Repository: hadoop Updated Branches: refs/heads/branch-2 15760a148 -> 796b94df1 refs/heads/branch-2.6 3c9c2b404 -> 597521fcf refs/heads/branch-2.7 90f364172 -> d8d33055b refs/heads/trunk 08bd4edf4 -> 05ed69058
HADOOP-11333. Fix deadlock in DomainSocketWatcher when the notification pipe is full (zhaoyunjiong via cmccabe) (cherry picked from commit 86e3993def01223f92b8d1dd35f6c1f8ab6033f5) (cherry picked from commit f6d1bf5ed1cf647d82e676df15587de42b1faa42) Project: http://git-wip-us.apache.org/repos/asf/hadoop/repo Commit: http://git-wip-us.apache.org/repos/asf/hadoop/commit/597521fc Tree: http://git-wip-us.apache.org/repos/asf/hadoop/tree/597521fc Diff: http://git-wip-us.apache.org/repos/asf/hadoop/diff/597521fc Branch: refs/heads/branch-2.6 Commit: 597521fcf8f678c27b7b5c2b11bb855695d60413 Parents: 3c9c2b4 Author: Colin Patrick Mccabe <cmcc...@cloudera.com> Authored: Mon Dec 1 11:42:10 2014 -0800 Committer: Vinayakumar B <vinayakum...@apache.org> Committed: Fri Aug 14 12:49:52 2015 +0530 ---------------------------------------------------------------------- hadoop-common-project/hadoop-common/CHANGES.txt | 3 +++ .../apache/hadoop/net/unix/DomainSocketWatcher.java | 15 +++++++++++++++ 2 files changed, 18 insertions(+) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/hadoop/blob/597521fc/hadoop-common-project/hadoop-common/CHANGES.txt ---------------------------------------------------------------------- diff --git a/hadoop-common-project/hadoop-common/CHANGES.txt b/hadoop-common-project/hadoop-common/CHANGES.txt index 19bc188..c790af5 100644 --- a/hadoop-common-project/hadoop-common/CHANGES.txt +++ b/hadoop-common-project/hadoop-common/CHANGES.txt @@ -18,6 +18,9 @@ Release 2.6.1 - UNRELEASED HADOOP-10786. Fix UGI#reloginFromKeytab on Java 8. (Stephen Chu via wheat9) + HADOOP-11333. Fix deadlock in DomainSocketWatcher when the notification + pipe is full (zhaoyunjiong via cmccabe) + Release 2.6.0 - 2014-11-18 INCOMPATIBLE CHANGES http://git-wip-us.apache.org/repos/asf/hadoop/blob/597521fc/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/net/unix/DomainSocketWatcher.java ---------------------------------------------------------------------- diff --git a/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/net/unix/DomainSocketWatcher.java b/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/net/unix/DomainSocketWatcher.java index 95ef30d..0172f6b 100644 --- a/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/net/unix/DomainSocketWatcher.java +++ b/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/net/unix/DomainSocketWatcher.java @@ -103,6 +103,7 @@ public final class DomainSocketWatcher implements Closeable { public boolean handle(DomainSocket sock) { assert(lock.isHeldByCurrentThread()); try { + kicked = false; if (LOG.isTraceEnabled()) { LOG.trace(this + ": NotificationHandler: doing a read on " + sock.fd); @@ -228,6 +229,14 @@ public final class DomainSocketWatcher implements Closeable { * Whether or not this DomainSocketWatcher is closed. */ private boolean closed = false; + + /** + * True if we have written a byte to the notification socket. We should not + * write anything else to the socket until the notification handler has had a + * chance to run. Otherwise, our thread might block, causing deadlock. + * See HADOOP-11333 for details. + */ + private boolean kicked = false; public DomainSocketWatcher(int interruptCheckPeriodMs) throws IOException { if (loadingFailureReason != null) { @@ -348,8 +357,14 @@ public final class DomainSocketWatcher implements Closeable { */ private void kick() { assert(lock.isHeldByCurrentThread()); + + if (kicked) { + return; + } + try { notificationSockets[0].getOutputStream().write(0); + kicked = true; } catch (IOException e) { if (!closed) { LOG.error(this + ": error writing to notificationSockets[0]", e);