Re: mmotm 2014-02-05 list_lru_add lockdep splat

[Andrew Morton <akpm@linux-foundation.org>]

On Thu, 6 Feb 2014 11:41:36 -0500 Johannes Weiner <hannes@cmpxchg.org> wrote:

> 
> Make the shadow lru->node[i].lock IRQ-safe to remove the order
> dictated by interruption.  This slightly increases the IRQ-disabled
> section in the shadow shrinker, but it still drops all locks and
> enables IRQ after every reclaimed shadow radix tree node.
> 
> ...
>
> --- a/mm/workingset.c
> +++ b/mm/workingset.c
> @@ -273,7 +273,10 @@ static unsigned long count_shadow_nodes(struct shrinker *shrinker,
>  	unsigned long max_nodes;
>  	unsigned long pages;
>  
> +	local_irq_disable();
>  	shadow_nodes = list_lru_count_node(&workingset_shadow_nodes, sc->nid);
> +	local_irq_enable();

This is a bit ugly-looking.

A reader will look at that and wonder why the heck we're disabling
interrupts here.  Against what?  Is there some way in which we can
clarify this?

Perhaps adding list_lru_count_node_irq[save] and
list_lru_walk_node_irq[save] would be better - is it reasonable to
assume this is the only caller of the list_lru code which will ever
want irq-safe treatment?

This is all somewhat a side-effect of list_lru implementing its own
locking rather than requiring caller-provided locking.  It's always a
mistake.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>