[Devel] [PATCH rh7 2/3] mm/tcache: rebalance tree is isolation failed.

[Kirill Tkhai]

On 18.10.2018 12:02, Andrey Ryabinin wrote:
> tcache_lru_isolate() doesn't rebalance reclaim tree if isolation
> failed. If the first pool in tree is empty, without rebalancing it
> may stay first indefinitely long preventing reclaim.
> 
> https://jira.sw.ru/browse/PSBM-89403
> Signed-off-by: Andrey Ryabinin <aryabinin at virtuozzo.com>

1 and 2 are brilliant :) 3 haven't seen yet.

Acked-by: Kirill Tkhai <ktkhai at virtuozzo.com>

> ---
>  mm/tcache.c | 5 +----
>  1 file changed, 1 insertion(+), 4 deletions(-)
> 
> diff --git a/mm/tcache.c b/mm/tcache.c
> index 960450bb825e..eb9c9dea4e51 100644
> --- a/mm/tcache.c
> +++ b/mm/tcache.c
> @@ -1109,9 +1109,6 @@ tcache_lru_isolate(int nid, struct page **pages, int nr_to_isolate)
>  	spin_lock_irq(&pni->lock);
>  	nr_isolated = __tcache_lru_isolate(pni, pages, nr_to_isolate);
>  
> -	if (!nr_isolated)
> -		goto unlock;
> -
>  	if (!RB_EMPTY_NODE(rbn) || !list_empty(&pni->lru)) {
>  		spin_lock(&ni->lock);
>  		if (!RB_EMPTY_NODE(rbn))
> @@ -1123,7 +1120,7 @@ tcache_lru_isolate(int nid, struct page **pages, int nr_to_isolate)
>  		update_ni_rb_first(ni);
>  		spin_unlock(&ni->lock);
>  	}
> -unlock:
> +
>  	spin_unlock_irq(&pni->lock);
>  	tcache_put_pool(pni->pool);
>  out:
>