| Message ID | 1546459533-36247-3-git-send-email-yang.shi@linux.alibaba.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [1/3] doc: memcontrol: fix the obsolete content about force empty | expand |
On Wed, Jan 2, 2019 at 12:06 PM Yang Shi <yang.shi@linux.alibaba.com> wrote: > > The typical usecase of force empty is to try to reclaim as much as > possible memory before offlining a memcg. Since there should be no > attached tasks to offlining memcg, the tasks anonymous pages would have > already been freed or uncharged. Anon pages can come from tmpfs files as well. > Even though anonymous pages get > swapped out, but they still get charged to swap space. So, it sounds > pointless to do swap for force empty. > I understand that force_empty is typically used before rmdir'ing a memcg but it might be used differently by some users. We use this interface to test memory reclaim behavior (anon and file). Anyways, I am not against changing the behavior, we can adapt internally but there might be other users using this interface differently. thanks, Shakeel
On 1/2/19 1:45 PM, Shakeel Butt wrote: > On Wed, Jan 2, 2019 at 12:06 PM Yang Shi <yang.shi@linux.alibaba.com> wrote: >> The typical usecase of force empty is to try to reclaim as much as >> possible memory before offlining a memcg. Since there should be no >> attached tasks to offlining memcg, the tasks anonymous pages would have >> already been freed or uncharged. > Anon pages can come from tmpfs files as well. Yes, but they are charged to swap space as regular anon pages. > >> Even though anonymous pages get >> swapped out, but they still get charged to swap space. So, it sounds >> pointless to do swap for force empty. >> > I understand that force_empty is typically used before rmdir'ing a > memcg but it might be used differently by some users. We use this > interface to test memory reclaim behavior (anon and file). Thanks for sharing your usecase. So, you uses this for test only? > > Anyways, I am not against changing the behavior, we can adapt > internally but there might be other users using this interface > differently. Thanks. Yang > > thanks, > Shakeel
On Thu, Jan 3, 2019 at 8:57 AM Yang Shi <yang.shi@linux.alibaba.com> wrote: > > > > On 1/2/19 1:45 PM, Shakeel Butt wrote: > > On Wed, Jan 2, 2019 at 12:06 PM Yang Shi <yang.shi@linux.alibaba.com> wrote: > >> The typical usecase of force empty is to try to reclaim as much as > >> possible memory before offlining a memcg. Since there should be no > >> attached tasks to offlining memcg, the tasks anonymous pages would have > >> already been freed or uncharged. > > Anon pages can come from tmpfs files as well. > > Yes, but they are charged to swap space as regular anon pages. > The point was the lifetime of tmpfs anon pages are not tied to any task. Even though there aren't any task attached to a memcg, the tmpfs anon pages will remain charged. Other than that, the old anon pages of a task which have migrated away might still be charged to the old memcg (if move_charge_at_immigrate is not set). > > > >> Even though anonymous pages get > >> swapped out, but they still get charged to swap space. So, it sounds > >> pointless to do swap for force empty. > >> > > I understand that force_empty is typically used before rmdir'ing a > > memcg but it might be used differently by some users. We use this > > interface to test memory reclaim behavior (anon and file). > > Thanks for sharing your usecase. So, you uses this for test only? > Yes. Shakeel
On 1/3/19 9:03 AM, Shakeel Butt wrote: > On Thu, Jan 3, 2019 at 8:57 AM Yang Shi <yang.shi@linux.alibaba.com> wrote: >> >> >> On 1/2/19 1:45 PM, Shakeel Butt wrote: >>> On Wed, Jan 2, 2019 at 12:06 PM Yang Shi <yang.shi@linux.alibaba.com> wrote: >>>> The typical usecase of force empty is to try to reclaim as much as >>>> possible memory before offlining a memcg. Since there should be no >>>> attached tasks to offlining memcg, the tasks anonymous pages would have >>>> already been freed or uncharged. >>> Anon pages can come from tmpfs files as well. >> Yes, but they are charged to swap space as regular anon pages. >> > The point was the lifetime of tmpfs anon pages are not tied to any > task. Even though there aren't any task attached to a memcg, the tmpfs > anon pages will remain charged. Other than that, the old anon pages of > a task which have migrated away might still be charged to the old > memcg (if move_charge_at_immigrate is not set). Yes, my understanding is even though they are swapped out but they are still charged to memsw for cgroupv1. force_empty is supposed to reclaim as much as possible memory, here I'm supposed "reclaim" also means "uncharge". Even though the anon pages are still charged to the old memcg, it will be moved the new memcg when the old one is deleted, or the pages will be just released if the task is killed. So, IMHO, I don't see the point why swapping anon pages when doing force_empty. Thanks, Yang >>>> Even though anonymous pages get >>>> swapped out, but they still get charged to swap space. So, it sounds >>>> pointless to do swap for force empty. >>>> >>> I understand that force_empty is typically used before rmdir'ing a >>> memcg but it might be used differently by some users. We use this >>> interface to test memory reclaim behavior (anon and file). >> Thanks for sharing your usecase. So, you uses this for test only? >> > Yes. > > Shakeel
diff --git a/mm/memcontrol.c b/mm/memcontrol.c index 6e1469b..bbf39b5 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -2872,7 +2872,7 @@ static int mem_cgroup_force_empty(struct mem_cgroup *memcg) return -EINTR; progress = try_to_free_mem_cgroup_pages(memcg, 1, - GFP_KERNEL, true); + GFP_KERNEL, false); if (!progress) { nr_retries--; /* maybe some writeback is necessary */
The typical usecase of force empty is to try to reclaim as much as possible memory before offlining a memcg. Since there should be no attached tasks to offlining memcg, the tasks anonymous pages would have already been freed or uncharged. Even though anonymous pages get swapped out, but they still get charged to swap space. So, it sounds pointless to do swap for force empty. I tried to dig into the history of this, it was introduced by commit 8c7c6e34a125 ("memcg: mem+swap controller core"), but there is not any clue about why it was done so at the first place. The below simple test script shows slight file cache reclaim improvement when swap is on. echo 3 > /proc/sys/vm/drop_caches mkdir /sys/fs/cgroup/memory/test echo 30 > /sys/fs/cgroup/memory/test/memory.swappiness echo $$ >/sys/fs/cgroup/memory/test/cgroup.procs cat /proc/meminfo | grep ^Cached|awk -F" " '{print $2}' dd if=/dev/zero of=/mnt/test bs=1M count=1024 ping localhost > /dev/null & echo 1 > /sys/fs/cgroup/memory/test/memory.force_empty killall ping echo $$ >/sys/fs/cgroup/memory/cgroup.procs cat /proc/meminfo | grep ^Cached|awk -F" " '{print $2}' rmdir /sys/fs/cgroup/memory/test cat /proc/meminfo | grep ^Cached|awk -F" " '{print $2}' The number of page cache is: w/o w/ before force empty 1088792 1088784 after force empty 41492 39428 reclaimed 1047300 1049356 Without doing swap, force empty can reclaim 2MB more memory in 1GB page cache. Cc: Michal Hocko <mhocko@suse.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> --- mm/memcontrol.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)