NFS: add nfs_sb_deactive_async to avoid deadlock

Use nfs_sb_deactive_async instead of nfs_sb_deactive when in a workqueue
context.  This avoids a deadlock where rpc_shutdown_client loops forever
in a workqueue kworker context, trying to kill all RPC tasks associated with
the client, while one or more of these tasks have already been assigned to the
same kworker (and will never run rpc_exit_task).

This approach is needed because RPC tasks that have already been assigned
to a kworker by queue_work cannot be canceled, as explained in the comment
for workqueue.c:insert_wq_barrier.

Signed-off-by: Weston Andros Adamson <dros@netapp.com>
[Trond: add module_get/put.]
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

5 files changed, 56 insertions, 3 deletions

diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index 5c7325c5c5e..6fa01aea248 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -685,7 +685,10 @@ static void __put_nfs_open_context(struct nfs_open_context *ctx, int is_sync)
 	if (ctx->cred != NULL)
 		put_rpccred(ctx->cred);
 	dput(ctx->dentry);
-	nfs_sb_deactive(sb);
+	if (is_sync)
+		nfs_sb_deactive(sb);
+	else
+		nfs_sb_deactive_async(sb);
 	kfree(ctx->mdsthreshold);
 	kfree(ctx);
 }
diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index a54fe51c1df..05521cadac2 100644
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -351,6 +351,7 @@ extern int __init register_nfs_fs(void);
 extern void __exit unregister_nfs_fs(void);
 extern void nfs_sb_active(struct super_block *sb);
 extern void nfs_sb_deactive(struct super_block *sb);
+extern void nfs_sb_deactive_async(struct super_block *sb);
 
 /* namespace.c */
 #define NFS_PATH_CANONICAL 1
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 1465364501b..8cfbac1a8d5 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -2197,7 +2197,7 @@ static void nfs4_free_closedata(void *data)
 	nfs4_put_open_state(calldata->state);
 	nfs_free_seqid(calldata->arg.seqid);
 	nfs4_put_state_owner(sp);
-	nfs_sb_deactive(sb);
+	nfs_sb_deactive_async(sb);
 	kfree(calldata);
 }
 
diff --git a/fs/nfs/super.c b/fs/nfs/super.c
index 13c2a5be476..652d3f7176a 100644
--- a/fs/nfs/super.c
+++ b/fs/nfs/super.c
@@ -54,6 +54,7 @@
 #include <linux/parser.h>
 #include <linux/nsproxy.h>
 #include <linux/rcupdate.h>
+#include <linux/kthread.h>
 
 #include <asm/uaccess.h>
 
@@ -415,6 +416,54 @@ void nfs_sb_deactive(struct super_block *sb)
 }
 EXPORT_SYMBOL_GPL(nfs_sb_deactive);
 
+static int nfs_deactivate_super_async_work(void *ptr)
+{
+	struct super_block *sb = ptr;
+
+	deactivate_super(sb);
+	module_put_and_exit(0);
+	return 0;
+}
+
+/*
+ * same effect as deactivate_super, but will do final unmount in kthread
+ * context
+ */
+static void nfs_deactivate_super_async(struct super_block *sb)
+{
+	struct task_struct *task;
+	char buf[INET6_ADDRSTRLEN + 1];
+	struct nfs_server *server = NFS_SB(sb);
+	struct nfs_client *clp = server->nfs_client;
+
+	if (!atomic_add_unless(&sb->s_active, -1, 1)) {
+		rcu_read_lock();
+		snprintf(buf, sizeof(buf),
+			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
+		rcu_read_unlock();
+
+		__module_get(THIS_MODULE);
+		task = kthread_run(nfs_deactivate_super_async_work, sb,
+				"%s-deactivate-super", buf);
+		if (IS_ERR(task)) {
+			pr_err("%s: kthread_run: %ld\n",
+				__func__, PTR_ERR(task));
+			/* make synchronous call and hope for the best */
+			deactivate_super(sb);
+			module_put(THIS_MODULE);
+		}
+	}
+}
+
+void nfs_sb_deactive_async(struct super_block *sb)
+{
+	struct nfs_server *server = NFS_SB(sb);
+
+	if (atomic_dec_and_test(&server->active))
+		nfs_deactivate_super_async(sb);
+}
+EXPORT_SYMBOL_GPL(nfs_sb_deactive_async);
+
 /*
  * Deliver file system statistics to userspace
  */
diff --git a/fs/nfs/unlink.c b/fs/nfs/unlink.c
index 13cea637eff..3f79c77153b 100644
--- a/fs/nfs/unlink.c
+++ b/fs/nfs/unlink.c
@@ -95,7 +95,7 @@ static void nfs_async_unlink_release(void *calldata)
 
 	nfs_dec_sillycount(data->dir);
 	nfs_free_unlinkdata(data);
-	nfs_sb_deactive(sb);
+	nfs_sb_deactive_async(sb);
 }
 
 static void nfs_unlink_prepare(struct rpc_task *task, void *calldata)