Merge branch 'master' of /home/davem/src/GIT/linux-2.6/

24 files changed, 7729 insertions, 211 deletions

diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index eb140ff38c27..e02d74b1e892 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -111,6 +111,24 @@ config SH_DMAE
 	help
 	  Enable support for the Renesas SuperH DMA controllers.
 
+config COH901318
+	bool "ST-Ericsson COH901318 DMA support"
+	select DMA_ENGINE
+	depends on ARCH_U300
+	help
+	  Enable support for ST-Ericsson COH 901 318 DMA.
+
+config AMCC_PPC440SPE_ADMA
+	tristate "AMCC PPC440SPe ADMA support"
+	depends on 440SPe || 440SP
+	select DMA_ENGINE
+	select ARCH_HAS_ASYNC_TX_FIND_CHANNEL
+	help
+	  Enable support for the AMCC PPC440SPe RAID engines.
+
+config ARCH_HAS_ASYNC_TX_FIND_CHANNEL
+	bool
+
 config DMA_ENGINE
 	bool
 
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index eca71ba78ae9..807053d48232 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -10,3 +10,5 @@ obj-$(CONFIG_AT_HDMAC) += at_hdmac.o
 obj-$(CONFIG_MX3_IPU) += ipu/
 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/
diff --git a/drivers/dma/at_hdmac.c b/drivers/dma/at_hdmac.c
index c52ac9efd0bf..efc1a61ca231 100644
--- a/drivers/dma/at_hdmac.c
+++ b/drivers/dma/at_hdmac.c
@@ -815,7 +815,7 @@ atc_is_tx_complete(struct dma_chan *chan,
 	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);
+	spin_lock_bh(&atchan->lock);
 
 	last_complete = atchan->completed_cookie;
 	last_used = chan->cookie;
@@ -830,7 +830,7 @@ atc_is_tx_complete(struct dma_chan *chan,
 		ret = dma_async_is_complete(cookie, last_complete, last_used);
 	}
 
-	spin_unlock_bh(atchan->lock);
+	spin_unlock_bh(&atchan->lock);
 
 	if (done)
 		*done = last_complete;
@@ -1188,7 +1188,7 @@ static int at_dma_resume_noirq(struct device *dev)
 	return 0;
 }
 
-static struct dev_pm_ops at_dma_dev_pm_ops = {
+static const struct dev_pm_ops at_dma_dev_pm_ops = {
 	.suspend_noirq = at_dma_suspend_noirq,
 	.resume_noirq = at_dma_resume_noirq,
 };
diff --git a/drivers/dma/coh901318.c b/drivers/dma/coh901318.c
new file mode 100644
index 000000000000..b5f2ee0f8e2c
--- /dev/null
+++ b/drivers/dma/coh901318.c
@@ -0,0 +1,1325 @@
+/*
+ * driver/dma/coh901318.c
+ *
+ * Copyright (C) 2007-2009 ST-Ericsson
+ * License terms: GNU General Public License (GPL) version 2
+ * DMA driver for COH 901 318
+ * Author: Per Friden <per.friden@stericsson.com>
+ */
+
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/kernel.h> /* printk() */
+#include <linux/fs.h> /* everything... */
+#include <linux/slab.h> /* kmalloc() */
+#include <linux/dmaengine.h>
+#include <linux/platform_device.h>
+#include <linux/device.h>
+#include <linux/irqreturn.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/uaccess.h>
+#include <linux/debugfs.h>
+#include <mach/coh901318.h>
+
+#include "coh901318_lli.h"
+
+#define COHC_2_DEV(cohc) (&cohc->chan.dev->device)
+
+#ifdef VERBOSE_DEBUG
+#define COH_DBG(x) ({ if (1) x; 0; })
+#else
+#define COH_DBG(x) ({ if (0) x; 0; })
+#endif
+
+struct coh901318_desc {
+	struct dma_async_tx_descriptor desc;
+	struct list_head node;
+	struct scatterlist *sg;
+	unsigned int sg_len;
+	struct coh901318_lli *data;
+	enum dma_data_direction dir;
+	int pending_irqs;
+	unsigned long flags;
+};
+
+struct coh901318_base {
+	struct device *dev;
+	void __iomem *virtbase;
+	struct coh901318_pool pool;
+	struct powersave pm;
+	struct dma_device dma_slave;
+	struct dma_device dma_memcpy;
+	struct coh901318_chan *chans;
+	struct coh901318_platform *platform;
+};
+
+struct coh901318_chan {
+	spinlock_t lock;
+	int allocated;
+	int completed;
+	int id;
+	int stopped;
+
+	struct work_struct free_work;
+	struct dma_chan chan;
+
+	struct tasklet_struct tasklet;
+
+	struct list_head active;
+	struct list_head queue;
+	struct list_head free;
+
+	unsigned long nbr_active_done;
+	unsigned long busy;
+	int pending_irqs;
+
+	struct coh901318_base *base;
+};
+
+static void coh901318_list_print(struct coh901318_chan *cohc,
+				 struct coh901318_lli *lli)
+{
+	struct coh901318_lli *l;
+	dma_addr_t addr =  virt_to_phys(lli);
+	int i = 0;
+
+	while (addr) {
+		l = phys_to_virt(addr);
+		dev_vdbg(COHC_2_DEV(cohc), "i %d, lli %p, ctrl 0x%x, src 0x%x"
+			 ", dst 0x%x, link 0x%x link_virt 0x%p\n",
+			 i, l, l->control, l->src_addr, l->dst_addr,
+			 l->link_addr, phys_to_virt(l->link_addr));
+		i++;
+		addr = l->link_addr;
+	}
+}
+
+#ifdef CONFIG_DEBUG_FS
+
+#define COH901318_DEBUGFS_ASSIGN(x, y) (x = y)
+
+static struct coh901318_base *debugfs_dma_base;
+static struct dentry *dma_dentry;
+
+static int coh901318_debugfs_open(struct inode *inode, struct file *file)
+{
+
+	file->private_data = inode->i_private;
+	return 0;
+}
+
+static int coh901318_debugfs_read(struct file *file, char __user *buf,
+				  size_t count, loff_t *f_pos)
+{
+	u64 started_channels = debugfs_dma_base->pm.started_channels;
+	int pool_count = debugfs_dma_base->pool.debugfs_pool_counter;
+	int i;
+	int ret = 0;
+	char *dev_buf;
+	char *tmp;
+	int dev_size;
+
+	dev_buf = kmalloc(4*1024, GFP_KERNEL);
+	if (dev_buf == NULL)
+		goto err_kmalloc;
+	tmp = dev_buf;
+
+	tmp += sprintf(tmp, "DMA -- enable dma channels\n");
+
+	for (i = 0; i < debugfs_dma_base->platform->max_channels; i++)
+		if (started_channels & (1 << i))
+			tmp += sprintf(tmp, "channel %d\n", i);
+
+	tmp += sprintf(tmp, "Pool alloc nbr %d\n", pool_count);
+	dev_size = tmp  - dev_buf;
+
+	/* No more to read if offset != 0 */
+	if (*f_pos > dev_size)
+		goto out;
+
+	if (count > dev_size - *f_pos)
+		count = dev_size - *f_pos;
+
+	if (copy_to_user(buf, dev_buf + *f_pos, count))
+		ret = -EINVAL;
+	ret = count;
+	*f_pos += count;
+
+ out:
+	kfree(dev_buf);
+	return ret;
+
+ err_kmalloc:
+	return 0;
+}
+
+static const struct file_operations coh901318_debugfs_status_operations = {
+	.owner		= THIS_MODULE,
+	.open		= coh901318_debugfs_open,
+	.read		= coh901318_debugfs_read,
+};
+
+
+static int __init init_coh901318_debugfs(void)
+{
+
+	dma_dentry = debugfs_create_dir("dma", NULL);
+
+	(void) debugfs_create_file("status",
+				   S_IFREG | S_IRUGO,
+				   dma_dentry, NULL,
+				   &coh901318_debugfs_status_operations);
+	return 0;
+}
+
+static void __exit exit_coh901318_debugfs(void)
+{
+	debugfs_remove_recursive(dma_dentry);
+}
+
+module_init(init_coh901318_debugfs);
+module_exit(exit_coh901318_debugfs);
+#else
+
+#define COH901318_DEBUGFS_ASSIGN(x, y)
+
+#endif /* CONFIG_DEBUG_FS */
+
+static inline struct coh901318_chan *to_coh901318_chan(struct dma_chan *chan)
+{
+	return container_of(chan, struct coh901318_chan, chan);
+}
+
+static inline dma_addr_t
+cohc_dev_addr(struct coh901318_chan *cohc)
+{
+	return cohc->base->platform->chan_conf[cohc->id].dev_addr;
+}
+
+static inline const struct coh901318_params *
+cohc_chan_param(struct coh901318_chan *cohc)
+{
+	return &cohc->base->platform->chan_conf[cohc->id].param;
+}
+
+static inline const struct coh_dma_channel *
+cohc_chan_conf(struct coh901318_chan *cohc)
+{
+	return &cohc->base->platform->chan_conf[cohc->id];
+}
+
+static void enable_powersave(struct coh901318_chan *cohc)
+{
+	unsigned long flags;
+	struct powersave *pm = &cohc->base->pm;
+
+	spin_lock_irqsave(&pm->lock, flags);
+
+	pm->started_channels &= ~(1ULL << cohc->id);
+
+	if (!pm->started_channels) {
+		/* DMA no longer intends to access memory */
+		cohc->base->platform->access_memory_state(cohc->base->dev,
+							  false);
+	}
+
+	spin_unlock_irqrestore(&pm->lock, flags);
+}
+static void disable_powersave(struct coh901318_chan *cohc)
+{
+	unsigned long flags;
+	struct powersave *pm = &cohc->base->pm;
+
+	spin_lock_irqsave(&pm->lock, flags);
+
+	if (!pm->started_channels) {
+		/* DMA intends to access memory */
+		cohc->base->platform->access_memory_state(cohc->base->dev,
+							  true);
+	}
+
+	pm->started_channels |= (1ULL << cohc->id);
+
+	spin_unlock_irqrestore(&pm->lock, flags);
+}
+
+static inline int coh901318_set_ctrl(struct coh901318_chan *cohc, u32 control)
+{
+	int channel = cohc->id;
+	void __iomem *virtbase = cohc->base->virtbase;
+
+	writel(control,
+	       virtbase + COH901318_CX_CTRL +
+	       COH901318_CX_CTRL_SPACING * channel);
+	return 0;
+}
+
+static inline int coh901318_set_conf(struct coh901318_chan *cohc, u32 conf)
+{
+	int channel = cohc->id;
+	void __iomem *virtbase = cohc->base->virtbase;
+
+	writel(conf,
+	       virtbase + COH901318_CX_CFG +
+	       COH901318_CX_CFG_SPACING*channel);
+	return 0;
+}
+
+
+static int coh901318_start(struct coh901318_chan *cohc)
+{
+	u32 val;
+	int channel = cohc->id;
+	void __iomem *virtbase = cohc->base->virtbase;
+
+	disable_powersave(cohc);
+
+	val = readl(virtbase + COH901318_CX_CFG +
+		    COH901318_CX_CFG_SPACING * channel);
+
+	/* Enable channel */
+	val |= COH901318_CX_CFG_CH_ENABLE;
+	writel(val, virtbase + COH901318_CX_CFG +
+	       COH901318_CX_CFG_SPACING * channel);
+
+	return 0;
+}
+
+static int coh901318_prep_linked_list(struct coh901318_chan *cohc,
+				      struct coh901318_lli *data)
+{
+	int channel = cohc->id;
+	void __iomem *virtbase = cohc->base->virtbase;
+
+	BUG_ON(readl(virtbase + COH901318_CX_STAT +
+		     COH901318_CX_STAT_SPACING*channel) &
+	       COH901318_CX_STAT_ACTIVE);
+
+	writel(data->src_addr,
+	       virtbase + COH901318_CX_SRC_ADDR +
+	       COH901318_CX_SRC_ADDR_SPACING * channel);
+
+	writel(data->dst_addr, virtbase +
+	       COH901318_CX_DST_ADDR +
+	       COH901318_CX_DST_ADDR_SPACING * channel);
+
+	writel(data->link_addr, virtbase + COH901318_CX_LNK_ADDR +
+	       COH901318_CX_LNK_ADDR_SPACING * channel);
+
+	writel(data->control, virtbase + COH901318_CX_CTRL +
+	       COH901318_CX_CTRL_SPACING * channel);
+
+	return 0;
+}
+static dma_cookie_t
+coh901318_assign_cookie(struct coh901318_chan *cohc,
+			struct coh901318_desc *cohd)
+{
+	dma_cookie_t cookie = cohc->chan.cookie;
+
+	if (++cookie < 0)
+		cookie = 1;
+
+	cohc->chan.cookie = cookie;
+	cohd->desc.cookie = cookie;
+
+	return cookie;
+}
+
+static struct coh901318_desc *
+coh901318_desc_get(struct coh901318_chan *cohc)
+{
+	struct coh901318_desc *desc;
+
+	if (list_empty(&cohc->free)) {
+		/* alloc new desc because we're out of used ones
+		 * TODO: alloc a pile of descs instead of just one,
+		 * avoid many small allocations.
+		 */
+		desc = kmalloc(sizeof(struct coh901318_desc), GFP_NOWAIT);
+		if (desc == NULL)
+			goto out;
+		INIT_LIST_HEAD(&desc->node);
+	} else {
+		/* Reuse an old desc. */
+		desc = list_first_entry(&cohc->free,
+					struct coh901318_desc,
+					node);
+		list_del(&desc->node);
+	}
+
+ out:
+	return desc;
+}
+
+static void
+coh901318_desc_free(struct coh901318_chan *cohc, struct coh901318_desc *cohd)
+{
+	list_add_tail(&cohd->node, &cohc->free);
+}
+
+/* call with irq lock held */
+static void
+coh901318_desc_submit(struct coh901318_chan *cohc, struct coh901318_desc *desc)
+{
+	list_add_tail(&desc->node, &cohc->active);
+
+	BUG_ON(cohc->pending_irqs != 0);
+
+	cohc->pending_irqs = desc->pending_irqs;
+}
+
+static struct coh901318_desc *
+coh901318_first_active_get(struct coh901318_chan *cohc)
+{
+	struct coh901318_desc *d;
+
+	if (list_empty(&cohc->active))
+		return NULL;
+
+	d = list_first_entry(&cohc->active,
+			     struct coh901318_desc,
+			     node);
+	return d;
+}
+
+static void
+coh901318_desc_remove(struct coh901318_desc *cohd)
+{
+	list_del(&cohd->node);
+}
+
+static void
+coh901318_desc_queue(struct coh901318_chan *cohc, struct coh901318_desc *desc)
+{
+	list_add_tail(&desc->node, &cohc->queue);
+}
+
+static struct coh901318_desc *
+coh901318_first_queued(struct coh901318_chan *cohc)
+{
+	struct coh901318_desc *d;
+
+	if (list_empty(&cohc->queue))
+		return NULL;
+
+	d = list_first_entry(&cohc->queue,
+			     struct coh901318_desc,
+			     node);
+	return d;
+}
+
+/*
+ * DMA start/stop controls
+ */
+u32 coh901318_get_bytes_left(struct dma_chan *chan)
+{
+	unsigned long flags;
+	u32 ret;
+	struct coh901318_chan *cohc = to_coh901318_chan(chan);
+
+	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;
+
+	spin_unlock_irqrestore(&cohc->lock, flags);
+
+	return ret;
+}
+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)
+{
+	u32 val;
+	unsigned long flags;
+	struct coh901318_chan *cohc = to_coh901318_chan(chan);
+	int channel = cohc->id;
+	void __iomem *virtbase = cohc->base->virtbase;
+
+	spin_lock_irqsave(&cohc->lock, flags);
+
+	/* Disable channel in HW */
+	val = readl(virtbase + COH901318_CX_CFG +
+		    COH901318_CX_CFG_SPACING * channel);
+
+	/* Stopping infinit transfer */
+	if ((val & COH901318_CX_CTRL_TC_ENABLE) == 0 &&
+	    (val & COH901318_CX_CFG_CH_ENABLE))
+		cohc->stopped = 1;
+
+
+	val &= ~COH901318_CX_CFG_CH_ENABLE;
+	/* Enable twice, HW bug work around */
+	writel(val, virtbase + COH901318_CX_CFG +
+	       COH901318_CX_CFG_SPACING * channel);
+	writel(val, virtbase + COH901318_CX_CFG +
+	       COH901318_CX_CFG_SPACING * channel);
+
+	/* Spin-wait for it to actually go inactive */
+	while (readl(virtbase + COH901318_CX_STAT+COH901318_CX_STAT_SPACING *
+		     channel) & COH901318_CX_STAT_ACTIVE)
+		cpu_relax();
+
+	/* Check if we stopped an active job */
+	if ((readl(virtbase + COH901318_CX_CTRL+COH901318_CX_CTRL_SPACING *
+		   channel) & COH901318_CX_CTRL_TC_VALUE_MASK) > 0)
+		cohc->stopped = 1;
+
+	enable_powersave(cohc);
+
+	spin_unlock_irqrestore(&cohc->lock, flags);
+}
+EXPORT_SYMBOL(coh901318_stop);
+
+/* Continues a transfer that has been stopped via 300_dma_stop(..).
+   Power save is handled.
+*/
+void coh901318_continue(struct dma_chan *chan)
+{
+	u32 val;
+	unsigned long flags;
+	struct coh901318_chan *cohc = to_coh901318_chan(chan);
+	int channel = cohc->id;
+
+	spin_lock_irqsave(&cohc->lock, flags);
+
+	disable_powersave(cohc);
+
+	if (cohc->stopped) {
+		/* Enable channel in HW */
+		val = readl(cohc->base->virtbase + COH901318_CX_CFG +
+			    COH901318_CX_CFG_SPACING * channel);
+
+		val |= COH901318_CX_CFG_CH_ENABLE;
+
+		writel(val, cohc->base->virtbase + COH901318_CX_CFG +
+		       COH901318_CX_CFG_SPACING*channel);
+
+		cohc->stopped = 0;
+	}
+
+	spin_unlock_irqrestore(&cohc->lock, flags);
+}
+EXPORT_SYMBOL(coh901318_continue);
+
+bool coh901318_filter_id(struct dma_chan *chan, void *chan_id)
+{
+	unsigned int ch_nr = (unsigned int) chan_id;
+
+	if (ch_nr == to_coh901318_chan(chan)->id)
+		return true;
+
+	return false;
+}
+EXPORT_SYMBOL(coh901318_filter_id);
+
+/*
+ * DMA channel allocation
+ */
+static int coh901318_config(struct coh901318_chan *cohc,
+			    struct coh901318_params *param)
+{
+	unsigned long flags;
+	const struct coh901318_params *p;
+	int channel = cohc->id;
+	void __iomem *virtbase = cohc->base->virtbase;
+
+	spin_lock_irqsave(&cohc->lock, flags);
+
+	if (param)
+		p = param;
+	else
+		p = &cohc->base->platform->chan_conf[channel].param;
+
+	/* Clear any pending BE or TC interrupt */
+	if (channel < 32) {
+		writel(1 << channel, virtbase + COH901318_BE_INT_CLEAR1);
+		writel(1 << channel, virtbase + COH901318_TC_INT_CLEAR1);
+	} else {
+		writel(1 << (channel - 32), virtbase +
+		       COH901318_BE_INT_CLEAR2);
+		writel(1 << (channel - 32), virtbase +
+		       COH901318_TC_INT_CLEAR2);
+	}
+
+	coh901318_set_conf(cohc, p->config);
+	coh901318_set_ctrl(cohc, p->ctrl_lli_last);
+
+	spin_unlock_irqrestore(&cohc->lock, flags);
+
+	return 0;
+}
+
+/* must lock when calling this function
+ * start queued jobs, if any
+ * TODO: start all queued jobs in one go
+ *
+ * Returns descriptor if queued job is started otherwise NULL.
+ * If the queue is empty NULL is returned.
+ */
+static struct coh901318_desc *coh901318_queue_start(struct coh901318_chan *cohc)
+{
+	struct coh901318_desc *cohd_que;
+
+	/* start queued jobs, if any
+	 * TODO: transmit all queued jobs in one go
+	 */
+	cohd_que = coh901318_first_queued(cohc);
+
+	if (cohd_que != NULL) {
+		/* Remove from queue */
+		coh901318_desc_remove(cohd_que);
+		/* initiate DMA job */
+		cohc->busy = 1;
+
+		coh901318_desc_submit(cohc, cohd_que);
+
+		coh901318_prep_linked_list(cohc, cohd_que->data);
+
+		/* start dma job */
+		coh901318_start(cohc);
+
+	}
+
+	return cohd_que;
+}
+
+static void dma_tasklet(unsigned long data)
+{
+	struct coh901318_chan *cohc = (struct coh901318_chan *) data;
+	struct coh901318_desc *cohd_fin;
+	unsigned long flags;
+	dma_async_tx_callback callback;
+	void *callback_param;
+
+	spin_lock_irqsave(&cohc->lock, flags);
+
+	/* get first active entry from list */
+	cohd_fin = coh901318_first_active_get(cohc);
+
+	BUG_ON(cohd_fin->pending_irqs == 0);
+
+	if (cohd_fin == NULL)
+		goto err;
+
+	cohd_fin->pending_irqs--;
+	cohc->completed = cohd_fin->desc.cookie;
+
+	BUG_ON(cohc->nbr_active_done && cohd_fin == NULL);
+
+	if (cohc->nbr_active_done == 0)
+		return;
+
+	if (!cohd_fin->pending_irqs) {
+		/* release the lli allocation*/
+		coh901318_lli_free(&cohc->base->pool, &cohd_fin->data);
+	}
+
+	dev_vdbg(COHC_2_DEV(cohc), "[%s] chan_id %d pending_irqs %d"
+		 " nbr_active_done %ld\n", __func__,
+		 cohc->id, cohc->pending_irqs, cohc->nbr_active_done);
+
+	/* callback to client */
+	callback = cohd_fin->desc.callback;
+	callback_param = cohd_fin->desc.callback_param;
+
+	if (!cohd_fin->pending_irqs) {
+		coh901318_desc_remove(cohd_fin);
+
+		/* return desc to free-list */
+		coh901318_desc_free(cohc, cohd_fin);
+	}
+
+	if (cohc->nbr_active_done)
+		cohc->nbr_active_done--;
+
+	if (cohc->nbr_active_done) {
+		if (cohc_chan_conf(cohc)->priority_high)
+			tasklet_hi_schedule(&cohc->tasklet);
+		else
+			tasklet_schedule(&cohc->tasklet);
+	}
+	spin_unlock_irqrestore(&cohc->lock, flags);
+
+	if (callback)
+		callback(callback_param);
+
+	return;
+
+ err:
+	spin_unlock_irqrestore(&cohc->lock, flags);
+	dev_err(COHC_2_DEV(cohc), "[%s] No active dma desc\n", __func__);
+}
+
+
+/* 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)
+		return;
+
+	BUG_ON(cohc->pending_irqs == 0);
+
+	cohc->pending_irqs--;
+	cohc->nbr_active_done++;
+
+	if (cohc->pending_irqs == 0 && coh901318_queue_start(cohc) == NULL)
+		cohc->busy = 0;
+
+	BUG_ON(list_empty(&cohc->active));
+
+	if (cohc_chan_conf(cohc)->priority_high)
+		tasklet_hi_schedule(&cohc->tasklet);
+	else
+		tasklet_schedule(&cohc->tasklet);
+}
+
+
+static irqreturn_t dma_irq_handler(int irq, void *dev_id)
+{
+	u32 status1;
+	u32 status2;
+	int i;
+	int ch;
+	struct coh901318_base *base  = dev_id;
+	struct coh901318_chan *cohc;
+	void __iomem *virtbase = base->virtbase;
+
+	status1 = readl(virtbase + COH901318_INT_STATUS1);
+	status2 = readl(virtbase + COH901318_INT_STATUS2);
+
+	if (unlikely(status1 == 0 && status2 == 0)) {
+		dev_warn(base->dev, "spurious DMA IRQ from no channel!\n");
+		return IRQ_HANDLED;
+	}
+
+	/* TODO: consider handle IRQ in tasklet here to
+	 *       minimize interrupt latency */
+
+	/* Check the first 32 DMA channels for IRQ */
+	while (status1) {
+		/* Find first bit set, return as a number. */
+		i = ffs(status1) - 1;
+		ch = i;
+
+		cohc = &base->chans[ch];
+		spin_lock(&cohc->lock);
+
+		/* Mask off this bit */
+		status1 &= ~(1 << i);
+		/* Check the individual channel bits */
+		if (test_bit(i, virtbase + COH901318_BE_INT_STATUS1)) {
+			dev_crit(COHC_2_DEV(cohc),
+				 "DMA bus error on channel %d!\n", ch);
+			BUG_ON(1);
+			/* Clear BE interrupt */
+			__set_bit(i, virtbase + COH901318_BE_INT_CLEAR1);
+		} else {
+			/* Caused by TC, really? */
+			if (unlikely(!test_bit(i, virtbase +
+					       COH901318_TC_INT_STATUS1))) {
+				dev_warn(COHC_2_DEV(cohc),
+					 "ignoring interrupt not caused by terminal count on channel %d\n", ch);
+				/* Clear TC interrupt */
+				BUG_ON(1);
+				__set_bit(i, virtbase + COH901318_TC_INT_CLEAR1);
+			} else {
+				/* Enable powersave if transfer has finished */
+				if (!(readl(virtbase + COH901318_CX_STAT +
+					    COH901318_CX_STAT_SPACING*ch) &
+				      COH901318_CX_STAT_ENABLED)) {
+					enable_powersave(cohc);
+				}
+
+				/* Must clear TC interrupt before calling
+				 * dma_tc_handle
+				 * in case tc_handle initate a new dma job
+				 */
+				__set_bit(i, virtbase + COH901318_TC_INT_CLEAR1);
+
+				dma_tc_handle(cohc);
+			}
+		}
+		spin_unlock(&cohc->lock);
+	}
+
+	/* Check the remaining 32 DMA channels for IRQ */
+	while (status2) {
+		/* Find first bit set, return as a number. */
+		i = ffs(status2) - 1;
+		ch = i + 32;
+		cohc = &base->chans[ch];
+		spin_lock(&cohc->lock);
+
+		/* Mask off this bit */
+		status2 &= ~(1 << i);
+		/* Check the individual channel bits */
+		if (test_bit(i, virtbase + COH901318_BE_INT_STATUS2)) {
+			dev_crit(COHC_2_DEV(cohc),
+				 "DMA bus error on channel %d!\n", ch);
+			/* Clear BE interrupt */
+			BUG_ON(1);
+			__set_bit(i, virtbase + COH901318_BE_INT_CLEAR2);
+		} else {
+			/* Caused by TC, really? */
+			if (unlikely(!test_bit(i, virtbase +
+					       COH901318_TC_INT_STATUS2))) {
+				dev_warn(COHC_2_DEV(cohc),
+					 "ignoring interrupt not caused by terminal count on channel %d\n", ch);
+				/* Clear TC interrupt */
+				__set_bit(i, virtbase + COH901318_TC_INT_CLEAR2);
+				BUG_ON(1);
+			} else {
+				/* Enable powersave if transfer has finished */
+				if (!(readl(virtbase + COH901318_CX_STAT +
+					    COH901318_CX_STAT_SPACING*ch) &
+				      COH901318_CX_STAT_ENABLED)) {
+					enable_powersave(cohc);
+				}
+				/* Must clear TC interrupt before calling
+				 * dma_tc_handle
+				 * in case tc_handle initate a new dma job
+				 */
+				__set_bit(i, virtbase + COH901318_TC_INT_CLEAR2);
+
+				dma_tc_handle(cohc);
+			}
+		}
+		spin_unlock(&cohc->lock);
+	}
+
+	return IRQ_HANDLED;
+}
+
+static int coh901318_alloc_chan_resources(struct dma_chan *chan)
+{
+	struct coh901318_chan	*cohc = to_coh901318_chan(chan);
+
+	dev_vdbg(COHC_2_DEV(cohc), "[%s] DMA channel %d\n",
+		 __func__, cohc->id);
+
+	if (chan->client_count > 1)
+		return -EBUSY;
+
+	coh901318_config(cohc, NULL);
+
+	cohc->allocated = 1;
+	cohc->completed = chan->cookie = 1;
+
+	return 1;
+}
+
+static void
+coh901318_free_chan_resources(struct dma_chan *chan)
+{
+	struct coh901318_chan	*cohc = to_coh901318_chan(chan);
+	int channel = cohc->id;
+	unsigned long flags;
+
+	spin_lock_irqsave(&cohc->lock, flags);
+
+	/* Disable HW */
+	writel(0x00000000U, cohc->base->virtbase + COH901318_CX_CFG +
+	       COH901318_CX_CFG_SPACING*channel);
+	writel(0x00000000U, cohc->base->virtbase + COH901318_CX_CTRL +
+	       COH901318_CX_CTRL_SPACING*channel);
+
+	cohc->allocated = 0;
+
+	spin_unlock_irqrestore(&cohc->lock, flags);
+
+	chan->device->device_terminate_all(chan);
+}
+
+
+static dma_cookie_t
+coh901318_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+	struct coh901318_desc *cohd = container_of(tx, struct coh901318_desc,
+						   desc);
+	struct coh901318_chan *cohc = to_coh901318_chan(tx->chan);
+	unsigned long flags;
+
+	spin_lock_irqsave(&cohc->lock, flags);
+
+	tx->cookie = coh901318_assign_cookie(cohc, cohd);
+
+	coh901318_desc_queue(cohc, cohd);
+
+	spin_unlock_irqrestore(&cohc->lock, flags);
+
+	return tx->cookie;
+}
+
+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_desc *cohd;
+	unsigned long flg;
+	struct coh901318_chan *cohc = to_coh901318_chan(chan);
+	int lli_len;
+	u32 ctrl_last = cohc_chan_param(cohc)->ctrl_lli_last;
+
+	spin_lock_irqsave(&cohc->lock, flg);
+
+	dev_vdbg(COHC_2_DEV(cohc),
+		 "[%s] channel %d src 0x%x dest 0x%x size %d\n",
+		 __func__, cohc->id, src, dest, size);
+
+	if (flags & DMA_PREP_INTERRUPT)
+		/* Trigger interrupt after last lli */
+		ctrl_last |= COH901318_CX_CTRL_TC_IRQ_ENABLE;
+
+	lli_len = size >> MAX_DMA_PACKET_SIZE_SHIFT;
+	if ((lli_len << MAX_DMA_PACKET_SIZE_SHIFT) < size)
+		lli_len++;
+
+	data = coh901318_lli_alloc(&cohc->base->pool, lli_len);
+
+	if (data == NULL)
+		goto err;
+
+	cohd = coh901318_desc_get(cohc);
+	cohd->sg = NULL;
+	cohd->sg_len = 0;
+	cohd->data = data;
+
+	cohd->pending_irqs =
+		coh901318_lli_fill_memcpy(
+				&cohc->base->pool, data, src, size, dest,
+				cohc_chan_param(cohc)->ctrl_lli_chained,
+				ctrl_last);
+	cohd->flags = flags;
+
+	COH_DBG(coh901318_list_print(cohc, data));
+
+	dma_async_tx_descriptor_init(&cohd->desc, chan);
+
+	cohd->desc.tx_submit = coh901318_tx_submit;
+
+	spin_unlock_irqrestore(&cohc->lock, flg);
+
+	return &cohd->desc;
+ err:
+	spin_unlock_irqrestore(&cohc->lock, flg);
+	return NULL;
+}
+
+static struct dma_async_tx_descriptor *
+coh901318_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
+			unsigned int sg_len, enum dma_data_direction direction,
+			unsigned long flags)
+{
+	struct coh901318_chan *cohc = to_coh901318_chan(chan);
+	struct coh901318_lli *data;
+	struct coh901318_desc *cohd;
+	struct scatterlist *sg;
+	int len = 0;
+	int size;
+	int i;
+	u32 ctrl_chained = cohc_chan_param(cohc)->ctrl_lli_chained;
+	u32 ctrl = cohc_chan_param(cohc)->ctrl_lli;
+	u32 ctrl_last = cohc_chan_param(cohc)->ctrl_lli_last;
+	unsigned long flg;
+
+	if (!sgl)
+		goto out;
+	if (sgl->length == 0)
+		goto out;
+
+	spin_lock_irqsave(&cohc->lock, flg);
+
+	dev_vdbg(COHC_2_DEV(cohc), "[%s] sg_len %d dir %d\n",
+		 __func__, sg_len, direction);
+
+	if (flags & DMA_PREP_INTERRUPT)
+		/* Trigger interrupt after last lli */
+		ctrl_last |= COH901318_CX_CTRL_TC_IRQ_ENABLE;
+
+	cohd = coh901318_desc_get(cohc);
+	cohd->sg = NULL;
+	cohd->sg_len = 0;
+	cohd->dir = direction;
+
+	if (direction == DMA_TO_DEVICE) {
+		u32 tx_flags = COH901318_CX_CTRL_PRDD_SOURCE |
+			COH901318_CX_CTRL_SRC_ADDR_INC_ENABLE;
+
+		ctrl_chained |= tx_flags;
+		ctrl_last |= tx_flags;
+		ctrl |= tx_flags;
+	} else if (direction == DMA_FROM_DEVICE) {
+		u32 rx_flags = COH901318_CX_CTRL_PRDD_DEST |
+			COH901318_CX_CTRL_DST_ADDR_INC_ENABLE;
+
+		ctrl_chained |= rx_flags;
+		ctrl_last |= rx_flags;
+		ctrl |= rx_flags;
+	} else
+		goto err_direction;
+
+	dma_async_tx_descriptor_init(&cohd->desc, chan);
+
+	cohd->desc.tx_submit = coh901318_tx_submit;
+
+
+	/* The dma only supports transmitting packages up to
+	 * MAX_DMA_PACKET_SIZE. Calculate to total number of
+	 * dma elemts required to send the entire sg list
+	 */
+	for_each_sg(sgl, sg, sg_len, i) {
+		unsigned int factor;
+		size = sg_dma_len(sg);
+
+		if (size <= MAX_DMA_PACKET_SIZE) {
+			len++;
+			continue;
+		}
+
+		factor = size >> MAX_DMA_PACKET_SIZE_SHIFT;
+		if ((factor << MAX_DMA_PACKET_SIZE_SHIFT) < size)
+			factor++;
+
+		len += factor;
+	}
+
+	data = coh901318_lli_alloc(&cohc->base->pool, len);
+
+	if (data == NULL)
+		goto err_dma_alloc;
+
+	/* initiate allocated data list */
+	cohd->pending_irqs =
+		coh901318_lli_fill_sg(&cohc->base->pool, data, sgl, sg_len,
+				      cohc_dev_addr(cohc),
+				      ctrl_chained,
+				      ctrl,
+				      ctrl_last,
+				      direction, COH901318_CX_CTRL_TC_IRQ_ENABLE);
+	cohd->data = data;
+
+	cohd->flags = flags;
+
+	COH_DBG(coh901318_list_print(cohc, data));
+
+	spin_unlock_irqrestore(&cohc->lock, flg);
+
+	return &cohd->desc;
+ err_dma_alloc:
+ err_direction:
+	coh901318_desc_remove(cohd);
+	coh901318_desc_free(cohc, cohd);
+	spin_unlock_irqrestore(&cohc->lock, flg);
+ out:
+	return NULL;
+}
+
+static enum dma_status
+coh901318_is_tx_complete(struct dma_chan *chan,
+			 dma_cookie_t cookie, dma_cookie_t *done,
+			 dma_cookie_t *used)
+{
+	struct coh901318_chan *cohc = to_coh901318_chan(chan);
+	dma_cookie_t last_used;
+	dma_cookie_t last_complete;
+	int ret;
+
+	last_complete = cohc->completed;
+	last_used = chan->cookie;
+
+	ret = dma_async_is_complete(cookie, last_complete, last_used);
+
+	if (done)
+		*done = last_complete;
+	if (used)
+		*used = last_used;
+
+	return ret;
+}
+
+static void
+coh901318_issue_pending(struct dma_chan *chan)
+{
+	struct coh901318_chan *cohc = to_coh901318_chan(chan);
+	unsigned long flags;
+
+	spin_lock_irqsave(&cohc->lock, flags);
+
+	/* Busy means that pending jobs are already being processed */
+	if (!cohc->busy)
+		coh901318_queue_start(cohc);
+
+	spin_unlock_irqrestore(&cohc->lock, flags);
+}
+
+static void
+coh901318_terminate_all(struct dma_chan *chan)
+{
+	unsigned long flags;
+	struct coh901318_chan *cohc = to_coh901318_chan(chan);
+	struct coh901318_desc *cohd;
+	void __iomem *virtbase = cohc->base->virtbase;
+
+	coh901318_stop(chan);
+
+	spin_lock_irqsave(&cohc->lock, flags);
+
+	/* Clear any pending BE or TC interrupt */
+	if (cohc->id < 32) {
+		writel(1 << cohc->id, virtbase + COH901318_BE_INT_CLEAR1);
+		writel(1 << cohc->id, virtbase + COH901318_TC_INT_CLEAR1);
+	} else {
+		writel(1 << (cohc->id - 32), virtbase +
+		       COH901318_BE_INT_CLEAR2);
+		writel(1 << (cohc->id - 32), virtbase +
+		       COH901318_TC_INT_CLEAR2);
+	}
+
+	enable_powersave(cohc);
+
+	while ((cohd = coh901318_first_active_get(cohc))) {
+		/* release the lli allocation*/
+		coh901318_lli_free(&cohc->base->pool, &cohd->data);
+
+		coh901318_desc_remove(cohd);
+
+		/* return desc to free-list */
+		coh901318_desc_free(cohc, cohd);
+	}
+
+	while ((cohd = coh901318_first_queued(cohc))) {
+		/* release the lli allocation*/
+		coh901318_lli_free(&cohc->base->pool, &cohd->data);
+
+		coh901318_desc_remove(cohd);
+
+		/* return desc to free-list */
+		coh901318_desc_free(cohc, cohd);
+	}
+
+
+	cohc->nbr_active_done = 0;
+	cohc->busy = 0;
+	cohc->pending_irqs = 0;
+
+	spin_unlock_irqrestore(&cohc->lock, flags);
+}
+void coh901318_base_init(struct dma_device *dma, const int *pick_chans,
+			 struct coh901318_base *base)
+{
+	int chans_i;
+	int i = 0;
+	struct coh901318_chan *cohc;
+
+	INIT_LIST_HEAD(&dma->channels);
+
+	for (chans_i = 0; pick_chans[chans_i] != -1; chans_i += 2) {
+		for (i = pick_chans[chans_i]; i <= pick_chans[chans_i+1]; i++) {
+			cohc = &base->chans[i];
+
+			cohc->base = base;
+			cohc->chan.device = dma;
+			cohc->id = i;
+
+			/* TODO: do we really need this lock if only one
+			 * client is connected to each channel?
+			 */
+
+			spin_lock_init(&cohc->lock);
+
+			cohc->pending_irqs = 0;
+			cohc->nbr_active_done = 0;
+			cohc->busy = 0;
+			INIT_LIST_HEAD(&cohc->free);
+			INIT_LIST_HEAD(&cohc->active);
+			INIT_LIST_HEAD(&cohc->queue);
+
+			tasklet_init(&cohc->tasklet, dma_tasklet,
+				     (unsigned long) cohc);
+
+			list_add_tail(&cohc->chan.device_node,
+				      &dma->channels);
+		}
+	}
+}
+
+static int __init coh901318_probe(struct platform_device *pdev)
+{
+	int err = 0;
+	struct coh901318_platform *pdata;
+	struct coh901318_base *base;
+	int irq;
+	struct resource *io;
+
+	io = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!io)
+		goto err_get_resource;
+
+	/* Map DMA controller registers to virtual memory */
+	if (request_mem_region(io->start,
+			       resource_size(io),
+			       pdev->dev.driver->name) == NULL) {
+		err = -EBUSY;
+		goto err_request_mem;
+	}
+
+	pdata = pdev->dev.platform_data;
+	if (!pdata)
+		goto err_no_platformdata;
+
+	base = kmalloc(ALIGN(sizeof(struct coh901318_base), 4) +
+		       pdata->max_channels *
+		       sizeof(struct coh901318_chan),
+		       GFP_KERNEL);
+	if (!base)
+		goto err_alloc_coh_dma_channels;
+
+	base->chans = ((void *)base) + ALIGN(sizeof(struct coh901318_base), 4);
+
+	base->virtbase = ioremap(io->start, resource_size(io));
+	if (!base->virtbase) {
+		err = -ENOMEM;
+		goto err_no_ioremap;
+	}
+
+	base->dev = &pdev->dev;
+	base->platform = pdata;
+	spin_lock_init(&base->pm.lock);
+	base->pm.started_channels = 0;
+
+	COH901318_DEBUGFS_ASSIGN(debugfs_dma_base, base);
+
+	platform_set_drvdata(pdev, base);
+
+	irq = platform_get_irq(pdev, 0);
+	if (irq < 0)
+		goto err_no_irq;
+
+	err = request_irq(irq, dma_irq_handler, IRQF_DISABLED,
+			  "coh901318", base);
+	if (err) {
+		dev_crit(&pdev->dev,
+			 "Cannot allocate IRQ for DMA controller!\n");
+		goto err_request_irq;
+	}
+
+	err = coh901318_pool_create(&base->pool, &pdev->dev,
+				    sizeof(struct coh901318_lli),
+				    32);
+	if (err)
+		goto err_pool_create;
+
+	/* init channels for device transfers */
+	coh901318_base_init(&base->dma_slave,  base->platform->chans_slave,
+			    base);
+
+	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 = 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_issue_pending = coh901318_issue_pending;
+	base->dma_slave.device_terminate_all = coh901318_terminate_all;
+	base->dma_slave.dev = &pdev->dev;
+
+	err = dma_async_device_register(&base->dma_slave);
+
+	if (err)
+		goto err_register_slave;
+
+	/* init channels for memcpy */
+	coh901318_base_init(&base->dma_memcpy, base->platform->chans_memcpy,
+			    base);
+
+	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 = 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_issue_pending = coh901318_issue_pending;
+	base->dma_memcpy.device_terminate_all = coh901318_terminate_all;
+	base->dma_memcpy.dev = &pdev->dev;
+	err = dma_async_device_register(&base->dma_memcpy);
+
+	if (err)
+		goto err_register_memcpy;
+
+	dev_dbg(&pdev->dev, "Initialized COH901318 DMA on virtual base 0x%08x\n",
+		(u32) base->virtbase);
+
+	return err;
+
+ err_register_memcpy:
+	dma_async_device_unregister(&base->dma_slave);
+ err_register_slave:
+	coh901318_pool_destroy(&base->pool);
+ err_pool_create:
+	free_irq(platform_get_irq(pdev, 0), base);
+ err_request_irq:
+ err_no_irq:
+	iounmap(base->virtbase);
+ err_no_ioremap:
+	kfree(base);
+ err_alloc_coh_dma_channels:
+ err_no_platformdata:
+	release_mem_region(pdev->resource->start,
+			   resource_size(pdev->resource));
+ err_request_mem:
+ err_get_resource:
+	return err;
+}
+
+static int __exit coh901318_remove(struct platform_device *pdev)
+{
+	struct coh901318_base *base = platform_get_drvdata(pdev);
+
+	dma_async_device_unregister(&base->dma_memcpy);
+	dma_async_device_unregister(&base->dma_slave);
+	coh901318_pool_destroy(&base->pool);
+	free_irq(platform_get_irq(pdev, 0), base);
+	iounmap(base->virtbase);
+	kfree(base);
+	release_mem_region(pdev->resource->start,
+			   resource_size(pdev->resource));
+	return 0;
+}
+
+
+static struct platform_driver coh901318_driver = {
+	.remove = __exit_p(coh901318_remove),
+	.driver = {
+		.name	= "coh901318",
+	},
+};
+
+int __init coh901318_init(void)
+{
+	return platform_driver_probe(&coh901318_driver, coh901318_probe);
+}
+subsys_initcall(coh901318_init);
+
+void __exit coh901318_exit(void)
+{
+	platform_driver_unregister(&coh901318_driver);
+}
+module_exit(coh901318_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Per Friden");
diff --git a/drivers/dma/coh901318_lli.c b/drivers/dma/coh901318_lli.c
new file mode 100644
index 000000000000..f5120f238a4d
--- /dev/null
+++ b/drivers/dma/coh901318_lli.c
@@ -0,0 +1,318 @@
+/*
+ * driver/dma/coh901318_lli.c
+ *
+ * Copyright (C) 2007-2009 ST-Ericsson
+ * License terms: GNU General Public License (GPL) version 2
+ * Support functions for handling lli for dma
+ * Author: Per Friden <per.friden@stericsson.com>
+ */
+
+#include <linux/dma-mapping.h>
+#include <linux/spinlock.h>
+#include <linux/dmapool.h>
+#include <linux/memory.h>
+#include <mach/coh901318.h>
+
+#include "coh901318_lli.h"
+
+#if (defined(CONFIG_DEBUG_FS) && defined(CONFIG_U300_DEBUG))
+#define DEBUGFS_POOL_COUNTER_RESET(pool) (pool->debugfs_pool_counter = 0)
+#define DEBUGFS_POOL_COUNTER_ADD(pool, add) (pool->debugfs_pool_counter += add)
+#else
+#define DEBUGFS_POOL_COUNTER_RESET(pool)
+#define DEBUGFS_POOL_COUNTER_ADD(pool, add)
+#endif
+
+static struct coh901318_lli *
+coh901318_lli_next(struct coh901318_lli *data)
+{
+	if (data == NULL || data->link_addr == 0)
+		return NULL;
+
+	return (struct coh901318_lli *) data->virt_link_addr;
+}
+
+int coh901318_pool_create(struct coh901318_pool *pool,
+			  struct device *dev,
+			  size_t size, size_t align)
+{
+	spin_lock_init(&pool->lock);
+	pool->dev = dev;
+	pool->dmapool = dma_pool_create("lli_pool", dev, size, align, 0);
+
+	DEBUGFS_POOL_COUNTER_RESET(pool);
+	return 0;
+}
+
+int coh901318_pool_destroy(struct coh901318_pool *pool)
+{
+
+	dma_pool_destroy(pool->dmapool);
+	return 0;
+}
+
+struct coh901318_lli *
+coh901318_lli_alloc(struct coh901318_pool *pool, unsigned int len)
+{
+	int i;
+	struct coh901318_lli *head;
+	struct coh901318_lli *lli;
+	struct coh901318_lli *lli_prev;
+	dma_addr_t phy;
+
+	if (len == 0)
+		goto err;
+
+	spin_lock(&pool->lock);
+
+	head = dma_pool_alloc(pool->dmapool, GFP_NOWAIT, &phy);
+
+	if (head == NULL)
+		goto err;
+
+	DEBUGFS_POOL_COUNTER_ADD(pool, 1);
+
+	lli = head;
+	lli->phy_this = phy;
+
+	for (i = 1; i < len; i++) {
+		lli_prev = lli;
+
+		lli = dma_pool_alloc(pool->dmapool, GFP_NOWAIT, &phy);
+
+		if (lli == NULL)
+			goto err_clean_up;
+
+		DEBUGFS_POOL_COUNTER_ADD(pool, 1);
+		lli->phy_this = phy;
+
+		lli_prev->link_addr = phy;
+		lli_prev->virt_link_addr = lli;
+	}
+
+	lli->link_addr = 0x00000000U;
+
+	spin_unlock(&pool->lock);
+
+	return head;
+
+ err:
+	spin_unlock(&pool->lock);
+	return NULL;
+
+ err_clean_up:
+	lli_prev->link_addr = 0x00000000U;
+	spin_unlock(&pool->lock);
+	coh901318_lli_free(pool, &head);
+	return NULL;
+}
+
+void coh901318_lli_free(struct coh901318_pool *pool,
+			struct coh901318_lli **lli)
+{
+	struct coh901318_lli *l;
+	struct coh901318_lli *next;
+
+	if (lli == NULL)
+		return;
+
+	l = *lli;
+
+	if (l == NULL)
+		return;
+
+	spin_lock(&pool->lock);
+
+	while (l->link_addr) {
+		next = l->virt_link_addr;
+		dma_pool_free(pool->dmapool, l, l->phy_this);
+		DEBUGFS_POOL_COUNTER_ADD(pool, -1);
+		l = next;
+	}
+	dma_pool_free(pool->dmapool, l, l->phy_this);
+	DEBUGFS_POOL_COUNTER_ADD(pool, -1);
+
+	spin_unlock(&pool->lock);
+	*lli = NULL;
+}
+
+int
+coh901318_lli_fill_memcpy(struct coh901318_pool *pool,
+			  struct coh901318_lli *lli,
+			  dma_addr_t source, unsigned int size,
+			  dma_addr_t destination, u32 ctrl_chained,
+			  u32 ctrl_eom)
+{
+	int s = size;
+	dma_addr_t src = source;
+	dma_addr_t dst = destination;
+
+	lli->src_addr = src;
+	lli->dst_addr = dst;
+
+	while (lli->link_addr) {
+		lli->control = ctrl_chained | MAX_DMA_PACKET_SIZE;
+		lli->src_addr = src;
+		lli->dst_addr = dst;
+
+		s -= MAX_DMA_PACKET_SIZE;
+		lli = coh901318_lli_next(lli);
+
+		src += MAX_DMA_PACKET_SIZE;
+		dst += MAX_DMA_PACKET_SIZE;
+	}
+
+	lli->control = ctrl_eom | s;
+	lli->src_addr = src;
+	lli->dst_addr = dst;
+
+	/* One irq per single transfer */
+	return 1;
+}
+
+int
+coh901318_lli_fill_single(struct coh901318_pool *pool,
+			  struct coh901318_lli *lli,
+			  dma_addr_t buf, unsigned int size,
+			  dma_addr_t dev_addr, u32 ctrl_chained, u32 ctrl_eom,
+			  enum dma_data_direction dir)
+{
+	int s = size;
+	dma_addr_t src;
+	dma_addr_t dst;
+
+
+	if (dir == DMA_TO_DEVICE) {
+		src = buf;
+		dst = dev_addr;
+
+	} else if (dir == DMA_FROM_DEVICE) {
+
+		src = dev_addr;
+		dst = buf;
+	} else {
+		return -EINVAL;
+	}
+
+	while (lli->link_addr) {
+		size_t block_size = MAX_DMA_PACKET_SIZE;
+		lli->control = ctrl_chained | MAX_DMA_PACKET_SIZE;
+
+		/* If we are on the next-to-final block and there will
+		 * be less than half a DMA packet left for the last
+		 * block, then we want to make this block a little
+		 * smaller to balance the sizes. This is meant to
+		 * avoid too small transfers if the buffer size is
+		 * (MAX_DMA_PACKET_SIZE*N + 1) */
+		if (s < (MAX_DMA_PACKET_SIZE + MAX_DMA_PACKET_SIZE/2))
+			block_size = MAX_DMA_PACKET_SIZE/2;
+
+		s -= block_size;
+		lli->src_addr = src;
+		lli->dst_addr = dst;
+
+		lli = coh901318_lli_next(lli);
+
+		if (dir == DMA_TO_DEVICE)
+			src += block_size;
+		else if (dir == DMA_FROM_DEVICE)
+			dst += block_size;
+	}
+
+	lli->control = ctrl_eom | s;
+	lli->src_addr = src;
+	lli->dst_addr = dst;
+
+	/* One irq per single transfer */
+	return 1;
+}
+
+int
+coh901318_lli_fill_sg(struct coh901318_pool *pool,
+		      struct coh901318_lli *lli,
+		      struct scatterlist *sgl, unsigned int nents,
+		      dma_addr_t dev_addr, u32 ctrl_chained, u32 ctrl,
+		      u32 ctrl_last,
+		      enum dma_data_direction dir, u32 ctrl_irq_mask)
+{
+	int i;
+	struct scatterlist *sg;
+	u32 ctrl_sg;
+	dma_addr_t src = 0;
+	dma_addr_t dst = 0;
+	int nbr_of_irq = 0;
+	u32 bytes_to_transfer;
+	u32 elem_size;
+
+	if (lli == NULL)
+		goto err;
+
+	spin_lock(&pool->lock);
+
+	if (dir == DMA_TO_DEVICE)
+		dst = dev_addr;
+	else if (dir == DMA_FROM_DEVICE)
+		src = dev_addr;
+	else
+		goto err;
+
+	for_each_sg(sgl, sg, nents, i) {
+		if (sg_is_chain(sg)) {
+			/* sg continues to the next sg-element don't
+			 * send ctrl_finish until the last
+			 * sg-element in the chain
+			 */
+			ctrl_sg = ctrl_chained;
+		} else if (i == nents - 1)
+			ctrl_sg = ctrl_last;
+		else
+			ctrl_sg = ctrl ? ctrl : ctrl_last;
+
+
+		if ((ctrl_sg & ctrl_irq_mask))
+			nbr_of_irq++;
+
+		if (dir == DMA_TO_DEVICE)
+			/* increment source address */
+			src = sg_dma_address(sg);
+		else
+			/* increment destination address */
+			dst =  sg_dma_address(sg);
+
+		bytes_to_transfer = sg_dma_len(sg);
+
+		while (bytes_to_transfer) {
+			u32 val;
+
+			if (bytes_to_transfer > MAX_DMA_PACKET_SIZE) {
+				elem_size = MAX_DMA_PACKET_SIZE;
+				val = ctrl_chained;
+			} else {
+				elem_size = bytes_to_transfer;
+				val = ctrl_sg;
+			}
+
+			lli->control = val | elem_size;
+			lli->src_addr = src;
+			lli->dst_addr = dst;
+
+			if (dir == DMA_FROM_DEVICE)
+				dst += elem_size;
+			else
+				src += elem_size;
+
+			BUG_ON(lli->link_addr & 3);
+
+			bytes_to_transfer -= elem_size;
+			lli = coh901318_lli_next(lli);
+		}
+
+	}
+	spin_unlock(&pool->lock);
+
+	/* There can be many IRQs per sg transfer */
+	return nbr_of_irq;
+ err:
+	spin_unlock(&pool->lock);
+	return -EINVAL;
+}
diff --git a/drivers/dma/coh901318_lli.h b/drivers/dma/coh901318_lli.h
new file mode 100644
index 000000000000..7bf713b79c6b
--- /dev/null
+++ b/drivers/dma/coh901318_lli.h
@@ -0,0 +1,124 @@
+/*
+ * driver/dma/coh901318_lli.h
+ *
+ * Copyright (C) 2007-2009 ST-Ericsson
+ * License terms: GNU General Public License (GPL) version 2
+ * Support functions for handling lli for coh901318
+ * Author: Per Friden <per.friden@stericsson.com>
+ */
+
+#ifndef COH901318_LLI_H
+#define COH901318_LLI_H
+
+#include <mach/coh901318.h>
+
+struct device;
+
+struct coh901318_pool {
+	spinlock_t lock;
+	struct dma_pool *dmapool;
+	struct device *dev;
+
+#ifdef CONFIG_DEBUG_FS
+	int debugfs_pool_counter;
+#endif
+};
+
+struct device;
+/**
+ * coh901318_pool_create() - Creates an dma pool for lli:s
+ * @pool: pool handle
+ * @dev: dma device
+ * @lli_nbr: number of lli:s in the pool
+ * @algin: adress alignemtn of lli:s
+ * returns 0 on success otherwise none zero
+ */
+int coh901318_pool_create(struct coh901318_pool *pool,
+			  struct device *dev,
+			  size_t lli_nbr, size_t align);
+
+/**
+ * coh901318_pool_destroy() - Destroys the dma pool
+ * @pool: pool handle
+ * returns 0 on success otherwise none zero
+ */
+int coh901318_pool_destroy(struct coh901318_pool *pool);
+
+/**
+ * coh901318_lli_alloc() - Allocates a linked list
+ *
+ * @pool: pool handle
+ * @len: length to list
+ * return: none NULL if success otherwise NULL
+ */
+struct coh901318_lli *
+coh901318_lli_alloc(struct coh901318_pool *pool,
+		    unsigned int len);
+
+/**
+ * coh901318_lli_free() - Returns the linked list items to the pool
+ * @pool: pool handle
+ * @lli: reference to lli pointer to be freed
+ */
+void coh901318_lli_free(struct coh901318_pool *pool,
+			struct coh901318_lli **lli);
+
+/**
+ * coh901318_lli_fill_memcpy() - Prepares the lli:s for dma memcpy
+ * @pool: pool handle
+ * @lli: allocated lli
+ * @src: src address
+ * @size: transfer size
+ * @dst: destination address
+ * @ctrl_chained: ctrl for chained lli
+ * @ctrl_last: ctrl for the last lli
+ * returns number of CPU interrupts for the lli, negative on error.
+ */
+int
+coh901318_lli_fill_memcpy(struct coh901318_pool *pool,
+			  struct coh901318_lli *lli,
+			  dma_addr_t src, unsigned int size,
+			  dma_addr_t dst, u32 ctrl_chained, u32 ctrl_last);
+
+/**
+ * coh901318_lli_fill_single() - Prepares the lli:s for dma single transfer
+ * @pool: pool handle
+ * @lli: allocated lli
+ * @buf: transfer buffer
+ * @size: transfer size
+ * @dev_addr: address of periphal
+ * @ctrl_chained: ctrl for chained lli
+ * @ctrl_last: ctrl for the last lli
+ * @dir: direction of transfer (to or from device)
+ * returns number of CPU interrupts for the lli, negative on error.
+ */
+int
+coh901318_lli_fill_single(struct coh901318_pool *pool,
+			  struct coh901318_lli *lli,
+			  dma_addr_t buf, unsigned int size,
+			  dma_addr_t dev_addr, u32 ctrl_chained, u32 ctrl_last,
+			  enum dma_data_direction dir);
+
+/**
+ * coh901318_lli_fill_single() - Prepares the lli:s for dma scatter list transfer
+ * @pool: pool handle
+ * @lli: allocated lli
+ * @sg: scatter gather list
+ * @nents: number of entries in sg
+ * @dev_addr: address of periphal
+ * @ctrl_chained: ctrl for chained lli
+ * @ctrl: ctrl of middle lli
+ * @ctrl_last: ctrl for the last lli
+ * @dir: direction of transfer (to or from device)
+ * @ctrl_irq_mask: ctrl mask for CPU interrupt
+ * returns number of CPU interrupts for the lli, negative on error.
+ */
+int
+coh901318_lli_fill_sg(struct coh901318_pool *pool,
+		      struct coh901318_lli *lli,
+		      struct scatterlist *sg, unsigned int nents,
+		      dma_addr_t dev_addr, u32 ctrl_chained,
+		      u32 ctrl, u32 ctrl_last,
+		      enum dma_data_direction dir, u32 ctrl_irq_mask);
+
+#endif /* COH901318_LLI_H */
diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
index 8f99354082ce..6f51a0a7a8bb 100644
--- a/drivers/dma/dmaengine.c
+++ b/drivers/dma/dmaengine.c
@@ -326,14 +326,7 @@ arch_initcall(dma_channel_table_init);
  */
 struct dma_chan *dma_find_channel(enum dma_transaction_type tx_type)
 {
-	struct dma_chan *chan;
-	int cpu;
-
-	cpu = get_cpu();
-	chan = per_cpu_ptr(channel_table[tx_type], cpu)->chan;
-	put_cpu();
-
-	return chan;
+	return this_cpu_read(channel_table[tx_type]->chan);
 }
 EXPORT_SYMBOL(dma_find_channel);
 
@@ -857,7 +850,6 @@ dma_async_memcpy_buf_to_buf(struct dma_chan *chan, void *dest,
 	struct dma_async_tx_descriptor *tx;
 	dma_addr_t dma_dest, dma_src;
 	dma_cookie_t cookie;
-	int cpu;
 	unsigned long flags;
 
 	dma_src = dma_map_single(dev->dev, src, len, DMA_TO_DEVICE);
@@ -876,10 +868,10 @@ dma_async_memcpy_buf_to_buf(struct dma_chan *chan, void *dest,
 	tx->callback = NULL;
 	cookie = tx->tx_submit(tx);
 
-	cpu = get_cpu();
-	per_cpu_ptr(chan->local, cpu)->bytes_transferred += len;
-	per_cpu_ptr(chan->local, cpu)->memcpy_count++;
-	put_cpu();
+	preempt_disable();
+	__this_cpu_add(chan->local->bytes_transferred, len);
+	__this_cpu_inc(chan->local->memcpy_count);
+	preempt_enable();
 
 	return cookie;
 }
@@ -906,7 +898,6 @@ dma_async_memcpy_buf_to_pg(struct dma_chan *chan, struct page *page,
 	struct dma_async_tx_descriptor *tx;
 	dma_addr_t dma_dest, dma_src;
 	dma_cookie_t cookie;
-	int cpu;
 	unsigned long flags;
 
 	dma_src = dma_map_single(dev->dev, kdata, len, DMA_TO_DEVICE);
@@ -923,10 +914,10 @@ dma_async_memcpy_buf_to_pg(struct dma_chan *chan, struct page *page,
 	tx->callback = NULL;
 	cookie = tx->tx_submit(tx);
 
-	cpu = get_cpu();
-	per_cpu_ptr(chan->local, cpu)->bytes_transferred += len;
-	per_cpu_ptr(chan->local, cpu)->memcpy_count++;
-	put_cpu();
+	preempt_disable();
+	__this_cpu_add(chan->local->bytes_transferred, len);
+	__this_cpu_inc(chan->local->memcpy_count);
+	preempt_enable();
 
 	return cookie;
 }
@@ -955,7 +946,6 @@ dma_async_memcpy_pg_to_pg(struct dma_chan *chan, struct page *dest_pg,
 	struct dma_async_tx_descriptor *tx;
 	dma_addr_t dma_dest, dma_src;
 	dma_cookie_t cookie;
-	int cpu;
 	unsigned long flags;
 
 	dma_src = dma_map_page(dev->dev, src_pg, src_off, len, DMA_TO_DEVICE);
@@ -973,10 +963,10 @@ dma_async_memcpy_pg_to_pg(struct dma_chan *chan, struct page *dest_pg,
 	tx->callback = NULL;
 	cookie = tx->tx_submit(tx);
 
-	cpu = get_cpu();
-	per_cpu_ptr(chan->local, cpu)->bytes_transferred += len;
-	per_cpu_ptr(chan->local, cpu)->memcpy_count++;
-	put_cpu();
+	preempt_disable();
+	__this_cpu_add(chan->local->bytes_transferred, len);
+	__this_cpu_inc(chan->local->memcpy_count);
+	preempt_enable();
 
 	return cookie;
 }
diff --git a/drivers/dma/dmatest.c b/drivers/dma/dmatest.c
index a32a4cf7b1e0..8b905161fbf4 100644
--- a/drivers/dma/dmatest.c
+++ b/drivers/dma/dmatest.c
@@ -298,10 +298,6 @@ static int dmatest_func(void *data)
 
 		total_tests++;
 
-		len = dmatest_random() % test_buf_size + 1;
-		src_off = dmatest_random() % (test_buf_size - len + 1);
-		dst_off = dmatest_random() % (test_buf_size - len + 1);
-
 		/* honor alignment restrictions */
 		if (thread->type == DMA_MEMCPY)
 			align = dev->copy_align;
@@ -310,7 +306,19 @@ static int dmatest_func(void *data)
 		else if (thread->type == DMA_PQ)
 			align = dev->pq_align;
 
+		if (1 << align > test_buf_size) {
+			pr_err("%u-byte buffer too small for %d-byte alignment\n",
+			       test_buf_size, 1 << align);
+			break;
+		}
+
+		len = dmatest_random() % test_buf_size + 1;
 		len = (len >> align) << align;
+		if (!len)
+			len = 1 << align;
+		src_off = dmatest_random() % (test_buf_size - len + 1);
+		dst_off = dmatest_random() % (test_buf_size - len + 1);
+
 		src_off = (src_off >> align) << align;
 		dst_off = (dst_off >> align) << align;
 
diff --git a/drivers/dma/dw_dmac.c b/drivers/dma/dw_dmac.c
index 2eea823516a7..d28369f7afd2 100644
--- a/drivers/dma/dw_dmac.c
+++ b/drivers/dma/dw_dmac.c
@@ -1270,8 +1270,6 @@ static int __init dw_probe(struct platform_device *pdev)
 		goto err_kfree;
 	}
 
-	memset(dw, 0, sizeof *dw);
-
 	dw->regs = ioremap(io->start, DW_REGLEN);
 	if (!dw->regs) {
 		err = -ENOMEM;
@@ -1427,7 +1425,7 @@ static int dw_resume_noirq(struct device *dev)
 	return 0;
 }
 
-static struct dev_pm_ops dw_dev_pm_ops = {
+static const struct dev_pm_ops dw_dev_pm_ops = {
 	.suspend_noirq = dw_suspend_noirq,
 	.resume_noirq = dw_resume_noirq,
 };
diff --git a/drivers/dma/ioat/dma.c b/drivers/dma/ioat/dma.c
index c524d36d3c2e..dcc4ab78b32b 100644
--- a/drivers/dma/ioat/dma.c
+++ b/drivers/dma/ioat/dma.c
@@ -1032,7 +1032,7 @@ int __devinit ioat_probe(struct ioatdma_device *device)
 	dma->dev = &pdev->dev;
 
 	if (!dma->chancnt) {
-		dev_err(dev, "zero channels detected\n");
+		dev_err(dev, "channel enumeration error\n");
 		goto err_setup_interrupts;
 	}
 
diff --git a/drivers/dma/ioat/dma.h b/drivers/dma/ioat/dma.h
index 45edde996480..bbc3e78ef333 100644
--- a/drivers/dma/ioat/dma.h
+++ b/drivers/dma/ioat/dma.h
@@ -60,6 +60,7 @@
  * @dca: direct cache access context
  * @intr_quirk: interrupt setup quirk (for ioat_v1 devices)
  * @enumerate_channels: hw version specific channel enumeration
+ * @reset_hw: hw version specific channel (re)initialization
  * @cleanup_tasklet: select between the v2 and v3 cleanup routines
  * @timer_fn: select between the v2 and v3 timer watchdog routines
  * @self_test: hardware version specific self test for each supported op type
@@ -78,6 +79,7 @@ struct ioatdma_device {
 	struct dca_provider *dca;
 	void (*intr_quirk)(struct ioatdma_device *device);
 	int (*enumerate_channels)(struct ioatdma_device *device);
+	int (*reset_hw)(struct ioat_chan_common *chan);
 	void (*cleanup_tasklet)(unsigned long data);
 	void (*timer_fn)(unsigned long data);
 	int (*self_test)(struct ioatdma_device *device);
@@ -264,6 +266,22 @@ static inline void ioat_suspend(struct ioat_chan_common *chan)
 	writeb(IOAT_CHANCMD_SUSPEND, chan->reg_base + IOAT_CHANCMD_OFFSET(ver));
 }
 
+static inline void ioat_reset(struct ioat_chan_common *chan)
+{
+	u8 ver = chan->device->version;
+
+	writeb(IOAT_CHANCMD_RESET, chan->reg_base + IOAT_CHANCMD_OFFSET(ver));
+}
+
+static inline bool ioat_reset_pending(struct ioat_chan_common *chan)
+{
+	u8 ver = chan->device->version;
+	u8 cmd;
+
+	cmd = readb(chan->reg_base + IOAT_CHANCMD_OFFSET(ver));
+	return (cmd & IOAT_CHANCMD_RESET) == IOAT_CHANCMD_RESET;
+}
+
 static inline void ioat_set_chainaddr(struct ioat_dma_chan *ioat, u64 addr)
 {
 	struct ioat_chan_common *chan = &ioat->base;
diff --git a/drivers/dma/ioat/dma_v2.c b/drivers/dma/ioat/dma_v2.c
index 8f1f7f05deaa..5f7a500e18d0 100644
--- a/drivers/dma/ioat/dma_v2.c
+++ b/drivers/dma/ioat/dma_v2.c
@@ -239,20 +239,50 @@ void __ioat2_restart_chan(struct ioat2_dma_chan *ioat)
 		__ioat2_start_null_desc(ioat);
 }
 
-static void ioat2_restart_channel(struct ioat2_dma_chan *ioat)
+int ioat2_quiesce(struct ioat_chan_common *chan, unsigned long tmo)
 {
-	struct ioat_chan_common *chan = &ioat->base;
-	unsigned long phys_complete;
+	unsigned long end = jiffies + tmo;
+	int err = 0;
 	u32 status;
 
 	status = ioat_chansts(chan);
 	if (is_ioat_active(status) || is_ioat_idle(status))
 		ioat_suspend(chan);
 	while (is_ioat_active(status) || is_ioat_idle(status)) {
+		if (end && time_after(jiffies, end)) {
+			err = -ETIMEDOUT;
+			break;
+		}
 		status = ioat_chansts(chan);
 		cpu_relax();
 	}
 
+	return err;
+}
+
+int ioat2_reset_sync(struct ioat_chan_common *chan, unsigned long tmo)
+{
+	unsigned long end = jiffies + tmo;
+	int err = 0;
+
+	ioat_reset(chan);
+	while (ioat_reset_pending(chan)) {
+		if (end && time_after(jiffies, end)) {
+			err = -ETIMEDOUT;
+			break;
+		}
+		cpu_relax();
+	}
+
+	return err;
+}
+
+static void ioat2_restart_channel(struct ioat2_dma_chan *ioat)
+{
+	struct ioat_chan_common *chan = &ioat->base;
+	unsigned long phys_complete;
+
+	ioat2_quiesce(chan, 0);
 	if (ioat_cleanup_preamble(chan, &phys_complete))
 		__cleanup(ioat, phys_complete);
 
@@ -318,6 +348,19 @@ void ioat2_timer_event(unsigned long data)
 	spin_unlock_bh(&chan->cleanup_lock);
 }
 
+static int ioat2_reset_hw(struct ioat_chan_common *chan)
+{
+	/* throw away whatever the channel was doing and get it initialized */
+	u32 chanerr;
+
+	ioat2_quiesce(chan, msecs_to_jiffies(100));
+
+	chanerr = readl(chan->reg_base + IOAT_CHANERR_OFFSET);
+	writel(chanerr, chan->reg_base + IOAT_CHANERR_OFFSET);
+
+	return ioat2_reset_sync(chan, msecs_to_jiffies(200));
+}
+
 /**
  * ioat2_enumerate_channels - find and initialize the device's channels
  * @device: the device to be enumerated
@@ -360,6 +403,10 @@ int ioat2_enumerate_channels(struct ioatdma_device *device)
 				  (unsigned long) ioat);
 		ioat->xfercap_log = xfercap_log;
 		spin_lock_init(&ioat->ring_lock);
+		if (device->reset_hw(&ioat->base)) {
+			i = 0;
+			break;
+		}
 	}
 	dma->chancnt = i;
 	return i;
@@ -467,7 +514,6 @@ int ioat2_alloc_chan_resources(struct dma_chan *c)
 	struct ioat2_dma_chan *ioat = to_ioat2_chan(c);
 	struct ioat_chan_common *chan = &ioat->base;
 	struct ioat_ring_ent **ring;
-	u32 chanerr;
 	int order;
 
 	/* have we already been set up? */
@@ -477,12 +523,6 @@ int ioat2_alloc_chan_resources(struct dma_chan *c)
 	/* Setup register to interrupt and write completion status on error */
 	writew(IOAT_CHANCTRL_RUN, chan->reg_base + IOAT_CHANCTRL_OFFSET);
 
-	chanerr = readl(chan->reg_base + IOAT_CHANERR_OFFSET);
-	if (chanerr) {
-		dev_err(to_dev(chan), "CHANERR = %x, clearing\n", chanerr);
-		writel(chanerr, chan->reg_base + IOAT_CHANERR_OFFSET);
-	}
-
 	/* allocate a completion writeback area */
 	/* doing 2 32bit writes to mmio since 1 64b write doesn't work */
 	chan->completion = pci_pool_alloc(chan->device->completion_pool,
@@ -746,13 +786,7 @@ void ioat2_free_chan_resources(struct dma_chan *c)
 	tasklet_disable(&chan->cleanup_task);
 	del_timer_sync(&chan->timer);
 	device->cleanup_tasklet((unsigned long) ioat);
-
-	/* Delay 100ms after reset to allow internal DMA logic to quiesce
-	 * before removing DMA descriptor resources.
-	 */
-	writeb(IOAT_CHANCMD_RESET,
-	       chan->reg_base + IOAT_CHANCMD_OFFSET(chan->device->version));
-	mdelay(100);
+	device->reset_hw(chan);
 
 	spin_lock_bh(&ioat->ring_lock);
 	descs = ioat2_ring_space(ioat);
@@ -839,6 +873,7 @@ int __devinit ioat2_dma_probe(struct ioatdma_device *device, int dca)
 	int err;
 
 	device->enumerate_channels = ioat2_enumerate_channels;
+	device->reset_hw = ioat2_reset_hw;
 	device->cleanup_tasklet = ioat2_cleanup_tasklet;
 	device->timer_fn = ioat2_timer_event;
 	device->self_test = ioat_dma_self_test;
diff --git a/drivers/dma/ioat/dma_v2.h b/drivers/dma/ioat/dma_v2.h
index 1d849ef74d5f..3afad8da43cc 100644
--- a/drivers/dma/ioat/dma_v2.h
+++ b/drivers/dma/ioat/dma_v2.h
@@ -185,6 +185,8 @@ bool reshape_ring(struct ioat2_dma_chan *ioat, int order);
 void __ioat2_issue_pending(struct ioat2_dma_chan *ioat);
 void ioat2_cleanup_tasklet(unsigned long data);
 void ioat2_timer_event(unsigned long data);
+int ioat2_quiesce(struct ioat_chan_common *chan, unsigned long tmo);
+int ioat2_reset_sync(struct ioat_chan_common *chan, unsigned long tmo);
 extern struct kobj_type ioat2_ktype;
 extern struct kmem_cache *ioat2_cache;
 #endif /* IOATDMA_V2_H */
diff --git a/drivers/dma/ioat/dma_v3.c b/drivers/dma/ioat/dma_v3.c
index 42f6f10fb0cc..9908c9e94b2d 100644
--- a/drivers/dma/ioat/dma_v3.c
+++ b/drivers/dma/ioat/dma_v3.c
@@ -650,9 +650,11 @@ __ioat3_prep_pq_lock(struct dma_chan *c, enum sum_check_flags *result,
 
 	num_descs = ioat2_xferlen_to_descs(ioat, len);
 	/* we need 2x the number of descriptors to cover greater than 3
-	 * sources
+	 * sources (we need 1 extra source in the q-only continuation
+	 * case and 3 extra sources in the p+q continuation case.
 	 */
-	if (src_cnt > 3 || flags & DMA_PREP_CONTINUE) {
+	if (src_cnt + dmaf_p_disabled_continue(flags) > 3 ||
+	    (dmaf_continue(flags) && !dmaf_p_disabled_continue(flags))) {
 		with_ext = 1;
 		num_descs *= 2;
 	} else
@@ -1128,6 +1130,45 @@ static int __devinit ioat3_dma_self_test(struct ioatdma_device *device)
 	return 0;
 }
 
+static int ioat3_reset_hw(struct ioat_chan_common *chan)
+{
+	/* throw away whatever the channel was doing and get it
+	 * initialized, with ioat3 specific workarounds
+	 */
+	struct ioatdma_device *device = chan->device;
+	struct pci_dev *pdev = device->pdev;
+	u32 chanerr;
+	u16 dev_id;
+	int err;
+
+	ioat2_quiesce(chan, msecs_to_jiffies(100));
+
+	chanerr = readl(chan->reg_base + IOAT_CHANERR_OFFSET);
+	writel(chanerr, chan->reg_base + IOAT_CHANERR_OFFSET);
+
+	/* -= IOAT ver.3 workarounds =- */
+	/* Write CHANERRMSK_INT with 3E07h to mask out the errors
+	 * that can cause stability issues for IOAT ver.3, and clear any
+	 * pending errors
+	 */
+	pci_write_config_dword(pdev, IOAT_PCI_CHANERRMASK_INT_OFFSET, 0x3e07);
+	err = pci_read_config_dword(pdev, IOAT_PCI_CHANERR_INT_OFFSET, &chanerr);
+	if (err) {
+		dev_err(&pdev->dev, "channel error register unreachable\n");
+		return err;
+	}
+	pci_write_config_dword(pdev, IOAT_PCI_CHANERR_INT_OFFSET, chanerr);
+
+	/* Clear DMAUNCERRSTS Cfg-Reg Parity Error status bit
+	 * (workaround for spurious config parity error after restart)
+	 */
+	pci_read_config_word(pdev, IOAT_PCI_DEVICE_ID_OFFSET, &dev_id);
+	if (dev_id == PCI_DEVICE_ID_INTEL_IOAT_TBG0)
+		pci_write_config_dword(pdev, IOAT_PCI_DMAUNCERRSTS_OFFSET, 0x10);
+
+	return ioat2_reset_sync(chan, msecs_to_jiffies(200));
+}
+
 int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 {
 	struct pci_dev *pdev = device->pdev;
@@ -1137,10 +1178,10 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 	struct ioat_chan_common *chan;
 	bool is_raid_device = false;
 	int err;
-	u16 dev_id;
 	u32 cap;
 
 	device->enumerate_channels = ioat2_enumerate_channels;
+	device->reset_hw = ioat3_reset_hw;
 	device->self_test = ioat3_dma_self_test;
 	dma = &device->common;
 	dma->device_prep_dma_memcpy = ioat2_dma_prep_memcpy_lock;
@@ -1216,19 +1257,6 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 	dma->device_prep_dma_xor_val = NULL;
 	#endif
 
-	/* -= IOAT ver.3 workarounds =- */
-	/* Write CHANERRMSK_INT with 3E07h to mask out the errors
-	 * that can cause stability issues for IOAT ver.3
-	 */
-	pci_write_config_dword(pdev, IOAT_PCI_CHANERRMASK_INT_OFFSET, 0x3e07);
-
-	/* Clear DMAUNCERRSTS Cfg-Reg Parity Error status bit
-	 * (workaround for spurious config parity error after restart)
-	 */
-	pci_read_config_word(pdev, IOAT_PCI_DEVICE_ID_OFFSET, &dev_id);
-	if (dev_id == PCI_DEVICE_ID_INTEL_IOAT_TBG0)
-		pci_write_config_dword(pdev, IOAT_PCI_DMAUNCERRSTS_OFFSET, 0x10);
-
 	err = ioat_probe(device);
 	if (err)
 		return err;
diff --git a/drivers/dma/ioat/registers.h b/drivers/dma/ioat/registers.h
index f015ec196700..e8ae63baf588 100644
--- a/drivers/dma/ioat/registers.h
+++ b/drivers/dma/ioat/registers.h
@@ -27,6 +27,7 @@
 
 #define IOAT_PCI_DEVICE_ID_OFFSET		0x02
 #define IOAT_PCI_DMAUNCERRSTS_OFFSET		0x148
+#define IOAT_PCI_CHANERR_INT_OFFSET		0x180
 #define IOAT_PCI_CHANERRMASK_INT_OFFSET		0x184
 
 /* MMIO Device Registers */
diff --git a/drivers/dma/iop-adma.c b/drivers/dma/iop-adma.c
index 645ca8d54ec4..ca6e6a0cb793 100644
--- a/drivers/dma/iop-adma.c
+++ b/drivers/dma/iop-adma.c
@@ -1470,7 +1470,7 @@ static int __devinit iop_adma_probe(struct platform_device *pdev)
 		return -ENODEV;
 
 	if (!devm_request_mem_region(&pdev->dev, res->start,
-				res->end - res->start, pdev->name))
+				resource_size(res), pdev->name))
 		return -EBUSY;
 
 	adev = kzalloc(sizeof(*adev), GFP_KERNEL);
@@ -1542,7 +1542,7 @@ static int __devinit iop_adma_probe(struct platform_device *pdev)
 	iop_chan->device = adev;
 
 	iop_chan->mmr_base = devm_ioremap(&pdev->dev, res->start,
-					res->end - res->start);
+					resource_size(res));
 	if (!iop_chan->mmr_base) {
 		ret = -ENOMEM;
 		goto err_free_iop_chan;
diff --git a/drivers/dma/ppc4xx/Makefile b/drivers/dma/ppc4xx/Makefile
new file mode 100644
index 000000000000..b3d259b3e52a
--- /dev/null
+++ b/drivers/dma/ppc4xx/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_AMCC_PPC440SPE_ADMA) += adma.o
diff --git a/drivers/dma/ppc4xx/adma.c b/drivers/dma/ppc4xx/adma.c
new file mode 100644
index 000000000000..0a3478e910f0
--- /dev/null
+++ b/drivers/dma/ppc4xx/adma.c
@@ -0,0 +1,5027 @@
+/*
+ * Copyright (C) 2006-2009 DENX Software Engineering.
+ *
+ * Author: Yuri Tikhonov <yur@emcraft.com>
+ *
+ * Further porting to arch/powerpc by
+ * 	Anatolij Gustschin <agust@denx.de>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ * 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., 59
+ * Temple Place - Suite 330, Boston, MA  02111-1307, USA.
+ *
+ * The full GNU General Public License is included in this distribution in the
+ * file called COPYING.
+ */
+
+/*
+ * This driver supports the asynchrounous DMA copy and RAID engines available
+ * on the AMCC PPC440SPe Processors.
+ * Based on the Intel Xscale(R) family of I/O Processors (IOP 32x, 33x, 134x)
+ * ADMA driver written by D.Williams.
+ */
+
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/async_tx.h>
+#include <linux/delay.h>
+#include <linux/dma-mapping.h>
+#include <linux/spinlock.h>
+#include <linux/interrupt.h>
+#include <linux/uaccess.h>
+#include <linux/proc_fs.h>
+#include <linux/of.h>
+#include <linux/of_platform.h>
+#include <asm/dcr.h>
+#include <asm/dcr-regs.h>
+#include "adma.h"
+
+enum ppc_adma_init_code {
+	PPC_ADMA_INIT_OK = 0,
+	PPC_ADMA_INIT_MEMRES,
+	PPC_ADMA_INIT_MEMREG,
+	PPC_ADMA_INIT_ALLOC,
+	PPC_ADMA_INIT_COHERENT,
+	PPC_ADMA_INIT_CHANNEL,
+	PPC_ADMA_INIT_IRQ1,
+	PPC_ADMA_INIT_IRQ2,
+	PPC_ADMA_INIT_REGISTER
+};
+
+static char *ppc_adma_errors[] = {
+	[PPC_ADMA_INIT_OK] = "ok",
+	[PPC_ADMA_INIT_MEMRES] = "failed to get memory resource",
+	[PPC_ADMA_INIT_MEMREG] = "failed to request memory region",
+	[PPC_ADMA_INIT_ALLOC] = "failed to allocate memory for adev "
+				"structure",
+	[PPC_ADMA_INIT_COHERENT] = "failed to allocate coherent memory for "
+				   "hardware descriptors",
+	[PPC_ADMA_INIT_CHANNEL] = "failed to allocate memory for channel",
+	[PPC_ADMA_INIT_IRQ1] = "failed to request first irq",
+	[PPC_ADMA_INIT_IRQ2] = "failed to request second irq",
+	[PPC_ADMA_INIT_REGISTER] = "failed to register dma async device",
+};
+
+static enum ppc_adma_init_code
+ppc440spe_adma_devices[PPC440SPE_ADMA_ENGINES_NUM];
+
+struct ppc_dma_chan_ref {
+	struct dma_chan *chan;
+	struct list_head node;
+};
+
+/* The list of channels exported by ppc440spe ADMA */
+struct list_head
+ppc440spe_adma_chan_list = LIST_HEAD_INIT(ppc440spe_adma_chan_list);
+
+/* This flag is set when want to refetch the xor chain in the interrupt
+ * handler
+ */
+static u32 do_xor_refetch;
+
+/* Pointer to DMA0, DMA1 CP/CS FIFO */
+static void *ppc440spe_dma_fifo_buf;
+
+/* Pointers to last submitted to DMA0, DMA1 CDBs */
+static struct ppc440spe_adma_desc_slot *chan_last_sub[3];
+static struct ppc440spe_adma_desc_slot *chan_first_cdb[3];
+
+/* Pointer to last linked and submitted xor CB */
+static struct ppc440spe_adma_desc_slot *xor_last_linked;
+static struct ppc440spe_adma_desc_slot *xor_last_submit;
+
+/* This array is used in data-check operations for storing a pattern */
+static char ppc440spe_qword[16];
+
+static atomic_t ppc440spe_adma_err_irq_ref;
+static dcr_host_t ppc440spe_mq_dcr_host;
+static unsigned int ppc440spe_mq_dcr_len;
+
+/* Since RXOR operations use the common register (MQ0_CF2H) for setting-up
+ * the block size in transactions, then we do not allow to activate more than
+ * only one RXOR transactions simultaneously. So use this var to store
+ * the information about is RXOR currently active (PPC440SPE_RXOR_RUN bit is
+ * set) or not (PPC440SPE_RXOR_RUN is clear).
+ */
+static unsigned long ppc440spe_rxor_state;
+
+/* These are used in enable & check routines
+ */
+static u32 ppc440spe_r6_enabled;
+static struct ppc440spe_adma_chan *ppc440spe_r6_tchan;
+static struct completion ppc440spe_r6_test_comp;
+
+static int ppc440spe_adma_dma2rxor_prep_src(
+		struct ppc440spe_adma_desc_slot *desc,
+		struct ppc440spe_rxor *cursor, int index,
+		int src_cnt, u32 addr);
+static void ppc440spe_adma_dma2rxor_set_src(
+		struct ppc440spe_adma_desc_slot *desc,
+		int index, dma_addr_t addr);
+static void ppc440spe_adma_dma2rxor_set_mult(
+		struct ppc440spe_adma_desc_slot *desc,
+		int index, u8 mult);
+
+#ifdef ADMA_LL_DEBUG
+#define ADMA_LL_DBG(x) ({ if (1) x; 0; })
+#else
+#define ADMA_LL_DBG(x) ({ if (0) x; 0; })
+#endif
+
+static void print_cb(struct ppc440spe_adma_chan *chan, void *block)
+{
+	struct dma_cdb *cdb;
+	struct xor_cb *cb;
+	int i;
+
+	switch (chan->device->id) {
+	case 0:
+	case 1:
+		cdb = block;
+
+		pr_debug("CDB at %p [%d]:\n"
+			"\t attr 0x%02x opc 0x%02x cnt 0x%08x\n"
+			"\t sg1u 0x%08x sg1l 0x%08x\n"
+			"\t sg2u 0x%08x sg2l 0x%08x\n"
+			"\t sg3u 0x%08x sg3l 0x%08x\n",
+			cdb, chan->device->id,
+			cdb->attr, cdb->opc, le32_to_cpu(cdb->cnt),
+			le32_to_cpu(cdb->sg1u), le32_to_cpu(cdb->sg1l),
+			le32_to_cpu(cdb->sg2u), le32_to_cpu(cdb->sg2l),
+			le32_to_cpu(cdb->sg3u), le32_to_cpu(cdb->sg3l)
+		);
+		break;
+	case 2:
+		cb = block;
+
+		pr_debug("CB at %p [%d]:\n"
+			"\t cbc 0x%08x cbbc 0x%08x cbs 0x%08x\n"
+			"\t cbtah 0x%08x cbtal 0x%08x\n"
+			"\t cblah 0x%08x cblal 0x%08x\n",
+			cb, chan->device->id,
+			cb->cbc, cb->cbbc, cb->cbs,
+			cb->cbtah, cb->cbtal,
+			cb->cblah, cb->cblal);
+		for (i = 0; i < 16; i++) {
+			if (i && !cb->ops[i].h && !cb->ops[i].l)
+				continue;
+			pr_debug("\t ops[%2d]: h 0x%08x l 0x%08x\n",
+				i, cb->ops[i].h, cb->ops[i].l);
+		}
+		break;
+	}
+}
+
+static void print_cb_list(struct ppc440spe_adma_chan *chan,
+			  struct ppc440spe_adma_desc_slot *iter)
+{
+	for (; iter; iter = iter->hw_next)
+		print_cb(chan, iter->hw_desc);
+}
+
+static void prep_dma_xor_dbg(int id, dma_addr_t dst, dma_addr_t *src,
+			     unsigned int src_cnt)
+{
+	int i;
+
+	pr_debug("\n%s(%d):\nsrc: ", __func__, id);
+	for (i = 0; i < src_cnt; i++)
+		pr_debug("\t0x%016llx ", src[i]);
+	pr_debug("dst:\n\t0x%016llx\n", dst);
+}
+
+static void prep_dma_pq_dbg(int id, dma_addr_t *dst, dma_addr_t *src,
+			    unsigned int src_cnt)
+{
+	int i;
+
+	pr_debug("\n%s(%d):\nsrc: ", __func__, id);
+	for (i = 0; i < src_cnt; i++)
+		pr_debug("\t0x%016llx ", src[i]);
+	pr_debug("dst: ");
+	for (i = 0; i < 2; i++)
+		pr_debug("\t0x%016llx ", dst[i]);
+}
+
+static void prep_dma_pqzero_sum_dbg(int id, dma_addr_t *src,
+				    unsigned int src_cnt,
+				    const unsigned char *scf)
+{
+	int i;
+
+	pr_debug("\n%s(%d):\nsrc(coef): ", __func__, id);
+	if (scf) {
+		for (i = 0; i < src_cnt; i++)
+			pr_debug("\t0x%016llx(0x%02x) ", src[i], scf[i]);
+	} else {
+		for (i = 0; i < src_cnt; i++)
+			pr_debug("\t0x%016llx(no) ", src[i]);
+	}
+
+	pr_debug("dst: ");
+	for (i = 0; i < 2; i++)
+		pr_debug("\t0x%016llx ", src[src_cnt + i]);
+}
+
+/******************************************************************************
+ * Command (Descriptor) Blocks low-level routines
+ ******************************************************************************/
+/**
+ * ppc440spe_desc_init_interrupt - initialize the descriptor for INTERRUPT
+ * pseudo operation
+ */
+static void ppc440spe_desc_init_interrupt(struct ppc440spe_adma_desc_slot *desc,
+					  struct ppc440spe_adma_chan *chan)
+{
+	struct xor_cb *p;
+
+	switch (chan->device->id) {
+	case PPC440SPE_XOR_ID:
+		p = desc->hw_desc;
+		memset(desc->hw_desc, 0, sizeof(struct xor_cb));
+		/* NOP with Command Block Complete Enable */
+		p->cbc = XOR_CBCR_CBCE_BIT;
+		break;
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		memset(desc->hw_desc, 0, sizeof(struct dma_cdb));
+		/* NOP with interrupt */
+		set_bit(PPC440SPE_DESC_INT, &desc->flags);
+		break;
+	default:
+		printk(KERN_ERR "Unsupported id %d in %s\n", chan->device->id,
+				__func__);
+		break;
+	}
+}
+
+/**
+ * ppc440spe_desc_init_null_xor - initialize the descriptor for NULL XOR
+ * pseudo operation
+ */
+static void ppc440spe_desc_init_null_xor(struct ppc440spe_adma_desc_slot *desc)
+{
+	memset(desc->hw_desc, 0, sizeof(struct xor_cb));
+	desc->hw_next = NULL;
+	desc->src_cnt = 0;
+	desc->dst_cnt = 1;
+}
+
+/**
+ * ppc440spe_desc_init_xor - initialize the descriptor for XOR operation
+ */
+static void ppc440spe_desc_init_xor(struct ppc440spe_adma_desc_slot *desc,
+					 int src_cnt, unsigned long flags)
+{
+	struct xor_cb *hw_desc = desc->hw_desc;
+
+	memset(desc->hw_desc, 0, sizeof(struct xor_cb));
+	desc->hw_next = NULL;
+	desc->src_cnt = src_cnt;
+	desc->dst_cnt = 1;
+
+	hw_desc->cbc = XOR_CBCR_TGT_BIT | src_cnt;
+	if (flags & DMA_PREP_INTERRUPT)
+		/* Enable interrupt on completion */
+		hw_desc->cbc |= XOR_CBCR_CBCE_BIT;
+}
+
+/**
+ * ppc440spe_desc_init_dma2pq - initialize the descriptor for PQ
+ * operation in DMA2 controller
+ */
+static void ppc440spe_desc_init_dma2pq(struct ppc440spe_adma_desc_slot *desc,
+		int dst_cnt, int src_cnt, unsigned long flags)
+{
+	struct xor_cb *hw_desc = desc->hw_desc;
+
+	memset(desc->hw_desc, 0, sizeof(struct xor_cb));
+	desc->hw_next = NULL;
+	desc->src_cnt = src_cnt;
+	desc->dst_cnt = dst_cnt;
+	memset(desc->reverse_flags, 0, sizeof(desc->reverse_flags));
+	desc->descs_per_op = 0;
+
+	hw_desc->cbc = XOR_CBCR_TGT_BIT;
+	if (flags & DMA_PREP_INTERRUPT)
+		/* Enable interrupt on completion */
+		hw_desc->cbc |= XOR_CBCR_CBCE_BIT;
+}
+
+#define DMA_CTRL_FLAGS_LAST	DMA_PREP_FENCE
+#define DMA_PREP_ZERO_P		(DMA_CTRL_FLAGS_LAST << 1)
+#define DMA_PREP_ZERO_Q		(DMA_PREP_ZERO_P << 1)
+
+/**
+ * ppc440spe_desc_init_dma01pq - initialize the descriptors for PQ operation
+ * with DMA0/1
+ */
+static void ppc440spe_desc_init_dma01pq(struct ppc440spe_adma_desc_slot *desc,
+				int dst_cnt, int src_cnt, unsigned long flags,
+				unsigned long op)
+{
+	struct dma_cdb *hw_desc;
+	struct ppc440spe_adma_desc_slot *iter;
+	u8 dopc;
+
+	/* Common initialization of a PQ descriptors chain */
+	set_bits(op, &desc->flags);
+	desc->src_cnt = src_cnt;
+	desc->dst_cnt = dst_cnt;
+
+	/* WXOR MULTICAST if both P and Q are being computed
+	 * MV_SG1_SG2 if Q only
+	 */
+	dopc = (desc->dst_cnt == DMA_DEST_MAX_NUM) ?
+		DMA_CDB_OPC_MULTICAST : DMA_CDB_OPC_MV_SG1_SG2;
+
+	list_for_each_entry(iter, &desc->group_list, chain_node) {
+		hw_desc = iter->hw_desc;
+		memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+
+		if (likely(!list_is_last(&iter->chain_node,
+				&desc->group_list))) {
+			/* set 'next' pointer */
+			iter->hw_next = list_entry(iter->chain_node.next,
+				struct ppc440spe_adma_desc_slot, chain_node);
+			clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+		} else {
+			/* this is the last descriptor.
+			 * this slot will be pasted from ADMA level
+			 * each time it wants to configure parameters
+			 * of the transaction (src, dst, ...)
+			 */
+			iter->hw_next = NULL;
+			if (flags & DMA_PREP_INTERRUPT)
+				set_bit(PPC440SPE_DESC_INT, &iter->flags);
+			else
+				clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+		}
+	}
+
+	/* Set OPS depending on WXOR/RXOR type of operation */
+	if (!test_bit(PPC440SPE_DESC_RXOR, &desc->flags)) {
+		/* This is a WXOR only chain:
+		 * - first descriptors are for zeroing destinations
+		 *   if PPC440SPE_ZERO_P/Q set;
+		 * - descriptors remained are for GF-XOR operations.
+		 */
+		iter = list_first_entry(&desc->group_list,
+					struct ppc440spe_adma_desc_slot,
+					chain_node);
+
+		if (test_bit(PPC440SPE_ZERO_P, &desc->flags)) {
+			hw_desc = iter->hw_desc;
+			hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+			iter = list_first_entry(&iter->chain_node,
+					struct ppc440spe_adma_desc_slot,
+					chain_node);
+		}
+
+		if (test_bit(PPC440SPE_ZERO_Q, &desc->flags)) {
+			hw_desc = iter->hw_desc;
+			hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+			iter = list_first_entry(&iter->chain_node,
+					struct ppc440spe_adma_desc_slot,
+					chain_node);
+		}
+
+		list_for_each_entry_from(iter, &desc->group_list, chain_node) {
+			hw_desc = iter->hw_desc;
+			hw_desc->opc = dopc;
+		}
+	} else {
+		/* This is either RXOR-only or mixed RXOR/WXOR */
+
+		/* The first 1 or 2 slots in chain are always RXOR,
+		 * if need to calculate P & Q, then there are two
+		 * RXOR slots; if only P or only Q, then there is one
+		 */
+		iter = list_first_entry(&desc->group_list,
+					struct ppc440spe_adma_desc_slot,
+					chain_node);
+		hw_desc = iter->hw_desc;
+		hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+
+		if (desc->dst_cnt == DMA_DEST_MAX_NUM) {
+			iter = list_first_entry(&iter->chain_node,
+						struct ppc440spe_adma_desc_slot,
+						chain_node);
+			hw_desc = iter->hw_desc;
+			hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+		}
+
+		/* The remaining descs (if any) are WXORs */
+		if (test_bit(PPC440SPE_DESC_WXOR, &desc->flags)) {
+			iter = list_first_entry(&iter->chain_node,
+						struct ppc440spe_adma_desc_slot,
+						chain_node);
+			list_for_each_entry_from(iter, &desc->group_list,
+						chain_node) {
+				hw_desc = iter->hw_desc;
+				hw_desc->opc = dopc;
+			}
+		}
+	}
+}
+
+/**
+ * ppc440spe_desc_init_dma01pqzero_sum - initialize the descriptor
+ * for PQ_ZERO_SUM operation
+ */
+static void ppc440spe_desc_init_dma01pqzero_sum(
+				struct ppc440spe_adma_desc_slot *desc,
+				int dst_cnt, int src_cnt)
+{
+	struct dma_cdb *hw_desc;
+	struct ppc440spe_adma_desc_slot *iter;
+	int i = 0;
+	u8 dopc = (dst_cnt == 2) ? DMA_CDB_OPC_MULTICAST :
+				   DMA_CDB_OPC_MV_SG1_SG2;
+	/*
+	 * Initialize starting from 2nd or 3rd descriptor dependent
+	 * on dst_cnt. First one or two slots are for cloning P
+	 * and/or Q to chan->pdest and/or chan->qdest as we have
+	 * to preserve original P/Q.
+	 */
+	iter = list_first_entry(&desc->group_list,
+				struct ppc440spe_adma_desc_slot, chain_node);
+	iter = list_entry(iter->chain_node.next,
+			  struct ppc440spe_adma_desc_slot, chain_node);
+
+	if (dst_cnt > 1) {
+		iter = list_entry(iter->chain_node.next,
+				  struct ppc440spe_adma_desc_slot, chain_node);
+	}
+	/* initialize each source descriptor in chain */
+	list_for_each_entry_from(iter, &desc->group_list, chain_node) {
+		hw_desc = iter->hw_desc;
+		memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+		iter->src_cnt = 0;
+		iter->dst_cnt = 0;
+
+		/* This is a ZERO_SUM operation:
+		 * - <src_cnt> descriptors starting from 2nd or 3rd
+		 *   descriptor are for GF-XOR operations;
+		 * - remaining <dst_cnt> descriptors are for checking the result
+		 */
+		if (i++ < src_cnt)
+			/* MV_SG1_SG2 if only Q is being verified
+			 * MULTICAST if both P and Q are being verified
+			 */
+			hw_desc->opc = dopc;
+		else
+			/* DMA_CDB_OPC_DCHECK128 operation */
+			hw_desc->opc = DMA_CDB_OPC_DCHECK128;
+
+		if (likely(!list_is_last(&iter->chain_node,
+					 &desc->group_list))) {
+			/* set 'next' pointer */
+			iter->hw_next = list_entry(iter->chain_node.next,
+						struct ppc440spe_adma_desc_slot,
+						chain_node);
+		} else {
+			/* this is the last descriptor.
+			 * this slot will be pasted from ADMA level
+			 * each time it wants to configure parameters
+			 * of the transaction (src, dst, ...)
+			 */
+			iter->hw_next = NULL;
+			/* always enable interrupt generation since we get
+			 * the status of pqzero from the handler
+			 */
+			set_bit(PPC440SPE_DESC_INT, &iter->flags);
+		}
+	}
+	desc->src_cnt = src_cnt;
+	desc->dst_cnt = dst_cnt;
+}
+
+/**
+ * ppc440spe_desc_init_memcpy - initialize the descriptor for MEMCPY operation
+ */
+static void ppc440spe_desc_init_memcpy(struct ppc440spe_adma_desc_slot *desc,
+					unsigned long flags)
+{
+	struct dma_cdb *hw_desc = desc->hw_desc;
+
+	memset(desc->hw_desc, 0, sizeof(struct dma_cdb));
+	desc->hw_next = NULL;
+	desc->src_cnt = 1;
+	desc->dst_cnt = 1;
+
+	if (flags & DMA_PREP_INTERRUPT)
+		set_bit(PPC440SPE_DESC_INT, &desc->flags);
+	else
+		clear_bit(PPC440SPE_DESC_INT, &desc->flags);
+
+	hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+}
+
+/**
+ * ppc440spe_desc_init_memset - initialize the descriptor for MEMSET operation
+ */
+static void ppc440spe_desc_init_memset(struct ppc440spe_adma_desc_slot *desc,
+					int value, unsigned long flags)
+{
+	struct dma_cdb *hw_desc = desc->hw_desc;
+
+	memset(desc->hw_desc, 0, sizeof(struct dma_cdb));
+	desc->hw_next = NULL;
+	desc->src_cnt = 1;
+	desc->dst_cnt = 1;
+
+	if (flags & DMA_PREP_INTERRUPT)
+		set_bit(PPC440SPE_DESC_INT, &desc->flags);
+	else
+		clear_bit(PPC440SPE_DESC_INT, &desc->flags);
+
+	hw_desc->sg1u = hw_desc->sg1l = cpu_to_le32((u32)value);
+	hw_desc->sg3u = hw_desc->sg3l = cpu_to_le32((u32)value);
+	hw_desc->opc = DMA_CDB_OPC_DFILL128;
+}
+
+/**
+ * ppc440spe_desc_set_src_addr - set source address into the descriptor
+ */
+static void ppc440spe_desc_set_src_addr(struct ppc440spe_adma_desc_slot *desc,
+					struct ppc440spe_adma_chan *chan,
+					int src_idx, dma_addr_t addrh,
+					dma_addr_t addrl)
+{
+	struct dma_cdb *dma_hw_desc;
+	struct xor_cb *xor_hw_desc;
+	phys_addr_t addr64, tmplow, tmphi;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		if (!addrh) {
+			addr64 = addrl;
+			tmphi = (addr64 >> 32);
+			tmplow = (addr64 & 0xFFFFFFFF);
+		} else {
+			tmphi = addrh;
+			tmplow = addrl;
+		}
+		dma_hw_desc = desc->hw_desc;
+		dma_hw_desc->sg1l = cpu_to_le32((u32)tmplow);
+		dma_hw_desc->sg1u |= cpu_to_le32((u32)tmphi);
+		break;
+	case PPC440SPE_XOR_ID:
+		xor_hw_desc = desc->hw_desc;
+		xor_hw_desc->ops[src_idx].l = addrl;
+		xor_hw_desc->ops[src_idx].h |= addrh;
+		break;
+	}
+}
+
+/**
+ * ppc440spe_desc_set_src_mult - set source address mult into the descriptor
+ */
+static void ppc440spe_desc_set_src_mult(struct ppc440spe_adma_desc_slot *desc,
+			struct ppc440spe_adma_chan *chan, u32 mult_index,
+			int sg_index, unsigned char mult_value)
+{
+	struct dma_cdb *dma_hw_desc;
+	struct xor_cb *xor_hw_desc;
+	u32 *psgu;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		dma_hw_desc = desc->hw_desc;
+
+		switch (sg_index) {
+		/* for RXOR operations set multiplier
+		 * into source cued address
+		 */
+		case DMA_CDB_SG_SRC:
+			psgu = &dma_hw_desc->sg1u;
+			break;
+		/* for WXOR operations set multiplier
+		 * into destination cued address(es)
+		 */
+		case DMA_CDB_SG_DST1:
+			psgu = &dma_hw_desc->sg2u;
+			break;
+		case DMA_CDB_SG_DST2:
+			psgu = &dma_hw_desc->sg3u;
+			break;
+		default:
+			BUG();
+		}
+
+		*psgu |= cpu_to_le32(mult_value << mult_index);
+		break;
+	case PPC440SPE_XOR_ID:
+		xor_hw_desc = desc->hw_desc;
+		break;
+	default:
+		BUG();
+	}
+}
+
+/**
+ * ppc440spe_desc_set_dest_addr - set destination address into the descriptor
+ */
+static void ppc440spe_desc_set_dest_addr(struct ppc440spe_adma_desc_slot *desc,
+				struct ppc440spe_adma_chan *chan,
+				dma_addr_t addrh, dma_addr_t addrl,
+				u32 dst_idx)
+{
+	struct dma_cdb *dma_hw_desc;
+	struct xor_cb *xor_hw_desc;
+	phys_addr_t addr64, tmphi, tmplow;
+	u32 *psgu, *psgl;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		if (!addrh) {
+			addr64 = addrl;
+			tmphi = (addr64 >> 32);
+			tmplow = (addr64 & 0xFFFFFFFF);
+		} else {
+			tmphi = addrh;
+			tmplow = addrl;
+		}
+		dma_hw_desc = desc->hw_desc;
+
+		psgu = dst_idx ? &dma_hw_desc->sg3u : &dma_hw_desc->sg2u;
+		psgl = dst_idx ? &dma_hw_desc->sg3l : &dma_hw_desc->sg2l;
+
+		*psgl = cpu_to_le32((u32)tmplow);
+		*psgu |= cpu_to_le32((u32)tmphi);
+		break;
+	case PPC440SPE_XOR_ID:
+		xor_hw_desc = desc->hw_desc;
+		xor_hw_desc->cbtal = addrl;
+		xor_hw_desc->cbtah |= addrh;
+		break;
+	}
+}
+
+/**
+ * ppc440spe_desc_set_byte_count - set number of data bytes involved
+ * into the operation
+ */
+static void ppc440spe_desc_set_byte_count(struct ppc440spe_adma_desc_slot *desc,
+				struct ppc440spe_adma_chan *chan,
+				u32 byte_count)
+{
+	struct dma_cdb *dma_hw_desc;
+	struct xor_cb *xor_hw_desc;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		dma_hw_desc = desc->hw_desc;
+		dma_hw_desc->cnt = cpu_to_le32(byte_count);
+		break;
+	case PPC440SPE_XOR_ID:
+		xor_hw_desc = desc->hw_desc;
+		xor_hw_desc->cbbc = byte_count;
+		break;
+	}
+}
+
+/**
+ * ppc440spe_desc_set_rxor_block_size - set RXOR block size
+ */
+static inline void ppc440spe_desc_set_rxor_block_size(u32 byte_count)
+{
+	/* assume that byte_count is aligned on the 512-boundary;
+	 * thus write it directly to the register (bits 23:31 are
+	 * reserved there).
+	 */
+	dcr_write(ppc440spe_mq_dcr_host, DCRN_MQ0_CF2H, byte_count);
+}
+
+/**
+ * ppc440spe_desc_set_dcheck - set CHECK pattern
+ */
+static void ppc440spe_desc_set_dcheck(struct ppc440spe_adma_desc_slot *desc,
+				struct ppc440spe_adma_chan *chan, u8 *qword)
+{
+	struct dma_cdb *dma_hw_desc;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		dma_hw_desc = desc->hw_desc;
+		iowrite32(qword[0], &dma_hw_desc->sg3l);
+		iowrite32(qword[4], &dma_hw_desc->sg3u);
+		iowrite32(qword[8], &dma_hw_desc->sg2l);
+		iowrite32(qword[12], &dma_hw_desc->sg2u);
+		break;
+	default:
+		BUG();
+	}
+}
+
+/**
+ * ppc440spe_xor_set_link - set link address in xor CB
+ */
+static void ppc440spe_xor_set_link(struct ppc440spe_adma_desc_slot *prev_desc,
+				struct ppc440spe_adma_desc_slot *next_desc)
+{
+	struct xor_cb *xor_hw_desc = prev_desc->hw_desc;
+
+	if (unlikely(!next_desc || !(next_desc->phys))) {
+		printk(KERN_ERR "%s: next_desc=0x%p; next_desc->phys=0x%llx\n",
+			__func__, next_desc,
+			next_desc ? next_desc->phys : 0);
+		BUG();
+	}
+
+	xor_hw_desc->cbs = 0;
+	xor_hw_desc->cblal = next_desc->phys;
+	xor_hw_desc->cblah = 0;
+	xor_hw_desc->cbc |= XOR_CBCR_LNK_BIT;
+}
+
+/**
+ * ppc440spe_desc_set_link - set the address of descriptor following this
+ * descriptor in chain
+ */
+static void ppc440spe_desc_set_link(struct ppc440spe_adma_chan *chan,
+				struct ppc440spe_adma_desc_slot *prev_desc,
+				struct ppc440spe_adma_desc_slot *next_desc)
+{
+	unsigned long flags;
+	struct ppc440spe_adma_desc_slot *tail = next_desc;
+
+	if (unlikely(!prev_desc || !next_desc ||
+		(prev_desc->hw_next && prev_desc->hw_next != next_desc))) {
+		/* If previous next is overwritten something is wrong.
+		 * though we may refetch from append to initiate list
+		 * processing; in this case - it's ok.
+		 */
+		printk(KERN_ERR "%s: prev_desc=0x%p; next_desc=0x%p; "
+			"prev->hw_next=0x%p\n", __func__, prev_desc,
+			next_desc, prev_desc ? prev_desc->hw_next : 0);
+		BUG();
+	}
+
+	local_irq_save(flags);
+
+	/* do s/w chaining both for DMA and XOR descriptors */
+	prev_desc->hw_next = next_desc;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		break;
+	case PPC440SPE_XOR_ID:
+		/* bind descriptor to the chain */
+		while (tail->hw_next)
+			tail = tail->hw_next;
+		xor_last_linked = tail;
+
+		if (prev_desc == xor_last_submit)
+			/* do not link to the last submitted CB */
+			break;
+		ppc440spe_xor_set_link(prev_desc, next_desc);
+		break;
+	}
+
+	local_irq_restore(flags);
+}
+
+/**
+ * ppc440spe_desc_get_src_addr - extract the source address from the descriptor
+ */
+static u32 ppc440spe_desc_get_src_addr(struct ppc440spe_adma_desc_slot *desc,
+				struct ppc440spe_adma_chan *chan, int src_idx)
+{
+	struct dma_cdb *dma_hw_desc;
+	struct xor_cb *xor_hw_desc;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		dma_hw_desc = desc->hw_desc;
+		/* May have 0, 1, 2, or 3 sources */
+		switch (dma_hw_desc->opc) {
+		case DMA_CDB_OPC_NO_OP:
+		case DMA_CDB_OPC_DFILL128:
+			return 0;
+		case DMA_CDB_OPC_DCHECK128:
+			if (unlikely(src_idx)) {
+				printk(KERN_ERR "%s: try to get %d source for"
+				    " DCHECK128\n", __func__, src_idx);
+				BUG();
+			}
+			return le32_to_cpu(dma_hw_desc->sg1l);
+		case DMA_CDB_OPC_MULTICAST:
+		case DMA_CDB_OPC_MV_SG1_SG2:
+			if (unlikely(src_idx > 2)) {
+				printk(KERN_ERR "%s: try to get %d source from"
+				    " DMA descr\n", __func__, src_idx);
+				BUG();
+			}
+			if (src_idx) {
+				if (le32_to_cpu(dma_hw_desc->sg1u) &
+				    DMA_CUED_XOR_WIN_MSK) {
+					u8 region;
+
+					if (src_idx == 1)
+						return le32_to_cpu(
+						    dma_hw_desc->sg1l) +
+							desc->unmap_len;
+
+					region = (le32_to_cpu(
+					    dma_hw_desc->sg1u)) >>
+						DMA_CUED_REGION_OFF;
+
+					region &= DMA_CUED_REGION_MSK;
+					switch (region) {
+					case DMA_RXOR123:
+						return le32_to_cpu(
+						    dma_hw_desc->sg1l) +
+							(desc->unmap_len << 1);
+					case DMA_RXOR124:
+						return le32_to_cpu(
+						    dma_hw_desc->sg1l) +
+							(desc->unmap_len * 3);
+					case DMA_RXOR125:
+						return le32_to_cpu(
+						    dma_hw_desc->sg1l) +
+							(desc->unmap_len << 2);
+					default:
+						printk(KERN_ERR
+						    "%s: try to"
+						    " get src3 for region %02x"
+						    "PPC440SPE_DESC_RXOR12?\n",
+						    __func__, region);
+						BUG();
+					}
+				} else {
+					printk(KERN_ERR
+						"%s: try to get %d"
+						" source for non-cued descr\n",
+						__func__, src_idx);
+					BUG();
+				}
+			}
+			return le32_to_cpu(dma_hw_desc->sg1l);
+		default:
+			printk(KERN_ERR "%s: unknown OPC 0x%02x\n",
+				__func__, dma_hw_desc->opc);
+			BUG();
+		}
+		return le32_to_cpu(dma_hw_desc->sg1l);
+	case PPC440SPE_XOR_ID:
+		/* May have up to 16 sources */
+		xor_hw_desc = desc->hw_desc;
+		return xor_hw_desc->ops[src_idx].l;
+	}
+	return 0;
+}
+
+/**
+ * ppc440spe_desc_get_dest_addr - extract the destination address from the
+ * descriptor
+ */
+static u32 ppc440spe_desc_get_dest_addr(struct ppc440spe_adma_desc_slot *desc,
+				struct ppc440spe_adma_chan *chan, int idx)
+{
+	struct dma_cdb *dma_hw_desc;
+	struct xor_cb *xor_hw_desc;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		dma_hw_desc = desc->hw_desc;
+
+		if (likely(!idx))
+			return le32_to_cpu(dma_hw_desc->sg2l);
+		return le32_to_cpu(dma_hw_desc->sg3l);
+	case PPC440SPE_XOR_ID:
+		xor_hw_desc = desc->hw_desc;
+		return xor_hw_desc->cbtal;
+	}
+	return 0;
+}
+
+/**
+ * ppc440spe_desc_get_src_num - extract the number of source addresses from
+ * the descriptor
+ */
+static u32 ppc440spe_desc_get_src_num(struct ppc440spe_adma_desc_slot *desc,
+				struct ppc440spe_adma_chan *chan)
+{
+	struct dma_cdb *dma_hw_desc;
+	struct xor_cb *xor_hw_desc;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		dma_hw_desc = desc->hw_desc;
+
+		switch (dma_hw_desc->opc) {
+		case DMA_CDB_OPC_NO_OP:
+		case DMA_CDB_OPC_DFILL128:
+			return 0;
+		case DMA_CDB_OPC_DCHECK128:
+			return 1;
+		case DMA_CDB_OPC_MV_SG1_SG2:
+		case DMA_CDB_OPC_MULTICAST:
+			/*
+			 * Only for RXOR operations we have more than
+			 * one source
+			 */
+			if (le32_to_cpu(dma_hw_desc->sg1u) &
+			    DMA_CUED_XOR_WIN_MSK) {
+				/* RXOR op, there are 2 or 3 sources */
+				if (((le32_to_cpu(dma_hw_desc->sg1u) >>
+				    DMA_CUED_REGION_OFF) &
+				      DMA_CUED_REGION_MSK) == DMA_RXOR12) {
+					/* RXOR 1-2 */
+					return 2;
+				} else {
+					/* RXOR 1-2-3/1-2-4/1-2-5 */
+					return 3;
+				}
+			}
+			return 1;
+		default:
+			printk(KERN_ERR "%s: unknown OPC 0x%02x\n",
+				__func__, dma_hw_desc->opc);
+			BUG();
+		}
+	case PPC440SPE_XOR_ID:
+		/* up to 16 sources */
+		xor_hw_desc = desc->hw_desc;
+		return xor_hw_desc->cbc & XOR_CDCR_OAC_MSK;
+	default:
+		BUG();
+	}
+	return 0;
+}
+
+/**
+ * ppc440spe_desc_get_dst_num - get the number of destination addresses in
+ * this descriptor
+ */
+static u32 ppc440spe_desc_get_dst_num(struct ppc440spe_adma_desc_slot *desc,
+				struct ppc440spe_adma_chan *chan)
+{
+	struct dma_cdb *dma_hw_desc;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		/* May be 1 or 2 destinations */
+		dma_hw_desc = desc->hw_desc;
+		switch (dma_hw_desc->opc) {
+		case DMA_CDB_OPC_NO_OP:
+		case DMA_CDB_OPC_DCHECK128:
+			return 0;
+		case DMA_CDB_OPC_MV_SG1_SG2:
+		case DMA_CDB_OPC_DFILL128:
+			return 1;
+		case DMA_CDB_OPC_MULTICAST:
+			if (desc->dst_cnt == 2)
+				return 2;
+			else
+				return 1;
+		default:
+			printk(KERN_ERR "%s: unknown OPC 0x%02x\n",
+				__func__, dma_hw_desc->opc);
+			BUG();
+		}
+	case PPC440SPE_XOR_ID:
+		/* Always only 1 destination */
+		return 1;
+	default:
+		BUG();
+	}
+	return 0;
+}
+
+/**
+ * ppc440spe_desc_get_link - get the address of the descriptor that
+ * follows this one
+ */
+static inline u32 ppc440spe_desc_get_link(struct ppc440spe_adma_desc_slot *desc,
+					struct ppc440spe_adma_chan *chan)
+{
+	if (!desc->hw_next)
+		return 0;
+
+	return desc->hw_next->phys;
+}
+
+/**
+ * ppc440spe_desc_is_aligned - check alignment
+ */
+static inline int ppc440spe_desc_is_aligned(
+	struct ppc440spe_adma_desc_slot *desc, int num_slots)
+{
+	return (desc->idx & (num_slots - 1)) ? 0 : 1;
+}
+
+/**
+ * ppc440spe_chan_xor_slot_count - get the number of slots necessary for
+ * XOR operation
+ */
+static int ppc440spe_chan_xor_slot_count(size_t len, int src_cnt,
+			int *slots_per_op)
+{
+	int slot_cnt;
+
+	/* each XOR descriptor provides up to 16 source operands */
+	slot_cnt = *slots_per_op = (src_cnt + XOR_MAX_OPS - 1)/XOR_MAX_OPS;
+
+	if (likely(len <= PPC440SPE_ADMA_XOR_MAX_BYTE_COUNT))
+		return slot_cnt;
+
+	printk(KERN_ERR "%s: len %d > max %d !!\n",
+		__func__, len, PPC440SPE_ADMA_XOR_MAX_BYTE_COUNT);
+	BUG();
+	return slot_cnt;
+}
+
+/**
+ * ppc440spe_dma2_pq_slot_count - get the number of slots necessary for
+ * DMA2 PQ operation
+ */
+static int ppc440spe_dma2_pq_slot_count(dma_addr_t *srcs,
+		int src_cnt, size_t len)
+{
+	signed long long order = 0;
+	int state = 0;
+	int addr_count = 0;
+	int i;
+	for (i = 1; i < src_cnt; i++) {
+		dma_addr_t cur_addr = srcs[i];
+		dma_addr_t old_addr = srcs[i-1];
+		switch (state) {
+		case 0:
+			if (cur_addr == old_addr + len) {
+				/* direct RXOR */
+				order = 1;
+				state = 1;
+				if (i == src_cnt-1)
+					addr_count++;
+			} else if (old_addr == cur_addr + len) {
+				/* reverse RXOR */
+				order = -1;
+				state = 1;
+				if (i == src_cnt-1)
+					addr_count++;
+			} else {
+				state = 3;
+			}
+			break;
+		case 1:
+			if (i == src_cnt-2 || (order == -1
+				&& cur_addr != old_addr - len)) {
+				order = 0;
+				state = 0;
+				addr_count++;
+			} else if (cur_addr == old_addr + len*order) {
+				state = 2;
+				if (i == src_cnt-1)
+					addr_count++;
+			} else if (cur_addr == old_addr + 2*len) {
+				state = 2;
+				if (i == src_cnt-1)
+					addr_count++;
+			} else if (cur_addr == old_addr + 3*len) {
+				state = 2;
+				if (i == src_cnt-1)
+					addr_count++;
+			} else {
+				order = 0;
+				state = 0;
+				addr_count++;
+			}
+			break;
+		case 2:
+			order = 0;
+			state = 0;
+			addr_count++;
+				break;
+		}
+		if (state == 3)
+			break;
+	}
+	if (src_cnt <= 1 || (state != 1 && state != 2)) {
+		pr_err("%s: src_cnt=%d, state=%d, addr_count=%d, order=%lld\n",
+			__func__, src_cnt, state, addr_count, order);
+		for (i = 0; i < src_cnt; i++)
+			pr_err("\t[%d] 0x%llx \n", i, srcs[i]);
+		BUG();
+	}
+
+	return (addr_count + XOR_MAX_OPS - 1) / XOR_MAX_OPS;
+}
+
+
+/******************************************************************************
+ * ADMA channel low-level routines
+ ******************************************************************************/
+
+static u32
+ppc440spe_chan_get_current_descriptor(struct ppc440spe_adma_chan *chan);
+static void ppc440spe_chan_append(struct ppc440spe_adma_chan *chan);
+
+/**
+ * ppc440spe_adma_device_clear_eot_status - interrupt ack to XOR or DMA engine
+ */
+static void ppc440spe_adma_device_clear_eot_status(
+					struct ppc440spe_adma_chan *chan)
+{
+	struct dma_regs *dma_reg;
+	struct xor_regs *xor_reg;
+	u8 *p = chan->device->dma_desc_pool_virt;
+	struct dma_cdb *cdb;
+	u32 rv, i;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		/* read FIFO to ack */
+		dma_reg = chan->device->dma_reg;
+		while ((rv = ioread32(&dma_reg->csfpl))) {
+			i = rv & DMA_CDB_ADDR_MSK;
+			cdb = (struct dma_cdb *)&p[i -
+			    (u32)chan->device->dma_desc_pool];
+
+			/* Clear opcode to ack. This is necessary for
+			 * ZeroSum operations only
+			 */
+			cdb->opc = 0;
+
+			if (test_bit(PPC440SPE_RXOR_RUN,
+			    &ppc440spe_rxor_state)) {
+				/* probably this is a completed RXOR op,
+				 * get pointer to CDB using the fact that
+				 * physical and virtual addresses of CDB
+				 * in pools have the same offsets
+				 */
+				if (le32_to_cpu(cdb->sg1u) &
+				    DMA_CUED_XOR_BASE) {
+					/* this is a RXOR */
+					clear_bit(PPC440SPE_RXOR_RUN,
+						  &ppc440spe_rxor_state);
+				}
+			}
+
+			if (rv & DMA_CDB_STATUS_MSK) {
+				/* ZeroSum check failed
+				 */
+				struct ppc440spe_adma_desc_slot *iter;
+				dma_addr_t phys = rv & ~DMA_CDB_MSK;
+
+				/*
+				 * Update the status of corresponding
+				 * descriptor.
+				 */
+				list_for_each_entry(iter, &chan->chain,
+				    chain_node) {
+					if (iter->phys == phys)
+						break;
+				}
+				/*
+				 * if cannot find the corresponding
+				 * slot it's a bug
+				 */
+				BUG_ON(&iter->chain_node == &chan->chain);
+
+				if (iter->xor_check_result) {
+					if (test_bit(PPC440SPE_DESC_PCHECK,
+						     &iter->flags)) {
+						*iter->xor_check_result |=
+							SUM_CHECK_P_RESULT;
+					} else
+					if (test_bit(PPC440SPE_DESC_QCHECK,
+						     &iter->flags)) {
+						*iter->xor_check_result |=
+							SUM_CHECK_Q_RESULT;
+					} else
+						BUG();
+				}
+			}
+		}
+
+		rv = ioread32(&dma_reg->dsts);
+		if (rv) {
+			pr_err("DMA%d err status: 0x%x\n",
+			       chan->device->id, rv);
+			/* write back to clear */
+			iowrite32(rv, &dma_reg->dsts);
+		}
+		break;
+	case PPC440SPE_XOR_ID:
+		/* reset status bits to ack */
+		xor_reg = chan->device->xor_reg;
+		rv = ioread32be(&xor_reg->sr);
+		iowrite32be(rv, &xor_reg->sr);
+
+		if (rv & (XOR_IE_ICBIE_BIT|XOR_IE_ICIE_BIT|XOR_IE_RPTIE_BIT)) {
+			if (rv & XOR_IE_RPTIE_BIT) {
+				/* Read PLB Timeout Error.
+				 * Try to resubmit the CB
+				 */
+				u32 val = ioread32be(&xor_reg->ccbalr);
+
+				iowrite32be(val, &xor_reg->cblalr);
+
+				val = ioread32be(&xor_reg->crsr);
+				iowrite32be(val | XOR_CRSR_XAE_BIT,
+					    &xor_reg->crsr);
+			} else
+				pr_err("XOR ERR 0x%x status\n", rv);
+			break;
+		}
+
+		/*  if the XORcore is idle, but there are unprocessed CBs
+		 * then refetch the s/w chain here
+		 */
+		if (!(ioread32be(&xor_reg->sr) & XOR_SR_XCP_BIT) &&
+		    do_xor_refetch)
+			ppc440spe_chan_append(chan);
+		break;
+	}
+}
+
+/**
+ * ppc440spe_chan_is_busy - get the channel status
+ */
+static int ppc440spe_chan_is_busy(struct ppc440spe_adma_chan *chan)
+{
+	struct dma_regs *dma_reg;
+	struct xor_regs *xor_reg;
+	int busy = 0;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		dma_reg = chan->device->dma_reg;
+		/*  if command FIFO's head and tail pointers are equal and
+		 * status tail is the same as command, then channel is free
+		 */
+		if (ioread16(&dma_reg->cpfhp) != ioread16(&dma_reg->cpftp) ||
+		    ioread16(&dma_reg->cpftp) != ioread16(&dma_reg->csftp))
+			busy = 1;
+		break;
+	case PPC440SPE_XOR_ID:
+		/* use the special status bit for the XORcore
+		 */
+		xor_reg = chan->device->xor_reg;
+		busy = (ioread32be(&xor_reg->sr) & XOR_SR_XCP_BIT) ? 1 : 0;
+		break;
+	}
+
+	return busy;
+}
+
+/**
+ * ppc440spe_chan_set_first_xor_descriptor -  init XORcore chain
+ */
+static void ppc440spe_chan_set_first_xor_descriptor(
+				struct ppc440spe_adma_chan *chan,
+				struct ppc440spe_adma_desc_slot *next_desc)
+{
+	struct xor_regs *xor_reg = chan->device->xor_reg;
+
+	if (ioread32be(&xor_reg->sr) & XOR_SR_XCP_BIT)
+		printk(KERN_INFO "%s: Warn: XORcore is running "
+			"when try to set the first CDB!\n",
+			__func__);
+
+	xor_last_submit = xor_last_linked = next_desc;
+
+	iowrite32be(XOR_CRSR_64BA_BIT, &xor_reg->crsr);
+
+	iowrite32be(next_desc->phys, &xor_reg->cblalr);
+	iowrite32be(0, &xor_reg->cblahr);
+	iowrite32be(ioread32be(&xor_reg->cbcr) | XOR_CBCR_LNK_BIT,
+		    &xor_reg->cbcr);
+
+	chan->hw_chain_inited = 1;
+}
+
+/**
+ * ppc440spe_dma_put_desc - put DMA0,1 descriptor to FIFO.
+ * called with irqs disabled
+ */
+static void ppc440spe_dma_put_desc(struct ppc440spe_adma_chan *chan,
+		struct ppc440spe_adma_desc_slot *desc)
+{
+	u32 pcdb;
+	struct dma_regs *dma_reg = chan->device->dma_reg;
+
+	pcdb = desc->phys;
+	if (!test_bit(PPC440SPE_DESC_INT, &desc->flags))
+		pcdb |= DMA_CDB_NO_INT;
+
+	chan_last_sub[chan->device->id] = desc;
+
+	ADMA_LL_DBG(print_cb(chan, desc->hw_desc));
+
+	iowrite32(pcdb, &dma_reg->cpfpl);
+}
+
+/**
+ * ppc440spe_chan_append - update the h/w chain in the channel
+ */
+static void ppc440spe_chan_append(struct ppc440spe_adma_chan *chan)
+{
+	struct xor_regs *xor_reg;
+	struct ppc440spe_adma_desc_slot *iter;
+	struct xor_cb *xcb;
+	u32 cur_desc;
+	unsigned long flags;
+
+	local_irq_save(flags);
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		cur_desc = ppc440spe_chan_get_current_descriptor(chan);
+
+		if (likely(cur_desc)) {
+			iter = chan_last_sub[chan->device->id];
+			BUG_ON(!iter);
+		} else {
+			/* first peer */
+			iter = chan_first_cdb[chan->device->id];
+			BUG_ON(!iter);
+			ppc440spe_dma_put_desc(chan, iter);
+			chan->hw_chain_inited = 1;
+		}
+
+		/* is there something new to append */
+		if (!iter->hw_next)
+			break;
+
+		/* flush descriptors from the s/w queue to fifo */
+		list_for_each_entry_continue(iter, &chan->chain, chain_node) {
+			ppc440spe_dma_put_desc(chan, iter);
+			if (!iter->hw_next)
+				break;
+		}
+		break;
+	case PPC440SPE_XOR_ID:
+		/* update h/w links and refetch */
+		if (!xor_last_submit->hw_next)
+			break;
+
+		xor_reg = chan->device->xor_reg;
+		/* the last linked CDB has to generate an interrupt
+		 * that we'd be able to append the next lists to h/w
+		 * regardless of the XOR engine state at the moment of
+		 * appending of these next lists
+		 */
+		xcb = xor_last_linked->hw_desc;
+		xcb->cbc |= XOR_CBCR_CBCE_BIT;
+
+		if (!(ioread32be(&xor_reg->sr) & XOR_SR_XCP_BIT)) {
+			/* XORcore is idle. Refetch now */
+			do_xor_refetch = 0;
+			ppc440spe_xor_set_link(xor_last_submit,
+				xor_last_submit->hw_next);
+
+			ADMA_LL_DBG(print_cb_list(chan,
+				xor_last_submit->hw_next));
+
+			xor_last_submit = xor_last_linked;
+			iowrite32be(ioread32be(&xor_reg->crsr) |
+				    XOR_CRSR_RCBE_BIT | XOR_CRSR_64BA_BIT,
+				    &xor_reg->crsr);
+		} else {
+			/* XORcore is running. Refetch later in the handler */
+			do_xor_refetch = 1;
+		}
+
+		break;
+	}
+
+	local_irq_restore(flags);
+}
+
+/**
+ * ppc440spe_chan_get_current_descriptor - get the currently executed descriptor
+ */
+static u32
+ppc440spe_chan_get_current_descriptor(struct ppc440spe_adma_chan *chan)
+{
+	struct dma_regs *dma_reg;
+	struct xor_regs *xor_reg;
+
+	if (unlikely(!chan->hw_chain_inited))
+		/* h/w descriptor chain is not initialized yet */
+		return 0;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		dma_reg = chan->device->dma_reg;
+		return ioread32(&dma_reg->acpl) & (~DMA_CDB_MSK);
+	case PPC440SPE_XOR_ID:
+		xor_reg = chan->device->xor_reg;
+		return ioread32be(&xor_reg->ccbalr);
+	}
+	return 0;
+}
+
+/**
+ * ppc440spe_chan_run - enable the channel
+ */
+static void ppc440spe_chan_run(struct ppc440spe_adma_chan *chan)
+{
+	struct xor_regs *xor_reg;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		/* DMAs are always enabled, do nothing */
+		break;
+	case PPC440SPE_XOR_ID:
+		/* drain write buffer */
+		xor_reg = chan->device->xor_reg;
+
+		/* fetch descriptor pointed to in <link> */
+		iowrite32be(XOR_CRSR_64BA_BIT | XOR_CRSR_XAE_BIT,
+			    &xor_reg->crsr);
+		break;
+	}
+}
+
+/******************************************************************************
+ * ADMA device level
+ ******************************************************************************/
+
+static void ppc440spe_chan_start_null_xor(struct ppc440spe_adma_chan *chan);
+static int ppc440spe_adma_alloc_chan_resources(struct dma_chan *chan);
+
+static dma_cookie_t
+ppc440spe_adma_tx_submit(struct dma_async_tx_descriptor *tx);
+
+static void ppc440spe_adma_set_dest(struct ppc440spe_adma_desc_slot *tx,
+				    dma_addr_t addr, int index);
+static void
+ppc440spe_adma_memcpy_xor_set_src(struct ppc440spe_adma_desc_slot *tx,
+				  dma_addr_t addr, int index);
+
+static void
+ppc440spe_adma_pq_set_dest(struct ppc440spe_adma_desc_slot *tx,
+			   dma_addr_t *paddr, unsigned long flags);
+static void
+ppc440spe_adma_pq_set_src(struct ppc440spe_adma_desc_slot *tx,
+			  dma_addr_t addr, int index);
+static void
+ppc440spe_adma_pq_set_src_mult(struct ppc440spe_adma_desc_slot *tx,
+			       unsigned char mult, int index, int dst_pos);
+static void
+ppc440spe_adma_pqzero_sum_set_dest(struct ppc440spe_adma_desc_slot *tx,
+				   dma_addr_t paddr, dma_addr_t qaddr);
+
+static struct page *ppc440spe_rxor_srcs[32];
+
+/**
+ * ppc440spe_can_rxor - check if the operands may be processed with RXOR
+ */
+static int ppc440spe_can_rxor(struct page **srcs, int src_cnt, size_t len)
+{
+	int i, order = 0, state = 0;
+	int idx = 0;
+
+	if (unlikely(!(src_cnt > 1)))
+		return 0;
+
+	BUG_ON(src_cnt > ARRAY_SIZE(ppc440spe_rxor_srcs));
+
+	/* Skip holes in the source list before checking */
+	for (i = 0; i < src_cnt; i++) {
+		if (!srcs[i])
+			continue;
+		ppc440spe_rxor_srcs[idx++] = srcs[i];
+	}
+	src_cnt = idx;
+
+	for (i = 1; i < src_cnt; i++) {
+		char *cur_addr = page_address(ppc440spe_rxor_srcs[i]);
+		char *old_addr = page_address(ppc440spe_rxor_srcs[i - 1]);
+
+		switch (state) {
+		case 0:
+			if (cur_addr == old_addr + len) {
+				/* direct RXOR */
+				order = 1;
+				state = 1;
+			} else if (old_addr == cur_addr + len) {
+				/* reverse RXOR */
+				order = -1;
+				state = 1;
+			} else
+				goto out;
+			break;
+		case 1:
+			if ((i == src_cnt - 2) ||
+			    (order == -1 && cur_addr != old_addr - len)) {
+				order = 0;
+				state = 0;
+			} else if ((cur_addr == old_addr + len * order) ||
+				   (cur_addr == old_addr + 2 * len) ||
+				   (cur_addr == old_addr + 3 * len)) {
+				state = 2;
+			} else {
+				order = 0;
+				state = 0;
+			}
+			break;
+		case 2:
+			order = 0;
+			state = 0;
+			break;
+		}
+	}
+
+out:
+	if (state == 1 || state == 2)
+		return 1;
+
+	return 0;
+}
+
+/**
+ * ppc440spe_adma_device_estimate - estimate the efficiency of processing
+ *	the operation given on this channel. It's assumed that 'chan' is
+ *	capable to process 'cap' type of operation.
+ * @chan: channel to use
+ * @cap: type of transaction
+ * @dst_lst: array of destination pointers
+ * @dst_cnt: number of destination operands
+ * @src_lst: array of source pointers
+ * @src_cnt: number of source operands
+ * @src_sz: size of each source operand
+ */
+static int ppc440spe_adma_estimate(struct dma_chan *chan,
+	enum dma_transaction_type cap, struct page **dst_lst, int dst_cnt,
+	struct page **src_lst, int src_cnt, size_t src_sz)
+{
+	int ef = 1;
+
+	if (cap == DMA_PQ || cap == DMA_PQ_VAL) {
+		/* If RAID-6 capabilities were not activated don't try
+		 * to use them
+		 */
+		if (unlikely(!ppc440spe_r6_enabled))
+			return -1;
+	}
+	/*  In the current implementation of ppc440spe ADMA driver it
+	 * makes sense to pick out only pq case, because it may be
+	 * processed:
+	 * (1) either using Biskup method on DMA2;
+	 * (2) or on DMA0/1.
+	 *  Thus we give a favour to (1) if the sources are suitable;
+	 * else let it be processed on one of the DMA0/1 engines.
+	 *  In the sum_product case where destination is also the
+	 * source process it on DMA0/1 only.
+	 */
+	if (cap == DMA_PQ && chan->chan_id == PPC440SPE_XOR_ID) {
+
+		if (dst_cnt == 1 && src_cnt == 2 && dst_lst[0] == src_lst[1])
+			ef = 0; /* sum_product case, process on DMA0/1 */
+		else if (ppc440spe_can_rxor(src_lst, src_cnt, src_sz))
+			ef = 3; /* override (DMA0/1 + idle) */
+		else
+			ef = 0; /* can't process on DMA2 if !rxor */
+	}
+
+	/* channel idleness increases the priority */
+	if (likely(ef) &&
+	    !ppc440spe_chan_is_busy(to_ppc440spe_adma_chan(chan)))
+		ef++;
+
+	return ef;
+}
+
+struct dma_chan *
+ppc440spe_async_tx_find_best_channel(enum dma_transaction_type cap,
+	struct page **dst_lst, int dst_cnt, struct page **src_lst,
+	int src_cnt, size_t src_sz)
+{
+	struct dma_chan *best_chan = NULL;
+	struct ppc_dma_chan_ref *ref;
+	int best_rank = -1;
+
+	if (unlikely(!src_sz))
+		return NULL;
+	if (src_sz > PAGE_SIZE) {
+		/*
+		 * should a user of the api ever pass > PAGE_SIZE requests
+		 * we sort out cases where temporary page-sized buffers
+		 * are used.
+		 */
+		switch (cap) {
+		case DMA_PQ:
+			if (src_cnt == 1 && dst_lst[1] == src_lst[0])
+				return NULL;
+			if (src_cnt == 2 && dst_lst[1] == src_lst[1])
+				return NULL;
+			break;
+		case DMA_PQ_VAL:
+		case DMA_XOR_VAL:
+			return NULL;
+		default:
+			break;
+		}
+	}
+
+	list_for_each_entry(ref, &ppc440spe_adma_chan_list, node) {
+		if (dma_has_cap(cap, ref->chan->device->cap_mask)) {
+			int rank;
+
+			rank = ppc440spe_adma_estimate(ref->chan, cap, dst_lst,
+					dst_cnt, src_lst, src_cnt, src_sz);
+			if (rank > best_rank) {
+				best_rank = rank;
+				best_chan = ref->chan;
+			}
+		}
+	}
+
+	return best_chan;
+}
+EXPORT_SYMBOL_GPL(ppc440spe_async_tx_find_best_channel);
+
+/**
+ * ppc440spe_get_group_entry - get group entry with index idx
+ * @tdesc: is the last allocated slot in the group.
+ */
+static struct ppc440spe_adma_desc_slot *
+ppc440spe_get_group_entry(struct ppc440spe_adma_desc_slot *tdesc, u32 entry_idx)
+{
+	struct ppc440spe_adma_desc_slot *iter = tdesc->group_head;
+	int i = 0;
+
+	if (entry_idx < 0 || entry_idx >= (tdesc->src_cnt + tdesc->dst_cnt)) {
+		printk("%s: entry_idx %d, src_cnt %d, dst_cnt %d\n",
+			__func__, entry_idx, tdesc->src_cnt, tdesc->dst_cnt);
+		BUG();
+	}
+
+	list_for_each_entry(iter, &tdesc->group_list, chain_node) {
+		if (i++ == entry_idx)
+			break;
+	}
+	return iter;
+}
+
+/**
+ * ppc440spe_adma_free_slots - flags descriptor slots for reuse
+ * @slot: Slot to free
+ * Caller must hold &ppc440spe_chan->lock while calling this function
+ */
+static void ppc440spe_adma_free_slots(struct ppc440spe_adma_desc_slot *slot,
+				      struct ppc440spe_adma_chan *chan)
+{
+	int stride = slot->slots_per_op;
+
+	while (stride--) {
+		slot->slots_per_op = 0;
+		slot = list_entry(slot->slot_node.next,
+				struct ppc440spe_adma_desc_slot,
+				slot_node);
+	}
+}
+
+static void ppc440spe_adma_unmap(struct ppc440spe_adma_chan *chan,
+				 struct ppc440spe_adma_desc_slot *desc)
+{
+	u32 src_cnt, dst_cnt;
+	dma_addr_t addr;
+
+	/*
+	 * get the number of sources & destination
+	 * included in this descriptor and unmap
+	 * them all
+	 */
+	src_cnt = ppc440spe_desc_get_src_num(desc, chan);
+	dst_cnt = ppc440spe_desc_get_dst_num(desc, chan);
+
+	/* unmap destinations */
+	if (!(desc->async_tx.flags & DMA_COMPL_SKIP_DEST_UNMAP)) {
+		while (dst_cnt--) {
+			addr = ppc440spe_desc_get_dest_addr(
+				desc, chan, dst_cnt);
+			dma_unmap_page(chan->device->dev,
+					addr, desc->unmap_len,
+					DMA_FROM_DEVICE);
+		}
+	}
+
+	/* unmap sources */
+	if (!(desc->async_tx.flags & DMA_COMPL_SKIP_SRC_UNMAP)) {
+		while (src_cnt--) {
+			addr = ppc440spe_desc_get_src_addr(
+				desc, chan, src_cnt);
+			dma_unmap_page(chan->device->dev,
+					addr, desc->unmap_len,
+					DMA_TO_DEVICE);
+		}
+	}
+}
+
+/**
+ * ppc440spe_adma_run_tx_complete_actions - call functions to be called
+ * upon completion
+ */
+static dma_cookie_t ppc440spe_adma_run_tx_complete_actions(
+		struct ppc440spe_adma_desc_slot *desc,
+		struct ppc440spe_adma_chan *chan,
+		dma_cookie_t cookie)
+{
+	int i;
+
+	BUG_ON(desc->async_tx.cookie < 0);
+	if (desc->async_tx.cookie > 0) {
+		cookie = desc->async_tx.cookie;
+		desc->async_tx.cookie = 0;
+
+		/* call the callback (must not sleep or submit new
+		 * operations to this channel)
+		 */
+		if (desc->async_tx.callback)
+			desc->async_tx.callback(
+				desc->async_tx.callback_param);
+
+		/* unmap dma addresses
+		 * (unmap_single vs unmap_page?)
+		 *
+		 * actually, ppc's dma_unmap_page() functions are empty, so
+		 * the following code is just for the sake of completeness
+		 */
+		if (chan && chan->needs_unmap && desc->group_head &&
+		     desc->unmap_len) {
+			struct ppc440spe_adma_desc_slot *unmap =
+							desc->group_head;
+			/* assume 1 slot per op always */
+			u32 slot_count = unmap->slot_cnt;
+
+			/* Run through the group list and unmap addresses */
+			for (i = 0; i < slot_count; i++) {
+				BUG_ON(!unmap);
+				ppc440spe_adma_unmap(chan, unmap);
+				unmap = unmap->hw_next;
+			}
+		}
+	}
+
+	/* run dependent operations */
+	dma_run_dependencies(&desc->async_tx);
+
+	return cookie;
+}
+
+/**
+ * ppc440spe_adma_clean_slot - clean up CDB slot (if ack is set)
+ */
+static int ppc440spe_adma_clean_slot(struct ppc440spe_adma_desc_slot *desc,
+		struct ppc440spe_adma_chan *chan)
+{
+	/* the client is allowed to attach dependent operations
+	 * until 'ack' is set
+	 */
+	if (!async_tx_test_ack(&desc->async_tx))
+		return 0;
+
+	/* leave the last descriptor in the chain
+	 * so we can append to it
+	 */
+	if (list_is_last(&desc->chain_node, &chan->chain) ||
+	    desc->phys == ppc440spe_chan_get_current_descriptor(chan))
+		return 1;
+
+	if (chan->device->id != PPC440SPE_XOR_ID) {
+		/* our DMA interrupt handler clears opc field of
+		 * each processed descriptor. For all types of
+		 * operations except for ZeroSum we do not actually
+		 * need ack from the interrupt handler. ZeroSum is a
+		 * special case since the result of this operation
+		 * is available from the handler only, so if we see
+		 * such type of descriptor (which is unprocessed yet)
+		 * then leave it in chain.
+		 */
+		struct dma_cdb *cdb = desc->hw_desc;
+		if (cdb->opc == DMA_CDB_OPC_DCHECK128)
+			return 1;
+	}
+
+	dev_dbg(chan->device->common.dev, "\tfree slot %llx: %d stride: %d\n",
+		desc->phys, desc->idx, desc->slots_per_op);
+
+	list_del(&desc->chain_node);
+	ppc440spe_adma_free_slots(desc, chan);
+	return 0;
+}
+
+/**
+ * __ppc440spe_adma_slot_cleanup - this is the common clean-up routine
+ *	which runs through the channel CDBs list until reach the descriptor
+ *	currently processed. When routine determines that all CDBs of group
+ *	are completed then corresponding callbacks (if any) are called and slots
+ *	are freed.
+ */
+static void __ppc440spe_adma_slot_cleanup(struct ppc440spe_adma_chan *chan)
+{
+	struct ppc440spe_adma_desc_slot *iter, *_iter, *group_start = NULL;
+	dma_cookie_t cookie = 0;
+	u32 current_desc = ppc440spe_chan_get_current_descriptor(chan);
+	int busy = ppc440spe_chan_is_busy(chan);
+	int seen_current = 0, slot_cnt = 0, slots_per_op = 0;
+
+	dev_dbg(chan->device->common.dev, "ppc440spe adma%d: %s\n",
+		chan->device->id, __func__);
+
+	if (!current_desc) {
+		/*  There were no transactions yet, so
+		 * nothing to clean
+		 */
+		return;
+	}
+
+	/* free completed slots from the chain starting with
+	 * the oldest descriptor
+	 */
+	list_for_each_entry_safe(iter, _iter, &chan->chain,
+					chain_node) {
+		dev_dbg(chan->device->common.dev, "\tcookie: %d slot: %d "
+		    "busy: %d this_desc: %#llx next_desc: %#x "
+		    "cur: %#x ack: %d\n",
+		    iter->async_tx.cookie, iter->idx, busy, iter->phys,
+		    ppc440spe_desc_get_link(iter, chan), current_desc,
+		    async_tx_test_ack(&iter->async_tx));
+		prefetch(_iter);
+		prefetch(&_iter->async_tx);
+
+		/* do not advance past the current descriptor loaded into the
+		 * hardware channel,subsequent descriptors are either in process
+		 * or have not been submitted
+		 */
+		if (seen_current)
+			break;
+
+		/* stop the search if we reach the current descriptor and the
+		 * channel is busy, or if it appears that the current descriptor
+		 * needs to be re-read (i.e. has been appended to)
+		 */
+		if (iter->phys == current_desc) {
+			BUG_ON(seen_current++);
+			if (busy || ppc440spe_desc_get_link(iter, chan)) {
+				/* not all descriptors of the group have
+				 * been completed; exit.
+				 */
+				break;
+			}
+		}
+
+		/* detect the start of a group transaction */
+		if (!slot_cnt && !slots_per_op) {
+			slot_cnt = iter->slot_cnt;
+			slots_per_op = iter->slots_per_op;
+			if (slot_cnt <= slots_per_op) {
+				slot_cnt = 0;
+				slots_per_op = 0;
+			}
+		}
+
+		if (slot_cnt) {
+			if (!group_start)
+				group_start = iter;
+			slot_cnt -= slots_per_op;
+		}
+
+		/* all the members of a group are complete */
+		if (slots_per_op != 0 && slot_cnt == 0) {
+			struct ppc440spe_adma_desc_slot *grp_iter, *_grp_iter;
+			int end_of_chain = 0;
+
+			/* clean up the group */
+			slot_cnt = group_start->slot_cnt;
+			grp_iter = group_start;
+			list_for_each_entry_safe_from(grp_iter, _grp_iter,
+				&chan->chain, chain_node) {
+
+				cookie = ppc440spe_adma_run_tx_complete_actions(
+					grp_iter, chan, cookie);
+
+				slot_cnt -= slots_per_op;
+				end_of_chain = ppc440spe_adma_clean_slot(
+				    grp_iter, chan);
+				if (end_of_chain && slot_cnt) {
+					/* Should wait for ZeroSum completion */
+					if (cookie > 0)
+						chan->completed_cookie = cookie;
+					return;
+				}
+
+				if (slot_cnt == 0 || end_of_chain)
+					break;
+			}
+
+			/* the group should be complete at this point */
+			BUG_ON(slot_cnt);
+
+			slots_per_op = 0;
+			group_start = NULL;
+			if (end_of_chain)
+				break;
+			else
+				continue;
+		} else if (slots_per_op) /* wait for group completion */
+			continue;
+
+		cookie = ppc440spe_adma_run_tx_complete_actions(iter, chan,
+		    cookie);
+
+		if (ppc440spe_adma_clean_slot(iter, chan))
+			break;
+	}
+
+	BUG_ON(!seen_current);
+
+	if (cookie > 0) {
+		chan->completed_cookie = cookie;
+		pr_debug("\tcompleted cookie %d\n", cookie);
+	}
+
+}
+
+/**
+ * ppc440spe_adma_tasklet - clean up watch-dog initiator
+ */
+static void ppc440spe_adma_tasklet(unsigned long data)
+{
+	struct ppc440spe_adma_chan *chan = (struct ppc440spe_adma_chan *) data;
+
+	spin_lock_nested(&chan->lock, SINGLE_DEPTH_NESTING);
+	__ppc440spe_adma_slot_cleanup(chan);
+	spin_unlock(&chan->lock);
+}
+
+/**
+ * ppc440spe_adma_slot_cleanup - clean up scheduled initiator
+ */
+static void ppc440spe_adma_slot_cleanup(struct ppc440spe_adma_chan *chan)
+{
+	spin_lock_bh(&chan->lock);
+	__ppc440spe_adma_slot_cleanup(chan);
+	spin_unlock_bh(&chan->lock);
+}
+
+/**
+ * ppc440spe_adma_alloc_slots - allocate free slots (if any)
+ */
+static struct ppc440spe_adma_desc_slot *ppc440spe_adma_alloc_slots(
+		struct ppc440spe_adma_chan *chan, int num_slots,
+		int slots_per_op)
+{
+	struct ppc440spe_adma_desc_slot *iter = NULL, *_iter;
+	struct ppc440spe_adma_desc_slot *alloc_start = NULL;
+	struct list_head chain = LIST_HEAD_INIT(chain);
+	int slots_found, retry = 0;
+
+
+	BUG_ON(!num_slots || !slots_per_op);
+	/* start search from the last allocated descrtiptor
+	 * if a contiguous allocation can not be found start searching
+	 * from the beginning of the list
+	 */
+retry:
+	slots_found = 0;
+	if (retry == 0)
+		iter = chan->last_used;
+	else
+		iter = list_entry(&chan->all_slots,
+				  struct ppc440spe_adma_desc_slot,
+				  slot_node);
+	list_for_each_entry_safe_continue(iter, _iter, &chan->all_slots,
+	    slot_node) {
+		prefetch(_iter);
+		prefetch(&_iter->async_tx);
+		if (iter->slots_per_op) {
+			slots_found = 0;
+			continue;
+		}
+
+		/* start the allocation if the slot is correctly aligned */
+		if (!slots_found++)
+			alloc_start = iter;
+
+		if (slots_found == num_slots) {
+			struct ppc440spe_adma_desc_slot *alloc_tail = NULL;
+			struct ppc440spe_adma_desc_slot *last_used = NULL;
+
+			iter = alloc_start;
+			while (num_slots) {
+				int i;
+				/* pre-ack all but the last descriptor */
+				if (num_slots != slots_per_op)
+					async_tx_ack(&iter->async_tx);
+
+				list_add_tail(&iter->chain_node, &chain);
+				alloc_tail = iter;
+				iter->async_tx.cookie = 0;
+				iter->hw_next = NULL;
+				iter->flags = 0;
+				iter->slot_cnt = num_slots;
+				iter->xor_check_result = NULL;
+				for (i = 0; i < slots_per_op; i++) {
+					iter->slots_per_op = slots_per_op - i;
+					last_used = iter;
+					iter = list_entry(iter->slot_node.next,
+						struct ppc440spe_adma_desc_slot,
+						slot_node);
+				}
+				num_slots -= slots_per_op;
+			}
+			alloc_tail->group_head = alloc_start;
+			alloc_tail->async_tx.cookie = -EBUSY;
+			list_splice(&chain, &alloc_tail->group_list);
+			chan->last_used = last_used;
+			return alloc_tail;
+		}
+	}
+	if (!retry++)
+		goto retry;
+
+	/* try to free some slots if the allocation fails */
+	tasklet_schedule(&chan->irq_tasklet);
+	return NULL;
+}
+
+/**
+ * ppc440spe_adma_alloc_chan_resources -  allocate pools for CDB slots
+ */
+static int ppc440spe_adma_alloc_chan_resources(struct dma_chan *chan)
+{
+	struct ppc440spe_adma_chan *ppc440spe_chan;
+	struct ppc440spe_adma_desc_slot *slot = NULL;
+	char *hw_desc;
+	int i, db_sz;
+	int init;
+
+	ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+	init = ppc440spe_chan->slots_allocated ? 0 : 1;
+	chan->chan_id = ppc440spe_chan->device->id;
+
+	/* Allocate descriptor slots */
+	i = ppc440spe_chan->slots_allocated;
+	if (ppc440spe_chan->device->id != PPC440SPE_XOR_ID)
+		db_sz = sizeof(struct dma_cdb);
+	else
+		db_sz = sizeof(struct xor_cb);
+
+	for (; i < (ppc440spe_chan->device->pool_size / db_sz); i++) {
+		slot = kzalloc(sizeof(struct ppc440spe_adma_desc_slot),
+			       GFP_KERNEL);
+		if (!slot) {
+			printk(KERN_INFO "SPE ADMA Channel only initialized"
+				" %d descriptor slots", i--);
+			break;
+		}
+
+		hw_desc = (char *) ppc440spe_chan->device->dma_desc_pool_virt;
+		slot->hw_desc = (void *) &hw_desc[i * db_sz];
+		dma_async_tx_descriptor_init(&slot->async_tx, chan);
+		slot->async_tx.tx_submit = ppc440spe_adma_tx_submit;
+		INIT_LIST_HEAD(&slot->chain_node);
+		INIT_LIST_HEAD(&slot->slot_node);
+		INIT_LIST_HEAD(&slot->group_list);
+		slot->phys = ppc440spe_chan->device->dma_desc_pool + i * db_sz;
+		slot->idx = i;
+
+		spin_lock_bh(&ppc440spe_chan->lock);
+		ppc440spe_chan->slots_allocated++;
+		list_add_tail(&slot->slot_node, &ppc440spe_chan->all_slots);
+		spin_unlock_bh(&ppc440spe_chan->lock);
+	}
+
+	if (i && !ppc440spe_chan->last_used) {
+		ppc440spe_chan->last_used =
+			list_entry(ppc440spe_chan->all_slots.next,
+				struct ppc440spe_adma_desc_slot,
+				slot_node);
+	}
+
+	dev_dbg(ppc440spe_chan->device->common.dev,
+		"ppc440spe adma%d: allocated %d descriptor slots\n",
+		ppc440spe_chan->device->id, i);
+
+	/* initialize the channel and the chain with a null operation */
+	if (init) {
+		switch (ppc440spe_chan->device->id) {
+		case PPC440SPE_DMA0_ID:
+		case PPC440SPE_DMA1_ID:
+			ppc440spe_chan->hw_chain_inited = 0;
+			/* Use WXOR for self-testing */
+			if (!ppc440spe_r6_tchan)
+				ppc440spe_r6_tchan = ppc440spe_chan;
+			break;
+		case PPC440SPE_XOR_ID:
+			ppc440spe_chan_start_null_xor(ppc440spe_chan);
+			break;
+		default:
+			BUG();
+		}
+		ppc440spe_chan->needs_unmap = 1;
+	}
+
+	return (i > 0) ? i : -ENOMEM;
+}
+
+/**
+ * ppc440spe_desc_assign_cookie - assign a cookie
+ */
+static dma_cookie_t ppc440spe_desc_assign_cookie(
+		struct ppc440spe_adma_chan *chan,
+		struct ppc440spe_adma_desc_slot *desc)
+{
+	dma_cookie_t cookie = chan->common.cookie;
+
+	cookie++;
+	if (cookie < 0)
+		cookie = 1;
+	chan->common.cookie = desc->async_tx.cookie = cookie;
+	return cookie;
+}
+
+/**
+ * ppc440spe_rxor_set_region_data -
+ */
+static void ppc440spe_rxor_set_region(struct ppc440spe_adma_desc_slot *desc,
+	u8 xor_arg_no, u32 mask)
+{
+	struct xor_cb *xcb = desc->hw_desc;
+
+	xcb->ops[xor_arg_no].h |= mask;
+}
+
+/**
+ * ppc440spe_rxor_set_src -
+ */
+static void ppc440spe_rxor_set_src(struct ppc440spe_adma_desc_slot *desc,
+	u8 xor_arg_no, dma_addr_t addr)
+{
+	struct xor_cb *xcb = desc->hw_desc;
+
+	xcb->ops[xor_arg_no].h |= DMA_CUED_XOR_BASE;
+	xcb->ops[xor_arg_no].l = addr;
+}
+
+/**
+ * ppc440spe_rxor_set_mult -
+ */
+static void ppc440spe_rxor_set_mult(struct ppc440spe_adma_desc_slot *desc,
+	u8 xor_arg_no, u8 idx, u8 mult)
+{
+	struct xor_cb *xcb = desc->hw_desc;
+
+	xcb->ops[xor_arg_no].h |= mult << (DMA_CUED_MULT1_OFF + idx * 8);
+}
+
+/**
+ * ppc440spe_adma_check_threshold - append CDBs to h/w chain if threshold
+ *	has been achieved
+ */
+static void ppc440spe_adma_check_threshold(struct ppc440spe_adma_chan *chan)
+{
+	dev_dbg(chan->device->common.dev, "ppc440spe adma%d: pending: %d\n",
+		chan->device->id, chan->pending);
+
+	if (chan->pending >= PPC440SPE_ADMA_THRESHOLD) {
+		chan->pending = 0;
+		ppc440spe_chan_append(chan);
+	}
+}
+
+/**
+ * ppc440spe_adma_tx_submit - submit new descriptor group to the channel
+ *	(it's not necessary that descriptors will be submitted to the h/w
+ *	chains too right now)
+ */
+static dma_cookie_t ppc440spe_adma_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+	struct ppc440spe_adma_desc_slot *sw_desc;
+	struct ppc440spe_adma_chan *chan = to_ppc440spe_adma_chan(tx->chan);
+	struct ppc440spe_adma_desc_slot *group_start, *old_chain_tail;
+	int slot_cnt;
+	int slots_per_op;
+	dma_cookie_t cookie;
+
+	sw_desc = tx_to_ppc440spe_adma_slot(tx);
+
+	group_start = sw_desc->group_head;
+	slot_cnt = group_start->slot_cnt;
+	slots_per_op = group_start->slots_per_op;
+
+	spin_lock_bh(&chan->lock);
+
+	cookie = ppc440spe_desc_assign_cookie(chan, sw_desc);
+
+	if (unlikely(list_empty(&chan->chain))) {
+		/* first peer */
+		list_splice_init(&sw_desc->group_list, &chan->chain);
+		chan_first_cdb[chan->device->id] = group_start;
+	} else {
+		/* isn't first peer, bind CDBs to chain */
+		old_chain_tail = list_entry(chan->chain.prev,
+					struct ppc440spe_adma_desc_slot,
+					chain_node);
+		list_splice_init(&sw_desc->group_list,
+		    &old_chain_tail->chain_node);
+		/* fix up the hardware chain */
+		ppc440spe_desc_set_link(chan, old_chain_tail, group_start);
+	}
+
+	/* increment the pending count by the number of operations */
+	chan->pending += slot_cnt / slots_per_op;
+	ppc440spe_adma_check_threshold(chan);
+	spin_unlock_bh(&chan->lock);
+
+	dev_dbg(chan->device->common.dev,
+		"ppc440spe adma%d: %s cookie: %d slot: %d tx %p\n",
+		chan->device->id, __func__,
+		sw_desc->async_tx.cookie, sw_desc->idx, sw_desc);
+
+	return cookie;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_interrupt - prepare CDB for a pseudo DMA operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_interrupt(
+		struct dma_chan *chan, unsigned long flags)
+{
+	struct ppc440spe_adma_chan *ppc440spe_chan;
+	struct ppc440spe_adma_desc_slot *sw_desc, *group_start;
+	int slot_cnt, slots_per_op;
+
+	ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+
+	dev_dbg(ppc440spe_chan->device->common.dev,
+		"ppc440spe adma%d: %s\n", ppc440spe_chan->device->id,
+		__func__);
+
+	spin_lock_bh(&ppc440spe_chan->lock);
+	slot_cnt = slots_per_op = 1;
+	sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt,
+			slots_per_op);
+	if (sw_desc) {
+		group_start = sw_desc->group_head;
+		ppc440spe_desc_init_interrupt(group_start, ppc440spe_chan);
+		group_start->unmap_len = 0;
+		sw_desc->async_tx.flags = flags;
+	}
+	spin_unlock_bh(&ppc440spe_chan->lock);
+
+	return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_memcpy - prepare CDB for a MEMCPY operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_memcpy(
+		struct dma_chan *chan, dma_addr_t dma_dest,
+		dma_addr_t dma_src, size_t len, unsigned long flags)
+{
+	struct ppc440spe_adma_chan *ppc440spe_chan;
+	struct ppc440spe_adma_desc_slot *sw_desc, *group_start;
+	int slot_cnt, slots_per_op;
+
+	ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+
+	if (unlikely(!len))
+		return NULL;
+
+	BUG_ON(unlikely(len > PPC440SPE_ADMA_DMA_MAX_BYTE_COUNT));
+
+	spin_lock_bh(&ppc440spe_chan->lock);
+
+	dev_dbg(ppc440spe_chan->device->common.dev,
+		"ppc440spe adma%d: %s len: %u int_en %d\n",
+		ppc440spe_chan->device->id, __func__, len,
+		flags & DMA_PREP_INTERRUPT ? 1 : 0);
+	slot_cnt = slots_per_op = 1;
+	sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt,
+		slots_per_op);
+	if (sw_desc) {
+		group_start = sw_desc->group_head;
+		ppc440spe_desc_init_memcpy(group_start, flags);
+		ppc440spe_adma_set_dest(group_start, dma_dest, 0);
+		ppc440spe_adma_memcpy_xor_set_src(group_start, dma_src, 0);
+		ppc440spe_desc_set_byte_count(group_start, ppc440spe_chan, len);
+		sw_desc->unmap_len = len;
+		sw_desc->async_tx.flags = flags;
+	}
+	spin_unlock_bh(&ppc440spe_chan->lock);
+
+	return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_memset - prepare CDB for a MEMSET operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_memset(
+		struct dma_chan *chan, dma_addr_t dma_dest, int value,
+		size_t len, unsigned long flags)
+{
+	struct ppc440spe_adma_chan *ppc440spe_chan;
+	struct ppc440spe_adma_desc_slot *sw_desc, *group_start;
+	int slot_cnt, slots_per_op;
+
+	ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+
+	if (unlikely(!len))
+		return NULL;
+
+	BUG_ON(unlikely(len > PPC440SPE_ADMA_DMA_MAX_BYTE_COUNT));
+
+	spin_lock_bh(&ppc440spe_chan->lock);
+
+	dev_dbg(ppc440spe_chan->device->common.dev,
+		"ppc440spe adma%d: %s cal: %u len: %u int_en %d\n",
+		ppc440spe_chan->device->id, __func__, value, len,
+		flags & DMA_PREP_INTERRUPT ? 1 : 0);
+
+	slot_cnt = slots_per_op = 1;
+	sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt,
+		slots_per_op);
+	if (sw_desc) {
+		group_start = sw_desc->group_head;
+		ppc440spe_desc_init_memset(group_start, value, flags);
+		ppc440spe_adma_set_dest(group_start, dma_dest, 0);
+		ppc440spe_desc_set_byte_count(group_start, ppc440spe_chan, len);
+		sw_desc->unmap_len = len;
+		sw_desc->async_tx.flags = flags;
+	}
+	spin_unlock_bh(&ppc440spe_chan->lock);
+
+	return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_xor - prepare CDB for a XOR operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_xor(
+		struct dma_chan *chan, dma_addr_t dma_dest,
+		dma_addr_t *dma_src, u32 src_cnt, size_t len,
+		unsigned long flags)
+{
+	struct ppc440spe_adma_chan *ppc440spe_chan;
+	struct ppc440spe_adma_desc_slot *sw_desc, *group_start;
+	int slot_cnt, slots_per_op;
+
+	ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+
+	ADMA_LL_DBG(prep_dma_xor_dbg(ppc440spe_chan->device->id,
+				     dma_dest, dma_src, src_cnt));
+	if (unlikely(!len))
+		return NULL;
+	BUG_ON(unlikely(len > PPC440SPE_ADMA_XOR_MAX_BYTE_COUNT));
+
+	dev_dbg(ppc440spe_chan->device->common.dev,
+		"ppc440spe adma%d: %s src_cnt: %d len: %u int_en: %d\n",
+		ppc440spe_chan->device->id, __func__, src_cnt, len,
+		flags & DMA_PREP_INTERRUPT ? 1 : 0);
+
+	spin_lock_bh(&ppc440spe_chan->lock);
+	slot_cnt = ppc440spe_chan_xor_slot_count(len, src_cnt, &slots_per_op);
+	sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt,
+			slots_per_op);
+	if (sw_desc) {
+		group_start = sw_desc->group_head;
+		ppc440spe_desc_init_xor(group_start, src_cnt, flags);
+		ppc440spe_adma_set_dest(group_start, dma_dest, 0);
+		while (src_cnt--)
+			ppc440spe_adma_memcpy_xor_set_src(group_start,
+				dma_src[src_cnt], src_cnt);
+		ppc440spe_desc_set_byte_count(group_start, ppc440spe_chan, len);
+		sw_desc->unmap_len = len;
+		sw_desc->async_tx.flags = flags;
+	}
+	spin_unlock_bh(&ppc440spe_chan->lock);
+
+	return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+static inline void
+ppc440spe_desc_set_xor_src_cnt(struct ppc440spe_adma_desc_slot *desc,
+				int src_cnt);
+static void ppc440spe_init_rxor_cursor(struct ppc440spe_rxor *cursor);
+
+/**
+ * ppc440spe_adma_init_dma2rxor_slot -
+ */
+static void ppc440spe_adma_init_dma2rxor_slot(
+		struct ppc440spe_adma_desc_slot *desc,
+		dma_addr_t *src, int src_cnt)
+{
+	int i;
+
+	/* initialize CDB */
+	for (i = 0; i < src_cnt; i++) {
+		ppc440spe_adma_dma2rxor_prep_src(desc, &desc->rxor_cursor, i,
+						 desc->src_cnt, (u32)src[i]);
+	}
+}
+
+/**
+ * ppc440spe_dma01_prep_mult -
+ * for Q operation where destination is also the source
+ */
+static struct ppc440spe_adma_desc_slot *ppc440spe_dma01_prep_mult(
+		struct ppc440spe_adma_chan *ppc440spe_chan,
+		dma_addr_t *dst, int dst_cnt, dma_addr_t *src, int src_cnt,
+		const unsigned char *scf, size_t len, unsigned long flags)
+{
+	struct ppc440spe_adma_desc_slot *sw_desc = NULL;
+	unsigned long op = 0;
+	int slot_cnt;
+
+	set_bit(PPC440SPE_DESC_WXOR, &op);
+	slot_cnt = 2;
+
+	spin_lock_bh(&ppc440spe_chan->lock);
+
+	/* use WXOR, each descriptor occupies one slot */
+	sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt, 1);
+	if (sw_desc) {
+		struct ppc440spe_adma_chan *chan;
+		struct ppc440spe_adma_desc_slot *iter;
+		struct dma_cdb *hw_desc;
+
+		chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+		set_bits(op, &sw_desc->flags);
+		sw_desc->src_cnt = src_cnt;
+		sw_desc->dst_cnt = dst_cnt;
+		/* First descriptor, zero data in the destination and copy it
+		 * to q page using MULTICAST transfer.
+		 */
+		iter = list_first_entry(&sw_desc->group_list,
+					struct ppc440spe_adma_desc_slot,
+					chain_node);
+		memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+		/* set 'next' pointer */
+		iter->hw_next = list_entry(iter->chain_node.next,
+					   struct ppc440spe_adma_desc_slot,
+					   chain_node);
+		clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+		hw_desc = iter->hw_desc;
+		hw_desc->opc = DMA_CDB_OPC_MULTICAST;
+
+		ppc440spe_desc_set_dest_addr(iter, chan,
+					     DMA_CUED_XOR_BASE, dst[0], 0);
+		ppc440spe_desc_set_dest_addr(iter, chan, 0, dst[1], 1);
+		ppc440spe_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB,
+					    src[0]);
+		ppc440spe_desc_set_byte_count(iter, ppc440spe_chan, len);
+		iter->unmap_len = len;
+
+		/*
+		 * Second descriptor, multiply data from the q page
+		 * and store the result in real destination.
+		 */
+		iter = list_first_entry(&iter->chain_node,
+					struct ppc440spe_adma_desc_slot,
+					chain_node);
+		memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+		iter->hw_next = NULL;
+		if (flags & DMA_PREP_INTERRUPT)
+			set_bit(PPC440SPE_DESC_INT, &iter->flags);
+		else
+			clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+
+		hw_desc = iter->hw_desc;
+		hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+		ppc440spe_desc_set_src_addr(iter, chan, 0,
+					    DMA_CUED_XOR_HB, dst[1]);
+		ppc440spe_desc_set_dest_addr(iter, chan,
+					     DMA_CUED_XOR_BASE, dst[0], 0);
+
+		ppc440spe_desc_set_src_mult(iter, chan, DMA_CUED_MULT1_OFF,
+					    DMA_CDB_SG_DST1, scf[0]);
+		ppc440spe_desc_set_byte_count(iter, ppc440spe_chan, len);
+		iter->unmap_len = len;
+		sw_desc->async_tx.flags = flags;
+	}
+
+	spin_unlock_bh(&ppc440spe_chan->lock);
+
+	return sw_desc;
+}
+
+/**
+ * ppc440spe_dma01_prep_sum_product -
+ * Dx = A*(P+Pxy) + B*(Q+Qxy) operation where destination is also
+ * the source.
+ */
+static struct ppc440spe_adma_desc_slot *ppc440spe_dma01_prep_sum_product(
+		struct ppc440spe_adma_chan *ppc440spe_chan,
+		dma_addr_t *dst, dma_addr_t *src, int src_cnt,
+		const unsigned char *scf, size_t len, unsigned long flags)
+{
+	struct ppc440spe_adma_desc_slot *sw_desc = NULL;
+	unsigned long op = 0;
+	int slot_cnt;
+
+	set_bit(PPC440SPE_DESC_WXOR, &op);
+	slot_cnt = 3;
+
+	spin_lock_bh(&ppc440spe_chan->lock);
+
+	/* WXOR, each descriptor occupies one slot */
+	sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt, 1);
+	if (sw_desc) {
+		struct ppc440spe_adma_chan *chan;
+		struct ppc440spe_adma_desc_slot *iter;
+		struct dma_cdb *hw_desc;
+
+		chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+		set_bits(op, &sw_desc->flags);
+		sw_desc->src_cnt = src_cnt;
+		sw_desc->dst_cnt = 1;
+		/* 1st descriptor, src[1] data to q page and zero destination */
+		iter = list_first_entry(&sw_desc->group_list,
+					struct ppc440spe_adma_desc_slot,
+					chain_node);
+		memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+		iter->hw_next = list_entry(iter->chain_node.next,
+					   struct ppc440spe_adma_desc_slot,
+					   chain_node);
+		clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+		hw_desc = iter->hw_desc;
+		hw_desc->opc = DMA_CDB_OPC_MULTICAST;
+
+		ppc440spe_desc_set_dest_addr(iter, chan, DMA_CUED_XOR_BASE,
+					     *dst, 0);
+		ppc440spe_desc_set_dest_addr(iter, chan, 0,
+					     ppc440spe_chan->qdest, 1);
+		ppc440spe_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB,
+					    src[1]);
+		ppc440spe_desc_set_byte_count(iter, ppc440spe_chan, len);
+		iter->unmap_len = len;
+
+		/* 2nd descriptor, multiply src[1] data and store the
+		 * result in destination */
+		iter = list_first_entry(&iter->chain_node,
+					struct ppc440spe_adma_desc_slot,
+					chain_node);
+		memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+		/* set 'next' pointer */
+		iter->hw_next = list_entry(iter->chain_node.next,
+					   struct ppc440spe_adma_desc_slot,
+					   chain_node);
+		if (flags & DMA_PREP_INTERRUPT)
+			set_bit(PPC440SPE_DESC_INT, &iter->flags);
+		else
+			clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+
+		hw_desc = iter->hw_desc;
+		hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+		ppc440spe_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB,
+					    ppc440spe_chan->qdest);
+		ppc440spe_desc_set_dest_addr(iter, chan, DMA_CUED_XOR_BASE,
+					     *dst, 0);
+		ppc440spe_desc_set_src_mult(iter, chan,	DMA_CUED_MULT1_OFF,
+					    DMA_CDB_SG_DST1, scf[1]);
+		ppc440spe_desc_set_byte_count(iter, ppc440spe_chan, len);
+		iter->unmap_len = len;
+
+		/*
+		 * 3rd descriptor, multiply src[0] data and xor it
+		 * with destination
+		 */
+		iter = list_first_entry(&iter->chain_node,
+					struct ppc440spe_adma_desc_slot,
+					chain_node);
+		memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+		iter->hw_next = NULL;
+		if (flags & DMA_PREP_INTERRUPT)
+			set_bit(PPC440SPE_DESC_INT, &iter->flags);
+		else
+			clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+
+		hw_desc = iter->hw_desc;
+		hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+		ppc440spe_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB,
+					    src[0]);
+		ppc440spe_desc_set_dest_addr(iter, chan, DMA_CUED_XOR_BASE,
+					     *dst, 0);
+		ppc440spe_desc_set_src_mult(iter, chan, DMA_CUED_MULT1_OFF,
+					    DMA_CDB_SG_DST1, scf[0]);
+		ppc440spe_desc_set_byte_count(iter, ppc440spe_chan, len);
+		iter->unmap_len = len;
+		sw_desc->async_tx.flags = flags;
+	}
+
+	spin_unlock_bh(&ppc440spe_chan->lock);
+
+	return sw_desc;
+}
+
+static struct ppc440spe_adma_desc_slot *ppc440spe_dma01_prep_pq(
+		struct ppc440spe_adma_chan *ppc440spe_chan,
+		dma_addr_t *dst, int dst_cnt, dma_addr_t *src, int src_cnt,
+		const unsigned char *scf, size_t len, unsigned long flags)
+{
+	int slot_cnt;
+	struct ppc440spe_adma_desc_slot *sw_desc = NULL, *iter;
+	unsigned long op = 0;
+	unsigned char mult = 1;
+
+	pr_debug("%s: dst_cnt %d, src_cnt %d, len %d\n",
+		 __func__, dst_cnt, src_cnt, len);
+	/*  select operations WXOR/RXOR depending on the
+	 * source addresses of operators and the number
+	 * of destinations (RXOR support only Q-parity calculations)
+	 */
+	set_bit(PPC440SPE_DESC_WXOR, &op);
+	if (!test_and_set_bit(PPC440SPE_RXOR_RUN, &ppc440spe_rxor_state)) {
+		/* no active RXOR;
+		 * do RXOR if:
+		 * - there are more than 1 source,
+		 * - len is aligned on 512-byte boundary,
+		 * - source addresses fit to one of 4 possible regions.
+		 */
+		if (src_cnt > 1 &&
+		    !(len & MQ0_CF2H_RXOR_BS_MASK) &&
+		    (src[0] + len) == src[1]) {
+			/* may do RXOR R1 R2 */
+			set_bit(PPC440SPE_DESC_RXOR, &op);
+			if (src_cnt != 2) {
+				/* may try to enhance region of RXOR */
+				if ((src[1] + len) == src[2]) {
+					/* do RXOR R1 R2 R3 */
+					set_bit(PPC440SPE_DESC_RXOR123,
+						&op);
+				} else if ((src[1] + len * 2) == src[2]) {
+					/* do RXOR R1 R2 R4 */
+					set_bit(PPC440SPE_DESC_RXOR124, &op);
+				} else if ((src[1] + len * 3) == src[2]) {
+					/* do RXOR R1 R2 R5 */
+					set_bit(PPC440SPE_DESC_RXOR125,
+						&op);
+				} else {
+					/* do RXOR R1 R2 */
+					set_bit(PPC440SPE_DESC_RXOR12,
+						&op);
+				}
+			} else {
+				/* do RXOR R1 R2 */
+				set_bit(PPC440SPE_DESC_RXOR12, &op);
+			}
+		}
+
+		if (!test_bit(PPC440SPE_DESC_RXOR, &op)) {
+			/* can not do this operation with RXOR */
+			clear_bit(PPC440SPE_RXOR_RUN,
+				&ppc440spe_rxor_state);
+		} else {
+			/* can do; set block size right now */
+			ppc440spe_desc_set_rxor_block_size(len);
+		}
+	}
+
+	/* Number of necessary slots depends on operation type selected */
+	if (!test_bit(PPC440SPE_DESC_RXOR, &op)) {
+		/*  This is a WXOR only chain. Need descriptors for each
+		 * source to GF-XOR them with WXOR, and need descriptors
+		 * for each destination to zero them with WXOR
+		 */
+		slot_cnt = src_cnt;
+
+		if (flags & DMA_PREP_ZERO_P) {
+			slot_cnt++;
+			set_bit(PPC440SPE_ZERO_P, &op);
+		}
+		if (flags & DMA_PREP_ZERO_Q) {
+			slot_cnt++;
+			set_bit(PPC440SPE_ZERO_Q, &op);
+		}
+	} else {
+		/*  Need 1/2 descriptor for RXOR operation, and
+		 * need (src_cnt - (2 or 3)) for WXOR of sources
+		 * remained (if any)
+		 */
+		slot_cnt = dst_cnt;
+
+		if (flags & DMA_PREP_ZERO_P)
+			set_bit(PPC440SPE_ZERO_P, &op);
+		if (flags & DMA_PREP_ZERO_Q)
+			set_bit(PPC440SPE_ZERO_Q, &op);
+
+		if (test_bit(PPC440SPE_DESC_RXOR12, &op))
+			slot_cnt += src_cnt - 2;
+		else
+			slot_cnt += src_cnt - 3;
+
+		/*  Thus we have either RXOR only chain or
+		 * mixed RXOR/WXOR
+		 */
+		if (slot_cnt == dst_cnt)
+			/* RXOR only chain */
+			clear_bit(PPC440SPE_DESC_WXOR, &op);
+	}
+
+	spin_lock_bh(&ppc440spe_chan->lock);
+	/* for both RXOR/WXOR each descriptor occupies one slot */
+	sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt, 1);
+	if (sw_desc) {
+		ppc440spe_desc_init_dma01pq(sw_desc, dst_cnt, src_cnt,
+				flags, op);
+
+		/* setup dst/src/mult */
+		pr_debug("%s: set dst descriptor 0, 1: 0x%016llx, 0x%016llx\n",
+			 __func__, dst[0], dst[1]);
+		ppc440spe_adma_pq_set_dest(sw_desc, dst, flags);
+		while (src_cnt--) {
+			ppc440spe_adma_pq_set_src(sw_desc, src[src_cnt],
+						  src_cnt);
+
+			/* NOTE: "Multi = 0 is equivalent to = 1" as it
+			 * stated in 440SPSPe_RAID6_Addendum_UM_1_17.pdf
+			 * doesn't work for RXOR with DMA0/1! Instead, multi=0
+			 * leads to zeroing source data after RXOR.
+			 * So, for P case set-up mult=1 explicitly.
+			 */
+			if (!(flags & DMA_PREP_PQ_DISABLE_Q))
+				mult = scf[src_cnt];
+			ppc440spe_adma_pq_set_src_mult(sw_desc,
+				mult, src_cnt,  dst_cnt - 1);
+		}
+
+		/* Setup byte count foreach slot just allocated */
+		sw_desc->async_tx.flags = flags;
+		list_for_each_entry(iter, &sw_desc->group_list,
+				chain_node) {
+			ppc440spe_desc_set_byte_count(iter,
+				ppc440spe_chan, len);
+			iter->unmap_len = len;
+		}
+	}
+	spin_unlock_bh(&ppc440spe_chan->lock);
+
+	return sw_desc;
+}
+
+static struct ppc440spe_adma_desc_slot *ppc440spe_dma2_prep_pq(
+		struct ppc440spe_adma_chan *ppc440spe_chan,
+		dma_addr_t *dst, int dst_cnt, dma_addr_t *src, int src_cnt,
+		const unsigned char *scf, size_t len, unsigned long flags)
+{
+	int slot_cnt, descs_per_op;
+	struct ppc440spe_adma_desc_slot *sw_desc = NULL, *iter;
+	unsigned long op = 0;
+	unsigned char mult = 1;
+
+	BUG_ON(!dst_cnt);
+	/*pr_debug("%s: dst_cnt %d, src_cnt %d, len %d\n",
+		 __func__, dst_cnt, src_cnt, len);*/
+
+	spin_lock_bh(&ppc440spe_chan->lock);
+	descs_per_op = ppc440spe_dma2_pq_slot_count(src, src_cnt, len);
+	if (descs_per_op < 0) {
+		spin_unlock_bh(&ppc440spe_chan->lock);
+		return NULL;
+	}
+
+	/* depending on number of sources we have 1 or 2 RXOR chains */
+	slot_cnt = descs_per_op * dst_cnt;
+
+	sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt, 1);
+	if (sw_desc) {
+		op = slot_cnt;
+		sw_desc->async_tx.flags = flags;
+		list_for_each_entry(iter, &sw_desc->group_list, chain_node) {
+			ppc440spe_desc_init_dma2pq(iter, dst_cnt, src_cnt,
+				--op ? 0 : flags);
+			ppc440spe_desc_set_byte_count(iter, ppc440spe_chan,
+				len);
+			iter->unmap_len = len;
+
+			ppc440spe_init_rxor_cursor(&(iter->rxor_cursor));
+			iter->rxor_cursor.len = len;
+			iter->descs_per_op = descs_per_op;
+		}
+		op = 0;
+		list_for_each_entry(iter, &sw_desc->group_list, chain_node) {
+			op++;
+			if (op % descs_per_op == 0)
+				ppc440spe_adma_init_dma2rxor_slot(iter, src,
+								  src_cnt);
+			if (likely(!list_is_last(&iter->chain_node,
+						 &sw_desc->group_list))) {
+				/* set 'next' pointer */
+				iter->hw_next =
+					list_entry(iter->chain_node.next,
+						struct ppc440spe_adma_desc_slot,
+						chain_node);
+				ppc440spe_xor_set_link(iter, iter->hw_next);
+			} else {
+				/* this is the last descriptor. */
+				iter->hw_next = NULL;
+			}
+		}
+
+		/* fixup head descriptor */
+		sw_desc->dst_cnt = dst_cnt;
+		if (flags & DMA_PREP_ZERO_P)
+			set_bit(PPC440SPE_ZERO_P, &sw_desc->flags);
+		if (flags & DMA_PREP_ZERO_Q)
+			set_bit(PPC440SPE_ZERO_Q, &sw_desc->flags);
+
+		/* setup dst/src/mult */
+		ppc440spe_adma_pq_set_dest(sw_desc, dst, flags);
+
+		while (src_cnt--) {
+			/* handle descriptors (if dst_cnt == 2) inside
+			 * the ppc440spe_adma_pq_set_srcxxx() functions
+			 */
+			ppc440spe_adma_pq_set_src(sw_desc, src[src_cnt],
+						  src_cnt);
+			if (!(flags & DMA_PREP_PQ_DISABLE_Q))
+				mult = scf[src_cnt];
+			ppc440spe_adma_pq_set_src_mult(sw_desc,
+					mult, src_cnt, dst_cnt - 1);
+		}
+	}
+	spin_unlock_bh(&ppc440spe_chan->lock);
+	ppc440spe_desc_set_rxor_block_size(len);
+	return sw_desc;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_pq - prepare CDB (group) for a GF-XOR operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_pq(
+		struct dma_chan *chan, dma_addr_t *dst, dma_addr_t *src,
+		unsigned int src_cnt, const unsigned char *scf,
+		size_t len, unsigned long flags)
+{
+	struct ppc440spe_adma_chan *ppc440spe_chan;
+	struct ppc440spe_adma_desc_slot *sw_desc = NULL;
+	int dst_cnt = 0;
+
+	ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+
+	ADMA_LL_DBG(prep_dma_pq_dbg(ppc440spe_chan->device->id,
+				    dst, src, src_cnt));
+	BUG_ON(!len);
+	BUG_ON(unlikely(len > PPC440SPE_ADMA_XOR_MAX_BYTE_COUNT));
+	BUG_ON(!src_cnt);
+
+	if (src_cnt == 1 && dst[1] == src[0]) {
+		dma_addr_t dest[2];
+
+		/* dst[1] is real destination (Q) */
+		dest[0] = dst[1];
+		/* this is the page to multicast source data to */
+		dest[1] = ppc440spe_chan->qdest;
+		sw_desc = ppc440spe_dma01_prep_mult(ppc440spe_chan,
+				dest, 2, src, src_cnt, scf, len, flags);
+		return sw_desc ? &sw_desc->async_tx : NULL;
+	}
+
+	if (src_cnt == 2 && dst[1] == src[1]) {
+		sw_desc = ppc440spe_dma01_prep_sum_product(ppc440spe_chan,
+					&dst[1], src, 2, scf, len, flags);
+		return sw_desc ? &sw_desc->async_tx : NULL;
+	}
+
+	if (!(flags & DMA_PREP_PQ_DISABLE_P)) {
+		BUG_ON(!dst[0]);
+		dst_cnt++;
+		flags |= DMA_PREP_ZERO_P;
+	}
+
+	if (!(flags & DMA_PREP_PQ_DISABLE_Q)) {
+		BUG_ON(!dst[1]);
+		dst_cnt++;
+		flags |= DMA_PREP_ZERO_Q;
+	}
+
+	BUG_ON(!dst_cnt);
+
+	dev_dbg(ppc440spe_chan->device->common.dev,
+		"ppc440spe adma%d: %s src_cnt: %d len: %u int_en: %d\n",
+		ppc440spe_chan->device->id, __func__, src_cnt, len,
+		flags & DMA_PREP_INTERRUPT ? 1 : 0);
+
+	switch (ppc440spe_chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		sw_desc = ppc440spe_dma01_prep_pq(ppc440spe_chan,
+				dst, dst_cnt, src, src_cnt, scf,
+				len, flags);
+		break;
+
+	case PPC440SPE_XOR_ID:
+		sw_desc = ppc440spe_dma2_prep_pq(ppc440spe_chan,
+				dst, dst_cnt, src, src_cnt, scf,
+				len, flags);
+		break;
+	}
+
+	return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_pqzero_sum - prepare CDB group for
+ * a PQ_ZERO_SUM operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_pqzero_sum(
+		struct dma_chan *chan, dma_addr_t *pq, dma_addr_t *src,
+		unsigned int src_cnt, const unsigned char *scf, size_t len,
+		enum sum_check_flags *pqres, unsigned long flags)
+{
+	struct ppc440spe_adma_chan *ppc440spe_chan;
+	struct ppc440spe_adma_desc_slot *sw_desc, *iter;
+	dma_addr_t pdest, qdest;
+	int slot_cnt, slots_per_op, idst, dst_cnt;
+
+	ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+
+	if (flags & DMA_PREP_PQ_DISABLE_P)
+		pdest = 0;
+	else
+		pdest = pq[0];
+
+	if (flags & DMA_PREP_PQ_DISABLE_Q)
+		qdest = 0;
+	else
+		qdest = pq[1];
+
+	ADMA_LL_DBG(prep_dma_pqzero_sum_dbg(ppc440spe_chan->device->id,
+					    src, src_cnt, scf));
+
+	/* Always use WXOR for P/Q calculations (two destinations).
+	 * Need 1 or 2 extra slots to verify results are zero.
+	 */
+	idst = dst_cnt = (pdest && qdest) ? 2 : 1;
+
+	/* One additional slot per destination to clone P/Q
+	 * before calculation (we have to preserve destinations).
+	 */
+	slot_cnt = src_cnt + dst_cnt * 2;
+	slots_per_op = 1;
+
+	spin_lock_bh(&ppc440spe_chan->lock);
+	sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt,
+					     slots_per_op);
+	if (sw_desc) {
+		ppc440spe_desc_init_dma01pqzero_sum(sw_desc, dst_cnt, src_cnt);
+
+		/* Setup byte count for each slot just allocated */
+		sw_desc->async_tx.flags = flags;
+		list_for_each_entry(iter, &sw_desc->group_list, chain_node) {
+			ppc440spe_desc_set_byte_count(iter, ppc440spe_chan,
+						      len);
+			iter->unmap_len = len;
+		}
+
+		if (pdest) {
+			struct dma_cdb *hw_desc;
+			struct ppc440spe_adma_chan *chan;
+
+			iter = sw_desc->group_head;
+			chan = to_ppc440spe_adma_chan(iter->async_tx.chan);
+			memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+			iter->hw_next = list_entry(iter->chain_node.next,
+						struct ppc440spe_adma_desc_slot,
+						chain_node);
+			hw_desc = iter->hw_desc;
+			hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+			iter->src_cnt = 0;
+			iter->dst_cnt = 0;
+			ppc440spe_desc_set_dest_addr(iter, chan, 0,
+						     ppc440spe_chan->pdest, 0);
+			ppc440spe_desc_set_src_addr(iter, chan, 0, 0, pdest);
+			ppc440spe_desc_set_byte_count(iter, ppc440spe_chan,
+						      len);
+			iter->unmap_len = 0;
+			/* override pdest to preserve original P */
+			pdest = ppc440spe_chan->pdest;
+		}
+		if (qdest) {
+			struct dma_cdb *hw_desc;
+			struct ppc440spe_adma_chan *chan;
+
+			iter = list_first_entry(&sw_desc->group_list,
+						struct ppc440spe_adma_desc_slot,
+						chain_node);
+			chan = to_ppc440spe_adma_chan(iter->async_tx.chan);
+
+			if (pdest) {
+				iter = list_entry(iter->chain_node.next,
+						struct ppc440spe_adma_desc_slot,
+						chain_node);
+			}
+
+			memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+			iter->hw_next = list_entry(iter->chain_node.next,
+						struct ppc440spe_adma_desc_slot,
+						chain_node);
+			hw_desc = iter->hw_desc;
+			hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+			iter->src_cnt = 0;
+			iter->dst_cnt = 0;
+			ppc440spe_desc_set_dest_addr(iter, chan, 0,
+						     ppc440spe_chan->qdest, 0);
+			ppc440spe_desc_set_src_addr(iter, chan, 0, 0, qdest);
+			ppc440spe_desc_set_byte_count(iter, ppc440spe_chan,
+						      len);
+			iter->unmap_len = 0;
+			/* override qdest to preserve original Q */
+			qdest = ppc440spe_chan->qdest;
+		}
+
+		/* Setup destinations for P/Q ops */
+		ppc440spe_adma_pqzero_sum_set_dest(sw_desc, pdest, qdest);
+
+		/* Setup zero QWORDs into DCHECK CDBs */
+		idst = dst_cnt;
+		list_for_each_entry_reverse(iter, &sw_desc->group_list,
+					    chain_node) {
+			/*
+			 * The last CDB corresponds to Q-parity check,
+			 * the one before last CDB corresponds
+			 * P-parity check
+			 */
+			if (idst == DMA_DEST_MAX_NUM) {
+				if (idst == dst_cnt) {
+					set_bit(PPC440SPE_DESC_QCHECK,
+						&iter->flags);
+				} else {
+					set_bit(PPC440SPE_DESC_PCHECK,
+						&iter->flags);
+				}
+			} else {
+				if (qdest) {
+					set_bit(PPC440SPE_DESC_QCHECK,
+						&iter->flags);
+				} else {
+					set_bit(PPC440SPE_DESC_PCHECK,
+						&iter->flags);
+				}
+			}
+			iter->xor_check_result = pqres;
+
+			/*
+			 * set it to zero, if check fail then result will
+			 * be updated
+			 */
+			*iter->xor_check_result = 0;
+			ppc440spe_desc_set_dcheck(iter, ppc440spe_chan,
+				ppc440spe_qword);
+
+			if (!(--dst_cnt))
+				break;
+		}
+
+		/* Setup sources and mults for P/Q ops */
+		list_for_each_entry_continue_reverse(iter, &sw_desc->group_list,
+						     chain_node) {
+			struct ppc440spe_adma_chan *chan;
+			u32 mult_dst;
+
+			chan = to_ppc440spe_adma_chan(iter->async_tx.chan);
+			ppc440spe_desc_set_src_addr(iter, chan, 0,
+						    DMA_CUED_XOR_HB,
+						    src[src_cnt - 1]);
+			if (qdest) {
+				mult_dst = (dst_cnt - 1) ? DMA_CDB_SG_DST2 :
+							   DMA_CDB_SG_DST1;
+				ppc440spe_desc_set_src_mult(iter, chan,
+							    DMA_CUED_MULT1_OFF,
+							    mult_dst,
+							    scf[src_cnt - 1]);
+			}
+			if (!(--src_cnt))
+				break;
+		}
+	}
+	spin_unlock_bh(&ppc440spe_chan->lock);
+	return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_xor_zero_sum - prepare CDB group for
+ * XOR ZERO_SUM operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_xor_zero_sum(
+		struct dma_chan *chan, dma_addr_t *src, unsigned int src_cnt,
+		size_t len, enum sum_check_flags *result, unsigned long flags)
+{
+	struct dma_async_tx_descriptor *tx;
+	dma_addr_t pq[2];
+
+	/* validate P, disable Q */
+	pq[0] = src[0];
+	pq[1] = 0;
+	flags |= DMA_PREP_PQ_DISABLE_Q;
+
+	tx = ppc440spe_adma_prep_dma_pqzero_sum(chan, pq, &src[1],
+						src_cnt - 1, 0, len,
+						result, flags);
+	return tx;
+}
+
+/**
+ * ppc440spe_adma_set_dest - set destination address into descriptor
+ */
+static void ppc440spe_adma_set_dest(struct ppc440spe_adma_desc_slot *sw_desc,
+		dma_addr_t addr, int index)
+{
+	struct ppc440spe_adma_chan *chan;
+
+	BUG_ON(index >= sw_desc->dst_cnt);
+
+	chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		/* to do: support transfers lengths >
+		 * PPC440SPE_ADMA_DMA/XOR_MAX_BYTE_COUNT
+		 */
+		ppc440spe_desc_set_dest_addr(sw_desc->group_head,
+			chan, 0, addr, index);
+		break;
+	case PPC440SPE_XOR_ID:
+		sw_desc = ppc440spe_get_group_entry(sw_desc, index);
+		ppc440spe_desc_set_dest_addr(sw_desc,
+			chan, 0, addr, index);
+		break;
+	}
+}
+
+static void ppc440spe_adma_pq_zero_op(struct ppc440spe_adma_desc_slot *iter,
+		struct ppc440spe_adma_chan *chan, dma_addr_t addr)
+{
+	/*  To clear destinations update the descriptor
+	 * (P or Q depending on index) as follows:
+	 * addr is destination (0 corresponds to SG2):
+	 */
+	ppc440spe_desc_set_dest_addr(iter, chan, DMA_CUED_XOR_BASE, addr, 0);
+
+	/* ... and the addr is source: */
+	ppc440spe_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB, addr);
+
+	/* addr is always SG2 then the mult is always DST1 */
+	ppc440spe_desc_set_src_mult(iter, chan, DMA_CUED_MULT1_OFF,
+				    DMA_CDB_SG_DST1, 1);
+}
+
+/**
+ * ppc440spe_adma_pq_set_dest - set destination address into descriptor
+ * for the PQXOR operation
+ */
+static void ppc440spe_adma_pq_set_dest(struct ppc440spe_adma_desc_slot *sw_desc,
+		dma_addr_t *addrs, unsigned long flags)
+{
+	struct ppc440spe_adma_desc_slot *iter;
+	struct ppc440spe_adma_chan *chan;
+	dma_addr_t paddr, qaddr;
+	dma_addr_t addr = 0, ppath, qpath;
+	int index = 0, i;
+
+	chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+
+	if (flags & DMA_PREP_PQ_DISABLE_P)
+		paddr = 0;
+	else
+		paddr = addrs[0];
+
+	if (flags & DMA_PREP_PQ_DISABLE_Q)
+		qaddr = 0;
+	else
+		qaddr = addrs[1];
+
+	if (!paddr || !qaddr)
+		addr = paddr ? paddr : qaddr;
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		/* walk through the WXOR source list and set P/Q-destinations
+		 * for each slot:
+		 */
+		if (!test_bit(PPC440SPE_DESC_RXOR, &sw_desc->flags)) {
+			/* This is WXOR-only chain; may have 1/2 zero descs */
+			if (test_bit(PPC440SPE_ZERO_P, &sw_desc->flags))
+				index++;
+			if (test_bit(PPC440SPE_ZERO_Q, &sw_desc->flags))
+				index++;
+
+			iter = ppc440spe_get_group_entry(sw_desc, index);
+			if (addr) {
+				/* one destination */
+				list_for_each_entry_from(iter,
+					&sw_desc->group_list, chain_node)
+					ppc440spe_desc_set_dest_addr(iter, chan,
+						DMA_CUED_XOR_BASE, addr, 0);
+			} else {
+				/* two destinations */
+				list_for_each_entry_from(iter,
+					&sw_desc->group_list, chain_node) {
+					ppc440spe_desc_set_dest_addr(iter, chan,
+						DMA_CUED_XOR_BASE, paddr, 0);
+					ppc440spe_desc_set_dest_addr(iter, chan,
+						DMA_CUED_XOR_BASE, qaddr, 1);
+				}
+			}
+
+			if (index) {
+				/*  To clear destinations update the descriptor
+				 * (1st,2nd, or both depending on flags)
+				 */
+				index = 0;
+				if (test_bit(PPC440SPE_ZERO_P,
+						&sw_desc->flags)) {
+					iter = ppc440spe_get_group_entry(
+							sw_desc, index++);
+					ppc440spe_adma_pq_zero_op(iter, chan,
+							paddr);
+				}
+
+				if (test_bit(PPC440SPE_ZERO_Q,
+						&sw_desc->flags)) {
+					iter = ppc440spe_get_group_entry(
+							sw_desc, index++);
+					ppc440spe_adma_pq_zero_op(iter, chan,
+							qaddr);
+				}
+
+				return;
+			}
+		} else {
+			/* This is RXOR-only or RXOR/WXOR mixed chain */
+
+			/* If we want to include destination into calculations,
+			 * then make dest addresses cued with mult=1 (XOR).
+			 */
+			ppath = test_bit(PPC440SPE_ZERO_P, &sw_desc->flags) ?
+					DMA_CUED_XOR_HB :
+					DMA_CUED_XOR_BASE |
+						(1 << DMA_CUED_MULT1_OFF);
+			qpath = test_bit(PPC440SPE_ZERO_Q, &sw_desc->flags) ?
+					DMA_CUED_XOR_HB :
+					DMA_CUED_XOR_BASE |
+						(1 << DMA_CUED_MULT1_OFF);
+
+			/* Setup destination(s) in RXOR slot(s) */
+			iter = ppc440spe_get_group_entry(sw_desc, index++);
+			ppc440spe_desc_set_dest_addr(iter, chan,
+						paddr ? ppath : qpath,
+						paddr ? paddr : qaddr, 0);
+			if (!addr) {
+				/* two destinations */
+				iter = ppc440spe_get_group_entry(sw_desc,
+								 index++);
+				ppc440spe_desc_set_dest_addr(iter, chan,
+						qpath, qaddr, 0);
+			}
+
+			if (test_bit(PPC440SPE_DESC_WXOR, &sw_desc->flags)) {
+				/* Setup destination(s) in remaining WXOR
+				 * slots
+				 */
+				iter = ppc440spe_get_group_entry(sw_desc,
+								 index);
+				if (addr) {
+					/* one destination */
+					list_for_each_entry_from(iter,
+					    &sw_desc->group_list,
+					    chain_node)
+						ppc440spe_desc_set_dest_addr(
+							iter, chan,
+							DMA_CUED_XOR_BASE,
+							addr, 0);
+
+				} else {
+					/* two destinations */
+					list_for_each_entry_from(iter,
+					    &sw_desc->group_list,
+					    chain_node) {
+						ppc440spe_desc_set_dest_addr(
+							iter, chan,
+							DMA_CUED_XOR_BASE,
+							paddr, 0);
+						ppc440spe_desc_set_dest_addr(
+							iter, chan,
+							DMA_CUED_XOR_BASE,
+							qaddr, 1);
+					}
+				}
+			}
+
+		}
+		break;
+
+	case PPC440SPE_XOR_ID:
+		/* DMA2 descriptors have only 1 destination, so there are
+		 * two chains - one for each dest.
+		 * If we want to include destination into calculations,
+		 * then make dest addresses cued with mult=1 (XOR).
+		 */
+		ppath = test_bit(PPC440SPE_ZERO_P, &sw_desc->flags) ?
+				DMA_CUED_XOR_HB :
+				DMA_CUED_XOR_BASE |
+					(1 << DMA_CUED_MULT1_OFF);
+
+		qpath = test_bit(PPC440SPE_ZERO_Q, &sw_desc->flags) ?
+				DMA_CUED_XOR_HB :
+				DMA_CUED_XOR_BASE |
+					(1 << DMA_CUED_MULT1_OFF);
+
+		iter = ppc440spe_get_group_entry(sw_desc, 0);
+		for (i = 0; i < sw_desc->descs_per_op; i++) {
+			ppc440spe_desc_set_dest_addr(iter, chan,
+				paddr ? ppath : qpath,
+				paddr ? paddr : qaddr, 0);
+			iter = list_entry(iter->chain_node.next,
+					  struct ppc440spe_adma_desc_slot,
+					  chain_node);
+		}
+
+		if (!addr) {
+			/* Two destinations; setup Q here */
+			iter = ppc440spe_get_group_entry(sw_desc,
+				sw_desc->descs_per_op);
+			for (i = 0; i < sw_desc->descs_per_op; i++) {
+				ppc440spe_desc_set_dest_addr(iter,
+					chan, qpath, qaddr, 0);
+				iter = list_entry(iter->chain_node.next,
+						struct ppc440spe_adma_desc_slot,
+						chain_node);
+			}
+		}
+
+		break;
+	}
+}
+
+/**
+ * ppc440spe_adma_pq_zero_sum_set_dest - set destination address into descriptor
+ * for the PQ_ZERO_SUM operation
+ */
+static void ppc440spe_adma_pqzero_sum_set_dest(
+		struct ppc440spe_adma_desc_slot *sw_desc,
+		dma_addr_t paddr, dma_addr_t qaddr)
+{
+	struct ppc440spe_adma_desc_slot *iter, *end;
+	struct ppc440spe_adma_chan *chan;
+	dma_addr_t addr = 0;
+	int idx;
+
+	chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+
+	/* walk through the WXOR source list and set P/Q-destinations
+	 * for each slot
+	 */
+	idx = (paddr && qaddr) ? 2 : 1;
+	/* set end */
+	list_for_each_entry_reverse(end, &sw_desc->group_list,
+				    chain_node) {
+		if (!(--idx))
+			break;
+	}
+	/* set start */
+	idx = (paddr && qaddr) ? 2 : 1;
+	iter = ppc440spe_get_group_entry(sw_desc, idx);
+
+	if (paddr && qaddr) {
+		/* two destinations */
+		list_for_each_entry_from(iter, &sw_desc->group_list,
+					 chain_node) {
+			if (unlikely(iter == end))
+				break;
+			ppc440spe_desc_set_dest_addr(iter, chan,
+						DMA_CUED_XOR_BASE, paddr, 0);
+			ppc440spe_desc_set_dest_addr(iter, chan,
+						DMA_CUED_XOR_BASE, qaddr, 1);
+		}
+	} else {
+		/* one destination */
+		addr = paddr ? paddr : qaddr;
+		list_for_each_entry_from(iter, &sw_desc->group_list,
+					 chain_node) {
+			if (unlikely(iter == end))
+				break;
+			ppc440spe_desc_set_dest_addr(iter, chan,
+						DMA_CUED_XOR_BASE, addr, 0);
+		}
+	}
+
+	/*  The remaining descriptors are DATACHECK. These have no need in
+	 * destination. Actually, these destinations are used there
+	 * as sources for check operation. So, set addr as source.
+	 */
+	ppc440spe_desc_set_src_addr(end, chan, 0, 0, addr ? addr : paddr);
+
+	if (!addr) {
+		end = list_entry(end->chain_node.next,
+				 struct ppc440spe_adma_desc_slot, chain_node);
+		ppc440spe_desc_set_src_addr(end, chan, 0, 0, qaddr);
+	}
+}
+
+/**
+ * ppc440spe_desc_set_xor_src_cnt - set source count into descriptor
+ */
+static inline void ppc440spe_desc_set_xor_src_cnt(
+			struct ppc440spe_adma_desc_slot *desc,
+			int src_cnt)
+{
+	struct xor_cb *hw_desc = desc->hw_desc;
+
+	hw_desc->cbc &= ~XOR_CDCR_OAC_MSK;
+	hw_desc->cbc |= src_cnt;
+}
+
+/**
+ * ppc440spe_adma_pq_set_src - set source address into descriptor
+ */
+static void ppc440spe_adma_pq_set_src(struct ppc440spe_adma_desc_slot *sw_desc,
+		dma_addr_t addr, int index)
+{
+	struct ppc440spe_adma_chan *chan;
+	dma_addr_t haddr = 0;
+	struct ppc440spe_adma_desc_slot *iter = NULL;
+
+	chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		/* DMA0,1 may do: WXOR, RXOR, RXOR+WXORs chain
+		 */
+		if (test_bit(PPC440SPE_DESC_RXOR, &sw_desc->flags)) {
+			/* RXOR-only or RXOR/WXOR operation */
+			int iskip = test_bit(PPC440SPE_DESC_RXOR12,
+				&sw_desc->flags) ?  2 : 3;
+
+			if (index == 0) {
+				/* 1st slot (RXOR) */
+				/* setup sources region (R1-2-3, R1-2-4,
+				 * or R1-2-5)
+				 */
+				if (test_bit(PPC440SPE_DESC_RXOR12,
+						&sw_desc->flags))
+					haddr = DMA_RXOR12 <<
+						DMA_CUED_REGION_OFF;
+				else if (test_bit(PPC440SPE_DESC_RXOR123,
+				    &sw_desc->flags))
+					haddr = DMA_RXOR123 <<
+						DMA_CUED_REGION_OFF;
+				else if (test_bit(PPC440SPE_DESC_RXOR124,
+				    &sw_desc->flags))
+					haddr = DMA_RXOR124 <<
+						DMA_CUED_REGION_OFF;
+				else if (test_bit(PPC440SPE_DESC_RXOR125,
+				    &sw_desc->flags))
+					haddr = DMA_RXOR125 <<
+						DMA_CUED_REGION_OFF;
+				else
+					BUG();
+				haddr |= DMA_CUED_XOR_BASE;
+				iter = ppc440spe_get_group_entry(sw_desc, 0);
+			} else if (index < iskip) {
+				/* 1st slot (RXOR)
+				 * shall actually set source address only once
+				 * instead of first <iskip>
+				 */
+				iter = NULL;
+			} else {
+				/* 2nd/3d and next slots (WXOR);
+				 * skip first slot with RXOR
+				 */
+				haddr = DMA_CUED_XOR_HB;
+				iter = ppc440spe_get_group_entry(sw_desc,
+				    index - iskip + sw_desc->dst_cnt);
+			}
+		} else {
+			int znum = 0;
+
+			/* WXOR-only operation; skip first slots with
+			 * zeroing destinations
+			 */
+			if (test_bit(PPC440SPE_ZERO_P, &sw_desc->flags))
+				znum++;
+			if (test_bit(PPC440SPE_ZERO_Q, &sw_desc->flags))
+				znum++;
+
+			haddr = DMA_CUED_XOR_HB;
+			iter = ppc440spe_get_group_entry(sw_desc,
+					index + znum);
+		}
+
+		if (likely(iter)) {
+			ppc440spe_desc_set_src_addr(iter, chan, 0, haddr, addr);
+
+			if (!index &&
+			    test_bit(PPC440SPE_DESC_RXOR, &sw_desc->flags) &&
+			    sw_desc->dst_cnt == 2) {
+				/* if we have two destinations for RXOR, then
+				 * setup source in the second descr too
+				 */
+				iter = ppc440spe_get_group_entry(sw_desc, 1);
+				ppc440spe_desc_set_src_addr(iter, chan, 0,
+					haddr, addr);
+			}
+		}
+		break;
+
+	case PPC440SPE_XOR_ID:
+		/* DMA2 may do Biskup */
+		iter = sw_desc->group_head;
+		if (iter->dst_cnt == 2) {
+			/* both P & Q calculations required; set P src here */
+			ppc440spe_adma_dma2rxor_set_src(iter, index, addr);
+
+			/* this is for Q */
+			iter = ppc440spe_get_group_entry(sw_desc,
+				sw_desc->descs_per_op);
+		}
+		ppc440spe_adma_dma2rxor_set_src(iter, index, addr);
+		break;
+	}
+}
+
+/**
+ * ppc440spe_adma_memcpy_xor_set_src - set source address into descriptor
+ */
+static void ppc440spe_adma_memcpy_xor_set_src(
+		struct ppc440spe_adma_desc_slot *sw_desc,
+		dma_addr_t addr, int index)
+{
+	struct ppc440spe_adma_chan *chan;
+
+	chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+	sw_desc = sw_desc->group_head;
+
+	if (likely(sw_desc))
+		ppc440spe_desc_set_src_addr(sw_desc, chan, index, 0, addr);
+}
+
+/**
+ * ppc440spe_adma_dma2rxor_inc_addr  -
+ */
+static void ppc440spe_adma_dma2rxor_inc_addr(
+		struct ppc440spe_adma_desc_slot *desc,
+		struct ppc440spe_rxor *cursor, int index, int src_cnt)
+{
+	cursor->addr_count++;
+	if (index == src_cnt - 1) {
+		ppc440spe_desc_set_xor_src_cnt(desc, cursor->addr_count);
+	} else if (cursor->addr_count == XOR_MAX_OPS) {
+		ppc440spe_desc_set_xor_src_cnt(desc, cursor->addr_count);
+		cursor->addr_count = 0;
+		cursor->desc_count++;
+	}
+}
+
+/**
+ * ppc440spe_adma_dma2rxor_prep_src - setup RXOR types in DMA2 CDB
+ */
+static int ppc440spe_adma_dma2rxor_prep_src(
+		struct ppc440spe_adma_desc_slot *hdesc,
+		struct ppc440spe_rxor *cursor, int index,
+		int src_cnt, u32 addr)
+{
+	int rval = 0;
+	u32 sign;
+	struct ppc440spe_adma_desc_slot *desc = hdesc;
+	int i;
+
+	for (i = 0; i < cursor->desc_count; i++) {
+		desc = list_entry(hdesc->chain_node.next,
+				  struct ppc440spe_adma_desc_slot,
+				  chain_node);
+	}
+
+	switch (cursor->state) {
+	case 0:
+		if (addr == cursor->addrl + cursor->len) {
+			/* direct RXOR */
+			cursor->state = 1;
+			cursor->xor_count++;
+			if (index == src_cnt-1) {
+				ppc440spe_rxor_set_region(desc,
+					cursor->addr_count,
+					DMA_RXOR12 << DMA_CUED_REGION_OFF);
+				ppc440spe_adma_dma2rxor_inc_addr(
+					desc, cursor, index, src_cnt);
+			}
+		} else if (cursor->addrl == addr + cursor->len) {
+			/* reverse RXOR */
+			cursor->state = 1;
+			cursor->xor_count++;
+			set_bit(cursor->addr_count, &desc->reverse_flags[0]);
+			if (index == src_cnt-1) {
+				ppc440spe_rxor_set_region(desc,
+					cursor->addr_count,
+					DMA_RXOR12 << DMA_CUED_REGION_OFF);
+				ppc440spe_adma_dma2rxor_inc_addr(
+					desc, cursor, index, src_cnt);
+			}
+		} else {
+			printk(KERN_ERR "Cannot build "
+				"DMA2 RXOR command block.\n");
+			BUG();
+		}
+		break;
+	case 1:
+		sign = test_bit(cursor->addr_count,
+				desc->reverse_flags)
+			? -1 : 1;
+		if (index == src_cnt-2 || (sign == -1
+			&& addr != cursor->addrl - 2*cursor->len)) {
+			cursor->state = 0;
+			cursor->xor_count = 1;
+			cursor->addrl = addr;
+			ppc440spe_rxor_set_region(desc,
+				cursor->addr_count,
+				DMA_RXOR12 << DMA_CUED_REGION_OFF);
+			ppc440spe_adma_dma2rxor_inc_addr(
+				desc, cursor, index, src_cnt);
+		} else if (addr == cursor->addrl + 2*sign*cursor->len) {
+			cursor->state = 2;
+			cursor->xor_count = 0;
+			ppc440spe_rxor_set_region(desc,
+				cursor->addr_count,
+				DMA_RXOR123 << DMA_CUED_REGION_OFF);
+			if (index == src_cnt-1) {
+				ppc440spe_adma_dma2rxor_inc_addr(
+					desc, cursor, index, src_cnt);
+			}
+		} else if (addr == cursor->addrl + 3*cursor->len) {
+			cursor->state = 2;
+			cursor->xor_count = 0;
+			ppc440spe_rxor_set_region(desc,
+				cursor->addr_count,
+				DMA_RXOR124 << DMA_CUED_REGION_OFF);
+			if (index == src_cnt-1) {
+				ppc440spe_adma_dma2rxor_inc_addr(
+					desc, cursor, index, src_cnt);
+			}
+		} else if (addr == cursor->addrl + 4*cursor->len) {
+			cursor->state = 2;
+			cursor->xor_count = 0;
+			ppc440spe_rxor_set_region(desc,
+				cursor->addr_count,
+				DMA_RXOR125 << DMA_CUED_REGION_OFF);
+			if (index == src_cnt-1) {
+				ppc440spe_adma_dma2rxor_inc_addr(
+					desc, cursor, index, src_cnt);
+			}
+		} else {
+			cursor->state = 0;
+			cursor->xor_count = 1;
+			cursor->addrl = addr;
+			ppc440spe_rxor_set_region(desc,
+				cursor->addr_count,
+				DMA_RXOR12 << DMA_CUED_REGION_OFF);
+			ppc440spe_adma_dma2rxor_inc_addr(
+				desc, cursor, index, src_cnt);
+		}
+		break;
+	case 2:
+		cursor->state = 0;
+		cursor->addrl = addr;
+		cursor->xor_count++;
+		if (index) {
+			ppc440spe_adma_dma2rxor_inc_addr(
+				desc, cursor, index, src_cnt);
+		}
+		break;
+	}
+
+	return rval;
+}
+
+/**
+ * ppc440spe_adma_dma2rxor_set_src - set RXOR source address; it's assumed that
+ *	ppc440spe_adma_dma2rxor_prep_src() has already done prior this call
+ */
+static void ppc440spe_adma_dma2rxor_set_src(
+		struct ppc440spe_adma_desc_slot *desc,
+		int index, dma_addr_t addr)
+{
+	struct xor_cb *xcb = desc->hw_desc;
+	int k = 0, op = 0, lop = 0;
+
+	/* get the RXOR operand which corresponds to index addr */
+	while (op <= index) {
+		lop = op;
+		if (k == XOR_MAX_OPS) {
+			k = 0;
+			desc = list_entry(desc->chain_node.next,
+				struct ppc440spe_adma_desc_slot, chain_node);
+			xcb = desc->hw_desc;
+
+		}
+		if ((xcb->ops[k++].h & (DMA_RXOR12 << DMA_CUED_REGION_OFF)) ==
+		    (DMA_RXOR12 << DMA_CUED_REGION_OFF))
+			op += 2;
+		else
+			op += 3;
+	}
+
+	BUG_ON(k < 1);
+
+	if (test_bit(k-1, desc->reverse_flags)) {
+		/* reverse operand order; put last op in RXOR group */
+		if (index == op - 1)
+			ppc440spe_rxor_set_src(desc, k - 1, addr);
+	} else {
+		/* direct operand order; put first op in RXOR group */
+		if (index == lop)
+			ppc440spe_rxor_set_src(desc, k - 1, addr);
+	}
+}
+
+/**
+ * ppc440spe_adma_dma2rxor_set_mult - set RXOR multipliers; it's assumed that
+ *	ppc440spe_adma_dma2rxor_prep_src() has already done prior this call
+ */
+static void ppc440spe_adma_dma2rxor_set_mult(
+		struct ppc440spe_adma_desc_slot *desc,
+		int index, u8 mult)
+{
+	struct xor_cb *xcb = desc->hw_desc;
+	int k = 0, op = 0, lop = 0;
+
+	/* get the RXOR operand which corresponds to index mult */
+	while (op <= index) {
+		lop = op;
+		if (k == XOR_MAX_OPS) {
+			k = 0;
+			desc = list_entry(desc->chain_node.next,
+					  struct ppc440spe_adma_desc_slot,
+					  chain_node);
+			xcb = desc->hw_desc;
+
+		}
+		if ((xcb->ops[k++].h & (DMA_RXOR12 << DMA_CUED_REGION_OFF)) ==
+		    (DMA_RXOR12 << DMA_CUED_REGION_OFF))
+			op += 2;
+		else
+			op += 3;
+	}
+
+	BUG_ON(k < 1);
+	if (test_bit(k-1, desc->reverse_flags)) {
+		/* reverse order */
+		ppc440spe_rxor_set_mult(desc, k - 1, op - index - 1, mult);
+	} else {
+		/* direct order */
+		ppc440spe_rxor_set_mult(desc, k - 1, index - lop, mult);
+	}
+}
+
+/**
+ * ppc440spe_init_rxor_cursor -
+ */
+static void ppc440spe_init_rxor_cursor(struct ppc440spe_rxor *cursor)
+{
+	memset(cursor, 0, sizeof(struct ppc440spe_rxor));
+	cursor->state = 2;
+}
+
+/**
+ * ppc440spe_adma_pq_set_src_mult - set multiplication coefficient into
+ * descriptor for the PQXOR operation
+ */
+static void ppc440spe_adma_pq_set_src_mult(
+		struct ppc440spe_adma_desc_slot *sw_desc,
+		unsigned char mult, int index, int dst_pos)
+{
+	struct ppc440spe_adma_chan *chan;
+	u32 mult_idx, mult_dst;
+	struct ppc440spe_adma_desc_slot *iter = NULL, *iter1 = NULL;
+
+	chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+
+	switch (chan->device->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		if (test_bit(PPC440SPE_DESC_RXOR, &sw_desc->flags)) {
+			int region = test_bit(PPC440SPE_DESC_RXOR12,
+					&sw_desc->flags) ? 2 : 3;
+
+			if (index < region) {
+				/* RXOR multipliers */
+				iter = ppc440spe_get_group_entry(sw_desc,
+					sw_desc->dst_cnt - 1);
+				if (sw_desc->dst_cnt == 2)
+					iter1 = ppc440spe_get_group_entry(
+							sw_desc, 0);
+
+				mult_idx = DMA_CUED_MULT1_OFF + (index << 3);
+				mult_dst = DMA_CDB_SG_SRC;
+			} else {
+				/* WXOR multiplier */
+				iter = ppc440spe_get_group_entry(sw_desc,
+							index - region +
+							sw_desc->dst_cnt);
+				mult_idx = DMA_CUED_MULT1_OFF;
+				mult_dst = dst_pos ? DMA_CDB_SG_DST2 :
+						     DMA_CDB_SG_DST1;
+			}
+		} else {
+			int znum = 0;
+
+			/* WXOR-only;
+			 * skip first slots with destinations (if ZERO_DST has
+			 * place)
+			 */
+			if (test_bit(PPC440SPE_ZERO_P, &sw_desc->flags))
+				znum++;
+			if (test_bit(PPC440SPE_ZERO_Q, &sw_desc->flags))
+				znum++;
+
+			iter = ppc440spe_get_group_entry(sw_desc, index + znum);
+			mult_idx = DMA_CUED_MULT1_OFF;
+			mult_dst = dst_pos ? DMA_CDB_SG_DST2 : DMA_CDB_SG_DST1;
+		}
+
+		if (likely(iter)) {
+			ppc440spe_desc_set_src_mult(iter, chan,
+				mult_idx, mult_dst, mult);
+
+			if (unlikely(iter1)) {
+				/* if we have two destinations for RXOR, then
+				 * we've just set Q mult. Set-up P now.
+				 */
+				ppc440spe_desc_set_src_mult(iter1, chan,
+					mult_idx, mult_dst, 1);
+			}
+
+		}
+		break;
+
+	case PPC440SPE_XOR_ID:
+		iter = sw_desc->group_head;
+		if (sw_desc->dst_cnt == 2) {
+			/* both P & Q calculations required; set P mult here */
+			ppc440spe_adma_dma2rxor_set_mult(iter, index, 1);
+
+			/* and then set Q mult */
+			iter = ppc440spe_get_group_entry(sw_desc,
+			       sw_desc->descs_per_op);
+		}
+		ppc440spe_adma_dma2rxor_set_mult(iter, index, mult);
+		break;
+	}
+}
+
+/**
+ * ppc440spe_adma_free_chan_resources - free the resources allocated
+ */
+static void ppc440spe_adma_free_chan_resources(struct dma_chan *chan)
+{
+	struct ppc440spe_adma_chan *ppc440spe_chan;
+	struct ppc440spe_adma_desc_slot *iter, *_iter;
+	int in_use_descs = 0;
+
+	ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+	ppc440spe_adma_slot_cleanup(ppc440spe_chan);
+
+	spin_lock_bh(&ppc440spe_chan->lock);
+	list_for_each_entry_safe(iter, _iter, &ppc440spe_chan->chain,
+					chain_node) {
+		in_use_descs++;
+		list_del(&iter->chain_node);
+	}
+	list_for_each_entry_safe_reverse(iter, _iter,
+			&ppc440spe_chan->all_slots, slot_node) {
+		list_del(&iter->slot_node);
+		kfree(iter);
+		ppc440spe_chan->slots_allocated--;
+	}
+	ppc440spe_chan->last_used = NULL;
+
+	dev_dbg(ppc440spe_chan->device->common.dev,
+		"ppc440spe adma%d %s slots_allocated %d\n",
+		ppc440spe_chan->device->id,
+		__func__, ppc440spe_chan->slots_allocated);
+	spin_unlock_bh(&ppc440spe_chan->lock);
+
+	/* one is ok since we left it on there on purpose */
+	if (in_use_descs > 1)
+		printk(KERN_ERR "SPE: Freeing %d in use descriptors!\n",
+			in_use_descs - 1);
+}
+
+/**
+ * ppc440spe_adma_is_complete - poll the status of an ADMA transaction
+ * @chan: ADMA channel handle
+ * @cookie: ADMA transaction identifier
+ */
+static enum dma_status ppc440spe_adma_is_complete(struct dma_chan *chan,
+	dma_cookie_t cookie, dma_cookie_t *done, dma_cookie_t *used)
+{
+	struct ppc440spe_adma_chan *ppc440spe_chan;
+	dma_cookie_t last_used;
+	dma_cookie_t last_complete;
+	enum dma_status ret;
+
+	ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+	last_used = chan->cookie;
+	last_complete = ppc440spe_chan->completed_cookie;
+
+	if (done)
+		*done = last_complete;
+	if (used)
+		*used = last_used;
+
+	ret = dma_async_is_complete(cookie, last_complete, last_used);
+	if (ret == DMA_SUCCESS)
+		return ret;
+
+	ppc440spe_adma_slot_cleanup(ppc440spe_chan);
+
+	last_used = chan->cookie;
+	last_complete = ppc440spe_chan->completed_cookie;
+
+	if (done)
+		*done = last_complete;
+	if (used)
+		*used = last_used;
+
+	return dma_async_is_complete(cookie, last_complete, last_used);
+}
+
+/**
+ * ppc440spe_adma_eot_handler - end of transfer interrupt handler
+ */
+static irqreturn_t ppc440spe_adma_eot_handler(int irq, void *data)
+{
+	struct ppc440spe_adma_chan *chan = data;
+
+	dev_dbg(chan->device->common.dev,
+		"ppc440spe adma%d: %s\n", chan->device->id, __func__);
+
+	tasklet_schedule(&chan->irq_tasklet);
+	ppc440spe_adma_device_clear_eot_status(chan);
+
+	return IRQ_HANDLED;
+}
+
+/**
+ * ppc440spe_adma_err_handler - DMA error interrupt handler;
+ *	do the same things as a eot handler
+ */
+static irqreturn_t ppc440spe_adma_err_handler(int irq, void *data)
+{
+	struct ppc440spe_adma_chan *chan = data;
+
+	dev_dbg(chan->device->common.dev,
+		"ppc440spe adma%d: %s\n", chan->device->id, __func__);
+
+	tasklet_schedule(&chan->irq_tasklet);
+	ppc440spe_adma_device_clear_eot_status(chan);
+
+	return IRQ_HANDLED;
+}
+
+/**
+ * ppc440spe_test_callback - called when test operation has been done
+ */
+static void ppc440spe_test_callback(void *unused)
+{
+	complete(&ppc440spe_r6_test_comp);
+}
+
+/**
+ * ppc440spe_adma_issue_pending - flush all pending descriptors to h/w
+ */
+static void ppc440spe_adma_issue_pending(struct dma_chan *chan)
+{
+	struct ppc440spe_adma_chan *ppc440spe_chan;
+
+	ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+	dev_dbg(ppc440spe_chan->device->common.dev,
+		"ppc440spe adma%d: %s %d \n", ppc440spe_chan->device->id,
+		__func__, ppc440spe_chan->pending);
+
+	if (ppc440spe_chan->pending) {
+		ppc440spe_chan->pending = 0;
+		ppc440spe_chan_append(ppc440spe_chan);
+	}
+}
+
+/**
+ * ppc440spe_chan_start_null_xor - initiate the first XOR operation (DMA engines
+ *	use FIFOs (as opposite to chains used in XOR) so this is a XOR
+ *	specific operation)
+ */
+static void ppc440spe_chan_start_null_xor(struct ppc440spe_adma_chan *chan)
+{
+	struct ppc440spe_adma_desc_slot *sw_desc, *group_start;
+	dma_cookie_t cookie;
+	int slot_cnt, slots_per_op;
+
+	dev_dbg(chan->device->common.dev,
+		"ppc440spe adma%d: %s\n", chan->device->id, __func__);
+
+	spin_lock_bh(&chan->lock);
+	slot_cnt = ppc440spe_chan_xor_slot_count(0, 2, &slots_per_op);
+	sw_desc = ppc440spe_adma_alloc_slots(chan, slot_cnt, slots_per_op);
+	if (sw_desc) {
+		group_start = sw_desc->group_head;
+		list_splice_init(&sw_desc->group_list, &chan->chain);
+		async_tx_ack(&sw_desc->async_tx);
+		ppc440spe_desc_init_null_xor(group_start);
+
+		cookie = chan->common.cookie;
+		cookie++;
+		if (cookie <= 1)
+			cookie = 2;
+
+		/* initialize the completed cookie to be less than
+		 * the most recently used cookie
+		 */
+		chan->completed_cookie = cookie - 1;
+		chan->common.cookie = sw_desc->async_tx.cookie = cookie;
+
+		/* channel should not be busy */
+		BUG_ON(ppc440spe_chan_is_busy(chan));
+
+		/* set the descriptor address */
+		ppc440spe_chan_set_first_xor_descriptor(chan, sw_desc);
+
+		/* run the descriptor */
+		ppc440spe_chan_run(chan);
+	} else
+		printk(KERN_ERR "ppc440spe adma%d"
+			" failed to allocate null descriptor\n",
+			chan->device->id);
+	spin_unlock_bh(&chan->lock);
+}
+
+/**
+ * ppc440spe_test_raid6 - test are RAID-6 capabilities enabled successfully.
+ *	For this we just perform one WXOR operation with the same source
+ *	and destination addresses, the GF-multiplier is 1; so if RAID-6
+ *	capabilities are enabled then we'll get src/dst filled with zero.
+ */
+static int ppc440spe_test_raid6(struct ppc440spe_adma_chan *chan)
+{
+	struct ppc440spe_adma_desc_slot *sw_desc, *iter;
+	struct page *pg;
+	char *a;
+	dma_addr_t dma_addr, addrs[2];
+	unsigned long op = 0;
+	int rval = 0;
+
+	set_bit(PPC440SPE_DESC_WXOR, &op);
+
+	pg = alloc_page(GFP_KERNEL);
+	if (!pg)
+		return -ENOMEM;
+
+	spin_lock_bh(&chan->lock);
+	sw_desc = ppc440spe_adma_alloc_slots(chan, 1, 1);
+	if (sw_desc) {
+		/* 1 src, 1 dsr, int_ena, WXOR */
+		ppc440spe_desc_init_dma01pq(sw_desc, 1, 1, 1, op);
+		list_for_each_entry(iter, &sw_desc->group_list, chain_node) {
+			ppc440spe_desc_set_byte_count(iter, chan, PAGE_SIZE);
+			iter->unmap_len = PAGE_SIZE;
+		}
+	} else {
+		rval = -EFAULT;
+		spin_unlock_bh(&chan->lock);
+		goto exit;
+	}
+	spin_unlock_bh(&chan->lock);
+
+	/* Fill the test page with ones */
+	memset(page_address(pg), 0xFF, PAGE_SIZE);
+	dma_addr = dma_map_page(chan->device->dev, pg, 0,
+				PAGE_SIZE, DMA_BIDIRECTIONAL);
+
+	/* Setup addresses */
+	ppc440spe_adma_pq_set_src(sw_desc, dma_addr, 0);
+	ppc440spe_adma_pq_set_src_mult(sw_desc, 1, 0, 0);
+	addrs[0] = dma_addr;
+	addrs[1] = 0;
+	ppc440spe_adma_pq_set_dest(sw_desc, addrs, DMA_PREP_PQ_DISABLE_Q);
+
+	async_tx_ack(&sw_desc->async_tx);
+	sw_desc->async_tx.callback = ppc440spe_test_callback;
+	sw_desc->async_tx.callback_param = NULL;
+
+	init_completion(&ppc440spe_r6_test_comp);
+
+	ppc440spe_adma_tx_submit(&sw_desc->async_tx);
+	ppc440spe_adma_issue_pending(&chan->common);
+
+	wait_for_completion(&ppc440spe_r6_test_comp);
+
+	/* Now check if the test page is zeroed */
+	a = page_address(pg);
+	if ((*(u32 *)a) == 0 && memcmp(a, a+4, PAGE_SIZE-4) == 0) {
+		/* page is zero - RAID-6 enabled */
+		rval = 0;
+	} else {
+		/* RAID-6 was not enabled */
+		rval = -EINVAL;
+	}
+exit:
+	__free_page(pg);
+	return rval;
+}
+
+static void ppc440spe_adma_init_capabilities(struct ppc440spe_adma_device *adev)
+{
+	switch (adev->id) {
+	case PPC440SPE_DMA0_ID:
+	case PPC440SPE_DMA1_ID:
+		dma_cap_set(DMA_MEMCPY, adev->common.cap_mask);
+		dma_cap_set(DMA_INTERRUPT, adev->common.cap_mask);
+		dma_cap_set(DMA_MEMSET, adev->common.cap_mask);
+		dma_cap_set(DMA_PQ, adev->common.cap_mask);
+		dma_cap_set(DMA_PQ_VAL, adev->common.cap_mask);
+		dma_cap_set(DMA_XOR_VAL, adev->common.cap_mask);
+		break;
+	case PPC440SPE_XOR_ID:
+		dma_cap_set(DMA_XOR, adev->common.cap_mask);
+		dma_cap_set(DMA_PQ, adev->common.cap_mask);
+		dma_cap_set(DMA_INTERRUPT, adev->common.cap_mask);
+		adev->common.cap_mask = adev->common.cap_mask;
+		break;
+	}
+
+	/* Set base routines */
+	adev->common.device_alloc_chan_resources =
+				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_issue_pending = ppc440spe_adma_issue_pending;
+
+	/* Set prep routines based on capability */
+	if (dma_has_cap(DMA_MEMCPY, adev->common.cap_mask)) {
+		adev->common.device_prep_dma_memcpy =
+			ppc440spe_adma_prep_dma_memcpy;
+	}
+	if (dma_has_cap(DMA_MEMSET, adev->common.cap_mask)) {
+		adev->common.device_prep_dma_memset =
+			ppc440spe_adma_prep_dma_memset;
+	}
+	if (dma_has_cap(DMA_XOR, adev->common.cap_mask)) {
+		adev->common.max_xor = XOR_MAX_OPS;
+		adev->common.device_prep_dma_xor =
+			ppc440spe_adma_prep_dma_xor;
+	}
+	if (dma_has_cap(DMA_PQ, adev->common.cap_mask)) {
+		switch (adev->id) {
+		case PPC440SPE_DMA0_ID:
+			dma_set_maxpq(&adev->common,
+				DMA0_FIFO_SIZE / sizeof(struct dma_cdb), 0);
+			break;
+		case PPC440SPE_DMA1_ID:
+			dma_set_maxpq(&adev->common,
+				DMA1_FIFO_SIZE / sizeof(struct dma_cdb), 0);
+			break;
+		case PPC440SPE_XOR_ID:
+			adev->common.max_pq = XOR_MAX_OPS * 3;
+			break;
+		}
+		adev->common.device_prep_dma_pq =
+			ppc440spe_adma_prep_dma_pq;
+	}
+	if (dma_has_cap(DMA_PQ_VAL, adev->common.cap_mask)) {
+		switch (adev->id) {
+		case PPC440SPE_DMA0_ID:
+			adev->common.max_pq = DMA0_FIFO_SIZE /
+						sizeof(struct dma_cdb);
+			break;
+		case PPC440SPE_DMA1_ID:
+			adev->common.max_pq = DMA1_FIFO_SIZE /
+						sizeof(struct dma_cdb);
+			break;
+		}
+		adev->common.device_prep_dma_pq_val =
+			ppc440spe_adma_prep_dma_pqzero_sum;
+	}
+	if (dma_has_cap(DMA_XOR_VAL, adev->common.cap_mask)) {
+		switch (adev->id) {
+		case PPC440SPE_DMA0_ID:
+			adev->common.max_xor = DMA0_FIFO_SIZE /
+						sizeof(struct dma_cdb);
+			break;
+		case PPC440SPE_DMA1_ID:
+			adev->common.max_xor = DMA1_FIFO_SIZE /
+						sizeof(struct dma_cdb);
+			break;
+		}
+		adev->common.device_prep_dma_xor_val =
+			ppc440spe_adma_prep_dma_xor_zero_sum;
+	}
+	if (dma_has_cap(DMA_INTERRUPT, adev->common.cap_mask)) {
+		adev->common.device_prep_dma_interrupt =
+			ppc440spe_adma_prep_dma_interrupt;
+	}
+	pr_info("%s: AMCC(R) PPC440SP(E) ADMA Engine: "
+	  "( %s%s%s%s%s%s%s)\n",
+	  dev_name(adev->dev),
+	  dma_has_cap(DMA_PQ, adev->common.cap_mask) ? "pq " : "",
+	  dma_has_cap(DMA_PQ_VAL, adev->common.cap_mask) ? "pq_val " : "",
+	  dma_has_cap(DMA_XOR, adev->common.cap_mask) ? "xor " : "",
+	  dma_has_cap(DMA_XOR_VAL, adev->common.cap_mask) ? "xor_val " : "",
+	  dma_has_cap(DMA_MEMCPY, adev->common.cap_mask) ? "memcpy " : "",
+	  dma_has_cap(DMA_MEMSET, adev->common.cap_mask)  ? "memset " : "",
+	  dma_has_cap(DMA_INTERRUPT, adev->common.cap_mask) ? "intr " : "");
+}
+
+static int ppc440spe_adma_setup_irqs(struct ppc440spe_adma_device *adev,
+				     struct ppc440spe_adma_chan *chan,
+				     int *initcode)
+{
+	struct device_node *np;
+	int ret;
+
+	np = container_of(adev->dev, struct of_device, dev)->node;
+	if (adev->id != PPC440SPE_XOR_ID) {
+		adev->err_irq = irq_of_parse_and_map(np, 1);
+		if (adev->err_irq == NO_IRQ) {
+			dev_warn(adev->dev, "no err irq resource?\n");
+			*initcode = PPC_ADMA_INIT_IRQ2;
+			adev->err_irq = -ENXIO;
+		} else
+			atomic_inc(&ppc440spe_adma_err_irq_ref);
+	} else {
+		adev->err_irq = -ENXIO;
+	}
+
+	adev->irq = irq_of_parse_and_map(np, 0);
+	if (adev->irq == NO_IRQ) {
+		dev_err(adev->dev, "no irq resource\n");
+		*initcode = PPC_ADMA_INIT_IRQ1;
+		ret = -ENXIO;
+		goto err_irq_map;
+	}
+	dev_dbg(adev->dev, "irq %d, err irq %d\n",
+		adev->irq, adev->err_irq);
+
+	ret = request_irq(adev->irq, ppc440spe_adma_eot_handler,
+			  0, dev_driver_string(adev->dev), chan);
+	if (ret) {
+		dev_err(adev->dev, "can't request irq %d\n",
+			adev->irq);
+		*initcode = PPC_ADMA_INIT_IRQ1;
+		ret = -EIO;
+		goto err_req1;
+	}
+
+	/* only DMA engines have a separate error IRQ
+	 * so it's Ok if err_irq < 0 in XOR engine case.
+	 */
+	if (adev->err_irq > 0) {
+		/* both DMA engines share common error IRQ */
+		ret = request_irq(adev->err_irq,
+				  ppc440spe_adma_err_handler,
+				  IRQF_SHARED,
+				  dev_driver_string(adev->dev),
+				  chan);
+		if (ret) {
+			dev_err(adev->dev, "can't request irq %d\n",
+				adev->err_irq);
+			*initcode = PPC_ADMA_INIT_IRQ2;
+			ret = -EIO;
+			goto err_req2;
+		}
+	}
+
+	if (adev->id == PPC440SPE_XOR_ID) {
+		/* enable XOR engine interrupts */
+		iowrite32be(XOR_IE_CBCIE_BIT | XOR_IE_ICBIE_BIT |
+			    XOR_IE_ICIE_BIT | XOR_IE_RPTIE_BIT,
+			    &adev->xor_reg->ier);
+	} else {
+		u32 mask, enable;
+
+		np = of_find_compatible_node(NULL, NULL, "ibm,i2o-440spe");
+		if (!np) {
+			pr_err("%s: can't find I2O device tree node\n",
+				__func__);
+			ret = -ENODEV;
+			goto err_req2;
+		}
+		adev->i2o_reg = of_iomap(np, 0);
+		if (!adev->i2o_reg) {
+			pr_err("%s: failed to map I2O registers\n", __func__);
+			of_node_put(np);
+			ret = -EINVAL;
+			goto err_req2;
+		}
+		of_node_put(np);
+		/* Unmask 'CS FIFO Attention' interrupts and
+		 * enable generating interrupts on errors
+		 */
+		enable = (adev->id == PPC440SPE_DMA0_ID) ?
+			 ~(I2O_IOPIM_P0SNE | I2O_IOPIM_P0EM) :
+			 ~(I2O_IOPIM_P1SNE | I2O_IOPIM_P1EM);
+		mask = ioread32(&adev->i2o_reg->iopim) & enable;
+		iowrite32(mask, &adev->i2o_reg->iopim);
+	}
+	return 0;
+
+err_req2:
+	free_irq(adev->irq, chan);
+err_req1:
+	irq_dispose_mapping(adev->irq);
+err_irq_map:
+	if (adev->err_irq > 0) {
+		if (atomic_dec_and_test(&ppc440spe_adma_err_irq_ref))
+			irq_dispose_mapping(adev->err_irq);
+	}
+	return ret;
+}
+
+static void ppc440spe_adma_release_irqs(struct ppc440spe_adma_device *adev,
+					struct ppc440spe_adma_chan *chan)
+{
+	u32 mask, disable;
+
+	if (adev->id == PPC440SPE_XOR_ID) {
+		/* disable XOR engine interrupts */
+		mask = ioread32be(&adev->xor_reg->ier);
+		mask &= ~(XOR_IE_CBCIE_BIT | XOR_IE_ICBIE_BIT |
+			  XOR_IE_ICIE_BIT | XOR_IE_RPTIE_BIT);
+		iowrite32be(mask, &adev->xor_reg->ier);
+	} else {
+		/* disable DMAx engine interrupts */
+		disable = (adev->id == PPC440SPE_DMA0_ID) ?
+			  (I2O_IOPIM_P0SNE | I2O_IOPIM_P0EM) :
+			  (I2O_IOPIM_P1SNE | I2O_IOPIM_P1EM);
+		mask = ioread32(&adev->i2o_reg->iopim) | disable;
+		iowrite32(mask, &adev->i2o_reg->iopim);
+	}
+	free_irq(adev->irq, chan);
+	irq_dispose_mapping(adev->irq);
+	if (adev->err_irq > 0) {
+		free_irq(adev->err_irq, chan);
+		if (atomic_dec_and_test(&ppc440spe_adma_err_irq_ref)) {
+			irq_dispose_mapping(adev->err_irq);
+			iounmap(adev->i2o_reg);
+		}
+	}
+}
+
+/**
+ * ppc440spe_adma_probe - probe the asynch device
+ */
+static int __devinit ppc440spe_adma_probe(struct of_device *ofdev,
+					  const struct of_device_id *match)
+{
+	struct device_node *np = ofdev->node;
+	struct resource res;
+	struct ppc440spe_adma_device *adev;
+	struct ppc440spe_adma_chan *chan;
+	struct ppc_dma_chan_ref *ref, *_ref;
+	int ret = 0, initcode = PPC_ADMA_INIT_OK;
+	const u32 *idx;
+	int len;
+	void *regs;
+	u32 id, pool_size;
+
+	if (of_device_is_compatible(np, "amcc,xor-accelerator")) {
+		id = PPC440SPE_XOR_ID;
+		/* As far as the XOR engine is concerned, it does not
+		 * use FIFOs but uses linked list. So there is no dependency
+		 * between pool size to allocate and the engine configuration.
+		 */
+		pool_size = PAGE_SIZE << 1;
+	} else {
+		/* it is DMA0 or DMA1 */
+		idx = of_get_property(np, "cell-index", &len);
+		if (!idx || (len != sizeof(u32))) {
+			dev_err(&ofdev->dev, "Device node %s has missing "
+				"or invalid cell-index property\n",
+				np->full_name);
+			return -EINVAL;
+		}
+		id = *idx;
+		/* DMA0,1 engines use FIFO to maintain CDBs, so we
+		 * should allocate the pool accordingly to size of this
+		 * FIFO. Thus, the pool size depends on the FIFO depth:
+		 * how much CDBs pointers the FIFO may contain then so
+		 * much CDBs we should provide in the pool.
+		 * That is
+		 *   CDB size = 32B;
+		 *   CDBs number = (DMA0_FIFO_SIZE >> 3);
+		 *   Pool size = CDBs number * CDB size =
+		 *      = (DMA0_FIFO_SIZE >> 3) << 5 = DMA0_FIFO_SIZE << 2.
+		 */
+		pool_size = (id == PPC440SPE_DMA0_ID) ?
+			    DMA0_FIFO_SIZE : DMA1_FIFO_SIZE;
+		pool_size <<= 2;
+	}
+
+	if (of_address_to_resource(np, 0, &res)) {
+		dev_err(&ofdev->dev, "failed to get memory resource\n");
+		initcode = PPC_ADMA_INIT_MEMRES;
+		ret = -ENODEV;
+		goto out;
+	}
+
+	if (!request_mem_region(res.start, resource_size(&res),
+				dev_driver_string(&ofdev->dev))) {
+		dev_err(&ofdev->dev, "failed to request memory region "
+			"(0x%016llx-0x%016llx)\n",
+			(u64)res.start, (u64)res.end);
+		initcode = PPC_ADMA_INIT_MEMREG;
+		ret = -EBUSY;
+		goto out;
+	}
+
+	/* create a device */
+	adev = kzalloc(sizeof(*adev), GFP_KERNEL);
+	if (!adev) {
+		dev_err(&ofdev->dev, "failed to allocate device\n");
+		initcode = PPC_ADMA_INIT_ALLOC;
+		ret = -ENOMEM;
+		goto err_adev_alloc;
+	}
+
+	adev->id = id;
+	adev->pool_size = pool_size;
+	/* allocate coherent memory for hardware descriptors */
+	adev->dma_desc_pool_virt = dma_alloc_coherent(&ofdev->dev,
+					adev->pool_size, &adev->dma_desc_pool,
+					GFP_KERNEL);
+	if (adev->dma_desc_pool_virt == NULL) {
+		dev_err(&ofdev->dev, "failed to allocate %d bytes of coherent "
+			"memory for hardware descriptors\n",
+			adev->pool_size);
+		initcode = PPC_ADMA_INIT_COHERENT;
+		ret = -ENOMEM;
+		goto err_dma_alloc;
+	}
+	dev_dbg(&ofdev->dev, "allocted descriptor pool virt 0x%p phys 0x%llx\n",
+		adev->dma_desc_pool_virt, (u64)adev->dma_desc_pool);
+
+	regs = ioremap(res.start, resource_size(&res));
+	if (!regs) {
+		dev_err(&ofdev->dev, "failed to ioremap regs!\n");
+		goto err_regs_alloc;
+	}
+
+	if (adev->id == PPC440SPE_XOR_ID) {
+		adev->xor_reg = regs;
+		/* Reset XOR */
+		iowrite32be(XOR_CRSR_XASR_BIT, &adev->xor_reg->crsr);
+		iowrite32be(XOR_CRSR_64BA_BIT, &adev->xor_reg->crrr);
+	} else {
+		size_t fifo_size = (adev->id == PPC440SPE_DMA0_ID) ?
+				   DMA0_FIFO_SIZE : DMA1_FIFO_SIZE;
+		adev->dma_reg = regs;
+		/* DMAx_FIFO_SIZE is defined in bytes,
+		 * <fsiz> - is defined in number of CDB pointers (8byte).
+		 * DMA FIFO Length = CSlength + CPlength, where
+		 * CSlength = CPlength = (fsiz + 1) * 8.
+		 */
+		iowrite32(DMA_FIFO_ENABLE | ((fifo_size >> 3) - 2),
+			  &adev->dma_reg->fsiz);
+		/* Configure DMA engine */
+		iowrite32(DMA_CFG_DXEPR_HP | DMA_CFG_DFMPP_HP | DMA_CFG_FALGN,
+			  &adev->dma_reg->cfg);
+		/* Clear Status */
+		iowrite32(~0, &adev->dma_reg->dsts);
+	}
+
+	adev->dev = &ofdev->dev;
+	adev->common.dev = &ofdev->dev;
+	INIT_LIST_HEAD(&adev->common.channels);
+	dev_set_drvdata(&ofdev->dev, adev);
+
+	/* create a channel */
+	chan = kzalloc(sizeof(*chan), GFP_KERNEL);
+	if (!chan) {
+		dev_err(&ofdev->dev, "can't allocate channel structure\n");
+		initcode = PPC_ADMA_INIT_CHANNEL;
+		ret = -ENOMEM;
+		goto err_chan_alloc;
+	}
+
+	spin_lock_init(&chan->lock);
+	INIT_LIST_HEAD(&chan->chain);
+	INIT_LIST_HEAD(&chan->all_slots);
+	chan->device = adev;
+	chan->common.device = &adev->common;
+	list_add_tail(&chan->common.device_node, &adev->common.channels);
+	tasklet_init(&chan->irq_tasklet, ppc440spe_adma_tasklet,
+		     (unsigned long)chan);
+
+	/* allocate and map helper pages for async validation or
+	 * async_mult/async_sum_product operations on DMA0/1.
+	 */
+	if (adev->id != PPC440SPE_XOR_ID) {
+		chan->pdest_page = alloc_page(GFP_KERNEL);
+		chan->qdest_page = alloc_page(GFP_KERNEL);
+		if (!chan->pdest_page ||
+		    !chan->qdest_page) {
+			if (chan->pdest_page)
+				__free_page(chan->pdest_page);
+			if (chan->qdest_page)
+				__free_page(chan->qdest_page);
+			ret = -ENOMEM;
+			goto err_page_alloc;
+		}
+		chan->pdest = dma_map_page(&ofdev->dev, chan->pdest_page, 0,
+					   PAGE_SIZE, DMA_BIDIRECTIONAL);
+		chan->qdest = dma_map_page(&ofdev->dev, chan->qdest_page, 0,
+					   PAGE_SIZE, DMA_BIDIRECTIONAL);
+	}
+
+	ref = kmalloc(sizeof(*ref), GFP_KERNEL);
+	if (ref) {
+		ref->chan = &chan->common;
+		INIT_LIST_HEAD(&ref->node);
+		list_add_tail(&ref->node, &ppc440spe_adma_chan_list);
+	} else {
+		dev_err(&ofdev->dev, "failed to allocate channel reference!\n");
+		ret = -ENOMEM;
+		goto err_ref_alloc;
+	}
+
+	ret = ppc440spe_adma_setup_irqs(adev, chan, &initcode);
+	if (ret)
+		goto err_irq;
+
+	ppc440spe_adma_init_capabilities(adev);
+
+	ret = dma_async_device_register(&adev->common);
+	if (ret) {
+		initcode = PPC_ADMA_INIT_REGISTER;
+		dev_err(&ofdev->dev, "failed to register dma device\n");
+		goto err_dev_reg;
+	}
+
+	goto out;
+
+err_dev_reg:
+	ppc440spe_adma_release_irqs(adev, chan);
+err_irq:
+	list_for_each_entry_safe(ref, _ref, &ppc440spe_adma_chan_list, node) {
+		if (chan == to_ppc440spe_adma_chan(ref->chan)) {
+			list_del(&ref->node);
+			kfree(ref);
+		}
+	}
+err_ref_alloc:
+	if (adev->id != PPC440SPE_XOR_ID) {
+		dma_unmap_page(&ofdev->dev, chan->pdest,
+			       PAGE_SIZE, DMA_BIDIRECTIONAL);
+		dma_unmap_page(&ofdev->dev, chan->qdest,
+			       PAGE_SIZE, DMA_BIDIRECTIONAL);
+		__free_page(chan->pdest_page);
+		__free_page(chan->qdest_page);
+	}
+err_page_alloc:
+	kfree(chan);
+err_chan_alloc:
+	if (adev->id == PPC440SPE_XOR_ID)
+		iounmap(adev->xor_reg);
+	else
+		iounmap(adev->dma_reg);
+err_regs_alloc:
+	dma_free_coherent(adev->dev, adev->pool_size,
+			  adev->dma_desc_pool_virt,
+			  adev->dma_desc_pool);
+err_dma_alloc:
+	kfree(adev);
+err_adev_alloc:
+	release_mem_region(res.start, resource_size(&res));
+out:
+	if (id < PPC440SPE_ADMA_ENGINES_NUM)
+		ppc440spe_adma_devices[id] = initcode;
+
+	return ret;
+}
+
+/**
+ * ppc440spe_adma_remove - remove the asynch device
+ */
+static int __devexit ppc440spe_adma_remove(struct of_device *ofdev)
+{
+	struct ppc440spe_adma_device *adev = dev_get_drvdata(&ofdev->dev);
+	struct device_node *np = ofdev->node;
+	struct resource res;
+	struct dma_chan *chan, *_chan;
+	struct ppc_dma_chan_ref *ref, *_ref;
+	struct ppc440spe_adma_chan *ppc440spe_chan;
+
+	dev_set_drvdata(&ofdev->dev, NULL);
+	if (adev->id < PPC440SPE_ADMA_ENGINES_NUM)
+		ppc440spe_adma_devices[adev->id] = -1;
+
+	dma_async_device_unregister(&adev->common);
+
+	list_for_each_entry_safe(chan, _chan, &adev->common.channels,
+				 device_node) {
+		ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+		ppc440spe_adma_release_irqs(adev, ppc440spe_chan);
+		tasklet_kill(&ppc440spe_chan->irq_tasklet);
+		if (adev->id != PPC440SPE_XOR_ID) {
+			dma_unmap_page(&ofdev->dev, ppc440spe_chan->pdest,
+					PAGE_SIZE, DMA_BIDIRECTIONAL);
+			dma_unmap_page(&ofdev->dev, ppc440spe_chan->qdest,
+					PAGE_SIZE, DMA_BIDIRECTIONAL);
+			__free_page(ppc440spe_chan->pdest_page);
+			__free_page(ppc440spe_chan->qdest_page);
+		}
+		list_for_each_entry_safe(ref, _ref, &ppc440spe_adma_chan_list,
+					 node) {
+			if (ppc440spe_chan ==
+			    to_ppc440spe_adma_chan(ref->chan)) {
+				list_del(&ref->node);
+				kfree(ref);
+			}
+		}
+		list_del(&chan->device_node);
+		kfree(ppc440spe_chan);
+	}
+
+	dma_free_coherent(adev->dev, adev->pool_size,
+			  adev->dma_desc_pool_virt, adev->dma_desc_pool);
+	if (adev->id == PPC440SPE_XOR_ID)
+		iounmap(adev->xor_reg);
+	else
+		iounmap(adev->dma_reg);
+	of_address_to_resource(np, 0, &res);
+	release_mem_region(res.start, resource_size(&res));
+	kfree(adev);
+	return 0;
+}
+
+/*
+ * /sys driver interface to enable h/w RAID-6 capabilities
+ * Files created in e.g. /sys/devices/plb.0/400100100.dma0/driver/
+ * directory are "devices", "enable" and "poly".
+ * "devices" shows available engines.
+ * "enable" is used to enable RAID-6 capabilities or to check
+ * whether these has been activated.
+ * "poly" allows setting/checking used polynomial (for PPC440SPe only).
+ */
+
+static ssize_t show_ppc440spe_devices(struct device_driver *dev, char *buf)
+{
+	ssize_t size = 0;
+	int i;
+
+	for (i = 0; i < PPC440SPE_ADMA_ENGINES_NUM; i++) {
+		if (ppc440spe_adma_devices[i] == -1)
+			continue;
+		size += snprintf(buf + size, PAGE_SIZE - size,
+				 "PPC440SP(E)-ADMA.%d: %s\n", i,
+				 ppc_adma_errors[ppc440spe_adma_devices[i]]);
+	}
+	return size;
+}
+
+static ssize_t show_ppc440spe_r6enable(struct device_driver *dev, char *buf)
+{
+	return snprintf(buf, PAGE_SIZE,
+			"PPC440SP(e) RAID-6 capabilities are %sABLED.\n",
+			ppc440spe_r6_enabled ? "EN" : "DIS");
+}
+
+static ssize_t store_ppc440spe_r6enable(struct device_driver *dev,
+					const char *buf, size_t count)
+{
+	unsigned long val;
+
+	if (!count || count > 11)
+		return -EINVAL;
+
+	if (!ppc440spe_r6_tchan)
+		return -EFAULT;
+
+	/* Write a key */
+	sscanf(buf, "%lx", &val);
+	dcr_write(ppc440spe_mq_dcr_host, DCRN_MQ0_XORBA, val);
+	isync();
+
+	/* Verify whether it really works now */
+	if (ppc440spe_test_raid6(ppc440spe_r6_tchan) == 0) {
+		pr_info("PPC440SP(e) RAID-6 has been activated "
+			"successfully\n");
+		ppc440spe_r6_enabled = 1;
+	} else {
+		pr_info("PPC440SP(e) RAID-6 hasn't been activated!"
+			" Error key ?\n");
+		ppc440spe_r6_enabled = 0;
+	}
+	return count;
+}
+
+static ssize_t show_ppc440spe_r6poly(struct device_driver *dev, char *buf)
+{
+	ssize_t size = 0;
+	u32 reg;
+
+#ifdef CONFIG_440SP
+	/* 440SP has fixed polynomial */
+	reg = 0x4d;
+#else
+	reg = dcr_read(ppc440spe_mq_dcr_host, DCRN_MQ0_CFBHL);
+	reg >>= MQ0_CFBHL_POLY;
+	reg &= 0xFF;
+#endif
+
+	size = snprintf(buf, PAGE_SIZE, "PPC440SP(e) RAID-6 driver "
+			"uses 0x1%02x polynomial.\n", reg);
+	return size;
+}
+
+static ssize_t store_ppc440spe_r6poly(struct device_driver *dev,
+				      const char *buf, size_t count)
+{
+	unsigned long reg, val;
+
+#ifdef CONFIG_440SP
+	/* 440SP uses default 0x14D polynomial only */
+	return -EINVAL;
+#endif
+
+	if (!count || count > 6)
+		return -EINVAL;
+
+	/* e.g., 0x14D or 0x11D */
+	sscanf(buf, "%lx", &val);
+
+	if (val & ~0x1FF)
+		return -EINVAL;
+
+	val &= 0xFF;
+	reg = dcr_read(ppc440spe_mq_dcr_host, DCRN_MQ0_CFBHL);
+	reg &= ~(0xFF << MQ0_CFBHL_POLY);
+	reg |= val << MQ0_CFBHL_POLY;
+	dcr_write(ppc440spe_mq_dcr_host, DCRN_MQ0_CFBHL, reg);
+
+	return count;
+}
+
+static DRIVER_ATTR(devices, S_IRUGO, show_ppc440spe_devices, NULL);
+static DRIVER_ATTR(enable, S_IRUGO | S_IWUSR, show_ppc440spe_r6enable,
+		   store_ppc440spe_r6enable);
+static DRIVER_ATTR(poly, S_IRUGO | S_IWUSR, show_ppc440spe_r6poly,
+		   store_ppc440spe_r6poly);
+
+/*
+ * Common initialisation for RAID engines; allocate memory for
+ * DMAx FIFOs, perform configuration common for all DMA engines.
+ * Further DMA engine specific configuration is done at probe time.
+ */
+static int ppc440spe_configure_raid_devices(void)
+{
+	struct device_node *np;
+	struct resource i2o_res;
+	struct i2o_regs __iomem *i2o_reg;
+	dcr_host_t i2o_dcr_host;
+	unsigned int dcr_base, dcr_len;
+	int i, ret;
+
+	np = of_find_compatible_node(NULL, NULL, "ibm,i2o-440spe");
+	if (!np) {
+		pr_err("%s: can't find I2O device tree node\n",
+			__func__);
+		return -ENODEV;
+	}
+
+	if (of_address_to_resource(np, 0, &i2o_res)) {
+		of_node_put(np);
+		return -EINVAL;
+	}
+
+	i2o_reg = of_iomap(np, 0);
+	if (!i2o_reg) {
+		pr_err("%s: failed to map I2O registers\n", __func__);
+		of_node_put(np);
+		return -EINVAL;
+	}
+
+	/* Get I2O DCRs base */
+	dcr_base = dcr_resource_start(np, 0);
+	dcr_len = dcr_resource_len(np, 0);
+	if (!dcr_base && !dcr_len) {
+		pr_err("%s: can't get DCR registers base/len!\n",
+			np->full_name);
+		of_node_put(np);
+		iounmap(i2o_reg);
+		return -ENODEV;
+	}
+
+	i2o_dcr_host = dcr_map(np, dcr_base, dcr_len);
+	if (!DCR_MAP_OK(i2o_dcr_host)) {
+		pr_err("%s: failed to map DCRs!\n", np->full_name);
+		of_node_put(np);
+		iounmap(i2o_reg);
+		return -ENODEV;
+	}
+	of_node_put(np);
+
+	/* Provide memory regions for DMA's FIFOs: I2O, DMA0 and DMA1 share
+	 * the base address of FIFO memory space.
+	 * Actually we need twice more physical memory than programmed in the
+	 * <fsiz> register (because there are two FIFOs for each DMA: CP and CS)
+	 */
+	ppc440spe_dma_fifo_buf = kmalloc((DMA0_FIFO_SIZE + DMA1_FIFO_SIZE) << 1,
+					 GFP_KERNEL);
+	if (!ppc440spe_dma_fifo_buf) {
+		pr_err("%s: DMA FIFO buffer allocation failed.\n", __func__);
+		iounmap(i2o_reg);
+		dcr_unmap(i2o_dcr_host, dcr_len);
+		return -ENOMEM;
+	}
+
+	/*
+	 * Configure h/w
+	 */
+	/* Reset I2O/DMA */
+	mtdcri(SDR0, DCRN_SDR0_SRST, DCRN_SDR0_SRST_I2ODMA);
+	mtdcri(SDR0, DCRN_SDR0_SRST, 0);
+
+	/* Setup the base address of mmaped registers */
+	dcr_write(i2o_dcr_host, DCRN_I2O0_IBAH, (u32)(i2o_res.start >> 32));
+	dcr_write(i2o_dcr_host, DCRN_I2O0_IBAL, (u32)(i2o_res.start) |
+						I2O_REG_ENABLE);
+	dcr_unmap(i2o_dcr_host, dcr_len);
+
+	/* Setup FIFO memory space base address */
+	iowrite32(0, &i2o_reg->ifbah);
+	iowrite32(((u32)__pa(ppc440spe_dma_fifo_buf)), &i2o_reg->ifbal);
+
+	/* set zero FIFO size for I2O, so the whole
+	 * ppc440spe_dma_fifo_buf is used by DMAs.
+	 * DMAx_FIFOs will be configured while probe.
+	 */
+	iowrite32(0, &i2o_reg->ifsiz);
+	iounmap(i2o_reg);
+
+	/* To prepare WXOR/RXOR functionality we need access to
+	 * Memory Queue Module DCRs (finally it will be enabled
+	 * via /sys interface of the ppc440spe ADMA driver).
+	 */
+	np = of_find_compatible_node(NULL, NULL, "ibm,mq-440spe");
+	if (!np) {
+		pr_err("%s: can't find MQ device tree node\n",
+			__func__);
+		ret = -ENODEV;
+		goto out_free;
+	}
+
+	/* Get MQ DCRs base */
+	dcr_base = dcr_resource_start(np, 0);
+	dcr_len = dcr_resource_len(np, 0);
+	if (!dcr_base && !dcr_len) {
+		pr_err("%s: can't get DCR registers base/len!\n",
+			np->full_name);
+		ret = -ENODEV;
+		goto out_mq;
+	}
+
+	ppc440spe_mq_dcr_host = dcr_map(np, dcr_base, dcr_len);
+	if (!DCR_MAP_OK(ppc440spe_mq_dcr_host)) {
+		pr_err("%s: failed to map DCRs!\n", np->full_name);
+		ret = -ENODEV;
+		goto out_mq;
+	}
+	of_node_put(np);
+	ppc440spe_mq_dcr_len = dcr_len;
+
+	/* Set HB alias */
+	dcr_write(ppc440spe_mq_dcr_host, DCRN_MQ0_BAUH, DMA_CUED_XOR_HB);
+
+	/* Set:
+	 * - LL transaction passing limit to 1;
+	 * - Memory controller cycle limit to 1;
+	 * - Galois Polynomial to 0x14d (default)
+	 */
+	dcr_write(ppc440spe_mq_dcr_host, DCRN_MQ0_CFBHL,
+		  (1 << MQ0_CFBHL_TPLM) | (1 << MQ0_CFBHL_HBCL) |
+		  (PPC440SPE_DEFAULT_POLY << MQ0_CFBHL_POLY));
+
+	atomic_set(&ppc440spe_adma_err_irq_ref, 0);
+	for (i = 0; i < PPC440SPE_ADMA_ENGINES_NUM; i++)
+		ppc440spe_adma_devices[i] = -1;
+
+	return 0;
+
+out_mq:
+	of_node_put(np);
+out_free:
+	kfree(ppc440spe_dma_fifo_buf);
+	return ret;
+}
+
+static struct of_device_id __devinitdata ppc440spe_adma_of_match[] = {
+	{ .compatible	= "ibm,dma-440spe", },
+	{ .compatible	= "amcc,xor-accelerator", },
+	{},
+};
+MODULE_DEVICE_TABLE(of, ppc440spe_adma_of_match);
+
+static struct of_platform_driver ppc440spe_adma_driver = {
+	.match_table = ppc440spe_adma_of_match,
+	.probe = ppc440spe_adma_probe,
+	.remove = __devexit_p(ppc440spe_adma_remove),
+	.driver = {
+		.name = "PPC440SP(E)-ADMA",
+		.owner = THIS_MODULE,
+	},
+};
+
+static __init int ppc440spe_adma_init(void)
+{
+	int ret;
+
+	ret = ppc440spe_configure_raid_devices();
+	if (ret)
+		return ret;
+
+	ret = of_register_platform_driver(&ppc440spe_adma_driver);
+	if (ret) {
+		pr_err("%s: failed to register platform driver\n",
+			__func__);
+		goto out_reg;
+	}
+
+	/* Initialization status */
+	ret = driver_create_file(&ppc440spe_adma_driver.driver,
+				 &driver_attr_devices);
+	if (ret)
+		goto out_dev;
+
+	/* RAID-6 h/w enable entry */
+	ret = driver_create_file(&ppc440spe_adma_driver.driver,
+				 &driver_attr_enable);
+	if (ret)
+		goto out_en;
+
+	/* GF polynomial to use */
+	ret = driver_create_file(&ppc440spe_adma_driver.driver,
+				 &driver_attr_poly);
+	if (!ret)
+		return ret;
+
+	driver_remove_file(&ppc440spe_adma_driver.driver,
+			   &driver_attr_enable);
+out_en:
+	driver_remove_file(&ppc440spe_adma_driver.driver,
+			   &driver_attr_devices);
+out_dev:
+	/* User will not be able to enable h/w RAID-6 */
+	pr_err("%s: failed to create RAID-6 driver interface\n",
+		__func__);
+	of_unregister_platform_driver(&ppc440spe_adma_driver);
+out_reg:
+	dcr_unmap(ppc440spe_mq_dcr_host, ppc440spe_mq_dcr_len);
+	kfree(ppc440spe_dma_fifo_buf);
+	return ret;
+}
+
+static void __exit ppc440spe_adma_exit(void)
+{
+	driver_remove_file(&ppc440spe_adma_driver.driver,
+			   &driver_attr_poly);
+	driver_remove_file(&ppc440spe_adma_driver.driver,
+			   &driver_attr_enable);
+	driver_remove_file(&ppc440spe_adma_driver.driver,
+			   &driver_attr_devices);
+	of_unregister_platform_driver(&ppc440spe_adma_driver);
+	dcr_unmap(ppc440spe_mq_dcr_host, ppc440spe_mq_dcr_len);
+	kfree(ppc440spe_dma_fifo_buf);
+}
+
+arch_initcall(ppc440spe_adma_init);
+module_exit(ppc440spe_adma_exit);
+
+MODULE_AUTHOR("Yuri Tikhonov <yur@emcraft.com>");
+MODULE_DESCRIPTION("PPC440SPE ADMA Engine Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/dma/ppc4xx/adma.h b/drivers/dma/ppc4xx/adma.h
new file mode 100644
index 000000000000..8ada5a812e3b
--- /dev/null
+++ b/drivers/dma/ppc4xx/adma.h
@@ -0,0 +1,195 @@
+/*
+ * 2006-2009 (C) DENX Software Engineering.
+ *
+ * Author: Yuri Tikhonov <yur@emcraft.com>
+ *
+ * This file is licensed under the terms of the GNU General Public License
+ * version 2.  This program is licensed "as is" without any warranty of
+ * any kind, whether express or implied.
+ */
+
+#ifndef _PPC440SPE_ADMA_H
+#define _PPC440SPE_ADMA_H
+
+#include <linux/types.h>
+#include "dma.h"
+#include "xor.h"
+
+#define to_ppc440spe_adma_chan(chan) \
+		container_of(chan, struct ppc440spe_adma_chan, common)
+#define to_ppc440spe_adma_device(dev) \
+		container_of(dev, struct ppc440spe_adma_device, common)
+#define tx_to_ppc440spe_adma_slot(tx) \
+		container_of(tx, struct ppc440spe_adma_desc_slot, async_tx)
+
+/* Default polynomial (for 440SP is only available) */
+#define PPC440SPE_DEFAULT_POLY	0x4d
+
+#define PPC440SPE_ADMA_ENGINES_NUM	(XOR_ENGINES_NUM + DMA_ENGINES_NUM)
+
+#define PPC440SPE_ADMA_WATCHDOG_MSEC	3
+#define PPC440SPE_ADMA_THRESHOLD	1
+
+#define PPC440SPE_DMA0_ID	0
+#define PPC440SPE_DMA1_ID	1
+#define PPC440SPE_XOR_ID	2
+
+#define PPC440SPE_ADMA_DMA_MAX_BYTE_COUNT	0xFFFFFFUL
+/* this is the XOR_CBBCR width */
+#define PPC440SPE_ADMA_XOR_MAX_BYTE_COUNT	(1 << 31)
+#define PPC440SPE_ADMA_ZERO_SUM_MAX_BYTE_COUNT PPC440SPE_ADMA_XOR_MAX_BYTE_COUNT
+
+#define PPC440SPE_RXOR_RUN	0
+
+#define MQ0_CF2H_RXOR_BS_MASK	0x1FF
+
+#undef ADMA_LL_DEBUG
+
+/**
+ * struct ppc440spe_adma_device - internal representation of an ADMA device
+ * @dev: device
+ * @dma_reg: base for DMAx register access
+ * @xor_reg: base for XOR register access
+ * @i2o_reg: base for I2O register access
+ * @id: HW ADMA Device selector
+ * @dma_desc_pool_virt: base of DMA descriptor region (CPU address)
+ * @dma_desc_pool: base of DMA descriptor region (DMA address)
+ * @pool_size: size of the pool
+ * @irq: DMAx or XOR irq number
+ * @err_irq: DMAx error irq number
+ * @common: embedded struct dma_device
+ */
+struct ppc440spe_adma_device {
+	struct device *dev;
+	struct dma_regs __iomem *dma_reg;
+	struct xor_regs __iomem *xor_reg;
+	struct i2o_regs __iomem *i2o_reg;
+	int id;
+	void *dma_desc_pool_virt;
+	dma_addr_t dma_desc_pool;
+	size_t pool_size;
+	int irq;
+	int err_irq;
+	struct dma_device common;
+};
+
+/**
+ * struct ppc440spe_adma_chan - internal representation of an ADMA channel
+ * @lock: serializes enqueue/dequeue operations to the slot pool
+ * @device: parent device
+ * @chain: device chain view of the descriptors
+ * @common: common dmaengine channel object members
+ * @all_slots: complete domain of slots usable by the channel
+ * @pending: allows batching of hardware operations
+ * @completed_cookie: identifier for the most recently completed operation
+ * @slots_allocated: records the actual size of the descriptor slot pool
+ * @hw_chain_inited: h/w descriptor chain initialization flag
+ * @irq_tasklet: bottom half where ppc440spe_adma_slot_cleanup runs
+ * @needs_unmap: if buffers should not be unmapped upon final processing
+ * @pdest_page: P destination page for async validate operation
+ * @qdest_page: Q destination page for async validate operation
+ * @pdest: P dma addr for async validate operation
+ * @qdest: Q dma addr for async validate operation
+ */
+struct ppc440spe_adma_chan {
+	spinlock_t lock;
+	struct ppc440spe_adma_device *device;
+	struct list_head chain;
+	struct dma_chan common;
+	struct list_head all_slots;
+	struct ppc440spe_adma_desc_slot *last_used;
+	int pending;
+	dma_cookie_t completed_cookie;
+	int slots_allocated;
+	int hw_chain_inited;
+	struct tasklet_struct irq_tasklet;
+	u8 needs_unmap;
+	struct page *pdest_page;
+	struct page *qdest_page;
+	dma_addr_t pdest;
+	dma_addr_t qdest;
+};
+
+struct ppc440spe_rxor {
+	u32 addrl;
+	u32 addrh;
+	int len;
+	int xor_count;
+	int addr_count;
+	int desc_count;
+	int state;
+};
+
+/**
+ * struct ppc440spe_adma_desc_slot - PPC440SPE-ADMA software descriptor
+ * @phys: hardware address of the hardware descriptor chain
+ * @group_head: first operation in a transaction
+ * @hw_next: pointer to the next descriptor in chain
+ * @async_tx: support for the async_tx api
+ * @slot_node: node on the iop_adma_chan.all_slots list
+ * @chain_node: node on the op_adma_chan.chain list
+ * @group_list: list of slots that make up a multi-descriptor transaction
+ *              for example transfer lengths larger than the supported hw max
+ * @unmap_len: transaction bytecount
+ * @hw_desc: virtual address of the hardware descriptor chain
+ * @stride: currently chained or not
+ * @idx: pool index
+ * @slot_cnt: total slots used in an transaction (group of operations)
+ * @src_cnt: number of sources set in this descriptor
+ * @dst_cnt: number of destinations set in the descriptor
+ * @slots_per_op: number of slots per operation
+ * @descs_per_op: number of slot per P/Q operation see comment
+ *                for ppc440spe_prep_dma_pqxor function
+ * @flags: desc state/type
+ * @reverse_flags: 1 if a corresponding rxor address uses reversed address order
+ * @xor_check_result: result of zero sum
+ * @crc32_result: result crc calculation
+ */
+struct ppc440spe_adma_desc_slot {
+	dma_addr_t phys;
+	struct ppc440spe_adma_desc_slot *group_head;
+	struct ppc440spe_adma_desc_slot *hw_next;
+	struct dma_async_tx_descriptor async_tx;
+	struct list_head slot_node;
+	struct list_head chain_node; /* node in channel ops list */
+	struct list_head group_list; /* list */
+	unsigned int unmap_len;
+	void *hw_desc;
+	u16 stride;
+	u16 idx;
+	u16 slot_cnt;
+	u8 src_cnt;
+	u8 dst_cnt;
+	u8 slots_per_op;
+	u8 descs_per_op;
+	unsigned long flags;
+	unsigned long reverse_flags[8];
+
+#define PPC440SPE_DESC_INT	0	/* generate interrupt on complete */
+#define PPC440SPE_ZERO_P	1	/* clear P destionaion */
+#define PPC440SPE_ZERO_Q	2	/* clear Q destination */
+#define PPC440SPE_COHERENT	3	/* src/dst are coherent */
+
+#define PPC440SPE_DESC_WXOR	4	/* WXORs are in chain */
+#define PPC440SPE_DESC_RXOR	5	/* RXOR is in chain */
+
+#define PPC440SPE_DESC_RXOR123	8	/* CDB for RXOR123 operation */
+#define PPC440SPE_DESC_RXOR124	9	/* CDB for RXOR124 operation */
+#define PPC440SPE_DESC_RXOR125	10	/* CDB for RXOR125 operation */
+#define PPC440SPE_DESC_RXOR12	11	/* CDB for RXOR12 operation */
+#define PPC440SPE_DESC_RXOR_REV	12	/* CDB has srcs in reversed order */
+
+#define PPC440SPE_DESC_PCHECK	13
+#define PPC440SPE_DESC_QCHECK	14
+
+#define PPC440SPE_DESC_RXOR_MSK	0x3
+
+	struct ppc440spe_rxor rxor_cursor;
+
+	union {
+		u32 *xor_check_result;
+		u32 *crc32_result;
+	};
+};
+
+#endif /* _PPC440SPE_ADMA_H */
diff --git a/drivers/dma/ppc4xx/dma.h b/drivers/dma/ppc4xx/dma.h
new file mode 100644
index 000000000000..bcde2df2f373
--- /dev/null
+++ b/drivers/dma/ppc4xx/dma.h
@@ -0,0 +1,223 @@
+/*
+ * 440SPe's DMA engines support header file
+ *
+ * 2006-2009 (C) DENX Software Engineering.
+ *
+ * Author: Yuri Tikhonov <yur@emcraft.com>
+ *
+ * This file is licensed under the term of  the GNU General Public License
+ * version 2. The program licensed "as is" without any warranty of any
+ * kind, whether express or implied.
+ */
+
+#ifndef	_PPC440SPE_DMA_H
+#define _PPC440SPE_DMA_H
+
+#include <linux/types.h>
+
+/* Number of elements in the array with statical CDBs */
+#define	MAX_STAT_DMA_CDBS	16
+/* Number of DMA engines available on the contoller */
+#define DMA_ENGINES_NUM		2
+
+/* Maximum h/w supported number of destinations */
+#define DMA_DEST_MAX_NUM	2
+
+/* FIFO's params */
+#define DMA0_FIFO_SIZE		0x1000
+#define DMA1_FIFO_SIZE		0x1000
+#define DMA_FIFO_ENABLE		(1<<12)
+
+/* DMA Configuration Register. Data Transfer Engine PLB Priority: */
+#define DMA_CFG_DXEPR_LP	(0<<26)
+#define DMA_CFG_DXEPR_HP	(3<<26)
+#define DMA_CFG_DXEPR_HHP	(2<<26)
+#define DMA_CFG_DXEPR_HHHP	(1<<26)
+
+/* DMA Configuration Register. DMA FIFO Manager PLB Priority: */
+#define DMA_CFG_DFMPP_LP	(0<<23)
+#define DMA_CFG_DFMPP_HP	(3<<23)
+#define DMA_CFG_DFMPP_HHP	(2<<23)
+#define DMA_CFG_DFMPP_HHHP	(1<<23)
+
+/* DMA Configuration Register. Force 64-byte Alignment */
+#define DMA_CFG_FALGN		(1 << 19)
+
+/*UIC0:*/
+#define D0CPF_INT		(1<<12)
+#define D0CSF_INT		(1<<11)
+#define D1CPF_INT		(1<<10)
+#define D1CSF_INT		(1<<9)
+/*UIC1:*/
+#define DMAE_INT		(1<<9)
+
+/* I2O IOP Interrupt Mask Register */
+#define I2O_IOPIM_P0SNE		(1<<3)
+#define I2O_IOPIM_P0EM		(1<<5)
+#define I2O_IOPIM_P1SNE		(1<<6)
+#define I2O_IOPIM_P1EM		(1<<8)
+
+/* DMA CDB fields */
+#define DMA_CDB_MSK		(0xF)
+#define DMA_CDB_64B_ADDR	(1<<2)
+#define DMA_CDB_NO_INT		(1<<3)
+#define DMA_CDB_STATUS_MSK	(0x3)
+#define DMA_CDB_ADDR_MSK	(0xFFFFFFF0)
+
+/* DMA CDB OpCodes */
+#define DMA_CDB_OPC_NO_OP	(0x00)
+#define DMA_CDB_OPC_MV_SG1_SG2	(0x01)
+#define DMA_CDB_OPC_MULTICAST	(0x05)
+#define DMA_CDB_OPC_DFILL128	(0x24)
+#define DMA_CDB_OPC_DCHECK128	(0x23)
+
+#define DMA_CUED_XOR_BASE	(0x10000000)
+#define DMA_CUED_XOR_HB		(0x00000008)
+
+#ifdef CONFIG_440SP
+#define DMA_CUED_MULT1_OFF	0
+#define DMA_CUED_MULT2_OFF	8
+#define DMA_CUED_MULT3_OFF	16
+#define DMA_CUED_REGION_OFF	24
+#define DMA_CUED_XOR_WIN_MSK	(0xFC000000)
+#else
+#define DMA_CUED_MULT1_OFF	2
+#define DMA_CUED_MULT2_OFF	10
+#define DMA_CUED_MULT3_OFF	18
+#define DMA_CUED_REGION_OFF	26
+#define DMA_CUED_XOR_WIN_MSK	(0xF0000000)
+#endif
+
+#define DMA_CUED_REGION_MSK	0x3
+#define DMA_RXOR123		0x0
+#define DMA_RXOR124		0x1
+#define DMA_RXOR125		0x2
+#define DMA_RXOR12		0x3
+
+/* S/G addresses */
+#define DMA_CDB_SG_SRC		1
+#define DMA_CDB_SG_DST1		2
+#define DMA_CDB_SG_DST2		3
+
+/*
+ * DMAx engines Command Descriptor Block Type
+ */
+struct dma_cdb {
+	/*
+	 * Basic CDB structure (Table 20-17, p.499, 440spe_um_1_22.pdf)
+	 */
+	u8	pad0[2];        /* reserved */
+	u8	attr;		/* attributes */
+	u8	opc;		/* opcode */
+	u32	sg1u;		/* upper SG1 address */
+	u32	sg1l;		/* lower SG1 address */
+	u32	cnt;		/* SG count, 3B used */
+	u32	sg2u;		/* upper SG2 address */
+	u32	sg2l;		/* lower SG2 address */
+	u32	sg3u;		/* upper SG3 address */
+	u32	sg3l;		/* lower SG3 address */
+};
+
+/*
+ * DMAx hardware registers (p.515 in 440SPe UM 1.22)
+ */
+struct dma_regs {
+	u32	cpfpl;
+	u32	cpfph;
+	u32	csfpl;
+	u32	csfph;
+	u32	dsts;
+	u32	cfg;
+	u8	pad0[0x8];
+	u16	cpfhp;
+	u16	cpftp;
+	u16	csfhp;
+	u16	csftp;
+	u8	pad1[0x8];
+	u32	acpl;
+	u32	acph;
+	u32	s1bpl;
+	u32	s1bph;
+	u32	s2bpl;
+	u32	s2bph;
+	u32	s3bpl;
+	u32	s3bph;
+	u8	pad2[0x10];
+	u32	earl;
+	u32	earh;
+	u8	pad3[0x8];
+	u32	seat;
+	u32	sead;
+	u32	op;
+	u32	fsiz;
+};
+
+/*
+ * I2O hardware registers (p.528 in 440SPe UM 1.22)
+ */
+struct i2o_regs {
+	u32	ists;
+	u32	iseat;
+	u32	isead;
+	u8	pad0[0x14];
+	u32	idbel;
+	u8	pad1[0xc];
+	u32	ihis;
+	u32	ihim;
+	u8	pad2[0x8];
+	u32	ihiq;
+	u32	ihoq;
+	u8	pad3[0x8];
+	u32	iopis;
+	u32	iopim;
+	u32	iopiq;
+	u8	iopoq;
+	u8	pad4[3];
+	u16	iiflh;
+	u16	iiflt;
+	u16	iiplh;
+	u16	iiplt;
+	u16	ioflh;
+	u16	ioflt;
+	u16	ioplh;
+	u16	ioplt;
+	u32	iidc;
+	u32	ictl;
+	u32	ifcpp;
+	u8	pad5[0x4];
+	u16	mfac0;
+	u16	mfac1;
+	u16	mfac2;
+	u16	mfac3;
+	u16	mfac4;
+	u16	mfac5;
+	u16	mfac6;
+	u16	mfac7;
+	u16	ifcfh;
+	u16	ifcht;
+	u8	pad6[0x4];
+	u32	iifmc;
+	u32	iodb;
+	u32	iodbc;
+	u32	ifbal;
+	u32	ifbah;
+	u32	ifsiz;
+	u32	ispd0;
+	u32	ispd1;
+	u32	ispd2;
+	u32	ispd3;
+	u32	ihipl;
+	u32	ihiph;
+	u32	ihopl;
+	u32	ihoph;
+	u32	iiipl;
+	u32	iiiph;
+	u32	iiopl;
+	u32	iioph;
+	u32	ifcpl;
+	u32	ifcph;
+	u8	pad7[0x8];
+	u32	iopt;
+};
+
+#endif /* _PPC440SPE_DMA_H */
diff --git a/drivers/dma/ppc4xx/xor.h b/drivers/dma/ppc4xx/xor.h
new file mode 100644
index 000000000000..daed7384daac
--- /dev/null
+++ b/drivers/dma/ppc4xx/xor.h
@@ -0,0 +1,110 @@
+/*
+ * 440SPe's XOR engines support header file
+ *
+ * 2006-2009 (C) DENX Software Engineering.
+ *
+ * Author: Yuri Tikhonov <yur@emcraft.com>
+ *
+ * This file is licensed under the term of  the GNU General Public License
+ * version 2. The program licensed "as is" without any warranty of any
+ * kind, whether express or implied.
+ */
+
+#ifndef _PPC440SPE_XOR_H
+#define _PPC440SPE_XOR_H
+
+#include <linux/types.h>
+
+/* Number of XOR engines available on the contoller */
+#define XOR_ENGINES_NUM		1
+
+/* Number of operands supported in the h/w */
+#define XOR_MAX_OPS		16
+
+/*
+ * XOR Command Block Control Register bits
+ */
+#define XOR_CBCR_LNK_BIT        (1<<31) /* link present */
+#define XOR_CBCR_TGT_BIT        (1<<30) /* target present */
+#define XOR_CBCR_CBCE_BIT       (1<<29) /* command block compete enable */
+#define XOR_CBCR_RNZE_BIT       (1<<28) /* result not zero enable */
+#define XOR_CBCR_XNOR_BIT       (1<<15) /* XOR/XNOR */
+#define XOR_CDCR_OAC_MSK        (0x7F)  /* operand address count */
+
+/*
+ * XORCore Status Register bits
+ */
+#define XOR_SR_XCP_BIT		(1<<31)	/* core processing */
+#define XOR_SR_ICB_BIT		(1<<17)	/* invalid CB */
+#define XOR_SR_IC_BIT		(1<<16)	/* invalid command */
+#define XOR_SR_IPE_BIT		(1<<15)	/* internal parity error */
+#define XOR_SR_RNZ_BIT		(1<<2)	/* result not Zero */
+#define XOR_SR_CBC_BIT		(1<<1)	/* CB complete */
+#define XOR_SR_CBLC_BIT		(1<<0)	/* CB list complete */
+
+/*
+ * XORCore Control Set and Reset Register bits
+ */
+#define XOR_CRSR_XASR_BIT	(1<<31)	/* soft reset */
+#define XOR_CRSR_XAE_BIT	(1<<30)	/* enable */
+#define XOR_CRSR_RCBE_BIT	(1<<29)	/* refetch CB enable */
+#define XOR_CRSR_PAUS_BIT	(1<<28)	/* pause */
+#define XOR_CRSR_64BA_BIT	(1<<27) /* 64/32 CB format */
+#define XOR_CRSR_CLP_BIT	(1<<25)	/* continue list processing */
+
+/*
+ * XORCore Interrupt Enable Register
+ */
+#define XOR_IE_ICBIE_BIT	(1<<17)	/* Invalid Command Block IRQ Enable */
+#define XOR_IE_ICIE_BIT		(1<<16)	/* Invalid Command IRQ Enable */
+#define XOR_IE_RPTIE_BIT	(1<<14)	/* Read PLB Timeout Error IRQ Enable */
+#define XOR_IE_CBCIE_BIT	(1<<1)	/* CB complete interrupt enable */
+#define XOR_IE_CBLCI_BIT	(1<<0)	/* CB list complete interrupt enable */
+
+/*
+ * XOR Accelerator engine Command Block Type
+ */
+struct xor_cb {
+	/*
+	 * Basic 64-bit format XOR CB (Table 19-1, p.463, 440spe_um_1_22.pdf)
+	 */
+	u32	cbc;		/* control */
+	u32	cbbc;		/* byte count */
+	u32	cbs;		/* status */
+	u8	pad0[4];	/* reserved */
+	u32	cbtah;		/* target address high */
+	u32	cbtal;		/* target address low */
+	u32	cblah;		/* link address high */
+	u32	cblal;		/* link address low */
+	struct {
+		u32 h;
+		u32 l;
+	} __attribute__ ((packed)) ops[16];
+} __attribute__ ((packed));
+
+/*
+ * XOR hardware registers Table 19-3, UM 1.22
+ */
+struct xor_regs {
+	u32	op_ar[16][2];	/* operand address[0]-high,[1]-low registers */
+	u8	pad0[352];	/* reserved */
+	u32	cbcr;		/* CB control register */
+	u32	cbbcr;		/* CB byte count register */
+	u32	cbsr;		/* CB status register */
+	u8	pad1[4];	/* reserved */
+	u32	cbtahr;		/* operand target address high register */
+	u32	cbtalr;		/* operand target address low register */
+	u32	cblahr;		/* CB link address high register */
+	u32	cblalr;		/* CB link address low register */
+	u32	crsr;		/* control set register */
+	u32	crrr;		/* control reset register */
+	u32	ccbahr;		/* current CB address high register */
+	u32	ccbalr;		/* current CB address low register */
+	u32	plbr;		/* PLB configuration register */
+	u32	ier;		/* interrupt enable register */
+	u32	pecr;		/* parity error count register */
+	u32	sr;		/* status register */
+	u32	revidr;		/* revision ID register */
+};
+
+#endif /* _PPC440SPE_XOR_H */
diff --git a/drivers/dma/shdma.c b/drivers/dma/shdma.c
index 034ecf0ace03..d10cc899c460 100644
--- a/drivers/dma/shdma.c
+++ b/drivers/dma/shdma.c
@@ -23,16 +23,19 @@
 #include <linux/dmaengine.h>
 #include <linux/delay.h>
 #include <linux/dma-mapping.h>
-#include <linux/dmapool.h>
 #include <linux/platform_device.h>
 #include <cpu/dma.h>
 #include <asm/dma-sh.h>
 #include "shdma.h"
 
 /* DMA descriptor control */
-#define DESC_LAST	(-1)
-#define DESC_COMP	(1)
-#define DESC_NCOMP	(0)
+enum sh_dmae_desc_status {
+	DESC_IDLE,
+	DESC_PREPARED,
+	DESC_SUBMITTED,
+	DESC_COMPLETED,	/* completed, have to call callback */
+	DESC_WAITING,	/* callback called, waiting for ack / re-submit */
+};
 
 #define NR_DESCS_PER_CHANNEL 32
 /*
@@ -45,6 +48,8 @@
  */
 #define RS_DEFAULT  (RS_DUAL)
 
+static void sh_dmae_chan_ld_cleanup(struct sh_dmae_chan *sh_chan, bool all);
+
 #define SH_DMAC_CHAN_BASE(id) (dma_base_addr[id])
 static void sh_dmae_writel(struct sh_dmae_chan *sh_dc, u32 data, u32 reg)
 {
@@ -80,17 +85,17 @@ static int sh_dmae_rst(int id)
 	unsigned short dmaor;
 
 	sh_dmae_ctl_stop(id);
-	dmaor = (dmaor_read_reg(id)|DMAOR_INIT);
+	dmaor = dmaor_read_reg(id) | DMAOR_INIT;
 
 	dmaor_write_reg(id, dmaor);
-	if ((dmaor_read_reg(id) & (DMAOR_AE | DMAOR_NMIF))) {
+	if (dmaor_read_reg(id) & (DMAOR_AE | DMAOR_NMIF)) {
 		pr_warning(KERN_ERR "dma-sh: Can't initialize DMAOR.\n");
 		return -EINVAL;
 	}
 	return 0;
 }
 
-static int dmae_is_idle(struct sh_dmae_chan *sh_chan)
+static int dmae_is_busy(struct sh_dmae_chan *sh_chan)
 {
 	u32 chcr = sh_dmae_readl(sh_chan, CHCR);
 	if (chcr & CHCR_DE) {
@@ -106,19 +111,18 @@ static inline unsigned int calc_xmit_shift(struct sh_dmae_chan *sh_chan)
 	return ts_shift[(chcr & CHCR_TS_MASK) >> CHCR_TS_SHIFT];
 }
 
-static void dmae_set_reg(struct sh_dmae_chan *sh_chan, struct sh_dmae_regs hw)
+static void dmae_set_reg(struct sh_dmae_chan *sh_chan, struct sh_dmae_regs *hw)
 {
-	sh_dmae_writel(sh_chan, hw.sar, SAR);
-	sh_dmae_writel(sh_chan, hw.dar, DAR);
-	sh_dmae_writel(sh_chan,
-		(hw.tcr >> calc_xmit_shift(sh_chan)), TCR);
+	sh_dmae_writel(sh_chan, hw->sar, SAR);
+	sh_dmae_writel(sh_chan, hw->dar, DAR);
+	sh_dmae_writel(sh_chan, hw->tcr >> calc_xmit_shift(sh_chan), TCR);
 }
 
 static void dmae_start(struct sh_dmae_chan *sh_chan)
 {
 	u32 chcr = sh_dmae_readl(sh_chan, CHCR);
 
-	chcr |= (CHCR_DE|CHCR_IE);
+	chcr |= CHCR_DE | CHCR_IE;
 	sh_dmae_writel(sh_chan, chcr, CHCR);
 }
 
@@ -132,7 +136,7 @@ static void dmae_halt(struct sh_dmae_chan *sh_chan)
 
 static int dmae_set_chcr(struct sh_dmae_chan *sh_chan, u32 val)
 {
-	int ret = dmae_is_idle(sh_chan);
+	int ret = dmae_is_busy(sh_chan);
 	/* When DMA was working, can not set data to CHCR */
 	if (ret)
 		return ret;
@@ -149,7 +153,7 @@ static int dmae_set_dmars(struct sh_dmae_chan *sh_chan, u16 val)
 {
 	u32 addr;
 	int shift = 0;
-	int ret = dmae_is_idle(sh_chan);
+	int ret = dmae_is_busy(sh_chan);
 	if (ret)
 		return ret;
 
@@ -185,8 +189,9 @@ static int dmae_set_dmars(struct sh_dmae_chan *sh_chan, u16 val)
 
 static dma_cookie_t sh_dmae_tx_submit(struct dma_async_tx_descriptor *tx)
 {
-	struct sh_desc *desc = tx_to_sh_desc(tx);
+	struct sh_desc *desc = tx_to_sh_desc(tx), *chunk, *last = desc, *c;
 	struct sh_dmae_chan *sh_chan = to_sh_chan(tx->chan);
+	dma_async_tx_callback callback = tx->callback;
 	dma_cookie_t cookie;
 
 	spin_lock_bh(&sh_chan->desc_lock);
@@ -196,45 +201,53 @@ static dma_cookie_t sh_dmae_tx_submit(struct dma_async_tx_descriptor *tx)
 	if (cookie < 0)
 		cookie = 1;
 
-	/* If desc only in the case of 1 */
-	if (desc->async_tx.cookie != -EBUSY)
-		desc->async_tx.cookie = cookie;
-	sh_chan->common.cookie = desc->async_tx.cookie;
+	sh_chan->common.cookie = cookie;
+	tx->cookie = cookie;
+
+	/* Mark all chunks of this descriptor as submitted, move to the queue */
+	list_for_each_entry_safe(chunk, c, desc->node.prev, node) {
+		/*
+		 * All chunks are on the global ld_free, so, we have to find
+		 * the end of the chain ourselves
+		 */
+		if (chunk != desc && (chunk->mark == DESC_IDLE ||
+				      chunk->async_tx.cookie > 0 ||
+				      chunk->async_tx.cookie == -EBUSY ||
+				      &chunk->node == &sh_chan->ld_free))
+			break;
+		chunk->mark = DESC_SUBMITTED;
+		/* Callback goes to the last chunk */
+		chunk->async_tx.callback = NULL;
+		chunk->cookie = cookie;
+		list_move_tail(&chunk->node, &sh_chan->ld_queue);
+		last = chunk;
+	}
+
+	last->async_tx.callback = callback;
+	last->async_tx.callback_param = tx->callback_param;
 
-	list_splice_init(&desc->tx_list, sh_chan->ld_queue.prev);
+	dev_dbg(sh_chan->dev, "submit #%d@%p on %d: %x[%d] -> %x\n",
+		tx->cookie, &last->async_tx, sh_chan->id,
+		desc->hw.sar, desc->hw.tcr, desc->hw.dar);
 
 	spin_unlock_bh(&sh_chan->desc_lock);
 
 	return cookie;
 }
 
+/* Called with desc_lock held */
 static struct sh_desc *sh_dmae_get_desc(struct sh_dmae_chan *sh_chan)
 {
-	struct sh_desc *desc, *_desc, *ret = NULL;
+	struct sh_desc *desc;
 
-	spin_lock_bh(&sh_chan->desc_lock);
-	list_for_each_entry_safe(desc, _desc, &sh_chan->ld_free, node) {
-		if (async_tx_test_ack(&desc->async_tx)) {
+	list_for_each_entry(desc, &sh_chan->ld_free, node)
+		if (desc->mark != DESC_PREPARED) {
+			BUG_ON(desc->mark != DESC_IDLE);
 			list_del(&desc->node);
-			ret = desc;
-			break;
+			return desc;
 		}
-	}
-	spin_unlock_bh(&sh_chan->desc_lock);
-
-	return ret;
-}
-
-static void sh_dmae_put_desc(struct sh_dmae_chan *sh_chan, struct sh_desc *desc)
-{
-	if (desc) {
-		spin_lock_bh(&sh_chan->desc_lock);
-
-		list_splice_init(&desc->tx_list, &sh_chan->ld_free);
-		list_add(&desc->node, &sh_chan->ld_free);
 
-		spin_unlock_bh(&sh_chan->desc_lock);
-	}
+	return NULL;
 }
 
 static int sh_dmae_alloc_chan_resources(struct dma_chan *chan)
@@ -253,11 +266,10 @@ static int sh_dmae_alloc_chan_resources(struct dma_chan *chan)
 		dma_async_tx_descriptor_init(&desc->async_tx,
 					&sh_chan->common);
 		desc->async_tx.tx_submit = sh_dmae_tx_submit;
-		desc->async_tx.flags = DMA_CTRL_ACK;
-		INIT_LIST_HEAD(&desc->tx_list);
-		sh_dmae_put_desc(sh_chan, desc);
+		desc->mark = DESC_IDLE;
 
 		spin_lock_bh(&sh_chan->desc_lock);
+		list_add(&desc->node, &sh_chan->ld_free);
 		sh_chan->descs_allocated++;
 	}
 	spin_unlock_bh(&sh_chan->desc_lock);
@@ -274,7 +286,10 @@ static void sh_dmae_free_chan_resources(struct dma_chan *chan)
 	struct sh_desc *desc, *_desc;
 	LIST_HEAD(list);
 
-	BUG_ON(!list_empty(&sh_chan->ld_queue));
+	/* Prepared and not submitted descriptors can still be on the queue */
+	if (!list_empty(&sh_chan->ld_queue))
+		sh_dmae_chan_ld_cleanup(sh_chan, true);
+
 	spin_lock_bh(&sh_chan->desc_lock);
 
 	list_splice_init(&sh_chan->ld_free, &list);
@@ -293,6 +308,8 @@ static struct dma_async_tx_descriptor *sh_dmae_prep_memcpy(
 	struct sh_dmae_chan *sh_chan;
 	struct sh_desc *first = NULL, *prev = NULL, *new;
 	size_t copy_size;
+	LIST_HEAD(tx_list);
+	int chunks = (len + SH_DMA_TCR_MAX) / (SH_DMA_TCR_MAX + 1);
 
 	if (!chan)
 		return NULL;
@@ -302,108 +319,189 @@ static struct dma_async_tx_descriptor *sh_dmae_prep_memcpy(
 
 	sh_chan = to_sh_chan(chan);
 
+	/* Have to lock the whole loop to protect against concurrent release */
+	spin_lock_bh(&sh_chan->desc_lock);
+
+	/*
+	 * Chaining:
+	 * first descriptor is what user is dealing with in all API calls, its
+	 *	cookie is at first set to -EBUSY, at tx-submit to a positive
+	 *	number
+	 * if more than one chunk is needed further chunks have cookie = -EINVAL
+	 * the last chunk, if not equal to the first, has cookie = -ENOSPC
+	 * all chunks are linked onto the tx_list head with their .node heads
+	 *	only during this function, then they are immediately spliced
+	 *	back onto the free list in form of a chain
+	 */
 	do {
-		/* Allocate the link descriptor from DMA pool */
+		/* Allocate the link descriptor from the free list */
 		new = sh_dmae_get_desc(sh_chan);
 		if (!new) {
 			dev_err(sh_chan->dev,
-					"No free memory for link descriptor\n");
-			goto err_get_desc;
+				"No free memory for link descriptor\n");
+			list_for_each_entry(new, &tx_list, node)
+				new->mark = DESC_IDLE;
+			list_splice(&tx_list, &sh_chan->ld_free);
+			spin_unlock_bh(&sh_chan->desc_lock);
+			return NULL;
 		}
 
-		copy_size = min(len, (size_t)SH_DMA_TCR_MAX);
+		copy_size = min(len, (size_t)SH_DMA_TCR_MAX + 1);
 
 		new->hw.sar = dma_src;
 		new->hw.dar = dma_dest;
 		new->hw.tcr = copy_size;
-		if (!first)
+		if (!first) {
+			/* First desc */
+			new->async_tx.cookie = -EBUSY;
 			first = new;
+		} else {
+			/* Other desc - invisible to the user */
+			new->async_tx.cookie = -EINVAL;
+		}
 
-		new->mark = DESC_NCOMP;
-		async_tx_ack(&new->async_tx);
+		dev_dbg(sh_chan->dev,
+			"chaining %u of %u with %p, dst %x, cookie %d\n",
+			copy_size, len, &new->async_tx, dma_dest,
+			new->async_tx.cookie);
+
+		new->mark = DESC_PREPARED;
+		new->async_tx.flags = flags;
+		new->chunks = chunks--;
 
 		prev = new;
 		len -= copy_size;
 		dma_src += copy_size;
 		dma_dest += copy_size;
 		/* Insert the link descriptor to the LD ring */
-		list_add_tail(&new->node, &first->tx_list);
+		list_add_tail(&new->node, &tx_list);
 	} while (len);
 
-	new->async_tx.flags = flags; /* client is in control of this ack */
-	new->async_tx.cookie = -EBUSY; /* Last desc */
+	if (new != first)
+		new->async_tx.cookie = -ENOSPC;
 
-	return &first->async_tx;
+	/* Put them back on the free list, so, they don't get lost */
+	list_splice_tail(&tx_list, &sh_chan->ld_free);
 
-err_get_desc:
-	sh_dmae_put_desc(sh_chan, first);
-	return NULL;
+	spin_unlock_bh(&sh_chan->desc_lock);
 
+	return &first->async_tx;
 }
 
-/*
- * sh_chan_ld_cleanup - Clean up link descriptors
- *
- * This function clean up the ld_queue of DMA channel.
- */
-static void sh_dmae_chan_ld_cleanup(struct sh_dmae_chan *sh_chan)
+static dma_async_tx_callback __ld_cleanup(struct sh_dmae_chan *sh_chan, bool all)
 {
 	struct sh_desc *desc, *_desc;
+	/* Is the "exposed" head of a chain acked? */
+	bool head_acked = false;
+	dma_cookie_t cookie = 0;
+	dma_async_tx_callback callback = NULL;
+	void *param = NULL;
 
 	spin_lock_bh(&sh_chan->desc_lock);
 	list_for_each_entry_safe(desc, _desc, &sh_chan->ld_queue, node) {
-		dma_async_tx_callback callback;
-		void *callback_param;
-
-		/* non send data */
-		if (desc->mark == DESC_NCOMP)
+		struct dma_async_tx_descriptor *tx = &desc->async_tx;
+
+		BUG_ON(tx->cookie > 0 && tx->cookie != desc->cookie);
+		BUG_ON(desc->mark != DESC_SUBMITTED &&
+		       desc->mark != DESC_COMPLETED &&
+		       desc->mark != DESC_WAITING);
+
+		/*
+		 * queue is ordered, and we use this loop to (1) clean up all
+		 * completed descriptors, and to (2) update descriptor flags of
+		 * any chunks in a (partially) completed chain
+		 */
+		if (!all && desc->mark == DESC_SUBMITTED &&
+		    desc->cookie != cookie)
 			break;
 
-		/* send data sesc */
-		callback = desc->async_tx.callback;
-		callback_param = desc->async_tx.callback_param;
+		if (tx->cookie > 0)
+			cookie = tx->cookie;
 
-		/* Remove from ld_queue list */
-		list_splice_init(&desc->tx_list, &sh_chan->ld_free);
+		if (desc->mark == DESC_COMPLETED && desc->chunks == 1) {
+			BUG_ON(sh_chan->completed_cookie != desc->cookie - 1);
+			sh_chan->completed_cookie = desc->cookie;
+		}
 
-		dev_dbg(sh_chan->dev, "link descriptor %p will be recycle.\n",
-				desc);
+		/* Call callback on the last chunk */
+		if (desc->mark == DESC_COMPLETED && tx->callback) {
+			desc->mark = DESC_WAITING;
+			callback = tx->callback;
+			param = tx->callback_param;
+			dev_dbg(sh_chan->dev, "descriptor #%d@%p on %d callback\n",
+				tx->cookie, tx, sh_chan->id);
+			BUG_ON(desc->chunks != 1);
+			break;
+		}
 
-		list_move(&desc->node, &sh_chan->ld_free);
-		/* Run the link descriptor callback function */
-		if (callback) {
-			spin_unlock_bh(&sh_chan->desc_lock);
-			dev_dbg(sh_chan->dev, "link descriptor %p callback\n",
-					desc);
-			callback(callback_param);
-			spin_lock_bh(&sh_chan->desc_lock);
+		if (tx->cookie > 0 || tx->cookie == -EBUSY) {
+			if (desc->mark == DESC_COMPLETED) {
+				BUG_ON(tx->cookie < 0);
+				desc->mark = DESC_WAITING;
+			}
+			head_acked = async_tx_test_ack(tx);
+		} else {
+			switch (desc->mark) {
+			case DESC_COMPLETED:
+				desc->mark = DESC_WAITING;
+				/* Fall through */
+			case DESC_WAITING:
+				if (head_acked)
+					async_tx_ack(&desc->async_tx);
+			}
+		}
+
+		dev_dbg(sh_chan->dev, "descriptor %p #%d completed.\n",
+			tx, tx->cookie);
+
+		if (((desc->mark == DESC_COMPLETED ||
+		      desc->mark == DESC_WAITING) &&
+		     async_tx_test_ack(&desc->async_tx)) || all) {
+			/* Remove from ld_queue list */
+			desc->mark = DESC_IDLE;
+			list_move(&desc->node, &sh_chan->ld_free);
 		}
 	}
 	spin_unlock_bh(&sh_chan->desc_lock);
+
+	if (callback)
+		callback(param);
+
+	return callback;
+}
+
+/*
+ * sh_chan_ld_cleanup - Clean up link descriptors
+ *
+ * This function cleans up the ld_queue of DMA channel.
+ */
+static void sh_dmae_chan_ld_cleanup(struct sh_dmae_chan *sh_chan, bool all)
+{
+	while (__ld_cleanup(sh_chan, all))
+		;
 }
 
 static void sh_chan_xfer_ld_queue(struct sh_dmae_chan *sh_chan)
 {
-	struct list_head *ld_node;
-	struct sh_dmae_regs hw;
+	struct sh_desc *sd;
 
+	spin_lock_bh(&sh_chan->desc_lock);
 	/* DMA work check */
-	if (dmae_is_idle(sh_chan))
+	if (dmae_is_busy(sh_chan)) {
+		spin_unlock_bh(&sh_chan->desc_lock);
 		return;
+	}
 
 	/* Find the first un-transfer desciptor */
-	for (ld_node = sh_chan->ld_queue.next;
-		(ld_node != &sh_chan->ld_queue)
-			&& (to_sh_desc(ld_node)->mark == DESC_COMP);
-		ld_node = ld_node->next)
-		cpu_relax();
-
-	if (ld_node != &sh_chan->ld_queue) {
-		/* Get the ld start address from ld_queue */
-		hw = to_sh_desc(ld_node)->hw;
-		dmae_set_reg(sh_chan, hw);
-		dmae_start(sh_chan);
-	}
+	list_for_each_entry(sd, &sh_chan->ld_queue, node)
+		if (sd->mark == DESC_SUBMITTED) {
+			/* Get the ld start address from ld_queue */
+			dmae_set_reg(sh_chan, &sd->hw);
+			dmae_start(sh_chan);
+			break;
+		}
+
+	spin_unlock_bh(&sh_chan->desc_lock);
 }
 
 static void sh_dmae_memcpy_issue_pending(struct dma_chan *chan)
@@ -421,12 +519,11 @@ static enum dma_status sh_dmae_is_complete(struct dma_chan *chan,
 	dma_cookie_t last_used;
 	dma_cookie_t last_complete;
 
-	sh_dmae_chan_ld_cleanup(sh_chan);
+	sh_dmae_chan_ld_cleanup(sh_chan, false);
 
 	last_used = chan->cookie;
 	last_complete = sh_chan->completed_cookie;
-	if (last_complete == -EBUSY)
-		last_complete = last_used;
+	BUG_ON(last_complete < 0);
 
 	if (done)
 		*done = last_complete;
@@ -481,11 +578,13 @@ static irqreturn_t sh_dmae_err(int irq, void *data)
 		err = sh_dmae_rst(0);
 		if (err)
 			return err;
+#ifdef SH_DMAC_BASE1
 		if (shdev->pdata.mode & SHDMA_DMAOR1) {
 			err = sh_dmae_rst(1);
 			if (err)
 				return err;
 		}
+#endif
 		disable_irq(irq);
 		return IRQ_HANDLED;
 	}
@@ -495,34 +594,25 @@ static irqreturn_t sh_dmae_err(int irq, void *data)
 static void dmae_do_tasklet(unsigned long data)
 {
 	struct sh_dmae_chan *sh_chan = (struct sh_dmae_chan *)data;
-	struct sh_desc *desc, *_desc, *cur_desc = NULL;
+	struct sh_desc *desc;
 	u32 sar_buf = sh_dmae_readl(sh_chan, SAR);
-	list_for_each_entry_safe(desc, _desc,
-					&sh_chan->ld_queue, node) {
-		if ((desc->hw.sar + desc->hw.tcr) == sar_buf) {
-			cur_desc = desc;
-			break;
-		}
-	}
 
-	if (cur_desc) {
-		switch (cur_desc->async_tx.cookie) {
-		case 0: /* other desc data */
-			break;
-		case -EBUSY: /* last desc */
-		sh_chan->completed_cookie =
-				cur_desc->async_tx.cookie;
-			break;
-		default: /* first desc ( 0 < )*/
-			sh_chan->completed_cookie =
-				cur_desc->async_tx.cookie - 1;
+	spin_lock(&sh_chan->desc_lock);
+	list_for_each_entry(desc, &sh_chan->ld_queue, node) {
+		if ((desc->hw.sar + desc->hw.tcr) == sar_buf &&
+		    desc->mark == DESC_SUBMITTED) {
+			dev_dbg(sh_chan->dev, "done #%d@%p dst %u\n",
+				desc->async_tx.cookie, &desc->async_tx,
+				desc->hw.dar);
+			desc->mark = DESC_COMPLETED;
 			break;
 		}
-		cur_desc->mark = DESC_COMP;
 	}
+	spin_unlock(&sh_chan->desc_lock);
+
 	/* Next desc */
 	sh_chan_xfer_ld_queue(sh_chan);
-	sh_dmae_chan_ld_cleanup(sh_chan);
+	sh_dmae_chan_ld_cleanup(sh_chan, false);
 }
 
 static unsigned int get_dmae_irq(unsigned int id)
@@ -543,8 +633,8 @@ static int __devinit sh_dmae_chan_probe(struct sh_dmae_device *shdev, int id)
 	/* alloc channel */
 	new_sh_chan = kzalloc(sizeof(struct sh_dmae_chan), GFP_KERNEL);
 	if (!new_sh_chan) {
-		dev_err(shdev->common.dev, "No free memory for allocating "
-				"dma channels!\n");
+		dev_err(shdev->common.dev,
+			"No free memory for allocating dma channels!\n");
 		return -ENOMEM;
 	}
 
@@ -586,8 +676,8 @@ static int __devinit sh_dmae_chan_probe(struct sh_dmae_device *shdev, int id)
 			"sh-dmae%d", new_sh_chan->id);
 
 	/* set up channel irq */
-	err = request_irq(irq, &sh_dmae_interrupt,
-		irqflags, new_sh_chan->dev_id, new_sh_chan);
+	err = request_irq(irq, &sh_dmae_interrupt, irqflags,
+			  new_sh_chan->dev_id, new_sh_chan);
 	if (err) {
 		dev_err(shdev->common.dev, "DMA channel %d request_irq error "
 			"with return %d\n", id, err);
@@ -676,6 +766,8 @@ static int __init sh_dmae_probe(struct platform_device *pdev)
 	shdev->common.device_is_tx_complete = sh_dmae_is_complete;
 	shdev->common.device_issue_pending = sh_dmae_memcpy_issue_pending;
 	shdev->common.dev = &pdev->dev;
+	/* Default transfer size of 32 bytes requires 32-byte alignment */
+	shdev->common.copy_align = 5;
 
 #if defined(CONFIG_CPU_SH4)
 	/* Non Mix IRQ mode SH7722/SH7730 etc... */
@@ -688,8 +780,8 @@ static int __init sh_dmae_probe(struct platform_device *pdev)
 	}
 
 	for (ecnt = 0 ; ecnt < ARRAY_SIZE(eirq); ecnt++) {
-		err = request_irq(eirq[ecnt], sh_dmae_err,
-			irqflags, "DMAC Address Error", shdev);
+		err = request_irq(eirq[ecnt], sh_dmae_err, irqflags,
+				  "DMAC Address Error", shdev);
 		if (err) {
 			dev_err(&pdev->dev, "DMA device request_irq"
 				"error (irq %d) with return %d\n",
diff --git a/drivers/dma/shdma.h b/drivers/dma/shdma.h
index 2b4bc15a2c0a..108f1cffb6f5 100644
--- a/drivers/dma/shdma.h
+++ b/drivers/dma/shdma.h
@@ -13,9 +13,9 @@
 #ifndef __DMA_SHDMA_H
 #define __DMA_SHDMA_H
 
-#include <linux/device.h>
-#include <linux/dmapool.h>
 #include <linux/dmaengine.h>
+#include <linux/interrupt.h>
+#include <linux/list.h>
 
 #define SH_DMA_TCR_MAX 0x00FFFFFF	/* 16MB */
 
@@ -26,24 +26,27 @@ struct sh_dmae_regs {
 };
 
 struct sh_desc {
-	struct list_head tx_list;
 	struct sh_dmae_regs hw;
 	struct list_head node;
 	struct dma_async_tx_descriptor async_tx;
+	dma_cookie_t cookie;
+	int chunks;
 	int mark;
 };
 
+struct device;
+
 struct sh_dmae_chan {
 	dma_cookie_t completed_cookie;	/* The maximum cookie completed */
-	spinlock_t desc_lock;			/* Descriptor operation lock */
-	struct list_head ld_queue;		/* Link descriptors queue */
-	struct list_head ld_free;		/* Link descriptors free */
-	struct dma_chan common;			/* DMA common channel */
-	struct device *dev;				/* Channel device */
+	spinlock_t desc_lock;		/* Descriptor operation lock */
+	struct list_head ld_queue;	/* Link descriptors queue */
+	struct list_head ld_free;	/* Link descriptors free */
+	struct dma_chan common;		/* DMA common channel */
+	struct device *dev;		/* Channel device */
 	struct tasklet_struct tasklet;	/* Tasklet */
-	int descs_allocated;			/* desc count */
+	int descs_allocated;		/* desc count */
 	int id;				/* Raw id of this channel */
-	char dev_id[16];	/* unique name per DMAC of channel */
+	char dev_id[16];		/* unique name per DMAC of channel */
 
 	/* Set chcr */
 	int (*set_chcr)(struct sh_dmae_chan *sh_chan, u32 regs);
diff --git a/drivers/dma/txx9dmac.c b/drivers/dma/txx9dmac.c
index fb6bb64e8861..3ebc61067e54 100644
--- a/drivers/dma/txx9dmac.c
+++ b/drivers/dma/txx9dmac.c
@@ -1313,7 +1313,7 @@ static int txx9dmac_resume_noirq(struct device *dev)
 
 }
 
-static struct dev_pm_ops txx9dmac_dev_pm_ops = {
+static const struct dev_pm_ops txx9dmac_dev_pm_ops = {
 	.suspend_noirq = txx9dmac_suspend_noirq,
 	.resume_noirq = txx9dmac_resume_noirq,
 };