4 * kernel/hung_task.c - kernel thread for detecting tasks stuck in D state
10 #include <linux/nmi.h>
11 #include <linux/init.h>
12 #include <linux/delay.h>
13 #include <linux/freezer.h>
14 #include <linux/kthread.h>
15 #include <linux/lockdep.h>
16 #include <linux/module.h>
17 #include <linux/sysctl.h>
20 * The number of tasks checked:
22 unsigned long __read_mostly sysctl_hung_task_check_count = PID_MAX_LIMIT;
25 * Limit number of tasks checked in a batch.
27 * This value controls the preemptibility of khungtaskd since preemption
28 * is disabled during the critical section. It also controls the size of
29 * the RCU grace period. So it needs to be upper-bound.
31 #define HUNG_TASK_BATCHING 1024
34 * Zero means infinite timeout - no checking done:
36 unsigned long __read_mostly sysctl_hung_task_timeout_secs = 120;
38 unsigned long __read_mostly sysctl_hung_task_warnings = 10;
40 static int __read_mostly did_panic;
42 static struct task_struct *watchdog_task;
45 * Should we panic (and reboot, if panic_timeout= is set) when a
46 * hung task is detected:
48 unsigned int __read_mostly sysctl_hung_task_panic =
49 CONFIG_BOOTPARAM_HUNG_TASK_PANIC_VALUE;
51 static int __init hung_task_panic_setup(char *str)
53 sysctl_hung_task_panic = simple_strtoul(str, NULL, 0);
57 __setup("hung_task_panic=", hung_task_panic_setup);
60 hung_task_panic(struct notifier_block *this, unsigned long event, void *ptr)
67 static struct notifier_block panic_block = {
68 .notifier_call = hung_task_panic,
71 static void check_hung_task(struct task_struct *t, unsigned long timeout)
73 unsigned long switch_count = t->nvcsw + t->nivcsw;
75 if (t->flags & PF_FROZEN)
78 if (switch_count != t->last_switch_count) {
79 t->last_switch_count = switch_count;
82 if (!sysctl_hung_task_warnings)
84 sysctl_hung_task_warnings--;
87 * Ok, the task did not get scheduled for more than 2 minutes,
90 printk(KERN_ERR "INFO: task %s:%d blocked for more than "
91 "%ld seconds.\n", t->comm, t->pid, timeout);
92 printk(KERN_ERR "\"echo 0 > /proc/sys/kernel/hung_task_timeout_secs\""
93 " disables this message.\n");
95 __debug_show_held_locks(t);
99 if (sysctl_hung_task_panic)
100 panic("hung_task: blocked tasks");
104 * To avoid extending the RCU grace period for an unbounded amount of time,
105 * periodically exit the critical section and enter a new one.
107 * For preemptible RCU it is sufficient to call rcu_read_unlock in order
108 * exit the grace period. For classic RCU, a reschedule is required.
110 static void rcu_lock_break(struct task_struct *g, struct task_struct *t)
122 * Check whether a TASK_UNINTERRUPTIBLE does not get woken up for
123 * a really long time (120 seconds). If that happens, print out
126 static void check_hung_uninterruptible_tasks(unsigned long timeout)
128 int max_count = sysctl_hung_task_check_count;
129 int batch_count = HUNG_TASK_BATCHING;
130 struct task_struct *g, *t;
133 * If the system crashed already then all bets are off,
134 * do not report extra hung tasks:
136 if (test_taint(TAINT_DIE) || did_panic)
140 do_each_thread(g, t) {
143 if (!--batch_count) {
144 batch_count = HUNG_TASK_BATCHING;
145 rcu_lock_break(g, t);
146 /* Exit if t or g was unhashed during refresh. */
147 if (t->state == TASK_DEAD || g->state == TASK_DEAD)
150 /* use "==" to skip the TASK_KILLABLE tasks waiting on NFS */
151 if (t->state == TASK_UNINTERRUPTIBLE)
152 check_hung_task(t, timeout);
153 } while_each_thread(g, t);
158 static unsigned long timeout_jiffies(unsigned long timeout)
160 /* timeout of 0 will disable the watchdog */
161 return timeout ? timeout * HZ : MAX_SCHEDULE_TIMEOUT;
165 * Process updating of timeout sysctl
167 int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,
168 struct file *filp, void __user *buffer,
169 size_t *lenp, loff_t *ppos)
173 ret = proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos);
178 wake_up_process(watchdog_task);
185 * kthread which checks for tasks stuck in D state
187 static int watchdog(void *dummy)
189 set_user_nice(current, 0);
192 unsigned long timeout = sysctl_hung_task_timeout_secs;
194 while (schedule_timeout_interruptible(timeout_jiffies(timeout)))
195 timeout = sysctl_hung_task_timeout_secs;
197 check_hung_uninterruptible_tasks(timeout);
203 static int __init hung_task_init(void)
205 atomic_notifier_chain_register(&panic_notifier_list, &panic_block);
206 watchdog_task = kthread_run(watchdog, NULL, "khungtaskd");
211 module_init(hung_task_init);