aboutsummaryrefslogtreecommitdiffstats
path: root/mm
diff options
context:
space:
mode:
authorAndrew Morton <akpm@linux-foundation.org>2008-03-24 15:29:52 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2008-03-24 22:22:19 -0400
commit4dd4b920218326231156c7991ce5b94afad841c3 (patch)
tree021d883ec96cd67f5a348eb739a5c6596fcf6d04 /mm
parent537878d2c988fa12bbfaec19ee060c7603a48230 (diff)
revert "kswapd should only wait on IO if there is IO"
Revert commit f1a9ee758de7de1e040de849fdef46e6802ea117: Author: Rik van Riel <riel@redhat.com> Date: Thu Feb 7 00:14:08 2008 -0800 kswapd should only wait on IO if there is IO The current kswapd (and try_to_free_pages) code has an oddity where the code will wait on IO, even if there is no IO in flight. This problem is notable especially when the system scans through many unfreeable pages, causing unnecessary stalls in the VM. Additionally, tasks without __GFP_FS or __GFP_IO in the direct reclaim path will sleep if a significant number of pages are encountered that should be written out. This gives kswapd a chance to write out those pages, while the direct reclaim task sleeps. Signed-off-by: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Because of large latencies and interactivity problems reported by Carlos, here: http://lkml.org/lkml/2008/3/22/211 Cc: Rik van Riel <riel@redhat.com> Cc: "Carlos R. Mafra" <crmafra2@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'mm')
-rw-r--r--mm/vmscan.c27
1 files changed, 5 insertions, 22 deletions
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 45711585684e..4046434046e6 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -70,13 +70,6 @@ struct scan_control {
70 70
71 int order; 71 int order;
72 72
73 /*
74 * Pages that have (or should have) IO pending. If we run into
75 * a lot of these, we're better off waiting a little for IO to
76 * finish rather than scanning more pages in the VM.
77 */
78 int nr_io_pages;
79
80 /* Which cgroup do we reclaim from */ 73 /* Which cgroup do we reclaim from */
81 struct mem_cgroup *mem_cgroup; 74 struct mem_cgroup *mem_cgroup;
82 75
@@ -512,10 +505,8 @@ static unsigned long shrink_page_list(struct list_head *page_list,
512 */ 505 */
513 if (sync_writeback == PAGEOUT_IO_SYNC && may_enter_fs) 506 if (sync_writeback == PAGEOUT_IO_SYNC && may_enter_fs)
514 wait_on_page_writeback(page); 507 wait_on_page_writeback(page);
515 else { 508 else
516 sc->nr_io_pages++;
517 goto keep_locked; 509 goto keep_locked;
518 }
519 } 510 }
520 511
521 referenced = page_referenced(page, 1, sc->mem_cgroup); 512 referenced = page_referenced(page, 1, sc->mem_cgroup);
@@ -554,10 +545,8 @@ static unsigned long shrink_page_list(struct list_head *page_list,
554 if (PageDirty(page)) { 545 if (PageDirty(page)) {
555 if (sc->order <= PAGE_ALLOC_COSTLY_ORDER && referenced) 546 if (sc->order <= PAGE_ALLOC_COSTLY_ORDER && referenced)
556 goto keep_locked; 547 goto keep_locked;
557 if (!may_enter_fs) { 548 if (!may_enter_fs)
558 sc->nr_io_pages++;
559 goto keep_locked; 549 goto keep_locked;
560 }
561 if (!sc->may_writepage) 550 if (!sc->may_writepage)
562 goto keep_locked; 551 goto keep_locked;
563 552
@@ -568,10 +557,8 @@ static unsigned long shrink_page_list(struct list_head *page_list,
568 case PAGE_ACTIVATE: 557 case PAGE_ACTIVATE:
569 goto activate_locked; 558 goto activate_locked;
570 case PAGE_SUCCESS: 559 case PAGE_SUCCESS:
571 if (PageWriteback(page) || PageDirty(page)) { 560 if (PageWriteback(page) || PageDirty(page))
572 sc->nr_io_pages++;
573 goto keep; 561 goto keep;
574 }
575 /* 562 /*
576 * A synchronous write - probably a ramdisk. Go 563 * A synchronous write - probably a ramdisk. Go
577 * ahead and try to reclaim the page. 564 * ahead and try to reclaim the page.
@@ -1344,7 +1331,6 @@ static unsigned long do_try_to_free_pages(struct zone **zones, gfp_t gfp_mask,
1344 1331
1345 for (priority = DEF_PRIORITY; priority >= 0; priority--) { 1332 for (priority = DEF_PRIORITY; priority >= 0; priority--) {
1346 sc->nr_scanned = 0; 1333 sc->nr_scanned = 0;
1347 sc->nr_io_pages = 0;
1348 if (!priority) 1334 if (!priority)
1349 disable_swap_token(); 1335 disable_swap_token();
1350 nr_reclaimed += shrink_zones(priority, zones, sc); 1336 nr_reclaimed += shrink_zones(priority, zones, sc);
@@ -1379,8 +1365,7 @@ static unsigned long do_try_to_free_pages(struct zone **zones, gfp_t gfp_mask,
1379 } 1365 }
1380 1366
1381 /* Take a nap, wait for some writeback to complete */ 1367 /* Take a nap, wait for some writeback to complete */
1382 if (sc->nr_scanned && priority < DEF_PRIORITY - 2 && 1368 if (sc->nr_scanned && priority < DEF_PRIORITY - 2)
1383 sc->nr_io_pages > sc->swap_cluster_max)
1384 congestion_wait(WRITE, HZ/10); 1369 congestion_wait(WRITE, HZ/10);
1385 } 1370 }
1386 /* top priority shrink_caches still had more to do? don't OOM, then */ 1371 /* top priority shrink_caches still had more to do? don't OOM, then */
@@ -1514,7 +1499,6 @@ loop_again:
1514 if (!priority) 1499 if (!priority)
1515 disable_swap_token(); 1500 disable_swap_token();
1516 1501
1517 sc.nr_io_pages = 0;
1518 all_zones_ok = 1; 1502 all_zones_ok = 1;
1519 1503
1520 /* 1504 /*
@@ -1607,8 +1591,7 @@ loop_again:
1607 * OK, kswapd is getting into trouble. Take a nap, then take 1591 * OK, kswapd is getting into trouble. Take a nap, then take
1608 * another pass across the zones. 1592 * another pass across the zones.
1609 */ 1593 */
1610 if (total_scanned && priority < DEF_PRIORITY - 2 && 1594 if (total_scanned && priority < DEF_PRIORITY - 2)
1611 sc.nr_io_pages > sc.swap_cluster_max)
1612 congestion_wait(WRITE, HZ/10); 1595 congestion_wait(WRITE, HZ/10);
1613 1596
1614 /* 1597 /*