mm: list_lru: transpose the array of per-node per-memcg lru lists

mm: list_lru: transpose the array of per-node per-memcg lru lists

Patch series "Optimize list lru memory consumption", v6.

In our server, we found a suspected memory leak problem.  The kmalloc-32
consumes more than 6GB of memory.  Other kmem_caches consume less than
2GB memory.

After our in-depth analysis, the memory consumption of kmalloc-32 slab
cache is the cause of list_lru_one allocation.

  crash> p
  memcg_nr_cache_ids memcg_nr_cache_ids = $2 = 24574

memcg_nr_cache_ids is very large and memory consumption of each list_lru
can be calculated with the following formula.

  num_numa_node * memcg_nr_cache_ids * 32 (kmalloc-32)

There are 4 numa nodes in our system, so each list_lru consumes ~3MB.

  crash> list super_blocks | wc -l
  952

Every mount will register 2 list lrus, one is for inode, another is for
dentry.  There are 952 super_blocks.  So the total memory is 952 * 2 * 3
MB (~5.6GB).  But now the number of memory cgroups is less than 500.  So
I guess more than 12286 memory cgroups have been created on this machine
(I do not know why there are so many cgroups, it may be a user's bug or
the user really want to do that).  Because memcg_nr_cache_ids has not
been reduced to a suitable value.  It leads to waste a lot of memory.
If we want to reduce memcg_nr_cache_ids, we have to *reboot* the server.
This is not what we want.

In order to reduce memcg_nr_cache_ids, I had posted a patchset [1] to do
this.  But this did not fundamentally solve the problem.

We currently allocate scope for every memcg to be able to tracked on
every superblock instantiated in the system, regardless of whether that
superblock is even accessible to that memcg.

These huge memcg counts come from container hosts where memcgs are
confined to just a small subset of the total number of superblocks that
instantiated at any given point in time.

For these systems with huge container counts, list_lru does not need the
capability of tracking every memcg on every superblock.

What it comes down to is that the list_lru is only needed for a given
memcg if that memcg is instatiating and freeing objects on a given
list_lru.

As Dave said, "Which makes me think we should be moving more towards 'add
the memcg to the list_lru at the first insert' model rather than
'instantiate all at memcg init time just in case'."

This patchset aims to optimize the list lru memory consumption from
different aspects.

I had done a easy test to show the optimization.  I create 10k memory
cgroups and mount 10k filesystems in the systems.  We use free command to
show how many memory does the systems comsumes after this operation (There
are 2 numa nodes in the system).

        +-----------------------+------------------------+
        |      condition        |   memory consumption   |
        +-----------------------+------------------------+
        | without this patchset |        24464 MB        |
        +-----------------------+------------------------+
        |     after patch 1     |        21957 MB        | <--------+
        +-----------------------+------------------------+          |
        |     after patch 10    |         6895 MB        |          |
        +-----------------------+------------------------+          |
        |     after patch 12    |         4367 MB        |          |
        +-----------------------+------------------------+          |
                                                                    |
        The more the number of nodes, the more obvious the effect---+

BTW, there was a recent discussion [2] on the same issue.

[1] https://lore.kernel.org/all/20210428094949.43579-1-songmuchun@bytedance.com/
[2] https://lore.kernel.org/all/20210405054848.GA1077931@in.ibm.com/

This series not only optimizes the memory usage of list_lru but also
simplifies the code.

This patch (of 16):

The current scheme of maintaining per-node per-memcg lru lists looks like:
  struct list_lru {
    struct list_lru_node *node;           (for each node)
      struct list_lru_memcg *memcg_lrus;
        struct list_lru_one *lru[];       (for each memcg)
  }

By effectively transposing the two-dimension array of list_lru_one's structures
(per-node per-memcg => per-memcg per-node) it's possible to save some memory
and simplify alloc/dealloc paths. The new scheme looks like:
  struct list_lru {
    struct list_lru_memcg *mlrus;
      struct list_lru_per_memcg *mlru[];  (for each memcg)
        struct list_lru_one node[0];      (for each node)
  }

Memory savings are coming from not only 'struct rcu_head' but also some
pointer arrays used to store the pointer to 'struct list_lru_one'.  The
array is per node and its size is 8 (a pointer) * num_memcgs.  So the
total size of the arrays is 8 * num_nodes * memcg_nr_cache_ids.  After
this patch, the size becomes 8 * memcg_nr_cache_ids.

Link: https://lkml.kernel.org/r/20220228122126.37293-1-songmuchun@bytedance.com
Link: https://lkml.kernel.org/r/20220228122126.37293-2-songmuchun@bytedance.com
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Alex Shi <alexs@kernel.org>
Cc: Wei Yang <richard.weiyang@gmail.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Trond Myklebust <trond.myklebust@hammerspace.com>
Cc: Anna Schumaker <Anna.Schumaker@Netapp.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Chao Yu <chao@kernel.org>
Cc: Kari Argillander <kari.argillander@gmail.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Xiongchun Duan <duanxiongchun@bytedance.com>
Cc: Fam Zheng <fam.zheng@bytedance.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>