RFC: hung task: Check specific tasks for long uninterruptible sleep state

Hi, Imran,

I think a "unmonitored list" is better than "monitor list", because we want khungtaskd can find out the "unexpected" hung task, but not few in a list. Then, for the fg tasks, which can put it in the "unmonitored list", for the bg tasks, I think we can tweak the timeout to control the duration.

Thanks!

BR, Shile

-----Original Message-----
From: Imran Khan [mailto:[email protected]]
Sent: Monday, August 21, 2017 6:26 PM
To: [email protected]
Cc: [email protected]; Imran Khan <[email protected]>; Luis R. Rodriguez <[email protected]>; Kees Cook <[email protected]>; Peter Zijlstra (Intel) <[email protected]>; Zhang, Shile (NSB - CN/Hangzhou) <[email protected]>; Matt Fleming <[email protected]>; Andrew Morton <[email protected]>; Vegard Nossum <[email protected]>; Tetsuo Handa <[email protected]>; John Siddle <[email protected]>; open list <[email protected]>; open list:PROC SYSCTL <[email protected]>
Subject: [PATCH] RFC: hung task: Check specific tasks for long uninterruptible sleep state

khungtask by default monitors either all tasks or no tasks at all for long unterruptible sleeps.
For Android like environments this arrangement is not optimal because on one hand it may be permissible to have some background(bg) task in uninterruptible sleep state for long duration while on the other hand it may not be permissible to have some foreground(fg) task like surfaceflinger in uninterruptible sleep state for long duration. So it would be good to have some arrangement so that few specified tasks can be monitored by khungtaskd, on a need basis.
This change introduces a sysctl option, /proc/sys/kernel/ hung_task_check_selected, to enable monitoring of selected tasks using khungtask daemon. If this sysctl option is enabled then only the tasks specified in /proc/hung_task_monitor_list are monitored otherwise all tasks are monitored, just like the default case.

Signed-off-by Imran Khan <[email protected]>

---
 include/linux/sched/sysctl.h |   1 +
 kernel/hung_task.c           | 121 ++++++++++++++++++++++++++++++++++++++++++-
 kernel/sysctl.c              |   8 +++
 3 files changed, 128 insertions(+), 2 deletions(-)

include/linux/sched/sysctl.h +1 -0

index 0f5ecd4..05892f1 100644
@@ -10,6 +10,7 @@
 extern unsigned int  sysctl_hung_task_panic;
 extern unsigned long sysctl_hung_task_timeout_secs;
 extern int sysctl_hung_task_warnings;
+extern int sysctl_hung_task_check_selected;
 extern int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,
 					 void __user *buffer,
 					 size_t *lenp, loff_t *ppos);

kernel/hung_task.c +119 -2

index 751593e..49f13fb 100644
@@ -16,12 +16,28 @@
 #include <linux/export.h>
 #include <linux/sysctl.h>
 #include <linux/utsname.h>
+#include <linux/uaccess.h>
+#include <linux/slab.h>
+#include <linux/proc_fs.h>
 #include <linux/sched/signal.h>
 #include <linux/sched/debug.h>
 
 #include <trace/events/sched.h>
 
 /*
+ * Hung task that needs monitoring
+ */
+struct hung_task {
+	struct list_head list;
+	char comm[TASK_COMM_LEN];
+};
+
+static struct hung_task	*monitor_list;
+int sysctl_hung_task_check_selected;
+
+
+
+/*
  * The number of tasks checked:
  */
 int __read_mostly sysctl_hung_task_check_count = PID_MAX_LIMIT;

@@ -76,6 +92,92 @@ static int __init hung_task_panic_setup(char *str)
 	.notifier_call = hung_task_panic,
 };
 
+static void hung_task_monitor_setup(void)
+{
+	monitor_list = kmalloc(sizeof(*monitor_list), GFP_KERNEL);
+	if (monitor_list) {
+		INIT_LIST_HEAD(&monitor_list->list);
+		memset(monitor_list->comm, 0, TASK_COMM_LEN);
+	}
+}
+
+
+static  int hung_task_info_show(struct seq_file *m, void *v)
+{
+	struct hung_task *ht;
+
+	ht = list_entry(v, struct hung_task, list);
+	seq_puts(m, ht->comm);
+
+	return 0;
+}
+
+static void *hung_task_info_start(struct seq_file *m, loff_t *pos)
+{
+	return seq_list_start_head(&monitor_list->list, *pos);
+}
+
+static void *hung_task_info_next(struct seq_file *m, void *v, loff_t *pos)
+{
+	return seq_list_next(v, &monitor_list->list, pos);
+}
+
+static void hung_task_info_stop(struct seq_file *m, void *v)
+{
+}
+
+const struct seq_operations hung_task_info_op = {
+	.start	= hung_task_info_start,
+	.next	= hung_task_info_next,
+	.stop	= hung_task_info_stop,
+	.show	= hung_task_info_show
+};
+
+static int hung_task_info_open(struct inode *inode, struct file *file)
+{
+	return seq_open(file, &hung_task_info_op);
+}
+
+static ssize_t
+hung_task_info_write(struct file *file, const char __user *buf, size_t count,
+	     loff_t *offs)
+{
+	struct task_struct *g, *t;
+	struct hung_task *ht = kmalloc(sizeof(*ht), GFP_KERNEL);
+
+	if (!ht)
+		return -ENOMEM;
+
+	if (copy_from_user(ht->comm, buf, count))
+		return -EFAULT;
+	ht->comm[count] = '\0';
+
+	for_each_process_thread(g, t) {
+		if (!strncmp(t->comm, ht->comm, strlen(t->comm))) {
+			list_add_tail(&ht->list, &monitor_list->list);
+			return count;
+		}
+	}
+
+	pr_err("Non-existing task: %s can't be monitored\n", ht->comm);
+	return count;
+}
+
+static const struct file_operations hung_task_info_operations = {
+	.open		= hung_task_info_open,
+	.read		= seq_read,
+	.write		= hung_task_info_write,
+	.llseek		= seq_lseek,
+	.release	= seq_release,
+};
+
+static int __init proc_hung_task_info_init(void)
+{
+	proc_create("hung_task_monitor_list", 0644, NULL,
+		    &hung_task_info_operations);
+	return 0;
+}
+
 static void check_hung_task(struct task_struct *t, unsigned long timeout)
 {
 	unsigned long switch_count = t->nvcsw + t->nivcsw;

@@ -167,6 +269,7 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout)
 	int max_count = sysctl_hung_task_check_count;
 	int batch_count = HUNG_TASK_BATCHING;
 	struct task_struct *g, *t;
+	struct hung_task *ht, *tmp;
 
 	/*
 	 * If the system crashed already then all bets are off,

@@ -186,9 +289,21 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout)
 				goto unlock;
 		}
 		/* use "==" to skip the TASK_KILLABLE tasks waiting on NFS */
-		if (t->state == TASK_UNINTERRUPTIBLE)
-			check_hung_task(t, timeout);
+		if (t->state == TASK_UNINTERRUPTIBLE) {
+			if (sysctl_hung_task_check_selected) {
+				list_for_each_entry_safe(ht, tmp,
+							 &monitor_list->list,
+							 list)
+					if (!strncmp(ht->comm, t->comm,
+					    strlen(t->comm)))
+					/* Task belongs to the selected group */
+						check_hung_task(t, timeout);
+			} else {
+				check_hung_task(t, timeout);
+			}
+		}
 	}
+
  unlock:
 	rcu_read_unlock();
 	if (hung_task_show_lock)

@@ -259,6 +374,8 @@ static int watchdog(void *dummy)
 static int __init hung_task_init(void)
 {
 	atomic_notifier_chain_register(&panic_notifier_list, &panic_block);
+	hung_task_monitor_setup();
+	proc_hung_task_info_init();
 	watchdog_task = kthread_run(watchdog, NULL, "khungtaskd");
 
 	return 0;

kernel/sysctl.c +8 -0

index 6648fbb..ab45774 100644
@@ -1096,6 +1096,14 @@ static int sysrq_sysctl_handler(struct ctl_table *table, int write,
 		.proc_handler	= proc_dointvec_minmax,
 		.extra1		= &neg_one,
 	},
+	{
+		.procname	= "hung_task_check_selected",
+		.data		= &sysctl_hung_task_check_selected,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &neg_one,
+	},
 #endif
 #ifdef CONFIG_RT_MUTEXES
 	{

Hi, Imran,

I think a "unmonitored list" is better than "monitor list", because we want khungtaskd can find out the "unexpected" hung task, but not few in a list. Then, for the fg tasks, which can put it in the "unmonitored list", for the bg tasks, I think we can tweak the timeout to control the duration.

Thanks!

BR, Shile

-----Original Message-----
From: Imran Khan [mailto:[email protected]]
Sent: Monday, August 21, 2017 6:26 PM
To: [email protected]
Cc: [email protected]; Imran Khan <[email protected]>; Luis R. Rodriguez <[email protected]>; Kees Cook <[email protected]>; Peter Zijlstra (Intel) <[email protected]>; Zhang, Shile (NSB - CN/Hangzhou) <[email protected]>; Matt Fleming <[email protected]>; Andrew Morton <[email protected]>; Vegard Nossum <[email protected]>; Tetsuo Handa <[email protected]>; John Siddle <[email protected]>; open list <[email protected]>; open list:PROC SYSCTL <[email protected]>
Subject: [PATCH] RFC: hung task: Check specific tasks for long uninterruptible sleep state

khungtask by default monitors either all tasks or no tasks at all for long unterruptible sleeps.
For Android like environments this arrangement is not optimal because on one hand it may be permissible to have some background(bg) task in uninterruptible sleep state for long duration while on the other hand it may not be permissible to have some foreground(fg) task like surfaceflinger in uninterruptible sleep state for long duration. So it would be good to have some arrangement so that few specified tasks can be monitored by khungtaskd, on a need basis.
This change introduces a sysctl option, /proc/sys/kernel/ hung_task_check_selected, to enable monitoring of selected tasks using khungtask daemon. If this sysctl option is enabled then only the tasks specified in /proc/hung_task_monitor_list are monitored otherwise all tasks are monitored, just like the default case.

Signed-off-by Imran Khan <[email protected]>

---
 include/linux/sched/sysctl.h |   1 +
 kernel/hung_task.c           | 121 ++++++++++++++++++++++++++++++++++++++++++-
 kernel/sysctl.c              |   8 +++
 3 files changed, 128 insertions(+), 2 deletions(-)

include/linux/sched/sysctl.h +1 -0

index 0f5ecd4..05892f1 100644
@@ -10,6 +10,7 @@
 extern unsigned int  sysctl_hung_task_panic;
 extern unsigned long sysctl_hung_task_timeout_secs;
 extern int sysctl_hung_task_warnings;
+extern int sysctl_hung_task_check_selected;
 extern int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,
 					 void __user *buffer,
 					 size_t *lenp, loff_t *ppos);

kernel/hung_task.c +119 -2

index 751593e..49f13fb 100644
@@ -16,12 +16,28 @@
 #include <linux/export.h>
 #include <linux/sysctl.h>
 #include <linux/utsname.h>
+#include <linux/uaccess.h>
+#include <linux/slab.h>
+#include <linux/proc_fs.h>
 #include <linux/sched/signal.h>
 #include <linux/sched/debug.h>
 
 #include <trace/events/sched.h>
 
 /*
+ * Hung task that needs monitoring
+ */
+struct hung_task {
+	struct list_head list;
+	char comm[TASK_COMM_LEN];
+};
+
+static struct hung_task	*monitor_list;
+int sysctl_hung_task_check_selected;
+
+
+
+/*
  * The number of tasks checked:
  */
 int __read_mostly sysctl_hung_task_check_count = PID_MAX_LIMIT;

@@ -76,6 +92,92 @@ static int __init hung_task_panic_setup(char *str)
 	.notifier_call = hung_task_panic,
 };
 
+static void hung_task_monitor_setup(void)
+{
+	monitor_list = kmalloc(sizeof(*monitor_list), GFP_KERNEL);
+	if (monitor_list) {
+		INIT_LIST_HEAD(&monitor_list->list);
+		memset(monitor_list->comm, 0, TASK_COMM_LEN);
+	}
+}
+
+
+static  int hung_task_info_show(struct seq_file *m, void *v)
+{
+	struct hung_task *ht;
+
+	ht = list_entry(v, struct hung_task, list);
+	seq_puts(m, ht->comm);
+
+	return 0;
+}
+
+static void *hung_task_info_start(struct seq_file *m, loff_t *pos)
+{
+	return seq_list_start_head(&monitor_list->list, *pos);
+}
+
+static void *hung_task_info_next(struct seq_file *m, void *v, loff_t *pos)
+{
+	return seq_list_next(v, &monitor_list->list, pos);
+}
+
+static void hung_task_info_stop(struct seq_file *m, void *v)
+{
+}
+
+const struct seq_operations hung_task_info_op = {
+	.start	= hung_task_info_start,
+	.next	= hung_task_info_next,
+	.stop	= hung_task_info_stop,
+	.show	= hung_task_info_show
+};
+
+static int hung_task_info_open(struct inode *inode, struct file *file)
+{
+	return seq_open(file, &hung_task_info_op);
+}
+
+static ssize_t
+hung_task_info_write(struct file *file, const char __user *buf, size_t count,
+	     loff_t *offs)
+{
+	struct task_struct *g, *t;
+	struct hung_task *ht = kmalloc(sizeof(*ht), GFP_KERNEL);
+
+	if (!ht)
+		return -ENOMEM;
+
+	if (copy_from_user(ht->comm, buf, count))
+		return -EFAULT;
+	ht->comm[count] = '\0';
+
+	for_each_process_thread(g, t) {
+		if (!strncmp(t->comm, ht->comm, strlen(t->comm))) {
+			list_add_tail(&ht->list, &monitor_list->list);
+			return count;
+		}
+	}
+
+	pr_err("Non-existing task: %s can't be monitored\n", ht->comm);
+	return count;
+}
+
+static const struct file_operations hung_task_info_operations = {
+	.open		= hung_task_info_open,
+	.read		= seq_read,
+	.write		= hung_task_info_write,
+	.llseek		= seq_lseek,
+	.release	= seq_release,
+};
+
+static int __init proc_hung_task_info_init(void)
+{
+	proc_create("hung_task_monitor_list", 0644, NULL,
+		    &hung_task_info_operations);
+	return 0;
+}
+
 static void check_hung_task(struct task_struct *t, unsigned long timeout)
 {
 	unsigned long switch_count = t->nvcsw + t->nivcsw;

@@ -167,6 +269,7 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout)
 	int max_count = sysctl_hung_task_check_count;
 	int batch_count = HUNG_TASK_BATCHING;
 	struct task_struct *g, *t;
+	struct hung_task *ht, *tmp;
 
 	/*
 	 * If the system crashed already then all bets are off,

@@ -186,9 +289,21 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout)
 				goto unlock;
 		}
 		/* use "==" to skip the TASK_KILLABLE tasks waiting on NFS */
-		if (t->state == TASK_UNINTERRUPTIBLE)
-			check_hung_task(t, timeout);
+		if (t->state == TASK_UNINTERRUPTIBLE) {
+			if (sysctl_hung_task_check_selected) {
+				list_for_each_entry_safe(ht, tmp,
+							 &monitor_list->list,
+							 list)
+					if (!strncmp(ht->comm, t->comm,
+					    strlen(t->comm)))
+					/* Task belongs to the selected group */
+						check_hung_task(t, timeout);
+			} else {
+				check_hung_task(t, timeout);
+			}
+		}
 	}
+
  unlock:
 	rcu_read_unlock();
 	if (hung_task_show_lock)

@@ -259,6 +374,8 @@ static int watchdog(void *dummy)
 static int __init hung_task_init(void)
 {
 	atomic_notifier_chain_register(&panic_notifier_list, &panic_block);
+	hung_task_monitor_setup();
+	proc_hung_task_info_init();
 	watchdog_task = kthread_run(watchdog, NULL, "khungtaskd");
 
 	return 0;

kernel/sysctl.c +8 -0

index 6648fbb..ab45774 100644
@@ -1096,6 +1096,14 @@ static int sysrq_sysctl_handler(struct ctl_table *table, int write,
 		.proc_handler	= proc_dointvec_minmax,
 		.extra1		= &neg_one,
 	},
+	{
+		.procname	= "hung_task_check_selected",
+		.data		= &sysctl_hung_task_check_selected,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &neg_one,
+	},
 #endif
 #ifdef CONFIG_RT_MUTEXES
 	{

khungtask by default monitors either all tasks or no tasks at all for long unterruptible sleeps.
For Android like environments this arrangement is not optimal because on one hand it may be permissible to have some background(bg) task in uninterruptible sleep state for long duration while on the other hand it may not be permissible to have some foreground(fg) task like surfaceflinger in uninterruptible sleep state for long duration.

So it would be good to have some arrangement so that few specified tasks can be monitored by khungtaskd, on a need basis.
This change introduces a sysctl option, /proc/sys/kernel/ hung_task_check_selected, to enable monitoring of selected tasks using khungtask daemon. If this sysctl option is enabled then only the tasks specified in /proc/hung_task_monitor_list are monitored otherwise all tasks are monitored, just like the default case.

Quoted 7 lines · Imran Khan

khungtask by default monitors either all tasks or no tasks at all for long unterruptible sleeps.
For Android like environments this arrangement is not optimal because on one hand it may be permissible to have some background(bg) task in uninterruptible sleep state for long duration while on the other hand it may not be permissible to have some foreground(fg) task like surfaceflinger in uninterruptible sleep state for long duration.

How are you getting tasks in UNINTERRUPTIBLE state for a long time to begin with? That's not a good thing, background or not.

Quoted 7 lines

So it would be good to have some arrangement so that few specified tasks can be monitored by khungtaskd, on a need basis.
This change introduces a sysctl option, /proc/sys/kernel/ hung_task_check_selected, to enable monitoring of selected tasks using khungtask daemon. If this sysctl option is enabled then only the tasks specified in /proc/hung_task_monitor_list are monitored otherwise all tasks are monitored, just like the default case.

That's horrible. Why not a file in /proc/$PID/ that excludes it from things?

Some tasks may intentionally moves to uninterruptable sleep state, which shouldn't leads to khungtask panics, as those are recoverable hungs. So to avoid false hung reports, add an option to select tasks to be monitored and report/panic them only.

Quoted 4 lines · Lingutla Chandrasekhar

Some tasks may intentionally moves to uninterruptable sleep state, which shouldn't leads to khungtask panics, as those are recoverable hungs. So to avoid false hung reports, add an option to select tasks to be monitored and report/panic them only.

What are backtraces of such tasks? Please point the locations in the code.

If they are absolutely recoverable, why can't we let themselves declare that "I'm intentionally in uninterruptible state. But there is no dependency that prevents me from recovering. So, please ignore me." using per "struct task_struct" flags rather than introducing userspace controlled interface?

khungtask by default monitors all tasks for long unterruptible sleep. This change introduces a sysctl option, /proc/sys/kernel/ hung_task_selective_monitoring, to enable monitoring selected tasks. If this sysctl option is enabled then only the tasks with /proc/$PID/hang_detection_enabled

set are to be monitored,
otherwise all tasks are monitored as default case.

Some tasks may intentionally moves to uninterruptable sleep state, which shouldn't leads to khungtask panics, as those are recoverable hungs. So to avoid false hung reports, add an option to select tasks to be monitored and report/panic them only.

Signed-off-by Imran Khan <[email protected]>

Signed-off-by Lingutla Chandrasekhar <[email protected]>

fs/proc/base.c +54 -0

index 9d357b2ea6cb..810f9a5e209e 100644
@@ -2733,6 +2733,52 @@ static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
 }
 #endif /* CONFIG_TASK_IO_ACCOUNTING */
 
+#ifdef CONFIG_DETECT_HUNG_TASK
+static ssize_t proc_hung_task_detection_enabled_read(struct file *file,
+				char __user *buf, size_t count, loff_t *ppos)
+{
+	struct task_struct *task = get_proc_task(file_inode(file));
+	char buffer[PROC_NUMBUF];
+	size_t len;
+	bool hang_detection_enabled;
+
+	if (!task)
+		return -ESRCH;
+	hang_detection_enabled = task->hang_detection_enabled;
+	put_task_struct(task);
+
+	len = snprintf(buffer, sizeof(buffer), "%d\n", hang_detection_enabled);
+
+	return simple_read_from_buffer(buf, sizeof(buffer), ppos, buffer, len);
+}
+
+static ssize_t proc_hung_task_detection_enabled_write(struct file *file,
+			const char __user *buf, size_t count, loff_t *ppos)
+{
+	struct task_struct *task;
+	bool hang_detection_enabled;
+	int rv;
+
+	rv = kstrtobool_from_user(buf, count, &hang_detection_enabled);
+	if (rv < 0)
+		return rv;
+
+	task = get_proc_task(file_inode(file));
+	if (!task)
+		return -ESRCH;
+	task->hang_detection_enabled = hang_detection_enabled;
+	put_task_struct(task);
+
+	return count;
+}
+
+static const struct file_operations proc_hung_task_detection_enabled_operations = {
+	.read		= proc_hung_task_detection_enabled_read,
+	.write		= proc_hung_task_detection_enabled_write,
+	.llseek		= generic_file_llseek,
+};
+#endif
+
 #ifdef CONFIG_USER_NS
 static int proc_id_map_open(struct inode *inode, struct file *file,
 	const struct seq_operations *seq_ops)

@@ -2976,6 +3022,10 @@ static const struct pid_entry tgid_base_stuff[] = {
 #ifdef CONFIG_HARDWALL
 	ONE("hardwall",   S_IRUGO, proc_pid_hardwall),
 #endif
+#ifdef CONFIG_DETECT_HUNG_TASK
+	REG("hang_detection_enabled", 0666,
+		proc_hung_task_detection_enabled_operations),
+#endif
 #ifdef CONFIG_USER_NS
 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),

@@ -3367,6 +3417,10 @@ static const struct pid_entry tid_base_stuff[] = {
 #ifdef CONFIG_HARDWALL
 	ONE("hardwall",   S_IRUGO, proc_pid_hardwall),
 #endif
+#ifdef CONFIG_DETECT_HUNG_TASK
+	REG("hang_detection_enabled", 0666,
+		proc_hung_task_detection_enabled_operations),
+#endif
 #ifdef CONFIG_USER_NS
 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),

include/linux/sched.h +1 -0

index fdf74f27acf1..1280df0a6708 100644
@@ -767,6 +767,7 @@ struct task_struct {
 #endif
 #ifdef CONFIG_DETECT_HUNG_TASK
 	unsigned long			last_switch_count;
+	bool				hang_detection_enabled;
 #endif
 	/* Filesystem information: */
 	struct fs_struct		*fs;

include/linux/sched/sysctl.h +1 -0

index d6a18a3839cc..09d9f1c65fd2 100644
@@ -11,6 +11,7 @@ extern int	     sysctl_hung_task_check_count;
 extern unsigned int  sysctl_hung_task_panic;
 extern unsigned long sysctl_hung_task_timeout_secs;
 extern int sysctl_hung_task_warnings;
+extern int sysctl_hung_task_selective_monitoring;
 extern int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,
 					 void __user *buffer,
 					 size_t *lenp, loff_t *ppos);

kernel/hung_task.c +13 -1

index 751593ed7c0b..321bcfa2e34a 100644
@@ -20,12 +20,21 @@
 #include <linux/sched/debug.h>
 
 #include <trace/events/sched.h>
+#include <linux/sched/sysctl.h>
 
 /*
  * The number of tasks checked:
  */
 int __read_mostly sysctl_hung_task_check_count = PID_MAX_LIMIT;
 
+/*
+ * Selective monitoring of hung tasks.
+ *
+ * if set to 1, khungtaskd only monitors tasks with 'hang_detection_enabled'
+ * value set, else monitors all tasks.
+ */
+int sysctl_hung_task_selective_monitoring;
+
 /*
  * Limit number of tasks checked in a batch.
  *

@@ -187,7 +196,10 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout)
 		}
 		/* use "==" to skip the TASK_KILLABLE tasks waiting on NFS */
 		if (t->state == TASK_UNINTERRUPTIBLE)
-			check_hung_task(t, timeout);
+			/* Check for selective monitoring */
+			if (!sysctl_hung_task_selective_monitoring ||
+			    t->hang_detection_enabled)
+				check_hung_task(t, timeout);
 	}
  unlock:
 	rcu_read_unlock();

kernel/sysctl.c +10 -0

index d9c31bc2eaea..30cc108ce2cc 100644
@@ -1093,6 +1093,16 @@ static struct ctl_table kern_table[] = {
 		.proc_handler	= proc_dointvec_minmax,
 		.extra1		= &neg_one,
 	},
+	{
+		.procname	= "hung_task_selective_monitoring",
+		.data		= &sysctl_hung_task_selective_monitoring,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &zero,
+		.extra2		= &one,
+	},
+
 #endif
 #ifdef CONFIG_RT_MUTEXES
 	{

Signature · 2 lines

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
 a Linux Foundation Collaborative Project.