async_tx: fix multiple dependency submission

Shrink struct dma_async_tx_descriptor and introduce
async_tx_channel_switch to properly inject a channel switch interrupt in
the descriptor stream.  This simplifies the locking model as drivers no
longer need to handle dma_async_tx_descriptor.lock.

Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
diff --git a/drivers/dma/iop-adma.c b/drivers/dma/iop-adma.c
index f82b090..21854cd 100644
--- a/drivers/dma/iop-adma.c
+++ b/drivers/dma/iop-adma.c
@@ -63,7 +63,6 @@
 	struct iop_adma_chan *iop_chan, dma_cookie_t cookie)
 {
 	BUG_ON(desc->async_tx.cookie < 0);
-	spin_lock_bh(&desc->async_tx.lock);
 	if (desc->async_tx.cookie > 0) {
 		cookie = desc->async_tx.cookie;
 		desc->async_tx.cookie = 0;
@@ -101,7 +100,6 @@
 
 	/* run dependent operations */
 	async_tx_run_dependencies(&desc->async_tx);
-	spin_unlock_bh(&desc->async_tx.lock);
 
 	return cookie;
 }
@@ -275,8 +273,11 @@
 
 static void iop_adma_tasklet(unsigned long data)
 {
-	struct iop_adma_chan *chan = (struct iop_adma_chan *) data;
-	__iop_adma_slot_cleanup(chan);
+	struct iop_adma_chan *iop_chan = (struct iop_adma_chan *) data;
+
+	spin_lock(&iop_chan->lock);
+	__iop_adma_slot_cleanup(iop_chan);
+	spin_unlock(&iop_chan->lock);
 }
 
 static struct iop_adma_desc_slot *