Merge git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm

* git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm:
  dm snapshot: extend exception store functions
  dm snapshot: split out exception store implementations
  dm snapshot: rename struct exception_store
  dm snapshot: separate out exception store interface
  dm mpath: move trigger_event to system workqueue
  dm: add name and uuid to sysfs
  dm table: rework reference counting
  dm: support barriers on simple devices
  dm request: extend target interface
  dm request: add caches
  dm ioctl: allow dm_copy_name_and_uuid to return only one field
  dm log: ensure log bitmap fits on log device
  dm log: move region_size validation
  dm log: avoid reinitialising io_req on every operation
  dm: consolidate target deregistration error handling
  dm raid1: fix error count
  dm log: fix dm_io_client leak on error paths
  dm snapshot: change yield to msleep
  dm table: drop reference at unbind

drivers/md/Makefile

@@ -3,9 +3,10 @@
 #
 
 dm-mod-objs	:= dm.o dm-table.o dm-target.o dm-linear.o dm-stripe.o \
-		   dm-ioctl.o dm-io.o dm-kcopyd.o
+		   dm-ioctl.o dm-io.o dm-kcopyd.o dm-sysfs.o
 dm-multipath-objs := dm-path-selector.o dm-mpath.o
-dm-snapshot-objs := dm-snap.o dm-exception-store.o
+dm-snapshot-objs := dm-snap.o dm-exception-store.o dm-snap-transient.o \
+		    dm-snap-persistent.o
 dm-mirror-objs	:= dm-raid1.o
 md-mod-objs     := md.o bitmap.o
 raid456-objs	:= raid5.o raid6algos.o raid6recov.o raid6tables.o \

drivers/md/dm-crypt.c

@@ -1322,11 +1322,7 @@ static int __init dm_crypt_init(void)
 
 static void __exit dm_crypt_exit(void)
 {
-	int r = dm_unregister_target(&crypt_target);
-
-	if (r < 0)
-		DMERR("unregister failed %d", r);
-
+	dm_unregister_target(&crypt_target);
 	kmem_cache_destroy(_crypt_io_pool);
 }
 

drivers/md/dm-delay.c

@@ -364,11 +364,7 @@ bad_queue:
 
 static void __exit dm_delay_exit(void)
 {
-	int r = dm_unregister_target(&delay_target);
-
-	if (r < 0)
-		DMERR("unregister failed %d", r);
-
+	dm_unregister_target(&delay_target);
 	kmem_cache_destroy(delayed_cache);
 	destroy_workqueue(kdelayd_wq);
 }

drivers/md/dm-exception-store.c

@@ -1,756 +1,45 @@
 /*
- * dm-exception-store.c
- *
  * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
- * Copyright (C) 2006 Red Hat GmbH
+ * Copyright (C) 2006-2008 Red Hat GmbH
  *
  * This file is released under the GPL.
  */
 
-#include "dm-snap.h"
+#include "dm-exception-store.h"
 
 #include <linux/mm.h>
 #include <linux/pagemap.h>
 #include <linux/vmalloc.h>
 #include <linux/slab.h>
-#include <linux/dm-io.h>
-#include <linux/dm-kcopyd.h>
-
-#define DM_MSG_PREFIX "snapshots"
-#define DM_CHUNK_SIZE_DEFAULT_SECTORS 32	/* 16KB */
-
-/*-----------------------------------------------------------------
- * Persistent snapshots, by persistent we mean that the snapshot
- * will survive a reboot.
- *---------------------------------------------------------------*/
-
-/*
- * We need to store a record of which parts of the origin have
- * been copied to the snapshot device.  The snapshot code
- * requires that we copy exception chunks to chunk aligned areas
- * of the COW store.  It makes sense therefore, to store the
- * metadata in chunk size blocks.
- *
- * There is no backward or forward compatibility implemented,
- * snapshots with different disk versions than the kernel will
- * not be usable.  It is expected that "lvcreate" will blank out
- * the start of a fresh COW device before calling the snapshot
- * constructor.
- *
- * The first chunk of the COW device just contains the header.
- * After this there is a chunk filled with exception metadata,
- * followed by as many exception chunks as can fit in the
- * metadata areas.
- *
- * All on disk structures are in little-endian format.  The end
- * of the exceptions info is indicated by an exception with a
- * new_chunk of 0, which is invalid since it would point to the
- * header chunk.
- */
-
-/*
- * Magic for persistent snapshots: "SnAp" - Feeble isn't it.
- */
-#define SNAP_MAGIC 0x70416e53
-
-/*
- * The on-disk version of the metadata.
- */
-#define SNAPSHOT_DISK_VERSION 1
-
-struct disk_header {
-	uint32_t magic;
-
-	/*
-	 * Is this snapshot valid.  There is no way of recovering
-	 * an invalid snapshot.
-	 */
-	uint32_t valid;
-
-	/*
-	 * Simple, incrementing version. no backward
-	 * compatibility.
-	 */
-	uint32_t version;
-
-	/* In sectors */
-	uint32_t chunk_size;
-};
-
-struct disk_exception {
-	uint64_t old_chunk;
-	uint64_t new_chunk;
-};
-
-struct commit_callback {
-	void (*callback)(void *, int success);
-	void *context;
-};
-
-/*
- * The top level structure for a persistent exception store.
- */
-struct pstore {
-	struct dm_snapshot *snap;	/* up pointer to my snapshot */
-	int version;
-	int valid;
-	uint32_t exceptions_per_area;
-
-	/*
-	 * Now that we have an asynchronous kcopyd there is no
-	 * need for large chunk sizes, so it wont hurt to have a
-	 * whole chunks worth of metadata in memory at once.
-	 */
-	void *area;
-
-	/*
-	 * An area of zeros used to clear the next area.
-	 */
-	void *zero_area;
-
-	/*
-	 * Used to keep track of which metadata area the data in
-	 * 'chunk' refers to.
-	 */
-	chunk_t current_area;
-
-	/*
-	 * The next free chunk for an exception.
-	 */
-	chunk_t next_free;
-
-	/*
-	 * The index of next free exception in the current
-	 * metadata area.
-	 */
-	uint32_t current_committed;
-
-	atomic_t pending_count;
-	uint32_t callback_count;
-	struct commit_callback *callbacks;
-	struct dm_io_client *io_client;
-
-	struct workqueue_struct *metadata_wq;
-};
-
-static unsigned sectors_to_pages(unsigned sectors)
-{
-	return DIV_ROUND_UP(sectors, PAGE_SIZE >> 9);
-}
-
-static int alloc_area(struct pstore *ps)
-{
-	int r = -ENOMEM;
-	size_t len;
-
-	len = ps->snap->chunk_size << SECTOR_SHIFT;
-
-	/*
-	 * Allocate the chunk_size block of memory that will hold
-	 * a single metadata area.
-	 */
-	ps->area = vmalloc(len);
-	if (!ps->area)
-		return r;
-
-	ps->zero_area = vmalloc(len);
-	if (!ps->zero_area) {
-		vfree(ps->area);
-		return r;
-	}
-	memset(ps->zero_area, 0, len);
-
-	return 0;
-}
-
-static void free_area(struct pstore *ps)
-{
-	vfree(ps->area);
-	ps->area = NULL;
-	vfree(ps->zero_area);
-	ps->zero_area = NULL;
-}
-
-struct mdata_req {
-	struct dm_io_region *where;
-	struct dm_io_request *io_req;
-	struct work_struct work;
-	int result;
-};
-
-static void do_metadata(struct work_struct *work)
-{
-	struct mdata_req *req = container_of(work, struct mdata_req, work);
-
-	req->result = dm_io(req->io_req, 1, req->where, NULL);
-}
-
-/*
- * Read or write a chunk aligned and sized block of data from a device.
- */
-static int chunk_io(struct pstore *ps, chunk_t chunk, int rw, int metadata)
-{
-	struct dm_io_region where = {
-		.bdev = ps->snap->cow->bdev,
-		.sector = ps->snap->chunk_size * chunk,
-		.count = ps->snap->chunk_size,
-	};
-	struct dm_io_request io_req = {
-		.bi_rw = rw,
-		.mem.type = DM_IO_VMA,
-		.mem.ptr.vma = ps->area,
-		.client = ps->io_client,
-		.notify.fn = NULL,
-	};
-	struct mdata_req req;
-
-	if (!metadata)
-		return dm_io(&io_req, 1, &where, NULL);
-
-	req.where = &where;
-	req.io_req = &io_req;
-
-	/*
-	 * Issue the synchronous I/O from a different thread
-	 * to avoid generic_make_request recursion.
-	 */
-	INIT_WORK(&req.work, do_metadata);
-	queue_work(ps->metadata_wq, &req.work);
-	flush_workqueue(ps->metadata_wq);
-
-	return req.result;
-}
-
-/*
- * Convert a metadata area index to a chunk index.
- */
-static chunk_t area_location(struct pstore *ps, chunk_t area)
-{
-	return 1 + ((ps->exceptions_per_area + 1) * area);
-}
-
-/*
- * Read or write a metadata area.  Remembering to skip the first
- * chunk which holds the header.
- */
-static int area_io(struct pstore *ps, int rw)
-{
-	int r;
-	chunk_t chunk;
-
-	chunk = area_location(ps, ps->current_area);
-
-	r = chunk_io(ps, chunk, rw, 0);
-	if (r)
-		return r;
-
-	return 0;
-}
-
-static void zero_memory_area(struct pstore *ps)
-{
-	memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT);
-}
-
-static int zero_disk_area(struct pstore *ps, chunk_t area)
-{
-	struct dm_io_region where = {
-		.bdev = ps->snap->cow->bdev,
-		.sector = ps->snap->chunk_size * area_location(ps, area),
-		.count = ps->snap->chunk_size,
-	};
-	struct dm_io_request io_req = {
-		.bi_rw = WRITE,
-		.mem.type = DM_IO_VMA,
-		.mem.ptr.vma = ps->zero_area,
-		.client = ps->io_client,
-		.notify.fn = NULL,
-	};
-
-	return dm_io(&io_req, 1, &where, NULL);
-}
-
-static int read_header(struct pstore *ps, int *new_snapshot)
-{
-	int r;
-	struct disk_header *dh;
-	chunk_t chunk_size;
-	int chunk_size_supplied = 1;
-
-	/*
-	 * Use default chunk size (or hardsect_size, if larger) if none supplied
-	 */
-	if (!ps->snap->chunk_size) {
-        	ps->snap->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
-		    bdev_hardsect_size(ps->snap->cow->bdev) >> 9);
-		ps->snap->chunk_mask = ps->snap->chunk_size - 1;
-		ps->snap->chunk_shift = ffs(ps->snap->chunk_size) - 1;
-		chunk_size_supplied = 0;
-	}
-
-	ps->io_client = dm_io_client_create(sectors_to_pages(ps->snap->
-							     chunk_size));
-	if (IS_ERR(ps->io_client))
-		return PTR_ERR(ps->io_client);
-
-	r = alloc_area(ps);
-	if (r)
-		return r;
-
-	r = chunk_io(ps, 0, READ, 1);
-	if (r)
-		goto bad;
-
-	dh = (struct disk_header *) ps->area;
-
-	if (le32_to_cpu(dh->magic) == 0) {
-		*new_snapshot = 1;
-		return 0;
-	}
-
-	if (le32_to_cpu(dh->magic) != SNAP_MAGIC) {
-		DMWARN("Invalid or corrupt snapshot");
-		r = -ENXIO;
-		goto bad;
-	}
-
-	*new_snapshot = 0;
-	ps->valid = le32_to_cpu(dh->valid);
-	ps->version = le32_to_cpu(dh->version);
-	chunk_size = le32_to_cpu(dh->chunk_size);
-
-	if (!chunk_size_supplied || ps->snap->chunk_size == chunk_size)
-		return 0;
-
-	DMWARN("chunk size %llu in device metadata overrides "
-	       "table chunk size of %llu.",
-	       (unsigned long long)chunk_size,
-	       (unsigned long long)ps->snap->chunk_size);
-
-	/* We had a bogus chunk_size. Fix stuff up. */
-	free_area(ps);
-
-	ps->snap->chunk_size = chunk_size;
-	ps->snap->chunk_mask = chunk_size - 1;
-	ps->snap->chunk_shift = ffs(chunk_size) - 1;
-
-	r = dm_io_client_resize(sectors_to_pages(ps->snap->chunk_size),
-				ps->io_client);
-	if (r)
-		return r;
-
-	r = alloc_area(ps);
-	return r;
-
-bad:
-	free_area(ps);
-	return r;
-}
-
-static int write_header(struct pstore *ps)
-{
-	struct disk_header *dh;
-
-	memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT);
-
-	dh = (struct disk_header *) ps->area;
-	dh->magic = cpu_to_le32(SNAP_MAGIC);
-	dh->valid = cpu_to_le32(ps->valid);
-	dh->version = cpu_to_le32(ps->version);
-	dh->chunk_size = cpu_to_le32(ps->snap->chunk_size);
-
-	return chunk_io(ps, 0, WRITE, 1);
-}
-
-/*
- * Access functions for the disk exceptions, these do the endian conversions.
- */
-static struct disk_exception *get_exception(struct pstore *ps, uint32_t index)
-{
-	BUG_ON(index >= ps->exceptions_per_area);
-
-	return ((struct disk_exception *) ps->area) + index;
-}
 
-static void read_exception(struct pstore *ps,
-			   uint32_t index, struct disk_exception *result)
-{
-	struct disk_exception *e = get_exception(ps, index);
-
-	/* copy it */
-	result->old_chunk = le64_to_cpu(e->old_chunk);
-	result->new_chunk = le64_to_cpu(e->new_chunk);
-}
-
-static void write_exception(struct pstore *ps,
-			    uint32_t index, struct disk_exception *de)
-{
-	struct disk_exception *e = get_exception(ps, index);
-
-	/* copy it */
-	e->old_chunk = cpu_to_le64(de->old_chunk);
-	e->new_chunk = cpu_to_le64(de->new_chunk);
-}
+#define DM_MSG_PREFIX "snapshot exception stores"
 
-/*
- * Registers the exceptions that are present in the current area.
- * 'full' is filled in to indicate if the area has been
- * filled.
- */
-static int insert_exceptions(struct pstore *ps, int *full)
+int dm_exception_store_init(void)
 {
 	int r;
-	unsigned int i;
-	struct disk_exception de;
-
-	/* presume the area is full */
-	*full = 1;
-
-	for (i = 0; i < ps->exceptions_per_area; i++) {
-		read_exception(ps, i, &de);
-
-		/*
-		 * If the new_chunk is pointing at the start of
-		 * the COW device, where the first metadata area
-		 * is we know that we've hit the end of the
-		 * exceptions.  Therefore the area is not full.
-		 */
-		if (de.new_chunk == 0LL) {
-			ps->current_committed = i;
-			*full = 0;
-			break;
-		}
-
-		/*
-		 * Keep track of the start of the free chunks.
-		 */
-		if (ps->next_free <= de.new_chunk)
-			ps->next_free = de.new_chunk + 1;
-
-		/*
-		 * Otherwise we add the exception to the snapshot.
-		 */
-		r = dm_add_exception(ps->snap, de.old_chunk, de.new_chunk);
-		if (r)
-			return r;
-	}
-
-	return 0;
-}
-
-static int read_exceptions(struct pstore *ps)
-{
-	int r, full = 1;
-
-	/*
-	 * Keeping reading chunks and inserting exceptions until
-	 * we find a partially full area.
-	 */
-	for (ps->current_area = 0; full; ps->current_area++) {
-		r = area_io(ps, READ);
-		if (r)
-			return r;
 
-		r = insert_exceptions(ps, &full);
-		if (r)
-			return r;
+	r = dm_transient_snapshot_init();
+	if (r) {
+		DMERR("Unable to register transient exception store type.");
+		goto transient_fail;
 	}
 
-	ps->current_area--;
-
-	return 0;
-}
-
-static struct pstore *get_info(struct exception_store *store)
-{
-	return (struct pstore *) store->context;
-}
-
-static void persistent_fraction_full(struct exception_store *store,
-				     sector_t *numerator, sector_t *denominator)
-{
-	*numerator = get_info(store)->next_free * store->snap->chunk_size;
-	*denominator = get_dev_size(store->snap->cow->bdev);
-}
-
-static void persistent_destroy(struct exception_store *store)
-{
-	struct pstore *ps = get_info(store);
-
-	destroy_workqueue(ps->metadata_wq);
-	dm_io_client_destroy(ps->io_client);
-	vfree(ps->callbacks);
-	free_area(ps);
-	kfree(ps);
-}
-
-static int persistent_read_metadata(struct exception_store *store)
-{
-	int r, uninitialized_var(new_snapshot);
-	struct pstore *ps = get_info(store);
-
-	/*
-	 * Read the snapshot header.
-	 */
-	r = read_header(ps, &new_snapshot);
-	if (r)
-		return r;
-
-	/*
-	 * Now we know correct chunk_size, complete the initialisation.
-	 */
-	ps->exceptions_per_area = (ps->snap->chunk_size << SECTOR_SHIFT) /
-				  sizeof(struct disk_exception);
-	ps->callbacks = dm_vcalloc(ps->exceptions_per_area,
-			sizeof(*ps->callbacks));
-	if (!ps->callbacks)
-		return -ENOMEM;
-
-	/*
-	 * Do we need to setup a new snapshot ?
-	 */
-	if (new_snapshot) {
-		r = write_header(ps);
-		if (r) {
-			DMWARN("write_header failed");
-			return r;
-		}
-
-		ps->current_area = 0;
-		zero_memory_area(ps);
-		r = zero_disk_area(ps, 0);
-		if (r) {
-			DMWARN("zero_disk_area(0) failed");
-			return r;
-		}
-	} else {
-		/*
-		 * Sanity checks.
-		 */
-		if (ps->version != SNAPSHOT_DISK_VERSION) {
-			DMWARN("unable to handle snapshot disk version %d",
-			       ps->version);
-			return -EINVAL;
-		}
-
-		/*
-		 * Metadata are valid, but snapshot is invalidated
-		 */
-		if (!ps->valid)
-			return 1;
-
-		/*
-		 * Read the metadata.
-		 */
-		r = read_exceptions(ps);
-		if (r)
-			return r;
+	r = dm_persistent_snapshot_init();
+	if (r) {
+		DMERR("Unable to register persistent exception store type");
+		goto persistent_fail;
 	}
 
 	return 0;
-}
-
-static int persistent_prepare(struct exception_store *store,
-			      struct dm_snap_exception *e)
-{
-	struct pstore *ps = get_info(store);
-	uint32_t stride;
-	chunk_t next_free;
-	sector_t size = get_dev_size(store->snap->cow->bdev);
-
-	/* Is there enough room ? */
-	if (size < ((ps->next_free + 1) * store->snap->chunk_size))
-		return -ENOSPC;
 
-	e->new_chunk = ps->next_free;
-
-	/*
-	 * Move onto the next free pending, making sure to take
-	 * into account the location of the metadata chunks.
-	 */
-	stride = (ps->exceptions_per_area + 1);
-	next_free = ++ps->next_free;
-	if (sector_div(next_free, stride) == 1)
-		ps->next_free++;
-
-	atomic_inc(&ps->pending_count);
-	return 0;
-}
-
-static void persistent_commit(struct exception_store *store,
-			      struct dm_snap_exception *e,
-			      void (*callback) (void *, int success),
-			      void *callback_context)
-{
-	unsigned int i;
-	struct pstore *ps = get_info(store);
-	struct disk_exception de;
-	struct commit_callback *cb;
-
-	de.old_chunk = e->old_chunk;
-	de.new_chunk = e->new_chunk;
-	write_exception(ps, ps->current_committed++, &de);
-
-	/*
-	 * Add the callback to the back of the array.  This code
-	 * is the only place where the callback array is
-	 * manipulated, and we know that it will never be called
-	 * multiple times concurrently.
-	 */
-	cb = ps->callbacks + ps->callback_count++;
-	cb->callback = callback;
-	cb->context = callback_context;
-
-	/*
-	 * If there are exceptions in flight and we have not yet
-	 * filled this metadata area there's nothing more to do.
-	 */
-	if (!atomic_dec_and_test(&ps->pending_count) &&
-	    (ps->current_committed != ps->exceptions_per_area))
-		return;
-
-	/*
-	 * If we completely filled the current area, then wipe the next one.
-	 */
-	if ((ps->current_committed == ps->exceptions_per_area) &&
-	     zero_disk_area(ps, ps->current_area + 1))
-		ps->valid = 0;
-
-	/*
-	 * Commit exceptions to disk.
-	 */
-	if (ps->valid && area_io(ps, WRITE))
-		ps->valid = 0;
-
-	/*
-	 * Advance to the next area if this one is full.
-	 */
-	if (ps->current_committed == ps->exceptions_per_area) {
-		ps->current_committed = 0;
-		ps->current_area++;
-		zero_memory_area(ps);
-	}
-
-	for (i = 0; i < ps->callback_count; i++) {
-		cb = ps->callbacks + i;
-		cb->callback(cb->context, ps->valid);
-	}
-
-	ps->callback_count = 0;
-}
-
-static void persistent_drop(struct exception_store *store)
-{
-	struct pstore *ps = get_info(store);
-
-	ps->valid = 0;
-	if (write_header(ps))
-		DMWARN("write header failed");
-}
-
-int dm_create_persistent(struct exception_store *store)
-{
-	struct pstore *ps;
-
-	/* allocate the pstore */
-	ps = kmalloc(sizeof(*ps), GFP_KERNEL);
-	if (!ps)
-		return -ENOMEM;
-
-	ps->snap = store->snap;
-	ps->valid = 1;
-	ps->version = SNAPSHOT_DISK_VERSION;
-	ps->area = NULL;
-	ps->next_free = 2;	/* skipping the header and first area */
-	ps->current_committed = 0;
-
-	ps->callback_count = 0;
-	atomic_set(&ps->pending_count, 0);
-	ps->callbacks = NULL;
-
-	ps->metadata_wq = create_singlethread_workqueue("ksnaphd");
-	if (!ps->metadata_wq) {
-		kfree(ps);
-		DMERR("couldn't start header metadata update thread");
-		return -ENOMEM;
-	}
-
-	store->destroy = persistent_destroy;
-	store->read_metadata = persistent_read_metadata;
-	store->prepare_exception = persistent_prepare;
-	store->commit_exception = persistent_commit;
-	store->drop_snapshot = persistent_drop;
-	store->fraction_full = persistent_fraction_full;
-	store->context = ps;
-
-	return 0;
-}
-
-/*-----------------------------------------------------------------
- * Implementation of the store for non-persistent snapshots.
- *---------------------------------------------------------------*/
-struct transient_c {
-	sector_t next_free;
-};
-
-static void transient_destroy(struct exception_store *store)
-{
-	kfree(store->context);
-}
-
-static int transient_read_metadata(struct exception_store *store)
-{
-	return 0;
-}
-
-static int transient_prepare(struct exception_store *store,
-			     struct dm_snap_exception *e)
-{
-	struct transient_c *tc = (struct transient_c *) store->context;
-	sector_t size = get_dev_size(store->snap->cow->bdev);
-
-	if (size < (tc->next_free + store->snap->chunk_size))
-		return -1;
-
-	e->new_chunk = sector_to_chunk(store->snap, tc->next_free);
-	tc->next_free += store->snap->chunk_size;
-
-	return 0;
-}
-
-static void transient_commit(struct exception_store *store,
-			     struct dm_snap_exception *e,
-			     void (*callback) (void *, int success),
-			     void *callback_context)
-{
-	/* Just succeed */
-	callback(callback_context, 1);
-}
-
-static void transient_fraction_full(struct exception_store *store,
-				    sector_t *numerator, sector_t *denominator)
-{
-	*numerator = ((struct transient_c *) store->context)->next_free;
-	*denominator = get_dev_size(store->snap->cow->bdev);
+persistent_fail:
+	dm_persistent_snapshot_exit();
+transient_fail:
+	return r;
 }
 
-int dm_create_transient(struct exception_store *store)
+void dm_exception_store_exit(void)
 {
-	struct transient_c *tc;
-
-	store->destroy = transient_destroy;
-	store->read_metadata = transient_read_metadata;
-	store->prepare_exception = transient_prepare;
-	store->commit_exception = transient_commit;
-	store->drop_snapshot = NULL;
-	store->fraction_full = transient_fraction_full;
-
-	tc = kmalloc(sizeof(struct transient_c), GFP_KERNEL);
-	if (!tc)
-		return -ENOMEM;
-
-	tc->next_free = 0;
-	store->context = tc;
-
-	return 0;
+	dm_persistent_snapshot_exit();
+	dm_transient_snapshot_exit();
 }

drivers/md/dm-exception-store.h

@@ -0,0 +1,148 @@
+/*
+ * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
+ * Copyright (C) 2008 Red Hat, Inc. All rights reserved.
+ *
+ * Device-mapper snapshot exception store.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef _LINUX_DM_EXCEPTION_STORE
+#define _LINUX_DM_EXCEPTION_STORE
+
+#include <linux/blkdev.h>
+#include <linux/device-mapper.h>
+
+/*
+ * The snapshot code deals with largish chunks of the disk at a
+ * time. Typically 32k - 512k.
+ */
+typedef sector_t chunk_t;
+
+/*
+ * An exception is used where an old chunk of data has been
+ * replaced by a new one.
+ * If chunk_t is 64 bits in size, the top 8 bits of new_chunk hold the number
+ * of chunks that follow contiguously.  Remaining bits hold the number of the
+ * chunk within the device.
+ */
+struct dm_snap_exception {
+	struct list_head hash_list;
+
+	chunk_t old_chunk;
+	chunk_t new_chunk;
+};
+
+/*
+ * Abstraction to handle the meta/layout of exception stores (the
+ * COW device).
+ */
+struct dm_exception_store {
+	/*
+	 * Destroys this object when you've finished with it.
+	 */
+	void (*destroy) (struct dm_exception_store *store);
+
+	/*
+	 * The target shouldn't read the COW device until this is
+	 * called.  As exceptions are read from the COW, they are
+	 * reported back via the callback.
+	 */
+	int (*read_metadata) (struct dm_exception_store *store,
+			      int (*callback)(void *callback_context,
+					      chunk_t old, chunk_t new),
+			      void *callback_context);
+
+	/*
+	 * Find somewhere to store the next exception.
+	 */
+	int (*prepare_exception) (struct dm_exception_store *store,
+				  struct dm_snap_exception *e);
+
+	/*
+	 * Update the metadata with this exception.
+	 */
+	void (*commit_exception) (struct dm_exception_store *store,
+				  struct dm_snap_exception *e,
+				  void (*callback) (void *, int success),
+				  void *callback_context);
+
+	/*
+	 * The snapshot is invalid, note this in the metadata.
+	 */
+	void (*drop_snapshot) (struct dm_exception_store *store);
+
+	int (*status) (struct dm_exception_store *store, status_type_t status,
+		       char *result, unsigned int maxlen);
+
+	/*
+	 * Return how full the snapshot is.
+	 */
+	void (*fraction_full) (struct dm_exception_store *store,
+			       sector_t *numerator,
+			       sector_t *denominator);
+
+	struct dm_snapshot *snap;
+	void *context;
+};
+
+/*
+ * Funtions to manipulate consecutive chunks
+ */
+#  if defined(CONFIG_LBD) || (BITS_PER_LONG == 64)
+#    define DM_CHUNK_CONSECUTIVE_BITS 8
+#    define DM_CHUNK_NUMBER_BITS 56
+
+static inline chunk_t dm_chunk_number(chunk_t chunk)
+{
+	return chunk & (chunk_t)((1ULL << DM_CHUNK_NUMBER_BITS) - 1ULL);
+}
+
+static inline unsigned dm_consecutive_chunk_count(struct dm_snap_exception *e)
+{
+	return e->new_chunk >> DM_CHUNK_NUMBER_BITS;
+}
+
+static inline void dm_consecutive_chunk_count_inc(struct dm_snap_exception *e)
+{
+	e->new_chunk += (1ULL << DM_CHUNK_NUMBER_BITS);
+
+	BUG_ON(!dm_consecutive_chunk_count(e));
+}
+
+#  else
+#    define DM_CHUNK_CONSECUTIVE_BITS 0
+
+static inline chunk_t dm_chunk_number(chunk_t chunk)
+{
+	return chunk;
+}
+
+static inline unsigned dm_consecutive_chunk_count(struct dm_snap_exception *e)
+{
+	return 0;
+}
+
+static inline void dm_consecutive_chunk_count_inc(struct dm_snap_exception *e)
+{
+}
+
+#  endif
+
+int dm_exception_store_init(void);
+void dm_exception_store_exit(void);
+
+/*
+ * Two exception store implementations.
+ */
+int dm_persistent_snapshot_init(void);
+void dm_persistent_snapshot_exit(void);
+
+int dm_transient_snapshot_init(void);
+void dm_transient_snapshot_exit(void);
+
+int dm_create_persistent(struct dm_exception_store *store);
+
+int dm_create_transient(struct dm_exception_store *store);
+
+#endif /* _LINUX_DM_EXCEPTION_STORE */

drivers/md/dm-ioctl.c

@@ -233,7 +233,7 @@ static void __hash_remove(struct hash_cell *hc)
 	}
 
 	if (hc->new_map)
-		dm_table_put(hc->new_map);
+		dm_table_destroy(hc->new_map);
 	dm_put(hc->md);
 	free_cell(hc);
 }
@@ -827,8 +827,8 @@ static int do_resume(struct dm_ioctl *param)
 
 		r = dm_swap_table(md, new_map);
 		if (r) {
+			dm_table_destroy(new_map);
 			dm_put(md);
-			dm_table_put(new_map);
 			return r;
 		}
 
@@ -836,8 +836,6 @@ static int do_resume(struct dm_ioctl *param)
 			set_disk_ro(dm_disk(md), 0);
 		else
 			set_disk_ro(dm_disk(md), 1);
-
-		dm_table_put(new_map);
 	}
 
 	if (dm_suspended(md))
@@ -1080,7 +1078,7 @@ static int table_load(struct dm_ioctl *param, size_t param_size)
 	}
 
 	if (hc->new_map)
-		dm_table_put(hc->new_map);
+		dm_table_destroy(hc->new_map);
 	hc->new_map = t;
 	up_write(&_hash_lock);
 
@@ -1109,7 +1107,7 @@ static int table_clear(struct dm_ioctl *param, size_t param_size)
 	}
 
 	if (hc->new_map) {
-		dm_table_put(hc->new_map);
+		dm_table_destroy(hc->new_map);
 		hc->new_map = NULL;
 	}
 
@@ -1550,8 +1548,10 @@ int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid)
 		goto out;
 	}
 
-	strcpy(name, hc->name);
-	strcpy(uuid, hc->uuid ? : "");
+	if (name)
+		strcpy(name, hc->name);
+	if (uuid)
+		strcpy(uuid, hc->uuid ? : "");
 
 out:
 	up_read(&_hash_lock);

drivers/md/dm-linear.c

@@ -142,6 +142,7 @@ static struct target_type linear_target = {
 	.status = linear_status,
 	.ioctl  = linear_ioctl,
 	.merge  = linear_merge,
+	.features = DM_TARGET_SUPPORTS_BARRIERS,
 };
 
 int __init dm_linear_init(void)
@@ -156,8 +157,5 @@ int __init dm_linear_init(void)
 
 void dm_linear_exit(void)
 {
-	int r = dm_unregister_target(&linear_target);
-
-	if (r < 0)
-		DMERR("unregister failed %d", r);
+	dm_unregister_target(&linear_target);
 }

drivers/md/dm-log.c

@@ -326,8 +326,6 @@ static void header_from_disk(struct log_header *core, struct log_header *disk)
 static int rw_header(struct log_c *lc, int rw)
 {
 	lc->io_req.bi_rw = rw;
-	lc->io_req.mem.ptr.vma = lc->disk_header;
-	lc->io_req.notify.fn = NULL;
 
 	return dm_io(&lc->io_req, 1, &lc->header_location, NULL);
 }
@@ -362,10 +360,15 @@ static int read_header(struct log_c *log)
 	return 0;
 }
 
-static inline int write_header(struct log_c *log)
+static int _check_region_size(struct dm_target *ti, uint32_t region_size)
 {
-	header_to_disk(&log->header, log->disk_header);
-	return rw_header(log, WRITE);
+	if (region_size < 2 || region_size > ti->len)
+		return 0;
+
+	if (!is_power_of_2(region_size))
+		return 0;
+
+	return 1;
 }
 
 /*----------------------------------------------------------------
@@ -403,8 +406,9 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 		}
 	}
 
-	if (sscanf(argv[0], "%u", &region_size) != 1) {
-		DMWARN("invalid region size string");
+	if (sscanf(argv[0], "%u", &region_size) != 1 ||
+	    !_check_region_size(ti, region_size)) {
+		DMWARN("invalid region size %s", argv[0]);
 		return -EINVAL;
 	}
 
@@ -453,8 +457,18 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 		 */
 		buf_size = dm_round_up((LOG_OFFSET << SECTOR_SHIFT) +
 				       bitset_size, ti->limits.hardsect_size);
+
+		if (buf_size > dev->bdev->bd_inode->i_size) {
+			DMWARN("log device %s too small: need %llu bytes",
+				dev->name, (unsigned long long)buf_size);
+			kfree(lc);
+			return -EINVAL;
+		}
+
 		lc->header_location.count = buf_size >> SECTOR_SHIFT;
+
 		lc->io_req.mem.type = DM_IO_VMA;
+		lc->io_req.notify.fn = NULL;
 		lc->io_req.client = dm_io_client_create(dm_div_up(buf_size,
 								   PAGE_SIZE));
 		if (IS_ERR(lc->io_req.client)) {
@@ -467,10 +481,12 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 		lc->disk_header = vmalloc(buf_size);
 		if (!lc->disk_header) {
 			DMWARN("couldn't allocate disk log buffer");
+			dm_io_client_destroy(lc->io_req.client);
 			kfree(lc);
 			return -ENOMEM;
 		}
 
+		lc->io_req.mem.ptr.vma = lc->disk_header;
 		lc->clean_bits = (void *)lc->disk_header +
 				 (LOG_OFFSET << SECTOR_SHIFT);
 	}
@@ -482,6 +498,8 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 		DMWARN("couldn't allocate sync bitset");
 		if (!dev)
 			vfree(lc->clean_bits);
+		else
+			dm_io_client_destroy(lc->io_req.client);
 		vfree(lc->disk_header);
 		kfree(lc);
 		return -ENOMEM;
@@ -495,6 +513,8 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 		vfree(lc->sync_bits);
 		if (!dev)
 			vfree(lc->clean_bits);
+		else
+			dm_io_client_destroy(lc->io_req.client);
 		vfree(lc->disk_header);
 		kfree(lc);
 		return -ENOMEM;
@@ -631,8 +651,10 @@ static int disk_resume(struct dm_dirty_log *log)
 	/* set the correct number of regions in the header */
 	lc->header.nr_regions = lc->region_count;
 
+	header_to_disk(&lc->header, lc->disk_header);
+
 	/* write the new header */
-	r = write_header(lc);
+	r = rw_header(lc, WRITE);
 	if (r) {
 		DMWARN("%s: Failed to write header on dirty region log device",
 		       lc->log_dev->name);
@@ -682,7 +704,7 @@ static int disk_flush(struct dm_dirty_log *log)
 	if (!lc->touched)
 		return 0;
 
-	r = write_header(lc);
+	r = rw_header(lc, WRITE);
 	if (r)
 		fail_log_device(lc);
 	else

drivers/md/dm-mpath.c

@@ -889,7 +889,7 @@ static int fail_path(struct pgpath *pgpath)
 	dm_path_uevent(DM_UEVENT_PATH_FAILED, m->ti,
 		      pgpath->path.dev->name, m->nr_valid_paths);
 
-	queue_work(kmultipathd, &m->trigger_event);
+	schedule_work(&m->trigger_event);
 	queue_work(kmultipathd, &pgpath->deactivate_path);
 
 out:
@@ -932,7 +932,7 @@ static int reinstate_path(struct pgpath *pgpath)
 	dm_path_uevent(DM_UEVENT_PATH_REINSTATED, m->ti,
 		      pgpath->path.dev->name, m->nr_valid_paths);
 
-	queue_work(kmultipathd, &m->trigger_event);
+	schedule_work(&m->trigger_event);
 
 out:
 	spin_unlock_irqrestore(&m->lock, flags);
@@ -976,7 +976,7 @@ static void bypass_pg(struct multipath *m, struct priority_group *pg,
 
 	spin_unlock_irqrestore(&m->lock, flags);
 
-	queue_work(kmultipathd, &m->trigger_event);
+	schedule_work(&m->trigger_event);
 }
 
 /*
@@ -1006,7 +1006,7 @@ static int switch_pg_num(struct multipath *m, const char *pgstr)
 	}
 	spin_unlock_irqrestore(&m->lock, flags);
 
-	queue_work(kmultipathd, &m->trigger_event);
+	schedule_work(&m->trigger_event);
 	return 0;
 }
 
@@ -1495,14 +1495,10 @@ static int __init dm_multipath_init(void)
 
 static void __exit dm_multipath_exit(void)
 {
-	int r;
-
 	destroy_workqueue(kmpath_handlerd);
 	destroy_workqueue(kmultipathd);
 
-	r = dm_unregister_target(&multipath_target);
-	if (r < 0)
-		DMERR("target unregister failed %d", r);
+	dm_unregister_target(&multipath_target);
 	kmem_cache_destroy(_mpio_cache);
 }
 

drivers/md/dm-raid1.c

@@ -197,9 +197,6 @@ static void fail_mirror(struct mirror *m, enum dm_raid1_error error_type)
 	struct mirror_set *ms = m->ms;
 	struct mirror *new;
 
-	if (!errors_handled(ms))
-		return;
-
 	/*
 	 * error_count is used for nothing more than a
 	 * simple way to tell if a device has encountered
@@ -210,6 +207,9 @@ static void fail_mirror(struct mirror *m, enum dm_raid1_error error_type)
 	if (test_and_set_bit(error_type, &m->error_type))
 		return;
 
+	if (!errors_handled(ms))
+		return;
+
 	if (m != get_default_mirror(ms))
 		goto out;
 
@@ -808,12 +808,6 @@ static void free_context(struct mirror_set *ms, struct dm_target *ti,
 	kfree(ms);
 }
 
-static inline int _check_region_size(struct dm_target *ti, uint32_t size)
-{
-	return !(size % (PAGE_SIZE >> 9) || !is_power_of_2(size) ||
-		 size > ti->len);
-}
-
 static int get_mirror(struct mirror_set *ms, struct dm_target *ti,
 		      unsigned int mirror, char **argv)
 {
@@ -872,12 +866,6 @@ static struct dm_dirty_log *create_dirty_log(struct dm_target *ti,
 		return NULL;
 	}
 
-	if (!_check_region_size(ti, dl->type->get_region_size(dl))) {
-		ti->error = "Invalid region size";
-		dm_dirty_log_destroy(dl);
-		return NULL;
-	}
-
 	return dl;
 }
 
@@ -1300,11 +1288,7 @@ static int __init dm_mirror_init(void)
 
 static void __exit dm_mirror_exit(void)
 {
-	int r;
-
-	r = dm_unregister_target(&mirror_target);
-	if (r < 0)
-		DMERR("unregister failed %d", r);
+	dm_unregister_target(&mirror_target);
 }
 
 /* Module hooks */

drivers/md/dm-snap-persistent.c

@@ -0,0 +1,704 @@
+/*
+ * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
+ * Copyright (C) 2006-2008 Red Hat GmbH
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-exception-store.h"
+#include "dm-snap.h"
+
+#include <linux/mm.h>
+#include <linux/pagemap.h>
+#include <linux/vmalloc.h>
+#include <linux/slab.h>
+#include <linux/dm-io.h>
+
+#define DM_MSG_PREFIX "persistent snapshot"
+#define DM_CHUNK_SIZE_DEFAULT_SECTORS 32	/* 16KB */
+
+/*-----------------------------------------------------------------
+ * Persistent snapshots, by persistent we mean that the snapshot
+ * will survive a reboot.
+ *---------------------------------------------------------------*/
+
+/*
+ * We need to store a record of which parts of the origin have
+ * been copied to the snapshot device.  The snapshot code
+ * requires that we copy exception chunks to chunk aligned areas
+ * of the COW store.  It makes sense therefore, to store the
+ * metadata in chunk size blocks.
+ *
+ * There is no backward or forward compatibility implemented,
+ * snapshots with different disk versions than the kernel will
+ * not be usable.  It is expected that "lvcreate" will blank out
+ * the start of a fresh COW device before calling the snapshot
+ * constructor.
+ *
+ * The first chunk of the COW device just contains the header.
+ * After this there is a chunk filled with exception metadata,
+ * followed by as many exception chunks as can fit in the
+ * metadata areas.
+ *
+ * All on disk structures are in little-endian format.  The end
+ * of the exceptions info is indicated by an exception with a
+ * new_chunk of 0, which is invalid since it would point to the
+ * header chunk.
+ */
+
+/*
+ * Magic for persistent snapshots: "SnAp" - Feeble isn't it.
+ */
+#define SNAP_MAGIC 0x70416e53
+
+/*
+ * The on-disk version of the metadata.
+ */
+#define SNAPSHOT_DISK_VERSION 1
+
+struct disk_header {
+	uint32_t magic;
+
+	/*
+	 * Is this snapshot valid.  There is no way of recovering
+	 * an invalid snapshot.
+	 */
+	uint32_t valid;
+
+	/*
+	 * Simple, incrementing version. no backward
+	 * compatibility.
+	 */
+	uint32_t version;
+
+	/* In sectors */
+	uint32_t chunk_size;
+};
+
+struct disk_exception {
+	uint64_t old_chunk;
+	uint64_t new_chunk;
+};
+
+struct commit_callback {
+	void (*callback)(void *, int success);
+	void *context;
+};
+
+/*
+ * The top level structure for a persistent exception store.
+ */
+struct pstore {
+	struct dm_snapshot *snap;	/* up pointer to my snapshot */
+	int version;
+	int valid;
+	uint32_t exceptions_per_area;
+
+	/*
+	 * Now that we have an asynchronous kcopyd there is no
+	 * need for large chunk sizes, so it wont hurt to have a
+	 * whole chunks worth of metadata in memory at once.
+	 */
+	void *area;
+
+	/*
+	 * An area of zeros used to clear the next area.
+	 */
+	void *zero_area;
+
+	/*
+	 * Used to keep track of which metadata area the data in
+	 * 'chunk' refers to.
+	 */
+	chunk_t current_area;
+
+	/*
+	 * The next free chunk for an exception.
+	 */
+	chunk_t next_free;
+
+	/*
+	 * The index of next free exception in the current
+	 * metadata area.
+	 */
+	uint32_t current_committed;
+
+	atomic_t pending_count;
+	uint32_t callback_count;
+	struct commit_callback *callbacks;
+	struct dm_io_client *io_client;
+
+	struct workqueue_struct *metadata_wq;
+};
+
+static unsigned sectors_to_pages(unsigned sectors)
+{
+	return DIV_ROUND_UP(sectors, PAGE_SIZE >> 9);
+}
+
+static int alloc_area(struct pstore *ps)
+{
+	int r = -ENOMEM;
+	size_t len;
+
+	len = ps->snap->chunk_size << SECTOR_SHIFT;
+
+	/*
+	 * Allocate the chunk_size block of memory that will hold
+	 * a single metadata area.
+	 */
+	ps->area = vmalloc(len);
+	if (!ps->area)
+		return r;
+
+	ps->zero_area = vmalloc(len);
+	if (!ps->zero_area) {
+		vfree(ps->area);
+		return r;
+	}
+	memset(ps->zero_area, 0, len);
+
+	return 0;
+}
+
+static void free_area(struct pstore *ps)
+{
+	vfree(ps->area);
+	ps->area = NULL;
+	vfree(ps->zero_area);
+	ps->zero_area = NULL;
+}
+
+struct mdata_req {
+	struct dm_io_region *where;
+	struct dm_io_request *io_req;
+	struct work_struct work;
+	int result;
+};
+
+static void do_metadata(struct work_struct *work)
+{
+	struct mdata_req *req = container_of(work, struct mdata_req, work);
+
+	req->result = dm_io(req->io_req, 1, req->where, NULL);
+}
+
+/*
+ * Read or write a chunk aligned and sized block of data from a device.
+ */
+static int chunk_io(struct pstore *ps, chunk_t chunk, int rw, int metadata)
+{
+	struct dm_io_region where = {
+		.bdev = ps->snap->cow->bdev,
+		.sector = ps->snap->chunk_size * chunk,
+		.count = ps->snap->chunk_size,
+	};
+	struct dm_io_request io_req = {
+		.bi_rw = rw,
+		.mem.type = DM_IO_VMA,
+		.mem.ptr.vma = ps->area,
+		.client = ps->io_client,
+		.notify.fn = NULL,
+	};
+	struct mdata_req req;
+
+	if (!metadata)
+		return dm_io(&io_req, 1, &where, NULL);
+
+	req.where = &where;
+	req.io_req = &io_req;
+
+	/*
+	 * Issue the synchronous I/O from a different thread
+	 * to avoid generic_make_request recursion.
+	 */
+	INIT_WORK(&req.work, do_metadata);
+	queue_work(ps->metadata_wq, &req.work);
+	flush_workqueue(ps->metadata_wq);
+
+	return req.result;
+}
+
+/*
+ * Convert a metadata area index to a chunk index.
+ */
+static chunk_t area_location(struct pstore *ps, chunk_t area)
+{
+	return 1 + ((ps->exceptions_per_area + 1) * area);
+}
+
+/*
+ * Read or write a metadata area.  Remembering to skip the first
+ * chunk which holds the header.
+ */
+static int area_io(struct pstore *ps, int rw)
+{
+	int r;
+	chunk_t chunk;
+
+	chunk = area_location(ps, ps->current_area);
+
+	r = chunk_io(ps, chunk, rw, 0);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static void zero_memory_area(struct pstore *ps)
+{
+	memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT);
+}
+
+static int zero_disk_area(struct pstore *ps, chunk_t area)
+{
+	struct dm_io_region where = {
+		.bdev = ps->snap->cow->bdev,
+		.sector = ps->snap->chunk_size * area_location(ps, area),
+		.count = ps->snap->chunk_size,
+	};
+	struct dm_io_request io_req = {
+		.bi_rw = WRITE,
+		.mem.type = DM_IO_VMA,
+		.mem.ptr.vma = ps->zero_area,
+		.client = ps->io_client,
+		.notify.fn = NULL,
+	};
+
+	return dm_io(&io_req, 1, &where, NULL);
+}
+
+static int read_header(struct pstore *ps, int *new_snapshot)
+{
+	int r;
+	struct disk_header *dh;
+	chunk_t chunk_size;
+	int chunk_size_supplied = 1;
+
+	/*
+	 * Use default chunk size (or hardsect_size, if larger) if none supplied
+	 */
+	if (!ps->snap->chunk_size) {
+		ps->snap->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
+		    bdev_hardsect_size(ps->snap->cow->bdev) >> 9);
+		ps->snap->chunk_mask = ps->snap->chunk_size - 1;
+		ps->snap->chunk_shift = ffs(ps->snap->chunk_size) - 1;
+		chunk_size_supplied = 0;
+	}
+
+	ps->io_client = dm_io_client_create(sectors_to_pages(ps->snap->
+							     chunk_size));
+	if (IS_ERR(ps->io_client))
+		return PTR_ERR(ps->io_client);
+
+	r = alloc_area(ps);
+	if (r)
+		return r;
+
+	r = chunk_io(ps, 0, READ, 1);
+	if (r)
+		goto bad;
+
+	dh = (struct disk_header *) ps->area;
+
+	if (le32_to_cpu(dh->magic) == 0) {
+		*new_snapshot = 1;
+		return 0;
+	}
+
+	if (le32_to_cpu(dh->magic) != SNAP_MAGIC) {
+		DMWARN("Invalid or corrupt snapshot");
+		r = -ENXIO;
+		goto bad;
+	}
+
+	*new_snapshot = 0;
+	ps->valid = le32_to_cpu(dh->valid);
+	ps->version = le32_to_cpu(dh->version);
+	chunk_size = le32_to_cpu(dh->chunk_size);
+
+	if (!chunk_size_supplied || ps->snap->chunk_size == chunk_size)
+		return 0;
+
+	DMWARN("chunk size %llu in device metadata overrides "
+	       "table chunk size of %llu.",
+	       (unsigned long long)chunk_size,
+	       (unsigned long long)ps->snap->chunk_size);
+
+	/* We had a bogus chunk_size. Fix stuff up. */
+	free_area(ps);
+
+	ps->snap->chunk_size = chunk_size;
+	ps->snap->chunk_mask = chunk_size - 1;
+	ps->snap->chunk_shift = ffs(chunk_size) - 1;
+
+	r = dm_io_client_resize(sectors_to_pages(ps->snap->chunk_size),
+				ps->io_client);
+	if (r)
+		return r;
+
+	r = alloc_area(ps);
+	return r;
+
+bad:
+	free_area(ps);
+	return r;
+}
+
+static int write_header(struct pstore *ps)
+{
+	struct disk_header *dh;
+
+	memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT);
+
+	dh = (struct disk_header *) ps->area;
+	dh->magic = cpu_to_le32(SNAP_MAGIC);
+	dh->valid = cpu_to_le32(ps->valid);
+	dh->version = cpu_to_le32(ps->version);
+	dh->chunk_size = cpu_to_le32(ps->snap->chunk_size);
+
+	return chunk_io(ps, 0, WRITE, 1);
+}
+
+/*
+ * Access functions for the disk exceptions, these do the endian conversions.
+ */
+static struct disk_exception *get_exception(struct pstore *ps, uint32_t index)
+{
+	BUG_ON(index >= ps->exceptions_per_area);
+
+	return ((struct disk_exception *) ps->area) + index;
+}
+
+static void read_exception(struct pstore *ps,
+			   uint32_t index, struct disk_exception *result)
+{
+	struct disk_exception *e = get_exception(ps, index);
+
+	/* copy it */
+	result->old_chunk = le64_to_cpu(e->old_chunk);
+	result->new_chunk = le64_to_cpu(e->new_chunk);
+}
+
+static void write_exception(struct pstore *ps,
+			    uint32_t index, struct disk_exception *de)
+{
+	struct disk_exception *e = get_exception(ps, index);
+
+	/* copy it */
+	e->old_chunk = cpu_to_le64(de->old_chunk);
+	e->new_chunk = cpu_to_le64(de->new_chunk);
+}
+
+/*
+ * Registers the exceptions that are present in the current area.
+ * 'full' is filled in to indicate if the area has been
+ * filled.
+ */
+static int insert_exceptions(struct pstore *ps,
+			     int (*callback)(void *callback_context,
+					     chunk_t old, chunk_t new),
+			     void *callback_context,
+			     int *full)
+{
+	int r;
+	unsigned int i;
+	struct disk_exception de;
+
+	/* presume the area is full */
+	*full = 1;
+
+	for (i = 0; i < ps->exceptions_per_area; i++) {
+		read_exception(ps, i, &de);
+
+		/*
+		 * If the new_chunk is pointing at the start of
+		 * the COW device, where the first metadata area
+		 * is we know that we've hit the end of the
+		 * exceptions.  Therefore the area is not full.
+		 */
+		if (de.new_chunk == 0LL) {
+			ps->current_committed = i;
+			*full = 0;
+			break;
+		}
+
+		/*
+		 * Keep track of the start of the free chunks.
+		 */
+		if (ps->next_free <= de.new_chunk)
+			ps->next_free = de.new_chunk + 1;
+
+		/*
+		 * Otherwise we add the exception to the snapshot.
+		 */
+		r = callback(callback_context, de.old_chunk, de.new_chunk);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static int read_exceptions(struct pstore *ps,
+			   int (*callback)(void *callback_context, chunk_t old,
+					   chunk_t new),
+			   void *callback_context)
+{
+	int r, full = 1;
+
+	/*
+	 * Keeping reading chunks and inserting exceptions until
+	 * we find a partially full area.
+	 */
+	for (ps->current_area = 0; full; ps->current_area++) {
+		r = area_io(ps, READ);
+		if (r)
+			return r;
+
+		r = insert_exceptions(ps, callback, callback_context, &full);
+		if (r)
+			return r;
+	}
+
+	ps->current_area--;
+
+	return 0;
+}
+
+static struct pstore *get_info(struct dm_exception_store *store)
+{
+	return (struct pstore *) store->context;
+}
+
+static void persistent_fraction_full(struct dm_exception_store *store,
+				     sector_t *numerator, sector_t *denominator)
+{
+	*numerator = get_info(store)->next_free * store->snap->chunk_size;
+	*denominator = get_dev_size(store->snap->cow->bdev);
+}
+
+static void persistent_destroy(struct dm_exception_store *store)
+{
+	struct pstore *ps = get_info(store);
+
+	destroy_workqueue(ps->metadata_wq);
+	dm_io_client_destroy(ps->io_client);
+	vfree(ps->callbacks);
+	free_area(ps);
+	kfree(ps);
+}
+
+static int persistent_read_metadata(struct dm_exception_store *store,
+				    int (*callback)(void *callback_context,
+						    chunk_t old, chunk_t new),
+				    void *callback_context)
+{
+	int r, uninitialized_var(new_snapshot);
+	struct pstore *ps = get_info(store);
+
+	/*
+	 * Read the snapshot header.
+	 */
+	r = read_header(ps, &new_snapshot);
+	if (r)
+		return r;
+
+	/*
+	 * Now we know correct chunk_size, complete the initialisation.
+	 */
+	ps->exceptions_per_area = (ps->snap->chunk_size << SECTOR_SHIFT) /
+				  sizeof(struct disk_exception);
+	ps->callbacks = dm_vcalloc(ps->exceptions_per_area,
+			sizeof(*ps->callbacks));
+	if (!ps->callbacks)
+		return -ENOMEM;
+
+	/*
+	 * Do we need to setup a new snapshot ?
+	 */
+	if (new_snapshot) {
+		r = write_header(ps);
+		if (r) {
+			DMWARN("write_header failed");
+			return r;
+		}
+
+		ps->current_area = 0;
+		zero_memory_area(ps);
+		r = zero_disk_area(ps, 0);
+		if (r) {
+			DMWARN("zero_disk_area(0) failed");
+			return r;
+		}
+	} else {
+		/*
+		 * Sanity checks.
+		 */
+		if (ps->version != SNAPSHOT_DISK_VERSION) {
+			DMWARN("unable to handle snapshot disk version %d",
+			       ps->version);
+			return -EINVAL;
+		}
+
+		/*
+		 * Metadata are valid, but snapshot is invalidated
+		 */
+		if (!ps->valid)
+			return 1;
+
+		/*
+		 * Read the metadata.
+		 */
+		r = read_exceptions(ps, callback, callback_context);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static int persistent_prepare_exception(struct dm_exception_store *store,
+					struct dm_snap_exception *e)
+{
+	struct pstore *ps = get_info(store);
+	uint32_t stride;
+	chunk_t next_free;
+	sector_t size = get_dev_size(store->snap->cow->bdev);
+
+	/* Is there enough room ? */
+	if (size < ((ps->next_free + 1) * store->snap->chunk_size))
+		return -ENOSPC;
+
+	e->new_chunk = ps->next_free;
+
+	/*
+	 * Move onto the next free pending, making sure to take
+	 * into account the location of the metadata chunks.
+	 */
+	stride = (ps->exceptions_per_area + 1);
+	next_free = ++ps->next_free;
+	if (sector_div(next_free, stride) == 1)
+		ps->next_free++;
+
+	atomic_inc(&ps->pending_count);
+	return 0;
+}
+
+static void persistent_commit_exception(struct dm_exception_store *store,
+					struct dm_snap_exception *e,
+					void (*callback) (void *, int success),
+					void *callback_context)
+{
+	unsigned int i;
+	struct pstore *ps = get_info(store);
+	struct disk_exception de;
+	struct commit_callback *cb;
+
+	de.old_chunk = e->old_chunk;
+	de.new_chunk = e->new_chunk;
+	write_exception(ps, ps->current_committed++, &de);
+
+	/*
+	 * Add the callback to the back of the array.  This code
+	 * is the only place where the callback array is
+	 * manipulated, and we know that it will never be called
+	 * multiple times concurrently.
+	 */
+	cb = ps->callbacks + ps->callback_count++;
+	cb->callback = callback;
+	cb->context = callback_context;
+
+	/*
+	 * If there are exceptions in flight and we have not yet
+	 * filled this metadata area there's nothing more to do.
+	 */
+	if (!atomic_dec_and_test(&ps->pending_count) &&
+	    (ps->current_committed != ps->exceptions_per_area))
+		return;
+
+	/*
+	 * If we completely filled the current area, then wipe the next one.
+	 */
+	if ((ps->current_committed == ps->exceptions_per_area) &&
+	     zero_disk_area(ps, ps->current_area + 1))
+		ps->valid = 0;
+
+	/*
+	 * Commit exceptions to disk.
+	 */
+	if (ps->valid && area_io(ps, WRITE))
+		ps->valid = 0;
+
+	/*
+	 * Advance to the next area if this one is full.
+	 */
+	if (ps->current_committed == ps->exceptions_per_area) {
+		ps->current_committed = 0;
+		ps->current_area++;
+		zero_memory_area(ps);
+	}
+
+	for (i = 0; i < ps->callback_count; i++) {
+		cb = ps->callbacks + i;
+		cb->callback(cb->context, ps->valid);
+	}
+
+	ps->callback_count = 0;
+}
+
+static void persistent_drop_snapshot(struct dm_exception_store *store)
+{
+	struct pstore *ps = get_info(store);
+
+	ps->valid = 0;
+	if (write_header(ps))
+		DMWARN("write header failed");
+}
+
+int dm_create_persistent(struct dm_exception_store *store)
+{
+	struct pstore *ps;
+
+	/* allocate the pstore */
+	ps = kmalloc(sizeof(*ps), GFP_KERNEL);
+	if (!ps)
+		return -ENOMEM;
+
+	ps->snap = store->snap;
+	ps->valid = 1;
+	ps->version = SNAPSHOT_DISK_VERSION;
+	ps->area = NULL;
+	ps->next_free = 2;	/* skipping the header and first area */
+	ps->current_committed = 0;
+
+	ps->callback_count = 0;
+	atomic_set(&ps->pending_count, 0);
+	ps->callbacks = NULL;
+
+	ps->metadata_wq = create_singlethread_workqueue("ksnaphd");
+	if (!ps->metadata_wq) {
+		kfree(ps);
+		DMERR("couldn't start header metadata update thread");
+		return -ENOMEM;
+	}
+
+	store->destroy = persistent_destroy;
+	store->read_metadata = persistent_read_metadata;
+	store->prepare_exception = persistent_prepare_exception;
+	store->commit_exception = persistent_commit_exception;
+	store->drop_snapshot = persistent_drop_snapshot;
+	store->fraction_full = persistent_fraction_full;
+	store->context = ps;
+
+	return 0;
+}
+
+int dm_persistent_snapshot_init(void)
+{
+	return 0;
+}
+
+void dm_persistent_snapshot_exit(void)
+{
+}

drivers/md/dm-snap-transient.c

@@ -0,0 +1,98 @@
+/*
+ * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
+ * Copyright (C) 2006-2008 Red Hat GmbH
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-exception-store.h"
+#include "dm-snap.h"
+
+#include <linux/mm.h>
+#include <linux/pagemap.h>
+#include <linux/vmalloc.h>
+#include <linux/slab.h>
+#include <linux/dm-io.h>
+
+#define DM_MSG_PREFIX "transient snapshot"
+
+/*-----------------------------------------------------------------
+ * Implementation of the store for non-persistent snapshots.
+ *---------------------------------------------------------------*/
+struct transient_c {
+	sector_t next_free;
+};
+
+static void transient_destroy(struct dm_exception_store *store)
+{
+	kfree(store->context);
+}
+
+static int transient_read_metadata(struct dm_exception_store *store,
+				   int (*callback)(void *callback_context,
+						   chunk_t old, chunk_t new),
+				   void *callback_context)
+{
+	return 0;
+}
+
+static int transient_prepare_exception(struct dm_exception_store *store,
+				       struct dm_snap_exception *e)
+{
+	struct transient_c *tc = (struct transient_c *) store->context;
+	sector_t size = get_dev_size(store->snap->cow->bdev);
+
+	if (size < (tc->next_free + store->snap->chunk_size))
+		return -1;
+
+	e->new_chunk = sector_to_chunk(store->snap, tc->next_free);
+	tc->next_free += store->snap->chunk_size;
+
+	return 0;
+}
+
+static void transient_commit_exception(struct dm_exception_store *store,
+				       struct dm_snap_exception *e,
+				       void (*callback) (void *, int success),
+				       void *callback_context)
+{
+	/* Just succeed */
+	callback(callback_context, 1);
+}
+
+static void transient_fraction_full(struct dm_exception_store *store,
+				    sector_t *numerator, sector_t *denominator)
+{
+	*numerator = ((struct transient_c *) store->context)->next_free;
+	*denominator = get_dev_size(store->snap->cow->bdev);
+}
+
+int dm_create_transient(struct dm_exception_store *store)
+{
+	struct transient_c *tc;
+
+	store->destroy = transient_destroy;
+	store->read_metadata = transient_read_metadata;
+	store->prepare_exception = transient_prepare_exception;
+	store->commit_exception = transient_commit_exception;
+	store->drop_snapshot = NULL;
+	store->fraction_full = transient_fraction_full;
+
+	tc = kmalloc(sizeof(struct transient_c), GFP_KERNEL);
+	if (!tc)
+		return -ENOMEM;
+
+	tc->next_free = 0;
+	store->context = tc;
+
+	return 0;
+}
+
+int dm_transient_snapshot_init(void)
+{
+	return 0;
+}
+
+void dm_transient_snapshot_exit(void)
+{
+}

drivers/md/dm-snap.c

@@ -9,6 +9,7 @@
 #include <linux/blkdev.h>
 #include <linux/ctype.h>
 #include <linux/device-mapper.h>
+#include <linux/delay.h>
 #include <linux/fs.h>
 #include <linux/init.h>
 #include <linux/kdev_t.h>
@@ -20,6 +21,7 @@
 #include <linux/log2.h>
 #include <linux/dm-kcopyd.h>
 
+#include "dm-exception-store.h"
 #include "dm-snap.h"
 #include "dm-bio-list.h"
 
@@ -428,8 +430,13 @@ out:
 	list_add(&new_e->hash_list, e ? &e->hash_list : l);
 }
 
-int dm_add_exception(struct dm_snapshot *s, chunk_t old, chunk_t new)
+/*
+ * Callback used by the exception stores to load exceptions when
+ * initialising.
+ */
+static int dm_add_exception(void *context, chunk_t old, chunk_t new)
 {
+	struct dm_snapshot *s = context;
 	struct dm_snap_exception *e;
 
 	e = alloc_exception();
@@ -658,7 +665,7 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	spin_lock_init(&s->tracked_chunk_lock);
 
 	/* Metadata must only be loaded into one table at once */
-	r = s->store.read_metadata(&s->store);
+	r = s->store.read_metadata(&s->store, dm_add_exception, (void *)s);
 	if (r < 0) {
 		ti->error = "Failed to read snapshot metadata";
 		goto bad_load_and_register;
@@ -735,7 +742,7 @@ static void snapshot_dtr(struct dm_target *ti)
 	unregister_snapshot(s);
 
 	while (atomic_read(&s->pending_exceptions_count))
-		yield();
+		msleep(1);
 	/*
 	 * Ensure instructions in mempool_destroy aren't reordered
 	 * before atomic_read.
@@ -888,10 +895,10 @@ static void pending_complete(struct dm_snap_pending_exception *pe, int success)
 
 	/*
 	 * Check for conflicting reads. This is extremely improbable,
-	 * so yield() is sufficient and there is no need for a wait queue.
+	 * so msleep(1) is sufficient and there is no need for a wait queue.
 	 */
 	while (__chunk_is_tracked(s, pe->e.old_chunk))
-		yield();
+		msleep(1);
 
 	/*
 	 * Add a proper exception, and remove the
@@ -1404,6 +1411,12 @@ static int __init dm_snapshot_init(void)
 {
 	int r;
 
+	r = dm_exception_store_init();
+	if (r) {
+		DMERR("Failed to initialize exception stores");
+		return r;
+	}
+
 	r = dm_register_target(&snapshot_target);
 	if (r) {
 		DMERR("snapshot target register failed %d", r);
@@ -1452,39 +1465,34 @@ static int __init dm_snapshot_init(void)
 
 	return 0;
 
-      bad_pending_pool:
+bad_pending_pool:
 	kmem_cache_destroy(tracked_chunk_cache);
-      bad5:
+bad5:
 	kmem_cache_destroy(pending_cache);
-      bad4:
+bad4:
 	kmem_cache_destroy(exception_cache);
-      bad3:
+bad3:
 	exit_origin_hash();
-      bad2:
+bad2:
 	dm_unregister_target(&origin_target);
-      bad1:
+bad1:
 	dm_unregister_target(&snapshot_target);
 	return r;
 }
 
 static void __exit dm_snapshot_exit(void)
 {
-	int r;
-
 	destroy_workqueue(ksnapd);
 
-	r = dm_unregister_target(&snapshot_target);
-	if (r)
-		DMERR("snapshot unregister failed %d", r);
-
-	r = dm_unregister_target(&origin_target);
-	if (r)
-		DMERR("origin unregister failed %d", r);
+	dm_unregister_target(&snapshot_target);
+	dm_unregister_target(&origin_target);
 
 	exit_origin_hash();
 	kmem_cache_destroy(pending_cache);
 	kmem_cache_destroy(exception_cache);
 	kmem_cache_destroy(tracked_chunk_cache);
+
+	dm_exception_store_exit();
 }
 
 /* Module hooks */

drivers/md/dm-snap.h

@@ -1,6 +1,4 @@
 /*
- * dm-snapshot.c
- *
  * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
  *
  * This file is released under the GPL.
@@ -10,6 +8,7 @@
 #define DM_SNAPSHOT_H
 
 #include <linux/device-mapper.h>
+#include "dm-exception-store.h"
 #include "dm-bio-list.h"
 #include <linux/blkdev.h>
 #include <linux/workqueue.h>
@@ -20,116 +19,6 @@ struct exception_table {
 	struct list_head *table;
 };
 
-/*
- * The snapshot code deals with largish chunks of the disk at a
- * time. Typically 32k - 512k.
- */
-typedef sector_t chunk_t;
-
-/*
- * An exception is used where an old chunk of data has been
- * replaced by a new one.
- * If chunk_t is 64 bits in size, the top 8 bits of new_chunk hold the number
- * of chunks that follow contiguously.  Remaining bits hold the number of the
- * chunk within the device.
- */
-struct dm_snap_exception {
-	struct list_head hash_list;
-
-	chunk_t old_chunk;
-	chunk_t new_chunk;
-};
-
-/*
- * Funtions to manipulate consecutive chunks
- */
-#  if defined(CONFIG_LBD) || (BITS_PER_LONG == 64)
-#    define DM_CHUNK_CONSECUTIVE_BITS 8
-#    define DM_CHUNK_NUMBER_BITS 56
-
-static inline chunk_t dm_chunk_number(chunk_t chunk)
-{
-	return chunk & (chunk_t)((1ULL << DM_CHUNK_NUMBER_BITS) - 1ULL);
-}
-
-static inline unsigned dm_consecutive_chunk_count(struct dm_snap_exception *e)
-{
-	return e->new_chunk >> DM_CHUNK_NUMBER_BITS;
-}
-
-static inline void dm_consecutive_chunk_count_inc(struct dm_snap_exception *e)
-{
-	e->new_chunk += (1ULL << DM_CHUNK_NUMBER_BITS);
-
-	BUG_ON(!dm_consecutive_chunk_count(e));
-}
-
-#  else
-#    define DM_CHUNK_CONSECUTIVE_BITS 0
-
-static inline chunk_t dm_chunk_number(chunk_t chunk)
-{
-	return chunk;
-}
-
-static inline unsigned dm_consecutive_chunk_count(struct dm_snap_exception *e)
-{
-	return 0;
-}
-
-static inline void dm_consecutive_chunk_count_inc(struct dm_snap_exception *e)
-{
-}
-
-#  endif
-
-/*
- * Abstraction to handle the meta/layout of exception stores (the
- * COW device).
- */
-struct exception_store {
-
-	/*
-	 * Destroys this object when you've finished with it.
-	 */
-	void (*destroy) (struct exception_store *store);
-
-	/*
-	 * The target shouldn't read the COW device until this is
-	 * called.
-	 */
-	int (*read_metadata) (struct exception_store *store);
-
-	/*
-	 * Find somewhere to store the next exception.
-	 */
-	int (*prepare_exception) (struct exception_store *store,
-				  struct dm_snap_exception *e);
-
-	/*
-	 * Update the metadata with this exception.
-	 */
-	void (*commit_exception) (struct exception_store *store,
-				  struct dm_snap_exception *e,
-				  void (*callback) (void *, int success),
-				  void *callback_context);
-
-	/*
-	 * The snapshot is invalid, note this in the metadata.
-	 */
-	void (*drop_snapshot) (struct exception_store *store);
-
-	/*
-	 * Return how full the snapshot is.
-	 */
-	void (*fraction_full) (struct exception_store *store,
-			       sector_t *numerator,
-			       sector_t *denominator);
-
-	struct dm_snapshot *snap;
-	void *context;
-};
-
 #define DM_TRACKED_CHUNK_HASH_SIZE	16
 #define DM_TRACKED_CHUNK_HASH(x)	((unsigned long)(x) & \
 					 (DM_TRACKED_CHUNK_HASH_SIZE - 1))
@@ -172,7 +61,7 @@ struct dm_snapshot {
 	spinlock_t pe_lock;
 
 	/* The on disk metadata handler */
-	struct exception_store store;
+	struct dm_exception_store store;
 
 	struct dm_kcopyd_client *kcopyd_client;
 
@@ -186,20 +75,6 @@ struct dm_snapshot {
 	struct hlist_head tracked_chunk_hash[DM_TRACKED_CHUNK_HASH_SIZE];
 };
 
-/*
- * Used by the exception stores to load exceptions hen
- * initialising.
- */
-int dm_add_exception(struct dm_snapshot *s, chunk_t old, chunk_t new);
-
-/*
- * Constructor and destructor for the default persistent
- * store.
- */
-int dm_create_persistent(struct exception_store *store);
-
-int dm_create_transient(struct exception_store *store);
-
 /*
  * Return the number of sectors in the device.
  */

drivers/md/dm-stripe.c

@@ -337,9 +337,7 @@ int __init dm_stripe_init(void)
 
 void dm_stripe_exit(void)
 {
-	if (dm_unregister_target(&stripe_target))
-		DMWARN("target unregistration failed");
-
+	dm_unregister_target(&stripe_target);
 	destroy_workqueue(kstriped);
 
 	return;

drivers/md/dm-sysfs.c

@@ -0,0 +1,99 @@
+/*
+ * Copyright (C) 2008 Red Hat, Inc. All rights reserved.
+ *
+ * This file is released under the GPL.
+ */
+
+#include <linux/sysfs.h>
+#include <linux/dm-ioctl.h>
+#include "dm.h"
+
+struct dm_sysfs_attr {
+	struct attribute attr;
+	ssize_t (*show)(struct mapped_device *, char *);
+	ssize_t (*store)(struct mapped_device *, char *);
+};
+
+#define DM_ATTR_RO(_name) \
+struct dm_sysfs_attr dm_attr_##_name = \
+	__ATTR(_name, S_IRUGO, dm_attr_##_name##_show, NULL)
+
+static ssize_t dm_attr_show(struct kobject *kobj, struct attribute *attr,
+			    char *page)
+{
+	struct dm_sysfs_attr *dm_attr;
+	struct mapped_device *md;
+	ssize_t ret;
+
+	dm_attr = container_of(attr, struct dm_sysfs_attr, attr);
+	if (!dm_attr->show)
+		return -EIO;
+
+	md = dm_get_from_kobject(kobj);
+	if (!md)
+		return -EINVAL;
+
+	ret = dm_attr->show(md, page);
+	dm_put(md);
+
+	return ret;
+}
+
+static ssize_t dm_attr_name_show(struct mapped_device *md, char *buf)
+{
+	if (dm_copy_name_and_uuid(md, buf, NULL))
+		return -EIO;
+
+	strcat(buf, "\n");
+	return strlen(buf);
+}
+
+static ssize_t dm_attr_uuid_show(struct mapped_device *md, char *buf)
+{
+	if (dm_copy_name_and_uuid(md, NULL, buf))
+		return -EIO;
+
+	strcat(buf, "\n");
+	return strlen(buf);
+}
+
+static DM_ATTR_RO(name);
+static DM_ATTR_RO(uuid);
+
+static struct attribute *dm_attrs[] = {
+	&dm_attr_name.attr,
+	&dm_attr_uuid.attr,
+	NULL,
+};
+
+static struct sysfs_ops dm_sysfs_ops = {
+	.show	= dm_attr_show,
+};
+
+/*
+ * dm kobject is embedded in mapped_device structure
+ * no need to define release function here
+ */
+static struct kobj_type dm_ktype = {
+	.sysfs_ops	= &dm_sysfs_ops,
+	.default_attrs	= dm_attrs,
+};
+
+/*
+ * Initialize kobj
+ * because nobody using md yet, no need to call explicit dm_get/put
+ */
+int dm_sysfs_init(struct mapped_device *md)
+{
+	return kobject_init_and_add(dm_kobject(md), &dm_ktype,
+				    &disk_to_dev(dm_disk(md))->kobj,
+				    "%s", "dm");
+}
+
+/*
+ * Remove kobj, called after all references removed
+ */
+void dm_sysfs_exit(struct mapped_device *md)
+{
+	kobject_put(dm_kobject(md));
+}

drivers/md/dm-table.c

@@ -1,6 +1,6 @@
 /*
  * Copyright (C) 2001 Sistina Software (UK) Limited.
- * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
  *
  * This file is released under the GPL.
  */
@@ -15,6 +15,7 @@
 #include <linux/slab.h>
 #include <linux/interrupt.h>
 #include <linux/mutex.h>
+#include <linux/delay.h>
 #include <asm/atomic.h>
 
 #define DM_MSG_PREFIX "table"
@@ -24,6 +25,19 @@
 #define KEYS_PER_NODE (NODE_SIZE / sizeof(sector_t))
 #define CHILDREN_PER_NODE (KEYS_PER_NODE + 1)
 
+/*
+ * The table has always exactly one reference from either mapped_device->map
+ * or hash_cell->new_map. This reference is not counted in table->holders.
+ * A pair of dm_create_table/dm_destroy_table functions is used for table
+ * creation/destruction.
+ *
+ * Temporary references from the other code increase table->holders. A pair
+ * of dm_table_get/dm_table_put functions is used to manipulate it.
+ *
+ * When the table is about to be destroyed, we wait for table->holders to
+ * drop to zero.
+ */
+
 struct dm_table {
 	struct mapped_device *md;
 	atomic_t holders;
@@ -38,6 +52,8 @@ struct dm_table {
 	sector_t *highs;
 	struct dm_target *targets;
 
+	unsigned barriers_supported:1;
+
 	/*
 	 * Indicates the rw permissions for the new logical
 	 * device.  This should be a combination of FMODE_READ
@@ -226,7 +242,8 @@ int dm_table_create(struct dm_table **result, fmode_t mode,
 		return -ENOMEM;
 
 	INIT_LIST_HEAD(&t->devices);
-	atomic_set(&t->holders, 1);
+	atomic_set(&t->holders, 0);
+	t->barriers_supported = 1;
 
 	if (!num_targets)
 		num_targets = KEYS_PER_NODE;
@@ -256,10 +273,14 @@ static void free_devices(struct list_head *devices)
 	}
 }
 
-static void table_destroy(struct dm_table *t)
+void dm_table_destroy(struct dm_table *t)
 {
 	unsigned int i;
 
+	while (atomic_read(&t->holders))
+		msleep(1);
+	smp_mb();
+
 	/* free the indexes (see dm_table_complete) */
 	if (t->depth >= 2)
 		vfree(t->index[t->depth - 2]);
@@ -297,8 +318,8 @@ void dm_table_put(struct dm_table *t)
 	if (!t)
 		return;
 
-	if (atomic_dec_and_test(&t->holders))
-		table_destroy(t);
+	smp_mb__before_atomic_dec();
+	atomic_dec(&t->holders);
 }
 
 /*
@@ -728,6 +749,10 @@ int dm_table_add_target(struct dm_table *t, const char *type,
 	/* FIXME: the plan is to combine high here and then have
 	 * the merge fn apply the target level restrictions. */
 	combine_restrictions_low(&t->limits, &tgt->limits);
+
+	if (!(tgt->type->features & DM_TARGET_SUPPORTS_BARRIERS))
+		t->barriers_supported = 0;
+
 	return 0;
 
  bad:
@@ -772,6 +797,12 @@ int dm_table_complete(struct dm_table *t)
 
 	check_for_valid_limits(&t->limits);
 
+	/*
+	 * We only support barriers if there is exactly one underlying device.
+	 */
+	if (!list_is_singular(&t->devices))
+		t->barriers_supported = 0;
+
 	/* how many indexes will the btree have ? */
 	leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
 	t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
@@ -986,6 +1017,12 @@ struct mapped_device *dm_table_get_md(struct dm_table *t)
 	return t->md;
 }
 
+int dm_table_barrier_ok(struct dm_table *t)
+{
+	return t->barriers_supported;
+}
+EXPORT_SYMBOL(dm_table_barrier_ok);
+
 EXPORT_SYMBOL(dm_vcalloc);
 EXPORT_SYMBOL(dm_get_device);
 EXPORT_SYMBOL(dm_put_device);

drivers/md/dm-target.c

@@ -130,26 +130,26 @@ int dm_register_target(struct target_type *t)
 	return rv;
 }
 
-int dm_unregister_target(struct target_type *t)
+void dm_unregister_target(struct target_type *t)
 {
 	struct tt_internal *ti;
 
 	down_write(&_lock);
 	if (!(ti = __find_target_type(t->name))) {
-		up_write(&_lock);
-		return -EINVAL;
+		DMCRIT("Unregistering unrecognised target: %s", t->name);
+		BUG();
 	}
 
 	if (ti->use) {
-		up_write(&_lock);
-		return -ETXTBSY;
+		DMCRIT("Attempt to unregister target still in use: %s",
+		       t->name);
+		BUG();
 	}
 
 	list_del(&ti->list);
 	kfree(ti);
 
 	up_write(&_lock);
-	return 0;
 }
 
 /*
@@ -187,8 +187,7 @@ int __init dm_target_init(void)
 
 void dm_target_exit(void)
 {
-	if (dm_unregister_target(&error_target))
-		DMWARN("error target unregistration failed");
+	dm_unregister_target(&error_target);
 }
 
 EXPORT_SYMBOL(dm_register_target);

drivers/md/dm-zero.c

@@ -69,10 +69,7 @@ static int __init dm_zero_init(void)
 
 static void __exit dm_zero_exit(void)
 {
-	int r = dm_unregister_target(&zero_target);
-
-	if (r < 0)
-		DMERR("unregister failed %d", r);
+	dm_unregister_target(&zero_target);
 }
 
 module_init(dm_zero_init)

drivers/md/dm.c

@@ -1,6 +1,6 @@
 /*
  * Copyright (C) 2001, 2002 Sistina Software (UK) Limited.
- * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
  *
  * This file is released under the GPL.
  */
@@ -32,6 +32,7 @@ static unsigned int _major = 0;
 
 static DEFINE_SPINLOCK(_minor_lock);
 /*
+ * For bio-based dm.
  * One of these is allocated per bio.
  */
 struct dm_io {
@@ -43,6 +44,7 @@ struct dm_io {
 };
 
 /*
+ * For bio-based dm.
  * One of these is allocated per target within a bio.  Hopefully
  * this will be simplified out one day.
  */
@@ -54,6 +56,27 @@ struct dm_target_io {
 
 DEFINE_TRACE(block_bio_complete);
 
+/*
+ * For request-based dm.
+ * One of these is allocated per request.
+ */
+struct dm_rq_target_io {
+	struct mapped_device *md;
+	struct dm_target *ti;
+	struct request *orig, clone;
+	int error;
+	union map_info info;
+};
+
+/*
+ * For request-based dm.
+ * One of these is allocated per bio.
+ */
+struct dm_rq_clone_bio_info {
+	struct bio *orig;
+	struct request *rq;
+};
+
 union map_info *dm_get_mapinfo(struct bio *bio)
 {
 	if (bio && bio->bi_private)
@@ -144,11 +167,16 @@ struct mapped_device {
 
 	/* forced geometry settings */
 	struct hd_geometry geometry;
+
+	/* sysfs handle */
+	struct kobject kobj;
 };
 
 #define MIN_IOS 256
 static struct kmem_cache *_io_cache;
 static struct kmem_cache *_tio_cache;
+static struct kmem_cache *_rq_tio_cache;
+static struct kmem_cache *_rq_bio_info_cache;
 
 static int __init local_init(void)
 {
@@ -164,9 +192,17 @@ static int __init local_init(void)
 	if (!_tio_cache)
 		goto out_free_io_cache;
 
+	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
+	if (!_rq_tio_cache)
+		goto out_free_tio_cache;
+
+	_rq_bio_info_cache = KMEM_CACHE(dm_rq_clone_bio_info, 0);
+	if (!_rq_bio_info_cache)
+		goto out_free_rq_tio_cache;
+
 	r = dm_uevent_init();
 	if (r)
-		goto out_free_tio_cache;
+		goto out_free_rq_bio_info_cache;
 
 	_major = major;
 	r = register_blkdev(_major, _name);
@@ -180,6 +216,10 @@ static int __init local_init(void)
 
 out_uevent_exit:
 	dm_uevent_exit();
+out_free_rq_bio_info_cache:
+	kmem_cache_destroy(_rq_bio_info_cache);
+out_free_rq_tio_cache:
+	kmem_cache_destroy(_rq_tio_cache);
 out_free_tio_cache:
 	kmem_cache_destroy(_tio_cache);
 out_free_io_cache:
@@ -190,6 +230,8 @@ out_free_io_cache:
 
 static void local_exit(void)
 {
+	kmem_cache_destroy(_rq_bio_info_cache);
+	kmem_cache_destroy(_rq_tio_cache);
 	kmem_cache_destroy(_tio_cache);
 	kmem_cache_destroy(_io_cache);
 	unregister_blkdev(_major, _name);
@@ -796,7 +838,11 @@ static int __split_bio(struct mapped_device *md, struct bio *bio)
 	ci.map = dm_get_table(md);
 	if (unlikely(!ci.map))
 		return -EIO;
-
+	if (unlikely(bio_barrier(bio) && !dm_table_barrier_ok(ci.map))) {
+		dm_table_put(ci.map);
+		bio_endio(bio, -EOPNOTSUPP);
+		return 0;
+	}
 	ci.md = md;
 	ci.bio = bio;
 	ci.io = alloc_io(md);
@@ -880,15 +926,6 @@ static int dm_request(struct request_queue *q, struct bio *bio)
 	struct mapped_device *md = q->queuedata;
 	int cpu;
 
-	/*
-	 * There is no use in forwarding any barrier request since we can't
-	 * guarantee it is (or can be) handled by the targets correctly.
-	 */
-	if (unlikely(bio_barrier(bio))) {
-		bio_endio(bio, -EOPNOTSUPP);
-		return 0;
-	}
-
 	down_read(&md->io_lock);
 
 	cpu = part_stat_lock();
@@ -943,8 +980,6 @@ static int dm_any_congested(void *congested_data, int bdi_bits)
 	struct mapped_device *md = congested_data;
 	struct dm_table *map;
 
-	atomic_inc(&md->pending);
-
 	if (!test_bit(DMF_BLOCK_IO, &md->flags)) {
 		map = dm_get_table(md);
 		if (map) {
@@ -953,10 +988,6 @@ static int dm_any_congested(void *congested_data, int bdi_bits)
 		}
 	}
 
-	if (!atomic_dec_return(&md->pending))
-		/* nudge anyone waiting on suspend queue */
-		wake_up(&md->wait);
-
 	return r;
 }
 
@@ -1216,10 +1247,12 @@ static int __bind(struct mapped_device *md, struct dm_table *t)
 
 	if (md->suspended_bdev)
 		__set_size(md, size);
-	if (size == 0)
+
+	if (!size) {
+		dm_table_destroy(t);
 		return 0;
+	}
 
-	dm_table_get(t);
 	dm_table_event_callback(t, event_callback, md);
 
 	write_lock(&md->map_lock);
@@ -1241,7 +1274,7 @@ static void __unbind(struct mapped_device *md)
 	write_lock(&md->map_lock);
 	md->map = NULL;
 	write_unlock(&md->map_lock);
-	dm_table_put(map);
+	dm_table_destroy(map);
 }
 
 /*
@@ -1255,6 +1288,8 @@ int dm_create(int minor, struct mapped_device **result)
 	if (!md)
 		return -ENXIO;
 
+	dm_sysfs_init(md);
+
 	*result = md;
 	return 0;
 }
@@ -1330,8 +1365,9 @@ void dm_put(struct mapped_device *md)
 			dm_table_presuspend_targets(map);
 			dm_table_postsuspend_targets(map);
 		}
-		__unbind(md);
+		dm_sysfs_exit(md);
 		dm_table_put(map);
+		__unbind(md);
 		free_dev(md);
 	}
 }
@@ -1669,6 +1705,27 @@ struct gendisk *dm_disk(struct mapped_device *md)
 	return md->disk;
 }
 
+struct kobject *dm_kobject(struct mapped_device *md)
+{
+	return &md->kobj;
+}
+
+/*
+ * struct mapped_device should not be exported outside of dm.c
+ * so use this check to verify that kobj is part of md structure
+ */
+struct mapped_device *dm_get_from_kobject(struct kobject *kobj)
+{
+	struct mapped_device *md;
+
+	md = container_of(kobj, struct mapped_device, kobj);
+	if (&md->kobj != kobj)
+		return NULL;
+
+	dm_get(md);
+	return md;
+}
+
 int dm_suspended(struct mapped_device *md)
 {
 	return test_bit(DMF_SUSPENDED, &md->flags);

drivers/md/dm.h

@@ -36,6 +36,7 @@ struct dm_table;
 /*-----------------------------------------------------------------
  * Internal table functions.
  *---------------------------------------------------------------*/
+void dm_table_destroy(struct dm_table *t);
 void dm_table_event_callback(struct dm_table *t,
 			     void (*fn)(void *), void *context);
 struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index);
@@ -51,6 +52,7 @@ int dm_table_any_congested(struct dm_table *t, int bdi_bits);
  * To check the return value from dm_table_find_target().
  */
 #define dm_target_is_valid(t) ((t)->table)
+int dm_table_barrier_ok(struct dm_table *t);
 
 /*-----------------------------------------------------------------
  * A registry of target types.
@@ -71,6 +73,14 @@ int dm_split_args(int *argc, char ***argvp, char *input);
 int dm_interface_init(void);
 void dm_interface_exit(void);
 
+/*
+ * sysfs interface
+ */
+int dm_sysfs_init(struct mapped_device *md);
+void dm_sysfs_exit(struct mapped_device *md);
+struct kobject *dm_kobject(struct mapped_device *md);
+struct mapped_device *dm_get_from_kobject(struct kobject *kobj);
+
 /*
  * Targets for linear and striped mappings
  */

include/linux/device-mapper.h

@@ -45,6 +45,8 @@ typedef void (*dm_dtr_fn) (struct dm_target *ti);
  */
 typedef int (*dm_map_fn) (struct dm_target *ti, struct bio *bio,
 			  union map_info *map_context);
+typedef int (*dm_map_request_fn) (struct dm_target *ti, struct request *clone,
+				  union map_info *map_context);
 
 /*
  * Returns:
@@ -57,6 +59,9 @@ typedef int (*dm_map_fn) (struct dm_target *ti, struct bio *bio,
 typedef int (*dm_endio_fn) (struct dm_target *ti,
 			    struct bio *bio, int error,
 			    union map_info *map_context);
+typedef int (*dm_request_endio_fn) (struct dm_target *ti,
+				    struct request *clone, int error,
+				    union map_info *map_context);
 
 typedef void (*dm_flush_fn) (struct dm_target *ti);
 typedef void (*dm_presuspend_fn) (struct dm_target *ti);
@@ -75,6 +80,13 @@ typedef int (*dm_ioctl_fn) (struct dm_target *ti, unsigned int cmd,
 typedef int (*dm_merge_fn) (struct dm_target *ti, struct bvec_merge_data *bvm,
 			    struct bio_vec *biovec, int max_size);
 
+/*
+ * Returns:
+ *    0: The target can handle the next I/O immediately.
+ *    1: The target can't handle the next I/O immediately.
+ */
+typedef int (*dm_busy_fn) (struct dm_target *ti);
+
 void dm_error(const char *message);
 
 /*
@@ -100,14 +112,23 @@ void dm_put_device(struct dm_target *ti, struct dm_dev *d);
 /*
  * Information about a target type
  */
+
+/*
+ * Target features
+ */
+#define DM_TARGET_SUPPORTS_BARRIERS 0x00000001
+
 struct target_type {
+	uint64_t features;
 	const char *name;
 	struct module *module;
 	unsigned version[3];
 	dm_ctr_fn ctr;
 	dm_dtr_fn dtr;
 	dm_map_fn map;
+	dm_map_request_fn map_rq;
 	dm_endio_fn end_io;
+	dm_request_endio_fn rq_end_io;
 	dm_flush_fn flush;
 	dm_presuspend_fn presuspend;
 	dm_postsuspend_fn postsuspend;
@@ -117,6 +138,7 @@ struct target_type {
 	dm_message_fn message;
 	dm_ioctl_fn ioctl;
 	dm_merge_fn merge;
+	dm_busy_fn busy;
 };
 
 struct io_restrictions {
@@ -157,8 +179,7 @@ struct dm_target {
 };
 
 int dm_register_target(struct target_type *t);
-int dm_unregister_target(struct target_type *t);
-
+void dm_unregister_target(struct target_type *t);
 
 /*-----------------------------------------------------------------
  * Functions for creating and manipulating mapped devices.
@@ -276,6 +297,9 @@ void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size);
  *---------------------------------------------------------------*/
 #define DM_NAME "device-mapper"
 
+#define DMCRIT(f, arg...) \
+	printk(KERN_CRIT DM_NAME ": " DM_MSG_PREFIX ": " f "\n", ## arg)
+
 #define DMERR(f, arg...) \
 	printk(KERN_ERR DM_NAME ": " DM_MSG_PREFIX ": " f "\n", ## arg)
 #define DMERR_LIMIT(f, arg...) \