NTB: Default to CPU memcpy for performance

Disable DMA usage by default, since the CPU provides much better
performance with write combining.  Provide a module parameter to enable
DMA usage when offloading the memcpy is preferred.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Allen Hubbe <Allen.Hubbe@emc.com>
Signed-off-by: Jon Mason <jdmason@kudzu.us>
diff --git a/drivers/ntb/ntb_transport.c b/drivers/ntb/ntb_transport.c
index 7a765d3..e07b056 100644
--- a/drivers/ntb/ntb_transport.c
+++ b/drivers/ntb/ntb_transport.c
@@ -88,6 +88,10 @@
 module_param(copy_bytes, uint, 0644);
 MODULE_PARM_DESC(copy_bytes, "Threshold under which NTB will use the CPU to copy instead of DMA");
 
+static bool use_dma;
+module_param(use_dma, bool, 0644);
+MODULE_PARM_DESC(use_dma, "Use DMA engine to perform large data copy");
+
 static struct dentry *nt_debugfs_dir;
 
 struct ntb_queue_entry {
@@ -1589,10 +1593,15 @@
 	dma_cap_zero(dma_mask);
 	dma_cap_set(DMA_MEMCPY, dma_mask);
 
-	qp->dma_chan = dma_request_channel(dma_mask, ntb_dma_filter_fn,
-					   (void *)(unsigned long)node);
-	if (!qp->dma_chan)
-		dev_info(&pdev->dev, "Unable to allocate DMA channel, using CPU instead\n");
+	if (use_dma) {
+		qp->dma_chan = dma_request_channel(dma_mask, ntb_dma_filter_fn,
+						   (void *)(unsigned long)node);
+		if (!qp->dma_chan)
+			dev_info(&pdev->dev, "Unable to allocate DMA channel\n");
+	} else {
+		qp->dma_chan = NULL;
+	}
+	dev_dbg(&pdev->dev, "Using %s memcpy\n", qp->dma_chan ? "DMA" : "CPU");
 
 	for (i = 0; i < NTB_QP_DEF_NUM_ENTRIES; i++) {
 		entry = kzalloc_node(sizeof(*entry), GFP_ATOMIC, node);