]> pilppa.com Git - linux-2.6-omap-h63xx.git/commitdiff
dmaengine: DMA engine driver for Marvell XOR engine
authorSaeed Bishara <saeed@marvell.com>
Tue, 8 Jul 2008 18:58:36 +0000 (11:58 -0700)
committerDan Williams <dan.j.williams@intel.com>
Tue, 8 Jul 2008 18:58:36 +0000 (11:58 -0700)
The XOR engine found in Marvell's SoCs and system controllers
provides XOR and DMA operation, iSCSI CRC32C calculation, memory
initialization, and memory ECC error cleanup operation support.

This driver implements the DMA engine API and supports the following
capabilities:
- memcpy
- xor
- memset

The XOR engine can be used by DMA engine clients implemented in the
kernel, one of those clients is the RAID module.  In that case, I
observed 20% improvement in the raid5 write throughput, and 40%
decrease in the CPU utilization when doing array construction, those
results obtained on an 5182 running at 500Mhz.

When enabling the NET DMA client, the performance decreased, so
meanwhile it is recommended to keep this client off.

Signed-off-by: Saeed Bishara <saeed@marvell.com>
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: Nicolas Pitre <nico@marvell.com>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
drivers/dma/Kconfig
drivers/dma/Makefile
drivers/dma/mv_xor.c [new file with mode: 0644]
drivers/dma/mv_xor.h [new file with mode: 0644]
include/asm-arm/plat-orion/mv_xor.h [new file with mode: 0644]

index e4dd0065da33e001de7ce80de980434b49d3758b..5af8b1cfc1e9ce714eceb358d62d37b2841d8af1 100644 (file)
@@ -47,6 +47,14 @@ config FSL_DMA
          MPC8560/40, MPC8555, MPC8548 and MPC8641 processors.
          The MPC8349, MPC8360 is also supported.
 
+config MV_XOR
+       bool "Marvell XOR engine support"
+       depends on PLAT_ORION
+       select ASYNC_CORE
+       select DMA_ENGINE
+       ---help---
+         Enable support for the Marvell XOR engine.
+
 config DMA_ENGINE
        bool
 
index c8036d94590277d24f994a82c61fccbe51a89f07..ee272fd329c916996f8e5bd89978b90c7f95ca28 100644 (file)
@@ -4,3 +4,4 @@ obj-$(CONFIG_INTEL_IOATDMA) += ioatdma.o
 ioatdma-objs := ioat.o ioat_dma.o ioat_dca.o
 obj-$(CONFIG_INTEL_IOP_ADMA) += iop-adma.o
 obj-$(CONFIG_FSL_DMA) += fsldma.o
+obj-$(CONFIG_MV_XOR) += mv_xor.o
diff --git a/drivers/dma/mv_xor.c b/drivers/dma/mv_xor.c
new file mode 100644 (file)
index 0000000..f0c123c
--- /dev/null
@@ -0,0 +1,1364 @@
+/*
+ * offload engine driver for the Marvell XOR engine
+ * Copyright (C) 2007, 2008, Marvell International Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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.,
+ * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+ */
+
+#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/platform_device.h>
+#include <linux/memory.h>
+#include <asm/plat-orion/mv_xor.h>
+#include "mv_xor.h"
+
+static void mv_xor_issue_pending(struct dma_chan *chan);
+
+#define to_mv_xor_chan(chan)           \
+       container_of(chan, struct mv_xor_chan, common)
+
+#define to_mv_xor_device(dev)          \
+       container_of(dev, struct mv_xor_device, common)
+
+#define to_mv_xor_slot(tx)             \
+       container_of(tx, struct mv_xor_desc_slot, async_tx)
+
+static void mv_desc_init(struct mv_xor_desc_slot *desc, unsigned long flags)
+{
+       struct mv_xor_desc *hw_desc = desc->hw_desc;
+
+       hw_desc->status = (1 << 31);
+       hw_desc->phy_next_desc = 0;
+       hw_desc->desc_command = (1 << 31);
+}
+
+static u32 mv_desc_get_dest_addr(struct mv_xor_desc_slot *desc)
+{
+       struct mv_xor_desc *hw_desc = desc->hw_desc;
+       return hw_desc->phy_dest_addr;
+}
+
+static u32 mv_desc_get_src_addr(struct mv_xor_desc_slot *desc,
+                               int src_idx)
+{
+       struct mv_xor_desc *hw_desc = desc->hw_desc;
+       return hw_desc->phy_src_addr[src_idx];
+}
+
+
+static void mv_desc_set_byte_count(struct mv_xor_desc_slot *desc,
+                                  u32 byte_count)
+{
+       struct mv_xor_desc *hw_desc = desc->hw_desc;
+       hw_desc->byte_count = byte_count;
+}
+
+static void mv_desc_set_next_desc(struct mv_xor_desc_slot *desc,
+                                 u32 next_desc_addr)
+{
+       struct mv_xor_desc *hw_desc = desc->hw_desc;
+       BUG_ON(hw_desc->phy_next_desc);
+       hw_desc->phy_next_desc = next_desc_addr;
+}
+
+static void mv_desc_clear_next_desc(struct mv_xor_desc_slot *desc)
+{
+       struct mv_xor_desc *hw_desc = desc->hw_desc;
+       hw_desc->phy_next_desc = 0;
+}
+
+static void mv_desc_set_block_fill_val(struct mv_xor_desc_slot *desc, u32 val)
+{
+       desc->value = val;
+}
+
+static void mv_desc_set_dest_addr(struct mv_xor_desc_slot *desc,
+                                 dma_addr_t addr)
+{
+       struct mv_xor_desc *hw_desc = desc->hw_desc;
+       hw_desc->phy_dest_addr = addr;
+}
+
+static int mv_chan_memset_slot_count(size_t len)
+{
+       return 1;
+}
+
+#define mv_chan_memcpy_slot_count(c) mv_chan_memset_slot_count(c)
+
+static void mv_desc_set_src_addr(struct mv_xor_desc_slot *desc,
+                                int index, dma_addr_t addr)
+{
+       struct mv_xor_desc *hw_desc = desc->hw_desc;
+       hw_desc->phy_src_addr[index] = addr;
+       if (desc->type == DMA_XOR)
+               hw_desc->desc_command |= (1 << index);
+}
+
+static u32 mv_chan_get_current_desc(struct mv_xor_chan *chan)
+{
+       return __raw_readl(XOR_CURR_DESC(chan));
+}
+
+static void mv_chan_set_next_descriptor(struct mv_xor_chan *chan,
+                                       u32 next_desc_addr)
+{
+       __raw_writel(next_desc_addr, XOR_NEXT_DESC(chan));
+}
+
+static void mv_chan_set_dest_pointer(struct mv_xor_chan *chan, u32 desc_addr)
+{
+       __raw_writel(desc_addr, XOR_DEST_POINTER(chan));
+}
+
+static void mv_chan_set_block_size(struct mv_xor_chan *chan, u32 block_size)
+{
+       __raw_writel(block_size, XOR_BLOCK_SIZE(chan));
+}
+
+static void mv_chan_set_value(struct mv_xor_chan *chan, u32 value)
+{
+       __raw_writel(value, XOR_INIT_VALUE_LOW(chan));
+       __raw_writel(value, XOR_INIT_VALUE_HIGH(chan));
+}
+
+static void mv_chan_unmask_interrupts(struct mv_xor_chan *chan)
+{
+       u32 val = __raw_readl(XOR_INTR_MASK(chan));
+       val |= XOR_INTR_MASK_VALUE << (chan->idx * 16);
+       __raw_writel(val, XOR_INTR_MASK(chan));
+}
+
+static u32 mv_chan_get_intr_cause(struct mv_xor_chan *chan)
+{
+       u32 intr_cause = __raw_readl(XOR_INTR_CAUSE(chan));
+       intr_cause = (intr_cause >> (chan->idx * 16)) & 0xFFFF;
+       return intr_cause;
+}
+
+static int mv_is_err_intr(u32 intr_cause)
+{
+       if (intr_cause & ((1<<4)|(1<<5)|(1<<6)|(1<<7)|(1<<8)|(1<<9)))
+               return 1;
+
+       return 0;
+}
+
+static void mv_xor_device_clear_eoc_cause(struct mv_xor_chan *chan)
+{
+       u32 val = (1 << (1 + (chan->idx * 16)));
+       dev_dbg(chan->device->common.dev, "%s, val 0x%08x\n", __func__, val);
+       __raw_writel(val, XOR_INTR_CAUSE(chan));
+}
+
+static void mv_xor_device_clear_err_status(struct mv_xor_chan *chan)
+{
+       u32 val = 0xFFFF0000 >> (chan->idx * 16);
+       __raw_writel(val, XOR_INTR_CAUSE(chan));
+}
+
+static int mv_can_chain(struct mv_xor_desc_slot *desc)
+{
+       struct mv_xor_desc_slot *chain_old_tail = list_entry(
+               desc->chain_node.prev, struct mv_xor_desc_slot, chain_node);
+
+       if (chain_old_tail->type != desc->type)
+               return 0;
+       if (desc->type == DMA_MEMSET)
+               return 0;
+
+       return 1;
+}
+
+static void mv_set_mode(struct mv_xor_chan *chan,
+                              enum dma_transaction_type type)
+{
+       u32 op_mode;
+       u32 config = __raw_readl(XOR_CONFIG(chan));
+
+       switch (type) {
+       case DMA_XOR:
+               op_mode = XOR_OPERATION_MODE_XOR;
+               break;
+       case DMA_MEMCPY:
+               op_mode = XOR_OPERATION_MODE_MEMCPY;
+               break;
+       case DMA_MEMSET:
+               op_mode = XOR_OPERATION_MODE_MEMSET;
+               break;
+       default:
+               dev_printk(KERN_ERR, chan->device->common.dev,
+                          "error: unsupported operation %d.\n",
+                          type);
+               BUG();
+               return;
+       }
+
+       config &= ~0x7;
+       config |= op_mode;
+       __raw_writel(config, XOR_CONFIG(chan));
+       chan->current_type = type;
+}
+
+static void mv_chan_activate(struct mv_xor_chan *chan)
+{
+       u32 activation;
+
+       dev_dbg(chan->device->common.dev, " activate chan.\n");
+       activation = __raw_readl(XOR_ACTIVATION(chan));
+       activation |= 0x1;
+       __raw_writel(activation, XOR_ACTIVATION(chan));
+}
+
+static char mv_chan_is_busy(struct mv_xor_chan *chan)
+{
+       u32 state = __raw_readl(XOR_ACTIVATION(chan));
+
+       state = (state >> 4) & 0x3;
+
+       return (state == 1) ? 1 : 0;
+}
+
+static int mv_chan_xor_slot_count(size_t len, int src_cnt)
+{
+       return 1;
+}
+
+/**
+ * mv_xor_free_slots - flags descriptor slots for reuse
+ * @slot: Slot to free
+ * Caller must hold &mv_chan->lock while calling this function
+ */
+static void mv_xor_free_slots(struct mv_xor_chan *mv_chan,
+                             struct mv_xor_desc_slot *slot)
+{
+       dev_dbg(mv_chan->device->common.dev, "%s %d slot %p\n",
+               __func__, __LINE__, slot);
+
+       slot->slots_per_op = 0;
+
+}
+
+/*
+ * mv_xor_start_new_chain - program the engine to operate on new chain headed by
+ * sw_desc
+ * Caller must hold &mv_chan->lock while calling this function
+ */
+static void mv_xor_start_new_chain(struct mv_xor_chan *mv_chan,
+                                  struct mv_xor_desc_slot *sw_desc)
+{
+       dev_dbg(mv_chan->device->common.dev, "%s %d: sw_desc %p\n",
+               __func__, __LINE__, sw_desc);
+       if (sw_desc->type != mv_chan->current_type)
+               mv_set_mode(mv_chan, sw_desc->type);
+
+       if (sw_desc->type == DMA_MEMSET) {
+               /* for memset requests we need to program the engine, no
+                * descriptors used.
+                */
+               struct mv_xor_desc *hw_desc = sw_desc->hw_desc;
+               mv_chan_set_dest_pointer(mv_chan, hw_desc->phy_dest_addr);
+               mv_chan_set_block_size(mv_chan, sw_desc->unmap_len);
+               mv_chan_set_value(mv_chan, sw_desc->value);
+       } else {
+               /* set the hardware chain */
+               mv_chan_set_next_descriptor(mv_chan, sw_desc->async_tx.phys);
+       }
+       mv_chan->pending += sw_desc->slot_cnt;
+       mv_xor_issue_pending(&mv_chan->common);
+}
+
+static dma_cookie_t
+mv_xor_run_tx_complete_actions(struct mv_xor_desc_slot *desc,
+       struct mv_xor_chan *mv_chan, dma_cookie_t cookie)
+{
+       BUG_ON(desc->async_tx.cookie < 0);
+
+       if (desc->async_tx.cookie > 0) {
+               cookie = desc->async_tx.cookie;
+
+               /* 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?)
+                */
+               if (desc->group_head && desc->unmap_len) {
+                       struct mv_xor_desc_slot *unmap = desc->group_head;
+                       struct device *dev =
+                               &mv_chan->device->pdev->dev;
+                       u32 len = unmap->unmap_len;
+                       u32 src_cnt = unmap->unmap_src_cnt;
+                       dma_addr_t addr = mv_desc_get_dest_addr(unmap);
+
+                       dma_unmap_page(dev, addr, len, DMA_FROM_DEVICE);
+                       while (src_cnt--) {
+                               addr = mv_desc_get_src_addr(unmap, src_cnt);
+                               dma_unmap_page(dev, addr, len, DMA_TO_DEVICE);
+                       }
+                       desc->group_head = NULL;
+               }
+       }
+
+       /* run dependent operations */
+       async_tx_run_dependencies(&desc->async_tx);
+
+       return cookie;
+}
+
+static int
+mv_xor_clean_completed_slots(struct mv_xor_chan *mv_chan)
+{
+       struct mv_xor_desc_slot *iter, *_iter;
+
+       dev_dbg(mv_chan->device->common.dev, "%s %d\n", __func__, __LINE__);
+       list_for_each_entry_safe(iter, _iter, &mv_chan->completed_slots,
+                                completed_node) {
+
+               if (async_tx_test_ack(&iter->async_tx)) {
+                       list_del(&iter->completed_node);
+                       mv_xor_free_slots(mv_chan, iter);
+               }
+       }
+       return 0;
+}
+
+static int
+mv_xor_clean_slot(struct mv_xor_desc_slot *desc,
+       struct mv_xor_chan *mv_chan)
+{
+       dev_dbg(mv_chan->device->common.dev, "%s %d: desc %p flags %d\n",
+               __func__, __LINE__, desc, desc->async_tx.flags);
+       list_del(&desc->chain_node);
+       /* the client is allowed to attach dependent operations
+        * until 'ack' is set
+        */
+       if (!async_tx_test_ack(&desc->async_tx)) {
+               /* move this slot to the completed_slots */
+               list_add_tail(&desc->completed_node, &mv_chan->completed_slots);
+               return 0;
+       }
+
+       mv_xor_free_slots(mv_chan, desc);
+       return 0;
+}
+
+static void __mv_xor_slot_cleanup(struct mv_xor_chan *mv_chan)
+{
+       struct mv_xor_desc_slot *iter, *_iter;
+       dma_cookie_t cookie = 0;
+       int busy = mv_chan_is_busy(mv_chan);
+       u32 current_desc = mv_chan_get_current_desc(mv_chan);
+       int seen_current = 0;
+
+       dev_dbg(mv_chan->device->common.dev, "%s %d\n", __func__, __LINE__);
+       dev_dbg(mv_chan->device->common.dev, "current_desc %x\n", current_desc);
+       mv_xor_clean_completed_slots(mv_chan);
+
+       /* free completed slots from the chain starting with
+        * the oldest descriptor
+        */
+
+       list_for_each_entry_safe(iter, _iter, &mv_chan->chain,
+                                       chain_node) {
+               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
+                */
+               if (iter->async_tx.phys == current_desc) {
+                       seen_current = 1;
+                       if (busy)
+                               break;
+               }
+
+               cookie = mv_xor_run_tx_complete_actions(iter, mv_chan, cookie);
+
+               if (mv_xor_clean_slot(iter, mv_chan))
+                       break;
+       }
+
+       if ((busy == 0) && !list_empty(&mv_chan->chain)) {
+               struct mv_xor_desc_slot *chain_head;
+               chain_head = list_entry(mv_chan->chain.next,
+                                       struct mv_xor_desc_slot,
+                                       chain_node);
+
+               mv_xor_start_new_chain(mv_chan, chain_head);
+       }
+
+       if (cookie > 0)
+               mv_chan->completed_cookie = cookie;
+}
+
+static void
+mv_xor_slot_cleanup(struct mv_xor_chan *mv_chan)
+{
+       spin_lock_bh(&mv_chan->lock);
+       __mv_xor_slot_cleanup(mv_chan);
+       spin_unlock_bh(&mv_chan->lock);
+}
+
+static void mv_xor_tasklet(unsigned long data)
+{
+       struct mv_xor_chan *chan = (struct mv_xor_chan *) data;
+       __mv_xor_slot_cleanup(chan);
+}
+
+static struct mv_xor_desc_slot *
+mv_xor_alloc_slots(struct mv_xor_chan *mv_chan, int num_slots,
+                   int slots_per_op)
+{
+       struct mv_xor_desc_slot *iter, *_iter, *alloc_start = NULL;
+       LIST_HEAD(chain);
+       int slots_found, retry = 0;
+
+       /* 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 = mv_chan->last_used;
+       else
+               iter = list_entry(&mv_chan->all_slots,
+                       struct mv_xor_desc_slot,
+                       slot_node);
+
+       list_for_each_entry_safe_continue(
+               iter, _iter, &mv_chan->all_slots, slot_node) {
+               prefetch(_iter);
+               prefetch(&_iter->async_tx);
+               if (iter->slots_per_op) {
+                       /* give up after finding the first busy slot
+                        * on the second pass through the list
+                        */
+                       if (retry)
+                               break;
+
+                       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 mv_xor_desc_slot *alloc_tail = NULL;
+                       struct mv_xor_desc_slot *last_used = NULL;
+                       iter = alloc_start;
+                       while (num_slots) {
+                               int i;
+
+                               /* pre-ack all but the last descriptor */
+                               async_tx_ack(&iter->async_tx);
+
+                               list_add_tail(&iter->chain_node, &chain);
+                               alloc_tail = iter;
+                               iter->async_tx.cookie = 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 mv_xor_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->async_tx.tx_list);
+                       mv_chan->last_used = last_used;
+                       mv_desc_clear_next_desc(alloc_start);
+                       mv_desc_clear_next_desc(alloc_tail);
+                       return alloc_tail;
+               }
+       }
+       if (!retry++)
+               goto retry;
+
+       /* try to free some slots if the allocation fails */
+       tasklet_schedule(&mv_chan->irq_tasklet);
+
+       return NULL;
+}
+
+static dma_cookie_t
+mv_desc_assign_cookie(struct mv_xor_chan *mv_chan,
+                     struct mv_xor_desc_slot *desc)
+{
+       dma_cookie_t cookie = mv_chan->common.cookie;
+
+       if (++cookie < 0)
+               cookie = 1;
+       mv_chan->common.cookie = desc->async_tx.cookie = cookie;
+       return cookie;
+}
+
+/************************ DMA engine API functions ****************************/
+static dma_cookie_t
+mv_xor_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+       struct mv_xor_desc_slot *sw_desc = to_mv_xor_slot(tx);
+       struct mv_xor_chan *mv_chan = to_mv_xor_chan(tx->chan);
+       struct mv_xor_desc_slot *grp_start, *old_chain_tail;
+       dma_cookie_t cookie;
+       int new_hw_chain = 1;
+
+       dev_dbg(mv_chan->device->common.dev,
+               "%s sw_desc %p: async_tx %p\n",
+               __func__, sw_desc, &sw_desc->async_tx);
+
+       grp_start = sw_desc->group_head;
+
+       spin_lock_bh(&mv_chan->lock);
+       cookie = mv_desc_assign_cookie(mv_chan, sw_desc);
+
+       if (list_empty(&mv_chan->chain))
+               list_splice_init(&sw_desc->async_tx.tx_list, &mv_chan->chain);
+       else {
+               new_hw_chain = 0;
+
+               old_chain_tail = list_entry(mv_chan->chain.prev,
+                                           struct mv_xor_desc_slot,
+                                           chain_node);
+               list_splice_init(&grp_start->async_tx.tx_list,
+                                &old_chain_tail->chain_node);
+
+               if (!mv_can_chain(grp_start))
+                       goto submit_done;
+
+               dev_dbg(mv_chan->device->common.dev, "Append to last desc %x\n",
+                       old_chain_tail->async_tx.phys);
+
+               /* fix up the hardware chain */
+               mv_desc_set_next_desc(old_chain_tail, grp_start->async_tx.phys);
+
+               /* if the channel is not busy */
+               if (!mv_chan_is_busy(mv_chan)) {
+                       u32 current_desc = mv_chan_get_current_desc(mv_chan);
+                       /*
+                        * and the curren desc is the end of the chain before
+                        * the append, then we need to start the channel
+                        */
+                       if (current_desc == old_chain_tail->async_tx.phys)
+                               new_hw_chain = 1;
+               }
+       }
+
+       if (new_hw_chain)
+               mv_xor_start_new_chain(mv_chan, grp_start);
+
+submit_done:
+       spin_unlock_bh(&mv_chan->lock);
+
+       return cookie;
+}
+
+/* returns the number of allocated descriptors */
+static int mv_xor_alloc_chan_resources(struct dma_chan *chan)
+{
+       char *hw_desc;
+       int idx;
+       struct mv_xor_chan *mv_chan = to_mv_xor_chan(chan);
+       struct mv_xor_desc_slot *slot = NULL;
+       struct mv_xor_platform_data *plat_data =
+               mv_chan->device->pdev->dev.platform_data;
+       int num_descs_in_pool = plat_data->pool_size/MV_XOR_SLOT_SIZE;
+
+       /* Allocate descriptor slots */
+       idx = mv_chan->slots_allocated;
+       while (idx < num_descs_in_pool) {
+               slot = kzalloc(sizeof(*slot), GFP_KERNEL);
+               if (!slot) {
+                       printk(KERN_INFO "MV XOR Channel only initialized"
+                               " %d descriptor slots", idx);
+                       break;
+               }
+               hw_desc = (char *) mv_chan->device->dma_desc_pool_virt;
+               slot->hw_desc = (void *) &hw_desc[idx * MV_XOR_SLOT_SIZE];
+
+               dma_async_tx_descriptor_init(&slot->async_tx, chan);
+               slot->async_tx.tx_submit = mv_xor_tx_submit;
+               INIT_LIST_HEAD(&slot->chain_node);
+               INIT_LIST_HEAD(&slot->slot_node);
+               INIT_LIST_HEAD(&slot->async_tx.tx_list);
+               hw_desc = (char *) mv_chan->device->dma_desc_pool;
+               slot->async_tx.phys =
+                       (dma_addr_t) &hw_desc[idx * MV_XOR_SLOT_SIZE];
+               slot->idx = idx++;
+
+               spin_lock_bh(&mv_chan->lock);
+               mv_chan->slots_allocated = idx;
+               list_add_tail(&slot->slot_node, &mv_chan->all_slots);
+               spin_unlock_bh(&mv_chan->lock);
+       }
+
+       if (mv_chan->slots_allocated && !mv_chan->last_used)
+               mv_chan->last_used = list_entry(mv_chan->all_slots.next,
+                                       struct mv_xor_desc_slot,
+                                       slot_node);
+
+       dev_dbg(mv_chan->device->common.dev,
+               "allocated %d descriptor slots last_used: %p\n",
+               mv_chan->slots_allocated, mv_chan->last_used);
+
+       return mv_chan->slots_allocated ? : -ENOMEM;
+}
+
+static struct dma_async_tx_descriptor *
+mv_xor_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src,
+               size_t len, unsigned long flags)
+{
+       struct mv_xor_chan *mv_chan = to_mv_xor_chan(chan);
+       struct mv_xor_desc_slot *sw_desc, *grp_start;
+       int slot_cnt;
+
+       dev_dbg(mv_chan->device->common.dev,
+               "%s dest: %x src %x len: %u flags: %ld\n",
+               __func__, dest, src, len, flags);
+       if (unlikely(len < MV_XOR_MIN_BYTE_COUNT))
+               return NULL;
+
+       BUG_ON(unlikely(len > MV_XOR_MAX_BYTE_COUNT));
+
+       spin_lock_bh(&mv_chan->lock);
+       slot_cnt = mv_chan_memcpy_slot_count(len);
+       sw_desc = mv_xor_alloc_slots(mv_chan, slot_cnt, 1);
+       if (sw_desc) {
+               sw_desc->type = DMA_MEMCPY;
+               sw_desc->async_tx.flags = flags;
+               grp_start = sw_desc->group_head;
+               mv_desc_init(grp_start, flags);
+               mv_desc_set_byte_count(grp_start, len);
+               mv_desc_set_dest_addr(sw_desc->group_head, dest);
+               mv_desc_set_src_addr(grp_start, 0, src);
+               sw_desc->unmap_src_cnt = 1;
+               sw_desc->unmap_len = len;
+       }
+       spin_unlock_bh(&mv_chan->lock);
+
+       dev_dbg(mv_chan->device->common.dev,
+               "%s sw_desc %p async_tx %p\n",
+               __func__, sw_desc, sw_desc ? &sw_desc->async_tx : 0);
+
+       return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+static struct dma_async_tx_descriptor *
+mv_xor_prep_dma_memset(struct dma_chan *chan, dma_addr_t dest, int value,
+                      size_t len, unsigned long flags)
+{
+       struct mv_xor_chan *mv_chan = to_mv_xor_chan(chan);
+       struct mv_xor_desc_slot *sw_desc, *grp_start;
+       int slot_cnt;
+
+       dev_dbg(mv_chan->device->common.dev,
+               "%s dest: %x len: %u flags: %ld\n",
+               __func__, dest, len, flags);
+       if (unlikely(len < MV_XOR_MIN_BYTE_COUNT))
+               return NULL;
+
+       BUG_ON(unlikely(len > MV_XOR_MAX_BYTE_COUNT));
+
+       spin_lock_bh(&mv_chan->lock);
+       slot_cnt = mv_chan_memset_slot_count(len);
+       sw_desc = mv_xor_alloc_slots(mv_chan, slot_cnt, 1);
+       if (sw_desc) {
+               sw_desc->type = DMA_MEMSET;
+               sw_desc->async_tx.flags = flags;
+               grp_start = sw_desc->group_head;
+               mv_desc_init(grp_start, flags);
+               mv_desc_set_byte_count(grp_start, len);
+               mv_desc_set_dest_addr(sw_desc->group_head, dest);
+               mv_desc_set_block_fill_val(grp_start, value);
+               sw_desc->unmap_src_cnt = 1;
+               sw_desc->unmap_len = len;
+       }
+       spin_unlock_bh(&mv_chan->lock);
+       dev_dbg(mv_chan->device->common.dev,
+               "%s sw_desc %p async_tx %p \n",
+               __func__, sw_desc, &sw_desc->async_tx);
+       return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+static struct dma_async_tx_descriptor *
+mv_xor_prep_dma_xor(struct dma_chan *chan, dma_addr_t dest, dma_addr_t *src,
+                   unsigned int src_cnt, size_t len, unsigned long flags)
+{
+       struct mv_xor_chan *mv_chan = to_mv_xor_chan(chan);
+       struct mv_xor_desc_slot *sw_desc, *grp_start;
+       int slot_cnt;
+
+       if (unlikely(len < MV_XOR_MIN_BYTE_COUNT))
+               return NULL;
+
+       BUG_ON(unlikely(len > MV_XOR_MAX_BYTE_COUNT));
+
+       dev_dbg(mv_chan->device->common.dev,
+               "%s src_cnt: %d len: dest %x %u flags: %ld\n",
+               __func__, src_cnt, len, dest, flags);
+
+       spin_lock_bh(&mv_chan->lock);
+       slot_cnt = mv_chan_xor_slot_count(len, src_cnt);
+       sw_desc = mv_xor_alloc_slots(mv_chan, slot_cnt, 1);
+       if (sw_desc) {
+               sw_desc->type = DMA_XOR;
+               sw_desc->async_tx.flags = flags;
+               grp_start = sw_desc->group_head;
+               mv_desc_init(grp_start, flags);
+               /* the byte count field is the same as in memcpy desc*/
+               mv_desc_set_byte_count(grp_start, len);
+               mv_desc_set_dest_addr(sw_desc->group_head, dest);
+               sw_desc->unmap_src_cnt = src_cnt;
+               sw_desc->unmap_len = len;
+               while (src_cnt--)
+                       mv_desc_set_src_addr(grp_start, src_cnt, src[src_cnt]);
+       }
+       spin_unlock_bh(&mv_chan->lock);
+       dev_dbg(mv_chan->device->common.dev,
+               "%s sw_desc %p async_tx %p \n",
+               __func__, sw_desc, &sw_desc->async_tx);
+       return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+static void mv_xor_free_chan_resources(struct dma_chan *chan)
+{
+       struct mv_xor_chan *mv_chan = to_mv_xor_chan(chan);
+       struct mv_xor_desc_slot *iter, *_iter;
+       int in_use_descs = 0;
+
+       mv_xor_slot_cleanup(mv_chan);
+
+       spin_lock_bh(&mv_chan->lock);
+       list_for_each_entry_safe(iter, _iter, &mv_chan->chain,
+                                       chain_node) {
+               in_use_descs++;
+               list_del(&iter->chain_node);
+       }
+       list_for_each_entry_safe(iter, _iter, &mv_chan->completed_slots,
+                                completed_node) {
+               in_use_descs++;
+               list_del(&iter->completed_node);
+       }
+       list_for_each_entry_safe_reverse(
+               iter, _iter, &mv_chan->all_slots, slot_node) {
+               list_del(&iter->slot_node);
+               kfree(iter);
+               mv_chan->slots_allocated--;
+       }
+       mv_chan->last_used = NULL;
+
+       dev_dbg(mv_chan->device->common.dev, "%s slots_allocated %d\n",
+               __func__, mv_chan->slots_allocated);
+       spin_unlock_bh(&mv_chan->lock);
+
+       if (in_use_descs)
+               dev_err(mv_chan->device->common.dev,
+                       "freeing %d in use descriptors!\n", in_use_descs);
+}
+
+/**
+ * mv_xor_is_complete - poll the status of an XOR transaction
+ * @chan: XOR channel handle
+ * @cookie: XOR transaction identifier
+ */
+static enum dma_status mv_xor_is_complete(struct dma_chan *chan,
+                                         dma_cookie_t cookie,
+                                         dma_cookie_t *done,
+                                         dma_cookie_t *used)
+{
+       struct mv_xor_chan *mv_chan = to_mv_xor_chan(chan);
+       dma_cookie_t last_used;
+       dma_cookie_t last_complete;
+       enum dma_status ret;
+
+       last_used = chan->cookie;
+       last_complete = mv_chan->completed_cookie;
+       mv_chan->is_complete_cookie = cookie;
+       if (done)
+               *done = last_complete;
+       if (used)
+               *used = last_used;
+
+       ret = dma_async_is_complete(cookie, last_complete, last_used);
+       if (ret == DMA_SUCCESS) {
+               mv_xor_clean_completed_slots(mv_chan);
+               return ret;
+       }
+       mv_xor_slot_cleanup(mv_chan);
+
+       last_used = chan->cookie;
+       last_complete = mv_chan->completed_cookie;
+
+       if (done)
+               *done = last_complete;
+       if (used)
+               *used = last_used;
+
+       return dma_async_is_complete(cookie, last_complete, last_used);
+}
+
+static void mv_dump_xor_regs(struct mv_xor_chan *chan)
+{
+       u32 val;
+
+       val = __raw_readl(XOR_CONFIG(chan));
+       dev_printk(KERN_ERR, chan->device->common.dev,
+                  "config       0x%08x.\n", val);
+
+       val = __raw_readl(XOR_ACTIVATION(chan));
+       dev_printk(KERN_ERR, chan->device->common.dev,
+                  "activation   0x%08x.\n", val);
+
+       val = __raw_readl(XOR_INTR_CAUSE(chan));
+       dev_printk(KERN_ERR, chan->device->common.dev,
+                  "intr cause   0x%08x.\n", val);
+
+       val = __raw_readl(XOR_INTR_MASK(chan));
+       dev_printk(KERN_ERR, chan->device->common.dev,
+                  "intr mask    0x%08x.\n", val);
+
+       val = __raw_readl(XOR_ERROR_CAUSE(chan));
+       dev_printk(KERN_ERR, chan->device->common.dev,
+                  "error cause  0x%08x.\n", val);
+
+       val = __raw_readl(XOR_ERROR_ADDR(chan));
+       dev_printk(KERN_ERR, chan->device->common.dev,
+                  "error addr   0x%08x.\n", val);
+}
+
+static void mv_xor_err_interrupt_handler(struct mv_xor_chan *chan,
+                                        u32 intr_cause)
+{
+       if (intr_cause & (1 << 4)) {
+            dev_dbg(chan->device->common.dev,
+                    "ignore this error\n");
+            return;
+       }
+
+       dev_printk(KERN_ERR, chan->device->common.dev,
+                  "error on chan %d. intr cause 0x%08x.\n",
+                  chan->idx, intr_cause);
+
+       mv_dump_xor_regs(chan);
+       BUG();
+}
+
+static irqreturn_t mv_xor_interrupt_handler(int irq, void *data)
+{
+       struct mv_xor_chan *chan = data;
+       u32 intr_cause = mv_chan_get_intr_cause(chan);
+
+       dev_dbg(chan->device->common.dev, "intr cause %x\n", intr_cause);
+
+       if (mv_is_err_intr(intr_cause))
+               mv_xor_err_interrupt_handler(chan, intr_cause);
+
+       tasklet_schedule(&chan->irq_tasklet);
+
+       mv_xor_device_clear_eoc_cause(chan);
+
+       return IRQ_HANDLED;
+}
+
+static void mv_xor_issue_pending(struct dma_chan *chan)
+{
+       struct mv_xor_chan *mv_chan = to_mv_xor_chan(chan);
+
+       if (mv_chan->pending >= MV_XOR_THRESHOLD) {
+               mv_chan->pending = 0;
+               mv_chan_activate(mv_chan);
+       }
+}
+
+/*
+ * Perform a transaction to verify the HW works.
+ */
+#define MV_XOR_TEST_SIZE 2000
+
+static int __devinit mv_xor_memcpy_self_test(struct mv_xor_device *device)
+{
+       int i;
+       void *src, *dest;
+       dma_addr_t src_dma, dest_dma;
+       struct dma_chan *dma_chan;
+       dma_cookie_t cookie;
+       struct dma_async_tx_descriptor *tx;
+       int err = 0;
+       struct mv_xor_chan *mv_chan;
+
+       src = kmalloc(sizeof(u8) * MV_XOR_TEST_SIZE, GFP_KERNEL);
+       if (!src)
+               return -ENOMEM;
+
+       dest = kzalloc(sizeof(u8) * MV_XOR_TEST_SIZE, GFP_KERNEL);
+       if (!dest) {
+               kfree(src);
+               return -ENOMEM;
+       }
+
+       /* Fill in src buffer */
+       for (i = 0; i < MV_XOR_TEST_SIZE; i++)
+               ((u8 *) src)[i] = (u8)i;
+
+       /* Start copy, using first DMA channel */
+       dma_chan = container_of(device->common.channels.next,
+                               struct dma_chan,
+                               device_node);
+       if (mv_xor_alloc_chan_resources(dma_chan) < 1) {
+               err = -ENODEV;
+               goto out;
+       }
+
+       dest_dma = dma_map_single(dma_chan->device->dev, dest,
+                                 MV_XOR_TEST_SIZE, DMA_FROM_DEVICE);
+
+       src_dma = dma_map_single(dma_chan->device->dev, src,
+                                MV_XOR_TEST_SIZE, DMA_TO_DEVICE);
+
+       tx = mv_xor_prep_dma_memcpy(dma_chan, dest_dma, src_dma,
+                                   MV_XOR_TEST_SIZE, 0);
+       cookie = mv_xor_tx_submit(tx);
+       mv_xor_issue_pending(dma_chan);
+       async_tx_ack(tx);
+       msleep(1);
+
+       if (mv_xor_is_complete(dma_chan, cookie, NULL, NULL) !=
+           DMA_SUCCESS) {
+               dev_printk(KERN_ERR, dma_chan->device->dev,
+                          "Self-test copy timed out, disabling\n");
+               err = -ENODEV;
+               goto free_resources;
+       }
+
+       mv_chan = to_mv_xor_chan(dma_chan);
+       dma_sync_single_for_cpu(&mv_chan->device->pdev->dev, dest_dma,
+                               MV_XOR_TEST_SIZE, DMA_FROM_DEVICE);
+       if (memcmp(src, dest, MV_XOR_TEST_SIZE)) {
+               dev_printk(KERN_ERR, dma_chan->device->dev,
+                          "Self-test copy failed compare, disabling\n");
+               err = -ENODEV;
+               goto free_resources;
+       }
+
+free_resources:
+       mv_xor_free_chan_resources(dma_chan);
+out:
+       kfree(src);
+       kfree(dest);
+       return err;
+}
+
+#define MV_XOR_NUM_SRC_TEST 4 /* must be <= 15 */
+static int __devinit
+mv_xor_xor_self_test(struct mv_xor_device *device)
+{
+       int i, src_idx;
+       struct page *dest;
+       struct page *xor_srcs[MV_XOR_NUM_SRC_TEST];
+       dma_addr_t dma_srcs[MV_XOR_NUM_SRC_TEST];
+       dma_addr_t dest_dma;
+       struct dma_async_tx_descriptor *tx;
+       struct dma_chan *dma_chan;
+       dma_cookie_t cookie;
+       u8 cmp_byte = 0;
+       u32 cmp_word;
+       int err = 0;
+       struct mv_xor_chan *mv_chan;
+
+       for (src_idx = 0; src_idx < MV_XOR_NUM_SRC_TEST; src_idx++) {
+               xor_srcs[src_idx] = alloc_page(GFP_KERNEL);
+               if (!xor_srcs[src_idx])
+                       while (src_idx--) {
+                               __free_page(xor_srcs[src_idx]);
+                               return -ENOMEM;
+                       }
+       }
+
+       dest = alloc_page(GFP_KERNEL);
+       if (!dest)
+               while (src_idx--) {
+                       __free_page(xor_srcs[src_idx]);
+                       return -ENOMEM;
+               }
+
+       /* Fill in src buffers */
+       for (src_idx = 0; src_idx < MV_XOR_NUM_SRC_TEST; src_idx++) {
+               u8 *ptr = page_address(xor_srcs[src_idx]);
+               for (i = 0; i < PAGE_SIZE; i++)
+                       ptr[i] = (1 << src_idx);
+       }
+
+       for (src_idx = 0; src_idx < MV_XOR_NUM_SRC_TEST; src_idx++)
+               cmp_byte ^= (u8) (1 << src_idx);
+
+       cmp_word = (cmp_byte << 24) | (cmp_byte << 16) |
+               (cmp_byte << 8) | cmp_byte;
+
+       memset(page_address(dest), 0, PAGE_SIZE);
+
+       dma_chan = container_of(device->common.channels.next,
+                               struct dma_chan,
+                               device_node);
+       if (mv_xor_alloc_chan_resources(dma_chan) < 1) {
+               err = -ENODEV;
+               goto out;
+       }
+
+       /* test xor */
+       dest_dma = dma_map_page(dma_chan->device->dev, dest, 0, PAGE_SIZE,
+                               DMA_FROM_DEVICE);
+
+       for (i = 0; i < MV_XOR_NUM_SRC_TEST; i++)
+               dma_srcs[i] = dma_map_page(dma_chan->device->dev, xor_srcs[i],
+                                          0, PAGE_SIZE, DMA_TO_DEVICE);
+
+       tx = mv_xor_prep_dma_xor(dma_chan, dest_dma, dma_srcs,
+                                MV_XOR_NUM_SRC_TEST, PAGE_SIZE, 0);
+
+       cookie = mv_xor_tx_submit(tx);
+       mv_xor_issue_pending(dma_chan);
+       async_tx_ack(tx);
+       msleep(8);
+
+       if (mv_xor_is_complete(dma_chan, cookie, NULL, NULL) !=
+           DMA_SUCCESS) {
+               dev_printk(KERN_ERR, dma_chan->device->dev,
+                          "Self-test xor timed out, disabling\n");
+               err = -ENODEV;
+               goto free_resources;
+       }
+
+       mv_chan = to_mv_xor_chan(dma_chan);
+       dma_sync_single_for_cpu(&mv_chan->device->pdev->dev, dest_dma,
+                               PAGE_SIZE, DMA_FROM_DEVICE);
+       for (i = 0; i < (PAGE_SIZE / sizeof(u32)); i++) {
+               u32 *ptr = page_address(dest);
+               if (ptr[i] != cmp_word) {
+                       dev_printk(KERN_ERR, dma_chan->device->dev,
+                                  "Self-test xor failed compare, disabling."
+                                  " index %d, data %x, expected %x\n", i,
+                                  ptr[i], cmp_word);
+                       err = -ENODEV;
+                       goto free_resources;
+               }
+       }
+
+free_resources:
+       mv_xor_free_chan_resources(dma_chan);
+out:
+       src_idx = MV_XOR_NUM_SRC_TEST;
+       while (src_idx--)
+               __free_page(xor_srcs[src_idx]);
+       __free_page(dest);
+       return err;
+}
+
+static int __devexit mv_xor_remove(struct platform_device *dev)
+{
+       struct mv_xor_device *device = platform_get_drvdata(dev);
+       struct dma_chan *chan, *_chan;
+       struct mv_xor_chan *mv_chan;
+       struct mv_xor_platform_data *plat_data = dev->dev.platform_data;
+
+       dma_async_device_unregister(&device->common);
+
+       dma_free_coherent(&dev->dev, plat_data->pool_size,
+                       device->dma_desc_pool_virt, device->dma_desc_pool);
+
+       list_for_each_entry_safe(chan, _chan, &device->common.channels,
+                               device_node) {
+               mv_chan = to_mv_xor_chan(chan);
+               list_del(&chan->device_node);
+       }
+
+       return 0;
+}
+
+static int __devinit mv_xor_probe(struct platform_device *pdev)
+{
+       int ret = 0;
+       int irq;
+       struct mv_xor_device *adev;
+       struct mv_xor_chan *mv_chan;
+       struct dma_device *dma_dev;
+       struct mv_xor_platform_data *plat_data = pdev->dev.platform_data;
+
+
+       adev = devm_kzalloc(&pdev->dev, sizeof(*adev), GFP_KERNEL);
+       if (!adev)
+               return -ENOMEM;
+
+       dma_dev = &adev->common;
+
+       /* allocate coherent memory for hardware descriptors
+        * note: writecombine gives slightly better performance, but
+        * requires that we explicitly flush the writes
+        */
+       adev->dma_desc_pool_virt = dma_alloc_writecombine(&pdev->dev,
+                                                         plat_data->pool_size,
+                                                         &adev->dma_desc_pool,
+                                                         GFP_KERNEL);
+       if (!adev->dma_desc_pool_virt)
+               return -ENOMEM;
+
+       adev->id = plat_data->hw_id;
+
+       /* discover transaction capabilites from the platform data */
+       dma_dev->cap_mask = plat_data->cap_mask;
+       adev->pdev = pdev;
+       platform_set_drvdata(pdev, adev);
+
+       adev->shared = platform_get_drvdata(plat_data->shared);
+
+       INIT_LIST_HEAD(&dma_dev->channels);
+
+       /* set base routines */
+       dma_dev->device_alloc_chan_resources = mv_xor_alloc_chan_resources;
+       dma_dev->device_free_chan_resources = mv_xor_free_chan_resources;
+       dma_dev->device_is_tx_complete = mv_xor_is_complete;
+       dma_dev->device_issue_pending = mv_xor_issue_pending;
+       dma_dev->dev = &pdev->dev;
+
+       /* set prep routines based on capability */
+       if (dma_has_cap(DMA_MEMCPY, dma_dev->cap_mask))
+               dma_dev->device_prep_dma_memcpy = mv_xor_prep_dma_memcpy;
+       if (dma_has_cap(DMA_MEMSET, dma_dev->cap_mask))
+               dma_dev->device_prep_dma_memset = mv_xor_prep_dma_memset;
+       if (dma_has_cap(DMA_XOR, dma_dev->cap_mask)) {
+               dma_dev->max_xor = 8;                  ;
+               dma_dev->device_prep_dma_xor = mv_xor_prep_dma_xor;
+       }
+
+       mv_chan = devm_kzalloc(&pdev->dev, sizeof(*mv_chan), GFP_KERNEL);
+       if (!mv_chan) {
+               ret = -ENOMEM;
+               goto err_free_dma;
+       }
+       mv_chan->device = adev;
+       mv_chan->idx = plat_data->hw_id;
+       mv_chan->mmr_base = adev->shared->xor_base;
+
+       if (!mv_chan->mmr_base) {
+               ret = -ENOMEM;
+               goto err_free_dma;
+       }
+       tasklet_init(&mv_chan->irq_tasklet, mv_xor_tasklet, (unsigned long)
+                    mv_chan);
+
+       /* clear errors before enabling interrupts */
+       mv_xor_device_clear_err_status(mv_chan);
+
+       irq = platform_get_irq(pdev, 0);
+       if (irq < 0) {
+               ret = irq;
+               goto err_free_dma;
+       }
+       ret = devm_request_irq(&pdev->dev, irq,
+                              mv_xor_interrupt_handler,
+                              0, dev_name(&pdev->dev), mv_chan);
+       if (ret)
+               goto err_free_dma;
+
+       mv_chan_unmask_interrupts(mv_chan);
+
+       mv_set_mode(mv_chan, DMA_MEMCPY);
+
+       spin_lock_init(&mv_chan->lock);
+       INIT_LIST_HEAD(&mv_chan->chain);
+       INIT_LIST_HEAD(&mv_chan->completed_slots);
+       INIT_LIST_HEAD(&mv_chan->all_slots);
+       INIT_RCU_HEAD(&mv_chan->common.rcu);
+       mv_chan->common.device = dma_dev;
+
+       list_add_tail(&mv_chan->common.device_node, &dma_dev->channels);
+
+       if (dma_has_cap(DMA_MEMCPY, dma_dev->cap_mask)) {
+               ret = mv_xor_memcpy_self_test(adev);
+               dev_dbg(&pdev->dev, "memcpy self test returned %d\n", ret);
+               if (ret)
+                       goto err_free_dma;
+       }
+
+       if (dma_has_cap(DMA_XOR, dma_dev->cap_mask)) {
+               ret = mv_xor_xor_self_test(adev);
+               dev_dbg(&pdev->dev, "xor self test returned %d\n", ret);
+               if (ret)
+                       goto err_free_dma;
+       }
+
+       dev_printk(KERN_INFO, &pdev->dev, "Marvell XOR: "
+         "( %s%s%s%s)\n",
+         dma_has_cap(DMA_XOR, dma_dev->cap_mask) ? "xor " : "",
+         dma_has_cap(DMA_MEMSET, dma_dev->cap_mask)  ? "fill " : "",
+         dma_has_cap(DMA_MEMCPY, dma_dev->cap_mask) ? "cpy " : "",
+         dma_has_cap(DMA_INTERRUPT, dma_dev->cap_mask) ? "intr " : "");
+
+       dma_async_device_register(dma_dev);
+       goto out;
+
+ err_free_dma:
+       dma_free_coherent(&adev->pdev->dev, plat_data->pool_size,
+                       adev->dma_desc_pool_virt, adev->dma_desc_pool);
+ out:
+       return ret;
+}
+
+static void
+mv_xor_conf_mbus_windows(struct mv_xor_shared_private *msp,
+                        struct mbus_dram_target_info *dram)
+{
+       void __iomem *base = msp->xor_base;
+       u32 win_enable = 0;
+       int i;
+
+       for (i = 0; i < 8; i++) {
+               writel(0, base + WINDOW_BASE(i));
+               writel(0, base + WINDOW_SIZE(i));
+               if (i < 4)
+                       writel(0, base + WINDOW_REMAP_HIGH(i));
+       }
+
+       for (i = 0; i < dram->num_cs; i++) {
+               struct mbus_dram_window *cs = dram->cs + i;
+
+               writel((cs->base & 0xffff0000) |
+                      (cs->mbus_attr << 8) |
+                      dram->mbus_dram_target_id, base + WINDOW_BASE(i));
+               writel((cs->size - 1) & 0xffff0000, base + WINDOW_SIZE(i));
+
+               win_enable |= (1 << i);
+               win_enable |= 3 << (16 + (2 * i));
+       }
+
+       writel(win_enable, base + WINDOW_BAR_ENABLE(0));
+       writel(win_enable, base + WINDOW_BAR_ENABLE(1));
+}
+
+static struct platform_driver mv_xor_driver = {
+       .probe          = mv_xor_probe,
+       .remove         = mv_xor_remove,
+       .driver         = {
+               .owner  = THIS_MODULE,
+               .name   = MV_XOR_NAME,
+       },
+};
+
+static int mv_xor_shared_probe(struct platform_device *pdev)
+{
+       struct mv_xor_platform_shared_data *msd = pdev->dev.platform_data;
+       struct mv_xor_shared_private *msp;
+       struct resource *res;
+
+       dev_printk(KERN_NOTICE, &pdev->dev, "Marvell shared XOR driver\n");
+
+       msp = devm_kzalloc(&pdev->dev, sizeof(*msp), GFP_KERNEL);
+       if (!msp)
+               return -ENOMEM;
+
+       res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+       if (!res)
+               return -ENODEV;
+
+       msp->xor_base = devm_ioremap(&pdev->dev, res->start,
+                                    res->end - res->start + 1);
+       if (!msp->xor_base)
+               return -EBUSY;
+
+       res = platform_get_resource(pdev, IORESOURCE_MEM, 1);
+       if (!res)
+               return -ENODEV;
+
+       msp->xor_high_base = devm_ioremap(&pdev->dev, res->start,
+                                         res->end - res->start + 1);
+       if (!msp->xor_high_base)
+               return -EBUSY;
+
+       platform_set_drvdata(pdev, msp);
+
+       /*
+        * (Re-)program MBUS remapping windows if we are asked to.
+        */
+       if (msd != NULL && msd->dram != NULL)
+               mv_xor_conf_mbus_windows(msp, msd->dram);
+
+       return 0;
+}
+
+static int mv_xor_shared_remove(struct platform_device *pdev)
+{
+       return 0;
+}
+
+static struct platform_driver mv_xor_shared_driver = {
+       .probe          = mv_xor_shared_probe,
+       .remove         = mv_xor_shared_remove,
+       .driver         = {
+               .owner  = THIS_MODULE,
+               .name   = MV_XOR_SHARED_NAME,
+       },
+};
+
+
+static int __init mv_xor_init(void)
+{
+       int rc;
+
+       rc = platform_driver_register(&mv_xor_shared_driver);
+       if (!rc) {
+               rc = platform_driver_register(&mv_xor_driver);
+               if (rc)
+                       platform_driver_unregister(&mv_xor_shared_driver);
+       }
+       return rc;
+}
+module_init(mv_xor_init);
+
+/* it's currently unsafe to unload this module */
+#if 0
+static void __exit mv_xor_exit(void)
+{
+       platform_driver_unregister(&mv_xor_driver);
+       platform_driver_unregister(&mv_xor_shared_driver);
+       return;
+}
+
+module_exit(mv_xor_exit);
+#endif
+
+MODULE_AUTHOR("Saeed Bishara <saeed@marvell.com>");
+MODULE_DESCRIPTION("DMA engine driver for Marvell's XOR engine");
+MODULE_LICENSE("GPL");
diff --git a/drivers/dma/mv_xor.h b/drivers/dma/mv_xor.h
new file mode 100644 (file)
index 0000000..06cafe1
--- /dev/null
@@ -0,0 +1,183 @@
+/*
+ * Copyright (C) 2007, 2008, Marvell International Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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., 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+ */
+
+#ifndef MV_XOR_H
+#define MV_XOR_H
+
+#include <linux/types.h>
+#include <linux/io.h>
+#include <linux/dmaengine.h>
+#include <linux/interrupt.h>
+
+#define USE_TIMER
+#define MV_XOR_SLOT_SIZE               64
+#define MV_XOR_THRESHOLD               1
+
+#define XOR_OPERATION_MODE_XOR         0
+#define XOR_OPERATION_MODE_MEMCPY      2
+#define XOR_OPERATION_MODE_MEMSET      4
+
+#define XOR_CURR_DESC(chan)    (chan->mmr_base + 0x210 + (chan->idx * 4))
+#define XOR_NEXT_DESC(chan)    (chan->mmr_base + 0x200 + (chan->idx * 4))
+#define XOR_BYTE_COUNT(chan)   (chan->mmr_base + 0x220 + (chan->idx * 4))
+#define XOR_DEST_POINTER(chan) (chan->mmr_base + 0x2B0 + (chan->idx * 4))
+#define XOR_BLOCK_SIZE(chan)   (chan->mmr_base + 0x2C0 + (chan->idx * 4))
+#define XOR_INIT_VALUE_LOW(chan)       (chan->mmr_base + 0x2E0)
+#define XOR_INIT_VALUE_HIGH(chan)      (chan->mmr_base + 0x2E4)
+
+#define XOR_CONFIG(chan)       (chan->mmr_base + 0x10 + (chan->idx * 4))
+#define XOR_ACTIVATION(chan)   (chan->mmr_base + 0x20 + (chan->idx * 4))
+#define XOR_INTR_CAUSE(chan)   (chan->mmr_base + 0x30)
+#define XOR_INTR_MASK(chan)    (chan->mmr_base + 0x40)
+#define XOR_ERROR_CAUSE(chan)  (chan->mmr_base + 0x50)
+#define XOR_ERROR_ADDR(chan)   (chan->mmr_base + 0x60)
+#define XOR_INTR_MASK_VALUE    0x3F5
+
+#define WINDOW_BASE(w)         (0x250 + ((w) << 2))
+#define WINDOW_SIZE(w)         (0x270 + ((w) << 2))
+#define WINDOW_REMAP_HIGH(w)   (0x290 + ((w) << 2))
+#define WINDOW_BAR_ENABLE(chan)        (0x240 + ((chan) << 2))
+
+struct mv_xor_shared_private {
+       void __iomem    *xor_base;
+       void __iomem    *xor_high_base;
+};
+
+
+/**
+ * struct mv_xor_device - internal representation of a XOR device
+ * @pdev: Platform device
+ * @id: HW XOR Device selector
+ * @dma_desc_pool: base of DMA descriptor region (DMA address)
+ * @dma_desc_pool_virt: base of DMA descriptor region (CPU address)
+ * @common: embedded struct dma_device
+ */
+struct mv_xor_device {
+       struct platform_device          *pdev;
+       int                             id;
+       dma_addr_t                      dma_desc_pool;
+       void                            *dma_desc_pool_virt;
+       struct dma_device               common;
+       struct mv_xor_shared_private    *shared;
+};
+
+/**
+ * struct mv_xor_chan - internal representation of a XOR channel
+ * @pending: allows batching of hardware operations
+ * @completed_cookie: identifier for the most recently completed operation
+ * @lock: serializes enqueue/dequeue operations to the descriptors pool
+ * @mmr_base: memory mapped register base
+ * @idx: the index of the xor channel
+ * @chain: device chain view of the descriptors
+ * @completed_slots: slots completed by HW but still need to be acked
+ * @device: parent device
+ * @common: common dmaengine channel object members
+ * @last_used: place holder for allocation to continue from where it left off
+ * @all_slots: complete domain of slots usable by the channel
+ * @slots_allocated: records the actual size of the descriptor slot pool
+ * @irq_tasklet: bottom half where mv_xor_slot_cleanup runs
+ */
+struct mv_xor_chan {
+       int                     pending;
+       dma_cookie_t            completed_cookie;
+       spinlock_t              lock; /* protects the descriptor slot pool */
+       void __iomem            *mmr_base;
+       unsigned int            idx;
+       enum dma_transaction_type       current_type;
+       struct list_head        chain;
+       struct list_head        completed_slots;
+       struct mv_xor_device    *device;
+       struct dma_chan         common;
+       struct mv_xor_desc_slot *last_used;
+       struct list_head        all_slots;
+       int                     slots_allocated;
+       struct tasklet_struct   irq_tasklet;
+#ifdef USE_TIMER
+       unsigned long           cleanup_time;
+       u32                     current_on_last_cleanup;
+       dma_cookie_t            is_complete_cookie;
+#endif
+};
+
+/**
+ * struct mv_xor_desc_slot - software descriptor
+ * @slot_node: node on the mv_xor_chan.all_slots list
+ * @chain_node: node on the mv_xor_chan.chain list
+ * @completed_node: node on the mv_xor_chan.completed_slots list
+ * @hw_desc: virtual address of the hardware descriptor chain
+ * @phys: hardware address of the hardware descriptor chain
+ * @group_head: first operation in a transaction
+ * @slot_cnt: total slots used in an transaction (group of operations)
+ * @slots_per_op: number of slots per operation
+ * @idx: pool index
+ * @unmap_src_cnt: number of xor sources
+ * @unmap_len: transaction bytecount
+ * @async_tx: support for the async_tx api
+ * @group_list: list of slots that make up a multi-descriptor transaction
+ *     for example transfer lengths larger than the supported hw max
+ * @xor_check_result: result of zero sum
+ * @crc32_result: result crc calculation
+ */
+struct mv_xor_desc_slot {
+       struct list_head        slot_node;
+       struct list_head        chain_node;
+       struct list_head        completed_node;
+       enum dma_transaction_type       type;
+       void                    *hw_desc;
+       struct mv_xor_desc_slot *group_head;
+       u16                     slot_cnt;
+       u16                     slots_per_op;
+       u16                     idx;
+       u16                     unmap_src_cnt;
+       u32                     value;
+       size_t                  unmap_len;
+       struct dma_async_tx_descriptor  async_tx;
+       union {
+               u32             *xor_check_result;
+               u32             *crc32_result;
+       };
+#ifdef USE_TIMER
+       unsigned long           arrival_time;
+       struct timer_list       timeout;
+#endif
+};
+
+/* This structure describes XOR descriptor size 64bytes        */
+struct mv_xor_desc {
+       u32 status;             /* descriptor execution status */
+       u32 crc32_result;       /* result of CRC-32 calculation */
+       u32 desc_command;       /* type of operation to be carried out */
+       u32 phy_next_desc;      /* next descriptor address pointer */
+       u32 byte_count;         /* size of src/dst blocks in bytes */
+       u32 phy_dest_addr;      /* destination block address */
+       u32 phy_src_addr[8];    /* source block addresses */
+       u32 reserved0;
+       u32 reserved1;
+};
+
+#define to_mv_sw_desc(addr_hw_desc)            \
+       container_of(addr_hw_desc, struct mv_xor_desc_slot, hw_desc)
+
+#define mv_hw_desc_slot_idx(hw_desc, idx)      \
+       ((void *)(((unsigned long)hw_desc) + ((idx) << 5)))
+
+#define MV_XOR_MIN_BYTE_COUNT  (128)
+#define XOR_MAX_BYTE_COUNT     ((16 * 1024 * 1024) - 1)
+#define MV_XOR_MAX_BYTE_COUNT  XOR_MAX_BYTE_COUNT
+
+
+#endif
diff --git a/include/asm-arm/plat-orion/mv_xor.h b/include/asm-arm/plat-orion/mv_xor.h
new file mode 100644 (file)
index 0000000..c349e8f
--- /dev/null
@@ -0,0 +1,28 @@
+/*
+ * Marvell XOR platform device data definition file.
+ */
+
+#ifndef __ASM_PLAT_ORION_MV_XOR_H
+#define __ASM_PLAT_ORION_MV_XOR_H
+
+#include <linux/dmaengine.h>
+#include <linux/mbus.h>
+
+#define MV_XOR_SHARED_NAME     "mv_xor_shared"
+#define MV_XOR_NAME            "mv_xor"
+
+struct mbus_dram_target_info;
+
+struct mv_xor_platform_shared_data {
+       struct mbus_dram_target_info    *dram;
+};
+
+struct mv_xor_platform_data {
+       struct platform_device          *shared;
+       int                             hw_id;
+       dma_cap_mask_t                  cap_mask;
+       size_t                          pool_size;
+};
+
+
+#endif