Btrfs: Extent based page cache code. This uses an rbtree of extents and tests

instead of buffer heads.

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

9 files changed, 2203 insertions, 669 deletions

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 9321438c5eef..ea7a22870350 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -4,7 +4,8 @@ ifneq ($(KERNELRELEASE),)
 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
+           transaction.o bit-radix.o inode.o file.o tree-defrag.o \
+           extent_map.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 \
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index f446f161f736..d1d5af471c3c 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -19,11 +19,14 @@
 #ifndef __BTRFS_I__
 #define __BTRFS_I__
 
+#include "extent_map.h"
+
 /* in memory btrfs inode */
 struct btrfs_inode {
         struct btrfs_root *root;
         struct btrfs_block_group_cache *block_group;
         struct btrfs_key location;
+        struct extent_map_tree extent_tree;
         struct inode vfs_inode;
 
         /*
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index c7f5161271d3..947f061ed118 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1176,7 +1176,14 @@ int btrfs_commit_write(struct file *file, struct page *page,
                        unsigned from, unsigned to);
 int btrfs_get_block(struct inode *inode, sector_t iblock,
                     struct buffer_head *result, int create);
+struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
+                                    size_t page_offset, u64 start, u64 end,
+                                    int create);
+int btrfs_update_inode(struct btrfs_trans_handle *trans,
+                              struct btrfs_root *root,
+                              struct inode *inode);
 /* file.c */
+int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end);
 extern struct file_operations btrfs_file_operations;
 int btrfs_drop_extents(struct btrfs_trans_handle *trans,
                        struct btrfs_root *root, struct inode *inode,
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 81fd18cbd824..5261733b8735 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -77,7 +77,6 @@ int btrfs_insert_dev_radix(struct btrfs_root *root,
                            u64 num_blocks);
 int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
                              u64 logical);
-int btrfs_releasepage(struct page *page, gfp_t flags);
 void btrfs_btree_balance_dirty(struct btrfs_root *root);
 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root);
 void btrfs_mark_buffer_dirty(struct buffer_head *bh);
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
new file mode 100644
index 000000000000..d378edf0964e
--- /dev/null
+++ b/fs/btrfs/extent_map.c
@@ -0,0 +1,1656 @@
+#include <linux/bitops.h>
+#include <linux/slab.h>
+#include <linux/bio.h>
+#include <linux/mm.h>
+#include <linux/gfp.h>
+#include <linux/pagemap.h>
+#include <linux/page-flags.h>
+#include <linux/module.h>
+#include <linux/spinlock.h>
+#include <linux/blkdev.h>
+#include "extent_map.h"
+
+static struct kmem_cache *extent_map_cache;
+static struct kmem_cache *extent_state_cache;
+
+struct tree_entry {
+        u64 start;
+        u64 end;
+        int in_tree;
+        struct rb_node rb_node;
+};
+
+/* bits for the extent state */
+#define EXTENT_DIRTY 1
+#define EXTENT_WRITEBACK (1 << 1)
+#define EXTENT_UPTODATE (1 << 2)
+#define EXTENT_LOCKED (1 << 3)
+#define EXTENT_NEW (1 << 4)
+#define EXTENT_DELALLOC (1 << 5)
+
+#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
+
+static LIST_HEAD(all_states);
+spinlock_t state_lock = SPIN_LOCK_UNLOCKED;
+
+void __init extent_map_init(void)
+{
+        extent_map_cache = kmem_cache_create("extent_map",
+                                            sizeof(struct extent_map), 0,
+                                            SLAB_RECLAIM_ACCOUNT |
+                                            SLAB_DESTROY_BY_RCU,
+                                            NULL);
+        extent_state_cache = kmem_cache_create("extent_state",
+                                            sizeof(struct extent_state), 0,
+                                            SLAB_RECLAIM_ACCOUNT |
+                                            SLAB_DESTROY_BY_RCU,
+                                            NULL);
+}
+
+void __exit extent_map_exit(void)
+{
+        while(!list_empty(&all_states)) {
+                struct extent_state *state;
+                struct list_head *cur = all_states.next;
+                state = list_entry(cur, struct extent_state, list);
+                printk("found leaked state %Lu %Lu state %d in_tree %d\n",
+                       state->start, state->end, state->state, state->in_tree);
+                list_del(&state->list);
+                kfree(state);
+        }
+        if (extent_map_cache)
+                kmem_cache_destroy(extent_map_cache);
+        if (extent_state_cache)
+                kmem_cache_destroy(extent_state_cache);
+}
+
+void extent_map_tree_init(struct extent_map_tree *tree,
+                          struct address_space *mapping, gfp_t mask)
+{
+        tree->map.rb_node = NULL;
+        tree->state.rb_node = NULL;
+        rwlock_init(&tree->lock);
+        tree->mapping = mapping;
+}
+EXPORT_SYMBOL(extent_map_tree_init);
+
+struct extent_map *alloc_extent_map(gfp_t mask)
+{
+        struct extent_map *em;
+        em = kmem_cache_alloc(extent_map_cache, mask);
+        if (!em || IS_ERR(em))
+                return em;
+        em->in_tree = 0;
+        atomic_set(&em->refs, 1);
+        return em;
+}
+EXPORT_SYMBOL(alloc_extent_map);
+
+void free_extent_map(struct extent_map *em)
+{
+        if (atomic_dec_and_test(&em->refs)) {
+                WARN_ON(em->in_tree);
+                kmem_cache_free(extent_map_cache, em);
+        }
+}
+EXPORT_SYMBOL(free_extent_map);
+
+
+struct extent_state *alloc_extent_state(gfp_t mask)
+{
+        struct extent_state *state;
+        state = kmem_cache_alloc(extent_state_cache, mask);
+        if (!state || IS_ERR(state))
+                return state;
+        state->state = 0;
+        state->in_tree = 0;
+        atomic_set(&state->refs, 1);
+        init_waitqueue_head(&state->wq);
+        spin_lock_irq(&state_lock);
+        list_add(&state->list, &all_states);
+        spin_unlock_irq(&state_lock);
+        return state;
+}
+EXPORT_SYMBOL(alloc_extent_state);
+
+void free_extent_state(struct extent_state *state)
+{
+        if (atomic_dec_and_test(&state->refs)) {
+                WARN_ON(state->in_tree);
+                spin_lock_irq(&state_lock);
+                list_del_init(&state->list);
+                spin_unlock_irq(&state_lock);
+                kmem_cache_free(extent_state_cache, state);
+        }
+}
+EXPORT_SYMBOL(free_extent_state);
+
+static struct rb_node *tree_insert(struct rb_root *root, u64 offset,
+                                   struct rb_node *node)
+{
+        struct rb_node ** p = &root->rb_node;
+        struct rb_node * parent = NULL;
+        struct tree_entry *entry;
+
+        while(*p) {
+                parent = *p;
+                entry = rb_entry(parent, struct tree_entry, rb_node);
+
+                if (offset < entry->start)
+                        p = &(*p)->rb_left;
+                else if (offset > entry->end)
+                        p = &(*p)->rb_right;
+                else
+                        return parent;
+        }
+
+        entry = rb_entry(node, struct tree_entry, rb_node);
+        entry->in_tree = 1;
+        rb_link_node(node, parent, p);
+        rb_insert_color(node, root);
+        return NULL;
+}
+
+static struct rb_node *__tree_search(struct rb_root *root, u64 offset,
+                                   struct rb_node **prev_ret)
+{
+        struct rb_node * n = root->rb_node;
+        struct rb_node *prev = NULL;
+        struct tree_entry *entry;
+        struct tree_entry *prev_entry = NULL;
+
+        while(n) {
+                entry = rb_entry(n, struct tree_entry, rb_node);
+                prev = n;
+                prev_entry = entry;
+
+                if (offset < entry->start)
+                        n = n->rb_left;
+                else if (offset > entry->end)
+                        n = n->rb_right;
+                else
+                        return n;
+        }
+        if (!prev_ret)
+                return NULL;
+        while(prev && offset > prev_entry->end) {
+                prev = rb_next(prev);
+                prev_entry = rb_entry(prev, struct tree_entry, rb_node);
+        }
+        *prev_ret = prev;
+        return NULL;
+}
+
+static inline struct rb_node *tree_search(struct rb_root *root, u64 offset)
+{
+        struct rb_node *prev;
+        struct rb_node *ret;
+        ret = __tree_search(root, offset, &prev);
+        if (!ret)
+                return prev;
+        return ret;
+}
+
+static int tree_delete(struct rb_root *root, u64 offset)
+{
+        struct rb_node *node;
+        struct tree_entry *entry;
+
+        node = __tree_search(root, offset, NULL);
+        if (!node)
+                return -ENOENT;
+        entry = rb_entry(node, struct tree_entry, rb_node);
+        entry->in_tree = 0;
+        rb_erase(node, root);
+        return 0;
+}
+
+/*
+ * add_extent_mapping tries a simple backward merge with existing
+ * mappings.  The extent_map struct passed in will be inserted into
+ * the tree directly (no copies made, just a reference taken).
+ */
+int add_extent_mapping(struct extent_map_tree *tree,
+                       struct extent_map *em)
+{
+        int ret = 0;
+        struct extent_map *prev = NULL;
+        struct rb_node *rb;
+
+        write_lock_irq(&tree->lock);
+        rb = tree_insert(&tree->map, em->end, &em->rb_node);
+        if (rb) {
+                prev = rb_entry(rb, struct extent_map, rb_node);
+                printk("found extent map %Lu %Lu on insert of %Lu %Lu\n", prev->start, prev->end, em->start, em->end);
+                ret = -EEXIST;
+                goto out;
+        }
+        atomic_inc(&em->refs);
+        if (em->start != 0) {
+                rb = rb_prev(&em->rb_node);
+                if (rb)
+                        prev = rb_entry(rb, struct extent_map, rb_node);
+                if (prev && prev->end + 1 == em->start &&
+                    ((em->block_start == 0 && prev->block_start == 0) ||
+                             (em->block_start == prev->block_end + 1))) {
+                        em->start = prev->start;
+                        em->block_start = prev->block_start;
+                        rb_erase(&prev->rb_node, &tree->map);
+                        prev->in_tree = 0;
+                        free_extent_map(prev);
+                }
+         }
+out:
+        write_unlock_irq(&tree->lock);
+        return ret;
+}
+EXPORT_SYMBOL(add_extent_mapping);
+
+/*
+ * lookup_extent_mapping returns the first extent_map struct in the
+ * tree that intersects the [start, end] (inclusive) range.  There may
+ * be additional objects in the tree that intersect, so check the object
+ * returned carefully to make sure you don't need additional lookups.
+ */
+struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree,
+                                         u64 start, u64 end)
+{
+        struct extent_map *em;
+        struct rb_node *rb_node;
+
+        read_lock_irq(&tree->lock);
+        rb_node = tree_search(&tree->map, start);
+        if (!rb_node) {
+                em = NULL;
+                goto out;
+        }
+        if (IS_ERR(rb_node)) {
+                em = ERR_PTR(PTR_ERR(rb_node));
+                goto out;
+        }
+        em = rb_entry(rb_node, struct extent_map, rb_node);
+        if (em->end < start || em->start > end) {
+                em = NULL;
+                goto out;
+        }
+        atomic_inc(&em->refs);
+out:
+        read_unlock_irq(&tree->lock);
+        return em;
+}
+EXPORT_SYMBOL(lookup_extent_mapping);
+
+/*
+ * removes an extent_map struct from the tree.  No reference counts are
+ * dropped, and no checks are done to  see if the range is in use
+ */
+int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em)
+{
+        int ret;
+
+        write_lock_irq(&tree->lock);
+        ret = tree_delete(&tree->map, em->end);
+        write_unlock_irq(&tree->lock);
+        return ret;
+}
+EXPORT_SYMBOL(remove_extent_mapping);
+
+/*
+ * utility function to look for merge candidates inside a given range.
+ * Any extents with matching state are merged together into a single
+ * extent in the tree.  Extents with EXTENT_IO in their state field
+ * are not merged because the end_io handlers need to be able to do
+ * operations on them without sleeping (or doing allocations/splits).
+ *
+ * This should be called with the tree lock held.
+ */
+static int merge_state(struct extent_map_tree *tree,
+                       struct extent_state *state)
+{
+        struct extent_state *other;
+        struct rb_node *other_node;
+
+        if (state->state & EXTENT_IOBITS)
+                return 0;
+
+        other_node = rb_prev(&state->rb_node);
+        if (other_node) {
+                other = rb_entry(other_node, struct extent_state, rb_node);
+                if (other->end == state->start - 1 &&
+                    other->state == state->state) {
+                        state->start = other->start;
+                        other->in_tree = 0;
+                        rb_erase(&other->rb_node, &tree->state);
+                        free_extent_state(other);
+                }
+        }
+        other_node = rb_next(&state->rb_node);
+        if (other_node) {
+                other = rb_entry(other_node, struct extent_state, rb_node);
+                if (other->start == state->end + 1 &&
+                    other->state == state->state) {
+                        other->start = state->start;
+                        state->in_tree = 0;
+                        rb_erase(&state->rb_node, &tree->state);
+                        free_extent_state(state);
+                }
+        }
+        return 0;
+}
+
+/*
+ * insert an extent_state struct into the tree.  'bits' are set on the
+ * struct before it is inserted.
+ *
+ * This may return -EEXIST if the extent is already there, in which case the
+ * state struct is freed.
+ *
+ * The tree lock is not taken internally.  This is a utility function and
+ * probably isn't what you want to call (see set/clear_extent_bit).
+ */
+static int insert_state(struct extent_map_tree *tree,
+                        struct extent_state *state, u64 start, u64 end,
+                        int bits)
+{
+        struct rb_node *node;
+
+        if (end < start) {
+                printk("end < start %Lu %Lu\n", end, start);
+                WARN_ON(1);
+        }
+        state->state |= bits;
+        state->start = start;
+        state->end = end;
+        if ((end & 4095) == 0) {
+                printk("insert state %Lu %Lu strange end\n", start, end);
+                WARN_ON(1);
+        }
+        node = tree_insert(&tree->state, end, &state->rb_node);
+        if (node) {
+                struct extent_state *found;
+                found = rb_entry(node, struct extent_state, rb_node);
+printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, start, end);
+                free_extent_state(state);
+                return -EEXIST;
+        }
+        merge_state(tree, state);
+        return 0;
+}
+
+/*
+ * split a given extent state struct in two, inserting the preallocated
+ * struct 'prealloc' as the newly created second half.  'split' indicates an
+ * offset inside 'orig' where it should be split.
+ *
+ * Before calling,
+ * the tree has 'orig' at [orig->start, orig->end].  After calling, there
+ * are two extent state structs in the tree:
+ * prealloc: [orig->start, split - 1]
+ * orig: [ split, orig->end ]
+ *
+ * The tree locks are not taken by this function. They need to be held
+ * by the caller.
+ */
+static int split_state(struct extent_map_tree *tree, struct extent_state *orig,
+                       struct extent_state *prealloc, u64 split)
+{
+        struct rb_node *node;
+        prealloc->start = orig->start;
+        prealloc->end = split - 1;
+        prealloc->state = orig->state;
+        orig->start = split;
+        if ((prealloc->end & 4095) == 0) {
+                printk("insert state %Lu %Lu strange end\n", prealloc->start,
+                       prealloc->end);
+                WARN_ON(1);
+        }
+        node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node);
+        if (node) {
+                struct extent_state *found;
+                found = rb_entry(node, struct extent_state, rb_node);
+printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, prealloc->start, prealloc->end);
+                free_extent_state(prealloc);
+                return -EEXIST;
+        }
+        return 0;
+}
+
+/*
+ * utility function to clear some bits in an extent state struct.
+ * it will optionally wake up any one waiting on this state (wake == 1), or
+ * forcibly remove the state from the tree (delete == 1).
+ *
+ * If no bits are set on the state struct after clearing things, the
+ * struct is freed and removed from the tree
+ */
+static int clear_state_bit(struct extent_map_tree *tree,
+                            struct extent_state *state, int bits, int wake,
+                            int delete)
+{
+        int ret = state->state & bits;
+        state->state &= ~bits;
+        if (wake)
+                wake_up(&state->wq);
+        if (delete || state->state == 0) {
+                if (state->in_tree) {
+                        rb_erase(&state->rb_node, &tree->state);
+                        state->in_tree = 0;
+                        free_extent_state(state);
+                } else {
+                        WARN_ON(1);
+                }
+        } else {
+                merge_state(tree, state);
+        }
+        return ret;
+}
+
+/*
+ * clear some bits on a range in the tree.  This may require splitting
+ * or inserting elements in the tree, so the gfp mask is used to
+ * indicate which allocations or sleeping are allowed.
+ *
+ * pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove
+ * the given range from the tree regardless of state (ie for truncate).
+ *
+ * the range [start, end] is inclusive.
+ *
+ * This takes the tree lock, and returns < 0 on error, > 0 if any of the
+ * bits were already set, or zero if none of the bits were already set.
+ */
+int clear_extent_bit(struct extent_map_tree *tree, u64 start, u64 end,
+                     int bits, int wake, int delete, gfp_t mask)
+{
+        struct extent_state *state;
+        struct extent_state *prealloc = NULL;
+        struct rb_node *node;
+        int err;
+        int set = 0;
+
+again:
+        if (!prealloc && (mask & __GFP_WAIT)) {
+                prealloc = alloc_extent_state(mask);
+                if (!prealloc)
+                        return -ENOMEM;
+        }
+
+        write_lock_irq(&tree->lock);
+        /*
+         * this search will find the extents that end after
+         * our range starts
+         */
+        node = tree_search(&tree->state, start);
+        if (!node)
+                goto out;
+        state = rb_entry(node, struct extent_state, rb_node);
+        if (state->start > end)
+                goto out;
+        WARN_ON(state->end < start);
+
+        /*
+         *     | ---- desired range ---- |
+         *  | state | or
+         *  | ------------- state -------------- |
+         *
+         * We need to split the extent we found, and may flip
+         * bits on second half.
+         *
+         * If the extent we found extends past our range, we
+         * just split and search again.  It'll get split again
+         * the next time though.
+         *
+         * If the extent we found is inside our range, we clear
+         * the desired bit on it.
+         */
+
+        if (state->start < start) {
+                err = split_state(tree, state, prealloc, start);
+                BUG_ON(err == -EEXIST);
+                prealloc = NULL;
+                if (err)
+                        goto out;
+                if (state->end <= end) {
+                        start = state->end + 1;
+                        set |= clear_state_bit(tree, state, bits,
+                                        wake, delete);
+                } else {
+                        start = state->start;
+                }
+                goto search_again;
+        }
+        /*
+         * | ---- desired range ---- |
+         *                        | state |
+         * We need to split the extent, and clear the bit
+         * on the first half
+         */
+        if (state->start <= end && state->end > end) {
+                err = split_state(tree, state, prealloc, end + 1);
+                BUG_ON(err == -EEXIST);
+
+                if (wake)
+                        wake_up(&state->wq);
+                set |= clear_state_bit(tree, prealloc, bits,
+                                       wake, delete);
+                prealloc = NULL;
+                goto out;
+        }
+
+        start = state->end + 1;
+        set |= clear_state_bit(tree, state, bits, wake, delete);
+        goto search_again;
+
+out:
+        write_unlock_irq(&tree->lock);
+        if (prealloc)
+                free_extent_state(prealloc);
+
+        return set;
+
+search_again:
+        if (start >= end)
+                goto out;
+        write_unlock_irq(&tree->lock);
+        if (mask & __GFP_WAIT)
+                cond_resched();
+        goto again;
+}
+EXPORT_SYMBOL(clear_extent_bit);
+
+static int wait_on_state(struct extent_map_tree *tree,
+                         struct extent_state *state)
+{
+        DEFINE_WAIT(wait);
+        prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
+        read_unlock_irq(&tree->lock);
+        schedule();
+        read_lock_irq(&tree->lock);
+        finish_wait(&state->wq, &wait);
+        return 0;
+}
+
+/*
+ * waits for one or more bits to clear on a range in the state tree.
+ * The range [start, end] is inclusive.
+ * The tree lock is taken by this function
+ */
+int wait_extent_bit(struct extent_map_tree *tree, u64 start, u64 end, int bits)
+{
+        struct extent_state *state;
+        struct rb_node *node;
+
+        read_lock_irq(&tree->lock);
+again:
+        while (1) {
+                /*
+                 * this search will find all the extents that end after
+                 * our range starts
+                 */
+                node = tree_search(&tree->state, start);
+                if (!node)
+                        break;
+
+                state = rb_entry(node, struct extent_state, rb_node);
+
+                if (state->start > end)
+                        goto out;
+
+                if (state->state & bits) {
+                        start = state->start;
+                        atomic_inc(&state->refs);
+                        wait_on_state(tree, state);
+                        free_extent_state(state);
+                        goto again;
+                }
+                start = state->end + 1;
+
+                if (start > end)
+                        break;
+
+                if (need_resched()) {
+                        read_unlock_irq(&tree->lock);
+                        cond_resched();
+                        read_lock_irq(&tree->lock);
+                }
+        }
+out:
+        read_unlock_irq(&tree->lock);
+        return 0;
+}
+EXPORT_SYMBOL(wait_extent_bit);
+
+/*
+ * set some bits on a range in the tree.  This may require allocations
+ * or sleeping, so the gfp mask is used to indicate what is allowed.
+ *
+ * If 'exclusive' == 1, this will fail with -EEXIST if some part of the
+ * range already has the desired bits set.  The start of the existing
+ * range is returned in failed_start in this case.
+ *
+ * [start, end] is inclusive
+ * This takes the tree lock.
+ */
+int set_extent_bit(struct extent_map_tree *tree, u64 start, u64 end, int bits,
+                   int exclusive, u64 *failed_start, gfp_t mask)
+{
+        struct extent_state *state;
+        struct extent_state *prealloc = NULL;
+        struct rb_node *node;
+        int err = 0;
+        int set;
+        u64 last_start;
+        u64 last_end;
+again:
+        if (!prealloc && (mask & __GFP_WAIT)) {
+                prealloc = alloc_extent_state(mask);
+                if (!prealloc)
+                        return -ENOMEM;
+        }
+
+        write_lock_irq(&tree->lock);
+        /*
+         * this search will find all the extents that end after
+         * our range starts.
+         */
+        node = tree_search(&tree->state, start);
+        if (!node) {
+                err = insert_state(tree, prealloc, start, end, bits);
+                prealloc = NULL;
+                BUG_ON(err == -EEXIST);
+                goto out;
+        }
+
+        state = rb_entry(node, struct extent_state, rb_node);
+        last_start = state->start;
+        last_end = state->end;
+
+        /*
+         * | ---- desired range ---- |
+         * | state |
+         *
+         * Just lock what we found and keep going
+         */
+        if (state->start == start && state->end <= end) {
+                set = state->state & bits;
+                if (set && exclusive) {
+                        *failed_start = state->start;
+                        err = -EEXIST;
+                        goto out;
+                }
+                state->state |= bits;
+                start = state->end + 1;
+                merge_state(tree, state);
+                goto search_again;
+        }
+
+        /*
+         *     | ---- desired range ---- |
+         * | state |
+         *   or
+         * | ------------- state -------------- |
+         *
+         * We need to split the extent we found, and may flip bits on
+         * second half.
+         *
+         * If the extent we found extends past our
+         * range, we just split and search again.  It'll get split
+         * again the next time though.
+         *
+         * If the extent we found is inside our range, we set the
+         * desired bit on it.
+         */
+        if (state->start < start) {
+                set = state->state & bits;
+                if (exclusive && set) {
+                        *failed_start = start;
+                        err = -EEXIST;
+                        goto out;
+                }
+                err = split_state(tree, state, prealloc, start);
+                BUG_ON(err == -EEXIST);
+                prealloc = NULL;
+                if (err)
+                        goto out;
+                if (state->end <= end) {
+                        state->state |= bits;
+                        start = state->end + 1;
+                        merge_state(tree, state);
+                } else {
+                        start = state->start;
+                }
+                goto search_again;
+        }
+        /*
+         * | ---- desired range ---- |
+         *                        | state |
+         * We need to split the extent, and set the bit
+         * on the first half
+         */
+        if (state->start <= end && state->end > end) {
+                set = state->state & bits;
+                if (exclusive && set) {
+                        *failed_start = start;
+                        err = -EEXIST;
+                        goto out;
+                }
+                err = split_state(tree, state, prealloc, end + 1);
+                BUG_ON(err == -EEXIST);
+
+                prealloc->state |= bits;
+                merge_state(tree, prealloc);
+                prealloc = NULL;
+                goto out;
+        }
+
+        /*
+         * | ---- desired range ---- |
+         *     | state | or               | state |
+         *
+         * There's a hole, we need to insert something in it and
+         * ignore the extent we found.
+         */
+        if (state->start > start) {
+                u64 this_end;
+                if (end < last_start)
+                        this_end = end;
+                else
+                        this_end = last_start -1;
+                err = insert_state(tree, prealloc, start, this_end,
+                                   bits);
+                prealloc = NULL;
+                BUG_ON(err == -EEXIST);
+                if (err)
+                        goto out;
+                start = this_end + 1;
+                goto search_again;
+        }
+        goto search_again;
+
+out:
+        write_unlock_irq(&tree->lock);
+        if (prealloc)
+                free_extent_state(prealloc);
+
+        return err;
+
+search_again:
+        if (start > end)
+                goto out;
+        write_unlock_irq(&tree->lock);
+        if (mask & __GFP_WAIT)
+                cond_resched();
+        goto again;
+}
+EXPORT_SYMBOL(set_extent_bit);
+
+/* wrappers around set/clear extent bit */
+int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end,
+                     gfp_t mask)
+{
+        return set_extent_bit(tree, start, end, EXTENT_DIRTY, 0, NULL,
+                              mask);
+}
+EXPORT_SYMBOL(set_extent_dirty);
+
+int clear_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end,
+                       gfp_t mask)
+{
+        return clear_extent_bit(tree, start, end, EXTENT_DIRTY, 0, 0, mask);
+}
+EXPORT_SYMBOL(clear_extent_dirty);
+
+int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end,
+                     gfp_t mask)
+{
+        return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL,
+                              mask);
+}
+EXPORT_SYMBOL(set_extent_new);
+
+int clear_extent_new(struct extent_map_tree *tree, u64 start, u64 end,
+                       gfp_t mask)
+{
+        return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0, mask);
+}
+EXPORT_SYMBOL(clear_extent_new);
+
+int set_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end,
+                        gfp_t mask)
+{
+        return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL,
+                              mask);
+}
+EXPORT_SYMBOL(set_extent_uptodate);
+
+int clear_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end,
+                          gfp_t mask)
+{
+        return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, mask);
+}
+EXPORT_SYMBOL(clear_extent_uptodate);
+
+int set_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end,
+                         gfp_t mask)
+{
+        return set_extent_bit(tree, start, end, EXTENT_WRITEBACK,
+                              0, NULL, mask);
+}
+EXPORT_SYMBOL(set_extent_writeback);
+
+int clear_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end,
+                           gfp_t mask)
+{
+        return clear_extent_bit(tree, start, end, EXTENT_WRITEBACK, 1, 0, mask);
+}
+EXPORT_SYMBOL(clear_extent_writeback);
+
+int wait_on_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end)
+{
+        return wait_extent_bit(tree, start, end, EXTENT_WRITEBACK);
+}
+EXPORT_SYMBOL(wait_on_extent_writeback);
+
+/*
+ * locks a range in ascending order, waiting for any locked regions
+ * it hits on the way.  [start,end] are inclusive, and this will sleep.
+ */
+int lock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask)
+{
+        int err;
+        u64 failed_start;
+        while (1) {
+                err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1,
+                                     &failed_start, mask);
+                if (err == -EEXIST && (mask & __GFP_WAIT)) {
+                        wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
+                        start = failed_start;
+                } else {
+                        break;
+                }
+                WARN_ON(start > end);
+        }
+        return err;
+}
+EXPORT_SYMBOL(lock_extent);
+
+int unlock_extent(struct extent_map_tree *tree, u64 start, u64 end,
+                  gfp_t mask)
+{
+        return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, mask);
+}
+EXPORT_SYMBOL(unlock_extent);
+
+/*
+ * helper function to set pages and extents in the tree dirty
+ */
+int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end)
+{
+        unsigned long index = start >> PAGE_CACHE_SHIFT;
+        unsigned long end_index = end >> PAGE_CACHE_SHIFT;
+        struct page *page;
+
+        while (index <= end_index) {
+                page = find_get_page(tree->mapping, index);
+                BUG_ON(!page);
+                __set_page_dirty_nobuffers(page);
+                page_cache_release(page);
+                index++;
+        }
+        set_extent_dirty(tree, start, end, GFP_NOFS);
+        return 0;
+}
+EXPORT_SYMBOL(set_range_dirty);
+
+/*
+ * helper function to set both pages and extents in the tree writeback
+ */
+int set_range_writeback(struct extent_map_tree *tree, u64 start, u64 end)
+{
+        unsigned long index = start >> PAGE_CACHE_SHIFT;
+        unsigned long end_index = end >> PAGE_CACHE_SHIFT;
+        struct page *page;
+
+        while (index <= end_index) {
+                page = find_get_page(tree->mapping, index);
+                BUG_ON(!page);
+                set_page_writeback(page);
+                page_cache_release(page);
+                index++;
+        }
+        set_extent_writeback(tree, start, end, GFP_NOFS);
+        return 0;
+}
+EXPORT_SYMBOL(set_range_writeback);
+
+/*
+ * helper function to lock both pages and extents in the tree.
+ * pages must be locked first.
+ */
+int lock_range(struct extent_map_tree *tree, u64 start, u64 end)
+{
+        unsigned long index = start >> PAGE_CACHE_SHIFT;
+        unsigned long end_index = end >> PAGE_CACHE_SHIFT;
+        struct page *page;
+        int err;
+
+        while (index <= end_index) {
+                page = grab_cache_page(tree->mapping, index);
+                if (!page) {
+                        err = -ENOMEM;
+                        goto failed;
+                }
+                if (IS_ERR(page)) {
+                        err = PTR_ERR(page);
+                        goto failed;
+                }
+                index++;
+        }
+        lock_extent(tree, start, end, GFP_NOFS);
+        return 0;
+
+failed:
+        /*
+         * we failed above in getting the page at 'index', so we undo here
+         * up to but not including the page at 'index'
+         */
+        end_index = index;
+        index = start >> PAGE_CACHE_SHIFT;
+        while (index < end_index) {
+                page = find_get_page(tree->mapping, index);
+                unlock_page(page);
+                page_cache_release(page);
+                index++;
+        }
+        return err;
+}
+EXPORT_SYMBOL(lock_range);
+
+/*
+ * helper function to unlock both pages and extents in the tree.
+ */
+int unlock_range(struct extent_map_tree *tree, u64 start, u64 end)
+{
+        unsigned long index = start >> PAGE_CACHE_SHIFT;
+        unsigned long end_index = end >> PAGE_CACHE_SHIFT;
+        struct page *page;
+
+        while (index <= end_index) {
+                page = find_get_page(tree->mapping, index);
+                unlock_page(page);
+                page_cache_release(page);
+                index++;
+        }
+        unlock_extent(tree, start, end, GFP_NOFS);
+        return 0;
+}
+EXPORT_SYMBOL(unlock_range);
+
+/*
+ * searches a range in the state tree for a given mask.
+ * If 'filled' == 1, this returns 1 only if ever extent in the tree
+ * has the bits set.  Otherwise, 1 is returned if any bit in the
+ * range is found set.
+ */
+static int test_range_bit(struct extent_map_tree *tree, u64 start, u64 end,
+                          int bits, int filled)
+{
+        struct extent_state *state = NULL;
+        struct rb_node *node;
+        int bitset = 0;
+
+        read_lock_irq(&tree->lock);
+        node = tree_search(&tree->state, start);
+        while (node && start <= end) {
+                state = rb_entry(node, struct extent_state, rb_node);
+                if (state->start > end)
+                        break;
+
+                if (filled && state->start > start) {
+                        bitset = 0;
+                        break;
+                }
+                if (state->state & bits) {
+                        bitset = 1;
+                        if (!filled)
+                                break;
+                } else if (filled) {
+                        bitset = 0;
+                        break;
+                }
+                start = state->end + 1;
+                if (start > end)
+                        break;
+                node = rb_next(node);
+        }
+        read_unlock_irq(&tree->lock);
+        return bitset;
+}
+
+/*
+ * helper function to set a given page up to date if all the
+ * extents in the tree for that page are up to date
+ */
+static int check_page_uptodate(struct extent_map_tree *tree,
+                               struct page *page)
+{
+        u64 start = page->index << PAGE_CACHE_SHIFT;
+        u64 end = start + PAGE_CACHE_SIZE - 1;
+        if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1))
+                SetPageUptodate(page);
+        return 0;
+}
+
+/*
+ * helper function to unlock a page if all the extents in the tree
+ * for that page are unlocked
+ */
+static int check_page_locked(struct extent_map_tree *tree,
+                             struct page *page)
+{
+        u64 start = page->index << PAGE_CACHE_SHIFT;
+        u64 end = start + PAGE_CACHE_SIZE - 1;
+        if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0))
+                unlock_page(page);
+        return 0;
+}
+
+/*
+ * helper function to end page writeback if all the extents
+ * in the tree for that page are done with writeback
+ */
+static int check_page_writeback(struct extent_map_tree *tree,
+                             struct page *page)
+{
+        u64 start = page->index << PAGE_CACHE_SHIFT;
+        u64 end = start + PAGE_CACHE_SIZE - 1;
+        if (!test_range_bit(tree, start, end, EXTENT_WRITEBACK, 0))
+                end_page_writeback(page);
+        return 0;
+}
+
+/* lots and lots of room for performance fixes in the end_bio funcs */
+
+/*
+ * after a writepage IO is done, we need to:
+ * clear the uptodate bits on error
+ * clear the writeback bits in the extent tree for this IO
+ * end_page_writeback if the page has no more pending IO
+ *
+ * Scheduling is not allowed, so the extent state tree is expected
+ * to have one and only one object corresponding to this IO.
+ */
+static int end_bio_extent_writepage(struct bio *bio,
+                                   unsigned int bytes_done, int err)
+{
+        const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+        struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
+        struct extent_map_tree *tree = bio->bi_private;
+        u64 start;
+        u64 end;
+        int whole_page;
+
+        if (bio->bi_size)
+                return 1;
+
+        do {
+                struct page *page = bvec->bv_page;
+                start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
+                end = start + bvec->bv_len - 1;
+
+                if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
+                        whole_page = 1;
+                else
+                        whole_page = 0;
+
+                if (--bvec >= bio->bi_io_vec)
+                        prefetchw(&bvec->bv_page->flags);
+
+                if (!uptodate) {
+                        clear_extent_uptodate(tree, start, end, GFP_ATOMIC);
+                        ClearPageUptodate(page);
+                        SetPageError(page);
+                }
+                clear_extent_writeback(tree, start, end, GFP_ATOMIC);
+
+                if (whole_page)
+                        end_page_writeback(page);
+                else
+                        check_page_writeback(tree, page);
+        } while (bvec >= bio->bi_io_vec);
+
+        bio_put(bio);
+        return 0;
+}
+
+/*
+ * after a readpage IO is done, we need to:
+ * clear the uptodate bits on error
+ * set the uptodate bits if things worked
+ * set the page up to date if all extents in the tree are uptodate
+ * clear the lock bit in the extent tree
+ * unlock the page if there are no other extents locked for it
+ *
+ * Scheduling is not allowed, so the extent state tree is expected
+ * to have one and only one object corresponding to this IO.
+ */
+static int end_bio_extent_readpage(struct bio *bio,
+                                   unsigned int bytes_done, int err)
+{
+        const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+        struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
+        struct extent_map_tree *tree = bio->bi_private;
+        u64 start;
+        u64 end;
+        int whole_page;
+
+        if (bio->bi_size)
+                return 1;
+
+        do {
+                struct page *page = bvec->bv_page;
+                start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
+                end = start + bvec->bv_len - 1;
+
+                if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
+                        whole_page = 1;
+                else
+                        whole_page = 0;
+
+                if (--bvec >= bio->bi_io_vec)
+                        prefetchw(&bvec->bv_page->flags);
+
+                if (uptodate) {
+                        set_extent_uptodate(tree, start, end, GFP_ATOMIC);
+                        if (whole_page)
+                                SetPageUptodate(page);
+                        else
+                                check_page_uptodate(tree, page);
+                } else {
+                        ClearPageUptodate(page);
+                        SetPageError(page);
+                }
+
+                unlock_extent(tree, start, end, GFP_ATOMIC);
+
+                if (whole_page)
+                        unlock_page(page);
+                else
+                        check_page_locked(tree, page);
+        } while (bvec >= bio->bi_io_vec);
+
+        bio_put(bio);
+        return 0;
+}
+
+/*
+ * IO done from prepare_write is pretty simple, we just unlock
+ * the structs in the extent tree when done, and set the uptodate bits
+ * as appropriate.
+ */
+static int end_bio_extent_preparewrite(struct bio *bio,
+                                       unsigned int bytes_done, int err)
+{
+        const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+        struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
+        struct extent_map_tree *tree = bio->bi_private;
+        u64 start;
+        u64 end;
+
+        if (bio->bi_size)
+                return 1;
+
+        do {
+                struct page *page = bvec->bv_page;
+                start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
+                end = start + bvec->bv_len - 1;
+
+                if (--bvec >= bio->bi_io_vec)
+                        prefetchw(&bvec->bv_page->flags);
+
+                if (uptodate) {
+                        set_extent_uptodate(tree, start, end, GFP_ATOMIC);
+                } else {
+                        ClearPageUptodate(page);
+                        SetPageError(page);
+                }
+
+                unlock_extent(tree, start, end, GFP_ATOMIC);
+
+        } while (bvec >= bio->bi_io_vec);
+
+        bio_put(bio);
+        return 0;
+}
+
+static int submit_extent_page(int rw, struct extent_map_tree *tree,
+                              struct page *page, sector_t sector,
+                              size_t size, unsigned long offset,
+                              struct block_device *bdev,
+                              bio_end_io_t end_io_func)
+{
+        struct bio *bio;
+        int ret = 0;
+
+        bio = bio_alloc(GFP_NOIO, 1);
+
+        bio->bi_sector = sector;
+        bio->bi_bdev = bdev;
+        bio->bi_io_vec[0].bv_page = page;
+        bio->bi_io_vec[0].bv_len = size;
+        bio->bi_io_vec[0].bv_offset = offset;
+
+        bio->bi_vcnt = 1;
+        bio->bi_idx = 0;
+        bio->bi_size = size;
+
+        bio->bi_end_io = end_io_func;
+        bio->bi_private = tree;
+
+        bio_get(bio);
+        submit_bio(rw, bio);
+
+        if (bio_flagged(bio, BIO_EOPNOTSUPP))
+                ret = -EOPNOTSUPP;
+
+        bio_put(bio);
+        return ret;
+}
+
+/*
+ * basic readpage implementation.  Locked extent state structs are inserted
+ * into the tree that are removed when the IO is done (by the end_io
+ * handlers)
+ */
+int extent_read_full_page(struct extent_map_tree *tree, struct page *page,
+                          get_extent_t *get_extent)
+{
+        struct inode *inode = page->mapping->host;
+        u64 start = page->index << PAGE_CACHE_SHIFT;
+        u64 page_end = start + PAGE_CACHE_SIZE - 1;
+        u64 end;
+        u64 cur = start;
+        u64 extent_offset;
+        u64 last_byte = i_size_read(inode);
+        u64 block_start;
+        u64 cur_end;
+        sector_t sector;
+        struct extent_map *em;
+        struct block_device *bdev;
+        int ret;
+        int nr = 0;
+        size_t page_offset = 0;
+        size_t iosize;
+        size_t blocksize = inode->i_sb->s_blocksize;
+
+        if (!PagePrivate(page)) {
+                SetPagePrivate(page);
+                set_page_private(page, 1);
+                page_cache_get(page);
+        }
+
+        end = page_end;
+        lock_extent(tree, start, end, GFP_NOFS);
+
+        while (cur <= end) {
+                if (cur >= last_byte) {
+                        iosize = PAGE_CACHE_SIZE - page_offset;
+                        zero_user_page(page, page_offset, iosize, KM_USER0);
+                        set_extent_uptodate(tree, cur, cur + iosize - 1,
+                                            GFP_NOFS);
+                        unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
+                        break;
+                }
+                em = get_extent(inode, page, page_offset, cur, end, 0);
+                if (IS_ERR(em) || !em) {
+                        SetPageError(page);
+                        unlock_extent(tree, cur, end, GFP_NOFS);
+                        break;
+                }
+
+                extent_offset = cur - em->start;
+                BUG_ON(em->end < cur);
+                BUG_ON(end < cur);
+
+                iosize = min(em->end - cur, end - cur) + 1;
+                cur_end = min(em->end, end);
+                iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
+                sector = (em->block_start + extent_offset) >> 9;
+                bdev = em->bdev;
+                block_start = em->block_start;
+                free_extent_map(em);
+                em = NULL;
+
+                /* we've found a hole, just zero and go on */
+                if (block_start == 0) {
+                        zero_user_page(page, page_offset, iosize, KM_USER0);
+                        set_extent_uptodate(tree, cur, cur + iosize - 1,
+                                            GFP_NOFS);
+                        unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
+                        cur = cur + iosize;
+                        page_offset += iosize;
+                        continue;
+                }
+                /* the get_extent function already copied into the page */
+                if (test_range_bit(tree, cur, cur_end, EXTENT_UPTODATE, 1)) {
+                        unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
+                        cur = cur + iosize;
+                        page_offset += iosize;
+                        continue;
+                }
+
+                ret = submit_extent_page(READ, tree, page,
+                                         sector, iosize, page_offset, bdev,
+                                         end_bio_extent_readpage);
+                if (ret)
+                        SetPageError(page);
+                cur = cur + iosize;
+                page_offset += iosize;
+                nr++;
+        }
+        if (!nr) {
+                if (!PageError(page))
+                        SetPageUptodate(page);
+                unlock_page(page);
+        }
+        return 0;
+}
+EXPORT_SYMBOL(extent_read_full_page);
+
+/*
+ * the writepage semantics are similar to regular writepage.  extent
+ * records are inserted to lock ranges in the tree, and as dirty areas
+ * are found, they are marked writeback.  Then the lock bits are removed
+ * and the end_io handler clears the writeback ranges
+ */
+int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
+                          get_extent_t *get_extent,
+                          struct writeback_control *wbc)
+{
+        struct inode *inode = page->mapping->host;
+        u64 start = page->index << PAGE_CACHE_SHIFT;
+        u64 page_end = start + PAGE_CACHE_SIZE - 1;
+        u64 end;
+        u64 cur = start;
+        u64 extent_offset;
+        u64 last_byte = i_size_read(inode);
+        u64 block_start;
+        sector_t sector;
+        struct extent_map *em;
+        struct block_device *bdev;
+        int ret;
+        int nr = 0;
+        size_t page_offset = 0;
+        size_t iosize;
+        size_t blocksize;
+        loff_t i_size = i_size_read(inode);
+        unsigned long end_index = i_size >> PAGE_CACHE_SHIFT;
+
+        if (page->index > end_index) {
+                clear_extent_dirty(tree, start, page_end, GFP_NOFS);
+                unlock_page(page);
+                return 0;
+        }
+
+        if (page->index == end_index) {
+                size_t offset = i_size & (PAGE_CACHE_SIZE - 1);
+                zero_user_page(page, offset,
+                               PAGE_CACHE_SIZE - offset, KM_USER0);
+        }
+
+        if (!PagePrivate(page)) {
+                SetPagePrivate(page);
+                set_page_private(page, 1);
+                page_cache_get(page);
+        }
+
+        end = page_end;
+        lock_extent(tree, start, page_end, GFP_NOFS);
+
+        if (last_byte <= start) {
+                clear_extent_dirty(tree, start, page_end, GFP_NOFS);
+                goto done;
+        }
+
+        set_extent_uptodate(tree, start, page_end, GFP_NOFS);
+        blocksize = inode->i_sb->s_blocksize;
+
+        while (cur <= end) {
+                if (cur >= last_byte) {
+                        clear_extent_dirty(tree, cur, page_end, GFP_NOFS);
+                        break;
+                }
+                em = get_extent(inode, page, page_offset, cur, end, 1);
+                if (IS_ERR(em) || !em) {
+                        SetPageError(page);
+                        break;
+                }
+
+                extent_offset = cur - em->start;
+                BUG_ON(em->end < cur);
+                BUG_ON(end < cur);
+                iosize = min(em->end - cur, end - cur) + 1;
+                iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
+                sector = (em->block_start + extent_offset) >> 9;
+                bdev = em->bdev;
+                block_start = em->block_start;
+                free_extent_map(em);
+                em = NULL;
+
+                if (block_start == 0 || block_start == EXTENT_MAP_INLINE) {
+                        clear_extent_dirty(tree, cur,
+                                           cur + iosize - 1, GFP_NOFS);
+                        cur = cur + iosize;
+                        page_offset += iosize;
+                        continue;
+                }
+
+                /* leave this out until we have a page_mkwrite call */
+                if (0 && !test_range_bit(tree, cur, cur + iosize - 1,
+                                   EXTENT_DIRTY, 0)) {
+                        cur = cur + iosize;
+                        page_offset += iosize;
+                        continue;
+                }
+                clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS);
+                set_range_writeback(tree, cur, cur + iosize - 1);
+                ret = submit_extent_page(WRITE, tree, page,
+                                         sector, iosize, page_offset, bdev,
+                                         end_bio_extent_writepage);
+                if (ret)
+                        SetPageError(page);
+                cur = cur + iosize;
+                page_offset += iosize;
+                nr++;
+        }
+done:
+        WARN_ON(test_range_bit(tree, start, page_end, EXTENT_DIRTY, 0));
+        unlock_extent(tree, start, page_end, GFP_NOFS);
+        unlock_page(page);
+        return 0;
+}
+EXPORT_SYMBOL(extent_write_full_page);
+
+/*
+ * basic invalidatepage code, this waits on any locked or writeback
+ * ranges corresponding to the page, and then deletes any extent state
+ * records from the tree
+ */
+int extent_invalidatepage(struct extent_map_tree *tree,
+                          struct page *page, unsigned long offset)
+{
+        u64 start = (page->index << PAGE_CACHE_SHIFT);
+        u64 end = start + PAGE_CACHE_SIZE - 1;
+        size_t blocksize = page->mapping->host->i_sb->s_blocksize;
+
+        start += (offset + blocksize -1) & ~(blocksize - 1);
+        if (start > end)
+                return 0;
+
+        lock_extent(tree, start, end, GFP_NOFS);
+        wait_on_extent_writeback(tree, start, end);
+        clear_extent_bit(tree, start, end, EXTENT_LOCKED | EXTENT_DIRTY,
+                         1, 1, GFP_NOFS);
+        return 0;
+}
+EXPORT_SYMBOL(extent_invalidatepage);
+
+/*
+ * simple commit_write call, set_range_dirty is used to mark both
+ * the pages and the extent records as dirty
+ */
+int extent_commit_write(struct extent_map_tree *tree,
+                        struct inode *inode, struct page *page,
+                        unsigned from, unsigned to)
+{
+        loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
+
+        if (!PagePrivate(page)) {
+                SetPagePrivate(page);
+                set_page_private(page, 1);
+                page_cache_get(page);
+        }
+
+        set_page_dirty(page);
+
+        if (pos > inode->i_size) {
+                i_size_write(inode, pos);
+                mark_inode_dirty(inode);
+        }
+        return 0;
+}
+EXPORT_SYMBOL(extent_commit_write);
+
+int extent_prepare_write(struct extent_map_tree *tree,
+                         struct inode *inode, struct page *page,
+                         unsigned from, unsigned to, get_extent_t *get_extent)
+{
+        u64 page_start = page->index << PAGE_CACHE_SHIFT;
+        u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
+        u64 block_start;
+        u64 orig_block_start;
+        u64 block_end;
+        u64 cur_end;
+        struct extent_map *em;
+        unsigned blocksize = 1 << inode->i_blkbits;
+        size_t page_offset = 0;
+        size_t block_off_start;
+        size_t block_off_end;
+        int err = 0;
+        int iocount = 0;
+        int ret = 0;
+        int isnew;
+
+        if (!PagePrivate(page)) {
+                SetPagePrivate(page);
+                set_page_private(page, 1);
+                page_cache_get(page);
+        }
+        block_start = (page_start + from) & ~((u64)blocksize - 1);
+        block_end = (page_start + to - 1) | (blocksize - 1);
+        orig_block_start = block_start;
+
+        lock_extent(tree, page_start, page_end, GFP_NOFS);
+        while(block_start <= block_end) {
+                em = get_extent(inode, page, page_offset, block_start,
+                                block_end, 1);
+                if (IS_ERR(em) || !em) {
+                        goto err;
+                }
+                cur_end = min(block_end, em->end);
+                block_off_start = block_start & (PAGE_CACHE_SIZE - 1);
+                block_off_end = block_off_start + blocksize;
+                isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS);
+
+                if (!PageUptodate(page) && isnew &&
+                    (block_off_end > to || block_off_start < from)) {
+                        void *kaddr;
+
+                        kaddr = kmap_atomic(page, KM_USER0);
+                        if (block_off_end > to)
+                                memset(kaddr + to, 0, block_off_end - to);
+                        if (block_off_start < from)
+                                memset(kaddr + block_off_start, 0,
+                                       from - block_off_start);
+                        flush_dcache_page(page);
+                        kunmap_atomic(kaddr, KM_USER0);
+                }
+                if (!isnew && !PageUptodate(page) &&
+                    (block_off_end > to || block_off_start < from) &&
+                    !test_range_bit(tree, block_start, cur_end,
+                                    EXTENT_UPTODATE, 1)) {
+                        u64 sector;
+                        u64 extent_offset = block_start - em->start;
+                        size_t iosize;
+                        sector = (em->block_start + extent_offset) >> 9;
+                        iosize = (cur_end - block_start + blocksize - 1) &
+                                ~((u64)blocksize - 1);
+                        /*
+                         * we've already got the extent locked, but we
+                         * need to split the state such that our end_bio
+                         * handler can clear the lock.
+                         */
+                        set_extent_bit(tree, block_start,
+                                       block_start + iosize - 1,
+                                       EXTENT_LOCKED, 0, NULL, GFP_NOFS);
+                        ret = submit_extent_page(READ, tree, page,
+                                         sector, iosize, page_offset, em->bdev,
+                                         end_bio_extent_preparewrite);
+                        iocount++;
+                        block_start = block_start + iosize;
+                } else {
+                        set_extent_uptodate(tree, block_start, cur_end,
+                                            GFP_NOFS);
+                        unlock_extent(tree, block_start, cur_end, GFP_NOFS);
+                        block_start = cur_end + 1;
+                }
+                page_offset = block_start & (PAGE_CACHE_SIZE - 1);
+                free_extent_map(em);
+        }
+        if (iocount) {
+                wait_extent_bit(tree, orig_block_start,
+                                block_end, EXTENT_LOCKED);
+        }
+        check_page_uptodate(tree, page);
+err:
+        /* FIXME, zero out newly allocated blocks on error */
+        return err;
+}
+EXPORT_SYMBOL(extent_prepare_write);
+
+/*
+ * a helper for releasepage.  As long as there are no locked extents
+ * in the range corresponding to the page, both state records and extent
+ * map records are removed
+ */
+int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page)
+{
+        struct extent_map *em;
+        u64 start = page->index << PAGE_CACHE_SHIFT;
+        u64 end = start + PAGE_CACHE_SIZE - 1;
+        u64 orig_start = start;
+
+        while (start <= end) {
+                em = lookup_extent_mapping(tree, start, end);
+                if (!em || IS_ERR(em))
+                        break;
+                if (test_range_bit(tree, em->start, em->end,
+                                   EXTENT_LOCKED, 0)) {
+                        free_extent_map(em);
+                        start = em->end + 1;
+printk("range still locked %Lu %Lu\n", em->start, em->end);
+                        break;
+                }
+                remove_extent_mapping(tree, em);
+                start = em->end + 1;
+                /* once for the rb tree */
+                free_extent_map(em);
+                /* once for us */
+                free_extent_map(em);
+        }
+        WARN_ON(test_range_bit(tree, orig_start, end, EXTENT_WRITEBACK, 0));
+        clear_extent_bit(tree, orig_start, end, EXTENT_UPTODATE,
+                         1, 1, GFP_NOFS);
+        return 1;
+}
+EXPORT_SYMBOL(try_release_extent_mapping);
+
diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h
new file mode 100644
index 000000000000..108944aab4bd
--- /dev/null
+++ b/fs/btrfs/extent_map.h
@@ -0,0 +1,89 @@
+#ifndef __EXTENTMAP__
+#define __EXTENTMAP__
+
+#include <linux/rbtree.h>
+
+#define EXTENT_MAP_INLINE (u64)-2
+#define EXTENT_MAP_DELALLOC (u64)-1
+
+struct extent_map_tree {
+        struct rb_root map;
+        struct rb_root state;
+        struct address_space *mapping;
+        rwlock_t lock;
+};
+
+/* note, this must start with the same fields as fs/extent_map.c:tree_entry */
+struct extent_map {
+        u64 start;
+        u64 end; /* inclusive */
+        int in_tree;
+        struct rb_node rb_node;
+        /* block_start and block_end are in bytes */
+        u64 block_start;
+        u64 block_end; /* inclusive */
+        struct block_device *bdev;
+        atomic_t refs;
+};
+
+/* note, this must start with the same fields as fs/extent_map.c:tree_entry */
+struct extent_state {
+        u64 start;
+        u64 end; /* inclusive */
+        int in_tree;
+        struct rb_node rb_node;
+        wait_queue_head_t wq;
+        atomic_t refs;
+        unsigned long state;
+        struct list_head list;
+};
+
+struct extent_buffer {
+        u64 start;
+        u64 end; /* inclusive */
+        char *addr;
+        struct page *pages[];
+};
+
+typedef struct extent_map *(get_extent_t)(struct inode *inode,
+                                          struct page *page,
+                                          size_t page_offset,
+                                          u64 start, u64 end,
+                                          int create);
+
+void extent_map_tree_init(struct extent_map_tree *tree,
+                          struct address_space *mapping, gfp_t mask);
+struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree,
+                                         u64 start, u64 end);
+int add_extent_mapping(struct extent_map_tree *tree,
+                       struct extent_map *em);
+int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em);
+int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page);
+int lock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask);
+int unlock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask);
+struct extent_map *alloc_extent_map(gfp_t mask);
+void free_extent_map(struct extent_map *em);
+int extent_read_full_page(struct extent_map_tree *tree, struct page *page,
+                          get_extent_t *get_extent);
+void __init extent_map_init(void);
+void __exit extent_map_exit(void);
+int extent_clean_all_trees(struct extent_map_tree *tree);
+int set_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end,
+                        gfp_t mask);
+int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end,
+                   gfp_t mask);
+int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end,
+                     gfp_t mask);
+int extent_invalidatepage(struct extent_map_tree *tree,
+                          struct page *page, unsigned long offset);
+int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
+                          get_extent_t *get_extent,
+                          struct writeback_control *wbc);
+int extent_prepare_write(struct extent_map_tree *tree,
+                         struct inode *inode, struct page *page,
+                         unsigned from, unsigned to, get_extent_t *get_extent);
+int extent_commit_write(struct extent_map_tree *tree,
+                        struct inode *inode, struct page *page,
+                        unsigned from, unsigned to);
+int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end);
+#endif
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 6933ab11a5cd..71a481894ab6 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -81,14 +81,14 @@ static void btrfs_drop_pages(struct page **pages, size_t num_pages)
         }
 }
 
-static int insert_inline_extent(struct btrfs_root *root, struct inode *inode,
+static int insert_inline_extent(struct btrfs_trans_handle *trans,
+                                struct btrfs_root *root, struct inode *inode,
                                 u64 offset, ssize_t size,
-                                struct buffer_head *bh)
+                                struct page *page, size_t page_offset)
 {
         struct btrfs_key key;
         struct btrfs_path *path;
         char *ptr, *kaddr;
-        struct btrfs_trans_handle *trans;
         struct btrfs_file_extent_item *ei;
         u32 datasize;
         int err = 0;
@@ -98,8 +98,6 @@ static int insert_inline_extent(struct btrfs_root *root, struct inode *inode,
         if (!path)
                 return -ENOMEM;
 
-        mutex_lock(&root->fs_info->fs_mutex);
-        trans = btrfs_start_transaction(root, 1);
         btrfs_set_trans_block_group(trans, inode);
 
         key.objectid = inode->i_ino;
@@ -122,18 +120,13 @@ static int insert_inline_extent(struct btrfs_root *root, struct inode *inode,
                                    BTRFS_FILE_EXTENT_INLINE);
         ptr = btrfs_file_extent_inline_start(ei);
 
-        kaddr = kmap_atomic(bh->b_page, KM_USER0);
+        kaddr = kmap_atomic(page, KM_USER0);
         btrfs_memcpy(root, path->nodes[0]->b_data,
-                     ptr, kaddr + bh_offset(bh),
-                     size);
+                     ptr, kaddr + page_offset, size);
         kunmap_atomic(kaddr, KM_USER0);
         btrfs_mark_buffer_dirty(path->nodes[0]);
 fail:
         btrfs_free_path(path);
-        ret = btrfs_end_transaction(trans, root);
-        if (ret && !err)
-                err = ret;
-        mutex_unlock(&root->fs_info->fs_mutex);
         return err;
 }
 
@@ -145,45 +138,143 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
                                    loff_t pos,
                                    size_t write_bytes)
 {
-        int i;
-        int offset;
         int err = 0;
-        int ret;
-        int this_write;
+        int i;
         struct inode *inode = file->f_path.dentry->d_inode;
-        struct buffer_head *bh;
+        struct extent_map *em;
+        struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+        struct btrfs_key ins;
+        u64 hint_block;
+        u64 num_blocks;
+        u64 start_pos;
+        u64 end_of_last_block;
+        u64 end_pos = pos + write_bytes;
+        loff_t isize = i_size_read(inode);
 
-        for (i = 0; i < num_pages; i++) {
-                offset = pos & (PAGE_CACHE_SIZE -1);
-                this_write = min((size_t)PAGE_CACHE_SIZE - offset, write_bytes);
+        em = alloc_extent_map(GFP_NOFS);
+        if (!em)
+                return -ENOMEM;
 
-                /* FIXME, one block at a time */
-                bh = page_buffers(pages[i]);
+        em->bdev = inode->i_sb->s_bdev;
 
-                if (buffer_mapped(bh) && bh->b_blocknr == 0) {
-                        ret = insert_inline_extent(root, inode,
-                                        pages[i]->index << PAGE_CACHE_SHIFT,
-                                        offset + this_write, bh);
-                        if (ret) {
-                                err = ret;
-                                goto failed;
-                        }
-                }
+        start_pos = pos & ~((u64)root->blocksize - 1);
+        num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >>
+                        inode->i_blkbits;
 
-                ret = btrfs_commit_write(file, pages[i], offset,
-                                         offset + this_write);
-                pos += this_write;
-                if (ret) {
-                        err = ret;
+        end_of_last_block = start_pos + (num_blocks << inode->i_blkbits) - 1;
+        mutex_lock(&root->fs_info->fs_mutex);
+        trans = btrfs_start_transaction(root, 1);
+        if (!trans) {
+                err = -ENOMEM;
+                goto out_unlock;
+        }
+        btrfs_set_trans_block_group(trans, inode);
+        inode->i_blocks += num_blocks << 3;
+        hint_block = 0;
+
+        if ((end_of_last_block & 4095) == 0) {
+                printk("strange end of last %Lu %lu %Lu\n", start_pos, write_bytes, end_of_last_block);
+        }
+        set_extent_uptodate(em_tree, start_pos, end_of_last_block, GFP_NOFS);
+
+        /* FIXME...EIEIO, ENOSPC and more */
+
+        /* step one, delete the existing extents in this range */
+        /* FIXME blocksize != pagesize */
+        if (start_pos < inode->i_size) {
+                err = btrfs_drop_extents(trans, root, inode,
+                         start_pos, (pos + write_bytes + root->blocksize -1) &
+                         ~((u64)root->blocksize - 1), &hint_block);
+                if (err)
+                        goto failed;
+        }
+
+        /* insert any holes we need to create */
+        if (inode->i_size < start_pos) {
+                u64 last_pos_in_file;
+                u64 hole_size;
+                u64 mask = root->blocksize - 1;
+                last_pos_in_file = (isize + mask) & ~mask;
+                hole_size = (start_pos - last_pos_in_file + mask) & ~mask;
+                hole_size >>= inode->i_blkbits;
+                if (last_pos_in_file < start_pos) {
+                        err = btrfs_insert_file_extent(trans, root,
+                                                       inode->i_ino,
+                                                       last_pos_in_file,
+                                                       0, 0, hole_size);
+                }
+                if (err)
                         goto failed;
+        }
+
+        /*
+         * either allocate an extent for the new bytes or setup the key
+         * to show we are doing inline data in the extent
+         */
+        if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size ||
+            pos + write_bytes - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) {
+                err = btrfs_alloc_extent(trans, root, inode->i_ino,
+                                         num_blocks, 0, hint_block, (u64)-1,
+                                         &ins, 1);
+                BUG_ON(err);
+                err = btrfs_insert_file_extent(trans, root, inode->i_ino,
+                                       start_pos, ins.objectid, ins.offset,
+                                       ins.offset);
+                BUG_ON(err);
+                em->start = start_pos;
+                em->end = end_of_last_block;
+                em->block_start = ins.objectid << inode->i_blkbits;
+                em->block_end = em->block_start +
+                        (ins.offset << inode->i_blkbits) - 1;
+                set_extent_dirty(em_tree, start_pos, end_of_last_block,
+                                 GFP_NOFS);
+                err = add_extent_mapping(em_tree, em);
+                for (i = 0; i < num_pages; i++) {
+                        struct page *p = pages[i];
+                        SetPageUptodate(p);
+                        __set_page_dirty_nobuffers(p);
                 }
-                WARN_ON(this_write > write_bytes);
-                write_bytes -= this_write;
+        } else {
+                struct page *p = pages[0];
+                err = insert_inline_extent(trans, root, inode, start_pos,
+                                           end_pos - start_pos, p, 0);
+                BUG_ON(err);
+                em->start = start_pos;
+                em->end = end_pos;
+                em->block_start = EXTENT_MAP_INLINE;
+                em->block_end = EXTENT_MAP_INLINE;
+                add_extent_mapping(em_tree, em);
+        }
+        if (end_pos > isize) {
+                i_size_write(inode, end_pos);
+                btrfs_update_inode(trans, root, inode);
         }
 failed:
+        err = btrfs_end_transaction(trans, root);
+out_unlock:
+        mutex_unlock(&root->fs_info->fs_mutex);
+        free_extent_map(em);
         return err;
 }
 
+int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end)
+{
+        struct extent_map *em;
+        struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+
+        while(1) {
+                em = lookup_extent_mapping(em_tree, start, end);
+                if (!em)
+                        break;
+                remove_extent_mapping(em_tree, em);
+                /* once for us */
+                free_extent_map(em);
+                /* once for the tree*/
+                free_extent_map(em);
+        }
+        return 0;
+}
+
 /*
  * this is very complex, but the basic idea is to drop all extents
  * in the range start - end.  hint_block is filled in with a block number
@@ -213,6 +304,8 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans,
         int found_inline;
         int recow;
 
+        btrfs_drop_extent_cache(inode, start, end - 1);
+
         path = btrfs_alloc_path();
         if (!path)
                 return -ENOMEM;
@@ -434,18 +527,9 @@ static int prepare_pages(struct btrfs_root *root,
         int i;
         unsigned long index = pos >> PAGE_CACHE_SHIFT;
         struct inode *inode = file->f_path.dentry->d_inode;
-        int offset;
         int err = 0;
-        int this_write;
-        struct buffer_head *bh;
-        struct buffer_head *head;
-        loff_t isize = i_size_read(inode);
-        struct btrfs_trans_handle *trans;
-        u64 hint_block;
         u64 num_blocks;
-        u64 alloc_extent_start;
         u64 start_pos;
-        struct btrfs_key ins;
 
         start_pos = pos & ~((u64)PAGE_CACHE_SIZE - 1);
         num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >>
@@ -457,119 +541,17 @@ static int prepare_pages(struct btrfs_root *root,
                 pages[i] = grab_cache_page(inode->i_mapping, index + i);
                 if (!pages[i]) {
                         err = -ENOMEM;
-                        goto failed_release;
+                        BUG_ON(1);
                 }
                 cancel_dirty_page(pages[i], PAGE_CACHE_SIZE);
                 wait_on_page_writeback(pages[i]);
-        }
-
-        mutex_lock(&root->fs_info->fs_mutex);
-        trans = btrfs_start_transaction(root, 1);
-        if (!trans) {
-                err = -ENOMEM;
-                mutex_unlock(&root->fs_info->fs_mutex);
-                goto out_unlock;
-        }
-        btrfs_set_trans_block_group(trans, inode);
-        /* FIXME blocksize != 4096 */
-        inode->i_blocks += num_blocks << 3;
-        hint_block = 0;
-
-        /* FIXME...EIEIO, ENOSPC and more */
-
-        /* step one, delete the existing extents in this range */
-        /* FIXME blocksize != pagesize */
-        if (start_pos < inode->i_size) {
-                err = btrfs_drop_extents(trans, root, inode,
-                         start_pos, (pos + write_bytes + root->blocksize -1) &
-                         ~((u64)root->blocksize - 1), &hint_block);
-                if (err)
-                        goto failed_release;
-        }
-
-        /* insert any holes we need to create */
-        if (inode->i_size < start_pos) {
-                u64 last_pos_in_file;
-                u64 hole_size;
-                u64 mask = root->blocksize - 1;
-                last_pos_in_file = (isize + mask) & ~mask;
-                hole_size = (start_pos - last_pos_in_file + mask) & ~mask;
-                hole_size >>= inode->i_blkbits;
-                if (last_pos_in_file < start_pos) {
-                        err = btrfs_insert_file_extent(trans, root,
-                                                       inode->i_ino,
-                                                       last_pos_in_file,
-                                                       0, 0, hole_size);
-                }
-                if (err)
-                        goto failed_release;
-        }
-
-        /*
-         * either allocate an extent for the new bytes or setup the key
-         * to show we are doing inline data in the extent
-         */
-        if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size ||
-            pos + write_bytes - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) {
-                err = btrfs_alloc_extent(trans, root, inode->i_ino,
-                                         num_blocks, 0, hint_block, (u64)-1,
-                                         &ins, 1);
-                if (err)
-                        goto failed_truncate;
-                err = btrfs_insert_file_extent(trans, root, inode->i_ino,
-                                       start_pos, ins.objectid, ins.offset,
-                                       ins.offset);
-                if (err)
-                        goto failed_truncate;
-        } else {
-                ins.offset = 0;
-                ins.objectid = 0;
-        }
-        BUG_ON(err);
-        alloc_extent_start = ins.objectid;
-        err = btrfs_end_transaction(trans, root);
-        mutex_unlock(&root->fs_info->fs_mutex);
-
-        for (i = 0; i < num_pages; i++) {
-                offset = pos & (PAGE_CACHE_SIZE -1);
-                this_write = min((size_t)PAGE_CACHE_SIZE - offset, write_bytes);
-                if (!page_has_buffers(pages[i])) {
-                        create_empty_buffers(pages[i],
-                                             root->fs_info->sb->s_blocksize,
-                                             (1 << BH_Uptodate));
+                if (!PagePrivate(pages[i])) {
+                        SetPagePrivate(pages[i]);
+                        set_page_private(pages[i], 1);
+                        page_cache_get(pages[i]);
                 }
-                head = page_buffers(pages[i]);
-                bh = head;
-                do {
-                        err = btrfs_map_bh_to_logical(root, bh,
-                                                      alloc_extent_start);
-                        BUG_ON(err);
-                        if (err)
-                                goto failed_truncate;
-                        bh = bh->b_this_page;
-                        if (alloc_extent_start)
-                                alloc_extent_start++;
-                } while (bh != head);
-                pos += this_write;
-                WARN_ON(this_write > write_bytes);
-                write_bytes -= this_write;
         }
         return 0;
-
-failed_release:
-        btrfs_drop_pages(pages, num_pages);
-        return err;
-
-failed_truncate:
-        btrfs_drop_pages(pages, num_pages);
-        if (pos > isize)
-                vmtruncate(inode, isize);
-        return err;
-
-out_unlock:
-        mutex_unlock(&root->fs_info->fs_mutex);
-        goto failed_release;
-
 }
 
 static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
@@ -685,7 +667,6 @@ out:
                 page_cache_release(pinned[1]);
         *ppos = pos;
         current->backing_dev_info = NULL;
-        mark_inode_dirty(inode);
         return num_written ? num_written : err;
 }
 
@@ -714,8 +695,8 @@ static int btrfs_sync_file(struct file *file,
         mutex_unlock(&root->fs_info->trans_mutex);
 
         /*
-         * ok we haven't committed the transaction yet, lets do a commit
-         */
+         * ok we haven't committed the transaction yet, lets do a commit
+         */
         trans = btrfs_start_transaction(root, 1);
         if (!trans) {
                 ret = -ENOMEM;
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 504d37dfa6c9..fd4cb65fcf1f 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -121,6 +121,8 @@ void btrfs_read_locked_inode(struct inode *inode)
         switch (inode->i_mode & S_IFMT) {
         case S_IFREG:
                 inode->i_mapping->a_ops = &btrfs_aops;
+                extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
+                                     inode->i_mapping, GFP_NOFS);
                 inode->i_fop = &btrfs_file_operations;
                 inode->i_op = &btrfs_file_inode_operations;
                 break;
@@ -169,7 +171,7 @@ static void fill_inode_item(struct btrfs_inode_item *item,
                                     BTRFS_I(inode)->block_group->key.objectid);
 }
 
-static int btrfs_update_inode(struct btrfs_trans_handle *trans,
+int btrfs_update_inode(struct btrfs_trans_handle *trans,
                               struct btrfs_root *root,
                               struct inode *inode)
 {
@@ -400,6 +402,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
         int found_extent;
         int del_item;
 
+        btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
         path = btrfs_alloc_path();
         path->reada = -1;
         BUG_ON(!path);
@@ -511,6 +514,50 @@ error:
         return ret;
 }
 
+static int btrfs_cow_one_page(struct btrfs_trans_handle *trans,
+                              struct inode *inode, struct page *page,
+                              size_t zero_start)
+{
+        char *kaddr;
+        int ret = 0;
+        struct btrfs_root *root = BTRFS_I(inode)->root;
+        u64 alloc_hint = 0;
+        u64 page_start = page->index << PAGE_CACHE_SHIFT;
+        struct btrfs_key ins;
+
+        if (!PagePrivate(page)) {
+                SetPagePrivate(page);
+                set_page_private(page, 1);
+                page_cache_get(page);
+        }
+
+        btrfs_set_trans_block_group(trans, inode);
+
+        ret = btrfs_drop_extents(trans, root, inode,
+                                 page_start, page_start + PAGE_CACHE_SIZE,
+                                 &alloc_hint);
+        if (ret)
+                goto out;
+        ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1, 0,
+                                 alloc_hint, (u64)-1, &ins, 1);
+        if (ret)
+                goto out;
+        ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
+                                       page_start, ins.objectid, 1, 1);
+        if (ret)
+                goto out;
+        SetPageChecked(page);
+        kaddr = kmap(page);
+        if (zero_start != PAGE_CACHE_SIZE) {
+                memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
+                flush_dcache_page(page);
+        }
+        kunmap(page);
+
+out:
+        return ret;
+}
+
 /*
  * taken from block_truncate_page, but does cow as it zeros out
  * any bytes left in the last page in the file.
@@ -518,16 +565,14 @@ error:
 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 {
         struct inode *inode = mapping->host;
+        struct btrfs_root *root = BTRFS_I(inode)->root;
+        struct btrfs_trans_handle *trans;
         unsigned blocksize = 1 << inode->i_blkbits;
         pgoff_t index = from >> PAGE_CACHE_SHIFT;
         unsigned offset = from & (PAGE_CACHE_SIZE-1);
         struct page *page;
-        char *kaddr;
         int ret = 0;
-        struct btrfs_root *root = BTRFS_I(inode)->root;
-        u64 alloc_hint = 0;
-        struct btrfs_key ins;
-        struct btrfs_trans_handle *trans;
+        u64 page_start;
 
         if ((offset & (blocksize - 1)) == 0)
                 goto out;
@@ -536,7 +581,6 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
         page = grab_cache_page(mapping, index);
         if (!page)
                 goto out;
-
         if (!PageUptodate(page)) {
                 ret = btrfs_readpage(NULL, page);
                 lock_page(page);
@@ -545,37 +589,24 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
                         goto out;
                 }
         }
+        page_start = page->index << PAGE_CACHE_SHIFT;
+
         mutex_lock(&root->fs_info->fs_mutex);
         trans = btrfs_start_transaction(root, 1);
-        btrfs_set_trans_block_group(trans, inode);
-
-        ret = btrfs_drop_extents(trans, root, inode,
-                                 page->index << PAGE_CACHE_SHIFT,
-                                 (page->index + 1) << PAGE_CACHE_SHIFT,
-                                 &alloc_hint);
-        if (ret)
-                goto out;
-        ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1, 0,
-                                 alloc_hint, (u64)-1, &ins, 1);
-        if (ret)
-                goto out;
-        ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
-                                       page->index << PAGE_CACHE_SHIFT,
-                                       ins.objectid, 1, 1);
-        if (ret)
-                goto out;
-        SetPageChecked(page);
-        kaddr = kmap(page);
-        memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
-        flush_dcache_page(page);
-        ret = btrfs_csum_file_block(trans, root, inode->i_ino,
-                              page->index << PAGE_CACHE_SHIFT,
-                              kaddr, PAGE_CACHE_SIZE);
-        kunmap(page);
+        ret = btrfs_cow_one_page(trans, inode, page, offset);
+        if (!ret) {
+                char *kaddr = kmap(page);
+                ret = btrfs_csum_file_block(trans, root, inode->i_ino,
+                                            page_start, kaddr, PAGE_CACHE_SIZE);
+                kunmap(page);
+        }
+        set_extent_dirty(&BTRFS_I(inode)->extent_tree,
+                         page_start, page_start + PAGE_CACHE_SIZE - 1,
+                         GFP_NOFS);
+        set_page_dirty(page);
         btrfs_end_transaction(trans, root);
         mutex_unlock(&root->fs_info->fs_mutex);
 
-        set_page_dirty(page);
         unlock_page(page);
         page_cache_release(page);
 out:
@@ -1095,6 +1126,8 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
                 inode->i_mapping->a_ops = &btrfs_aops;
                 inode->i_fop = &btrfs_file_operations;
                 inode->i_op = &btrfs_file_inode_operations;
+                extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
+                                     inode->i_mapping, GFP_NOFS);
         }
         dir->i_sb->s_dirt = 1;
         btrfs_update_inode_block_group(trans, inode);
@@ -1238,6 +1271,182 @@ out_unlock:
         return err;
 }
 
+struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
+                                    size_t page_offset, u64 start, u64 end,
+                                    int create)
+{
+        int ret;
+        int err = 0;
+        u64 blocknr;
+        u64 extent_start = 0;
+        u64 extent_end = 0;
+        u64 objectid = inode->i_ino;
+        u32 found_type;
+        int failed_insert = 0;
+        struct btrfs_path *path;
+        struct btrfs_root *root = BTRFS_I(inode)->root;
+        struct btrfs_file_extent_item *item;
+        struct btrfs_leaf *leaf;
+        struct btrfs_disk_key *found_key;
+        struct extent_map *em = NULL;
+        struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+        struct btrfs_trans_handle *trans = NULL;
+
+        path = btrfs_alloc_path();
+        BUG_ON(!path);
+        mutex_lock(&root->fs_info->fs_mutex);
+
+again:
+        em = lookup_extent_mapping(em_tree, start, end);
+        if (em) {
+                goto out;
+        }
+        if (!em) {
+                em = alloc_extent_map(GFP_NOFS);
+                if (!em) {
+                        err = -ENOMEM;
+                        goto out;
+                }
+                em->start = 0;
+                em->end = 0;
+        }
+        em->bdev = inode->i_sb->s_bdev;
+        ret = btrfs_lookup_file_extent(NULL, root, path,
+                                       objectid, start, 0);
+        if (ret < 0) {
+                err = ret;
+                goto out;
+        }
+
+        if (ret != 0) {
+                if (path->slots[0] == 0)
+                        goto not_found;
+                path->slots[0]--;
+        }
+
+        item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
+                              struct btrfs_file_extent_item);
+        leaf = btrfs_buffer_leaf(path->nodes[0]);
+        blocknr = btrfs_file_extent_disk_blocknr(item);
+        blocknr += btrfs_file_extent_offset(item);
+
+        /* are we inside the extent that was found? */
+        found_key = &leaf->items[path->slots[0]].key;
+        found_type = btrfs_disk_key_type(found_key);
+        if (btrfs_disk_key_objectid(found_key) != objectid ||
+            found_type != BTRFS_EXTENT_DATA_KEY) {
+                goto not_found;
+        }
+
+        found_type = btrfs_file_extent_type(item);
+        extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
+        if (found_type == BTRFS_FILE_EXTENT_REG) {
+                extent_end = extent_start +
+                       (btrfs_file_extent_num_blocks(item) << inode->i_blkbits);
+                err = 0;
+                if (start < extent_start || start > extent_end) {
+                        em->start = start;
+                        if (start < extent_start) {
+                                em->end = extent_end - 1;
+                        } else {
+                                em->end = end;
+                        }
+                        goto not_found_em;
+                }
+                if (btrfs_file_extent_disk_blocknr(item) == 0) {
+                        em->start = extent_start;
+                        em->end = extent_end - 1;
+                        em->block_start = 0;
+                        em->block_end = 0;
+                        goto insert;
+                }
+                em->block_start = blocknr << inode->i_blkbits;
+                em->block_end = em->block_start +
+                        (btrfs_file_extent_num_blocks(item) <<
+                         inode->i_blkbits) - 1;
+                em->start = extent_start;
+                em->end = extent_end - 1;
+                goto insert;
+        } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
+                char *ptr;
+                char *map;
+                u32 size;
+
+                size = btrfs_file_extent_inline_len(leaf->items +
+                                                    path->slots[0]);
+                extent_end = extent_start + size;
+                if (start < extent_start || start > extent_end) {
+                        em->start = start;
+                        if (start < extent_start) {
+                                em->end = extent_end - 1;
+                        } else {
+                                em->end = end;
+                        }
+                        goto not_found_em;
+                }
+                em->block_start = EXTENT_MAP_INLINE;
+                em->block_end = EXTENT_MAP_INLINE;
+                em->start = extent_start;
+                em->end = extent_end - 1;
+                if (!page) {
+                        goto insert;
+                }
+                ptr = btrfs_file_extent_inline_start(item);
+                map = kmap(page);
+                memcpy(map + page_offset, ptr, size);
+                flush_dcache_page(result->b_page);
+                kunmap(page);
+                set_extent_uptodate(em_tree, extent_start,
+                                    extent_end, GFP_NOFS);
+                goto insert;
+        } else {
+                printk("unkknown found_type %d\n", found_type);
+                WARN_ON(1);
+        }
+not_found:
+        em->start = start;
+        em->end = end;
+not_found_em:
+        em->block_start = 0;
+        em->block_end = 0;
+insert:
+        btrfs_release_path(root, path);
+        if (em->start > start || em->end < start) {
+                printk("bad extent! %Lu %Lu start %Lu end %Lu\n", em->start, em->end, start, end);
+                WARN_ON(1);
+                err = -EIO;
+                goto out;
+        }
+        ret = add_extent_mapping(em_tree, em);
+        if (ret == -EEXIST) {
+                free_extent_map(em);
+                failed_insert++;
+                if (failed_insert > 5) {
+                        printk("failing to insert %Lu %Lu\n", start, end);
+                        err = -EIO;
+                        goto out;
+                }
+                em = NULL;
+                goto again;
+        }
+        err = 0;
+out:
+        btrfs_free_path(path);
+        if (trans) {
+                ret = btrfs_end_transaction(trans, root);
+                if (!err)
+                        err = ret;
+        }
+        mutex_unlock(&root->fs_info->fs_mutex);
+        if (err) {
+                free_extent_map(em);
+                WARN_ON(1);
+                return ERR_PTR(err);
+        }
+        return em;
+}
+
+
 /*
  * FIBMAP and others want to pass in a fake buffer head.  They need to
  * use BTRFS_GET_BLOCK_NO_DIRECT to make sure we don't try to memcpy
@@ -1398,46 +1607,22 @@ int btrfs_get_block(struct inode *inode, sector_t iblock,
         return err;
 }
 
-static int btrfs_get_block_csum(struct inode *inode, sector_t iblock,
-                                struct buffer_head *result, int create)
-{
-        int ret;
-        struct btrfs_root *root = BTRFS_I(inode)->root;
-        struct page *page = result->b_page;
-        u64 offset = (page->index << PAGE_CACHE_SHIFT) + bh_offset(result);
-        struct btrfs_csum_item *item;
-        struct btrfs_path *path = NULL;
-
-        mutex_lock(&root->fs_info->fs_mutex);
-        ret = btrfs_get_block_lock(inode, iblock, result, create);
-        if (ret)
-                goto out;
-
-        path = btrfs_alloc_path();
-        item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, offset, 0);
-        if (IS_ERR(item)) {
-                ret = PTR_ERR(item);
-                /* a csum that isn't present is a preallocated region. */
-                if (ret == -ENOENT || ret == -EFBIG)
-                        ret = 0;
-                result->b_private = NULL;
-                goto out;
-        }
-        memcpy((char *)&result->b_private, &item->csum, BTRFS_CRC32_SIZE);
-out:
-        if (path)
-                btrfs_free_path(path);
-        mutex_unlock(&root->fs_info->fs_mutex);
-        return ret;
-}
-
 static int btrfs_get_block_bmap(struct inode *inode, sector_t iblock,
                            struct buffer_head *result, int create)
 {
         struct btrfs_root *root = BTRFS_I(inode)->root;
-        mutex_lock(&root->fs_info->fs_mutex);
-        btrfs_get_block_lock(inode, iblock, result, BTRFS_GET_BLOCK_NO_DIRECT);
-        mutex_unlock(&root->fs_info->fs_mutex);
+        u64 start = iblock << inode->i_blkbits;
+        u64 end = start + root->blocksize -1;
+        struct extent_map *em;
+
+        em = btrfs_get_extent(inode, NULL, 0, start, end, 0);
+        if (em && !IS_ERR(em) && em->block_start != EXTENT_MAP_INLINE &&
+            em->block_start != 0) {
+                u64 offset;
+                offset = start - em->start;
+                start = (em->block_start + offset) >> inode->i_blkbits;
+                btrfs_map_bh_to_logical(root, result, start);
+        }
         return 0;
 }
 
@@ -1449,442 +1634,50 @@ static sector_t btrfs_bmap(struct address_space *as, sector_t block)
 static int btrfs_prepare_write(struct file *file, struct page *page,
                                unsigned from, unsigned to)
 {
-        return block_prepare_write(page, from, to, btrfs_get_block);
+        return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
+                                    page->mapping->host, page, from, to,
+                                    btrfs_get_extent);
 }
 
-static void buffer_io_error(struct buffer_head *bh)
+int btrfs_readpage(struct file *file, struct page *page)
 {
-        char b[BDEVNAME_SIZE];
-
-        printk(KERN_ERR "Buffer I/O error on device %s, logical block %Lu\n",
-                        bdevname(bh->b_bdev, b),
-                        (unsigned long long)bh->b_blocknr);
+        struct extent_map_tree *tree;
+        tree = &BTRFS_I(page->mapping->host)->extent_tree;
+        return extent_read_full_page(tree, page, btrfs_get_extent);
 }
-
-/*
- * I/O completion handler for block_read_full_page() - pages
- * which come unlocked at the end of I/O.
- */
-static void btrfs_end_buffer_async_read(struct buffer_head *bh, int uptodate)
+static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
 {
-        unsigned long flags;
-        struct buffer_head *first;
-        struct buffer_head *tmp;
-        struct page *page;
-        int page_uptodate = 1;
-        struct inode *inode;
-        int ret;
-
-        BUG_ON(!buffer_async_read(bh));
-
-        page = bh->b_page;
-        inode = page->mapping->host;
-        if (uptodate) {
-                void *kaddr;
-                struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
-                if (bh->b_private) {
-                        char csum[BTRFS_CRC32_SIZE];
-                        kaddr = kmap_atomic(page, KM_IRQ0);
-                        ret = btrfs_csum_data(root, kaddr + bh_offset(bh),
-                                              bh->b_size, csum);
-                        BUG_ON(ret);
-                        if (memcmp(csum, &bh->b_private, BTRFS_CRC32_SIZE)) {
-                                u64 offset;
-                                offset = (page->index << PAGE_CACHE_SHIFT) +
-                                        bh_offset(bh);
-                                printk("btrfs csum failed ino %lu off %llu\n",
-                                       page->mapping->host->i_ino,
-                                       (unsigned long long)offset);
-                                memset(kaddr + bh_offset(bh), 1, bh->b_size);
-                                flush_dcache_page(page);
-                        }
-                        kunmap_atomic(kaddr, KM_IRQ0);
-                }
-                set_buffer_uptodate(bh);
-        } else {
-                clear_buffer_uptodate(bh);
-                if (printk_ratelimit())
-                        buffer_io_error(bh);
-                SetPageError(page);
-        }
-
-        /*
-         * Be _very_ careful from here on. Bad things can happen if
-         * two buffer heads end IO at almost the same time and both
-         * decide that the page is now completely done.
-         */
-        first = page_buffers(page);
-        local_irq_save(flags);
-        bit_spin_lock(BH_Uptodate_Lock, &first->b_state);
-        clear_buffer_async_read(bh);
-        unlock_buffer(bh);
-        tmp = bh;
-        do {
-                if (!buffer_uptodate(tmp))
-                        page_uptodate = 0;
-                if (buffer_async_read(tmp)) {
-                        BUG_ON(!buffer_locked(tmp));
-                        goto still_busy;
-                }
-                tmp = tmp->b_this_page;
-        } while (tmp != bh);
-        bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
-        local_irq_restore(flags);
-
-        /*
-         * If none of the buffers had errors and they are all
-         * uptodate then we can set the page uptodate.
-         */
-        if (page_uptodate && !PageError(page))
-                SetPageUptodate(page);
-        unlock_page(page);
-        return;
-
-still_busy:
-        bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
-        local_irq_restore(flags);
-        return;
+        struct extent_map_tree *tree;
+        tree = &BTRFS_I(page->mapping->host)->extent_tree;
+        return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
 }
 
-/*
- * Generic "read page" function for block devices that have the normal
- * get_block functionality. This is most of the block device filesystems.
- * Reads the page asynchronously --- the unlock_buffer() and
- * set/clear_buffer_uptodate() functions propagate buffer state into the
- * page struct once IO has completed.
- */
-int btrfs_readpage(struct file *file, struct page *page)
+static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
 {
-        struct inode *inode = page->mapping->host;
-        sector_t iblock, lblock;
-        struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
-        unsigned int blocksize;
-        int nr, i;
-        int fully_mapped = 1;
-
-        BUG_ON(!PageLocked(page));
-        blocksize = 1 << inode->i_blkbits;
-        if (!page_has_buffers(page))
-                create_empty_buffers(page, blocksize, 0);
-        head = page_buffers(page);
-
-        iblock = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
-        lblock = (i_size_read(inode)+blocksize-1) >> inode->i_blkbits;
-        bh = head;
-        nr = 0;
-        i = 0;
-
-        do {
-                if (buffer_uptodate(bh))
-                        continue;
-
-                if (!buffer_mapped(bh)) {
-                        int err = 0;
-
-                        fully_mapped = 0;
-                        if (iblock < lblock) {
-                                WARN_ON(bh->b_size != blocksize);
-                                err = btrfs_get_block_csum(inode, iblock,
-                                                           bh, 0);
-                                if (err)
-                                        SetPageError(page);
-                        }
-                        if (!buffer_mapped(bh)) {
-                                void *kaddr = kmap_atomic(page, KM_USER0);
-                                memset(kaddr + i * blocksize, 0, blocksize);
-                                flush_dcache_page(page);
-                                kunmap_atomic(kaddr, KM_USER0);
-                                if (!err)
-                                        set_buffer_uptodate(bh);
-                                continue;
-                        }
-                        /*
-                         * get_block() might have updated the buffer
-                         * synchronously
-                         */
-                        if (buffer_uptodate(bh))
-                                continue;
-                }
-                arr[nr++] = bh;
-        } while (i++, iblock++, (bh = bh->b_this_page) != head);
-
-        if (fully_mapped)
-                SetPageMappedToDisk(page);
-
-        if (!nr) {
-                /*
-                 * All buffers are uptodate - we can set the page uptodate
-                 * as well. But not if get_block() returned an error.
-                 */
-                if (!PageError(page))
-                        SetPageUptodate(page);
-                unlock_page(page);
-                return 0;
-        }
-
-        /* Stage two: lock the buffers */
-        for (i = 0; i < nr; i++) {
-                bh = arr[i];
-                lock_buffer(bh);
-                bh->b_end_io = btrfs_end_buffer_async_read;
-                set_buffer_async_read(bh);
-        }
-
-        /*
-         * Stage 3: start the IO.  Check for uptodateness
-         * inside the buffer lock in case another process reading
-         * the underlying blockdev brought it uptodate (the sct fix).
-         */
-        for (i = 0; i < nr; i++) {
-                bh = arr[i];
-                if (buffer_uptodate(bh))
-                        btrfs_end_buffer_async_read(bh, 1);
-                else
-                        submit_bh(READ, bh);
-        }
-        return 0;
-}
-
-/*
- * Aside from a tiny bit of packed file data handling, this is the
- * same as the generic code.
- *
- * While block_write_full_page is writing back the dirty buffers under
- * the page lock, whoever dirtied the buffers may decide to clean them
- * again at any time.  We handle that by only looking at the buffer
- * state inside lock_buffer().
- *
- * If block_write_full_page() is called for regular writeback
- * (wbc->sync_mode == WB_SYNC_NONE) then it will redirty a page which has a
- * locked buffer.   This only can happen if someone has written the buffer
- * directly, with submit_bh().  At the address_space level PageWriteback
- * prevents this contention from occurring.
- */
-static int __btrfs_write_full_page(struct inode *inode, struct page *page,
-                                   struct writeback_control *wbc)
-{
-        int err;
-        sector_t block;
-        sector_t last_block;
-        struct buffer_head *bh, *head;
-        const unsigned blocksize = 1 << inode->i_blkbits;
-        int nr_underway = 0;
-        struct btrfs_root *root = BTRFS_I(inode)->root;
-
-        BUG_ON(!PageLocked(page));
-
-        last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
-
-        /* no csumming allowed when from PF_MEMALLOC */
-        if (current->flags & PF_MEMALLOC) {
-                redirty_page_for_writepage(wbc, page);
-                unlock_page(page);
-                return 0;
-        }
+        struct extent_map_tree *tree;
+        int ret;
 
-        if (!page_has_buffers(page)) {
-                create_empty_buffers(page, blocksize,
-                                        (1 << BH_Dirty)|(1 << BH_Uptodate));
+        if (page->private != 1) {
+                WARN_ON(1);
+                return try_to_free_buffers(page);
         }
-
-        /*
-         * Be very careful.  We have no exclusion from __set_page_dirty_buffers
-         * here, and the (potentially unmapped) buffers may become dirty at
-         * any time.  If a buffer becomes dirty here after we've inspected it
-         * then we just miss that fact, and the page stays dirty.
-         *
-         * Buffers outside i_size may be dirtied by __set_page_dirty_buffers;
-         * handle that here by just cleaning them.
-         */
-
-        block = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
-        head = page_buffers(page);
-        bh = head;
-
-        /*
-         * Get all the dirty buffers mapped to disk addresses and
-         * handle any aliases from the underlying blockdev's mapping.
-         */
-        do {
-                if (block > last_block) {
-                        /*
-                         * mapped buffers outside i_size will occur, because
-                         * this page can be outside i_size when there is a
-                         * truncate in progress.
-                         */
-                        /*
-                         * The buffer was zeroed by block_write_full_page()
-                         */
-                        clear_buffer_dirty(bh);
-                        set_buffer_uptodate(bh);
-                } else if (!buffer_mapped(bh) && buffer_dirty(bh)) {
-                        WARN_ON(bh->b_size != blocksize);
-                        err = btrfs_get_block(inode, block, bh, 0);
-                        if (err) {
-                                goto recover;
-                        }
-                        if (buffer_new(bh)) {
-                                /* blockdev mappings never come here */
-                                clear_buffer_new(bh);
-                        }
-                }
-                bh = bh->b_this_page;
-                block++;
-        } while (bh != head);
-
-        do {
-                if (!buffer_mapped(bh))
-                        continue;
-                /*
-                 * If it's a fully non-blocking write attempt and we cannot
-                 * lock the buffer then redirty the page.  Note that this can
-                 * potentially cause a busy-wait loop from pdflush and kswapd
-                 * activity, but those code paths have their own higher-level
-                 * throttling.
-                 */
-                if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
-                        lock_buffer(bh);
-                } else if (test_set_buffer_locked(bh)) {
-                        redirty_page_for_writepage(wbc, page);
-                        continue;
-                }
-                if (test_clear_buffer_dirty(bh) && bh->b_blocknr != 0) {
-                        struct btrfs_trans_handle *trans;
-                        int ret;
-                        u64 off = page->index << PAGE_CACHE_SHIFT;
-                        char *kaddr;
-
-                        off += bh_offset(bh);
-                        mutex_lock(&root->fs_info->fs_mutex);
-                        trans = btrfs_start_transaction(root, 1);
-                        btrfs_set_trans_block_group(trans, inode);
-                        kaddr = kmap(page);
-                        btrfs_csum_file_block(trans, root, inode->i_ino,
-                                                    off, kaddr + bh_offset(bh),
-                                                    bh->b_size);
-                        kunmap(page);
-                        ret = btrfs_end_transaction(trans, root);
-                        BUG_ON(ret);
-                        mutex_unlock(&root->fs_info->fs_mutex);
-                        mark_buffer_async_write(bh);
-                } else {
-                        unlock_buffer(bh);
-                }
-        } while ((bh = bh->b_this_page) != head);
-
-        /*
-         * The page and its buffers are protected by PageWriteback(), so we can
-         * drop the bh refcounts early.
-         */
-        BUG_ON(PageWriteback(page));
-        set_page_writeback(page);
-
-        do {
-                struct buffer_head *next = bh->b_this_page;
-                if (buffer_async_write(bh)) {
-                        submit_bh(WRITE, bh);
-                        nr_underway++;
-                }
-                bh = next;
-        } while (bh != head);
-        unlock_page(page);
-
-        err = 0;
-done:
-        if (nr_underway == 0) {
-                /*
-                 * The page was marked dirty, but the buffers were
-                 * clean.  Someone wrote them back by hand with
-                 * ll_rw_block/submit_bh.  A rare case.
-                 */
-                int uptodate = 1;
-                do {
-                        if (!buffer_uptodate(bh)) {
-                                uptodate = 0;
-                                break;
-                        }
-                        bh = bh->b_this_page;
-                } while (bh != head);
-                if (uptodate)
-                        SetPageUptodate(page);
-                end_page_writeback(page);
+        tree = &BTRFS_I(page->mapping->host)->extent_tree;
+        ret = try_release_extent_mapping(tree, page);
+        if (ret == 1) {
+                ClearPagePrivate(page);
+                set_page_private(page, 0);
+                page_cache_release(page);
         }
-        return err;
-
-recover:
-        /*
-         * ENOSPC, or some other error.  We may already have added some
-         * blocks to the file, so we need to write these out to avoid
-         * exposing stale data.
-         * The page is currently locked and not marked for writeback
-         */
-        bh = head;
-        /* Recovery: lock and submit the mapped buffers */
-        do {
-                if (buffer_mapped(bh) && buffer_dirty(bh)) {
-                        lock_buffer(bh);
-                        mark_buffer_async_write(bh);
-                } else {
-                        /*
-                         * The buffer may have been set dirty during
-                         * attachment to a dirty page.
-                         */
-                        clear_buffer_dirty(bh);
-                }
-        } while ((bh = bh->b_this_page) != head);
-        SetPageError(page);
-        BUG_ON(PageWriteback(page));
-        set_page_writeback(page);
-        do {
-                struct buffer_head *next = bh->b_this_page;
-                if (buffer_async_write(bh)) {
-                        clear_buffer_dirty(bh);
-                        submit_bh(WRITE, bh);
-                        nr_underway++;
-                }
-                bh = next;
-        } while (bh != head);
-        unlock_page(page);
-        goto done;
+        return ret;
 }
 
-static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
+static void btrfs_invalidatepage(struct page *page, unsigned long offset)
 {
-        struct inode * const inode = page->mapping->host;
-        loff_t i_size = i_size_read(inode);
-        const pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
-        unsigned offset;
-        void *kaddr;
-
-        /* Is the page fully inside i_size? */
-        if (page->index < end_index)
-                return __btrfs_write_full_page(inode, page, wbc);
-
-        /* Is the page fully outside i_size? (truncate in progress) */
-        offset = i_size & (PAGE_CACHE_SIZE-1);
-        if (page->index >= end_index+1 || !offset) {
-                /*
-                 * The page may have dirty, unmapped buffers.  For example,
-                 * they may have been added in ext3_writepage().  Make them
-                 * freeable here, so the page does not leak.
-                 */
-                block_invalidatepage(page, 0);
-                unlock_page(page);
-                return 0; /* don't care */
-        }
+        struct extent_map_tree *tree;
 
-        /*
-         * The page straddles i_size.  It must be zeroed out on each and every
-         * writepage invokation because it may be mmapped.  "A file is mapped
-         * in multiples of the page size.  For a file that is not a multiple of
-         * the  page size, the remaining memory is zeroed when mapped, and
-         * writes to that region are not written out to the file."
-         */
-        kaddr = kmap_atomic(page, KM_USER0);
-        memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
-        flush_dcache_page(page);
-        kunmap_atomic(kaddr, KM_USER0);
-        return __btrfs_write_full_page(inode, page, wbc);
+        tree = &BTRFS_I(page->mapping->host)->extent_tree;
+        extent_invalidatepage(tree, page, offset);
+        btrfs_releasepage(page, GFP_NOFS);
 }
 
 /*
@@ -1905,28 +1698,39 @@ static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
 {
         struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
+        struct btrfs_root *root = BTRFS_I(inode)->root;
+        struct btrfs_trans_handle *trans;
         unsigned long end;
         loff_t size;
         int ret = -EINVAL;
+        u64 page_start;
 
         lock_page(page);
         wait_on_page_writeback(page);
         size = i_size_read(inode);
+        page_start = page->index << PAGE_CACHE_SHIFT;
+
         if ((page->mapping != inode->i_mapping) ||
-            ((page->index << PAGE_CACHE_SHIFT) > size)) {
+            (page_start > size)) {
                 /* page got truncated out from underneath us */
                 goto out_unlock;
         }
 
         /* page is wholly or partially inside EOF */
-        if (((page->index + 1) << PAGE_CACHE_SHIFT) > size)
+        if (page_start + PAGE_CACHE_SIZE > size)
                 end = size & ~PAGE_CACHE_MASK;
         else
                 end = PAGE_CACHE_SIZE;
 
-        ret = btrfs_prepare_write(NULL, page, 0, end);
-        if (!ret)
-                ret = btrfs_commit_write(NULL, page, 0, end);
+        mutex_lock(&root->fs_info->fs_mutex);
+        trans = btrfs_start_transaction(root, 1);
+        ret = btrfs_cow_one_page(trans, inode, page, end);
+        btrfs_end_transaction(trans, root);
+        mutex_unlock(&root->fs_info->fs_mutex);
+        set_extent_dirty(&BTRFS_I(inode)->extent_tree,
+                         page_start, page_start + PAGE_CACHE_SIZE - 1,
+                         GFP_NOFS);
+        set_page_dirty(page);
 
 out_unlock:
         unlock_page(page);
@@ -1962,21 +1766,8 @@ static void btrfs_truncate(struct inode *inode)
 int btrfs_commit_write(struct file *file, struct page *page,
                        unsigned from, unsigned to)
 {
-        struct inode *inode = page->mapping->host;
-        struct buffer_head *bh;
-        loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
-
-        SetPageUptodate(page);
-        bh = page_buffers(page);
-        set_buffer_uptodate(bh);
-        if (buffer_mapped(bh) && bh->b_blocknr != 0) {
-                set_page_dirty(page);
-        }
-        if (pos > inode->i_size) {
-                i_size_write(inode, pos);
-                mark_inode_dirty(inode);
-        }
-        return 0;
+        return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
+                                   page->mapping->host, page, from, to);
 }
 
 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
@@ -2471,6 +2262,8 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
                 inode->i_mapping->a_ops = &btrfs_aops;
                 inode->i_fop = &btrfs_file_operations;
                 inode->i_op = &btrfs_file_inode_operations;
+                extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
+                                     inode->i_mapping, GFP_NOFS);
         }
         dir->i_sb->s_dirt = 1;
         btrfs_update_inode_block_group(trans, inode);
@@ -2553,6 +2346,9 @@ static struct address_space_operations btrfs_aops = {
         .prepare_write  = btrfs_prepare_write,
         .commit_write   = btrfs_commit_write,
         .bmap           = btrfs_bmap,
+        .invalidatepage = btrfs_invalidatepage,
+        .releasepage    = btrfs_releasepage,
+        .set_page_dirty = __set_page_dirty_nobuffers,
 };
 
 static struct address_space_operations btrfs_symlink_aops = {
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 74f3de47423c..7b7c32c73a2c 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -185,6 +185,7 @@ static int __init init_btrfs_fs(void)
         err = btrfs_init_cachep();
         if (err)
                 return err;
+        extent_map_init();
         return register_filesystem(&btrfs_fs_type);
 }
 
@@ -192,6 +193,7 @@ static void __exit exit_btrfs_fs(void)
 {
         btrfs_exit_transaction_sys();
         btrfs_destroy_cachep();
+        extent_map_exit();
         unregister_filesystem(&btrfs_fs_type);
 }