On Thu, Jul 16, 2020 at 2:28 PM David Rientjes <rientjes@xxxxxxxxxx> wrote:
>
> On Thu, 16 Jul 2020, Shakeel Butt wrote:
>
> > > Userspace can lack insight into the amount of memory that can be reclaimed
> > > from a memcg based on values from memory.stat. Two specific examples:
> > >
> > > - Lazy freeable memory (MADV_FREE) that are clean anonymous pages on the
> > > inactive file LRU that can be quickly reclaimed under memory pressure
> > > but otherwise shows up as mapped anon in memory.stat, and
> > >
> > > - Memory on deferred split queues (thp) that are compound pages that can
> > > be split and uncharged from the memcg under memory pressure, but
> > > otherwise shows up as charged anon LRU memory in memory.stat.
> > >
> > > Both of this anonymous usage is also charged to memory.current.
> > >
> > > Userspace can currently derive this information but it depends on kernel
> > > implementation details for how this memory is handled for the purposes of
> > > reclaim (anon on inactive file LRU or unmapped anon on the LRU).
> > >
> > > For the purposes of writing portable userspace code that does not need to
> > > have insight into the kernel implementation for reclaimable memory, this
> > > exports a stat that reveals the amount of anonymous memory that can be
> > > reclaimed and uncharged from the memcg to start new applications.
> > >
> > > As the kernel implementation evolves for memory that can be reclaimed
> > > under memory pressure, this stat can be kept consistent.
> > >
> > > Signed-off-by: David Rientjes <rientjes@xxxxxxxxxx>
> > > ---
> > > Documentation/admin-guide/cgroup-v2.rst | 6 +++++
> > > mm/memcontrol.c | 31 +++++++++++++++++++++++++
> > > 2 files changed, 37 insertions(+)
> > >
> > > diff --git a/Documentation/admin-guide/cgroup-v2.rst b/Documentation/admin-guide/cgroup-v2.rst
> > > --- a/Documentation/admin-guide/cgroup-v2.rst
> > > +++ b/Documentation/admin-guide/cgroup-v2.rst
> > > @@ -1296,6 +1296,12 @@ PAGE_SIZE multiple when read back.
> > > Amount of memory used in anonymous mappings backed by
> > > transparent hugepages
> > >
> > > + anon_reclaimable
> > > + The amount of charged anonymous memory that can be reclaimed
> > > + under memory pressure without swap. This currently includes
> > > + lazy freeable memory (MADV_FREE) and compound pages that can be
> > > + split and uncharged.
> > > +
> > > inactive_anon, active_anon, inactive_file, active_file, unevictable
> > > Amount of memory, swap-backed and filesystem-backed,
> > > on the internal memory management lists used by the
> > > diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> > > --- a/mm/memcontrol.c
> > > +++ b/mm/memcontrol.c
> > > @@ -1350,6 +1350,32 @@ static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
> > > return false;
> > > }
> > >
> > > +/*
> > > + * Returns the amount of anon memory that is charged to the memcg that is
> > > + * reclaimable under memory pressure without swap, in pages.
> > > + */
> > > +static unsigned long memcg_anon_reclaimable(struct mem_cgroup *memcg)
> > > +{
> > > + long deferred, lazyfree;
> > > +
> > > + /*
> > > + * Deferred pages are charged anonymous pages that are on the LRU but
> > > + * are unmapped. These compound pages are split under memory pressure.
> > > + */
> > > + deferred = max_t(long, memcg_page_state(memcg, NR_ACTIVE_ANON) +
> > > + memcg_page_state(memcg, NR_INACTIVE_ANON) -
> > > + memcg_page_state(memcg, NR_ANON_MAPPED), 0);
> >
> > Please note that the NR_ANON_MAPPED does not include tmpfs memory but
> > NR_[IN]ACTIVE_ANON does include the tmpfs.
> >
> > > + /*
> > > + * Lazyfree pages are charged clean anonymous pages that are on the file
> > > + * LRU and can be reclaimed under memory pressure.
> > > + */
> > > + lazyfree = max_t(long, memcg_page_state(memcg, NR_ACTIVE_FILE) +
> > > + memcg_page_state(memcg, NR_INACTIVE_FILE) -
> > > + memcg_page_state(memcg, NR_FILE_PAGES), 0);
> >
> > Similarly NR_FILE_PAGES includes tmpfs memory but NR_[IN]ACTIVE_FILE does not.
> >
>
> Ah, so this adds to the motivation of providing the anon_reclaimable stat
> because the calculation becomes even more convoluted and completely based
> on the kernel implementation details for both lazyfree memory and deferred
> split queues.
Yes, I agree.
> Did you have a calculation in mind for
> memcg_anon_reclaimable()?
For deferred, "memcg->deferred_split_queue.split_queue_len" should be usable.
For lazyfree, NR_ACTIVE_FILE + NR_INACTIVE_FILE + NR_SHMEM -
NR_FILE_PAGES seems like the right formula.
