From: Michal Hocko <mhocko@xxxxxxxx>
Both oom and the allocation failure reports are not providing any
information about the compaction/migration counters which might give us
a clue what went wrong and why we are OOM for the particular order -
e.g. the compaction fails constantly because it cannot isolate any pages
or that the migration fails. So far we have been asking for /proc/vmstat
content before and after the OOM which is rather clumsy, especially when
the OOM is not 100% reproducible.
Extend show_mem() to understand a new filter (SHOW_COMPACTION_STATS)
which is enabled only for higer order paths.
Signed-off-by: Michal Hocko <mhocko@xxxxxxxx>
---
include/linux/mm.h | 1 +
lib/show_mem.c | 14 ++++++++++++++
mm/oom_kill.c | 2 +-
mm/page_alloc.c | 2 ++
4 files changed, 18 insertions(+), 1 deletion(-)
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 7e44613c5078..b4859547acc4 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1146,6 +1146,7 @@ extern void pagefault_out_of_memory(void);
* various contexts.
*/
#define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
+#define SHOW_COMPACTION_STATS (0x0002u)
extern void show_free_areas(unsigned int flags);
extern bool skip_free_areas_node(unsigned int flags, int nid);
diff --git a/lib/show_mem.c b/lib/show_mem.c
index 1feed6a2b12a..595ba00f5836 100644
--- a/lib/show_mem.c
+++ b/lib/show_mem.c
@@ -8,6 +8,7 @@
#include <linux/mm.h>
#include <linux/quicklist.h>
#include <linux/cma.h>
+#include <linux/vm_event_item.h>
void show_mem(unsigned int filter)
{
@@ -17,6 +18,19 @@ void show_mem(unsigned int filter)
printk("Mem-Info:\n");
show_free_areas(filter);
+#ifdef CONFIG_COMPACTION
+ if (filter & SHOW_COMPACTION_STATS) {
+ printk("compaction_stall:%lu compaction_fail:%lu "
+ "compact_migrate_scanned:%lu compact_free_scanned:%lu "
+ "compact_isolated:%lu "
+ "pgmigrate_success:%lu pgmigrate_fail:%lu\n",
+ global_page_state(COMPACTSTALL), global_page_state(COMPACTFAIL),
+ global_page_state(COMPACTMIGRATE_SCANNED), global_page_state(COMPACTFAIL),
+ global_page_state(COMPACTISOLATED),
+ global_page_state(PGMIGRATE_SUCCESS), global_page_state(PGMIGRATE_FAIL));
+ }
+#endif
+
for_each_online_pgdat(pgdat) {
unsigned long flags;
int zoneid;
diff --git a/mm/oom_kill.c b/mm/oom_kill.c
index 463cdd22d4e0..5e7a09f4dbc9 100644
--- a/mm/oom_kill.c
+++ b/mm/oom_kill.c
@@ -419,7 +419,7 @@ static void dump_header(struct oom_control *oc, struct task_struct *p)
if (oc->memcg)
mem_cgroup_print_oom_info(oc->memcg, p);
else
- show_mem(SHOW_MEM_FILTER_NODES);
+ show_mem(SHOW_MEM_FILTER_NODES | (oc->order)?SHOW_COMPACTION_STATS:0);
if (sysctl_oom_dump_tasks)
dump_tasks(oc->memcg, oc->nodemask);
}
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 9d46b65061be..adf0cb655827 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -2999,6 +2999,8 @@ void warn_alloc_failed(gfp_t gfp_mask, unsigned int order, const char *fmt, ...)
pr_warn("%s: page allocation failure: order:%u, mode:%#x(%pGg)\n",
current->comm, order, gfp_mask, &gfp_mask);
dump_stack();
+ if (order)
+ filter |= SHOW_COMPACTION_STATS;
if (!should_suppress_show_mem())
show_mem(filter);
}
--
2.8.1
--
Michal Hocko
SUSE Labs
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
