Merge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx

* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx:
  DMAENGINE: extend the control command to include an arg
  async_tx: trim dma_async_tx_descriptor in 'no channel switch' case
  DMAENGINE: DMA40 fix for allocation of logical channel 0
  DMAENGINE: DMA40 support paused channel status
  dmaengine: mpc512x: Use resource_size
  DMA ENGINE: Do not reset 'private' of channel
  ioat: Remove duplicated devm_kzalloc() calls for ioatdma_device
  ioat3: disable cacheline-unaligned transfers for raid operations
  ioat2,3: convert to producer/consumer locking
  ioat: convert to circ_buf
  DMAENGINE: Support for ST-Ericssons DMA40 block v3
  async_tx: use of kzalloc/kfree requires the include of slab.h
  dmaengine: provide helper for setting txstate
  DMAENGINE: generic channel status v2
  DMAENGINE: generic slave control v2
  dma: timb-dma: Update comment and fix compiler warning
  dma: Add timb-dma
  DMAENGINE: COH 901 318 fix bytesleft
  DMAENGINE: COH 901 318 rename confusing vars

24 files changed, 4845 insertions, 450 deletions

diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index c27f80e5d53..1b8877922fb 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -141,6 +141,13 @@ config COH901318
 	help
 	  Enable support for ST-Ericsson COH 901 318 DMA.
 
+config STE_DMA40
+	bool "ST-Ericsson DMA40 support"
+	depends on ARCH_U8500
+	select DMA_ENGINE
+	help
+	  Support for ST-Ericsson DMA40 controller
+
 config AMCC_PPC440SPE_ADMA
 	tristate "AMCC PPC440SPe ADMA support"
 	depends on 440SPe || 440SP
@@ -149,6 +156,13 @@ config AMCC_PPC440SPE_ADMA
 	help
 	  Enable support for the AMCC PPC440SPe RAID engines.
 
+config TIMB_DMA
+	tristate "Timberdale FPGA DMA support"
+	depends on MFD_TIMBERDALE || HAS_IOMEM
+	select DMA_ENGINE
+	help
+	  Enable support for the Timberdale FPGA DMA engine.
+
 config ARCH_HAS_ASYNC_TX_FIND_CHANNEL
 	bool
 
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index 22bba3d5e2b..20881426c1a 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -20,3 +20,5 @@ obj-$(CONFIG_TXX9_DMAC) += txx9dmac.o
 obj-$(CONFIG_SH_DMAE) += shdma.o
 obj-$(CONFIG_COH901318) += coh901318.o coh901318_lli.o
 obj-$(CONFIG_AMCC_PPC440SPE_ADMA) += ppc4xx/
+obj-$(CONFIG_TIMB_DMA) += timb_dma.o
+obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
diff --git a/drivers/dma/at_hdmac.c b/drivers/dma/at_hdmac.c
index 278cf5bceef..bd5250e8c00 100644
--- a/drivers/dma/at_hdmac.c
+++ b/drivers/dma/at_hdmac.c
@@ -760,13 +760,18 @@ err_desc_get:
 	return NULL;
 }
 
-static void atc_terminate_all(struct dma_chan *chan)
+static int atc_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+		       unsigned long arg)
 {
 	struct at_dma_chan	*atchan = to_at_dma_chan(chan);
 	struct at_dma		*atdma = to_at_dma(chan->device);
 	struct at_desc		*desc, *_desc;
 	LIST_HEAD(list);
 
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
 	/*
 	 * This is only called when something went wrong elsewhere, so
 	 * we don't really care about the data. Just disable the
@@ -790,32 +795,30 @@ static void atc_terminate_all(struct dma_chan *chan)
 	/* Flush all pending and queued descriptors */
 	list_for_each_entry_safe(desc, _desc, &list, desc_node)
 		atc_chain_complete(atchan, desc);
+
+	return 0;
 }
 
 /**
- * atc_is_tx_complete - poll for transaction completion
+ * atc_tx_status - poll for transaction completion
  * @chan: DMA channel
  * @cookie: transaction identifier to check status of
- * @done: if not %NULL, updated with last completed transaction
- * @used: if not %NULL, updated with last used transaction
+ * @txstate: if not %NULL updated with transaction state
  *
- * If @done and @used are passed in, upon return they reflect the driver
+ * If @txstate is passed in, upon return it reflect the driver
  * internal state and can be used with dma_async_is_complete() to check
  * the status of multiple cookies without re-checking hardware state.
  */
 static enum dma_status
-atc_is_tx_complete(struct dma_chan *chan,
+atc_tx_status(struct dma_chan *chan,
 		dma_cookie_t cookie,
-		dma_cookie_t *done, dma_cookie_t *used)
+		struct dma_tx_state *txstate)
 {
 	struct at_dma_chan	*atchan = to_at_dma_chan(chan);
 	dma_cookie_t		last_used;
 	dma_cookie_t		last_complete;
 	enum dma_status		ret;
 
-	dev_vdbg(chan2dev(chan), "is_tx_complete: %d (d%d, u%d)\n",
-			cookie, done ? *done : 0, used ? *used : 0);
-
 	spin_lock_bh(&atchan->lock);
 
 	last_complete = atchan->completed_cookie;
@@ -833,10 +836,10 @@ atc_is_tx_complete(struct dma_chan *chan,
 
 	spin_unlock_bh(&atchan->lock);
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
+	dev_vdbg(chan2dev(chan), "tx_status: %d (d%d, u%d)\n",
+		 cookie, last_complete ? last_complete : 0,
+		 last_used ? last_used : 0);
 
 	return ret;
 }
@@ -1082,7 +1085,7 @@ static int __init at_dma_probe(struct platform_device *pdev)
 	/* set base routines */
 	atdma->dma_common.device_alloc_chan_resources = atc_alloc_chan_resources;
 	atdma->dma_common.device_free_chan_resources = atc_free_chan_resources;
-	atdma->dma_common.device_is_tx_complete = atc_is_tx_complete;
+	atdma->dma_common.device_tx_status = atc_tx_status;
 	atdma->dma_common.device_issue_pending = atc_issue_pending;
 	atdma->dma_common.dev = &pdev->dev;
 
@@ -1092,7 +1095,7 @@ static int __init at_dma_probe(struct platform_device *pdev)
 
 	if (dma_has_cap(DMA_SLAVE, atdma->dma_common.cap_mask)) {
 		atdma->dma_common.device_prep_slave_sg = atc_prep_slave_sg;
-		atdma->dma_common.device_terminate_all = atc_terminate_all;
+		atdma->dma_common.device_control = atc_control;
 	}
 
 	dma_writel(atdma, EN, AT_DMA_ENABLE);
diff --git a/drivers/dma/coh901318.c b/drivers/dma/coh901318.c
index 1656fdcdb6c..a724e6be1b4 100644
--- a/drivers/dma/coh901318.c
+++ b/drivers/dma/coh901318.c
@@ -37,7 +37,7 @@ struct coh901318_desc {
 	struct list_head node;
 	struct scatterlist *sg;
 	unsigned int sg_len;
-	struct coh901318_lli *data;
+	struct coh901318_lli *lli;
 	enum dma_data_direction dir;
 	unsigned long flags;
 };
@@ -283,7 +283,7 @@ static int coh901318_start(struct coh901318_chan *cohc)
 }
 
 static int coh901318_prep_linked_list(struct coh901318_chan *cohc,
-				      struct coh901318_lli *data)
+				      struct coh901318_lli *lli)
 {
 	int channel = cohc->id;
 	void __iomem *virtbase = cohc->base->virtbase;
@@ -292,18 +292,18 @@ static int coh901318_prep_linked_list(struct coh901318_chan *cohc,
 		     COH901318_CX_STAT_SPACING*channel) &
 	       COH901318_CX_STAT_ACTIVE);
 
-	writel(data->src_addr,
+	writel(lli->src_addr,
 	       virtbase + COH901318_CX_SRC_ADDR +
 	       COH901318_CX_SRC_ADDR_SPACING * channel);
 
-	writel(data->dst_addr, virtbase +
+	writel(lli->dst_addr, virtbase +
 	       COH901318_CX_DST_ADDR +
 	       COH901318_CX_DST_ADDR_SPACING * channel);
 
-	writel(data->link_addr, virtbase + COH901318_CX_LNK_ADDR +
+	writel(lli->link_addr, virtbase + COH901318_CX_LNK_ADDR +
 	       COH901318_CX_LNK_ADDR_SPACING * channel);
 
-	writel(data->control, virtbase + COH901318_CX_CTRL +
+	writel(lli->control, virtbase + COH901318_CX_CTRL +
 	       COH901318_CX_CTRL_SPACING * channel);
 
 	return 0;
@@ -408,33 +408,107 @@ coh901318_first_queued(struct coh901318_chan *cohc)
 	return d;
 }
 
+static inline u32 coh901318_get_bytes_in_lli(struct coh901318_lli *in_lli)
+{
+	struct coh901318_lli *lli = in_lli;
+	u32 bytes = 0;
+
+	while (lli) {
+		bytes += lli->control & COH901318_CX_CTRL_TC_VALUE_MASK;
+		lli = lli->virt_link_addr;
+	}
+	return bytes;
+}
+
 /*
- * DMA start/stop controls
+ * Get the number of bytes left to transfer on this channel,
+ * it is unwise to call this before stopping the channel for
+ * absolute measures, but for a rough guess you can still call
+ * it.
  */
-u32 coh901318_get_bytes_left(struct dma_chan *chan)
+static u32 coh901318_get_bytes_left(struct dma_chan *chan)
 {
-	unsigned long flags;
-	u32 ret;
 	struct coh901318_chan *cohc = to_coh901318_chan(chan);
+	struct coh901318_desc *cohd;
+	struct list_head *pos;
+	unsigned long flags;
+	u32 left = 0;
+	int i = 0;
 
 	spin_lock_irqsave(&cohc->lock, flags);
 
-	/* Read transfer count value */
-	ret = readl(cohc->base->virtbase +
-		    COH901318_CX_CTRL+COH901318_CX_CTRL_SPACING *
-		    cohc->id) & COH901318_CX_CTRL_TC_VALUE_MASK;
+	/*
+	 * If there are many queued jobs, we iterate and add the
+	 * size of them all. We take a special look on the first
+	 * job though, since it is probably active.
+	 */
+	list_for_each(pos, &cohc->active) {
+		/*
+		 * The first job in the list will be working on the
+		 * hardware. The job can be stopped but still active,
+		 * so that the transfer counter is somewhere inside
+		 * the buffer.
+		 */
+		cohd = list_entry(pos, struct coh901318_desc, node);
+
+		if (i == 0) {
+			struct coh901318_lli *lli;
+			dma_addr_t ladd;
+
+			/* Read current transfer count value */
+			left = readl(cohc->base->virtbase +
+				     COH901318_CX_CTRL +
+				     COH901318_CX_CTRL_SPACING * cohc->id) &
+				COH901318_CX_CTRL_TC_VALUE_MASK;
+
+			/* See if the transfer is linked... */
+			ladd = readl(cohc->base->virtbase +
+				     COH901318_CX_LNK_ADDR +
+				     COH901318_CX_LNK_ADDR_SPACING *
+				     cohc->id) &
+				~COH901318_CX_LNK_LINK_IMMEDIATE;
+			/* Single transaction */
+			if (!ladd)
+				continue;
+
+			/*
+			 * Linked transaction, follow the lli, find the
+			 * currently processing lli, and proceed to the next
+			 */
+			lli = cohd->lli;
+			while (lli && lli->link_addr != ladd)
+				lli = lli->virt_link_addr;
+
+			if (lli)
+				lli = lli->virt_link_addr;
+
+			/*
+			 * Follow remaining lli links around to count the total
+			 * number of bytes left
+			 */
+			left += coh901318_get_bytes_in_lli(lli);
+		} else {
+			left += coh901318_get_bytes_in_lli(cohd->lli);
+		}
+		i++;
+	}
+
+	/* Also count bytes in the queued jobs */
+	list_for_each(pos, &cohc->queue) {
+		cohd = list_entry(pos, struct coh901318_desc, node);
+		left += coh901318_get_bytes_in_lli(cohd->lli);
+	}
 
 	spin_unlock_irqrestore(&cohc->lock, flags);
 
-	return ret;
+	return left;
 }
-EXPORT_SYMBOL(coh901318_get_bytes_left);
-
 
-/* Stops a transfer without losing data. Enables power save.
-   Use this function in conjunction with coh901318_continue(..)
-*/
-void coh901318_stop(struct dma_chan *chan)
+/*
+ * Pauses a transfer without losing data. Enables power save.
+ * Use this function in conjunction with coh901318_resume.
+ */
+static void coh901318_pause(struct dma_chan *chan)
 {
 	u32 val;
 	unsigned long flags;
@@ -475,12 +549,11 @@ void coh901318_stop(struct dma_chan *chan)
 
 	spin_unlock_irqrestore(&cohc->lock, flags);
 }
-EXPORT_SYMBOL(coh901318_stop);
 
-/* Continues a transfer that has been stopped via 300_dma_stop(..).
+/* Resumes a transfer that has been stopped via 300_dma_stop(..).
    Power save is handled.
 */
-void coh901318_continue(struct dma_chan *chan)
+static void coh901318_resume(struct dma_chan *chan)
 {
 	u32 val;
 	unsigned long flags;
@@ -506,7 +579,6 @@ void coh901318_continue(struct dma_chan *chan)
 
 	spin_unlock_irqrestore(&cohc->lock, flags);
 }
-EXPORT_SYMBOL(coh901318_continue);
 
 bool coh901318_filter_id(struct dma_chan *chan, void *chan_id)
 {
@@ -565,29 +637,30 @@ static int coh901318_config(struct coh901318_chan *cohc,
  */
 static struct coh901318_desc *coh901318_queue_start(struct coh901318_chan *cohc)
 {
-	struct coh901318_desc *cohd_que;
+	struct coh901318_desc *cohd;
 
-	/* start queued jobs, if any
+	/*
+	 * start queued jobs, if any
 	 * TODO: transmit all queued jobs in one go
 	 */
-	cohd_que = coh901318_first_queued(cohc);
+	cohd = coh901318_first_queued(cohc);
 
-	if (cohd_que != NULL) {
+	if (cohd != NULL) {
 		/* Remove from queue */
-		coh901318_desc_remove(cohd_que);
+		coh901318_desc_remove(cohd);
 		/* initiate DMA job */
 		cohc->busy = 1;
 
-		coh901318_desc_submit(cohc, cohd_que);
+		coh901318_desc_submit(cohc, cohd);
 
-		coh901318_prep_linked_list(cohc, cohd_que->data);
+		coh901318_prep_linked_list(cohc, cohd->lli);
 
-		/* start dma job */
+		/* start dma job on this channel */
 		coh901318_start(cohc);
 
 	}
 
-	return cohd_que;
+	return cohd;
 }
 
 /*
@@ -622,7 +695,7 @@ static void dma_tasklet(unsigned long data)
 	cohc->completed = cohd_fin->desc.cookie;
 
 	/* release the lli allocation and remove the descriptor */
-	coh901318_lli_free(&cohc->base->pool, &cohd_fin->data);
+	coh901318_lli_free(&cohc->base->pool, &cohd_fin->lli);
 
 	/* return desc to free-list */
 	coh901318_desc_remove(cohd_fin);
@@ -666,23 +739,44 @@ static void dma_tasklet(unsigned long data)
 /* called from interrupt context */
 static void dma_tc_handle(struct coh901318_chan *cohc)
 {
-	BUG_ON(!cohc->allocated && (list_empty(&cohc->active) ||
-				    list_empty(&cohc->queue)));
-
-	if (!cohc->allocated)
+	/*
+	 * If the channel is not allocated, then we shouldn't have
+	 * any TC interrupts on it.
+	 */
+	if (!cohc->allocated) {
+		dev_err(COHC_2_DEV(cohc), "spurious interrupt from "
+			"unallocated channel\n");
 		return;
+	}
 
 	spin_lock(&cohc->lock);
 
+	/*
+	 * When we reach this point, at least one queue item
+	 * should have been moved over from cohc->queue to
+	 * cohc->active and run to completion, that is why we're
+	 * getting a terminal count interrupt is it not?
+	 * If you get this BUG() the most probable cause is that
+	 * the individual nodes in the lli chain have IRQ enabled,
+	 * so check your platform config for lli chain ctrl.
+	 */
+	BUG_ON(list_empty(&cohc->active));
+
 	cohc->nbr_active_done++;
 
+	/*
+	 * This attempt to take a job from cohc->queue, put it
+	 * into cohc->active and start it.
+	 */
 	if (coh901318_queue_start(cohc) == NULL)
 		cohc->busy = 0;
 
-	BUG_ON(list_empty(&cohc->active));
-
 	spin_unlock(&cohc->lock);
 
+	/*
+	 * This tasklet will remove items from cohc->active
+	 * and thus terminates them.
+	 */
 	if (cohc_chan_conf(cohc)->priority_high)
 		tasklet_hi_schedule(&cohc->tasklet);
 	else
@@ -809,6 +903,7 @@ static irqreturn_t dma_irq_handler(int irq, void *dev_id)
 static int coh901318_alloc_chan_resources(struct dma_chan *chan)
 {
 	struct coh901318_chan	*cohc = to_coh901318_chan(chan);
+	unsigned long flags;
 
 	dev_vdbg(COHC_2_DEV(cohc), "[%s] DMA channel %d\n",
 		 __func__, cohc->id);
@@ -816,11 +911,15 @@ static int coh901318_alloc_chan_resources(struct dma_chan *chan)
 	if (chan->client_count > 1)
 		return -EBUSY;
 
+	spin_lock_irqsave(&cohc->lock, flags);
+
 	coh901318_config(cohc, NULL);
 
 	cohc->allocated = 1;
 	cohc->completed = chan->cookie = 1;
 
+	spin_unlock_irqrestore(&cohc->lock, flags);
+
 	return 1;
 }
 
@@ -843,7 +942,7 @@ coh901318_free_chan_resources(struct dma_chan *chan)
 
 	spin_unlock_irqrestore(&cohc->lock, flags);
 
-	chan->device->device_terminate_all(chan);
+	chan->device->device_control(chan, DMA_TERMINATE_ALL, 0);
 }
 
 
@@ -870,7 +969,7 @@ static struct dma_async_tx_descriptor *
 coh901318_prep_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src,
 		      size_t size, unsigned long flags)
 {
-	struct coh901318_lli *data;
+	struct coh901318_lli *lli;
 	struct coh901318_desc *cohd;
 	unsigned long flg;
 	struct coh901318_chan *cohc = to_coh901318_chan(chan);
@@ -892,23 +991,23 @@ coh901318_prep_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src,
 	if ((lli_len << MAX_DMA_PACKET_SIZE_SHIFT) < size)
 		lli_len++;
 
-	data = coh901318_lli_alloc(&cohc->base->pool, lli_len);
+	lli = coh901318_lli_alloc(&cohc->base->pool, lli_len);
 
-	if (data == NULL)
+	if (lli == NULL)
 		goto err;
 
 	ret = coh901318_lli_fill_memcpy(
-		&cohc->base->pool, data, src, size, dest,
+		&cohc->base->pool, lli, src, size, dest,
 		cohc_chan_param(cohc)->ctrl_lli_chained,
 		ctrl_last);
 	if (ret)
 		goto err;
 
-	COH_DBG(coh901318_list_print(cohc, data));
+	COH_DBG(coh901318_list_print(cohc, lli));
 
 	/* Pick a descriptor to handle this transfer */
 	cohd = coh901318_desc_get(cohc);
-	cohd->data = data;
+	cohd->lli = lli;
 	cohd->flags = flags;
 	cohd->desc.tx_submit = coh901318_tx_submit;
 
@@ -926,7 +1025,7 @@ coh901318_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 			unsigned long flags)
 {
 	struct coh901318_chan *cohc = to_coh901318_chan(chan);
-	struct coh901318_lli *data;
+	struct coh901318_lli *lli;
 	struct coh901318_desc *cohd;
 	const struct coh901318_params *params;
 	struct scatterlist *sg;
@@ -999,13 +1098,13 @@ coh901318_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 	}
 
 	pr_debug("Allocate %d lli:s for this transfer\n", len);
-	data = coh901318_lli_alloc(&cohc->base->pool, len);
+	lli = coh901318_lli_alloc(&cohc->base->pool, len);
 
-	if (data == NULL)
+	if (lli == NULL)
 		goto err_dma_alloc;
 
-	/* initiate allocated data list */
-	ret = coh901318_lli_fill_sg(&cohc->base->pool, data, sgl, sg_len,
+	/* initiate allocated lli list */
+	ret = coh901318_lli_fill_sg(&cohc->base->pool, lli, sgl, sg_len,
 				    cohc_dev_addr(cohc),
 				    ctrl_chained,
 				    ctrl,
@@ -1014,14 +1113,14 @@ coh901318_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 	if (ret)
 		goto err_lli_fill;
 
-	COH_DBG(coh901318_list_print(cohc, data));
+	COH_DBG(coh901318_list_print(cohc, lli));
 
 	/* Pick a descriptor to handle this transfer */
 	cohd = coh901318_desc_get(cohc);
 	cohd->dir = direction;
 	cohd->flags = flags;
 	cohd->desc.tx_submit = coh901318_tx_submit;
-	cohd->data = data;
+	cohd->lli = lli;
 
 	spin_unlock_irqrestore(&cohc->lock, flg);
 
@@ -1035,9 +1134,8 @@ coh901318_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 }
 
 static enum dma_status
-coh901318_is_tx_complete(struct dma_chan *chan,
-			 dma_cookie_t cookie, dma_cookie_t *done,
-			 dma_cookie_t *used)
+coh901318_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
+		 struct dma_tx_state *txstate)
 {
 	struct coh901318_chan *cohc = to_coh901318_chan(chan);
 	dma_cookie_t last_used;
@@ -1049,10 +1147,10 @@ coh901318_is_tx_complete(struct dma_chan *chan,
 
 	ret = dma_async_is_complete(cookie, last_complete, last_used);
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used,
+			 coh901318_get_bytes_left(chan));
+	if (ret == DMA_IN_PROGRESS && cohc->stopped)
+		ret = DMA_PAUSED;
 
 	return ret;
 }
@@ -1065,23 +1163,42 @@ coh901318_issue_pending(struct dma_chan *chan)
 
 	spin_lock_irqsave(&cohc->lock, flags);
 
-	/* Busy means that pending jobs are already being processed */
+	/*
+	 * Busy means that pending jobs are already being processed,
+	 * and then there is no point in starting the queue: the
+	 * terminal count interrupt on the channel will take the next
+	 * job on the queue and execute it anyway.
+	 */
 	if (!cohc->busy)
 		coh901318_queue_start(cohc);
 
 	spin_unlock_irqrestore(&cohc->lock, flags);
 }
 
-static void
-coh901318_terminate_all(struct dma_chan *chan)
+static int
+coh901318_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+		  unsigned long arg)
 {
 	unsigned long flags;
 	struct coh901318_chan *cohc = to_coh901318_chan(chan);
 	struct coh901318_desc *cohd;
 	void __iomem *virtbase = cohc->base->virtbase;
 
-	coh901318_stop(chan);
+	if (cmd == DMA_PAUSE) {
+		coh901318_pause(chan);
+		return 0;
+	}
+
+	if (cmd == DMA_RESUME) {
+		coh901318_resume(chan);
+		return 0;
+	}
 
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
+	/* The remainder of this function terminates the transfer */
+	coh901318_pause(chan);
 	spin_lock_irqsave(&cohc->lock, flags);
 
 	/* Clear any pending BE or TC interrupt */
@@ -1099,7 +1216,7 @@ coh901318_terminate_all(struct dma_chan *chan)
 
 	while ((cohd = coh901318_first_active_get(cohc))) {
 		/* release the lli allocation*/
-		coh901318_lli_free(&cohc->base->pool, &cohd->data);
+		coh901318_lli_free(&cohc->base->pool, &cohd->lli);
 
 		/* return desc to free-list */
 		coh901318_desc_remove(cohd);
@@ -1108,7 +1225,7 @@ coh901318_terminate_all(struct dma_chan *chan)
 
 	while ((cohd = coh901318_first_queued(cohc))) {
 		/* release the lli allocation*/
-		coh901318_lli_free(&cohc->base->pool, &cohd->data);
+		coh901318_lli_free(&cohc->base->pool, &cohd->lli);
 
 		/* return desc to free-list */
 		coh901318_desc_remove(cohd);
@@ -1120,6 +1237,8 @@ coh901318_terminate_all(struct dma_chan *chan)
 	cohc->busy = 0;
 
 	spin_unlock_irqrestore(&cohc->lock, flags);
+
+	return 0;
 }
 void coh901318_base_init(struct dma_device *dma, const int *pick_chans,
 			 struct coh901318_base *base)
@@ -1235,9 +1354,9 @@ static int __init coh901318_probe(struct platform_device *pdev)
 	base->dma_slave.device_alloc_chan_resources = coh901318_alloc_chan_resources;
 	base->dma_slave.device_free_chan_resources = coh901318_free_chan_resources;
 	base->dma_slave.device_prep_slave_sg = coh901318_prep_slave_sg;
-	base->dma_slave.device_is_tx_complete = coh901318_is_tx_complete;
+	base->dma_slave.device_tx_status = coh901318_tx_status;
 	base->dma_slave.device_issue_pending = coh901318_issue_pending;
-	base->dma_slave.device_terminate_all = coh901318_terminate_all;
+	base->dma_slave.device_control = coh901318_control;
 	base->dma_slave.dev = &pdev->dev;
 
 	err = dma_async_device_register(&base->dma_slave);
@@ -1255,9 +1374,9 @@ static int __init coh901318_probe(struct platform_device *pdev)
 	base->dma_memcpy.device_alloc_chan_resources = coh901318_alloc_chan_resources;
 	base->dma_memcpy.device_free_chan_resources = coh901318_free_chan_resources;
 	base->dma_memcpy.device_prep_dma_memcpy = coh901318_prep_memcpy;
-	base->dma_memcpy.device_is_tx_complete = coh901318_is_tx_complete;
+	base->dma_memcpy.device_tx_status = coh901318_tx_status;
 	base->dma_memcpy.device_issue_pending = coh901318_issue_pending;
-	base->dma_memcpy.device_terminate_all = coh901318_terminate_all;
+	base->dma_memcpy.device_control = coh901318_control;
 	base->dma_memcpy.dev = &pdev->dev;
 	/*
 	 * This controller can only access address at even 32bit boundaries,
diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
index d18b5d069d7..9d31d5eb95c 100644
--- a/drivers/dma/dmaengine.c
+++ b/drivers/dma/dmaengine.c
@@ -515,7 +515,6 @@ struct dma_chan *__dma_request_channel(dma_cap_mask_t *mask, dma_filter_fn fn, v
 				break;
 			if (--device->privatecnt == 0)
 				dma_cap_clear(DMA_PRIVATE, device->cap_mask);
-			chan->private = NULL;
 			chan = NULL;
 		}
 	}
@@ -537,7 +536,6 @@ void dma_release_channel(struct dma_chan *chan)
 	/* drop PRIVATE cap enabled by __dma_request_channel() */
 	if (--chan->device->privatecnt == 0)
 		dma_cap_clear(DMA_PRIVATE, chan->device->cap_mask);
-	chan->private = NULL;
 	mutex_unlock(&dma_list_mutex);
 }
 EXPORT_SYMBOL_GPL(dma_release_channel);
@@ -695,11 +693,11 @@ int dma_async_device_register(struct dma_device *device)
 	BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
 		!device->device_prep_slave_sg);
 	BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
-		!device->device_terminate_all);
+		!device->device_control);
 
 	BUG_ON(!device->device_alloc_chan_resources);
 	BUG_ON(!device->device_free_chan_resources);
-	BUG_ON(!device->device_is_tx_complete);
+	BUG_ON(!device->device_tx_status);
 	BUG_ON(!device->device_issue_pending);
 	BUG_ON(!device->dev);
 
@@ -978,7 +976,9 @@ void dma_async_tx_descriptor_init(struct dma_async_tx_descriptor *tx,
 	struct dma_chan *chan)
 {
 	tx->chan = chan;
+	#ifndef CONFIG_ASYNC_TX_DISABLE_CHANNEL_SWITCH
 	spin_lock_init(&tx->lock);
+	#endif
 }
 EXPORT_SYMBOL(dma_async_tx_descriptor_init);
 
@@ -1011,7 +1011,7 @@ EXPORT_SYMBOL_GPL(dma_wait_for_async_tx);
  */
 void dma_run_dependencies(struct dma_async_tx_descriptor *tx)
 {
-	struct dma_async_tx_descriptor *dep = tx->next;
+	struct dma_async_tx_descriptor *dep = txd_next(tx);
 	struct dma_async_tx_descriptor *dep_next;
 	struct dma_chan *chan;
 
@@ -1019,7 +1019,7 @@ void dma_run_dependencies(struct dma_async_tx_descriptor *tx)
 		return;
 
 	/* we'll submit tx->next now, so clear the link */
-	tx->next = NULL;
+	txd_clear_next(tx);
 	chan = dep->chan;
 
 	/* keep submitting up until a channel switch is detected
@@ -1027,14 +1027,14 @@ void dma_run_dependencies(struct dma_async_tx_descriptor *tx)
 	 * processing the interrupt from async_tx_channel_switch
 	 */
 	for (; dep; dep = dep_next) {
-		spin_lock_bh(&dep->lock);
-		dep->parent = NULL;
-		dep_next = dep->next;
+		txd_lock(dep);
+		txd_clear_parent(dep);
+		dep_next = txd_next(dep);
 		if (dep_next && dep_next->chan == chan)
-			dep->next = NULL; /* ->next will be submitted */
+			txd_clear_next(dep); /* ->next will be submitted */
 		else
 			dep_next = NULL; /* submit current dep and terminate */
-		spin_unlock_bh(&dep->lock);
+		txd_unlock(dep);
 
 		dep->tx_submit(dep);
 	}
diff --git a/drivers/dma/dw_dmac.c b/drivers/dma/dw_dmac.c
index d28369f7afd..a3991ab0d67 100644
--- a/drivers/dma/dw_dmac.c
+++ b/drivers/dma/dw_dmac.c
@@ -781,13 +781,18 @@ err_desc_get:
 	return NULL;
 }
 
-static void dwc_terminate_all(struct dma_chan *chan)
+static int dwc_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+		       unsigned long arg)
 {
 	struct dw_dma_chan	*dwc = to_dw_dma_chan(chan);
 	struct dw_dma		*dw = to_dw_dma(chan->device);
 	struct dw_desc		*desc, *_desc;
 	LIST_HEAD(list);
 
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
 	/*
 	 * This is only called when something went wrong elsewhere, so
 	 * we don't really care about the data. Just disable the
@@ -810,12 +815,14 @@ static void dwc_terminate_all(struct dma_chan *chan)
 	/* Flush all pending and queued descriptors */
 	list_for_each_entry_safe(desc, _desc, &list, desc_node)
 		dwc_descriptor_complete(dwc, desc);
+
+	return 0;
 }
 
 static enum dma_status
-dwc_is_tx_complete(struct dma_chan *chan,
-		dma_cookie_t cookie,
-		dma_cookie_t *done, dma_cookie_t *used)
+dwc_tx_status(struct dma_chan *chan,
+	      dma_cookie_t cookie,
+	      struct dma_tx_state *txstate)
 {
 	struct dw_dma_chan	*dwc = to_dw_dma_chan(chan);
 	dma_cookie_t		last_used;
@@ -835,10 +842,7 @@ dwc_is_tx_complete(struct dma_chan *chan,
 		ret = dma_async_is_complete(cookie, last_complete, last_used);
 	}
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	return ret;
 }
@@ -1338,9 +1342,9 @@ static int __init dw_probe(struct platform_device *pdev)
 	dw->dma.device_prep_dma_memcpy = dwc_prep_dma_memcpy;
 
 	dw->dma.device_prep_slave_sg = dwc_prep_slave_sg;
-	dw->dma.device_terminate_all = dwc_terminate_all;
+	dw->dma.device_control = dwc_control;
 
-	dw->dma.device_is_tx_complete = dwc_is_tx_complete;
+	dw->dma.device_tx_status = dwc_tx_status;
 	dw->dma.device_issue_pending = dwc_issue_pending;
 
 	dma_writel(dw, CFG, DW_CFG_DMA_EN);
diff --git a/drivers/dma/fsldma.c b/drivers/dma/fsldma.c
index 88f470f0d82..1fdf180cbd6 100644
--- a/drivers/dma/fsldma.c
+++ b/drivers/dma/fsldma.c
@@ -775,13 +775,18 @@ fail:
 	return NULL;
 }
 
-static void fsl_dma_device_terminate_all(struct dma_chan *dchan)
+static int fsl_dma_device_control(struct dma_chan *dchan,
+				  enum dma_ctrl_cmd cmd, unsigned long arg)
 {
 	struct fsldma_chan *chan;
 	unsigned long flags;
 
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
 	if (!dchan)
-		return;
+		return -EINVAL;
 
 	chan = to_fsl_chan(dchan);
 
@@ -795,6 +800,8 @@ static void fsl_dma_device_terminate_all(struct dma_chan *dchan)
 	fsldma_free_desc_list(chan, &chan->ld_running);
 
 	spin_unlock_irqrestore(&chan->desc_lock, flags);
+
+	return 0;
 }
 
 /**
@@ -965,13 +972,12 @@ static void fsl_dma_memcpy_issue_pending(struct dma_chan *dchan)
 }
 
 /**
- * fsl_dma_is_complete - Determine the DMA status
+ * fsl_tx_status - Determine the DMA status
  * @chan : Freescale DMA channel
  */
-static enum dma_status fsl_dma_is_complete(struct dma_chan *dchan,
+static enum dma_status fsl_tx_status(struct dma_chan *dchan,
 					dma_cookie_t cookie,
-					dma_cookie_t *done,
-					dma_cookie_t *used)
+					struct dma_tx_state *txstate)
 {
 	struct fsldma_chan *chan = to_fsl_chan(dchan);
 	dma_cookie_t last_used;
@@ -982,11 +988,7 @@ static enum dma_status fsl_dma_is_complete(struct dma_chan *dchan,
 	last_used = dchan->cookie;
 	last_complete = chan->completed_cookie;
 
-	if (done)
-		*done = last_complete;
-
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	return dma_async_is_complete(cookie, last_complete, last_used);
 }
@@ -1330,10 +1332,10 @@ static int __devinit fsldma_of_probe(struct of_device *op,
 	fdev->common.device_free_chan_resources = fsl_dma_free_chan_resources;
 	fdev->common.device_prep_dma_interrupt = fsl_dma_prep_interrupt;
 	fdev->common.device_prep_dma_memcpy = fsl_dma_prep_memcpy;
-	fdev->common.device_is_tx_complete = fsl_dma_is_complete;
+	fdev->common.device_tx_status = fsl_tx_status;
 	fdev->common.device_issue_pending = fsl_dma_memcpy_issue_pending;
 	fdev->common.device_prep_slave_sg = fsl_dma_prep_slave_sg;
-	fdev->common.device_terminate_all = fsl_dma_device_terminate_all;
+	fdev->common.device_control = fsl_dma_device_control;
 	fdev->common.dev = &op->dev;
 
 	dev_set_drvdata(&op->dev, fdev);
diff --git a/drivers/dma/ioat/dma.c b/drivers/dma/ioat/dma.c
index 3e5a8005c62..c9213ead4a2 100644
--- a/drivers/dma/ioat/dma.c
+++ b/drivers/dma/ioat/dma.c
@@ -727,18 +727,18 @@ static void ioat1_timer_event(unsigned long data)
 }
 
 enum dma_status
-ioat_is_dma_complete(struct dma_chan *c, dma_cookie_t cookie,
-		      dma_cookie_t *done, dma_cookie_t *used)
+ioat_dma_tx_status(struct dma_chan *c, dma_cookie_t cookie,
+		   struct dma_tx_state *txstate)
 {
 	struct ioat_chan_common *chan = to_chan_common(c);
 	struct ioatdma_device *device = chan->device;
 
-	if (ioat_is_complete(c, cookie, done, used) == DMA_SUCCESS)
+	if (ioat_tx_status(c, cookie, txstate) == DMA_SUCCESS)
 		return DMA_SUCCESS;
 
 	device->cleanup_fn((unsigned long) c);
 
-	return ioat_is_complete(c, cookie, done, used);
+	return ioat_tx_status(c, cookie, txstate);
 }
 
 static void ioat1_dma_start_null_desc(struct ioat_dma_chan *ioat)
@@ -858,7 +858,7 @@ int __devinit ioat_dma_self_test(struct ioatdma_device *device)
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
 	if (tmo == 0 ||
-	    dma->device_is_tx_complete(dma_chan, cookie, NULL, NULL)
+	    dma->device_tx_status(dma_chan, cookie, NULL)
 					!= DMA_SUCCESS) {
 		dev_err(dev, "Self-test copy timed out, disabling\n");
 		err = -ENODEV;
@@ -1199,7 +1199,7 @@ int __devinit ioat1_dma_probe(struct ioatdma_device *device, int dca)
 	dma->device_issue_pending = ioat1_dma_memcpy_issue_pending;
 	dma->device_alloc_chan_resources = ioat1_dma_alloc_chan_resources;
 	dma->device_free_chan_resources = ioat1_dma_free_chan_resources;
-	dma->device_is_tx_complete = ioat_is_dma_complete;
+	dma->device_tx_status = ioat_dma_tx_status;
 
 	err = ioat_probe(device);
 	if (err)
diff --git a/drivers/dma/ioat/dma.h b/drivers/dma/ioat/dma.h
index 86b97ac8774..6d3a73b57e5 100644
--- a/drivers/dma/ioat/dma.h
+++ b/drivers/dma/ioat/dma.h
@@ -96,6 +96,7 @@ struct ioat_chan_common {
 	#define IOAT_COMPLETION_ACK 1
 	#define IOAT_RESET_PENDING 2
 	#define IOAT_KOBJ_INIT_FAIL 3
+	#define IOAT_RESHAPE_PENDING 4
 	struct timer_list timer;
 	#define COMPLETION_TIMEOUT msecs_to_jiffies(100)
 	#define IDLE_TIMEOUT msecs_to_jiffies(2000)
@@ -142,15 +143,14 @@ static inline struct ioat_dma_chan *to_ioat_chan(struct dma_chan *c)
 }
 
 /**
- * ioat_is_complete - poll the status of an ioat transaction
+ * ioat_tx_status - poll the status of an ioat transaction
  * @c: channel handle
  * @cookie: transaction identifier
- * @done: if set, updated with last completed transaction
- * @used: if set, updated with last used transaction
+ * @txstate: if set, updated with the transaction state
  */
 static inline enum dma_status
-ioat_is_complete(struct dma_chan *c, dma_cookie_t cookie,
-		 dma_cookie_t *done, dma_cookie_t *used)
+ioat_tx_status(struct dma_chan *c, dma_cookie_t cookie,
+		 struct dma_tx_state *txstate)
 {
 	struct ioat_chan_common *chan = to_chan_common(c);
 	dma_cookie_t last_used;
@@ -159,10 +159,7 @@ ioat_is_complete(struct dma_chan *c, dma_cookie_t cookie,
 	last_used = c->cookie;
 	last_complete = chan->completed_cookie;
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	return dma_async_is_complete(cookie, last_complete, last_used);
 }
@@ -338,8 +335,8 @@ struct dca_provider * __devinit ioat_dca_init(struct pci_dev *pdev,
 unsigned long ioat_get_current_completion(struct ioat_chan_common *chan);
 void ioat_init_channel(struct ioatdma_device *device,
 		       struct ioat_chan_common *chan, int idx);
-enum dma_status ioat_is_dma_complete(struct dma_chan *c, dma_cookie_t cookie,
-				     dma_cookie_t *done, dma_cookie_t *used);
+enum dma_status ioat_dma_tx_status(struct dma_chan *c, dma_cookie_t cookie,
+				   struct dma_tx_state *txstate);
 void ioat_dma_unmap(struct ioat_chan_common *chan, enum dma_ctrl_flags flags,
 		    size_t len, struct ioat_dma_descriptor *hw);
 bool ioat_cleanup_preamble(struct ioat_chan_common *chan,
diff --git a/drivers/dma/ioat/dma_v2.c b/drivers/dma/ioat/dma_v2.c
index b5ae56c211e..3c8b32a8379 100644
--- a/drivers/dma/ioat/dma_v2.c
+++ b/drivers/dma/ioat/dma_v2.c
@@ -56,8 +56,6 @@ void __ioat2_issue_pending(struct ioat2_dma_chan *ioat)
 
 	ioat->dmacount += ioat2_ring_pending(ioat);
 	ioat->issued = ioat->head;
-	/* make descriptor updates globally visible before notifying channel */
-	wmb();
 	writew(ioat->dmacount, chan->reg_base + IOAT_CHAN_DMACOUNT_OFFSET);
 	dev_dbg(to_dev(chan),
 		"%s: head: %#x tail: %#x issued: %#x count: %#x\n",
@@ -69,9 +67,9 @@ void ioat2_issue_pending(struct dma_chan *c)
 	struct ioat2_dma_chan *ioat = to_ioat2_chan(c);
 
 	if (ioat2_ring_pending(ioat)) {
-		spin_lock_bh(&ioat->ring_lock);
+		spin_lock_bh(&ioat->prep_lock);
 		__ioat2_issue_pending(ioat);
-		spin_unlock_bh(&ioat->ring_lock);
+		spin_unlock_bh(&ioat->prep_lock);
 	}
 }
 
@@ -80,7 +78,7 @@ void ioat2_issue_pending(struct dma_chan *c)
  * @ioat: ioat2+ channel
  *
  * Check if the number of unsubmitted descriptors has exceeded the
- * watermark.  Called with ring_lock held
+ * watermark.  Called with prep_lock held
  */
 static void ioat2_update_pending(struct ioat2_dma_chan *ioat)
 {
@@ -92,7 +90,6 @@ static void __ioat2_start_null_desc(struct ioat2_dma_chan *ioat)
 {
 	struct ioat_ring_ent *desc;
 	struct ioat_dma_descriptor *hw;
-	int idx;
 
 	if (ioat2_ring_space(ioat) < 1) {
 		dev_err(to_dev(&ioat->base),
@@ -102,8 +99,7 @@ static void __ioat2_start_null_desc(struct ioat2_dma_chan *ioat)
 
 	dev_dbg(to_dev(&ioat->base), "%s: head: %#x tail: %#x issued: %#x\n",
 		__func__, ioat->head, ioat->tail, ioat->issued);
-	idx = ioat2_desc_alloc(ioat, 1);
-	desc = ioat2_get_ring_ent(ioat, idx);
+	desc = ioat2_get_ring_ent(ioat, ioat->head);
 
 	hw = desc->hw;
 	hw->ctl = 0;
@@ -117,14 +113,16 @@ static void __ioat2_start_null_desc(struct ioat2_dma_chan *ioat)
 	async_tx_ack(&desc->txd);
 	ioat2_set_chainaddr(ioat, desc->txd.phys);
 	dump_desc_dbg(ioat, desc);
+	wmb();
+	ioat->head += 1;
 	__ioat2_issue_pending(ioat);
 }
 
 static void ioat2_start_null_desc(struct ioat2_dma_chan *ioat)
 {
-	spin_lock_bh(&ioat->ring_lock);
+	spin_lock_bh(&ioat->prep_lock);
 	__ioat2_start_null_desc(ioat);
-	spin_unlock_bh(&ioat->ring_lock);
+	spin_unlock_bh(&ioat->prep_lock);
 }
 
 static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
@@ -134,15 +132,16 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 	struct ioat_ring_ent *desc;
 	bool seen_current = false;
 	u16 active;
-	int i;
+	int idx = ioat->tail, i;
 
 	dev_dbg(to_dev(chan), "%s: head: %#x tail: %#x issued: %#x\n",
 		__func__, ioat->head, ioat->tail, ioat->issued);
 
 	active = ioat2_ring_active(ioat);
 	for (i = 0; i < active && !seen_current; i++) {
-		prefetch(ioat2_get_ring_ent(ioat, ioat->tail + i + 1));
-		desc = ioat2_get_ring_ent(ioat, ioat->tail + i);
+		smp_read_barrier_depends();
+		prefetch(ioat2_get_ring_ent(ioat, idx + i + 1));
+		desc = ioat2_get_ring_ent(ioat, idx + i);
 		tx = &desc->txd;
 		dump_desc_dbg(ioat, desc);
 		if (tx->cookie) {
@@ -158,11 +157,12 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 		if (tx->phys == phys_complete)
 			seen_current = true;
 	}
-	ioat->tail += i;
+	smp_mb(); /* finish all descriptor reads before incrementing tail */
+	ioat->tail = idx + i;
 	BUG_ON(active && !seen_current); /* no active descs have written a completion? */
 
 	chan->last_completion = phys_complete;
-	if (ioat->head == ioat->tail) {
+	if (active - i == 0) {
 		dev_dbg(to_dev(chan), "%s: cancel completion timeout\n",
 			__func__);
 		clear_bit(IOAT_COMPLETION_PENDING, &chan->state);
@@ -179,24 +179,9 @@ static void ioat2_cleanup(struct ioat2_dma_chan *ioat)
 	struct ioat_chan_common *chan = &ioat->base;
 	unsigned long phys_complete;
 
-	prefetch(chan->completion);
-
-	if (!spin_trylock_bh(&chan->cleanup_lock))
-		return;
-
-	if (!ioat_cleanup_preamble(chan, &phys_complete)) {
-		spin_unlock_bh(&chan->cleanup_lock);
-		return;
-	}
-
-	if (!spin_trylock_bh(&ioat->ring_lock)) {
-		spin_unlock_bh(&chan->cleanup_lock);
-		return;
-	}
-
-	__cleanup(ioat, phys_complete);
-
-	spin_unlock_bh(&ioat->ring_lock);
+	spin_lock_bh(&chan->cleanup_lock);
+	if (ioat_cleanup_preamble(chan, &phys_complete))
+		__cleanup(ioat, phys_complete);
 	spin_unlock_bh(&chan->cleanup_lock);
 }
 
@@ -287,12 +272,10 @@ void ioat2_timer_event(unsigned long data)
 	struct ioat2_dma_chan *ioat = to_ioat2_chan((void *) data);
 	struct ioat_chan_common *chan = &ioat->base;
 
-	spin_lock_bh(&chan->cleanup_lock);
 	if (test_bit(IOAT_COMPLETION_PENDING, &chan->state)) {
 		unsigned long phys_complete;
 		u64 status;
 
-		spin_lock_bh(&ioat->ring_lock);
 		status = ioat_chansts(chan);
 
 		/* when halted due to errors check for channel
@@ -311,26 +294,31 @@ void ioat2_timer_event(unsigned long data)
 		 * acknowledged a pending completion once, then be more
 		 * forceful with a restart
 		 */
-		if (ioat_cleanup_preamble(chan, &phys_complete))
+		spin_lock_bh(&chan->cleanup_lock);
+		if (ioat_cleanup_preamble(chan, &phys_complete)) {
 			__cleanup(ioat, phys_complete);
-		else if (test_bit(IOAT_COMPLETION_ACK, &chan->state))
+		} else if (test_bit(IOAT_COMPLETION_ACK, &chan->state)) {
+			spin_lock_bh(&ioat->prep_lock);
 			ioat2_restart_channel(ioat);
-		else {
+			spin_unlock_bh(&ioat->prep_lock);
+		} else {
 			set_bit(IOAT_COMPLETION_ACK, &chan->state);
 			mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
 		}
-		spin_unlock_bh(&ioat->ring_lock);
+		spin_unlock_bh(&chan->cleanup_lock);
 	} else {
 		u16 active;
 
 		/* if the ring is idle, empty, and oversized try to step
 		 * down the size
 		 */
-		spin_lock_bh(&ioat->ring_lock);
+		spin_lock_bh(&chan->cleanup_lock);
+		spin_lock_bh(&ioat->prep_lock);
 		active = ioat2_ring_active(ioat);
 		if (active == 0 && ioat->alloc_order > ioat_get_alloc_order())
 			reshape_ring(ioat, ioat->alloc_order-1);
-		spin_unlock_bh(&ioat->ring_lock);
+		spin_unlock_bh(&ioat->prep_lock);
+		spin_unlock_bh(&chan->cleanup_lock);
 
 		/* keep shrinking until we get back to our minimum
 		 * default size
@@ -338,7 +326,6 @@ void ioat2_timer_event(unsigned long data)
 		if (ioat->alloc_order > ioat_get_alloc_order())
 			mod_timer(&chan->timer, jiffies + IDLE_TIMEOUT);
 	}
-	spin_unlock_bh(&chan->cleanup_lock);
 }
 
 static int ioat2_reset_hw(struct ioat_chan_common *chan)
@@ -392,7 +379,7 @@ int ioat2_enumerate_channels(struct ioatdma_device *device)
 
 		ioat_init_channel(device, &ioat->base, i);
 		ioat->xfercap_log = xfercap_log;
-		spin_lock_init(&ioat->ring_lock);
+		spin_lock_init(&ioat->prep_lock);
 		if (device->reset_hw(&ioat->base)) {
 			i = 0;
 			break;
@@ -418,8 +405,17 @@ static dma_cookie_t ioat2_tx_submit_unlock(struct dma_async_tx_descriptor *tx)
 
 	if (!test_and_set_bit(IOAT_COMPLETION_PENDING, &chan->state))
 		mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
+
+	/* make descriptor updates visible before advancing ioat->head,
+	 * this is purposefully not smp_wmb() since we are also
+	 * publishing the descriptor updates to a dma device
+	 */
+	wmb();
+
+	ioat->head += ioat->produce;
+
 	ioat2_update_pending(ioat);
-	spin_unlock_bh(&ioat->ring_lock);
+	spin_unlock_bh(&ioat->prep_lock);
 
 	return cookie;
 }
@@ -531,13 +527,15 @@ int ioat2_alloc_chan_resources(struct dma_chan *c)
 	if (!ring)
 		return -ENOMEM;
 
-	spin_lock_bh(&ioat->ring_lock);
+	spin_lock_bh(&chan->cleanup_lock);
+	spin_lock_bh(&ioat->prep_lock);
 	ioat->ring = ring;
 	ioat->head = 0;
 	ioat->issued = 0;
 	ioat->tail = 0;
 	ioat->alloc_order = order;
-	spin_unlock_bh(&ioat->ring_lock);
+	spin_unlock_bh(&ioat->prep_lock);
+	spin_unlock_bh(&chan->cleanup_lock);
 
 	tasklet_enable(&chan->cleanup_task);
 	ioat2_start_null_desc(ioat);
@@ -553,7 +551,7 @@ bool reshape_ring(struct ioat2_dma_chan *ioat, int order)
 	 */
 	struct ioat_chan_common *chan = &ioat->base;
 	struct dma_chan *c = &chan->common;
-	const u16 curr_size = ioat2_ring_mask(ioat) + 1;
+	const u16 curr_size = ioat2_ring_size(ioat);
 	const u16 active = ioat2_ring_active(ioat);
 	const u16 new_size = 1 << order;
 	struct ioat_ring_ent **ring;
@@ -653,54 +651,61 @@ bool reshape_ring(struct ioat2_dma_chan *ioat, int order)
 }
 
 /**
- * ioat2_alloc_and_lock - common descriptor alloc boilerplate for ioat2,3 ops
- * @idx: gets starting descriptor index on successful allocation
+ * ioat2_check_space_lock - verify space and grab ring producer lock
  * @ioat: ioat2,3 channel (ring) to operate on
  * @num_descs: allocation length
  */
-int ioat2_alloc_and_lock(u16 *idx, struct ioat2_dma_chan *ioat, int num_descs)
+int ioat2_check_space_lock(struct ioat2_dma_chan *ioat, int num_descs)
 {
 	struct ioat_chan_common *chan = &ioat->base;
+	bool retry;
 
-	spin_lock_bh(&ioat->ring_lock);
+ retry:
+	spin_lock_bh(&ioat->prep_lock);
 	/* never allow the last descriptor to be consumed, we need at
 	 * least one free at all times to allow for on-the-fly ring
 	 * resizing.
 	 */
-	while (unlikely(ioat2_ring_space(ioat) <= num_descs)) {
-		if (reshape_ring(ioat, ioat->alloc_order + 1) &&
-		    ioat2_ring_space(ioat) > num_descs)
-				break;
-
-		if (printk_ratelimit())
-			dev_dbg(to_dev(chan),
-				"%s: ring full! num_descs: %d (%x:%x:%x)\n",
-				__func__, num_descs, ioat->head, ioat->tail,
-				ioat->issued);
-		spin_unlock_bh(&ioat->ring_lock);
-
-		/* progress reclaim in the allocation failure case we
-		 * may be called under bh_disabled so we need to trigger
-		 * the timer event directly
-		 */
-		spin_lock_bh(&chan->cleanup_lock);
-		if (jiffies > chan->timer.expires &&
-		    timer_pending(&chan->timer)) {
-			struct ioatdma_device *device = chan->device;
-
-			mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
-			spin_unlock_bh(&chan->cleanup_lock);
-			device->timer_fn((unsigned long) &chan->common);
-		} else
-			spin_unlock_bh(&chan->cleanup_lock);
-		return -ENOMEM;
+	if (likely(ioat2_ring_space(ioat) > num_descs)) {
+		dev_dbg(to_dev(chan), "%s: num_descs: %d (%x:%x:%x)\n",
+			__func__, num_descs, ioat->head, ioat->tail, ioat->issued);
+		ioat->produce = num_descs;
+		return 0;  /* with ioat->prep_lock held */
 	}
+	retry = test_and_set_bit(IOAT_RESHAPE_PENDING, &chan->state);
+	spin_unlock_bh(&ioat->prep_lock);
 
-	dev_dbg(to_dev(chan), "%s: num_descs: %d (%x:%x:%x)\n",
-		__func__, num_descs, ioat->head, ioat->tail, ioat->issued);
+	/* is another cpu already trying to expand the ring? */
+	if (retry)
+		goto retry;
 
-	*idx = ioat2_desc_alloc(ioat, num_descs);
-	return 0;  /* with ioat->ring_lock held */
+	spin_lock_bh(&chan->cleanup_lock);
+	spin_lock_bh(&ioat->prep_lock);
+	retry = reshape_ring(ioat, ioat->alloc_order + 1);
+	clear_bit(IOAT_RESHAPE_PENDING, &chan->state);
+	spin_unlock_bh(&ioat->prep_lock);
+	spin_unlock_bh(&chan->cleanup_lock);
+
+	/* if we were able to expand the ring retry the allocation */
+	if (retry)
+		goto retry;
+
+	if (printk_ratelimit())
+		dev_dbg(to_dev(chan), "%s: ring full! num_descs: %d (%x:%x:%x)\n",
+			__func__, num_descs, ioat->head, ioat->tail, ioat->issued);
+
+	/* progress reclaim in the allocation failure case we may be
+	 * called under bh_disabled so we need to trigger the timer
+	 * event directly
+	 */
+	if (jiffies > chan->timer.expires && timer_pending(&chan->timer)) {
+		struct ioatdma_device *device = chan->device;
+
+		mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
+		device->timer_fn((unsigned long) &chan->common);
+	}
+
+	return -ENOMEM;
 }
 
 struct dma_async_tx_descriptor *
@@ -713,14 +718,11 @@ ioat2_dma_prep_memcpy_lock(struct dma_chan *c, dma_addr_t dma_dest,
 	dma_addr_t dst = dma_dest;
 	dma_addr_t src = dma_src;
 	size_t total_len = len;
-	int num_descs;
-	u16 idx;
-	int i;
+	int num_descs, idx, i;
 
 	num_descs = ioat2_xferlen_to_descs(ioat, len);
-	if (likely(num_descs) &&
-	    ioat2_alloc_and_lock(&idx, ioat, num_descs) == 0)
-		/* pass */;
+	if (likely(num_descs) && ioat2_check_space_lock(ioat, num_descs) == 0)
+		idx = ioat->head;
 	else
 		return NULL;
 	i = 0;
@@ -777,7 +779,8 @@ void ioat2_free_chan_resources(struct dma_chan *c)
 	device->cleanup_fn((unsigned long) c);
 	device->reset_hw(chan);
 
-	spin_lock_bh(&ioat->ring_lock);
+	spin_lock_bh(&chan->cleanup_lock);
+	spin_lock_bh(&ioat->prep_lock);
 	descs = ioat2_ring_space(ioat);
 	dev_dbg(to_dev(chan), "freeing %d idle descriptors\n", descs);
 	for (i = 0; i < descs; i++) {
@@ -800,7 +803,8 @@ void ioat2_free_chan_resources(struct dma_chan *c)
 	ioat->alloc_order = 0;
 	pci_pool_free(device->completion_pool, chan->completion,
 		      chan->completion_dma);
-	spin_unlock_bh(&ioat->ring_lock);
+	spin_unlock_bh(&ioat->prep_lock);
+	spin_unlock_bh(&chan->cleanup_lock);
 
 	chan->last_completion = 0;
 	chan->completion_dma = 0;
@@ -855,7 +859,7 @@ int __devinit ioat2_dma_probe(struct ioatdma_device *device, int dca)
 	dma->device_issue_pending = ioat2_issue_pending;
 	dma->device_alloc_chan_resources = ioat2_alloc_chan_resources;
 	dma->device_free_chan_resources = ioat2_free_chan_resources;
-	dma->device_is_tx_complete = ioat_is_dma_complete;
+	dma->device_tx_status = ioat_tx_status;
 
 	err = ioat_probe(device);
 	if (err)
diff --git a/drivers/dma/ioat/dma_v2.h b/drivers/dma/ioat/dma_v2.h
index ef2871fd786..a2c413b2b8d 100644
--- a/drivers/dma/ioat/dma_v2.h
+++ b/drivers/dma/ioat/dma_v2.h
@@ -22,6 +22,7 @@
 #define IOATDMA_V2_H
 
 #include <linux/dmaengine.h>
+#include <linux/circ_buf.h>
 #include "dma.h"
 #include "hw.h"
 
@@ -49,8 +50,9 @@ extern int ioat_ring_alloc_order;
  * @tail: cleanup index
  * @dmacount: identical to 'head' except for occasionally resetting to zero
  * @alloc_order: log2 of the number of allocated descriptors
+ * @produce: number of descriptors to produce at submit time
  * @ring: software ring buffer implementation of hardware ring
- * @ring_lock: protects ring attributes
+ * @prep_lock: serializes descriptor preparation (producers)
  */
 struct ioat2_dma_chan {
 	struct ioat_chan_common base;
@@ -60,8 +62,9 @@ struct ioat2_dma_chan {
 	u16 tail;
 	u16 dmacount;
 	u16 alloc_order;
+	u16 produce;
 	struct ioat_ring_ent **ring;
-	spinlock_t ring_lock;
+	spinlock_t prep_lock;
 };
 
 static inline struct ioat2_dma_chan *to_ioat2_chan(struct dma_chan *c)
@@ -71,38 +74,26 @@ static inline struct ioat2_dma_chan *to_ioat2_chan(struct dma_chan *c)
 	return container_of(chan, struct ioat2_dma_chan, base);
 }
 
-static inline u16 ioat2_ring_mask(struct ioat2_dma_chan *ioat)
+static inline u16 ioat2_ring_size(struct ioat2_dma_chan *ioat)
 {
-	return (1 << ioat->alloc_order) - 1;
+	return 1 << ioat->alloc_order;
 }
 
 /* count of descriptors in flight with the engine */
 static inline u16 ioat2_ring_active(struct ioat2_dma_chan *ioat)
 {
-	return (ioat->head - ioat->tail) & ioat2_ring_mask(ioat);
+	return CIRC_CNT(ioat->head, ioat->tail, ioat2_ring_size(ioat));
 }
 
 /* count of descriptors pending submission to hardware */
 static inline u16 ioat2_ring_pending(struct ioat2_dma_chan *ioat)
 {
-	return (ioat->head - ioat->issued) & ioat2_ring_mask(ioat);
+	return CIRC_CNT(ioat->head, ioat->issued, ioat2_ring_size(ioat));
 }
 
 static inline u16 ioat2_ring_space(struct ioat2_dma_chan *ioat)
 {
-	u16 num_descs = ioat2_ring_mask(ioat) + 1;
-	u16 active = ioat2_ring_active(ioat);
-
-	BUG_ON(active > num_descs);
-
-	return num_descs - active;
-}
-
-/* assumes caller already checked space */
-static inline u16 ioat2_desc_alloc(struct ioat2_dma_chan *ioat, u16 len)
-{
-	ioat->head += len;
-	return ioat->head - len;
+	return ioat2_ring_size(ioat) - ioat2_ring_active(ioat);
 }
 
 static inline u16 ioat2_xferlen_to_descs(struct ioat2_dma_chan *ioat, size_t len)
@@ -151,7 +142,7 @@ struct ioat_ring_ent {
 static inline struct ioat_ring_ent *
 ioat2_get_ring_ent(struct ioat2_dma_chan *ioat, u16 idx)
 {
-	return ioat->ring[idx & ioat2_ring_mask(ioat)];
+	return ioat->ring[idx & (ioat2_ring_size(ioat) - 1)];
 }
 
 static inline void ioat2_set_chainaddr(struct ioat2_dma_chan *ioat, u64 addr)
@@ -168,7 +159,7 @@ int __devinit ioat2_dma_probe(struct ioatdma_device *dev, int dca);
 int __devinit ioat3_dma_probe(struct ioatdma_device *dev, int dca);
 struct dca_provider * __devinit ioat2_dca_init(struct pci_dev *pdev, void __iomem *iobase);
 struct dca_provider * __devinit ioat3_dca_init(struct pci_dev *pdev, void __iomem *iobase);
-int ioat2_alloc_and_lock(u16 *idx, struct ioat2_dma_chan *ioat, int num_descs);
+int ioat2_check_space_lock(struct ioat2_dma_chan *ioat, int num_descs);
 int ioat2_enumerate_channels(struct ioatdma_device *device);
 struct dma_async_tx_descriptor *
 ioat2_dma_prep_memcpy_lock(struct dma_chan *c, dma_addr_t dma_dest,
diff --git a/drivers/dma/ioat/dma_v3.c b/drivers/dma/ioat/dma_v3.c
index 6740e319c9c..1cdd22e1051 100644
--- a/drivers/dma/ioat/dma_v3.c
+++ b/drivers/dma/ioat/dma_v3.c
@@ -260,8 +260,8 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 	struct ioat_chan_common *chan = &ioat->base;
 	struct ioat_ring_ent *desc;
 	bool seen_current = false;
+	int idx = ioat->tail, i;
 	u16 active;
-	int i;
 
 	dev_dbg(to_dev(chan), "%s: head: %#x tail: %#x issued: %#x\n",
 		__func__, ioat->head, ioat->tail, ioat->issued);
@@ -270,13 +270,14 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 	for (i = 0; i < active && !seen_current; i++) {
 		struct dma_async_tx_descriptor *tx;
 
-		prefetch(ioat2_get_ring_ent(ioat, ioat->tail + i + 1));
-		desc = ioat2_get_ring_ent(ioat, ioat->tail + i);
+		smp_read_barrier_depends();
+		prefetch(ioat2_get_ring_ent(ioat, idx + i + 1));
+		desc = ioat2_get_ring_ent(ioat, idx + i);
 		dump_desc_dbg(ioat, desc);
 		tx = &desc->txd;
 		if (tx->cookie) {
 			chan->completed_cookie = tx->cookie;
-			ioat3_dma_unmap(ioat, desc, ioat->tail + i);
+			ioat3_dma_unmap(ioat, desc, idx + i);
 			tx->cookie = 0;
 			if (tx->callback) {
 				tx->callback(tx->callback_param);
@@ -293,69 +294,30 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 			i++;
 		}
 	}
-	ioat->tail += i;
+	smp_mb(); /* finish all descriptor reads before incrementing tail */
+	ioat->tail = idx + i;
 	BUG_ON(active && !seen_current); /* no active descs have written a completion? */
 	chan->last_completion = phys_complete;
 
-	active = ioat2_ring_active(ioat);
-	if (active == 0) {
+	if (active - i == 0) {
 		dev_dbg(to_dev(chan), "%s: cancel completion timeout\n",
 			__func__);
 		clear_bit(IOAT_COMPLETION_PENDING, &chan->state);
 		mod_timer(&chan->timer, jiffies + IDLE_TIMEOUT);
 	}
 	/* 5 microsecond delay per pending descriptor */
-	writew(min((5 * active), IOAT_INTRDELAY_MASK),
+	writew(min((5 * (active - i)), IOAT_INTRDELAY_MASK),
 	       chan->device->reg_base + IOAT_INTRDELAY_OFFSET);
 }
 
-/* try to cleanup, but yield (via spin_trylock) to incoming submissions
- * with the expectation that we will immediately poll again shortly
- */
-static void ioat3_cleanup_poll(struct ioat2_dma_chan *ioat)
+static void ioat3_cleanup(struct ioat2_dma_chan *ioat)
 {
 	struct ioat_chan_common *chan = &ioat->base;
 	unsigned long phys_complete;
 
-	prefetch(chan->completion);
-
-	if (!spin_trylock_bh(&chan->cleanup_lock))
-		return;
-
-	if (!ioat_cleanup_preamble(chan, &phys_complete)) {
-		spin_unlock_bh(&chan->cleanup_lock);
-		return;
-	}
-
-	if (!spin_trylock_bh(&ioat->ring_lock)) {
-		spin_unlock_bh(&chan->cleanup_lock);
-		return;
-	}
-
-	__cleanup(ioat, phys_complete);
-
-	spin_unlock_bh(&ioat->ring_lock);
-	spin_unlock_bh(&chan->cleanup_lock);
-}
-
-/* run cleanup now because we already delayed the interrupt via INTRDELAY */
-static void ioat3_cleanup_sync(struct ioat2_dma_chan *ioat)
-{
-	struct ioat_chan_common *chan = &ioat->base;
-	unsigned long phys_complete;
-
-	prefetch(chan->completion);
-
 	spin_lock_bh(&chan->cleanup_lock);
-	if (!ioat_cleanup_preamble(chan, &phys_complete)) {
-		spin_unlock_bh(&chan->cleanup_lock);
-		return;
-	}
-	spin_lock_bh(&ioat->ring_lock);
-
-	__cleanup(ioat, phys_complete);
-
-	spin_unlock_bh(&ioat->ring_lock);
+	if (ioat_cleanup_preamble(chan, &phys_complete))
+		__cleanup(ioat, phys_complete);
 	spin_unlock_bh(&chan->cleanup_lock);
 }
 
@@ -363,7 +325,7 @@ static void ioat3_cleanup_event(unsigned long data)
 {
 	struct ioat2_dma_chan *ioat = to_ioat2_chan((void *) data);
 
-	ioat3_cleanup_sync(ioat);
+	ioat3_cleanup(ioat);
 	writew(IOAT_CHANCTRL_RUN, ioat->base.reg_base + IOAT_CHANCTRL_OFFSET);
 }
 
@@ -384,12 +346,10 @@ static void ioat3_timer_event(unsigned long data)
 	struct ioat2_dma_chan *ioat = to_ioat2_chan((void *) data);
 	struct ioat_chan_common *chan = &ioat->base;
 
-	spin_lock_bh(&chan->cleanup_lock);
 	if (test_bit(IOAT_COMPLETION_PENDING, &chan->state)) {
 		unsigned long phys_complete;
 		u64 status;
 
-		spin_lock_bh(&ioat->ring_lock);
 		status = ioat_chansts(chan);
 
 		/* when halted due to errors check for channel
@@ -408,26 +368,31 @@ static void ioat3_timer_event(unsigned long data)
 		 * acknowledged a pending completion once, then be more
 		 * forceful with a restart
 		 */
+		spin_lock_bh(&chan->cleanup_lock);
 		if (ioat_cleanup_preamble(chan, &phys_complete))
 			__cleanup(ioat, phys_complete);
-		else if (test_bit(IOAT_COMPLETION_ACK, &chan->state))
+		else if (test_bit(IOAT_COMPLETION_ACK, &chan->state)) {
+			spin_lock_bh(&ioat->prep_lock);
 			ioat3_restart_channel(ioat);
-		else {
+			spin_unlock_bh(&ioat->prep_lock);
+		} else {
 			set_bit(IOAT_COMPLETION_ACK, &chan->state);
 			mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
 		}
-		spin_unlock_bh(&ioat->ring_lock);
+		spin_unlock_bh(&chan->cleanup_lock);
 	} else {
 		u16 active;
 
 		/* if the ring is idle, empty, and oversized try to step
 		 * down the size
 		 */
-		spin_lock_bh(&ioat->ring_lock);
+		spin_lock_bh(&chan->cleanup_lock);
+		spin_lock_bh(&ioat->prep_lock);
 		active = ioat2_ring_active(ioat);
 		if (active == 0 && ioat->alloc_order > ioat_get_alloc_order())
 			reshape_ring(ioat, ioat->alloc_order-1);
-		spin_unlock_bh(&ioat->ring_lock);
+		spin_unlock_bh(&ioat->prep_lock);
+		spin_unlock_bh(&chan->cleanup_lock);
 
 		/* keep shrinking until we get back to our minimum
 		 * default size
@@ -435,21 +400,20 @@ static void ioat3_timer_event(unsigned long data)
 		if (ioat->alloc_order > ioat_get_alloc_order())
 			mod_timer(&chan->timer, jiffies + IDLE_TIMEOUT);
 	}
-	spin_unlock_bh(&chan->cleanup_lock);
 }
 
 static enum dma_status
-ioat3_is_complete(struct dma_chan *c, dma_cookie_t cookie,
-		  dma_cookie_t *done, dma_cookie_t *used)
+ioat3_tx_status(struct dma_chan *c, dma_cookie_t cookie,
+		struct dma_tx_state *txstate)
 {
 	struct ioat2_dma_chan *ioat = to_ioat2_chan(c);
 
-	if (ioat_is_complete(c, cookie, done, used) == DMA_SUCCESS)
+	if (ioat_tx_status(c, cookie, txstate) == DMA_SUCCESS)
 		return DMA_SUCCESS;
 
-	ioat3_cleanup_poll(ioat);
+	ioat3_cleanup(ioat);
 
-	return ioat_is_complete(c, cookie, done, used);
+	return ioat_tx_status(c, cookie, txstate);
 }
 
 static struct dma_async_tx_descriptor *
@@ -460,15 +424,12 @@ ioat3_prep_memset_lock(struct dma_chan *c, dma_addr_t dest, int value,
 	struct ioat_ring_ent *desc;
 	size_t total_len = len;
 	struct ioat_fill_descriptor *fill;
-	int num_descs;
 	u64 src_data = (0x0101010101010101ULL) * (value & 0xff);
-	u16 idx;
-	int i;
+	int num_descs, idx, i;
 
 	num_descs = ioat2_xferlen_to_descs(ioat, len);
-	if (likely(num_descs) &&
-	    ioat2_alloc_and_lock(&idx, ioat, num_descs) == 0)
-		/* pass */;
+	if (likely(num_descs) && ioat2_check_space_lock(ioat, num_descs) == 0)
+		idx = ioat->head;
 	else
 		return NULL;
 	i = 0;
@@ -513,11 +474,8 @@ __ioat3_prep_xor_lock(struct dma_chan *c, enum sum_check_flags *result,
 	struct ioat_xor_descriptor *xor;
 	struct ioat_xor_ext_descriptor *xor_ex = NULL;
 	struct ioat_dma_descriptor *hw;
+	int num_descs, with_ext, idx, i;
 	u32 offset = 0;
-	int num_descs;
-	int with_ext;
-	int i;
-	u16 idx;
 	u8 op = result ? IOAT_OP_XOR_VAL : IOAT_OP_XOR;
 
 	BUG_ON(src_cnt < 2);
@@ -537,9 +495,8 @@ __ioat3_prep_xor_lock(struct dma_chan *c, enum sum_check_flags *result,
 	 * (legacy) descriptor to ensure all completion writes arrive in
 	 * order.
 	 */
-	if (likely(num_descs) &&
-	    ioat2_alloc_and_lock(&idx, ioat, num_descs+1) == 0)
-		/* pass */;
+	if (likely(num_descs) && ioat2_check_space_lock(ioat, num_descs+1) == 0)
+		idx = ioat->head;
 	else
 		return NULL;
 	i = 0;
@@ -657,11 +614,8 @@ __ioat3_prep_pq_lock(struct dma_chan *c, enum sum_check_flags *result,
 	struct ioat_pq_ext_descriptor *pq_ex = NULL;
 	struct ioat_dma_descriptor *hw;
 	u32 offset = 0;
-	int num_descs;
-	int with_ext;
-	int i, s;
-	u16 idx;
 	u8 op = result ? IOAT_OP_PQ_VAL : IOAT_OP_PQ;
+	int i, s, idx, with_ext, num_descs;
 
 	dev_dbg(to_dev(chan), "%s\n", __func__);
 	/* the engine requires at least two sources (we provide
@@ -687,8 +641,8 @@ __ioat3_prep_pq_lock(struct dma_chan *c, enum sum_check_flags *result,
 	 * order.
 	 */
 	if (likely(num_descs) &&
-	    ioat2_alloc_and_lock(&idx, ioat, num_descs+1) == 0)
-		/* pass */;
+	    ioat2_check_space_lock(ioat, num_descs+1) == 0)
+		idx = ioat->head;
 	else
 		return NULL;
 	i = 0;
@@ -851,10 +805,9 @@ ioat3_prep_interrupt_lock(struct dma_chan *c, unsigned long flags)
 	struct ioat2_dma_chan *ioat = to_ioat2_chan(c);
 	struct ioat_ring_ent *desc;
 	struct ioat_dma_descriptor *hw;
-	u16 idx;
 
-	if (ioat2_alloc_and_lock(&idx, ioat, 1) == 0)
-		desc = ioat2_get_ring_ent(ioat, idx);
+	if (ioat2_check_space_lock(ioat, 1) == 0)
+		desc = ioat2_get_ring_ent(ioat, ioat->head);
 	else
 		return NULL;
 
@@ -977,7 +930,7 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
-	if (dma->device_is_tx_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_err(dev, "Self-test xor timed out\n");
 		err = -ENODEV;
 		goto free_resources;
@@ -1031,7 +984,7 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
-	if (dma->device_is_tx_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_err(dev, "Self-test validate timed out\n");
 		err = -ENODEV;
 		goto free_resources;
@@ -1072,7 +1025,7 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
-	if (dma->device_is_tx_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_err(dev, "Self-test memset timed out\n");
 		err = -ENODEV;
 		goto free_resources;
@@ -1115,7 +1068,7 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
-	if (dma->device_is_tx_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_err(dev, "Self-test 2nd validate timed out\n");
 		err = -ENODEV;
 		goto free_resources;
@@ -1222,7 +1175,7 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 	if (cap & IOAT_CAP_XOR) {
 		is_raid_device = true;
 		dma->max_xor = 8;
-		dma->xor_align = 2;
+		dma->xor_align = 6;
 
 		dma_cap_set(DMA_XOR, dma->cap_mask);
 		dma->device_prep_dma_xor = ioat3_prep_xor;
@@ -1233,7 +1186,7 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 	if (cap & IOAT_CAP_PQ) {
 		is_raid_device = true;
 		dma_set_maxpq(dma, 8, 0);
-		dma->pq_align = 2;
+		dma->pq_align = 6;
 
 		dma_cap_set(DMA_PQ, dma->cap_mask);
 		dma->device_prep_dma_pq = ioat3_prep_pq;
@@ -1243,7 +1196,7 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 
 		if (!(cap & IOAT_CAP_XOR)) {
 			dma->max_xor = 8;
-			dma->xor_align = 2;
+			dma->xor_align = 6;
 
 			dma_cap_set(DMA_XOR, dma->cap_mask);
 			dma->device_prep_dma_xor = ioat3_prep_pqxor;
@@ -1259,11 +1212,11 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 
 
 	if (is_raid_device) {
-		dma->device_is_tx_complete = ioat3_is_complete;
+		dma->device_tx_status = ioat3_tx_status;
 		device->cleanup_fn = ioat3_cleanup_event;
 		device->timer_fn = ioat3_timer_event;
 	} else {
-		dma->device_is_tx_complete = ioat_is_dma_complete;
+		dma->device_tx_status = ioat_dma_tx_status;
 		device->cleanup_fn = ioat2_cleanup_event;
 		device->timer_fn = ioat2_timer_event;
 	}
diff --git a/drivers/dma/ioat/pci.c b/drivers/dma/ioat/pci.c
index 99ec26725ba..fab37d1cf48 100644
--- a/drivers/dma/ioat/pci.c
+++ b/drivers/dma/ioat/pci.c
@@ -138,15 +138,10 @@ static int __devinit ioat_pci_probe(struct pci_dev *pdev, const struct pci_devic
 	if (err)
 		return err;
 
-	device = devm_kzalloc(dev, sizeof(*device), GFP_KERNEL);
-	if (!device)
-		return -ENOMEM;
-
-	pci_set_master(pdev);
-
 	device = alloc_ioatdma(pdev, iomap[IOAT_MMIO_BAR]);
 	if (!device)
 		return -ENOMEM;
+	pci_set_master(pdev);
 	pci_set_drvdata(pdev, device);
 
 	device->version = readb(device->reg_base + IOAT_VER_OFFSET);
diff --git a/drivers/dma/iop-adma.c b/drivers/dma/iop-adma.c
index 1ebc801678b..161c452923b 100644
--- a/drivers/dma/iop-adma.c
+++ b/drivers/dma/iop-adma.c
@@ -894,14 +894,14 @@ static void iop_adma_free_chan_resources(struct dma_chan *chan)
 }
 
 /**
- * iop_adma_is_complete - poll the status of an ADMA transaction
+ * iop_adma_status - poll the status of an ADMA transaction
  * @chan: ADMA channel handle
  * @cookie: ADMA transaction identifier
+ * @txstate: a holder for the current state of the channel or NULL
  */
-static enum dma_status iop_adma_is_complete(struct dma_chan *chan,
+static enum dma_status iop_adma_status(struct dma_chan *chan,
 					dma_cookie_t cookie,
-					dma_cookie_t *done,
-					dma_cookie_t *used)
+					struct dma_tx_state *txstate)
 {
 	struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan);
 	dma_cookie_t last_used;
@@ -910,12 +910,7 @@ static enum dma_status iop_adma_is_complete(struct dma_chan *chan,
 
 	last_used = chan->cookie;
 	last_complete = iop_chan->completed_cookie;
-
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
-
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 	ret = dma_async_is_complete(cookie, last_complete, last_used);
 	if (ret == DMA_SUCCESS)
 		return ret;
@@ -924,11 +919,7 @@ static enum dma_status iop_adma_is_complete(struct dma_chan *chan,
 
 	last_used = chan->cookie;
 	last_complete = iop_chan->completed_cookie;
-
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	return dma_async_is_complete(cookie, last_complete, last_used);
 }
@@ -1043,7 +1034,7 @@ static int __devinit iop_adma_memcpy_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(1);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (iop_adma_status(dma_chan, cookie, NULL) !=
 			DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			"Self-test copy timed out, disabling\n");
@@ -1143,7 +1134,7 @@ iop_adma_xor_val_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (iop_adma_status(dma_chan, cookie, NULL) !=
 		DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			"Self-test xor timed out, disabling\n");
@@ -1190,7 +1181,7 @@ iop_adma_xor_val_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (iop_adma_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			"Self-test zero sum timed out, disabling\n");
 		err = -ENODEV;
@@ -1214,7 +1205,7 @@ iop_adma_xor_val_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (iop_adma_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			"Self-test memset timed out, disabling\n");
 		err = -ENODEV;
@@ -1246,7 +1237,7 @@ iop_adma_xor_val_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (iop_adma_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			"Self-test non-zero sum timed out, disabling\n");
 		err = -ENODEV;
@@ -1341,7 +1332,7 @@ iop_adma_pq_zero_sum_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (iop_adma_status(dma_chan, cookie, NULL) !=
 		DMA_SUCCESS) {
 		dev_err(dev, "Self-test pq timed out, disabling\n");
 		err = -ENODEV;
@@ -1378,7 +1369,7 @@ iop_adma_pq_zero_sum_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (iop_adma_status(dma_chan, cookie, NULL) !=
 		DMA_SUCCESS) {
 		dev_err(dev, "Self-test pq-zero-sum timed out, disabling\n");
 		err = -ENODEV;
@@ -1410,7 +1401,7 @@ iop_adma_pq_zero_sum_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (iop_adma_status(dma_chan, cookie, NULL) !=
 		DMA_SUCCESS) {
 		dev_err(dev, "Self-test !pq-zero-sum timed out, disabling\n");
 		err = -ENODEV;
@@ -1508,7 +1499,7 @@ static int __devinit iop_adma_probe(struct platform_device *pdev)
 	/* set base routines */
 	dma_dev->device_alloc_chan_resources = iop_adma_alloc_chan_resources;
 	dma_dev->device_free_chan_resources = iop_adma_free_chan_resources;
-	dma_dev->device_is_tx_complete = iop_adma_is_complete;
+	dma_dev->device_tx_status = iop_adma_status;
 	dma_dev->device_issue_pending = iop_adma_issue_pending;
 	dma_dev->dev = &pdev->dev;
 
diff --git a/drivers/dma/ipu/ipu_idmac.c b/drivers/dma/ipu/ipu_idmac.c
index 2a446397c88..cb26ee9773d 100644
--- a/drivers/dma/ipu/ipu_idmac.c
+++ b/drivers/dma/ipu/ipu_idmac.c
@@ -1472,13 +1472,18 @@ static void idmac_issue_pending(struct dma_chan *chan)
 	 */
 }
 
-static void __idmac_terminate_all(struct dma_chan *chan)
+static int __idmac_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+			   unsigned long arg)
 {
 	struct idmac_channel *ichan = to_idmac_chan(chan);
 	struct idmac *idmac = to_idmac(chan->device);
 	unsigned long flags;
 	int i;
 
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
 	ipu_disable_channel(idmac, ichan,
 			    ichan->status >= IPU_CHANNEL_ENABLED);
 
@@ -1505,17 +1510,23 @@ static void __idmac_terminate_all(struct dma_chan *chan)
 	tasklet_enable(&to_ipu(idmac)->tasklet);
 
 	ichan->status = IPU_CHANNEL_INITIALIZED;
+
+	return 0;
 }
 
-static void idmac_terminate_all(struct dma_chan *chan)
+static int idmac_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+			 unsigned long arg)
 {
 	struct idmac_channel *ichan = to_idmac_chan(chan);
+	int ret;
 
 	mutex_lock(&ichan->chan_mutex);
 
-	__idmac_terminate_all(chan);
+	ret = __idmac_control(chan, cmd, arg);
 
 	mutex_unlock(&ichan->chan_mutex);
+
+	return ret;
 }
 
 #ifdef DEBUG
@@ -1607,7 +1618,7 @@ static void idmac_free_chan_resources(struct dma_chan *chan)
 
 	mutex_lock(&ichan->chan_mutex);
 
-	__idmac_terminate_all(chan);
+	__idmac_control(chan, DMA_TERMINATE_ALL, 0);
 
 	if (ichan->status > IPU_CHANNEL_FREE) {
 #ifdef DEBUG
@@ -1637,15 +1648,12 @@ static void idmac_free_chan_resources(struct dma_chan *chan)
 	tasklet_schedule(&to_ipu(idmac)->tasklet);
 }
 
-static enum dma_status idmac_is_tx_complete(struct dma_chan *chan,
-		dma_cookie_t cookie, dma_cookie_t *done, dma_cookie_t *used)
+static enum dma_status idmac_tx_status(struct dma_chan *chan,
+		       dma_cookie_t cookie, struct dma_tx_state *txstate)
 {
 	struct idmac_channel *ichan = to_idmac_chan(chan);
 
-	if (done)
-		*done = ichan->completed;
-	if (used)
-		*used = chan->cookie;
+	dma_set_tx_state(txstate, ichan->completed, chan->cookie, 0);
 	if (cookie != chan->cookie)
 		return DMA_ERROR;
 	return DMA_SUCCESS;
@@ -1664,12 +1672,12 @@ static int __init ipu_idmac_init(struct ipu *ipu)
 	dma->dev				= ipu->dev;
 	dma->device_alloc_chan_resources	= idmac_alloc_chan_resources;
 	dma->device_free_chan_resources		= idmac_free_chan_resources;
-	dma->device_is_tx_complete		= idmac_is_tx_complete;
+	dma->device_tx_status			= idmac_tx_status;
 	dma->device_issue_pending		= idmac_issue_pending;
 
 	/* Compulsory for DMA_SLAVE fields */
 	dma->device_prep_slave_sg		= idmac_prep_slave_sg;
-	dma->device_terminate_all		= idmac_terminate_all;
+	dma->device_control			= idmac_control;
 
 	INIT_LIST_HEAD(&dma->channels);
 	for (i = 0; i < IPU_CHANNELS_NUM; i++) {
@@ -1703,7 +1711,7 @@ static void __exit ipu_idmac_exit(struct ipu *ipu)
 	for (i = 0; i < IPU_CHANNELS_NUM; i++) {
 		struct idmac_channel *ichan = ipu->channel + i;
 
-		idmac_terminate_all(&ichan->dma_chan);
+		idmac_control(&ichan->dma_chan, DMA_TERMINATE_ALL, 0);
 		idmac_prep_slave_sg(&ichan->dma_chan, NULL, 0, DMA_NONE, 0);
 	}
 
diff --git a/drivers/dma/mpc512x_dma.c b/drivers/dma/mpc512x_dma.c
index bbbd5856662..201e6e19c34 100644
--- a/drivers/dma/mpc512x_dma.c
+++ b/drivers/dma/mpc512x_dma.c
@@ -541,8 +541,8 @@ static void mpc_dma_issue_pending(struct dma_chan *chan)
 
 /* Check request completion status */
 static enum dma_status
-mpc_dma_is_tx_complete(struct dma_chan *chan, dma_cookie_t cookie,
-					dma_cookie_t *done, dma_cookie_t *used)
+mpc_dma_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
+	       struct dma_tx_state *txstate)
 {
 	struct mpc_dma_chan *mchan = dma_chan_to_mpc_dma_chan(chan);
 	unsigned long flags;
@@ -554,12 +554,7 @@ mpc_dma_is_tx_complete(struct dma_chan *chan, dma_cookie_t cookie,
 	last_complete = mchan->completed_cookie;
 	spin_unlock_irqrestore(&mchan->lock, flags);
 
-	if (done)
-		*done = last_complete;
-
-	if (used)
-		*used = last_used;
-
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 	return dma_async_is_complete(cookie, last_complete, last_used);
 }
 
@@ -663,7 +658,7 @@ static int __devinit mpc_dma_probe(struct of_device *op,
 	}
 
 	regs_start = res.start;
-	regs_size = res.end - res.start + 1;
+	regs_size = resource_size(&res);
 
 	if (!devm_request_mem_region(dev, regs_start, regs_size, DRV_NAME)) {
 		dev_err(dev, "Error requesting memory region!\n");
@@ -694,7 +689,7 @@ static int __devinit mpc_dma_probe(struct of_device *op,
 	dma->device_alloc_chan_resources = mpc_dma_alloc_chan_resources;
 	dma->device_free_chan_resources = mpc_dma_free_chan_resources;
 	dma->device_issue_pending = mpc_dma_issue_pending;
-	dma->device_is_tx_complete = mpc_dma_is_tx_complete;
+	dma->device_tx_status = mpc_dma_tx_status;
 	dma->device_prep_dma_memcpy = mpc_dma_prep_memcpy;
 
 	INIT_LIST_HEAD(&dma->channels);
diff --git a/drivers/dma/mv_xor.c b/drivers/dma/mv_xor.c
index e2fd34da64f..86c5ae9fde3 100644
--- a/drivers/dma/mv_xor.c
+++ b/drivers/dma/mv_xor.c
@@ -810,14 +810,14 @@ static void mv_xor_free_chan_resources(struct dma_chan *chan)
 }
 
 /**
- * mv_xor_is_complete - poll the status of an XOR transaction
+ * mv_xor_status - poll the status of an XOR transaction
  * @chan: XOR channel handle
  * @cookie: XOR transaction identifier
+ * @txstate: XOR transactions state holder (or NULL)
  */
-static enum dma_status mv_xor_is_complete(struct dma_chan *chan,
+static enum dma_status mv_xor_status(struct dma_chan *chan,
 					  dma_cookie_t cookie,
-					  dma_cookie_t *done,
-					  dma_cookie_t *used)
+					  struct dma_tx_state *txstate)
 {
 	struct mv_xor_chan *mv_chan = to_mv_xor_chan(chan);
 	dma_cookie_t last_used;
@@ -827,10 +827,7 @@ static enum dma_status mv_xor_is_complete(struct dma_chan *chan,
 	last_used = chan->cookie;
 	last_complete = mv_chan->completed_cookie;
 	mv_chan->is_complete_cookie = cookie;
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	ret = dma_async_is_complete(cookie, last_complete, last_used);
 	if (ret == DMA_SUCCESS) {
@@ -842,11 +839,7 @@ static enum dma_status mv_xor_is_complete(struct dma_chan *chan,
 	last_used = chan->cookie;
 	last_complete = mv_chan->completed_cookie;
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
-
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 	return dma_async_is_complete(cookie, last_complete, last_used);
 }
 
@@ -975,7 +968,7 @@ static int __devinit mv_xor_memcpy_self_test(struct mv_xor_device *device)
 	async_tx_ack(tx);
 	msleep(1);
 
-	if (mv_xor_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (mv_xor_status(dma_chan, cookie, NULL) !=
 	    DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			   "Self-test copy timed out, disabling\n");
@@ -1073,7 +1066,7 @@ mv_xor_xor_self_test(struct mv_xor_device *device)
 	async_tx_ack(tx);
 	msleep(8);
 
-	if (mv_xor_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (mv_xor_status(dma_chan, cookie, NULL) !=
 	    DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			   "Self-test xor timed out, disabling\n");
@@ -1168,7 +1161,7 @@ static int __devinit mv_xor_probe(struct platform_device *pdev)
 	/* set base routines */
 	dma_dev->device_alloc_chan_resources = mv_xor_alloc_chan_resources;
 	dma_dev->device_free_chan_resources = mv_xor_free_chan_resources;
-	dma_dev->device_is_tx_complete = mv_xor_is_complete;
+	dma_dev->device_tx_status = mv_xor_status;
 	dma_dev->device_issue_pending = mv_xor_issue_pending;
 	dma_dev->dev = &pdev->dev;
 
diff --git a/drivers/dma/ppc4xx/adma.c b/drivers/dma/ppc4xx/adma.c
index d44626fa35a..c6079fcca13 100644
--- a/drivers/dma/ppc4xx/adma.c
+++ b/drivers/dma/ppc4xx/adma.c
@@ -3935,12 +3935,13 @@ static void ppc440spe_adma_free_chan_resources(struct dma_chan *chan)
 }
 
 /**
- * ppc440spe_adma_is_complete - poll the status of an ADMA transaction
+ * ppc440spe_adma_tx_status - poll the status of an ADMA transaction
  * @chan: ADMA channel handle
  * @cookie: ADMA transaction identifier
+ * @txstate: a holder for the current state of the channel
  */
-static enum dma_status ppc440spe_adma_is_complete(struct dma_chan *chan,
-	dma_cookie_t cookie, dma_cookie_t *done, dma_cookie_t *used)
+static enum dma_status ppc440spe_adma_tx_status(struct dma_chan *chan,
+			dma_cookie_t cookie, struct dma_tx_state *txstate)
 {
 	struct ppc440spe_adma_chan *ppc440spe_chan;
 	dma_cookie_t last_used;
@@ -3951,10 +3952,7 @@ static enum dma_status ppc440spe_adma_is_complete(struct dma_chan *chan,
 	last_used = chan->cookie;
 	last_complete = ppc440spe_chan->completed_cookie;
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	ret = dma_async_is_complete(cookie, last_complete, last_used);
 	if (ret == DMA_SUCCESS)
@@ -3965,10 +3963,7 @@ static enum dma_status ppc440spe_adma_is_complete(struct dma_chan *chan,
 	last_used = chan->cookie;
 	last_complete = ppc440spe_chan->completed_cookie;
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	return dma_async_is_complete(cookie, last_complete, last_used);
 }
@@ -4180,7 +4175,7 @@ static void ppc440spe_adma_init_capabilities(struct ppc440spe_adma_device *adev)
 				ppc440spe_adma_alloc_chan_resources;
 	adev->common.device_free_chan_resources =
 				ppc440spe_adma_free_chan_resources;
-	adev->common.device_is_tx_complete = ppc440spe_adma_is_complete;
+	adev->common.device_tx_status = ppc440spe_adma_tx_status;
 	adev->common.device_issue_pending = ppc440spe_adma_issue_pending;
 
 	/* Set prep routines based on capability */
diff --git a/drivers/dma/shdma.c b/drivers/dma/shdma.c
index 323afef7780..a1727522343 100644
--- a/drivers/dma/shdma.c
+++ b/drivers/dma/shdma.c
@@ -597,12 +597,17 @@ static struct dma_async_tx_descriptor *sh_dmae_prep_slave_sg(
 			       direction, flags);
 }
 
-static void sh_dmae_terminate_all(struct dma_chan *chan)
+static int sh_dmae_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+			   unsigned long arg)
 {
 	struct sh_dmae_chan *sh_chan = to_sh_chan(chan);
 
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
 	if (!chan)
-		return;
+		return -EINVAL;
 
 	dmae_halt(sh_chan);
 
@@ -618,6 +623,8 @@ static void sh_dmae_terminate_all(struct dma_chan *chan)
 	spin_unlock_bh(&sh_chan->desc_lock);
 
 	sh_dmae_chan_ld_cleanup(sh_chan, true);
+
+	return 0;
 }
 
 static dma_async_tx_callback __ld_cleanup(struct sh_dmae_chan *sh_chan, bool all)
@@ -749,10 +756,9 @@ static void sh_dmae_memcpy_issue_pending(struct dma_chan *chan)
 	sh_chan_xfer_ld_queue(sh_chan);
 }
 
-static enum dma_status sh_dmae_is_complete(struct dma_chan *chan,
+static enum dma_status sh_dmae_tx_status(struct dma_chan *chan,
 					dma_cookie_t cookie,
-					dma_cookie_t *done,
-					dma_cookie_t *used)
+					struct dma_tx_state *txstate)
 {
 	struct sh_dmae_chan *sh_chan = to_sh_chan(chan);
 	dma_cookie_t last_used;
@@ -764,12 +770,7 @@ static enum dma_status sh_dmae_is_complete(struct dma_chan *chan,
 	last_used = chan->cookie;
 	last_complete = sh_chan->completed_cookie;
 	BUG_ON(last_complete < 0);
-
-	if (done)
-		*done = last_complete;
-
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	spin_lock_bh(&sh_chan->desc_lock);
 
@@ -1041,12 +1042,12 @@ static int __init sh_dmae_probe(struct platform_device *pdev)
 		= sh_dmae_alloc_chan_resources;
 	shdev->common.device_free_chan_resources = sh_dmae_free_chan_resources;
 	shdev->common.device_prep_dma_memcpy = sh_dmae_prep_memcpy;
-	shdev->common.device_is_tx_complete = sh_dmae_is_complete;
+	shdev->common.device_tx_status = sh_dmae_tx_status;
 	shdev->common.device_issue_pending = sh_dmae_memcpy_issue_pending;
 
 	/* Compulsory for DMA_SLAVE fields */
 	shdev->common.device_prep_slave_sg = sh_dmae_prep_slave_sg;
-	shdev->common.device_terminate_all = sh_dmae_terminate_all;
+	shdev->common.device_control = sh_dmae_control;
 
 	shdev->common.dev = &pdev->dev;
 	/* Default transfer size of 32 bytes requires 32-byte alignment */
diff --git a/drivers/dma/ste_dma40.c b/drivers/dma/ste_dma40.c
new file mode 100644
index 00000000000..c426829f6ab
--- /dev/null
+++ b/drivers/dma/ste_dma40.c
@@ -0,0 +1,2657 @@
+/*
+ * driver/dma/ste_dma40.c
+ *
+ * Copyright (C) ST-Ericsson 2007-2010
+ * License terms: GNU General Public License (GPL) version 2
+ * Author: Per Friden <per.friden@stericsson.com>
+ * Author: Jonas Aaberg <jonas.aberg@stericsson.com>
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/dmaengine.h>
+#include <linux/platform_device.h>
+#include <linux/clk.h>
+#include <linux/delay.h>
+
+#include <plat/ste_dma40.h>
+
+#include "ste_dma40_ll.h"
+
+#define D40_NAME "dma40"
+
+#define D40_PHY_CHAN -1
+
+/* For masking out/in 2 bit channel positions */
+#define D40_CHAN_POS(chan)  (2 * (chan / 2))
+#define D40_CHAN_POS_MASK(chan) (0x3 << D40_CHAN_POS(chan))
+
+/* Maximum iterations taken before giving up suspending a channel */
+#define D40_SUSPEND_MAX_IT 500
+
+#define D40_ALLOC_FREE		(1 << 31)
+#define D40_ALLOC_PHY		(1 << 30)
+#define D40_ALLOC_LOG_FREE	0
+
+/* The number of free d40_desc to keep in memory before starting
+ * to kfree() them */
+#define D40_DESC_CACHE_SIZE 50
+
+/* Hardware designer of the block */
+#define D40_PERIPHID2_DESIGNER 0x8
+
+/**
+ * enum 40_command - The different commands and/or statuses.
+ *
+ * @D40_DMA_STOP: DMA channel command STOP or status STOPPED,
+ * @D40_DMA_RUN: The DMA channel is RUNNING of the command RUN.
+ * @D40_DMA_SUSPEND_REQ: Request the DMA to SUSPEND as soon as possible.
+ * @D40_DMA_SUSPENDED: The DMA channel is SUSPENDED.
+ */
+enum d40_command {
+	D40_DMA_STOP		= 0,
+	D40_DMA_RUN		= 1,
+	D40_DMA_SUSPEND_REQ	= 2,
+	D40_DMA_SUSPENDED	= 3
+};
+
+/**
+ * struct d40_lli_pool - Structure for keeping LLIs in memory
+ *
+ * @base: Pointer to memory area when the pre_alloc_lli's are not large
+ * enough, IE bigger than the most common case, 1 dst and 1 src. NULL if
+ * pre_alloc_lli is used.
+ * @size: The size in bytes of the memory at base or the size of pre_alloc_lli.
+ * @pre_alloc_lli: Pre allocated area for the most common case of transfers,
+ * one buffer to one buffer.
+ */
+struct d40_lli_pool {
+	void	*base;
+	int	size;
+	/* Space for dst and src, plus an extra for padding */
+	u8	pre_alloc_lli[3 * sizeof(struct d40_phy_lli)];
+};
+
+/**
+ * struct d40_desc - A descriptor is one DMA job.
+ *
+ * @lli_phy: LLI settings for physical channel. Both src and dst=
+ * points into the lli_pool, to base if lli_len > 1 or to pre_alloc_lli if
+ * lli_len equals one.
+ * @lli_log: Same as above but for logical channels.
+ * @lli_pool: The pool with two entries pre-allocated.
+ * @lli_len: Number of LLI's in lli_pool
+ * @lli_tcount: Number of LLIs processed in the transfer. When equals lli_len
+ * then this transfer job is done.
+ * @txd: DMA engine struct. Used for among other things for communication
+ * during a transfer.
+ * @node: List entry.
+ * @dir: The transfer direction of this job.
+ * @is_in_client_list: true if the client owns this descriptor.
+ *
+ * This descriptor is used for both logical and physical transfers.
+ */
+
+struct d40_desc {
+	/* LLI physical */
+	struct d40_phy_lli_bidir	 lli_phy;
+	/* LLI logical */
+	struct d40_log_lli_bidir	 lli_log;
+
+	struct d40_lli_pool		 lli_pool;
+	u32				 lli_len;
+	u32				 lli_tcount;
+
+	struct dma_async_tx_descriptor	 txd;
+	struct list_head		 node;
+
+	enum dma_data_direction		 dir;
+	bool				 is_in_client_list;
+};
+
+/**
+ * struct d40_lcla_pool - LCLA pool settings and data.
+ *
+ * @base: The virtual address of LCLA.
+ * @phy: Physical base address of LCLA.
+ * @base_size: size of lcla.
+ * @lock: Lock to protect the content in this struct.
+ * @alloc_map: Mapping between physical channel and LCLA entries.
+ * @num_blocks: The number of entries of alloc_map. Equals to the
+ * number of physical channels.
+ */
+struct d40_lcla_pool {
+	void		*base;
+	dma_addr_t	 phy;
+	resource_size_t  base_size;
+	spinlock_t	 lock;
+	u32		*alloc_map;
+	int		 num_blocks;
+};
+
+/**
+ * struct d40_phy_res - struct for handling eventlines mapped to physical
+ * channels.
+ *
+ * @lock: A lock protection this entity.
+ * @num: The physical channel number of this entity.
+ * @allocated_src: Bit mapped to show which src event line's are mapped to
+ * this physical channel. Can also be free or physically allocated.
+ * @allocated_dst: Same as for src but is dst.
+ * allocated_dst and allocated_src uses the D40_ALLOC* defines as well as
+ * event line number. Both allocated_src and allocated_dst can not be
+ * allocated to a physical channel, since the interrupt handler has then
+ * no way of figure out which one the interrupt belongs to.
+ */
+struct d40_phy_res {
+	spinlock_t lock;
+	int	   num;
+	u32	   allocated_src;
+	u32	   allocated_dst;
+};
+
+struct d40_base;
+
+/**
+ * struct d40_chan - Struct that describes a channel.
+ *
+ * @lock: A spinlock to protect this struct.
+ * @log_num: The logical number, if any of this channel.
+ * @completed: Starts with 1, after first interrupt it is set to dma engine's
+ * current cookie.
+ * @pending_tx: The number of pending transfers. Used between interrupt handler
+ * and tasklet.
+ * @busy: Set to true when transfer is ongoing on this channel.
+ * @phy_chan: Pointer to physical channel which this instance runs on.
+ * @chan: DMA engine handle.
+ * @tasklet: Tasklet that gets scheduled from interrupt context to complete a
+ * transfer and call client callback.
+ * @client: Cliented owned descriptor list.
+ * @active: Active descriptor.
+ * @queue: Queued jobs.
+ * @free: List of free descripts, ready to be reused.
+ * @free_len: Number of descriptors in the free list.
+ * @dma_cfg: The client configuration of this dma channel.
+ * @base: Pointer to the device instance struct.
+ * @src_def_cfg: Default cfg register setting for src.
+ * @dst_def_cfg: Default cfg register setting for dst.
+ * @log_def: Default logical channel settings.
+ * @lcla: Space for one dst src pair for logical channel transfers.
+ * @lcpa: Pointer to dst and src lcpa settings.
+ *
+ * This struct can either "be" a logical or a physical channel.
+ */
+struct d40_chan {
+	spinlock_t			 lock;
+	int				 log_num;
+	/* ID of the most recent completed transfer */
+	int				 completed;
+	int				 pending_tx;
+	bool				 busy;
+	struct d40_phy_res		*phy_chan;
+	struct dma_chan			 chan;
+	struct tasklet_struct		 tasklet;
+	struct list_head		 client;
+	struct list_head		 active;
+	struct list_head		 queue;
+	struct list_head		 free;
+	int				 free_len;
+	struct stedma40_chan_cfg	 dma_cfg;
+	struct d40_base			*base;
+	/* Default register configurations */
+	u32				 src_def_cfg;
+	u32				 dst_def_cfg;
+	struct d40_def_lcsp		 log_def;
+	struct d40_lcla_elem		 lcla;
+	struct d40_log_lli_full		*lcpa;
+};
+
+/**
+ * struct d40_base - The big global struct, one for each probe'd instance.
+ *
+ * @interrupt_lock: Lock used to make sure one interrupt is handle a time.
+ * @execmd_lock: Lock for execute command usage since several channels share
+ * the same physical register.
+ * @dev: The device structure.
+ * @virtbase: The virtual base address of the DMA's register.
+ * @clk: Pointer to the DMA clock structure.
+ * @phy_start: Physical memory start of the DMA registers.
+ * @phy_size: Size of the DMA register map.
+ * @irq: The IRQ number.
+ * @num_phy_chans: The number of physical channels. Read from HW. This
+ * is the number of available channels for this driver, not counting "Secure
+ * mode" allocated physical channels.
+ * @num_log_chans: The number of logical channels. Calculated from
+ * num_phy_chans.
+ * @dma_both: dma_device channels that can do both memcpy and slave transfers.
+ * @dma_slave: dma_device channels that can do only do slave transfers.
+ * @dma_memcpy: dma_device channels that can do only do memcpy transfers.
+ * @phy_chans: Room for all possible physical channels in system.
+ * @log_chans: Room for all possible logical channels in system.
+ * @lookup_log_chans: Used to map interrupt number to logical channel. Points
+ * to log_chans entries.
+ * @lookup_phy_chans: Used to map interrupt number to physical channel. Points
+ * to phy_chans entries.
+ * @plat_data: Pointer to provided platform_data which is the driver
+ * configuration.
+ * @phy_res: Vector containing all physical channels.
+ * @lcla_pool: lcla pool settings and data.
+ * @lcpa_base: The virtual mapped address of LCPA.
+ * @phy_lcpa: The physical address of the LCPA.
+ * @lcpa_size: The size of the LCPA area.
+ */
+struct d40_base {
+	spinlock_t			 interrupt_lock;
+	spinlock_t			 execmd_lock;
+	struct device			 *dev;
+	void __iomem			 *virtbase;
+	struct clk			 *clk;
+	phys_addr_t			  phy_start;
+	resource_size_t			  phy_size;
+	int				  irq;
+	int				  num_phy_chans;
+	int				  num_log_chans;
+	struct dma_device		  dma_both;
+	struct dma_device		  dma_slave;
+	struct dma_device		  dma_memcpy;
+	struct d40_chan			 *phy_chans;
+	struct d40_chan			 *log_chans;
+	struct d40_chan			**lookup_log_chans;
+	struct d40_chan			**lookup_phy_chans;
+	struct stedma40_platform_data	 *plat_data;
+	/* Physical half channels */
+	struct d40_phy_res		 *phy_res;
+	struct d40_lcla_pool		  lcla_pool;
+	void				 *lcpa_base;
+	dma_addr_t			  phy_lcpa;
+	resource_size_t			  lcpa_size;
+};
+
+/**
+ * struct d40_interrupt_lookup - lookup table for interrupt handler
+ *
+ * @src: Interrupt mask register.
+ * @clr: Interrupt clear register.
+ * @is_error: true if this is an error interrupt.
+ * @offset: start delta in the lookup_log_chans in d40_base. If equals to
+ * D40_PHY_CHAN, the lookup_phy_chans shall be used instead.
+ */
+struct d40_interrupt_lookup {
+	u32 src;
+	u32 clr;
+	bool is_error;
+	int offset;
+};
+
+/**
+ * struct d40_reg_val - simple lookup struct
+ *
+ * @reg: The register.
+ * @val: The value that belongs to the register in reg.
+ */
+struct d40_reg_val {
+	unsigned int reg;
+	unsigned int val;
+};
+
+static int d40_pool_lli_alloc(struct d40_desc *d40d,
+			      int lli_len, bool is_log)
+{
+	u32 align;
+	void *base;
+
+	if (is_log)
+		align = sizeof(struct d40_log_lli);
+	else
+		align = sizeof(struct d40_phy_lli);
+
+	if (lli_len == 1) {
+		base = d40d->lli_pool.pre_alloc_lli;
+		d40d->lli_pool.size = sizeof(d40d->lli_pool.pre_alloc_lli);
+		d40d->lli_pool.base = NULL;
+	} else {
+		d40d->lli_pool.size = ALIGN(lli_len * 2 * align, align);
+
+		base = kmalloc(d40d->lli_pool.size + align, GFP_NOWAIT);
+		d40d->lli_pool.base = base;
+
+		if (d40d->lli_pool.base == NULL)
+			return -ENOMEM;
+	}
+
+	if (is_log) {
+		d40d->lli_log.src = PTR_ALIGN((struct d40_log_lli *) base,
+					      align);
+		d40d->lli_log.dst = PTR_ALIGN(d40d->lli_log.src + lli_len,
+					      align);
+	} else {
+		d40d->lli_phy.src = PTR_ALIGN((struct d40_phy_lli *)base,
+					      align);
+		d40d->lli_phy.dst = PTR_ALIGN(d40d->lli_phy.src + lli_len,
+					      align);
+
+		d40d->lli_phy.src_addr = virt_to_phys(d40d->lli_phy.src);
+		d40d->lli_phy.dst_addr = virt_to_phys(d40d->lli_phy.dst);
+	}
+
+	return 0;
+}
+
+static void d40_pool_lli_free(struct d40_desc *d40d)
+{
+	kfree(d40d->lli_pool.base);
+	d40d->lli_pool.base = NULL;
+	d40d->lli_pool.size = 0;
+	d40d->lli_log.src = NULL;
+	d40d->lli_log.dst = NULL;
+	d40d->lli_phy.src = NULL;
+	d40d->lli_phy.dst = NULL;
+	d40d->lli_phy.src_addr = 0;
+	d40d->lli_phy.dst_addr = 0;
+}
+
+static dma_cookie_t d40_assign_cookie(struct d40_chan *d40c,
+				      struct d40_desc *desc)
+{
+	dma_cookie_t cookie = d40c->chan.cookie;
+
+	if (++cookie < 0)
+		cookie = 1;
+
+	d40c->chan.cookie = cookie;
+	desc->txd.cookie = cookie;
+
+	return cookie;
+}
+
+static void d40_desc_reset(struct d40_desc *d40d)
+{
+	d40d->lli_tcount = 0;
+}
+
+static void d40_desc_remove(struct d40_desc *d40d)
+{
+	list_del(&d40d->node);
+}
+
+static struct d40_desc *d40_desc_get(struct d40_chan *d40c)
+{
+	struct d40_desc *desc;
+	struct d40_desc *d;
+	struct d40_desc *_d;
+
+	if (!list_empty(&d40c->client)) {
+		list_for_each_entry_safe(d, _d, &d40c->client, node)
+			if (async_tx_test_ack(&d->txd)) {
+				d40_pool_lli_free(d);
+				d40_desc_remove(d);
+				desc = d;
+				goto out;
+			}
+	}
+
+	if (list_empty(&d40c->free)) {
+		/* Alloc new desc because we're out of used ones */
+		desc = kzalloc(sizeof(struct d40_desc), GFP_NOWAIT);
+		if (desc == NULL)
+			goto out;
+		INIT_LIST_HEAD(&desc->node);
+	} else {
+		/* Reuse an old desc. */
+		desc = list_first_entry(&d40c->free,
+					struct d40_desc,
+					node);
+		list_del(&desc->node);
+		d40c->free_len--;
+	}
+out:
+	return desc;
+}
+
+static void d40_desc_free(struct d40_chan *d40c, struct d40_desc *d40d)
+{
+	if (d40c->free_len < D40_DESC_CACHE_SIZE) {
+		list_add_tail(&d40d->node, &d40c->free);
+		d40c->free_len++;
+	} else
+		kfree(d40d);
+}
+
+static void d40_desc_submit(struct d40_chan *d40c, struct d40_desc *desc)
+{
+	list_add_tail(&desc->node, &d40c->active);
+}
+
+static struct d40_desc *d40_first_active_get(struct d40_chan *d40c)
+{
+	struct d40_desc *d;
+
+	if (list_empty(&d40c->active))
+		return NULL;
+
+	d = list_first_entry(&d40c->active,
+			     struct d40_desc,
+			     node);
+	return d;
+}
+
+static void d40_desc_queue(struct d40_chan *d40c, struct d40_desc *desc)
+{
+	list_add_tail(&desc->node, &d40c->queue);
+}
+
+static struct d40_desc *d40_first_queued(struct d40_chan *d40c)
+{
+	struct d40_desc *d;
+
+	if (list_empty(&d40c->queue))
+		return NULL;
+
+	d = list_first_entry(&d40c->queue,
+			     struct d40_desc,
+			     node);
+	return d;
+}
+
+/* Support functions for logical channels */
+
+static int d40_lcla_id_get(struct d40_chan *d40c,
+			   struct d40_lcla_pool *pool)
+{
+	int src_id = 0;
+	int dst_id = 0;
+	struct d40_log_lli *lcla_lidx_base =
+		pool->base + d40c->phy_chan->num * 1024;
+	int i;
+	int lli_per_log = d40c->base->plat_data->llis_per_log;
+
+	if (d40c->lcla.src_id >= 0 && d40c->lcla.dst_id >= 0)
+		return 0;
+
+	if (pool->num_blocks > 32)
+		return -EINVAL;
+
+	spin_lock(&pool->lock);
+
+	for (i = 0; i < pool->num_blocks; i++) {
+		if (!(pool->alloc_map[d40c->phy_chan->num] & (0x1 << i))) {
+			pool->alloc_map[d40c->phy_chan->num] |= (0x1 << i);
+			break;
+		}
+	}
+	src_id = i;
+	if (src_id >= pool->num_blocks)
+		goto err;
+
+	for (; i < pool->num_blocks; i++) {
+		if (!(pool->alloc_map[d40c->phy_chan->num] & (0x1 << i))) {
+			pool->alloc_map[d40c->phy_chan->num] |= (0x1 << i);
+			break;
+		}
+	}
+
+	dst_id = i;
+	if (dst_id == src_id)
+		goto err;
+
+	d40c->lcla.src_id = src_id;
+	d40c->lcla.dst_id = dst_id;
+	d40c->lcla.dst = lcla_lidx_base + dst_id * lli_per_log + 1;
+	d40c->lcla.src = lcla_lidx_base + src_id * lli_per_log + 1;
+
+
+	spin_unlock(&pool->lock);
+	return 0;
+err:
+	spin_unlock(&pool->lock);
+	return -EINVAL;
+}
+
+static void d40_lcla_id_put(struct d40_chan *d40c,
+			    struct d40_lcla_pool *pool,
+			    int id)
+{
+	if (id < 0)
+		return;
+
+	d40c->lcla.src_id = -1;
+	d40c->lcla.dst_id = -1;
+
+	spin_lock(&pool->lock);
+	pool->alloc_map[d40c->phy_chan->num] &= (~(0x1 << id));
+	spin_unlock(&pool->lock);
+}
+
+static int d40_channel_execute_command(struct d40_chan *d40c,
+				       enum d40_command command)
+{
+	int status, i;
+	void __iomem *active_reg;
+	int ret = 0;
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->base->execmd_lock, flags);
+
+	if (d40c->phy_chan->num % 2 == 0)
+		active_reg = d40c->base->virtbase + D40_DREG_ACTIVE;
+	else
+		active_reg = d40c->base->virtbase + D40_DREG_ACTIVO;
+
+	if (command == D40_DMA_SUSPEND_REQ) {
+		status = (readl(active_reg) &
+			  D40_CHAN_POS_MASK(d40c->phy_chan->num)) >>
+			D40_CHAN_POS(d40c->phy_chan->num);
+
+		if (status == D40_DMA_SUSPENDED || status == D40_DMA_STOP)
+			goto done;
+	}
+
+	writel(command << D40_CHAN_POS(d40c->phy_chan->num), active_reg);
+
+	if (command == D40_DMA_SUSPEND_REQ) {
+
+		for (i = 0 ; i < D40_SUSPEND_MAX_IT; i++) {
+			status = (readl(active_reg) &
+				  D40_CHAN_POS_MASK(d40c->phy_chan->num)) >>
+				D40_CHAN_POS(d40c->phy_chan->num);
+
+			cpu_relax();
+			/*
+			 * Reduce the number of bus accesses while
+			 * waiting for the DMA to suspend.
+			 */
+			udelay(3);
+
+			if (status == D40_DMA_STOP ||
+			    status == D40_DMA_SUSPENDED)
+				break;
+		}
+
+		if (i == D40_SUSPEND_MAX_IT) {
+			dev_err(&d40c->chan.dev->device,
+				"[%s]: unable to suspend the chl %d (log: %d) status %x\n",
+				__func__, d40c->phy_chan->num, d40c->log_num,
+				status);
+			dump_stack();
+			ret = -EBUSY;
+		}
+
+	}
+done:
+	spin_unlock_irqrestore(&d40c->base->execmd_lock, flags);
+	return ret;
+}
+
+static void d40_term_all(struct d40_chan *d40c)
+{
+	struct d40_desc *d40d;
+	struct d40_desc *d;
+	struct d40_desc *_d;
+
+	/* Release active descriptors */
+	while ((d40d = d40_first_active_get(d40c))) {
+		d40_desc_remove(d40d);
+
+		/* Return desc to free-list */
+		d40_desc_free(d40c, d40d);
+	}
+
+	/* Release queued descriptors waiting for transfer */
+	while ((d40d = d40_first_queued(d40c))) {
+		d40_desc_remove(d40d);
+
+		/* Return desc to free-list */
+		d40_desc_free(d40c, d40d);
+	}
+
+	/* Release client owned descriptors */
+	if (!list_empty(&d40c->client))
+		list_for_each_entry_safe(d, _d, &d40c->client, node) {
+			d40_pool_lli_free(d);
+			d40_desc_remove(d);
+			/* Return desc to free-list */
+			d40_desc_free(d40c, d40d);
+		}
+
+	d40_lcla_id_put(d40c, &d40c->base->lcla_pool,
+			d40c->lcla.src_id);
+	d40_lcla_id_put(d40c, &d40c->base->lcla_pool,
+			d40c->lcla.dst_id);
+
+	d40c->pending_tx = 0;
+	d40c->busy = false;
+}
+
+static void d40_config_set_event(struct d40_chan *d40c, bool do_enable)
+{
+	u32 val;
+	unsigned long flags;
+
+	if (do_enable)
+		val = D40_ACTIVATE_EVENTLINE;
+	else
+		val = D40_DEACTIVATE_EVENTLINE;
+
+	spin_lock_irqsave(&d40c->phy_chan->lock, flags);
+
+	/* Enable event line connected to device (or memcpy) */
+	if ((d40c->dma_cfg.dir ==  STEDMA40_PERIPH_TO_MEM) ||
+	    (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_PERIPH)) {
+		u32 event = D40_TYPE_TO_EVENT(d40c->dma_cfg.src_dev_type);
+
+		writel((val << D40_EVENTLINE_POS(event)) |
+		       ~D40_EVENTLINE_MASK(event),
+		       d40c->base->virtbase + D40_DREG_PCBASE +
+		       d40c->phy_chan->num * D40_DREG_PCDELTA +
+		       D40_CHAN_REG_SSLNK);
+	}
+	if (d40c->dma_cfg.dir !=  STEDMA40_PERIPH_TO_MEM) {
+		u32 event = D40_TYPE_TO_EVENT(d40c->dma_cfg.dst_dev_type);
+
+		writel((val << D40_EVENTLINE_POS(event)) |
+		       ~D40_EVENTLINE_MASK(event),
+		       d40c->base->virtbase + D40_DREG_PCBASE +
+		       d40c->phy_chan->num * D40_DREG_PCDELTA +
+		       D40_CHAN_REG_SDLNK);
+	}
+
+	spin_unlock_irqrestore(&d40c->phy_chan->lock, flags);
+}
+
+static u32 d40_chan_has_events(struct d40_chan *d40c)
+{
+	u32 val = 0;
+
+	/* If SSLNK or SDLNK is zero all events are disabled */
+	if ((d40c->dma_cfg.dir ==  STEDMA40_PERIPH_TO_MEM) ||
+	    (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_PERIPH))
+		val = readl(d40c->base->virtbase + D40_DREG_PCBASE +
+			    d40c->phy_chan->num * D40_DREG_PCDELTA +
+			    D40_CHAN_REG_SSLNK);
+
+	if (d40c->dma_cfg.dir !=  STEDMA40_PERIPH_TO_MEM)
+		val = readl(d40c->base->virtbase + D40_DREG_PCBASE +
+			    d40c->phy_chan->num * D40_DREG_PCDELTA +
+			    D40_CHAN_REG_SDLNK);
+	return val;
+}
+
+static void d40_config_enable_lidx(struct d40_chan *d40c)
+{
+	/* Set LIDX for lcla */
+	writel((d40c->phy_chan->num << D40_SREG_ELEM_LOG_LIDX_POS) &
+	       D40_SREG_ELEM_LOG_LIDX_MASK,
+	       d40c->base->virtbase + D40_DREG_PCBASE +
+	       d40c->phy_chan->num * D40_DREG_PCDELTA + D40_CHAN_REG_SDELT);
+
+	writel((d40c->phy_chan->num << D40_SREG_ELEM_LOG_LIDX_POS) &
+	       D40_SREG_ELEM_LOG_LIDX_MASK,
+	       d40c->base->virtbase + D40_DREG_PCBASE +
+	       d40c->phy_chan->num * D40_DREG_PCDELTA + D40_CHAN_REG_SSELT);
+}
+
+static int d40_config_write(struct d40_chan *d40c)
+{
+	u32 addr_base;
+	u32 var;
+	int res;
+
+	res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
+	if (res)
+		return res;
+
+	/* Odd addresses are even addresses + 4 */
+	addr_base = (d40c->phy_chan->num % 2) * 4;
+	/* Setup channel mode to logical or physical */
+	var = ((u32)(d40c->log_num != D40_PHY_CHAN) + 1) <<
+		D40_CHAN_POS(d40c->phy_chan->num);
+	writel(var, d40c->base->virtbase + D40_DREG_PRMSE + addr_base);
+
+	/* Setup operational mode option register */
+	var = ((d40c->dma_cfg.channel_type >> STEDMA40_INFO_CH_MODE_OPT_POS) &
+	       0x3) << D40_CHAN_POS(d40c->phy_chan->num);
+
+	writel(var, d40c->base->virtbase + D40_DREG_PRMOE + addr_base);
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		/* Set default config for CFG reg */
+		writel(d40c->src_def_cfg,
+		       d40c->base->virtbase + D40_DREG_PCBASE +
+		       d40c->phy_chan->num * D40_DREG_PCDELTA +
+		       D40_CHAN_REG_SSCFG);
+		writel(d40c->dst_def_cfg,
+		       d40c->base->virtbase + D40_DREG_PCBASE +
+		       d40c->phy_chan->num * D40_DREG_PCDELTA +
+		       D40_CHAN_REG_SDCFG);
+
+		d40_config_enable_lidx(d40c);
+	}
+	return res;
+}
+
+static void d40_desc_load(struct d40_chan *d40c, struct d40_desc *d40d)
+{
+
+	if (d40d->lli_phy.dst && d40d->lli_phy.src) {
+		d40_phy_lli_write(d40c->base->virtbase,
+				  d40c->phy_chan->num,
+				  d40d->lli_phy.dst,
+				  d40d->lli_phy.src);
+		d40d->lli_tcount = d40d->lli_len;
+	} else if (d40d->lli_log.dst && d40d->lli_log.src) {
+		u32 lli_len;
+		struct d40_log_lli *src = d40d->lli_log.src;
+		struct d40_log_lli *dst = d40d->lli_log.dst;
+
+		src += d40d->lli_tcount;
+		dst += d40d->lli_tcount;
+
+		if (d40d->lli_len <= d40c->base->plat_data->llis_per_log)
+			lli_len = d40d->lli_len;
+		else
+			lli_len = d40c->base->plat_data->llis_per_log;
+		d40d->lli_tcount += lli_len;
+		d40_log_lli_write(d40c->lcpa, d40c->lcla.src,
+				  d40c->lcla.dst,
+				  dst, src,
+				  d40c->base->plat_data->llis_per_log);
+	}
+}
+
+static dma_cookie_t d40_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+	struct d40_chan *d40c = container_of(tx->chan,
+					     struct d40_chan,
+					     chan);
+	struct d40_desc *d40d = container_of(tx, struct d40_desc, txd);
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	tx->cookie = d40_assign_cookie(d40c, d40d);
+
+	d40_desc_queue(d40c, d40d);
+
+	spin_unlock_irqrestore(&d40c->lock, flags);
+
+	return tx->cookie;
+}
+
+static int d40_start(struct d40_chan *d40c)
+{
+	int err;
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		err = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
+		if (err)
+			return err;
+		d40_config_set_event(d40c, true);
+	}
+
+	err = d40_channel_execute_command(d40c, D40_DMA_RUN);
+
+	return err;
+}
+
+static struct d40_desc *d40_queue_start(struct d40_chan *d40c)
+{
+	struct d40_desc *d40d;
+	int err;
+
+	/* Start queued jobs, if any */
+	d40d = d40_first_queued(d40c);
+
+	if (d40d != NULL) {
+		d40c->busy = true;
+
+		/* Remove from queue */
+		d40_desc_remove(d40d);
+
+		/* Add to active queue */
+		d40_desc_submit(d40c, d40d);
+
+		/* Initiate DMA job */
+		d40_desc_load(d40c, d40d);
+
+		/* Start dma job */
+		err = d40_start(d40c);
+
+		if (err)
+			return NULL;
+	}
+
+	return d40d;
+}
+
+/* called from interrupt context */
+static void dma_tc_handle(struct d40_chan *d40c)
+{
+	struct d40_desc *d40d;
+
+	if (!d40c->phy_chan)
+		return;
+
+	/* Get first active entry from list */
+	d40d = d40_first_active_get(d40c);
+
+	if (d40d == NULL)
+		return;
+
+	if (d40d->lli_tcount < d40d->lli_len) {
+
+		d40_desc_load(d40c, d40d);
+		/* Start dma job */
+		(void) d40_start(d40c);
+		return;
+	}
+
+	if (d40_queue_start(d40c) == NULL)
+		d40c->busy = false;
+
+	d40c->pending_tx++;
+	tasklet_schedule(&d40c->tasklet);
+
+}
+
+static void dma_tasklet(unsigned long data)
+{
+	struct d40_chan *d40c = (struct d40_chan *) data;
+	struct d40_desc *d40d_fin;
+	unsigned long flags;
+	dma_async_tx_callback callback;
+	void *callback_param;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	/* Get first active entry from list */
+	d40d_fin = d40_first_active_get(d40c);
+
+	if (d40d_fin == NULL)
+		goto err;
+
+	d40c->completed = d40d_fin->txd.cookie;
+
+	/*
+	 * If terminating a channel pending_tx is set to zero.
+	 * This prevents any finished active jobs to return to the client.
+	 */
+	if (d40c->pending_tx == 0) {
+		spin_unlock_irqrestore(&d40c->lock, flags);
+		return;
+	}
+
+	/* Callback to client */
+	callback = d40d_fin->txd.callback;
+	callback_param = d40d_fin->txd.callback_param;
+
+	if (async_tx_test_ack(&d40d_fin->txd)) {
+		d40_pool_lli_free(d40d_fin);
+		d40_desc_remove(d40d_fin);
+		/* Return desc to free-list */
+		d40_desc_free(d40c, d40d_fin);
+	} else {
+		d40_desc_reset(d40d_fin);
+		if (!d40d_fin->is_in_client_list) {
+			d40_desc_remove(d40d_fin);
+			list_add_tail(&d40d_fin->node, &d40c->client);
+			d40d_fin->is_in_client_list = true;
+		}
+	}
+
+	d40c->pending_tx--;
+
+	if (d40c->pending_tx)
+		tasklet_schedule(&d40c->tasklet);
+
+	spin_unlock_irqrestore(&d40c->lock, flags);
+
+	if (callback)
+		callback(callback_param);
+
+	return;
+
+ err:
+	/* Rescue manouver if receiving double interrupts */
+	if (d40c->pending_tx > 0)
+		d40c->pending_tx--;
+	spin_unlock_irqrestore(&d40c->lock, flags);
+}
+
+static irqreturn_t d40_handle_interrupt(int irq, void *data)
+{
+	static const struct d40_interrupt_lookup il[] = {
+		{D40_DREG_LCTIS0, D40_DREG_LCICR0, false,  0},
+		{D40_DREG_LCTIS1, D40_DREG_LCICR1, false, 32},
+		{D40_DREG_LCTIS2, D40_DREG_LCICR2, false, 64},
+		{D40_DREG_LCTIS3, D40_DREG_LCICR3, false, 96},
+		{D40_DREG_LCEIS0, D40_DREG_LCICR0, true,   0},
+		{D40_DREG_LCEIS1, D40_DREG_LCICR1, true,  32},
+		{D40_DREG_LCEIS2, D40_DREG_LCICR2, true,  64},
+		{D40_DREG_LCEIS3, D40_DREG_LCICR3, true,  96},
+		{D40_DREG_PCTIS,  D40_DREG_PCICR,  false, D40_PHY_CHAN},
+		{D40_DREG_PCEIS,  D40_DREG_PCICR,  true,  D40_PHY_CHAN},
+	};
+
+	int i;
+	u32 regs[ARRAY_SIZE(il)];
+	u32 tmp;
+	u32 idx;
+	u32 row;
+	long chan = -1;
+	struct d40_chan *d40c;
+	unsigned long flags;
+	struct d40_base *base = data;
+
+	spin_lock_irqsave(&base->interrupt_lock, flags);
+
+	/* Read interrupt status of both logical and physical channels */
+	for (i = 0; i < ARRAY_SIZE(il); i++)
+		regs[i] = readl(base->virtbase + il[i].src);
+
+	for (;;) {
+
+		chan = find_next_bit((unsigned long *)regs,
+				     BITS_PER_LONG * ARRAY_SIZE(il), chan + 1);
+
+		/* No more set bits found? */
+		if (chan == BITS_PER_LONG * ARRAY_SIZE(il))
+			break;
+
+		row = chan / BITS_PER_LONG;
+		idx = chan & (BITS_PER_LONG - 1);
+
+		/* ACK interrupt */
+		tmp = readl(base->virtbase + il[row].clr);
+		tmp |= 1 << idx;
+		writel(tmp, base->virtbase + il[row].clr);
+
+		if (il[row].offset == D40_PHY_CHAN)
+			d40c = base->lookup_phy_chans[idx];
+		else
+			d40c = base->lookup_log_chans[il[row].offset + idx];
+		spin_lock(&d40c->lock);
+
+		if (!il[row].is_error)
+			dma_tc_handle(d40c);
+		else
+			dev_err(base->dev, "[%s] IRQ chan: %ld offset %d idx %d\n",
+				__func__, chan, il[row].offset, idx);
+
+		spin_unlock(&d40c->lock);
+	}
+
+	spin_unlock_irqrestore(&base->interrupt_lock, flags);
+
+	return IRQ_HANDLED;
+}
+
+
+static int d40_validate_conf(struct d40_chan *d40c,
+			     struct stedma40_chan_cfg *conf)
+{
+	int res = 0;
+	u32 dst_event_group = D40_TYPE_TO_GROUP(conf->dst_dev_type);
+	u32 src_event_group = D40_TYPE_TO_GROUP(conf->src_dev_type);
+	bool is_log = (conf->channel_type & STEDMA40_CHANNEL_IN_OPER_MODE)
+		== STEDMA40_CHANNEL_IN_LOG_MODE;
+
+	if (d40c->dma_cfg.dir == STEDMA40_MEM_TO_PERIPH &&
+	    dst_event_group == STEDMA40_DEV_DST_MEMORY) {
+		dev_err(&d40c->chan.dev->device, "[%s] Invalid dst\n",
+			__func__);
+		res = -EINVAL;
+	}
+
+	if (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM &&
+	    src_event_group == STEDMA40_DEV_SRC_MEMORY) {
+		dev_err(&d40c->chan.dev->device, "[%s] Invalid src\n",
+			__func__);
+		res = -EINVAL;
+	}
+
+	if (src_event_group == STEDMA40_DEV_SRC_MEMORY &&
+	    dst_event_group == STEDMA40_DEV_DST_MEMORY && is_log) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] No event line\n", __func__);
+		res = -EINVAL;
+	}
+
+	if (conf->dir == STEDMA40_PERIPH_TO_PERIPH &&
+	    (src_event_group != dst_event_group)) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Invalid event group\n", __func__);
+		res = -EINVAL;
+	}
+
+	if (conf->dir == STEDMA40_PERIPH_TO_PERIPH) {
+		/*
+		 * DMAC HW supports it. Will be added to this driver,
+		 * in case any dma client requires it.
+		 */
+		dev_err(&d40c->chan.dev->device,
+			"[%s] periph to periph not supported\n",
+			__func__);
+		res = -EINVAL;
+	}
+
+	return res;
+}
+
+static bool d40_alloc_mask_set(struct d40_phy_res *phy, bool is_src,
+			       int log_event_line, bool is_log)
+{
+	unsigned long flags;
+	spin_lock_irqsave(&phy->lock, flags);
+	if (!is_log) {
+		/* Physical interrupts are masked per physical full channel */
+		if (phy->allocated_src == D40_ALLOC_FREE &&
+		    phy->allocated_dst == D40_ALLOC_FREE) {
+			phy->allocated_dst = D40_ALLOC_PHY;
+			phy->allocated_src = D40_ALLOC_PHY;
+			goto found;
+		} else
+			goto not_found;
+	}
+
+	/* Logical channel */
+	if (is_src) {
+		if (phy->allocated_src == D40_ALLOC_PHY)
+			goto not_found;
+
+		if (phy->allocated_src == D40_ALLOC_FREE)
+			phy->allocated_src = D40_ALLOC_LOG_FREE;
+
+		if (!(phy->allocated_src & (1 << log_event_line))) {
+			phy->allocated_src |= 1 << log_event_line;
+			goto found;
+		} else
+			goto not_found;
+	} else {
+		if (phy->allocated_dst == D40_ALLOC_PHY)
+			goto not_found;
+
+		if (phy->allocated_dst == D40_ALLOC_FREE)
+			phy->allocated_dst = D40_ALLOC_LOG_FREE;
+
+		if (!(phy->allocated_dst & (1 << log_event_line))) {
+			phy->allocated_dst |= 1 << log_event_line;
+			goto found;
+		} else
+			goto not_found;
+	}
+
+not_found:
+	spin_unlock_irqrestore(&phy->lock, flags);
+	return false;
+found:
+	spin_unlock_irqrestore(&phy->lock, flags);
+	return true;
+}
+
+static bool d40_alloc_mask_free(struct d40_phy_res *phy, bool is_src,
+			       int log_event_line)
+{
+	unsigned long flags;
+	bool is_free = false;
+
+	spin_lock_irqsave(&phy->lock, flags);
+	if (!log_event_line) {
+		/* Physical interrupts are masked per physical full channel */
+		phy->allocated_dst = D40_ALLOC_FREE;
+		phy->allocated_src = D40_ALLOC_FREE;
+		is_free = true;
+		goto out;
+	}
+
+	/* Logical channel */
+	if (is_src) {
+		phy->allocated_src &= ~(1 << log_event_line);
+		if (phy->allocated_src == D40_ALLOC_LOG_FREE)
+			phy->allocated_src = D40_ALLOC_FREE;
+	} else {
+		phy->allocated_dst &= ~(1 << log_event_line);
+		if (phy->allocated_dst == D40_ALLOC_LOG_FREE)
+			phy->allocated_dst = D40_ALLOC_FREE;
+	}
+
+	is_free = ((phy->allocated_src | phy->allocated_dst) ==
+		   D40_ALLOC_FREE);
+
+out:
+	spin_unlock_irqrestore(&phy->lock, flags);
+
+	return is_free;
+}
+
+static int d40_allocate_channel(struct d40_chan *d40c)
+{
+	int dev_type;
+	int event_group;
+	int event_line;
+	struct d40_phy_res *phys;
+	int i;
+	int j;
+	int log_num;
+	bool is_src;
+	bool is_log = (d40c->dma_cfg.channel_type & STEDMA40_CHANNEL_IN_OPER_MODE)
+		== STEDMA40_CHANNEL_IN_LOG_MODE;
+
+
+	phys = d40c->base->phy_res;
+
+	if (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM) {
+		dev_type = d40c->dma_cfg.src_dev_type;
+		log_num = 2 * dev_type;
+		is_src = true;
+	} else if (d40c->dma_cfg.dir == STEDMA40_MEM_TO_PERIPH ||
+		   d40c->dma_cfg.dir == STEDMA40_MEM_TO_MEM) {
+		/* dst event lines are used for logical memcpy */
+		dev_type = d40c->dma_cfg.dst_dev_type;
+		log_num = 2 * dev_type + 1;
+		is_src = false;
+	} else
+		return -EINVAL;
+
+	event_group = D40_TYPE_TO_GROUP(dev_type);
+	event_line = D40_TYPE_TO_EVENT(dev_type);
+
+	if (!is_log) {
+		if (d40c->dma_cfg.dir == STEDMA40_MEM_TO_MEM) {
+			/* Find physical half channel */
+			for (i = 0; i < d40c->base->num_phy_chans; i++) {
+
+				if (d40_alloc_mask_set(&phys[i], is_src,
+						       0, is_log))
+					goto found_phy;
+			}
+		} else
+			for (j = 0; j < d40c->base->num_phy_chans; j += 8) {
+				int phy_num = j  + event_group * 2;
+				for (i = phy_num; i < phy_num + 2; i++) {
+					if (d40_alloc_mask_set(&phys[i], is_src,
+							       0, is_log))
+						goto found_phy;
+				}
+			}
+		return -EINVAL;
+found_phy:
+		d40c->phy_chan = &phys[i];
+		d40c->log_num = D40_PHY_CHAN;
+		goto out;
+	}
+	if (dev_type == -1)
+		return -EINVAL;
+
+	/* Find logical channel */
+	for (j = 0; j < d40c->base->num_phy_chans; j += 8) {
+		int phy_num = j + event_group * 2;
+		/*
+		 * Spread logical channels across all available physical rather
+		 * than pack every logical channel at the first available phy
+		 * channels.
+		 */
+		if (is_src) {
+			for (i = phy_num; i < phy_num + 2; i++) {
+				if (d40_alloc_mask_set(&phys[i], is_src,
+						       event_line, is_log))
+					goto found_log;
+			}
+		} else {
+			for (i = phy_num + 1; i >= phy_num; i--) {
+				if (d40_alloc_mask_set(&phys[i], is_src,
+						       event_line, is_log))
+					goto found_log;
+			}
+		}
+	}
+	return -EINVAL;
+
+found_log:
+	d40c->phy_chan = &phys[i];
+	d40c->log_num = log_num;
+out:
+
+	if (is_log)
+		d40c->base->lookup_log_chans[d40c->log_num] = d40c;
+	else
+		d40c->base->lookup_phy_chans[d40c->phy_chan->num] = d40c;
+
+	return 0;
+
+}
+
+static int d40_config_chan(struct d40_chan *d40c,
+			   struct stedma40_chan_cfg *info)
+{
+
+	/* Fill in basic CFG register values */
+	d40_phy_cfg(&d40c->dma_cfg, &d40c->src_def_cfg,
+		    &d40c->dst_def_cfg, d40c->log_num != D40_PHY_CHAN);
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		d40_log_cfg(&d40c->dma_cfg,
+			    &d40c->log_def.lcsp1, &d40c->log_def.lcsp3);
+
+		if (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM)
+			d40c->lcpa = d40c->base->lcpa_base +
+				d40c->dma_cfg.src_dev_type * 32;
+		else
+			d40c->lcpa = d40c->base->lcpa_base +
+				d40c->dma_cfg.dst_dev_type * 32 + 16;
+	}
+
+	/* Write channel configuration to the DMA */
+	return d40_config_write(d40c);
+}
+
+static int d40_config_memcpy(struct d40_chan *d40c)
+{
+	dma_cap_mask_t cap = d40c->chan.device->cap_mask;
+
+	if (dma_has_cap(DMA_MEMCPY, cap) && !dma_has_cap(DMA_SLAVE, cap)) {
+		d40c->dma_cfg = *d40c->base->plat_data->memcpy_conf_log;
+		d40c->dma_cfg.src_dev_type = STEDMA40_DEV_SRC_MEMORY;
+		d40c->dma_cfg.dst_dev_type = d40c->base->plat_data->
+			memcpy[d40c->chan.chan_id];
+
+	} else if (dma_has_cap(DMA_MEMCPY, cap) &&
+		   dma_has_cap(DMA_SLAVE, cap)) {
+		d40c->dma_cfg = *d40c->base->plat_data->memcpy_conf_phy;
+	} else {
+		dev_err(&d40c->chan.dev->device, "[%s] No memcpy\n",
+			__func__);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+
+static int d40_free_dma(struct d40_chan *d40c)
+{
+
+	int res = 0;
+	u32 event, dir;
+	struct d40_phy_res *phy = d40c->phy_chan;
+	bool is_src;
+
+	/* Terminate all queued and active transfers */
+	d40_term_all(d40c);
+
+	if (phy == NULL) {
+		dev_err(&d40c->chan.dev->device, "[%s] phy == null\n",
+			__func__);
+		return -EINVAL;
+	}
+
+	if (phy->allocated_src == D40_ALLOC_FREE &&
+	    phy->allocated_dst == D40_ALLOC_FREE) {
+		dev_err(&d40c->chan.dev->device, "[%s] channel already free\n",
+			__func__);
+		return -EINVAL;
+	}
+
+
+	res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
+	if (res) {
+		dev_err(&d40c->chan.dev->device, "[%s] suspend\n",
+			__func__);
+		return res;
+	}
+
+	if (d40c->dma_cfg.dir == STEDMA40_MEM_TO_PERIPH ||
+	    d40c->dma_cfg.dir == STEDMA40_MEM_TO_MEM) {
+		event = D40_TYPE_TO_EVENT(d40c->dma_cfg.dst_dev_type);
+		dir = D40_CHAN_REG_SDLNK;
+		is_src = false;
+	} else if (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM) {
+		event = D40_TYPE_TO_EVENT(d40c->dma_cfg.src_dev_type);
+		dir = D40_CHAN_REG_SSLNK;
+		is_src = true;
+	} else {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Unknown direction\n", __func__);
+		return -EINVAL;
+	}
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		/*
+		 * Release logical channel, deactivate the event line during
+		 * the time physical res is suspended.
+		 */
+		writel((D40_DEACTIVATE_EVENTLINE << D40_EVENTLINE_POS(event)) &
+		       D40_EVENTLINE_MASK(event),
+		       d40c->base->virtbase + D40_DREG_PCBASE +
+		       phy->num * D40_DREG_PCDELTA + dir);
+
+		d40c->base->lookup_log_chans[d40c->log_num] = NULL;
+
+		/*
+		 * Check if there are more logical allocation
+		 * on this phy channel.
+		 */
+		if (!d40_alloc_mask_free(phy, is_src, event)) {
+			/* Resume the other logical channels if any */
+			if (d40_chan_has_events(d40c)) {
+				res = d40_channel_execute_command(d40c,
+								  D40_DMA_RUN);
+				if (res) {
+					dev_err(&d40c->chan.dev->device,
+						"[%s] Executing RUN command\n",
+						__func__);
+					return res;
+				}
+			}
+			return 0;
+		}
+	} else
+		d40_alloc_mask_free(phy, is_src, 0);
+
+	/* Release physical channel */
+	res = d40_channel_execute_command(d40c, D40_DMA_STOP);
+	if (res) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Failed to stop channel\n", __func__);
+		return res;
+	}
+	d40c->phy_chan = NULL;
+	/* Invalidate channel type */
+	d40c->dma_cfg.channel_type = 0;
+	d40c->base->lookup_phy_chans[phy->num] = NULL;
+
+	return 0;
+
+
+}
+
+static int d40_pause(struct dma_chan *chan)
+{
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+	int res;
+
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
+	if (res == 0) {
+		if (d40c->log_num != D40_PHY_CHAN) {
+			d40_config_set_event(d40c, false);
+			/* Resume the other logical channels if any */
+			if (d40_chan_has_events(d40c))
+				res = d40_channel_execute_command(d40c,
+								  D40_DMA_RUN);
+		}
+	}
+
+	spin_unlock_irqrestore(&d40c->lock, flags);
+	return res;
+}
+
+static bool d40_is_paused(struct d40_chan *d40c)
+{
+	bool is_paused = false;
+	unsigned long flags;
+	void __iomem *active_reg;
+	u32 status;
+	u32 event;
+	int res;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	if (d40c->log_num == D40_PHY_CHAN) {
+		if (d40c->phy_chan->num % 2 == 0)
+			active_reg = d40c->base->virtbase + D40_DREG_ACTIVE;
+		else
+			active_reg = d40c->base->virtbase + D40_DREG_ACTIVO;
+
+		status = (readl(active_reg) &
+			  D40_CHAN_POS_MASK(d40c->phy_chan->num)) >>
+			D40_CHAN_POS(d40c->phy_chan->num);
+		if (status == D40_DMA_SUSPENDED || status == D40_DMA_STOP)
+			is_paused = true;
+
+		goto _exit;
+	}
+
+	res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
+	if (res != 0)
+		goto _exit;
+
+	if (d40c->dma_cfg.dir == STEDMA40_MEM_TO_PERIPH ||
+	    d40c->dma_cfg.dir == STEDMA40_MEM_TO_MEM)
+		event = D40_TYPE_TO_EVENT(d40c->dma_cfg.dst_dev_type);
+	else if (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM)
+		event = D40_TYPE_TO_EVENT(d40c->dma_cfg.src_dev_type);
+	else {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Unknown direction\n", __func__);
+		goto _exit;
+	}
+	status = d40_chan_has_events(d40c);
+	status = (status & D40_EVENTLINE_MASK(event)) >>
+		D40_EVENTLINE_POS(event);
+
+	if (status != D40_DMA_RUN)
+		is_paused = true;
+
+	/* Resume the other logical channels if any */
+	if (d40_chan_has_events(d40c))
+		res = d40_channel_execute_command(d40c,
+						  D40_DMA_RUN);
+
+_exit:
+	spin_unlock_irqrestore(&d40c->lock, flags);
+	return is_paused;
+
+}
+
+
+static bool d40_tx_is_linked(struct d40_chan *d40c)
+{
+	bool is_link;
+
+	if (d40c->log_num != D40_PHY_CHAN)
+		is_link = readl(&d40c->lcpa->lcsp3) &  D40_MEM_LCSP3_DLOS_MASK;
+	else
+		is_link = readl(d40c->base->virtbase + D40_DREG_PCBASE +
+				d40c->phy_chan->num * D40_DREG_PCDELTA +
+				D40_CHAN_REG_SDLNK) &
+			D40_SREG_LNK_PHYS_LNK_MASK;
+	return is_link;
+}
+
+static u32 d40_residue(struct d40_chan *d40c)
+{
+	u32 num_elt;
+
+	if (d40c->log_num != D40_PHY_CHAN)
+		num_elt = (readl(&d40c->lcpa->lcsp2) &  D40_MEM_LCSP2_ECNT_MASK)
+			>> D40_MEM_LCSP2_ECNT_POS;
+	else
+		num_elt = (readl(d40c->base->virtbase + D40_DREG_PCBASE +
+				 d40c->phy_chan->num * D40_DREG_PCDELTA +
+				 D40_CHAN_REG_SDELT) &
+			   D40_SREG_ELEM_PHY_ECNT_MASK) >> D40_SREG_ELEM_PHY_ECNT_POS;
+	return num_elt * (1 << d40c->dma_cfg.dst_info.data_width);
+}
+
+static int d40_resume(struct dma_chan *chan)
+{
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+	int res = 0;
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
+		if (res)
+			goto out;
+
+		/* If bytes left to transfer or linked tx resume job */
+		if (d40_residue(d40c) || d40_tx_is_linked(d40c)) {
+			d40_config_set_event(d40c, true);
+			res = d40_channel_execute_command(d40c, D40_DMA_RUN);
+		}
+	} else if (d40_residue(d40c) || d40_tx_is_linked(d40c))
+		res = d40_channel_execute_command(d40c, D40_DMA_RUN);
+
+out:
+	spin_unlock_irqrestore(&d40c->lock, flags);
+	return res;
+}
+
+static u32 stedma40_residue(struct dma_chan *chan)
+{
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+	u32 bytes_left;
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+	bytes_left = d40_residue(d40c);
+	spin_unlock_irqrestore(&d40c->lock, flags);
+
+	return bytes_left;
+}
+
+/* Public DMA functions in addition to the DMA engine framework */
+
+int stedma40_set_psize(struct dma_chan *chan,
+		       int src_psize,
+		       int dst_psize)
+{
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		d40c->log_def.lcsp1 &= ~D40_MEM_LCSP1_SCFG_PSIZE_MASK;
+		d40c->log_def.lcsp3 &= ~D40_MEM_LCSP1_SCFG_PSIZE_MASK;
+		d40c->log_def.lcsp1 |= src_psize << D40_MEM_LCSP1_SCFG_PSIZE_POS;
+		d40c->log_def.lcsp3 |= dst_psize << D40_MEM_LCSP1_SCFG_PSIZE_POS;
+		goto out;
+	}
+
+	if (src_psize == STEDMA40_PSIZE_PHY_1)
+		d40c->src_def_cfg &= ~(1 << D40_SREG_CFG_PHY_PEN_POS);
+	else {
+		d40c->src_def_cfg |= 1 << D40_SREG_CFG_PHY_PEN_POS;
+		d40c->src_def_cfg &= ~(STEDMA40_PSIZE_PHY_16 <<
+				       D40_SREG_CFG_PSIZE_POS);
+		d40c->src_def_cfg |= src_psize << D40_SREG_CFG_PSIZE_POS;
+	}
+
+	if (dst_psize == STEDMA40_PSIZE_PHY_1)
+		d40c->dst_def_cfg &= ~(1 << D40_SREG_CFG_PHY_PEN_POS);
+	else {
+		d40c->dst_def_cfg |= 1 << D40_SREG_CFG_PHY_PEN_POS;
+		d40c->dst_def_cfg &= ~(STEDMA40_PSIZE_PHY_16 <<
+				       D40_SREG_CFG_PSIZE_POS);
+		d40c->dst_def_cfg |= dst_psize << D40_SREG_CFG_PSIZE_POS;
+	}
+out:
+	spin_unlock_irqrestore(&d40c->lock, flags);
+	return 0;
+}
+EXPORT_SYMBOL(stedma40_set_psize);
+
+struct dma_async_tx_descriptor *stedma40_memcpy_sg(struct dma_chan *chan,
+						   struct scatterlist *sgl_dst,
+						   struct scatterlist *sgl_src,
+						   unsigned int sgl_len,
+						   unsigned long flags)
+{
+	int res;
+	struct d40_desc *d40d;
+	struct d40_chan *d40c = container_of(chan, struct d40_chan,
+					     chan);
+	unsigned long flg;
+	int lli_max = d40c->base->plat_data->llis_per_log;
+
+
+	spin_lock_irqsave(&d40c->lock, flg);
+	d40d = d40_desc_get(d40c);
+
+	if (d40d == NULL)
+		goto err;
+
+	memset(d40d, 0, sizeof(struct d40_desc));
+	d40d->lli_len = sgl_len;
+
+	d40d->txd.flags = flags;
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		if (sgl_len > 1)
+			/*
+			 * Check if there is space available in lcla. If not,
+			 * split list into 1-length and run only in lcpa
+			 * space.
+			 */
+			if (d40_lcla_id_get(d40c,
+					    &d40c->base->lcla_pool) != 0)
+				lli_max = 1;
+
+		if (d40_pool_lli_alloc(d40d, sgl_len, true) < 0) {
+			dev_err(&d40c->chan.dev->device,
+				"[%s] Out of memory\n", __func__);
+			goto err;
+		}
+
+		(void) d40_log_sg_to_lli(d40c->lcla.src_id,
+					 sgl_src,
+					 sgl_len,
+					 d40d->lli_log.src,
+					 d40c->log_def.lcsp1,
+					 d40c->dma_cfg.src_info.data_width,
+					 flags & DMA_PREP_INTERRUPT, lli_max,
+					 d40c->base->plat_data->llis_per_log);
+
+		(void) d40_log_sg_to_lli(d40c->lcla.dst_id,
+					 sgl_dst,
+					 sgl_len,
+					 d40d->lli_log.dst,
+					 d40c->log_def.lcsp3,
+					 d40c->dma_cfg.dst_info.data_width,
+					 flags & DMA_PREP_INTERRUPT, lli_max,
+					 d40c->base->plat_data->llis_per_log);
+
+
+	} else {
+		if (d40_pool_lli_alloc(d40d, sgl_len, false) < 0) {
+			dev_err(&d40c->chan.dev->device,
+				"[%s] Out of memory\n", __func__);
+			goto err;
+		}
+
+		res = d40_phy_sg_to_lli(sgl_src,
+					sgl_len,
+					0,
+					d40d->lli_phy.src,
+					d40d->lli_phy.src_addr,
+					d40c->src_def_cfg,
+					d40c->dma_cfg.src_info.data_width,
+					d40c->dma_cfg.src_info.psize,
+					true);
+
+		if (res < 0)
+			goto err;
+
+		res = d40_phy_sg_to_lli(sgl_dst,
+					sgl_len,
+					0,
+					d40d->lli_phy.dst,
+					d40d->lli_phy.dst_addr,
+					d40c->dst_def_cfg,
+					d40c->dma_cfg.dst_info.data_width,
+					d40c->dma_cfg.dst_info.psize,
+					true);
+
+		if (res < 0)
+			goto err;
+
+		(void) dma_map_single(d40c->base->dev, d40d->lli_phy.src,
+				      d40d->lli_pool.size, DMA_TO_DEVICE);
+	}
+
+	dma_async_tx_descriptor_init(&d40d->txd, chan);
+
+	d40d->txd.tx_submit = d40_tx_submit;
+
+	spin_unlock_irqrestore(&d40c->lock, flg);
+
+	return &d40d->txd;
+err:
+	spin_unlock_irqrestore(&d40c->lock, flg);
+	return NULL;
+}
+EXPORT_SYMBOL(stedma40_memcpy_sg);
+
+bool stedma40_filter(struct dma_chan *chan, void *data)
+{
+	struct stedma40_chan_cfg *info = data;
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+	int err;
+
+	if (data) {
+		err = d40_validate_conf(d40c, info);
+		if (!err)
+			d40c->dma_cfg = *info;
+	} else
+		err = d40_config_memcpy(d40c);
+
+	return err == 0;
+}
+EXPORT_SYMBOL(stedma40_filter);
+
+/* DMA ENGINE functions */
+static int d40_alloc_chan_resources(struct dma_chan *chan)
+{
+	int err;
+	unsigned long flags;
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	d40c->completed = chan->cookie = 1;
+
+	/*
+	 * If no dma configuration is set (channel_type == 0)
+	 * use default configuration
+	 */
+	if (d40c->dma_cfg.channel_type == 0) {
+		err = d40_config_memcpy(d40c);
+		if (err)
+			goto err_alloc;
+	}
+
+	err = d40_allocate_channel(d40c);
+	if (err) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Failed to allocate channel\n", __func__);
+		goto err_alloc;
+	}
+
+	err = d40_config_chan(d40c, &d40c->dma_cfg);
+	if (err) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Failed to configure channel\n",
+			__func__);
+		goto err_config;
+	}
+
+	spin_unlock_irqrestore(&d40c->lock, flags);
+	return 0;
+
+ err_config:
+	(void) d40_free_dma(d40c);
+ err_alloc:
+	spin_unlock_irqrestore(&d40c->lock, flags);
+	dev_err(&d40c->chan.dev->device,
+		"[%s] Channel allocation failed\n", __func__);
+	return -EINVAL;
+}
+
+static void d40_free_chan_resources(struct dma_chan *chan)
+{
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+	int err;
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	err = d40_free_dma(d40c);
+
+	if (err)
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Failed to free channel\n", __func__);
+	spin_unlock_irqrestore(&d40c->lock, flags);
+}
+
+static struct dma_async_tx_descriptor *d40_prep_memcpy(struct dma_chan *chan,
+						       dma_addr_t dst,
+						       dma_addr_t src,
+						       size_t size,
+						       unsigned long flags)
+{
+	struct d40_desc *d40d;
+	struct d40_chan *d40c = container_of(chan, struct d40_chan,
+					     chan);
+	unsigned long flg;
+	int err = 0;
+
+	spin_lock_irqsave(&d40c->lock, flg);
+	d40d = d40_desc_get(d40c);
+
+	if (d40d == NULL) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Descriptor is NULL\n", __func__);
+		goto err;
+	}
+
+	memset(d40d, 0, sizeof(struct d40_desc));
+
+	d40d->txd.flags = flags;
+
+	dma_async_tx_descriptor_init(&d40d->txd, chan);
+
+	d40d->txd.tx_submit = d40_tx_submit;
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+
+		if (d40_pool_lli_alloc(d40d, 1, true) < 0) {
+			dev_err(&d40c->chan.dev->device,
+				"[%s] Out of memory\n", __func__);
+			goto err;
+		}
+		d40d->lli_len = 1;
+
+		d40_log_fill_lli(d40d->lli_log.src,
+				 src,
+				 size,
+				 0,
+				 d40c->log_def.lcsp1,
+				 d40c->dma_cfg.src_info.data_width,
+				 true, true);
+
+		d40_log_fill_lli(d40d->lli_log.dst,
+				 dst,
+				 size,
+				 0,
+				 d40c->log_def.lcsp3,
+				 d40c->dma_cfg.dst_info.data_width,
+				 true, true);
+
+	} else {
+
+		if (d40_pool_lli_alloc(d40d, 1, false) < 0) {
+			dev_err(&d40c->chan.dev->device,
+				"[%s] Out of memory\n", __func__);
+			goto err;
+		}
+
+		err = d40_phy_fill_lli(d40d->lli_phy.src,
+				       src,
+				       size,
+				       d40c->dma_cfg.src_info.psize,
+				       0,
+				       d40c->src_def_cfg,
+				       true,
+				       d40c->dma_cfg.src_info.data_width,
+				       false);
+		if (err)
+			goto err_fill_lli;
+
+		err = d40_phy_fill_lli(d40d->lli_phy.dst,
+				       dst,
+				       size,
+				       d40c->dma_cfg.dst_info.psize,
+				       0,
+				       d40c->dst_def_cfg,
+				       true,
+				       d40c->dma_cfg.dst_info.data_width,
+				       false);
+
+		if (err)
+			goto err_fill_lli;
+
+		(void) dma_map_single(d40c->base->dev, d40d->lli_phy.src,
+				      d40d->lli_pool.size, DMA_TO_DEVICE);
+	}
+
+	spin_unlock_irqrestore(&d40c->lock, flg);
+	return &d40d->txd;
+
+err_fill_lli:
+	dev_err(&d40c->chan.dev->device,
+		"[%s] Failed filling in PHY LLI\n", __func__);
+	d40_pool_lli_free(d40d);
+err:
+	spin_unlock_irqrestore(&d40c->lock, flg);
+	return NULL;
+}
+
+static int d40_prep_slave_sg_log(struct d40_desc *d40d,
+				 struct d40_chan *d40c,
+				 struct scatterlist *sgl,
+				 unsigned int sg_len,
+				 enum dma_data_direction direction,
+				 unsigned long flags)
+{
+	dma_addr_t dev_addr = 0;
+	int total_size;
+	int lli_max = d40c->base->plat_data->llis_per_log;
+
+	if (d40_pool_lli_alloc(d40d, sg_len, true) < 0) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Out of memory\n", __func__);
+		return -ENOMEM;
+	}
+
+	d40d->lli_len = sg_len;
+	d40d->lli_tcount = 0;
+
+	if (sg_len > 1)
+		/*
+		 * Check if there is space available in lcla.
+		 * If not, split list into 1-length and run only
+		 * in lcpa space.
+		 */
+		if (d40_lcla_id_get(d40c, &d40c->base->lcla_pool) != 0)
+			lli_max = 1;
+
+	if (direction == DMA_FROM_DEVICE) {
+		dev_addr = d40c->base->plat_data->dev_rx[d40c->dma_cfg.src_dev_type];
+		total_size = d40_log_sg_to_dev(&d40c->lcla,
+					       sgl, sg_len,
+					       &d40d->lli_log,
+					       &d40c->log_def,
+					       d40c->dma_cfg.src_info.data_width,
+					       d40c->dma_cfg.dst_info.data_width,
+					       direction,
+					       flags & DMA_PREP_INTERRUPT,
+					       dev_addr, lli_max,
+					       d40c->base->plat_data->llis_per_log);
+	} else if (direction == DMA_TO_DEVICE) {
+		dev_addr = d40c->base->plat_data->dev_tx[d40c->dma_cfg.dst_dev_type];
+		total_size = d40_log_sg_to_dev(&d40c->lcla,
+					       sgl, sg_len,
+					       &d40d->lli_log,
+					       &d40c->log_def,
+					       d40c->dma_cfg.src_info.data_width,
+					       d40c->dma_cfg.dst_info.data_width,
+					       direction,
+					       flags & DMA_PREP_INTERRUPT,
+					       dev_addr, lli_max,
+					       d40c->base->plat_data->llis_per_log);
+	} else
+		return -EINVAL;
+	if (total_size < 0)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int d40_prep_slave_sg_phy(struct d40_desc *d40d,
+				 struct d40_chan *d40c,
+				 struct scatterlist *sgl,
+				 unsigned int sgl_len,
+				 enum dma_data_direction direction,
+				 unsigned long flags)
+{
+	dma_addr_t src_dev_addr;
+	dma_addr_t dst_dev_addr;
+	int res;
+
+	if (d40_pool_lli_alloc(d40d, sgl_len, false) < 0) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Out of memory\n", __func__);
+		return -ENOMEM;
+	}
+
+	d40d->lli_len = sgl_len;
+	d40d->lli_tcount = 0;
+
+	if (direction == DMA_FROM_DEVICE) {
+		dst_dev_addr = 0;
+		src_dev_addr = d40c->base->plat_data->dev_rx[d40c->dma_cfg.src_dev_type];
+	} else if (direction == DMA_TO_DEVICE) {
+		dst_dev_addr = d40c->base->plat_data->dev_tx[d40c->dma_cfg.dst_dev_type];
+		src_dev_addr = 0;
+	} else
+		return -EINVAL;
+
+	res = d40_phy_sg_to_lli(sgl,
+				sgl_len,
+				src_dev_addr,
+				d40d->lli_phy.src,
+				d40d->lli_phy.src_addr,
+				d40c->src_def_cfg,
+				d40c->dma_cfg.src_info.data_width,
+				d40c->dma_cfg.src_info.psize,
+				true);
+	if (res < 0)
+		return res;
+
+	res = d40_phy_sg_to_lli(sgl,
+				sgl_len,
+				dst_dev_addr,
+				d40d->lli_phy.dst,
+				d40d->lli_phy.dst_addr,
+				d40c->dst_def_cfg,
+				d40c->dma_cfg.dst_info.data_width,
+				d40c->dma_cfg.dst_info.psize,
+				 true);
+	if (res < 0)
+		return res;
+
+	(void) dma_map_single(d40c->base->dev, d40d->lli_phy.src,
+			      d40d->lli_pool.size, DMA_TO_DEVICE);
+	return 0;
+}
+
+static struct dma_async_tx_descriptor *d40_prep_slave_sg(struct dma_chan *chan,
+							 struct scatterlist *sgl,
+							 unsigned int sg_len,
+							 enum dma_data_direction direction,
+							 unsigned long flags)
+{
+	struct d40_desc *d40d;
+	struct d40_chan *d40c = container_of(chan, struct d40_chan,
+					     chan);
+	unsigned long flg;
+	int err;
+
+	if (d40c->dma_cfg.pre_transfer)
+		d40c->dma_cfg.pre_transfer(chan,
+					   d40c->dma_cfg.pre_transfer_data,
+					   sg_dma_len(sgl));
+
+	spin_lock_irqsave(&d40c->lock, flg);
+	d40d = d40_desc_get(d40c);
+	spin_unlock_irqrestore(&d40c->lock, flg);
+
+	if (d40d == NULL)
+		return NULL;
+
+	memset(d40d, 0, sizeof(struct d40_desc));
+
+	if (d40c->log_num != D40_PHY_CHAN)
+		err = d40_prep_slave_sg_log(d40d, d40c, sgl, sg_len,
+					    direction, flags);
+	else
+		err = d40_prep_slave_sg_phy(d40d, d40c, sgl, sg_len,
+					    direction, flags);
+	if (err) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Failed to prepare %s slave sg job: %d\n",
+			__func__,
+			d40c->log_num != D40_PHY_CHAN ? "log" : "phy", err);
+		return NULL;
+	}
+
+	d40d->txd.flags = flags;
+
+	dma_async_tx_descriptor_init(&d40d->txd, chan);
+
+	d40d->txd.tx_submit = d40_tx_submit;
+
+	return &d40d->txd;
+}
+
+static enum dma_status d40_tx_status(struct dma_chan *chan,
+				     dma_cookie_t cookie,
+				     struct dma_tx_state *txstate)
+{
+	struct d40_chan *d40c = container_of(chan, struct d40_chan, chan);
+	dma_cookie_t last_used;
+	dma_cookie_t last_complete;
+	int ret;
+
+	last_complete = d40c->completed;
+	last_used = chan->cookie;
+
+	if (d40_is_paused(d40c))
+		ret = DMA_PAUSED;
+	else
+		ret = dma_async_is_complete(cookie, last_complete, last_used);
+
+	dma_set_tx_state(txstate, last_complete, last_used,
+			 stedma40_residue(chan));
+
+	return ret;
+}
+
+static void d40_issue_pending(struct dma_chan *chan)
+{
+	struct d40_chan *d40c = container_of(chan, struct d40_chan, chan);
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	/* Busy means that pending jobs are already being processed */
+	if (!d40c->busy)
+		(void) d40_queue_start(d40c);
+
+	spin_unlock_irqrestore(&d40c->lock, flags);
+}
+
+static int d40_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+		       unsigned long arg)
+{
+	unsigned long flags;
+	struct d40_chan *d40c = container_of(chan, struct d40_chan, chan);
+
+	switch (cmd) {
+	case DMA_TERMINATE_ALL:
+		spin_lock_irqsave(&d40c->lock, flags);
+		d40_term_all(d40c);
+		spin_unlock_irqrestore(&d40c->lock, flags);
+		return 0;
+	case DMA_PAUSE:
+		return d40_pause(chan);
+	case DMA_RESUME:
+		return d40_resume(chan);
+	}
+
+	/* Other commands are unimplemented */
+	return -ENXIO;
+}
+
+/* Initialization functions */
+
+static void __init d40_chan_init(struct d40_base *base, struct dma_device *dma,
+				 struct d40_chan *chans, int offset,
+				 int num_chans)
+{
+	int i = 0;
+	struct d40_chan *d40c;
+
+	INIT_LIST_HEAD(&dma->channels);
+
+	for (i = offset; i < offset + num_chans; i++) {
+		d40c = &chans[i];
+		d40c->base = base;
+		d40c->chan.device = dma;
+
+		/* Invalidate lcla element */
+		d40c->lcla.src_id = -1;
+		d40c->lcla.dst_id = -1;
+
+		spin_lock_init(&d40c->lock);
+
+		d40c->log_num = D40_PHY_CHAN;
+
+		INIT_LIST_HEAD(&d40c->free);
+		INIT_LIST_HEAD(&d40c->active);
+		INIT_LIST_HEAD(&d40c->queue);
+		INIT_LIST_HEAD(&d40c->client);
+
+		d40c->free_len = 0;
+
+		tasklet_init(&d40c->tasklet, dma_tasklet,
+			     (unsigned long) d40c);
+
+		list_add_tail(&d40c->chan.device_node,
+			      &dma->channels);
+	}
+}
+
+static int __init d40_dmaengine_init(struct d40_base *base,
+				     int num_reserved_chans)
+{
+	int err ;
+
+	d40_chan_init(base, &base->dma_slave, base->log_chans,
+		      0, base->num_log_chans);
+
+	dma_cap_zero(base->dma_slave.cap_mask);
+	dma_cap_set(DMA_SLAVE, base->dma_slave.cap_mask);
+
+	base->dma_slave.device_alloc_chan_resources = d40_alloc_chan_resources;
+	base->dma_slave.device_free_chan_resources = d40_free_chan_resources;
+	base->dma_slave.device_prep_dma_memcpy = d40_prep_memcpy;
+	base->dma_slave.device_prep_slave_sg = d40_prep_slave_sg;
+	base->dma_slave.device_tx_status = d40_tx_status;
+	base->dma_slave.device_issue_pending = d40_issue_pending;
+	base->dma_slave.device_control = d40_control;
+	base->dma_slave.dev = base->dev;
+
+	err = dma_async_device_register(&base->dma_slave);
+
+	if (err) {
+		dev_err(base->dev,
+			"[%s] Failed to register slave channels\n",
+			__func__);
+		goto failure1;
+	}
+
+	d40_chan_init(base, &base->dma_memcpy, base->log_chans,
+		      base->num_log_chans, base->plat_data->memcpy_len);
+
+	dma_cap_zero(base->dma_memcpy.cap_mask);
+	dma_cap_set(DMA_MEMCPY, base->dma_memcpy.cap_mask);
+
+	base->dma_memcpy.device_alloc_chan_resources = d40_alloc_chan_resources;
+	base->dma_memcpy.device_free_chan_resources = d40_free_chan_resources;
+	base->dma_memcpy.device_prep_dma_memcpy = d40_prep_memcpy;
+	base->dma_memcpy.device_prep_slave_sg = d40_prep_slave_sg;
+	base->dma_memcpy.device_tx_status = d40_tx_status;
+	base->dma_memcpy.device_issue_pending = d40_issue_pending;
+	base->dma_memcpy.device_control = d40_control;
+	base->dma_memcpy.dev = base->dev;
+	/*
+	 * This controller can only access address at even
+	 * 32bit boundaries, i.e. 2^2
+	 */
+	base->dma_memcpy.copy_align = 2;
+
+	err = dma_async_device_register(&base->dma_memcpy);
+
+	if (err) {
+		dev_err(base->dev,
+			"[%s] Failed to regsiter memcpy only channels\n",
+			__func__);
+		goto failure2;
+	}
+
+	d40_chan_init(base, &base->dma_both, base->phy_chans,
+		      0, num_reserved_chans);
+
+	dma_cap_zero(base->dma_both.cap_mask);
+	dma_cap_set(DMA_SLAVE, base->dma_both.cap_mask);
+	dma_cap_set(DMA_MEMCPY, base->dma_both.cap_mask);
+
+	base->dma_both.device_alloc_chan_resources = d40_alloc_chan_resources;
+	base->dma_both.device_free_chan_resources = d40_free_chan_resources;
+	base->dma_both.device_prep_dma_memcpy = d40_prep_memcpy;
+	base->dma_both.device_prep_slave_sg = d40_prep_slave_sg;
+	base->dma_both.device_tx_status = d40_tx_status;
+	base->dma_both.device_issue_pending = d40_issue_pending;
+	base->dma_both.device_control = d40_control;
+	base->dma_both.dev = base->dev;
+	base->dma_both.copy_align = 2;
+	err = dma_async_device_register(&base->dma_both);
+
+	if (err) {
+		dev_err(base->dev,
+			"[%s] Failed to register logical and physical capable channels\n",
+			__func__);
+		goto failure3;
+	}
+	return 0;
+failure3:
+	dma_async_device_unregister(&base->dma_memcpy);
+failure2:
+	dma_async_device_unregister(&base->dma_slave);
+failure1:
+	return err;
+}
+
+/* Initialization functions. */
+
+static int __init d40_phy_res_init(struct d40_base *base)
+{
+	int i;
+	int num_phy_chans_avail = 0;
+	u32 val[2];
+	int odd_even_bit = -2;
+
+	val[0] = readl(base->virtbase + D40_DREG_PRSME);
+	val[1] = readl(base->virtbase + D40_DREG_PRSMO);
+
+	for (i = 0; i < base->num_phy_chans; i++) {
+		base->phy_res[i].num = i;
+		odd_even_bit += 2 * ((i % 2) == 0);
+		if (((val[i % 2] >> odd_even_bit) & 3) == 1) {
+			/* Mark security only channels as occupied */
+			base->phy_res[i].allocated_src = D40_ALLOC_PHY;
+			base->phy_res[i].allocated_dst = D40_ALLOC_PHY;
+		} else {
+			base->phy_res[i].allocated_src = D40_ALLOC_FREE;
+			base->phy_res[i].allocated_dst = D40_ALLOC_FREE;
+			num_phy_chans_avail++;
+		}
+		spin_lock_init(&base->phy_res[i].lock);
+	}
+	dev_info(base->dev, "%d of %d physical DMA channels available\n",
+		 num_phy_chans_avail, base->num_phy_chans);
+
+	/* Verify settings extended vs standard */
+	val[0] = readl(base->virtbase + D40_DREG_PRTYP);
+
+	for (i = 0; i < base->num_phy_chans; i++) {
+
+		if (base->phy_res[i].allocated_src == D40_ALLOC_FREE &&
+		    (val[0] & 0x3) != 1)
+			dev_info(base->dev,
+				 "[%s] INFO: channel %d is misconfigured (%d)\n",
+				 __func__, i, val[0] & 0x3);
+
+		val[0] = val[0] >> 2;
+	}
+
+	return num_phy_chans_avail;
+}
+
+static struct d40_base * __init d40_hw_detect_init(struct platform_device *pdev)
+{
+	static const struct d40_reg_val dma_id_regs[] = {
+		/* Peripheral Id */
+		{ .reg = D40_DREG_PERIPHID0, .val = 0x0040},
+		{ .reg = D40_DREG_PERIPHID1, .val = 0x0000},
+		/*
+		 * D40_DREG_PERIPHID2 Depends on HW revision:
+		 *  MOP500/HREF ED has 0x0008,
+		 *  ? has 0x0018,
+		 *  HREF V1 has 0x0028
+		 */
+		{ .reg = D40_DREG_PERIPHID3, .val = 0x0000},
+
+		/* PCell Id */
+		{ .reg = D40_DREG_CELLID0, .val = 0x000d},
+		{ .reg = D40_DREG_CELLID1, .val = 0x00f0},
+		{ .reg = D40_DREG_CELLID2, .val = 0x0005},
+		{ .reg = D40_DREG_CELLID3, .val = 0x00b1}
+	};
+	struct stedma40_platform_data *plat_data;
+	struct clk *clk = NULL;
+	void __iomem *virtbase = NULL;
+	struct resource *res = NULL;
+	struct d40_base *base = NULL;
+	int num_log_chans = 0;
+	int num_phy_chans;
+	int i;
+
+	clk = clk_get(&pdev->dev, NULL);
+
+	if (IS_ERR(clk)) {
+		dev_err(&pdev->dev, "[%s] No matching clock found\n",
+			__func__);
+		goto failure;
+	}
+
+	clk_enable(clk);
+
+	/* Get IO for DMAC base address */
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "base");
+	if (!res)
+		goto failure;
+
+	if (request_mem_region(res->start, resource_size(res),
+			       D40_NAME " I/O base") == NULL)
+		goto failure;
+
+	virtbase = ioremap(res->start, resource_size(res));
+	if (!virtbase)
+		goto failure;
+
+	/* HW version check */
+	for (i = 0; i < ARRAY_SIZE(dma_id_regs); i++) {
+		if (dma_id_regs[i].val !=
+		    readl(virtbase + dma_id_regs[i].reg)) {
+			dev_err(&pdev->dev,
+				"[%s] Unknown hardware! Expected 0x%x at 0x%x but got 0x%x\n",
+				__func__,
+				dma_id_regs[i].val,
+				dma_id_regs[i].reg,
+				readl(virtbase + dma_id_regs[i].reg));
+			goto failure;
+		}
+	}
+
+	i = readl(virtbase + D40_DREG_PERIPHID2);
+
+	if ((i & 0xf) != D40_PERIPHID2_DESIGNER) {
+		dev_err(&pdev->dev,
+			"[%s] Unknown designer! Got %x wanted %x\n",
+			__func__, i & 0xf, D40_PERIPHID2_DESIGNER);
+		goto failure;
+	}
+
+	/* The number of physical channels on this HW */
+	num_phy_chans = 4 * (readl(virtbase + D40_DREG_ICFG) & 0x7) + 4;
+
+	dev_info(&pdev->dev, "hardware revision: %d @ 0x%x\n",
+		 (i >> 4) & 0xf, res->start);
+
+	plat_data = pdev->dev.platform_data;
+
+	/* Count the number of logical channels in use */
+	for (i = 0; i < plat_data->dev_len; i++)
+		if (plat_data->dev_rx[i] != 0)
+			num_log_chans++;
+
+	for (i = 0; i < plat_data->dev_len; i++)
+		if (plat_data->dev_tx[i] != 0)
+			num_log_chans++;
+
+	base = kzalloc(ALIGN(sizeof(struct d40_base), 4) +
+		       (num_phy_chans + num_log_chans + plat_data->memcpy_len) *
+		       sizeof(struct d40_chan), GFP_KERNEL);
+
+	if (base == NULL) {
+		dev_err(&pdev->dev, "[%s] Out of memory\n", __func__);
+		goto failure;
+	}
+
+	base->clk = clk;
+	base->num_phy_chans = num_phy_chans;
+	base->num_log_chans = num_log_chans;
+	base->phy_start = res->start;
+	base->phy_size = resource_size(res);
+	base->virtbase = virtbase;
+	base->plat_data = plat_data;
+	base->dev = &pdev->dev;
+	base->phy_chans = ((void *)base) + ALIGN(sizeof(struct d40_base), 4);
+	base->log_chans = &base->phy_chans[num_phy_chans];
+
+	base->phy_res = kzalloc(num_phy_chans * sizeof(struct d40_phy_res),
+				GFP_KERNEL);
+	if (!base->phy_res)
+		goto failure;
+
+	base->lookup_phy_chans = kzalloc(num_phy_chans *
+					 sizeof(struct d40_chan *),
+					 GFP_KERNEL);
+	if (!base->lookup_phy_chans)
+		goto failure;
+
+	if (num_log_chans + plat_data->memcpy_len) {
+		/*
+		 * The max number of logical channels are event lines for all
+		 * src devices and dst devices
+		 */
+		base->lookup_log_chans = kzalloc(plat_data->dev_len * 2 *
+						 sizeof(struct d40_chan *),
+						 GFP_KERNEL);
+		if (!base->lookup_log_chans)
+			goto failure;
+	}
+	base->lcla_pool.alloc_map = kzalloc(num_phy_chans * sizeof(u32),
+					    GFP_KERNEL);
+	if (!base->lcla_pool.alloc_map)
+		goto failure;
+
+	return base;
+
+failure:
+	if (clk) {
+		clk_disable(clk);
+		clk_put(clk);
+	}
+	if (virtbase)
+		iounmap(virtbase);
+	if (res)
+		release_mem_region(res->start,
+				   resource_size(res));
+	if (virtbase)
+		iounmap(virtbase);
+
+	if (base) {
+		kfree(base->lcla_pool.alloc_map);
+		kfree(base->lookup_log_chans);
+		kfree(base->lookup_phy_chans);
+		kfree(base->phy_res);
+		kfree(base);
+	}
+
+	return NULL;
+}
+
+static void __init d40_hw_init(struct d40_base *base)
+{
+
+	static const struct d40_reg_val dma_init_reg[] = {
+		/* Clock every part of the DMA block from start */
+		{ .reg = D40_DREG_GCC,    .val = 0x0000ff01},
+
+		/* Interrupts on all logical channels */
+		{ .reg = D40_DREG_LCMIS0, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCMIS1, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCMIS2, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCMIS3, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCICR0, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCICR1, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCICR2, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCICR3, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCTIS0, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCTIS1, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCTIS2, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCTIS3, .val = 0xFFFFFFFF}
+	};
+	int i;
+	u32 prmseo[2] = {0, 0};
+	u32 activeo[2] = {0xFFFFFFFF, 0xFFFFFFFF};
+	u32 pcmis = 0;
+	u32 pcicr = 0;
+
+	for (i = 0; i < ARRAY_SIZE(dma_init_reg); i++)
+		writel(dma_init_reg[i].val,
+		       base->virtbase + dma_init_reg[i].reg);
+
+	/* Configure all our dma channels to default settings */
+	for (i = 0; i < base->num_phy_chans; i++) {
+
+		activeo[i % 2] = activeo[i % 2] << 2;
+
+		if (base->phy_res[base->num_phy_chans - i - 1].allocated_src
+		    == D40_ALLOC_PHY) {
+			activeo[i % 2] |= 3;
+			continue;
+		}
+
+		/* Enable interrupt # */
+		pcmis = (pcmis << 1) | 1;
+
+		/* Clear interrupt # */
+		pcicr = (pcicr << 1) | 1;
+
+		/* Set channel to physical mode */
+		prmseo[i % 2] = prmseo[i % 2] << 2;
+		prmseo[i % 2] |= 1;
+
+	}
+
+	writel(prmseo[1], base->virtbase + D40_DREG_PRMSE);
+	writel(prmseo[0], base->virtbase + D40_DREG_PRMSO);
+	writel(activeo[1], base->virtbase + D40_DREG_ACTIVE);
+	writel(activeo[0], base->virtbase + D40_DREG_ACTIVO);
+
+	/* Write which interrupt to enable */
+	writel(pcmis, base->virtbase + D40_DREG_PCMIS);
+
+	/* Write which interrupt to clear */
+	writel(pcicr, base->virtbase + D40_DREG_PCICR);
+
+}
+
+static int __init d40_probe(struct platform_device *pdev)
+{
+	int err;
+	int ret = -ENOENT;
+	struct d40_base *base;
+	struct resource *res = NULL;
+	int num_reserved_chans;
+	u32 val;
+
+	base = d40_hw_detect_init(pdev);
+
+	if (!base)
+		goto failure;
+
+	num_reserved_chans = d40_phy_res_init(base);
+
+	platform_set_drvdata(pdev, base);
+
+	spin_lock_init(&base->interrupt_lock);
+	spin_lock_init(&base->execmd_lock);
+
+	/* Get IO for logical channel parameter address */
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "lcpa");
+	if (!res) {
+		ret = -ENOENT;
+		dev_err(&pdev->dev,
+			"[%s] No \"lcpa\" memory resource\n",
+			__func__);
+		goto failure;
+	}
+	base->lcpa_size = resource_size(res);
+	base->phy_lcpa = res->start;
+
+	if (request_mem_region(res->start, resource_size(res),
+			       D40_NAME " I/O lcpa") == NULL) {
+		ret = -EBUSY;
+		dev_err(&pdev->dev,
+			"[%s] Failed to request LCPA region 0x%x-0x%x\n",
+			__func__, res->start, res->end);
+		goto failure;
+	}
+
+	/* We make use of ESRAM memory for this. */
+	val = readl(base->virtbase + D40_DREG_LCPA);
+	if (res->start != val && val != 0) {
+		dev_warn(&pdev->dev,
+			 "[%s] Mismatch LCPA dma 0x%x, def 0x%x\n",
+			 __func__, val, res->start);
+	} else
+		writel(res->start, base->virtbase + D40_DREG_LCPA);
+
+	base->lcpa_base = ioremap(res->start, resource_size(res));
+	if (!base->lcpa_base) {
+		ret = -ENOMEM;
+		dev_err(&pdev->dev,
+			"[%s] Failed to ioremap LCPA region\n",
+			__func__);
+		goto failure;
+	}
+	/* Get IO for logical channel link address */
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "lcla");
+	if (!res) {
+		ret = -ENOENT;
+		dev_err(&pdev->dev,
+			"[%s] No \"lcla\" resource defined\n",
+			__func__);
+		goto failure;
+	}
+
+	base->lcla_pool.base_size = resource_size(res);
+	base->lcla_pool.phy = res->start;
+
+	if (request_mem_region(res->start, resource_size(res),
+			       D40_NAME " I/O lcla") == NULL) {
+		ret = -EBUSY;
+		dev_err(&pdev->dev,
+			"[%s] Failed to request LCLA region 0x%x-0x%x\n",
+			__func__, res->start, res->end);
+		goto failure;
+	}
+	val = readl(base->virtbase + D40_DREG_LCLA);
+	if (res->start != val && val != 0) {
+		dev_warn(&pdev->dev,
+			 "[%s] Mismatch LCLA dma 0x%x, def 0x%x\n",
+			 __func__, val, res->start);
+	} else
+		writel(res->start, base->virtbase + D40_DREG_LCLA);
+
+	base->lcla_pool.base = ioremap(res->start, resource_size(res));
+	if (!base->lcla_pool.base) {
+		ret = -ENOMEM;
+		dev_err(&pdev->dev,
+			"[%s] Failed to ioremap LCLA 0x%x-0x%x\n",
+			__func__, res->start, res->end);
+		goto failure;
+	}
+
+	spin_lock_init(&base->lcla_pool.lock);
+
+	base->lcla_pool.num_blocks = base->num_phy_chans;
+
+	base->irq = platform_get_irq(pdev, 0);
+
+	ret = request_irq(base->irq, d40_handle_interrupt, 0, D40_NAME, base);
+
+	if (ret) {
+		dev_err(&pdev->dev, "[%s] No IRQ defined\n", __func__);
+		goto failure;
+	}
+
+	err = d40_dmaengine_init(base, num_reserved_chans);
+	if (err)
+		goto failure;
+
+	d40_hw_init(base);
+
+	dev_info(base->dev, "initialized\n");
+	return 0;
+
+failure:
+	if (base) {
+		if (base->virtbase)
+			iounmap(base->virtbase);
+		if (base->lcla_pool.phy)
+			release_mem_region(base->lcla_pool.phy,
+					   base->lcla_pool.base_size);
+		if (base->phy_lcpa)
+			release_mem_region(base->phy_lcpa,
+					   base->lcpa_size);
+		if (base->phy_start)
+			release_mem_region(base->phy_start,
+					   base->phy_size);
+		if (base->clk) {
+			clk_disable(base->clk);
+			clk_put(base->clk);
+		}
+
+		kfree(base->lcla_pool.alloc_map);
+		kfree(base->lookup_log_chans);
+		kfree(base->lookup_phy_chans);
+		kfree(base->phy_res);
+		kfree(base);
+	}
+
+	dev_err(&pdev->dev, "[%s] probe failed\n", __func__);
+	return ret;
+}
+
+static struct platform_driver d40_driver = {
+	.driver = {
+		.owner = THIS_MODULE,
+		.name  = D40_NAME,
+	},
+};
+
+int __init stedma40_init(void)
+{
+	return platform_driver_probe(&d40_driver, d40_probe);
+}
+arch_initcall(stedma40_init);
diff --git a/drivers/dma/ste_dma40_ll.c b/drivers/dma/ste_dma40_ll.c
new file mode 100644
index 00000000000..561fdd8a80c
--- /dev/null
+++ b/drivers/dma/ste_dma40_ll.c
@@ -0,0 +1,454 @@
+/*
+ * driver/dma/ste_dma40_ll.c
+ *
+ * Copyright (C) ST-Ericsson 2007-2010
+ * License terms: GNU General Public License (GPL) version 2
+ * Author: Per Friden <per.friden@stericsson.com>
+ * Author: Jonas Aaberg <jonas.aberg@stericsson.com>
+ */
+
+#include <linux/kernel.h>
+#include <plat/ste_dma40.h>
+
+#include "ste_dma40_ll.h"
+
+/* Sets up proper LCSP1 and LCSP3 register for a logical channel */
+void d40_log_cfg(struct stedma40_chan_cfg *cfg,
+		 u32 *lcsp1, u32 *lcsp3)
+{
+	u32 l3 = 0; /* dst */
+	u32 l1 = 0; /* src */
+
+	/* src is mem? -> increase address pos */
+	if (cfg->dir ==  STEDMA40_MEM_TO_PERIPH ||
+	    cfg->dir ==  STEDMA40_MEM_TO_MEM)
+		l1 |= 1 << D40_MEM_LCSP1_SCFG_INCR_POS;
+
+	/* dst is mem? -> increase address pos */
+	if (cfg->dir ==  STEDMA40_PERIPH_TO_MEM ||
+	    cfg->dir ==  STEDMA40_MEM_TO_MEM)
+		l3 |= 1 << D40_MEM_LCSP3_DCFG_INCR_POS;
+
+	/* src is hw? -> master port 1 */
+	if (cfg->dir ==  STEDMA40_PERIPH_TO_MEM ||
+	    cfg->dir ==  STEDMA40_PERIPH_TO_PERIPH)
+		l1 |= 1 << D40_MEM_LCSP1_SCFG_MST_POS;
+
+	/* dst is hw? -> master port 1 */
+	if (cfg->dir ==  STEDMA40_MEM_TO_PERIPH ||
+	    cfg->dir ==  STEDMA40_PERIPH_TO_PERIPH)
+		l3 |= 1 << D40_MEM_LCSP3_DCFG_MST_POS;
+
+	l3 |= 1 << D40_MEM_LCSP3_DCFG_TIM_POS;
+	l3 |= 1 << D40_MEM_LCSP3_DCFG_EIM_POS;
+	l3 |= cfg->dst_info.psize << D40_MEM_LCSP3_DCFG_PSIZE_POS;
+	l3 |= cfg->dst_info.data_width << D40_MEM_LCSP3_DCFG_ESIZE_POS;
+	l3 |= 1 << D40_MEM_LCSP3_DTCP_POS;
+
+	l1 |= 1 << D40_MEM_LCSP1_SCFG_EIM_POS;
+	l1 |= cfg->src_info.psize << D40_MEM_LCSP1_SCFG_PSIZE_POS;
+	l1 |= cfg->src_info.data_width << D40_MEM_LCSP1_SCFG_ESIZE_POS;
+	l1 |= 1 << D40_MEM_LCSP1_STCP_POS;
+
+	*lcsp1 = l1;
+	*lcsp3 = l3;
+
+}
+
+/* Sets up SRC and DST CFG register for both logical and physical channels */
+void d40_phy_cfg(struct stedma40_chan_cfg *cfg,
+		 u32 *src_cfg, u32 *dst_cfg, bool is_log)
+{
+	u32 src = 0;
+	u32 dst = 0;
+
+	if (!is_log) {
+		/* Physical channel */
+		if ((cfg->dir ==  STEDMA40_PERIPH_TO_MEM) ||
+		    (cfg->dir == STEDMA40_PERIPH_TO_PERIPH)) {
+			/* Set master port to 1 */
+			src |= 1 << D40_SREG_CFG_MST_POS;
+			src |= D40_TYPE_TO_EVENT(cfg->src_dev_type);
+
+			if (cfg->src_info.flow_ctrl == STEDMA40_NO_FLOW_CTRL)
+				src |= 1 << D40_SREG_CFG_PHY_TM_POS;
+			else
+				src |= 3 << D40_SREG_CFG_PHY_TM_POS;
+		}
+		if ((cfg->dir ==  STEDMA40_MEM_TO_PERIPH) ||
+		    (cfg->dir == STEDMA40_PERIPH_TO_PERIPH)) {
+			/* Set master port to 1 */
+			dst |= 1 << D40_SREG_CFG_MST_POS;
+			dst |= D40_TYPE_TO_EVENT(cfg->dst_dev_type);
+
+			if (cfg->dst_info.flow_ctrl == STEDMA40_NO_FLOW_CTRL)
+				dst |= 1 << D40_SREG_CFG_PHY_TM_POS;
+			else
+				dst |= 3 << D40_SREG_CFG_PHY_TM_POS;
+		}
+		/* Interrupt on end of transfer for destination */
+		dst |= 1 << D40_SREG_CFG_TIM_POS;
+
+		/* Generate interrupt on error */
+		src |= 1 << D40_SREG_CFG_EIM_POS;
+		dst |= 1 << D40_SREG_CFG_EIM_POS;
+
+		/* PSIZE */
+		if (cfg->src_info.psize != STEDMA40_PSIZE_PHY_1) {
+			src |= 1 << D40_SREG_CFG_PHY_PEN_POS;
+			src |= cfg->src_info.psize << D40_SREG_CFG_PSIZE_POS;
+		}
+		if (cfg->dst_info.psize != STEDMA40_PSIZE_PHY_1) {
+			dst |= 1 << D40_SREG_CFG_PHY_PEN_POS;
+			dst |= cfg->dst_info.psize << D40_SREG_CFG_PSIZE_POS;
+		}
+
+		/* Element size */
+		src |= cfg->src_info.data_width << D40_SREG_CFG_ESIZE_POS;
+		dst |= cfg->dst_info.data_width << D40_SREG_CFG_ESIZE_POS;
+
+	} else {
+		/* Logical channel */
+		dst |= 1 << D40_SREG_CFG_LOG_GIM_POS;
+		src |= 1 << D40_SREG_CFG_LOG_GIM_POS;
+	}
+
+	if (cfg->channel_type & STEDMA40_HIGH_PRIORITY_CHANNEL) {
+		src |= 1 << D40_SREG_CFG_PRI_POS;
+		dst |= 1 << D40_SREG_CFG_PRI_POS;
+	}
+
+	src |= cfg->src_info.endianess << D40_SREG_CFG_LBE_POS;
+	dst |= cfg->dst_info.endianess << D40_SREG_CFG_LBE_POS;
+
+	*src_cfg = src;
+	*dst_cfg = dst;
+}
+
+int d40_phy_fill_lli(struct d40_phy_lli *lli,
+		     dma_addr_t data,
+		     u32 data_size,
+		     int psize,
+		     dma_addr_t next_lli,
+		     u32 reg_cfg,
+		     bool term_int,
+		     u32 data_width,
+		     bool is_device)
+{
+	int num_elems;
+
+	if (psize == STEDMA40_PSIZE_PHY_1)
+		num_elems = 1;
+	else
+		num_elems = 2 << psize;
+
+	/*
+	 * Size is 16bit. data_width is 8, 16, 32 or 64 bit
+	 * Block large than 64 KiB must be split.
+	 */
+	if (data_size > (0xffff << data_width))
+		return -EINVAL;
+
+	/* Must be aligned */
+	if (!IS_ALIGNED(data, 0x1 << data_width))
+		return -EINVAL;
+
+	/* Transfer size can't be smaller than (num_elms * elem_size) */
+	if (data_size < num_elems * (0x1 << data_width))
+		return -EINVAL;
+
+	/* The number of elements. IE now many chunks */
+	lli->reg_elt = (data_size >> data_width) << D40_SREG_ELEM_PHY_ECNT_POS;
+
+	/*
+	 * Distance to next element sized entry.
+	 * Usually the size of the element unless you want gaps.
+	 */
+	if (!is_device)
+		lli->reg_elt |= (0x1 << data_width) <<
+			D40_SREG_ELEM_PHY_EIDX_POS;
+
+	/* Where the data is */
+	lli->reg_ptr = data;
+	lli->reg_cfg = reg_cfg;
+
+	/* If this scatter list entry is the last one, no next link */
+	if (next_lli == 0)
+		lli->reg_lnk = 0x1 << D40_SREG_LNK_PHY_TCP_POS;
+	else
+		lli->reg_lnk = next_lli;
+
+	/* Set/clear interrupt generation on this link item.*/
+	if (term_int)
+		lli->reg_cfg |= 0x1 << D40_SREG_CFG_TIM_POS;
+	else
+		lli->reg_cfg &= ~(0x1 << D40_SREG_CFG_TIM_POS);
+
+	/* Post link */
+	lli->reg_lnk |= 0 << D40_SREG_LNK_PHY_PRE_POS;
+
+	return 0;
+}
+
+int d40_phy_sg_to_lli(struct scatterlist *sg,
+		      int sg_len,
+		      dma_addr_t target,
+		      struct d40_phy_lli *lli,
+		      dma_addr_t lli_phys,
+		      u32 reg_cfg,
+		      u32 data_width,
+		      int psize,
+		      bool term_int)
+{
+	int total_size = 0;
+	int i;
+	struct scatterlist *current_sg = sg;
+	dma_addr_t next_lli_phys;
+	dma_addr_t dst;
+	int err = 0;
+
+	for_each_sg(sg, current_sg, sg_len, i) {
+
+		total_size += sg_dma_len(current_sg);
+
+		/* If this scatter list entry is the last one, no next link */
+		if (sg_len - 1 == i)
+			next_lli_phys = 0;
+		else
+			next_lli_phys = ALIGN(lli_phys + (i + 1) *
+					      sizeof(struct d40_phy_lli),
+					      D40_LLI_ALIGN);
+
+		if (target)
+			dst = target;
+		else
+			dst = sg_phys(current_sg);
+
+		err = d40_phy_fill_lli(&lli[i],
+				       dst,
+				       sg_dma_len(current_sg),
+				       psize,
+				       next_lli_phys,
+				       reg_cfg,
+				       !next_lli_phys,
+				       data_width,
+				       target == dst);
+		if (err)
+			goto err;
+	}
+
+	return total_size;
+ err:
+	return err;
+}
+
+
+void d40_phy_lli_write(void __iomem *virtbase,
+		       u32 phy_chan_num,
+		       struct d40_phy_lli *lli_dst,
+		       struct d40_phy_lli *lli_src)
+{
+
+	writel(lli_src->reg_cfg, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SSCFG);
+	writel(lli_src->reg_elt, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SSELT);
+	writel(lli_src->reg_ptr, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SSPTR);
+	writel(lli_src->reg_lnk, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SSLNK);
+
+	writel(lli_dst->reg_cfg, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SDCFG);
+	writel(lli_dst->reg_elt, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SDELT);
+	writel(lli_dst->reg_ptr, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SDPTR);
+	writel(lli_dst->reg_lnk, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SDLNK);
+
+}
+
+/* DMA logical lli operations */
+
+void d40_log_fill_lli(struct d40_log_lli *lli,
+		      dma_addr_t data, u32 data_size,
+		      u32 lli_next_off, u32 reg_cfg,
+		      u32 data_width,
+		      bool term_int, bool addr_inc)
+{
+	lli->lcsp13 = reg_cfg;
+
+	/* The number of elements to transfer */
+	lli->lcsp02 = ((data_size >> data_width) <<
+		       D40_MEM_LCSP0_ECNT_POS) & D40_MEM_LCSP0_ECNT_MASK;
+	/* 16 LSBs address of the current element */
+	lli->lcsp02 |= data & D40_MEM_LCSP0_SPTR_MASK;
+	/* 16 MSBs address of the current element */
+	lli->lcsp13 |= data & D40_MEM_LCSP1_SPTR_MASK;
+
+	if (addr_inc)
+		lli->lcsp13 |= D40_MEM_LCSP1_SCFG_INCR_MASK;
+
+	lli->lcsp13 |= D40_MEM_LCSP3_DTCP_MASK;
+	/* If this scatter list entry is the last one, no next link */
+	lli->lcsp13 |= (lli_next_off << D40_MEM_LCSP1_SLOS_POS) &
+		D40_MEM_LCSP1_SLOS_MASK;
+
+	if (term_int)
+		lli->lcsp13 |= D40_MEM_LCSP1_SCFG_TIM_MASK;
+	else
+		lli->lcsp13 &= ~D40_MEM_LCSP1_SCFG_TIM_MASK;
+}
+
+int d40_log_sg_to_dev(struct d40_lcla_elem *lcla,
+		      struct scatterlist *sg,
+		      int sg_len,
+		      struct d40_log_lli_bidir *lli,
+		      struct d40_def_lcsp *lcsp,
+		      u32 src_data_width,
+		      u32 dst_data_width,
+		      enum dma_data_direction direction,
+		      bool term_int, dma_addr_t dev_addr, int max_len,
+		      int llis_per_log)
+{
+	int total_size = 0;
+	struct scatterlist *current_sg = sg;
+	int i;
+	u32 next_lli_off_dst;
+	u32 next_lli_off_src;
+
+	next_lli_off_src = 0;
+	next_lli_off_dst = 0;
+
+	for_each_sg(sg, current_sg, sg_len, i) {
+		total_size += sg_dma_len(current_sg);
+
+		/*
+		 * If this scatter list entry is the last one or
+		 * max length, terminate link.
+		 */
+		if (sg_len - 1 == i || ((i+1) % max_len == 0)) {
+			next_lli_off_src = 0;
+			next_lli_off_dst = 0;
+		} else {
+			if (next_lli_off_dst == 0 &&
+			    next_lli_off_src == 0) {
+				/* The first lli will be at next_lli_off */
+				next_lli_off_dst = (lcla->dst_id *
+						    llis_per_log + 1);
+				next_lli_off_src = (lcla->src_id *
+						    llis_per_log + 1);
+			} else {
+				next_lli_off_dst++;
+				next_lli_off_src++;
+			}
+		}
+
+		if (direction == DMA_TO_DEVICE) {
+			d40_log_fill_lli(&lli->src[i],
+					 sg_phys(current_sg),
+					 sg_dma_len(current_sg),
+					 next_lli_off_src,
+					 lcsp->lcsp1, src_data_width,
+					 term_int && !next_lli_off_src,
+					 true);
+			d40_log_fill_lli(&lli->dst[i],
+					 dev_addr,
+					 sg_dma_len(current_sg),
+					 next_lli_off_dst,
+					 lcsp->lcsp3, dst_data_width,
+					 /* No next == terminal interrupt */
+					 term_int && !next_lli_off_dst,
+					 false);
+		} else {
+			d40_log_fill_lli(&lli->dst[i],
+					 sg_phys(current_sg),
+					 sg_dma_len(current_sg),
+					 next_lli_off_dst,
+					 lcsp->lcsp3, dst_data_width,
+					 /* No next == terminal interrupt */
+					 term_int && !next_lli_off_dst,
+					 true);
+			d40_log_fill_lli(&lli->src[i],
+					 dev_addr,
+					 sg_dma_len(current_sg),
+					 next_lli_off_src,
+					 lcsp->lcsp1, src_data_width,
+					 term_int && !next_lli_off_src,
+					 false);
+		}
+	}
+	return total_size;
+}
+
+int d40_log_sg_to_lli(int lcla_id,
+		      struct scatterlist *sg,
+		      int sg_len,
+		      struct d40_log_lli *lli_sg,
+		      u32 lcsp13, /* src or dst*/
+		      u32 data_width,
+		      bool term_int, int max_len, int llis_per_log)
+{
+	int total_size = 0;
+	struct scatterlist *current_sg = sg;
+	int i;
+	u32 next_lli_off = 0;
+
+	for_each_sg(sg, current_sg, sg_len, i) {
+		total_size += sg_dma_len(current_sg);
+
+		/*
+		 * If this scatter list entry is the last one or
+		 * max length, terminate link.
+		 */
+		if (sg_len - 1 == i || ((i+1) % max_len == 0))
+			next_lli_off = 0;
+		else {
+			if (next_lli_off == 0)
+				/* The first lli will be at next_lli_off */
+				next_lli_off = lcla_id * llis_per_log + 1;
+			else
+				next_lli_off++;
+		}
+
+		d40_log_fill_lli(&lli_sg[i],
+				 sg_phys(current_sg),
+				 sg_dma_len(current_sg),
+				 next_lli_off,
+				 lcsp13, data_width,
+				 term_int && !next_lli_off,
+				 true);
+	}
+	return total_size;
+}
+
+void d40_log_lli_write(struct d40_log_lli_full *lcpa,
+		       struct d40_log_lli *lcla_src,
+		       struct d40_log_lli *lcla_dst,
+		       struct d40_log_lli *lli_dst,
+		       struct d40_log_lli *lli_src,
+		       int llis_per_log)
+{
+	u32 slos = 0;
+	u32 dlos = 0;
+	int i;
+
+	lcpa->lcsp0 = lli_src->lcsp02;
+	lcpa->lcsp1 = lli_src->lcsp13;
+	lcpa->lcsp2 = lli_dst->lcsp02;
+	lcpa->lcsp3 = lli_dst->lcsp13;
+
+	slos = lli_src->lcsp13 & D40_MEM_LCSP1_SLOS_MASK;
+	dlos = lli_dst->lcsp13 & D40_MEM_LCSP3_DLOS_MASK;
+
+	for (i = 0; (i < llis_per_log) && slos && dlos; i++) {
+		writel(lli_src[i+1].lcsp02, &lcla_src[i].lcsp02);
+		writel(lli_src[i+1].lcsp13, &lcla_src[i].lcsp13);
+		writel(lli_dst[i+1].lcsp02, &lcla_dst[i].lcsp02);
+		writel(lli_dst[i+1].lcsp13, &lcla_dst[i].lcsp13);
+
+		slos = lli_src[i+1].lcsp13 & D40_MEM_LCSP1_SLOS_MASK;
+		dlos = lli_dst[i+1].lcsp13 & D40_MEM_LCSP3_DLOS_MASK;
+	}
+}
diff --git a/drivers/dma/ste_dma40_ll.h b/drivers/dma/ste_dma40_ll.h
new file mode 100644
index 00000000000..2029280cb33
--- /dev/null
+++ b/drivers/dma/ste_dma40_ll.h
@@ -0,0 +1,354 @@
+/*
+ * driver/dma/ste_dma40_ll.h
+ *
+ * Copyright (C) ST-Ericsson 2007-2010
+ * License terms: GNU General Public License (GPL) version 2
+ * Author: Per Friden <per.friden@stericsson.com>
+ * Author: Jonas Aaberg <jonas.aberg@stericsson.com>
+ */
+#ifndef STE_DMA40_LL_H
+#define STE_DMA40_LL_H
+
+#define D40_DREG_PCBASE		0x400
+#define D40_DREG_PCDELTA	(8 * 4)
+#define D40_LLI_ALIGN		16 /* LLI alignment must be 16 bytes. */
+
+#define D40_TYPE_TO_GROUP(type) (type / 16)
+#define D40_TYPE_TO_EVENT(type) (type % 16)
+
+/* Most bits of the CFG register are the same in log as in phy mode */
+#define D40_SREG_CFG_MST_POS		15
+#define D40_SREG_CFG_TIM_POS		14
+#define D40_SREG_CFG_EIM_POS		13
+#define D40_SREG_CFG_LOG_INCR_POS	12
+#define D40_SREG_CFG_PHY_PEN_POS	12
+#define D40_SREG_CFG_PSIZE_POS		10
+#define D40_SREG_CFG_ESIZE_POS		 8
+#define D40_SREG_CFG_PRI_POS		 7
+#define D40_SREG_CFG_LBE_POS		 6
+#define D40_SREG_CFG_LOG_GIM_POS	 5
+#define D40_SREG_CFG_LOG_MFU_POS	 4
+#define D40_SREG_CFG_PHY_TM_POS		 4
+#define D40_SREG_CFG_PHY_EVTL_POS	 0
+
+
+/* Standard channel parameters - basic mode (element register) */
+#define D40_SREG_ELEM_PHY_ECNT_POS	16
+#define D40_SREG_ELEM_PHY_EIDX_POS	 0
+
+#define D40_SREG_ELEM_PHY_ECNT_MASK	(0xFFFF << D40_SREG_ELEM_PHY_ECNT_POS)
+
+/* Standard channel parameters - basic mode (Link register) */
+#define D40_SREG_LNK_PHY_TCP_POS	0
+#define D40_SREG_LNK_PHY_LMP_POS	1
+#define D40_SREG_LNK_PHY_PRE_POS	2
+/*
+ * Source  destination link address. Contains the
+ * 29-bit byte word aligned address of the reload area.
+ */
+#define D40_SREG_LNK_PHYS_LNK_MASK	0xFFFFFFF8UL
+
+/* Standard basic channel logical mode */
+
+/* Element register */
+#define D40_SREG_ELEM_LOG_ECNT_POS	16
+#define D40_SREG_ELEM_LOG_LIDX_POS	 8
+#define D40_SREG_ELEM_LOG_LOS_POS	 1
+#define D40_SREG_ELEM_LOG_TCP_POS	 0
+
+#define D40_SREG_ELEM_LOG_LIDX_MASK	(0xFF << D40_SREG_ELEM_LOG_LIDX_POS)
+
+/* Link register */
+#define D40_DEACTIVATE_EVENTLINE	0x0
+#define D40_ACTIVATE_EVENTLINE		0x1
+#define D40_EVENTLINE_POS(i)		(2 * i)
+#define D40_EVENTLINE_MASK(i)		(0x3 << D40_EVENTLINE_POS(i))
+
+/* Standard basic channel logical params in memory */
+
+/* LCSP0 */
+#define D40_MEM_LCSP0_ECNT_POS		16
+#define D40_MEM_LCSP0_SPTR_POS		 0
+
+#define D40_MEM_LCSP0_ECNT_MASK		(0xFFFF << D40_MEM_LCSP0_ECNT_POS)
+#define D40_MEM_LCSP0_SPTR_MASK		(0xFFFF << D40_MEM_LCSP0_SPTR_POS)
+
+/* LCSP1 */
+#define D40_MEM_LCSP1_SPTR_POS		16
+#define D40_MEM_LCSP1_SCFG_MST_POS	15
+#define D40_MEM_LCSP1_SCFG_TIM_POS	14
+#define D40_MEM_LCSP1_SCFG_EIM_POS	13
+#define D40_MEM_LCSP1_SCFG_INCR_POS	12
+#define D40_MEM_LCSP1_SCFG_PSIZE_POS	10
+#define D40_MEM_LCSP1_SCFG_ESIZE_POS	 8
+#define D40_MEM_LCSP1_SLOS_POS		 1
+#define D40_MEM_LCSP1_STCP_POS		 0
+
+#define D40_MEM_LCSP1_SPTR_MASK		(0xFFFF << D40_MEM_LCSP1_SPTR_POS)
+#define D40_MEM_LCSP1_SCFG_TIM_MASK	(0x1 << D40_MEM_LCSP1_SCFG_TIM_POS)
+#define D40_MEM_LCSP1_SCFG_INCR_MASK	(0x1 << D40_MEM_LCSP1_SCFG_INCR_POS)
+#define D40_MEM_LCSP1_SCFG_PSIZE_MASK	(0x3 << D40_MEM_LCSP1_SCFG_PSIZE_POS)
+#define D40_MEM_LCSP1_SLOS_MASK		(0x7F << D40_MEM_LCSP1_SLOS_POS)
+#define D40_MEM_LCSP1_STCP_MASK		(0x1 << D40_MEM_LCSP1_STCP_POS)
+
+/* LCSP2 */
+#define D40_MEM_LCSP2_ECNT_POS		16
+
+#define D40_MEM_LCSP2_ECNT_MASK		(0xFFFF << D40_MEM_LCSP2_ECNT_POS)
+
+/* LCSP3 */
+#define D40_MEM_LCSP3_DCFG_MST_POS	15
+#define D40_MEM_LCSP3_DCFG_TIM_POS	14
+#define D40_MEM_LCSP3_DCFG_EIM_POS	13
+#define D40_MEM_LCSP3_DCFG_INCR_POS	12
+#define D40_MEM_LCSP3_DCFG_PSIZE_POS	10
+#define D40_MEM_LCSP3_DCFG_ESIZE_POS	 8
+#define D40_MEM_LCSP3_DLOS_POS		 1
+#define D40_MEM_LCSP3_DTCP_POS		 0
+
+#define D40_MEM_LCSP3_DLOS_MASK		(0x7F << D40_MEM_LCSP3_DLOS_POS)
+#define D40_MEM_LCSP3_DTCP_MASK		(0x1 << D40_MEM_LCSP3_DTCP_POS)
+
+
+/* Standard channel parameter register offsets */
+#define D40_CHAN_REG_SSCFG	0x00
+#define D40_CHAN_REG_SSELT	0x04
+#define D40_CHAN_REG_SSPTR	0x08
+#define D40_CHAN_REG_SSLNK	0x0C
+#define D40_CHAN_REG_SDCFG	0x10
+#define D40_CHAN_REG_SDELT	0x14
+#define D40_CHAN_REG_SDPTR	0x18
+#define D40_CHAN_REG_SDLNK	0x1C
+
+/* DMA Register Offsets */
+#define D40_DREG_GCC		0x000
+#define D40_DREG_PRTYP		0x004
+#define D40_DREG_PRSME		0x008
+#define D40_DREG_PRSMO		0x00C
+#define D40_DREG_PRMSE		0x010
+#define D40_DREG_PRMSO		0x014
+#define D40_DREG_PRMOE		0x018
+#define D40_DREG_PRMOO		0x01C
+#define D40_DREG_LCPA		0x020
+#define D40_DREG_LCLA		0x024
+#define D40_DREG_ACTIVE		0x050
+#define D40_DREG_ACTIVO		0x054
+#define D40_DREG_FSEB1		0x058
+#define D40_DREG_FSEB2		0x05C
+#define D40_DREG_PCMIS		0x060
+#define D40_DREG_PCICR		0x064
+#define D40_DREG_PCTIS		0x068
+#define D40_DREG_PCEIS		0x06C
+#define D40_DREG_LCMIS0		0x080
+#define D40_DREG_LCMIS1		0x084
+#define D40_DREG_LCMIS2		0x088
+#define D40_DREG_LCMIS3		0x08C
+#define D40_DREG_LCICR0		0x090
+#define D40_DREG_LCICR1		0x094
+#define D40_DREG_LCICR2		0x098
+#define D40_DREG_LCICR3		0x09C
+#define D40_DREG_LCTIS0		0x0A0
+#define D40_DREG_LCTIS1		0x0A4
+#define D40_DREG_LCTIS2		0x0A8
+#define D40_DREG_LCTIS3		0x0AC
+#define D40_DREG_LCEIS0		0x0B0
+#define D40_DREG_LCEIS1		0x0B4
+#define D40_DREG_LCEIS2		0x0B8
+#define D40_DREG_LCEIS3		0x0BC
+#define D40_DREG_STFU		0xFC8
+#define D40_DREG_ICFG		0xFCC
+#define D40_DREG_PERIPHID0	0xFE0
+#define D40_DREG_PERIPHID1	0xFE4
+#define D40_DREG_PERIPHID2	0xFE8
+#define D40_DREG_PERIPHID3	0xFEC
+#define D40_DREG_CELLID0	0xFF0
+#define D40_DREG_CELLID1	0xFF4
+#define D40_DREG_CELLID2	0xFF8
+#define D40_DREG_CELLID3	0xFFC
+
+/* LLI related structures */
+
+/**
+ * struct d40_phy_lli - The basic configration register for each physical
+ * channel.
+ *
+ * @reg_cfg: The configuration register.
+ * @reg_elt: The element register.
+ * @reg_ptr: The pointer register.
+ * @reg_lnk: The link register.
+ *
+ * These registers are set up for both physical and logical transfers
+ * Note that the bit in each register means differently in logical and
+ * physical(standard) mode.
+ *
+ * This struct must be 16 bytes aligned, and only contain physical registers
+ * since it will be directly accessed by the DMA.
+ */
+struct d40_phy_lli {
+	u32 reg_cfg;
+	u32 reg_elt;
+	u32 reg_ptr;
+	u32 reg_lnk;
+};
+
+/**
+ * struct d40_phy_lli_bidir - struct for a transfer.
+ *
+ * @src: Register settings for src channel.
+ * @dst: Register settings for dst channel.
+ * @dst_addr: Physical destination address.
+ * @src_addr: Physical source address.
+ *
+ * All DMA transfers have a source and a destination.
+ */
+
+struct d40_phy_lli_bidir {
+	struct d40_phy_lli	*src;
+	struct d40_phy_lli	*dst;
+	dma_addr_t		 dst_addr;
+	dma_addr_t		 src_addr;
+};
+
+
+/**
+ * struct d40_log_lli - logical lli configuration
+ *
+ * @lcsp02: Either maps to register lcsp0 if src or lcsp2 if dst.
+ * @lcsp13: Either maps to register lcsp1 if src or lcsp3 if dst.
+ *
+ * This struct must be 8 bytes aligned since it will be accessed directy by
+ * the DMA. Never add any none hw mapped registers to this struct.
+ */
+
+struct d40_log_lli {
+	u32 lcsp02;
+	u32 lcsp13;
+};
+
+/**
+ * struct d40_log_lli_bidir - For both src and dst
+ *
+ * @src: pointer to src lli configuration.
+ * @dst: pointer to dst lli configuration.
+ *
+ * You always have a src and a dst when doing DMA transfers.
+ */
+
+struct d40_log_lli_bidir {
+	struct d40_log_lli *src;
+	struct d40_log_lli *dst;
+};
+
+/**
+ * struct d40_log_lli_full - LCPA layout
+ *
+ * @lcsp0: Logical Channel Standard Param 0 - Src.
+ * @lcsp1: Logical Channel Standard Param 1 - Src.
+ * @lcsp2: Logical Channel Standard Param 2 - Dst.
+ * @lcsp3: Logical Channel Standard Param 3 - Dst.
+ *
+ * This struct maps to LCPA physical memory layout. Must map to
+ * the hw.
+ */
+struct d40_log_lli_full {
+	u32 lcsp0;
+	u32 lcsp1;
+	u32 lcsp2;
+	u32 lcsp3;
+};
+
+/**
+ * struct d40_def_lcsp - Default LCSP1 and LCSP3 settings
+ *
+ * @lcsp3: The default configuration for dst.
+ * @lcsp1: The default configuration for src.
+ */
+struct d40_def_lcsp {
+	u32 lcsp3;
+	u32 lcsp1;
+};
+
+/**
+ * struct d40_lcla_elem - Info for one LCA element.
+ *
+ * @src_id: logical channel src id
+ * @dst_id: logical channel dst id
+ * @src: LCPA formated src parameters
+ * @dst: LCPA formated dst parameters
+ *
+ */
+struct d40_lcla_elem {
+	int			src_id;
+	int			dst_id;
+	struct d40_log_lli     *src;
+	struct d40_log_lli     *dst;
+};
+
+/* Physical channels */
+
+void d40_phy_cfg(struct stedma40_chan_cfg *cfg,
+		 u32 *src_cfg, u32 *dst_cfg, bool is_log);
+
+void d40_log_cfg(struct stedma40_chan_cfg *cfg,
+		 u32 *lcsp1, u32 *lcsp2);
+
+int d40_phy_sg_to_lli(struct scatterlist *sg,
+		      int sg_len,
+		      dma_addr_t target,
+		      struct d40_phy_lli *lli,
+		      dma_addr_t lli_phys,
+		      u32 reg_cfg,
+		      u32 data_width,
+		      int psize,
+		      bool term_int);
+
+int d40_phy_fill_lli(struct d40_phy_lli *lli,
+		     dma_addr_t data,
+		     u32 data_size,
+		     int psize,
+		     dma_addr_t next_lli,
+		     u32 reg_cfg,
+		     bool term_int,
+		     u32 data_width,
+		     bool is_device);
+
+void d40_phy_lli_write(void __iomem *virtbase,
+		       u32 phy_chan_num,
+		       struct d40_phy_lli *lli_dst,
+		       struct d40_phy_lli *lli_src);
+
+/* Logical channels */
+
+void d40_log_fill_lli(struct d40_log_lli *lli,
+		      dma_addr_t data, u32 data_size,
+		      u32 lli_next_off, u32 reg_cfg,
+		      u32 data_width,
+		      bool term_int, bool addr_inc);
+
+int d40_log_sg_to_dev(struct d40_lcla_elem *lcla,
+		      struct scatterlist *sg,
+		      int sg_len,
+		      struct d40_log_lli_bidir *lli,
+		      struct d40_def_lcsp *lcsp,
+		      u32 src_data_width,
+		      u32 dst_data_width,
+		      enum dma_data_direction direction,
+		      bool term_int, dma_addr_t dev_addr, int max_len,
+		      int llis_per_log);
+
+void d40_log_lli_write(struct d40_log_lli_full *lcpa,
+		       struct d40_log_lli *lcla_src,
+		       struct d40_log_lli *lcla_dst,
+		       struct d40_log_lli *lli_dst,
+		       struct d40_log_lli *lli_src,
+		       int llis_per_log);
+
+int d40_log_sg_to_lli(int lcla_id,
+		      struct scatterlist *sg,
+		      int sg_len,
+		      struct d40_log_lli *lli_sg,
+		      u32 lcsp13, /* src or dst*/
+		      u32 data_width,
+		      bool term_int, int max_len, int llis_per_log);
+
+#endif /* STE_DMA40_LLI_H */
diff --git a/drivers/dma/timb_dma.c b/drivers/dma/timb_dma.c
new file mode 100644
index 00000000000..0172fa3c7a2
--- /dev/null
+++ b/drivers/dma/timb_dma.c
@@ -0,0 +1,860 @@
+/*
+ * timb_dma.c timberdale FPGA DMA driver
+ * Copyright (c) 2010 Intel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+/* Supports:
+ * Timberdale FPGA DMA engine
+ */
+
+#include <linux/dmaengine.h>
+#include <linux/dma-mapping.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+
+#include <linux/timb_dma.h>
+
+#define DRIVER_NAME "timb-dma"
+
+/* Global DMA registers */
+#define TIMBDMA_ACR		0x34
+#define TIMBDMA_32BIT_ADDR	0x01
+
+#define TIMBDMA_ISR		0x080000
+#define TIMBDMA_IPR		0x080004
+#define TIMBDMA_IER		0x080008
+
+/* Channel specific registers */
+/* RX instances base addresses are 0x00, 0x40, 0x80 ...
+ * TX instances base addresses are 0x18, 0x58, 0x98 ...
+ */
+#define TIMBDMA_INSTANCE_OFFSET		0x40
+#define TIMBDMA_INSTANCE_TX_OFFSET	0x18
+
+/* RX registers, relative the instance base */
+#define TIMBDMA_OFFS_RX_DHAR	0x00
+#define TIMBDMA_OFFS_RX_DLAR	0x04
+#define TIMBDMA_OFFS_RX_LR	0x0C
+#define TIMBDMA_OFFS_RX_BLR	0x10
+#define TIMBDMA_OFFS_RX_ER	0x14
+#define TIMBDMA_RX_EN		0x01
+/* bytes per Row, video specific register
+ * which is placed after the TX registers...
+ */
+#define TIMBDMA_OFFS_RX_BPRR	0x30
+
+/* TX registers, relative the instance base */
+#define TIMBDMA_OFFS_TX_DHAR	0x00
+#define TIMBDMA_OFFS_TX_DLAR	0x04
+#define TIMBDMA_OFFS_TX_BLR	0x0C
+#define TIMBDMA_OFFS_TX_LR	0x14
+
+
+#define TIMB_DMA_DESC_SIZE	8
+
+struct timb_dma_desc {
+	struct list_head		desc_node;
+	struct dma_async_tx_descriptor	txd;
+	u8				*desc_list;
+	unsigned int			desc_list_len;
+	bool				interrupt;
+};
+
+struct timb_dma_chan {
+	struct dma_chan		chan;
+	void __iomem		*membase;
+	spinlock_t		lock; /* Used to protect data structures,
+					especially the lists and descriptors,
+					from races between the tasklet and calls
+					from above */
+	dma_cookie_t		last_completed_cookie;
+	bool			ongoing;
+	struct list_head	active_list;
+	struct list_head	queue;
+	struct list_head	free_list;
+	unsigned int		bytes_per_line;
+	enum dma_data_direction	direction;
+	unsigned int		descs; /* Descriptors to allocate */
+	unsigned int		desc_elems; /* number of elems per descriptor */
+};
+
+struct timb_dma {
+	struct dma_device	dma;
+	void __iomem		*membase;
+	struct tasklet_struct	tasklet;
+	struct timb_dma_chan	channels[0];
+};
+
+static struct device *chan2dev(struct dma_chan *chan)
+{
+	return &chan->dev->device;
+}
+static struct device *chan2dmadev(struct dma_chan *chan)
+{
+	return chan2dev(chan)->parent->parent;
+}
+
+static struct timb_dma *tdchantotd(struct timb_dma_chan *td_chan)
+{
+	int id = td_chan->chan.chan_id;
+	return (struct timb_dma *)((u8 *)td_chan -
+		id * sizeof(struct timb_dma_chan) - sizeof(struct timb_dma));
+}
+
+/* Must be called with the spinlock held */
+static void __td_enable_chan_irq(struct timb_dma_chan *td_chan)
+{
+	int id = td_chan->chan.chan_id;
+	struct timb_dma *td = tdchantotd(td_chan);
+	u32 ier;
+
+	/* enable interrupt for this channel */
+	ier = ioread32(td->membase + TIMBDMA_IER);
+	ier |= 1 << id;
+	dev_dbg(chan2dev(&td_chan->chan), "Enabling irq: %d, IER: 0x%x\n", id,
+		ier);
+	iowrite32(ier, td->membase + TIMBDMA_IER);
+}
+
+/* Should be called with the spinlock held */
+static bool __td_dma_done_ack(struct timb_dma_chan *td_chan)
+{
+	int id = td_chan->chan.chan_id;
+	struct timb_dma *td = (struct timb_dma *)((u8 *)td_chan -
+		id * sizeof(struct timb_dma_chan) - sizeof(struct timb_dma));
+	u32 isr;
+	bool done = false;
+
+	dev_dbg(chan2dev(&td_chan->chan), "Checking irq: %d, td: %p\n", id, td);
+
+	isr = ioread32(td->membase + TIMBDMA_ISR) & (1 << id);
+	if (isr) {
+		iowrite32(isr, td->membase + TIMBDMA_ISR);
+		done = true;
+	}
+
+	return done;
+}
+
+static void __td_unmap_desc(struct timb_dma_chan *td_chan, const u8 *dma_desc,
+	bool single)
+{
+	dma_addr_t addr;
+	int len;
+
+	addr = (dma_desc[7] << 24) | (dma_desc[6] << 16) | (dma_desc[5] << 8) |
+		dma_desc[4];
+
+	len = (dma_desc[3] << 8) | dma_desc[2];
+
+	if (single)
+		dma_unmap_single(chan2dev(&td_chan->chan), addr, len,
+			td_chan->direction);
+	else
+		dma_unmap_page(chan2dev(&td_chan->chan), addr, len,
+			td_chan->direction);
+}
+
+static void __td_unmap_descs(struct timb_dma_desc *td_desc, bool single)
+{
+	struct timb_dma_chan *td_chan = container_of(td_desc->txd.chan,
+		struct timb_dma_chan, chan);
+	u8 *descs;
+
+	for (descs = td_desc->desc_list; ; descs += TIMB_DMA_DESC_SIZE) {
+		__td_unmap_desc(td_chan, descs, single);
+		if (descs[0] & 0x02)
+			break;
+	}
+}
+
+static int td_fill_desc(struct timb_dma_chan *td_chan, u8 *dma_desc,
+	struct scatterlist *sg, bool last)
+{
+	if (sg_dma_len(sg) > USHORT_MAX) {
+		dev_err(chan2dev(&td_chan->chan), "Too big sg element\n");
+		return -EINVAL;
+	}
+
+	/* length must be word aligned */
+	if (sg_dma_len(sg) % sizeof(u32)) {
+		dev_err(chan2dev(&td_chan->chan), "Incorrect length: %d\n",
+			sg_dma_len(sg));
+		return -EINVAL;
+	}
+
+	dev_dbg(chan2dev(&td_chan->chan), "desc: %p, addr: %p\n",
+		dma_desc, (void *)sg_dma_address(sg));
+
+	dma_desc[7] = (sg_dma_address(sg) >> 24) & 0xff;
+	dma_desc[6] = (sg_dma_address(sg) >> 16) & 0xff;
+	dma_desc[5] = (sg_dma_address(sg) >> 8) & 0xff;
+	dma_desc[4] = (sg_dma_address(sg) >> 0) & 0xff;
+
+	dma_desc[3] = (sg_dma_len(sg) >> 8) & 0xff;
+	dma_desc[2] = (sg_dma_len(sg) >> 0) & 0xff;
+
+	dma_desc[1] = 0x00;
+	dma_desc[0] = 0x21 | (last ? 0x02 : 0); /* tran, valid */
+
+	return 0;
+}
+
+/* Must be called with the spinlock held */
+static void __td_start_dma(struct timb_dma_chan *td_chan)
+{
+	struct timb_dma_desc *td_desc;
+
+	if (td_chan->ongoing) {
+		dev_err(chan2dev(&td_chan->chan),
+			"Transfer already ongoing\n");
+		return;
+	}
+
+	td_desc = list_entry(td_chan->active_list.next, struct timb_dma_desc,
+		desc_node);
+
+	dev_dbg(chan2dev(&td_chan->chan),
+		"td_chan: %p, chan: %d, membase: %p\n",
+		td_chan, td_chan->chan.chan_id, td_chan->membase);
+
+	if (td_chan->direction == DMA_FROM_DEVICE) {
+
+		/* descriptor address */
+		iowrite32(0, td_chan->membase + TIMBDMA_OFFS_RX_DHAR);
+		iowrite32(td_desc->txd.phys, td_chan->membase +
+			TIMBDMA_OFFS_RX_DLAR);
+		/* Bytes per line */
+		iowrite32(td_chan->bytes_per_line, td_chan->membase +
+			TIMBDMA_OFFS_RX_BPRR);
+		/* enable RX */
+		iowrite32(TIMBDMA_RX_EN, td_chan->membase + TIMBDMA_OFFS_RX_ER);
+	} else {
+		/* address high */
+		iowrite32(0, td_chan->membase + TIMBDMA_OFFS_TX_DHAR);
+		iowrite32(td_desc->txd.phys, td_chan->membase +
+			TIMBDMA_OFFS_TX_DLAR);
+	}
+
+	td_chan->ongoing = true;
+
+	if (td_desc->interrupt)
+		__td_enable_chan_irq(td_chan);
+}
+
+static void __td_finish(struct timb_dma_chan *td_chan)
+{
+	dma_async_tx_callback		callback;
+	void				*param;
+	struct dma_async_tx_descriptor	*txd;
+	struct timb_dma_desc		*td_desc;
+
+	/* can happen if the descriptor is canceled */
+	if (list_empty(&td_chan->active_list))
+		return;
+
+	td_desc = list_entry(td_chan->active_list.next, struct timb_dma_desc,
+		desc_node);
+	txd = &td_desc->txd;
+
+	dev_dbg(chan2dev(&td_chan->chan), "descriptor %u complete\n",
+		txd->cookie);
+
+	/* make sure to stop the transfer */
+	if (td_chan->direction == DMA_FROM_DEVICE)
+		iowrite32(0, td_chan->membase + TIMBDMA_OFFS_RX_ER);
+/* Currently no support for stopping DMA transfers
+	else
+		iowrite32(0, td_chan->membase + TIMBDMA_OFFS_TX_DLAR);
+*/
+	td_chan->last_completed_cookie = txd->cookie;
+	td_chan->ongoing = false;
+
+	callback = txd->callback;
+	param = txd->callback_param;
+
+	list_move(&td_desc->desc_node, &td_chan->free_list);
+
+	if (!(txd->flags & DMA_COMPL_SKIP_SRC_UNMAP))
+		__td_unmap_descs(td_desc,
+			txd->flags & DMA_COMPL_SRC_UNMAP_SINGLE);
+
+	/*
+	 * The API requires that no submissions are done from a
+	 * callback, so we don't need to drop the lock here
+	 */
+	if (callback)
+		callback(param);
+}
+
+static u32 __td_ier_mask(struct timb_dma *td)
+{
+	int i;
+	u32 ret = 0;
+
+	for (i = 0; i < td->dma.chancnt; i++) {
+		struct timb_dma_chan *td_chan = td->channels + i;
+		if (td_chan->ongoing) {
+			struct timb_dma_desc *td_desc =
+				list_entry(td_chan->active_list.next,
+				struct timb_dma_desc, desc_node);
+			if (td_desc->interrupt)
+				ret |= 1 << i;
+		}
+	}
+
+	return ret;
+}
+
+static void __td_start_next(struct timb_dma_chan *td_chan)
+{
+	struct timb_dma_desc *td_desc;
+
+	BUG_ON(list_empty(&td_chan->queue));
+	BUG_ON(td_chan->ongoing);
+
+	td_desc = list_entry(td_chan->queue.next, struct timb_dma_desc,
+		desc_node);
+
+	dev_dbg(chan2dev(&td_chan->chan), "%s: started %u\n",
+		__func__, td_desc->txd.cookie);
+
+	list_move(&td_desc->desc_node, &td_chan->active_list);
+	__td_start_dma(td_chan);
+}
+
+static dma_cookie_t td_tx_submit(struct dma_async_tx_descriptor *txd)
+{
+	struct timb_dma_desc *td_desc = container_of(txd, struct timb_dma_desc,
+		txd);
+	struct timb_dma_chan *td_chan = container_of(txd->chan,
+		struct timb_dma_chan, chan);
+	dma_cookie_t cookie;
+
+	spin_lock_bh(&td_chan->lock);
+
+	cookie = txd->chan->cookie;
+	if (++cookie < 0)
+		cookie = 1;
+	txd->chan->cookie = cookie;
+	txd->cookie = cookie;
+
+	if (list_empty(&td_chan->active_list)) {
+		dev_dbg(chan2dev(txd->chan), "%s: started %u\n", __func__,
+			txd->cookie);
+		list_add_tail(&td_desc->desc_node, &td_chan->active_list);
+		__td_start_dma(td_chan);
+	} else {
+		dev_dbg(chan2dev(txd->chan), "tx_submit: queued %u\n",
+			txd->cookie);
+
+		list_add_tail(&td_desc->desc_node, &td_chan->queue);
+	}
+
+	spin_unlock_bh(&td_chan->lock);
+
+	return cookie;
+}
+
+static struct timb_dma_desc *td_alloc_init_desc(struct timb_dma_chan *td_chan)
+{
+	struct dma_chan *chan = &td_chan->chan;
+	struct timb_dma_desc *td_desc;
+	int err;
+
+	td_desc = kzalloc(sizeof(struct timb_dma_desc), GFP_KERNEL);
+	if (!td_desc) {
+		dev_err(chan2dev(chan), "Failed to alloc descriptor\n");
+		goto err;
+	}
+
+	td_desc->desc_list_len = td_chan->desc_elems * TIMB_DMA_DESC_SIZE;
+
+	td_desc->desc_list = kzalloc(td_desc->desc_list_len, GFP_KERNEL);
+	if (!td_desc->desc_list) {
+		dev_err(chan2dev(chan), "Failed to alloc descriptor\n");
+		goto err;
+	}
+
+	dma_async_tx_descriptor_init(&td_desc->txd, chan);
+	td_desc->txd.tx_submit = td_tx_submit;
+	td_desc->txd.flags = DMA_CTRL_ACK;
+
+	td_desc->txd.phys = dma_map_single(chan2dmadev(chan),
+		td_desc->desc_list, td_desc->desc_list_len, DMA_TO_DEVICE);
+
+	err = dma_mapping_error(chan2dmadev(chan), td_desc->txd.phys);
+	if (err) {
+		dev_err(chan2dev(chan), "DMA mapping error: %d\n", err);
+		goto err;
+	}
+
+	return td_desc;
+err:
+	kfree(td_desc->desc_list);
+	kfree(td_desc);
+
+	return NULL;
+
+}
+
+static void td_free_desc(struct timb_dma_desc *td_desc)
+{
+	dev_dbg(chan2dev(td_desc->txd.chan), "Freeing desc: %p\n", td_desc);
+	dma_unmap_single(chan2dmadev(td_desc->txd.chan), td_desc->txd.phys,
+		td_desc->desc_list_len, DMA_TO_DEVICE);
+
+	kfree(td_desc->desc_list);
+	kfree(td_desc);
+}
+
+static void td_desc_put(struct timb_dma_chan *td_chan,
+	struct timb_dma_desc *td_desc)
+{
+	dev_dbg(chan2dev(&td_chan->chan), "Putting desc: %p\n", td_desc);
+
+	spin_lock_bh(&td_chan->lock);
+	list_add(&td_desc->desc_node, &td_chan->free_list);
+	spin_unlock_bh(&td_chan->lock);
+}
+
+static struct timb_dma_desc *td_desc_get(struct timb_dma_chan *td_chan)
+{
+	struct timb_dma_desc *td_desc, *_td_desc;
+	struct timb_dma_desc *ret = NULL;
+
+	spin_lock_bh(&td_chan->lock);
+	list_for_each_entry_safe(td_desc, _td_desc, &td_chan->free_list,
+		desc_node) {
+		if (async_tx_test_ack(&td_desc->txd)) {
+			list_del(&td_desc->desc_node);
+			ret = td_desc;
+			break;
+		}
+		dev_dbg(chan2dev(&td_chan->chan), "desc %p not ACKed\n",
+			td_desc);
+	}
+	spin_unlock_bh(&td_chan->lock);
+
+	return ret;
+}
+
+static int td_alloc_chan_resources(struct dma_chan *chan)
+{
+	struct timb_dma_chan *td_chan =
+		container_of(chan, struct timb_dma_chan, chan);
+	int i;
+
+	dev_dbg(chan2dev(chan), "%s: entry\n", __func__);
+
+	BUG_ON(!list_empty(&td_chan->free_list));
+	for (i = 0; i < td_chan->descs; i++) {
+		struct timb_dma_desc *td_desc = td_alloc_init_desc(td_chan);
+		if (!td_desc) {
+			if (i)
+				break;
+			else {
+				dev_err(chan2dev(chan),
+					"Couldnt allocate any descriptors\n");
+				return -ENOMEM;
+			}
+		}
+
+		td_desc_put(td_chan, td_desc);
+	}
+
+	spin_lock_bh(&td_chan->lock);
+	td_chan->last_completed_cookie = 1;
+	chan->cookie = 1;
+	spin_unlock_bh(&td_chan->lock);
+
+	return 0;
+}
+
+static void td_free_chan_resources(struct dma_chan *chan)
+{
+	struct timb_dma_chan *td_chan =
+		container_of(chan, struct timb_dma_chan, chan);
+	struct timb_dma_desc *td_desc, *_td_desc;
+	LIST_HEAD(list);
+
+	dev_dbg(chan2dev(chan), "%s: Entry\n", __func__);
+
+	/* check that all descriptors are free */
+	BUG_ON(!list_empty(&td_chan->active_list));
+	BUG_ON(!list_empty(&td_chan->queue));
+
+	spin_lock_bh(&td_chan->lock);
+	list_splice_init(&td_chan->free_list, &list);
+	spin_unlock_bh(&td_chan->lock);
+
+	list_for_each_entry_safe(td_desc, _td_desc, &list, desc_node) {
+		dev_dbg(chan2dev(chan), "%s: Freeing desc: %p\n", __func__,
+			td_desc);
+		td_free_desc(td_desc);
+	}
+}
+
+static enum dma_status td_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
+				    struct dma_tx_state *txstate)
+{
+	struct timb_dma_chan *td_chan =
+		container_of(chan, struct timb_dma_chan, chan);
+	dma_cookie_t		last_used;
+	dma_cookie_t		last_complete;
+	int			ret;
+
+	dev_dbg(chan2dev(chan), "%s: Entry\n", __func__);
+
+	last_complete = td_chan->last_completed_cookie;
+	last_used = chan->cookie;
+
+	ret = dma_async_is_complete(cookie, last_complete, last_used);
+
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
+
+	dev_dbg(chan2dev(chan),
+		"%s: exit, ret: %d, last_complete: %d, last_used: %d\n",
+		__func__, ret, last_complete, last_used);
+
+	return ret;
+}
+
+static void td_issue_pending(struct dma_chan *chan)
+{
+	struct timb_dma_chan *td_chan =
+		container_of(chan, struct timb_dma_chan, chan);
+
+	dev_dbg(chan2dev(chan), "%s: Entry\n", __func__);
+	spin_lock_bh(&td_chan->lock);
+
+	if (!list_empty(&td_chan->active_list))
+		/* transfer ongoing */
+		if (__td_dma_done_ack(td_chan))
+			__td_finish(td_chan);
+
+	if (list_empty(&td_chan->active_list) && !list_empty(&td_chan->queue))
+		__td_start_next(td_chan);
+
+	spin_unlock_bh(&td_chan->lock);
+}
+
+static struct dma_async_tx_descriptor *td_prep_slave_sg(struct dma_chan *chan,
+	struct scatterlist *sgl, unsigned int sg_len,
+	enum dma_data_direction direction, unsigned long flags)
+{
+	struct timb_dma_chan *td_chan =
+		container_of(chan, struct timb_dma_chan, chan);
+	struct timb_dma_desc *td_desc;
+	struct scatterlist *sg;
+	unsigned int i;
+	unsigned int desc_usage = 0;
+
+	if (!sgl || !sg_len) {
+		dev_err(chan2dev(chan), "%s: No SG list\n", __func__);
+		return NULL;
+	}
+
+	/* even channels are for RX, odd for TX */
+	if (td_chan->direction != direction) {
+		dev_err(chan2dev(chan),
+			"Requesting channel in wrong direction\n");
+		return NULL;
+	}
+
+	td_desc = td_desc_get(td_chan);
+	if (!td_desc) {
+		dev_err(chan2dev(chan), "Not enough descriptors available\n");
+		return NULL;
+	}
+
+	td_desc->interrupt = (flags & DMA_PREP_INTERRUPT) != 0;
+
+	for_each_sg(sgl, sg, sg_len, i) {
+		int err;
+		if (desc_usage > td_desc->desc_list_len) {
+			dev_err(chan2dev(chan), "No descriptor space\n");
+			return NULL;
+		}
+
+		err = td_fill_desc(td_chan, td_desc->desc_list + desc_usage, sg,
+			i == (sg_len - 1));
+		if (err) {
+			dev_err(chan2dev(chan), "Failed to update desc: %d\n",
+				err);
+			td_desc_put(td_chan, td_desc);
+			return NULL;
+		}
+		desc_usage += TIMB_DMA_DESC_SIZE;
+	}
+
+	dma_sync_single_for_device(chan2dmadev(chan), td_desc->txd.phys,
+		td_desc->desc_list_len, DMA_TO_DEVICE);
+
+	return &td_desc->txd;
+}
+
+static int td_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+		      unsigned long arg)
+{
+	struct timb_dma_chan *td_chan =
+		container_of(chan, struct timb_dma_chan, chan);
+	struct timb_dma_desc *td_desc, *_td_desc;
+
+	dev_dbg(chan2dev(chan), "%s: Entry\n", __func__);
+
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
+	/* first the easy part, put the queue into the free list */
+	spin_lock_bh(&td_chan->lock);
+	list_for_each_entry_safe(td_desc, _td_desc, &td_chan->queue,
+		desc_node)
+		list_move(&td_desc->desc_node, &td_chan->free_list);
+
+	/* now tear down the runnning */
+	__td_finish(td_chan);
+	spin_unlock_bh(&td_chan->lock);
+
+	return 0;
+}
+
+static void td_tasklet(unsigned long data)
+{
+	struct timb_dma *td = (struct timb_dma *)data;
+	u32 isr;
+	u32 ipr;
+	u32 ier;
+	int i;
+
+	isr = ioread32(td->membase + TIMBDMA_ISR);
+	ipr = isr & __td_ier_mask(td);
+
+	/* ack the interrupts */
+	iowrite32(ipr, td->membase + TIMBDMA_ISR);
+
+	for (i = 0; i < td->dma.chancnt; i++)
+		if (ipr & (1 << i)) {
+			struct timb_dma_chan *td_chan = td->channels + i;
+			spin_lock(&td_chan->lock);
+			__td_finish(td_chan);
+			if (!list_empty(&td_chan->queue))
+				__td_start_next(td_chan);
+			spin_unlock(&td_chan->lock);
+		}
+
+	ier = __td_ier_mask(td);
+	iowrite32(ier, td->membase + TIMBDMA_IER);
+}
+
+
+static irqreturn_t td_irq(int irq, void *devid)
+{
+	struct timb_dma *td = devid;
+	u32 ipr = ioread32(td->membase + TIMBDMA_IPR);
+
+	if (ipr) {
+		/* disable interrupts, will be re-enabled in tasklet */
+		iowrite32(0, td->membase + TIMBDMA_IER);
+
+		tasklet_schedule(&td->tasklet);
+
+		return IRQ_HANDLED;
+	} else
+		return IRQ_NONE;
+}
+
+
+static int __devinit td_probe(struct platform_device *pdev)
+{
+	struct timb_dma_platform_data *pdata = pdev->dev.platform_data;
+	struct timb_dma *td;
+	struct resource *iomem;
+	int irq;
+	int err;
+	int i;
+
+	if (!pdata) {
+		dev_err(&pdev->dev, "No platform data\n");
+		return -EINVAL;
+	}
+
+	iomem = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!iomem)
+		return -EINVAL;
+
+	irq = platform_get_irq(pdev, 0);
+	if (irq < 0)
+		return irq;
+
+	if (!request_mem_region(iomem->start, resource_size(iomem),
+		DRIVER_NAME))
+		return -EBUSY;
+
+	td  = kzalloc(sizeof(struct timb_dma) +
+		sizeof(struct timb_dma_chan) * pdata->nr_channels, GFP_KERNEL);
+	if (!td) {
+		err = -ENOMEM;
+		goto err_release_region;
+	}
+
+	dev_dbg(&pdev->dev, "Allocated TD: %p\n", td);
+
+	td->membase = ioremap(iomem->start, resource_size(iomem));
+	if (!td->membase) {
+		dev_err(&pdev->dev, "Failed to remap I/O memory\n");
+		err = -ENOMEM;
+		goto err_free_mem;
+	}
+
+	/* 32bit addressing */
+	iowrite32(TIMBDMA_32BIT_ADDR, td->membase + TIMBDMA_ACR);
+
+	/* disable and clear any interrupts */
+	iowrite32(0x0, td->membase + TIMBDMA_IER);
+	iowrite32(0xFFFFFFFF, td->membase + TIMBDMA_ISR);
+
+	tasklet_init(&td->tasklet, td_tasklet, (unsigned long)td);
+
+	err = request_irq(irq, td_irq, IRQF_SHARED, DRIVER_NAME, td);
+	if (err) {
+		dev_err(&pdev->dev, "Failed to request IRQ\n");
+		goto err_tasklet_kill;
+	}
+
+	td->dma.device_alloc_chan_resources	= td_alloc_chan_resources;
+	td->dma.device_free_chan_resources	= td_free_chan_resources;
+	td->dma.device_tx_status		= td_tx_status;
+	td->dma.device_issue_pending		= td_issue_pending;
+
+	dma_cap_set(DMA_SLAVE, td->dma.cap_mask);
+	dma_cap_set(DMA_PRIVATE, td->dma.cap_mask);
+	td->dma.device_prep_slave_sg = td_prep_slave_sg;
+	td->dma.device_control = td_control;
+
+	td->dma.dev = &pdev->dev;
+
+	INIT_LIST_HEAD(&td->dma.channels);
+
+	for (i = 0; i < pdata->nr_channels; i++, td->dma.chancnt++) {
+		struct timb_dma_chan *td_chan = &td->channels[i];
+		struct timb_dma_platform_data_channel *pchan =
+			pdata->channels + i;
+
+		/* even channels are RX, odd are TX */
+		if (((i % 2) && pchan->rx) || (!(i % 2) && !pchan->rx)) {
+			dev_err(&pdev->dev, "Wrong channel configuration\n");
+			err = -EINVAL;
+			goto err_tasklet_kill;
+		}
+
+		td_chan->chan.device = &td->dma;
+		td_chan->chan.cookie = 1;
+		td_chan->chan.chan_id = i;
+		spin_lock_init(&td_chan->lock);
+		INIT_LIST_HEAD(&td_chan->active_list);
+		INIT_LIST_HEAD(&td_chan->queue);
+		INIT_LIST_HEAD(&td_chan->free_list);
+
+		td_chan->descs = pchan->descriptors;
+		td_chan->desc_elems = pchan->descriptor_elements;
+		td_chan->bytes_per_line = pchan->bytes_per_line;
+		td_chan->direction = pchan->rx ? DMA_FROM_DEVICE :
+			DMA_TO_DEVICE;
+
+		td_chan->membase = td->membase +
+			(i / 2) * TIMBDMA_INSTANCE_OFFSET +
+			(pchan->rx ? 0 : TIMBDMA_INSTANCE_TX_OFFSET);
+
+		dev_dbg(&pdev->dev, "Chan: %d, membase: %p\n",
+			i, td_chan->membase);
+
+		list_add_tail(&td_chan->chan.device_node, &td->dma.channels);
+	}
+
+	err = dma_async_device_register(&td->dma);
+	if (err) {
+		dev_err(&pdev->dev, "Failed to register async device\n");
+		goto err_free_irq;
+	}
+
+	platform_set_drvdata(pdev, td);
+
+	dev_dbg(&pdev->dev, "Probe result: %d\n", err);
+	return err;
+
+err_free_irq:
+	free_irq(irq, td);
+err_tasklet_kill:
+	tasklet_kill(&td->tasklet);
+	iounmap(td->membase);
+err_free_mem:
+	kfree(td);
+err_release_region:
+	release_mem_region(iomem->start, resource_size(iomem));
+
+	return err;
+
+}
+
+static int __devexit td_remove(struct platform_device *pdev)
+{
+	struct timb_dma *td = platform_get_drvdata(pdev);
+	struct resource *iomem = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	int irq = platform_get_irq(pdev, 0);
+
+	dma_async_device_unregister(&td->dma);
+	free_irq(irq, td);
+	tasklet_kill(&td->tasklet);
+	iounmap(td->membase);
+	kfree(td);
+	release_mem_region(iomem->start, resource_size(iomem));
+
+	platform_set_drvdata(pdev, NULL);
+
+	dev_dbg(&pdev->dev, "Removed...\n");
+	return 0;
+}
+
+static struct platform_driver td_driver = {
+	.driver = {
+		.name	= DRIVER_NAME,
+		.owner  = THIS_MODULE,
+	},
+	.probe	= td_probe,
+	.remove	= __exit_p(td_remove),
+};
+
+static int __init td_init(void)
+{
+	return platform_driver_register(&td_driver);
+}
+module_init(td_init);
+
+static void __exit td_exit(void)
+{
+	platform_driver_unregister(&td_driver);
+}
+module_exit(td_exit);
+
+MODULE_LICENSE("GPL v2");
+MODULE_DESCRIPTION("Timberdale DMA controller driver");
+MODULE_AUTHOR("Pelagicore AB <info@pelagicore.com>");
+MODULE_ALIAS("platform:"DRIVER_NAME);
diff --git a/drivers/dma/txx9dmac.c b/drivers/dma/txx9dmac.c
index 75fcf1ac8bb..cbd83e362b5 100644
--- a/drivers/dma/txx9dmac.c
+++ b/drivers/dma/txx9dmac.c
@@ -938,12 +938,17 @@ txx9dmac_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 	return &first->txd;
 }
 
-static void txx9dmac_terminate_all(struct dma_chan *chan)
+static int txx9dmac_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+			    unsigned long arg)
 {
 	struct txx9dmac_chan *dc = to_txx9dmac_chan(chan);
 	struct txx9dmac_desc *desc, *_desc;
 	LIST_HEAD(list);
 
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -EINVAL;
+
 	dev_vdbg(chan2dev(chan), "terminate_all\n");
 	spin_lock_bh(&dc->lock);
 
@@ -958,12 +963,13 @@ static void txx9dmac_terminate_all(struct dma_chan *chan)
 	/* Flush all pending and queued descriptors */
 	list_for_each_entry_safe(desc, _desc, &list, desc_node)
 		txx9dmac_descriptor_complete(dc, desc);
+
+	return 0;
 }
 
 static enum dma_status
-txx9dmac_is_tx_complete(struct dma_chan *chan,
-			dma_cookie_t cookie,
-		dma_cookie_t *done, dma_cookie_t *used)
+txx9dmac_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
+		   struct dma_tx_state *txstate)
 {
 	struct txx9dmac_chan *dc = to_txx9dmac_chan(chan);
 	dma_cookie_t last_used;
@@ -985,10 +991,7 @@ txx9dmac_is_tx_complete(struct dma_chan *chan,
 		ret = dma_async_is_complete(cookie, last_complete, last_used);
 	}
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	return ret;
 }
@@ -1153,8 +1156,8 @@ static int __init txx9dmac_chan_probe(struct platform_device *pdev)
 	dc->dma.dev = &pdev->dev;
 	dc->dma.device_alloc_chan_resources = txx9dmac_alloc_chan_resources;
 	dc->dma.device_free_chan_resources = txx9dmac_free_chan_resources;
-	dc->dma.device_terminate_all = txx9dmac_terminate_all;
-	dc->dma.device_is_tx_complete = txx9dmac_is_tx_complete;
+	dc->dma.device_control = txx9dmac_control;
+	dc->dma.device_tx_status = txx9dmac_tx_status;
 	dc->dma.device_issue_pending = txx9dmac_issue_pending;
 	if (pdata && pdata->memcpy_chan == ch) {
 		dc->dma.device_prep_dma_memcpy = txx9dmac_prep_dma_memcpy;