mm: hide per-cpu lists in output of show_mem()
authorKonstantin Khlebnikov <khlebnikov@yandex-team.ru>
Tue, 14 Apr 2015 22:45:30 +0000 (15:45 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 14 Apr 2015 23:49:01 +0000 (16:49 -0700)
This makes show_mem() much less verbose on huge machines.  Instead of huge
and almost useless dump of counters for each per-zone per-cpu lists this
patch prints the sum of these counters for each zone (free_pcp) and size
of per-cpu list for current cpu (local_pcp).

The filter flag SHOW_MEM_PERCPU_LISTS reverts to the old verbose mode.

[akpm@linux-foundation.org: update show_free_areas comment]
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Acked-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/linux/mm.h
mm/page_alloc.c

index 6571dd7..9c21b42 100644 (file)
@@ -1126,6 +1126,7 @@ extern void pagefault_out_of_memory(void);
  * various contexts.
  */
 #define SHOW_MEM_FILTER_NODES          (0x0001u)       /* disallowed nodes */
+#define SHOW_MEM_PERCPU_LISTS          (0x0002u)       /* per-zone per-cpu */
 
 extern void show_free_areas(unsigned int flags);
 extern bool skip_free_areas_node(unsigned int flags, int nid);
index 6dfa5b2..eab8e20 100644 (file)
@@ -3251,25 +3251,37 @@ static void show_migration_types(unsigned char type)
  * Show free area list (used inside shift_scroll-lock stuff)
  * We also calculate the percentage fragmentation. We do this by counting the
  * memory on each free list with the exception of the first item on the list.
- * Suppresses nodes that are not allowed by current's cpuset if
- * SHOW_MEM_FILTER_NODES is passed.
+ *
+ * Bits in @filter:
+ * SHOW_MEM_FILTER_NODES: suppress nodes that are not allowed by current's
+ *   cpuset.
+ * SHOW_MEM_PERCPU_LISTS: display full per-node per-cpu pcp lists
  */
 void show_free_areas(unsigned int filter)
 {
+       unsigned long free_pcp = 0;
        int cpu;
        struct zone *zone;
 
        for_each_populated_zone(zone) {
                if (skip_free_areas_node(filter, zone_to_nid(zone)))
                        continue;
-               show_node(zone);
-               printk("%s per-cpu:\n", zone->name);
+
+               if (filter & SHOW_MEM_PERCPU_LISTS) {
+                       show_node(zone);
+                       printk("%s per-cpu:\n", zone->name);
+               }
 
                for_each_online_cpu(cpu) {
                        struct per_cpu_pageset *pageset;
 
                        pageset = per_cpu_ptr(zone->pageset, cpu);
 
+                       free_pcp += pageset->pcp.count;
+
+                       if (!(filter & SHOW_MEM_PERCPU_LISTS))
+                               continue;
+
                        printk("CPU %4d: hi:%5d, btch:%4d usd:%4d\n",
                               cpu, pageset->pcp.high,
                               pageset->pcp.batch, pageset->pcp.count);
@@ -3278,11 +3290,10 @@ void show_free_areas(unsigned int filter)
 
        printk("active_anon:%lu inactive_anon:%lu isolated_anon:%lu\n"
                " active_file:%lu inactive_file:%lu isolated_file:%lu\n"
-               " unevictable:%lu"
-               " dirty:%lu writeback:%lu unstable:%lu\n"
-               " free:%lu slab_reclaimable:%lu slab_unreclaimable:%lu\n"
+               " unevictable:%lu dirty:%lu writeback:%lu unstable:%lu\n"
+               " slab_reclaimable:%lu slab_unreclaimable:%lu\n"
                " mapped:%lu shmem:%lu pagetables:%lu bounce:%lu\n"
-               " free_cma:%lu\n",
+               " free:%lu free_pcp:%lu free_cma:%lu\n",
                global_page_state(NR_ACTIVE_ANON),
                global_page_state(NR_INACTIVE_ANON),
                global_page_state(NR_ISOLATED_ANON),
@@ -3293,13 +3304,14 @@ void show_free_areas(unsigned int filter)
                global_page_state(NR_FILE_DIRTY),
                global_page_state(NR_WRITEBACK),
                global_page_state(NR_UNSTABLE_NFS),
-               global_page_state(NR_FREE_PAGES),
                global_page_state(NR_SLAB_RECLAIMABLE),
                global_page_state(NR_SLAB_UNRECLAIMABLE),
                global_page_state(NR_FILE_MAPPED),
                global_page_state(NR_SHMEM),
                global_page_state(NR_PAGETABLE),
                global_page_state(NR_BOUNCE),
+               global_page_state(NR_FREE_PAGES),
+               free_pcp,
                global_page_state(NR_FREE_CMA_PAGES));
 
        for_each_populated_zone(zone) {
@@ -3307,6 +3319,11 @@ void show_free_areas(unsigned int filter)
 
                if (skip_free_areas_node(filter, zone_to_nid(zone)))
                        continue;
+
+               free_pcp = 0;
+               for_each_online_cpu(cpu)
+                       free_pcp += per_cpu_ptr(zone->pageset, cpu)->pcp.count;
+
                show_node(zone);
                printk("%s"
                        " free:%lukB"
@@ -3333,6 +3350,8 @@ void show_free_areas(unsigned int filter)
                        " pagetables:%lukB"
                        " unstable:%lukB"
                        " bounce:%lukB"
+                       " free_pcp:%lukB"
+                       " local_pcp:%ukB"
                        " free_cma:%lukB"
                        " writeback_tmp:%lukB"
                        " pages_scanned:%lu"
@@ -3364,6 +3383,8 @@ void show_free_areas(unsigned int filter)
                        K(zone_page_state(zone, NR_PAGETABLE)),
                        K(zone_page_state(zone, NR_UNSTABLE_NFS)),
                        K(zone_page_state(zone, NR_BOUNCE)),
+                       K(free_pcp),
+                       K(this_cpu_read(zone->pageset->pcp.count)),
                        K(zone_page_state(zone, NR_FREE_CMA_PAGES)),
                        K(zone_page_state(zone, NR_WRITEBACK_TEMP)),
                        K(zone_page_state(zone, NR_PAGES_SCANNED)),