[PATCH 3.2 087/107] ocfs2/dlm: fix deadlock when dispatch assert master
3.2.72-rc1 review patch. If anyone has any objections, please let me know. -- From: Joseph Qi commit 012572d4fc2e4ddd5c8ec8614d51414ec6cae02a upstream. The order of the following three spinlocks should be: dlm_domain_lock < dlm_ctxt->spinlock < dlm_lock_resource->spinlock But dlm_dispatch_assert_master() is called while holding dlm_ctxt->spinlock and dlm_lock_resource->spinlock, and then it calls dlm_grab() which will take dlm_domain_lock. Once another thread (for example, dlm_query_join_handler) has already taken dlm_domain_lock, and tries to take dlm_ctxt->spinlock deadlock happens. Signed-off-by: Joseph Qi Cc: Joel Becker Cc: Mark Fasheh Cc: "Junxiao Bi" Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds [bwh: Backported to 3.2: adjust context] Signed-off-by: Ben Hutchings --- --- a/fs/ocfs2/dlm/dlmmaster.c +++ b/fs/ocfs2/dlm/dlmmaster.c @@ -1411,6 +1411,7 @@ int dlm_master_request_handler(struct o2 int found, ret; int set_maybe; int dispatch_assert = 0; + int dispatched = 0; if (!dlm_grab(dlm)) return DLM_MASTER_RESP_NO; @@ -1617,13 +1618,16 @@ send_response: mlog(ML_ERROR, "failed to dispatch assert master work\n"); response = DLM_MASTER_RESP_ERROR; dlm_lockres_put(res); + } else { + dispatched = 1; } } else { if (res) dlm_lockres_put(res); } - dlm_put(dlm); + if (!dispatched) + dlm_put(dlm); return response; } @@ -2041,7 +2045,6 @@ int dlm_dispatch_assert_master(struct dl /* queue up work for dlm_assert_master_worker */ - dlm_grab(dlm); /* get an extra ref for the work item */ dlm_init_work_item(dlm, item, dlm_assert_master_worker, NULL); item->u.am.lockres = res; /* already have a ref */ /* can optionally ignore node numbers higher than this node */ --- a/fs/ocfs2/dlm/dlmrecovery.c +++ b/fs/ocfs2/dlm/dlmrecovery.c @@ -1689,6 +1689,7 @@ int dlm_master_requery_handler(struct o2 unsigned int hash; int master = DLM_LOCK_RES_OWNER_UNKNOWN; u32 flags = DLM_ASSERT_MASTER_REQUERY; + int dispatched = 0; if (!dlm_grab(dlm)) { /* since the domain has gone away on this @@ -1710,6 +1711,8 @@ int dlm_master_requery_handler(struct o2 mlog_errno(-ENOMEM); /* retry!? */ BUG(); + } else { + dispatched = 1; } } else /* put.. incase we are not the master */ dlm_lockres_put(res); @@ -1717,7 +1720,8 @@ int dlm_master_requery_handler(struct o2 } spin_unlock(>spinlock); - dlm_put(dlm); + if (!dispatched) + dlm_put(dlm); return master; } -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[PATCH 3.2 087/107] ocfs2/dlm: fix deadlock when dispatch assert master
3.2.72-rc1 review patch. If anyone has any objections, please let me know. -- From: Joseph Qicommit 012572d4fc2e4ddd5c8ec8614d51414ec6cae02a upstream. The order of the following three spinlocks should be: dlm_domain_lock < dlm_ctxt->spinlock < dlm_lock_resource->spinlock But dlm_dispatch_assert_master() is called while holding dlm_ctxt->spinlock and dlm_lock_resource->spinlock, and then it calls dlm_grab() which will take dlm_domain_lock. Once another thread (for example, dlm_query_join_handler) has already taken dlm_domain_lock, and tries to take dlm_ctxt->spinlock deadlock happens. Signed-off-by: Joseph Qi Cc: Joel Becker Cc: Mark Fasheh Cc: "Junxiao Bi" Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds [bwh: Backported to 3.2: adjust context] Signed-off-by: Ben Hutchings --- --- a/fs/ocfs2/dlm/dlmmaster.c +++ b/fs/ocfs2/dlm/dlmmaster.c @@ -1411,6 +1411,7 @@ int dlm_master_request_handler(struct o2 int found, ret; int set_maybe; int dispatch_assert = 0; + int dispatched = 0; if (!dlm_grab(dlm)) return DLM_MASTER_RESP_NO; @@ -1617,13 +1618,16 @@ send_response: mlog(ML_ERROR, "failed to dispatch assert master work\n"); response = DLM_MASTER_RESP_ERROR; dlm_lockres_put(res); + } else { + dispatched = 1; } } else { if (res) dlm_lockres_put(res); } - dlm_put(dlm); + if (!dispatched) + dlm_put(dlm); return response; } @@ -2041,7 +2045,6 @@ int dlm_dispatch_assert_master(struct dl /* queue up work for dlm_assert_master_worker */ - dlm_grab(dlm); /* get an extra ref for the work item */ dlm_init_work_item(dlm, item, dlm_assert_master_worker, NULL); item->u.am.lockres = res; /* already have a ref */ /* can optionally ignore node numbers higher than this node */ --- a/fs/ocfs2/dlm/dlmrecovery.c +++ b/fs/ocfs2/dlm/dlmrecovery.c @@ -1689,6 +1689,7 @@ int dlm_master_requery_handler(struct o2 unsigned int hash; int master = DLM_LOCK_RES_OWNER_UNKNOWN; u32 flags = DLM_ASSERT_MASTER_REQUERY; + int dispatched = 0; if (!dlm_grab(dlm)) { /* since the domain has gone away on this @@ -1710,6 +1711,8 @@ int dlm_master_requery_handler(struct o2 mlog_errno(-ENOMEM); /* retry!? */ BUG(); + } else { + dispatched = 1; } } else /* put.. incase we are not the master */ dlm_lockres_put(res); @@ -1717,7 +1720,8 @@ int dlm_master_requery_handler(struct o2 } spin_unlock(>spinlock); - dlm_put(dlm); + if (!dispatched) + dlm_put(dlm); return master; } -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/