ddsetup-2.6.23.12 - zumastor - Project Hosting on Google Code

r1735

Source path: svn/ www/ ddsetup/ ddsetup-2.6.23.12

--- 2.6.23.12.base/block/ll_rw_blk.c	2008-02-14 02:01:41.000000000 -0800
+++ 2.6.23.12/block/ll_rw_blk.c	2008-02-13 02:13:13.000000000 -0800
@@ -606,7 +606,7 @@ void blk_queue_bounce_limit(struct reque
 		dma = 1;
 	q->bounce_pfn = bounce_pfn;
 #endif
-	if (dma) {
+	if (0 && dma) { // !!! level violation !!! (generic block layer should not know about isa)
 		init_emergency_isa_pool();
 		q->bounce_gfp = GFP_NOIO | GFP_DMA;
 		q->bounce_pfn = bounce_pfn;
--- 2.6.23.12.base/Documentation/ioctl-number.txt	2007-12-18 13:55:57.000000000 -0800
+++ 2.6.23.12/Documentation/ioctl-number.txt	2008-02-17 00:07:52.000000000 -0800
@@ -65,6 +65,7 @@ Code	Seq#	Include File		Comments
 0x00	00-1F	scsi/scsi_ioctl.h	conflict!
 0x00	00-1F	linux/fb.h		conflict!
 0x00	00-1F	linux/wavefront.h	conflict!
+0x00	DD-DD   linux/ddlink.h		Device driver link
 0x02	all	linux/fd.h
 0x03	all	linux/hdreg.h
 0x04	D2-DC	linux/umsdos_fs.h	Dead since 2.6.11, but don't reuse these.
--- 2.6.23.12.base/drivers/md/ddlink.c	2008-02-14 02:01:41.000000000 -0800
+++ 2.6.23.12/drivers/md/ddlink.c	2008-02-17 00:11:48.000000000 -0800
@@ -0,0 +1,294 @@
+#include <linux/file.h>
+#include <linux/poll.h>
+#include <linux/mount.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/ddlink.h>
+
+/*
+ * ddlink: Device Driver link.  A userspace to kernel device control
+ * interface useful for virtual block devices and other fancy things.
+ * (c) 2007 Daniel Phillips <daniel@phlyps.net>
+ */
+
+// To do:
+//  * need spinlocks on push, pop, queue, ready and ???
+
+int ddlink_ready(struct ddinode *dd)
+{
+	return !list_empty(&dd->list);
+}
+EXPORT_SYMBOL_GPL(ddlink_ready);
+
+struct dditem *ddlink_pop(struct ddinode *dd)
+{
+	struct dditem *item;
+	BUG_ON(!ddlink_ready(dd));
+	item = list_entry(dd->list.next, struct dditem, link);
+	list_del(&item->link);
+	return item;
+}
+EXPORT_SYMBOL_GPL(ddlink_pop);
+
+void ddlink_clear(struct ddinode *dd)
+{
+	while (ddlink_ready(dd))
+		dd->destroy_item(ddlink_pop(dd));
+}
+EXPORT_SYMBOL_GPL(ddlink_clear);
+
+static struct inode *ddlink_alloc_inode(struct super_block *sb)
+{
+	struct ddinode *dd = kmalloc(sizeof(struct ddinode), GFP_KERNEL);
+	if (!dd)
+		return NULL;
+	*dd = (typeof(*dd)){
+		.wait = __WAIT_QUEUE_HEAD_INITIALIZER(dd->wait),
+		.list = LIST_HEAD_INIT(dd->list) };
+	inode_init_once(&dd->inode);
+	return &dd->inode;
+}
+
+static void ddlink_free_inode(struct inode *inode)
+{
+	struct ddinode *dd = ddinode(inode);
+	ddlink_clear(dd);
+	if (ddinfo(inode))
+		dd->destroy_info(ddinfo(inode));
+	kfree(ddinode(inode));
+}
+
+static struct super_operations ddlink_sops = {
+	.alloc_inode = ddlink_alloc_inode,
+	.destroy_inode = ddlink_free_inode
+};
+
+static struct vfsmount *ddlink_mnt;
+
+static int ddlink_get_sb(struct file_system_type *fs_type, int flags, const char *dev_name, void *data, struct vfsmount *mnt)
+{
+	return get_sb_pseudo(fs_type, "ddlink", &ddlink_sops, 0x20070318, mnt);
+}
+
+static struct file_system_type ddlink_type = {
+	.name = "ddlink",
+	.get_sb = ddlink_get_sb,
+	.kill_sb = kill_anon_super };
+
+static int ddlink_delete_dentry(struct dentry *dentry) { return 1; } // vfs bogon
+static struct dentry_operations ddlink_dops = { .d_delete = ddlink_delete_dentry };
+
+int ddlink(struct file_operations *fops, void *(*create)(struct ddinode *dd, void *info), void *info)
+{
+	int err = -ENFILE, fd;
+	struct file *file = get_empty_filp();
+	struct inode *inode;
+	struct dentry *dentry;
+	void *private;
+
+	if (!file)
+		goto no_file;
+	inode = new_inode(ddlink_mnt->mnt_sb);
+	if (!inode)
+		goto no_inode;
+	ddinode(inode)->destroy_info = kfree;
+	ddinode(inode)->create_item = kmalloc;
+	ddinode(inode)->destroy_item = kfree;
+	if (IS_ERR(private = create(ddinode(inode), info))) {
+		err = PTR_ERR(private);
+		goto no_inode;
+	}
+	inode->i_private = private;
+
+	dentry = d_alloc(ddlink_mnt->mnt_sb->s_root, &(struct qstr){ });
+	if (!dentry)
+		goto no_dentry;
+	if ((fd = err = get_unused_fd()) < 0)
+		goto no_fd;
+	/*
+	 * Mark the inode dirty from the very beginning,
+	 * that way it will never be moved to the dirty
+	 * list because "mark_inode_dirty()" will think
+	 * that it already _is_ on the dirty list (gag)
+	 */
+	inode->i_state = I_DIRTY;
+	inode->i_mode = S_IFCHR|S_IRWXUGO;
+	inode->i_uid = current->fsuid;
+	inode->i_gid = current->fsgid;
+	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+	inode->i_fop = fops;
+	dentry->d_op = &ddlink_dops;
+	d_instantiate(dentry, inode); // d_add?
+
+	file->f_dentry = dentry;
+	file->f_vfsmnt = mntget(ddlink_mnt);
+	file->f_mapping = inode->i_mapping;
+	file->f_mode = FMODE_READ|FMODE_WRITE;
+//	file->f_flags = O_RDWR; // what is this for?
+	file->f_op = inode->i_fop;
+	fd_install(fd, file);
+	return fd;
+no_fd:
+	dput(dentry);
+no_dentry:
+	iput(inode);
+no_inode:
+	put_filp(file);
+no_file:
+	return err;
+}
+EXPORT_SYMBOL_GPL(ddlink);
+
+static int __init ddlink_init(void)
+{
+	struct vfsmount *mnt;
+	register_filesystem(&ddlink_type);
+	mnt = kern_mount(&ddlink_type);
+	if (IS_ERR(mnt))
+		return PTR_ERR(mnt);
+	ddlink_mnt = mnt;
+	return 0;
+}
+
+static void __exit ddlink_exit(void)
+{
+	mntput(ddlink_mnt);
+	unregister_filesystem(&ddlink_type);
+}
+
+module_init(ddlink_init);
+module_exit(ddlink_exit);
+
+/* library methods */
+
+void ddlink_push(struct ddinode *dd, struct dditem *item)
+{
+	list_add(&item->link, &dd->list);
+	wake_up_interruptible(&dd->wait);
+}
+EXPORT_SYMBOL_GPL(ddlink_push);
+
+void ddlink_queue(struct ddinode *dd, struct dditem *item)
+{
+	list_add_tail(&item->link, &dd->list);
+	wake_up_interruptible(&dd->wait);
+}
+EXPORT_SYMBOL_GPL(ddlink_queue);
+
+struct dditem *dditem_new(struct ddinode *dd, size_t size)
+{
+	struct dditem *item;
+	if (!(item = dd->create_item(sizeof(*item) + size, __GFP_NOFAIL))) // NOFAIL??
+		return NULL;
+	*item = (typeof(*item)){ .size = size };
+	return item;
+}
+EXPORT_SYMBOL_GPL(dditem_new);
+
+struct dditem *dditem_in(struct ddinode *dd, const void *buf, size_t len, int z)
+{
+	struct dditem *item = dditem_new(dd, len + z);
+
+	if (!item)
+		return ERR_PTR(-ENOMEM);
+	if (copy_from_user(item->data, buf, len)) {
+		dd->destroy_item(item);
+		return ERR_PTR(-EFAULT);
+	}
+	if (z) {
+		item->size -= z;
+		item->data[len] = 0;
+	}
+	return item;
+}
+EXPORT_SYMBOL_GPL(dditem_in);
+
+int dditem_out(struct ddinode *dd, void *buf, size_t len, struct dditem *item) // !!! not tested
+{
+	if (len < item->size)
+		return -EINVAL;
+	if (copy_to_user(buf, item->data, len = item->size))
+		return -EFAULT;
+	dd->destroy_item(item);
+	return len;
+}
+EXPORT_SYMBOL_GPL(dditem_out);
+
+int ddlink_post(struct ddinode *dd, void *data, unsigned len)
+{
+	struct dditem *item = dditem_new(dd, len);
+	if (!item)
+		return -ENOMEM;
+	memcpy(item->data, data, len);
+	ddlink_push(dd, item);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(ddlink_post);
+
+int ddlink_error(struct ddinode *dd, int err, const char *fmt, ...)
+{
+	int size = 200; /* enough for any error? */
+	struct dditem *item, *work;
+	va_list args;
+	if (err >= 0)
+		return err;
+	if (!(work = dditem_new(dd, size)))
+		return -ENOMEM;
+	va_start(args, fmt);
+	size = vsnprintf(work->data, size, fmt, args);
+	va_end(args);
+	if (!(item = dditem_new(dd, size))) {
+		err = -ENOMEM;
+		goto fail;
+	}
+	memcpy(item->data, work->data, size);
+	ddlink_push(dd, item);
+fail:
+	dd->destroy_item(work);
+	return err;
+}
+EXPORT_SYMBOL_GPL(ddlink_error);
+
+unsigned ddlink_poll(struct file *file, poll_table *table)
+{
+	struct ddinode *dd = ddinode(file->f_dentry->d_inode);
+	poll_wait(file, &dd->wait, table);
+	return ddlink_ready(dd) ? POLLIN : 0;
+}
+EXPORT_SYMBOL_GPL(ddlink_poll);
+
+#if 0
+/* ddlink example */
+
+static int ddlink_example_ioctl(struct inode *inode, struct file *file, unsigned cmd, unsigned long ptr)
+{
+	struct ddinode *dd = ddinode(inode);
+	printk(">>> ddlink_ioctl %i (%lx) on %p <<<\n", cmd, ptr, dd);
+	return 0;
+}
+
+static ssize_t ddlink_example_write(struct file *file, const char *buf, size_t len, loff_t *offset)
+{
+	struct ddinode *dd = ddinode(file->f_dentry->d_inode);
+	struct dditem *item;
+
+	if (IS_ERR(item = dditem_in(dd, buf, len, 1)))
+		return PTR_ERR(item);
+	ddlink_queue(dd, item);
+	return len;
+}
+
+static ssize_t ddlink_example_read(struct file *file, char *buf, size_t len, loff_t *offset)
+{
+	struct ddinode *dd = ddinode(file->f_dentry->d_inode);
+	if (!dd->head)
+		return EFAULT;
+	return dditem_out(dd, buf, len, ddlink_pop(dd));
+}
+
+static struct file_operations ddlink_example_fops = {
+	.read = ddlink_example_read,
+	.write = ddlink_example_write,
+	.ioctl = ddlink_example_ioctl,
+	.poll = ddlink_poll };
+#endif
--- 2.6.23.12.base/drivers/md/dm.c	2008-02-14 02:01:41.000000000 -0800
+++ 2.6.23.12/drivers/md/dm.c	2008-02-13 02:13:13.000000000 -0800
@@ -70,59 +70,6 @@ union map_info *dm_get_mapinfo(struct bi
 #define DMF_DELETING 4
 #define DMF_NOFLUSH_SUSPENDING 5
 
-struct mapped_device {
-	struct rw_semaphore io_lock;
-	struct semaphore suspend_lock;
-	spinlock_t pushback_lock;
-	rwlock_t map_lock;
-	atomic_t holders;
-	atomic_t open_count;
-
-	unsigned long flags;
-
-	struct request_queue *queue;
-	struct gendisk *disk;
-	char name[16];
-
-	void *interface_ptr;
-
-	/*
-	 * A list of ios that arrived while we were suspended.
-	 */
-	atomic_t pending;
-	wait_queue_head_t wait;
-	struct bio_list deferred;
-	struct bio_list pushback;
-
-	/*
-	 * The current mapping.
-	 */
-	struct dm_table *map;
-
-	/*
-	 * io objects are allocated from here.
-	 */
-	mempool_t *io_pool;
-	mempool_t *tio_pool;
-
-	struct bio_set *bs;
-
-	/*
-	 * Event handling.
-	 */
-	atomic_t event_nr;
-	wait_queue_head_t eventq;
-
-	/*
-	 * freeze/thaw support require holding onto a super block
-	 */
-	struct super_block *frozen_sb;
-	struct block_device *suspended_bdev;
-
-	/* forced geometry settings */
-	struct hd_geometry geometry;
-};
-
 #define MIN_IOS 256
 static struct kmem_cache *_io_cache;
 static struct kmem_cache *_tio_cache;
--- 2.6.23.12.base/drivers/md/dm.h	2008-02-14 02:01:41.000000000 -0800
+++ 2.6.23.12/drivers/md/dm.h	2008-02-13 02:13:13.000000000 -0800
@@ -95,7 +95,83 @@ struct dm_dev {
 	char name[16];
 };
 
-struct dm_table;
+#include <linux/blkpg.h>
+#include <linux/bio.h>
+#include "dm-bio-list.h"
+#include <linux/buffer_head.h>
+#include <linux/mempool.h>
+#include <linux/slab.h>
+#include <linux/idr.h>
+#include <linux/hdreg.h>
+#include <linux/blktrace_api.h>
+#include <linux/smp_lock.h>
+
+struct mapped_device {
+	struct rw_semaphore io_lock;

+	struct semaphore suspend_lock;
+	spinlock_t pushback_lock;
+	rwlock_t map_lock;
+	atomic_t holders;
+	atomic_t open_count;
+
+	unsigned long flags;
+
+	struct request_queue *queue;
+	struct gendisk *disk;
+	char name[16];
+
+	void *interface_ptr;
+
+	/*
+	 * A list of ios that arrived while we were suspended.
+	 */
+	atomic_t pending;
+	wait_queue_head_t wait;
+	struct bio_list deferred;
+	struct bio_list pushback;
+
+	/*
+	 * The current mapping.
+	 */
+	struct dm_table *map;
+
+	/*
+	 * io objects are allocated from here.
+	 */
+	mempool_t *io_pool;
+	mempool_t *tio_pool;
+
+	struct bio_set *bs;
+
+	/*
+	 * Event handling.
+	 */
+	atomic_t event_nr;
+	wait_queue_head_t eventq;
+
+	/*
+	 * freeze/thaw support require holding onto a super block
+	 */
+	struct super_block *frozen_sb;
+	struct block_device *suspended_bdev;
+
+	/* forced geometry settings */
+	struct hd_geometry geometry;
+};
+
+int adjoin(struct dm_table *table, struct dm_target *ti);
+void combine_restrictions_low(struct io_restrictions *lhs, struct io_restrictions *rhs);
+int alloc_targets(struct dm_table *t, unsigned int num);
+/*
+ * 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;
+}
 
 /*-----------------------------------------------------------------
  * Internal table functions.
--- 2.6.23.12.base/drivers/md/dm-ioctl.c	2008-02-14 02:01:41.000000000 -0800
+++ 2.6.23.12/drivers/md/dm-ioctl.c	2008-02-17 03:05:51.000000000 -0800
@@ -275,11 +275,25 @@ retry:
 	up_write(&_hash_lock);
 }
 
+static int check_name(const char *name)
+{
+	if (strchr(name, '/')) {
+		DMWARN("invalid device name");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 static int dm_hash_rename(const char *old, const char *new)
 {
 	char *new_name, *old_name;
-	struct hash_cell *hc;
+	struct hash_cell *hc, *nh;
 	struct dm_table *table;
+	int err;
+
+	if ((err = check_name(new)))
+		return err;
 
 	/*
 	 * duplicate new.
@@ -291,28 +305,28 @@ static int dm_hash_rename(const char *ol
 	down_write(&_hash_lock);
 
 	/*
-	 * Is new free ?
+	 * Is there such a device as 'old' ?
 	 */
-	hc = __get_name_cell(new);
-	if (hc) {
-		DMWARN("asked to rename to an already existing name %s -> %s",
+	hc = __get_name_cell(old);
+	if (!hc) {
+		DMWARN("asked to rename a non existent device %s -> %s",
 		       old, new);
-		dm_put(hc->md);
 		up_write(&_hash_lock);
 		kfree(new_name);
-		return -EBUSY;
+		return -ENXIO;
 	}
 
 	/*
-	 * Is there such a device as 'old' ?
+	 * Is new free ?
 	 */
-	hc = __get_name_cell(old);
-	if (!hc) {
-		DMWARN("asked to rename a non existent device %s -> %s",
+	nh = __get_name_cell(new);
+	if (nh) {
+		DMWARN("asked to rename to an already existing name %s -> %s",
 		       old, new);
+		dm_put(nh->md);
 		up_write(&_hash_lock);
 		kfree(new_name);
-		return -ENXIO;
+		return -EBUSY;
 	}
 
 	/*
@@ -509,18 +523,6 @@ static int list_versions(struct dm_ioctl
 	return 0;
 }
 
-
-
-static int check_name(const char *name)
-{
-	if (strchr(name, '/')) {
-		DMWARN("invalid device name");
-		return -EINVAL;
-	}
-
-	return 0;
-}
-
 /*
  * Fills in a dm_ioctl structure, ready for sending back to
  * userland.
@@ -697,7 +699,6 @@ static int invalid_str(char *str, void *
 
 static int dev_rename(struct dm_ioctl *param, size_t param_size)
 {
-	int r;
 	char *new_name = (char *) param + param->data_start;
 
 	if (new_name < (char *) (param + 1) ||
@@ -706,10 +707,6 @@ static int dev_rename(struct dm_ioctl *p
 		return -EINVAL;
 	}
 
-	r = check_name(new_name);
-	if (r)
-		return r;
-
 	param->data_size = 0;
 	return dm_hash_rename(param->name, new_name);
 }
@@ -1399,6 +1396,541 @@ static int validate_params(uint cmd, str
 	return 0;
 }
 
+static int get_devnums(struct gendisk *disk, unsigned *major, unsigned *minor)
+{
+	struct block_device *bdev;
+	if (!(bdev = bdget_disk(disk, 0)))
+		return -ENXIO;
+	*major = disk->major;
+	*minor = disk->first_minor;
+	bdput(bdev);
+	return 0;
+}
+
+static void destroy_table(struct dm_table *table)
+{
+	struct mapped_device *md = table->md;
+	/* handle dm object count madness */
+	if (atomic_read(&table->holders) != 1)
+		printk("table %p not destroyed!\n", table);
+	dm_table_put(table);
+	if (atomic_read(&md->holders) != 1)
+		printk("md %p not destroyed!\n", md);
+	dm_put(md); /* does dm_table_put on final */
+}
+
+/* alternative devmapper interface using ddlink */
+
+#include <linux/ddlink.h>
+#include <linux/ddsetup.h>
+#include <linux/err.h>
+
+struct ddmap {
+	unsigned depth, state;
+	struct list_head stack;
+	struct ddinode *ddi;
+	struct dm_table *dmt;
+};
+
+static inline struct ddmap *ddmap(struct inode *inode)
+{
+	return ddinfo(inode);
+}
+
+static struct dditem *dditem_from_list(struct list_head *head)
+{
+	return list_entry(head, struct dditem, link);
+}
+
+static char *ddtop(struct ddmap *ddm)
+{
+	BUG_ON(!ddm->depth);
+	return dditem_from_list(ddm->stack.next)->data;
+}
+
+static int ddmap_drop(struct ddmap *ddm, int n)
+{
+	BUG_ON(n > ddm->depth);
+	while (n--) {
+		struct dditem *item; // seen this code before... (ddlink_pop)
+		BUG_ON(list_empty(&ddm->stack));
+		item = dditem_from_list(ddm->stack.next);
+		list_del(&item->link);
+		ddm->ddi->destroy_item(item);
+		ddm->depth--;
+	}
+	return 0;
+}
+
+static int create(struct ddmap *ddm, int targets, int ddmode)
+{
+	struct dm_table *table;
+	struct mapped_device *md;
+	int err, mode;
+	mode = (ddmode & DMREAD) ? FMODE_READ : 0;
+	mode |= (ddmode & DMWRITE) ? FMODE_WRITE : 0;
+	if (ddm->dmt)
+		return ddlink_error(ddm->ddi, -EINVAL, "table already created");
+	if ((err = dm_create(DM_ANY_MINOR, &md))) // dm_create_with_minor
+		goto fail;
+	if ((err = dm_table_create(&table, mode, targets, md)))
+		goto fail_md;
+	ddm->dmt = table;
+	return 0;
+fail_md:
+	dm_put(md);
+fail:
+	return ddlink_error(ddm->ddi, err, "table create failed");
+}
+
+static int target(struct ddmap *ddm, sector_t offset, sector_t sectors)
+{
+	int err, i, argc = ddm->depth - 1;
+	char *type, **argv, *error = "out of memory";
+	struct dm_table *table = ddm->dmt;
+	struct dm_target *target;
+	struct list_head *next = ddm->stack.next;
+
+	if (!table)
+		return ddlink_error(ddm->ddi, -EINVAL,
+			"create table before target");
+	error = "no space for target";
+	if ((err = check_space(table)))
+		goto fail;
+	err = -EINVAL;
+	error = "zero size target not allowed";
+	if (!sectors)
+		goto fail;
+	err = -ENODATA;
+	error = "target type is missing";
+	if (ddm->depth < 1)
+		goto fail;
+	type = dditem_from_list(next)->data;
+	target = table->targets + table->num_targets;
+	*target = (typeof(*target)){
+		.table = table, .begin = offset, .len = sectors,
+		.type = dm_get_target_type(type), .error = "" };
+	if (!target->type)
+		return ddlink_error(ddm->ddi, -EINVAL,
+			"'%s' is not a target type", type);
+	/* from here on need to release target type on failure */
+
+	error = "gap in table";
+	if (!adjoin(table, target)) // why even have the offset field then???
+		goto fail_type;
+	argv = kmalloc(argc * sizeof(char *), __GFP_NOFAIL);
+	for (i = 0; i < argc; i++)
+		argv[i] = dditem_from_list(next = next->next)->data;
+	err = target->type->ctr(target, argc, argv);
+	error = target->error;
+	kfree(argv);
+	if (err)
+		goto fail_type;
+	table->highs[table->num_targets++] = target->begin + target->len - 1;
+	combine_restrictions_low(&table->limits, &target->limits);
+	return 0;
+fail_type:
+	dm_put_target_type(target->type);
+fail:
+	return ddlink_error(ddm->ddi, err, error);
+}
+
+static int install(struct ddmap *ddm)
+{
+	char *name = ddtop(ddm);
+	struct dm_table *table;
+	struct mapped_device *md;
+	struct ddresult result;
+	unsigned major, minor;
+	char *error;
+	int err;
+
+	if (!(table = ddm->dmt))
+		return ddlink_error(ddm->ddi, -EINVAL,
+			"create table before creating device");
+	md = table->md;
+	ddm->dmt = NULL;
+	error = "table complete failed";
+	if ((err = dm_table_complete(table)))
+		goto fail;
+	error = "invalid device name";
+	if ((err = check_name(name)))
+		goto fail;
+	error = "could not get a device number";
+	if ((err = get_devnums(dm_disk(md), &major, &minor)))
+		goto fail;
+	switch ((err = dm_hash_insert(name, NULL/*uuid*/, md))) {
+	case 0:
+		break;
+	case -EBUSY:
+		error = "device already exists";
+		goto fail;
+	default:
+		error = "hash insert failed";
+		goto fail;
+	}
+	if (!dm_suspended(md)) // wacky table install
+		dm_suspend(md, DM_SUSPEND_LOCKFS_FLAG);
+	error = "could not swap in table";
+	if ((err = dm_swap_table(md, table)))
+		goto fail;
+	set_disk_ro(dm_disk(md), !(dm_table_get_mode(table) & FMODE_WRITE)); // do it before installing!!!
+	dm_table_put(table);
+	if ((err = dm_resume(md)))
+		DMWARN("device %s is installed but not resumed", name);
+	dm_put(md);
+
+	result = (struct ddresult){ .dev.major = major, .dev.minor = minor };
+	error = "failed to post result";
+	if (!(err = ddlink_post(ddm->ddi, &result, sizeof(result))))
+		return 0;
+fail:
+	destroy_table(table);
+	return ddlink_error(ddm->ddi, err, "%s for %s", error, name);
+}
+
+static int remove(struct ddmap *ddm)
+{
+	char *name = ddtop(ddm);
+	struct hash_cell *hc;
+	struct mapped_device *md = NULL;
+	char *error = "%s not defined";
+	int err = -ENXIO;
+
+	down_write(&_hash_lock);
+	if (!(hc = __get_name_cell(name)))
+		goto fail;
+	md = hc->md;
+	err = -EBUSY;
+	error = "%s is still open";
+	if ((err = dm_lock_for_deletion(md)))
+		goto fail;
+	__hash_remove(hc);
+	up_write(&_hash_lock);
+	dm_put(md);
+	return 0;
+fail:
+	up_write(&_hash_lock);
+	if (md)
+		dm_put(md);
+	return ddlink_error(ddm->ddi, err, error, name);
+}
+
+struct mapped_device *lookup_device(char *name)
+{

+	struct mapped_device *md = NULL;
+	struct hash_cell *hc;
+	down_read(&_hash_lock);
+	if ((hc = __get_name_cell(name)))
+		md = hc->md;
+	up_read(&_hash_lock);
+	return md;
+}
+
+static int dependencies(struct ddmap *ddm)
+{
+	char *name = ddtop(ddm);
+	struct mapped_device *md;
+	struct dm_table *table;
+	struct dm_dev *dmd;
+	int err = 0, num = 0;
+
+	if (!(md = lookup_device(name)))
+		return -ENXIO;
+	if (!(table = dm_get_table(md)))
+		goto out;
+	err = -ENOMEM;
+	list_for_each_entry(dmd, dm_table_get_devices(table), list) {
+		dev_t dev = dmd->bdev->bd_dev;
+		struct dditem *item = dditem_new(ddm->ddi, sizeof(struct devnum));
+		if (!item)
+			goto out_table;
+		*(struct devnum *)item->data = (struct devnum){MAJOR(dev), MINOR(dev)};
+		ddlink_queue(ddm->ddi, item);
+		num++;
+	}
+	err = num;
+out_table:
+	dm_table_put(table);
+out:
+	dm_put(md);
+	return err;
+}
+
+static int suspend(struct ddmap *ddm, int suspend, int lockfs, int noflush)
+{
+	char *name = ddtop(ddm);
+	struct hash_cell *hc;
+	struct mapped_device *md;
+	struct dm_table *new_map = NULL;
+	char *error = "%s not found";
+	int err = 0, flags = 0;
+	if (lockfs)
+		flags |= DM_SUSPEND_LOCKFS_FLAG;
+	if (noflush)
+		flags |= DM_SUSPEND_NOFLUSH_FLAG;
+	if (suspend) {
+		if (!(md = lookup_device(name))) {
+			err = -ENXIO;
+			goto out;
+		}
+		error = "could not suspend %s";
+		if (!dm_suspended(md))
+			err = dm_suspend(md, flags);
+	} else {
+		down_write(&_hash_lock);
+		if (!(hc = __get_name_cell(name)) || !(md = hc->md)) {
+			up_write(&_hash_lock);
+			err = -ENXIO;
+			goto out;
+		}
+		new_map = hc->new_map;
+		hc->new_map = NULL;
+		up_write(&_hash_lock);
+		if (new_map) {
+			error = "could not suspend %s";
+			if (!dm_suspended(md) && (err = dm_suspend(md, flags)))
+				goto out_put;
+			error = "could not swap table for %s";
+			if ((err = dm_swap_table(md, new_map)))
+				goto out_put;
+			set_disk_ro(dm_disk(md), !!(dm_table_get_mode(new_map) & FMODE_WRITE));
+			dm_table_put(new_map);
+		}
+		error = "could not resume %s";
+		if (dm_suspended(md))
+			err = dm_resume(md);
+	}
+out_put:
+	dm_put(md);
+out:
+	return ddlink_error(ddm->ddi, err, error, name);
+}
+
+struct dumb { struct ddmap *ddm; int *err; }; /* dumb way to err out iterator */
+
+static void per_target(struct target_type *tt, void *data)
+{
+	unsigned len = sizeof(struct ddtype) + strlen(tt->name) + 1;
+	struct dumb *info = data;
+	struct ddinode *dd = info->ddm->ddi;
+	struct dditem *item = dditem_new(dd, len);
+	struct ddversion version = { tt->version[0], tt->version[1], tt->version[2] };
+
+	if (!item) {
+		*(info->err) = -ENOMEM; /* oh my... */
+		return;
+	}
+	strcpy(((struct ddtype *)item->data)->name, tt->name);
+	((struct ddtype *)item->data)->version = version;
+	ddlink_queue(dd, item);
+}
+
+static int ddlist(struct ddmap *ddm)
+{
+	struct ddinode *dd = ddm->ddi;
+	struct hash_cell *hc;
+	int i, err = -ENOMEM;
+	down_write(&_hash_lock);
+	for (i = 0; i < NUM_BUCKETS; i++)
+		list_for_each_entry(hc, _name_buckets + i, name_list) {
+			struct gendisk *disk = dm_disk(hc->md);
+			int len = strlen(hc->name);
+			struct dditem *item = dditem_new(dd, sizeof(struct ddname) + len);
+			struct ddname *info;
+			if (!item)
+				goto bleah;
+			info = (void *)item->data;
+			*info = (struct ddname){{disk->major, disk->first_minor}};
+			memcpy(info->name, hc->name, len);
+			ddlink_queue(dd, item);
+		}
+	err = 0;
+bleah:
+	up_write(&_hash_lock);
+	return ddlink_error(dd, err, "why me?");
+}
+
+static int dmstatus(struct ddinode *dd, struct mapped_device *md)
+{
+	struct gendisk *disk = dm_disk(md);
+	struct dm_table *table;
+	unsigned flags = 0, targets = 0;
+
+	if (dm_suspended(md))
+		flags |= DMFLAG_SUSPEND;
+	if (disk->policy)
+		flags |= DMFLAG_READONLY;
+	if ((table = dm_get_table(md))) {
+		flags |= DMFLAG_PRESENT;
+		targets = dm_table_get_num_targets(table);
+		dm_table_put(table);
+	}
+	return ddlink_post(dd, &(struct dmstatus){
+			.dev = {disk->major, disk->first_minor},
+			.event = dm_get_event_nr(md),
+			.targets = targets,
+			.opens = dm_open_count(md), /* racy */
+			.flags = flags},
+		sizeof(struct dmstatus));
+}
+
+static int ddmap_ioctl(struct inode *inode, struct file *file, unsigned cmd, unsigned long u)
+{
+	struct ddmap *ddm = ddmap(inode);
+	struct ddinode *dd = ddm->ddi;
+	const char __user *user = (void *)u;
+	int err = 0, need = 0, depth = ddm->depth;
+
+	switch (cmd) {
+	case DMTARGET:
+	case DMCREATE:
+	case DMREMOVE:
+	case DMSUSPEND:
+	case DMRESUME:
+	case DMDEPS:
+	case DMSTATUS:
+		need = 1;
+		break;
+	case DMRENAME:
+		need = 2;
+		break;
+	}
+	if (depth < need)
+		return ddlink_error(dd, -ENODATA, "need %u args", need);
+
+	switch (cmd) {
+	case DMTABLE:
+	{
+		struct ddcreate got;
+		if (copy_from_user(&got, user, sizeof(got)))
+			return -EFAULT;
+		err = create(ddm, got.targets, got.mode);
+		break;
+	}
+	case DMTARGET:
+	{
+		struct ddtarget got;
+		if (copy_from_user(&got, user, sizeof(got)))
+			return -EFAULT;
+		err = target(ddm, got.offset, got.sectors);
+		break;
+	}
+	case DMCREATE:
+		err = install(ddm);
+		break;
+	case DMREMOVE:
+		err = remove(ddm);
+		break;
+	case DMSUSPEND:
+	case DMRESUME:
+		err = suspend(ddm, cmd == DMSUSPEND, 0, 0);
+		break;
+	case DMDEPS:
+		err = dependencies(ddm);
+		break;
+	case DMTYPES:
+		dm_target_iterate(per_target, &(struct dumb){ ddm, &err });
+		if (err)
+			return ddlink_error(dd, err, "iteration failed");
+		break;
+	case DMNAMES:
+		return ddlink_error(dd, ddlist(ddm), "failed to list devices");
+	case DMSTATUS:
+	{
+		char *name = ddtop(ddm);
+		struct mapped_device *md = lookup_device(name);
+		err = -ENXIO;
+		if (!md)
+			break;
+		err = dmstatus(dd, md);
+		break;
+	}
+	case DMRENAME:
+	{
+		struct dditem *old = dditem_from_list(ddm->stack.next);
+		struct dditem *new = dditem_from_list(ddm->stack.next->next);
+		err = dm_hash_rename(old->data, new->data);
+		if (err == -EINVAL)
+			ddlink_error(dd, err, "invalid name %s", new->data);
+		if (err == -ENXIO)
+			ddlink_error(dd, err, "no device %s", old->data);
+		if (err == -EBUSY)
+			ddlink_error(dd, err, "%s already exists", new->data);
+		break;
+	}
+	case DMVERSION:
+		return ddlink_post(dd, &(struct ddversion){0, 0, 0}, sizeof(struct ddversion));
+	default:
+		return ddlink_error(dd, -ENOTTY, "invalid ioctl %u", cmd);
+	}
+	ddmap_drop(ddm, depth);
+	return err;
+}
+
+/* ddmap file methods */
+
+static ssize_t ddmap_write(struct file *file, const char *buf, size_t len, loff_t *offset)
+{
+	struct ddinode *dd = ddinode(file->f_dentry->d_inode);
+	struct ddmap *ddm = ddmap(file->f_dentry->d_inode);
+	struct dditem *item;
+
+	if (ddm->depth >= 1000)
+		return ddlink_error(ddm->ddi, -E2BIG, "%i arguments", ddm->depth + 1);
+	if (IS_ERR(item = dditem_in(dd, buf, len, 1)))
+		return ddlink_error(ddm->ddi, PTR_ERR(item), "write failed");
+	list_add_tail(&item->link, &ddm->stack);
+	ddm->depth++;
+	return len;
+
+}
+
+static ssize_t ddmap_read(struct file *file, char *buf, size_t len, loff_t *offset)
+{
+	struct ddinode *dd = ddinode(file->f_dentry->d_inode);
+	struct dditem *item;
+	unsigned size;
+	if (!ddlink_ready(dd))
+		return 0;
+	item = ddlink_pop(dd);
+	size = item->size;
+	if (size > len)
+		return ddlink_error(dd, -EINVAL,
+			"tried to read %zi bytes of %u byte item", len, size);
+	if (copy_to_user(buf, item->data, size))
+		return ddlink_error(dd, -EFAULT, "invalid address");
+	dd->destroy_item(item);
+	return size;
+}
+
+static void ddmap_destroy(struct ddmap *ddm)
+{
+	ddmap_drop(ddm, ddm->depth);
+	if (ddm->dmt)
+		destroy_table(ddm->dmt);
+	kfree(ddm);
+}
+
+void *ddmap_create(struct ddinode *dd, void *info)
+{
+	struct ddmap *ddm;
+	if (!(ddm = kmalloc(sizeof(*ddm), __GFP_NOFAIL)))
+		return ERR_PTR(-ENOMEM);
+	*ddm = (typeof(*ddm)){ .ddi = dd };
+	dd->destroy_info = (typeof(&kfree))ddmap_destroy;
+	dd->create_item = kmalloc;
+	dd->destroy_item = kfree;
+	INIT_LIST_HEAD(&ddm->stack);
+	return ddm;
+}
+
+static struct file_operations ddmap_fops = {
+	.read = ddmap_read,
+	.write = ddmap_write,
+	.ioctl = ddmap_ioctl,
+	.poll = ddlink_poll };
+
 static int ctl_ioctl(struct inode *inode, struct file *file,
 		     uint command, ulong u)
 {
@@ -1413,6 +1945,9 @@ static int ctl_ioctl(struct inode *inode
 	if (!capable(CAP_SYS_ADMIN))
 		return -EACCES;
 
+	if (command == DDLINK)
+		return ddlink(&ddmap_fops, ddmap_create, NULL);
+
 	if (_IOC_TYPE(command) != DM_IOCTL)
 		return -ENOTTY;
 
--- 2.6.23.12.base/drivers/md/dm-table.c	2008-02-14 02:01:41.000000000 -0800
+++ 2.6.23.12/drivers/md/dm-table.c	2008-02-13 02:13:13.000000000 -0800
@@ -24,41 +24,6 @@
 #define KEYS_PER_NODE (NODE_SIZE / sizeof(sector_t))
 #define CHILDREN_PER_NODE (KEYS_PER_NODE + 1)
 
-struct dm_table {
-	struct mapped_device *md;
-	atomic_t holders;
-
-	/* btree table */
-	unsigned int depth;
-	unsigned int counts[MAX_DEPTH];	/* in nodes */
-	sector_t *index[MAX_DEPTH];
-
-	unsigned int num_targets;
-	unsigned int num_allocated;
-	sector_t *highs;
-	struct dm_target *targets;
-
-	/*
-	 * Indicates the rw permissions for the new logical
-	 * device.  This should be a combination of FMODE_READ
-	 * and FMODE_WRITE.
-	 */
-	int mode;
-
-	/* a list of devices used by this table */
-	struct list_head devices;
-
-	/*
-	 * These are optimistic limits taken from all the
-	 * targets, some targets will need smaller limits.
-	 */
-	struct io_restrictions limits;
-
-	/* events get handed up using this callback */
-	void (*event_fn)(void *);
-	void *event_context;
-};
-
 /*
  * Similar to ceiling(log_size(n))
  */
@@ -82,7 +47,7 @@ static unsigned int int_log(unsigned int
 /*
  * Combine two io_restrictions, always taking the lower value.
  */
-static void combine_restrictions_low(struct io_restrictions *lhs,
+void combine_restrictions_low(struct io_restrictions *lhs,
 				     struct io_restrictions *rhs)
 {
 	lhs->max_sectors =
@@ -179,7 +144,7 @@ void *dm_vcalloc(unsigned long nmemb, un
  * highs, and targets are managed as dynamic arrays during a
  * table load.
  */
-static int alloc_targets(struct dm_table *t, unsigned int num)
+int alloc_targets(struct dm_table *t, unsigned int num)
 {
 	sector_t *n_highs;
 	struct dm_target *n_targets;
@@ -334,17 +299,6 @@ void dm_table_put(struct dm_table *t)
 }
 
 /*
- * 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;

-}
-
-/*
  * Convert a device path to a dev_t.
  */
 static int lookup_device(const char *path, dev_t *dev)
@@ -598,7 +552,7 @@ void dm_put_device(struct dm_target *ti,
 /*
  * Checks to see if the target joins onto the end of the table.
  */
-static int adjoin(struct dm_table *table, struct dm_target *ti)
+int adjoin(struct dm_table *table, struct dm_target *ti)
 {
 	struct dm_target *prev;
 
--- 2.6.23.12.base/drivers/md/Makefile	2008-02-14 02:01:41.000000000 -0800
+++ 2.6.23.12/drivers/md/Makefile	2008-02-17 20:38:22.000000000 -0800
@@ -22,6 +22,7 @@ hostprogs-y	:= mktables
 # themselves, and md.o may use the personalities when it 
 # auto-initialised.
 
+obj-y				+= ddlink.o
 obj-$(CONFIG_MD_LINEAR)		+= linear.o
 obj-$(CONFIG_MD_RAID0)		+= raid0.o
 obj-$(CONFIG_MD_RAID1)		+= raid1.o
@@ -39,6 +40,7 @@ obj-$(CONFIG_DM_MULTIPATH_RDAC)	+= dm-rd
 obj-$(CONFIG_DM_SNAPSHOT)	+= dm-snapshot.o
 obj-$(CONFIG_DM_MIRROR)		+= dm-mirror.o
 obj-$(CONFIG_DM_ZERO)		+= dm-zero.o
+obj-y				+= dm-ramback.o
 
 quiet_cmd_unroll = UNROLL  $@
       cmd_unroll = $(PERL) $(srctree)/$(src)/unroll.pl $(UNROLL) \
--- 2.6.23.12.base/include/linux/ddlink.h	2008-02-14 02:01:41.000000000 -0800
+++ 2.6.23.12/include/linux/ddlink.h	2008-02-17 00:10:26.000000000 -0800
@@ -0,0 +1,41 @@
+#define DDLINK _IO(0, 0xdd)
+
+struct dditem {
+	struct list_head link;
+	unsigned type; // not actually used
+	unsigned size;
+	unsigned char data[];
+};
+
+struct ddinode {
+	struct list_head list;
+	wait_queue_head_t wait;
+	typeof(kmalloc) *create_item;
+	typeof(kfree) *destroy_item;
+	typeof(kfree) *destroy_info;
+	struct inode inode;
+};
+
+int ddlink(struct file_operations *fops,
+	void *(*create)(struct ddinode *dd, void *info), void *info);
+void ddlink_queue(struct ddinode *dd, struct dditem *item);
+void ddlink_push(struct ddinode *dd, struct dditem *item);
+struct dditem *ddlink_pop(struct ddinode *dd);
+struct dditem *dditem_new(struct ddinode *dd, size_t size);
+void ddlink_clear(struct ddinode *dd);
+int ddlink_ready(struct ddinode *dd);
+
+struct dditem *dditem_in(struct ddinode *dd, const void *buf, size_t len, int z);
+int ddlink_post(struct ddinode *dd, void *data, unsigned len);
+unsigned ddlink_poll(struct file *file, poll_table *table);
+int ddlink_error(struct ddinode *dd, int err, const char *fmt, ...);
+
+static inline struct ddinode *ddinode(struct inode *inode)
+{
+	return container_of(inode, struct ddinode, inode); 
+}
+
+static inline void *ddinfo(struct inode *inode)
+{
+	return inode->i_private;
+}
--- 2.6.23.12.base/include/linux/ddsetup.h	2008-02-14 02:01:41.000000000 -0800
+++ 2.6.23.12/include/linux/ddsetup.h	2008-02-17 01:47:22.000000000 -0800
@@ -0,0 +1,36 @@
+#ifndef DMSETUP_H
+#define DMSETUP_H
+
+#define DDLINK _IO(0, 0xdd)
+#define PACKED __attribute__ ((packed))
+
+enum { DMREAD = 1, DMWRITE = 2 };
+enum { DMRESULT = 0xDDC0DE00 };
+
+#define DMVERSION	_IO(0xdc, 0x0)
+#define DMSTRING	_IO(0xdc, 0x1)
+#define DMTABLE		_IO(0xdc, 0x2)
+#define DMTARGET	_IO(0xdc, 0x3)
+#define DMCREATE	_IO(0xdc, 0x4)
+#define DMREMOVE	_IO(0xdc, 0x5)
+#define DMSUSPEND	_IO(0xdc, 0x6)
+#define DMRESUME	_IO(0xdc, 0x7)
+#define DMRENAME	_IO(0xdc, 0x8)
+#define DMDEPS		_IO(0xdc, 0x9)
+#define DMNAMES		_IO(0xdc, 0xa)
+#define DMTYPES		_IO(0xdc, 0xb)
+#define DMSTATUS	_IO(0xdc, 0xc)
+
+#define DMFLAG_READONLY 1
+#define DMFLAG_PRESENT 2
+#define DMFLAG_SUSPEND 4
+
+struct devnum { uint32_t major, minor; } PACKED;
+struct ddcreate { uint32_t  mode, targets; } PACKED;
+struct ddtarget { uint64_t offset, sectors; uint32_t mode; } PACKED;
+struct ddresult { uint32_t magic; struct devnum dev; } PACKED;
+struct ddversion { uint32_t major, minor, point; } PACKED;
+struct ddtype { struct ddversion version; char name[]; } PACKED;
+struct ddname { struct devnum dev; char name[]; } PACKED;
+struct dmstatus { struct devnum dev; uint32_t flags, targets, opens, event; } PACKED;
+#endif
--- 2.6.23.12.base/include/linux/device-mapper.h	2008-02-14 02:01:41.000000000 -0800
+++ 2.6.23.12/include/linux/device-mapper.h	2008-02-13 02:13:13.000000000 -0800
@@ -147,6 +147,43 @@ struct dm_target {
 int dm_register_target(struct target_type *t);
 int dm_unregister_target(struct target_type *t);
 
+#define DM_MAX_DEPTH 16
+
+struct dm_table {
+	struct mapped_device *md;
+	atomic_t holders;
+
+	/* btree table */
+	unsigned int depth;
+	unsigned int counts[DM_MAX_DEPTH];	/* in nodes */
+	sector_t *index[DM_MAX_DEPTH];
+
+	unsigned int num_targets;
+	unsigned int num_allocated;
+	sector_t *highs;
+	struct dm_target *targets;
+
+	/*
+	 * Indicates the rw permissions for the new logical
+	 * device.  This should be a combination of FMODE_READ
+	 * and FMODE_WRITE.
+	 */
+	int mode;
+
+	/* a list of devices used by this table */
+	struct list_head devices;
+
+	/*
+	 * These are optimistic limits taken from all the
+	 * targets, some targets will need smaller limits.
+	 */
+	struct io_restrictions limits;
+
+	/* events get handed up using this callback */
+	void (*event_fn)(void *);
+	void *event_context;
+};
+
 
 /*-----------------------------------------------------------------
  * Functions for creating and manipulating mapped devices.

Show details Hide details

Change log

r1427 by Daniel.Raymond.Phillips on Mar 04, 2008 Diff

Really add the ddsetup files

Go to:

Project members, sign in to write a code review

Older revisions

All revisions of this file

File info

Size: 35936 bytes, 1362 lines

View raw file