[PATCH] lktdm: Bring support for hardlockup, softlockup and hung task crashes

[PATCH] lktdm: Bring support for hardlockup, softlockup and hung task crashes

[Frederic Weisbecker]

This adds three new types of kernel "crashes" in the lkdtm
driver to trigger hardlockups, softlockups and task hung
states at will.

The two formers are useful to test the new generic lockup
detector and check its further regressions. The latter one
is a bonus to check the hung task detector regressions even
though it's not currently in rework.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Simon Kagstrom <simon.kagstrom@netinsight.net>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
---
 drivers/misc/lkdtm.c |   20 ++++++++++++++++++++
 1 files changed, 20 insertions(+), 0 deletions(-)

diff --git a/drivers/misc/lkdtm.c b/drivers/misc/lkdtm.c
index 31a9911..5bfb2a2 100644
--- a/drivers/misc/lkdtm.c
+++ b/drivers/misc/lkdtm.c
@@ -75,6 +75,9 @@ enum ctype {
 	UNALIGNED_LOAD_STORE_WRITE,
 	OVERWRITE_ALLOCATION,
 	WRITE_AFTER_FREE,
+	SOFTLOCKUP,
+	HARDLOCKUP,
+	HUNG_TASK,
 };
 
 static char* cp_name[] = {
@@ -99,6 +102,9 @@ static char* cp_type[] = {
 	"UNALIGNED_LOAD_STORE_WRITE",
 	"OVERWRITE_ALLOCATION",
 	"WRITE_AFTER_FREE",
+	"SOFTLOCKUP",
+	"HARDLOCKUP",
+	"HUNG_TASK",
 };
 
 static struct jprobe lkdtm;
@@ -320,6 +326,20 @@ static void lkdtm_do_action(enum ctype which)
 		memset(data, 0x78, len);
 		break;
 	}
+	case SOFTLOCKUP:
+		preempt_disable();
+		for (;;)
+			cpu_relax();
+		break;
+	case HARDLOCKUP:
+		local_irq_disable();
+		for (;;)
+			cpu_relax();
+		break;
+	case HUNG_TASK:
+		set_current_state(TASK_UNINTERRUPTIBLE);
+		schedule();
+		break;
 	case NONE:
 	default:
 		break;
-- 
1.6.2.3

Re: [PATCH] lktdm: Bring support for hardlockup, softlockup and hung task crashes

[Cyrill Gorcunov]

On Sun, May 16, 2010 at 07:39:51AM +0200, Frederic Weisbecker wrote:
...  
>  static struct jprobe lkdtm;
> @@ -320,6 +326,20 @@ static void lkdtm_do_action(enum ctype which)
>  		memset(data, 0x78, len);
>  		break;
>  	}
> +	case SOFTLOCKUP:
> +		preempt_disable();
> +		for (;;)
> +			cpu_relax();
> +		break;
> +	case HARDLOCKUP:
> +		local_irq_disable();
> +		for (;;)
> +			cpu_relax();
> +		break;
> +	case HUNG_TASK:
> +		set_current_state(TASK_UNINTERRUPTIBLE);
> +		schedule();
> +		break;
>  	case NONE:
>  	default:
>  		break;

Looks good to me. Btw perhaps we may simplify it a bit:

	case HARDLOCKUP:
		local_irq_disable();
	case SOFTLOCKUP:
		preempt_disable();
		for (;;)
			cpu_relax();
		break;

since it'll save a few bytes. What do you think? Did I miss
something?

	-- Cyrill

Re: [PATCH] lktdm: Bring support for hardlockup, softlockup and hung task crashes

[Frederic Weisbecker]

On Sun, May 16, 2010 at 01:36:36PM +0400, Cyrill Gorcunov wrote:
> On Sun, May 16, 2010 at 07:39:51AM +0200, Frederic Weisbecker wrote:
> ...  
> >  static struct jprobe lkdtm;
> > @@ -320,6 +326,20 @@ static void lkdtm_do_action(enum ctype which)
> >  		memset(data, 0x78, len);
> >  		break;
> >  	}
> > +	case SOFTLOCKUP:
> > +		preempt_disable();
> > +		for (;;)
> > +			cpu_relax();
> > +		break;
> > +	case HARDLOCKUP:
> > +		local_irq_disable();
> > +		for (;;)
> > +			cpu_relax();
> > +		break;
> > +	case HUNG_TASK:
> > +		set_current_state(TASK_UNINTERRUPTIBLE);
> > +		schedule();
> > +		break;
> >  	case NONE:
> >  	default:
> >  		break;
> 
> Looks good to me. Btw perhaps we may simplify it a bit:
> 
> 	case HARDLOCKUP:
> 		local_irq_disable();
> 	case SOFTLOCKUP:
> 		preempt_disable();
> 		for (;;)
> 			cpu_relax();
> 		break;
> 
> since it'll save a few bytes. What do you think? Did I miss
> something?


It would make the code a bit less clear in that people
might stick on the reason to disable preemption after disabling
irq, especially with a code that already does something rather
unusual ;)

Re: [PATCH] lktdm: Bring support for hardlockup, softlockup and hung task crashes

[Cyrill Gorcunov]

On Sun, May 16, 2010 at 12:35:08PM +0200, Frederic Weisbecker wrote:
...
> > Looks good to me. Btw perhaps we may simplify it a bit:
> > 
> > 	case HARDLOCKUP:
> > 		local_irq_disable();
> > 	case SOFTLOCKUP:
> > 		preempt_disable();
> > 		for (;;)
> > 			cpu_relax();
> > 		break;
> > 
> > since it'll save a few bytes. What do you think? Did I miss
> > something?
> 
> 
> It would make the code a bit less clear in that people
> might stick on the reason to disable preemption after disabling
> irq, especially with a code that already does something rather
> unusual ;)
> 

ok, convinced

	-- Cyrill