Re: [PATCH v2] memcg: add vmscan_stat

[KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>]

On Tue, 12 Jul 2011 16:02:02 -0700
Andrew Bresticker <abrestic@google.com> wrote:

> On Mon, Jul 11, 2011 at 3:30 AM, KAMEZAWA Hiroyuki <
> kamezawa.hiroyu@jp.fujitsu.com> wrote:
> 
> >
> > This patch is onto mmotm-0710... got bigger than expected ;(
> > ==
> > [PATCH] add memory.vmscan_stat
> >
> > commit log of commit 0ae5e89 " memcg: count the soft_limit reclaim in..."
> > says it adds scanning stats to memory.stat file. But it doesn't because
> > we considered we needed to make a concensus for such new APIs.
> >
> > This patch is a trial to add memory.scan_stat. This shows
> >  - the number of scanned pages(total, anon, file)
> >  - the number of rotated pages(total, anon, file)
> >  - the number of freed pages(total, anon, file)
> >  - the number of elaplsed time (including sleep/pause time)
> >
> >  for both of direct/soft reclaim.
> >
> > The biggest difference with oringinal Ying's one is that this file
> > can be reset by some write, as
> >
> >  # echo 0 ...../memory.scan_stat
> >
> > Example of output is here. This is a result after make -j 6 kernel
> > under 300M limit.
> >
> > [kamezawa@bluextal ~]$ cat /cgroup/memory/A/memory.scan_stat
> > [kamezawa@bluextal ~]$ cat /cgroup/memory/A/memory.vmscan_stat
> > scanned_pages_by_limit 9471864
> > scanned_anon_pages_by_limit 6640629
> > scanned_file_pages_by_limit 2831235
> > rotated_pages_by_limit 4243974
> > rotated_anon_pages_by_limit 3971968
> > rotated_file_pages_by_limit 272006
> > freed_pages_by_limit 2318492
> > freed_anon_pages_by_limit 962052
> > freed_file_pages_by_limit 1356440
> > elapsed_ns_by_limit 351386416101
> > scanned_pages_by_system 0
> > scanned_anon_pages_by_system 0
> > scanned_file_pages_by_system 0
> > rotated_pages_by_system 0
> > rotated_anon_pages_by_system 0
> > rotated_file_pages_by_system 0
> > freed_pages_by_system 0
> > freed_anon_pages_by_system 0
> > freed_file_pages_by_system 0
> > elapsed_ns_by_system 0
> > scanned_pages_by_limit_under_hierarchy 9471864
> > scanned_anon_pages_by_limit_under_hierarchy 6640629
> > scanned_file_pages_by_limit_under_hierarchy 2831235
> > rotated_pages_by_limit_under_hierarchy 4243974
> > rotated_anon_pages_by_limit_under_hierarchy 3971968
> > rotated_file_pages_by_limit_under_hierarchy 272006
> > freed_pages_by_limit_under_hierarchy 2318492
> > freed_anon_pages_by_limit_under_hierarchy 962052
> > freed_file_pages_by_limit_under_hierarchy 1356440
> > elapsed_ns_by_limit_under_hierarchy 351386416101
> > scanned_pages_by_system_under_hierarchy 0
> > scanned_anon_pages_by_system_under_hierarchy 0
> > scanned_file_pages_by_system_under_hierarchy 0
> > rotated_pages_by_system_under_hierarchy 0
> > rotated_anon_pages_by_system_under_hierarchy 0
> > rotated_file_pages_by_system_under_hierarchy 0
> > freed_pages_by_system_under_hierarchy 0
> > freed_anon_pages_by_system_under_hierarchy 0
> > freed_file_pages_by_system_under_hierarchy 0
> > elapsed_ns_by_system_under_hierarchy 0
> >
> >
> > total_xxxx is for hierarchy management.
> >
> > This will be useful for further memcg developments and need to be
> > developped before we do some complicated rework on LRU/softlimit
> > management.
> >
> > This patch adds a new struct memcg_scanrecord into scan_control struct.
> > sc->nr_scanned at el is not designed for exporting information. For
> > example,
> > nr_scanned is reset frequentrly and incremented +2 at scanning mapped
> > pages.
> >
> > For avoiding complexity, I added a new param in scan_control which is for
> > exporting scanning score.
> >
> > Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> >
> > Changelog:
> >  - renamed as vmscan_stat
> >  - handle file/anon
> >  - added "rotated"
> >  - changed names of param in vmscan_stat.
> > ---
> >  Documentation/cgroups/memory.txt |   85 +++++++++++++++++++
> >  include/linux/memcontrol.h       |   19 ++++
> >  include/linux/swap.h             |    6 -
> >  mm/memcontrol.c                  |  172
> > +++++++++++++++++++++++++++++++++++++--
> >  mm/vmscan.c                      |   39 +++++++-
> >  5 files changed, 303 insertions(+), 18 deletions(-)
> >
> > Index: mmotm-0710/Documentation/cgroups/memory.txt
> > ===================================================================
> > --- mmotm-0710.orig/Documentation/cgroups/memory.txt
> > +++ mmotm-0710/Documentation/cgroups/memory.txt
> > @@ -380,7 +380,7 @@ will be charged as a new owner of it.
> >
> >  5.2 stat file
> >
> > -memory.stat file includes following statistics
> > +5.2.1 memory.stat file includes following statistics
> >
> >  # per-memory cgroup local status
> >  cache          - # of bytes of page cache memory.
> > @@ -438,6 +438,89 @@ Note:
> >         file_mapped is accounted only when the memory cgroup is owner of
> > page
> >         cache.)
> >
> > +5.2.2 memory.vmscan_stat
> > +
> > +memory.vmscan_stat includes statistics information for memory scanning and
> > +freeing, reclaiming. The statistics shows memory scanning information
> > since
> > +memory cgroup creation and can be reset to 0 by writing 0 as
> > +
> > + #echo 0 > ../memory.vmscan_stat
> > +
> > +This file contains following statistics.
> > +
> > +[param]_[file_or_anon]_pages_by_[reason]_[under_heararchy]
> > +[param]_elapsed_ns_by_[reason]_[under_hierarchy]
> > +
> > +For example,
> > +
> > +  scanned_file_pages_by_limit indicates the number of scanned
> > +  file pages at vmscan.
> > +
> > +Now, 3 parameters are supported
> > +
> > +  scanned - the number of pages scanned by vmscan
> > +  rotated - the number of pages activated at vmscan
> > +  freed   - the number of pages freed by vmscan
> > +
> > +If "rotated" is high against scanned/freed, the memcg seems busy.
> > +
> > +Now, 2 reason are supported
> > +
> > +  limit - the memory cgroup's limit
> > +  system - global memory pressure + softlimit
> > +           (global memory pressure not under softlimit is not handled now)
> > +
> > +When under_hierarchy is added in the tail, the number indicates the
> > +total memcg scan of its children and itself.
> > +
> > +elapsed_ns is a elapsed time in nanosecond. This may include sleep time
> > +and not indicates CPU usage. So, please take this as just showing
> > +latency.
> > +
> > +Here is an example.
> > +
> > +# cat /cgroup/memory/A/memory.vmscan_stat
> > +scanned_pages_by_limit 9471864
> > +scanned_anon_pages_by_limit 6640629
> > +scanned_file_pages_by_limit 2831235
> > +rotated_pages_by_limit 4243974
> > +rotated_anon_pages_by_limit 3971968
> > +rotated_file_pages_by_limit 272006
> > +freed_pages_by_limit 2318492
> > +freed_anon_pages_by_limit 962052
> > +freed_file_pages_by_limit 1356440
> > +elapsed_ns_by_limit 351386416101
> > +scanned_pages_by_system 0
> > +scanned_anon_pages_by_system 0
> > +scanned_file_pages_by_system 0
> > +rotated_pages_by_system 0
> > +rotated_anon_pages_by_system 0
> > +rotated_file_pages_by_system 0
> > +freed_pages_by_system 0
> > +freed_anon_pages_by_system 0
> > +freed_file_pages_by_system 0
> > +elapsed_ns_by_system 0
> > +scanned_pages_by_limit_under_hierarchy 9471864
> > +scanned_anon_pages_by_limit_under_hierarchy 6640629
> > +scanned_file_pages_by_limit_under_hierarchy 2831235
> > +rotated_pages_by_limit_under_hierarchy 4243974
> > +rotated_anon_pages_by_limit_under_hierarchy 3971968
> > +rotated_file_pages_by_limit_under_hierarchy 272006
> > +freed_pages_by_limit_under_hierarchy 2318492
> > +freed_anon_pages_by_limit_under_hierarchy 962052
> > +freed_file_pages_by_limit_under_hierarchy 1356440
> > +elapsed_ns_by_limit_under_hierarchy 351386416101
> > +scanned_pages_by_system_under_hierarchy 0
> > +scanned_anon_pages_by_system_under_hierarchy 0
> > +scanned_file_pages_by_system_under_hierarchy 0
> > +rotated_pages_by_system_under_hierarchy 0
> > +rotated_anon_pages_by_system_under_hierarchy 0
> > +rotated_file_pages_by_system_under_hierarchy 0
> > +freed_pages_by_system_under_hierarchy 0
> > +freed_anon_pages_by_system_under_hierarchy 0
> > +freed_file_pages_by_system_under_hierarchy 0
> > +elapsed_ns_by_system_under_hierarchy 0
> > +
> >  5.3 swappiness
> >
> >  Similar to /proc/sys/vm/swappiness, but affecting a hierarchy of groups
> > only.
> > Index: mmotm-0710/include/linux/memcontrol.h
> > ===================================================================
> > --- mmotm-0710.orig/include/linux/memcontrol.h
> > +++ mmotm-0710/include/linux/memcontrol.h
> > @@ -39,6 +39,16 @@ extern unsigned long mem_cgroup_isolate_
> >                                        struct mem_cgroup *mem_cont,
> >                                        int active, int file);
> >
> > +struct memcg_scanrecord {
> > +       struct mem_cgroup *mem; /* scanend memory cgroup */
> > +       struct mem_cgroup *root; /* scan target hierarchy root */
> > +       int context;            /* scanning context (see memcontrol.c) */
> > +       unsigned long nr_scanned[2]; /* the number of scanned pages */
> > +       unsigned long nr_rotated[2]; /* the number of rotated pages */
> > +       unsigned long nr_freed[2]; /* the number of freed pages */
> > +       unsigned long elapsed; /* nsec of time elapsed while scanning */
> > +};
> > +
> >  #ifdef CONFIG_CGROUP_MEM_RES_CTLR
> >  /*
> >  * All "charge" functions with gfp_mask should use GFP_KERNEL or
> > @@ -117,6 +127,15 @@ mem_cgroup_get_reclaim_stat_from_page(st
> >  extern void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
> >                                        struct task_struct *p);
> >
> > +extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *mem,
> > +                                                 gfp_t gfp_mask, bool
> > noswap,
> > +                                                 struct memcg_scanrecord
> > *rec);
> > +extern unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *mem,
> > +                                               gfp_t gfp_mask, bool
> > noswap,
> > +                                               struct zone *zone,
> > +                                               struct memcg_scanrecord
> > *rec,
> > +                                               unsigned long *nr_scanned);
> > +
> >  #ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
> >  extern int do_swap_account;
> >  #endif
> > Index: mmotm-0710/include/linux/swap.h
> > ===================================================================
> > --- mmotm-0710.orig/include/linux/swap.h
> > +++ mmotm-0710/include/linux/swap.h
> > @@ -253,12 +253,6 @@ static inline void lru_cache_add_file(st
> >  /* linux/mm/vmscan.c */
> >  extern unsigned long try_to_free_pages(struct zonelist *zonelist, int
> > order,
> >                                        gfp_t gfp_mask, nodemask_t *mask);
> > -extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *mem,
> > -                                                 gfp_t gfp_mask, bool
> > noswap);
> > -extern unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *mem,
> > -                                               gfp_t gfp_mask, bool
> > noswap,
> > -                                               struct zone *zone,
> > -                                               unsigned long *nr_scanned);
> >  extern int __isolate_lru_page(struct page *page, int mode, int file);
> >  extern unsigned long shrink_all_memory(unsigned long nr_pages);
> >  extern int vm_swappiness;
> > Index: mmotm-0710/mm/memcontrol.c
> > ===================================================================
> > --- mmotm-0710.orig/mm/memcontrol.c
> > +++ mmotm-0710/mm/memcontrol.c
> > @@ -204,6 +204,50 @@ struct mem_cgroup_eventfd_list {
> >  static void mem_cgroup_threshold(struct mem_cgroup *mem);
> >  static void mem_cgroup_oom_notify(struct mem_cgroup *mem);
> >
> > +enum {
> > +       SCAN_BY_LIMIT,
> > +       SCAN_BY_SYSTEM,
> > +       NR_SCAN_CONTEXT,
> > +       SCAN_BY_SHRINK, /* not recorded now */
> > +};
> > +
> > +enum {
> > +       SCAN,
> > +       SCAN_ANON,
> > +       SCAN_FILE,
> > +       ROTATE,
> > +       ROTATE_ANON,
> > +       ROTATE_FILE,
> > +       FREED,
> > +       FREED_ANON,
> > +       FREED_FILE,
> > +       ELAPSED,
> > +       NR_SCANSTATS,
> > +};
> > +
> > +struct scanstat {
> > +       spinlock_t      lock;
> > +       unsigned long   stats[NR_SCAN_CONTEXT][NR_SCANSTATS];
> > +       unsigned long   rootstats[NR_SCAN_CONTEXT][NR_SCANSTATS];
> > +};
> >
> 
> I'm working on a similar effort with Ying here at Google and so far we've
> been using per-cpu counters for these statistics instead of spin-lock
> protected counters.  Clearly the spin-lock protected counters have less
> memory overhead and make reading the stat file faster, but our concern is
> that this method is inconsistent with the other memory stat files such
> /proc/vmstat and /dev/cgroup/memory/.../memory.stat.  Is there any
> particular reason you chose to use spin-lock protected counters instead of
> per-cpu counters?
> 

In my experience, if we do "batch" enouch, it works always better than
percpu-counter. percpu counter is effective when batching is difficult.
This patch's implementation does enough batching and it's much coarse
grained than percpu counter. Then, this patch is better than percpu.


> I've also modified your patch to use per-cpu counters instead of spin-lock
> protected counters.  I tested it by doing streaming I/O from a ramdisk:
> 
> $ mke2fs /dev/ram1
> $ mkdir /tmp/swapram
> $ mkdir /tmp/swapram/ram1
> $ mount -t ext2 /dev/ram1 /tmp/swapram/ram1
> $ dd if=/dev/urandom of=/tmp/swapram/ram1/file_16m bs=4096 count=4096
> $ mkdir /dev/cgroup/memory/1
> $ echo 8m > /dev/cgroup/memory/1
> $ ./ramdisk_load.sh 7
> $ echo $$ > /dev/cgroup/memory/1/tasks
> $ time for ((z=0; z<=2000; z++)); do cat /tmp/swapram/ram1/file_16m >
> /dev/zero; done
> 
> Where ramdisk_load.sh is:
> for ((i=0; i<=$1; i++))
> do
>   echo $$ >/dev/cgroup/memory/1/tasks
>   for ((z=0; z<=2000; z++)); do cat /tmp/swapram/ram1/file_16m > /dev/zero;
> done &
> done
> 
> Surprisingly, the per-cpu counters perform worse than the spin-lock
> protected counters.  Over 10 runs of the test above, the per-cpu counters
> were 1.60% slower in both real time and sys time.  I'm wondering if you have
> any insight as to why this is.  I can provide my diff against your patch if
> necessary.
> 

The percpu counte works effectively only when we use +1/-1 at each change of
counters. It uses "batch" to merge the per-cpu value to the counter.
I think you use default "batch" value but the scan/rotate/free/elapsed value
is always larger than "batch" and you just added memory overhead and "if"
to pure spinlock counters.

Determining this "batch" threshold for percpu counter is difficult.

Thanks,
-Kame


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>