diff options
author | Jens Axboe <jens.axboe@oracle.com> | 2009-07-09 08:52:32 -0400 |
---|---|---|
committer | Jens Axboe <axboe@carl.(none)> | 2009-07-10 14:31:53 -0400 |
commit | 8aa7e847d834ed937a9ad37a0f2ad5b8584c1ab0 (patch) | |
tree | 76c8b4f1362a928d426f2201790ab5d128f57724 /mm | |
parent | c2cc49a2f8a479dde96a599646d30b6cc9dbed78 (diff) |
Fix congestion_wait() sync/async vs read/write confusion
Commit 1faa16d22877f4839bd433547d770c676d1d964c accidentally broke
the bdi congestion wait queue logic, causing us to wait on congestion
for WRITE (== 1) when we really wanted BLK_RW_ASYNC (== 0) instead.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Diffstat (limited to 'mm')
-rw-r--r-- | mm/backing-dev.c | 7 | ||||
-rw-r--r-- | mm/memcontrol.c | 2 | ||||
-rw-r--r-- | mm/page-writeback.c | 8 | ||||
-rw-r--r-- | mm/page_alloc.c | 4 | ||||
-rw-r--r-- | mm/vmscan.c | 8 |
5 files changed, 14 insertions, 15 deletions
diff --git a/mm/backing-dev.c b/mm/backing-dev.c index 493b468a5035..c86edd244294 100644 --- a/mm/backing-dev.c +++ b/mm/backing-dev.c | |||
@@ -283,7 +283,6 @@ static wait_queue_head_t congestion_wqh[2] = { | |||
283 | __WAIT_QUEUE_HEAD_INITIALIZER(congestion_wqh[1]) | 283 | __WAIT_QUEUE_HEAD_INITIALIZER(congestion_wqh[1]) |
284 | }; | 284 | }; |
285 | 285 | ||
286 | |||
287 | void clear_bdi_congested(struct backing_dev_info *bdi, int sync) | 286 | void clear_bdi_congested(struct backing_dev_info *bdi, int sync) |
288 | { | 287 | { |
289 | enum bdi_state bit; | 288 | enum bdi_state bit; |
@@ -308,18 +307,18 @@ EXPORT_SYMBOL(set_bdi_congested); | |||
308 | 307 | ||
309 | /** | 308 | /** |
310 | * congestion_wait - wait for a backing_dev to become uncongested | 309 | * congestion_wait - wait for a backing_dev to become uncongested |
311 | * @rw: READ or WRITE | 310 | * @sync: SYNC or ASYNC IO |
312 | * @timeout: timeout in jiffies | 311 | * @timeout: timeout in jiffies |
313 | * | 312 | * |
314 | * Waits for up to @timeout jiffies for a backing_dev (any backing_dev) to exit | 313 | * Waits for up to @timeout jiffies for a backing_dev (any backing_dev) to exit |
315 | * write congestion. If no backing_devs are congested then just wait for the | 314 | * write congestion. If no backing_devs are congested then just wait for the |
316 | * next write to be completed. | 315 | * next write to be completed. |
317 | */ | 316 | */ |
318 | long congestion_wait(int rw, long timeout) | 317 | long congestion_wait(int sync, long timeout) |
319 | { | 318 | { |
320 | long ret; | 319 | long ret; |
321 | DEFINE_WAIT(wait); | 320 | DEFINE_WAIT(wait); |
322 | wait_queue_head_t *wqh = &congestion_wqh[rw]; | 321 | wait_queue_head_t *wqh = &congestion_wqh[sync]; |
323 | 322 | ||
324 | prepare_to_wait(wqh, &wait, TASK_UNINTERRUPTIBLE); | 323 | prepare_to_wait(wqh, &wait, TASK_UNINTERRUPTIBLE); |
325 | ret = io_schedule_timeout(timeout); | 324 | ret = io_schedule_timeout(timeout); |
diff --git a/mm/memcontrol.c b/mm/memcontrol.c index e2fa20dadf40..e717964cb5a0 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c | |||
@@ -1973,7 +1973,7 @@ try_to_free: | |||
1973 | if (!progress) { | 1973 | if (!progress) { |
1974 | nr_retries--; | 1974 | nr_retries--; |
1975 | /* maybe some writeback is necessary */ | 1975 | /* maybe some writeback is necessary */ |
1976 | congestion_wait(WRITE, HZ/10); | 1976 | congestion_wait(BLK_RW_ASYNC, HZ/10); |
1977 | } | 1977 | } |
1978 | 1978 | ||
1979 | } | 1979 | } |
diff --git a/mm/page-writeback.c b/mm/page-writeback.c index 7687879253b9..81627ebcd313 100644 --- a/mm/page-writeback.c +++ b/mm/page-writeback.c | |||
@@ -575,7 +575,7 @@ static void balance_dirty_pages(struct address_space *mapping) | |||
575 | if (pages_written >= write_chunk) | 575 | if (pages_written >= write_chunk) |
576 | break; /* We've done our duty */ | 576 | break; /* We've done our duty */ |
577 | 577 | ||
578 | congestion_wait(WRITE, HZ/10); | 578 | congestion_wait(BLK_RW_ASYNC, HZ/10); |
579 | } | 579 | } |
580 | 580 | ||
581 | if (bdi_nr_reclaimable + bdi_nr_writeback < bdi_thresh && | 581 | if (bdi_nr_reclaimable + bdi_nr_writeback < bdi_thresh && |
@@ -669,7 +669,7 @@ void throttle_vm_writeout(gfp_t gfp_mask) | |||
669 | if (global_page_state(NR_UNSTABLE_NFS) + | 669 | if (global_page_state(NR_UNSTABLE_NFS) + |
670 | global_page_state(NR_WRITEBACK) <= dirty_thresh) | 670 | global_page_state(NR_WRITEBACK) <= dirty_thresh) |
671 | break; | 671 | break; |
672 | congestion_wait(WRITE, HZ/10); | 672 | congestion_wait(BLK_RW_ASYNC, HZ/10); |
673 | 673 | ||
674 | /* | 674 | /* |
675 | * The caller might hold locks which can prevent IO completion | 675 | * The caller might hold locks which can prevent IO completion |
@@ -715,7 +715,7 @@ static void background_writeout(unsigned long _min_pages) | |||
715 | if (wbc.nr_to_write > 0 || wbc.pages_skipped > 0) { | 715 | if (wbc.nr_to_write > 0 || wbc.pages_skipped > 0) { |
716 | /* Wrote less than expected */ | 716 | /* Wrote less than expected */ |
717 | if (wbc.encountered_congestion || wbc.more_io) | 717 | if (wbc.encountered_congestion || wbc.more_io) |
718 | congestion_wait(WRITE, HZ/10); | 718 | congestion_wait(BLK_RW_ASYNC, HZ/10); |
719 | else | 719 | else |
720 | break; | 720 | break; |
721 | } | 721 | } |
@@ -787,7 +787,7 @@ static void wb_kupdate(unsigned long arg) | |||
787 | writeback_inodes(&wbc); | 787 | writeback_inodes(&wbc); |
788 | if (wbc.nr_to_write > 0) { | 788 | if (wbc.nr_to_write > 0) { |
789 | if (wbc.encountered_congestion || wbc.more_io) | 789 | if (wbc.encountered_congestion || wbc.more_io) |
790 | congestion_wait(WRITE, HZ/10); | 790 | congestion_wait(BLK_RW_ASYNC, HZ/10); |
791 | else | 791 | else |
792 | break; /* All the old data is written */ | 792 | break; /* All the old data is written */ |
793 | } | 793 | } |
diff --git a/mm/page_alloc.c b/mm/page_alloc.c index ad7cd1c56b07..a35eeab2724c 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c | |||
@@ -1666,7 +1666,7 @@ __alloc_pages_high_priority(gfp_t gfp_mask, unsigned int order, | |||
1666 | preferred_zone, migratetype); | 1666 | preferred_zone, migratetype); |
1667 | 1667 | ||
1668 | if (!page && gfp_mask & __GFP_NOFAIL) | 1668 | if (!page && gfp_mask & __GFP_NOFAIL) |
1669 | congestion_wait(WRITE, HZ/50); | 1669 | congestion_wait(BLK_RW_ASYNC, HZ/50); |
1670 | } while (!page && (gfp_mask & __GFP_NOFAIL)); | 1670 | } while (!page && (gfp_mask & __GFP_NOFAIL)); |
1671 | 1671 | ||
1672 | return page; | 1672 | return page; |
@@ -1831,7 +1831,7 @@ rebalance: | |||
1831 | pages_reclaimed += did_some_progress; | 1831 | pages_reclaimed += did_some_progress; |
1832 | if (should_alloc_retry(gfp_mask, order, pages_reclaimed)) { | 1832 | if (should_alloc_retry(gfp_mask, order, pages_reclaimed)) { |
1833 | /* Wait for some write requests to complete then retry */ | 1833 | /* Wait for some write requests to complete then retry */ |
1834 | congestion_wait(WRITE, HZ/50); | 1834 | congestion_wait(BLK_RW_ASYNC, HZ/50); |
1835 | goto rebalance; | 1835 | goto rebalance; |
1836 | } | 1836 | } |
1837 | 1837 | ||
diff --git a/mm/vmscan.c b/mm/vmscan.c index 54155268dfca..dea7abd31098 100644 --- a/mm/vmscan.c +++ b/mm/vmscan.c | |||
@@ -1104,7 +1104,7 @@ static unsigned long shrink_inactive_list(unsigned long max_scan, | |||
1104 | */ | 1104 | */ |
1105 | if (nr_freed < nr_taken && !current_is_kswapd() && | 1105 | if (nr_freed < nr_taken && !current_is_kswapd() && |
1106 | lumpy_reclaim) { | 1106 | lumpy_reclaim) { |
1107 | congestion_wait(WRITE, HZ/10); | 1107 | congestion_wait(BLK_RW_ASYNC, HZ/10); |
1108 | 1108 | ||
1109 | /* | 1109 | /* |
1110 | * The attempt at page out may have made some | 1110 | * The attempt at page out may have made some |
@@ -1721,7 +1721,7 @@ static unsigned long do_try_to_free_pages(struct zonelist *zonelist, | |||
1721 | 1721 | ||
1722 | /* Take a nap, wait for some writeback to complete */ | 1722 | /* Take a nap, wait for some writeback to complete */ |
1723 | if (sc->nr_scanned && priority < DEF_PRIORITY - 2) | 1723 | if (sc->nr_scanned && priority < DEF_PRIORITY - 2) |
1724 | congestion_wait(WRITE, HZ/10); | 1724 | congestion_wait(BLK_RW_ASYNC, HZ/10); |
1725 | } | 1725 | } |
1726 | /* top priority shrink_zones still had more to do? don't OOM, then */ | 1726 | /* top priority shrink_zones still had more to do? don't OOM, then */ |
1727 | if (!sc->all_unreclaimable && scanning_global_lru(sc)) | 1727 | if (!sc->all_unreclaimable && scanning_global_lru(sc)) |
@@ -1960,7 +1960,7 @@ loop_again: | |||
1960 | * another pass across the zones. | 1960 | * another pass across the zones. |
1961 | */ | 1961 | */ |
1962 | if (total_scanned && priority < DEF_PRIORITY - 2) | 1962 | if (total_scanned && priority < DEF_PRIORITY - 2) |
1963 | congestion_wait(WRITE, HZ/10); | 1963 | congestion_wait(BLK_RW_ASYNC, HZ/10); |
1964 | 1964 | ||
1965 | /* | 1965 | /* |
1966 | * We do this so kswapd doesn't build up large priorities for | 1966 | * We do this so kswapd doesn't build up large priorities for |
@@ -2233,7 +2233,7 @@ unsigned long shrink_all_memory(unsigned long nr_pages) | |||
2233 | goto out; | 2233 | goto out; |
2234 | 2234 | ||
2235 | if (sc.nr_scanned && prio < DEF_PRIORITY - 2) | 2235 | if (sc.nr_scanned && prio < DEF_PRIORITY - 2) |
2236 | congestion_wait(WRITE, HZ / 10); | 2236 | congestion_wait(BLK_RW_ASYNC, HZ / 10); |
2237 | } | 2237 | } |
2238 | } | 2238 | } |
2239 | 2239 | ||