diff --git a/src/basic/limits-util.c b/src/basic/limits-util.c
index dc0a2a6c94..9597c4c41c 100644
--- a/src/basic/limits-util.c
+++ b/src/basic/limits-util.c
@@ -156,8 +156,11 @@ uint64_t system_tasks_max(void) {
         if (r < 0)
                 log_debug_errno(r, "Failed to determine cgroup root path, ignoring: %m");
         else {
+                /* We'll have the "pids.max" attribute on the our root cgroup only if we are in a
+                 * CLONE_NEWCGROUP namespace. On the top-level namespace this attribute is missing, hence
+                 * suppress any message about that */
                 r = cg_get_attribute_as_uint64("pids", root, "pids.max", &c);
-                if (r < 0)
+                if (r < 0 && r != -ENODATA)
                         log_debug_errno(r, "Failed to read pids.max attribute of root cgroup, ignoring: %m");
         }
 
diff --git a/src/core/bpf-lsm.c b/src/core/bpf-lsm.c
index 5f614ead04..22200bd939 100644
--- a/src/core/bpf-lsm.c
+++ b/src/core/bpf-lsm.c
@@ -228,11 +228,14 @@ int lsm_bpf_cleanup(const Unit *u) {
         if (!u->manager->restrict_fs)
                 return 0;
 
+        if (u->cgroup_id == 0)
+                return 0;
+
         int fd = sym_bpf_map__fd(u->manager->restrict_fs->maps.cgroup_hash);
         if (fd < 0)
                 return log_unit_error_errno(u, errno, "bpf-lsm: Failed to get BPF map fd: %m");
 
-        if (sym_bpf_map_delete_elem(fd, &u->cgroup_id) != 0)
+        if (sym_bpf_map_delete_elem(fd, &u->cgroup_id) != 0 && errno != ENOENT)
                 return log_unit_debug_errno(u, errno, "bpf-lsm: Failed to delete cgroup entry from LSM BPF map: %m");
 
         return 0;
diff --git a/src/shared/fdset.c b/src/shared/fdset.c
index f5dcdce218..b62f15c649 100644
--- a/src/shared/fdset.c
+++ b/src/shared/fdset.c
@@ -54,6 +54,8 @@ void fdset_close(FDSet *s) {
         void *p;
 
         while ((p = set_steal_first(MAKE_SET(s)))) {
+                int fd = PTR_TO_FD(p);
+
                 /* Valgrind's fd might have ended up in this set here, due to fdset_new_fill(). We'll ignore
                  * all failures here, so that the EBADFD that valgrind will return us on close() doesn't
                  * influence us */
@@ -62,8 +64,14 @@ void fdset_close(FDSet *s) {
                  * which has no effect at all, since they are only duplicates. So don't be surprised about
                  * these log messages. */
 
-                log_debug("Closing set fd %i", PTR_TO_FD(p));
-                (void) close_nointr(PTR_TO_FD(p));
+                if (DEBUG_LOGGING) {
+                        _cleanup_free_ char *path = NULL;
+
+                        (void) fd_get_path(fd, &path);
+                        log_debug("Closing set fd %i (%s)", fd, strna(path));
+                }
+
+                (void) close_nointr(fd);
         }
 }
 
diff --git a/src/shared/killall.c b/src/shared/killall.c
index 1011cd83d3..ef715dbdfe 100644
--- a/src/shared/killall.c
+++ b/src/shared/killall.c
@@ -60,7 +60,7 @@ static bool is_survivor_cgroup(pid_t pid) {
         /* user xattr support was added to kernel v5.7, try with the trusted namespace as a fallback */
         if (ERRNO_IS_NEG_XATTR_ABSENT(r))
                 r = cg_get_xattr_bool(cgroup_path, "trusted.survive_final_kill_signal");
-        if (r < 0)
+        if (r < 0 && !ERRNO_IS_NEG_XATTR_ABSENT(r))
                 log_debug_errno(r,
                                 "Failed to get survive_final_kill_signal xattr of %s, ignoring: %m",
                                 cgroup_path);