dcavalca / rpms / systemd

Forked from rpms/systemd 3 months ago
Clone
Blob Blame History Raw
From 25b93538eba0275d35ef4b0792c2cd63d63d5e8d Mon Sep 17 00:00:00 2001
From: Franck Bui <fbui@suse.com>
Date: Tue, 19 Mar 2019 10:59:26 +0100
Subject: [PATCH] core: only watch processes when it's really necessary
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

If we know that main pid is our child then it's unnecessary to watch all
other processes of a unit since in this case we will get SIGCHLD when the main
process will exit and will act upon accordingly.

So let's watch all processes only if the main process is not our child since in
this case we need to detect when the cgroup will become empty in order to
figure out when the service becomes dead. This is only needed by cgroupv1.

Thanks Renaud Métrich for backporting this to RHEL.
Resolves: #1744972
---
 src/core/service.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/src/core/service.c b/src/core/service.c
index 310838a5f6..b1ec52d220 100644
--- a/src/core/service.c
+++ b/src/core/service.c
@@ -3410,8 +3410,7 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) {
                                 if (main_pid_good(s) <= 0)
                                         service_enter_stop_post(s, f);
 
-                                /* If there is still a service
-                                 * process around, wait until
+                                /* If there is still a service process around, wait until
                                  * that one quit, too */
                                 break;
 
@@ -3433,10 +3432,14 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) {
         if (notify_dbus)
                 unit_add_to_dbus_queue(u);
 
-        /* If we get a SIGCHLD event for one of the processes we were interested in, then we look for others to watch,
-         * under the assumption that we'll sooner or later get a SIGCHLD for them, as the original process we watched
-         * was probably the parent of them, and they are hence now our children. */
-        (void) unit_enqueue_rewatch_pids(u);
+        /* We watch the main/control process otherwise we can't retrieve the unit they
+         * belong to with cgroupv1. But if they are not our direct child, we won't get a
+         * SIGCHLD for them. Therefore we need to look for others to watch so we can
+         * detect when the cgroup becomes empty. Note that the control process is always
+         * our child so it's pointless to watch all other processes. */
+        if (!control_pid_good(s))
+                if (!s->main_pid_known || s->main_pid_alien)
+                        (void) unit_enqueue_rewatch_pids(u);
 }
 
 static int service_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata) {