NFS: Don't accept more reads/writes if the open context recovery failed

If the state recovery failed, we want to ensure that the application
doesn't try to use the same file descriptor for more reads or writes.

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

6 files changed, 31 insertions, 0 deletions

diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c
index 1ee5737211d7..4ba32e23eddd 100644
--- a/fs/nfs/nfs4filelayout.c
+++ b/fs/nfs/nfs4filelayout.c
@@ -305,6 +305,10 @@ static void filelayout_read_prepare(struct rpc_task *task, void *data)
 {
 	struct nfs_read_data *rdata = data;
 
+	if (unlikely(test_bit(NFS_CONTEXT_BAD, &rdata->args.context->flags))) {
+		rpc_exit(task, -EIO);
+		return;
+	}
 	if (filelayout_reset_to_mds(rdata->header->lseg)) {
 		dprintk("%s task %u reset io to MDS\n", __func__, task->tk_pid);
 		filelayout_reset_read(rdata);
@@ -407,6 +411,10 @@ static void filelayout_write_prepare(struct rpc_task *task, void *data)
 {
 	struct nfs_write_data *wdata = data;
 
+	if (unlikely(test_bit(NFS_CONTEXT_BAD, &wdata->args.context->flags))) {
+		rpc_exit(task, -EIO);
+		return;
+	}
 	if (filelayout_reset_to_mds(wdata->header->lseg)) {
 		dprintk("%s task %u reset io to MDS\n", __func__, task->tk_pid);
 		filelayout_reset_write(wdata);
diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c
index fec1c5bb4863..8db102c7add6 100644
--- a/fs/nfs/nfs4state.c
+++ b/fs/nfs/nfs4state.c
@@ -1328,9 +1328,25 @@ void nfs_inode_find_state_and_recover(struct inode *inode,
 		nfs4_schedule_state_manager(clp);
 }
 
+static void nfs4_state_mark_open_context_bad(struct nfs4_state *state)
+{
+	struct inode *inode = state->inode;
+	struct nfs_inode *nfsi = NFS_I(inode);
+	struct nfs_open_context *ctx;
+
+	spin_lock(&inode->i_lock);
+	list_for_each_entry(ctx, &nfsi->open_files, list) {
+		if (ctx->state != state)
+			continue;
+		set_bit(NFS_CONTEXT_BAD, &ctx->flags);
+	}
+	spin_unlock(&inode->i_lock);
+}
+
 static void nfs4_state_mark_recovery_failed(struct nfs4_state *state, int error)
 {
 	set_bit(NFS_STATE_RECOVERY_FAILED, &state->flags);
+	nfs4_state_mark_open_context_bad(state);
 }
 
 
diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c
index e56e846e9d2d..7f0933086b36 100644
--- a/fs/nfs/pagelist.c
+++ b/fs/nfs/pagelist.c
@@ -104,6 +104,8 @@ nfs_create_request(struct nfs_open_context *ctx, struct inode *inode,
 	struct nfs_page		*req;
 	struct nfs_lock_context *l_ctx;
 
+	if (test_bit(NFS_CONTEXT_BAD, &ctx->flags))
+		return ERR_PTR(-EBADF);
 	/* try to allocate the request struct */
 	req = nfs_page_alloc();
 	if (req == NULL)
diff --git a/fs/nfs/read.c b/fs/nfs/read.c
index a5e5d9899d56..70a26c651f09 100644
--- a/fs/nfs/read.c
+++ b/fs/nfs/read.c
@@ -514,6 +514,8 @@ void nfs_read_prepare(struct rpc_task *task, void *calldata)
 {
 	struct nfs_read_data *data = calldata;
 	NFS_PROTO(data->header->inode)->read_rpc_prepare(task, data);
+	if (unlikely(test_bit(NFS_CONTEXT_BAD, &data->args.context->flags)))
+		rpc_exit(task, -EIO);
 }
 
 static const struct rpc_call_ops nfs_read_common_ops = {
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index c483cc50b82e..a2c7c28049d5 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -1251,6 +1251,8 @@ void nfs_write_prepare(struct rpc_task *task, void *calldata)
 {
 	struct nfs_write_data *data = calldata;
 	NFS_PROTO(data->header->inode)->write_rpc_prepare(task, data);
+	if (unlikely(test_bit(NFS_CONTEXT_BAD, &data->args.context->flags)))
+		rpc_exit(task, -EIO);
 }
 
 void nfs_commit_prepare(struct rpc_task *task, void *calldata)
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index 1cc25682b20b..f6b1956f3c86 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -77,6 +77,7 @@ struct nfs_open_context {
 	unsigned long flags;
 #define NFS_CONTEXT_ERROR_WRITE		(0)
 #define NFS_CONTEXT_RESEND_WRITES	(1)
+#define NFS_CONTEXT_BAD			(2)
 	int error;
 
 	struct list_head list;