rcu: fix rcutorture behavior during reboot

Impact: fix very rare reboot hang Because rcutorture ignored all signals, it does not terminate in response to the signals sent at shutdown time. This can cause strange failures due to its continuing to make use of kernel function too late in the shutdown sequence. This patch therefore adds a shutdown notifier to rcutorture, causing it to shut down in response to a reboot or an orderly shutdown. Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>

rcu: fix rcutorture behavior during reboot
Impact: fix very rare reboot hang Because rcutorture ignored all signals, it does not terminate in response to the signals sent at shutdown time. This can cause strange failures due to its continuing to make use of kernel function too late in the shutdown sequence. This patch therefore adds a shutdown notifier to rcutorture, causing it to shut down in response to a reboot or an orderly shutdown. Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Paul E. McKenney · Ingo Molnar
1 parent 29cbda77a6
Showing 1 changed file with 56 additions and 10 deletions Side-by-side Diff
kernel/rcutorture.c
@@ -39,6 +39,7 @@
 #include <linux/moduleparam.h>
 #include <linux/percpu.h>
 #include <linux/notifier.h>
+#include <linux/reboot.h>
 #include <linux/freezer.h>
 #include <linux/cpu.h>
 #include <linux/delay.h>
@@ -108,7 +109,6 @@
 	int rtort_mbtest;
 };
  
-static int fullstop = 0;	/* stop generating callbacks at test end. */
 static LIST_HEAD(rcu_torture_freelist);
 static struct rcu_torture *rcu_torture_current = NULL;
 static long rcu_torture_current_version = 0;
  
@@ -136,7 +136,31 @@
 #endif
 int rcutorture_runnable = RCUTORTURE_RUNNABLE_INIT;
  
+#define FULLSTOP_SIGNALED 1	/* Bail due to signal. */
+#define FULLSTOP_CLEANUP  2	/* Orderly shutdown. */
+static int fullstop;		/* stop generating callbacks at test end. */
+DEFINE_MUTEX(fullstop_mutex);	/* protect fullstop transitions and */
+				/*  spawning of kthreads. */
+
 /*
+ * Detect and respond to a signal-based shutdown.
+ */
+static int
+rcutorture_shutdown_notify(struct notifier_block *unused1,
+			   unsigned long unused2, void *unused3)
+{
+	if (fullstop)
+		return NOTIFY_DONE;
+	if (signal_pending(current)) {
+		mutex_lock(&fullstop_mutex);
+		if (!ACCESS_ONCE(fullstop))
+			fullstop = FULLSTOP_SIGNALED;
+		mutex_unlock(&fullstop_mutex);
+	}
+	return NOTIFY_DONE;
+}
+
+/*
  * Allocate an element from the rcu_tortures pool.
  */
 static struct rcu_torture *
  
@@ -199,11 +223,12 @@
 static void
 rcu_stutter_wait(void)
 {
-	while (stutter_pause_test || !rcutorture_runnable)
+	while ((stutter_pause_test || !rcutorture_runnable) && !fullstop) {
 		if (rcutorture_runnable)
 			schedule_timeout_interruptible(1);
 		else
 			schedule_timeout_interruptible(round_jiffies_relative(HZ));
+	}
 }
  
 /*
@@ -599,7 +624,7 @@
 		rcu_stutter_wait();
 	} while (!kthread_should_stop() && !fullstop);
 	VERBOSE_PRINTK_STRING("rcu_torture_writer task stopping");
-	while (!kthread_should_stop())
+	while (!kthread_should_stop() && fullstop != FULLSTOP_SIGNALED)
 		schedule_timeout_uninterruptible(1);
 	return 0;
 }
@@ -624,7 +649,7 @@
 	} while (!kthread_should_stop() && !fullstop);
  
 	VERBOSE_PRINTK_STRING("rcu_torture_fakewriter task stopping");
-	while (!kthread_should_stop())
+	while (!kthread_should_stop() && fullstop != FULLSTOP_SIGNALED)
 		schedule_timeout_uninterruptible(1);
 	return 0;
 }
@@ -734,7 +759,7 @@
 	VERBOSE_PRINTK_STRING("rcu_torture_reader task stopping");
 	if (irqreader && cur_ops->irqcapable)
 		del_timer_sync(&t);
-	while (!kthread_should_stop())
+	while (!kthread_should_stop() && fullstop != FULLSTOP_SIGNALED)
 		schedule_timeout_uninterruptible(1);
 	return 0;
 }
@@ -831,7 +856,7 @@
 	do {
 		schedule_timeout_interruptible(stat_interval * HZ);
 		rcu_torture_stats_print();
-	} while (!kthread_should_stop());
+	} while (!kthread_should_stop() && !fullstop);
 	VERBOSE_PRINTK_STRING("rcu_torture_stats task stopping");
 	return 0;
 }
@@ -899,7 +924,7 @@
 	do {
 		schedule_timeout_interruptible(shuffle_interval * HZ);
 		rcu_torture_shuffle_tasks();
-	} while (!kthread_should_stop());
+	} while (!kthread_should_stop() && !fullstop);
 	VERBOSE_PRINTK_STRING("rcu_torture_shuffle task stopping");
 	return 0;
 }
  
@@ -914,10 +939,10 @@
 	do {
 		schedule_timeout_interruptible(stutter * HZ);
 		stutter_pause_test = 1;
-		if (!kthread_should_stop())
+		if (!kthread_should_stop() && !fullstop)
 			schedule_timeout_interruptible(stutter * HZ);
 		stutter_pause_test = 0;
-	} while (!kthread_should_stop());
+	} while (!kthread_should_stop() && !fullstop);
 	VERBOSE_PRINTK_STRING("rcu_torture_stutter task stopping");
 	return 0;
 }
  
@@ -934,12 +959,27 @@
 		stutter, irqreader);
 }
  
+static struct notifier_block rcutorture_nb = {
+	.notifier_call = rcutorture_shutdown_notify,
+};
+
 static void
 rcu_torture_cleanup(void)
 {
 	int i;
  
-	fullstop = 1;
+	mutex_lock(&fullstop_mutex);
+	if (!fullstop) {
+		/* If being signaled, let it happen, then exit. */
+		mutex_unlock(&fullstop_mutex);
+		schedule_timeout_interruptible(10 * HZ);
+		if (cur_ops->cb_barrier != NULL)
+			cur_ops->cb_barrier();
+		return;
+	}
+	fullstop = FULLSTOP_CLEANUP;
+	mutex_unlock(&fullstop_mutex);
+	unregister_reboot_notifier(&rcutorture_nb);
 	if (stutter_task) {
 		VERBOSE_PRINTK_STRING("Stopping rcu_torture_stutter task");
 		kthread_stop(stutter_task);
@@ -1015,6 +1055,8 @@
 		{ &rcu_ops, &rcu_sync_ops, &rcu_bh_ops, &rcu_bh_sync_ops,
 		  &srcu_ops, &sched_ops, &sched_ops_sync, };
  
+	mutex_lock(&fullstop_mutex);
+
 	/* Process args and tell the world that the torturer is on the job. */
 	for (i = 0; i < ARRAY_SIZE(torture_ops); i++) {
 		cur_ops = torture_ops[i];
@@ -1024,6 +1066,7 @@
 	if (i == ARRAY_SIZE(torture_ops)) {
 		printk(KERN_ALERT "rcutorture: invalid torture type: \"%s\"\n",
 		       torture_type);
+		mutex_unlock(&fullstop_mutex);
 		return (-EINVAL);
 	}
 	if (cur_ops->init)
  
@@ -1146,9 +1189,12 @@
 			goto unwind;
 		}
 	}
+	register_reboot_notifier(&rcutorture_nb);
+	mutex_unlock(&fullstop_mutex);
 	return 0;
  
 unwind:
+	mutex_unlock(&fullstop_mutex);
 	rcu_torture_cleanup();
 	return firsterr;
 }
...	...	@@ -39,6 +39,7 @@
39	39	#include <linux/moduleparam.h>
40	40	#include <linux/percpu.h>
41	41	#include <linux/notifier.h>
	42	+#include <linux/reboot.h>
42	43	#include <linux/freezer.h>
43	44	#include <linux/cpu.h>
44	45	#include <linux/delay.h>
...	...	@@ -108,7 +109,6 @@
108	109	int rtort_mbtest;
109	110	};
110	111
111		-static int fullstop = 0; /* stop generating callbacks at test end. */
112	112	static LIST_HEAD(rcu_torture_freelist);
113	113	static struct rcu_torture *rcu_torture_current = NULL;
114	114	static long rcu_torture_current_version = 0;
115	115
...	...	@@ -136,7 +136,31 @@
136	136	#endif
137	137	int rcutorture_runnable = RCUTORTURE_RUNNABLE_INIT;
138	138
	139	+#define FULLSTOP_SIGNALED 1 /* Bail due to signal. */
	140	+#define FULLSTOP_CLEANUP 2 /* Orderly shutdown. */
	141	+static int fullstop; /* stop generating callbacks at test end. */
	142	+DEFINE_MUTEX(fullstop_mutex); /* protect fullstop transitions and */
	143	+ /* spawning of kthreads. */
	144	+
139	145	/*
	146	+ * Detect and respond to a signal-based shutdown.
	147	+ */
	148	+static int
	149	+rcutorture_shutdown_notify(struct notifier_block *unused1,
	150	+ unsigned long unused2, void *unused3)
	151	+{
	152	+ if (fullstop)
	153	+ return NOTIFY_DONE;
	154	+ if (signal_pending(current)) {
	155	+ mutex_lock(&fullstop_mutex);
	156	+ if (!ACCESS_ONCE(fullstop))
	157	+ fullstop = FULLSTOP_SIGNALED;
	158	+ mutex_unlock(&fullstop_mutex);
	159	+ }
	160	+ return NOTIFY_DONE;
	161	+}
	162	+
	163	+/*
140	164	* Allocate an element from the rcu_tortures pool.
141	165	*/
142	166	static struct rcu_torture *
143	167
...	...	@@ -199,11 +223,12 @@
199	223	static void
200	224	rcu_stutter_wait(void)
201	225	{
202		- while (stutter_pause_test \|\| !rcutorture_runnable)
	226	+ while ((stutter_pause_test \|\| !rcutorture_runnable) && !fullstop) {
203	227	if (rcutorture_runnable)
204	228	schedule_timeout_interruptible(1);
205	229	else
206	230	schedule_timeout_interruptible(round_jiffies_relative(HZ));
	231	+ }
207	232	}
208	233
209	234	/*
...	...	@@ -599,7 +624,7 @@
599	624	rcu_stutter_wait();
600	625	} while (!kthread_should_stop() && !fullstop);
601	626	VERBOSE_PRINTK_STRING("rcu_torture_writer task stopping");
602		- while (!kthread_should_stop())
	627	+ while (!kthread_should_stop() && fullstop != FULLSTOP_SIGNALED)
603	628	schedule_timeout_uninterruptible(1);
604	629	return 0;
605	630	}
...	...	@@ -624,7 +649,7 @@
624	649	} while (!kthread_should_stop() && !fullstop);
625	650
626	651	VERBOSE_PRINTK_STRING("rcu_torture_fakewriter task stopping");
627		- while (!kthread_should_stop())
	652	+ while (!kthread_should_stop() && fullstop != FULLSTOP_SIGNALED)
628	653	schedule_timeout_uninterruptible(1);
629	654	return 0;
630	655	}
...	...	@@ -734,7 +759,7 @@
734	759	VERBOSE_PRINTK_STRING("rcu_torture_reader task stopping");
735	760	if (irqreader && cur_ops->irqcapable)
736	761	del_timer_sync(&t);
737		- while (!kthread_should_stop())
	762	+ while (!kthread_should_stop() && fullstop != FULLSTOP_SIGNALED)
738	763	schedule_timeout_uninterruptible(1);
739	764	return 0;
740	765	}
...	...	@@ -831,7 +856,7 @@
831	856	do {
832	857	schedule_timeout_interruptible(stat_interval * HZ);
833	858	rcu_torture_stats_print();
834		- } while (!kthread_should_stop());
	859	+ } while (!kthread_should_stop() && !fullstop);
835	860	VERBOSE_PRINTK_STRING("rcu_torture_stats task stopping");
836	861	return 0;
837	862	}
...	...	@@ -899,7 +924,7 @@
899	924	do {
900	925	schedule_timeout_interruptible(shuffle_interval * HZ);
901	926	rcu_torture_shuffle_tasks();
902		- } while (!kthread_should_stop());
	927	+ } while (!kthread_should_stop() && !fullstop);
903	928	VERBOSE_PRINTK_STRING("rcu_torture_shuffle task stopping");
904	929	return 0;
905	930	}
906	931
...	...	@@ -914,10 +939,10 @@
914	939	do {
915	940	schedule_timeout_interruptible(stutter * HZ);
916	941	stutter_pause_test = 1;
917		- if (!kthread_should_stop())
	942	+ if (!kthread_should_stop() && !fullstop)
918	943	schedule_timeout_interruptible(stutter * HZ);
919	944	stutter_pause_test = 0;
920		- } while (!kthread_should_stop());
	945	+ } while (!kthread_should_stop() && !fullstop);
921	946	VERBOSE_PRINTK_STRING("rcu_torture_stutter task stopping");
922	947	return 0;
923	948	}
924	949
...	...	@@ -934,12 +959,27 @@
934	959	stutter, irqreader);
935	960	}
936	961
	962	+static struct notifier_block rcutorture_nb = {
	963	+ .notifier_call = rcutorture_shutdown_notify,
	964	+};
	965	+
937	966	static void
938	967	rcu_torture_cleanup(void)
939	968	{
940	969	int i;
941	970
942		- fullstop = 1;
	971	+ mutex_lock(&fullstop_mutex);
	972	+ if (!fullstop) {
	973	+ /* If being signaled, let it happen, then exit. */
	974	+ mutex_unlock(&fullstop_mutex);
	975	+ schedule_timeout_interruptible(10 * HZ);
	976	+ if (cur_ops->cb_barrier != NULL)
	977	+ cur_ops->cb_barrier();
	978	+ return;
	979	+ }
	980	+ fullstop = FULLSTOP_CLEANUP;
	981	+ mutex_unlock(&fullstop_mutex);
	982	+ unregister_reboot_notifier(&rcutorture_nb);
943	983	if (stutter_task) {
944	984	VERBOSE_PRINTK_STRING("Stopping rcu_torture_stutter task");
945	985	kthread_stop(stutter_task);
...	...	@@ -1015,6 +1055,8 @@
1015	1055	{ &rcu_ops, &rcu_sync_ops, &rcu_bh_ops, &rcu_bh_sync_ops,
1016	1056	&srcu_ops, &sched_ops, &sched_ops_sync, };
1017	1057
	1058	+ mutex_lock(&fullstop_mutex);
	1059	+
1018	1060	/* Process args and tell the world that the torturer is on the job. */
1019	1061	for (i = 0; i < ARRAY_SIZE(torture_ops); i++) {
1020	1062	cur_ops = torture_ops[i];
...	...	@@ -1024,6 +1066,7 @@
1024	1066	if (i == ARRAY_SIZE(torture_ops)) {
1025	1067	printk(KERN_ALERT "rcutorture: invalid torture type: \"%s\"\n",
1026	1068	torture_type);
	1069	+ mutex_unlock(&fullstop_mutex);
1027	1070	return (-EINVAL);
1028	1071	}
1029	1072	if (cur_ops->init)
1030	1073
...	...	@@ -1146,9 +1189,12 @@
1146	1189	goto unwind;
1147	1190	}
1148	1191	}
	1192	+ register_reboot_notifier(&rcutorture_nb);
	1193	+ mutex_unlock(&fullstop_mutex);
1149	1194	return 0;
1150	1195
1151	1196	unwind:
	1197	+ mutex_unlock(&fullstop_mutex);
1152	1198	rcu_torture_cleanup();
1153	1199	return firsterr;
1154	1200	}