aboutsummaryrefslogtreecommitdiffstats
path: root/mm/swap.c
diff options
context:
space:
mode:
authorKOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>2009-04-30 18:08:51 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2009-05-02 18:36:10 -0400
commit00a62ce91e554198ef28234c91c36f850f5a3bc9 (patch)
tree367ef134219deef91903c3fa0eb108c13658f2c7 /mm/swap.c
parent0763ed2355198cdef2f6a2098e9d52eb1fe4365d (diff)
mm: fix Committed_AS underflow on large NR_CPUS environment
The Committed_AS field can underflow in certain situations: > # while true; do cat /proc/meminfo | grep _AS; sleep 1; done | uniq -c > 1 Committed_AS: 18446744073709323392 kB > 11 Committed_AS: 18446744073709455488 kB > 6 Committed_AS: 35136 kB > 5 Committed_AS: 18446744073709454400 kB > 7 Committed_AS: 35904 kB > 3 Committed_AS: 18446744073709453248 kB > 2 Committed_AS: 34752 kB > 9 Committed_AS: 18446744073709453248 kB > 8 Committed_AS: 34752 kB > 3 Committed_AS: 18446744073709320960 kB > 7 Committed_AS: 18446744073709454080 kB > 3 Committed_AS: 18446744073709320960 kB > 5 Committed_AS: 18446744073709454080 kB > 6 Committed_AS: 18446744073709320960 kB Because NR_CPUS can be greater than 1000 and meminfo_proc_show() does not check for underflow. But NR_CPUS proportional isn't good calculation. In general, possibility of lock contention is proportional to the number of online cpus, not theorical maximum cpus (NR_CPUS). The current kernel has generic percpu-counter stuff. using it is right way. it makes code simplify and percpu_counter_read_positive() don't make underflow issue. Reported-by: Dave Hansen <dave@linux.vnet.ibm.com> Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Eric B Munson <ebmunson@us.ibm.com> Cc: Mel Gorman <mel@csn.ul.ie> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: <stable@kernel.org> [All kernel versions] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'mm/swap.c')
-rw-r--r--mm/swap.c46
1 files changed, 0 insertions, 46 deletions
diff --git a/mm/swap.c b/mm/swap.c
index bede23ce64ea..cb29ae5d33ab 100644
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -491,49 +491,6 @@ unsigned pagevec_lookup_tag(struct pagevec *pvec, struct address_space *mapping,
491 491
492EXPORT_SYMBOL(pagevec_lookup_tag); 492EXPORT_SYMBOL(pagevec_lookup_tag);
493 493
494#ifdef CONFIG_SMP
495/*
496 * We tolerate a little inaccuracy to avoid ping-ponging the counter between
497 * CPUs
498 */
499#define ACCT_THRESHOLD max(16, NR_CPUS * 2)
500
501static DEFINE_PER_CPU(long, committed_space);
502
503void vm_acct_memory(long pages)
504{
505 long *local;
506
507 preempt_disable();
508 local = &__get_cpu_var(committed_space);
509 *local += pages;
510 if (*local > ACCT_THRESHOLD || *local < -ACCT_THRESHOLD) {
511 atomic_long_add(*local, &vm_committed_space);
512 *local = 0;
513 }
514 preempt_enable();
515}
516
517#ifdef CONFIG_HOTPLUG_CPU
518
519/* Drop the CPU's cached committed space back into the central pool. */
520static int cpu_swap_callback(struct notifier_block *nfb,
521 unsigned long action,
522 void *hcpu)
523{
524 long *committed;
525
526 committed = &per_cpu(committed_space, (long)hcpu);
527 if (action == CPU_DEAD || action == CPU_DEAD_FROZEN) {
528 atomic_long_add(*committed, &vm_committed_space);
529 *committed = 0;
530 drain_cpu_pagevecs((long)hcpu);
531 }
532 return NOTIFY_OK;
533}
534#endif /* CONFIG_HOTPLUG_CPU */
535#endif /* CONFIG_SMP */
536
537/* 494/*
538 * Perform any setup for the swap system 495 * Perform any setup for the swap system
539 */ 496 */
@@ -554,7 +511,4 @@ void __init swap_setup(void)
554 * Right now other parts of the system means that we 511 * Right now other parts of the system means that we
555 * _really_ don't want to cluster much more 512 * _really_ don't want to cluster much more
556 */ 513 */
557#ifdef CONFIG_HOTPLUG_CPU
558 hotcpu_notifier(cpu_swap_callback, 0);
559#endif
560} 514}