/*
- * Copyright (C) 2007 Oracle. All rights reserved.
+ * Copyright (C) 2007,2008 Oracle. All rights reserved.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public
return path;
}
+/* this also releases the path */
void btrfs_free_path(struct btrfs_path *p)
{
btrfs_release_path(NULL, p);
kmem_cache_free(btrfs_path_cachep, p);
}
+/*
+ * path release drops references on the extent buffers in the path
+ * and it drops any locks held by this path
+ *
+ * It is safe to call this on paths that no locks or extent buffers held.
+ */
void noinline btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
{
int i;
}
}
+/*
+ * safely gets a reference on the root node of a tree. A lock
+ * is not taken, so a concurrent writer may put a different node
+ * at the root of the tree. See btrfs_lock_root_node for the
+ * looping required.
+ *
+ * The extent buffer returned by this has a reference taken, so
+ * it won't disappear. It may stop being the root of the tree
+ * at any time because there are no locks held.
+ */
struct extent_buffer *btrfs_root_node(struct btrfs_root *root)
{
struct extent_buffer *eb;
return eb;
}
+/* loop around taking references on and locking the root node of the
+ * tree until you end up with a lock on the root. A locked buffer
+ * is returned, with a reference held.
+ */
struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root)
{
struct extent_buffer *eb;
return eb;
}
+/* cowonly root (everything not a reference counted cow subvolume), just get
+ * put onto a simple dirty list. transaction.c walks this to make sure they
+ * get properly updated on disk.
+ */
static void add_root_to_dirty_list(struct btrfs_root *root)
{
if (root->track_dirty && list_empty(&root->dirty_list)) {
}
}
+/*
+ * used by snapshot creation to make a copy of a root for a tree with
+ * a given objectid. The buffer with the new root node is returned in
+ * cow_ret, and this func returns zero on success or a negative error code.
+ */
int btrfs_copy_root(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct extent_buffer *buf,
u32 nritems;
int ret = 0;
int level;
- struct btrfs_key first_key;
struct btrfs_root *new_root;
new_root = kmalloc(sizeof(*new_root), GFP_NOFS);
level = btrfs_header_level(buf);
nritems = btrfs_header_nritems(buf);
- if (nritems) {
- if (level == 0)
- btrfs_item_key_to_cpu(buf, &first_key, 0);
- else
- btrfs_node_key_to_cpu(buf, &first_key, 0);
- } else {
- first_key.objectid = 0;
- }
- cow = btrfs_alloc_free_block(trans, new_root, buf->len,
- new_root_objectid,
- trans->transid, first_key.objectid,
- level, buf->start, 0);
+
+ cow = btrfs_alloc_free_block(trans, new_root, buf->len, 0,
+ new_root_objectid, trans->transid,
+ level, buf->start, 0);
if (IS_ERR(cow)) {
kfree(new_root);
return PTR_ERR(cow);
btrfs_set_header_owner(cow, new_root_objectid);
btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN);
+ write_extent_buffer(cow, root->fs_info->fsid,
+ (unsigned long)btrfs_header_fsid(cow),
+ BTRFS_FSID_SIZE);
+
WARN_ON(btrfs_header_generation(buf) > trans->transid);
- ret = btrfs_inc_ref(trans, new_root, buf, 0);
+ ret = btrfs_inc_ref(trans, new_root, buf, cow, NULL);
kfree(new_root);
if (ret)
return 0;
}
+/*
+ * does the dirty work in cow of a single block. The parent block
+ * (if supplied) is updated to point to the new cow copy. The new
+ * buffer is marked dirty and returned locked. If you modify the block
+ * it needs to be marked dirty again.
+ *
+ * search_start -- an allocation hint for the new block
+ *
+ * empty_size -- a hint that you plan on doing more cow. This is the size in bytes
+ * the allocator should try to find free next to the block it returns. This is
+ * just a hint and may be ignored by the allocator.
+ *
+ * prealloc_dest -- if you have already reserved a destination for the cow,
+ * this uses that block instead of allocating a new one. btrfs_alloc_reserved_extent
+ * is used to finish the allocation.
+ */
int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct extent_buffer *buf,
u64 search_start, u64 empty_size,
u64 prealloc_dest)
{
- u64 root_gen;
+ u64 parent_start;
struct extent_buffer *cow;
u32 nritems;
int ret = 0;
- int different_trans = 0;
int level;
int unlock_orig = 0;
- struct btrfs_key first_key;
if (*cow_ret == buf)
unlock_orig = 1;
WARN_ON(!btrfs_tree_locked(buf));
- if (root->ref_cows) {
- root_gen = trans->transid;
- } else {
- root_gen = 0;
- }
+ if (parent)
+ parent_start = parent->start;
+ else
+ parent_start = 0;
+
WARN_ON(root->ref_cows && trans->transid !=
root->fs_info->running_transaction->transid);
WARN_ON(root->ref_cows && trans->transid != root->last_trans);
level = btrfs_header_level(buf);
nritems = btrfs_header_nritems(buf);
- if (nritems) {
- if (level == 0)
- btrfs_item_key_to_cpu(buf, &first_key, 0);
- else
- btrfs_node_key_to_cpu(buf, &first_key, 0);
- } else {
- first_key.objectid = 0;
- }
+
if (prealloc_dest) {
struct btrfs_key ins;
ins.offset = buf->len;
ins.type = BTRFS_EXTENT_ITEM_KEY;
- ret = btrfs_alloc_reserved_extent(trans, root,
+ ret = btrfs_alloc_reserved_extent(trans, root, parent_start,
root->root_key.objectid,
- root_gen, level,
- first_key.objectid,
- &ins);
+ trans->transid, level, &ins);
BUG_ON(ret);
cow = btrfs_init_new_buffer(trans, root, prealloc_dest,
buf->len);
} else {
cow = btrfs_alloc_free_block(trans, root, buf->len,
+ parent_start,
root->root_key.objectid,
- root_gen, first_key.objectid,
- level, search_start, empty_size);
+ trans->transid, level,
+ search_start, empty_size);
}
if (IS_ERR(cow))
return PTR_ERR(cow);
btrfs_set_header_owner(cow, root->root_key.objectid);
btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN);
+ write_extent_buffer(cow, root->fs_info->fsid,
+ (unsigned long)btrfs_header_fsid(cow),
+ BTRFS_FSID_SIZE);
+
WARN_ON(btrfs_header_generation(buf) > trans->transid);
if (btrfs_header_generation(buf) != trans->transid) {
- different_trans = 1;
- ret = btrfs_inc_ref(trans, root, buf, 1);
+ u32 nr_extents;
+ ret = btrfs_inc_ref(trans, root, buf, cow, &nr_extents);
if (ret)
return ret;
+
+ ret = btrfs_cache_ref(trans, root, buf, nr_extents);
+ WARN_ON(ret);
+ } else if (btrfs_header_owner(buf) == BTRFS_TREE_RELOC_OBJECTID) {
+ /*
+ * There are only two places that can drop reference to
+ * tree blocks owned by living reloc trees, one is here,
+ * the other place is btrfs_drop_subtree. In both places,
+ * we check reference count while tree block is locked.
+ * Furthermore, if reference count is one, it won't get
+ * increased by someone else.
+ */
+ u32 refs;
+ ret = btrfs_lookup_extent_ref(trans, root, buf->start,
+ buf->len, &refs);
+ BUG_ON(ret);
+ if (refs == 1) {
+ ret = btrfs_update_ref(trans, root, buf, cow,
+ 0, nritems);
+ clean_tree_block(trans, root, buf);
+ } else {
+ ret = btrfs_inc_ref(trans, root, buf, cow, NULL);
+ }
+ BUG_ON(ret);
} else {
+ ret = btrfs_update_ref(trans, root, buf, cow, 0, nritems);
+ if (ret)
+ return ret;
clean_tree_block(trans, root, buf);
}
+ if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) {
+ ret = btrfs_reloc_tree_cache_ref(trans, root, cow, buf->start);
+ WARN_ON(ret);
+ }
+
if (buf == root->node) {
WARN_ON(parent && parent != buf);
- root_gen = btrfs_header_generation(buf);
spin_lock(&root->node_lock);
root->node = cow;
if (buf != root->commit_root) {
btrfs_free_extent(trans, root, buf->start,
- buf->len, root->root_key.objectid,
- root_gen, 0, 0, 1);
+ buf->len, buf->start,
+ root->root_key.objectid,
+ btrfs_header_generation(buf),
+ level, 1);
}
free_extent_buffer(buf);
add_root_to_dirty_list(root);
} else {
- root_gen = btrfs_header_generation(parent);
btrfs_set_node_blockptr(parent, parent_slot,
cow->start);
WARN_ON(trans->transid == 0);
btrfs_mark_buffer_dirty(parent);
WARN_ON(btrfs_header_generation(parent) != trans->transid);
btrfs_free_extent(trans, root, buf->start, buf->len,
- btrfs_header_owner(parent), root_gen,
- 0, 0, 1);
+ parent_start, btrfs_header_owner(parent),
+ btrfs_header_generation(parent), level, 1);
}
if (unlock_orig)
btrfs_tree_unlock(buf);
return 0;
}
+/*
+ * cows a single block, see __btrfs_cow_block for the real work.
+ * This version of it has extra checks so that a block isn't cow'd more than
+ * once per transaction, as long as it hasn't been written yet
+ */
int noinline btrfs_cow_block(struct btrfs_trans_handle *trans,
struct btrfs_root *root, struct extent_buffer *buf,
struct extent_buffer *parent, int parent_slot,
struct extent_buffer **cow_ret, u64 prealloc_dest)
{
u64 search_start;
- u64 header_trans;
int ret;
if (trans->transaction != root->fs_info->running_transaction) {
WARN_ON(1);
}
- header_trans = btrfs_header_generation(buf);
spin_lock(&root->fs_info->hash_lock);
- if (header_trans == trans->transid &&
+ if (btrfs_header_generation(buf) == trans->transid &&
+ btrfs_header_owner(buf) == root->root_key.objectid &&
!btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
*cow_ret = buf;
spin_unlock(&root->fs_info->hash_lock);
return ret;
}
+/*
+ * helper function for defrag to decide if two blocks pointed to by a
+ * node are actually close by
+ */
static int close_blocks(u64 blocknr, u64 other, u32 blocksize)
{
if (blocknr < other && other - (blocknr + blocksize) < 32768)
return 0;
}
+/*
+ * same as comp_keys only with two btrfs_key's
+ */
+static int comp_cpu_keys(struct btrfs_key *k1, struct btrfs_key *k2)
+{
+ if (k1->objectid > k2->objectid)
+ return 1;
+ if (k1->objectid < k2->objectid)
+ return -1;
+ if (k1->type > k2->type)
+ return 1;
+ if (k1->type < k2->type)
+ return -1;
+ if (k1->offset > k2->offset)
+ return 1;
+ if (k1->offset < k2->offset)
+ return -1;
+ return 0;
+}
+/*
+ * this is used by the defrag code to go through all the
+ * leaves pointed to by a node and reallocate them so that
+ * disk order is close to key order
+ */
int btrfs_realloc_node(struct btrfs_trans_handle *trans,
struct btrfs_root *root, struct extent_buffer *parent,
int start_slot, int cache_only, u64 *last_ret,
return btrfs_item_offset_nr(leaf, nr - 1);
}
+/*
+ * extra debugging checks to make sure all the items in a key are
+ * well formed and in the proper order
+ */
static int check_node(struct btrfs_root *root, struct btrfs_path *path,
int level)
{
return 0;
}
+/*
+ * extra checking to make sure all the items in a leaf are
+ * well formed and in the proper order
+ */
static int check_leaf(struct btrfs_root *root, struct btrfs_path *path,
int level)
{
return -1;
}
+/* given a node and slot number, this reads the blocks it points to. The
+ * extent buffer is returned with a reference taken (but unlocked).
+ * NULL is returned on error.
+ */
static noinline struct extent_buffer *read_node_slot(struct btrfs_root *root,
struct extent_buffer *parent, int slot)
{
btrfs_node_ptr_generation(parent, slot));
}
+/*
+ * node level balancing, used to make sure nodes are in proper order for
+ * item deletion. We balance from the top down, so we have to make sure
+ * that a deletion won't leave an node completely empty later on.
+ */
static noinline int balance_level(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path, int level)
root->node = child;
spin_unlock(&root->node_lock);
+ ret = btrfs_update_extent_ref(trans, root, child->start,
+ mid->start, child->start,
+ root->root_key.objectid,
+ trans->transid, level - 1);
+ BUG_ON(ret);
+
add_root_to_dirty_list(root);
btrfs_tree_unlock(child);
path->locks[level] = 0;
/* once for the path */
free_extent_buffer(mid);
ret = btrfs_free_extent(trans, root, mid->start, mid->len,
- root->root_key.objectid,
- btrfs_header_generation(mid), 0, 0, 1);
+ mid->start, root->root_key.objectid,
+ btrfs_header_generation(mid),
+ level, 1);
/* once for the root ptr */
free_extent_buffer(mid);
return ret;
if (wret)
ret = wret;
wret = btrfs_free_extent(trans, root, bytenr,
- blocksize,
+ blocksize, parent->start,
btrfs_header_owner(parent),
- generation, 0, 0, 1);
+ generation, level, 1);
if (wret)
ret = wret;
} else {
if (wret)
ret = wret;
wret = btrfs_free_extent(trans, root, bytenr, blocksize,
+ parent->start,
btrfs_header_owner(parent),
- root_gen, 0, 0, 1);
+ root_gen, level, 1);
if (wret)
ret = wret;
} else {
return ret;
}
-/* returns zero if the push worked, non-zero otherwise */
+/* Node balancing for insertion. Here we only split or push nodes around
+ * when they are completely full. This is also done top down, so we
+ * have to be pessimistic.
+ */
static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path, int level)
}
/*
- * readahead one full node of leaves
+ * readahead one full node of leaves, finding things that are close
+ * to the block in 'slot', and triggering ra on them.
*/
static noinline void reada_for_search(struct btrfs_root *root,
struct btrfs_path *path,
}
search = btrfs_node_blockptr(node, nr);
if ((search >= lowest_read && search <= highest_read) ||
- (search < lowest_read && lowest_read - search <= 32768) ||
- (search > highest_read && search - highest_read <= 32768)) {
+ (search < lowest_read && lowest_read - search <= 16384) ||
+ (search > highest_read && search - highest_read <= 16384)) {
readahead_tree_block(root, search, blocksize,
btrfs_node_ptr_generation(node, nr));
nread += blocksize;
}
nscan++;
- if (path->reada < 2 && (nread > (256 * 1024) || nscan > 32))
+ if (path->reada < 2 && (nread > (64 * 1024) || nscan > 32))
break;
- if(nread > (1024 * 1024) || nscan > 128)
+ if(nread > (256 * 1024) || nscan > 128)
break;
if (search < lowest_read)
}
}
+/*
+ * when we walk down the tree, it is usually safe to unlock the higher layers in
+ * the tree. The exceptions are when our path goes through slot 0, because operations
+ * on the tree might require changing key pointers higher up in the tree.
+ *
+ * callers might also have set path->keep_locks, which tells this code to
+ * keep the lock if the path points to the last slot in the block. This is
+ * part of walking through the tree, and selecting the next slot in the higher
+ * block.
+ *
+ * lowest_unlock sets the lowest level in the tree we're allowed to unlock.
+ * so if lowest_unlock is 1, level 0 won't be unlocked
+ */
static noinline void unlock_up(struct btrfs_path *path, int level,
int lowest_unlock)
{
struct btrfs_key prealloc_block;
lowest_level = p->lowest_level;
- WARN_ON(lowest_level && ins_len);
+ WARN_ON(lowest_level && ins_len > 0);
WARN_ON(p->nodes[0] != NULL);
- WARN_ON(cow && root == root->fs_info->extent_root &&
- !mutex_is_locked(&root->fs_info->alloc_mutex));
+
if (ins_len < 0)
lowest_unlock = 2;
/* is a cow on this block not required */
spin_lock(&root->fs_info->hash_lock);
if (btrfs_header_generation(b) == trans->transid &&
+ btrfs_header_owner(b) == root->root_key.objectid &&
!btrfs_header_flag(b, BTRFS_HEADER_FLAG_WRITTEN)) {
spin_unlock(&root->fs_info->hash_lock);
goto cow_done;
/* this is only true while dropping a snapshot */
if (level == lowest_level) {
- break;
+ ret = 0;
+ goto done;
}
blocknr = btrfs_node_blockptr(b, slot);
return ret;
}
+int btrfs_merge_path(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_key *node_keys,
+ u64 *nodes, int lowest_level)
+{
+ struct extent_buffer *eb;
+ struct extent_buffer *parent;
+ struct btrfs_key key;
+ u64 bytenr;
+ u64 generation;
+ u32 blocksize;
+ int level;
+ int slot;
+ int key_match;
+ int ret;
+
+ eb = btrfs_lock_root_node(root);
+ ret = btrfs_cow_block(trans, root, eb, NULL, 0, &eb, 0);
+ BUG_ON(ret);
+
+ parent = eb;
+ while (1) {
+ level = btrfs_header_level(parent);
+ if (level == 0 || level <= lowest_level)
+ break;
+
+ ret = bin_search(parent, &node_keys[lowest_level], level,
+ &slot);
+ if (ret && slot > 0)
+ slot--;
+
+ bytenr = btrfs_node_blockptr(parent, slot);
+ if (nodes[level - 1] == bytenr)
+ break;
+
+ blocksize = btrfs_level_size(root, level - 1);
+ generation = btrfs_node_ptr_generation(parent, slot);
+ btrfs_node_key_to_cpu(eb, &key, slot);
+ key_match = !memcmp(&key, &node_keys[level - 1], sizeof(key));
+
+ if (generation == trans->transid) {
+ eb = read_tree_block(root, bytenr, blocksize,
+ generation);
+ btrfs_tree_lock(eb);
+ }
+
+ /*
+ * if node keys match and node pointer hasn't been modified
+ * in the running transaction, we can merge the path. for
+ * blocks owened by reloc trees, the node pointer check is
+ * skipped, this is because these blocks are fully controlled
+ * by the space balance code, no one else can modify them.
+ */
+ if (!nodes[level - 1] || !key_match ||
+ (generation == trans->transid &&
+ btrfs_header_owner(eb) != BTRFS_TREE_RELOC_OBJECTID)) {
+ if (level == 1 || level == lowest_level + 1) {
+ if (generation == trans->transid) {
+ btrfs_tree_unlock(eb);
+ free_extent_buffer(eb);
+ }
+ break;
+ }
+
+ if (generation != trans->transid) {
+ eb = read_tree_block(root, bytenr, blocksize,
+ generation);
+ btrfs_tree_lock(eb);
+ }
+
+ ret = btrfs_cow_block(trans, root, eb, parent, slot,
+ &eb, 0);
+ BUG_ON(ret);
+
+ if (root->root_key.objectid ==
+ BTRFS_TREE_RELOC_OBJECTID) {
+ if (!nodes[level - 1]) {
+ nodes[level - 1] = eb->start;
+ memcpy(&node_keys[level - 1], &key,
+ sizeof(node_keys[0]));
+ } else {
+ WARN_ON(1);
+ }
+ }
+
+ btrfs_tree_unlock(parent);
+ free_extent_buffer(parent);
+ parent = eb;
+ continue;
+ }
+
+ btrfs_set_node_blockptr(parent, slot, nodes[level - 1]);
+ btrfs_set_node_ptr_generation(parent, slot, trans->transid);
+ btrfs_mark_buffer_dirty(parent);
+
+ ret = btrfs_inc_extent_ref(trans, root,
+ nodes[level - 1],
+ blocksize, parent->start,
+ btrfs_header_owner(parent),
+ btrfs_header_generation(parent),
+ level - 1);
+ BUG_ON(ret);
+
+ /*
+ * If the block was created in the running transaction,
+ * it's possible this is the last reference to it, so we
+ * should drop the subtree.
+ */
+ if (generation == trans->transid) {
+ ret = btrfs_drop_subtree(trans, root, eb, parent);
+ BUG_ON(ret);
+ btrfs_tree_unlock(eb);
+ free_extent_buffer(eb);
+ } else {
+ ret = btrfs_free_extent(trans, root, bytenr,
+ blocksize, parent->start,
+ btrfs_header_owner(parent),
+ btrfs_header_generation(parent),
+ level - 1, 1);
+ BUG_ON(ret);
+ }
+ break;
+ }
+ btrfs_tree_unlock(parent);
+ free_extent_buffer(parent);
+ return 0;
+}
+
/*
* adjust the pointers going up the tree, starting at level
* making sure the right key of each node is points to 'key'.
return ret;
}
+/*
+ * update item key.
+ *
+ * This function isn't completely safe. It's the caller's responsibility
+ * that the new key won't break the order
+ */
+int btrfs_set_item_key_safe(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, struct btrfs_path *path,
+ struct btrfs_key *new_key)
+{
+ struct btrfs_disk_key disk_key;
+ struct extent_buffer *eb;
+ int slot;
+
+ eb = path->nodes[0];
+ slot = path->slots[0];
+ if (slot > 0) {
+ btrfs_item_key(eb, &disk_key, slot - 1);
+ if (comp_keys(&disk_key, new_key) >= 0)
+ return -1;
+ }
+ if (slot < btrfs_header_nritems(eb) - 1) {
+ btrfs_item_key(eb, &disk_key, slot + 1);
+ if (comp_keys(&disk_key, new_key) <= 0)
+ return -1;
+ }
+
+ btrfs_cpu_key_to_disk(&disk_key, new_key);
+ btrfs_set_item_key(eb, &disk_key, slot);
+ btrfs_mark_buffer_dirty(eb);
+ if (slot == 0)
+ fixup_low_keys(trans, root, path, &disk_key, 1);
+ return 0;
+}
+
/*
* try to push data from one node into the next node left in the
* tree.
btrfs_set_header_nritems(dst, dst_nritems + push_items);
btrfs_mark_buffer_dirty(src);
btrfs_mark_buffer_dirty(dst);
+
+ ret = btrfs_update_ref(trans, root, src, dst, dst_nritems, push_items);
+ BUG_ON(ret);
+
return ret;
}
btrfs_mark_buffer_dirty(src);
btrfs_mark_buffer_dirty(dst);
+
+ ret = btrfs_update_ref(trans, root, src, dst, 0, push_items);
+ BUG_ON(ret);
+
return ret;
}
struct btrfs_root *root,
struct btrfs_path *path, int level)
{
- u64 root_gen;
u64 lower_gen;
struct extent_buffer *lower;
struct extent_buffer *c;
struct extent_buffer *old;
struct btrfs_disk_key lower_key;
+ int ret;
BUG_ON(path->nodes[level]);
BUG_ON(path->nodes[level-1] != root->node);
- if (root->ref_cows)
- root_gen = trans->transid;
- else
- root_gen = 0;
-
lower = path->nodes[level-1];
if (level == 1)
btrfs_item_key(lower, &lower_key, 0);
else
btrfs_node_key(lower, &lower_key, 0);
- c = btrfs_alloc_free_block(trans, root, root->nodesize,
- root->root_key.objectid,
- root_gen, le64_to_cpu(lower_key.objectid),
+ c = btrfs_alloc_free_block(trans, root, root->nodesize, 0,
+ root->root_key.objectid, trans->transid,
level, root->node->start, 0);
if (IS_ERR(c))
return PTR_ERR(c);
btrfs_set_node_key(c, &lower_key, 0);
btrfs_set_node_blockptr(c, 0, lower->start);
lower_gen = btrfs_header_generation(lower);
- WARN_ON(lower_gen == 0);
+ WARN_ON(lower_gen != trans->transid);
btrfs_set_node_ptr_generation(c, 0, lower_gen);
root->node = c;
spin_unlock(&root->node_lock);
+ ret = btrfs_update_extent_ref(trans, root, lower->start,
+ lower->start, c->start,
+ root->root_key.objectid,
+ trans->transid, level - 1);
+ BUG_ON(ret);
+
/* the super has an extra ref to root->node */
free_extent_buffer(old);
path->nodes[level] = c;
path->locks[level] = 1;
path->slots[level] = 0;
-
- if (root->ref_cows && lower_gen != trans->transid) {
- struct btrfs_path *back_path = btrfs_alloc_path();
- int ret;
- mutex_lock(&root->fs_info->alloc_mutex);
- ret = btrfs_insert_extent_backref(trans,
- root->fs_info->extent_root,
- path, lower->start,
- root->root_key.objectid,
- trans->transid, 0, 0);
- BUG_ON(ret);
- mutex_unlock(&root->fs_info->alloc_mutex);
- btrfs_free_path(back_path);
- }
return 0;
}
struct btrfs_root *root,
struct btrfs_path *path, int level)
{
- u64 root_gen;
struct extent_buffer *c;
struct extent_buffer *split;
struct btrfs_disk_key disk_key;
}
c_nritems = btrfs_header_nritems(c);
- if (root->ref_cows)
- root_gen = trans->transid;
- else
- root_gen = 0;
- btrfs_node_key(c, &disk_key, 0);
split = btrfs_alloc_free_block(trans, root, root->nodesize,
- root->root_key.objectid,
- root_gen,
- btrfs_disk_key_objectid(&disk_key),
- level, c->start, 0);
+ path->nodes[level + 1]->start,
+ root->root_key.objectid,
+ trans->transid, level, c->start, 0);
if (IS_ERR(split))
return PTR_ERR(split);
if (wret)
ret = wret;
+ ret = btrfs_update_ref(trans, root, c, split, 0, c_nritems - mid);
+ BUG_ON(ret);
+
if (path->slots[level] >= mid) {
path->slots[level] -= mid;
btrfs_tree_unlock(c);
else
nr = 1;
+ if (path->slots[0] >= left_nritems)
+ push_space += data_size + sizeof(*item);
+
i = left_nritems - 1;
while (i >= nr) {
item = btrfs_item_nr(left, i);
+ if (!empty && push_items > 0) {
+ if (path->slots[0] > i)
+ break;
+ if (path->slots[0] == i) {
+ int space = btrfs_leaf_free_space(root, left);
+ if (space + push_space * 2 > free_space)
+ break;
+ }
+ }
+
if (path->slots[0] == i)
push_space += data_size + sizeof(*item);
this_item_size = btrfs_item_size(left, item);
if (this_item_size + sizeof(*item) + push_space > free_space)
break;
+
push_items++;
push_space += this_item_size + sizeof(*item);
if (i == 0)
btrfs_mark_buffer_dirty(left);
btrfs_mark_buffer_dirty(right);
+ ret = btrfs_update_ref(trans, root, left, right, 0, push_items);
+ BUG_ON(ret);
+
btrfs_item_key(right, &disk_key, 0);
btrfs_set_node_key(upper, &disk_key, slot + 1);
btrfs_mark_buffer_dirty(upper);
KM_USER1);
}
+ if (!empty && push_items > 0) {
+ if (path->slots[0] < i)
+ break;
+ if (path->slots[0] == i) {
+ int space = btrfs_leaf_free_space(root, right);
+ if (space + push_space * 2 > free_space)
+ break;
+ }
+ }
+
if (path->slots[0] == i)
push_space += data_size + sizeof(*item);
if (right_nritems)
btrfs_mark_buffer_dirty(right);
+ ret = btrfs_update_ref(trans, root, right, left,
+ old_left_nritems, push_items);
+ BUG_ON(ret);
+
btrfs_item_key(right, &disk_key, 0);
wret = fixup_low_keys(trans, root, path, &disk_key, 1);
if (wret)
struct btrfs_path *path, int data_size,
int extend)
{
- u64 root_gen;
struct extent_buffer *l;
u32 nritems;
int mid;
if (extend)
space_needed = data_size;
- if (root->ref_cows)
- root_gen = trans->transid;
- else
- root_gen = 0;
-
/* first try to make some room by pushing left and right */
if (ins_key->type != BTRFS_DIR_ITEM_KEY) {
wret = push_leaf_right(trans, root, path, data_size, 0);
nritems = btrfs_header_nritems(l);
mid = (nritems + 1)/ 2;
- btrfs_item_key(l, &disk_key, 0);
-
right = btrfs_alloc_free_block(trans, root, root->leafsize,
- root->root_key.objectid,
- root_gen,
- le64_to_cpu(disk_key.objectid),
- 0, l->start, 0);
+ path->nodes[1]->start,
+ root->root_key.objectid,
+ trans->transid, 0, l->start, 0);
if (IS_ERR(right)) {
BUG_ON(1);
return PTR_ERR(right);
btrfs_mark_buffer_dirty(l);
BUG_ON(path->slots[0] != slot);
+ ret = btrfs_update_ref(trans, root, l, right, 0, nritems);
+ BUG_ON(ret);
+
if (mid <= slot) {
btrfs_tree_unlock(path->nodes[0]);
free_extent_buffer(path->nodes[0]);
return ret;
}
+/*
+ * make the item pointed to by the path smaller. new_size indicates
+ * how small to make it, and from_end tells us if we just chop bytes
+ * off the end of the item or if we shift the item to chop bytes off
+ * the front.
+ */
int btrfs_truncate_item(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
return ret;
}
+/*
+ * make the item pointed to by the path bigger, data_size is the new size.
+ */
int btrfs_extend_item(struct btrfs_trans_handle *trans,
struct btrfs_root *root, struct btrfs_path *path,
u32 data_size)
}
/*
- * Given a key and some data, insert an item into the tree.
+ * Given a key and some data, insert items into the tree.
+ * This does all the path init required, making room in the tree if needed.
+ * Returns the number of keys that were inserted.
+ */
+int btrfs_insert_some_items(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_path *path,
+ struct btrfs_key *cpu_key, u32 *data_size,
+ int nr)
+{
+ struct extent_buffer *leaf;
+ struct btrfs_item *item;
+ int ret = 0;
+ int slot;
+ int slot_orig;
+ int i;
+ u32 nritems;
+ u32 total_data = 0;
+ u32 total_size = 0;
+ unsigned int data_end;
+ struct btrfs_disk_key disk_key;
+ struct btrfs_key found_key;
+
+ found_key.objectid = 0;
+ nr = min_t(int, nr, BTRFS_NODEPTRS_PER_BLOCK(root));
+
+ for (i = 0; i < nr; i++)
+ total_data += data_size[i];
+
+ total_data = min_t(u32, total_data, BTRFS_LEAF_DATA_SIZE(root));
+ total_size = total_data + (nr * sizeof(struct btrfs_item));
+ ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1);
+ if (ret == 0)
+ return -EEXIST;
+ if (ret < 0)
+ goto out;
+
+ slot_orig = path->slots[0];
+ leaf = path->nodes[0];
+
+ nritems = btrfs_header_nritems(leaf);
+ data_end = leaf_data_end(root, leaf);
+
+ if (btrfs_leaf_free_space(root, leaf) < total_size) {
+ for (i = nr; i >= 0; i--) {
+ total_data -= data_size[i];
+ total_size -= data_size[i] + sizeof(struct btrfs_item);
+ if (total_size < btrfs_leaf_free_space(root, leaf))
+ break;
+ }
+ nr = i;
+ }
+
+ slot = path->slots[0];
+ BUG_ON(slot < 0);
+
+ if (slot != nritems) {
+ unsigned int old_data = btrfs_item_end_nr(leaf, slot);
+
+ item = btrfs_item_nr(leaf, slot);
+ btrfs_item_key_to_cpu(leaf, &found_key, slot);
+
+ /* figure out how many keys we can insert in here */
+ total_data = data_size[0];
+ for (i = 1; i < nr; i++) {
+ if (comp_cpu_keys(&found_key, cpu_key + i) <= 0)
+ break;
+ total_data += data_size[i];
+ }
+ nr = i;
+
+ if (old_data < data_end) {
+ btrfs_print_leaf(root, leaf);
+ printk("slot %d old_data %d data_end %d\n",
+ slot, old_data, data_end);
+ BUG_ON(1);
+ }
+ /*
+ * item0..itemN ... dataN.offset..dataN.size .. data0.size
+ */
+ /* first correct the data pointers */
+ WARN_ON(leaf->map_token);
+ for (i = slot; i < nritems; i++) {
+ u32 ioff;
+
+ item = btrfs_item_nr(leaf, i);
+ if (!leaf->map_token) {
+ map_extent_buffer(leaf, (unsigned long)item,
+ sizeof(struct btrfs_item),
+ &leaf->map_token, &leaf->kaddr,
+ &leaf->map_start, &leaf->map_len,
+ KM_USER1);
+ }
+
+ ioff = btrfs_item_offset(leaf, item);
+ btrfs_set_item_offset(leaf, item, ioff - total_data);
+ }
+ if (leaf->map_token) {
+ unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
+ leaf->map_token = NULL;
+ }
+
+ /* shift the items */
+ memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + nr),
+ btrfs_item_nr_offset(slot),
+ (nritems - slot) * sizeof(struct btrfs_item));
+
+ /* shift the data */
+ memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
+ data_end - total_data, btrfs_leaf_data(leaf) +
+ data_end, old_data - data_end);
+ data_end = old_data;
+ } else {
+ /*
+ * this sucks but it has to be done, if we are inserting at
+ * the end of the leaf only insert 1 of the items, since we
+ * have no way of knowing whats on the next leaf and we'd have
+ * to drop our current locks to figure it out
+ */
+ nr = 1;
+ }
+
+ /* setup the item for the new data */
+ for (i = 0; i < nr; i++) {
+ btrfs_cpu_key_to_disk(&disk_key, cpu_key + i);
+ btrfs_set_item_key(leaf, &disk_key, slot + i);
+ item = btrfs_item_nr(leaf, slot + i);
+ btrfs_set_item_offset(leaf, item, data_end - data_size[i]);
+ data_end -= data_size[i];
+ btrfs_set_item_size(leaf, item, data_size[i]);
+ }
+ btrfs_set_header_nritems(leaf, nritems + nr);
+ btrfs_mark_buffer_dirty(leaf);
+
+ ret = 0;
+ if (slot == 0) {
+ btrfs_cpu_key_to_disk(&disk_key, cpu_key);
+ ret = fixup_low_keys(trans, root, path, &disk_key, 1);
+ }
+
+ if (btrfs_leaf_free_space(root, leaf) < 0) {
+ btrfs_print_leaf(root, leaf);
+ BUG();
+ }
+out:
+ if (!ret)
+ ret = nr;
+ return ret;
+}
+
+/*
+ * Given a key and some data, insert items into the tree.
* This does all the path init required, making room in the tree if needed.
*/
int btrfs_insert_empty_items(struct btrfs_trans_handle *trans,
total_size = total_data + (nr * sizeof(struct btrfs_item));
ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1);
- if (ret == 0) {
+ if (ret == 0)
return -EEXIST;
- }
if (ret < 0)
goto out;
nritems = btrfs_header_nritems(leaf);
data_end = leaf_data_end(root, leaf);
- if (btrfs_leaf_free_space(root, leaf) <
- sizeof(struct btrfs_item) + total_size) {
+ if (btrfs_leaf_free_space(root, leaf) < total_size) {
btrfs_print_leaf(root, leaf);
printk("not enough freespace need %u have %d\n",
total_size, btrfs_leaf_free_space(root, leaf));
BUG_ON(slot < 0);
if (slot != nritems) {
- int i;
unsigned int old_data = btrfs_item_end_nr(leaf, slot);
if (old_data < data_end) {
/*
* delete the pointer from a given node.
*
- * If the delete empties a node, the node is removed from the tree,
- * continuing all the way the root if required. The root is converted into
- * a leaf if all the nodes are emptied.
+ * the tree should have been previously balanced so the deletion does not
+ * empty a node.
*/
static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
struct btrfs_path *path, int level, int slot)
return ret;
}
+/*
+ * a helper function to delete the leaf pointed to by path->slots[1] and
+ * path->nodes[1]. bytenr is the node block pointer, but since the callers
+ * already know it, it is faster to have them pass it down than to
+ * read it out of the node again.
+ *
+ * This deletes the pointer in path->nodes[1] and frees the leaf
+ * block extent. zero is returned if it all worked out, < 0 otherwise.
+ *
+ * The path must have already been setup for deleting the leaf, including
+ * all the proper balancing. path->nodes[1] must be locked.
+ */
+noinline int btrfs_del_leaf(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_path *path, u64 bytenr)
+{
+ int ret;
+ u64 root_gen = btrfs_header_generation(path->nodes[1]);
+
+ ret = del_ptr(trans, root, path, 1, path->slots[1]);
+ if (ret)
+ return ret;
+
+ ret = btrfs_free_extent(trans, root, bytenr,
+ btrfs_level_size(root, 0),
+ path->nodes[1]->start,
+ btrfs_header_owner(path->nodes[1]),
+ root_gen, 0, 1);
+ return ret;
+}
/*
* delete the item at the leaf level in path. If that empties
* the leaf, remove it from the tree
nritems = btrfs_header_nritems(leaf);
if (slot + nr != nritems) {
- int i;
int data_end = leaf_data_end(root, leaf);
memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
if (leaf == root->node) {
btrfs_set_header_level(leaf, 0);
} else {
- u64 root_gen = btrfs_header_generation(path->nodes[1]);
- wret = del_ptr(trans, root, path, 1, path->slots[1]);
- if (wret)
- ret = wret;
- wret = btrfs_free_extent(trans, root,
- leaf->start, leaf->len,
- btrfs_header_owner(path->nodes[1]),
- root_gen, 0, 0, 1);
- if (wret)
- ret = wret;
+ ret = btrfs_del_leaf(trans, root, path, leaf->start);
+ BUG_ON(ret);
}
} else {
int used = leaf_space_used(leaf, 0, nritems);
}
if (btrfs_header_nritems(leaf) == 0) {
- u64 root_gen;
- u64 bytenr = leaf->start;
- u32 blocksize = leaf->len;
-
- root_gen = btrfs_header_generation(
- path->nodes[1]);
-
- wret = del_ptr(trans, root, path, 1, slot);
- if (wret)
- ret = wret;
-
+ path->slots[1] = slot;
+ ret = btrfs_del_leaf(trans, root, path, leaf->start);
+ BUG_ON(ret);
free_extent_buffer(leaf);
- wret = btrfs_free_extent(trans, root, bytenr,
- blocksize,
- btrfs_header_owner(path->nodes[1]),
- root_gen, 0, 0, 1);
- if (wret)
- ret = wret;
} else {
/* if we're still in the path, make sure
* we're dirty. Otherwise, one of the
* search the tree again to find a leaf with lesser keys
* returns 0 if it found something or 1 if there are no lesser leaves.
* returns < 0 on io errors.
+ *
+ * This may release the path, and so you may lose any locks held at the
+ * time you call it.
*/
int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path)
{
/*
* A helper function to walk down the tree starting at min_key, and looking
* for nodes or leaves that are either in cache or have a minimum
- * transaction id. This is used by the btree defrag code, but could
- * also be used to search for blocks that have changed since a given
- * transaction id.
+ * transaction id. This is used by the btree defrag code, and tree logging
*
* This does not cow, but it does stuff the starting key it finds back
* into min_key, so you can call btrfs_search_slot with cow=1 on the
* This honors path->lowest_level to prevent descent past a given level
* of the tree.
*
+ * min_trans indicates the oldest transaction that you are interested
+ * in walking through. Any nodes or leaves older than min_trans are
+ * skipped over (without reading them).
+ *
* returns zero if something useful was found, < 0 on error and 1 if there
* was nothing in the tree that matched the search criteria.
*/
level = btrfs_header_level(cur);
sret = bin_search(cur, min_key, level, &slot);
- /* at level = 0, we're done, setup the path and exit */
- if (level == 0) {
+ /* at the lowest level, we're done, setup the path and exit */
+ if (level == path->lowest_level) {
if (slot >= nritems)
goto find_next_key;
ret = 0;
projects / linux-2.6-block.git / blobdiff