From 9b27811d49709b13f6255061089c9632572ee1ce Mon Sep 17 00:00:00 2001
From: Lennart Poettering <lennart@poettering.net>
Date: Mon, 27 May 2024 15:37:51 +0200
Subject: [PATCH] machined: GC machines during runtime too

One major omission in machine's logic so far was that the GC was only
run at startup and on the check-idle timeout, which is really slow.

Let's make this more like the GC logic in homed or logind: make sure we
run it in a close-by event loop cycle.
---
 src/machine/machine.c       |  2 ++
 src/machine/machined-core.c | 52 +++++++++++++++++++++++++++++++++++++
 src/machine/machined.c      | 27 ++-----------------
 src/machine/machined.h      |  5 ++++
 4 files changed, 61 insertions(+), 25 deletions(-)

diff --git a/src/machine/machine.c b/src/machine/machine.c
index 1ef2be27ab..25ecc4d197 100644
--- a/src/machine/machine.c
+++ b/src/machine/machine.c
@@ -592,6 +592,8 @@ void machine_add_to_gc_queue(Machine *m) {
 
         LIST_PREPEND(gc_queue, m->manager->machine_gc_queue, m);
         m->in_gc_queue = true;
+
+        manager_enqueue_gc(m->manager);
 }
 
 MachineState machine_get_state(Machine *s) {
diff --git a/src/machine/machined-core.c b/src/machine/machined-core.c
index ffca209494..a7e1853b53 100644
--- a/src/machine/machined-core.c
+++ b/src/machine/machined-core.c
@@ -104,3 +104,55 @@ int manager_find_machine_for_gid(Manager *m, gid_t gid, Machine **ret_machine, g
 
         return false;
 }
+
+void manager_gc(Manager *m, bool drop_not_started) {
+        Machine *machine;
+
+        assert(m);
+
+        while ((machine = LIST_POP(gc_queue, m->machine_gc_queue))) {
+                machine->in_gc_queue = false;
+
+                /* First, if we are not closing yet, initiate stopping */
+                if (machine_may_gc(machine, drop_not_started) &&
+                    machine_get_state(machine) != MACHINE_CLOSING)
+                        machine_stop(machine);
+
+                /* Now, the stop probably made this referenced
+                 * again, but if it didn't, then it's time to let it
+                 * go entirely. */
+                if (machine_may_gc(machine, drop_not_started)) {
+                        machine_finalize(machine);
+                        machine_free(machine);
+                }
+        }
+}
+
+static int on_deferred_gc(sd_event_source *s, void *userdata) {
+        manager_gc(userdata, /* drop_not_started= */ true);
+        return 0;
+}
+
+void manager_enqueue_gc(Manager *m) {
+        int r;
+
+        assert(m);
+
+        if (m->deferred_gc_event_source) {
+                r = sd_event_source_set_enabled(m->deferred_gc_event_source, SD_EVENT_ONESHOT);
+                if (r < 0)
+                        log_warning_errno(r, "Failed to enable GC event source, ignoring: %m");
+
+                return;
+        }
+
+        r = sd_event_add_defer(m->event, &m->deferred_gc_event_source, on_deferred_gc, m);
+        if (r < 0)
+                return (void) log_warning_errno(r, "Failed to allocate GC event source, ignoring: %m");
+
+        r = sd_event_source_set_priority(m->deferred_gc_event_source, SD_EVENT_PRIORITY_IDLE);
+        if (r < 0)
+                log_warning_errno(r, "Failed to tweak priority of event source, ignoring: %m");
+
+        (void) sd_event_source_set_description(m->deferred_gc_event_source, "deferred-gc");
+}
diff --git a/src/machine/machined.c b/src/machine/machined.c
index 60badbf5d0..4f2ca626e2 100644
--- a/src/machine/machined.c
+++ b/src/machine/machined.c
@@ -97,6 +97,8 @@ static Manager* manager_unref(Manager *m) {
         sd_event_source_unref(m->nscd_cache_flush_event);
 #endif
 
+        sd_event_source_disable_unref(m->deferred_gc_event_source);
+
         hashmap_free(m->polkit_registry);
 
         manager_varlink_done(m);
@@ -264,29 +266,6 @@ static int manager_connect_bus(Manager *m) {
         return 0;
 }
 
-static void manager_gc(Manager *m, bool drop_not_started) {
-        Machine *machine;
-
-        assert(m);
-
-        while ((machine = LIST_POP(gc_queue, m->machine_gc_queue))) {
-                machine->in_gc_queue = false;
-
-                /* First, if we are not closing yet, initiate stopping */
-                if (machine_may_gc(machine, drop_not_started) &&
-                    machine_get_state(machine) != MACHINE_CLOSING)
-                        machine_stop(machine);
-
-                /* Now, the stop probably made this referenced
-                 * again, but if it didn't, then it's time to let it
-                 * go entirely. */
-                if (machine_may_gc(machine, drop_not_started)) {
-                        machine_finalize(machine);
-                        machine_free(machine);
-                }
-        }
-}
-
 static int manager_startup(Manager *m) {
         Machine *machine;
         int r;
@@ -331,8 +310,6 @@ static bool check_idle(void *userdata) {
         if (!hashmap_isempty(m->polkit_registry))
                 return false;
 
-        manager_gc(m, true);
-
         return hashmap_isempty(m->machines);
 }
 
diff --git a/src/machine/machined.h b/src/machine/machined.h
index 67abed0fd6..4c22382f2a 100644
--- a/src/machine/machined.h
+++ b/src/machine/machined.h
@@ -24,6 +24,8 @@ struct Manager {
         Hashmap *machine_units;
         Hashmap *machine_leaders;
 
+        sd_event_source *deferred_gc_event_source;
+
         Hashmap *polkit_registry;
 
         Hashmap *image_cache;
@@ -68,3 +70,6 @@ static inline void manager_enqueue_nscd_cache_flush(Manager *m) {}
 
 int manager_find_machine_for_uid(Manager *m, uid_t host_uid, Machine **ret_machine, uid_t *ret_internal_uid);
 int manager_find_machine_for_gid(Manager *m, gid_t host_gid, Machine **ret_machine, gid_t *ret_internal_gid);
+
+void manager_gc(Manager *m, bool drop_not_started);
+void manager_enqueue_gc(Manager *m);