EFI_STATUS efi_main(EFI_HANDLE image, EFI_SYSTEM_TABLE *sys_table) {
                 UINT64 key;
 
                 err = console_key_read(&key, FALSE);
+
+                if (err == EFI_NOT_READY) {
+                        uefi_call_wrapper(BS->Stall, 1, 100 * 1000);
+                        err = console_key_read(&key, FALSE);
+                }
+
                 if (!EFI_ERROR(err)) {
                         INT16 idx;
 

src/core/execute.c

@@ -3595,8 +3595,7 @@ static int exec_child(
 
                 /* This is done before enforce_user, but ambient set
                  * does not survive over setresuid() if keep_caps is not set. */
-                if (!needs_ambient_hack &&
-                    context->capability_ambient_set != 0) {
+                if (!needs_ambient_hack) {
                         r = capability_ambient_set_apply(context->capability_ambient_set, true);
                         if (r < 0) {
                                 *exit_status = EXIT_CAPABILITIES;

src/core/main.c

@@ -475,6 +475,15 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
                 if (arg_default_timeout_start_usec <= 0)
                         arg_default_timeout_start_usec = USEC_INFINITY;
 
+        } else if (proc_cmdline_key_streq(key, "systemd.cpu_affinity")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
+                r = parse_cpu_set(value, &arg_cpu_affinity);
+                if (r < 0)
+                        log_warning_errno(r, "Faile to parse CPU affinity mask '%s', ignoring: %m", value);
+
         } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
 
                 if (proc_cmdline_value_missing(key, value))
@@ -1738,6 +1747,8 @@ static int invoke_main_loop(
                         saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
                         saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
 
+                        mac_selinux_reload();
+
                         (void) parse_configuration(saved_rlimit_nofile, saved_rlimit_memlock);
 
                         set_manager_defaults(m);
@@ -1986,20 +1997,36 @@ static int do_queue_default_job(
                 const char **ret_error_message) {
 
         _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
+        const char* default_unit;
         Job *default_unit_job;
         Unit *target = NULL;
         int r;
 
-        log_debug("Activating default unit: %s", arg_default_unit);
+        if (arg_default_unit)
+                default_unit = arg_default_unit;
+        else if (in_initrd())
+                default_unit = SPECIAL_INITRD_TARGET;
+        else
+                default_unit = SPECIAL_DEFAULT_TARGET;
 
-        r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
+        log_debug("Activating default unit: %s", default_unit);
+
+        r = manager_load_startable_unit_or_warn(m, default_unit, NULL, &target);
+        if (r < 0 && in_initrd() && !arg_default_unit) {
+                /* Fall back to default.target, which we used to always use by default. Only do this if no
+                 * explicit configuration was given. */
+
+                log_info("Falling back to " SPECIAL_DEFAULT_TARGET ".");
+
+                r = manager_load_startable_unit_or_warn(m, SPECIAL_DEFAULT_TARGET, NULL, &target);
+        }
         if (r < 0) {
-                log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
+                log_info("Falling back to " SPECIAL_RESCUE_TARGET ".");
 
                 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
                 if (r < 0) {
-                        *ret_error_message = r == -ERFKILL ? "Rescue target masked"
-                                                           : "Failed to load rescue target";
+                        *ret_error_message = r == -ERFKILL ? SPECIAL_RESCUE_TARGET " masked"
+                                                           : "Failed to load " SPECIAL_RESCUE_TARGET;
                         return r;
                 }
         }
@@ -2211,15 +2238,6 @@ static int load_configuration(
                 return r;
         }
 
-        /* Initialize default unit */
-        if (!arg_default_unit) {
-                arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
-                if (!arg_default_unit) {
-                        *ret_error_message = "Failed to set default unit";
-                        return log_oom();
-                }
-        }
-
         /* Initialize the show status setting if it hasn't been set explicitly yet */
         if (arg_show_status == _SHOW_STATUS_INVALID)
                 arg_show_status = SHOW_STATUS_YES;

src/core/namespace.c

@@ -29,6 +29,7 @@
 #include "string-table.h"
 #include "string-util.h"
 #include "strv.h"
+#include "tmpfile-util.h"
 #include "umask-util.h"
 #include "user-util.h"
 
@@ -1640,6 +1641,44 @@ int temporary_filesystem_add(
         return 0;
 }
 
+static int make_tmp_prefix(const char *prefix) {
+        _cleanup_free_ char *t = NULL;
+        int r;
+
+        /* Don't do anything unless we know the dir is actually missing */
+        r = access(prefix, F_OK);
+        if (r >= 0)
+                return 0;
+        if (errno != ENOENT)
+                return -errno;
+
+        r = mkdir_parents(prefix, 0755);
+        if (r < 0)
+                return r;
+
+        r = tempfn_random(prefix, NULL, &t);
+        if (r < 0)
+                return r;
+
+        if (mkdir(t, 0777) < 0)
+                return -errno;
+
+        if (chmod(t, 01777) < 0) {
+                r = -errno;
+                (void) rmdir(t);
+                return r;
+        }
+
+        if (rename(t, prefix) < 0) {
+                r = -errno;
+                (void) rmdir(t);
+                return r == -EEXIST ? 0 : r; /* it's fine if someone else created the dir by now */
+        }
+
+        return 0;
+
+}
+
 static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) {
         _cleanup_free_ char *x = NULL;
         char bid[SD_ID128_STRING_MAX];
@@ -1661,6 +1700,10 @@ static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) {
         if (!x)
                 return -ENOMEM;
 
+        r = make_tmp_prefix(prefix);
+        if (r < 0)
+                return r;
+
         RUN_WITH_UMASK(0077)
                 if (!mkdtemp(x))
                         return -errno;

src/debug-generator/debug-generator.c

@@ -126,7 +126,15 @@ static int generate_wants_symlinks(void) {
 
         STRV_FOREACH(u, arg_wants) {
                 _cleanup_free_ char *p = NULL, *f = NULL;
-                const char *target = arg_default_unit ?: SPECIAL_DEFAULT_TARGET;
+                const char *target;
+
+                /* This should match what do_queue_default_job() in core/main.c does. */
+                if (arg_default_unit)
+                        target = arg_default_unit;
+                else if (in_initrd())
+                        target = SPECIAL_INITRD_TARGET;
+                else
+                        target = SPECIAL_DEFAULT_TARGET;
 
                 p = strjoin(arg_dest, "/", target, ".wants/", *u);
                 if (!p)

src/journal-remote/journal-upload.h

@@ -4,6 +4,7 @@
 
 #include "sd-event.h"
 #include "sd-journal.h"
+
 #include "time-util.h"
 
 typedef enum {

src/libsystemd-network/sd-dhcp-server.c

@@ -6,6 +6,7 @@
 #include <sys/ioctl.h>
 
 #include "sd-dhcp-server.h"
+#include "sd-id128.h"
 
 #include "alloc-util.h"
 #include "dhcp-internal.h"
@@ -13,7 +14,6 @@
 #include "fd-util.h"
 #include "in-addr-util.h"
 #include "io-util.h"
-#include "sd-id128.h"
 #include "siphash24.h"
 #include "string-util.h"
 #include "unaligned.h"

src/libsystemd/libsystemd.sym

@@ -682,3 +682,14 @@ global:
         sd_bus_object_vtable_format;
         sd_event_source_disable_unref;
 } LIBSYSTEMD_241;
+
+LIBSYSTEMD_245 {
+global:
+        sd_event_add_child_pidfd;
+        sd_event_source_get_child_pidfd;
+        sd_event_source_get_child_pidfd_own;
+        sd_event_source_set_child_pidfd_own;
+        sd_event_source_get_child_process_own;
+        sd_event_source_set_child_process_own;
+        sd_event_source_send_child_signal;
+} LIBSYSTEMD_243;

src/libsystemd/sd-bus/bus-common-errors.c

@@ -102,7 +102,6 @@ BUS_ERROR_MAP_ELF_REGISTER const sd_bus_error_map bus_common_errors[] = {
 
         SD_BUS_ERROR_MAP(BUS_ERROR_NO_PRODUCT_UUID,              EOPNOTSUPP),
 
-        SD_BUS_ERROR_MAP(BUS_ERROR_SPEED_METER_INACTIVE,         EOPNOTSUPP),
         SD_BUS_ERROR_MAP(BUS_ERROR_UNMANAGED_INTERFACE,          EOPNOTSUPP),
 
         SD_BUS_ERROR_MAP_END

src/libsystemd/sd-bus/bus-common-errors.h

@@ -81,7 +81,6 @@
 
 #define BUS_ERROR_NO_PRODUCT_UUID "org.freedesktop.hostname1.NoProductUUID"
 
-#define BUS_ERROR_SPEED_METER_INACTIVE "org.freedesktop.network1.SpeedMeterInactive"
 #define BUS_ERROR_UNMANAGED_INTERFACE "org.freedesktop.network1.UnmanagedInterface"
 
 BUS_ERROR_MAP_ELF_USE(bus_common_errors);

src/libsystemd/sd-event/event-source.h

@@ -34,7 +34,7 @@ typedef enum EventSourceType {
  * we know how to dispatch it */
 typedef enum WakeupType {
         WAKEUP_NONE,
-        WAKEUP_EVENT_SOURCE,
+        WAKEUP_EVENT_SOURCE, /* either I/O or pidfd wakeup */
         WAKEUP_CLOCK_DATA,
         WAKEUP_SIGNAL_DATA,
         WAKEUP_INOTIFY_DATA,
@@ -96,6 +96,12 @@ struct sd_event_source {
                         siginfo_t siginfo;
                         pid_t pid;
                         int options;
+                        int pidfd;
+                        bool registered:1; /* whether the pidfd is registered in the epoll */
+                        bool pidfd_owned:1; /* close pidfd when event source is freed */
+                        bool process_owned:1; /* kill+reap process when event source is freed */
+                        bool exited:1; /* true if process exited (i.e. if there's value in SIGKILLing it if we want to get rid of it) */
+                        bool waited:1; /* true if process was waited for (i.e. if there's value in waitid(P_PID)'ing it if we want to get rid of it) */
                 } child;
                 struct {
                         sd_event_handler_t callback;

src/libsystemd/sd-event/sd-event.c

@@ -9,6 +9,7 @@
 #include "sd-id128.h"
 
 #include "alloc-util.h"
+#include "env-util.h"
 #include "event-source.h"
 #include "fd-util.h"
 #include "fs-util.h"
@@ -28,6 +29,14 @@
 
 #define DEFAULT_ACCURACY_USEC (250 * USEC_PER_MSEC)
 
+static bool EVENT_SOURCE_WATCH_PIDFD(sd_event_source *s) {
+        /* Returns true if this is a PID event source and can be implemented by watching EPOLLIN */
+        return s &&
+                s->type == SOURCE_CHILD &&
+                s->child.pidfd >= 0 &&
+                s->child.options == WEXITED;
+}
+
 static const char* const event_source_type_table[_SOURCE_EVENT_SOURCE_TYPE_MAX] = {
         [SOURCE_IO] = "io",
         [SOURCE_TIME_REALTIME] = "realtime",
@@ -356,8 +365,6 @@ static bool event_pid_changed(sd_event *e) {
 }
 
 static void source_io_unregister(sd_event_source *s) {
-        int r;
-
         assert(s);
         assert(s->type == SOURCE_IO);
 
@@ -367,8 +374,7 @@ static void source_io_unregister(sd_event_source *s) {
         if (!s->io.registered)
                 return;
 
-        r = epoll_ctl(s->event->epoll_fd, EPOLL_CTL_DEL, s->io.fd, NULL);
-        if (r < 0)
+        if (epoll_ctl(s->event->epoll_fd, EPOLL_CTL_DEL, s->io.fd, NULL) < 0)
                 log_debug_errno(errno, "Failed to remove source %s (type %s) from epoll: %m",
                                 strna(s->description), event_source_type_to_string(s->type));
 
@@ -404,6 +410,51 @@ static int source_io_register(
         return 0;
 }
 
+static void source_child_pidfd_unregister(sd_event_source *s) {
+        assert(s);
+        assert(s->type == SOURCE_CHILD);
+
+        if (event_pid_changed(s->event))
+                return;
+
+        if (!s->child.registered)
+                return;
+
+        if (EVENT_SOURCE_WATCH_PIDFD(s))
+                if (epoll_ctl(s->event->epoll_fd, EPOLL_CTL_DEL, s->child.pidfd, NULL) < 0)
+                        log_debug_errno(errno, "Failed to remove source %s (type %s) from epoll: %m",
+                                        strna(s->description), event_source_type_to_string(s->type));
+
+        s->child.registered = false;
+}
+
+static int source_child_pidfd_register(sd_event_source *s, int enabled) {
+        int r;
+
+        assert(s);
+        assert(s->type == SOURCE_CHILD);
+        assert(enabled != SD_EVENT_OFF);
+
+        if (EVENT_SOURCE_WATCH_PIDFD(s)) {
+                struct epoll_event ev;
+
+                ev = (struct epoll_event) {
+                        .events = EPOLLIN | (enabled == SD_EVENT_ONESHOT ? EPOLLONESHOT : 0),
+                        .data.ptr = s,
+                };
+
+                if (s->child.registered)
+                        r = epoll_ctl(s->event->epoll_fd, EPOLL_CTL_MOD, s->child.pidfd, &ev);
+                else
+                        r = epoll_ctl(s->event->epoll_fd, EPOLL_CTL_ADD, s->child.pidfd, &ev);
+                if (r < 0)
+                        return -errno;
+        }
+
+        s->child.registered = true;
+        return 0;
+}
+
 static clockid_t event_source_type_to_clock(EventSourceType t) {
 
         switch (t) {
@@ -614,9 +665,8 @@ static void event_gc_signal_data(sd_event *e, const int64_t *priority, int sig)
 
         assert(e);
 
-        /* Rechecks if the specified signal is still something we are
-         * interested in. If not, we'll unmask it, and possibly drop
-         * the signalfd for it. */
+        /* Rechecks if the specified signal is still something we are interested in. If not, we'll unmask it,
+         * and possibly drop the signalfd for it. */
 
         if (sig == SIGCHLD &&
             e->n_enabled_child_sources > 0)
@@ -707,9 +757,13 @@ static void source_disconnect(sd_event_source *s) {
                         }
 
                         (void) hashmap_remove(s->event->child_sources, PID_TO_PTR(s->child.pid));
-                        event_gc_signal_data(s->event, &s->priority, SIGCHLD);
                 }
 
+                if (EVENT_SOURCE_WATCH_PIDFD(s))
+                        source_child_pidfd_unregister(s);
+                else
+                        event_gc_signal_data(s->event, &s->priority, SIGCHLD);
+
                 break;
 
         case SOURCE_DEFER:
@@ -790,6 +844,44 @@ static void source_free(sd_event_source *s) {
         if (s->type == SOURCE_IO && s->io.owned)
                 s->io.fd = safe_close(s->io.fd);
 
+        if (s->type == SOURCE_CHILD) {
+                /* Eventually the kernel will do this automatically for us, but for now let's emulate this (unreliably) in userspace. */
+
+                if (s->child.process_owned) {
+
+                        if (!s->child.exited) {
+                                bool sent = false;
+
+                                if (s->child.pidfd >= 0) {
+                                        if (pidfd_send_signal(s->child.pidfd, SIGKILL, NULL, 0) < 0) {
+                                                if (errno == ESRCH) /* Already dead */
+                                                        sent = true;
+                                                else if (!ERRNO_IS_NOT_SUPPORTED(errno))
+                                                        log_debug_errno(errno, "Failed to kill process " PID_FMT " via pidfd_send_signal(), re-trying via kill(): %m",
+                                                                        s->child.pid);
+                                        } else
+                                                sent = true;
+                                }
+
+                                if (!sent)
+                                        if (kill(s->child.pid, SIGKILL) < 0)
+                                                if (errno != ESRCH) /* Already dead */
+                                                        log_debug_errno(errno, "Failed to kill process " PID_FMT " via kill(), ignoring: %m",
+                                                                        s->child.pid);
+                        }
+
+                        if (!s->child.waited) {
+                                siginfo_t si = {};
+
+                                /* Reap the child if we can */
+                                (void) waitid(P_PID, s->child.pid, &si, WEXITED);
+                        }
+                }
+
+                if (s->child.pidfd_owned)
+                        s->child.pidfd = safe_close(s->child.pidfd);
+        }
+
         if (s->destroy_callback)
                 s->destroy_callback(s->userdata);
 
@@ -1073,7 +1165,6 @@ _public_ int sd_event_add_signal(
 
         _cleanup_(source_freep) sd_event_source *s = NULL;
         struct signal_data *d;
-        sigset_t ss;
         int r;
 
         assert_return(e, -EINVAL);
@@ -1085,11 +1176,10 @@ _public_ int sd_event_add_signal(
         if (!callback)
                 callback = signal_exit_callback;
 
-        r = pthread_sigmask(SIG_SETMASK, NULL, &ss);
-        if (r != 0)
-                return -r;
-
-        if (!sigismember(&ss, sig))
+        r = signal_is_blocked(sig);
+        if (r < 0)
+                return r;
+        if (r == 0)
                 return -EBUSY;
 
         if (!e->signal_sources) {
@@ -1124,6 +1214,11 @@ _public_ int sd_event_add_signal(
         return 0;
 }
 
+static bool shall_use_pidfd(void) {
+        /* Mostly relevant for debugging, i.e. this is used in test-event.c to test the event loop once with and once without pidfd */
+        return getenv_bool_secure("SYSTEMD_PIDFD") != 0;
+}
+
 _public_ int sd_event_add_child(
                 sd_event *e,
                 sd_event_source **ret,
@@ -1144,6 +1239,20 @@ _public_ int sd_event_add_child(
         assert_return(e->state != SD_EVENT_FINISHED, -ESTALE);
         assert_return(!event_pid_changed(e), -ECHILD);
 
+        if (e->n_enabled_child_sources == 0) {
+                /* Caller must block SIGCHLD before using us to watch children, even if pidfd is available,
+                 * for compatibility with pre-pidfd and because we don't want the reap the child processes
+                 * ourselves, i.e. call waitid(), and don't want Linux' default internal logic for that to
+                 * take effect.
+                 *
+                 * (As an optimization we only do this check on the first child event source created.) */
+                r = signal_is_blocked(SIGCHLD);
+                if (r < 0)
+                        return r;
+                if (r == 0)
+                        return -EBUSY;
+        }
+
         r = hashmap_ensure_allocated(&e->child_sources, NULL);
         if (r < 0)
                 return r;
@@ -1155,30 +1264,144 @@ _public_ int sd_event_add_child(
         if (!s)
                 return -ENOMEM;
 
+        s->wakeup = WAKEUP_EVENT_SOURCE;
         s->child.pid = pid;
         s->child.options = options;
         s->child.callback = callback;
         s->userdata = userdata;
         s->enabled = SD_EVENT_ONESHOT;
 
+        /* We always take a pidfd here if we can, even if we wait for anything else than WEXITED, so that we
+         * pin the PID, and make regular waitid() handling race-free. */
+
+        if (shall_use_pidfd()) {
+                s->child.pidfd = pidfd_open(s->child.pid, 0);
+                if (s->child.pidfd < 0) {
+                        /* Propagate errors unless the syscall is not supported or blocked */
+                        if (!ERRNO_IS_NOT_SUPPORTED(errno) && !ERRNO_IS_PRIVILEGE(errno))
+                                return -errno;
+                } else
+                        s->child.pidfd_owned = true; /* If we allocate the pidfd we own it by default */
+        } else
+                s->child.pidfd = -1;
+
+        r = hashmap_put(e->child_sources, PID_TO_PTR(pid), s);
+        if (r < 0)
+                return r;
+
+        e->n_enabled_child_sources++;
+
+        if (EVENT_SOURCE_WATCH_PIDFD(s)) {
+                /* We have a pidfd and we only want to watch for exit */
+
+                r = source_child_pidfd_register(s, s->enabled);
+                if (r < 0) {
+                        e->n_enabled_child_sources--;
+                        return r;
+                }
+        } else {
+                /* We have no pidfd or we shall wait for some other event than WEXITED */
+
+                r = event_make_signal_data(e, SIGCHLD, NULL);
+                if (r < 0) {
+                        e->n_enabled_child_sources--;
+                        return r;
+                }
+
+                e->need_process_child = true;
+        }
+
+        if (ret)
+                *ret = s;
+
+        TAKE_PTR(s);
+        return 0;
+}
+
+_public_ int sd_event_add_child_pidfd(
+                sd_event *e,
+                sd_event_source **ret,
+                int pidfd,
+                int options,
+                sd_event_child_handler_t callback,
+                void *userdata) {
+
+
+        _cleanup_(source_freep) sd_event_source *s = NULL;
+        pid_t pid;
+        int r;
+
+        assert_return(e, -EINVAL);
+        assert_return(e = event_resolve(e), -ENOPKG);
+        assert_return(pidfd >= 0, -EBADF);
+        assert_return(!(options & ~(WEXITED|WSTOPPED|WCONTINUED)), -EINVAL);
+        assert_return(options != 0, -EINVAL);
+        assert_return(callback, -EINVAL);
+        assert_return(e->state != SD_EVENT_FINISHED, -ESTALE);
+        assert_return(!event_pid_changed(e), -ECHILD);
+
+        if (e->n_enabled_child_sources == 0) {
+                r = signal_is_blocked(SIGCHLD);
+                if (r < 0)
+                        return r;
+                if (r == 0)
+                        return -EBUSY;
+        }
+
+        r = hashmap_ensure_allocated(&e->child_sources, NULL);
+        if (r < 0)
+                return r;
+
+        r = pidfd_get_pid(pidfd, &pid);
+        if (r < 0)
+                return r;
+
+        if (hashmap_contains(e->child_sources, PID_TO_PTR(pid)))
+                return -EBUSY;
+
+        s = source_new(e, !ret, SOURCE_CHILD);
+        if (!s)
+                return -ENOMEM;
+
+        s->wakeup = WAKEUP_EVENT_SOURCE;
+        s->child.pidfd = pidfd;
+        s->child.pid = pid;
+        s->child.options = options;
+        s->child.callback = callback;
+        s->child.pidfd_owned = false; /* If we got the pidfd passed in we don't own it by default (similar to the IO fd case) */
+        s->userdata = userdata;
+        s->enabled = SD_EVENT_ONESHOT;
+
         r = hashmap_put(e->child_sources, PID_TO_PTR(pid), s);
         if (r < 0)
                 return r;
 
         e->n_enabled_child_sources++;
 
-        r = event_make_signal_data(e, SIGCHLD, NULL);
-        if (r < 0) {
-                e->n_enabled_child_sources--;
-                return r;
-        }
+        if (EVENT_SOURCE_WATCH_PIDFD(s)) {
+                /* We only want to watch for WEXITED */
 
-        e->need_process_child = true;
+                r = source_child_pidfd_register(s, s->enabled);
+                if (r < 0) {
+                        e->n_enabled_child_sources--;
+                        return r;
+                }
+        } else {
+                /* We shall wait for some other event than WEXITED */
+
+                r = event_make_signal_data(e, SIGCHLD, NULL);
+                if (r < 0) {
+                        e->n_enabled_child_sources--;
+                        return r;
+                }
+
+                e->need_process_child = true;
+        }
 
         if (ret)
                 *ret = s;
-        TAKE_PTR(s);
 
+        TAKE_PTR(s);
         return 0;
 }
 
@@ -1765,7 +1988,7 @@ _public_ int sd_event_source_set_io_fd(sd_event_source *s, int fd) {
                         return r;
                 }
 
-                epoll_ctl(s->event->epoll_fd, EPOLL_CTL_DEL, saved_fd, NULL);
+                (void) epoll_ctl(s->event->epoll_fd, EPOLL_CTL_DEL, saved_fd, NULL);
         }
 
         return 0;
@@ -2026,7 +2249,11 @@ _public_ int sd_event_source_set_enabled(sd_event_source *s, int m) {
                         assert(s->event->n_enabled_child_sources > 0);
                         s->event->n_enabled_child_sources--;
 
-                        event_gc_signal_data(s->event, &s->priority, SIGCHLD);
+                        if (EVENT_SOURCE_WATCH_PIDFD(s))
+                                source_child_pidfd_unregister(s);
+                        else
+                                event_gc_signal_data(s->event, &s->priority, SIGCHLD);
+
                         break;
 
                 case SOURCE_EXIT:
@@ -2100,12 +2327,25 @@ _public_ int sd_event_source_set_enabled(sd_event_source *s, int m) {
 
                         s->enabled = m;
 
-                        r = event_make_signal_data(s->event, SIGCHLD, NULL);
-                        if (r < 0) {
-                                s->enabled = SD_EVENT_OFF;
-                                s->event->n_enabled_child_sources--;
-                                event_gc_signal_data(s->event, &s->priority, SIGCHLD);
-                                return r;
+                        if (EVENT_SOURCE_WATCH_PIDFD(s)) {
+                                /* yes, we have pidfd */
+
+                                r = source_child_pidfd_register(s, s->enabled);
+                                if (r < 0) {
+                                        s->enabled = SD_EVENT_OFF;
+                                        s->event->n_enabled_child_sources--;
+                                        return r;
+                                }
+                        } else {
+                                /* no pidfd, or something other to watch for than WEXITED */
+
+                                r = event_make_signal_data(s->event, SIGCHLD, NULL);
+                                if (r < 0) {
+                                        s->enabled = SD_EVENT_OFF;
+                                        s->event->n_enabled_child_sources--;
+                                        event_gc_signal_data(s->event, &s->priority, SIGCHLD);
+                                        return r;
+                                }
                         }
 
                         break;
@@ -2228,6 +2468,98 @@ _public_ int sd_event_source_get_child_pid(sd_event_source *s, pid_t *pid) {
         return 0;
 }
 
+_public_ int sd_event_source_get_child_pidfd(sd_event_source *s) {
+        assert_return(s, -EINVAL);
+        assert_return(s->type == SOURCE_CHILD, -EDOM);
+        assert_return(!event_pid_changed(s->event), -ECHILD);
+
+        if (s->child.pidfd < 0)
+                return -EOPNOTSUPP;
+
+        return s->child.pidfd;
+}
+
+_public_ int sd_event_source_send_child_signal(sd_event_source *s, int sig, const siginfo_t *si, unsigned flags) {
+        assert_return(s, -EINVAL);
+        assert_return(s->type == SOURCE_CHILD, -EDOM);
+        assert_return(!event_pid_changed(s->event), -ECHILD);
+        assert_return(SIGNAL_VALID(sig), -EINVAL);
+
+        /* If we already have seen indication the process exited refuse sending a signal early. This way we
+         * can be sure we don't accidentally kill the wrong process on PID reuse when pidfds are not
+         * available. */
+        if (s->child.exited)
+                return -ESRCH;
+
+        if (s->child.pidfd >= 0) {
+                siginfo_t copy;
+
+                /* pidfd_send_signal() changes the siginfo_t argument. This is weird, let's hence copy the
+                 * structure here */
+                if (si)
+                        copy = *si;
+
+                if (pidfd_send_signal(s->child.pidfd, sig, si ? &copy : NULL, 0) < 0) {
+                        /* Let's propagate the error only if the system call is not implemented or prohibited */
+                        if (!ERRNO_IS_NOT_SUPPORTED(errno) && !ERRNO_IS_PRIVILEGE(errno))
+                                return -errno;
+                } else
+                        return 0;
+        }
+
+        /* Flags are only supported for pidfd_send_signal(), not for rt_sigqueueinfo(), hence let's refuse
+         * this here. */
+        if (flags != 0)
+                return -EOPNOTSUPP;
+
+        if (si) {
+                /* We use rt_sigqueueinfo() only if siginfo_t is specified. */
+                siginfo_t copy = *si;
+
+                if (rt_sigqueueinfo(s->child.pid, sig, &copy) < 0)
+                        return -errno;
+        } else if (kill(s->child.pid, sig) < 0)
+                return -errno;
+
+        return 0;
+}
+
+_public_ int sd_event_source_get_child_pidfd_own(sd_event_source *s) {
+        assert_return(s, -EINVAL);
+        assert_return(s->type == SOURCE_CHILD, -EDOM);
+
+        if (s->child.pidfd < 0)
+                return -EOPNOTSUPP;
+
+        return s->child.pidfd_owned;
+}
+
+_public_ int sd_event_source_set_child_pidfd_own(sd_event_source *s, int own) {
+        assert_return(s, -EINVAL);
+        assert_return(s->type == SOURCE_CHILD, -EDOM);
+
+        if (s->child.pidfd < 0)
+                return -EOPNOTSUPP;
+
+        s->child.pidfd_owned = own;
+        return 0;
+}
+
+_public_ int sd_event_source_get_child_process_own(sd_event_source *s) {
+        assert_return(s, -EINVAL);
+        assert_return(s->type == SOURCE_CHILD, -EDOM);
+
+        return s->child.process_owned;
+}
+
+_public_ int sd_event_source_set_child_process_own(sd_event_source *s, int own) {
+        assert_return(s, -EINVAL);
+        assert_return(s->type == SOURCE_CHILD, -EDOM);
+
+        s->child.process_owned = own;
+        return 0;
+}
+
 _public_ int sd_event_source_get_inotify_mask(sd_event_source *s, uint32_t *mask) {
         assert_return(s, -EINVAL);
         assert_return(mask, -EINVAL);
@@ -2538,6 +2870,12 @@ static int process_child(sd_event *e) {
                 if (s->enabled == SD_EVENT_OFF)
                         continue;
 
+                if (s->child.exited)
+                        continue;
+
+                if (EVENT_SOURCE_WATCH_PIDFD(s)) /* There's a usable pidfd known for this event source? then don't waitid() for it here */
+                        continue;
+
                 zero(s->child.siginfo);
                 r = waitid(P_PID, s->child.pid, &s->child.siginfo,
                            WNOHANG | (s->child.options & WEXITED ? WNOWAIT : 0) | s->child.options);
@@ -2547,6 +2885,9 @@ static int process_child(sd_event *e) {
                 if (s->child.siginfo.si_pid != 0) {
                         bool zombie = IN_SET(s->child.siginfo.si_code, CLD_EXITED, CLD_KILLED, CLD_DUMPED);
 
+                        if (zombie)
+                                s->child.exited = true;
+
                         if (!zombie && (s->child.options & WEXITED)) {
                                 /* If the child isn't dead then let's
                                  * immediately remove the state change
@@ -2566,6 +2907,33 @@ static int process_child(sd_event *e) {
         return 0;
 }
 
+static int process_pidfd(sd_event *e, sd_event_source *s, uint32_t revents) {
+        assert(e);
+        assert(s);
+        assert(s->type == SOURCE_CHILD);
+
+        if (s->pending)
+                return 0;
+
+        if (s->enabled == SD_EVENT_OFF)
+                return 0;
+
+        if (!EVENT_SOURCE_WATCH_PIDFD(s))
+                return 0;
+
+        zero(s->child.siginfo);
+        if (waitid(P_PID, s->child.pid, &s->child.siginfo, WNOHANG | WNOWAIT | s->child.options) < 0)
+                return -errno;
+
+        if (s->child.siginfo.si_pid == 0)
+                return 0;
+
+        if (IN_SET(s->child.siginfo.si_code, CLD_EXITED, CLD_KILLED, CLD_DUMPED))
+                s->child.exited = true;
+
+        return source_set_pending(s, true);
+}
+
 static int process_signal(sd_event *e, struct signal_data *d, uint32_t events) {
         bool read_one = false;
         int r;
@@ -2850,8 +3218,10 @@ static int source_dispatch(sd_event_source *s) {
                 r = s->child.callback(s, &s->child.siginfo, s->userdata);
 
                 /* Now, reap the PID for good. */
-                if (zombie)
+                if (zombie) {
                         (void) waitid(P_PID, s->child.pid, &s->child.siginfo, WNOHANG|WEXITED);
+                        s->child.waited = true;
+                }
 
                 break;
         }
@@ -3052,6 +3422,11 @@ _public_ int sd_event_prepare(sd_event *e) {
         assert_return(e->state != SD_EVENT_FINISHED, -ESTALE);
         assert_return(e->state == SD_EVENT_INITIAL, -EBUSY);
 
+        /* Let's check that if we are a default event loop we are executed in the correct thread. We only do
+         * this check here once, since gettid() is typically not cached, and thus want to minimize
+         * syscalls */
+        assert_return(!e->default_event_ptr || e->tid == gettid(), -EREMOTEIO);
+
         if (e->exit_requested)
                 goto pending;
 
@@ -3147,12 +3522,33 @@ _public_ int sd_event_wait(sd_event *e, uint64_t timeout) {
 
                         switch (*t) {
 
-                        case WAKEUP_EVENT_SOURCE:
-                                r = process_io(e, ev_queue[i].data.ptr, ev_queue[i].events);
+                        case WAKEUP_EVENT_SOURCE: {
+                                sd_event_source *s = ev_queue[i].data.ptr;
+
+                                assert(s);
+
+                                switch (s->type) {
+
+                                case SOURCE_IO:
+                                        r = process_io(e, s, ev_queue[i].events);
+                                        break;
+
+                                case SOURCE_CHILD:
+                                        r = process_pidfd(e, s, ev_queue[i].events);
+                                        break;
+
+                                default:
+                                        assert_not_reached("Unexpected event source type");
+                                }
+
                                 break;
+                        }
 
                         case WAKEUP_CLOCK_DATA: {
                                 struct clock_data *d = ev_queue[i].data.ptr;
+
+                                assert(d);
+
                                 r = flush_timer(e, d->fd, ev_queue[i].events, &d->next);
                                 break;
                         }
@@ -3476,7 +3872,7 @@ _public_ int sd_event_set_watchdog(sd_event *e, int b) {
 
         } else {
                 if (e->watchdog_fd >= 0) {
-                        epoll_ctl(e->epoll_fd, EPOLL_CTL_DEL, e->watchdog_fd, NULL);
+                        (void) epoll_ctl(e->epoll_fd, EPOLL_CTL_DEL, e->watchdog_fd, NULL);
                         e->watchdog_fd = safe_close(e->watchdog_fd);
                 }
         }

src/libsystemd/sd-event/test-event.c

@@ -9,6 +9,7 @@
 #include "fs-util.h"
 #include "log.h"
 #include "macro.h"
+#include "missing_syscall.h"
 #include "parse-util.h"
 #include "path-util.h"
 #include "process-util.h"
@@ -62,6 +63,11 @@ static int child_handler(sd_event_source *s, const siginfo_t *si, void *userdata
         assert_se(s);
         assert_se(si);
 
+        assert_se(si->si_uid == getuid());
+        assert_se(si->si_signo == SIGCHLD);
+        assert_se(si->si_code == CLD_EXITED);
+        assert_se(si->si_status == 78);
+
         log_info("got child on %c", PTR_TO_INT(userdata));
 
         assert_se(userdata == INT_TO_PTR('f'));
@@ -75,6 +81,7 @@ static int child_handler(sd_event_source *s, const siginfo_t *si, void *userdata
 static int signal_handler(sd_event_source *s, const struct signalfd_siginfo *si, void *userdata) {
         sd_event_source *p = NULL;
         pid_t pid;
+        siginfo_t plain_si;
 
         assert_se(s);
         assert_se(si);
@@ -83,16 +90,41 @@ static int signal_handler(sd_event_source *s, const struct signalfd_siginfo *si,
 
         assert_se(userdata == INT_TO_PTR('e'));
 
-        assert_se(sigprocmask_many(SIG_BLOCK, NULL, SIGCHLD, -1) >= 0);
+        assert_se(sigprocmask_many(SIG_BLOCK, NULL, SIGCHLD, SIGUSR2, -1) >= 0);
 
         pid = fork();
         assert_se(pid >= 0);
 
-        if (pid == 0)
-                _exit(EXIT_SUCCESS);
+        if (pid == 0) {
+                sigset_t ss;
+
+                assert_se(sigemptyset(&ss) >= 0);
+                assert_se(sigaddset(&ss, SIGUSR2) >= 0);
+
+                zero(plain_si);
+                assert_se(sigwaitinfo(&ss, &plain_si) >= 0);
+
+                assert_se(plain_si.si_signo == SIGUSR2);
+                assert_se(plain_si.si_value.sival_int == 4711);
+
+                _exit(78);
+        }
 
         assert_se(sd_event_add_child(sd_event_source_get_event(s), &p, pid, WEXITED, child_handler, INT_TO_PTR('f')) >= 0);
         assert_se(sd_event_source_set_enabled(p, SD_EVENT_ONESHOT) >= 0);
+        assert_se(sd_event_source_set_child_process_own(p, true) >= 0);
+
+        /* We can't use structured initialization here, since the structure contains various unions and these
+         * fields lie in overlapping (carefully aligned) unions that LLVM is allergic to allow assignments
+         * to */
+        zero(plain_si);
+        plain_si.si_signo = SIGUSR2;
+        plain_si.si_code = SI_QUEUE;
+        plain_si.si_pid = getpid();
+        plain_si.si_uid = getuid();
+        plain_si.si_value.sival_int = 4711;
+
+        assert_se(sd_event_source_send_child_signal(p, SIGUSR2, &plain_si, 0) >= 0);
 
         sd_event_source_unref(s);
 
@@ -119,7 +151,7 @@ static int defer_handler(sd_event_source *s, void *userdata) {
         return 1;
 }
 
-static bool do_quit = false;
+static bool do_quit;
 
 static int time_handler(sd_event_source *s, uint64_t usec, void *userdata) {
         log_info("got timer on %c", PTR_TO_INT(userdata));
@@ -161,7 +193,7 @@ static int post_handler(sd_event_source *s, void *userdata) {
         return 2;
 }
 
-static void test_basic(void) {
+static void test_basic(bool with_pidfd) {
         sd_event *e = NULL;
         sd_event_source *w = NULL, *x = NULL, *y = NULL, *z = NULL, *q = NULL, *t = NULL;
         static const char ch = 'x';
@@ -169,6 +201,8 @@ static void test_basic(void) {
         uint64_t event_now;
         int64_t priority;
 
+        assert_se(setenv("SYSTEMD_PIDFD", yes_no(with_pidfd), 1) >= 0);
+
         assert_se(pipe(a) >= 0);
         assert_se(pipe(b) >= 0);
         assert_se(pipe(d) >= 0);
@@ -201,6 +235,8 @@ static void test_basic(void) {
 
         assert_se(sd_event_add_io(e, &x, a[0], EPOLLIN, io_handler, INT_TO_PTR('a')) >= 0);
         assert_se(sd_event_add_io(e, &y, b[0], EPOLLIN, io_handler, INT_TO_PTR('b')) >= 0);
+
+        do_quit = false;
         assert_se(sd_event_add_time(e, &z, CLOCK_MONOTONIC, 0, 0, time_handler, INT_TO_PTR('c')) >= 0);
         assert_se(sd_event_add_exit(e, &q, exit_handler, INT_TO_PTR('g')) >= 0);
 
@@ -258,6 +294,8 @@ static void test_basic(void) {
         safe_close_pair(b);
         safe_close_pair(d);
         safe_close_pair(k);
+
+        assert_se(unsetenv("SYSTEMD_PIDFD") >= 0);
 }
 
 static void test_sd_event_now(void) {
@@ -482,15 +520,89 @@ static void test_inotify(unsigned n_create_events) {
         sd_event_unref(e);
 }
 
+static int pidfd_handler(sd_event_source *s, const siginfo_t *si, void *userdata) {
+        assert_se(s);
+        assert_se(si);
+
+        assert_se(si->si_uid == getuid());
+        assert_se(si->si_signo == SIGCHLD);
+        assert_se(si->si_code == CLD_EXITED);
+        assert_se(si->si_status == 66);
+
+        log_info("got pidfd on %c", PTR_TO_INT(userdata));
+
+        assert_se(userdata == INT_TO_PTR('p'));
+
+        assert_se(sd_event_exit(sd_event_source_get_event(s), 0) >= 0);
+        sd_event_source_unref(s);
+
+        return 0;
+}
+
+static void test_pidfd(void) {
+        sd_event_source *s = NULL, *t = NULL;
+        sd_event *e = NULL;
+        int pidfd;
+        pid_t pid, pid2;
+
+        assert_se(sigprocmask_many(SIG_BLOCK, NULL, SIGCHLD, -1) >= 0);
+
+        pid = fork();
+        if (pid == 0) {
+                /* child */
+                _exit(66);
+        }
+
+        assert_se(pid > 1);
+
+        pidfd = pidfd_open(pid, 0);
+        if (pidfd < 0) {
+                /* No pidfd_open() supported or blocked? */
+                assert_se(ERRNO_IS_NOT_SUPPORTED(errno) || ERRNO_IS_PRIVILEGE(errno));
+                (void) wait_for_terminate(pid, NULL);
+                return;
+        }
+
+        pid2 = fork();
+        if (pid2 == 0)
+                freeze();
+
+        assert_se(pid > 2);
+
+        assert_se(sd_event_default(&e) >= 0);
+        assert_se(sd_event_add_child_pidfd(e, &s, pidfd, WEXITED, pidfd_handler, INT_TO_PTR('p')) >= 0);
+        assert_se(sd_event_source_set_child_pidfd_own(s, true) >= 0);
+
+        /* This one should never trigger, since our second child lives forever */
+        assert_se(sd_event_add_child(e, &t, pid2, WEXITED, pidfd_handler, INT_TO_PTR('q')) >= 0);
+        assert_se(sd_event_source_set_child_process_own(t, true) >= 0);
+
+        assert_se(sd_event_loop(e) >= 0);
+
+        /* Child should still be alive */
+        assert_se(kill(pid2, 0) >= 0);
+
+        t = sd_event_source_unref(t);
+
+        /* Child should now be dead, since we dropped the ref */
+        assert_se(kill(pid2, 0) < 0 && errno == ESRCH);
+
+        sd_event_unref(e);
+}
+
 int main(int argc, char *argv[]) {
         test_setup_logging(LOG_INFO);
 
-        test_basic();
+        test_basic(true);   /* test with pidfd */
+        test_basic(false);  /* test without pidfd */
+
         test_sd_event_now();
         test_rtqueue();
 
         test_inotify(100); /* should work without overflow */
         test_inotify(33000); /* should trigger a q overflow */
 
+        test_pidfd();
+
         return 0;
 }

src/libsystemd/sd-netlink/netlink-message.c

@@ -575,7 +575,9 @@ static int netlink_message_read_internal(sd_netlink_message *m, unsigned short t
         assert_return(data, -EINVAL);
 
         assert(m->n_containers < RTNL_CONTAINER_DEPTH);
-        assert(m->containers[m->n_containers].attributes);
+
+        if (!m->containers[m->n_containers].attributes)
+                return -ENODATA;
 
         if (type >= m->containers[m->n_containers].n_attributes)
                 return -ENODATA;
@@ -1021,6 +1023,27 @@ int sd_netlink_message_get_errno(const sd_netlink_message *m) {
         return err->error;
 }
 
+static int netlink_message_parse_error(sd_netlink_message *m) {
+        struct nlmsgerr *err = NLMSG_DATA(m->hdr);
+        size_t hlen = sizeof(struct nlmsgerr);
+
+        /* no TLVs, nothing to do here */
+        if (!(m->hdr->nlmsg_flags & NLM_F_ACK_TLVS))
+                return 0;
+
+        /* if NLM_F_CAPPED is set then the inner err msg was capped */
+        if (!(m->hdr->nlmsg_flags & NLM_F_CAPPED))
+                hlen += err->msg.nlmsg_len - sizeof(struct nlmsghdr);
+
+        if (m->hdr->nlmsg_len <= NLMSG_SPACE(hlen))
+                return 0;
+
+        return netlink_container_parse(m,
+                                       &m->containers[m->n_containers],
+                                       (struct rtattr*)((uint8_t*) NLMSG_DATA(m->hdr) + hlen),
+                                       NLMSG_PAYLOAD(m->hdr, hlen));
+}
+
 int sd_netlink_message_rewind(sd_netlink_message *m, sd_netlink *genl) {
         const NLType *nl_type;
         uint16_t type;
@@ -1060,10 +1083,13 @@ int sd_netlink_message_rewind(sd_netlink_message *m, sd_netlink *genl) {
 
                 m->containers[0].type_system = type_system;
 
-                r = netlink_container_parse(m,
-                                            &m->containers[m->n_containers],
-                                            (struct rtattr*)((uint8_t*)NLMSG_DATA(m->hdr) + NLMSG_ALIGN(size)),
-                                            NLMSG_PAYLOAD(m->hdr, size));
+                if (sd_netlink_message_is_error(m))
+                        r = netlink_message_parse_error(m);
+                else
+                        r = netlink_container_parse(m,
+                                                    &m->containers[m->n_containers],
+                                                    (struct rtattr*)((uint8_t*) NLMSG_DATA(m->hdr) + NLMSG_ALIGN(size)),
+                                                    NLMSG_PAYLOAD(m->hdr, size));
                 if (r < 0)
                         return r;
         }

src/libsystemd/sd-netlink/netlink-types.c

@@ -745,9 +745,19 @@ static const NLTypeSystem rtnl_qdisc_type_system = {
         .types = rtnl_qdisc_types,
 };
 
+static const NLType error_types[] = {
+        [NLMSGERR_ATTR_MSG]  = { .type = NETLINK_TYPE_STRING },
+        [NLMSGERR_ATTR_OFFS] = { .type = NETLINK_TYPE_U32 },
+};
+
+static const NLTypeSystem error_type_system = {
+        .count = ELEMENTSOF(error_types),
+        .types = error_types,
+};
+
 static const NLType rtnl_types[] = {
         [NLMSG_DONE]       = { .type = NETLINK_TYPE_NESTED, .type_system = &empty_type_system, .size = 0 },
-        [NLMSG_ERROR]      = { .type = NETLINK_TYPE_NESTED, .type_system = &empty_type_system, .size = sizeof(struct nlmsgerr) },
+        [NLMSG_ERROR]      = { .type = NETLINK_TYPE_NESTED, .type_system = &error_type_system, .size = sizeof(struct nlmsgerr) },
         [RTM_NEWLINK]      = { .type = NETLINK_TYPE_NESTED, .type_system = &rtnl_link_type_system, .size = sizeof(struct ifinfomsg) },
         [RTM_DELLINK]      = { .type = NETLINK_TYPE_NESTED, .type_system = &rtnl_link_type_system, .size = sizeof(struct ifinfomsg) },
         [RTM_GETLINK]      = { .type = NETLINK_TYPE_NESTED, .type_system = &rtnl_link_type_system, .size = sizeof(struct ifinfomsg) },
@@ -1052,7 +1062,7 @@ const NLTypeSystem genl_family_type_system_root = {
 };
 
 static const NLType genl_types[] = {
-        [SD_GENL_ERROR]   = { .type = NETLINK_TYPE_NESTED, .type_system = &empty_type_system, .size = sizeof(struct nlmsgerr) },
+        [SD_GENL_ERROR]   = { .type = NETLINK_TYPE_NESTED, .type_system = &error_type_system, .size = sizeof(struct nlmsgerr) },
         [SD_GENL_DONE]    = { .type = NETLINK_TYPE_NESTED, .type_system = &empty_type_system },
         [SD_GENL_ID_CTRL] = { .type = NETLINK_TYPE_NESTED, .type_system = &genl_get_family_type_system, .size = sizeof(struct genlmsghdr) },
         [SD_GENL_NL80211] = { .type = NETLINK_TYPE_NESTED, .type_system = &genl_nl80211_type_system, .size = sizeof(struct genlmsghdr) },

src/libsystemd/sd-netlink/sd-netlink.c

@@ -107,6 +107,10 @@ int sd_netlink_open_fd(sd_netlink **ret, int fd) {
         rtnl->fd = fd;
         rtnl->protocol = protocol;
 
+        r = setsockopt_int(fd, SOL_NETLINK, NETLINK_EXT_ACK, 1);
+        if (r < 0)
+                log_debug_errno(r, "sd-netlink: Failed to enable NETLINK_EXT_ACK option, ignoring: %m");
+
         r = socket_bind(rtnl);
         if (r < 0) {
                 rtnl->fd = -1; /* on failure, the caller remains owner of the fd, hence don't close it here */

src/network/netdev/bond.c

@@ -587,7 +587,7 @@ const NetDevVTable bond_vtable = {
         .object_size = sizeof(Bond),
         .init = bond_init,
         .done = bond_done,
-        .sections = "Match\0NetDev\0Bond\0",
+        .sections = NETDEV_COMMON_SECTIONS "Bond\0",
         .fill_message_create = netdev_bond_fill_message_create,
         .create_type = NETDEV_CREATE_MASTER,
         .generate_mac = true,

src/network/netdev/bridge.c

@@ -355,7 +355,7 @@ static void bridge_init(NetDev *n) {
 const NetDevVTable bridge_vtable = {
         .object_size = sizeof(Bridge),
         .init = bridge_init,
-        .sections = "Match\0NetDev\0Bridge\0",
+        .sections = NETDEV_COMMON_SECTIONS "Bridge\0",
         .post_create = netdev_bridge_post_create,
         .create_type = NETDEV_CREATE_MASTER,
 };

src/network/netdev/dummy.c

@@ -4,7 +4,7 @@
 
 const NetDevVTable dummy_vtable = {
         .object_size = sizeof(Dummy),
-        .sections = "Match\0NetDev\0",
+        .sections = NETDEV_COMMON_SECTIONS,
         .create_type = NETDEV_CREATE_INDEPENDENT,
         .generate_mac = true,
 };

src/network/netdev/fou-tunnel.c

@@ -262,7 +262,7 @@ static void fou_tunnel_init(NetDev *netdev) {
 const NetDevVTable foutnl_vtable = {
         .object_size = sizeof(FouTunnel),
         .init = fou_tunnel_init,
-        .sections = "Match\0NetDev\0FooOverUDP\0",
+        .sections = NETDEV_COMMON_SECTIONS "FooOverUDP\0",
         .create = netdev_fou_tunnel_create,
         .create_type = NETDEV_CREATE_INDEPENDENT,
         .config_verify = netdev_fou_tunnel_verify,

src/network/netdev/geneve.c

@@ -348,7 +348,7 @@ static void geneve_init(NetDev *netdev) {
 const NetDevVTable geneve_vtable = {
         .object_size = sizeof(Geneve),
         .init = geneve_init,
-        .sections = "Match\0NetDev\0GENEVE\0",
+        .sections = NETDEV_COMMON_SECTIONS "GENEVE\0",
         .create = netdev_geneve_create,
         .create_type = NETDEV_CREATE_INDEPENDENT,
         .config_verify = netdev_geneve_verify,

src/network/netdev/ipvlan.c

@@ -74,7 +74,7 @@ static void ipvlan_init(NetDev *n) {
 const NetDevVTable ipvlan_vtable = {
         .object_size = sizeof(IPVlan),
         .init = ipvlan_init,
-        .sections = "Match\0NetDev\0IPVLAN\0",
+        .sections = NETDEV_COMMON_SECTIONS "IPVLAN\0",
         .fill_message_create = netdev_ipvlan_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .generate_mac = true,
@@ -83,7 +83,7 @@ const NetDevVTable ipvlan_vtable = {
 const NetDevVTable ipvtap_vtable = {
         .object_size = sizeof(IPVlan),
         .init = ipvlan_init,
-        .sections = "Match\0NetDev\0IPVTAP\0",
+        .sections = NETDEV_COMMON_SECTIONS "IPVTAP\0",
         .fill_message_create = netdev_ipvlan_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .generate_mac = true,

src/network/netdev/l2tp-tunnel.c

@@ -723,7 +723,7 @@ static void l2tp_tunnel_done(NetDev *netdev) {
 const NetDevVTable l2tptnl_vtable = {
         .object_size = sizeof(L2tpTunnel),
         .init = l2tp_tunnel_init,
-        .sections = "Match\0NetDev\0L2TP\0L2TPSession\0",
+        .sections = NETDEV_COMMON_SECTIONS "L2TP\0L2TPSession\0",
         .create_after_configured = l2tp_create_tunnel,
         .done = l2tp_tunnel_done,
         .create_type = NETDEV_CREATE_AFTER_CONFIGURED,

src/network/netdev/macsec.c

@@ -1235,7 +1235,7 @@ static void macsec_done(NetDev *netdev) {
 const NetDevVTable macsec_vtable = {
         .object_size = sizeof(MACsec),
         .init = macsec_init,
-        .sections = "Match\0NetDev\0MACsec\0MACsecReceiveChannel\0MACsecTransmitAssociation\0MACsecReceiveAssociation\0",
+        .sections = NETDEV_COMMON_SECTIONS "MACsec\0MACsecReceiveChannel\0MACsecTransmitAssociation\0MACsecReceiveAssociation\0",
         .fill_message_create = netdev_macsec_fill_message_create,
         .post_create = netdev_macsec_configure,
         .done = macsec_done,

src/network/netdev/macvlan.c

@@ -58,7 +58,7 @@ static void macvlan_init(NetDev *n) {
 const NetDevVTable macvtap_vtable = {
         .object_size = sizeof(MacVlan),
         .init = macvlan_init,
-        .sections = "Match\0NetDev\0MACVTAP\0",
+        .sections = NETDEV_COMMON_SECTIONS "MACVTAP\0",
         .fill_message_create = netdev_macvlan_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .generate_mac = true,
@@ -67,7 +67,7 @@ const NetDevVTable macvtap_vtable = {
 const NetDevVTable macvlan_vtable = {
         .object_size = sizeof(MacVlan),
         .init = macvlan_init,
-        .sections = "Match\0NetDev\0MACVLAN\0",
+        .sections = NETDEV_COMMON_SECTIONS "MACVLAN\0",
         .fill_message_create = netdev_macvlan_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .generate_mac = true,

src/network/netdev/netdev.c

@@ -682,9 +682,9 @@ int netdev_load_one(Manager *manager, const char *filename) {
 
         dropin_dirname = strjoina(basename(filename), ".d");
         r = config_parse_many(filename, NETWORK_DIRS, dropin_dirname,
-                              "Match\0NetDev\0",
+                              NETDEV_COMMON_SECTIONS NETDEV_OTHER_SECTIONS,
                               config_item_perf_lookup, network_netdev_gperf_lookup,
-                              CONFIG_PARSE_WARN|CONFIG_PARSE_RELAXED, netdev_raw);
+                              CONFIG_PARSE_WARN, netdev_raw);
         if (r < 0)
                 return r;
 

src/network/netdev/netdev.h

@@ -8,6 +8,35 @@
 #include "networkd-link.h"
 #include "time-util.h"
 
+#define NETDEV_COMMON_SECTIONS "Match\0NetDev\0"
+/* This is the list of known sections. We need to ignore them in the initial parsing phase. */
+#define NETDEV_OTHER_SECTIONS                     \
+        "-Bond\0"                                 \
+        "-Bridge\0"                               \
+        "-FooOverUDP\0"                           \
+        "-GENEVE\0"                               \
+        "-IPVLAN\0"                               \
+        "-IPVTAP\0"                               \
+        "-L2TP\0"                                 \
+        "-L2TPSession\0"                          \
+        "-MACsec\0"                               \
+        "-MACsecReceiveChannel\0"                 \
+        "-MACsecTransmitAssociation\0"            \
+        "-MACsecReceiveAssociation\0"             \
+        "-MACVTAP\0"                              \
+        "-MACVLAN\0"                              \
+        "-Tunnel\0"                               \
+        "-Tun\0"                                  \
+        "-Tap\0"                                  \
+        "-Peer\0"                                 \
+        "-VLAN\0"                                 \
+        "-VRF\0"                                  \
+        "-VXCAN\0"                                \
+        "-VXLAN\0"                                \
+        "-WireGuard\0"                            \
+        "-WireGuardPeer\0"                        \
+        "-Xfrm\0"
+
 typedef struct netdev_join_callback netdev_join_callback;
 
 struct netdev_join_callback {

src/network/netdev/netdevsim.c

@@ -4,7 +4,7 @@
 
 const NetDevVTable netdevsim_vtable = {
         .object_size = sizeof(NetDevSim),
-        .sections = "Match\0NetDev\0",
+        .sections = NETDEV_COMMON_SECTIONS,
         .create_type = NETDEV_CREATE_INDEPENDENT,
         .generate_mac = true,
 };

src/network/netdev/nlmon.c

@@ -16,7 +16,7 @@ static int netdev_nlmon_verify(NetDev *netdev, const char *filename) {
 
 const NetDevVTable nlmon_vtable = {
         .object_size = sizeof(NLMon),
-        .sections = "Match\0NetDev\0",
+        .sections = NETDEV_COMMON_SECTIONS,
         .create_type = NETDEV_CREATE_INDEPENDENT,
         .config_verify = netdev_nlmon_verify,
 };

src/network/netdev/tunnel.c

@@ -805,7 +805,7 @@ static void ip6tnl_init(NetDev *n) {
 const NetDevVTable ipip_vtable = {
         .object_size = sizeof(Tunnel),
         .init = ipip_sit_init,
-        .sections = "Match\0NetDev\0Tunnel\0",
+        .sections = NETDEV_COMMON_SECTIONS "Tunnel\0",
         .fill_message_create = netdev_ipip_sit_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .config_verify = netdev_tunnel_verify,
@@ -815,7 +815,7 @@ const NetDevVTable ipip_vtable = {
 const NetDevVTable sit_vtable = {
         .object_size = sizeof(Tunnel),
         .init = ipip_sit_init,
-        .sections = "Match\0NetDev\0Tunnel\0",
+        .sections = NETDEV_COMMON_SECTIONS "Tunnel\0",
         .fill_message_create = netdev_ipip_sit_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .config_verify = netdev_tunnel_verify,
@@ -825,7 +825,7 @@ const NetDevVTable sit_vtable = {
 const NetDevVTable vti_vtable = {
         .object_size = sizeof(Tunnel),
         .init = vti_init,
-        .sections = "Match\0NetDev\0Tunnel\0",
+        .sections = NETDEV_COMMON_SECTIONS "Tunnel\0",
         .fill_message_create = netdev_vti_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .config_verify = netdev_tunnel_verify,
@@ -835,7 +835,7 @@ const NetDevVTable vti_vtable = {
 const NetDevVTable vti6_vtable = {
         .object_size = sizeof(Tunnel),
         .init = vti_init,
-        .sections = "Match\0NetDev\0Tunnel\0",
+        .sections = NETDEV_COMMON_SECTIONS "Tunnel\0",
         .fill_message_create = netdev_vti_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .config_verify = netdev_tunnel_verify,
@@ -845,7 +845,7 @@ const NetDevVTable vti6_vtable = {
 const NetDevVTable gre_vtable = {
         .object_size = sizeof(Tunnel),
         .init = gre_erspan_init,
-        .sections = "Match\0NetDev\0Tunnel\0",
+        .sections = NETDEV_COMMON_SECTIONS "Tunnel\0",
         .fill_message_create = netdev_gre_erspan_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .config_verify = netdev_tunnel_verify,
@@ -855,7 +855,7 @@ const NetDevVTable gre_vtable = {
 const NetDevVTable gretap_vtable = {
         .object_size = sizeof(Tunnel),
         .init = gre_erspan_init,
-        .sections = "Match\0NetDev\0Tunnel\0",
+        .sections = NETDEV_COMMON_SECTIONS "Tunnel\0",
         .fill_message_create = netdev_gre_erspan_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .config_verify = netdev_tunnel_verify,
@@ -865,7 +865,7 @@ const NetDevVTable gretap_vtable = {
 const NetDevVTable ip6gre_vtable = {
         .object_size = sizeof(Tunnel),
         .init = ip6gre_init,
-        .sections = "Match\0NetDev\0Tunnel\0",
+        .sections = NETDEV_COMMON_SECTIONS "Tunnel\0",
         .fill_message_create = netdev_ip6gre_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .config_verify = netdev_tunnel_verify,
@@ -875,7 +875,7 @@ const NetDevVTable ip6gre_vtable = {
 const NetDevVTable ip6gretap_vtable = {
         .object_size = sizeof(Tunnel),
         .init = ip6gre_init,
-        .sections = "Match\0NetDev\0Tunnel\0",
+        .sections = NETDEV_COMMON_SECTIONS "Tunnel\0",
         .fill_message_create = netdev_ip6gre_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .config_verify = netdev_tunnel_verify,
@@ -885,7 +885,7 @@ const NetDevVTable ip6gretap_vtable = {
 const NetDevVTable ip6tnl_vtable = {
         .object_size = sizeof(Tunnel),
         .init = ip6tnl_init,
-        .sections = "Match\0NetDev\0Tunnel\0",
+        .sections = NETDEV_COMMON_SECTIONS "Tunnel\0",
         .fill_message_create = netdev_ip6tnl_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .config_verify = netdev_tunnel_verify,
@@ -895,7 +895,7 @@ const NetDevVTable ip6tnl_vtable = {
 const NetDevVTable erspan_vtable = {
         .object_size = sizeof(Tunnel),
         .init = gre_erspan_init,
-        .sections = "Match\0NetDev\0Tunnel\0",
+        .sections = NETDEV_COMMON_SECTIONS "Tunnel\0",
         .fill_message_create = netdev_gre_erspan_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .config_verify = netdev_tunnel_verify,

src/network/netdev/tuntap.c

@@ -147,7 +147,7 @@ static int tuntap_verify(NetDev *netdev, const char *filename) {
 
 const NetDevVTable tun_vtable = {
         .object_size = sizeof(TunTap),
-        .sections = "Match\0NetDev\0Tun\0",
+        .sections = NETDEV_COMMON_SECTIONS "Tun\0",
         .config_verify = tuntap_verify,
         .done = tuntap_done,
         .create = netdev_create_tuntap,
@@ -156,7 +156,7 @@ const NetDevVTable tun_vtable = {
 
 const NetDevVTable tap_vtable = {
         .object_size = sizeof(TunTap),
-        .sections = "Match\0NetDev\0Tap\0",
+        .sections = NETDEV_COMMON_SECTIONS "Tap\0",
         .config_verify = tuntap_verify,
         .done = tuntap_done,
         .create = netdev_create_tuntap,

src/network/netdev/vcan.c

@@ -4,7 +4,7 @@
 
 const NetDevVTable vcan_vtable = {
         .object_size = sizeof(VCan),
-        .sections = "Match\0NetDev\0",
+        .sections = NETDEV_COMMON_SECTIONS,
         .create_type = NETDEV_CREATE_INDEPENDENT,
         .generate_mac = true,
 };

src/network/netdev/veth.c

@@ -85,7 +85,7 @@ static void veth_done(NetDev *n) {
 
 const NetDevVTable veth_vtable = {
         .object_size = sizeof(Veth),
-        .sections = "Match\0NetDev\0Peer\0",
+        .sections = NETDEV_COMMON_SECTIONS "Peer\0",
         .done = veth_done,
         .fill_message_create = netdev_veth_fill_message_create,
         .create_type = NETDEV_CREATE_INDEPENDENT,

src/network/netdev/vlan.c

@@ -85,7 +85,7 @@ static void vlan_init(NetDev *netdev) {
 const NetDevVTable vlan_vtable = {
         .object_size = sizeof(VLan),
         .init = vlan_init,
-        .sections = "Match\0NetDev\0VLAN\0",
+        .sections = NETDEV_COMMON_SECTIONS "VLAN\0",
         .fill_message_create = netdev_vlan_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .config_verify = netdev_vlan_verify,

src/network/netdev/vrf.c

@@ -25,7 +25,7 @@ static int netdev_vrf_fill_message_create(NetDev *netdev, Link *link, sd_netlink
 
 const NetDevVTable vrf_vtable = {
         .object_size = sizeof(Vrf),
-        .sections = "Match\0NetDev\0VRF\0",
+        .sections = NETDEV_COMMON_SECTIONS "VRF\0",
         .fill_message_create = netdev_vrf_fill_message_create,
         .create_type = NETDEV_CREATE_MASTER,
         .generate_mac = true,

src/network/netdev/vxcan.c

@@ -65,7 +65,7 @@ static void vxcan_done(NetDev *n) {
 
 const NetDevVTable vxcan_vtable = {
         .object_size = sizeof(VxCan),
-        .sections = "Match\0NetDev\0VXCAN\0",
+        .sections = NETDEV_COMMON_SECTIONS "VXCAN\0",
         .done = vxcan_done,
         .fill_message_create = netdev_vxcan_fill_message_create,
         .create_type = NETDEV_CREATE_INDEPENDENT,

src/network/netdev/vxlan.c

@@ -371,7 +371,7 @@ static void vxlan_init(NetDev *netdev) {
 const NetDevVTable vxlan_vtable = {
         .object_size = sizeof(VxLan),
         .init = vxlan_init,
-        .sections = "Match\0NetDev\0VXLAN\0",
+        .sections = NETDEV_COMMON_SECTIONS "VXLAN\0",
         .fill_message_create = netdev_vxlan_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED,
         .config_verify = netdev_vxlan_verify,

src/network/netdev/wireguard.c

@@ -967,7 +967,7 @@ static int wireguard_verify(NetDev *netdev, const char *filename) {
 
 const NetDevVTable wireguard_vtable = {
         .object_size = sizeof(Wireguard),
-        .sections = "Match\0NetDev\0WireGuard\0WireGuardPeer\0",
+        .sections = NETDEV_COMMON_SECTIONS "WireGuard\0WireGuardPeer\0",
         .post_create = netdev_wireguard_post_create,
         .init = wireguard_init,
         .done = wireguard_done,

src/network/netdev/xfrm.c

@@ -27,7 +27,7 @@ static int xfrm_fill_message_create(NetDev *netdev, Link *link, sd_netlink_messa
 
 const NetDevVTable xfrm_vtable = {
         .object_size = sizeof(Xfrm),
-        .sections = "Match\0NetDev\0Xfrm\0",
+        .sections = NETDEV_COMMON_SECTIONS "Xfrm\0",
         .fill_message_create = xfrm_fill_message_create,
         .create_type = NETDEV_CREATE_STACKED
 };

src/network/networkctl.c

@@ -349,8 +349,7 @@ static int acquire_link_bitrates(sd_bus *bus, LinkInfo *link) {
                         "org.freedesktop.network1.Link",
                         "BitRates");
         if (r < 0) {
-                bool quiet = sd_bus_error_has_name(&error, SD_BUS_ERROR_UNKNOWN_PROPERTY) ||
-                             sd_bus_error_has_name(&error, BUS_ERROR_SPEED_METER_INACTIVE);
+                bool quiet = sd_bus_error_has_name(&error, SD_BUS_ERROR_UNKNOWN_PROPERTY);
 
                 return log_full_errno(quiet ? LOG_DEBUG : LOG_WARNING,
                                       r, "Failed to query link bit rates: %s", bus_error_message(&error, r));
@@ -368,7 +367,7 @@ static int acquire_link_bitrates(sd_bus *bus, LinkInfo *link) {
         if (r < 0)
                 return bus_log_parse_error(r);
 
-        link->has_bitrates = true;
+        link->has_bitrates = link->tx_bitrate != UINT64_MAX && link->rx_bitrate != UINT64_MAX;
 
         return 0;
 }

src/network/networkd-address-label.c

@@ -95,7 +95,7 @@ static int address_label_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_warning_errno(link, r, "could not set address label: %m");
+                log_link_message_warning_errno(link, m, r, "Could not set address label");
                 link_enter_failed(link);
                 return 1;
         } else if (r >= 0)

src/network/networkd-address.c

@@ -431,7 +431,7 @@ static int address_remove_handler(sd_netlink *rtnl, sd_netlink_message *m, Link
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EADDRNOTAVAIL)
-                log_link_warning_errno(link, r, "Could not drop address: %m");
+                log_link_message_warning_errno(link, m, r, "Could not drop address");
 
         return 1;
 }

src/network/networkd-brvlan.c

@@ -141,7 +141,7 @@ static int set_brvlan_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *lin
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST)
-                log_link_error_errno(link, r, "Could not add VLAN to bridge port: %m");
+                log_link_message_warning_errno(link, m, r, "Could not add VLAN to bridge port");
 
         return 1;
 }

src/network/networkd-can.c

@@ -20,7 +20,7 @@ static int link_up_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link)
         r = sd_netlink_message_get_errno(m);
         if (r < 0)
                 /* we warn but don't fail the link, as it may be brought up later */
-                log_link_warning_errno(link, r, "Could not bring up interface: %m");
+                log_link_message_warning_errno(link, m, r, "Could not bring up interface");
 
         return 1;
 }
@@ -60,7 +60,7 @@ static int link_set_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link)
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_error_errno(link, r, "Failed to configure CAN link: %m");
+                log_link_message_warning_errno(link, m, r, "Failed to configure CAN link");
                 link_enter_failed(link);
         }
 
@@ -183,7 +183,7 @@ static int link_down_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0) {
-                log_link_warning_errno(link, r, "Could not bring down interface: %m");
+                log_link_message_warning_errno(link, m, r, "Could not bring down interface");
                 link_enter_failed(link);
                 return 1;
         }

src/network/networkd-dhcp4.c

@@ -65,11 +65,11 @@ static int dhcp4_route_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *li
                 /* It seems kernel does not support that the prefix route cannot be configured with
                  * route table. Let's once drop the config and reconfigure them later. */
 
-                log_link_debug_errno(link, r, "Could not set DHCPv4 route, retrying later: %m");
+                log_link_message_debug_errno(link, m, r, "Could not set DHCPv4 route, retrying later: %m");
                 link->dhcp4_route_failed = true;
                 link->manager->dhcp4_prefix_root_cannot_set_table = true;
         } else if (r < 0 && r != -EEXIST) {
-                log_link_error_errno(link, r, "Could not set DHCPv4 route: %m");
+                log_link_message_warning_errno(link, m, r, "Could not set DHCPv4 route: %m");
                 link_enter_failed(link);
                 return 1;
         }
@@ -543,7 +543,7 @@ static int dhcp_remove_address_handler(sd_netlink *rtnl, sd_netlink_message *m,
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0)
-                log_link_debug_errno(link, r, "Failed to remove DHCPv4 address, ignoring: %m");
+                log_link_message_warning_errno(link, m, r, "Failed to remove DHCPv4 address, ignoring");
         else
                 (void) manager_rtnl_process_address(rtnl, m, link->manager);
 
@@ -665,7 +665,7 @@ static int dhcp4_address_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_error_errno(link, r, "Could not set DHCPv4 address: %m");
+                log_link_message_warning_errno(link, m, r, "Could not set DHCPv4 address");
                 link_enter_failed(link);
                 return 1;
         } else if (r >= 0)

src/network/networkd-dhcp6.c

@@ -112,7 +112,7 @@ static int dhcp6_route_remove_handler(sd_netlink *nl, sd_netlink_message *m, Lin
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0)
-                log_link_debug_errno(link, r, "Received error on unreachable route removal for DHCPv6 delegated subnet: %m");
+                log_link_message_warning_errno(link, m, r, "Received error on unreachable route removal for DHCPv6 delegated subnet");
 
         return 1;
 }
@@ -251,7 +251,7 @@ static int dhcp6_route_handler(sd_netlink *nl, sd_netlink_message *m, Link *link
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST)
-                log_link_debug_errno(link, r, "Received error when adding unreachable route for DHCPv6 delegated subnet: %m");
+                log_link_message_warning_errno(link, m, r, "Received error when adding unreachable route for DHCPv6 delegated subnet");
 
         return 1;
 }
@@ -402,7 +402,7 @@ static int dhcp6_address_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_error_errno(link, r, "Could not set DHCPv6 address: %m");
+                log_link_message_warning_errno(link, m, r, "Could not set DHCPv6 address");
                 link_enter_failed(link);
                 return 1;
         } else if (r >= 0)
@@ -714,7 +714,7 @@ static int dhcp6_route_add_handler(sd_netlink *nl, sd_netlink_message *m, Link *
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_debug_errno(link, r, "Received error adding DHCPv6 Prefix Delegation route: %m");
+                log_link_message_warning_errno(link, m, r, "Received error adding DHCPv6 Prefix Delegation route");
                 link_enter_failed(link);
                 return 1;
         }
@@ -780,7 +780,7 @@ static int dhcp6_prefix_remove_handler(sd_netlink *nl, sd_netlink_message *m, Li
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0) {
-                log_link_debug_errno(link, r, "Received error on DHCPv6 Prefix Delegation route removal: %m");
+                log_link_message_warning_errno(link, m, r, "Received error on DHCPv6 Prefix Delegation route removal");
                 link_enter_failed(link);
                 return 1;
         }

src/network/networkd-fdb.c

@@ -105,7 +105,7 @@ static int set_fdb_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link)
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_error_errno(link, r, "Could not add FDB entry: %m");
+                log_link_message_warning_errno(link, m, r, "Could not add FDB entry");
                 link_enter_failed(link);
                 return 1;
         }

src/network/networkd-ipv4ll.c

@@ -51,7 +51,7 @@ static int ipv4ll_address_handler(sd_netlink *rtnl, sd_netlink_message *m, Link
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_error_errno(link, r, "could not set ipv4ll address: %m");
+                log_link_message_warning_errno(link, m, r, "could not set ipv4ll address");
                 link_enter_failed(link);
                 return 1;
         } else if (r >= 0)

src/network/networkd-ipv6-proxy-ndp.c

@@ -141,7 +141,7 @@ static int set_ipv6_proxy_ndp_address_handler(sd_netlink *rtnl, sd_netlink_messa
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST)
-                log_link_error_errno(link, r, "Could not add IPv6 proxy ndp address entry: %m");
+                log_link_message_warning_errno(link, m, r, "Could not add IPv6 proxy ndp address entry, ignoring");
 
         return 1;
 }

src/network/networkd-link-bus.c

@@ -41,14 +41,10 @@ static int property_get_bit_rates(
 
         manager = link->manager;
 
-        if (!manager->use_speed_meter)
-                return sd_bus_error_set(error, BUS_ERROR_SPEED_METER_INACTIVE, "Speed meter is disabled.");
-
-        if (manager->speed_meter_usec_old == 0)
-                return sd_bus_error_set(error, BUS_ERROR_SPEED_METER_INACTIVE, "Speed meter is not active.");
-
-        if (!link->stats_updated)
-                return sd_bus_error_set(error, BUS_ERROR_SPEED_METER_INACTIVE, "Failed to measure bit-rates.");
+        if (!manager->use_speed_meter ||
+            manager->speed_meter_usec_old == 0 ||
+            !link->stats_updated)
+                return sd_bus_message_append(reply, "(tt)", UINT64_MAX, UINT64_MAX);
 
         assert(manager->speed_meter_usec_new > manager->speed_meter_usec_old);
         interval_sec = (manager->speed_meter_usec_new - manager->speed_meter_usec_old) / USEC_PER_SEC;

src/network/networkd-link.c

@@ -929,7 +929,7 @@ static int nexthop_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link)
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_warning_errno(link, r, "Could not set nexthop: %m");
+                log_link_message_warning_errno(link, m, r, "Could not set nexthop");
                 link_enter_failed(link);
                 return 1;
         }
@@ -981,7 +981,7 @@ static int route_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link) {
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_warning_errno(link, r, "Could not set route: %m");
+                log_link_message_warning_errno(link, m, r, "Could not set route");
                 link_enter_failed(link);
                 return 1;
         }
@@ -1170,7 +1170,7 @@ static int address_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link)
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_warning_errno(link, r, "could not set address: %m");
+                log_link_message_warning_errno(link, m, r, "Could not set address");
                 link_enter_failed(link);
                 return 1;
         } else if (r >= 0)
@@ -1303,7 +1303,7 @@ static int set_mtu_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link)
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0)
-                log_link_warning_errno(link, r, "Could not set MTU, ignoring: %m");
+                log_link_message_warning_errno(link, m, r, "Could not set MTU, ignoring");
         else
                 log_link_debug(link, "Setting MTU done.");
 
@@ -1415,7 +1415,7 @@ static int set_flags_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0)
-                log_link_warning_errno(link, r, "Could not set link flags, ignoring: %m");
+                log_link_message_warning_errno(link, m, r, "Could not set link flags, ignoring");
 
         return 1;
 }
@@ -1582,7 +1582,7 @@ static int link_address_genmode_handler(sd_netlink *rtnl, sd_netlink_message *m,
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0)
-                log_link_warning_errno(link, r, "Could not set address genmode for interface, ignoring: %m");
+                log_link_message_warning_errno(link, m, r, "Could not set address genmode for interface, ignoring");
 
         return 1;
 }
@@ -1656,7 +1656,7 @@ static int link_up_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link)
         r = sd_netlink_message_get_errno(m);
         if (r < 0)
                 /* we warn but don't fail the link, as it may be brought up later */
-                log_link_warning_errno(link, r, "Could not bring up interface: %m");
+                log_link_message_warning_errno(link, m, r, "Could not bring up interface");
 
         return 1;
 }
@@ -1713,7 +1713,7 @@ static int link_down_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0)
-                log_link_warning_errno(link, r, "Could not bring down interface: %m");
+                log_link_message_warning_errno(link, m, r, "Could not bring down interface");
 
         return 1;
 }
@@ -2118,7 +2118,7 @@ static int netdev_join_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *li
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_error_errno(link, r, "Could not join netdev: %m");
+                log_link_message_warning_errno(link, m, r, "Could not join netdev");
                 link_enter_failed(link);
                 return 1;
         }
@@ -4053,3 +4053,10 @@ static const char* const link_state_table[_LINK_STATE_MAX] = {
 };
 
 DEFINE_STRING_TABLE_LOOKUP(link_state, LinkState);
+
+int log_link_message_full_errno(Link *link, sd_netlink_message *m, int level, int err, const char *msg) {
+        const char *err_msg = NULL;
+
+        (void) sd_netlink_message_read_string(m, NLMSGERR_ATTR_MSG, &err_msg);
+        return log_link_full(link, level, err, "%s: %s%s%m", msg, strempty(err_msg), err_msg ? " " : "");
+}

src/network/networkd-link.h

@@ -213,6 +213,13 @@ int link_request_set_nexthop(Link *link);
 
 int link_reconfigure(Link *link, bool force);
 
+int log_link_message_full_errno(Link *link, sd_netlink_message *m, int level, int err, const char *msg);
+#define log_link_message_error_errno(link, m, err, msg)   log_link_message_full_errno(link, m, LOG_ERR, err, msg)
+#define log_link_message_warning_errno(link, m, err, msg) log_link_message_full_errno(link, m, LOG_WARNING, err, msg)
+#define log_link_message_notice_errno(link, m, err, msg)  log_link_message_full_errno(link, m, LOG_NOTICE, err, msg)
+#define log_link_message_info_errno(link, m, err, msg)    log_link_message_full_errno(link, m, LOG_INFO, err, msg)
+#define log_link_message_debug_errno(link, m, err, msg)   log_link_message_full_errno(link, m, LOG_DEBUG, err, msg)
+
 #define ADDRESS_FMT_VAL(address)                   \
         be32toh((address).s_addr) >> 24,           \
         (be32toh((address).s_addr) >> 16) & 0xFFu, \

src/network/networkd-manager.c

@@ -41,6 +41,13 @@
 /* use 8 MB for receive socket kernel queue. */
 #define RCVBUF_SIZE    (8*1024*1024)
 
+static int log_message_warning_errno(sd_netlink_message *m, int err, const char *msg) {
+        const char *err_msg = NULL;
+
+        (void) sd_netlink_message_read_string(m, NLMSGERR_ATTR_MSG, &err_msg);
+        return log_warning_errno(err, "%s: %s%s%m", msg, strempty(err_msg), err_msg ? " " : "");
+}
+
 static int setup_default_address_pool(Manager *m) {
         AddressPool *p;
         int r;
@@ -283,7 +290,7 @@ int manager_rtnl_process_route(sd_netlink *rtnl, sd_netlink_message *message, vo
         if (sd_netlink_message_is_error(message)) {
                 r = sd_netlink_message_get_errno(message);
                 if (r < 0)
-                        log_warning_errno(r, "rtnl: failed to receive route message, ignoring: %m");
+                        log_message_warning_errno(message, r, "rtnl: failed to receive route message, ignoring");
 
                 return 0;
         }
@@ -576,7 +583,7 @@ int manager_rtnl_process_neighbor(sd_netlink *rtnl, sd_netlink_message *message,
         if (sd_netlink_message_is_error(message)) {
                 r = sd_netlink_message_get_errno(message);
                 if (r < 0)
-                        log_warning_errno(r, "rtnl: failed to receive neighbor message, ignoring: %m");
+                        log_message_warning_errno(message, r, "rtnl: failed to receive neighbor message, ignoring");
 
                 return 0;
         }
@@ -714,7 +721,7 @@ int manager_rtnl_process_address(sd_netlink *rtnl, sd_netlink_message *message,
         if (sd_netlink_message_is_error(message)) {
                 r = sd_netlink_message_get_errno(message);
                 if (r < 0)
-                        log_warning_errno(r, "rtnl: failed to receive address message, ignoring: %m");
+                        log_message_warning_errno(message, r, "rtnl: failed to receive address message, ignoring");
 
                 return 0;
         }
@@ -867,7 +874,7 @@ static int manager_rtnl_process_link(sd_netlink *rtnl, sd_netlink_message *messa
         if (sd_netlink_message_is_error(message)) {
                 r = sd_netlink_message_get_errno(message);
                 if (r < 0)
-                        log_warning_errno(r, "rtnl: Could not receive link message, ignoring: %m");
+                        log_message_warning_errno(message, r, "rtnl: Could not receive link message, ignoring");
 
                 return 0;
         }
@@ -957,7 +964,7 @@ int manager_rtnl_process_rule(sd_netlink *rtnl, sd_netlink_message *message, voi
         if (sd_netlink_message_is_error(message)) {
                 r = sd_netlink_message_get_errno(message);
                 if (r < 0)
-                        log_warning_errno(r, "rtnl: failed to receive rule message, ignoring: %m");
+                        log_message_warning_errno(message, r, "rtnl: failed to receive rule message, ignoring");
 
                 return 0;
         }
@@ -1170,7 +1177,7 @@ int manager_rtnl_process_nexthop(sd_netlink *rtnl, sd_netlink_message *message,
         if (sd_netlink_message_is_error(message)) {
                 r = sd_netlink_message_get_errno(message);
                 if (r < 0)
-                        log_warning_errno(r, "rtnl: failed to receive rule message, ignoring: %m");
+                        log_message_warning_errno(message, r, "rtnl: failed to receive rule message, ignoring");
 
                 return 0;
         }

src/network/networkd-ndisc.c

@@ -32,7 +32,7 @@ static int ndisc_netlink_route_message_handler(sd_netlink *rtnl, sd_netlink_mess
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_error_errno(link, r, "Could not set NDisc route or address: %m");
+                log_link_message_error_errno(link, m, r, "Could not set NDisc route or address");
                 link_enter_failed(link);
                 return 1;
         }
@@ -63,7 +63,7 @@ static int ndisc_netlink_address_message_handler(sd_netlink *rtnl, sd_netlink_me
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_error_errno(link, r, "Could not set NDisc route or address: %m");
+                log_link_message_error_errno(link, m, r, "Could not set NDisc route or address");
                 link_enter_failed(link);
                 return 1;
         } else if (r >= 0)

src/network/networkd-neighbor.c

@@ -102,7 +102,7 @@ static int neighbor_configure_handler(sd_netlink *rtnl, sd_netlink_message *m, L
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST)
                 /* Neighbor may not exist yet. So, do not enter failed state here. */
-                log_link_warning_errno(link, r, "Could not set neighbor, ignoring: %m");
+                log_link_message_warning_errno(link, m, r, "Could not set neighbor, ignoring");
 
         if (link->neighbor_messages == 0) {
                 log_link_debug(link, "Neighbors set");
@@ -171,7 +171,7 @@ static int neighbor_remove_handler(sd_netlink *rtnl, sd_netlink_message *m, Link
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -ESRCH)
                 /* Neighbor may not exist because it already got deleted, ignore that. */
-                log_link_warning_errno(link, r, "Could not remove neighbor: %m");
+                log_link_message_warning_errno(link, m, r, "Could not remove neighbor");
 
         return 1;
 }

src/network/networkd-nexthop.c

@@ -278,7 +278,7 @@ static int nexthop_remove_handler(sd_netlink *rtnl, sd_netlink_message *m, Link
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -ESRCH)
-                log_link_warning_errno(link, r, "Could not drop nexthop: %m");
+                log_link_message_warning_errno(link, m, r, "Could not drop nexthop, ignoring");
 
         return 1;
 }

src/network/networkd-route.c

@@ -398,7 +398,7 @@ static int route_remove_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *l
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -ESRCH)
-                log_link_warning_errno(link, r, "Could not drop route: %m");
+                log_link_message_warning_errno(link, m, r, "Could not drop route, ignoring");
 
         return 1;
 }

src/network/networkd-routing-policy-rule.c

@@ -317,7 +317,7 @@ static int routing_policy_rule_remove_handler(sd_netlink *rtnl, sd_netlink_messa
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0)
-                log_link_warning_errno(link, r, "Could not drop routing policy rule: %m");
+                log_link_message_warning_errno(link, m, r, "Could not drop routing policy rule");
 
         return 1;
 }
@@ -431,7 +431,7 @@ static int routing_policy_rule_handler(sd_netlink *rtnl, sd_netlink_message *m,
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_warning_errno(link, r, "Could not add routing policy rule: %m");
+                log_link_message_warning_errno(link, m, r, "Could not add routing policy rule");
                 link_enter_failed(link);
                 return 1;
         }

src/network/tc/qdisc.c

@@ -100,7 +100,7 @@ static int qdisc_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link) {
 
         r = sd_netlink_message_get_errno(m);
         if (r < 0 && r != -EEXIST) {
-                log_link_error_errno(link, r, "Could not set QDisc: %m");
+                log_link_message_error_errno(link, m, r, "Could not set QDisc: %m");
                 link_enter_failed(link);
                 return 1;
         }

src/resolve/meson.build

@@ -64,6 +64,8 @@ systemd_resolved_sources = files('''
         resolved-etc-hosts.h
         resolved-etc-hosts.c
         resolved-dnstls.h
+        resolved-util.c
+        resolved-util.h
 '''.split())
 
 resolvectl_sources = files('''
@@ -228,4 +230,10 @@ tests += [
          [],
          [],
          'ENABLE_RESOLVE', 'manual'],
+
+        [['src/resolve/test-resolved-util.c',
+          'src/resolve/resolved-util.c',
+          'src/resolve/resolved-util.h'],
+         [],
+         []],
 ]

src/resolve/resolved-conf.c

@@ -8,6 +8,7 @@
 #include "parse-util.h"
 #include "resolved-conf.h"
 #include "resolved-dnssd.h"
+#include "resolved-util.h"
 #include "specifier.h"
 #include "string-table.h"
 #include "string-util.h"
@@ -27,11 +28,12 @@ static int manager_add_dns_server_by_string(Manager *m, DnsServerType type, cons
         union in_addr_union address;
         int family, r, ifindex = 0;
         DnsServer *s;
+        _cleanup_free_ char *server_name = NULL;
 
         assert(m);
         assert(word);
 
-        r = in_addr_ifindex_from_string_auto(word, &family, &address, &ifindex);
+        r = in_addr_ifindex_name_from_string_auto(word, &family, &address, &ifindex, &server_name);
         if (r < 0)
                 return r;
 
@@ -52,7 +54,7 @@ static int manager_add_dns_server_by_string(Manager *m, DnsServerType type, cons
                 return 0;
         }
 
-        return dns_server_new(m, NULL, type, NULL, family, &address, ifindex);
+        return dns_server_new(m, NULL, type, NULL, family, &address, ifindex, server_name);
 }
 
 int manager_parse_dns_server_string_and_warn(Manager *m, DnsServerType type, const char *string) {

src/resolve/resolved-dns-server.c

@@ -25,8 +25,10 @@ int dns_server_new(
                 Link *l,
                 int family,
                 const union in_addr_union *in_addr,
-                int ifindex) {
+                int ifindex,
+                const char *server_name) {
 
+        _cleanup_free_ char *name = NULL;
         DnsServer *s;
 
         assert(m);
@@ -44,6 +46,12 @@ int dns_server_new(
                         return -E2BIG;
         }
 
+        if (server_name) {
+                name = strdup(server_name);
+                if (!name)
+                        return -ENOMEM;
+        }
+
         s = new(DnsServer, 1);
         if (!s)
                 return -ENOMEM;
@@ -55,6 +63,7 @@ int dns_server_new(
                 .family = family,
                 .address = *in_addr,
                 .ifindex = ifindex,
+                .server_name = TAKE_PTR(name),
         };
 
         dns_server_reset_features(s);
@@ -107,6 +116,7 @@ static DnsServer* dns_server_free(DnsServer *s)  {
 #endif
 
         free(s->server_string);
+        free(s->server_name);
         return mfree(s);
 }
 

src/resolve/resolved-dns-server.h

@@ -53,6 +53,8 @@ struct DnsServer {
 
         char *server_string;
 
+        char *server_name;
+
         /* The long-lived stream towards this server. */
         DnsStream *stream;
 
@@ -94,7 +96,8 @@ int dns_server_new(
                 Link *link,
                 int family,
                 const union in_addr_union *address,
-                int ifindex);
+                int ifindex,
+                const char *server_string);
 
 DnsServer* dns_server_ref(DnsServer *s);
 DnsServer* dns_server_unref(DnsServer *s);

src/resolve/resolved-dnstls-gnutls.c

@@ -67,6 +67,12 @@ int dnstls_stream_connect_tls(DnsStream *stream, DnsServer *server) {
                 gnutls_session_set_verify_cert2(gs, &stream->dnstls_data.validation, 1, 0);
         }
 
+        if (server->server_name) {
+                r = gnutls_server_name_set(gs, GNUTLS_NAME_DNS, server->server_name, strlen(server->server_name));
+                if (r < 0)
+                        return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Failed to set server name: %s", gnutls_strerror(r));
+        }
+
         gnutls_handshake_set_timeout(gs, GNUTLS_DEFAULT_HANDSHAKE_TIMEOUT);
 
         gnutls_transport_set_ptr2(gs, (gnutls_transport_ptr_t) (long) stream->fd, stream);

src/resolve/resolved-dnstls-openssl.c

@@ -87,6 +87,17 @@ int dnstls_stream_connect_tls(DnsStream *stream, DnsServer *server) {
                         return -ECONNREFUSED;
         }
 
+        if (server->server_name) {
+                r = SSL_set_tlsext_host_name(s, server->server_name);
+                if (r <= 0) {
+                        char errbuf[256];
+
+                        error = ERR_get_error();
+                        ERR_error_string_n(error, errbuf, sizeof(errbuf));
+                        return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Failed to set server name: %s", errbuf);
+                }
+        }
+
         ERR_clear_error();
         stream->dnstls_data.handshake = SSL_do_handshake(s);
         if (stream->dnstls_data.handshake <= 0) {

src/resolve/resolved-link-bus.c

@@ -284,7 +284,7 @@ int bus_link_method_set_dns_servers(sd_bus_message *message, void *userdata, sd_
                 if (s)
                         dns_server_move_back_and_unmark(s);
                 else {
-                        r = dns_server_new(l->manager, NULL, DNS_SERVER_LINK, l, dns[i].family, &dns[i].address, 0);
+                        r = dns_server_new(l->manager, NULL, DNS_SERVER_LINK, l, dns[i].family, &dns[i].address, 0, NULL);
                         if (r < 0)
                                 goto clear;
                 }

src/resolve/resolved-link.c

@@ -269,7 +269,7 @@ static int link_update_dns_server_one(Link *l, const char *name) {
                 return 0;
         }
 
-        return dns_server_new(l->manager, NULL, DNS_SERVER_LINK, l, family, &a, 0);
+        return dns_server_new(l->manager, NULL, DNS_SERVER_LINK, l, family, &a, 0, NULL);
 }
 
 static int link_update_dns_servers(Link *l) {

src/resolve/resolved-util.c

@@ -0,0 +1,36 @@
+/* SPDX-License-Identifier: LGPL-2.1+ */
+
+#include "alloc-util.h"
+#include "in-addr-util.h"
+#include "macro.h"
+#include "resolved-util.h"
+
+int in_addr_ifindex_name_from_string_auto(const char *s, int *family, union in_addr_union *ret, int *ifindex, char **server_name) {
+        _cleanup_free_ char *buf = NULL, *name = NULL;
+        const char *m;
+        int r;
+
+        assert(s);
+
+        m = strchr(s, '#');
+        if (m) {
+                name = strdup(m+1);
+                if (!name)
+                        return -ENOMEM;
+
+                buf = strndup(s, m - s);
+                if (!buf)
+                        return -ENOMEM;
+
+                s = buf;
+        }
+
+        r = in_addr_ifindex_from_string_auto(s, family, ret, ifindex);
+        if (r < 0)
+                return r;
+
+        if (server_name)
+                *server_name = TAKE_PTR(name);
+
+        return r;
+}

src/resolve/resolved-util.h

@@ -0,0 +1,6 @@
+/* SPDX-License-Identifier: LGPL-2.1+ */
+#pragma once
+
+#include "in-addr-util.h"
+
+int in_addr_ifindex_name_from_string_auto(const char *s, int *family, union in_addr_union *ret, int *ifindex, char **server_name);