Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block: (41 commits)
  Revert "Seperate read and write statistics of in_flight requests"
  cfq-iosched: don't delay async queue if it hasn't dispatched at all
  block: Topology ioctls
  cfq-iosched: use assigned slice sync value, not default
  cfq-iosched: rename 'desktop' sysfs entry to 'low_latency'
  cfq-iosched: implement slower async initiate and queue ramp up
  cfq-iosched: delay async IO dispatch, if sync IO was just done
  cfq-iosched: add a knob for desktop interactiveness
  Add a tracepoint for block request remapping
  block: allow large discard requests
  block: use normal I/O path for discard requests
  swapfile: avoid NULL pointer dereference in swapon when s_bdev is NULL
  fs/bio.c: move EXPORT* macros to line after function
  Add missing blk_trace_remove_sysfs to be in pair with blk_trace_init_sysfs
  cciss: fix build when !PROC_FS
  block: Do not clamp max_hw_sectors for stacking devices
  block: Set max_sectors correctly for stacking devices
  cciss: cciss_host_attr_groups should be const
  cciss: Dynamically allocate the drive_info_struct for each logical drive.
  cciss: Add usage_count attribute to each logical drive in /sys
  ...

7 files changed, 667 insertions, 354 deletions

diff --git a/drivers/block/DAC960.c b/drivers/block/DAC960.c
index 6fa7b0fdbdf..eb4fa194394 100644
--- a/drivers/block/DAC960.c
+++ b/drivers/block/DAC960.c
@@ -38,6 +38,7 @@
 #include <linux/slab.h>
 #include <linux/smp_lock.h>
 #include <linux/proc_fs.h>
+#include <linux/seq_file.h>
 #include <linux/reboot.h>
 #include <linux/spinlock.h>
 #include <linux/timer.h>
@@ -6422,16 +6423,10 @@ static bool DAC960_V2_ExecuteUserCommand(DAC960_Controller_T *Controller,
   return true;
 }
 
-
-/*
-  DAC960_ProcReadStatus implements reading /proc/rd/status.
-*/
-
-static int DAC960_ProcReadStatus(char *Page, char **Start, off_t Offset,
-				 int Count, int *EOF, void *Data)
+static int dac960_proc_show(struct seq_file *m, void *v)
 {
   unsigned char *StatusMessage = "OK\n";
-  int ControllerNumber, BytesAvailable;
+  int ControllerNumber;
   for (ControllerNumber = 0;
        ControllerNumber < DAC960_ControllerCount;
        ControllerNumber++)
@@ -6444,52 +6439,49 @@ static int DAC960_ProcReadStatus(char *Page, char **Start, off_t Offset,
 	  break;
 	}
     }
-  BytesAvailable = strlen(StatusMessage) - Offset;
-  if (Count >= BytesAvailable)
-    {
-      Count = BytesAvailable;
-      *EOF = true;
-    }
-  if (Count <= 0) return 0;
-  *Start = Page;
-  memcpy(Page, &StatusMessage[Offset], Count);
-  return Count;
+  seq_puts(m, StatusMessage);
+  return 0;
 }
 
+static int dac960_proc_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, dac960_proc_show, NULL);
+}
 
-/*
-  DAC960_ProcReadInitialStatus implements reading /proc/rd/cN/initial_status.
-*/
+static const struct file_operations dac960_proc_fops = {
+	.owner		= THIS_MODULE,
+	.open		= dac960_proc_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
 
-static int DAC960_ProcReadInitialStatus(char *Page, char **Start, off_t Offset,
-					int Count, int *EOF, void *Data)
+static int dac960_initial_status_proc_show(struct seq_file *m, void *v)
 {
-  DAC960_Controller_T *Controller = (DAC960_Controller_T *) Data;
-  int BytesAvailable = Controller->InitialStatusLength - Offset;
-  if (Count >= BytesAvailable)
-    {
-      Count = BytesAvailable;
-      *EOF = true;
-    }
-  if (Count <= 0) return 0;
-  *Start = Page;
-  memcpy(Page, &Controller->CombinedStatusBuffer[Offset], Count);
-  return Count;
+	DAC960_Controller_T *Controller = (DAC960_Controller_T *)m->private;
+	seq_printf(m, "%.*s", Controller->InitialStatusLength, Controller->CombinedStatusBuffer);
+	return 0;
 }
 
+static int dac960_initial_status_proc_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, dac960_initial_status_proc_show, PDE(inode)->data);
+}
 
-/*
-  DAC960_ProcReadCurrentStatus implements reading /proc/rd/cN/current_status.
-*/
+static const struct file_operations dac960_initial_status_proc_fops = {
+	.owner		= THIS_MODULE,
+	.open		= dac960_initial_status_proc_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
 
-static int DAC960_ProcReadCurrentStatus(char *Page, char **Start, off_t Offset,
-					int Count, int *EOF, void *Data)
+static int dac960_current_status_proc_show(struct seq_file *m, void *v)
 {
-  DAC960_Controller_T *Controller = (DAC960_Controller_T *) Data;
+  DAC960_Controller_T *Controller = (DAC960_Controller_T *) m->private;
   unsigned char *StatusMessage =
     "No Rebuild or Consistency Check in Progress\n";
   int ProgressMessageLength = strlen(StatusMessage);
-  int BytesAvailable;
   if (jiffies != Controller->LastCurrentStatusTime)
     {
       Controller->CurrentStatusLength = 0;
@@ -6513,49 +6505,41 @@ static int DAC960_ProcReadCurrentStatus(char *Page, char **Start, off_t Offset,
 	}
       Controller->LastCurrentStatusTime = jiffies;
     }
-  BytesAvailable = Controller->CurrentStatusLength - Offset;
-  if (Count >= BytesAvailable)
-    {
-      Count = BytesAvailable;
-      *EOF = true;
-    }
-  if (Count <= 0) return 0;
-  *Start = Page;
-  memcpy(Page, &Controller->CurrentStatusBuffer[Offset], Count);
-  return Count;
+	seq_printf(m, "%.*s", Controller->CurrentStatusLength, Controller->CurrentStatusBuffer);
+	return 0;
 }
 
+static int dac960_current_status_proc_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, dac960_current_status_proc_show, PDE(inode)->data);
+}
 
-/*
-  DAC960_ProcReadUserCommand implements reading /proc/rd/cN/user_command.
-*/
+static const struct file_operations dac960_current_status_proc_fops = {
+	.owner		= THIS_MODULE,
+	.open		= dac960_current_status_proc_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
 
-static int DAC960_ProcReadUserCommand(char *Page, char **Start, off_t Offset,
-				      int Count, int *EOF, void *Data)
+static int dac960_user_command_proc_show(struct seq_file *m, void *v)
 {
-  DAC960_Controller_T *Controller = (DAC960_Controller_T *) Data;
-  int BytesAvailable = Controller->UserStatusLength - Offset;
-  if (Count >= BytesAvailable)
-    {
-      Count = BytesAvailable;
-      *EOF = true;
-    }
-  if (Count <= 0) return 0;
-  *Start = Page;
-  memcpy(Page, &Controller->UserStatusBuffer[Offset], Count);
-  return Count;
-}
+	DAC960_Controller_T *Controller = (DAC960_Controller_T *)m->private;
 
+	seq_printf(m, "%.*s", Controller->UserStatusLength, Controller->UserStatusBuffer);
+	return 0;
+}
 
-/*
-  DAC960_ProcWriteUserCommand implements writing /proc/rd/cN/user_command.
-*/
+static int dac960_user_command_proc_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, dac960_user_command_proc_show, PDE(inode)->data);
+}
 
-static int DAC960_ProcWriteUserCommand(struct file *file,
+static ssize_t dac960_user_command_proc_write(struct file *file,
 				       const char __user *Buffer,
-				       unsigned long Count, void *Data)
+				       size_t Count, loff_t *pos)
 {
-  DAC960_Controller_T *Controller = (DAC960_Controller_T *) Data;
+  DAC960_Controller_T *Controller = (DAC960_Controller_T *) PDE(file->f_path.dentry->d_inode)->data;
   unsigned char CommandBuffer[80];
   int Length;
   if (Count > sizeof(CommandBuffer)-1) return -EINVAL;
@@ -6572,6 +6556,14 @@ static int DAC960_ProcWriteUserCommand(struct file *file,
 	    ? Count : -EBUSY);
 }
 
+static const struct file_operations dac960_user_command_proc_fops = {
+	.owner		= THIS_MODULE,
+	.open		= dac960_user_command_proc_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+	.write		= dac960_user_command_proc_write,
+};
 
 /*
   DAC960_CreateProcEntries creates the /proc/rd/... entries for the
@@ -6586,23 +6578,17 @@ static void DAC960_CreateProcEntries(DAC960_Controller_T *Controller)
 
 	if (DAC960_ProcDirectoryEntry == NULL) {
   		DAC960_ProcDirectoryEntry = proc_mkdir("rd", NULL);
-  		StatusProcEntry = create_proc_read_entry("status", 0,
+  		StatusProcEntry = proc_create("status", 0,
 					   DAC960_ProcDirectoryEntry,
-					   DAC960_ProcReadStatus, NULL);
+					   &dac960_proc_fops);
 	}
 
       sprintf(Controller->ControllerName, "c%d", Controller->ControllerNumber);
       ControllerProcEntry = proc_mkdir(Controller->ControllerName,
 				       DAC960_ProcDirectoryEntry);
-      create_proc_read_entry("initial_status", 0, ControllerProcEntry,
-			     DAC960_ProcReadInitialStatus, Controller);
-      create_proc_read_entry("current_status", 0, ControllerProcEntry,
-			     DAC960_ProcReadCurrentStatus, Controller);
-      UserCommandProcEntry =
-	create_proc_read_entry("user_command", S_IWUSR | S_IRUSR,
-			       ControllerProcEntry, DAC960_ProcReadUserCommand,
-			       Controller);
-      UserCommandProcEntry->write_proc = DAC960_ProcWriteUserCommand;
+      proc_create_data("initial_status", 0, ControllerProcEntry, &dac960_initial_status_proc_fops, Controller);
+      proc_create_data("current_status", 0, ControllerProcEntry, &dac960_current_status_proc_fops, Controller);
+      UserCommandProcEntry = proc_create_data("user_command", S_IWUSR | S_IRUSR, ControllerProcEntry, &dac960_user_command_proc_fops, Controller);
       Controller->ControllerProcEntry = ControllerProcEntry;
 }
 
diff --git a/drivers/block/cciss.c b/drivers/block/cciss.c
index 1ece0b47b58..fb5be2d95d5 100644
--- a/drivers/block/cciss.c
+++ b/drivers/block/cciss.c
@@ -36,9 +36,11 @@
 #include <linux/proc_fs.h>
 #include <linux/seq_file.h>
 #include <linux/init.h>
+#include <linux/jiffies.h>
 #include <linux/hdreg.h>
 #include <linux/spinlock.h>
 #include <linux/compat.h>
+#include <linux/mutex.h>
 #include <asm/uaccess.h>
 #include <asm/io.h>
 
@@ -155,6 +157,10 @@ static struct board_type products[] = {
 
 static ctlr_info_t *hba[MAX_CTLR];
 
+static struct task_struct *cciss_scan_thread;
+static DEFINE_MUTEX(scan_mutex);
+static LIST_HEAD(scan_q);
+
 static void do_cciss_request(struct request_queue *q);
 static irqreturn_t do_cciss_intr(int irq, void *dev_id);
 static int cciss_open(struct block_device *bdev, fmode_t mode);
@@ -164,9 +170,9 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode,
 static int cciss_getgeo(struct block_device *bdev, struct hd_geometry *geo);
 
 static int cciss_revalidate(struct gendisk *disk);
-static int rebuild_lun_table(ctlr_info_t *h, int first_time);
+static int rebuild_lun_table(ctlr_info_t *h, int first_time, int via_ioctl);
 static int deregister_disk(ctlr_info_t *h, int drv_index,
-			   int clear_all);
+			   int clear_all, int via_ioctl);
 
 static void cciss_read_capacity(int ctlr, int logvol, int withirq,
 			sector_t *total_size, unsigned int *block_size);
@@ -189,8 +195,13 @@ static int sendcmd_withirq_core(ctlr_info_t *h, CommandList_struct *c,
 static int process_sendcmd_error(ctlr_info_t *h, CommandList_struct *c);
 
 static void fail_all_cmds(unsigned long ctlr);
+static int add_to_scan_list(struct ctlr_info *h);
 static int scan_thread(void *data);
 static int check_for_unit_attention(ctlr_info_t *h, CommandList_struct *c);
+static void cciss_hba_release(struct device *dev);
+static void cciss_device_release(struct device *dev);
+static void cciss_free_gendisk(ctlr_info_t *h, int drv_index);
+static void cciss_free_drive_info(ctlr_info_t *h, int drv_index);
 
 #ifdef CONFIG_PROC_FS
 static void cciss_procinit(int i);
@@ -245,7 +256,10 @@ static inline void removeQ(CommandList_struct *c)
 
 #include "cciss_scsi.c"		/* For SCSI tape support */
 
-#define RAID_UNKNOWN 6
+static const char *raid_label[] = { "0", "4", "1(1+0)", "5", "5+1", "ADG",
+	"UNKNOWN"
+};
+#define RAID_UNKNOWN (sizeof(raid_label) / sizeof(raid_label[0])-1)
 
 #ifdef CONFIG_PROC_FS
 
@@ -255,9 +269,6 @@ static inline void removeQ(CommandList_struct *c)
 #define ENG_GIG 1000000000
 #define ENG_GIG_FACTOR (ENG_GIG/512)
 #define ENGAGE_SCSI	"engage scsi"
-static const char *raid_label[] = { "0", "4", "1(1+0)", "5", "5+1", "ADG",
-	"UNKNOWN"
-};
 
 static struct proc_dir_entry *proc_cciss;
 
@@ -318,7 +329,7 @@ static int cciss_seq_show(struct seq_file *seq, void *v)
 	ctlr_info_t *h = seq->private;
 	unsigned ctlr = h->ctlr;
 	loff_t *pos = v;
-	drive_info_struct *drv = &h->drv[*pos];
+	drive_info_struct *drv = h->drv[*pos];
 
 	if (*pos > h->highest_lun)
 		return 0;
@@ -331,7 +342,7 @@ static int cciss_seq_show(struct seq_file *seq, void *v)
 	vol_sz_frac *= 100;
 	sector_div(vol_sz_frac, ENG_GIG_FACTOR);
 
-	if (drv->raid_level > 5)
+	if (drv->raid_level < 0 || drv->raid_level > RAID_UNKNOWN)
 		drv->raid_level = RAID_UNKNOWN;
 	seq_printf(seq, "cciss/c%dd%d:"
 			"\t%4u.%02uGB\tRAID %s\n",
@@ -454,9 +465,19 @@ static void __devinit cciss_procinit(int i)
 #define to_hba(n) container_of(n, struct ctlr_info, dev)
 #define to_drv(n) container_of(n, drive_info_struct, dev)
 
-static struct device_type cciss_host_type = {
-	.name		= "cciss_host",
-};
+static ssize_t host_store_rescan(struct device *dev,
+				 struct device_attribute *attr,
+				 const char *buf, size_t count)
+{
+	struct ctlr_info *h = to_hba(dev);
+
+	add_to_scan_list(h);
+	wake_up_process(cciss_scan_thread);
+	wait_for_completion_interruptible(&h->scan_wait);
+
+	return count;
+}
+DEVICE_ATTR(rescan, S_IWUSR, NULL, host_store_rescan);
 
 static ssize_t dev_show_unique_id(struct device *dev,
 				 struct device_attribute *attr,
@@ -560,11 +581,101 @@ static ssize_t dev_show_rev(struct device *dev,
 }
 DEVICE_ATTR(rev, S_IRUGO, dev_show_rev, NULL);
 
+static ssize_t cciss_show_lunid(struct device *dev,
+				struct device_attribute *attr, char *buf)
+{
+	drive_info_struct *drv = to_drv(dev);
+	struct ctlr_info *h = to_hba(drv->dev.parent);
+	unsigned long flags;
+	unsigned char lunid[8];
+
+	spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags);
+	if (h->busy_configuring) {
+		spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
+		return -EBUSY;
+	}
+	if (!drv->heads) {
+		spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
+		return -ENOTTY;
+	}
+	memcpy(lunid, drv->LunID, sizeof(lunid));
+	spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), 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]);
+}
+DEVICE_ATTR(lunid, S_IRUGO, cciss_show_lunid, NULL);
+
+static ssize_t cciss_show_raid_level(struct device *dev,
+				     struct device_attribute *attr, char *buf)
+{
+	drive_info_struct *drv = to_drv(dev);
+	struct ctlr_info *h = to_hba(drv->dev.parent);
+	int raid;
+	unsigned long flags;
+
+	spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags);
+	if (h->busy_configuring) {
+		spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
+		return -EBUSY;
+	}
+	raid = drv->raid_level;
+	spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
+	if (raid < 0 || raid > RAID_UNKNOWN)
+		raid = RAID_UNKNOWN;
+
+	return snprintf(buf, strlen(raid_label[raid]) + 7, "RAID %s\n",
+			raid_label[raid]);
+}
+DEVICE_ATTR(raid_level, S_IRUGO, cciss_show_raid_level, NULL);
+
+static ssize_t cciss_show_usage_count(struct device *dev,
+				      struct device_attribute *attr, char *buf)
+{
+	drive_info_struct *drv = to_drv(dev);
+	struct ctlr_info *h = to_hba(drv->dev.parent);
+	unsigned long flags;
+	int count;
+
+	spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags);
+	if (h->busy_configuring) {
+		spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
+		return -EBUSY;
+	}
+	count = drv->usage_count;
+	spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
+	return snprintf(buf, 20, "%d\n", count);
+}
+DEVICE_ATTR(usage_count, S_IRUGO, cciss_show_usage_count, NULL);
+
+static struct attribute *cciss_host_attrs[] = {
+	&dev_attr_rescan.attr,
+	NULL
+};
+
+static struct attribute_group cciss_host_attr_group = {
+	.attrs = cciss_host_attrs,
+};
+
+static const struct attribute_group *cciss_host_attr_groups[] = {
+	&cciss_host_attr_group,
+	NULL
+};
+
+static struct device_type cciss_host_type = {
+	.name		= "cciss_host",
+	.groups		= cciss_host_attr_groups,
+	.release	= cciss_hba_release,
+};
+
 static struct attribute *cciss_dev_attrs[] = {
 	&dev_attr_unique_id.attr,
 	&dev_attr_model.attr,
 	&dev_attr_vendor.attr,
 	&dev_attr_rev.attr,
+	&dev_attr_lunid.attr,
+	&dev_attr_raid_level.attr,
+	&dev_attr_usage_count.attr,
 	NULL
 };
 
@@ -580,12 +691,24 @@ static const struct attribute_group *cciss_dev_attr_groups[] = {
 static struct device_type cciss_dev_type = {
 	.name		= "cciss_device",
 	.groups		= cciss_dev_attr_groups,
+	.release	= cciss_device_release,
 };
 
 static struct bus_type cciss_bus_type = {
 	.name		= "cciss",
 };
 
+/*
+ * cciss_hba_release is called when the reference count
+ * of h->dev goes to zero.
+ */
+static void cciss_hba_release(struct device *dev)
+{
+	/*
+	 * nothing to do, but need this to avoid a warning
+	 * about not having a release handler from lib/kref.c.
+	 */
+}
 
 /*
  * Initialize sysfs entry for each controller.  This sets up and registers
@@ -609,6 +732,16 @@ static int cciss_create_hba_sysfs_entry(struct ctlr_info *h)
 static void cciss_destroy_hba_sysfs_entry(struct ctlr_info *h)
 {
 	device_del(&h->dev);
+	put_device(&h->dev); /* final put. */
+}
+
+/* cciss_device_release is called when the reference count
+ * of h->drv[x]dev goes to zero.
+ */
+static void cciss_device_release(struct device *dev)
+{
+	drive_info_struct *drv = to_drv(dev);
+	kfree(drv);
 }
 
 /*
@@ -617,24 +750,39 @@ static void cciss_destroy_hba_sysfs_entry(struct ctlr_info *h)
  * /sys/bus/pci/devices/<dev/ccis#/. We also create a link from
  * /sys/block/cciss!c#d# to this entry.
  */
-static int cciss_create_ld_sysfs_entry(struct ctlr_info *h,
-				       drive_info_struct *drv,
+static long cciss_create_ld_sysfs_entry(struct ctlr_info *h,
 				       int drv_index)
 {
-	device_initialize(&drv->dev);
-	drv->dev.type = &cciss_dev_type;
-	drv->dev.bus = &cciss_bus_type;
-	dev_set_name(&drv->dev, "c%dd%d", h->ctlr, drv_index);
-	drv->dev.parent = &h->dev;
-	return device_add(&drv->dev);
+	struct device *dev;
+
+	if (h->drv[drv_index]->device_initialized)
+		return 0;
+
+	dev = &h->drv[drv_index]->dev;
+	device_initialize(dev);
+	dev->type = &cciss_dev_type;
+	dev->bus = &cciss_bus_type;
+	dev_set_name(dev, "c%dd%d", h->ctlr, drv_index);
+	dev->parent = &h->dev;
+	h->drv[drv_index]->device_initialized = 1;
+	return device_add(dev);
 }
 
 /*
  * Remove sysfs entries for a logical drive.
  */
-static void cciss_destroy_ld_sysfs_entry(drive_info_struct *drv)
+static void cciss_destroy_ld_sysfs_entry(struct ctlr_info *h, int drv_index,
+	int ctlr_exiting)
 {
-	device_del(&drv->dev);
+	struct device *dev = &h->drv[drv_index]->dev;
+
+	/* special case for c*d0, we only destroy it on controller exit */
+	if (drv_index == 0 && !ctlr_exiting)
+		return;
+
+	device_del(dev);
+	put_device(dev); /* the "final" put. */
+	h->drv[drv_index] = NULL;
 }
 
 /*
@@ -751,7 +899,7 @@ static int cciss_open(struct block_device *bdev, fmode_t mode)
 	printk(KERN_DEBUG "cciss_open %s\n", bdev->bd_disk->disk_name);
 #endif				/* CCISS_DEBUG */
 
-	if (host->busy_initializing || drv->busy_configuring)
+	if (drv->busy_configuring)
 		return -EBUSY;
 	/*
 	 * Root is allowed to open raw volume zero even if it's not configured
@@ -767,7 +915,8 @@ static int cciss_open(struct block_device *bdev, fmode_t mode)
 			if (MINOR(bdev->bd_dev) & 0x0f) {
 				return -ENXIO;
 				/* if it is, make sure we have a LUN ID */
-			} else if (drv->LunID == 0) {
+			} else if (memcmp(drv->LunID, CTLR_LUNID,
+				sizeof(drv->LunID))) {
 				return -ENXIO;
 			}
 		}
@@ -1132,12 +1281,13 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode,
 	case CCISS_DEREGDISK:
 	case CCISS_REGNEWD:
 	case CCISS_REVALIDVOLS:
-		return rebuild_lun_table(host, 0);
+		return rebuild_lun_table(host, 0, 1);
 
 	case CCISS_GETLUNINFO:{
 			LogvolInfo_struct luninfo;
 
-			luninfo.LunID = drv->LunID;
+			memcpy(&luninfo.LunID, drv->LunID,
+				sizeof(luninfo.LunID));
 			luninfo.num_opens = drv->usage_count;
 			luninfo.num_parts = 0;
 			if (copy_to_user(argp, &luninfo,
@@ -1475,7 +1625,10 @@ static void cciss_check_queues(ctlr_info_t *h)
 		/* make sure the disk has been added and the drive is real
 		 * because this can be called from the middle of init_one.
 		 */
-		if (!(h->drv[curr_queue].queue) || !(h->drv[curr_queue].heads))
+		if (!h->drv[curr_queue])
+			continue;
+		if (!(h->drv[curr_queue]->queue) ||
+			!(h->drv[curr_queue]->heads))
 			continue;
 		blk_start_queue(h->gendisk[curr_queue]->queue);
 
@@ -1532,13 +1685,11 @@ static void cciss_softirq_done(struct request *rq)
 	spin_unlock_irqrestore(&h->lock, flags);
 }
 
-static void log_unit_to_scsi3addr(ctlr_info_t *h, unsigned char scsi3addr[],
-	uint32_t log_unit)
+static inline void log_unit_to_scsi3addr(ctlr_info_t *h,
+	unsigned char scsi3addr[], uint32_t log_unit)
 {
-	log_unit = h->drv[log_unit].LunID & 0x03fff;
-	memset(&scsi3addr[4], 0, 4);
-	memcpy(&scsi3addr[0], &log_unit, 4);
-	scsi3addr[3] |= 0x40;
+	memcpy(scsi3addr, h->drv[log_unit]->LunID,
+		sizeof(h->drv[log_unit]->LunID));
 }
 
 /* This function gets the SCSI vendor, model, and revision of a logical drive
@@ -1615,16 +1766,23 @@ static void cciss_get_serial_no(int ctlr, int logvol, int withirq,
 	return;
 }
 
-static void cciss_add_disk(ctlr_info_t *h, struct gendisk *disk,
+/*
+ * cciss_add_disk sets up the block device queue for a logical drive
+ */
+static int cciss_add_disk(ctlr_info_t *h, struct gendisk *disk,
 				int drv_index)
 {
 	disk->queue = blk_init_queue(do_cciss_request, &h->lock);
+	if (!disk->queue)
+		goto init_queue_failure;
 	sprintf(disk->disk_name, "cciss/c%dd%d", h->ctlr, drv_index);
 	disk->major = h->major;
 	disk->first_minor = drv_index << NWD_SHIFT;
 	disk->fops = &cciss_fops;
-	disk->private_data = &h->drv[drv_index];
-	disk->driverfs_dev = &h->drv[drv_index].dev;
+	if (cciss_create_ld_sysfs_entry(h, drv_index))
+		goto cleanup_queue;
+	disk->private_data = h->drv[drv_index];
+	disk->driverfs_dev = &h->drv[drv_index]->dev;
 
 	/* Set up queue information */
 	blk_queue_bounce_limit(disk->queue, h->pdev->dma_mask);
@@ -1642,14 +1800,21 @@ static void cciss_add_disk(ctlr_info_t *h, struct gendisk *disk,
 	disk->queue->queuedata = h;
 
 	blk_queue_logical_block_size(disk->queue,
-				     h->drv[drv_index].block_size);
+				     h->drv[drv_index]->block_size);
 
 	/* Make sure all queue data is written out before */
-	/* setting h->drv[drv_index].queue, as setting this */
+	/* setting h->drv[drv_index]->queue, as setting this */
 	/* allows the interrupt handler to start the queue */
 	wmb();
-	h->drv[drv_index].queue = disk->queue;
+	h->drv[drv_index]->queue = disk->queue;
 	add_disk(disk);
+	return 0;
+
+cleanup_queue:
+	blk_cleanup_queue(disk->queue);
+	disk->queue = NULL;
+init_queue_failure:
+	return -1;
 }
 
 /* This function will check the usage_count of the drive to be updated/added.
@@ -1662,7 +1827,8 @@ static void cciss_add_disk(ctlr_info_t *h, struct gendisk *disk,
  * is also the controller node.  Any changes to disk 0 will show up on
  * the next reboot.
  */
-static void cciss_update_drive_info(int ctlr, int drv_index, int first_time)
+static void cciss_update_drive_info(int ctlr, int drv_index, int first_time,
+	int via_ioctl)
 {
 	ctlr_info_t *h = hba[ctlr];
 	struct gendisk *disk;
@@ -1672,21 +1838,13 @@ static void cciss_update_drive_info(int ctlr, int drv_index, int first_time)
 	unsigned long flags = 0;
 	int ret = 0;
 	drive_info_struct *drvinfo;
-	int was_only_controller_node;
 
 	/* Get information about the disk and modify the driver structure */
 	inq_buff = kmalloc(sizeof(InquiryData_struct), GFP_KERNEL);
-	drvinfo = kmalloc(sizeof(*drvinfo), GFP_KERNEL);
+	drvinfo = kzalloc(sizeof(*drvinfo), GFP_KERNEL);
 	if (inq_buff == NULL || drvinfo == NULL)
 		goto mem_msg;
 
-	/* See if we're trying to update the "controller node"
-	 * this will happen the when the first logical drive gets
-	 * created by ACU.
-	 */
-	was_only_controller_node = (drv_index == 0 &&
-				h->drv[0].raid_level == -1);
-
 	/* testing to see if 16-byte CDBs are already being used */
 	if (h->cciss_read == CCISS_READ_16) {
 		cciss_read_capacity_16(h->ctlr, drv_index, 1,
@@ -1719,16 +1877,19 @@ static void cciss_update_drive_info(int ctlr, int drv_index, int first_time)
 				drvinfo->model, drvinfo->rev);
 	cciss_get_serial_no(ctlr, drv_index, 1, drvinfo->serial_no,
 			sizeof(drvinfo->serial_no));
+	/* Save the lunid in case we deregister the disk, below. */
+	memcpy(drvinfo->LunID, h->drv[drv_index]->LunID,
+		sizeof(drvinfo->LunID));
 
 	/* Is it the same disk we already know, and nothing's changed? */
-	if (h->drv[drv_index].raid_level != -1 &&
+	if (h->drv[drv_index]->raid_level != -1 &&
 		((memcmp(drvinfo->serial_no,
-				h->drv[drv_index].serial_no, 16) == 0) &&
-		drvinfo->block_size == h->drv[drv_index].block_size &&
-		drvinfo->nr_blocks == h->drv[drv_index].nr_blocks &&
-		drvinfo->heads == h->drv[drv_index].heads &&
-		drvinfo->sectors == h->drv[drv_index].sectors &&
-		drvinfo->cylinders == h->drv[drv_index].cylinders))
+				h->drv[drv_index]->serial_no, 16) == 0) &&
+		drvinfo->block_size == h->drv[drv_index]->block_size &&
+		drvinfo->nr_blocks == h->drv[drv_index]->nr_blocks &&
+		drvinfo->heads == h->drv[drv_index]->heads &&
+		drvinfo->sectors == h->drv[drv_index]->sectors &&
+		drvinfo->cylinders == h->drv[drv_index]->cylinders))
 			/* The disk is unchanged, nothing to update */
 			goto freeret;
 
@@ -1738,18 +1899,17 @@ static void cciss_update_drive_info(int ctlr, int drv_index, int first_time)
 	 * If the disk already exists then deregister it before proceeding
 	 * (unless it's the first disk (for the controller node).
 	 */
-	if (h->drv[drv_index].raid_level != -1 && drv_index != 0) {
+	if (h->drv[drv_index]->raid_level != -1 && drv_index != 0) {
 		printk(KERN_WARNING "disk %d has changed.\n", drv_index);
 		spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags);
-		h->drv[drv_index].busy_configuring = 1;
+		h->drv[drv_index]->busy_configuring = 1;
 		spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
 
-		/* deregister_disk sets h->drv[drv_index].queue = NULL
+		/* deregister_disk sets h->drv[drv_index]->queue = NULL
 		 * which keeps the interrupt handler from starting
 		 * the queue.
 		 */
-		ret = deregister_disk(h, drv_index, 0);
-		h->drv[drv_index].busy_configuring = 0;
+		ret = deregister_disk(h, drv_index, 0, via_ioctl);
 	}
 
 	/* If the disk is in use return */
@@ -1757,22 +1917,31 @@ static void cciss_update_drive_info(int ctlr, int drv_index, int first_time)
 		goto freeret;
 
 	/* Save the new information from cciss_geometry_inquiry
-	 * and serial number inquiry.
+	 * and serial number inquiry.  If the disk was deregistered
+	 * above, then h->drv[drv_index] will be NULL.
 	 */
-	h->drv[drv_index].block_size = drvinfo->block_size;
-	h->drv[drv_index].nr_blocks = drvinfo->nr_blocks;
-	h->drv[drv_index].heads = drvinfo->heads;
-	h->drv[drv_index].sectors = drvinfo->sectors;
-	h->drv[drv_index].cylinders = drvinfo->cylinders;
-	h->drv[drv_index].raid_level = drvinfo->raid_level;
-	memcpy(h->drv[drv_index].serial_no, drvinfo->serial_no, 16);
-	memcpy(h->drv[drv_index].vendor, drvinfo->vendor, VENDOR_LEN + 1);
-	memcpy(h->drv[drv_index].model, drvinfo->model, MODEL_LEN + 1);
-	memcpy(h->drv[drv_index].rev, drvinfo->rev, REV_LEN + 1);
+	if (h->drv[drv_index] == NULL) {
+		drvinfo->device_initialized = 0;
+		h->drv[drv_index] = drvinfo;
+		drvinfo = NULL; /* so it won't be freed below. */
+	} else {
+		/* special case for cxd0 */
+		h->drv[drv_index]->block_size = drvinfo->block_size;
+		h->drv[drv_index]->nr_blocks = drvinfo->nr_blocks;
+		h->drv[drv_index]->heads = drvinfo->heads;
+		h->drv[drv_index]->sectors = drvinfo->sectors;
+		h->drv[drv_index]->cylinders = drvinfo->cylinders;
+		h->drv[drv_index]->raid_level = drvinfo->raid_level;
+		memcpy(h->drv[drv_index]->serial_no, drvinfo->serial_no, 16);
+		memcpy(h->drv[drv_index]->vendor, drvinfo->vendor,
+			VENDOR_LEN + 1);
+		memcpy(h->drv[drv_index]->model, drvinfo->model, MODEL_LEN + 1);
+		memcpy(h->drv[drv_index]->rev, drvinfo->rev, REV_LEN + 1);
+	}
 
 	++h->num_luns;
 	disk = h->gendisk[drv_index];
-	set_capacity(disk, h->drv[drv_index].nr_blocks);
+	set_capacity(disk, h->drv[drv_index]->nr_blocks);
 
 	/* If it's not disk 0 (drv_index != 0)
 	 * or if it was disk 0, but there was previously
@@ -1780,8 +1949,15 @@ static void cciss_update_drive_info(int ctlr, int drv_index, int first_time)
 	 * (raid_leve == -1) then we want to update the
 	 * logical drive's information.
 	 */
-	if (drv_index || first_time)
-		cciss_add_disk(h, disk, drv_index);
+	if (drv_index || first_time) {
+		if (cciss_add_disk(h, disk, drv_index) != 0) {
+			cciss_free_gendisk(h, drv_index);
+			cciss_free_drive_info(h, drv_index);
+			printk(KERN_WARNING "cciss:%d could not update "
+				"disk %d\n", h->ctlr, drv_index);
+			--h->num_luns;
+		}
+	}
 
 freeret:
 	kfree(inq_buff);
@@ -1793,28 +1969,70 @@ mem_msg:
 }
 
 /* This function will find the first index of the controllers drive array
- * that has a -1 for the raid_level and will return that index.  This is
- * where new drives will be added.  If the index to be returned is greater
- * than the highest_lun index for the controller then highest_lun is set
- * to this new index.  If there are no available indexes then -1 is returned.
- * "controller_node" is used to know if this is a real logical drive, or just
- * the controller node, which determines if this counts towards highest_lun.
+ * that has a null drv pointer and allocate the drive info struct and
+ * will return that index   This is where new drives will be added.
+ * If the index to be returned is greater than the highest_lun index for
+ * the controller then highest_lun is set * to this new index.
+ * If there are no available indexes or if tha allocation fails, then -1
+ * is returned.  * "controller_node" is used to know if this is a real
+ * logical drive, or just the controller node, which determines if this
+ * counts towards highest_lun.
  */
-static int cciss_find_free_drive_index(int ctlr, int controller_node)
+static int cciss_alloc_drive_info(ctlr_info_t *h, int controller_node)
 {
 	int i;
+	drive_info_struct *drv;
 
+	/* Search for an empty slot for our drive info */
 	for (i = 0; i < CISS_MAX_LUN; i++) {
-		if (hba[ctlr]->drv[i].raid_level == -1) {
-			if (i > hba[ctlr]->highest_lun)
-				if (!controller_node)
-					hba[ctlr]->highest_lun = i;
+
+		/* if not cxd0 case, and it's occupied, skip it. */
+		if (h->drv[i] && i != 0)
+			continue;
+		/*
+		 * If it's cxd0 case, and drv is alloc'ed already, and a
+		 * disk is configured there, skip it.
+		 */
+		if (i == 0 && h->drv[i] && h->drv[i]->raid_level != -1)
+			continue;
+
+		/*
+		 * We've found an empty slot.  Update highest_lun
+		 * provided this isn't just the fake cxd0 controller node.
+		 */
+		if (i > h->highest_lun && !controller_node)
+			h->highest_lun = i;
+
+		/* If adding a real disk at cxd0, and it's already alloc'ed */
+		if (i == 0 && h->drv[i] != NULL)
 			return i;
-		}
+
+		/*
+		 * Found an empty slot, not already alloc'ed.  Allocate it.
+		 * Mark it with raid_level == -1, so we know it's new later on.
+		 */
+		drv = kzalloc(sizeof(*drv), GFP_KERNEL);
+		if (!drv)
+			return -1;
+		drv->raid_level = -1; /* so we know it's new */
+		h->drv[i] = drv;
+		return i;
 	}
 	return -1;
 }
 
+static void cciss_free_drive_info(ctlr_info_t *h, int drv_index)
+{
+	kfree(h->drv[drv_index]);
+	h->drv[drv_index] = NULL;
+}
+
+static void cciss_free_gendisk(ctlr_info_t *h, int drv_index)
+{
+	put_disk(h->gendisk[drv_index]);
+	h->gendisk[drv_index] = NULL;
+}
+
 /* cciss_add_gendisk finds a free hba[]->drv structure
  * and allocates a gendisk if needed, and sets the lunid
  * in the drvinfo structure.   It returns the index into
@@ -1824,13 +2042,15 @@ static int cciss_find_free_drive_index(int ctlr, int controller_node)
  * a means to talk to the controller in case no logical
  * drives have yet been configured.
  */
-static int cciss_add_gendisk(ctlr_info_t *h, __u32 lunid, int controller_node)
+static int cciss_add_gendisk(ctlr_info_t *h, unsigned char lunid[],
+	int controller_node)
 {
 	int drv_index;
 
-	drv_index = cciss_find_free_drive_index(h->ctlr, controller_node);
+	drv_index = cciss_alloc_drive_info(h, controller_node);
 	if (drv_index == -1)
 		return -1;
+
 	/*Check if the gendisk needs to be allocated */
 	if (!h->gendisk[drv_index]) {
 		h->gendisk[drv_index] =
@@ -1839,23 +2059,24 @@ static int cciss_add_gendisk(ctlr_info_t *h, __u32 lunid, int controller_node)
 			printk(KERN_ERR "cciss%d: could not "
 				"allocate a new disk %d\n",
 				h->ctlr, drv_index);
-			return -1;
+			goto err_free_drive_info;
 		}
 	}
-	h->drv[drv_index].LunID = lunid;
-	if (cciss_create_ld_sysfs_entry(h, &h->drv[drv_index], drv_index))
+	memcpy(h->drv[drv_index]->LunID, lunid,
+		sizeof(h->drv[drv_index]->LunID));
+	if (cciss_create_ld_sysfs_entry(h, drv_index))
 		goto err_free_disk;
-
 	/* Don't need to mark this busy because nobody */
 	/* else knows about this disk yet to contend */
 	/* for access to it. */
-	h->drv[drv_index].busy_configuring = 0;
+	h->drv[drv_index]->busy_configuring = 0;
 	wmb();
 	return drv_index;
 
 err_free_disk:
-	put_disk(h->gendisk[drv_index]);
-	h->gendisk[drv_index] = NULL;
+	cciss_free_gendisk(h, drv_index);
+err_free_drive_info:
+	cciss_free_drive_info(h, drv_index);
 	return -1;
 }
 
@@ -1872,21 +2093,25 @@ static void cciss_add_controller_node(ctlr_info_t *h)
 	if (h->gendisk[0] != NULL) /* already did this? Then bail. */
 		return;
 
-	drv_index = cciss_add_gendisk(h, 0, 1);
-	if (drv_index == -1) {
-		printk(KERN_WARNING "cciss%d: could not "
-			"add disk 0.\n", h->ctlr);
-		return;
-	}
-	h->drv