On 25.09.2015 15:14, Vladimir Davydov wrote:
> Currently, we try to create a tcache node even if we are not going to
> actually store a page in tcache (tcache is disabled or we are called by
> global reclaimer). This is suboptimal, because we will drop the node
> anyway then.
>
> Signed-off-by: Vladimir Davydov <vdavy...@parallels.com>
Reviewed-by: Kirill Tkhai <ktk...@odin.com>
> ---
> mm/tcache.c | 9 +++++++--
> 1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/mm/tcache.c b/mm/tcache.c
> index b078c84f9564..ce5b0edd8253 100644
> --- a/mm/tcache.c
> +++ b/mm/tcache.c
> @@ -835,10 +835,15 @@ static void tcache_cleancache_put_page(int pool_id,
> {
> struct tcache_node *node;
> struct page *cache_page = NULL;
> + bool may_put = ACCESS_ONCE(tcache_active);
>
> - node = tcache_get_node_and_pool(pool_id, &key, true);
> + /* It makes no sense to populate tcache when we are short on memory */
> + if (current->flags & PF_MEMALLOC)
> + may_put = false;
> +
> + node = tcache_get_node_and_pool(pool_id, &key, may_put);
> if (node) {
> - if (tcache_active && !(current->flags & PF_MEMALLOC))
> + if (may_put)
> cache_page = tcache_alloc_page();
> if (cache_page) {
> copy_highpage(cache_page, page);
>
_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
