SUNRPC: Use the ENOTCONN error on socket disconnect

Trond Myklebust · Trond Myklebust · commit 27adc785928a · 2019-03-15T13:08:20.000-04:00
When the socket is closed, we currently send an EAGAIN error to all
pending requests in order to ask them to retransmit. Use ENOTCONN
instead, to ensure that they try to reconnect before attempting to
transmit.
This also helps SOFTCONN tasks to behave correctly in this
situation.

Signed-off-by: Trond Myklebust &lt;trond.myklebust@hammerspace.com&gt;
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
@@ -664,7 +664,7 @@ void xprt_disconnect_done(struct rpc_xprt *xprt)
 	spin_lock_bh(&xprt->transport_lock);
 	xprt_clear_connected(xprt);
 	xprt_clear_write_space_locked(xprt);
-	xprt_wake_pending_tasks(xprt, -EAGAIN);
+	xprt_wake_pending_tasks(xprt, -ENOTCONN);
 	spin_unlock_bh(&xprt->transport_lock);
 }
 EXPORT_SYMBOL_GPL(xprt_disconnect_done);

Original file line number	Diff line number	Diff line change
`@@ -664,7 +664,7 @@ void xprt_disconnect_done(struct rpc_xprt *xprt)`
`664`	`664`	`spin_lock_bh(&xprt->transport_lock);`
`665`	`665`	`xprt_clear_connected(xprt);`
`666`	`666`	`xprt_clear_write_space_locked(xprt);`
`667`		`- xprt_wake_pending_tasks(xprt, -EAGAIN);`
	`667`	`+ xprt_wake_pending_tasks(xprt, -ENOTCONN);`
`668`	`668`	`spin_unlock_bh(&xprt->transport_lock);`
`669`	`669`	`}`
`670`	`670`	`EXPORT_SYMBOL_GPL(xprt_disconnect_done);`