From: Xiaoxi Chen <xiaoxi.chen@intel.com>
Date: Wed, 29 Apr 2015 05:59:16 +0000 (+0800)
Subject: os/newstore: fix deadlock when newstore_sync_transaction=true
X-Git-Tag: v9.1.0~242^2~29
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=e3abf245ba4a555b695a437fa488ca41b36c2c77;p=ceph.git

os/newstore: fix deadlock when newstore_sync_transaction=true

There is a deadlock issue in Newstore when newstore_sync_transaction = true.
With sync_transaction to true, the txc state machine will go all the way down
from STATE_IO_DONE to STATE_FINISHING in the same thread, while holding the osr->qlock().
The deadlock is caused in _txc_finish and _osr_reap_done, when trying to
lock osr->qlock again.

Since the _txc_finish can be called with(in sync transaction mode) or without
(in async transaction mode) holding the qlock, so fix this by setting the qlock
to PTHREAD_MUTEX_RECURSIVE, thus we can recursive acquire the qlock.

Signed-off-by: Xiaoxi Chen <xiaoxi.chen@intel.com>
---

diff --git a/src/os/newstore/NewStore.h b/src/os/newstore/NewStore.h
index 95122a1c4662..c126873e970e 100644
--- a/src/os/newstore/NewStore.h
+++ b/src/os/newstore/NewStore.h
@@ -279,7 +279,8 @@ public:
     Mutex wal_apply_lock;
 
     OpSequencer()
-      : qlock("NewStore::OpSequencer::qlock", false, false),
+	//set the qlock to to PTHREAD_MUTEX_RECURSIVE mode
+      : qlock("NewStore::OpSequencer::qlock", true, false),
 	parent(NULL),
 	wal_apply_lock("NewStore::OpSequencer::wal_apply_lock") {
     }