X-Git-Url: https://git.openpandora.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=fs%2Fbtrfs%2Fdisk-io.c;h=ffc363d2fb24ada506b1baa8cd15530e8d058ec7;hb=051e1b9f748ae673b7325d3fc049bb838606cffa;hp=3e16cca72b4962d04d82308b1c2cd28caa65c13d;hpb=324ae4df00fdc1a6a179bf584d8addf027bb75fb;p=pandora-kernel.git

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 3e16cca72b49..ffc363d2fb24 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -18,17 +18,21 @@
 
 #include <linux/fs.h>
 #include <linux/blkdev.h>
-#include <linux/crc32c.h>
 #include <linux/scatterlist.h>
 #include <linux/swap.h>
 #include <linux/radix-tree.h>
 #include <linux/writeback.h>
 #include <linux/buffer_head.h> // for block_sync_page
+#include <linux/workqueue.h>
+#include "crc32c.h"
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
+#include "volumes.h"
 #include "print-tree.h"
+#include "async-thread.h"
+#include "locking.h"
 
 #if 0
 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
@@ -43,67 +47,98 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
 }
 #endif
 
-static struct extent_map_ops btree_extent_map_ops;
-
-struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
-					    u64 bytenr, u32 blocksize)
-{
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	struct extent_buffer *eb;
-	eb = find_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
-				bytenr, blocksize, GFP_NOFS);
-	return eb;
-}
-
-struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
-						 u64 bytenr, u32 blocksize)
-{
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	struct extent_buffer *eb;
+static struct extent_io_ops btree_extent_io_ops;
+static void end_workqueue_fn(struct btrfs_work *work);
+
+struct end_io_wq {
+	struct bio *bio;
+	bio_end_io_t *end_io;
+	void *private;
+	struct btrfs_fs_info *info;
+	int error;
+	int metadata;
+	struct list_head list;
+	struct btrfs_work work;
+};
 
-	eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
-				 bytenr, blocksize, NULL, GFP_NOFS);
-	return eb;
-}
+struct async_submit_bio {
+	struct inode *inode;
+	struct bio *bio;
+	struct list_head list;
+	extent_submit_bio_hook_t *submit_bio_hook;
+	int rw;
+	int mirror_num;
+	struct btrfs_work work;
+};
 
 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
-				    size_t page_offset, u64 start, u64 end,
+				    size_t page_offset, u64 start, u64 len,
 				    int create)
 {
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	struct extent_map *em;
 	int ret;
 
-again:
-	em = lookup_extent_mapping(em_tree, start, end);
+	spin_lock(&em_tree->lock);
+	em = lookup_extent_mapping(em_tree, start, len);
 	if (em) {
+		em->bdev =
+			BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
+		spin_unlock(&em_tree->lock);
 		goto out;
 	}
+	spin_unlock(&em_tree->lock);
+
 	em = alloc_extent_map(GFP_NOFS);
 	if (!em) {
 		em = ERR_PTR(-ENOMEM);
 		goto out;
 	}
 	em->start = 0;
-	em->end = (i_size_read(inode) & ~((u64)PAGE_CACHE_SIZE -1)) - 1;
+	em->len = (u64)-1;
 	em->block_start = 0;
-	em->block_end = em->end;
-	em->bdev = inode->i_sb->s_bdev;
+	em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
+
+	spin_lock(&em_tree->lock);
 	ret = add_extent_mapping(em_tree, em);
 	if (ret == -EEXIST) {
+		u64 failed_start = em->start;
+		u64 failed_len = em->len;
+
+		printk("failed to insert %Lu %Lu -> %Lu into tree\n",
+		       em->start, em->len, em->block_start);
 		free_extent_map(em);
-		em = NULL;
-		goto again;
+		em = lookup_extent_mapping(em_tree, start, len);
+		if (em) {
+			printk("after failing, found %Lu %Lu %Lu\n",
+			       em->start, em->len, em->block_start);
+			ret = 0;
+		} else {
+			em = lookup_extent_mapping(em_tree, failed_start,
+						   failed_len);
+			if (em) {
+				printk("double failure lookup gives us "
+				       "%Lu %Lu -> %Lu\n", em->start,
+				       em->len, em->block_start);
+				free_extent_map(em);
+			}
+			ret = -EIO;
+		}
 	} else if (ret) {
-		em = ERR_PTR(ret);
+		free_extent_map(em);
+		em = NULL;
 	}
+	spin_unlock(&em_tree->lock);
+
+	if (ret)
+		em = ERR_PTR(ret);
 out:
 	return em;
 }
 
 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
 {
-	return crc32c(seed, data, len);
+	return btrfs_crc32c(seed, data, len);
 }
 
 void btrfs_csum_final(u32 crc, char *result)
@@ -145,10 +180,26 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 	btrfs_csum_final(crc, result);
 
 	if (verify) {
+		int from_this_trans = 0;
+
+		if (root->fs_info->running_transaction &&
+		    btrfs_header_generation(buf) ==
+		    root->fs_info->running_transaction->transid)
+			from_this_trans = 1;
+
+		/* FIXME, this is not good */
 		if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
-			printk("btrfs: %s checksum verify failed on %llu\n",
+			u32 val;
+			u32 found = 0;
+			memcpy(&found, result, BTRFS_CRC32_SIZE);
+
+			read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE);
+			printk("btrfs: %s checksum verify failed on %llu "
+			       "wanted %X found %X from_this_trans %d "
+			       "level %d\n",
 			       root->fs_info->sb->s_id,
-			       buf->start);
+			       buf->start, val, found, from_this_trans,
+			       btrfs_header_level(buf));
 			return 1;
 		}
 	} else {
@@ -157,16 +208,73 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 	return 0;
 }
 
+static int verify_parent_transid(struct extent_io_tree *io_tree,
+				 struct extent_buffer *eb, u64 parent_transid)
+{
+	int ret;
+
+	if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
+		return 0;
+
+	lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS);
+	if (extent_buffer_uptodate(io_tree, eb) &&
+	    btrfs_header_generation(eb) == parent_transid) {
+		ret = 0;
+		goto out;
+	}
+	printk("parent transid verify failed on %llu wanted %llu found %llu\n",
+	       (unsigned long long)eb->start,
+	       (unsigned long long)parent_transid,
+	       (unsigned long long)btrfs_header_generation(eb));
+	ret = 1;
+out:
+	clear_extent_buffer_uptodate(io_tree, eb);
+	unlock_extent(io_tree, eb->start, eb->start + eb->len - 1,
+		      GFP_NOFS);
+	return ret;
+
+}
+
+static int btree_read_extent_buffer_pages(struct btrfs_root *root,
+					  struct extent_buffer *eb,
+					  u64 start, u64 parent_transid)
+{
+	struct extent_io_tree *io_tree;
+	int ret;
+	int num_copies = 0;
+	int mirror_num = 0;
+
+	io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
+	while (1) {
+		ret = read_extent_buffer_pages(io_tree, eb, start, 1,
+					       btree_get_extent, mirror_num);
+		if (!ret &&
+		    !verify_parent_transid(io_tree, eb, parent_transid))
+			return ret;
+
+		num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
+					      eb->start, eb->len);
+		if (num_copies == 1)
+			return ret;
+
+		mirror_num++;
+		if (mirror_num > num_copies)
+			return ret;
+	}
+	return -EIO;
+}
 
 int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
 {
-	struct extent_map_tree *tree;
+	struct extent_io_tree *tree;
 	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
 	u64 found_start;
 	int found_level;
 	unsigned long len;
 	struct extent_buffer *eb;
-	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	int ret;
+
+	tree = &BTRFS_I(page->mapping->host)->io_tree;
 
 	if (page->private == EXTENT_PAGE_PRIVATE)
 		goto out;
@@ -177,14 +285,34 @@ int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
 		WARN_ON(1);
 	}
 	eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
-	read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1);
+	ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
+					     btrfs_header_generation(eb));
+	BUG_ON(ret);
+	btrfs_clear_buffer_defrag(eb);
 	found_start = btrfs_header_bytenr(eb);
 	if (found_start != start) {
 		printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
 		       start, found_start, len);
+		WARN_ON(1);
+		goto err;
+	}
+	if (eb->first_page != page) {
+		printk("bad first page %lu %lu\n", eb->first_page->index,
+		       page->index);
+		WARN_ON(1);
+		goto err;
+	}
+	if (!PageUptodate(page)) {
+		printk("csum not up to date page %lu\n", page->index);
+		WARN_ON(1);
+		goto err;
 	}
 	found_level = btrfs_header_level(eb);
+	spin_lock(&root->fs_info->hash_lock);
+	btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
+	spin_unlock(&root->fs_info->hash_lock);
 	csum_tree_block(root, eb, 0);
+err:
 	free_extent_buffer(eb);
 out:
 	return 0;
@@ -198,36 +326,245 @@ static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
 	return 0;
 }
 
+int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
+			       struct extent_state *state)
+{
+	struct extent_io_tree *tree;
+	u64 found_start;
+	int found_level;
+	unsigned long len;
+	struct extent_buffer *eb;
+	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
+	int ret = 0;
+
+	tree = &BTRFS_I(page->mapping->host)->io_tree;
+	if (page->private == EXTENT_PAGE_PRIVATE)
+		goto out;
+	if (!page->private)
+		goto out;
+	len = page->private >> 2;
+	if (len == 0) {
+		WARN_ON(1);
+	}
+	eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
+
+	btrfs_clear_buffer_defrag(eb);
+	found_start = btrfs_header_bytenr(eb);
+	if (found_start != start) {
+		ret = -EIO;
+		goto err;
+	}
+	if (eb->first_page != page) {
+		printk("bad first page %lu %lu\n", eb->first_page->index,
+		       page->index);
+		WARN_ON(1);
+		ret = -EIO;
+		goto err;
+	}
+	if (memcmp_extent_buffer(eb, root->fs_info->fsid,
+				 (unsigned long)btrfs_header_fsid(eb),
+				 BTRFS_FSID_SIZE)) {
+		printk("bad fsid on block %Lu\n", eb->start);
+		ret = -EIO;
+		goto err;
+	}
+	found_level = btrfs_header_level(eb);
+
+	ret = csum_tree_block(root, eb, 1);
+	if (ret)
+		ret = -EIO;
+
+	end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
+	end = eb->start + end - 1;
+	release_extent_buffer_tail_pages(eb);
+err:
+	free_extent_buffer(eb);
+out:
+	return ret;
+}
+
+#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
+static void end_workqueue_bio(struct bio *bio, int err)
+#else
+static int end_workqueue_bio(struct bio *bio,
+				   unsigned int bytes_done, int err)
+#endif
+{
+	struct end_io_wq *end_io_wq = bio->bi_private;
+	struct btrfs_fs_info *fs_info;
+
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
+	if (bio->bi_size)
+		return 1;
+#endif
+
+	fs_info = end_io_wq->info;
+	end_io_wq->error = err;
+	end_io_wq->work.func = end_workqueue_fn;
+	end_io_wq->work.flags = 0;
+	btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work);
+
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
+	return 0;
+#endif
+}
+
+int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
+			int metadata)
+{
+	struct end_io_wq *end_io_wq;
+	end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
+	if (!end_io_wq)
+		return -ENOMEM;
+
+	end_io_wq->private = bio->bi_private;
+	end_io_wq->end_io = bio->bi_end_io;
+	end_io_wq->info = info;
+	end_io_wq->error = 0;
+	end_io_wq->bio = bio;
+	end_io_wq->metadata = metadata;
+
+	bio->bi_private = end_io_wq;
+	bio->bi_end_io = end_workqueue_bio;
+	return 0;
+}
+
+static void run_one_async_submit(struct btrfs_work *work)
+{
+	struct btrfs_fs_info *fs_info;
+	struct async_submit_bio *async;
+
+	async = container_of(work, struct  async_submit_bio, work);
+	fs_info = BTRFS_I(async->inode)->root->fs_info;
+	atomic_dec(&fs_info->nr_async_submits);
+	async->submit_bio_hook(async->inode, async->rw, async->bio,
+			       async->mirror_num);
+	kfree(async);
+}
+
+int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
+			int rw, struct bio *bio, int mirror_num,
+			extent_submit_bio_hook_t *submit_bio_hook)
+{
+	struct async_submit_bio *async;
+
+	async = kmalloc(sizeof(*async), GFP_NOFS);
+	if (!async)
+		return -ENOMEM;
+
+	async->inode = inode;
+	async->rw = rw;
+	async->bio = bio;
+	async->mirror_num = mirror_num;
+	async->submit_bio_hook = submit_bio_hook;
+	async->work.func = run_one_async_submit;
+	async->work.flags = 0;
+	atomic_inc(&fs_info->nr_async_submits);
+	btrfs_queue_worker(&fs_info->workers, &async->work);
+	return 0;
+}
+
+static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+				 int mirror_num)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 offset;
+	int ret;
+
+	offset = bio->bi_sector << 9;
+
+	/*
+	 * when we're called for a write, we're already in the async
+	 * submission context.  Just jump ingo btrfs_map_bio
+	 */
+	if (rw & (1 << BIO_RW)) {
+		return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
+				     mirror_num, 0);
+	}
+
+	/*
+	 * called for a read, do the setup so that checksum validation
+	 * can happen in the async kernel threads
+	 */
+	ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1);
+	BUG_ON(ret);
+
+	return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
+}
+
+static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+				 int mirror_num)
+{
+	/*
+	 * kthread helpers are used to submit writes so that checksumming
+	 * can happen in parallel across all CPUs
+	 */
+	if (!(rw & (1 << BIO_RW))) {
+		return __btree_submit_bio_hook(inode, rw, bio, mirror_num);
+	}
+	return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
+				   inode, rw, bio, mirror_num,
+				   __btree_submit_bio_hook);
+}
+
 static int btree_writepage(struct page *page, struct writeback_control *wbc)
 {
-	struct extent_map_tree *tree;
-	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	struct extent_io_tree *tree;
+	tree = &BTRFS_I(page->mapping->host)->io_tree;
 	return extent_write_full_page(tree, page, btree_get_extent, wbc);
 }
 
 static int btree_writepages(struct address_space *mapping,
 			    struct writeback_control *wbc)
 {
-	struct extent_map_tree *tree;
-	tree = &BTRFS_I(mapping->host)->extent_tree;
+	struct extent_io_tree *tree;
+	tree = &BTRFS_I(mapping->host)->io_tree;
+	if (wbc->sync_mode == WB_SYNC_NONE) {
+		u64 num_dirty;
+		u64 start = 0;
+		unsigned long thresh = 96 * 1024 * 1024;
+
+		if (wbc->for_kupdate)
+			return 0;
+
+		if (current_is_pdflush()) {
+			thresh = 96 * 1024 * 1024;
+		} else {
+			thresh = 8 * 1024 * 1024;
+		}
+		num_dirty = count_range_bits(tree, &start, (u64)-1,
+					     thresh, EXTENT_DIRTY);
+		if (num_dirty < thresh) {
+			return 0;
+		}
+	}
 	return extent_writepages(tree, mapping, btree_get_extent, wbc);
 }
 
 int btree_readpage(struct file *file, struct page *page)
 {
-	struct extent_map_tree *tree;
-	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	struct extent_io_tree *tree;
+	tree = &BTRFS_I(page->mapping->host)->io_tree;
 	return extent_read_full_page(tree, page, btree_get_extent);
 }
 
-static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags)
+static int btree_releasepage(struct page *page, gfp_t gfp_flags)
 {
-	struct extent_map_tree *tree;
+	struct extent_io_tree *tree;
+	struct extent_map_tree *map;
 	int ret;
 
-	tree = &BTRFS_I(page->mapping->host)->extent_tree;
-	ret = try_release_extent_mapping(tree, page);
+	if (page_count(page) > 3) {
+		/* once for page->private, once for the caller, once
+		 * once for the page cache
+		 */
+		return 0;
+	}
+	tree = &BTRFS_I(page->mapping->host)->io_tree;
+	map = &BTRFS_I(page->mapping->host)->extent_tree;
+	ret = try_release_extent_state(map, tree, page, gfp_flags);
 	if (ret == 1) {
+		invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE);
 		ClearPagePrivate(page);
 		set_page_private(page, 0);
 		page_cache_release(page);
@@ -237,10 +574,16 @@ static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags)
 
 static void btree_invalidatepage(struct page *page, unsigned long offset)
 {
-	struct extent_map_tree *tree;
-	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	struct extent_io_tree *tree;
+	tree = &BTRFS_I(page->mapping->host)->io_tree;
 	extent_invalidatepage(tree, page, offset);
 	btree_releasepage(page, GFP_NOFS);
+	if (PagePrivate(page)) {
+		invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE);
+		ClearPagePrivate(page);
+		set_page_private(page, 0);
+		page_cache_release(page);
+	}
 }
 
 #if 0
@@ -273,7 +616,8 @@ static struct address_space_operations btree_aops = {
 	.sync_page	= block_sync_page,
 };
 
-int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
+int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
+			 u64 parent_transid)
 {
 	struct extent_buffer *buf = NULL;
 	struct inode *btree_inode = root->fs_info->btree_inode;
@@ -282,48 +626,67 @@ int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
 	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
 	if (!buf)
 		return 0;
-	read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
-				 buf, 0, 0);
+	read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
+				 buf, 0, 0, btree_get_extent, 0);
 	free_extent_buffer(buf);
 	return ret;
 }
 
+struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
+					    u64 bytenr, u32 blocksize)
+{
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	struct extent_buffer *eb;
+	eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
+				bytenr, blocksize, GFP_NOFS);
+	return eb;
+}
+
+struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
+						 u64 bytenr, u32 blocksize)
+{
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	struct extent_buffer *eb;
+
+	eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
+				 bytenr, blocksize, NULL, GFP_NOFS);
+	return eb;
+}
+
+
 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
-				      u32 blocksize)
+				      u32 blocksize, u64 parent_transid)
 {
 	struct extent_buffer *buf = NULL;
 	struct inode *btree_inode = root->fs_info->btree_inode;
-	struct extent_map_tree *extent_tree;
+	struct extent_io_tree *io_tree;
 	int ret;
 
-	extent_tree = &BTRFS_I(btree_inode)->extent_tree;
+	io_tree = &BTRFS_I(btree_inode)->io_tree;
 
 	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
 	if (!buf)
 		return NULL;
-	read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
-				 buf, 0, 1);
-	if (buf->flags & EXTENT_CSUM) {
-		return buf;
-	}
-	if (test_range_bit(extent_tree, buf->start, buf->start + buf->len - 1,
-			   EXTENT_CSUM, 1)) {
-		buf->flags |= EXTENT_CSUM;
-		return buf;
-	}
-	ret = csum_tree_block(root, buf, 1);
-	set_extent_bits(extent_tree, buf->start,
-			buf->start + buf->len - 1,
-			EXTENT_CSUM, GFP_NOFS);
-	buf->flags |= EXTENT_CSUM;
+
+	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
+
+	if (ret == 0) {
+		buf->flags |= EXTENT_UPTODATE;
+	}
 	return buf;
+
 }
 
 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		     struct extent_buffer *buf)
 {
 	struct inode *btree_inode = root->fs_info->btree_inode;
-	clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
+	if (btrfs_header_generation(buf) ==
+	    root->fs_info->running_transaction->transid) {
+		WARN_ON(!btrfs_tree_locked(buf));
+		clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
+					  buf);
+	}
 	return 0;
 }
 
@@ -331,13 +694,13 @@ int wait_on_tree_block_writeback(struct btrfs_root *root,
 				 struct extent_buffer *buf)
 {
 	struct inode *btree_inode = root->fs_info->btree_inode;
-	wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->extent_tree,
+	wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree,
 					buf);
 	return 0;
 }
 
 static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
-			struct btrfs_root *root,
+			u32 stripesize, struct btrfs_root *root,
 			struct btrfs_fs_info *fs_info,
 			u64 objectid)
 {
@@ -347,19 +710,26 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	root->sectorsize = sectorsize;
 	root->nodesize = nodesize;
 	root->leafsize = leafsize;
+	root->stripesize = stripesize;
 	root->ref_cows = 0;
+	root->track_dirty = 0;
+
 	root->fs_info = fs_info;
 	root->objectid = objectid;
 	root->last_trans = 0;
 	root->highest_inode = 0;
 	root->last_inode_alloc = 0;
 	root->name = NULL;
+	root->in_sysfs = 0;
+
+	INIT_LIST_HEAD(&root->dirty_list);
+	spin_lock_init(&root->node_lock);
+	mutex_init(&root->objectid_mutex);
 	memset(&root->root_key, 0, sizeof(root->root_key));
 	memset(&root->root_item, 0, sizeof(root->root_item));
 	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;
@@ -375,14 +745,15 @@ static int find_and_setup_root(struct btrfs_root *tree_root,
 	u32 blocksize;
 
 	__setup_root(tree_root->nodesize, tree_root->leafsize,
-		     tree_root->sectorsize, root, fs_info, objectid);
+		     tree_root->sectorsize, tree_root->stripesize,
+		     root, fs_info, objectid);
 	ret = btrfs_find_last_root(tree_root, objectid,
 				   &root->root_item, &root->root_key);
 	BUG_ON(ret);
 
 	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
-				     blocksize);
+				     blocksize, 0);
 	BUG_ON(!root->node);
 	return 0;
 }
@@ -412,8 +783,8 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 	}
 
 	__setup_root(tree_root->nodesize, tree_root->leafsize,
-		     tree_root->sectorsize, root, fs_info,
-		     location->objectid);
+		     tree_root->sectorsize, tree_root->stripesize,
+		     root, fs_info, location->objectid);
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
@@ -438,7 +809,7 @@ out:
 	}
 	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
-				     blocksize);
+				     blocksize, 0);
 	BUG_ON(!root->node);
 insert:
 	root->ref_cows = 1;
@@ -450,13 +821,36 @@ insert:
 	return root;
 }
 
-struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
-				      struct btrfs_key *location,
-				      const char *name, int namelen)
+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)
 {
 	struct btrfs_root *root;
 	int ret;
 
+	if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
+		return fs_info->tree_root;
+	if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
+		return fs_info->extent_root;
+	if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
+		return fs_info->chunk_root;
+	if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
+		return fs_info->dev_root;
+
 	root = radix_tree_lookup(&fs_info->fs_roots_radix,
 				 (unsigned long)location->objectid);
 	if (root)
@@ -473,6 +867,26 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 		kfree(root);
 		return ERR_PTR(ret);
 	}
+	ret = btrfs_find_dead_roots(fs_info->tree_root,
+				    root->root_key.objectid, root);
+	BUG_ON(ret);
+
+	return root;
+}
+
+struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
+				      struct btrfs_key *location,
+				      const char *name, int namelen)
+{
+	struct btrfs_root *root;
+	int ret;
+
+	root = btrfs_read_fs_root_no_name(fs_info, location);
+	if (!root)
+		return NULL;
+
+	if (root->in_sysfs)
+		return root;
 
 	ret = btrfs_set_root_name(root, name, namelen);
 	if (ret) {
@@ -488,11 +902,7 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 		kfree(root);
 		return ERR_PTR(ret);
 	}
-
-	ret = btrfs_find_dead_roots(fs_info->tree_root,
-				    root->root_key.objectid, root);
-	BUG_ON(ret);
-
+	root->in_sysfs = 1;
 	return root;
 }
 #if 0
@@ -513,20 +923,206 @@ static int add_hasher(struct btrfs_fs_info *info, char *type) {
 	return 0;
 }
 #endif
-struct btrfs_root *open_ctree(struct super_block *sb)
+
+static int btrfs_congested_fn(void *congested_data, int bdi_bits)
+{
+	struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
+	int ret = 0;
+	int limit = 256 * info->fs_devices->open_devices;
+	struct list_head *cur;
+	struct btrfs_device *device;
+	struct backing_dev_info *bdi;
+
+	if ((bdi_bits & (1 << BDI_write_congested)) &&
+	    atomic_read(&info->nr_async_submits) > limit) {
+		return 1;
+	}
+
+	list_for_each(cur, &info->fs_devices->devices) {
+		device = list_entry(cur, struct btrfs_device, dev_list);
+		if (!device->bdev)
+			continue;
+		bdi = blk_get_backing_dev_info(device->bdev);
+		if (bdi && bdi_congested(bdi, bdi_bits)) {
+			ret = 1;
+			break;
+		}
+	}
+	return ret;
+}
+
+/*
+ * this unplugs every device on the box, and it is only used when page
+ * is null
+ */
+static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
+{
+	struct list_head *cur;
+	struct btrfs_device *device;
+	struct btrfs_fs_info *info;
+
+	info = (struct btrfs_fs_info *)bdi->unplug_io_data;
+	list_for_each(cur, &info->fs_devices->devices) {
+		device = list_entry(cur, struct btrfs_device, dev_list);
+		bdi = blk_get_backing_dev_info(device->bdev);
+		if (bdi->unplug_io_fn) {
+			bdi->unplug_io_fn(bdi, page);
+		}
+	}
+}
+
+void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
+{
+	struct inode *inode;
+	struct extent_map_tree *em_tree;
+	struct extent_map *em;
+	struct address_space *mapping;
+	u64 offset;
+
+	/* the generic O_DIRECT read code does this */
+	if (!page) {
+		__unplug_io_fn(bdi, page);
+		return;
+	}
+
+	/*
+	 * page->mapping may change at any time.  Get a consistent copy
+	 * and use that for everything below
+	 */
+	smp_mb();
+	mapping = page->mapping;
+	if (!mapping)
+		return;
+
+	inode = mapping->host;
+	offset = page_offset(page);
+
+	em_tree = &BTRFS_I(inode)->extent_tree;
+	spin_lock(&em_tree->lock);
+	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
+	spin_unlock(&em_tree->lock);
+	if (!em)
+		return;
+
+	offset = offset - em->start;
+	btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree,
+			  em->block_start + offset, page);
+	free_extent_map(em);
+}
+
+static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
+{
+#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
+	bdi_init(bdi);
+#endif
+	bdi->ra_pages	= default_backing_dev_info.ra_pages;
+	bdi->state		= 0;
+	bdi->capabilities	= default_backing_dev_info.capabilities;
+	bdi->unplug_io_fn	= btrfs_unplug_io_fn;
+	bdi->unplug_io_data	= info;
+	bdi->congested_fn	= btrfs_congested_fn;
+	bdi->congested_data	= info;
+	return 0;
+}
+
+static int bio_ready_for_csum(struct bio *bio)
+{
+	u64 length = 0;
+	u64 buf_len = 0;
+	u64 start = 0;
+	struct page *page;
+	struct extent_io_tree *io_tree = NULL;
+	struct btrfs_fs_info *info = NULL;
+	struct bio_vec *bvec;
+	int i;
+	int ret;
+
+	bio_for_each_segment(bvec, bio, i) {
+		page = bvec->bv_page;
+		if (page->private == EXTENT_PAGE_PRIVATE) {
+			length += bvec->bv_len;
+			continue;
+		}
+		if (!page->private) {
+			length += bvec->bv_len;
+			continue;
+		}
+		length = bvec->bv_len;
+		buf_len = page->private >> 2;
+		start = page_offset(page) + bvec->bv_offset;
+		io_tree = &BTRFS_I(page->mapping->host)->io_tree;
+		info = BTRFS_I(page->mapping->host)->root->fs_info;
+	}
+	/* are we fully contained in this bio? */
+	if (buf_len <= length)
+		return 1;
+
+	ret = extent_range_uptodate(io_tree, start + length,
+				    start + buf_len - 1);
+	if (ret == 1)
+		return ret;
+	return ret;
+}
+
+/*
+ * called by the kthread helper functions to finally call the bio end_io
+ * functions.  This is where read checksum verification actually happens
+ */
+static void end_workqueue_fn(struct btrfs_work *work)
+{
+	struct bio *bio;
+	struct end_io_wq *end_io_wq;
+	struct btrfs_fs_info *fs_info;
+	int error;
+
+	end_io_wq = container_of(work, struct end_io_wq, work);
+	bio = end_io_wq->bio;
+	fs_info = end_io_wq->info;
+
+	/* metadata bios are special because the whole tree block must
+	 * be checksummed at once.  This makes sure the entire block is in
+	 * ram and up to date before trying to verify things.  For
+	 * blocksize <= pagesize, it is basically a noop
+	 */
+	if (end_io_wq->metadata && !bio_ready_for_csum(bio)) {
+		btrfs_queue_worker(&fs_info->endio_workers,
+				   &end_io_wq->work);
+		return;
+	}
+	error = end_io_wq->error;
+	bio->bi_private = end_io_wq->private;
+	bio->bi_end_io = end_io_wq->end_io;
+	kfree(end_io_wq);
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
+	bio_endio(bio, bio->bi_size, error);
+#else
+	bio_endio(bio, error);
+#endif
+}
+
+struct btrfs_root *open_ctree(struct super_block *sb,
+			      struct btrfs_fs_devices *fs_devices,
+			      char *options)
 {
 	u32 sectorsize;
 	u32 nodesize;
 	u32 leafsize;
 	u32 blocksize;
+	u32 stripesize;
+	struct buffer_head *bh;
 	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
 						 GFP_NOFS);
 	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
 					       GFP_NOFS);
-	struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
+	struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
+						GFP_NOFS);
+	struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root),
 						GFP_NOFS);
+	struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
+					      GFP_NOFS);
 	int ret;
-	int err = -EIO;
+	int err = -EINVAL;
+
 	struct btrfs_super_block *disk_super;
 
 	if (!extent_root || !tree_root || !fs_info) {
@@ -538,40 +1134,66 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	INIT_LIST_HEAD(&fs_info->dead_roots);
 	INIT_LIST_HEAD(&fs_info->hashers);
 	spin_lock_init(&fs_info->hash_lock);
+	spin_lock_init(&fs_info->delalloc_lock);
+	spin_lock_init(&fs_info->new_trans_lock);
 
-	memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj));
 	init_completion(&fs_info->kobj_unregister);
-	sb_set_blocksize(sb, 4096);
-	fs_info->running_transaction = NULL;
-	fs_info->last_trans_committed = 0;
 	fs_info->tree_root = tree_root;
 	fs_info->extent_root = extent_root;
+	fs_info->chunk_root = chunk_root;
+	fs_info->dev_root = dev_root;
+	fs_info->fs_devices = fs_devices;
+	INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
+	INIT_LIST_HEAD(&fs_info->space_info);
+	btrfs_mapping_init(&fs_info->mapping_tree);
+	atomic_set(&fs_info->nr_async_submits, 0);
+	atomic_set(&fs_info->throttles, 0);
 	fs_info->sb = sb;
+	fs_info->max_extent = (u64)-1;
+	fs_info->max_inline = 8192 * 1024;
+	setup_bdi(fs_info, &fs_info->bdi);
 	fs_info->btree_inode = new_inode(sb);
 	fs_info->btree_inode->i_ino = 1;
 	fs_info->btree_inode->i_nlink = 1;
-	fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
+	fs_info->thread_pool_size = min(num_online_cpus() + 2, 8);
+
+	sb->s_blocksize = 4096;
+	sb->s_blocksize_bits = blksize_bits(4096);
+
+	/*
+	 * we set the i_size on the btree inode to the max possible int.
+	 * the real end of the address space is determined by all of
+	 * the devices in the system
+	 */
+	fs_info->btree_inode->i_size = OFFSET_MAX;
 	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
-	extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
+	fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
+
+	extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
 			     fs_info->btree_inode->i_mapping,
 			     GFP_NOFS);
-	BTRFS_I(fs_info->btree_inode)->extent_tree.ops = &btree_extent_map_ops;
+	extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
+			     GFP_NOFS);
+
+	BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
 
-	extent_map_tree_init(&fs_info->free_space_cache,
+	extent_io_tree_init(&fs_info->free_space_cache,
 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
-	extent_map_tree_init(&fs_info->block_group_cache,
+	extent_io_tree_init(&fs_info->block_group_cache,
 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
-	extent_map_tree_init(&fs_info->pinned_extents,
+	extent_io_tree_init(&fs_info->pinned_extents,
 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
-	extent_map_tree_init(&fs_info->pending_del,
+	extent_io_tree_init(&fs_info->pending_del,
 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
-	extent_map_tree_init(&fs_info->extent_ins,
+	extent_io_tree_init(&fs_info->extent_ins,
 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
 	fs_info->do_barriers = 1;
-	fs_info->closing = 0;
-	fs_info->total_pinned = 0;
 
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+	INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
+#else
 	INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
+#endif
 	BTRFS_I(fs_info->btree_inode)->root = tree_root;
 	memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
 	       sizeof(struct btrfs_key));
@@ -579,7 +1201,9 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
 
 	mutex_init(&fs_info->trans_mutex);
-	mutex_init(&fs_info->fs_mutex);
+	mutex_init(&fs_info->drop_mutex);
+	mutex_init(&fs_info->alloc_mutex);
+	mutex_init(&fs_info->chunk_mutex);
 
 #if 0
 	ret = add_hasher(fs_info, "crc32c");
@@ -589,36 +1213,66 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 		goto fail_iput;
 	}
 #endif
-	__setup_root(512, 512, 512, tree_root,
+	__setup_root(4096, 4096, 4096, 4096, tree_root,
 		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
 
-	fs_info->sb_buffer = read_tree_block(tree_root,
-					     BTRFS_SUPER_INFO_OFFSET,
-					     512);
 
-	if (!fs_info->sb_buffer)
+	bh = __bread(fs_devices->latest_bdev,
+		     BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
+	if (!bh)
 		goto fail_iput;
 
-	read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
-			   sizeof(fs_info->super_copy));
+	memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
+	brelse(bh);
+
+	memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
 
-	read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
-			   (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
-			   BTRFS_FSID_SIZE);
 	disk_super = &fs_info->super_copy;
 	if (!btrfs_super_root(disk_super))
 		goto fail_sb_buffer;
 
+	err = btrfs_parse_options(tree_root, options);
+	if (err)
+		goto fail_sb_buffer;
+
+	/*
+	 * we need to start all the end_io workers up front because the
+	 * queue work function gets called at interrupt time, and so it
+	 * cannot dynamically grow.
+	 */
+	btrfs_init_workers(&fs_info->workers, fs_info->thread_pool_size);
+	btrfs_init_workers(&fs_info->submit_workers, fs_info->thread_pool_size);
+	btrfs_init_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
+	btrfs_start_workers(&fs_info->workers, 1);
+	btrfs_start_workers(&fs_info->submit_workers, 1);
+	btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
+
+
+	err = -EINVAL;
+	if (btrfs_super_num_devices(disk_super) > fs_devices->open_devices) {
+		printk("Btrfs: wanted %llu devices, but found %llu\n",
+		       (unsigned long long)btrfs_super_num_devices(disk_super),
+		       (unsigned long long)fs_devices->open_devices);
+		if (btrfs_test_opt(tree_root, DEGRADED))
+			printk("continuing in degraded mode\n");
+		else {
+			goto fail_sb_buffer;
+		}
+	}
+
+	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
+
 	nodesize = btrfs_super_nodesize(disk_super);
 	leafsize = btrfs_super_leafsize(disk_super);
 	sectorsize = btrfs_super_sectorsize(disk_super);
+	stripesize = btrfs_super_stripesize(disk_super);
 	tree_root->nodesize = nodesize;
 	tree_root->leafsize = leafsize;
 	tree_root->sectorsize = sectorsize;
-	sb_set_blocksize(sb, sectorsize);
+	tree_root->stripesize = stripesize;
 
-	i_size_write(fs_info->btree_inode,
-		     btrfs_super_total_bytes(disk_super));
+	sb->s_blocksize = sectorsize;
+	sb->s_blocksize_bits = blksize_bits(sectorsize);
 
 	if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
 		    sizeof(disk_super->magic))) {
@@ -626,53 +1280,240 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 		goto fail_sb_buffer;
 	}
 
+	mutex_lock(&fs_info->chunk_mutex);
+	ret = btrfs_read_sys_array(tree_root);
+	mutex_unlock(&fs_info->chunk_mutex);
+	if (ret) {
+		printk("btrfs: failed to read the system array on %s\n",
+		       sb->s_id);
+		goto fail_sys_array;
+	}
+
+	blocksize = btrfs_level_size(tree_root,
+				     btrfs_super_chunk_root_level(disk_super));
+
+	__setup_root(nodesize, leafsize, sectorsize, stripesize,
+		     chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
+
+	chunk_root->node = read_tree_block(chunk_root,
+					   btrfs_super_chunk_root(disk_super),
+					   blocksize, 0);
+	BUG_ON(!chunk_root->node);
+
+	read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
+	         (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
+		 BTRFS_UUID_SIZE);
+
+	mutex_lock(&fs_info->chunk_mutex);
+	ret = btrfs_read_chunk_tree(chunk_root);
+	mutex_unlock(&fs_info->chunk_mutex);
+	BUG_ON(ret);
+
+	btrfs_close_extra_devices(fs_devices);
+
 	blocksize = btrfs_level_size(tree_root,
 				     btrfs_super_root_level(disk_super));
 
+
 	tree_root->node = read_tree_block(tree_root,
 					  btrfs_super_root(disk_super),
-					  blocksize);
+					  blocksize, 0);
 	if (!tree_root->node)
 		goto fail_sb_buffer;
 
-	mutex_lock(&fs_info->fs_mutex);
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
-	if (ret) {
-		mutex_unlock(&fs_info->fs_mutex);
+	if (ret)
 		goto fail_tree_root;
-	}
+	extent_root->track_dirty = 1;
+
+	ret = find_and_setup_root(tree_root, fs_info,
+				  BTRFS_DEV_TREE_OBJECTID, dev_root);
+	dev_root->track_dirty = 1;
+
+	if (ret)
+		goto fail_extent_root;
 
 	btrfs_read_block_groups(extent_root);
 
 	fs_info->generation = btrfs_super_generation(disk_super) + 1;
-	mutex_unlock(&fs_info->fs_mutex);
+	fs_info->data_alloc_profile = (u64)-1;
+	fs_info->metadata_alloc_profile = (u64)-1;
+	fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
+
 	return tree_root;
 
+fail_extent_root:
+	free_extent_buffer(extent_root->node);
 fail_tree_root:
 	free_extent_buffer(tree_root->node);
+fail_sys_array:
 fail_sb_buffer:
-	free_extent_buffer(fs_info->sb_buffer);
+	extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
+	btrfs_stop_workers(&fs_info->workers);
+	btrfs_stop_workers(&fs_info->endio_workers);
+	btrfs_stop_workers(&fs_info->submit_workers);
 fail_iput:
 	iput(fs_info->btree_inode);
 fail:
+	btrfs_close_devices(fs_info->fs_devices);
+	btrfs_mapping_tree_free(&fs_info->mapping_tree);
+
 	kfree(extent_root);
 	kfree(tree_root);
+#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
+	bdi_destroy(&fs_info->bdi);
+#endif
 	kfree(fs_info);
 	return ERR_PTR(err);
 }
 
+static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
+{
+	char b[BDEVNAME_SIZE];
+
+	if (uptodate) {
+		set_buffer_uptodate(bh);
+	} else {
+		if (!buffer_eopnotsupp(bh) && printk_ratelimit()) {
+			printk(KERN_WARNING "lost page write due to "
+					"I/O error on %s\n",
+				       bdevname(bh->b_bdev, b));
+		}
+		/* note, we dont' set_buffer_write_io_error because we have
+		 * our own ways of dealing with the IO errors
+		 */
+		clear_buffer_uptodate(bh);
+	}
+	unlock_buffer(bh);
+	put_bh(bh);
+}
+
+int write_all_supers(struct btrfs_root *root)
+{
+	struct list_head *cur;
+	struct list_head *head = &root->fs_info->fs_devices->devices;
+	struct btrfs_device *dev;
+	struct btrfs_super_block *sb;
+	struct btrfs_dev_item *dev_item;
+	struct buffer_head *bh;
+	int ret;
+	int do_barriers;
+	int max_errors;
+	int total_errors = 0;
+	u32 crc;
+	u64 flags;
+
+	max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
+	do_barriers = !btrfs_test_opt(root, NOBARRIER);
+
+	sb = &root->fs_info->super_for_commit;
+	dev_item = &sb->dev_item;
+	list_for_each(cur, head) {
+		dev = list_entry(cur, struct btrfs_device, dev_list);
+		if (!dev->bdev) {
+			total_errors++;
+			continue;
+		}
+		if (!dev->in_fs_metadata)
+			continue;
+
+		btrfs_set_stack_device_type(dev_item, dev->type);
+		btrfs_set_stack_device_id(dev_item, dev->devid);
+		btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
+		btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
+		btrfs_set_stack_device_io_align(dev_item, dev->io_align);
+		btrfs_set_stack_device_io_width(dev_item, dev->io_width);
+		btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
+		memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
+		flags = btrfs_super_flags(sb);
+		btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
+
+
+		crc = ~(u32)0;
+		crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc,
+				      BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
+		btrfs_csum_final(crc, sb->csum);
+
+		bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096,
+			      BTRFS_SUPER_INFO_SIZE);
+
+		memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
+		dev->pending_io = bh;
+
+		get_bh(bh);
+		set_buffer_uptodate(bh);
+		lock_buffer(bh);
+		bh->b_end_io = btrfs_end_buffer_write_sync;
+
+		if (do_barriers && dev->barriers) {
+			ret = submit_bh(WRITE_BARRIER, bh);
+			if (ret == -EOPNOTSUPP) {
+				printk("btrfs: disabling barriers on dev %s\n",
+				       dev->name);
+				set_buffer_uptodate(bh);
+				dev->barriers = 0;
+				get_bh(bh);
+				lock_buffer(bh);
+				ret = submit_bh(WRITE, bh);
+			}
+		} else {
+			ret = submit_bh(WRITE, bh);
+		}
+		if (ret)
+			total_errors++;
+	}
+	if (total_errors > max_errors) {
+		printk("btrfs: %d errors while writing supers\n", total_errors);
+		BUG();
+	}
+	total_errors = 0;
+
+	list_for_each(cur, head) {
+		dev = list_entry(cur, struct btrfs_device, dev_list);
+		if (!dev->bdev)
+			continue;
+		if (!dev->in_fs_metadata)
+			continue;
+
+		BUG_ON(!dev->pending_io);
+		bh = dev->pending_io;
+		wait_on_buffer(bh);
+		if (!buffer_uptodate(dev->pending_io)) {
+			if (do_barriers && dev->barriers) {
+				printk("btrfs: disabling barriers on dev %s\n",
+				       dev->name);
+				set_buffer_uptodate(bh);
+				get_bh(bh);
+				lock_buffer(bh);
+				dev->barriers = 0;
+				ret = submit_bh(WRITE, bh);
+				BUG_ON(ret);
+				wait_on_buffer(bh);
+				if (!buffer_uptodate(bh))
+					total_errors++;
+			} else {
+				total_errors++;
+			}
+
+		}
+		dev->pending_io = NULL;
+		brelse(bh);
+	}
+	if (total_errors > max_errors) {
+		printk("btrfs: %d errors while writing supers\n", total_errors);
+		BUG();
+	}
+	return 0;
+}
+
 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root)
 {
 	int ret;
-	struct extent_buffer *super = root->fs_info->sb_buffer;
-	struct inode *btree_inode = root->fs_info->btree_inode;
 
-	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, super);
-	ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
-				     super->start, super->len);
+	ret = write_all_supers(root);
 	return ret;
 }
 
@@ -680,7 +1521,8 @@ int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
 {
 	radix_tree_delete(&fs_info->fs_roots_radix,
 			  (unsigned long)root->root_key.objectid);
-	btrfs_sysfs_del_root(root);
+	if (root->in_sysfs)
+		btrfs_sysfs_del_root(root);
 	if (root->inode)
 		iput(root->inode);
 	if (root->node)
@@ -718,8 +1560,9 @@ int close_ctree(struct btrfs_root *root)
 	struct btrfs_fs_info *fs_info = root->fs_info;
 
 	fs_info->closing = 1;
+	smp_mb();
+
 	btrfs_transaction_flush_work(root);
-	mutex_lock(&fs_info->fs_mutex);
 	btrfs_defrag_dirty_roots(root->fs_info);
 	trans = btrfs_start_transaction(root, 1);
 	ret = btrfs_commit_transaction(trans, root);
@@ -728,21 +1571,45 @@ int close_ctree(struct btrfs_root *root)
 	btrfs_commit_transaction(trans, root);
 	ret = btrfs_write_and_wait_transaction(NULL, root);
 	BUG_ON(ret);
+
 	write_ctree_super(NULL, root);
-	mutex_unlock(&fs_info->fs_mutex);
 
+	btrfs_transaction_flush_work(root);
+
+	if (fs_info->delalloc_bytes) {
+		printk("btrfs: at unmount delalloc count %Lu\n",
+		       fs_info->delalloc_bytes);
+	}
 	if (fs_info->extent_root->node)
 		free_extent_buffer(fs_info->extent_root->node);
 
 	if (fs_info->tree_root->node)
 		free_extent_buffer(fs_info->tree_root->node);
 
-	free_extent_buffer(fs_info->sb_buffer);
+	if (root->fs_info->chunk_root->node);
+		free_extent_buffer(root->fs_info->chunk_root->node);
+
+	if (root->fs_info->dev_root->node);
+		free_extent_buffer(root->fs_info->dev_root->node);
 
 	btrfs_free_block_groups(root->fs_info);
 	del_fs_roots(fs_info);
-	extent_map_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->extent_tree);
+
+	filemap_write_and_wait(fs_info->btree_inode->i_mapping);
+
+	extent_io_tree_empty_lru(&fs_info->free_space_cache);
+	extent_io_tree_empty_lru(&fs_info->block_group_cache);
+	extent_io_tree_empty_lru(&fs_info->pinned_extents);
+	extent_io_tree_empty_lru(&fs_info->pending_del);
+	extent_io_tree_empty_lru(&fs_info->extent_ins);
+	extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
+
 	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
+
+	btrfs_stop_workers(&fs_info->workers);
+	btrfs_stop_workers(&fs_info->endio_workers);
+	btrfs_stop_workers(&fs_info->submit_workers);
+
 	iput(fs_info->btree_inode);
 #if 0
 	while(!list_empty(&fs_info->hashers)) {
@@ -754,21 +1621,38 @@ int close_ctree(struct btrfs_root *root)
 		kfree(hasher);
 	}
 #endif
+	btrfs_close_devices(fs_info->fs_devices);
+	btrfs_mapping_tree_free(&fs_info->mapping_tree);
+
+#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
+	bdi_destroy(&fs_info->bdi);
+#endif
+
 	kfree(fs_info->extent_root);
 	kfree(fs_info->tree_root);
+	kfree(fs_info->chunk_root);
+	kfree(fs_info->dev_root);
 	return 0;
 }
 
-int btrfs_buffer_uptodate(struct extent_buffer *buf)
+int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
 {
+	int ret;
 	struct inode *btree_inode = buf->first_page->mapping->host;
-	return extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree, buf);
+
+	ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
+	if (!ret)
+		return ret;
+
+	ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
+				    parent_transid);
+	return !ret;
 }
 
 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
 {
 	struct inode *btree_inode = buf->first_page->mapping->host;
-	return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree,
+	return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
 					  buf);
 }
 
@@ -778,26 +1662,61 @@ void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
 	u64 transid = btrfs_header_generation(buf);
 	struct inode *btree_inode = root->fs_info->btree_inode;
 
+	WARN_ON(!btrfs_tree_locked(buf));
 	if (transid != root->fs_info->generation) {
 		printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
 			(unsigned long long)buf->start,
 			transid, root->fs_info->generation);
 		WARN_ON(1);
 	}
-	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
+	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
+}
+
+void btrfs_throttle(struct btrfs_root *root)
+{
+	struct backing_dev_info *bdi;
+
+	bdi = &root->fs_info->bdi;
+	if (atomic_read(&root->fs_info->throttles) &&
+	    bdi_write_congested(bdi)) {
+#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
+		congestion_wait(WRITE, HZ/20);
+#else
+		blk_congestion_wait(WRITE, HZ/20);
+#endif
+
+	}
 }
 
 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
 {
-	balance_dirty_pages_ratelimited_nr(
-			root->fs_info->btree_inode->i_mapping, 1);
+	/*
+	 * looks as though older kernels can get into trouble with
+	 * this code, they end up stuck in balance_dirty_pages forever
+	 */
+	struct extent_io_tree *tree;
+	u64 num_dirty;
+	u64 start = 0;
+	unsigned long thresh = 16 * 1024 * 1024;
+	tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
+
+	if (current_is_pdflush())
+		return;
+
+	num_dirty = count_range_bits(tree, &start, (u64)-1,
+				     thresh, EXTENT_DIRTY);
+	if (num_dirty > thresh) {
+		balance_dirty_pages_ratelimited_nr(
+				   root->fs_info->btree_inode->i_mapping, 1);
+	}
+	return;
 }
 
 void btrfs_set_buffer_defrag(struct extent_buffer *buf)
 {
 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
 	struct inode *btree_inode = root->fs_info->btree_inode;
-	set_extent_bits(&BTRFS_I(btree_inode)->extent_tree, buf->start,
+	set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
 			buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS);
 }
 
@@ -805,7 +1724,7 @@ void btrfs_set_buffer_defrag_done(struct extent_buffer *buf)
 {
 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
 	struct inode *btree_inode = root->fs_info->btree_inode;
-	set_extent_bits(&BTRFS_I(btree_inode)->extent_tree, buf->start,
+	set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
 			buf->start + buf->len - 1, EXTENT_DEFRAG_DONE,
 			GFP_NOFS);
 }
@@ -814,7 +1733,7 @@ int btrfs_buffer_defrag(struct extent_buffer *buf)
 {
 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
 	struct inode *btree_inode = root->fs_info->btree_inode;
-	return test_range_bit(&BTRFS_I(btree_inode)->extent_tree,
+	return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
 		     buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0);
 }
 
@@ -822,7 +1741,7 @@ int btrfs_buffer_defrag_done(struct extent_buffer *buf)
 {
 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
 	struct inode *btree_inode = root->fs_info->btree_inode;
-	return test_range_bit(&BTRFS_I(btree_inode)->extent_tree,
+	return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
 		     buf->start, buf->start + buf->len - 1,
 		     EXTENT_DEFRAG_DONE, 0);
 }
@@ -831,7 +1750,7 @@ int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf)
 {
 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
 	struct inode *btree_inode = root->fs_info->btree_inode;
-	return clear_extent_bits(&BTRFS_I(btree_inode)->extent_tree,
+	return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
 		     buf->start, buf->start + buf->len - 1,
 		     EXTENT_DEFRAG_DONE, GFP_NOFS);
 }
@@ -840,19 +1759,26 @@ int btrfs_clear_buffer_defrag(struct extent_buffer *buf)
 {
 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
 	struct inode *btree_inode = root->fs_info->btree_inode;
-	return clear_extent_bits(&BTRFS_I(btree_inode)->extent_tree,
+	return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
 		     buf->start, buf->start + buf->len - 1,
 		     EXTENT_DEFRAG, GFP_NOFS);
 }
 
-int btrfs_read_buffer(struct extent_buffer *buf)
+int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
 {
 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	return read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
-					buf, 0, 1);
+	int ret;
+	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
+	if (ret == 0) {
+		buf->flags |= EXTENT_UPTODATE;
+	}
+	return ret;
 }
 
-static struct extent_map_ops btree_extent_map_ops = {
+static struct extent_io_ops btree_extent_io_ops = {
 	.writepage_io_hook = btree_writepage_io_hook,
+	.readpage_end_io_hook = btree_readpage_end_io_hook,
+	.submit_bio_hook = btree_submit_bio_hook,
+	/* note we're sharing with inode.c for the merge bio hook */
+	.merge_bio_hook = btrfs_merge_bio_hook,
 };