Command queueing is now generally disabled in PIO modes. No matter what

the cause is, it didn't work over here at all in this combination.


git-svn-id: file:///srv/svn/repos/haiku/trunk/current@10629 a95241bf-73f2-0310-859d-f6bbb57e9c96

src/add-ons/kernel/bus_managers/ide/ide_sim.c

@@ -1,5 +1,7 @@
 /*
+ * Copyright 2005, Axel Dörfler, [email protected]. All rights reserved.
  * Copyright 2002/03, Thomas Kurschel. All rights reserved.
+ *
  * Distributed under the terms of the MIT License.
  */
 
@@ -617,12 +619,21 @@ ide_sim_init_bus(pnp_node_handle node, void *user_cookie, void **cookie)
 
 	SHOW_FLOW(2, "can_dma: %d", bus->can_DMA);
 
-	if (pnp->get_attr_uint8(node, IDE_CONTROLLER_CAN_CQ_ITEM, &bus->can_CQ, true) != B_OK)
+	if (bus->can_DMA) {
+		if (pnp->get_attr_uint8(node, IDE_CONTROLLER_CAN_CQ_ITEM, &bus->can_CQ, true) != B_OK) {
 			// per default, command queuing is supported unless the driver
 			// reports problems (queuing should be transparent to
 			// controller, but for sure there is some buggy, over-optimizing
 			// controller out there)
 			bus->can_CQ = true;
+		}
+	} else {
+		// I am not sure if it's a problem of the driver or the drive (probably the
+		// former), but we're generally disable command queueing in case of PIO
+		// transfers. Since those should be rare on a real system (as is CQ support
+		// in the drive), it's not really worth investigating, though.
+		bus->can_CQ = false;
+	}
 
 	res = pnp->load_driver(pnp->get_parent(node), bus, 
 			(pnp_driver_info **)&bus->controller,