Skip to content

L
linux
Commit 182608c8 authored by David Sterba's avatar David Sterba
Browse files
Options

btrfs: remove old unused commented out code 

Remove code which has been #if0-ed out for a very long time and does not
seem to be related to current codebase anymore.
Signed-off-by: default avatarDavid Sterba <dsterba@suse.cz>
parent f2a97a9d
Hide whitespace changes
Inline Side-by-side
Showing with 1 addition and 2071 deletions
fs/btrfs/delayed-ref.c
View file @ 182608c8
......@@ -709,79 +709,3 @@ btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr)
return btrfs_delayed_node_to_head(ref);
return NULL;
}
/*
* add a delayed ref to the tree. This does all of the accounting required
* to make sure the delayed ref is eventually processed before this
* transaction commits.
*
* The main point of this call is to add and remove a backreference in a single
* shot, taking the lock only once, and only searching for the head node once.
*
* It is the same as doing a ref add and delete in two separate calls.
*/
#if 0
int btrfs_update_delayed_ref(struct btrfs_trans_handle *trans,
u64 bytenr, u64 num_bytes, u64 orig_parent,
u64 parent, u64 orig_ref_root, u64 ref_root,
u64 orig_ref_generation, u64 ref_generation,
u64 owner_objectid, int pin)
{
struct btrfs_delayed_ref *ref;
struct btrfs_delayed_ref *old_ref;
struct btrfs_delayed_ref_head *head_ref;
struct btrfs_delayed_ref_root *delayed_refs;
int ret;
ref = kmalloc(sizeof(*ref), GFP_NOFS);
if (!ref)
return -ENOMEM;
old_ref = kmalloc(sizeof(*old_ref), GFP_NOFS);
if (!old_ref) {
kfree(ref);
return -ENOMEM;
}
/*
* the parent = 0 case comes from cases where we don't actually
* know the parent yet. It will get updated later via a add/drop
* pair.
*/
if (parent == 0)
parent = bytenr;
if (orig_parent == 0)
orig_parent = bytenr;
head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS);
if (!head_ref) {
kfree(ref);
kfree(old_ref);
return -ENOMEM;
}
delayed_refs = &trans->transaction->delayed_refs;
spin_lock(&delayed_refs->lock);
/*
* insert both the head node and the new ref without dropping
* the spin lock
*/
ret = __btrfs_add_delayed_ref(trans, &head_ref->node, bytenr, num_bytes,
(u64)-1, 0, 0, 0,
BTRFS_UPDATE_DELAYED_HEAD, 0);
BUG_ON(ret);
ret = __btrfs_add_delayed_ref(trans, &ref->node, bytenr, num_bytes,
parent, ref_root, ref_generation,
owner_objectid, BTRFS_ADD_DELAYED_REF, 0);
BUG_ON(ret);
ret = __btrfs_add_delayed_ref(trans, &old_ref->node, bytenr, num_bytes,
orig_parent, orig_ref_root,
orig_ref_generation, owner_objectid,
BTRFS_DROP_DELAYED_REF, pin);
BUG_ON(ret);
spin_unlock(&delayed_refs->lock);
return 0;
}
#endif
fs/btrfs/disk-io.c
View file @ 182608c8
......@@ -1348,35 +1348,6 @@ struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
return ERR_PTR(ret);
}
#if 0
struct btrfs_root *root;
int ret;
root = btrfs_read_fs_root_no_name(fs_info, location);
if (!root)
return NULL;
if (root->in_sysfs)
return root;
ret = btrfs_set_root_name(root, name, namelen);
if (ret) {
free_extent_buffer(root->node);
kfree(root);
return ERR_PTR(ret);
}
ret = btrfs_sysfs_add_root(root);
if (ret) {
free_extent_buffer(root->node);
kfree(root->name);
kfree(root);
return ERR_PTR(ret);
}
root->in_sysfs = 1;
return root;
#endif
static int btrfs_congested_fn(void *congested_data, int bdi_bits)
{
struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
......
fs/btrfs/extent-tree.c
View file @ 182608c8
......@@ -2522,126 +2522,6 @@ int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
return ret;
}
#if 0
int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
struct extent_buffer *buf, u32 nr_extents)
{
struct btrfs_key key;
struct btrfs_file_extent_item *fi;
u64 root_gen;
u32 nritems;
int i;
int level;
int ret = 0;
int shared = 0;
if (!root->ref_cows)
return 0;
if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
shared = 0;
root_gen = root->root_key.offset;
} else {
shared = 1;
root_gen = trans->transid - 1;
}
level = btrfs_header_level(buf);
nritems = btrfs_header_nritems(buf);
if (level == 0) {
struct btrfs_leaf_ref *ref;
struct btrfs_extent_info *info;
ref = btrfs_alloc_leaf_ref(root, nr_extents);
if (!ref) {
ret = -ENOMEM;
goto out;
}
ref->root_gen = root_gen;
ref->bytenr = buf->start;
ref->owner = btrfs_header_owner(buf);
ref->generation = btrfs_header_generation(buf);
ref->nritems = nr_extents;
info = ref->extents;
for (i = 0; nr_extents > 0 && i < nritems; i++) {
u64 disk_bytenr;
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;
disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
if (disk_bytenr == 0)
continue;
info->bytenr = disk_bytenr;
info->num_bytes =
btrfs_file_extent_disk_num_bytes(buf, fi);
info->objectid = key.objectid;
info->offset = key.offset;
info++;
}
ret = btrfs_add_leaf_ref(root, ref, shared);
if (ret == -EEXIST && shared) {
struct btrfs_leaf_ref *old;
old = btrfs_lookup_leaf_ref(root, ref->bytenr);
BUG_ON(!old);
btrfs_remove_leaf_ref(root, old);
btrfs_free_leaf_ref(root, old);
ret = btrfs_add_leaf_ref(root, ref, shared);
}
WARN_ON(ret);
btrfs_free_leaf_ref(root, ref);
}
out:
return ret;
}
/* when a block goes through cow, we update the reference counts of
* everything that block points to. The internal pointers of the block
* can be in just about any order, and it is likely to have clusters of
* things that are close together and clusters of things that are not.
*
* To help reduce the seeks that come with updating all of these reference
* counts, sort them by byte number before actual updates are done.
*
* struct refsort is used to match byte number to slot in the btree block.
* we sort based on the byte number and then use the slot to actually
* find the item.
*
* struct refsort is smaller than strcut btrfs_item and smaller than
* struct btrfs_key_ptr. Since we're currently limited to the page size
* for a btree block, there's no way for a kmalloc of refsorts for a
* single node to be bigger than a page.
*/
struct refsort {
u64 bytenr;
u32 slot;
};
/*
* for passing into sort()
*/
static int refsort_cmp(const void *a_void, const void *b_void)
{
const struct refsort *a = a_void;
const struct refsort *b = b_void;
if (a->bytenr < b->bytenr)
return -1;
if (a->bytenr > b->bytenr)
return 1;
return 0;
}
#endif
static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct extent_buffer *buf,
......@@ -3223,18 +3103,6 @@ int btrfs_check_data_free_space(struct inode *inode, u64 bytes)
goto again;
}
#if 0 /* I hope we never need this code again, just in case */
printk(KERN_ERR "no space left, need %llu, %llu bytes_used, "
"%llu bytes_reserved, " "%llu bytes_pinned, "
"%llu bytes_readonly, %llu may use %llu total\n",
(unsigned long long)bytes,
(unsigned long long)data_sinfo->bytes_used,
(unsigned long long)data_sinfo->bytes_reserved,
(unsigned long long)data_sinfo->bytes_pinned,
(unsigned long long)data_sinfo->bytes_readonly,
(unsigned long long)data_sinfo->bytes_may_use,
(unsigned long long)data_sinfo->total_bytes);
#endif
return -ENOSPC;
}
data_sinfo->bytes_may_use += bytes;
......@@ -3867,23 +3735,7 @@ static u64 calc_global_metadata_size(struct btrfs_fs_info *fs_info)
u64 meta_used;
u64 data_used;
int csum_size = btrfs_super_csum_size(&fs_info->super_copy);
#if 0
/*
* per tree used space accounting can be inaccuracy, so we
* can't rely on it.
*/
spin_lock(&fs_info->extent_root->accounting_lock);
num_bytes = btrfs_root_used(&fs_info->extent_root->root_item);
spin_unlock(&fs_info->extent_root->accounting_lock);
spin_lock(&fs_info->csum_root->accounting_lock);
num_bytes += btrfs_root_used(&fs_info->csum_root->root_item);
spin_unlock(&fs_info->csum_root->accounting_lock);
spin_lock(&fs_info->tree_root->accounting_lock);
num_bytes += btrfs_root_used(&fs_info->tree_root->root_item);
spin_unlock(&fs_info->tree_root->accounting_lock);
#endif
sinfo = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_DATA);
spin_lock(&sinfo->lock);
data_used = sinfo->bytes_used;
......@@ -3936,10 +3788,7 @@ static void update_global_block_rsv(struct btrfs_fs_info *fs_info)
block_rsv->reserved = block_rsv->size;
block_rsv->full = 1;
}
#if 0
printk(KERN_INFO"global block rsv size %llu reserved %llu\n",
block_rsv->size, block_rsv->reserved);
#endif
spin_unlock(&sinfo->lock);
spin_unlock(&block_rsv->lock);
}
......@@ -6596,1514 +6445,6 @@ int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
return ret;
}
#if 0
static unsigned long calc_ra(unsigned long start, unsigned long last,
unsigned long nr)
{
return min(last, start + nr - 1);
}
static noinline int relocate_inode_pages(struct inode *inode, u64 start,
u64 len)
{
u64 page_start;
u64 page_end;
unsigned long first_index;
unsigned long last_index;
unsigned long i;
struct page *page;
struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
struct file_ra_state *ra;
struct btrfs_ordered_extent *ordered;
unsigned int total_read = 0;
unsigned int total_dirty = 0;
int ret = 0;
ra = kzalloc(sizeof(*ra), GFP_NOFS);
if (!ra)
return -ENOMEM;
mutex_lock(&inode->i_mutex);
first_index = start >> PAGE_CACHE_SHIFT;
last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
/* make sure the dirty trick played by the caller work */
ret = invalidate_inode_pages2_range(inode->i_mapping,
first_index, last_index);
if (ret)
goto out_unlock;
file_ra_state_init(ra, inode->i_mapping);
for (i = first_index ; i <= last_index; i++) {
if (total_read % ra->ra_pages == 0) {
btrfs_force_ra(inode->i_mapping, ra, NULL, i,
calc_ra(i, last_index, ra->ra_pages));
}
total_read++;
again:
if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode))
BUG_ON(1);
page = grab_cache_page(inode->i_mapping, i);
if (!page) {
ret = -ENOMEM;
goto out_unlock;
}
if (!PageUptodate(page)) {
btrfs_readpage(NULL, page);
lock_page(page);
if (!PageUptodate(page)) {
unlock_page(page);
page_cache_release(page);
ret = -EIO;
goto out_unlock;
}
}
wait_on_page_writeback(page);
page_start = (u64)page->index << PAGE_CACHE_SHIFT;
page_end = page_start + PAGE_CACHE_SIZE - 1;
lock_extent(io_tree, page_start, page_end, GFP_NOFS);
ordered = btrfs_lookup_ordered_extent(inode, page_start);
if (ordered) {
unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
unlock_page(page);
page_cache_release(page);
btrfs_start_ordered_extent(inode, ordered, 1);
btrfs_put_ordered_extent(ordered);
goto again;
}
set_page_extent_mapped(page);
if (i == first_index)
set_extent_bits(io_tree, page_start, page_end,
EXTENT_BOUNDARY, GFP_NOFS);
btrfs_set_extent_delalloc(inode, page_start, page_end);
set_page_dirty(page);
total_dirty++;
unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
unlock_page(page);
page_cache_release(page);
}
out_unlock:
kfree(ra);
mutex_unlock(&inode->i_mutex);
balance_dirty_pages_ratelimited_nr(inode->i_mapping, total_dirty);
return ret;
}
static noinline int relocate_data_extent(struct inode *reloc_inode,
struct btrfs_key *extent_key,
u64 offset)
{
struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree;
struct extent_map *em;
u64 start = extent_key->objectid - offset;
u64 end = start + extent_key->offset - 1;
em = alloc_extent_map();
BUG_ON(!em);
em->start = start;
em->len = extent_key->offset;
em->block_len = extent_key->offset;
em->block_start = extent_key->objectid;
em->bdev = root->fs_info->fs_devices->latest_bdev;
set_bit(EXTENT_FLAG_PINNED, &em->flags);
/* setup extent map to cheat btrfs_readpage */
lock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
while (1) {
int ret;
write_lock(&em_tree->lock);
ret = add_extent_mapping(em_tree, em);
write_unlock(&em_tree->lock);
if (ret != -EEXIST) {
free_extent_map(em);
break;
}
btrfs_drop_extent_cache(reloc_inode, start, end, 0);
}
unlock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
return relocate_inode_pages(reloc_inode, start, extent_key->offset);
}
struct btrfs_ref_path {
u64 extent_start;
u64 nodes[BTRFS_MAX_LEVEL];
u64 root_objectid;
u64 root_generation;
u64 owner_objectid;
u32 num_refs;
int lowest_level;
int current_level;
int shared_level;
struct btrfs_key node_keys[BTRFS_MAX_LEVEL];
u64 new_nodes[BTRFS_MAX_LEVEL];
};
struct disk_extent {
u64 ram_bytes;
u64 disk_bytenr;
u64 disk_num_bytes;
u64 offset;
u64 num_bytes;
u8 compression;
u8 encryption;
u16 other_encoding;
};
static int is_cowonly_root(u64 root_objectid)
{
if (root_objectid == BTRFS_ROOT_TREE_OBJECTID ||
root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
root_objectid == BTRFS_DEV_TREE_OBJECTID ||
root_objectid == BTRFS_TREE_LOG_OBJECTID ||
root_objectid == BTRFS_CSUM_TREE_OBJECTID)
return 1;
return 0;
}
static noinline int __next_ref_path(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct btrfs_ref_path *ref_path,
int first_time)
{
struct extent_buffer *leaf;
struct btrfs_path *path;
struct btrfs_extent_ref *ref;
struct btrfs_key key;
struct btrfs_key found_key;
u64 bytenr;
u32 nritems;
int level;
int ret = 1;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
if (first_time) {
ref_path->lowest_level = -1;
ref_path->current_level = -1;
ref_path->shared_level = -1;
goto walk_up;
}
walk_down:
level = ref_path->current_level - 1;
while (level >= -1) {
u64 parent;
if (level < ref_path->lowest_level)
break;
if (level >= 0)
bytenr = ref_path->nodes[level];
else
bytenr = ref_path->extent_start;
BUG_ON(bytenr == 0);
parent = ref_path->nodes[level + 1];
ref_path->nodes[level + 1] = 0;
ref_path->current_level = level;
BUG_ON(parent == 0);
key.objectid = bytenr;
key.offset = parent + 1;
key.type = BTRFS_EXTENT_REF_KEY;
ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
if (ret < 0)
goto out;
BUG_ON(ret == 0);
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
if (path->slots[0] >= nritems) {
ret = btrfs_next_leaf(extent_root, path);
if (ret < 0)
goto out;
if (ret > 0)
goto next;
leaf = path->nodes[0];
}
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
if (found_key.objectid == bytenr &&
found_key.type == BTRFS_EXTENT_REF_KEY) {
if (level < ref_path->shared_level)
ref_path->shared_level = level;
goto found;
}
next:
level--;
btrfs_release_path(extent_root, path);
cond_resched();
}
/* reached lowest level */
ret = 1;
goto out;
walk_up:
level = ref_path->current_level;
while (level < BTRFS_MAX_LEVEL - 1) {
u64 ref_objectid;
if (level >= 0)
bytenr = ref_path->nodes[level];
else
bytenr = ref_path->extent_start;
BUG_ON(bytenr == 0);
key.objectid = bytenr;
key.offset = 0;
key.type = BTRFS_EXTENT_REF_KEY;
ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
if (ret < 0)
goto out;
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
if (path->slots[0] >= nritems) {
ret = btrfs_next_leaf(extent_root, path);
if (ret < 0)
goto out;
if (ret > 0) {
/* the extent was freed by someone */
if (ref_path->lowest_level == level)
goto out;
btrfs_release_path(extent_root, path);
goto walk_down;
}
leaf = path->nodes[0];
}
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
if (found_key.objectid != bytenr ||
found_key.type != BTRFS_EXTENT_REF_KEY) {
/* the extent was freed by someone */
if (ref_path->lowest_level == level) {
ret = 1;
goto out;
}
btrfs_release_path(extent_root, path);
goto walk_down;
}
found:
ref = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_ref);
ref_objectid = btrfs_ref_objectid(leaf, ref);
if (ref_objectid < BTRFS_FIRST_FREE_OBJECTID) {
if (first_time) {
level = (int)ref_objectid;
BUG_ON(level >= BTRFS_MAX_LEVEL);
ref_path->lowest_level = level;
ref_path->current_level = level;
ref_path->nodes[level] = bytenr;
} else {
WARN_ON(ref_objectid != level);
}
} else {
WARN_ON(level != -1);
}
first_time = 0;
if (ref_path->lowest_level == level) {
ref_path->owner_objectid = ref_objectid;
ref_path->num_refs = btrfs_ref_num_refs(leaf, ref);
}
/*
* the block is tree root or the block isn't in reference
* counted tree.
*/
if (found_key.objectid == found_key.offset ||
is_cowonly_root(btrfs_ref_root(leaf, ref))) {
ref_path->root_objectid = btrfs_ref_root(leaf, ref);
ref_path->root_generation =
btrfs_ref_generation(leaf, ref);
if (level < 0) {
/* special reference from the tree log */
ref_path->nodes[0] = found_key.offset;
ref_path->current_level = 0;
}
ret = 0;
goto out;
}
level++;
BUG_ON(ref_path->nodes[level] != 0);
ref_path->nodes[level] = found_key.offset;
ref_path->current_level = level;
/*
* the reference was created in the running transaction,
* no need to continue walking up.
*/
if (btrfs_ref_generation(leaf, ref) == trans->transid) {
ref_path->root_objectid = btrfs_ref_root(leaf, ref);
ref_path->root_generation =
btrfs_ref_generation(leaf, ref);
ret = 0;
goto out;
}
btrfs_release_path(extent_root, path);
cond_resched();
}
/* reached max tree level, but no tree root found. */
BUG();
out:
btrfs_free_path(path);
return ret;
}
static int btrfs_first_ref_path(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct btrfs_ref_path *ref_path,
u64 extent_start)
{
memset(ref_path, 0, sizeof(*ref_path));
ref_path->extent_start = extent_start;
return __next_ref_path(trans, extent_root, ref_path, 1);
}
static int btrfs_next_ref_path(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct btrfs_ref_path *ref_path)
{
return __next_ref_path(trans, extent_root, ref_path, 0);
}
static noinline int get_new_locations(struct inode *reloc_inode,
struct btrfs_key *extent_key,
u64 offset, int no_fragment,
struct disk_extent **extents,
int *nr_extents)
{
struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
struct btrfs_path *path;
struct btrfs_file_extent_item *fi;
struct extent_buffer *leaf;
struct disk_extent *exts = *extents;
struct btrfs_key found_key;
u64 cur_pos;
u64 last_byte;
u32 nritems;
int nr = 0;
int max = *nr_extents;
int ret;
WARN_ON(!no_fragment && *extents);
if (!exts) {
max = 1;
exts = kmalloc(sizeof(*exts) * max, GFP_NOFS);
if (!exts)
return -ENOMEM;
}
path = btrfs_alloc_path();
if (!path) {
if (exts != *extents)
kfree(exts);
return -ENOMEM;
}
cur_pos = extent_key->objectid - offset;
last_byte = extent_key->objectid + extent_key->offset;
ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
cur_pos, 0);
if (ret < 0)
goto out;
if (ret > 0) {
ret = -ENOENT;
goto out;
}
while (1) {
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
if (path->slots[0] >= nritems) {
ret = btrfs_next_leaf(root, path);
if (ret < 0)
goto out;
if (ret > 0)
break;
leaf = path->nodes[0];
}
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
if (found_key.offset != cur_pos ||
found_key.type != BTRFS_EXTENT_DATA_KEY ||
found_key.objectid != reloc_inode->i_ino)
break;
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
if (btrfs_file_extent_type(leaf, fi) !=
BTRFS_FILE_EXTENT_REG ||
btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
break;
if (nr == max) {
struct disk_extent *old = exts;
max *= 2;
exts = kzalloc(sizeof(*exts) * max, GFP_NOFS);
if (!exts) {
ret = -ENOMEM;
goto out;
}
memcpy(exts, old, sizeof(*exts) * nr);
if (old != *extents)
kfree(old);
}
exts[nr].disk_bytenr =
btrfs_file_extent_disk_bytenr(leaf, fi);
exts[nr].disk_num_bytes =
btrfs_file_extent_disk_num_bytes(leaf, fi);
exts[nr].offset = btrfs_file_extent_offset(leaf, fi);
exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
exts[nr].ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
exts[nr].compression = btrfs_file_extent_compression(leaf, fi);
exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi);
exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf,
fi);
BUG_ON(exts[nr].offset > 0);
BUG_ON(exts[nr].compression || exts[nr].encryption);
BUG_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
cur_pos += exts[nr].num_bytes;
nr++;
if (cur_pos + offset >= last_byte)
break;
if (no_fragment) {
ret = 1;
goto out;
}
path->slots[0]++;
}
BUG_ON(cur_pos + offset > last_byte);
if (cur_pos + offset < last_byte) {
ret = -ENOENT;
goto out;
}
ret = 0;
out:
btrfs_free_path(path);
if (ret) {
if (exts != *extents)
kfree(exts);
} else {
*extents = exts;
*nr_extents = nr;
}
return ret;
}
static noinline int replace_one_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
struct btrfs_key *extent_key,
struct btrfs_key *leaf_key,
struct btrfs_ref_path *ref_path,
struct disk_extent *new_extents,
int nr_extents)
{
struct extent_buffer *leaf;
struct btrfs_file_extent_item *fi;
struct inode *inode = NULL;
struct btrfs_key key;
u64 lock_start = 0;
u64 lock_end = 0;
u64 num_bytes;
u64 ext_offset;
u64 search_end = (u64)-1;
u32 nritems;
int nr_scaned = 0;
int extent_locked = 0;
int extent_type;
int ret;
memcpy(&key, leaf_key, sizeof(key));
if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
if (key.objectid < ref_path->owner_objectid ||
(key.objectid == ref_path->owner_objectid &&
key.type < BTRFS_EXTENT_DATA_KEY)) {
key.objectid = ref_path->owner_objectid;
key.type = BTRFS_EXTENT_DATA_KEY;
key.offset = 0;
}
}
while (1) {
ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
if (ret < 0)
goto out;
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
next:
if (extent_locked && ret > 0) {
/*
* the file extent item was modified by someone
* before the extent got locked.
*/
unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
lock_end, GFP_NOFS);
extent_locked = 0;
}
if (path->slots[0] >= nritems) {
if (++nr_scaned > 2)
break;
BUG_ON(extent_locked);
ret = btrfs_next_leaf(root, path);
if (ret < 0)
goto out;
if (ret > 0)
break;
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
}
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
if ((key.objectid > ref_path->owner_objectid) ||
(key.objectid == ref_path->owner_objectid &&
key.type > BTRFS_EXTENT_DATA_KEY) ||
key.offset >= search_end)
break;
}
if (inode && key.objectid != inode->i_ino) {
BUG_ON(extent_locked);
btrfs_release_path(root, path);
mutex_unlock(&inode->i_mutex);
iput(inode);
inode = NULL;
continue;
}
if (key.type != BTRFS_EXTENT_DATA_KEY) {
path->slots[0]++;
ret = 1;
goto next;
}
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
extent_type = btrfs_file_extent_type(leaf, fi);
if ((extent_type != BTRFS_FILE_EXTENT_REG &&
extent_type != BTRFS_FILE_EXTENT_PREALLOC) ||
(btrfs_file_extent_disk_bytenr(leaf, fi) !=
extent_key->objectid)) {
path->slots[0]++;
ret = 1;
goto next;
}
num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
ext_offset = btrfs_file_extent_offset(leaf, fi);
if (search_end == (u64)-1) {
search_end = key.offset - ext_offset +
btrfs_file_extent_ram_bytes(leaf, fi);
}
if (!extent_locked) {
lock_start = key.offset;
lock_end = lock_start + num_bytes - 1;
} else {
if (lock_start > key.offset ||
lock_end + 1 < key.offset + num_bytes) {
unlock_extent(&BTRFS_I(inode)->io_tree,
lock_start, lock_end, GFP_NOFS);
extent_locked = 0;
}
}
if (!inode) {
btrfs_release_path(root, path);
inode = btrfs_iget_locked(root->fs_info->sb,
key.objectid, root);
if (inode->i_state & I_NEW) {
BTRFS_I(inode)->root = root;
BTRFS_I(inode)->location.objectid =
key.objectid;
BTRFS_I(inode)->location.type =
BTRFS_INODE_ITEM_KEY;
BTRFS_I(inode)->location.offset = 0;
btrfs_read_locked_inode(inode);
unlock_new_inode(inode);
}
/*
* some code call btrfs_commit_transaction while
* holding the i_mutex, so we can't use mutex_lock
* here.
*/
if (is_bad_inode(inode) ||
!mutex_trylock(&inode->i_mutex)) {
iput(inode);
inode = NULL;
key.offset = (u64)-1;
goto skip;
}
}
if (!extent_locked) {
struct btrfs_ordered_extent *ordered;
btrfs_release_path(root, path);
lock_extent(&BTRFS_I(inode)->io_tree, lock_start,
lock_end, GFP_NOFS);
ordered = btrfs_lookup_first_ordered_extent(inode,
lock_end);
if (ordered &&
ordered->file_offset <= lock_end &&
ordered->file_offset + ordered->len > lock_start) {
unlock_extent(&BTRFS_I(inode)->io_tree,
lock_start, lock_end, GFP_NOFS);
btrfs_start_ordered_extent(inode, ordered, 1);
btrfs_put_ordered_extent(ordered);
key.offset += num_bytes;
goto skip;
}
if (ordered)
btrfs_put_ordered_extent(ordered);
extent_locked = 1;
continue;
}
if (nr_extents == 1) {
/* update extent pointer in place */
btrfs_set_file_extent_disk_bytenr(leaf, fi,
new_extents[0].disk_bytenr);
btrfs_set_file_extent_disk_num_bytes(leaf, fi,
new_extents[0].disk_num_bytes);
btrfs_mark_buffer_dirty(leaf);
btrfs_drop_extent_cache(inode, key.offset,
key.offset + num_bytes - 1, 0);
ret = btrfs_inc_extent_ref(trans, root,
new_extents[0].disk_bytenr,
new_extents[0].disk_num_bytes,
leaf->start,
root->root_key.objectid,
trans->transid,
key.objectid);
BUG_ON(ret);
ret = btrfs_free_extent(trans, root,
extent_key->objectid,
extent_key->offset,
leaf->start,
btrfs_header_owner(leaf),
btrfs_header_generation(leaf),
key.objectid, 0);
BUG_ON(ret);
btrfs_release_path(root, path);
key.offset += num_bytes;
} else {
BUG_ON(1);
#if 0
u64 alloc_hint;
u64 extent_len;
int i;
/*
* drop old extent pointer at first, then insert the
* new pointers one bye one
*/
btrfs_release_path(root, path);
ret = btrfs_drop_extents(trans, root, inode, key.offset,
key.offset + num_bytes,
key.offset, &alloc_hint);
BUG_ON(ret);
for (i = 0; i < nr_extents; i++) {
if (ext_offset >= new_extents[i].num_bytes) {
ext_offset -= new_extents[i].num_bytes;
continue;
}
extent_len = min(new_extents[i].num_bytes -
ext_offset, num_bytes);
ret = btrfs_insert_empty_item(trans, root,
path, &key,
sizeof(*fi));
BUG_ON(ret);
leaf = path->nodes[0];
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
btrfs_set_file_extent_generation(leaf, fi,
trans->transid);
btrfs_set_file_extent_type(leaf, fi,
BTRFS_FILE_EXTENT_REG);
btrfs_set_file_extent_disk_bytenr(leaf, fi,
new_extents[i].disk_bytenr);
btrfs_set_file_extent_disk_num_bytes(leaf, fi,
new_extents[i].disk_num_bytes);
btrfs_set_file_extent_ram_bytes(leaf, fi,
new_extents[i].ram_bytes);
btrfs_set_file_extent_compression(leaf, fi,
new_extents[i].compression);
btrfs_set_file_extent_encryption(leaf, fi,
new_extents[i].encryption);
btrfs_set_file_extent_other_encoding(leaf, fi,
new_extents[i].other_encoding);
btrfs_set_file_extent_num_bytes(leaf, fi,
extent_len);
ext_offset += new_extents[i].offset;
btrfs_set_file_extent_offset(leaf, fi,
ext_offset);
btrfs_mark_buffer_dirty(leaf);
btrfs_drop_extent_cache(inode, key.offset,
key.offset + extent_len - 1, 0);
ret = btrfs_inc_extent_ref(trans, root,
new_extents[i].disk_bytenr,
new_extents[i].disk_num_bytes,
leaf->start,
root->root_key.objectid,
trans->transid, key.objectid);
BUG_ON(ret);
btrfs_release_path(root, path);
inode_add_bytes(inode, extent_len);
ext_offset = 0;
num_bytes -= extent_len;
key.offset += extent_len;
if (num_bytes == 0)
break;
}
BUG_ON(i >= nr_extents);
#endif
}
if (extent_locked) {
unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
lock_end, GFP_NOFS);
extent_locked = 0;
}
skip:
if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS &&
key.offset >= search_end)
break;
cond_resched();
}
ret = 0;
out:
btrfs_release_path(root, path);
if (inode) {
mutex_unlock(&inode->i_mutex);
if (extent_locked) {
unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
lock_end, GFP_NOFS);
}
iput(inode);
}
return ret;
}
int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct extent_buffer *buf, u64 orig_start)
{
int level;
int ret;
BUG_ON(btrfs_header_generation(buf) != trans->transid);
BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
level = btrfs_header_level(buf);
if (level == 0) {
struct btrfs_leaf_ref *ref;
struct btrfs_leaf_ref *orig_ref;
orig_ref = btrfs_lookup_leaf_ref(root, orig_start);
if (!orig_ref)
return -ENOENT;
ref = btrfs_alloc_leaf_ref(root, orig_ref->nritems);
if (!ref) {
btrfs_free_leaf_ref(root, orig_ref);
return -ENOMEM;
}
ref->nritems = orig_ref->nritems;
memcpy(ref->extents, orig_ref->extents,
sizeof(ref->extents[0]) * ref->nritems);
btrfs_free_leaf_ref(root, orig_ref);
ref->root_gen = trans->transid;
ref->bytenr = buf->start;
ref->owner = btrfs_header_owner(buf);
ref->generation = btrfs_header_generation(buf);
ret = btrfs_add_leaf_ref(root, ref, 0);
WARN_ON(ret);
btrfs_free_leaf_ref(root, ref);
}
return 0;
}
static noinline int invalidate_extent_cache(struct btrfs_root *root,
struct extent_buffer *leaf,
struct btrfs_block_group_cache *group,
struct btrfs_root *target_root)
{
struct btrfs_key key;
struct inode *inode = NULL;
struct btrfs_file_extent_item *fi;
struct extent_state *cached_state = NULL;
u64 num_bytes;
u64 skip_objectid = 0;
u32 nritems;
u32 i;
nritems = btrfs_header_nritems(leaf);
for (i = 0; i < nritems; i++) {
btrfs_item_key_to_cpu(leaf, &key, i);
if (key.objectid == skip_objectid ||
key.type != BTRFS_EXTENT_DATA_KEY)
continue;
fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
if (btrfs_file_extent_type(leaf, fi) ==
BTRFS_FILE_EXTENT_INLINE)
continue;
if (btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
continue;
if (!inode || inode->i_ino != key.objectid) {
iput(inode);
inode = btrfs_ilookup(target_root->fs_info->sb,
key.objectid, target_root, 1);
}
if (!inode) {
skip_objectid = key.objectid;
continue;
}
num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
lock_extent_bits(&BTRFS_I(inode)->io_tree, key.offset,
key.offset + num_bytes - 1, 0, &cached_state,
GFP_NOFS);
btrfs_drop_extent_cache(inode, key.offset,
key.offset + num_bytes - 1, 1);
unlock_extent_cached(&BTRFS_I(inode)->io_tree, key.offset,
key.offset + num_bytes - 1, &cached_state,
GFP_NOFS);
cond_resched();
}
iput(inode);
return 0;
}
static noinline int replace_extents_in_leaf(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct extent_buffer *leaf,
struct btrfs_block_group_cache *group,
struct inode *reloc_inode)
{
struct btrfs_key key;
struct btrfs_key extent_key;
struct btrfs_file_extent_item *fi;
struct btrfs_leaf_ref *ref;
struct disk_extent *new_extent;
u64 bytenr;
u64 num_bytes;
u32 nritems;
u32 i;
int ext_index;
int nr_extent;
int ret;
new_extent = kmalloc(sizeof(*new_extent), GFP_NOFS);
if (!new_extent)
return -ENOMEM;
ref = btrfs_lookup_leaf_ref(root, leaf->start);
BUG_ON(!ref);
ext_index = -1;
nritems = btrfs_header_nritems(leaf);
for (i = 0; i < nritems; i++) {
btrfs_item_key_to_cpu(leaf, &key, i);
if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
continue;
fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
if (btrfs_file_extent_type(leaf, fi) ==
BTRFS_FILE_EXTENT_INLINE)
continue;
bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
if (bytenr == 0)
continue;
ext_index++;
if (bytenr >= group->key.objectid + group->key.offset ||
bytenr + num_bytes <= group->key.objectid)
continue;
extent_key.objectid = bytenr;
extent_key.offset = num_bytes;
extent_key.type = BTRFS_EXTENT_ITEM_KEY;
nr_extent = 1;
ret = get_new_locations(reloc_inode, &extent_key,
group->key.objectid, 1,
&new_extent, &nr_extent);
if (ret > 0)
continue;
BUG_ON(ret < 0);
BUG_ON(ref->extents[ext_index].bytenr != bytenr);
BUG_ON(ref->extents[ext_index].num_bytes != num_bytes);
ref->extents[ext_index].bytenr = new_extent->disk_bytenr;
ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
btrfs_set_file_extent_disk_bytenr(leaf, fi,
new_extent->disk_bytenr);
btrfs_set_file_extent_disk_num_bytes(leaf, fi,
new_extent->disk_num_bytes);
btrfs_mark_buffer_dirty(leaf);
ret = btrfs_inc_extent_ref(trans, root,
new_extent->disk_bytenr,
new_extent->disk_num_bytes,
leaf->start,
root->root_key.objectid,
trans->transid, key.objectid);
BUG_ON(ret);
ret = btrfs_free_extent(trans, root,
bytenr, num_bytes, leaf->start,
btrfs_header_owner(leaf),
btrfs_header_generation(leaf),
key.objectid, 0);
BUG_ON(ret);
cond_resched();
}
kfree(new_extent);
BUG_ON(ext_index + 1 != ref->nritems);
btrfs_free_leaf_ref(root, ref);
return 0;
}
int btrfs_free_reloc_root(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
struct btrfs_root *reloc_root;
int ret;
if (root->reloc_root) {
reloc_root = root->reloc_root;
root->reloc_root = NULL;
list_add(&reloc_root->dead_list,
&root->fs_info->dead_reloc_roots);
btrfs_set_root_bytenr(&reloc_root->root_item,
reloc_root->node->start);
btrfs_set_root_level(&root->root_item,
btrfs_header_level(reloc_root->node));
memset(&reloc_root->root_item.drop_progress, 0,
sizeof(struct btrfs_disk_key));
reloc_root->root_item.drop_level = 0;
ret = btrfs_update_root(trans, root->fs_info->tree_root,
&reloc_root->root_key,
&reloc_root->root_item);
BUG_ON(ret);
}
return 0;
}
int btrfs_drop_dead_reloc_roots(struct btrfs_root *root)
{
struct btrfs_trans_handle *trans;
struct btrfs_root *reloc_root;
struct btrfs_root *prev_root = NULL;
struct list_head dead_roots;
int ret;
unsigned long nr;
INIT_LIST_HEAD(&dead_roots);
list_splice_init(&root->fs_info->dead_reloc_roots, &dead_roots);
while (!list_empty(&dead_roots)) {
reloc_root = list_entry(dead_roots.prev,
struct btrfs_root, dead_list);
list_del_init(&reloc_root->dead_list);
BUG_ON(reloc_root->commit_root != NULL);
while (1) {
trans = btrfs_join_transaction(root, 1);
BUG_ON(IS_ERR(trans));
mutex_lock(&root->fs_info->drop_mutex);
ret = btrfs_drop_snapshot(trans, reloc_root);
if (ret != -EAGAIN)
break;
mutex_unlock(&root->fs_info->drop_mutex);
nr = trans->blocks_used;
ret = btrfs_end_transaction(trans, root);
BUG_ON(ret);
btrfs_btree_balance_dirty(root, nr);
}
free_extent_buffer(reloc_root->node);
ret = btrfs_del_root(trans, root->fs_info->tree_root,
&reloc_root->root_key);
BUG_ON(ret);
mutex_unlock(&root->fs_info->drop_mutex);
nr = trans->blocks_used;
ret = btrfs_end_transaction(trans, root);
BUG_ON(ret);
btrfs_btree_balance_dirty(root, nr);
kfree(prev_root);
prev_root = reloc_root;
}
if (prev_root) {
btrfs_remove_leaf_refs(prev_root, (u64)-1, 0);
kfree(prev_root);
}
return 0;
}
int btrfs_add_dead_reloc_root(struct btrfs_root *root)
{
list_add(&root->dead_list, &root->fs_info->dead_reloc_roots);
return 0;
}
int btrfs_cleanup_reloc_trees(struct btrfs_root *root)
{
struct btrfs_root *reloc_root;
struct btrfs_trans_handle *trans;
struct btrfs_key location;
int found;
int ret;
mutex_lock(&root->fs_info->tree_reloc_mutex);
ret = btrfs_find_dead_roots(root, BTRFS_TREE_RELOC_OBJECTID, NULL);
BUG_ON(ret);
found = !list_empty(&root->fs_info->dead_reloc_roots);
mutex_unlock(&root->fs_info->tree_reloc_mutex);
if (found) {
trans = btrfs_start_transaction(root, 1);
BUG_ON(IS_ERR(trans));
ret = btrfs_commit_transaction(trans, root);
BUG_ON(ret);
}
location.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
location.offset = (u64)-1;
location.type = BTRFS_ROOT_ITEM_KEY;
reloc_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
BUG_ON(!reloc_root);
ret = btrfs_orphan_cleanup(reloc_root);
BUG_ON(ret);
return 0;
}
static noinline int init_reloc_tree(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
struct btrfs_root *reloc_root;
struct extent_buffer *eb;
struct btrfs_root_item *root_item;
struct btrfs_key root_key;
int ret;
BUG_ON(!root->ref_cows);
if (root->reloc_root)
return 0;
root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
if (!root_item)
return -ENOMEM;
ret = btrfs_copy_root(trans, root, root->commit_root,
&eb, BTRFS_TREE_RELOC_OBJECTID);
BUG_ON(ret);
root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
root_key.offset = root->root_key.objectid;
root_key.type = BTRFS_ROOT_ITEM_KEY;
memcpy(root_item, &root->root_item, sizeof(root_item));
btrfs_set_root_refs(root_item, 0);
btrfs_set_root_bytenr(root_item, eb->start);
btrfs_set_root_level(root_item, btrfs_header_level(eb));
btrfs_set_root_generation(root_item, trans->transid);
btrfs_tree_unlock(eb);
free_extent_buffer(eb);
ret = btrfs_insert_root(trans, root->fs_info->tree_root,
&root_key, root_item);
BUG_ON(ret);
kfree(root_item);
reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
&root_key);
BUG_ON(IS_ERR(reloc_root));
reloc_root->last_trans = trans->transid;
reloc_root->commit_root = NULL;
reloc_root->ref_tree = &root->fs_info->reloc_ref_tree;
root->reloc_root = reloc_root;
return 0;
}
/*
* Core function of space balance.
*
* The idea is using reloc trees to relocate tree blocks in reference
* counted roots. There is one reloc tree for each subvol, and all
* reloc trees share same root key objectid. Reloc trees are snapshots
* of the latest committed roots of subvols (root->commit_root).
*
* To relocate a tree block referenced by a subvol, there are two steps.
* COW the block through subvol's reloc tree, then update block pointer
* in the subvol to point to the new block. Since all reloc trees share
* same root key objectid, doing special handing for tree blocks owned
* by them is easy. Once a tree block has been COWed in one reloc tree,
* we can use the resulting new block directly when the same block is
* required to COW again through other reloc trees. By this way, relocated
* tree blocks are shared between reloc trees, so they are also shared
* between subvols.
*/
static noinline int relocate_one_path(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
struct btrfs_key *first_key,
struct btrfs_ref_path *ref_path,
struct btrfs_block_group_cache *group,
struct inode *reloc_inode)
{
struct btrfs_root *reloc_root;
struct extent_buffer *eb = NULL;
struct btrfs_key *keys;
u64 *nodes;
int level;
int shared_level;
int lowest_level = 0;
int ret;
if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
lowest_level = ref_path->owner_objectid;
if (!root->ref_cows) {
path->lowest_level = lowest_level;
ret = btrfs_search_slot(trans, root, first_key, path, 0, 1);
BUG_ON(ret < 0);
path->lowest_level = 0;
btrfs_release_path(root, path);
return 0;
}
mutex_lock(&root->fs_info->tree_reloc_mutex);
ret = init_reloc_tree(trans, root);
BUG_ON(ret);
reloc_root = root->reloc_root;
shared_level = ref_path->shared_level;
ref_path->shared_level = BTRFS_MAX_LEVEL - 1;
keys = ref_path->node_keys;
nodes = ref_path->new_nodes;
memset(&keys[shared_level + 1], 0,
sizeof(*keys) * (BTRFS_MAX_LEVEL - shared_level - 1));
memset(&nodes[shared_level + 1], 0,
sizeof(*nodes) * (BTRFS_MAX_LEVEL - shared_level - 1));
if (nodes[lowest_level] == 0) {
path->lowest_level = lowest_level;
ret = btrfs_search_slot(trans, reloc_root, first_key, path,
0, 1);
BUG_ON(ret);
for (level = lowest_level; level < BTRFS_MAX_LEVEL; level++) {
eb = path->nodes[level];
if (!eb || eb == reloc_root->node)
break;
nodes[level] = eb->start;
if (level == 0)
btrfs_item_key_to_cpu(eb, &keys[level], 0);
else
btrfs_node_key_to_cpu(eb, &keys[level], 0);
}
if (nodes[0] &&
ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
eb = path->nodes[0];
ret = replace_extents_in_leaf(trans, reloc_root, eb,
group, reloc_inode);
BUG_ON(ret);
}
btrfs_release_path(reloc_root, path);
} else {
ret = btrfs_merge_path(trans, reloc_root, keys, nodes,
lowest_level);
BUG_ON(ret);
}
/*
* replace tree blocks in the fs tree with tree blocks in
* the reloc tree.
*/
ret = btrfs_merge_path(trans, root, keys, nodes, lowest_level);
BUG_ON(ret < 0);
if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
ret = btrfs_search_slot(trans, reloc_root, first_key, path,
0, 0);
BUG_ON(ret);
extent_buffer_get(path->nodes[0]);
eb = path->nodes[0];
btrfs_release_path(reloc_root, path);
ret = invalidate_extent_cache(reloc_root, eb, group, root);
BUG_ON(ret);
free_extent_buffer(eb);
}
mutex_unlock(&root->fs_info->tree_reloc_mutex);
path->lowest_level = 0;
return 0;
}
static noinline int relocate_tree_block(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
struct btrfs_key *first_key,
struct btrfs_ref_path *ref_path)
{
int ret;
ret = relocate_one_path(trans, root, path, first_key,
ref_path, NULL, NULL);
BUG_ON(ret);
return 0;
}
static noinline int del_extent_zero(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct btrfs_path *path,
struct btrfs_key *extent_key)
{
int ret;
ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1);
if (ret)
goto out;
ret = btrfs_del_item(trans, extent_root, path);
out:
btrfs_release_path(extent_root, path);
return ret;
}
static noinline struct btrfs_root *read_ref_root(struct btrfs_fs_info *fs_info,
struct btrfs_ref_path *ref_path)
{
struct btrfs_key root_key;
root_key.objectid = ref_path->root_objectid;
root_key.type = BTRFS_ROOT_ITEM_KEY;
if (is_cowonly_root(ref_path->root_objectid))
root_key.offset = 0;
else
root_key.offset = (u64)-1;
return btrfs_read_fs_root_no_name(fs_info, &root_key);
}
static noinline int relocate_one_extent(struct btrfs_root *extent_root,
struct btrfs_path *path,
struct btrfs_key *extent_key,
struct btrfs_block_group_cache *group,
struct inode *reloc_inode, int pass)
{
struct btrfs_trans_handle *trans;
struct btrfs_root *found_root;
struct btrfs_ref_path *ref_path = NULL;
struct disk_extent *new_extents = NULL;
int nr_extents = 0;
int loops;
int ret;
int level;
struct btrfs_key first_key;
u64 prev_block = 0;
trans = btrfs_start_transaction(extent_root, 1);
BUG_ON(IS_ERR(trans));
if (extent_key->objectid == 0) {
ret = del_extent_zero(trans, extent_root, path, extent_key);
goto out;
}
ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS);
if (!ref_path) {
ret = -ENOMEM;
goto out;
}
for (loops = 0; ; loops++) {
if (loops == 0) {
ret = btrfs_first_ref_path(trans, extent_root, ref_path,
extent_key->objectid);
} else {
ret = btrfs_next_ref_path(trans, extent_root, ref_path);
}
if (ret < 0)
goto out;
if (ret > 0)
break;
if (ref_path->root_objectid == BTRFS_TREE_LOG_OBJECTID ||
ref_path->root_objectid == BTRFS_TREE_RELOC_OBJECTID)
continue;
found_root = read_ref_root(extent_root->fs_info, ref_path);
BUG_ON(!found_root);
/*
* for reference counted tree, only process reference paths
* rooted at the latest committed root.
*/
if (found_root->ref_cows &&
ref_path->root_generation != found_root->root_key.offset)
continue;
if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
if (pass == 0) {
/*
* copy data extents to new locations
*/
u64 group_start = group->key.objectid;
ret = relocate_data_extent(reloc_inode,
extent_key,
group_start);
if (ret < 0)
goto out;
break;
}
level = 0;
} else {
level = ref_path->owner_objectid;
}
if (prev_block != ref_path->nodes[level]) {
struct extent_buffer *eb;
u64 block_start = ref_path->nodes[level];
u64 block_size = btrfs_level_size(found_root, level);
eb = read_tree_block(found_root, block_start,
block_size, 0);
if (!eb) {
ret = -EIO;
goto out;
}
btrfs_tree_lock(eb);
BUG_ON(level != btrfs_header_level(eb));
if (level == 0)
btrfs_item_key_to_cpu(eb, &first_key, 0);
else
btrfs_node_key_to_cpu(eb, &first_key, 0);
btrfs_tree_unlock(eb);
free_extent_buffer(eb);
prev_block = block_start;
}
mutex_lock(&extent_root->fs_info->trans_mutex);
btrfs_record_root_in_trans(found_root);
mutex_unlock(&extent_root->fs_info->trans_mutex);
if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
/*
* try to update data extent references while
* keeping metadata shared between snapshots.
*/
if (pass == 1) {
ret = relocate_one_path(trans, found_root,
path, &first_key, ref_path,
group, reloc_inode);
if (ret < 0)
goto out;
continue;
}
/*
* use fallback method to process the remaining
* references.
*/
if (!new_extents) {
u64 group_start = group->key.objectid;
new_extents = kmalloc(sizeof(*new_extents),
GFP_NOFS);
if (!new_extents) {
ret = -ENOMEM;
goto out;
}
nr_extents = 1;
ret = get_new_locations(reloc_inode,
extent_key,
group_start, 1,
&new_extents,
&nr_extents);
if (ret)
goto out;
}
ret = replace_one_extent(trans, found_root,
path, extent_key,
&first_key, ref_path,
new_extents, nr_extents);
} else {
ret = relocate_tree_block(trans, found_root, path,
&first_key, ref_path);
}
if (ret < 0)
goto out;
}
ret = 0;
out:
btrfs_end_transaction(trans, extent_root);
kfree(new_extents);
kfree(ref_path);
return ret;
}
#endif
static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
{
u64 num_devices;
......
fs/btrfs/inode.c
View file @ 182608c8
......@@ -3093,178 +3093,6 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
return err;
}
#if 0
/*
* when truncating bytes in a file, it is possible to avoid reading
* the leaves that contain only checksum items. This can be the
* majority of the IO required to delete a large file, but it must
* be done carefully.
*
* The keys in the level just above the leaves are checked to make sure
* the lowest key in a given leaf is a csum key, and starts at an offset
* after the new size.
*
* Then the key for the next leaf is checked to make sure it also has
* a checksum item for the same file. If it does, we know our target leaf
* contains only checksum items, and it can be safely freed without reading
* it.
*
* This is just an optimization targeted at large files. It may do
* nothing. It will return 0 unless things went badly.
*/
static noinline int drop_csum_leaves(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
struct inode *inode, u64 new_size)
{
struct btrfs_key key;
int ret;
int nritems;
struct btrfs_key found_key;
struct btrfs_key other_key;
struct btrfs_leaf_ref *ref;
u64 leaf_gen;
u64 leaf_start;
path->lowest_level = 1;
key.objectid = inode->i_ino;
key.type = BTRFS_CSUM_ITEM_KEY;
key.offset = new_size;
again:
ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
if (ret < 0)
goto out;
if (path->nodes[1] == NULL) {
ret = 0;
goto out;
}
ret = 0;
btrfs_node_key_to_cpu(path->nodes[1], &found_key, path->slots[1]);
nritems = btrfs_header_nritems(path->nodes[1]);
if (!nritems)
goto out;
if (path->slots[1] >= nritems)
goto next_node;
/* did we find a key greater than anything we want to delete? */
if (found_key.objectid > inode->i_ino ||
(found_key.objectid == inode->i_ino && found_key.type > key.type))
goto out;
/* we check the next key in the node to make sure the leave contains
* only checksum items. This comparison doesn't work if our
* leaf is the last one in the node
*/
if (path->slots[1] + 1 >= nritems) {
next_node:
/* search forward from the last key in the node, this
* will bring us into the next node in the tree
*/
btrfs_node_key_to_cpu(path->nodes[1], &found_key, nritems - 1);
/* unlikely, but we inc below, so check to be safe */
if (found_key.offset == (u64)-1)
goto out;
/* search_forward needs a path with locks held, do the
* search again for the original key. It is possible
* this will race with a balance and return a path that
* we could modify, but this drop is just an optimization
* and is allowed to miss some leaves.
*/
btrfs_release_path(root, path);
found_key.offset++;
/* setup a max key for search_forward */
other_key.offset = (u64)-1;
other_key.type = key.type;
other_key.objectid = key.objectid;
path->keep_locks = 1;
ret = btrfs_search_forward(root, &found_key, &other_key,
path, 0, 0);
path->keep_locks = 0;
if (ret || found_key.objectid != key.objectid ||
found_key.type != key.type) {
ret = 0;
goto out;
}
key.offset = found_key.offset;
btrfs_release_path(root, path);
cond_resched();
goto again;
}
/* we know there's one more slot after us in the tree,
* read that key so we can verify it is also a checksum item
*/
btrfs_node_key_to_cpu(path->nodes[1], &other_key, path->slots[1] + 1);
if (found_key.objectid < inode->i_ino)
goto next_key;
if (found_key.type != key.type || found_key.offset < new_size)
goto next_key;
/*
* if the key for the next leaf isn't a csum key from this objectid,
* we can't be sure there aren't good items inside this leaf.
* Bail out
*/
if (other_key.objectid != inode->i_ino || other_key.type != key.type)
goto out;
leaf_start = btrfs_node_blockptr(path->nodes[1], path->slots[1]);
leaf_gen = btrfs_node_ptr_generation(path->nodes[1], path->slots[1]);
/*
* it is safe to delete this leaf, it contains only
* csum items from this inode at an offset >= new_size
*/
ret = btrfs_del_leaf(trans, root, path, leaf_start);
BUG_ON(ret);
if (root->ref_cows && leaf_gen < trans->transid) {
ref = btrfs_alloc_leaf_ref(root, 0);
if (ref) {
ref->root_gen = root->root_key.offset;
ref->bytenr = leaf_start;
ref->owner = 0;
ref->generation = leaf_gen;
ref->nritems = 0;
btrfs_sort_leaf_ref(ref);
ret = btrfs_add_leaf_ref(root, ref, 0);
WARN_ON(ret);
btrfs_free_leaf_ref(root, ref);
} else {
WARN_ON(1);
}
}
next_key:
btrfs_release_path(root, path);
if (other_key.objectid == inode->i_ino &&
other_key.type == key.type && other_key.offset > key.offset) {
key.offset = other_key.offset;
cond_resched();
goto again;
}
ret = 0;
out:
/* fixup any changes we've made to the path */
path->lowest_level = 0;
path->keep_locks = 0;
btrfs_release_path(root, path);
return ret;
}
#endif
/*
* this can truncate away extent items, csum items and directory items.
* It starts at a high offset and removes keys until it can't find
......
fs/btrfs/transaction.c
View file @ 182608c8
......@@ -346,49 +346,6 @@ int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid)
return ret;
}
#if 0
/*
* rate limit against the drop_snapshot code. This helps to slow down new
* operations if the drop_snapshot code isn't able to keep up.
*/
static void throttle_on_drops(struct btrfs_root *root)
{
struct btrfs_fs_info *info = root->fs_info;
int harder_count = 0;
harder:
if (atomic_read(&info->throttles)) {
DEFINE_WAIT(wait);
int thr;
thr = atomic_read(&info->throttle_gen);
do {
prepare_to_wait(&info->transaction_throttle,
&wait, TASK_UNINTERRUPTIBLE);
if (!atomic_read(&info->throttles)) {
finish_wait(&info->transaction_throttle, &wait);
break;
}
schedule();
finish_wait(&info->transaction_throttle, &wait);
} while (thr == atomic_read(&info->throttle_gen));
harder_count++;
if (root->fs_info->total_ref_cache_size > 1 * 1024 * 1024 &&
harder_count < 2)
goto harder;
if (root->fs_info->total_ref_cache_size > 5 * 1024 * 1024 &&
harder_count < 10)
goto harder;
if (root->fs_info->total_ref_cache_size > 10 * 1024 * 1024 &&
harder_count < 20)
goto harder;
}
}
#endif
void btrfs_throttle(struct btrfs_root *root)
{
mutex_lock(&root->fs_info->trans_mutex);
......@@ -808,97 +765,6 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly)
return ret;
}
#if 0
/*
* when dropping snapshots, we generate a ton of delayed refs, and it makes
* sense not to join the transaction while it is trying to flush the current
* queue of delayed refs out.
*
* This is used by the drop snapshot code only
*/
static noinline int wait_transaction_pre_flush(struct btrfs_fs_info *info)
{
DEFINE_WAIT(wait);
mutex_lock(&info->trans_mutex);
while (info->running_transaction &&
info->running_transaction->delayed_refs.flushing) {
prepare_to_wait(&info->transaction_wait, &wait,
TASK_UNINTERRUPTIBLE);
mutex_unlock(&info->trans_mutex);
schedule();
mutex_lock(&info->trans_mutex);
finish_wait(&info->transaction_wait, &wait);
}
mutex_unlock(&info->trans_mutex);
return 0;
}
/*
* Given a list of roots that need to be deleted, call btrfs_drop_snapshot on
* all of them
*/
int btrfs_drop_dead_root(struct btrfs_root *root)
{
struct btrfs_trans_handle *trans;
struct btrfs_root *tree_root = root->fs_info->tree_root;
unsigned long nr;
int ret;
while (1) {
/*
* we don't want to jump in and create a bunch of
* delayed refs if the transaction is starting to close
*/
wait_transaction_pre_flush(tree_root->fs_info);
trans = btrfs_start_transaction(tree_root, 1);
/*
* we've joined a transaction, make sure it isn't
* closing right now
*/
if (trans->transaction->delayed_refs.flushing) {
btrfs_end_transaction(trans, tree_root);
continue;
}
ret = btrfs_drop_snapshot(trans, root);
if (ret != -EAGAIN)
break;
ret = btrfs_update_root(trans, tree_root,
&root->root_key,
&root->root_item);
if (ret)
break;
nr = trans->blocks_used;
ret = btrfs_end_transaction(trans, tree_root);
BUG_ON(ret);
btrfs_btree_balance_dirty(tree_root, nr);
cond_resched();
}
BUG_ON(ret);
ret = btrfs_del_root(trans, tree_root, &root->root_key);
BUG_ON(ret);
nr = trans->blocks_used;
ret = btrfs_end_transaction(trans, tree_root);
BUG_ON(ret);
free_extent_buffer(root->node);
free_extent_buffer(root->commit_root);
kfree(root);
btrfs_btree_balance_dirty(tree_root, nr);
return ret;
}
#endif
/*
* new snapshots need to be created at a very specific time in the
* transaction commit. This does the actual creation
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7