History

mm: /proc/sys/vm/stat_refresh to force vmstat update

Provide /proc/sys/vm/stat_refresh to force an immediate update of per-cpu into global vmstats: useful to avoid a sleep(2) or whatever before checking counts when testing. Originally added to work around a bug which left counts stranded indefinitely on a cpu going idle (an inaccuracy magnified when small below-batch numbers represent "huge" amounts of memory), but I believe that bug is now fixed: nonetheless, this is still a useful knob. Its schedule_on_each_cpu() is probably too expensive just to fold into reading /proc/meminfo itself: give this mode 0600 to prevent abuse. Allow a write or a read to do the same: nothing to read, but "grep -h Shmem /proc/sys/vm/stat_refresh /proc/meminfo" is convenient. Oh, and since global_page_state() itself is careful to disguise any underflow as 0, hack in an "Invalid argument" and pr_warn() if a counter is negative after the refresh - this helped to fix a misaccounting of NR_ISOLATED_FILE in my migration code. But on recent kernels, I find that NR_ALLOC_BATCH and NR_PAGES_SCANNED often go negative some of the time. I have not yet worked out why, but have no evidence that it's actually harmful. Punt for the moment by just ignoring the anomaly on those.

Hugh Dickins committed 10 years ago

Commit 52b6f461 parent 9e18eb24 files · +85 −0Browse files

Unified · Split

4 files
  1. Documentation/sysctl/vm.txt +14 −0
  2. include/linux/vmstat.h +4 −0
  3. kernel/sysctl.c +7 −0
  4. mm/vmstat.c +60 −0
Documentation/sysctl/vm.txt+14−0
@@ -57,6 +57,7 @@ Currently, these files are in /proc/sys/vm: · 56 unmodified lines
5757 - panic_on_oom
5858 - percpu_pagelist_fraction
5959 - stat_interval
60+- stat_refresh
6061 - swappiness
6162 - user_reserve_kbytes
6263 - vfs_cache_pressure
@@ -755,6 +756,19 @@ is 1 second. · 692 unmodified lines
755756
756757 ==============================================================
757758
759+stat_refresh
760+
761+Any read or write (by root only) flushes all the per-cpu vm statistics
762+into their global totals, for more accurate reports when testing
763+e.g. cat /proc/sys/vm/stat_refresh /proc/meminfo
764+
765+As a side-effect, it also checks for negative totals (elsewhere reported
766+as 0) and "fails" with EINVAL if any are found, with a warning in dmesg.
767+(At time of writing, a few stats are known sometimes to be found negative,
768+with no ill effects: errors and warnings on these stats are suppressed.)
769+
770+==============================================================
771+
758772 swappiness
759773
760774 This control is used to define how aggressive the kernel will swap
include/linux/vmstat.h+4−0
@@ -193,6 +193,10 @@ void quiet_vmstat(void); · 192 unmodified lines
193193 void cpu_vm_stats_fold(int cpu);
194194 void refresh_zone_stat_thresholds(void);
195195
196+struct ctl_table;
197+int vmstat_refresh(struct ctl_table *, int write,
198+ void __user *buffer, size_t *lenp, loff_t *ppos);
199+
196200 void drain_zonestat(struct zone *zone, struct per_cpu_pageset *);
197201
198202 int calculate_pressure_threshold(struct zone *zone);
kernel/sysctl.c+7−0
@@ -1521,6 +1521,13 @@ static struct ctl_table vm_table[] = { · 1520 unmodified lines
15211521 .mode = 0644,
15221522 .proc_handler = proc_dointvec_jiffies,
15231523 },
1524+ {
1525+ .procname = "stat_refresh",
1526+ .data = NULL,
1527+ .maxlen = 0,
1528+ .mode = 0600,
1529+ .proc_handler = vmstat_refresh,
1530+ },
15241531 #endif
15251532 #ifdef CONFIG_MMU
15261533 {
mm/vmstat.c+60−0
@@ -1379,6 +1379,66 @@ static DEFINE_PER_CPU(struct delayed_work, vmstat_work); · 1378 unmodified lines
13791379 int sysctl_stat_interval __read_mostly = HZ;
13801380 static cpumask_var_t cpu_stat_off;
13811381
1382+#ifdef CONFIG_PROC_FS
1383+static void refresh_vm_stats(struct work_struct *work)
1384+{
1385+ refresh_cpu_vm_stats(true);
1386+}
1387+
1388+int vmstat_refresh(struct ctl_table *table, int write,
1389+ void __user *buffer, size_t *lenp, loff_t *ppos)
1390+{
1391+ long val;
1392+ int err;
1393+ int i;
1394+
1395+ /*
1396+ * The regular update, every sysctl_stat_interval, may come later
1397+ * than expected: leaving a significant amount in per_cpu buckets.
1398+ * This is particularly misleading when checking a quantity of HUGE
1399+ * pages, immediately after running a test. /proc/sys/vm/stat_refresh,
1400+ * which can equally be echo'ed to or cat'ted from (by root),
1401+ * can be used to update the stats just before reading them.
1402+ *
1403+ * Oh, and since global_page_state() etc. are so careful to hide
1404+ * transiently negative values, report an error here if any of
1405+ * the stats is negative, so we know to go looking for imbalance.
1406+ */
1407+ err = schedule_on_each_cpu(refresh_vm_stats);
1408+ if (err)
1409+ return err;
1410+ for (i = 0; i < NR_VM_ZONE_STAT_ITEMS; i++) {
1411+ val = atomic_long_read(&vm_stat[i]);
1412+ if (val < 0) {
1413+ switch (i) {
1414+ case NR_ALLOC_BATCH:
1415+ case NR_PAGES_SCANNED:
1416+ /*
1417+ * These are often seen to go negative in
1418+ * recent kernels, but not to go permanently
1419+ * negative. Whilst it would be nicer not to
1420+ * have exceptions, rooting them out would be
1421+ * another task, of rather low priority.
1422+ */
1423+ break;
1424+ default:
1425+ pr_warn("%s: %s %ld\n",
1426+ __func__, vmstat_text[i], val);
1427+ err = -EINVAL;
1428+ break;
1429+ }
1430+ }
1431+ }
1432+ if (err)
1433+ return err;
1434+ if (write)
1435+ *ppos += *lenp;
1436+ else
1437+ *lenp = 0;
1438+ return 0;
1439+}
1440+#endif /* CONFIG_PROC_FS */
1441+
13821442 static void vmstat_update(struct work_struct *w)
13831443 {
13841444 if (refresh_cpu_vm_stats(true)) {

Keyboard shortcuts

General

Show keyboard shortcuts
?
Focus the current search
/

Issues and pull requests

Open filters
f

Issues, pull requests and Inbox

Select next item
j↓
Select previous item
k↑
Open selected item
Enter
Dismiss filters or clear selection
Esc

Repository code

Go to file
t