NFS: Fix a deadlock with lazy umount

We can't allow rpc callback functions like task->tk_ops->rpc_call_prepare()
and task->tk_ops->rpc_call_done() to call mntput() in any way, since
that will cause a deadlock when the call to rpc_shutdown_client() attempts
to wait on 'task' to complete.

We can avoid the above deadlock by moving calls to mntput to
task->tk_ops->rpc_release() callback, since at that time the task will be
marked as completed, and so rpc_shutdown_client won't attempt to wait on
it.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

1 files changed, 5 insertions, 2 deletions

diff --git a/fs/nfs/read.c b/fs/nfs/read.c
index 3d7d9631e125..fab0d3720a03 100644
--- a/fs/nfs/read.c
+++ b/fs/nfs/read.c
@@ -73,7 +73,10 @@ static void nfs_readdata_free(struct nfs_read_data *rdata)
 
 void nfs_readdata_release(void *data)
 {
-        nfs_readdata_free(data);
+        struct nfs_read_data *rdata = data;
+
+        put_nfs_open_context(rdata->args.context);
+        nfs_readdata_free(rdata);
 }
 
 static
@@ -186,7 +189,7 @@ static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data,
         data->args.pgbase = req->wb_pgbase + offset;
         data->args.pages  = data->pagevec;
         data->args.count  = count;
-        data->args.context = req->wb_context;
+        data->args.context = get_nfs_open_context(req->wb_context);
 
         data->res.fattr   = &data->fattr;
         data->res.count   = count;