isci: move core/controller to host
Dan Williams [Mon, 9 May 2011 00:34:44 +0000 (17:34 -0700)]
Now that the data structures are unified unify the implementation in
host.[ch] and cleanup namespace pollution.

Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>

28 files changed:
drivers/scsi/isci/Makefile
drivers/scsi/isci/core/sci_util.c
drivers/scsi/isci/core/scic_config_parameters.h
drivers/scsi/isci/core/scic_controller.h [deleted file]
drivers/scsi/isci/core/scic_sds_controller.c [deleted file]
drivers/scsi/isci/core/scic_sds_controller.h [deleted file]
drivers/scsi/isci/core/scic_sds_phy.c
drivers/scsi/isci/core/scic_sds_port.c
drivers/scsi/isci/core/scic_sds_port.h
drivers/scsi/isci/core/scic_sds_port_configuration_agent.c
drivers/scsi/isci/core/scic_sds_request.c
drivers/scsi/isci/core/scic_sds_smp_request.c
drivers/scsi/isci/core/scic_sds_ssp_request.c
drivers/scsi/isci/core/scic_sds_stp_request.c
drivers/scsi/isci/core/scic_sds_unsolicited_frame_control.c
drivers/scsi/isci/host.c
drivers/scsi/isci/host.h
drivers/scsi/isci/init.c
drivers/scsi/isci/isci.h
drivers/scsi/isci/pool.h [moved from drivers/scsi/isci/core/sci_pool.h with 100% similarity]
drivers/scsi/isci/port.c
drivers/scsi/isci/probe_roms.c
drivers/scsi/isci/remote_device.c
drivers/scsi/isci/remote_node_context.c
drivers/scsi/isci/remote_node_table.c
drivers/scsi/isci/request.c
drivers/scsi/isci/sci_environment.h [deleted file]
drivers/scsi/isci/task.c

index 78ba0fc..45f8f68 100644 (file)
@@ -5,7 +5,6 @@ isci-objs := init.o phy.o request.o sata.o \
             host.o task.o probe_roms.o \
             remote_node_context.o \
             remote_node_table.o \
-            core/scic_sds_controller.o  \
             core/scic_sds_request.o \
             core/scic_sds_stp_request.o \
             core/scic_sds_port.o \
index 0101fec..595d8da 100644 (file)
@@ -55,7 +55,6 @@
 
 #include <linux/kernel.h>
 #include "sci_util.h"
-#include "sci_environment.h"
 #include "request.h"
 
 void *scic_request_get_virt_addr(struct scic_sds_request *sci_req, dma_addr_t phys_addr)
index 8b8c925..15e7744 100644 (file)
@@ -229,44 +229,6 @@ union scic_oem_parameters {
        struct scic_sds_oem_params sds1;
 };
 
-/**
- * scic_user_parameters_set() - This method allows the user to attempt to
- *    change the user parameters utilized by the controller.
- * @controller: This parameter specifies the controller on which to set the
- *    user parameters.
- * @user_parameters: This parameter specifies the USER_PARAMETERS object
- *    containing the potential new values.
- *
- * Indicate if the update of the user parameters was successful. SCI_SUCCESS
- * This value is returned if the operation succeeded. SCI_FAILURE_INVALID_STATE
- * This value is returned if the attempt to change the user parameter failed,
- * because changing one of the parameters is not currently allowed.
- * SCI_FAILURE_INVALID_PARAMETER_VALUE This value is returned if the user
- * supplied an invalid interrupt coalescence time, spin up delay interval, etc.
- */
-enum sci_status scic_user_parameters_set(
-       struct scic_sds_controller *controller,
-       union scic_user_parameters *user_parameters);
-
-/**
- * scic_oem_parameters_set() - This method allows the user to attempt to change
- *    the OEM parameters utilized by the controller.
- * @controller: This parameter specifies the controller on which to set the
- *    user parameters.
- * @oem_parameters: This parameter specifies the OEM parameters object
- *    containing the potential new values.
- *
- * Indicate if the update of the user parameters was successful. SCI_SUCCESS
- * This value is returned if the operation succeeded. SCI_FAILURE_INVALID_STATE
- * This value is returned if the attempt to change the user parameter failed,
- * because changing one of the parameters is not currently allowed.
- * SCI_FAILURE_INVALID_PARAMETER_VALUE This value is returned if the user
- * supplied an unsupported value for one of the OEM parameters.
- */
-enum sci_status scic_oem_parameters_set(
-       struct scic_sds_controller *controller,
-       union scic_oem_parameters *oem_parameters);
-
 int scic_oem_parameters_validate(struct scic_sds_oem_params *oem);
 
 /**
diff --git a/drivers/scsi/isci/core/scic_controller.h b/drivers/scsi/isci/core/scic_controller.h
deleted file mode 100644 (file)
index bd08f30..0000000
+++ /dev/null
@@ -1,130 +0,0 @@
-/*
- * This file is provided under a dual BSD/GPLv2 license.  When using or
- * redistributing this file, you may do so under either license.
- *
- * GPL LICENSE SUMMARY
- *
- * Copyright(c) 2008 - 2011 Intel Corporation. All rights reserved.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
- * The full GNU General Public License is included in this distribution
- * in the file called LICENSE.GPL.
- *
- * BSD LICENSE
- *
- * Copyright(c) 2008 - 2011 Intel Corporation. All rights reserved.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- *   * Redistributions of source code must retain the above copyright
- *     notice, this list of conditions and the following disclaimer.
- *   * Redistributions in binary form must reproduce the above copyright
- *     notice, this list of conditions and the following disclaimer in
- *     the documentation and/or other materials provided with the
- *     distribution.
- *   * Neither the name of Intel Corporation nor the names of its
- *     contributors may be used to endorse or promote products derived
- *     from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-#ifndef _SCIC_CONTROLLER_H_
-#define _SCIC_CONTROLLER_H_
-
-#include "scic_config_parameters.h"
-
-struct scic_sds_request;
-struct scic_sds_phy;
-struct scic_sds_port;
-struct scic_sds_remote_device;
-
-enum sci_status scic_controller_construct(struct scic_sds_controller *c,
-                                         void __iomem *scu_base,
-                                         void __iomem *smu_base);
-
-void scic_controller_enable_interrupts(
-       struct scic_sds_controller *controller);
-
-void scic_controller_disable_interrupts(
-       struct scic_sds_controller *controller);
-
-enum sci_status scic_controller_initialize(
-       struct scic_sds_controller *controller);
-
-u32 scic_controller_get_suggested_start_timeout(
-       struct scic_sds_controller *controller);
-
-enum sci_status scic_controller_start(
-       struct scic_sds_controller *controller,
-       u32 timeout);
-
-enum sci_status scic_controller_stop(
-       struct scic_sds_controller *controller,
-       u32 timeout);
-
-enum sci_status scic_controller_reset(
-       struct scic_sds_controller *controller);
-
-enum sci_status scic_controller_start_io(
-       struct scic_sds_controller *controller,
-       struct scic_sds_remote_device *remote_device,
-       struct scic_sds_request *io_request,
-       u16 io_tag);
-
-enum sci_task_status scic_controller_start_task(
-       struct scic_sds_controller *controller,
-       struct scic_sds_remote_device *remote_device,
-       struct scic_sds_request *task_request,
-       u16 io_tag);
-
-enum sci_status scic_controller_terminate_request(
-       struct scic_sds_controller *controller,
-       struct scic_sds_remote_device *remote_device,
-       struct scic_sds_request *request);
-
-enum sci_status scic_controller_complete_io(
-       struct scic_sds_controller *controller,
-       struct scic_sds_remote_device *remote_device,
-       struct scic_sds_request *io_request);
-
-enum sci_status scic_controller_get_phy_handle(
-       struct scic_sds_controller *controller,
-       u8 phy_index,
-       struct scic_sds_phy **phy_handle);
-
-u16 scic_controller_allocate_io_tag(
-       struct scic_sds_controller *controller);
-
-enum sci_status scic_controller_free_io_tag(
-       struct scic_sds_controller *controller,
-       u16 io_tag);
-
-struct device;
-struct scic_sds_controller *scic_controller_alloc(struct device *dev);
-int scic_controller_mem_init(struct scic_sds_controller *scic);
-#endif  /* _SCIC_CONTROLLER_H_ */
diff --git a/drivers/scsi/isci/core/scic_sds_controller.c b/drivers/scsi/isci/core/scic_sds_controller.c
deleted file mode 100644 (file)
index e77265b..0000000
+++ /dev/null
@@ -1,2973 +0,0 @@
-/*
- * This file is provided under a dual BSD/GPLv2 license.  When using or
- * redistributing this file, you may do so under either license.
- *
- * GPL LICENSE SUMMARY
- *
- * Copyright(c) 2008 - 2011 Intel Corporation. All rights reserved.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
- * The full GNU General Public License is included in this distribution
- * in the file called LICENSE.GPL.
- *
- * BSD LICENSE
- *
- * Copyright(c) 2008 - 2011 Intel Corporation. All rights reserved.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- *   * Redistributions of source code must retain the above copyright
- *     notice, this list of conditions and the following disclaimer.
- *   * Redistributions in binary form must reproduce the above copyright
- *     notice, this list of conditions and the following disclaimer in
- *     the documentation and/or other materials provided with the
- *     distribution.
- *   * Neither the name of Intel Corporation nor the names of its
- *     contributors may be used to endorse or promote products derived
- *     from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-#include <linux/device.h>
-#include <scsi/sas.h>
-#include "scic_controller.h"
-#include "scic_phy.h"
-#include "scic_port.h"
-#include "scic_sds_controller.h"
-#include "scu_registers.h"
-#include "scic_sds_phy.h"
-#include "scic_sds_port_configuration_agent.h"
-#include "scic_sds_port.h"
-#include "remote_device.h"
-#include "scic_sds_request.h"
-#include "sci_environment.h"
-#include "sci_util.h"
-#include "scu_completion_codes.h"
-#include "scu_event_codes.h"
-#include "scu_remote_node_context.h"
-#include "scu_task_context.h"
-#include "scu_unsolicited_frame.h"
-#include "timers.h"
-
-#define SCU_CONTEXT_RAM_INIT_STALL_TIME      200
-
-/**
- * smu_dcc_get_max_ports() -
- *
- * This macro returns the maximum number of logical ports supported by the
- * hardware. The caller passes in the value read from the device context
- * capacity register and this macro will mash and shift the value appropriately.
- */
-#define smu_dcc_get_max_ports(dcc_value) \
-       (\
-               (((dcc_value) & SMU_DEVICE_CONTEXT_CAPACITY_MAX_LP_MASK) \
-                >> SMU_DEVICE_CONTEXT_CAPACITY_MAX_LP_SHIFT) + 1 \
-       )
-
-/**
- * smu_dcc_get_max_task_context() -
- *
- * This macro returns the maximum number of task contexts supported by the
- * hardware. The caller passes in the value read from the device context
- * capacity register and this macro will mash and shift the value appropriately.
- */
-#define smu_dcc_get_max_task_context(dcc_value)        \
-       (\
-               (((dcc_value) & SMU_DEVICE_CONTEXT_CAPACITY_MAX_TC_MASK) \
-                >> SMU_DEVICE_CONTEXT_CAPACITY_MAX_TC_SHIFT) + 1 \
-       )
-
-/**
- * smu_dcc_get_max_remote_node_context() -
- *
- * This macro returns the maximum number of remote node contexts supported by
- * the hardware. The caller passes in the value read from the device context
- * capacity register and this macro will mash and shift the value appropriately.
- */
-#define smu_dcc_get_max_remote_node_context(dcc_value) \
-       (\
-               (((dcc_value) & SMU_DEVICE_CONTEXT_CAPACITY_MAX_RNC_MASK) \
-                >> SMU_DEVICE_CONTEXT_CAPACITY_MAX_RNC_SHIFT) + 1 \
-       )
-
-
-static void scic_sds_controller_power_control_timer_handler(
-       void *controller);
-#define SCIC_SDS_CONTROLLER_MIN_TIMER_COUNT  3
-#define SCIC_SDS_CONTROLLER_MAX_TIMER_COUNT  3
-
-/**
- *
- *
- * The number of milliseconds to wait for a phy to start.
- */
-#define SCIC_SDS_CONTROLLER_PHY_START_TIMEOUT      100
-
-/**
- *
- *
- * The number of milliseconds to wait while a given phy is consuming power
- * before allowing another set of phys to consume power. Ultimately, this will
- * be specified by OEM parameter.
- */
-#define SCIC_SDS_CONTROLLER_POWER_CONTROL_INTERVAL 500
-
-/**
- * COMPLETION_QUEUE_CYCLE_BIT() -
- *
- * This macro will return the cycle bit of the completion queue entry
- */
-#define COMPLETION_QUEUE_CYCLE_BIT(x) ((x) & 0x80000000)
-
-/**
- * NORMALIZE_GET_POINTER() -
- *
- * This macro will normalize the completion queue get pointer so its value can
- * be used as an index into an array
- */
-#define NORMALIZE_GET_POINTER(x) \
-       ((x) & SMU_COMPLETION_QUEUE_GET_POINTER_MASK)
-
-/**
- * NORMALIZE_PUT_POINTER() -
- *
- * This macro will normalize the completion queue put pointer so its value can
- * be used as an array inde
- */
-#define NORMALIZE_PUT_POINTER(x) \
-       ((x) & SMU_COMPLETION_QUEUE_PUT_POINTER_MASK)
-
-
-/**
- * NORMALIZE_GET_POINTER_CYCLE_BIT() -
- *
- * This macro will normalize the completion queue cycle pointer so it matches
- * the completion queue cycle bit
- */
-#define NORMALIZE_GET_POINTER_CYCLE_BIT(x) \
-       ((SMU_CQGR_CYCLE_BIT & (x)) << (31 - SMU_COMPLETION_QUEUE_GET_CYCLE_BIT_SHIFT))
-
-/**
- * NORMALIZE_EVENT_POINTER() -
- *
- * This macro will normalize the completion queue event entry so its value can
- * be used as an index.
- */
-#define NORMALIZE_EVENT_POINTER(x) \
-       (\
-               ((x) & SMU_COMPLETION_QUEUE_GET_EVENT_POINTER_MASK) \
-               >> SMU_COMPLETION_QUEUE_GET_EVENT_POINTER_SHIFT \
-       )
-
-/**
- * INCREMENT_COMPLETION_QUEUE_GET() -
- *
- * This macro will increment the controllers completion queue index value and
- * possibly toggle the cycle bit if the completion queue index wraps back to 0.
- */
-#define INCREMENT_COMPLETION_QUEUE_GET(controller, index, cycle) \
-       INCREMENT_QUEUE_GET(\
-               (index), \
-               (cycle), \
-               (controller)->completion_queue_entries, \
-               SMU_CQGR_CYCLE_BIT \
-               )
-
-/**
- * INCREMENT_EVENT_QUEUE_GET() -
- *
- * This macro will increment the controllers event queue index value and
- * possibly toggle the event cycle bit if the event queue index wraps back to 0.
- */
-#define INCREMENT_EVENT_QUEUE_GET(controller, index, cycle) \
-       INCREMENT_QUEUE_GET(\
-               (index), \
-               (cycle), \
-               (controller)->completion_event_entries, \
-               SMU_CQGR_EVENT_CYCLE_BIT \
-               )
-
-static void scic_sds_controller_initialize_power_control(struct scic_sds_controller *scic)
-{
-       struct isci_host *ihost = scic_to_ihost(scic);
-       scic->power_control.timer = isci_timer_create(ihost,
-                                                     scic,
-                                       scic_sds_controller_power_control_timer_handler);
-
-       memset(scic->power_control.requesters, 0,
-              sizeof(scic->power_control.requesters));
-
-       scic->power_control.phys_waiting = 0;
-       scic->power_control.phys_granted_power = 0;
-}
-
-int scic_controller_mem_init(struct scic_sds_controller *scic)
-{
-       struct device *dev = scic_to_dev(scic);
-       dma_addr_t dma_handle;
-       enum sci_status result;
-
-       scic->completion_queue = dmam_alloc_coherent(dev,
-                       scic->completion_queue_entries * sizeof(u32),
-                       &dma_handle, GFP_KERNEL);
-       if (!scic->completion_queue)
-               return -ENOMEM;
-
-       writel(lower_32_bits(dma_handle),
-               &scic->smu_registers->completion_queue_lower);
-       writel(upper_32_bits(dma_handle),
-               &scic->smu_registers->completion_queue_upper);
-
-       scic->remote_node_context_table = dmam_alloc_coherent(dev,
-                       scic->remote_node_entries *
-                               sizeof(union scu_remote_node_context),
-                       &dma_handle, GFP_KERNEL);
-       if (!scic->remote_node_context_table)
-               return -ENOMEM;
-
-       writel(lower_32_bits(dma_handle),
-               &scic->smu_registers->remote_node_context_lower);
-       writel(upper_32_bits(dma_handle),
-               &scic->smu_registers->remote_node_context_upper);
-
-       scic->task_context_table = dmam_alloc_coherent(dev,
-                       scic->task_context_entries *
-                               sizeof(struct scu_task_context),
-                       &dma_handle, GFP_KERNEL);
-       if (!scic->task_context_table)
-               return -ENOMEM;
-
-       writel(lower_32_bits(dma_handle),
-               &scic->smu_registers->host_task_table_lower);
-       writel(upper_32_bits(dma_handle),
-               &scic->smu_registers->host_task_table_upper);
-
-       result = scic_sds_unsolicited_frame_control_construct(scic);
-       if (result)
-               return result;
-
-       /*
-        * Inform the silicon as to the location of the UF headers and
-        * address table.
-        */
-       writel(lower_32_bits(scic->uf_control.headers.physical_address),
-               &scic->scu_registers->sdma.uf_header_base_address_lower);
-       writel(upper_32_bits(scic->uf_control.headers.physical_address),
-               &scic->scu_registers->sdma.uf_header_base_address_upper);
-
-       writel(lower_32_bits(scic->uf_control.address_table.physical_address),
-               &scic->scu_registers->sdma.uf_address_table_lower);
-       writel(upper_32_bits(scic->uf_control.address_table.physical_address),
-               &scic->scu_registers->sdma.uf_address_table_upper);
-
-       return 0;
-}
-
-/**
- * This method initializes the task context data for the controller.
- * @scic:
- *
- */
-static void
-scic_sds_controller_assign_task_entries(struct scic_sds_controller *controller)
-{
-       u32 task_assignment;
-
-       /*
-        * Assign all the TCs to function 0
-        * TODO: Do we actually need to read this register to write it back?
-        */
-
-       task_assignment =
-               readl(&controller->smu_registers->task_context_assignment[0]);
-
-       task_assignment |= (SMU_TCA_GEN_VAL(STARTING, 0)) |
-               (SMU_TCA_GEN_VAL(ENDING,  controller->task_context_entries - 1)) |
-               (SMU_TCA_GEN_BIT(RANGE_CHECK_ENABLE));
-
-       writel(task_assignment,
-               &controller->smu_registers->task_context_assignment[0]);
-
-}
-
-/**
- * This method initializes the hardware completion queue.
- *
- *
- */
-static void scic_sds_controller_initialize_completion_queue(
-       struct scic_sds_controller *scic)
-{
-       u32 index;
-       u32 completion_queue_control_value;
-       u32 completion_queue_get_value;
-       u32 completion_queue_put_value;
-
-       scic->completion_queue_get = 0;
-
-       completion_queue_control_value = (
-               SMU_CQC_QUEUE_LIMIT_SET(scic->completion_queue_entries - 1)
-               | SMU_CQC_EVENT_LIMIT_SET(scic->completion_event_entries - 1)
-               );
-
-       writel(completion_queue_control_value,
-              &scic->smu_registers->completion_queue_control);
-
-
-       /* Set the completion queue get pointer and enable the queue */
-       completion_queue_get_value = (
-               (SMU_CQGR_GEN_VAL(POINTER, 0))
-               | (SMU_CQGR_GEN_VAL(EVENT_POINTER, 0))
-               | (SMU_CQGR_GEN_BIT(ENABLE))
-               | (SMU_CQGR_GEN_BIT(EVENT_ENABLE))
-               );
-
-       writel(completion_queue_get_value,
-              &scic->smu_registers->completion_queue_get);
-
-       /* Set the completion queue put pointer */
-       completion_queue_put_value = (
-               (SMU_CQPR_GEN_VAL(POINTER, 0))
-               | (SMU_CQPR_GEN_VAL(EVENT_POINTER, 0))
-               );
-
-       writel(completion_queue_put_value,
-              &scic->smu_registers->completion_queue_put);
-
-       /* Initialize the cycle bit of the completion queue entries */
-       for (index = 0; index < scic->completion_queue_entries; index++) {
-               /*
-                * If get.cycle_bit != completion_queue.cycle_bit
-                * its not a valid completion queue entry
-                * so at system start all entries are invalid */
-               scic->completion_queue[index] = 0x80000000;
-       }
-}
-
-/**
- * This method initializes the hardware unsolicited frame queue.
- *
- *
- */
-static void scic_sds_controller_initialize_unsolicited_frame_queue(
-       struct scic_sds_controller *scic)
-{
-       u32 frame_queue_control_value;
-       u32 frame_queue_get_value;
-       u32 frame_queue_put_value;
-
-       /* Write the queue size */
-       frame_queue_control_value =
-               SCU_UFQC_GEN_VAL(QUEUE_SIZE,
-                                scic->uf_control.address_table.count);
-
-       writel(frame_queue_control_value,
-              &scic->scu_registers->sdma.unsolicited_frame_queue_control);
-
-       /* Setup the get pointer for the unsolicited frame queue */
-       frame_queue_get_value = (
-               SCU_UFQGP_GEN_VAL(POINTER, 0)
-               |  SCU_UFQGP_GEN_BIT(ENABLE_BIT)
-               );
-
-       writel(frame_queue_get_value,
-              &scic->scu_registers->sdma.unsolicited_frame_get_pointer);
-       /* Setup the put pointer for the unsolicited frame queue */
-       frame_queue_put_value = SCU_UFQPP_GEN_VAL(POINTER, 0);
-       writel(frame_queue_put_value,
-              &scic->scu_registers->sdma.unsolicited_frame_put_pointer);
-}
-
-/**
- * This method enables the hardware port task scheduler.
- *
- *
- */
-static void scic_sds_controller_enable_port_task_scheduler(
-       struct scic_sds_controller *scic)
-{
-       u32 port_task_scheduler_value;
-
-       port_task_scheduler_value =
-               readl(&scic->scu_registers->peg0.ptsg.control);
-       port_task_scheduler_value |=
-               (SCU_PTSGCR_GEN_BIT(ETM_ENABLE) |
-                SCU_PTSGCR_GEN_BIT(PTSG_ENABLE));
-       writel(port_task_scheduler_value,
-              &scic->scu_registers->peg0.ptsg.control);
-}
-
-/**
- *
- *
- * This macro is used to delay between writes to the AFE registers during AFE
- * initialization.
- */
-#define AFE_REGISTER_WRITE_DELAY 10
-
-/* Initialize the AFE for this phy index. We need to read the AFE setup from
- * the OEM parameters none
- */
-static void scic_sds_controller_afe_initialization(struct scic_sds_controller *scic)
-{
-       const struct scic_sds_oem_params *oem = &scic->oem_parameters.sds1;
-       u32 afe_status;
-       u32 phy_id;
-
-       /* Clear DFX Status registers */
-       writel(0x0081000f, &scic->scu_registers->afe.afe_dfx_master_control0);
-       udelay(AFE_REGISTER_WRITE_DELAY);
-
-       if (is_b0()) {
-               /* PM Rx Equalization Save, PM SPhy Rx Acknowledgement
-                * Timer, PM Stagger Timer */
-               writel(0x0007BFFF, &scic->scu_registers->afe.afe_pmsn_master_control2);
-               udelay(AFE_REGISTER_WRITE_DELAY);
-       }
-
-       /* Configure bias currents to normal */
-       if (is_a0())
-               writel(0x00005500, &scic->scu_registers->afe.afe_bias_control);
-       else if (is_a2())
-               writel(0x00005A00, &scic->scu_registers->afe.afe_bias_control);
-       else if (is_b0())
-               writel(0x00005F00, &scic->scu_registers->afe.afe_bias_control);
-
-       udelay(AFE_REGISTER_WRITE_DELAY);
-
-       /* Enable PLL */
-       if (is_b0())
-               writel(0x80040A08, &scic->scu_registers->afe.afe_pll_control0);
-       else
-               writel(0x80040908, &scic->scu_registers->afe.afe_pll_control0);
-
-       udelay(AFE_REGISTER_WRITE_DELAY);
-
-       /* Wait for the PLL to lock */
-       do {
-               afe_status = readl(&scic->scu_registers->afe.afe_common_block_status);
-               udelay(AFE_REGISTER_WRITE_DELAY);
-       } while ((afe_status & 0x00001000) == 0);
-
-       if (is_a0() || is_a2()) {
-               /* Shorten SAS SNW lock time (RxLock timer value from 76 us to 50 us) */
-               writel(0x7bcc96ad, &scic->scu_registers->afe.afe_pmsn_master_control0);
-               udelay(AFE_REGISTER_WRITE_DELAY);
-       }
-
-       for (phy_id = 0; phy_id < SCI_MAX_PHYS; phy_id++) {
-               const struct sci_phy_oem_params *oem_phy = &oem->phys[phy_id];
-
-               if (is_b0()) {
-                        /* Configure transmitter SSC parameters */
-                       writel(0x00030000, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_ssc_control);
-                       udelay(AFE_REGISTER_WRITE_DELAY);
-               } else {
-                       /*
-                        * All defaults, except the Receive Word Alignament/Comma Detect
-                        * Enable....(0xe800) */
-                       writel(0x00004512, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_xcvr_control0);
-                       udelay(AFE_REGISTER_WRITE_DELAY);
-
-                       writel(0x0050100F, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_xcvr_control1);
-                       udelay(AFE_REGISTER_WRITE_DELAY);
-               }
-
-               /*
-                * Power up TX and RX out from power down (PWRDNTX and PWRDNRX)
-                * & increase TX int & ext bias 20%....(0xe85c) */
-               if (is_a0())
-                       writel(0x000003D4, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_channel_control);
-               else if (is_a2())
-                       writel(0x000003F0, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_channel_control);
-               else {
-                        /* Power down TX and RX (PWRDNTX and PWRDNRX) */
-                       writel(0x000003d7, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_channel_control);
-                       udelay(AFE_REGISTER_WRITE_DELAY);
-
-                       /*
-                        * Power up TX and RX out from power down (PWRDNTX and PWRDNRX)
-                        * & increase TX int & ext bias 20%....(0xe85c) */
-                       writel(0x000003d4, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_channel_control);
-               }
-               udelay(AFE_REGISTER_WRITE_DELAY);
-
-               if (is_a0() || is_a2()) {
-                       /* Enable TX equalization (0xe824) */
-                       writel(0x00040000, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_control);
-                       udelay(AFE_REGISTER_WRITE_DELAY);
-               }
-
-               /*
-                * RDPI=0x0(RX Power On), RXOOBDETPDNC=0x0, TPD=0x0(TX Power On),
-                * RDD=0x0(RX Detect Enabled) ....(0xe800) */
-               writel(0x00004100, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_xcvr_control0);
-               udelay(AFE_REGISTER_WRITE_DELAY);
-
-               /* Leave DFE/FFE on */
-               if (is_a0())
-                       writel(0x3F09983F, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_rx_ssc_control0);
-               else if (is_a2())
-                       writel(0x3F11103F, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_rx_ssc_control0);
-               else {
-                       writel(0x3F11103F, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_rx_ssc_control0);
-                       udelay(AFE_REGISTER_WRITE_DELAY);
-                       /* Enable TX equalization (0xe824) */
-                       writel(0x00040000, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_control);
-               }
-               udelay(AFE_REGISTER_WRITE_DELAY);
-
-               writel(oem_phy->afe_tx_amp_control0,
-                       &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_amp_control0);
-               udelay(AFE_REGISTER_WRITE_DELAY);
-
-               writel(oem_phy->afe_tx_amp_control1,
-                       &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_amp_control1);
-               udelay(AFE_REGISTER_WRITE_DELAY);
-
-               writel(oem_phy->afe_tx_amp_control2,
-                       &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_amp_control2);
-               udelay(AFE_REGISTER_WRITE_DELAY);
-
-               writel(oem_phy->afe_tx_amp_control3,
-                       &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_amp_control3);
-               udelay(AFE_REGISTER_WRITE_DELAY);
-       }
-
-       /* Transfer control to the PEs */
-       writel(0x00010f00, &scic->scu_registers->afe.afe_dfx_master_control0);
-       udelay(AFE_REGISTER_WRITE_DELAY);
-}
-
-/*
- * ****************************************************************************-
- * * SCIC SDS Controller Internal Start/Stop Routines
- * ****************************************************************************- */
-
-
-/**
- * This method will attempt to transition into the ready state for the
- *    controller and indicate that the controller start operation has completed
- *    if all criteria are met.
- * @scic: This parameter indicates the controller object for which
- *    to transition to ready.
- * @status: This parameter indicates the status value to be pass into the call
- *    to scic_cb_controller_start_complete().
- *
- * none.
- */
-static void scic_sds_controller_transition_to_ready(
-       struct scic_sds_controller *scic,
-       enum sci_status status)
-{
-       struct isci_host *ihost = scic_to_ihost(scic);
-
-       if (scic->state_machine.current_state_id ==
-           SCI_BASE_CONTROLLER_STATE_STARTING) {
-               /*
-                * We move into the ready state, because some of the phys/ports
-                * may be up and operational.
-                */
-               sci_base_state_machine_change_state(&scic->state_machine,
-                                                   SCI_BASE_CONTROLLER_STATE_READY);
-
-               isci_host_start_complete(ihost, status);
-       }
-}
-
-static void scic_sds_controller_timeout_handler(void *_scic)
-{
-       struct scic_sds_controller *scic = _scic;
-       struct isci_host *ihost = scic_to_ihost(scic);
-       struct sci_base_state_machine *sm = &scic->state_machine;
-
-       if (sm->current_state_id == SCI_BASE_CONTROLLER_STATE_STARTING)
-               scic_sds_controller_transition_to_ready(scic, SCI_FAILURE_TIMEOUT);
-       else if (sm->current_state_id == SCI_BASE_CONTROLLER_STATE_STOPPING) {
-               sci_base_state_machine_change_state(sm, SCI_BASE_CONTROLLER_STATE_FAILED);
-               isci_host_stop_complete(ihost, SCI_FAILURE_TIMEOUT);
-       } else  /* / @todo Now what do we want to do in this case? */
-               dev_err(scic_to_dev(scic),
-                       "%s: Controller timer fired when controller was not "
-                       "in a state being timed.\n",
-                       __func__);
-}
-
-static enum sci_status scic_sds_controller_stop_ports(struct scic_sds_controller *scic)
-{
-       u32 index;
-       enum sci_status port_status;
-       enum sci_status status = SCI_SUCCESS;
-       struct isci_host *ihost = scic_to_ihost(scic);
-
-       for (index = 0; index < scic->logical_port_entries; index++) {
-               struct scic_sds_port *sci_port = &ihost->ports[index].sci;
-               scic_sds_port_handler_t stop;
-
-               stop = sci_port->state_handlers->stop_handler;
-               port_status = stop(sci_port);
-
-               if ((port_status != SCI_SUCCESS) &&
-                   (port_status != SCI_FAILURE_INVALID_STATE)) {
-                       status = SCI_FAILURE;
-
-                       dev_warn(scic_to_dev(scic),
-                                "%s: Controller stop operation failed to "
-                                "stop port %d because of status %d.\n",
-                                __func__,
-                                sci_port->logical_port_index,
-                                port_status);
-               }
-       }
-
-       return status;
-}
-
-static inline void scic_sds_controller_phy_timer_start(
-               struct scic_sds_controller *scic)
-{
-       isci_timer_start(scic->phy_startup_timer,
-                        SCIC_SDS_CONTROLLER_PHY_START_TIMEOUT);
-
-       scic->phy_startup_timer_pending = true;
-}
-
-static void scic_sds_controller_phy_timer_stop(struct scic_sds_controller *scic)
-{
-       isci_timer_stop(scic->phy_startup_timer);
-
-       scic->phy_startup_timer_pending = false;
-}
-
-/**
- * scic_sds_controller_start_next_phy - start phy
- * @scic: controller
- *
- * If all the phys have been started, then attempt to transition the
- * controller to the READY state and inform the user
- * (scic_cb_controller_start_complete()).
- */
-static enum sci_status scic_sds_controller_start_next_phy(struct scic_sds_controller *scic)
-{
-       struct isci_host *ihost = scic_to_ihost(scic);
-       struct scic_sds_oem_params *oem = &scic->oem_parameters.sds1;
-       struct scic_sds_phy *sci_phy;
-       enum sci_status status;
-
-       status = SCI_SUCCESS;
-
-       if (scic->phy_startup_timer_pending)
-               return status;
-
-       if (scic->next_phy_to_start >= SCI_MAX_PHYS) {
-               bool is_controller_start_complete = true;
-               u32 state;
-               u8 index;
-
-               for (index = 0; index < SCI_MAX_PHYS; index++) {
-                       sci_phy = &ihost->phys[index].sci;
-                       state = sci_phy->state_machine.current_state_id;
-
-                       if (!scic_sds_phy_get_port(sci_phy))
-                               continue;
-
-                       /* The controller start operation is complete iff:
-                        * - all links have been given an opportunity to start
-                        * - have no indication of a connected device
-                        * - have an indication of a connected device and it has
-                        *   finished the link training process.
-                        */
-                       if ((sci_phy->is_in_link_training == false &&
-                            state == SCI_BASE_PHY_STATE_INITIAL) ||
-                           (sci_phy->is_in_link_training == false &&
-                            state == SCI_BASE_PHY_STATE_STOPPED) ||
-                           (sci_phy->is_in_link_training == true &&
-                            state == SCI_BASE_PHY_STATE_STARTING)) {
-                               is_controller_start_complete = false;
-                               break;
-                       }
-               }
-
-               /*
-                * The controller has successfully finished the start process.
-                * Inform the SCI Core user and transition to the READY state. */
-               if (is_controller_start_complete == true) {
-                       scic_sds_controller_transition_to_ready(scic, SCI_SUCCESS);
-                       scic_sds_controller_phy_timer_stop(scic);
-               }
-       } else {
-               sci_phy = &ihost->phys[scic->next_phy_to_start].sci;
-
-               if (oem->controller.mode_type == SCIC_PORT_MANUAL_CONFIGURATION_MODE) {
-                       if (scic_sds_phy_get_port(sci_phy) == NULL) {
-                               scic->next_phy_to_start++;
-
-                               /* Caution recursion ahead be forwarned
-                                *
-                                * The PHY was never added to a PORT in MPC mode
-                                * so start the next phy in sequence This phy
-                                * will never go link up and will not draw power
-                                * the OEM parameters either configured the phy
-                                * incorrectly for the PORT or it was never
-                                * assigned to a PORT
-                                */
-                               return scic_sds_controller_start_next_phy(scic);
-                       }
-               }
-
-               status = scic_sds_phy_start(sci_phy);
-
-               if (status == SCI_SUCCESS) {
-                       scic_sds_controller_phy_timer_start(scic);
-               } else {
-                       dev_warn(scic_to_dev(scic),
-                                "%s: Controller stop operation failed "
-                                "to stop phy %d because of status "
-                                "%d.\n",
-                                __func__,
-                                ihost->phys[scic->next_phy_to_start].sci.phy_index,
-                                status);
-               }
-
-               scic->next_phy_to_start++;
-       }
-
-       return status;
-}
-
-static void scic_sds_controller_phy_startup_timeout_handler(void *_scic)
-{
-       struct scic_sds_controller *scic = _scic;
-       enum sci_status status;
-
-       scic->phy_startup_timer_pending = false;
-       status = SCI_FAILURE;
-       while (status != SCI_SUCCESS)
-               status = scic_sds_controller_start_next_phy(scic);
-}
-
-static enum sci_status scic_sds_controller_initialize_phy_startup(struct scic_sds_controller *scic)
-{
-       struct isci_host *ihost = scic_to_ihost(scic);
-
-       scic->phy_startup_timer = isci_timer_create(ihost,
-                                                   scic,
-                                                   scic_sds_controller_phy_startup_timeout_handler);
-
-       if (scic->phy_startup_timer == NULL)
-               return SCI_FAILURE_INSUFFICIENT_RESOURCES;
-       else {
-               scic->next_phy_to_start = 0;
-               scic->phy_startup_timer_pending = false;
-       }
-
-       return SCI_SUCCESS;
-}
-
-static enum sci_status scic_sds_controller_stop_phys(struct scic_sds_controller *scic)
-{
-       u32 index;
-       enum sci_status status;
-       enum sci_status phy_status;
-       struct isci_host *ihost = scic_to_ihost(scic);
-
-       status = SCI_SUCCESS;
-
-       for (index = 0; index < SCI_MAX_PHYS; index++) {
-               phy_status = scic_sds_phy_stop(&ihost->phys[index].sci);
-
-               if (phy_status != SCI_SUCCESS &&
-                   phy_status != SCI_FAILURE_INVALID_STATE) {
-                       status = SCI_FAILURE;
-
-                       dev_warn(scic_to_dev(scic),
-                                "%s: Controller stop operation failed to stop "
-                                "phy %d because of status %d.\n",
-                                __func__,
-                                ihost->phys[index].sci.phy_index, phy_status);
-               }
-       }
-
-       return status;
-}
-
-static enum sci_status scic_sds_controller_stop_devices(struct scic_sds_controller *scic)
-{
-       u32 index;
-       enum sci_status status;
-       enum sci_status device_status;
-
-       status = SCI_SUCCESS;
-
-       for (index = 0; index < scic->remote_node_entries; index++) {
-               if (scic->device_table[index] != NULL) {
-                       /* / @todo What timeout value do we want to provide to this request? */
-                       device_status = scic_remote_device_stop(scic->device_table[index], 0);
-
-                       if ((device_status != SCI_SUCCESS) &&
-                           (device_status != SCI_FAILURE_INVALID_STATE)) {
-                               dev_warn(scic_to_dev(scic),
-                                        "%s: Controller stop operation failed "
-                                        "to stop device 0x%p because of "
-                                        "status %d.\n",
-                                        __func__,
-                                        scic->device_table[index], device_status);
-                       }
-               }
-       }
-
-       return status;
-}
-
-static void scic_sds_controller_power_control_timer_start(struct scic_sds_controller *scic)
-{
-       isci_timer_start(scic->power_control.timer,
-                        SCIC_SDS_CONTROLLER_POWER_CONTROL_INTERVAL);
-
-       scic->power_control.timer_started = true;
-}
-
-static void scic_sds_controller_power_control_timer_stop(struct scic_sds_controller *scic)
-{
-       if (scic->power_control.timer_started) {
-               isci_timer_stop(scic->power_control.timer);
-               scic->power_control.timer_started = false;
-       }
-}
-
-static void scic_sds_controller_power_control_timer_restart(struct scic_sds_controller *scic)
-{
-       scic_sds_controller_power_control_timer_stop(scic);
-       scic_sds_controller_power_control_timer_start(scic);
-}
-
-static void scic_sds_controller_power_control_timer_handler(
-       void *controller)
-{
-       struct scic_sds_controller *scic;
-
-       scic = (struct scic_sds_controller *)controller;
-
-       scic->power_control.phys_granted_power = 0;
-
-       if (scic->power_control.phys_waiting == 0) {
-               scic->power_control.timer_started = false;
-       } else {
-               struct scic_sds_phy *sci_phy = NULL;
-               u8 i;
-
-               for (i = 0;
-                    (i < SCI_MAX_PHYS)
-                    && (scic->power_control.phys_waiting != 0);
-                    i++) {
-                       if (scic->power_control.requesters[i] != NULL) {
-                               if (scic->power_control.phys_granted_power <
-                                   scic->oem_parameters.sds1.controller.max_concurrent_dev_spin_up) {
-                                       sci_phy = scic->power_control.requesters[i];
-                                       scic->power_control.requesters[i] = NULL;
-                                       scic->power_control.phys_waiting--;
-                                       scic->power_control.phys_granted_power++;
-                                       scic_sds_phy_consume_power_handler(sci_phy);
-                               } else {
-                                       break;
-                               }
-                       }
-               }
-
-               /*
-                * It doesn't matter if the power list is empty, we need to start the
-                * timer in case another phy becomes ready.
-                */
-               scic_sds_controller_power_control_timer_start(scic);
-       }
-}
-
-/**
- * This method inserts the phy in the stagger spinup control queue.
- * @scic:
- *
- *
- */
-void scic_sds_controller_power_control_queue_insert(
-       struct scic_sds_controller *scic,
-       struct scic_sds_phy *sci_phy)
-{
-       BUG_ON(sci_phy == NULL);
-
-       if (scic->power_control.phys_granted_power <
-           scic->oem_parameters.sds1.controller.max_concurrent_dev_spin_up) {
-               scic->power_control.phys_granted_power++;
-               scic_sds_phy_consume_power_handler(sci_phy);
-
-               /*
-                * stop and start the power_control timer. When the timer fires, the
-                * no_of_phys_granted_power will be set to 0
-                */
-               scic_sds_controller_power_control_timer_restart(scic);
-       } else {
-               /* Add the phy in the waiting list */
-               scic->power_control.requesters[sci_phy->phy_index] = sci_phy;
-               scic->power_control.phys_waiting++;
-       }
-}
-
-/**
- * This method removes the phy from the stagger spinup control queue.
- * @scic:
- *
- *
- */
-void scic_sds_controller_power_control_queue_remove(
-       struct scic_sds_controller *scic,
-       struct scic_sds_phy *sci_phy)
-{
-       BUG_ON(sci_phy == NULL);
-
-       if (scic->power_control.requesters[sci_phy->phy_index] != NULL) {
-               scic->power_control.phys_waiting--;
-       }
-
-       scic->power_control.requesters[sci_phy->phy_index] = NULL;
-}
-
-/*
- * ****************************************************************************-
- * * SCIC SDS Controller Completion Routines
- * ****************************************************************************- */
-
-/**
- * This method returns a true value if the completion queue has entries that
- *    can be processed
- * @scic:
- *
- * bool true if the completion queue has entries to process false if the
- * completion queue has no entries to process
- */
-static bool scic_sds_controller_completion_queue_has_entries(
-       struct scic_sds_controller *scic)
-{
-       u32 get_value = scic->completion_queue_get;
-       u32 get_index = get_value & SMU_COMPLETION_QUEUE_GET_POINTER_MASK;
-
-       if (NORMALIZE_GET_POINTER_CYCLE_BIT(get_value) ==
-           COMPLETION_QUEUE_CYCLE_BIT(scic->completion_queue[get_index]))
-               return true;
-
-       return false;
-}
-
-/**
- * This method processes a task completion notification.  This is called from
- *    within the controller completion handler.
- * @scic:
- * @completion_entry:
- *
- */
-static void scic_sds_controller_task_completion(
-       struct scic_sds_controller *scic,
-       u32 completion_entry)
-{
-       u32 index;
-       struct scic_sds_request *io_request;
-
-       index = SCU_GET_COMPLETION_INDEX(completion_entry);
-       io_request = scic->io_request_table[index];
-
-       /* Make sure that we really want to process this IO request */
-       if (
-               (io_request != NULL)
-               && (io_request->io_tag != SCI_CONTROLLER_INVALID_IO_TAG)
-               && (
-                       scic_sds_io_tag_get_sequence(io_request->io_tag)
-                       == scic->io_request_sequence[index]
-                       )
-               ) {
-               /* Yep this is a valid io request pass it along to the io request handler */
-               scic_sds_io_request_tc_completion(io_request, completion_entry);
-       }
-}
-
-/**
- * This method processes an SDMA completion event.  This is called from within
- *    the controller completion handler.
- * @scic:
- * @completion_entry:
- *
- */
-static void scic_sds_controller_sdma_completion(
-       struct scic_sds_controller *scic,
-       u32 completion_entry)
-{
-       u32 index;
-       struct scic_sds_request *io_request;
-       struct scic_sds_remote_device *device;
-
-       index = SCU_GET_COMPLETION_INDEX(completion_entry);
-
-       switch (scu_get_command_request_type(completion_entry)) {
-       case SCU_CONTEXT_COMMAND_REQUEST_TYPE_POST_TC:
-       case SCU_CONTEXT_COMMAND_REQUEST_TYPE_DUMP_TC:
-               io_request = scic->io_request_table[index];
-               dev_warn(scic_to_dev(scic),
-                        "%s: SCIC SDS Completion type SDMA %x for io request "
-                        "%p\n",
-                        __func__,
-                        completion_entry,
-                        io_request);
-               /* @todo For a post TC operation we need to fail the IO
-                * request
-                */
-               break;
-
-       case SCU_CONTEXT_COMMAND_REQUEST_TYPE_DUMP_RNC:
-       case SCU_CONTEXT_COMMAND_REQUEST_TYPE_OTHER_RNC:
-       case SCU_CONTEXT_COMMAND_REQUEST_TYPE_POST_RNC:
-               device = scic->device_table[index];
-               dev_warn(scic_to_dev(scic),
-                        "%s: SCIC SDS Completion type SDMA %x for remote "
-                        "device %p\n",
-                        __func__,
-                        completion_entry,
-                        device);
-               /* @todo For a port RNC operation we need to fail the
-                * device
-                */
-               break;
-
-       default:
-               dev_warn(scic_to_dev(scic),
-                        "%s: SCIC SDS Completion unknown SDMA completion "
-                        "type %x\n",
-                        __func__,
-                        completion_entry);
-               break;
-
-       }
-}
-
-static void scic_sds_controller_unsolicited_frame(struct scic_sds_controller *scic,
-                                                 u32 completion_entry)
-{
-       u32 index;
-       u32 frame_index;
-
-       struct isci_host *ihost = scic_to_ihost(scic);
-       struct scu_unsolicited_frame_header *frame_header;
-       struct scic_sds_phy *phy;
-       struct scic_sds_remote_device *device;
-
-       enum sci_status result = SCI_FAILURE;
-
-       frame_index = SCU_GET_FRAME_INDEX(completion_entry);
-
-       frame_header = scic->uf_control.buffers.array[frame_index].header;
-       scic->uf_control.buffers.array[frame_index].state = UNSOLICITED_FRAME_IN_USE;
-
-       if (SCU_GET_FRAME_ERROR(completion_entry)) {
-               /*
-                * / @todo If the IAF frame or SIGNATURE FIS frame has an error will
-                * /       this cause a problem? We expect the phy initialization will
-                * /       fail if there is an error in the frame. */
-               scic_sds_controller_release_frame(scic, frame_index);
-               return;
-       }
-
-       if (frame_header->is_address_frame) {
-               index = SCU_GET_PROTOCOL_ENGINE_INDEX(completion_entry);
-               phy = &ihost->phys[index].sci;
-               result = scic_sds_phy_frame_handler(phy, frame_index);
-       } else {
-
-               index = SCU_GET_COMPLETION_INDEX(completion_entry);
-
-               if (index == SCIC_SDS_REMOTE_NODE_CONTEXT_INVALID_INDEX) {
-                       /*
-                        * This is a signature fis or a frame from a direct attached SATA
-                        * device that has not yet been created.  In either case forwared
-                        * the frame to the PE and let it take care of the frame data. */
-                       index = SCU_GET_PROTOCOL_ENGINE_INDEX(completion_entry);
-                       phy = &ihost->phys[index].sci;
-                       result = scic_sds_phy_frame_handler(phy, frame_index);
-               } else {
-                       if (index < scic->remote_node_entries)
-                               device = scic->device_table[index];
-                       else
-                               device = NULL;
-
-                       if (device != NULL)
-                               result = scic_sds_remote_device_frame_handler(device, frame_index);
-                       else
-                               scic_sds_controller_release_frame(scic, frame_index);
-               }
-       }
-
-       if (result != SCI_SUCCESS) {
-               /*
-                * / @todo Is there any reason to report some additional error message
-                * /       when we get this failure notifiction? */
-       }
-}
-
-static void scic_sds_controller_event_completion(struct scic_sds_controller *scic,
-                                                u32 completion_entry)
-{
-       struct isci_host *ihost = scic_to_ihost(scic);
-       struct scic_sds_request *io_request;
-       struct scic_sds_remote_device *device;
-       struct scic_sds_phy *phy;
-       u32 index;
-
-       index = SCU_GET_COMPLETION_INDEX(completion_entry);
-
-       switch (scu_get_event_type(completion_entry)) {
-       case SCU_EVENT_TYPE_SMU_COMMAND_ERROR:
-               /* / @todo The driver did something wrong and we need to fix the condtion. */
-               dev_err(scic_to_dev(scic),
-                       "%s: SCIC Controller 0x%p received SMU command error "
-                       "0x%x\n",
-                       __func__,
-                       scic,
-                       completion_entry);
-               break;
-
-       case SCU_EVENT_TYPE_SMU_PCQ_ERROR:
-       case SCU_EVENT_TYPE_SMU_ERROR:
-       case SCU_EVENT_TYPE_FATAL_MEMORY_ERROR:
-               /*
-                * / @todo This is a hardware failure and its likely that we want to
-                * /       reset the controller. */
-               dev_err(scic_to_dev(scic),
-                       "%s: SCIC Controller 0x%p received fatal controller "
-                       "event  0x%x\n",
-                       __func__,
-                       scic,
-                       completion_entry);
-               break;
-
-       case SCU_EVENT_TYPE_TRANSPORT_ERROR:
-               io_request = scic->io_request_table[index];
-               scic_sds_io_request_event_handler(io_request, completion_entry);
-               break;
-
-       case SCU_EVENT_TYPE_PTX_SCHEDULE_EVENT:
-               switch (scu_get_event_specifier(completion_entry)) {
-               case SCU_EVENT_SPECIFIC_SMP_RESPONSE_NO_PE:
-               case SCU_EVENT_SPECIFIC_TASK_TIMEOUT:
-                       io_request = scic->io_request_table[index];
-                       if (io_request != NULL)
-                               scic_sds_io_request_event_handler(io_request, completion_entry);
-                       else
-                               dev_warn(scic_to_dev(scic),
-                                        "%s: SCIC Controller 0x%p received "
-                                        "event 0x%x for io request object "
-                                        "that doesnt exist.\n",
-                                        __func__,
-                                        scic,
-                                        completion_entry);
-
-                       break;
-
-               case SCU_EVENT_SPECIFIC_IT_NEXUS_TIMEOUT:
-                       device = scic->device_table[index];
-                       if (device != NULL)
-                               scic_sds_remote_device_event_handler(device, completion_entry);
-                       else
-                               dev_warn(scic_to_dev(scic),
-                                        "%s: SCIC Controller 0x%p received "
-                                        "event 0x%x for remote device object "
-                                        "that doesnt exist.\n",
-                                        __func__,
-                                        scic,
-                                        completion_entry);
-
-                       break;
-               }
-               break;
-
-       case SCU_EVENT_TYPE_BROADCAST_CHANGE:
-       /*
-        * direct the broadcast change event to the phy first and then let
-        * the phy redirect the broadcast change to the port object */
-       case SCU_EVENT_TYPE_ERR_CNT_EVENT:
-       /*
-        * direct error counter event to the phy object since that is where
-        * we get the event notification.  This is a type 4 event. */
-       case SCU_EVENT_TYPE_OSSP_EVENT:
-               index = SCU_GET_PROTOCOL_ENGINE_INDEX(completion_entry);
-               phy = &ihost->phys[index].sci;
-               scic_sds_phy_event_handler(phy, completion_entry);
-               break;
-
-       case SCU_EVENT_TYPE_RNC_SUSPEND_TX:
-       case SCU_EVENT_TYPE_RNC_SUSPEND_TX_RX:
-       case SCU_EVENT_TYPE_RNC_OPS_MISC:
-               if (index < scic->remote_node_entries) {
-                       device = scic->device_table[index];
-
-                       if (device != NULL)
-                               scic_sds_remote_device_event_handler(device, completion_entry);
-               } else
-                       dev_err(scic_to_dev(scic),
-                               "%s: SCIC Controller 0x%p received event 0x%x "
-                               "for remote device object 0x%0x that doesnt "
-                               "exist.\n",
-                               __func__,
-                               scic,
-                               completion_entry,
-                               index);
-
-               break;
-
-       default:
-               dev_warn(scic_to_dev(scic),
-                        "%s: SCIC Controller received unknown event code %x\n",
-                        __func__,
-                        completion_entry);
-               break;
-       }
-}
-
-/**
- * This method is a private routine for processing the completion queue entries.
- * @scic:
- *
- */
-static void scic_sds_controller_process_completions(
-       struct scic_sds_controller *scic)
-{
-       u32 completion_count = 0;
-       u32 completion_entry;
-       u32 get_index;
-       u32 get_cycle;
-       u32 event_index;
-       u32 event_cycle;
-
-       dev_dbg(scic_to_dev(scic),
-               "%s: completion queue begining get:0x%08x\n",
-               __func__,
-               scic->completion_queue_get);
-
-       /* Get the component parts of the completion queue */
-       get_index = NORMALIZE_GET_POINTER(scic->completion_queue_get);
-       get_cycle = SMU_CQGR_CYCLE_BIT & scic->completion_queue_get;
-
-       event_index = NORMALIZE_EVENT_POINTER(scic->completion_queue_get);
-       event_cycle = SMU_CQGR_EVENT_CYCLE_BIT & scic->completion_queue_get;
-
-       while (
-               NORMALIZE_GET_POINTER_CYCLE_BIT(get_cycle)
-               == COMPLETION_QUEUE_CYCLE_BIT(scic->completion_queue[get_index])
-               ) {
-               completion_count++;
-
-               completion_entry = scic->completion_queue[get_index];
-               INCREMENT_COMPLETION_QUEUE_GET(scic, get_index, get_cycle);
-
-               dev_dbg(scic_to_dev(scic),
-                       "%s: completion queue entry:0x%08x\n",
-                       __func__,
-                       completion_entry);
-
-               switch (SCU_GET_COMPLETION_TYPE(completion_entry)) {
-               case SCU_COMPLETION_TYPE_TASK:
-                       scic_sds_controller_task_completion(scic, completion_entry);
-                       break;
-
-               case SCU_COMPLETION_TYPE_SDMA:
-                       scic_sds_controller_sdma_completion(scic, completion_entry);
-                       break;
-
-               case SCU_COMPLETION_TYPE_UFI:
-                       scic_sds_controller_unsolicited_frame(scic, completion_entry);
-                       break;
-
-               case SCU_COMPLETION_TYPE_EVENT:
-                       INCREMENT_EVENT_QUEUE_GET(scic, event_index, event_cycle);
-                       scic_sds_controller_event_completion(scic, completion_entry);
-                       break;
-
-               case SCU_COMPLETION_TYPE_NOTIFY:
-                       /*
-                        * Presently we do the same thing with a notify event that we do with the
-                        * other event codes. */
-                       INCREMENT_EVENT_QUEUE_GET(scic, event_index, event_cycle);
-                       scic_sds_controller_event_completion(scic, completion_entry);
-                       break;
-
-               default:
-                       dev_warn(scic_to_dev(scic),
-                                "%s: SCIC Controller received unknown "
-                                "completion type %x\n",
-                                __func__,
-                                completion_entry);
-                       break;
-               }
-       }
-
-       /* Update the get register if we completed one or more entries */
-       if (completion_count > 0) {
-               scic->completion_queue_get =
-                       SMU_CQGR_GEN_BIT(ENABLE) |
-                       SMU_CQGR_GEN_BIT(EVENT_ENABLE) |
-                       event_cycle |
-                       SMU_CQGR_GEN_VAL(EVENT_POINTER, event_index) |
-                       get_cycle |
-                       SMU_CQGR_GEN_VAL(POINTER, get_index);
-
-               writel(scic->completion_queue_get,
-                      &scic->smu_registers->completion_queue_get);
-
-       }
-
-       dev_dbg(scic_to_dev(scic),
-               "%s: completion queue ending get:0x%08x\n",
-               __func__,
-               scic->completion_queue_get);
-
-}
-
-bool scic_sds_controller_isr(struct scic_sds_controller *scic)
-{
-       if (scic_sds_controller_completion_queue_has_entries(scic)) {
-               return true;
-       } else {
-               /*
-                * we have a spurious interrupt it could be that we have already
-                * emptied the completion queue from a previous interrupt */
-               writel(SMU_ISR_COMPLETION, &scic->smu_registers->interrupt_status);
-
-               /*
-                * There is a race in the hardware that could cause us not to be notified
-                * of an interrupt completion if we do not take this step.  We will mask
-                * then unmask the interrupts so if there is another interrupt pending
-                * the clearing of the interrupt source we get the next interrupt message. */
-               writel(0xFF000000, &scic->smu_registers->interrupt_mask);
-               writel(0, &scic->smu_registers->interrupt_mask);
-       }
-
-       return false;
-}
-
-void scic_sds_controller_completion_handler(struct scic_sds_controller *scic)
-{
-       /* Empty out the completion queue */
-       if (scic_sds_controller_completion_queue_has_entries(scic))
-               scic_sds_controller_process_completions(scic);
-
-       /* Clear the interrupt and enable all interrupts again */
-       writel(SMU_ISR_COMPLETION, &scic->smu_registers->interrupt_status);
-       /* Could we write the value of SMU_ISR_COMPLETION? */
-       writel(0xFF000000, &scic->smu_registers->interrupt_mask);
-       writel(0, &scic->smu_registers->interrupt_mask);
-}
-
-bool scic_sds_controller_error_isr(struct scic_sds_controller *scic)
-{
-       u32 interrupt_status;
-
-       interrupt_status =
-               readl(&scic->smu_registers->interrupt_status);
-       interrupt_status &= (SMU_ISR_QUEUE_ERROR | SMU_ISR_QUEUE_SUSPEND);
-
-       if (interrupt_status != 0) {
-               /*
-                * There is an error interrupt pending so let it through and handle
-                * in the callback */
-               return true;
-       }
-
-       /*
-        * There is a race in the hardware that could cause us not to be notified
-        * of an interrupt completion if we do not take this step.  We will mask
-        * then unmask the error interrupts so if there was another interrupt
-        * pending we will be notified.
-        * Could we write the value of (SMU_ISR_QUEUE_ERROR | SMU_ISR_QUEUE_SUSPEND)? */
-       writel(0xff, &scic->smu_registers->interrupt_mask);
-       writel(0, &scic->smu_registers->interrupt_mask);
-
-       return false;
-}
-
-void scic_sds_controller_error_handler(struct scic_sds_controller *scic)
-{
-       u32 interrupt_status;
-
-       interrupt_status =
-               readl(&scic->smu_registers->interrupt_status);
-
-       if ((interrupt_status & SMU_ISR_QUEUE_SUSPEND) &&
-           scic_sds_controller_completion_queue_has_entries(scic)) {
-
-               scic_sds_controller_process_completions(scic);
-               writel(SMU_ISR_QUEUE_SUSPEND, &scic->smu_registers->interrupt_status);
-       } else {
-               dev_err(scic_to_dev(scic), "%s: status: %#x\n", __func__,
-                       interrupt_status);
-
-               sci_base_state_machine_change_state(&scic->state_machine,
-                                                   SCI_BASE_CONTROLLER_STATE_FAILED);
-
-               return;
-       }
-
-       /* If we dont process any completions I am not sure that we want to do this.
-        * We are in the middle of a hardware fault and should probably be reset.
-        */
-       writel(0, &scic->smu_registers->interrupt_mask);
-}
-
-
-
-
-void scic_sds_controller_link_up(struct scic_sds_controller *scic,
-               struct scic_sds_port *port, struct scic_sds_phy *phy)
-{
-       switch (scic->state_machine.current_state_id) {
-       case SCI_BASE_CONTROLLER_STATE_STARTING:
-               scic_sds_controller_phy_timer_stop(scic);
-               scic->port_agent.link_up_handler(scic, &scic->port_agent,
-                                                port, phy);
-               scic_sds_controller_start_next_phy(scic);
-               break;
-       case SCI_BASE_CONTROLLER_STATE_READY:
-               scic->port_agent.link_up_handler(scic, &scic->port_agent,
-                                                port, phy);
-               break;
-       default:
-               dev_dbg(scic_to_dev(scic),
-                       "%s: SCIC Controller linkup event from phy %d in "
-                       "unexpected state %d\n", __func__, phy->phy_index,
-                       scic->state_machine.current_state_id);
-       }
-}
-
-void scic_sds_controller_link_down(struct scic_sds_controller *scic,
-               struct scic_sds_port *port, struct scic_sds_phy *phy)
-{
-       switch (scic->state_machine.current_state_id) {
-       case SCI_BASE_CONTROLLER_STATE_STARTING:
-       case SCI_BASE_CONTROLLER_STATE_READY:
-               scic->port_agent.link_down_handler(scic, &scic->port_agent,
-                                                  port, phy);
-               break;
-       default:
-               dev_dbg(scic_to_dev(scic),
-                       "%s: SCIC Controller linkdown event from phy %d in "
-                       "unexpected state %d\n",
-                       __func__,
-                       phy->phy_index,
-                       scic->state_machine.current_state_id);
-       }
-}
-
-/**
- * This is a helper method to determine if any remote devices on this
- * controller are still in the stopping state.
- *
- */
-static bool scic_sds_controller_has_remote_devices_stopping(
-       struct scic_sds_controller *controller)
-{
-       u32 index;
-
-       for (index = 0; index < controller->remote_node_entries; index++) {
-               if ((controller->device_table[index] != NULL) &&
-                  (controller->device_table[index]->state_machine.current_state_id
-                   == SCI_BASE_REMOTE_DEVICE_STATE_STOPPING))
-                       return true;
-       }
-
-       return false;
-}
-
-/**
- * This method is called by the remote device to inform the controller
- * object that the remote device has stopped.
- */
-void scic_sds_controller_remote_device_stopped(struct scic_sds_controller *scic,
-                                              struct scic_sds_remote_device *sci_dev)
-{
-       if (scic->state_machine.current_state_id !=
-           SCI_BASE_CONTROLLER_STATE_STOPPING) {
-               dev_dbg(scic_to_dev(scic),
-                       "SCIC Controller 0x%p remote device stopped event "
-                       "from device 0x%p in unexpected state %d\n",
-                       scic, sci_dev,
-                       scic->state_machine.current_state_id);
-               return;
-       }
-
-       if (!scic_sds_controller_has_remote_devices_stopping(scic)) {
-               sci_base_state_machine_change_state(&scic->state_machine,
-                               SCI_BASE_CONTROLLER_STATE_STOPPED);
-       }
-}
-
-/**
- * This method will write to the SCU PCP register the request value. The method
- *    is used to suspend/resume ports, devices, and phys.
- * @scic:
- *
- *
- */
-void scic_sds_controller_post_request(
-       struct scic_sds_controller *scic,
-       u32 request)
-{
-       dev_dbg(scic_to_dev(scic),
-               "%s: SCIC Controller 0x%p post request 0x%08x\n",
-               __func__,
-               scic,
-               request);
-
-       writel(request, &scic->smu_registers->post_context_port);
-}
-
-/**
- * This method will copy the soft copy of the task context into the physical
- *    memory accessible by the controller.
- * @scic: This parameter specifies the controller for which to copy
- *    the task context.
- * @sci_req: This parameter specifies the request for which the task
- *    context is being copied.
- *
- * After this call is made the SCIC_SDS_IO_REQUEST object will always point to
- * the physical memory version of the task context. Thus, all subsequent
- * updates to the task context are performed in the TC table (i.e. DMAable
- * memory). none
- */
-void scic_sds_controller_copy_task_context(
-       struct scic_sds_controller *scic,
-       struct scic_sds_request *sci_req)
-{
-       struct scu_task_context *task_context_buffer;
-
-       task_context_buffer = scic_sds_controller_get_task_context_buffer(
-               scic, sci_req->io_tag);
-
-       memcpy(task_context_buffer,
-              sci_req->task_context_buffer,
-              offsetof(struct scu_task_context, sgl_snapshot_ac));
-
-       /*
-        * Now that the soft copy of the TC has been copied into the TC
-        * table accessible by the silicon.  Thus, any further changes to
-        * the TC (e.g. TC termination) occur in the appropriate location. */
-       sci_req->task_context_buffer = task_context_buffer;
-}
-
-/**
- * This method returns the task context buffer for the given io tag.
- * @scic:
- * @io_tag:
- *
- * struct scu_task_context*
- */
-struct scu_task_context *scic_sds_controller_get_task_context_buffer(
-       struct scic_sds_controller *scic,
-       u16 io_tag
-       ) {
-       u16 task_index = scic_sds_io_tag_get_index(io_tag);
-
-       if (task_index < scic->task_context_entries) {
-               return &scic->task_context_table[task_index];
-       }
-
-       return NULL;
-}
-
-struct scic_sds_request *scic_request_by_tag(struct scic_sds_controller *scic,
-                                            u16 io_tag)
-{
-       u16 task_index;
-       u16 task_sequence;
-
-       task_index = scic_sds_io_tag_get_index(io_tag);
-
-       if (task_index  < scic->task_context_entries) {
-               if (scic->io_request_table[task_index] != NULL) {
-                       task_sequence = scic_sds_io_tag_get_sequence(io_tag);
-
-                       if (task_sequence == scic->io_request_sequence[task_index]) {
-                               return scic->io_request_table[task_index];
-                       }
-               }
-       }
-
-       return NULL;
-}
-
-/**
- * This method allocates remote node index and the reserves the remote node
- *    context space for use. This method can fail if there are no more remote
- *    node index available.
- * @scic: This is the controller object which contains the set of
- *    free remote node ids
- * @sci_dev: This is the device object which is requesting the a remote node
- *    id
- * @node_id: This is the remote node id that is assinged to the device if one
- *    is available
- *
- * enum sci_status SCI_FAILURE_OUT_OF_RESOURCES if there are no available remote
- * node index available.
- */
-enum sci_status scic_sds_controller_allocate_remote_node_context(
-       struct scic_sds_controller *scic,
-       struct scic_sds_remote_device *sci_dev,
-       u16 *node_id)
-{
-       u16 node_index;
-       u32 remote_node_count = scic_sds_remote_device_node_count(sci_dev);
-
-       node_index = scic_sds_remote_node_table_allocate_remote_node(
-               &scic->available_remote_nodes, remote_node_count
-               );
-
-       if (node_index != SCIC_SDS_REMOTE_NODE_CONTEXT_INVALID_INDEX) {
-               scic->device_table[node_index] = sci_dev;
-
-               *node_id = node_index;
-
-               return SCI_SUCCESS;
-       }
-
-       return SCI_FAILURE_INSUFFICIENT_RESOURCES;
-}
-
-/**
- * This method frees the remote node index back to the available pool.  Once
- *    this is done the remote node context buffer is no longer valid and can
- *    not be used.
- * @scic:
- * @sci_dev:
- * @node_id:
- *
- */
-void scic_sds_controller_free_remote_node_context(
-       struct scic_sds_controller *scic,
-       struct scic_sds_remote_device *sci_dev,
-       u16 node_id)
-{
-       u32 remote_node_count = scic_sds_remote_device_node_count(sci_dev);
-
-       if (scic->device_table[node_id] == sci_dev) {
-               scic->device_table[node_id] = NULL;
-
-               scic_sds_remote_node_table_release_remote_node_index(
-                       &scic->available_remote_nodes, remote_node_count, node_id
-                       );
-       }
-}
-
-/**
- * This method returns the union scu_remote_node_context for the specified remote
- *    node id.
- * @scic:
- * @node_id:
- *
- * union scu_remote_node_context*
- */
-union scu_remote_node_context *scic_sds_controller_get_remote_node_context_buffer(
-       struct scic_sds_controller *scic,
-       u16 node_id
-       ) {
-       if (
-               (node_id < scic->remote_node_entries)
-               && (scic->device_table[node_id] != NULL)
-               ) {
-               return &scic->remote_node_context_table[node_id];
-       }
-
-       return NULL;
-}
-
-/**
- *
- * @resposne_buffer: This is the buffer into which the D2H register FIS will be
- *    constructed.
- * @frame_header: This is the frame header returned by the hardware.
- * @frame_buffer: This is the frame buffer returned by the hardware.
- *
- * This method will combind the frame header and frame buffer to create a SATA
- * D2H register FIS none
- */
-void scic_sds_controller_copy_sata_response(
-       void *response_buffer,
-       void *frame_header,
-       void *frame_buffer)
-{
-       memcpy(response_buffer, frame_header, sizeof(u32));
-
-       memcpy(response_buffer + sizeof(u32),
-              frame_buffer,
-              sizeof(struct dev_to_host_fis) - sizeof(u32));
-}
-
-/**
- * This method releases the frame once this is done the frame is available for
- *    re-use by the hardware.  The data contained in the frame header and frame
- *    buffer is no longer valid. The UF queue get pointer is only updated if UF
- *    control indicates this is appropriate.
- * @scic:
- * @frame_index:
- *
- */
-void scic_sds_controller_release_frame(
-       struct scic_sds_controller *scic,
-       u32 frame_index)
-{
-       if (scic_sds_unsolicited_frame_control_release_frame(
-                   &scic->uf_control, frame_index) == true)
-               writel(scic->uf_control.get,
-                       &scic->scu_registers->sdma.unsolicited_frame_get_pointer);
-}
-
-/**
- * This method sets user parameters and OEM parameters to default values.
- *    Users can override these values utilizing the scic_user_parameters_set()
- *    and scic_oem_parameters_set() methods.
- * @scic: This parameter specifies the controller for which to set the
- *    configuration parameters to their default values.
- *
- */
-static void scic_sds_controller_set_default_config_parameters(struct scic_sds_controller *scic)
-{
-       struct isci_host *ihost = scic_to_ihost(scic);
-       u16 index;
-
-       /* Default to APC mode. */
-       scic->oem_parameters.sds1.controller.mode_type = SCIC_PORT_AUTOMATIC_CONFIGURATION_MODE;
-
-       /* Default to APC mode. */
-       scic->oem_parameters.sds1.controller.max_concurrent_dev_spin_up = 1;
-
-       /* Default to no SSC operation. */
-       scic->oem_parameters.sds1.controller.do_enable_ssc = false;
-
-       /* Initialize all of the port parameter information to narrow ports. */
-       for (index = 0; index < SCI_MAX_PORTS; index++) {
-               scic->oem_parameters.sds1.ports[index].phy_mask = 0;
-       }
-
-       /* Initialize all of the phy parameter information. */
-       for (index = 0; index < SCI_MAX_PHYS; index++) {
-               /* Default to 6G (i.e. Gen 3) for now. */
-               scic->user_parameters.sds1.phys[index].max_speed_generation = 3;
-
-               /* the frequencies cannot be 0 */
-               scic->user_parameters.sds1.phys[index].align_insertion_frequency = 0x7f;
-               scic->user_parameters.sds1.phys[index].in_connection_align_insertion_frequency = 0xff;
-               scic->user_parameters.sds1.phys[index].notify_enable_spin_up_insertion_frequency = 0x33;
-
-               /*
-                * Previous Vitesse based expanders had a arbitration issue that
-                * is worked around by having the upper 32-bits of SAS address
-                * with a value greater then the Vitesse company identifier.
-                * Hence, usage of 0x5FCFFFFF. */
-               scic->oem_parameters.sds1.phys[index].sas_address.low = 0x1 + ihost->id;
-               scic->oem_parameters.sds1.phys[index].sas_address.high = 0x5FCFFFFF;
-       }
-
-       scic->user_parameters.sds1.stp_inactivity_timeout = 5;
-       scic->user_parameters.sds1.ssp_inactivity_timeout = 5;
-       scic->user_parameters.sds1.stp_max_occupancy_timeout = 5;
-       scic->user_parameters.sds1.ssp_max_occupancy_timeout = 20;
-       scic->user_parameters.sds1.no_outbound_task_timeout = 20;
-}
-
-/**
- * scic_controller_get_suggested_start_timeout() - This method returns the
- *    suggested scic_controller_start() timeout amount.  The user is free to
- *    use any timeout value, but this method provides the suggested minimum
- *    start timeout value.  The returned value is based upon empirical
- *    information determined as a result of interoperability testing.
- * @controller: the handle to the controller object for which to return the
- *    suggested start timeout.
- *
- * This method returns the number of milliseconds for the suggested start
- * operation timeout.
- */
-u32 scic_controller_get_suggested_start_timeout(
-       struct scic_sds_controller *sc)
-{
-       /* Validate the user supplied parameters. */
-       if (sc == NULL)
-               return 0;
-
-       /*
-        * The suggested minimum timeout value for a controller start operation:
-        *
-        *     Signature FIS Timeout
-        *   + Phy Start Timeout
-        *   + Number of Phy Spin Up Intervals
-        *   ---------------------------------
-        *   Number of milliseconds for the controller start operation.
-        *
-        * NOTE: The number of phy spin up intervals will be equivalent
-        *       to the number of phys divided by the number phys allowed
-        *       per interval - 1 (once OEM parameters are supported).
-        *       Currently we assume only 1 phy per interval. */
-
-       return SCIC_SDS_SIGNATURE_FIS_TIMEOUT
-               + SCIC_SDS_CONTROLLER_PHY_START_TIMEOUT
-               + ((SCI_MAX_PHYS - 1) * SCIC_SDS_CONTROLLER_POWER_CONTROL_INTERVAL);
-}
-
-/**
- * scic_controller_stop() - This method will stop an individual controller
- *    object.This method will invoke the associated user callback upon
- *    completion.  The completion callback is called when the following
- *    conditions are met: -# the method return status is SCI_SUCCESS. -# the
- *    controller has been quiesced. This method will ensure that all IO
- *    requests are quiesced, phys are stopped, and all additional operation by
- *    the hardware is halted.
- * @controller: the handle to the controller object to stop.
- * @timeout: This parameter specifies the number of milliseconds in which the
- *    stop operation should complete.
- *
- * The controller must be in the STARTED or STOPPED state. Indicate if the
- * controller stop method succeeded or failed in some way. SCI_SUCCESS if the
- * stop operation successfully began. SCI_WARNING_ALREADY_IN_STATE if the
- * controller is already in the STOPPED state. SCI_FAILURE_INVALID_STATE if the
- * controller is not either in the STARTED or STOPPED states.
- */
-enum sci_status scic_controller_stop(
-       struct scic_sds_controller *scic,
-       u32 timeout)
-{
-       if (scic->state_machine.current_state_id !=
-           SCI_BASE_CONTROLLER_STATE_READY) {
-               dev_warn(scic_to_dev(scic),
-                        "SCIC Controller stop operation requested in "
-                        "invalid state\n");
-               return SCI_FAILURE_INVALID_STATE;
-       }
-
-       isci_timer_start(scic->timeout_timer, timeout);
-       sci_base_state_machine_change_state(&scic->state_machine,
-                                           SCI_BASE_CONTROLLER_STATE_STOPPING);
-       return SCI_SUCCESS;
-}
-
-/**
- * scic_controller_reset() - This method will reset the supplied core
- *    controller regardless of the state of said controller.  This operation is
- *    considered destructive.  In other words, all current operations are wiped
- *    out.  No IO completions for outstanding devices occur.  Outstanding IO
- *    requests are not aborted or completed at the actual remote device.
- * @controller: the handle to the controller object to reset.
- *
- * Indicate if the controller reset method succeeded or failed in some way.
- * SCI_SUCCESS if the reset operation successfully started. SCI_FATAL_ERROR if
- * the controller reset operation is unable to complete.
- */
-enum sci_status scic_controller_reset(
-       struct scic_sds_controller *scic)
-{
-       switch (scic->state_machine.current_state_id) {
-       case SCI_BASE_CONTROLLER_STATE_RESET:
-       case SCI_BASE_CONTROLLER_STATE_READY:
-       case SCI_BASE_CONTROLLER_STATE_STOPPED:
-       case SCI_BASE_CONTROLLER_STATE_FAILED:
-               /*
-                * The reset operation is not a graceful cleanup, just
-                * perform the state transition.
-                */
-               sci_base_state_machine_change_state(&scic->state_machine,
-                               SCI_BASE_CONTROLLER_STATE_RESETTING);
-               return SCI_SUCCESS;
-       default:
-               dev_warn(scic_to_dev(scic),
-                        "SCIC Controller reset operation requested in "
-                        "invalid state\n");
-               return SCI_FAILURE_INVALID_STATE;
-       }
-}
-
-/**
- * scic_controller_start_io() - This method is called by the SCI user to
- *    send/start an IO request. If the method invocation is successful, then
- *    the IO request has been queued to the hardware for processing.
- * @controller: the handle to the controller object for which to start an IO
- *    request.
- * @remote_device: the handle to the remote device object for which to start an
- *    IO request.
- * @io_request: the handle to the io request object to start.
- * @io_tag: This parameter specifies a previously allocated IO tag that the
- *    user desires to be utilized for this request. This parameter is optional.
- *     The user is allowed to supply SCI_CONTROLLER_INVALID_IO_TAG as the value
- *    for this parameter.
- *
- * - IO tags are a protected resource.  It is incumbent upon the SCI Core user
- * to ensure that each of the methods that may allocate or free available IO
- * tags are handled in a mutually exclusive manner.  This method is one of said
- * methods requiring proper critical code section protection (e.g. semaphore,
- * spin-lock, etc.). - For SATA, the user is required to manage NCQ tags.  As a
- * result, it is expected the user will have set the NCQ tag field in the host
- * to device register FIS prior to calling this method.  There is also a
- * requirement for the user to call scic_stp_io_set_ncq_tag() prior to invoking
- * the scic_controller_start_io() method. scic_controller_allocate_tag() for
- * more information on allocating a tag. Indicate if the controller
- * successfully started the IO request. SCI_SUCCESS if the IO request was
- * successfully started. Determine the failure situations and return values.
- */
-enum sci_status scic_controller_start_io(
-       struct scic_sds_controller *scic,
-       struct scic_sds_remote_device *rdev,
-       struct scic_sds_request *req,
-       u16 io_tag)
-{
-       enum sci_status status;
-
-       if (scic->state_machine.current_state_id !=
-           SCI_BASE_CONTROLLER_STATE_READY) {
-               dev_warn(scic_to_dev(scic), "invalid state to start I/O");
-               return SCI_FAILURE_INVALID_STATE;
-       }
-
-       status = scic_sds_remote_device_start_io(scic, rdev, req);
-       if (status != SCI_SUCCESS)
-               return status;
-
-       scic->io_request_table[scic_sds_io_tag_get_index(req->io_tag)] = req;
-       scic_sds_controller_post_request(scic, scic_sds_request_get_post_context(req));
-       return SCI_SUCCESS;
-}
-
-/**
- * scic_controller_terminate_request() - This method is called by the SCI Core
- *    user to terminate an ongoing (i.e. started) core IO request.  This does
- *    not abort the IO request at the target, but rather removes the IO request
- *    from the host controller.
- * @controller: the handle to the controller object for which to terminate a
- *    request.
- * @remote_device: the handle to the remote device object for which to
- *    terminate a request.
- * @request: the handle to the io or task management request object to
- *    terminate.
- *
- * Indicate if the controller successfully began the terminate process for the
- * IO request. SCI_SUCCESS if the terminate process was successfully started
- * for the request. Determine the failure situations and return values.
- */
-enum sci_status scic_controller_terminate_request(
-       struct scic_sds_controller *scic,
-       struct scic_sds_remote_device *rdev,
-       struct scic_sds_request *req)
-{
-       enum sci_status status;
-
-       if (scic->state_machine.current_state_id !=
-           SCI_BASE_CONTROLLER_STATE_READY) {
-               dev_warn(scic_to_dev(scic),
-                        "invalid state to terminate request\n");
-               return SCI_FAILURE_INVALID_STATE;
-       }
-
-       status = scic_sds_io_request_terminate(req);
-       if (status != SCI_SUCCESS)
-               return status;
-
-       /*
-        * Utilize the original post context command and or in the POST_TC_ABORT
-        * request sub-type.
-        */
-       scic_sds_controller_post_request(scic,
-               scic_sds_request_get_post_context(req) |
-               SCU_CONTEXT_COMMAND_REQUEST_POST_TC_ABORT);
-       return SCI_SUCCESS;
-}
-
-/**
- * scic_controller_complete_io() - This method will perform core specific
- *    completion operations for an IO request.  After this method is invoked,
- *    the user should consider the IO request as invalid until it is properly
- *    reused (i.e. re-constructed).
- * @controller: The handle to the controller object for which to complete the
- *    IO request.
- * @remote_device: The handle to the remote device object for which to complete
- *    the IO request.
- * @io_request: the handle to the io request object to complete.
- *
- * - IO tags are a protected resource.  It is incumbent upon the SCI Core user
- * to ensure that each of the methods that may allocate or free available IO
- * tags are handled in a mutually exclusive manner.  This method is one of said
- * methods requiring proper critical code section protection (e.g. semaphore,
- * spin-lock, etc.). - If the IO tag for a request was allocated, by the SCI
- * Core user, using the scic_controller_allocate_io_tag() method, then it is
- * the responsibility of the caller to invoke the scic_controller_free_io_tag()
- * method to free the tag (i.e. this method will not free the IO tag). Indicate
- * if the controller successfully completed the IO request. SCI_SUCCESS if the
- * completion process was successful.
- */
-enum sci_status scic_controller_complete_io(
-       struct scic_sds_controller *scic,
-       struct scic_sds_remote_device *rdev,
-       struct scic_sds_request *request)
-{
-       enum sci_status status;
-       u16 index;
-
-       switch (scic->state_machine.current_state_id) {
-       case SCI_BASE_CONTROLLER_STATE_STOPPING:
-               /* XXX: Implement this function */
-               return SCI_FAILURE;
-       case SCI_BASE_CONTROLLER_STATE_READY:
-               status = scic_sds_remote_device_complete_io(scic, rdev, request);
-               if (status != SCI_SUCCESS)
-                       return status;
-
-               index = scic_sds_io_tag_get_index(request->io_tag);
-               scic->io_request_table[index] = NULL;
-               return SCI_SUCCESS;
-       default:
-               dev_warn(scic_to_dev(scic), "invalid state to complete I/O");
-               return SCI_FAILURE_INVALID_STATE;
-       }
-
-}
-
-enum sci_status scic_controller_continue_io(struct scic_sds_request *sci_req)
-{
-       struct scic_sds_controller *scic = sci_req->owning_controller;
-
-       if (scic->state_machine.current_state_id !=
-           SCI_BASE_CONTROLLER_STATE_READY) {
-               dev_warn(scic_to_dev(scic), "invalid state to continue I/O");
-               return SCI_FAILURE_INVALID_STATE;
-       }
-
-       scic->io_request_table[scic_sds_io_tag_get_index(sci_req->io_tag)] = sci_req;
-       scic_sds_controller_post_request(scic, scic_sds_request_get_post_context(sci_req));
-       return SCI_SUCCESS;
-}
-
-/**
- * scic_controller_start_task() - This method is called by the SCIC user to
- *    send/start a framework task management request.
- * @controller: the handle to the controller object for which to start the task
- *    management request.
- * @remote_device: the handle to the remote device object for which to start
- *    the task management request.
- * @task_request: the handle to the task request object to start.
- * @io_tag: This parameter specifies a previously allocated IO tag that the
- *    user desires to be utilized for this request.  Note this not the io_tag
- *    of the request being managed.  It is to be utilized for the task request
- *    itself. This parameter is optional.  The user is allowed to supply
- *    SCI_CONTROLLER_INVALID_IO_TAG as the value for this parameter.
- *
- * - IO tags are a protected resource.  It is incumbent upon the SCI Core user
- * to ensure that each of the methods that may allocate or free available IO
- * tags are handled in a mutually exclusive manner.  This method is one of said
- * methods requiring proper critical code section protection (e.g. semaphore,
- * spin-lock, etc.). - The user must synchronize this task with completion
- * queue processing.  If they are not synchronized then it is possible for the
- * io requests that are being managed by the task request can complete before
- * starting the task request. scic_controller_allocate_tag() for more
- * information on allocating a tag. Indicate if the controller successfully
- * started the IO request. SCI_TASK_SUCCESS if the task request was
- * successfully started. SCI_TASK_FAILURE_REQUIRES_SCSI_ABORT This value is
- * returned if there is/are task(s) outstanding that require termination or
- * completion before this request can succeed.
- */
-enum sci_task_status scic_controller_start_task(
-       struct scic_sds_controller *scic,
-       struct scic_sds_remote_device *rdev,
-       struct scic_sds_request *req,
-       u16 task_tag)
-{
-       enum sci_status status;
-
-       if (scic->state_machine.current_state_id !=
-           SCI_BASE_CONTROLLER_STATE_READY) {
-               dev_warn(scic_to_dev(scic),
-                        "%s: SCIC Controller starting task from invalid "
-                        "state\n",
-                        __func__);
-               return SCI_TASK_FAILURE_INVALID_STATE;
-       }
-
-       status = scic_sds_remote_device_start_task(scic, rdev, req);
-       switch (status) {
-       case SCI_FAILURE_RESET_DEVICE_PARTIAL_SUCCESS:
-               scic->io_request_table[scic_sds_io_tag_get_index(req->io_tag)] = req;
-
-               /*
-                * We will let framework know this task request started successfully,
-                * although core is still woring on starting the request (to post tc when
-                * RNC is resumed.)
-                */
-               return SCI_SUCCESS;
-       case SCI_SUCCESS:
-               scic->io_request_table[scic_sds_io_tag_get_index(req->io_tag)] = req;
-
-               scic_sds_controller_post_request(scic,
-                       scic_sds_request_get_post_context(req));
-               break;
-       default:
-               break;
-       }
-
-       return status;
-}
-
-/**
- * scic_controller_allocate_io_tag() - This method will allocate a tag from the
- *    pool of free IO tags. Direct allocation of IO tags by the SCI Core user
- *    is optional. The scic_controller_start_io() method will allocate an IO
- *    tag if this method is not utilized and the tag is not supplied to the IO
- *    construct routine.  Direct allocation of IO tags may provide additional
- *    performance improvements in environments capable of supporting this usage
- *    model.  Additionally, direct allocation of IO tags also provides
- *    additional flexibility to the SCI Core user.  Specifically, the user may
- *    retain IO tags across the lives of multiple IO requests.
- * @controller: the handle to the controller object for which to allocate the
- *    tag.
- *
- * IO tags are a protected resource.  It is incumbent upon the SCI Core user to
- * ensure that each of the methods that may allocate or free available IO tags
- * are handled in a mutually exclusive manner.  This method is one of said
- * methods requiring proper critical code section protection (e.g. semaphore,
- * spin-lock, etc.). An unsigned integer representing an available IO tag.
- * SCI_CONTROLLER_INVALID_IO_TAG This value is returned if there are no
- * currently available tags to be allocated. All return other values indicate a
- * legitimate tag.
- */
-u16 scic_controller_allocate_io_tag(
-       struct scic_sds_controller *scic)
-{
-       u16 task_context;
-       u16 sequence_count;
-
-       if (!sci_pool_empty(scic->tci_pool)) {
-               sci_pool_get(scic->tci_pool, task_context);
-
-               sequence_count = scic->io_request_sequence[task_context];
-
-               return scic_sds_io_tag_construct(sequence_count, task_context);
-       }
-
-       return SCI_CONTROLLER_INVALID_IO_TAG;
-}
-
-/**
- * scic_controller_free_io_tag() - This method will free an IO tag to the pool
- *    of free IO tags. This method provides the SCI Core user more flexibility
- *    with regards to IO tags.  The user may desire to keep an IO tag after an
- *    IO request has completed, because they plan on re-using the tag for a
- *    subsequent IO request.  This method is only legal if the tag was
- *    allocated via scic_controller_allocate_io_tag().
- * @controller: This parameter specifies the handle to the controller object
- *    for which to free/return the tag.
- * @io_tag: This parameter represents the tag to be freed to the pool of
- *    available tags.
- *
- * - IO tags are a protected resource.  It is incumbent upon the SCI Core user
- * to ensure that each of the methods that may allocate or free available IO
- * tags are handled in a mutually exclusive manner.  This method is one of said
- * methods requiring proper critical code section protection (e.g. semaphore,
- * spin-lock, etc.). - If the IO tag for a request was allocated, by the SCI
- * Core user, using the scic_controller_allocate_io_tag() method, then it is
- * the responsibility of the caller to invoke this method to free the tag. This
- * method returns an indication of whether the tag was successfully put back
- * (freed) to the pool of available tags. SCI_SUCCESS This return value
- * indicates the tag was successfully placed into the pool of available IO
- * tags. SCI_FAILURE_INVALID_IO_TAG This value is returned if the supplied tag
- * is not a valid IO tag value.
- */
-enum sci_status scic_controller_free_io_tag(
-       struct scic_sds_controller *scic,
-       u16 io_tag)
-{
-       u16 sequence;
-       u16 index;
-
-       BUG_ON(io_tag == SCI_CONTROLLER_INVALID_IO_TAG);
-
-       sequence = scic_sds_io_tag_get_sequence(io_tag);
-       index    = scic_sds_io_tag_get_index(io_tag);
-
-       if (!sci_pool_full(scic->tci_pool)) {
-               if (sequence == scic->io_request_sequence[index]) {
-                       scic_sds_io_sequence_increment(
-                               scic->io_request_sequence[index]);
-
-                       sci_pool_put(scic->tci_pool, index);
-
-                       return SCI_SUCCESS;
-               }
-       }
-
-       return SCI_FAILURE_INVALID_IO_TAG;
-}
-
-void scic_controller_enable_interrupts(
-       struct scic_sds_controller *scic)
-{
-       BUG_ON(scic->smu_registers == NULL);
-       writel(0, &scic->smu_registers->interrupt_mask);
-}
-
-void scic_controller_disable_interrupts(
-       struct scic_sds_controller *scic)
-{
-       BUG_ON(scic->smu_registers == NULL);
-       writel(0xffffffff, &scic->smu_registers->interrupt_mask);
-}
-
-static enum sci_status scic_controller_set_mode(
-       struct scic_sds_controller *scic,
-       enum sci_controller_mode operating_mode)
-{
-       enum sci_status status          = SCI_SUCCESS;
-
-       if ((scic->state_machine.current_state_id ==
-                               SCI_BASE_CONTROLLER_STATE_INITIALIZING) ||
-           (scic->state_machine.current_state_id ==
-                               SCI_BASE_CONTROLLER_STATE_INITIALIZED)) {
-               switch (operating_mode) {
-               case SCI_MODE_SPEED:
-                       scic->remote_node_entries      = SCI_MAX_REMOTE_DEVICES;
-                       scic->task_context_entries     = SCU_IO_REQUEST_COUNT;
-                       scic->uf_control.buffers.count =
-                               SCU_UNSOLICITED_FRAME_COUNT;
-                       scic->completion_event_entries = SCU_EVENT_COUNT;
-                       scic->completion_queue_entries =
-                               SCU_COMPLETION_QUEUE_COUNT;
-                       break;
-
-               case SCI_MODE_SIZE:
-                       scic->remote_node_entries      = SCI_MIN_REMOTE_DEVICES;
-                       scic->task_context_entries     = SCI_MIN_IO_REQUESTS;
-                       scic->uf_control.buffers.count =
-                               SCU_MIN_UNSOLICITED_FRAMES;
-                       scic->completion_event_entries = SCU_MIN_EVENTS;
-                       scic->completion_queue_entries =
-                               SCU_MIN_COMPLETION_QUEUE_ENTRIES;
-                       break;
-
-               default:
-                       status = SCI_FAILURE_INVALID_PARAMETER_VALUE;
-                       break;
-               }
-       } else
-               status = SCI_FAILURE_INVALID_STATE;
-
-       return status;
-}
-
-/**
- * scic_sds_controller_reset_hardware() -
- *
- * This method will reset the controller hardware.
- */
-static void scic_sds_controller_reset_hardware(
-       struct scic_sds_controller *scic)
-{
-       /* Disable interrupts so we dont take any spurious interrupts */
-       scic_controller_disable_interrupts(scic);
-
-       /* Reset the SCU */
-       writel(0xFFFFFFFF, &scic->smu_registers->soft_reset_control);
-
-       /* Delay for 1ms to before clearing the CQP and UFQPR. */
-       udelay(1000);
-
-       /* The write to the CQGR clears the CQP */
-       writel(0x00000000, &scic->smu_registers->completion_queue_get);
-
-       /* The write to the UFQGP clears the UFQPR */
-       writel(0, &scic->scu_registers->sdma.unsolicited_frame_get_pointer);
-}
-
-enum sci_status scic_user_parameters_set(
-       struct scic_sds_controller *scic,
-       union scic_user_parameters *scic_parms)
-{
-       u32 state = scic->state_machine.current_state_id;
-
-       if (state == SCI_BASE_CONTROLLER_STATE_RESET ||
-           state == SCI_BASE_CONTROLLER_STATE_INITIALIZING ||
-           state == SCI_BASE_CONTROLLER_STATE_INITIALIZED) {
-               u16 index;
-
-               /*
-                * Validate the user parameters.  If they are not legal, then
-                * return a failure.
-                */
-               for (index = 0; index < SCI_MAX_PHYS; index++) {
-                       struct sci_phy_user_params *user_phy;
-
-                       user_phy = &scic_parms->sds1.phys[index];
-
-                       if (!((user_phy->max_speed_generation <=
-                                               SCIC_SDS_PARM_MAX_SPEED) &&
-                             (user_phy->max_speed_generation >
-                                               SCIC_SDS_PARM_NO_SPEED)))
-                               return SCI_FAILURE_INVALID_PARAMETER_VALUE;
-
-                       if (user_phy->in_connection_align_insertion_frequency <
-                                       3)
-                               return SCI_FAILURE_INVALID_PARAMETER_VALUE;
-
-                       if ((user_phy->in_connection_align_insertion_frequency <
-                                               3) ||
-                           (user_phy->align_insertion_frequency == 0) ||
-                           (user_phy->
-                               notify_enable_spin_up_insertion_frequency ==
-                                               0))
-                               return SCI_FAILURE_INVALID_PARAMETER_VALUE;
-               }
-
-               if ((scic_parms->sds1.stp_inactivity_timeout == 0) ||
-                   (scic_parms->sds1.ssp_inactivity_timeout == 0) ||
-                   (scic_parms->sds1.stp_max_occupancy_timeout == 0) ||
-                   (scic_parms->sds1.ssp_max_occupancy_timeout == 0) ||
-                   (scic_parms->sds1.no_outbound_task_timeout == 0))
-                       return SCI_FAILURE_INVALID_PARAMETER_VALUE;
-
-               memcpy(&scic->user_parameters, scic_parms, sizeof(*scic_parms));
-
-               return SCI_SUCCESS;
-       }
-
-       return SCI_FAILURE_INVALID_STATE;
-}
-
-int scic_oem_parameters_validate(struct scic_sds_oem_params *oem)
-{
-       int i;
-
-       for (i = 0; i < SCI_MAX_PORTS; i++)
-               if (oem->ports[i].phy_mask > SCIC_SDS_PARM_PHY_MASK_MAX)
-                       return -EINVAL;
-
-       for (i = 0; i < SCI_MAX_PHYS; i++)
-               if (oem->phys[i].sas_address.high == 0 &&
-                   oem->phys[i].sas_address.low == 0)
-                       return -EINVAL;
-
-       if (oem->controller.mode_type == SCIC_PORT_AUTOMATIC_CONFIGURATION_MODE) {
-               for (i = 0; i < SCI_MAX_PHYS; i++)
-                       if (oem->ports[i].phy_mask != 0)
-                               return -EINVAL;
-       } else if (oem->controller.mode_type == SCIC_PORT_MANUAL_CONFIGURATION_MODE) {
-               u8 phy_mask = 0;
-
-               for (i = 0; i < SCI_MAX_PHYS; i++)
-                       phy_mask |= oem->ports[i].phy_mask;
-
-               if (phy_mask == 0)
-                       return -EINVAL;
-       } else
-               return -EINVAL;
-
-       if (oem->controller.max_concurrent_dev_spin_up > MAX_CONCURRENT_DEVICE_SPIN_UP_COUNT)
-               return -EINVAL;
-
-       return 0;
-}
-
-enum sci_status scic_oem_parameters_set(struct scic_sds_controller *scic,
-                                       union scic_oem_parameters *scic_parms)
-{
-       u32 state = scic->state_machine.current_state_id;
-
-       if (state == SCI_BASE_CONTROLLER_STATE_RESET ||
-           state == SCI_BASE_CONTROLLER_STATE_INITIALIZING ||
-           state == SCI_BASE_CONTROLLER_STATE_INITIALIZED) {
-
-               if (scic_oem_parameters_validate(&scic_parms->sds1))
-                       return SCI_FAILURE_INVALID_PARAMETER_VALUE;
-               scic->oem_parameters.sds1 = scic_parms->sds1;
-
-               return SCI_SUCCESS;
-       }
-
-       return SCI_FAILURE_INVALID_STATE;
-}
-
-void scic_oem_parameters_get(
-       struct scic_sds_controller *scic,
-       union scic_oem_parameters *scic_parms)
-{
-       memcpy(scic_parms, (&scic->oem_parameters), sizeof(*scic_parms));
-}
-
-#define INTERRUPT_COALESCE_TIMEOUT_BASE_RANGE_LOWER_BOUND_NS 853
-#define INTERRUPT_COALESCE_TIMEOUT_BASE_RANGE_UPPER_BOUND_NS 1280
-#define INTERRUPT_COALESCE_TIMEOUT_MAX_US                    2700000
-#define INTERRUPT_COALESCE_NUMBER_MAX                        256
-#define INTERRUPT_COALESCE_TIMEOUT_ENCODE_MIN                7
-#define INTERRUPT_COALESCE_TIMEOUT_ENCODE_MAX                28
-
-/**
- * scic_controller_set_interrupt_coalescence() - This method allows the user to
- *    configure the interrupt coalescence.
- * @controller: This parameter represents the handle to the controller object
- *    for which its interrupt coalesce register is overridden.
- * @coalesce_number: Used to control the number of entries in the Completion
- *    Queue before an interrupt is generated. If the number of entries exceed
- *    this number, an interrupt will be generated. The valid range of the input
- *    is [0, 256]. A setting of 0 results in coalescing being disabled.
- * @coalesce_timeout: Timeout value in microseconds. The valid range of the
- *    input is [0, 2700000] . A setting of 0 is allowed and results in no
- *    interrupt coalescing timeout.
- *
- * Indicate if the user successfully set the interrupt coalesce parameters.
- * SCI_SUCCESS The user successfully updated the interrutp coalescence.
- * SCI_FAILURE_INVALID_PARAMETER_VALUE The user input value is out of range.
- */
-static enum sci_status scic_controller_set_interrupt_coalescence(
-       struct scic_sds_controller *scic_controller,
-       u32 coalesce_number,
-       u32 coalesce_timeout)
-{
-       u8 timeout_encode = 0;
-       u32 min = 0;
-       u32 max = 0;
-
-       /* Check if the input parameters fall in the range. */
-       if (coalesce_number > INTERRUPT_COALESCE_NUMBER_MAX)
-               return SCI_FAILURE_INVALID_PARAMETER_VALUE;
-
-       /*
-        *  Defined encoding for interrupt coalescing timeout:
-        *              Value   Min      Max     Units
-        *              -----   ---      ---     -----
-        *              0       -        -       Disabled
-        *              1       13.3     20.0    ns
-        *              2       26.7     40.0
-        *              3       53.3     80.0
-        *              4       106.7    160.0
-        *              5       213.3    320.0
-        *              6       426.7    640.0
-        *              7       853.3    1280.0
-        *              8       1.7      2.6     us
-        *              9       3.4      5.1
-        *              10      6.8      10.2
-        *              11      13.7     20.5
-        *              12      27.3     41.0
-        *              13      54.6     81.9
-        *              14      109.2    163.8
-        *              15      218.5    327.7
-        *              16      436.9    655.4
-        *              17      873.8    1310.7
-        *              18      1.7      2.6     ms
-        *              19      3.5      5.2
-        *              20      7.0      10.5
-        *              21      14.0     21.0
-        *              22      28.0     41.9
-        *              23      55.9     83.9
-        *              24      111.8    167.8
-        *              25      223.7    335.5
-        *              26      447.4    671.1
-        *              27      894.8    1342.2
-        *              28      1.8      2.7     s
-        *              Others Undefined */
-
-       /*
-        * Use the table above to decide the encode of interrupt coalescing timeout
-        * value for register writing. */
-       if (coalesce_timeout == 0)
-               timeout_encode = 0;
-       else{
-               /* make the timeout value in unit of (10 ns). */
-               coalesce_timeout = coalesce_timeout * 100;
-               min = INTERRUPT_COALESCE_TIMEOUT_BASE_RANGE_LOWER_BOUND_NS / 10;
-               max = INTERRUPT_COALESCE_TIMEOUT_BASE_RANGE_UPPER_BOUND_NS / 10;
-
-               /* get the encode of timeout for register writing. */
-               for (timeout_encode = INTERRUPT_COALESCE_TIMEOUT_ENCODE_MIN;
-                     timeout_encode <= INTERRUPT_COALESCE_TIMEOUT_ENCODE_MAX;
-                     timeout_encode++) {
-                       if (min <= coalesce_timeout &&  max > coalesce_timeout)
-                               break;
-                       else if (coalesce_timeout >= max && coalesce_timeout < min * 2
-                                && coalesce_timeout <= INTERRUPT_COALESCE_TIMEOUT_MAX_US * 100) {
-                               if ((coalesce_timeout - max) < (2 * min - coalesce_timeout))
-                                       break;
-                               else{
-                                       timeout_encode++;
-                                       break;
-                               }
-                       } else {
-                               max = max * 2;
-                               min = min * 2;
-                       }
-               }
-
-               if (timeout_encode == INTERRUPT_COALESCE_TIMEOUT_ENCODE_MAX + 1)
-                       /* the value is out of range. */
-                       return SCI_FAILURE_INVALID_PARAMETER_VALUE;
-       }
-
-       writel(SMU_ICC_GEN_VAL(NUMBER, coalesce_number) |
-              SMU_ICC_GEN_VAL(TIMER, timeout_encode),
-              &scic_controller->smu_registers->interrupt_coalesce_control);
-
-
-       scic_controller->interrupt_coalesce_number = (u16)coalesce_number;
-       scic_controller->interrupt_coalesce_timeout = coalesce_timeout / 100;
-
-       return SCI_SUCCESS;
-}
-
-
-
-enum sci_status scic_controller_initialize(struct scic_sds_controller *scic)
-{
-       struct sci_base_state_machine *sm = &scic->state_machine;
-       enum sci_status result = SCI_SUCCESS;
-       struct isci_host *ihost = scic_to_ihost(scic);
-       u32 index, state;
-
-       if (scic->state_machine.current_state_id !=
-           SCI_BASE_CONTROLLER_STATE_RESET) {
-               dev_warn(scic_to_dev(scic),
-                        "SCIC Controller initialize operation requested "
-                        "in invalid state\n");
-               return SCI_FAILURE_INVALID_STATE;
-       }
-
-       sci_base_state_machine_change_state(sm, SCI_BASE_CONTROLLER_STATE_INITIALIZING);
-
-       scic->timeout_timer = isci_timer_create(ihost,
-                                               scic,
-                                               scic_sds_controller_timeout_handler);
-
-       scic_sds_controller_initialize_phy_startup(scic);
-
-       scic_sds_controller_initialize_power_control(scic);
-
-       /*
-        * There is nothing to do here for B0 since we do not have to
-        * program the AFE registers.
-        * / @todo The AFE settings are supposed to be correct for the B0 but
-        * /       presently they seem to be wrong. */
-       scic_sds_controller_afe_initialization(scic);
-
-       if (result == SCI_SUCCESS) {
-               u32 status;
-               u32 terminate_loop;
-
-               /* Take the hardware out of reset */
-               writel(0, &scic->smu_registers->soft_reset_control);
-
-               /*
-                * / @todo Provide meaningfull error code for hardware failure
-                * result = SCI_FAILURE_CONTROLLER_HARDWARE; */
-               result = SCI_FAILURE;
-               terminate_loop = 100;
-
-               while (terminate_loop-- && (result != SCI_SUCCESS)) {
-                       /* Loop until the hardware reports success */
-                       udelay(SCU_CONTEXT_RAM_INIT_STALL_TIME);
-                       status = readl(&scic->smu_registers->control_status);
-
-                       if ((status & SCU_RAM_INIT_COMPLETED) ==
-                                       SCU_RAM_INIT_COMPLETED)
-                               result = SCI_SUCCESS;
-               }
-       }
-
-       if (result == SCI_SUCCESS) {
-               u32 max_supported_ports;
-               u32 max_supported_devices;
-               u32 max_supported_io_requests;
-               u32 device_context_capacity;
-
-               /*
-                * Determine what are the actaul device capacities that the
-                * hardware will support */
-               device_context_capacity =
-                       readl(&scic->smu_registers->device_context_capacity);
-
-
-               max_supported_ports = smu_dcc_get_max_ports(device_context_capacity);
-               max_supported_devices = smu_dcc_get_max_remote_node_context(device_context_capacity);
-               max_supported_io_requests = smu_dcc_get_max_task_context(device_context_capacity);
-
-               /*
-                * Make all PEs that are unassigned match up with the
-                * logical ports
-                */
-               for (index = 0; index < max_supported_ports; index++) {
-                       struct scu_port_task_scheduler_group_registers __iomem
-                               *ptsg = &scic->scu_registers->peg0.ptsg;
-
-                       writel(index, &ptsg->protocol_engine[index]);
-               }
-
-               /* Record the smaller of the two capacity values */
-               scic->logical_port_entries =
-                       min(max_supported_ports, scic->logical_port_entries);
-
-               scic->task_context_entries =
-                       min(max_supported_io_requests,
-                           scic->task_context_entries);
-
-               scic->remote_node_entries =
-                       min(max_supported_devices, scic->remote_node_entries);
-
-               /*
-                * Now that we have the correct hardware reported minimum values
-                * build the MDL for the controller.  Default to a performance
-                * configuration.
-                */
-               scic_controller_set_mode(scic, SCI_MODE_SPEED);
-       }
-
-       /* Initialize hardware PCI Relaxed ordering in DMA engines */
-       if (result == SCI_SUCCESS) {
-               u32 dma_configuration;
-
-               /* Configure the payload DMA */
-               dma_configuration =
-                       readl(&scic->scu_registers->sdma.pdma_configuration);
-               dma_configuration |=
-                       SCU_PDMACR_GEN_BIT(PCI_RELAXED_ORDERING_ENABLE);
-               writel(dma_configuration,
-                       &scic->scu_registers->sdma.pdma_configuration);
-
-               /* Configure the control DMA */
-               dma_configuration =
-                       readl(&scic->scu_registers->sdma.cdma_configuration);
-               dma_configuration |=
-                       SCU_CDMACR_GEN_BIT(PCI_RELAXED_ORDERING_ENABLE);
-               writel(dma_configuration,
-                       &scic->scu_registers->sdma.cdma_configuration);
-       }
-
-       /*
-        * Initialize the PHYs before the PORTs because the PHY registers
-        * are accessed during the port initialization.
-        */
-       if (result == SCI_SUCCESS) {
-               /* Initialize the phys */
-               for (index = 0;
-                    (result == SCI_SUCCESS) && (index < SCI_MAX_PHYS);
-                    index++) {
-                       result = scic_sds_phy_initialize(
-                               &ihost->phys[index].sci,
-                               &scic->scu_registers->peg0.pe[index].tl,
-                               &scic->scu_registers->peg0.pe[index].ll);
-               }
-       }
-
-       if (result == SCI_SUCCESS) {
-               /* Initialize the logical ports */
-               for (index = 0;
-                    (index < scic->logical_port_entries) &&
-                    (result == SCI_SUCCESS);
-                    index++) {
-                       result = scic_sds_port_initialize(
-                               &ihost->ports[index].sci,
-                               &scic->scu_registers->peg0.ptsg.port[index],
-                               &scic->scu_registers->peg0.ptsg.protocol_engine,
-                               &scic->scu_registers->peg0.viit[index]);
-               }
-       }
-
-       if (result == SCI_SUCCESS)
-               result = scic_sds_port_configuration_agent_initialize(
-                               scic,
-                               &scic->port_agent);
-
-       /* Advance the controller state machine */
-       if (result == SCI_SUCCESS)
-               state = SCI_BASE_CONTROLLER_STATE_INITIALIZED;
-       else
-               state = SCI_BASE_CONTROLLER_STATE_FAILED;
-       sci_base_state_machine_change_state(sm, state);
-
-       return result;
-}
-
-enum sci_status scic_controller_start(struct scic_sds_controller *scic,
-                                     u32 timeout)
-{
-       struct isci_host *ihost = scic_to_ihost(scic);
-       enum sci_status result;
-       u16 index;
-
-       if (scic->state_machine.current_state_id !=
-           SCI_BASE_CONTROLLER_STATE_INITIALIZED) {
-               dev_warn(scic_to_dev(scic),
-                        "SCIC Controller start operation requested in "
-                        "invalid state\n");
-               return SCI_FAILURE_INVALID_STATE;
-       }
-
-       /* Build the TCi free pool */
-       sci_pool_initialize(scic->tci_pool);
-       for (index = 0; index < scic->task_context_entries; index++)
-               sci_pool_put(scic->tci_pool, index);
-
-       /* Build the RNi free pool */
-       scic_sds_remote_node_table_initialize(
-                       &scic->available_remote_nodes,
-                       scic->remote_node_entries);
-
-       /*
-        * Before anything else lets make sure we will not be
-        * interrupted by the hardware.
-        */
-       scic_controller_disable_interrupts(scic);
-
-       /* Enable the port task scheduler */
-       scic_sds_controller_enable_port_task_scheduler(scic);
-
-       /* Assign all the task entries to scic physical function */
-       scic_sds_controller_assign_task_entries(scic);
-
-       /* Now initialize the completion queue */
-       scic_sds_controller_initialize_completion_queue(scic);
-
-       /* Initialize the unsolicited frame queue for use */
-       scic_sds_controller_initialize_unsolicited_frame_queue(scic);
-
-       /* Start all of the ports on this controller */
-       for (index = 0; index < scic->logical_port_entries; index++) {
-               struct scic_sds_port *sci_port = &ihost->ports[index].sci;
-
-               result = sci_port->state_handlers->start_handler(sci_port);
-               if (result)
-                       return result;
-       }
-
-       scic_sds_controller_start_next_phy(scic);
-
-       isci_timer_start(scic->timeout_timer, timeout);
-
-       sci_base_state_machine_change_state(&scic->state_machine,
-                                           SCI_BASE_CONTROLLER_STATE_STARTING);
-
-       return SCI_SUCCESS;
-}
-
-/**
- *
- * @object: This is the object which is cast to a struct scic_sds_controller
- *    object.
- *
- * This method implements the actions taken by the struct scic_sds_controller on entry
- * to the SCI_BASE_CONTROLLER_STATE_INITIAL. - Set the state handlers to the
- * controllers initial state. none This function should initialize the
- * controller object.
- */
-static void scic_sds_controller_initial_state_enter(void *object)
-{
-       struct scic_sds_controller *scic = object;
-
-       sci_base_state_machine_change_state(&scic->state_machine,
-                       SCI_BASE_CONTROLLER_STATE_RESET);
-}
-
-/**
- *
- * @object: This is the object which is cast to a struct scic_sds_controller
- *    object.
- *
- * This method implements the actions taken by the struct scic_sds_controller on exit
- * from the SCI_BASE_CONTROLLER_STATE_STARTING. - This function stops the
- * controller starting timeout timer. none
- */
-static inline void scic_sds_controller_starting_state_exit(void *object)
-{
-       struct scic_sds_controller *scic = object;
-
-       isci_timer_stop(scic->timeout_timer);
-}
-
-/**
- *
- * @object: This is the object which is cast to a struct scic_sds_controller
- *    object.
- *
- * This method implements the actions taken by the struct scic_sds_controller on entry
- * to the SCI_BASE_CONTROLLER_STATE_READY. - Set the state handlers to the
- * controllers ready state. none
- */
-static void scic_sds_controller_ready_state_enter(void *object)
-{
-       struct scic_sds_controller *scic = object;
-
-       /* set the default interrupt coalescence number and timeout value. */
-       scic_controller_set_interrupt_coalescence(
-               scic, 0x10, 250);
-}
-
-/**
- *
- * @object: This is the object which is cast to a struct scic_sds_controller
- *    object.
- *
- * This method implements the actions taken by the struct scic_sds_controller on exit
- * from the SCI_BASE_CONTROLLER_STATE_READY. - This function does nothing. none
- */
-static void scic_sds_controller_ready_state_exit(void *object)
-{
-       struct scic_sds_controller *scic = object;
-
-       /* disable interrupt coalescence. */
-       scic_controller_set_interrupt_coalescence(scic, 0, 0);
-}
-
-/**
- *
- * @object: This is the object which is cast to a struct scic_sds_controller
- *    object.
- *
- * This method implements the actions taken by the struct scic_sds_controller on entry
- * to the SCI_BASE_CONTROLLER_STATE_READY. - Set the state handlers to the
- * controllers ready state. - Stop the phys on this controller - Stop the ports
- * on this controller - Stop all of the remote devices on this controller none
- */
-static void scic_sds_controller_stopping_state_enter(void *object)
-{
-       struct scic_sds_controller *scic = object;
-
-       /* Stop all of the components for this controller */
-       scic_sds_controller_stop_phys(scic);
-       scic_sds_controller_stop_ports(scic);
-       scic_sds_controller_stop_devices(scic);
-}
-
-/**
- *
- * @object: This is the object which is cast to a struct
- * scic_sds_controller object.
- *
- * This function implements the actions taken by the struct scic_sds_controller
- * on exit from the SCI_BASE_CONTROLLER_STATE_STOPPING. -
- * This function stops the controller stopping timeout timer.
- */
-static inline void scic_sds_controller_stopping_state_exit(void *object)
-{
-       struct scic_sds_controller *scic = object;
-
-       isci_timer_stop(scic->timeout_timer);
-}
-
-static void scic_sds_controller_resetting_state_enter(void *object)
-{
-       struct scic_sds_controller *scic = object;
-
-       scic_sds_controller_reset_hardware(scic);
-       sci_base_state_machine_change_state(&scic->state_machine,
-                                           SCI_BASE_CONTROLLER_STATE_RESET);
-}
-
-static const struct sci_base_state scic_sds_controller_state_table[] = {
-       [SCI_BASE_CONTROLLER_STATE_INITIAL] = {
-               .enter_state = scic_sds_controller_initial_state_enter,
-       },
-       [SCI_BASE_CONTROLLER_STATE_RESET] = {},
-       [SCI_BASE_CONTROLLER_STATE_INITIALIZING] = {},
-       [SCI_BASE_CONTROLLER_STATE_INITIALIZED] = {},
-       [SCI_BASE_CONTROLLER_STATE_STARTING] = {
-               .exit_state  = scic_sds_controller_starting_state_exit,
-       },
-       [SCI_BASE_CONTROLLER_STATE_READY] = {
-               .enter_state = scic_sds_controller_ready_state_enter,
-               .exit_state  = scic_sds_controller_ready_state_exit,
-       },
-       [SCI_BASE_CONTROLLER_STATE_RESETTING] = {
-               .enter_state = scic_sds_controller_resetting_state_enter,
-       },
-       [SCI_BASE_CONTROLLER_STATE_STOPPING] = {
-               .enter_state = scic_sds_controller_stopping_state_enter,
-               .exit_state = scic_sds_controller_stopping_state_exit,
-       },
-       [SCI_BASE_CONTROLLER_STATE_STOPPED] = {},
-       [SCI_BASE_CONTROLLER_STATE_FAILED] = {}
-};
-
-/**
- * scic_controller_construct() - This method will attempt to construct a
- *    controller object utilizing the supplied parameter information.
- * @c: This parameter specifies the controller to be constructed.
- * @scu_base: mapped base address of the scu registers
- * @smu_base: mapped base address of the smu registers
- *
- * Indicate if the controller was successfully constructed or if it failed in
- * some way. SCI_SUCCESS This value is returned if the controller was
- * successfully constructed. SCI_WARNING_TIMER_CONFLICT This value is returned
- * if the interrupt coalescence timer may cause SAS compliance issues for SMP
- * Target mode response processing. SCI_FAILURE_UNSUPPORTED_CONTROLLER_TYPE
- * This value is returned if the controller does not support the supplied type.
- * SCI_FAILURE_UNSUPPORTED_INIT_DATA_VERSION This value is returned if the
- * controller does not support the supplied initialization data version.
- */
-enum sci_status scic_controller_construct(struct scic_sds_controller *scic,
-                                         void __iomem *scu_base,
-                                         void __iomem *smu_base)
-{
-       struct isci_host *ihost = scic_to_ihost(scic);
-       u8 i;
-
-       sci_base_state_machine_construct(&scic->state_machine,
-               scic, scic_sds_controller_state_table,
-               SCI_BASE_CONTROLLER_STATE_INITIAL);
-
-       sci_base_state_machine_start(&scic->state_machine);
-
-       scic->scu_registers = scu_base;
-       scic->smu_registers = smu_base;
-
-       scic_sds_port_configuration_agent_construct(&scic->port_agent);
-
-       /* Construct the ports for this controller */
-       for (i = 0; i < SCI_MAX_PORTS; i++)
-               scic_sds_port_construct(&ihost->ports[i].sci, i, scic);
-       scic_sds_port_construct(&ihost->ports[i].sci, SCIC_SDS_DUMMY_PORT, scic);
-
-       /* Construct the phys for this controller */
-       for (i = 0; i < SCI_MAX_PHYS; i++) {
-               /* Add all the PHYs to the dummy port */
-               scic_sds_phy_construct(&ihost->phys[i].sci,
-                                      &ihost->ports[SCI_MAX_PORTS].sci, i);
-       }
-
-       scic->invalid_phy_mask = 0;
-
-       /* Set the default maximum values */
-       scic->completion_event_entries      = SCU_EVENT_COUNT;
-       scic->completion_queue_entries      = SCU_COMPLETION_QUEUE_COUNT;
-       scic->remote_node_entries           = SCI_MAX_REMOTE_DEVICES;
-       scic->logical_port_entries          = SCI_MAX_PORTS;
-       scic->task_context_entries          = SCU_IO_REQUEST_COUNT;
-       scic->uf_control.buffers.count      = SCU_UNSOLICITED_FRAME_COUNT;
-       scic->uf_control.address_table.count = SCU_UNSOLICITED_FRAME_COUNT;
-
-       /* Initialize the User and OEM parameters to default values. */
-       scic_sds_controller_set_default_config_parameters(scic);
-
-       return scic_controller_reset(scic);
-}
diff --git a/drivers/scsi/isci/core/scic_sds_controller.h b/drivers/scsi/isci/core/scic_sds_controller.h
deleted file mode 100644 (file)
index 5c00f96..0000000
+++ /dev/null
@@ -1,576 +0,0 @@
-/*
- * This file is provided under a dual BSD/GPLv2 license.  When using or
- * redistributing this file, you may do so under either license.
- *
- * GPL LICENSE SUMMARY
- *
- * Copyright(c) 2008 - 2011 Intel Corporation. All rights reserved.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
- * The full GNU General Public License is included in this distribution
- * in the file called LICENSE.GPL.
- *
- * BSD LICENSE
- *
- * Copyright(c) 2008 - 2011 Intel Corporation. All rights reserved.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- *   * Redistributions of source code must retain the above copyright
- *     notice, this list of conditions and the following disclaimer.
- *   * Redistributions in binary form must reproduce the above copyright
- *     notice, this list of conditions and the following disclaimer in
- *     the documentation and/or other materials provided with the
- *     distribution.
- *   * Neither the name of Intel Corporation nor the names of its
- *     contributors may be used to endorse or promote products derived
- *     from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-#ifndef _SCIC_SDS_CONTROLLER_H_
-#define _SCIC_SDS_CONTROLLER_H_
-
-#include <linux/string.h>
-#include <linux/io.h>
-
-/**
- * This file contains the structures, constants and prototypes used for the
- *    core controller object.
- *
- *
- */
-
-#include "sci_pool.h"
-#include "sci_base_state.h"
-#include "sci_base_state_machine.h"
-#include "scic_config_parameters.h"
-#include "scic_sds_port.h"
-#include "scic_sds_phy.h"
-#include "remote_node_table.h"
-#include "remote_device.h"
-#include "scu_registers.h"
-#include "scu_task_context.h"
-#include "scu_unsolicited_frame.h"
-#include "scic_sds_unsolicited_frame_control.h"
-#include "scic_sds_port_configuration_agent.h"
-
-struct sci_base_remote_device;
-struct scic_sds_remote_device;
-struct scic_sds_request;
-struct scic_sds_controller;
-
-/**
- * struct scic_power_control -
- *
- * This structure defines the fields for managing power control for direct
- * attached disk devices.
- */
-struct scic_power_control {
-       /**
-        * This field is set when the power control timer is running and cleared when
-        * it is not.
-        */
-       bool timer_started;
-
-       /**
-        * This field is the handle to the driver timer object.  This timer is used to
-        * control when the directed attached disks can consume power.
-        */
-       void *timer;
-
-       /**
-        * This field is used to keep track of how many phys are put into the
-        * requesters field.
-        */
-       u8 phys_waiting;
-
-       /**
-        * This field is used to keep track of how many phys have been granted to consume power
-        */
-       u8 phys_granted_power;
-
-       /**
-        * This field is an array of phys that we are waiting on. The phys are direct
-        * mapped into requesters via struct scic_sds_phy.phy_index
-        */
-       struct scic_sds_phy *requesters[SCI_MAX_PHYS];
-
-};
-
-/**
- * struct scic_sds_controller -
- *
- * This structure represents the SCU controller object.
- */
-struct scic_sds_controller {
-       /**
-        * This field contains the information for the base controller state
-        * machine.
-        */
-       struct sci_base_state_machine state_machine;
-
-       /**
-        * This field is the driver timer object handler used to time the controller
-        * object start and stop requests.
-        */
-       void *timeout_timer;
-
-       /**
-        * This field contains the user parameters to be utilized for this
-        * core controller object.
-        */
-       union scic_user_parameters user_parameters;
-
-       /**
-        * This field contains the OEM parameters to be utilized for this
-        * core controller object.
-        */
-       union scic_oem_parameters oem_parameters;
-
-       /**
-        * This field contains the port configuration agent for this controller.
-        */
-       struct scic_sds_port_configuration_agent port_agent;
-
-       /**
-        * This field is the array of device objects that are currently constructed
-        * for this controller object.  This table is used as a fast lookup of device
-        * objects that need to handle device completion notifications from the
-        * hardware. The table is RNi based.
-        */
-       struct scic_sds_remote_device *device_table[SCI_MAX_REMOTE_DEVICES];
-
-       /**
-        * This field is the array of IO request objects that are currently active for
-        * this controller object.  This table is used as a fast lookup of the io
-        * request object that need to handle completion queue notifications.  The
-        * table is TCi based.
-        */
-       struct scic_sds_request *io_request_table[SCI_MAX_IO_REQUESTS];
-
-       /**
-        * This field is the free RNi data structure
-        */
-       struct scic_remote_node_table available_remote_nodes;
-
-       /**
-        * This field is the TCi pool used to manage the task context index.
-        */
-       SCI_POOL_CREATE(tci_pool, u16, SCI_MAX_IO_REQUESTS);
-
-       /**
-        * This filed is the struct scic_power_control data used to controll when direct
-        * attached devices can consume power.
-        */
-       struct scic_power_control power_control;
-
-       /**
-        * This field is the array of sequence values for the IO Tag fields.  Even
-        * though only 4 bits of the field is used for the sequence the sequence is 16
-        * bits in size so the sequence can be bitwise or'd with the TCi to build the
-        * IO Tag value.
-        */
-       u16 io_request_sequence[SCI_MAX_IO_REQUESTS];
-
-       /**
-        * This field in the array of sequence values for the RNi.  These are used
-        * to control io request build to io request start operations.  The sequence
-        * value is recorded into an io request when it is built and is checked on
-        * the io request start operation to make sure that there was not a device
-        * hot plug between the build and start operation.
-        */
-       u8 remote_device_sequence[SCI_MAX_REMOTE_DEVICES];
-
-       /**
-        * This field is a pointer to the memory allocated by the driver for the task
-        * context table.  This data is shared between the hardware and software.
-        */
-       struct scu_task_context *task_context_table;
-
-       /**
-        * This field is a pointer to the memory allocated by the driver for the
-        * remote node context table.  This table is shared between the hardware and
-        * software.
-        */
-       union scu_remote_node_context *remote_node_context_table;
-
-       /**
-        * This field is a pointer to the completion queue.  This memory is
-        * written to by the hardware and read by the software.
-        */
-       u32 *completion_queue;
-
-       /**
-        * This field is the software copy of the completion queue get pointer.  The
-        * controller object writes this value to the hardware after processing the
-        * completion entries.
-        */
-       u32 completion_queue_get;
-
-       /**
-        * This field is the minimum of the number of hardware supported port entries
-        * and the software requested port entries.
-        */
-       u32 logical_port_entries;
-
-       /**
-        * This field is the minimum number of hardware supported completion queue
-        * entries and the software requested completion queue entries.
-        */
-       u32 completion_queue_entries;
-
-       /**
-        * This field is the minimum number of hardware supported event entries and
-        * the software requested event entries.
-        */
-       u32 completion_event_entries;
-
-       /**
-        * This field is the minimum number of devices supported by the hardware and
-        * the number of devices requested by the software.
-        */
-       u32 remote_node_entries;
-
-       /**
-        * This field is the minimum number of IO requests supported by the hardware
-        * and the number of IO requests requested by the software.
-        */
-       u32 task_context_entries;
-
-       /**
-        * This object contains all of the unsolicited frame specific
-        * data utilized by the core controller.
-        */
-       struct scic_sds_unsolicited_frame_control uf_control;
-
-       /* Phy Startup Data */
-       /**
-        * This field is the driver timer handle for controller phy request startup.
-        * On controller start the controller will start each PHY individually in
-        * order of phy index.
-        */
-       void *phy_startup_timer;
-
-       /**
-        * This field is set when the phy_startup_timer is running and is cleared when
-        * the phy_startup_timer is stopped.
-        */
-       bool phy_startup_timer_pending;
-
-       /**
-        * This field is the index of the next phy start.  It is initialized to 0 and
-        * increments for each phy index that is started.
-        */
-       u32 next_phy_to_start;
-
-       /**
-        * This field controlls the invalid link up notifications to the SCI_USER.  If
-        * an invalid_link_up notification is reported a bit for the PHY index is set
-        * so further notifications are not made.  Once the PHY object reports link up
-        * and is made part of a port then this bit for the PHY index is cleared.
-        */
-       u8 invalid_phy_mask;
-
-       /*
-        * This field saves the current interrupt coalescing number of the controller.
-        */
-       u16 interrupt_coalesce_number;
-
-       /*
-        * This field saves the current interrupt coalescing timeout value in microseconds.
-        */
-       u32 interrupt_coalesce_timeout;
-
-       /**
-        * This field is a pointer to the memory mapped register space for the
-        * struct smu_registers.
-        */
-       struct smu_registers __iomem *smu_registers;
-
-       /**
-        * This field is a pointer to the memory mapped register space for the
-        * struct scu_registers.
-        */
-       struct scu_registers __iomem *scu_registers;
-
-};
-
-/**
- * enum scic_sds_controller_states - This enumeration depicts all the states
- *    for the common controller state machine.
- */
-enum scic_sds_controller_states {
-       /**
-        * Simply the initial state for the base controller state machine.
-        */
-       SCI_BASE_CONTROLLER_STATE_INITIAL = 0,
-
-       /**
-        * This state indicates that the controller is reset.  The memory for
-        * the controller is in it's initial state, but the controller requires
-        * initialization.
-        * This state is entered from the INITIAL state.
-        * This state is entered from the RESETTING state.
-        */
-       SCI_BASE_CONTROLLER_STATE_RESET,
-
-       /**
-        * This state is typically an action state that indicates the controller
-        * is in the process of initialization.  In this state no new IO operations
-        * are permitted.
-        * This state is entered from the RESET state.
-        */
-       SCI_BASE_CONTROLLER_STATE_INITIALIZING,
-
-       /**
-        * This state indicates that the controller has been successfully
-        * initialized.  In this state no new IO operations are permitted.
-        * This state is entered from the INITIALIZING state.
-        */
-       SCI_BASE_CONTROLLER_STATE_INITIALIZED,
-
-       /**
-        * This state indicates the the controller is in the process of becoming
-        * ready (i.e. starting).  In this state no new IO operations are permitted.
-        * This state is entered from the INITIALIZED state.
-        */
-       SCI_BASE_CONTROLLER_STATE_STARTING,
-
-       /**
-        * This state indicates the controller is now ready.  Thus, the user
-        * is able to perform IO operations on the controller.
-        * This state is entered from the STARTING state.
-        */
-       SCI_BASE_CONTROLLER_STATE_READY,
-
-       /**
-        * This state is typically an action state that indicates the controller
-        * is in the process of resetting.  Thus, the user is unable to perform
-        * IO operations on the controller.  A reset is considered destructive in
-        * most cases.
-        * This state is entered from the READY state.
-        * This state is entered from the FAILED state.
-        * This state is entered from the STOPPED state.
-        */
-       SCI_BASE_CONTROLLER_STATE_RESETTING,
-
-       /**
-        * This state indicates that the controller is in the process of stopping.
-        * In this state no new IO operations are permitted, but existing IO
-        * operations are allowed to complete.
-        * This state is entered from the READY state.
-        */
-       SCI_BASE_CONTROLLER_STATE_STOPPING,
-
-       /**
-        * This state indicates that the controller has successfully been stopped.
-        * In this state no new IO operations are permitted.
-        * This state is entered from the STOPPING state.
-        */
-       SCI_BASE_CONTROLLER_STATE_STOPPED,
-
-       /**
-        * This state indicates that the controller could not successfully be
-        * initialized.  In this state no new IO operations are permitted.
-        * This state is entered from the INITIALIZING state.
-        * This state is entered from the STARTING state.
-        * This state is entered from the STOPPING state.
-        * This state is entered from the RESETTING state.
-        */
-       SCI_BASE_CONTROLLER_STATE_FAILED,
-
-       SCI_BASE_CONTROLLER_MAX_STATES
-
-};
-
-/**
- * INCREMENT_QUEUE_GET() -
- *
- * This macro will increment the specified index to and if the index wraps to 0
- * it will toggel the cycle bit.
- */
-#define INCREMENT_QUEUE_GET(index, cycle, entry_count, bit_toggle) \
-       { \
-               if ((index) + 1 == entry_count) {       \
-                       (index) = 0; \
-                       (cycle) = (cycle) ^ (bit_toggle); \
-               } else { \
-                       index = index + 1; \
-               } \
-       }
-
-/**
- * scic_sds_controller_get_port_configuration_agent() -
- *
- * This is a helper macro to get the port configuration agent from the
- * controller object.
- */
-#define scic_sds_controller_get_port_configuration_agent(controller) \
-       (&(controller)->port_agent)
-
-/**
- * scic_sds_controller_get_protocol_engine_group() -
- *
- * This macro returns the protocol engine group for this controller object.
- * Presently we only support protocol engine group 0 so just return that
- */
-#define scic_sds_controller_get_protocol_engine_group(controller) 0
-
-/**
- * scic_sds_io_tag_construct() -
- *
- * This macro constructs an IO tag from the sequence and index values.
- */
-#define scic_sds_io_tag_construct(sequence, task_index)        \
-       ((sequence) << 12 | (task_index))
-
-/**
- * scic_sds_io_tag_get_sequence() -
- *
- * This macro returns the IO sequence from the IO tag value.
- */
-#define scic_sds_io_tag_get_sequence(io_tag) \
-       (((io_tag) & 0xF000) >> 12)
-
-/**
- * scic_sds_io_tag_get_index() -
- *
- * This macro returns the TCi from the io tag value
- */
-#define scic_sds_io_tag_get_index(io_tag) \
-       ((io_tag) & 0x0FFF)
-
-/**
- * scic_sds_io_sequence_increment() -
- *
- * This is a helper macro to increment the io sequence count. We may find in
- * the future that it will be faster to store the sequence count in such a way
- * as we dont perform the shift operation to build io tag values so therefore
- * need a way to incrment them correctly
- */
-#define scic_sds_io_sequence_increment(value) \
-       ((value) = (((value) + 1) & 0x000F))
-
-/* expander attached sata devices require 3 rnc slots */
-static inline int scic_sds_remote_device_node_count(struct scic_sds_remote_device *sci_dev)
-{
-       struct domain_device *dev = sci_dev_to_domain(sci_dev);
-
-       if ((dev->dev_type == SATA_DEV || (dev->tproto & SAS_PROTOCOL_STP)) &&
-           !sci_dev->is_direct_attached)
-               return SCU_STP_REMOTE_NODE_COUNT;
-       return SCU_SSP_REMOTE_NODE_COUNT;
-}
-
-/**
- * scic_sds_controller_set_invalid_phy() -
- *
- * This macro will set the bit in the invalid phy mask for this controller
- * object.  This is used to control messages reported for invalid link up
- * notifications.
- */
-#define scic_sds_controller_set_invalid_phy(controller, phy) \
-       ((controller)->invalid_phy_mask |= (1 << (phy)->phy_index))
-
-/**
- * scic_sds_controller_clear_invalid_phy() -
- *
- * This macro will clear the bit in the invalid phy mask for this controller
- * object.  This is used to control messages reported for invalid link up
- * notifications.
- */
-#define scic_sds_controller_clear_invalid_phy(controller, phy) \
-       ((controller)->invalid_phy_mask &= ~(1 << (phy)->phy_index))
-
-void scic_sds_controller_post_request(
-       struct scic_sds_controller *this_controller,
-       u32 request);
-
-void scic_sds_controller_release_frame(
-       struct scic_sds_controller *this_controller,
-       u32 frame_index);
-
-void scic_sds_controller_copy_sata_response(
-       void *response_buffer,
-       void *frame_header,
-       void *frame_buffer);
-
-enum sci_status scic_sds_controller_allocate_remote_node_context(
-       struct scic_sds_controller *this_controller,
-       struct scic_sds_remote_device *sci_dev,
-       u16 *node_id);
-
-void scic_sds_controller_free_remote_node_context(
-       struct scic_sds_controller *this_controller,
-       struct scic_sds_remote_device *sci_dev,
-       u16 node_id);
-
-union scu_remote_node_context *scic_sds_controller_get_remote_node_context_buffer(
-       struct scic_sds_controller *this_controller,
-       u16 node_id);
-
-struct scic_sds_request *scic_request_by_tag(struct scic_sds_controller *scic,
-                                            u16 io_tag);
-
-struct scu_task_context *scic_sds_controller_get_task_context_buffer(
-       struct scic_sds_controller *this_controller,
-       u16 io_tag);
-
-void scic_sds_controller_power_control_queue_insert(
-       struct scic_sds_controller *this_controller,
-       struct scic_sds_phy *sci_phy);
-
-void scic_sds_controller_power_control_queue_remove(
-       struct scic_sds_controller *this_controller,
-       struct scic_sds_phy *sci_phy);
-
-void scic_sds_controller_link_up(
-       struct scic_sds_controller *this_controller,
-       struct scic_sds_port *sci_port,
-       struct scic_sds_phy *sci_phy);
-
-void scic_sds_controller_link_down(
-       struct scic_sds_controller *this_controller,
-       struct scic_sds_port *sci_port,
-       struct scic_sds_phy *sci_phy);
-
-void scic_sds_controller_remote_device_stopped(
-       struct scic_sds_controller *this_controller,
-       struct scic_sds_remote_device *sci_dev);
-
-void scic_sds_controller_copy_task_context(
-       struct scic_sds_controller *this_controller,
-       struct scic_sds_request *this_request);
-
-void scic_sds_controller_register_setup(
-       struct scic_sds_controller *this_controller);
-
-enum sci_status scic_controller_continue_io(struct scic_sds_request *sci_req);
-
-#endif /* _SCIC_SDS_CONTROLLER_H_ */
index c6df0e2..c82ccb9 100644 (file)
 
 #include <scsi/sas.h>
 #include "sas.h"
+#include "host.h"
 #include "sci_base_state.h"
 #include "sci_base_state_machine.h"
 #include "scic_phy.h"
-#include "scic_sds_controller.h"
 #include "scic_sds_phy.h"
 #include "scic_sds_port.h"
 #include "remote_node_context.h"
-#include "sci_environment.h"
 #include "sci_util.h"
 #include "scu_event_codes.h"
 #include "timers.h"
index 9302e39..652d823 100644 (file)
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-#include "scic_controller.h"
+#include "host.h"
 #include "scic_phy.h"
 #include "scic_port.h"
-#include "scic_sds_controller.h"
 #include "scic_sds_phy.h"
 #include "scic_sds_port.h"
 #include "remote_device.h"
 #include "remote_node_context.h"
 #include "scic_sds_request.h"
-#include "sci_environment.h"
 #include "scu_registers.h"
 #include "timers.h"
 
index 3696deb..bd612d5 100644 (file)
@@ -57,6 +57,7 @@
 #define _SCIC_SDS_PORT_H_
 
 #include <linux/kernel.h>
+#include "isci.h"
 #include "sas.h"
 #include "scu_registers.h"
 #include "sci_base_state_machine.h"
index 3fad8c1..a5871fd 100644 (file)
@@ -53,9 +53,7 @@
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-#include "sci_environment.h"
-#include "scic_controller.h"
-#include "scic_sds_controller.h"
+#include "host.h"
 #include "scic_sds_port_configuration_agent.h"
 #include "timers.h"
 
index de35885..1405aa7 100644 (file)
@@ -54,9 +54,7 @@
  */
 
 #include <scsi/sas.h>
-#include "scic_controller.h"
 #include "scic_io_request.h"
-#include "scic_sds_controller.h"
 #include "scu_registers.h"
 #include "scic_sds_port.h"
 #include "remote_device.h"
@@ -64,7 +62,6 @@
 #include "scic_sds_smp_request.h"
 #include "scic_sds_stp_request.h"
 #include "scic_sds_unsolicited_frame_control.h"
-#include "sci_environment.h"
 #include "sci_util.h"
 #include "scu_completion_codes.h"
 #include "scu_task_context.h"
index 2b91120..7f33894 100644 (file)
 
 #include <scsi/sas.h>
 #include "sci_base_state_machine.h"
-#include "scic_controller.h"
-#include "scic_sds_controller.h"
 #include "remote_device.h"
 #include "scic_sds_request.h"
 #include "scic_sds_smp_request.h"
-#include "sci_environment.h"
 #include "sci_util.h"
 #include "scu_completion_codes.h"
 #include "scu_task_context.h"
+#include "host.h"
 
 static void scu_smp_request_construct_task_context(
        struct scic_sds_request *sci_req,
index 18bf387..137f6dd 100644 (file)
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
+#include "host.h"
 #include "sci_base_state_machine.h"
-#include "scic_controller.h"
-#include "scic_sds_controller.h"
 #include "scic_sds_request.h"
-#include "sci_environment.h"
 #include "scu_completion_codes.h"
 #include "scu_task_context.h"
 
index c1c316c..2f50951 100644 (file)
 #include "sci_base_state.h"
 #include "sci_base_state_machine.h"
 #include "scic_io_request.h"
-#include "scic_sds_controller.h"
 #include "remote_device.h"
 #include "scic_sds_request.h"
 #include "scic_sds_stp_pio_request.h"
 #include "scic_sds_stp_request.h"
 #include "scic_sds_unsolicited_frame_control.h"
-#include "sci_environment.h"
 #include "sci_util.h"
 #include "scu_completion_codes.h"
 #include "scu_event_codes.h"
index 9e393e5..d0e0373 100644 (file)
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-/**
- * This file contains the implementation of the
- *    struct scic_sds_unsolicited_frame_control object and it's public, protected, and
- *    private methods.
- *
- *
- */
-
+#include "host.h"
 #include "scic_sds_unsolicited_frame_control.h"
 #include "scu_registers.h"
-#include "scic_sds_controller.h"
 #include "sci_util.h"
-#include "sci_environment.h"
 
 /**
  * This method will program the unsolicited frames (UFs) into the UF address
index 5847149..43a5d7a 100644 (file)
  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
-
+#include <linux/device.h>
+#include <scsi/sas.h>
+#include "host.h"
 #include "isci.h"
-#include "scic_io_request.h"
-#include "scic_port.h"
 #include "port.h"
-#include "request.h"
 #include "host.h"
 #include "probe_roms.h"
-#include "scic_controller.h"
-#include "scic_sds_controller.h"
+#include "remote_device.h"
+#include "request.h"
+#include "scic_io_request.h"
+#include "scic_sds_port_configuration_agent.h"
+#include "sci_util.h"
+#include "scu_completion_codes.h"
+#include "scu_event_codes.h"
+#include "scu_registers.h"
+#include "scu_remote_node_context.h"
+#include "scu_task_context.h"
+#include "scu_unsolicited_frame.h"
 #include "timers.h"
 
+#define SCU_CONTEXT_RAM_INIT_STALL_TIME      200
+
+/**
+ * smu_dcc_get_max_ports() -
+ *
+ * This macro returns the maximum number of logical ports supported by the
+ * hardware. The caller passes in the value read from the device context
+ * capacity register and this macro will mash and shift the value appropriately.
+ */
+#define smu_dcc_get_max_ports(dcc_value) \
+       (\
+               (((dcc_value) & SMU_DEVICE_CONTEXT_CAPACITY_MAX_LP_MASK) \
+                >> SMU_DEVICE_CONTEXT_CAPACITY_MAX_LP_SHIFT) + 1 \
+       )
+
+/**
+ * smu_dcc_get_max_task_context() -
+ *
+ * This macro returns the maximum number of task contexts supported by the
+ * hardware. The caller passes in the value read from the device context
+ * capacity register and this macro will mash and shift the value appropriately.
+ */
+#define smu_dcc_get_max_task_context(dcc_value)        \
+       (\
+               (((dcc_value) & SMU_DEVICE_CONTEXT_CAPACITY_MAX_TC_MASK) \
+                >> SMU_DEVICE_CONTEXT_CAPACITY_MAX_TC_SHIFT) + 1 \
+       )
+
+/**
+ * smu_dcc_get_max_remote_node_context() -
+ *
+ * This macro returns the maximum number of remote node contexts supported by
+ * the hardware. The caller passes in the value read from the device context
+ * capacity register and this macro will mash and shift the value appropriately.
+ */
+#define smu_dcc_get_max_remote_node_context(dcc_value) \
+       (\
+               (((dcc_value) & SMU_DEVICE_CONTEXT_CAPACITY_MAX_RNC_MASK) \
+                >> SMU_DEVICE_CONTEXT_CAPACITY_MAX_RNC_SHIFT) + 1 \
+       )
+
+
+#define SCIC_SDS_CONTROLLER_MIN_TIMER_COUNT  3
+#define SCIC_SDS_CONTROLLER_MAX_TIMER_COUNT  3
+
+/**
+ *
+ *
+ * The number of milliseconds to wait for a phy to start.
+ */
+#define SCIC_SDS_CONTROLLER_PHY_START_TIMEOUT      100
+
+/**
+ *
+ *
+ * The number of milliseconds to wait while a given phy is consuming power
+ * before allowing another set of phys to consume power. Ultimately, this will
+ * be specified by OEM parameter.
+ */
+#define SCIC_SDS_CONTROLLER_POWER_CONTROL_INTERVAL 500
+
+/**
+ * NORMALIZE_PUT_POINTER() -
+ *
+ * This macro will normalize the completion queue put pointer so its value can
+ * be used as an array inde
+ */
+#define NORMALIZE_PUT_POINTER(x) \
+       ((x) & SMU_COMPLETION_QUEUE_PUT_POINTER_MASK)
+
+
+/**
+ * NORMALIZE_EVENT_POINTER() -
+ *
+ * This macro will normalize the completion queue event entry so its value can
+ * be used as an index.
+ */
+#define NORMALIZE_EVENT_POINTER(x) \
+       (\
+               ((x) & SMU_COMPLETION_QUEUE_GET_EVENT_POINTER_MASK) \
+               >> SMU_COMPLETION_QUEUE_GET_EVENT_POINTER_SHIFT \
+       )
+
+/**
+ * INCREMENT_COMPLETION_QUEUE_GET() -
+ *
+ * This macro will increment the controllers completion queue index value and
+ * possibly toggle the cycle bit if the completion queue index wraps back to 0.
+ */
+#define INCREMENT_COMPLETION_QUEUE_GET(controller, index, cycle) \
+       INCREMENT_QUEUE_GET(\
+               (index), \
+               (cycle), \
+               (controller)->completion_queue_entries, \
+               SMU_CQGR_CYCLE_BIT \
+               )
+
+/**
+ * INCREMENT_EVENT_QUEUE_GET() -
+ *
+ * This macro will increment the controllers event queue index value and
+ * possibly toggle the event cycle bit if the event queue index wraps back to 0.
+ */
+#define INCREMENT_EVENT_QUEUE_GET(controller, index, cycle) \
+       INCREMENT_QUEUE_GET(\
+               (index), \
+               (cycle), \
+               (controller)->completion_event_entries, \
+               SMU_CQGR_EVENT_CYCLE_BIT \
+               )
+
+
+/**
+ * NORMALIZE_GET_POINTER() -
+ *
+ * This macro will normalize the completion queue get pointer so its value can
+ * be used as an index into an array
+ */
+#define NORMALIZE_GET_POINTER(x) \
+       ((x) & SMU_COMPLETION_QUEUE_GET_POINTER_MASK)
+
+/**
+ * NORMALIZE_GET_POINTER_CYCLE_BIT() -
+ *
+ * This macro will normalize the completion queue cycle pointer so it matches
+ * the completion queue cycle bit
+ */
+#define NORMALIZE_GET_POINTER_CYCLE_BIT(x) \
+       ((SMU_CQGR_CYCLE_BIT & (x)) << (31 - SMU_COMPLETION_QUEUE_GET_CYCLE_BIT_SHIFT))
+
+/**
+ * COMPLETION_QUEUE_CYCLE_BIT() -
+ *
+ * This macro will return the cycle bit of the completion queue entry
+ */
+#define COMPLETION_QUEUE_CYCLE_BIT(x) ((x) & 0x80000000)
+
+static bool scic_sds_controller_completion_queue_has_entries(
+       struct scic_sds_controller *scic)
+{
+       u32 get_value = scic->completion_queue_get;
+       u32 get_index = get_value & SMU_COMPLETION_QUEUE_GET_POINTER_MASK;
+
+       if (NORMALIZE_GET_POINTER_CYCLE_BIT(get_value) ==
+           COMPLETION_QUEUE_CYCLE_BIT(scic->completion_queue[get_index]))
+               return true;
+
+       return false;
+}
+
+static bool scic_sds_controller_isr(struct scic_sds_controller *scic)
+{
+       if (scic_sds_controller_completion_queue_has_entries(scic)) {
+               return true;
+       } else {
+               /*
+                * we have a spurious interrupt it could be that we have already
+                * emptied the completion queue from a previous interrupt */
+               writel(SMU_ISR_COMPLETION, &scic->smu_registers->interrupt_status);
+
+               /*
+                * There is a race in the hardware that could cause us not to be notified
+                * of an interrupt completion if we do not take this step.  We will mask
+                * then unmask the interrupts so if there is another interrupt pending
+                * the clearing of the interrupt source we get the next interrupt message. */
+               writel(0xFF000000, &scic->smu_registers->interrupt_mask);
+               writel(0, &scic->smu_registers->interrupt_mask);
+       }
+
+       return false;
+}
+
 irqreturn_t isci_msix_isr(int vec, void *data)
 {
        struct isci_host *ihost = data;
@@ -74,6 +254,411 @@ irqreturn_t isci_msix_isr(int vec, void *data)
        return IRQ_HANDLED;
 }
 
+static bool scic_sds_controller_error_isr(struct scic_sds_controller *scic)
+{
+       u32 interrupt_status;
+
+       interrupt_status =
+               readl(&scic->smu_registers->interrupt_status);
+       interrupt_status &= (SMU_ISR_QUEUE_ERROR | SMU_ISR_QUEUE_SUSPEND);
+
+       if (interrupt_status != 0) {
+               /*
+                * There is an error interrupt pending so let it through and handle
+                * in the callback */
+               return true;
+       }
+
+       /*
+        * There is a race in the hardware that could cause us not to be notified
+        * of an interrupt completion if we do not take this step.  We will mask
+        * then unmask the error interrupts so if there was another interrupt
+        * pending we will be notified.
+        * Could we write the value of (SMU_ISR_QUEUE_ERROR | SMU_ISR_QUEUE_SUSPEND)? */
+       writel(0xff, &scic->smu_registers->interrupt_mask);
+       writel(0, &scic->smu_registers->interrupt_mask);
+
+       return false;
+}
+
+static void scic_sds_controller_task_completion(struct scic_sds_controller *scic,
+                                               u32 completion_entry)
+{
+       u32 index;
+       struct scic_sds_request *io_request;
+
+       index = SCU_GET_COMPLETION_INDEX(completion_entry);
+       io_request = scic->io_request_table[index];
+
+       /* Make sure that we really want to process this IO request */
+       if (
+               (io_request != NULL)
+               && (io_request->io_tag != SCI_CONTROLLER_INVALID_IO_TAG)
+               && (
+                       scic_sds_io_tag_get_sequence(io_request->io_tag)
+                       == scic->io_request_sequence[index]
+                       )
+               ) {
+               /* Yep this is a valid io request pass it along to the io request handler */
+               scic_sds_io_request_tc_completion(io_request, completion_entry);
+       }
+}
+
+static void scic_sds_controller_sdma_completion(struct scic_sds_controller *scic,
+                                               u32 completion_entry)
+{
+       u32 index;
+       struct scic_sds_request *io_request;
+       struct scic_sds_remote_device *device;
+
+       index = SCU_GET_COMPLETION_INDEX(completion_entry);
+
+       switch (scu_get_command_request_type(completion_entry)) {
+       case SCU_CONTEXT_COMMAND_REQUEST_TYPE_POST_TC:
+       case SCU_CONTEXT_COMMAND_REQUEST_TYPE_DUMP_TC:
+               io_request = scic->io_request_table[index];
+               dev_warn(scic_to_dev(scic),
+                        "%s: SCIC SDS Completion type SDMA %x for io request "
+                        "%p\n",
+                        __func__,
+                        completion_entry,
+                        io_request);
+               /* @todo For a post TC operation we need to fail the IO
+                * request
+                */
+               break;
+
+       case SCU_CONTEXT_COMMAND_REQUEST_TYPE_DUMP_RNC:
+       case SCU_CONTEXT_COMMAND_REQUEST_TYPE_OTHER_RNC:
+       case SCU_CONTEXT_COMMAND_REQUEST_TYPE_POST_RNC:
+               device = scic->device_table[index];
+               dev_warn(scic_to_dev(scic),
+                        "%s: SCIC SDS Completion type SDMA %x for remote "
+                        "device %p\n",
+                        __func__,
+                        completion_entry,
+                        device);
+               /* @todo For a port RNC operation we need to fail the
+                * device
+                */
+               break;
+
+       default:
+               dev_warn(scic_to_dev(scic),
+                        "%s: SCIC SDS Completion unknown SDMA completion "
+                        "type %x\n",
+                        __func__,
+                        completion_entry);
+               break;
+
+       }
+}
+
+static void scic_sds_controller_unsolicited_frame(struct scic_sds_controller *scic,
+                                                 u32 completion_entry)
+{
+       u32 index;
+       u32 frame_index;
+
+       struct isci_host *ihost = scic_to_ihost(scic);
+       struct scu_unsolicited_frame_header *frame_header;
+       struct scic_sds_phy *phy;
+       struct scic_sds_remote_device *device;
+
+       enum sci_status result = SCI_FAILURE;
+
+       frame_index = SCU_GET_FRAME_INDEX(completion_entry);
+
+       frame_header = scic->uf_control.buffers.array[frame_index].header;
+       scic->uf_control.buffers.array[frame_index].state = UNSOLICITED_FRAME_IN_USE;
+
+       if (SCU_GET_FRAME_ERROR(completion_entry)) {
+               /*
+                * / @todo If the IAF frame or SIGNATURE FIS frame has an error will
+                * /       this cause a problem? We expect the phy initialization will
+                * /       fail if there is an error in the frame. */
+               scic_sds_controller_release_frame(scic, frame_index);
+               return;
+       }
+
+       if (frame_header->is_address_frame) {
+               index = SCU_GET_PROTOCOL_ENGINE_INDEX(completion_entry);
+               phy = &ihost->phys[index].sci;
+               result = scic_sds_phy_frame_handler(phy, frame_index);
+       } else {
+
+               index = SCU_GET_COMPLETION_INDEX(completion_entry);
+
+               if (index == SCIC_SDS_REMOTE_NODE_CONTEXT_INVALID_INDEX) {
+                       /*
+                        * This is a signature fis or a frame from a direct attached SATA
+                        * device that has not yet been created.  In either case forwared
+                        * the frame to the PE and let it take care of the frame data. */
+                       index = SCU_GET_PROTOCOL_ENGINE_INDEX(completion_entry);
+                       phy = &ihost->phys[index].sci;
+                       result = scic_sds_phy_frame_handler(phy, frame_index);
+               } else {
+                       if (index < scic->remote_node_entries)
+                               device = scic->device_table[index];
+                       else
+                               device = NULL;
+
+                       if (device != NULL)
+                               result = scic_sds_remote_device_frame_handler(device, frame_index);
+                       else
+                               scic_sds_controller_release_frame(scic, frame_index);
+               }
+       }
+
+       if (result != SCI_SUCCESS) {
+               /*
+                * / @todo Is there any reason to report some additional error message
+                * /       when we get this failure notifiction? */
+       }
+}
+
+static void scic_sds_controller_event_completion(struct scic_sds_controller *scic,
+                                                u32 completion_entry)
+{
+       struct isci_host *ihost = scic_to_ihost(scic);
+       struct scic_sds_request *io_request;
+       struct scic_sds_remote_device *device;
+       struct scic_sds_phy *phy;
+       u32 index;
+
+       index = SCU_GET_COMPLETION_INDEX(completion_entry);
+
+       switch (scu_get_event_type(completion_entry)) {
+       case SCU_EVENT_TYPE_SMU_COMMAND_ERROR:
+               /* / @todo The driver did something wrong and we need to fix the condtion. */
+               dev_err(scic_to_dev(scic),
+                       "%s: SCIC Controller 0x%p received SMU command error "
+                       "0x%x\n",
+                       __func__,
+                       scic,
+                       completion_entry);
+               break;
+
+       case SCU_EVENT_TYPE_SMU_PCQ_ERROR:
+       case SCU_EVENT_TYPE_SMU_ERROR:
+       case SCU_EVENT_TYPE_FATAL_MEMORY_ERROR:
+               /*
+                * / @todo This is a hardware failure and its likely that we want to
+                * /       reset the controller. */
+               dev_err(scic_to_dev(scic),
+                       "%s: SCIC Controller 0x%p received fatal controller "
+                       "event  0x%x\n",
+                       __func__,
+                       scic,
+                       completion_entry);
+               break;
+
+       case SCU_EVENT_TYPE_TRANSPORT_ERROR:
+               io_request = scic->io_request_table[index];
+               scic_sds_io_request_event_handler(io_request, completion_entry);
+               break;
+
+       case SCU_EVENT_TYPE_PTX_SCHEDULE_EVENT:
+               switch (scu_get_event_specifier(completion_entry)) {
+               case SCU_EVENT_SPECIFIC_SMP_RESPONSE_NO_PE:
+               case SCU_EVENT_SPECIFIC_TASK_TIMEOUT:
+                       io_request = scic->io_request_table[index];
+                       if (io_request != NULL)
+                               scic_sds_io_request_event_handler(io_request, completion_entry);
+                       else
+                               dev_warn(scic_to_dev(scic),
+                                        "%s: SCIC Controller 0x%p received "
+                                        "event 0x%x for io request object "
+                                        "that doesnt exist.\n",
+                                        __func__,
+                                        scic,
+                                        completion_entry);
+
+                       break;
+
+               case SCU_EVENT_SPECIFIC_IT_NEXUS_TIMEOUT:
+                       device = scic->device_table[index];
+                       if (device != NULL)
+                               scic_sds_remote_device_event_handler(device, completion_entry);
+                       else
+                               dev_warn(scic_to_dev(scic),
+                                        "%s: SCIC Controller 0x%p received "
+                                        "event 0x%x for remote device object "
+                                        "that doesnt exist.\n",
+                                        __func__,
+                                        scic,
+                                        completion_entry);
+
+                       break;
+               }
+               break;
+
+       case SCU_EVENT_TYPE_BROADCAST_CHANGE:
+       /*
+        * direct the broadcast change event to the phy first and then let
+        * the phy redirect the broadcast change to the port object */
+       case SCU_EVENT_TYPE_ERR_CNT_EVENT:
+       /*
+        * direct error counter event to the phy object since that is where
+        * we get the event notification.  This is a type 4 event. */
+       case SCU_EVENT_TYPE_OSSP_EVENT:
+               index = SCU_GET_PROTOCOL_ENGINE_INDEX(completion_entry);
+               phy = &ihost->phys[index].sci;
+               scic_sds_phy_event_handler(phy, completion_entry);
+               break;
+
+       case SCU_EVENT_TYPE_RNC_SUSPEND_TX:
+       case SCU_EVENT_TYPE_RNC_SUSPEND_TX_RX:
+       case SCU_EVENT_TYPE_RNC_OPS_MISC:
+               if (index < scic->remote_node_entries) {
+                       device = scic->device_table[index];
+
+                       if (device != NULL)
+                               scic_sds_remote_device_event_handler(device, completion_entry);
+               } else
+                       dev_err(scic_to_dev(scic),
+                               "%s: SCIC Controller 0x%p received event 0x%x "
+                               "for remote device object 0x%0x that doesnt "
+                               "exist.\n",
+                               __func__,
+                               scic,
+                               completion_entry,
+                               index);
+
+               break;
+
+       default:
+               dev_warn(scic_to_dev(scic),
+                        "%s: SCIC Controller received unknown event code %x\n",
+                        __func__,
+                        completion_entry);
+               break;
+       }
+}
+
+
+
+static void scic_sds_controller_process_completions(struct scic_sds_controller *scic)
+{
+       u32 completion_count = 0;
+       u32 completion_entry;
+       u32 get_index;
+       u32 get_cycle;
+       u32 event_index;
+       u32 event_cycle;
+
+       dev_dbg(scic_to_dev(scic),
+               "%s: completion queue begining get:0x%08x\n",
+               __func__,
+               scic->completion_queue_get);
+
+       /* Get the component parts of the completion queue */
+       get_index = NORMALIZE_GET_POINTER(scic->completion_queue_get);
+       get_cycle = SMU_CQGR_CYCLE_BIT & scic->completion_queue_get;
+
+       event_index = NORMALIZE_EVENT_POINTER(scic->completion_queue_get);
+       event_cycle = SMU_CQGR_EVENT_CYCLE_BIT & scic->completion_queue_get;
+
+       while (
+               NORMALIZE_GET_POINTER_CYCLE_BIT(get_cycle)
+               == COMPLETION_QUEUE_CYCLE_BIT(scic->completion_queue[get_index])
+               ) {
+               completion_count++;
+
+               completion_entry = scic->completion_queue[get_index];
+               INCREMENT_COMPLETION_QUEUE_GET(scic, get_index, get_cycle);
+
+               dev_dbg(scic_to_dev(scic),
+                       "%s: completion queue entry:0x%08x\n",
+                       __func__,
+                       completion_entry);
+
+               switch (SCU_GET_COMPLETION_TYPE(completion_entry)) {
+               case SCU_COMPLETION_TYPE_TASK:
+                       scic_sds_controller_task_completion(scic, completion_entry);
+                       break;
+
+               case SCU_COMPLETION_TYPE_SDMA:
+                       scic_sds_controller_sdma_completion(scic, completion_entry);
+                       break;
+
+               case SCU_COMPLETION_TYPE_UFI:
+                       scic_sds_controller_unsolicited_frame(scic, completion_entry);
+                       break;
+
+               case SCU_COMPLETION_TYPE_EVENT:
+                       INCREMENT_EVENT_QUEUE_GET(scic, event_index, event_cycle);
+                       scic_sds_controller_event_completion(scic, completion_entry);
+                       break;
+
+               case SCU_COMPLETION_TYPE_NOTIFY:
+                       /*
+                        * Presently we do the same thing with a notify event that we do with the
+                        * other event codes. */
+                       INCREMENT_EVENT_QUEUE_GET(scic, event_index, event_cycle);
+                       scic_sds_controller_event_completion(scic, completion_entry);
+                       break;
+
+               default:
+                       dev_warn(scic_to_dev(scic),
+                                "%s: SCIC Controller received unknown "
+                                "completion type %x\n",
+                                __func__,
+                                completion_entry);
+                       break;
+               }
+       }
+
+       /* Update the get register if we completed one or more entries */
+       if (completion_count > 0) {
+               scic->completion_queue_get =
+                       SMU_CQGR_GEN_BIT(ENABLE) |
+                       SMU_CQGR_GEN_BIT(EVENT_ENABLE) |
+                       event_cycle |
+                       SMU_CQGR_GEN_VAL(EVENT_POINTER, event_index) |
+                       get_cycle |
+                       SMU_CQGR_GEN_VAL(POINTER, get_index);
+
+               writel(scic->completion_queue_get,
+                      &scic->smu_registers->completion_queue_get);
+
+       }
+
+       dev_dbg(scic_to_dev(scic),
+               "%s: completion queue ending get:0x%08x\n",
+               __func__,
+               scic->completion_queue_get);
+
+}
+
+static void scic_sds_controller_error_handler(struct scic_sds_controller *scic)
+{
+       u32 interrupt_status;
+
+       interrupt_status =
+               readl(&scic->smu_registers->interrupt_status);
+
+       if ((interrupt_status & SMU_ISR_QUEUE_SUSPEND) &&
+           scic_sds_controller_completion_queue_has_entries(scic)) {
+
+               scic_sds_controller_process_completions(scic);
+               writel(SMU_ISR_QUEUE_SUSPEND, &scic->smu_registers->interrupt_status);
+       } else {
+               dev_err(scic_to_dev(scic), "%s: status: %#x\n", __func__,
+                       interrupt_status);
+
+               sci_base_state_machine_change_state(&scic->state_machine,
+                                                   SCI_BASE_CONTROLLER_STATE_FAILED);
+
+               return;
+       }
+
+       /* If we dont process any completions I am not sure that we want to do this.
+        * We are in the middle of a hardware fault and should probably be reset.
+        */
+       writel(0, &scic->smu_registers->interrupt_mask);
+}
+
 irqreturn_t isci_intx_isr(int vec, void *data)
 {
        irqreturn_t ret = IRQ_NONE;
@@ -112,7 +697,7 @@ irqreturn_t isci_error_isr(int vec, void *data)
  *    core library.
  *
  */
-void isci_host_start_complete(struct isci_host *ihost, enum sci_status completion_status)
+static void isci_host_start_complete(struct isci_host *ihost, enum sci_status completion_status)
 {
        if (completion_status != SCI_SUCCESS)
                dev_info(&ihost->pdev->dev,
@@ -142,6 +727,383 @@ int isci_host_scan_finished(struct Scsi_Host *shost, unsigned long time)
 
 }
 
+/**
+ * scic_controller_get_suggested_start_timeout() - This method returns the
+ *    suggested scic_controller_start() timeout amount.  The user is free to
+ *    use any timeout value, but this method provides the suggested minimum
+ *    start timeout value.  The returned value is based upon empirical
+ *    information determined as a result of interoperability testing.
+ * @controller: the handle to the controller object for which to return the
+ *    suggested start timeout.
+ *
+ * This method returns the number of milliseconds for the suggested start
+ * operation timeout.
+ */
+static u32 scic_controller_get_suggested_start_timeout(
+       struct scic_sds_controller *sc)
+{
+       /* Validate the user supplied parameters. */
+       if (sc == NULL)
+               return 0;
+
+       /*
+        * The suggested minimum timeout value for a controller start operation:
+        *
+        *     Signature FIS Timeout
+        *   + Phy Start Timeout
+        *   + Number of Phy Spin Up Intervals
+        *   ---------------------------------
+        *   Number of milliseconds for the controller start operation.
+        *
+        * NOTE: The number of phy spin up intervals will be equivalent
+        *       to the number of phys divided by the number phys allowed
+        *       per interval - 1 (once OEM parameters are supported).
+        *       Currently we assume only 1 phy per interval. */
+
+       return SCIC_SDS_SIGNATURE_FIS_TIMEOUT
+               + SCIC_SDS_CONTROLLER_PHY_START_TIMEOUT
+               + ((SCI_MAX_PHYS - 1) * SCIC_SDS_CONTROLLER_POWER_CONTROL_INTERVAL);
+}
+
+static void scic_controller_enable_interrupts(
+       struct scic_sds_controller *scic)
+{
+       BUG_ON(scic->smu_registers == NULL);
+       writel(0, &scic->smu_registers->interrupt_mask);
+}
+
+void scic_controller_disable_interrupts(
+       struct scic_sds_controller *scic)
+{
+       BUG_ON(scic->smu_registers == NULL);
+       writel(0xffffffff, &scic->smu_registers->interrupt_mask);
+}
+
+static void scic_sds_controller_enable_port_task_scheduler(
+       struct scic_sds_controller *scic)
+{
+       u32 port_task_scheduler_value;
+
+       port_task_scheduler_value =
+               readl(&scic->scu_registers->peg0.ptsg.control);
+       port_task_scheduler_value |=
+               (SCU_PTSGCR_GEN_BIT(ETM_ENABLE) |
+                SCU_PTSGCR_GEN_BIT(PTSG_ENABLE));
+       writel(port_task_scheduler_value,
+              &scic->scu_registers->peg0.ptsg.control);
+}
+
+static void scic_sds_controller_assign_task_entries(struct scic_sds_controller *scic)
+{
+       u32 task_assignment;
+
+       /*
+        * Assign all the TCs to function 0
+        * TODO: Do we actually need to read this register to write it back?
+        */
+
+       task_assignment =
+               readl(&scic->smu_registers->task_context_assignment[0]);
+
+       task_assignment |= (SMU_TCA_GEN_VAL(STARTING, 0)) |
+               (SMU_TCA_GEN_VAL(ENDING,  scic->task_context_entries - 1)) |
+               (SMU_TCA_GEN_BIT(RANGE_CHECK_ENABLE));
+
+       writel(task_assignment,
+               &scic->smu_registers->task_context_assignment[0]);
+
+}
+
+static void scic_sds_controller_initialize_completion_queue(struct scic_sds_controller *scic)
+{
+       u32 index;
+       u32 completion_queue_control_value;
+       u32 completion_queue_get_value;
+       u32 completion_queue_put_value;
+
+       scic->completion_queue_get = 0;
+
+       completion_queue_control_value = (
+               SMU_CQC_QUEUE_LIMIT_SET(scic->completion_queue_entries - 1)
+               | SMU_CQC_EVENT_LIMIT_SET(scic->completion_event_entries - 1)
+               );
+
+       writel(completion_queue_control_value,
+              &scic->smu_registers->completion_queue_control);
+
+
+       /* Set the completion queue get pointer and enable the queue */
+       completion_queue_get_value = (
+               (SMU_CQGR_GEN_VAL(POINTER, 0))
+               | (SMU_CQGR_GEN_VAL(EVENT_POINTER, 0))
+               | (SMU_CQGR_GEN_BIT(ENABLE))
+               | (SMU_CQGR_GEN_BIT(EVENT_ENABLE))
+               );
+
+       writel(completion_queue_get_value,
+              &scic->smu_registers->completion_queue_get);
+
+       /* Set the completion queue put pointer */
+       completion_queue_put_value = (
+               (SMU_CQPR_GEN_VAL(POINTER, 0))
+               | (SMU_CQPR_GEN_VAL(EVENT_POINTER, 0))
+               );
+
+       writel(completion_queue_put_value,
+              &scic->smu_registers->completion_queue_put);
+
+       /* Initialize the cycle bit of the completion queue entries */
+       for (index = 0; index < scic->completion_queue_entries; index++) {
+               /*
+                * If get.cycle_bit != completion_queue.cycle_bit
+                * its not a valid completion queue entry
+                * so at system start all entries are invalid */
+               scic->completion_queue[index] = 0x80000000;
+       }
+}
+
+static void scic_sds_controller_initialize_unsolicited_frame_queue(struct scic_sds_controller *scic)
+{
+       u32 frame_queue_control_value;
+       u32 frame_queue_get_value;
+       u32 frame_queue_put_value;
+
+       /* Write the queue size */
+       frame_queue_control_value =
+               SCU_UFQC_GEN_VAL(QUEUE_SIZE,
+                                scic->uf_control.address_table.count);
+
+       writel(frame_queue_control_value,
+              &scic->scu_registers->sdma.unsolicited_frame_queue_control);
+
+       /* Setup the get pointer for the unsolicited frame queue */
+       frame_queue_get_value = (
+               SCU_UFQGP_GEN_VAL(POINTER, 0)
+               |  SCU_UFQGP_GEN_BIT(ENABLE_BIT)
+               );
+
+       writel(frame_queue_get_value,
+              &scic->scu_registers->sdma.unsolicited_frame_get_pointer);
+       /* Setup the put pointer for the unsolicited frame queue */
+       frame_queue_put_value = SCU_UFQPP_GEN_VAL(POINTER, 0);
+       writel(frame_queue_put_value,
+              &scic->scu_registers->sdma.unsolicited_frame_put_pointer);
+}
+
+/**
+ * This method will attempt to transition into the ready state for the
+ *    controller and indicate that the controller start operation has completed
+ *    if all criteria are met.
+ * @scic: This parameter indicates the controller object for which
+ *    to transition to ready.
+ * @status: This parameter indicates the status value to be pass into the call
+ *    to scic_cb_controller_start_complete().
+ *
+ * none.
+ */
+static void scic_sds_controller_transition_to_ready(
+       struct scic_sds_controller *scic,
+       enum sci_status status)
+{
+       struct isci_host *ihost = scic_to_ihost(scic);
+
+       if (scic->state_machine.current_state_id ==
+           SCI_BASE_CONTROLLER_STATE_STARTING) {
+               /*
+                * We move into the ready state, because some of the phys/ports
+                * may be up and operational.
+                */
+               sci_base_state_machine_change_state(&scic->state_machine,
+                                                   SCI_BASE_CONTROLLER_STATE_READY);
+
+               isci_host_start_complete(ihost, status);
+       }
+}
+
+static void scic_sds_controller_phy_timer_stop(struct scic_sds_controller *scic)
+{
+       isci_timer_stop(scic->phy_startup_timer);
+
+       scic->phy_startup_timer_pending = false;
+}
+
+static void scic_sds_controller_phy_timer_start(struct scic_sds_controller *scic)
+{
+       isci_timer_start(scic->phy_startup_timer,
+                        SCIC_SDS_CONTROLLER_PHY_START_TIMEOUT);
+
+       scic->phy_startup_timer_pending = true;
+}
+
+/**
+ * scic_sds_controller_start_next_phy - start phy
+ * @scic: controller
+ *
+ * If all the phys have been started, then attempt to transition the
+ * controller to the READY state and inform the user
+ * (scic_cb_controller_start_complete()).
+ */
+static enum sci_status scic_sds_controller_start_next_phy(struct scic_sds_controller *scic)
+{
+       struct isci_host *ihost = scic_to_ihost(scic);
+       struct scic_sds_oem_params *oem = &scic->oem_parameters.sds1;
+       struct scic_sds_phy *sci_phy;
+       enum sci_status status;
+
+       status = SCI_SUCCESS;
+
+       if (scic->phy_startup_timer_pending)
+               return status;
+
+       if (scic->next_phy_to_start >= SCI_MAX_PHYS) {
+               bool is_controller_start_complete = true;
+               u32 state;
+               u8 index;
+
+               for (index = 0; index < SCI_MAX_PHYS; index++) {
+                       sci_phy = &ihost->phys[index].sci;
+                       state = sci_phy->state_machine.current_state_id;
+
+                       if (!scic_sds_phy_get_port(sci_phy))
+                               continue;
+
+                       /* The controller start operation is complete iff:
+                        * - all links have been given an opportunity to start
+                        * - have no indication of a connected device
+                        * - have an indication of a connected device and it has
+                        *   finished the link training process.
+                        */
+                       if ((sci_phy->is_in_link_training == false &&
+                            state == SCI_BASE_PHY_STATE_INITIAL) ||
+                           (sci_phy->is_in_link_training == false &&
+                            state == SCI_BASE_PHY_STATE_STOPPED) ||
+                           (sci_phy->is_in_link_training == true &&
+                            state == SCI_BASE_PHY_STATE_STARTING)) {
+                               is_controller_start_complete = false;
+                               break;
+                       }
+               }
+
+               /*
+                * The controller has successfully finished the start process.
+                * Inform the SCI Core user and transition to the READY state. */
+               if (is_controller_start_complete == true) {
+                       scic_sds_controller_transition_to_ready(scic, SCI_SUCCESS);
+                       scic_sds_controller_phy_timer_stop(scic);
+               }
+       } else {
+               sci_phy = &ihost->phys[scic->next_phy_to_start].sci;
+
+               if (oem->controller.mode_type == SCIC_PORT_MANUAL_CONFIGURATION_MODE) {
+                       if (scic_sds_phy_get_port(sci_phy) == NULL) {
+                               scic->next_phy_to_start++;
+
+                               /* Caution recursion ahead be forwarned
+                                *
+                                * The PHY was never added to a PORT in MPC mode
+                                * so start the next phy in sequence This phy
+                                * will never go link up and will not draw power
+                                * the OEM parameters either configured the phy
+                                * incorrectly for the PORT or it was never
+                                * assigned to a PORT
+                                */
+                               return scic_sds_controller_start_next_phy(scic);
+                       }
+               }
+
+               status = scic_sds_phy_start(sci_phy);
+
+               if (status == SCI_SUCCESS) {
+                       scic_sds_controller_phy_timer_start(scic);
+               } else {
+                       dev_warn(scic_to_dev(scic),
+                                "%s: Controller stop operation failed "
+                                "to stop phy %d because of status "
+                                "%d.\n",
+                                __func__,
+                                ihost->phys[scic->next_phy_to_start].sci.phy_index,
+                                status);
+               }
+
+               scic->next_phy_to_start++;
+       }
+
+       return status;
+}
+
+static void scic_sds_controller_phy_startup_timeout_handler(void *_scic)
+{
+       struct scic_sds_controller *scic = _scic;
+       enum sci_status status;
+
+       scic->phy_startup_timer_pending = false;
+       status = SCI_FAILURE;
+       while (status != SCI_SUCCESS)
+               status = scic_sds_controller_start_next_phy(scic);
+}
+
+static enum sci_status scic_controller_start(struct scic_sds_controller *scic,
+                                            u32 timeout)
+{
+       struct isci_host *ihost = scic_to_ihost(scic);
+       enum sci_status result;
+       u16 index;
+
+       if (scic->state_machine.current_state_id !=
+           SCI_BASE_CONTROLLER_STATE_INITIALIZED) {
+               dev_warn(scic_to_dev(scic),
+                        "SCIC Controller start operation requested in "
+                        "invalid state\n");
+               return SCI_FAILURE_INVALID_STATE;
+       }
+
+       /* Build the TCi free pool */
+       sci_pool_initialize(scic->tci_pool);
+       for (index = 0; index < scic->task_context_entries; index++)
+               sci_pool_put(scic->tci_pool, index);
+
+       /* Build the RNi free pool */
+       scic_sds_remote_node_table_initialize(
+                       &scic->available_remote_nodes,
+                       scic->remote_node_entries);
+
+       /*
+        * Before anything else lets make sure we will not be
+        * interrupted by the hardware.
+        */
+       scic_controller_disable_interrupts(scic);
+
+       /* Enable the port task scheduler */
+       scic_sds_controller_enable_port_task_scheduler(scic);
+
+       /* Assign all the task entries to scic physical function */
+       scic_sds_controller_assign_task_entries(scic);
+
+       /* Now initialize the completion queue */
+       scic_sds_controller_initialize_completion_queue(scic);
+
+       /* Initialize the unsolicited frame queue for use */
+       scic_sds_controller_initialize_unsolicited_frame_queue(scic);
+
+       /* Start all of the ports on this controller */
+       for (index = 0; index < scic->logical_port_entries; index++) {
+               struct scic_sds_port *sci_port = &ihost->ports[index].sci;
+
+               result = sci_port->state_handlers->start_handler(sci_port);
+               if (result)
+                       return result;
+       }
+
+       scic_sds_controller_start_next_phy(scic);
+
+       isci_timer_start(scic->timeout_timer, timeout);
+
+       sci_base_state_machine_change_state(&scic->state_machine,
+                                           SCI_BASE_CONTROLLER_STATE_STARTING);
+
+       return SCI_SUCCESS;
+}
+
 void isci_host_scan_start(struct Scsi_Host *shost)
 {
        struct isci_host *ihost = SHOST_TO_SAS_HA(shost)->lldd_ha;
@@ -155,7 +1117,7 @@ void isci_host_scan_start(struct Scsi_Host *shost)
        spin_unlock_irq(&ihost->scic_lock);
 }
 
-void isci_host_stop_complete(struct isci_host *ihost, enum sci_status completion_status)
+static void isci_host_stop_complete(struct isci_host *ihost, enum sci_status completion_status)
 {
        isci_host_change_state(ihost, isci_stopped);
        scic_controller_disable_interrupts(&ihost->sci);
@@ -163,6 +1125,19 @@ void isci_host_stop_complete(struct isci_host *ihost, enum sci_status completion
        wake_up(&ihost->eventq);
 }
 
+static void scic_sds_controller_completion_handler(struct scic_sds_controller *scic)
+{
+       /* Empty out the completion queue */
+       if (scic_sds_controller_completion_queue_has_entries(scic))
+               scic_sds_controller_process_completions(scic);
+
+       /* Clear the interrupt and enable all interrupts again */
+       writel(SMU_ISR_COMPLETION, &scic->smu_registers->interrupt_status);
+       /* Could we write the value of SMU_ISR_COMPLETION? */
+       writel(0xFF000000, &scic->smu_registers->interrupt_mask);
+       writel(0, &scic->smu_registers->interrupt_mask);
+}
+
 /**
  * isci_host_completion_routine() - This function is the delayed service
  *    routine that calls the sci core library's completion handler. It's
@@ -273,6 +1248,75 @@ static void isci_host_completion_routine(unsigned long data)
 
 }
 
+/**
+ * scic_controller_stop() - This method will stop an individual controller
+ *    object.This method will invoke the associated user callback upon
+ *    completion.  The completion callback is called when the following
+ *    conditions are met: -# the method return status is SCI_SUCCESS. -# the
+ *    controller has been quiesced. This method will ensure that all IO
+ *    requests are quiesced, phys are stopped, and all additional operation by
+ *    the hardware is halted.
+ * @controller: the handle to the controller object to stop.
+ * @timeout: This parameter specifies the number of milliseconds in which the
+ *    stop operation should complete.
+ *
+ * The controller must be in the STARTED or STOPPED state. Indicate if the
+ * controller stop method succeeded or failed in some way. SCI_SUCCESS if the
+ * stop operation successfully began. SCI_WARNING_ALREADY_IN_STATE if the
+ * controller is already in the STOPPED state. SCI_FAILURE_INVALID_STATE if the
+ * controller is not either in the STARTED or STOPPED states.
+ */
+static enum sci_status scic_controller_stop(struct scic_sds_controller *scic,
+                                           u32 timeout)
+{
+       if (scic->state_machine.current_state_id !=
+           SCI_BASE_CONTROLLER_STATE_READY) {
+               dev_warn(scic_to_dev(scic),
+                        "SCIC Controller stop operation requested in "
+                        "invalid state\n");
+               return SCI_FAILURE_INVALID_STATE;
+       }
+
+       isci_timer_start(scic->timeout_timer, timeout);
+       sci_base_state_machine_change_state(&scic->state_machine,
+                                           SCI_BASE_CONTROLLER_STATE_STOPPING);
+       return SCI_SUCCESS;
+}
+
+/**
+ * scic_controller_reset() - This method will reset the supplied core
+ *    controller regardless of the state of said controller.  This operation is
+ *    considered destructive.  In other words, all current operations are wiped
+ *    out.  No IO completions for outstanding devices occur.  Outstanding IO
+ *    requests are not aborted or completed at the actual remote device.
+ * @controller: the handle to the controller object to reset.
+ *
+ * Indicate if the controller reset method succeeded or failed in some way.
+ * SCI_SUCCESS if the reset operation successfully started. SCI_FATAL_ERROR if
+ * the controller reset operation is unable to complete.
+ */
+static enum sci_status scic_controller_reset(struct scic_sds_controller *scic)
+{
+       switch (scic->state_machine.current_state_id) {
+       case SCI_BASE_CONTROLLER_STATE_RESET:
+       case SCI_BASE_CONTROLLER_STATE_READY:
+       case SCI_BASE_CONTROLLER_STATE_STOPPED:
+       case SCI_BASE_CONTROLLER_STATE_FAILED:
+               /*
+                * The reset operation is not a graceful cleanup, just
+                * perform the state transition.
+                */
+               sci_base_state_machine_change_state(&scic->state_machine,
+                               SCI_BASE_CONTROLLER_STATE_RESETTING);
+               return SCI_SUCCESS;
+       default:
+               dev_warn(scic_to_dev(scic),
+                        "SCIC Controller reset operation requested in "
+                        "invalid state\n");
+               return SCI_FAILURE_INVALID_STATE;
+       }
+}
+
 void isci_host_deinit(struct isci_host *ihost)
 {
        int i;
@@ -341,6 +1385,1109 @@ static void isci_user_parameters_get(
        u->max_number_concurrent_device_spin_up = max_concurr_spinup;
 }
 
+static void scic_sds_controller_initial_state_enter(void *object)
+{
+       struct scic_sds_controller *scic = object;
+
+       sci_base_state_machine_change_state(&scic->state_machine,
+                       SCI_BASE_CONTROLLER_STATE_RESET);
+}
+
+static inline void scic_sds_controller_starting_state_exit(void *object)
+{
+       struct scic_sds_controller *scic = object;
+
+       isci_timer_stop(scic->timeout_timer);
+}
+
+#define INTERRUPT_COALESCE_TIMEOUT_BASE_RANGE_LOWER_BOUND_NS 853
+#define INTERRUPT_COALESCE_TIMEOUT_BASE_RANGE_UPPER_BOUND_NS 1280
+#define INTERRUPT_COALESCE_TIMEOUT_MAX_US                    2700000
+#define INTERRUPT_COALESCE_NUMBER_MAX                        256
+#define INTERRUPT_COALESCE_TIMEOUT_ENCODE_MIN                7
+#define INTERRUPT_COALESCE_TIMEOUT_ENCODE_MAX                28
+
+/**
+ * scic_controller_set_interrupt_coalescence() - This method allows the user to
+ *    configure the interrupt coalescence.
+ * @controller: This parameter represents the handle to the controller object
+ *    for which its interrupt coalesce register is overridden.
+ * @coalesce_number: Used to control the number of entries in the Completion
+ *    Queue before an interrupt is generated. If the number of entries exceed
+ *    this number, an interrupt will be generated. The valid range of the input
+ *    is [0, 256]. A setting of 0 results in coalescing being disabled.
+ * @coalesce_timeout: Timeout value in microseconds. The valid range of the
+ *    input is [0, 2700000] . A setting of 0 is allowed and results in no
+ *    interrupt coalescing timeout.
+ *
+ * Indicate if the user successfully set the interrupt coalesce parameters.
+ * SCI_SUCCESS The user successfully updated the interrutp coalescence.
+ * SCI_FAILURE_INVALID_PARAMETER_VALUE The user input value is out of range.
+ */
+static enum sci_status scic_controller_set_interrupt_coalescence(
+       struct scic_sds_controller *scic_controller,
+       u32 coalesce_number,
+       u32 coalesce_timeout)
+{
+       u8 timeout_encode = 0;
+       u32 min = 0;
+       u32 max = 0;
+
+       /* Check if the input parameters fall in the range. */
+       if (coalesce_number > INTERRUPT_COALESCE_NUMBER_MAX)
+               return SCI_FAILURE_INVALID_PARAMETER_VALUE;
+
+       /*
+        *  Defined encoding for interrupt coalescing timeout:
+        *              Value   Min      Max     Units
+        *              -----   ---      ---     -----
+        *              0       -        -       Disabled
+        *              1       13.3     20.0    ns
+        *              2       26.7     40.0
+        *              3       53.3     80.0
+        *              4       106.7    160.0
+        *              5       213.3    320.0
+        *              6       426.7    640.0
+        *              7       853.3    1280.0
+        *              8       1.7      2.6     us
+        *              9       3.4      5.1
+        *              10      6.8      10.2
+        *              11      13.7     20.5
+        *              12      27.3     41.0
+        *              13      54.6     81.9
+        *              14      109.2    163.8
+        *              15      218.5    327.7
+        *              16      436.9    655.4
+        *              17      873.8    1310.7
+        *              18      1.7      2.6     ms
+        *              19      3.5      5.2
+        *              20      7.0      10.5
+        *              21      14.0     21.0
+        *              22      28.0     41.9
+        *              23      55.9     83.9
+        *              24      111.8    167.8
+        *              25      223.7    335.5
+        *              26      447.4    671.1
+        *              27      894.8    1342.2
+        *              28      1.8      2.7     s
+        *              Others Undefined */
+
+       /*
+        * Use the table above to decide the encode of interrupt coalescing timeout
+        * value for register writing. */
+       if (coalesce_timeout == 0)
+               timeout_encode = 0;
+       else{
+               /* make the timeout value in unit of (10 ns). */
+               coalesce_timeout = coalesce_timeout * 100;
+               min = INTERRUPT_COALESCE_TIMEOUT_BASE_RANGE_LOWER_BOUND_NS / 10;
+               max = INTERRUPT_COALESCE_TIMEOUT_BASE_RANGE_UPPER_BOUND_NS / 10;
+
+               /* get the encode of timeout for register writing. */
+               for (timeout_encode = INTERRUPT_COALESCE_TIMEOUT_ENCODE_MIN;
+                     timeout_encode <= INTERRUPT_COALESCE_TIMEOUT_ENCODE_MAX;
+                     timeout_encode++) {
+                       if (min <= coalesce_timeout &&  max > coalesce_timeout)
+                               break;
+                       else if (coalesce_timeout >= max && coalesce_timeout < min * 2
+                                && coalesce_timeout <= INTERRUPT_COALESCE_TIMEOUT_MAX_US * 100) {
+                               if ((coalesce_timeout - max) < (2 * min - coalesce_timeout))
+                                       break;
+                               else{
+                                       timeout_encode++;
+                                       break;
+                               }
+                       } else {
+                               max = max * 2;
+                               min = min * 2;
+                       }
+               }
+
+               if (timeout_encode == INTERRUPT_COALESCE_TIMEOUT_ENCODE_MAX + 1)
+                       /* the value is out of range. */
+                       return SCI_FAILURE_INVALID_PARAMETER_VALUE;
+       }
+
+       writel(SMU_ICC_GEN_VAL(NUMBER, coalesce_number) |
+              SMU_ICC_GEN_VAL(TIMER, timeout_encode),
+              &scic_controller->smu_registers->interrupt_coalesce_control);
+
+
+       scic_controller->interrupt_coalesce_number = (u16)coalesce_number;
+       scic_controller->interrupt_coalesce_timeout = coalesce_timeout / 100;
+
+       return SCI_SUCCESS;
+}
+
+
+static void scic_sds_controller_ready_state_enter(void *object)
+{
+       struct scic_sds_controller *scic = object;
+
+       /* set the default interrupt coalescence number and timeout value. */
+       scic_controller_set_interrupt_coalescence(scic, 0x10, 250);
+}
+
+static void scic_sds_controller_ready_state_exit(void *object)
+{
+       struct scic_sds_controller *scic = object;
+
+       /* disable interrupt coalescence. */
+       scic_controller_set_interrupt_coalescence(scic, 0, 0);
+}
+
+static enum sci_status scic_sds_controller_stop_phys(struct scic_sds_controller *scic)
+{
+       u32 index;
+       enum sci_status status;
+       enum sci_status phy_status;
+       struct isci_host *ihost = scic_to_ihost(scic);
+
+       status = SCI_SUCCESS;
+
+       for (index = 0; index < SCI_MAX_PHYS; index++) {
+               phy_status = scic_sds_phy_stop(&ihost->phys[index].sci);
+
+               if (phy_status != SCI_SUCCESS &&
+                   phy_status != SCI_FAILURE_INVALID_STATE) {
+                       status = SCI_FAILURE;
+
+                       dev_warn(scic_to_dev(scic),
+                                "%s: Controller stop operation failed to stop "
+                                "phy %d because of status %d.\n",
+                                __func__,
+                                ihost->phys[index].sci.phy_index, phy_status);
+               }
+       }
+
+       return status;
+}
+
+static enum sci_status scic_sds_controller_stop_ports(struct scic_sds_controller *scic)
+{
+       u32 index;
+       enum sci_status port_status;
+       enum sci_status status = SCI_SUCCESS;
+       struct isci_host *ihost = scic_to_ihost(scic);
+
+       for (index = 0; index < scic->logical_port_entries; index++) {
+               struct scic_sds_port *sci_port = &ihost->ports[index].sci;
+               scic_sds_port_handler_t stop;
+
+               stop = sci_port->state_handlers->stop_handler;
+               port_status = stop(sci_port);
+
+               if ((port_status != SCI_SUCCESS) &&
+                   (port_status != SCI_FAILURE_INVALID_STATE)) {
+                       status = SCI_FAILURE;
+
+                       dev_warn(scic_to_dev(scic),
+                                "%s: Controller stop operation failed to "
+                                "stop port %d because of status %d.\n",
+                                __func__,
+                                sci_port->logical_port_index,
+                                port_status);
+               }
+       }
+
+       return status;
+}
+
+static enum sci_status scic_sds_controller_stop_devices(struct scic_sds_controller *scic)
+{
+       u32 index;
+       enum sci_status status;
+       enum sci_status device_status;
+
+       status = SCI_SUCCESS;
+
+       for (index = 0; index < scic->remote_node_entries; index++) {
+               if (scic->device_table[index] != NULL) {
+                       /* / @todo What timeout value do we want to provide to this request? */
+                       device_status = scic_remote_device_stop(scic->device_table[index], 0);
+
+                       if ((device_status != SCI_SUCCESS) &&
+                           (device_status != SCI_FAILURE_INVALID_STATE)) {
+                               dev_warn(scic_to_dev(scic),
+                                        "%s: Controller stop operation failed "
+                                        "to stop device 0x%p because of "
+                                        "status %d.\n",
+                                        __func__,
+                                        scic->device_table[index], device_status);
+                       }
+               }
+       }
+
+       return status;
+}
+
+static void scic_sds_controller_stopping_state_enter(void *object)
+{
+       struct scic_sds_controller *scic = object;
+
+       /* Stop all of the components for this controller */
+       scic_sds_controller_stop_phys(scic);
+       scic_sds_controller_stop_ports(scic);
+       scic_sds_controller_stop_devices(scic);
+}
+
+static void scic_sds_controller_stopping_state_exit(void *object)
+{
+       struct scic_sds_controller *scic = object;
+
+       isci_timer_stop(scic->timeout_timer);
+}
+
+
+/**
+ * scic_sds_controller_reset_hardware() -
+ *
+ * This method will reset the controller hardware.
+ */
+static void scic_sds_controller_reset_hardware(struct scic_sds_controller *scic)
+{
+       /* Disable interrupts so we dont take any spurious interrupts */
+       scic_controller_disable_interrupts(scic);
+
+       /* Reset the SCU */
+       writel(0xFFFFFFFF, &scic->smu_registers->soft_reset_control);
+
+       /* Delay for 1ms to before clearing the CQP and UFQPR. */
+       udelay(1000);
+
+       /* The write to the CQGR clears the CQP */
+       writel(0x00000000, &scic->smu_registers->completion_queue_get);
+
+       /* The write to the UFQGP clears the UFQPR */
+       writel(0, &scic->scu_registers->sdma.unsolicited_frame_get_pointer);
+}
+
+static void scic_sds_controller_resetting_state_enter(void *object)
+{
+       struct scic_sds_controller *scic = object;
+
+       scic_sds_controller_reset_hardware(scic);
+       sci_base_state_machine_change_state(&scic->state_machine,
+                                           SCI_BASE_CONTROLLER_STATE_RESET);
+}
+
+static const struct sci_base_state scic_sds_controller_state_table[] = {
+       [SCI_BASE_CONTROLLER_STATE_INITIAL] = {
+               .enter_state = scic_sds_controller_initial_state_enter,
+       },
+       [SCI_BASE_CONTROLLER_STATE_RESET] = {},
+       [SCI_BASE_CONTROLLER_STATE_INITIALIZING] = {},
+       [SCI_BASE_CONTROLLER_STATE_INITIALIZED] = {},
+       [SCI_BASE_CONTROLLER_STATE_STARTING] = {
+               .exit_state  = scic_sds_controller_starting_state_exit,
+       },
+       [SCI_BASE_CONTROLLER_STATE_READY] = {
+               .enter_state = scic_sds_controller_ready_state_enter,
+               .exit_state  = scic_sds_controller_ready_state_exit,
+       },
+       [SCI_BASE_CONTROLLER_STATE_RESETTING] = {
+               .enter_state = scic_sds_controller_resetting_state_enter,
+       },
+       [SCI_BASE_CONTROLLER_STATE_STOPPING] = {
+               .enter_state = scic_sds_controller_stopping_state_enter,
+               .exit_state = scic_sds_controller_stopping_state_exit,
+       },
+       [SCI_BASE_CONTROLLER_STATE_STOPPED] = {},
+       [SCI_BASE_CONTROLLER_STATE_FAILED] = {}
+};
+
+static void scic_sds_controller_set_default_config_parameters(struct scic_sds_controller *scic)
+{
+       /* these defaults are overridden by the platform / firmware */
+       struct isci_host *ihost = scic_to_ihost(scic);
+       u16 index;
+
+       /* Default to APC mode. */
+       scic->oem_parameters.sds1.controller.mode_type = SCIC_PORT_AUTOMATIC_CONFIGURATION_MODE;
+
+       /* Default to APC mode. */
+       scic->oem_parameters.sds1.controller.max_concurrent_dev_spin_up = 1;
+
+       /* Default to no SSC operation. */
+       scic->oem_parameters.sds1.controller.do_enable_ssc = false;
+
+       /* Initialize all of the port parameter information to narrow ports. */
+       for (index = 0; index < SCI_MAX_PORTS; index++) {
+               scic->oem_parameters.sds1.ports[index].phy_mask = 0;
+       }
+
+       /* Initialize all of the phy parameter information. */
+       for (index = 0; index < SCI_MAX_PHYS; index++) {
+               /* Default to 6G (i.e. Gen 3) for now. */
+               scic->user_parameters.sds1.phys[index].max_speed_generation = 3;
+
+               /* the frequencies cannot be 0 */
+               scic->user_parameters.sds1.phys[index].align_insertion_frequency = 0x7f;
+               scic->user_parameters.sds1.phys[index].in_connection_align_insertion_frequency = 0xff;
+               scic->user_parameters.sds1.phys[index].notify_enable_spin_up_insertion_frequency = 0x33;
+
+               /*
+                * Previous Vitesse based expanders had a arbitration issue that
+                * is worked around by having the upper 32-bits of SAS address
+                * with a value greater then the Vitesse company identifier.
+                * Hence, usage of 0x5FCFFFFF. */
+               scic->oem_parameters.sds1.phys[index].sas_address.low = 0x1 + ihost->id;
+               scic->oem_parameters.sds1.phys[index].sas_address.high = 0x5FCFFFFF;
+       }
+
+       scic->user_parameters.sds1.stp_inactivity_timeout = 5;
+       scic->user_parameters.sds1.ssp_inactivity_timeout = 5;
+       scic->user_parameters.sds1.stp_max_occupancy_timeout = 5;
+       scic->user_parameters.sds1.ssp_max_occupancy_timeout = 20;
+       scic->user_parameters.sds1.no_outbound_task_timeout = 20;
+}
+
+
+
+/**
+ * scic_controller_construct() - This method will attempt to construct a
+ *    controller object utilizing the supplied parameter information.
+ * @c: This parameter specifies the controller to be constructed.
+ * @scu_base: mapped base address of the scu registers
+ * @smu_base: mapped base address of the smu registers
+ *
+ * Indicate if the controller was successfully constructed or if it failed in
+ * some way. SCI_SUCCESS This value is returned if the controller was
+ * successfully constructed. SCI_WARNING_TIMER_CONFLICT This value is returned
+ * if the interrupt coalescence timer may cause SAS compliance issues for SMP
+ * Target mode response processing. SCI_FAILURE_UNSUPPORTED_CONTROLLER_TYPE
+ * This value is returned if the controller does not support the supplied type.
+ * SCI_FAILURE_UNSUPPORTED_INIT_DATA_VERSION This value is returned if the
+ * controller does not support the supplied initialization data version.
+ */
+static enum sci_status scic_controller_construct(struct scic_sds_controller *scic,
+                                         void __iomem *scu_base,
+                                         void __iomem *smu_base)
+{
+       struct isci_host *ihost = scic_to_ihost(scic);
+       u8 i;
+
+       sci_base_state_machine_construct(&scic->state_machine,
+               scic, scic_sds_controller_state_table,
+               SCI_BASE_CONTROLLER_STATE_INITIAL);
+
+       sci_base_state_machine_start(&scic->state_machine);
+
+       scic->scu_registers = scu_base;
+       scic->smu_registers = smu_base;
+
+       scic_sds_port_configuration_agent_construct(&scic->port_agent);
+
+       /* Construct the ports for this controller */
+       for (i = 0; i < SCI_MAX_PORTS; i++)
+               scic_sds_port_construct(&ihost->ports[i].sci, i, scic);
+       scic_sds_port_construct(&ihost->ports[i].sci, SCIC_SDS_DUMMY_PORT, scic);
+
+       /* Construct the phys for this controller */
+       for (i = 0; i < SCI_MAX_PHYS; i++) {
+               /* Add all the PHYs to the dummy port */
+               scic_sds_phy_construct(&ihost->phys[i].sci,
+                                      &ihost->ports[SCI_MAX_PORTS].sci, i);
+       }
+
+       scic->invalid_phy_mask = 0;
+
+       /* Set the default maximum values */
+       scic->completion_event_entries      = SCU_EVENT_COUNT;
+       scic->completion_queue_entries      = SCU_COMPLETION_QUEUE_COUNT;
+       scic->remote_node_entries           = SCI_MAX_REMOTE_DEVICES;
+       scic->logical_port_entries          = SCI_MAX_PORTS;
+       scic->task_context_entries          = SCU_IO_REQUEST_COUNT;
+       scic->uf_control.buffers.count      = SCU_UNSOLICITED_FRAME_COUNT;
+       scic->uf_control.address_table.count = SCU_UNSOLICITED_FRAME_COUNT;
+
+       /* Initialize the User and OEM parameters to default values. */
+       scic_sds_controller_set_default_config_parameters(scic);
+
+       return scic_controller_reset(scic);
+}
+
+int scic_oem_parameters_validate(struct scic_sds_oem_params *oem)
+{
+       int i;
+
+       for (i = 0; i < SCI_MAX_PORTS; i++)
+               if (oem->ports[i].phy_mask > SCIC_SDS_PARM_PHY_MASK_MAX)
+                       return -EINVAL;
+
+       for (i = 0; i < SCI_MAX_PHYS; i++)
+               if (oem->phys[i].sas_address.high == 0 &&
+                   oem->phys[i].sas_address.low == 0)
+                       return -EINVAL;
+
+       if (oem->controller.mode_type == SCIC_PORT_AUTOMATIC_CONFIGURATION_MODE) {
+               for (i = 0; i < SCI_MAX_PHYS; i++)
+                       if (oem->ports[i].phy_mask != 0)
+                               return -EINVAL;
+       } else if (oem->controller.mode_type == SCIC_PORT_MANUAL_CONFIGURATION_MODE) {
+               u8 phy_mask = 0;
+
+               for (i = 0; i < SCI_MAX_PHYS; i++)
+                       phy_mask |= oem->ports[i].phy_mask;
+
+               if (phy_mask == 0)
+                       return -EINVAL;
+       } else
+               return -EINVAL;
+
+       if (oem->controller.max_concurrent_dev_spin_up > MAX_CONCURRENT_DEVICE_SPIN_UP_COUNT)
+               return -EINVAL;
+
+       return 0;
+}
+
+static enum sci_status scic_oem_parameters_set(struct scic_sds_controller *scic,
+                                       union scic_oem_parameters *scic_parms)
+{
+       u32 state = scic->state_machine.current_state_id;
+
+       if (state == SCI_BASE_CONTROLLER_STATE_RESET ||
+           state == SCI_BASE_CONTROLLER_STATE_INITIALIZING ||
+           state == SCI_BASE_CONTROLLER_STATE_INITIALIZED) {
+
+               if (scic_oem_parameters_validate(&scic_parms->sds1))
+                       return SCI_FAILURE_INVALID_PARAMETER_VALUE;
+               scic->oem_parameters.sds1 = scic_parms->sds1;
+
+               return SCI_SUCCESS;
+       }
+
+       return SCI_FAILURE_INVALID_STATE;
+}
+
+void scic_oem_parameters_get(
+       struct scic_sds_controller *scic,
+       union scic_oem_parameters *scic_parms)
+{
+       memcpy(scic_parms, (&scic->oem_parameters), sizeof(*scic_parms));
+}
+
+static void scic_sds_controller_timeout_handler(void *_scic)
+{
+       struct scic_sds_controller *scic = _scic;
+       struct isci_host *ihost = scic_to_ihost(scic);
+       struct sci_base_state_machine *sm = &scic->state_machine;
+
+       if (sm->current_state_id == SCI_BASE_CONTROLLER_STATE_STARTING)
+               scic_sds_controller_transition_to_ready(scic, SCI_FAILURE_TIMEOUT);
+       else if (sm->current_state_id == SCI_BASE_CONTROLLER_STATE_STOPPING) {
+               sci_base_state_machine_change_state(sm, SCI_BASE_CONTROLLER_STATE_FAILED);
+               isci_host_stop_complete(ihost, SCI_FAILURE_TIMEOUT);
+       } else  /* / @todo Now what do we want to do in this case? */
+               dev_err(scic_to_dev(scic),
+                       "%s: Controller timer fired when controller was not "
+                       "in a state being timed.\n",
+                       __func__);
+}
+
+static enum sci_status scic_sds_controller_initialize_phy_startup(struct scic_sds_controller *scic)
+{
+       struct isci_host *ihost = scic_to_ihost(scic);
+
+       scic->phy_startup_timer = isci_timer_create(ihost,
+                                                   scic,
+                                                   scic_sds_controller_phy_startup_timeout_handler);
+
+       if (scic->phy_startup_timer == NULL)
+               return SCI_FAILURE_INSUFFICIENT_RESOURCES;
+       else {
+               scic->next_phy_to_start = 0;
+               scic->phy_startup_timer_pending = false;
+       }
+
+       return SCI_SUCCESS;
+}
+
+static void scic_sds_controller_power_control_timer_start(struct scic_sds_controller *scic)
+{
+       isci_timer_start(scic->power_control.timer,
+                        SCIC_SDS_CONTROLLER_POWER_CONTROL_INTERVAL);
+
+       scic->power_control.timer_started = true;
+}
+
+static void scic_sds_controller_power_control_timer_stop(struct scic_sds_controller *scic)
+{
+       if (scic->power_control.timer_started) {
+               isci_timer_stop(scic->power_control.timer);
+               scic->power_control.timer_started = false;
+       }
+}
+
+static void scic_sds_controller_power_control_timer_restart(struct scic_sds_controller *scic)
+{
+       scic_sds_controller_power_control_timer_stop(scic);
+       scic_sds_controller_power_control_timer_start(scic);
+}
+
+static void scic_sds_controller_power_control_timer_handler(
+       void *controller)
+{
+       struct scic_sds_controller *scic;
+
+       scic = (struct scic_sds_controller *)controller;
+
+       scic->power_control.phys_granted_power = 0;
+
+       if (scic->power_control.phys_waiting == 0) {
+               scic->power_control.timer_started = false;
+       } else {
+               struct scic_sds_phy *sci_phy = NULL;
+               u8 i;
+
+               for (i = 0;
+                    (i < SCI_MAX_PHYS)
+                    && (scic->power_control.phys_waiting != 0);
+                    i++) {
+                       if (scic->power_control.requesters[i] != NULL) {
+                               if (scic->power_control.phys_granted_power <
+                                   scic->oem_parameters.sds1.controller.max_concurrent_dev_spin_up) {
+                                       sci_phy = scic->power_control.requesters[i];
+                                       scic->power_control.requesters[i] = NULL;
+                                       scic->power_control.phys_waiting--;
+                                       scic->power_control.phys_granted_power++;
+                                       scic_sds_phy_consume_power_handler(sci_phy);
+                               } else {
+                                       break;
+                               }
+                       }
+               }
+
+               /*
+                * It doesn't matter if the power list is empty, we need to start the
+                * timer in case another phy becomes ready.
+                */
+               scic_sds_controller_power_control_timer_start(scic);
+       }
+}
+
+/**
+ * This method inserts the phy in the stagger spinup control queue.
+ * @scic:
+ *
+ *
+ */
+void scic_sds_controller_power_control_queue_insert(
+       struct scic_sds_controller *scic,
+       struct scic_sds_phy *sci_phy)
+{
+       BUG_ON(sci_phy == NULL);
+
+       if (scic->power_control.phys_granted_power <
+           scic->oem_parameters.sds1.controller.max_concurrent_dev_spin_up) {
+               scic->power_control.phys_granted_power++;
+               scic_sds_phy_consume_power_handler(sci_phy);
+
+               /*
+                * stop and start the power_control timer. When the timer fires, the
+                * no_of_phys_granted_power will be set to 0
+                */
+               scic_sds_controller_power_control_timer_restart(scic);
+       } else {
+               /* Add the phy in the waiting list */
+               scic->power_control.requesters[sci_phy->phy_index] = sci_phy;
+               scic->power_control.phys_waiting++;
+       }
+}
+
+/**
+ * This method removes the phy from the stagger spinup control queue.
+ * @scic:
+ *
+ *
+ */
+void scic_sds_controller_power_control_queue_remove(
+       struct scic_sds_controller *scic,
+       struct scic_sds_phy *sci_phy)
+{
+       BUG_ON(sci_phy == NULL);
+
+       if (scic->power_control.requesters[sci_phy->phy_index] != NULL) {
+               scic->power_control.phys_waiting--;
+       }
+
+       scic->power_control.requesters[sci_phy->phy_index] = NULL;
+}
+
+#define AFE_REGISTER_WRITE_DELAY 10
+
+/* Initialize the AFE for this phy index. We need to read the AFE setup from
+ * the OEM parameters
+ */
+static void scic_sds_controller_afe_initialization(struct scic_sds_controller *scic)
+{
+       const struct scic_sds_oem_params *oem = &scic->oem_parameters.sds1;
+       u32 afe_status;
+       u32 phy_id;
+
+       /* Clear DFX Status registers */
+       writel(0x0081000f, &scic->scu_registers->afe.afe_dfx_master_control0);
+       udelay(AFE_REGISTER_WRITE_DELAY);
+
+       if (is_b0()) {
+               /* PM Rx Equalization Save, PM SPhy Rx Acknowledgement
+                * Timer, PM Stagger Timer */
+               writel(0x0007BFFF, &scic->scu_registers->afe.afe_pmsn_master_control2);
+               udelay(AFE_REGISTER_WRITE_DELAY);
+       }
+
+       /* Configure bias currents to normal */
+       if (is_a0())
+               writel(0x00005500, &scic->scu_registers->afe.afe_bias_control);
+       else if (is_a2())
+               writel(0x00005A00, &scic->scu_registers->afe.afe_bias_control);
+       else if (is_b0())
+               writel(0x00005F00, &scic->scu_registers->afe.afe_bias_control);
+
+       udelay(AFE_REGISTER_WRITE_DELAY);
+
+       /* Enable PLL */
+       if (is_b0())
+               writel(0x80040A08, &scic->scu_registers->afe.afe_pll_control0);
+       else
+               writel(0x80040908, &scic->scu_registers->afe.afe_pll_control0);
+
+       udelay(AFE_REGISTER_WRITE_DELAY);
+
+       /* Wait for the PLL to lock */
+       do {
+               afe_status = readl(&scic->scu_registers->afe.afe_common_block_status);
+               udelay(AFE_REGISTER_WRITE_DELAY);
+       } while ((afe_status & 0x00001000) == 0);
+
+       if (is_a0() || is_a2()) {
+               /* Shorten SAS SNW lock time (RxLock timer value from 76 us to 50 us) */
+               writel(0x7bcc96ad, &scic->scu_registers->afe.afe_pmsn_master_control0);
+               udelay(AFE_REGISTER_WRITE_DELAY);
+       }
+
+       for (phy_id = 0; phy_id < SCI_MAX_PHYS; phy_id++) {
+               const struct sci_phy_oem_params *oem_phy = &oem->phys[phy_id];
+
+               if (is_b0()) {
+                        /* Configure transmitter SSC parameters */
+                       writel(0x00030000, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_ssc_control);
+                       udelay(AFE_REGISTER_WRITE_DELAY);
+               } else {
+                       /*
+                        * All defaults, except the Receive Word Alignament/Comma Detect
+                        * Enable....(0xe800) */
+                       writel(0x00004512, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_xcvr_control0);
+                       udelay(AFE_REGISTER_WRITE_DELAY);
+
+                       writel(0x0050100F, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_xcvr_control1);
+                       udelay(AFE_REGISTER_WRITE_DELAY);
+               }
+
+               /*
+                * Power up TX and RX out from power down (PWRDNTX and PWRDNRX)
+                * & increase TX int & ext bias 20%....(0xe85c) */
+               if (is_a0())
+                       writel(0x000003D4, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_channel_control);
+               else if (is_a2())
+                       writel(0x000003F0, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_channel_control);
+               else {
+                        /* Power down TX and RX (PWRDNTX and PWRDNRX) */
+                       writel(0x000003d7, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_channel_control);
+                       udelay(AFE_REGISTER_WRITE_DELAY);
+
+                       /*
+                        * Power up TX and RX out from power down (PWRDNTX and PWRDNRX)
+                        * & increase TX int & ext bias 20%....(0xe85c) */
+                       writel(0x000003d4, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_channel_control);
+               }
+               udelay(AFE_REGISTER_WRITE_DELAY);
+
+               if (is_a0() || is_a2()) {
+                       /* Enable TX equalization (0xe824) */
+                       writel(0x00040000, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_control);
+                       udelay(AFE_REGISTER_WRITE_DELAY);
+               }
+
+               /*
+                * RDPI=0x0(RX Power On), RXOOBDETPDNC=0x0, TPD=0x0(TX Power On),
+                * RDD=0x0(RX Detect Enabled) ....(0xe800) */
+               writel(0x00004100, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_xcvr_control0);
+               udelay(AFE_REGISTER_WRITE_DELAY);
+
+               /* Leave DFE/FFE on */
+               if (is_a0())
+                       writel(0x3F09983F, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_rx_ssc_control0);
+               else if (is_a2())
+                       writel(0x3F11103F, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_rx_ssc_control0);
+               else {
+                       writel(0x3F11103F, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_rx_ssc_control0);
+                       udelay(AFE_REGISTER_WRITE_DELAY);
+                       /* Enable TX equalization (0xe824) */
+                       writel(0x00040000, &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_control);
+               }
+               udelay(AFE_REGISTER_WRITE_DELAY);
+
+               writel(oem_phy->afe_tx_amp_control0,
+                       &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_amp_control0);
+               udelay(AFE_REGISTER_WRITE_DELAY);
+
+               writel(oem_phy->afe_tx_amp_control1,
+                       &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_amp_control1);
+               udelay(AFE_REGISTER_WRITE_DELAY);
+
+               writel(oem_phy->afe_tx_amp_control2,
+                       &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_amp_control2);
+               udelay(AFE_REGISTER_WRITE_DELAY);
+
+               writel(oem_phy->afe_tx_amp_control3,
+                       &scic->scu_registers->afe.scu_afe_xcvr[phy_id].afe_tx_amp_control3);
+               udelay(AFE_REGISTER_WRITE_DELAY);
+       }
+
+       /* Transfer control to the PEs */
+       writel(0x00010f00, &scic->scu_registers->afe.afe_dfx_master_control0);
+       udelay(AFE_REGISTER_WRITE_DELAY);
+}
+
+static enum sci_status scic_controller_set_mode(struct scic_sds_controller *scic,
+                                               enum sci_controller_mode operating_mode)
+{
+       enum sci_status status          = SCI_SUCCESS;
+
+       if ((scic->state_machine.current_state_id ==
+                               SCI_BASE_CONTROLLER_STATE_INITIALIZING) ||
+           (scic->state_machine.current_state_id ==
+                               SCI_BASE_CONTROLLER_STATE_INITIALIZED)) {
+               switch (operating_mode) {
+               case SCI_MODE_SPEED:
+                       scic->remote_node_entries      = SCI_MAX_REMOTE_DEVICES;
+                       scic->task_context_entries     = SCU_IO_REQUEST_COUNT;
+                       scic->uf_control.buffers.count =
+                               SCU_UNSOLICITED_FRAME_COUNT;
+                       scic->completion_event_entries = SCU_EVENT_COUNT;
+                       scic->completion_queue_entries =
+                               SCU_COMPLETION_QUEUE_COUNT;
+                       break;
+
+               case SCI_MODE_SIZE:
+                       scic->remote_node_entries      = SCI_MIN_REMOTE_DEVICES;
+                       scic->task_context_entries     = SCI_MIN_IO_REQUESTS;
+                       scic->uf_control.buffers.count =
+                               SCU_MIN_UNSOLICITED_FRAMES;
+                       scic->completion_event_entries = SCU_MIN_EVENTS;
+                       scic->completion_queue_entries =
+                               SCU_MIN_COMPLETION_QUEUE_ENTRIES;
+                       break;
+
+               default:
+                       status = SCI_FAILURE_INVALID_PARAMETER_VALUE;
+                       break;
+               }
+       } else
+               status = SCI_FAILURE_INVALID_STATE;
+
+       return status;
+}
+
+static void scic_sds_controller_initialize_power_control(struct scic_sds_controller *scic)
+{
+       struct isci_host *ihost = scic_to_ihost(scic);
+       scic->power_control.timer = isci_timer_create(ihost,
+                                                     scic,
+                                       scic_sds_controller_power_control_timer_handler);
+
+       memset(scic->power_control.requesters, 0,
+              sizeof(scic->power_control.requesters));
+
+       scic->power_control.phys_waiting = 0;
+       scic->power_control.phys_granted_power = 0;
+}
+
+static enum sci_status scic_controller_initialize(struct scic_sds_controller *scic)
+{
+       struct sci_base_state_machine *sm = &scic->state_machine;
+       enum sci_status result = SCI_SUCCESS;
+       struct isci_host *ihost = scic_to_ihost(scic);
+       u32 index, state;
+
+       if (scic->state_machine.current_state_id !=
+           SCI_BASE_CONTROLLER_STATE_RESET) {
+               dev_warn(scic_to_dev(scic),
+                        "SCIC Controller initialize operation requested "
+                        "in invalid state\n");
+               return SCI_FAILURE_INVALID_STATE;
+       }
+
+       sci_base_state_machine_change_state(sm, SCI_BASE_CONTROLLER_STATE_INITIALIZING);
+
+       scic->timeout_timer = isci_timer_create(ihost, scic,
+                                               scic_sds_controller_timeout_handler);
+
+       scic_sds_controller_initialize_phy_startup(scic);
+
+       scic_sds_controller_initialize_power_control(scic);
+
+       /*
+        * There is nothing to do here for B0 since we do not have to
+        * program the AFE registers.
+        * / @todo The AFE settings are supposed to be correct for the B0 but
+        * /       presently they seem to be wrong. */
+       scic_sds_controller_afe_initialization(scic);
+
+       if (result == SCI_SUCCESS) {
+               u32 status;
+               u32 terminate_loop;
+
+               /* Take the hardware out of reset */
+               writel(0, &scic->smu_registers->soft_reset_control);
+
+               /*
+                * / @todo Provide meaningfull error code for hardware failure
+                * result = SCI_FAILURE_CONTROLLER_HARDWARE; */
+               result = SCI_FAILURE;
+               terminate_loop = 100;
+
+               while (terminate_loop-- && (result != SCI_SUCCESS)) {
+                       /* Loop until the hardware reports success */
+                       udelay(SCU_CONTEXT_RAM_INIT_STALL_TIME);
+                       status = readl(&scic->smu_registers->control_status);
+
+                       if ((status & SCU_RAM_INIT_COMPLETED) ==
+                                       SCU_RAM_INIT_COMPLETED)
+                               result = SCI_SUCCESS;
+               }
+       }
+
+       if (result == SCI_SUCCESS) {
+               u32 max_supported_ports;
+               u32 max_supported_devices;
+               u32 max_supported_io_requests;
+               u32 device_context_capacity;
+
+               /*
+                * Determine what are the actaul device capacities that the
+                * hardware will support */
+               device_context_capacity =
+                       readl(&scic->smu_registers->device_context_capacity);
+
+
+               max_supported_ports = smu_dcc_get_max_ports(device_context_capacity);
+               max_supported_devices = smu_dcc_get_max_remote_node_context(device_context_capacity);
+               max_supported_io_requests = smu_dcc_get_max_task_context(device_context_capacity);
+
+               /*
+                * Make all PEs that are unassigned match up with the
+                * logical ports
+                */
+               for (index = 0; index < max_supported_ports; index++) {
+                       struct scu_port_task_scheduler_group_registers __iomem
+                               *ptsg = &scic->scu_registers->peg0.ptsg;
+
+                       writel(index, &ptsg->protocol_engine[index]);
+               }
+
+               /* Record the smaller of the two capacity values */
+               scic->logical_port_entries =
+                       min(max_supported_ports, scic->logical_port_entries);
+
+               scic->task_context_entries =
+                       min(max_supported_io_requests,
+                           scic->task_context_entries);
+
+               scic->remote_node_entries =
+                       min(max_supported_devices, scic->remote_node_entries);
+
+               /*
+                * Now that we have the correct hardware reported minimum values
+                * build the MDL for the controller.  Default to a performance
+                * configuration.
+                */
+               scic_controller_set_mode(scic, SCI_MODE_SPEED);
+       }
+
+       /* Initialize hardware PCI Relaxed ordering in DMA engines */
+       if (result == SCI_SUCCESS) {
+               u32 dma_configuration;
+
+               /* Configure the payload DMA */
+               dma_configuration =
+                       readl(&scic->scu_registers->sdma.pdma_configuration);
+               dma_configuration |=
+                       SCU_PDMACR_GEN_BIT(PCI_RELAXED_ORDERING_ENABLE);
+               writel(dma_configuration,
+                       &scic->scu_registers->sdma.pdma_configuration);
+
+               /* Configure the control DMA */
+               dma_configuration =
+                       readl(&scic->scu_registers->sdma.cdma_configuration);
+               dma_configuration |=
+                       SCU_CDMACR_GEN_BIT(PCI_RELAXED_ORDERING_ENABLE);
+               writel(dma_configuration,
+                       &scic->scu_registers->sdma.cdma_configuration);
+       }
+
+       /*
+        * Initialize the PHYs before the PORTs because the PHY registers
+        * are accessed during the port initialization.
+        */
+       if (result == SCI_SUCCESS) {
+               /* Initialize the phys */
+               for (index = 0;
+                    (result == SCI_SUCCESS) && (index < SCI_MAX_PHYS);
+                    index++) {
+                       result = scic_sds_phy_initialize(
+                               &ihost->phys[index].sci,
+                               &scic->scu_registers->peg0.pe[index].tl,
+                               &scic->scu_registers->peg0.pe[index].ll);
+               }
+       }
+
+       if (result == SCI_SUCCESS) {
+               /* Initialize the logical ports */
+               for (index = 0;
+                    (index < scic->logical_port_entries) &&
+                    (result == SCI_SUCCESS);
+                    index++) {
+                       result = scic_sds_port_initialize(
+                               &ihost->ports[index].sci,
+                               &scic->scu_registers->peg0.ptsg.port[index],
+                               &scic->scu_registers->peg0.ptsg.protocol_engine,
+                               &scic->scu_registers->peg0.viit[index]);
+               }
+       }
+
+       if (result == SCI_SUCCESS)
+               result = scic_sds_port_configuration_agent_initialize(
+                               scic,
+                               &scic->port_agent);
+
+       /* Advance the controller state machine */
+       if (result == SCI_SUCCESS)
+               state = SCI_BASE_CONTROLLER_STATE_INITIALIZED;
+       else
+               state = SCI_BASE_CONTROLLER_STATE_FAILED;
+       sci_base_state_machine_change_state(sm, state);
+
+       return result;
+}
+
+static enum sci_status scic_user_parameters_set(
+       struct scic_sds_controller *scic,
+       union scic_user_parameters *scic_parms)
+{
+       u32 state = scic->state_machine.current_state_id;
+
+       if (state == SCI_BASE_CONTROLLER_STATE_RESET ||
+           state == SCI_BASE_CONTROLLER_STATE_INITIALIZING ||
+           state == SCI_BASE_CONTROLLER_STATE_INITIALIZED) {
+               u16 index;
+
+               /*
+                * Validate the user parameters.  If they are not legal, then
+                * return a failure.
+                */
+               for (index = 0; index < SCI_MAX_PHYS; index++) {
+                       struct sci_phy_user_params *user_phy;
+
+                       user_phy = &scic_parms->sds1.phys[index];
+
+                       if (!((user_phy->max_speed_generation <=
+                                               SCIC_SDS_PARM_MAX_SPEED) &&
+                             (user_phy->max_speed_generation >
+                                               SCIC_SDS_PARM_NO_SPEED)))
+                               return SCI_FAILURE_INVALID_PARAMETER_VALUE;
+
+                       if (user_phy->in_connection_align_insertion_frequency <
+                                       3)
+                               return SCI_FAILURE_INVALID_PARAMETER_VALUE;
+
+                       if ((user_phy->in_connection_align_insertion_frequency <
+                                               3) ||
+                           (user_phy->align_insertion_frequency == 0) ||
+                           (user_phy->
+                               notify_enable_spin_up_insertion_frequency ==
+                                               0))
+                               return SCI_FAILURE_INVALID_PARAMETER_VALUE;
+               }
+
+               if ((scic_parms->sds1.stp_inactivity_timeout == 0) ||
+                   (scic_parms->sds1.ssp_inactivity_timeout == 0) ||
+                   (scic_parms->sds1.stp_max_occupancy_timeout == 0) ||
+                   (scic_parms->sds1.ssp_max_occupancy_timeout == 0) ||
+                   (scic_parms->sds1.no_outbound_task_timeout == 0))
+                       return SCI_FAILURE_INVALID_PARAMETER_VALUE;
+
+               memcpy(&scic->user_parameters, scic_parms, sizeof(*scic_parms));
+
+               return SCI_SUCCESS;
+       }
+
+       return SCI_FAILURE_INVALID_STATE;
+}
+
+static int scic_controller_mem_init(struct scic_sds_controller *scic)
+{
+       struct device *dev = scic_to_dev(scic);
+       dma_addr_t dma_handle;
+       enum sci_status result;
+
+       scic->completion_queue = dmam_alloc_coherent(dev,
+                       scic->completion_queue_entries * sizeof(u32),
+                       &dma_handle, GFP_KERNEL);
+       if (!scic->completion_queue)
+               return -ENOMEM;
+
+       writel(lower_32_bits(dma_handle),
+               &scic->smu_registers->completion_queue_lower);
+       writel(upper_32_bits(dma_handle),
+               &scic->smu_registers->completion_queue_upper);
+
+       scic->remote_node_context_table = dmam_alloc_coherent(dev,
+                       scic->remote_node_entries *
+                               sizeof(union scu_remote_node_context),
+                       &dma_handle, GFP_KERNEL);
+       if (!scic->remote_node_context_table)
+               return -ENOMEM;
+
+       writel(lower_32_bits(dma_handle),
+               &scic->smu_registers->remote_node_context_lower);
+       writel(upper_32_bits(dma_handle),
+               &scic->smu_registers->remote_node_context_upper);
+
+       scic->task_context_table = dmam_alloc_coherent(dev,
+                       scic->task_context_entries *
+                               sizeof(struct scu_task_context),
+                       &dma_handle, GFP_KERNEL);
+       if (!scic->task_context_table)
+               return -ENOMEM;
+
+       writel(lower_32_bits(dma_handle),
+               &scic->smu_registers->host_task_table_lower);
+       writel(upper_32_bits(dma_handle),
+               &scic->smu_registers->host_task_table_upper);
+
+       result = scic_sds_unsolicited_frame_control_construct(scic);
+       if (result)
+               return result;
+
+       /*
+        * Inform the silicon as to the location of the UF headers and
+        * address table.
+        */
+       writel(lower_32_bits(scic->uf_control.headers.physical_address),
+               &scic->scu_registers->sdma.uf_header_base_address_lower);
+       writel(upper_32_bits(scic->uf_control.headers.physical_address),
+               &scic->scu_registers->sdma.uf_header_base_address_upper);
+
+       writel(lower_32_bits(scic->uf_control.address_table.physical_address),
+               &scic->scu_registers->sdma.uf_address_table_lower);
+       writel(upper_32_bits(scic->uf_control.address_table.physical_address),
+               &scic->scu_registers->sdma.uf_address_table_upper);
+
+       return 0;
+}
+
 int isci_host_init(struct isci_host *isci_host)
 {
        int err = 0, i;
@@ -453,3 +2600,625 @@ int isci_host_init(struct isci_host *isci_host)
 
        return 0;
 }
+
+void scic_sds_controller_link_up(struct scic_sds_controller *scic,
+               struct scic_sds_port *port, struct scic_sds_phy *phy)
+{
+       switch (scic->state_machine.current_state_id) {
+       case SCI_BASE_CONTROLLER_STATE_STARTING:
+               scic_sds_controller_phy_timer_stop(scic);
+               scic->port_agent.link_up_handler(scic, &scic->port_agent,
+                                                port, phy);
+               scic_sds_controller_start_next_phy(scic);
+               break;
+       case SCI_BASE_CONTROLLER_STATE_READY:
+               scic->port_agent.link_up_handler(scic, &scic->port_agent,
+                                                port, phy);
+               break;
+       default:
+               dev_dbg(scic_to_dev(scic),
+                       "%s: SCIC Controller linkup event from phy %d in "
+                       "unexpected state %d\n", __func__, phy->phy_index,
+                       scic->state_machine.current_state_id);
+       }
+}
+
+void scic_sds_controller_link_down(struct scic_sds_controller *scic,
+               struct scic_sds_port *port, struct scic_sds_phy *phy)
+{
+       switch (scic->state_machine.current_state_id) {
+       case SCI_BASE_CONTROLLER_STATE_STARTING:
+       case SCI_BASE_CONTROLLER_STATE_READY:
+               scic->port_agent.link_down_handler(scic, &scic->port_agent,
+                                                  port, phy);
+               break;
+       default:
+               dev_dbg(scic_to_dev(scic),
+                       "%s: SCIC Controller linkdown event from phy %d in "
+                       "unexpected state %d\n",
+                       __func__,
+                       phy->phy_index,
+                       scic->state_machine.current_state_id);
+       }
+}
+
+/**
+ * This is a helper method to determine if any remote devices on this
+ * controller are still in the stopping state.
+ *
+ */
+static bool scic_sds_controller_has_remote_devices_stopping(
+       struct scic_sds_controller *controller)
+{
+       u32 index;
+
+       for (index = 0; index < controller->remote_node_entries; index++) {
+               if ((controller->device_table[index] != NULL) &&
+                  (controller->device_table[index]->state_machine.current_state_id
+                   == SCI_BASE_REMOTE_DEVICE_STATE_STOPPING))
+                       return true;
+       }
+
+       return false;
+}
+
+/**
+ * This method is called by the remote device to inform the controller
+ * object that the remote device has stopped.
+ */
+void scic_sds_controller_remote_device_stopped(struct scic_sds_controller *scic,
+                                              struct scic_sds_remote_device *sci_dev)
+{
+       if (scic->state_machine.current_state_id !=
+           SCI_BASE_CONTROLLER_STATE_STOPPING) {
+               dev_dbg(scic_to_dev(scic),
+                       "SCIC Controller 0x%p remote device stopped event "
+                       "from device 0x%p in unexpected state %d\n",
+                       scic, sci_dev,
+                       scic->state_machine.current_state_id);
+               return;
+       }
+
+       if (!scic_sds_controller_has_remote_devices_stopping(scic)) {
+               sci_base_state_machine_change_state(&scic->state_machine,
+                               SCI_BASE_CONTROLLER_STATE_STOPPED);
+       }
+}
+
+/**
+ * This method will write to the SCU PCP register the request value. The method
+ *    is used to suspend/resume ports, devices, and phys.
+ * @scic:
+ *
+ *
+ */
+void scic_sds_controller_post_request(
+       struct scic_sds_controller *scic,
+       u32 request)
+{
+       dev_dbg(scic_to_dev(scic),
+               "%s: SCIC Controller 0x%p post request 0x%08x\n",
+               __func__,
+               scic,
+               request);
+
+       writel(request, &scic->smu_registers->post_context_port);
+}
+
+/**
+ * This method will copy the soft copy of the task context into the physical
+ *    memory accessible by the controller.
+ * @scic: This parameter specifies the controller for which to copy
+ *    the task context.
+ * @sci_req: This parameter specifies the request for which the task
+ *    context is being copied.
+ *
+ * After this call is made the SCIC_SDS_IO_REQUEST object will always point to
+ * the physical memory version of the task context. Thus, all subsequent
+ * updates to the task context are performed in the TC table (i.e. DMAable
+ * memory). none
+ */
+void scic_sds_controller_copy_task_context(
+       struct scic_sds_controller *scic,
+       struct scic_sds_request *sci_req)
+{
+       struct scu_task_context *task_context_buffer;
+
+       task_context_buffer = scic_sds_controller_get_task_context_buffer(
+               scic, sci_req->io_tag);
+
+       memcpy(task_context_buffer,
+              sci_req->task_context_buffer,
+              offsetof(struct scu_task_context, sgl_snapshot_ac));
+
+       /*
+        * Now that the soft copy of the TC has been copied into the TC
+        * table accessible by the silicon.  Thus, any further changes to
+        * the TC (e.g. TC termination) occur in the appropriate location. */
+       sci_req->task_context_buffer = task_context_buffer;
+}
+
+/**
+ * This method returns the task context buffer for the given io tag.
+ * @scic:
+ * @io_tag:
+ *
+ * struct scu_task_context*
+ */
+struct scu_task_context *scic_sds_controller_get_task_context_buffer(
+       struct scic_sds_controller *scic,
+       u16 io_tag
+       ) {
+       u16 task_index = scic_sds_io_tag_get_index(io_tag);
+
+       if (task_index < scic->task_context_entries) {
+               return &scic->task_context_table[task_index];
+       }
+
+       return NULL;
+}
+
+struct scic_sds_request *scic_request_by_tag(struct scic_sds_controller *scic,
+                                            u16 io_tag)
+{
+       u16 task_index;
+       u16 task_sequence;
+
+       task_index = scic_sds_io_tag_get_index(io_tag);
+
+       if (task_index  < scic->task_context_entries) {
+               if (scic->io_request_table[task_index] != NULL) {
+                       task_sequence = scic_sds_io_tag_get_sequence(io_tag);
+
+                       if (task_sequence == scic->io_request_sequence[task_index]) {
+                               return scic->io_request_table[task_index];
+                       }
+               }
+       }
+
+       return NULL;
+}
+
+/**
+ * This method allocates remote node index and the reserves the remote node
+ *    context space for use. This method can fail if there are no more remote
+ *    node index available.
+ * @scic: This is the controller object which contains the set of
+ *    free remote node ids
+ * @sci_dev: This is the device object which is requesting the a remote node
+ *    id
+ * @node_id: This is the remote node id that is assinged to the device if one
+ *    is available
+ *
+ * enum sci_status SCI_FAILURE_OUT_OF_RESOURCES if there are no available remote
+ * node index available.
+ */
+enum sci_status scic_sds_controller_allocate_remote_node_context(
+       struct scic_sds_controller *scic,
+       struct scic_sds_remote_device *sci_dev,
+       u16 *node_id)
+{
+       u16 node_index;
+       u32 remote_node_count = scic_sds_remote_device_node_count(sci_dev);
+
+       node_index = scic_sds_remote_node_table_allocate_remote_node(
+               &scic->available_remote_nodes, remote_node_count
+               );
+
+       if (node_index != SCIC_SDS_REMOTE_NODE_CONTEXT_INVALID_INDEX) {
+               scic->device_table[node_index] = sci_dev;
+
+               *node_id = node_index;
+
+               return SCI_SUCCESS;
+       }
+
+       return SCI_FAILURE_INSUFFICIENT_RESOURCES;
+}
+
+/**
+ * This method frees the remote node index back to the available pool.  Once
+ *    this is done the remote node context buffer is no longer valid and can
+ *    not be used.
+ * @scic:
+ * @sci_dev:
+ * @node_id:
+ *
+ */
+void scic_sds_controller_free_remote_node_context(
+       struct scic_sds_controller *scic,
+       struct scic_sds_remote_device *sci_dev,
+       u16 node_id)
+{
+       u32 remote_node_count = scic_sds_remote_device_node_count(sci_dev);
+
+       if (scic->device_table[node_id] == sci_dev) {
+               scic->device_table[node_id] = NULL;
+
+               scic_sds_remote_node_table_release_remote_node_index(
+                       &scic->available_remote_nodes, remote_node_count, node_id
+                       );
+       }
+}
+
+/**
+ * This method returns the union scu_remote_node_context for the specified remote
+ *    node id.
+ * @scic:
+ * @node_id:
+ *
+ * union scu_remote_node_context*
+ */
+union scu_remote_node_context *scic_sds_controller_get_remote_node_context_buffer(
+       struct scic_sds_controller *scic,
+       u16 node_id
+       ) {
+       if (
+               (node_id < scic->remote_node_entries)
+               && (scic->device_table[node_id] != NULL)
+               ) {
+               return &scic->remote_node_context_table[node_id];
+       }
+
+       return NULL;
+}
+
+/**
+ *
+ * @resposne_buffer: This is the buffer into which the D2H register FIS will be
+ *    constructed.
+ * @frame_header: This is the frame header returned by the hardware.
+ * @frame_buffer: This is the frame buffer returned by the hardware.
+ *
+ * This method will combind the frame header and frame buffer to create a SATA
+ * D2H register FIS none
+ */
+void scic_sds_controller_copy_sata_response(
+       void *response_buffer,
+       void *frame_header,
+       void *frame_buffer)
+{
+       memcpy(response_buffer, frame_header, sizeof(u32));
+
+       memcpy(response_buffer + sizeof(u32),
+              frame_buffer,
+              sizeof(struct dev_to_host_fis) - sizeof(u32));
+}
+
+/**
+ * This method releases the frame once this is done the frame is available for
+ *    re-use by the hardware.  The data contained in the frame header and frame
+ *    buffer is no longer valid. The UF queue get pointer is only updated if UF
+ *    control indicates this is appropriate.
+ * @scic:
+ * @frame_index:
+ *
+ */
+void scic_sds_controller_release_frame(
+       struct scic_sds_controller *scic,
+       u32 frame_index)
+{
+       if (scic_sds_unsolicited_frame_control_release_frame(
+                   &scic->uf_control, frame_index) == true)
+               writel(scic->uf_control.get,
+                       &scic->scu_registers->sdma.unsolicited_frame_get_pointer);
+}
+
+/**
+ * scic_controller_start_io() - This method is called by the SCI user to
+ *    send/start an IO request. If the method invocation is successful, then
+ *    the IO request has been queued to the hardware for processing.
+ * @controller: the handle to the controller object for which to start an IO
+ *    request.
+ * @remote_device: the handle to the remote device object for which to start an
+ *    IO request.
+ * @io_request: the handle to the io request object to start.
+ * @io_tag: This parameter specifies a previously allocated IO tag that the
+ *    user desires to be utilized for this request. This parameter is optional.
+ *     The user is allowed to supply SCI_CONTROLLER_INVALID_IO_TAG as the value
+ *    for this parameter.
+ *
+ * - IO tags are a protected resource.  It is incumbent upon the SCI Core user
+ * to ensure that each of the methods that may allocate or free available IO
+ * tags are handled in a mutually exclusive manner.  This method is one of said
+ * methods requiring proper critical code section protection (e.g. semaphore,
+ * spin-lock, etc.). - For SATA, the user is required to manage NCQ tags.  As a
+ * result, it is expected the user will have set the NCQ tag field in the host
+ * to device register FIS prior to calling this method.  There is also a
+ * requirement for the user to call scic_stp_io_set_ncq_tag() prior to invoking
+ * the scic_controller_start_io() method. scic_controller_allocate_tag() for
+ * more information on allocating a tag. Indicate if the controller
+ * successfully started the IO request. SCI_SUCCESS if the IO request was
+ * successfully started. Determine the failure situations and return values.
+ */
+enum sci_status scic_controller_start_io(
+       struct scic_sds_controller *scic,
+       struct scic_sds_remote_device *rdev,
+       struct scic_sds_request *req,
+       u16 io_tag)
+{
+       enum sci_status status;
+
+       if (scic->state_machine.current_state_id !=
+           SCI_BASE_CONTROLLER_STATE_READY) {
+               dev_warn(scic_to_dev(scic), "invalid state to start I/O");
+               return SCI_FAILURE_INVALID_STATE;
+       }
+
+       status = scic_sds_remote_device_start_io(scic, rdev, req);
+       if (status != SCI_SUCCESS)
+               return status;
+
+       scic->io_request_table[scic_sds_io_tag_get_index(req->io_tag)] = req;
+       scic_sds_controller_post_request(scic, scic_sds_request_get_post_context(req));
+       return SCI_SUCCESS;
+}
+
+/**
+ * scic_controller_terminate_request() - This method is called by the SCI Core
+ *    user to terminate an ongoing (i.e. started) core IO request.  This does
+ *    not abort the IO request at the target, but rather removes the IO request
+ *    from the host controller.
+ * @controller: the handle to the controller object for which to terminate a
+ *    request.
+ * @remote_device: the handle to the remote device object for which to
+ *    terminate a request.
+ * @request: the handle to the io or task management request object to
+ *    terminate.
+ *
+ * Indicate if the controller successfully began the terminate process for the
+ * IO request. SCI_SUCCESS if the terminate process was successfully started
+ * for the request. Determine the failure situations and return values.
+ */
+enum sci_status scic_controller_terminate_request(
+       struct scic_sds_controller *scic,
+       struct scic_sds_remote_device *rdev,
+       struct scic_sds_request *req)
+{
+       enum sci_status status;
+
+       if (scic->state_machine.current_state_id !=
+           SCI_BASE_CONTROLLER_STATE_READY) {
+               dev_warn(scic_to_dev(scic),
+                        "invalid state to terminate request\n");
+               return SCI_FAILURE_INVALID_STATE;
+       }
+
+       status = scic_sds_io_request_terminate(req);
+       if (status != SCI_SUCCESS)
+               return status;
+
+       /*
+        * Utilize the original post context command and or in the POST_TC_ABORT
+        * request sub-type.
+        */
+       scic_sds_controller_post_request(scic,
+               scic_sds_request_get_post_context(req) |
+               SCU_CONTEXT_COMMAND_REQUEST_POST_TC_ABORT);
+       return SCI_SUCCESS;
+}
+
+/**
+ * scic_controller_complete_io() - This method will perform core specific
+ *    completion operations for an IO request.  After this method is invoked,
+ *    the user should consider the IO request as invalid until it is properly
+ *    reused (i.e. re-constructed).
+ * @controller: The handle to the controller object for which to complete the
+ *    IO request.
+ * @remote_device: The handle to the remote device object for which to complete
+ *    the IO request.
+ * @io_request: the handle to the io request object to complete.
+ *
+ * - IO tags are a protected resource.  It is incumbent upon the SCI Core user
+ * to ensure that each of the methods that may allocate or free available IO
+ * tags are handled in a mutually exclusive manner.  This method is one of said
+ * methods requiring proper critical code section protection (e.g. semaphore,
+ * spin-lock, etc.). - If the IO tag for a request was allocated, by the SCI
+ * Core user, using the scic_controller_allocate_io_tag() method, then it is
+ * the responsibility of the caller to invoke the scic_controller_free_io_tag()
+ * method to free the tag (i.e. this method will not free the IO tag). Indicate
+ * if the controller successfully completed the IO request. SCI_SUCCESS if the
+ * completion process was successful.
+ */
+enum sci_status scic_controller_complete_io(
+       struct scic_sds_controller *scic,
+       struct scic_sds_remote_device *rdev,
+       struct scic_sds_request *request)
+{
+       enum sci_status status;
+       u16 index;
+
+       switch (scic->state_machine.current_state_id) {
+       case SCI_BASE_CONTROLLER_STATE_STOPPING:
+               /* XXX: Implement this function */
+               return SCI_FAILURE;
+       case SCI_BASE_CONTROLLER_STATE_READY:
+               status = scic_sds_remote_device_complete_io(scic, rdev, request);
+               if (status != SCI_SUCCESS)
+                       return status;
+
+               index = scic_sds_io_tag_get_index(request->io_tag);
+               scic->io_request_table[index] = NULL;
+               return SCI_SUCCESS;
+       default:
+               dev_warn(scic_to_dev(scic), "invalid state to complete I/O");
+               return SCI_FAILURE_INVALID_STATE;
+       }
+
+}
+
+enum sci_status scic_controller_continue_io(struct scic_sds_request *sci_req)
+{
+       struct scic_sds_controller *scic = sci_req->owning_controller;
+
+       if (scic->state_machine.current_state_id !=
+           SCI_BASE_CONTROLLER_STATE_READY) {
+               dev_warn(scic_to_dev(scic), "invalid state to continue I/O");
+               return SCI_FAILURE_INVALID_STATE;
+       }
+
+       scic->io_request_table[scic_sds_io_tag_get_index(sci_req->io_tag)] = sci_req;
+       scic_sds_controller_post_request(scic, scic_sds_request_get_post_context(sci_req));
+       return SCI_SUCCESS;
+}
+
+/**
+ * scic_controller_start_task() - This method is called by the SCIC user to
+ *    send/start a framework task management request.
+ * @controller: the handle to the controller object for which to start the task
+ *    management request.
+ * @remote_device: the handle to the remote device object for which to start
+ *    the task management request.
+ * @task_request: the handle to the task request object to start.
+ * @io_tag: This parameter specifies a previously allocated IO tag that the
+ *    user desires to be utilized for this request.  Note this not the io_tag
+ *    of the request being managed.  It is to be utilized for the task request
+ *    itself. This parameter is optional.  The user is allowed to supply
+ *    SCI_CONTROLLER_INVALID_IO_TAG as the value for this parameter.
+ *
+ * - IO tags are a protected resource.  It is incumbent upon the SCI Core user
+ * to ensure that each of the methods that may allocate or free available IO
+ * tags are handled in a mutually exclusive manner.  This method is one of said
+ * methods requiring proper critical code section protection (e.g. semaphore,
+ * spin-lock, etc.). - The user must synchronize this task with completion
+ * queue processing.  If they are not synchronized then it is possible for the
+ * io requests that are being managed by the task request can complete before
+ * starting the task request. scic_controller_allocate_tag() for more
+ * information on allocating a tag. Indicate if the controller successfully
+ * started the IO request. SCI_TASK_SUCCESS if the task request was
+ * successfully started. SCI_TASK_FAILURE_REQUIRES_SCSI_ABORT This value is
+ * returned if there is/are task(s) outstanding that require termination or
+ * completion before this request can succeed.
+ */
+enum sci_task_status scic_controller_start_task(
+       struct scic_sds_controller *scic,
+       struct scic_sds_remote_device *rdev,
+       struct scic_sds_request *req,
+       u16 task_tag)
+{
+       enum sci_status status;
+
+       if (scic->state_machine.current_state_id !=
+           SCI_BASE_CONTROLLER_STATE_READY) {
+               dev_warn(scic_to_dev(scic),
+                        "%s: SCIC Controller starting task from invalid "
+                        "state\n",
+                        __func__);
+               return SCI_TASK_FAILURE_INVALID_STATE;
+       }
+
+       status = scic_sds_remote_device_start_task(scic, rdev, req);
+       switch (status) {
+       case SCI_FAILURE_RESET_DEVICE_PARTIAL_SUCCESS:
+               scic->io_request_table[scic_sds_io_tag_get_index(req->io_tag)] = req;
+
+               /*
+                * We will let framework know this task request started successfully,
+                * although core is still woring on starting the request (to post tc when
+                * RNC is resumed.)
+                */
+               return SCI_SUCCESS;
+       case SCI_SUCCESS:
+               scic->io_request_table[scic_sds_io_tag_get_index(req->io_tag)] = req;
+
+               scic_sds_controller_post_request(scic,
+                       scic_sds_request_get_post_context(req));
+               break;
+       default:
+               break;
+       }
+
+       return status;
+}
+
+/**
+ * scic_controller_allocate_io_tag() - This method will allocate a tag from the
+ *    pool of free IO tags. Direct allocation of IO tags by the SCI Core user
+ *    is optional. The scic_controller_start_io() method will allocate an IO
+ *    tag if this method is not utilized and the tag is not supplied to the IO
+ *    construct routine.  Direct allocation of IO tags may provide additional
+ *    performance improvements in environments capable of supporting this usage
+ *    model.  Additionally, direct allocation of IO tags also provides
+ *    additional flexibility to the SCI Core user.  Specifically, the user may
+ *    retain IO tags across the lives of multiple IO requests.
+ * @controller: the handle to the controller object for which to allocate the
+ *    tag.
+ *
+ * IO tags are a protected resource.  It is incumbent upon the SCI Core user to
+ * ensure that each of the methods that may allocate or free available IO tags
+ * are handled in a mutually exclusive manner.  This method is one of said
+ * methods requiring proper critical code section protection (e.g. semaphore,
+ * spin-lock, etc.). An unsigned integer representing an available IO tag.
+ * SCI_CONTROLLER_INVALID_IO_TAG This value is returned if there are no
+ * currently available tags to be allocated. All return other values indicate a
+ * legitimate tag.
+ */
+u16 scic_controller_allocate_io_tag(
+       struct scic_sds_controller *scic)
+{
+       u16 task_context;
+       u16 sequence_count;
+
+       if (!sci_pool_empty(scic->tci_pool)) {
+               sci_pool_get(scic->tci_pool, task_context);
+
+               sequence_count = scic->io_request_sequence[task_context];
+
+               return scic_sds_io_tag_construct(sequence_count, task_context);
+       }
+
+       return SCI_CONTROLLER_INVALID_IO_TAG;
+}
+
+/**
+ * scic_controller_free_io_tag() - This method will free an IO tag to the pool
+ *    of free IO tags. This method provides the SCI Core user more flexibility
+ *    with regards to IO tags.  The user may desire to keep an IO tag after an
+ *    IO request has completed, because they plan on re-using the tag for a
+ *    subsequent IO request.  This method is only legal if the tag was
+ *    allocated via scic_controller_allocate_io_tag().
+ * @controller: This parameter specifies the handle to the controller object
+ *    for which to free/return the tag.
+ * @io_tag: This parameter represents the tag to be freed to the pool of
+ *    available tags.
+ *
+ * - IO tags are a protected resource.  It is incumbent upon the SCI Core user
+ * to ensure that each of the methods that may allocate or free available IO
+ * tags are handled in a mutually exclusive manner.  This method is one of said
+ * methods requiring proper critical code section protection (e.g. semaphore,
+ * spin-lock, etc.). - If the IO tag for a request was allocated, by the SCI
+ * Core user, using the scic_controller_allocate_io_tag() method, then it is
+ * the responsibility of the caller to invoke this method to free the tag. This
+ * method returns an indication of whether the tag was successfully put back
+ * (freed) to the pool of available tags. SCI_SUCCESS This return value
+ * indicates the tag was successfully placed into the pool of available IO
+ * tags. SCI_FAILURE_INVALID_IO_TAG This value is returned if the supplied tag
+ * is not a valid IO tag value.
+ */
+enum sci_status scic_controller_free_io_tag(
+       struct scic_sds_controller *scic,
+       u16 io_tag)
+{
+       u16 sequence;
+       u16 index;
+
+       BUG_ON(io_tag == SCI_CONTROLLER_INVALID_IO_TAG);
+
+       sequence = scic_sds_io_tag_get_sequence(io_tag);
+       index    = scic_sds_io_tag_get_index(io_tag);
+
+       if (!sci_pool_full(scic->tci_pool)) {
+               if (sequence == scic->io_request_sequence[index]) {
+                       scic_sds_io_sequence_increment(
+                               scic->io_request_sequence[index]);
+
+                       sci_pool_put(scic->tci_pool, index);
+
+                       return SCI_SUCCESS;
+               }
+       }
+
+       return SCI_FAILURE_INVALID_IO_TAG;
+}
+
+
index 13c1c99..1f542c4 100644 (file)
  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
-
 #ifndef _SCI_HOST_H_
 #define _SCI_HOST_H_
 
-#include "scic_sds_controller.h"
+#include "scic_config_parameters.h"
 #include "remote_device.h"
 #include "phy.h"
+#include "pool.h"
+#include "sci_base_state_machine.h"
+#include "remote_node_table.h"
+#include "scu_registers.h"
+#include "scu_unsolicited_frame.h"
+#include "scic_sds_unsolicited_frame_control.h"
+#include "scic_sds_port_configuration_agent.h"
+
+struct scic_sds_request;
+struct scu_task_context;
+
+/**
+ * struct scic_power_control -
+ *
+ * This structure defines the fields for managing power control for direct
+ * attached disk devices.
+ */
+struct scic_power_control {
+       /**
+        * This field is set when the power control timer is running and cleared when
+        * it is not.
+        */
+       bool timer_started;
+
+       /**
+        * This field is the handle to the driver timer object.  This timer is used to
+        * control when the directed attached disks can consume power.
+        */
+       void *timer;
+
+       /**
+        * This field is used to keep track of how many phys are put into the
+        * requesters field.
+        */
+       u8 phys_waiting;
+
+       /**
+        * This field is used to keep track of how many phys have been granted to consume power
+        */
+       u8 phys_granted_power;
+
+       /**
+        * This field is an array of phys that we are waiting on. The phys are direct
+        * mapped into requesters via struct scic_sds_phy.phy_index
+        */
+       struct scic_sds_phy *requesters[SCI_MAX_PHYS];
+
+};
+
+/**
+ * struct scic_sds_controller -
+ *
+ * This structure represents the SCU controller object.
+ */
+struct scic_sds_controller {
+       /**
+        * This field contains the information for the base controller state
+        * machine.
+        */
+       struct sci_base_state_machine state_machine;
+
+       /**
+        * This field is the driver timer object handler used to time the controller
+        * object start and stop requests.
+        */
+       void *timeout_timer;
+
+       /**
+        * This field contains the user parameters to be utilized for this
+        * core controller object.
+        */
+       union scic_user_parameters user_parameters;
+
+       /**
+        * This field contains the OEM parameters to be utilized for this
+        * core controller object.
+        */
+       union scic_oem_parameters oem_parameters;
+
+       /**
+        * This field contains the port configuration agent for this controller.
+        */
+       struct scic_sds_port_configuration_agent port_agent;
+
+       /**
+        * This field is the array of device objects that are currently constructed
+        * for this controller object.  This table is used as a fast lookup of device
+        * objects that need to handle device completion notifications from the
+        * hardware. The table is RNi based.
+        */
+       struct scic_sds_remote_device *device_table[SCI_MAX_REMOTE_DEVICES];
+
+       /**
+        * This field is the array of IO request objects that are currently active for
+        * this controller object.  This table is used as a fast lookup of the io
+        * request object that need to handle completion queue notifications.  The
+        * table is TCi based.
+        */
+       struct scic_sds_request *io_request_table[SCI_MAX_IO_REQUESTS];
+
+       /**
+        * This field is the free RNi data structure
+        */
+       struct scic_remote_node_table available_remote_nodes;
+
+       /**
+        * This field is the TCi pool used to manage the task context index.
+        */
+       SCI_POOL_CREATE(tci_pool, u16, SCI_MAX_IO_REQUESTS);
+
+       /**
+        * This filed is the struct scic_power_control data used to controll when direct
+        * attached devices can consume power.
+        */
+       struct scic_power_control power_control;
+
+       /**
+        * This field is the array of sequence values for the IO Tag fields.  Even
+        * though only 4 bits of the field is used for the sequence the sequence is 16
+        * bits in size so the sequence can be bitwise or'd with the TCi to build the
+        * IO Tag value.
+        */
+       u16 io_request_sequence[SCI_MAX_IO_REQUESTS];
+
+       /**
+        * This field in the array of sequence values for the RNi.  These are used
+        * to control io request build to io request start operations.  The sequence
+        * value is recorded into an io request when it is built and is checked on
+        * the io request start operation to make sure that there was not a device
+        * hot plug between the build and start operation.
+        */
+       u8 remote_device_sequence[SCI_MAX_REMOTE_DEVICES];
+
+       /**
+        * This field is a pointer to the memory allocated by the driver for the task
+        * context table.  This data is shared between the hardware and software.
+        */
+       struct scu_task_context *task_context_table;
+
+       /**
+        * This field is a pointer to the memory allocated by the driver for the
+        * remote node context table.  This table is shared between the hardware and
+        * software.
+        */
+       union scu_remote_node_context *remote_node_context_table;
+
+       /**
+        * This field is a pointer to the completion queue.  This memory is
+        * written to by the hardware and read by the software.
+        */
+       u32 *completion_queue;
+
+       /**
+        * This field is the software copy of the completion queue get pointer.  The
+        * controller object writes this value to the hardware after processing the
+        * completion entries.
+        */
+       u32 completion_queue_get;
+
+       /**
+        * This field is the minimum of the number of hardware supported port entries
+        * and the software requested port entries.
+        */
+       u32 logical_port_entries;
+
+       /**
+        * This field is the minimum number of hardware supported completion queue
+        * entries and the software requested completion queue entries.
+        */
+       u32 completion_queue_entries;
+
+       /**
+        * This field is the minimum number of hardware supported event entries and
+        * the software requested event entries.
+        */
+       u32 completion_event_entries;
+
+       /**
+        * This field is the minimum number of devices supported by the hardware and
+        * the number of devices requested by the software.
+        */
+       u32 remote_node_entries;
+
+       /**
+        * This field is the minimum number of IO requests supported by the hardware
+        * and the number of IO requests requested by the software.
+        */
+       u32 task_context_entries;
+
+       /**
+        * This object contains all of the unsolicited frame specific
+        * data utilized by the core controller.
+        */
+       struct scic_sds_unsolicited_frame_control uf_control;
+
+       /* Phy Startup Data */
+       /**
+        * This field is the driver timer handle for controller phy request startup.
+        * On controller start the controller will start each PHY individually in
+        * order of phy index.
+        */
+       void *phy_startup_timer;
+
+       /**
+        * This field is set when the phy_startup_timer is running and is cleared when
+        * the phy_startup_timer is stopped.
+        */
+       bool phy_startup_timer_pending;
+
+       /**
+        * This field is the index of the next phy start.  It is initialized to 0 and
+        * increments for each phy index that is started.
+        */
+       u32 next_phy_to_start;
+
+       /**
+        * This field controlls the invalid link up notifications to the SCI_USER.  If
+        * an invalid_link_up notification is reported a bit for the PHY index is set
+        * so further notifications are not made.  Once the PHY object reports link up
+        * and is made part of a port then this bit for the PHY index is cleared.
+        */
+       u8 invalid_phy_mask;
+
+       /*
+        * This field saves the current interrupt coalescing number of the controller.
+        */
+       u16 interrupt_coalesce_number;
+
+       /*
+        * This field saves the current interrupt coalescing timeout value in microseconds.
+        */
+       u32 interrupt_coalesce_timeout;
+
+       /**
+        * This field is a pointer to the memory mapped register space for the
+        * struct smu_registers.
+        */
+       struct smu_registers __iomem *smu_registers;
+
+       /**
+        * This field is a pointer to the memory mapped register space for the
+        * struct scu_registers.
+        */
+       struct scu_registers __iomem *scu_registers;
+
+};
 
 struct isci_host {
        struct scic_sds_controller sci;
@@ -93,6 +338,96 @@ struct isci_host {
 };
 
 /**
+ * enum scic_sds_controller_states - This enumeration depicts all the states
+ *    for the common controller state machine.
+ */
+enum scic_sds_controller_states {
+       /**
+        * Simply the initial state for the base controller state machine.
+        */
+       SCI_BASE_CONTROLLER_STATE_INITIAL = 0,
+
+       /**
+        * This state indicates that the controller is reset.  The memory for
+        * the controller is in it's initial state, but the controller requires
+        * initialization.
+        * This state is entered from the INITIAL state.
+        * This state is entered from the RESETTING state.
+        */
+       SCI_BASE_CONTROLLER_STATE_RESET,
+
+       /**
+        * This state is typically an action state that indicates the controller
+        * is in the process of initialization.  In this state no new IO operations
+        * are permitted.
+        * This state is entered from the RESET state.
+        */
+       SCI_BASE_CONTROLLER_STATE_INITIALIZING,
+
+       /**
+        * This state indicates that the controller has been successfully
+        * initialized.  In this state no new IO operations are permitted.
+        * This state is entered from the INITIALIZING state.
+        */
+       SCI_BASE_CONTROLLER_STATE_INITIALIZED,
+
+       /**
+        * This state indicates the the controller is in the process of becoming
+        * ready (i.e. starting).  In this state no new IO operations are permitted.
+        * This state is entered from the INITIALIZED state.
+        */
+       SCI_BASE_CONTROLLER_STATE_STARTING,
+
+       /**
+        * This state indicates the controller is now ready.  Thus, the user
+        * is able to perform IO operations on the controller.
+        * This state is entered from the STARTING state.
+        */
+       SCI_BASE_CONTROLLER_STATE_READY,
+
+       /**
+        * This state is typically an action state that indicates the controller
+        * is in the process of resetting.  Thus, the user is unable to perform
+        * IO operations on the controller.  A reset is considered destructive in
+        * most cases.
+        * This state is entered from the READY state.
+        * This state is entered from the FAILED state.
+        * This state is entered from the STOPPED state.
+        */
+       SCI_BASE_CONTROLLER_STATE_RESETTING,
+
+       /**
+        * This state indicates that the controller is in the process of stopping.
+        * In this state no new IO operations are permitted, but existing IO
+        * operations are allowed to complete.
+        * This state is entered from the READY state.
+        */
+       SCI_BASE_CONTROLLER_STATE_STOPPING,
+
+       /**
+        * This state indicates that the controller has successfully been stopped.
+        * In this state no new IO operations are permitted.
+        * This state is entered from the STOPPING state.
+        */
+       SCI_BASE_CONTROLLER_STATE_STOPPED,
+
+       /**
+        * This state indicates that the controller could not successfully be
+        * initialized.  In this state no new IO operations are permitted.
+        * This state is entered from the INITIALIZING state.
+        * This state is entered from the STARTING state.
+        * This state is entered from the STOPPING state.
+        * This state is entered from the RESETTING state.
+        */
+       SCI_BASE_CONTROLLER_STATE_FAILED,
+
+       SCI_BASE_CONTROLLER_MAX_STATES
+
+};
+
+
+
+/**
  * struct isci_pci_info - This class represents the pci function containing the
  *    controllers. Depending on PCI SKU, there could be up to 2 controllers in
  *    the PCI function.
@@ -115,17 +450,13 @@ static inline struct isci_pci_info *to_pci_info(struct pci_dev *pdev)
             id < ARRAY_SIZE(to_pci_info(pdev)->hosts) && ihost; \
             ihost = to_pci_info(pdev)->hosts[++id])
 
-static inline
-enum isci_status isci_host_get_state(
-       struct isci_host *isci_host)
+static inline enum isci_status isci_host_get_state(struct isci_host *isci_host)
 {
        return isci_host->status;
 }
 
-
-static inline void isci_host_change_state(
-       struct isci_host *isci_host,
-       enum isci_status status)
+static inline void isci_host_change_state(struct isci_host *isci_host,
+                                         enum isci_status status)
 {
        unsigned long flags;
 
@@ -140,9 +471,7 @@ static inline void isci_host_change_state(
 
 }
 
-static inline int isci_host_can_queue(
-       struct isci_host *isci_host,
-       int num)
+static inline int isci_host_can_queue(struct isci_host *isci_host, int num)
 {
        int ret = 0;
        unsigned long flags;
@@ -163,9 +492,7 @@ static inline int isci_host_can_queue(
        return ret;
 }
 
-static inline void isci_host_can_dequeue(
-       struct isci_host *isci_host,
-       int num)
+static inline void isci_host_can_dequeue(struct isci_host *isci_host, int num)
 {
        unsigned long flags;
 
@@ -208,39 +535,219 @@ static inline struct isci_host *scic_to_ihost(struct scic_sds_controller *scic)
 }
 
 /**
- * isci_host_scan_finished() -
+ * INCREMENT_QUEUE_GET() -
  *
- * This function is one of the SCSI Host Template functions. The SCSI midlayer
- * calls this function during a target scan, approx. once every 10 millisecs.
+ * This macro will increment the specified index to and if the index wraps to 0
+ * it will toggel the cycle bit.
  */
-int isci_host_scan_finished(
-       struct Scsi_Host *,
-       unsigned long);
+#define INCREMENT_QUEUE_GET(index, cycle, entry_count, bit_toggle) \
+       { \
+               if ((index) + 1 == entry_count) {       \
+                       (index) = 0; \
+                       (cycle) = (cycle) ^ (bit_toggle); \
+               } else { \
+                       index = index + 1; \
+               } \
+       }
 
+/**
+ * scic_sds_controller_get_port_configuration_agent() -
+ *
+ * This is a helper macro to get the port configuration agent from the
+ * controller object.
+ */
+#define scic_sds_controller_get_port_configuration_agent(controller) \
+       (&(controller)->port_agent)
 
 /**
- * isci_host_scan_start() -
+ * scic_sds_controller_get_protocol_engine_group() -
  *
- * This function is one of the SCSI Host Template function, called by the SCSI
- * mid layer berfore a target scan begins. The core library controller start
- * routine is called from here.
+ * This macro returns the protocol engine group for this controller object.
+ * Presently we only support protocol engine group 0 so just return that
  */
-void isci_host_scan_start(
-       struct Scsi_Host *);
+#define scic_sds_controller_get_protocol_engine_group(controller) 0
 
 /**
- * isci_host_start_complete() -
+ * scic_sds_io_tag_construct() -
  *
- * This function is called by the core library, through the ISCI Module, to
- * indicate controller start status.
+ * This macro constructs an IO tag from the sequence and index values.
  */
-void isci_host_start_complete(
-       struct isci_host *,
-       enum sci_status);
+#define scic_sds_io_tag_construct(sequence, task_index)        \
+       ((sequence) << 12 | (task_index))
 
-void isci_host_stop_complete(
-       struct isci_host *isci_host,
-       enum sci_status completion_status);
+/**
+ * scic_sds_io_tag_get_sequence() -
+ *
+ * This macro returns the IO sequence from the IO tag value.
+ */
+#define scic_sds_io_tag_get_sequence(io_tag) \
+       (((io_tag) & 0xF000) >> 12)
+
+/**
+ * scic_sds_io_tag_get_index() -
+ *
+ * This macro returns the TCi from the io tag value
+ */
+#define scic_sds_io_tag_get_index(io_tag) \
+       ((io_tag) & 0x0FFF)
+
+/**
+ * scic_sds_io_sequence_increment() -
+ *
+ * This is a helper macro to increment the io sequence count. We may find in
+ * the future that it will be faster to store the sequence count in such a way
+ * as we dont perform the shift operation to build io tag values so therefore
+ * need a way to incrment them correctly
+ */
+#define scic_sds_io_sequence_increment(value) \
+       ((value) = (((value) + 1) & 0x000F))
+
+/* expander attached sata devices require 3 rnc slots */
+static inline int scic_sds_remote_device_node_count(struct scic_sds_remote_device *sci_dev)
+{
+       struct domain_device *dev = sci_dev_to_domain(sci_dev);
+
+       if ((dev->dev_type == SATA_DEV || (dev->tproto & SAS_PROTOCOL_STP)) &&
+           !sci_dev->is_direct_attached)
+               return SCU_STP_REMOTE_NODE_COUNT;
+       return SCU_SSP_REMOTE_NODE_COUNT;
+}
+
+/**
+ * scic_sds_controller_set_invalid_phy() -
+ *
+ * This macro will set the bit in the invalid phy mask for this controller
+ * object.  This is used to control messages reported for invalid link up
+ * notifications.
+ */
+#define scic_sds_controller_set_invalid_phy(controller, phy) \
+       ((controller)->invalid_phy_mask |= (1 << (phy)->phy_index))
+
+/**
+ * scic_sds_controller_clear_invalid_phy() -
+ *
+ * This macro will clear the bit in the invalid phy mask for this controller
+ * object.  This is used to control messages reported for invalid link up
+ * notifications.
+ */
+#define scic_sds_controller_clear_invalid_phy(controller, phy) \
+       ((controller)->invalid_phy_mask &= ~(1 << (phy)->phy_index))
+
+static inline struct device *scic_to_dev(struct scic_sds_controller *scic)
+{
+       return &scic_to_ihost(scic)->pdev->dev;
+}
+
+static inline struct device *sciphy_to_dev(struct scic_sds_phy *sci_phy)
+{
+       struct isci_phy *iphy = sci_phy_to_iphy(sci_phy);
+
+       if (!iphy || !iphy->isci_port || !iphy->isci_port->isci_host)
+               return NULL;
+
+       return &iphy->isci_port->isci_host->pdev->dev;
+}
+
+static inline struct device *sciport_to_dev(struct scic_sds_port *sci_port)
+{
+       struct isci_port *iport = sci_port_to_iport(sci_port);
+
+       if (!iport || !iport->isci_host)
+               return NULL;
+
+       return &iport->isci_host->pdev->dev;
+}
+
+static inline struct device *scirdev_to_dev(struct scic_sds_remote_device *sci_dev)
+{
+       struct isci_remote_device *idev =
+                       container_of(sci_dev, typeof(*idev), sci);
+
+       if (!idev || !idev->isci_port || !idev->isci_port->isci_host)
+               return NULL;
+
+       return &idev->isci_port->isci_host->pdev->dev;
+}
+
+enum {
+       ISCI_SI_REVA0,
+       ISCI_SI_REVA2,
+       ISCI_SI_REVB0,
+};
+
+extern int isci_si_rev;
+
+static inline bool is_a0(void)
+{
+       return isci_si_rev == ISCI_SI_REVA0;
+}
+
+static inline bool is_a2(void)
+{
+       return isci_si_rev == ISCI_SI_REVA2;
+}
+
+static inline bool is_b0(void)
+{
+       return isci_si_rev > ISCI_SI_REVA2;
+}
+
+void scic_sds_controller_post_request(struct scic_sds_controller *scic,
+                                     u32 request);
+void scic_sds_controller_release_frame(struct scic_sds_controller *scic,
+                                      u32 frame_index);
+void scic_sds_controller_copy_sata_response(void *response_buffer,
+                                           void *frame_header,
+                                           void *frame_buffer);
+enum sci_status scic_sds_controller_allocate_remote_node_context(struct scic_sds_controller *scic,
+                                                                struct scic_sds_remote_device *sci_dev,
+                                                                u16 *node_id);
+void scic_sds_controller_free_remote_node_context(
+       struct scic_sds_controller *scic,
+       struct scic_sds_remote_device *sci_dev,
+       u16 node_id);
+union scu_remote_node_context *scic_sds_controller_get_remote_node_context_buffer(
+       struct scic_sds_controller *scic,
+       u16 node_id);
+
+struct scic_sds_request *scic_request_by_tag(struct scic_sds_controller *scic,
+                                            u16 io_tag);
+
+struct scu_task_context *scic_sds_controller_get_task_context_buffer(
+       struct scic_sds_controller *scic,
+       u16 io_tag);
+
+void scic_sds_controller_power_control_queue_insert(
+       struct scic_sds_controller *scic,
+       struct scic_sds_phy *sci_phy);
+
+void scic_sds_controller_power_control_queue_remove(
+       struct scic_sds_controller *scic,
+       struct scic_sds_phy *sci_phy);
+
+void scic_sds_controller_link_up(
+       struct scic_sds_controller *scic,
+       struct scic_sds_port *sci_port,
+       struct scic_sds_phy *sci_phy);
+
+void scic_sds_controller_link_down(
+       struct scic_sds_controller *scic,
+       struct scic_sds_port *sci_port,
+       struct scic_sds_phy *sci_phy);
+
+void scic_sds_controller_remote_device_stopped(
+       struct scic_sds_controller *scic,
+       struct scic_sds_remote_device *sci_dev);
+
+void scic_sds_controller_copy_task_context(
+       struct scic_sds_controller *scic,
+       struct scic_sds_request *this_request);
+
+void scic_sds_controller_register_setup(struct scic_sds_controller *scic);
+
+enum sci_status scic_controller_continue_io(struct scic_sds_request *sci_req);
+int isci_host_scan_finished(struct Scsi_Host *, unsigned long);
+void isci_host_scan_start(struct Scsi_Host *);
 
 int isci_host_init(struct isci_host *);
 
@@ -262,4 +769,35 @@ void isci_host_remote_device_start_complete(
        struct isci_remote_device *,
        enum sci_status);
 
-#endif /* !defined(_SCI_HOST_H_) */
+void scic_controller_disable_interrupts(
+       struct scic_sds_controller *scic);
+
+enum sci_status scic_controller_start_io(
+       struct scic_sds_controller *scic,
+       struct scic_sds_remote_device *remote_device,
+       struct scic_sds_request *io_request,
+       u16 io_tag);
+
+enum sci_task_status scic_controller_start_task(
+       struct scic_sds_controller *scic,
+       struct scic_sds_remote_device *remote_device,
+       struct scic_sds_request *task_request,
+       u16 io_tag);
+
+enum sci_status scic_controller_terminate_request(
+       struct scic_sds_controller *scic,
+       struct scic_sds_remote_device *remote_device,
+       struct scic_sds_request *request);
+
+enum sci_status scic_controller_complete_io(
+       struct scic_sds_controller *scic,
+       struct scic_sds_remote_device *remote_device,
+       struct scic_sds_request *io_request);
+
+u16 scic_controller_allocate_io_tag(
+       struct scic_sds_controller *scic);
+
+enum sci_status scic_controller_free_io_tag(
+       struct scic_sds_controller *scic,
+       u16 io_tag);
+#endif
index df132c0..bda7016 100644 (file)
@@ -61,9 +61,7 @@
 #include <asm/string.h>
 #include "isci.h"
 #include "task.h"
-#include "sci_environment.h"
 #include "probe_roms.h"
-#include "scic_controller.h"
 
 static struct scsi_transport_template *isci_transport_template;
 
index 800f233..d288897 100644 (file)
@@ -532,10 +532,4 @@ extern unsigned char max_concurr_spinup;
 irqreturn_t isci_msix_isr(int vec, void *data);
 irqreturn_t isci_intx_isr(int vec, void *data);
 irqreturn_t isci_error_isr(int vec, void *data);
-
-struct scic_sds_controller;
-bool scic_sds_controller_isr(struct scic_sds_controller *scic);
-void scic_sds_controller_completion_handler(struct scic_sds_controller *scic);
-bool scic_sds_controller_error_isr(struct scic_sds_controller *scic);
-void scic_sds_controller_error_handler(struct scic_sds_controller *scic);
 #endif  /* __ISCI_H__ */
index 35e2e51..f44fa20 100644 (file)
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-/**
- * This file contains the isci port implementation.
- *
- *
- */
-
-
 #include <linux/workqueue.h>
 #include "isci.h"
 #include "scic_io_request.h"
@@ -68,7 +61,6 @@
 #include "scic_port.h"
 #include "port.h"
 #include "request.h"
-#include "core/scic_sds_controller.h"
 
 static void isci_port_change_state(struct isci_port *iport, enum isci_status status)
 {
index 7055174..9bc173f 100644 (file)
@@ -32,7 +32,6 @@
 
 #include "isci.h"
 #include "task.h"
-#include "sci_environment.h"
 #include "probe_roms.h"
 
 struct efi_variable {
index 8b1ef19..00334b9 100644 (file)
 #include "port.h"
 #include "remote_device.h"
 #include "request.h"
-#include "scic_controller.h"
 #include "scic_io_request.h"
 #include "scic_phy.h"
 #include "scic_port.h"
-#include "scic_sds_controller.h"
 #include "scic_sds_phy.h"
 #include "scic_sds_port.h"
 #include "remote_node_context.h"
 #include "scic_sds_request.h"
-#include "sci_environment.h"
 #include "sci_util.h"
 #include "scu_event_codes.h"
 #include "task.h"
index 5e85a18..af47593 100644 (file)
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
+#include "host.h"
 #include "sci_base_state_machine.h"
-#include "scic_sds_controller.h"
 #include "scic_sds_port.h"
 #include "remote_device.h"
 #include "remote_node_context.h"
-#include "sci_environment.h"
 #include "sci_util.h"
 #include "scu_event_codes.h"
 #include "scu_task_context.h"
index 8886146..80f44c2 100644 (file)
@@ -60,7 +60,6 @@
  *
  */
 #include "sci_util.h"
-#include "sci_environment.h"
 #include "remote_node_table.h"
 #include "remote_node_context.h"
 
index a580728..4961ee3 100644 (file)
@@ -62,7 +62,6 @@
 #include "sata.h"
 #include "scu_completion_codes.h"
 #include "scic_sds_request.h"
-#include "scic_controller.h"
 #include "sas.h"
 
 static enum sci_status isci_request_ssp_request_construct(
diff --git a/drivers/scsi/isci/sci_environment.h b/drivers/scsi/isci/sci_environment.h
deleted file mode 100644 (file)
index 30addba..0000000
+++ /dev/null
@@ -1,122 +0,0 @@
-/*
- * This file is provided under a dual BSD/GPLv2 license.  When using or
- * redistributing this file, you may do so under either license.
- *
- * GPL LICENSE SUMMARY
- *
- * Copyright(c) 2008 - 2011 Intel Corporation. All rights reserved.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
- * The full GNU General Public License is included in this distribution
- * in the file called LICENSE.GPL.
- *
- * BSD LICENSE
- *
- * Copyright(c) 2008 - 2011 Intel Corporation. All rights reserved.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- *   * Redistributions of source code must retain the above copyright
- *     notice, this list of conditions and the following disclaimer.
- *   * Redistributions in binary form must reproduce the above copyright
- *     notice, this list of conditions and the following disclaimer in
- *     the documentation and/or other materials provided with the
- *     distribution.
- *   * Neither the name of Intel Corporation nor the names of its
- *     contributors may be used to endorse or promote products derived
- *     from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-#ifndef _SCI_ENVIRONMENT_H_
-#define _SCI_ENVIRONMENT_H_
-
-#include "host.h"
-
-
-static inline struct device *scic_to_dev(struct scic_sds_controller *scic)
-{
-       return &scic_to_ihost(scic)->pdev->dev;
-}
-
-static inline struct device *sciphy_to_dev(struct scic_sds_phy *sci_phy)
-{
-       struct isci_phy *iphy = sci_phy_to_iphy(sci_phy);
-
-       if (!iphy || !iphy->isci_port || !iphy->isci_port->isci_host)
-               return NULL;
-
-       return &iphy->isci_port->isci_host->pdev->dev;
-}
-
-static inline struct device *sciport_to_dev(struct scic_sds_port *sci_port)
-{
-       struct isci_port *iport = sci_port_to_iport(sci_port);
-
-       if (!iport || !iport->isci_host)
-               return NULL;
-
-       return &iport->isci_host->pdev->dev;
-}
-
-static inline struct device *scirdev_to_dev(
-               struct scic_sds_remote_device *sci_dev)
-{
-       struct isci_remote_device *idev =
-                       container_of(sci_dev, typeof(*idev), sci);
-
-       if (!idev || !idev->isci_port || !idev->isci_port->isci_host)
-               return NULL;
-
-       return &idev->isci_port->isci_host->pdev->dev;
-}
-
-enum {
-       ISCI_SI_REVA0,
-       ISCI_SI_REVA2,
-       ISCI_SI_REVB0,
-};
-
-extern int isci_si_rev;
-
-static inline bool is_a0(void)
-{
-       return isci_si_rev == ISCI_SI_REVA0;
-}
-
-static inline bool is_a2(void)
-{
-       return isci_si_rev == ISCI_SI_REVA2;
-}
-
-static inline bool is_b0(void)
-{
-       return isci_si_rev > ISCI_SI_REVA2;
-}
-
-#endif
index 597c490..2a86038 100644 (file)
@@ -65,7 +65,6 @@
 #include "sata.h"
 #include "task.h"
 #include "scic_sds_request.h"
-#include "scic_controller.h"
 #include "timers.h"
 
 /**