From: Jon Mason <jon.mason@intel.com>
Date: Tue, 19 Nov 2013 16:34:47 +0000 (-0700)
Subject: NTB: correct dmaengine_get/put usage
X-Git-Tag: v3.13-rc2~28^2~4
X-Git-Url: http://git.cascardo.eti.br/?a=commitdiff_plain;h=94681194e667c33ba303f1225814197594cbfc34;p=cascardo%2Flinux.git

NTB: correct dmaengine_get/put usage

dmaengine_get() causes the initialization of the per-cpu channel tables.
It needs to be called prior to dma_find_channel().

Initial version by Dan Williams <dan.j.williams@intel.com>

Signed-off-by: Jon Mason <jon.mason@intel.com>
---

diff --git a/drivers/ntb/ntb_transport.c b/drivers/ntb/ntb_transport.c
index 7010f23dab18..172501d47600 100644
--- a/drivers/ntb/ntb_transport.c
+++ b/drivers/ntb/ntb_transport.c
@@ -1419,11 +1419,12 @@ ntb_transport_create_queue(void *data, struct pci_dev *pdev,
 	qp->tx_handler = handlers->tx_handler;
 	qp->event_handler = handlers->event_handler;
 
+	dmaengine_get();
 	qp->dma_chan = dma_find_channel(DMA_MEMCPY);
-	if (!qp->dma_chan)
+	if (!qp->dma_chan) {
+		dmaengine_put();
 		dev_info(&pdev->dev, "Unable to allocate DMA channel, using CPU instead\n");
-	else
-		dmaengine_get();
+	}
 
 	for (i = 0; i < NTB_QP_DEF_NUM_ENTRIES; i++) {
 		entry = kzalloc(sizeof(struct ntb_queue_entry), GFP_ATOMIC);
@@ -1464,6 +1465,8 @@ err2:
 err1:
 	while ((entry = ntb_list_rm(&qp->ntb_rx_free_q_lock, &qp->rx_free_q)))
 		kfree(entry);
+	if (qp->dma_chan)
+		dmaengine_put();
 	set_bit(free_queue, &nt->qp_bitmap);
 err:
 	return NULL;