o split struct mapped_device into mapped_device and dm_table

o seperated loading of a table from binding a table to the device

These should allow multiple tables to be managed by dm-fs

driver/device-mapper/device-mapper.h

@@ -35,11 +35,11 @@
 /* FIXME: Use value from local range for now, for co-existence with LVM 1 */
 #define DM_BLK_MAJOR 124
 
-struct mapped_device;
+struct dm_table;
 typedef unsigned int offset_t;
 
 /* constructor, destructor and map fn types */
-typedef int (*dm_ctr_fn)(offset_t b, offset_t e, struct mapped_device *md,
+typedef int (*dm_ctr_fn)(offset_t b, offset_t e, struct dm_table *t,
 			 const char *cb, const char *ce, void **result);
 typedef void (*dm_dtr_fn)(void *c);
 typedef int (*dm_map_fn)(struct buffer_head *bh, void *context);
@@ -47,9 +47,12 @@ typedef int (*dm_map_fn)(struct buffer_head *bh, void *context);
 int register_map_target(const char *name, dm_ctr_fn ctr, dm_dtr_fn dtr,
 			dm_map_fn map);
 
-/* contructors should call this to make sure any destination devices
+
-   are handled correctly (ie. opened/closed) */
+/* contructors should call this to make sure any
-int dm_add_device(struct mapped_device *md, kdev_t dev);
+ * destination devices are handled correctly
+ * (ie. opened/closed).
+ */
+int dm_table_add_device(struct dm_table *t, kdev_t dev);
 
 #endif
 #endif

driver/device-mapper/dm-fs.c

@@ -49,21 +49,25 @@ static struct proc_dir_entry *_control;
 
 static devfs_handle_t _dev_dir;
 
+static int process_control(const char *b, const char *e, int minor,
+			   struct dm_table **map);
+static int process_table(const char *b, const char *e, int minor,
+			 struct dm_table **map);
 static int line_splitter(struct file *file, const char *buffer,
 			 unsigned long count, void *data);
-static int process_control(const char *b, const char *e, int minor);
-static int process_table(const char *b, const char *e, int minor);
 static int get_word(const char *b, const char *e,
 		    const char **wb, const char **we);
 static int tok_cmp(const char *str, const char *b, const char *e);
 static void tok_cpy(char *dest, size_t max,
 		    const char *b, const char *e);
 
-typedef int (*process_fn)(const char *b, const char *e, int minor);
+typedef int (*process_fn)(const char *b, const char *e, int minor,
+			  struct dm_table **map);
 
 struct pf_data {
 	process_fn fn;
 	int minor;
+	struct dm_table *map;
 };
 
 int dm_fs_init(void)
@@ -119,6 +123,7 @@ int dm_fs_add(struct mapped_device *md)
 
 	pfd->fn = process_table;
 	pfd->minor = MINOR(md->dev);
+	pfd->map = 0;
 
 	if (!(md->pde = create_proc_entry(md->name, S_IRUGO | S_IWUSR,
 					_proc_dir))) {
@@ -148,6 +153,11 @@ int dm_fs_add(struct mapped_device *md)
 int dm_fs_remove(struct mapped_device *md)
 {
 	if (md->pde) {
+		struct pf_data *pfd = (struct pf_data *) md->pde->data;
+
+		if (pfd->map)
+			dm_table_destroy(pfd->map);
+
 		kfree(md->pde->data);
 		remove_proc_entry(md->name, _proc_dir);
 		md->pde = 0;
@@ -158,7 +168,8 @@ int dm_fs_remove(struct mapped_device *md)
 	return 0;
 }
 
-static int process_control(const char *b, const char *e, int minor)
+static int process_control(const char *b, const char *e, int minor,
+			   struct dm_table **map)
 {
 	const char *wb, *we;
 	char name[64];
@@ -201,10 +212,12 @@ static int process_control(const char *b, const char *e, int minor)
 	return -EINVAL;
 }
 
-static int process_table(const char *b, const char *e, int minor)
+static int process_table(const char *b, const char *e, int minor,
+			 struct dm_table **map)
 {
 	const char *wb, *we;
 	struct mapped_device *md = dm_find_by_minor(minor);
+	struct dm_table *table = *map;
 	void *context;
 	int r;
 
@@ -219,12 +232,15 @@ static int process_table(const char *b, const char *e, int minor)
 		dm_suspend(md);
 
 		/* start loading a table */
-		dm_table_start(md);
+		table = *map = dm_table_create();
 
 	} else if (!tok_cmp("end", b, e)) {
 		/* activate the device ... <evil chuckle> ... */
-		dm_table_complete(md);
+		if (table) {
+			dm_table_complete(table);
+			dm_bind(md, table);
 			dm_activate(md);
+		}
 
 	} else {
 		/* add the new entry */
@@ -253,18 +269,18 @@ static int process_table(const char *b, const char *e, int minor)
 			return -EINVAL;
 
 		/* check there isn't a gap */
-		if ((md->num_targets &&
+		if ((table->num_targets &&
-		     start != md->highs[md->num_targets - 1] + 1) ||
+		     start != table->highs[table->num_targets - 1] + 1) ||
-		    (!md->num_targets && start)) {
+		    (!table->num_targets && start)) {
 			WARN("gap in target ranges");
 			return -EINVAL;
 		}
 
 		high = start + (size - 1);
-		if ((r = t->ctr(start, high, md, we, e, &context)))
+		if ((r = t->ctr(start, high, table, we, e, &context)))
 			return r;
 
-		if ((r = dm_table_add_entry(md, high, t->map, context)))
+		if ((r = dm_table_add_entry(table, high, t->map, context)))
 			return r;
 	}
 
@@ -302,7 +318,7 @@ static int line_splitter(struct file *file, const char *buffer,
 		while((b != e) && *b != '\n')
 			b++;
 
-		if ((r = pfd->fn(lb, b, pfd->minor)))
+		if ((r = pfd->fn(lb, b, pfd->minor, &pfd->map)))
 			return r;
 	}
 

driver/device-mapper/dm-table.c

@@ -41,7 +41,7 @@ static inline ulong div_up(ulong n, ulong size)
 }
 
 /* ceiling(log_size(n)) */
-static uint int_log(ulong base, ulong n)
+static uint int_log(ulong n, ulong base)
 {
 	int result = 0;
 
@@ -57,14 +57,14 @@ static uint int_log(ulong base, ulong n)
  * return the highest key that you could lookup
  * from the n'th node on level l of the btree.
  */
-static offset_t high(struct mapped_device *md, int l, int n)
+static offset_t high(struct dm_table *t, int l, int n)
 {
 	while (1) {
-		if (n >= md->counts[l])
+		if (n >= t->counts[l])
 			return (offset_t) -1;
 
-		if (l == md->depth - 1)
+		if (l == t->depth - 1)
-			return md->index[l][((n + 1) * KEYS_PER_NODE) - 1];
+			return t->index[l][((n + 1) * KEYS_PER_NODE) - 1];
 
 		l++;
 		n = (n + 1) * (KEYS_PER_NODE + 1) - 1;
@@ -77,15 +77,15 @@ static offset_t high(struct mapped_device *md, int l, int n)
  * fills in a level of the btree based on the
  * highs of the level below it.
  */
-static int setup_btree_index(int l, struct mapped_device *md)
+static int setup_btree_index(int l, struct dm_table *t)
 {
 	int n, c, cn;
 
-	for (n = 0, cn = 0; n < md->counts[l]; n++) {
+	for (n = 0, cn = 0; n < t->counts[l]; n++) {
-		offset_t *k = md->index[l] + (n * KEYS_PER_NODE);
+		offset_t *k = t->index[l] + (n * KEYS_PER_NODE);
 
 		for (c = 0; c < KEYS_PER_NODE; c++)
-			k[c] = high(md, l + 1, cn++);
+			k[c] = high(t, l + 1, cn++);
 		cn++;		/* one extra for the child that's
                                    greater than all keys */
 	}
@@ -93,31 +93,15 @@ static int setup_btree_index(int l, struct mapped_device *md)
 	return 0;
 }
 
-void dm_table_free(struct mapped_device *md)
-{
-	int i;
-	for (i = 0; i < md->depth; i++) {
-		vfree(md->index[i]);
-		md->index[i] = 0;
-	}
-
-	vfree(md->targets);
-
-	md->highs = 0;
-	md->targets = 0;
-
-	md->num_targets = 0;
-	md->num_allocated = 0;
-}
-
 /*
- * md->highs, and md->targets are managed as
+ * highs, and targets are managed as dynamic
- * dynamic arrays during a table load.
+ * arrays during a table load.
  */
-static int alloc_targets(struct mapped_device *md, int num)
+static int alloc_targets(struct dm_table *t, int num)
 {
 	offset_t *n_highs;
 	struct target *n_targets;
+	int n = t->num_targets;
 
 	if (!(n_highs = vmalloc(sizeof(*n_highs) * num)))
 		return -ENOMEM;
@@ -127,93 +111,143 @@ static int alloc_targets(struct mapped_device *md, int num)
 		return -ENOMEM;
 	}
 
-	if (md->num_targets) {
+	if (n) {
-		memcpy(n_highs, md->highs,
+		memcpy(n_highs, t->highs, sizeof(*n_highs) * n);
-		       sizeof(*n_highs) * md->num_targets);
+		memcpy(n_targets, t->targets, sizeof(*n_targets) * n);
-
-		memcpy(n_targets, md->targets,
-		       sizeof(*n_targets) * md->num_targets);
 	}
 
-	vfree(md->highs);
+	vfree(t->highs);
-	vfree(md->targets);
+	vfree(t->targets);
 
-	md->num_allocated = num;
+	t->num_allocated = num;
-	md->highs = n_highs;
+	t->highs = n_highs;
-	md->targets = n_targets;
+	t->targets = n_targets;
 
 	return 0;
 }
 
-int dm_table_start(struct mapped_device *md)
+struct dm_table *dm_table_create(void)
 {
-	int r;
+	struct dm_table *t = kmalloc(sizeof(struct dm_table), GFP_NOIO);
-	set_bit(DM_LOADING, &md->state);
 
-	dm_table_free(md);
+	if (!t)
+		return 0;
 
-	/* allocate a single nodes worth to start with */
+	memset(t, 0, sizeof(*t));
-	if ((r = alloc_targets(md, KEYS_PER_NODE)))
+
+	/* allocate a single nodes worth of targets to
+	   begin with */
+	if (t && alloc_targets(t, KEYS_PER_NODE)) {
+		kfree(t);
+		t = 0;
+	}
+
+	return t;
+}
+
+void dm_table_destroy(struct dm_table *t)
+{
+	struct dev_list *d, *n;
+	int i;
+
+	if (!t)
+		return;
+
+	/* free the indexes */
+	for (i = 0; i < t->depth; i++) {
+		vfree(t->index[i]);
+		t->index[i] = 0;
+	}
+
+	/* t->highs was already freed as t->index[t->depth - 1] */
+	vfree(t->targets);
+	kfree(t);
+
+	/* free the device list */
+	for (d = t->devices; d; d = n) {
+		n = d->next;
+		kfree(d);
+	}
+}
+
+/*
+ * checks to see if we need to extend highs or targets
+ */
+static inline int check_space(struct dm_table *t)
+{
+	if (t->num_targets >= t->num_allocated)
+		return alloc_targets(t, t->num_allocated * 2);
+
+	return 0;
+}
+
+/*
+ * adds a target to the map
+ */
+int dm_table_add_entry(struct dm_table *t, offset_t high,
+		       dm_map_fn target, void *private)
+{
+	int r, n;
+
+	if ((r = check_space(t)))
 		return r;
 
+	n = t->num_targets++;
+	t->highs[n] = high;
+	t->targets[n].map = target;
+	t->targets[n].private = private;
+
 	return 0;
 }
 
-static inline int check_space(struct mapped_device *md)
+int dm_table_add_device(struct dm_table *t, kdev_t dev)
 {
-	if (md->num_targets >= md->num_allocated)
+	struct dev_list *d = kmalloc(sizeof(*d), GFP_KERNEL);
-		return alloc_targets(md, md->num_allocated * 2);
+
+	if (!d)
+		return -ENOMEM;
+
+	d->dev = dev;
+	d->next = t->devices;
+	t->devices = d;
 
 	return 0;
 }
 
-int dm_table_add_entry(struct mapped_device *md, offset_t high,
+/*
-		       dm_map_fn target, void *context)
+ * builds the btree to index the map
-{
+ */
-	int r;
+int dm_table_complete(struct dm_table *t)
-
-	if ((r = check_space(md)))
-		return r;
-
-	md->highs[md->num_targets] = high;
-	md->targets[md->num_targets].map = target;
-	md->targets[md->num_targets].private = context;
-	md->num_targets++;
-
-	return 0;
-}
-
-int dm_table_complete(struct mapped_device *md)
 {
 	int i, leaf_nodes;
 
-	clear_bit(DM_LOADING, &md->state);
-
 	/* how many indexes will the btree have ? */
-	leaf_nodes = div_up(md->num_targets, KEYS_PER_NODE);
+	leaf_nodes = div_up(t->num_targets, KEYS_PER_NODE);
 	i = 1 + int_log(leaf_nodes, KEYS_PER_NODE + 1);
 
-	md->depth = i;
+	/* work out how many nodes are in each layer */
-	md->counts[md->depth - 1] = div_up(md->num_targets, KEYS_PER_NODE);
+	t->depth = i;
+	t->counts[t->depth - 1] = div_up(t->num_targets, KEYS_PER_NODE);
 
 	while (--i)
-		md->counts[i - 1] = div_up(md->counts[i], KEYS_PER_NODE + 1);
+		t->counts[i - 1] = div_up(t->counts[i], KEYS_PER_NODE + 1);
 
-	for (i = 0; i < (md->depth - 1); i++) {
+	/* allocate memory for the internal nodes */
-		size_t s = NODE_SIZE * md->counts[i];
+	for (i = 0; i < (t->depth - 1); i++) {
-		md->index[i] = vmalloc(s);
+		size_t s = NODE_SIZE * t->counts[i];
-		memset(md->index[i], -1, s);
+		t->index[i] = vmalloc(s);
+		memset(t->index[i], -1, s);
 	}
 
-	/* bottom layer is easy */
+	/* leaf layer has already been set up */
-	md->index[md->depth - 1] = md->highs;
+	t->index[t->depth - 1] = t->highs;
 
 	/* fill in higher levels */
-	for (i = md->depth - 1; i; i--)
+	for (i = t->depth - 1; i; i--)
-		setup_btree_index(i - 1, md);
+		setup_btree_index(i - 1, t);
 
-	set_bit(DM_LOADED, &md->state);
 	return 0;
 }
 
 
+EXPORT_SYMBOL(dm_table_add_device);

driver/device-mapper/dm-target.c

@@ -47,6 +47,9 @@ struct target_type *dm_get_target(const char *name)
 	return t;
 }
 
+/*
+ * register a new target_type.
+ */
 int register_map_target(const char *name, dm_ctr_fn ctr,
 			dm_dtr_fn dtr, dm_map_fn map)
 {
@@ -82,7 +85,7 @@ int register_map_target(const char *name, dm_ctr_fn ctr,
  * io-err: always fails an io, useful for bringing
  * up LV's that have holes in them.
  */
-static int io_err_ctr(offset_t b, offset_t e, struct mapped_device *md,
+static int io_err_ctr(offset_t b, offset_t e, struct dm_table *t,
 		      const char *cb, const char *ce, void **result)
 {
 	/* this takes no arguments */
@@ -109,11 +112,13 @@ struct linear_c {
 	int offset;		/* FIXME: we need a signed offset type */
 };
 
-static int linear_ctr(offset_t low, offset_t high, struct mapped_device *md,
+/*
+ * construct a linear mapping.
+ * <major> <minor> <offset>
+ */
+static int linear_ctr(offset_t low, offset_t high, struct dm_table *t,
 		      const char *cb, const char *ce, void **result)
 {
-	/* <major> <minor> <offset> */
-
 	struct linear_c *lc;
 	unsigned int major, minor, start;
 	int r;
@@ -135,7 +140,7 @@ static int linear_ctr(offset_t low, offset_t high, struct mapped_device *md,
 	lc->dev = MKDEV((int) major, (int) minor);
 	lc->offset = (int) start - (int) low;
 
-	if ((r = dm_add_device(md, lc->dev))) {
+	if ((r = dm_table_add_device(t, lc->dev))) {
 		kfree(lc);
 		return r;
 	}

driver/device-mapper/dm.c

@@ -63,23 +63,12 @@ static int _block_size[MAX_DEVICES];
 static int _blksize_size[MAX_DEVICES];
 static int _hardsect_size[MAX_DEVICES];
 
-static int blk_open(struct inode *inode, struct file *file);
-static int blk_close(struct inode *inode, struct file *file);
-static int blk_ioctl(struct inode *inode, struct file *file,
-		     uint command, ulong a);
-
-struct block_device_operations dm_blk_dops = {
-	open:     blk_open,
-	release:  blk_close,
-	ioctl:    blk_ioctl
-};
-
 static int request(request_queue_t *q, int rw, struct buffer_head *bh);
 
 /*
  * setup and teardown the driver
  */
-static int init_dm(void)
+static int dm_init(void)
 {
 	int ret;
 
@@ -112,7 +101,7 @@ static int init_dm(void)
 	return 0;
 }
 
-static void exit_dm(void)
+static void dm_exit(void)
 {
 	if(kmem_cache_shrink(_io_hook_cache))
 		WARN("it looks like there are still some io_hooks allocated");
@@ -252,6 +241,11 @@ inline static void free_deferred(struct deferred_io *di)
 	kfree(di);
 }
 
+/*
+ * bh->b_end_io routine that decrements the
+ * pending count and then calls the original
+ * bh->b_end_io fn.
+ */
 static void dec_pending(struct buffer_head *bh, int uptodate)
 {
 	struct io_hook *ih = bh->b_private;
@@ -267,6 +261,9 @@ static void dec_pending(struct buffer_head *bh, int uptodate)
 	bh->b_end_io(bh, uptodate);
 }
 
+/*
+ * add the bh to the list of deferred io.
+ */
 static int queue_io(struct mapped_device *md, struct buffer_head *bh, int rw)
 {
 	struct deferred_io *di = alloc_deferred();
@@ -289,15 +286,17 @@ static int queue_io(struct mapped_device *md, struct buffer_head *bh, int rw)
 	return 1;
 }
 
-
+/*
+ * do the bh mapping for a given leaf
+ */
 inline static int __map_buffer(struct mapped_device *md,
-			       struct buffer_head *bh, int node)
+			       struct buffer_head *bh, int leaf)
 {
 	dm_map_fn fn;
 	void *context;
 	struct io_hook *ih = 0;
 	int r;
-	struct target *ti = md->targets + node;
+	struct target *ti = md->map->targets + leaf;
 
 	fn = ti->map;
 	context = ti->private;
@@ -334,15 +333,17 @@ inline static int __map_buffer(struct mapped_device *md,
 	return 1;
 }
 
-inline static int __find_node(struct mapped_device *md, struct buffer_head *bh)
+/*
+ * search the btree for the correct target.
+ */
+inline static int __find_node(struct dm_table *t, struct buffer_head *bh)
 {
 	int i = 0, l, r = 0;
 	offset_t *node;
 
-	/* search the btree for the correct target */
+	for (l = 0; l < t->depth; l++) {
-	for (l = 0; l < md->depth; l++) {
 		r = ((KEYS_PER_NODE + 1) * r) + i;
-		node = md->index[l] + (r * KEYS_PER_NODE);
+		node = t->index[l] + (r * KEYS_PER_NODE);
 
 		for (i = 0; i < KEYS_PER_NODE; i++)
 			if (node[i] >= bh->b_rsector)
@@ -363,7 +364,7 @@ static int request(request_queue_t *q, int rw, struct buffer_head *bh)
 	rl;
 	md = _devs[minor];
 
-	if (!md || !test_bit(DM_LOADED, &md->state))
+	if (!md || !md->map)
 		goto bad;
 
 	/* if we're suspended we have to queue this io for later */
@@ -380,7 +381,7 @@ static int request(request_queue_t *q, int rw, struct buffer_head *bh)
 		rl;	/* FIXME: there's still a race here */
 	}
 
-	if (!__map_buffer(md, bh, __find_node(md, bh)))
+	if (!__map_buffer(md, bh, __find_node(md->map, bh)))
 		goto bad;
 
 	ru;
@@ -394,6 +395,10 @@ static int request(request_queue_t *q, int rw, struct buffer_head *bh)
 	return 0;
 }
 
+/*
+ * see if the device with a specific minor # is
+ * free.
+ */
 static inline int __specific_dev(int minor)
 {
 	if (minor > MAX_DEVICES) {
@@ -407,6 +412,9 @@ static inline int __specific_dev(int minor)
 	return -1;
 }
 
+/*
+ * find the first free device.
+ */
 static inline int __any_old_dev(void)
 {
 	int i;
@@ -418,6 +426,9 @@ static inline int __any_old_dev(void)
 	return -1;
 }
 
+/*
+ * allocate and initialise a blank device.
+ */
 static struct mapped_device *alloc_dev(int minor)
 {
 	struct mapped_device *md = kmalloc(sizeof(*md), GFP_KERNEL);
@@ -445,16 +456,10 @@ static struct mapped_device *alloc_dev(int minor)
 	return md;
 }
 
-static inline struct mapped_device *__find_by_name(const char *name)
+/*
-{
+ * open a device so we can use it as a map
-	int i;
+ * destination.
-	for (i = 0; i < MAX_DEVICES; i++)
+ */
-		if (_devs[i] && !strcmp(_devs[i]->name, name))
-			return _devs[i];
-
-	return 0;
-}
-
 static int open_dev(struct dev_list *d)
 {
 	int err;
@@ -470,6 +475,9 @@ static int open_dev(struct dev_list *d)
 	return 0;
 }
 
+/*
+ * close a device that we've been using.
+ */
 static void close_dev(struct dev_list *d)
 {
 	blkdev_put(d->bd, BDEV_FILE);
@@ -477,18 +485,14 @@ static void close_dev(struct dev_list *d)
 	d->bd = 0;
 }
 
-static int __find_hardsect_size(struct mapped_device *md)
+static inline struct mapped_device *__find_by_name(const char *name)
 {
-	int r = INT_MAX, s;
+	int i;
-	struct dev_list *dl;
+	for (i = 0; i < MAX_DEVICES; i++)
+		if (_devs[i] && !strcmp(_devs[i]->name, name))
+			return _devs[i];
 
-	for (dl = md->devices; dl; dl = dl->next) {
+	return 0;
-		s = get_hardsect_size(dl->dev);
-		if (s < r)
-			r = s;
-	}
-
-	return r;
 }
 
 struct mapped_device *dm_find_by_name(const char *name)
@@ -513,6 +517,9 @@ struct mapped_device *dm_find_by_minor(int minor)
 	return md;
 }
 
+/*
+ * constructor for a new device
+ */
 int dm_create(const char *name, int minor)
 {
 	int r;
@@ -544,10 +551,14 @@ int dm_create(const char *name, int minor)
 	return 0;
 }
 
+/*
+ * destructor for the device.  md->map is
+ * deliberately not destroyed, dm-fs should manage
+ * table objects.
+ */
 int dm_remove(const char *name)
 {
 	struct mapped_device *md;
-	struct dev_list *d, *n;
 	int minor, r;
 
 	wl;
@@ -566,12 +577,6 @@ int dm_remove(const char *name)
 		return r;
 	}
 
-	dm_table_free(md);
-	for (d = md->devices; d; d = n) {
-		n = d->next;
-		kfree(d);
-	}
-
 	minor = MINOR(md->dev);
 	kfree(md);
 	_devs[minor] = 0;
@@ -580,20 +585,57 @@ int dm_remove(const char *name)
 	return 0;
 }
 
-int dm_add_device(struct mapped_device *md, kdev_t dev)
+/*
+ * the hardsect size for a mapped device is the
+ * smallest hard sect size from the devices it
+ * maps onto.
+ */
+static int __find_hardsect_size(struct dev_list *dl)
 {
-	struct dev_list *d = kmalloc(sizeof(*d), GFP_KERNEL);
+	int result = INT_MAX, size;
 
-	if (!d)
+	while(dl) {
-		return -EINVAL;
+		size = get_hardsect_size(dl->dev);
+		if (size < result)
+			result = size;
+		dl = dl->next;
+	}
 
-	d->dev = dev;
+	return result;
-	d->next = md->devices;
+}
-	md->devices = d;
+
+/*
+ * bind a table to the device, the device must not
+ * be active, though it could have another table
+ * aready bound.
+ */
+int dm_bind(struct mapped_device *md, struct dm_table *t)
+{
+	int minor = MINOR(md->dev);
+
+	wl;
+	if (is_active(md)) {
+		wu;
+		return -EPERM;
+	}
+
+	md->map = t;
+
+	_block_size[minor] = (t->highs[t->num_targets - 1] + 1) >> 1;
+
+	/* FIXME: block size depends on the mapping table */
+	_blksize_size[minor] = BLOCK_SIZE;
+	_hardsect_size[minor] = __find_hardsect_size(t->devices);
+	register_disk(NULL, md->dev, 1, &dm_blk_dops, _block_size[minor]);
+	wu;
 
 	return 0;
 }
 
+/*
+ * requeue the deferred buffer_heads by calling
+ * generic_make_request.
+ */
 static void __flush_deferred_io(struct mapped_device *md)
 {
 	struct deferred_io *c, *n;
@@ -605,9 +647,14 @@ static void __flush_deferred_io(struct mapped_device *md)
 	}
 }
 
+/*
+ * make the device available for use, if was
+ * previously suspended rather than newly created
+ * then all queued io is flushed
+ */
 int dm_activate(struct mapped_device *md)
 {
-	int ret, minor;
+	int ret;
 	struct dev_list *d, *od;
 
 	wl;
@@ -617,27 +664,17 @@ int dm_activate(struct mapped_device *md)
 		return 0;
 	}
 
-	if (!md->num_targets) {
+	if (!md->map) {
 		wu;
 		return -ENXIO;
 	}
 
 	/* open all the devices */
-	for (d = md->devices; d; d = d->next)
+	for (d = md->map->devices; d; d = d->next)
 		if ((ret = open_dev(d)))
 			goto bad;
 
-	minor = MINOR(md->dev);
-
-	_block_size[minor] = (md->highs[md->num_targets - 1] + 1) >> 1;
-	_blksize_size[minor] = BLOCK_SIZE; /* FIXME: this depends on
-                                              the mapping table */
-	_hardsect_size[minor] = __find_hardsect_size(md);
-
-	register_disk(NULL, md->dev, 1, &dm_blk_dops, _block_size[minor]);
-
 	set_bit(DM_ACTIVE, &md->state);
-
 	__flush_deferred_io(md);
 	wu;
 
@@ -645,17 +682,27 @@ int dm_activate(struct mapped_device *md)
 
  bad:
 	od = d;
-	for (d = md->devices; d != od; d = d->next)
+	for (d = md->map->devices; d != od; d = d->next)
 		close_dev(d);
 	ru;
 
 	return ret;
 }
 
+/*
+ * we need to be able to change a mapping table
+ * under a mounted filesystem.  for example we
+ * might want to move some data in the background.
+ * Before the table can be swapped with
+ * dm_bind_table, dm_suspend must be called to
+ * flush any in flight buffer_heads and ensure
+ * that any further io gets deferred.
+ */
 void dm_suspend(struct mapped_device *md)
 {
 	DECLARE_WAITQUEUE(wait, current);
 	struct dev_list *d;
+
 	if (!is_active(md))
 		return;
 
@@ -676,7 +723,7 @@ void dm_suspend(struct mapped_device *md)
 	remove_wait_queue(&md->wait, &wait);
 
 	/* close all the devices */
-	for (d = md->devices; d; d = d->next)
+	for (d = md->map->devices; d; d = d->next)
 		close_dev(d);
 
 	clear_bit(DM_ACTIVE, &md->state);
@@ -684,11 +731,17 @@ void dm_suspend(struct mapped_device *md)
 }
 
 
+struct block_device_operations dm_blk_dops = {
+	open:     blk_open,
+	release:  blk_close,
+	ioctl:    blk_ioctl
+};
+
 /*
  * module hooks
  */
-module_init(init_dm);
+module_init(dm_init);
-module_exit(exit_dm);
+module_exit(dm_exit);
 
 /*
  * Local variables:

driver/device-mapper/dm.h

@@ -137,9 +137,8 @@
 #define DM_NAME_LEN 64
 
 enum {
-	DM_LOADED = 0,
+	DM_BOUND = 0,		/* device has been bound to a table */
-	DM_LOADING,
+	DM_ACTIVE,		/* device is running */
-	DM_ACTIVE,
 };
 
 /*
@@ -182,6 +181,24 @@ struct target {
 	void *private;
 };
 
+/*
+ * the btree
+ */
+struct dm_table {
+	/* btree table */
+	int depth;
+	int counts[MAX_DEPTH];	/* in nodes */
+	offset_t *index[MAX_DEPTH];
+
+	int num_targets;
+	int num_allocated;
+	offset_t *highs;
+	struct target *targets;
+
+	/* a list of devices used by this table */
+	struct dev_list *devices;
+};
+
 /*
  * the actual device struct
  */
@@ -198,46 +215,40 @@ struct mapped_device {
 	/* a list of io's that arrived while we were suspended */
 	struct deferred_io *deferred;
 
-	/* btree table */
+	struct dm_table *map;
-	int depth;
-	int counts[MAX_DEPTH];	/* in nodes */
-	offset_t *index[MAX_DEPTH];
-
-	int num_targets;
-	int num_allocated;
-	offset_t *highs;
-	struct target *targets;
 
 	/* used by dm-fs.c */
 	devfs_handle_t devfs_entry;
 	struct proc_dir_entry *pde;
-
-	/* a list of devices used by this md */
-	struct dev_list *devices;
 };
 
 extern struct block_device_operations dm_blk_dops;
 
+
 /* dm-target.c */
 int dm_target_init(void);
 struct target_type *dm_get_target(const char *name);
 
+
 /* dm.c */
 struct mapped_device *dm_find_by_name(const char *name);
 struct mapped_device *dm_find_by_minor(int minor);
 
 int dm_create(const char *name, int minor);
 int dm_remove(const char *name);
-
+int dm_bind(struct mapped_device *md, struct dm_table *t);
 int dm_activate(struct mapped_device *md);
 void dm_suspend(struct mapped_device *md);
 
+
 /* dm-table.c */
-int dm_table_start(struct mapped_device *md);
+struct dm_table *dm_table_create(void);
-int dm_table_add_entry(struct mapped_device *md, offset_t high,
+void dm_table_destroy(struct dm_table *t);
-		       dm_map_fn target, void *context);
+
-int dm_table_complete(struct mapped_device *md);
+int dm_table_add_entry(struct dm_table *t, offset_t high,
-void dm_table_free(struct mapped_device *md);
+		       dm_map_fn target, void *private);
+int dm_table_complete(struct dm_table *t);
+
 
 /* dm-fs.c */
 int dm_fs_init(void);