[SCSI] mvumi: Add Marvell UMI driver
authorJianyun Li <jyli@marvell.com>
Wed, 11 May 2011 15:22:44 +0000 (23:22 +0800)
committerJames Bottomley <JBottomley@Parallels.com>
Sat, 27 Aug 2011 14:36:58 +0000 (08:36 -0600)
The Marvell Universal Message Interface (UMI) defines a messaging
interface between host and Marvell products (Plato, for example). It
considers situations of limited system resource and optimized system
performance.
UMI driver translates host request to message and sends message
to FW via UMI, FW receives message and processes it, then sends response
to UMI driver.
FW generates an interrupt when it needs to send information or
response to UMI driver

Signed-off-by: Jianyun Li <jyli@marvell.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
drivers/scsi/Kconfig
drivers/scsi/Makefile
drivers/scsi/mvumi.c [new file with mode: 0644]
drivers/scsi/mvumi.h [new file with mode: 0644]

index 4a79b9d..74c967e 100644 (file)
@@ -560,6 +560,15 @@ source "drivers/scsi/aic7xxx/Kconfig.aic79xx"
 source "drivers/scsi/aic94xx/Kconfig"
 source "drivers/scsi/mvsas/Kconfig"
 
+config SCSI_MVUMI
+       tristate "Marvell UMI driver"
+       depends on SCSI && PCI
+       help
+         Module for Marvell Universal Message Interface(UMI) driver
+
+         To compile this driver as a module, choose M here: the
+         module will be called mvumi.
+
 config SCSI_DPT_I2O
        tristate "Adaptec I2O RAID support "
        depends on SCSI && PCI && VIRT_TO_BUS
index 3c08f53..2b46997 100644 (file)
@@ -134,6 +134,7 @@ obj-$(CONFIG_SCSI_IBMVFC)   += ibmvscsi/
 obj-$(CONFIG_SCSI_HPTIOP)      += hptiop.o
 obj-$(CONFIG_SCSI_STEX)                += stex.o
 obj-$(CONFIG_SCSI_MVSAS)       += mvsas/
+obj-$(CONFIG_SCSI_MVUMI)       += mvumi.o
 obj-$(CONFIG_PS3_ROM)          += ps3rom.o
 obj-$(CONFIG_SCSI_CXGB3_ISCSI) += libiscsi.o libiscsi_tcp.o cxgbi/
 obj-$(CONFIG_SCSI_CXGB4_ISCSI) += libiscsi.o libiscsi_tcp.o cxgbi/
diff --git a/drivers/scsi/mvumi.c b/drivers/scsi/mvumi.c
new file mode 100644 (file)
index 0000000..88cf1db
--- /dev/null
@@ -0,0 +1,2018 @@
+/*
+ * Marvell UMI driver
+ *
+ * Copyright 2011 Marvell. <jyli@marvell.com>
+ *
+ * This file is licensed under GPLv2.
+ *
+ * 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; version 2 of the
+ * License.
+ *
+ * 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
+*/
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/init.h>
+#include <linux/device.h>
+#include <linux/pci.h>
+#include <linux/list.h>
+#include <linux/spinlock.h>
+#include <linux/interrupt.h>
+#include <linux/delay.h>
+#include <linux/blkdev.h>
+#include <linux/io.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_host.h>
+#include <scsi/scsi_transport.h>
+#include <scsi/scsi_eh.h>
+#include <linux/uaccess.h>
+
+#include "mvumi.h"
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("jyli@marvell.com");
+MODULE_DESCRIPTION("Marvell UMI Driver");
+
+static DEFINE_PCI_DEVICE_TABLE(mvumi_pci_table) = {
+       { PCI_DEVICE(PCI_VENDOR_ID_MARVELL_2, PCI_DEVICE_ID_MARVELL_MV9143) },
+       { 0 }
+};
+
+MODULE_DEVICE_TABLE(pci, mvumi_pci_table);
+
+static void tag_init(struct mvumi_tag *st, unsigned short size)
+{
+       unsigned short i;
+       BUG_ON(size != st->size);
+       st->top = size;
+       for (i = 0; i < size; i++)
+               st->stack[i] = size - 1 - i;
+}
+
+static unsigned short tag_get_one(struct mvumi_hba *mhba, struct mvumi_tag *st)
+{
+       BUG_ON(st->top <= 0);
+       return st->stack[--st->top];
+}
+
+static void tag_release_one(struct mvumi_hba *mhba, struct mvumi_tag *st,
+                                                       unsigned short tag)
+{
+       BUG_ON(st->top >= st->size);
+       st->stack[st->top++] = tag;
+}
+
+static bool tag_is_empty(struct mvumi_tag *st)
+{
+       if (st->top == 0)
+               return 1;
+       else
+               return 0;
+}
+
+static void mvumi_unmap_pci_addr(struct pci_dev *dev, void **addr_array)
+{
+       int i;
+
+       for (i = 0; i < MAX_BASE_ADDRESS; i++)
+               if ((pci_resource_flags(dev, i) & IORESOURCE_MEM) &&
+                                                               addr_array[i])
+                       pci_iounmap(dev, addr_array[i]);
+}
+
+static int mvumi_map_pci_addr(struct pci_dev *dev, void **addr_array)
+{
+       int i;
+
+       for (i = 0; i < MAX_BASE_ADDRESS; i++) {
+               if (pci_resource_flags(dev, i) & IORESOURCE_MEM) {
+                       addr_array[i] = pci_iomap(dev, i, 0);
+                       if (!addr_array[i]) {
+                               dev_err(&dev->dev, "failed to map Bar[%d]\n",
+                                                                       i);
+                               mvumi_unmap_pci_addr(dev, addr_array);
+                               return -ENOMEM;
+                       }
+               } else
+                       addr_array[i] = NULL;
+
+               dev_dbg(&dev->dev, "Bar %d : %p.\n", i, addr_array[i]);
+       }
+
+       return 0;
+}
+
+static struct mvumi_res *mvumi_alloc_mem_resource(struct mvumi_hba *mhba,
+                               enum resource_type type, unsigned int size)
+{
+       struct mvumi_res *res = kzalloc(sizeof(*res), GFP_KERNEL);
+
+       if (!res) {
+               dev_err(&mhba->pdev->dev,
+                       "Failed to allocate memory for resouce manager.\n");
+               return NULL;
+       }
+
+       switch (type) {
+       case RESOURCE_CACHED_MEMORY:
+               res->virt_addr = kzalloc(size, GFP_KERNEL);
+               if (!res->virt_addr) {
+                       dev_err(&mhba->pdev->dev,
+                               "unable to allocate memory,size = %d.\n", size);
+                       kfree(res);
+                       return NULL;
+               }
+               break;
+
+       case RESOURCE_UNCACHED_MEMORY:
+               size = round_up(size, 8);
+               res->virt_addr = pci_alloc_consistent(mhba->pdev, size,
+                                                       &res->bus_addr);
+               if (!res->virt_addr) {
+                       dev_err(&mhba->pdev->dev,
+                                       "unable to allocate consistent mem,"
+                                                       "size = %d.\n", size);
+                       kfree(res);
+                       return NULL;
+               }
+               memset(res->virt_addr, 0, size);
+               break;
+
+       default:
+               dev_err(&mhba->pdev->dev, "unknown resource type %d.\n", type);
+               kfree(res);
+               return NULL;
+       }
+
+       res->type = type;
+       res->size = size;
+       INIT_LIST_HEAD(&res->entry);
+       list_add_tail(&res->entry, &mhba->res_list);
+
+       return res;
+}
+
+static void mvumi_release_mem_resource(struct mvumi_hba *mhba)
+{
+       struct mvumi_res *res, *tmp;
+
+       list_for_each_entry_safe(res, tmp, &mhba->res_list, entry) {
+               switch (res->type) {
+               case RESOURCE_UNCACHED_MEMORY:
+                       pci_free_consistent(mhba->pdev, res->size,
+                                               res->virt_addr, res->bus_addr);
+                       break;
+               case RESOURCE_CACHED_MEMORY:
+                       kfree(res->virt_addr);
+                       break;
+               default:
+                       dev_err(&mhba->pdev->dev,
+                               "unknown resource type %d\n", res->type);
+                       break;
+               }
+               list_del(&res->entry);
+               kfree(res);
+       }
+       mhba->fw_flag &= ~MVUMI_FW_ALLOC;
+}
+
+/**
+ * mvumi_make_sgl -    Prepares  SGL
+ * @mhba:              Adapter soft state
+ * @scmd:              SCSI command from the mid-layer
+ * @sgl_p:             SGL to be filled in
+ * @sg_count           return the number of SG elements
+ *
+ * If successful, this function returns 0. otherwise, it returns -1.
+ */
+static int mvumi_make_sgl(struct mvumi_hba *mhba, struct scsi_cmnd *scmd,
+                                       void *sgl_p, unsigned char *sg_count)
+{
+       struct scatterlist *sg;
+       struct mvumi_sgl *m_sg = (struct mvumi_sgl *) sgl_p;
+       unsigned int i;
+       unsigned int sgnum = scsi_sg_count(scmd);
+       dma_addr_t busaddr;
+
+       if (sgnum) {
+               sg = scsi_sglist(scmd);
+               *sg_count = pci_map_sg(mhba->pdev, sg, sgnum,
+                               (int) scmd->sc_data_direction);
+               if (*sg_count > mhba->max_sge) {
+                       dev_err(&mhba->pdev->dev, "sg count[0x%x] is bigger "
+                                               "than max sg[0x%x].\n",
+                                               *sg_count, mhba->max_sge);
+                       return -1;
+               }
+               for (i = 0; i < *sg_count; i++) {
+                       busaddr = sg_dma_address(&sg[i]);
+                       m_sg->baseaddr_l = cpu_to_le32(lower_32_bits(busaddr));
+                       m_sg->baseaddr_h = cpu_to_le32(upper_32_bits(busaddr));
+                       m_sg->flags = 0;
+                       m_sg->size = cpu_to_le32(sg_dma_len(&sg[i]));
+                       if ((i + 1) == *sg_count)
+                               m_sg->flags |= SGD_EOT;
+
+                       m_sg++;
+               }
+       } else {
+               scmd->SCp.dma_handle = scsi_bufflen(scmd) ?
+                       pci_map_single(mhba->pdev, scsi_sglist(scmd),
+                               scsi_bufflen(scmd),
+                               (int) scmd->sc_data_direction)
+                       : 0;
+               busaddr = scmd->SCp.dma_handle;
+               m_sg->baseaddr_l = cpu_to_le32(lower_32_bits(busaddr));
+               m_sg->baseaddr_h = cpu_to_le32(upper_32_bits(busaddr));
+               m_sg->flags = SGD_EOT;
+               m_sg->size = cpu_to_le32(scsi_bufflen(scmd));
+               *sg_count = 1;
+       }
+
+       return 0;
+}
+
+static int mvumi_internal_cmd_sgl(struct mvumi_hba *mhba, struct mvumi_cmd *cmd,
+                                                       unsigned int size)
+{
+       struct mvumi_sgl *m_sg;
+       void *virt_addr;
+       dma_addr_t phy_addr;
+
+       if (size == 0)
+               return 0;
+
+       virt_addr = pci_alloc_consistent(mhba->pdev, size, &phy_addr);
+       if (!virt_addr)
+               return -1;
+
+       memset(virt_addr, 0, size);
+
+       m_sg = (struct mvumi_sgl *) &cmd->frame->payload[0];
+       cmd->frame->sg_counts = 1;
+       cmd->data_buf = virt_addr;
+
+       m_sg->baseaddr_l = cpu_to_le32(lower_32_bits(phy_addr));
+       m_sg->baseaddr_h = cpu_to_le32(upper_32_bits(phy_addr));
+       m_sg->flags = SGD_EOT;
+       m_sg->size = cpu_to_le32(size);
+
+       return 0;
+}
+
+static struct mvumi_cmd *mvumi_create_internal_cmd(struct mvumi_hba *mhba,
+                               unsigned int buf_size)
+{
+       struct mvumi_cmd *cmd;
+
+       cmd = kzalloc(sizeof(*cmd), GFP_KERNEL);
+       if (!cmd) {
+               dev_err(&mhba->pdev->dev, "failed to create a internal cmd\n");
+               return NULL;
+       }
+       INIT_LIST_HEAD(&cmd->queue_pointer);
+
+       cmd->frame = kzalloc(mhba->ib_max_size, GFP_KERNEL);
+       if (!cmd->frame) {
+               dev_err(&mhba->pdev->dev, "failed to allocate memory for FW"
+                       " frame,size = %d.\n", mhba->ib_max_size);
+               kfree(cmd);
+               return NULL;
+       }
+
+       if (buf_size) {
+               if (mvumi_internal_cmd_sgl(mhba, cmd, buf_size)) {
+                       dev_err(&mhba->pdev->dev, "failed to allocate memory"
+                                               " for internal frame\n");
+                       kfree(cmd->frame);
+                       kfree(cmd);
+                       return NULL;
+               }
+       } else
+               cmd->frame->sg_counts = 0;
+
+       return cmd;
+}
+
+static void mvumi_delete_internal_cmd(struct mvumi_hba *mhba,
+                                               struct mvumi_cmd *cmd)
+{
+       struct mvumi_sgl *m_sg;
+       unsigned int size;
+       dma_addr_t phy_addr;
+
+       if (cmd && cmd->frame) {
+               if (cmd->frame->sg_counts) {
+                       m_sg = (struct mvumi_sgl *) &cmd->frame->payload[0];
+                       size = m_sg->size;
+
+                       phy_addr = (dma_addr_t) m_sg->baseaddr_l |
+                               (dma_addr_t) ((m_sg->baseaddr_h << 16) << 16);
+
+                       pci_free_consistent(mhba->pdev, size, cmd->data_buf,
+                                                               phy_addr);
+               }
+               kfree(cmd->frame);
+               kfree(cmd);
+       }
+}
+
+/**
+ * mvumi_get_cmd -     Get a command from the free pool
+ * @mhba:              Adapter soft state
+ *
+ * Returns a free command from the pool
+ */
+static struct mvumi_cmd *mvumi_get_cmd(struct mvumi_hba *mhba)
+{
+       struct mvumi_cmd *cmd = NULL;
+
+       if (likely(!list_empty(&mhba->cmd_pool))) {
+               cmd = list_entry((&mhba->cmd_pool)->next,
+                               struct mvumi_cmd, queue_pointer);
+               list_del_init(&cmd->queue_pointer);
+       } else
+               dev_warn(&mhba->pdev->dev, "command pool is empty!\n");
+
+       return cmd;
+}
+
+/**
+ * mvumi_return_cmd -  Return a cmd to free command pool
+ * @mhba:              Adapter soft state
+ * @cmd:               Command packet to be returned to free command pool
+ */
+static inline void mvumi_return_cmd(struct mvumi_hba *mhba,
+                                               struct mvumi_cmd *cmd)
+{
+       cmd->scmd = NULL;
+       list_add_tail(&cmd->queue_pointer, &mhba->cmd_pool);
+}
+
+/**
+ * mvumi_free_cmds -   Free all the cmds in the free cmd pool
+ * @mhba:              Adapter soft state
+ */
+static void mvumi_free_cmds(struct mvumi_hba *mhba)
+{
+       struct mvumi_cmd *cmd;
+
+       while (!list_empty(&mhba->cmd_pool)) {
+               cmd = list_first_entry(&mhba->cmd_pool, struct mvumi_cmd,
+                                                       queue_pointer);
+               list_del(&cmd->queue_pointer);
+               kfree(cmd->frame);
+               kfree(cmd);
+       }
+}
+
+/**
+ * mvumi_alloc_cmds -  Allocates the command packets
+ * @mhba:              Adapter soft state
+ *
+ */
+static int mvumi_alloc_cmds(struct mvumi_hba *mhba)
+{
+       int i;
+       struct mvumi_cmd *cmd;
+
+       for (i = 0; i < mhba->max_io; i++) {
+               cmd = kzalloc(sizeof(*cmd), GFP_KERNEL);
+               if (!cmd)
+                       goto err_exit;
+
+               INIT_LIST_HEAD(&cmd->queue_pointer);
+               list_add_tail(&cmd->queue_pointer, &mhba->cmd_pool);
+               cmd->frame = kzalloc(mhba->ib_max_size, GFP_KERNEL);
+               if (!cmd->frame)
+                       goto err_exit;
+       }
+       return 0;
+
+err_exit:
+       dev_err(&mhba->pdev->dev,
+                       "failed to allocate memory for cmd[0x%x].\n", i);
+       while (!list_empty(&mhba->cmd_pool)) {
+               cmd = list_first_entry(&mhba->cmd_pool, struct mvumi_cmd,
+                                               queue_pointer);
+               list_del(&cmd->queue_pointer);
+               kfree(cmd->frame);
+               kfree(cmd);
+       }
+       return -ENOMEM;
+}
+
+static int mvumi_get_ib_list_entry(struct mvumi_hba *mhba, void **ib_entry)
+{
+       unsigned int ib_rp_reg, cur_ib_entry;
+
+       if (atomic_read(&mhba->fw_outstanding) >= mhba->max_io) {
+               dev_warn(&mhba->pdev->dev, "firmware io overflow.\n");
+               return -1;
+       }
+       ib_rp_reg = ioread32(mhba->mmio + CLA_INB_READ_POINTER);
+
+       if (unlikely(((ib_rp_reg & CL_SLOT_NUM_MASK) ==
+                       (mhba->ib_cur_slot & CL_SLOT_NUM_MASK)) &&
+                       ((ib_rp_reg & CL_POINTER_TOGGLE) !=
+                       (mhba->ib_cur_slot & CL_POINTER_TOGGLE)))) {
+               dev_warn(&mhba->pdev->dev, "no free slot to use.\n");
+               return -1;
+       }
+
+       cur_ib_entry = mhba->ib_cur_slot & CL_SLOT_NUM_MASK;
+       cur_ib_entry++;
+       if (cur_ib_entry >= mhba->list_num_io) {
+               cur_ib_entry -= mhba->list_num_io;
+               mhba->ib_cur_slot ^= CL_POINTER_TOGGLE;
+       }
+       mhba->ib_cur_slot &= ~CL_SLOT_NUM_MASK;
+       mhba->ib_cur_slot |= (cur_ib_entry & CL_SLOT_NUM_MASK);
+       *ib_entry = mhba->ib_list + cur_ib_entry * mhba->ib_max_size;
+       atomic_inc(&mhba->fw_outstanding);
+
+       return 0;
+}
+
+static void mvumi_send_ib_list_entry(struct mvumi_hba *mhba)
+{
+       iowrite32(0xfff, mhba->ib_shadow);
+       iowrite32(mhba->ib_cur_slot, mhba->mmio + CLA_INB_WRITE_POINTER);
+}
+
+static char mvumi_check_ob_frame(struct mvumi_hba *mhba,
+               unsigned int cur_obf, struct mvumi_rsp_frame *p_outb_frame)
+{
+       unsigned short tag, request_id;
+
+       udelay(1);
+       p_outb_frame = mhba->ob_list + cur_obf * mhba->ob_max_size;
+       request_id = p_outb_frame->request_id;
+       tag = p_outb_frame->tag;
+       if (tag > mhba->tag_pool.size) {
+               dev_err(&mhba->pdev->dev, "ob frame data error\n");
+               return -1;
+       }
+       if (mhba->tag_cmd[tag] == NULL) {
+               dev_err(&mhba->pdev->dev, "tag[0x%x] with NO command\n", tag);
+               return -1;
+       } else if (mhba->tag_cmd[tag]->request_id != request_id &&
+                                               mhba->request_id_enabled) {
+                       dev_err(&mhba->pdev->dev, "request ID from FW:0x%x,"
+                                       "cmd request ID:0x%x\n", request_id,
+                                       mhba->tag_cmd[tag]->request_id);
+                       return -1;
+       }
+
+       return 0;
+}
+
+static void mvumi_receive_ob_list_entry(struct mvumi_hba *mhba)
+{
+       unsigned int ob_write_reg, ob_write_shadow_reg;
+       unsigned int cur_obf, assign_obf_end, i;
+       struct mvumi_ob_data *ob_data;
+       struct mvumi_rsp_frame *p_outb_frame;
+
+       do {
+               ob_write_reg = ioread32(mhba->mmio + CLA_OUTB_COPY_POINTER);
+               ob_write_shadow_reg = ioread32(mhba->ob_shadow);
+       } while ((ob_write_reg & CL_SLOT_NUM_MASK) != ob_write_shadow_reg);
+
+       cur_obf = mhba->ob_cur_slot & CL_SLOT_NUM_MASK;
+       assign_obf_end = ob_write_reg & CL_SLOT_NUM_MASK;
+
+       if ((ob_write_reg & CL_POINTER_TOGGLE) !=
+                               (mhba->ob_cur_slot & CL_POINTER_TOGGLE)) {
+               assign_obf_end += mhba->list_num_io;
+       }
+
+       for (i = (assign_obf_end - cur_obf); i != 0; i--) {
+               cur_obf++;
+               if (cur_obf >= mhba->list_num_io) {
+                       cur_obf -= mhba->list_num_io;
+                       mhba->ob_cur_slot ^= CL_POINTER_TOGGLE;
+               }
+
+               p_outb_frame = mhba->ob_list + cur_obf * mhba->ob_max_size;
+
+               /* Copy pointer may point to entry in outbound list
+               *  before entry has valid data
+               */
+               if (unlikely(p_outb_frame->tag > mhba->tag_pool.size ||
+                       mhba->tag_cmd[p_outb_frame->tag] == NULL ||
+                       p_outb_frame->request_id !=
+                               mhba->tag_cmd[p_outb_frame->tag]->request_id))
+                       if (mvumi_check_ob_frame(mhba, cur_obf, p_outb_frame))
+                               continue;
+
+               if (!list_empty(&mhba->ob_data_list)) {
+                       ob_data = (struct mvumi_ob_data *)
+                               list_first_entry(&mhba->ob_data_list,
+                                       struct mvumi_ob_data, list);
+                       list_del_init(&ob_data->list);
+               } else {
+                       ob_data = NULL;
+                       if (cur_obf == 0) {
+                               cur_obf = mhba->list_num_io - 1;
+                               mhba->ob_cur_slot ^= CL_POINTER_TOGGLE;
+                       } else
+                               cur_obf -= 1;
+                       break;
+               }
+
+               memcpy(ob_data->data, p_outb_frame, mhba->ob_max_size);
+               p_outb_frame->tag = 0xff;
+
+               list_add_tail(&ob_data->list, &mhba->free_ob_list);
+       }
+       mhba->ob_cur_slot &= ~CL_SLOT_NUM_MASK;
+       mhba->ob_cur_slot |= (cur_obf & CL_SLOT_NUM_MASK);
+       iowrite32(mhba->ob_cur_slot, mhba->mmio + CLA_OUTB_READ_POINTER);
+}
+
+static void mvumi_reset(void *regs)
+{
+       iowrite32(0, regs + CPU_ENPOINTA_MASK_REG);
+       if (ioread32(regs + CPU_ARM_TO_PCIEA_MSG1) != HANDSHAKE_DONESTATE)
+               return;
+
+       iowrite32(DRBL_SOFT_RESET, regs + CPU_PCIEA_TO_ARM_DRBL_REG);
+}
+
+static unsigned char mvumi_start(struct mvumi_hba *mhba);
+
+static int mvumi_wait_for_outstanding(struct mvumi_hba *mhba)
+{
+       mhba->fw_state = FW_STATE_ABORT;
+       mvumi_reset(mhba->mmio);
+
+       if (mvumi_start(mhba))
+               return FAILED;
+       else
+               return SUCCESS;
+}
+
+static int mvumi_host_reset(struct scsi_cmnd *scmd)
+{
+       struct mvumi_hba *mhba;
+
+       mhba = (struct mvumi_hba *) scmd->device->host->hostdata;
+
+       scmd_printk(KERN_NOTICE, scmd, "RESET -%ld cmd=%x retries=%x\n",
+                       scmd->serial_number, scmd->cmnd[0], scmd->retries);
+
+       return mvumi_wait_for_outstanding(mhba);
+}
+
+static int mvumi_issue_blocked_cmd(struct mvumi_hba *mhba,
+                                               struct mvumi_cmd *cmd)
+{
+       unsigned long flags;
+
+       cmd->cmd_status = REQ_STATUS_PENDING;
+
+       if (atomic_read(&cmd->sync_cmd)) {
+               dev_err(&mhba->pdev->dev,
+                       "last blocked cmd not finished, sync_cmd = %d\n",
+                                               atomic_read(&cmd->sync_cmd));
+               BUG_ON(1);
+               return -1;
+       }
+       atomic_inc(&cmd->sync_cmd);
+       spin_lock_irqsave(mhba->shost->host_lock, flags);
+       mhba->instancet->fire_cmd(mhba, cmd);
+       spin_unlock_irqrestore(mhba->shost->host_lock, flags);
+
+       wait_event_timeout(mhba->int_cmd_wait_q,
+               (cmd->cmd_status != REQ_STATUS_PENDING),
+               MVUMI_INTERNAL_CMD_WAIT_TIME * HZ);
+
+       /* command timeout */
+       if (atomic_read(&cmd->sync_cmd)) {
+               spin_lock_irqsave(mhba->shost->host_lock, flags);
+               atomic_dec(&cmd->sync_cmd);
+               if (mhba->tag_cmd[cmd->frame->tag]) {
+                       mhba->tag_cmd[cmd->frame->tag] = 0;
+                       dev_warn(&mhba->pdev->dev, "TIMEOUT:release tag [%d]\n",
+                                                       cmd->frame->tag);
+                       tag_release_one(mhba, &mhba->tag_pool, cmd->frame->tag);
+               }
+               if (!list_empty(&cmd->queue_pointer)) {
+                       dev_warn(&mhba->pdev->dev,
+                               "TIMEOUT:A internal command doesn't send!\n");
+                       list_del_init(&cmd->queue_pointer);
+               } else
+                       atomic_dec(&mhba->fw_outstanding);
+
+               spin_unlock_irqrestore(mhba->shost->host_lock, flags);
+       }
+       return 0;
+}
+
+static void mvumi_release_fw(struct mvumi_hba *mhba)
+{
+       mvumi_free_cmds(mhba);
+       mvumi_release_mem_resource(mhba);
+       mvumi_unmap_pci_addr(mhba->pdev, mhba->base_addr);
+       kfree(mhba->handshake_page);
+       pci_release_regions(mhba->pdev);
+}
+
+static unsigned char mvumi_flush_cache(struct mvumi_hba *mhba)
+{
+       struct mvumi_cmd *cmd;
+       struct mvumi_msg_frame *frame;
+       unsigned char device_id, retry = 0;
+       unsigned char bitcount = sizeof(unsigned char) * 8;
+
+       for (device_id = 0; device_id < mhba->max_target_id; device_id++) {
+               if (!(mhba->target_map[device_id / bitcount] &
+                               (1 << (device_id % bitcount))))
+                       continue;
+get_cmd:       cmd = mvumi_create_internal_cmd(mhba, 0);
+               if (!cmd) {
+                       if (retry++ >= 5) {
+                               dev_err(&mhba->pdev->dev, "failed to get memory"
+                                       " for internal flush cache cmd for "
+                                       "device %d", device_id);
+                               retry = 0;
+                               continue;
+                       } else
+                               goto get_cmd;
+               }
+               cmd->scmd = NULL;
+               cmd->cmd_status = REQ_STATUS_PENDING;
+               atomic_set(&cmd->sync_cmd, 0);
+               frame = cmd->frame;
+               frame->req_function = CL_FUN_SCSI_CMD;
+               frame->device_id = device_id;
+               frame->cmd_flag = CMD_FLAG_NON_DATA;
+               frame->data_transfer_length = 0;
+               frame->cdb_length = MAX_COMMAND_SIZE;
+               memset(frame->cdb, 0, MAX_COMMAND_SIZE);
+               frame->cdb[0] = SCSI_CMD_MARVELL_SPECIFIC;
+               frame->cdb[2] = CDB_CORE_SHUTDOWN;
+
+               mvumi_issue_blocked_cmd(mhba, cmd);
+               if (cmd->cmd_status != SAM_STAT_GOOD) {
+                       dev_err(&mhba->pdev->dev,
+                               "device %d flush cache failed, status=0x%x.\n",
+                               device_id, cmd->cmd_status);
+               }
+
+               mvumi_delete_internal_cmd(mhba, cmd);
+       }
+       return 0;
+}
+
+static unsigned char
+mvumi_calculate_checksum(struct mvumi_hs_header *p_header,
+                                                       unsigned short len)
+{
+       unsigned char *ptr;
+       unsigned char ret = 0, i;
+
+       ptr = (unsigned char *) p_header->frame_content;
+       for (i = 0; i < len; i++) {
+               ret ^= *ptr;
+               ptr++;
+       }
+
+       return ret;
+}
+
+void mvumi_hs_build_page(struct mvumi_hba *mhba,
+                               struct mvumi_hs_header *hs_header)
+{
+       struct mvumi_hs_page2 *hs_page2;
+       struct mvumi_hs_page4 *hs_page4;
+       struct mvumi_hs_page3 *hs_page3;
+       struct timeval time;
+       unsigned int local_time;
+
+       switch (hs_header->page_code) {
+       case HS_PAGE_HOST_INFO:
+               hs_page2 = (struct mvumi_hs_page2 *) hs_header;
+               hs_header->frame_length = sizeof(*hs_page2) - 4;
+               memset(hs_header->frame_content, 0, hs_header->frame_length);
+               hs_page2->host_type = 3; /* 3 mean linux*/
+               hs_page2->host_ver.ver_major = VER_MAJOR;
+               hs_page2->host_ver.ver_minor = VER_MINOR;
+               hs_page2->host_ver.ver_oem = VER_OEM;
+               hs_page2->host_ver.ver_build = VER_BUILD;
+               hs_page2->system_io_bus = 0;
+               hs_page2->slot_number = 0;
+               hs_page2->intr_level = 0;
+               hs_page2->intr_vector = 0;
+               do_gettimeofday(&time);
+               local_time = (unsigned int) (time.tv_sec -
+                                               (sys_tz.tz_minuteswest * 60));
+               hs_page2->seconds_since1970 = local_time;
+               hs_header->checksum = mvumi_calculate_checksum(hs_header,
+                                               hs_header->frame_length);
+               break;
+
+       case HS_PAGE_FIRM_CTL:
+               hs_page3 = (struct mvumi_hs_page3 *) hs_header;
+               hs_header->frame_length = sizeof(*hs_page3) - 4;
+               memset(hs_header->frame_content, 0, hs_header->frame_length);
+               hs_header->checksum = mvumi_calculate_checksum(hs_header,
+                                               hs_header->frame_length);
+               break;
+
+       case HS_PAGE_CL_INFO:
+               hs_page4 = (struct mvumi_hs_page4 *) hs_header;
+               hs_header->frame_length = sizeof(*hs_page4) - 4;
+               memset(hs_header->frame_content, 0, hs_header->frame_length);
+               hs_page4->ib_baseaddr_l = lower_32_bits(mhba->ib_list_phys);
+               hs_page4->ib_baseaddr_h = upper_32_bits(mhba->ib_list_phys);
+
+               hs_page4->ob_baseaddr_l = lower_32_bits(mhba->ob_list_phys);
+               hs_page4->ob_baseaddr_h = upper_32_bits(mhba->ob_list_phys);
+               hs_page4->ib_entry_size = mhba->ib_max_size_setting;
+               hs_page4->ob_entry_size = mhba->ob_max_size_setting;
+               hs_page4->ob_depth = mhba->list_num_io;
+               hs_page4->ib_depth = mhba->list_num_io;
+               hs_header->checksum = mvumi_calculate_checksum(hs_header,
+                                               hs_header->frame_length);
+               break;
+
+       default:
+               dev_err(&mhba->pdev->dev, "cannot build page, code[0x%x]\n",
+                       hs_header->page_code);
+               break;
+       }
+}
+
+/**
+ * mvumi_init_data -   Initialize requested date for FW
+ * @mhba:                      Adapter soft state
+ */
+static int mvumi_init_data(struct mvumi_hba *mhba)
+{
+       struct mvumi_ob_data *ob_pool;
+       struct mvumi_res *res_mgnt;
+       unsigned int tmp_size, offset, i;
+       void *virmem, *v;
+       dma_addr_t p;
+
+       if (mhba->fw_flag & MVUMI_FW_ALLOC)
+               return 0;
+
+       tmp_size = mhba->ib_max_size * mhba->max_io;
+       tmp_size += 128 + mhba->ob_max_size * mhba->max_io;
+       tmp_size += 8 + sizeof(u32) + 16;
+
+       res_mgnt = mvumi_alloc_mem_resource(mhba,
+                                       RESOURCE_UNCACHED_MEMORY, tmp_size);
+       if (!res_mgnt) {
+               dev_err(&mhba->pdev->dev,
+                       "failed to allocate memory for inbound list\n");
+               goto fail_alloc_dma_buf;
+       }
+
+       p = res_mgnt->bus_addr;
+       v = res_mgnt->virt_addr;
+       /* ib_list */
+       offset = round_up(p, 128) - p;
+       p += offset;
+       v += offset;
+       mhba->ib_list = v;
+       mhba->ib_list_phys = p;
+       v += mhba->ib_max_size * mhba->max_io;
+       p += mhba->ib_max_size * mhba->max_io;
+       /* ib shadow */
+       offset = round_up(p, 8) - p;
+       p += offset;
+       v += offset;
+       mhba->ib_shadow = v;
+       mhba->ib_shadow_phys = p;
+       p += sizeof(u32);
+       v += sizeof(u32);
+       /* ob shadow */
+       offset = round_up(p, 8) - p;
+       p += offset;
+       v += offset;
+       mhba->ob_shadow = v;
+       mhba->ob_shadow_phys = p;
+       p += 8;
+       v += 8;
+
+       /* ob list */
+       offset = round_up(p, 128) - p;
+       p += offset;
+       v += offset;
+
+       mhba->ob_list = v;
+       mhba->ob_list_phys = p;
+
+       /* ob data pool */
+       tmp_size = mhba->max_io * (mhba->ob_max_size + sizeof(*ob_pool));
+       tmp_size = round_up(tmp_size, 8);
+
+       res_mgnt = mvumi_alloc_mem_resource(mhba,
+                               RESOURCE_CACHED_MEMORY, tmp_size);
+       if (!res_mgnt) {
+               dev_err(&mhba->pdev->dev,
+                       "failed to allocate memory for outbound data buffer\n");
+               goto fail_alloc_dma_buf;
+       }
+       virmem = res_mgnt->virt_addr;
+
+       for (i = mhba->max_io; i != 0; i--) {
+               ob_pool = (struct mvumi_ob_data *) virmem;
+               list_add_tail(&ob_pool->list, &mhba->ob_data_list);
+               virmem += mhba->ob_max_size + sizeof(*ob_pool);
+       }
+
+       tmp_size = sizeof(unsigned short) * mhba->max_io +
+                               sizeof(struct mvumi_cmd *) * mhba->max_io;
+       tmp_size += round_up(mhba->max_target_id, sizeof(unsigned char) * 8) /
+                                               (sizeof(unsigned char) * 8);
+
+       res_mgnt = mvumi_alloc_mem_resource(mhba,
+                               RESOURCE_CACHED_MEMORY, tmp_size);
+       if (!res_mgnt) {
+               dev_err(&mhba->pdev->dev,
+                       "failed to allocate memory for tag and target map\n");
+               goto fail_alloc_dma_buf;
+       }
+
+       virmem = res_mgnt->virt_addr;
+       mhba->tag_pool.stack = virmem;
+       mhba->tag_pool.size = mhba->max_io;
+       tag_init(&mhba->tag_pool, mhba->max_io);
+       virmem += sizeof(unsigned short) * mhba->max_io;
+
+       mhba->tag_cmd = virmem;
+       virmem += sizeof(struct mvumi_cmd *) * mhba->max_io;
+
+       mhba->target_map = virmem;
+
+       mhba->fw_flag |= MVUMI_FW_ALLOC;
+       return 0;
+
+fail_alloc_dma_buf:
+       mvumi_release_mem_resource(mhba);
+       return -1;
+}
+
+static int mvumi_hs_process_page(struct mvumi_hba *mhba,
+                               struct mvumi_hs_header *hs_header)
+{
+       struct mvumi_hs_page1 *hs_page1;
+       unsigned char page_checksum;
+
+       page_checksum = mvumi_calculate_checksum(hs_header,
+                                               hs_header->frame_length);
+       if (page_checksum != hs_header->checksum) {
+               dev_err(&mhba->pdev->dev, "checksum error\n");
+               return -1;
+       }
+
+       switch (hs_header->page_code) {
+       case HS_PAGE_FIRM_CAP:
+               hs_page1 = (struct mvumi_hs_page1 *) hs_header;
+
+               mhba->max_io = hs_page1->max_io_support;
+               mhba->list_num_io = hs_page1->cl_inout_list_depth;
+               mhba->max_transfer_size = hs_page1->max_transfer_size;
+               mhba->max_target_id = hs_page1->max_devices_support;
+               mhba->hba_capability = hs_page1->capability;
+               mhba->ib_max_size_setting = hs_page1->cl_in_max_entry_size;
+               mhba->ib_max_size = (1 << hs_page1->cl_in_max_entry_size) << 2;
+
+               mhba->ob_max_size_setting = hs_page1->cl_out_max_entry_size;
+               mhba->ob_max_size = (1 << hs_page1->cl_out_max_entry_size) << 2;
+
+               dev_dbg(&mhba->pdev->dev, "FW version:%d\n",
+                                               hs_page1->fw_ver.ver_build);
+
+               break;
+       default:
+               dev_err(&mhba->pdev->dev, "handshake: page code error\n");
+               return -1;
+       }
+       return 0;
+}
+
+/**
+ * mvumi_handshake -   Move the FW to READY state
+ * @mhba:                              Adapter soft state
+ *
+ * During the initialization, FW passes can potentially be in any one of
+ * several possible states. If the FW in operational, waiting-for-handshake
+ * states, driver must take steps to bring it to ready state. Otherwise, it
+ * has to wait for the ready state.
+ */
+static int mvumi_handshake(struct mvumi_hba *mhba)
+{
+       unsigned int hs_state, tmp, hs_fun;
+       struct mvumi_hs_header *hs_header;
+       void *regs = mhba->mmio;
+
+       if (mhba->fw_state == FW_STATE_STARTING)
+               hs_state = HS_S_START;
+       else {
+               tmp = ioread32(regs + CPU_ARM_TO_PCIEA_MSG0);
+               hs_state = HS_GET_STATE(tmp);
+               dev_dbg(&mhba->pdev->dev, "handshake state[0x%x].\n", hs_state);
+               if (HS_GET_STATUS(tmp) != HS_STATUS_OK) {
+                       mhba->fw_state = FW_STATE_STARTING;
+                       return -1;
+               }
+       }
+
+       hs_fun = 0;
+       switch (hs_state) {
+       case HS_S_START:
+               mhba->fw_state = FW_STATE_HANDSHAKING;
+               HS_SET_STATUS(hs_fun, HS_STATUS_OK);
+               HS_SET_STATE(hs_fun, HS_S_RESET);
+               iowrite32(HANDSHAKE_SIGNATURE, regs + CPU_PCIEA_TO_ARM_MSG1);
+               iowrite32(hs_fun, regs + CPU_PCIEA_TO_ARM_MSG0);
+               iowrite32(DRBL_HANDSHAKE, regs + CPU_PCIEA_TO_ARM_DRBL_REG);
+               break;
+
+       case HS_S_RESET:
+               iowrite32(lower_32_bits(mhba->handshake_page_phys),
+                                       regs + CPU_PCIEA_TO_ARM_MSG1);
+               iowrite32(upper_32_bits(mhba->handshake_page_phys),
+                                       regs + CPU_ARM_TO_PCIEA_MSG1);
+               HS_SET_STATUS(hs_fun, HS_STATUS_OK);
+               HS_SET_STATE(hs_fun, HS_S_PAGE_ADDR);
+               iowrite32(hs_fun, regs + CPU_PCIEA_TO_ARM_MSG0);
+               iowrite32(DRBL_HANDSHAKE, regs + CPU_PCIEA_TO_ARM_DRBL_REG);
+
+               break;
+
+       case HS_S_PAGE_ADDR:
+       case HS_S_QUERY_PAGE:
+       case HS_S_SEND_PAGE:
+               hs_header = (struct mvumi_hs_header *) mhba->handshake_page;
+               if (hs_header->page_code == HS_PAGE_FIRM_CAP) {
+                       mhba->hba_total_pages =
+                       ((struct mvumi_hs_page1 *) hs_header)->total_pages;
+
+                       if (mhba->hba_total_pages == 0)
+                               mhba->hba_total_pages = HS_PAGE_TOTAL-1;
+               }
+
+               if (hs_state == HS_S_QUERY_PAGE) {
+                       if (mvumi_hs_process_page(mhba, hs_header)) {
+                               HS_SET_STATE(hs_fun, HS_S_ABORT);
+                               return -1;
+                       }
+                       if (mvumi_init_data(mhba)) {
+                               HS_SET_STATE(hs_fun, HS_S_ABORT);
+                               return -1;
+                       }
+               } else if (hs_state == HS_S_PAGE_ADDR) {
+                       hs_header->page_code = 0;
+                       mhba->hba_total_pages = HS_PAGE_TOTAL-1;
+               }
+
+               if ((hs_header->page_code + 1) <= mhba->hba_total_pages) {
+                       hs_header->page_code++;
+                       if (hs_header->page_code != HS_PAGE_FIRM_CAP) {
+                               mvumi_hs_build_page(mhba, hs_header);
+                               HS_SET_STATE(hs_fun, HS_S_SEND_PAGE);
+                       } else
+                               HS_SET_STATE(hs_fun, HS_S_QUERY_PAGE);
+               } else
+                       HS_SET_STATE(hs_fun, HS_S_END);
+
+               HS_SET_STATUS(hs_fun, HS_STATUS_OK);
+               iowrite32(hs_fun, regs + CPU_PCIEA_TO_ARM_MSG0);
+               iowrite32(DRBL_HANDSHAKE, regs + CPU_PCIEA_TO_ARM_DRBL_REG);
+               break;
+
+       case HS_S_END:
+               /* Set communication list ISR */
+               tmp = ioread32(regs + CPU_ENPOINTA_MASK_REG);
+               tmp |= INT_MAP_COMAOUT | INT_MAP_COMAERR;
+               iowrite32(tmp, regs + CPU_ENPOINTA_MASK_REG);
+               iowrite32(mhba->list_num_io, mhba->ib_shadow);
+               /* Set InBound List Avaliable count shadow */
+               iowrite32(lower_32_bits(mhba->ib_shadow_phys),
+                                       regs + CLA_INB_AVAL_COUNT_BASEL);
+               iowrite32(upper_32_bits(mhba->ib_shadow_phys),
+                                       regs + CLA_INB_AVAL_COUNT_BASEH);
+
+               /* Set OutBound List Avaliable count shadow */
+               iowrite32((mhba->list_num_io-1) | CL_POINTER_TOGGLE,
+                                               mhba->ob_shadow);
+               iowrite32(lower_32_bits(mhba->ob_shadow_phys), regs + 0x5B0);
+               iowrite32(upper_32_bits(mhba->ob_shadow_phys), regs + 0x5B4);
+
+               mhba->ib_cur_slot = (mhba->list_num_io - 1) | CL_POINTER_TOGGLE;
+               mhba->ob_cur_slot = (mhba->list_num_io - 1) | CL_POINTER_TOGGLE;
+               mhba->fw_state = FW_STATE_STARTED;
+
+               break;
+       default:
+               dev_err(&mhba->pdev->dev, "unknown handshake state [0x%x].\n",
+                                                               hs_state);
+               return -1;
+       }
+       return 0;
+}
+
+static unsigned char mvumi_handshake_event(struct mvumi_hba *mhba)
+{
+       unsigned int isr_status;
+       unsigned long before;
+
+       before = jiffies;
+       mvumi_handshake(mhba);
+       do {
+               isr_status = mhba->instancet->read_fw_status_reg(mhba->mmio);
+
+               if (mhba->fw_state == FW_STATE_STARTED)
+                       return 0;
+               if (time_after(jiffies, before + FW_MAX_DELAY * HZ)) {
+                       dev_err(&mhba->pdev->dev,
+                               "no handshake response at state 0x%x.\n",
+                                 mhba->fw_state);
+                       dev_err(&mhba->pdev->dev,
+                               "isr : global=0x%x,status=0x%x.\n",
+                                       mhba->global_isr, isr_status);
+                       return -1;
+               }
+               rmb();
+               usleep_range(1000, 2000);
+       } while (!(isr_status & DRBL_HANDSHAKE_ISR));
+
+       return 0;
+}
+
+static unsigned char mvumi_check_handshake(struct mvumi_hba *mhba)
+{
+       void *regs = mhba->mmio;
+       unsigned int tmp;
+       unsigned long before;
+
+       before = jiffies;
+       tmp = ioread32(regs + CPU_ARM_TO_PCIEA_MSG1);
+       while ((tmp != HANDSHAKE_READYSTATE) && (tmp != HANDSHAKE_DONESTATE)) {
+               if (tmp != HANDSHAKE_READYSTATE)
+                       iowrite32(DRBL_MU_RESET,
+                                       regs + CPU_PCIEA_TO_ARM_DRBL_REG);
+               if (time_after(jiffies, before + FW_MAX_DELAY * HZ)) {
+                       dev_err(&mhba->pdev->dev,
+                               "invalid signature [0x%x].\n", tmp);
+                       return -1;
+               }
+               usleep_range(1000, 2000);
+               rmb();
+               tmp = ioread32(regs + CPU_ARM_TO_PCIEA_MSG1);
+       }
+
+       mhba->fw_state = FW_STATE_STARTING;
+       dev_dbg(&mhba->pdev->dev, "start firmware handshake...\n");
+       do {
+               if (mvumi_handshake_event(mhba)) {
+                       dev_err(&mhba->pdev->dev,
+                                       "handshake failed at state 0x%x.\n",
+                                               mhba->fw_state);
+                       return -1;
+               }
+       } while (mhba->fw_state != FW_STATE_STARTED);
+
+       dev_dbg(&mhba->pdev->dev, "firmware handshake done\n");
+
+       return 0;
+}
+
+static unsigned char mvumi_start(struct mvumi_hba *mhba)
+{
+       void *regs = mhba->mmio;
+       unsigned int tmp;
+       /* clear Door bell */
+       tmp = ioread32(regs + CPU_ARM_TO_PCIEA_DRBL_REG);
+       iowrite32(tmp, regs + CPU_ARM_TO_PCIEA_DRBL_REG);
+
+       iowrite32(0x3FFFFFFF, regs + CPU_ARM_TO_PCIEA_MASK_REG);
+       tmp = ioread32(regs + CPU_ENPOINTA_MASK_REG) | INT_MAP_DL_CPU2PCIEA;
+       iowrite32(tmp, regs + CPU_ENPOINTA_MASK_REG);
+       if (mvumi_check_handshake(mhba))
+               return -1;
+
+       return 0;
+}
+
+/**
+ * mvumi_complete_cmd -        Completes a command
+ * @mhba:                      Adapter soft state
+ * @cmd:                       Command to be completed
+ */
+static void mvumi_complete_cmd(struct mvumi_hba *mhba, struct mvumi_cmd *cmd,
+                                       struct mvumi_rsp_frame *ob_frame)
+{
+       struct scsi_cmnd *scmd = cmd->scmd;
+
+       cmd->scmd->SCp.ptr = NULL;
+       scmd->result = ob_frame->req_status;
+
+       switch (ob_frame->req_status) {
+       case SAM_STAT_GOOD:
+               scmd->result |= DID_OK << 16;
+               break;
+       case SAM_STAT_BUSY:
+               scmd->result |= DID_BUS_BUSY << 16;
+               break;
+       case SAM_STAT_CHECK_CONDITION:
+               scmd->result |= (DID_OK << 16);
+               if (ob_frame->rsp_flag & CL_RSP_FLAG_SENSEDATA) {
+                       memcpy(cmd->scmd->sense_buffer, ob_frame->payload,
+                               sizeof(struct mvumi_sense_data));
+                       scmd->result |=  (DRIVER_SENSE << 24);
+               }
+               break;
+       default:
+               scmd->result |= (DRIVER_INVALID << 24) | (DID_ABORT << 16);
+               break;
+       }
+
+       if (scsi_bufflen(scmd)) {
+               if (scsi_sg_count(scmd)) {
+                       pci_unmap_sg(mhba->pdev,
+                               scsi_sglist(scmd),
+                               scsi_sg_count(scmd),
+                               (int) scmd->sc_data_direction);
+               } else {
+                       pci_unmap_single(mhba->pdev,
+                               scmd->SCp.dma_handle,
+                               scsi_bufflen(scmd),
+                               (int) scmd->sc_data_direction);
+
+                       scmd->SCp.dma_handle = 0;
+               }
+       }
+       cmd->scmd->scsi_done(scmd);
+       mvumi_return_cmd(mhba, cmd);
+}
+static void mvumi_complete_internal_cmd(struct mvumi_hba *mhba,
+                                               struct mvumi_cmd *cmd,
+                                       struct mvumi_rsp_frame *ob_frame)
+{
+       if (atomic_read(&cmd->sync_cmd)) {
+               cmd->cmd_status = ob_frame->req_status;
+
+               if ((ob_frame->req_status == SAM_STAT_CHECK_CONDITION) &&
+                               (ob_frame->rsp_flag & CL_RSP_FLAG_SENSEDATA) &&
+                               cmd->data_buf) {
+                       memcpy(cmd->data_buf, ob_frame->payload,
+                                       sizeof(struct mvumi_sense_data));
+               }
+               atomic_dec(&cmd->sync_cmd);
+               wake_up(&mhba->int_cmd_wait_q);
+       }
+}
+
+static void mvumi_show_event(struct mvumi_hba *mhba,
+                       struct mvumi_driver_event *ptr)
+{
+       unsigned int i;
+
+       dev_warn(&mhba->pdev->dev,
+               "Event[0x%x] id[0x%x] severity[0x%x] device id[0x%x]\n",
+               ptr->sequence_no, ptr->event_id, ptr->severity, ptr->device_id);
+       if (ptr->param_count) {
+               printk(KERN_WARNING "Event param(len 0x%x): ",
+                                               ptr->param_count);
+               for (i = 0; i < ptr->param_count; i++)
+                       printk(KERN_WARNING "0x%x ", ptr->params[i]);
+
+               printk(KERN_WARNING "\n");
+       }
+
+       if (ptr->sense_data_length) {
+               printk(KERN_WARNING "Event sense data(len 0x%x): ",
+                                               ptr->sense_data_length);
+               for (i = 0; i < ptr->sense_data_length; i++)
+                       printk(KERN_WARNING "0x%x ", ptr->sense_data[i]);
+               printk(KERN_WARNING "\n");
+       }
+}
+
+static void mvumi_notification(struct mvumi_hba *mhba, u8 msg, void *buffer)
+{
+       if (msg == APICDB1_EVENT_GETEVENT) {
+               int i, count;
+               struct mvumi_driver_event *param = NULL;
+               struct mvumi_event_req *er = buffer;
+               count = er->count;
+               if (count > MAX_EVENTS_RETURNED) {
+                       dev_err(&mhba->pdev->dev, "event count[0x%x] is bigger"
+                                       " than max event count[0x%x].\n",
+                                       count, MAX_EVENTS_RETURNED);
+                       return;
+               }
+               for (i = 0; i < count; i++) {
+                       param = &er->events[i];
+                       mvumi_show_event(mhba, param);
+               }
+       }
+}
+
+static int mvumi_get_event(struct mvumi_hba *mhba, unsigned char msg)
+{
+       struct mvumi_cmd *cmd;
+       struct mvumi_msg_frame *frame;
+
+       cmd = mvumi_create_internal_cmd(mhba, 512);
+       if (!cmd)
+               return -1;
+       cmd->scmd = NULL;
+       cmd->cmd_status = REQ_STATUS_PENDING;
+       atomic_set(&cmd->sync_cmd, 0);
+       frame = cmd->frame;
+       frame->device_id = 0;
+       frame->cmd_flag = CMD_FLAG_DATA_IN;
+       frame->req_function = CL_FUN_SCSI_CMD;
+       frame->cdb_length = MAX_COMMAND_SIZE;
+       frame->data_transfer_length = sizeof(struct mvumi_event_req);
+       memset(frame->cdb, 0, MAX_COMMAND_SIZE);
+       frame->cdb[0] = APICDB0_EVENT;
+       frame->cdb[1] = msg;
+       mvumi_issue_blocked_cmd(mhba, cmd);
+
+       if (cmd->cmd_status != SAM_STAT_GOOD)
+               dev_err(&mhba->pdev->dev, "get event failed, status=0x%x.\n",
+                                                       cmd->cmd_status);
+       else
+               mvumi_notification(mhba, cmd->frame->cdb[1], cmd->data_buf);
+
+       mvumi_delete_internal_cmd(mhba, cmd);
+       return 0;
+}
+
+static void mvumi_scan_events(struct work_struct *work)
+{
+       struct mvumi_events_wq *mu_ev =
+               container_of(work, struct mvumi_events_wq, work_q);
+
+       mvumi_get_event(mu_ev->mhba, mu_ev->event);
+       kfree(mu_ev);
+}
+
+static void mvumi_launch_events(struct mvumi_hba *mhba, u8 msg)
+{
+       struct mvumi_events_wq *mu_ev;
+
+       mu_ev = kzalloc(sizeof(*mu_ev), GFP_ATOMIC);
+       if (mu_ev) {
+               INIT_WORK(&mu_ev->work_q, mvumi_scan_events);
+               mu_ev->mhba = mhba;
+               mu_ev->event = msg;
+               mu_ev->param = NULL;
+               schedule_work(&mu_ev->work_q);
+       }
+}
+
+static void mvumi_handle_clob(struct mvumi_hba *mhba)
+{
+       struct mvumi_rsp_frame *ob_frame;
+       struct mvumi_cmd *cmd;
+       struct mvumi_ob_data *pool;
+
+       while (!list_empty(&mhba->free_ob_list)) {
+               pool = list_first_entry(&mhba->free_ob_list,
+                                               struct mvumi_ob_data, list);
+               list_del_init(&pool->list);
+               list_add_tail(&pool->list, &mhba->ob_data_list);
+
+               ob_frame = (struct mvumi_rsp_frame *) &pool->data[0];
+               cmd = mhba->tag_cmd[ob_frame->tag];
+
+               atomic_dec(&mhba->fw_outstanding);
+               mhba->tag_cmd[ob_frame->tag] = 0;
+               tag_release_one(mhba, &mhba->tag_pool, ob_frame->tag);
+               if (cmd->scmd)
+                       mvumi_complete_cmd(mhba, cmd, ob_frame);
+               else
+                       mvumi_complete_internal_cmd(mhba, cmd, ob_frame);
+       }
+       mhba->instancet->fire_cmd(mhba, NULL);
+}
+
+static irqreturn_t mvumi_isr_handler(int irq, void *devp)
+{
+       struct mvumi_hba *mhba = (struct mvumi_hba *) devp;
+       unsigned long flags;
+
+       spin_lock_irqsave(mhba->shost->host_lock, flags);
+       if (unlikely(mhba->instancet->clear_intr(mhba) || !mhba->global_isr)) {
+               spin_unlock_irqrestore(mhba->shost->host_lock, flags);
+               return IRQ_NONE;
+       }
+
+       if (mhba->global_isr & INT_MAP_DL_CPU2PCIEA) {
+               if (mhba->isr_status & DRBL_HANDSHAKE_ISR) {
+                       dev_warn(&mhba->pdev->dev, "enter handshake again!\n");
+                       mvumi_handshake(mhba);
+               }
+               if (mhba->isr_status & DRBL_EVENT_NOTIFY)
+                       mvumi_launch_events(mhba, APICDB1_EVENT_GETEVENT);
+       }
+
+       if (mhba->global_isr & INT_MAP_COMAOUT)
+               mvumi_receive_ob_list_entry(mhba);
+
+       mhba->global_isr = 0;
+       mhba->isr_status = 0;
+       if (mhba->fw_state == FW_STATE_STARTED)
+               mvumi_handle_clob(mhba);
+       spin_unlock_irqrestore(mhba->shost->host_lock, flags);
+       return IRQ_HANDLED;
+}
+
+static enum mvumi_qc_result mvumi_send_command(struct mvumi_hba *mhba,
+                                               struct mvumi_cmd *cmd)
+{
+       void *ib_entry;
+       struct mvumi_msg_frame *ib_frame;
+       unsigned int frame_len;
+
+       ib_frame = cmd->frame;
+       if (unlikely(mhba->fw_state != FW_STATE_STARTED)) {
+               dev_dbg(&mhba->pdev->dev, "firmware not ready.\n");
+               return MV_QUEUE_COMMAND_RESULT_NO_RESOURCE;
+       }
+       if (tag_is_empty(&mhba->tag_pool)) {
+               dev_dbg(&mhba->pdev->dev, "no free tag.\n");
+               return MV_QUEUE_COMMAND_RESULT_NO_RESOURCE;
+       }
+       if (mvumi_get_ib_list_entry(mhba, &ib_entry))
+               return MV_QUEUE_COMMAND_RESULT_NO_RESOURCE;
+
+       cmd->frame->tag = tag_get_one(mhba, &mhba->tag_pool);
+       cmd->frame->request_id = mhba->io_seq++;
+       cmd->request_id = cmd->frame->request_id;
+       mhba->tag_cmd[cmd->frame->tag] = cmd;
+       frame_len = sizeof(*ib_frame) - 4 +
+                               ib_frame->sg_counts * sizeof(struct mvumi_sgl);
+       memcpy(ib_entry, ib_frame, frame_len);
+       return MV_QUEUE_COMMAND_RESULT_SENT;
+}
+
+static void mvumi_fire_cmd(struct mvumi_hba *mhba, struct mvumi_cmd *cmd)
+{
+       unsigned short num_of_cl_sent = 0;
+       enum mvumi_qc_result result;
+
+       if (cmd)
+               list_add_tail(&cmd->queue_pointer, &mhba->waiting_req_list);
+
+       while (!list_empty(&mhba->waiting_req_list)) {
+               cmd = list_first_entry(&mhba->waiting_req_list,
+                                        struct mvumi_cmd, queue_pointer);
+               list_del_init(&cmd->queue_pointer);
+               result = mvumi_send_command(mhba, cmd);
+               switch (result) {
+               case MV_QUEUE_COMMAND_RESULT_SENT:
+                       num_of_cl_sent++;
+                       break;
+               case MV_QUEUE_COMMAND_RESULT_NO_RESOURCE:
+                       list_add(&cmd->queue_pointer, &mhba->waiting_req_list);
+                       if (num_of_cl_sent > 0)
+                               mvumi_send_ib_list_entry(mhba);
+
+                       return;
+               }
+       }
+       if (num_of_cl_sent > 0)
+               mvumi_send_ib_list_entry(mhba);
+}
+
+/**
+ * mvumi_enable_intr - Enables interrupts
+ * @regs:                      FW register set
+ */
+static void mvumi_enable_intr(void *regs)
+{
+       unsigned int mask;
+
+       iowrite32(0x3FFFFFFF, regs + CPU_ARM_TO_PCIEA_MASK_REG);
+       mask = ioread32(regs + CPU_ENPOINTA_MASK_REG);
+       mask |= INT_MAP_DL_CPU2PCIEA | INT_MAP_COMAOUT | INT_MAP_COMAERR;
+       iowrite32(mask, regs + CPU_ENPOINTA_MASK_REG);
+}
+
+/**
+ * mvumi_disable_intr -Disables interrupt
+ * @regs:                      FW register set
+ */
+static void mvumi_disable_intr(void *regs)
+{
+       unsigned int mask;
+
+       iowrite32(0, regs + CPU_ARM_TO_PCIEA_MASK_REG);
+       mask = ioread32(regs + CPU_ENPOINTA_MASK_REG);
+       mask &= ~(INT_MAP_DL_CPU2PCIEA | INT_MAP_COMAOUT | INT_MAP_COMAERR);
+       iowrite32(mask, regs + CPU_ENPOINTA_MASK_REG);
+}
+
+static int mvumi_clear_intr(void *extend)
+{
+       struct mvumi_hba *mhba = (struct mvumi_hba *) extend;
+       unsigned int status, isr_status = 0, tmp = 0;
+       void *regs = mhba->mmio;
+
+       status = ioread32(regs + CPU_MAIN_INT_CAUSE_REG);
+       if (!(status & INT_MAP_MU) || status == 0xFFFFFFFF)
+               return 1;
+       if (unlikely(status & INT_MAP_COMAERR)) {
+               tmp = ioread32(regs + CLA_ISR_CAUSE);
+               if (tmp & (CLIC_IN_ERR_IRQ | CLIC_OUT_ERR_IRQ))
+                       iowrite32(tmp & (CLIC_IN_ERR_IRQ | CLIC_OUT_ERR_IRQ),
+                                       regs + CLA_ISR_CAUSE);
+               status ^= INT_MAP_COMAERR;
+               /* inbound or outbound parity error, command will timeout */
+       }
+       if (status & INT_MAP_COMAOUT) {
+               tmp = ioread32(regs + CLA_ISR_CAUSE);
+               if (tmp & CLIC_OUT_IRQ)
+                       iowrite32(tmp & CLIC_OUT_IRQ, regs + CLA_ISR_CAUSE);
+       }
+       if (status & INT_MAP_DL_CPU2PCIEA) {
+               isr_status = ioread32(regs + CPU_ARM_TO_PCIEA_DRBL_REG);
+               if (isr_status)
+                       iowrite32(isr_status, regs + CPU_ARM_TO_PCIEA_DRBL_REG);
+       }
+
+       mhba->global_isr = status;
+       mhba->isr_status = isr_status;
+
+       return 0;
+}
+
+/**
+ * mvumi_read_fw_status_reg - returns the current FW status value
+ * @regs:                      FW register set
+ */
+static unsigned int mvumi_read_fw_status_reg(void *regs)
+{
+       unsigned int status;
+
+       status = ioread32(regs + CPU_ARM_TO_PCIEA_DRBL_REG);
+       if (status)
+               iowrite32(status, regs + CPU_ARM_TO_PCIEA_DRBL_REG);
+       return status;
+}
+
+static struct mvumi_instance_template mvumi_instance_template = {
+       .fire_cmd = mvumi_fire_cmd,
+       .enable_intr = mvumi_enable_intr,
+       .disable_intr = mvumi_disable_intr,
+       .clear_intr = mvumi_clear_intr,
+       .read_fw_status_reg = mvumi_read_fw_status_reg,
+};
+
+static int mvumi_slave_configure(struct scsi_device *sdev)
+{
+       struct mvumi_hba *mhba;
+       unsigned char bitcount = sizeof(unsigned char) * 8;
+
+       mhba = (struct mvumi_hba *) sdev->host->hostdata;
+       if (sdev->id >= mhba->max_target_id)
+               return -EINVAL;
+
+       mhba->target_map[sdev->id / bitcount] |= (1 << (sdev->id % bitcount));
+       return 0;
+}
+
+/**
+ * mvumi_build_frame - Prepares a direct cdb (DCDB) command
+ * @mhba:              Adapter soft state
+ * @scmd:              SCSI command
+ * @cmd:               Command to be prepared in
+ *
+ * This function prepares CDB commands. These are typcially pass-through
+ * commands to the devices.
+ */
+static unsigned char mvumi_build_frame(struct mvumi_hba *mhba,
+                               struct scsi_cmnd *scmd, struct mvumi_cmd *cmd)
+{
+       struct mvumi_msg_frame *pframe;
+
+       cmd->scmd = scmd;
+       cmd->cmd_status = REQ_STATUS_PENDING;
+       pframe = cmd->frame;
+       pframe->device_id = ((unsigned short) scmd->device->id) |
+                               (((unsigned short) scmd->device->lun) << 8);
+       pframe->cmd_flag = 0;
+
+       switch (scmd->sc_data_direction) {
+       case DMA_NONE:
+               pframe->cmd_flag |= CMD_FLAG_NON_DATA;
+               break;
+       case DMA_FROM_DEVICE:
+               pframe->cmd_flag |= CMD_FLAG_DATA_IN;
+               break;
+       case DMA_TO_DEVICE:
+               pframe->cmd_flag |= CMD_FLAG_DATA_OUT;
+               break;
+       case DMA_BIDIRECTIONAL:
+       default:
+               dev_warn(&mhba->pdev->dev, "unexpected data direction[%d] "
+                       "cmd[0x%x]\n", scmd->sc_data_direction, scmd->cmnd[0]);
+               goto error;
+       }
+
+       pframe->cdb_length = scmd->cmd_len;
+       memcpy(pframe->cdb, scmd->cmnd, pframe->cdb_length);
+       pframe->req_function = CL_FUN_SCSI_CMD;
+       if (scsi_bufflen(scmd)) {
+               if (mvumi_make_sgl(mhba, scmd, &pframe->payload[0],
+                       &pframe->sg_counts))
+                       goto error;
+
+               pframe->data_transfer_length = scsi_bufflen(scmd);
+       } else {
+               pframe->sg_counts = 0;
+               pframe->data_transfer_length = 0;
+       }
+       return 0;
+
+error:
+       scmd->result = (DID_OK << 16) | (DRIVER_SENSE << 24) |
+               SAM_STAT_CHECK_CONDITION;
+       scsi_build_sense_buffer(0, scmd->sense_buffer, ILLEGAL_REQUEST, 0x24,
+                                                                       0);
+       return -1;
+}
+
+/**
+ * mvumi_queue_command -       Queue entry point
+ * @scmd:                      SCSI command to be queued
+ * @done:                      Callback entry point
+ */
+static int mvumi_queue_command(struct Scsi_Host *shost,
+                                       struct scsi_cmnd *scmd)
+{
+       struct mvumi_cmd *cmd;
+       struct mvumi_hba *mhba;
+       unsigned long irq_flags;
+
+       spin_lock_irqsave(shost->host_lock, irq_flags);
+       scsi_cmd_get_serial(shost, scmd);
+
+       mhba = (struct mvumi_hba *) shost->hostdata;
+       scmd->result = 0;
+       cmd = mvumi_get_cmd(mhba);
+       if (unlikely(!cmd)) {
+               spin_unlock_irqrestore(shost->host_lock, irq_flags);
+               return SCSI_MLQUEUE_HOST_BUSY;
+       }
+
+       if (unlikely(mvumi_build_frame(mhba, scmd, cmd)))
+               goto out_return_cmd;
+
+       cmd->scmd = scmd;
+       scmd->SCp.ptr = (char *) cmd;
+       mhba->instancet->fire_cmd(mhba, cmd);
+       spin_unlock_irqrestore(shost->host_lock, irq_flags);
+       return 0;
+
+out_return_cmd:
+       mvumi_return_cmd(mhba, cmd);
+       scmd->scsi_done(scmd);
+       spin_unlock_irqrestore(shost->host_lock, irq_flags);
+       return 0;
+}
+
+static enum blk_eh_timer_return mvumi_timed_out(struct scsi_cmnd *scmd)
+{
+       struct mvumi_cmd *cmd = (struct mvumi_cmd *) scmd->SCp.ptr;
+       struct Scsi_Host *host = scmd->device->host;
+       struct mvumi_hba *mhba = shost_priv(host);
+       unsigned long flags;
+
+       spin_lock_irqsave(mhba->shost->host_lock, flags);
+
+       if (mhba->tag_cmd[cmd->frame->tag]) {
+               mhba->tag_cmd[cmd->frame->tag] = 0;
+               tag_release_one(mhba, &mhba->tag_pool, cmd->frame->tag);
+       }
+       if (!list_empty(&cmd->queue_pointer))
+               list_del_init(&cmd->queue_pointer);
+       else
+               atomic_dec(&mhba->fw_outstanding);
+
+       scmd->result = (DRIVER_INVALID << 24) | (DID_ABORT << 16);
+       scmd->SCp.ptr = NULL;
+       if (scsi_bufflen(scmd)) {
+               if (scsi_sg_count(scmd)) {
+                       pci_unmap_sg(mhba->pdev,
+                               scsi_sglist(scmd),
+                               scsi_sg_count(scmd),
+                               (int)scmd->sc_data_direction);
+               } else {
+                       pci_unmap_single(mhba->pdev,
+                               scmd->SCp.dma_handle,
+                               scsi_bufflen(scmd),
+                               (int)scmd->sc_data_direction);
+
+                       scmd->SCp.dma_handle = 0;
+               }
+       }
+       mvumi_return_cmd(mhba, cmd);
+       spin_unlock_irqrestore(mhba->shost->host_lock, flags);
+
+       return BLK_EH_NOT_HANDLED;
+}
+
+static int
+mvumi_bios_param(struct scsi_device *sdev, struct block_device *bdev,
+                       sector_t capacity, int geom[])
+{
+       int heads, sectors;
+       sector_t cylinders;
+       unsigned long tmp;
+
+       heads = 64;
+       sectors = 32;
+       tmp = heads * sectors;
+       cylinders = capacity;
+       sector_div(cylinders, tmp);
+
+       if (capacity >= 0x200000) {
+               heads = 255;
+               sectors = 63;
+               tmp = heads * sectors;
+               cylinders = capacity;
+               sector_div(cylinders, tmp);
+       }
+       geom[0] = heads;
+       geom[1] = sectors;
+       geom[2] = cylinders;
+
+       return 0;
+}
+
+static struct scsi_host_template mvumi_template = {
+
+       .module = THIS_MODULE,
+       .name = "Marvell Storage Controller",
+       .slave_configure = mvumi_slave_configure,
+       .queuecommand = mvumi_queue_command,
+       .eh_host_reset_handler = mvumi_host_reset,
+       .bios_param = mvumi_bios_param,
+       .this_id = -1,
+};
+
+static struct scsi_transport_template mvumi_transport_template = {
+       .eh_timed_out = mvumi_timed_out,
+};
+
+/**
+ * mvumi_init_fw -     Initializes the FW
+ * @mhba:              Adapter soft state
+ *
+ * This is the main function for initializing firmware.
+ */
+static int mvumi_init_fw(struct mvumi_hba *mhba)
+{
+       int ret = 0;
+
+       if (pci_request_regions(mhba->pdev, MV_DRIVER_NAME)) {
+               dev_err(&mhba->pdev->dev, "IO memory region busy!\n");
+               return -EBUSY;
+       }
+       ret = mvumi_map_pci_addr(mhba->pdev, mhba->base_addr);
+       if (ret)
+               goto fail_ioremap;
+
+       mhba->mmio = mhba->base_addr[0];
+
+       switch (mhba->pdev->device) {
+       case PCI_DEVICE_ID_MARVELL_MV9143:
+               mhba->instancet = &mvumi_instance_template;
+               mhba->io_seq = 0;
+               mhba->max_sge = MVUMI_MAX_SG_ENTRY;
+               mhba->request_id_enabled = 1;
+               break;
+       default:
+               dev_err(&mhba->pdev->dev, "device 0x%x not supported!\n",
+                                                       mhba->pdev->device);
+               mhba->instancet = NULL;
+               ret = -EINVAL;
+               goto fail_alloc_mem;
+       }
+       dev_dbg(&mhba->pdev->dev, "device id : %04X is found.\n",
+                                                       mhba->pdev->device);
+
+       mhba->handshake_page = kzalloc(HSP_MAX_SIZE, GFP_KERNEL);
+       if (!mhba->handshake_page) {
+               dev_err(&mhba->pdev->dev,
+                       "failed to allocate memory for handshake\n");
+               ret = -ENOMEM;
+               goto fail_alloc_mem;
+       }
+       mhba->handshake_page_phys = virt_to_phys(mhba->handshake_page);
+
+       if (mvumi_start(mhba)) {
+               ret = -EINVAL;
+               goto fail_ready_state;
+       }
+       ret = mvumi_alloc_cmds(mhba);
+       if (ret)
+               goto fail_ready_state;
+
+       return 0;
+
+fail_ready_state:
+       mvumi_release_mem_resource(mhba);
+       kfree(mhba->handshake_page);
+fail_alloc_mem:
+       mvumi_unmap_pci_addr(mhba->pdev, mhba->base_addr);
+fail_ioremap:
+       pci_release_regions(mhba->pdev);
+
+       return ret;
+}
+
+/**
+ * mvumi_io_attach -   Attaches this driver to SCSI mid-layer
+ * @mhba:              Adapter soft state
+ */
+static int mvumi_io_attach(struct mvumi_hba *mhba)
+{
+       struct Scsi_Host *host = mhba->shost;
+       int ret;
+       unsigned int max_sg = (mhba->ib_max_size + 4 -
+               sizeof(struct mvumi_msg_frame)) / sizeof(struct mvumi_sgl);
+
+       host->irq = mhba->pdev->irq;
+       host->unique_id = mhba->unique_id;
+       host->can_queue = (mhba->max_io - 1) ? (mhba->max_io - 1) : 1;
+       host->sg_tablesize = mhba->max_sge > max_sg ? max_sg : mhba->max_sge;
+       host->max_sectors = mhba->max_transfer_size / 512;
+       host->cmd_per_lun =  (mhba->max_io - 1) ? (mhba->max_io - 1) : 1;
+       host->max_id = mhba->max_target_id;
+       host->max_cmd_len = MAX_COMMAND_SIZE;
+       host->transportt = &mvumi_transport_template;
+
+       ret = scsi_add_host(host, &mhba->pdev->dev);
+       if (ret) {
+               dev_err(&mhba->pdev->dev, "scsi_add_host failed\n");
+               return ret;
+       }
+       mhba->fw_flag |= MVUMI_FW_ATTACH;
+       scsi_scan_host(host);
+
+       return 0;
+}
+
+/**
+ * mvumi_probe_one -   PCI hotplug entry point
+ * @pdev:              PCI device structure
+ * @id:                        PCI ids of supported hotplugged adapter
+ */
+static int __devinit mvumi_probe_one(struct pci_dev *pdev,
+                                       const struct pci_device_id *id)
+{
+       struct Scsi_Host *host;
+       struct mvumi_hba *mhba;
+       int ret;
+
+       dev_dbg(&pdev->dev, " %#4.04x:%#4.04x:%#4.04x:%#4.04x: ",
+                       pdev->vendor, pdev->device, pdev->subsystem_vendor,
+                       pdev->subsystem_device);
+
+       ret = pci_enable_device(pdev);
+       if (ret)
+               return ret;
+
+       pci_set_master(pdev);
+
+       if (IS_DMA64) {
+               ret = pci_set_dma_mask(pdev, DMA_BIT_MASK(64));
+               if (ret) {
+                       ret = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
+                       if (ret)
+                               goto fail_set_dma_mask;
+               }
+       } else {
+               ret = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
+               if (ret)
+                       goto fail_set_dma_mask;
+       }
+
+       host = scsi_host_alloc(&mvumi_template, sizeof(*mhba));
+       if (!host) {
+               dev_err(&pdev->dev, "scsi_host_alloc failed\n");
+               ret = -ENOMEM;
+               goto fail_alloc_instance;
+       }
+       mhba = shost_priv(host);
+
+       INIT_LIST_HEAD(&mhba->cmd_pool);
+       INIT_LIST_HEAD(&mhba->ob_data_list);
+       INIT_LIST_HEAD(&mhba->free_ob_list);
+       INIT_LIST_HEAD(&mhba->res_list);
+       INIT_LIST_HEAD(&mhba->waiting_req_list);
+       atomic_set(&mhba->fw_outstanding, 0);
+       init_waitqueue_head(&mhba->int_cmd_wait_q);
+
+       mhba->pdev = pdev;
+       mhba->shost = host;
+       mhba->unique_id = pdev->bus->number << 8 | pdev->devfn;
+
+       ret = mvumi_init_fw(mhba);
+       if (ret)
+               goto fail_init_fw;
+
+       ret = request_irq(mhba->pdev->irq, mvumi_isr_handler, IRQF_SHARED,
+                               "mvumi", mhba);
+       if (ret) {
+               dev_err(&pdev->dev, "failed to register IRQ\n");
+               goto fail_init_irq;
+       }
+       mhba->instancet->enable_intr(mhba->mmio);
+       pci_set_drvdata(pdev, mhba);
+
+       ret = mvumi_io_attach(mhba);
+       if (ret)
+               goto fail_io_attach;
+       dev_dbg(&pdev->dev, "probe mvumi driver successfully.\n");
+
+       return 0;
+
+fail_io_attach:
+       pci_set_drvdata(pdev, NULL);
+       mhba->instancet->disable_intr(mhba->mmio);
+       free_irq(mhba->pdev->irq, mhba);
+fail_init_irq:
+       mvumi_release_fw(mhba);
+fail_init_fw:
+       scsi_host_put(host);
+
+fail_alloc_instance:
+fail_set_dma_mask:
+       pci_disable_device(pdev);
+
+       return ret;
+}
+
+static void mvumi_detach_one(struct pci_dev *pdev)
+{
+       struct Scsi_Host *host;
+       struct mvumi_hba *mhba;
+
+       mhba = pci_get_drvdata(pdev);
+       host = mhba->shost;
+       scsi_remove_host(mhba->shost);
+       mvumi_flush_cache(mhba);
+
+       mhba->instancet->disable_intr(mhba->mmio);
+       free_irq(mhba->pdev->irq, mhba);
+       mvumi_release_fw(mhba);
+       scsi_host_put(host);
+       pci_set_drvdata(pdev, NULL);
+       pci_disable_device(pdev);
+       dev_dbg(&pdev->dev, "driver is removed!\n");
+}
+
+/**
+ * mvumi_shutdown -    Shutdown entry point
+ * @device:            Generic device structure
+ */
+static void mvumi_shutdown(struct pci_dev *pdev)
+{
+       struct mvumi_hba *mhba = pci_get_drvdata(pdev);
+
+       mvumi_flush_cache(mhba);
+}
+
+static int mvumi_suspend(struct pci_dev *pdev, pm_message_t state)
+{
+       struct mvumi_hba *mhba = NULL;
+
+       mhba = pci_get_drvdata(pdev);
+       mvumi_flush_cache(mhba);
+
+       pci_set_drvdata(pdev, mhba);
+       mhba->instancet->disable_intr(mhba->mmio);
+       free_irq(mhba->pdev->irq, mhba);
+       mvumi_unmap_pci_addr(pdev, mhba->base_addr);
+       pci_release_regions(pdev);
+       pci_save_state(pdev);
+       pci_disable_device(pdev);
+       pci_set_power_state(pdev, pci_choose_state(pdev, state));
+
+       return 0;
+}
+
+static int mvumi_resume(struct pci_dev *pdev)
+{
+       int ret;
+       struct mvumi_hba *mhba = NULL;
+
+       mhba = pci_get_drvdata(pdev);
+
+       pci_set_power_state(pdev, PCI_D0);
+       pci_enable_wake(pdev, PCI_D0, 0);
+       pci_restore_state(pdev);
+
+       ret = pci_enable_device(pdev);
+       if (ret) {
+               dev_err(&pdev->dev, "enable device failed\n");
+               return ret;
+       }
+       pci_set_master(pdev);
+       if (IS_DMA64) {
+               ret = pci_set_dma_mask(pdev, DMA_BIT_MASK(64));
+               if (ret) {
+                       ret = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
+                       if (ret)
+                               goto fail;
+               }
+       } else {
+               ret = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
+               if (ret)
+                       goto fail;
+       }
+       ret = pci_request_regions(mhba->pdev, MV_DRIVER_NAME);
+       if (ret)
+               goto fail;
+       ret = mvumi_map_pci_addr(mhba->pdev, mhba->base_addr);
+       if (ret)
+               goto release_regions;
+
+       mhba->mmio = mhba->base_addr[0];
+       mvumi_reset(mhba->mmio);
+
+       if (mvumi_start(mhba)) {
+               ret = -EINVAL;
+               goto unmap_pci_addr;
+       }
+
+       ret = request_irq(mhba->pdev->irq, mvumi_isr_handler, IRQF_SHARED,
+                               "mvumi", mhba);
+       if (ret) {
+               dev_err(&pdev->dev, "failed to register IRQ\n");
+               goto unmap_pci_addr;
+       }
+       mhba->instancet->enable_intr(mhba->mmio);
+
+       return 0;
+
+unmap_pci_addr:
+       mvumi_unmap_pci_addr(pdev, mhba->base_addr);
+release_regions:
+       pci_release_regions(pdev);
+fail:
+       pci_disable_device(pdev);
+
+       return ret;
+}
+
+static struct pci_driver mvumi_pci_driver = {
+
+       .name = MV_DRIVER_NAME,
+       .id_table = mvumi_pci_table,
+       .probe = mvumi_probe_one,
+       .remove = __devexit_p(mvumi_detach_one),
+       .shutdown = mvumi_shutdown,
+#ifdef CONFIG_PM
+       .suspend = mvumi_suspend,
+       .resume = mvumi_resume,
+#endif
+};
+
+/**
+ * mvumi_init - Driver load entry point
+ */
+static int __init mvumi_init(void)
+{
+       return pci_register_driver(&mvumi_pci_driver);
+}
+
+/**
+ * mvumi_exit - Driver unload entry point
+ */
+static void __exit mvumi_exit(void)
+{
+
+       pci_unregister_driver(&mvumi_pci_driver);
+}
+
+module_init(mvumi_init);
+module_exit(mvumi_exit);
diff --git a/drivers/scsi/mvumi.h b/drivers/scsi/mvumi.h
new file mode 100644 (file)
index 0000000..10b9237
--- /dev/null
@@ -0,0 +1,505 @@
+/*
+  * Marvell UMI head file
+  *
+  * Copyright 2011 Marvell. <jyli@marvell.com>
+  *
+  * This file is licensed under GPLv2.
+  *
+  * 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; version 2 of the
+  * License.
+  *
+  * 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
+ */
+
+#ifndef MVUMI_H
+#define MVUMI_H
+
+#define MAX_BASE_ADDRESS       6
+
+#define VER_MAJOR              1
+#define VER_MINOR              1
+#define VER_OEM                        0
+#define VER_BUILD              1500
+
+#define MV_DRIVER_NAME                 "mvumi"
+#define PCI_VENDOR_ID_MARVELL_2                0x1b4b
+#define PCI_DEVICE_ID_MARVELL_MV9143   0x9143
+
+#define MVUMI_INTERNAL_CMD_WAIT_TIME   45
+
+#define IS_DMA64                       (sizeof(dma_addr_t) == 8)
+
+enum mvumi_qc_result {
+       MV_QUEUE_COMMAND_RESULT_SENT    = 0,
+       MV_QUEUE_COMMAND_RESULT_NO_RESOURCE,
+};
+
+enum {
+       /*******************************************/
+
+       /* ARM Mbus Registers Map       */
+
+       /*******************************************/
+       CPU_MAIN_INT_CAUSE_REG  = 0x20200,
+       CPU_MAIN_IRQ_MASK_REG   = 0x20204,
+       CPU_MAIN_FIQ_MASK_REG   = 0x20208,
+       CPU_ENPOINTA_MASK_REG   = 0x2020C,
+       CPU_ENPOINTB_MASK_REG   = 0x20210,
+
+       INT_MAP_COMAERR         = 1 << 6,
+       INT_MAP_COMAIN          = 1 << 7,
+       INT_MAP_COMAOUT         = 1 << 8,
+       INT_MAP_COMBERR         = 1 << 9,
+       INT_MAP_COMBIN          = 1 << 10,
+       INT_MAP_COMBOUT         = 1 << 11,
+
+       INT_MAP_COMAINT = (INT_MAP_COMAOUT | INT_MAP_COMAERR),
+       INT_MAP_COMBINT = (INT_MAP_COMBOUT | INT_MAP_COMBIN | INT_MAP_COMBERR),
+
+       INT_MAP_DL_PCIEA2CPU    = 1 << 0,
+       INT_MAP_DL_CPU2PCIEA    = 1 << 1,
+
+       /***************************************/
+
+       /* ARM Doorbell Registers Map           */
+
+       /***************************************/
+       CPU_PCIEA_TO_ARM_DRBL_REG       = 0x20400,
+       CPU_PCIEA_TO_ARM_MASK_REG       = 0x20404,
+       CPU_ARM_TO_PCIEA_DRBL_REG       = 0x20408,
+       CPU_ARM_TO_PCIEA_MASK_REG       = 0x2040C,
+
+       DRBL_HANDSHAKE                  = 1 << 0,
+       DRBL_SOFT_RESET                 = 1 << 1,
+       DRBL_BUS_CHANGE                 = 1 << 2,
+       DRBL_EVENT_NOTIFY               = 1 << 3,
+       DRBL_MU_RESET                   = 1 << 4,
+       DRBL_HANDSHAKE_ISR              = DRBL_HANDSHAKE,
+
+       CPU_PCIEA_TO_ARM_MSG0           = 0x20430,
+       CPU_PCIEA_TO_ARM_MSG1           = 0x20434,
+       CPU_ARM_TO_PCIEA_MSG0           = 0x20438,
+       CPU_ARM_TO_PCIEA_MSG1           = 0x2043C,
+
+       /*******************************************/
+
+       /* ARM Communication List Registers Map    */
+
+       /*******************************************/
+       CLA_INB_LIST_BASEL              = 0x500,
+       CLA_INB_LIST_BASEH              = 0x504,
+       CLA_INB_AVAL_COUNT_BASEL        = 0x508,
+       CLA_INB_AVAL_COUNT_BASEH        = 0x50C,
+       CLA_INB_DESTI_LIST_BASEL        = 0x510,
+       CLA_INB_DESTI_LIST_BASEH        = 0x514,
+       CLA_INB_WRITE_POINTER           = 0x518,
+       CLA_INB_READ_POINTER            = 0x51C,
+
+       CLA_OUTB_LIST_BASEL             = 0x530,
+       CLA_OUTB_LIST_BASEH             = 0x534,
+       CLA_OUTB_SOURCE_LIST_BASEL      = 0x538,
+       CLA_OUTB_SOURCE_LIST_BASEH      = 0x53C,
+       CLA_OUTB_COPY_POINTER           = 0x544,
+       CLA_OUTB_READ_POINTER           = 0x548,
+
+       CLA_ISR_CAUSE                   = 0x560,
+       CLA_ISR_MASK                    = 0x564,
+
+       INT_MAP_MU              = (INT_MAP_DL_CPU2PCIEA | INT_MAP_COMAINT),
+
+       CL_POINTER_TOGGLE               = 1 << 12,
+
+       CLIC_IN_IRQ                     = 1 << 0,
+       CLIC_OUT_IRQ                    = 1 << 1,
+       CLIC_IN_ERR_IRQ                 = 1 << 8,
+       CLIC_OUT_ERR_IRQ                = 1 << 12,
+
+       CL_SLOT_NUM_MASK                = 0xFFF,
+
+       /*
+       * Command flag is the flag for the CDB command itself
+       */
+       /* 1-non data; 0-data command */
+       CMD_FLAG_NON_DATA               = 1 << 0,
+       CMD_FLAG_DMA                    = 1 << 1,
+       CMD_FLAG_PIO                    = 1 << 2,
+       /* 1-host read data */
+       CMD_FLAG_DATA_IN                = 1 << 3,
+       /* 1-host write data */
+       CMD_FLAG_DATA_OUT               = 1 << 4,
+
+       SCSI_CMD_MARVELL_SPECIFIC       = 0xE1,
+       CDB_CORE_SHUTDOWN               = 0xB,
+};
+
+#define APICDB0_EVENT                  0xF4
+#define APICDB1_EVENT_GETEVENT         0
+#define MAX_EVENTS_RETURNED            6
+
+struct mvumi_driver_event {
+       u32     time_stamp;
+       u32     sequence_no;
+       u32     event_id;
+       u8      severity;
+       u8      param_count;
+       u16     device_id;
+       u32     params[4];
+       u8      sense_data_length;
+       u8      Reserved1;
+       u8      sense_data[30];
+};
+
+struct mvumi_event_req {
+       unsigned char   count;
+       unsigned char   reserved[3];
+       struct mvumi_driver_event  events[MAX_EVENTS_RETURNED];
+};
+
+struct mvumi_events_wq {
+       struct work_struct work_q;
+       struct mvumi_hba *mhba;
+       unsigned int event;
+       void *param;
+};
+
+#define MVUMI_MAX_SG_ENTRY     32
+#define SGD_EOT                        (1L << 27)
+
+struct mvumi_sgl {
+       u32     baseaddr_l;
+       u32     baseaddr_h;
+       u32     flags;
+       u32     size;
+};
+
+struct mvumi_res {
+       struct list_head entry;
+       dma_addr_t bus_addr;
+       void *virt_addr;
+       unsigned int size;
+       unsigned short type;    /* enum Resource_Type */
+};
+
+/* Resource type */
+enum resource_type {
+       RESOURCE_CACHED_MEMORY = 0,
+       RESOURCE_UNCACHED_MEMORY
+};
+
+struct mvumi_sense_data {
+       u8 error_eode:7;
+       u8 valid:1;
+       u8 segment_number;
+       u8 sense_key:4;
+       u8 reserved:1;
+       u8 incorrect_length:1;
+       u8 end_of_media:1;
+       u8 file_mark:1;
+       u8 information[4];
+       u8 additional_sense_length;
+       u8 command_specific_information[4];
+       u8 additional_sense_code;
+       u8 additional_sense_code_qualifier;
+       u8 field_replaceable_unit_code;
+       u8 sense_key_specific[3];
+};
+
+/* Request initiator must set the status to REQ_STATUS_PENDING. */
+#define REQ_STATUS_PENDING             0x80
+
+struct mvumi_cmd {
+       struct list_head queue_pointer;
+       struct mvumi_msg_frame *frame;
+       struct scsi_cmnd *scmd;
+       atomic_t sync_cmd;
+       void *data_buf;
+       unsigned short request_id;
+       unsigned char cmd_status;
+};
+
+/*
+ * the function type of the in bound frame
+ */
+#define CL_FUN_SCSI_CMD                        0x1
+
+struct mvumi_msg_frame {
+       u16 device_id;
+       u16 tag;
+       u8 cmd_flag;
+       u8 req_function;
+       u8 cdb_length;
+       u8 sg_counts;
+       u32 data_transfer_length;
+       u16 request_id;
+       u16 reserved1;
+       u8 cdb[MAX_COMMAND_SIZE];
+       u32 payload[1];
+};
+
+/*
+ * the respond flag for data_payload of the out bound frame
+ */
+#define CL_RSP_FLAG_NODATA             0x0
+#define CL_RSP_FLAG_SENSEDATA          0x1
+
+struct mvumi_rsp_frame {
+       u16 device_id;
+       u16 tag;
+       u8 req_status;
+       u8 rsp_flag;    /* Indicates the type of Data_Payload.*/
+       u16 request_id;
+       u32 payload[1];
+};
+
+struct mvumi_ob_data {
+       struct list_head list;
+       unsigned char data[0];
+};
+
+struct version_info {
+       u32 ver_major;
+       u32 ver_minor;
+       u32 ver_oem;
+       u32 ver_build;
+};
+
+#define FW_MAX_DELAY                   30
+#define MVUMI_FW_BUSY                  (1U << 0)
+#define MVUMI_FW_ATTACH                        (1U << 1)
+#define MVUMI_FW_ALLOC                 (1U << 2)
+
+/*
+ * State is the state of the MU
+ */
+#define FW_STATE_IDLE                  0
+#define FW_STATE_STARTING              1
+#define FW_STATE_HANDSHAKING           2
+#define FW_STATE_STARTED               3
+#define FW_STATE_ABORT                 4
+
+#define HANDSHAKE_SIGNATURE            0x5A5A5A5AL
+#define HANDSHAKE_READYSTATE           0x55AA5AA5L
+#define HANDSHAKE_DONESTATE            0x55AAA55AL
+
+/* HandShake Status definition */
+#define HS_STATUS_OK                   1
+#define HS_STATUS_ERR                  2
+#define HS_STATUS_INVALID              3
+
+/* HandShake State/Cmd definition */
+#define HS_S_START                     1
+#define HS_S_RESET                     2
+#define HS_S_PAGE_ADDR                 3
+#define HS_S_QUERY_PAGE                        4
+#define HS_S_SEND_PAGE                 5
+#define HS_S_END                       6
+#define HS_S_ABORT                     7
+#define HS_PAGE_VERIFY_SIZE            128
+
+#define HS_GET_STATE(a)                        (a & 0xFFFF)
+#define HS_GET_STATUS(a)               ((a & 0xFFFF0000) >> 16)
+#define HS_SET_STATE(a, b)             (a |= (b & 0xFFFF))
+#define HS_SET_STATUS(a, b)            (a |= ((b & 0xFFFF) << 16))
+
+/* handshake frame */
+struct mvumi_hs_frame {
+       u16 size;
+       /* host information */
+       u8 host_type;
+       u8 reserved_1[1];
+       struct version_info host_ver; /* bios or driver version */
+
+       /* controller information */
+       u32 system_io_bus;
+       u32 slot_number;
+       u32 intr_level;
+       u32 intr_vector;
+
+       /* communication list configuration */
+       u32 ib_baseaddr_l;
+       u32 ib_baseaddr_h;
+       u32 ob_baseaddr_l;
+       u32 ob_baseaddr_h;
+
+       u8 ib_entry_size;
+       u8 ob_entry_size;
+       u8 ob_depth;
+       u8 ib_depth;
+
+       /* system time */
+       u64 seconds_since1970;
+};
+
+struct mvumi_hs_header {
+       u8      page_code;
+       u8      checksum;
+       u16     frame_length;
+       u32     frame_content[1];
+};
+
+/*
+ * the page code type of the handshake header
+ */
+#define HS_PAGE_FIRM_CAP       0x1
+#define HS_PAGE_HOST_INFO      0x2
+#define HS_PAGE_FIRM_CTL       0x3
+#define HS_PAGE_CL_INFO                0x4
+#define HS_PAGE_TOTAL          0x5
+
+#define HSP_SIZE(i)    sizeof(struct mvumi_hs_page##i)
+
+#define HSP_MAX_SIZE ({                                        \
+       int size, m1, m2;                               \
+       m1 = max(HSP_SIZE(1), HSP_SIZE(3));             \
+       m2 = max(HSP_SIZE(2), HSP_SIZE(4));             \
+       size = max(m1, m2);                             \
+       size;                                           \
+})
+
+/* The format of the page code for Firmware capability */
+struct mvumi_hs_page1 {
+       u8 pagecode;
+       u8 checksum;
+       u16 frame_length;
+
+       u16 number_of_ports;
+       u16 max_devices_support;
+       u16 max_io_support;
+       u16 umi_ver;
+       u32 max_transfer_size;
+       struct version_info fw_ver;
+       u8 cl_in_max_entry_size;
+       u8 cl_out_max_entry_size;
+       u8 cl_inout_list_depth;
+       u8 total_pages;
+       u16 capability;
+       u16 reserved1;
+};
+
+/* The format of the page code for Host information */
+struct mvumi_hs_page2 {
+       u8 pagecode;
+       u8 checksum;
+       u16 frame_length;
+
+       u8 host_type;
+       u8 reserved[3];
+       struct version_info host_ver;
+       u32 system_io_bus;
+       u32 slot_number;
+       u32 intr_level;
+       u32 intr_vector;
+       u64 seconds_since1970;
+};
+
+/* The format of the page code for firmware control  */
+struct mvumi_hs_page3 {
+       u8      pagecode;
+       u8      checksum;
+       u16     frame_length;
+       u16     control;
+       u8      reserved[2];
+       u32     host_bufferaddr_l;
+       u32     host_bufferaddr_h;
+       u32     host_eventaddr_l;
+       u32     host_eventaddr_h;
+};
+
+struct mvumi_hs_page4 {
+       u8      pagecode;
+       u8      checksum;
+       u16     frame_length;
+       u32     ib_baseaddr_l;
+       u32     ib_baseaddr_h;
+       u32     ob_baseaddr_l;
+       u32     ob_baseaddr_h;
+       u8      ib_entry_size;
+       u8      ob_entry_size;
+       u8      ob_depth;
+       u8      ib_depth;
+};
+
+struct mvumi_tag {
+       unsigned short *stack;
+       unsigned short top;
+       unsigned short size;
+};
+
+struct mvumi_hba {
+       void *base_addr[MAX_BASE_ADDRESS];
+       void *mmio;
+       struct list_head cmd_pool;
+       struct Scsi_Host *shost;
+       wait_queue_head_t int_cmd_wait_q;
+       struct pci_dev *pdev;
+       unsigned int unique_id;
+       atomic_t fw_outstanding;
+       struct mvumi_instance_template *instancet;
+
+       void *ib_list;
+       dma_addr_t ib_list_phys;
+
+       void *ob_list;
+       dma_addr_t ob_list_phys;
+
+       void *ib_shadow;
+       dma_addr_t ib_shadow_phys;
+
+       void *ob_shadow;
+       dma_addr_t ob_shadow_phys;
+
+       void *handshake_page;
+       dma_addr_t handshake_page_phys;
+
+       unsigned int global_isr;
+       unsigned int isr_status;
+
+       unsigned short max_sge;
+       unsigned short max_target_id;
+       unsigned char *target_map;
+       unsigned int max_io;
+       unsigned int list_num_io;
+       unsigned int ib_max_size;
+       unsigned int ob_max_size;
+       unsigned int ib_max_size_setting;
+       unsigned int ob_max_size_setting;
+       unsigned int max_transfer_size;
+       unsigned char hba_total_pages;
+       unsigned char fw_flag;
+       unsigned char request_id_enabled;
+       unsigned short hba_capability;
+       unsigned short io_seq;
+
+       unsigned int ib_cur_slot;
+       unsigned int ob_cur_slot;
+       unsigned int fw_state;
+
+       struct list_head ob_data_list;
+       struct list_head free_ob_list;
+       struct list_head res_list;
+       struct list_head waiting_req_list;
+
+       struct mvumi_tag tag_pool;
+       struct mvumi_cmd **tag_cmd;
+};
+
+struct mvumi_instance_template {
+       void (*fire_cmd)(struct mvumi_hba *, struct mvumi_cmd *);
+       void (*enable_intr)(void *) ;
+       void (*disable_intr)(void *);
+       int (*clear_intr)(void *);
+       unsigned int (*read_fw_status_reg)(void *);
+};
+
+extern struct timezone sys_tz;
+#endif