Three new sd-event features + various fixes (#39608)

This avoids if conditions at the callsite which mess up stack based
lifetimes.

man/run0.xml

@@ -302,6 +302,15 @@
         <xi:include href="version-info.xml" xpointer="v259"/></listitem>
       </varlistentry>
 
+      <varlistentry>
+        <term><option>--same-root-dir</option></term>
+
+        <listitem><para>Execute the <command>run0</command> session in the same root directory that the
+        <command>run0</command> command is executed in.</para>
+
+        <xi:include href="version-info.xml" xpointer="v259"/></listitem>
+      </varlistentry>
+
       <varlistentry>
         <term><option>--machine=</option></term>
 

man/sd_event_add_child.xml

@@ -116,8 +116,9 @@
     parameter specifies the PID of the process to watch, which must be a direct child process of the invoking
     process. 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>WSTOPPED</constant> (watch for the child process being stopped by a signal),
-    <constant>WCONTINUED</constant> (watch for the child process being resumed by a signal). See
+    <constant>WCONTINUED</constant> (watch for the child process being resumed by a signal) and
+    <constant>WNOWAIT</constant> (Do not reap the child process after it exits). See
     <citerefentry project='man-pages'><refentrytitle>waitid</refentrytitle><manvolnum>2</manvolnum></citerefentry>
     for further information.</para>
 

man/sd_event_set_exit_on_idle.xml

@@ -0,0 +1,116 @@
+<?xml version='1.0'?>
+<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.5//EN"
+  "http://www.oasis-open.org/docbook/xml/4.5/docbookx.dtd">
+<!-- SPDX-License-Identifier: LGPL-2.1-or-later -->
+
+<refentry id="sd_event_set_exit_on_idle" xmlns:xi="http://www.w3.org/2001/XInclude">
+
+  <refentryinfo>
+    <title>sd_event_set_exit_on_idle</title>
+    <productname>systemd</productname>
+  </refentryinfo>
+
+  <refmeta>
+    <refentrytitle>sd_event_set_exit_on_idle</refentrytitle>
+    <manvolnum>3</manvolnum>
+  </refmeta>
+
+  <refnamediv>
+    <refname>sd_event_set_exit_on_idle</refname>
+    <refname>sd_event_get_exit_on_idle</refname>
+
+    <refpurpose>Enable event loop exit-on-idle support</refpurpose>
+  </refnamediv>
+
+  <refsynopsisdiv>
+    <funcsynopsis>
+      <funcsynopsisinfo>#include &lt;systemd/sd-event.h&gt;</funcsynopsisinfo>
+
+      <funcprototype>
+        <funcdef>int <function>sd_event_set_exit_on_idle</function></funcdef>
+        <paramdef>sd_event *<parameter>event</parameter></paramdef>
+        <paramdef>int b</paramdef>
+      </funcprototype>
+
+      <funcprototype>
+        <funcdef>int <function>sd_event_get_exit_on_idle</function></funcdef>
+        <paramdef>sd_event *<parameter>event</parameter></paramdef>
+      </funcprototype>
+
+    </funcsynopsis>
+  </refsynopsisdiv>
+
+  <refsect1>
+    <title>Description</title>
+
+    <para><function>sd_event_set_exit_on_idle()</function> may be used to
+    enable or disable the exit-on-idle support in the
+    event loop object specified in the <parameter>event</parameter>
+    parameter. If enabled, the event loop will exit with a zero exit code
+    there are no more enabled (<constant>SD_EVENT_ON</constant>, <constant>SD_EVENT_ONESHOT</constant>),
+    non-exit, non-post event sources.</para>
+
+    <para><function>sd_event_get_exit_on_idle()</function> may be used to
+    determine whether exit-on-idle support was previously requested by a
+    call to <function>sd_event_set_exit_on_idle()</function> with a true
+    <parameter>b</parameter> parameter and successfully enabled.</para>
+  </refsect1>
+
+  <refsect1>
+    <title>Return Value</title>
+
+    <para>On success, <function>sd_event_set_exit_on_idle()</function> and
+    <function>sd_event_get_exit_on_idle()</function> return a non-zero positive integer if the exit-on-idle
+    support was successfully enabled. They return zero if the exit-on-idle support was explicitly disabled
+    with a false <parameter>b</parameter> parameter. On failure, they return a negative errno-style error
+    code.</para>
+
+    <refsect2>
+      <title>Errors</title>
+
+      <para>Returned errors may indicate the following problems:</para>
+
+      <variablelist>
+
+        <varlistentry>
+          <term><constant>-ECHILD</constant></term>
+
+          <listitem><para>The event loop has been created in a different process, library or module instance.</para></listitem>
+        </varlistentry>
+
+        <varlistentry>
+          <term><constant>-EINVAL</constant></term>
+
+          <listitem><para>The passed event loop object was invalid.</para></listitem>
+        </varlistentry>
+
+      </variablelist>
+    </refsect2>
+  </refsect1>
+
+  <xi:include href="libsystemd-pkgconfig.xml" />
+
+  <refsect1>
+    <title>History</title>
+    <para><function>sd_event_set_exit_on_idle()</function> and
+    <function>sd_event_get_exit_on_idle()</function> were added in version 259.</para>
+  </refsect1>
+
+  <refsect1>
+    <title>See Also</title>
+
+    <para><simplelist type="inline">
+      <member><citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry></member>
+      <member><citerefentry><refentrytitle>sd-event</refentrytitle><manvolnum>3</manvolnum></citerefentry></member>
+      <member><citerefentry><refentrytitle>sd_event_new</refentrytitle><manvolnum>3</manvolnum></citerefentry></member>
+      <member><citerefentry><refentrytitle>sd_event_add_io</refentrytitle><manvolnum>3</manvolnum></citerefentry></member>
+      <member><citerefentry><refentrytitle>sd_event_add_time</refentrytitle><manvolnum>3</manvolnum></citerefentry></member>
+      <member><citerefentry><refentrytitle>sd_event_add_signal</refentrytitle><manvolnum>3</manvolnum></citerefentry></member>
+      <member><citerefentry><refentrytitle>sd_event_add_child</refentrytitle><manvolnum>3</manvolnum></citerefentry></member>
+      <member><citerefentry><refentrytitle>sd_event_add_inotify</refentrytitle><manvolnum>3</manvolnum></citerefentry></member>
+      <member><citerefentry><refentrytitle>sd_event_add_defer</refentrytitle><manvolnum>3</manvolnum></citerefentry></member>
+      <member><citerefentry><refentrytitle>systemd.service</refentrytitle><manvolnum>5</manvolnum></citerefentry></member>
+    </simplelist></para>
+  </refsect1>
+
+</refentry>

src/basic/basic-forward.h

@@ -94,6 +94,7 @@ typedef enum TimestampStyle TimestampStyle;
 typedef enum UnitActiveState UnitActiveState;
 typedef enum UnitDependency UnitDependency;
 typedef enum UnitType UnitType;
+typedef enum WaitFlags WaitFlags;
 
 typedef struct Hashmap Hashmap;
 typedef struct HashmapBase HashmapBase;

src/basic/log-context.c

@@ -58,7 +58,9 @@ static LogContext* log_context_detach(LogContext *c) {
 LogContext* log_context_new(const char *key, const char *value) {
         assert(key);
         assert(endswith(key, "="));
-        assert(value);
+
+        if (!value)
+                return NULL;
 
         LIST_FOREACH(ll, i, _log_context)
                 if (i->key == key && i->value == value)

src/libsystemd/libsystemd.sym

@@ -1081,5 +1081,7 @@ global:
 
 LIBSYSTEMD_259 {
 global:
+        sd_event_set_exit_on_idle;
+        sd_event_get_exit_on_idle;
         sd_varlink_is_connected;
 } LIBSYSTEMD_258;

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

@@ -48,7 +48,7 @@ static bool EVENT_SOURCE_WATCH_PIDFD(const 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.options == WEXITED;
+                (s->child.options & ~WNOWAIT) == WEXITED;
 }
 
 static bool event_source_is_online(sd_event_source *s) {
@@ -157,6 +157,7 @@ struct sd_event {
         bool need_process_child:1;
         bool watchdog:1;
         bool profile_delays:1;
+        bool exit_on_idle:1;
 
         int exit_code;
 
@@ -1583,7 +1584,7 @@ _public_ int sd_event_add_child(
         assert_return(e, -EINVAL);
         assert_return(e = event_resolve(e), -ENOPKG);
         assert_return(pid > 1, -EINVAL);
-        assert_return(!(options & ~(WEXITED|WSTOPPED|WCONTINUED)), -EINVAL);
+        assert_return(!(options & ~(WEXITED|WSTOPPED|WCONTINUED|WNOWAIT)), -EINVAL);
         assert_return(options != 0, -EINVAL);
         assert_return(e->state != SD_EVENT_FINISHED, -ESTALE);
         assert_return(!event_origin_changed(e), -ECHILD);
@@ -1675,7 +1676,7 @@ _public_ int sd_event_add_child_pidfd(
         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 & ~(WEXITED|WSTOPPED|WCONTINUED|WNOWAIT)), -EINVAL);
         assert_return(options != 0, -EINVAL);
         assert_return(e->state != SD_EVENT_FINISHED, -ESTALE);
         assert_return(!event_origin_changed(e), -ECHILD);
@@ -2987,9 +2988,11 @@ static int event_source_online(
                 break;
 
         case SOURCE_MEMORY_PRESSURE:
+                if (s->memory_pressure.write_buffer_size == 0) {
                         r = source_memory_pressure_register(s, enabled);
                         if (r < 0)
                                 return r;
+                }
 
                 break;
 
@@ -3042,16 +3045,8 @@ _public_ int sd_event_source_set_enabled(sd_event_source *s, int m) {
 
         if (m == SD_EVENT_OFF)
                 r = event_source_offline(s, m, s->ratelimited);
-        else {
+        else
-                if (s->enabled != SD_EVENT_OFF) {
-                        /* Switching from "on" to "oneshot" or back? If that's the case, we can take a shortcut, the
-                         * event source is already enabled after all. */
-                        s->enabled = m;
-                        return 0;
-                }
-
                 r = event_source_online(s, m, s->ratelimited);
-        }
         if (r < 0)
                 return r;
 
@@ -3695,7 +3690,7 @@ static int process_child(sd_event *e, int64_t threshold, int64_t *ret_min_priori
 
                 zero(s->child.siginfo);
                 if (waitid(P_PIDFD, s->child.pidfd, &s->child.siginfo,
-                           WNOHANG | (s->child.options & WEXITED ? WNOWAIT : 0) | s->child.options) < 0)
+                           WNOHANG | (s->child.options & WEXITED ? WNOWAIT : 0) | (s->child.options & ~WNOWAIT)) < 0)
                         return negative_errno();
 
                 if (s->child.siginfo.si_pid != 0) {
@@ -3743,7 +3738,6 @@ static int process_pidfd(sd_event *e, sd_event_source *s, uint32_t revents) {
         /* Note that pidfd would also generate EPOLLHUP when the process gets reaped. But at this point we
          * only permit EPOLLIN, under the assumption that upon EPOLLHUP the child source should already
          * be set to pending, and we would have returned early above. */
-        assert(!s->child.exited);
 
         zero(s->child.siginfo);
         if (waitid(P_PIDFD, s->child.pidfd, &s->child.siginfo, WNOHANG | WNOWAIT | s->child.options) < 0)
@@ -4083,6 +4077,22 @@ static int source_memory_pressure_initiate_dispatch(sd_event_source *s) {
         return 0; /* go on, dispatch to user callback */
 }
 
+static int mark_post_sources_pending(sd_event *e) {
+        sd_event_source *z;
+        int r;
+
+        SET_FOREACH(z, e->post_sources) {
+                if (event_source_is_offline(z))
+                        continue;
+
+                r = source_set_pending(z, true);
+                if (r < 0)
+                        return r;
+        }
+
+        return 0;
+}
+
 static int source_dispatch(sd_event_source *s) {
         EventSourceType saved_type;
         sd_event *saved_event;
@@ -4110,26 +4120,24 @@ static int source_dispatch(sd_event_source *s) {
                 return 1;
         }
 
-        if (!IN_SET(s->type, SOURCE_DEFER, SOURCE_EXIT)) {
+        if (IN_SET(s->type, SOURCE_DEFER, SOURCE_EXIT)) {
+                /* Make sure this event source is moved to the end of the priority list now. We do this here
+                 * because defer and exit event sources are always pending from the moment they're added so
+                 * the same logic in source_set_pending() is never triggered. */
+                s->pending_iteration = s->event->iteration;
+                event_source_pp_prioq_reshuffle(s);
+        } else {
                 r = source_set_pending(s, false);
                 if (r < 0)
                         return r;
         }
 
         if (s->type != SOURCE_POST) {
-                sd_event_source *z;
-
                 /* If we execute a non-post source, let's mark all post sources as pending. */
-
+                r = mark_post_sources_pending(s->event);
-                SET_FOREACH(z, s->event->post_sources) {
-                        if (event_source_is_offline(z))
-                                continue;
-
-                        r = source_set_pending(z, true);
                 if (r < 0)
                         return r;
         }
-        }
 
         if (s->type == SOURCE_MEMORY_PRESSURE) {
                 r = source_memory_pressure_initiate_dispatch(s);
@@ -4172,9 +4180,10 @@ static int source_dispatch(sd_event_source *s) {
 
                 r = s->child.callback(s, &s->child.siginfo, s->userdata);
 
-                /* Now, reap the PID for good. */
+                /* Now, reap the PID for good (unless WNOWAIT was specified by the caller). */
                 if (zombie) {
-                        (void) waitid(P_PIDFD, s->child.pidfd, &s->child.siginfo, WNOHANG|WEXITED);
+                        (void) waitid(P_PIDFD, s->child.pidfd, &s->child.siginfo, WNOHANG|WEXITED|(s->child.options & WNOWAIT));
+                        if (!FLAGS_SET(s->child.options, WNOWAIT))
                                 s->child.waited = true;
                 }
 
@@ -4237,6 +4246,14 @@ static int source_dispatch(sd_event_source *s) {
 
         s->dispatching = false;
 
+        if (saved_type != SOURCE_POST) {
+                /* More post sources might have been added while executing the callback, let's make sure
+                 * those are marked pending as well. */
+                r = mark_post_sources_pending(saved_event);
+                if (r < 0)
+                        return r;
+        }
+
 finish:
         if (r < 0) {
                 log_debug_errno(r, "Event source %s (type %s) returned error, %s: %m",
@@ -4412,6 +4429,28 @@ static int event_memory_pressure_write_list(sd_event *e) {
         return 0;
 }
 
+static bool event_loop_idle(sd_event *e) {
+        assert(e);
+
+        LIST_FOREACH(sources, s, e->sources) {
+                /* Exit sources only trigger on exit, so whether they're enabled or not doesn't matter when
+                 * we're deciding if the event loop is idle or not. */
+                if (s->type == SOURCE_EXIT)
+                        continue;
+
+                if (s->enabled == SD_EVENT_OFF)
+                        continue;
+
+                /* Post event sources always need another active event source to become pending. */
+                if (s->type == SOURCE_POST && !s->pending)
+                        continue;
+
+                return false;
+        }
+
+        return true;
+}
+
 _public_ int sd_event_prepare(sd_event *e) {
         int r;
 
@@ -4429,6 +4468,9 @@ _public_ int sd_event_prepare(sd_event *e) {
         /* Make sure that none of the preparation callbacks ends up freeing the event source under our feet */
         PROTECT_EVENT(e);
 
+        if (!e->exit_requested && e->exit_on_idle && event_loop_idle(e))
+                (void) sd_event_exit(e, 0);
+
         if (e->exit_requested)
                 goto pending;
 
@@ -5233,6 +5275,22 @@ _public_ int sd_event_set_signal_exit(sd_event *e, int b) {
         return change;
 }
 
+_public_ int sd_event_set_exit_on_idle(sd_event *e, int b) {
+        assert_return(e, -EINVAL);
+        assert_return(e = event_resolve(e), -ENOPKG);
+        assert_return(!event_origin_changed(e), -ECHILD);
+
+        return e->exit_on_idle = b;
+}
+
+_public_ int sd_event_get_exit_on_idle(sd_event *e) {
+        assert_return(e, -EINVAL);
+        assert_return(e = event_resolve(e), -ENOPKG);
+        assert_return(!event_origin_changed(e), -ECHILD);
+
+        return e->exit_on_idle;
+}
+
 _public_ int sd_event_source_set_memory_pressure_type(sd_event_source *s, const char *ty) {
         _cleanup_free_ char *b = NULL;
         _cleanup_free_ void *w = NULL;

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

@@ -946,4 +946,225 @@ TEST(leave_ratelimit) {
         ASSERT_TRUE(manually_left_ratelimit);
 }
 
+static int defer_post_handler(sd_event_source *s, void *userdata) {
+        bool *dispatched_post = ASSERT_PTR(userdata);
+
+        *dispatched_post = true;
+
+        return 0;
+}
+
+static int defer_adds_post_handler(sd_event_source *s, void *userdata) {
+        sd_event *e = sd_event_source_get_event(s);
+
+        /* Add a post event source from within the defer handler */
+        ASSERT_OK(sd_event_add_post(e, NULL, defer_post_handler, userdata));
+
+        return 0;
+}
+
+TEST(defer_add_post) {
+        _cleanup_(sd_event_unrefp) sd_event *e = NULL;
+        bool dispatched_post = false;
+
+        ASSERT_OK(sd_event_default(&e));
+
+        /* Add a oneshot defer event source that will add a post event source */
+        ASSERT_OK(sd_event_add_defer(e, NULL, defer_adds_post_handler, &dispatched_post));
+
+        /* Run one iteration - this should dispatch the defer handler */
+        ASSERT_OK_POSITIVE(sd_event_run(e, UINT64_MAX));
+
+        /* The post handler should have been added but not yet dispatched */
+        ASSERT_FALSE(dispatched_post);
+
+        /* Run another iteration - this should dispatch the post handler */
+        ASSERT_OK_POSITIVE(sd_event_run(e, 0));
+
+        /* Now the post handler should have been dispatched */
+        ASSERT_TRUE(dispatched_post);
+}
+
+static int child_handler_wnowait(sd_event_source *s, const siginfo_t *si, void *userdata) {
+        int *counter = ASSERT_PTR(userdata);
+
+        (*counter)++;
+
+        if (*counter == 5)
+                ASSERT_OK(sd_event_exit(sd_event_source_get_event(s), 0));
+
+        return 0;
+}
+
+TEST(child_wnowait) {
+        _cleanup_(sd_event_unrefp) sd_event *e = NULL;
+
+        ASSERT_OK(sigprocmask_many(SIG_BLOCK, NULL, SIGCHLD));
+
+        ASSERT_OK(sd_event_default(&e));
+
+        /* Fork a subprocess */
+        pid_t pid;
+        ASSERT_OK_ERRNO(pid = fork());
+
+        if (pid == 0)
+                /* Child process - exit with a specific code */
+                _exit(42);
+
+        /* Add a child source with WNOWAIT flag */
+        _cleanup_(sd_event_source_unrefp) sd_event_source *s = NULL;
+        int counter = 0;
+        ASSERT_OK(sd_event_add_child(e, &s, pid, WEXITED|WNOWAIT, child_handler_wnowait, &counter));
+        ASSERT_OK(sd_event_source_set_enabled(s, SD_EVENT_ON));
+
+        /* Run the event loop - this should call the handler */
+        ASSERT_OK(sd_event_loop(e));
+        ASSERT_EQ(counter, 5);
+
+        /* Since we used WNOWAIT, the child should still be waitable */
+        siginfo_t si = {};
+        ASSERT_OK_ERRNO(waitid(P_PID, pid, &si, WEXITED|WNOHANG));
+        ASSERT_EQ(si.si_pid, pid);
+        ASSERT_EQ(si.si_code, CLD_EXITED);
+        ASSERT_EQ(si.si_status, 42);
+}
+
+TEST(child_pidfd_wnowait) {
+        _cleanup_(sd_event_unrefp) sd_event *e = NULL;
+
+        ASSERT_OK(sigprocmask_many(SIG_BLOCK, NULL, SIGCHLD));
+
+        ASSERT_OK(sd_event_default(&e));
+
+        /* Fork a subprocess */
+        pid_t pid;
+        ASSERT_OK_ERRNO(pid = fork());
+
+        if (pid == 0)
+                /* Child process - exit with a specific code */
+                _exit(42);
+
+        _cleanup_close_ int pidfd = -EBADF;
+        ASSERT_OK_ERRNO(pidfd = pidfd_open(pid, 0));
+
+        /* Add a child source with WNOWAIT flag */
+        _cleanup_(sd_event_source_unrefp) sd_event_source *s = NULL;
+        int counter = 0;
+        ASSERT_OK(sd_event_add_child_pidfd(e, &s, pidfd, WEXITED|WNOWAIT, child_handler_wnowait, &counter));
+        ASSERT_OK(sd_event_source_set_enabled(s, SD_EVENT_ON));
+
+        /* Run the event loop - this should call the handler */
+        ASSERT_OK(sd_event_loop(e));
+        ASSERT_EQ(counter, 5);
+
+        /* Since we used WNOWAIT, the child should still be waitable */
+        siginfo_t si = {};
+        ASSERT_OK_ERRNO(waitid(P_PIDFD, pidfd, &si, WEXITED|WNOHANG));
+        ASSERT_EQ(si.si_pid, pid);
+        ASSERT_EQ(si.si_code, CLD_EXITED);
+        ASSERT_EQ(si.si_status, 42);
+}
+
+static int exit_on_idle_defer_handler(sd_event_source *s, void *userdata) {
+        unsigned *c = ASSERT_PTR(userdata);
+
+        /* Should not be reached on third call because the event loop should exit before */
+        ASSERT_LT(*c, 2u);
+
+        (*c)++;
+
+        /* Disable ourselves, which should trigger exit-on-idle after the second iteration */
+        if (*c == 2)
+                sd_event_source_set_enabled(s, SD_EVENT_OFF);
+
+        return 0;
+}
+
+static int exit_on_idle_post_handler(sd_event_source *s, void *userdata) {
+        unsigned *c = ASSERT_PTR(userdata);
+
+        /* Should not be reached on third call because the event loop should exit before */
+        ASSERT_LT(*c, 2u);
+
+        (*c)++;
+        return 0;
+}
+
+static int exit_on_idle_exit_handler(sd_event_source *s, void *userdata) {
+        return 0;
+}
+
+TEST(exit_on_idle) {
+        _cleanup_(sd_event_unrefp) sd_event *e = NULL;
+        ASSERT_OK(sd_event_new(&e));
+        ASSERT_OK(sd_event_set_exit_on_idle(e, true));
+        ASSERT_OK_POSITIVE(sd_event_get_exit_on_idle(e));
+
+        /* Create a recurring defer event source. */
+        _cleanup_(sd_event_source_unrefp) sd_event_source *d = NULL;
+        unsigned dc = 0;
+        ASSERT_OK(sd_event_add_defer(e, &d, exit_on_idle_defer_handler, &dc));
+        ASSERT_OK(sd_event_source_set_enabled(d, SD_EVENT_ON));
+
+        /* This post event source should not keep the event loop running after the defer source is disabled. */
+        _cleanup_(sd_event_source_unrefp) sd_event_source *p = NULL;
+        unsigned pc = 0;
+        ASSERT_OK(sd_event_add_post(e, &p, exit_on_idle_post_handler, &pc));
+        ASSERT_OK(sd_event_source_set_enabled(p, SD_EVENT_ON));
+        ASSERT_OK(sd_event_source_set_priority(p, SD_EVENT_PRIORITY_IMPORTANT));
+
+        /* And neither should this exit event source. */
+        ASSERT_OK(sd_event_add_exit(e, NULL, exit_on_idle_exit_handler, NULL));
+
+        /* Run the event loop - it should exit after we disable the event source */
+        ASSERT_OK(sd_event_loop(e));
+        ASSERT_EQ(dc, 2u);
+        ASSERT_EQ(pc, 2u);
+}
+
+TEST(exit_on_idle_no_sources) {
+        _cleanup_(sd_event_unrefp) sd_event *e = NULL;
+        ASSERT_OK(sd_event_new(&e));
+        ASSERT_OK(sd_event_set_exit_on_idle(e, true));
+
+        /* Running loop with no sources should return immediately with success */
+        ASSERT_OK(sd_event_loop(e));
+}
+
+static int defer_fair_handler(sd_event_source *s, void *userdata) {
+        unsigned *counter = ASSERT_PTR(userdata);
+
+        /* If we're about to increment above 5, exit the event loop */
+        if (*counter >= 5)
+                return sd_event_exit(sd_event_source_get_event(s), 0);
+
+        (*counter)++;
+
+        return 0;
+}
+
+TEST(defer_fair_scheduling) {
+        _cleanup_(sd_event_unrefp) sd_event *e = NULL;
+        sd_event_source *sources[5] = {};
+        unsigned counters[5] = {};
+
+        ASSERT_OK(sd_event_new(&e));
+        ASSERT_OK(sd_event_set_exit_on_idle(e, true));
+
+        /* Create 5 defer sources with equal priority */
+        for (unsigned i = 0; i < 5; i++) {
+                ASSERT_OK(sd_event_add_defer(e, &sources[i], defer_fair_handler, &counters[i]));
+                ASSERT_OK(sd_event_source_set_enabled(sources[i], SD_EVENT_ON));
+        }
+
+        /* Run the event loop until one of the handlers exits */
+        ASSERT_OK(sd_event_loop(e));
+
+        /* All counters should be equal to 5, demonstrating fair scheduling */
+        for (unsigned i = 0; i < 5; i++) {
+                ASSERT_EQ(counters[i], 5u);
+                sd_event_source_unref(sources[i]);
+        }
+}
+
 DEFINE_TEST_MAIN(LOG_DEBUG);

src/run/run.c

@@ -899,6 +899,7 @@ static int parse_argv_sudo_mode(int argc, char *argv[]) {
                 ARG_AREA,
                 ARG_VIA_SHELL,
                 ARG_EMPOWER,
+                ARG_SAME_ROOT_DIR,
         };
 
         /* If invoked as "run0" binary, let's expose a more sudo-like interface. We add various extensions
@@ -929,6 +930,7 @@ static int parse_argv_sudo_mode(int argc, char *argv[]) {
                 { "lightweight",         required_argument, NULL, ARG_LIGHTWEIGHT         },
                 { "area",                required_argument, NULL, ARG_AREA                },
                 { "empower",             no_argument,       NULL, ARG_EMPOWER             },
+                { "same-root-dir",       no_argument,       NULL, ARG_SAME_ROOT_DIR       },
                 {},
         };
 
@@ -1072,6 +1074,13 @@ static int parse_argv_sudo_mode(int argc, char *argv[]) {
                         arg_empower = true;
                         break;
 
+                case ARG_SAME_ROOT_DIR:
+                        r = free_and_strdup_warn(&arg_root_directory, "/");
+                        if (r < 0)
+                                return r;
+
+                        break;
+
                 case '?':
                         return -EINVAL;
 

src/systemd/sd-event.h

@@ -116,6 +116,8 @@ int sd_event_set_watchdog(sd_event *e, int b);
 int sd_event_get_watchdog(sd_event *e);
 int sd_event_get_iteration(sd_event *e, uint64_t *ret);
 int sd_event_set_signal_exit(sd_event *e, int b);
+int sd_event_set_exit_on_idle(sd_event *e, int b);
+int sd_event_get_exit_on_idle(sd_event *e);
 
 sd_event_source* sd_event_source_ref(sd_event_source *s);
 sd_event_source* sd_event_source_unref(sd_event_source *s);

src/test/test-cgroup-util.c

@@ -527,7 +527,9 @@ TEST(cgroupid) {
 
         fd2 = cg_cgroupid_open(fd, id);
 
-        if (ERRNO_IS_NEG_PRIVILEGE(fd2))
+        /* The kernel converts a bunch of errors to ESTALE in the open_by_handle_at() codepath so we treat
+         * it as missing privs but it could be absolutely anything really. */
+        if (ERRNO_IS_NEG_PRIVILEGE(fd2) || fd2 == -ESTALE)
                 log_notice("Skipping open-by-cgroup-id test because lacking privs.");
         else if (ERRNO_IS_NEG_NOT_SUPPORTED(fd2))
                 log_notice("Skipping open-by-cgroup-id test because syscall is missing or blocked.");

src/test/test-id128.c

@@ -8,6 +8,7 @@
 #include "sd-id128.h"
 
 #include "alloc-util.h"
+#include "capability-util.h"
 #include "fd-util.h"
 #include "id128-util.h"
 #include "path-util.h"
@@ -278,7 +279,7 @@ TEST(id128_at) {
         ASSERT_OK(sd_id128_randomize(&id));
 
         ASSERT_OK(id128_write_at(tfd, "etc/machine-id", ID128_FORMAT_PLAIN, id));
-        if (geteuid() == 0)
+        if (have_effective_cap(CAP_DAC_OVERRIDE))
                 ASSERT_OK(id128_write_at(tfd, "etc/machine-id", ID128_FORMAT_PLAIN, id));
         else
                 ASSERT_ERROR(id128_write_at(tfd, "etc/machine-id", ID128_FORMAT_PLAIN, id), EACCES);

src/test/test-rm-rf.c

@@ -3,6 +3,7 @@
 #include <sys/stat.h>
 #include <unistd.h>
 
+#include "capability-util.h"
 #include "process-util.h"
 #include "rm-rf.h"
 #include "string-util.h"
@@ -29,6 +30,7 @@ static void test_rm_rf_chmod_inner(void) {
         ASSERT_OK_ERRNO(chmod(x, 0500));
         ASSERT_OK_ERRNO(chmod(d, 0500));
 
+        if (!have_effective_cap(CAP_DAC_OVERRIDE))
                 ASSERT_ERROR(rm_rf(d, REMOVE_PHYSICAL), EACCES);
 
         ASSERT_OK_ERRNO(access(d, F_OK));