On Sat 20-09-14 16:00:33, Johannes Weiner wrote:
> Charges currently pin the css indirectly by playing tricks during
> css_offline(): user pages stall the offlining process until all of
> them have been reparented, whereas kmemcg acquires a keep-alive
> reference if outstanding kernel pages are detected at that point.
>
> In preparation for removing all this complexity, make the pinning
> explicit and acquire a css references for every charged page.
OK, all the added {un}charges/atomics happen in a page_counter paths so
there shouldn't be any noticeable overhead.
I cannot judge the percpu counter part but the rest seems OK to me.
Two minor suggestions below.
> Signed-off-by: Johannes Weiner <hannes@xxxxxxxxxxx>
For the memcg part
Acked-by: Michal Hocko <mhocko@xxxxxxx>
> ---
> include/linux/cgroup.h | 26 +++++++++++++++++++++++++
> include/linux/percpu-refcount.h | 43 ++++++++++++++++++++++++++++++++---------
> mm/memcontrol.c | 17 +++++++++++++++-
> 3 files changed, 76 insertions(+), 10 deletions(-)
>
[...]
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index 154161bb7d4c..b832c87ec43b 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -2317,6 +2317,7 @@ static void drain_stock(struct memcg_stock_pcp *stock)
> page_counter_uncharge(&old->memory, stock->nr_pages);
> if (do_swap_account)
> page_counter_uncharge(&old->memsw, stock->nr_pages);
/* pairs with css_get_many in try_charge */
> + css_put_many(&old->css, stock->nr_pages);
> stock->nr_pages = 0;
> }
> stock->cached = NULL;
[...]
> @@ -2803,8 +2808,10 @@ static void memcg_uncharge_kmem(struct mem_cgroup *memcg,
> page_counter_uncharge(&memcg->memsw, nr_pages);
>
Wouldn't a single out_css_put be more readable? I was quite confused
when I start reading the patch before I saw the next hunk.
> /* Not down to 0 */
> - if (page_counter_uncharge(&memcg->kmem, nr_pages))
goto out_css_put;
> + if (page_counter_uncharge(&memcg->kmem, nr_pages)) {
> + css_put_many(&memcg->css, nr_pages);
> return;
> + }
>
> /*
> * Releases a reference taken in kmem_cgroup_css_offline in case
> @@ -2816,6 +2823,8 @@ static void memcg_uncharge_kmem(struct mem_cgroup *memcg,
> */
> if (memcg_kmem_test_and_clear_dead(memcg))
> css_put(&memcg->css);
> +
out_css_put:
> + css_put_many(&memcg->css, nr_pages);
> }
>
> /*
--
Michal Hocko
SUSE Labs
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>