cgroup-util: Properly handle conditions where cgroup.threads is empty after SIGKILL but processes still remain

After sending a SIGKILL to a process, the process might disappear from
`cgroup.threads` but still show up in `cgroup.procs` and still remains in the
cgroup and cause migrating new processes to `Delegate=yes` cgroups to fail with
`-EBUSY`. This is especially likely for heavyweight processes that consume more
kernel CPU time to clean up.

Fix this by only returning 0 when both `cgroup.threads` and
`cgroup.procs` are empty.

1 files changed, 12 insertions, 3 deletions

diff --git a/src/basic/cgroup-util.c b/src/basic/cgroup-util.c
index 95bf177a6b..b03cc70e2e 100644
--- a/src/basic/cgroup-util.c
+++ b/src/basic/cgroup-util.c
@@ -357,20 +357,29 @@ int cg_kill(
                 Set *s,
                 cg_kill_log_func_t log_kill,
                 void *userdata) {
-        int r;
+
+        int r, ret;
 
         r = cg_kill_items(controller, path, sig, flags, s, log_kill, userdata, "cgroup.procs");
         if (r < 0 || sig != SIGKILL)
                 return r;
 
+        ret = r;
+
         /* Only in case of killing with SIGKILL and when using cgroupsv2, kill remaining threads manually as
            a workaround for kernel bug. It was fixed in 5.2-rc5 (c03cd7738a83), backported to 4.19.66
            (4340d175b898) and 4.14.138 (feb6b123b7dd). */
         r = cg_unified_controller(controller);
-        if (r <= 0)
+        if (r < 0)
+                return r;
+        if (r == 0)
+                return ret;
+
+        r = cg_kill_items(controller, path, sig, flags, s, log_kill, userdata, "cgroup.threads");
+        if (r < 0)
                 return r;
 
-        return cg_kill_items(controller, path, sig, flags, s, log_kill, userdata, "cgroup.threads");
+        return r > 0 || ret > 0;
 }
 
 int cg_kill_kernel_sigkill(const char *controller, const char *path) {