Merge branch 'for-3.3/drivers' of git://git.kernel.dk/linux-block

* 'for-3.3/drivers' of git://git.kernel.dk/linux-block: mtip32xx: do rebuild monitoring asynchronously xen-blkfront: Use kcalloc instead of kzalloc to allocate array mtip32xx: uninitialized variable in mtip_quiesce_io() mtip32xx: updates based on feedback xen-blkback: convert hole punching to discard request on loop devices xen/blkback: Move processing of BLKIF_OP_DISCARD from dispatch_rw_block_io xen/blk[front|back]: Enhance discard support with secure erasing support. xen/blk[front|back]: Squash blkif_request_rw and blkif_request_discard together mtip32xx: update to new ->make_request() API mtip32xx: add module.h include to avoid conflict with moduleh tree mtip32xx: mark a few more items static mtip32xx: ensure that all local functions are static mtip32xx: cleanup compat ioctl handling mtip32xx: fix warnings/errors on 32-bit compiles block: Add driver for Micron RealSSD pcie flash cards

Merge branch 'for-3.3/drivers' of git://git.kernel.dk/linux-block
* 'for-3.3/drivers' of git://git.kernel.dk/linux-block: mtip32xx: do rebuild monitoring asynchronously xen-blkfront: Use kcalloc instead of kzalloc to allocate array mtip32xx: uninitialized variable in mtip_quiesce_io() mtip32xx: updates based on feedback xen-blkback: convert hole punching to discard request on loop devices xen/blkback: Move processing of BLKIF_OP_DISCARD from dispatch_rw_block_io xen/blk[front|back]: Enhance discard support with secure erasing support. xen/blk[front|back]: Squash blkif_request_rw and blkif_request_discard together mtip32xx: update to new ->make_request() API mtip32xx: add module.h include to avoid conflict with moduleh tree mtip32xx: mark a few more items static mtip32xx: ensure that all local functions are static mtip32xx: cleanup compat ioctl handling mtip32xx: fix warnings/errors on 32-bit compiles block: Add driver for Micron RealSSD pcie flash cards
Linus Torvalds
2 parents b3c9dd182e 85a0f7b220
Showing 11 changed files Side-by-side Diff
drivers/block/Kconfig
drivers/block/Makefile
drivers/block/mtip32xx/Kconfig
drivers/block/mtip32xx/Makefile
drivers/block/mtip32xx/mtip32xx.c
drivers/block/mtip32xx/mtip32xx.h
drivers/block/xen-blkback/blkback.c
drivers/block/xen-blkback/common.h
drivers/block/xen-blkback/xenbus.c
drivers/block/xen-blkfront.c
include/xen/interface/io/blkif.h
@@ -116,6 +116,8 @@
  
 source "drivers/block/paride/Kconfig"
  
+source "drivers/block/mtip32xx/Kconfig"
+
 config BLK_CPQ_DA
 	tristate "Compaq SMART2 support"
 	depends on PCI && VIRT_TO_BUS
@@ -39,6 +39,7 @@
 obj-$(CONFIG_XEN_BLKDEV_BACKEND)	+= xen-blkback/
 obj-$(CONFIG_BLK_DEV_DRBD)     += drbd/
 obj-$(CONFIG_BLK_DEV_RBD)     += rbd.o
+obj-$(CONFIG_BLK_DEV_PCIESSD_MTIP32XX)	+= mtip32xx/
  
 swim_mod-y	:= swim.o swim_asm.o
+#
+# mtip32xx device driver configuration
+#
+
+config BLK_DEV_PCIESSD_MTIP32XX
+	tristate "Block Device Driver for Micron PCIe SSDs"
+	depends on HOTPLUG_PCI_PCIE
+	help
+          This enables the block driver for Micron PCIe SSDs.
+#
+# Makefile for  Block device driver for Micron PCIe SSD
+#
+
+obj-$(CONFIG_BLK_DEV_PCIESSD_MTIP32XX) += mtip32xx.o
+/*
+ * Driver for the Micron P320 SSD
+ *   Copyright (C) 2011 Micron Technology, Inc.
+ *
+ * Portions of this code were derived from works subjected to the
+ * following copyright:
+ *    Copyright (C) 2009 Integrated Device Technology, Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * 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.
+ *
+ */
+
+#include <linux/pci.h>
+#include <linux/interrupt.h>
+#include <linux/ata.h>
+#include <linux/delay.h>
+#include <linux/hdreg.h>
+#include <linux/uaccess.h>
+#include <linux/random.h>
+#include <linux/smp.h>
+#include <linux/compat.h>
+#include <linux/fs.h>
+#include <linux/module.h>
+#include <linux/genhd.h>
+#include <linux/blkdev.h>
+#include <linux/bio.h>
+#include <linux/dma-mapping.h>
+#include <linux/idr.h>
+#include <linux/kthread.h>
+#include <../drivers/ata/ahci.h>
+#include "mtip32xx.h"
+
+#define HW_CMD_SLOT_SZ		(MTIP_MAX_COMMAND_SLOTS * 32)
+#define HW_CMD_TBL_SZ		(AHCI_CMD_TBL_HDR_SZ + (MTIP_MAX_SG * 16))
+#define HW_CMD_TBL_AR_SZ	(HW_CMD_TBL_SZ * MTIP_MAX_COMMAND_SLOTS)
+#define HW_PORT_PRIV_DMA_SZ \
+		(HW_CMD_SLOT_SZ + HW_CMD_TBL_AR_SZ + AHCI_RX_FIS_SZ)
+
+#define HOST_HSORG		0xFC
+#define HSORG_DISABLE_SLOTGRP_INTR (1<<24)
+#define HSORG_DISABLE_SLOTGRP_PXIS (1<<16)
+#define HSORG_HWREV		0xFF00
+#define HSORG_STYLE		0x8
+#define HSORG_SLOTGROUPS	0x7
+
+#define PORT_COMMAND_ISSUE	0x38
+#define PORT_SDBV		0x7C
+
+#define PORT_OFFSET		0x100
+#define PORT_MEM_SIZE		0x80
+
+#define PORT_IRQ_ERR \
+	(PORT_IRQ_HBUS_ERR | PORT_IRQ_IF_ERR | PORT_IRQ_CONNECT | \
+	 PORT_IRQ_PHYRDY | PORT_IRQ_UNK_FIS | PORT_IRQ_BAD_PMP | \
+	 PORT_IRQ_TF_ERR | PORT_IRQ_HBUS_DATA_ERR | PORT_IRQ_IF_NONFATAL | \
+	 PORT_IRQ_OVERFLOW)
+#define PORT_IRQ_LEGACY \
+	(PORT_IRQ_PIOS_FIS | PORT_IRQ_D2H_REG_FIS)
+#define PORT_IRQ_HANDLED \
+	(PORT_IRQ_SDB_FIS | PORT_IRQ_LEGACY | \
+	 PORT_IRQ_TF_ERR | PORT_IRQ_IF_ERR | \
+	 PORT_IRQ_CONNECT | PORT_IRQ_PHYRDY)
+#define DEF_PORT_IRQ \
+	(PORT_IRQ_ERR | PORT_IRQ_LEGACY | PORT_IRQ_SDB_FIS)
+
+/* product numbers */
+#define MTIP_PRODUCT_UNKNOWN	0x00
+#define MTIP_PRODUCT_ASICFPGA	0x11
+
+/* Device instance number, incremented each time a device is probed. */
+static int instance;
+
+/*
+ * Global variable used to hold the major block device number
+ * allocated in mtip_init().
+ */
+static int mtip_major;
+
+static DEFINE_SPINLOCK(rssd_index_lock);
+static DEFINE_IDA(rssd_index_ida);
+
+static int mtip_block_initialize(struct driver_data *dd);
+
+#ifdef CONFIG_COMPAT
+struct mtip_compat_ide_task_request_s {
+	__u8		io_ports[8];
+	__u8		hob_ports[8];
+	ide_reg_valid_t	out_flags;
+	ide_reg_valid_t	in_flags;
+	int		data_phase;
+	int		req_cmd;
+	compat_ulong_t	out_size;
+	compat_ulong_t	in_size;
+};
+#endif
+
+/*
+ * This function check_for_surprise_removal is called
+ * while card is removed from the system and it will
+ * read the vendor id from the configration space
+ *
+ * @pdev Pointer to the pci_dev structure.
+ *
+ * return value
+ *	 true if device removed, else false
+ */
+static bool mtip_check_surprise_removal(struct pci_dev *pdev)
+{
+	u16 vendor_id = 0;
+
+       /* Read the vendorID from the configuration space */
+	pci_read_config_word(pdev, 0x00, &vendor_id);
+	if (vendor_id == 0xFFFF)
+		return true; /* device removed */
+
+	return false; /* device present */
+}
+
+/*
+ * This function is called for clean the pending command in the
+ * command slot during the surprise removal of device and return
+ * error to the upper layer.
+ *
+ * @dd Pointer to the DRIVER_DATA structure.
+ *
+ * return value
+ *	None
+ */
+static void mtip_command_cleanup(struct driver_data *dd)
+{
+	int group = 0, commandslot = 0, commandindex = 0;
+	struct mtip_cmd *command;
+	struct mtip_port *port = dd->port;
+
+	for (group = 0; group < 4; group++) {
+		for (commandslot = 0; commandslot < 32; commandslot++) {
+			if (!(port->allocated[group] & (1 << commandslot)))
+				continue;
+
+			commandindex = group << 5 | commandslot;
+			command = &port->commands[commandindex];
+
+			if (atomic_read(&command->active)
+			    && (command->async_callback)) {
+				command->async_callback(command->async_data,
+					-ENODEV);
+				command->async_callback = NULL;
+				command->async_data = NULL;
+			}
+
+			dma_unmap_sg(&port->dd->pdev->dev,
+				command->sg,
+				command->scatter_ents,
+				command->direction);
+		}
+	}
+
+	up(&port->cmd_slot);
+
+	atomic_set(&dd->drv_cleanup_done, true);
+}
+
+/*
+ * Obtain an empty command slot.
+ *
+ * This function needs to be reentrant since it could be called
+ * at the same time on multiple CPUs. The allocation of the
+ * command slot must be atomic.
+ *
+ * @port Pointer to the port data structure.
+ *
+ * return value
+ *	>= 0	Index of command slot obtained.
+ *	-1	No command slots available.
+ */
+static int get_slot(struct mtip_port *port)
+{
+	int slot, i;
+	unsigned int num_command_slots = port->dd->slot_groups * 32;
+
+	/*
+	 * Try 10 times, because there is a small race here.
+	 *  that's ok, because it's still cheaper than a lock.
+	 *
+	 * Race: Since this section is not protected by lock, same bit
+	 * could be chosen by different process contexts running in
+	 * different processor. So instead of costly lock, we are going
+	 * with loop.
+	 */
+	for (i = 0; i < 10; i++) {
+		slot = find_next_zero_bit(port->allocated,
+					 num_command_slots, 1);
+		if ((slot < num_command_slots) &&
+		    (!test_and_set_bit(slot, port->allocated)))
+			return slot;
+	}
+	dev_warn(&port->dd->pdev->dev, "Failed to get a tag.\n");
+
+	if (mtip_check_surprise_removal(port->dd->pdev)) {
+		/* Device not present, clean outstanding commands */
+		mtip_command_cleanup(port->dd);
+	}
+	return -1;
+}
+
+/*
+ * Release a command slot.
+ *
+ * @port Pointer to the port data structure.
+ * @tag  Tag of command to release
+ *
+ * return value
+ *	None
+ */
+static inline void release_slot(struct mtip_port *port, int tag)
+{
+	smp_mb__before_clear_bit();
+	clear_bit(tag, port->allocated);
+	smp_mb__after_clear_bit();
+}
+
+/*
+ * Reset the HBA (without sleeping)
+ *
+ * Just like hba_reset, except does not call sleep, so can be
+ * run from interrupt/tasklet context.
+ *
+ * @dd Pointer to the driver data structure.
+ *
+ * return value
+ *	0	The reset was successful.
+ *	-1	The HBA Reset bit did not clear.
+ */
+static int hba_reset_nosleep(struct driver_data *dd)
+{
+	unsigned long timeout;
+
+	/* Chip quirk: quiesce any chip function */
+	mdelay(10);
+
+	/* Set the reset bit */
+	writel(HOST_RESET, dd->mmio + HOST_CTL);
+
+	/* Flush */
+	readl(dd->mmio + HOST_CTL);
+
+	/*
+	 * Wait 10ms then spin for up to 1 second
+	 * waiting for reset acknowledgement
+	 */
+	timeout = jiffies + msecs_to_jiffies(1000);
+	mdelay(10);
+	while ((readl(dd->mmio + HOST_CTL) & HOST_RESET)
+		 && time_before(jiffies, timeout))
+		mdelay(1);
+
+	if (readl(dd->mmio + HOST_CTL) & HOST_RESET)
+		return -1;
+
+	return 0;
+}
+
+/*
+ * Issue a command to the hardware.
+ *
+ * Set the appropriate bit in the s_active and Command Issue hardware
+ * registers, causing hardware command processing to begin.
+ *
+ * @port Pointer to the port structure.
+ * @tag  The tag of the command to be issued.
+ *
+ * return value
+ *      None
+ */
+static inline void mtip_issue_ncq_command(struct mtip_port *port, int tag)
+{
+	unsigned long flags = 0;
+
+	atomic_set(&port->commands[tag].active, 1);
+
+	spin_lock_irqsave(&port->cmd_issue_lock, flags);
+
+	writel((1 << MTIP_TAG_BIT(tag)),
+			port->s_active[MTIP_TAG_INDEX(tag)]);
+	writel((1 << MTIP_TAG_BIT(tag)),
+			port->cmd_issue[MTIP_TAG_INDEX(tag)]);
+
+	spin_unlock_irqrestore(&port->cmd_issue_lock, flags);
+}
+
+/*
+ * Enable/disable the reception of FIS
+ *
+ * @port   Pointer to the port data structure
+ * @enable 1 to enable, 0 to disable
+ *
+ * return value
+ *	Previous state: 1 enabled, 0 disabled
+ */
+static int mtip_enable_fis(struct mtip_port *port, int enable)
+{
+	u32 tmp;
+
+	/* enable FIS reception */
+	tmp = readl(port->mmio + PORT_CMD);
+	if (enable)
+		writel(tmp | PORT_CMD_FIS_RX, port->mmio + PORT_CMD);
+	else
+		writel(tmp & ~PORT_CMD_FIS_RX, port->mmio + PORT_CMD);
+
+	/* Flush */
+	readl(port->mmio + PORT_CMD);
+
+	return (((tmp & PORT_CMD_FIS_RX) == PORT_CMD_FIS_RX));
+}
+
+/*
+ * Enable/disable the DMA engine
+ *
+ * @port   Pointer to the port data structure
+ * @enable 1 to enable, 0 to disable
+ *
+ * return value
+ *	Previous state: 1 enabled, 0 disabled.
+ */
+static int mtip_enable_engine(struct mtip_port *port, int enable)
+{
+	u32 tmp;
+
+	/* enable FIS reception */
+	tmp = readl(port->mmio + PORT_CMD);
+	if (enable)
+		writel(tmp | PORT_CMD_START, port->mmio + PORT_CMD);
+	else
+		writel(tmp & ~PORT_CMD_START, port->mmio + PORT_CMD);
+
+	readl(port->mmio + PORT_CMD);
+	return (((tmp & PORT_CMD_START) == PORT_CMD_START));
+}
+
+/*
+ * Enables the port DMA engine and FIS reception.
+ *
+ * return value
+ *	None
+ */
+static inline void mtip_start_port(struct mtip_port *port)
+{
+	/* Enable FIS reception */
+	mtip_enable_fis(port, 1);
+
+	/* Enable the DMA engine */
+	mtip_enable_engine(port, 1);
+}
+
+/*
+ * Deinitialize a port by disabling port interrupts, the DMA engine,
+ * and FIS reception.
+ *
+ * @port Pointer to the port structure
+ *
+ * return value
+ *	None
+ */
+static inline void mtip_deinit_port(struct mtip_port *port)
+{
+	/* Disable interrupts on this port */
+	writel(0, port->mmio + PORT_IRQ_MASK);
+
+	/* Disable the DMA engine */
+	mtip_enable_engine(port, 0);
+
+	/* Disable FIS reception */
+	mtip_enable_fis(port, 0);
+}
+
+/*
+ * Initialize a port.
+ *
+ * This function deinitializes the port by calling mtip_deinit_port() and
+ * then initializes it by setting the command header and RX FIS addresses,
+ * clearing the SError register and any pending port interrupts before
+ * re-enabling the default set of port interrupts.
+ *
+ * @port Pointer to the port structure.
+ *
+ * return value
+ *	None
+ */
+static void mtip_init_port(struct mtip_port *port)
+{
+	int i;
+	mtip_deinit_port(port);
+
+	/* Program the command list base and FIS base addresses */
+	if (readl(port->dd->mmio + HOST_CAP) & HOST_CAP_64) {
+		writel((port->command_list_dma >> 16) >> 16,
+			 port->mmio + PORT_LST_ADDR_HI);
+		writel((port->rxfis_dma >> 16) >> 16,
+			 port->mmio + PORT_FIS_ADDR_HI);
+	}
+
+	writel(port->command_list_dma & 0xFFFFFFFF,
+			port->mmio + PORT_LST_ADDR);
+	writel(port->rxfis_dma & 0xFFFFFFFF, port->mmio + PORT_FIS_ADDR);
+
+	/* Clear SError */
+	writel(readl(port->mmio + PORT_SCR_ERR), port->mmio + PORT_SCR_ERR);
+
+	/* reset the completed registers.*/
+	for (i = 0; i < port->dd->slot_groups; i++)
+		writel(0xFFFFFFFF, port->completed[i]);
+
+	/* Clear any pending interrupts for this port */
+	writel(readl(port->mmio + PORT_IRQ_STAT), port->mmio + PORT_IRQ_STAT);
+
+	/* Enable port interrupts */
+	writel(DEF_PORT_IRQ, port->mmio + PORT_IRQ_MASK);
+}
+
+/*
+ * Restart a port
+ *
+ * @port Pointer to the port data structure.
+ *
+ * return value
+ *	None
+ */
+static void mtip_restart_port(struct mtip_port *port)
+{
+	unsigned long timeout;
+
+	/* Disable the DMA engine */
+	mtip_enable_engine(port, 0);
+
+	/* Chip quirk: wait up to 500ms for PxCMD.CR == 0 */
+	timeout = jiffies + msecs_to_jiffies(500);
+	while ((readl(port->mmio + PORT_CMD) & PORT_CMD_LIST_ON)
+		 && time_before(jiffies, timeout))
+		;
+
+	/*
+	 * Chip quirk: escalate to hba reset if
+	 * PxCMD.CR not clear after 500 ms
+	 */
+	if (readl(port->mmio + PORT_CMD) & PORT_CMD_LIST_ON) {
+		dev_warn(&port->dd->pdev->dev,
+			"PxCMD.CR not clear, escalating reset\n");
+
+		if (hba_reset_nosleep(port->dd))
+			dev_err(&port->dd->pdev->dev,
+				"HBA reset escalation failed.\n");
+
+		/* 30 ms delay before com reset to quiesce chip */
+		mdelay(30);
+	}
+
+	dev_warn(&port->dd->pdev->dev, "Issuing COM reset\n");
+
+	/* Set PxSCTL.DET */
+	writel(readl(port->mmio + PORT_SCR_CTL) |
+			 1, port->mmio + PORT_SCR_CTL);
+	readl(port->mmio + PORT_SCR_CTL);
+
+	/* Wait 1 ms to quiesce chip function */
+	timeout = jiffies + msecs_to_jiffies(1);
+	while (time_before(jiffies, timeout))
+		;
+
+	/* Clear PxSCTL.DET */
+	writel(readl(port->mmio + PORT_SCR_CTL) & ~1,
+			 port->mmio + PORT_SCR_CTL);
+	readl(port->mmio + PORT_SCR_CTL);
+
+	/* Wait 500 ms for bit 0 of PORT_SCR_STS to be set */
+	timeout = jiffies + msecs_to_jiffies(500);
+	while (((readl(port->mmio + PORT_SCR_STAT) & 0x01) == 0)
+			 && time_before(jiffies, timeout))
+		;
+
+	if ((readl(port->mmio + PORT_SCR_STAT) & 0x01) == 0)
+		dev_warn(&port->dd->pdev->dev,
+			"COM reset failed\n");
+
+	/* Clear SError, the PxSERR.DIAG.x should be set so clear it */
+	writel(readl(port->mmio + PORT_SCR_ERR), port->mmio + PORT_SCR_ERR);
+
+	/* Enable the DMA engine */
+	mtip_enable_engine(port, 1);
+}
+
+/*
+ * Called periodically to see if any read/write commands are
+ * taking too long to complete.
+ *
+ * @data Pointer to the PORT data structure.
+ *
+ * return value
+ *	None
+ */
+static void mtip_timeout_function(unsigned long int data)
+{
+	struct mtip_port *port = (struct mtip_port *) data;
+	struct host_to_dev_fis *fis;
+	struct mtip_cmd *command;
+	int tag, cmdto_cnt = 0;
+	unsigned int bit, group;
+	unsigned int num_command_slots = port->dd->slot_groups * 32;
+
+	if (unlikely(!port))
+		return;
+
+	if (atomic_read(&port->dd->resumeflag) == true) {
+		mod_timer(&port->cmd_timer,
+			jiffies + msecs_to_jiffies(30000));
+		return;
+	}
+
+	for (tag = 0; tag < num_command_slots; tag++) {
+		/*
+		 * Skip internal command slot as it has
+		 * its own timeout mechanism
+		 */
+		if (tag == MTIP_TAG_INTERNAL)
+			continue;
+
+		if (atomic_read(&port->commands[tag].active) &&
+		   (time_after(jiffies, port->commands[tag].comp_time))) {
+			group = tag >> 5;
+			bit = tag & 0x1F;
+
+			command = &port->commands[tag];
+			fis = (struct host_to_dev_fis *) command->command;
+
+			dev_warn(&port->dd->pdev->dev,
+				"Timeout for command tag %d\n", tag);
+
+			cmdto_cnt++;
+			if (cmdto_cnt == 1)
+				set_bit(MTIP_FLAG_EH_ACTIVE_BIT, &port->flags);
+
+			/*
+			 * Clear the completed bit. This should prevent
+			 *  any interrupt handlers from trying to retire
+			 *  the command.
+			 */
+			writel(1 << bit, port->completed[group]);
+
+			/* Call the async completion callback. */
+			if (likely(command->async_callback))
+				command->async_callback(command->async_data,
+							 -EIO);
+			command->async_callback = NULL;
+			command->comp_func = NULL;
+
+			/* Unmap the DMA scatter list entries */
+			dma_unmap_sg(&port->dd->pdev->dev,
+					command->sg,
+					command->scatter_ents,
+					command->direction);
+
+			/*
+			 * Clear the allocated bit and active tag for the
+			 * command.
+			 */
+			atomic_set(&port->commands[tag].active, 0);
+			release_slot(port, tag);
+
+			up(&port->cmd_slot);
+		}
+	}
+
+	if (cmdto_cnt) {
+		dev_warn(&port->dd->pdev->dev,
+			"%d commands timed out: restarting port",
+			cmdto_cnt);
+		mtip_restart_port(port);
+		clear_bit(MTIP_FLAG_EH_ACTIVE_BIT, &port->flags);
+		wake_up_interruptible(&port->svc_wait);
+	}
+
+	/* Restart the timer */
+	mod_timer(&port->cmd_timer,
+		jiffies + msecs_to_jiffies(MTIP_TIMEOUT_CHECK_PERIOD));
+}
+
+/*
+ * IO completion function.
+ *
+ * This completion function is called by the driver ISR when a
+ * command that was issued by the kernel completes. It first calls the
+ * asynchronous completion function which normally calls back into the block
+ * layer passing the asynchronous callback data, then unmaps the
+ * scatter list associated with the completed command, and finally
+ * clears the allocated bit associated with the completed command.
+ *
+ * @port   Pointer to the port data structure.
+ * @tag    Tag of the command.
+ * @data   Pointer to driver_data.
+ * @status Completion status.
+ *
+ * return value
+ *	None
+ */
+static void mtip_async_complete(struct mtip_port *port,
+				int tag,
+				void *data,
+				int status)
+{
+	struct mtip_cmd *command;
+	struct driver_data *dd = data;
+	int cb_status = status ? -EIO : 0;
+
+	if (unlikely(!dd) || unlikely(!port))
+		return;
+
+	command = &port->commands[tag];
+
+	if (unlikely(status == PORT_IRQ_TF_ERR)) {
+		dev_warn(&port->dd->pdev->dev,
+			"Command tag %d failed due to TFE\n", tag);
+	}
+
+	/* Upper layer callback */
+	if (likely(command->async_callback))
+		command->async_callback(command->async_data, cb_status);
+
+	command->async_callback = NULL;
+	command->comp_func = NULL;
+
+	/* Unmap the DMA scatter list entries */
+	dma_unmap_sg(&dd->pdev->dev,
+		command->sg,
+		command->scatter_ents,
+		command->direction);
+
+	/* Clear the allocated and active bits for the command */
+	atomic_set(&port->commands[tag].active, 0);
+	release_slot(port, tag);
+
+	up(&port->cmd_slot);
+}
+
+/*
+ * Internal command completion callback function.
+ *
+ * This function is normally called by the driver ISR when an internal
+ * command completed. This function signals the command completion by
+ * calling complete().
+ *
+ * @port   Pointer to the port data structure.
+ * @tag    Tag of the command that has completed.
+ * @data   Pointer to a completion structure.
+ * @status Completion status.
+ *
+ * return value
+ *	None
+ */
+static void mtip_completion(struct mtip_port *port,
+			    int tag,
+			    void *data,
+			    int status)
+{
+	struct mtip_cmd *command = &port->commands[tag];
+	struct completion *waiting = data;
+	if (unlikely(status == PORT_IRQ_TF_ERR))
+		dev_warn(&port->dd->pdev->dev,
+			"Internal command %d completed with TFE\n", tag);
+
+	command->async_callback = NULL;
+	command->comp_func = NULL;
+
+	complete(waiting);
+}
+
+/*
+ * Helper function for tag logging
+ */
+static void print_tags(struct driver_data *dd,
+			char *msg,
+			unsigned long *tagbits)
+{
+	unsigned int tag, count = 0;
+
+	for (tag = 0; tag < (dd->slot_groups) * 32; tag++) {
+		if (test_bit(tag, tagbits))
+			count++;
+	}
+	if (count)
+		dev_info(&dd->pdev->dev, "%s [%i tags]\n", msg, count);
+}
+
+/*
+ * Handle an error.
+ *
+ * @dd Pointer to the DRIVER_DATA structure.
+ *
+ * return value
+ *	None
+ */
+static void mtip_handle_tfe(struct driver_data *dd)
+{
+	int group, tag, bit, reissue;
+	struct mtip_port *port;
+	struct mtip_cmd  *command;
+	u32 completed;
+	struct host_to_dev_fis *fis;
+	unsigned long tagaccum[SLOTBITS_IN_LONGS];
+
+	dev_warn(&dd->pdev->dev, "Taskfile error\n");
+
+	port = dd->port;
+
+	/* Stop the timer to prevent command timeouts. */
+	del_timer(&port->cmd_timer);
+
+	/* Set eh_active */
+	set_bit(MTIP_FLAG_EH_ACTIVE_BIT, &port->flags);
+
+	/* Loop through all the groups */
+	for (group = 0; group < dd->slot_groups; group++) {
+		completed = readl(port->completed[group]);
+
+		/* clear completed status register in the hardware.*/
+		writel(completed, port->completed[group]);
+
+		/* clear the tag accumulator */
+		memset(tagaccum, 0, SLOTBITS_IN_LONGS * sizeof(long));
+
+		/* Process successfully completed commands */
+		for (bit = 0; bit < 32 && completed; bit++) {
+			if (!(completed & (1<<bit)))
+				continue;
+			tag = (group << 5) + bit;
+
+			/* Skip the internal command slot */
+			if (tag == MTIP_TAG_INTERNAL)
+				continue;
+
+			command = &port->commands[tag];
+			if (likely(command->comp_func)) {
+				set_bit(tag, tagaccum);
+				atomic_set(&port->commands[tag].active, 0);
+				command->comp_func(port,
+					 tag,
+					 command->comp_data,
+					 0);
+			} else {
+				dev_err(&port->dd->pdev->dev,
+					"Missing completion func for tag %d",
+					tag);
+				if (mtip_check_surprise_removal(dd->pdev)) {
+					mtip_command_cleanup(dd);
+					/* don't proceed further */
+					return;
+				}
+			}
+		}
+	}
+	print_tags(dd, "TFE tags completed:", tagaccum);
+
+	/* Restart the port */
+	mdelay(20);
+	mtip_restart_port(port);
+
+	/* clear the tag accumulator */
+	memset(tagaccum, 0, SLOTBITS_IN_LONGS * sizeof(long));
+
+	/* Loop through all the groups */
+	for (group = 0; group < dd->slot_groups; group++) {
+		for (bit = 0; bit < 32; bit++) {
+			reissue = 1;
+			tag = (group << 5) + bit;
+
+			/* If the active bit is set re-issue the command */
+			if (atomic_read(&port->commands[tag].active) == 0)
+				continue;
+
+			fis = (struct host_to_dev_fis *)
+				port->commands[tag].command;
+
+			/* Should re-issue? */
+			if (tag == MTIP_TAG_INTERNAL ||
+			    fis->command == ATA_CMD_SET_FEATURES)
+				reissue = 0;
+
+			/*
+			 * First check if this command has
+			 *  exceeded its retries.
+			 */
+			if (reissue &&
+			    (port->commands[tag].retries-- > 0)) {
+
+				set_bit(tag, tagaccum);
+
+				/* Update the timeout value. */
+				port->commands[tag].comp_time =
+					jiffies + msecs_to_jiffies(
+					MTIP_NCQ_COMMAND_TIMEOUT_MS);
+				/* Re-issue the command. */
+				mtip_issue_ncq_command(port, tag);
+
+				continue;
+			}
+
+			/* Retire a command that will not be reissued */
+			dev_warn(&port->dd->pdev->dev,
+				"retiring tag %d\n", tag);
+			atomic_set(&port->commands[tag].active, 0);
+
+			if (port->commands[tag].comp_func)
+				port->commands[tag].comp_func(
+					port,
+					tag,
+					port->commands[tag].comp_data,
+					PORT_IRQ_TF_ERR);
+			else
+				dev_warn(&port->dd->pdev->dev,
+					"Bad completion for tag %d\n",
+					tag);
+		}
+	}
+	print_tags(dd, "TFE tags reissued:", tagaccum);
+
+	/* clear eh_active */
+	clear_bit(MTIP_FLAG_EH_ACTIVE_BIT, &port->flags);
+	wake_up_interruptible(&port->svc_wait);
+
+	mod_timer(&port->cmd_timer,
+		 jiffies + msecs_to_jiffies(MTIP_TIMEOUT_CHECK_PERIOD));
+}
+
+/*
+ * Handle a set device bits interrupt
+ */
+static inline void mtip_process_sdbf(struct driver_data *dd)
+{
+	struct mtip_port  *port = dd->port;
+	int group, tag, bit;
+	u32 completed;
+	struct mtip_cmd *command;
+
+	/* walk all bits in all slot groups */
+	for (group = 0; group < dd->slot_groups; group++) {
+		completed = readl(port->completed[group]);
+
+		/* clear completed status register in the hardware.*/
+		writel(completed, port->completed[group]);
+
+		/* Process completed commands. */
+		for (bit = 0;
+		     (bit < 32) && completed;
+		     bit++, completed >>= 1) {
+			if (completed & 0x01) {
+				tag = (group << 5) | bit;
+
+				/* skip internal command slot. */
+				if (unlikely(tag == MTIP_TAG_INTERNAL))
+					continue;
+
+				command = &port->commands[tag];
+				/* make internal callback */
+				if (likely(command->comp_func)) {
+					command->comp_func(
+						port,
+						tag,
+						command->comp_data,
+						0);
+				} else {
+					dev_warn(&dd->pdev->dev,
+						"Null completion "
+						"for tag %d",
+						tag);
+
+					if (mtip_check_surprise_removal(
+						dd->pdev)) {
+						mtip_command_cleanup(dd);
+						return;
+					}
+				}
+			}
+		}
+	}
+}
+
+/*
+ * Process legacy pio and d2h interrupts
+ */
+static inline void mtip_process_legacy(struct driver_data *dd, u32 port_stat)
+{
+	struct mtip_port *port = dd->port;
+	struct mtip_cmd *cmd = &port->commands[MTIP_TAG_INTERNAL];
+
+	if (test_bit(MTIP_FLAG_IC_ACTIVE_BIT, &port->flags) &&
+	    (cmd != NULL) && !(readl(port->cmd_issue[MTIP_TAG_INTERNAL])
+		& (1 << MTIP_TAG_INTERNAL))) {
+		if (cmd->comp_func) {
+			cmd->comp_func(port,
+				MTIP_TAG_INTERNAL,
+				cmd->comp_data,
+				0);
+			return;
+		}
+	}
+
+	dev_warn(&dd->pdev->dev, "IRQ status 0x%x ignored.\n", port_stat);
+
+	return;
+}
+
+/*
+ * Demux and handle errors
+ */
+static inline void mtip_process_errors(struct driver_data *dd, u32 port_stat)
+{
+	if (likely(port_stat & (PORT_IRQ_TF_ERR | PORT_IRQ_IF_ERR)))
+		mtip_handle_tfe(dd);
+
+	if (unlikely(port_stat & PORT_IRQ_CONNECT)) {
+		dev_warn(&dd->pdev->dev,
+			"Clearing PxSERR.DIAG.x\n");
+		writel((1 << 26), dd->port->mmio + PORT_SCR_ERR);
+	}
+
+	if (unlikely(port_stat & PORT_IRQ_PHYRDY)) {
+		dev_warn(&dd->pdev->dev,
+			"Clearing PxSERR.DIAG.n\n");
+		writel((1 << 16), dd->port->mmio + PORT_SCR_ERR);
+	}
+
+	if (unlikely(port_stat & ~PORT_IRQ_HANDLED)) {
+		dev_warn(&dd->pdev->dev,
+			"Port stat errors %x unhandled\n",
+			(port_stat & ~PORT_IRQ_HANDLED));
+	}
+}
+
+static inline irqreturn_t mtip_handle_irq(struct driver_data *data)
+{
+	struct driver_data *dd = (struct driver_data *) data;
+	struct mtip_port *port = dd->port;
+	u32 hba_stat, port_stat;
+	int rv = IRQ_NONE;
+
+	hba_stat = readl(dd->mmio + HOST_IRQ_STAT);
+	if (hba_stat) {
+		rv = IRQ_HANDLED;
+
+		/* Acknowledge the interrupt status on the port.*/
+		port_stat = readl(port->mmio + PORT_IRQ_STAT);
+		writel(port_stat, port->mmio + PORT_IRQ_STAT);
+
+		/* Demux port status */
+		if (likely(port_stat & PORT_IRQ_SDB_FIS))
+			mtip_process_sdbf(dd);
+
+		if (unlikely(port_stat & PORT_IRQ_ERR)) {
+			if (unlikely(mtip_check_surprise_removal(dd->pdev))) {
+				mtip_command_cleanup(dd);
+				/* don't proceed further */
+				return IRQ_HANDLED;
+			}
+
+			mtip_process_errors(dd, port_stat & PORT_IRQ_ERR);
+		}
+
+		if (unlikely(port_stat & PORT_IRQ_LEGACY))
+			mtip_process_legacy(dd, port_stat & PORT_IRQ_LEGACY);
+	}
+
+	/* acknowledge interrupt */
+	writel(hba_stat, dd->mmio + HOST_IRQ_STAT);
+
+	return rv;
+}
+
+/*
+ * Wrapper for mtip_handle_irq
+ * (ignores return code)
+ */
+static void mtip_tasklet(unsigned long data)
+{
+	mtip_handle_irq((struct driver_data *) data);
+}
+
+/*
+ * HBA interrupt subroutine.
+ *
+ * @irq		IRQ number.
+ * @instance	Pointer to the driver data structure.
+ *
+ * return value
+ *	IRQ_HANDLED	A HBA interrupt was pending and handled.
+ *	IRQ_NONE	This interrupt was not for the HBA.
+ */
+static irqreturn_t mtip_irq_handler(int irq, void *instance)
+{
+	struct driver_data *dd = instance;
+	tasklet_schedule(&dd->tasklet);
+	return IRQ_HANDLED;
+}
+
+static void mtip_issue_non_ncq_command(struct mtip_port *port, int tag)
+{
+	atomic_set(&port->commands[tag].active, 1);
+	writel(1 << MTIP_TAG_BIT(tag),
+		port->cmd_issue[MTIP_TAG_INDEX(tag)]);
+}
+
+/*
+ * Wait for port to quiesce
+ *
+ * @port    Pointer to port data structure
+ * @timeout Max duration to wait (ms)
+ *
+ * return value
+ *	0	Success
+ *	-EBUSY  Commands still active
+ */
+static int mtip_quiesce_io(struct mtip_port *port, unsigned long timeout)
+{
+	unsigned long to;
+	unsigned int n;
+	unsigned int active = 1;
+
+	to = jiffies + msecs_to_jiffies(timeout);
+	do {
+		if (test_bit(MTIP_FLAG_SVC_THD_ACTIVE_BIT, &port->flags) &&
+			test_bit(MTIP_FLAG_ISSUE_CMDS_BIT, &port->flags)) {
+			msleep(20);
+			continue; /* svc thd is actively issuing commands */
+		}
+		/*
+		 * Ignore s_active bit 0 of array element 0.
+		 * This bit will always be set
+		 */
+		active = readl(port->s_active[0]) & 0xFFFFFFFE;
+		for (n = 1; n < port->dd->slot_groups; n++)
+			active |= readl(port->s_active[n]);
+
+		if (!active)
+			break;
+
+		msleep(20);
+	} while (time_before(jiffies, to));
+
+	return active ? -EBUSY : 0;
+}
+
+/*
+ * Execute an internal command and wait for the completion.
+ *
+ * @port    Pointer to the port data structure.
+ * @fis     Pointer to the FIS that describes the command.
+ * @fis_len  Length in WORDS of the FIS.
+ * @buffer  DMA accessible for command data.
+ * @buf_len  Length, in bytes, of the data buffer.
+ * @opts    Command header options, excluding the FIS length
+ *             and the number of PRD entries.
+ * @timeout Time in ms to wait for the command to complete.
+ *
+ * return value
+ *	0	 Command completed successfully.
+ *	-EFAULT  The buffer address is not correctly aligned.
+ *	-EBUSY   Internal command or other IO in progress.
+ *	-EAGAIN  Time out waiting for command to complete.
+ */
+static int mtip_exec_internal_command(struct mtip_port *port,
+					void *fis,
+					int fis_len,
+					dma_addr_t buffer,
+					int buf_len,
+					u32 opts,
+					gfp_t atomic,
+					unsigned long timeout)
+{
+	struct mtip_cmd_sg *command_sg;
+	DECLARE_COMPLETION_ONSTACK(wait);
+	int rv = 0;
+	struct mtip_cmd *int_cmd = &port->commands[MTIP_TAG_INTERNAL];
+
+	/* Make sure the buffer is 8 byte aligned. This is asic specific. */
+	if (buffer & 0x00000007) {
+		dev_err(&port->dd->pdev->dev,
+			"SG buffer is not 8 byte aligned\n");
+		return -EFAULT;
+	}
+
+	/* Only one internal command should be running at a time */
+	if (test_and_set_bit(MTIP_TAG_INTERNAL, port->allocated)) {
+		dev_warn(&port->dd->pdev->dev,
+			"Internal command already active\n");
+		return -EBUSY;
+	}
+	set_bit(MTIP_FLAG_IC_ACTIVE_BIT, &port->flags);
+
+	if (atomic == GFP_KERNEL) {
+		/* wait for io to complete if non atomic */
+		if (mtip_quiesce_io(port, 5000) < 0) {
+			dev_warn(&port->dd->pdev->dev,
+				"Failed to quiesce IO\n");
+			release_slot(port, MTIP_TAG_INTERNAL);
+			clear_bit(MTIP_FLAG_IC_ACTIVE_BIT, &port->flags);
+			wake_up_interruptible(&port->svc_wait);
+			return -EBUSY;
+		}
+
+		/* Set the completion function and data for the command. */
+		int_cmd->comp_data = &wait;
+		int_cmd->comp_func = mtip_completion;
+
+	} else {
+		/* Clear completion - we're going to poll */
+		int_cmd->comp_data = NULL;
+		int_cmd->comp_func = NULL;
+	}
+
+	/* Copy the command to the command table */
+	memcpy(int_cmd->command, fis, fis_len*4);
+
+	/* Populate the SG list */
+	int_cmd->command_header->opts =
+		 __force_bit2int cpu_to_le32(opts | fis_len);
+	if (buf_len) {
+		command_sg = int_cmd->command + AHCI_CMD_TBL_HDR_SZ;
+
+		command_sg->info =
+			__force_bit2int cpu_to_le32((buf_len-1) & 0x3FFFFF);
+		command_sg->dba	=
+			__force_bit2int cpu_to_le32(buffer & 0xFFFFFFFF);
+		command_sg->dba_upper =
+			__force_bit2int cpu_to_le32((buffer >> 16) >> 16);
+
+		int_cmd->command_header->opts |=
+			__force_bit2int cpu_to_le32((1 << 16));
+	}
+
+	/* Populate the command header */
+	int_cmd->command_header->byte_count = 0;
+
+	/* Issue the command to the hardware */
+	mtip_issue_non_ncq_command(port, MTIP_TAG_INTERNAL);
+
+	/* Poll if atomic, wait_for_completion otherwise */
+	if (atomic == GFP_KERNEL) {
+		/* Wait for the command to complete or timeout. */
+		if (wait_for_completion_timeout(
+				&wait,
+				msecs_to_jiffies(timeout)) == 0) {
+			dev_err(&port->dd->pdev->dev,
+				"Internal command did not complete [%d] "
+				"within timeout of  %lu ms\n",
+				atomic, timeout);
+			rv = -EAGAIN;
+		}
+
+		if (readl(port->cmd_issue[MTIP_TAG_INTERNAL])
+			& (1 << MTIP_TAG_INTERNAL)) {
+			dev_warn(&port->dd->pdev->dev,
+				"Retiring internal command but CI is 1.\n");
+		}
+
+	} else {
+		/* Spin for <timeout> checking if command still outstanding */
+		timeout = jiffies + msecs_to_jiffies(timeout);
+
+		while ((readl(
+			port->cmd_issue[MTIP_TAG_INTERNAL])
+			& (1 << MTIP_TAG_INTERNAL))
+			&& time_before(jiffies, timeout))
+			;
+
+		if (readl(port->cmd_issue[MTIP_TAG_INTERNAL])
+			& (1 << MTIP_TAG_INTERNAL)) {
+			dev_err(&port->dd->pdev->dev,
+				"Internal command did not complete [%d]\n",
+				atomic);
+			rv = -EAGAIN;
+		}
+	}
+
+	/* Clear the allocated and active bits for the internal command. */
+	atomic_set(&int_cmd->active, 0);
+	release_slot(port, MTIP_TAG_INTERNAL);
+	clear_bit(MTIP_FLAG_IC_ACTIVE_BIT, &port->flags);
+	wake_up_interruptible(&port->svc_wait);
+
+	return rv;
+}
+
+/*
+ * Byte-swap ATA ID strings.
+ *
+ * ATA identify data contains strings in byte-swapped 16-bit words.
+ * They must be swapped (on all architectures) to be usable as C strings.
+ * This function swaps bytes in-place.
+ *
+ * @buf The buffer location of the string
+ * @len The number of bytes to swap
+ *
+ * return value
+ *	None
+ */
+static inline void ata_swap_string(u16 *buf, unsigned int len)
+{
+	int i;
+	for (i = 0; i < (len/2); i++)
+		be16_to_cpus(&buf[i]);
+}
+
+/*
+ * Request the device identity information.
+ *
+ * If a user space buffer is not specified, i.e. is NULL, the
+ * identify information is still read from the drive and placed
+ * into the identify data buffer (@e port->identify) in the
+ * port data structure.
+ * When the identify buffer contains valid identify information @e
+ * port->identify_valid is non-zero.
+ *
+ * @port	 Pointer to the port structure.
+ * @user_buffer  A user space buffer where the identify data should be
+ *                    copied.
+ *
+ * return value
+ *	0	Command completed successfully.
+ *	-EFAULT An error occurred while coping data to the user buffer.
+ *	-1	Command failed.
+ */
+static int mtip_get_identify(struct mtip_port *port, void __user *user_buffer)
+{
+	int rv = 0;
+	struct host_to_dev_fis fis;
+
+	/* Build the FIS. */
+	memset(&fis, 0, sizeof(struct host_to_dev_fis));
+	fis.type	= 0x27;
+	fis.opts	= 1 << 7;
+	fis.command	= ATA_CMD_ID_ATA;
+
+	/* Set the identify information as invalid. */
+	port->identify_valid = 0;
+
+	/* Clear the identify information. */
+	memset(port->identify, 0, sizeof(u16) * ATA_ID_WORDS);
+
+	/* Execute the command. */
+	if (mtip_exec_internal_command(port,
+				&fis,
+				5,
+				port->identify_dma,
+				sizeof(u16) * ATA_ID_WORDS,
+				0,
+				GFP_KERNEL,
+				MTIP_INTERNAL_COMMAND_TIMEOUT_MS)
+				< 0) {
+		rv = -1;
+		goto out;
+	}
+
+	/*
+	 * Perform any necessary byte-swapping.  Yes, the kernel does in fact
+	 * perform field-sensitive swapping on the string fields.
+	 * See the kernel use of ata_id_string() for proof of this.
+	 */
+#ifdef __LITTLE_ENDIAN
+	ata_swap_string(port->identify + 27, 40);  /* model string*/
+	ata_swap_string(port->identify + 23, 8);   /* firmware string*/
+	ata_swap_string(port->identify + 10, 20);  /* serial# string*/
+#else
+	{
+		int i;
+		for (i = 0; i < ATA_ID_WORDS; i++)
+			port->identify[i] = le16_to_cpu(port->identify[i]);
+	}
+#endif
+
+	/* Set the identify buffer as valid. */
+	port->identify_valid = 1;
+
+	if (user_buffer) {
+		if (copy_to_user(
+			user_buffer,
+			port->identify,
+			ATA_ID_WORDS * sizeof(u16))) {
+			rv = -EFAULT;
+			goto out;
+		}
+	}
+
+out:
+	return rv;
+}
+
+/*
+ * Issue a standby immediate command to the device.
+ *
+ * @port Pointer to the port structure.
+ *
+ * return value
+ *	0	Command was executed successfully.
+ *	-1	An error occurred while executing the command.
+ */
+static int mtip_standby_immediate(struct mtip_port *port)
+{
+	int rv;
+	struct host_to_dev_fis	fis;
+
+	/* Build the FIS. */
+	memset(&fis, 0, sizeof(struct host_to_dev_fis));
+	fis.type	= 0x27;
+	fis.opts	= 1 << 7;
+	fis.command	= ATA_CMD_STANDBYNOW1;
+
+	/* Execute the command.  Use a 15-second timeout for large drives. */
+	rv = mtip_exec_internal_command(port,
+					&fis,
+					5,
+					0,
+					0,
+					0,
+					GFP_KERNEL,
+					15000);
+
+	return rv;
+}
+
+/*
+ * Get the drive capacity.
+ *
+ * @dd      Pointer to the device data structure.
+ * @sectors Pointer to the variable that will receive the sector count.
+ *
+ * return value
+ *	1 Capacity was returned successfully.
+ *	0 The identify information is invalid.
+ */
+static bool mtip_hw_get_capacity(struct driver_data *dd, sector_t *sectors)
+{
+	struct mtip_port *port = dd->port;
+	u64 total, raw0, raw1, raw2, raw3;
+	raw0 = port->identify[100];
+	raw1 = port->identify[101];
+	raw2 = port->identify[102];
+	raw3 = port->identify[103];
+	total = raw0 | raw1<<16 | raw2<<32 | raw3<<48;
+	*sectors = total;
+	return (bool) !!port->identify_valid;
+}
+
+/*
+ * Reset the HBA.
+ *
+ * Resets the HBA by setting the HBA Reset bit in the Global
+ * HBA Control register. After setting the HBA Reset bit the
+ * function waits for 1 second before reading the HBA Reset
+ * bit to make sure it has cleared. If HBA Reset is not clear
+ * an error is returned. Cannot be used in non-blockable
+ * context.
+ *
+ * @dd Pointer to the driver data structure.
+ *
+ * return value
+ *	0  The reset was successful.
+ *	-1 The HBA Reset bit did not clear.
+ */
+static int mtip_hba_reset(struct driver_data *dd)
+{
+	mtip_deinit_port(dd->port);
+
+	/* Set the reset bit */
+	writel(HOST_RESET, dd->mmio + HOST_CTL);
+
+	/* Flush */
+	readl(dd->mmio + HOST_CTL);
+
+	/* Wait for reset to clear */
+	ssleep(1);
+
+	/* Check the bit has cleared */
+	if (readl(dd->mmio + HOST_CTL) & HOST_RESET) {
+		dev_err(&dd->pdev->dev,
+			"Reset bit did not clear.\n");
+		return -1;
+	}
+
+	return 0;
+}
+
+/*
+ * Display the identify command data.
+ *
+ * @port Pointer to the port data structure.
+ *
+ * return value
+ *	None
+ */
+static void mtip_dump_identify(struct mtip_port *port)
+{
+	sector_t sectors;
+	unsigned short revid;
+	char cbuf[42];
+
+	if (!port->identify_valid)
+		return;
+
+	strlcpy(cbuf, (char *)(port->identify+10), 21);
+	dev_info(&port->dd->pdev->dev,
+		"Serial No.: %s\n", cbuf);
+
+	strlcpy(cbuf, (char *)(port->identify+23), 9);
+	dev_info(&port->dd->pdev->dev,
+		"Firmware Ver.: %s\n", cbuf);
+
+	strlcpy(cbuf, (char *)(port->identify+27), 41);
+	dev_info(&port->dd->pdev->dev, "Model: %s\n", cbuf);
+
+	if (mtip_hw_get_capacity(port->dd, &sectors))
+		dev_info(&port->dd->pdev->dev,
+			"Capacity: %llu sectors (%llu MB)\n",
+			 (u64)sectors,
+			 ((u64)sectors) * ATA_SECT_SIZE >> 20);
+
+	pci_read_config_word(port->dd->pdev, PCI_REVISION_ID, &revid);
+	switch (revid & 0xFF) {
+	case 0x1:
+		strlcpy(cbuf, "A0", 3);
+		break;
+	case 0x3:
+		strlcpy(cbuf, "A2", 3);
+		break;
+	default:
+		strlcpy(cbuf, "?", 2);
+		break;
+	}
+	dev_info(&port->dd->pdev->dev,
+		"Card Type: %s\n", cbuf);
+}
+
+/*
+ * Map the commands scatter list into the command table.
+ *
+ * @command Pointer to the command.
+ * @nents Number of scatter list entries.
+ *
+ * return value
+ *	None
+ */
+static inline void fill_command_sg(struct driver_data *dd,
+				struct mtip_cmd *command,
+				int nents)
+{
+	int n;
+	unsigned int dma_len;
+	struct mtip_cmd_sg *command_sg;
+	struct scatterlist *sg = command->sg;
+
+	command_sg = command->command + AHCI_CMD_TBL_HDR_SZ;
+
+	for (n = 0; n < nents; n++) {
+		dma_len = sg_dma_len(sg);
+		if (dma_len > 0x400000)
+			dev_err(&dd->pdev->dev,
+				"DMA segment length truncated\n");
+		command_sg->info = __force_bit2int
+			cpu_to_le32((dma_len-1) & 0x3FFFFF);
+		command_sg->dba	= __force_bit2int
+			cpu_to_le32(sg_dma_address(sg));
+		command_sg->dba_upper = __force_bit2int
+			cpu_to_le32((sg_dma_address(sg) >> 16) >> 16);
+		command_sg++;
+		sg++;
+	}
+}
+
+/*
+ * @brief Execute a drive command.
+ *
+ * return value 0 The command completed successfully.
+ * return value -1 An error occurred while executing the command.
+ */
+static int exec_drive_task(struct mtip_port *port, u8 *command)
+{
+	struct host_to_dev_fis	fis;
+	struct host_to_dev_fis *reply = (port->rxfis + RX_FIS_D2H_REG);
+
+	/* Build the FIS. */
+	memset(&fis, 0, sizeof(struct host_to_dev_fis));
+	fis.type	= 0x27;
+	fis.opts	= 1 << 7;
+	fis.command	= command[0];
+	fis.features	= command[1];
+	fis.sect_count	= command[2];
+	fis.sector	= command[3];
+	fis.cyl_low	= command[4];
+	fis.cyl_hi	= command[5];
+	fis.device	= command[6] & ~0x10; /* Clear the dev bit*/
+
+
+	dbg_printk(MTIP_DRV_NAME "%s: User Command: cmd %x, feat %x, "
+		"nsect %x, sect %x, lcyl %x, "
+		"hcyl %x, sel %x\n",
+		__func__,
+		command[0],
+		command[1],
+		command[2],
+		command[3],
+		command[4],
+		command[5],
+		command[6]);
+
+	/* Execute the command. */
+	if (mtip_exec_internal_command(port,
+				 &fis,
+				 5,
+				 0,
+				 0,
+				 0,
+				 GFP_KERNEL,
+				 MTIP_IOCTL_COMMAND_TIMEOUT_MS) < 0) {
+		return -1;
+	}
+
+	command[0] = reply->command; /* Status*/
+	command[1] = reply->features; /* Error*/
+	command[4] = reply->cyl_low;
+	command[5] = reply->cyl_hi;
+
+	dbg_printk(MTIP_DRV_NAME "%s: Completion Status: stat %x, "
+		"err %x , cyl_lo %x cyl_hi %x\n",
+		__func__,
+		command[0],
+		command[1],
+		command[4],
+		command[5]);
+
+	return 0;
+}
+
+/*
+ * @brief Execute a drive command.
+ *
+ * @param port Pointer to the port data structure.
+ * @param command Pointer to the user specified command parameters.
+ * @param user_buffer Pointer to the user space buffer where read sector
+ *                   data should be copied.
+ *
+ * return value 0 The command completed successfully.
+ * return value -EFAULT An error occurred while copying the completion
+ *                 data to the user space buffer.
+ * return value -1 An error occurred while executing the command.
+ */
+static int exec_drive_command(struct mtip_port *port, u8 *command,
+				void __user *user_buffer)
+{
+	struct host_to_dev_fis	fis;
+	struct host_to_dev_fis *reply = (port->rxfis + RX_FIS_D2H_REG);
+
+	/* Build the FIS. */
+	memset(&fis, 0, sizeof(struct host_to_dev_fis));
+	fis.type		= 0x27;
+	fis.opts		= 1 << 7;
+	fis.command		= command[0];
+	fis.features	= command[2];
+	fis.sect_count	= command[3];
+	if (fis.command == ATA_CMD_SMART) {
+		fis.sector	= command[1];
+		fis.cyl_low	= 0x4F;
+		fis.cyl_hi	= 0xC2;
+	}
+
+	dbg_printk(MTIP_DRV_NAME
+		"%s: User Command: cmd %x, sect %x, "
+		"feat %x, sectcnt %x\n",
+		__func__,
+		command[0],
+		command[1],
+		command[2],
+		command[3]);
+
+	memset(port->sector_buffer, 0x00, ATA_SECT_SIZE);
+
+	/* Execute the command. */
+	if (mtip_exec_internal_command(port,
+				&fis,
+				 5,
+				 port->sector_buffer_dma,
+				 (command[3] != 0) ? ATA_SECT_SIZE : 0,
+				 0,
+				 GFP_KERNEL,
+				 MTIP_IOCTL_COMMAND_TIMEOUT_MS)
+				 < 0) {
+		return -1;
+	}
+
+	/* Collect the completion status. */
+	command[0] = reply->command; /* Status*/
+	command[1] = reply->features; /* Error*/
+	command[2] = command[3];
+
+	dbg_printk(MTIP_DRV_NAME
+		"%s: Completion Status: stat %x, "
+		"err %x, cmd %x\n",
+		__func__,
+		command[0],
+		command[1],
+		command[2]);
+
+	if (user_buffer && command[3]) {
+		if (copy_to_user(user_buffer,
+				 port->sector_buffer,
+				 ATA_SECT_SIZE * command[3])) {
+			return -EFAULT;
+		}
+	}
+
+	return 0;
+}
+
+/*
+ *  Indicates whether a command has a single sector payload.
+ *
+ *  @command passed to the device to perform the certain event.
+ *  @features passed to the device to perform the certain event.
+ *
+ *  return value
+ *	1	command is one that always has a single sector payload,
+ *		regardless of the value in the Sector Count field.
+ *      0       otherwise
+ *
+ */
+static unsigned int implicit_sector(unsigned char command,
+				    unsigned char features)
+{
+	unsigned int rv = 0;
+
+	/* list of commands that have an implicit sector count of 1 */
+	switch (command) {
+	case ATA_CMD_SEC_SET_PASS:
+	case ATA_CMD_SEC_UNLOCK:
+	case ATA_CMD_SEC_ERASE_PREP:
+	case ATA_CMD_SEC_ERASE_UNIT:
+	case ATA_CMD_SEC_FREEZE_LOCK:
+	case ATA_CMD_SEC_DISABLE_PASS:
+	case ATA_CMD_PMP_READ:
+	case ATA_CMD_PMP_WRITE:
+		rv = 1;
+		break;
+	case ATA_CMD_SET_MAX:
+		if (features == ATA_SET_MAX_UNLOCK)
+			rv = 1;
+		break;
+	case ATA_CMD_SMART:
+		if ((features == ATA_SMART_READ_VALUES) ||
+				(features == ATA_SMART_READ_THRESHOLDS))
+			rv = 1;
+		break;
+	case ATA_CMD_CONF_OVERLAY:
+		if ((features == ATA_DCO_IDENTIFY) ||
+				(features == ATA_DCO_SET))
+			rv = 1;
+		break;
+	}
+	return rv;
+}
+
+/*
+ * Executes a taskfile
+ * See ide_taskfile_ioctl() for derivation
+ */
+static int exec_drive_taskfile(struct driver_data *dd,
+			       void __user *buf,
+			       ide_task_request_t *req_task,
+			       int outtotal)
+{
+	struct host_to_dev_fis	fis;
+	struct host_to_dev_fis *reply;
+	u8 *outbuf = NULL;
+	u8 *inbuf = NULL;
+	dma_addr_t outbuf_dma = 0;
+	dma_addr_t inbuf_dma = 0;
+	dma_addr_t dma_buffer = 0;
+	int err = 0;
+	unsigned int taskin = 0;
+	unsigned int taskout = 0;
+	u8 nsect = 0;
+	unsigned int timeout = MTIP_IOCTL_COMMAND_TIMEOUT_MS;
+	unsigned int force_single_sector;
+	unsigned int transfer_size;
+	unsigned long task_file_data;
+	int intotal = outtotal + req_task->out_size;
+
+	taskout = req_task->out_size;
+	taskin = req_task->in_size;
+	/* 130560 = 512 * 0xFF*/
+	if (taskin > 130560 || taskout > 130560) {
+		err = -EINVAL;
+		goto abort;
+	}
+
+	if (taskout) {
+		outbuf = kzalloc(taskout, GFP_KERNEL);
+		if (outbuf == NULL) {
+			err = -ENOMEM;
+			goto abort;
+		}
+		if (copy_from_user(outbuf, buf + outtotal, taskout)) {
+			err = -EFAULT;
+			goto abort;
+		}
+		outbuf_dma = pci_map_single(dd->pdev,
+					 outbuf,
+					 taskout,
+					 DMA_TO_DEVICE);
+		if (outbuf_dma == 0) {
+			err = -ENOMEM;
+			goto abort;
+		}
+		dma_buffer = outbuf_dma;
+	}
+
+	if (taskin) {
+		inbuf = kzalloc(taskin, GFP_KERNEL);
+		if (inbuf == NULL) {
+			err = -ENOMEM;
+			goto abort;
+		}
+
+		if (copy_from_user(inbuf, buf + intotal, taskin)) {
+			err = -EFAULT;
+			goto abort;
+		}
+		inbuf_dma = pci_map_single(dd->pdev,
+					 inbuf,
+					 taskin, DMA_FROM_DEVICE);
+		if (inbuf_dma == 0) {
+			err = -ENOMEM;
+			goto abort;
+		}
+		dma_buffer = inbuf_dma;
+	}
+
+	/* only supports PIO and non-data commands from this ioctl. */
+	switch (req_task->data_phase) {
+	case TASKFILE_OUT:
+		nsect = taskout / ATA_SECT_SIZE;
+		reply = (dd->port->rxfis + RX_FIS_PIO_SETUP);
+		break;
+	case TASKFILE_IN:
+		reply = (dd->port->rxfis + RX_FIS_PIO_SETUP);
+		break;
+	case TASKFILE_NO_DATA:
+		reply = (dd->port->rxfis + RX_FIS_D2H_REG);
+		break;
+	default:
+		err = -EINVAL;
+		goto abort;
+	}
+
+	/* Build the FIS. */
+	memset(&fis, 0, sizeof(struct host_to_dev_fis));
+
+	fis.type	= 0x27;
+	fis.opts	= 1 << 7;
+	fis.command	= req_task->io_ports[7];
+	fis.features	= req_task->io_ports[1];
+	fis.sect_count	= req_task->io_ports[2];
+	fis.lba_low	= req_task->io_ports[3];
+	fis.lba_mid	= req_task->io_ports[4];
+	fis.lba_hi	= req_task->io_ports[5];
+	 /* Clear the dev bit*/
+	fis.device	= req_task->io_ports[6] & ~0x10;
+
+	if ((req_task->in_flags.all == 0) && (req_task->out_flags.all & 1)) {
+		req_task->in_flags.all	=
+			IDE_TASKFILE_STD_IN_FLAGS |
+			(IDE_HOB_STD_IN_FLAGS << 8);
+		fis.lba_low_ex		= req_task->hob_ports[3];
+		fis.lba_mid_ex		= req_task->hob_ports[4];
+		fis.lba_hi_ex		= req_task->hob_ports[5];
+		fis.features_ex		= req_task->hob_ports[1];
+		fis.sect_cnt_ex		= req_task->hob_ports[2];
+
+	} else {
+		req_task->in_flags.all = IDE_TASKFILE_STD_IN_FLAGS;
+	}
+
+	force_single_sector = implicit_sector(fis.command, fis.features);
+
+	if ((taskin || taskout) && (!fis.sect_count)) {
+		if (nsect)
+			fis.sect_count = nsect;
+		else {
+			if (!force_single_sector) {
+				dev_warn(&dd->pdev->dev,
+					"data movement but "
+					"sect_count is 0\n");
+					err = -EINVAL;
+					goto abort;
+			}
+		}
+	}
+
+	dbg_printk(MTIP_DRV_NAME
+		"taskfile: cmd %x, feat %x, nsect %x,"
+		" sect/lbal %x, lcyl/lbam %x, hcyl/lbah %x,"
+		" head/dev %x\n",
+		fis.command,
+		fis.features,
+		fis.sect_count,
+		fis.lba_low,
+		fis.lba_mid,
+		fis.lba_hi,
+		fis.device);
+
+	switch (fis.command) {
+	case ATA_CMD_DOWNLOAD_MICRO:
+		/* Change timeout for Download Microcode to 60 seconds.*/
+		timeout = 60000;
+		break;
+	case ATA_CMD_SEC_ERASE_UNIT:
+		/* Change timeout for Security Erase Unit to 4 minutes.*/
+		timeout = 240000;
+		break;
+	case ATA_CMD_STANDBYNOW1:
+		/* Change timeout for standby immediate to 10 seconds.*/
+		timeout = 10000;
+		break;
+	case 0xF7:
+	case 0xFA:
+		/* Change timeout for vendor unique command to 10 secs */
+		timeout = 10000;
+		break;
+	case ATA_CMD_SMART:
+		/* Change timeout for vendor unique command to 10 secs */
+		timeout = 10000;
+		break;
+	default:
+		timeout = MTIP_IOCTL_COMMAND_TIMEOUT_MS;
+		break;
+	}
+
+	/* Determine the correct transfer size.*/
+	if (force_single_sector)
+		transfer_size = ATA_SECT_SIZE;
+	else
+		transfer_size = ATA_SECT_SIZE * fis.sect_count;
+
+	/* Execute the command.*/
+	if (mtip_exec_internal_command(dd->port,
+				 &fis,
+				 5,
+				 dma_buffer,
+				 transfer_size,
+				 0,
+				 GFP_KERNEL,
+				 timeout) < 0) {
+		err = -EIO;
+		goto abort;
+	}
+
+	task_file_data = readl(dd->port->mmio+PORT_TFDATA);
+
+	if ((req_task->data_phase == TASKFILE_IN) && !(task_file_data & 1)) {
+		reply = dd->port->rxfis + RX_FIS_PIO_SETUP;
+		req_task->io_ports[7] = reply->control;
+	} else {
+		reply = dd->port->rxfis + RX_FIS_D2H_REG;
+		req_task->io_ports[7] = reply->command;
+	}
+
+	/* reclaim the DMA buffers.*/
+	if (inbuf_dma)
+		pci_unmap_single(dd->pdev, inbuf_dma,
+			taskin, DMA_FROM_DEVICE);
+	if (outbuf_dma)
+		pci_unmap_single(dd->pdev, outbuf_dma,
+			taskout, DMA_TO_DEVICE);
+	inbuf_dma  = 0;
+	outbuf_dma = 0;
+
+	/* return the ATA registers to the caller.*/
+	req_task->io_ports[1] = reply->features;
+	req_task->io_ports[2] = reply->sect_count;
+	req_task->io_ports[3] = reply->lba_low;
+	req_task->io_ports[4] = reply->lba_mid;
+	req_task->io_ports[5] = reply->lba_hi;
+	req_task->io_ports[6] = reply->device;
+
+	if (req_task->out_flags.all & 1)  {
+
+		req_task->hob_ports[3] = reply->lba_low_ex;
+		req_task->hob_ports[4] = reply->lba_mid_ex;
+		req_task->hob_ports[5] = reply->lba_hi_ex;
+		req_task->hob_ports[1] = reply->features_ex;
+		req_task->hob_ports[2] = reply->sect_cnt_ex;
+	}
+
+	/* Com rest after secure erase or lowlevel format */
+	if (((fis.command == ATA_CMD_SEC_ERASE_UNIT) ||
+		((fis.command == 0xFC) &&
+			(fis.features == 0x27 || fis.features == 0x72 ||
+			 fis.features == 0x62 || fis.features == 0x26))) &&
+			 !(reply->command & 1)) {
+		mtip_restart_port(dd->port);
+	}
+
+	dbg_printk(MTIP_DRV_NAME
+		"%s: Completion: stat %x,"
+		"err %x, sect_cnt %x, lbalo %x,"
+		"lbamid %x, lbahi %x, dev %x\n",
+		__func__,
+		req_task->io_ports[7],
+		req_task->io_ports[1],
+		req_task->io_ports[2],
+		req_task->io_ports[3],
+		req_task->io_ports[4],
+		req_task->io_ports[5],
+		req_task->io_ports[6]);
+
+	if (taskout) {
+		if (copy_to_user(buf + outtotal, outbuf, taskout)) {
+			err = -EFAULT;
+			goto abort;
+		}
+	}
+	if (taskin) {
+		if (copy_to_user(buf + intotal, inbuf, taskin)) {
+			err = -EFAULT;
+			goto abort;
+		}
+	}
+abort:
+	if (inbuf_dma)
+		pci_unmap_single(dd->pdev, inbuf_dma,
+					taskin, DMA_FROM_DEVICE);
+	if (outbuf_dma)
+		pci_unmap_single(dd->pdev, outbuf_dma,
+					taskout, DMA_TO_DEVICE);
+	kfree(outbuf);
+	kfree(inbuf);
+
+	return err;
+}
+
+/*
+ * Handle IOCTL calls from the Block Layer.
+ *
+ * This function is called by the Block Layer when it receives an IOCTL
+ * command that it does not understand. If the IOCTL command is not supported
+ * this function returns -ENOTTY.
+ *
+ * @dd  Pointer to the driver data structure.
+ * @cmd IOCTL command passed from the Block Layer.
+ * @arg IOCTL argument passed from the Block Layer.
+ *
+ * return value
+ *	0	The IOCTL completed successfully.
+ *	-ENOTTY The specified command is not supported.
+ *	-EFAULT An error occurred copying data to a user space buffer.
+ *	-EIO	An error occurred while executing the command.
+ */
+static int mtip_hw_ioctl(struct driver_data *dd, unsigned int cmd,
+			 unsigned long arg)
+{
+	switch (cmd) {
+	case HDIO_GET_IDENTITY:
+		if (mtip_get_identify(dd->port, (void __user *) arg) < 0) {
+			dev_warn(&dd->pdev->dev,
+				"Unable to read identity\n");
+			return -EIO;
+		}
+
+		break;
+	case HDIO_DRIVE_CMD:
+	{
+		u8 drive_command[4];
+
+		/* Copy the user command info to our buffer. */
+		if (copy_from_user(drive_command,
+					 (void __user *) arg,
+					 sizeof(drive_command)))
+			return -EFAULT;
+
+		/* Execute the drive command. */
+		if (exec_drive_command(dd->port,
+					 drive_command,
+					 (void __user *) (arg+4)))
+			return -EIO;
+
+		/* Copy the status back to the users buffer. */
+		if (copy_to_user((void __user *) arg,
+					 drive_command,
+					 sizeof(drive_command)))
+			return -EFAULT;
+
+		break;
+	}
+	case HDIO_DRIVE_TASK:
+	{
+		u8 drive_command[7];
+
+		/* Copy the user command info to our buffer. */
+		if (copy_from_user(drive_command,
+					 (void __user *) arg,
+					 sizeof(drive_command)))
+			return -EFAULT;
+
+		/* Execute the drive command. */
+		if (exec_drive_task(dd->port, drive_command))
+			return -EIO;
+
+		/* Copy the status back to the users buffer. */
+		if (copy_to_user((void __user *) arg,
+					 drive_command,
+					 sizeof(drive_command)))
+			return -EFAULT;
+
+		break;
+	}
+	case HDIO_DRIVE_TASKFILE: {
+		ide_task_request_t req_task;
+		int ret, outtotal;
+
+		if (copy_from_user(&req_task, (void __user *) arg,
+					sizeof(req_task)))
+			return -EFAULT;
+
+		outtotal = sizeof(req_task);
+
+		ret = exec_drive_taskfile(dd, (void __user *) arg,
+						&req_task, outtotal);
+
+		if (copy_to_user((void __user *) arg, &req_task,
+							sizeof(req_task)))
+			return -EFAULT;
+
+		return ret;
+	}
+
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+/*
+ * Submit an IO to the hw
+ *
+ * This function is called by the block layer to issue an io
+ * to the device. Upon completion, the callback function will
+ * be called with the data parameter passed as the callback data.
+ *
+ * @dd       Pointer to the driver data structure.
+ * @start    First sector to read.
+ * @nsect    Number of sectors to read.
+ * @nents    Number of entries in scatter list for the read command.
+ * @tag      The tag of this read command.
+ * @callback Pointer to the function that should be called
+ *	     when the read completes.
+ * @data     Callback data passed to the callback function
+ *	     when the read completes.
+ * @barrier  If non-zero, this command must be completed before
+ *	     issuing any other commands.
+ * @dir      Direction (read or write)
+ *
+ * return value
+ *	None
+ */
+static void mtip_hw_submit_io(struct driver_data *dd, sector_t start,
+			      int nsect, int nents, int tag, void *callback,
+			      void *data, int barrier, int dir)
+{
+	struct host_to_dev_fis	*fis;
+	struct mtip_port *port = dd->port;
+	struct mtip_cmd *command = &port->commands[tag];
+
+	/* Map the scatter list for DMA access */
+	if (dir == READ)
+		nents = dma_map_sg(&dd->pdev->dev, command->sg,
+					nents, DMA_FROM_DEVICE);
+	else
+		nents = dma_map_sg(&dd->pdev->dev, command->sg,
+					nents, DMA_TO_DEVICE);
+
+	command->scatter_ents = nents;
+
+	/*
+	 * The number of retries for this command before it is
+	 * reported as a failure to the upper layers.
+	 */
+	command->retries = MTIP_MAX_RETRIES;
+
+	/* Fill out fis */
+	fis = command->command;
+	fis->type        = 0x27;
+	fis->opts        = 1 << 7;
+	fis->command     =
+		(dir == READ ? ATA_CMD_FPDMA_READ : ATA_CMD_FPDMA_WRITE);
+	*((unsigned int *) &fis->lba_low) = (start & 0xFFFFFF);
+	*((unsigned int *) &fis->lba_low_ex) = ((start >> 24) & 0xFFFFFF);
+	fis->device	 = 1 << 6;
+	if (barrier)
+		fis->device |= FUA_BIT;
+	fis->features    = nsect & 0xFF;
+	fis->features_ex = (nsect >> 8) & 0xFF;
+	fis->sect_count  = ((tag << 3) | (tag >> 5));
+	fis->sect_cnt_ex = 0;
+	fis->control     = 0;
+	fis->res2        = 0;
+	fis->res3        = 0;
+	fill_command_sg(dd, command, nents);
+
+	/* Populate the command header */
+	command->command_header->opts =
+			__force_bit2int cpu_to_le32(
+				(nents << 16) | 5 | AHCI_CMD_PREFETCH);
+	command->command_header->byte_count = 0;
+
+	/*
+	 * Set the completion function and data for the command
+	 * within this layer.
+	 */
+	command->comp_data = dd;
+	command->comp_func = mtip_async_complete;
+	command->direction = (dir == READ ? DMA_FROM_DEVICE : DMA_TO_DEVICE);
+
+	/*
+	 * Set the completion function and data for the command passed
+	 * from the upper layer.
+	 */
+	command->async_data = data;
+	command->async_callback = callback;
+
+	/*
+	 * To prevent this command from being issued
+	 * if an internal command is in progress or error handling is active.
+	 */
+	if (unlikely(test_bit(MTIP_FLAG_IC_ACTIVE_BIT, &port->flags) ||
+			test_bit(MTIP_FLAG_EH_ACTIVE_BIT, &port->flags))) {
+		set_bit(tag, port->cmds_to_issue);
+		set_bit(MTIP_FLAG_ISSUE_CMDS_BIT, &port->flags);
+		return;
+	}
+
+	/* Issue the command to the hardware */
+	mtip_issue_ncq_command(port, tag);
+
+	/* Set the command's timeout value.*/
+	port->commands[tag].comp_time = jiffies + msecs_to_jiffies(
+					MTIP_NCQ_COMMAND_TIMEOUT_MS);
+}
+
+/*
+ * Release a command slot.
+ *
+ * @dd  Pointer to the driver data structure.
+ * @tag Slot tag
+ *
+ * return value
+ *      None
+ */
+static void mtip_hw_release_scatterlist(struct driver_data *dd, int tag)
+{
+	release_slot(dd->port, tag);
+}
+
+/*
+ * Obtain a command slot and return its associated scatter list.
+ *
+ * @dd  Pointer to the driver data structure.
+ * @tag Pointer to an int that will receive the allocated command
+ *            slot tag.
+ *
+ * return value
+ *	Pointer to the scatter list for the allocated command slot
+ *	or NULL if no command slots are available.
+ */
+static struct scatterlist *mtip_hw_get_scatterlist(struct driver_data *dd,
+						   int *tag)
+{
+	/*
+	 * It is possible that, even with this semaphore, a thread
+	 * may think that no command slots are available. Therefore, we
+	 * need to make an attempt to get_slot().
+	 */
+	down(&dd->port->cmd_slot);
+	*tag = get_slot(dd->port);
+
+	if (unlikely(*tag < 0))
+		return NULL;
+
+	return dd->port->commands[*tag].sg;
+}
+
+/*
+ * Sysfs register/status dump.
+ *
+ * @dev  Pointer to the device structure, passed by the kernrel.
+ * @attr Pointer to the device_attribute structure passed by the kernel.
+ * @buf  Pointer to the char buffer that will receive the stats info.
+ *
+ * return value
+ *	The size, in bytes, of the data copied into buf.
+ */
+static ssize_t hw_show_registers(struct device *dev,
+				struct device_attribute *attr,
+				char *buf)
+{
+	u32 group_allocated;
+	struct driver_data *dd = dev_to_disk(dev)->private_data;
+	int size = 0;
+	int n;
+
+	size += sprintf(&buf[size], "%s:\ns_active:\n", __func__);
+
+	for (n = 0; n < dd->slot_groups; n++)
+		size += sprintf(&buf[size], "0x%08x\n",
+					 readl(dd->port->s_active[n]));
+
+	size += sprintf(&buf[size], "Command Issue:\n");
+
+	for (n = 0; n < dd->slot_groups; n++)
+		size += sprintf(&buf[size], "0x%08x\n",
+					readl(dd->port->cmd_issue[n]));
+
+	size += sprintf(&buf[size], "Allocated:\n");
+
+	for (n = 0; n < dd->slot_groups; n++) {
+		if (sizeof(long) > sizeof(u32))
+			group_allocated =
+				dd->port->allocated[n/2] >> (32*(n&1));
+		else
+			group_allocated = dd->port->allocated[n];
+		size += sprintf(&buf[size], "0x%08x\n",
+				 group_allocated);
+	}
+
+	size += sprintf(&buf[size], "completed:\n");
+
+	for (n = 0; n < dd->slot_groups; n++)
+		size += sprintf(&buf[size], "0x%08x\n",
+				readl(dd->port->completed[n]));
+
+	size += sprintf(&buf[size], "PORT_IRQ_STAT 0x%08x\n",
+				readl(dd->port->mmio + PORT_IRQ_STAT));
+	size += sprintf(&buf[size], "HOST_IRQ_STAT 0x%08x\n",
+				readl(dd->mmio + HOST_IRQ_STAT));
+
+	return size;
+}
+static DEVICE_ATTR(registers, S_IRUGO, hw_show_registers, NULL);
+
+/*
+ * Create the sysfs related attributes.
+ *
+ * @dd   Pointer to the driver data structure.
+ * @kobj Pointer to the kobj for the block device.
+ *
+ * return value
+ *	0	Operation completed successfully.
+ *	-EINVAL Invalid parameter.
+ */
+static int mtip_hw_sysfs_init(struct driver_data *dd, struct kobject *kobj)
+{
+	if (!kobj || !dd)
+		return -EINVAL;
+
+	if (sysfs_create_file(kobj, &dev_attr_registers.attr))
+		dev_warn(&dd->pdev->dev,
+			"Error creating registers sysfs entry\n");
+	return 0;
+}
+
+/*
+ * Remove the sysfs related attributes.
+ *
+ * @dd   Pointer to the driver data structure.
+ * @kobj Pointer to the kobj for the block device.
+ *
+ * return value
+ *	0	Operation completed successfully.
+ *	-EINVAL Invalid parameter.
+ */
+static int mtip_hw_sysfs_exit(struct driver_data *dd, struct kobject *kobj)
+{
+	if (!kobj || !dd)
+		return -EINVAL;
+
+	sysfs_remove_file(kobj, &dev_attr_registers.attr);
+
+	return 0;
+}
+
+/*
+ * Perform any init/resume time hardware setup
+ *
+ * @dd Pointer to the driver data structure.
+ *
+ * return value
+ *	None
+ */
+static inline void hba_setup(struct driver_data *dd)
+{
+	u32 hwdata;
+	hwdata = readl(dd->mmio + HOST_HSORG);
+
+	/* interrupt bug workaround: use only 1 IS bit.*/
+	writel(hwdata |
+		HSORG_DISABLE_SLOTGRP_INTR |
+		HSORG_DISABLE_SLOTGRP_PXIS,
+		dd->mmio + HOST_HSORG);
+}
+
+/*
+ * Detect the details of the product, and store anything needed
+ * into the driver data structure.  This includes product type and
+ * version and number of slot groups.
+ *
+ * @dd Pointer to the driver data structure.
+ *
+ * return value
+ *	None
+ */
+static void mtip_detect_product(struct driver_data *dd)
+{
+	u32 hwdata;
+	unsigned int rev, slotgroups;
+
+	/*
+	 * HBA base + 0xFC [15:0] - vendor-specific hardware interface
+	 * info register:
+	 * [15:8] hardware/software interface rev#
+	 * [   3] asic-style interface
+	 * [ 2:0] number of slot groups, minus 1 (only valid for asic-style).
+	 */
+	hwdata = readl(dd->mmio + HOST_HSORG);
+
+	dd->product_type = MTIP_PRODUCT_UNKNOWN;
+	dd->slot_groups = 1;
+
+	if (hwdata & 0x8) {
+		dd->product_type = MTIP_PRODUCT_ASICFPGA;
+		rev = (hwdata & HSORG_HWREV) >> 8;
+		slotgroups = (hwdata & HSORG_SLOTGROUPS) + 1;
+		dev_info(&dd->pdev->dev,
+			"ASIC-FPGA design, HS rev 0x%x, "
+			"%i slot groups [%i slots]\n",
+			 rev,
+			 slotgroups,
+			 slotgroups * 32);
+
+		if (slotgroups > MTIP_MAX_SLOT_GROUPS) {
+			dev_warn(&dd->pdev->dev,
+				"Warning: driver only supports "
+				"%i slot groups.\n", MTIP_MAX_SLOT_GROUPS);
+			slotgroups = MTIP_MAX_SLOT_GROUPS;
+		}
+		dd->slot_groups = slotgroups;
+		return;
+	}
+
+	dev_warn(&dd->pdev->dev, "Unrecognized product id\n");
+}
+
+/*
+ * Blocking wait for FTL rebuild to complete
+ *
+ * @dd Pointer to the DRIVER_DATA structure.
+ *
+ * return value
+ *	0	FTL rebuild completed successfully
+ *	-EFAULT FTL rebuild error/timeout/interruption
+ */
+static int mtip_ftl_rebuild_poll(struct driver_data *dd)
+{
+	unsigned long timeout, cnt = 0, start;
+
+	dev_warn(&dd->pdev->dev,
+		"FTL rebuild in progress. Polling for completion.\n");
+
+	start = jiffies;
+	dd->ftlrebuildflag = 1;
+	timeout = jiffies + msecs_to_jiffies(MTIP_FTL_REBUILD_TIMEOUT_MS);
+
+	do {
+		if (mtip_check_surprise_removal(dd->pdev))
+			return -EFAULT;
+
+		if (mtip_get_identify(dd->port, NULL) < 0)
+			return -EFAULT;
+
+		if (*(dd->port->identify + MTIP_FTL_REBUILD_OFFSET) ==
+			MTIP_FTL_REBUILD_MAGIC) {
+			ssleep(1);
+			/* Print message every 3 minutes */
+			if (cnt++ >= 180) {
+				dev_warn(&dd->pdev->dev,
+				"FTL rebuild in progress (%d secs).\n",
+				jiffies_to_msecs(jiffies - start) / 1000);
+				cnt = 0;
+			}
+		} else {
+			dev_warn(&dd->pdev->dev,
+				"FTL rebuild complete (%d secs).\n",
+			jiffies_to_msecs(jiffies - start) / 1000);
+			dd->ftlrebuildflag = 0;
+			mtip_block_initialize(dd);
+			break;
+		}
+		ssleep(10);
+	} while (time_before(jiffies, timeout));
+
+	/* Check for timeout */
+	if (dd->ftlrebuildflag) {
+		dev_err(&dd->pdev->dev,
+		"Timed out waiting for FTL rebuild to complete (%d secs).\n",
+		jiffies_to_msecs(jiffies - start) / 1000);
+		return -EFAULT;
+	}
+
+	return 0;
+}
+
+/*
+ * service thread to issue queued commands
+ *
+ * @data Pointer to the driver data structure.
+ *
+ * return value
+ *	0
+ */
+
+static int mtip_service_thread(void *data)
+{
+	struct driver_data *dd = (struct driver_data *)data;
+	unsigned long slot, slot_start, slot_wrap;
+	unsigned int num_cmd_slots = dd->slot_groups * 32;
+	struct mtip_port *port = dd->port;
+
+	while (1) {
+		/*
+		 * the condition is to check neither an internal command is
+		 * is in progress nor error handling is active
+		 */
+		wait_event_interruptible(port->svc_wait, (port->flags) &&
+			!test_bit(MTIP_FLAG_IC_ACTIVE_BIT, &port->flags) &&
+			!test_bit(MTIP_FLAG_EH_ACTIVE_BIT, &port->flags));
+
+		if (kthread_should_stop())
+			break;
+
+		set_bit(MTIP_FLAG_SVC_THD_ACTIVE_BIT, &port->flags);
+		if (test_bit(MTIP_FLAG_ISSUE_CMDS_BIT, &port->flags)) {
+			slot = 1;
+			/* used to restrict the loop to one iteration */
+			slot_start = num_cmd_slots;
+			slot_wrap = 0;
+			while (1) {
+				slot = find_next_bit(port->cmds_to_issue,
+						num_cmd_slots, slot);
+				if (slot_wrap == 1) {
+					if ((slot_start >= slot) ||
+						(slot >= num_cmd_slots))
+						break;
+				}
+				if (unlikely(slot_start == num_cmd_slots))
+					slot_start = slot;
+
+				if (unlikely(slot == num_cmd_slots)) {
+					slot = 1;
+					slot_wrap = 1;
+					continue;
+				}
+
+				/* Issue the command to the hardware */
+				mtip_issue_ncq_command(port, slot);
+
+				/* Set the command's timeout value.*/
+				port->commands[slot].comp_time = jiffies +
+				msecs_to_jiffies(MTIP_NCQ_COMMAND_TIMEOUT_MS);
+
+				clear_bit(slot, port->cmds_to_issue);
+			}
+
+			clear_bit(MTIP_FLAG_ISSUE_CMDS_BIT, &port->flags);
+		} else if (test_bit(MTIP_FLAG_REBUILD_BIT, &port->flags)) {
+			mtip_ftl_rebuild_poll(dd);
+			clear_bit(MTIP_FLAG_REBUILD_BIT, &port->flags);
+		}
+		clear_bit(MTIP_FLAG_SVC_THD_ACTIVE_BIT, &port->flags);
+
+		if (test_bit(MTIP_FLAG_SVC_THD_SHOULD_STOP_BIT, &port->flags))
+			break;
+	}
+	return 0;
+}
+
+/*
+ * Called once for each card.
+ *
+ * @dd Pointer to the driver data structure.
+ *
+ * return value
+ *	0 on success, else an error code.
+ */
+static int mtip_hw_init(struct driver_data *dd)
+{
+	int i;
+	int rv;
+	unsigned int num_command_slots;
+
+	dd->mmio = pcim_iomap_table(dd->pdev)[MTIP_ABAR];
+
+	mtip_detect_product(dd);
+	if (dd->product_type == MTIP_PRODUCT_UNKNOWN) {
+		rv = -EIO;
+		goto out1;
+	}
+	num_command_slots = dd->slot_groups * 32;
+
+	hba_setup(dd);
+
+	tasklet_init(&dd->tasklet, mtip_tasklet, (unsigned long)dd);
+
+	dd->port = kzalloc(sizeof(struct mtip_port), GFP_KERNEL);
+	if (!dd->port) {
+		dev_err(&dd->pdev->dev,
+			"Memory allocation: port structure\n");
+		return -ENOMEM;
+	}
+
+	/* Counting semaphore to track command slot usage */
+	sema_init(&dd->port->cmd_slot, num_command_slots - 1);
+
+	/* Spinlock to prevent concurrent issue */
+	spin_lock_init(&dd->port->cmd_issue_lock);
+
+	/* Set the port mmio base address. */
+	dd->port->mmio	= dd->mmio + PORT_OFFSET;
+	dd->port->dd	= dd;
+
+	/* Allocate memory for the command list. */
+	dd->port->command_list =
+		dmam_alloc_coherent(&dd->pdev->dev,
+			HW_PORT_PRIV_DMA_SZ + (ATA_SECT_SIZE * 2),
+			&dd->port->command_list_dma,
+			GFP_KERNEL);
+	if (!dd->port->command_list) {
+		dev_err(&dd->pdev->dev,
+			"Memory allocation: command list\n");
+		rv = -ENOMEM;
+		goto out1;
+	}
+
+	/* Clear the memory we have allocated. */
+	memset(dd->port->command_list,
+		0,
+		HW_PORT_PRIV_DMA_SZ + (ATA_SECT_SIZE * 2));
+
+	/* Setup the addresse of the RX FIS. */
+	dd->port->rxfis	    = dd->port->command_list + HW_CMD_SLOT_SZ;
+	dd->port->rxfis_dma = dd->port->command_list_dma + HW_CMD_SLOT_SZ;
+
+	/* Setup the address of the command tables. */
+	dd->port->command_table	  = dd->port->rxfis + AHCI_RX_FIS_SZ;
+	dd->port->command_tbl_dma = dd->port->rxfis_dma + AHCI_RX_FIS_SZ;
+
+	/* Setup the address of the identify data. */
+	dd->port->identify     = dd->port->command_table +
+					HW_CMD_TBL_AR_SZ;
+	dd->port->identify_dma = dd->port->command_tbl_dma +
+					HW_CMD_TBL_AR_SZ;
+
+	/* Setup the address of the sector buffer. */
+	dd->port->sector_buffer	= (void *) dd->port->identify + ATA_SECT_SIZE;
+	dd->port->sector_buffer_dma = dd->port->identify_dma + ATA_SECT_SIZE;
+
+	/* Point the command headers at the command tables. */
+	for (i = 0; i < num_command_slots; i++) {
+		dd->port->commands[i].command_header =
+					dd->port->command_list +
+					(sizeof(struct mtip_cmd_hdr) * i);
+		dd->port->commands[i].command_header_dma =
+					dd->port->command_list_dma +
+					(sizeof(struct mtip_cmd_hdr) * i);
+
+		dd->port->commands[i].command =
+			dd->port->command_table + (HW_CMD_TBL_SZ * i);
+		dd->port->commands[i].command_dma =
+			dd->port->command_tbl_dma + (HW_CMD_TBL_SZ * i);
+
+		if (readl(dd->mmio + HOST_CAP) & HOST_CAP_64)
+			dd->port->commands[i].command_header->ctbau =
+			__force_bit2int cpu_to_le32(
+			(dd->port->commands[i].command_dma >> 16) >> 16);
+		dd->port->commands[i].command_header->ctba =
+			__force_bit2int cpu_to_le32(
+			dd->port->commands[i].command_dma & 0xFFFFFFFF);
+
+		/*
+		 * If this is not done, a bug is reported by the stock
+		 * FC11 i386. Due to the fact that it has lots of kernel
+		 * debugging enabled.
+		 */
+		sg_init_table(dd->port->commands[i].sg, MTIP_MAX_SG);
+
+		/* Mark all commands as currently inactive.*/
+		atomic_set(&dd->port->commands[i].active, 0);
+	}
+
+	/* Setup the pointers to the extended s_active and CI registers. */
+	for (i = 0; i < dd->slot_groups; i++) {
+		dd->port->s_active[i] =
+			dd->port->mmio + i*0x80 + PORT_SCR_ACT;
+		dd->port->cmd_issue[i] =
+			dd->port->mmio + i*0x80 + PORT_COMMAND_ISSUE;
+		dd->port->completed[i] =
+			dd->port->mmio + i*0x80 + PORT_SDBV;
+	}
+
+	/* Reset the HBA. */
+	if (mtip_hba_reset(dd) < 0) {
+		dev_err(&dd->pdev->dev,
+			"Card did not reset within timeout\n");
+		rv = -EIO;
+		goto out2;
+	}
+
+	mtip_init_port(dd->port);
+	mtip_start_port(dd->port);
+
+	/* Setup the ISR and enable interrupts. */
+	rv = devm_request_irq(&dd->pdev->dev,
+				dd->pdev->irq,
+				mtip_irq_handler,
+				IRQF_SHARED,
+				dev_driver_string(&dd->pdev->dev),
+				dd);
+
+	if (rv) {
+		dev_err(&dd->pdev->dev,
+			"Unable to allocate IRQ %d\n", dd->pdev->irq);
+		goto out2;
+	}
+
+	/* Enable interrupts on the HBA. */
+	writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
+					dd->mmio + HOST_CTL);
+
+	init_timer(&dd->port->cmd_timer);
+	init_waitqueue_head(&dd->port->svc_wait);
+
+	dd->port->cmd_timer.data = (unsigned long int) dd->port;
+	dd->port->cmd_timer.function = mtip_timeout_function;
+	mod_timer(&dd->port->cmd_timer,
+		jiffies + msecs_to_jiffies(MTIP_TIMEOUT_CHECK_PERIOD));
+
+	if (mtip_get_identify(dd->port, NULL) < 0) {
+		rv = -EFAULT;
+		goto out3;
+	}
+
+	if (*(dd->port->identify + MTIP_FTL_REBUILD_OFFSET) ==
+		MTIP_FTL_REBUILD_MAGIC) {
+		set_bit(MTIP_FLAG_REBUILD_BIT, &dd->port->flags);
+		return MTIP_FTL_REBUILD_MAGIC;
+	}
+	mtip_dump_identify(dd->port);
+	return rv;
+
+out3:
+	del_timer_sync(&dd->port->cmd_timer);
+
+	/* Disable interrupts on the HBA. */
+	writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
+			dd->mmio + HOST_CTL);
+
+	/*Release the IRQ. */
+	devm_free_irq(&dd->pdev->dev, dd->pdev->irq, dd);
+
+out2:
+	mtip_deinit_port(dd->port);
+
+	/* Free the command/command header memory. */
+	dmam_free_coherent(&dd->pdev->dev,
+				HW_PORT_PRIV_DMA_SZ + (ATA_SECT_SIZE * 2),
+				dd->port->command_list,
+				dd->port->command_list_dma);
+out1:
+	/* Free the memory allocated for the for structure. */
+	kfree(dd->port);
+
+	return rv;
+}
+
+/*
+ * Called to deinitialize an interface.
+ *
+ * @dd Pointer to the driver data structure.
+ *
+ * return value
+ *	0
+ */
+static int mtip_hw_exit(struct driver_data *dd)
+{
+	/*
+	 * Send standby immediate (E0h) to the drive so that it
+	 * saves its state.
+	 */
+	if (atomic_read(&dd->drv_cleanup_done) != true) {
+
+		mtip_standby_immediate(dd->port);
+
+		/* de-initialize the port. */
+		mtip_deinit_port(dd->port);
+
+		/* Disable interrupts on the HBA. */
+		writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
+				dd->mmio + HOST_CTL);
+	}
+
+	del_timer_sync(&dd->port->cmd_timer);
+
+	/* Release the IRQ. */
+	devm_free_irq(&dd->pdev->dev, dd->pdev->irq, dd);
+
+	/* Stop the bottom half tasklet. */
+	tasklet_kill(&dd->tasklet);
+
+	/* Free the command/command header memory. */
+	dmam_free_coherent(&dd->pdev->dev,
+			HW_PORT_PRIV_DMA_SZ + (ATA_SECT_SIZE * 2),
+			dd->port->command_list,
+			dd->port->command_list_dma);
+	/* Free the memory allocated for the for structure. */
+	kfree(dd->port);
+
+	return 0;
+}
+
+/*
+ * Issue a Standby Immediate command to the device.
+ *
+ * This function is called by the Block Layer just before the
+ * system powers off during a shutdown.
+ *
+ * @dd Pointer to the driver data structure.
+ *
+ * return value
+ *	0
+ */
+static int mtip_hw_shutdown(struct driver_data *dd)
+{
+	/*
+	 * Send standby immediate (E0h) to the drive so that it
+	 * saves its state.
+	 */
+	mtip_standby_immediate(dd->port);
+
+	return 0;
+}
+
+/*
+ * Suspend function
+ *
+ * This function is called by the Block Layer just before the
+ * system hibernates.
+ *
+ * @dd Pointer to the driver data structure.
+ *
+ * return value
+ *	0	Suspend was successful
+ *	-EFAULT Suspend was not successful
+ */
+static int mtip_hw_suspend(struct driver_data *dd)
+{
+	/*
+	 * Send standby immediate (E0h) to the drive
+	 * so that it saves its state.
+	 */
+	if (mtip_standby_immediate(dd->port) != 0) {
+		dev_err(&dd->pdev->dev,
+			"Failed standby-immediate command\n");
+		return -EFAULT;
+	}
+
+	/* Disable interrupts on the HBA.*/
+	writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
+			dd->mmio + HOST_CTL);
+	mtip_deinit_port(dd->port);
+
+	return 0;
+}
+
+/*
+ * Resume function
+ *
+ * This function is called by the Block Layer as the
+ * system resumes.
+ *
+ * @dd Pointer to the driver data structure.
+ *
+ * return value
+ *	0	Resume was successful
+ *      -EFAULT Resume was not successful
+ */
+static int mtip_hw_resume(struct driver_data *dd)
+{
+	/* Perform any needed hardware setup steps */
+	hba_setup(dd);
+
+	/* Reset the HBA */
+	if (mtip_hba_reset(dd) != 0) {
+		dev_err(&dd->pdev->dev,
+			"Unable to reset the HBA\n");
+		return -EFAULT;
+	}
+
+	/*
+	 * Enable the port, DMA engine, and FIS reception specific
+	 * h/w in controller.
+	 */
+	mtip_init_port(dd->port);
+	mtip_start_port(dd->port);
+
+	/* Enable interrupts on the HBA.*/
+	writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
+			dd->mmio + HOST_CTL);
+
+	return 0;
+}
+
+/*
+ * Helper function for reusing disk name
+ * upon hot insertion.
+ */
+static int rssd_disk_name_format(char *prefix,
+				 int index,
+				 char *buf,
+				 int buflen)
+{
+	const int base = 'z' - 'a' + 1;
+	char *begin = buf + strlen(prefix);
+	char *end = buf + buflen;
+	char *p;
+	int unit;
+
+	p = end - 1;
+	*p = '\0';
+	unit = base;
+	do {
+		if (p == begin)
+			return -EINVAL;
+		*--p = 'a' + (index % unit);
+		index = (index / unit) - 1;
+	} while (index >= 0);
+
+	memmove(begin, p, end - p);
+	memcpy(buf, prefix, strlen(prefix));
+
+	return 0;
+}
+
+/*
+ * Block layer IOCTL handler.
+ *
+ * @dev Pointer to the block_device structure.
+ * @mode ignored
+ * @cmd IOCTL command passed from the user application.
+ * @arg Argument passed from the user application.
+ *
+ * return value
+ *	0        IOCTL completed successfully.
+ *	-ENOTTY  IOCTL not supported or invalid driver data
+ *                 structure pointer.
+ */
+static int mtip_block_ioctl(struct block_device *dev,
+			    fmode_t mode,
+			    unsigned cmd,
+			    unsigned long arg)
+{
+	struct driver_data *dd = dev->bd_disk->private_data;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EACCES;
+
+	if (!dd)
+		return -ENOTTY;
+
+	switch (cmd) {
+	case BLKFLSBUF:
+		return -ENOTTY;
+	default:
+		return mtip_hw_ioctl(dd, cmd, arg);
+	}
+}
+
+#ifdef CONFIG_COMPAT
+/*
+ * Block layer compat IOCTL handler.
+ *
+ * @dev Pointer to the block_device structure.
+ * @mode ignored
+ * @cmd IOCTL command passed from the user application.
+ * @arg Argument passed from the user application.
+ *
+ * return value
+ *	0        IOCTL completed successfully.
+ *	-ENOTTY  IOCTL not supported or invalid driver data
+ *                 structure pointer.
+ */
+static int mtip_block_compat_ioctl(struct block_device *dev,
+			    fmode_t mode,
+			    unsigned cmd,
+			    unsigned long arg)
+{
+	struct driver_data *dd = dev->bd_disk->private_data;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EACCES;
+
+	if (!dd)
+		return -ENOTTY;
+
+	switch (cmd) {
+	case BLKFLSBUF:
+		return -ENOTTY;
+	case HDIO_DRIVE_TASKFILE: {
+		struct mtip_compat_ide_task_request_s __user *compat_req_task;
+		ide_task_request_t req_task;
+		int compat_tasksize, outtotal, ret;
+
+		compat_tasksize =
+			sizeof(struct mtip_compat_ide_task_request_s);
+
+		compat_req_task =
+			(struct mtip_compat_ide_task_request_s __user *) arg;
+
+		if (copy_from_user(&req_task, (void __user *) arg,
+			compat_tasksize - (2 * sizeof(compat_long_t))))
+			return -EFAULT;
+
+		if (get_user(req_task.out_size, &compat_req_task->out_size))
+			return -EFAULT;
+
+		if (get_user(req_task.in_size, &compat_req_task->in_size))
+			return -EFAULT;
+
+		outtotal = sizeof(struct mtip_compat_ide_task_request_s);
+
+		ret = exec_drive_taskfile(dd, (void __user *) arg,
+						&req_task, outtotal);
+
+		if (copy_to_user((void __user *) arg, &req_task,
+				compat_tasksize -
+				(2 * sizeof(compat_long_t))))
+			return -EFAULT;
+
+		if (put_user(req_task.out_size, &compat_req_task->out_size))
+			return -EFAULT;
+
+		if (put_user(req_task.in_size, &compat_req_task->in_size))
+			return -EFAULT;
+
+		return ret;
+	}
+	default:
+		return mtip_hw_ioctl(dd, cmd, arg);
+	}
+}
+#endif
+
+/*
+ * Obtain the geometry of the device.
+ *
+ * You may think that this function is obsolete, but some applications,
+ * fdisk for example still used CHS values. This function describes the
+ * device as having 224 heads and 56 sectors per cylinder. These values are
+ * chosen so that each cylinder is aligned on a 4KB boundary. Since a
+ * partition is described in terms of a start and end cylinder this means
+ * that each partition is also 4KB aligned. Non-aligned partitions adversely
+ * affects performance.
+ *
+ * @dev Pointer to the block_device strucutre.
+ * @geo Pointer to a hd_geometry structure.
+ *
+ * return value
+ *	0       Operation completed successfully.
+ *	-ENOTTY An error occurred while reading the drive capacity.
+ */
+static int mtip_block_getgeo(struct block_device *dev,
+				struct hd_geometry *geo)
+{
+	struct driver_data *dd = dev->bd_disk->private_data;
+	sector_t capacity;
+
+	if (!dd)
+		return -ENOTTY;
+
+	if (!(mtip_hw_get_capacity(dd, &capacity))) {
+		dev_warn(&dd->pdev->dev,
+			"Could not get drive capacity.\n");
+		return -ENOTTY;
+	}
+
+	geo->heads = 224;
+	geo->sectors = 56;
+	sector_div(capacity, (geo->heads * geo->sectors));
+	geo->cylinders = capacity;
+	return 0;
+}
+
+/*
+ * Block device operation function.
+ *
+ * This structure contains pointers to the functions required by the block
+ * layer.
+ */
+static const struct block_device_operations mtip_block_ops = {
+	.ioctl		= mtip_block_ioctl,
+#ifdef CONFIG_COMPAT
+	.compat_ioctl	= mtip_block_compat_ioctl,
+#endif
+	.getgeo		= mtip_block_getgeo,
+	.owner		= THIS_MODULE
+};
+
+/*
+ * Block layer make request function.
+ *
+ * This function is called by the kernel to process a BIO for
+ * the P320 device.
+ *
+ * @queue Pointer to the request queue. Unused other than to obtain
+ *              the driver data structure.
+ * @bio   Pointer to the BIO.
+ *
+ */
+static void mtip_make_request(struct request_queue *queue, struct bio *bio)
+{
+	struct driver_data *dd = queue->queuedata;
+	struct scatterlist *sg;
+	struct bio_vec *bvec;
+	int nents = 0;
+	int tag = 0;
+
+	if (unlikely(!bio_has_data(bio))) {
+		blk_queue_flush(queue, 0);
+		bio_endio(bio, 0);
+		return;
+	}
+
+	sg = mtip_hw_get_scatterlist(dd, &tag);
+	if (likely(sg != NULL)) {
+		blk_queue_bounce(queue, &bio);
+
+		if (unlikely((bio)->bi_vcnt > MTIP_MAX_SG)) {
+			dev_warn(&dd->pdev->dev,
+				"Maximum number of SGL entries exceeded");
+			bio_io_error(bio);
+			mtip_hw_release_scatterlist(dd, tag);
+			return;
+		}
+
+		/* Create the scatter list for this bio. */
+		bio_for_each_segment(bvec, bio, nents) {
+			sg_set_page(&sg[nents],
+					bvec->bv_page,
+					bvec->bv_len,
+					bvec->bv_offset);
+		}
+
+		/* Issue the read/write. */
+		mtip_hw_submit_io(dd,
+				bio->bi_sector,
+				bio_sectors(bio),
+				nents,
+				tag,
+				bio_endio,
+				bio,
+				bio->bi_rw & REQ_FUA,
+				bio_data_dir(bio));
+	} else
+		bio_io_error(bio);
+}
+
+/*
+ * Block layer initialization function.
+ *
+ * This function is called once by the PCI layer for each P320
+ * device that is connected to the system.
+ *
+ * @dd Pointer to the driver data structure.
+ *
+ * return value
+ *	0 on success else an error code.
+ */
+static int mtip_block_initialize(struct driver_data *dd)
+{
+	int rv = 0, wait_for_rebuild = 0;
+	sector_t capacity;
+	unsigned int index = 0;
+	struct kobject *kobj;
+	unsigned char thd_name[16];
+
+	if (dd->disk)
+		goto skip_create_disk; /* hw init done, before rebuild */
+
+	/* Initialize the protocol layer. */
+	wait_for_rebuild = mtip_hw_init(dd);
+	if (wait_for_rebuild < 0) {
+		dev_err(&dd->pdev->dev,
+			"Protocol layer initialization failed\n");
+		rv = -EINVAL;
+		goto protocol_init_error;
+	}
+
+	dd->disk = alloc_disk(MTIP_MAX_MINORS);
+	if (dd->disk  == NULL) {
+		dev_err(&dd->pdev->dev,
+			"Unable to allocate gendisk structure\n");
+		rv = -EINVAL;
+		goto alloc_disk_error;
+	}
+
+	/* Generate the disk name, implemented same as in sd.c */
+	do {
+		if (!ida_pre_get(&rssd_index_ida, GFP_KERNEL))
+			goto ida_get_error;
+
+		spin_lock(&rssd_index_lock);
+		rv = ida_get_new(&rssd_index_ida, &index);
+		spin_unlock(&rssd_index_lock);
+	} while (rv == -EAGAIN);
+
+	if (rv)
+		goto ida_get_error;
+
+	rv = rssd_disk_name_format("rssd",
+				index,
+				dd->disk->disk_name,
+				DISK_NAME_LEN);
+	if (rv)
+		goto disk_index_error;
+
+	dd->disk->driverfs_dev	= &dd->pdev->dev;
+	dd->disk->major		= dd->major;
+	dd->disk->first_minor	= dd->instance * MTIP_MAX_MINORS;
+	dd->disk->fops		= &mtip_block_ops;
+	dd->disk->private_data	= dd;
+	dd->index		= index;
+
+	/*
+	 * if rebuild pending, start the service thread, and delay the block
+	 * queue creation and add_disk()
+	 */
+	if (wait_for_rebuild == MTIP_FTL_REBUILD_MAGIC)
+		goto start_service_thread;
+
+skip_create_disk:
+	/* Allocate the request queue. */
+	dd->queue = blk_alloc_queue(GFP_KERNEL);
+	if (dd->queue == NULL) {
+		dev_err(&dd->pdev->dev,
+			"Unable to allocate request queue\n");
+		rv = -ENOMEM;
+		goto block_queue_alloc_init_error;
+	}
+
+	/* Attach our request function to the request queue. */
+	blk_queue_make_request(dd->queue, mtip_make_request);
+
+	dd->disk->queue		= dd->queue;
+	dd->queue->queuedata	= dd;
+
+	/* Set device limits. */
+	set_bit(QUEUE_FLAG_NONROT, &dd->queue->queue_flags);
+	blk_queue_max_segments(dd->queue, MTIP_MAX_SG);
+	blk_queue_physical_block_size(dd->queue, 4096);
+	blk_queue_io_min(dd->queue, 4096);
+	blk_queue_flush(dd->queue, 0);
+
+	/* Set the capacity of the device in 512 byte sectors. */
+	if (!(mtip_hw_get_capacity(dd, &capacity))) {
+		dev_warn(&dd->pdev->dev,
+			"Could not read drive capacity\n");
+		rv = -EIO;
+		goto read_capacity_error;
+	}
+	set_capacity(dd->disk, capacity);
+
+	/* Enable the block device and add it to /dev */
+	add_disk(dd->disk);
+
+	/*
+	 * Now that the disk is active, initialize any sysfs attributes
+	 * managed by the protocol layer.
+	 */
+	kobj = kobject_get(&disk_to_dev(dd->disk)->kobj);
+	if (kobj) {
+		mtip_hw_sysfs_init(dd, kobj);
+		kobject_put(kobj);
+	}
+
+	if (dd->mtip_svc_handler)
+		return rv; /* service thread created for handling rebuild */
+
+start_service_thread:
+	sprintf(thd_name, "mtip_svc_thd_%02d", index);
+
+	dd->mtip_svc_handler = kthread_run(mtip_service_thread,
+						dd, thd_name);
+
+	if (IS_ERR(dd->mtip_svc_handler)) {
+		printk(KERN_ERR "mtip32xx: service thread failed to start\n");
+		dd->mtip_svc_handler = NULL;
+		rv = -EFAULT;
+		goto kthread_run_error;
+	}
+
+	return rv;
+
+kthread_run_error:
+	/* Delete our gendisk. This also removes the device from /dev */
+	del_gendisk(dd->disk);
+
+read_capacity_error:
+	blk_cleanup_queue(dd->queue);
+
+block_queue_alloc_init_error:
+disk_index_error:
+	spin_lock(&rssd_index_lock);
+	ida_remove(&rssd_index_ida, index);
+	spin_unlock(&rssd_index_lock);
+
+ida_get_error:
+	put_disk(dd->disk);
+
+alloc_disk_error:
+	mtip_hw_exit(dd); /* De-initialize the protocol layer. */
+
+protocol_init_error:
+	return rv;
+}
+
+/*
+ * Block layer deinitialization function.
+ *
+ * Called by the PCI layer as each P320 device is removed.
+ *
+ * @dd Pointer to the driver data structure.
+ *
+ * return value
+ *	0
+ */
+static int mtip_block_remove(struct driver_data *dd)
+{
+	struct kobject *kobj;
+
+	if (dd->mtip_svc_handler) {
+		set_bit(MTIP_FLAG_SVC_THD_SHOULD_STOP_BIT, &dd->port->flags);
+		wake_up_interruptible(&dd->port->svc_wait);
+		kthread_stop(dd->mtip_svc_handler);
+	}
+
+	/* Clean up the sysfs attributes managed by the protocol layer. */
+	kobj = kobject_get(&disk_to_dev(dd->disk)->kobj);
+	if (kobj) {
+		mtip_hw_sysfs_exit(dd, kobj);
+		kobject_put(kobj);
+	}
+
+	/*
+	 * Delete our gendisk structure. This also removes the device
+	 * from /dev
+	 */
+	del_gendisk(dd->disk);
+	blk_cleanup_queue(dd->queue);
+	dd->disk  = NULL;
+	dd->queue = NULL;
+
+	/* De-initialize the protocol layer. */
+	mtip_hw_exit(dd);
+
+	return 0;
+}
+
+/*
+ * Function called by the PCI layer when just before the
+ * machine shuts down.
+ *
+ * If a protocol layer shutdown function is present it will be called
+ * by this function.
+ *
+ * @dd Pointer to the driver data structure.
+ *
+ * return value
+ *	0
+ */
+static int mtip_block_shutdown(struct driver_data *dd)
+{
+	dev_info(&dd->pdev->dev,
+		"Shutting down %s ...\n", dd->disk->disk_name);
+
+	/* Delete our gendisk structure, and cleanup the blk queue. */
+	del_gendisk(dd->disk);
+	blk_cleanup_queue(dd->queue);
+	dd->disk  = NULL;
+	dd->queue = NULL;
+
+	mtip_hw_shutdown(dd);
+	return 0;
+}
+
+static int mtip_block_suspend(struct driver_data *dd)
+{
+	dev_info(&dd->pdev->dev,
+		"Suspending %s ...\n", dd->disk->disk_name);
+	mtip_hw_suspend(dd);
+	return 0;
+}
+
+static int mtip_block_resume(struct driver_data *dd)
+{
+	dev_info(&dd->pdev->dev, "Resuming %s ...\n",
+		dd->disk->disk_name);
+	mtip_hw_resume(dd);
+	return 0;
+}
+
+/*
+ * Called for each supported PCI device detected.
+ *
+ * This function allocates the private data structure, enables the
+ * PCI device and then calls the block layer initialization function.
+ *
+ * return value
+ *	0 on success else an error code.
+ */
+static int mtip_pci_probe(struct pci_dev *pdev,
+			const struct pci_device_id *ent)
+{
+	int rv = 0;
+	struct driver_data *dd = NULL;
+
+	/* Allocate memory for this devices private data. */
+	dd = kzalloc(sizeof(struct driver_data), GFP_KERNEL);
+	if (dd == NULL) {
+		dev_err(&pdev->dev,
+			"Unable to allocate memory for driver data\n");
+		return -ENOMEM;
+	}
+
+	/* Set the atomic variable as 1 in case of SRSI */
+	atomic_set(&dd->drv_cleanup_done, true);
+
+	atomic_set(&dd->resumeflag, false);
+
+	/* Attach the private data to this PCI device.  */
+	pci_set_drvdata(pdev, dd);
+
+	rv = pcim_enable_device(pdev);
+	if (rv < 0) {
+		dev_err(&pdev->dev, "Unable to enable device\n");
+		goto iomap_err;
+	}
+
+	/* Map BAR5 to memory. */
+	rv = pcim_iomap_regions(pdev, 1 << MTIP_ABAR, MTIP_DRV_NAME);
+	if (rv < 0) {
+		dev_err(&pdev->dev, "Unable to map regions\n");
+		goto iomap_err;
+	}
+
+	if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) {
+		rv = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
+
+		if (rv) {
+			rv = pci_set_consistent_dma_mask(pdev,
+						DMA_BIT_MASK(32));
+			if (rv) {
+				dev_warn(&pdev->dev,
+					"64-bit DMA enable failed\n");
+				goto setmask_err;
+			}
+		}
+	}
+
+	pci_set_master(pdev);
+
+	if (pci_enable_msi(pdev)) {
+		dev_warn(&pdev->dev,
+			"Unable to enable MSI interrupt.\n");
+		goto block_initialize_err;
+	}
+
+	/* Copy the info we may need later into the private data structure. */
+	dd->major	= mtip_major;
+	dd->instance	= instance;
+	dd->pdev	= pdev;
+
+	/* Initialize the block layer. */
+	rv = mtip_block_initialize(dd);
+	if (rv < 0) {
+		dev_err(&pdev->dev,
+			"Unable to initialize block layer\n");
+		goto block_initialize_err;
+	}
+
+	/*
+	 * Increment the instance count so that each device has a unique
+	 * instance number.
+	 */
+	instance++;
+
+	goto done;
+
+block_initialize_err:
+	pci_disable_msi(pdev);
+
+setmask_err:
+	pcim_iounmap_regions(pdev, 1 << MTIP_ABAR);
+
+iomap_err:
+	kfree(dd);
+	pci_set_drvdata(pdev, NULL);
+	return rv;
+done:
+	/* Set the atomic variable as 0 in case of SRSI */
+	atomic_set(&dd->drv_cleanup_done, true);
+
+	return rv;
+}
+
+/*
+ * Called for each probed device when the device is removed or the
+ * driver is unloaded.
+ *
+ * return value
+ *	None
+ */
+static void mtip_pci_remove(struct pci_dev *pdev)
+{
+	struct driver_data *dd = pci_get_drvdata(pdev);
+	int counter = 0;
+
+	if (mtip_check_surprise_removal(pdev)) {
+		while (atomic_read(&dd->drv_cleanup_done) == false) {
+			counter++;
+			msleep(20);
+			if (counter == 10) {
+				/* Cleanup the outstanding commands */
+				mtip_command_cleanup(dd);
+				break;
+			}
+		}
+	}
+	/* Set the atomic variable as 1 in case of SRSI */
+	atomic_set(&dd->drv_cleanup_done, true);
+
+	/* Clean up the block layer. */
+	mtip_block_remove(dd);
+
+	pci_disable_msi(pdev);
+
+	kfree(dd);
+	pcim_iounmap_regions(pdev, 1 << MTIP_ABAR);
+}
+
+/*
+ * Called for each probed device when the device is suspended.
+ *
+ * return value
+ *	0  Success
+ *	<0 Error
+ */
+static int mtip_pci_suspend(struct pci_dev *pdev, pm_message_t mesg)
+{
+	int rv = 0;
+	struct driver_data *dd = pci_get_drvdata(pdev);
+
+	if (!dd) {
+		dev_err(&pdev->dev,
+			"Driver private datastructure is NULL\n");
+		return -EFAULT;
+	}
+
+	atomic_set(&dd->resumeflag, true);
+
+	/* Disable ports & interrupts then send standby immediate */
+	rv = mtip_block_suspend(dd);
+	if (rv < 0) {
+		dev_err(&pdev->dev,
+			"Failed to suspend controller\n");
+		return rv;
+	}
+
+	/*
+	 * Save the pci config space to pdev structure &
+	 * disable the device
+	 */
+	pci_save_state(pdev);
+	pci_disable_device(pdev);
+
+	/* Move to Low power state*/
+	pci_set_power_state(pdev, PCI_D3hot);
+
+	return rv;
+}
+
+/*
+ * Called for each probed device when the device is resumed.
+ *
+ * return value
+ *      0  Success
+ *      <0 Error
+ */
+static int mtip_pci_resume(struct pci_dev *pdev)
+{
+	int rv = 0;
+	struct driver_data *dd;
+
+	dd = pci_get_drvdata(pdev);
+	if (!dd) {
+		dev_err(&pdev->dev,
+			"Driver private datastructure is NULL\n");
+		return -EFAULT;
+	}
+
+	/* Move the device to active State */
+	pci_set_power_state(pdev, PCI_D0);
+
+	/* Restore PCI configuration space */
+	pci_restore_state(pdev);
+
+	/* Enable the PCI device*/
+	rv = pcim_enable_device(pdev);
+	if (rv < 0) {
+		dev_err(&pdev->dev,
+			"Failed to enable card during resume\n");
+		goto err;
+	}
+	pci_set_master(pdev);
+
+	/*
+	 * Calls hbaReset, initPort, & startPort function
+	 * then enables interrupts
+	 */
+	rv = mtip_block_resume(dd);
+	if (rv < 0)
+		dev_err(&pdev->dev, "Unable to resume\n");
+
+err:
+	atomic_set(&dd->resumeflag, false);
+
+	return rv;
+}
+
+/*
+ * Shutdown routine
+ *
+ * return value
+ *      None
+ */
+static void mtip_pci_shutdown(struct pci_dev *pdev)
+{
+	struct driver_data *dd = pci_get_drvdata(pdev);
+	if (dd)
+		mtip_block_shutdown(dd);
+}
+
+/* Table of device ids supported by this driver. */
+static DEFINE_PCI_DEVICE_TABLE(mtip_pci_tbl) = {
+	{  PCI_DEVICE(PCI_VENDOR_ID_MICRON, P320_DEVICE_ID) },
+	{ 0 }
+};
+
+/* Structure that describes the PCI driver functions. */
+static struct pci_driver mtip_pci_driver = {
+	.name			= MTIP_DRV_NAME,
+	.id_table		= mtip_pci_tbl,
+	.probe			= mtip_pci_probe,
+	.remove			= mtip_pci_remove,
+	.suspend		= mtip_pci_suspend,
+	.resume			= mtip_pci_resume,
+	.shutdown		= mtip_pci_shutdown,
+};
+
+MODULE_DEVICE_TABLE(pci, mtip_pci_tbl);
+
+/*
+ * Module initialization function.
+ *
+ * Called once when the module is loaded. This function allocates a major
+ * block device number to the Cyclone devices and registers the PCI layer
+ * of the driver.
+ *
+ * Return value
+ *      0 on success else error code.
+ */
+static int __init mtip_init(void)
+{
+	printk(KERN_INFO MTIP_DRV_NAME " Version " MTIP_DRV_VERSION "\n");
+
+	/* Allocate a major block device number to use with this driver. */
+	mtip_major = register_blkdev(0, MTIP_DRV_NAME);
+	if (mtip_major < 0) {
+		printk(KERN_ERR "Unable to register block device (%d)\n",
+		mtip_major);
+		return -EBUSY;
+	}
+
+	/* Register our PCI operations. */
+	return pci_register_driver(&mtip_pci_driver);
+}
+
+/*
+ * Module de-initialization function.
+ *
+ * Called once when the module is unloaded. This function deallocates
+ * the major block device number allocated by mtip_init() and
+ * unregisters the PCI layer of the driver.
+ *
+ * Return value
+ *      none
+ */
+static void __exit mtip_exit(void)
+{
+	/* Release the allocated major block device number. */
+	unregister_blkdev(mtip_major, MTIP_DRV_NAME);
+
+	/* Unregister the PCI driver. */
+	pci_unregister_driver(&mtip_pci_driver);
+}
+
+MODULE_AUTHOR("Micron Technology, Inc");
+MODULE_DESCRIPTION("Micron RealSSD PCIe Block Driver");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(MTIP_DRV_VERSION);
+
+module_init(mtip_init);
+module_exit(mtip_exit);
+/*
+ * mtip32xx.h - Header file for the P320 SSD Block Driver
+ *   Copyright (C) 2011 Micron Technology, Inc.
+ *
+ * Portions of this code were derived from works subjected to the
+ * following copyright:
+ *    Copyright (C) 2009 Integrated Device Technology, Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * 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.
+ *
+ */
+
+#ifndef __MTIP32XX_H__
+#define __MTIP32XX_H__
+
+#include <linux/spinlock.h>
+#include <linux/rwsem.h>
+#include <linux/ata.h>
+#include <linux/interrupt.h>
+#include <linux/genhd.h>
+#include <linux/version.h>
+
+/* Offset of Subsystem Device ID in pci confoguration space */
+#define PCI_SUBSYSTEM_DEVICEID	0x2E
+
+/* offset of Device Control register in PCIe extended capabilites space */
+#define PCIE_CONFIG_EXT_DEVICE_CONTROL_OFFSET	0x48
+
+/* # of times to retry timed out IOs */
+#define MTIP_MAX_RETRIES	5
+
+/* Various timeout values in ms */
+#define MTIP_NCQ_COMMAND_TIMEOUT_MS       5000
+#define MTIP_IOCTL_COMMAND_TIMEOUT_MS     5000
+#define MTIP_INTERNAL_COMMAND_TIMEOUT_MS  5000
+
+/* check for timeouts every 500ms */
+#define MTIP_TIMEOUT_CHECK_PERIOD	500
+
+/* ftl rebuild */
+#define MTIP_FTL_REBUILD_OFFSET		142
+#define MTIP_FTL_REBUILD_MAGIC		0xED51
+#define MTIP_FTL_REBUILD_TIMEOUT_MS	2400000
+
+/* Macro to extract the tag bit number from a tag value. */
+#define MTIP_TAG_BIT(tag)	(tag & 0x1F)
+
+/*
+ * Macro to extract the tag index from a tag value. The index
+ * is used to access the correct s_active/Command Issue register based
+ * on the tag value.
+ */
+#define MTIP_TAG_INDEX(tag)	(tag >> 5)
+
+/*
+ * Maximum number of scatter gather entries
+ * a single command may have.
+ */
+#define MTIP_MAX_SG		128
+
+/*
+ * Maximum number of slot groups (Command Issue & s_active registers)
+ * NOTE: This is the driver maximum; check dd->slot_groups for actual value.
+ */
+#define MTIP_MAX_SLOT_GROUPS	8
+
+/* Internal command tag. */
+#define MTIP_TAG_INTERNAL	0
+
+/* Micron Vendor ID & P320x SSD Device ID */
+#define PCI_VENDOR_ID_MICRON    0x1344
+#define P320_DEVICE_ID		0x5150
+
+/* Driver name and version strings */
+#define MTIP_DRV_NAME		"mtip32xx"
+#define MTIP_DRV_VERSION	"1.2.6os3"
+
+/* Maximum number of minor device numbers per device. */
+#define MTIP_MAX_MINORS		16
+
+/* Maximum number of supported command slots. */
+#define MTIP_MAX_COMMAND_SLOTS	(MTIP_MAX_SLOT_GROUPS * 32)
+
+/*
+ * Per-tag bitfield size in longs.
+ * Linux bit manipulation functions
+ * (i.e. test_and_set_bit, find_next_zero_bit)
+ * manipulate memory in longs, so we try to make the math work.
+ * take the slot groups and find the number of longs, rounding up.
+ * Careful! i386 and x86_64 use different size longs!
+ */
+#define U32_PER_LONG	(sizeof(long) / sizeof(u32))
+#define SLOTBITS_IN_LONGS ((MTIP_MAX_SLOT_GROUPS + \
+					(U32_PER_LONG-1))/U32_PER_LONG)
+
+/* BAR number used to access the HBA registers. */
+#define MTIP_ABAR		5
+
+/* Forced Unit Access Bit */
+#define FUA_BIT			0x80
+
+#ifdef DEBUG
+ #define dbg_printk(format, arg...)	\
+	printk(pr_fmt(format), ##arg);
+#else
+ #define dbg_printk(format, arg...)
+#endif
+
+#define __force_bit2int (unsigned int __force)
+
+/* below are bit numbers in 'flags' defined in mtip_port */
+#define MTIP_FLAG_IC_ACTIVE_BIT			0
+#define MTIP_FLAG_EH_ACTIVE_BIT			1
+#define MTIP_FLAG_SVC_THD_ACTIVE_BIT		2
+#define MTIP_FLAG_ISSUE_CMDS_BIT		4
+#define MTIP_FLAG_REBUILD_BIT			5
+#define MTIP_FLAG_SVC_THD_SHOULD_STOP_BIT	8
+
+/* Register Frame Information Structure (FIS), host to device. */
+struct host_to_dev_fis {
+	/*
+	 * FIS type.
+	 * - 27h Register FIS, host to device.
+	 * - 34h Register FIS, device to host.
+	 * - 39h DMA Activate FIS, device to host.
+	 * - 41h DMA Setup FIS, bi-directional.
+	 * - 46h Data FIS, bi-directional.
+	 * - 58h BIST Activate FIS, bi-directional.
+	 * - 5Fh PIO Setup FIS, device to host.
+	 * - A1h Set Device Bits FIS, device to host.
+	 */
+	unsigned char type;
+	unsigned char opts;
+	unsigned char command;
+	unsigned char features;
+
+	union {
+		unsigned char lba_low;
+		unsigned char sector;
+	};
+	union {
+		unsigned char lba_mid;
+		unsigned char cyl_low;
+	};
+	union {
+		unsigned char lba_hi;
+		unsigned char cyl_hi;
+	};
+	union {
+		unsigned char device;
+		unsigned char head;
+	};
+
+	union {
+		unsigned char lba_low_ex;
+		unsigned char sector_ex;
+	};
+	union {
+		unsigned char lba_mid_ex;
+		unsigned char cyl_low_ex;
+	};
+	union {
+		unsigned char lba_hi_ex;
+		unsigned char cyl_hi_ex;
+	};
+	unsigned char features_ex;
+
+	unsigned char sect_count;
+	unsigned char sect_cnt_ex;
+	unsigned char res2;
+	unsigned char control;
+
+	unsigned int res3;
+};
+
+/* Command header structure. */
+struct mtip_cmd_hdr {
+	/*
+	 * Command options.
+	 * - Bits 31:16 Number of PRD entries.
+	 * - Bits 15:8 Unused in this implementation.
+	 * - Bit 7 Prefetch bit, informs the drive to prefetch PRD entries.
+	 * - Bit 6 Write bit, should be set when writing data to the device.
+	 * - Bit 5 Unused in this implementation.
+	 * - Bits 4:0 Length of the command FIS in DWords (DWord = 4 bytes).
+	 */
+	unsigned int opts;
+	/* This field is unsed when using NCQ. */
+	union {
+		unsigned int byte_count;
+		unsigned int status;
+	};
+	/*
+	 * Lower 32 bits of the command table address associated with this
+	 * header. The command table addresses must be 128 byte aligned.
+	 */
+	unsigned int ctba;
+	/*
+	 * If 64 bit addressing is used this field is the upper 32 bits
+	 * of the command table address associated with this command.
+	 */
+	unsigned int ctbau;
+	/* Reserved and unused. */
+	unsigned int res[4];
+};
+
+/* Command scatter gather structure (PRD). */
+struct mtip_cmd_sg {
+	/*
+	 * Low 32 bits of the data buffer address. For P320 this
+	 * address must be 8 byte aligned signified by bits 2:0 being
+	 * set to 0.
+	 */
+	unsigned int dba;
+	/*
+	 * When 64 bit addressing is used this field is the upper
+	 * 32 bits of the data buffer address.
+	 */
+	unsigned int dba_upper;
+	/* Unused. */
+	unsigned int reserved;
+	/*
+	 * Bit 31: interrupt when this data block has been transferred.
+	 * Bits 30..22: reserved
+	 * Bits 21..0: byte count (minus 1).  For P320 the byte count must be
+	 * 8 byte aligned signified by bits 2:0 being set to 1.
+	 */
+	unsigned int info;
+};
+struct mtip_port;
+
+/* Structure used to describe a command. */
+struct mtip_cmd {
+
+	struct mtip_cmd_hdr *command_header; /* ptr to command header entry */
+
+	dma_addr_t command_header_dma; /* corresponding physical address */
+
+	void *command; /* ptr to command table entry */
+
+	dma_addr_t command_dma; /* corresponding physical address */
+
+	void *comp_data; /* data passed to completion function comp_func() */
+	/*
+	 * Completion function called by the ISR upon completion of
+	 * a command.
+	 */
+	void (*comp_func)(struct mtip_port *port,
+				int tag,
+				void *data,
+				int status);
+	/* Additional callback function that may be called by comp_func() */
+	void (*async_callback)(void *data, int status);
+
+	void *async_data; /* Addl. data passed to async_callback() */
+
+	int scatter_ents; /* Number of scatter list entries used */
+
+	struct scatterlist sg[MTIP_MAX_SG]; /* Scatter list entries */
+
+	int retries; /* The number of retries left for this command. */
+
+	int direction; /* Data transfer direction */
+
+	unsigned long comp_time; /* command completion time, in jiffies */
+
+	atomic_t active; /* declares if this command sent to the drive. */
+};
+
+/* Structure used to describe a port. */
+struct mtip_port {
+	/* Pointer back to the driver data for this port. */
+	struct driver_data *dd;
+	/*
+	 * Used to determine if the data pointed to by the
+	 * identify field is valid.
+	 */
+	unsigned long identify_valid;
+	/* Base address of the memory mapped IO for the port. */
+	void __iomem *mmio;
+	/* Array of pointers to the memory mapped s_active registers. */
+	void __iomem *s_active[MTIP_MAX_SLOT_GROUPS];
+	/* Array of pointers to the memory mapped completed registers. */
+	void __iomem *completed[MTIP_MAX_SLOT_GROUPS];
+	/* Array of pointers to the memory mapped Command Issue registers. */
+	void __iomem *cmd_issue[MTIP_MAX_SLOT_GROUPS];
+	/*
+	 * Pointer to the beginning of the command header memory as used
+	 * by the driver.
+	 */
+	void *command_list;
+	/*
+	 * Pointer to the beginning of the command header memory as used
+	 * by the DMA.
+	 */
+	dma_addr_t command_list_dma;
+	/*
+	 * Pointer to the beginning of the RX FIS memory as used
+	 * by the driver.
+	 */
+	void *rxfis;
+	/*
+	 * Pointer to the beginning of the RX FIS memory as used
+	 * by the DMA.
+	 */
+	dma_addr_t rxfis_dma;
+	/*
+	 * Pointer to the beginning of the command table memory as used
+	 * by the driver.
+	 */
+	void *command_table;
+	/*
+	 * Pointer to the beginning of the command table memory as used
+	 * by the DMA.
+	 */
+	dma_addr_t command_tbl_dma;
+	/*
+	 * Pointer to the beginning of the identify data memory as used
+	 * by the driver.
+	 */
+	u16 *identify;
+	/*
+	 * Pointer to the beginning of the identify data memory as used
+	 * by the DMA.
+	 */
+	dma_addr_t identify_dma;
+	/*
+	 * Pointer to the beginning of a sector buffer that is used
+	 * by the driver when issuing internal commands.
+	 */
+	u16 *sector_buffer;
+	/*
+	 * Pointer to the beginning of a sector buffer that is used
+	 * by the DMA when the driver issues internal commands.
+	 */
+	dma_addr_t sector_buffer_dma;
+	/*
+	 * Bit significant, used to determine if a command slot has
+	 * been allocated. i.e. the slot is in use.  Bits are cleared
+	 * when the command slot and all associated data structures
+	 * are no longer needed.
+	 */
+	unsigned long allocated[SLOTBITS_IN_LONGS];
+	/*
+	 * used to queue commands when an internal command is in progress
+	 * or error handling is active
+	 */
+	unsigned long cmds_to_issue[SLOTBITS_IN_LONGS];
+	/*
+	 * Array of command slots. Structure includes pointers to the
+	 * command header and command table, and completion function and data
+	 * pointers.
+	 */
+	struct mtip_cmd commands[MTIP_MAX_COMMAND_SLOTS];
+	/* Used by mtip_service_thread to wait for an event */
+	wait_queue_head_t svc_wait;
+	/*
+	 * indicates the state of the port. Also, helps the service thread
+	 * to determine its action on wake up.
+	 */
+	unsigned long flags;
+	/*
+	 * Timer used to complete commands that have been active for too long.
+	 */
+	struct timer_list cmd_timer;
+	/*
+	 * Semaphore used to block threads if there are no
+	 * command slots available.
+	 */
+	struct semaphore cmd_slot;
+	/* Spinlock for working around command-issue bug. */
+	spinlock_t cmd_issue_lock;
+};
+
+/*
+ * Driver private data structure.
+ *
+ * One structure is allocated per probed device.
+ */
+struct driver_data {
+	void __iomem *mmio; /* Base address of the HBA registers. */
+
+	int major; /* Major device number. */
+
+	int instance; /* Instance number. First device probed is 0, ... */
+
+	struct gendisk *disk; /* Pointer to our gendisk structure. */
+
+	struct pci_dev *pdev; /* Pointer to the PCI device structure. */
+
+	struct request_queue *queue; /* Our request queue. */
+
+	struct mtip_port *port; /* Pointer to the port data structure. */
+
+	/* Tasklet used to process the bottom half of the ISR. */
+	struct tasklet_struct tasklet;
+
+	unsigned product_type; /* magic value declaring the product type */
+
+	unsigned slot_groups; /* number of slot groups the product supports */
+
+	atomic_t drv_cleanup_done; /* Atomic variable for SRSI */
+
+	unsigned long index; /* Index to determine the disk name */
+
+	unsigned int ftlrebuildflag; /* FTL rebuild flag */
+
+	atomic_t resumeflag; /* Atomic variable to track suspend/resume */
+
+	atomic_t eh_active; /* Flag for error handling tracking */
+
+	struct task_struct *mtip_svc_handler; /* task_struct of svc thd */
+};
+
+#endif
@@ -39,9 +39,6 @@
 #include <linux/list.h>
 #include <linux/delay.h>
 #include <linux/freezer.h>
-#include <linux/loop.h>
-#include <linux/falloc.h>
-#include <linux/fs.h>
  
 #include <xen/events.h>
 #include <xen/page.h>
@@ -362,7 +359,7 @@
 {
 	struct gnttab_map_grant_ref map[BLKIF_MAX_SEGMENTS_PER_REQUEST];
 	int i;
-	int nseg = req->nr_segments;
+	int nseg = req->u.rw.nr_segments;
 	int ret = 0;
  
 	/*
  
  
@@ -416,30 +413,25 @@
 	return ret;
 }
  
-static void xen_blk_discard(struct xen_blkif *blkif, struct blkif_request *req)
+static int dispatch_discard_io(struct xen_blkif *blkif,
+				struct blkif_request *req)
 {
 	int err = 0;
 	int status = BLKIF_RSP_OKAY;
 	struct block_device *bdev = blkif->vbd.bdev;
  
-	if (blkif->blk_backend_type == BLKIF_BACKEND_PHY)
-		/* just forward the discard request */
+	blkif->st_ds_req++;
+
+	xen_blkif_get(blkif);
+	if (blkif->blk_backend_type == BLKIF_BACKEND_PHY ||
+	    blkif->blk_backend_type == BLKIF_BACKEND_FILE) {
+		unsigned long secure = (blkif->vbd.discard_secure &&
+			(req->u.discard.flag & BLKIF_DISCARD_SECURE)) ?
+			BLKDEV_DISCARD_SECURE : 0;
 		err = blkdev_issue_discard(bdev,
 				req->u.discard.sector_number,
 				req->u.discard.nr_sectors,
-				GFP_KERNEL, 0);
-	else if (blkif->blk_backend_type == BLKIF_BACKEND_FILE) {
-		/* punch a hole in the backing file */
-		struct loop_device *lo = bdev->bd_disk->private_data;
-		struct file *file = lo->lo_backing_file;
-
-		if (file->f_op->fallocate)
-			err = file->f_op->fallocate(file,
-				FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE,
-				req->u.discard.sector_number << 9,
-				req->u.discard.nr_sectors << 9);
-		else
-			err = -EOPNOTSUPP;
+				GFP_KERNEL, secure);
 	} else
 		err = -EOPNOTSUPP;
  
@@ -449,7 +441,9 @@
 	} else if (err)
 		status = BLKIF_RSP_ERROR;
  
-	make_response(blkif, req->id, req->operation, status);
+	make_response(blkif, req->u.discard.id, req->operation, status);
+	xen_blkif_put(blkif);
+	return err;
 }
  
 static void xen_blk_drain_io(struct xen_blkif *blkif)
@@ -573,8 +567,11 @@
  
 		/* Apply all sanity checks to /private copy/ of request. */
 		barrier();
-
-		if (dispatch_rw_block_io(blkif, &req, pending_req))
+		if (unlikely(req.operation == BLKIF_OP_DISCARD)) {
+			free_req(pending_req);
+			if (dispatch_discard_io(blkif, &req))
+				break;
+		} else if (dispatch_rw_block_io(blkif, &req, pending_req))
 			break;
  
 		/* Yield point for this unbounded loop. */
@@ -633,10 +630,6 @@
 		blkif->st_f_req++;
 		operation = WRITE_FLUSH;
 		break;
-	case BLKIF_OP_DISCARD:
-		blkif->st_ds_req++;
-		operation = REQ_DISCARD;
-		break;
 	default:
 		operation = 0; /* make gcc happy */
 		goto fail_response;
@@ -644,9 +637,9 @@
 	}
  
 	/* Check that the number of segments is sane. */
-	nseg = req->nr_segments;
-	if (unlikely(nseg == 0 && operation != WRITE_FLUSH &&
-				operation != REQ_DISCARD) ||
+	nseg = req->u.rw.nr_segments;
+
+	if (unlikely(nseg == 0 && operation != WRITE_FLUSH) ||
 	    unlikely(nseg > BLKIF_MAX_SEGMENTS_PER_REQUEST)) {
 		pr_debug(DRV_PFX "Bad number of segments in request (%d)\n",
 			 nseg);
  
@@ -654,12 +647,12 @@
 		goto fail_response;
 	}
  
-	preq.dev           = req->handle;
+	preq.dev           = req->u.rw.handle;
 	preq.sector_number = req->u.rw.sector_number;
 	preq.nr_sects      = 0;
  
 	pending_req->blkif     = blkif;
-	pending_req->id        = req->id;
+	pending_req->id        = req->u.rw.id;
 	pending_req->operation = req->operation;
 	pending_req->status    = BLKIF_RSP_OKAY;
 	pending_req->nr_pages  = nseg;
@@ -707,7 +700,7 @@
 	 * the hypercall to unmap the grants - that is all done in
 	 * xen_blkbk_unmap.
 	 */
-	if (operation != REQ_DISCARD && xen_blkbk_map(req, pending_req, seg))
+	if (xen_blkbk_map(req, pending_req, seg))
 		goto fail_flush;
  
 	/*
  
  
@@ -739,23 +732,16 @@
  
 	/* This will be hit if the operation was a flush or discard. */
 	if (!bio) {
-		BUG_ON(operation != WRITE_FLUSH && operation != REQ_DISCARD);
+		BUG_ON(operation != WRITE_FLUSH);
  
-		if (operation == WRITE_FLUSH) {
-			bio = bio_alloc(GFP_KERNEL, 0);
-			if (unlikely(bio == NULL))
-				goto fail_put_bio;
+		bio = bio_alloc(GFP_KERNEL, 0);
+		if (unlikely(bio == NULL))
+			goto fail_put_bio;
  
-			biolist[nbio++] = bio;
-			bio->bi_bdev    = preq.bdev;
-			bio->bi_private = pending_req;
-			bio->bi_end_io  = end_block_io_op;
-		} else if (operation == REQ_DISCARD) {
-			xen_blk_discard(blkif, req);
-			xen_blkif_put(blkif);
-			free_req(pending_req);
-			return 0;
-		}
+		biolist[nbio++] = bio;
+		bio->bi_bdev    = preq.bdev;
+		bio->bi_private = pending_req;
+		bio->bi_end_io  = end_block_io_op;
 	}
  
 	/*
@@ -784,7 +770,7 @@
 	xen_blkbk_unmap(pending_req);
  fail_response:
 	/* Haven't submitted any bio's yet. */
-	make_response(blkif, req->id, req->operation, BLKIF_RSP_ERROR);
+	make_response(blkif, req->u.rw.id, req->operation, BLKIF_RSP_ERROR);
 	free_req(pending_req);
 	msleep(1); /* back off a bit */
 	return -EIO;
@@ -60,58 +60,66 @@
 	char dummy;
 };
  
-/* i386 protocol version */
-#pragma pack(push, 4)
-
 struct blkif_x86_32_request_rw {
+	uint8_t        nr_segments;  /* number of segments                   */
+	blkif_vdev_t   handle;       /* only for read/write requests         */
+	uint64_t       id;           /* private guest value, echoed in resp  */
 	blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
 	struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
-};
+} __attribute__((__packed__));
  
 struct blkif_x86_32_request_discard {
+	uint8_t        flag;         /* BLKIF_DISCARD_SECURE or zero         */
+	blkif_vdev_t   _pad1;        /* was "handle" for read/write requests */
+	uint64_t       id;           /* private guest value, echoed in resp  */
 	blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
-	uint64_t nr_sectors;
-};
+	uint64_t       nr_sectors;
+} __attribute__((__packed__));
  
 struct blkif_x86_32_request {
 	uint8_t        operation;    /* BLKIF_OP_???                         */
-	uint8_t        nr_segments;  /* number of segments                   */
-	blkif_vdev_t   handle;       /* only for read/write requests         */
-	uint64_t       id;           /* private guest value, echoed in resp  */
 	union {
 		struct blkif_x86_32_request_rw rw;
 		struct blkif_x86_32_request_discard discard;
 	} u;
-};
+} __attribute__((__packed__));
+
+/* i386 protocol version */
+#pragma pack(push, 4)
 struct blkif_x86_32_response {
 	uint64_t        id;              /* copied from request */
 	uint8_t         operation;       /* copied from request */
 	int16_t         status;          /* BLKIF_RSP_???       */
 };
 #pragma pack(pop)
-
 /* x86_64 protocol version */
  
 struct blkif_x86_64_request_rw {
+	uint8_t        nr_segments;  /* number of segments                   */
+	blkif_vdev_t   handle;       /* only for read/write requests         */
+	uint32_t       _pad1;        /* offsetof(blkif_reqest..,u.rw.id)==8  */
+	uint64_t       id;
 	blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
 	struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
-};
+} __attribute__((__packed__));
  
 struct blkif_x86_64_request_discard {
+	uint8_t        flag;         /* BLKIF_DISCARD_SECURE or zero         */
+	blkif_vdev_t   _pad1;        /* was "handle" for read/write requests */
+        uint32_t       _pad2;        /* offsetof(blkif_..,u.discard.id)==8   */
+	uint64_t       id;
 	blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
-	uint64_t nr_sectors;
-};
+	uint64_t       nr_sectors;
+} __attribute__((__packed__));
  
 struct blkif_x86_64_request {
 	uint8_t        operation;    /* BLKIF_OP_???                         */
-	uint8_t        nr_segments;  /* number of segments                   */
-	blkif_vdev_t   handle;       /* only for read/write requests         */
-	uint64_t       __attribute__((__aligned__(8))) id;
 	union {
 		struct blkif_x86_64_request_rw rw;
 		struct blkif_x86_64_request_discard discard;
 	} u;
-};
+} __attribute__((__packed__));
+
 struct blkif_x86_64_response {
 	uint64_t       __attribute__((__aligned__(8))) id;
 	uint8_t         operation;       /* copied from request */
@@ -156,6 +164,7 @@
 	/* Cached size parameter. */
 	sector_t		size;
 	bool			flush_support;
+	bool			discard_secure;
 };
  
 struct backend_info;
  
  
  
@@ -237,22 +246,23 @@
 {
 	int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST;
 	dst->operation = src->operation;
-	dst->nr_segments = src->nr_segments;
-	dst->handle = src->handle;
-	dst->id = src->id;
 	switch (src->operation) {
 	case BLKIF_OP_READ:
 	case BLKIF_OP_WRITE:
 	case BLKIF_OP_WRITE_BARRIER:
 	case BLKIF_OP_FLUSH_DISKCACHE:
+		dst->u.rw.nr_segments = src->u.rw.nr_segments;
+		dst->u.rw.handle = src->u.rw.handle;
+		dst->u.rw.id = src->u.rw.id;
 		dst->u.rw.sector_number = src->u.rw.sector_number;
 		barrier();
-		if (n > dst->nr_segments)
-			n = dst->nr_segments;
+		if (n > dst->u.rw.nr_segments)
+			n = dst->u.rw.nr_segments;
 		for (i = 0; i < n; i++)
 			dst->u.rw.seg[i] = src->u.rw.seg[i];
 		break;
 	case BLKIF_OP_DISCARD:
+		dst->u.discard.flag = src->u.discard.flag;
 		dst->u.discard.sector_number = src->u.discard.sector_number;
 		dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
 		break;
  
  
  
@@ -266,22 +276,23 @@
 {
 	int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST;
 	dst->operation = src->operation;
-	dst->nr_segments = src->nr_segments;
-	dst->handle = src->handle;
-	dst->id = src->id;
 	switch (src->operation) {
 	case BLKIF_OP_READ:
 	case BLKIF_OP_WRITE:
 	case BLKIF_OP_WRITE_BARRIER:
 	case BLKIF_OP_FLUSH_DISKCACHE:
+		dst->u.rw.nr_segments = src->u.rw.nr_segments;
+		dst->u.rw.handle = src->u.rw.handle;
+		dst->u.rw.id = src->u.rw.id;
 		dst->u.rw.sector_number = src->u.rw.sector_number;
 		barrier();
-		if (n > dst->nr_segments)
-			n = dst->nr_segments;
+		if (n > dst->u.rw.nr_segments)
+			n = dst->u.rw.nr_segments;
 		for (i = 0; i < n; i++)
 			dst->u.rw.seg[i] = src->u.rw.seg[i];
 		break;
 	case BLKIF_OP_DISCARD:
+		dst->u.discard.flag = src->u.discard.flag;
 		dst->u.discard.sector_number = src->u.discard.sector_number;
 		dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
 		break;
@@ -338,6 +338,9 @@
 	if (q && q->flush_flags)
 		vbd->flush_support = true;
  
+	if (q && blk_queue_secdiscard(q))
+		vbd->discard_secure = true;
+
 	DPRINTK("Successful creation of handle=%04x (dom=%u)\n",
 		handle, blkif->domid);
 	return 0;
@@ -419,6 +422,15 @@
 				}
 				state = 1;
 				blkif->blk_backend_type = BLKIF_BACKEND_PHY;
+			}
+			/* Optional. */
+			err = xenbus_printf(xbt, dev->nodename,
+				"discard-secure", "%d",
+				blkif->vbd.discard_secure);
+			if (err) {
+				xenbus_dev_fatal(dev, err,
+					"writting discard-secure");
+				goto kfree;
 			}
 		}
 	} else {
@@ -98,7 +98,8 @@
 	unsigned long shadow_free;
 	unsigned int feature_flush;
 	unsigned int flush_op;
-	unsigned int feature_discard;
+	unsigned int feature_discard:1;
+	unsigned int feature_secdiscard:1;
 	unsigned int discard_granularity;
 	unsigned int discard_alignment;
 	int is_ready;
  
@@ -135,15 +136,15 @@
 {
 	unsigned long free = info->shadow_free;
 	BUG_ON(free >= BLK_RING_SIZE);
-	info->shadow_free = info->shadow[free].req.id;
-	info->shadow[free].req.id = 0x0fffffee; /* debug */
+	info->shadow_free = info->shadow[free].req.u.rw.id;
+	info->shadow[free].req.u.rw.id = 0x0fffffee; /* debug */
 	return free;
 }
  
 static void add_id_to_freelist(struct blkfront_info *info,
 			       unsigned long id)
 {
-	info->shadow[id].req.id  = info->shadow_free;
+	info->shadow[id].req.u.rw.id  = info->shadow_free;
 	info->shadow[id].request = NULL;
 	info->shadow_free = id;
 }
@@ -156,7 +157,7 @@
 	if (end > nr_minors) {
 		unsigned long *bitmap, *old;
  
-		bitmap = kzalloc(BITS_TO_LONGS(end) * sizeof(*bitmap),
+		bitmap = kcalloc(BITS_TO_LONGS(end), sizeof(*bitmap),
 				 GFP_KERNEL);
 		if (bitmap == NULL)
 			return -ENOMEM;
  
@@ -287,9 +288,9 @@
 	id = get_id_from_freelist(info);
 	info->shadow[id].request = req;
  
-	ring_req->id = id;
+	ring_req->u.rw.id = id;
 	ring_req->u.rw.sector_number = (blkif_sector_t)blk_rq_pos(req);
-	ring_req->handle = info->handle;
+	ring_req->u.rw.handle = info->handle;
  
 	ring_req->operation = rq_data_dir(req) ?
 		BLKIF_OP_WRITE : BLKIF_OP_READ;
  
  
  
  
@@ -305,16 +306,21 @@
 		ring_req->operation = info->flush_op;
 	}
  
-	if (unlikely(req->cmd_flags & REQ_DISCARD)) {
+	if (unlikely(req->cmd_flags & (REQ_DISCARD | REQ_SECURE))) {
 		/* id, sector_number and handle are set above. */
 		ring_req->operation = BLKIF_OP_DISCARD;
-		ring_req->nr_segments = 0;
 		ring_req->u.discard.nr_sectors = blk_rq_sectors(req);
+		if ((req->cmd_flags & REQ_SECURE) && info->feature_secdiscard)
+			ring_req->u.discard.flag = BLKIF_DISCARD_SECURE;
+		else
+			ring_req->u.discard.flag = 0;
 	} else {
-		ring_req->nr_segments = blk_rq_map_sg(req->q, req, info->sg);
-		BUG_ON(ring_req->nr_segments > BLKIF_MAX_SEGMENTS_PER_REQUEST);
+		ring_req->u.rw.nr_segments = blk_rq_map_sg(req->q, req,
+							   info->sg);
+		BUG_ON(ring_req->u.rw.nr_segments >
+		       BLKIF_MAX_SEGMENTS_PER_REQUEST);
  
-		for_each_sg(info->sg, sg, ring_req->nr_segments, i) {
+		for_each_sg(info->sg, sg, ring_req->u.rw.nr_segments, i) {
 			buffer_mfn = pfn_to_mfn(page_to_pfn(sg_page(sg)));
 			fsect = sg->offset >> 9;
 			lsect = fsect + (sg->length >> 9) - 1;
@@ -424,6 +430,8 @@
 		blk_queue_max_discard_sectors(rq, get_capacity(gd));
 		rq->limits.discard_granularity = info->discard_granularity;
 		rq->limits.discard_alignment = info->discard_alignment;
+		if (info->feature_secdiscard)
+			queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD, rq);
 	}
  
 	/* Hard sector size and max sectors impersonate the equiv. hardware. */
@@ -705,7 +713,9 @@
 static void blkif_completion(struct blk_shadow *s)
 {
 	int i;
-	for (i = 0; i < s->req.nr_segments; i++)
+	/* Do not let BLKIF_OP_DISCARD as nr_segment is in the same place
+	 * flag. */
+	for (i = 0; i < s->req.u.rw.nr_segments; i++)
 		gnttab_end_foreign_access(s->req.u.rw.seg[i].gref, 0, 0UL);
 }
  
@@ -736,7 +746,8 @@
 		id   = bret->id;
 		req  = info->shadow[id].request;
  
-		blkif_completion(&info->shadow[id]);
+		if (bret->operation != BLKIF_OP_DISCARD)
+			blkif_completion(&info->shadow[id]);
  
 		add_id_to_freelist(info, id);
  
  
@@ -749,7 +760,9 @@
 					   info->gd->disk_name);
 				error = -EOPNOTSUPP;
 				info->feature_discard = 0;
+				info->feature_secdiscard = 0;
 				queue_flag_clear(QUEUE_FLAG_DISCARD, rq);
+				queue_flag_clear(QUEUE_FLAG_SECDISCARD, rq);
 			}
 			__blk_end_request_all(req, error);
 			break;
@@ -763,7 +776,7 @@
 				error = -EOPNOTSUPP;
 			}
 			if (unlikely(bret->status == BLKIF_RSP_ERROR &&
-				     info->shadow[id].req.nr_segments == 0)) {
+				     info->shadow[id].req.u.rw.nr_segments == 0)) {
 				printk(KERN_WARNING "blkfront: %s: empty write %s op failed\n",
 				       info->flush_op == BLKIF_OP_WRITE_BARRIER ?
 				       "barrier" :  "flush disk cache",
@@ -984,8 +997,8 @@
 	INIT_WORK(&info->work, blkif_restart_queue);
  
 	for (i = 0; i < BLK_RING_SIZE; i++)
-		info->shadow[i].req.id = i+1;
-	info->shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
+		info->shadow[i].req.u.rw.id = i+1;
+	info->shadow[BLK_RING_SIZE-1].req.u.rw.id = 0x0fffffff;
  
 	/* Front end dir is a number, which is used as the id. */
 	info->handle = simple_strtoul(strrchr(dev->nodename, '/')+1, NULL, 0);
  
@@ -1019,9 +1032,9 @@
 	/* Stage 2: Set up free list. */
 	memset(&info->shadow, 0, sizeof(info->shadow));
 	for (i = 0; i < BLK_RING_SIZE; i++)
-		info->shadow[i].req.id = i+1;
+		info->shadow[i].req.u.rw.id = i+1;
 	info->shadow_free = info->ring.req_prod_pvt;
-	info->shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
+	info->shadow[BLK_RING_SIZE-1].req.u.rw.id = 0x0fffffff;
  
 	/* Stage 3: Find pending requests and requeue them. */
 	for (i = 0; i < BLK_RING_SIZE; i++) {
  
  
@@ -1034,17 +1047,19 @@
 		*req = copy[i].req;
  
 		/* We get a new request id, and must reset the shadow state. */
-		req->id = get_id_from_freelist(info);
-		memcpy(&info->shadow[req->id], &copy[i], sizeof(copy[i]));
+		req->u.rw.id = get_id_from_freelist(info);
+		memcpy(&info->shadow[req->u.rw.id], &copy[i], sizeof(copy[i]));
  
+		if (req->operation != BLKIF_OP_DISCARD) {
 		/* Rewrite any grant references invalidated by susp/resume. */
-		for (j = 0; j < req->nr_segments; j++)
-			gnttab_grant_foreign_access_ref(
-				req->u.rw.seg[j].gref,
-				info->xbdev->otherend_id,
-				pfn_to_mfn(info->shadow[req->id].frame[j]),
-				rq_data_dir(info->shadow[req->id].request));
-		info->shadow[req->id].req = *req;
+			for (j = 0; j < req->u.rw.nr_segments; j++)
+				gnttab_grant_foreign_access_ref(
+					req->u.rw.seg[j].gref,
+					info->xbdev->otherend_id,
+					pfn_to_mfn(info->shadow[req->u.rw.id].frame[j]),
+					rq_data_dir(info->shadow[req->u.rw.id].request));
+		}
+		info->shadow[req->u.rw.id].req = *req;
  
 		info->ring.req_prod_pvt++;
 	}
  
@@ -1135,11 +1150,13 @@
 	char *type;
 	unsigned int discard_granularity;
 	unsigned int discard_alignment;
+	unsigned int discard_secure;
  
 	type = xenbus_read(XBT_NIL, info->xbdev->otherend, "type", NULL);
 	if (IS_ERR(type))
 		return;
  
+	info->feature_secdiscard = 0;
 	if (strncmp(type, "phy", 3) == 0) {
 		err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
 			"discard-granularity", "%u", &discard_granularity,
@@ -1150,6 +1167,12 @@
 			info->discard_granularity = discard_granularity;
 			info->discard_alignment = discard_alignment;
 		}
+		err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
+			    "discard-secure", "%d", &discard_secure,
+			    NULL);
+		if (!err)
+			info->feature_secdiscard = discard_secure;
+
 	} else if (strncmp(type, "file", 4) == 0)
 		info->feature_discard = 1;
  
@@ -84,6 +84,21 @@
  *     e07154r6-Data_Set_Management_Proposal_for_ATA-ACS2.doc
  * http://www.seagate.com/staticfiles/support/disc/manuals/
  *     Interface%20manuals/100293068c.pdf
+ * The backend can optionally provide three extra XenBus attributes to
+ * further optimize the discard functionality:
+ * 'discard-aligment' - Devices that support discard functionality may
+ * internally allocate space in units that are bigger than the exported
+ * logical block size. The discard-alignment parameter indicates how many bytes
+ * the beginning of the partition is offset from the internal allocation unit's
+ * natural alignment.
+ * 'discard-granularity'  - Devices that support discard functionality may
+ * internally allocate space using units that are bigger than the logical block
+ * size. The discard-granularity parameter indicates the size of the internal
+ * allocation unit in bytes if reported by the device. Otherwise the
+ * discard-granularity will be set to match the device's physical block size.
+ * 'discard-secure' - All copies of the discarded sectors (potentially created
+ * by garbage collection) must also be erased.  To use this feature, the flag
+ * BLKIF_DISCARD_SECURE must be set in the blkif_request_trim.
  */
 #define BLKIF_OP_DISCARD           5
  
@@ -95,6 +110,12 @@
 #define BLKIF_MAX_SEGMENTS_PER_REQUEST 11
  
 struct blkif_request_rw {
+	uint8_t        nr_segments;  /* number of segments                   */
+	blkif_vdev_t   handle;       /* only for read/write requests         */
+#ifdef CONFIG_X86_64
+	uint32_t       _pad1;	     /* offsetof(blkif_request,u.rw.id) == 8 */
+#endif
+	uint64_t       id;           /* private guest value, echoed in resp  */
 	blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
 	struct blkif_request_segment {
 		grant_ref_t gref;        /* reference to I/O buffer frame        */
  
  
  
  
@@ -102,23 +123,28 @@
 		/* @last_sect: last sector in frame to transfer (inclusive).     */
 		uint8_t     first_sect, last_sect;
 	} seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
-};
+} __attribute__((__packed__));
  
 struct blkif_request_discard {
+	uint8_t        flag;         /* BLKIF_DISCARD_SECURE or zero.        */
+#define BLKIF_DISCARD_SECURE (1<<0)  /* ignored if discard-secure=0          */
+	blkif_vdev_t   _pad1;        /* only for read/write requests         */
+#ifdef CONFIG_X86_64
+	uint32_t       _pad2;        /* offsetof(blkif_req..,u.discard.id)==8*/
+#endif
+	uint64_t       id;           /* private guest value, echoed in resp  */
 	blkif_sector_t sector_number;
-	uint64_t nr_sectors;
-};
+	uint64_t       nr_sectors;
+	uint8_t        _pad3;
+} __attribute__((__packed__));
  
 struct blkif_request {
 	uint8_t        operation;    /* BLKIF_OP_???                         */
-	uint8_t        nr_segments;  /* number of segments                   */
-	blkif_vdev_t   handle;       /* only for read/write requests         */
-	uint64_t       id;           /* private guest value, echoed in resp  */
 	union {
 		struct blkif_request_rw rw;
 		struct blkif_request_discard discard;
 	} u;
-};
+} __attribute__((__packed__));
  
 struct blkif_response {
 	uint64_t        id;              /* copied from request */
...	...	@@ -116,6 +116,8 @@
116	116
117	117	source "drivers/block/paride/Kconfig"
118	118
	119	+source "drivers/block/mtip32xx/Kconfig"
	120	+
119	121	config BLK_CPQ_DA
120	122	tristate "Compaq SMART2 support"
121	123	depends on PCI && VIRT_TO_BUS
...	...	@@ -39,6 +39,7 @@
39	39	obj-$(CONFIG_XEN_BLKDEV_BACKEND) += xen-blkback/
40	40	obj-$(CONFIG_BLK_DEV_DRBD) += drbd/
41	41	obj-$(CONFIG_BLK_DEV_RBD) += rbd.o
	42	+obj-$(CONFIG_BLK_DEV_PCIESSD_MTIP32XX) += mtip32xx/
42	43
43	44	swim_mod-y := swim.o swim_asm.o
	1	+#
	2	+# mtip32xx device driver configuration
	3	+#
	4	+
	5	+config BLK_DEV_PCIESSD_MTIP32XX
	6	+ tristate "Block Device Driver for Micron PCIe SSDs"
	7	+ depends on HOTPLUG_PCI_PCIE
	8	+ help
	9	+ This enables the block driver for Micron PCIe SSDs.
	1	+#
	2	+# Makefile for Block device driver for Micron PCIe SSD
	3	+#
	4	+
	5	+obj-$(CONFIG_BLK_DEV_PCIESSD_MTIP32XX) += mtip32xx.o
	1	+/*
	2	+ * mtip32xx.h - Header file for the P320 SSD Block Driver
	3	+ * Copyright (C) 2011 Micron Technology, Inc.
	4	+ *
	5	+ * Portions of this code were derived from works subjected to the
	6	+ * following copyright:
	7	+ * Copyright (C) 2009 Integrated Device Technology, Inc.
	8	+ *
	9	+ * This program is free software; you can redistribute it and/or modify
	10	+ * it under the terms of the GNU General Public License as published by
	11	+ * the Free Software Foundation; either version 2 of the License, or
	12	+ * (at your option) any later version.
	13	+ *
	14	+ * This program is distributed in the hope that it will be useful,
	15	+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
	16	+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	17	+ * GNU General Public License for more details.
	18	+ *
	19	+ */
	20	+
	21	+#ifndef __MTIP32XX_H__
	22	+#define __MTIP32XX_H__
	23	+
	24	+#include <linux/spinlock.h>
	25	+#include <linux/rwsem.h>
	26	+#include <linux/ata.h>
	27	+#include <linux/interrupt.h>
	28	+#include <linux/genhd.h>
	29	+#include <linux/version.h>
	30	+
	31	+/* Offset of Subsystem Device ID in pci confoguration space */
	32	+#define PCI_SUBSYSTEM_DEVICEID 0x2E
	33	+
	34	+/* offset of Device Control register in PCIe extended capabilites space */
	35	+#define PCIE_CONFIG_EXT_DEVICE_CONTROL_OFFSET 0x48
	36	+
	37	+/* # of times to retry timed out IOs */
	38	+#define MTIP_MAX_RETRIES 5
	39	+
	40	+/* Various timeout values in ms */
	41	+#define MTIP_NCQ_COMMAND_TIMEOUT_MS 5000
	42	+#define MTIP_IOCTL_COMMAND_TIMEOUT_MS 5000
	43	+#define MTIP_INTERNAL_COMMAND_TIMEOUT_MS 5000
	44	+
	45	+/* check for timeouts every 500ms */
	46	+#define MTIP_TIMEOUT_CHECK_PERIOD 500
	47	+
	48	+/* ftl rebuild */
	49	+#define MTIP_FTL_REBUILD_OFFSET 142
	50	+#define MTIP_FTL_REBUILD_MAGIC 0xED51
	51	+#define MTIP_FTL_REBUILD_TIMEOUT_MS 2400000
	52	+
	53	+/* Macro to extract the tag bit number from a tag value. */
	54	+#define MTIP_TAG_BIT(tag) (tag & 0x1F)
	55	+
	56	+/*
	57	+ * Macro to extract the tag index from a tag value. The index
	58	+ * is used to access the correct s_active/Command Issue register based
	59	+ * on the tag value.
	60	+ */
	61	+#define MTIP_TAG_INDEX(tag) (tag >> 5)
	62	+
	63	+/*
	64	+ * Maximum number of scatter gather entries
	65	+ * a single command may have.
	66	+ */
	67	+#define MTIP_MAX_SG 128
	68	+
	69	+/*
	70	+ * Maximum number of slot groups (Command Issue & s_active registers)
	71	+ * NOTE: This is the driver maximum; check dd->slot_groups for actual value.
	72	+ */
	73	+#define MTIP_MAX_SLOT_GROUPS 8
	74	+
	75	+/* Internal command tag. */
	76	+#define MTIP_TAG_INTERNAL 0
	77	+
	78	+/* Micron Vendor ID & P320x SSD Device ID */
	79	+#define PCI_VENDOR_ID_MICRON 0x1344
	80	+#define P320_DEVICE_ID 0x5150
	81	+
	82	+/* Driver name and version strings */
	83	+#define MTIP_DRV_NAME "mtip32xx"
	84	+#define MTIP_DRV_VERSION "1.2.6os3"
	85	+
	86	+/* Maximum number of minor device numbers per device. */
	87	+#define MTIP_MAX_MINORS 16
	88	+
	89	+/* Maximum number of supported command slots. */
	90	+#define MTIP_MAX_COMMAND_SLOTS (MTIP_MAX_SLOT_GROUPS * 32)
	91	+
	92	+/*
	93	+ * Per-tag bitfield size in longs.
	94	+ * Linux bit manipulation functions
	95	+ * (i.e. test_and_set_bit, find_next_zero_bit)
	96	+ * manipulate memory in longs, so we try to make the math work.
	97	+ * take the slot groups and find the number of longs, rounding up.
	98	+ * Careful! i386 and x86_64 use different size longs!
	99	+ */
	100	+#define U32_PER_LONG (sizeof(long) / sizeof(u32))
	101	+#define SLOTBITS_IN_LONGS ((MTIP_MAX_SLOT_GROUPS + \
	102	+ (U32_PER_LONG-1))/U32_PER_LONG)
	103	+
	104	+/* BAR number used to access the HBA registers. */
	105	+#define MTIP_ABAR 5
	106	+
	107	+/* Forced Unit Access Bit */
	108	+#define FUA_BIT 0x80
	109	+
	110	+#ifdef DEBUG
	111	+ #define dbg_printk(format, arg...) \
	112	+ printk(pr_fmt(format), ##arg);
	113	+#else
	114	+ #define dbg_printk(format, arg...)
	115	+#endif
	116	+
	117	+#define __force_bit2int (unsigned int __force)
	118	+
	119	+/* below are bit numbers in 'flags' defined in mtip_port */
	120	+#define MTIP_FLAG_IC_ACTIVE_BIT 0
	121	+#define MTIP_FLAG_EH_ACTIVE_BIT 1
	122	+#define MTIP_FLAG_SVC_THD_ACTIVE_BIT 2
	123	+#define MTIP_FLAG_ISSUE_CMDS_BIT 4
	124	+#define MTIP_FLAG_REBUILD_BIT 5
	125	+#define MTIP_FLAG_SVC_THD_SHOULD_STOP_BIT 8
	126	+
	127	+/* Register Frame Information Structure (FIS), host to device. */
	128	+struct host_to_dev_fis {
	129	+ /*
	130	+ * FIS type.
	131	+ * - 27h Register FIS, host to device.
	132	+ * - 34h Register FIS, device to host.
	133	+ * - 39h DMA Activate FIS, device to host.
	134	+ * - 41h DMA Setup FIS, bi-directional.
	135	+ * - 46h Data FIS, bi-directional.
	136	+ * - 58h BIST Activate FIS, bi-directional.
	137	+ * - 5Fh PIO Setup FIS, device to host.
	138	+ * - A1h Set Device Bits FIS, device to host.
	139	+ */
	140	+ unsigned char type;
	141	+ unsigned char opts;
	142	+ unsigned char command;
	143	+ unsigned char features;
	144	+
	145	+ union {
	146	+ unsigned char lba_low;
	147	+ unsigned char sector;
	148	+ };
	149	+ union {
	150	+ unsigned char lba_mid;
	151	+ unsigned char cyl_low;
	152	+ };
	153	+ union {
	154	+ unsigned char lba_hi;
	155	+ unsigned char cyl_hi;
	156	+ };
	157	+ union {
	158	+ unsigned char device;
	159	+ unsigned char head;
	160	+ };
	161	+
	162	+ union {
	163	+ unsigned char lba_low_ex;
	164	+ unsigned char sector_ex;
	165	+ };
	166	+ union {
	167	+ unsigned char lba_mid_ex;
	168	+ unsigned char cyl_low_ex;
	169	+ };
	170	+ union {
	171	+ unsigned char lba_hi_ex;
	172	+ unsigned char cyl_hi_ex;
	173	+ };
	174	+ unsigned char features_ex;
	175	+
	176	+ unsigned char sect_count;
	177	+ unsigned char sect_cnt_ex;
	178	+ unsigned char res2;
	179	+ unsigned char control;
	180	+
	181	+ unsigned int res3;
	182	+};
	183	+
	184	+/* Command header structure. */
	185	+struct mtip_cmd_hdr {
	186	+ /*
	187	+ * Command options.
	188	+ * - Bits 31:16 Number of PRD entries.
	189	+ * - Bits 15:8 Unused in this implementation.
	190	+ * - Bit 7 Prefetch bit, informs the drive to prefetch PRD entries.
	191	+ * - Bit 6 Write bit, should be set when writing data to the device.
	192	+ * - Bit 5 Unused in this implementation.
	193	+ * - Bits 4:0 Length of the command FIS in DWords (DWord = 4 bytes).
	194	+ */
	195	+ unsigned int opts;
	196	+ /* This field is unsed when using NCQ. */
	197	+ union {
	198	+ unsigned int byte_count;
	199	+ unsigned int status;
	200	+ };
	201	+ /*
	202	+ * Lower 32 bits of the command table address associated with this
	203	+ * header. The command table addresses must be 128 byte aligned.
	204	+ */
	205	+ unsigned int ctba;
	206	+ /*
	207	+ * If 64 bit addressing is used this field is the upper 32 bits
	208	+ * of the command table address associated with this command.
	209	+ */
	210	+ unsigned int ctbau;
	211	+ /* Reserved and unused. */
	212	+ unsigned int res[4];
	213	+};
	214	+
	215	+/* Command scatter gather structure (PRD). */
	216	+struct mtip_cmd_sg {
	217	+ /*
	218	+ * Low 32 bits of the data buffer address. For P320 this
	219	+ * address must be 8 byte aligned signified by bits 2:0 being
	220	+ * set to 0.
	221	+ */
	222	+ unsigned int dba;
	223	+ /*
	224	+ * When 64 bit addressing is used this field is the upper
	225	+ * 32 bits of the data buffer address.
	226	+ */
	227	+ unsigned int dba_upper;
	228	+ /* Unused. */
	229	+ unsigned int reserved;
	230	+ /*
	231	+ * Bit 31: interrupt when this data block has been transferred.
	232	+ * Bits 30..22: reserved
	233	+ * Bits 21..0: byte count (minus 1). For P320 the byte count must be
	234	+ * 8 byte aligned signified by bits 2:0 being set to 1.
	235	+ */
	236	+ unsigned int info;
	237	+};
	238	+struct mtip_port;
	239	+
	240	+/* Structure used to describe a command. */
	241	+struct mtip_cmd {
	242	+
	243	+ struct mtip_cmd_hdr command_header; / ptr to command header entry */
	244	+
	245	+ dma_addr_t command_header_dma; /* corresponding physical address */
	246	+
	247	+ void command; / ptr to command table entry */
	248	+
	249	+ dma_addr_t command_dma; /* corresponding physical address */
	250	+
	251	+ void comp_data; / data passed to completion function comp_func() */
	252	+ /*
	253	+ * Completion function called by the ISR upon completion of
	254	+ * a command.
	255	+ */
	256	+ void (comp_func)(struct mtip_port port,
	257	+ int tag,
	258	+ void *data,
	259	+ int status);
	260	+ /* Additional callback function that may be called by comp_func() */
	261	+ void (async_callback)(void data, int status);
	262	+
	263	+ void async_data; / Addl. data passed to async_callback() */
	264	+
	265	+ int scatter_ents; /* Number of scatter list entries used */
	266	+
	267	+ struct scatterlist sg[MTIP_MAX_SG]; /* Scatter list entries */
	268	+
	269	+ int retries; /* The number of retries left for this command. */
	270	+
	271	+ int direction; /* Data transfer direction */
	272	+
	273	+ unsigned long comp_time; /* command completion time, in jiffies */
	274	+
	275	+ atomic_t active; /* declares if this command sent to the drive. */
	276	+};
	277	+
	278	+/* Structure used to describe a port. */
	279	+struct mtip_port {
	280	+ /* Pointer back to the driver data for this port. */
	281	+ struct driver_data *dd;
	282	+ /*
	283	+ * Used to determine if the data pointed to by the
	284	+ * identify field is valid.
	285	+ */
	286	+ unsigned long identify_valid;
	287	+ /* Base address of the memory mapped IO for the port. */
	288	+ void __iomem *mmio;
	289	+ /* Array of pointers to the memory mapped s_active registers. */
	290	+ void __iomem *s_active[MTIP_MAX_SLOT_GROUPS];
	291	+ /* Array of pointers to the memory mapped completed registers. */
	292	+ void __iomem *completed[MTIP_MAX_SLOT_GROUPS];
	293	+ /* Array of pointers to the memory mapped Command Issue registers. */
	294	+ void __iomem *cmd_issue[MTIP_MAX_SLOT_GROUPS];
	295	+ /*
	296	+ * Pointer to the beginning of the command header memory as used
	297	+ * by the driver.
	298	+ */
	299	+ void *command_list;
	300	+ /*
	301	+ * Pointer to the beginning of the command header memory as used
	302	+ * by the DMA.
	303	+ */
	304	+ dma_addr_t command_list_dma;
	305	+ /*
	306	+ * Pointer to the beginning of the RX FIS memory as used
	307	+ * by the driver.
	308	+ */
	309	+ void *rxfis;
	310	+ /*
	311	+ * Pointer to the beginning of the RX FIS memory as used
	312	+ * by the DMA.
	313	+ */
	314	+ dma_addr_t rxfis_dma;
	315	+ /*
	316	+ * Pointer to the beginning of the command table memory as used
	317	+ * by the driver.
	318	+ */
	319	+ void *command_table;
	320	+ /*
	321	+ * Pointer to the beginning of the command table memory as used
	322	+ * by the DMA.
	323	+ */
	324	+ dma_addr_t command_tbl_dma;
	325	+ /*
	326	+ * Pointer to the beginning of the identify data memory as used
	327	+ * by the driver.
	328	+ */
	329	+ u16 *identify;
	330	+ /*
	331	+ * Pointer to the beginning of the identify data memory as used
	332	+ * by the DMA.
	333	+ */
	334	+ dma_addr_t identify_dma;
	335	+ /*
	336	+ * Pointer to the beginning of a sector buffer that is used
	337	+ * by the driver when issuing internal commands.
	338	+ */
	339	+ u16 *sector_buffer;
	340	+ /*
	341	+ * Pointer to the beginning of a sector buffer that is used
	342	+ * by the DMA when the driver issues internal commands.
	343	+ */
	344	+ dma_addr_t sector_buffer_dma;
	345	+ /*
	346	+ * Bit significant, used to determine if a command slot has
	347	+ * been allocated. i.e. the slot is in use. Bits are cleared
	348	+ * when the command slot and all associated data structures
	349	+ * are no longer needed.
	350	+ */
	351	+ unsigned long allocated[SLOTBITS_IN_LONGS];
	352	+ /*
	353	+ * used to queue commands when an internal command is in progress
	354	+ * or error handling is active
	355	+ */
	356	+ unsigned long cmds_to_issue[SLOTBITS_IN_LONGS];
	357	+ /*
	358	+ * Array of command slots. Structure includes pointers to the
	359	+ * command header and command table, and completion function and data
	360	+ * pointers.
	361	+ */
	362	+ struct mtip_cmd commands[MTIP_MAX_COMMAND_SLOTS];
	363	+ /* Used by mtip_service_thread to wait for an event */
	364	+ wait_queue_head_t svc_wait;
	365	+ /*
	366	+ * indicates the state of the port. Also, helps the service thread
	367	+ * to determine its action on wake up.
	368	+ */
	369	+ unsigned long flags;
	370	+ /*
	371	+ * Timer used to complete commands that have been active for too long.
	372	+ */
	373	+ struct timer_list cmd_timer;
	374	+ /*
	375	+ * Semaphore used to block threads if there are no
	376	+ * command slots available.
	377	+ */
	378	+ struct semaphore cmd_slot;
	379	+ /* Spinlock for working around command-issue bug. */
	380	+ spinlock_t cmd_issue_lock;
	381	+};
	382	+
	383	+/*
	384	+ * Driver private data structure.
	385	+ *
	386	+ * One structure is allocated per probed device.
	387	+ */
	388	+struct driver_data {
	389	+ void __iomem mmio; / Base address of the HBA registers. */
	390	+
	391	+ int major; /* Major device number. */
	392	+
	393	+ int instance; /* Instance number. First device probed is 0, ... */
	394	+
	395	+ struct gendisk disk; / Pointer to our gendisk structure. */
	396	+
	397	+ struct pci_dev pdev; / Pointer to the PCI device structure. */
	398	+
	399	+ struct request_queue queue; / Our request queue. */
	400	+
	401	+ struct mtip_port port; / Pointer to the port data structure. */
	402	+
	403	+ /* Tasklet used to process the bottom half of the ISR. */
	404	+ struct tasklet_struct tasklet;
	405	+
	406	+ unsigned product_type; /* magic value declaring the product type */
	407	+
	408	+ unsigned slot_groups; /* number of slot groups the product supports */
	409	+
	410	+ atomic_t drv_cleanup_done; /* Atomic variable for SRSI */
	411	+
	412	+ unsigned long index; /* Index to determine the disk name */
	413	+
	414	+ unsigned int ftlrebuildflag; /* FTL rebuild flag */
	415	+
	416	+ atomic_t resumeflag; /* Atomic variable to track suspend/resume */
	417	+
	418	+ atomic_t eh_active; /* Flag for error handling tracking */
	419	+
	420	+ struct task_struct mtip_svc_handler; / task_struct of svc thd */
	421	+};
	422	+
	423	+#endif
...	...	@@ -39,9 +39,6 @@
39	39	#include <linux/list.h>
40	40	#include <linux/delay.h>
41	41	#include <linux/freezer.h>
42		-#include <linux/loop.h>
43		-#include <linux/falloc.h>
44		-#include <linux/fs.h>
45	42
46	43	#include <xen/events.h>
47	44	#include <xen/page.h>
...	...	@@ -362,7 +359,7 @@
362	359	{
363	360	struct gnttab_map_grant_ref map[BLKIF_MAX_SEGMENTS_PER_REQUEST];
364	361	int i;
365		- int nseg = req->nr_segments;
	362	+ int nseg = req->u.rw.nr_segments;
366	363	int ret = 0;
367	364
368	365	/*
369	366
370	367
...	...	@@ -416,30 +413,25 @@
416	413	return ret;
417	414	}
418	415
419		-static void xen_blk_discard(struct xen_blkif blkif, struct blkif_request req)
	416	+static int dispatch_discard_io(struct xen_blkif *blkif,
	417	+ struct blkif_request *req)
420	418	{
421	419	int err = 0;
422	420	int status = BLKIF_RSP_OKAY;
423	421	struct block_device *bdev = blkif->vbd.bdev;
424	422
425		- if (blkif->blk_backend_type == BLKIF_BACKEND_PHY)
426		- /* just forward the discard request */
	423	+ blkif->st_ds_req++;
	424	+
	425	+ xen_blkif_get(blkif);
	426	+ if (blkif->blk_backend_type == BLKIF_BACKEND_PHY \|\|
	427	+ blkif->blk_backend_type == BLKIF_BACKEND_FILE) {
	428	+ unsigned long secure = (blkif->vbd.discard_secure &&
	429	+ (req->u.discard.flag & BLKIF_DISCARD_SECURE)) ?
	430	+ BLKDEV_DISCARD_SECURE : 0;
427	431	err = blkdev_issue_discard(bdev,
428	432	req->u.discard.sector_number,
429	433	req->u.discard.nr_sectors,
430		- GFP_KERNEL, 0);
431		- else if (blkif->blk_backend_type == BLKIF_BACKEND_FILE) {
432		- /* punch a hole in the backing file */
433		- struct loop_device *lo = bdev->bd_disk->private_data;
434		- struct file *file = lo->lo_backing_file;
435		-
436		- if (file->f_op->fallocate)
437		- err = file->f_op->fallocate(file,
438		- FALLOC_FL_KEEP_SIZE \| FALLOC_FL_PUNCH_HOLE,
439		- req->u.discard.sector_number << 9,
440		- req->u.discard.nr_sectors << 9);
441		- else
442		- err = -EOPNOTSUPP;
	434	+ GFP_KERNEL, secure);
443	435	} else
444	436	err = -EOPNOTSUPP;
445	437
...	...	@@ -449,7 +441,9 @@
449	441	} else if (err)
450	442	status = BLKIF_RSP_ERROR;
451	443
452		- make_response(blkif, req->id, req->operation, status);
	444	+ make_response(blkif, req->u.discard.id, req->operation, status);
	445	+ xen_blkif_put(blkif);
	446	+ return err;
453	447	}
454	448
455	449	static void xen_blk_drain_io(struct xen_blkif *blkif)
...	...	@@ -573,8 +567,11 @@
573	567
574	568	/* Apply all sanity checks to /private copy/ of request. */
575	569	barrier();
576		-
577		- if (dispatch_rw_block_io(blkif, &req, pending_req))
	570	+ if (unlikely(req.operation == BLKIF_OP_DISCARD)) {
	571	+ free_req(pending_req);
	572	+ if (dispatch_discard_io(blkif, &req))
	573	+ break;
	574	+ } else if (dispatch_rw_block_io(blkif, &req, pending_req))
578	575	break;
579	576
580	577	/* Yield point for this unbounded loop. */
...	...	@@ -633,10 +630,6 @@
633	630	blkif->st_f_req++;
634	631	operation = WRITE_FLUSH;
635	632	break;
636		- case BLKIF_OP_DISCARD:
637		- blkif->st_ds_req++;
638		- operation = REQ_DISCARD;
639		- break;
640	633	default:
641	634	operation = 0; /* make gcc happy */
642	635	goto fail_response;
...	...	@@ -644,9 +637,9 @@
644	637	}
645	638
646	639	/* Check that the number of segments is sane. */
647		- nseg = req->nr_segments;
648		- if (unlikely(nseg == 0 && operation != WRITE_FLUSH &&
649		- operation != REQ_DISCARD) \|\|
	640	+ nseg = req->u.rw.nr_segments;
	641	+
	642	+ if (unlikely(nseg == 0 && operation != WRITE_FLUSH) \|\|
650	643	unlikely(nseg > BLKIF_MAX_SEGMENTS_PER_REQUEST)) {
651	644	pr_debug(DRV_PFX "Bad number of segments in request (%d)\n",
652	645	nseg);
653	646
...	...	@@ -654,12 +647,12 @@
654	647	goto fail_response;
655	648	}
656	649
657		- preq.dev = req->handle;
	650	+ preq.dev = req->u.rw.handle;
658	651	preq.sector_number = req->u.rw.sector_number;
659	652	preq.nr_sects = 0;
660	653
661	654	pending_req->blkif = blkif;
662		- pending_req->id = req->id;
	655	+ pending_req->id = req->u.rw.id;
663	656	pending_req->operation = req->operation;
664	657	pending_req->status = BLKIF_RSP_OKAY;
665	658	pending_req->nr_pages = nseg;
...	...	@@ -707,7 +700,7 @@
707	700	* the hypercall to unmap the grants - that is all done in
708	701	* xen_blkbk_unmap.
709	702	*/
710		- if (operation != REQ_DISCARD && xen_blkbk_map(req, pending_req, seg))
	703	+ if (xen_blkbk_map(req, pending_req, seg))
711	704	goto fail_flush;
712	705
713	706	/*
714	707
715	708
...	...	@@ -739,23 +732,16 @@
739	732
740	733	/* This will be hit if the operation was a flush or discard. */
741	734	if (!bio) {
742		- BUG_ON(operation != WRITE_FLUSH && operation != REQ_DISCARD);
	735	+ BUG_ON(operation != WRITE_FLUSH);
743	736
744		- if (operation == WRITE_FLUSH) {
745		- bio = bio_alloc(GFP_KERNEL, 0);
746		- if (unlikely(bio == NULL))
747		- goto fail_put_bio;
	737	+ bio = bio_alloc(GFP_KERNEL, 0);
	738	+ if (unlikely(bio == NULL))
	739	+ goto fail_put_bio;
748	740
749		- biolist[nbio++] = bio;
750		- bio->bi_bdev = preq.bdev;
751		- bio->bi_private = pending_req;
752		- bio->bi_end_io = end_block_io_op;
753		- } else if (operation == REQ_DISCARD) {
754		- xen_blk_discard(blkif, req);
755		- xen_blkif_put(blkif);
756		- free_req(pending_req);
757		- return 0;
758		- }
	741	+ biolist[nbio++] = bio;
	742	+ bio->bi_bdev = preq.bdev;
	743	+ bio->bi_private = pending_req;
	744	+ bio->bi_end_io = end_block_io_op;
759	745	}
760	746
761	747	/*
...	...	@@ -784,7 +770,7 @@
784	770	xen_blkbk_unmap(pending_req);
785	771	fail_response:
786	772	/* Haven't submitted any bio's yet. */
787		- make_response(blkif, req->id, req->operation, BLKIF_RSP_ERROR);
	773	+ make_response(blkif, req->u.rw.id, req->operation, BLKIF_RSP_ERROR);
788	774	free_req(pending_req);
789	775	msleep(1); /* back off a bit */
790	776	return -EIO;
...	...	@@ -60,58 +60,66 @@
60	60	char dummy;
61	61	};
62	62
63		-/* i386 protocol version */
64		-#pragma pack(push, 4)
65		-
66	63	struct blkif_x86_32_request_rw {
	64	+ uint8_t nr_segments; /* number of segments */
	65	+ blkif_vdev_t handle; /* only for read/write requests */
	66	+ uint64_t id; /* private guest value, echoed in resp */
67	67	blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
68	68	struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
69		-};
	69	+} __attribute__((__packed__));
70	70
71	71	struct blkif_x86_32_request_discard {
	72	+ uint8_t flag; /* BLKIF_DISCARD_SECURE or zero */
	73	+ blkif_vdev_t _pad1; /* was "handle" for read/write requests */
	74	+ uint64_t id; /* private guest value, echoed in resp */
72	75	blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
73		- uint64_t nr_sectors;
74		-};
	76	+ uint64_t nr_sectors;
	77	+} __attribute__((__packed__));
75	78
76	79	struct blkif_x86_32_request {
77	80	uint8_t operation; /* BLKIF_OP_??? */
78		- uint8_t nr_segments; /* number of segments */
79		- blkif_vdev_t handle; /* only for read/write requests */
80		- uint64_t id; /* private guest value, echoed in resp */
81	81	union {
82	82	struct blkif_x86_32_request_rw rw;
83	83	struct blkif_x86_32_request_discard discard;
84	84	} u;
85		-};
	85	+} __attribute__((__packed__));
	86	+
	87	+/* i386 protocol version */
	88	+#pragma pack(push, 4)
86	89	struct blkif_x86_32_response {
87	90	uint64_t id; /* copied from request */
88	91	uint8_t operation; /* copied from request */
89	92	int16_t status; /* BLKIF_RSP_??? */
90	93	};
91	94	#pragma pack(pop)
92		-
93	95	/* x86_64 protocol version */
94	96
95	97	struct blkif_x86_64_request_rw {
	98	+ uint8_t nr_segments; /* number of segments */
	99	+ blkif_vdev_t handle; /* only for read/write requests */
	100	+ uint32_t _pad1; /* offsetof(blkif_reqest..,u.rw.id)==8 */
	101	+ uint64_t id;
96	102	blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
97	103	struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
98		-};
	104	+} __attribute__((__packed__));
99	105
100	106	struct blkif_x86_64_request_discard {
	107	+ uint8_t flag; /* BLKIF_DISCARD_SECURE or zero */
	108	+ blkif_vdev_t _pad1; /* was "handle" for read/write requests */
	109	+ uint32_t _pad2; /* offsetof(blkif_..,u.discard.id)==8 */
	110	+ uint64_t id;
101	111	blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
102		- uint64_t nr_sectors;
103		-};
	112	+ uint64_t nr_sectors;
	113	+} __attribute__((__packed__));
104	114
105	115	struct blkif_x86_64_request {
106	116	uint8_t operation; /* BLKIF_OP_??? */
107		- uint8_t nr_segments; /* number of segments */
108		- blkif_vdev_t handle; /* only for read/write requests */
109		- uint64_t __attribute__((__aligned__(8))) id;
110	117	union {
111	118	struct blkif_x86_64_request_rw rw;
112	119	struct blkif_x86_64_request_discard discard;
113	120	} u;
114		-};
	121	+} __attribute__((__packed__));
	122	+
115	123	struct blkif_x86_64_response {
116	124	uint64_t __attribute__((__aligned__(8))) id;
117	125	uint8_t operation; /* copied from request */
...	...	@@ -156,6 +164,7 @@
156	164	/* Cached size parameter. */
157	165	sector_t size;
158	166	bool flush_support;
	167	+ bool discard_secure;
159	168	};
160	169
161	170	struct backend_info;
162	171
163	172
164	173
...	...	@@ -237,22 +246,23 @@
237	246	{
238	247	int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST;
239	248	dst->operation = src->operation;
240		- dst->nr_segments = src->nr_segments;
241		- dst->handle = src->handle;
242		- dst->id = src->id;
243	249	switch (src->operation) {
244	250	case BLKIF_OP_READ:
245	251	case BLKIF_OP_WRITE:
246	252	case BLKIF_OP_WRITE_BARRIER:
247	253	case BLKIF_OP_FLUSH_DISKCACHE:
	254	+ dst->u.rw.nr_segments = src->u.rw.nr_segments;
	255	+ dst->u.rw.handle = src->u.rw.handle;
	256	+ dst->u.rw.id = src->u.rw.id;
248	257	dst->u.rw.sector_number = src->u.rw.sector_number;
249	258	barrier();
250		- if (n > dst->nr_segments)
251		- n = dst->nr_segments;
	259	+ if (n > dst->u.rw.nr_segments)
	260	+ n = dst->u.rw.nr_segments;
252	261	for (i = 0; i < n; i++)
253	262	dst->u.rw.seg[i] = src->u.rw.seg[i];
254	263	break;
255	264	case BLKIF_OP_DISCARD:
	265	+ dst->u.discard.flag = src->u.discard.flag;
256	266	dst->u.discard.sector_number = src->u.discard.sector_number;
257	267	dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
258	268	break;
259	269
260	270
261	271
...	...	@@ -266,22 +276,23 @@
266	276	{
267	277	int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST;
268	278	dst->operation = src->operation;
269		- dst->nr_segments = src->nr_segments;
270		- dst->handle = src->handle;
271		- dst->id = src->id;
272	279	switch (src->operation) {
273	280	case BLKIF_OP_READ:
274	281	case BLKIF_OP_WRITE:
275	282	case BLKIF_OP_WRITE_BARRIER:
276	283	case BLKIF_OP_FLUSH_DISKCACHE:
	284	+ dst->u.rw.nr_segments = src->u.rw.nr_segments;
	285	+ dst->u.rw.handle = src->u.rw.handle;
	286	+ dst->u.rw.id = src->u.rw.id;
277	287	dst->u.rw.sector_number = src->u.rw.sector_number;
278	288	barrier();
279		- if (n > dst->nr_segments)
280		- n = dst->nr_segments;
	289	+ if (n > dst->u.rw.nr_segments)
	290	+ n = dst->u.rw.nr_segments;
281	291	for (i = 0; i < n; i++)
282	292	dst->u.rw.seg[i] = src->u.rw.seg[i];
283	293	break;
284	294	case BLKIF_OP_DISCARD:
	295	+ dst->u.discard.flag = src->u.discard.flag;
285	296	dst->u.discard.sector_number = src->u.discard.sector_number;
286	297	dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
287	298	break;
...	...	@@ -338,6 +338,9 @@
338	338	if (q && q->flush_flags)
339	339	vbd->flush_support = true;
340	340
	341	+ if (q && blk_queue_secdiscard(q))
	342	+ vbd->discard_secure = true;
	343	+
341	344	DPRINTK("Successful creation of handle=%04x (dom=%u)\n",
342	345	handle, blkif->domid);
343	346	return 0;
...	...	@@ -419,6 +422,15 @@
419	422	}
420	423	state = 1;
421	424	blkif->blk_backend_type = BLKIF_BACKEND_PHY;
	425	+ }
	426	+ /* Optional. */
	427	+ err = xenbus_printf(xbt, dev->nodename,
	428	+ "discard-secure", "%d",
	429	+ blkif->vbd.discard_secure);
	430	+ if (err) {
	431	+ xenbus_dev_fatal(dev, err,
	432	+ "writting discard-secure");
	433	+ goto kfree;
422	434	}
423	435	}
424	436	} else {
...	...	@@ -98,7 +98,8 @@
98	98	unsigned long shadow_free;
99	99	unsigned int feature_flush;
100	100	unsigned int flush_op;
101		- unsigned int feature_discard;
	101	+ unsigned int feature_discard:1;
	102	+ unsigned int feature_secdiscard:1;
102	103	unsigned int discard_granularity;
103	104	unsigned int discard_alignment;
104	105	int is_ready;
105	106
...	...	@@ -135,15 +136,15 @@
135	136	{
136	137	unsigned long free = info->shadow_free;
137	138	BUG_ON(free >= BLK_RING_SIZE);
138		- info->shadow_free = info->shadow[free].req.id;
139		- info->shadow[free].req.id = 0x0fffffee; /* debug */
	139	+ info->shadow_free = info->shadow[free].req.u.rw.id;
	140	+ info->shadow[free].req.u.rw.id = 0x0fffffee; /* debug */
140	141	return free;
141	142	}
142	143
143	144	static void add_id_to_freelist(struct blkfront_info *info,
144	145	unsigned long id)
145	146	{
146		- info->shadow[id].req.id = info->shadow_free;
	147	+ info->shadow[id].req.u.rw.id = info->shadow_free;
147	148	info->shadow[id].request = NULL;
148	149	info->shadow_free = id;
149	150	}
...	...	@@ -156,7 +157,7 @@
156	157	if (end > nr_minors) {
157	158	unsigned long bitmap, old;
158	159
159		- bitmap = kzalloc(BITS_TO_LONGS(end) * sizeof(*bitmap),
	160	+ bitmap = kcalloc(BITS_TO_LONGS(end), sizeof(*bitmap),
160	161	GFP_KERNEL);
161	162	if (bitmap == NULL)
162	163	return -ENOMEM;
163	164
...	...	@@ -287,9 +288,9 @@
287	288	id = get_id_from_freelist(info);
288	289	info->shadow[id].request = req;
289	290
290		- ring_req->id = id;
	291	+ ring_req->u.rw.id = id;
291	292	ring_req->u.rw.sector_number = (blkif_sector_t)blk_rq_pos(req);
292		- ring_req->handle = info->handle;
	293	+ ring_req->u.rw.handle = info->handle;
293	294
294	295	ring_req->operation = rq_data_dir(req) ?
295	296	BLKIF_OP_WRITE : BLKIF_OP_READ;
296	297
297	298
298	299
299	300
...	...	@@ -305,16 +306,21 @@
305	306	ring_req->operation = info->flush_op;
306	307	}
307	308
308		- if (unlikely(req->cmd_flags & REQ_DISCARD)) {
	309	+ if (unlikely(req->cmd_flags & (REQ_DISCARD \| REQ_SECURE))) {
309	310	/* id, sector_number and handle are set above. */
310	311	ring_req->operation = BLKIF_OP_DISCARD;
311		- ring_req->nr_segments = 0;
312	312	ring_req->u.discard.nr_sectors = blk_rq_sectors(req);
	313	+ if ((req->cmd_flags & REQ_SECURE) && info->feature_secdiscard)
	314	+ ring_req->u.discard.flag = BLKIF_DISCARD_SECURE;
	315	+ else
	316	+ ring_req->u.discard.flag = 0;
313	317	} else {
314		- ring_req->nr_segments = blk_rq_map_sg(req->q, req, info->sg);
315		- BUG_ON(ring_req->nr_segments > BLKIF_MAX_SEGMENTS_PER_REQUEST);
	318	+ ring_req->u.rw.nr_segments = blk_rq_map_sg(req->q, req,
	319	+ info->sg);
	320	+ BUG_ON(ring_req->u.rw.nr_segments >
	321	+ BLKIF_MAX_SEGMENTS_PER_REQUEST);
316	322
317		- for_each_sg(info->sg, sg, ring_req->nr_segments, i) {
	323	+ for_each_sg(info->sg, sg, ring_req->u.rw.nr_segments, i) {
318	324	buffer_mfn = pfn_to_mfn(page_to_pfn(sg_page(sg)));
319	325	fsect = sg->offset >> 9;
320	326	lsect = fsect + (sg->length >> 9) - 1;
...	...	@@ -424,6 +430,8 @@
424	430	blk_queue_max_discard_sectors(rq, get_capacity(gd));
425	431	rq->limits.discard_granularity = info->discard_granularity;
426	432	rq->limits.discard_alignment = info->discard_alignment;
	433	+ if (info->feature_secdiscard)
	434	+ queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD, rq);
427	435	}
428	436
429	437	/* Hard sector size and max sectors impersonate the equiv. hardware. */
...	...	@@ -705,7 +713,9 @@
705	713	static void blkif_completion(struct blk_shadow *s)
706	714	{
707	715	int i;
708		- for (i = 0; i < s->req.nr_segments; i++)
	716	+ /* Do not let BLKIF_OP_DISCARD as nr_segment is in the same place
	717	+ * flag. */
	718	+ for (i = 0; i < s->req.u.rw.nr_segments; i++)
709	719	gnttab_end_foreign_access(s->req.u.rw.seg[i].gref, 0, 0UL);
710	720	}
711	721
...	...	@@ -736,7 +746,8 @@
736	746	id = bret->id;
737	747	req = info->shadow[id].request;
738	748
739		- blkif_completion(&info->shadow[id]);
	749	+ if (bret->operation != BLKIF_OP_DISCARD)
	750	+ blkif_completion(&info->shadow[id]);
740	751
741	752	add_id_to_freelist(info, id);
742	753
743	754
...	...	@@ -749,7 +760,9 @@
749	760	info->gd->disk_name);
750	761	error = -EOPNOTSUPP;
751	762	info->feature_discard = 0;
	763	+ info->feature_secdiscard = 0;
752	764	queue_flag_clear(QUEUE_FLAG_DISCARD, rq);
	765	+ queue_flag_clear(QUEUE_FLAG_SECDISCARD, rq);
753	766	}
754	767	__blk_end_request_all(req, error);
755	768	break;
...	...	@@ -763,7 +776,7 @@
763	776	error = -EOPNOTSUPP;
764	777	}
765	778	if (unlikely(bret->status == BLKIF_RSP_ERROR &&
766		- info->shadow[id].req.nr_segments == 0)) {
	779	+ info->shadow[id].req.u.rw.nr_segments == 0)) {
767	780	printk(KERN_WARNING "blkfront: %s: empty write %s op failed\n",
768	781	info->flush_op == BLKIF_OP_WRITE_BARRIER ?
769	782	"barrier" : "flush disk cache",
...	...	@@ -984,8 +997,8 @@
984	997	INIT_WORK(&info->work, blkif_restart_queue);
985	998
986	999	for (i = 0; i < BLK_RING_SIZE; i++)
987		- info->shadow[i].req.id = i+1;
988		- info->shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
	1000	+ info->shadow[i].req.u.rw.id = i+1;
	1001	+ info->shadow[BLK_RING_SIZE-1].req.u.rw.id = 0x0fffffff;
989	1002
990	1003	/* Front end dir is a number, which is used as the id. */
991	1004	info->handle = simple_strtoul(strrchr(dev->nodename, '/')+1, NULL, 0);
992	1005
...	...	@@ -1019,9 +1032,9 @@
1019	1032	/* Stage 2: Set up free list. */
1020	1033	memset(&info->shadow, 0, sizeof(info->shadow));
1021	1034	for (i = 0; i < BLK_RING_SIZE; i++)
1022		- info->shadow[i].req.id = i+1;
	1035	+ info->shadow[i].req.u.rw.id = i+1;
1023	1036	info->shadow_free = info->ring.req_prod_pvt;
1024		- info->shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
	1037	+ info->shadow[BLK_RING_SIZE-1].req.u.rw.id = 0x0fffffff;
1025	1038
1026	1039	/* Stage 3: Find pending requests and requeue them. */
1027	1040	for (i = 0; i < BLK_RING_SIZE; i++) {
1028	1041
1029	1042
...	...	@@ -1034,17 +1047,19 @@
1034	1047	*req = copy[i].req;
1035	1048
1036	1049	/* We get a new request id, and must reset the shadow state. */
1037		- req->id = get_id_from_freelist(info);
1038		- memcpy(&info->shadow[req->id], &copy[i], sizeof(copy[i]));
	1050	+ req->u.rw.id = get_id_from_freelist(info);
	1051	+ memcpy(&info->shadow[req->u.rw.id], &copy[i], sizeof(copy[i]));
1039	1052
	1053	+ if (req->operation != BLKIF_OP_DISCARD) {
1040	1054	/* Rewrite any grant references invalidated by susp/resume. */
1041		- for (j = 0; j < req->nr_segments; j++)
1042		- gnttab_grant_foreign_access_ref(
1043		- req->u.rw.seg[j].gref,
1044		- info->xbdev->otherend_id,
1045		- pfn_to_mfn(info->shadow[req->id].frame[j]),
1046		- rq_data_dir(info->shadow[req->id].request));
1047		- info->shadow[req->id].req = *req;
	1055	+ for (j = 0; j < req->u.rw.nr_segments; j++)
	1056	+ gnttab_grant_foreign_access_ref(
	1057	+ req->u.rw.seg[j].gref,
	1058	+ info->xbdev->otherend_id,
	1059	+ pfn_to_mfn(info->shadow[req->u.rw.id].frame[j]),
	1060	+ rq_data_dir(info->shadow[req->u.rw.id].request));
	1061	+ }
	1062	+ info->shadow[req->u.rw.id].req = *req;
1048	1063
1049	1064	info->ring.req_prod_pvt++;
1050	1065	}
1051	1066
...	...	@@ -1135,11 +1150,13 @@
1135	1150	char *type;
1136	1151	unsigned int discard_granularity;
1137	1152	unsigned int discard_alignment;
	1153	+ unsigned int discard_secure;
1138	1154
1139	1155	type = xenbus_read(XBT_NIL, info->xbdev->otherend, "type", NULL);
1140	1156	if (IS_ERR(type))
1141	1157	return;
1142	1158
	1159	+ info->feature_secdiscard = 0;
1143	1160	if (strncmp(type, "phy", 3) == 0) {
1144	1161	err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1145	1162	"discard-granularity", "%u", &discard_granularity,
...	...	@@ -1150,6 +1167,12 @@
1150	1167	info->discard_granularity = discard_granularity;
1151	1168	info->discard_alignment = discard_alignment;
1152	1169	}
	1170	+ err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
	1171	+ "discard-secure", "%d", &discard_secure,
	1172	+ NULL);
	1173	+ if (!err)
	1174	+ info->feature_secdiscard = discard_secure;
	1175	+
1153	1176	} else if (strncmp(type, "file", 4) == 0)
1154	1177	info->feature_discard = 1;
1155	1178
...	...	@@ -84,6 +84,21 @@
84	84	* e07154r6-Data_Set_Management_Proposal_for_ATA-ACS2.doc
85	85	* http://www.seagate.com/staticfiles/support/disc/manuals/
86	86	* Interface%20manuals/100293068c.pdf
	87	+ * The backend can optionally provide three extra XenBus attributes to
	88	+ * further optimize the discard functionality:
	89	+ * 'discard-aligment' - Devices that support discard functionality may
	90	+ * internally allocate space in units that are bigger than the exported
	91	+ * logical block size. The discard-alignment parameter indicates how many bytes
	92	+ * the beginning of the partition is offset from the internal allocation unit's
	93	+ * natural alignment.
	94	+ * 'discard-granularity' - Devices that support discard functionality may
	95	+ * internally allocate space using units that are bigger than the logical block
	96	+ * size. The discard-granularity parameter indicates the size of the internal
	97	+ * allocation unit in bytes if reported by the device. Otherwise the
	98	+ * discard-granularity will be set to match the device's physical block size.
	99	+ * 'discard-secure' - All copies of the discarded sectors (potentially created
	100	+ * by garbage collection) must also be erased. To use this feature, the flag
	101	+ * BLKIF_DISCARD_SECURE must be set in the blkif_request_trim.
87	102	*/
88	103	#define BLKIF_OP_DISCARD 5
89	104
...	...	@@ -95,6 +110,12 @@
95	110	#define BLKIF_MAX_SEGMENTS_PER_REQUEST 11
96	111
97	112	struct blkif_request_rw {
	113	+ uint8_t nr_segments; /* number of segments */
	114	+ blkif_vdev_t handle; /* only for read/write requests */
	115	+#ifdef CONFIG_X86_64
	116	+ uint32_t _pad1; /* offsetof(blkif_request,u.rw.id) == 8 */
	117	+#endif
	118	+ uint64_t id; /* private guest value, echoed in resp */
98	119	blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
99	120	struct blkif_request_segment {
100	121	grant_ref_t gref; /* reference to I/O buffer frame */
101	122
102	123
103	124
104	125
...	...	@@ -102,23 +123,28 @@
102	123	/* @last_sect: last sector in frame to transfer (inclusive). */
103	124	uint8_t first_sect, last_sect;
104	125	} seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
105		-};
	126	+} __attribute__((__packed__));
106	127
107	128	struct blkif_request_discard {
	129	+ uint8_t flag; /* BLKIF_DISCARD_SECURE or zero. */
	130	+#define BLKIF_DISCARD_SECURE (1<<0) /* ignored if discard-secure=0 */
	131	+ blkif_vdev_t _pad1; /* only for read/write requests */
	132	+#ifdef CONFIG_X86_64
	133	+ uint32_t _pad2; /* offsetof(blkif_req..,u.discard.id)==8*/
	134	+#endif
	135	+ uint64_t id; /* private guest value, echoed in resp */
108	136	blkif_sector_t sector_number;
109		- uint64_t nr_sectors;
110		-};
	137	+ uint64_t nr_sectors;
	138	+ uint8_t _pad3;
	139	+} __attribute__((__packed__));
111	140
112	141	struct blkif_request {
113	142	uint8_t operation; /* BLKIF_OP_??? */
114		- uint8_t nr_segments; /* number of segments */
115		- blkif_vdev_t handle; /* only for read/write requests */
116		- uint64_t id; /* private guest value, echoed in resp */
117	143	union {
118	144	struct blkif_request_rw rw;
119	145	struct blkif_request_discard discard;
120	146	} u;
121		-};
	147	+} __attribute__((__packed__));
122	148
123	149	struct blkif_response {
124	150	uint64_t id; /* copied from request */