Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / scsi / mvsas / mv_init.c
diff --git a/kernel/drivers/scsi/mvsas/mv_init.c b/kernel/drivers/scsi/mvsas/mv_init.c
new file mode 100644 (file)
index 0000000..53030b0
--- /dev/null
@@ -0,0 +1,855 @@
+/*
+ * Marvell 88SE64xx/88SE94xx pci init
+ *
+ * Copyright 2007 Red Hat, Inc.
+ * Copyright 2008 Marvell. <kewei@marvell.com>
+ * Copyright 2009-2011 Marvell. <yuxiangl@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 "mv_sas.h"
+
+int interrupt_coalescing = 0x80;
+
+static struct scsi_transport_template *mvs_stt;
+static const struct mvs_chip_info mvs_chips[] = {
+       [chip_6320] =   { 1, 2, 0x400, 17, 16, 6,  9, &mvs_64xx_dispatch, },
+       [chip_6440] =   { 1, 4, 0x400, 17, 16, 6,  9, &mvs_64xx_dispatch, },
+       [chip_6485] =   { 1, 8, 0x800, 33, 32, 6, 10, &mvs_64xx_dispatch, },
+       [chip_9180] =   { 2, 4, 0x800, 17, 64, 8,  9, &mvs_94xx_dispatch, },
+       [chip_9480] =   { 2, 4, 0x800, 17, 64, 8,  9, &mvs_94xx_dispatch, },
+       [chip_9445] =   { 1, 4, 0x800, 17, 64, 8, 11, &mvs_94xx_dispatch, },
+       [chip_9485] =   { 2, 4, 0x800, 17, 64, 8, 11, &mvs_94xx_dispatch, },
+       [chip_1300] =   { 1, 4, 0x400, 17, 16, 6,  9, &mvs_64xx_dispatch, },
+       [chip_1320] =   { 2, 4, 0x800, 17, 64, 8,  9, &mvs_94xx_dispatch, },
+};
+
+struct device_attribute *mvst_host_attrs[];
+
+#define SOC_SAS_NUM 2
+
+static struct scsi_host_template mvs_sht = {
+       .module                 = THIS_MODULE,
+       .name                   = DRV_NAME,
+       .queuecommand           = sas_queuecommand,
+       .target_alloc           = sas_target_alloc,
+       .slave_configure        = sas_slave_configure,
+       .scan_finished          = mvs_scan_finished,
+       .scan_start             = mvs_scan_start,
+       .change_queue_depth     = sas_change_queue_depth,
+       .bios_param             = sas_bios_param,
+       .can_queue              = 1,
+       .cmd_per_lun            = 1,
+       .this_id                = -1,
+       .sg_tablesize           = SG_ALL,
+       .max_sectors            = SCSI_DEFAULT_MAX_SECTORS,
+       .use_clustering         = ENABLE_CLUSTERING,
+       .eh_device_reset_handler = sas_eh_device_reset_handler,
+       .eh_bus_reset_handler   = sas_eh_bus_reset_handler,
+       .target_destroy         = sas_target_destroy,
+       .ioctl                  = sas_ioctl,
+       .shost_attrs            = mvst_host_attrs,
+       .use_blk_tags           = 1,
+       .track_queue_depth      = 1,
+};
+
+static struct sas_domain_function_template mvs_transport_ops = {
+       .lldd_dev_found         = mvs_dev_found,
+       .lldd_dev_gone          = mvs_dev_gone,
+       .lldd_execute_task      = mvs_queue_command,
+       .lldd_control_phy       = mvs_phy_control,
+
+       .lldd_abort_task        = mvs_abort_task,
+       .lldd_abort_task_set    = mvs_abort_task_set,
+       .lldd_clear_aca         = mvs_clear_aca,
+       .lldd_clear_task_set    = mvs_clear_task_set,
+       .lldd_I_T_nexus_reset   = mvs_I_T_nexus_reset,
+       .lldd_lu_reset          = mvs_lu_reset,
+       .lldd_query_task        = mvs_query_task,
+       .lldd_port_formed       = mvs_port_formed,
+       .lldd_port_deformed     = mvs_port_deformed,
+
+};
+
+static void mvs_phy_init(struct mvs_info *mvi, int phy_id)
+{
+       struct mvs_phy *phy = &mvi->phy[phy_id];
+       struct asd_sas_phy *sas_phy = &phy->sas_phy;
+
+       phy->mvi = mvi;
+       phy->port = NULL;
+       init_timer(&phy->timer);
+       sas_phy->enabled = (phy_id < mvi->chip->n_phy) ? 1 : 0;
+       sas_phy->class = SAS;
+       sas_phy->iproto = SAS_PROTOCOL_ALL;
+       sas_phy->tproto = 0;
+       sas_phy->type = PHY_TYPE_PHYSICAL;
+       sas_phy->role = PHY_ROLE_INITIATOR;
+       sas_phy->oob_mode = OOB_NOT_CONNECTED;
+       sas_phy->linkrate = SAS_LINK_RATE_UNKNOWN;
+
+       sas_phy->id = phy_id;
+       sas_phy->sas_addr = &mvi->sas_addr[0];
+       sas_phy->frame_rcvd = &phy->frame_rcvd[0];
+       sas_phy->ha = (struct sas_ha_struct *)mvi->shost->hostdata;
+       sas_phy->lldd_phy = phy;
+}
+
+static void mvs_free(struct mvs_info *mvi)
+{
+       struct mvs_wq *mwq;
+       int slot_nr;
+
+       if (!mvi)
+               return;
+
+       if (mvi->flags & MVF_FLAG_SOC)
+               slot_nr = MVS_SOC_SLOTS;
+       else
+               slot_nr = MVS_CHIP_SLOT_SZ;
+
+       if (mvi->dma_pool)
+               pci_pool_destroy(mvi->dma_pool);
+
+       if (mvi->tx)
+               dma_free_coherent(mvi->dev,
+                                 sizeof(*mvi->tx) * MVS_CHIP_SLOT_SZ,
+                                 mvi->tx, mvi->tx_dma);
+       if (mvi->rx_fis)
+               dma_free_coherent(mvi->dev, MVS_RX_FISL_SZ,
+                                 mvi->rx_fis, mvi->rx_fis_dma);
+       if (mvi->rx)
+               dma_free_coherent(mvi->dev,
+                                 sizeof(*mvi->rx) * (MVS_RX_RING_SZ + 1),
+                                 mvi->rx, mvi->rx_dma);
+       if (mvi->slot)
+               dma_free_coherent(mvi->dev,
+                                 sizeof(*mvi->slot) * slot_nr,
+                                 mvi->slot, mvi->slot_dma);
+
+       if (mvi->bulk_buffer)
+               dma_free_coherent(mvi->dev, TRASH_BUCKET_SIZE,
+                                 mvi->bulk_buffer, mvi->bulk_buffer_dma);
+       if (mvi->bulk_buffer1)
+               dma_free_coherent(mvi->dev, TRASH_BUCKET_SIZE,
+                                 mvi->bulk_buffer1, mvi->bulk_buffer_dma1);
+
+       MVS_CHIP_DISP->chip_iounmap(mvi);
+       if (mvi->shost)
+               scsi_host_put(mvi->shost);
+       list_for_each_entry(mwq, &mvi->wq_list, entry)
+               cancel_delayed_work(&mwq->work_q);
+       kfree(mvi->tags);
+       kfree(mvi);
+}
+
+#ifdef CONFIG_SCSI_MVSAS_TASKLET
+static void mvs_tasklet(unsigned long opaque)
+{
+       u32 stat;
+       u16 core_nr, i = 0;
+
+       struct mvs_info *mvi;
+       struct sas_ha_struct *sha = (struct sas_ha_struct *)opaque;
+
+       core_nr = ((struct mvs_prv_info *)sha->lldd_ha)->n_host;
+       mvi = ((struct mvs_prv_info *)sha->lldd_ha)->mvi[0];
+
+       if (unlikely(!mvi))
+               BUG_ON(1);
+
+       stat = MVS_CHIP_DISP->isr_status(mvi, mvi->pdev->irq);
+       if (!stat)
+               goto out;
+
+       for (i = 0; i < core_nr; i++) {
+               mvi = ((struct mvs_prv_info *)sha->lldd_ha)->mvi[i];
+               MVS_CHIP_DISP->isr(mvi, mvi->pdev->irq, stat);
+       }
+out:
+       MVS_CHIP_DISP->interrupt_enable(mvi);
+
+}
+#endif
+
+static irqreturn_t mvs_interrupt(int irq, void *opaque)
+{
+       u32 core_nr;
+       u32 stat;
+       struct mvs_info *mvi;
+       struct sas_ha_struct *sha = opaque;
+#ifndef CONFIG_SCSI_MVSAS_TASKLET
+       u32 i;
+#endif
+
+       core_nr = ((struct mvs_prv_info *)sha->lldd_ha)->n_host;
+       mvi = ((struct mvs_prv_info *)sha->lldd_ha)->mvi[0];
+
+       if (unlikely(!mvi))
+               return IRQ_NONE;
+#ifdef CONFIG_SCSI_MVSAS_TASKLET
+       MVS_CHIP_DISP->interrupt_disable(mvi);
+#endif
+
+       stat = MVS_CHIP_DISP->isr_status(mvi, irq);
+       if (!stat) {
+       #ifdef CONFIG_SCSI_MVSAS_TASKLET
+               MVS_CHIP_DISP->interrupt_enable(mvi);
+       #endif
+               return IRQ_NONE;
+       }
+
+#ifdef CONFIG_SCSI_MVSAS_TASKLET
+       tasklet_schedule(&((struct mvs_prv_info *)sha->lldd_ha)->mv_tasklet);
+#else
+       for (i = 0; i < core_nr; i++) {
+               mvi = ((struct mvs_prv_info *)sha->lldd_ha)->mvi[i];
+               MVS_CHIP_DISP->isr(mvi, irq, stat);
+       }
+#endif
+       return IRQ_HANDLED;
+}
+
+static int mvs_alloc(struct mvs_info *mvi, struct Scsi_Host *shost)
+{
+       int i = 0, slot_nr;
+       char pool_name[32];
+
+       if (mvi->flags & MVF_FLAG_SOC)
+               slot_nr = MVS_SOC_SLOTS;
+       else
+               slot_nr = MVS_CHIP_SLOT_SZ;
+
+       spin_lock_init(&mvi->lock);
+       for (i = 0; i < mvi->chip->n_phy; i++) {
+               mvs_phy_init(mvi, i);
+               mvi->port[i].wide_port_phymap = 0;
+               mvi->port[i].port_attached = 0;
+               INIT_LIST_HEAD(&mvi->port[i].list);
+       }
+       for (i = 0; i < MVS_MAX_DEVICES; i++) {
+               mvi->devices[i].taskfileset = MVS_ID_NOT_MAPPED;
+               mvi->devices[i].dev_type = SAS_PHY_UNUSED;
+               mvi->devices[i].device_id = i;
+               mvi->devices[i].dev_status = MVS_DEV_NORMAL;
+               init_timer(&mvi->devices[i].timer);
+       }
+
+       /*
+        * alloc and init our DMA areas
+        */
+       mvi->tx = dma_alloc_coherent(mvi->dev,
+                                    sizeof(*mvi->tx) * MVS_CHIP_SLOT_SZ,
+                                    &mvi->tx_dma, GFP_KERNEL);
+       if (!mvi->tx)
+               goto err_out;
+       memset(mvi->tx, 0, sizeof(*mvi->tx) * MVS_CHIP_SLOT_SZ);
+       mvi->rx_fis = dma_alloc_coherent(mvi->dev, MVS_RX_FISL_SZ,
+                                        &mvi->rx_fis_dma, GFP_KERNEL);
+       if (!mvi->rx_fis)
+               goto err_out;
+       memset(mvi->rx_fis, 0, MVS_RX_FISL_SZ);
+
+       mvi->rx = dma_alloc_coherent(mvi->dev,
+                                    sizeof(*mvi->rx) * (MVS_RX_RING_SZ + 1),
+                                    &mvi->rx_dma, GFP_KERNEL);
+       if (!mvi->rx)
+               goto err_out;
+       memset(mvi->rx, 0, sizeof(*mvi->rx) * (MVS_RX_RING_SZ + 1));
+       mvi->rx[0] = cpu_to_le32(0xfff);
+       mvi->rx_cons = 0xfff;
+
+       mvi->slot = dma_alloc_coherent(mvi->dev,
+                                      sizeof(*mvi->slot) * slot_nr,
+                                      &mvi->slot_dma, GFP_KERNEL);
+       if (!mvi->slot)
+               goto err_out;
+       memset(mvi->slot, 0, sizeof(*mvi->slot) * slot_nr);
+
+       mvi->bulk_buffer = dma_alloc_coherent(mvi->dev,
+                                      TRASH_BUCKET_SIZE,
+                                      &mvi->bulk_buffer_dma, GFP_KERNEL);
+       if (!mvi->bulk_buffer)
+               goto err_out;
+
+       mvi->bulk_buffer1 = dma_alloc_coherent(mvi->dev,
+                                      TRASH_BUCKET_SIZE,
+                                      &mvi->bulk_buffer_dma1, GFP_KERNEL);
+       if (!mvi->bulk_buffer1)
+               goto err_out;
+
+       sprintf(pool_name, "%s%d", "mvs_dma_pool", mvi->id);
+       mvi->dma_pool = pci_pool_create(pool_name, mvi->pdev, MVS_SLOT_BUF_SZ, 16, 0);
+       if (!mvi->dma_pool) {
+                       printk(KERN_DEBUG "failed to create dma pool %s.\n", pool_name);
+                       goto err_out;
+       }
+       mvi->tags_num = slot_nr;
+
+       /* Initialize tags */
+       mvs_tag_init(mvi);
+       return 0;
+err_out:
+       return 1;
+}
+
+
+int mvs_ioremap(struct mvs_info *mvi, int bar, int bar_ex)
+{
+       unsigned long res_start, res_len, res_flag, res_flag_ex = 0;
+       struct pci_dev *pdev = mvi->pdev;
+       if (bar_ex != -1) {
+               /*
+                * ioremap main and peripheral registers
+                */
+               res_start = pci_resource_start(pdev, bar_ex);
+               res_len = pci_resource_len(pdev, bar_ex);
+               if (!res_start || !res_len)
+                       goto err_out;
+
+               res_flag_ex = pci_resource_flags(pdev, bar_ex);
+               if (res_flag_ex & IORESOURCE_MEM) {
+                       if (res_flag_ex & IORESOURCE_CACHEABLE)
+                               mvi->regs_ex = ioremap(res_start, res_len);
+                       else
+                               mvi->regs_ex = ioremap_nocache(res_start,
+                                               res_len);
+               } else
+                       mvi->regs_ex = (void *)res_start;
+               if (!mvi->regs_ex)
+                       goto err_out;
+       }
+
+       res_start = pci_resource_start(pdev, bar);
+       res_len = pci_resource_len(pdev, bar);
+       if (!res_start || !res_len)
+               goto err_out;
+
+       res_flag = pci_resource_flags(pdev, bar);
+       if (res_flag & IORESOURCE_CACHEABLE)
+               mvi->regs = ioremap(res_start, res_len);
+       else
+               mvi->regs = ioremap_nocache(res_start, res_len);
+
+       if (!mvi->regs) {
+               if (mvi->regs_ex && (res_flag_ex & IORESOURCE_MEM))
+                       iounmap(mvi->regs_ex);
+               mvi->regs_ex = NULL;
+               goto err_out;
+       }
+
+       return 0;
+err_out:
+       return -1;
+}
+
+void mvs_iounmap(void __iomem *regs)
+{
+       iounmap(regs);
+}
+
+static struct mvs_info *mvs_pci_alloc(struct pci_dev *pdev,
+                               const struct pci_device_id *ent,
+                               struct Scsi_Host *shost, unsigned int id)
+{
+       struct mvs_info *mvi = NULL;
+       struct sas_ha_struct *sha = SHOST_TO_SAS_HA(shost);
+
+       mvi = kzalloc(sizeof(*mvi) +
+               (1L << mvs_chips[ent->driver_data].slot_width) *
+               sizeof(struct mvs_slot_info), GFP_KERNEL);
+       if (!mvi)
+               return NULL;
+
+       mvi->pdev = pdev;
+       mvi->dev = &pdev->dev;
+       mvi->chip_id = ent->driver_data;
+       mvi->chip = &mvs_chips[mvi->chip_id];
+       INIT_LIST_HEAD(&mvi->wq_list);
+
+       ((struct mvs_prv_info *)sha->lldd_ha)->mvi[id] = mvi;
+       ((struct mvs_prv_info *)sha->lldd_ha)->n_phy = mvi->chip->n_phy;
+
+       mvi->id = id;
+       mvi->sas = sha;
+       mvi->shost = shost;
+
+       mvi->tags = kzalloc(MVS_CHIP_SLOT_SZ>>3, GFP_KERNEL);
+       if (!mvi->tags)
+               goto err_out;
+
+       if (MVS_CHIP_DISP->chip_ioremap(mvi))
+               goto err_out;
+       if (!mvs_alloc(mvi, shost))
+               return mvi;
+err_out:
+       mvs_free(mvi);
+       return NULL;
+}
+
+static int pci_go_64(struct pci_dev *pdev)
+{
+       int rc;
+
+       if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) {
+               rc = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
+               if (rc) {
+                       rc = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
+                       if (rc) {
+                               dev_printk(KERN_ERR, &pdev->dev,
+                                          "64-bit DMA enable failed\n");
+                               return rc;
+                       }
+               }
+       } else {
+               rc = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
+               if (rc) {
+                       dev_printk(KERN_ERR, &pdev->dev,
+                                  "32-bit DMA enable failed\n");
+                       return rc;
+               }
+               rc = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
+               if (rc) {
+                       dev_printk(KERN_ERR, &pdev->dev,
+                                  "32-bit consistent DMA enable failed\n");
+                       return rc;
+               }
+       }
+
+       return rc;
+}
+
+static int mvs_prep_sas_ha_init(struct Scsi_Host *shost,
+                               const struct mvs_chip_info *chip_info)
+{
+       int phy_nr, port_nr; unsigned short core_nr;
+       struct asd_sas_phy **arr_phy;
+       struct asd_sas_port **arr_port;
+       struct sas_ha_struct *sha = SHOST_TO_SAS_HA(shost);
+
+       core_nr = chip_info->n_host;
+       phy_nr  = core_nr * chip_info->n_phy;
+       port_nr = phy_nr;
+
+       memset(sha, 0x00, sizeof(struct sas_ha_struct));
+       arr_phy  = kcalloc(phy_nr, sizeof(void *), GFP_KERNEL);
+       arr_port = kcalloc(port_nr, sizeof(void *), GFP_KERNEL);
+       if (!arr_phy || !arr_port)
+               goto exit_free;
+
+       sha->sas_phy = arr_phy;
+       sha->sas_port = arr_port;
+       sha->core.shost = shost;
+
+       sha->lldd_ha = kzalloc(sizeof(struct mvs_prv_info), GFP_KERNEL);
+       if (!sha->lldd_ha)
+               goto exit_free;
+
+       ((struct mvs_prv_info *)sha->lldd_ha)->n_host = core_nr;
+
+       shost->transportt = mvs_stt;
+       shost->max_id = MVS_MAX_DEVICES;
+       shost->max_lun = ~0;
+       shost->max_channel = 1;
+       shost->max_cmd_len = 16;
+
+       return 0;
+exit_free:
+       kfree(arr_phy);
+       kfree(arr_port);
+       return -1;
+
+}
+
+static void  mvs_post_sas_ha_init(struct Scsi_Host *shost,
+                       const struct mvs_chip_info *chip_info)
+{
+       int can_queue, i = 0, j = 0;
+       struct mvs_info *mvi = NULL;
+       struct sas_ha_struct *sha = SHOST_TO_SAS_HA(shost);
+       unsigned short nr_core = ((struct mvs_prv_info *)sha->lldd_ha)->n_host;
+
+       for (j = 0; j < nr_core; j++) {
+               mvi = ((struct mvs_prv_info *)sha->lldd_ha)->mvi[j];
+               for (i = 0; i < chip_info->n_phy; i++) {
+                       sha->sas_phy[j * chip_info->n_phy  + i] =
+                               &mvi->phy[i].sas_phy;
+                       sha->sas_port[j * chip_info->n_phy + i] =
+                               &mvi->port[i].sas_port;
+               }
+       }
+
+       sha->sas_ha_name = DRV_NAME;
+       sha->dev = mvi->dev;
+       sha->lldd_module = THIS_MODULE;
+       sha->sas_addr = &mvi->sas_addr[0];
+
+       sha->num_phys = nr_core * chip_info->n_phy;
+
+       if (mvi->flags & MVF_FLAG_SOC)
+               can_queue = MVS_SOC_CAN_QUEUE;
+       else
+               can_queue = MVS_CHIP_SLOT_SZ;
+
+       shost->sg_tablesize = min_t(u16, SG_ALL, MVS_MAX_SG);
+       shost->can_queue = can_queue;
+       mvi->shost->cmd_per_lun = MVS_QUEUE_SIZE;
+       sha->core.shost = mvi->shost;
+}
+
+static void mvs_init_sas_add(struct mvs_info *mvi)
+{
+       u8 i;
+       for (i = 0; i < mvi->chip->n_phy; i++) {
+               mvi->phy[i].dev_sas_addr = 0x5005043011ab0000ULL;
+               mvi->phy[i].dev_sas_addr =
+                       cpu_to_be64((u64)(*(u64 *)&mvi->phy[i].dev_sas_addr));
+       }
+
+       memcpy(mvi->sas_addr, &mvi->phy[0].dev_sas_addr, SAS_ADDR_SIZE);
+}
+
+static int mvs_pci_init(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+       unsigned int rc, nhost = 0;
+       struct mvs_info *mvi;
+       struct mvs_prv_info *mpi;
+       irq_handler_t irq_handler = mvs_interrupt;
+       struct Scsi_Host *shost = NULL;
+       const struct mvs_chip_info *chip;
+
+       dev_printk(KERN_INFO, &pdev->dev,
+               "mvsas: driver version %s\n", DRV_VERSION);
+       rc = pci_enable_device(pdev);
+       if (rc)
+               goto err_out_enable;
+
+       pci_set_master(pdev);
+
+       rc = pci_request_regions(pdev, DRV_NAME);
+       if (rc)
+               goto err_out_disable;
+
+       rc = pci_go_64(pdev);
+       if (rc)
+               goto err_out_regions;
+
+       shost = scsi_host_alloc(&mvs_sht, sizeof(void *));
+       if (!shost) {
+               rc = -ENOMEM;
+               goto err_out_regions;
+       }
+
+       chip = &mvs_chips[ent->driver_data];
+       SHOST_TO_SAS_HA(shost) =
+               kcalloc(1, sizeof(struct sas_ha_struct), GFP_KERNEL);
+       if (!SHOST_TO_SAS_HA(shost)) {
+               kfree(shost);
+               rc = -ENOMEM;
+               goto err_out_regions;
+       }
+
+       rc = mvs_prep_sas_ha_init(shost, chip);
+       if (rc) {
+               kfree(shost);
+               rc = -ENOMEM;
+               goto err_out_regions;
+       }
+
+       pci_set_drvdata(pdev, SHOST_TO_SAS_HA(shost));
+
+       do {
+               mvi = mvs_pci_alloc(pdev, ent, shost, nhost);
+               if (!mvi) {
+                       rc = -ENOMEM;
+                       goto err_out_regions;
+               }
+
+               memset(&mvi->hba_info_param, 0xFF,
+                       sizeof(struct hba_info_page));
+
+               mvs_init_sas_add(mvi);
+
+               mvi->instance = nhost;
+               rc = MVS_CHIP_DISP->chip_init(mvi);
+               if (rc) {
+                       mvs_free(mvi);
+                       goto err_out_regions;
+               }
+               nhost++;
+       } while (nhost < chip->n_host);
+       mpi = (struct mvs_prv_info *)(SHOST_TO_SAS_HA(shost)->lldd_ha);
+#ifdef CONFIG_SCSI_MVSAS_TASKLET
+       tasklet_init(&(mpi->mv_tasklet), mvs_tasklet,
+                    (unsigned long)SHOST_TO_SAS_HA(shost));
+#endif
+
+       mvs_post_sas_ha_init(shost, chip);
+
+       rc = scsi_add_host(shost, &pdev->dev);
+       if (rc)
+               goto err_out_shost;
+
+       rc = sas_register_ha(SHOST_TO_SAS_HA(shost));
+       if (rc)
+               goto err_out_shost;
+       rc = request_irq(pdev->irq, irq_handler, IRQF_SHARED,
+               DRV_NAME, SHOST_TO_SAS_HA(shost));
+       if (rc)
+               goto err_not_sas;
+
+       MVS_CHIP_DISP->interrupt_enable(mvi);
+
+       scsi_scan_host(mvi->shost);
+
+       return 0;
+
+err_not_sas:
+       sas_unregister_ha(SHOST_TO_SAS_HA(shost));
+err_out_shost:
+       scsi_remove_host(mvi->shost);
+err_out_regions:
+       pci_release_regions(pdev);
+err_out_disable:
+       pci_disable_device(pdev);
+err_out_enable:
+       return rc;
+}
+
+static void mvs_pci_remove(struct pci_dev *pdev)
+{
+       unsigned short core_nr, i = 0;
+       struct sas_ha_struct *sha = pci_get_drvdata(pdev);
+       struct mvs_info *mvi = NULL;
+
+       core_nr = ((struct mvs_prv_info *)sha->lldd_ha)->n_host;
+       mvi = ((struct mvs_prv_info *)sha->lldd_ha)->mvi[0];
+
+#ifdef CONFIG_SCSI_MVSAS_TASKLET
+       tasklet_kill(&((struct mvs_prv_info *)sha->lldd_ha)->mv_tasklet);
+#endif
+
+       sas_unregister_ha(sha);
+       sas_remove_host(mvi->shost);
+       scsi_remove_host(mvi->shost);
+
+       MVS_CHIP_DISP->interrupt_disable(mvi);
+       free_irq(mvi->pdev->irq, sha);
+       for (i = 0; i < core_nr; i++) {
+               mvi = ((struct mvs_prv_info *)sha->lldd_ha)->mvi[i];
+               mvs_free(mvi);
+       }
+       kfree(sha->sas_phy);
+       kfree(sha->sas_port);
+       kfree(sha);
+       pci_release_regions(pdev);
+       pci_disable_device(pdev);
+       return;
+}
+
+static struct pci_device_id mvs_pci_table[] = {
+       { PCI_VDEVICE(MARVELL, 0x6320), chip_6320 },
+       { PCI_VDEVICE(MARVELL, 0x6340), chip_6440 },
+       {
+               .vendor         = PCI_VENDOR_ID_MARVELL,
+               .device         = 0x6440,
+               .subvendor      = PCI_ANY_ID,
+               .subdevice      = 0x6480,
+               .class          = 0,
+               .class_mask     = 0,
+               .driver_data    = chip_6485,
+       },
+       { PCI_VDEVICE(MARVELL, 0x6440), chip_6440 },
+       { PCI_VDEVICE(MARVELL, 0x6485), chip_6485 },
+       { PCI_VDEVICE(MARVELL, 0x9480), chip_9480 },
+       { PCI_VDEVICE(MARVELL, 0x9180), chip_9180 },
+       { PCI_VDEVICE(ARECA, PCI_DEVICE_ID_ARECA_1300), chip_1300 },
+       { PCI_VDEVICE(ARECA, PCI_DEVICE_ID_ARECA_1320), chip_1320 },
+       { PCI_VDEVICE(ADAPTEC2, 0x0450), chip_6440 },
+       { PCI_VDEVICE(TTI, 0x2710), chip_9480 },
+       { PCI_VDEVICE(TTI, 0x2720), chip_9480 },
+       { PCI_VDEVICE(TTI, 0x2721), chip_9480 },
+       { PCI_VDEVICE(TTI, 0x2722), chip_9480 },
+       { PCI_VDEVICE(TTI, 0x2740), chip_9480 },
+       { PCI_VDEVICE(TTI, 0x2744), chip_9480 },
+       { PCI_VDEVICE(TTI, 0x2760), chip_9480 },
+       {
+               .vendor         = PCI_VENDOR_ID_MARVELL_EXT,
+               .device         = 0x9480,
+               .subvendor      = PCI_ANY_ID,
+               .subdevice      = 0x9480,
+               .class          = 0,
+               .class_mask     = 0,
+               .driver_data    = chip_9480,
+       },
+       {
+               .vendor         = PCI_VENDOR_ID_MARVELL_EXT,
+               .device         = 0x9445,
+               .subvendor      = PCI_ANY_ID,
+               .subdevice      = 0x9480,
+               .class          = 0,
+               .class_mask     = 0,
+               .driver_data    = chip_9445,
+       },
+       {
+               .vendor         = PCI_VENDOR_ID_MARVELL_EXT,
+               .device         = 0x9485,
+               .subvendor      = PCI_ANY_ID,
+               .subdevice      = 0x9480,
+               .class          = 0,
+               .class_mask     = 0,
+               .driver_data    = chip_9485,
+       },
+       {
+               .vendor         = PCI_VENDOR_ID_MARVELL_EXT,
+               .device         = 0x9485,
+               .subvendor      = PCI_ANY_ID,
+               .subdevice      = 0x9485,
+               .class          = 0,
+               .class_mask     = 0,
+               .driver_data    = chip_9485,
+       },
+       { PCI_VDEVICE(OCZ, 0x1021), chip_9485}, /* OCZ RevoDrive3 */
+       { PCI_VDEVICE(OCZ, 0x1022), chip_9485}, /* OCZ RevoDrive3/zDriveR4 (exact model unknown) */
+       { PCI_VDEVICE(OCZ, 0x1040), chip_9485}, /* OCZ RevoDrive3/zDriveR4 (exact model unknown) */
+       { PCI_VDEVICE(OCZ, 0x1041), chip_9485}, /* OCZ RevoDrive3/zDriveR4 (exact model unknown) */
+       { PCI_VDEVICE(OCZ, 0x1042), chip_9485}, /* OCZ RevoDrive3/zDriveR4 (exact model unknown) */
+       { PCI_VDEVICE(OCZ, 0x1043), chip_9485}, /* OCZ RevoDrive3/zDriveR4 (exact model unknown) */
+       { PCI_VDEVICE(OCZ, 0x1044), chip_9485}, /* OCZ RevoDrive3/zDriveR4 (exact model unknown) */
+       { PCI_VDEVICE(OCZ, 0x1080), chip_9485}, /* OCZ RevoDrive3/zDriveR4 (exact model unknown) */
+       { PCI_VDEVICE(OCZ, 0x1083), chip_9485}, /* OCZ RevoDrive3/zDriveR4 (exact model unknown) */
+       { PCI_VDEVICE(OCZ, 0x1084), chip_9485}, /* OCZ RevoDrive3/zDriveR4 (exact model unknown) */
+
+       { }     /* terminate list */
+};
+
+static struct pci_driver mvs_pci_driver = {
+       .name           = DRV_NAME,
+       .id_table       = mvs_pci_table,
+       .probe          = mvs_pci_init,
+       .remove         = mvs_pci_remove,
+};
+
+static ssize_t
+mvs_show_driver_version(struct device *cdev,
+               struct device_attribute *attr,  char *buffer)
+{
+       return snprintf(buffer, PAGE_SIZE, "%s\n", DRV_VERSION);
+}
+
+static DEVICE_ATTR(driver_version,
+                        S_IRUGO,
+                        mvs_show_driver_version,
+                        NULL);
+
+static ssize_t
+mvs_store_interrupt_coalescing(struct device *cdev,
+                       struct device_attribute *attr,
+                       const char *buffer, size_t size)
+{
+       int val = 0;
+       struct mvs_info *mvi = NULL;
+       struct Scsi_Host *shost = class_to_shost(cdev);
+       struct sas_ha_struct *sha = SHOST_TO_SAS_HA(shost);
+       u8 i, core_nr;
+       if (buffer == NULL)
+               return size;
+
+       if (sscanf(buffer, "%d", &val) != 1)
+               return -EINVAL;
+
+       if (val >= 0x10000) {
+               mv_dprintk("interrupt coalescing timer %d us is"
+                       "too long\n", val);
+               return strlen(buffer);
+       }
+
+       interrupt_coalescing = val;
+
+       core_nr = ((struct mvs_prv_info *)sha->lldd_ha)->n_host;
+       mvi = ((struct mvs_prv_info *)sha->lldd_ha)->mvi[0];
+
+       if (unlikely(!mvi))
+               return -EINVAL;
+
+       for (i = 0; i < core_nr; i++) {
+               mvi = ((struct mvs_prv_info *)sha->lldd_ha)->mvi[i];
+               if (MVS_CHIP_DISP->tune_interrupt)
+                       MVS_CHIP_DISP->tune_interrupt(mvi,
+                               interrupt_coalescing);
+       }
+       mv_dprintk("set interrupt coalescing time to %d us\n",
+               interrupt_coalescing);
+       return strlen(buffer);
+}
+
+static ssize_t mvs_show_interrupt_coalescing(struct device *cdev,
+                       struct device_attribute *attr, char *buffer)
+{
+       return snprintf(buffer, PAGE_SIZE, "%d\n", interrupt_coalescing);
+}
+
+static DEVICE_ATTR(interrupt_coalescing,
+                        S_IRUGO|S_IWUSR,
+                        mvs_show_interrupt_coalescing,
+                        mvs_store_interrupt_coalescing);
+
+/* task handler */
+struct task_struct *mvs_th;
+static int __init mvs_init(void)
+{
+       int rc;
+       mvs_stt = sas_domain_attach_transport(&mvs_transport_ops);
+       if (!mvs_stt)
+               return -ENOMEM;
+
+       rc = pci_register_driver(&mvs_pci_driver);
+       if (rc)
+               goto err_out;
+
+       return 0;
+
+err_out:
+       sas_release_transport(mvs_stt);
+       return rc;
+}
+
+static void __exit mvs_exit(void)
+{
+       pci_unregister_driver(&mvs_pci_driver);
+       sas_release_transport(mvs_stt);
+}
+
+struct device_attribute *mvst_host_attrs[] = {
+       &dev_attr_driver_version,
+       &dev_attr_interrupt_coalescing,
+       NULL,
+};
+
+module_init(mvs_init);
+module_exit(mvs_exit);
+
+MODULE_AUTHOR("Jeff Garzik <jgarzik@pobox.com>");
+MODULE_DESCRIPTION("Marvell 88SE6440 SAS/SATA controller driver");
+MODULE_VERSION(DRV_VERSION);
+MODULE_LICENSE("GPL");
+#ifdef CONFIG_PCI
+MODULE_DEVICE_TABLE(pci, mvs_pci_table);
+#endif