Re: [PATCH] hung_task: Allow hung_task_panic when hung_task_warnings is 0.

[Andrew Morton]

On Fri,  9 Sep 2016 15:43:34 -0400 jsiddle@xxxxxxxxxx wrote:

> From: John Siddle <jsiddle@xxxxxxxxxx>
> 
> Previously hung_task_panic would not be respected if enabled after
> hung_task_warnings had already been decremented to 0.
> 
> Permit the kernel to panic if hung_task_panic is enabled after
> hung_task_warnings has already been decremented to 0 and another task
> hangs for hung_task_timeout_secs seconds.
> 
> Check if hung_task_panic is enabled so we don't return prematurely, and
> check if hung_task_warnings is non-zero so we don't print the warning
> unnecessarily.
> 
> ...
>
> --- a/kernel/hung_task.c
> +++ b/kernel/hung_task.c
> @@ -98,7 +98,7 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
>  
>  	trace_sched_process_hang(t);
>  
> -	if (!sysctl_hung_task_warnings)
> +	if (!sysctl_hung_task_warnings && !sysctl_hung_task_panic)
>  		return;
>  
>  	if (sysctl_hung_task_warnings > 0)
> @@ -108,16 +108,18 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
>  	 * Ok, the task did not get scheduled for more than 2 minutes,
>  	 * complain:
>  	 */
> -	pr_err("INFO: task %s:%d blocked for more than %ld seconds.\n",
> -		t->comm, t->pid, timeout);
> -	pr_err("      %s %s %.*s\n",
> -		print_tainted(), init_utsname()->release,
> -		(int)strcspn(init_utsname()->version, " "),
> -		init_utsname()->version);
> -	pr_err("\"echo 0 > /proc/sys/kernel/hung_task_timeout_secs\""
> -		" disables this message.\n");
> -	sched_show_task(t);
> -	debug_show_held_locks(t);
> +	if (sysctl_hung_task_warnings) {
> +		pr_err("INFO: task %s:%d blocked for more than %ld seconds.\n",
> +			t->comm, t->pid, timeout);
> +		pr_err("      %s %s %.*s\n",
> +			print_tainted(), init_utsname()->release,
> +			(int)strcspn(init_utsname()->version, " "),
> +			init_utsname()->version);
> +		pr_err("\"echo 0 > /proc/sys/kernel/hung_task_timeout_secs\""
> +			" disables this message.\n");
> +		sched_show_task(t);
> +		debug_show_held_locks(t);
> +	}

This introduces an off-by-one error.  In the old code, if
sysctl_hung_task_warnings==1 on entry, we warn.  With the new code, we
no longer warn.

This?

--- a/kernel/hung_task.c~hung_task-allow-hung_task_panic-when-hung_task_warnings-is-0-fix
+++ a/kernel/hung_task.c
@@ -101,14 +101,12 @@ static void check_hung_task(struct task_
 	if (!sysctl_hung_task_warnings && !sysctl_hung_task_panic)
 		return;
 
-	if (sysctl_hung_task_warnings > 0)
-		sysctl_hung_task_warnings--;
-
 	/*
 	 * Ok, the task did not get scheduled for more than 2 minutes,
 	 * complain:
 	 */
 	if (sysctl_hung_task_warnings) {
+		sysctl_hung_task_warnings--;
 		pr_err("INFO: task %s:%d blocked for more than %ld seconds.\n",
 			t->comm, t->pid, timeout);
 		pr_err("      %s %s %.*s\n",
_