oom: suppress nodes that are not allowed from meminfo on oom kill

The oom killer is extremely verbose for machines with a large number of cpus and/or nodes. This verbosity can often be harmful if it causes other important messages to be scrolled from the kernel log and incurs a signicant time delay, specifically for kernels with CONFIG_NODES_SHIFT > 8. This patch causes only memory information to be displayed for nodes that are allowed by current's cpuset when dumping the VM state. Information for all other nodes is irrelevant to the oom condition; we don't care if there's an abundance of memory elsewhere if we can't access it. This only affects the behavior of dumping memory information when an oom is triggered. Other dumps, such as for sysrq+m, still display the unfiltered form when using the existing show_mem() interface. Additionally, the per-cpu pageset statistics are extremely verbose in oom killer output, so it is now suppressed. This removes nodes_weight(current->mems_allowed) * (1 + nr_cpus) lines from the oom killer output. Callers may use __show_mem(SHOW_MEM_FILTER_NODES) to filter disallowed nodes. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Mel Gorman <mel@csn.ul.ie> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

oom: suppress nodes that are not allowed from meminfo on oom kill
The oom killer is extremely verbose for machines with a large number of cpus and/or nodes. This verbosity can often be harmful if it causes other important messages to be scrolled from the kernel log and incurs a signicant time delay, specifically for kernels with CONFIG_NODES_SHIFT > 8. This patch causes only memory information to be displayed for nodes that are allowed by current's cpuset when dumping the VM state. Information for all other nodes is irrelevant to the oom condition; we don't care if there's an abundance of memory elsewhere if we can't access it. This only affects the behavior of dumping memory information when an oom is triggered. Other dumps, such as for sysrq+m, still display the unfiltered form when using the existing show_mem() interface. Additionally, the per-cpu pageset statistics are extremely verbose in oom killer output, so it is now suppressed. This removes nodes_weight(current->mems_allowed) * (1 + nr_cpus) lines from the oom killer output. Callers may use __show_mem(SHOW_MEM_FILTER_NODES) to filter disallowed nodes. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Mel Gorman <mel@csn.ul.ie> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
David Rientjes · Linus Torvalds
1 parent 94dcf29a11
Showing 4 changed files with 49 additions and 4 deletions Side-by-side Diff
include/linux/mm.h
lib/show_mem.c
mm/oom_kill.c
mm/page_alloc.c
@@ -859,7 +859,14 @@
  
 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
  
+/*
+ * Flags passed to __show_mem() and __show_free_areas() to suppress output in
+ * various contexts.
+ */
+#define SHOW_MEM_FILTER_NODES	(0x0001u)	/* filter disallowed nodes */
+
 extern void show_free_areas(void);
+extern void __show_free_areas(unsigned int flags);
  
 int shmem_lock(struct file *file, int lock, struct user_struct *user);
 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
@@ -1348,6 +1355,7 @@
 extern void mem_init(void);
 extern void __init mmap_init(void);
 extern void show_mem(void);
+extern void __show_mem(unsigned int flags);
 extern void si_meminfo(struct sysinfo * val);
 extern void si_meminfo_node(struct sysinfo *val, int nid);
 extern int after_bootmem;
@@ -9,14 +9,14 @@
 #include <linux/nmi.h>
 #include <linux/quicklist.h>
  
-void show_mem(void)
+void __show_mem(unsigned int filter)
 {
 	pg_data_t *pgdat;
 	unsigned long total = 0, reserved = 0, shared = 0,
 		nonshared = 0, highmem = 0;
  
 	printk("Mem-Info:\n");
-	show_free_areas();
+	__show_free_areas(filter);
  
 	for_each_online_pgdat(pgdat) {
 		unsigned long i, flags;
@@ -60,5 +60,10 @@
 	printk("%lu pages in pagetable cache\n",
 		quicklist_total_size());
 #endif
+}
+
+void show_mem(void)
+{
+	__show_mem(0);
 }
@@ -406,7 +406,7 @@
 	task_unlock(current);
 	dump_stack();
 	mem_cgroup_print_oom_info(mem, p);
-	show_mem();
+	__show_mem(SHOW_MEM_FILTER_NODES);
 	if (sysctl_oom_dump_tasks)
 		dump_tasks(mem, nodemask);
 }
@@ -2411,19 +2411,42 @@
 }
 #endif
  
+/*
+ * Determine whether the zone's node should be displayed or not, depending on
+ * whether SHOW_MEM_FILTER_NODES was passed to __show_free_areas().
+ */
+static bool skip_free_areas_zone(unsigned int flags, const struct zone *zone)
+{
+	bool ret = false;
+
+	if (!(flags & SHOW_MEM_FILTER_NODES))
+		goto out;
+
+	get_mems_allowed();
+	ret = !node_isset(zone->zone_pgdat->node_id,
+				cpuset_current_mems_allowed);
+	put_mems_allowed();
+out:
+	return ret;
+}
+
 #define K(x) ((x) << (PAGE_SHIFT-10))
  
 /*
  * 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.
  */
-void show_free_areas(void)
+void __show_free_areas(unsigned int filter)
 {
 	int cpu;
 	struct zone *zone;
  
 	for_each_populated_zone(zone) {
+		if (skip_free_areas_zone(filter, zone))
+			continue;
 		show_node(zone);
 		printk("%s per-cpu:\n", zone->name);
  
@@ -2465,6 +2488,8 @@
 	for_each_populated_zone(zone) {
 		int i;
  
+		if (skip_free_areas_zone(filter, zone))
+			continue;
 		show_node(zone);
 		printk("%s"
 			" free:%lukB"
@@ -2532,6 +2557,8 @@
 	for_each_populated_zone(zone) {
  		unsigned long nr[MAX_ORDER], flags, order, total = 0;
  
+		if (skip_free_areas_zone(filter, zone))
+			continue;
 		show_node(zone);
 		printk("%s: ", zone->name);
  
@@ -2549,6 +2576,11 @@
 	printk("%ld total pagecache pages\n", global_page_state(NR_FILE_PAGES));
  
 	show_swap_cache_info();
+}
+
+void show_free_areas(void)
+{
+	__show_free_areas(0);
 }
  
 static void zoneref_set_zone(struct zone *zone, struct zoneref *zoneref)
...	...	@@ -859,7 +859,14 @@
859	859
860	860	#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
861	861
	862	+/*
	863	+ * Flags passed to __show_mem() and __show_free_areas() to suppress output in
	864	+ * various contexts.
	865	+ */
	866	+#define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */
	867	+
862	868	extern void show_free_areas(void);
	869	+extern void __show_free_areas(unsigned int flags);
863	870
864	871	int shmem_lock(struct file file, int lock, struct user_struct user);
865	872	struct file shmem_file_setup(const char name, loff_t size, unsigned long flags);
...	...	@@ -1348,6 +1355,7 @@
1348	1355	extern void mem_init(void);
1349	1356	extern void __init mmap_init(void);
1350	1357	extern void show_mem(void);
	1358	+extern void __show_mem(unsigned int flags);
1351	1359	extern void si_meminfo(struct sysinfo * val);
1352	1360	extern void si_meminfo_node(struct sysinfo *val, int nid);
1353	1361	extern int after_bootmem;
...	...	@@ -9,14 +9,14 @@
9	9	#include <linux/nmi.h>
10	10	#include <linux/quicklist.h>
11	11
12		-void show_mem(void)
	12	+void __show_mem(unsigned int filter)
13	13	{
14	14	pg_data_t *pgdat;
15	15	unsigned long total = 0, reserved = 0, shared = 0,
16	16	nonshared = 0, highmem = 0;
17	17
18	18	printk("Mem-Info:\n");
19		- show_free_areas();
	19	+ __show_free_areas(filter);
20	20
21	21	for_each_online_pgdat(pgdat) {
22	22	unsigned long i, flags;
...	...	@@ -60,5 +60,10 @@
60	60	printk("%lu pages in pagetable cache\n",
61	61	quicklist_total_size());
62	62	#endif
	63	+}
	64	+
	65	+void show_mem(void)
	66	+{
	67	+ __show_mem(0);
63	68	}
...	...	@@ -406,7 +406,7 @@
406	406	task_unlock(current);
407	407	dump_stack();
408	408	mem_cgroup_print_oom_info(mem, p);
409		- show_mem();
	409	+ __show_mem(SHOW_MEM_FILTER_NODES);
410	410	if (sysctl_oom_dump_tasks)
411	411	dump_tasks(mem, nodemask);
412	412	}
...	...	@@ -2411,19 +2411,42 @@
2411	2411	}
2412	2412	#endif
2413	2413
	2414	+/*
	2415	+ * Determine whether the zone's node should be displayed or not, depending on
	2416	+ * whether SHOW_MEM_FILTER_NODES was passed to __show_free_areas().
	2417	+ */
	2418	+static bool skip_free_areas_zone(unsigned int flags, const struct zone *zone)
	2419	+{
	2420	+ bool ret = false;
	2421	+
	2422	+ if (!(flags & SHOW_MEM_FILTER_NODES))
	2423	+ goto out;
	2424	+
	2425	+ get_mems_allowed();
	2426	+ ret = !node_isset(zone->zone_pgdat->node_id,
	2427	+ cpuset_current_mems_allowed);
	2428	+ put_mems_allowed();
	2429	+out:
	2430	+ return ret;
	2431	+}
	2432	+
2414	2433	#define K(x) ((x) << (PAGE_SHIFT-10))
2415	2434
2416	2435	/*
2417	2436	* Show free area list (used inside shift_scroll-lock stuff)
2418	2437	* We also calculate the percentage fragmentation. We do this by counting the
2419	2438	* memory on each free list with the exception of the first item on the list.
	2439	+ * Suppresses nodes that are not allowed by current's cpuset if
	2440	+ * SHOW_MEM_FILTER_NODES is passed.
2420	2441	*/
2421		-void show_free_areas(void)
	2442	+void __show_free_areas(unsigned int filter)
2422	2443	{
2423	2444	int cpu;
2424	2445	struct zone *zone;
2425	2446
2426	2447	for_each_populated_zone(zone) {
	2448	+ if (skip_free_areas_zone(filter, zone))
	2449	+ continue;
2427	2450	show_node(zone);
2428	2451	printk("%s per-cpu:\n", zone->name);
2429	2452
...	...	@@ -2465,6 +2488,8 @@
2465	2488	for_each_populated_zone(zone) {
2466	2489	int i;
2467	2490
	2491	+ if (skip_free_areas_zone(filter, zone))
	2492	+ continue;
2468	2493	show_node(zone);
2469	2494	printk("%s"
2470	2495	" free:%lukB"
...	...	@@ -2532,6 +2557,8 @@
2532	2557	for_each_populated_zone(zone) {
2533	2558	unsigned long nr[MAX_ORDER], flags, order, total = 0;
2534	2559
	2560	+ if (skip_free_areas_zone(filter, zone))
	2561	+ continue;
2535	2562	show_node(zone);
2536	2563	printk("%s: ", zone->name);
2537	2564
...	...	@@ -2549,6 +2576,11 @@
2549	2576	printk("%ld total pagecache pages\n", global_page_state(NR_FILE_PAGES));
2550	2577
2551	2578	show_swap_cache_info();
	2579	+}
	2580	+
	2581	+void show_free_areas(void)
	2582	+{
	2583	+ __show_free_areas(0);
2552	2584	}
2553	2585
2554	2586	static void zoneref_set_zone(struct zone zone, struct zoneref zoneref)