aboutsummaryrefslogtreecommitdiff
path: root/fs/btrfs/inode.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/btrfs/inode.c')
-rw-r--r--fs/btrfs/inode.c700
1 files changed, 282 insertions, 418 deletions
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 7cd8ab0ef04..bb51bb1fa44 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -37,6 +37,7 @@
#include <linux/posix_acl.h>
#include <linux/falloc.h>
#include <linux/slab.h>
+#include <linux/ratelimit.h>
#include "compat.h"
#include "ctree.h"
#include "disk-io.h"
@@ -51,6 +52,7 @@
#include "compression.h"
#include "locking.h"
#include "free-space-cache.h"
+#include "inode-map.h"
struct btrfs_iget_args {
u64 ino;
@@ -138,7 +140,7 @@ static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
path->leave_spinning = 1;
btrfs_set_trans_block_group(trans, inode);
- key.objectid = inode->i_ino;
+ key.objectid = btrfs_ino(inode);
key.offset = start;
btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
datasize = btrfs_file_extent_calc_inline_size(cur_size);
@@ -340,6 +342,10 @@ static noinline int compress_file_range(struct inode *inode,
int will_compress;
int compress_type = root->fs_info->compress_type;
+ /* if this is a small write inside eof, kick off a defragbot */
+ if (end <= BTRFS_I(inode)->disk_i_size && (end - start + 1) < 16 * 1024)
+ btrfs_add_inode_defrag(NULL, inode);
+
actual_end = min_t(u64, isize, end + 1);
again:
will_compress = 0;
@@ -649,7 +655,7 @@ retry:
async_extent->start +
async_extent->ram_size - 1, 0);
- em = alloc_extent_map(GFP_NOFS);
+ em = alloc_extent_map();
BUG_ON(!em);
em->start = async_extent->start;
em->len = async_extent->ram_size;
@@ -745,6 +751,15 @@ static u64 get_extent_allocation_hint(struct inode *inode, u64 start,
return alloc_hint;
}
+static inline bool is_free_space_inode(struct btrfs_root *root,
+ struct inode *inode)
+{
+ if (root == root->fs_info->tree_root ||
+ BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID)
+ return true;
+ return false;
+}
+
/*
* when extent_io.c finds a delayed allocation range in the file,
* the call backs end up in this code. The basic idea is to
@@ -777,7 +792,7 @@ static noinline int cow_file_range(struct inode *inode,
struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
int ret = 0;
- BUG_ON(root == root->fs_info->tree_root);
+ BUG_ON(is_free_space_inode(root, inode));
trans = btrfs_join_transaction(root, 1);
BUG_ON(IS_ERR(trans));
btrfs_set_trans_block_group(trans, inode);
@@ -788,6 +803,10 @@ static noinline int cow_file_range(struct inode *inode,
disk_num_bytes = num_bytes;
ret = 0;
+ /* if this is a small write inside eof, kick off defrag */
+ if (end <= BTRFS_I(inode)->disk_i_size && num_bytes < 64 * 1024)
+ btrfs_add_inode_defrag(trans, inode);
+
if (start == 0) {
/* lets try to make an inline extent */
ret = cow_file_range_inline(trans, root, inode,
@@ -826,7 +845,7 @@ static noinline int cow_file_range(struct inode *inode,
(u64)-1, &ins, 1);
BUG_ON(ret);
- em = alloc_extent_map(GFP_NOFS);
+ em = alloc_extent_map();
BUG_ON(!em);
em->start = start;
em->orig_start = em->start;
@@ -1008,7 +1027,7 @@ static noinline int csum_exist_in_range(struct btrfs_root *root,
LIST_HEAD(list);
ret = btrfs_lookup_csums_range(root->fs_info->csum_root, bytenr,
- bytenr + num_bytes - 1, &list);
+ bytenr + num_bytes - 1, &list, 0);
if (ret == 0 && list_empty(&list))
return 0;
@@ -1049,29 +1068,31 @@ static noinline int run_delalloc_nocow(struct inode *inode,
int type;
int nocow;
int check_prev = 1;
- bool nolock = false;
+ bool nolock;
+ u64 ino = btrfs_ino(inode);
path = btrfs_alloc_path();
BUG_ON(!path);
- if (root == root->fs_info->tree_root) {
- nolock = true;
+
+ nolock = is_free_space_inode(root, inode);
+
+ if (nolock)
trans = btrfs_join_transaction_nolock(root, 1);
- } else {
+ else
trans = btrfs_join_transaction(root, 1);
- }
BUG_ON(IS_ERR(trans));
cow_start = (u64)-1;
cur_offset = start;
while (1) {
- ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
+ ret = btrfs_lookup_file_extent(trans, root, path, ino,
cur_offset, 0);
BUG_ON(ret < 0);
if (ret > 0 && path->slots[0] > 0 && check_prev) {
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &found_key,
path->slots[0] - 1);
- if (found_key.objectid == inode->i_ino &&
+ if (found_key.objectid == ino &&
found_key.type == BTRFS_EXTENT_DATA_KEY)
path->slots[0]--;
}
@@ -1092,7 +1113,7 @@ next_slot:
num_bytes = 0;
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
- if (found_key.objectid > inode->i_ino ||
+ if (found_key.objectid > ino ||
found_key.type > BTRFS_EXTENT_DATA_KEY ||
found_key.offset > end)
break;
@@ -1127,7 +1148,7 @@ next_slot:
goto out_check;
if (btrfs_extent_readonly(root, disk_bytenr))
goto out_check;
- if (btrfs_cross_ref_exist(trans, root, inode->i_ino,
+ if (btrfs_cross_ref_exist(trans, root, ino,
found_key.offset -
extent_offset, disk_bytenr))
goto out_check;
@@ -1164,7 +1185,7 @@ out_check:
goto next_slot;
}
- btrfs_release_path(root, path);
+ btrfs_release_path(path);
if (cow_start != (u64)-1) {
ret = cow_file_range(inode, locked_page, cow_start,
found_key.offset - 1, page_started,
@@ -1177,7 +1198,7 @@ out_check:
struct extent_map *em;
struct extent_map_tree *em_tree;
em_tree = &BTRFS_I(inode)->extent_tree;
- em = alloc_extent_map(GFP_NOFS);
+ em = alloc_extent_map();
BUG_ON(!em);
em->start = cur_offset;
em->orig_start = em->start;
@@ -1222,7 +1243,7 @@ out_check:
if (cur_offset > end)
break;
}
- btrfs_release_path(root, path);
+ btrfs_release_path(path);
if (cur_offset <= end && cow_start == (u64)-1)
cow_start = cur_offset;
@@ -1310,14 +1331,13 @@ static int btrfs_set_bit_hook(struct inode *inode,
/*
* set_bit and clear bit hooks normally require _irqsave/restore
- * but in this case, we are only testeing for the DELALLOC
+ * but in this case, we are only testing for the DELALLOC
* bit, which is only set or cleared with irqs on
*/
if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
struct btrfs_root *root = BTRFS_I(inode)->root;
u64 len = state->end + 1 - state->start;
- int do_list = (root->root_key.objectid !=
- BTRFS_ROOT_TREE_OBJECTID);
+ bool do_list = !is_free_space_inode(root, inode);
if (*bits & EXTENT_FIRST_DELALLOC)
*bits &= ~EXTENT_FIRST_DELALLOC;
@@ -1344,14 +1364,13 @@ static int btrfs_clear_bit_hook(struct inode *inode,
{
/*
* set_bit and clear bit hooks normally require _irqsave/restore
- * but in this case, we are only testeing for the DELALLOC
+ * but in this case, we are only testing for the DELALLOC
* bit, which is only set or cleared with irqs on
*/
if ((state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
struct btrfs_root *root = BTRFS_I(inode)->root;
u64 len = state->end + 1 - state->start;
- int do_list = (root->root_key.objectid !=
- BTRFS_ROOT_TREE_OBJECTID);
+ bool do_list = !is_free_space_inode(root, inode);
if (*bits & EXTENT_FIRST_DELALLOC)
*bits &= ~EXTENT_FIRST_DELALLOC;
@@ -1458,7 +1477,7 @@ static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
- if (root == root->fs_info->tree_root)
+ if (is_free_space_inode(root, inode))
ret = btrfs_bio_wq_end_io(root->fs_info, bio, 2);
else
ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
@@ -1644,7 +1663,7 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
&hint, 0);
BUG_ON(ret);
- ins.objectid = inode->i_ino;
+ ins.objectid = btrfs_ino(inode);
ins.offset = file_pos;
ins.type = BTRFS_EXTENT_DATA_KEY;
ret = btrfs_insert_empty_item(trans, root, path, &ins, sizeof(*fi));
@@ -1675,7 +1694,7 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
ins.type = BTRFS_EXTENT_ITEM_KEY;
ret = btrfs_alloc_reserved_file_extent(trans, root,
root->root_key.objectid,
- inode->i_ino, file_pos, &ins);
+ btrfs_ino(inode), file_pos, &ins);
BUG_ON(ret);
btrfs_free_path(path);
@@ -1701,7 +1720,7 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
struct extent_state *cached_state = NULL;
int compress_type = 0;
int ret;
- bool nolock = false;
+ bool nolock;
ret = btrfs_dec_test_ordered_pending(inode, &ordered_extent, start,
end - start + 1);
@@ -1709,7 +1728,7 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
return 0;
BUG_ON(!ordered_extent);
- nolock = (root == root->fs_info->tree_root);
+ nolock = is_free_space_inode(root, inode);
if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) {
BUG_ON(!list_empty(&ordered_extent->list));
@@ -1855,7 +1874,7 @@ static int btrfs_io_failed_hook(struct bio *failed_bio,
}
read_unlock(&em_tree->lock);
- if (!em || IS_ERR(em)) {
+ if (IS_ERR_OR_NULL(em)) {
kfree(failrec);
return -EIO;
}
@@ -2004,12 +2023,11 @@ good:
return 0;
zeroit:
- if (printk_ratelimit()) {
- printk(KERN_INFO "btrfs csum failed ino %lu off %llu csum %u "
- "private %llu\n", page->mapping->host->i_ino,
+ printk_ratelimited(KERN_INFO "btrfs csum failed ino %llu off %llu csum %u "
+ "private %llu\n",
+ (unsigned long long)btrfs_ino(page->mapping->host),
(unsigned long long)start, csum,
(unsigned long long)private);
- }
memset(kaddr + offset, 1, end - start + 1);
flush_dcache_page(page);
kunmap_atomic(kaddr, KM_USER0);
@@ -2244,7 +2262,7 @@ int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode)
/* insert an orphan item to track this unlinked/truncated file */
if (insert >= 1) {
- ret = btrfs_insert_orphan_item(trans, root, inode->i_ino);
+ ret = btrfs_insert_orphan_item(trans, root, btrfs_ino(inode));
BUG_ON(ret);
}
@@ -2281,7 +2299,7 @@ int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode)
spin_unlock(&root->orphan_lock);
if (trans && delete_item) {
- ret = btrfs_del_orphan_item(trans, root, inode->i_ino);
+ ret = btrfs_del_orphan_item(trans, root, btrfs_ino(inode));
BUG_ON(ret);
}
@@ -2346,7 +2364,7 @@ int btrfs_orphan_cleanup(struct btrfs_root *root)
break;
/* release the path since we're done with it */
- btrfs_release_path(root, path);
+ btrfs_release_path(path);
/*
* this is where we are basically btrfs_lookup, without the
@@ -2543,7 +2561,8 @@ static void btrfs_read_locked_inode(struct inode *inode)
* try to precache a NULL acl entry for files that don't have
* any xattrs or acls
*/
- maybe_acls = acls_after_inode_item(leaf, path->slots[0], inode->i_ino);
+ maybe_acls = acls_after_inode_item(leaf, path->slots[0],
+ btrfs_ino(inode));
if (!maybe_acls)
cache_no_acl(inode);
@@ -2647,11 +2666,26 @@ noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
struct extent_buffer *leaf;
int ret;
+ /*
+ * If root is tree root, it means this inode is used to
+ * store free space information. And these inodes are updated
+ * when committing the transaction, so they needn't delaye to
+ * be updated, or deadlock will occured.
+ */
+ if (!is_free_space_inode(root, inode)) {
+ ret = btrfs_delayed_update_inode(trans, root, inode);
+ if (!ret)
+ btrfs_set_inode_last_trans(trans, inode);
+ return ret;
+ }
+
path = btrfs_alloc_path();
- BUG_ON(!path);
+ if (!path)
+ return -ENOMEM;
+
path->leave_spinning = 1;
- ret = btrfs_lookup_inode(trans, root, path,
- &BTRFS_I(inode)->location, 1);
+ ret = btrfs_lookup_inode(trans, root, path, &BTRFS_I(inode)->location,
+ 1);
if (ret) {
if (ret > 0)
ret = -ENOENT;
@@ -2661,7 +2695,7 @@ noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
btrfs_unlock_up_safe(path, 1);
leaf = path->nodes[0];
inode_item = btrfs_item_ptr(leaf, path->slots[0],
- struct btrfs_inode_item);
+ struct btrfs_inode_item);
fill_inode_item(trans, leaf, inode_item, inode);
btrfs_mark_buffer_dirty(leaf);
@@ -2672,7 +2706,6 @@ failed:
return ret;
}
-
/*
* unlink helper that gets used here in inode.c and in the tree logging
* recovery code. It remove a link in a directory with a given name, and
@@ -2689,6 +2722,8 @@ static int __btrfs_unlink_inode(struct btrfs_trans_handle *trans,
struct btrfs_dir_item *di;
struct btrfs_key key;
u64 index;
+ u64 ino = btrfs_ino(inode);
+ u64 dir_ino = btrfs_ino(dir);
path = btrfs_alloc_path();
if (!path) {
@@ -2697,7 +2732,7 @@ static int __btrfs_unlink_inode(struct btrfs_trans_handle *trans,
}
path->leave_spinning = 1;
- di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
+ di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
name, name_len, -1);
if (IS_ERR(di)) {
ret = PTR_ERR(di);
@@ -2712,33 +2747,23 @@ static int __btrfs_unlink_inode(struct btrfs_trans_handle *trans,
ret = btrfs_delete_one_dir_name(trans, root, path, di);
if (ret)
goto err;
- btrfs_release_path(root, path);
+ btrfs_release_path(path);
- ret = btrfs_del_inode_ref(trans, root, name, name_len,
- inode->i_ino,
- dir->i_ino, &index);
+ ret = btrfs_del_inode_ref(trans, root, name, name_len, ino,
+ dir_ino, &index);
if (ret) {
printk(KERN_INFO "btrfs failed to delete reference to %.*s, "
- "inode %lu parent %lu\n", name_len, name,
- inode->i_ino, dir->i_ino);
+ "inode %llu parent %llu\n", name_len, name,
+ (unsigned long long)ino, (unsigned long long)dir_ino);
goto err;
}
- di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
- index, name, name_len, -1);
- if (IS_ERR(di)) {
- ret = PTR_ERR(di);
- goto err;
- }
- if (!di) {
- ret = -ENOENT;
+ ret = btrfs_delete_delayed_dir_index(trans, root, dir, index);
+ if (ret)
goto err;
- }
- ret = btrfs_delete_one_dir_name(trans, root, path, di);
- btrfs_release_path(root, path);
ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len,
- inode, dir->i_ino);
+ inode, dir_ino);
BUG_ON(ret != 0 && ret != -ENOENT);
ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len,
@@ -2816,12 +2841,14 @@ static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
int check_link = 1;
int err = -ENOSPC;
int ret;
+ u64 ino = btrfs_ino(inode);
+ u64 dir_ino = btrfs_ino(dir);
trans = btrfs_start_transaction(root, 10);
if (!IS_ERR(trans) || PTR_ERR(trans) != -ENOSPC)
return trans;
- if (inode->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
+ if (ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
return ERR_PTR(-ENOSPC);
/* check if there is someone else holds reference */
@@ -2862,7 +2889,7 @@ static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
} else {
check_link = 0;
}
- btrfs_release_path(root, path);
+ btrfs_release_path(path);
ret = btrfs_lookup_inode(trans, root, path,
&BTRFS_I(inode)->location, 0);
@@ -2876,11 +2903,11 @@ static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
} else {
check_link = 0;
}
- btrfs_release_path(root, path);
+ btrfs_release_path(path);
if (ret == 0 && S_ISREG(inode->i_mode)) {
ret = btrfs_lookup_file_extent(trans, root, path,
- inode->i_ino, (u64)-1, 0);
+ ino, (u64)-1, 0);
if (ret < 0) {
err = ret;
goto out;
@@ -2888,7 +2915,7 @@ static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
BUG_ON(ret == 0);
if (check_path_shared(root, path))
goto out;
- btrfs_release_path(root, path);
+ btrfs_release_path(path);
}
if (!check_link) {
@@ -2896,7 +2923,7 @@ static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
goto out;
}
- di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
+ di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
dentry->d_name.name, dentry->d_name.len, 0);
if (IS_ERR(di)) {
err = PTR_ERR(di);
@@ -2909,11 +2936,11 @@ static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
err = 0;
goto out;
}
- btrfs_release_path(root, path);
+ btrfs_release_path(path);
ref = btrfs_lookup_inode_ref(trans, root, path,
dentry->d_name.name, dentry->d_name.len,
- inode->i_ino, dir->i_ino, 0);
+ ino, dir_ino, 0);
if (IS_ERR(ref)) {
err = PTR_ERR(ref);
goto out;
@@ -2922,9 +2949,17 @@ static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
if (check_path_shared(root, path))
goto out;
index = btrfs_inode_ref_index(path->nodes[0], ref);
- btrfs_release_path(root, path);
+ btrfs_release_path(path);
- di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino, index,
+ /*
+ * This is a commit root search, if we can lookup inode item and other
+ * relative items in the commit root, it means the transaction of
+ * dir/file creation has been committed, and the dir index item that we
+ * delay to insert has also been inserted into the commit root. So
+ * we needn't worry about the delayed insertion of the dir index item
+ * here.
+ */
+ di = btrfs_lookup_dir_index_item(trans, root, path, dir_ino, index,
dentry->d_name.name, dentry->d_name.len, 0);
if (IS_ERR(di)) {
err = PTR_ERR(di);
@@ -2999,54 +3034,47 @@ int btrfs_unlink_subvol(struct btrfs_trans_handle *trans,
struct btrfs_key key;
u64 index;
int ret;
+ u64 dir_ino = btrfs_ino(dir);
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
- di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
+ di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
name, name_len, -1);
- BUG_ON(!di || IS_ERR(di));
+ BUG_ON(IS_ERR_OR_NULL(di));
leaf = path->nodes[0];
btrfs_dir_item_key_to_cpu(leaf, di, &key);
WARN_ON(key.type != BTRFS_ROOT_ITEM_KEY || key.objectid != objectid);
ret = btrfs_delete_one_dir_name(trans, root, path, di);
BUG_ON(ret);
- btrfs_release_path(root, path);
+ btrfs_release_path(path);
ret = btrfs_del_root_ref(trans, root->fs_info->tree_root,
objectid, root->root_key.objectid,
- dir->i_ino, &index, name, name_len);
+ dir_ino, &index, name, name_len);
if (ret < 0) {
BUG_ON(ret != -ENOENT);
- di = btrfs_search_dir_index_item(root, path, dir->i_ino,
+ di = btrfs_search_dir_index_item(root, path, dir_ino,
name, name_len);
- BUG_ON(!di || IS_ERR(di));
+ BUG_ON(IS_ERR_OR_NULL(di));
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
- btrfs_release_path(root, path);
+ btrfs_release_path(path);
index = key.offset;
}
+ btrfs_release_path(path);
- di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
- index, name, name_len, -1);
- BUG_ON(!di || IS_ERR(di));
-
- leaf = path->nodes[0];
- btrfs_dir_item_key_to_cpu(leaf, di, &key);
- WARN_ON(key.type != BTRFS_ROOT_ITEM_KEY || key.objectid != objectid);
- ret = btrfs_delete_one_dir_name(trans, root, path, di);
+ ret = btrfs_delete_delayed_dir_index(trans, root, dir, index);
BUG_ON(ret);
- btrfs_release_path(root, path);
btrfs_i_size_write(dir, dir->i_size - name_len * 2);
dir->i_mtime = dir->i_ctime = CURRENT_TIME;
ret = btrfs_update_inode(trans, root, dir);
BUG_ON(ret);
- btrfs_free_path(path);
return 0;
}
@@ -3059,7 +3087,7 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
unsigned long nr = 0;
if (inode->i_size > BTRFS_EMPTY_DIR_SIZE ||
- inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
+ btrfs_ino(inode) == BTRFS_FIRST_FREE_OBJECTID)
return -ENOTEMPTY;
trans = __unlink_start_trans(dir, dentry);
@@ -3068,7 +3096,7 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
btrfs_set_trans_block_group(trans, dir);
- if (unlikely(inode->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
+ if (unlikely(btrfs_ino(inode) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
err = btrfs_unlink_subvol(trans, root, dir,
BTRFS_I(inode)->location.objectid,
dentry->d_name.name,
@@ -3093,178 +3121,6 @@ out:
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
@@ -3300,17 +3156,27 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
int encoding;
int ret;
int err = 0;
+ u64 ino = btrfs_ino(inode);
BUG_ON(new_size > 0 && min_type != BTRFS_EXTENT_DATA_KEY);
if (root->ref_cows || root == root->fs_info->tree_root)
btrfs_drop_extent_cache(inode, new_size & (~mask), (u64)-1, 0);
+ /*
+ * This function is also used to drop the items in the log tree before
+ * we relog the inode, so if root != BTRFS_I(inode)->root, it means
+ * it is used to drop the loged items. So we shouldn't kill the delayed
+ * items.
+ */
+ if (min_type == 0 && root == BTRFS_I(inode)->root)
+ btrfs_kill_delayed_inode_items(inode);
+
path = btrfs_alloc_path();
BUG_ON(!path);
path->reada = -1;
- key.objectid = inode->i_ino;
+ key.objectid = ino;
key.offset = (u64)-1;
key.type = (u8)-1;
@@ -3338,7 +3204,7 @@ search_again:
found_type = btrfs_key_type(&found_key);
encoding = 0;
- if (found_key.objectid != inode->i_ino)
+ if (found_key.objectid != ino)
break;
if (found_type < min_type)
@@ -3428,7 +3294,6 @@ search_again:
btrfs_file_extent_calc_inline_size(size);
ret = btrfs_truncate_item(trans, root, path,
size, 1);
- BUG_ON(ret);
} else if (root->ref_cows) {
inode_sub_bytes(inode, item_end + 1 -
found_key.offset);
@@ -3457,7 +3322,7 @@ delete:
ret = btrfs_free_extent(trans, root, extent_start,
extent_num_bytes, 0,
btrfs_header_owner(leaf),
- inode->i_ino, extent_offset);
+ ino, extent_offset);
BUG_ON(ret);
}
@@ -3466,7 +3331,9 @@ delete:
if (path->slots[0] == 0 ||
path->slots[0] != pending_del_slot) {
- if (root->ref_cows) {
+ if (root->ref_cows &&
+ BTRFS_I(inode)->location.objectid !=
+ BTRFS_FREE_INO_OBJECTID) {
err = -EAGAIN;
goto out;
}
@@ -3477,7 +3344,7 @@ delete:
BUG_ON(ret);
pending_del_nr = 0;
}
- btrfs_release_path(root, path);
+ btrfs_release_path(path);
goto search_again;
} else {
path->slots[0]--;
@@ -3635,7 +3502,7 @@ int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size)
while (1) {
em = btrfs_get_extent(inode, NULL, 0, cur_offset,
block_end - cur_offset, 0);
- BUG_ON(IS_ERR(em) || !em);
+ BUG_ON(IS_ERR_OR_NULL(em));
last_byte = min(extent_map_end(em), block_end);
last_byte = (last_byte + mask) & ~mask;
if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) {
@@ -3656,7 +3523,7 @@ int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size)
break;
err = btrfs_insert_file_extent(trans, root,
- inode->i_ino, cur_offset, 0,
+ btrfs_ino(inode), cur_offset, 0,
0, hole_size, 0, hole_size,
0, 0, 0);
if (err)
@@ -3758,7 +3625,7 @@ void btrfs_evict_inode(struct inode *inode)
truncate_inode_pages(&inode->i_data, 0);
if (inode->i_nlink && (btrfs_root_refs(&root->root_item) != 0 ||
- root == root->fs_info->tree_root))
+ is_free_space_inode(root, inode)))
goto no_delete;
if (is_bad_inode(inode)) {
@@ -3811,6 +3678,10 @@ void btrfs_evict_inode(struct inode *inode)
BUG_ON(ret);
}
+ if (!(root == root->fs_info->tree_root ||
+ root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID))
+ btrfs_return_ino(root, btrfs_ino(inode));
+
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
btrfs_btree_balance_dirty(root, nr);
@@ -3836,12 +3707,12 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
path = btrfs_alloc_path();
BUG_ON(!path);
- di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
+ di = btrfs_lookup_dir_item(NULL, root, path, btrfs_ino(dir), name,
namelen, 0);
if (IS_ERR(di))
ret = PTR_ERR(di);
- if (!di || IS_ERR(di))
+ if (IS_ERR_OR_NULL(di))
goto out_err;
btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
@@ -3889,7 +3760,7 @@ static int fixup_tree_root_location(struct btrfs_root *root,
leaf = path->nodes[0];
ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
- if (btrfs_root_ref_dirid(leaf, ref) != dir->i_ino ||
+ if (btrfs_root_ref_dirid(leaf, ref) != btrfs_ino(dir) ||
btrfs_root_ref_name_len(leaf, ref) != dentry->d_name.len)
goto out;
@@ -3899,7 +3770,7 @@ static int fixup_tree_root_location(struct btrfs_root *root,
if (ret)
goto out;
- btrfs_release_path(root->fs_info->tree_root, path);
+ btrfs_release_path(path);
new_root = btrfs_read_fs_root_no_name(root->fs_info, location);
if (IS_ERR(new_root)) {
@@ -3928,6 +3799,7 @@ static void inode_tree_add(struct inode *inode)
struct btrfs_inode *entry;
struct rb_node **p;
struct rb_node *parent;
+ u64 ino = btrfs_ino(inode);
again:
p = &root->inode_tree.rb_node;
parent = NULL;
@@ -3940,9 +3812,9 @@ again:
parent = *p;
entry = rb_entry(parent, struct btrfs_inode, rb_node);
- if (inode->i_ino < entry->vfs_inode.i_ino)
+ if (ino < btrfs_ino(&entry->vfs_inode))
p = &parent->rb_left;
- else if (inode->i_ino > entry->vfs_inode.i_ino)
+ else if (ino > btrfs_ino(&entry->vfs_inode))
p = &parent->rb_right;
else {
WARN_ON(!(entry->vfs_inode.i_state &
@@ -4006,9 +3878,9 @@ again:
prev = node;
entry = rb_entry(node, struct btrfs_inode, rb_node);
- if (objectid < entry->vfs_inode.i_ino)
+ if (objectid < btrfs_ino(&entry->vfs_inode))
node = node->rb_left;
- else if (objectid > entry->vfs_inode.i_ino)
+ else if (objectid > btrfs_ino(&entry->vfs_inode))
node = node->rb_right;
else
break;
@@ -4016,7 +3888,7 @@ again:
if (!node) {
while (prev) {
entry = rb_entry(prev, struct btrfs_inode, rb_node);
- if (objectid <= entry->vfs_inode.i_ino) {
+ if (objectid <= btrfs_ino(&entry->vfs_inode)) {
node = prev;
break;
}
@@ -4025,7 +3897,7 @@ again:
}
while (node) {
entry = rb_entry(node, struct btrfs_inode, rb_node);
- objectid = entry->vfs_inode.i_ino + 1;
+ objectid = btrfs_ino(&entry->vfs_inode) + 1;
inode = igrab(&entry->vfs_inode);
if (inode) {
spin_unlock(&root->inode_lock);
@@ -4063,7 +3935,7 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p)
static int btrfs_find_actor(struct inode *inode, void *opaque)
{
struct btrfs_iget_args *args = opaque;
- return args->ino == inode->i_ino &&
+ return args->ino == btrfs_ino(inode) &&
args->root == BTRFS_I(inode)->root;
}
@@ -4208,7 +4080,7 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
return d_splice_alias(inode, dentry);
}
-static unsigned char btrfs_filetype_table[] = {
+unsigned char btrfs_filetype_table[] = {
DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
};
@@ -4222,6 +4094,8 @@ static int btrfs_real_readdir(struct file *filp, void *dirent,
struct btrfs_key key;
struct btrfs_key found_key;
struct btrfs_path *path;
+ struct list_head ins_list;
+ struct list_head del_list;
int ret;
struct extent_buffer *leaf;
int slot;
@@ -4234,6 +4108,7 @@ static int btrfs_real_readdir(struct file *filp, void *dirent,
char tmp_name[32];
char *name_ptr;
int name_len;
+ int is_curr = 0; /* filp->f_pos points to the current index? */
/* FIXME, use a real flag for deciding about the key type */
if (root->fs_info->tree_root == root)
@@ -4241,9 +4116,7 @@ static int btrfs_real_readdir(struct file *filp, void *dirent,
/* special case for "." */
if (filp->f_pos == 0) {
- over = filldir(dirent, ".", 1,
- 1, inode->i_ino,
- DT_DIR);
+ over = filldir(dirent, ".", 1, 1, btrfs_ino(inode), DT_DIR);
if (over)
return 0;
filp->f_pos = 1;
@@ -4258,11 +4131,19 @@ static int btrfs_real_readdir(struct file *filp, void *dirent,
filp->f_pos = 2;
}
path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
path->reada = 2;
+ if (key_type == BTRFS_DIR_INDEX_KEY) {
+ INIT_LIST_HEAD(&ins_list);
+ INIT_LIST_HEAD(&del_list);
+ btrfs_get_delayed_items(inode, &ins_list, &del_list);
+ }
+
btrfs_set_key_type(&key, key_type);
key.offset = filp->f_pos;
- key.objectid = inode->i_ino;
+ key.objectid = btrfs_ino(inode);
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
if (ret < 0)
@@ -4289,8 +4170,13 @@ static int btrfs_real_readdir(struct file *filp, void *dirent,
break;
if (found_key.offset < filp->f_pos)
goto next;
+ if (key_type == BTRFS_DIR_INDEX_KEY &&
+ btrfs_should_delete_dir_index(&del_list,
+ found_key.offset))
+ goto next;
filp->f_pos = found_key.offset;
+ is_curr = 1;
di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
di_cur = 0;
@@ -4345,6 +4231,15 @@ next:
path->slots[0]++;
}
+ if (key_type == BTRFS_DIR_INDEX_KEY) {
+ if (is_curr)
+ filp->f_pos++;
+ ret = btrfs_readdir_delayed_dir_index(filp, dirent, filldir,
+ &ins_list);
+ if (ret)
+ goto nopos;
+ }
+
/* Reached end of directory/root. Bump pos past the last item. */
if (key_type == BTRFS_DIR_INDEX_KEY)
/*
@@ -4357,6 +4252,8 @@ next:
nopos:
ret = 0;
err:
+ if (key_type == BTRFS_DIR_INDEX_KEY)
+ btrfs_put_delayed_items(&ins_list, &del_list);
btrfs_free_path(path);
return ret;
}
@@ -4372,7 +4269,8 @@ int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc)
return 0;
smp_mb();
- nolock = (root->fs_info->closing && root == root->fs_info->tree_root);
+ if (root->fs_info->closing && is_free_space_inode(root, inode))
+ nolock = true;
if (wbc->sync_mode == WB_SYNC_ALL) {
if (nolock)
@@ -4415,25 +4313,25 @@ void btrfs_dirty_inode(struct inode *inode)
btrfs_end_transaction(trans, root);
trans = btrfs_start_tra