Re: kmemleak not tainted

[Catalin Marinas]

On Tue, 2009-07-07 at 14:51 +0300, Sergey Senozhatsky wrote:
> kernel: [ 1917.133154] INFO: RCU detected CPU 0 stall (t=485140/3000 jiffies)

That's the relevant message. With CONFIG_RCU_CPU_STALL_DETECTOR you may
get these messages.

> static struct kmemleak_object *find_and_get_object(unsigned long ptr, int alias)
> {
> 	unsigned long flags;
> 	struct kmemleak_object *object = NULL;
> 
> 	rcu_read_lock();
> 	read_lock_irqsave(&kmemleak_lock, flags);
> 	if (ptr >= min_addr && ptr < max_addr)
> 		object = lookup_object(ptr, alias);
> >>	read_unlock_irqrestore(&kmemleak_lock, flags);
> 
> 	/* check whether the object is still available */
> 	if (object && !get_object(object))
> 		object = NULL;
> 	rcu_read_unlock();
> 
> 	return object;
> }

It just happened here because that's where the interrupts were enabled
and the timer routine invoked. The rcu-locked region above should be
pretty short (just a tree look-up).

What I think happens is that the kmemleak thread runs for several
seconds for scanning the memory and there may not be any context
switches. I have a patch to add more cond_resched() calls throughout the
kmemleak_scan() function which I hope will get merged. I don't get any
of these messages with CONFIG_PREEMPT enabled.

-- 
Catalin

--
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/