Re: [PATCH 3.4 61/92] ocfs2/dlm: fix deadlock when dispatch assert master

[Joseph Qi]

Hi Zefan,
In dlm_master_request_handler, it has missed the following change :
-	dlm_put(dlm);
+	if (!dispatched)
+		dlm_put(dlm);

Thanks,
Joseph

On 2016/4/18 18:46, lizf@xxxxxxxxxx wrote:
> From: Joseph Qi <joseph.qi@xxxxxxxxxx>
> 
> 3.4.112-rc1 review patch.  If anyone has any objections, please let me know.
> 
> ------------------
> 
> 
> 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 <joseph.qi@xxxxxxxxxx>
> Cc: Joel Becker <jlbec@xxxxxxxxxxxx>
> Cc: Mark Fasheh <mfasheh@xxxxxxxx>
> Cc: "Junxiao Bi" <junxiao.bi@xxxxxxxxxx>
> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> Signed-off-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
> [lizf: Backported to 3.4: adjust context]
> Signed-off-by: Zefan Li <lizefan@xxxxxxxxxx>
> ---
>  fs/ocfs2/dlm/dlmmaster.c   | 4 +++-
>  fs/ocfs2/dlm/dlmrecovery.c | 6 +++++-
>  2 files changed, 8 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/ocfs2/dlm/dlmmaster.c b/fs/ocfs2/dlm/dlmmaster.c
> index 7ba6ac1..751efa8 100644
> --- a/fs/ocfs2/dlm/dlmmaster.c
> +++ b/fs/ocfs2/dlm/dlmmaster.c
> @@ -1411,6 +1411,7 @@ int dlm_master_request_handler(struct o2net_msg *msg, u32 len, void *data,
>  	int found, ret;
>  	int set_maybe;
>  	int dispatch_assert = 0;
> +	int dispatched = 0;
>  
>  	if (!dlm_grab(dlm))
>  		return DLM_MASTER_RESP_NO;
> @@ -1617,6 +1618,8 @@ 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)
> @@ -2041,7 +2044,6 @@ int dlm_dispatch_assert_master(struct dlm_ctxt *dlm,
>  
>  
>  	/* 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 */
> diff --git a/fs/ocfs2/dlm/dlmrecovery.c b/fs/ocfs2/dlm/dlmrecovery.c
> index d15b071..0e5013e 100644
> --- a/fs/ocfs2/dlm/dlmrecovery.c
> +++ b/fs/ocfs2/dlm/dlmrecovery.c
> @@ -1689,6 +1689,7 @@ int dlm_master_requery_handler(struct o2net_msg *msg, u32 len, void *data,
>  	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 o2net_msg *msg, u32 len, void *data,
>  				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 o2net_msg *msg, u32 len, void *data,
>  	}
>  	spin_unlock(&dlm->spinlock);
>  
> -	dlm_put(dlm);
> +	if (!dispatched)
> +		dlm_put(dlm);
>  	return master;
>  }
>  
>