+ u64 bytenr;
+ u64 ref_root;
+ u64 orig_root;
+ u64 ref_generation;
+ u64 orig_generation;
+ struct refsort *sorted;
+ u32 nritems;
+ u32 nr_file_extents = 0;
+ struct btrfs_key key;
+ struct btrfs_file_extent_item *fi;
+ int i;
+ int level;
+ int ret = 0;
+ int faili = 0;
+ int refi = 0;
+ int slot;
+ int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *,
+ u64, u64, u64, u64, u64, u64, u64, u64);
+
+ ref_root = btrfs_header_owner(buf);
+ ref_generation = btrfs_header_generation(buf);
+ orig_root = btrfs_header_owner(orig_buf);
+ orig_generation = btrfs_header_generation(orig_buf);
+
+ nritems = btrfs_header_nritems(buf);
+ level = btrfs_header_level(buf);
+
+ sorted = kmalloc(sizeof(struct refsort) * nritems, GFP_NOFS);
+ BUG_ON(!sorted);
+
+ if (root->ref_cows) {
+ process_func = __btrfs_inc_extent_ref;
+ } else {
+ if (level == 0 &&
+ root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
+ goto out;
+ if (level != 0 &&
+ root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID)
+ goto out;
+ process_func = __btrfs_update_extent_ref;
+ }
+
+ /*
+ * we make two passes through the items. In the first pass we
+ * only record the byte number and slot. Then we sort based on
+ * byte number and do the actual work based on the sorted results
+ */
+ for (i = 0; i < nritems; i++) {
+ cond_resched();
+ if (level == 0) {
+ btrfs_item_key_to_cpu(buf, &key, i);
+ if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
+ continue;
+ fi = btrfs_item_ptr(buf, i,
+ struct btrfs_file_extent_item);
+ if (btrfs_file_extent_type(buf, fi) ==
+ BTRFS_FILE_EXTENT_INLINE)
+ continue;
+ bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
+ if (bytenr == 0)
+ continue;
+
+ nr_file_extents++;
+ sorted[refi].bytenr = bytenr;
+ sorted[refi].slot = i;
+ refi++;
+ } else {
+ bytenr = btrfs_node_blockptr(buf, i);
+ sorted[refi].bytenr = bytenr;
+ sorted[refi].slot = i;
+ refi++;
+ }
+ }
+ /*
+ * if refi == 0, we didn't actually put anything into the sorted
+ * array and we're done
+ */
+ if (refi == 0)
+ goto out;
+
+ sort(sorted, refi, sizeof(struct refsort), refsort_cmp, NULL);
+
+ for (i = 0; i < refi; i++) {
+ cond_resched();
+ slot = sorted[i].slot;
+ bytenr = sorted[i].bytenr;
+
+ if (level == 0) {
+ btrfs_item_key_to_cpu(buf, &key, slot);
+
+ ret = process_func(trans, root, bytenr,
+ orig_buf->start, buf->start,
+ orig_root, ref_root,
+ orig_generation, ref_generation,
+ key.objectid);
+
+ if (ret) {
+ faili = slot;
+ WARN_ON(1);
+ goto fail;
+ }
+ } else {
+ ret = process_func(trans, root, bytenr,
+ orig_buf->start, buf->start,
+ orig_root, ref_root,
+ orig_generation, ref_generation,
+ level - 1);
+ if (ret) {
+ faili = slot;
+ WARN_ON(1);
+ goto fail;
+ }
+ }
+ }
+out:
+ kfree(sorted);
+ if (nr_extents) {
+ if (level == 0)
+ *nr_extents = nr_file_extents;
+ else
+ *nr_extents = nritems;
+ }
+ return 0;
+fail:
+ kfree(sorted);
+ WARN_ON(1);
+ return ret;
+}
+
+int btrfs_update_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, struct extent_buffer *orig_buf,
+ struct extent_buffer *buf, int start_slot, int nr)
+
+{
+ u64 bytenr;
+ u64 ref_root;
+ u64 orig_root;
+ u64 ref_generation;
+ u64 orig_generation;
+ struct btrfs_key key;
+ struct btrfs_file_extent_item *fi;
+ int i;
+ int ret;
+ int slot;
+ int level;
+
+ BUG_ON(start_slot < 0);
+ BUG_ON(start_slot + nr > btrfs_header_nritems(buf));
+
+ ref_root = btrfs_header_owner(buf);
+ ref_generation = btrfs_header_generation(buf);
+ orig_root = btrfs_header_owner(orig_buf);
+ orig_generation = btrfs_header_generation(orig_buf);
+ level = btrfs_header_level(buf);
+
+ if (!root->ref_cows) {
+ if (level == 0 &&
+ root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
+ return 0;
+ if (level != 0 &&
+ root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID)
+ return 0;
+ }
+
+ for (i = 0, slot = start_slot; i < nr; i++, slot++) {
+ cond_resched();
+ if (level == 0) {
+ btrfs_item_key_to_cpu(buf, &key, slot);
+ if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
+ continue;
+ fi = btrfs_item_ptr(buf, slot,
+ struct btrfs_file_extent_item);
+ if (btrfs_file_extent_type(buf, fi) ==
+ BTRFS_FILE_EXTENT_INLINE)
+ continue;
+ bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
+ if (bytenr == 0)
+ continue;
+ ret = __btrfs_update_extent_ref(trans, root, bytenr,
+ orig_buf->start, buf->start,
+ orig_root, ref_root,
+ orig_generation, ref_generation,
+ key.objectid);
+ if (ret)
+ goto fail;
+ } else {
+ bytenr = btrfs_node_blockptr(buf, slot);
+ ret = __btrfs_update_extent_ref(trans, root, bytenr,
+ orig_buf->start, buf->start,
+ orig_root, ref_root,
+ orig_generation, ref_generation,
+ level - 1);
+ if (ret)
+ goto fail;
+ }
+ }
+ return 0;
+fail:
+ WARN_ON(1);
+ return -1;
+}
+
+static int write_one_cache_group(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_path *path,
+ struct btrfs_block_group_cache *cache)
+{
+ int ret;
+ int pending_ret;
+ struct btrfs_root *extent_root = root->fs_info->extent_root;
+ unsigned long bi;
+ struct extent_buffer *leaf;
+
+ ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
+ if (ret < 0)
+ goto fail;
+ BUG_ON(ret);
+
+ leaf = path->nodes[0];
+ bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
+ write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
+ btrfs_mark_buffer_dirty(leaf);
+ btrfs_release_path(extent_root, path);
+fail:
+ finish_current_insert(trans, extent_root, 0);
+ pending_ret = del_pending_extents(trans, extent_root, 0);
+ if (ret)
+ return ret;
+ if (pending_ret)
+ return pending_ret;
+ return 0;
+
+}
+
+int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root)
+{
+ struct btrfs_block_group_cache *cache, *entry;
+ struct rb_node *n;
+ int err = 0;
+ int werr = 0;
+ struct btrfs_path *path;
+ u64 last = 0;
+
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
+
+ while (1) {
+ cache = NULL;
+ spin_lock(&root->fs_info->block_group_cache_lock);
+ for (n = rb_first(&root->fs_info->block_group_cache_tree);
+ n; n = rb_next(n)) {
+ entry = rb_entry(n, struct btrfs_block_group_cache,
+ cache_node);
+ if (entry->dirty) {
+ cache = entry;
+ break;
+ }
+ }
+ spin_unlock(&root->fs_info->block_group_cache_lock);
+
+ if (!cache)
+ break;
+
+ cache->dirty = 0;
+ last += cache->key.offset;
+
+ err = write_one_cache_group(trans, root,
+ path, cache);
+ /*
+ * if we fail to write the cache group, we want
+ * to keep it marked dirty in hopes that a later
+ * write will work
+ */
+ if (err) {
+ werr = err;
+ continue;
+ }
+ }
+ btrfs_free_path(path);
+ return werr;
+}
+
+int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
+{
+ struct btrfs_block_group_cache *block_group;
+ int readonly = 0;
+
+ block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
+ if (!block_group || block_group->ro)
+ readonly = 1;
+ if (block_group)
+ put_block_group(block_group);
+ return readonly;
+}
+
+static int update_space_info(struct btrfs_fs_info *info, u64 flags,
+ u64 total_bytes, u64 bytes_used,
+ struct btrfs_space_info **space_info)
+{
+ struct btrfs_space_info *found;
+
+ found = __find_space_info(info, flags);
+ if (found) {
+ spin_lock(&found->lock);
+ found->total_bytes += total_bytes;
+ found->bytes_used += bytes_used;
+ found->full = 0;
+ spin_unlock(&found->lock);
+ *space_info = found;
+ return 0;
+ }
+ found = kzalloc(sizeof(*found), GFP_NOFS);
+ if (!found)
+ return -ENOMEM;
+
+ list_add(&found->list, &info->space_info);
+ INIT_LIST_HEAD(&found->block_groups);
+ init_rwsem(&found->groups_sem);
+ spin_lock_init(&found->lock);
+ found->flags = flags;
+ found->total_bytes = total_bytes;
+ found->bytes_used = bytes_used;
+ found->bytes_pinned = 0;
+ found->bytes_reserved = 0;
+ found->bytes_readonly = 0;
+ found->bytes_delalloc = 0;
+ found->full = 0;
+ found->force_alloc = 0;
+ *space_info = found;
+ return 0;
+}
+
+static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
+{
+ u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
+ BTRFS_BLOCK_GROUP_RAID1 |
+ BTRFS_BLOCK_GROUP_RAID10 |
+ BTRFS_BLOCK_GROUP_DUP);
+ if (extra_flags) {
+ if (flags & BTRFS_BLOCK_GROUP_DATA)
+ fs_info->avail_data_alloc_bits |= extra_flags;
+ if (flags & BTRFS_BLOCK_GROUP_METADATA)
+ fs_info->avail_metadata_alloc_bits |= extra_flags;
+ if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
+ fs_info->avail_system_alloc_bits |= extra_flags;
+ }
+}
+
+static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
+{
+ spin_lock(&cache->space_info->lock);
+ spin_lock(&cache->lock);
+ if (!cache->ro) {
+ cache->space_info->bytes_readonly += cache->key.offset -
+ btrfs_block_group_used(&cache->item);
+ cache->ro = 1;
+ }
+ spin_unlock(&cache->lock);
+ spin_unlock(&cache->space_info->lock);
+}
+
+u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
+{
+ u64 num_devices = root->fs_info->fs_devices->rw_devices;
+
+ if (num_devices == 1)
+ flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
+ if (num_devices < 4)
+ flags &= ~BTRFS_BLOCK_GROUP_RAID10;
+
+ if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
+ (flags & (BTRFS_BLOCK_GROUP_RAID1 |
+ BTRFS_BLOCK_GROUP_RAID10))) {
+ flags &= ~BTRFS_BLOCK_GROUP_DUP;
+ }
+
+ if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
+ (flags & BTRFS_BLOCK_GROUP_RAID10)) {
+ flags &= ~BTRFS_BLOCK_GROUP_RAID1;
+ }
+
+ if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
+ ((flags & BTRFS_BLOCK_GROUP_RAID1) |
+ (flags & BTRFS_BLOCK_GROUP_RAID10) |
+ (flags & BTRFS_BLOCK_GROUP_DUP)))
+ flags &= ~BTRFS_BLOCK_GROUP_RAID0;
+ return flags;
+}
+
+static u64 btrfs_get_alloc_profile(struct btrfs_root *root, u64 data)
+{
+ struct btrfs_fs_info *info = root->fs_info;
+ u64 alloc_profile;
+
+ if (data) {
+ alloc_profile = info->avail_data_alloc_bits &
+ info->data_alloc_profile;
+ data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
+ } else if (root == root->fs_info->chunk_root) {
+ alloc_profile = info->avail_system_alloc_bits &
+ info->system_alloc_profile;
+ data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
+ } else {
+ alloc_profile = info->avail_metadata_alloc_bits &
+ info->metadata_alloc_profile;
+ data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
+ }
+
+ return btrfs_reduce_alloc_profile(root, data);
+}
+
+void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode)
+{
+ u64 alloc_target;
+
+ alloc_target = btrfs_get_alloc_profile(root, 1);
+ BTRFS_I(inode)->space_info = __find_space_info(root->fs_info,
+ alloc_target);
+}
+
+/*
+ * for now this just makes sure we have at least 5% of our metadata space free
+ * for use.
+ */
+int btrfs_check_metadata_free_space(struct btrfs_root *root)
+{
+ struct btrfs_fs_info *info = root->fs_info;
+ struct btrfs_space_info *meta_sinfo;
+ u64 alloc_target, thresh;
+ int committed = 0, ret;
+
+ /* get the space info for where the metadata will live */
+ alloc_target = btrfs_get_alloc_profile(root, 0);
+ meta_sinfo = __find_space_info(info, alloc_target);
+
+again:
+ spin_lock(&meta_sinfo->lock);
+ if (!meta_sinfo->full)
+ thresh = meta_sinfo->total_bytes * 80;
+ else
+ thresh = meta_sinfo->total_bytes * 95;
+
+ do_div(thresh, 100);
+
+ if (meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
+ meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly > thresh) {
+ struct btrfs_trans_handle *trans;
+ if (!meta_sinfo->full) {
+ meta_sinfo->force_alloc = 1;
+ spin_unlock(&meta_sinfo->lock);
+
+ trans = btrfs_start_transaction(root, 1);
+ if (!trans)
+ return -ENOMEM;
+
+ ret = do_chunk_alloc(trans, root->fs_info->extent_root,
+ 2 * 1024 * 1024, alloc_target, 0);
+ btrfs_end_transaction(trans, root);
+ goto again;
+ }
+ spin_unlock(&meta_sinfo->lock);
+
+ if (!committed) {
+ committed = 1;
+ trans = btrfs_join_transaction(root, 1);
+ if (!trans)
+ return -ENOMEM;
+ ret = btrfs_commit_transaction(trans, root);
+ if (ret)
+ return ret;
+ goto again;
+ }
+ return -ENOSPC;
+ }
+ spin_unlock(&meta_sinfo->lock);
+
+ return 0;
+}
+
+/*
+ * This will check the space that the inode allocates from to make sure we have
+ * enough space for bytes.
+ */
+int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode,
+ u64 bytes)
+{
+ struct btrfs_space_info *data_sinfo;
+ int ret = 0, committed = 0;
+
+ /* make sure bytes are sectorsize aligned */
+ bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
+
+ data_sinfo = BTRFS_I(inode)->space_info;
+again:
+ /* make sure we have enough space to handle the data first */
+ spin_lock(&data_sinfo->lock);
+ if (data_sinfo->total_bytes - data_sinfo->bytes_used -
+ data_sinfo->bytes_delalloc - data_sinfo->bytes_reserved -
+ data_sinfo->bytes_pinned - data_sinfo->bytes_readonly -
+ data_sinfo->bytes_may_use < bytes) {
+ struct btrfs_trans_handle *trans;
+
+ /*
+ * if we don't have enough free bytes in this space then we need
+ * to alloc a new chunk.
+ */
+ if (!data_sinfo->full) {
+ u64 alloc_target;
+
+ data_sinfo->force_alloc = 1;
+ spin_unlock(&data_sinfo->lock);
+
+ alloc_target = btrfs_get_alloc_profile(root, 1);
+ trans = btrfs_start_transaction(root, 1);
+ if (!trans)
+ return -ENOMEM;
+
+ ret = do_chunk_alloc(trans, root->fs_info->extent_root,
+ bytes + 2 * 1024 * 1024,
+ alloc_target, 0);
+ btrfs_end_transaction(trans, root);
+ if (ret)
+ return ret;
+ goto again;
+ }
+ spin_unlock(&data_sinfo->lock);
+
+ /* commit the current transaction and try again */
+ if (!committed) {
+ committed = 1;
+ trans = btrfs_join_transaction(root, 1);
+ if (!trans)
+ return -ENOMEM;
+ ret = btrfs_commit_transaction(trans, root);
+ if (ret)
+ return ret;
+ goto again;
+ }
+
+ printk(KERN_ERR "no space left, need %llu, %llu delalloc bytes"
+ ", %llu bytes_used, %llu bytes_reserved, "
+ "%llu bytes_pinned, %llu bytes_readonly, %llu may use"
+ "%llu total\n", bytes, data_sinfo->bytes_delalloc,
+ data_sinfo->bytes_used, data_sinfo->bytes_reserved,
+ data_sinfo->bytes_pinned, data_sinfo->bytes_readonly,
+ data_sinfo->bytes_may_use, data_sinfo->total_bytes);
+ return -ENOSPC;
+ }
+ data_sinfo->bytes_may_use += bytes;
+ BTRFS_I(inode)->reserved_bytes += bytes;
+ spin_unlock(&data_sinfo->lock);
+
+ return btrfs_check_metadata_free_space(root);
+}
+
+/*
+ * if there was an error for whatever reason after calling
+ * btrfs_check_data_free_space, call this so we can cleanup the counters.
+ */
+void btrfs_free_reserved_data_space(struct btrfs_root *root,
+ struct inode *inode, u64 bytes)
+{
+ struct btrfs_space_info *data_sinfo;
+
+ /* make sure bytes are sectorsize aligned */
+ bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
+
+ data_sinfo = BTRFS_I(inode)->space_info;
+ spin_lock(&data_sinfo->lock);
+ data_sinfo->bytes_may_use -= bytes;
+ BTRFS_I(inode)->reserved_bytes -= bytes;
+ spin_unlock(&data_sinfo->lock);
+}
+
+/* called when we are adding a delalloc extent to the inode's io_tree */
+void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode,
+ u64 bytes)
+{
+ struct btrfs_space_info *data_sinfo;
+
+ /* get the space info for where this inode will be storing its data */
+ data_sinfo = BTRFS_I(inode)->space_info;
+
+ /* make sure we have enough space to handle the data first */
+ spin_lock(&data_sinfo->lock);
+ data_sinfo->bytes_delalloc += bytes;
+
+ /*
+ * we are adding a delalloc extent without calling
+ * btrfs_check_data_free_space first. This happens on a weird
+ * writepage condition, but shouldn't hurt our accounting
+ */
+ if (unlikely(bytes > BTRFS_I(inode)->reserved_bytes)) {
+ data_sinfo->bytes_may_use -= BTRFS_I(inode)->reserved_bytes;
+ BTRFS_I(inode)->reserved_bytes = 0;
+ } else {
+ data_sinfo->bytes_may_use -= bytes;
+ BTRFS_I(inode)->reserved_bytes -= bytes;
+ }
+
+ spin_unlock(&data_sinfo->lock);
+}
+
+/* called when we are clearing an delalloc extent from the inode's io_tree */
+void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
+ u64 bytes)
+{
+ struct btrfs_space_info *info;
+
+ info = BTRFS_I(inode)->space_info;
+
+ spin_lock(&info->lock);
+ info->bytes_delalloc -= bytes;
+ spin_unlock(&info->lock);
+}
+
+static int do_chunk_alloc(struct btrfs_trans_handle *trans,
+ struct btrfs_root *extent_root, u64 alloc_bytes,
+ u64 flags, int force)
+{
+ struct btrfs_space_info *space_info;
+ u64 thresh;
+ int ret = 0;
+
+ mutex_lock(&extent_root->fs_info->chunk_mutex);
+
+ flags = btrfs_reduce_alloc_profile(extent_root, flags);
+
+ space_info = __find_space_info(extent_root->fs_info, flags);
+ if (!space_info) {
+ ret = update_space_info(extent_root->fs_info, flags,
+ 0, 0, &space_info);
+ BUG_ON(ret);
+ }
+ BUG_ON(!space_info);
+
+ spin_lock(&space_info->lock);
+ if (space_info->force_alloc) {
+ force = 1;
+ space_info->force_alloc = 0;
+ }
+ if (space_info->full) {
+ spin_unlock(&space_info->lock);
+ goto out;
+ }
+
+ thresh = space_info->total_bytes - space_info->bytes_readonly;
+ thresh = div_factor(thresh, 6);
+ if (!force &&
+ (space_info->bytes_used + space_info->bytes_pinned +
+ space_info->bytes_reserved + alloc_bytes) < thresh) {
+ spin_unlock(&space_info->lock);
+ goto out;
+ }
+ spin_unlock(&space_info->lock);
+
+ ret = btrfs_alloc_chunk(trans, extent_root, flags);
+ if (ret)
+ space_info->full = 1;
+out:
+ mutex_unlock(&extent_root->fs_info->chunk_mutex);
+ return ret;
+}
+
+static int update_block_group(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ u64 bytenr, u64 num_bytes, int alloc,
+ int mark_free)
+{
+ struct btrfs_block_group_cache *cache;
+ struct btrfs_fs_info *info = root->fs_info;
+ u64 total = num_bytes;
+ u64 old_val;
+ u64 byte_in_group;
+
+ while (total) {
+ cache = btrfs_lookup_block_group(info, bytenr);
+ if (!cache)
+ return -1;
+ byte_in_group = bytenr - cache->key.objectid;
+ WARN_ON(byte_in_group > cache->key.offset);
+
+ spin_lock(&cache->space_info->lock);
+ spin_lock(&cache->lock);
+ cache->dirty = 1;
+ old_val = btrfs_block_group_used(&cache->item);
+ num_bytes = min(total, cache->key.offset - byte_in_group);
+ if (alloc) {
+ old_val += num_bytes;
+ cache->space_info->bytes_used += num_bytes;
+ if (cache->ro)
+ cache->space_info->bytes_readonly -= num_bytes;
+ btrfs_set_block_group_used(&cache->item, old_val);
+ spin_unlock(&cache->lock);
+ spin_unlock(&cache->space_info->lock);
+ } else {
+ old_val -= num_bytes;
+ cache->space_info->bytes_used -= num_bytes;
+ if (cache->ro)
+ cache->space_info->bytes_readonly += num_bytes;
+ btrfs_set_block_group_used(&cache->item, old_val);
+ spin_unlock(&cache->lock);
+ spin_unlock(&cache->space_info->lock);
+ if (mark_free) {
+ int ret;
+
+ ret = btrfs_discard_extent(root, bytenr,
+ num_bytes);
+ WARN_ON(ret);
+
+ ret = btrfs_add_free_space(cache, bytenr,
+ num_bytes);
+ WARN_ON(ret);
+ }
+ }
+ put_block_group(cache);
+ total -= num_bytes;
+ bytenr += num_bytes;
+ }
+ return 0;
+}
+
+static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
+{
+ struct btrfs_block_group_cache *cache;
+ u64 bytenr;
+
+ cache = btrfs_lookup_first_block_group(root->fs_info, search_start);
+ if (!cache)
+ return 0;
+
+ bytenr = cache->key.objectid;
+ put_block_group(cache);
+
+ return bytenr;
+}
+
+int btrfs_update_pinned_extents(struct btrfs_root *root,
+ u64 bytenr, u64 num, int pin)
+{
+ u64 len;
+ struct btrfs_block_group_cache *cache;
+ struct btrfs_fs_info *fs_info = root->fs_info;
+
+ WARN_ON(!mutex_is_locked(&root->fs_info->pinned_mutex));
+ if (pin) {
+ set_extent_dirty(&fs_info->pinned_extents,
+ bytenr, bytenr + num - 1, GFP_NOFS);
+ } else {
+ clear_extent_dirty(&fs_info->pinned_extents,
+ bytenr, bytenr + num - 1, GFP_NOFS);
+ }
+ while (num > 0) {
+ cache = btrfs_lookup_block_group(fs_info, bytenr);
+ BUG_ON(!cache);
+ len = min(num, cache->key.offset -
+ (bytenr - cache->key.objectid));
+ if (pin) {
+ spin_lock(&cache->space_info->lock);
+ spin_lock(&cache->lock);
+ cache->pinned += len;
+ cache->space_info->bytes_pinned += len;
+ spin_unlock(&cache->lock);
+ spin_unlock(&cache->space_info->lock);
+ fs_info->total_pinned += len;
+ } else {
+ spin_lock(&cache->space_info->lock);
+ spin_lock(&cache->lock);
+ cache->pinned -= len;
+ cache->space_info->bytes_pinned -= len;
+ spin_unlock(&cache->lock);
+ spin_unlock(&cache->space_info->lock);
+ fs_info->total_pinned -= len;
+ if (cache->cached)
+ btrfs_add_free_space(cache, bytenr, len);
+ }
+ put_block_group(cache);
+ bytenr += len;
+ num -= len;
+ }
+ return 0;
+}
+
+static int update_reserved_extents(struct btrfs_root *root,
+ u64 bytenr, u64 num, int reserve)
+{
+ u64 len;
+ struct btrfs_block_group_cache *cache;
+ struct btrfs_fs_info *fs_info = root->fs_info;
+
+ while (num > 0) {
+ cache = btrfs_lookup_block_group(fs_info, bytenr);
+ BUG_ON(!cache);
+ len = min(num, cache->key.offset -
+ (bytenr - cache->key.objectid));
+
+ spin_lock(&cache->space_info->lock);
+ spin_lock(&cache->lock);
+ if (reserve) {
+ cache->reserved += len;
+ cache->space_info->bytes_reserved += len;
+ } else {
+ cache->reserved -= len;
+ cache->space_info->bytes_reserved -= len;
+ }
+ spin_unlock(&cache->lock);
+ spin_unlock(&cache->space_info->lock);
+ put_block_group(cache);
+ bytenr += len;
+ num -= len;
+ }
+ return 0;
+}
+
+int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy)
+{
+ u64 last = 0;
+ u64 start;
+ u64 end;
+ struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents;
+ int ret;
+
+ mutex_lock(&root->fs_info->pinned_mutex);
+ while (1) {
+ ret = find_first_extent_bit(pinned_extents, last,
+ &start, &end, EXTENT_DIRTY);
+ if (ret)
+ break;
+ set_extent_dirty(copy, start, end, GFP_NOFS);
+ last = end + 1;
+ }
+ mutex_unlock(&root->fs_info->pinned_mutex);
+ return 0;
+}
+
+int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct extent_io_tree *unpin)
+{
+ u64 start;
+ u64 end;
+ int ret;
+
+ mutex_lock(&root->fs_info->pinned_mutex);
+ while (1) {
+ ret = find_first_extent_bit(unpin, 0, &start, &end,
+ EXTENT_DIRTY);
+ if (ret)
+ break;
+
+ ret = btrfs_discard_extent(root, start, end + 1 - start);
+
+ btrfs_update_pinned_extents(root, start, end + 1 - start, 0);
+ clear_extent_dirty(unpin, start, end, GFP_NOFS);
+
+ if (need_resched()) {
+ mutex_unlock(&root->fs_info->pinned_mutex);
+ cond_resched();
+ mutex_lock(&root->fs_info->pinned_mutex);
+ }
+ }
+ mutex_unlock(&root->fs_info->pinned_mutex);
+ return ret;
+}
+
+static int finish_current_insert(struct btrfs_trans_handle *trans,
+ struct btrfs_root *extent_root, int all)
+{
+ u64 start;
+ u64 end;
+ u64 priv;
+ u64 search = 0;
+ struct btrfs_fs_info *info = extent_root->fs_info;
+ struct btrfs_path *path;
+ struct pending_extent_op *extent_op, *tmp;
+ struct list_head insert_list, update_list;
+ int ret;
+ int num_inserts = 0, max_inserts, restart = 0;
+
+ path = btrfs_alloc_path();
+ INIT_LIST_HEAD(&insert_list);
+ INIT_LIST_HEAD(&update_list);
+
+ max_inserts = extent_root->leafsize /
+ (2 * sizeof(struct btrfs_key) + 2 * sizeof(struct btrfs_item) +
+ sizeof(struct btrfs_extent_ref) +
+ sizeof(struct btrfs_extent_item));
+again:
+ mutex_lock(&info->extent_ins_mutex);
+ while (1) {
+ ret = find_first_extent_bit(&info->extent_ins, search, &start,
+ &end, EXTENT_WRITEBACK);
+ if (ret) {
+ if (restart && !num_inserts &&
+ list_empty(&update_list)) {
+ restart = 0;
+ search = 0;
+ continue;
+ }
+ break;
+ }
+
+ ret = try_lock_extent(&info->extent_ins, start, end, GFP_NOFS);
+ if (!ret) {
+ if (all)
+ restart = 1;
+ search = end + 1;
+ if (need_resched()) {
+ mutex_unlock(&info->extent_ins_mutex);
+ cond_resched();
+ mutex_lock(&info->extent_ins_mutex);
+ }
+ continue;
+ }
+
+ ret = get_state_private(&info->extent_ins, start, &priv);
+ BUG_ON(ret);
+ extent_op = (struct pending_extent_op *)(unsigned long) priv;
+
+ if (extent_op->type == PENDING_EXTENT_INSERT) {
+ num_inserts++;
+ list_add_tail(&extent_op->list, &insert_list);
+ search = end + 1;
+ if (num_inserts == max_inserts) {
+ restart = 1;
+ break;
+ }
+ } else if (extent_op->type == PENDING_BACKREF_UPDATE) {
+ list_add_tail(&extent_op->list, &update_list);
+ search = end + 1;
+ } else {
+ BUG();
+ }
+ }
+
+ /*
+ * process the update list, clear the writeback bit for it, and if
+ * somebody marked this thing for deletion then just unlock it and be
+ * done, the free_extents will handle it
+ */
+ list_for_each_entry_safe(extent_op, tmp, &update_list, list) {
+ clear_extent_bits(&info->extent_ins, extent_op->bytenr,
+ extent_op->bytenr + extent_op->num_bytes - 1,
+ EXTENT_WRITEBACK, GFP_NOFS);
+ if (extent_op->del) {
+ list_del_init(&extent_op->list);
+ unlock_extent(&info->extent_ins, extent_op->bytenr,
+ extent_op->bytenr + extent_op->num_bytes
+ - 1, GFP_NOFS);
+ kfree(extent_op);
+ }
+ }
+ mutex_unlock(&info->extent_ins_mutex);
+
+ /*
+ * still have things left on the update list, go ahead an update
+ * everything
+ */
+ if (!list_empty(&update_list)) {
+ ret = update_backrefs(trans, extent_root, path, &update_list);
+ BUG_ON(ret);
+
+ /* we may have COW'ed new blocks, so lets start over */
+ if (all)
+ restart = 1;
+ }
+
+ /*
+ * if no inserts need to be done, but we skipped some extents and we
+ * need to make sure everything is cleaned then reset everything and
+ * go back to the beginning
+ */
+ if (!num_inserts && restart) {
+ search = 0;
+ restart = 0;
+ INIT_LIST_HEAD(&update_list);
+ INIT_LIST_HEAD(&insert_list);
+ goto again;
+ } else if (!num_inserts) {
+ goto out;
+ }
+
+ /*
+ * process the insert extents list. Again if we are deleting this
+ * extent, then just unlock it, pin down the bytes if need be, and be
+ * done with it. Saves us from having to actually insert the extent
+ * into the tree and then subsequently come along and delete it
+ */
+ mutex_lock(&info->extent_ins_mutex);
+ list_for_each_entry_safe(extent_op, tmp, &insert_list, list) {
+ clear_extent_bits(&info->extent_ins, extent_op->bytenr,
+ extent_op->bytenr + extent_op->num_bytes - 1,
+ EXTENT_WRITEBACK, GFP_NOFS);
+ if (extent_op->del) {
+ u64 used;
+ list_del_init(&extent_op->list);
+ unlock_extent(&info->extent_ins, extent_op->bytenr,
+ extent_op->bytenr + extent_op->num_bytes
+ - 1, GFP_NOFS);
+
+ mutex_lock(&extent_root->fs_info->pinned_mutex);
+ ret = pin_down_bytes(trans, extent_root,
+ extent_op->bytenr,
+ extent_op->num_bytes, 0);
+ mutex_unlock(&extent_root->fs_info->pinned_mutex);
+
+ spin_lock(&info->delalloc_lock);
+ used = btrfs_super_bytes_used(&info->super_copy);
+ btrfs_set_super_bytes_used(&info->super_copy,
+ used - extent_op->num_bytes);
+ used = btrfs_root_used(&extent_root->root_item);
+ btrfs_set_root_used(&extent_root->root_item,
+ used - extent_op->num_bytes);
+ spin_unlock(&info->delalloc_lock);
+
+ ret = update_block_group(trans, extent_root,
+ extent_op->bytenr,
+ extent_op->num_bytes,
+ 0, ret > 0);
+ BUG_ON(ret);
+ kfree(extent_op);
+ num_inserts--;
+ }
+ }
+ mutex_unlock(&info->extent_ins_mutex);
+
+ ret = insert_extents(trans, extent_root, path, &insert_list,
+ num_inserts);
+ BUG_ON(ret);
+
+ /*
+ * if restart is set for whatever reason we need to go back and start
+ * searching through the pending list again.
+ *
+ * We just inserted some extents, which could have resulted in new
+ * blocks being allocated, which would result in new blocks needing
+ * updates, so if all is set we _must_ restart to get the updated
+ * blocks.
+ */
+ if (restart || all) {
+ INIT_LIST_HEAD(&insert_list);
+ INIT_LIST_HEAD(&update_list);
+ search = 0;
+ restart = 0;
+ num_inserts = 0;
+ goto again;
+ }
+out:
+ btrfs_free_path(path);
+ return 0;
+}
+
+static int pin_down_bytes(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ u64 bytenr, u64 num_bytes, int is_data)
+{
+ int err = 0;
+ struct extent_buffer *buf;
+
+ if (is_data)
+ goto pinit;
+
+ buf = btrfs_find_tree_block(root, bytenr, num_bytes);
+ if (!buf)
+ goto pinit;
+
+ /* we can reuse a block if it hasn't been written
+ * and it is from this transaction. We can't
+ * reuse anything from the tree log root because
+ * it has tiny sub-transactions.
+ */
+ if (btrfs_buffer_uptodate(buf, 0) &&
+ btrfs_try_tree_lock(buf)) {
+ u64 header_owner = btrfs_header_owner(buf);
+ u64 header_transid = btrfs_header_generation(buf);
+ if (header_owner != BTRFS_TREE_LOG_OBJECTID &&
+ header_owner != BTRFS_TREE_RELOC_OBJECTID &&
+ header_transid == trans->transid &&
+ !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
+ clean_tree_block(NULL, root, buf);
+ btrfs_tree_unlock(buf);
+ free_extent_buffer(buf);
+ return 1;
+ }
+ btrfs_tree_unlock(buf);
+ }
+ free_extent_buffer(buf);
+pinit:
+ btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
+
+ BUG_ON(err < 0);
+ return 0;
+}
+
+/*
+ * remove an extent from the root, returns 0 on success
+ */
+static int __free_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ u64 bytenr, u64 num_bytes, u64 parent,
+ u64 root_objectid, u64 ref_generation,
+ u64 owner_objectid, int pin, int mark_free)
+{
+ struct btrfs_path *path;
+ struct btrfs_key key;
+ struct btrfs_fs_info *info = root->fs_info;
+ struct btrfs_root *extent_root = info->extent_root;
+ struct extent_buffer *leaf;
+ int ret;
+ int extent_slot = 0;
+ int found_extent = 0;
+ int num_to_del = 1;
+ struct btrfs_extent_item *ei;
+ u32 refs;
+
+ key.objectid = bytenr;
+ btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
+ key.offset = num_bytes;
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
+
+ path->reada = 1;
+ ret = lookup_extent_backref(trans, extent_root, path,
+ bytenr, parent, root_objectid,
+ ref_generation, owner_objectid, 1);
+ if (ret == 0) {
+ struct btrfs_key found_key;
+ extent_slot = path->slots[0];
+ while (extent_slot > 0) {
+ extent_slot--;
+ btrfs_item_key_to_cpu(path->nodes[0], &found_key,
+ extent_slot);
+ if (found_key.objectid != bytenr)
+ break;
+ if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
+ found_key.offset == num_bytes) {
+ found_extent = 1;
+ break;
+ }
+ if (path->slots[0] - extent_slot > 5)
+ break;
+ }
+ if (!found_extent) {
+ ret = remove_extent_backref(trans, extent_root, path);
+ BUG_ON(ret);
+ btrfs_release_path(extent_root, path);
+ ret = btrfs_search_slot(trans, extent_root,
+ &key, path, -1, 1);
+ if (ret) {
+ printk(KERN_ERR "umm, got %d back from search"
+ ", was looking for %llu\n", ret,
+ (unsigned long long)bytenr);
+ btrfs_print_leaf(extent_root, path->nodes[0]);
+ }
+ BUG_ON(ret);
+ extent_slot = path->slots[0];
+ }
+ } else {
+ btrfs_print_leaf(extent_root, path->nodes[0]);
+ WARN_ON(1);
+ printk(KERN_ERR "btrfs unable to find ref byte nr %llu "
+ "root %llu gen %llu owner %llu\n",
+ (unsigned long long)bytenr,
+ (unsigned long long)root_objectid,
+ (unsigned long long)ref_generation,
+ (unsigned long long)owner_objectid);
+ }
+
+ leaf = path->nodes[0];
+ ei = btrfs_item_ptr(leaf, extent_slot,
+ struct btrfs_extent_item);
+ refs = btrfs_extent_refs(leaf, ei);
+ BUG_ON(refs == 0);
+ refs -= 1;
+ btrfs_set_extent_refs(leaf, ei, refs);
+
+ btrfs_mark_buffer_dirty(leaf);
+
+ if (refs == 0 && found_extent && path->slots[0] == extent_slot + 1) {
+ struct btrfs_extent_ref *ref;
+ ref = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_extent_ref);
+ BUG_ON(btrfs_ref_num_refs(leaf, ref) != 1);
+ /* if the back ref and the extent are next to each other
+ * they get deleted below in one shot
+ */
+ path->slots[0] = extent_slot;
+ num_to_del = 2;
+ } else if (found_extent) {
+ /* otherwise delete the extent back ref */
+ ret = remove_extent_backref(trans, extent_root, path);
+ BUG_ON(ret);
+ /* if refs are 0, we need to setup the path for deletion */
+ if (refs == 0) {
+ btrfs_release_path(extent_root, path);
+ ret = btrfs_search_slot(trans, extent_root, &key, path,
+ -1, 1);
+ BUG_ON(ret);
+ }
+ }
+
+ if (refs == 0) {
+ u64 super_used;
+ u64 root_used;
+
+ if (pin) {
+ mutex_lock(&root->fs_info->pinned_mutex);
+ ret = pin_down_bytes(trans, root, bytenr, num_bytes,
+ owner_objectid >= BTRFS_FIRST_FREE_OBJECTID);
+ mutex_unlock(&root->fs_info->pinned_mutex);
+ if (ret > 0)
+ mark_free = 1;
+ BUG_ON(ret < 0);
+ }
+ /* block accounting for super block */
+ spin_lock(&info->delalloc_lock);
+ super_used = btrfs_super_bytes_used(&info->super_copy);
+ btrfs_set_super_bytes_used(&info->super_copy,
+ super_used - num_bytes);
+
+ /* block accounting for root item */
+ root_used = btrfs_root_used(&root->root_item);
+ btrfs_set_root_used(&root->root_item,
+ root_used - num_bytes);
+ spin_unlock(&info->delalloc_lock);
+ ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
+ num_to_del);
+ BUG_ON(ret);
+ btrfs_release_path(extent_root, path);
+
+ if (owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
+ ret = btrfs_del_csums(trans, root, bytenr, num_bytes);
+ BUG_ON(ret);
+ }
+
+ ret = update_block_group(trans, root, bytenr, num_bytes, 0,
+ mark_free);
+ BUG_ON(ret);
+ }
+ btrfs_free_path(path);
+ finish_current_insert(trans, extent_root, 0);
+ return ret;
+}
+
+/*
+ * find all the blocks marked as pending in the radix tree and remove
+ * them from the extent map
+ */
+static int del_pending_extents(struct btrfs_trans_handle *trans,
+ struct btrfs_root *extent_root, int all)
+{
+ int ret;
+ int err = 0;
+ u64 start;
+ u64 end;
+ u64 priv;
+ u64 search = 0;
+ int nr = 0, skipped = 0;
+ struct extent_io_tree *pending_del;
+ struct extent_io_tree *extent_ins;
+ struct pending_extent_op *extent_op;
+ struct btrfs_fs_info *info = extent_root->fs_info;
+ struct list_head delete_list;
+
+ INIT_LIST_HEAD(&delete_list);
+ extent_ins = &extent_root->fs_info->extent_ins;
+ pending_del = &extent_root->fs_info->pending_del;
+
+again:
+ mutex_lock(&info->extent_ins_mutex);
+ while (1) {
+ ret = find_first_extent_bit(pending_del, search, &start, &end,
+ EXTENT_WRITEBACK);
+ if (ret) {
+ if (all && skipped && !nr) {
+ search = 0;
+ skipped = 0;
+ continue;
+ }
+ mutex_unlock(&info->extent_ins_mutex);
+ break;
+ }
+
+ ret = try_lock_extent(extent_ins, start, end, GFP_NOFS);
+ if (!ret) {
+ search = end+1;
+ skipped = 1;
+
+ if (need_resched()) {
+ mutex_unlock(&info->extent_ins_mutex);
+ cond_resched();
+ mutex_lock(&info->extent_ins_mutex);
+ }
+
+ continue;
+ }
+ BUG_ON(ret < 0);
+
+ ret = get_state_private(pending_del, start, &priv);
+ BUG_ON(ret);
+ extent_op = (struct pending_extent_op *)(unsigned long)priv;
+
+ clear_extent_bits(pending_del, start, end, EXTENT_WRITEBACK,
+ GFP_NOFS);
+ if (!test_range_bit(extent_ins, start, end,
+ EXTENT_WRITEBACK, 0)) {
+ list_add_tail(&extent_op->list, &delete_list);
+ nr++;
+ } else {
+ kfree(extent_op);
+
+ ret = get_state_private(&info->extent_ins, start,
+ &priv);
+ BUG_ON(ret);
+ extent_op = (struct pending_extent_op *)
+ (unsigned long)priv;
+
+ clear_extent_bits(&info->extent_ins, start, end,
+ EXTENT_WRITEBACK, GFP_NOFS);
+
+ if (extent_op->type == PENDING_BACKREF_UPDATE) {
+ list_add_tail(&extent_op->list, &delete_list);
+ search = end + 1;
+ nr++;
+ continue;
+ }
+
+ mutex_lock(&extent_root->fs_info->pinned_mutex);
+ ret = pin_down_bytes(trans, extent_root, start,
+ end + 1 - start, 0);
+ mutex_unlock(&extent_root->fs_info->pinned_mutex);
+
+ ret = update_block_group(trans, extent_root, start,
+ end + 1 - start, 0, ret > 0);
+
+ unlock_extent(extent_ins, start, end, GFP_NOFS);
+ BUG_ON(ret);
+ kfree(extent_op);
+ }
+ if (ret)
+ err = ret;
+
+ search = end + 1;
+
+ if (need_resched()) {
+ mutex_unlock(&info->extent_ins_mutex);
+ cond_resched();
+ mutex_lock(&info->extent_ins_mutex);
+ }
+ }
+
+ if (nr) {
+ ret = free_extents(trans, extent_root, &delete_list);
+ BUG_ON(ret);
+ }
+
+ if (all && skipped) {
+ INIT_LIST_HEAD(&delete_list);
+ search = 0;
+ nr = 0;
+ goto again;
+ }
+
+ if (!err)
+ finish_current_insert(trans, extent_root, 0);
+ return err;
+}
+
+/*
+ * remove an extent from the root, returns 0 on success
+ */
+static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ u64 bytenr, u64 num_bytes, u64 parent,
+ u64 root_objectid, u64 ref_generation,
+ u64 owner_objectid, int pin)
+{
+ struct btrfs_root *extent_root = root->fs_info->extent_root;
+ int pending_ret;
+ int ret;
+
+ WARN_ON(num_bytes < root->sectorsize);
+ if (root == extent_root) {
+ struct pending_extent_op *extent_op = NULL;
+
+ mutex_lock(&root->fs_info->extent_ins_mutex);
+ if (test_range_bit(&root->fs_info->extent_ins, bytenr,
+ bytenr + num_bytes - 1, EXTENT_WRITEBACK, 0)) {
+ u64 priv;
+ ret = get_state_private(&root->fs_info->extent_ins,
+ bytenr, &priv);
+ BUG_ON(ret);
+ extent_op = (struct pending_extent_op *)
+ (unsigned long)priv;
+
+ extent_op->del = 1;
+ if (extent_op->type == PENDING_EXTENT_INSERT) {
+ mutex_unlock(&root->fs_info->extent_ins_mutex);
+ return 0;
+ }
+ }
+
+ if (extent_op) {
+ ref_generation = extent_op->orig_generation;
+ parent = extent_op->orig_parent;
+ }
+
+ extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
+ BUG_ON(!extent_op);
+
+ extent_op->type = PENDING_EXTENT_DELETE;
+ extent_op->bytenr = bytenr;
+ extent_op->num_bytes = num_bytes;
+ extent_op->parent = parent;
+ extent_op->orig_parent = parent;
+ extent_op->generation = ref_generation;
+ extent_op->orig_generation = ref_generation;
+ extent_op->level = (int)owner_objectid;
+ INIT_LIST_HEAD(&extent_op->list);
+ extent_op->del = 0;
+
+ set_extent_bits(&root->fs_info->pending_del,
+ bytenr, bytenr + num_bytes - 1,
+ EXTENT_WRITEBACK, GFP_NOFS);
+ set_state_private(&root->fs_info->pending_del,
+ bytenr, (unsigned long)extent_op);
+ mutex_unlock(&root->fs_info->extent_ins_mutex);
+ return 0;
+ }
+ /* if metadata always pin */
+ if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
+ if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
+ mutex_lock(&root->fs_info->pinned_mutex);
+ btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
+ mutex_unlock(&root->fs_info->pinned_mutex);
+ update_reserved_extents(root, bytenr, num_bytes, 0);
+ return 0;
+ }
+ pin = 1;
+ }
+
+ /* if data pin when any transaction has committed this */
+ if (ref_generation != trans->transid)
+ pin = 1;
+
+ ret = __free_extent(trans, root, bytenr, num_bytes, parent,
+ root_objectid, ref_generation,
+ owner_objectid, pin, pin == 0);
+
+ finish_current_insert(trans, root->fs_info->extent_root, 0);
+ pending_ret = del_pending_extents(trans, root->fs_info->extent_root, 0);
+ return ret ? ret : pending_ret;
+}
+
+int btrfs_free_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ u64 bytenr, u64 num_bytes, u64 parent,
+ u64 root_objectid, u64 ref_generation,
+ u64 owner_objectid, int pin)
+{
+ int ret;
+
+ ret = __btrfs_free_extent(trans, root, bytenr, num_bytes, parent,
+ root_objectid, ref_generation,
+ owner_objectid, pin);
+ return ret;
+}
+
+static u64 stripe_align(struct btrfs_root *root, u64 val)
+{
+ u64 mask = ((u64)root->stripesize - 1);
+ u64 ret = (val + mask) & ~mask;
+ return ret;
+}
+
+/*
+ * walks the btree of allocated extents and find a hole of a given size.
+ * The key ins is changed to record the hole:
+ * ins->objectid == block start
+ * ins->flags = BTRFS_EXTENT_ITEM_KEY
+ * ins->offset == number of blocks
+ * Any available blocks before search_start are skipped.
+ */
+static noinline int find_free_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_root *orig_root,
+ u64 num_bytes, u64 empty_size,
+ u64 search_start, u64 search_end,
+ u64 hint_byte, struct btrfs_key *ins,
+ u64 exclude_start, u64 exclude_nr,
+ int data)
+{
+ int ret = 0;
+ struct btrfs_root *root = orig_root->fs_info->extent_root;
+ u64 total_needed = num_bytes;
+ u64 *last_ptr = NULL;
+ u64 last_wanted = 0;
+ struct btrfs_block_group_cache *block_group = NULL;
+ int chunk_alloc_done = 0;
+ int empty_cluster = 2 * 1024 * 1024;
+ int allowed_chunk_alloc = 0;
+ struct list_head *head = NULL, *cur = NULL;
+ int loop = 0;
+ int extra_loop = 0;
+ struct btrfs_space_info *space_info;
+
+ WARN_ON(num_bytes < root->sectorsize);
+ btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
+ ins->objectid = 0;
+ ins->offset = 0;
+
+ if (orig_root->ref_cows || empty_size)
+ allowed_chunk_alloc = 1;
+
+ if (data & BTRFS_BLOCK_GROUP_METADATA) {
+ last_ptr = &root->fs_info->last_alloc;
+ if (!btrfs_test_opt(root, SSD))
+ empty_cluster = 64 * 1024;
+ }
+
+ if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD))
+ last_ptr = &root->fs_info->last_data_alloc;
+
+ if (last_ptr) {
+ if (*last_ptr) {
+ hint_byte = *last_ptr;
+ last_wanted = *last_ptr;
+ } else
+ empty_size += empty_cluster;
+ } else {
+ empty_cluster = 0;
+ }
+ search_start = max(search_start, first_logical_byte(root, 0));
+ search_start = max(search_start, hint_byte);
+
+ if (last_wanted && search_start != last_wanted) {
+ last_wanted = 0;
+ empty_size += empty_cluster;
+ }
+
+ total_needed += empty_size;
+ block_group = btrfs_lookup_block_group(root->fs_info, search_start);
+ if (!block_group)
+ block_group = btrfs_lookup_first_block_group(root->fs_info,
+ search_start);
+ space_info = __find_space_info(root->fs_info, data);
+
+ down_read(&space_info->groups_sem);
+ while (1) {
+ struct btrfs_free_space *free_space;
+ /*
+ * the only way this happens if our hint points to a block
+ * group thats not of the proper type, while looping this
+ * should never happen
+ */
+ if (empty_size)
+ extra_loop = 1;
+
+ if (!block_group)
+ goto new_group_no_lock;
+
+ if (unlikely(!block_group->cached)) {
+ mutex_lock(&block_group->cache_mutex);
+ ret = cache_block_group(root, block_group);
+ mutex_unlock(&block_group->cache_mutex);
+ if (ret)
+ break;
+ }
+
+ mutex_lock(&block_group->alloc_mutex);
+ if (unlikely(!block_group_bits(block_group, data)))
+ goto new_group;
+
+ if (unlikely(block_group->ro))
+ goto new_group;
+
+ free_space = btrfs_find_free_space(block_group, search_start,
+ total_needed);
+ if (free_space) {
+ u64 start = block_group->key.objectid;
+ u64 end = block_group->key.objectid +
+ block_group->key.offset;
+
+ search_start = stripe_align(root, free_space->offset);
+
+ /* move on to the next group */
+ if (search_start + num_bytes >= search_end)
+ goto new_group;
+
+ /* move on to the next group */
+ if (search_start + num_bytes > end)
+ goto new_group;
+
+ if (last_wanted && search_start != last_wanted) {
+ total_needed += empty_cluster;
+ empty_size += empty_cluster;
+ last_wanted = 0;
+ /*
+ * if search_start is still in this block group
+ * then we just re-search this block group
+ */
+ if (search_start >= start &&
+ search_start < end) {
+ mutex_unlock(&block_group->alloc_mutex);
+ continue;
+ }
+
+ /* else we go to the next block group */
+ goto new_group;
+ }
+
+ if (exclude_nr > 0 &&
+ (search_start + num_bytes > exclude_start &&
+ search_start < exclude_start + exclude_nr)) {
+ search_start = exclude_start + exclude_nr;
+ /*
+ * if search_start is still in this block group
+ * then we just re-search this block group
+ */
+ if (search_start >= start &&
+ search_start < end) {
+ mutex_unlock(&block_group->alloc_mutex);
+ last_wanted = 0;
+ continue;
+ }
+
+ /* else we go to the next block group */
+ goto new_group;
+ }
+
+ ins->objectid = search_start;
+ ins->offset = num_bytes;
+
+ btrfs_remove_free_space_lock(block_group, search_start,
+ num_bytes);
+ /* we are all good, lets return */
+ mutex_unlock(&block_group->alloc_mutex);
+ break;
+ }
+new_group:
+ mutex_unlock(&block_group->alloc_mutex);
+ put_block_group(block_group);
+ block_group = NULL;
+new_group_no_lock:
+ /* don't try to compare new allocations against the
+ * last allocation any more
+ */
+ last_wanted = 0;
+
+ /*
+ * Here's how this works.
+ * loop == 0: we were searching a block group via a hint
+ * and didn't find anything, so we start at
+ * the head of the block groups and keep searching
+ * loop == 1: we're searching through all of the block groups
+ * if we hit the head again we have searched
+ * all of the block groups for this space and we
+ * need to try and allocate, if we cant error out.
+ * loop == 2: we allocated more space and are looping through
+ * all of the block groups again.
+ */
+ if (loop == 0) {
+ head = &space_info->block_groups;
+ cur = head->next;
+ loop++;
+ } else if (loop == 1 && cur == head) {
+ int keep_going;
+
+ /* at this point we give up on the empty_size
+ * allocations and just try to allocate the min
+ * space.
+ *
+ * The extra_loop field was set if an empty_size
+ * allocation was attempted above, and if this
+ * is try we need to try the loop again without
+ * the additional empty_size.
+ */
+ total_needed -= empty_size;
+ empty_size = 0;
+ keep_going = extra_loop;
+ loop++;
+
+ if (allowed_chunk_alloc && !chunk_alloc_done) {
+ up_read(&space_info->groups_sem);
+ ret = do_chunk_alloc(trans, root, num_bytes +
+ 2 * 1024 * 1024, data, 1);
+ down_read(&space_info->groups_sem);
+ if (ret < 0)
+ goto loop_check;
+ head = &space_info->block_groups;
+ /*
+ * we've allocated a new chunk, keep
+ * trying
+ */
+ keep_going = 1;
+ chunk_alloc_done = 1;
+ } else if (!allowed_chunk_alloc) {
+ space_info->force_alloc = 1;
+ }
+loop_check:
+ if (keep_going) {
+ cur = head->next;
+ extra_loop = 0;
+ } else {
+ break;
+ }
+ } else if (cur == head) {
+ break;
+ }
+
+ block_group = list_entry(cur, struct btrfs_block_group_cache,
+ list);
+ atomic_inc(&block_group->count);
+
+ search_start = block_group->key.objectid;
+ cur = cur->next;
+ }
+
+ /* we found what we needed */
+ if (ins->objectid) {
+ if (!(data & BTRFS_BLOCK_GROUP_DATA))
+ trans->block_group = block_group->key.objectid;
+
+ if (last_ptr)
+ *last_ptr = ins->objectid + ins->offset;
+ ret = 0;
+ } else if (!ret) {
+ printk(KERN_ERR "btrfs searching for %llu bytes, "
+ "num_bytes %llu, loop %d, allowed_alloc %d\n",
+ (unsigned long long)total_needed,
+ (unsigned long long)num_bytes,
+ loop, allowed_chunk_alloc);
+ ret = -ENOSPC;
+ }
+ if (block_group)
+ put_block_group(block_group);
+
+ up_read(&space_info->groups_sem);
+ return ret;
+}
+
+static void dump_space_info(struct btrfs_space_info *info, u64 bytes)
+{
+ struct btrfs_block_group_cache *cache;
+
+ printk(KERN_INFO "space_info has %llu free, is %sfull\n",
+ (unsigned long long)(info->total_bytes - info->bytes_used -
+ info->bytes_pinned - info->bytes_reserved),
+ (info->full) ? "" : "not ");
+ printk(KERN_INFO "space_info total=%llu, pinned=%llu, delalloc=%llu,"
+ " may_use=%llu, used=%llu\n", info->total_bytes,
+ info->bytes_pinned, info->bytes_delalloc, info->bytes_may_use,
+ info->bytes_used);
+
+ down_read(&info->groups_sem);
+ list_for_each_entry(cache, &info->block_groups, list) {
+ spin_lock(&cache->lock);
+ printk(KERN_INFO "block group %llu has %llu bytes, %llu used "
+ "%llu pinned %llu reserved\n",
+ (unsigned long long)cache->key.objectid,
+ (unsigned long long)cache->key.offset,
+ (unsigned long long)btrfs_block_group_used(&cache->item),
+ (unsigned long long)cache->pinned,
+ (unsigned long long)cache->reserved);
+ btrfs_dump_free_space(cache, bytes);
+ spin_unlock(&cache->lock);
+ }
+ up_read(&info->groups_sem);
+}
+
+static int __btrfs_reserve_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ u64 num_bytes, u64 min_alloc_size,
+ u64 empty_size, u64 hint_byte,
+ u64 search_end, struct btrfs_key *ins,
+ u64 data)
+{
+ int ret;
+ u64 search_start = 0;
+ struct btrfs_fs_info *info = root->fs_info;
+
+ data = btrfs_get_alloc_profile(root, data);
+again:
+ /*
+ * the only place that sets empty_size is btrfs_realloc_node, which
+ * is not called recursively on allocations
+ */
+ if (empty_size || root->ref_cows) {
+ if (!(data & BTRFS_BLOCK_GROUP_METADATA)) {
+ ret = do_chunk_alloc(trans, root->fs_info->extent_root,
+ 2 * 1024 * 1024,
+ BTRFS_BLOCK_GROUP_METADATA |
+ (info->metadata_alloc_profile &
+ info->avail_metadata_alloc_bits), 0);
+ }
+ ret = do_chunk_alloc(trans, root->fs_info->extent_root,
+ num_bytes + 2 * 1024 * 1024, data, 0);
+ }
+
+ WARN_ON(num_bytes < root->sectorsize);
+ ret = find_free_extent(trans, root, num_bytes, empty_size,
+ search_start, search_end, hint_byte, ins,
+ trans->alloc_exclude_start,
+ trans->alloc_exclude_nr, data);
+
+ if (ret == -ENOSPC && num_bytes > min_alloc_size) {
+ num_bytes = num_bytes >> 1;
+ num_bytes = num_bytes & ~(root->sectorsize - 1);
+ num_bytes = max(num_bytes, min_alloc_size);
+ do_chunk_alloc(trans, root->fs_info->extent_root,
+ num_bytes, data, 1);
+ goto again;
+ }
+ if (ret) {
+ struct btrfs_space_info *sinfo;
+
+ sinfo = __find_space_info(root->fs_info, data);
+ printk(KERN_ERR "btrfs allocation failed flags %llu, "
+ "wanted %llu\n", (unsigned long long)data,
+ (unsigned long long)num_bytes);
+ dump_space_info(sinfo, num_bytes);
+ BUG();
+ }
+
+ return ret;
+}
+
+int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
+{
+ struct btrfs_block_group_cache *cache;
+ int ret = 0;
+
+ cache = btrfs_lookup_block_group(root->fs_info, start);
+ if (!cache) {
+ printk(KERN_ERR "Unable to find block group for %llu\n",
+ (unsigned long long)start);
+ return -ENOSPC;
+ }
+
+ ret = btrfs_discard_extent(root, start, len);
+
+ btrfs_add_free_space(cache, start, len);
+ put_block_group(cache);
+ update_reserved_extents(root, start, len, 0);
+
+ return ret;
+}
+
+int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ u64 num_bytes, u64 min_alloc_size,
+ u64 empty_size, u64 hint_byte,
+ u64 search_end, struct btrfs_key *ins,
+ u64 data)
+{
+ int ret;
+ ret = __btrfs_reserve_extent(trans, root, num_bytes, min_alloc_size,
+ empty_size, hint_byte, search_end, ins,
+ data);
+ update_reserved_extents(root, ins->objectid, ins->offset, 1);
+ return ret;
+}
+
+static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, u64 parent,
+ u64 root_objectid, u64 ref_generation,
+ u64 owner, struct btrfs_key *ins)
+{
+ int ret;
+ int pending_ret;
+ u64 super_used;
+ u64 root_used;
+ u64 num_bytes = ins->offset;
+ u32 sizes[2];
+ struct btrfs_fs_info *info = root->fs_info;
+ struct btrfs_root *extent_root = info->extent_root;
+ struct btrfs_extent_item *extent_item;
+ struct btrfs_extent_ref *ref;
+ struct btrfs_path *path;
+ struct btrfs_key keys[2];
+
+ if (parent == 0)
+ parent = ins->objectid;
+
+ /* block accounting for super block */
+ spin_lock(&info->delalloc_lock);
+ super_used = btrfs_super_bytes_used(&info->super_copy);
+ btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes);
+
+ /* block accounting for root item */
+ root_used = btrfs_root_used(&root->root_item);
+ btrfs_set_root_used(&root->root_item, root_used + num_bytes);
+ spin_unlock(&info->delalloc_lock);
+
+ if (root == extent_root) {
+ struct pending_extent_op *extent_op;
+
+ extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
+ BUG_ON(!extent_op);
+
+ extent_op->type = PENDING_EXTENT_INSERT;
+ extent_op->bytenr = ins->objectid;
+ extent_op->num_bytes = ins->offset;
+ extent_op->parent = parent;
+ extent_op->orig_parent = 0;
+ extent_op->generation = ref_generation;
+ extent_op->orig_generation = 0;
+ extent_op->level = (int)owner;
+ INIT_LIST_HEAD(&extent_op->list);
+ extent_op->del = 0;
+
+ mutex_lock(&root->fs_info->extent_ins_mutex);
+ set_extent_bits(&root->fs_info->extent_ins, ins->objectid,
+ ins->objectid + ins->offset - 1,
+ EXTENT_WRITEBACK, GFP_NOFS);
+ set_state_private(&root->fs_info->extent_ins,
+ ins->objectid, (unsigned long)extent_op);
+ mutex_unlock(&root->fs_info->extent_ins_mutex);
+ goto update_block;
+ }
+
+ memcpy(&keys[0], ins, sizeof(*ins));
+ keys[1].objectid = ins->objectid;
+ keys[1].type = BTRFS_EXTENT_REF_KEY;
+ keys[1].offset = parent;
+ sizes[0] = sizeof(*extent_item);
+ sizes[1] = sizeof(*ref);
+
+ path = btrfs_alloc_path();
+ BUG_ON(!path);
+
+ ret = btrfs_insert_empty_items(trans, extent_root, path, keys,
+ sizes, 2);
+ BUG_ON(ret);
+
+ extent_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+ struct btrfs_extent_item);
+ btrfs_set_extent_refs(path->nodes[0], extent_item, 1);
+ ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
+ struct btrfs_extent_ref);
+
+ btrfs_set_ref_root(path->nodes[0], ref, root_objectid);
+ btrfs_set_ref_generation(path->nodes[0], ref, ref_generation);
+ btrfs_set_ref_objectid(path->nodes[0], ref, owner);
+ btrfs_set_ref_num_refs(path->nodes[0], ref, 1);
+
+ btrfs_mark_buffer_dirty(path->nodes[0]);
+
+ trans->alloc_exclude_start = 0;
+ trans->alloc_exclude_nr = 0;
+ btrfs_free_path(path);
+ finish_current_insert(trans, extent_root, 0);
+ pending_ret = del_pending_extents(trans, extent_root, 0);
+
+ if (ret)
+ goto out;
+ if (pending_ret) {
+ ret = pending_ret;
+ goto out;
+ }
+
+update_block:
+ ret = update_block_group(trans, root, ins->objectid,
+ ins->offset, 1, 0);
+ if (ret) {
+ printk(KERN_ERR "btrfs update block group failed for %llu "
+ "%llu\n", (unsigned long long)ins->objectid,
+ (unsigned long long)ins->offset);
+ BUG();
+ }
+out:
+ return ret;
+}
+
+int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, u64 parent,
+ u64 root_objectid, u64 ref_generation,
+ u64 owner, struct btrfs_key *ins)
+{
+ int ret;
+
+ if (root_objectid == BTRFS_TREE_LOG_OBJECTID)
+ return 0;
+ ret = __btrfs_alloc_reserved_extent(trans, root, parent, root_objectid,
+ ref_generation, owner, ins);
+ update_reserved_extents(root, ins->objectid, ins->offset, 0);
+ return ret;
+}
+
+/*
+ * this is used by the tree logging recovery code. It records that
+ * an extent has been allocated and makes sure to clear the free
+ * space cache bits as well
+ */
+int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, u64 parent,
+ u64 root_objectid, u64 ref_generation,
+ u64 owner, struct btrfs_key *ins)
+{
+ int ret;
+ struct btrfs_block_group_cache *block_group;
+
+ block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
+ mutex_lock(&block_group->cache_mutex);
+ cache_block_group(root, block_group);
+ mutex_unlock(&block_group->cache_mutex);
+
+ ret = btrfs_remove_free_space(block_group, ins->objectid,
+ ins->offset);
+ BUG_ON(ret);
+ put_block_group(block_group);
+ ret = __btrfs_alloc_reserved_extent(trans, root, parent, root_objectid,
+ ref_generation, owner, ins);
+ return ret;
+}
+
+/*
+ * finds a free extent and does all the dirty work required for allocation
+ * returns the key for the extent through ins, and a tree buffer for
+ * the first block of the extent through buf.
+ *
+ * returns 0 if everything worked, non-zero otherwise.
+ */
+int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ u64 num_bytes, u64 parent, u64 min_alloc_size,
+ u64 root_objectid, u64 ref_generation,
+ u64 owner_objectid, u64 empty_size, u64 hint_byte,
+ u64 search_end, struct btrfs_key *ins, u64 data)
+{
+ int ret;
+
+ ret = __btrfs_reserve_extent(trans, root, num_bytes,
+ min_alloc_size, empty_size, hint_byte,
+ search_end, ins, data);
+ BUG_ON(ret);
+ if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
+ ret = __btrfs_alloc_reserved_extent(trans, root, parent,
+ root_objectid, ref_generation,
+ owner_objectid, ins);
+ BUG_ON(ret);
+
+ } else {
+ update_reserved_extents(root, ins->objectid, ins->offset, 1);
+ }
+ return ret;
+}
+
+struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ u64 bytenr, u32 blocksize,
+ int level)
+{
+ struct extent_buffer *buf;
+
+ buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
+ if (!buf)
+ return ERR_PTR(-ENOMEM);
+ btrfs_set_header_generation(buf, trans->transid);
+ btrfs_set_buffer_lockdep_class(buf, level);
+ btrfs_tree_lock(buf);
+ clean_tree_block(trans, root, buf);
+
+ btrfs_set_lock_blocking(buf);
+ btrfs_set_buffer_uptodate(buf);
+
+ if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
+ set_extent_dirty(&root->dirty_log_pages, buf->start,
+ buf->start + buf->len - 1, GFP_NOFS);
+ } else {
+ set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
+ buf->start + buf->len - 1, GFP_NOFS);