Btrfs: Add data=ordered support

This forces file data extents down the disk along with the metadata that
references them.  The current implementation is fairly simple, and just
writes out all of the dirty pages in an inode before the commit.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

fs/btrfs/Makefile

@@ -5,7 +5,7 @@ obj-m  := btrfs.o
 btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   hash.o file-item.o inode-item.o inode-map.o disk-io.o \
 	   transaction.o bit-radix.o inode.o file.o tree-defrag.o \
-	   extent_map.o sysfs.o struct-funcs.o xattr.o acl.o
+	   extent_map.o sysfs.o struct-funcs.o xattr.o acl.o ordered-data.o
 
 #btrfs-y := ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \
 #	  root-tree.o dir-item.o hash.o file-item.o inode-item.o \

fs/btrfs/btrfs_inode.h

@@ -29,6 +29,7 @@ struct btrfs_inode {
 	struct extent_map_tree extent_tree;
 	struct inode vfs_inode;
 
+	u64 ordered_trans;
 	/*
 	 * transid of the trans_handle that last modified this inode
 	 */

fs/btrfs/ctree.c

@@ -221,7 +221,9 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		    struct extent_buffer **cow_ret)
 {
 	u64 search_start;
+	u64 header_trans;
 	int ret;
+
 	if (trans->transaction != root->fs_info->running_transaction) {
 		printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid,
 		       root->fs_info->running_transaction->transid);
@@ -232,7 +234,9 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		       root->fs_info->generation);
 		WARN_ON(1);
 	}
-	if (btrfs_header_generation(buf) == trans->transid) {
+
+	header_trans = btrfs_header_generation(buf);
+	if (header_trans == trans->transid) {
 		*cow_ret = buf;
 		return 0;
 	}

fs/btrfs/ctree.h

@@ -16,8 +16,8 @@
  * Boston, MA 021110-1307, USA.
  */
 
-#ifndef __BTRFS__
-#define __BTRFS__
+#ifndef __BTRFS_CTREE__
+#define __BTRFS_CTREE__
 
 #include <linux/version.h>
 #include <linux/mm.h>
@@ -363,7 +363,6 @@ struct btrfs_root {
 	struct inode *inode;
 	struct kobject root_kobj;
 	struct completion kobj_unregister;
-	struct rw_semaphore snap_sem;
 	u64 objectid;
 	u64 last_trans;
 
@@ -1142,6 +1141,8 @@ void btrfs_destroy_cachep(void);
 long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
 				struct btrfs_root *root);
+struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
+			    u64 root_objectid);
 int btrfs_commit_write(struct file *file, struct page *page,
 		       unsigned from, unsigned to);
 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,

fs/btrfs/disk-io.c

@@ -406,7 +406,6 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
 	memset(&root->root_kobj, 0, sizeof(root->root_kobj));
 	init_completion(&root->kobj_unregister);
-	init_rwsem(&root->snap_sem);
 	root->defrag_running = 0;
 	root->defrag_level = 0;
 	root->root_key.objectid = objectid;
@@ -498,6 +497,21 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 	return root;
 }
 
+struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
+					u64 root_objectid)
+{
+	struct btrfs_root *root;
+
+	if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
+		return fs_info->tree_root;
+	if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
+		return fs_info->extent_root;
+
+	root = radix_tree_lookup(&fs_info->fs_roots_radix,
+				 (unsigned long)root_objectid);
+	return root;
+}
+
 struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
 					      struct btrfs_key *location)
 {

fs/btrfs/disk-io.h

@@ -34,6 +34,8 @@ int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root);
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
 					    u64 bytenr, u32 blocksize);
+struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
+					u64 root_objectid);
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				      struct btrfs_key *location,
 				      const char *name, int namelen);

fs/btrfs/extent-tree.c

@@ -1195,7 +1195,9 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes,
 			if (btrfs_buffer_uptodate(buf)) {
 				u64 transid =
 				    root->fs_info->running_transaction->transid;
-				if (btrfs_header_generation(buf) == transid) {
+				u64 header_transid =
+					btrfs_header_generation(buf);
+				if (header_transid == transid) {
 					free_extent_buffer(buf);
 					return 1;
 				}

fs/btrfs/file.c

@@ -34,6 +34,7 @@
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
+#include "ordered-data.h"
 #include "ioctl.h"
 #include "print-tree.h"
 
@@ -329,6 +330,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans,
 		root->fs_info->delalloc_bytes += (end_of_last_block + 1 -
 					  start_pos) - existing_delalloc;
 		spin_unlock(&root->fs_info->delalloc_lock);
+		btrfs_add_ordered_inode(inode);
 	} else {
 		u64 aligned_end;
 		/* step one, delete the existing extents in this range */
@@ -724,8 +726,6 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 
 	pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL);
 
-	down_read(&BTRFS_I(inode)->root->snap_sem);
-
 	mutex_lock(&inode->i_mutex);
 	first_index = pos >> PAGE_CACHE_SHIFT;
 	last_index = (pos + count) >> PAGE_CACHE_SHIFT;
@@ -804,7 +804,6 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 	}
 out:
 	mutex_unlock(&inode->i_mutex);
-	up_read(&BTRFS_I(inode)->root->snap_sem);
 
 out_nolock:
 	kfree(pages);

fs/btrfs/inode.c

@@ -135,6 +135,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end)
 		alloc_hint = ins.objectid + ins.offset;
 		start += cur_alloc_size;
 	}
+	btrfs_add_ordered_inode(inode);
 out:
 	btrfs_end_transaction(trans, root);
 	return ret;
@@ -367,8 +368,8 @@ void btrfs_read_locked_inode(struct inode *inode)
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
 	mutex_lock(&root->fs_info->fs_mutex);
-
 	memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
+
 	ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
 	if (ret)
 		goto make_bad;
@@ -898,7 +899,6 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 	if ((offset & (blocksize - 1)) == 0)
 		goto out;
 
-	down_read(&root->snap_sem);
 	ret = -ENOMEM;
 	page = grab_cache_page(mapping, index);
 	if (!page)
@@ -917,7 +917,6 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 
 	unlock_page(page);
 	page_cache_release(page);
-	up_read(&BTRFS_I(inode)->root->snap_sem);
 out:
 	return ret;
 }
@@ -1146,6 +1145,19 @@ static int btrfs_find_actor(struct inode *inode, void *opaque)
 		args->root == BTRFS_I(inode)->root);
 }
 
+struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
+			    u64 root_objectid)
+{
+	struct btrfs_iget_args args;
+	args.ino = objectid;
+	args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
+
+	if (!args.root)
+		return NULL;
+
+	return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
+}
+
 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
 				struct btrfs_root *root)
 {
@@ -1336,7 +1348,6 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 
 			d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
 			btrfs_dir_item_key_to_cpu(leaf, di, &location);
-
 			over = filldir(dirent, name_ptr, name_len,
 				       found_key.offset,
 				       location.objectid,
@@ -2054,7 +2065,6 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
 
 	ret = -EINVAL;
 
-	down_read(&BTRFS_I(inode)->root->snap_sem);
 	lock_page(page);
 	wait_on_page_writeback(page);
 	size = i_size_read(inode);
@@ -2075,7 +2085,6 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
 	ret = btrfs_cow_one_page(inode, page, end);
 
 out_unlock:
-	up_read(&BTRFS_I(inode)->root->snap_sem);
 	unlock_page(page);
 out:
 	return ret;
@@ -2118,7 +2127,7 @@ static int noinline create_subvol(struct btrfs_root *root, char *name,
 	struct btrfs_root_item root_item;
 	struct btrfs_inode_item *inode_item;
 	struct extent_buffer *leaf;
-	struct btrfs_root *new_root;
+	struct btrfs_root *new_root = root;
 	struct inode *inode;
 	struct inode *dir;
 	int ret;
@@ -2230,7 +2239,7 @@ static int noinline create_subvol(struct btrfs_root *root, char *name,
 		goto fail;
 fail:
 	nr = trans->blocks_used;
-	err = btrfs_commit_transaction(trans, root);
+	err = btrfs_commit_transaction(trans, new_root);
 	if (err && !ret)
 		ret = err;
 fail_commit:
@@ -2253,17 +2262,16 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	if (!root->ref_cows)
 		return -EINVAL;
 
-	down_write(&root->snap_sem);
-	freeze_bdev(root->fs_info->sb->s_bdev);
-	thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb);
-
 	mutex_lock(&root->fs_info->fs_mutex);
 	ret = btrfs_check_free_space(root, 1, 0);
 	if (ret)
 		goto fail_unlock;
 
 	trans = btrfs_start_transaction(root, 1);
 	BUG_ON(!trans);
+	err = btrfs_commit_transaction(trans, root);
+
+	trans = btrfs_start_transaction(root, 1);
 
 	ret = btrfs_update_inode(trans, root, root->inode);
 	if (ret)
@@ -2272,9 +2280,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
 				       0, &objectid);
 	if (ret)
-		goto fail;
-
-	memcpy(&new_root_item, &root->root_item,
+		goto fail; memcpy(&new_root_item, &root->root_item,
 	       sizeof(new_root_item));
 
 	key.objectid = objectid;
@@ -2285,12 +2291,20 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
 	free_extent_buffer(tmp);
 
+	/* write the ordered inodes to force all delayed allocations to
+	 * be filled.  Once this is done, we can copy the root
+	 */
+	mutex_lock(&root->fs_info->trans_mutex);
+	btrfs_write_ordered_inodes(trans, root);
+	mutex_unlock(&root->fs_info->trans_mutex);
+
 	btrfs_copy_root(trans, root, root->node, &tmp, objectid);
 
 	btrfs_set_root_bytenr(&new_root_item, tmp->start);
 	btrfs_set_root_level(&new_root_item, btrfs_header_level(tmp));
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 				&new_root_item);
+printk("new root %Lu node %Lu\n", objectid, tmp->start);
 	free_extent_buffer(tmp);
 	if (ret)
 		goto fail;
@@ -2321,7 +2335,6 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 		ret = err;
 fail_unlock:
 	mutex_unlock(&root->fs_info->fs_mutex);
-	up_write(&root->snap_sem);
 	btrfs_btree_balance_dirty(root, nr);
 	return ret;
 }
@@ -2608,6 +2621,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
 	if (!ei)
 		return NULL;
 	ei->last_trans = 0;
+	ei->ordered_trans = 0;
 	return &ei->vfs_inode;
 }