block: IBM RamSan 70/80 device driver

This patch includes the device driver for the IBM RamSan
family of PCI SSD flash storage cards. This driver will
include support for the RamSan 70 and 80. The driver
presents a block device for device I/O.

Signed-off-by: Philip J Kelleher <pjk1939@linux.vnet.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

9 files changed, 3496 insertions, 0 deletions

diff --git a/drivers/block/rsxx/Makefile b/drivers/block/rsxx/Makefile
new file mode 100644
index 000000000000..f35cd0b71f7b
--- /dev/null
+++ b/drivers/block/rsxx/Makefile
@@ -0,0 +1,2 @@
+obj-$(CONFIG_BLK_DEV_RSXX) += rsxx.o
+rsxx-y := config.o core.o cregs.o dev.o dma.o
diff --git a/drivers/block/rsxx/config.c b/drivers/block/rsxx/config.c
new file mode 100644
index 000000000000..c8829cd4db11
--- /dev/null
+++ b/drivers/block/rsxx/config.c
@@ -0,0 +1,213 @@
+/*
+* Filename: config.c
+*
+*
+* Authors: Joshua Morris <josh.h.morris@us.ibm.com>
+*	Philip Kelleher <pjk1939@linux.vnet.ibm.com>
+*
+* (C) Copyright 2013 IBM Corporation
+*
+* 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.
+*
+* You should have received a copy of the GNU General Public License
+* along with this program; if not, write to the Free Software Foundation,
+* Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+*/
+
+#include <linux/types.h>
+#include <linux/crc32.h>
+#include <linux/swab.h>
+
+#include "rsxx_priv.h"
+#include "rsxx_cfg.h"
+
+static void initialize_config(void *config)
+{
+	struct rsxx_card_cfg *cfg = (struct rsxx_card_cfg *) config;
+
+	cfg->hdr.version = RSXX_CFG_VERSION;
+
+	cfg->data.block_size        = RSXX_HW_BLK_SIZE;
+	cfg->data.stripe_size       = RSXX_HW_BLK_SIZE;
+	cfg->data.vendor_id         = RSXX_VENDOR_ID_TMS_IBM;
+	cfg->data.cache_order       = (-1);
+	cfg->data.intr_coal.mode    = RSXX_INTR_COAL_DISABLED;
+	cfg->data.intr_coal.count   = 0;
+	cfg->data.intr_coal.latency = 0;
+}
+
+static u32 config_data_crc32(struct rsxx_card_cfg *cfg)
+{
+	/*
+	 * Return the compliment of the CRC to ensure compatibility
+	 * (i.e. this is how early rsxx drivers did it.)
+	 */
+
+	return ~crc32(~0, &cfg->data, sizeof(cfg->data));
+}
+
+
+/*----------------- Config Byte Swap Functions -------------------*/
+static void config_hdr_be_to_cpu(struct card_cfg_hdr *hdr)
+{
+	hdr->version = be32_to_cpu((__force __be32) hdr->version);
+	hdr->crc     = be32_to_cpu((__force __be32) hdr->crc);
+}
+
+static void config_hdr_cpu_to_be(struct card_cfg_hdr *hdr)
+{
+	hdr->version = (__force u32) cpu_to_be32(hdr->version);
+	hdr->crc     = (__force u32) cpu_to_be32(hdr->crc);
+}
+
+static void config_data_swab(struct rsxx_card_cfg *cfg)
+{
+	u32 *data = (u32 *) &cfg->data;
+	int i;
+
+	for (i = 0; i < (sizeof(cfg->data) / 4); i++)
+		data[i] = swab32(data[i]);
+}
+
+static void config_data_le_to_cpu(struct rsxx_card_cfg *cfg)
+{
+	u32 *data = (u32 *) &cfg->data;
+	int i;
+
+	for (i = 0; i < (sizeof(cfg->data) / 4); i++)
+		data[i] = le32_to_cpu((__force __le32) data[i]);
+}
+
+static void config_data_cpu_to_le(struct rsxx_card_cfg *cfg)
+{
+	u32 *data = (u32 *) &cfg->data;
+	int i;
+
+	for (i = 0; i < (sizeof(cfg->data) / 4); i++)
+		data[i] = (__force u32) cpu_to_le32(data[i]);
+}
+
+
+/*----------------- Config Operations ------------------*/
+int rsxx_save_config(struct rsxx_cardinfo *card)
+{
+	struct rsxx_card_cfg cfg;
+	int st;
+
+	memcpy(&cfg, &card->config, sizeof(cfg));
+
+	if (unlikely(cfg.hdr.version != RSXX_CFG_VERSION)) {
+		dev_err(CARD_TO_DEV(card),
+			"Cannot save config with invalid version %d\n",
+			cfg.hdr.version);
+		return -EINVAL;
+	}
+
+	/* Convert data to little endian for the CRC calculation. */
+	config_data_cpu_to_le(&cfg);
+
+	cfg.hdr.crc = config_data_crc32(&cfg);
+
+	/*
+	 * Swap the data from little endian to big endian so it can be
+	 * stored.
+	 */
+	config_data_swab(&cfg);
+	config_hdr_cpu_to_be(&cfg.hdr);
+
+	st = rsxx_creg_write(card, CREG_ADD_CONFIG, sizeof(cfg), &cfg, 1);
+	if (st)
+		return st;
+
+	return 0;
+}
+
+int rsxx_load_config(struct rsxx_cardinfo *card)
+{
+	int st;
+	u32 crc;
+
+	st = rsxx_creg_read(card, CREG_ADD_CONFIG, sizeof(card->config),
+				&card->config, 1);
+	if (st) {
+		dev_err(CARD_TO_DEV(card),
+			"Failed reading card config.\n");
+		return st;
+	}
+
+	config_hdr_be_to_cpu(&card->config.hdr);
+
+	if (card->config.hdr.version == RSXX_CFG_VERSION) {
+		/*
+		 * We calculate the CRC with the data in little endian, because
+		 * early drivers did not take big endian CPUs into account.
+		 * The data is always stored in big endian, so we need to byte
+		 * swap it before calculating the CRC.
+		 */
+
+		config_data_swab(&card->config);
+
+		/* Check the CRC */
+		crc = config_data_crc32(&card->config);
+		if (crc != card->config.hdr.crc) {
+			dev_err(CARD_TO_DEV(card),
+				"Config corruption detected!\n");
+			dev_info(CARD_TO_DEV(card),
+				"CRC (sb x%08x is x%08x)\n",
+				card->config.hdr.crc, crc);
+			return -EIO;
+		}
+
+		/* Convert the data to CPU byteorder */
+		config_data_le_to_cpu(&card->config);
+
+	} else if (card->config.hdr.version != 0) {
+		dev_err(CARD_TO_DEV(card),
+			"Invalid config version %d.\n",
+			card->config.hdr.version);
+		/*
+		 * Config version changes require special handling from the
+		 * user
+		 */
+		return -EINVAL;
+	} else {
+		dev_info(CARD_TO_DEV(card),
+			"Initializing card configuration.\n");
+		initialize_config(card);
+		st = rsxx_save_config(card);
+		if (st)
+			return st;
+	}
+
+	card->config_valid = 1;
+
+	dev_dbg(CARD_TO_DEV(card), "version:     x%08x\n",
+		card->config.hdr.version);
+	dev_dbg(CARD_TO_DEV(card), "crc:         x%08x\n",
+		card->config.hdr.crc);
+	dev_dbg(CARD_TO_DEV(card), "block_size:  x%08x\n",
+		card->config.data.block_size);
+	dev_dbg(CARD_TO_DEV(card), "stripe_size: x%08x\n",
+		card->config.data.stripe_size);
+	dev_dbg(CARD_TO_DEV(card), "vendor_id:   x%08x\n",
+		card->config.data.vendor_id);
+	dev_dbg(CARD_TO_DEV(card), "cache_order: x%08x\n",
+		card->config.data.cache_order);
+	dev_dbg(CARD_TO_DEV(card), "mode:        x%08x\n",
+		card->config.data.intr_coal.mode);
+	dev_dbg(CARD_TO_DEV(card), "count:       x%08x\n",
+		card->config.data.intr_coal.count);
+	dev_dbg(CARD_TO_DEV(card), "latency:     x%08x\n",
+		 card->config.data.intr_coal.latency);
+
+	return 0;
+}
+
diff --git a/drivers/block/rsxx/core.c b/drivers/block/rsxx/core.c
new file mode 100644
index 000000000000..f75219140e70
--- /dev/null
+++ b/drivers/block/rsxx/core.c
@@ -0,0 +1,651 @@
+/*
+* Filename: core.c
+*
+*
+* Authors: Joshua Morris <josh.h.morris@us.ibm.com>
+*	Philip Kelleher <pjk1939@linux.vnet.ibm.com>
+*
+* (C) Copyright 2013 IBM Corporation
+*
+* 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.
+*
+* You should have received a copy of the GNU General Public License
+* along with this program; if not, write to the Free Software Foundation,
+* Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+*/
+
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+#include <linux/reboot.h>
+#include <linux/slab.h>
+#include <linux/bitops.h>
+
+#include <linux/genhd.h>
+#include <linux/idr.h>
+
+#include "rsxx_priv.h"
+#include "rsxx_cfg.h"
+
+#define NO_LEGACY 0
+
+MODULE_DESCRIPTION("IBM RamSan PCIe Flash SSD Device Driver");
+MODULE_AUTHOR("IBM <support@ramsan.com>");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(DRIVER_VERSION);
+
+static unsigned int force_legacy = NO_LEGACY;
+module_param(force_legacy, uint, 0444);
+MODULE_PARM_DESC(force_legacy, "Force the use of legacy type PCI interrupts");
+
+static DEFINE_IDA(rsxx_disk_ida);
+static DEFINE_SPINLOCK(rsxx_ida_lock);
+
+/*----------------- Interrupt Control & Handling -------------------*/
+static void __enable_intr(unsigned int *mask, unsigned int intr)
+{
+	*mask |= intr;
+}
+
+static void __disable_intr(unsigned int *mask, unsigned int intr)
+{
+	*mask &= ~intr;
+}
+
+/*
+ * NOTE: Disabling the IER will disable the hardware interrupt.
+ * Disabling the ISR will disable the software handling of the ISR bit.
+ *
+ * Enable/Disable interrupt functions assume the card->irq_lock
+ * is held by the caller.
+ */
+void rsxx_enable_ier(struct rsxx_cardinfo *card, unsigned int intr)
+{
+	if (unlikely(card->halt))
+		return;
+
+	__enable_intr(&card->ier_mask, intr);
+	iowrite32(card->ier_mask, card->regmap + IER);
+}
+
+void rsxx_disable_ier(struct rsxx_cardinfo *card, unsigned int intr)
+{
+	__disable_intr(&card->ier_mask, intr);
+	iowrite32(card->ier_mask, card->regmap + IER);
+}
+
+void rsxx_enable_ier_and_isr(struct rsxx_cardinfo *card,
+				 unsigned int intr)
+{
+	if (unlikely(card->halt))
+		return;
+
+	__enable_intr(&card->isr_mask, intr);
+	__enable_intr(&card->ier_mask, intr);
+	iowrite32(card->ier_mask, card->regmap + IER);
+}
+void rsxx_disable_ier_and_isr(struct rsxx_cardinfo *card,
+				  unsigned int intr)
+{
+	__disable_intr(&card->isr_mask, intr);
+	__disable_intr(&card->ier_mask, intr);
+	iowrite32(card->ier_mask, card->regmap + IER);
+}
+
+irqreturn_t rsxx_isr(int irq, void *pdata)
+{
+	struct rsxx_cardinfo *card = (struct rsxx_cardinfo *) pdata;
+	unsigned int isr;
+	int handled = 0;
+	int reread_isr;
+	int i;
+
+	spin_lock(&card->irq_lock);
+
+	do {
+		reread_isr = 0;
+
+		isr = ioread32(card->regmap + ISR);
+		if (isr == 0xffffffff) {
+			/*
+			 * A few systems seem to have an intermittent issue
+			 * where PCI reads return all Fs, but retrying the read
+			 * a little later will return as expected.
+			 */
+			dev_info(CARD_TO_DEV(card),
+				"ISR = 0xFFFFFFFF, retrying later\n");
+			break;
+		}
+
+		isr &= card->isr_mask;
+		if (!isr)
+			break;
+
+		for (i = 0; i < card->n_targets; i++) {
+			if (isr & CR_INTR_DMA(i)) {
+				if (card->ier_mask & CR_INTR_DMA(i)) {
+					rsxx_disable_ier(card, CR_INTR_DMA(i));
+					reread_isr = 1;
+				}
+				queue_work(card->ctrl[i].done_wq,
+					   &card->ctrl[i].dma_done_work);
+				handled++;
+			}
+		}
+
+		if (isr & CR_INTR_CREG) {
+			schedule_work(&card->creg_ctrl.done_work);
+			handled++;
+		}
+
+		if (isr & CR_INTR_EVENT) {
+			schedule_work(&card->event_work);
+			rsxx_disable_ier_and_isr(card, CR_INTR_EVENT);
+			handled++;
+		}
+	} while (reread_isr);
+
+	spin_unlock(&card->irq_lock);
+
+	return handled ? IRQ_HANDLED : IRQ_NONE;
+}
+
+/*----------------- Card Event Handler -------------------*/
+static void card_state_change(struct rsxx_cardinfo *card,
+			      unsigned int new_state)
+{
+	int st;
+
+	dev_info(CARD_TO_DEV(card),
+		"card state change detected.(%s -> %s)\n",
+		rsxx_card_state_to_str(card->state),
+		rsxx_card_state_to_str(new_state));
+
+	card->state = new_state;
+
+	/* Don't attach DMA interfaces if the card has an invalid config */
+	if (!card->config_valid)
+		return;
+
+	switch (new_state) {
+	case CARD_STATE_RD_ONLY_FAULT:
+		dev_crit(CARD_TO_DEV(card),
+			"Hardware has entered read-only mode!\n");
+		/*
+		 * Fall through so the DMA devices can be attached and
+		 * the user can attempt to pull off their data.
+		 */
+	case CARD_STATE_GOOD:
+		st = rsxx_get_card_size8(card, &card->size8);
+		if (st)
+			dev_err(CARD_TO_DEV(card),
+				"Failed attaching DMA devices\n");
+
+		if (card->config_valid)
+			set_capacity(card->gendisk, card->size8 >> 9);
+		break;
+
+	case CARD_STATE_FAULT:
+		dev_crit(CARD_TO_DEV(card),
+			"Hardware Fault reported!\n");
+		/* Fall through. */
+
+	/* Everything else, detach DMA interface if it's attached. */
+	case CARD_STATE_SHUTDOWN:
+	case CARD_STATE_STARTING:
+	case CARD_STATE_FORMATTING:
+	case CARD_STATE_UNINITIALIZED:
+	case CARD_STATE_SHUTTING_DOWN:
+	/*
+	 * dStroy is a term coined by marketing to represent the low level
+	 * secure erase.
+	 */
+	case CARD_STATE_DSTROYING:
+		set_capacity(card->gendisk, 0);
+		break;
+	}
+}
+
+static void card_event_handler(struct work_struct *work)
+{
+	struct rsxx_cardinfo *card;
+	unsigned int state;
+	unsigned long flags;
+	int st;
+
+	card = container_of(work, struct rsxx_cardinfo, event_work);
+
+	if (unlikely(card->halt))
+		return;
+
+	/*
+	 * Enable the interrupt now to avoid any weird race conditions where a
+	 * state change might occur while rsxx_get_card_state() is
+	 * processing a returned creg cmd.
+	 */
+	spin_lock_irqsave(&card->irq_lock, flags);
+	rsxx_enable_ier_and_isr(card, CR_INTR_EVENT);
+	spin_unlock_irqrestore(&card->irq_lock, flags);
+
+	st = rsxx_get_card_state(card, &state);
+	if (st) {
+		dev_info(CARD_TO_DEV(card),
+			"Failed reading state after event.\n");
+		return;
+	}
+
+	if (card->state != state)
+		card_state_change(card, state);
+
+	if (card->creg_ctrl.creg_stats.stat & CREG_STAT_LOG_PENDING)
+		rsxx_read_hw_log(card);
+}
+
+
+char *rsxx_card_state_to_str(unsigned int state)
+{
+	static char *state_strings[] = {
+		"Unknown", "Shutdown", "Starting", "Formatting",
+		"Uninitialized", "Good", "Shutting Down",
+		"Fault", "Read Only Fault", "dStroying"
+	};
+
+	return state_strings[ffs(state)];
+}
+
+/*----------------- Card Operations -------------------*/
+static int card_shutdown(struct rsxx_cardinfo *card)
+{
+	unsigned int state;
+	signed long start;
+	const int timeout = msecs_to_jiffies(120000);
+	int st;
+
+	/* We can't issue a shutdown if the card is in a transition state */
+	start = jiffies;
+	do {
+		st = rsxx_get_card_state(card, &state);
+		if (st)
+			return st;
+	} while (state == CARD_STATE_STARTING &&
+		 (jiffies - start < timeout));
+
+	if (state == CARD_STATE_STARTING)
+		return -ETIMEDOUT;
+
+	/* Only issue a shutdown if we need to */
+	if ((state != CARD_STATE_SHUTTING_DOWN) &&
+	    (state != CARD_STATE_SHUTDOWN)) {
+		st = rsxx_issue_card_cmd(card, CARD_CMD_SHUTDOWN);
+		if (st)
+			return st;
+	}
+
+	start = jiffies;
+	do {
+		st = rsxx_get_card_state(card, &state);
+		if (st)
+			return st;
+	} while (state != CARD_STATE_SHUTDOWN &&
+		 (jiffies - start < timeout));
+
+	if (state != CARD_STATE_SHUTDOWN)
+		return -ETIMEDOUT;
+
+	return 0;
+}
+
+/*----------------- Driver Initialization & Setup -------------------*/
+/* Returns:   0 if the driver is compatible with the device
+	     -1 if the driver is NOT compatible with the device */
+static int rsxx_compatibility_check(struct rsxx_cardinfo *card)
+{
+	unsigned char pci_rev;
+
+	pci_read_config_byte(card->dev, PCI_REVISION_ID, &pci_rev);
+
+	if (pci_rev > RS70_PCI_REV_SUPPORTED)
+		return -1;
+	return 0;
+}
+
+static int __devinit rsxx_pci_probe(struct pci_dev *dev,
+					const struct pci_device_id *id)
+{
+	struct rsxx_cardinfo *card;
+	unsigned long flags;
+	int st;
+
+	dev_info(&dev->dev, "PCI-Flash SSD discovered\n");
+
+	card = kzalloc(sizeof(*card), GFP_KERNEL);
+	if (!card)
+		return -ENOMEM;
+
+	card->dev = dev;
+	pci_set_drvdata(dev, card);
+
+	do {
+		if (!ida_pre_get(&rsxx_disk_ida, GFP_KERNEL)) {
+			st = -ENOMEM;
+			goto failed_ida_get;
+		}
+
+		spin_lock(&rsxx_ida_lock);
+		st = ida_get_new(&rsxx_disk_ida, &card->disk_id);
+		spin_unlock(&rsxx_ida_lock);
+	} while (st == -EAGAIN);
+
+	if (st)
+		goto failed_ida_get;
+
+	st = pci_enable_device(dev);
+	if (st)
+		goto failed_enable;
+
+	pci_set_master(dev);
+	pci_set_dma_max_seg_size(dev, RSXX_HW_BLK_SIZE);
+
+	st = pci_set_dma_mask(dev, DMA_BIT_MASK(64));
+	if (st) {
+		dev_err(CARD_TO_DEV(card),
+			"No usable DMA configuration,aborting\n");
+		goto failed_dma_mask;
+	}
+
+	st = pci_request_regions(dev, DRIVER_NAME);
+	if (st) {
+		dev_err(CARD_TO_DEV(card),
+			"Failed to request memory region\n");
+		goto failed_request_regions;
+	}
+
+	if (pci_resource_len(dev, 0) == 0) {
+		dev_err(CARD_TO_DEV(card), "BAR0 has length 0!\n");
+		st = -ENOMEM;
+		goto failed_iomap;
+	}
+
+	card->regmap = pci_iomap(dev, 0, 0);
+	if (!card->regmap) {
+		dev_err(CARD_TO_DEV(card), "Failed to map BAR0\n");
+		st = -ENOMEM;
+		goto failed_iomap;
+	}
+
+	spin_lock_init(&card->irq_lock);
+	card->halt = 0;
+
+	spin_lock_irqsave(&card->irq_lock, flags);
+	rsxx_disable_ier_and_isr(card, CR_INTR_ALL);
+	spin_unlock_irqrestore(&card->irq_lock, flags);
+
+	if (!force_legacy) {
+		st = pci_enable_msi(dev);
+		if (st)
+			dev_warn(CARD_TO_DEV(card),
+				"Failed to enable MSI\n");
+	}
+
+	st = request_irq(dev->irq, rsxx_isr, IRQF_DISABLED | IRQF_SHARED,
+			 DRIVER_NAME, card);
+	if (st) {
+		dev_err(CARD_TO_DEV(card),
+			"Failed requesting IRQ%d\n", dev->irq);
+		goto failed_irq;
+	}
+
+	/************* Setup Processor Command Interface *************/
+	rsxx_creg_setup(card);
+
+	spin_lock_irqsave(&card->irq_lock, flags);
+	rsxx_enable_ier_and_isr(card, CR_INTR_CREG);
+	spin_unlock_irqrestore(&card->irq_lock, flags);
+
+	st = rsxx_compatibility_check(card);
+	if (st) {
+		dev_warn(CARD_TO_DEV(card),
+			"Incompatible driver detected. Please update the driver.\n");
+		st = -EINVAL;
+		goto failed_compatiblity_check;
+	}
+
+	/************* Load Card Config *************/
+	st = rsxx_load_config(card);
+	if (st)
+		dev_err(CARD_TO_DEV(card),
+			"Failed loading card config\n");
+
+	/************* Setup DMA Engine *************/
+	st = rsxx_get_num_targets(card, &card->n_targets);
+	if (st)
+		dev_info(CARD_TO_DEV(card),
+			"Failed reading the number of DMA targets\n");
+
+	card->ctrl = kzalloc(card->n_targets * sizeof(*card->ctrl), GFP_KERNEL);
+	if (!card->ctrl) {
+		st = -ENOMEM;
+		goto failed_dma_setup;
+	}
+
+	st = rsxx_dma_setup(card);
+	if (st) {
+		dev_info(CARD_TO_DEV(card),
+			"Failed to setup DMA engine\n");
+		goto failed_dma_setup;
+	}
+
+	/************* Setup Card Event Handler *************/
+	INIT_WORK(&card->event_work, card_event_handler);
+
+	st = rsxx_setup_dev(card);
+	if (st)
+		goto failed_create_dev;
+
+	rsxx_get_card_state(card, &card->state);
+
+	dev_info(CARD_TO_DEV(card),
+		"card state: %s\n",
+		rsxx_card_state_to_str(card->state));
+
+	/*
+	 * Now that the DMA Engine and devices have been setup,
+	 * we can enable the event interrupt(it kicks off actions in
+	 * those layers so we couldn't enable it right away.)
+	 */
+	spin_lock_irqsave(&card->irq_lock, flags);
+	rsxx_enable_ier_and_isr(card, CR_INTR_EVENT);
+	spin_unlock_irqrestore(&card->irq_lock, flags);
+
+	if (card->state == CARD_STATE_SHUTDOWN) {
+		st = rsxx_issue_card_cmd(card, CARD_CMD_STARTUP);
+		if (st)
+			dev_crit(CARD_TO_DEV(card),
+				"Failed issuing card startup\n");
+	} else if (card->state == CARD_STATE_GOOD ||
+		   card->state == CARD_STATE_RD_ONLY_FAULT) {
+		st = rsxx_get_card_size8(card, &card->size8);
+		if (st)
+			card->size8 = 0;
+	}
+
+	rsxx_attach_dev(card);
+
+	return 0;
+
+failed_create_dev:
+	rsxx_dma_destroy(card);
+failed_dma_setup:
+failed_compatiblity_check:
+	spin_lock_irqsave(&card->irq_lock, flags);
+	rsxx_disable_ier_and_isr(card, CR_INTR_ALL);
+	spin_unlock_irqrestore(&card->irq_lock, flags);
+	free_irq(dev->irq, card);
+	if (!force_legacy)
+		pci_disable_msi(dev);
+failed_irq:
+	pci_iounmap(dev, card->regmap);
+failed_iomap:
+	pci_release_regions(dev);
+failed_request_regions:
+failed_dma_mask:
+	pci_disable_device(dev);
+failed_enable:
+	spin_lock(&rsxx_ida_lock);
+	ida_remove(&rsxx_disk_ida, card->disk_id);
+	spin_unlock(&rsxx_ida_lock);
+failed_ida_get:
+	kfree(card);
+
+	return st;
+}
+
+static void __devexit rsxx_pci_remove(struct pci_dev *dev)
+{
+	struct rsxx_cardinfo *card = pci_get_drvdata(dev);
+	unsigned long flags;
+	int st;
+	int i;
+
+	if (!card)
+		return;
+
+	dev_info(CARD_TO_DEV(card),
+		"Removing PCI-Flash SSD.\n");
+
+	rsxx_detach_dev(card);
+
+	for (i = 0; i < card->n_targets; i++) {
+		spin_lock_irqsave(&card->irq_lock, flags);
+		rsxx_disable_ier_and_isr(card, CR_INTR_DMA(i));
+		spin_unlock_irqrestore(&card->irq_lock, flags);
+	}
+
+	st = card_shutdown(card);
+	if (st)
+		dev_crit(CARD_TO_DEV(card), "Shutdown failed!\n");
+
+	/* Sync outstanding event handlers. */
+	spin_lock_irqsave(&card->irq_lock, flags);
+	rsxx_disable_ier_and_isr(card, CR_INTR_EVENT);
+	spin_unlock_irqrestore(&card->irq_lock, flags);
+
+	/* Prevent work_structs from re-queuing themselves. */
+	card->halt = 1;
+
+	cancel_work_sync(&card->event_work);
+
+	rsxx_destroy_dev(card);
+	rsxx_dma_destroy(card);
+
+	spin_lock_irqsave(&card->irq_lock, flags);
+	rsxx_disable_ier_and_isr(card, CR_INTR_ALL);
+	spin_unlock_irqrestore(&card->irq_lock, flags);
+	free_irq(dev->irq, card);
+
+	if (!force_legacy)
+		pci_disable_msi(dev);
+
+	rsxx_creg_destroy(card);
+
+	pci_iounmap(dev, card->regmap);
+
+	pci_disable_device(dev);
+	pci_release_regions(dev);
+
+	kfree(card);
+}
+
+static int rsxx_pci_suspend(struct pci_dev *dev, pm_message_t state)
+{
+	/* We don't support suspend at this time. */
+	return -ENOSYS;
+}
+
+static void rsxx_pci_shutdown(struct pci_dev *dev)
+{
+	struct rsxx_cardinfo *card = pci_get_drvdata(dev);
+	unsigned long flags;
+	int i;
+
+	if (!card)
+		return;
+
+	dev_info(CARD_TO_DEV(card), "Shutting down PCI-Flash SSD.\n");
+
+	rsxx_detach_dev(card);
+
+	for (i = 0; i < card->n_targets; i++) {
+		spin_lock_irqsave(&card->irq_lock, flags);
+		rsxx_disable_ier_and_isr(card, CR_INTR_DMA(i));
+		spin_unlock_irqrestore(&card->irq_lock, flags);
+	}
+
+	card_shutdown(card);
+}
+
+static DEFINE_PCI_DEVICE_TABLE(rsxx_pci_ids) = {
+	{PCI_DEVICE(PCI_VENDOR_ID_TMS_IBM, PCI_DEVICE_ID_RS70_FLASH)},
+	{PCI_DEVICE(PCI_VENDOR_ID_TMS_IBM, PCI_DEVICE_ID_RS70D_FLASH)},
+	{PCI_DEVICE(PCI_VENDOR_ID_TMS_IBM, PCI_DEVICE_ID_RS80_FLASH)},
+	{PCI_DEVICE(PCI_VENDOR_ID_TMS_IBM, PCI_DEVICE_ID_RS81_FLASH)},
+	{0,},
+};
+
+MODULE_DEVICE_TABLE(pci, rsxx_pci_ids);
+
+static struct pci_driver rsxx_pci_driver = {
+	.name		= DRIVER_NAME,
+	.id_table	= rsxx_pci_ids,
+	.probe		= rsxx_pci_probe,
+	.remove		= __devexit_p(rsxx_pci_remove),
+	.suspend	= rsxx_pci_suspend,
+	.shutdown	= rsxx_pci_shutdown,
+};
+
+static int __init rsxx_core_init(void)
+{
+	int st;
+
+	st = rsxx_dev_init();
+	if (st)
+		return st;
+
+	st = rsxx_dma_init();
+	if (st)
+		goto dma_init_failed;
+
+	st = rsxx_creg_init();
+	if (st)
+		goto creg_init_failed;
+
+	return pci_register_driver(&rsxx_pci_driver);
+
+creg_init_failed:
+	rsxx_dma_cleanup();
+dma_init_failed:
+	rsxx_dev_cleanup();
+
+	return st;
+}
+
+static void __exit rsxx_core_cleanup(void)
+{
+	pci_unregister_driver(&rsxx_pci_driver);
+	rsxx_creg_cleanup();
+	rsxx_dma_cleanup();
+	rsxx_dev_cleanup();
+}
+
+module_init(rsxx_core_init);
+module_exit(rsxx_core_cleanup);
diff --git a/drivers/block/rsxx/cregs.c b/drivers/block/rsxx/cregs.c
new file mode 100644
index 000000000000..a31fd727e804
--- /dev/null
+++ b/drivers/block/rsxx/cregs.c
@@ -0,0 +1,743 @@
+/*
+* Filename: cregs.c
+*
+*
+* Authors: Joshua Morris <josh.h.morris@us.ibm.com>
+*	Philip Kelleher <pjk1939@linux.vnet.ibm.com>
+*
+* (C) Copyright 2013 IBM Corporation
+*
+* 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.
+*
+* You should have received a copy of the GNU General Public License
+* along with this program; if not, write to the Free Software Foundation,
+* Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+*/
+
+#include <linux/completion.h>
+#include <linux/slab.h>
+
+#include "rsxx_priv.h"
+
+#define CREG_TIMEOUT_MSEC	10000
+
+typedef void (*creg_cmd_cb)(struct rsxx_cardinfo *card,
+			    struct creg_cmd *cmd,
+			    int st);
+
+struct creg_cmd {
+	struct list_head list;
+	creg_cmd_cb cb;
+	void *cb_private;
+	unsigned int op;
+	unsigned int addr;
+	int cnt8;
+	void *buf;
+	unsigned int stream;
+	unsigned int status;
+};
+
+static struct kmem_cache *creg_cmd_pool;
+
+
+/*------------ Private Functions --------------*/
+
+#if defined(__LITTLE_ENDIAN)
+#define LITTLE_ENDIAN 1
+#elif defined(__BIG_ENDIAN)
+#define LITTLE_ENDIAN 0
+#else
+#error Unknown endianess!!! Aborting...
+#endif
+
+static void copy_to_creg_data(struct rsxx_cardinfo *card,
+			      int cnt8,
+			      void *buf,
+			      unsigned int stream)
+{
+	int i = 0;
+	u32 *data = buf;
+
+	for (i = 0; cnt8 > 0; i++, cnt8 -= 4) {
+		/*
+		 * Firmware implementation makes it necessary to byte swap on
+		 * little endian processors.
+		 */
+		if (LITTLE_ENDIAN && stream)
+			iowrite32be(data[i], card->regmap + CREG_DATA(i));
+		else
+			iowrite32(data[i], card->regmap + CREG_DATA(i));
+	}
+}
+
+
+static void copy_from_creg_data(struct rsxx_cardinfo *card,
+				int cnt8,
+				void *buf,
+				unsigned int stream)
+{
+	int i = 0;
+	u32 *data = buf;
+
+	for (i = 0; cnt8 > 0; i++, cnt8 -= 4) {
+		/*
+		 * Firmware implementation makes it necessary to byte swap on
+		 * little endian processors.
+		 */
+		if (LITTLE_ENDIAN && stream)
+			data[i] = ioread32be(card->regmap + CREG_DATA(i));
+		else
+			data[i] = ioread32(card->regmap + CREG_DATA(i));
+	}
+}
+
+static struct creg_cmd *pop_active_cmd(struct rsxx_cardinfo *card)
+{
+	struct creg_cmd *cmd;
+
+	/*
+	 * Spin lock is needed because this can be called in atomic/interrupt
+	 * context.
+	 */
+	spin_lock_bh(&card->creg_ctrl.pop_lock);
+	cmd = card->creg_ctrl.active_cmd;
+	card->creg_ctrl.active_cmd = NULL;
+	spin_unlock_bh(&card->creg_ctrl.pop_lock);
+
+	return cmd;
+}
+
+static void creg_issue_cmd(struct rsxx_cardinfo *card, struct creg_cmd *cmd)
+{
+	iowrite32(cmd->addr, card->regmap + CREG_ADD);
+	iowrite32(cmd->cnt8, card->regmap + CREG_CNT);
+
+	if (cmd->op == CREG_OP_WRITE) {
+		if (cmd->buf)
+			copy_to_creg_data(card, cmd->cnt8,
+					  cmd->buf, cmd->stream);
+	}
+
+	/* Data copy must complete before initiating the command. */
+	wmb();
+
+	/* Setting the valid bit will kick off the command. */
+	iowrite32(cmd->op, card->regmap + CREG_CMD);
+}
+
+static void creg_kick_queue(struct rsxx_cardinfo *card)
+{
+	if (card->creg_ctrl.active || list_empty(&card->creg_ctrl.queue))
+		return;
+
+	card->creg_ctrl.active = 1;
+	card->creg_ctrl.active_cmd = list_first_entry(&card->creg_ctrl.queue,
+						      struct creg_cmd, list);
+	list_del(&card->creg_ctrl.active_cmd->list);
+	card->creg_ctrl.q_depth--;
+
+	/*
+	 * We have to set the timer before we push the new command. Otherwise,
+	 * we could create a race condition that would occur if the timer
+	 * was not canceled, and expired after the new command was pushed,
+	 * but before the command was issued to hardware.
+	 */
+	mod_timer(&card->creg_ctrl.cmd_timer,
+				jiffies + msecs_to_jiffies(CREG_TIMEOUT_MSEC));
+
+	creg_issue_cmd(card, card->creg_ctrl.active_cmd);
+}
+
+static int creg_queue_cmd(struct rsxx_cardinfo *card,
+			  unsigned int op,
+			  unsigned int addr,
+			  unsigned int cnt8,
+			  void *buf,
+			  int stream,
+			  creg_cmd_cb callback,
+			  void *cb_private)
+{
+	struct creg_cmd *cmd;
+
+	/* Don't queue stuff up if we're halted. */
+	if (unlikely(card->halt))
+		return -EINVAL;
+
+	if (card->creg_ctrl.reset)
+		return -EAGAIN;
+
+	if (cnt8 > MAX_CREG_DATA8)
+		return -EINVAL;
+
+	cmd = kmem_cache_alloc(creg_cmd_pool, GFP_KERNEL);
+	if (!cmd)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&cmd->list);
+
+	cmd->op		= op;
+	cmd->addr	= addr;
+	cmd->cnt8	= cnt8;
+	cmd->buf	= buf;
+	cmd->stream	= stream;
+	cmd->cb		= callback;
+	cmd->cb_private = cb_private;
+	cmd->status	= 0;
+
+	mutex_lock(&card->creg_ctrl.lock);
+	list_add_tail(&cmd->list, &card->creg_ctrl.queue);
+	card->creg_ctrl.q_depth++;
+	creg_kick_queue(card);
+	mutex_unlock(&card->creg_ctrl.lock);
+
+	return 0;
+}
+
+static void creg_cmd_timed_out(unsigned long data)
+{
+	struct rsxx_cardinfo *card = (struct rsxx_cardinfo *) data;
+	struct creg_cmd *cmd;
+
+	cmd = pop_active_cmd(card);
+	if (cmd == NULL) {
+		card->creg_ctrl.creg_stats.creg_timeout++;
+		dev_warn(CARD_TO_DEV(card),
+			"No active command associated with timeout!\n");
+		return;
+	}
+
+	if (cmd->cb)
+		cmd->cb(card, cmd, -ETIMEDOUT);
+
+	kmem_cache_free(creg_cmd_pool, cmd);
+
+	spin_lock(&card->creg_ctrl.pop_lock);
+	card->creg_ctrl.active = 0;
+	creg_kick_queue(card);
+	spin_unlock(&card->creg_ctrl.pop_lock);
+}
+
+
+static void creg_cmd_done(struct work_struct *work)
+{
+	struct rsxx_cardinfo *card;
+	struct creg_cmd *cmd;
+	int st = 0;
+
+	card = container_of(work, struct rsxx_cardinfo,
+			    creg_ctrl.done_work);
+
+	/*
+	 * The timer could not be cancelled for some reason,
+	 * race to pop the active command.
+	 */
+	if (del_timer_sync(&card->creg_ctrl.cmd_timer) == 0)
+		card->creg_ctrl.creg_stats.failed_cancel_timer++;
+
+	cmd = pop_active_cmd(card);
+	if (cmd == NULL) {
+		dev_err(CARD_TO_DEV(card),
+			"Spurious creg interru