MicroSemi Switchtec management interface driver

Microsemi's "Switchtec" line of PCI switch devices is already well
supported by the kernel with standard PCI switch drivers.  However, the
Switchtec device advertises a special management endpoint with a separate
PCI function address and class code.  This endpoint enables some additional
functionality which includes:

 * Packet and Byte Counters
 * Switch Firmware Upgrades
 * Event and Error logs
 * Querying port link status
 * Custom user firmware commands

Add a switchtec kernel module which provides PCI driver that exposes a char
device.  The char device provides userspace access to this interface
through read, write and (optionally) poll calls.

A userspace tool and library which utilizes this interface is available
at [1].  This tool takes inspiration (and borrows some code) from
nvme-cli [2].  The tool is largely complete at this time but additional
features may be added in the future.

[1] https://github.com/sbates130272/switchtec-user
[2] https://github.com/linux-nvme/nvme-cli

[Dan Carpenter <dan.carpenter@oracle.com>: don't invert error codes]
[Christophe JAILLET <christophe.jaillet@wanadoo.fr>: fix
switchtec_dev_open() error handling]
Tested-by: Krishna Dhulipala <krishnad@fb.com>
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Signed-off-by: Stephen Bates <stephen.bates@microsemi.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Wei Zhang <wzhang@fb.com>
Reviewed-by: Jens Axboe <axboe@fb.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

5 files changed, 1022 insertions, 0 deletions

diff --git a/drivers/pci/Kconfig b/drivers/pci/Kconfig
index df141420c902..c0bb4cbd1edb 100644
--- a/drivers/pci/Kconfig
+++ b/drivers/pci/Kconfig
@@ -134,3 +134,4 @@ config PCI_HYPERV
 source "drivers/pci/hotplug/Kconfig"
 source "drivers/pci/dwc/Kconfig"
 source "drivers/pci/host/Kconfig"
+source "drivers/pci/switch/Kconfig"
diff --git a/drivers/pci/Makefile b/drivers/pci/Makefile
index 8db5079f09a7..15b46dd5074b 100644
--- a/drivers/pci/Makefile
+++ b/drivers/pci/Makefile
@@ -68,3 +68,4 @@ ccflags-$(CONFIG_PCI_DEBUG) := -DDEBUG
 
 # PCI host controller drivers
 obj-y += host/
+obj-y += switch/
diff --git a/drivers/pci/switch/Kconfig b/drivers/pci/switch/Kconfig
new file mode 100644
index 000000000000..4c49648e0646
--- /dev/null
+++ b/drivers/pci/switch/Kconfig
@@ -0,0 +1,13 @@
+menu "PCI switch controller drivers"
+	depends on PCI
+
+config PCI_SW_SWITCHTEC
+	tristate "MicroSemi Switchtec PCIe Switch Management Driver"
+	help
+	 Enables support for the management interface for the MicroSemi
+	 Switchtec series of PCIe switches. Supports userspace access
+	 to submit MRPC commands to the switch via /dev/switchtecX
+	 devices. See <file:Documentation/switchtec.txt> for more
+	 information.
+
+endmenu
diff --git a/drivers/pci/switch/Makefile b/drivers/pci/switch/Makefile
new file mode 100644
index 000000000000..37d8cfb03f3f
--- /dev/null
+++ b/drivers/pci/switch/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_PCI_SW_SWITCHTEC) += switchtec.o
diff --git a/drivers/pci/switch/switchtec.c b/drivers/pci/switch/switchtec.c
new file mode 100644
index 000000000000..0bbe90df968d
--- /dev/null
+++ b/drivers/pci/switch/switchtec.c
@@ -0,0 +1,1006 @@
+/*
+ * Microsemi Switchtec(tm) PCIe Management Driver
+ * Copyright (c) 2017, Microsemi Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ */
+
+#include <linux/interrupt.h>
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/uaccess.h>
+#include <linux/poll.h>
+#include <linux/pci.h>
+#include <linux/cdev.h>
+#include <linux/wait.h>
+
+MODULE_DESCRIPTION("Microsemi Switchtec(tm) PCIe Management Driver");
+MODULE_VERSION("0.1");
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Microsemi Corporation");
+
+static int max_devices = 16;
+module_param(max_devices, int, 0644);
+MODULE_PARM_DESC(max_devices, "max number of switchtec device instances");
+
+static dev_t switchtec_devt;
+static struct class *switchtec_class;
+static DEFINE_IDA(switchtec_minor_ida);
+
+#define MICROSEMI_VENDOR_ID         0x11f8
+#define MICROSEMI_NTB_CLASSCODE     0x068000
+#define MICROSEMI_MGMT_CLASSCODE    0x058000
+
+#define SWITCHTEC_MRPC_PAYLOAD_SIZE 1024
+#define SWITCHTEC_MAX_PFF_CSR 48
+
+#define SWITCHTEC_EVENT_OCCURRED BIT(0)
+#define SWITCHTEC_EVENT_CLEAR    BIT(0)
+#define SWITCHTEC_EVENT_EN_LOG   BIT(1)
+#define SWITCHTEC_EVENT_EN_CLI   BIT(2)
+#define SWITCHTEC_EVENT_EN_IRQ   BIT(3)
+#define SWITCHTEC_EVENT_FATAL    BIT(4)
+
+enum {
+	SWITCHTEC_GAS_MRPC_OFFSET       = 0x0000,
+	SWITCHTEC_GAS_TOP_CFG_OFFSET    = 0x1000,
+	SWITCHTEC_GAS_SW_EVENT_OFFSET   = 0x1800,
+	SWITCHTEC_GAS_SYS_INFO_OFFSET   = 0x2000,
+	SWITCHTEC_GAS_FLASH_INFO_OFFSET = 0x2200,
+	SWITCHTEC_GAS_PART_CFG_OFFSET   = 0x4000,
+	SWITCHTEC_GAS_NTB_OFFSET        = 0x10000,
+	SWITCHTEC_GAS_PFF_CSR_OFFSET    = 0x134000,
+};
+
+struct mrpc_regs {
+	u8 input_data[SWITCHTEC_MRPC_PAYLOAD_SIZE];
+	u8 output_data[SWITCHTEC_MRPC_PAYLOAD_SIZE];
+	u32 cmd;
+	u32 status;
+	u32 ret_value;
+} __packed;
+
+enum mrpc_status {
+	SWITCHTEC_MRPC_STATUS_INPROGRESS = 1,
+	SWITCHTEC_MRPC_STATUS_DONE = 2,
+	SWITCHTEC_MRPC_STATUS_ERROR = 0xFF,
+	SWITCHTEC_MRPC_STATUS_INTERRUPTED = 0x100,
+};
+
+struct sw_event_regs {
+	u64 event_report_ctrl;
+	u64 reserved1;
+	u64 part_event_bitmap;
+	u64 reserved2;
+	u32 global_summary;
+	u32 reserved3[3];
+	u32 stack_error_event_hdr;
+	u32 stack_error_event_data;
+	u32 reserved4[4];
+	u32 ppu_error_event_hdr;
+	u32 ppu_error_event_data;
+	u32 reserved5[4];
+	u32 isp_error_event_hdr;
+	u32 isp_error_event_data;
+	u32 reserved6[4];
+	u32 sys_reset_event_hdr;
+	u32 reserved7[5];
+	u32 fw_exception_hdr;
+	u32 reserved8[5];
+	u32 fw_nmi_hdr;
+	u32 reserved9[5];
+	u32 fw_non_fatal_hdr;
+	u32 reserved10[5];
+	u32 fw_fatal_hdr;
+	u32 reserved11[5];
+	u32 twi_mrpc_comp_hdr;
+	u32 twi_mrpc_comp_data;
+	u32 reserved12[4];
+	u32 twi_mrpc_comp_async_hdr;
+	u32 twi_mrpc_comp_async_data;
+	u32 reserved13[4];
+	u32 cli_mrpc_comp_hdr;
+	u32 cli_mrpc_comp_data;
+	u32 reserved14[4];
+	u32 cli_mrpc_comp_async_hdr;
+	u32 cli_mrpc_comp_async_data;
+	u32 reserved15[4];
+	u32 gpio_interrupt_hdr;
+	u32 gpio_interrupt_data;
+	u32 reserved16[4];
+} __packed;
+
+struct sys_info_regs {
+	u32 device_id;
+	u32 device_version;
+	u32 firmware_version;
+	u32 reserved1;
+	u32 vendor_table_revision;
+	u32 table_format_version;
+	u32 partition_id;
+	u32 cfg_file_fmt_version;
+	u32 reserved2[58];
+	char vendor_id[8];
+	char product_id[16];
+	char product_revision[4];
+	char component_vendor[8];
+	u16 component_id;
+	u8 component_revision;
+} __packed;
+
+struct flash_info_regs {
+	u32 flash_part_map_upd_idx;
+
+	struct active_partition_info {
+		u32 address;
+		u32 build_version;
+		u32 build_string;
+	} active_img;
+
+	struct active_partition_info active_cfg;
+	struct active_partition_info inactive_img;
+	struct active_partition_info inactive_cfg;
+
+	u32 flash_length;
+
+	struct partition_info {
+		u32 address;
+		u32 length;
+	} cfg0;
+
+	struct partition_info cfg1;
+	struct partition_info img0;
+	struct partition_info img1;
+	struct partition_info nvlog;
+	struct partition_info vendor[8];
+};
+
+struct ntb_info_regs {
+	u8  partition_count;
+	u8  partition_id;
+	u16 reserved1;
+	u64 ep_map;
+	u16 requester_id;
+} __packed;
+
+struct part_cfg_regs {
+	u32 status;
+	u32 state;
+	u32 port_cnt;
+	u32 usp_port_mode;
+	u32 usp_pff_inst_id;
+	u32 vep_pff_inst_id;
+	u32 dsp_pff_inst_id[47];
+	u32 reserved1[11];
+	u16 vep_vector_number;
+	u16 usp_vector_number;
+	u32 port_event_bitmap;
+	u32 reserved2[3];
+	u32 part_event_summary;
+	u32 reserved3[3];
+	u32 part_reset_hdr;
+	u32 part_reset_data[5];
+	u32 mrpc_comp_hdr;
+	u32 mrpc_comp_data[5];
+	u32 mrpc_comp_async_hdr;
+	u32 mrpc_comp_async_data[5];
+	u32 dyn_binding_hdr;
+	u32 dyn_binding_data[5];
+	u32 reserved4[159];
+} __packed;
+
+enum {
+	SWITCHTEC_PART_CFG_EVENT_RESET = 1 << 0,
+	SWITCHTEC_PART_CFG_EVENT_MRPC_CMP = 1 << 1,
+	SWITCHTEC_PART_CFG_EVENT_MRPC_ASYNC_CMP = 1 << 2,
+	SWITCHTEC_PART_CFG_EVENT_DYN_PART_CMP = 1 << 3,
+};
+
+struct pff_csr_regs {
+	u16 vendor_id;
+	u16 device_id;
+	u32 pci_cfg_header[15];
+	u32 pci_cap_region[48];
+	u32 pcie_cap_region[448];
+	u32 indirect_gas_window[128];
+	u32 indirect_gas_window_off;
+	u32 reserved[127];
+	u32 pff_event_summary;
+	u32 reserved2[3];
+	u32 aer_in_p2p_hdr;
+	u32 aer_in_p2p_data[5];
+	u32 aer_in_vep_hdr;
+	u32 aer_in_vep_data[5];
+	u32 dpc_hdr;
+	u32 dpc_data[5];
+	u32 cts_hdr;
+	u32 cts_data[5];
+	u32 reserved3[6];
+	u32 hotplug_hdr;
+	u32 hotplug_data[5];
+	u32 ier_hdr;
+	u32 ier_data[5];
+	u32 threshold_hdr;
+	u32 threshold_data[5];
+	u32 power_mgmt_hdr;
+	u32 power_mgmt_data[5];
+	u32 tlp_throttling_hdr;
+	u32 tlp_throttling_data[5];
+	u32 force_speed_hdr;
+	u32 force_speed_data[5];
+	u32 credit_timeout_hdr;
+	u32 credit_timeout_data[5];
+	u32 link_state_hdr;
+	u32 link_state_data[5];
+	u32 reserved4[174];
+} __packed;
+
+struct switchtec_dev {
+	struct pci_dev *pdev;
+	struct device dev;
+	struct cdev cdev;
+
+	int partition;
+	int partition_count;
+	int pff_csr_count;
+	char pff_local[SWITCHTEC_MAX_PFF_CSR];
+
+	void __iomem *mmio;
+	struct mrpc_regs __iomem *mmio_mrpc;
+	struct sw_event_regs __iomem *mmio_sw_event;
+	struct sys_info_regs __iomem *mmio_sys_info;
+	struct flash_info_regs __iomem *mmio_flash_info;
+	struct ntb_info_regs __iomem *mmio_ntb;
+	struct part_cfg_regs __iomem *mmio_part_cfg;
+	struct part_cfg_regs __iomem *mmio_part_cfg_all;
+	struct pff_csr_regs __iomem *mmio_pff_csr;
+
+	/*
+	 * The mrpc mutex must be held when accessing the other
+	 * mrpc_ fields, alive flag and stuser->state field
+	 */
+	struct mutex mrpc_mutex;
+	struct list_head mrpc_queue;
+	int mrpc_busy;
+	struct work_struct mrpc_work;
+	struct delayed_work mrpc_timeout;
+	bool alive;
+
+	wait_queue_head_t event_wq;
+	atomic_t event_cnt;
+};
+
+static struct switchtec_dev *to_stdev(struct device *dev)
+{
+	return container_of(dev, struct switchtec_dev, dev);
+}
+
+enum mrpc_state {
+	MRPC_IDLE = 0,
+	MRPC_QUEUED,
+	MRPC_RUNNING,
+	MRPC_DONE,
+};
+
+struct switchtec_user {
+	struct switchtec_dev *stdev;
+
+	enum mrpc_state state;
+
+	struct completion comp;
+	struct kref kref;
+	struct list_head list;
+
+	u32 cmd;
+	u32 status;
+	u32 return_code;
+	size_t data_len;
+	size_t read_len;
+	unsigned char data[SWITCHTEC_MRPC_PAYLOAD_SIZE];
+	int event_cnt;
+};
+
+static struct switchtec_user *stuser_create(struct switchtec_dev *stdev)
+{
+	struct switchtec_user *stuser;
+
+	stuser = kzalloc(sizeof(*stuser), GFP_KERNEL);
+	if (!stuser)
+		return ERR_PTR(-ENOMEM);
+
+	get_device(&stdev->dev);
+	stuser->stdev = stdev;
+	kref_init(&stuser->kref);
+	INIT_LIST_HEAD(&stuser->list);
+	init_completion(&stuser->comp);
+	stuser->event_cnt = atomic_read(&stdev->event_cnt);
+
+	dev_dbg(&stdev->dev, "%s: %p\n", __func__, stuser);
+
+	return stuser;
+}
+
+static void stuser_free(struct kref *kref)
+{
+	struct switchtec_user *stuser;
+
+	stuser = container_of(kref, struct switchtec_user, kref);
+
+	dev_dbg(&stuser->stdev->dev, "%s: %p\n", __func__, stuser);
+
+	put_device(&stuser->stdev->dev);
+	kfree(stuser);
+}
+
+static void stuser_put(struct switchtec_user *stuser)
+{
+	kref_put(&stuser->kref, stuser_free);
+}
+
+static void stuser_set_state(struct switchtec_user *stuser,
+			     enum mrpc_state state)
+{
+	/* requires the mrpc_mutex to already be held when called */
+
+	const char * const state_names[] = {
+		[MRPC_IDLE] = "IDLE",
+		[MRPC_QUEUED] = "QUEUED",
+		[MRPC_RUNNING] = "RUNNING",
+		[MRPC_DONE] = "DONE",
+	};
+
+	stuser->state = state;
+
+	dev_dbg(&stuser->stdev->dev, "stuser state %p -> %s",
+		stuser, state_names[state]);
+}
+
+static void mrpc_complete_cmd(struct switchtec_dev *stdev);
+
+static void mrpc_cmd_submit(struct switchtec_dev *stdev)
+{
+	/* requires the mrpc_mutex to already be held when called */
+
+	struct switchtec_user *stuser;
+
+	if (stdev->mrpc_busy)
+		return;
+
+	if (list_empty(&stdev->mrpc_queue))
+		return;
+
+	stuser = list_entry(stdev->mrpc_queue.next, struct switchtec_user,
+			    list);
+
+	stuser_set_state(stuser, MRPC_RUNNING);
+	stdev->mrpc_busy = 1;
+	memcpy_toio(&stdev->mmio_mrpc->input_data,
+		    stuser->data, stuser->data_len);
+	iowrite32(stuser->cmd, &stdev->mmio_mrpc->cmd);
+
+	stuser->status = ioread32(&stdev->mmio_mrpc->status);
+	if (stuser->status != SWITCHTEC_MRPC_STATUS_INPROGRESS)
+		mrpc_complete_cmd(stdev);
+
+	schedule_delayed_work(&stdev->mrpc_timeout,
+			      msecs_to_jiffies(500));
+}
+
+static int mrpc_queue_cmd(struct switchtec_user *stuser)
+{
+	/* requires the mrpc_mutex to already be held when called */
+
+	struct switchtec_dev *stdev = stuser->stdev;
+
+	kref_get(&stuser->kref);
+	stuser->read_len = sizeof(stuser->data);
+	stuser_set_state(stuser, MRPC_QUEUED);
+	init_completion(&stuser->comp);
+	list_add_tail(&stuser->list, &stdev->mrpc_queue);
+
+	mrpc_cmd_submit(stdev);
+
+	return 0;
+}
+
+static void mrpc_complete_cmd(struct switchtec_dev *stdev)
+{
+	/* requires the mrpc_mutex to already be held when called */
+	struct switchtec_user *stuser;
+
+	if (list_empty(&stdev->mrpc_queue))
+		return;
+
+	stuser = list_entry(stdev->mrpc_queue.next, struct switchtec_user,
+			    list);
+
+	stuser->status = ioread32(&stdev->mmio_mrpc->status);
+	if (stuser->status == SWITCHTEC_MRPC_STATUS_INPROGRESS)
+		return;
+
+	stuser_set_state(stuser, MRPC_DONE);
+	stuser->return_code = 0;
+
+	if (stuser->status != SWITCHTEC_MRPC_STATUS_DONE)
+		goto out;
+
+	stuser->return_code = ioread32(&stdev->mmio_mrpc->ret_value);
+	if (stuser->return_code != 0)
+		goto out;
+
+	memcpy_fromio(stuser->data, &stdev->mmio_mrpc->output_data,
+		      stuser->read_len);
+
+out:
+	complete_all(&stuser->comp);
+	list_del_init(&stuser->list);
+	stuser_put(stuser);
+	stdev->mrpc_busy = 0;
+
+	mrpc_cmd_submit(stdev);
+}
+
+static void mrpc_event_work(struct work_struct *work)
+{
+	struct switchtec_dev *stdev;
+
+	stdev = container_of(work, struct switchtec_dev, mrpc_work);
+
+	dev_dbg(&stdev->dev, "%s\n", __func__);
+
+	mutex_lock(&stdev->mrpc_mutex);
+	cancel_delayed_work(&stdev->mrpc_timeout);
+	mrpc_complete_cmd(stdev);
+	mutex_unlock(&stdev->mrpc_mutex);
+}
+
+static void mrpc_timeout_work(struct work_struct *work)
+{
+	struct switchtec_dev *stdev;
+	u32 status;
+
+	stdev = container_of(work, struct switchtec_dev, mrpc_timeout.work);
+
+	dev_dbg(&stdev->dev, "%s\n", __func__);
+
+	mutex_lock(&stdev->mrpc_mutex);
+
+	status = ioread32(&stdev->mmio_mrpc->status);
+	if (status == SWITCHTEC_MRPC_STATUS_INPROGRESS) {
+		schedule_delayed_work(&stdev->mrpc_timeout,
+				      msecs_to_jiffies(500));
+		goto out;
+	}
+
+	mrpc_complete_cmd(stdev);
+
+out:
+	mutex_unlock(&stdev->mrpc_mutex);
+}
+
+static int switchtec_dev_open(struct inode *inode, struct file *filp)
+{
+	struct switchtec_dev *stdev;
+	struct switchtec_user *stuser;
+
+	stdev = container_of(inode->i_cdev, struct switchtec_dev, cdev);
+
+	stuser = stuser_create(stdev);
+	if (IS_ERR(stuser))
+		return PTR_ERR(stuser);
+
+	filp->private_data = stuser;
+	nonseekable_open(inode, filp);
+
+	dev_dbg(&stdev->dev, "%s: %p\n", __func__, stuser);
+
+	return 0;
+}
+
+static int switchtec_dev_release(struct inode *inode, struct file *filp)
+{
+	struct switchtec_user *stuser = filp->private_data;
+
+	stuser_put(stuser);
+
+	return 0;
+}
+
+static int lock_mutex_and_test_alive(struct switchtec_dev *stdev)
+{
+	if (mutex_lock_interruptible(&stdev->mrpc_mutex))
+		return -EINTR;
+
+	if (!stdev->alive) {
+		mutex_unlock(&stdev->mrpc_mutex);
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+static ssize_t switchtec_dev_write(struct file *filp, const char __user *data,
+				   size_t size, loff_t *off)
+{
+	struct switchtec_user *stuser = filp->private_data;
+	struct switchtec_dev *stdev = stuser->stdev;
+	int rc;
+
+	if (size < sizeof(stuser->cmd) ||
+	    size > sizeof(stuser->cmd) + sizeof(stuser->data))
+		return -EINVAL;
+
+	stuser->data_len = size - sizeof(stuser->cmd);
+
+	rc = lock_mutex_and_test_alive(stdev);
+	if (rc)
+		return rc;
+
+	if (stuser->state != MRPC_IDLE) {
+		rc = -EBADE;
+		goto out;
+	}
+
+	rc = copy_from_user(&stuser->cmd, data, sizeof(stuser->cmd));
+	if (rc) {
+		rc = -EFAULT;
+		goto out;
+	}
+
+	data += sizeof(stuser->cmd);
+	rc = copy_from_user(&stuser->data, data, size - sizeof(stuser->cmd));
+	if (rc) {
+		rc = -EFAULT;
+		goto out;
+	}
+
+	rc = mrpc_queue_cmd(stuser);
+
+out:
+	mutex_unlock(&stdev->mrpc_mutex);
+
+	if (rc)
+		return rc;
+
+	return size;
+}
+
+static ssize_t switchtec_dev_read(struct file *filp, char __user *data,
+				  size_t size, loff_t *off)
+{
+	struct switchtec_user *stuser = filp->private_data;
+	struct switchtec_dev *stdev = stuser->stdev;
+	int rc;
+
+	if (size < sizeof(stuser->cmd) ||
+	    size > sizeof(stuser->cmd) + sizeof(stuser->data))
+		return -EINVAL;
+
+	rc = lock_mutex_and_test_alive(stdev);
+	if (rc)
+		return rc;
+
+	if (stuser->state == MRPC_IDLE) {
+		mutex_unlock(&stdev->mrpc_mutex);
+		return -EBADE;
+	}
+
+	stuser->read_len = size - sizeof(stuser->return_code);
+
+	mutex_unlock(&stdev->mrpc_mutex);
+
+	if (filp->f_flags & O_NONBLOCK) {
+		if (!try_wait_for_completion(&stuser->comp))
+			return -EAGAIN;
+	} else {
+		rc = wait_for_completion_interruptible(&stuser->comp);
+		if (rc < 0)
+			return rc;
+	}
+
+	rc = lock_mutex_and_test_alive(stdev);
+	if (rc)
+		return rc;
+
+	if (stuser->state != MRPC_DONE) {
+		mutex_unlock(&stdev->mrpc_mutex);
+		return -EBADE;
+	}
+
+	rc = copy_to_user(data, &stuser->return_code,
+			  sizeof(stuser->return_code));
+	if (rc) {
+		rc = -EFAULT;
+		goto out;
+	}
+
+	data += sizeof(stuser->return_code);
+	rc = copy_to_user(data, &stuser->data,
+			  size - sizeof(stuser->return_code));
+	if (rc) {
+		rc = -EFAULT;
+		goto out;
+	}
+
+	stuser_set_state(stuser, MRPC_IDLE);
+
+out:
+	mutex_unlock(&stdev->mrpc_mutex);
+
+	if (stuser->status == SWITCHTEC_MRPC_STATUS_DONE)
+		return size;
+	else if (stuser->status == SWITCHTEC_MRPC_STATUS_INTERRUPTED)
+		return -ENXIO;
+	else
+		return -EBADMSG;
+}
+
+static unsigned int switchtec_dev_poll(struct file *filp, poll_table *wait)
+{
+	struct switchtec_user *stuser = filp->private_data;
+	struct switchtec_dev *stdev = stuser->stdev;
+	int ret = 0;
+
+	poll_wait(filp, &stuser->comp.wait, wait);
+	poll_wait(filp, &stdev->event_wq, wait);
+
+	if (lock_mutex_and_test_alive(stdev))
+		return POLLIN | POLLRDHUP | POLLOUT | POLLERR | POLLHUP;
+
+	mutex_unlock(&stdev->mrpc_mutex);
+
+	if (try_wait_for_completion(&stuser->comp))
+		ret |= POLLIN | POLLRDNORM;
+
+	if (stuser->event_cnt != atomic_read(&stdev->event_cnt))
+		ret |= POLLPRI | POLLRDBAND;
+
+	return ret;
+}
+
+static const struct file_operations switchtec_fops = {
+	.owner = THIS_MODULE,
+	.open = switchtec_dev_open,
+	.release = switchtec_dev_release,
+	.write = switchtec_dev_write,
+	.read = switchtec_dev_read,
+	.poll = switchtec_dev_poll,
+};
+
+static void stdev_release(struct device *dev)
+{
+	struct switchtec_dev *stdev = to_stdev(dev);
+
+	kfree(stdev);
+}
+
+static void stdev_kill(struct switchtec_dev *stdev)
+{
+	struct switchtec_user *stuser, *tmpuser;
+
+	pci_clear_master(stdev->pdev);
+
+	cancel_delayed_work_sync(&stdev->mrpc_timeout);
+
+	/* Mark the hardware as unavailable and complete all completions */
+	mutex_lock(&stdev->mrpc_mutex);
+	stdev->alive = false;
+
+	/* Wake up and kill any users waiting on an MRPC request */
+	list_for_each_entry_safe(stuser, tmpuser, &stdev->mrpc_queue, list) {
+		complete_all(&stuser->comp);
+		list_del_init(&stuser->list);
+		stuser_put(stuser);
+	}
+
+	mutex_unlock(&stdev->mrpc_mutex);
+
+	/* Wake up any users waiting on event_wq */
+	wake_up_interruptible(&stdev->event_wq);
+}
+
+static struct switchtec_dev *stdev_create(struct pci_dev *pdev)
+{
+	struct switchtec_dev *stdev;
+	int minor;
+	struct device *dev;
+	struct cdev *cdev;
+	int rc;
+
+	stdev = kzalloc_node(sizeof(*stdev), GFP_KERNEL,
+			     dev_to_node(&pdev->dev));
+	if (!stdev)
+		return ERR_PTR(-ENOMEM);
+
+	stdev->alive = true;
+	stdev->pdev = pdev;
+	INIT_LIST_HEAD(&stdev->mrpc_queue);
+	mutex_init(&stdev->mrpc_mutex);
+	stdev->mrpc_busy = 0;
+	INIT_WORK(&stdev->mrpc_work, mrpc_event_work);
+	INIT_DELAYED_WORK(&stdev->mrpc_timeout, mrpc_timeout_work);
+	init_waitqueue_head(&stdev->event_wq);
+	atomic_set(&stdev->event_cnt, 0);
+
+	dev = &stdev->dev;
+	device_initialize(dev);
+	dev->class = switchtec_class;
+	dev->parent = &pdev->dev;
+	dev->release = stdev_release;
+
+	minor = ida_simple_get(&switchtec_minor_ida, 0, 0,
+			       GFP_KERNEL);
+	if (minor < 0) {
+		rc = minor;
+		goto err_put;
+	}
+
+	dev->devt = MKDEV(MAJOR(switchtec_devt), minor);
+	dev_set_name(dev, "switchtec%d", minor);
+
+	cdev = &stdev->cdev;
+	cdev_init(cdev, &switchtec_fops);
+	cdev->owner = THIS_MODULE;
+	cdev->kobj.parent = &dev->kobj;
+
+	return stdev;
+
+err_put:
+	put_device(&stdev->dev);
+	return ERR_PTR(rc);
+}
+
+static irqreturn_t switchtec_event_isr(int irq, void *dev)
+{
+	struct switchtec_dev *stdev = dev;
+	u32 reg;
+	irqreturn_t ret = IRQ_NONE;
+
+	reg = ioread32(&stdev->mmio_part_cfg->mrpc_comp_hdr);
+	if (reg & SWITCHTEC_EVENT_OCCURRED) {
+		dev_dbg(&stdev->dev, "%s: mrpc comp\n", __func__);
+		ret = IRQ_HANDLED;
+		schedule_work(&stdev->mrpc_work);
+		iowrite32(reg, &stdev->mmio_part_cfg->mrpc_comp_hdr);
+	}
+
+	return ret;
+}
+
+static int switchtec_init_isr(struct switchtec_dev *stdev)
+{
+	int nvecs;
+	int event_irq;
+
+	nvecs = pci_alloc_irq_vectors(stdev->pdev, 1, 4,
+				      PCI_IRQ_MSIX | PCI_IRQ_MSI);
+	if (nvecs < 0)
+		return nvecs;
+
+	event_irq = ioread32(&stdev->mmio_part_cfg->vep_vector_number);
+	if (event_irq < 0 || event_irq >= nvecs)
+		return -EFAULT;
+
+	event_irq = pci_irq_vector(stdev->pdev, event_irq);
+	if (event_irq < 0)
+		return event_irq;
+
+	return devm_request_irq(&stdev->pdev->dev, event_irq,
+				switchtec_event_isr, 0,
+				KBUILD_MODNAME, stdev);
+}
+
+static void init_pff(struct switchtec_dev *stdev)
+{
+	int i;
+	u32 reg;
+	struct part_cfg_regs *pcfg = stdev->mmio_part_cfg;
+
+	for (i = 0; i < SWITCHTEC_MAX_PFF_CSR; i++) {
+		reg = ioread16(&stdev->mmio_pff_csr[i].vendor_id);
+		if (reg != MICROSEMI_VENDOR_ID)
+			break;
+	}
+
+	stdev->pff_csr_count = i;
+
+	reg = ioread32(&pcfg->usp_pff_inst_id);
+	if (reg < SWITCHTEC_MAX_PFF_CSR)
+		stdev->pff_local[reg] = 1;
+
+	reg = ioread32(&pcfg->vep_pff_inst_id);
+	if (reg < SWITCHTEC_MAX_PFF_CSR)
+		stdev->pff_local[reg] = 1;
+
+	for (i = 0; i < ARRAY_SIZE(pcfg->dsp_pff_inst_id); i++) {
+		reg = ioread32(&pcfg->dsp_pff_inst_id[i]);
+		if (reg < SWITCHTEC_MAX_PFF_CSR)
+			stdev->pff_local[reg] = 1;
+	}
+}
+
+static int switchtec_init_pci(struct switchtec_dev *stdev,
+			      struct pci_dev *pdev)
+{
+	int rc;
+
+	rc = pcim_enable_device(pdev);
+	if (rc)
+		return rc;
+
+	rc = pcim_iomap_regions(pdev, 0x1, KBUILD_MODNAME);
+	if (rc)
+		return rc;
+
+	pci_set_master(pdev);
+
+	stdev->mmio = pcim_iomap_table(pdev)[0];
+	stdev->mmio_mrpc = stdev->mmio + SWITCHTEC_GAS_MRPC_OFFSET;
+	stdev->mmio_sw_event = stdev->mmio + SWITCHTEC_GAS_SW_EVENT_OFFSET;
+	stdev->mmio_sys_info = stdev->mmio + SWITCHTEC_GAS_SYS_INFO_OFFSET;
+	stdev->mmio_flash_info = stdev->mmio + SWITCHTEC_GAS_FLASH_INFO_OFFSET;
+	stdev->mmio_ntb = stdev->mmio + SWITCHTEC_GAS_NTB_OFFSET;
+	stdev->partition = ioread8(&stdev->mmio_ntb->partition_id);
+	stdev->partition_count = ioread8(&stdev->mmio_ntb->partition_count);
+	stdev->mmio_part_cfg_all = stdev->mmio + SWITCHTEC_GAS_PART_CFG_OFFSET;
+	stdev->mmio_part_cfg = &stdev->mmio_part_cfg_all[stdev->partition];
+	stdev->mmio_pff_csr = stdev->mmio + SWITCHTEC_GAS_PFF_CSR_OFFSET;
+
+	init_pff(stdev);
+
+	pci_set_drvdata(pdev, stdev);
+
+	return 0;
+}
+
+static int switchtec_pci_probe(struct pci_dev *pdev,
+			       const struct pci_device_id *id)
+{
+	struct switchtec_dev *stdev;
+	int rc;
+
+	stdev = stdev_create(pdev);
+	if (IS_ERR(stdev))
+		return PTR_ERR(stdev);
+
+	rc = switchtec_init_pci(stdev, pdev);
+	if (rc)
+		goto err_put;
+
+	rc = switchtec_init_isr(stdev);
+	if (rc) {
+		dev_err(&stdev->dev, "failed to init isr.\n");
+		goto err_put;
+	}
+
+	iowrite32(SWITCHTEC_EVENT_CLEAR |
+		  SWITCHTEC_EVENT_EN_IRQ,
+		  &stdev->mmio_part_cfg->mrpc_comp_hdr);
+
+	rc = cdev_add(&stdev->cdev, stdev->dev.devt, 1);
+	if (rc)
+		goto err_put;
+
+	rc = device_add(&stdev->dev);
+	if (rc)
+		goto err_devadd;
+
+	dev_info(&stdev->dev, "Management device registered.\n");
+
+	return 0;
+
+err_devadd:
+	cdev_del(&stdev->cdev);
+	stdev_kill(stdev);
+err_put:
+	ida_simple_remove(&switchtec_minor_ida, MINOR(stdev->dev.devt));
+	put_device(&stdev->dev);
+	return rc;
+}
+
+static void switchtec_pci_remove(struct pci_dev *pdev)
+{
+	struct switchtec_dev *stdev = pci_get_drvdata(pdev);
+
+	pci_set_drvdata(pdev, NULL);
+
+	device_del(&stdev->dev);
+	cdev_del(&stdev->cdev);
+	ida_simple_remove(&switchtec_minor_ida, MINOR(stdev->dev.devt));
+	dev_info(&stdev->dev, "unregistered.\n");
+
+	stdev_kill(stdev);
+	put_device(&stdev->dev);
+}
+
+#define SWITCHTEC_PCI_DEVICE(device_id) \
+	{ \
+		.vendor     = MICROSEMI_VENDOR_ID, \
+		.device     = device_id, \
+		.subvendor  = PCI_ANY_ID, \
+		.subdevice  = PCI_ANY_ID, \
+		.class      = MICROSEMI_MGMT_CLASSCODE, \
+		.class_mask = 0xFFFFFFFF, \
+	}, \
+	{ \
+		.vendor     = MICROSEMI_VENDOR_ID, \
+		.device     = device_id, \
+		.subvendor  = PCI_ANY_ID, \
+		.subdevice  = PCI_ANY_ID, \
+		.class      = MICROSEMI_NTB_CLASSCODE, \
+		.class_mask = 0xFFFFFFFF, \
+	}
+
+static const struct pci_device_id switchtec_pci_tbl[] = {
+	SWITCHTEC_PCI_DEVICE(0x8531),  //PFX 24xG3
+	SWITCHTEC_PCI_DEVICE(0x8532),  //PFX 32xG3
+	SWITCHTEC_PCI_DEVICE(0x8533),  //PFX 48xG3
+	SWITCHTEC_PCI_DEVICE(0x8534),  //PFX 64xG3
+	SWITCHTEC_PCI_DEVICE(0x8535),  //PFX 80xG3
+	SWITCHTEC_PCI_DEVICE(0x8536),  //PFX 96xG3
+	SWITCHTEC_PCI_DEVICE(0x8543),  //PSX 48xG3
+	SWITCHTEC_PCI_DEVICE(0x8544),  //PSX 64xG3
+	SWITCHTEC_PCI_DEVICE(0x8545),  //PSX 80xG3
+	SWITCHTEC_PCI_DEVICE(0x8546),  //PSX 96xG3
+	{0}
+};
+MODULE_DEVICE_TABLE(pci, switchtec_pci_tbl);
+
+static struct pci_driver switchtec_pci_driver = {
+	.name		= KBUILD_MODNAME,
+	.id_table	= switchtec_pci_tbl,
+	.probe		= switchtec_pci_probe,
+	.remove		= switchtec_pci_remove,
+};
+
+static int __init switchtec_init(void)
+{
+	int rc;
+
+	rc = alloc_chrdev_region(&switchtec_devt, 0, max_devices,
+				 "switchtec");
+	if (rc)
+		return rc;
+
+	switchtec_class = class_create(THIS_MODULE, "switchtec");
+	if (IS_ERR(switchtec_class)) {
+		rc = PTR_ERR(switchtec_class);
+		goto err_create_class;
+	}
+
+	rc = pci_register_driver(&switchtec_pci_driver);
+	if (rc)
+		goto err_pci_register;
+
+	pr_info(KBUILD_MODNAME ": loaded.\n");
+
+	return 0;
+
+err_pci_register:
+	class_destroy(switchtec_class);
+
+err_create_class:
+	unregister_chrdev_region(switchtec_devt, max_devices);
+
+	return rc;
+}
+module_init(switchtec_init);
+
+static void __exit switchtec_exit(void)
+{
+	pci_unregister_driver(&switchtec_pci_driver);
+	class_destroy(switchtec_class);
+	unregister_chrdev_region(switchtec_devt, max_devices);
+	ida_destroy(&switchtec_minor_ida);
+
+	pr_info(KBUILD_MODNAME ": unloaded.\n");
+}
+module_exit(switchtec_exit);