sched/fair: fix kabi broken due to adding idle_h_nr_running in cfs_rq

hulk inclusion
category: bugfix
bugzilla: 38260, https://bugzilla.openeuler.org/show_bug.cgi?id=22
CVE: NA

---------------------------

Commit 92010bb6d6b8 ("sched/fair: Start tracking SCHED_IDLE tasks count in cfs_rq")
introduced kabi broken, just fix it.

Fixes: 92010bb6d6b8 ("sched/fair: Start tracking SCHED_IDLE tasks count in cfs_rq")
Signed-off-by: NCheng Jian <cj.chengjian@huawei.com>
Reviewed-by: NXie XiuQi <xiexiuqi@huawei.com>

kernel/sched/sched.h

@@ -490,7 +490,6 @@ struct cfs_rq {
 	unsigned long		runnable_weight;
 	unsigned int		nr_running;
 	unsigned int		h_nr_running;      /* SCHED_{NORMAL,BATCH,IDLE} */
-	unsigned int		idle_h_nr_running; /* SCHED_IDLE */
 
 	u64			exec_clock;
 	u64			min_vruntime;
@@ -574,7 +573,15 @@ struct cfs_rq {
 #endif /* CONFIG_CFS_BANDWIDTH */
 #endif /* CONFIG_FAIR_GROUP_SCHED */
 
+#ifndef __GENKSYMS__
+	union {
+		unsigned int		idle_h_nr_running; /* SCHED_IDLE */
+		unsigned long		idle_h_nr_running_padding; /* SCHED_IDLE padding for KABI */
+	};
+#else
 	KABI_RESERVE(1)
+#endif
+
 	KABI_RESERVE(2)
 };