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>

diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c
index 16844f98f50e4b6f6ad24eea74470fb6da263adf..e0170407a885a58374bde64718b0661557475c77 100644 (file)
--- a/fs/nfs/direct.c
+++ b/fs/nfs/direct.c
@@ -323,7 +323,7 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq,
                data->inode = inode;
                data->cred = msg.rpc_cred;
                data->args.fh = NFS_FH(inode);
-               data->args.context = ctx;
+               data->args.context = get_nfs_open_context(ctx);
                data->args.offset = pos;
                data->args.pgbase = pgbase;
                data->args.pages = data->pagevec;
@@ -546,6 +546,7 @@ static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
        data->args.fh = NFS_FH(data->inode);
        data->args.offset = 0;
        data->args.count = 0;
+       data->args.context = get_nfs_open_context(dreq->ctx);
        data->res.count = 0;
        data->res.fattr = &data->fattr;
        data->res.verf = &data->verf;
@@ -728,7 +729,7 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq,
                data->inode = inode;
                data->cred = msg.rpc_cred;
                data->args.fh = NFS_FH(inode);
-               data->args.context = ctx;
+               data->args.context = get_nfs_open_context(ctx);
                data->args.offset = pos;
                data->args.pgbase = pgbase;
                data->args.pages = data->pagevec;

diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index 966a8850aa30be5330a524699069718cd00932ce..a499fb58d8583f82356b35ce189a6d0682414bb5 100644 (file)
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -521,8 +521,12 @@ struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
 
 static void __put_nfs_open_context(struct nfs_open_context *ctx, int wait)
 {
-       struct inode *inode = ctx->path.dentry->d_inode;
+       struct inode *inode;
 
+       if (ctx == NULL)
+               return;
+
+       inode = ctx->path.dentry->d_inode;
        if (!atomic_dec_and_lock(&ctx->count, &inode->i_lock))
                return;
        list_del(&ctx->list);

diff --git a/fs/nfs/read.c b/fs/nfs/read.c
index 3d7d9631e1254f7ad898863e66519740af1e34e9..fab0d3720a03b44a1cd09d5feea86a94a4357c36 100644 (file)
--- 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;

diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 80c61fdb2720c4b05f86ff7c590cd0af566e2b8d..69b4158d9a1001122c2950ced3055f4cbc656d0e 100644 (file)
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -105,8 +105,11 @@ static void nfs_writedata_free(struct nfs_write_data *wdata)
        call_rcu_bh(&wdata->task.u.tk_rcu, nfs_writedata_rcu_free);
 }
 
-void nfs_writedata_release(void *wdata)
+void nfs_writedata_release(void *data)
 {
+       struct nfs_write_data *wdata = data;
+
+       put_nfs_open_context(wdata->args.context);
        nfs_writedata_free(wdata);
 }
 
@@ -816,7 +819,7 @@ static void nfs_write_rpcsetup(struct nfs_page *req,
        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->args.stable  = NFS_UNSTABLE;
        if (how & FLUSH_STABLE) {
                data->args.stable = NFS_DATA_SYNC;
@@ -1153,8 +1156,11 @@ int nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
 
 
 #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
-void nfs_commit_release(void *wdata)
+void nfs_commit_release(void *data)
 {
+       struct nfs_write_data *wdata = data;
+
+       put_nfs_open_context(wdata->args.context);
        nfs_commit_free(wdata);
 }
 
@@ -1197,6 +1203,7 @@ static void nfs_commit_rpcsetup(struct list_head *head,
        /* Note: we always request a commit of the entire inode */
        data->args.offset = 0;
        data->args.count  = 0;
+       data->args.context = get_nfs_open_context(first->wb_context);
        data->res.count   = 0;
        data->res.fattr   = &data->fattr;
        data->res.verf    = &data->verf;