Message-ID: <20070226012512.GM5049@linux.vnet.ibm.com>
Date:	Sun, 25 Feb 2007 17:25:13 -0800
From:	"Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
To:	Ingo Molnar <mingo@...e.hu>
Cc:	tglx@...utronix.de, linux-kernel@...r.kernel.org,
	linux-rt-users@...r.kernel.org
Subject: Re: BUG in 2.6.20-rt8

On Sun, Feb 25, 2007 at 07:27:47AM +0100, Ingo Molnar wrote:
> 
> * Paul E. McKenney <paulmck@...ux.vnet.ibm.com> wrote:
> 
> > I got the following running stock 2.6.20-rt8 on an 4-CPU 1.8GHz 
> > Opteron box.  The machine continued to run a few rounds of kernbench 
> > and LTP. Looks a bit scary -- a tasklet was "stolen" from 
> > __tasklet_action().
> > 
> > Thoughts?  In the meantime, kicking it off again to see if it repeats.
> 
> > BUG: at kernel/softirq.c:559 __tasklet_action()
> 
> this seems to happen very sporadically. Seems to happen more likely on 
> hyperthreading CPUs. It is very likely caused by the 
> redesign-tasklet-locking-to-be-sane patch below - which is a quick hack 
> of mine from early -rt days. Can you see any obvious bug in it? The 
> cmpxchg logic is certainly a bit ... tricky, locking-wise.

One thing thus far...

There is the repeated-reschedule case that looks like it can result
in a tasklet being dropped on the floor:

o	__tasklet_action() runs out of "loops" (at which point,
	we know TASKLET_STATE_SCHED is clear due to the prior test).

o	Someone calls tasklet_schedule(), which sets TASKLET_STATE_SCHED
	and calls __tasklet_schedule(), which fails in tasklet_trylock()
	due to __tasklet_action() still having the TASKLET_STATE_RUN
	bit set.  So the tasklet is not queued.

o	__tasklet_action() then forcibly unlocks, and the tasklet can
	never run again, because its TASKLET_STATE_SCHED is already
	set.

But this would have printed the "hm" message, which I didn't see.
And been a low-probability race on top of that.

And I have not yet come up with a better way of handling this -- and
if I did, it would likely involve more rather than less cmpxchg.  ;-)

If I can figure out a reasonable assertion, I will throw it at Promela
and see what happens.

						Thanx, Paul

> ---------------->
> From: Ingo Molnar <mingo@...e.hu>
> 
> tasklet redesign: make it saner and make it easier to thread.
> 
> Signed-off-by: Ingo Molnar <mingo@...e.hu>
> 
> ----
>  include/linux/interrupt.h |   39 ++++++-----
>  kernel/softirq.c          |  155 +++++++++++++++++++++++++++++++---------------
>  2 files changed, 128 insertions(+), 66 deletions(-)
> 
> Index: linux/include/linux/interrupt.h
> ===================================================================
> --- linux.orig/include/linux/interrupt.h
> +++ linux/include/linux/interrupt.h
> @@ -276,8 +276,9 @@ extern void wait_for_softirq(int softirq
>       to be executed on some cpu at least once after this.
>     * If the tasklet is already scheduled, but its excecution is still not
>       started, it will be executed only once.
> -   * If this tasklet is already running on another CPU (or schedule is called
> -     from tasklet itself), it is rescheduled for later.
> +   * If this tasklet is already running on another CPU, it is rescheduled
> +     for later.
> +   * Schedule must not be called from the tasklet itself (a lockup occurs)
>     * Tasklet is strictly serialized wrt itself, but not
>       wrt another tasklets. If client needs some intertask synchronization,
>       he makes it with spinlocks.
> @@ -302,15 +303,25 @@ struct tasklet_struct name = { NULL, 0, 
>  enum
>  {
>  	TASKLET_STATE_SCHED,	/* Tasklet is scheduled for execution */
> -	TASKLET_STATE_RUN	/* Tasklet is running (SMP only) */
> +	TASKLET_STATE_RUN,	/* Tasklet is running (SMP only) */
> +	TASKLET_STATE_PENDING	/* Tasklet is pending */
>  };
> 
> -#ifdef CONFIG_SMP
> +#define TASKLET_STATEF_SCHED	(1 << TASKLET_STATE_SCHED)
> +#define TASKLET_STATEF_RUN	(1 << TASKLET_STATE_RUN)
> +#define TASKLET_STATEF_PENDING	(1 << TASKLET_STATE_PENDING)
> +
> +#if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT_RT)
>  static inline int tasklet_trylock(struct tasklet_struct *t)
>  {
>  	return !test_and_set_bit(TASKLET_STATE_RUN, &(t)->state);
>  }
> 
> +static inline int tasklet_tryunlock(struct tasklet_struct *t)
> +{
> +	return cmpxchg(&t->state, TASKLET_STATEF_RUN, 0) == TASKLET_STATEF_RUN;
> +}
> +
>  static inline void tasklet_unlock(struct tasklet_struct *t)
>  {
>  	smp_mb__before_clear_bit(); 
> @@ -322,9 +333,10 @@ static inline void tasklet_unlock_wait(s
>  	while (test_bit(TASKLET_STATE_RUN, &(t)->state)) { barrier(); }
>  }
>  #else
> -#define tasklet_trylock(t) 1
> -#define tasklet_unlock_wait(t) do { } while (0)
> -#define tasklet_unlock(t) do { } while (0)
> +# define tasklet_trylock(t)		1
> +# define tasklet_tryunlock(t)		1
> +# define tasklet_unlock_wait(t)		do { } while (0)
> +# define tasklet_unlock(t)		do { } while (0)
>  #endif
> 
>  extern void FASTCALL(__tasklet_schedule(struct tasklet_struct *t));
> @@ -357,17 +369,8 @@ static inline void tasklet_disable(struc
>  	smp_mb();
>  }
> 
> -static inline void tasklet_enable(struct tasklet_struct *t)
> -{
> -	smp_mb__before_atomic_dec();
> -	atomic_dec(&t->count);
> -}
> -
> -static inline void tasklet_hi_enable(struct tasklet_struct *t)
> -{
> -	smp_mb__before_atomic_dec();
> -	atomic_dec(&t->count);
> -}
> +extern fastcall void tasklet_enable(struct tasklet_struct *t);
> +extern fastcall void tasklet_hi_enable(struct tasklet_struct *t);
> 
>  extern void tasklet_kill(struct tasklet_struct *t);
>  extern void tasklet_kill_immediate(struct tasklet_struct *t, unsigned int cpu);
> Index: linux/kernel/softirq.c
> ===================================================================
> --- linux.orig/kernel/softirq.c
> +++ linux/kernel/softirq.c
> @@ -464,14 +464,24 @@ struct tasklet_head
>  static DEFINE_PER_CPU(struct tasklet_head, tasklet_vec) = { NULL };
>  static DEFINE_PER_CPU(struct tasklet_head, tasklet_hi_vec) = { NULL };
> 
> +static void inline
> +__tasklet_common_schedule(struct tasklet_struct *t, struct tasklet_head *head, unsigned int nr)
> +{
> +	if (tasklet_trylock(t)) {
> +		WARN_ON(t->next != NULL);
> +		t->next = head->list;
> +		head->list = t;
> +		raise_softirq_irqoff(nr);
> +		tasklet_unlock(t);
> +	}
> +}
> +
>  void fastcall __tasklet_schedule(struct tasklet_struct *t)
>  {
>  	unsigned long flags;
> 
>  	local_irq_save(flags);
> -	t->next = __get_cpu_var(tasklet_vec).list;
> -	__get_cpu_var(tasklet_vec).list = t;
> -	raise_softirq_irqoff(TASKLET_SOFTIRQ);
> +	__tasklet_common_schedule(t, &__get_cpu_var(tasklet_vec), TASKLET_SOFTIRQ);
>  	local_irq_restore(flags);
>  }
> 
> @@ -482,81 +492,130 @@ void fastcall __tasklet_hi_schedule(stru
>  	unsigned long flags;
> 
>  	local_irq_save(flags);
> -	t->next = __get_cpu_var(tasklet_hi_vec).list;
> -	__get_cpu_var(tasklet_hi_vec).list = t;
> -	raise_softirq_irqoff(HI_SOFTIRQ);
> +	__tasklet_common_schedule(t, &__get_cpu_var(tasklet_hi_vec), HI_SOFTIRQ);
>  	local_irq_restore(flags);
>  }
> 
>  EXPORT_SYMBOL(__tasklet_hi_schedule);
> 
> -static void tasklet_action(struct softirq_action *a)
> +void fastcall tasklet_enable(struct tasklet_struct *t)
>  {
> -	struct tasklet_struct *list;
> +	if (!atomic_dec_and_test(&t->count))
> +		return;
> +	if (test_and_clear_bit(TASKLET_STATE_PENDING, &t->state))
> +		tasklet_schedule(t);
> +}
> 
> -	local_irq_disable();
> -	list = __get_cpu_var(tasklet_vec).list;
> -	__get_cpu_var(tasklet_vec).list = NULL;
> -	local_irq_enable();
> +EXPORT_SYMBOL(tasklet_enable);
> +
> +void fastcall tasklet_hi_enable(struct tasklet_struct *t)
> +{
> +	if (!atomic_dec_and_test(&t->count))
> +		return;
> +	if (test_and_clear_bit(TASKLET_STATE_PENDING, &t->state))
> +		tasklet_hi_schedule(t);
> +}
> +
> +EXPORT_SYMBOL(tasklet_hi_enable);
> +
> +static void
> +__tasklet_action(struct softirq_action *a, struct tasklet_struct *list)
> +{
> +	int loops = 1000000;
> 
>  	while (list) {
>  		struct tasklet_struct *t = list;
> 
>  		list = list->next;
> +		/*
> +		 * Should always succeed - after a tasklist got on the
> +		 * list (after getting the SCHED bit set from 0 to 1),
> +		 * nothing but the tasklet softirq it got queued to can
> +		 * lock it:
> +		 */
> +		if (!tasklet_trylock(t)) {
> +			WARN_ON(1);
> +			continue;
> +		}
> +
> +		t->next = NULL;
> +
> +		/*
> +		 * If we cannot handle the tasklet because it's disabled,
> +		 * mark it as pending. tasklet_enable() will later
> +		 * re-schedule the tasklet.
> +		 */
> +		if (unlikely(atomic_read(&t->count))) {
> +out_disabled:
> +			/* implicit unlock: */
> +			wmb();
> +			t->state = TASKLET_STATEF_PENDING;
> +			continue;
> +		}
> 
> -		if (tasklet_trylock(t)) {
> -			if (!atomic_read(&t->count)) {
> -				if (!test_and_clear_bit(TASKLET_STATE_SCHED, &t->state))
> -					BUG();
> -				t->func(t->data);
> +		/*
> +		 * After this point on the tasklet might be rescheduled
> +		 * on another CPU, but it can only be added to another
> +		 * CPU's tasklet list if we unlock the tasklet (which we
> +		 * dont do yet).
> +		 */
> +		if (!test_and_clear_bit(TASKLET_STATE_SCHED, &t->state))
> +			WARN_ON(1);
> +
> +again:
> +		t->func(t->data);
> +
> +		/*
> +		 * Try to unlock the tasklet. We must use cmpxchg, because
> +		 * another CPU might have scheduled or disabled the tasklet.
> +		 * We only allow the STATE_RUN -> 0 transition here.
> +		 */
> +		while (!tasklet_tryunlock(t)) {
> +			/*
> +			 * If it got disabled meanwhile, bail out:
> +			 */
> +			if (atomic_read(&t->count))
> +				goto out_disabled;
> +			/*
> +			 * If it got scheduled meanwhile, re-execute
> +			 * the tasklet function:
> +			 */
> +			if (test_and_clear_bit(TASKLET_STATE_SCHED, &t->state))
> +				goto again;
> +			if (!--loops) {
> +				printk("hm, tasklet state: %08lx\n", t->state);
> +				WARN_ON(1);
>  				tasklet_unlock(t);
> -				continue;
> +				break;
>  			}
> -			tasklet_unlock(t);
>  		}
> -
> -		local_irq_disable();
> -		t->next = __get_cpu_var(tasklet_vec).list;
> -		__get_cpu_var(tasklet_vec).list = t;
> -		__do_raise_softirq_irqoff(TASKLET_SOFTIRQ);
> -		local_irq_enable();
>  	}
>  }
> 
> -static void tasklet_hi_action(struct softirq_action *a)
> +static void tasklet_action(struct softirq_action *a)
>  {
>  	struct tasklet_struct *list;
> 
>  	local_irq_disable();
> -	list = __get_cpu_var(tasklet_hi_vec).list;
> -	__get_cpu_var(tasklet_hi_vec).list = NULL;
> +	list = __get_cpu_var(tasklet_vec).list;
> +	__get_cpu_var(tasklet_vec).list = NULL;
>  	local_irq_enable();
> 
> -	while (list) {
> -		struct tasklet_struct *t = list;
> +	__tasklet_action(a, list);
> +}
> 
> -		list = list->next;
> +static void tasklet_hi_action(struct softirq_action *a)
> +{
> +	struct tasklet_struct *list;
> 
> -		if (tasklet_trylock(t)) {
> -			if (!atomic_read(&t->count)) {
> -				if (!test_and_clear_bit(TASKLET_STATE_SCHED, &t->state))
> -					BUG();
> -				t->func(t->data);
> -				tasklet_unlock(t);
> -				continue;
> -			}
> -			tasklet_unlock(t);
> -		}
> +	local_irq_disable();
> +	list = __get_cpu_var(tasklet_hi_vec).list;
> +	__get_cpu_var(tasklet_hi_vec).list = NULL;
> +	local_irq_enable();
> 
> -		local_irq_disable();
> -		t->next = __get_cpu_var(tasklet_hi_vec).list;
> -		__get_cpu_var(tasklet_hi_vec).list = t;
> -		__do_raise_softirq_irqoff(HI_SOFTIRQ);
> -		local_irq_enable();
> -	}
> +	__tasklet_action(a, list);
>  }
> 
> -
>  void tasklet_init(struct tasklet_struct *t,
>  		  void (*func)(unsigned long), unsigned long data)
>  {
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists