17 files changed, 819 insertions, 317 deletions

diff --git a/block/genhd.c b/block/genhd.c
index 16ccc0d2d5d..74331738163 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -18,6 +18,7 @@
 #include <linux/buffer_head.h>
 #include <linux/mutex.h>
 #include <linux/idr.h>
+#include <linux/log2.h>
 
 #include "blk.h"
 
@@ -35,6 +36,10 @@ static DEFINE_IDR(ext_devt_idr);
 
 static struct device_type disk_type;
 
+static void disk_add_events(struct gendisk *disk);
+static void disk_del_events(struct gendisk *disk);
+static void disk_release_events(struct gendisk *disk);
+
 /**
  * disk_get_part - get partition
  * @disk: disk to look partition from
@@ -502,6 +507,64 @@ static int exact_lock(dev_t devt, void *data)
 	return 0;
 }
 
+void register_disk(struct gendisk *disk)
+{
+	struct device *ddev = disk_to_dev(disk);
+	struct block_device *bdev;
+	struct disk_part_iter piter;
+	struct hd_struct *part;
+	int err;
+
+	ddev->parent = disk->driverfs_dev;
+
+	dev_set_name(ddev, disk->disk_name);
+
+	/* delay uevents, until we scanned partition table */
+	dev_set_uevent_suppress(ddev, 1);
+
+	if (device_add(ddev))
+		return;
+	if (!sysfs_deprecated) {
+		err = sysfs_create_link(block_depr, &ddev->kobj,
+					kobject_name(&ddev->kobj));
+		if (err) {
+			device_del(ddev);
+			return;
+		}
+	}
+	disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
+	disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
+
+	/* No minors to use for partitions */
+	if (!disk_partitionable(disk))
+		goto exit;
+
+	/* No such device (e.g., media were just removed) */
+	if (!get_capacity(disk))
+		goto exit;
+
+	bdev = bdget_disk(disk, 0);
+	if (!bdev)
+		goto exit;
+
+	bdev->bd_invalidated = 1;
+	err = blkdev_get(bdev, FMODE_READ, NULL);
+	if (err < 0)
+		goto exit;
+	blkdev_put(bdev, FMODE_READ);
+
+exit:
+	/* announce disk after possible partitions are created */
+	dev_set_uevent_suppress(ddev, 0);
+	kobject_uevent(&ddev->kobj, KOBJ_ADD);
+
+	/* announce possible partitions */
+	disk_part_iter_init(&piter, disk, 0);
+	while ((part = disk_part_iter_next(&piter)))
+		kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
+	disk_part_iter_exit(&piter);
+}
+
 /**
  * add_disk - add partitioning information to kernel list
  * @disk: per-device partitioning information
@@ -551,18 +614,48 @@ void add_disk(struct gendisk *disk)
 	retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
 				   "bdi");
 	WARN_ON(retval);
-}
 
+	disk_add_events(disk);
+}
 EXPORT_SYMBOL(add_disk);
-EXPORT_SYMBOL(del_gendisk);	/* in partitions/check.c */
 
-void unlink_gendisk(struct gendisk *disk)
+void del_gendisk(struct gendisk *disk)
 {
+	struct disk_part_iter piter;
+	struct hd_struct *part;
+
+	disk_del_events(disk);
+
+	/* invalidate stuff */
+	disk_part_iter_init(&piter, disk,
+			     DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
+	while ((part = disk_part_iter_next(&piter))) {
+		invalidate_partition(disk, part->partno);
+		delete_partition(disk, part->partno);
+	}
+	disk_part_iter_exit(&piter);
+
+	invalidate_partition(disk, 0);
+	blk_free_devt(disk_to_dev(disk)->devt);
+	set_capacity(disk, 0);
+	disk->flags &= ~GENHD_FL_UP;
+
 	sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
 	bdi_unregister(&disk->queue->backing_dev_info);
 	blk_unregister_queue(disk);
 	blk_unregister_region(disk_devt(disk), disk->minors);
+
+	part_stat_set_all(&disk->part0, 0);
+	disk->part0.stamp = 0;
+
+	kobject_put(disk->part0.holder_dir);
+	kobject_put(disk->slave_dir);
+	disk->driverfs_dev = NULL;
+	if (!sysfs_deprecated)
+		sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
+	device_del(disk_to_dev(disk));
 }
+EXPORT_SYMBOL(del_gendisk);
 
 /**
  * get_gendisk - get partitioning information for a given device
@@ -1005,6 +1098,7 @@ static void disk_release(struct device *dev)
 {
 	struct gendisk *disk = dev_to_disk(dev);
 
+	disk_release_events(disk);
 	kfree(disk->random);
 	disk_replace_part_tbl(disk, NULL);
 	free_part_stats(&disk->part0);
@@ -1110,29 +1204,6 @@ static int __init proc_genhd_init(void)
 module_init(proc_genhd_init);
 #endif /* CONFIG_PROC_FS */
 
-static void media_change_notify_thread(struct work_struct *work)
-{
-	struct gendisk *gd = container_of(work, struct gendisk, async_notify);
-	char event[] = "MEDIA_CHANGE=1";
-	char *envp[] = { event, NULL };
-
-	/*
-	 * set enviroment vars to indicate which event this is for
-	 * so that user space will know to go check the media status.
-	 */
-	kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
-	put_device(gd->driverfs_dev);
-}
-
-#if 0
-void genhd_media_change_notify(struct gendisk *disk)
-{
-	get_device(disk->driverfs_dev);
-	schedule_work(&disk->async_notify);
-}
-EXPORT_SYMBOL_GPL(genhd_media_change_notify);
-#endif  /*  0  */
-
 dev_t blk_lookup_devt(const char *name, int partno)
 {
 	dev_t devt = MKDEV(0, 0);
@@ -1198,8 +1269,6 @@ struct gendisk *alloc_disk_node(int minors, int node_id)
 		disk_to_dev(disk)->class = &block_class;
 		disk_to_dev(disk)->type = &disk_type;
 		device_initialize(disk_to_dev(disk));
-		INIT_WORK(&disk->async_notify,
-			media_change_notify_thread);
 	}
 	return disk;
 }
@@ -1291,3 +1360,422 @@ int invalidate_partition(struct gendisk *disk, int partno)
 }
 
 EXPORT_SYMBOL(invalidate_partition);
+
+/*
+ * Disk events - monitor disk events like media change and eject request.
+ */
+struct disk_events {
+	struct list_head	node;		/* all disk_event's */
+	struct gendisk		*disk;		/* the associated disk */
+	spinlock_t		lock;
+
+	int			block;		/* event blocking depth */
+	unsigned int		pending;	/* events already sent out */
+	unsigned int		clearing;	/* events being cleared */
+
+	long			poll_msecs;	/* interval, -1 for default */
+	struct delayed_work	dwork;
+};
+
+static const char *disk_events_strs[] = {
+	[ilog2(DISK_EVENT_MEDIA_CHANGE)]	= "media_change",
+	[ilog2(DISK_EVENT_EJECT_REQUEST)]	= "eject_request",
+};
+
+static char *disk_uevents[] = {
+	[ilog2(DISK_EVENT_MEDIA_CHANGE)]	= "DISK_MEDIA_CHANGE=1",
+	[ilog2(DISK_EVENT_EJECT_REQUEST)]	= "DISK_EJECT_REQUEST=1",
+};
+
+/* list of all disk_events */
+static DEFINE_MUTEX(disk_events_mutex);
+static LIST_HEAD(disk_events);
+
+/* disable in-kernel polling by default */
+static unsigned long disk_events_dfl_poll_msecs	= 0;
+
+static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
+{
+	struct disk_events *ev = disk->ev;
+	long intv_msecs = 0;
+
+	/*
+	 * If device-specific poll interval is set, always use it.  If
+	 * the default is being used, poll iff there are events which
+	 * can't be monitored asynchronously.
+	 */
+	if (ev->poll_msecs >= 0)
+		intv_msecs = ev->poll_msecs;
+	else if (disk->events & ~disk->async_events)
+		intv_msecs = disk_events_dfl_poll_msecs;
+
+	return msecs_to_jiffies(intv_msecs);
+}
+
+static void __disk_block_events(struct gendisk *disk, bool sync)
+{
+	struct disk_events *ev = disk->ev;
+	unsigned long flags;
+	bool cancel;
+
+	spin_lock_irqsave(&ev->lock, flags);
+	cancel = !ev->block++;
+	spin_unlock_irqrestore(&ev->lock, flags);
+
+	if (cancel) {
+		if (sync)
+			cancel_delayed_work_sync(&disk->ev->dwork);
+		else
+			cancel_delayed_work(&disk->ev->dwork);
+	}
+}
+
+static void __disk_unblock_events(struct gendisk *disk, bool check_now)
+{
+	struct disk_events *ev = disk->ev;
+	unsigned long intv;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ev->lock, flags);
+
+	if (WARN_ON_ONCE(ev->block <= 0))
+		goto out_unlock;
+
+	if (--ev->block)
+		goto out_unlock;
+
+	/*
+	 * Not exactly a latency critical operation, set poll timer
+	 * slack to 25% and kick event check.
+	 */
+	intv = disk_events_poll_jiffies(disk);
+	set_timer_slack(&ev->dwork.timer, intv / 4);
+	if (check_now)
+		queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
+	else if (intv)
+		queue_delayed_work(system_nrt_wq, &ev->dwork, intv);
+out_unlock:
+	spin_unlock_irqrestore(&ev->lock, flags);
+}
+
+/**
+ * disk_block_events - block and flush disk event checking
+ * @disk: disk to block events for
+ *
+ * On return from this function, it is guaranteed that event checking
+ * isn't in progress and won't happen until unblocked by
+ * disk_unblock_events().  Events blocking is counted and the actual
+ * unblocking happens after the matching number of unblocks are done.
+ *
+ * Note that this intentionally does not block event checking from
+ * disk_clear_events().
+ *
+ * CONTEXT:
+ * Might sleep.
+ */
+void disk_block_events(struct gendisk *disk)
+{
+	if (disk->ev)
+		__disk_block_events(disk, true);
+}
+
+/**
+ * disk_unblock_events - unblock disk event checking
+ * @disk: disk to unblock events for
+ *
+ * Undo disk_block_events().  When the block count reaches zero, it
+ * starts events polling if configured.
+ *
+ * CONTEXT:
+ * Don't care.  Safe to call from irq context.
+ */
+void disk_unblock_events(struct gendisk *disk)
+{
+	if (disk->ev)
+		__disk_unblock_events(disk, true);
+}
+
+/**
+ * disk_check_events - schedule immediate event checking
+ * @disk: disk to check events for
+ *
+ * Schedule immediate event checking on @disk if not blocked.
+ *
+ * CONTEXT:
+ * Don't care.  Safe to call from irq context.
+ */
+void disk_check_events(struct gendisk *disk)
+{
+	if (disk->ev) {
+		__disk_block_events(disk, false);
+		__disk_unblock_events(disk, true);
+	}
+}
+EXPORT_SYMBOL_GPL(disk_check_events);
+
+/**
+ * disk_clear_events - synchronously check, clear and return pending events
+ * @disk: disk to fetch and clear events from
+ * @mask: mask of events to be fetched and clearted
+ *
+ * Disk events are synchronously checked and pending events in @mask
+ * are cleared and returned.  This ignores the block count.
+ *
+ * CONTEXT:
+ * Might sleep.
+ */
+unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
+{
+	const struct block_device_operations *bdops = disk->fops;
+	struct disk_events *ev = disk->ev;
+	unsigned int pending;
+
+	if (!ev) {
+		/* for drivers still using the old ->media_changed method */
+		if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
+		    bdops->media_changed && bdops->media_changed(disk))
+			return DISK_EVENT_MEDIA_CHANGE;
+		return 0;
+	}
+
+	/* tell the workfn about the events being cleared */
+	spin_lock_irq(&ev->lock);
+	ev->clearing |= mask;
+	spin_unlock_irq(&ev->lock);
+
+	/* uncondtionally schedule event check and wait for it to finish */
+	__disk_block_events(disk, true);
+	queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
+	flush_delayed_work(&ev->dwork);
+	__disk_unblock_events(disk, false);
+
+	/* then, fetch and clear pending events */
+	spin_lock_irq(&ev->lock);
+	WARN_ON_ONCE(ev->clearing & mask);	/* cleared by workfn */
+	pending = ev->pending & mask;
+	ev->pending &= ~mask;
+	spin_unlock_irq(&ev->lock);
+
+	return pending;
+}
+
+static void disk_events_workfn(struct work_struct *work)
+{
+	struct delayed_work *dwork = to_delayed_work(work);
+	struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
+	struct gendisk *disk = ev->disk;
+	char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
+	unsigned int clearing = ev->clearing;
+	unsigned int events;
+	unsigned long intv;
+	int nr_events = 0, i;
+
+	/* check events */
+	events = disk->fops->check_events(disk, clearing);
+
+	/* accumulate pending events and schedule next poll if necessary */
+	spin_lock_irq(&ev->lock);
+
+	events &= ~ev->pending;
+	ev->pending |= events;
+	ev->clearing &= ~clearing;
+
+	intv = disk_events_poll_jiffies(disk);
+	if (!ev->block && intv)
+		queue_delayed_work(system_nrt_wq, &ev->dwork, intv);
+
+	spin_unlock_irq(&ev->lock);
+
+	/* tell userland about new events */
+	for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
+		if (events & (1 << i))
+			envp[nr_events++] = disk_uevents[i];
+
+	if (nr_events)
+		kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
+}
+
+/*
+ * A disk events enabled device has the following sysfs nodes under
+ * its /sys/block/X/ directory.
+ *
+ * events		: list of all supported events
+ * events_async		: list of events which can be detected w/o polling
+ * events_poll_msecs	: polling interval, 0: disable, -1: system default
+ */
+static ssize_t __disk_events_show(unsigned int events, char *buf)
+{
+	const char *delim = "";
+	ssize_t pos = 0;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
+		if (events & (1 << i)) {
+			pos += sprintf(buf + pos, "%s%s",
+				       delim, disk_events_strs[i]);
+			delim = " ";
+		}
+	if (pos)
+		pos += sprintf(buf + pos, "\n");
+	return pos;
+}
+
+static ssize_t disk_events_show(struct device *dev,
+				struct device_attribute *attr, char *buf)
+{
+	struct gendisk *disk = dev_to_disk(dev);
+
+	return __disk_events_show(disk->events, buf);
+}
+
+static ssize_t disk_events_async_show(struct device *dev,
+				      struct device_attribute *attr, char *buf)
+{
+	struct gendisk *disk = dev_to_disk(dev);
+
+	return __disk_events_show(disk->async_events, buf);
+}
+
+static ssize_t disk_events_poll_msecs_show(struct device *dev,
+					   struct device_attribute *attr,
+					   char *buf)
+{
+	struct gendisk *disk = dev_to_disk(dev);
+
+	return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
+}
+
+static ssize_t disk_events_poll_msecs_store(struct device *dev,
+					    struct device_attribute *attr,
+					    const char *buf, size_t count)
+{
+	struct gendisk *disk = dev_to_disk(dev);
+	long intv;
+
+	if (!count || !sscanf(buf, "%ld", &intv))
+		return -EINVAL;
+
+	if (intv < 0 && intv != -1)
+		return -EINVAL;
+
+	__disk_block_events(disk, true);
+	disk->ev->poll_msecs = intv;
+	__disk_unblock_events(disk, true);
+
+	return count;
+}
+
+static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
+static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
+static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
+			 disk_events_poll_msecs_show,
+			 disk_events_poll_msecs_store);
+
+static const struct attribute *disk_events_attrs[] = {
+	&dev_attr_events.attr,
+	&dev_attr_events_async.attr,
+	&dev_attr_events_poll_msecs.attr,
+	NULL,
+};
+
+/*
+ * The default polling interval can be specified by the kernel
+ * parameter block.events_dfl_poll_msecs which defaults to 0
+ * (disable).  This can also be modified runtime by writing to
+ * /sys/module/block/events_dfl_poll_msecs.
+ */
+static int disk_events_set_dfl_poll_msecs(const char *val,
+					  const struct kernel_param *kp)
+{
+	struct disk_events *ev;
+	int ret;
+
+	ret = param_set_ulong(val, kp);
+	if (ret < 0)
+		return ret;
+
+	mutex_lock(&disk_events_mutex);
+
+	list_for_each_entry(ev, &disk_events, node)
+		disk_check_events(ev->disk);
+
+	mutex_unlock(&disk_events_mutex);
+
+	return 0;
+}
+
+static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
+	.set	= disk_events_set_dfl_poll_msecs,
+	.get	= param_get_ulong,
+};
+
+#undef MODULE_PARAM_PREFIX
+#define MODULE_PARAM_PREFIX	"block."
+
+module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
+		&disk_events_dfl_poll_msecs, 0644);
+
+/*
+ * disk_{add|del|release}_events - initialize and destroy disk_events.
+ */
+static void disk_add_events(struct gendisk *disk)
+{
+	struct disk_events *ev;
+
+	if (!disk->fops->check_events || !(disk->events | disk->async_events))
+		return;
+
+	ev = kzalloc(sizeof(*ev), GFP_KERNEL);
+	if (!ev) {
+		pr_warn("%s: failed to initialize events\n", disk->disk_name);
+		return;
+	}
+
+	if (sysfs_create_files(&disk_to_dev(disk)->kobj,
+			       disk_events_attrs) < 0) {
+		pr_warn("%s: failed to create sysfs files for events\n",
+			disk->disk_name);
+		kfree(ev);
+		return;
+	}
+
+	disk->ev = ev;
+
+	INIT_LIST_HEAD(&ev->node);
+	ev->disk = disk;
+	spin_lock_init(&ev->lock);
+	ev->block = 1;
+	ev->poll_msecs = -1;
+	INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
+
+	mutex_lock(&disk_events_mutex);
+	list_add_tail(&ev->node, &disk_events);
+	mutex_unlock(&disk_events_mutex);
+
+	/*
+	 * Block count is initialized to 1 and the following initial
+	 * unblock kicks it into action.
+	 */
+	__disk_unblock_events(disk, true);
+}
+
+static void disk_del_events(struct gendisk *disk)
+{
+	if (!disk->ev)
+		return;
+
+	__disk_block_events(disk, true);
+
+	mutex_lock(&disk_events_mutex);
+	list_del_init(&disk->ev->node);
+	mutex_unlock(&disk_events_mutex);
+
+	sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
+}
+
+static void disk_release_events(struct gendisk *disk)
+{
+	/* the block count should be 1 from disk_del_events() */
+	WARN_ON_ONCE(disk->ev && disk->ev->block != 1);
+	kfree(disk->ev);
+}
diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c
index 255035cfc88..4f9e22f2913 100644
--- a/drivers/block/xen-blkfront.c
+++ b/drivers/block/xen-blkfront.c
@@ -65,7 +65,7 @@ enum blkif_state {
 
 struct blk_shadow {
 	struct blkif_request req;
-	unsigned long request;
+	struct request *request;
 	unsigned long frame[BLKIF_MAX_SEGMENTS_PER_REQUEST];
 };
 
@@ -136,7 +136,7 @@ static void add_id_to_freelist(struct blkfront_info *info,
 			       unsigned long id)
 {
 	info->shadow[id].req.id  = info->shadow_free;
-	info->shadow[id].request = 0;
+	info->shadow[id].request = NULL;
 	info->shadow_free = id;
 }
 
@@ -245,14 +245,11 @@ static int blkif_ioctl(struct block_device *bdev, fmode_t mode,
 }
 
 /*
- * blkif_queue_request
+ * Generate a Xen blkfront IO request from a blk layer request.  Reads
+ * and writes are handled as expected.  Since we lack a loose flush
+ * request, we map flushes into a full ordered barrier.
  *
- * request block io
- *
- * id: for guest use only.
- * operation: BLKIF_OP_{READ,WRITE,PROBE}
- * buffer: buffer to read/write into. this should be a
- *   virtual address in the guest os.
+ * @req: a request struct
  */
 static int blkif_queue_request(struct request *req)
 {
@@ -281,7 +278,7 @@ static int blkif_queue_request(struct request *req)
 	/* Fill out a communications ring structure. */
 	ring_req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt);
 	id = get_id_from_freelist(info);
-	info->shadow[id].request = (unsigned long)req;
+	info->shadow[id].request = req;
 
 	ring_req->id = id;
 	ring_req->sector_number = (blkif_sector_t)blk_rq_pos(req);
@@ -290,6 +287,18 @@ static int blkif_queue_request(struct request *req)
 	ring_req->operation = rq_data_dir(req) ?
 		BLKIF_OP_WRITE : BLKIF_OP_READ;
 
+	if (req->cmd_flags & (REQ_FLUSH | REQ_FUA)) {
+		/*
+		 * Ideally we could just do an unordered
+		 * flush-to-disk, but all we have is a full write
+		 * barrier at the moment.  However, a barrier write is
+		 * a superset of FUA, so we can implement it the same
+		 * way.  (It's also a FLUSH+FUA, since it is
+		 * guaranteed ordered WRT previous writes.)
+		 */
+		ring_req->operation = BLKIF_OP_WRITE_BARRIER;
+	}
+
 	ring_req->nr_segments = blk_rq_map_sg(req->q, req, info->sg);
 	BUG_ON(ring_req->nr_segments > BLKIF_MAX_SEGMENTS_PER_REQUEST);
 
@@ -634,7 +643,7 @@ static irqreturn_t blkif_interrupt(int irq, void *dev_id)
 
 		bret = RING_GET_RESPONSE(&info->ring, i);
 		id   = bret->id;
-		req  = (struct request *)info->shadow[id].request;
+		req  = info->shadow[id].request;
 
 		blkif_completion(&info->shadow[id]);
 
@@ -647,6 +656,16 @@ static irqreturn_t blkif_interrupt(int irq, void *dev_id)
 				printk(KERN_WARNING "blkfront: %s: write barrier op failed\n",
 				       info->gd->disk_name);
 				error = -EOPNOTSUPP;
+			}
+			if (unlikely(bret->status == BLKIF_RSP_ERROR &&
+				     info->shadow[id].req.nr_segments == 0)) {
+				printk(KERN_WARNING "blkfront: %s: empty write barrier op failed\n",
+				       info->gd->disk_name);
+				error = -EOPNOTSUPP;
+			}
+			if (unlikely(error)) {
+				if (error == -EOPNOTSUPP)
+					error = 0;
 				info->feature_flush = 0;
 				xlvbd_flush(info);
 			}
@@ -899,7 +918,7 @@ static int blkif_recover(struct blkfront_info *info)
 	/* Stage 3: Find pending requests and requeue them. */
 	for (i = 0; i < BLK_RING_SIZE; i++) {
 		/* Not in use? */
-		if (copy[i].request == 0)
+		if (!copy[i].request)
 			continue;
 
 		/* Grab a request slot and copy shadow state into it. */
@@ -916,9 +935,7 @@ static int blkif_recover(struct blkfront_info *info)
 				req->seg[j].gref,
 				info->xbdev->otherend_id,
 				pfn_to_mfn(info->shadow[req->id].frame[j]),
-				rq_data_dir(
-					(struct request *)
-					info->shadow[req->id].request));
+				rq_data_dir(info->shadow[req->id].request));
 		info->shadow[req->id].req = *req;
 
 		info->ring.req_prod_pvt++;
@@ -1067,14 +1084,8 @@ static void blkfront_connect(struct blkfront_info *info)
 	 */
 	info->feature_flush = 0;
 
-	/*
-	 * The driver doesn't properly handled empty flushes, so
-	 * lets disable barrier support for now.
-	 */
-#if 0
 	if (!err && barrier)
-		info->feature_flush = REQ_FLUSH;
-#endif
+		info->feature_flush = REQ_FLUSH | REQ_FUA;
 
 	err = xlvbd_alloc_gendisk(sectors, info, binfo, sector_size);
 	if (err) {
diff --git a/drivers/cdrom/cdrom.c b/drivers/cdrom/cdrom.c
index af13c62dc47..14033a36bcd 100644
--- a/drivers/cdrom/cdrom.c
+++ b/drivers/cdrom/cdrom.c
@@ -1348,7 +1348,10 @@ static int cdrom_select_disc(struct cdrom_device_info *cdi, int slot)
 	if (!CDROM_CAN(CDC_SELECT_DISC))
 		return -EDRIVE_CANT_DO_THIS;
 
-	(void) cdi->ops->media_changed(cdi, slot);
+	if (cdi->ops->check_events)
+		cdi->ops->check_events(cdi, 0, slot);
+	else
+		cdi->ops->media_changed(cdi, slot);
 
 	if (slot == CDSL_NONE) {
 		/* set media changed bits, on both queues */
@@ -1392,6 +1395,42 @@ static int cdrom_select_disc(struct cdrom_device_info *cdi, int slot)
 	return slot;
 }
 
+/*
+ * As cdrom implements an extra ioctl consumer for media changed
+ * event, it needs to buffer ->check_events() output, such that event
+ * is not lost for both the usual VFS and ioctl paths.
+ * cdi->{vfs|ioctl}_events are used to buffer pending events for each
+ * path.
+ *
+ * XXX: Locking is non-existent.  cdi->ops->check_events() can be
+ * called in parallel and buffering fields are accessed without any
+ * exclusion.  The original media_changed code had the same problem.
+ * It might be better to simply deprecate CDROM_MEDIA_CHANGED ioctl
+ * and remove this cruft altogether.  It doesn't have much usefulness
+ * at this point.
+ */
+static void cdrom_update_events(struct cdrom_device_info *cdi,
+				unsigned int clearing)
+{
+	unsigned int events;
+
+	events = cdi->ops->check_events(cdi, clearing, CDSL_CURRENT);
+	cdi->vfs_events |= events;
+	cdi->ioctl_events |= events;
+}
+
+unsigned int cdrom_check_events(struct cdrom_device_info *cdi,
+				unsigned int clearing)
+{
+	unsigned int events;
+
+	cdrom_update_events(cdi, clearing);
+	events = cdi->vfs_events;
+	cdi->vfs_events = 0;
+	return events;
+}
+EXPORT_SYMBOL(cdrom_check_events);
+
 /* We want to make media_changed accessible to the user through an
  * ioctl. The main problem now is that we must double-buffer the
  * low-level implementation, to assure that the VFS and the user both
@@ -1403,15 +1442,26 @@ int media_changed(struct cdrom_device_info *cdi, int queue)
 {
 	unsigned int mask = (1 << (queue & 1));
 	int ret = !!(cdi->mc_flags & mask);
+	bool changed;
 
 	if (!CDROM_CAN(CDC_MEDIA_CHANGED))
-	    return ret;
+		return ret;
+
 	/* changed since last call? */
-	if (cdi->ops->media_changed(cdi, CDSL_CURRENT)) {
+	if (cdi->ops->check_events) {
+		BUG_ON(!queue);	/* shouldn't be called from VFS path */
+		cdrom_update_events(cdi, DISK_EVENT_MEDIA_CHANGE);
+		changed = cdi->ioctl_events & DISK_EVENT_MEDIA_CHANGE;
+		cdi->ioctl_events = 0;
+	} else
+		changed = cdi->ops->media_changed(cdi, CDSL_CURRENT);
+
+	if (changed) {
 		cdi->mc_flags = 0x3;    /* set bit on both queues */
 		ret |= 1;
 		cdi->media_written = 0;
 	}
+
 	cdi->mc_flags &= ~mask;         /* clear bit */
 	return ret;
 }
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index eafeeda6e19..13bf89145b1 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -1984,8 +1984,7 @@ EXPORT_SYMBOL(scsi_mode_sense);
  *		in.
  *
  *	Returns zero if unsuccessful or an error if TUR failed.  For
- *	removable media, a return of NOT_READY or UNIT_ATTENTION is
- *	translated to success, with the ->changed flag updated.
+ *	removable media, UNIT_ATTENTION sets ->changed flag.
  **/
 int
 scsi_test_unit_ready(struct scsi_device *sdev, int timeout, int retries,
@@ -2012,16 +2011,6 @@ scsi_test_unit_ready(struct scsi_device *sdev, int timeout, int retries,
 	} while (scsi_sense_valid(sshdr) &&
 		 sshdr->sense_key == UNIT_ATTENTION && --retries);
 
-	if (!sshdr)
-		/* could not allocate sense buffer, so can't process it */
-		return result;
-
-	if (sdev->removable && scsi_sense_valid(sshdr) &&
-	    (sshdr->sense_key == UNIT_ATTENTION ||
-	     sshdr->sense_key == NOT_READY)) {
-		sdev->changed = 1;
-		result = 0;
-	}
 	if (!sshdr_external)
 		kfree(sshdr);
 	return result;
diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
index b9ab3a590e4..b179b0e39a3 100644
--- a/drivers/scsi/sd.c
+++ b/drivers/scsi/sd.c
@@ -991,30 +991,50 @@ out:
 
 static void set_media_not_present(struct scsi_disk *sdkp)
 {
+	if (sdkp->media_present)
+		sdkp->device->changed = 1;
 	sdkp->media_present = 0;
 	sdkp->capacity = 0;
-	sdkp->device->changed = 1;
+}
+
+static int media_not_present(struct scsi_disk *sdkp,
+			     struct scsi_sense_hdr *sshdr)
+{
+	if (!scsi_sense_valid(sshdr))
+		return 0;
+
+	/* not invoked for commands that could return deferred errors */
+	switch (sshdr->sense_key) {
+	case UNIT_ATTENTION:
+		sdkp->device->changed = 1;
+		/* fall through */
+	case NOT_READY:
+		/* medium not present */
+		if (sshdr->asc == 0x3A) {
+			set_media_not_present(sdkp);
+			return 1;
+		}
+	}
+	return 0;
 }
 
 /**
- *	sd_media_changed - check if our medium changed
- *	@disk: kernel device descriptor 
+ *	sd_check_events - check media events
+ *	@disk: kernel device descriptor
+ *	@clearing: disk events currently being cleared
  *
- *	Returns 0 if not applicable or no change; 1 if change
+ *	Returns mask of DISK_EVENT_*.
  *
  *	Note: this function is invoked from the block subsystem.
  **/
-static int sd_media_changed(struct gendisk *disk)
+static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
 {
 	struct scsi_disk *sdkp = scsi_disk(disk);
 	struct scsi_device *sdp = sdkp->device;
 	struct scsi_sense_hdr *sshdr = NULL;
 	int retval;
 
-	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_media_changed\n"));
-
-	if (!sdp->removable)
-		return 0;
+	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
 
 	/*
 	 * If the device is offline, don't send any commands - just pretend as
@@ -1024,7 +1044,6 @@ static int sd_media_changed(struct gendisk *disk)
 	 */
 	if (!scsi_device_online(sdp)) {
 		set_media_not_present(sdkp);
-		retval = 1;
 		goto out;
 	}
 
@@ -1045,34 +1064,30 @@ static int sd_media_changed(struct gendisk *disk)
 					      sshdr);
 	}
 
-	/*
-	 * Unable to test, unit probably not ready.   This usually
-	 * means there is no disc in the drive.  Mark as changed,
-	 * and we will figure it out later once the drive is
-	 * available again.
-	 */
-	if (retval || (scsi_sense_valid(sshdr) &&
-		       /* 0x3a is medium not present */
-		       sshdr->asc == 0x3a)) {
+	/* failed to execute TUR, assume media not present */
+	if (host_byte(retval)) {
 		set_media_not_present(sdkp);
-		retval = 1;
 		goto out;
 	}
 
+	if (media_not_present(sdkp, sshdr))
+		goto out;
+
 	/*
 	 * For removable scsi disk we have to recognise the presence
-	 * of a disk in the drive. This is kept in the struct scsi_disk
-	 * struct and tested at open !  Daniel Roche (dan@lectra.fr)
+	 * of a disk in the drive.
 	 */
+	if (!sdkp->media_present)
+		sdp->changed = 1;
 	sdkp->media_present = 1;
-
-	retval = sdp->changed;
-	sdp->changed = 0;
 out:
-	if (retval != sdkp->previous_state)
+	/* for backward compatibility */
+	if (sdp->changed)
 		sdev_evt_send_simple(sdp, SDEV_EVT_MEDIA_CHANGE, GFP_KERNEL);
-	sdkp->previous_state = retval;
 	kfree(sshdr);
+
+	retval = sdp->changed ? DISK_EVENT_MEDIA_CHANGE : 0;
+	sdp->changed = 0;
 	return retval;
 }
 
@@ -1165,7 +1180,7 @@ static const struct block_device_operations sd_fops = {
 #ifdef CONFIG_COMPAT
 	.compat_ioctl		= sd_compat_ioctl,
 #endif
-	.media_changed		= sd_media_changed,
+	.check_events		= sd_check_events,
 	.revalidate_disk	= sd_revalidate_disk,
 	.unlock_native_capacity	= sd_unlock_native_capacity,
 };
@@ -1301,23 +1316,6 @@ static int sd_done(struct scsi_cmnd *SCpnt)
 	return good_bytes;
 }
 
-static int media_not_present(struct scsi_disk *sdkp,
-			     struct scsi_sense_hdr *sshdr)
-{
-
-	if (!scsi_sense_valid(sshdr))
-		return 0;
-	/* not invoked for commands that could return deferred errors */
-	if (sshdr->sense_key != NOT_READY &&
-	    sshdr->sense_key != UNIT_ATTENTION)
-		return 0;
-	if (sshdr->asc != 0x3A) /* medium not present */
-		return 0;
-
-	set_media_not_present(sdkp);
-	return 1;
-}
-
 /*
  * spinup disk - called only in sd_revalidate_disk()
  */
@@ -1492,7 +1490,7 @@ static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
 	 */
 	if (sdp->removable &&
 	    sense_valid && sshdr->sense_key == NOT_READY)
-		sdp->changed = 1;
+		set_media_not_present(sdkp);
 
 	/*
 	 * We used to set media_present to 0 here to indicate no media
@@ -2347,8 +2345,10 @@ static void sd_probe_async(void *data, async_cookie_t coo