[PATCH] NFS4: Apply delay_retrans to async operations

[PATCH] NFS4: Apply delay_retrans to async operations

[Joshua Watt]

From: Joshua Watt <jpewhacker@gmail.com>

The setting of delay_retrans is applied to synchronous RPC operations
because the retransmit count is stored in same struct nfs4_exception
that is passed each time an error is checked. However, for asynchronous
operations (READ, WRITE, LOCKU, CLOSE, DELEGRETURN), a new struct
nfs4_exception is made on the stack each time the task callback is
invoked. This means that the retransmit count is always zero and thus
delay_retrans never takes effect.

Apply delay_retrans to these operations by tracking and updating their
retransmit count.

Change-Id: Ieb33e046c2b277cb979caa3faca7f52faf0568c9
Signed-off-by: Joshua Watt <jpewhacker@gmail.com>
---
 fs/nfs/nfs4proc.c       | 13 +++++++++++++
 include/linux/nfs_xdr.h |  1 +
 2 files changed, 14 insertions(+)

diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index f58098417142..411776718494 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -3636,6 +3636,7 @@ struct nfs4_closedata {
 	} lr;
 	struct nfs_fattr fattr;
 	unsigned long timestamp;
+	unsigned short retrans;
 };
 
 static void nfs4_free_closedata(void *data)
@@ -3664,6 +3665,7 @@ static void nfs4_close_done(struct rpc_task *task, void *data)
 		.state = state,
 		.inode = calldata->inode,
 		.stateid = &calldata->arg.stateid,
+		.retrans = calldata->retrans,
 	};
 
 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
@@ -3711,6 +3713,7 @@ static void nfs4_close_done(struct rpc_task *task, void *data)
 		default:
 			task->tk_status = nfs4_async_handle_exception(task,
 					server, task->tk_status, &exception);
+			calldata->retrans = exception.retrans;
 			if (exception.retry)
 				goto out_restart;
 	}
@@ -5593,9 +5596,11 @@ static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
 			.inode = hdr->inode,
 			.state = hdr->args.context->state,
 			.stateid = &hdr->args.stateid,
+			.retrans = hdr->retrans,
 		};
 		task->tk_status = nfs4_async_handle_exception(task,
 				server, task->tk_status, &exception);
+		hdr->retrans = exception.retrans;
 		if (exception.retry) {
 			rpc_restart_call_prepare(task);
 			return -EAGAIN;
@@ -5709,10 +5714,12 @@ static int nfs4_write_done_cb(struct rpc_task *task,
 			.inode = hdr->inode,
 			.state = hdr->args.context->state,
 			.stateid = &hdr->args.stateid,
+			.retrans = hdr->retrans,
 		};
 		task->tk_status = nfs4_async_handle_exception(task,
 				NFS_SERVER(inode), task->tk_status,
 				&exception);
+		hdr->retrans = exception.retrans;
 		if (exception.retry) {
 			rpc_restart_call_prepare(task);
 			return -EAGAIN;
@@ -6726,6 +6733,7 @@ struct nfs4_delegreturndata {
 	struct nfs_fh fh;
 	nfs4_stateid stateid;
 	unsigned long timestamp;
+	unsigned short retrans;
 	struct {
 		struct nfs4_layoutreturn_args arg;
 		struct nfs4_layoutreturn_res res;
@@ -6746,6 +6754,7 @@ static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
 		.inode = data->inode,
 		.stateid = &data->stateid,
 		.task_is_privileged = data->args.seq_args.sa_privileged,
+		.retrans = data->retrans,
 	};
 
 	if (!nfs4_sequence_done(task, &data->res.seq_res))
@@ -6817,6 +6826,7 @@ static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
 		task->tk_status = nfs4_async_handle_exception(task,
 				data->res.server, task->tk_status,
 				&exception);
+		data->retrans = exception.retrans;
 		if (exception.retry)
 			goto out_restart;
 	}
@@ -7093,6 +7103,7 @@ struct nfs4_unlockdata {
 	struct file_lock fl;
 	struct nfs_server *server;
 	unsigned long timestamp;
+	unsigned short retrans;
 };
 
 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
@@ -7147,6 +7158,7 @@ static void nfs4_locku_done(struct rpc_task *task, void *data)
 	struct nfs4_exception exception = {
 		.inode = calldata->lsp->ls_state->inode,
 		.stateid = &calldata->arg.stateid,
+		.retrans = calldata->retrans,
 	};
 
 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
@@ -7180,6 +7192,7 @@ static void nfs4_locku_done(struct rpc_task *task, void *data)
 			task->tk_status = nfs4_async_handle_exception(task,
 					calldata->server, task->tk_status,
 					&exception);
+			calldata->retrans = exception.retrans;
 			if (exception.retry)
 				rpc_restart_call_prepare(task);
 	}
diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h
index d56583572c98..31463286402f 100644
--- a/include/linux/nfs_xdr.h
+++ b/include/linux/nfs_xdr.h
@@ -1659,6 +1659,7 @@ struct nfs_pgio_header {
 	void			*netfs;
 #endif
 
+	unsigned short		retrans;
 	int			pnfs_error;
 	int			error;		/* merge with pnfs_error */
 	unsigned int		good_bytes;	/* boundary of good data */
-- 
2.51.0

Re: [PATCH] NFS4: Apply delay_retrans to async operations

[Benjamin Coddington]

On 7 Oct 2025, at 17:22, Joshua Watt wrote:

> From: Joshua Watt <jpewhacker@gmail.com>
>
> The setting of delay_retrans is applied to synchronous RPC operations
> because the retransmit count is stored in same struct nfs4_exception
> that is passed each time an error is checked. However, for asynchronous
> operations (READ, WRITE, LOCKU, CLOSE, DELEGRETURN), a new struct
> nfs4_exception is made on the stack each time the task callback is
> invoked. This means that the retransmit count is always zero and thus
> delay_retrans never takes effect.
>
> Apply delay_retrans to these operations by tracking and updating their
> retransmit count.

Looks correct.  I wish we didn't have to assign these values back and forth,
but I don't see a better way to fix this.

Reviewed-by: Benjamin Coddington <bcodding@redhat.com>

Ben