Re: [PATCH v4 2/2] fs/dcache.c: fix spin lockup issue on nlru->lock

From: Andrew Morton
Date: Thu Jun 29 2017 - 18:49:00 EST


On Thu, 29 Jun 2017 09:09:35 +0530 Sahitya Tummala <stummala@xxxxxxxxxxxxxx> wrote:

> __list_lru_walk_one() acquires nlru spin lock (nlru->lock) for
> longer duration if there are more number of items in the lru list.
> As per the current code, it can hold the spin lock for upto maximum
> UINT_MAX entries at a time. So if there are more number of items in
> the lru list, then "BUG: spinlock lockup suspected" is observed in
> the below path -
>
> ...
>
> Fix this lockup by reducing the number of entries to be shrinked
> from the lru list to 1024 at once. Also, add cond_resched() before
> processing the lru list again.
>
> ...
>
> --- a/fs/dcache.c
> +++ b/fs/dcache.c
> @@ -1133,11 +1133,12 @@ void shrink_dcache_sb(struct super_block *sb)
> LIST_HEAD(dispose);
>
> freed = list_lru_walk(&sb->s_dentry_lru,
> - dentry_lru_isolate_shrink, &dispose, UINT_MAX);
> + dentry_lru_isolate_shrink, &dispose, 1024);
>
> this_cpu_sub(nr_dentry_unused, freed);
> shrink_dentry_list(&dispose);
> - } while (freed > 0);
> + cond_resched();
> + } while (list_lru_count(&sb->s_dentry_lru) > 0);
> }
> EXPORT_SYMBOL(shrink_dcache_sb);

I'll add a cc:stable to this one - a large dentry list is a relatively
common thing.

I'm assumng that [1/2] does not need to be backported, OK?