iscsi-target: Fix immediate queue starvation regression with DATAIN

This patch addresses a v3.5+ regression in iscsi-target where TX thread
process context -> handle_response_queue() execution is allowed to run
unbounded while servicing constant outgoing flow of ISTATE_SEND_DATAIN
response state.

This ends up preventing memory release of StatSN acknowledged commands
in a timely manner when under heavy large block streaming DATAIN
workloads.

The regression bug was initially introduced with:

commit 6f3c0e6
Author: Andy Grover <agrover@redhat.com>
Date:   Tue Apr 3 15:51:09 2012 -0700

    target/iscsi: Refactor target_tx_thread immediate+response queue loops

Go ahead and follow original iscsi_target_tx_thread() logic and check
to break for immediate queue processing after each DataIN Sequence and/or
Response PDU has been sent.

Reported-by: Benjamin ESTRABAUD <be@mpstor.com>
Cc: Andy Grover <agrover@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

drivers/target/iscsi/iscsi_target.c

@@ -3583,6 +3583,10 @@ static int handle_response_queue(struct iscsi_conn *conn)
 				spin_lock_bh(&cmd->istate_lock);
 				cmd->i_state = ISTATE_SENT_STATUS;
 				spin_unlock_bh(&cmd->istate_lock);
+
+				if (atomic_read(&conn->check_immediate_queue))
+					return 1;
+
 				continue;
 			} else if (ret == 2) {
 				/* Still must send status,
@@ -3672,7 +3676,7 @@ static int handle_response_queue(struct iscsi_conn *conn)
 		}
 
 		if (atomic_read(&conn->check_immediate_queue))
-			break;
+			return 1;
 	}
 
 	return 0;
@@ -3716,12 +3720,15 @@ int iscsi_target_tx_thread(void *arg)
 		     signal_pending(current))
 			goto transport_err;
 
+get_immediate:
 		ret = handle_immediate_queue(conn);
 		if (ret < 0)
 			goto transport_err;
 
 		ret = handle_response_queue(conn);
-		if (ret == -EAGAIN)
+		if (ret == 1)
+			goto get_immediate;
+		else if (ret == -EAGAIN)
 			goto restart;
 		else if (ret < 0)
 			goto transport_err;