[SCSI] hpsa: add driver for HP Smart Array controllers.
Stephen M. Cameron [Tue, 8 Dec 2009 22:09:11 +0000 (14:09 -0800)]
This driver supports a subset of HP Smart Array Controllers.
It is a SCSI alternative to the cciss driver.

[akpm@linux-foundation.org: avoid helpful cleanup patches]
[achiang@hp.com: make device attrs static]
[akpm@linux-foundation.org: msleep() does set_current_state() itself]
Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: Mike Miller <mikem@beardog.cce.hp.com>
Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

drivers/scsi/Kconfig
drivers/scsi/Makefile
drivers/scsi/hpsa.c [new file with mode: 0644]
drivers/scsi/hpsa.h [new file with mode: 0644]
drivers/scsi/hpsa_cmd.h [new file with mode: 0644]

index 36900c7..9191d1e 100644 (file)
@@ -388,6 +388,16 @@ config BLK_DEV_3W_XXXX_RAID
          Please read the comments at the top of
          <file:drivers/scsi/3w-xxxx.c>.
 
+config SCSI_HPSA
+       tristate "HP Smart Array SCSI driver"
+       depends on PCI && SCSI
+       help
+         This driver supports HP Smart Array Controllers (circa 2009).
+         It is a SCSI alternative to the cciss driver, which is a block
+         driver.  Anyone wishing to use HP Smart Array controllers who
+         would prefer the devices be presented to linux as SCSI devices,
+         rather than as generic block devices should say Y here.
+
 config SCSI_3W_9XXX
        tristate "3ware 9xxx SATA-RAID support"
        depends on PCI && SCSI
index 280d3c6..92a8c50 100644 (file)
@@ -91,6 +91,7 @@ obj-$(CONFIG_SCSI_BFA_FC)     += bfa/
 obj-$(CONFIG_SCSI_PAS16)       += pas16.o
 obj-$(CONFIG_SCSI_T128)                += t128.o
 obj-$(CONFIG_SCSI_DMX3191D)    += dmx3191d.o
+obj-$(CONFIG_SCSI_HPSA)                += hpsa.o
 obj-$(CONFIG_SCSI_DTC3280)     += dtc.o
 obj-$(CONFIG_SCSI_SYM53C8XX_2) += sym53c8xx_2/
 obj-$(CONFIG_SCSI_ZALON)       += zalon7xx.o
diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c
new file mode 100644 (file)
index 0000000..bb96fdd
--- /dev/null
@@ -0,0 +1,3531 @@
+/*
+ *    Disk Array driver for HP Smart Array SAS controllers
+ *    Copyright 2000, 2009 Hewlett-Packard Development Company, L.P.
+ *
+ *    This program is free software; you can redistribute it and/or modify
+ *    it under the terms of the GNU General Public License as published by
+ *    the Free Software Foundation; version 2 of the License.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *    NON INFRINGEMENT.  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., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ *    Questions/Comments/Bugfixes to iss_storagedev@hp.com
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/interrupt.h>
+#include <linux/types.h>
+#include <linux/pci.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/fs.h>
+#include <linux/timer.h>
+#include <linux/seq_file.h>
+#include <linux/init.h>
+#include <linux/spinlock.h>
+#include <linux/smp_lock.h>
+#include <linux/compat.h>
+#include <linux/blktrace_api.h>
+#include <linux/uaccess.h>
+#include <linux/io.h>
+#include <linux/dma-mapping.h>
+#include <linux/completion.h>
+#include <linux/moduleparam.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_device.h>
+#include <scsi/scsi_host.h>
+#include <linux/cciss_ioctl.h>
+#include <linux/string.h>
+#include <linux/bitmap.h>
+#include <asm/atomic.h>
+#include <linux/kthread.h>
+#include "hpsa_cmd.h"
+#include "hpsa.h"
+
+/* HPSA_DRIVER_VERSION must be 3 byte values (0-255) separated by '.' */
+#define HPSA_DRIVER_VERSION "1.0.0"
+#define DRIVER_NAME "HP HPSA Driver (v " HPSA_DRIVER_VERSION ")"
+
+/* How long to wait (in milliseconds) for board to go into simple mode */
+#define MAX_CONFIG_WAIT 30000
+#define MAX_IOCTL_CONFIG_WAIT 1000
+
+/*define how many times we will try a command because of bus resets */
+#define MAX_CMD_RETRIES 3
+
+/* Embedded module documentation macros - see modules.h */
+MODULE_AUTHOR("Hewlett-Packard Company");
+MODULE_DESCRIPTION("Driver for HP Smart Array Controller version " \
+       HPSA_DRIVER_VERSION);
+MODULE_SUPPORTED_DEVICE("HP Smart Array Controllers");
+MODULE_VERSION(HPSA_DRIVER_VERSION);
+MODULE_LICENSE("GPL");
+
+static int hpsa_allow_any;
+module_param(hpsa_allow_any, int, S_IRUGO|S_IWUSR);
+MODULE_PARM_DESC(hpsa_allow_any,
+               "Allow hpsa driver to access unknown HP Smart Array hardware");
+
+/* define the PCI info for the cards we can control */
+static const struct pci_device_id hpsa_pci_device_id[] = {
+       {PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSC,     0x103C, 0x3223},
+       {PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSC,     0x103C, 0x3234},
+       {PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSC,     0x103C, 0x323D},
+       {PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x3241},
+       {PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x3243},
+       {PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x3245},
+       {PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x3247},
+       {PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x3249},
+       {PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x324a},
+       {PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x324b},
+       {PCI_VENDOR_ID_HP,     PCI_ANY_ID,             PCI_ANY_ID, PCI_ANY_ID,
+               PCI_CLASS_STORAGE_RAID << 8, 0xffff << 8, 0},
+       {0,}
+};
+
+MODULE_DEVICE_TABLE(pci, hpsa_pci_device_id);
+
+/*  board_id = Subsystem Device ID & Vendor ID
+ *  product = Marketing Name for the board
+ *  access = Address of the struct of function pointers
+ */
+static struct board_type products[] = {
+       {0x3223103C, "Smart Array P800", &SA5_access},
+       {0x3234103C, "Smart Array P400", &SA5_access},
+       {0x323d103c, "Smart Array P700M", &SA5_access},
+       {0x3241103C, "Smart Array P212", &SA5_access},
+       {0x3243103C, "Smart Array P410", &SA5_access},
+       {0x3245103C, "Smart Array P410i", &SA5_access},
+       {0x3247103C, "Smart Array P411", &SA5_access},
+       {0x3249103C, "Smart Array P812", &SA5_access},
+       {0x324a103C, "Smart Array P712m", &SA5_access},
+       {0x324b103C, "Smart Array P711m", &SA5_access},
+       {0xFFFF103C, "Unknown Smart Array", &SA5_access},
+};
+
+static int number_of_controllers;
+
+static irqreturn_t do_hpsa_intr(int irq, void *dev_id);
+static int hpsa_ioctl(struct scsi_device *dev, int cmd, void *arg);
+static void start_io(struct ctlr_info *h);
+
+#ifdef CONFIG_COMPAT
+static int hpsa_compat_ioctl(struct scsi_device *dev, int cmd, void *arg);
+#endif
+
+static void cmd_free(struct ctlr_info *h, struct CommandList *c);
+static void cmd_special_free(struct ctlr_info *h, struct CommandList *c);
+static struct CommandList *cmd_alloc(struct ctlr_info *h);
+static struct CommandList *cmd_special_alloc(struct ctlr_info *h);
+static void fill_cmd(struct CommandList *c, __u8 cmd, struct ctlr_info *h,
+       void *buff, size_t size, __u8 page_code, unsigned char *scsi3addr,
+       int cmd_type);
+
+static int hpsa_scsi_queue_command(struct scsi_cmnd *cmd,
+               void (*done)(struct scsi_cmnd *));
+
+static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd);
+static int hpsa_slave_alloc(struct scsi_device *sdev);
+static void hpsa_slave_destroy(struct scsi_device *sdev);
+
+static ssize_t raid_level_show(struct device *dev,
+       struct device_attribute *attr, char *buf);
+static ssize_t lunid_show(struct device *dev,
+       struct device_attribute *attr, char *buf);
+static ssize_t unique_id_show(struct device *dev,
+       struct device_attribute *attr, char *buf);
+static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno);
+static ssize_t host_store_rescan(struct device *dev,
+        struct device_attribute *attr, const char *buf, size_t count);
+static int check_for_unit_attention(struct ctlr_info *h,
+       struct CommandList *c);
+static void check_ioctl_unit_attention(struct ctlr_info *h,
+       struct CommandList *c);
+
+static DEVICE_ATTR(raid_level, S_IRUGO, raid_level_show, NULL);
+static DEVICE_ATTR(lunid, S_IRUGO, lunid_show, NULL);
+static DEVICE_ATTR(unique_id, S_IRUGO, unique_id_show, NULL);
+static DEVICE_ATTR(rescan, S_IWUSR, NULL, host_store_rescan);
+
+static struct device_attribute *hpsa_sdev_attrs[] = {
+       &dev_attr_raid_level,
+       &dev_attr_lunid,
+       &dev_attr_unique_id,
+       NULL,
+};
+
+static struct device_attribute *hpsa_shost_attrs[] = {
+       &dev_attr_rescan,
+       NULL,
+};
+
+static struct scsi_host_template hpsa_driver_template = {
+       .module                 = THIS_MODULE,
+       .name                   = "hpsa",
+       .proc_name              = "hpsa",
+       .queuecommand           = hpsa_scsi_queue_command,
+       .can_queue              = 512,
+       .this_id                = -1,
+       .sg_tablesize           = MAXSGENTRIES,
+       .cmd_per_lun            = 512,
+       .use_clustering         = ENABLE_CLUSTERING,
+       .eh_device_reset_handler = hpsa_eh_device_reset_handler,
+       .ioctl                  = hpsa_ioctl,
+       .slave_alloc            = hpsa_slave_alloc,
+       .slave_destroy          = hpsa_slave_destroy,
+#ifdef CONFIG_COMPAT
+       .compat_ioctl           = hpsa_compat_ioctl,
+#endif
+       .sdev_attrs = hpsa_sdev_attrs,
+       .shost_attrs = hpsa_shost_attrs,
+};
+
+static inline struct ctlr_info *sdev_to_hba(struct scsi_device *sdev)
+{
+       unsigned long *priv = shost_priv(sdev->host);
+       return (struct ctlr_info *) *priv;
+}
+
+static struct task_struct *hpsa_scan_thread;
+static DEFINE_MUTEX(hpsa_scan_mutex);
+static LIST_HEAD(hpsa_scan_q);
+static int hpsa_scan_func(void *data);
+
+/**
+ * add_to_scan_list() - add controller to rescan queue
+ * @h:               Pointer to the controller.
+ *
+ * Adds the controller to the rescan queue if not already on the queue.
+ *
+ * returns 1 if added to the queue, 0 if skipped (could be on the
+ * queue already, or the controller could be initializing or shutting
+ * down).
+ **/
+static int add_to_scan_list(struct ctlr_info *h)
+{
+       struct ctlr_info *test_h;
+       int found = 0;
+       int ret = 0;
+
+       if (h->busy_initializing)
+               return 0;
+
+       /*
+        * If we don't get the lock, it means the driver is unloading
+        * and there's no point in scheduling a new scan.
+        */
+       if (!mutex_trylock(&h->busy_shutting_down))
+               return 0;
+
+       mutex_lock(&hpsa_scan_mutex);
+       list_for_each_entry(test_h, &hpsa_scan_q, scan_list) {
+               if (test_h == h) {
+                       found = 1;
+                       break;
+               }
+       }
+       if (!found && !h->busy_scanning) {
+               INIT_COMPLETION(h->scan_wait);
+               list_add_tail(&h->scan_list, &hpsa_scan_q);
+               ret = 1;
+       }
+       mutex_unlock(&hpsa_scan_mutex);
+       mutex_unlock(&h->busy_shutting_down);
+
+       return ret;
+}
+
+/**
+ * remove_from_scan_list() - remove controller from rescan queue
+ * @h:                    Pointer to the controller.
+ *
+ * Removes the controller from the rescan queue if present. Blocks if
+ * the controller is currently conducting a rescan.  The controller
+ * can be in one of three states:
+ * 1. Doesn't need a scan
+ * 2. On the scan list, but not scanning yet (we remove it)
+ * 3. Busy scanning (and not on the list). In this case we want to wait for
+ *    the scan to complete to make sure the scanning thread for this
+ *    controller is completely idle.
+ **/
+static void remove_from_scan_list(struct ctlr_info *h)
+{
+       struct ctlr_info *test_h, *tmp_h;
+
+       mutex_lock(&hpsa_scan_mutex);
+       list_for_each_entry_safe(test_h, tmp_h, &hpsa_scan_q, scan_list) {
+               if (test_h == h) { /* state 2. */
+                       list_del(&h->scan_list);
+                       complete_all(&h->scan_wait);
+                       mutex_unlock(&hpsa_scan_mutex);
+                       return;
+               }
+       }
+       if (h->busy_scanning) { /* state 3. */
+               mutex_unlock(&hpsa_scan_mutex);
+               wait_for_completion(&h->scan_wait);
+       } else { /* state 1, nothing to do. */
+               mutex_unlock(&hpsa_scan_mutex);
+       }
+}
+
+/* hpsa_scan_func() - kernel thread used to rescan controllers
+ * @data:       Ignored.
+ *
+ * A kernel thread used scan for drive topology changes on
+ * controllers. The thread processes only one controller at a time
+ * using a queue.  Controllers are added to the queue using
+ * add_to_scan_list() and removed from the queue either after done
+ * processing or using remove_from_scan_list().
+ *
+ * returns 0.
+ **/
+static int hpsa_scan_func(__attribute__((unused)) void *data)
+{
+       struct ctlr_info *h;
+       int host_no;
+
+       while (1) {
+               set_current_state(TASK_INTERRUPTIBLE);
+               schedule();
+               if (kthread_should_stop())
+                       break;
+
+               while (1) {
+                       mutex_lock(&hpsa_scan_mutex);
+                       if (list_empty(&hpsa_scan_q)) {
+                               mutex_unlock(&hpsa_scan_mutex);
+                               break;
+                       }
+                       h = list_entry(hpsa_scan_q.next, struct ctlr_info,
+                                       scan_list);
+                       list_del(&h->scan_list);
+                       h->busy_scanning = 1;
+                       mutex_unlock(&hpsa_scan_mutex);
+                       host_no = h->scsi_host ?  h->scsi_host->host_no : -1;
+                       hpsa_update_scsi_devices(h, host_no);
+                       complete_all(&h->scan_wait);
+                       mutex_lock(&hpsa_scan_mutex);
+                       h->busy_scanning = 0;
+                       mutex_unlock(&hpsa_scan_mutex);
+               }
+       }
+       return 0;
+}
+
+static int check_for_unit_attention(struct ctlr_info *h,
+       struct CommandList *c)
+{
+       if (c->err_info->SenseInfo[2] != UNIT_ATTENTION)
+               return 0;
+
+       switch (c->err_info->SenseInfo[12]) {
+       case STATE_CHANGED:
+               dev_warn(&h->pdev->dev, "hpsa%d: a state change "
+                       "detected, command retried\n", h->ctlr);
+               break;
+       case LUN_FAILED:
+               dev_warn(&h->pdev->dev, "hpsa%d: LUN failure "
+                       "detected, action required\n", h->ctlr);
+               break;
+       case REPORT_LUNS_CHANGED:
+               dev_warn(&h->pdev->dev, "hpsa%d: report LUN data "
+                       "changed\n", h->ctlr);
+       /*
+        * Here, we could call add_to_scan_list and wake up the scan thread,
+        * except that it's quite likely that we will get more than one
+        * REPORT_LUNS_CHANGED condition in quick succession, which means
+        * that those which occur after the first one will likely happen
+        * *during* the hpsa_scan_thread's rescan.  And the rescan code is not
+        * robust enough to restart in the middle, undoing what it has already
+        * done, and it's not clear that it's even possible to do this, since
+        * part of what it does is notify the SCSI mid layer, which starts
+        * doing it's own i/o to read partition tables and so on, and the
+        * driver doesn't have visibility to know what might need undoing.
+        * In any event, if possible, it is horribly complicated to get right
+        * so we just don't do it for now.
+        *
+        * Note: this REPORT_LUNS_CHANGED condition only occurs on the MSA2012.
+        */
+               break;
+       case POWER_OR_RESET:
+               dev_warn(&h->pdev->dev, "hpsa%d: a power on "
+                       "or device reset detected\n", h->ctlr);
+               break;
+       case UNIT_ATTENTION_CLEARED:
+               dev_warn(&h->pdev->dev, "hpsa%d: unit attention "
+                   "cleared by another initiator\n", h->ctlr);
+               break;
+       default:
+               dev_warn(&h->pdev->dev, "hpsa%d: unknown "
+                       "unit attention detected\n", h->ctlr);
+               break;
+       }
+       return 1;
+}
+
+static ssize_t host_store_rescan(struct device *dev,
+                                struct device_attribute *attr,
+                                const char *buf, size_t count)
+{
+       struct ctlr_info *h;
+       struct Scsi_Host *shost = class_to_shost(dev);
+       unsigned long *priv = shost_priv(shost);
+       h = (struct ctlr_info *) *priv;
+       if (add_to_scan_list(h)) {
+               wake_up_process(hpsa_scan_thread);
+               wait_for_completion_interruptible(&h->scan_wait);
+       }
+       return count;
+}
+
+/* Enqueuing and dequeuing functions for cmdlists. */
+static inline void addQ(struct hlist_head *list, struct CommandList *c)
+{
+       hlist_add_head(&c->list, list);
+}
+
+static void enqueue_cmd_and_start_io(struct ctlr_info *h,
+       struct CommandList *c)
+{
+       unsigned long flags;
+       spin_lock_irqsave(&h->lock, flags);
+       addQ(&h->reqQ, c);
+       h->Qdepth++;
+       start_io(h);
+       spin_unlock_irqrestore(&h->lock, flags);
+}
+
+static inline void removeQ(struct CommandList *c)
+{
+       if (WARN_ON(hlist_unhashed(&c->list)))
+               return;
+       hlist_del_init(&c->list);
+}
+
+static inline int is_hba_lunid(unsigned char scsi3addr[])
+{
+       return memcmp(scsi3addr, RAID_CTLR_LUNID, 8) == 0;
+}
+
+static inline int is_logical_dev_addr_mode(unsigned char scsi3addr[])
+{
+       return (scsi3addr[3] & 0xC0) == 0x40;
+}
+
+static const char *raid_label[] = { "0", "4", "1(1+0)", "5", "5+1", "ADG",
+       "UNKNOWN"
+};
+#define RAID_UNKNOWN (ARRAY_SIZE(raid_label) - 1)
+
+static ssize_t raid_level_show(struct device *dev,
+            struct device_attribute *attr, char *buf)
+{
+       ssize_t l = 0;
+       int rlevel;
+       struct ctlr_info *h;
+       struct scsi_device *sdev;
+       struct hpsa_scsi_dev_t *hdev;
+       unsigned long flags;
+
+       sdev = to_scsi_device(dev);
+       h = sdev_to_hba(sdev);
+       spin_lock_irqsave(&h->lock, flags);
+       hdev = sdev->hostdata;
+       if (!hdev) {
+               spin_unlock_irqrestore(&h->lock, flags);
+               return -ENODEV;
+       }
+
+       /* Is this even a logical drive? */
+       if (!is_logical_dev_addr_mode(hdev->scsi3addr)) {
+               spin_unlock_irqrestore(&h->lock, flags);
+               l = snprintf(buf, PAGE_SIZE, "N/A\n");
+               return l;
+       }
+
+       rlevel = hdev->raid_level;
+       spin_unlock_irqrestore(&h->lock, flags);
+       if (rlevel < 0 || rlevel > RAID_UNKNOWN)
+               rlevel = RAID_UNKNOWN;
+       l = snprintf(buf, PAGE_SIZE, "RAID %s\n", raid_label[rlevel]);
+       return l;
+}
+
+static ssize_t lunid_show(struct device *dev,
+            struct device_attribute *attr, char *buf)
+{
+       struct ctlr_info *h;
+       struct scsi_device *sdev;
+       struct hpsa_scsi_dev_t *hdev;
+       unsigned long flags;
+       unsigned char lunid[8];
+
+       sdev = to_scsi_device(dev);
+       h = sdev_to_hba(sdev);
+       spin_lock_irqsave(&h->lock, flags);
+       hdev = sdev->hostdata;
+       if (!hdev) {
+               spin_unlock_irqrestore(&h->lock, flags);
+               return -ENODEV;
+       }
+       memcpy(lunid, hdev->scsi3addr, sizeof(lunid));
+       spin_unlock_irqrestore(&h->lock, flags);
+       return snprintf(buf, 20, "0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
+               lunid[0], lunid[1], lunid[2], lunid[3],
+               lunid[4], lunid[5], lunid[6], lunid[7]);
+}
+
+static ssize_t unique_id_show(struct device *dev,
+            struct device_attribute *attr, char *buf)
+{
+       struct ctlr_info *h;
+       struct scsi_device *sdev;
+       struct hpsa_scsi_dev_t *hdev;
+       unsigned long flags;
+       unsigned char sn[16];
+
+       sdev = to_scsi_device(dev);
+       h = sdev_to_hba(sdev);
+       spin_lock_irqsave(&h->lock, flags);
+       hdev = sdev->hostdata;
+       if (!hdev) {
+               spin_unlock_irqrestore(&h->lock, flags);
+               return -ENODEV;
+       }
+       memcpy(sn, hdev->device_id, sizeof(sn));
+       spin_unlock_irqrestore(&h->lock, flags);
+       return snprintf(buf, 16 * 2 + 2,
+                       "%02X%02X%02X%02X%02X%02X%02X%02X"
+                       "%02X%02X%02X%02X%02X%02X%02X%02X\n",
+                       sn[0], sn[1], sn[2], sn[3],
+                       sn[4], sn[5], sn[6], sn[7],
+                       sn[8], sn[9], sn[10], sn[11],
+                       sn[12], sn[13], sn[14], sn[15]);
+}
+
+static int hpsa_find_target_lun(struct ctlr_info *h,
+       unsigned char scsi3addr[], int bus, int *target, int *lun)
+{
+       /* finds an unused bus, target, lun for a new physical device
+        * assumes h->devlock is held
+        */
+       int i, found = 0;
+       DECLARE_BITMAP(lun_taken, HPSA_MAX_SCSI_DEVS_PER_HBA);
+
+       memset(&lun_taken[0], 0, HPSA_MAX_SCSI_DEVS_PER_HBA >> 3);
+
+       for (i = 0; i < h->ndevices; i++) {
+               if (h->dev[i]->bus == bus && h->dev[i]->target != -1)
+                       set_bit(h->dev[i]->target, lun_taken);
+       }
+
+       for (i = 0; i < HPSA_MAX_SCSI_DEVS_PER_HBA; i++) {
+               if (!test_bit(i, lun_taken)) {
+                       /* *bus = 1; */
+                       *target = i;
+                       *lun = 0;
+                       found = 1;
+                       break;
+               }
+       }
+       return !found;
+}
+
+/* Add an entry into h->dev[] array. */
+static int hpsa_scsi_add_entry(struct ctlr_info *h, int hostno,
+               struct hpsa_scsi_dev_t *device,
+               struct hpsa_scsi_dev_t *added[], int *nadded)
+{
+       /* assumes h->devlock is held */
+       int n = h->ndevices;
+       int i;
+       unsigned char addr1[8], addr2[8];
+       struct hpsa_scsi_dev_t *sd;
+
+       if (n >= HPSA_MAX_SCSI_DEVS_PER_HBA) {
+               dev_err(&h->pdev->dev, "too many devices, some will be "
+                       "inaccessible.\n");
+               return -1;
+       }
+
+       /* physical devices do not have lun or target assigned until now. */
+       if (device->lun != -1)
+               /* Logical device, lun is already assigned. */
+               goto lun_assigned;
+
+       /* If this device a non-zero lun of a multi-lun device
+        * byte 4 of the 8-byte LUN addr will contain the logical
+        * unit no, zero otherise.
+        */
+       if (device->scsi3addr[4] == 0) {
+               /* This is not a non-zero lun of a multi-lun device */
+               if (hpsa_find_target_lun(h, device->scsi3addr,
+                       device->bus, &device->target, &device->lun) != 0)
+                       return -1;
+               goto lun_assigned;
+       }
+
+       /* This is a non-zero lun of a multi-lun device.
+        * Search through our list and find the device which
+        * has the same 8 byte LUN address, excepting byte 4.
+        * Assign the same bus and target for this new LUN.
+        * Use the logical unit number from the firmware.
+        */
+       memcpy(addr1, device->scsi3addr, 8);
+       addr1[4] = 0;
+       for (i = 0; i < n; i++) {
+               sd = h->dev[i];
+               memcpy(addr2, sd->scsi3addr, 8);
+               addr2[4] = 0;
+               /* differ only in byte 4? */
+               if (memcmp(addr1, addr2, 8) == 0) {
+                       device->bus = sd->bus;
+                       device->target = sd->target;
+                       device->lun = device->scsi3addr[4];
+                       break;
+               }
+       }
+       if (device->lun == -1) {
+               dev_warn(&h->pdev->dev, "physical device with no LUN=0,"
+                       " suspect firmware bug or unsupported hardware "
+                       "configuration.\n");
+                       return -1;
+       }
+
+lun_assigned:
+
+       h->dev[n] = device;
+       h->ndevices++;
+       added[*nadded] = device;
+       (*nadded)++;
+
+       /* initially, (before registering with scsi layer) we don't
+        * know our hostno and we don't want to print anything first
+        * time anyway (the scsi layer's inquiries will show that info)
+        */
+       /* if (hostno != -1) */
+               dev_info(&h->pdev->dev, "%s device c%db%dt%dl%d added.\n",
+                       scsi_device_type(device->devtype), hostno,
+                       device->bus, device->target, device->lun);
+       return 0;
+}
+
+/* Remove an entry from h->dev[] array. */
+static void hpsa_scsi_remove_entry(struct ctlr_info *h, int hostno, int entry,
+       struct hpsa_scsi_dev_t *removed[], int *nremoved)
+{
+       /* assumes h->devlock is held */
+       int i;
+       struct hpsa_scsi_dev_t *sd;
+
+       if (entry < 0 || entry >= HPSA_MAX_SCSI_DEVS_PER_HBA)
+               BUG();
+
+       sd = h->dev[entry];
+       removed[*nremoved] = h->dev[entry];
+       (*nremoved)++;
+
+       for (i = entry; i < h->ndevices-1; i++)
+               h->dev[i] = h->dev[i+1];
+       h->ndevices--;
+       dev_info(&h->pdev->dev, "%s device c%db%dt%dl%d removed.\n",
+               scsi_device_type(sd->devtype), hostno, sd->bus, sd->target,
+               sd->lun);
+}
+
+#define SCSI3ADDR_EQ(a, b) ( \
+       (a)[7] == (b)[7] && \
+       (a)[6] == (b)[6] && \
+       (a)[5] == (b)[5] && \
+       (a)[4] == (b)[4] && \
+       (a)[3] == (b)[3] && \
+       (a)[2] == (b)[2] && \
+       (a)[1] == (b)[1] && \
+       (a)[0] == (b)[0])
+
+static void fixup_botched_add(struct ctlr_info *h,
+       struct hpsa_scsi_dev_t *added)
+{
+       /* called when scsi_add_device fails in order to re-adjust
+        * h->dev[] to match the mid layer's view.
+        */
+       unsigned long flags;
+       int i, j;
+
+       spin_lock_irqsave(&h->lock, flags);
+       for (i = 0; i < h->ndevices; i++) {
+               if (h->dev[i] == added) {
+                       for (j = i; j < h->ndevices-1; j++)
+                               h->dev[j] = h->dev[j+1];
+                       h->ndevices--;
+                       break;
+               }
+       }
+       spin_unlock_irqrestore(&h->lock, flags);
+       kfree(added);
+}
+
+static inline int device_is_the_same(struct hpsa_scsi_dev_t *dev1,
+       struct hpsa_scsi_dev_t *dev2)
+{
+       if ((is_logical_dev_addr_mode(dev1->scsi3addr) ||
+               (dev1->lun != -1 && dev2->lun != -1)) &&
+               dev1->devtype != 0x0C)
+               return (memcmp(dev1, dev2, sizeof(*dev1)) == 0);
+
+       /* we compare everything except lun and target as these
+        * are not yet assigned.  Compare parts likely
+        * to differ first
+        */
+       if (memcmp(dev1->scsi3addr, dev2->scsi3addr,
+               sizeof(dev1->scsi3addr)) != 0)
+               return 0;
+       if (memcmp(dev1->device_id, dev2->device_id,
+               sizeof(dev1->device_id)) != 0)
+               return 0;
+       if (memcmp(dev1->model, dev2->model, sizeof(dev1->model)) != 0)
+               return 0;
+       if (memcmp(dev1->vendor, dev2->vendor, sizeof(dev1->vendor)) != 0)
+               return 0;
+       if (memcmp(dev1->revision, dev2->revision, sizeof(dev1->revision)) != 0)
+               return 0;
+       if (dev1->devtype != dev2->devtype)
+               return 0;
+       if (dev1->raid_level != dev2->raid_level)
+               return 0;
+       if (dev1->bus != dev2->bus)
+               return 0;
+       return 1;
+}
+
+/* Find needle in haystack.  If exact match found, return DEVICE_SAME,
+ * and return needle location in *index.  If scsi3addr matches, but not
+ * vendor, model, serial num, etc. return DEVICE_CHANGED, and return needle
+ * location in *index.  If needle not found, return DEVICE_NOT_FOUND.
+ */
+static int hpsa_scsi_find_entry(struct hpsa_scsi_dev_t *needle,
+       struct hpsa_scsi_dev_t *haystack[], int haystack_size,
+       int *index)
+{
+       int i;
+#define DEVICE_NOT_FOUND 0
+#define DEVICE_CHANGED 1
+#define DEVICE_SAME 2
+       for (i = 0; i < haystack_size; i++) {
+               if (SCSI3ADDR_EQ(needle->scsi3addr, haystack[i]->scsi3addr)) {
+                       *index = i;
+                       if (device_is_the_same(needle, haystack[i]))
+                               return DEVICE_SAME;
+                       else
+                               return DEVICE_CHANGED;
+               }
+       }
+       *index = -1;
+       return DEVICE_NOT_FOUND;
+}
+
+static int adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno,
+       struct hpsa_scsi_dev_t *sd[], int nsds)
+{
+       /* sd contains scsi3 addresses and devtypes, and inquiry
+        * data.  This function takes what's in sd to be the current
+        * reality and updates h->dev[] to reflect that reality.
+        */
+       int i, entry, device_change, changes = 0;
+       struct hpsa_scsi_dev_t *csd;
+       unsigned long flags;
+       struct hpsa_scsi_dev_t **added, **removed;
+       int nadded, nremoved;
+       struct Scsi_Host *sh = NULL;
+
+       added = kzalloc(sizeof(*added) * HPSA_MAX_SCSI_DEVS_PER_HBA,
+               GFP_KERNEL);
+       removed = kzalloc(sizeof(*removed) * HPSA_MAX_SCSI_DEVS_PER_HBA,
+               GFP_KERNEL);
+
+       if (!added || !removed) {
+               dev_warn(&h->pdev->dev, "out of memory in "
+                       "adjust_hpsa_scsi_table\n");
+               goto free_and_out;
+       }
+
+       spin_lock_irqsave(&h->devlock, flags);
+
+       /* find any devices in h->dev[] that are not in
+        * sd[] and remove them from h->dev[], and for any
+        * devices which have changed, remove the old device
+        * info and add the new device info.
+        */
+       i = 0;
+       nremoved = 0;
+       nadded = 0;
+       while (i < h->ndevices) {
+               csd = h->dev[i];
+               device_change = hpsa_scsi_find_entry(csd, sd, nsds, &entry);
+               if (device_change == DEVICE_NOT_FOUND) {
+                       changes++;
+                       hpsa_scsi_remove_entry(h, hostno, i,
+                               removed, &nremoved);
+                       continue; /* remove ^^^, hence i not incremented */
+               } else if (device_change == DEVICE_CHANGED) {
+                       changes++;
+                       hpsa_scsi_remove_entry(h, hostno, i,
+                               removed, &nremoved);
+                       (void) hpsa_scsi_add_entry(h, hostno, sd[entry],
+                               added, &nadded);
+                       /* add can't fail, we just removed one. */
+                       sd[entry] = NULL; /* prevent it from being freed */
+               }
+               i++;
+       }
+
+       /* Now, make sure every device listed in sd[] is also
+        * listed in h->dev[], adding them if they aren't found
+        */
+
+       for (i = 0; i < nsds; i++) {
+               if (!sd[i]) /* if already added above. */
+                       continue;
+               device_change = hpsa_scsi_find_entry(sd[i], h->dev,
+                                       h->ndevices, &entry);
+               if (device_change == DEVICE_NOT_FOUND) {
+                       changes++;
+                       if (hpsa_scsi_add_entry(h, hostno, sd[i],
+                               added, &nadded) != 0)
+                               break;
+                       sd[i] = NULL; /* prevent from being freed later. */
+               } else if (device_change == DEVICE_CHANGED) {
+                       /* should never happen... */
+                       changes++;
+                       dev_warn(&h->pdev->dev,
+                               "device unexpectedly changed.\n");
+                       /* but if it does happen, we just ignore that device */
+               }
+       }
+       spin_unlock_irqrestore(&h->devlock, flags);
+
+       /* Don't notify scsi mid layer of any changes the first time through
+        * (or if there are no changes) scsi_scan_host will do it later the
+        * first time through.
+        */
+       if (hostno == -1 || !changes)
+               goto free_and_out;
+
+       sh = h->scsi_host;
+       /* Notify scsi mid layer of any removed devices */
+       for (i = 0; i < nremoved; i++) {
+               struct scsi_device *sdev =
+                       scsi_device_lookup(sh, removed[i]->bus,
+                               removed[i]->target, removed[i]->lun);
+               if (sdev != NULL) {
+                       scsi_remove_device(sdev);
+                       scsi_device_put(sdev);
+               } else {
+                       /* We don't expect to get here.
+                        * future cmds to this device will get selection
+                        * timeout as if the device was gone.
+                        */
+                       dev_warn(&h->pdev->dev, "didn't find c%db%dt%dl%d "
+                               " for removal.", hostno, removed[i]->bus,
+                               removed[i]->target, removed[i]->lun);
+               }
+               kfree(removed[i]);
+               removed[i] = NULL;
+       }
+
+       /* Notify scsi mid layer of any added devices */
+       for (i = 0; i < nadded; i++) {
+               if (scsi_add_device(sh, added[i]->bus,
+                       added[i]->target, added[i]->lun) == 0)
+                       continue;
+               dev_warn(&h->pdev->dev, "scsi_add_device c%db%dt%dl%d failed, "
+                       "device not added.\n", hostno, added[i]->bus,
+                       added[i]->target, added[i]->lun);
+               /* now we have to remove it from h->dev,
+                * since it didn't get added to scsi mid layer
+                */
+               fixup_botched_add(h, added[i]);
+       }
+
+free_and_out:
+       kfree(added);
+       kfree(removed);
+       return 0;
+}
+
+/*
+ * Lookup bus/target/lun and retrun corresponding struct hpsa_scsi_dev_t *
+ * Assume's h->devlock is held.
+ */
+static struct hpsa_scsi_dev_t *lookup_hpsa_scsi_dev(struct ctlr_info *h,
+       int bus, int target, int lun)
+{
+       int i;
+       struct hpsa_scsi_dev_t *sd;
+
+       for (i = 0; i < h->ndevices; i++) {
+               sd = h->dev[i];
+               if (sd->bus == bus && sd->target == target && sd->lun == lun)
+                       return sd;
+       }
+       return NULL;
+}
+
+/* link sdev->hostdata to our per-device structure. */
+static int hpsa_slave_alloc(struct scsi_device *sdev)
+{
+       struct hpsa_scsi_dev_t *sd;
+       unsigned long flags;
+       struct ctlr_info *h;
+
+       h = sdev_to_hba(sdev);
+       spin_lock_irqsave(&h->devlock, flags);
+       sd = lookup_hpsa_scsi_dev(h, sdev_channel(sdev),
+               sdev_id(sdev), sdev->lun);
+       if (sd != NULL)
+               sdev->hostdata = sd;
+       spin_unlock_irqrestore(&h->devlock, flags);
+       return 0;
+}
+
+static void hpsa_slave_destroy(struct scsi_device *sdev)
+{
+       return; /* nothing to do. */
+}
+
+static void hpsa_scsi_setup(struct ctlr_info *h)
+{
+       h->ndevices = 0;
+       h->scsi_host = NULL;
+       spin_lock_init(&h->devlock);
+       return;
+}
+
+static void complete_scsi_command(struct CommandList *cp,
+       int timeout, __u32 tag)
+{
+       struct scsi_cmnd *cmd;
+       struct ctlr_info *h;
+       struct ErrorInfo *ei;
+
+       unsigned char sense_key;
+       unsigned char asc;      /* additional sense code */
+       unsigned char ascq;     /* additional sense code qualifier */
+
+       ei = cp->err_info;
+       cmd = (struct scsi_cmnd *) cp->scsi_cmd;
+       h = cp->h;
+
+       scsi_dma_unmap(cmd); /* undo the DMA mappings */
+
+       cmd->result = (DID_OK << 16);           /* host byte */
+       cmd->result |= (COMMAND_COMPLETE << 8); /* msg byte */
+       cmd->result |= (ei->ScsiStatus << 1);
+
+       /* copy the sense data whether we need to or not. */
+       memcpy(cmd->sense_buffer, ei->SenseInfo,
+               ei->SenseLen > SCSI_SENSE_BUFFERSIZE ?
+                       SCSI_SENSE_BUFFERSIZE :
+                       ei->SenseLen);
+       scsi_set_resid(cmd, ei->ResidualCnt);
+
+       if (ei->CommandStatus == 0) {
+               cmd->scsi_done(cmd);
+               cmd_free(h, cp);
+               return;
+       }
+
+       /* an error has occurred */
+       switch (ei->CommandStatus) {
+
+       case CMD_TARGET_STATUS:
+               if (ei->ScsiStatus) {
+                       /* Get sense key */
+                       sense_key = 0xf & ei->SenseInfo[2];
+                       /* Get additional sense code */
+                       asc = ei->SenseInfo[12];
+                       /* Get addition sense code qualifier */
+                       ascq = ei->SenseInfo[13];
+               }
+
+               if (ei->ScsiStatus == SAM_STAT_CHECK_CONDITION) {
+                       if (check_for_unit_attention(h, cp)) {
+                               cmd->result = DID_SOFT_ERROR << 16;
+                               break;
+                       }
+                       if (sense_key == ILLEGAL_REQUEST) {
+                               /*
+                                * SCSI REPORT_LUNS is commonly unsupported on
+                                * Smart Array.  Suppress noisy complaint.
+                                */
+                               if (cp->Request.CDB[0] == REPORT_LUNS)
+                                       break;
+
+                               /* If ASC/ASCQ indicate Logical Unit
+                                * Not Supported condition,
+                                */
+                               if ((asc == 0x25) && (ascq == 0x0)) {
+                                       dev_warn(&h->pdev->dev, "cp %p "
+                                               "has check condition\n", cp);
+                                       break;
+                               }
+                       }
+
+                       if (sense_key == NOT_READY) {
+                               /* If Sense is Not Ready, Logical Unit
+                                * Not ready, Manual Intervention
+                                * required
+                                */
+                               if ((asc == 0x04) && (ascq == 0x03)) {
+                                       cmd->result = DID_NO_CONNECT << 16;
+                                       dev_warn(&h->pdev->dev, "cp %p "
+                                               "has check condition: unit "
+                                               "not ready, manual "
+                                               "intervention required\n", cp);
+                                       break;
+                               }
+                       }
+
+
+                       /* Must be some other type of check condition */
+                       dev_warn(&h->pdev->dev, "cp %p has check condition: "
+                                       "unknown type: "
+                                       "Sense: 0x%x, ASC: 0x%x, ASCQ: 0x%x, "
+                                       "Returning result: 0x%x, "
+                                       "cmd=[%02x %02x %02x %02x %02x "
+                                       "%02x %02x %02x %02x %02x]\n",
+                                       cp, sense_key, asc, ascq,
+                                       cmd->result,
+                                       cmd->cmnd[0], cmd->cmnd[1],
+                                       cmd->cmnd[2], cmd->cmnd[3],
+                                       cmd->cmnd[4], cmd->cmnd[5],
+                                       cmd->cmnd[6], cmd->cmnd[7],
+                                       cmd->cmnd[8], cmd->cmnd[9]);
+                       break;
+               }
+
+
+               /* Problem was not a check condition
+                * Pass it up to the upper layers...
+                */
+               if (ei->ScsiStatus) {
+                       dev_warn(&h->pdev->dev, "cp %p has status 0x%x "
+                               "Sense: 0x%x, ASC: 0x%x, ASCQ: 0x%x, "
+                               "Returning result: 0x%x\n",
+                               cp, ei->ScsiStatus,
+                               sense_key, asc, ascq,
+                               cmd->result);
+               } else {  /* scsi status is zero??? How??? */
+                       dev_warn(&h->pdev->dev, "cp %p SCSI status was 0. "
+                               "Returning no connection.\n", cp),
+
+                       /* Ordinarily, this case should never happen,
+                        * but there is a bug in some released firmware
+                        * revisions that allows it to happen if, for
+                        * example, a 4100 backplane loses power and
+                        * the tape drive is in it.  We assume that
+                        * it's a fatal error of some kind because we
+                        * can't show that it wasn't. We will make it
+                        * look like selection timeout since that is
+                        * the most common reason for this to occur,
+                        * and it's severe enough.
+                        */
+
+                       cmd->result = DID_NO_CONNECT << 16;
+               }
+               break;
+
+       case CMD_DATA_UNDERRUN: /* let mid layer handle it. */
+               break;
+       case CMD_DATA_OVERRUN:
+               dev_warn(&h->pdev->dev, "cp %p has"
+                       " completed with data overrun "
+                       "reported\n", cp);
+               break;
+       case CMD_INVALID: {
+               /* print_bytes(cp, sizeof(*cp), 1, 0);
+               print_cmd(cp); */
+               /* We get CMD_INVALID if you address a non-existent device
+                * instead of a selection timeout (no response).  You will
+                * see this if you yank out a drive, then try to access it.
+                * This is kind of a shame because it means that any other
+                * CMD_INVALID (e.g. driver bug) will get interpreted as a
+                * missing target. */
+               cmd->result = DID_NO_CONNECT << 16;
+       }
+               break;
+       case CMD_PROTOCOL_ERR:
+               dev_warn(&h->pdev->dev, "cp %p has "
+                       "protocol error \n", cp);
+               break;
+       case CMD_HARDWARE_ERR:
+               cmd->result = DID_ERROR << 16;
+               dev_warn(&h->pdev->dev, "cp %p had  hardware error\n", cp);
+               break;
+       case CMD_CONNECTION_LOST:
+               cmd->result = DID_ERROR << 16;
+               dev_warn(&h->pdev->dev, "cp %p had connection lost\n", cp);
+               break;
+       case CMD_ABORTED:
+               cmd->result = DID_ABORT << 16;
+               dev_warn(&h->pdev->dev, "cp %p was aborted with status 0x%x\n",
+                               cp, ei->ScsiStatus);
+               break;
+       case CMD_ABORT_FAILED:
+               cmd->result = DID_ERROR << 16;
+               dev_warn(&h->pdev->dev, "cp %p reports abort failed\n", cp);
+               break;
+       case CMD_UNSOLICITED_ABORT:
+               cmd->result = DID_ABORT << 16;
+               dev_warn(&h->pdev->dev, "cp %p aborted do to an unsolicited "
+                       "abort\n", cp);
+               break;
+       case CMD_TIMEOUT:
+               cmd->result = DID_TIME_OUT << 16;
+               dev_warn(&h->pdev->dev, "cp %p timedout\n", cp);
+               break;
+       default:
+               cmd->result = DID_ERROR << 16;
+               dev_warn(&h->pdev->dev, "cp %p returned unknown status %x\n",
+                               cp, ei->CommandStatus);
+       }
+       cmd->scsi_done(cmd);
+       cmd_free(h, cp);
+}
+
+static int hpsa_scsi_detect(struct ctlr_info *h)
+{
+       struct Scsi_Host *sh;
+       int error;
+
+       sh = scsi_host_alloc(&hpsa_driver_template, sizeof(h));
+       if (sh == NULL)
+               goto fail;
+
+       sh->io_port = 0;
+       sh->n_io_port = 0;
+       sh->this_id = -1;
+       sh->max_channel = 3;
+       sh->max_cmd_len = MAX_COMMAND_SIZE;
+       sh->max_lun = HPSA_MAX_LUN;
+       sh->max_id = HPSA_MAX_LUN;
+       h->scsi_host = sh;
+       sh->hostdata[0] = (unsigned long) h;
+       sh->irq = h->intr[SIMPLE_MODE_INT];
+       sh->unique_id = sh->irq;
+       error = scsi_add_host(sh, &h->pdev->dev);
+       if (error)
+               goto fail_host_put;
+       scsi_scan_host(sh);
+       return 0;
+
+ fail_host_put:
+       dev_err(&h->pdev->dev, "hpsa_scsi_detect: scsi_add_host"
+               " failed for controller %d\n", h->ctlr);
+       scsi_host_put(sh);
+       return -1;
+ fail:
+       dev_err(&h->pdev->dev, "hpsa_scsi_detect: scsi_host_alloc"
+               " failed for controller %d\n", h->ctlr);
+       return -1;
+}
+
+static void hpsa_pci_unmap(struct pci_dev *pdev,
+       struct CommandList *c, int sg_used, int data_direction)
+{
+       int i;
+       union u64bit addr64;
+
+       for (i = 0; i < sg_used; i++) {
+               addr64.val32.lower = c->SG[i].Addr.lower;
+               addr64.val32.upper = c->SG[i].Addr.upper;
+               pci_unmap_single(pdev, (dma_addr_t) addr64.val, c->SG[i].Len,
+                       data_direction);
+       }
+}
+
+static void hpsa_map_one(struct pci_dev *pdev,
+               struct CommandList *cp,
+               unsigned char *buf,
+               size_t buflen,
+               int data_direction)
+{
+       __u64 addr64;
+
+       if (buflen == 0 || data_direction == PCI_DMA_NONE) {
+               cp->Header.SGList = 0;
+               cp->Header.SGTotal = 0;
+               return;
+       }
+
+       addr64 = (__u64) pci_map_single(pdev, buf, buflen, data_direction);
+       cp->SG[0].Addr.lower =
+         (__u32) (addr64 & (__u64) 0x00000000FFFFFFFF);
+       cp->SG[0].Addr.upper =
+         (__u32) ((addr64 >> 32) & (__u64) 0x00000000FFFFFFFF);
+       cp->SG[0].Len = buflen;
+       cp->Header.SGList = (__u8) 1;   /* no. SGs contig in this cmd */
+       cp->Header.SGTotal = (__u16) 1; /* total sgs in this cmd list */
+}
+
+static inline void hpsa_scsi_do_simple_cmd_core(struct ctlr_info *h,
+       struct CommandList *c)
+{
+       DECLARE_COMPLETION_ONSTACK(wait);
+
+       c->waiting = &wait;
+       enqueue_cmd_and_start_io(h, c);
+       wait_for_completion(&wait);
+}
+
+static void hpsa_scsi_do_simple_cmd_with_retry(struct ctlr_info *h,
+       struct CommandList *c, int data_direction)
+{
+       int retry_count = 0;
+
+       do {
+               memset(c->err_info, 0, sizeof(c->err_info));
+               hpsa_scsi_do_simple_cmd_core(h, c);
+               retry_count++;
+       } while (check_for_unit_attention(h, c) && retry_count <= 3);
+       hpsa_pci_unmap(h->pdev, c, 1, data_direction);
+}
+
+static void hpsa_scsi_interpret_error(struct CommandList *cp)
+{
+       struct ErrorInfo *ei;
+       struct device *d = &cp->h->pdev->dev;
+
+       ei = cp->err_info;
+       switch (ei->CommandStatus) {
+       case CMD_TARGET_STATUS:
+               dev_warn(d, "cmd %p has completed with errors\n", cp);
+               dev_warn(d, "cmd %p has SCSI Status = %x\n", cp,
+                               ei->ScsiStatus);
+               if (ei->ScsiStatus == 0)
+                       dev_warn(d, "SCSI status is abnormally zero.  "
+                       "(probably indicates selection timeout "
+                       "reported incorrectly due to a known "
+                       "firmware bug, circa July, 2001.)\n");
+               break;
+       case CMD_DATA_UNDERRUN: /* let mid layer handle it. */
+                       dev_info(d, "UNDERRUN\n");
+               break;
+       case CMD_DATA_OVERRUN:
+               dev_warn(d, "cp %p has completed with data overrun\n", cp);
+               break;
+       case CMD_INVALID: {
+               /* controller unfortunately reports SCSI passthru's
+                * to non-existent targets as invalid commands.
+                */
+               dev_warn(d, "cp %p is reported invalid (probably means "
+                       "target device no longer present)\n", cp);
+               /* print_bytes((unsigned char *) cp, sizeof(*cp), 1, 0);
+               print_cmd(cp);  */
+               }
+               break;
+       case CMD_PROTOCOL_ERR:
+               dev_warn(d, "cp %p has protocol error \n", cp);
+               break;
+       case CMD_HARDWARE_ERR:
+               /* cmd->result = DID_ERROR << 16; */
+               dev_warn(d, "cp %p had hardware error\n", cp);
+               break;
+       case CMD_CONNECTION_LOST:
+               dev_warn(d, "cp %p had connection lost\n", cp);
+               break;
+       case CMD_ABORTED:
+               dev_warn(d, "cp %p was aborted\n", cp);
+               break;
+       case CMD_ABORT_FAILED:
+               dev_warn(d, "cp %p reports abort failed\n", cp);
+               break;
+       case CMD_UNSOLICITED_ABORT:
+               dev_warn(d, "cp %p aborted due to an unsolicited abort\n", cp);
+               break;
+       case CMD_TIMEOUT:
+               dev_warn(d, "cp %p timed out\n", cp);
+               break;
+       default:
+               dev_warn(d, "cp %p returned unknown status %x\n", cp,
+                               ei->CommandStatus);
+       }
+}
+
+static int hpsa_scsi_do_inquiry(struct ctlr_info *h, unsigned char *scsi3addr,
+                       unsigned char page, unsigned char *buf,
+                       unsigned char bufsize)
+{
+       int rc = IO_OK;
+       struct CommandList *c;
+       struct ErrorInfo *ei;
+
+       c = cmd_special_alloc(h);
+
+       if (c == NULL) {                        /* trouble... */
+               dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n");
+               return -1;
+       }
+
+       fill_cmd(c, HPSA_INQUIRY, h, buf, bufsize, page, scsi3addr, TYPE_CMD);
+       hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE);
+       ei = c->err_info;
+       if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) {
+               hpsa_scsi_interpret_error(c);
+               rc = -1;
+       }
+       cmd_special_free(h, c);
+       return rc;
+}
+
+static int hpsa_send_reset(struct ctlr_info *h, unsigned char *scsi3addr)
+{
+       int rc = IO_OK;
+       struct CommandList *c;
+       struct ErrorInfo *ei;
+
+       c = cmd_special_alloc(h);
+
+       if (c == NULL) {                        /* trouble... */
+               dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n");
+               return -1;
+       }
+
+       fill_cmd(c, HPSA_DEVICE_RESET_MSG, h, NULL, 0, 0, scsi3addr, TYPE_MSG);
+       hpsa_scsi_do_simple_cmd_core(h, c);
+       /* no unmap needed here because no data xfer. */
+
+       ei = c->err_info;
+       if (ei->CommandStatus != 0) {
+               hpsa_scsi_interpret_error(c);
+               rc = -1;
+       }
+       cmd_special_free(h, c);
+       return rc;
+}
+
+static void hpsa_get_raid_level(struct ctlr_info *h,
+       unsigned char *scsi3addr, unsigned char *raid_level)
+{
+       int rc;
+       unsigned char *buf;
+
+       *raid_level = RAID_UNKNOWN;
+       buf = kzalloc(64, GFP_KERNEL);
+       if (!buf)
+               return;
+       rc = hpsa_scsi_do_inquiry(h, scsi3addr, 0xC1, buf, 64);
+       if (rc == 0)
+               *raid_level = buf[8];
+       if (*raid_level > RAID_UNKNOWN)
+               *raid_level = RAID_UNKNOWN;
+       kfree(buf);
+       return;
+}
+
+/* Get the device id from inquiry page 0x83 */
+static int hpsa_get_device_id(struct ctlr_info *h, unsigned char *scsi3addr,
+       unsigned char *device_id, int buflen)
+{
+       int rc;
+       unsigned char *buf;
+
+       if (buflen > 16)
+               buflen = 16;
+       buf = kzalloc(64, GFP_KERNEL);
+       if (!buf)
+               return -1;
+       rc = hpsa_scsi_do_inquiry(h, scsi3addr, 0x83, buf, 64);
+       if (rc == 0)
+               memcpy(device_id, &buf[8], buflen);
+       kfree(buf);
+       return rc != 0;
+}
+
+static int hpsa_scsi_do_report_luns(struct ctlr_info *h, int logical,
+               struct ReportLUNdata *buf, int bufsize,
+               int extended_response)
+{
+       int rc = IO_OK;
+       struct CommandList *c;
+       unsigned char scsi3addr[8];
+       struct ErrorInfo *ei;
+
+       c = cmd_special_alloc(h);
+       if (c == NULL) {                        /* trouble... */
+               dev_err(&h->pdev->dev, "cmd_special_alloc returned NULL!\n");
+               return -1;
+       }
+
+       memset(&scsi3addr[0], 0, 8); /* address the controller */
+
+       fill_cmd(c, logical ? HPSA_REPORT_LOG : HPSA_REPORT_PHYS, h,
+               buf, bufsize, 0, scsi3addr, TYPE_CMD);
+       if (extended_response)
+               c->Request.CDB[1] = extended_response;
+       hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE);
+       ei = c->err_info;
+       if (ei->CommandStatus != 0 &&
+           ei->CommandStatus != CMD_DATA_UNDERRUN) {
+               hpsa_scsi_interpret_error(c);
+               rc = -1;
+       }
+       cmd_special_free(h, c);
+       return rc;
+}
+
+static inline int hpsa_scsi_do_report_phys_luns(struct ctlr_info *h,
+               struct ReportLUNdata *buf,
+               int bufsize, int extended_response)
+{
+       return hpsa_scsi_do_report_luns(h, 0, buf, bufsize, extended_response);
+}
+
+static inline int hpsa_scsi_do_report_log_luns(struct ctlr_info *h,
+               struct ReportLUNdata *buf, int bufsize)
+{
+       return hpsa_scsi_do_report_luns(h, 1, buf, bufsize, 0);
+}
+
+static inline void hpsa_set_bus_target_lun(struct hpsa_scsi_dev_t *device,
+       int bus, int target, int lun)
+{
+       device->bus = bus;
+       device->target = target;
+       device->lun = lun;
+}
+
+static int hpsa_update_device_info(struct ctlr_info *h,
+       unsigned char scsi3addr[], struct hpsa_scsi_dev_t *this_device)
+{
+#define OBDR_TAPE_INQ_SIZE 49
+       unsigned char *inq_buff = NULL;
+
+       inq_buff = kmalloc(OBDR_TAPE_INQ_SIZE, GFP_KERNEL);
+       if (!inq_buff)
+               goto bail_out;
+
+       memset(inq_buff, 0, OBDR_TAPE_INQ_SIZE);
+       /* Do an inquiry to the device to see what it is. */
+       if (hpsa_scsi_do_inquiry(h, scsi3addr, 0, inq_buff,
+               (unsigned char) OBDR_TAPE_INQ_SIZE) != 0) {
+               /* Inquiry failed (msg printed already) */
+               dev_err(&h->pdev->dev,
+                       "hpsa_update_device_info: inquiry failed\n");
+               goto bail_out;
+       }
+
+       /* As a side effect, record the firmware version number
+        * if we happen to be talking to the RAID controller.
+        */
+       if (is_hba_lunid(scsi3addr))
+               memcpy(h->firm_ver, &inq_buff[32], 4);
+
+       this_device->devtype = (inq_buff[0] & 0x1f);
+       memcpy(this_device->scsi3addr, scsi3addr, 8);
+       memcpy(this_device->vendor, &inq_buff[8],
+               sizeof(this_device->vendor));
+       memcpy(this_device->model, &inq_buff[16],
+               sizeof(this_device->model));
+       memcpy(this_device->revision, &inq_buff[32],
+               sizeof(this_device->revision));
+       memset(this_device->device_id, 0,
+               sizeof(this_device->device_id));
+       hpsa_get_device_id(h, scsi3addr, this_device->device_id,
+               sizeof(this_device->device_id));
+
+       if (this_device->devtype == TYPE_DISK &&
+               is_logical_dev_addr_mode(scsi3addr))
+               hpsa_get_raid_level(h, scsi3addr, &this_device->raid_level);
+       else
+               this_device->raid_level = RAID_UNKNOWN;
+
+       kfree(inq_buff);
+       return 0;
+
+bail_out:
+       kfree(inq_buff);
+       return 1;
+}
+
+static unsigned char *msa2xxx_model[] = {
+       "MSA2012",
+       "MSA2024",
+       "MSA2312",
+       "MSA2324",
+       NULL,
+};
+
+static int is_msa2xxx(struct ctlr_info *h, struct hpsa_scsi_dev_t *device)
+{
+       int i;
+
+       for (i = 0; msa2xxx_model[i]; i++)
+               if (strncmp(device->model, msa2xxx_model[i],
+                       strlen(msa2xxx_model[i])) == 0)
+                       return 1;
+       return 0;
+}
+
+/* Helper function to assign bus, target, lun mapping of devices.
+ * Puts non-msa2xxx logical volumes on bus 0, msa2xxx logical
+ * volumes on bus 1, physical devices on bus 2. and the hba on bus 3.
+ * Logical drive target and lun are assigned at this time, but
+ * physical device lun and target assignment are deferred (assigned
+ * in hpsa_find_target_lun, called by hpsa_scsi_add_entry.)
+ */
+static void figure_bus_target_lun(struct ctlr_info *h,
+       __u8 *lunaddrbytes, int *bus, int *target, int *lun,
+       struct hpsa_scsi_dev_t *device)
+{
+
+       __u32 lunid;
+
+       if (is_logical_dev_addr_mode(lunaddrbytes)) {
+               /* logical device */
+               memcpy(&lunid, lunaddrbytes, sizeof(lunid));
+               lunid = le32_to_cpu(lunid);
+
+               if (is_msa2xxx(h, device)) {
+                       *bus = 1;
+                       *target = (lunid >> 16) & 0x3fff;
+                       *lun = lunid & 0x00ff;
+               } else {
+                       *bus = 0;
+                       *lun = 0;
+                       *target = lunid & 0x3fff;
+               }
+       } else {
+               /* physical device */
+               if (is_hba_lunid(lunaddrbytes))
+                       *bus = 3;
+               else
+                       *bus = 2;
+               *target = -1;
+               *lun = -1; /* we will fill these in later. */
+       }
+}
+
+/*
+ * If there is no lun 0 on a target, linux won't find any devices.
+ * For the MSA2xxx boxes, we have to manually detect the enclosure
+ * which is at lun zero, as CCISS_REPORT_PHYSICAL_LUNS doesn't report
+ * it for some reason.  *tmpdevice is the target we're adding,
+ * this_device is a pointer into the current element of currentsd[]
+ * that we're building up in update_scsi_devices(), below.
+ * lunzerobits is a bitmap that tracks which targets already have a
+ * lun 0 assigned.
+ * Returns 1 if an enclosure was added, 0 if not.
+ */
+static int add_msa2xxx_enclosure_device(struct ctlr_info *h,
+       struct hpsa_scsi_dev_t *tmpdevice,
+       struct hpsa_scsi_dev_t *this_device, __u8 *lunaddrbytes,
+       int bus, int target, int lun, unsigned long lunzerobits[],
+       int *nmsa2xxx_enclosures)
+{
+       unsigned char scsi3addr[8];
+
+       if (test_bit(target, lunzerobits))
+               return 0; /* There is already a lun 0 on this target. */
+
+       if (!is_logical_dev_addr_mode(lunaddrbytes))
+               return 0; /* It's the logical targets that may lack lun 0. */
+
+       if (!is_msa2xxx(h, tmpdevice))
+               return 0; /* It's only the MSA2xxx that have this problem. */
+
+       if (lun == 0) /* if lun is 0, then obviously we have a lun 0. */
+               return 0;
+
+       if (is_hba_lunid(scsi3addr))
+               return 0; /* Don't add the RAID controller here. */
+
+#define MAX_MSA2XXX_ENCLOSURES 32
+       if (*nmsa2xxx_enclosures >= MAX_MSA2XXX_ENCLOSURES) {
+               dev_warn(&h->pdev->dev, "Maximum number of MSA2XXX "
+                       "enclosures exceeded.  Check your hardware "
+                       "configuration.");
+               return 0;
+       }
+
+       memset(scsi3addr, 0, 8);
+       scsi3addr[3] = target;
+       if (hpsa_update_device_info(h, scsi3addr, this_device))
+               return 0;
+       (*nmsa2xxx_enclosures)++;
+       hpsa_set_bus_target_lun(this_device, bus, target, 0);
+       set_bit(target, lunzerobits);
+       return 1;
+}
+
+/*
+ * Do CISS_REPORT_PHYS and CISS_REPORT_LOG.  Data is returned in physdev,
+ * logdev.  The number of luns in physdev and logdev are returned in
+ * *nphysicals and *nlogicals, respectively.
+ * Returns 0 on success, -1 otherwise.
+ */
+static int hpsa_gather_lun_info(struct ctlr_info *h,
+       int reportlunsize,
+       struct ReportLUNdata *physdev, __u32 *nphysicals,
+       struct ReportLUNdata *logdev, __u32 *nlogicals)
+{
+       if (hpsa_scsi_do_report_phys_luns(h, physdev, reportlunsize, 0)) {
+               dev_err(&h->pdev->dev, "report physical LUNs failed.\n");
+               return -1;
+       }
+       memcpy(nphysicals, &physdev->LUNListLength[0], sizeof(*nphysicals));
+       *nphysicals = be32_to_cpu(*nphysicals) / 8;
+#ifdef DEBUG
+       dev_info(&h->pdev->dev, "number of physical luns is %d\n", *nphysicals);
+#endif
+       if (*nphysicals > HPSA_MAX_PHYS_LUN) {
+               dev_warn(&h->pdev->dev, "maximum physical LUNs (%d) exceeded."
+                       "  %d LUNs ignored.\n", HPSA_MAX_PHYS_LUN,
+                       *nphysicals - HPSA_MAX_PHYS_LUN);
+               *nphysicals = HPSA_MAX_PHYS_LUN;
+       }
+       if (hpsa_scsi_do_report_log_luns(h, logdev, reportlunsize)) {
+               dev_err(&h->pdev->dev, "report logical LUNs failed.\n");
+               return -1;
+       }
+       memcpy(nlogicals, &logdev->LUNListLength[0], sizeof(*nlogicals));
+       *nlogicals = be32_to_cpu(*nlogicals) / 8;
+#ifdef DEBUG
+       dev_info(&h->pdev->dev, "number of logical luns is %d\n", *nlogicals);
+#endif
+       /* Reject Logicals in excess of our max capability. */
+       if (*nlogicals > HPSA_MAX_LUN) {
+               dev_warn(&h->pdev->dev,
+                       "maximum logical LUNs (%d) exceeded.  "
+                       "%d LUNs ignored.\n", HPSA_MAX_LUN,
+                       *nlogicals - HPSA_MAX_LUN);
+                       *nlogicals = HPSA_MAX_LUN;
+       }
+       if (*nlogicals + *nphysicals > HPSA_MAX_PHYS_LUN) {
+               dev_warn(&h->pdev->dev,
+                       "maximum logical + physical LUNs (%d) exceeded. "
+                       "%d LUNs ignored.\n", HPSA_MAX_PHYS_LUN,
+                       *nphysicals + *nlogicals - HPSA_MAX_PHYS_LUN);
+               *nlogicals = HPSA_MAX_PHYS_LUN - *nphysicals;
+       }
+       return 0;
+}
+
+static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno)
+{
+       /* the idea here is we could get notified
+        * that some devices have changed, so we do a report
+        * physical luns and report logical luns cmd, and adjust
+        * our list of devices accordingly.
+        *
+        * The scsi3addr's of devices won't change so long as the
+        * adapter is not reset.  That means we can rescan and
+        * tell which devices we already know about, vs. new
+        * devices, vs.  disappearing devices.
+        */
+       struct ReportLUNdata *physdev_list = NULL;
+       struct ReportLUNdata *logdev_list = NULL;
+       unsigned char *inq_buff = NULL;
+       __u32 nphysicals = 0;
+       __u32 nlogicals = 0;
+       __u32 ndev_allocated = 0;
+       struct hpsa_scsi_dev_t **currentsd, *this_device, *tmpdevice;
+       int ncurrent = 0;
+       int reportlunsize = sizeof(*physdev_list) + HPSA_MAX_PHYS_LUN * 8;
+       int i, nmsa2xxx_enclosures, ndevs_to_allocate;
+       int bus, target, lun;
+       DECLARE_BITMAP(lunzerobits, HPSA_MAX_TARGETS_PER_CTLR);
+
+       currentsd = kzalloc(sizeof(*currentsd) * HPSA_MAX_SCSI_DEVS_PER_HBA,
+               GFP_KERNEL);
+       physdev_list = kzalloc(reportlunsize, GFP_KERNEL);
+       logdev_list = kzalloc(reportlunsize, GFP_KERNEL);
+       inq_buff = kmalloc(OBDR_TAPE_INQ_SIZE, GFP_KERNEL);
+       tmpdevice = kzalloc(sizeof(*tmpdevice), GFP_KERNEL);
+
+       if (!currentsd || !physdev_list || !logdev_list ||
+               !inq_buff || !tmpdevice) {
+               dev_err(&h->pdev->dev, "out of memory\n");
+               goto out;
+       }
+       memset(lunzerobits, 0, sizeof(lunzerobits));
+
+       if (hpsa_gather_lun_info(h, reportlunsize, physdev_list, &nphysicals,
+                       logdev_list, &nlogicals))
+               goto out;
+
+       /* We might see up to 32 MSA2xxx enclosures, actually 8 of them
+        * but each of them 4 times through different paths.  The plus 1
+        * is for the RAID controller.
+        */
+       ndevs_to_allocate = nphysicals + nlogicals + MAX_MSA2XXX_ENCLOSURES + 1;
+
+       /* Allocate the per device structures */
+       for (i = 0; i < ndevs_to_allocate; i++) {
+               currentsd[i] = kzalloc(sizeof(*currentsd[i]), GFP_KERNEL);
+               if (!currentsd[i]) {
+                       dev_warn(&h->pdev->dev, "out of memory at %s:%d\n",
+                               __FILE__, __LINE__);
+                       goto out;
+               }
+               ndev_allocated++;
+       }
+
+       /* adjust our table of devices */
+       nmsa2xxx_enclosures = 0;
+       for (i = 0; i < nphysicals + nlogicals + 1; i++) {
+               __u8 *lunaddrbytes;
+
+               /* Figure out where the LUN ID info is coming from */
+               if (i < nphysicals)
+                       lunaddrbytes = &physdev_list->LUN[i][0];
+               else
+                       if (i < nphysicals + nlogicals)
+                               lunaddrbytes =
+                                       &logdev_list->LUN[i-nphysicals][0];
+                       else /* jam in the RAID controller at the end */
+                               lunaddrbytes = RAID_CTLR_LUNID;
+
+               /* skip masked physical devices. */
+               if (lunaddrbytes[3] & 0xC0 && i < nphysicals)
+                       continue;
+
+               /* Get device type, vendor, model, device id */
+               if (hpsa_update_device_info(h, lunaddrbytes, tmpdevice))
+                       continue; /* skip it if we can't talk to it. */
+               figure_bus_target_lun(h, lunaddrbytes, &bus, &target, &lun,
+                       tmpdevice);
+               this_device = currentsd[ncurrent];
+
+               /*
+                * For the msa2xxx boxes, we have to insert a LUN 0 which
+                * doesn't show up in CCISS_REPORT_PHYSICAL data, but there
+                * is nonetheless an enclosure device there.  We have to
+                * present that otherwise linux won't find anything if
+                * there is no lun 0.
+                */
+               if (add_msa2xxx_enclosure_device(h, tmpdevice, this_device,
+                               lunaddrbytes, bus, target, lun, lunzerobits,
+                               &nmsa2xxx_enclosures)) {
+                       ncurrent++;
+                       this_device = currentsd[ncurrent];
+               }
+
+               *this_device = *tmpdevice;
+               hpsa_set_bus_target_lun(this_device, bus, target, lun);
+
+               switch (this_device->devtype) {
+               case TYPE_ROM: {
+                       /* We don't *really* support actual CD-ROM devices,
+                        * just "One Button Disaster Recovery" tape drive
+                        * which temporarily pretends to be a CD-ROM drive.
+                        * So we check that the device is really an OBDR tape
+                        * device by checking for "$DR-10" in bytes 43-48 of
+                        * the inquiry data.
+                        */
+                               char obdr_sig[7];
+#define OBDR_TAPE_SIG "$DR-10"
+                               strncpy(obdr_sig, &inq_buff[43], 6);
+                               obdr_sig[6] = '\0';
+                               if (strncmp(obdr_sig, OBDR_TAPE_SIG, 6) != 0)
+                                       /* Not OBDR device, ignore it. */
+                                       break;
+                       }
+                       ncurrent++;
+                       break;
+               case TYPE_DISK:
+                       if (i < nphysicals)
+                               break;
+                       ncurrent++;
+                       break;
+               case TYPE_TAPE:
+               case TYPE_MEDIUM_CHANGER:
+                       ncurrent++;
+                       break;
+               case TYPE_RAID:
+                       /* Only present the Smartarray HBA as a RAID controller.
+                        * If it's a RAID controller other than the HBA itself
+                        * (an external RAID controller, MSA500 or similar)
+                        * don't present it.
+                        */
+                       if (!is_hba_lunid(lunaddrbytes))
+                               break;
+                       ncurrent++;
+                       break;
+               default:
+                       break;
+               }
+               if (ncurrent >= HPSA_MAX_SCSI_DEVS_PER_HBA)
+                       break;
+       }
+       adjust_hpsa_scsi_table(h, hostno, currentsd, ncurrent);
+out:
+       kfree(tmpdevice);
+       for (i = 0; i < ndev_allocated; i++)
+               kfree(currentsd[i]);
+       kfree(currentsd);
+       kfree(inq_buff);
+       kfree(physdev_list);
+       kfree(logdev_list);
+       return;
+}
+
+/* hpsa_scatter_gather takes a struct scsi_cmnd, (cmd), and does the pci
+ * dma mapping  and fills in the scatter gather entries of the
+ * hpsa command, cp.
+ */
+static int hpsa_scatter_gather(struct pci_dev *pdev,
+               struct CommandList *cp,
+               struct scsi_cmnd *cmd)
+{
+       unsigned int len;
+       struct scatterlist *sg;
+       __u64 addr64;
+       int use_sg, i;
+
+       BUG_ON(scsi_sg_count(cmd) > MAXSGENTRIES);
+
+       use_sg = scsi_dma_map(cmd);
+       if (use_sg < 0)
+               return use_sg;
+
+       if (!use_sg)
+               goto sglist_finished;
+
+       scsi_for_each_sg(cmd, sg, use_sg, i) {
+               addr64 = (__u64) sg_dma_address(sg);
+               len  = sg_dma_len(sg);
+               cp->SG[i].Addr.lower =
+                       (__u32) (addr64 & (__u64) 0x00000000FFFFFFFF);
+               cp->SG[i].Addr.upper =
+                       (__u32) ((addr64 >> 32) & (__u64) 0x00000000FFFFFFFF);
+               cp->SG[i].Len = len;
+               cp->SG[i].Ext = 0;  /* we are not chaining */
+       }
+
+sglist_finished:
+
+       cp->Header.SGList = (__u8) use_sg;   /* no. SGs contig in this cmd */
+       cp->Header.SGTotal = (__u16) use_sg; /* total sgs in this cmd list */
+       return 0;
+}
+
+
+static int hpsa_scsi_queue_command(struct scsi_cmnd *cmd,
+       void (*done)(struct scsi_cmnd *))
+{
+       struct ctlr_info *h;
+       struct hpsa_scsi_dev_t *dev;
+       unsigned char scsi3addr[8];
+       struct CommandList *c;
+       unsigned long flags;
+
+       /* Get the ptr to our adapter structure out of cmd->host. */
+       h = sdev_to_hba(cmd->device);
+       dev = cmd->device->hostdata;
+       if (!dev) {
+               cmd->result = DID_NO_CONNECT << 16;
+               done(cmd);
+               return 0;
+       }
+       memcpy(scsi3addr, dev->scsi3addr, sizeof(scsi3addr));
+
+       /* Need a lock as this is being allocated from the pool */
+       spin_lock_irqsave(&h->lock, flags);
+       c = cmd_alloc(h);
+       spin_unlock_irqrestore(&h->lock, flags);
+       if (c == NULL) {                        /* trouble... */
+               dev_err(&h->pdev->dev, "cmd_alloc returned NULL!\n");
+               return SCSI_MLQUEUE_HOST_BUSY;
+       }
+
+       /* Fill in the command list header */
+
+       cmd->scsi_done = done;    /* save this for use by completion code */
+
+       /* save c in case we have to abort it  */
+       cmd->host_scribble = (unsigned char *) c;
+
+       c->cmd_type = CMD_SCSI;
+       c->scsi_cmd = cmd;
+       c->Header.ReplyQueue = 0;  /* unused in simple mode */
+       memcpy(&c->Header.LUN.LunAddrBytes[0], &scsi3addr[0], 8);
+       c->Header.Tag.lower = c->busaddr;  /* Use k. address of cmd as tag */
+
+       /* Fill in the request block... */
+
+       c->Request.Timeout = 0;
+       memset(c->Request.CDB, 0, sizeof(c->Request.CDB));
+       BUG_ON(cmd->cmd_len > sizeof(c->Request.CDB));
+       c->Request.CDBLen = cmd->cmd_len;
+       memcpy(c->Request.CDB, cmd->cmnd, cmd->cmd_len);
+       c->Request.Type.Type = TYPE_CMD;
+       c->Request.Type.Attribute = ATTR_SIMPLE;
+       switch (cmd->sc_data_direction) {
+       case DMA_TO_DEVICE:
+               c->Request.Type.Direction = XFER_WRITE;
+               break;
+       case DMA_FROM_DEVICE:
+               c->Request.Type.Direction = XFER_READ;
+               break;
+       case DMA_NONE:
+               c->Request.Type.Direction = XFER_NONE;
+               break;
+       case DMA_BIDIRECTIONAL:
+               /* This can happen if a buggy application does a scsi passthru
+                * and sets both inlen and outlen to non-zero. ( see
+                * ../scsi/scsi_ioctl.c:scsi_ioctl_send_command() )
+                */
+
+               c->Request.Type.Direction = XFER_RSVD;
+               /* This is technically wrong, and hpsa controllers should
+                * reject it with CMD_INVALID, which is the most correct
+                * response, but non-fibre backends appear to let it
+                * slide by, and give the same results as if this field
+                * were set correctly.  Either way is acceptable for
+                * our purposes here.
+                */
+
+               break;
+
+       default:
+               dev_err(&h->pdev->dev, "unknown data direction: %d\n",
+                       cmd->sc_data_direction);
+               BUG();
+               break;
+       }
+
+       if (hpsa_scatter_gather(h->pdev, c, cmd) < 0) { /* Fill SG list */
+               cmd_free(h, c);
+               return SCSI_MLQUEUE_HOST_BUSY;
+       }
+       enqueue_cmd_and_start_io(h, c);
+       /* the cmd'll come back via intr handler in complete_scsi_command()  */
+       return 0;
+}
+
+static void hpsa_unregister_scsi(struct ctlr_info *h)
+{
+       /* we are being forcibly unloaded, and may not refuse. */
+       scsi_remove_host(h->scsi_host);
+       scsi_host_put(h->scsi_host);
+       h->scsi_host = NULL;
+}
+
+static int hpsa_register_scsi(struct ctlr_info *h)
+{
+       int rc;
+
+       hpsa_update_scsi_devices(h, -1);
+       rc = hpsa_scsi_detect(h);
+       if (rc != 0)
+               dev_err(&h->pdev->dev, "hpsa_register_scsi: failed"
+                       " hpsa_scsi_detect(), rc is %d\n", rc);
+       return rc;
+}
+
+static int wait_for_device_to_become_ready(struct ctlr_info *h,
+       unsigned char lunaddr[])
+{
+       int rc = 0;
+       int count = 0;
+       int waittime = 1; /* seconds */
+       struct CommandList *c;
+
+       c = cmd_special_alloc(h);
+       if (!c) {
+               dev_warn(&h->pdev->dev, "out of memory in "
+                       "wait_for_device_to_become_ready.\n");
+               return IO_ERROR;
+       }
+
+       /* Send test unit ready until device ready, or give up. */
+       while (count < HPSA_TUR_RETRY_LIMIT) {
+
+               /* Wait for a bit.  do this first, because if we send
+                * the TUR right away, the reset will just abort it.
+                */
+               msleep(1000 * waittime);
+               count++;
+
+               /* Increase wait time with each try, up to a point. */
+               if (waittime < HPSA_MAX_WAIT_INTERVAL_SECS)
+                       waittime = waittime * 2;
+
+               /* Send the Test Unit Ready */
+               fill_cmd(c, TEST_UNIT_READY, h, NULL, 0, 0, lunaddr, TYPE_CMD);
+               hpsa_scsi_do_simple_cmd_core(h, c);
+               /* no unmap needed here because no data xfer. */
+
+               if (c->err_info->CommandStatus == CMD_SUCCESS)
+                       break;
+
+               if (c->err_info->CommandStatus == CMD_TARGET_STATUS &&
+                       c->err_info->ScsiStatus == SAM_STAT_CHECK_CONDITION &&
+                       (c->err_info->SenseInfo[2] == NO_SENSE ||
+                       c->err_info->SenseInfo[2] == UNIT_ATTENTION))
+                       break;
+
+               dev_warn(&h->pdev->dev, "waiting %d secs "
+                       "for device to become ready.\n", waittime);
+               rc = 1; /* device not ready. */
+       }
+
+       if (rc)
+               dev_warn(&h->pdev->dev, "giving up on device.\n");
+       else
+               dev_warn(&h->pdev->dev, "device is ready.\n");
+
+       cmd_special_free(h, c);
+       return rc;
+}
+
+/* Need at least one of these error handlers to keep ../scsi/hosts.c from
+ * complaining.  Doing a host- or bus-reset can't do anything good here.
+ */
+static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd)
+{
+       int rc;
+       struct ctlr_info *h;
+       struct hpsa_scsi_dev_t *dev;
+
+       /* find the controller to which the command to be aborted was sent */
+       h = sdev_to_hba(scsicmd->device);
+       if (h == NULL) /* paranoia */
+               return FAILED;
+       dev_warn(&h->pdev->dev, "resetting drive\n");
+
+       dev = scsicmd->device->hostdata;
+       if (!dev) {
+               dev_err(&h->pdev->dev, "hpsa_eh_device_reset_handler: "
+                       "device lookup failed.\n");
+               return FAILED;
+       }
+       /* send a reset to the SCSI LUN which the command was sent to */
+       rc = hpsa_send_reset(h, dev->scsi3addr);
+       if (rc == 0 && wait_for_device_to_become_ready(h, dev->scsi3addr) == 0)
+               return SUCCESS;
+
+       dev_warn(&h->pdev->dev, "resetting device failed.\n");
+       return FAILED;
+}
+
+/*
+ * For operations that cannot sleep, a command block is allocated at init,
+ * and managed by cmd_alloc() and cmd_free() using a simple bitmap to track
+ * which ones are free or in use.  Lock must be held when calling this.
+ * cmd_free() is the complement.
+ */
+static struct CommandList *cmd_alloc(struct ctlr_info *h)
+{
+       struct CommandList *c;
+       int i;
+       union u64bit temp64;
+       dma_addr_t cmd_dma_handle, err_dma_handle;
+
+       do {
+               i = find_first_zero_bit(h->cmd_pool_bits, h->nr_cmds);
+               if (i == h->nr_cmds)
+                       return NULL;
+       } while (test_and_set_bit
+                (i & (BITS_PER_LONG - 1),
+                 h->cmd_pool_bits + (i / BITS_PER_LONG)) != 0);
+       c = h->cmd_pool + i;
+       memset(c, 0, sizeof(*c));
+       cmd_dma_handle = h->cmd_pool_dhandle
+           + i * sizeof(*c);
+       c->err_info = h->errinfo_pool + i;
+       memset(c->err_info, 0, sizeof(*c->err_info));
+       err_dma_handle = h->errinfo_pool_dhandle
+           + i * sizeof(*c->err_info);
+       h->nr_allocs++;
+
+       c->cmdindex = i;
+
+       INIT_HLIST_NODE(&c->list);
+       c->busaddr = (__u32) cmd_dma_handle;
+       temp64.val = (__u64) err_dma_handle;
+       c->ErrDesc.Addr.lower = temp64.val32.lower;
+       c->ErrDesc.Addr.upper = temp64.val32.upper;
+       c->ErrDesc.Len = sizeof(*c->err_info);
+
+       c->h = h;
+       return c;
+}
+
+/* For operations that can wait for kmalloc to possibly sleep,
+ * this routine can be called. Lock need not be held to call
+ * cmd_special_alloc. cmd_special_free() is the complement.
+ */
+static struct CommandList *cmd_special_alloc(struct ctlr_info *h)
+{
+       struct CommandList *c;
+       union u64bit temp64;
+       dma_addr_t cmd_dma_handle, err_dma_handle;
+
+       c = pci_alloc_consistent(h->pdev, sizeof(*c), &cmd_dma_handle);
+       if (c == NULL)
+               return NULL;
+       memset(c, 0, sizeof(*c));
+
+       c->cmdindex = -1;
+
+       c->err_info = pci_alloc_consistent(h->pdev, sizeof(*c->err_info),
+                   &err_dma_handle);
+
+       if (c->err_info == NULL) {
+               pci_free_consistent(h->pdev,
+                       sizeof(*c), c, cmd_dma_handle);
+               return NULL;
+       }
+       memset(c->err_info, 0, sizeof(*c->err_info));
+
+       INIT_HLIST_NODE(&c->list);
+       c->busaddr = (__u32) cmd_dma_handle;
+       temp64.val = (__u64) err_dma_handle;
+       c->ErrDesc.Addr.lower = temp64.val32.lower;
+       c->ErrDesc.Addr.upper = temp64.val32.upper;
+       c->ErrDesc.Len = sizeof(*c->err_info);
+
+       c->h = h;
+       return c;
+}
+
+static void cmd_free(struct ctlr_info *h, struct CommandList *c)
+{
+       int i;
+
+       i = c - h->cmd_pool;
+       clear_bit(i & (BITS_PER_LONG - 1),
+                 h->cmd_pool_bits + (i / BITS_PER_LONG));
+       h->nr_frees++;
+}
+
+static void cmd_special_free(struct ctlr_info *h, struct CommandList *c)
+{
+       union u64bit temp64;
+
+       temp64.val32.lower = c->ErrDesc.Addr.lower;
+       temp64.val32.upper = c->ErrDesc.Addr.upper;
+       pci_free_consistent(h->pdev, sizeof(*c->err_info),
+                           c->err_info, (dma_addr_t) temp64.val);
+       pci_free_consistent(h->pdev, sizeof(*c),
+                           c, (dma_addr_t) c->busaddr);
+}
+
+#ifdef CONFIG_COMPAT
+
+static int do_ioctl(struct scsi_device *dev, int cmd, void *arg)
+{
+       int ret;
+
+       lock_kernel();
+       ret = hpsa_ioctl(dev, cmd, arg);
+       unlock_kernel();
+       return ret;
+}
+
+static int hpsa_ioctl32_passthru(struct scsi_device *dev, int cmd, void *arg);
+static int hpsa_ioctl32_big_passthru(struct scsi_device *dev,
+       int cmd, void *arg);
+
+static int hpsa_compat_ioctl(struct scsi_device *dev, int cmd, void *arg)
+{
+       switch (cmd) {
+       case CCISS_GETPCIINFO:
+       case CCISS_GETINTINFO:
+       case CCISS_SETINTINFO:
+       case CCISS_GETNODENAME:
+       case CCISS_SETNODENAME:
+       case CCISS_GETHEARTBEAT:
+       case CCISS_GETBUSTYPES:
+       case CCISS_GETFIRMVER:
+       case CCISS_GETDRIVVER:
+       case CCISS_REVALIDVOLS:
+       case CCISS_DEREGDISK:
+       case CCISS_REGNEWDISK:
+       case CCISS_REGNEWD:
+       case CCISS_RESCANDISK:
+       case CCISS_GETLUNINFO:
+               return do_ioctl(dev, cmd, arg);
+
+       case CCISS_PASSTHRU32:
+               return hpsa_ioctl32_passthru(dev, cmd, arg);
+       case CCISS_BIG_PASSTHRU32:
+               return hpsa_ioctl32_big_passthru(dev, cmd, arg);
+
+       default:
+               return -ENOIOCTLCMD;
+       }
+}
+
+static int hpsa_ioctl32_passthru(struct scsi_device *dev, int cmd, void *arg)
+{
+       IOCTL32_Command_struct __user *arg32 =
+           (IOCTL32_Command_struct __user *) arg;
+       IOCTL_Command_struct arg64;
+       IOCTL_Command_struct __user *p = compat_alloc_user_space(sizeof(arg64));
+       int err;
+       u32 cp;
+
+       err = 0;
+       err |= copy_from_user(&arg64.LUN_info, &arg32->LUN_info,
+                          sizeof(arg64.LUN_info));
+       err |= copy_from_user(&arg64.Request, &arg32->Request,
+                          sizeof(arg64.Request));
+       err |= copy_from_user(&arg64.error_info, &arg32->error_info,
+                          sizeof(arg64.error_info));
+       err |= get_user(arg64.buf_size, &arg32->buf_size);
+       err |= get_user(cp, &arg32->buf);
+       arg64.buf = compat_ptr(cp);
+       err |= copy_to_user(p, &arg64, sizeof(arg64));
+
+       if (err)
+               return -EFAULT;
+
+       err = do_ioctl(dev, CCISS_PASSTHRU, (void *)p);
+       if (err)
+               return err;
+       err |= copy_in_user(&arg32->error_info, &p->error_info,
+                        sizeof(arg32->error_info));
+       if (err)
+               return -EFAULT;
+       return err;
+}
+
+static int hpsa_ioctl32_big_passthru(struct scsi_device *dev,
+       int cmd, void *arg)
+{
+       BIG_IOCTL32_Command_struct __user *arg32 =
+           (BIG_IOCTL32_Command_struct __user *) arg;
+       BIG_IOCTL_Command_struct arg64;
+       BIG_IOCTL_Command_struct __user *p =
+           compat_alloc_user_space(sizeof(arg64));
+       int err;
+       u32 cp;
+
+       err = 0;
+       err |= copy_from_user(&arg64.LUN_info, &arg32->LUN_info,
+                          sizeof(arg64.LUN_info));
+       err |= copy_from_user(&arg64.Request, &arg32->Request,
+                          sizeof(arg64.Request));
+       err |= copy_from_user(&arg64.error_info, &arg32->error_info,
+                          sizeof(arg64.error_info));
+       err |= get_user(arg64.buf_size, &arg32->buf_size);
+       err |= get_user(arg64.malloc_size, &arg32->malloc_size);
+       err |= get_user(cp, &arg32->buf);
+       arg64.buf = compat_ptr(cp);
+       err |= copy_to_user(p, &arg64, sizeof(arg64));
+
+       if (err)
+               return -EFAULT;
+
+       err = do_ioctl(dev, CCISS_BIG_PASSTHRU, (void *)p);
+       if (err)
+               return err;
+       err |= copy_in_user(&arg32->error_info, &p->error_info,
+                        sizeof(arg32->error_info));
+       if (err)
+               return -EFAULT;
+       return err;
+}
+#endif
+
+static int hpsa_getpciinfo_ioctl(struct ctlr_info *h, void __user *argp)
+{
+       struct hpsa_pci_info pciinfo;
+
+       if (!argp)
+               return -EINVAL;
+       pciinfo.domain = pci_domain_nr(h->pdev->bus);
+       pciinfo.bus = h->pdev->bus->number;
+       pciinfo.dev_fn = h->pdev->devfn;
+       pciinfo.board_id = h->board_id;
+       if (copy_to_user(argp, &pciinfo, sizeof(pciinfo)))
+               return -EFAULT;
+       return 0;
+}
+
+static int hpsa_getdrivver_ioctl(struct ctlr_info *h, void __user *argp)
+{
+       DriverVer_type DriverVer;
+       unsigned char vmaj, vmin, vsubmin;
+       int rc;
+
+       rc = sscanf(HPSA_DRIVER_VERSION, "%hhu.%hhu.%hhu",
+               &vmaj, &vmin, &vsubmin);
+       if (rc != 3) {
+               dev_info(&h->pdev->dev, "driver version string '%s' "
+                       "unrecognized.", HPSA_DRIVER_VERSION);
+               vmaj = 0;
+               vmin = 0;
+               vsubmin = 0;
+       }
+       DriverVer = (vmaj << 16) | (vmin << 8) | vsubmin;
+       if (!argp)
+               return -EINVAL;
+       if (copy_to_user(argp, &DriverVer, sizeof(DriverVer_type)))
+               return -EFAULT;
+       return 0;
+}
+
+static int hpsa_passthru_ioctl(struct ctlr_info *h, void __user *argp)
+{
+       IOCTL_Command_struct iocommand;
+       struct CommandList *c;
+       char *buff = NULL;
+       union u64bit temp64;
+
+       if (!argp)
+               return -EINVAL;
+       if (!capable(CAP_SYS_RAWIO))
+               return -EPERM;
+       if (copy_from_user(&iocommand, argp, sizeof(iocommand)))
+               return -EFAULT;
+       if ((iocommand.buf_size < 1) &&
+           (iocommand.Request.Type.Direction != XFER_NONE)) {
+               return -EINVAL;
+       }
+       if (iocommand.buf_size > 0) {
+               buff = kmalloc(iocommand.buf_size, GFP_KERNEL);
+               if (buff == NULL)
+                       return -EFAULT;
+       }
+       if (iocommand.Request.Type.Direction == XFER_WRITE) {
+               /* Copy the data into the buffer we created */
+               if (copy_from_user(buff, iocommand.buf, iocommand.buf_size)) {
+                       kfree(buff);
+                       return -EFAULT;
+               }
+       } else
+               memset(buff, 0, iocommand.buf_size);
+       c = cmd_special_alloc(h);
+       if (c == NULL) {
+               kfree(buff);
+               return -ENOMEM;
+       }
+       /* Fill in the command type */
+       c->cmd_type = CMD_IOCTL_PEND;
+       /* Fill in Command Header */
+       c->Header.ReplyQueue = 0; /* unused in simple mode */
+       if (iocommand.buf_size > 0) {   /* buffer to fill */
+               c->Header.SGList = 1;
+               c->Header.SGTotal = 1;
+       } else  { /* no buffers to fill */
+               c->Header.SGList = 0;
+               c->Header.SGTotal = 0;
+       }
+       memcpy(&c->Header.LUN, &iocommand.LUN_info, sizeof(c->Header.LUN));
+       /* use the kernel address the cmd block for tag */
+       c->Header.Tag.lower = c->busaddr;
+
+       /* Fill in Request block */
+       memcpy(&c->Request, &iocommand.Request,
+               sizeof(c->Request));
+
+       /* Fill in the scatter gather information */
+       if (iocommand.buf_size > 0) {
+               temp64.val = pci_map_single(h->pdev, buff,
+                       iocommand.buf_size, PCI_DMA_BIDIRECTIONAL);
+               c->SG[0].Addr.lower = temp64.val32.lower;
+               c->SG[0].Addr.upper = temp64.val32.upper;
+               c->SG[0].Len = iocommand.buf_size;
+               c->SG[0].Ext = 0; /* we are not chaining*/
+       }
+       hpsa_scsi_do_simple_cmd_core(h, c);
+       hpsa_pci_unmap(h->pdev, c, 1, PCI_DMA_BIDIRECTIONAL);
+       check_ioctl_unit_attention(h, c);
+
+       /* Copy the error information out */
+       memcpy(&iocommand.error_info, c->err_info,
+               sizeof(iocommand.error_info));
+       if (copy_to_user(argp, &iocommand, sizeof(iocommand))) {
+               kfree(buff);
+               cmd_special_free(h, c);
+               return -EFAULT;
+       }
+
+       if (iocommand.Request.Type.Direction == XFER_READ) {
+               /* Copy the data out of the buffer we created */
+               if (copy_to_user(iocommand.buf, buff, iocommand.buf_size)) {
+                       kfree(buff);
+                       cmd_special_free(h, c);
+                       return -EFAULT;
+               }
+       }
+       kfree(buff);
+       cmd_special_free(h, c);
+       return 0;
+}
+
+static int hpsa_big_passthru_ioctl(struct ctlr_info *h, void __user *argp)
+{
+       BIG_IOCTL_Command_struct *ioc;
+       struct CommandList *c;
+       unsigned char **buff = NULL;
+       int *buff_size = NULL;
+       union u64bit temp64;
+       BYTE sg_used = 0;
+       int status = 0;
+       int i;
+       __u32 left;
+       __u32 sz;
+       BYTE __user *data_ptr;
+
+       if (!argp)
+               return -EINVAL;
+       if (!capable(CAP_SYS_RAWIO))
+               return -EPERM;
+       ioc = (BIG_IOCTL_Command_struct *)
+           kmalloc(sizeof(*ioc), GFP_KERNEL);
+       if (!ioc) {
+               status = -ENOMEM;
+               goto cleanup1;
+       }
+       if (copy_from_user(ioc, argp, sizeof(*ioc))) {
+               status = -EFAULT;
+               goto cleanup1;
+       }
+       if ((ioc->buf_size < 1) &&
+           (ioc->Request.Type.Direction != XFER_NONE)) {
+               status = -EINVAL;
+               goto cleanup1;
+       }
+       /* Check kmalloc limits  using all SGs */
+       if (ioc->malloc_size > MAX_KMALLOC_SIZE) {
+               status = -EINVAL;
+               goto cleanup1;
+       }
+       if (ioc->buf_size > ioc->malloc_size * MAXSGENTRIES) {
+               status = -EINVAL;
+               goto cleanup1;
+       }
+       buff = kzalloc(MAXSGENTRIES * sizeof(char *), GFP_KERNEL);
+       if (!buff) {
+               status = -ENOMEM;
+               goto cleanup1;
+       }
+       buff_size = kmalloc(MAXSGENTRIES * sizeof(int), GFP_KERNEL);
+       if (!buff_size) {
+               status = -ENOMEM;
+               goto cleanup1;
+       }
+       left = ioc->buf_size;
+       data_ptr = ioc->buf;
+       while (left) {
+               sz = (left > ioc->malloc_size) ? ioc->malloc_size : left;
+               buff_size[sg_used] = sz;
+               buff[sg_used] = kmalloc(sz, GFP_KERNEL);
+               if (buff[sg_used] == NULL) {
+                       status = -ENOMEM;
+                       goto cleanup1;
+               }
+               if (ioc->Request.Type.Direction == XFER_WRITE) {
+                       if (copy_from_user(buff[sg_used], data_ptr, sz)) {
+                               status = -ENOMEM;
+                               goto cleanup1;
+                       }
+               } else
+                       memset(buff[sg_used], 0, sz);
+               left -= sz;
+               data_ptr += sz;
+               sg_used++;
+       }
+       c = cmd_special_alloc(h);
+       if (c == NULL) {
+               status = -ENOMEM;
+               goto cleanup1;
+       }
+       c->cmd_type = CMD_IOCTL_PEND;
+       c->Header.ReplyQueue = 0;
+
+       if (ioc->buf_size > 0) {
+               c->Header.SGList = sg_used;
+               c->Header.SGTotal = sg_used;
+       } else {
+               c->Header.SGList = 0;
+               c->Header.SGTotal = 0;
+       }
+       memcpy(&c->Header.LUN, &ioc->LUN_info, sizeof(c->Header.LUN));
+       c->Header.Tag.lower = c->busaddr;
+       memcpy(&c->Request, &ioc->Request, sizeof(c->Request));
+       if (ioc->buf_size > 0) {
+               int i;
+               for (i = 0; i < sg_used; i++) {
+                       temp64.val = pci_map_single(h->pdev, buff[i],
+                                   buff_size[i], PCI_DMA_BIDIRECTIONAL);
+                       c->SG[i].Addr.lower = temp64.val32.lower;
+                       c->SG[i].Addr.upper = temp64.val32.upper;
+                       c->SG[i].Len = buff_size[i];
+                       /* we are not chaining */
+                       c->SG[i].Ext = 0;
+               }
+       }
+       hpsa_scsi_do_simple_cmd_core(h, c);
+       hpsa_pci_unmap(h->pdev, c, sg_used, PCI_DMA_BIDIRECTIONAL);
+       check_ioctl_unit_attention(h, c);
+       /* Copy the error information out */
+       memcpy(&ioc->error_info, c->err_info, sizeof(ioc->error_info));
+       if (copy_to_user(argp, ioc, sizeof(*ioc))) {
+               cmd_special_free(h, c);
+               status = -EFAULT;
+               goto cleanup1;
+       }
+       if (ioc->Request.Type.Direction == XFER_READ) {
+               /* Copy the data out of the buffer we created */
+               BYTE __user *ptr = ioc->buf;
+               for (i = 0; i < sg_used; i++) {
+                       if (copy_to_user(ptr, buff[i], buff_size[i])) {
+                               cmd_special_free(h, c);
+                               status = -EFAULT;
+                               goto cleanup1;
+                       }
+                       ptr += buff_size[i];
+               }
+       }
+       cmd_special_free(h, c);
+       status = 0;
+cleanup1:
+       if (buff) {
+               for (i = 0; i < sg_used; i++)
+                       kfree(buff[i]);
+               kfree(buff);
+       }
+       kfree(buff_size);
+       kfree(ioc);
+       return status;
+}
+
+static void check_ioctl_unit_attention(struct ctlr_info *h,
+       struct CommandList *c)
+{
+       if (c->err_info->CommandStatus == CMD_TARGET_STATUS &&
+                       c->err_info->ScsiStatus != SAM_STAT_CHECK_CONDITION)
+               (void) check_for_unit_attention(h, c);
+}
+/*
+ * ioctl
+ */
+static int hpsa_ioctl(struct scsi_device *dev, int cmd, void *arg)
+{
+       struct ctlr_info *h;
+       void __user *argp = (void __user *)arg;
+
+       h = sdev_to_hba(dev);
+
+       switch (cmd) {
+       case CCISS_DEREGDISK:
+       case CCISS_REGNEWDISK:
+       case CCISS_REGNEWD:
+               hpsa_update_scsi_devices(h, dev->host->host_no);
+               return 0;
+       case CCISS_GETPCIINFO:
+               return hpsa_getpciinfo_ioctl(h, argp);
+       case CCISS_GETDRIVVER:
+               return hpsa_getdrivver_ioctl(h, argp);
+       case CCISS_PASSTHRU:
+               return hpsa_passthru_ioctl(h, argp);
+       case CCISS_BIG_PASSTHRU:
+               return hpsa_big_passthru_ioctl(h, argp);
+       default:
+               return -ENOTTY;
+       }
+}
+
+static void fill_cmd(struct CommandList *c, __u8 cmd, struct ctlr_info *h,
+       void *buff, size_t size, __u8 page_code, unsigned char *scsi3addr,
+       int cmd_type)
+{
+       int pci_dir = XFER_NONE;
+
+       c->cmd_type = CMD_IOCTL_PEND;
+       c->Header.ReplyQueue = 0;
+       if (buff != NULL && size > 0) {
+               c->Header.SGList = 1;
+               c->Header.SGTotal = 1;
+       } else {
+               c->Header.SGList = 0;
+               c->Header.SGTotal = 0;
+       }
+       c->Header.Tag.lower = c->busaddr;
+       memcpy(c->Header.LUN.LunAddrBytes, scsi3addr, 8);
+
+       c->Request.Type.Type = cmd_type;
+       if (cmd_type == TYPE_CMD) {
+               switch (cmd) {
+               case HPSA_INQUIRY:
+                       /* are we trying to read a vital product page */
+                       if (page_code != 0) {
+                               c->Request.CDB[1] = 0x01;
+                               c->Request.CDB[2] = page_code;
+                       }
+                       c->Request.CDBLen = 6;
+                       c->Request.Type.Attribute = ATTR_SIMPLE;
+                       c->Request.Type.Direction = XFER_READ;
+                       c->Request.Timeout = 0;
+                       c->Request.CDB[0] = HPSA_INQUIRY;
+                       c->Request.CDB[4] = size & 0xFF;
+                       break;
+               case HPSA_REPORT_LOG:
+               case HPSA_REPORT_PHYS:
+                       /* Talking to controller so It's a physical command
+                          mode = 00 target = 0.  Nothing to write.
+                        */
+                       c->Request.CDBLen = 12;
+                       c->Request.Type.Attribute = ATTR_SIMPLE;
+                       c->Request.Type.Direction = XFER_READ;
+                       c->Request.Timeout = 0;
+                       c->Request.CDB[0] = cmd;
+                       c->Request.CDB[6] = (size >> 24) & 0xFF; /* MSB */
+                       c->Request.CDB[7] = (size >> 16) & 0xFF;
+                       c->Request.CDB[8] = (size >> 8) & 0xFF;
+                       c->Request.CDB[9] = size & 0xFF;
+                       break;
+
+               case HPSA_READ_CAPACITY:
+                       c->Request.CDBLen = 10;
+                       c->Request.Type.Attribute = ATTR_SIMPLE;
+                       c->Request.Type.Direction = XFER_READ;
+                       c->Request.Timeout = 0;
+                       c->Request.CDB[0] = cmd;
+                       break;
+               case HPSA_CACHE_FLUSH:
+                       c->Request.CDBLen = 12;
+                       c->Request.Type.Attribute = ATTR_SIMPLE;
+                       c->Request.Type.Direction = XFER_WRITE;
+                       c->Request.Timeout = 0;
+                       c->Request.CDB[0] = BMIC_WRITE;
+                       c->Request.CDB[6] = BMIC_CACHE_FLUSH;
+                       break;
+               case TEST_UNIT_READY:
+                       c->Request.CDBLen = 6;
+                       c->Request.Type.Attribute = ATTR_SIMPLE;
+                       c->Request.Type.Direction = XFER_NONE;
+                       c->Request.Timeout = 0;
+                       break;
+               default:
+                       dev_warn(&h->pdev->dev, "unknown command 0x%c\n", cmd);
+                       BUG();
+                       return;
+               }
+       } else if (cmd_type == TYPE_MSG) {
+               switch (cmd) {
+
+               case  HPSA_DEVICE_RESET_MSG:
+                       c->Request.CDBLen = 16;
+                       c->Request.Type.Type =  1; /* It is a MSG not a CMD */
+                       c->Request.Type.Attribute = ATTR_SIMPLE;
+                       c->Request.Type.Direction = XFER_NONE;
+                       c->Request.Timeout = 0; /* Don't time out */
+                       c->Request.CDB[0] =  0x01; /* RESET_MSG is 0x01 */
+                       c->Request.CDB[1] = 0x03;  /* Reset target above */
+                       /* If bytes 4-7 are zero, it means reset the */
+                       /* LunID device */
+                       c->Request.CDB[4] = 0x00;
+                       c->Request.CDB[5] = 0x00;
+                       c->Request.CDB[6] = 0x00;
+                       c->Request.CDB[7] = 0x00;
+               break;
+
+               default:
+                       dev_warn(&h->pdev->dev, "unknown message type %d\n",
+                               cmd);
+                       BUG();
+               }
+       } else {
+               dev_warn(&h->pdev->dev, "unknown command type %d\n", cmd_type);
+               BUG();
+       }
+
+       switch (c->Request.Type.Direction) {
+       case XFER_READ:
+               pci_dir = PCI_DMA_FROMDEVICE;
+               break;
+       case XFER_WRITE:
+               pci_dir = PCI_DMA_TODEVICE;
+               break;
+       case XFER_NONE:
+               pci_dir = PCI_DMA_NONE;
+               break;
+       default:
+               pci_dir = PCI_DMA_BIDIRECTIONAL;
+       }
+
+       hpsa_map_one(h->pdev, c, buff, size, pci_dir);
+
+       return;
+}
+
+/*
+ * Map (physical) PCI mem into (virtual) kernel space
+ */
+static void __iomem *remap_pci_mem(ulong base, ulong size)
+{
+       ulong page_base = ((ulong) base) & PAGE_MASK;
+       ulong page_offs = ((ulong) base) - page_base;
+       void __iomem *page_remapped = ioremap(page_base, page_offs + size);
+
+       return page_remapped ? (page_remapped + page_offs) : NULL;
+}
+
+/* Takes cmds off the submission queue and sends them to the hardware,
+ * then puts them on the queue of cmds waiting for completion.
+ */
+static void start_io(struct ctlr_info *h)
+{
+       struct CommandList *c;
+
+       while (!hlist_empty(&h->reqQ)) {
+               c = hlist_entry(h->reqQ.first, struct CommandList, list);
+               /* can't do anything if fifo is full */
+               if ((h->access.fifo_full(h))) {
+                       dev_warn(&h->pdev->dev, "fifo full\n");
+                       break;
+               }
+
+               /* Get the first entry from the Request Q */
+               removeQ(c);
+               h->Qdepth--;
+
+               /* Tell the controller execute command */
+               h->access.submit_command(h, c);
+
+               /* Put job onto the completed Q */
+               addQ(&h->cmpQ, c);
+       }
+}
+
+static inline unsigned long get_next_completion(struct ctlr_info *h)
+{
+       return h->access.command_completed(h);
+}
+
+static inline int interrupt_pending(struct ctlr_info *h)
+{
+       return h->access.intr_pending(h);
+}
+
+static inline long interrupt_not_for_us(struct ctlr_info *h)
+{
+       return ((h->access.intr_pending(h) == 0) ||
+                (h->interrupts_enabled == 0));
+}
+
+static inline int bad_tag(struct ctlr_info *h, __u32 tag_index,
+       __u32 raw_tag)
+{
+       if (unlikely(tag_index >= h->nr_cmds)) {
+               dev_warn(&h->pdev->dev, "bad tag 0x%08x ignored.\n", raw_tag);
+               return 1;
+       }
+       return 0;
+}
+
+static inline void finish_cmd(struct CommandList *c, __u32 raw_tag)
+{
+       removeQ(c);
+       if (likely(c->cmd_type == CMD_SCSI))
+               complete_scsi_command(c, 0, raw_tag);
+       else if (c->cmd_type == CMD_IOCTL_PEND)
+               complete(c->waiting);
+}
+
+static irqreturn_t do_hpsa_intr(int irq, void *dev_id)
+{
+       struct ctlr_info *h = dev_id;
+       struct CommandList *c;
+       unsigned long flags;
+       __u32 raw_tag, tag, tag_index;
+       struct hlist_node *tmp;
+
+       if (interrupt_not_for_us(h))
+               return IRQ_NONE;
+       spin_lock_irqsave(&h->lock, flags);
+       while (interrupt_pending(h)) {
+               while ((raw_tag = get_next_completion(h)) != FIFO_EMPTY) {
+                       if (likely(HPSA_TAG_CONTAINS_INDEX(raw_tag))) {
+                               tag_index = HPSA_TAG_TO_INDEX(raw_tag);
+                               if (bad_tag(h, tag_index, raw_tag))
+                                       return IRQ_HANDLED;
+                               c = h->cmd_pool + tag_index;
+                               finish_cmd(c, raw_tag);
+                               continue;
+                       }
+                       tag = HPSA_TAG_DISCARD_ERROR_BITS(raw_tag);
+                       c = NULL;
+                       hlist_for_each_entry(c, tmp, &h->cmpQ, list) {
+                               if (c->busaddr == tag) {
+                                       finish_cmd(c, raw_tag);
+                                       break;
+                               }
+                       }
+               }
+       }
+       spin_unlock_irqrestore(&h->lock, flags);
+       return IRQ_HANDLED;
+}
+
+/* Send a message CDB to the firmware. */
+static __devinit int hpsa_message(struct pci_dev *pdev, unsigned char opcode,
+                                               unsigned char type)
+{
+       struct Command {
+               struct CommandListHeader CommandHeader;
+               struct RequestBlock Request;
+               struct ErrDescriptor ErrorDescriptor;
+       };
+       struct Command *cmd;
+       static const size_t cmd_sz = sizeof(*cmd) +
+                                       sizeof(cmd->ErrorDescriptor);
+       dma_addr_t paddr64;
+       uint32_t paddr32, tag;
+       void __iomem *vaddr;
+       int i, err;
+
+       vaddr = pci_ioremap_bar(pdev, 0);
+       if (vaddr == NULL)
+               return -ENOMEM;
+
+       /* The Inbound Post Queue only accepts 32-bit physical addresses for the
+        * CCISS commands, so they must be allocated from the lower 4GiB of
+        * memory.
+        */
+       err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
+       if (err) {
+               iounmap(vaddr);
+               return -ENOMEM;
+       }
+
+       cmd = pci_alloc_consistent(pdev, cmd_sz, &paddr64);
+       if (cmd == NULL) {
+               iounmap(vaddr);
+               return -ENOMEM;
+       }
+
+       /* This must fit, because of the 32-bit consistent DMA mask.  Also,
+        * although there's no guarantee, we assume that the address is at
+        * least 4-byte aligned (most likely, it's page-aligned).
+        */
+       paddr32 = paddr64;
+
+       cmd->CommandHeader.ReplyQueue = 0;
+       cmd->CommandHeader.SGList = 0;
+       cmd->CommandHeader.SGTotal = 0;
+       cmd->CommandHeader.Tag.lower = paddr32;
+       cmd->CommandHeader.Tag.upper = 0;
+       memset(&cmd->CommandHeader.LUN.LunAddrBytes, 0, 8);
+
+       cmd->Request.CDBLen = 16;
+       cmd->Request.Type.Type = TYPE_MSG;
+       cmd->Request.Type.Attribute = ATTR_HEADOFQUEUE;
+       cmd->Request.Type.Direction = XFER_NONE;
+       cmd->Request.Timeout = 0; /* Don't time out */
+       cmd->Request.CDB[0] = opcode;
+       cmd->Request.CDB[1] = type;
+       memset(&cmd->Request.CDB[2], 0, 14); /* rest of the CDB is reserved */
+       cmd->ErrorDescriptor.Addr.lower = paddr32 + sizeof(*cmd);
+       cmd->ErrorDescriptor.Addr.upper = 0;
+       cmd->ErrorDescriptor.Len = sizeof(struct ErrorInfo);
+
+       writel(paddr32, vaddr + SA5_REQUEST_PORT_OFFSET);
+
+       for (i = 0; i < HPSA_MSG_SEND_RETRY_LIMIT; i++) {
+               tag = readl(vaddr + SA5_REPLY_PORT_OFFSET);
+               if (HPSA_TAG_DISCARD_ERROR_BITS(tag) == paddr32)
+                       break;
+               msleep(HPSA_MSG_SEND_RETRY_INTERVAL_MSECS);
+       }
+
+       iounmap(vaddr);
+
+       /* we leak the DMA buffer here ... no choice since the controller could
+        *  still complete the command.
+        */
+       if (i == HPSA_MSG_SEND_RETRY_LIMIT) {
+               dev_err(&pdev->dev, "controller message %02x:%02x timed out\n",
+                       opcode, type);
+               return -ETIMEDOUT;
+       }
+
+       pci_free_consistent(pdev, cmd_sz, cmd, paddr64);
+
+       if (tag & HPSA_ERROR_BIT) {
+               dev_err(&pdev->dev, "controller message %02x:%02x failed\n",
+                       opcode, type);
+               return -EIO;
+       }
+
+       dev_info(&pdev->dev, "controller message %02x:%02x succeeded\n",
+               opcode, type);
+       return 0;
+}
+
+#define hpsa_soft_reset_controller(p) hpsa_message(p, 1, 0)
+#define hpsa_noop(p) hpsa_message(p, 3, 0)
+
+static __devinit int hpsa_reset_msi(struct pci_dev *pdev)
+{
+/* the #defines are stolen from drivers/pci/msi.h. */
+#define msi_control_reg(base)          (base + PCI_MSI_FLAGS)
+#define PCI_MSIX_FLAGS_ENABLE          (1 << 15)
+
+       int pos;
+       u16 control = 0;
+
+       pos = pci_find_capability(pdev, PCI_CAP_ID_MSI);
+       if (pos) {
+               pci_read_config_word(pdev, msi_control_reg(pos), &control);
+               if (control & PCI_MSI_FLAGS_ENABLE) {
+                       dev_info(&pdev->dev, "resetting MSI\n");
+                       pci_write_config_word(pdev, msi_control_reg(pos),
+                                       control & ~PCI_MSI_FLAGS_ENABLE);
+               }
+       }
+
+       pos = pci_find_capability(pdev, PCI_CAP_ID_MSIX);
+       if (pos) {
+               pci_read_config_word(pdev, msi_control_reg(pos), &control);
+               if (control & PCI_MSIX_FLAGS_ENABLE) {
+                       dev_info(&pdev->dev, "resetting MSI-X\n");
+                       pci_write_config_word(pdev, msi_control_reg(pos),
+                                       control & ~PCI_MSIX_FLAGS_ENABLE);
+               }
+       }
+
+       return 0;
+}
+
+/* This does a hard reset of the controller using PCI power management
+ * states.
+ */
+static __devinit int hpsa_hard_reset_controller(struct pci_dev *pdev)
+{
+       u16 pmcsr, saved_config_space[32];
+       int i, pos;
+
+       dev_info(&pdev->dev, "using PCI PM to reset controller\n");
+
+       /* This is very nearly the same thing as
+        *
+        * pci_save_state(pci_dev);
+        * pci_set_power_state(pci_dev, PCI_D3hot);
+        * pci_set_power_state(pci_dev, PCI_D0);
+        * pci_restore_state(pci_dev);
+        *
+        * but we can't use these nice canned kernel routines on
+        * kexec, because they also check the MSI/MSI-X state in PCI
+        * configuration space and do the wrong thing when it is
+        * set/cleared.  Also, the pci_save/restore_state functions
+        * violate the ordering requirements for restoring the
+        * configuration space from the CCISS document (see the
+        * comment below).  So we roll our own ....
+        */
+
+       for (i = 0; i < 32; i++)
+               pci_read_config_word(pdev, 2*i, &saved_config_space[i]);
+
+       pos = pci_find_capability(pdev, PCI_CAP_ID_PM);
+       if (pos == 0) {
+               dev_err(&pdev->dev,
+                       "hpsa_reset_controller: PCI PM not supported\n");
+               return -ENODEV;
+       }
+
+       /* Quoting from the Open CISS Specification: "The Power
+        * Management Control/Status Register (CSR) controls the power
+        * state of the device.  The normal operating state is D0,
+        * CSR=00h.  The software off state is D3, CSR=03h.  To reset
+        * the controller, place the interface device in D3 then to
+        * D0, this causes a secondary PCI reset which will reset the
+        * controller."
+        */
+
+       /* enter the D3hot power management state */
+       pci_read_config_word(pdev, pos + PCI_PM_CTRL, &pmcsr);
+       pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
+       pmcsr |= PCI_D3hot;
+       pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr);
+
+       msleep(500);
+
+       /* enter the D0 power management state */
+       pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
+       pmcsr |= PCI_D0;
+       pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr);
+
+       msleep(500);
+
+       /* Restore the PCI configuration space.  The Open CISS
+        * Specification says, "Restore the PCI Configuration
+        * Registers, offsets 00h through 60h. It is important to
+        * restore the command register, 16-bits at offset 04h,
+        * last. Do not restore the configuration status register,
+        * 16-bits at offset 06h."  Note that the offset is 2*i.
+        */
+       for (i = 0; i < 32; i++) {
+               if (i == 2 || i == 3)
+                       continue;
+               pci_write_config_word(pdev, 2*i, saved_config_space[i]);
+       }
+       wmb();
+       pci_write_config_word(pdev, 4, saved_config_space[2]);
+
+       return 0;
+}
+
+/*
+ *  We cannot read the structure directly, for portability we must use
+ *   the io functions.
+ *   This is for debug only.
+ */
+#ifdef HPSA_DEBUG
+static void print_cfg_table(struct device *dev, struct CfgTable *tb)
+{
+       int i;
+       char temp_name[17];
+
+       dev_info(dev, "Controller Configuration information\n");
+       dev_info(dev, "------------------------------------\n");
+       for (i = 0; i < 4; i++)
+               temp_name[i] = readb(&(tb->Signature[i]));
+       temp_name[4] = '\0';
+       dev_info(dev, "   Signature = %s\n", temp_name);
+       dev_info(dev, "   Spec Number = %d\n", readl(&(tb->SpecValence)));
+       dev_info(dev, "   Transport methods supported = 0x%x\n",
+              readl(&(tb->TransportSupport)));
+       dev_info(dev, "   Transport methods active = 0x%x\n",
+              readl(&(tb->TransportActive)));
+       dev_info(dev, "   Requested transport Method = 0x%x\n",
+              readl(&(tb->HostWrite.TransportRequest)));
+       dev_info(dev, "   Coalesce Interrupt Delay = 0x%x\n",
+              readl(&(tb->HostWrite.CoalIntDelay)));
+       dev_info(dev, "   Coalesce Interrupt Count = 0x%x\n",
+              readl(&(tb->HostWrite.CoalIntCount)));
+       dev_info(dev, "   Max outstanding commands = 0x%d\n",
+              readl(&(tb->CmdsOutMax)));
+       dev_info(dev, "   Bus Types = 0x%x\n", readl(&(tb->BusTypes)));
+       for (i = 0; i < 16; i++)
+               temp_name[i] = readb(&(tb->ServerName[i]));
+       temp_name[16] = '\0';
+       dev_info(dev, "   Server Name = %s\n", temp_name);
+       dev_info(dev, "   Heartbeat Counter = 0x%x\n\n\n",
+               readl(&(tb->HeartBeat)));
+}
+#endif                         /* HPSA_DEBUG */
+
+static int find_PCI_BAR_index(struct pci_dev *pdev, unsigned long pci_bar_addr)
+{
+       int i, offset, mem_type, bar_type;
+
+       if (pci_bar_addr == PCI_BASE_ADDRESS_0) /* looking for BAR zero? */
+               return 0;
+       offset = 0;
+       for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) {
+               bar_type = pci_resource_flags(pdev, i) & PCI_BASE_ADDRESS_SPACE;
+               if (bar_type == PCI_BASE_ADDRESS_SPACE_IO)
+                       offset += 4;
+               else {
+                       mem_type = pci_resource_flags(pdev, i) &
+                           PCI_BASE_ADDRESS_MEM_TYPE_MASK;
+                       switch (mem_type) {
+                       case PCI_BASE_ADDRESS_MEM_TYPE_32:
+                       case PCI_BASE_ADDRESS_MEM_TYPE_1M:
+                               offset += 4;    /* 32 bit */
+                               break;
+                       case PCI_BASE_ADDRESS_MEM_TYPE_64:
+                               offset += 8;
+                               break;
+                       default:        /* reserved in PCI 2.2 */
+                               dev_warn(&pdev->dev,
+                                      "base address is invalid\n");
+                               return -1;
+                               break;
+                       }
+               }
+               if (offset == pci_bar_addr - PCI_BASE_ADDRESS_0)
+                       return i + 1;
+       }
+       return -1;
+}
+
+/* If MSI/MSI-X is supported by the kernel we will try to enable it on
+ * controllers that are capable. If not, we use IO-APIC mode.
+ */
+
+static void __devinit hpsa_interrupt_mode(struct ctlr_info *h,
+                                          struct pci_dev *pdev, __u32 board_id)
+{
+#ifdef CONFIG_PCI_MSI
+       int err;
+       struct msix_entry hpsa_msix_entries[4] = { {0, 0}, {0, 1},
+       {0, 2}, {0, 3}
+       };
+
+       /* Some boards advertise MSI but don't really support it */
+       if ((board_id == 0x40700E11) ||
+           (board_id == 0x40800E11) ||
+           (board_id == 0x40820E11) || (board_id == 0x40830E11))
+               goto default_int_mode;
+       if (pci_find_capability(pdev, PCI_CAP_ID_MSIX)) {
+               dev_info(&pdev->dev, "MSIX\n");
+               err = pci_enable_msix(pdev, hpsa_msix_entries, 4);
+               if (!err) {
+                       h->intr[0] = hpsa_msix_entries[0].vector;
+                       h->intr[1] = hpsa_msix_entries[1].vector;
+                       h->intr[2] = hpsa_msix_entries[2].vector;
+                       h->intr[3] = hpsa_msix_entries[3].vector;
+                       h->msix_vector = 1;
+                       return;
+               }
+               if (err > 0) {
+                       dev_warn(&pdev->dev, "only %d MSI-X vectors "
+                              "available\n", err);
+                       goto default_int_mode;
+               } else {
+                       dev_warn(&pdev->dev, "MSI-X init failed %d\n",
+                              err);
+                       goto default_int_mode;
+               }
+       }
+       if (pci_find_capability(pdev, PCI_CAP_ID_MSI)) {
+               dev_info(&pdev->dev, "MSI\n");
+               if (!pci_enable_msi(pdev))
+                       h->msi_vector = 1;
+               else
+                       dev_warn(&pdev->dev, "MSI init failed\n");
+       }
+default_int_mode:
+#endif                         /* CONFIG_PCI_MSI */
+       /* if we get here we're going to use the default interrupt mode */
+       h->intr[SIMPLE_MODE_INT] = pdev->irq;
+       return;
+}
+
+static int hpsa_pci_init(struct ctlr_info *h, struct pci_dev *pdev)
+{
+       ushort subsystem_vendor_id, subsystem_device_id, command;
+       __u32 board_id, scratchpad = 0;
+       __u64 cfg_offset;
+       __u32 cfg_base_addr;
+       __u64 cfg_base_addr_index;
+       int i, prod_index, err;
+
+       subsystem_vendor_id = pdev->subsystem_vendor;
+       subsystem_device_id = pdev->subsystem_device;
+       board_id = (((__u32) (subsystem_device_id << 16) & 0xffff0000) |
+                   subsystem_vendor_id);
+
+       for (i = 0; i < ARRAY_SIZE(products); i++)
+               if (board_id == products[i].board_id)
+                       break;
+
+       prod_index = i;
+
+       if (prod_index == ARRAY_SIZE(products)) {
+               prod_index--;
+               if (subsystem_vendor_id != PCI_VENDOR_ID_HP ||
+                               !hpsa_allow_any) {
+                       dev_warn(&pdev->dev, "unrecognized board ID:"
+                               " 0x%08lx, ignoring.\n",
+                               (unsigned long) board_id);
+                       return -ENODEV;
+               }
+       }
+       /* check to see if controller has been disabled
+        * BEFORE trying to enable it
+        */
+       (void)pci_read_config_word(pdev, PCI_COMMAND, &command);
+       if (!(command & 0x02)) {
+               dev_warn(&pdev->dev, "controller appears to be disabled\n");
+               return -ENODEV;
+       }
+
+       err = pci_enable_device(pdev);
+       if (err) {
+               dev_warn(&pdev->dev, "unable to enable PCI device\n");
+               return err;
+       }
+
+       err = pci_request_regions(pdev, "hpsa");
+       if (err) {
+               dev_err(&pdev->dev, "cannot obtain PCI resources, aborting\n");
+               return err;
+       }
+
+       /* If the kernel supports MSI/MSI-X we will try to enable that,
+        * else we use the IO-APIC interrupt assigned to us by system ROM.
+        */
+       hpsa_interrupt_mode(h, pdev, board_id);
+
+       /* find the memory BAR */
+       for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) {
+               if (pci_resource_flags(pdev, i) & IORESOURCE_MEM)
+                       break;
+       }
+       if (i == DEVICE_COUNT_RESOURCE) {
+               dev_warn(&pdev->dev, "no memory BAR found\n");
+               err = -ENODEV;
+               goto err_out_free_res;
+       }
+
+       h->paddr = pci_resource_start(pdev, i); /* addressing mode bits
+                                                * already removed
+                                                */
+
+       h->vaddr = remap_pci_mem(h->paddr, 0x250);
+
+       /* Wait for the board to become ready.  */
+       for (i = 0; i < HPSA_BOARD_READY_ITERATIONS; i++) {
+               scratchpad = readl(h->vaddr + SA5_SCRATCHPAD_OFFSET);
+               if (scratchpad == HPSA_FIRMWARE_READY)
+                       break;
+               msleep(HPSA_BOARD_READY_POLL_INTERVAL_MSECS);
+       }
+       if (scratchpad != HPSA_FIRMWARE_READY) {
+               dev_warn(&pdev->dev, "board not ready, timed out.\n");
+               err = -ENODEV;
+               goto err_out_free_res;
+       }
+
+       /* get the address index number */
+       cfg_base_addr = readl(h->vaddr + SA5_CTCFG_OFFSET);
+       cfg_base_addr &= (__u32) 0x0000ffff;
+       cfg_base_addr_index = find_PCI_BAR_index(pdev, cfg_base_addr);
+       if (cfg_base_addr_index == -1) {
+               dev_warn(&pdev->dev, "cannot find cfg_base_addr_index\n");
+               err = -ENODEV;
+               goto err_out_free_res;
+       }
+
+       cfg_offset = readl(h->vaddr + SA5_CTMEM_OFFSET);
+       h->cfgtable = remap_pci_mem(pci_resource_start(pdev,
+                              cfg_base_addr_index) + cfg_offset,
+                               sizeof(h->cfgtable));
+       h->board_id = board_id;
+
+       /* Query controller for max supported commands: */
+       h->max_commands = readl(&(h->cfgtable->CmdsOutMax));
+
+       h->product_name = products[prod_index].product_name;
+       h->access = *(products[prod_index].access);
+       /* Allow room for some ioctls */
+       h->nr_cmds = h->max_commands - 4;
+
+       if ((readb(&h->cfgtable->Signature[0]) != 'C') ||
+           (readb(&h->cfgtable->Signature[1]) != 'I') ||
+           (readb(&h->cfgtable->Signature[2]) != 'S') ||
+           (readb(&h->cfgtable->Signature[3]) != 'S')) {
+               dev_warn(&pdev->dev, "not a valid CISS config table\n");
+               err = -ENODEV;
+               goto err_out_free_res;
+       }
+#ifdef CONFIG_X86
+       {
+               /* Need to enable prefetch in the SCSI core for 6400 in x86 */
+               __u32 prefetch;
+               prefetch = readl(&(h->cfgtable->SCSI_Prefetch));
+               prefetch |= 0x100;
+               writel(prefetch, &(h->cfgtable->SCSI_Prefetch));
+       }
+#endif
+
+       /* Disabling DMA prefetch for the P600
+        * An ASIC bug may result in a prefetch beyond
+        * physical memory.
+        */
+       if (board_id == 0x3225103C) {
+               __u32 dma_prefetch;
+               dma_prefetch = readl(h->vaddr + I2O_DMA1_CFG);
+               dma_prefetch |= 0x8000;
+               writel(dma_prefetch, h->vaddr + I2O_DMA1_CFG);
+       }
+
+       h->max_commands = readl(&(h->cfgtable->CmdsOutMax));
+       /* Update the field, and then ring the doorbell */
+       writel(CFGTBL_Trans_Simple, &(h->cfgtable->HostWrite.TransportRequest));
+       writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL);
+
+       /* under certain very rare conditions, this can take awhile.
+        * (e.g.: hot replace a failed 144GB drive in a RAID 5 set right
+        * as we enter this code.)
+        */
+       for (i = 0; i < MAX_CONFIG_WAIT; i++) {
+               if (!(readl(h->vaddr + SA5_DOORBELL) & CFGTBL_ChangeReq))
+                       break;
+               /* delay and try again */
+               msleep(10);
+       }
+
+#ifdef HPSA_DEBUG
+       print_cfg_table(&pdev->dev, h->cfgtable);
+#endif                         /* HPSA_DEBUG */
+
+       if (!(readl(&(h->cfgtable->TransportActive)) & CFGTBL_Trans_Simple)) {
+               dev_warn(&pdev->dev, "unable to get board into simple mode\n");
+               err = -ENODEV;
+               goto err_out_free_res;
+       }
+       return 0;
+
+err_out_free_res:
+       /*
+        * Deliberately omit pci_disable_device(): it does something nasty to
+        * Smart Array controllers that pci_enable_device does not undo
+        */
+       pci_release_regions(pdev);
+       return err;
+}
+
+static int __devinit hpsa_init_one(struct pci_dev *pdev,
+                                   const struct pci_device_id *ent)
+{
+       int i;
+       int dac;
+       struct ctlr_info *h;
+
+       if (number_of_controllers == 0)
+               printk(KERN_INFO DRIVER_NAME "\n");
+       if (reset_devices) {
+               /* Reset the controller with a PCI power-cycle */
+               if (hpsa_hard_reset_controller(pdev) || hpsa_reset_msi(pdev))
+                       return -ENODEV;
+
+               /* Some devices (notably the HP Smart Array 5i Controller)
+                  need a little pause here */
+               msleep(HPSA_POST_RESET_PAUSE_MSECS);
+
+               /* Now try to get the controller to respond to a no-op */
+               for (i = 0; i < HPSA_POST_RESET_NOOP_RETRIES; i++) {
+                       if (hpsa_noop(pdev) == 0)
+                               break;
+                       else
+                               dev_warn(&pdev->dev, "no-op failed%s\n",
+                                               (i < 11 ? "; re-trying" : ""));
+               }
+       }
+
+       BUILD_BUG_ON(sizeof(struct CommandList) % 8);
+       h = kzalloc(sizeof(*h), GFP_KERNEL);
+       if (!h)
+               return -1;
+
+       h->busy_initializing = 1;
+       INIT_HLIST_HEAD(&h->cmpQ);
+       INIT_HLIST_HEAD(&h->reqQ);
+       mutex_init(&h->busy_shutting_down);
+       init_completion(&h->scan_wait);
+       if (hpsa_pci_init(h, pdev) != 0)
+               goto clean1;
+
+       sprintf(h->devname, "hpsa%d", number_of_controllers);
+       h->ctlr = number_of_controllers;
+       number_of_controllers++;
+       h->pdev = pdev;
+
+       /* configure PCI DMA stuff */
+       if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64)))
+               dac = 1;
+       else if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(32)))
+               dac = 0;
+       else {
+               dev_err(&pdev->dev, "no suitable DMA available\n");
+               goto clean1;
+       }
+
+       /* make sure the board interrupts are off */
+       h->access.set_intr_mask(h, HPSA_INTR_OFF);
+       if (request_irq(h->intr[SIMPLE_MODE_INT], do_hpsa_intr,
+                       IRQF_DISABLED | IRQF_SHARED, h->devname, h)) {
+               dev_err(&pdev->dev, "unable to get irq %d for %s\n",
+                      h->intr[SIMPLE_MODE_INT], h->devname);
+               goto clean2;
+       }
+
+       dev_info(&pdev->dev, "%s: <0x%x> at PCI %s IRQ %d%s using DAC\n",
+              h->devname, pdev->device, pci_name(pdev),
+              h->intr[SIMPLE_MODE_INT], dac ? "" : " not");
+
+       h->cmd_pool_bits =
+           kmalloc(((h->nr_cmds + BITS_PER_LONG -
+                     1) / BITS_PER_LONG) * sizeof(unsigned long), GFP_KERNEL);
+       h->cmd_pool = pci_alloc_consistent(h->pdev,
+                   h->nr_cmds * sizeof(*h->cmd_pool),
+                   &(h->cmd_pool_dhandle));
+       h->errinfo_pool = pci_alloc_consistent(h->pdev,
+                   h->nr_cmds * sizeof(*h->errinfo_pool),
+                   &(h->errinfo_pool_dhandle));
+       if ((h->cmd_pool_bits == NULL)
+           || (h->cmd_pool == NULL)
+           || (h->errinfo_pool == NULL)) {
+               dev_err(&pdev->dev, "out of memory");
+               goto clean4;
+       }
+       spin_lock_init(&h->lock);
+
+       pci_set_drvdata(pdev, h);
+       memset(h->cmd_pool_bits, 0,
+              ((h->nr_cmds + BITS_PER_LONG -
+                1) / BITS_PER_LONG) * sizeof(unsigned long));
+
+       hpsa_scsi_setup(h);
+
+       /* Turn the interrupts on so we can service requests */
+       h->access.set_intr_mask(h, HPSA_INTR_ON);
+
+       hpsa_register_scsi(h);  /* hook ourselves into SCSI subsystem */
+       h->busy_initializing = 0;
+       return 1;
+
+clean4:
+       kfree(h->cmd_pool_bits);
+       if (h->cmd_pool)
+               pci_free_consistent(h->pdev,
+                           h->nr_cmds * sizeof(struct CommandList),
+                           h->cmd_pool, h->cmd_pool_dhandle);
+       if (h->errinfo_pool)
+               pci_free_consistent(h->pdev,
+                           h->nr_cmds * sizeof(struct ErrorInfo),
+                           h->errinfo_pool,
+                           h->errinfo_pool_dhandle);
+       free_irq(h->intr[SIMPLE_MODE_INT], h);
+clean2:
+clean1:
+       h->busy_initializing = 0;
+       kfree(h);
+       return -1;
+}
+
+static void hpsa_flush_cache(struct ctlr_info *h)
+{
+       char *flush_buf;
+       struct CommandList *c;
+
+       flush_buf = kzalloc(4, GFP_KERNEL);
+       if (!flush_buf)
+               return;
+
+       c = cmd_special_alloc(h);
+       if (!c) {
+               dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n");
+               goto out_of_memory;
+       }
+       fill_cmd(c, HPSA_CACHE_FLUSH, h, flush_buf, 4, 0,
+               RAID_CTLR_LUNID, TYPE_CMD);
+       hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_TODEVICE);
+       if (c->err_info->CommandStatus != 0)
+               dev_warn(&h->pdev->dev,
+                       "error flushing cache on controller\n");
+       cmd_special_free(h, c);
+out_of_memory:
+       kfree(flush_buf);
+}
+
+static void hpsa_shutdown(struct pci_dev *pdev)
+{
+       struct ctlr_info *h;
+
+       h = pci_get_drvdata(pdev);
+       /* Turn board interrupts off  and send the flush cache command
+        * sendcmd will turn off interrupt, and send the flush...
+        * To write all data in the battery backed cache to disks
+        */
+       hpsa_flush_cache(h);
+       h->access.set_intr_mask(h, HPSA_INTR_OFF);
+       free_irq(h->intr[2], h);
+#ifdef CONFIG_PCI_MSI
+       if (h->msix_vector)
+               pci_disable_msix(h->pdev);
+       else if (h->msi_vector)
+               pci_disable_msi(h->pdev);
+#endif                         /* CONFIG_PCI_MSI */
+}
+
+static void __devexit hpsa_remove_one(struct pci_dev *pdev)
+{
+       struct ctlr_info *h;
+
+       if (pci_get_drvdata(pdev) == NULL) {
+               dev_err(&pdev->dev, "unable to remove device \n");
+               return;
+       }
+       h = pci_get_drvdata(pdev);
+       mutex_lock(&h->busy_shutting_down);
+       remove_from_scan_list(h);
+       hpsa_unregister_scsi(h);        /* unhook from SCSI subsystem */
+       hpsa_shutdown(pdev);
+       iounmap(h->vaddr);
+       pci_free_consistent(h->pdev,
+               h->nr_cmds * sizeof(struct CommandList),
+               h->cmd_pool, h->cmd_pool_dhandle);
+       pci_free_consistent(h->pdev,
+               h->nr_cmds * sizeof(struct ErrorInfo),
+               h->errinfo_pool, h->errinfo_pool_dhandle);
+       kfree(h->cmd_pool_bits);
+       /*
+        * Deliberately omit pci_disable_device(): it does something nasty to
+        * Smart Array controllers that pci_enable_device does not undo
+        */
+       pci_release_regions(pdev);
+       pci_set_drvdata(pdev, NULL);
+       mutex_unlock(&h->busy_shutting_down);
+       kfree(h);
+}
+
+static int hpsa_suspend(__attribute__((unused)) struct pci_dev *pdev,
+       __attribute__((unused)) pm_message_t state)
+{
+       return -ENOSYS;
+}
+
+static int hpsa_resume(__attribute__((unused)) struct pci_dev *pdev)
+{
+       return -ENOSYS;
+}
+
+static struct pci_driver hpsa_pci_driver = {
+       .name = "hpsa",
+       .probe = hpsa_init_one,
+       .remove = __devexit_p(hpsa_remove_one),
+       .id_table = hpsa_pci_device_id, /* id_table */
+       .shutdown = hpsa_shutdown,
+       .suspend = hpsa_suspend,
+       .resume = hpsa_resume,
+};
+
+/*
+ *  This is it.  Register the PCI driver information for the cards we control
+ *  the OS will call our registered routines when it finds one of our cards.
+ */
+static int __init hpsa_init(void)
+{
+       int err;
+       /* Start the scan thread */
+       hpsa_scan_thread = kthread_run(hpsa_scan_func, NULL, "hpsa_scan");
+       if (IS_ERR(hpsa_scan_thread)) {
+               err = PTR_ERR(hpsa_scan_thread);
+               return -ENODEV;
+       }
+       err = pci_register_driver(&hpsa_pci_driver);
+       if (err)
+               kthread_stop(hpsa_scan_thread);
+       return err;
+}
+
+static void __exit hpsa_cleanup(void)
+{
+       pci_unregister_driver(&hpsa_pci_driver);
+       kthread_stop(hpsa_scan_thread);
+}
+
+module_init(hpsa_init);
+module_exit(hpsa_cleanup);
diff --git a/drivers/scsi/hpsa.h b/drivers/scsi/hpsa.h
new file mode 100644 (file)
index 0000000..6bd1949
--- /dev/null
@@ -0,0 +1,273 @@
+/*
+ *    Disk Array driver for HP Smart Array SAS controllers
+ *    Copyright 2000, 2009 Hewlett-Packard Development Company, L.P.
+ *
+ *    This program is free software; you can redistribute it and/or modify
+ *    it under the terms of the GNU General Public License as published by
+ *    the Free Software Foundation; version 2 of the License.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *    NON INFRINGEMENT.  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., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ *    Questions/Comments/Bugfixes to iss_storagedev@hp.com
+ *
+ */
+#ifndef HPSA_H
+#define HPSA_H
+
+#include <scsi/scsicam.h>
+
+#define IO_OK          0
+#define IO_ERROR       1
+
+struct ctlr_info;
+
+struct access_method {
+       void (*submit_command)(struct ctlr_info *h,
+               struct CommandList *c);
+       void (*set_intr_mask)(struct ctlr_info *h, unsigned long val);
+       unsigned long (*fifo_full)(struct ctlr_info *h);
+       unsigned long (*intr_pending)(struct ctlr_info *h);
+       unsigned long (*command_completed)(struct ctlr_info *h);
+};
+
+struct hpsa_scsi_dev_t {
+       int devtype;
+       int bus, target, lun;           /* as presented to the OS */
+       unsigned char scsi3addr[8];     /* as presented to the HW */
+#define RAID_CTLR_LUNID "\0\0\0\0\0\0\0\0"
+       unsigned char device_id[16];    /* from inquiry pg. 0x83 */
+       unsigned char vendor[8];        /* bytes 8-15 of inquiry data */
+       unsigned char model[16];        /* bytes 16-31 of inquiry data */
+       unsigned char revision[4];      /* bytes 32-35 of inquiry data */
+       unsigned char raid_level;       /* from inquiry page 0xC1 */
+};
+
+struct ctlr_info {
+       int     ctlr;
+       char    devname[8];
+       char    *product_name;
+       char    firm_ver[4]; /* Firmware version */
+       struct pci_dev *pdev;
+       __u32   board_id;
+       void __iomem *vaddr;
+       unsigned long paddr;
+       int     nr_cmds; /* Number of commands allowed on this controller */
+       struct CfgTable __iomem *cfgtable;
+       int     interrupts_enabled;
+       int     major;
+       int     max_commands;
+       int     commands_outstanding;
+       int     max_outstanding; /* Debug */
+       int     usage_count;  /* number of opens all all minor devices */
+#      define DOORBELL_INT     0
+#      define PERF_MODE_INT    1
+#      define SIMPLE_MODE_INT  2
+#      define MEMQ_MODE_INT    3
+       unsigned int intr[4];
+       unsigned int msix_vector;
+       unsigned int msi_vector;
+       struct access_method access;
+
+       /* queue and queue Info */
+       struct hlist_head reqQ;
+       struct hlist_head cmpQ;
+       unsigned int Qdepth;
+       unsigned int maxQsinceinit;
+       unsigned int maxSG;
+       spinlock_t lock;
+
+       /* pointers to command and error info pool */
+       struct CommandList      *cmd_pool;
+       dma_addr_t              cmd_pool_dhandle;
+       struct ErrorInfo        *errinfo_pool;
+       dma_addr_t              errinfo_pool_dhandle;
+       unsigned long           *cmd_pool_bits;
+       int                     nr_allocs;
+       int                     nr_frees;
+       int                     busy_initializing;
+       int                     busy_scanning;
+       struct mutex            busy_shutting_down;
+       struct list_head        scan_list;
+       struct completion       scan_wait;
+
+       struct Scsi_Host *scsi_host;
+       spinlock_t devlock; /* to protect hba[ctlr]->dev[];  */
+       int ndevices; /* number of used elements in .dev[] array. */
+#define HPSA_MAX_SCSI_DEVS_PER_HBA 256
+       struct hpsa_scsi_dev_t *dev[HPSA_MAX_SCSI_DEVS_PER_HBA];
+};
+#define HPSA_ABORT_MSG 0
+#define HPSA_DEVICE_RESET_MSG 1
+#define HPSA_BUS_RESET_MSG 2
+#define HPSA_HOST_RESET_MSG 3
+#define HPSA_MSG_SEND_RETRY_LIMIT 10
+#define HPSA_MSG_SEND_RETRY_INTERVAL_MSECS 1000
+
+/* Maximum time in seconds driver will wait for command completions
+ * when polling before giving up.
+ */
+#define HPSA_MAX_POLL_TIME_SECS (20)
+
+/* During SCSI error recovery, HPSA_TUR_RETRY_LIMIT defines
+ * how many times to retry TEST UNIT READY on a device
+ * while waiting for it to become ready before giving up.
+ * HPSA_MAX_WAIT_INTERVAL_SECS is the max wait interval
+ * between sending TURs while waiting for a device
+ * to become ready.
+ */
+#define HPSA_TUR_RETRY_LIMIT (20)
+#define HPSA_MAX_WAIT_INTERVAL_SECS (30)
+
+/* HPSA_BOARD_READY_WAIT_SECS is how long to wait for a board
+ * to become ready, in seconds, before giving up on it.
+ * HPSA_BOARD_READY_POLL_INTERVAL_MSECS * is how long to wait
+ * between polling the board to see if it is ready, in
+ * milliseconds.  HPSA_BOARD_READY_POLL_INTERVAL and
+ * HPSA_BOARD_READY_ITERATIONS are derived from those.
+ */
+#define HPSA_BOARD_READY_WAIT_SECS (120)
+#define HPSA_BOARD_READY_POLL_INTERVAL_MSECS (100)
+#define HPSA_BOARD_READY_POLL_INTERVAL \
+       ((HPSA_BOARD_READY_POLL_INTERVAL_MSECS * HZ) / 1000)
+#define HPSA_BOARD_READY_ITERATIONS \
+       ((HPSA_BOARD_READY_WAIT_SECS * 1000) / \
+               HPSA_BOARD_READY_POLL_INTERVAL_MSECS)
+#define HPSA_POST_RESET_PAUSE_MSECS (3000)
+#define HPSA_POST_RESET_NOOP_RETRIES (12)
+
+/*  Defining the diffent access_menthods */
+/*
+ * Memory mapped FIFO interface (SMART 53xx cards)
+ */
+#define SA5_DOORBELL   0x20
+#define SA5_REQUEST_PORT_OFFSET        0x40
+#define SA5_REPLY_INTR_MASK_OFFSET     0x34
+#define SA5_REPLY_PORT_OFFSET          0x44
+#define SA5_INTR_STATUS                0x30
+#define SA5_SCRATCHPAD_OFFSET  0xB0
+
+#define SA5_CTCFG_OFFSET       0xB4
+#define SA5_CTMEM_OFFSET       0xB8
+
+#define SA5_INTR_OFF           0x08
+#define SA5B_INTR_OFF          0x04
+#define SA5_INTR_PENDING       0x08
+#define SA5B_INTR_PENDING      0x04
+#define FIFO_EMPTY             0xffffffff
+#define HPSA_FIRMWARE_READY    0xffff0000 /* value in scratchpad register */
+
+#define HPSA_ERROR_BIT         0x02
+#define HPSA_TAG_CONTAINS_INDEX(tag) ((tag) & 0x04)
+#define HPSA_TAG_TO_INDEX(tag) ((tag) >> 3)
+#define HPSA_TAG_DISCARD_ERROR_BITS(tag) ((tag) & ~3)
+
+#define HPSA_INTR_ON   1
+#define HPSA_INTR_OFF  0
+/*
+       Send the command to the hardware
+*/
+static void SA5_submit_command(struct ctlr_info *h,
+       struct CommandList *c)
+{
+#ifdef HPSA_DEBUG
+        printk(KERN_WARNING "hpsa: Sending %x - down to controller\n",
+               c->busaddr);
+#endif /* HPSA_DEBUG */
+       writel(c->busaddr, h->vaddr + SA5_REQUEST_PORT_OFFSET);
+       h->commands_outstanding++;
+       if (h->commands_outstanding > h->max_outstanding)
+               h->max_outstanding = h->commands_outstanding;
+}
+
+/*
+ *  This card is the opposite of the other cards.
+ *   0 turns interrupts on...
+ *   0x08 turns them off...
+ */
+static void SA5_intr_mask(struct ctlr_info *h, unsigned long val)
+{
+       if (val) { /* Turn interrupts on */
+               h->interrupts_enabled = 1;
+               writel(0, h->vaddr + SA5_REPLY_INTR_MASK_OFFSET);
+       } else { /* Turn them off */
+               h->interrupts_enabled = 0;
+               writel(SA5_INTR_OFF,
+                       h->vaddr + SA5_REPLY_INTR_MASK_OFFSET);
+       }
+}
+/*
+ *  Returns true if fifo is full.
+ *
+ */
+static unsigned long SA5_fifo_full(struct ctlr_info *h)
+{
+       if (h->commands_outstanding >= h->max_commands)
+               return 1;
+       else
+               return 0;
+
+}
+/*
+ *   returns value read from hardware.
+ *     returns FIFO_EMPTY if there is nothing to read
+ */
+static unsigned long SA5_completed(struct ctlr_info *h)
+{
+       unsigned long register_value
+               = readl(h->vaddr + SA5_REPLY_PORT_OFFSET);
+
+       if (register_value != FIFO_EMPTY)
+               h->commands_outstanding--;
+
+#ifdef HPSA_DEBUG
+       if (register_value != FIFO_EMPTY)
+               printk(KERN_INFO "hpsa:  Read %lx back from board\n",
+                       register_value);
+       else
+               printk(KERN_INFO "hpsa:  FIFO Empty read\n");
+#endif
+
+       return register_value;
+}
+/*
+ *     Returns true if an interrupt is pending..
+ */
+static unsigned long SA5_intr_pending(struct ctlr_info *h)
+{
+       unsigned long register_value  =
+               readl(h->vaddr + SA5_INTR_STATUS);
+#ifdef HPSA_DEBUG
+       printk(KERN_INFO "hpsa: intr_pending %lx\n", register_value);
+#endif  /* HPSA_DEBUG */
+       if (register_value &  SA5_INTR_PENDING)
+               return  1;
+       return 0 ;
+}
+
+
+static struct access_method SA5_access = {
+       SA5_submit_command,
+       SA5_intr_mask,
+       SA5_fifo_full,
+       SA5_intr_pending,
+       SA5_completed,
+};
+
+struct board_type {
+       __u32   board_id;
+       char    *product_name;
+       struct access_method *access;
+};
+
+
+/* end of old hpsa_scsi.h file */
+
+#endif /* HPSA_H */
+
diff --git a/drivers/scsi/hpsa_cmd.h b/drivers/scsi/hpsa_cmd.h
new file mode 100644 (file)
index 0000000..12d7138
--- /dev/null
@@ -0,0 +1,326 @@
+/*
+ *    Disk Array driver for HP Smart Array SAS controllers
+ *    Copyright 2000, 2009 Hewlett-Packard Development Company, L.P.
+ *
+ *    This program is free software; you can redistribute it and/or modify
+ *    it under the terms of the GNU General Public License as published by
+ *    the Free Software Foundation; version 2 of the License.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *    NON INFRINGEMENT.  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., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ *    Questions/Comments/Bugfixes to iss_storagedev@hp.com
+ *
+ */
+#ifndef HPSA_CMD_H
+#define HPSA_CMD_H
+
+/* general boundary defintions */
+#define SENSEINFOBYTES          32 /* may vary between hbas */
+#define MAXSGENTRIES            31
+#define MAXREPLYQS              256
+
+/* Command Status value */
+#define CMD_SUCCESS             0x0000
+#define CMD_TARGET_STATUS       0x0001
+#define CMD_DATA_UNDERRUN       0x0002
+#define CMD_DATA_OVERRUN        0x0003
+#define CMD_INVALID             0x0004
+#define CMD_PROTOCOL_ERR        0x0005
+#define CMD_HARDWARE_ERR        0x0006
+#define CMD_CONNECTION_LOST     0x0007
+#define CMD_ABORTED             0x0008
+#define CMD_ABORT_FAILED        0x0009
+#define CMD_UNSOLICITED_ABORT   0x000A
+#define CMD_TIMEOUT             0x000B
+#define CMD_UNABORTABLE                0x000C
+
+/* Unit Attentions ASC's as defined for the MSA2012sa */
+#define POWER_OR_RESET                 0x29
+#define STATE_CHANGED                  0x2a
+#define UNIT_ATTENTION_CLEARED         0x2f
+#define LUN_FAILED                     0x3e
+#define REPORT_LUNS_CHANGED            0x3f
+
+/* Unit Attentions ASCQ's as defined for the MSA2012sa */
+
+       /* These ASCQ's defined for ASC = POWER_OR_RESET */
+#define POWER_ON_RESET                 0x00
+#define POWER_ON_REBOOT                        0x01
+#define SCSI_BUS_RESET                 0x02
+#define MSA_TARGET_RESET               0x03
+#define CONTROLLER_FAILOVER            0x04
+#define TRANSCEIVER_SE                 0x05
+#define TRANSCEIVER_LVD                        0x06
+
+       /* These ASCQ's defined for ASC = STATE_CHANGED */
+#define RESERVATION_PREEMPTED          0x03
+#define ASYM_ACCESS_CHANGED            0x06
+#define LUN_CAPACITY_CHANGED           0x09
+
+/* transfer direction */
+#define XFER_NONE               0x00
+#define XFER_WRITE              0x01
+#define XFER_READ               0x02
+#define XFER_RSVD               0x03
+
+/* task attribute */
+#define ATTR_UNTAGGED           0x00
+#define ATTR_SIMPLE             0x04
+#define ATTR_HEADOFQUEUE        0x05
+#define ATTR_ORDERED            0x06
+#define ATTR_ACA                0x07
+
+/* cdb type */
+#define TYPE_CMD                               0x00
+#define TYPE_MSG                               0x01
+
+/* config space register offsets */
+#define CFG_VENDORID            0x00
+#define CFG_DEVICEID            0x02
+#define CFG_I2OBAR              0x10
+#define CFG_MEM1BAR             0x14
+
+/* i2o space register offsets */
+#define I2O_IBDB_SET            0x20
+#define I2O_IBDB_CLEAR          0x70
+#define I2O_INT_STATUS          0x30
+#define I2O_INT_MASK            0x34
+#define I2O_IBPOST_Q            0x40
+#define I2O_OBPOST_Q            0x44
+#define I2O_DMA1_CFG           0x214
+
+/* Configuration Table */
+#define CFGTBL_ChangeReq        0x00000001l
+#define CFGTBL_AccCmds          0x00000001l
+
+#define CFGTBL_Trans_Simple     0x00000002l
+
+#define CFGTBL_BusType_Ultra2   0x00000001l
+#define CFGTBL_BusType_Ultra3   0x00000002l
+#define CFGTBL_BusType_Fibre1G  0x00000100l
+#define CFGTBL_BusType_Fibre2G  0x00000200l
+struct vals32 {
+       __u32   lower;
+       __u32   upper;
+};
+
+union u64bit {
+       struct vals32 val32;
+       __u64 val;
+};
+
+/* FIXME this is a per controller value (barf!) */
+#define HPSA_MAX_TARGETS_PER_CTLR 16
+#define HPSA_MAX_LUN 256
+#define HPSA_MAX_PHYS_LUN 1024
+
+/* SCSI-3 Commands */
+#pragma pack(1)
+
+#define HPSA_INQUIRY 0x12
+struct InquiryData {
+       __u8 data_byte[36];
+};
+
+#define HPSA_REPORT_LOG 0xc2    /* Report Logical LUNs */
+#define HPSA_REPORT_PHYS 0xc3   /* Report Physical LUNs */
+struct ReportLUNdata {
+       __u8 LUNListLength[4];
+       __u32 reserved;
+       __u8 LUN[HPSA_MAX_LUN][8];
+};
+
+struct ReportExtendedLUNdata {
+       __u8 LUNListLength[4];
+       __u8 extended_response_flag;
+       __u8 reserved[3];
+       __u8 LUN[HPSA_MAX_LUN][24];
+};
+
+struct SenseSubsystem_info {
+       __u8 reserved[36];
+       __u8 portname[8];
+       __u8 reserved1[1108];
+};
+
+#define HPSA_READ_CAPACITY 0x25 /* Read Capacity */
+struct ReadCapdata {
+       __u8 total_size[4];     /* Total size in blocks */
+       __u8 block_size[4];     /* Size of blocks in bytes */
+};
+
+#if 0
+/* 12 byte commands not implemented in firmware yet. */
+#define HPSA_READ      0xa8
+#define HPSA_WRITE     0xaa
+#endif
+
+#define HPSA_READ   0x28    /* Read(10) */
+#define HPSA_WRITE  0x2a    /* Write(10) */
+
+/* BMIC commands */
+#define BMIC_READ 0x26
+#define BMIC_WRITE 0x27
+#define BMIC_CACHE_FLUSH 0xc2
+#define HPSA_CACHE_FLUSH 0x01  /* C2 was already being used by HPSA */
+
+/* Command List Structure */
+union SCSI3Addr {
+       struct {
+               __u8 Dev;
+               __u8 Bus:6;
+               __u8 Mode:2;        /* b00 */
+       } PeripDev;
+       struct {
+               __u8 DevLSB;
+               __u8 DevMSB:6;
+               __u8 Mode:2;        /* b01 */
+       } LogDev;
+       struct {
+               __u8 Dev:5;
+               __u8 Bus:3;
+               __u8 Targ:6;
+               __u8 Mode:2;        /* b10 */
+       } LogUnit;
+};
+
+struct PhysDevAddr {
+       __u32             TargetId:24;
+       __u32             Bus:6;
+       __u32             Mode:2;
+       /* 2 level target device addr */
+       union SCSI3Addr  Target[2];
+};
+
+struct LogDevAddr {
+       __u32            VolId:30;
+       __u32            Mode:2;
+       __u8             reserved[4];
+};
+
+union LUNAddr {
+       __u8               LunAddrBytes[8];
+       union SCSI3Addr    SCSI3Lun[4];
+       struct PhysDevAddr PhysDev;
+       struct LogDevAddr  LogDev;
+};
+
+struct CommandListHeader {
+       __u8              ReplyQueue;
+       __u8              SGList;
+       __u16             SGTotal;
+       struct vals32     Tag;
+       union LUNAddr     LUN;
+};
+
+struct RequestBlock {
+       __u8   CDBLen;
+       struct {
+               __u8 Type:3;
+               __u8 Attribute:3;
+               __u8 Direction:2;
+       } Type;
+       __u16  Timeout;
+       __u8   CDB[16];
+};
+
+struct ErrDescriptor {
+       struct vals32 Addr;
+       __u32  Len;
+};
+
+struct SGDescriptor {
+       struct vals32 Addr;
+       __u32  Len;
+       __u32  Ext;
+};
+
+union MoreErrInfo {
+       struct {
+               __u8  Reserved[3];
+               __u8  Type;
+               __u32 ErrorInfo;
+       } Common_Info;
+       struct {
+               __u8  Reserved[2];
+               __u8  offense_size; /* size of offending entry */
+               __u8  offense_num;  /* byte # of offense 0-base */
+               __u32 offense_value;
+       } Invalid_Cmd;
+};
+struct ErrorInfo {
+       __u8               ScsiStatus;
+       __u8               SenseLen;
+       __u16              CommandStatus;
+       __u32              ResidualCnt;
+       union MoreErrInfo  MoreErrInfo;
+       __u8               SenseInfo[SENSEINFOBYTES];
+};
+/* Command types */
+#define CMD_IOCTL_PEND  0x01
+#define CMD_SCSI       0x03
+
+struct ctlr_info; /* defined in hpsa.h */
+/* The size of this structure needs to be divisible by 8
+ * od on all architectures, because the controller uses 2
+ * lower bits of the address, and the driver uses 1 lower
+ * bit (3 bits total.)
+ */
+struct CommandList {
+       struct CommandListHeader Header;
+       struct RequestBlock      Request;
+       struct ErrDescriptor     ErrDesc;
+       struct SGDescriptor      SG[MAXSGENTRIES];
+       /* information associated with the command */
+       __u32                      busaddr; /* physical addr of this record */
+       struct ErrorInfo *err_info; /* pointer to the allocated mem */
+       struct ctlr_info           *h;
+       int                        cmd_type;
+       long                       cmdindex;
+       struct hlist_node list;
+       struct CommandList *prev;
+       struct CommandList *next;
+       struct request *rq;
+       struct completion *waiting;
+       int      retry_count;
+       void   *scsi_cmd;
+};
+
+/* Configuration Table Structure */
+struct HostWrite {
+       __u32 TransportRequest;
+       __u32 Reserved;
+       __u32 CoalIntDelay;
+       __u32 CoalIntCount;
+};
+
+struct CfgTable {
+       __u8             Signature[4];
+       __u32            SpecValence;
+       __u32            TransportSupport;
+       __u32            TransportActive;
+       struct HostWrite HostWrite;
+       __u32            CmdsOutMax;
+       __u32            BusTypes;
+       __u32            Reserved;
+       __u8             ServerName[16];
+       __u32            HeartBeat;
+       __u32            SCSI_Prefetch;
+};
+
+struct hpsa_pci_info {
+       unsigned char   bus;
+       unsigned char   dev_fn;
+       unsigned short  domain;
+       __u32           board_id;
+};
+
+#pragma pack()
+#endif /* HPSA_CMD_H */