Re: 2.6.21-rc2-git3 soft lockup detected on CPU#0 (crash dump kernel)

[Ingo Molnar]

* Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:

> > shouldnt that be 'R' instead of the question mark?
> 
> ? are dead ones,  

ok.

> The interesting part is here:
> 
> 1 lock held by udevd/918:
>  #0:  (&mm->mmap_sem){----}, at: [<c161135b>] do_page_fault+0x150/0x50d
> 
> But udevd/918 is nowhere in the sysrq-t output.

also below is a softlockup-debug patch that might be useful, which adds 
a tick/tock output so it spams the console once per second but can show 
the dynamics of lockups (and other delays).

	Ingo

------------------->
Subject: [patch] softlockup: add /proc/sys/kernel/softlockup_print_tick
From: Ingo Molnar <mingo@xxxxxxx>

/proc/sys/kernel/softlockup_print_tick tunable:

 0: off
 1: print tick/tock messages in softlockup irq and task.
 2: also print stack dump

Signed-off-by: Ingo Molnar <mingo@xxxxxxx>
---
 Documentation/kernel-parameters.txt |    5 +++++
 include/linux/sched.h               |    2 ++
 kernel/softlockup.c                 |   13 ++++++++++++-
 kernel/sysctl.c                     |   10 ++++++++++
 4 files changed, 29 insertions(+), 1 deletion(-)

Index: linux/Documentation/kernel-parameters.txt
===================================================================
--- linux.orig/Documentation/kernel-parameters.txt
+++ linux/Documentation/kernel-parameters.txt
@@ -1659,6 +1659,11 @@ and is between 256 and 4096 characters. 
 
 	snd-ymfpci=	[HW,ALSA]
 
+	softlockup_print_tick
+			[KNL] Print softlockup tick in the irq and in the
+			watchdog task. If value 2 then print stackdump too.
+			Format: <0/1/2>
+
 	sonycd535=	[HW,CD]
 			Format: <io>[,<irq>]
 
Index: linux/include/linux/sched.h
===================================================================
--- linux.orig/include/linux/sched.h
+++ linux/include/linux/sched.h
@@ -223,6 +223,8 @@ extern void update_process_times(int use
 extern void scheduler_tick(void);
 
 #ifdef CONFIG_DETECT_SOFTLOCKUP
+extern int softlockup_print_tick;
+
 extern void softlockup_tick(void);
 extern void spawn_softlockup_task(void);
 extern void touch_softlockup_watchdog(void);
Index: linux/kernel/softlockup.c
===================================================================
--- linux.orig/kernel/softlockup.c
+++ linux/kernel/softlockup.c
@@ -21,6 +21,7 @@ static DEFINE_PER_CPU(unsigned long, pri
 static DEFINE_PER_CPU(struct task_struct *, watchdog_task);
 
 static int did_panic = 0;
+int softlockup_print_tick = 2;
 
 static int
 softlock_panic(struct notifier_block *this, unsigned long event, void *ptr)
@@ -62,8 +63,15 @@ void softlockup_tick(void)
 	}
 
 	/* Wake up the high-prio watchdog task every second: */
-	if (time_after(jiffies, touch_timestamp + HZ))
+	if (time_after(jiffies, touch_timestamp + HZ)) {
+		if (softlockup_print_tick) {
+			printk("softlockup tick on CPU#%d, %s:%d\n",
+				this_cpu, current->comm, current->pid);
+			if (softlockup_print_tick == 2)
+				dump_stack();
+		}
 		wake_up_process(per_cpu(watchdog_task, this_cpu));
+	}
 
 	/* Warn about unreasonable 10+ seconds delays: */
 	if (time_after(jiffies, touch_timestamp + 10*HZ)) {
@@ -83,6 +91,7 @@ void softlockup_tick(void)
 static int watchdog(void * __bind_cpu)
 {
 	struct sched_param param = { .sched_priority = 99 };
+	int this_cpu = (int)(long)__bind_cpu;
 
 	sched_setscheduler(current, SCHED_FIFO, &param);
 	current->flags |= PF_NOFREEZE;
@@ -96,6 +105,8 @@ static int watchdog(void * __bind_cpu)
 		set_current_state(TASK_INTERRUPTIBLE);
 		touch_softlockup_watchdog();
 		schedule();
+		if (softlockup_print_tick)
+			printk("softlockup tock on CPU#%d\n", this_cpu);
 	}
 
 	return 0;
Index: linux/kernel/sysctl.c
===================================================================
--- linux.orig/kernel/sysctl.c
+++ linux/kernel/sysctl.c
@@ -231,6 +231,16 @@ static ctl_table kern_table[] = {
 		.proc_handler	= &proc_dostring,
 		.strategy	= &sysctl_string,
 	},
+#ifdef CONFIG_DETECT_SOFTLOCKUP
+	{
+		.ctl_name	= CTL_UNNUMBERED,
+		.procname	= "softlockup_print_tick",
+		.data		= &softlockup_print_tick,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+	},
+#endif
 #ifdef CONFIG_PROC_SYSCTL
 	{
 		.ctl_name	= KERN_TAINTED,
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/