Merge /spare/repo/netdev-2.6/ branch 'ieee80211'

16 files changed, 638 insertions, 512 deletions

diff --git a/drivers/block/aoe/aoechr.c b/drivers/block/aoe/aoechr.c
index 14aeca3e2e8c..45a243096187 100644
--- a/drivers/block/aoe/aoechr.c
+++ b/drivers/block/aoe/aoechr.c
@@ -36,7 +36,7 @@ static int emsgs_head_idx, emsgs_tail_idx;
 static struct semaphore emsgs_sema;
 static spinlock_t emsgs_lock;
 static int nblocked_emsgs_readers;
-static struct class_simple *aoe_class;
+static struct class *aoe_class;
 static struct aoe_chardev chardevs[] = {
 	{ MINOR_ERR, "err" },
 	{ MINOR_DISCOVER, "discover" },
@@ -218,13 +218,13 @@ aoechr_init(void)
 	}
 	sema_init(&emsgs_sema, 0);
 	spin_lock_init(&emsgs_lock);
-	aoe_class = class_simple_create(THIS_MODULE, "aoe");
+	aoe_class = class_create(THIS_MODULE, "aoe");
 	if (IS_ERR(aoe_class)) {
 		unregister_chrdev(AOE_MAJOR, "aoechr");
 		return PTR_ERR(aoe_class);
 	}
 	for (i = 0; i < ARRAY_SIZE(chardevs); ++i)
-		class_simple_device_add(aoe_class,
+		class_device_create(aoe_class,
 					MKDEV(AOE_MAJOR, chardevs[i].minor),
 					NULL, chardevs[i].name);
 
@@ -237,8 +237,8 @@ aoechr_exit(void)
 	int i;
 
 	for (i = 0; i < ARRAY_SIZE(chardevs); ++i)
-		class_simple_device_remove(MKDEV(AOE_MAJOR, chardevs[i].minor));
-	class_simple_destroy(aoe_class);
+		class_device_destroy(aoe_class, MKDEV(AOE_MAJOR, chardevs[i].minor));
+	class_destroy(aoe_class);
 	unregister_chrdev(AOE_MAJOR, "aoechr");
 }
 
diff --git a/drivers/block/as-iosched.c b/drivers/block/as-iosched.c
index a9575bb58a5e..3410b4d294b9 100644
--- a/drivers/block/as-iosched.c
+++ b/drivers/block/as-iosched.c
@@ -1871,20 +1871,22 @@ static int as_init_queue(request_queue_t *q, elevator_t *e)
 	if (!arq_pool)
 		return -ENOMEM;
 
-	ad = kmalloc(sizeof(*ad), GFP_KERNEL);
+	ad = kmalloc_node(sizeof(*ad), GFP_KERNEL, q->node);
 	if (!ad)
 		return -ENOMEM;
 	memset(ad, 0, sizeof(*ad));
 
 	ad->q = q; /* Identify what queue the data belongs to */
 
-	ad->hash = kmalloc(sizeof(struct list_head)*AS_HASH_ENTRIES,GFP_KERNEL);
+	ad->hash = kmalloc_node(sizeof(struct list_head)*AS_HASH_ENTRIES,
+				GFP_KERNEL, q->node);
 	if (!ad->hash) {
 		kfree(ad);
 		return -ENOMEM;
 	}
 
-	ad->arq_pool = mempool_create(BLKDEV_MIN_RQ, mempool_alloc_slab, mempool_free_slab, arq_pool);
+	ad->arq_pool = mempool_create_node(BLKDEV_MIN_RQ, mempool_alloc_slab,
+				mempool_free_slab, arq_pool, q->node);
 	if (!ad->arq_pool) {
 		kfree(ad->hash);
 		kfree(ad);
@@ -2044,7 +2046,7 @@ as_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
 	struct as_fs_entry *entry = to_as(attr);
 
 	if (!entry->show)
-		return 0;
+		return -EIO;
 
 	return entry->show(e->elevator_data, page);
 }
@@ -2057,7 +2059,7 @@ as_attr_store(struct kobject *kobj, struct attribute *attr,
 	struct as_fs_entry *entry = to_as(attr);
 
 	if (!entry->store)
-		return -EINVAL;
+		return -EIO;
 
 	return entry->store(e->elevator_data, page, length);
 }
diff --git a/drivers/block/cciss.c b/drivers/block/cciss.c
index 8f7c1a1ed7f4..abde27027c06 100644
--- a/drivers/block/cciss.c
+++ b/drivers/block/cciss.c
@@ -41,6 +41,7 @@
 #include <asm/uaccess.h>
 #include <asm/io.h>
 
+#include <linux/dma-mapping.h>
 #include <linux/blkdev.h>
 #include <linux/genhd.h>
 #include <linux/completion.h>
@@ -126,8 +127,6 @@ static struct board_type products[] = {
 #define MAX_CTLR_ORIG 	8
 
 
-#define CCISS_DMA_MASK	0xFFFFFFFF	/* 32 bit DMA */
-
 static ctlr_info_t *hba[MAX_CTLR];
 
 static void do_cciss_request(request_queue_t *q);
@@ -2393,11 +2392,6 @@ static int cciss_pci_init(ctlr_info_t *c, struct pci_dev *pdev)
 		printk(KERN_ERR "cciss: Unable to Enable PCI device\n");
 		return( -1);
 	}
-	if (pci_set_dma_mask(pdev, CCISS_DMA_MASK ) != 0)
-	{
-		printk(KERN_ERR "cciss:  Unable to set DMA mask\n");
-		return(-1);
-	}
 
 	subsystem_vendor_id = pdev->subsystem_vendor;
 	subsystem_device_id = pdev->subsystem_device;
@@ -2747,9 +2741,9 @@ static int __devinit cciss_init_one(struct pci_dev *pdev,
 	hba[i]->pdev = pdev;
 
 	/* configure PCI DMA stuff */
-	if (!pci_set_dma_mask(pdev, 0xffffffffffffffffULL))
+	if (!pci_set_dma_mask(pdev, DMA_64BIT_MASK))
 		printk("cciss: using DAC cycles\n");
-	else if (!pci_set_dma_mask(pdev, 0xffffffff))
+	else if (!pci_set_dma_mask(pdev, DMA_32BIT_MASK))
 		printk("cciss: not using DAC cycles\n");
 	else {
 		printk("cciss: no suitable DMA available\n");
diff --git a/drivers/block/cfq-iosched.c b/drivers/block/cfq-iosched.c
index 0ef7a0065ece..3ac47dde64da 100644
--- a/drivers/block/cfq-iosched.c
+++ b/drivers/block/cfq-iosched.c
@@ -1202,13 +1202,16 @@ retry:
 		if (new_cfqq) {
 			cfqq = new_cfqq;
 			new_cfqq = NULL;
-		} else if (gfp_mask & __GFP_WAIT) {
+		} else {
 			spin_unlock_irq(cfqd->queue->queue_lock);
 			new_cfqq = kmem_cache_alloc(cfq_pool, gfp_mask);
 			spin_lock_irq(cfqd->queue->queue_lock);
+
+			if (!new_cfqq && !(gfp_mask & __GFP_WAIT))
+				goto out;
+
 			goto retry;
-		} else
-			goto out;
+		}
 
 		memset(cfqq, 0, sizeof(*cfqq));
 
@@ -1772,7 +1775,7 @@ cfq_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
 	struct cfq_fs_entry *entry = to_cfq(attr);
 
 	if (!entry->show)
-		return 0;
+		return -EIO;
 
 	return entry->show(e->elevator_data, page);
 }
@@ -1785,7 +1788,7 @@ cfq_attr_store(struct kobject *kobj, struct attribute *attr,
 	struct cfq_fs_entry *entry = to_cfq(attr);
 
 	if (!entry->store)
-		return -EINVAL;
+		return -EIO;
 
 	return entry->store(e->elevator_data, page, length);
 }
diff --git a/drivers/block/deadline-iosched.c b/drivers/block/deadline-iosched.c
index d63d34c671f7..4bc2fea73273 100644
--- a/drivers/block/deadline-iosched.c
+++ b/drivers/block/deadline-iosched.c
@@ -711,18 +711,20 @@ static int deadline_init_queue(request_queue_t *q, elevator_t *e)
 	if (!drq_pool)
 		return -ENOMEM;
 
-	dd = kmalloc(sizeof(*dd), GFP_KERNEL);
+	dd = kmalloc_node(sizeof(*dd), GFP_KERNEL, q->node);
 	if (!dd)
 		return -ENOMEM;
 	memset(dd, 0, sizeof(*dd));
 
-	dd->hash = kmalloc(sizeof(struct list_head)*DL_HASH_ENTRIES,GFP_KERNEL);
+	dd->hash = kmalloc_node(sizeof(struct list_head)*DL_HASH_ENTRIES,
+				GFP_KERNEL, q->node);
 	if (!dd->hash) {
 		kfree(dd);
 		return -ENOMEM;
 	}
 
-	dd->drq_pool = mempool_create(BLKDEV_MIN_RQ, mempool_alloc_slab, mempool_free_slab, drq_pool);
+	dd->drq_pool = mempool_create_node(BLKDEV_MIN_RQ, mempool_alloc_slab,
+					mempool_free_slab, drq_pool, q->node);
 	if (!dd->drq_pool) {
 		kfree(dd->hash);
 		kfree(dd);
@@ -886,7 +888,7 @@ deadline_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
 	struct deadline_fs_entry *entry = to_deadline(attr);
 
 	if (!entry->show)
-		return 0;
+		return -EIO;
 
 	return entry->show(e->elevator_data, page);
 }
@@ -899,7 +901,7 @@ deadline_attr_store(struct kobject *kobj, struct attribute *attr,
 	struct deadline_fs_entry *entry = to_deadline(attr);
 
 	if (!entry->store)
-		return -EINVAL;
+		return -EIO;
 
 	return entry->store(e->elevator_data, page, length);
 }
diff --git a/drivers/block/elevator.c b/drivers/block/elevator.c
index 6b79b4314622..f831f08f839c 100644
--- a/drivers/block/elevator.c
+++ b/drivers/block/elevator.c
@@ -220,11 +220,6 @@ void elevator_exit(elevator_t *e)
 	kfree(e);
 }
 
-static int elevator_global_init(void)
-{
-	return 0;
-}
-
 int elv_merge(request_queue_t *q, struct request **req, struct bio *bio)
 {
 	elevator_t *e = q->elevator;
@@ -291,6 +286,13 @@ void elv_requeue_request(request_queue_t *q, struct request *rq)
 	}
 
 	/*
+	 * the request is prepped and may have some resources allocated.
+	 * allowing unprepped requests to pass this one may cause resource
+	 * deadlock.  turn on softbarrier.
+	 */
+	rq->flags |= REQ_SOFTBARRIER;
+
+	/*
 	 * if iosched has an explicit requeue hook, then use that. otherwise
 	 * just put the request at the front of the queue
 	 */
@@ -322,7 +324,7 @@ void __elv_add_request(request_queue_t *q, struct request *rq, int where,
 			int nrq = q->rq.count[READ] + q->rq.count[WRITE]
 				  - q->in_flight;
 
-			if (nrq == q->unplug_thresh)
+			if (nrq >= q->unplug_thresh)
 				__generic_unplug_device(q);
 		}
 	} else
@@ -386,6 +388,12 @@ struct request *elv_next_request(request_queue_t *q)
 		if (ret == BLKPREP_OK) {
 			break;
 		} else if (ret == BLKPREP_DEFER) {
+			/*
+			 * the request may have been (partially) prepped.
+			 * we need to keep this request in the front to
+			 * avoid resource deadlock.  turn on softbarrier.
+			 */
+			rq->flags |= REQ_SOFTBARRIER;
 			rq = NULL;
 			break;
 		} else if (ret == BLKPREP_KILL) {
@@ -692,8 +700,6 @@ ssize_t elv_iosched_show(request_queue_t *q, char *name)
 	return len;
 }
 
-module_init(elevator_global_init);
-
 EXPORT_SYMBOL(elv_add_request);
 EXPORT_SYMBOL(__elv_add_request);
 EXPORT_SYMBOL(elv_requeue_request);
diff --git a/drivers/block/genhd.c b/drivers/block/genhd.c
index 8bbe01d4b487..47fd3659a061 100644
--- a/drivers/block/genhd.c
+++ b/drivers/block/genhd.c
@@ -40,7 +40,7 @@ static inline int major_to_index(int major)
 
 #ifdef CONFIG_PROC_FS
 /* get block device names in somewhat random order */
-int get_blkdev_list(char *p)
+int get_blkdev_list(char *p, int used)
 {
 	struct blk_major_name *n;
 	int i, len;
@@ -49,10 +49,18 @@ int get_blkdev_list(char *p)
 
 	down(&block_subsys_sem);
 	for (i = 0; i < ARRAY_SIZE(major_names); i++) {
-		for (n = major_names[i]; n; n = n->next)
+		for (n = major_names[i]; n; n = n->next) {
+			/*
+			 * If the curent string plus the 5 extra characters
+			 * in the line would run us off the page, then we're done
+			 */
+			if ((len + used + strlen(n->name) + 5) >= PAGE_SIZE)
+				goto page_full;
 			len += sprintf(p+len, "%3d %s\n",
 				       n->major, n->name);
+		}
 	}
+page_full:
 	up(&block_subsys_sem);
 
 	return len;
@@ -322,7 +330,7 @@ static ssize_t disk_attr_show(struct kobject *kobj, struct attribute *attr,
 	struct gendisk *disk = to_disk(kobj);
 	struct disk_attribute *disk_attr =
 		container_of(attr,struct disk_attribute,attr);
-	ssize_t ret = 0;
+	ssize_t ret = -EIO;
 
 	if (disk_attr->show)
 		ret = disk_attr->show(disk,page);
@@ -582,10 +590,16 @@ struct seq_operations diskstats_op = {
 	.show	= diskstats_show
 };
 
-
 struct gendisk *alloc_disk(int minors)
 {
-	struct gendisk *disk = kmalloc(sizeof(struct gendisk), GFP_KERNEL);
+	return alloc_disk_node(minors, -1);
+}
+
+struct gendisk *alloc_disk_node(int minors, int node_id)
+{
+	struct gendisk *disk;
+
+	disk = kmalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
 	if (disk) {
 		memset(disk, 0, sizeof(struct gendisk));
 		if (!init_disk_stats(disk)) {
@@ -594,7 +608,7 @@ struct gendisk *alloc_disk(int minors)
 		}
 		if (minors > 1) {
 			int size = (minors - 1) * sizeof(struct hd_struct *);
-			disk->part = kmalloc(size, GFP_KERNEL);
+			disk->part = kmalloc_node(size, GFP_KERNEL, node_id);
 			if (!disk->part) {
 				kfree(disk);
 				return NULL;
@@ -610,6 +624,7 @@ struct gendisk *alloc_disk(int minors)
 }
 
 EXPORT_SYMBOL(alloc_disk);
+EXPORT_SYMBOL(alloc_disk_node);
 
 struct kobject *get_disk(struct gendisk *disk)
 {
diff --git a/drivers/block/ioctl.c b/drivers/block/ioctl.c
index 6d7bcc9da9e7..6e278474f9a8 100644
--- a/drivers/block/ioctl.c
+++ b/drivers/block/ioctl.c
@@ -133,11 +133,9 @@ static int put_u64(unsigned long arg, u64 val)
 	return put_user(val, (u64 __user *)arg);
 }
 
-int blkdev_ioctl(struct inode *inode, struct file *file, unsigned cmd,
-			unsigned long arg)
+static int blkdev_locked_ioctl(struct file *file, struct block_device *bdev,
+				unsigned cmd, unsigned long arg)
 {
-	struct block_device *bdev = inode->i_bdev;
-	struct gendisk *disk = bdev->bd_disk;
 	struct backing_dev_info *bdi;
 	int ret, n;
 
@@ -190,36 +188,72 @@ int blkdev_ioctl(struct inode *inode, struct file *file, unsigned cmd,
 		return put_ulong(arg, bdev->bd_inode->i_size >> 9);
 	case BLKGETSIZE64:
 		return put_u64(arg, bdev->bd_inode->i_size);
+	}
+	return -ENOIOCTLCMD;
+}
+
+static int blkdev_driver_ioctl(struct inode *inode, struct file *file,
+		struct gendisk *disk, unsigned cmd, unsigned long arg)
+{
+	int ret;
+	if (disk->fops->unlocked_ioctl)
+		return disk->fops->unlocked_ioctl(file, cmd, arg);
+
+	if (disk->fops->ioctl) {
+		lock_kernel();
+		ret = disk->fops->ioctl(inode, file, cmd, arg);
+		unlock_kernel();
+		return ret;
+	}
+
+	return -ENOTTY;
+}
+
+int blkdev_ioctl(struct inode *inode, struct file *file, unsigned cmd,
+			unsigned long arg)
+{
+	struct block_device *bdev = inode->i_bdev;
+	struct gendisk *disk = bdev->bd_disk;
+	int ret, n;
+
+	switch(cmd) {
 	case BLKFLSBUF:
 		if (!capable(CAP_SYS_ADMIN))
 			return -EACCES;
-		if (disk->fops->ioctl) {
-			ret = disk->fops->ioctl(inode, file, cmd, arg);
-			/* -EINVAL to handle old uncorrected drivers */
-			if (ret != -EINVAL && ret != -ENOTTY)
-				return ret;
-		}
+
+		ret = blkdev_driver_ioctl(inode, file, disk, cmd, arg);
+		/* -EINVAL to handle old uncorrected drivers */
+		if (ret != -EINVAL && ret != -ENOTTY)
+			return ret;
+
+		lock_kernel();
 		fsync_bdev(bdev);
 		invalidate_bdev(bdev, 0);
+		unlock_kernel();
 		return 0;
+
 	case BLKROSET:
-		if (disk->fops->ioctl) {
-			ret = disk->fops->ioctl(inode, file, cmd, arg);
-			/* -EINVAL to handle old uncorrected drivers */
-			if (ret != -EINVAL && ret != -ENOTTY)
-				return ret;
-		}
+		ret = blkdev_driver_ioctl(inode, file, disk, cmd, arg);
+		/* -EINVAL to handle old uncorrected drivers */
+		if (ret != -EINVAL && ret != -ENOTTY)
+			return ret;
 		if (!capable(CAP_SYS_ADMIN))
 			return -EACCES;
 		if (get_user(n, (int __user *)(arg)))
 			return -EFAULT;
+		lock_kernel();
 		set_device_ro(bdev, n);
+		unlock_kernel();
 		return 0;
-	default:
-		if (disk->fops->ioctl)
-			return disk->fops->ioctl(inode, file, cmd, arg);
 	}
-	return -ENOTTY;
+
+	lock_kernel();
+	ret = blkdev_locked_ioctl(file, bdev, cmd, arg);
+	unlock_kernel();
+	if (ret != -ENOIOCTLCMD)
+		return ret;
+
+	return blkdev_driver_ioctl(inode, file, disk, cmd, arg);
 }
 
 /* Most of the generic ioctls are handled in the normal fallback path.
diff --git a/drivers/block/ll_rw_blk.c b/drivers/block/ll_rw_blk.c
index 11ef9d9ea139..60e64091de1b 100644
--- a/drivers/block/ll_rw_blk.c
+++ b/drivers/block/ll_rw_blk.c
@@ -28,6 +28,7 @@
 #include <linux/slab.h>
 #include <linux/swap.h>
 #include <linux/writeback.h>
+#include <linux/blkdev.h>
 
 /*
  * for max sense size
@@ -36,6 +37,7 @@
 
 static void blk_unplug_work(void *data);
 static void blk_unplug_timeout(unsigned long data);
+static void drive_stat_acct(struct request *rq, int nr_sectors, int new_io);
 
 /*
  * For the allocated request tables
@@ -716,7 +718,7 @@ struct request *blk_queue_find_tag(request_queue_t *q, int tag)
 {
 	struct blk_queue_tag *bqt = q->queue_tags;
 
-	if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
+	if (unlikely(bqt == NULL || tag >= bqt->max_depth))
 		return NULL;
 
 	return bqt->tag_index[tag];
@@ -774,9 +776,9 @@ EXPORT_SYMBOL(blk_queue_free_tags);
 static int
 init_tag_map(request_queue_t *q, struct blk_queue_tag *tags, int depth)
 {
-	int bits, i;
 	struct request **tag_index;
 	unsigned long *tag_map;
+	int nr_ulongs;
 
 	if (depth > q->nr_requests * 2) {
 		depth = q->nr_requests * 2;
@@ -788,24 +790,17 @@ init_tag_map(request_queue_t *q, struct blk_queue_tag *tags, int depth)
 	if (!tag_index)
 		goto fail;
 
-	bits = (depth / BLK_TAGS_PER_LONG) + 1;
-	tag_map = kmalloc(bits * sizeof(unsigned long), GFP_ATOMIC);
+	nr_ulongs = ALIGN(depth, BITS_PER_LONG) / BITS_PER_LONG;
+	tag_map = kmalloc(nr_ulongs * sizeof(unsigned long), GFP_ATOMIC);
 	if (!tag_map)
 		goto fail;
 
 	memset(tag_index, 0, depth * sizeof(struct request *));
-	memset(tag_map, 0, bits * sizeof(unsigned long));
+	memset(tag_map, 0, nr_ulongs * sizeof(unsigned long));
 	tags->max_depth = depth;
-	tags->real_max_depth = bits * BITS_PER_LONG;
 	tags->tag_index = tag_index;
 	tags->tag_map = tag_map;
 
-	/*
-	 * set the upper bits if the depth isn't a multiple of the word size
-	 */
-	for (i = depth; i < bits * BLK_TAGS_PER_LONG; i++)
-		__set_bit(i, tag_map);
-
 	return 0;
 fail:
 	kfree(tag_index);
@@ -870,32 +865,24 @@ int blk_queue_resize_tags(request_queue_t *q, int new_depth)
 	struct blk_queue_tag *bqt = q->queue_tags;
 	struct request **tag_index;
 	unsigned long *tag_map;
-	int bits, max_depth;
+	int max_depth, nr_ulongs;
 
 	if (!bqt)
 		return -ENXIO;
 
 	/*
-	 * don't bother sizing down
-	 */
-	if (new_depth <= bqt->real_max_depth) {
-		bqt->max_depth = new_depth;
-		return 0;
-	}
-
-	/*
 	 * save the old state info, so we can copy it back
 	 */
 	tag_index = bqt->tag_index;
 	tag_map = bqt->tag_map;
-	max_depth = bqt->real_max_depth;
+	max_depth = bqt->max_depth;
 
 	if (init_tag_map(q, bqt, new_depth))
 		return -ENOMEM;
 
 	memcpy(bqt->tag_index, tag_index, max_depth * sizeof(struct request *));
-	bits = max_depth / BLK_TAGS_PER_LONG;
-	memcpy(bqt->tag_map, tag_map, bits * sizeof(unsigned long));
+	nr_ulongs = ALIGN(max_depth, BITS_PER_LONG) / BITS_PER_LONG;
+	memcpy(bqt->tag_map, tag_map, nr_ulongs * sizeof(unsigned long));
 
 	kfree(tag_index);
 	kfree(tag_map);
@@ -925,11 +912,16 @@ void blk_queue_end_tag(request_queue_t *q, struct request *rq)
 
 	BUG_ON(tag == -1);
 
-	if (unlikely(tag >= bqt->real_max_depth))
+	if (unlikely(tag >= bqt->max_depth))
+		/*
+		 * This can happen after tag depth has been reduced.
+		 * FIXME: how about a warning or info message here?
+		 */
 		return;
 
 	if (unlikely(!__test_and_clear_bit(tag, bqt->tag_map))) {
-		printk("attempt to clear non-busy tag (%d)\n", tag);
+		printk(KERN_ERR "%s: attempt to clear non-busy tag (%d)\n",
+		       __FUNCTION__, tag);
 		return;
 	}
 
@@ -938,7 +930,8 @@ void blk_queue_end_tag(request_queue_t *q, struct request *rq)
 	rq->tag = -1;
 
 	if (unlikely(bqt->tag_index[tag] == NULL))
-		printk("tag %d is missing\n", tag);
+		printk(KERN_ERR "%s: tag %d is missing\n",
+		       __FUNCTION__, tag);
 
 	bqt->tag_index[tag] = NULL;
 	bqt->busy--;
@@ -967,24 +960,20 @@ EXPORT_SYMBOL(blk_queue_end_tag);
 int blk_queue_start_tag(request_queue_t *q, struct request *rq)
 {
 	struct blk_queue_tag *bqt = q->queue_tags;
-	unsigned long *map = bqt->tag_map;
-	int tag = 0;
+	int tag;
 
 	if (unlikely((rq->flags & REQ_QUEUED))) {
 		printk(KERN_ERR 
-		       "request %p for device [%s] already tagged %d",
-		       rq, rq->rq_disk ? rq->rq_disk->disk_name : "?", rq->tag);
+		       "%s: request %p for device [%s] already tagged %d",
+		       __FUNCTION__, rq,
+		       rq->rq_disk ? rq->rq_disk->disk_name : "?", rq->tag);
 		BUG();
 	}
 
-	for (map = bqt->tag_map; *map == -1UL; map++) {
-		tag += BLK_TAGS_PER_LONG;
-
-		if (tag >= bqt->max_depth)
-			return 1;
-	}
+	tag = find_first_zero_bit(bqt->tag_map, bqt->max_depth);
+	if (tag >= bqt->max_depth)
+		return 1;
 
-	tag += ffz(*map);
 	__set_bit(tag, bqt->tag_map);
 
 	rq->flags |= REQ_QUEUED;
@@ -1020,7 +1009,8 @@ void blk_queue_invalidate_tags(request_queue_t *q)
 		rq = list_entry_rq(tmp);
 
 		if (rq->tag == -1) {
-			printk("bad tag found on list\n");
+			printk(KERN_ERR
+			       "%s: bad tag found on list\n", __FUNCTION__);
 			list_del_init(&rq->queuelist);
 			rq->flags &= ~REQ_QUEUED;
 		} else
@@ -1148,7 +1138,7 @@ new_hw_segment:
 }
 
 
-int blk_phys_contig_segment(request_queue_t *q, struct bio *bio,
+static int blk_phys_contig_segment(request_queue_t *q, struct bio *bio,
 				   struct bio *nxt)
 {
 	if (!(q->queue_flags & (1 << QUEUE_FLAG_CLUSTER)))
@@ -1169,9 +1159,7 @@ int blk_phys_contig_segment(request_queue_t *q, struct bio *bio,
 	return 0;
 }
 
-EXPORT_SYMBOL(blk_phys_contig_segment);
-
-int blk_hw_contig_segment(request_queue_t *q, struct bio *bio,
+static int blk_hw_contig_segment(request_queue_t *q, struct bio *bio,
 				 struct bio *nxt)
 {
 	if (unlikely(!bio_flagged(bio, BIO_SEG_VALID)))
@@ -1187,8 +1175,6 @@ int blk_hw_contig_segment(request_queue_t *q, struct bio *bio,
 	return 1;
 }
 
-EXPORT_SYMBOL(blk_hw_contig_segment);
-
 /*
  * map a request to scatterlist, return number of sg entries setup. Caller
  * must make sure sg can hold rq->nr_phys_segments entries
@@ -1358,8 +1344,8 @@ static int ll_front_merge_fn(request_queue_t *q, struct request *req,
 static int ll_merge_requests_fn(request_queue_t *q, struct request *req,
 				struct request *next)
 {
-	int total_phys_segments = req->nr_phys_segments +next->nr_phys_segments;
-	int total_hw_segments = req->nr_hw_segments + next->nr_hw_segments;
+	int total_phys_segments;
+	int total_hw_segments;
 
 	/*
 	 * First check if the either of the requests are re-queued
@@ -1369,7 +1355,7 @@ static int ll_merge_requests_fn(request_queue_t *q, struct request *req,
 		return 0;
 
 	/*
-	 * Will it become to large?
+	 * Will it become too large?
 	 */
 	if ((req->nr_sectors + next->nr_sectors) > q->max_sectors)
 		return 0;
@@ -1450,7 +1436,7 @@ EXPORT_SYMBOL(blk_remove_plug);
  */
 void __generic_unplug_device(request_queue_t *q)
 {
-	if (test_bit(QUEUE_FLAG_STOPPED, &q->queue_flags))
+	if (unlikely(test_bit(QUEUE_FLAG_STOPPED, &q->queue_flags)))
 		return;
 
 	if (!blk_remove_plug(q))
@@ -1645,7 +1631,8 @@ static int blk_init_free_list(request_queue_t *q)
 	init_waitqueue_head(&rl->wait[WRITE]);
 	init_waitqueue_head(&rl->drain);
 
-	rl->rq_pool = mempool_create(BLKDEV_MIN_RQ, mempool_alloc_slab, mempool_free_slab, request_cachep);
+	rl->rq_pool = mempool_create_node(BLKDEV_MIN_RQ, mempool_alloc_slab,
+				mempool_free_slab, request_cachep, q->node);
 
 	if (!rl->rq_pool)
 		return -ENOMEM;
@@ -1657,8 +1644,15 @@ static int __make_request(request_queue_t *, struct bio *);
 
 request_queue_t *blk_alloc_queue(int gfp_mask)
 {
-	request_queue_t *q = kmem_cache_alloc(requestq_cachep, gfp_mask);
+	return blk_alloc_queue_node(gfp_mask, -1);
+}
+EXPORT_SYMBOL(blk_alloc_queue);
 
+request_queue_t *blk_alloc_queue_node(int gfp_mask, int node_id)
+{
+	request_queue_t *q;
+
+	q = kmem_cache_alloc_node(requestq_cachep, gfp_mask, node_id);
 	if (!q)
 		return NULL;
 
@@ -1671,8 +1665,7 @@ request_queue_t *blk_alloc_queue(int gfp_mask)
 
 	return q;
 }
-
-EXPORT_SYMBOL(blk_alloc_queue);
+EXPORT_SYMBOL(blk_alloc_queue_node);
 
 /**
  * blk_init_queue  - prepare a request queue for use with a block device
@@ -1705,13 +1698,22 @@ EXPORT_SYMBOL(blk_alloc_queue);
  *    blk_init_queue() must be paired with a blk_cleanup_queue() call
  *    when the block device is deactivated (such as at module unload).
  **/
+
 request_queue_t *blk_init_queue(request_fn_proc *rfn, spinlock_t *lock)
 {
-	request_queue_t *q = blk_alloc_queue(GFP_KERNEL);
+	return blk_init_queue_node(rfn, lock, -1);
+}
+EXPORT_SYMBOL(blk_init_queue);
+
+request_queue_t *
+blk_init_queue_node(request_fn_proc *rfn, spinlock_t *lock, int node_id)
+{
+	request_queue_t *q = blk_alloc_queue_node(GFP_KERNEL, node_id);
 
 	if (!q)
 		return NULL;
 
+	q->node = node_id;
 	if (blk_init_free_list(q))
 		goto out_init;
 
@@ -1754,12 +1756,11 @@ out_init:
 	kmem_cache_free(requestq_cachep, q);
 	return NULL;
 }
-
-EXPORT_SYMBOL(blk_init_queue);
+EXPORT_SYMBOL(blk_init_queue_node);
 
 int blk_get_queue(request_queue_t *q)
 {
-	if (!test_bit(QUEUE_FLAG_DEAD, &q->queue_flags)) {
+	if (likely(!test_bit(QUEUE_FLAG_DEAD, &q->queue_flags))) {
 		atomic_inc(&q->refcnt);
 		return 0;
 	}
@@ -1821,7 +1822,7 @@ static inline int ioc_batching(request_queue_t *q, struct io_context *ioc)
  * is the behaviour we want though - once it gets a wakeup it should be given
  * a nice run.
  */
-void ioc_set_batching(request_queue_t *q, struct io_context *ioc)
+static void ioc_set_batching(request_queue_t *q, struct io_context *ioc)
 {
 	if (!ioc || ioc_batching(q, ioc))
 		return;
@@ -1838,7 +1839,6 @@ static void __freed_request(request_queue_t *q, int rw)
 		clear_queue_congested(q, rw);
 
 	if (rl->count[rw] + 1 <= q->nr_requests) {
-		smp_mb();
 		if (waitqueue_active(&rl->wait[rw]))
 			wake_up(&rl->wait[rw]);
 
@@ -1966,7 +1966,6 @@ static struct request *get_request_wait(request_queue_t *q, int rw)
 	DEFINE_WAIT(wait);
 	struct request *rq;
 
-	generic_unplug_device(q);
 	do {
 		struct request_list *rl = &q->rq;
 
@@ -1978,6 +1977,7 @@ static struct request *get_request_wait(request_queue_t *q, int rw)
 		if (!rq) {
 			struct io_context *ioc;
 
+			generic_unplug_device(q);
 			io_schedule();
 
 			/*
@@ -2038,7 +2038,6 @@ EXPORT_SYMBOL(blk_requeue_request);
  * @rq:		request to be inserted
  * @at_head:	insert request at head or tail of queue
  * @data:	private data
- * @reinsert:	true if request it a reinsertion of previously processed one
  *
  * Description:
  *    Many block devices need to execute commands asynchronously, so they don't
@@ -2053,8 +2052,9 @@ EXPORT_SYMBOL(blk_requeue_request);
  *    host that is unable to accept a particular command.
  */
 void blk_insert_request(request_queue_t *q, struct request *rq,
-			int at_head, void *data, int reinsert)
+			int at_head, void *data)
 {
+	int where = at_head ? ELEVATOR_INSERT_FRONT : ELEVATOR_INSERT_BACK;
 	unsigned long flags;
 
 	/*
@@ -2071,20 +2071,12 @@ void blk_insert_request(request_queue_t *q, struct request *rq,
 	/*
 	 * If command is tagged, release the tag
 	 */
-	if (reinsert)
-		blk_requeue_request(q, rq);
-	else {
-		int where = ELEVATOR_INSERT_BACK;
-
-		if (at_head)
-			where = ELEVATOR_INSERT_FRONT;
+	if (blk_rq_tagged(rq))
+		blk_queue_end_tag(q, rq);
 
-		if (blk_rq_tagged(rq))
-			blk_queue_end_tag(q, rq);
+	drive_stat_acct(rq, rq->nr_sectors, 1);
+	__elv_add_request(q, rq, where, 0);
 
-		drive_stat_acct(rq, rq->nr_sectors, 1);
-		__elv_add_request(q, rq, where, 0);
-	}
 	if (blk_queue_plugged(q))
 		__generic_unplug_device(q);
 	else
@@ -2259,45 +2251,7 @@ int blkdev_issue_flush(struct block_device *bdev, sector_t *error_sector)
 
 EXPORT_SYMBOL(blkdev_issue_flush);
 
-/**
- * blkdev_scsi_issue_flush_fn - issue flush for SCSI devices
- * @q:		device queue
- * @disk:	gendisk
- * @error_sector:	error offset
- *
- * Description:
- *    Devices understanding the SCSI command set, can use this function as
- *    a helper for issuing a cache flush. Note: driver is required to store
- *    the error offset (in case of error flushing) in ->sector of struct
- *    request.
- */
-int blkdev_scsi_issue_flush_fn(request_queue_t *q, struct gendisk *disk,
-			       sector_t *error_sector)
-{
-	struct request *rq = blk_get_request(q, WRITE, __GFP_WAIT);
-	int ret;
-
-	rq->flags |= REQ_BLOCK_PC | REQ_SOFTBARRIER;
-	rq->sector = 0;
-	memset(rq->cmd, 0, sizeof(rq->cmd));
-	rq->cmd[0] = 0x35;
-	rq->cmd_len = 12;
-	rq->data = NULL;
-	rq->data_len = 0;
-	rq->timeout = 60 * HZ;
-
-	ret = blk_execute_rq(q, disk, rq);
-
-	if (ret && error_sector)
-		*error_sector = rq->sector;
-
-	blk_put_request(rq);
-	return ret;
-}
-
-EXPORT_SYMBOL(blkdev_scsi_issue_flush_fn);
-
-void drive_stat_acct(struct request *rq, int nr_sectors, int new_io)
+static void drive_stat_acct(struct request *rq, int nr_sectors, int new_io)
 {
 	int rw = rq_data_dir(rq);
 
@@ -2556,16 +2510,6 @@ void blk_attempt_remerge(request_queue_t *q, struct request *rq)
 
 EXPORT_SYMBOL(blk_attempt_remerge);
 
-/*
- * Non-locking blk_attempt_remerge variant.
- */
-void __blk_attempt_remerge(request_queue_t *q, struct request *rq)
-{
-	attempt_back_merge(q, rq);
-}
-
-EXPORT_SYMBOL(__blk_attempt_remerge);
-
 static int __make_request(request_queue_t *q, struct bio *bio)
 {
 	struct request *req, *freereq = NULL;
@@ -2589,7 +2533,7 @@ static int __make_request(request_queue_t *q, struct bio *bio)
 	spin_lock_prefetch(q->queue_lock);
 
 	barrier = bio_barrier(bio);
-	if (barrier && (q->ordered == QUEUE_ORDERED_NONE)) {
+	if (unlikely(barrier) && (q->ordered == QUEUE_ORDERED_NONE)) {
 		err = -EOPNOTSUPP;
 		goto end_io;
 	}
@@ -2690,7 +2634,7 @@ get_rq:
 	/*
 	 * REQ_BARRIER implies no merging, but lets make it explicit
 	 */
-	if (barrier)
+	if (unlikely(barrier))
 		req->flags |= (REQ_HARDBARRIER | REQ_NOMERGE);
 
 	req->errors = 0;
@@ -2814,7 +2758,7 @@ static inline void block_wait_queue_running(request_queue_t *q)
 {
 	DEFINE_WAIT(wait);
 
-	while (test_bit(QUEUE_FLAG_DRAIN, &q->queue_flags)) {
+	while (unlikely(test_bit(QUEUE_FLAG_DRAIN, &q->queue_flags))) {
 		struct request_list *rl = &q->rq;
 
 		prepare_to_wait_exclusive(&rl->drain, &wait,
@@ -2923,7 +2867,7 @@ end_io:
 			goto end_io;
 		}
 
-		if (test_bit(QUEUE_FLAG_DEAD, &q->queue_flags))
+		if (unlikely(test_bit(QUEUE_FLAG_DEAD, &q->queue_flags)))
 			goto end_io;
 
 		block_wait_queue_running(q);
@@ -2976,7 +2920,7 @@ void submit_bio(int rw, struct bio *bio)
 
 EXPORT_SYMBOL(submit_bio);
 
-void blk_recalc_rq_segments(struct request *rq)
+static void blk_recalc_rq_segments(struct request *rq)
 {
 	struct bio *bio, *prevbio = NULL;
 	int nr_phys_segs, nr_hw_segs;
@@ -3018,7 +2962,7 @@ void blk_recalc_rq_segments(struct request *rq)
 	rq->nr_hw_segments = nr_hw_segs;
 }
 
-void blk_recalc_rq_sectors(struct request *rq, int nsect)
+static void blk_recalc_rq_sectors(struct request *rq, int nsect)
 {
 	if (blk_fs_request(rq)) {
 		rq->hard_sector += nsect;
@@ -3582,7 +3526,7 @@ queue_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
 
 	q = container_of(kobj, struct request_queue, kobj);
 	if (!entry->show)
-		return 0;
+		return -EIO;
 
 	return entry->show(q, page);
 }
@@ -3596,7 +3540,7 @@ queue_attr_store(struct kobject *kobj, struct attribute *attr,
 
 	q = container_of(kobj, struct request_queue, kobj);
 	if (!entry->store)
-		return -EINVAL;
+		return -EIO;
 
 	return entry->store(q, page, length);
 }
@@ -3606,7 +3550,7 @@ static struct sysfs_ops queue_sysfs_ops = {
 	.store	= queue_attr_store,
 };
 
-struct kobj_type queue_ktype = {
+static struct kobj_type queue_ktype = {
 	.sysfs_ops	= &queue_sysfs_ops,
 	.default_attrs	= default_attrs,
 };
diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index 6f011d0d8e97..b35e08876dd4 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -472,17 +472,11 @@ static int do_bio_filebacked(struct loop_device *lo, struct bio *bio)
  */
 static void loop_add_bio(struct loop_device *lo, struct bio *bio)
 {
-	unsigned long flags;
-
-	spin_lock_irqsave(&lo->lo_lock, flags);
 	if (lo->lo_biotail) {
 		lo->lo_biotail->bi_next = bio;
 		lo->lo_biotail = bio;
 	} else
 		lo->lo_bio = lo->lo_biotail = bio;
-	spin_unlock_irqrestore(&lo->lo_lock, flags);
-
-	up(&lo->lo_bh_mutex);
 }
 
 /*
@@ -492,14 +486,12 @@ static struct bio *loop_get_bio(struct loop_device *lo)
 {
 	struct bio *bio;
 
-	spin_lock_irq(&lo->lo_lock);
 	if ((bio = lo->lo_bio)) {
 		if (bio == lo->lo_biotail)
 			lo->lo_biotail = NULL;
 		lo->lo_bio = bio->bi_next;
 		bio->bi_next = NULL;
 	}
-	spin_unlock_irq(&lo->lo_lock);
 
 	return bio;
 }
@@ -509,35 +501,28 @@ static int loop_make_request(request_queue_t *q, struct bio *old_bio)
 	struct loop_device *lo = q->queuedata;
 	int rw = bio_rw(old_bio);
 
-	if (!lo)
-		goto out;
+	if (rw == READA)
+		rw = READ;
+
+	BUG_ON(!lo || (rw != READ && rw != WRITE));
 
 	spin_lock_irq(&lo->lo_lock);
 	if (lo->lo_state != Lo_bound)
-		goto inactive;
-	atomic_inc(&lo->lo_pending);
-	spin_unlock_irq(&lo->lo_lock);
-
-	if (rw == WRITE) {
-		if (lo->lo_flags & LO_FLAGS_READ_ONLY)
-			goto err;
-	} else if (rw == READA) {
-		rw = READ;
-	} else if (rw != READ) {
-		printk(KERN_ERR "loop: unknown command (%x)\n", rw);
-		goto err;
-	}
+		goto out;
+	if (unlikely(rw == WRITE && (lo->lo_flags & LO_FLAGS_READ_ONLY)))
+		goto out;
+	lo->lo_pending++;
 	loop_add_bio(lo, old_bio);
+	spin_unlock_irq(&lo->lo_lock);
+	up(&lo->lo_bh_mutex);
 	return 0;
-err:
-	if (atomic_dec_and_test(&lo->lo_pending))
-		up(&lo->lo_bh_mutex);
+
 out:
+	if (lo->lo_pending == 0)
+		up(&lo->lo_bh_mutex);
+	spin_unlock_irq(&lo->lo_lock);
 	bio_io_error(old_bio, old_bio->bi_size);
 	return 0;
-inactive:
-	spin_unlock_irq(&lo->lo_lock);
-	goto out;
 }
 
 /*
@@ -560,13 +545,11 @@ static void do_loop_switch(struct loop_device *, struct switch_request *);
 
 static inline void loop_handle_bio(struct loop_device *lo, struct bio *bio)
 {
-	int ret;
-
 	if (unlikely(!bio->bi_bdev)) {
 		do_loop_switch(lo, bio->bi_private);
 		bio_put(bio);
 	} else {
-		ret = do_bio_filebacked(lo, bio);
+		int ret = do_bio_filebacked(lo, bio);
 		bio_endio(bio, bio->bi_size, ret);
 	}
 }
@@ -594,7 +577,7 @@ static int loop_thread(void *data)
 	set_user_nice(current, -20);
 
 	lo->lo_state = Lo_bound;
-	atomic_inc(&lo->lo_pending);
+	lo->lo_pending = 1;
 
 	/*
 	 * up sem, we are running
@@ -602,26 +585,37 @@ static int loop_thread(void *data)
 	up(&lo->lo_sem);
 
 	for (;;) {
-		down_interruptible(&lo->lo_bh_mutex);
+		int pending;
+
 		/*
-		 * could be upped because of tear-down, not because of
-		 * pending work
+		 * interruptible just to not contribute to load avg
 		 */
-		if (!atomic_read(&lo->lo_pending))
+		if (down_interruptible(&lo->lo_bh_mutex))
+			continue;
+
+		spin_lock_irq(&lo->lo_lock);
+
+		/*
+		 * could be upped because of tear-down, not pending work
+		 */
+		if (unlikely(!lo->lo_pending)) {
+			spin_unlock_irq(&lo->lo_lock);
 			break;
+		}
 
 		bio = loop_get_bio(lo);
-		if (!bio) {
-			printk("loop: missing bio\n");
-			continue;
-		}
+		lo->lo_pending--;
+		pending = lo->lo_pending;
+		spin_unlock_irq(&lo->lo_lock);
+
+		BUG_ON(!bio);
 		loop_handle_bio(lo, bio);
 
 		/*
 		 * upped both for pending work and tear-down, lo_pending
 		 * will hit zero then
 		 */
-		if (atomic_dec_and_test(&lo->lo_pending))
+		if (unlikely(!pending))
 			break;
 	}
 
@@ -900,7 +894,8 @@ static int loop_clr_fd(struct loop_device *lo, struct block_device *bdev)
 
 	spin_lock_irq(&lo->lo_lock);
 	lo->lo_state = Lo_rundown;
-	if (atomic_dec_and_test(&lo->lo_pending))
+	lo->lo_pending--;
+	if (!lo->lo_pending)
 		up(&lo->lo_bh_mutex);
 	spin_unlock_irq(&lo->lo_lock);
 
diff --git a/drivers/block/paride/pd.c b/drivers/block/paride/pd.c
index 202a5a74ad37..fa49d62626ba 100644
--- a/drivers/block/paride/pd.c
+++ b/drivers/block/paride/pd.c
@@ -723,7 +723,7 @@ static int pd_special_command(struct pd_unit *disk,
 	rq.ref_count = 1;
 	rq.waiting = &wait;
 	rq.end_io = blk_end_sync_rq;
-	blk_insert_request(disk->gd->queue, &rq, 0, func, 0);
+	blk_insert_request(disk->gd->queue, &rq, 0, func);
 	wait_for_completion(&wait);
 	rq.waiting = NULL;
 	if (rq.errors)
diff --git a/drivers/block/paride/pg.c b/drivers/block/paride/pg.c
index dbeb107bb971..84d8e291ed96 100644
--- a/drivers/block/paride/pg.c
+++ b/drivers/block/paride/pg.c
@@ -222,7 +222,7 @@ static int pg_identify(struct pg *dev, int log);
 
 static char pg_scratch[512];	/* scratch block buffer */
 
-static struct class_simple *pg_class;
+static struct class *pg_class;
 
 /* kernel glue structures */
 
@@ -666,7 +666,7 @@ static int __init pg_init(void)
 		err = -1;
 		goto out;
 	}
-	pg_class = class_simple_create(THIS_MODULE, "pg");
+	pg_class = class_create(THIS_MODULE, "pg");
 	if (IS_ERR(pg_class)) {
 		err = PTR_ERR(pg_class);
 		goto out_chrdev;
@@ -675,7 +675,7 @@ static int __init pg_init(void)
 	for (unit = 0; unit < PG_UNITS; unit++) {
 		struct pg *dev = &devices[unit];
 		if (dev->present) {
-			class_simple_device_add(pg_class, MKDEV(major, unit), 
+			class_device_create(pg_class, MKDEV(major, unit),
 					NULL, "pg%u", unit);
 			err = devfs_mk_cdev(MKDEV(major, unit),
 				      S_IFCHR | S_IRUSR | S_IWUSR, "pg/%u",
@@ -688,8 +688,8 @@ static int __init pg_init(void)
 	goto out;
 
 out_class:
-	class_simple_device_remove(MKDEV(major, unit));
-	class_simple_destroy(pg_class);
+	class_device_destroy(pg_class, MKDEV(major, unit));
+	class_destroy(pg_class);
 out_chrdev:
 	unregister_chrdev(major, "pg");
 out:
@@ -703,11 +703,11 @@ static void __exit pg_exit(void)
 	for (unit = 0; unit < PG_UNITS; unit++) {
 		struct pg *dev = &devices[unit];
 		if (dev->present) {
-			class_simple_device_remove(MKDEV(major, unit));
+			class_device_destroy(pg_class, MKDEV(major, unit));
 			devfs_remove("pg/%u", unit);
 		}
 	}
-	class_simple_destroy(pg_class);
+	class_destroy(pg_class);
 	devfs_remove("pg");
 	unregister_chrdev(major, name);
 
diff --git a/drivers/block/paride/pt.c b/drivers/block/paride/pt.c
index 8fbd6922fe0d..5fe8ee86f095 100644
--- a/drivers/block/paride/pt.c
+++ b/drivers/block/paride/pt.c
@@ -242,7 +242,7 @@ static struct file_operations pt_fops = {
 };
 
 /* sysfs class support */
-static struct class_simple *pt_class;
+static struct class *pt_class;
 
 static inline int status_reg(struct pi_adapter *pi)
 {
@@ -963,7 +963,7 @@ static int __init pt_init(void)
 		err = -1;
 		goto out;
 	}
-	pt_class = class_simple_create(THIS_MODULE, "pt");
+	pt_class = class_create(THIS_MODULE, "pt");
 	if (IS_ERR(pt_class)) {
 		err = PTR_ERR(pt_class);
 		goto out_chrdev;
@@ -972,29 +972,29 @@ static int __init pt_init(void)
 	devfs_mk_dir("pt");
 	for (unit = 0; unit < PT_UNITS; unit++)
 		if (pt[unit].present) {
-			class_simple_device_add(pt_class, MKDEV(major, unit), 
+			class_device_create(pt_class, MKDEV(major, unit),
 					NULL, "pt%d", unit);
 			err = devfs_mk_cdev(MKDEV(major, unit),
 				      S_IFCHR | S_IRUSR | S_IWUSR,
 				      "pt/%d", unit);
 			if (err) {
-				class_simple_device_remove(MKDEV(major, unit));
+				class_device_destroy(pt_class, MKDEV(major, unit));
 				goto out_class;
 			}
-			class_simple_device_add(pt_class, MKDEV(major, unit + 128),
+			class_device_create(pt_class, MKDEV(major, unit + 128),
 					NULL, "pt%dn", unit);
 			err = devfs_mk_cdev(MKDEV(major, unit + 128),
 				      S_IFCHR | S_IRUSR | S_IWUSR,
 				      "pt/%dn", unit);
 			if (err) {
-				class_simple_device_remove(MKDEV(major, unit + 128));
+				class_device_destroy(pt_class, MKDEV(major, unit + 128));
 				goto out_class;
 			}
 		}
 	goto out;
 
 out_class:
-	class_simple_destroy(pt_class);
+	class_destroy(pt_class);
 out_chrdev:
 	unregister_chrdev(major, "pt");
 out:
@@ -1006,12 +1006,12 @@ static void __exit pt_exit(void)
 	int unit;
 	for (unit = 0; unit < PT_UNITS; unit++)
 		if (pt[unit].present) {
-			class_simple_device_remove(MKDEV(major, unit));
+			class_device_destroy(pt_class, MKDEV(major, unit));
 			devfs_remove("pt/%d", unit);
-			class_simple_device_remove(MKDEV(major, unit + 128));
+			class_device_destroy(pt_class, MKDEV(major, unit + 128));
 			devfs_remove("pt/%dn", unit);
 		}
-	class_simple_destroy(pt_class);
+	class_destroy(pt_class);
 	devfs_remove("pt");
 	unregister_chrdev(major, name);
 	for (unit = 0; unit < PT_UNITS; unit++)
diff --git a/drivers/block/pktcdvd.c b/drivers/block/pktcdvd.c
index bc56770bcc90..7b838342f0a3 100644
--- a/drivers/block/pktcdvd.c
+++ b/drivers/block/pktcdvd.c
@@ -467,14 +467,12 @@ static int pkt_set_speed(struct pktcdvd_device *pd, unsigned write_speed, unsign
  * Queue a bio for processing by the low-level CD device. Must be called
  * from process context.
  */
-static void pkt_queue_bio(struct pktcdvd_device *pd, struct bio *bio, int high_prio_read)
+static void pkt_queue_bio(struct pktcdvd_device *pd, struct bio *bio)
 {
 	spin_lock(&pd->iosched.lock);
 	if (bio_data_dir(bio) == READ) {
 		pkt_add_list_last(bio, &pd->iosched.read_queue,
 				  &pd->iosched.read_queue_tail);
-		if (high_prio_read)
-			pd->iosched.high_prio_read = 1;
 	} else {
 		pkt_add_list_last(bio, &pd->iosched.write_queue,
 				  &pd->iosched.write_queue_tail);
@@ -490,15 +488,16 @@ static void pkt_queue_bio(struct pktcdvd_device *pd, struct bio *bio, int high_p
  * requirements for CDRW drives:
  * - A cache flush command must be inserted before a read request if the
  *   previous request was a write.
- * - Switching between reading and writing is slow, so don't it more often
+ * - Switching between reading and writing is slow, so don't do it more often
  *   than necessary.
+ * - Optimize for throughput at the expense of latency. This means that streaming
+ *   writes will never be interrupted by a read, but if the drive has to seek
+ *   before the next write, switch to reading instead if there are any pending
+ *   read requests.
  * - Set the read speed according to current usage pattern. When only reading
  *   from the device, it's best to use the highest possible read speed, but
  *   when switching often between reading and writing, it's better to have the
  *   same read and write speeds.
- * - Reads originating from user space should have higher priority than reads
- *   originating from pkt_gather_data, because some process is usually waiting
- *   on reads of the first kind.
  */
 static void pkt_iosched_process_queue(struct pktcdvd_device *pd)
 {
@@ -512,21 +511,24 @@ static void pkt_iosched_process_queue(struct pktcdvd_device *pd)
 
 	for (;;) {
 		struct bio *bio;
-		int reads_queued, writes_queued, high_prio_read;
+		int reads_queued, writes_queued;
 
 		spin_lock(&pd->iosched.lock);
 		reads_queued = (pd->iosched.read_queue != NULL);
 		writes_queued = (pd->iosched.write_queue != NULL);
-		if (!reads_queued)
-			pd->iosched.high_prio_read = 0;
-		high_prio_read = pd->iosched.high_prio_read;
 		spin_unlock(&pd->iosched.lock);
 
 		if (!reads_queued && !writes_queued)
 			break;
 
 		if (pd->iosched.writing) {
-			if (high_prio_read || (!writes_queued && reads_queued)) {
+			int need_write_seek = 1;
+			spin_lock(&pd->iosched.lock);
+			bio = pd->iosched.write_queue;
+			spin_unlock(&pd->iosched.lock);
+			if (bio && (bio->bi_sector == pd->iosched.last_write))
+				need_write_seek = 0;
+			if (need_write_seek && reads_queued) {
 				if (atomic_read(&pd->cdrw.pending_bios) > 0) {
 					VPRINTK("pktcdvd: write, waiting\n");
 					break;
@@ -559,8 +561,10 @@ static void pkt_iosched_process_queue(struct pktcdvd_device *pd)
 
 		if (bio_data_dir(bio) == READ)
 			pd->iosched.successive_reads += bio->bi_size >> 10;
-		else
+		else {
 			pd->iosched.successive_reads = 0;
+			pd->iosched.last_write = bio->bi_sector + bio_sectors(bio);
+		}
 		if (pd->iosched.successive_reads >= HI_SPEED_SWITCH) {
 			if (pd->read_speed == pd->write_speed) {
 				pd->read_speed = MAX_SPEED;
@@ -765,7 +769,7 @@ static void pkt_gather_data(struct pktcdvd_device *pd, struct packet_data *pkt)
 
 		atomic_inc(&pkt->io_wait);
 		bio->bi_rw = READ;
-		pkt_queue_bio(pd, bio, 0);
+		pkt_queue_bio(pd, bio);
 		frames_read++;
 	}
 
@@ -1062,7 +1066,7 @@ static void pkt_start_write(struct pktcdvd_device *pd, struct packet_data *pkt)
 
 	atomic_set(&pkt->io_wait, 1);
 	pkt->w_bio->bi_rw = WRITE;
-	pkt_queue_bio(pd, pkt->w_bio, 0);
+	pkt_queue_bio(pd, pkt->w_bio);
 }
 
 static void pkt_finish_packet(struct packet_data *pkt, int uptodate)
@@ -1247,8 +1251,7 @@ static int kcdrwd(void *foobar)
 			VPRINTK("kcdrwd: wake up\n");
 
 			/* make swsusp happy with our thread */
-			if (current->flags & PF_FREEZE)
-				refrigerator(PF_FREEZE);
+			try_to_freeze();
 
 			list_for_each_entry(pkt, &pd->cdrw.pkt_active_list, list) {
 				if (!pkt->sleep_time)
@@ -2120,7 +2123,7 @@ static int pkt_make_request(request_queue_t *q, struct bio *bio)
 		cloned_bio->bi_private = psd;
 		cloned_bio->bi_end_io = pkt_end_io_read_cloned;
 		pd->stats.secs_r += bio->bi_size >> 9;
-		pkt_queue_bio(pd, cloned_bio, 1);
+		pkt_queue_bio(pd, cloned_bio);
 		return 0;
 	}
 
diff --git a/drivers/block/sx8.c b/drivers/block/sx8.c
index 797f5988c2b5..5ed3a6379452 100644
--- a/drivers/block/sx8.c
+++ b/drivers/block/sx8.c
@@ -614,7 +614,7 @@ static int carm_array_info (struct carm_host *host, unsigned int array_idx)
 	spin_unlock_irq(&host->lock);
 
 	DPRINTK("blk_insert_request, tag == %u\n", idx);
-	blk_insert_request(host->oob_q, crq->rq, 1, crq, 0);
+	blk_insert_request(host->oob_q, crq->rq, 1, crq);
 
 	return 0;
 
@@ -653,7 +653,7 @@ static int carm_send_special (struct carm_host *host, carm_sspc_t func)
 	crq->msg_bucket = (u32) rc;
 
 	DPRINTK("blk_insert_request, tag == %u\n", idx);
-	blk_insert_request(host->oob_q, crq->rq, 1, crq, 0);
+	blk_insert_request(host->oob_q, crq->rq, 1, crq);
 
 	return 0;
 }
diff --git a/drivers/block/ub.c b/drivers/block/ub.c
index ce42889f98fb..685f061e69b2 100644
--- a/drivers/block/ub.c
+++ b/drivers/block/ub.c
@@ -8,13 +8,12 @@
  * and is not licensed separately. See file COPYING for details.
  *
  * TODO (sorted by decreasing priority)
+ *  -- Kill first_open (Al Viro fixed the block layer now)
  *  -- Do resets with usb_device_reset (needs a thread context, use khubd)
  *  -- set readonly flag for CDs, set removable flag for CF readers
  *  -- do inquiry and verify we got a disk and not a tape (for LUN mismatch)
- *  -- support pphaneuf's SDDR-75 with two LUNs (also broken capacity...)
  *  -- special case some senses, e.g. 3a/0 -> no media present, reduce retries
  *  -- verify the 13 conditions and do bulk resets
- *  -- normal pool of commands instead of cmdv[]?
  *  -- kill last_pipe and simply do two-state clearing on both pipes
  *  -- verify protocol (bulk) from USB descriptors (maybe...)
  *  -- highmem and sg
@@ -49,7 +48,14 @@
 #define US_SC_SCSI	0x06		/* Transparent */
 
 /*
+ * This many LUNs per USB device.
+ * Every one of them takes a host, see UB_MAX_HOSTS.
  */
+#define UB_MAX_LUNS   9
+
+/*
+ */
+
 #define UB_MINORS_PER_MAJOR	8
 
 #define UB_MAX_CDB_SIZE      16		/* Corresponds to Bulk */
@@ -65,7 +71,7 @@ struct bulk_cb_wrap {
 	u32	Tag;			/* unique per command id */
 	__le32	DataTransferLength;	/* size of data */
 	u8	Flags;			/* direction in bit 0 */
-	u8	Lun;			/* LUN normally 0 */
+	u8	Lun;			/* LUN */
 	u8	Length;			/* of of the CDB */
 	u8	CDB[UB_MAX_CDB_SIZE];	/* max command */
 };
@@ -168,6 +174,7 @@ struct ub_scsi_cmd {
 	unsigned int len;		/* Requested length */
 	// struct scatterlist sgv[UB_MAX_REQ_SG];
 
+	struct ub_lun *lun;
 	void (*done)(struct ub_dev *, struct ub_scsi_cmd *);
 	void *back;
 };
@@ -252,25 +259,47 @@ struct ub_scsi_cmd_queue {
 };
 
 /*
- * The UB device instance.
+ * The block device instance (one per LUN).
+ */
+struct ub_lun {
+	struct ub_dev *udev;
+	struct list_head link;
+	struct gendisk *disk;
+	int id;				/* Host index */
+	int num;			/* LUN number */
+	char name[16];
+
+	int changed;			/* Media was changed */
+	int removable;
+	int readonly;
+	int first_open;			/* Kludge. See ub_bd_open. */
+
+	/* Use Ingo's mempool if or when we have more than one command. */
+	/*
+	 * Currently we never need more than one command for the whole device.
+	 * However, giving every LUN a command is a cheap and automatic way
+	 * to enforce fairness between them.
+	 */
+	int cmda[1];
+	struct ub_scsi_cmd cmdv[1];
+
+	struct ub_capacity capacity; 
+};
+
+/*
+ * The USB device instance.
  */
 struct ub_dev {
 	spinlock_t lock;
-	int id;				/* Number among ub's */
 	atomic_t poison;		/* The USB device is disconnected */
 	int openc;			/* protected by ub_lock! */
 					/* kref is too implicit for our taste */
 	unsigned int tagcnt;
-	int changed;			/* Media was changed */
-	int removable;
-	int readonly;
-	int first_open;			/* Kludge. See ub_bd_open. */
-	char name[8];
+	char name[12];
 	struct usb_device *dev;
 	struct usb_interface *intf;
 
-	struct ub_capacity capacity; 
-	struct gendisk *disk;
+	struct list_head luns;
 
 	unsigned int send_bulk_pipe;	/* cached pipe values */
 	unsigned int recv_bulk_pipe;
@@ -279,10 +308,6 @@ struct ub_dev {
 
 	struct tasklet_struct tasklet;
 
-	/* XXX Use Ingo's mempool (once we have more than one) */
-	int cmda[1];
-	struct ub_scsi_cmd cmdv[1];
-
 	struct ub_scsi_cmd_queue cmd_queue;
 	struct ub_scsi_cmd top_rqs_cmd;	/* REQUEST SENSE */
 	unsigned char top_sense[UB_SENSE_SIZE];
@@ -301,9 +326,9 @@ struct ub_dev {
 /*
  */
 static void ub_cleanup(struct ub_dev *sc);
-static int ub_bd_rq_fn_1(struct ub_dev *sc, struct request *rq);
-static int ub_cmd_build_block(struct ub_dev *sc, struct ub_scsi_cmd *cmd,
-    struct request *rq);
+static int ub_bd_rq_fn_1(struct ub_lun *lun, struct request *rq);
+static int ub_cmd_build_block(struct ub_dev *sc, struct ub_lun *lun,
+    struct ub_scsi_cmd *cmd, struct request *rq);
 static int ub_cmd_build_packet(struct ub_dev *sc, struct ub_scsi_cmd *cmd,
     struct request *rq);
 static void ub_rw_cmd_done(struct ub_dev *sc, struct ub_scsi_cmd *cmd);
@@ -320,8 +345,10 @@ static void ub_state_sense(struct ub_dev *sc, struct ub_scsi_cmd *cmd);
 static int ub_submit_clear_stall(struct ub_dev *sc, struct ub_scsi_cmd *cmd,
     int stalled_pipe);
 static void ub_top_sense_done(struct ub_dev *sc, struct ub_scsi_cmd *scmd);
-static int ub_sync_tur(struct ub_dev *sc);
-static int ub_sync_read_cap(struct ub_dev *sc, struct ub_capacity *ret);
+static int ub_sync_tur(struct ub_dev *sc, struct ub_lun *lun);
+static int ub_sync_read_cap(struct ub_dev *sc, struct ub_lun *lun,
+    struct ub_capacity *ret);
+static int ub_probe_lun(struct ub_dev *sc, int lnum);
 
 /*
  */
@@ -342,6 +369,7 @@ MODULE_DEVICE_TABLE(usb, ub_usb_ids);
  */
 #define UB_MAX_HOSTS  26
 static char ub_hostv[UB_MAX_HOSTS];
+
 static DEFINE_SPINLOCK(ub_lock);	/* Locks globals and ->openc */
 
 /*
@@ -402,10 +430,12 @@ static void ub_cmdtr_sense(struct ub_dev *sc, struct ub_scsi_cmd *cmd,
 	}
 }
 
-static ssize_t ub_diag_show(struct device *dev, char *page)
+static ssize_t ub_diag_show(struct device *dev, struct device_attribute *attr, char *page)
 {
 	struct usb_interface *intf;
 	struct ub_dev *sc;
+	struct list_head *p;
+	struct ub_lun *lun;
 	int cnt;
 	unsigned long flags;
 	int nc, nh;
@@ -421,9 +451,15 @@ static ssize_t ub_diag_show(struct device *dev, char *page)
 	spin_lock_irqsave(&sc->lock, flags);
 
 	cnt += sprintf(page + cnt,
-	    "qlen %d qmax %d changed %d removable %d readonly %d\n",
-	    sc->cmd_queue.qlen, sc->cmd_queue.qmax,
-	    sc->changed, sc->removable, sc->readonly);
+	    "qlen %d qmax %d\n",
+	    sc->cmd_queue.qlen, sc->cmd_queue.qmax);
+
+	list_for_each (p, &sc->luns) {
+		lun = list_entry(p, struct ub_lun, link);
+		cnt += sprintf(page + cnt,
+		    "lun %u changed %d removable %d readonly %d\n",
+		    lun->num, lun->changed, lun->removable, lun->readonly);
+	}
 
 	if ((nc = sc->tr.cur + 1) == SCMD_TRACE_SZ) nc = 0;
 	for (j = 0; j < SCMD_TRACE_SZ; j++) {
@@ -523,53 +559,63 @@ static void ub_put(struct ub_dev *sc)
  */
 static void ub_cleanup(struct ub_dev *sc)
 {
+	struct list_head *p;
+	struct ub_lun *lun;
 	request_queue_t *q;
 
-	/* I don't think queue can be NULL. But... Stolen from sx8.c */
-	if ((q = sc->disk->queue) != NULL)
-		blk_cleanup_queue(q);
+	while (!list_empty(&sc->luns)) {
+		p = sc->luns.next;
+		lun = list_entry(p, struct ub_lun, link);
+		list_del(p);
 
-	/*
-	 * If we zero disk->private_data BEFORE put_disk, we have to check
-	 * for NULL all over the place in open, release, check_media and
-	 * revalidate, because the block level semaphore is well inside the
-	 * put_disk. But we cannot zero after the call, because *disk is gone.
-	 * The sd.c is blatantly racy in this area.
-	 */
-	/* disk->private_data = NULL; */
-	put_disk(sc->disk);
-	sc->disk = NULL;
+		/* I don't think queue can be NULL. But... Stolen from sx8.c */
+		if ((q = lun->disk->queue) != NULL)
+			blk_cleanup_queue(q);
+		/*
+		 * If we zero disk->private_data BEFORE put_disk, we have
+		 * to check for NULL all over the place in open, release,
+		 * check_media and revalidate, because the block level
+		 * semaphore is well inside the put_disk.
+		 * But we cannot zero after the call, because *disk is gone.
+		 * The sd.c is blatantly racy in this area.
+		 */
+		/* disk->private_data = NULL; */
+		put_disk(lun->disk);
+		lun->disk = NULL;
+
+		ub_id_put(lun->id);
+		kfree(lun);
+	}
 
-	ub_id_put(sc->id);
 	kfree(sc);
 }
 
 /*
  * The "command allocator".
  */
-static struct ub_scsi_cmd *ub_get_cmd(struct ub_dev *sc)
+static struct ub_scsi_cmd *ub_get_cmd(struct ub_lun *lun)
 {
 	struct ub_scsi_cmd *ret;
 
-	if (sc->cmda[0])
+	if (lun->cmda[0])
 		return NULL;
-	ret = &sc->cmdv[0];
-	sc->cmda[0] = 1;
+	ret = &lun->cmdv[0];
+	lun->cmda[0] = 1;
 	return ret;
 }
 
-static void ub_put_cmd(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
+static void ub_put_cmd(struct ub_lun *lun, struct ub_scsi_cmd *cmd)
 {
-	if (cmd != &sc->cmdv[0]) {
+	if (cmd != &lun->cmdv[0]) {
 		printk(KERN_WARNING "%s: releasing a foreign cmd %p\n",
-		    sc->name, cmd);
+		    lun->name, cmd);
 		return;
 	}
-	if (!sc->cmda[0]) {
-		printk(KERN_WARNING "%s: releasing a free cmd\n", sc->name);
+	if (!lun->cmda[0]) {
+		printk(KERN_WARNING "%s: releasing a free cmd\n", lun->name);
 		return;
 	}
-	sc->cmda[0] = 0;
+	lun->cmda[0] = 0;
 }
 
 /*
@@ -630,29 +676,30 @@ static struct ub_scsi_cmd *ub_cmdq_pop(struct ub_dev *sc)
 
 static void ub_bd_rq_fn(request_queue_t *q)
 {
-	struct ub_dev *sc = q->queuedata;
+	struct ub_lun *lun = q->queuedata;
 	struct request *rq;
 
 	while ((rq = elv_next_request(q)) != NULL) {
-		if (ub_bd_rq_fn_1(sc, rq) != 0) {
+		if (ub_bd_rq_fn_1(lun, rq) != 0) {
 			blk_stop_queue(q);
 			break;
 		}
 	}
 }
 
-static int ub_bd_rq_fn_1(struct ub_dev *sc, struct request *rq)
+static int ub_bd_rq_fn_1(struct ub_lun *lun, struct request *rq)
 {
+	struct ub_dev *sc = lun->udev;
 	struct ub_scsi_cmd *cmd;
 	int rc;
 
-	if (atomic_read(&sc->poison) || sc->changed) {
+	if (atomic_read(&sc->poison) || lun->changed) {
 		blkdev_dequeue_request(rq);
 		ub_end_rq(rq, 0);
 		return 0;
 	}
 
-	if ((cmd = ub_get_cmd(sc)) == NULL)
+	if ((cmd = ub_get_cmd(lun)) == NULL)
 		return -1;
 	memset(cmd, 0, sizeof(struct ub_scsi_cmd));
 
@@ -661,32 +708,30 @@ static int ub_bd_rq_fn_1(struct ub_dev *sc, struct request *rq)
 	if (blk_pc_request(rq)) {
 		rc = ub_cmd_build_packet(sc, cmd, rq);
 	} else {
-		rc = ub_cmd_build_block(sc, cmd, rq);
+		rc = ub_cmd_build_block(sc, lun, cmd, rq);
 	}
 	if (rc != 0) {
-		ub_put_cmd(sc, cmd);
+		ub_put_cmd(lun, cmd);
 		ub_end_rq(rq, 0);
-		blk_start_queue(sc->disk->queue);
 		return 0;
 	}
-
 	cmd->state = UB_CMDST_INIT;
+	cmd->lun = lun;
 	cmd->done = ub_rw_cmd_done;
 	cmd->back = rq;
 
 	cmd->tag = sc->tagcnt++;
 	if ((rc = ub_submit_scsi(sc, cmd)) != 0) {
-		ub_put_cmd(sc, cmd);
+		ub_put_cmd(lun, cmd);
 		ub_end_rq(rq, 0);
-		blk_start_queue(sc->disk->queue);
 		return 0;
 	}
 
 	return 0;
 }
 
-static int ub_cmd_build_block(struct ub_dev *sc, struct ub_scsi_cmd *cmd,
-    struct request *rq)
+static int ub_cmd_build_block(struct ub_dev *sc, struct ub_lun *lun,
+    struct ub_scsi_cmd *cmd, struct request *rq)
 {
 	int ub_dir;
 #if 0 /* We use rq->buffer for now */
@@ -707,7 +752,7 @@ static int ub_cmd_build_block(struct ub_dev *sc, struct ub_scsi_cmd *cmd,
 	sg = &cmd->sgv[0];
 	n_elem = blk_rq_map_sg(q, rq, sg);
 	if (n_elem <= 0) {
-		ub_put_cmd(sc, cmd);
+		ub_put_cmd(lun, cmd);
 		ub_end_rq(rq, 0);
 		blk_start_queue(q);
 		return 0;		/* request with no s/g entries? */
@@ -716,7 +761,7 @@ static int ub_cmd_build_block(struct ub_dev *sc, struct ub_scsi_cmd *cmd,
 	if (n_elem != 1) {		/* Paranoia */
 		printk(KERN_WARNING "%s: request with %d segments\n",
 		    sc->name, n_elem);
-		ub_put_cmd(sc, cmd);
+		ub_put_cmd(lun, cmd);
 		ub_end_rq(rq, 0);
 		blk_start_queue(q);
 		return 0;
@@ -748,8 +793,8 @@ static int ub_cmd_build_block(struct ub_dev *sc, struct ub_scsi_cmd *cmd,
 	 * The call to blk_queue_hardsect_size() guarantees that request
 	 * is aligned, but it is given in terms of 512 byte units, always.
 	 */
-	block = rq->sector >> sc->capacity.bshift;
-	nblks = rq->nr_sectors >> sc->capacity.bshift;
+	block = rq->sector >> lun->capacity.bshift;
+	nblks = rq->nr_sectors >> lun->capacity.bshift;
 
 	cmd->cdb[0] = (ub_dir == UB_DIR_READ)? READ_10: WRITE_10;
 	/* 10-byte uses 4 bytes of LBA: 2147483648KB, 2097152MB, 2048GB */
@@ -803,7 +848,8 @@ static int ub_cmd_build_packet(struct ub_dev *sc, struct ub_scsi_cmd *cmd,
 static void ub_rw_cmd_done(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 {
 	struct request *rq = cmd->back;
-	struct gendisk *disk = sc->disk;
+	struct ub_lun *lun = cmd->lun;
+	struct gendisk *disk = lun->disk;
 	request_queue_t *q = disk->queue;
 	int uptodate;
 
@@ -818,7 +864,7 @@ static void ub_rw_cmd_done(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 	else
 		uptodate = 0;
 
-	ub_put_cmd(sc, cmd);
+	ub_put_cmd(lun, cmd);
 	ub_end_rq(rq, uptodate);
 	blk_start_queue(q);
 }
@@ -887,7 +933,7 @@ static int ub_scsi_cmd_start(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 	bcb->Tag = cmd->tag;		/* Endianness is not important */
 	bcb->DataTransferLength = cpu_to_le32(cmd->len);
 	bcb->Flags = (cmd->dir == UB_DIR_READ) ? 0x80 : 0;
-	bcb->Lun = 0;			/* No multi-LUN yet */
+	bcb->Lun = (cmd->lun != NULL) ? cmd->lun->num : 0;
 	bcb->Length = cmd->cdb_len;
 
 	/* copy the command payload */
@@ -1002,9 +1048,8 @@ static void ub_scsi_urb_compl(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 			 * The control pipe clears itself - nothing to do.
 			 * XXX Might try to reset the device here and retry.
 			 */
-			printk(KERN_NOTICE "%s: "
-			    "stall on control pipe for device %u\n",
-			    sc->name, sc->dev->devnum);
+			printk(KERN_NOTICE "%s: stall on control pipe\n",
+			    sc->name);
 			goto Bad_End;
 		}
 
@@ -1025,9 +1070,8 @@ static void ub_scsi_urb_compl(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 			 * The control pipe clears itself - nothing to do.
 			 * XXX Might try to reset the device here and retry.
 			 */
-			printk(KERN_NOTICE "%s: "
-			    "stall on control pipe for device %u\n",
-			    sc->name, sc->dev->devnum);
+			printk(KERN_NOTICE "%s: stall on control pipe\n",
+			    sc->name);
 			goto Bad_End;
 		}
 
@@ -1046,9 +1090,8 @@ static void ub_scsi_urb_compl(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 			rc = ub_submit_clear_stall(sc, cmd, sc->last_pipe);
 			if (rc != 0) {
 				printk(KERN_NOTICE "%s: "
-				    "unable to submit clear for device %u"
-				    " (code %d)\n",
-				    sc->name, sc->dev->devnum, rc);
+				    "unable to submit clear (%d)\n",
+				    sc->name, rc);
 				/*
 				 * This is typically ENOMEM or some other such shit.
 				 * Retrying is pointless. Just do Bad End on it...
@@ -1107,9 +1150,8 @@ static void ub_scsi_urb_compl(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 			rc = ub_submit_clear_stall(sc, cmd, sc->last_pipe);
 			if (rc != 0) {
 				printk(KERN_NOTICE "%s: "
-				    "unable to submit clear for device %u"
-				    " (code %d)\n",
-				    sc->name, sc->dev->devnum, rc);
+				    "unable to submit clear (%d)\n",
+				    sc->name, rc);
 				/*
 				 * This is typically ENOMEM or some other such shit.
 				 * Retrying is pointless. Just do Bad End on it...
@@ -1140,9 +1182,8 @@ static void ub_scsi_urb_compl(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 			rc = ub_submit_clear_stall(sc, cmd, sc->last_pipe);
 			if (rc != 0) {
 				printk(KERN_NOTICE "%s: "
-				    "unable to submit clear for device %u"
-				    " (code %d)\n",
-				    sc->name, sc->dev->devnum, rc);
+				    "unable to submit clear (%d)\n",
+				    sc->name, rc);
 				/*
 				 * This is typically ENOMEM or some other such shit.
 				 * Retrying is pointless. Just do Bad End on it...
@@ -1164,9 +1205,8 @@ static void ub_scsi_urb_compl(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 			 * encounter such a thing, try to read the CSW again.
 			 */
 			if (++cmd->stat_count >= 4) {
-				printk(KERN_NOTICE "%s: "
-				    "unable to get CSW on device %u\n",
-				    sc->name, sc->dev->devnum);
+				printk(KERN_NOTICE "%s: unable to get CSW\n",
+				    sc->name);
 				goto Bad_End;
 			}
 			__ub_state_stat(sc, cmd);
@@ -1207,10 +1247,8 @@ static void ub_scsi_urb_compl(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 			 */
 			if (++cmd->stat_count >= 4) {
 				printk(KERN_NOTICE "%s: "
-				    "tag mismatch orig 0x%x reply 0x%x "
-				    "on device %u\n",
-				    sc->name, cmd->tag, bcs->Tag,
-				    sc->dev->devnum);
+				    "tag mismatch orig 0x%x reply 0x%x\n",
+				    sc->name, cmd->tag, bcs->Tag);
 				goto Bad_End;
 			}
 			__ub_state_stat(sc, cmd);
@@ -1244,8 +1282,8 @@ static void ub_scsi_urb_compl(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 
 	} else {
 		printk(KERN_WARNING "%s: "
-		    "wrong command state %d on device %u\n",
-		    sc->name, cmd->state, sc->dev->devnum);
+		    "wrong command state %d\n",
+		    sc->name, cmd->state);
 		goto Bad_End;
 	}
 	return;
@@ -1288,7 +1326,6 @@ static void __ub_state_stat(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 
 	if ((rc = usb_submit_urb(&sc->work_urb, GFP_ATOMIC)) != 0) {
 		/* XXX Clear stalls */
-		printk("%s: CSW #%d submit failed (%d)\n", sc->name, cmd->tag, rc); /* P3 */
 		ub_complete(&sc->work_done);
 		ub_state_done(sc, cmd, rc);
 		return;
@@ -1333,6 +1370,7 @@ static void ub_state_sense(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 	scmd->state = UB_CMDST_INIT;
 	scmd->data = sc->top_sense;
 	scmd->len = UB_SENSE_SIZE;
+	scmd->lun = cmd->lun;
 	scmd->done = ub_top_sense_done;
 	scmd->back = cmd;
 
@@ -1411,14 +1449,14 @@ static void ub_top_sense_done(struct ub_dev *sc, struct ub_scsi_cmd *scmd)
 	}
 	if (cmd != scmd->back) {
 		printk(KERN_WARNING "%s: "
-		    "sense done for wrong command 0x%x on device %u\n",
-		    sc->name, cmd->tag, sc->dev->devnum);
+		    "sense done for wrong command 0x%x\n",
+		    sc->name, cmd->tag);
 		return;
 	}
 	if (cmd->state != UB_CMDST_SENSE) {
 		printk(KERN_WARNING "%s: "
-		    "sense done with bad cmd state %d on device %u\n",
-		    sc->name, cmd->state, sc->dev->devnum);
+		    "sense done with bad cmd state %d\n",
+		    sc->name, cmd->state);
 		return;
 	}
 
@@ -1429,68 +1467,32 @@ static void ub_top_sense_done(struct ub_dev *sc, struct ub_scsi_cmd *scmd)
 	ub_scsi_urb_compl(sc, cmd);
 }
 
-#if 0
-/* Determine what the maximum LUN supported is */
-int usb_stor_Bulk_max_lun(struct us_data *us)
-{
-	int result;
-
-	/* issue the command */
-	result = usb_stor_control_msg(us, us->recv_ctrl_pipe,
-				 US_BULK_GET_MAX_LUN, 
-				 USB_DIR_IN | USB_TYPE_CLASS | 
-				 USB_RECIP_INTERFACE,
-				 0, us->ifnum, us->iobuf, 1, HZ);
-
-	/* 
-	 * Some devices (i.e. Iomega Zip100) need this -- apparently
-	 * the bulk pipes get STALLed when the GetMaxLUN request is
-	 * processed.   This is, in theory, harmless to all other devices
-	 * (regardless of if they stall or not).
-	 */
-	if (result < 0) {
-		usb_stor_clear_halt(us, us->recv_bulk_pipe);
-		usb_stor_clear_halt(us, us->send_bulk_pipe);
-	}
-
-	US_DEBUGP("GetMaxLUN command result is %d, data is %d\n", 
-		  result, us->iobuf[0]);
-
-	/* if we have a successful request, return the result */
-	if (result == 1)
-		return us->iobuf[0];
-
-	/* return the default -- no LUNs */
-	return 0;
-}
-#endif
-
 /*
  * This is called from a process context.
  */
-static void ub_revalidate(struct ub_dev *sc)
+static void ub_revalidate(struct ub_dev *sc, struct ub_lun *lun)
 {
 
-	sc->readonly = 0;	/* XXX Query this from the device */
+	lun->readonly = 0;	/* XXX Query this from the device */
 
-	sc->capacity.nsec = 0;
-	sc->capacity.bsize = 512;
-	sc->capacity.bshift = 0;
+	lun->capacity.nsec = 0;
+	lun->capacity.bsize = 512;
+	lun->capacity.bshift = 0;
 
-	if (ub_sync_tur(sc) != 0)
+	if (ub_sync_tur(sc, lun) != 0)
 		return;			/* Not ready */
-	sc->changed = 0;
+	lun->changed = 0;
 
-	if (ub_sync_read_cap(sc, &sc->capacity) != 0) {
+	if (ub_sync_read_cap(sc, lun, &lun->capacity) != 0) {
 		/*
 		 * The retry here means something is wrong, either with the
 		 * device, with the transport, or with our code.
 		 * We keep this because sd.c has retries for capacity.
 		 */
-		if (ub_sync_read_cap(sc, &sc->capacity) != 0) {
-			sc->capacity.nsec = 0;
-			sc->capacity.bsize = 512;
-			sc->capacity.bshift = 0;
+		if (ub_sync_read_cap(sc, lun, &lun->capacity) != 0) {
+			lun->capacity.nsec = 0;
+			lun->capacity.bsize = 512;
+			lun->capacity.bshift = 0;
 		}
 	}
 }
@@ -1503,12 +1505,15 @@ static void ub_revalidate(struct ub_dev *sc)
 static int ub_bd_open(struct inode *inode, struct file *filp)
 {
 	struct gendisk *disk = inode->i_bdev->bd_disk;
+	struct ub_lun *lun;
 	struct ub_dev *sc;
 	unsigned long flags;
 	int rc;
 
-	if ((sc = disk->private_data) == NULL)
+	if ((lun = disk->private_data) == NULL)
 		return -ENXIO;
+	sc = lun->udev;
+
 	spin_lock_irqsave(&ub_lock, flags);
 	if (atomic_read(&sc->poison)) {
 		spin_unlock_irqrestore(&ub_lock, flags);
@@ -1529,15 +1534,15 @@ static int ub_bd_open(struct inode *inode, struct file *filp)
 	 * The bottom line is, Al Viro says that we should not allow
 	 * bdev->bd_invalidated to be set when doing add_disk no matter what.
 	 */
-	if (sc->first_open) {
-		if (sc->changed) {
-			sc->first_open = 0;
+	if (lun->first_open) {
+		lun->first_open = 0;
+		if (lun->changed) {
 			rc = -ENOMEDIUM;
 			goto err_open;
 		}
 	}
 
-	if (sc->removable || sc->readonly)
+	if (lun->removable || lun->readonly)
 		check_disk_change(inode->i_bdev);
 
 	/*
@@ -1545,12 +1550,12 @@ static int ub_bd_open(struct inode *inode, struct file *filp)
 	 * under some pretty murky conditions (a failure of READ CAPACITY).
 	 * We may need it one day.
 	 */
-	if (sc->removable && sc->changed && !(filp->f_flags & O_NDELAY)) {
+	if (lun->removable && lun->changed && !(filp->f_flags & O_NDELAY)) {
 		rc = -ENOMEDIUM;
 		goto err_open;
 	}
 
-	if (sc->readonly && (filp->f_mode & FMODE_WRITE)) {
+	if (lun->readonly && (filp->f_mode & FMODE_WRITE)) {
 		rc = -EROFS;
 		goto err_open;
 	}
@@ -1567,7 +1572,8 @@ err_open:
 static int ub_bd_release(struct inode *inode, struct file *filp)
 {
 	struct gendisk *disk = inode->i_bdev->bd_disk;
-	struct ub_dev *sc = disk->private_data;
+	struct ub_lun *lun = disk->private_data;
+	struct ub_dev *sc = lun->udev;
 
 	ub_put(sc);
 	return 0;
@@ -1597,20 +1603,14 @@ static int ub_bd_ioctl(struct inode *inode, struct file *filp,
  */
 static int ub_bd_revalidate(struct gendisk *disk)
 {
-	struct ub_dev *sc = disk->private_data;
-
-	ub_revalidate(sc);
-	/* This is pretty much a long term P3 */
-	if (!atomic_read(&sc->poison)) {		/* Cover sc->dev */
-		printk(KERN_INFO "%s: device %u capacity nsec %ld bsize %u\n",
-		    sc->name, sc->dev->devnum,
-		    sc->capacity.nsec, sc->capacity.bsize);
-	}
+	struct ub_lun *lun = disk->private_data;
+
+	ub_revalidate(lun->udev, lun);
 
 	/* XXX Support sector size switching like in sr.c */
-	blk_queue_hardsect_size(disk->queue, sc->capacity.bsize);
-	set_capacity(disk, sc->capacity.nsec);
-	// set_disk_ro(sdkp->disk, sc->readonly);
+	blk_queue_hardsect_size(disk->queue, lun->capacity.bsize);
+	set_capacity(disk, lun->capacity.nsec);
+	// set_disk_ro(sdkp->disk, lun->readonly);
 
 	return 0;
 }
@@ -1626,9 +1626,9 @@ static int ub_bd_revalidate(struct gendisk *disk)
  */
 static int ub_bd_media_changed(struct gendisk *disk)
 {
-	struct ub_dev *sc = disk->private_data;
+	struct ub_lun *lun = disk->private_data;
 
-	if (!sc->removable)
+	if (!lun->removable)
 		return 0;
 
 	/*
@@ -1640,12 +1640,12 @@ static int ub_bd_media_changed(struct gendisk *disk)
 	 * will fail, then block layer discards the data. Since we never
 	 * spin drives up, such devices simply cannot be used with ub anyway.
 	 */
-	if (ub_sync_tur(sc) != 0) {
-		sc->changed = 1;
+	if (ub_sync_tur(lun->udev, lun) != 0) {
+		lun->changed = 1;
 		return 1;
 	}
 
-	return sc->changed;
+	return lun->changed;
 }
 
 static struct block_device_operations ub_bd_fops = {
@@ -1669,7 +1669,7 @@ static void ub_probe_done(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
 /*
  * Test if the device has a check condition on it, synchronously.
  */
-static int ub_sync_tur(struct ub_dev *sc)
+static int ub_sync_tur(struct ub_dev *sc, struct ub_lun *lun)
 {
 	struct ub_scsi_cmd *cmd;
 	enum { ALLOC_SIZE = sizeof(struct ub_scsi_cmd) };
@@ -1688,6 +1688,7 @@ static int ub_sync_tur(struct ub_dev *sc)
 	cmd->cdb_len = 6;
 	cmd->dir = UB_DIR_NONE;
 	cmd->state = UB_CMDST_INIT;
+	cmd->lun = lun;			/* This may be NULL, but that's ok */
 	cmd->done = ub_probe_done;
 	cmd->back = &compl;
 
@@ -1718,7 +1719,8 @@ err_alloc:
 /*
  * Read the SCSI capacity synchronously (for probing).
  */
-static int ub_sync_read_cap(struct ub_dev *sc, struct ub_capacity *ret)
+static int ub_sync_read_cap(struct ub_dev *sc, struct ub_lun *lun,
+    struct ub_capacity *ret)
 {
 	struct ub_scsi_cmd *cmd;
 	char *p;
@@ -1743,6 +1745,7 @@ static int ub_sync_read_cap(struct ub_dev *sc, struct ub_capacity *ret)
 	cmd->state = UB_CMDST_INIT;
 	cmd->data = p;
 	cmd->len = 8;
+	cmd->lun = lun;
 	cmd->done = ub_probe_done;
 	cmd->back = &compl;
 
@@ -1812,6 +1815,90 @@ static void ub_probe_timeout(unsigned long arg)
 }
 
 /*
+ * Get number of LUNs by the way of Bulk GetMaxLUN command.
+ */
+static int ub_sync_getmaxlun(struct ub_dev *sc)
+{
+	int ifnum = sc->intf->cur_altsetting->desc.bInterfaceNumber;
+	unsigned char *p;
+	enum { ALLOC_SIZE = 1 };
+	struct usb_ctrlrequest *cr;
+	struct completion compl;
+	struct timer_list timer;
+	int nluns;
+	int rc;
+
+	init_completion(&compl);
+
+	rc = -ENOMEM;
+	if ((p = kmalloc(ALLOC_SIZE, GFP_KERNEL)) == NULL)
+		goto err_alloc;
+	*p = 55;
+
+	cr = &sc->work_cr;
+	cr->bRequestType = USB_DIR_IN | USB_TYPE_CLASS | USB_RECIP_INTERFACE;
+	cr->bRequest = US_BULK_GET_MAX_LUN;
+	cr->wValue = cpu_to_le16(0);
+	cr->wIndex = cpu_to_le16(ifnum);
+	cr->wLength = cpu_to_le16(1);
+
+	usb_fill_control_urb(&sc->work_urb, sc->dev, sc->recv_ctrl_pipe,
+	    (unsigned char*) cr, p, 1, ub_probe_urb_complete, &compl);
+	sc->work_urb.transfer_flags = 0;
+	sc->work_urb.actual_length = 0;
+	sc->work_urb.error_count = 0;
+	sc->work_urb.status = 0;
+
+	if ((rc = usb_submit_urb(&sc->work_urb, GFP_KERNEL)) != 0) {
+		if (rc == -EPIPE) {
+			printk("%s: Stall at GetMaxLUN, using 1 LUN\n",
+			     sc->name); /* P3 */
+		} else {
+			printk(KERN_WARNING
+			     "%s: Unable to submit GetMaxLUN (%d)\n",
+			     sc->name, rc);
+		}
+		goto err_submit;
+	}
+
+	init_timer(&timer);
+	timer.function = ub_probe_timeout;
+	timer.data = (unsigned long) &compl;
+	timer.expires = jiffies + UB_CTRL_TIMEOUT;
+	add_timer(&timer);
+
+	wait_for_completion(&compl);
+
+	del_timer_sync(&timer);
+	usb_kill_urb(&sc->work_urb);
+
+	if (sc->work_urb.actual_length != 1) {
+		printk("%s: GetMaxLUN returned %d bytes\n", sc->name,
+		    sc->work_urb.actual_length); /* P3 */
+		nluns = 0;
+	} else {
+		if ((nluns = *p) == 55) {
+			nluns = 0;
+		} else {
+  			/* GetMaxLUN returns the maximum LUN number */
+			nluns += 1;
+			if (nluns > UB_MAX_LUNS)
+				nluns = UB_MAX_LUNS;
+		}
+		printk("%s: GetMaxLUN returned %d, using %d LUNs\n", sc->name,
+		    *p, nluns); /* P3 */
+	}
+
+	kfree(p);
+	return nluns;
+
+err_submit:
+	kfree(p);
+err_alloc:
+	return rc;
+}
+
+/*
  * Clear initial stalls.
  */
 static int ub_probe_clear_stall(struct ub_dev *sc, int stalled_pipe)
@@ -1897,8 +1984,8 @@ static int ub_get_pipes(struct ub_dev *sc, struct usb_device *dev,
 	}
 
 	if (ep_in == NULL || ep_out == NULL) {
-		printk(KERN_NOTICE "%s: device %u failed endpoint check\n",
-		    sc->name, sc->dev->devnum);
+		printk(KERN_NOTICE "%s: failed endpoint check\n",
+		    sc->name);
 		return -EIO;
 	}
 
@@ -1921,8 +2008,7 @@ static int ub_probe(struct usb_interface *intf,
     const struct usb_device_id *dev_id)
 {
 	struct ub_dev *sc;
-	request_queue_t *q;
-	struct gendisk *disk;
+	int nluns;
 	int rc;
 	int i;
 
@@ -1931,6 +2017,7 @@ static int ub_probe(struct usb_interface *intf,
 		goto err_core;
 	memset(sc, 0, sizeof(struct ub_dev));
 	spin_lock_init(&sc->lock);
+	INIT_LIST_HEAD(&sc->luns);
 	usb_init_urb(&sc->work_urb);
 	tasklet_init(&sc->tasklet, ub_scsi_action, (unsigned long)sc);
 	atomic_set(&sc->poison, 0);
@@ -1942,19 +2029,16 @@ static int ub_probe(struct usb_interface *intf,
 	ub_init_completion(&sc->work_done);
 	sc->work_done.done = 1;		/* A little yuk, but oh well... */
 
-	rc = -ENOSR;
-	if ((sc->id = ub_id_get()) == -1)
-		goto err_id;
-	snprintf(sc->name, 8, DRV_NAME "%c", sc->id + 'a');
-
 	sc->dev = interface_to_usbdev(intf);
 	sc->intf = intf;
 	// sc->ifnum = intf->cur_altsetting->desc.bInterfaceNumber;
-
 	usb_set_intfdata(intf, sc);
 	usb_get_dev(sc->dev);
 	// usb_get_intf(sc->intf);	/* Do we need this? */
 
+	snprintf(sc->name, 12, DRV_NAME "(%d.%d)",
+	    sc->dev->bus->busnum, sc->dev->devnum);
+
 	/* XXX Verify that we can handle the device (from descriptors) */
 
 	ub_get_pipes(sc, sc->dev, intf);
@@ -1992,35 +2076,88 @@ static int ub_probe(struct usb_interface *intf,
 	 * In any case it's not our business how revaliadation is implemented.
 	 */
 	for (i = 0; i < 3; i++) {	/* Retries for benh's key */
-		if ((rc = ub_sync_tur(sc)) <= 0) break;
+		if ((rc = ub_sync_tur(sc, NULL)) <= 0) break;
 		if (rc != 0x6) break;
 		msleep(10);
 	}
 
-	sc->removable = 1;		/* XXX Query this from the device */
-	sc->changed = 1;		/* ub_revalidate clears only */
-	sc->first_open = 1;
+	nluns = 1;
+	for (i = 0; i < 3; i++) {
+		if ((rc = ub_sync_getmaxlun(sc)) < 0) {
+			/* 
+			 * Some devices (i.e. Iomega Zip100) need this --
+			 * apparently the bulk pipes get STALLed when the
+			 * GetMaxLUN request is processed.
+			 * XXX I have a ZIP-100, verify it does this.
+			 */
+			if (rc == -EPIPE) {
+				ub_probe_clear_stall(sc, sc->recv_bulk_pipe);
+				ub_probe_clear_stall(sc, sc->send_bulk_pipe);
+			}
+			break;
+		}
+		if (rc != 0) {
+			nluns = rc;
+			break;
+		}
+		msleep(100);
+	}
 
-	ub_revalidate(sc);
-	/* This is pretty much a long term P3 */
-	printk(KERN_INFO "%s: device %u capacity nsec %ld bsize %u\n",
-	    sc->name, sc->dev->devnum, sc->capacity.nsec, sc->capacity.bsize);
+	for (i = 0; i < nluns; i++) {
+		ub_probe_lun(sc, i);
+	}
+	return 0;
+
+	/* device_remove_file(&sc->intf->dev, &dev_attr_diag); */
+err_diag:
+	usb_set_intfdata(intf, NULL);
+	// usb_put_intf(sc->intf);
+	usb_put_dev(sc->dev);
+	kfree(sc);
+err_core:
+	return rc;
+}
+
+static int ub_probe_lun(struct ub_dev *sc, int lnum)
+{
+	struct ub_lun *lun;
+	request_queue_t *q;
+	struct gendisk *disk;
+	int rc;
+
+	rc = -ENOMEM;
+	if ((lun = kmalloc(sizeof(struct ub_lun), GFP_KERNEL)) == NULL)
+		goto err_alloc;
+	memset(lun, 0, sizeof(struct ub_lun));
+	lun->num = lnum;
+
+	rc = -ENOSR;
+	if ((lun->id = ub_id_get()) == -1)
+		goto err_id;
+
+	lun->udev = sc;
+	list_add(&lun->link, &sc->luns);
+
+	snprintf(lun->name, 16, DRV_NAME "%c(%d.%d.%d)",
+	    lun->id + 'a', sc->dev->bus->busnum, sc->dev->devnum, lun->num);
+
+	lun->removable = 1;		/* XXX Query this from the device */
+	lun->changed = 1;		/* ub_revalidate clears only */
+	lun->first_open = 1;
+	ub_revalidate(sc, lun);
 
-	/*
-	 * Just one disk per sc currently, but maybe more.
-	 */
 	rc = -ENOMEM;
 	if ((disk = alloc_disk(UB_MINORS_PER_MAJOR)) == NULL)
 		goto err_diskalloc;
 
-	sc->disk = disk;
-	sprintf(disk->disk_name, DRV_NAME "%c", sc->id + 'a');
-	sprintf(disk->devfs_name, DEVFS_NAME "/%c", sc->id + 'a');
+	lun->disk = disk;
+	sprintf(disk->disk_name, DRV_NAME "%c", lun->id + 'a');
+	sprintf(disk->devfs_name, DEVFS_NAME "/%c", lun->id + 'a');
 	disk->major = UB_MAJOR;
-	disk->first_minor = sc->id * UB_MINORS_PER_MAJOR;
+	disk->first_minor = lun->id * UB_MINORS_PER_MAJOR;
 	disk->fops = &ub_bd_fops;
-	disk->private_data = sc;
-	disk->driverfs_dev = &intf->dev;
+	disk->private_data = lun;
+	disk->driverfs_dev = &sc->intf->dev;	/* XXX Many to one ok? */
 
 	rc = -ENOMEM;
 	if ((q = blk_init_queue(ub_bd_rq_fn, &sc->lock)) == NULL)
@@ -2028,28 +2165,17 @@ static int ub_probe(struct usb_interface *intf,
 
 	disk->queue = q;
 
-        // blk_queue_bounce_limit(q, hba[i]->pdev->dma_mask);
+	blk_queue_bounce_limit(q, BLK_BOUNCE_HIGH);
 	blk_queue_max_hw_segments(q, UB_MAX_REQ_SG);
 	blk_queue_max_phys_segments(q, UB_MAX_REQ_SG);
-	// blk_queue_segment_boundary(q, CARM_SG_BOUNDARY);
+	blk_queue_segment_boundary(q, 0xffffffff);	/* Dubious. */
 	blk_queue_max_sectors(q, UB_MAX_SECTORS);
-	blk_queue_hardsect_size(q, sc->capacity.bsize);
-
-	/*
-	 * This is a serious infraction, caused by a deficiency in the
-	 * USB sg interface (usb_sg_wait()). We plan to remove this once
-	 * we get mileage on the driver and can justify a change to USB API.
-	 * See blk_queue_bounce_limit() to understand this part.
-	 *
-	 * XXX And I still need to be aware of the DMA mask in the HC.
-	 */
-	q->bounce_pfn = blk_max_low_pfn;
-	q->bounce_gfp = GFP_NOIO;
+	blk_queue_hardsect_size(q, lun->capacity.bsize);
 
-	q->queuedata = sc;
+	q->queuedata = lun;
 
-	set_capacity(disk, sc->capacity.nsec);
-	if (sc->removable)
+	set_capacity(disk, lun->capacity.nsec);
+	if (lun->removable)
 		disk->flags |= GENHD_FL_REMOVABLE;
 
 	add_disk(disk);
@@ -2059,22 +2185,20 @@ static int ub_probe(struct usb_interface *intf,
 err_blkqinit:
 	put_disk(disk);
 err_diskalloc:
-	device_remove_file(&sc->intf->dev, &dev_attr_diag);
-err_diag:
-	usb_set_intfdata(intf, NULL);
-	// usb_put_intf(sc->intf);
-	usb_put_dev(sc->dev);
-	ub_id_put(sc->id);
+	list_del(&lun->link);
+	ub_id_put(lun->id);
 err_id:
-	kfree(sc);
-err_core:
+	kfree(lun);
+err_alloc:
 	return rc;
 }
 
 static void ub_disconnect(struct usb_interface *intf)
 {
 	struct ub_dev *sc = usb_get_intfdata(intf);
-	struct gendisk *disk = sc->disk;
+	struct list_head *p;
+	struct ub_lun *lun;
+	struct gendisk *disk;
 	unsigned long flags;
 
 	/*
@@ -2124,14 +2248,18 @@ static void ub_disconnect(struct usb_interface *intf)
 	/*
 	 * Unregister the upper layer.
 	 */
-	if (disk->flags & GENHD_FL_UP)
-		del_gendisk(disk);
-	/*
-	 * I wish I could do:
-	 *    set_bit(QUEUE_FLAG_DEAD, &q->queue_flags);
-	 * As it is, we rely on our internal poisoning and let
-	 * the upper levels to spin furiously failing all the I/O.
-	 */
+	list_for_each (p, &sc->luns) {
+		lun = list_entry(p, struct ub_lun, link);
+		disk = lun->disk;
+		if (disk->flags & GENHD_FL_UP)
+			del_gendisk(disk);
+		/*
+		 * I wish I could do:
+		 *    set_bit(QUEUE_FLAG_DEAD, &q->queue_flags);
+		 * As it is, we rely on our internal poisoning and let
+		 * the upper levels to spin furiously failing all the I/O.
+		 */
+	}
 
 	/*
 	 * Taking a lock on a structure which is about to be freed
@@ -2182,8 +2310,8 @@ static int __init ub_init(void)
 {
 	int rc;
 
-	/* P3 */ printk("ub: sizeof ub_scsi_cmd %zu ub_dev %zu\n",
-			sizeof(struct ub_scsi_cmd), sizeof(struct ub_dev));
+	/* P3 */ printk("ub: sizeof ub_scsi_cmd %zu ub_dev %zu ub_lun %zu\n",
+			sizeof(struct ub_scsi_cmd), sizeof(struct ub_dev), sizeof(struct ub_lun));
 
 	if ((rc = register_blkdev(UB_MAJOR, DRV_NAME)) != 0)
 		goto err_regblkdev;